diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
135 files changed, 8942 insertions, 5004 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/Kconfig b/drivers/gpu/drm/amd/amdgpu/Kconfig index 26682454a446..e8af1f5e8a79 100644 --- a/drivers/gpu/drm/amd/amdgpu/Kconfig +++ b/drivers/gpu/drm/amd/amdgpu/Kconfig @@ -41,3 +41,4 @@ config DRM_AMDGPU_GART_DEBUGFS pages. Uses more memory for housekeeping, enable only for debugging. source "drivers/gpu/drm/amd/acp/Kconfig" +source "drivers/gpu/drm/amd/display/Kconfig" diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index faea6349228f..78d609123420 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -1,15 +1,22 @@ +# SPDX-License-Identifier: GPL-2.0 # # Makefile for the drm device driver. This driver provides support for the # Direct Rendering Infrastructure (DRI) in XFree86 4.1.0 and higher. FULL_AMD_PATH=$(src)/.. +DISPLAY_FOLDER_NAME=display +FULL_AMD_DISPLAY_PATH = $(FULL_AMD_PATH)/$(DISPLAY_FOLDER_NAME) ccflags-y := -I$(FULL_AMD_PATH)/include/asic_reg \ -I$(FULL_AMD_PATH)/include \ -I$(FULL_AMD_PATH)/amdgpu \ -I$(FULL_AMD_PATH)/scheduler \ -I$(FULL_AMD_PATH)/powerplay/inc \ - -I$(FULL_AMD_PATH)/acp/include + -I$(FULL_AMD_PATH)/acp/include \ + -I$(FULL_AMD_DISPLAY_PATH) \ + -I$(FULL_AMD_DISPLAY_PATH)/include \ + -I$(FULL_AMD_DISPLAY_PATH)/dc \ + -I$(FULL_AMD_DISPLAY_PATH)/amdgpu_dm amdgpu-y := amdgpu_drv.o @@ -25,7 +32,7 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \ amdgpu_prime.o amdgpu_vm.o amdgpu_ib.o amdgpu_pll.o \ amdgpu_ucode.o amdgpu_bo_list.o amdgpu_ctx.o amdgpu_sync.o \ amdgpu_gtt_mgr.o amdgpu_vram_mgr.o amdgpu_virt.o amdgpu_atomfirmware.o \ - amdgpu_queue_mgr.o + amdgpu_queue_mgr.o amdgpu_vf_error.o amdgpu_sched.o # add asic specific block amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o cik_ih.o kv_smc.o kv_dpm.o \ @@ -132,6 +139,13 @@ include $(FULL_AMD_PATH)/powerplay/Makefile amdgpu-y += $(AMD_POWERPLAY_FILES) -obj-$(CONFIG_DRM_AMDGPU)+= amdgpu.o +ifneq ($(CONFIG_DRM_AMD_DC),) + +RELATIVE_AMD_DISPLAY_PATH = ../$(DISPLAY_FOLDER_NAME) +include $(FULL_AMD_DISPLAY_PATH)/Makefile + +amdgpu-y += $(AMD_DISPLAY_FILES) -CFLAGS_amdgpu_trace_points.o := -I$(src) +endif + +obj-$(CONFIG_DRM_AMDGPU)+= amdgpu.o diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index ff7bf1a9f967..5afaf6016b4a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -65,16 +65,21 @@ #include "amdgpu_uvd.h" #include "amdgpu_vce.h" #include "amdgpu_vcn.h" +#include "amdgpu_mn.h" +#include "amdgpu_dm.h" #include "gpu_scheduler.h" #include "amdgpu_virt.h" +#include "amdgpu_gart.h" /* * Modules parameters. */ extern int amdgpu_modeset; extern int amdgpu_vram_limit; +extern int amdgpu_vis_vram_limit; extern int amdgpu_gart_size; +extern int amdgpu_gtt_size; extern int amdgpu_moverate; extern int amdgpu_benchmarking; extern int amdgpu_testing; @@ -88,25 +93,29 @@ extern int amdgpu_dpm; extern int amdgpu_fw_load_type; extern int amdgpu_aspm; extern int amdgpu_runtime_pm; -extern unsigned amdgpu_ip_block_mask; +extern uint amdgpu_ip_block_mask; extern int amdgpu_bapm; extern int amdgpu_deep_color; extern int amdgpu_vm_size; extern int amdgpu_vm_block_size; +extern int amdgpu_vm_fragment_size; extern int amdgpu_vm_fault_stop; extern int amdgpu_vm_debug; extern int amdgpu_vm_update_mode; +extern int amdgpu_dc; +extern int amdgpu_dc_log; extern int amdgpu_sched_jobs; extern int amdgpu_sched_hw_submission; extern int amdgpu_no_evict; extern int amdgpu_direct_gma_size; -extern unsigned amdgpu_pcie_gen_cap; -extern unsigned amdgpu_pcie_lane_cap; -extern unsigned amdgpu_cg_mask; -extern unsigned amdgpu_pg_mask; +extern uint amdgpu_pcie_gen_cap; +extern uint amdgpu_pcie_lane_cap; +extern uint amdgpu_cg_mask; +extern uint amdgpu_pg_mask; +extern uint amdgpu_sdma_phase_quantum; extern char *amdgpu_disable_cu; extern char *amdgpu_virtual_display; -extern unsigned amdgpu_pp_feature_mask; +extern uint amdgpu_pp_feature_mask; extern int amdgpu_vram_page_split; extern int amdgpu_ngg; extern int amdgpu_prim_buf_per_se; @@ -115,6 +124,7 @@ extern int amdgpu_cntl_sb_buf_per_se; extern int amdgpu_param_buf_per_se; extern int amdgpu_job_hang_limit; extern int amdgpu_lbpw; +extern int amdgpu_compute_multipipe; #ifdef CONFIG_DRM_AMDGPU_SI extern int amdgpu_si_support; @@ -173,6 +183,7 @@ struct amdgpu_cs_parser; struct amdgpu_job; struct amdgpu_irq_src; struct amdgpu_fpriv; +struct amdgpu_bo_va_mapping; enum amdgpu_cp_irq { AMDGPU_CP_IRQ_GFX_EOP = 0, @@ -287,14 +298,25 @@ struct amdgpu_buffer_funcs { /* provided by hw blocks that can write ptes, e.g., sdma */ struct amdgpu_vm_pte_funcs { + /* number of dw to reserve per operation */ + unsigned copy_pte_num_dw; + /* copy pte entries from GART */ void (*copy_pte)(struct amdgpu_ib *ib, uint64_t pe, uint64_t src, unsigned count); + /* write pte one entry at a time with addr mapping */ void (*write_pte)(struct amdgpu_ib *ib, uint64_t pe, uint64_t value, unsigned count, uint32_t incr); + + /* maximum nums of PTEs/PDEs in a single operation */ + uint32_t set_max_nums_pte_pde; + + /* number of dw to reserve per operation */ + unsigned set_pte_pde_num_dw; + /* for linear pte/pde updates without addr mapping */ void (*set_pte_pde)(struct amdgpu_ib *ib, uint64_t pe, @@ -327,6 +349,7 @@ struct amdgpu_gart_funcs { struct amdgpu_ih_funcs { /* ring read/write ptr handling, called from interrupt context */ u32 (*get_wptr)(struct amdgpu_device *adev); + bool (*prescreen_iv)(struct amdgpu_device *adev); void (*decode_iv)(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry); void (*set_rptr)(struct amdgpu_device *adev); @@ -369,78 +392,10 @@ struct amdgpu_clock { }; /* - * BO. + * GEM. */ -struct amdgpu_bo_list_entry { - struct amdgpu_bo *robj; - struct ttm_validate_buffer tv; - struct amdgpu_bo_va *bo_va; - uint32_t priority; - struct page **user_pages; - int user_invalidated; -}; - -struct amdgpu_bo_va_mapping { - struct list_head list; - struct rb_node rb; - uint64_t start; - uint64_t last; - uint64_t __subtree_last; - uint64_t offset; - uint64_t flags; -}; - -/* bo virtual addresses in a specific vm */ -struct amdgpu_bo_va { - /* protected by bo being reserved */ - struct list_head bo_list; - struct dma_fence *last_pt_update; - unsigned ref_count; - - /* protected by vm mutex and spinlock */ - struct list_head vm_status; - - /* mappings for this bo_va */ - struct list_head invalids; - struct list_head valids; - - /* constant after initialization */ - struct amdgpu_vm *vm; - struct amdgpu_bo *bo; -}; #define AMDGPU_GEM_DOMAIN_MAX 0x3 - -struct amdgpu_bo { - /* Protected by tbo.reserved */ - u32 prefered_domains; - u32 allowed_domains; - struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1]; - struct ttm_placement placement; - struct ttm_buffer_object tbo; - struct ttm_bo_kmap_obj kmap; - u64 flags; - unsigned pin_count; - void *kptr; - u64 tiling_flags; - u64 metadata_flags; - void *metadata; - u32 metadata_size; - unsigned prime_shared_count; - /* list of all virtual address to which this bo - * is associated to - */ - struct list_head va; - /* Constant after initialization */ - struct drm_gem_object gem_base; - struct amdgpu_bo *parent; - struct amdgpu_bo *shadow; - - struct ttm_bo_kmap_obj dma_buf_vmap; - struct amdgpu_mn *mn; - struct list_head mn_list; - struct list_head shadow_list; -}; #define gem_to_amdgpu_bo(gobj) container_of((gobj), struct amdgpu_bo, gem_base) void amdgpu_gem_object_free(struct drm_gem_object *obj); @@ -462,6 +417,7 @@ void amdgpu_gem_prime_unpin(struct drm_gem_object *obj); struct reservation_object *amdgpu_gem_prime_res_obj(struct drm_gem_object *); void *amdgpu_gem_prime_vmap(struct drm_gem_object *obj); void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr); +int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma); int amdgpu_gem_debugfs_init(struct amdgpu_device *adev); /* sub-allocation manager, it has to be protected by another lock. @@ -518,9 +474,10 @@ struct amdgpu_sa_bo { */ void amdgpu_gem_force_release(struct amdgpu_device *adev); int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, - int alignment, u32 initial_domain, - u64 flags, bool kernel, - struct drm_gem_object **obj); + int alignment, u32 initial_domain, + u64 flags, bool kernel, + struct reservation_object *resv, + struct drm_gem_object **obj); int amdgpu_mode_dumb_create(struct drm_file *file_priv, struct drm_device *dev, @@ -532,49 +489,6 @@ int amdgpu_fence_slab_init(void); void amdgpu_fence_slab_fini(void); /* - * GART structures, functions & helpers - */ -struct amdgpu_mc; - -#define AMDGPU_GPU_PAGE_SIZE 4096 -#define AMDGPU_GPU_PAGE_MASK (AMDGPU_GPU_PAGE_SIZE - 1) -#define AMDGPU_GPU_PAGE_SHIFT 12 -#define AMDGPU_GPU_PAGE_ALIGN(a) (((a) + AMDGPU_GPU_PAGE_MASK) & ~AMDGPU_GPU_PAGE_MASK) - -struct amdgpu_gart { - dma_addr_t table_addr; - struct amdgpu_bo *robj; - void *ptr; - unsigned num_gpu_pages; - unsigned num_cpu_pages; - unsigned table_size; -#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS - struct page **pages; -#endif - bool ready; - - /* Asic default pte flags */ - uint64_t gart_pte_flags; - - const struct amdgpu_gart_funcs *gart_funcs; -}; - -int amdgpu_gart_table_ram_alloc(struct amdgpu_device *adev); -void amdgpu_gart_table_ram_free(struct amdgpu_device *adev); -int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev); -void amdgpu_gart_table_vram_free(struct amdgpu_device *adev); -int amdgpu_gart_table_vram_pin(struct amdgpu_device *adev); -void amdgpu_gart_table_vram_unpin(struct amdgpu_device *adev); -int amdgpu_gart_init(struct amdgpu_device *adev); -void amdgpu_gart_fini(struct amdgpu_device *adev); -int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, - int pages); -int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset, - int pages, struct page **pagelist, - dma_addr_t *dma_addr, uint64_t flags); -int amdgpu_ttm_recover_gart(struct amdgpu_device *adev); - -/* * VMHUB structures, functions & helpers */ struct amdgpu_vmhub { @@ -598,22 +512,20 @@ struct amdgpu_mc { * about vram size near mc fb location */ u64 mc_vram_size; u64 visible_vram_size; - u64 gtt_size; - u64 gtt_start; - u64 gtt_end; + u64 gart_size; + u64 gart_start; + u64 gart_end; u64 vram_start; u64 vram_end; unsigned vram_width; u64 real_vram_size; int vram_mtrr; - u64 gtt_base_align; u64 mc_mask; const struct firmware *fw; /* MC firmware */ uint32_t fw_version; struct amdgpu_irq_src vm_fault; uint32_t vram_type; uint32_t srbm_soft_reset; - struct amdgpu_mode_mc_save save; bool prt_warning; uint64_t stolen_size; /* apertures */ @@ -719,15 +631,15 @@ typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT /* overlap the doorbell assignment with VCN as they are mutually exclusive * VCE engine's doorbell is 32 bit and two VCE ring share one QWORD */ - AMDGPU_DOORBELL64_RING0_1 = 0xF8, - AMDGPU_DOORBELL64_RING2_3 = 0xF9, - AMDGPU_DOORBELL64_RING4_5 = 0xFA, - AMDGPU_DOORBELL64_RING6_7 = 0xFB, + AMDGPU_DOORBELL64_UVD_RING0_1 = 0xF8, + AMDGPU_DOORBELL64_UVD_RING2_3 = 0xF9, + AMDGPU_DOORBELL64_UVD_RING4_5 = 0xFA, + AMDGPU_DOORBELL64_UVD_RING6_7 = 0xFB, - AMDGPU_DOORBELL64_UVD_RING0_1 = 0xFC, - AMDGPU_DOORBELL64_UVD_RING2_3 = 0xFD, - AMDGPU_DOORBELL64_UVD_RING4_5 = 0xFE, - AMDGPU_DOORBELL64_UVD_RING6_7 = 0xFF, + AMDGPU_DOORBELL64_VCE_RING0_1 = 0xFC, + AMDGPU_DOORBELL64_VCE_RING2_3 = 0xFD, + AMDGPU_DOORBELL64_VCE_RING4_5 = 0xFE, + AMDGPU_DOORBELL64_VCE_RING6_7 = 0xFF, AMDGPU_DOORBELL64_MAX_ASSIGNMENT = 0xFF, AMDGPU_DOORBELL64_INVALID = 0xFFFF @@ -823,10 +735,14 @@ struct amdgpu_ctx { struct amdgpu_device *adev; struct amdgpu_queue_mgr queue_mgr; unsigned reset_counter; + uint32_t vram_lost_counter; spinlock_t ring_lock; struct dma_fence **fences; struct amdgpu_ctx_ring rings[AMDGPU_MAX_RINGS]; - bool preamble_presented; + bool preamble_presented; + enum amd_sched_priority init_priority; + enum amd_sched_priority override_priority; + struct mutex lock; }; struct amdgpu_ctx_mgr { @@ -839,17 +755,22 @@ struct amdgpu_ctx_mgr { struct amdgpu_ctx *amdgpu_ctx_get(struct amdgpu_fpriv *fpriv, uint32_t id); int amdgpu_ctx_put(struct amdgpu_ctx *ctx); -uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, - struct dma_fence *fence); +int amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, + struct dma_fence *fence, uint64_t *seq); struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, uint64_t seq); +void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx, + enum amd_sched_priority priority); int amdgpu_ctx_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); +int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, unsigned ring_id); + void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr); void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr); + /* * file private structure */ @@ -857,15 +778,23 @@ void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr); struct amdgpu_fpriv { struct amdgpu_vm vm; struct amdgpu_bo_va *prt_va; + struct amdgpu_bo_va *csa_va; struct mutex bo_list_lock; struct idr bo_list_handles; struct amdgpu_ctx_mgr ctx_mgr; - u32 vram_lost_counter; }; /* * residency list */ +struct amdgpu_bo_list_entry { + struct amdgpu_bo *robj; + struct ttm_validate_buffer tv; + struct amdgpu_bo_va *bo_va; + uint32_t priority; + struct page **user_pages; + int user_invalidated; +}; struct amdgpu_bo_list { struct mutex lock; @@ -953,7 +882,7 @@ struct amdgpu_mec { struct amdgpu_kiq { u64 eop_gpu_addr; struct amdgpu_bo *eop_obj; - struct mutex ring_mutex; + spinlock_t ring_lock; struct amdgpu_ring ring; struct amdgpu_irq_src irq; }; @@ -1113,11 +1042,14 @@ struct amdgpu_gfx { /* reset mask */ uint32_t grbm_soft_reset; uint32_t srbm_soft_reset; - bool in_reset; /* s3/s4 mask */ bool in_suspend; /* NGG */ struct amdgpu_ngg ngg; + + /* pipe reservation */ + struct mutex pipe_reserve_mutex; + DECLARE_BITMAP (pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); }; int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, @@ -1155,11 +1087,14 @@ struct amdgpu_cs_parser { /* buffer objects */ struct ww_acquire_ctx ticket; struct amdgpu_bo_list *bo_list; + struct amdgpu_mn *mn; struct amdgpu_bo_list_entry vm_pd; struct list_head validated; struct dma_fence *fence; uint64_t bytes_moved_threshold; + uint64_t bytes_moved_vis_threshold; uint64_t bytes_moved; + uint64_t bytes_moved_vis; struct amdgpu_bo_list_entry *evictable; /* user fence */ @@ -1193,6 +1128,7 @@ struct amdgpu_job { uint32_t gds_base, gds_size; uint32_t gws_base, gws_size; uint32_t oa_base, oa_size; + uint32_t vram_lost_counter; /* user fence handling */ uint64_t uf_addr; @@ -1218,7 +1154,7 @@ static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p, /* * Writeback */ -#define AMDGPU_MAX_WB 1024 /* Reserve at most 1024 WB slots for amdgpu-owned rings. */ +#define AMDGPU_MAX_WB 512 /* Reserve at most 512 WB slots for amdgpu-owned rings. */ struct amdgpu_wb { struct amdgpu_bo *wb_obj; @@ -1230,8 +1166,6 @@ struct amdgpu_wb { int amdgpu_wb_get(struct amdgpu_device *adev, u32 *wb); void amdgpu_wb_free(struct amdgpu_device *adev, u32 wb); -int amdgpu_wb_get_64bit(struct amdgpu_device *adev, u32 *wb); -void amdgpu_wb_free_64bit(struct amdgpu_device *adev, u32 wb); void amdgpu_get_pcie_info(struct amdgpu_device *adev); @@ -1282,6 +1216,9 @@ struct amdgpu_firmware { /* gpu info firmware data pointer */ const struct firmware *gpu_info_fw; + + void *fw_buf_ptr; + uint64_t fw_buf_mc; }; /* @@ -1296,20 +1233,6 @@ void amdgpu_benchmark(struct amdgpu_device *adev, int test_number); void amdgpu_test_moves(struct amdgpu_device *adev); /* - * MMU Notifier - */ -#if defined(CONFIG_MMU_NOTIFIER) -int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr); -void amdgpu_mn_unregister(struct amdgpu_bo *bo); -#else -static inline int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) -{ - return -ENODEV; -} -static inline void amdgpu_mn_unregister(struct amdgpu_bo *bo) {} -#endif - -/* * Debugfs */ struct amdgpu_debugfs { @@ -1404,6 +1327,8 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); +int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); @@ -1470,6 +1395,18 @@ struct amdgpu_atcs { }; /* + * Firmware VRAM reservation + */ +struct amdgpu_fw_vram_usage { + u64 start_offset; + u64 size; + struct amdgpu_bo *reserved_bo; + void *va; +}; + +int amdgpu_fw_reserve_vram_init(struct amdgpu_device *adev); + +/* * CGS */ struct cgs_device *amdgpu_cgs_create_device(struct amdgpu_device *adev); @@ -1525,7 +1462,7 @@ struct amdgpu_device { bool is_atom_fw; uint8_t *bios; uint32_t bios_size; - struct amdgpu_bo *stollen_vga_memory; + struct amdgpu_bo *stolen_vga_memory; uint32_t bios_scratch_reg_offset; uint32_t bios_scratch[AMDGPU_BIOS_NUM_SCRATCH]; @@ -1557,6 +1494,10 @@ struct amdgpu_device { spinlock_t gc_cac_idx_lock; amdgpu_rreg_t gc_cac_rreg; amdgpu_wreg_t gc_cac_wreg; + /* protects concurrent se_cac register access */ + spinlock_t se_cac_idx_lock; + amdgpu_rreg_t se_cac_rreg; + amdgpu_wreg_t se_cac_wreg; /* protects concurrent ENDPOINT (audio) register access */ spinlock_t audio_endpt_idx_lock; amdgpu_block_rreg_t audio_endpt_rreg; @@ -1579,9 +1520,6 @@ struct amdgpu_device { struct amdgpu_mman mman; struct amdgpu_vram_scratch vram_scratch; struct amdgpu_wb wb; - atomic64_t vram_usage; - atomic64_t vram_vis_usage; - atomic64_t gtt_usage; atomic64_t num_bytes_moved; atomic64_t num_evictions; atomic64_t num_vram_cpu_page_faults; @@ -1593,12 +1531,14 @@ struct amdgpu_device { spinlock_t lock; s64 last_update_us; s64 accum_us; /* accumulated microseconds */ + s64 accum_us_vis; /* for visible VRAM */ u32 log2_max_MBps; } mm_stats; /* display */ bool enable_virtual_display; struct amdgpu_mode_info mode_info; + /* For pre-DCE11. DCE11 and later are in "struct amdgpu_device->dm" */ struct work_struct hotplug_work; struct amdgpu_irq_src crtc_irq; struct amdgpu_irq_src pageflip_irq; @@ -1616,7 +1556,6 @@ struct amdgpu_device { /* powerplay */ struct amd_powerplay powerplay; - bool pp_enabled; bool pp_force_state_enabled; /* dpm */ @@ -1655,6 +1594,9 @@ struct amdgpu_device { /* GDS */ struct amdgpu_gds gds; + /* display related functionality */ + struct amdgpu_display_manager dm; + struct amdgpu_ip_block ip_blocks[AMDGPU_MAX_IP_NUM]; int num_ip_blocks; struct mutex mn_lock; @@ -1672,6 +1614,8 @@ struct amdgpu_device { struct delayed_work late_init_work; struct amdgpu_virt virt; + /* firmware VRAM reservation */ + struct amdgpu_fw_vram_usage fw_vram_usage; /* link all shadow bo */ struct list_head shadow_list; @@ -1687,6 +1631,9 @@ struct amdgpu_device { bool has_hw_reset; u8 reset_magic[AMDGPU_RESET_MAGIC_NUM]; + /* record last mm index being written through WREG32*/ + unsigned long last_mm_index; + bool in_sriov_reset; }; static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_bo_device *bdev) @@ -1713,6 +1660,9 @@ void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v); u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index); void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v); +bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type); +bool amdgpu_device_has_dc_support(struct amdgpu_device *adev); + /* * Registers read & write functions. */ @@ -1742,6 +1692,8 @@ void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v); #define WREG32_DIDT(reg, v) adev->didt_wreg(adev, (reg), (v)) #define RREG32_GC_CAC(reg) adev->gc_cac_rreg(adev, (reg)) #define WREG32_GC_CAC(reg, v) adev->gc_cac_wreg(adev, (reg), (v)) +#define RREG32_SE_CAC(reg) adev->se_cac_rreg(adev, (reg)) +#define WREG32_SE_CAC(reg, v) adev->se_cac_wreg(adev, (reg), (v)) #define RREG32_AUDIO_ENDPT(block, reg) adev->audio_endpt_rreg(adev, (block), (reg)) #define WREG32_AUDIO_ENDPT(block, reg, v) adev->audio_endpt_wreg(adev, (block), (reg), (v)) #define WREG32_P(reg, val, mask) \ @@ -1792,50 +1744,6 @@ void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v); #define RBIOS16(i) (RBIOS8(i) | (RBIOS8((i)+1) << 8)) #define RBIOS32(i) ((RBIOS16(i)) | (RBIOS16((i)+2) << 16)) -/* - * RING helpers. - */ -static inline void amdgpu_ring_write(struct amdgpu_ring *ring, uint32_t v) -{ - if (ring->count_dw <= 0) - DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n"); - ring->ring[ring->wptr++ & ring->buf_mask] = v; - ring->wptr &= ring->ptr_mask; - ring->count_dw--; -} - -static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring, void *src, int count_dw) -{ - unsigned occupied, chunk1, chunk2; - void *dst; - - if (unlikely(ring->count_dw < count_dw)) { - DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n"); - return; - } - - occupied = ring->wptr & ring->buf_mask; - dst = (void *)&ring->ring[occupied]; - chunk1 = ring->buf_mask + 1 - occupied; - chunk1 = (chunk1 >= count_dw) ? count_dw: chunk1; - chunk2 = count_dw - chunk1; - chunk1 <<= 2; - chunk2 <<= 2; - - if (chunk1) - memcpy(dst, src, chunk1); - - if (chunk2) { - src += chunk1; - dst = (void *)ring->ring; - memcpy(dst, src, chunk2); - } - - ring->wptr += count_dw; - ring->wptr &= ring->ptr_mask; - ring->count_dw -= count_dw; -} - static inline struct amdgpu_sdma_instance * amdgpu_get_sdma_instance(struct amdgpu_ring *ring) { @@ -1896,9 +1804,9 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r)) #define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o)) #define amdgpu_ih_get_wptr(adev) (adev)->irq.ih_funcs->get_wptr((adev)) +#define amdgpu_ih_prescreen_iv(adev) (adev)->irq.ih_funcs->prescreen_iv((adev)) #define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv)) #define amdgpu_ih_set_rptr(adev) (adev)->irq.ih_funcs->set_rptr((adev)) -#define amdgpu_display_set_vga_render_state(adev, r) (adev)->mode_info.funcs->set_vga_render_state((adev), (r)) #define amdgpu_display_vblank_get_counter(adev, crtc) (adev)->mode_info.funcs->vblank_get_counter((adev), (crtc)) #define amdgpu_display_vblank_wait(adev, crtc) (adev)->mode_info.funcs->vblank_wait((adev), (crtc)) #define amdgpu_display_backlight_set_level(adev, e, l) (adev)->mode_info.funcs->backlight_set_level((e), (l)) @@ -1911,8 +1819,6 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_display_page_flip_get_scanoutpos(adev, crtc, vbl, pos) (adev)->mode_info.funcs->page_flip_get_scanoutpos((adev), (crtc), (vbl), (pos)) #define amdgpu_display_add_encoder(adev, e, s, c) (adev)->mode_info.funcs->add_encoder((adev), (e), (s), (c)) #define amdgpu_display_add_connector(adev, ci, sd, ct, ib, coi, h, r) (adev)->mode_info.funcs->add_connector((adev), (ci), (sd), (ct), (ib), (coi), (h), (r)) -#define amdgpu_display_stop_mc_access(adev, s) (adev)->mode_info.funcs->stop_mc_access((adev), (s)) -#define amdgpu_display_resume_mc_access(adev, s) (adev)->mode_info.funcs->resume_mc_access((adev), (s)) #define amdgpu_emit_copy_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_copy_buffer((ib), (s), (d), (b)) #define amdgpu_emit_fill_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_fill_buffer((ib), (s), (d), (b)) #define amdgpu_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev)) @@ -1927,23 +1833,12 @@ void amdgpu_pci_config_reset(struct amdgpu_device *adev); bool amdgpu_need_post(struct amdgpu_device *adev); void amdgpu_update_display_priority(struct amdgpu_device *adev); -void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes); +void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes, + u64 num_vis_bytes); void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain); bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo); -int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages); -int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, - uint32_t flags); -bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm); -struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm); -bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, - unsigned long end); -bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, - int *last_invalidated); -bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm); -uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, - struct ttm_mem_reg *mem); void amdgpu_vram_location(struct amdgpu_device *adev, struct amdgpu_mc *mc, u64 base); -void amdgpu_gtt_location(struct amdgpu_device *adev, struct amdgpu_mc *mc); +void amdgpu_gart_location(struct amdgpu_device *adev, struct amdgpu_mc *mc); void amdgpu_ttm_set_active_vram_size(struct amdgpu_device *adev, u64 size); int amdgpu_ttm_init(struct amdgpu_device *adev); void amdgpu_ttm_fini(struct amdgpu_device *adev); @@ -1975,8 +1870,6 @@ static inline bool amdgpu_has_atpx(void) { return false; } extern const struct drm_ioctl_desc amdgpu_ioctls_kms[]; extern const int amdgpu_max_kms_ioctl; -bool amdgpu_kms_vram_lost(struct amdgpu_device *adev, - struct amdgpu_fpriv *fpriv); int amdgpu_driver_load_kms(struct drm_device *dev, unsigned long flags); void amdgpu_driver_unload_kms(struct drm_device *dev); void amdgpu_driver_lastclose_kms(struct drm_device *dev); @@ -2024,10 +1917,15 @@ static inline int amdgpu_acpi_init(struct amdgpu_device *adev) { return 0; } static inline void amdgpu_acpi_fini(struct amdgpu_device *adev) { } #endif -struct amdgpu_bo_va_mapping * -amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, - uint64_t addr, struct amdgpu_bo **bo); -int amdgpu_cs_sysvm_access_required(struct amdgpu_cs_parser *parser); +int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, + uint64_t addr, struct amdgpu_bo **bo, + struct amdgpu_bo_va_mapping **mapping); + +#if defined(CONFIG_DRM_AMD_DC) +int amdgpu_dm_display_resume(struct amdgpu_device *adev ); +#else +static inline int amdgpu_dm_display_resume(struct amdgpu_device *adev) { return 0; } +#endif #include "amdgpu_object.h" #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c index 06879d1dcabd..c04f44a90392 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c @@ -35,41 +35,50 @@ #include "acp_gfx_if.h" -#define ACP_TILE_ON_MASK 0x03 -#define ACP_TILE_OFF_MASK 0x02 -#define ACP_TILE_ON_RETAIN_REG_MASK 0x1f -#define ACP_TILE_OFF_RETAIN_REG_MASK 0x20 - -#define ACP_TILE_P1_MASK 0x3e -#define ACP_TILE_P2_MASK 0x3d -#define ACP_TILE_DSP0_MASK 0x3b -#define ACP_TILE_DSP1_MASK 0x37 - -#define ACP_TILE_DSP2_MASK 0x2f - -#define ACP_DMA_REGS_END 0x146c0 -#define ACP_I2S_PLAY_REGS_START 0x14840 -#define ACP_I2S_PLAY_REGS_END 0x148b4 -#define ACP_I2S_CAP_REGS_START 0x148b8 -#define ACP_I2S_CAP_REGS_END 0x1496c - -#define ACP_I2S_COMP1_CAP_REG_OFFSET 0xac -#define ACP_I2S_COMP2_CAP_REG_OFFSET 0xa8 -#define ACP_I2S_COMP1_PLAY_REG_OFFSET 0x6c -#define ACP_I2S_COMP2_PLAY_REG_OFFSET 0x68 - -#define mmACP_PGFSM_RETAIN_REG 0x51c9 -#define mmACP_PGFSM_CONFIG_REG 0x51ca -#define mmACP_PGFSM_READ_REG_0 0x51cc - -#define mmACP_MEM_SHUT_DOWN_REQ_LO 0x51f8 -#define mmACP_MEM_SHUT_DOWN_REQ_HI 0x51f9 -#define mmACP_MEM_SHUT_DOWN_STS_LO 0x51fa -#define mmACP_MEM_SHUT_DOWN_STS_HI 0x51fb - -#define ACP_TIMEOUT_LOOP 0x000000FF -#define ACP_DEVS 3 -#define ACP_SRC_ID 162 +#define ACP_TILE_ON_MASK 0x03 +#define ACP_TILE_OFF_MASK 0x02 +#define ACP_TILE_ON_RETAIN_REG_MASK 0x1f +#define ACP_TILE_OFF_RETAIN_REG_MASK 0x20 + +#define ACP_TILE_P1_MASK 0x3e +#define ACP_TILE_P2_MASK 0x3d +#define ACP_TILE_DSP0_MASK 0x3b +#define ACP_TILE_DSP1_MASK 0x37 + +#define ACP_TILE_DSP2_MASK 0x2f + +#define ACP_DMA_REGS_END 0x146c0 +#define ACP_I2S_PLAY_REGS_START 0x14840 +#define ACP_I2S_PLAY_REGS_END 0x148b4 +#define ACP_I2S_CAP_REGS_START 0x148b8 +#define ACP_I2S_CAP_REGS_END 0x1496c + +#define ACP_I2S_COMP1_CAP_REG_OFFSET 0xac +#define ACP_I2S_COMP2_CAP_REG_OFFSET 0xa8 +#define ACP_I2S_COMP1_PLAY_REG_OFFSET 0x6c +#define ACP_I2S_COMP2_PLAY_REG_OFFSET 0x68 + +#define mmACP_PGFSM_RETAIN_REG 0x51c9 +#define mmACP_PGFSM_CONFIG_REG 0x51ca +#define mmACP_PGFSM_READ_REG_0 0x51cc + +#define mmACP_MEM_SHUT_DOWN_REQ_LO 0x51f8 +#define mmACP_MEM_SHUT_DOWN_REQ_HI 0x51f9 +#define mmACP_MEM_SHUT_DOWN_STS_LO 0x51fa +#define mmACP_MEM_SHUT_DOWN_STS_HI 0x51fb + +#define mmACP_CONTROL 0x5131 +#define mmACP_STATUS 0x5133 +#define mmACP_SOFT_RESET 0x5134 +#define ACP_CONTROL__ClkEn_MASK 0x1 +#define ACP_SOFT_RESET__SoftResetAud_MASK 0x100 +#define ACP_SOFT_RESET__SoftResetAudDone_MASK 0x1000000 +#define ACP_CLOCK_EN_TIME_OUT_VALUE 0x000000FF +#define ACP_SOFT_RESET_DONE_TIME_OUT_VALUE 0x000000FF + +#define ACP_TIMEOUT_LOOP 0x000000FF +#define ACP_DEVS 3 +#define ACP_SRC_ID 162 enum { ACP_TILE_P1 = 0, @@ -260,6 +269,8 @@ static int acp_hw_init(void *handle) { int r, i; uint64_t acp_base; + u32 val = 0; + u32 count = 0; struct device *dev; struct i2s_platform_data *i2s_pdata; @@ -285,19 +296,20 @@ static int acp_hw_init(void *handle) return 0; else if (r) return r; + if (adev->asic_type != CHIP_STONEY) { + adev->acp.acp_genpd = kzalloc(sizeof(struct acp_pm_domain), GFP_KERNEL); + if (adev->acp.acp_genpd == NULL) + return -ENOMEM; - adev->acp.acp_genpd = kzalloc(sizeof(struct acp_pm_domain), GFP_KERNEL); - if (adev->acp.acp_genpd == NULL) - return -ENOMEM; - - adev->acp.acp_genpd->gpd.name = "ACP_AUDIO"; - adev->acp.acp_genpd->gpd.power_off = acp_poweroff; - adev->acp.acp_genpd->gpd.power_on = acp_poweron; + adev->acp.acp_genpd->gpd.name = "ACP_AUDIO"; + adev->acp.acp_genpd->gpd.power_off = acp_poweroff; + adev->acp.acp_genpd->gpd.power_on = acp_poweron; - adev->acp.acp_genpd->cgs_dev = adev->acp.cgs_device; + adev->acp.acp_genpd->cgs_dev = adev->acp.cgs_device; - pm_genpd_init(&adev->acp.acp_genpd->gpd, NULL, false); + pm_genpd_init(&adev->acp.acp_genpd->gpd, NULL, false); + } adev->acp.acp_cell = kzalloc(sizeof(struct mfd_cell) * ACP_DEVS, GFP_KERNEL); @@ -319,14 +331,29 @@ static int acp_hw_init(void *handle) return -ENOMEM; } - i2s_pdata[0].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET; + switch (adev->asic_type) { + case CHIP_STONEY: + i2s_pdata[0].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET | + DW_I2S_QUIRK_16BIT_IDX_OVERRIDE; + break; + default: + i2s_pdata[0].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET; + } i2s_pdata[0].cap = DWC_I2S_PLAY; i2s_pdata[0].snd_rates = SNDRV_PCM_RATE_8000_96000; i2s_pdata[0].i2s_reg_comp1 = ACP_I2S_COMP1_PLAY_REG_OFFSET; i2s_pdata[0].i2s_reg_comp2 = ACP_I2S_COMP2_PLAY_REG_OFFSET; + switch (adev->asic_type) { + case CHIP_STONEY: + i2s_pdata[1].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET | + DW_I2S_QUIRK_COMP_PARAM1 | + DW_I2S_QUIRK_16BIT_IDX_OVERRIDE; + break; + default: + i2s_pdata[1].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET | + DW_I2S_QUIRK_COMP_PARAM1; + } - i2s_pdata[1].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET | - DW_I2S_QUIRK_COMP_PARAM1; i2s_pdata[1].cap = DWC_I2S_RECORD; i2s_pdata[1].snd_rates = SNDRV_PCM_RATE_8000_96000; i2s_pdata[1].i2s_reg_comp1 = ACP_I2S_COMP1_CAP_REG_OFFSET; @@ -355,6 +382,8 @@ static int acp_hw_init(void *handle) adev->acp.acp_cell[0].name = "acp_audio_dma"; adev->acp.acp_cell[0].num_resources = 4; adev->acp.acp_cell[0].resources = &adev->acp.acp_res[0]; + adev->acp.acp_cell[0].platform_data = &adev->asic_type; + adev->acp.acp_cell[0].pdata_size = sizeof(adev->asic_type); adev->acp.acp_cell[1].name = "designware-i2s"; adev->acp.acp_cell[1].num_resources = 1; @@ -373,14 +402,56 @@ static int acp_hw_init(void *handle) if (r) return r; - for (i = 0; i < ACP_DEVS ; i++) { - dev = get_mfd_cell_dev(adev->acp.acp_cell[i].name, i); - r = pm_genpd_add_device(&adev->acp.acp_genpd->gpd, dev); - if (r) { - dev_err(dev, "Failed to add dev to genpd\n"); - return r; + if (adev->asic_type != CHIP_STONEY) { + for (i = 0; i < ACP_DEVS ; i++) { + dev = get_mfd_cell_dev(adev->acp.acp_cell[i].name, i); + r = pm_genpd_add_device(&adev->acp.acp_genpd->gpd, dev); + if (r) { + dev_err(dev, "Failed to add dev to genpd\n"); + return r; + } + } + } + + /* Assert Soft reset of ACP */ + val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET); + + val |= ACP_SOFT_RESET__SoftResetAud_MASK; + cgs_write_register(adev->acp.cgs_device, mmACP_SOFT_RESET, val); + + count = ACP_SOFT_RESET_DONE_TIME_OUT_VALUE; + while (true) { + val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET); + if (ACP_SOFT_RESET__SoftResetAudDone_MASK == + (val & ACP_SOFT_RESET__SoftResetAudDone_MASK)) + break; + if (--count == 0) { + dev_err(&adev->pdev->dev, "Failed to reset ACP\n"); + return -ETIMEDOUT; } + udelay(100); } + /* Enable clock to ACP and wait until the clock is enabled */ + val = cgs_read_register(adev->acp.cgs_device, mmACP_CONTROL); + val = val | ACP_CONTROL__ClkEn_MASK; + cgs_write_register(adev->acp.cgs_device, mmACP_CONTROL, val); + + count = ACP_CLOCK_EN_TIME_OUT_VALUE; + + while (true) { + val = cgs_read_register(adev->acp.cgs_device, mmACP_STATUS); + if (val & (u32) 0x1) + break; + if (--count == 0) { + dev_err(&adev->pdev->dev, "Failed to reset ACP\n"); + return -ETIMEDOUT; + } + udelay(100); + } + /* Deassert the SOFT RESET flags */ + val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET); + val &= ~ACP_SOFT_RESET__SoftResetAud_MASK; + cgs_write_register(adev->acp.cgs_device, mmACP_SOFT_RESET, val); return 0; } @@ -394,24 +465,64 @@ static int acp_hw_init(void *handle) static int acp_hw_fini(void *handle) { int i, ret; + u32 val = 0; + u32 count = 0; struct device *dev; struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* return early if no ACP */ - if (!adev->acp.acp_genpd) + if (!adev->acp.acp_cell) return 0; - for (i = 0; i < ACP_DEVS ; i++) { - dev = get_mfd_cell_dev(adev->acp.acp_cell[i].name, i); - ret = pm_genpd_remove_device(&adev->acp.acp_genpd->gpd, dev); - /* If removal fails, dont giveup and try rest */ - if (ret) - dev_err(dev, "remove dev from genpd failed\n"); + /* Assert Soft reset of ACP */ + val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET); + + val |= ACP_SOFT_RESET__SoftResetAud_MASK; + cgs_write_register(adev->acp.cgs_device, mmACP_SOFT_RESET, val); + + count = ACP_SOFT_RESET_DONE_TIME_OUT_VALUE; + while (true) { + val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET); + if (ACP_SOFT_RESET__SoftResetAudDone_MASK == + (val & ACP_SOFT_RESET__SoftResetAudDone_MASK)) + break; + if (--count == 0) { + dev_err(&adev->pdev->dev, "Failed to reset ACP\n"); + return -ETIMEDOUT; + } + udelay(100); + } + /* Disable ACP clock */ + val = cgs_read_register(adev->acp.cgs_device, mmACP_CONTROL); + val &= ~ACP_CONTROL__ClkEn_MASK; + cgs_write_register(adev->acp.cgs_device, mmACP_CONTROL, val); + + count = ACP_CLOCK_EN_TIME_OUT_VALUE; + + while (true) { + val = cgs_read_register(adev->acp.cgs_device, mmACP_STATUS); + if (val & (u32) 0x1) + break; + if (--count == 0) { + dev_err(&adev->pdev->dev, "Failed to reset ACP\n"); + return -ETIMEDOUT; + } + udelay(100); + } + + if (adev->acp.acp_genpd) { + for (i = 0; i < ACP_DEVS ; i++) { + dev = get_mfd_cell_dev(adev->acp.acp_cell[i].name, i); + ret = pm_genpd_remove_device(&adev->acp.acp_genpd->gpd, dev); + /* If removal fails, dont giveup and try rest */ + if (ret) + dev_err(dev, "remove dev from genpd failed\n"); + } + kfree(adev->acp.acp_genpd); } mfd_remove_devices(adev->acp.parent); kfree(adev->acp.acp_res); - kfree(adev->acp.acp_genpd); kfree(adev->acp.acp_cell); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c index ef79551b4cb7..57afad79f55d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c @@ -30,10 +30,10 @@ #include <drm/drmP.h> #include <drm/drm_crtc_helper.h> #include "amdgpu.h" +#include "amdgpu_pm.h" #include "amd_acpi.h" #include "atom.h" -extern void amdgpu_pm_acpi_event_handler(struct amdgpu_device *adev); /* Call the ATIF method */ /** @@ -289,7 +289,7 @@ out: * handles it. * Returns NOTIFY code */ -int amdgpu_atif_handler(struct amdgpu_device *adev, +static int amdgpu_atif_handler(struct amdgpu_device *adev, struct acpi_bus_event *event) { struct amdgpu_atif *atif = &adev->atif; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 37971d9402e3..5432af39a674 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -27,16 +27,15 @@ #include "amdgpu_gfx.h" #include <linux/module.h> -const struct kfd2kgd_calls *kfd2kgd; const struct kgd2kfd_calls *kgd2kfd; -bool (*kgd2kfd_init_p)(unsigned, const struct kgd2kfd_calls**); +bool (*kgd2kfd_init_p)(unsigned int, const struct kgd2kfd_calls**); int amdgpu_amdkfd_init(void) { int ret; #if defined(CONFIG_HSA_AMD_MODULE) - int (*kgd2kfd_init_p)(unsigned, const struct kgd2kfd_calls**); + int (*kgd2kfd_init_p)(unsigned int, const struct kgd2kfd_calls**); kgd2kfd_init_p = symbol_request(kgd2kfd_init); @@ -61,8 +60,21 @@ int amdgpu_amdkfd_init(void) return ret; } -bool amdgpu_amdkfd_load_interface(struct amdgpu_device *adev) +void amdgpu_amdkfd_fini(void) +{ + if (kgd2kfd) { + kgd2kfd->exit(); + symbol_put(kgd2kfd_init); + } +} + +void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) { + const struct kfd2kgd_calls *kfd2kgd; + + if (!kgd2kfd) + return; + switch (adev->asic_type) { #ifdef CONFIG_DRM_AMDGPU_CIK case CHIP_KAVERI: @@ -73,25 +85,12 @@ bool amdgpu_amdkfd_load_interface(struct amdgpu_device *adev) kfd2kgd = amdgpu_amdkfd_gfx_8_0_get_functions(); break; default: - return false; + dev_info(adev->dev, "kfd not supported on this ASIC\n"); + return; } - return true; -} - -void amdgpu_amdkfd_fini(void) -{ - if (kgd2kfd) { - kgd2kfd->exit(); - symbol_put(kgd2kfd_init); - } -} - -void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) -{ - if (kgd2kfd) - adev->kfd = kgd2kfd->probe((struct kgd_dev *)adev, - adev->pdev, kfd2kgd); + adev->kfd = kgd2kfd->probe((struct kgd_dev *)adev, + adev->pdev, kfd2kgd); } void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) @@ -184,7 +183,8 @@ int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, return -ENOMEM; r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_CPU_GTT_USWC, NULL, NULL, &(*mem)->bo); + AMDGPU_GEM_CREATE_CPU_GTT_USWC, NULL, NULL, 0, + &(*mem)->bo); if (r) { dev_err(adev->dev, "failed to allocate BO for amdkfd (%d)\n", r); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index 73f83a10ae14..8d689ab7e429 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -26,6 +26,7 @@ #define AMDGPU_AMDKFD_H_INCLUDED #include <linux/types.h> +#include <linux/mmu_context.h> #include <kgd_kfd_interface.h> struct amdgpu_device; @@ -39,8 +40,6 @@ struct kgd_mem { int amdgpu_amdkfd_init(void); void amdgpu_amdkfd_fini(void); -bool amdgpu_amdkfd_load_interface(struct amdgpu_device *adev); - void amdgpu_amdkfd_suspend(struct amdgpu_device *adev); int amdgpu_amdkfd_resume(struct amdgpu_device *adev); void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev, @@ -62,4 +61,19 @@ uint64_t get_gpu_clock_counter(struct kgd_dev *kgd); uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd); +#define read_user_wptr(mmptr, wptr, dst) \ + ({ \ + bool valid = false; \ + if ((mmptr) && (wptr)) { \ + if ((mmptr) == current->mm) { \ + valid = !get_user((dst), (wptr)); \ + } else if (current->mm == NULL) { \ + use_mm(mmptr); \ + valid = !get_user((dst), (wptr)); \ + unuse_mm(mmptr); \ + } \ + } \ + valid; \ + }) + #endif /* AMDGPU_AMDKFD_H_INCLUDED */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index 5254562fd0f9..47d1c132ac40 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -39,6 +39,12 @@ #include "gmc/gmc_7_1_sh_mask.h" #include "cik_structs.h" +enum hqd_dequeue_request_type { + NO_ACTION = 0, + DRAIN_PIPE, + RESET_WAVES +}; + enum { MAX_TRAPID = 8, /* 3 bits in the bitfield. */ MAX_WATCH_ADDRESSES = 4 @@ -96,12 +102,15 @@ static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id, uint32_t hpd_size, uint64_t hpd_gpu_addr); static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id); static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, - uint32_t queue_id, uint32_t __user *wptr); + uint32_t queue_id, uint32_t __user *wptr, + uint32_t wptr_shift, uint32_t wptr_mask, + struct mm_struct *mm); static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd); static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id); -static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type, +static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, + enum kfd_preempt_type reset_type, unsigned int utimeout, uint32_t pipe_id, uint32_t queue_id); static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd); @@ -126,6 +135,33 @@ static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd, static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid); static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type); +static void set_scratch_backing_va(struct kgd_dev *kgd, + uint64_t va, uint32_t vmid); + +/* Because of REG_GET_FIELD() being used, we put this function in the + * asic specific file. + */ +static int get_tile_config(struct kgd_dev *kgd, + struct tile_config *config) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + + config->gb_addr_config = adev->gfx.config.gb_addr_config; + config->num_banks = REG_GET_FIELD(adev->gfx.config.mc_arb_ramcfg, + MC_ARB_RAMCFG, NOOFBANK); + config->num_ranks = REG_GET_FIELD(adev->gfx.config.mc_arb_ramcfg, + MC_ARB_RAMCFG, NOOFRANKS); + + config->tile_config_ptr = adev->gfx.config.tile_mode_array; + config->num_tile_configs = + ARRAY_SIZE(adev->gfx.config.tile_mode_array); + config->macro_tile_config_ptr = + adev->gfx.config.macrotile_mode_array; + config->num_macro_tile_configs = + ARRAY_SIZE(adev->gfx.config.macrotile_mode_array); + + return 0; +} static const struct kfd2kgd_calls kfd2kgd = { .init_gtt_mem_allocation = alloc_gtt_mem, @@ -133,6 +169,8 @@ static const struct kfd2kgd_calls kfd2kgd = { .get_vmem_size = get_vmem_size, .get_gpu_clock_counter = get_gpu_clock_counter, .get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz, + .alloc_pasid = amdgpu_vm_alloc_pasid, + .free_pasid = amdgpu_vm_free_pasid, .program_sh_mem_settings = kgd_program_sh_mem_settings, .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping, .init_pipeline = kgd_init_pipeline, @@ -150,7 +188,9 @@ static const struct kfd2kgd_calls kfd2kgd = { .get_atc_vmid_pasid_mapping_pasid = get_atc_vmid_pasid_mapping_pasid, .get_atc_vmid_pasid_mapping_valid = get_atc_vmid_pasid_mapping_valid, .write_vmid_invalidate_request = write_vmid_invalidate_request, - .get_fw_version = get_fw_version + .get_fw_version = get_fw_version, + .set_scratch_backing_va = set_scratch_backing_va, + .get_tile_config = get_tile_config, }; struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void) @@ -186,7 +226,7 @@ static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id, { struct amdgpu_device *adev = get_amdgpu_device(kgd); - uint32_t mec = (++pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; + uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); lock_srbm(kgd, mec, pipe, queue_id, 0); @@ -290,20 +330,46 @@ static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd) } static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, - uint32_t queue_id, uint32_t __user *wptr) + uint32_t queue_id, uint32_t __user *wptr, + uint32_t wptr_shift, uint32_t wptr_mask, + struct mm_struct *mm) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - uint32_t wptr_shadow, is_wptr_shadow_valid; struct cik_mqd *m; + uint32_t *mqd_hqd; + uint32_t reg, wptr_val, data; + bool valid_wptr = false; m = get_mqd(mqd); - is_wptr_shadow_valid = !get_user(wptr_shadow, wptr); - if (is_wptr_shadow_valid) - m->cp_hqd_pq_wptr = wptr_shadow; + acquire_queue(kgd, pipe_id, queue_id); + + /* HQD registers extend from CP_MQD_BASE_ADDR to CP_MQD_CONTROL. */ + mqd_hqd = &m->cp_mqd_base_addr_lo; + + for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_MQD_CONTROL; reg++) + WREG32(reg, mqd_hqd[reg - mmCP_MQD_BASE_ADDR]); + + /* Copy userspace write pointer value to register. + * Activate doorbell logic to monitor subsequent changes. + */ + data = REG_SET_FIELD(m->cp_hqd_pq_doorbell_control, + CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1); + WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, data); + /* read_user_ptr may take the mm->mmap_sem. + * release srbm_mutex to avoid circular dependency between + * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex. + */ + release_queue(kgd); + valid_wptr = read_user_wptr(mm, wptr, wptr_val); acquire_queue(kgd, pipe_id, queue_id); - gfx_v7_0_mqd_commit(adev, m); + if (valid_wptr) + WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask); + + data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1); + WREG32(mmCP_HQD_ACTIVE, data); + release_queue(kgd); return 0; @@ -382,30 +448,99 @@ static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) return false; } -static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type, +static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, + enum kfd_preempt_type reset_type, unsigned int utimeout, uint32_t pipe_id, uint32_t queue_id) { struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t temp; - int timeout = utimeout; + enum hqd_dequeue_request_type type; + unsigned long flags, end_jiffies; + int retry; acquire_queue(kgd, pipe_id, queue_id); WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, 0); - WREG32(mmCP_HQD_DEQUEUE_REQUEST, reset_type); + switch (reset_type) { + case KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN: + type = DRAIN_PIPE; + break; + case KFD_PREEMPT_TYPE_WAVEFRONT_RESET: + type = RESET_WAVES; + break; + default: + type = DRAIN_PIPE; + break; + } + /* Workaround: If IQ timer is active and the wait time is close to or + * equal to 0, dequeueing is not safe. Wait until either the wait time + * is larger or timer is cleared. Also, ensure that IQ_REQ_PEND is + * cleared before continuing. Also, ensure wait times are set to at + * least 0x3. + */ + local_irq_save(flags); + preempt_disable(); + retry = 5000; /* wait for 500 usecs at maximum */ + while (true) { + temp = RREG32(mmCP_HQD_IQ_TIMER); + if (REG_GET_FIELD(temp, CP_HQD_IQ_TIMER, PROCESSING_IQ)) { + pr_debug("HW is processing IQ\n"); + goto loop; + } + if (REG_GET_FIELD(temp, CP_HQD_IQ_TIMER, ACTIVE)) { + if (REG_GET_FIELD(temp, CP_HQD_IQ_TIMER, RETRY_TYPE) + == 3) /* SEM-rearm is safe */ + break; + /* Wait time 3 is safe for CP, but our MMIO read/write + * time is close to 1 microsecond, so check for 10 to + * leave more buffer room + */ + if (REG_GET_FIELD(temp, CP_HQD_IQ_TIMER, WAIT_TIME) + >= 10) + break; + pr_debug("IQ timer is active\n"); + } else + break; +loop: + if (!retry) { + pr_err("CP HQD IQ timer status time out\n"); + break; + } + ndelay(100); + --retry; + } + retry = 1000; + while (true) { + temp = RREG32(mmCP_HQD_DEQUEUE_REQUEST); + if (!(temp & CP_HQD_DEQUEUE_REQUEST__IQ_REQ_PEND_MASK)) + break; + pr_debug("Dequeue request is pending\n"); + + if (!retry) { + pr_err("CP HQD dequeue request time out\n"); + break; + } + ndelay(100); + --retry; + } + local_irq_restore(flags); + preempt_enable(); + + WREG32(mmCP_HQD_DEQUEUE_REQUEST, type); + + end_jiffies = (utimeout * HZ / 1000) + jiffies; while (true) { temp = RREG32(mmCP_HQD_ACTIVE); - if (temp & CP_HQD_ACTIVE__ACTIVE_MASK) + if (!(temp & CP_HQD_ACTIVE__ACTIVE_MASK)) break; - if (timeout <= 0) { - pr_err("kfd: cp queue preemption time out.\n"); + if (time_after(jiffies, end_jiffies)) { + pr_err("cp queue preemption time out\n"); release_queue(kgd); return -ETIME; } - msleep(20); - timeout -= 20; + usleep_range(500, 1000); } release_queue(kgd); @@ -556,6 +691,16 @@ static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid) WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid); } +static void set_scratch_backing_va(struct kgd_dev *kgd, + uint64_t va, uint32_t vmid) +{ + struct amdgpu_device *adev = (struct amdgpu_device *) kgd; + + lock_srbm(kgd, 0, 0, 0, vmid); + WREG32(mmSH_HIDDEN_PRIVATE_BASE_VMID, va); + unlock_srbm(kgd); +} + static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type) { struct amdgpu_device *adev = (struct amdgpu_device *) kgd; @@ -566,42 +711,42 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type) switch (type) { case KGD_ENGINE_PFP: hdr = (const union amdgpu_firmware_header *) - adev->gfx.pfp_fw->data; + adev->gfx.pfp_fw->data; break; case KGD_ENGINE_ME: hdr = (const union amdgpu_firmware_header *) - adev->gfx.me_fw->data; + adev->gfx.me_fw->data; break; case KGD_ENGINE_CE: hdr = (const union amdgpu_firmware_header *) - adev->gfx.ce_fw->data; + adev->gfx.ce_fw->data; break; case KGD_ENGINE_MEC1: hdr = (const union amdgpu_firmware_header *) - adev->gfx.mec_fw->data; + adev->gfx.mec_fw->data; break; case KGD_ENGINE_MEC2: hdr = (const union amdgpu_firmware_header *) - adev->gfx.mec2_fw->data; + adev->gfx.mec2_fw->data; break; case KGD_ENGINE_RLC: hdr = (const union amdgpu_firmware_header *) - adev->gfx.rlc_fw->data; + adev->gfx.rlc_fw->data; break; case KGD_ENGINE_SDMA1: hdr = (const union amdgpu_firmware_header *) - adev->sdma.instance[0].fw->data; + adev->sdma.instance[0].fw->data; break; case KGD_ENGINE_SDMA2: hdr = (const union amdgpu_firmware_header *) - adev->sdma.instance[1].fw->data; + adev->sdma.instance[1].fw->data; break; default: diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index 133d06671e46..056929b8ccd0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -39,6 +39,12 @@ #include "vi_structs.h" #include "vid.h" +enum hqd_dequeue_request_type { + NO_ACTION = 0, + DRAIN_PIPE, + RESET_WAVES +}; + struct cik_sdma_rlc_registers; /* @@ -55,12 +61,15 @@ static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id, uint32_t hpd_size, uint64_t hpd_gpu_addr); static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id); static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, - uint32_t queue_id, uint32_t __user *wptr); + uint32_t queue_id, uint32_t __user *wptr, + uint32_t wptr_shift, uint32_t wptr_mask, + struct mm_struct *mm); static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd); static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id); static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd); -static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type, +static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, + enum kfd_preempt_type reset_type, unsigned int utimeout, uint32_t pipe_id, uint32_t queue_id); static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, @@ -85,6 +94,33 @@ static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd, uint8_t vmid); static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid); static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type); +static void set_scratch_backing_va(struct kgd_dev *kgd, + uint64_t va, uint32_t vmid); + +/* Because of REG_GET_FIELD() being used, we put this function in the + * asic specific file. + */ +static int get_tile_config(struct kgd_dev *kgd, + struct tile_config *config) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + + config->gb_addr_config = adev->gfx.config.gb_addr_config; + config->num_banks = REG_GET_FIELD(adev->gfx.config.mc_arb_ramcfg, + MC_ARB_RAMCFG, NOOFBANK); + config->num_ranks = REG_GET_FIELD(adev->gfx.config.mc_arb_ramcfg, + MC_ARB_RAMCFG, NOOFRANKS); + + config->tile_config_ptr = adev->gfx.config.tile_mode_array; + config->num_tile_configs = + ARRAY_SIZE(adev->gfx.config.tile_mode_array); + config->macro_tile_config_ptr = + adev->gfx.config.macrotile_mode_array; + config->num_macro_tile_configs = + ARRAY_SIZE(adev->gfx.config.macrotile_mode_array); + + return 0; +} static const struct kfd2kgd_calls kfd2kgd = { .init_gtt_mem_allocation = alloc_gtt_mem, @@ -92,6 +128,8 @@ static const struct kfd2kgd_calls kfd2kgd = { .get_vmem_size = get_vmem_size, .get_gpu_clock_counter = get_gpu_clock_counter, .get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz, + .alloc_pasid = amdgpu_vm_alloc_pasid, + .free_pasid = amdgpu_vm_free_pasid, .program_sh_mem_settings = kgd_program_sh_mem_settings, .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping, .init_pipeline = kgd_init_pipeline, @@ -111,7 +149,9 @@ static const struct kfd2kgd_calls kfd2kgd = { .get_atc_vmid_pasid_mapping_valid = get_atc_vmid_pasid_mapping_valid, .write_vmid_invalidate_request = write_vmid_invalidate_request, - .get_fw_version = get_fw_version + .get_fw_version = get_fw_version, + .set_scratch_backing_va = set_scratch_backing_va, + .get_tile_config = get_tile_config, }; struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void) @@ -147,7 +187,7 @@ static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id, { struct amdgpu_device *adev = get_amdgpu_device(kgd); - uint32_t mec = (++pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; + uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); lock_srbm(kgd, mec, pipe, queue_id, 0); @@ -216,7 +256,7 @@ static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id) uint32_t mec; uint32_t pipe; - mec = (++pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; + mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); lock_srbm(kgd, mec, pipe, 0, 0); @@ -244,20 +284,75 @@ static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd) } static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, - uint32_t queue_id, uint32_t __user *wptr) + uint32_t queue_id, uint32_t __user *wptr, + uint32_t wptr_shift, uint32_t wptr_mask, + struct mm_struct *mm) { - struct vi_mqd *m; - uint32_t shadow_wptr, valid_wptr; struct amdgpu_device *adev = get_amdgpu_device(kgd); + struct vi_mqd *m; + uint32_t *mqd_hqd; + uint32_t reg, wptr_val, data; + bool valid_wptr = false; m = get_mqd(mqd); - valid_wptr = copy_from_user(&shadow_wptr, wptr, sizeof(shadow_wptr)); - if (valid_wptr == 0) - m->cp_hqd_pq_wptr = shadow_wptr; + acquire_queue(kgd, pipe_id, queue_id); + + /* HIQ is set during driver init period with vmid set to 0*/ + if (m->cp_hqd_vmid == 0) { + uint32_t value, mec, pipe; + + mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; + pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); + + pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n", + mec, pipe, queue_id); + value = RREG32(mmRLC_CP_SCHEDULERS); + value = REG_SET_FIELD(value, RLC_CP_SCHEDULERS, scheduler1, + ((mec << 5) | (pipe << 3) | queue_id | 0x80)); + WREG32(mmRLC_CP_SCHEDULERS, value); + } + /* HQD registers extend from CP_MQD_BASE_ADDR to CP_HQD_EOP_WPTR_MEM. */ + mqd_hqd = &m->cp_mqd_base_addr_lo; + + for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_HQD_EOP_CONTROL; reg++) + WREG32(reg, mqd_hqd[reg - mmCP_MQD_BASE_ADDR]); + + /* Tonga errata: EOP RPTR/WPTR should be left unmodified. + * This is safe since EOP RPTR==WPTR for any inactive HQD + * on ASICs that do not support context-save. + * EOP writes/reads can start anywhere in the ring. + */ + if (get_amdgpu_device(kgd)->asic_type != CHIP_TONGA) { + WREG32(mmCP_HQD_EOP_RPTR, m->cp_hqd_eop_rptr); + WREG32(mmCP_HQD_EOP_WPTR, m->cp_hqd_eop_wptr); + WREG32(mmCP_HQD_EOP_WPTR_MEM, m->cp_hqd_eop_wptr_mem); + } + + for (reg = mmCP_HQD_EOP_EVENTS; reg <= mmCP_HQD_ERROR; reg++) + WREG32(reg, mqd_hqd[reg - mmCP_MQD_BASE_ADDR]); + + /* Copy userspace write pointer value to register. + * Activate doorbell logic to monitor subsequent changes. + */ + data = REG_SET_FIELD(m->cp_hqd_pq_doorbell_control, + CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1); + WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, data); + + /* read_user_ptr may take the mm->mmap_sem. + * release srbm_mutex to avoid circular dependency between + * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex. + */ + release_queue(kgd); + valid_wptr = read_user_wptr(mm, wptr, wptr_val); acquire_queue(kgd, pipe_id, queue_id); - gfx_v8_0_mqd_commit(adev, mqd); + if (valid_wptr) + WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask); + + data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1); + WREG32(mmCP_HQD_ACTIVE, data); + release_queue(kgd); return 0; @@ -308,29 +403,102 @@ static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) return false; } -static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type, +static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, + enum kfd_preempt_type reset_type, unsigned int utimeout, uint32_t pipe_id, uint32_t queue_id) { struct amdgpu_device *adev = get_amdgpu_device(kgd); uint32_t temp; - int timeout = utimeout; + enum hqd_dequeue_request_type type; + unsigned long flags, end_jiffies; + int retry; + struct vi_mqd *m = get_mqd(mqd); acquire_queue(kgd, pipe_id, queue_id); - WREG32(mmCP_HQD_DEQUEUE_REQUEST, reset_type); + if (m->cp_hqd_vmid == 0) + WREG32_FIELD(RLC_CP_SCHEDULERS, scheduler1, 0); + + switch (reset_type) { + case KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN: + type = DRAIN_PIPE; + break; + case KFD_PREEMPT_TYPE_WAVEFRONT_RESET: + type = RESET_WAVES; + break; + default: + type = DRAIN_PIPE; + break; + } + /* Workaround: If IQ timer is active and the wait time is close to or + * equal to 0, dequeueing is not safe. Wait until either the wait time + * is larger or timer is cleared. Also, ensure that IQ_REQ_PEND is + * cleared before continuing. Also, ensure wait times are set to at + * least 0x3. + */ + local_irq_save(flags); + preempt_disable(); + retry = 5000; /* wait for 500 usecs at maximum */ + while (true) { + temp = RREG32(mmCP_HQD_IQ_TIMER); + if (REG_GET_FIELD(temp, CP_HQD_IQ_TIMER, PROCESSING_IQ)) { + pr_debug("HW is processing IQ\n"); + goto loop; + } + if (REG_GET_FIELD(temp, CP_HQD_IQ_TIMER, ACTIVE)) { + if (REG_GET_FIELD(temp, CP_HQD_IQ_TIMER, RETRY_TYPE) + == 3) /* SEM-rearm is safe */ + break; + /* Wait time 3 is safe for CP, but our MMIO read/write + * time is close to 1 microsecond, so check for 10 to + * leave more buffer room + */ + if (REG_GET_FIELD(temp, CP_HQD_IQ_TIMER, WAIT_TIME) + >= 10) + break; + pr_debug("IQ timer is active\n"); + } else + break; +loop: + if (!retry) { + pr_err("CP HQD IQ timer status time out\n"); + break; + } + ndelay(100); + --retry; + } + retry = 1000; + while (true) { + temp = RREG32(mmCP_HQD_DEQUEUE_REQUEST); + if (!(temp & CP_HQD_DEQUEUE_REQUEST__IQ_REQ_PEND_MASK)) + break; + pr_debug("Dequeue request is pending\n"); + + if (!retry) { + pr_err("CP HQD dequeue request time out\n"); + break; + } + ndelay(100); + --retry; + } + local_irq_restore(flags); + preempt_enable(); + + WREG32(mmCP_HQD_DEQUEUE_REQUEST, type); + + end_jiffies = (utimeout * HZ / 1000) + jiffies; while (true) { temp = RREG32(mmCP_HQD_ACTIVE); - if (temp & CP_HQD_ACTIVE__ACTIVE_MASK) + if (!(temp & CP_HQD_ACTIVE__ACTIVE_MASK)) break; - if (timeout <= 0) { - pr_err("kfd: cp queue preemption time out.\n"); + if (time_after(jiffies, end_jiffies)) { + pr_err("cp queue preemption time out.\n"); release_queue(kgd); return -ETIME; } - msleep(20); - timeout -= 20; + usleep_range(500, 1000); } release_queue(kgd); @@ -444,6 +612,16 @@ static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd, return 0; } +static void set_scratch_backing_va(struct kgd_dev *kgd, + uint64_t va, uint32_t vmid) +{ + struct amdgpu_device *adev = (struct amdgpu_device *) kgd; + + lock_srbm(kgd, 0, 0, 0, vmid); + WREG32(mmSH_HIDDEN_PRIVATE_BASE_VMID, va); + unlock_srbm(kgd); +} + static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type) { struct amdgpu_device *adev = (struct amdgpu_device *) kgd; @@ -454,42 +632,42 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type) switch (type) { case KGD_ENGINE_PFP: hdr = (const union amdgpu_firmware_header *) - adev->gfx.pfp_fw->data; + adev->gfx.pfp_fw->data; break; case KGD_ENGINE_ME: hdr = (const union amdgpu_firmware_header *) - adev->gfx.me_fw->data; + adev->gfx.me_fw->data; break; case KGD_ENGINE_CE: hdr = (const union amdgpu_firmware_header *) - adev->gfx.ce_fw->data; + adev->gfx.ce_fw->data; break; case KGD_ENGINE_MEC1: hdr = (const union amdgpu_firmware_header *) - adev->gfx.mec_fw->data; + adev->gfx.mec_fw->data; break; case KGD_ENGINE_MEC2: hdr = (const union amdgpu_firmware_header *) - adev->gfx.mec2_fw->data; + adev->gfx.mec2_fw->data; break; case KGD_ENGINE_RLC: hdr = (const union amdgpu_firmware_header *) - adev->gfx.rlc_fw->data; + adev->gfx.rlc_fw->data; break; case KGD_ENGINE_SDMA1: hdr = (const union amdgpu_firmware_header *) - adev->sdma.instance[0].fw->data; + adev->sdma.instance[0].fw->data; break; case KGD_ENGINE_SDMA2: hdr = (const union amdgpu_firmware_header *) - adev->sdma.instance[1].fw->data; + adev->sdma.instance[1].fw->data; break; default: diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c index 1e8e1123ddf4..f450b69323fa 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c @@ -1686,7 +1686,7 @@ void amdgpu_atombios_scratch_regs_lock(struct amdgpu_device *adev, bool lock) { uint32_t bios_6_scratch; - bios_6_scratch = RREG32(mmBIOS_SCRATCH_6); + bios_6_scratch = RREG32(adev->bios_scratch_reg_offset + 6); if (lock) { bios_6_scratch |= ATOM_S6_CRITICAL_STATE; @@ -1696,15 +1696,17 @@ void amdgpu_atombios_scratch_regs_lock(struct amdgpu_device *adev, bool lock) bios_6_scratch |= ATOM_S6_ACC_MODE; } - WREG32(mmBIOS_SCRATCH_6, bios_6_scratch); + WREG32(adev->bios_scratch_reg_offset + 6, bios_6_scratch); } void amdgpu_atombios_scratch_regs_init(struct amdgpu_device *adev) { uint32_t bios_2_scratch, bios_6_scratch; - bios_2_scratch = RREG32(mmBIOS_SCRATCH_2); - bios_6_scratch = RREG32(mmBIOS_SCRATCH_6); + adev->bios_scratch_reg_offset = mmBIOS_SCRATCH_0; + + bios_2_scratch = RREG32(adev->bios_scratch_reg_offset + 2); + bios_6_scratch = RREG32(adev->bios_scratch_reg_offset + 6); /* let the bios control the backlight */ bios_2_scratch &= ~ATOM_S2_VRI_BRIGHT_ENABLE; @@ -1715,8 +1717,8 @@ void amdgpu_atombios_scratch_regs_init(struct amdgpu_device *adev) /* clear the vbios dpms state */ bios_2_scratch &= ~ATOM_S2_DEVICE_DPMS_STATE; - WREG32(mmBIOS_SCRATCH_2, bios_2_scratch); - WREG32(mmBIOS_SCRATCH_6, bios_6_scratch); + WREG32(adev->bios_scratch_reg_offset + 2, bios_2_scratch); + WREG32(adev->bios_scratch_reg_offset + 6, bios_6_scratch); } void amdgpu_atombios_scratch_regs_save(struct amdgpu_device *adev) @@ -1724,7 +1726,7 @@ void amdgpu_atombios_scratch_regs_save(struct amdgpu_device *adev) int i; for (i = 0; i < AMDGPU_BIOS_NUM_SCRATCH; i++) - adev->bios_scratch[i] = RREG32(mmBIOS_SCRATCH_0 + i); + adev->bios_scratch[i] = RREG32(adev->bios_scratch_reg_offset + i); } void amdgpu_atombios_scratch_regs_restore(struct amdgpu_device *adev) @@ -1738,50 +1740,58 @@ void amdgpu_atombios_scratch_regs_restore(struct amdgpu_device *adev) adev->bios_scratch[7] &= ~ATOM_S7_ASIC_INIT_COMPLETE_MASK; for (i = 0; i < AMDGPU_BIOS_NUM_SCRATCH; i++) - WREG32(mmBIOS_SCRATCH_0 + i, adev->bios_scratch[i]); + WREG32(adev->bios_scratch_reg_offset + i, adev->bios_scratch[i]); } void amdgpu_atombios_scratch_regs_engine_hung(struct amdgpu_device *adev, bool hung) { - u32 tmp = RREG32(mmBIOS_SCRATCH_3); + u32 tmp = RREG32(adev->bios_scratch_reg_offset + 3); if (hung) tmp |= ATOM_S3_ASIC_GUI_ENGINE_HUNG; else tmp &= ~ATOM_S3_ASIC_GUI_ENGINE_HUNG; - WREG32(mmBIOS_SCRATCH_3, tmp); + WREG32(adev->bios_scratch_reg_offset + 3, tmp); +} + +bool amdgpu_atombios_scratch_need_asic_init(struct amdgpu_device *adev) +{ + u32 tmp = RREG32(adev->bios_scratch_reg_offset + 7); + + if (tmp & ATOM_S7_ASIC_INIT_COMPLETE_MASK) + return false; + else + return true; } -/* Atom needs data in little endian format - * so swap as appropriate when copying data to - * or from atom. Note that atom operates on - * dw units. +/* Atom needs data in little endian format so swap as appropriate when copying + * data to or from atom. Note that atom operates on dw units. + * + * Use to_le=true when sending data to atom and provide at least + * ALIGN(num_bytes,4) bytes in the dst buffer. + * + * Use to_le=false when receiving data from atom and provide ALIGN(num_bytes,4) + * byes in the src buffer. */ void amdgpu_atombios_copy_swap(u8 *dst, u8 *src, u8 num_bytes, bool to_le) { #ifdef __BIG_ENDIAN - u8 src_tmp[20], dst_tmp[20]; /* used for byteswapping */ - u32 *dst32, *src32; + u32 src_tmp[5], dst_tmp[5]; int i; + u8 align_num_bytes = ALIGN(num_bytes, 4); - memcpy(src_tmp, src, num_bytes); - src32 = (u32 *)src_tmp; - dst32 = (u32 *)dst_tmp; if (to_le) { - for (i = 0; i < ((num_bytes + 3) / 4); i++) - dst32[i] = cpu_to_le32(src32[i]); - memcpy(dst, dst_tmp, num_bytes); + memcpy(src_tmp, src, num_bytes); + for (i = 0; i < align_num_bytes / 4; i++) + dst_tmp[i] = cpu_to_le32(src_tmp[i]); + memcpy(dst, dst_tmp, align_num_bytes); } else { - u8 dws = num_bytes & ~3; - for (i = 0; i < ((num_bytes + 3) / 4); i++) - dst32[i] = le32_to_cpu(src32[i]); - memcpy(dst, dst_tmp, dws); - if (num_bytes % 4) { - for (i = 0; i < (num_bytes % 4); i++) - dst[dws+i] = dst_tmp[dws+i]; - } + memcpy(src_tmp, src, align_num_bytes); + for (i = 0; i < align_num_bytes / 4; i++) + dst_tmp[i] = le32_to_cpu(src_tmp[i]); + memcpy(dst, dst_tmp, num_bytes); } #else memcpy(dst, src, num_bytes); @@ -1795,6 +1805,8 @@ int amdgpu_atombios_allocate_fb_scratch(struct amdgpu_device *adev) uint16_t data_offset; int usage_bytes = 0; struct _ATOM_VRAM_USAGE_BY_FIRMWARE *firmware_usage; + u64 start_addr; + u64 size; if (amdgpu_atom_parse_data_header(ctx, index, NULL, NULL, NULL, &data_offset)) { firmware_usage = (struct _ATOM_VRAM_USAGE_BY_FIRMWARE *)(ctx->bios + data_offset); @@ -1803,7 +1815,21 @@ int amdgpu_atombios_allocate_fb_scratch(struct amdgpu_device *adev) le32_to_cpu(firmware_usage->asFirmwareVramReserveInfo[0].ulStartAddrUsedByFirmware), le16_to_cpu(firmware_usage->asFirmwareVramReserveInfo[0].usFirmwareUseInKb)); - usage_bytes = le16_to_cpu(firmware_usage->asFirmwareVramReserveInfo[0].usFirmwareUseInKb) * 1024; + start_addr = firmware_usage->asFirmwareVramReserveInfo[0].ulStartAddrUsedByFirmware; + size = firmware_usage->asFirmwareVramReserveInfo[0].usFirmwareUseInKb; + + if ((uint32_t)(start_addr & ATOM_VRAM_OPERATION_FLAGS_MASK) == + (uint32_t)(ATOM_VRAM_BLOCK_SRIOV_MSG_SHARE_RESERVATION << + ATOM_VRAM_OPERATION_FLAGS_SHIFT)) { + /* Firmware request VRAM reservation for SR-IOV */ + adev->fw_vram_usage.start_offset = (start_addr & + (~ATOM_VRAM_OPERATION_FLAGS_MASK)) << 10; + adev->fw_vram_usage.size = size << 10; + /* Use the default scratch size */ + usage_bytes = 0; + } else { + usage_bytes = le16_to_cpu(firmware_usage->asFirmwareVramReserveInfo[0].usFirmwareUseInKb) * 1024; + } } ctx->scratch_size_bytes = 0; if (usage_bytes == 0) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h index 38d0fe32e5cd..b0d5d1d7fdba 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h @@ -200,6 +200,7 @@ void amdgpu_atombios_scratch_regs_save(struct amdgpu_device *adev); void amdgpu_atombios_scratch_regs_restore(struct amdgpu_device *adev); void amdgpu_atombios_scratch_regs_engine_hung(struct amdgpu_device *adev, bool hung); +bool amdgpu_atombios_scratch_need_asic_init(struct amdgpu_device *adev); void amdgpu_atombios_copy_swap(u8 *dst, u8 *src, u8 num_bytes, bool to_le); int amdgpu_atombios_get_max_vddc(struct amdgpu_device *adev, u8 voltage_type, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c index 4bdda56fccee..ff8efd0f8fd5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c @@ -66,59 +66,38 @@ void amdgpu_atomfirmware_scratch_regs_init(struct amdgpu_device *adev) } } -void amdgpu_atomfirmware_scratch_regs_save(struct amdgpu_device *adev) -{ - int i; - - for (i = 0; i < AMDGPU_BIOS_NUM_SCRATCH; i++) - adev->bios_scratch[i] = RREG32(adev->bios_scratch_reg_offset + i); -} - -void amdgpu_atomfirmware_scratch_regs_restore(struct amdgpu_device *adev) -{ - int i; - - /* - * VBIOS will check ASIC_INIT_COMPLETE bit to decide if - * execute ASIC_Init posting via driver - */ - adev->bios_scratch[7] &= ~ATOM_S7_ASIC_INIT_COMPLETE_MASK; - - for (i = 0; i < AMDGPU_BIOS_NUM_SCRATCH; i++) - WREG32(adev->bios_scratch_reg_offset + i, adev->bios_scratch[i]); -} - -void amdgpu_atomfirmware_scratch_regs_engine_hung(struct amdgpu_device *adev, - bool hung) -{ - u32 tmp = RREG32(adev->bios_scratch_reg_offset + 3); - - if (hung) - tmp |= ATOM_S3_ASIC_GUI_ENGINE_HUNG; - else - tmp &= ~ATOM_S3_ASIC_GUI_ENGINE_HUNG; - - WREG32(adev->bios_scratch_reg_offset + 3, tmp); -} - int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev) { struct atom_context *ctx = adev->mode_info.atom_context; int index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, vram_usagebyfirmware); + struct vram_usagebyfirmware_v2_1 * firmware_usage; + uint32_t start_addr, size; uint16_t data_offset; int usage_bytes = 0; if (amdgpu_atom_parse_data_header(ctx, index, NULL, NULL, NULL, &data_offset)) { - struct vram_usagebyfirmware_v2_1 *firmware_usage = - (struct vram_usagebyfirmware_v2_1 *)(ctx->bios + data_offset); - + firmware_usage = (struct vram_usagebyfirmware_v2_1 *)(ctx->bios + data_offset); DRM_DEBUG("atom firmware requested %08x %dkb fw %dkb drv\n", le32_to_cpu(firmware_usage->start_address_in_kb), le16_to_cpu(firmware_usage->used_by_firmware_in_kb), le16_to_cpu(firmware_usage->used_by_driver_in_kb)); - usage_bytes = le16_to_cpu(firmware_usage->used_by_driver_in_kb) * 1024; + start_addr = le32_to_cpu(firmware_usage->start_address_in_kb); + size = le16_to_cpu(firmware_usage->used_by_firmware_in_kb); + + if ((uint32_t)(start_addr & ATOM_VRAM_OPERATION_FLAGS_MASK) == + (uint32_t)(ATOM_VRAM_BLOCK_SRIOV_MSG_SHARE_RESERVATION << + ATOM_VRAM_OPERATION_FLAGS_SHIFT)) { + /* Firmware request VRAM reservation for SR-IOV */ + adev->fw_vram_usage.start_offset = (start_addr & + (~ATOM_VRAM_OPERATION_FLAGS_MASK)) << 10; + adev->fw_vram_usage.size = size << 10; + /* Use the default scratch size */ + usage_bytes = 0; + } else { + usage_bytes = le16_to_cpu(firmware_usage->used_by_driver_in_kb) << 10; + } } ctx->scratch_size_bytes = 0; if (usage_bytes == 0) @@ -130,3 +109,129 @@ int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev) ctx->scratch_size_bytes = usage_bytes; return 0; } + +union igp_info { + struct atom_integrated_system_info_v1_11 v11; +}; + +/* + * Return vram width from integrated system info table, if available, + * or 0 if not. + */ +int amdgpu_atomfirmware_get_vram_width(struct amdgpu_device *adev) +{ + struct amdgpu_mode_info *mode_info = &adev->mode_info; + int index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, + integratedsysteminfo); + u16 data_offset, size; + union igp_info *igp_info; + u8 frev, crev; + + /* get any igp specific overrides */ + if (amdgpu_atom_parse_data_header(mode_info->atom_context, index, &size, + &frev, &crev, &data_offset)) { + igp_info = (union igp_info *) + (mode_info->atom_context->bios + data_offset); + switch (crev) { + case 11: + return igp_info->v11.umachannelnumber * 64; + default: + return 0; + } + } + + return 0; +} + +union firmware_info { + struct atom_firmware_info_v3_1 v31; +}; + +union smu_info { + struct atom_smu_info_v3_1 v31; +}; + +union umc_info { + struct atom_umc_info_v3_1 v31; +}; + +int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev) +{ + struct amdgpu_mode_info *mode_info = &adev->mode_info; + struct amdgpu_pll *spll = &adev->clock.spll; + struct amdgpu_pll *mpll = &adev->clock.mpll; + uint8_t frev, crev; + uint16_t data_offset; + int ret = -EINVAL, index; + + index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, + firmwareinfo); + if (amdgpu_atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) { + union firmware_info *firmware_info = + (union firmware_info *)(mode_info->atom_context->bios + + data_offset); + + adev->clock.default_sclk = + le32_to_cpu(firmware_info->v31.bootup_sclk_in10khz); + adev->clock.default_mclk = + le32_to_cpu(firmware_info->v31.bootup_mclk_in10khz); + + adev->pm.current_sclk = adev->clock.default_sclk; + adev->pm.current_mclk = adev->clock.default_mclk; + + /* not technically a clock, but... */ + adev->mode_info.firmware_flags = + le32_to_cpu(firmware_info->v31.firmware_capability); + + ret = 0; + } + + index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, + smu_info); + if (amdgpu_atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) { + union smu_info *smu_info = + (union smu_info *)(mode_info->atom_context->bios + + data_offset); + + /* system clock */ + spll->reference_freq = le32_to_cpu(smu_info->v31.core_refclk_10khz); + + spll->reference_div = 0; + spll->min_post_div = 1; + spll->max_post_div = 1; + spll->min_ref_div = 2; + spll->max_ref_div = 0xff; + spll->min_feedback_div = 4; + spll->max_feedback_div = 0xff; + spll->best_vco = 0; + + ret = 0; + } + + index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, + umc_info); + if (amdgpu_atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) { + union umc_info *umc_info = + (union umc_info *)(mode_info->atom_context->bios + + data_offset); + + /* memory clock */ + mpll->reference_freq = le32_to_cpu(umc_info->v31.mem_refclk_10khz); + + mpll->reference_div = 0; + mpll->min_post_div = 1; + mpll->max_post_div = 1; + mpll->min_ref_div = 2; + mpll->max_ref_div = 0xff; + mpll->min_feedback_div = 4; + mpll->max_feedback_div = 0xff; + mpll->best_vco = 0; + + ret = 0; + } + + return ret; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h index a2c3ebe22c71..288b97e54347 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h @@ -26,10 +26,8 @@ bool amdgpu_atomfirmware_gpu_supports_virtualization(struct amdgpu_device *adev); void amdgpu_atomfirmware_scratch_regs_init(struct amdgpu_device *adev); -void amdgpu_atomfirmware_scratch_regs_save(struct amdgpu_device *adev); -void amdgpu_atomfirmware_scratch_regs_restore(struct amdgpu_device *adev); -void amdgpu_atomfirmware_scratch_regs_engine_hung(struct amdgpu_device *adev, - bool hung); int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev); +int amdgpu_atomfirmware_get_vram_width(struct amdgpu_device *adev); +int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c index 1beae5b930d0..63ec1e1bb6aa 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c @@ -40,7 +40,7 @@ static int amdgpu_benchmark_do_move(struct amdgpu_device *adev, unsigned size, for (i = 0; i < n; i++) { struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; r = amdgpu_copy_buffer(ring, saddr, daddr, size, NULL, &fence, - false); + false, false); if (r) goto exit_do_move; r = dma_fence_wait(fence, false); @@ -81,7 +81,7 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, n = AMDGPU_BENCHMARK_ITERATIONS; r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, sdomain, 0, NULL, - NULL, &sobj); + NULL, 0, &sobj); if (r) { goto out_cleanup; } @@ -94,7 +94,7 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, goto out_cleanup; } r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, ddomain, 0, NULL, - NULL, &dobj); + NULL, 0, &dobj); if (r) { goto out_cleanup; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c index 365e735f6647..057e1ecd83ce 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c @@ -59,12 +59,6 @@ static bool check_atom_bios(uint8_t *bios, size_t size) return false; } - tmp = bios[0x18] | (bios[0x19] << 8); - if (bios[tmp + 0x14] != 0x0) { - DRM_INFO("Not an x86 BIOS ROM\n"); - return false; - } - bios_header_start = bios[0x48] | (bios[0x49] << 8); if (!bios_header_start) { DRM_INFO("Can't locate bios header\n"); @@ -86,19 +80,6 @@ static bool check_atom_bios(uint8_t *bios, size_t size) return false; } -static bool is_atom_fw(uint8_t *bios) -{ - uint16_t bios_header_start = bios[0x48] | (bios[0x49] << 8); - uint8_t frev = bios[bios_header_start + 2]; - uint8_t crev = bios[bios_header_start + 3]; - - if ((frev < 3) || - ((frev == 3) && (crev < 3))) - return false; - - return true; -} - /* If you boot an IGP board with a discrete card as the primary, * the IGP rom is not accessible via the rom bar as the IGP rom is * part of the system bios. On boot, the system bios puts a @@ -117,7 +98,7 @@ static bool igp_read_bios_from_vram(struct amdgpu_device *adev) adev->bios = NULL; vram_base = pci_resource_start(adev->pdev, 0); - bios = ioremap(vram_base, size); + bios = ioremap_wc(vram_base, size); if (!bios) { return false; } @@ -455,6 +436,6 @@ bool amdgpu_get_bios(struct amdgpu_device *adev) return false; success: - adev->is_atom_fw = is_atom_fw(adev->bios); + adev->is_atom_fw = (adev->asic_type >= CHIP_VEGA10) ? true : false; return true; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c index f621ee115c98..59089e027f4d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c @@ -83,7 +83,7 @@ static int amdgpu_bo_list_create(struct amdgpu_device *adev, r = idr_alloc(&fpriv->bo_list_handles, list, 1, 0, GFP_KERNEL); mutex_unlock(&fpriv->bo_list_lock); if (r < 0) { - kfree(list); + amdgpu_bo_list_free(list); return r; } *id = r; @@ -136,7 +136,7 @@ static int amdgpu_bo_list_set(struct amdgpu_device *adev, } bo = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj)); - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); usermm = amdgpu_ttm_tt_get_usermm(bo->tbo.ttm); if (usermm) { @@ -156,11 +156,11 @@ static int amdgpu_bo_list_set(struct amdgpu_device *adev, entry->tv.bo = &entry->robj->tbo; entry->tv.shared = !entry->robj->prime_shared_count; - if (entry->robj->prefered_domains == AMDGPU_GEM_DOMAIN_GDS) + if (entry->robj->preferred_domains == AMDGPU_GEM_DOMAIN_GDS) gds_obj = entry->robj; - if (entry->robj->prefered_domains == AMDGPU_GEM_DOMAIN_GWS) + if (entry->robj->preferred_domains == AMDGPU_GEM_DOMAIN_GWS) gws_obj = entry->robj; - if (entry->robj->prefered_domains == AMDGPU_GEM_DOMAIN_OA) + if (entry->robj->preferred_domains == AMDGPU_GEM_DOMAIN_OA) oa_obj = entry->robj; total_size += amdgpu_bo_size(entry->robj); @@ -198,12 +198,16 @@ amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id) result = idr_find(&fpriv->bo_list_handles, id); if (result) { - if (kref_get_unless_zero(&result->refcount)) + if (kref_get_unless_zero(&result->refcount)) { + rcu_read_unlock(); mutex_lock(&result->lock); - else + } else { + rcu_read_unlock(); result = NULL; + } + } else { + rcu_read_unlock(); } - rcu_read_unlock(); return result; } @@ -266,7 +270,7 @@ int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data, struct amdgpu_fpriv *fpriv = filp->driver_priv; union drm_amdgpu_bo_list *args = data; uint32_t handle = args->in.list_handle; - const void __user *uptr = (const void*)(uintptr_t)args->in.bo_info_ptr; + const void __user *uptr = u64_to_user_ptr(args->in.bo_info_ptr); struct drm_amdgpu_bo_list_entry *info; struct amdgpu_bo_list *list; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c index c0a806280257..f2b72c7c6857 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c @@ -42,10 +42,31 @@ struct amdgpu_cgs_device { struct amdgpu_device *adev = \ ((struct amdgpu_cgs_device *)cgs_device)->adev +static void *amdgpu_cgs_register_pp_handle(struct cgs_device *cgs_device, + int (*call_back_func)(struct amd_pp_init *, void **)) +{ + CGS_FUNC_ADEV; + struct amd_pp_init pp_init; + struct amd_powerplay *amd_pp; + + if (call_back_func == NULL) + return NULL; + + amd_pp = &(adev->powerplay); + pp_init.chip_family = adev->family; + pp_init.chip_id = adev->asic_type; + pp_init.pm_en = (amdgpu_dpm != 0 && !amdgpu_sriov_vf(adev)) ? true : false; + pp_init.feature_mask = amdgpu_pp_feature_mask; + pp_init.device = cgs_device; + if (call_back_func(&pp_init, &(amd_pp->pp_handle))) + return NULL; + + return adev->powerplay.pp_handle; +} + static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device, enum cgs_gpu_mem_type type, uint64_t size, uint64_t align, - uint64_t min_offset, uint64_t max_offset, cgs_handle_t *handle) { CGS_FUNC_ADEV; @@ -53,13 +74,6 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device, int ret = 0; uint32_t domain = 0; struct amdgpu_bo *obj; - struct ttm_placement placement; - struct ttm_place place; - - if (min_offset > max_offset) { - BUG_ON(1); - return -EINVAL; - } /* fail if the alignment is not a power of 2 */ if (((align != 1) && (align & (align - 1))) @@ -73,41 +87,19 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device, flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; domain = AMDGPU_GEM_DOMAIN_VRAM; - if (max_offset > adev->mc.real_vram_size) - return -EINVAL; - place.fpfn = min_offset >> PAGE_SHIFT; - place.lpfn = max_offset >> PAGE_SHIFT; - place.flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | - TTM_PL_FLAG_VRAM; break; case CGS_GPU_MEM_TYPE__INVISIBLE_CONTIG_FB: case CGS_GPU_MEM_TYPE__INVISIBLE_FB: flags = AMDGPU_GEM_CREATE_NO_CPU_ACCESS | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; domain = AMDGPU_GEM_DOMAIN_VRAM; - if (adev->mc.visible_vram_size < adev->mc.real_vram_size) { - place.fpfn = - max(min_offset, adev->mc.visible_vram_size) >> PAGE_SHIFT; - place.lpfn = - min(max_offset, adev->mc.real_vram_size) >> PAGE_SHIFT; - place.flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | - TTM_PL_FLAG_VRAM; - } - break; case CGS_GPU_MEM_TYPE__GART_CACHEABLE: domain = AMDGPU_GEM_DOMAIN_GTT; - place.fpfn = min_offset >> PAGE_SHIFT; - place.lpfn = max_offset >> PAGE_SHIFT; - place.flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_TT; break; case CGS_GPU_MEM_TYPE__GART_WRITECOMBINE: flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC; domain = AMDGPU_GEM_DOMAIN_GTT; - place.fpfn = min_offset >> PAGE_SHIFT; - place.lpfn = max_offset >> PAGE_SHIFT; - place.flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_TT | - TTM_PL_FLAG_UNCACHED; break; default: return -EINVAL; @@ -116,15 +108,8 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device, *handle = 0; - placement.placement = &place; - placement.num_placement = 1; - placement.busy_placement = &place; - placement.num_busy_placement = 1; - - ret = amdgpu_bo_create_restricted(adev, size, PAGE_SIZE, - true, domain, flags, - NULL, &placement, NULL, - &obj); + ret = amdgpu_bo_create(adev, size, align, true, domain, flags, + NULL, NULL, 0, &obj); if (ret) { DRM_ERROR("(%d) bo create failed\n", ret); return ret; @@ -155,19 +140,14 @@ static int amdgpu_cgs_gmap_gpu_mem(struct cgs_device *cgs_device, cgs_handle_t h uint64_t *mcaddr) { int r; - u64 min_offset, max_offset; struct amdgpu_bo *obj = (struct amdgpu_bo *)handle; WARN_ON_ONCE(obj->placement.num_placement > 1); - min_offset = obj->placements[0].fpfn << PAGE_SHIFT; - max_offset = obj->placements[0].lpfn << PAGE_SHIFT; - r = amdgpu_bo_reserve(obj, true); if (unlikely(r != 0)) return r; - r = amdgpu_bo_pin_restricted(obj, obj->prefered_domains, - min_offset, max_offset, mcaddr); + r = amdgpu_bo_pin(obj, obj->preferred_domains, mcaddr); amdgpu_bo_unreserve(obj); return r; } @@ -240,6 +220,8 @@ static uint32_t amdgpu_cgs_read_ind_register(struct cgs_device *cgs_device, return RREG32_DIDT(index); case CGS_IND_REG_GC_CAC: return RREG32_GC_CAC(index); + case CGS_IND_REG_SE_CAC: + return RREG32_SE_CAC(index); case CGS_IND_REG__AUDIO_ENDPT: DRM_ERROR("audio endpt register access not implemented.\n"); return 0; @@ -266,6 +248,8 @@ static void amdgpu_cgs_write_ind_register(struct cgs_device *cgs_device, return WREG32_DIDT(index, value); case CGS_IND_REG_GC_CAC: return WREG32_GC_CAC(index, value); + case CGS_IND_REG_SE_CAC: + return WREG32_SE_CAC(index, value); case CGS_IND_REG__AUDIO_ENDPT: DRM_ERROR("audio endpt register access not implemented.\n"); return; @@ -610,6 +594,17 @@ static int amdgpu_cgs_enter_safe_mode(struct cgs_device *cgs_device, return 0; } +static void amdgpu_cgs_lock_grbm_idx(struct cgs_device *cgs_device, + bool lock) +{ + CGS_FUNC_ADEV; + + if (lock) + mutex_lock(&adev->grbm_idx_mutex); + else + mutex_unlock(&adev->grbm_idx_mutex); +} + static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, enum cgs_ucode_id type, struct cgs_firmware_info *info) @@ -644,7 +639,7 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, info->version = (uint16_t)le32_to_cpu(header->header.ucode_version); if (CGS_UCODE_ID_CP_MEC == type) - info->image_size = (header->jt_offset) << 2; + info->image_size = le32_to_cpu(header->jt_offset) << 2; info->fw_version = amdgpu_get_firmware_version(cgs_device, type); info->feature_version = (uint16_t)le32_to_cpu(header->ucode_feature_version); @@ -660,6 +655,85 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, if (!adev->pm.fw) { switch (adev->asic_type) { + case CHIP_TAHITI: + strcpy(fw_name, "radeon/tahiti_smc.bin"); + break; + case CHIP_PITCAIRN: + if ((adev->pdev->revision == 0x81) && + ((adev->pdev->device == 0x6810) || + (adev->pdev->device == 0x6811))) { + info->is_kicker = true; + strcpy(fw_name, "radeon/pitcairn_k_smc.bin"); + } else { + strcpy(fw_name, "radeon/pitcairn_smc.bin"); + } + break; + case CHIP_VERDE: + if (((adev->pdev->device == 0x6820) && + ((adev->pdev->revision == 0x81) || + (adev->pdev->revision == 0x83))) || + ((adev->pdev->device == 0x6821) && + ((adev->pdev->revision == 0x83) || + (adev->pdev->revision == 0x87))) || + ((adev->pdev->revision == 0x87) && + ((adev->pdev->device == 0x6823) || + (adev->pdev->device == 0x682b)))) { + info->is_kicker = true; + strcpy(fw_name, "radeon/verde_k_smc.bin"); + } else { + strcpy(fw_name, "radeon/verde_smc.bin"); + } + break; + case CHIP_OLAND: + if (((adev->pdev->revision == 0x81) && + ((adev->pdev->device == 0x6600) || + (adev->pdev->device == 0x6604) || + (adev->pdev->device == 0x6605) || + (adev->pdev->device == 0x6610))) || + ((adev->pdev->revision == 0x83) && + (adev->pdev->device == 0x6610))) { + info->is_kicker = true; + strcpy(fw_name, "radeon/oland_k_smc.bin"); + } else { + strcpy(fw_name, "radeon/oland_smc.bin"); + } + break; + case CHIP_HAINAN: + if (((adev->pdev->revision == 0x81) && + (adev->pdev->device == 0x6660)) || + ((adev->pdev->revision == 0x83) && + ((adev->pdev->device == 0x6660) || + (adev->pdev->device == 0x6663) || + (adev->pdev->device == 0x6665) || + (adev->pdev->device == 0x6667)))) { + info->is_kicker = true; + strcpy(fw_name, "radeon/hainan_k_smc.bin"); + } else if ((adev->pdev->revision == 0xc3) && + (adev->pdev->device == 0x6665)) { + info->is_kicker = true; + strcpy(fw_name, "radeon/banks_k_2_smc.bin"); + } else { + strcpy(fw_name, "radeon/hainan_smc.bin"); + } + break; + case CHIP_BONAIRE: + if ((adev->pdev->revision == 0x80) || + (adev->pdev->revision == 0x81) || + (adev->pdev->device == 0x665f)) { + info->is_kicker = true; + strcpy(fw_name, "radeon/bonaire_k_smc.bin"); + } else { + strcpy(fw_name, "radeon/bonaire_smc.bin"); + } + break; + case CHIP_HAWAII: + if (adev->pdev->revision == 0x80) { + info->is_kicker = true; + strcpy(fw_name, "radeon/hawaii_k_smc.bin"); + } else { + strcpy(fw_name, "radeon/hawaii_smc.bin"); + } + break; case CHIP_TOPAZ: if (((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0x81)) || ((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0x83)) || @@ -719,7 +793,13 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, strcpy(fw_name, "amdgpu/polaris12_smc.bin"); break; case CHIP_VEGA10: - strcpy(fw_name, "amdgpu/vega10_smc.bin"); + if ((adev->pdev->device == 0x687f) && + ((adev->pdev->revision == 0xc0) || + (adev->pdev->revision == 0xc1) || + (adev->pdev->revision == 0xc3))) + strcpy(fw_name, "amdgpu/vega10_acg_smc.bin"); + else + strcpy(fw_name, "amdgpu/vega10_smc.bin"); break; default: DRM_ERROR("SMC firmware not supported\n"); @@ -817,6 +897,9 @@ static int amdgpu_cgs_query_system_info(struct cgs_device *cgs_device, case CGS_SYSTEM_INFO_PCIE_SUB_SYS_VENDOR_ID: sys_info->value = adev->pdev->subsystem_vendor; break; + case CGS_SYSTEM_INFO_PCIE_BUS_DEVFN: + sys_info->value = adev->pdev->devfn; + break; default: return -ENODEV; } @@ -828,10 +911,6 @@ static int amdgpu_cgs_get_active_displays_info(struct cgs_device *cgs_device, struct cgs_display_info *info) { CGS_FUNC_ADEV; - struct amdgpu_crtc *amdgpu_crtc; - struct drm_device *ddev = adev->ddev; - struct drm_crtc *crtc; - uint32_t line_time_us, vblank_lines; struct cgs_mode_info *mode_info; if (info == NULL) @@ -845,30 +924,43 @@ static int amdgpu_cgs_get_active_displays_info(struct cgs_device *cgs_device, mode_info->ref_clock = adev->clock.spll.reference_freq; } - if (adev->mode_info.num_crtc && adev->mode_info.mode_config_initialized) { - list_for_each_entry(crtc, - &ddev->mode_config.crtc_list, head) { - amdgpu_crtc = to_amdgpu_crtc(crtc); - if (crtc->enabled) { - info->active_display_mask |= (1 << amdgpu_crtc->crtc_id); - info->display_count++; - } - if (mode_info != NULL && - crtc->enabled && amdgpu_crtc->enabled && - amdgpu_crtc->hw_mode.clock) { - line_time_us = (amdgpu_crtc->hw_mode.crtc_htotal * 1000) / - amdgpu_crtc->hw_mode.clock; - vblank_lines = amdgpu_crtc->hw_mode.crtc_vblank_end - - amdgpu_crtc->hw_mode.crtc_vdisplay + - (amdgpu_crtc->v_border * 2); - mode_info->vblank_time_us = vblank_lines * line_time_us; - mode_info->refresh_rate = drm_mode_vrefresh(&amdgpu_crtc->hw_mode); - mode_info->ref_clock = adev->clock.spll.reference_freq; - mode_info = NULL; + if (!amdgpu_device_has_dc_support(adev)) { + struct amdgpu_crtc *amdgpu_crtc; + struct drm_device *ddev = adev->ddev; + struct drm_crtc *crtc; + uint32_t line_time_us, vblank_lines; + + if (adev->mode_info.num_crtc && adev->mode_info.mode_config_initialized) { + list_for_each_entry(crtc, + &ddev->mode_config.crtc_list, head) { + amdgpu_crtc = to_amdgpu_crtc(crtc); + if (crtc->enabled) { + info->active_display_mask |= (1 << amdgpu_crtc->crtc_id); + info->display_count++; + } + if (mode_info != NULL && + crtc->enabled && amdgpu_crtc->enabled && + amdgpu_crtc->hw_mode.clock) { + line_time_us = (amdgpu_crtc->hw_mode.crtc_htotal * 1000) / + amdgpu_crtc->hw_mode.clock; + vblank_lines = amdgpu_crtc->hw_mode.crtc_vblank_end - + amdgpu_crtc->hw_mode.crtc_vdisplay + + (amdgpu_crtc->v_border * 2); + mode_info->vblank_time_us = vblank_lines * line_time_us; + mode_info->refresh_rate = drm_mode_vrefresh(&amdgpu_crtc->hw_mode); + mode_info->ref_clock = adev->clock.spll.reference_freq; + mode_info = NULL; + } } } + } else { + info->display_count = adev->pm.pm_display_cfg.num_display; + if (mode_info != NULL) { + mode_info->vblank_time_us = adev->pm.pm_display_cfg.min_vblank_time; + mode_info->refresh_rate = adev->pm.pm_display_cfg.vrefresh; + mode_info->ref_clock = adev->clock.spll.reference_freq; + } } - return 0; } @@ -1117,6 +1209,8 @@ static const struct cgs_ops amdgpu_cgs_ops = { .query_system_info = amdgpu_cgs_query_system_info, .is_virtualization_enabled = amdgpu_cgs_is_virtualization_enabled, .enter_safe_mode = amdgpu_cgs_enter_safe_mode, + .lock_grbm_idx = amdgpu_cgs_lock_grbm_idx, + .register_pp_handle = amdgpu_cgs_register_pp_handle, }; static const struct cgs_os_ops amdgpu_cgs_os_ops = { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index 8d1cf2d3e663..df9cbc78e168 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -231,7 +231,7 @@ amdgpu_connector_update_scratch_regs(struct drm_connector *connector, if (connector->encoder_ids[i] == 0) break; - encoder = drm_encoder_find(connector->dev, + encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]); if (!encoder) continue; @@ -256,7 +256,7 @@ amdgpu_connector_find_encoder(struct drm_connector *connector, for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) { if (connector->encoder_ids[i] == 0) break; - encoder = drm_encoder_find(connector->dev, + encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]); if (!encoder) continue; @@ -346,10 +346,8 @@ static void amdgpu_connector_free_edid(struct drm_connector *connector) { struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector); - if (amdgpu_connector->edid) { - kfree(amdgpu_connector->edid); - amdgpu_connector->edid = NULL; - } + kfree(amdgpu_connector->edid); + amdgpu_connector->edid = NULL; } static int amdgpu_connector_ddc_get_modes(struct drm_connector *connector) @@ -374,7 +372,7 @@ amdgpu_connector_best_single_encoder(struct drm_connector *connector) /* pick the encoder ids */ if (enc_id) - return drm_encoder_find(connector->dev, enc_id); + return drm_encoder_find(connector->dev, NULL, enc_id); return NULL; } @@ -1079,7 +1077,7 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force) if (connector->encoder_ids[i] == 0) break; - encoder = drm_encoder_find(connector->dev, connector->encoder_ids[i]); + encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]); if (!encoder) continue; @@ -1136,7 +1134,7 @@ amdgpu_connector_dvi_encoder(struct drm_connector *connector) if (connector->encoder_ids[i] == 0) break; - encoder = drm_encoder_find(connector->dev, connector->encoder_ids[i]); + encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]); if (!encoder) continue; @@ -1155,7 +1153,7 @@ amdgpu_connector_dvi_encoder(struct drm_connector *connector) /* then check use digitial */ /* pick the first one */ if (enc_id) - return drm_encoder_find(connector->dev, enc_id); + return drm_encoder_find(connector->dev, NULL, enc_id); return NULL; } @@ -1296,7 +1294,7 @@ u16 amdgpu_connector_encoder_get_dp_bridge_encoder_id(struct drm_connector *conn if (connector->encoder_ids[i] == 0) break; - encoder = drm_encoder_find(connector->dev, + encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]); if (!encoder) continue; @@ -1325,7 +1323,7 @@ static bool amdgpu_connector_encoder_is_hbr2(struct drm_connector *connector) for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) { if (connector->encoder_ids[i] == 0) break; - encoder = drm_encoder_find(connector->dev, + encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]); if (!encoder) continue; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 5599c01b265d..a57cec737c18 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -25,6 +25,7 @@ * Jerome Glisse <glisse@freedesktop.org> */ #include <linux/pagemap.h> +#include <linux/sync_file.h> #include <drm/drmP.h> #include <drm/amdgpu_drm.h> #include <drm/drm_syncobj.h> @@ -54,7 +55,7 @@ static int amdgpu_cs_user_fence_chunk(struct amdgpu_cs_parser *p, *offset = data->offset; - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); if (amdgpu_ttm_tt_get_usermm(p->uf_entry.robj->tbo.ttm)) { amdgpu_bo_unref(&p->uf_entry.robj); @@ -89,12 +90,14 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data) goto free_chunk; } + mutex_lock(&p->ctx->lock); + /* get chunks */ - chunk_array_user = (uint64_t __user *)(uintptr_t)(cs->in.chunks); + chunk_array_user = u64_to_user_ptr(cs->in.chunks); if (copy_from_user(chunk_array, chunk_array_user, sizeof(uint64_t)*cs->in.num_chunks)) { ret = -EFAULT; - goto put_ctx; + goto free_chunk; } p->nchunks = cs->in.num_chunks; @@ -102,7 +105,7 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data) GFP_KERNEL); if (!p->chunks) { ret = -ENOMEM; - goto put_ctx; + goto free_chunk; } for (i = 0; i < p->nchunks; i++) { @@ -110,7 +113,7 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data) struct drm_amdgpu_cs_chunk user_chunk; uint32_t __user *cdata; - chunk_ptr = (void __user *)(uintptr_t)chunk_array[i]; + chunk_ptr = u64_to_user_ptr(chunk_array[i]); if (copy_from_user(&user_chunk, chunk_ptr, sizeof(struct drm_amdgpu_cs_chunk))) { ret = -EFAULT; @@ -121,7 +124,7 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data) p->chunks[i].length_dw = user_chunk.length_dw; size = p->chunks[i].length_dw; - cdata = (void __user *)(uintptr_t)user_chunk.chunk_data; + cdata = u64_to_user_ptr(user_chunk.chunk_data); p->chunks[i].kdata = kvmalloc_array(size, sizeof(uint32_t), GFP_KERNEL); if (p->chunks[i].kdata == NULL) { @@ -169,6 +172,11 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data) if (ret) goto free_all_kdata; + if (p->ctx->vram_lost_counter != p->job->vram_lost_counter) { + ret = -ECANCELED; + goto free_all_kdata; + } + if (p->uf_entry.robj) p->job->uf_addr = uf_offset; kfree(chunk_array); @@ -182,8 +190,6 @@ free_partial_kdata: kfree(p->chunks); p->chunks = NULL; p->nchunks = 0; -put_ctx: - amdgpu_ctx_put(p->ctx); free_chunk: kfree(chunk_array); @@ -223,10 +229,11 @@ static s64 bytes_to_us(struct amdgpu_device *adev, u64 bytes) * ticks. The accumulated microseconds (us) are converted to bytes and * returned. */ -static u64 amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev) +static void amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev, + u64 *max_bytes, + u64 *max_vis_bytes) { s64 time_us, increment_us; - u64 max_bytes; u64 free_vram, total_vram, used_vram; /* Allow a maximum of 200 accumulated ms. This is basically per-IB @@ -238,11 +245,14 @@ static u64 amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev) */ const s64 us_upper_bound = 200000; - if (!adev->mm_stats.log2_max_MBps) - return 0; + if (!adev->mm_stats.log2_max_MBps) { + *max_bytes = 0; + *max_vis_bytes = 0; + return; + } total_vram = adev->mc.real_vram_size - adev->vram_pin_size; - used_vram = atomic64_read(&adev->vram_usage); + used_vram = amdgpu_vram_mgr_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); free_vram = used_vram >= total_vram ? 0 : total_vram - used_vram; spin_lock(&adev->mm_stats.lock); @@ -280,23 +290,46 @@ static u64 amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev) adev->mm_stats.accum_us = max(min_us, adev->mm_stats.accum_us); } - /* This returns 0 if the driver is in debt to disallow (optional) + /* This is set to 0 if the driver is in debt to disallow (optional) * buffer moves. */ - max_bytes = us_to_bytes(adev, adev->mm_stats.accum_us); + *max_bytes = us_to_bytes(adev, adev->mm_stats.accum_us); + + /* Do the same for visible VRAM if half of it is free */ + if (adev->mc.visible_vram_size < adev->mc.real_vram_size) { + u64 total_vis_vram = adev->mc.visible_vram_size; + u64 used_vis_vram = + amdgpu_vram_mgr_vis_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); + + if (used_vis_vram < total_vis_vram) { + u64 free_vis_vram = total_vis_vram - used_vis_vram; + adev->mm_stats.accum_us_vis = min(adev->mm_stats.accum_us_vis + + increment_us, us_upper_bound); + + if (free_vis_vram >= total_vis_vram / 2) + adev->mm_stats.accum_us_vis = + max(bytes_to_us(adev, free_vis_vram / 2), + adev->mm_stats.accum_us_vis); + } + + *max_vis_bytes = us_to_bytes(adev, adev->mm_stats.accum_us_vis); + } else { + *max_vis_bytes = 0; + } spin_unlock(&adev->mm_stats.lock); - return max_bytes; } /* Report how many bytes have really been moved for the last command * submission. This can result in a debt that can stop buffer migrations * temporarily. */ -void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes) +void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes, + u64 num_vis_bytes) { spin_lock(&adev->mm_stats.lock); adev->mm_stats.accum_us -= bytes_to_us(adev, num_bytes); + adev->mm_stats.accum_us_vis -= bytes_to_us(adev, num_vis_bytes); spin_unlock(&adev->mm_stats.lock); } @@ -304,7 +337,7 @@ static int amdgpu_cs_bo_validate(struct amdgpu_cs_parser *p, struct amdgpu_bo *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - u64 initial_bytes_moved; + u64 initial_bytes_moved, bytes_moved; uint32_t domain; int r; @@ -314,17 +347,35 @@ static int amdgpu_cs_bo_validate(struct amdgpu_cs_parser *p, /* Don't move this buffer if we have depleted our allowance * to move it. Don't move anything if the threshold is zero. */ - if (p->bytes_moved < p->bytes_moved_threshold) - domain = bo->prefered_domains; - else + if (p->bytes_moved < p->bytes_moved_threshold) { + if (adev->mc.visible_vram_size < adev->mc.real_vram_size && + (bo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)) { + /* And don't move a CPU_ACCESS_REQUIRED BO to limited + * visible VRAM if we've depleted our allowance to do + * that. + */ + if (p->bytes_moved_vis < p->bytes_moved_vis_threshold) + domain = bo->preferred_domains; + else + domain = bo->allowed_domains; + } else { + domain = bo->preferred_domains; + } + } else { domain = bo->allowed_domains; + } retry: amdgpu_ttm_placement_from_domain(bo, domain); initial_bytes_moved = atomic64_read(&adev->num_bytes_moved); r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); - p->bytes_moved += atomic64_read(&adev->num_bytes_moved) - - initial_bytes_moved; + bytes_moved = atomic64_read(&adev->num_bytes_moved) - + initial_bytes_moved; + p->bytes_moved += bytes_moved; + if (adev->mc.visible_vram_size < adev->mc.real_vram_size && + bo->tbo.mem.mem_type == TTM_PL_VRAM && + bo->tbo.mem.start < adev->mc.visible_vram_size >> PAGE_SHIFT) + p->bytes_moved_vis += bytes_moved; if (unlikely(r == -ENOMEM) && domain != bo->allowed_domains) { domain = bo->allowed_domains; @@ -350,7 +401,8 @@ static bool amdgpu_cs_try_evict(struct amdgpu_cs_parser *p, struct amdgpu_bo_list_entry *candidate = p->evictable; struct amdgpu_bo *bo = candidate->robj; struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - u64 initial_bytes_moved; + u64 initial_bytes_moved, bytes_moved; + bool update_bytes_moved_vis; uint32_t other; /* If we reached our current BO we can forget it */ @@ -370,10 +422,17 @@ static bool amdgpu_cs_try_evict(struct amdgpu_cs_parser *p, /* Good we can try to move this BO somewhere else */ amdgpu_ttm_placement_from_domain(bo, other); + update_bytes_moved_vis = + adev->mc.visible_vram_size < adev->mc.real_vram_size && + bo->tbo.mem.mem_type == TTM_PL_VRAM && + bo->tbo.mem.start < adev->mc.visible_vram_size >> PAGE_SHIFT; initial_bytes_moved = atomic64_read(&adev->num_bytes_moved); r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); - p->bytes_moved += atomic64_read(&adev->num_bytes_moved) - + bytes_moved = atomic64_read(&adev->num_bytes_moved) - initial_bytes_moved; + p->bytes_moved += bytes_moved; + if (update_bytes_moved_vis) + p->bytes_moved_vis += bytes_moved; if (unlikely(r)) break; @@ -420,11 +479,16 @@ static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p, return -EPERM; /* Check if we have user pages and nobody bound the BO already */ - if (lobj->user_pages && bo->tbo.ttm->state != tt_bound) { - size_t size = sizeof(struct page *); - - size *= bo->tbo.ttm->num_pages; - memcpy(bo->tbo.ttm->pages, lobj->user_pages, size); + if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm) && + lobj->user_pages) { + amdgpu_ttm_placement_from_domain(bo, + AMDGPU_GEM_DOMAIN_CPU); + r = ttm_bo_validate(&bo->tbo, &bo->placement, true, + false); + if (r) + return r; + amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, + lobj->user_pages); binding_userptr = true; } @@ -449,7 +513,6 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, struct amdgpu_fpriv *fpriv = p->filp->driver_priv; struct amdgpu_bo_list_entry *e; struct list_head duplicates; - bool need_mmap_lock = false; unsigned i, tries = 10; int r; @@ -457,9 +520,9 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, p->bo_list = amdgpu_bo_list_get(fpriv, cs->in.bo_list_handle); if (p->bo_list) { - need_mmap_lock = p->bo_list->first_userptr != - p->bo_list->num_entries; amdgpu_bo_list_get_list(p->bo_list, &p->validated); + if (p->bo_list->first_userptr != p->bo_list->num_entries) + p->mn = amdgpu_mn_get(p->adev); } INIT_LIST_HEAD(&duplicates); @@ -468,9 +531,6 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, if (p->uf_entry.robj) list_add(&p->uf_entry.tv.head, &p->validated); - if (need_mmap_lock) - down_read(¤t->mm->mmap_sem); - while (1) { struct list_head need_pages; unsigned i; @@ -490,23 +550,24 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, INIT_LIST_HEAD(&need_pages); for (i = p->bo_list->first_userptr; i < p->bo_list->num_entries; ++i) { + struct amdgpu_bo *bo; e = &p->bo_list->array[i]; + bo = e->robj; - if (amdgpu_ttm_tt_userptr_invalidated(e->robj->tbo.ttm, + if (amdgpu_ttm_tt_userptr_invalidated(bo->tbo.ttm, &e->user_invalidated) && e->user_pages) { /* We acquired a page array, but somebody * invalidated it. Free it and try again */ release_pages(e->user_pages, - e->robj->tbo.ttm->num_pages, - false); + bo->tbo.ttm->num_pages); kvfree(e->user_pages); e->user_pages = NULL; } - if (e->robj->tbo.ttm->state != tt_bound && + if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm) && !e->user_pages) { list_del(&e->tv.head); list_add(&e->tv.head, &need_pages); @@ -554,8 +615,10 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, list_splice(&need_pages, &p->validated); } - p->bytes_moved_threshold = amdgpu_cs_get_threshold_for_moves(p->adev); + amdgpu_cs_get_threshold_for_moves(p->adev, &p->bytes_moved_threshold, + &p->bytes_moved_vis_threshold); p->bytes_moved = 0; + p->bytes_moved_vis = 0; p->evictable = list_last_entry(&p->validated, struct amdgpu_bo_list_entry, tv.head); @@ -579,11 +642,8 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, goto error_validate; } - amdgpu_cs_report_moved_bytes(p->adev, p->bytes_moved); - - fpriv->vm.last_eviction_counter = - atomic64_read(&p->adev->num_evictions); - + amdgpu_cs_report_moved_bytes(p->adev, p->bytes_moved, + p->bytes_moved_vis); if (p->bo_list) { struct amdgpu_bo *gds = p->bo_list->gds_obj; struct amdgpu_bo *gws = p->bo_list->gws_obj; @@ -619,16 +679,11 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, } error_validate: - if (r) { - amdgpu_vm_move_pt_bos_in_lru(p->adev, &fpriv->vm); + if (r) ttm_eu_backoff_reservation(&p->ticket, &p->validated); - } error_free_pages: - if (need_mmap_lock) - up_read(¤t->mm->mmap_sem); - if (p->bo_list) { for (i = p->bo_list->first_userptr; i < p->bo_list->num_entries; ++i) { @@ -638,8 +693,7 @@ error_free_pages: continue; release_pages(e->user_pages, - e->robj->tbo.ttm->num_pages, - false); + e->robj->tbo.ttm->num_pages); kvfree(e->user_pages); } } @@ -654,7 +708,8 @@ static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p) list_for_each_entry(e, &p->validated, tv.head) { struct reservation_object *resv = e->robj->tbo.resv; - r = amdgpu_sync_resv(p->adev, &p->job->sync, resv, p->filp); + r = amdgpu_sync_resv(p->adev, &p->job->sync, resv, p->filp, + amdgpu_bo_explicit_sync(e->robj)); if (r) return r; @@ -670,21 +725,14 @@ static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p) * If error is set than unvalidate buffer, otherwise just free memory * used by parsing context. **/ -static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error, bool backoff) +static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error, + bool backoff) { - struct amdgpu_fpriv *fpriv = parser->filp->driver_priv; unsigned i; - if (!error) { - amdgpu_vm_move_pt_bos_in_lru(parser->adev, &fpriv->vm); - - ttm_eu_fence_buffer_objects(&parser->ticket, - &parser->validated, - parser->fence); - } else if (backoff) { + if (error && backoff) ttm_eu_backoff_reservation(&parser->ticket, &parser->validated); - } for (i = 0; i < parser->num_post_dep_syncobjs; i++) drm_syncobj_put(parser->post_dep_syncobjs[i]); @@ -692,8 +740,10 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error, bo dma_fence_put(parser->fence); - if (parser->ctx) + if (parser->ctx) { + mutex_unlock(&parser->ctx->lock); amdgpu_ctx_put(parser->ctx); + } if (parser->bo_list) amdgpu_bo_list_put(parser->bo_list); @@ -718,10 +768,6 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) if (r) return r; - r = amdgpu_sync_fence(adev, &p->job->sync, vm->last_dir_update); - if (r) - return r; - r = amdgpu_vm_clear_freed(adev, vm, NULL); if (r) return r; @@ -737,7 +783,8 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) if (amdgpu_sriov_vf(adev)) { struct dma_fence *f; - bo_va = vm->csa_bo_va; + + bo_va = fpriv->csa_va; BUG_ON(!bo_va); r = amdgpu_vm_bo_update(adev, bo_va, false); if (r) @@ -774,7 +821,13 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) } - r = amdgpu_vm_clear_invalids(adev, vm, &p->job->sync); + r = amdgpu_vm_handle_moved(adev, vm); + if (r) + return r; + + r = amdgpu_sync_fence(adev, &p->job->sync, vm->last_update); + if (r) + return r; if (amdgpu_vm_debug && p->bo_list) { /* Invalidate all BOs to test for userspace bugs */ @@ -784,7 +837,7 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) if (!bo) continue; - amdgpu_vm_bo_invalidate(adev, bo); + amdgpu_vm_bo_invalidate(adev, bo, false); } } @@ -797,19 +850,63 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev, struct amdgpu_fpriv *fpriv = p->filp->driver_priv; struct amdgpu_vm *vm = &fpriv->vm; struct amdgpu_ring *ring = p->job->ring; - int i, r; + int r; /* Only for UVD/VCE VM emulation */ - if (ring->funcs->parse_cs) { - for (i = 0; i < p->job->num_ibs; i++) { - r = amdgpu_ring_parse_cs(ring, p, i); + if (p->job->ring->funcs->parse_cs) { + unsigned i, j; + + for (i = 0, j = 0; i < p->nchunks && j < p->job->num_ibs; i++) { + struct drm_amdgpu_cs_chunk_ib *chunk_ib; + struct amdgpu_bo_va_mapping *m; + struct amdgpu_bo *aobj = NULL; + struct amdgpu_cs_chunk *chunk; + struct amdgpu_ib *ib; + uint64_t offset; + uint8_t *kptr; + + chunk = &p->chunks[i]; + ib = &p->job->ibs[j]; + chunk_ib = chunk->kdata; + + if (chunk->chunk_id != AMDGPU_CHUNK_ID_IB) + continue; + + r = amdgpu_cs_find_mapping(p, chunk_ib->va_start, + &aobj, &m); + if (r) { + DRM_ERROR("IB va_start is invalid\n"); + return r; + } + + if ((chunk_ib->va_start + chunk_ib->ib_bytes) > + (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) { + DRM_ERROR("IB va_start+ib_bytes is invalid\n"); + return -EINVAL; + } + + /* the IB should be reserved at this point */ + r = amdgpu_bo_kmap(aobj, (void **)&kptr); + if (r) { + return r; + } + + offset = m->start * AMDGPU_GPU_PAGE_SIZE; + kptr += chunk_ib->va_start - offset; + + memcpy(ib->ptr, kptr, chunk_ib->ib_bytes); + amdgpu_bo_kunmap(aobj); + + r = amdgpu_ring_parse_cs(ring, p, j); if (r) return r; + + j++; } } if (p->job->vm) { - p->job->vm_pd_addr = amdgpu_bo_gpu_offset(vm->root.bo); + p->job->vm_pd_addr = amdgpu_bo_gpu_offset(vm->root.base.bo); r = amdgpu_bo_vm_update_pte(p); if (r) @@ -871,54 +968,18 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev, parser->job->ring = ring; - if (ring->funcs->parse_cs) { - struct amdgpu_bo_va_mapping *m; - struct amdgpu_bo *aobj = NULL; - uint64_t offset; - uint8_t *kptr; - - m = amdgpu_cs_find_mapping(parser, chunk_ib->va_start, - &aobj); - if (!aobj) { - DRM_ERROR("IB va_start is invalid\n"); - return -EINVAL; - } - - if ((chunk_ib->va_start + chunk_ib->ib_bytes) > - (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) { - DRM_ERROR("IB va_start+ib_bytes is invalid\n"); - return -EINVAL; - } - - /* the IB should be reserved at this point */ - r = amdgpu_bo_kmap(aobj, (void **)&kptr); - if (r) { - return r; - } - - offset = m->start * AMDGPU_GPU_PAGE_SIZE; - kptr += chunk_ib->va_start - offset; - - r = amdgpu_ib_get(adev, vm, chunk_ib->ib_bytes, ib); - if (r) { - DRM_ERROR("Failed to get ib !\n"); - return r; - } - - memcpy(ib->ptr, kptr, chunk_ib->ib_bytes); - amdgpu_bo_kunmap(aobj); - } else { - r = amdgpu_ib_get(adev, vm, 0, ib); - if (r) { - DRM_ERROR("Failed to get ib !\n"); - return r; - } - + r = amdgpu_ib_get(adev, vm, + ring->funcs->parse_cs ? chunk_ib->ib_bytes : 0, + ib); + if (r) { + DRM_ERROR("Failed to get ib !\n"); + return r; } ib->gpu_addr = chunk_ib->va_start; ib->length_dw = chunk_ib->ib_bytes / 4; ib->flags = chunk_ib->flags; + j++; } @@ -928,7 +989,7 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev, parser->job->ring->funcs->type == AMDGPU_RING_TYPE_VCE)) return -EINVAL; - return 0; + return amdgpu_ctx_wait_prev_fence(parser->ctx, parser->job->ring->idx); } static int amdgpu_cs_process_fence_dep(struct amdgpu_cs_parser *p, @@ -984,7 +1045,7 @@ static int amdgpu_syncobj_lookup_and_add_to_sync(struct amdgpu_cs_parser *p, { int r; struct dma_fence *fence; - r = drm_syncobj_fence_get(p->filp, handle, &fence); + r = drm_syncobj_find_fence(p->filp, handle, &fence); if (r) return r; @@ -1028,6 +1089,9 @@ static int amdgpu_cs_process_syncobj_out_dep(struct amdgpu_cs_parser *p, GFP_KERNEL); p->num_post_dep_syncobjs = 0; + if (!p->post_dep_syncobjs) + return -ENOMEM; + for (i = 0; i < num_deps; ++i) { p->post_dep_syncobjs[i] = drm_syncobj_find(p->filp, deps[i].handle); if (!p->post_dep_syncobjs[i]) @@ -1079,14 +1143,31 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, struct amdgpu_ring *ring = p->job->ring; struct amd_sched_entity *entity = &p->ctx->rings[ring->idx].entity; struct amdgpu_job *job; + unsigned i; + uint64_t seq; + int r; + amdgpu_mn_lock(p->mn); + if (p->bo_list) { + for (i = p->bo_list->first_userptr; + i < p->bo_list->num_entries; ++i) { + struct amdgpu_bo *bo = p->bo_list->array[i].robj; + + if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm)) { + amdgpu_mn_unlock(p->mn); + return -ERESTARTSYS; + } + } + } + job = p->job; p->job = NULL; r = amd_sched_job_init(&job->base, &ring->sched, entity, p->filp); if (r) { amdgpu_job_free(job); + amdgpu_mn_unlock(p->mn); return r; } @@ -1094,22 +1175,36 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, job->fence_ctx = entity->fence_context; p->fence = dma_fence_get(&job->base.s_fence->finished); + r = amdgpu_ctx_add_fence(p->ctx, ring, p->fence, &seq); + if (r) { + dma_fence_put(p->fence); + dma_fence_put(&job->base.s_fence->finished); + amdgpu_job_free(job); + amdgpu_mn_unlock(p->mn); + return r; + } + amdgpu_cs_post_dependencies(p); - cs->out.handle = amdgpu_ctx_add_fence(p->ctx, ring, p->fence); - job->uf_sequence = cs->out.handle; + cs->out.handle = seq; + job->uf_sequence = seq; + amdgpu_job_free_resources(job); - amdgpu_cs_parser_fini(p, 0, true); + amdgpu_ring_priority_get(job->ring, + amd_sched_get_job_priority(&job->base)); trace_amdgpu_cs_ioctl(job); amd_sched_entity_push_job(&job->base); + + ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence); + amdgpu_mn_unlock(p->mn); + return 0; } int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { struct amdgpu_device *adev = dev->dev_private; - struct amdgpu_fpriv *fpriv = filp->driver_priv; union drm_amdgpu_cs *cs = data; struct amdgpu_cs_parser parser = {}; bool reserved_buffers = false; @@ -1117,8 +1212,6 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) if (!adev->accel_working) return -EBUSY; - if (amdgpu_kms_vram_lost(adev, fpriv)) - return -ENODEV; parser.adev = adev; parser.filp = filp; @@ -1129,6 +1222,10 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) goto out; } + r = amdgpu_cs_ib_fill(adev, &parser); + if (r) + goto out; + r = amdgpu_cs_parser_bos(&parser, data); if (r) { if (r == -ENOMEM) @@ -1139,9 +1236,6 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) } reserved_buffers = true; - r = amdgpu_cs_ib_fill(adev, &parser); - if (r) - goto out; r = amdgpu_cs_dependencies(adev, &parser); if (r) { @@ -1157,10 +1251,7 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) goto out; r = amdgpu_cs_submit(&parser, cs); - if (r) - goto out; - return 0; out: amdgpu_cs_parser_fini(&parser, r, reserved_buffers); return r; @@ -1180,16 +1271,12 @@ int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data, { union drm_amdgpu_wait_cs *wait = data; struct amdgpu_device *adev = dev->dev_private; - struct amdgpu_fpriv *fpriv = filp->driver_priv; unsigned long timeout = amdgpu_gem_timeout(wait->in.timeout); struct amdgpu_ring *ring = NULL; struct amdgpu_ctx *ctx; struct dma_fence *fence; long r; - if (amdgpu_kms_vram_lost(adev, fpriv)) - return -ENODEV; - ctx = amdgpu_ctx_get(filp->driver_priv, wait->in.ctx_id); if (ctx == NULL) return -EINVAL; @@ -1207,6 +1294,8 @@ int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data, r = PTR_ERR(fence); else if (fence) { r = dma_fence_wait_timeout(fence, true, timeout); + if (r > 0 && fence->error) + r = fence->error; dma_fence_put(fence); } else r = 1; @@ -1254,6 +1343,62 @@ static struct dma_fence *amdgpu_cs_get_fence(struct amdgpu_device *adev, return fence; } +int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp) +{ + struct amdgpu_device *adev = dev->dev_private; + union drm_amdgpu_fence_to_handle *info = data; + struct dma_fence *fence; + struct drm_syncobj *syncobj; + struct sync_file *sync_file; + int fd, r; + + fence = amdgpu_cs_get_fence(adev, filp, &info->in.fence); + if (IS_ERR(fence)) + return PTR_ERR(fence); + + switch (info->in.what) { + case AMDGPU_FENCE_TO_HANDLE_GET_SYNCOBJ: + r = drm_syncobj_create(&syncobj, 0, fence); + dma_fence_put(fence); + if (r) + return r; + r = drm_syncobj_get_handle(filp, syncobj, &info->out.handle); + drm_syncobj_put(syncobj); + return r; + + case AMDGPU_FENCE_TO_HANDLE_GET_SYNCOBJ_FD: + r = drm_syncobj_create(&syncobj, 0, fence); + dma_fence_put(fence); + if (r) + return r; + r = drm_syncobj_get_fd(syncobj, (int*)&info->out.handle); + drm_syncobj_put(syncobj); + return r; + + case AMDGPU_FENCE_TO_HANDLE_GET_SYNC_FILE_FD: + fd = get_unused_fd_flags(O_CLOEXEC); + if (fd < 0) { + dma_fence_put(fence); + return fd; + } + + sync_file = sync_file_create(fence); + dma_fence_put(fence); + if (!sync_file) { + put_unused_fd(fd); + return -ENOMEM; + } + + fd_install(fd, sync_file->file); + info->out.handle = fd; + return 0; + + default: + return -EINVAL; + } +} + /** * amdgpu_cs_wait_all_fence - wait on all fences to signal * @@ -1288,6 +1433,9 @@ static int amdgpu_cs_wait_all_fences(struct amdgpu_device *adev, if (r == 0) break; + + if (fence->error) + return fence->error; } memset(wait, 0, sizeof(*wait)); @@ -1333,6 +1481,7 @@ static int amdgpu_cs_wait_any_fence(struct amdgpu_device *adev, array[i] = fence; } else { /* NULL, the fence has been already signaled */ r = 1; + first = i; goto out; } } @@ -1346,8 +1495,11 @@ out: memset(wait, 0, sizeof(*wait)); wait->out.status = (r > 0); wait->out.first_signaled = first; - /* set return value 0 to indicate success */ - r = 0; + + if (first < fence_count && array[first]) + r = array[first]->error; + else + r = 0; err_free_fence_array: for (i = 0; i < fence_count; i++) @@ -1368,22 +1520,19 @@ int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { struct amdgpu_device *adev = dev->dev_private; - struct amdgpu_fpriv *fpriv = filp->driver_priv; union drm_amdgpu_wait_fences *wait = data; uint32_t fence_count = wait->in.fence_count; struct drm_amdgpu_fence *fences_user; struct drm_amdgpu_fence *fences; int r; - if (amdgpu_kms_vram_lost(adev, fpriv)) - return -ENODEV; /* Get the fences from userspace */ fences = kmalloc_array(fence_count, sizeof(struct drm_amdgpu_fence), GFP_KERNEL); if (fences == NULL) return -ENOMEM; - fences_user = (void __user *)(uintptr_t)(wait->in.fences); + fences_user = u64_to_user_ptr(wait->in.fences); if (copy_from_user(fences, fences_user, sizeof(struct drm_amdgpu_fence) * fence_count)) { r = -EFAULT; @@ -1412,78 +1561,36 @@ err_free_fences: * virtual memory address. Returns allocation structure when found, NULL * otherwise. */ -struct amdgpu_bo_va_mapping * -amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, - uint64_t addr, struct amdgpu_bo **bo) +int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, + uint64_t addr, struct amdgpu_bo **bo, + struct amdgpu_bo_va_mapping **map) { + struct amdgpu_fpriv *fpriv = parser->filp->driver_priv; + struct amdgpu_vm *vm = &fpriv->vm; struct amdgpu_bo_va_mapping *mapping; - unsigned i; - - if (!parser->bo_list) - return NULL; - - addr /= AMDGPU_GPU_PAGE_SIZE; - - for (i = 0; i < parser->bo_list->num_entries; i++) { - struct amdgpu_bo_list_entry *lobj; - - lobj = &parser->bo_list->array[i]; - if (!lobj->bo_va) - continue; - - list_for_each_entry(mapping, &lobj->bo_va->valids, list) { - if (mapping->start > addr || - addr > mapping->last) - continue; - - *bo = lobj->bo_va->bo; - return mapping; - } - - list_for_each_entry(mapping, &lobj->bo_va->invalids, list) { - if (mapping->start > addr || - addr > mapping->last) - continue; - - *bo = lobj->bo_va->bo; - return mapping; - } - } - - return NULL; -} - -/** - * amdgpu_cs_sysvm_access_required - make BOs accessible by the system VM - * - * @parser: command submission parser context - * - * Helper for UVD/VCE VM emulation, make sure BOs are accessible by the system VM. - */ -int amdgpu_cs_sysvm_access_required(struct amdgpu_cs_parser *parser) -{ - unsigned i; int r; - if (!parser->bo_list) - return 0; + addr /= AMDGPU_GPU_PAGE_SIZE; - for (i = 0; i < parser->bo_list->num_entries; i++) { - struct amdgpu_bo *bo = parser->bo_list->array[i].robj; + mapping = amdgpu_vm_bo_lookup_mapping(vm, addr); + if (!mapping || !mapping->bo_va || !mapping->bo_va->base.bo) + return -EINVAL; - r = amdgpu_ttm_bind(&bo->tbo, &bo->tbo.mem); - if (unlikely(r)) - return r; + *bo = mapping->bo_va->base.bo; + *map = mapping; - if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS) - continue; + /* Double check that the BO is reserved by this CS */ + if (READ_ONCE((*bo)->tbo.resv->lock.ctx) != &parser->ticket) + return -EINVAL; - bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; - amdgpu_ttm_placement_from_domain(bo, bo->allowed_domains); - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); - if (unlikely(r)) + if (!((*bo)->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)) { + (*bo)->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; + amdgpu_ttm_placement_from_domain(*bo, (*bo)->allowed_domains); + r = ttm_bo_validate(&(*bo)->tbo, &(*bo)->placement, false, + false); + if (r) return r; } - return 0; + return amdgpu_ttm_bind(&(*bo)->tbo, &(*bo)->tbo.mem); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index a11e44340b23..c184468e2b2b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -23,13 +23,41 @@ */ #include <drm/drmP.h> +#include <drm/drm_auth.h> #include "amdgpu.h" +#include "amdgpu_sched.h" -static int amdgpu_ctx_init(struct amdgpu_device *adev, struct amdgpu_ctx *ctx) +static int amdgpu_ctx_priority_permit(struct drm_file *filp, + enum amd_sched_priority priority) +{ + /* NORMAL and below are accessible by everyone */ + if (priority <= AMD_SCHED_PRIORITY_NORMAL) + return 0; + + if (capable(CAP_SYS_NICE)) + return 0; + + if (drm_is_current_master(filp)) + return 0; + + return -EACCES; +} + +static int amdgpu_ctx_init(struct amdgpu_device *adev, + enum amd_sched_priority priority, + struct drm_file *filp, + struct amdgpu_ctx *ctx) { unsigned i, j; int r; + if (priority < 0 || priority >= AMD_SCHED_PRIORITY_MAX) + return -EINVAL; + + r = amdgpu_ctx_priority_permit(filp, priority); + if (r) + return r; + memset(ctx, 0, sizeof(*ctx)); ctx->adev = adev; kref_init(&ctx->refcount); @@ -39,19 +67,24 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, struct amdgpu_ctx *ctx) if (!ctx->fences) return -ENOMEM; + mutex_init(&ctx->lock); + for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { ctx->rings[i].sequence = 1; ctx->rings[i].fences = &ctx->fences[amdgpu_sched_jobs * i]; } ctx->reset_counter = atomic_read(&adev->gpu_reset_counter); + ctx->vram_lost_counter = atomic_read(&adev->vram_lost_counter); + ctx->init_priority = priority; + ctx->override_priority = AMD_SCHED_PRIORITY_UNSET; /* create context entity for each ring */ for (i = 0; i < adev->num_rings; i++) { struct amdgpu_ring *ring = adev->rings[i]; struct amd_sched_rq *rq; - rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL]; + rq = &ring->sched.sched_rq[priority]; if (ring == &adev->gfx.kiq.ring) continue; @@ -96,10 +129,14 @@ static void amdgpu_ctx_fini(struct amdgpu_ctx *ctx) &ctx->rings[i].entity); amdgpu_queue_mgr_fini(adev, &ctx->queue_mgr); + + mutex_destroy(&ctx->lock); } static int amdgpu_ctx_alloc(struct amdgpu_device *adev, struct amdgpu_fpriv *fpriv, + struct drm_file *filp, + enum amd_sched_priority priority, uint32_t *id) { struct amdgpu_ctx_mgr *mgr = &fpriv->ctx_mgr; @@ -117,8 +154,9 @@ static int amdgpu_ctx_alloc(struct amdgpu_device *adev, kfree(ctx); return r; } + *id = (uint32_t)r; - r = amdgpu_ctx_init(adev, ctx); + r = amdgpu_ctx_init(adev, priority, filp, ctx); if (r) { idr_remove(&mgr->ctx_handles, *id); *id = 0; @@ -193,6 +231,7 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data, { int r; uint32_t id; + enum amd_sched_priority priority; union drm_amdgpu_ctx *args = data; struct amdgpu_device *adev = dev->dev_private; @@ -200,10 +239,16 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data, r = 0; id = args->in.ctx_id; + priority = amdgpu_to_sched_priority(args->in.priority); + + /* For backwards compatibility reasons, we need to accept + * ioctls with garbage in the priority field */ + if (priority == AMD_SCHED_PRIORITY_INVALID) + priority = AMD_SCHED_PRIORITY_NORMAL; switch (args->in.op) { case AMDGPU_CTX_OP_ALLOC_CTX: - r = amdgpu_ctx_alloc(adev, fpriv, &id); + r = amdgpu_ctx_alloc(adev, fpriv, filp, priority, &id); args->out.alloc.ctx_id = id; break; case AMDGPU_CTX_OP_FREE_CTX: @@ -246,8 +291,8 @@ int amdgpu_ctx_put(struct amdgpu_ctx *ctx) return 0; } -uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, - struct dma_fence *fence) +int amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, + struct dma_fence *fence, uint64_t* handler) { struct amdgpu_ctx_ring *cring = & ctx->rings[ring->idx]; uint64_t seq = cring->sequence; @@ -256,12 +301,8 @@ uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, idx = seq & (amdgpu_sched_jobs - 1); other = cring->fences[idx]; - if (other) { - signed long r; - r = dma_fence_wait_timeout(other, false, MAX_SCHEDULE_TIMEOUT); - if (r < 0) - DRM_ERROR("Error (%ld) waiting for fence!\n", r); - } + if (other) + BUG_ON(!dma_fence_is_signaled(other)); dma_fence_get(fence); @@ -271,8 +312,10 @@ uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, spin_unlock(&ctx->ring_lock); dma_fence_put(other); + if (handler) + *handler = seq; - return seq; + return 0; } struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx, @@ -303,6 +346,51 @@ struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx, return fence; } +void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx, + enum amd_sched_priority priority) +{ + int i; + struct amdgpu_device *adev = ctx->adev; + struct amd_sched_rq *rq; + struct amd_sched_entity *entity; + struct amdgpu_ring *ring; + enum amd_sched_priority ctx_prio; + + ctx->override_priority = priority; + + ctx_prio = (ctx->override_priority == AMD_SCHED_PRIORITY_UNSET) ? + ctx->init_priority : ctx->override_priority; + + for (i = 0; i < adev->num_rings; i++) { + ring = adev->rings[i]; + entity = &ctx->rings[i].entity; + rq = &ring->sched.sched_rq[ctx_prio]; + + if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) + continue; + + amd_sched_entity_set_rq(entity, rq); + } +} + +int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, unsigned ring_id) +{ + struct amdgpu_ctx_ring *cring = &ctx->rings[ring_id]; + unsigned idx = cring->sequence & (amdgpu_sched_jobs - 1); + struct dma_fence *other = cring->fences[idx]; + + if (other) { + signed long r; + r = dma_fence_wait_timeout(other, false, MAX_SCHEDULE_TIMEOUT); + if (r < 0) { + DRM_ERROR("Error (%ld) waiting for fence!\n", r); + return r; + } + } + + return 0; +} + void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr) { mutex_init(&mgr->lock); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 4a8fc15467cf..2c574374d9b6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -31,6 +31,7 @@ #include <linux/debugfs.h> #include <drm/drmP.h> #include <drm/drm_crtc_helper.h> +#include <drm/drm_atomic_helper.h> #include <drm/amdgpu_drm.h> #include <linux/vgaarb.h> #include <linux/vga_switcheroo.h> @@ -53,6 +54,10 @@ #include "bif/bif_4_1_d.h" #include <linux/pci.h> #include <linux/firmware.h> +#include "amdgpu_vf_error.h" + +#include "amdgpu_amdkfd.h" +#include "amdgpu_pm.h" MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin"); MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin"); @@ -62,6 +67,7 @@ MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin"); static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev); static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev); static int amdgpu_debugfs_test_ib_ring_init(struct amdgpu_device *adev); +static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev); static const char *amdgpu_asic_name[] = { "TAHITI", @@ -104,10 +110,8 @@ uint32_t amdgpu_mm_rreg(struct amdgpu_device *adev, uint32_t reg, { uint32_t ret; - if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev)) { - BUG_ON(in_interrupt()); + if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev)) return amdgpu_virt_kiq_rreg(adev, reg); - } if ((reg * 4) < adev->rmmio_size && !(acc_flags & AMDGPU_REGS_IDX)) ret = readl(((void __iomem *)adev->rmmio) + (reg * 4)); @@ -128,11 +132,13 @@ void amdgpu_mm_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, { trace_amdgpu_mm_wreg(adev->pdev->device, reg, v); - if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev)) { - BUG_ON(in_interrupt()); - return amdgpu_virt_kiq_wreg(adev, reg, v); + if (adev->asic_type >= CHIP_VEGA10 && reg == 0) { + adev->last_mm_index = v; } + if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev)) + return amdgpu_virt_kiq_wreg(adev, reg, v); + if ((reg * 4) < adev->rmmio_size && !(acc_flags & AMDGPU_REGS_IDX)) writel(v, ((void __iomem *)adev->rmmio) + (reg * 4)); else { @@ -143,6 +149,10 @@ void amdgpu_mm_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, writel(v, ((void __iomem *)adev->rmmio) + (mmMM_DATA * 4)); spin_unlock_irqrestore(&adev->mmio_idx_lock, flags); } + + if (adev->asic_type >= CHIP_VEGA10 && reg == 1 && adev->last_mm_index == 0x5702C) { + udelay(500); + } } u32 amdgpu_io_rreg(struct amdgpu_device *adev, u32 reg) @@ -157,6 +167,9 @@ u32 amdgpu_io_rreg(struct amdgpu_device *adev, u32 reg) void amdgpu_io_wreg(struct amdgpu_device *adev, u32 reg, u32 v) { + if (adev->asic_type >= CHIP_VEGA10 && reg == 0) { + adev->last_mm_index = v; + } if ((reg * 4) < adev->rio_mem_size) iowrite32(v, adev->rio_mem + (reg * 4)); @@ -164,6 +177,10 @@ void amdgpu_io_wreg(struct amdgpu_device *adev, u32 reg, u32 v) iowrite32((reg * 4), adev->rio_mem + (mmMM_INDEX * 4)); iowrite32(v, adev->rio_mem + (mmMM_DATA * 4)); } + + if (adev->asic_type >= CHIP_VEGA10 && reg == 1 && adev->last_mm_index == 0x5702C) { + udelay(500); + } } /** @@ -318,51 +335,16 @@ static void amdgpu_block_invalid_wreg(struct amdgpu_device *adev, static int amdgpu_vram_scratch_init(struct amdgpu_device *adev) { - int r; - - if (adev->vram_scratch.robj == NULL) { - r = amdgpu_bo_create(adev, AMDGPU_GPU_PAGE_SIZE, - PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, &adev->vram_scratch.robj); - if (r) { - return r; - } - } - - r = amdgpu_bo_reserve(adev->vram_scratch.robj, false); - if (unlikely(r != 0)) - return r; - r = amdgpu_bo_pin(adev->vram_scratch.robj, - AMDGPU_GEM_DOMAIN_VRAM, &adev->vram_scratch.gpu_addr); - if (r) { - amdgpu_bo_unreserve(adev->vram_scratch.robj); - return r; - } - r = amdgpu_bo_kmap(adev->vram_scratch.robj, - (void **)&adev->vram_scratch.ptr); - if (r) - amdgpu_bo_unpin(adev->vram_scratch.robj); - amdgpu_bo_unreserve(adev->vram_scratch.robj); - - return r; + return amdgpu_bo_create_kernel(adev, AMDGPU_GPU_PAGE_SIZE, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, + &adev->vram_scratch.robj, + &adev->vram_scratch.gpu_addr, + (void **)&adev->vram_scratch.ptr); } static void amdgpu_vram_scratch_fini(struct amdgpu_device *adev) { - int r; - - if (adev->vram_scratch.robj == NULL) { - return; - } - r = amdgpu_bo_reserve(adev->vram_scratch.robj, true); - if (likely(r == 0)) { - amdgpu_bo_kunmap(adev->vram_scratch.robj); - amdgpu_bo_unpin(adev->vram_scratch.robj); - amdgpu_bo_unreserve(adev->vram_scratch.robj); - } - amdgpu_bo_unref(&adev->vram_scratch.robj); + amdgpu_bo_free_kernel(&adev->vram_scratch.robj, NULL, NULL); } /** @@ -419,6 +401,15 @@ void amdgpu_pci_config_reset(struct amdgpu_device *adev) */ static int amdgpu_doorbell_init(struct amdgpu_device *adev) { + /* No doorbell on SI hardware generation */ + if (adev->asic_type < CHIP_BONAIRE) { + adev->doorbell.base = 0; + adev->doorbell.size = 0; + adev->doorbell.num_doorbells = 0; + adev->doorbell.ptr = NULL; + return 0; + } + /* doorbell bar mapping */ adev->doorbell.base = pci_resource_start(adev->pdev, 2); adev->doorbell.size = pci_resource_len(adev->pdev, 2); @@ -521,7 +512,8 @@ static int amdgpu_wb_init(struct amdgpu_device *adev) int r; if (adev->wb.wb_obj == NULL) { - r = amdgpu_bo_create_kernel(adev, AMDGPU_MAX_WB * sizeof(uint32_t), + /* AMDGPU_MAX_WB * sizeof(uint32_t) * 8 = AMDGPU_MAX_WB 256bit slots */ + r = amdgpu_bo_create_kernel(adev, AMDGPU_MAX_WB * sizeof(uint32_t) * 8, PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, &adev->wb.wb_obj, &adev->wb.gpu_addr, (void **)&adev->wb.wb); @@ -552,32 +544,10 @@ static int amdgpu_wb_init(struct amdgpu_device *adev) int amdgpu_wb_get(struct amdgpu_device *adev, u32 *wb) { unsigned long offset = find_first_zero_bit(adev->wb.used, adev->wb.num_wb); - if (offset < adev->wb.num_wb) { - __set_bit(offset, adev->wb.used); - *wb = offset; - return 0; - } else { - return -EINVAL; - } -} -/** - * amdgpu_wb_get_64bit - Allocate a wb entry - * - * @adev: amdgpu_device pointer - * @wb: wb index - * - * Allocate a wb slot for use by the driver (all asics). - * Returns 0 on success or -EINVAL on failure. - */ -int amdgpu_wb_get_64bit(struct amdgpu_device *adev, u32 *wb) -{ - unsigned long offset = bitmap_find_next_zero_area_off(adev->wb.used, - adev->wb.num_wb, 0, 2, 7, 0); - if ((offset + 1) < adev->wb.num_wb) { + if (offset < adev->wb.num_wb) { __set_bit(offset, adev->wb.used); - __set_bit(offset + 1, adev->wb.used); - *wb = offset; + *wb = offset << 3; /* convert to dw offset */ return 0; } else { return -EINVAL; @@ -595,23 +565,7 @@ int amdgpu_wb_get_64bit(struct amdgpu_device *adev, u32 *wb) void amdgpu_wb_free(struct amdgpu_device *adev, u32 wb) { if (wb < adev->wb.num_wb) - __clear_bit(wb, adev->wb.used); -} - -/** - * amdgpu_wb_free_64bit - Free a wb entry - * - * @adev: amdgpu_device pointer - * @wb: wb index - * - * Free a wb slot allocated for use by the driver (all asics) - */ -void amdgpu_wb_free_64bit(struct amdgpu_device *adev, u32 wb) -{ - if ((wb + 1) < adev->wb.num_wb) { - __clear_bit(wb, adev->wb.used); - __clear_bit(wb + 1, adev->wb.used); - } + __clear_bit(wb >> 3, adev->wb.used); } /** @@ -665,7 +619,7 @@ void amdgpu_vram_location(struct amdgpu_device *adev, struct amdgpu_mc *mc, u64 } /** - * amdgpu_gtt_location - try to find GTT location + * amdgpu_gart_location - try to find GTT location * @adev: amdgpu device structure holding all necessary informations * @mc: memory controller structure holding memory informations * @@ -676,31 +630,106 @@ void amdgpu_vram_location(struct amdgpu_device *adev, struct amdgpu_mc *mc, u64 * * FIXME: when reducing GTT size align new size on power of 2. */ -void amdgpu_gtt_location(struct amdgpu_device *adev, struct amdgpu_mc *mc) +void amdgpu_gart_location(struct amdgpu_device *adev, struct amdgpu_mc *mc) { u64 size_af, size_bf; - size_af = ((adev->mc.mc_mask - mc->vram_end) + mc->gtt_base_align) & ~mc->gtt_base_align; - size_bf = mc->vram_start & ~mc->gtt_base_align; + size_af = adev->mc.mc_mask - mc->vram_end; + size_bf = mc->vram_start; if (size_bf > size_af) { - if (mc->gtt_size > size_bf) { + if (mc->gart_size > size_bf) { dev_warn(adev->dev, "limiting GTT\n"); - mc->gtt_size = size_bf; + mc->gart_size = size_bf; } - mc->gtt_start = 0; + mc->gart_start = 0; } else { - if (mc->gtt_size > size_af) { + if (mc->gart_size > size_af) { dev_warn(adev->dev, "limiting GTT\n"); - mc->gtt_size = size_af; + mc->gart_size = size_af; } - mc->gtt_start = (mc->vram_end + 1 + mc->gtt_base_align) & ~mc->gtt_base_align; + mc->gart_start = mc->vram_end + 1; } - mc->gtt_end = mc->gtt_start + mc->gtt_size - 1; + mc->gart_end = mc->gart_start + mc->gart_size - 1; dev_info(adev->dev, "GTT: %lluM 0x%016llX - 0x%016llX\n", - mc->gtt_size >> 20, mc->gtt_start, mc->gtt_end); + mc->gart_size >> 20, mc->gart_start, mc->gart_end); } /* + * Firmware Reservation functions + */ +/** + * amdgpu_fw_reserve_vram_fini - free fw reserved vram + * + * @adev: amdgpu_device pointer + * + * free fw reserved vram if it has been reserved. + */ +void amdgpu_fw_reserve_vram_fini(struct amdgpu_device *adev) +{ + amdgpu_bo_free_kernel(&adev->fw_vram_usage.reserved_bo, + NULL, &adev->fw_vram_usage.va); +} + +/** + * amdgpu_fw_reserve_vram_init - create bo vram reservation from fw + * + * @adev: amdgpu_device pointer + * + * create bo vram reservation from fw. + */ +int amdgpu_fw_reserve_vram_init(struct amdgpu_device *adev) +{ + int r = 0; + u64 gpu_addr; + u64 vram_size = adev->mc.visible_vram_size; + + adev->fw_vram_usage.va = NULL; + adev->fw_vram_usage.reserved_bo = NULL; + + if (adev->fw_vram_usage.size > 0 && + adev->fw_vram_usage.size <= vram_size) { + + r = amdgpu_bo_create(adev, adev->fw_vram_usage.size, + PAGE_SIZE, true, 0, + AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | + AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, NULL, NULL, 0, + &adev->fw_vram_usage.reserved_bo); + if (r) + goto error_create; + + r = amdgpu_bo_reserve(adev->fw_vram_usage.reserved_bo, false); + if (r) + goto error_reserve; + r = amdgpu_bo_pin_restricted(adev->fw_vram_usage.reserved_bo, + AMDGPU_GEM_DOMAIN_VRAM, + adev->fw_vram_usage.start_offset, + (adev->fw_vram_usage.start_offset + + adev->fw_vram_usage.size), &gpu_addr); + if (r) + goto error_pin; + r = amdgpu_bo_kmap(adev->fw_vram_usage.reserved_bo, + &adev->fw_vram_usage.va); + if (r) + goto error_kmap; + + amdgpu_bo_unreserve(adev->fw_vram_usage.reserved_bo); + } + return r; + +error_kmap: + amdgpu_bo_unpin(adev->fw_vram_usage.reserved_bo); +error_pin: + amdgpu_bo_unreserve(adev->fw_vram_usage.reserved_bo); +error_reserve: + amdgpu_bo_unref(&adev->fw_vram_usage.reserved_bo); +error_create: + adev->fw_vram_usage.va = NULL; + adev->fw_vram_usage.reserved_bo = NULL; + return r; +} + + +/* * GPU helpers function. */ /** @@ -716,22 +745,6 @@ bool amdgpu_need_post(struct amdgpu_device *adev) { uint32_t reg; - if (adev->has_hw_reset) { - adev->has_hw_reset = false; - return true; - } - /* then check MEM_SIZE, in case the crtcs are off */ - reg = amdgpu_asic_get_config_memsize(adev); - - if ((reg != 0) && (reg != 0xffffffff)) - return false; - - return true; - -} - -static bool amdgpu_vpost_needed(struct amdgpu_device *adev) -{ if (amdgpu_sriov_vf(adev)) return false; @@ -754,7 +767,23 @@ static bool amdgpu_vpost_needed(struct amdgpu_device *adev) return true; } } - return amdgpu_need_post(adev); + + if (adev->has_hw_reset) { + adev->has_hw_reset = false; + return true; + } + + /* bios scratch used on CIK+ */ + if (adev->asic_type >= CHIP_BONAIRE) + return amdgpu_atombios_scratch_need_asic_init(adev); + + /* check MEM_SIZE for older asics */ + reg = amdgpu_asic_get_config_memsize(adev); + + if ((reg != 0) && (reg != 0xffffffff)) + return false; + + return true; } /** @@ -936,6 +965,20 @@ static uint32_t cail_ioreg_read(struct card_info *info, uint32_t reg) return r; } +static ssize_t amdgpu_atombios_get_vbios_version(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + struct atom_context *ctx = adev->mode_info.atom_context; + + return snprintf(buf, PAGE_SIZE, "%s\n", ctx->vbios_version); +} + +static DEVICE_ATTR(vbios_version, 0444, amdgpu_atombios_get_vbios_version, + NULL); + /** * amdgpu_atombios_fini - free the driver info and callbacks for atombios * @@ -955,6 +998,7 @@ static void amdgpu_atombios_fini(struct amdgpu_device *adev) adev->mode_info.atom_context = NULL; kfree(adev->mode_info.atom_card_info); adev->mode_info.atom_card_info = NULL; + device_remove_file(adev->dev, &dev_attr_vbios_version); } /** @@ -971,6 +1015,7 @@ static int amdgpu_atombios_init(struct amdgpu_device *adev) { struct card_info *atom_card_info = kzalloc(sizeof(struct card_info), GFP_KERNEL); + int ret; if (!atom_card_info) return -ENOMEM; @@ -1007,6 +1052,13 @@ static int amdgpu_atombios_init(struct amdgpu_device *adev) amdgpu_atombios_scratch_regs_init(adev); amdgpu_atombios_allocate_fb_scratch(adev); } + + ret = device_create_file(adev->dev, &dev_attr_vbios_version); + if (ret) { + DRM_ERROR("Failed to create device file for VBIOS version\n"); + return ret; + } + return 0; } @@ -1031,19 +1083,6 @@ static unsigned int amdgpu_vga_set_decode(void *cookie, bool state) return VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM; } -/** - * amdgpu_check_pot_argument - check that argument is a power of two - * - * @arg: value to check - * - * Validates that a certain argument is a power of two (all asics). - * Returns true if argument is valid. - */ -static bool amdgpu_check_pot_argument(int arg) -{ - return (arg & (arg - 1)) == 0; -} - static void amdgpu_check_block_size(struct amdgpu_device *adev) { /* defines number of bits in page table versus page directory, @@ -1077,7 +1116,7 @@ static void amdgpu_check_vm_size(struct amdgpu_device *adev) if (amdgpu_vm_size == -1) return; - if (!amdgpu_check_pot_argument(amdgpu_vm_size)) { + if (!is_power_of_2(amdgpu_vm_size)) { dev_warn(adev->dev, "VM size (%d) must be a power of 2\n", amdgpu_vm_size); goto def_value; @@ -1118,19 +1157,31 @@ static void amdgpu_check_arguments(struct amdgpu_device *adev) dev_warn(adev->dev, "sched jobs (%d) must be at least 4\n", amdgpu_sched_jobs); amdgpu_sched_jobs = 4; - } else if (!amdgpu_check_pot_argument(amdgpu_sched_jobs)){ + } else if (!is_power_of_2(amdgpu_sched_jobs)){ dev_warn(adev->dev, "sched jobs (%d) must be a power of 2\n", amdgpu_sched_jobs); amdgpu_sched_jobs = roundup_pow_of_two(amdgpu_sched_jobs); } - if (amdgpu_gart_size != -1) { + if (amdgpu_gart_size != -1 && amdgpu_gart_size < 32) { + /* gart size must be greater or equal to 32M */ + dev_warn(adev->dev, "gart size (%d) too small\n", + amdgpu_gart_size); + amdgpu_gart_size = -1; + } + + if (amdgpu_gtt_size != -1 && amdgpu_gtt_size < 32) { /* gtt size must be greater or equal to 32M */ - if (amdgpu_gart_size < 32) { - dev_warn(adev->dev, "gart size (%d) too small\n", - amdgpu_gart_size); - amdgpu_gart_size = -1; - } + dev_warn(adev->dev, "gtt size (%d) too small\n", + amdgpu_gtt_size); + amdgpu_gtt_size = -1; + } + + /* valid range is between 4 and 9 inclusive */ + if (amdgpu_vm_fragment_size != -1 && + (amdgpu_vm_fragment_size > 9 || amdgpu_vm_fragment_size < 4)) { + dev_warn(adev->dev, "valid range is between 4 and 9\n"); + amdgpu_vm_fragment_size = -1; } amdgpu_check_vm_size(adev); @@ -1138,7 +1189,7 @@ static void amdgpu_check_arguments(struct amdgpu_device *adev) amdgpu_check_block_size(adev); if (amdgpu_vram_page_split != -1 && (amdgpu_vram_page_split < 16 || - !amdgpu_check_pot_argument(amdgpu_vram_page_split))) { + !is_power_of_2(amdgpu_vram_page_split))) { dev_warn(adev->dev, "invalid VRAM page split (%d)\n", amdgpu_vram_page_split); amdgpu_vram_page_split = 1024; @@ -1786,6 +1837,9 @@ static int amdgpu_fini(struct amdgpu_device *adev) adev->ip_blocks[i].status.hw = false; } + if (adev->firmware.load_type == AMDGPU_FW_LOAD_SMU) + amdgpu_ucode_fini_bo(adev); + for (i = adev->num_ip_blocks - 1; i >= 0; i--) { if (!adev->ip_blocks[i].status.sw) continue; @@ -1807,10 +1861,8 @@ static int amdgpu_fini(struct amdgpu_device *adev) adev->ip_blocks[i].status.late_initialized = false; } - if (amdgpu_sriov_vf(adev)) { - amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL); + if (amdgpu_sriov_vf(adev)) amdgpu_virt_release_full_gpu(adev, false); - } return 0; } @@ -1898,10 +1950,12 @@ static int amdgpu_sriov_reinit_late(struct amdgpu_device *adev) static enum amd_ip_block_type ip_order[] = { AMD_IP_BLOCK_TYPE_SMC, + AMD_IP_BLOCK_TYPE_PSP, AMD_IP_BLOCK_TYPE_DCE, AMD_IP_BLOCK_TYPE_GFX, AMD_IP_BLOCK_TYPE_SDMA, - AMD_IP_BLOCK_TYPE_VCE, + AMD_IP_BLOCK_TYPE_UVD, + AMD_IP_BLOCK_TYPE_VCE }; for (i = 0; i < ARRAY_SIZE(ip_order); i++) { @@ -1982,16 +2036,67 @@ static int amdgpu_resume(struct amdgpu_device *adev) static void amdgpu_device_detect_sriov_bios(struct amdgpu_device *adev) { - if (adev->is_atom_fw) { - if (amdgpu_atomfirmware_gpu_supports_virtualization(adev)) - adev->virt.caps |= AMDGPU_SRIOV_CAPS_SRIOV_VBIOS; - } else { - if (amdgpu_atombios_has_gpu_virtualization_table(adev)) - adev->virt.caps |= AMDGPU_SRIOV_CAPS_SRIOV_VBIOS; + if (amdgpu_sriov_vf(adev)) { + if (adev->is_atom_fw) { + if (amdgpu_atomfirmware_gpu_supports_virtualization(adev)) + adev->virt.caps |= AMDGPU_SRIOV_CAPS_SRIOV_VBIOS; + } else { + if (amdgpu_atombios_has_gpu_virtualization_table(adev)) + adev->virt.caps |= AMDGPU_SRIOV_CAPS_SRIOV_VBIOS; + } + + if (!(adev->virt.caps & AMDGPU_SRIOV_CAPS_SRIOV_VBIOS)) + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_NO_VBIOS, 0, 0); + } +} + +bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type) +{ + switch (asic_type) { +#if defined(CONFIG_DRM_AMD_DC) + case CHIP_BONAIRE: + case CHIP_HAWAII: + case CHIP_KAVERI: + case CHIP_CARRIZO: + case CHIP_STONEY: + case CHIP_POLARIS11: + case CHIP_POLARIS10: + case CHIP_POLARIS12: + case CHIP_TONGA: + case CHIP_FIJI: +#if defined(CONFIG_DRM_AMD_DC_PRE_VEGA) + return amdgpu_dc != 0; +#endif + case CHIP_KABINI: + case CHIP_MULLINS: + return amdgpu_dc > 0; + case CHIP_VEGA10: +#if defined(CONFIG_DRM_AMD_DC_DCN1_0) + case CHIP_RAVEN: +#endif + return amdgpu_dc != 0; +#endif + default: + return false; } } /** + * amdgpu_device_has_dc_support - check if dc is supported + * + * @adev: amdgpu_device_pointer + * + * Returns true for supported, false for not supported + */ +bool amdgpu_device_has_dc_support(struct amdgpu_device *adev) +{ + if (amdgpu_sriov_vf(adev)) + return false; + + return amdgpu_device_asic_has_dc_support(adev->asic_type); +} + +/** * amdgpu_device_init - initialize the driver * * @adev: amdgpu_device pointer @@ -2019,7 +2124,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, adev->flags = flags; adev->asic_type = flags & AMD_ASIC_MASK; adev->usec_timeout = AMDGPU_MAX_USEC_TIMEOUT; - adev->mc.gtt_size = 512 * 1024 * 1024; + adev->mc.gart_size = 512 * 1024 * 1024; adev->accel_working = false; adev->num_rings = 0; adev->mman.buffer_funcs = NULL; @@ -2028,6 +2133,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, adev->vm_manager.vm_pte_num_rings = 0; adev->gart.gart_funcs = NULL; adev->fence_context = dma_fence_context_alloc(AMDGPU_MAX_RINGS); + bitmap_zero(adev->gfx.pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); adev->smc_rreg = &amdgpu_invalid_rreg; adev->smc_wreg = &amdgpu_invalid_wreg; @@ -2044,7 +2150,6 @@ int amdgpu_device_init(struct amdgpu_device *adev, adev->audio_endpt_rreg = &amdgpu_block_invalid_rreg; adev->audio_endpt_wreg = &amdgpu_block_invalid_wreg; - DRM_INFO("initializing kernel modesetting (%s 0x%04X:0x%04X 0x%04X:0x%04X 0x%02X).\n", amdgpu_asic_name[adev->asic_type], pdev->vendor, pdev->device, pdev->subsystem_vendor, pdev->subsystem_device, pdev->revision); @@ -2056,8 +2161,10 @@ int amdgpu_device_init(struct amdgpu_device *adev, mutex_init(&adev->pm.mutex); mutex_init(&adev->gfx.gpu_clock_mutex); mutex_init(&adev->srbm_mutex); + mutex_init(&adev->gfx.pipe_reserve_mutex); mutex_init(&adev->grbm_idx_mutex); mutex_init(&adev->mn_lock); + mutex_init(&adev->virt.vf_errors.lock); hash_init(adev->mn_hash); amdgpu_check_arguments(adev); @@ -2068,6 +2175,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, spin_lock_init(&adev->uvd_ctx_idx_lock); spin_lock_init(&adev->didt_idx_lock); spin_lock_init(&adev->gc_cac_idx_lock); + spin_lock_init(&adev->se_cac_idx_lock); spin_lock_init(&adev->audio_endpt_idx_lock); spin_lock_init(&adev->mm_stats.lock); @@ -2099,9 +2207,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, DRM_INFO("register mmio base: 0x%08X\n", (uint32_t)adev->rmmio_base); DRM_INFO("register mmio size: %u\n", (unsigned)adev->rmmio_size); - if (adev->asic_type >= CHIP_BONAIRE) - /* doorbell bar mapping */ - amdgpu_doorbell_init(adev); + /* doorbell bar mapping */ + amdgpu_doorbell_init(adev); /* io port mapping */ for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) { @@ -2143,6 +2250,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_atombios_init(adev); if (r) { dev_err(adev->dev, "amdgpu_atombios_init failed\n"); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0); goto failed; } @@ -2150,7 +2258,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, amdgpu_device_detect_sriov_bios(adev); /* Post card if necessary */ - if (amdgpu_vpost_needed(adev)) { + if (amdgpu_need_post(adev)) { if (!adev->bios) { dev_err(adev->dev, "no vBIOS found\n"); r = -EINVAL; @@ -2166,21 +2274,32 @@ int amdgpu_device_init(struct amdgpu_device *adev, DRM_INFO("GPU post is not needed\n"); } - if (!adev->is_atom_fw) { + if (adev->is_atom_fw) { + /* Initialize clocks */ + r = amdgpu_atomfirmware_get_clock_info(adev); + if (r) { + dev_err(adev->dev, "amdgpu_atomfirmware_get_clock_info failed\n"); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0); + goto failed; + } + } else { /* Initialize clocks */ r = amdgpu_atombios_get_clock_info(adev); if (r) { dev_err(adev->dev, "amdgpu_atombios_get_clock_info failed\n"); - return r; + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0); + goto failed; } /* init i2c buses */ - amdgpu_atombios_i2c_init(adev); + if (!amdgpu_device_has_dc_support(adev)) + amdgpu_atombios_i2c_init(adev); } /* Fence driver */ r = amdgpu_fence_driver_init(adev); if (r) { dev_err(adev->dev, "amdgpu_fence_driver_init failed\n"); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_FENCE_INIT_FAIL, 0, 0); goto failed; } @@ -2190,6 +2309,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_init(adev); if (r) { dev_err(adev->dev, "amdgpu_init failed\n"); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0); amdgpu_fini(adev); goto failed; } @@ -2209,6 +2329,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_ib_pool_init(adev); if (r) { dev_err(adev->dev, "IB initialization failed (%d).\n", r); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r); goto failed; } @@ -2216,8 +2337,15 @@ int amdgpu_device_init(struct amdgpu_device *adev, if (r) DRM_ERROR("ib ring test failed (%d).\n", r); + if (amdgpu_sriov_vf(adev)) + amdgpu_virt_init_data_exchange(adev); + amdgpu_fbdev_init(adev); + r = amdgpu_pm_sysfs_init(adev); + if (r) + DRM_ERROR("registering pm debugfs failed (%d).\n", r); + r = amdgpu_gem_debugfs_init(adev); if (r) DRM_ERROR("registering gem debugfs failed (%d).\n", r); @@ -2234,6 +2362,10 @@ int amdgpu_device_init(struct amdgpu_device *adev, if (r) DRM_ERROR("registering firmware debugfs failed (%d).\n", r); + r = amdgpu_debugfs_vbios_dump_init(adev); + if (r) + DRM_ERROR("Creating vbios dump debugfs failed (%d).\n", r); + if ((amdgpu_testing & 1)) { if (adev->accel_working) amdgpu_test_moves(adev); @@ -2253,12 +2385,14 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_late_init(adev); if (r) { dev_err(adev->dev, "amdgpu_late_init failed\n"); + amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_LATE_INIT_FAIL, 0, r); goto failed; } return 0; failed: + amdgpu_vf_error_trans_all(adev); if (runtime) vga_switcheroo_fini_domain_pm_ops(adev->dev); return r; @@ -2283,6 +2417,7 @@ void amdgpu_device_fini(struct amdgpu_device *adev) /* evict vram memory */ amdgpu_bo_evict_vram(adev); amdgpu_ib_pool_fini(adev); + amdgpu_fw_reserve_vram_fini(adev); amdgpu_fence_driver_fini(adev); amdgpu_fbdev_fini(adev); r = amdgpu_fini(adev); @@ -2293,7 +2428,8 @@ void amdgpu_device_fini(struct amdgpu_device *adev) adev->accel_working = false; cancel_delayed_work_sync(&adev->late_init_work); /* free i2c buses */ - amdgpu_i2c_fini(adev); + if (!amdgpu_device_has_dc_support(adev)) + amdgpu_i2c_fini(adev); amdgpu_atombios_fini(adev); kfree(adev->bios); adev->bios = NULL; @@ -2307,8 +2443,8 @@ void amdgpu_device_fini(struct amdgpu_device *adev) adev->rio_mem = NULL; iounmap(adev->rmmio); adev->rmmio = NULL; - if (adev->asic_type >= CHIP_BONAIRE) - amdgpu_doorbell_fini(adev); + amdgpu_doorbell_fini(adev); + amdgpu_pm_sysfs_fini(adev); amdgpu_debugfs_regs_cleanup(adev); } @@ -2344,12 +2480,16 @@ int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon) drm_kms_helper_poll_disable(dev); - /* turn off display hw */ - drm_modeset_lock_all(dev); - list_for_each_entry(connector, &dev->mode_config.connector_list, head) { - drm_helper_connector_dpms(connector, DRM_MODE_DPMS_OFF); + if (!amdgpu_device_has_dc_support(adev)) { + /* turn off display hw */ + drm_modeset_lock_all(dev); + list_for_each_entry(connector, &dev->mode_config.connector_list, head) { + drm_helper_connector_dpms(connector, DRM_MODE_DPMS_OFF); + } + drm_modeset_unlock_all(dev); } - drm_modeset_unlock_all(dev); + + amdgpu_amdkfd_suspend(adev); /* unpin the front buffers and cursors */ list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { @@ -2392,10 +2532,7 @@ int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon) */ amdgpu_bo_evict_vram(adev); - if (adev->is_atom_fw) - amdgpu_atomfirmware_scratch_regs_save(adev); - else - amdgpu_atombios_scratch_regs_save(adev); + amdgpu_atombios_scratch_regs_save(adev); pci_save_state(dev->pdev); if (suspend) { /* Shut down the device */ @@ -2444,10 +2581,7 @@ int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon) if (r) goto unlock; } - if (adev->is_atom_fw) - amdgpu_atomfirmware_scratch_regs_restore(adev); - else - amdgpu_atombios_scratch_regs_restore(adev); + amdgpu_atombios_scratch_regs_restore(adev); /* post card */ if (amdgpu_need_post(adev)) { @@ -2490,16 +2624,31 @@ int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon) } } } + r = amdgpu_amdkfd_resume(adev); + if (r) + return r; /* blat the mode back in */ if (fbcon) { - drm_helper_resume_force_mode(dev); - /* turn on display hw */ - drm_modeset_lock_all(dev); - list_for_each_entry(connector, &dev->mode_config.connector_list, head) { - drm_helper_connector_dpms(connector, DRM_MODE_DPMS_ON); + if (!amdgpu_device_has_dc_support(adev)) { + /* pre DCE11 */ + drm_helper_resume_force_mode(dev); + + /* turn on display hw */ + drm_modeset_lock_all(dev); + list_for_each_entry(connector, &dev->mode_config.connector_list, head) { + drm_helper_connector_dpms(connector, DRM_MODE_DPMS_ON); + } + drm_modeset_unlock_all(dev); + } else { + /* + * There is no equivalent atomic helper to turn on + * display, so we defined our own function for this, + * once suspend resume is supported by the atomic + * framework this will be reworked + */ + amdgpu_dm_display_resume(adev); } - drm_modeset_unlock_all(dev); } drm_kms_helper_poll_enable(dev); @@ -2516,7 +2665,10 @@ int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon) #ifdef CONFIG_PM dev->dev->power.disable_depth++; #endif - drm_helper_hpd_irq_event(dev); + if (!amdgpu_device_has_dc_support(adev)) + drm_helper_hpd_irq_event(dev); + else + drm_kms_helper_hotplug_event(dev); #ifdef CONFIG_PM dev->dev->power.disable_depth--; #endif @@ -2536,6 +2688,9 @@ static bool amdgpu_check_soft_reset(struct amdgpu_device *adev) int i; bool asic_hang = false; + if (amdgpu_sriov_vf(adev)) + return true; + for (i = 0; i < adev->num_ip_blocks; i++) { if (!adev->ip_blocks[i].status.valid) continue; @@ -2578,7 +2733,8 @@ static bool amdgpu_need_full_reset(struct amdgpu_device *adev) if ((adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) || (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) || (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_ACP) || - (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE)) { + (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE) || + adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP) { if (adev->ip_blocks[i].status.hang) { DRM_INFO("Some block need full reset!\n"); return true; @@ -2654,12 +2810,6 @@ static int amdgpu_recover_vram_from_shadow(struct amdgpu_device *adev, goto err; } - r = amdgpu_ttm_bind(&bo->shadow->tbo, &bo->shadow->tbo.mem); - if (r) { - DRM_ERROR("%p bind failed\n", bo->shadow); - goto err; - } - r = amdgpu_bo_restore_from_shadow(adev, ring, bo, NULL, fence, true); if (r) { @@ -2692,7 +2842,7 @@ int amdgpu_sriov_gpu_reset(struct amdgpu_device *adev, struct amdgpu_job *job) mutex_lock(&adev->virt.lock_reset); atomic_inc(&adev->gpu_reset_counter); - adev->gfx.in_reset = true; + adev->in_sriov_reset = true; /* block TTM */ resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev); @@ -2803,7 +2953,7 @@ give_up_reset: dev_info(adev->dev, "GPU reset successed!\n"); } - adev->gfx.in_reset = false; + adev->in_sriov_reset = false; mutex_unlock(&adev->virt.lock_reset); return r; } @@ -2818,6 +2968,7 @@ give_up_reset: */ int amdgpu_gpu_reset(struct amdgpu_device *adev) { + struct drm_atomic_state *state = NULL; int i, r; int resched; bool need_full_reset, vram_lost = false; @@ -2831,6 +2982,9 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev) /* block TTM */ resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev); + /* store modesetting */ + if (amdgpu_device_has_dc_support(adev)) + state = drm_atomic_helper_suspend(adev->ddev); /* block scheduler */ for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { @@ -2860,21 +3014,9 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev) r = amdgpu_suspend(adev); retry: - /* Disable fb access */ - if (adev->mode_info.num_crtc) { - struct amdgpu_mode_mc_save save; - amdgpu_display_stop_mc_access(adev, &save); - amdgpu_wait_for_idle(adev, AMD_IP_BLOCK_TYPE_GMC); - } - if (adev->is_atom_fw) - amdgpu_atomfirmware_scratch_regs_save(adev); - else - amdgpu_atombios_scratch_regs_save(adev); + amdgpu_atombios_scratch_regs_save(adev); r = amdgpu_asic_reset(adev); - if (adev->is_atom_fw) - amdgpu_atomfirmware_scratch_regs_restore(adev); - else - amdgpu_atombios_scratch_regs_restore(adev); + amdgpu_atombios_scratch_regs_restore(adev); /* post card */ amdgpu_atom_asic_init(adev->mode_info.atom_context); @@ -2959,15 +3101,22 @@ out: } } - drm_helper_resume_force_mode(adev->ddev); + if (amdgpu_device_has_dc_support(adev)) { + r = drm_atomic_helper_resume(adev->ddev, state); + amdgpu_dm_display_resume(adev); + } else + drm_helper_resume_force_mode(adev->ddev); ttm_bo_unlock_delayed_workqueue(&adev->mman.bdev, resched); - if (r) + if (r) { /* bad news, how to tell it to userspace ? */ dev_info(adev->dev, "GPU reset failed\n"); - else + } + else { dev_info(adev->dev, "GPU reset successed!\n"); + } + amdgpu_vf_error_trans_all(adev); return r; } @@ -3115,9 +3264,9 @@ static ssize_t amdgpu_debugfs_regs_read(struct file *f, char __user *buf, pm_pg_lock = (*pos >> 23) & 1; if (*pos & (1ULL << 62)) { - se_bank = (*pos >> 24) & 0x3FF; - sh_bank = (*pos >> 34) & 0x3FF; - instance_bank = (*pos >> 44) & 0x3FF; + se_bank = (*pos & GENMASK_ULL(33, 24)) >> 24; + sh_bank = (*pos & GENMASK_ULL(43, 34)) >> 34; + instance_bank = (*pos & GENMASK_ULL(53, 44)) >> 44; if (se_bank == 0x3FF) se_bank = 0xFFFFFFFF; @@ -3191,9 +3340,9 @@ static ssize_t amdgpu_debugfs_regs_write(struct file *f, const char __user *buf, pm_pg_lock = (*pos >> 23) & 1; if (*pos & (1ULL << 62)) { - se_bank = (*pos >> 24) & 0x3FF; - sh_bank = (*pos >> 34) & 0x3FF; - instance_bank = (*pos >> 44) & 0x3FF; + se_bank = (*pos & GENMASK_ULL(33, 24)) >> 24; + sh_bank = (*pos & GENMASK_ULL(43, 34)) >> 34; + instance_bank = (*pos & GENMASK_ULL(53, 44)) >> 44; if (se_bank == 0x3FF) se_bank = 0xFFFFFFFF; @@ -3508,10 +3657,7 @@ static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf, valuesize = sizeof(values); if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->read_sensor) - r = adev->powerplay.pp_funcs->read_sensor(adev->powerplay.pp_handle, idx, &values[0], &valuesize); - else if (adev->pm.funcs && adev->pm.funcs->read_sensor) - r = adev->pm.funcs->read_sensor(adev, idx, &values[0], - &valuesize); + r = amdgpu_dpm_read_sensor(adev, idx, &values[0], &valuesize); else return -EINVAL; @@ -3544,12 +3690,12 @@ static ssize_t amdgpu_debugfs_wave_read(struct file *f, char __user *buf, return -EINVAL; /* decode offset */ - offset = (*pos & 0x7F); - se = ((*pos >> 7) & 0xFF); - sh = ((*pos >> 15) & 0xFF); - cu = ((*pos >> 23) & 0xFF); - wave = ((*pos >> 31) & 0xFF); - simd = ((*pos >> 37) & 0xFF); + offset = (*pos & GENMASK_ULL(6, 0)); + se = (*pos & GENMASK_ULL(14, 7)) >> 7; + sh = (*pos & GENMASK_ULL(22, 15)) >> 15; + cu = (*pos & GENMASK_ULL(30, 23)) >> 23; + wave = (*pos & GENMASK_ULL(36, 31)) >> 31; + simd = (*pos & GENMASK_ULL(44, 37)) >> 37; /* switch to the specific se/sh/cu */ mutex_lock(&adev->grbm_idx_mutex); @@ -3594,14 +3740,14 @@ static ssize_t amdgpu_debugfs_gpr_read(struct file *f, char __user *buf, return -EINVAL; /* decode offset */ - offset = (*pos & 0xFFF); /* in dwords */ - se = ((*pos >> 12) & 0xFF); - sh = ((*pos >> 20) & 0xFF); - cu = ((*pos >> 28) & 0xFF); - wave = ((*pos >> 36) & 0xFF); - simd = ((*pos >> 44) & 0xFF); - thread = ((*pos >> 52) & 0xFF); - bank = ((*pos >> 60) & 1); + offset = *pos & GENMASK_ULL(11, 0); + se = (*pos & GENMASK_ULL(19, 12)) >> 12; + sh = (*pos & GENMASK_ULL(27, 20)) >> 20; + cu = (*pos & GENMASK_ULL(35, 28)) >> 28; + wave = (*pos & GENMASK_ULL(43, 36)) >> 36; + simd = (*pos & GENMASK_ULL(51, 44)) >> 44; + thread = (*pos & GENMASK_ULL(59, 52)) >> 52; + bank = (*pos & GENMASK_ULL(61, 60)) >> 60; data = kmalloc_array(1024, sizeof(*data), GFP_KERNEL); if (!data) @@ -3799,6 +3945,28 @@ int amdgpu_debugfs_init(struct drm_minor *minor) { return 0; } + +static int amdgpu_debugfs_get_vbios_dump(struct seq_file *m, void *data) +{ + struct drm_info_node *node = (struct drm_info_node *) m->private; + struct drm_device *dev = node->minor->dev; + struct amdgpu_device *adev = dev->dev_private; + + seq_write(m, adev->bios, adev->bios_size); + return 0; +} + +static const struct drm_info_list amdgpu_vbios_dump_list[] = { + {"amdgpu_vbios", + amdgpu_debugfs_get_vbios_dump, + 0, NULL}, +}; + +static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev) +{ + return amdgpu_debugfs_add_files(adev, + amdgpu_vbios_dump_list, 1); +} #else static int amdgpu_debugfs_test_ib_ring_init(struct amdgpu_device *adev) { @@ -3808,5 +3976,9 @@ static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev) { return 0; } +static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev) +{ + return 0; +} static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev) { } #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index cdf2ab20166a..138beb550a58 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -482,7 +482,7 @@ static void amdgpu_user_framebuffer_destroy(struct drm_framebuffer *fb) { struct amdgpu_framebuffer *amdgpu_fb = to_amdgpu_framebuffer(fb); - drm_gem_object_unreference_unlocked(amdgpu_fb->obj); + drm_gem_object_put_unlocked(amdgpu_fb->obj); drm_framebuffer_cleanup(fb); kfree(amdgpu_fb); } @@ -518,7 +518,7 @@ amdgpu_framebuffer_init(struct drm_device *dev, return 0; } -static struct drm_framebuffer * +struct drm_framebuffer * amdgpu_user_framebuffer_create(struct drm_device *dev, struct drm_file *file_priv, const struct drm_mode_fb_cmd2 *mode_cmd) @@ -542,21 +542,21 @@ amdgpu_user_framebuffer_create(struct drm_device *dev, amdgpu_fb = kzalloc(sizeof(*amdgpu_fb), GFP_KERNEL); if (amdgpu_fb == NULL) { - drm_gem_object_unreference_unlocked(obj); + drm_gem_object_put_unlocked(obj); return ERR_PTR(-ENOMEM); } ret = amdgpu_framebuffer_init(dev, amdgpu_fb, mode_cmd, obj); if (ret) { kfree(amdgpu_fb); - drm_gem_object_unreference_unlocked(obj); + drm_gem_object_put_unlocked(obj); return ERR_PTR(ret); } return &amdgpu_fb->base; } -static void amdgpu_output_poll_changed(struct drm_device *dev) +void amdgpu_output_poll_changed(struct drm_device *dev) { struct amdgpu_device *adev = dev->dev_private; amdgpu_fb_output_poll_changed(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h new file mode 100644 index 000000000000..3cc0ef0c055e --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h @@ -0,0 +1,33 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __AMDGPU_DISPLAY_H__ +#define __AMDGPU_DISPLAY_H__ + +struct drm_framebuffer * +amdgpu_user_framebuffer_create(struct drm_device *dev, + struct drm_file *file_priv, + const struct drm_mode_fb_cmd2 *mode_cmd); + +void amdgpu_output_poll_changed(struct drm_device *dev); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c index 1cb52fd19060..e997ebbe43ea 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c @@ -960,8 +960,10 @@ u8 amdgpu_encode_pci_lane_width(u32 lanes) } struct amd_vce_state* -amdgpu_get_vce_clock_state(struct amdgpu_device *adev, unsigned idx) +amdgpu_get_vce_clock_state(void *handle, u32 idx) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (idx < adev->pm.dpm.num_of_vce_states) return &adev->pm.dpm.vce_states[idx]; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h index 8c96a4caa715..56caaeee6fea 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h @@ -241,179 +241,125 @@ enum amdgpu_pcie_gen { AMDGPU_PCIE_GEN_INVALID = 0xffff }; -struct amdgpu_dpm_funcs { - int (*get_temperature)(struct amdgpu_device *adev); - int (*pre_set_power_state)(struct amdgpu_device *adev); - int (*set_power_state)(struct amdgpu_device *adev); - void (*post_set_power_state)(struct amdgpu_device *adev); - void (*display_configuration_changed)(struct amdgpu_device *adev); - u32 (*get_sclk)(struct amdgpu_device *adev, bool low); - u32 (*get_mclk)(struct amdgpu_device *adev, bool low); - void (*print_power_state)(struct amdgpu_device *adev, struct amdgpu_ps *ps); - void (*debugfs_print_current_performance_level)(struct amdgpu_device *adev, struct seq_file *m); - int (*force_performance_level)(struct amdgpu_device *adev, enum amd_dpm_forced_level level); - bool (*vblank_too_short)(struct amdgpu_device *adev); - void (*powergate_uvd)(struct amdgpu_device *adev, bool gate); - void (*powergate_vce)(struct amdgpu_device *adev, bool gate); - void (*enable_bapm)(struct amdgpu_device *adev, bool enable); - void (*set_fan_control_mode)(struct amdgpu_device *adev, u32 mode); - u32 (*get_fan_control_mode)(struct amdgpu_device *adev); - int (*set_fan_speed_percent)(struct amdgpu_device *adev, u32 speed); - int (*get_fan_speed_percent)(struct amdgpu_device *adev, u32 *speed); - int (*force_clock_level)(struct amdgpu_device *adev, enum pp_clock_type type, uint32_t mask); - int (*print_clock_levels)(struct amdgpu_device *adev, enum pp_clock_type type, char *buf); - int (*get_sclk_od)(struct amdgpu_device *adev); - int (*set_sclk_od)(struct amdgpu_device *adev, uint32_t value); - int (*get_mclk_od)(struct amdgpu_device *adev); - int (*set_mclk_od)(struct amdgpu_device *adev, uint32_t value); - int (*check_state_equal)(struct amdgpu_device *adev, - struct amdgpu_ps *cps, - struct amdgpu_ps *rps, - bool *equal); - int (*read_sensor)(struct amdgpu_device *adev, int idx, void *value, - int *size); - - struct amd_vce_state* (*get_vce_clock_state)(struct amdgpu_device *adev, unsigned idx); - int (*reset_power_profile_state)(struct amdgpu_device *adev, - struct amd_pp_profile *request); - int (*get_power_profile_state)(struct amdgpu_device *adev, - struct amd_pp_profile *query); - int (*set_power_profile_state)(struct amdgpu_device *adev, - struct amd_pp_profile *request); - int (*switch_power_profile)(struct amdgpu_device *adev, - enum amd_pp_profile_type type); -}; +#define amdgpu_dpm_pre_set_power_state(adev) \ + ((adev)->powerplay.pp_funcs->pre_set_power_state((adev)->powerplay.pp_handle)) + +#define amdgpu_dpm_set_power_state(adev) \ + ((adev)->powerplay.pp_funcs->set_power_state((adev)->powerplay.pp_handle)) + +#define amdgpu_dpm_post_set_power_state(adev) \ + ((adev)->powerplay.pp_funcs->post_set_power_state((adev)->powerplay.pp_handle)) + +#define amdgpu_dpm_display_configuration_changed(adev) \ + ((adev)->powerplay.pp_funcs->display_configuration_changed((adev)->powerplay.pp_handle)) -#define amdgpu_dpm_pre_set_power_state(adev) (adev)->pm.funcs->pre_set_power_state((adev)) -#define amdgpu_dpm_set_power_state(adev) (adev)->pm.funcs->set_power_state((adev)) -#define amdgpu_dpm_post_set_power_state(adev) (adev)->pm.funcs->post_set_power_state((adev)) -#define amdgpu_dpm_display_configuration_changed(adev) (adev)->pm.funcs->display_configuration_changed((adev)) -#define amdgpu_dpm_print_power_state(adev, ps) (adev)->pm.funcs->print_power_state((adev), (ps)) -#define amdgpu_dpm_vblank_too_short(adev) (adev)->pm.funcs->vblank_too_short((adev)) -#define amdgpu_dpm_enable_bapm(adev, e) (adev)->pm.funcs->enable_bapm((adev), (e)) +#define amdgpu_dpm_print_power_state(adev, ps) \ + ((adev)->powerplay.pp_funcs->print_power_state((adev)->powerplay.pp_handle, (ps))) + +#define amdgpu_dpm_vblank_too_short(adev) \ + ((adev)->powerplay.pp_funcs->vblank_too_short((adev)->powerplay.pp_handle)) + +#define amdgpu_dpm_enable_bapm(adev, e) \ + ((adev)->powerplay.pp_funcs->enable_bapm((adev)->powerplay.pp_handle, (e))) #define amdgpu_dpm_read_sensor(adev, idx, value, size) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->read_sensor(adev->powerplay.pp_handle, (idx), (value), (size)) : \ - (adev)->pm.funcs->read_sensor((adev), (idx), (value), (size))) + ((adev)->powerplay.pp_funcs->read_sensor((adev)->powerplay.pp_handle, (idx), (value), (size))) #define amdgpu_dpm_get_temperature(adev) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_temperature((adev)->powerplay.pp_handle) : \ - (adev)->pm.funcs->get_temperature((adev))) + ((adev)->powerplay.pp_funcs->get_temperature((adev)->powerplay.pp_handle)) #define amdgpu_dpm_set_fan_control_mode(adev, m) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->set_fan_control_mode((adev)->powerplay.pp_handle, (m)) : \ - (adev)->pm.funcs->set_fan_control_mode((adev), (m))) + ((adev)->powerplay.pp_funcs->set_fan_control_mode((adev)->powerplay.pp_handle, (m))) #define amdgpu_dpm_get_fan_control_mode(adev) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_fan_control_mode((adev)->powerplay.pp_handle) : \ - (adev)->pm.funcs->get_fan_control_mode((adev))) + ((adev)->powerplay.pp_funcs->get_fan_control_mode((adev)->powerplay.pp_handle)) #define amdgpu_dpm_set_fan_speed_percent(adev, s) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->set_fan_speed_percent((adev)->powerplay.pp_handle, (s)) : \ - (adev)->pm.funcs->set_fan_speed_percent((adev), (s))) + ((adev)->powerplay.pp_funcs->set_fan_speed_percent((adev)->powerplay.pp_handle, (s))) #define amdgpu_dpm_get_fan_speed_percent(adev, s) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_fan_speed_percent((adev)->powerplay.pp_handle, (s)) : \ - (adev)->pm.funcs->get_fan_speed_percent((adev), (s))) + ((adev)->powerplay.pp_funcs->get_fan_speed_percent((adev)->powerplay.pp_handle, (s))) #define amdgpu_dpm_get_fan_speed_rpm(adev, s) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_fan_speed_rpm((adev)->powerplay.pp_handle, (s)) : \ - -EINVAL) + ((adev)->powerplay.pp_funcs->get_fan_speed_rpm)((adev)->powerplay.pp_handle, (s)) #define amdgpu_dpm_get_sclk(adev, l) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_sclk((adev)->powerplay.pp_handle, (l)) : \ - (adev)->pm.funcs->get_sclk((adev), (l))) + ((adev)->powerplay.pp_funcs->get_sclk((adev)->powerplay.pp_handle, (l))) #define amdgpu_dpm_get_mclk(adev, l) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_mclk((adev)->powerplay.pp_handle, (l)) : \ - (adev)->pm.funcs->get_mclk((adev), (l))) - + ((adev)->powerplay.pp_funcs->get_mclk((adev)->powerplay.pp_handle, (l))) #define amdgpu_dpm_force_performance_level(adev, l) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->force_performance_level((adev)->powerplay.pp_handle, (l)) : \ - (adev)->pm.funcs->force_performance_level((adev), (l))) + ((adev)->powerplay.pp_funcs->force_performance_level((adev)->powerplay.pp_handle, (l))) #define amdgpu_dpm_powergate_uvd(adev, g) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->powergate_uvd((adev)->powerplay.pp_handle, (g)) : \ - (adev)->pm.funcs->powergate_uvd((adev), (g))) + ((adev)->powerplay.pp_funcs->powergate_uvd((adev)->powerplay.pp_handle, (g))) #define amdgpu_dpm_powergate_vce(adev, g) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->powergate_vce((adev)->powerplay.pp_handle, (g)) : \ - (adev)->pm.funcs->powergate_vce((adev), (g))) + ((adev)->powerplay.pp_funcs->powergate_vce((adev)->powerplay.pp_handle, (g))) #define amdgpu_dpm_get_current_power_state(adev) \ - (adev)->powerplay.pp_funcs->get_current_power_state((adev)->powerplay.pp_handle) + ((adev)->powerplay.pp_funcs->get_current_power_state((adev)->powerplay.pp_handle)) #define amdgpu_dpm_get_pp_num_states(adev, data) \ - (adev)->powerplay.pp_funcs->get_pp_num_states((adev)->powerplay.pp_handle, data) + ((adev)->powerplay.pp_funcs->get_pp_num_states((adev)->powerplay.pp_handle, data)) #define amdgpu_dpm_get_pp_table(adev, table) \ - (adev)->powerplay.pp_funcs->get_pp_table((adev)->powerplay.pp_handle, table) + ((adev)->powerplay.pp_funcs->get_pp_table((adev)->powerplay.pp_handle, table)) #define amdgpu_dpm_set_pp_table(adev, buf, size) \ - (adev)->powerplay.pp_funcs->set_pp_table((adev)->powerplay.pp_handle, buf, size) + ((adev)->powerplay.pp_funcs->set_pp_table((adev)->powerplay.pp_handle, buf, size)) #define amdgpu_dpm_print_clock_levels(adev, type, buf) \ - (adev)->powerplay.pp_funcs->print_clock_levels((adev)->powerplay.pp_handle, type, buf) + ((adev)->powerplay.pp_funcs->print_clock_levels((adev)->powerplay.pp_handle, type, buf)) #define amdgpu_dpm_force_clock_level(adev, type, level) \ - (adev)->powerplay.pp_funcs->force_clock_level((adev)->powerplay.pp_handle, type, level) + ((adev)->powerplay.pp_funcs->force_clock_level((adev)->powerplay.pp_handle, type, level)) #define amdgpu_dpm_get_sclk_od(adev) \ - (adev)->powerplay.pp_funcs->get_sclk_od((adev)->powerplay.pp_handle) + ((adev)->powerplay.pp_funcs->get_sclk_od((adev)->powerplay.pp_handle)) #define amdgpu_dpm_set_sclk_od(adev, value) \ - (adev)->powerplay.pp_funcs->set_sclk_od((adev)->powerplay.pp_handle, value) + ((adev)->powerplay.pp_funcs->set_sclk_od((adev)->powerplay.pp_handle, value)) #define amdgpu_dpm_get_mclk_od(adev) \ - ((adev)->powerplay.pp_funcs->get_mclk_od((adev)->powerplay.pp_handle)) + ((adev)->powerplay.pp_funcs->get_mclk_od((adev)->powerplay.pp_handle)) #define amdgpu_dpm_set_mclk_od(adev, value) \ - ((adev)->powerplay.pp_funcs->set_mclk_od((adev)->powerplay.pp_handle, value)) + ((adev)->powerplay.pp_funcs->set_mclk_od((adev)->powerplay.pp_handle, value)) -#define amdgpu_dpm_dispatch_task(adev, event_id, input, output) \ - (adev)->powerplay.pp_funcs->dispatch_tasks((adev)->powerplay.pp_handle, (event_id), (input), (output)) +#define amdgpu_dpm_dispatch_task(adev, task_id, input, output) \ + ((adev)->powerplay.pp_funcs->dispatch_tasks)((adev)->powerplay.pp_handle, (task_id), (input), (output)) -#define amgdpu_dpm_check_state_equal(adev, cps, rps, equal) (adev)->pm.funcs->check_state_equal((adev), (cps),(rps),(equal)) +#define amdgpu_dpm_check_state_equal(adev, cps, rps, equal) \ + ((adev)->powerplay.pp_funcs->check_state_equal((adev)->powerplay.pp_handle, (cps), (rps), (equal))) #define amdgpu_dpm_get_vce_clock_state(adev, i) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_vce_clock_state((adev)->powerplay.pp_handle, (i)) : \ - (adev)->pm.funcs->get_vce_clock_state((adev), (i))) + ((adev)->powerplay.pp_funcs->get_vce_clock_state((adev)->powerplay.pp_handle, (i))) -#define amdgpu_dpm_get_performance_level(adev) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_performance_level((adev)->powerplay.pp_handle) : \ - (adev)->pm.dpm.forced_level) +#define amdgpu_dpm_get_performance_level(adev) \ + ((adev)->powerplay.pp_funcs->get_performance_level((adev)->powerplay.pp_handle)) #define amdgpu_dpm_reset_power_profile_state(adev, request) \ - ((adev)->powerplay.pp_funcs->reset_power_profile_state(\ + ((adev)->powerplay.pp_funcs->reset_power_profile_state(\ (adev)->powerplay.pp_handle, request)) #define amdgpu_dpm_get_power_profile_state(adev, query) \ - ((adev)->powerplay.pp_funcs->get_power_profile_state(\ + ((adev)->powerplay.pp_funcs->get_power_profile_state(\ (adev)->powerplay.pp_handle, query)) #define amdgpu_dpm_set_power_profile_state(adev, request) \ - ((adev)->powerplay.pp_funcs->set_power_profile_state(\ + ((adev)->powerplay.pp_funcs->set_power_profile_state(\ (adev)->powerplay.pp_handle, request)) #define amdgpu_dpm_switch_power_profile(adev, type) \ - ((adev)->powerplay.pp_funcs->switch_power_profile(\ + ((adev)->powerplay.pp_funcs->switch_power_profile(\ (adev)->powerplay.pp_handle, type)) +#define amdgpu_dpm_set_clockgating_by_smu(adev, msg_id) \ + ((adev)->powerplay.pp_funcs->set_clockgating_by_smu(\ + (adev)->powerplay.pp_handle, msg_id)) + struct amdgpu_dpm { struct amdgpu_ps *ps; /* number of valid power states */ @@ -485,10 +431,9 @@ struct amdgpu_pm { struct amdgpu_dpm dpm; const struct firmware *fw; /* SMC firmware */ uint32_t fw_version; - const struct amdgpu_dpm_funcs *funcs; uint32_t pcie_gen_mask; uint32_t pcie_mlw_mask; - struct amd_pp_display_configuration pm_display_cfg;/* set by DAL */ + struct amd_pp_display_configuration pm_display_cfg;/* set by dc */ }; #define R600_SSTU_DFLT 0 @@ -551,6 +496,6 @@ u16 amdgpu_get_pcie_lane_support(struct amdgpu_device *adev, u8 amdgpu_encode_pci_lane_width(u32 lanes); struct amd_vce_state* -amdgpu_get_vce_clock_state(struct amdgpu_device *adev, unsigned idx); +amdgpu_get_vce_clock_state(void *handle, u32 idx); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index b59f37c83fa6..ec96bb1f9eaf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -68,13 +68,20 @@ * - 3.16.0 - Add reserved vmid support * - 3.17.0 - Add AMDGPU_NUM_VRAM_CPU_PAGE_FAULTS. * - 3.18.0 - Export gpu always on cu bitmap + * - 3.19.0 - Add support for UVD MJPEG decode + * - 3.20.0 - Add support for local BOs + * - 3.21.0 - Add DRM_AMDGPU_FENCE_TO_HANDLE ioctl + * - 3.22.0 - Add DRM_AMDGPU_SCHED ioctl + * - 3.23.0 - Add query for VRAM lost counter */ #define KMS_DRIVER_MAJOR 3 -#define KMS_DRIVER_MINOR 18 +#define KMS_DRIVER_MINOR 23 #define KMS_DRIVER_PATCHLEVEL 0 int amdgpu_vram_limit = 0; +int amdgpu_vis_vram_limit = 0; int amdgpu_gart_size = -1; /* auto */ +int amdgpu_gtt_size = -1; /* auto */ int amdgpu_moverate = -1; /* auto */ int amdgpu_benchmarking = 0; int amdgpu_testing = 0; @@ -88,27 +95,31 @@ int amdgpu_dpm = -1; int amdgpu_fw_load_type = -1; int amdgpu_aspm = -1; int amdgpu_runtime_pm = -1; -unsigned amdgpu_ip_block_mask = 0xffffffff; +uint amdgpu_ip_block_mask = 0xffffffff; int amdgpu_bapm = -1; int amdgpu_deep_color = 0; int amdgpu_vm_size = -1; +int amdgpu_vm_fragment_size = -1; int amdgpu_vm_block_size = -1; int amdgpu_vm_fault_stop = 0; int amdgpu_vm_debug = 0; int amdgpu_vram_page_split = 512; int amdgpu_vm_update_mode = -1; int amdgpu_exp_hw_support = 0; +int amdgpu_dc = -1; +int amdgpu_dc_log = 0; int amdgpu_sched_jobs = 32; int amdgpu_sched_hw_submission = 2; int amdgpu_no_evict = 0; int amdgpu_direct_gma_size = 0; -unsigned amdgpu_pcie_gen_cap = 0; -unsigned amdgpu_pcie_lane_cap = 0; -unsigned amdgpu_cg_mask = 0xffffffff; -unsigned amdgpu_pg_mask = 0xffffffff; +uint amdgpu_pcie_gen_cap = 0; +uint amdgpu_pcie_lane_cap = 0; +uint amdgpu_cg_mask = 0xffffffff; +uint amdgpu_pg_mask = 0xffffffff; +uint amdgpu_sdma_phase_quantum = 32; char *amdgpu_disable_cu = NULL; char *amdgpu_virtual_display = NULL; -unsigned amdgpu_pp_feature_mask = 0xffffffff; +uint amdgpu_pp_feature_mask = 0xffffffff; int amdgpu_ngg = 0; int amdgpu_prim_buf_per_se = 0; int amdgpu_pos_buf_per_se = 0; @@ -116,12 +127,19 @@ int amdgpu_cntl_sb_buf_per_se = 0; int amdgpu_param_buf_per_se = 0; int amdgpu_job_hang_limit = 0; int amdgpu_lbpw = -1; +int amdgpu_compute_multipipe = -1; MODULE_PARM_DESC(vramlimit, "Restrict VRAM for testing, in megabytes"); module_param_named(vramlimit, amdgpu_vram_limit, int, 0600); -MODULE_PARM_DESC(gartsize, "Size of PCIE/IGP gart to setup in megabytes (32, 64, etc., -1 = auto)"); -module_param_named(gartsize, amdgpu_gart_size, int, 0600); +MODULE_PARM_DESC(vis_vramlimit, "Restrict visible VRAM for testing, in megabytes"); +module_param_named(vis_vramlimit, amdgpu_vis_vram_limit, int, 0444); + +MODULE_PARM_DESC(gartsize, "Size of GART to setup in megabytes (32, 64, etc., -1=auto)"); +module_param_named(gartsize, amdgpu_gart_size, uint, 0600); + +MODULE_PARM_DESC(gttsize, "Size of the GTT domain in megabytes (-1 = auto)"); +module_param_named(gttsize, amdgpu_gtt_size, int, 0600); MODULE_PARM_DESC(moverate, "Maximum buffer migration rate in MB/s. (32, 64, etc., -1=auto, 0=1=disabled)"); module_param_named(moverate, amdgpu_moverate, int, 0600); @@ -174,6 +192,9 @@ module_param_named(deep_color, amdgpu_deep_color, int, 0444); MODULE_PARM_DESC(vm_size, "VM address space size in gigabytes (default 64GB)"); module_param_named(vm_size, amdgpu_vm_size, int, 0444); +MODULE_PARM_DESC(vm_fragment_size, "VM fragment size in bits (4, 5, etc. 4 = 64K (default), Max 9 = 2M)"); +module_param_named(vm_fragment_size, amdgpu_vm_fragment_size, int, 0444); + MODULE_PARM_DESC(vm_block_size, "VM page table size in bits (default depending on vm_size)"); module_param_named(vm_block_size, amdgpu_vm_block_size, int, 0444); @@ -186,12 +207,18 @@ module_param_named(vm_debug, amdgpu_vm_debug, int, 0644); MODULE_PARM_DESC(vm_update_mode, "VM update using CPU (0 = never (default except for large BAR(LB)), 1 = Graphics only, 2 = Compute only (default for LB), 3 = Both"); module_param_named(vm_update_mode, amdgpu_vm_update_mode, int, 0444); -MODULE_PARM_DESC(vram_page_split, "Number of pages after we split VRAM allocations (default 1024, -1 = disable)"); +MODULE_PARM_DESC(vram_page_split, "Number of pages after we split VRAM allocations (default 512, -1 = disable)"); module_param_named(vram_page_split, amdgpu_vram_page_split, int, 0444); MODULE_PARM_DESC(exp_hw_support, "experimental hw support (1 = enable, 0 = disable (default))"); module_param_named(exp_hw_support, amdgpu_exp_hw_support, int, 0444); +MODULE_PARM_DESC(dc, "Display Core driver (1 = enable, 0 = disable, -1 = auto (default))"); +module_param_named(dc, amdgpu_dc, int, 0444); + +MODULE_PARM_DESC(dc, "Display Core Log Level (0 = minimal (default), 1 = chatty"); +module_param_named(dc_log, amdgpu_dc_log, int, 0444); + MODULE_PARM_DESC(sched_jobs, "the max number of jobs supported in the sw queue (default 32)"); module_param_named(sched_jobs, amdgpu_sched_jobs, int, 0444); @@ -199,7 +226,7 @@ MODULE_PARM_DESC(sched_hw_submission, "the max number of HW submissions (default module_param_named(sched_hw_submission, amdgpu_sched_hw_submission, int, 0444); MODULE_PARM_DESC(ppfeaturemask, "all power features enabled (default))"); -module_param_named(ppfeaturemask, amdgpu_pp_feature_mask, int, 0444); +module_param_named(ppfeaturemask, amdgpu_pp_feature_mask, uint, 0444); MODULE_PARM_DESC(no_evict, "Support pinning request from user space (1 = enable, 0 = disable (default))"); module_param_named(no_evict, amdgpu_no_evict, int, 0444); @@ -219,6 +246,9 @@ module_param_named(cg_mask, amdgpu_cg_mask, uint, 0444); MODULE_PARM_DESC(pg_mask, "Powergating flags mask (0 = disable power gating)"); module_param_named(pg_mask, amdgpu_pg_mask, uint, 0444); +MODULE_PARM_DESC(sdma_phase_quantum, "SDMA context switch phase quantum (x 1K GPU clock cycles, 0 = no change (default 32))"); +module_param_named(sdma_phase_quantum, amdgpu_sdma_phase_quantum, uint, 0444); + MODULE_PARM_DESC(disable_cu, "Disable CUs (se.sh.cu,...)"); module_param_named(disable_cu, amdgpu_disable_cu, charp, 0444); @@ -247,6 +277,9 @@ module_param_named(job_hang_limit, amdgpu_job_hang_limit, int ,0444); MODULE_PARM_DESC(lbpw, "Load Balancing Per Watt (LBPW) support (1 = enable, 0 = disable, -1 = auto)"); module_param_named(lbpw, amdgpu_lbpw, int, 0444); +MODULE_PARM_DESC(compute_multipipe, "Force compute queues to be spread across pipes (1 = enable, 0 = disable, -1 = auto)"); +module_param_named(compute_multipipe, amdgpu_compute_multipipe, int, 0444); + #ifdef CONFIG_DRM_AMDGPU_SI #if defined(CONFIG_DRM_RADEON) || defined(CONFIG_DRM_RADEON_MODULE) @@ -493,15 +526,15 @@ static const struct pci_device_id pciidlist[] = { {0x1002, 0x6997, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12}, {0x1002, 0x699F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12}, /* Vega 10 */ - {0x1002, 0x6860, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x6861, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x6862, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x6863, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x6864, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x6867, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x6868, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x686c, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x687f, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10|AMD_EXP_HW_SUPPORT}, + {0x1002, 0x6860, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, + {0x1002, 0x6861, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, + {0x1002, 0x6862, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, + {0x1002, 0x6863, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, + {0x1002, 0x6864, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, + {0x1002, 0x6867, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, + {0x1002, 0x6868, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, + {0x1002, 0x686c, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, + {0x1002, 0x687f, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, /* Raven */ {0x1002, 0x15dd, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RAVEN|AMD_IS_APU|AMD_EXP_HW_SUPPORT}, @@ -591,6 +624,8 @@ amdgpu_pci_remove(struct pci_dev *pdev) drm_dev_unregister(dev); drm_dev_unref(dev); + pci_disable_device(pdev); + pci_set_drvdata(pdev, NULL); } static void @@ -803,7 +838,6 @@ static struct drm_driver kms_driver = { .open = amdgpu_driver_open_kms, .postclose = amdgpu_driver_postclose_kms, .lastclose = amdgpu_driver_lastclose_kms, - .set_busid = drm_pci_set_busid, .unload = amdgpu_driver_unload_kms, .get_vblank_counter = amdgpu_get_vblank_counter_kms, .enable_vblank = amdgpu_enable_vblank_kms, @@ -823,7 +857,6 @@ static struct drm_driver kms_driver = { .gem_close_object = amdgpu_gem_object_close, .dumb_create = amdgpu_mode_dumb_create, .dumb_map_offset = amdgpu_mode_dumb_mmap, - .dumb_destroy = drm_gem_dumb_destroy, .fops = &amdgpu_driver_kms_fops, .prime_handle_to_fd = drm_gem_prime_handle_to_fd, @@ -837,6 +870,7 @@ static struct drm_driver kms_driver = { .gem_prime_import_sg_table = amdgpu_gem_prime_import_sg_table, .gem_prime_vmap = amdgpu_gem_prime_vmap, .gem_prime_vunmap = amdgpu_gem_prime_vunmap, + .gem_prime_mmap = amdgpu_gem_prime_mmap, .name = DRIVER_NAME, .desc = DRIVER_DESC, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c index c0d8c6ff6380..90fa8e8bc6fb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c @@ -42,11 +42,6 @@ this contains a helper + a amdgpu fb the helper contains a pointer to amdgpu framebuffer baseclass. */ -struct amdgpu_fbdev { - struct drm_fb_helper helper; - struct amdgpu_framebuffer rfb; - struct amdgpu_device *adev; -}; static int amdgpufb_open(struct fb_info *info, int user) @@ -118,7 +113,7 @@ static void amdgpufb_destroy_pinned_object(struct drm_gem_object *gobj) amdgpu_bo_unpin(abo); amdgpu_bo_unreserve(abo); } - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); } static int amdgpufb_create_pinned_object(struct amdgpu_fbdev *rfbdev, @@ -149,7 +144,7 @@ static int amdgpufb_create_pinned_object(struct amdgpu_fbdev *rfbdev, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | AMDGPU_GEM_CREATE_VRAM_CLEARED, - true, &gobj); + true, NULL, &gobj); if (ret) { pr_err("failed to allocate framebuffer (%d)\n", aligned_size); return -ENOMEM; @@ -245,13 +240,12 @@ static int amdgpufb_create(struct drm_fb_helper *helper, drm_fb_helper_fill_fix(info, fb->pitches[0], fb->format->depth); - info->flags = FBINFO_DEFAULT | FBINFO_CAN_FORCE_OUTPUT; info->fbops = &amdgpufb_ops; tmp = amdgpu_bo_gpu_offset(abo) - adev->mc.vram_start; info->fix.smem_start = adev->mc.aper_base + tmp; info->fix.smem_len = amdgpu_bo_size(abo); - info->screen_base = abo->kptr; + info->screen_base = amdgpu_bo_kptr(abo); info->screen_size = amdgpu_bo_size(abo); drm_fb_helper_fill_var(info, &rfbdev->helper, sizes->fb_width, sizes->fb_height); @@ -281,7 +275,7 @@ out: } if (fb && ret) { - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); drm_framebuffer_unregister_private(fb); drm_framebuffer_cleanup(fb); kfree(fb); @@ -304,39 +298,15 @@ static int amdgpu_fbdev_destroy(struct drm_device *dev, struct amdgpu_fbdev *rfb if (rfb->obj) { amdgpufb_destroy_pinned_object(rfb->obj); rfb->obj = NULL; + drm_framebuffer_unregister_private(&rfb->base); + drm_framebuffer_cleanup(&rfb->base); } drm_fb_helper_fini(&rfbdev->helper); - drm_framebuffer_unregister_private(&rfb->base); - drm_framebuffer_cleanup(&rfb->base); return 0; } -/** Sets the color ramps on behalf of fbcon */ -static void amdgpu_crtc_fb_gamma_set(struct drm_crtc *crtc, u16 red, u16 green, - u16 blue, int regno) -{ - struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - - amdgpu_crtc->lut_r[regno] = red >> 6; - amdgpu_crtc->lut_g[regno] = green >> 6; - amdgpu_crtc->lut_b[regno] = blue >> 6; -} - -/** Gets the color ramps on behalf of fbcon */ -static void amdgpu_crtc_fb_gamma_get(struct drm_crtc *crtc, u16 *red, u16 *green, - u16 *blue, int regno) -{ - struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - - *red = amdgpu_crtc->lut_r[regno] << 6; - *green = amdgpu_crtc->lut_g[regno] << 6; - *blue = amdgpu_crtc->lut_b[regno] << 6; -} - static const struct drm_fb_helper_funcs amdgpu_fb_helper_funcs = { - .gamma_set = amdgpu_crtc_fb_gamma_set, - .gamma_get = amdgpu_crtc_fb_gamma_get, .fb_probe = amdgpufb_create, }; @@ -378,7 +348,8 @@ int amdgpu_fbdev_init(struct amdgpu_device *adev) drm_fb_helper_single_add_all_connectors(&rfbdev->helper); /* disable all the possible outputs/crtcs before entering KMS mode */ - drm_helper_disable_unused_functions(adev->ddev); + if (!amdgpu_device_has_dc_support(adev)) + drm_helper_disable_unused_functions(adev->ddev); drm_fb_helper_initial_config(&rfbdev->helper, bpp_sel); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index 333bad749067..2fa95aef74d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -169,6 +169,32 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f) } /** + * amdgpu_fence_emit_polling - emit a fence on the requeste ring + * + * @ring: ring the fence is associated with + * @s: resulting sequence number + * + * Emits a fence command on the requested ring (all asics). + * Used For polling fence. + * Returns 0 on success, -ENOMEM on failure. + */ +int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s) +{ + uint32_t seq; + + if (!s) + return -EINVAL; + + seq = ++ring->fence_drv.sync_seq; + amdgpu_ring_emit_fence(ring, ring->fence_drv.gpu_addr, + seq, AMDGPU_FENCE_FLAG_INT); + + *s = seq; + + return 0; +} + +/** * amdgpu_fence_schedule_fallback - schedule fallback check * * @ring: pointer to struct amdgpu_ring @@ -242,9 +268,10 @@ void amdgpu_fence_process(struct amdgpu_ring *ring) * * Checks for fence activity. */ -static void amdgpu_fence_fallback(unsigned long arg) +static void amdgpu_fence_fallback(struct timer_list *t) { - struct amdgpu_ring *ring = (void *)arg; + struct amdgpu_ring *ring = from_timer(ring, t, + fence_drv.fallback_timer); amdgpu_fence_process(ring); } @@ -260,7 +287,7 @@ static void amdgpu_fence_fallback(unsigned long arg) */ int amdgpu_fence_wait_empty(struct amdgpu_ring *ring) { - uint64_t seq = ACCESS_ONCE(ring->fence_drv.sync_seq); + uint64_t seq = READ_ONCE(ring->fence_drv.sync_seq); struct dma_fence *fence, **ptr; int r; @@ -282,6 +309,30 @@ int amdgpu_fence_wait_empty(struct amdgpu_ring *ring) } /** + * amdgpu_fence_wait_polling - busy wait for givn sequence number + * + * @ring: ring index the fence is associated with + * @wait_seq: sequence number to wait + * @timeout: the timeout for waiting in usecs + * + * Wait for all fences on the requested ring to signal (all asics). + * Returns left time if no timeout, 0 or minus if timeout. + */ +signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring, + uint32_t wait_seq, + signed long timeout) +{ + uint32_t seq; + + do { + seq = amdgpu_fence_read(ring); + udelay(5); + timeout -= 5; + } while ((int32_t)(wait_seq - seq) > 0 && timeout > 0); + + return timeout > 0 ? timeout : 0; +} +/** * amdgpu_fence_count_emitted - get the count of emitted fences * * @ring: ring the fence is associated with @@ -300,7 +351,7 @@ unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring) amdgpu_fence_process(ring); emitted = 0x100000000ull; emitted -= atomic_read(&ring->fence_drv.last_seq); - emitted += ACCESS_ONCE(ring->fence_drv.sync_seq); + emitted += READ_ONCE(ring->fence_drv.sync_seq); return lower_32_bits(emitted); } @@ -372,8 +423,7 @@ int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring, atomic_set(&ring->fence_drv.last_seq, 0); ring->fence_drv.initialized = false; - setup_timer(&ring->fence_drv.fallback_timer, amdgpu_fence_fallback, - (unsigned long)ring); + timer_setup(&ring->fence_drv.fallback_timer, amdgpu_fence_fallback, 0); ring->fence_drv.num_fences_mask = num_hw_submission * 2 - 1; spin_lock_init(&ring->fence_drv.lock); @@ -641,6 +691,19 @@ static int amdgpu_debugfs_fence_info(struct seq_file *m, void *data) atomic_read(&ring->fence_drv.last_seq)); seq_printf(m, "Last emitted 0x%08x\n", ring->fence_drv.sync_seq); + + if (ring->funcs->type != AMDGPU_RING_TYPE_GFX) + continue; + + /* set in CP_VMID_PREEMPT and preemption occurred */ + seq_printf(m, "Last preempted 0x%08x\n", + le32_to_cpu(*(ring->fence_drv.cpu_addr + 2))); + /* set in CP_VMID_RESET and reset occurred */ + seq_printf(m, "Last reset 0x%08x\n", + le32_to_cpu(*(ring->fence_drv.cpu_addr + 4))); + /* Both preemption and reset occurred */ + seq_printf(m, "Last both 0x%08x\n", + le32_to_cpu(*(ring->fence_drv.cpu_addr + 6))); } return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index a57abc1a25fb..fe818501c520 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -55,6 +55,7 @@ /* * Common GART table functions. */ + /** * amdgpu_gart_table_ram_alloc - allocate system ram for gart page table * @@ -131,7 +132,7 @@ int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev) PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, &adev->gart.robj); + NULL, NULL, 0, &adev->gart.robj); if (r) { return r; } @@ -263,6 +264,41 @@ int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, } /** + * amdgpu_gart_map - map dma_addresses into GART entries + * + * @adev: amdgpu_device pointer + * @offset: offset into the GPU's gart aperture + * @pages: number of pages to bind + * @dma_addr: DMA addresses of pages + * + * Map the dma_addresses into GART entries (all asics). + * Returns 0 for success, -EINVAL for failure. + */ +int amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset, + int pages, dma_addr_t *dma_addr, uint64_t flags, + void *dst) +{ + uint64_t page_base; + unsigned i, j, t; + + if (!adev->gart.ready) { + WARN(1, "trying to bind memory to uninitialized GART !\n"); + return -EINVAL; + } + + t = offset / AMDGPU_GPU_PAGE_SIZE; + + for (i = 0; i < pages; i++) { + page_base = dma_addr[i]; + for (j = 0; j < (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); j++, t++) { + amdgpu_gart_set_pte_pde(adev, dst, t, page_base, flags); + page_base += AMDGPU_GPU_PAGE_SIZE; + } + } + return 0; +} + +/** * amdgpu_gart_bind - bind pages into the gart page table * * @adev: amdgpu_device pointer @@ -279,31 +315,31 @@ int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset, int pages, struct page **pagelist, dma_addr_t *dma_addr, uint64_t flags) { - unsigned t; - unsigned p; - uint64_t page_base; - int i, j; +#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS + unsigned i,t,p; +#endif + int r; if (!adev->gart.ready) { WARN(1, "trying to bind memory to uninitialized GART !\n"); return -EINVAL; } +#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS t = offset / AMDGPU_GPU_PAGE_SIZE; p = t / (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); - - for (i = 0; i < pages; i++, p++) { -#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS + for (i = 0; i < pages; i++, p++) adev->gart.pages[p] = pagelist[i]; #endif - if (adev->gart.ptr) { - page_base = dma_addr[i]; - for (j = 0; j < (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); j++, t++) { - amdgpu_gart_set_pte_pde(adev, adev->gart.ptr, t, page_base, flags); - page_base += AMDGPU_GPU_PAGE_SIZE; - } - } - } + + if (!adev->gart.ptr) + return 0; + + r = amdgpu_gart_map(adev, offset, pages, dma_addr, flags, + adev->gart.ptr); + if (r) + return r; + mb(); amdgpu_gart_flush_gpu_tlb(adev, 0); return 0; @@ -333,8 +369,8 @@ int amdgpu_gart_init(struct amdgpu_device *adev) if (r) return r; /* Compute table size */ - adev->gart.num_cpu_pages = adev->mc.gtt_size / PAGE_SIZE; - adev->gart.num_gpu_pages = adev->mc.gtt_size / AMDGPU_GPU_PAGE_SIZE; + adev->gart.num_cpu_pages = adev->mc.gart_size / PAGE_SIZE; + adev->gart.num_gpu_pages = adev->mc.gart_size / AMDGPU_GPU_PAGE_SIZE; DRM_INFO("GART: num cpu pages %u, num gpu pages %u\n", adev->gart.num_cpu_pages, adev->gart.num_gpu_pages); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h new file mode 100644 index 000000000000..afbe803b1a13 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h @@ -0,0 +1,76 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __AMDGPU_GART_H__ +#define __AMDGPU_GART_H__ + +#include <linux/types.h> + +/* + * GART structures, functions & helpers + */ +struct amdgpu_device; +struct amdgpu_bo; +struct amdgpu_gart_funcs; + +#define AMDGPU_GPU_PAGE_SIZE 4096 +#define AMDGPU_GPU_PAGE_MASK (AMDGPU_GPU_PAGE_SIZE - 1) +#define AMDGPU_GPU_PAGE_SHIFT 12 +#define AMDGPU_GPU_PAGE_ALIGN(a) (((a) + AMDGPU_GPU_PAGE_MASK) & ~AMDGPU_GPU_PAGE_MASK) + +struct amdgpu_gart { + dma_addr_t table_addr; + struct amdgpu_bo *robj; + void *ptr; + unsigned num_gpu_pages; + unsigned num_cpu_pages; + unsigned table_size; +#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS + struct page **pages; +#endif + bool ready; + + /* Asic default pte flags */ + uint64_t gart_pte_flags; + + const struct amdgpu_gart_funcs *gart_funcs; +}; + +int amdgpu_gart_table_ram_alloc(struct amdgpu_device *adev); +void amdgpu_gart_table_ram_free(struct amdgpu_device *adev); +int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev); +void amdgpu_gart_table_vram_free(struct amdgpu_device *adev); +int amdgpu_gart_table_vram_pin(struct amdgpu_device *adev); +void amdgpu_gart_table_vram_unpin(struct amdgpu_device *adev); +int amdgpu_gart_init(struct amdgpu_device *adev); +void amdgpu_gart_fini(struct amdgpu_device *adev); +int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, + int pages); +int amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset, + int pages, dma_addr_t *dma_addr, uint64_t flags, + void *dst); +int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset, + int pages, struct page **pagelist, + dma_addr_t *dma_addr, uint64_t flags); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 621f739103a6..e87eedcc0da9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -44,12 +44,12 @@ void amdgpu_gem_object_free(struct drm_gem_object *gobj) } int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, - int alignment, u32 initial_domain, - u64 flags, bool kernel, - struct drm_gem_object **obj) + int alignment, u32 initial_domain, + u64 flags, bool kernel, + struct reservation_object *resv, + struct drm_gem_object **obj) { - struct amdgpu_bo *robj; - unsigned long max_size; + struct amdgpu_bo *bo; int r; *obj = NULL; @@ -58,22 +58,16 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, alignment = PAGE_SIZE; } - if (!(initial_domain & (AMDGPU_GEM_DOMAIN_GDS | AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA))) { - /* Maximum bo size is the unpinned gtt size since we use the gtt to - * handle vram to system pool migrations. - */ - max_size = adev->mc.gtt_size - adev->gart_pin_size; - if (size > max_size) { - DRM_DEBUG("Allocation size %ldMb bigger than %ldMb limit\n", - size >> 20, max_size >> 20); - return -ENOMEM; - } - } retry: r = amdgpu_bo_create(adev, size, alignment, kernel, initial_domain, - flags, NULL, NULL, &robj); + flags, NULL, resv, 0, &bo); if (r) { if (r != -ERESTARTSYS) { + if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) { + flags &= ~AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; + goto retry; + } + if (initial_domain == AMDGPU_GEM_DOMAIN_VRAM) { initial_domain |= AMDGPU_GEM_DOMAIN_GTT; goto retry; @@ -83,7 +77,7 @@ retry: } return r; } - *obj = &robj->gem_base; + *obj = &bo->gem_base; return 0; } @@ -103,7 +97,7 @@ void amdgpu_gem_force_release(struct amdgpu_device *adev) spin_lock(&file->table_lock); idr_for_each_entry(&file->object_idr, gobj, handle) { WARN_ONCE(1, "And also active allocations!\n"); - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); } idr_destroy(&file->object_idr); spin_unlock(&file->table_lock); @@ -124,7 +118,17 @@ int amdgpu_gem_object_open(struct drm_gem_object *obj, struct amdgpu_fpriv *fpriv = file_priv->driver_priv; struct amdgpu_vm *vm = &fpriv->vm; struct amdgpu_bo_va *bo_va; + struct mm_struct *mm; int r; + + mm = amdgpu_ttm_tt_get_usermm(abo->tbo.ttm); + if (mm && mm != current->mm) + return -EPERM; + + if (abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID && + abo->tbo.resv != vm->root.base.bo->tbo.resv) + return -EPERM; + r = amdgpu_bo_reserve(abo, false); if (r) return r; @@ -139,35 +143,6 @@ int amdgpu_gem_object_open(struct drm_gem_object *obj, return 0; } -static int amdgpu_gem_vm_check(void *param, struct amdgpu_bo *bo) -{ - /* if anything is swapped out don't swap it in here, - just abort and wait for the next CS */ - if (!amdgpu_bo_gpu_accessible(bo)) - return -ERESTARTSYS; - - if (bo->shadow && !amdgpu_bo_gpu_accessible(bo->shadow)) - return -ERESTARTSYS; - - return 0; -} - -static bool amdgpu_gem_vm_ready(struct amdgpu_device *adev, - struct amdgpu_vm *vm, - struct list_head *list) -{ - struct ttm_validate_buffer *entry; - - list_for_each_entry(entry, list, head) { - struct amdgpu_bo *bo = - container_of(entry->bo, struct amdgpu_bo, tbo); - if (amdgpu_gem_vm_check(NULL, bo)) - return false; - } - - return !amdgpu_vm_validate_pt_bos(adev, vm, amdgpu_gem_vm_check, NULL); -} - void amdgpu_gem_object_close(struct drm_gem_object *obj, struct drm_file *file_priv) { @@ -177,13 +152,14 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj, struct amdgpu_vm *vm = &fpriv->vm; struct amdgpu_bo_list_entry vm_pd; - struct list_head list; + struct list_head list, duplicates; struct ttm_validate_buffer tv; struct ww_acquire_ctx ticket; struct amdgpu_bo_va *bo_va; int r; INIT_LIST_HEAD(&list); + INIT_LIST_HEAD(&duplicates); tv.bo = &bo->tbo; tv.shared = true; @@ -191,7 +167,7 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj, amdgpu_vm_get_pd_bo(vm, &list, &vm_pd); - r = ttm_eu_reserve_buffers(&ticket, &list, false, NULL); + r = ttm_eu_reserve_buffers(&ticket, &list, false, &duplicates); if (r) { dev_err(adev->dev, "leaking bo va because " "we fail to reserve bo (%d)\n", r); @@ -201,7 +177,7 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj, if (bo_va && --bo_va->ref_count == 0) { amdgpu_vm_bo_rmv(adev, bo_va); - if (amdgpu_gem_vm_ready(adev, vm, &list)) { + if (amdgpu_vm_ready(vm)) { struct dma_fence *fence = NULL; r = amdgpu_vm_clear_freed(adev, vm, &fence); @@ -226,20 +202,24 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { struct amdgpu_device *adev = dev->dev_private; + struct amdgpu_fpriv *fpriv = filp->driver_priv; + struct amdgpu_vm *vm = &fpriv->vm; union drm_amdgpu_gem_create *args = data; + uint64_t flags = args->in.domain_flags; uint64_t size = args->in.bo_size; + struct reservation_object *resv = NULL; struct drm_gem_object *gobj; uint32_t handle; - bool kernel = false; int r; /* reject invalid gem flags */ - if (args->in.domain_flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_NO_CPU_ACCESS | - AMDGPU_GEM_CREATE_CPU_GTT_USWC | - AMDGPU_GEM_CREATE_VRAM_CLEARED| - AMDGPU_GEM_CREATE_SHADOW | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)) + if (flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | + AMDGPU_GEM_CREATE_NO_CPU_ACCESS | + AMDGPU_GEM_CREATE_CPU_GTT_USWC | + AMDGPU_GEM_CREATE_VRAM_CLEARED | + AMDGPU_GEM_CREATE_VM_ALWAYS_VALID | + AMDGPU_GEM_CREATE_EXPLICIT_SYNC)) + return -EINVAL; /* reject invalid gem domains */ @@ -254,7 +234,7 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, /* create a gem object to contain this object in */ if (args->in.domains & (AMDGPU_GEM_DOMAIN_GDS | AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA)) { - kernel = true; + flags |= AMDGPU_GEM_CREATE_NO_CPU_ACCESS; if (args->in.domains == AMDGPU_GEM_DOMAIN_GDS) size = size << AMDGPU_GDS_SHIFT; else if (args->in.domains == AMDGPU_GEM_DOMAIN_GWS) @@ -266,16 +246,31 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, } size = roundup(size, PAGE_SIZE); + if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) { + r = amdgpu_bo_reserve(vm->root.base.bo, false); + if (r) + return r; + + resv = vm->root.base.bo->tbo.resv; + } + r = amdgpu_gem_object_create(adev, size, args->in.alignment, (u32)(0xffffffff & args->in.domains), - args->in.domain_flags, - kernel, &gobj); + flags, false, resv, &gobj); + if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) { + if (!r) { + struct amdgpu_bo *abo = gem_to_amdgpu_bo(gobj); + + abo->parent = amdgpu_bo_ref(vm->root.base.bo); + } + amdgpu_bo_unreserve(vm->root.base.bo); + } if (r) return r; r = drm_gem_handle_create(filp, gobj, &handle); /* drop reference from allocate - handle holds it now */ - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); if (r) return r; @@ -311,14 +306,13 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, } /* create a gem object to contain this object in */ - r = amdgpu_gem_object_create(adev, args->size, 0, - AMDGPU_GEM_DOMAIN_CPU, 0, - 0, &gobj); + r = amdgpu_gem_object_create(adev, args->size, 0, AMDGPU_GEM_DOMAIN_CPU, + 0, 0, NULL, &gobj); if (r) return r; bo = gem_to_amdgpu_bo(gobj); - bo->prefered_domains = AMDGPU_GEM_DOMAIN_GTT; + bo->preferred_domains = AMDGPU_GEM_DOMAIN_GTT; bo->allowed_domains = AMDGPU_GEM_DOMAIN_GTT; r = amdgpu_ttm_tt_set_userptr(bo->tbo.ttm, args->addr, args->flags); if (r) @@ -331,12 +325,10 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, } if (args->flags & AMDGPU_GEM_USERPTR_VALIDATE) { - down_read(¤t->mm->mmap_sem); - r = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm, bo->tbo.ttm->pages); if (r) - goto unlock_mmap_sem; + goto release_object; r = amdgpu_bo_reserve(bo, true); if (r) @@ -347,13 +339,11 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, amdgpu_bo_unreserve(bo); if (r) goto free_pages; - - up_read(¤t->mm->mmap_sem); } r = drm_gem_handle_create(filp, gobj, &handle); /* drop reference from allocate - handle holds it now */ - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); if (r) return r; @@ -361,13 +351,10 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, return 0; free_pages: - release_pages(bo->tbo.ttm->pages, bo->tbo.ttm->num_pages, false); - -unlock_mmap_sem: - up_read(¤t->mm->mmap_sem); + release_pages(bo->tbo.ttm->pages, bo->tbo.ttm->num_pages); release_object: - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); return r; } @@ -386,11 +373,11 @@ int amdgpu_mode_dumb_mmap(struct drm_file *filp, robj = gem_to_amdgpu_bo(gobj); if (amdgpu_ttm_tt_get_usermm(robj->tbo.ttm) || (robj->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) { - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); return -EPERM; } *offset_p = amdgpu_bo_mmap_offset(robj); - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); return 0; } @@ -460,7 +447,7 @@ int amdgpu_gem_wait_idle_ioctl(struct drm_device *dev, void *data, } else r = ret; - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); return r; } @@ -503,7 +490,7 @@ int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data, unreserve: amdgpu_bo_unreserve(robj); out: - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); return r; } @@ -525,10 +512,10 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev, struct list_head *list, uint32_t operation) { - int r = -ERESTARTSYS; + int r; - if (!amdgpu_gem_vm_ready(adev, vm, list)) - goto error; + if (!amdgpu_vm_ready(vm)) + return; r = amdgpu_vm_update_directories(adev, vm); if (r) @@ -565,15 +552,14 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, struct amdgpu_bo_list_entry vm_pd; struct ttm_validate_buffer tv; struct ww_acquire_ctx ticket; - struct list_head list; + struct list_head list, duplicates; uint64_t va_flags; int r = 0; if (args->va_address < AMDGPU_VA_RESERVED_SIZE) { dev_err(&dev->pdev->dev, - "va_address 0x%lX is in reserved area 0x%X\n", - (unsigned long)args->va_address, - AMDGPU_VA_RESERVED_SIZE); + "va_address 0x%LX is in reserved area 0x%LX\n", + args->va_address, AMDGPU_VA_RESERVED_SIZE); return -EINVAL; } @@ -594,13 +580,9 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, args->operation); return -EINVAL; } - if ((args->operation == AMDGPU_VA_OP_MAP) || - (args->operation == AMDGPU_VA_OP_REPLACE)) { - if (amdgpu_kms_vram_lost(adev, fpriv)) - return -ENODEV; - } INIT_LIST_HEAD(&list); + INIT_LIST_HEAD(&duplicates); if ((args->operation != AMDGPU_VA_OP_CLEAR) && !(args->flags & AMDGPU_VM_PAGE_PRT)) { gobj = drm_gem_object_lookup(filp, args->handle); @@ -617,7 +599,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, amdgpu_vm_get_pd_bo(&fpriv->vm, &list, &vm_pd); - r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL); + r = ttm_eu_reserve_buffers(&ticket, &list, true, &duplicates); if (r) goto error_unref; @@ -635,7 +617,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, switch (args->operation) { case AMDGPU_VA_OP_MAP: - r = amdgpu_vm_alloc_pts(adev, bo_va->vm, args->va_address, + r = amdgpu_vm_alloc_pts(adev, bo_va->base.vm, args->va_address, args->map_size); if (r) goto error_backoff; @@ -655,7 +637,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, args->map_size); break; case AMDGPU_VA_OP_REPLACE: - r = amdgpu_vm_alloc_pts(adev, bo_va->vm, args->va_address, + r = amdgpu_vm_alloc_pts(adev, bo_va->base.vm, args->va_address, args->map_size); if (r) goto error_backoff; @@ -676,13 +658,14 @@ error_backoff: ttm_eu_backoff_reservation(&ticket, &list); error_unref: - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); return r; } int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { + struct amdgpu_device *adev = dev->dev_private; struct drm_amdgpu_gem_op *args = data; struct drm_gem_object *gobj; struct amdgpu_bo *robj; @@ -701,11 +684,11 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, switch (args->op) { case AMDGPU_GEM_OP_GET_GEM_CREATE_INFO: { struct drm_amdgpu_gem_create_in info; - void __user *out = (void __user *)(uintptr_t)args->value; + void __user *out = u64_to_user_ptr(args->value); info.bo_size = robj->gem_base.size; info.alignment = robj->tbo.mem.page_alignment << PAGE_SHIFT; - info.domains = robj->prefered_domains; + info.domains = robj->preferred_domains; info.domain_flags = robj->flags; amdgpu_bo_unreserve(robj); if (copy_to_user(out, &info, sizeof(info))) @@ -723,13 +706,16 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, amdgpu_bo_unreserve(robj); break; } - robj->prefered_domains = args->value & (AMDGPU_GEM_DOMAIN_VRAM | + robj->preferred_domains = args->value & (AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT | AMDGPU_GEM_DOMAIN_CPU); - robj->allowed_domains = robj->prefered_domains; + robj->allowed_domains = robj->preferred_domains; if (robj->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM) robj->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT; + if (robj->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) + amdgpu_vm_bo_invalidate(adev, robj, true); + amdgpu_bo_unreserve(robj); break; default: @@ -738,7 +724,7 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, } out: - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); return r; } @@ -759,14 +745,13 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv, r = amdgpu_gem_object_create(adev, args->size, 0, AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED, - ttm_bo_type_device, - &gobj); + false, NULL, &gobj); if (r) return -ENOMEM; r = drm_gem_handle_create(file_priv, gobj, &handle); /* drop reference from allocate - handle holds it now */ - drm_gem_object_unreference_unlocked(gobj); + drm_gem_object_put_unlocked(gobj); if (r) { return r; } @@ -784,6 +769,7 @@ static int amdgpu_debugfs_gem_bo_info(int id, void *ptr, void *data) unsigned domain; const char *placement; unsigned pin_count; + uint64_t offset; domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type); switch (domain) { @@ -798,11 +784,14 @@ static int amdgpu_debugfs_gem_bo_info(int id, void *ptr, void *data) placement = " CPU"; break; } - seq_printf(m, "\t0x%08x: %12ld byte %s @ 0x%010Lx", - id, amdgpu_bo_size(bo), placement, - amdgpu_bo_gpu_offset(bo)); + seq_printf(m, "\t0x%08x: %12ld byte %s", + id, amdgpu_bo_size(bo), placement); + + offset = READ_ONCE(bo->tbo.mem.start); + if (offset != AMDGPU_BO_INVALID_OFFSET) + seq_printf(m, " @ 0x%010Lx", offset); - pin_count = ACCESS_ONCE(bo->pin_count); + pin_count = READ_ONCE(bo->pin_count); if (pin_count) seq_printf(m, " pin count %d", pin_count); seq_printf(m, "\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index e26108aad3fe..ef043361009f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -109,9 +109,26 @@ void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, unsigned max_s } } +static bool amdgpu_gfx_is_multipipe_capable(struct amdgpu_device *adev) +{ + if (amdgpu_compute_multipipe != -1) { + DRM_INFO("amdgpu: forcing compute pipe policy %d\n", + amdgpu_compute_multipipe); + return amdgpu_compute_multipipe == 1; + } + + /* FIXME: spreading the queues across pipes causes perf regressions + * on POLARIS11 compute workloads */ + if (adev->asic_type == CHIP_POLARIS11) + return false; + + return adev->gfx.mec.num_mec > 1; +} + void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev) { int i, queue, pipe, mec; + bool multipipe_policy = amdgpu_gfx_is_multipipe_capable(adev); /* policy for amdgpu compute queue ownership */ for (i = 0; i < AMDGPU_MAX_COMPUTE_QUEUES; ++i) { @@ -125,7 +142,7 @@ void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev) if (mec >= adev->gfx.mec.num_mec) break; - if (adev->gfx.mec.num_mec > 1) { + if (multipipe_policy) { /* policy: amdgpu owns the first two queues of the first MEC */ if (mec == 0 && queue < 2) set_bit(i, adev->gfx.mec.queue_bitmap); @@ -184,7 +201,7 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, struct amdgpu_kiq *kiq = &adev->gfx.kiq; int r = 0; - mutex_init(&kiq->ring_mutex); + spin_lock_init(&kiq->ring_lock); r = amdgpu_wb_get(adev, &adev->virt.reg_val_offs); if (r) @@ -259,8 +276,13 @@ int amdgpu_gfx_compute_mqd_sw_init(struct amdgpu_device *adev, /* create MQD for KIQ */ ring = &adev->gfx.kiq.ring; if (!ring->mqd_obj) { + /* originaly the KIQ MQD is put in GTT domain, but for SRIOV VRAM domain is a must + * otherwise hypervisor trigger SAVE_VF fail after driver unloaded which mean MQD + * deallocated and gart_unbind, to strict diverage we decide to use VRAM domain for + * KIQ MQD no matter SRIOV or Bare-metal + */ r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj, + AMDGPU_GEM_DOMAIN_VRAM, &ring->mqd_obj, &ring->mqd_gpu_addr, &ring->mqd_ptr); if (r) { dev_warn(adev->dev, "failed to create ring mqd ob (%d)", r); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c index f7d22c44034d..00e0ce10862f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c @@ -28,7 +28,7 @@ struct amdgpu_gtt_mgr { struct drm_mm mm; spinlock_t lock; - uint64_t available; + atomic64_t available; }; /** @@ -42,15 +42,19 @@ struct amdgpu_gtt_mgr { static int amdgpu_gtt_mgr_init(struct ttm_mem_type_manager *man, unsigned long p_size) { + struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev); struct amdgpu_gtt_mgr *mgr; + uint64_t start, size; mgr = kzalloc(sizeof(*mgr), GFP_KERNEL); if (!mgr) return -ENOMEM; - drm_mm_init(&mgr->mm, 0, p_size); + start = AMDGPU_GTT_MAX_TRANSFER_SIZE * AMDGPU_GTT_NUM_TRANSFER_WINDOWS; + size = (adev->mc.gart_size >> PAGE_SHIFT) - start; + drm_mm_init(&mgr->mm, start, size); spin_lock_init(&mgr->lock); - mgr->available = p_size; + atomic64_set(&mgr->available, p_size); man->priv = mgr; return 0; } @@ -67,12 +71,6 @@ static int amdgpu_gtt_mgr_fini(struct ttm_mem_type_manager *man) { struct amdgpu_gtt_mgr *mgr = man->priv; - spin_lock(&mgr->lock); - if (!drm_mm_clean(&mgr->mm)) { - spin_unlock(&mgr->lock); - return -EBUSY; - } - drm_mm_takedown(&mgr->mm); spin_unlock(&mgr->lock); kfree(mgr); @@ -81,6 +79,20 @@ static int amdgpu_gtt_mgr_fini(struct ttm_mem_type_manager *man) } /** + * amdgpu_gtt_mgr_is_allocated - Check if mem has address space + * + * @mem: the mem object to check + * + * Check if a mem object has already address space allocated. + */ +bool amdgpu_gtt_mgr_is_allocated(struct ttm_mem_reg *mem) +{ + struct drm_mm_node *node = mem->mm_node; + + return (node->start != AMDGPU_BO_INVALID_OFFSET); +} + +/** * amdgpu_gtt_mgr_alloc - allocate new ranges * * @man: TTM memory type manager @@ -90,18 +102,19 @@ static int amdgpu_gtt_mgr_fini(struct ttm_mem_type_manager *man) * * Allocate the address space for a node. */ -int amdgpu_gtt_mgr_alloc(struct ttm_mem_type_manager *man, - struct ttm_buffer_object *tbo, - const struct ttm_place *place, - struct ttm_mem_reg *mem) +static int amdgpu_gtt_mgr_alloc(struct ttm_mem_type_manager *man, + struct ttm_buffer_object *tbo, + const struct ttm_place *place, + struct ttm_mem_reg *mem) { + struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev); struct amdgpu_gtt_mgr *mgr = man->priv; struct drm_mm_node *node = mem->mm_node; enum drm_mm_insert_mode mode; unsigned long fpfn, lpfn; int r; - if (node->start != AMDGPU_BO_INVALID_OFFSET) + if (amdgpu_gtt_mgr_is_allocated(mem)) return 0; if (place) @@ -112,7 +125,7 @@ int amdgpu_gtt_mgr_alloc(struct ttm_mem_type_manager *man, if (place && place->lpfn) lpfn = place->lpfn; else - lpfn = man->size; + lpfn = adev->gart.num_cpu_pages; mode = DRM_MM_INSERT_BEST; if (place && place->flags & TTM_PL_FLAG_TOPDOWN) @@ -124,25 +137,12 @@ int amdgpu_gtt_mgr_alloc(struct ttm_mem_type_manager *man, fpfn, lpfn, mode); spin_unlock(&mgr->lock); - if (!r) { + if (!r) mem->start = node->start; - if (&tbo->mem == mem) - tbo->offset = (tbo->mem.start << PAGE_SHIFT) + - tbo->bdev->man[tbo->mem.mem_type].gpu_offset; - } return r; } -void amdgpu_gtt_mgr_print(struct seq_file *m, struct ttm_mem_type_manager *man) -{ - struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev); - struct amdgpu_gtt_mgr *mgr = man->priv; - - seq_printf(m, "man size:%llu pages, gtt available:%llu pages, usage:%lluMB\n", - man->size, mgr->available, (u64)atomic64_read(&adev->gtt_usage) >> 20); - -} /** * amdgpu_gtt_mgr_new - allocate a new node * @@ -163,11 +163,12 @@ static int amdgpu_gtt_mgr_new(struct ttm_mem_type_manager *man, int r; spin_lock(&mgr->lock); - if (mgr->available < mem->num_pages) { + if ((&tbo->mem == mem || tbo->mem.mem_type != TTM_PL_TT) && + atomic64_read(&mgr->available) < mem->num_pages) { spin_unlock(&mgr->lock); return 0; } - mgr->available -= mem->num_pages; + atomic64_sub(mem->num_pages, &mgr->available); spin_unlock(&mgr->lock); node = kzalloc(sizeof(*node), GFP_KERNEL); @@ -194,9 +195,7 @@ static int amdgpu_gtt_mgr_new(struct ttm_mem_type_manager *man, return 0; err_out: - spin_lock(&mgr->lock); - mgr->available += mem->num_pages; - spin_unlock(&mgr->lock); + atomic64_add(mem->num_pages, &mgr->available); return r; } @@ -223,30 +222,48 @@ static void amdgpu_gtt_mgr_del(struct ttm_mem_type_manager *man, spin_lock(&mgr->lock); if (node->start != AMDGPU_BO_INVALID_OFFSET) drm_mm_remove_node(node); - mgr->available += mem->num_pages; spin_unlock(&mgr->lock); + atomic64_add(mem->num_pages, &mgr->available); kfree(node); mem->mm_node = NULL; } /** + * amdgpu_gtt_mgr_usage - return usage of GTT domain + * + * @man: TTM memory type manager + * + * Return how many bytes are used in the GTT domain + */ +uint64_t amdgpu_gtt_mgr_usage(struct ttm_mem_type_manager *man) +{ + struct amdgpu_gtt_mgr *mgr = man->priv; + s64 result = man->size - atomic64_read(&mgr->available); + + return (result > 0 ? result : 0) * PAGE_SIZE; +} + +/** * amdgpu_gtt_mgr_debug - dump VRAM table * * @man: TTM memory type manager - * @prefix: text prefix + * @printer: DRM printer to use * * Dump the table content using printk. */ static void amdgpu_gtt_mgr_debug(struct ttm_mem_type_manager *man, - const char *prefix) + struct drm_printer *printer) { struct amdgpu_gtt_mgr *mgr = man->priv; - struct drm_printer p = drm_debug_printer(prefix); spin_lock(&mgr->lock); - drm_mm_print(&mgr->mm, &p); + drm_mm_print(&mgr->mm, printer); spin_unlock(&mgr->lock); + + drm_printf(printer, "man size:%llu pages, gtt available:%lld pages, usage:%lluMB\n", + man->size, (u64)atomic64_read(&mgr->available), + amdgpu_gtt_mgr_usage(man) >> 20); } const struct ttm_mem_type_manager_func amdgpu_gtt_mgr_func = { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index f774b3f497d2..659997bfff30 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -130,6 +130,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, unsigned i; int r = 0; + bool need_pipe_sync = false; if (num_ibs == 0) return -EINVAL; @@ -165,15 +166,15 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, if (ring->funcs->emit_pipeline_sync && job && ((tmp = amdgpu_sync_get_fence(&job->sched_sync)) || amdgpu_vm_need_pipeline_sync(ring, job))) { - amdgpu_ring_emit_pipeline_sync(ring); + need_pipe_sync = true; dma_fence_put(tmp); } if (ring->funcs->insert_start) ring->funcs->insert_start(ring); - if (vm) { - r = amdgpu_vm_flush(ring, job); + if (job) { + r = amdgpu_vm_flush(ring, job, need_pipe_sync); if (r) { amdgpu_ring_undo(ring); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c index 3ab4c65ecc8b..f5f27e4f0f7f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c @@ -169,6 +169,12 @@ restart_ih: while (adev->irq.ih.rptr != wptr) { u32 ring_index = adev->irq.ih.rptr >> 2; + /* Prescreening of high-frequency interrupts */ + if (!amdgpu_ih_prescreen_iv(adev)) { + adev->irq.ih.rptr &= adev->irq.ih.ptr_mask; + continue; + } + /* Before dispatching irq to IP blocks, send it to amdkfd */ amdgpu_amdkfd_interrupt(adev, (const void *) &adev->irq.ih.ring[ring_index]); @@ -190,3 +196,79 @@ restart_ih: return IRQ_HANDLED; } + +/** + * amdgpu_ih_add_fault - Add a page fault record + * + * @adev: amdgpu device pointer + * @key: 64-bit encoding of PASID and address + * + * This should be called when a retry page fault interrupt is + * received. If this is a new page fault, it will be added to a hash + * table. The return value indicates whether this is a new fault, or + * a fault that was already known and is already being handled. + * + * If there are too many pending page faults, this will fail. Retry + * interrupts should be ignored in this case until there is enough + * free space. + * + * Returns 0 if the fault was added, 1 if the fault was already known, + * -ENOSPC if there are too many pending faults. + */ +int amdgpu_ih_add_fault(struct amdgpu_device *adev, u64 key) +{ + unsigned long flags; + int r = -ENOSPC; + + if (WARN_ON_ONCE(!adev->irq.ih.faults)) + /* Should be allocated in <IP>_ih_sw_init on GPUs that + * support retry faults and require retry filtering. + */ + return r; + + spin_lock_irqsave(&adev->irq.ih.faults->lock, flags); + + /* Only let the hash table fill up to 50% for best performance */ + if (adev->irq.ih.faults->count >= (1 << (AMDGPU_PAGEFAULT_HASH_BITS-1))) + goto unlock_out; + + r = chash_table_copy_in(&adev->irq.ih.faults->hash, key, NULL); + if (!r) + adev->irq.ih.faults->count++; + + /* chash_table_copy_in should never fail unless we're losing count */ + WARN_ON_ONCE(r < 0); + +unlock_out: + spin_unlock_irqrestore(&adev->irq.ih.faults->lock, flags); + return r; +} + +/** + * amdgpu_ih_clear_fault - Remove a page fault record + * + * @adev: amdgpu device pointer + * @key: 64-bit encoding of PASID and address + * + * This should be called when a page fault has been handled. Any + * future interrupt with this key will be processed as a new + * page fault. + */ +void amdgpu_ih_clear_fault(struct amdgpu_device *adev, u64 key) +{ + unsigned long flags; + int r; + + if (!adev->irq.ih.faults) + return; + + spin_lock_irqsave(&adev->irq.ih.faults->lock, flags); + + r = chash_table_remove(&adev->irq.ih.faults->hash, key, NULL); + if (!WARN_ON_ONCE(r < 0)) { + adev->irq.ih.faults->count--; + WARN_ON_ONCE(adev->irq.ih.faults->count < 0); + } + + spin_unlock_irqrestore(&adev->irq.ih.faults->lock, flags); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h index 3de8e74e5b3a..ada89358e220 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h @@ -24,6 +24,8 @@ #ifndef __AMDGPU_IH_H__ #define __AMDGPU_IH_H__ +#include <linux/chash.h> + struct amdgpu_device; /* * vega10+ IH clients @@ -69,6 +71,13 @@ enum amdgpu_ih_clientid #define AMDGPU_IH_CLIENTID_LEGACY 0 +#define AMDGPU_PAGEFAULT_HASH_BITS 8 +struct amdgpu_retryfault_hashtable { + DECLARE_CHASH_TABLE(hash, AMDGPU_PAGEFAULT_HASH_BITS, 8, 0); + spinlock_t lock; + int count; +}; + /* * R6xx+ IH ring */ @@ -87,6 +96,7 @@ struct amdgpu_ih_ring { bool use_doorbell; bool use_bus_addr; dma_addr_t rb_dma_addr; /* only used when use_bus_addr = true */ + struct amdgpu_retryfault_hashtable *faults; }; #define AMDGPU_IH_SRC_DATA_MAX_SIZE_DW 4 @@ -109,5 +119,7 @@ int amdgpu_ih_ring_init(struct amdgpu_device *adev, unsigned ring_size, bool use_bus_addr); void amdgpu_ih_ring_fini(struct amdgpu_device *adev); int amdgpu_ih_process(struct amdgpu_device *adev); +int amdgpu_ih_add_fault(struct amdgpu_device *adev, u64 key); +void amdgpu_ih_clear_fault(struct amdgpu_device *adev, u64 key); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c index 62da6c5c6095..47c5ce9807db 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c @@ -37,6 +37,10 @@ #include <linux/pm_runtime.h> +#ifdef CONFIG_DRM_AMD_DC +#include "amdgpu_dm_irq.h" +#endif + #define AMDGPU_WAIT_IDLE_TIMEOUT 200 /* @@ -220,10 +224,6 @@ int amdgpu_irq_init(struct amdgpu_device *adev) int r = 0; spin_lock_init(&adev->irq.lock); - r = drm_vblank_init(adev->ddev, adev->mode_info.num_crtc); - if (r) { - return r; - } /* enable msi */ adev->irq.msi_enabled = false; @@ -236,7 +236,21 @@ int amdgpu_irq_init(struct amdgpu_device *adev) } } - INIT_WORK(&adev->hotplug_work, amdgpu_hotplug_work_func); + if (!amdgpu_device_has_dc_support(adev)) { + if (!adev->enable_virtual_display) + /* Disable vblank irqs aggressively for power-saving */ + /* XXX: can this be enabled for DC? */ + adev->ddev->vblank_disable_immediate = true; + + r = drm_vblank_init(adev->ddev, adev->mode_info.num_crtc); + if (r) + return r; + + /* pre DCE11 */ + INIT_WORK(&adev->hotplug_work, + amdgpu_hotplug_work_func); + } + INIT_WORK(&adev->reset_work, amdgpu_irq_reset_work_func); adev->irq.installed = true; @@ -263,7 +277,6 @@ void amdgpu_irq_fini(struct amdgpu_device *adev) { unsigned i, j; - drm_vblank_cleanup(adev->ddev); if (adev->irq.installed) { drm_irq_uninstall(adev->ddev); adev->irq.installed = false; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index 3d641e10e6b6..0cfc68db575b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -65,6 +65,7 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, amdgpu_sync_create(&(*job)->sync); amdgpu_sync_create(&(*job)->dep_sync); amdgpu_sync_create(&(*job)->sched_sync); + (*job)->vram_lost_counter = atomic_read(&adev->vram_lost_counter); return 0; } @@ -81,6 +82,8 @@ int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, r = amdgpu_ib_get(adev, NULL, size, &(*job)->ibs[0]); if (r) kfree(*job); + else + (*job)->vm_pd_addr = adev->gart.table_addr; return r; } @@ -101,6 +104,7 @@ static void amdgpu_job_free_cb(struct amd_sched_job *s_job) { struct amdgpu_job *job = container_of(s_job, struct amdgpu_job, base); + amdgpu_ring_priority_put(job->ring, amd_sched_get_job_priority(s_job)); dma_fence_put(job->fence); amdgpu_sync_free(&job->sync); amdgpu_sync_free(&job->dep_sync); @@ -137,6 +141,8 @@ int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring, job->fence_ctx = entity->fence_context; *f = dma_fence_get(&job->base.s_fence->finished); amdgpu_job_free_resources(job); + amdgpu_ring_priority_get(job->ring, + amd_sched_get_job_priority(&job->base)); amd_sched_entity_push_job(&job->base); return 0; @@ -175,8 +181,8 @@ static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job) static struct dma_fence *amdgpu_job_run(struct amd_sched_job *sched_job) { struct dma_fence *fence = NULL; + struct amdgpu_device *adev; struct amdgpu_job *job; - struct amdgpu_fpriv *fpriv = NULL; int r; if (!sched_job) { @@ -184,23 +190,25 @@ static struct dma_fence *amdgpu_job_run(struct amd_sched_job *sched_job) return NULL; } job = to_amdgpu_job(sched_job); + adev = job->adev; BUG_ON(amdgpu_sync_peek_fence(&job->sync, NULL)); trace_amdgpu_sched_run_job(job); - if (job->vm) - fpriv = container_of(job->vm, struct amdgpu_fpriv, vm); /* skip ib schedule when vram is lost */ - if (fpriv && amdgpu_kms_vram_lost(job->adev, fpriv)) + if (job->vram_lost_counter != atomic_read(&adev->vram_lost_counter)) { + dma_fence_set_error(&job->base.s_fence->finished, -ECANCELED); DRM_ERROR("Skip scheduling IBs!\n"); - else { - r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs, job, &fence); + } else { + r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs, job, + &fence); if (r) DRM_ERROR("Error scheduling IBs (%d)\n", r); } /* if gpu reset, hw fence will be replaced here */ dma_fence_put(job->fence); job->fence = dma_fence_get(fence); + amdgpu_job_free_resources(job); return fence; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index b0b23101d1c8..720139e182a3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -28,6 +28,7 @@ #include <drm/drmP.h> #include "amdgpu.h" #include <drm/amdgpu_drm.h> +#include "amdgpu_sched.h" #include "amdgpu_uvd.h" #include "amdgpu_vce.h" @@ -158,7 +159,6 @@ int amdgpu_driver_load_kms(struct drm_device *dev, unsigned long flags) "Error during ACPI methods call\n"); } - amdgpu_amdkfd_load_interface(adev); amdgpu_amdkfd_device_probe(adev); amdgpu_amdkfd_device_init(adev); @@ -270,7 +270,6 @@ static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info, static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { struct amdgpu_device *adev = dev->dev_private; - struct amdgpu_fpriv *fpriv = filp->driver_priv; struct drm_amdgpu_info *info = data; struct amdgpu_mode_info *minfo = &adev->mode_info; void __user *out = (void __user *)(uintptr_t)info->return_pointer; @@ -283,8 +282,6 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file if (!info->return_size || !info->return_pointer) return -EINVAL; - if (amdgpu_kms_vram_lost(adev, fpriv)) - return -ENODEV; switch (info->query) { case AMDGPU_INFO_ACCEL_WORKING: @@ -456,13 +453,13 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file ui64 = atomic64_read(&adev->num_vram_cpu_page_faults); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; case AMDGPU_INFO_VRAM_USAGE: - ui64 = atomic64_read(&adev->vram_usage); + ui64 = amdgpu_vram_mgr_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; case AMDGPU_INFO_VIS_VRAM_USAGE: - ui64 = atomic64_read(&adev->vram_vis_usage); + ui64 = amdgpu_vram_mgr_vis_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; case AMDGPU_INFO_GTT_USAGE: - ui64 = atomic64_read(&adev->gtt_usage); + ui64 = amdgpu_gtt_mgr_usage(&adev->mman.bdev.man[TTM_PL_TT]); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; case AMDGPU_INFO_GDS_CONFIG: { struct drm_amdgpu_info_gds gds_info; @@ -485,7 +482,8 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file vram_gtt.vram_size -= adev->vram_pin_size; vram_gtt.vram_cpu_accessible_size = adev->mc.visible_vram_size; vram_gtt.vram_cpu_accessible_size -= (adev->vram_pin_size - adev->invisible_pin_size); - vram_gtt.gtt_size = adev->mc.gtt_size; + vram_gtt.gtt_size = adev->mman.bdev.man[TTM_PL_TT].size; + vram_gtt.gtt_size *= PAGE_SIZE; vram_gtt.gtt_size -= adev->gart_pin_size; return copy_to_user(out, &vram_gtt, min((size_t)size, sizeof(vram_gtt))) ? -EFAULT : 0; @@ -497,7 +495,8 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file mem.vram.total_heap_size = adev->mc.real_vram_size; mem.vram.usable_heap_size = adev->mc.real_vram_size - adev->vram_pin_size; - mem.vram.heap_usage = atomic64_read(&adev->vram_usage); + mem.vram.heap_usage = + amdgpu_vram_mgr_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); mem.vram.max_allocation = mem.vram.usable_heap_size * 3 / 4; mem.cpu_accessible_vram.total_heap_size = @@ -506,14 +505,16 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file adev->mc.visible_vram_size - (adev->vram_pin_size - adev->invisible_pin_size); mem.cpu_accessible_vram.heap_usage = - atomic64_read(&adev->vram_vis_usage); + amdgpu_vram_mgr_vis_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); mem.cpu_accessible_vram.max_allocation = mem.cpu_accessible_vram.usable_heap_size * 3 / 4; - mem.gtt.total_heap_size = adev->mc.gtt_size; - mem.gtt.usable_heap_size = - adev->mc.gtt_size - adev->gart_pin_size; - mem.gtt.heap_usage = atomic64_read(&adev->gtt_usage); + mem.gtt.total_heap_size = adev->mman.bdev.man[TTM_PL_TT].size; + mem.gtt.total_heap_size *= PAGE_SIZE; + mem.gtt.usable_heap_size = mem.gtt.total_heap_size + - adev->gart_pin_size; + mem.gtt.heap_usage = + amdgpu_gtt_mgr_usage(&adev->mman.bdev.man[TTM_PL_TT]); mem.gtt.max_allocation = mem.gtt.usable_heap_size * 3 / 4; return copy_to_user(out, &mem, @@ -571,8 +572,8 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file dev_info.max_engine_clock = amdgpu_dpm_get_sclk(adev, false) * 10; dev_info.max_memory_clock = amdgpu_dpm_get_mclk(adev, false) * 10; } else { - dev_info.max_engine_clock = adev->pm.default_sclk * 10; - dev_info.max_memory_clock = adev->pm.default_mclk * 10; + dev_info.max_engine_clock = adev->clock.default_sclk * 10; + dev_info.max_memory_clock = adev->clock.default_mclk * 10; } dev_info.enabled_rb_pipes_mask = adev->gfx.config.backend_enable_mask; dev_info.num_rb_pipes = adev->gfx.config.max_backends_per_se * @@ -587,10 +588,8 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file dev_info.virtual_address_offset = AMDGPU_VA_RESERVED_SIZE; dev_info.virtual_address_max = (uint64_t)adev->vm_manager.max_pfn * AMDGPU_GPU_PAGE_SIZE; dev_info.virtual_address_alignment = max((int)PAGE_SIZE, AMDGPU_GPU_PAGE_SIZE); - dev_info.pte_fragment_size = (1 << AMDGPU_LOG2_PAGES_PER_FRAG) * - AMDGPU_GPU_PAGE_SIZE; + dev_info.pte_fragment_size = (1 << adev->vm_manager.fragment_size) * AMDGPU_GPU_PAGE_SIZE; dev_info.gart_page_size = AMDGPU_GPU_PAGE_SIZE; - dev_info.cu_active_number = adev->gfx.cu_info.number; dev_info.cu_ao_mask = adev->gfx.cu_info.ao_cu_mask; dev_info.ce_ram_size = adev->gfx.ce_ram_size; @@ -764,6 +763,9 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file } return copy_to_user(out, &ui32, min(size, 4u)) ? -EFAULT : 0; } + case AMDGPU_INFO_VRAM_LOST_COUNTER: + ui32 = atomic_read(&adev->vram_lost_counter); + return copy_to_user(out, &ui32, min(size, 4u)) ? -EFAULT : 0; default: DRM_DEBUG_KMS("Invalid request %d\n", info->query); return -EINVAL; @@ -790,12 +792,6 @@ void amdgpu_driver_lastclose_kms(struct drm_device *dev) vga_switcheroo_process_delayed_switch(); } -bool amdgpu_kms_vram_lost(struct amdgpu_device *adev, - struct amdgpu_fpriv *fpriv) -{ - return fpriv->vram_lost_counter != atomic_read(&adev->vram_lost_counter); -} - /** * amdgpu_driver_open_kms - drm callback for open * @@ -824,7 +820,7 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) } r = amdgpu_vm_init(adev, &fpriv->vm, - AMDGPU_VM_CONTEXT_GFX); + AMDGPU_VM_CONTEXT_GFX, 0); if (r) { kfree(fpriv); goto out_suspend; @@ -839,9 +835,12 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) } if (amdgpu_sriov_vf(adev)) { - r = amdgpu_map_static_csa(adev, &fpriv->vm); - if (r) + r = amdgpu_map_static_csa(adev, &fpriv->vm, &fpriv->csa_va); + if (r) { + amdgpu_vm_fini(adev, &fpriv->vm); + kfree(fpriv); goto out_suspend; + } } mutex_init(&fpriv->bo_list_lock); @@ -849,7 +848,6 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) amdgpu_ctx_mgr_init(&fpriv->ctx_mgr); - fpriv->vram_lost_counter = atomic_read(&adev->vram_lost_counter); file_priv->driver_priv = fpriv; out_suspend: @@ -892,8 +890,8 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, if (amdgpu_sriov_vf(adev)) { /* TODO: how to handle reserve failure */ BUG_ON(amdgpu_bo_reserve(adev->virt.csa_obj, true)); - amdgpu_vm_bo_rmv(adev, fpriv->vm.csa_bo_va); - fpriv->vm.csa_bo_va = NULL; + amdgpu_vm_bo_rmv(adev, fpriv->csa_va); + fpriv->csa_va = NULL; amdgpu_bo_unreserve(adev->virt.csa_obj); } @@ -1019,7 +1017,9 @@ const struct drm_ioctl_desc amdgpu_ioctls_kms[] = { DRM_IOCTL_DEF_DRV(AMDGPU_GEM_CREATE, amdgpu_gem_create_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(AMDGPU_CTX, amdgpu_ctx_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(AMDGPU_VM, amdgpu_vm_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), + DRM_IOCTL_DEF_DRV(AMDGPU_SCHED, amdgpu_sched_ioctl, DRM_MASTER), DRM_IOCTL_DEF_DRV(AMDGPU_BO_LIST, amdgpu_bo_list_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), + DRM_IOCTL_DEF_DRV(AMDGPU_FENCE_TO_HANDLE, amdgpu_cs_fence_to_handle_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), /* KMS */ DRM_IOCTL_DEF_DRV(AMDGPU_GEM_MMAP, amdgpu_gem_mmap_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(AMDGPU_GEM_WAIT_IDLE, amdgpu_gem_wait_idle_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), @@ -1030,7 +1030,7 @@ const struct drm_ioctl_desc amdgpu_ioctls_kms[] = { DRM_IOCTL_DEF_DRV(AMDGPU_GEM_METADATA, amdgpu_gem_metadata_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(AMDGPU_GEM_VA, amdgpu_gem_va_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), DRM_IOCTL_DEF_DRV(AMDGPU_GEM_OP, amdgpu_gem_op_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), - DRM_IOCTL_DEF_DRV(AMDGPU_GEM_USERPTR, amdgpu_gem_userptr_ioctl, DRM_AUTH|DRM_RENDER_ALLOW), + DRM_IOCTL_DEF_DRV(AMDGPU_GEM_USERPTR, amdgpu_gem_userptr_ioctl, DRM_AUTH|DRM_RENDER_ALLOW) }; const int amdgpu_max_kms_ioctl = ARRAY_SIZE(amdgpu_ioctls_kms); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c index 38f739fb727b..bd67f4cb8e6c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c @@ -50,8 +50,10 @@ struct amdgpu_mn { struct hlist_node node; /* objects protected by lock */ - struct mutex lock; - struct rb_root objects; + struct rw_semaphore lock; + struct rb_root_cached objects; + struct mutex read_lock; + atomic_t recursion; }; struct amdgpu_mn_node { @@ -74,17 +76,17 @@ static void amdgpu_mn_destroy(struct work_struct *work) struct amdgpu_bo *bo, *next_bo; mutex_lock(&adev->mn_lock); - mutex_lock(&rmn->lock); + down_write(&rmn->lock); hash_del(&rmn->node); - rbtree_postorder_for_each_entry_safe(node, next_node, &rmn->objects, - it.rb) { + rbtree_postorder_for_each_entry_safe(node, next_node, + &rmn->objects.rb_root, it.rb) { list_for_each_entry_safe(bo, next_bo, &node->bos, mn_list) { bo->mn = NULL; list_del_init(&bo->mn_list); } kfree(node); } - mutex_unlock(&rmn->lock); + up_write(&rmn->lock); mutex_unlock(&adev->mn_lock); mmu_notifier_unregister_no_release(&rmn->mn, rmn->mm); kfree(rmn); @@ -106,6 +108,53 @@ static void amdgpu_mn_release(struct mmu_notifier *mn, schedule_work(&rmn->work); } + +/** + * amdgpu_mn_lock - take the write side lock for this mn + */ +void amdgpu_mn_lock(struct amdgpu_mn *mn) +{ + if (mn) + down_write(&mn->lock); +} + +/** + * amdgpu_mn_unlock - drop the write side lock for this mn + */ +void amdgpu_mn_unlock(struct amdgpu_mn *mn) +{ + if (mn) + up_write(&mn->lock); +} + +/** + * amdgpu_mn_read_lock - take the rmn read lock + * + * @rmn: our notifier + * + * Take the rmn read side lock. + */ +static void amdgpu_mn_read_lock(struct amdgpu_mn *rmn) +{ + mutex_lock(&rmn->read_lock); + if (atomic_inc_return(&rmn->recursion) == 1) + down_read_non_owner(&rmn->lock); + mutex_unlock(&rmn->read_lock); +} + +/** + * amdgpu_mn_read_unlock - drop the rmn read lock + * + * @rmn: our notifier + * + * Drop the rmn read side lock. + */ +static void amdgpu_mn_read_unlock(struct amdgpu_mn *rmn) +{ + if (atomic_dec_return(&rmn->recursion) == 0) + up_read_non_owner(&rmn->lock); +} + /** * amdgpu_mn_invalidate_node - unmap all BOs of a node * @@ -126,54 +175,13 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node, if (!amdgpu_ttm_tt_affect_userptr(bo->tbo.ttm, start, end)) continue; - r = amdgpu_bo_reserve(bo, true); - if (r) { - DRM_ERROR("(%ld) failed to reserve user bo\n", r); - continue; - } - r = reservation_object_wait_timeout_rcu(bo->tbo.resv, true, false, MAX_SCHEDULE_TIMEOUT); if (r <= 0) DRM_ERROR("(%ld) failed to wait for user bo\n", r); - amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU); - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); - if (r) - DRM_ERROR("(%ld) failed to validate user bo\n", r); - - amdgpu_bo_unreserve(bo); - } -} - -/** - * amdgpu_mn_invalidate_page - callback to notify about mm change - * - * @mn: our notifier - * @mn: the mm this callback is about - * @address: address of invalidate page - * - * Invalidation of a single page. Blocks for all BOs mapping it - * and unmap them by move them into system domain again. - */ -static void amdgpu_mn_invalidate_page(struct mmu_notifier *mn, - struct mm_struct *mm, - unsigned long address) -{ - struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn); - struct interval_tree_node *it; - - mutex_lock(&rmn->lock); - - it = interval_tree_iter_first(&rmn->objects, address, address); - if (it) { - struct amdgpu_mn_node *node; - - node = container_of(it, struct amdgpu_mn_node, it); - amdgpu_mn_invalidate_node(node, address, address); + amdgpu_ttm_tt_mark_user_pages(bo->tbo.ttm); } - - mutex_unlock(&rmn->lock); } /** @@ -198,7 +206,7 @@ static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn, /* notification is exclusive, but interval is inclusive */ end -= 1; - mutex_lock(&rmn->lock); + amdgpu_mn_read_lock(rmn); it = interval_tree_iter_first(&rmn->objects, start, end); while (it) { @@ -209,14 +217,32 @@ static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn, amdgpu_mn_invalidate_node(node, start, end); } +} - mutex_unlock(&rmn->lock); +/** + * amdgpu_mn_invalidate_range_end - callback to notify about mm change + * + * @mn: our notifier + * @mn: the mm this callback is about + * @start: start of updated range + * @end: end of updated range + * + * Release the lock again to allow new command submissions. + */ +static void amdgpu_mn_invalidate_range_end(struct mmu_notifier *mn, + struct mm_struct *mm, + unsigned long start, + unsigned long end) +{ + struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn); + + amdgpu_mn_read_unlock(rmn); } static const struct mmu_notifier_ops amdgpu_mn_ops = { .release = amdgpu_mn_release, - .invalidate_page = amdgpu_mn_invalidate_page, .invalidate_range_start = amdgpu_mn_invalidate_range_start, + .invalidate_range_end = amdgpu_mn_invalidate_range_end, }; /** @@ -226,7 +252,7 @@ static const struct mmu_notifier_ops amdgpu_mn_ops = { * * Creates a notifier context for current->mm. */ -static struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) +struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) { struct mm_struct *mm = current->mm; struct amdgpu_mn *rmn; @@ -251,8 +277,10 @@ static struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) rmn->adev = adev; rmn->mm = mm; rmn->mn.ops = &amdgpu_mn_ops; - mutex_init(&rmn->lock); - rmn->objects = RB_ROOT; + init_rwsem(&rmn->lock); + rmn->objects = RB_ROOT_CACHED; + mutex_init(&rmn->read_lock); + atomic_set(&rmn->recursion, 0); r = __mmu_notifier_register(&rmn->mn, mm); if (r) @@ -298,7 +326,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) INIT_LIST_HEAD(&bos); - mutex_lock(&rmn->lock); + down_write(&rmn->lock); while ((it = interval_tree_iter_first(&rmn->objects, addr, end))) { kfree(node); @@ -312,7 +340,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) if (!node) { node = kmalloc(sizeof(struct amdgpu_mn_node), GFP_KERNEL); if (!node) { - mutex_unlock(&rmn->lock); + up_write(&rmn->lock); return -ENOMEM; } } @@ -327,7 +355,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) interval_tree_insert(&node->it, &rmn->objects); - mutex_unlock(&rmn->lock); + up_write(&rmn->lock); return 0; } @@ -353,13 +381,13 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo) return; } - mutex_lock(&rmn->lock); + down_write(&rmn->lock); /* save the next list entry for later */ head = bo->mn_list.next; bo->mn = NULL; - list_del(&bo->mn_list); + list_del_init(&bo->mn_list); if (list_empty(head)) { struct amdgpu_mn_node *node; @@ -368,6 +396,7 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo) kfree(node); } - mutex_unlock(&rmn->lock); + up_write(&rmn->lock); mutex_unlock(&adev->mn_lock); } + diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h new file mode 100644 index 000000000000..d0095a3793b8 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h @@ -0,0 +1,52 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Christian König + */ +#ifndef __AMDGPU_MN_H__ +#define __AMDGPU_MN_H__ + +/* + * MMU Notifier + */ +struct amdgpu_mn; + +#if defined(CONFIG_MMU_NOTIFIER) +void amdgpu_mn_lock(struct amdgpu_mn *mn); +void amdgpu_mn_unlock(struct amdgpu_mn *mn); +struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev); +int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr); +void amdgpu_mn_unregister(struct amdgpu_bo *bo); +#else +static inline void amdgpu_mn_lock(struct amdgpu_mn *mn) {} +static inline void amdgpu_mn_unlock(struct amdgpu_mn *mn) {} +static inline struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) +{ + return NULL; +} +static inline int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) +{ + return -ENODEV; +} +static inline void amdgpu_mn_unregister(struct amdgpu_bo *bo) {} +#endif + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index 43a9d3aec6c4..ffde1e9666e8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -38,11 +38,15 @@ #include <drm/drm_crtc_helper.h> #include <drm/drm_fb_helper.h> #include <drm/drm_plane_helper.h> +#include <drm/drm_fb_helper.h> #include <linux/i2c.h> #include <linux/i2c-algo-bit.h> #include <linux/hrtimer.h> #include "amdgpu_irq.h" +#include <drm/drm_dp_mst_helper.h> +#include "modules/inc/mod_freesync.h" + struct amdgpu_bo; struct amdgpu_device; struct amdgpu_encoder; @@ -53,9 +57,13 @@ struct amdgpu_hpd; #define to_amdgpu_connector(x) container_of(x, struct amdgpu_connector, base) #define to_amdgpu_encoder(x) container_of(x, struct amdgpu_encoder, base) #define to_amdgpu_framebuffer(x) container_of(x, struct amdgpu_framebuffer, base) +#define to_amdgpu_plane(x) container_of(x, struct amdgpu_plane, base) + +#define to_dm_plane_state(x) container_of(x, struct dm_plane_state, base); #define AMDGPU_MAX_HPD_PINS 6 #define AMDGPU_MAX_CRTCS 6 +#define AMDGPU_MAX_PLANES 6 #define AMDGPU_MAX_AFMT_BLOCKS 9 enum amdgpu_rmx_type { @@ -257,15 +265,7 @@ struct amdgpu_audio { int num_pins; }; -struct amdgpu_mode_mc_save { - u32 vga_render_control; - u32 vga_hdp_control; - bool crtc_enabled[AMDGPU_MAX_CRTCS]; -}; - struct amdgpu_display_funcs { - /* vga render */ - void (*set_vga_render_state)(struct amdgpu_device *adev, bool render); /* display watermarks */ void (*bandwidth_update)(struct amdgpu_device *adev); /* get frame count */ @@ -300,10 +300,30 @@ struct amdgpu_display_funcs { uint16_t connector_object_id, struct amdgpu_hpd *hpd, struct amdgpu_router *router); - void (*stop_mc_access)(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save); - void (*resume_mc_access)(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save); + /* it is used to enter or exit into free sync mode */ + int (*notify_freesync)(struct drm_device *dev, void *data, + struct drm_file *filp); + /* it is used to allow enablement of freesync mode */ + int (*set_freesync_property)(struct drm_connector *connector, + struct drm_property *property, + uint64_t val); + + +}; + +struct amdgpu_framebuffer { + struct drm_framebuffer base; + struct drm_gem_object *obj; + + /* caching for later use */ + uint64_t address; +}; + +struct amdgpu_fbdev { + struct drm_fb_helper helper; + struct amdgpu_framebuffer rfb; + struct list_head fbdev_list; + struct amdgpu_device *adev; }; struct amdgpu_mode_info { @@ -311,6 +331,7 @@ struct amdgpu_mode_info { struct card_info *atom_card_info; bool mode_config_initialized; struct amdgpu_crtc *crtcs[AMDGPU_MAX_CRTCS]; + struct amdgpu_plane *planes[AMDGPU_MAX_PLANES]; struct amdgpu_afmt *afmt[AMDGPU_MAX_AFMT_BLOCKS]; /* DVI-I properties */ struct drm_property *coherent_mode_property; @@ -340,6 +361,7 @@ struct amdgpu_mode_info { int num_dig; /* number of dig blocks */ int disp_priority; const struct amdgpu_display_funcs *funcs; + const enum drm_plane_type *plane_type; }; #define AMDGPU_MAX_BL_LEVEL 0xFF @@ -369,7 +391,6 @@ struct amdgpu_atom_ss { struct amdgpu_crtc { struct drm_crtc base; int crtc_id; - u16 lut_r[256], lut_g[256], lut_b[256]; bool enabled; bool can_tile; uint32_t crtc_offset; @@ -413,6 +434,14 @@ struct amdgpu_crtc { /* for virtual dce */ struct hrtimer vblank_timer; enum amdgpu_interrupt_state vsync_timer_enabled; + + int otg_inst; + struct drm_pending_vblank_event *event; +}; + +struct amdgpu_plane { + struct drm_plane base; + enum drm_plane_type plane_type; }; struct amdgpu_encoder_atom_dig { @@ -502,6 +531,19 @@ enum amdgpu_connector_dither { AMDGPU_FMT_DITHER_ENABLE = 1, }; +struct amdgpu_dm_dp_aux { + struct drm_dp_aux aux; + struct ddc_service *ddc_service; +}; + +struct amdgpu_i2c_adapter { + struct i2c_adapter base; + + struct ddc_service *ddc_service; +}; + +#define TO_DM_AUX(x) container_of((x), struct amdgpu_dm_dp_aux, aux) + struct amdgpu_connector { struct drm_connector base; uint32_t connector_id; @@ -513,6 +555,14 @@ struct amdgpu_connector { /* we need to mind the EDID between detect and get modes due to analog/digital/tvencoder */ struct edid *edid; + /* number of modes generated from EDID at 'dc_sink' */ + int num_modes; + /* The 'old' sink - before an HPD. + * The 'current' sink is in dc_link->sink. */ + struct dc_sink *dc_sink; + struct dc_link *dc_link; + struct dc_sink *dc_em_sink; + const struct dc_stream *stream; void *con_priv; bool dac_load_detect; bool detected_by_load; /* if the connection status was determined by load */ @@ -523,11 +573,39 @@ struct amdgpu_connector { enum amdgpu_connector_audio audio; enum amdgpu_connector_dither dither; unsigned pixelclock_for_modeset; + + struct drm_dp_mst_topology_mgr mst_mgr; + struct amdgpu_dm_dp_aux dm_dp_aux; + struct drm_dp_mst_port *port; + struct amdgpu_connector *mst_port; + struct amdgpu_encoder *mst_encoder; + struct semaphore mst_sem; + + /* TODO see if we can merge with ddc_bus or make a dm_connector */ + struct amdgpu_i2c_adapter *i2c; + + /* Monitor range limits */ + int min_vfreq ; + int max_vfreq ; + int pixel_clock_mhz; + + /*freesync caps*/ + struct mod_freesync_caps caps; + + struct mutex hpd_lock; + }; -struct amdgpu_framebuffer { - struct drm_framebuffer base; - struct drm_gem_object *obj; +/* TODO: start to use this struct and remove same field from base one */ +struct amdgpu_mst_connector { + struct amdgpu_connector base; + + struct drm_dp_mst_topology_mgr mst_mgr; + struct amdgpu_dm_dp_aux dm_dp_aux; + struct drm_dp_mst_port *port; + struct amdgpu_connector *mst_port; + bool is_mst_connector; + struct amdgpu_encoder *mst_encoder; }; #define ENCODER_MODE_IS_DP(em) (((em) == ATOM_ENCODER_MODE_DP) || \ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 8ee69652be8c..ea25164e7f4b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -37,63 +37,12 @@ #include "amdgpu.h" #include "amdgpu_trace.h" - - -static u64 amdgpu_get_vis_part_size(struct amdgpu_device *adev, - struct ttm_mem_reg *mem) -{ - if (mem->start << PAGE_SHIFT >= adev->mc.visible_vram_size) - return 0; - - return ((mem->start << PAGE_SHIFT) + mem->size) > - adev->mc.visible_vram_size ? - adev->mc.visible_vram_size - (mem->start << PAGE_SHIFT) : - mem->size; -} - -static void amdgpu_update_memory_usage(struct amdgpu_device *adev, - struct ttm_mem_reg *old_mem, - struct ttm_mem_reg *new_mem) -{ - u64 vis_size; - if (!adev) - return; - - if (new_mem) { - switch (new_mem->mem_type) { - case TTM_PL_TT: - atomic64_add(new_mem->size, &adev->gtt_usage); - break; - case TTM_PL_VRAM: - atomic64_add(new_mem->size, &adev->vram_usage); - vis_size = amdgpu_get_vis_part_size(adev, new_mem); - atomic64_add(vis_size, &adev->vram_vis_usage); - break; - } - } - - if (old_mem) { - switch (old_mem->mem_type) { - case TTM_PL_TT: - atomic64_sub(old_mem->size, &adev->gtt_usage); - break; - case TTM_PL_VRAM: - atomic64_sub(old_mem->size, &adev->vram_usage); - vis_size = amdgpu_get_vis_part_size(adev, old_mem); - atomic64_sub(vis_size, &adev->vram_vis_usage); - break; - } - } -} - static void amdgpu_ttm_bo_destroy(struct ttm_buffer_object *tbo) { struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev); - struct amdgpu_bo *bo; - - bo = container_of(tbo, struct amdgpu_bo, tbo); + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(tbo); - amdgpu_update_memory_usage(adev, &bo->tbo.mem, NULL); + amdgpu_bo_kunmap(bo); drm_gem_object_release(&bo->gem_base); amdgpu_bo_unref(&bo->parent); @@ -113,11 +62,12 @@ bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo) return false; } -static void amdgpu_ttm_placement_init(struct amdgpu_device *adev, - struct ttm_placement *placement, - struct ttm_place *places, - u32 domain, u64 flags) +void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain) { + struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev); + struct ttm_placement *placement = &abo->placement; + struct ttm_place *places = abo->placements; + u64 flags = abo->flags; u32 c = 0; if (domain & AMDGPU_GEM_DOMAIN_VRAM) { @@ -140,7 +90,10 @@ static void amdgpu_ttm_placement_init(struct amdgpu_device *adev, if (domain & AMDGPU_GEM_DOMAIN_GTT) { places[c].fpfn = 0; - places[c].lpfn = 0; + if (flags & AMDGPU_GEM_CREATE_SHADOW) + places[c].lpfn = adev->mc.gart_size >> PAGE_SHIFT; + else + places[c].lpfn = 0; places[c].flags = TTM_PL_FLAG_TT; if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) places[c].flags |= TTM_PL_FLAG_WC | @@ -197,29 +150,8 @@ static void amdgpu_ttm_placement_init(struct amdgpu_device *adev, placement->busy_placement = places; } -void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain) -{ - struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev); - - amdgpu_ttm_placement_init(adev, &abo->placement, abo->placements, - domain, abo->flags); -} - -static void amdgpu_fill_placement_to_bo(struct amdgpu_bo *bo, - struct ttm_placement *placement) -{ - BUG_ON(placement->num_placement > (AMDGPU_GEM_DOMAIN_MAX + 1)); - - memcpy(bo->placements, placement->placement, - placement->num_placement * sizeof(struct ttm_place)); - bo->placement.num_placement = placement->num_placement; - bo->placement.num_busy_placement = placement->num_busy_placement; - bo->placement.placement = bo->placements; - bo->placement.busy_placement = bo->placements; -} - /** - * amdgpu_bo_create_kernel - create BO for kernel use + * amdgpu_bo_create_reserved - create reserved BO for kernel use * * @adev: amdgpu device object * @size: size for the new BO @@ -229,24 +161,30 @@ static void amdgpu_fill_placement_to_bo(struct amdgpu_bo *bo, * @gpu_addr: GPU addr of the pinned BO * @cpu_addr: optional CPU address mapping * - * Allocates and pins a BO for kernel internal use. + * Allocates and pins a BO for kernel internal use, and returns it still + * reserved. * * Returns 0 on success, negative error code otherwise. */ -int amdgpu_bo_create_kernel(struct amdgpu_device *adev, - unsigned long size, int align, - u32 domain, struct amdgpu_bo **bo_ptr, - u64 *gpu_addr, void **cpu_addr) +int amdgpu_bo_create_reserved(struct amdgpu_device *adev, + unsigned long size, int align, + u32 domain, struct amdgpu_bo **bo_ptr, + u64 *gpu_addr, void **cpu_addr) { + bool free = false; int r; - r = amdgpu_bo_create(adev, size, align, true, domain, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, bo_ptr); - if (r) { - dev_err(adev->dev, "(%d) failed to allocate kernel bo\n", r); - return r; + if (!*bo_ptr) { + r = amdgpu_bo_create(adev, size, align, true, domain, + AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | + AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, + NULL, NULL, 0, bo_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to allocate kernel bo\n", + r); + return r; + } + free = true; } r = amdgpu_bo_reserve(*bo_ptr, false); @@ -269,20 +207,52 @@ int amdgpu_bo_create_kernel(struct amdgpu_device *adev, } } - amdgpu_bo_unreserve(*bo_ptr); - return 0; error_unreserve: amdgpu_bo_unreserve(*bo_ptr); error_free: - amdgpu_bo_unref(bo_ptr); + if (free) + amdgpu_bo_unref(bo_ptr); return r; } /** + * amdgpu_bo_create_kernel - create BO for kernel use + * + * @adev: amdgpu device object + * @size: size for the new BO + * @align: alignment for the new BO + * @domain: where to place it + * @bo_ptr: resulting BO + * @gpu_addr: GPU addr of the pinned BO + * @cpu_addr: optional CPU address mapping + * + * Allocates and pins a BO for kernel internal use. + * + * Returns 0 on success, negative error code otherwise. + */ +int amdgpu_bo_create_kernel(struct amdgpu_device *adev, + unsigned long size, int align, + u32 domain, struct amdgpu_bo **bo_ptr, + u64 *gpu_addr, void **cpu_addr) +{ + int r; + + r = amdgpu_bo_create_reserved(adev, size, align, domain, bo_ptr, + gpu_addr, cpu_addr); + + if (r) + return r; + + amdgpu_bo_unreserve(*bo_ptr); + + return 0; +} + +/** * amdgpu_bo_free_kernel - free BO for kernel use * * @bo: amdgpu BO to free @@ -311,18 +281,18 @@ void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr, *cpu_addr = NULL; } -int amdgpu_bo_create_restricted(struct amdgpu_device *adev, - unsigned long size, int byte_align, - bool kernel, u32 domain, u64 flags, - struct sg_table *sg, - struct ttm_placement *placement, - struct reservation_object *resv, - struct amdgpu_bo **bo_ptr) +static int amdgpu_bo_do_create(struct amdgpu_device *adev, + unsigned long size, int byte_align, + bool kernel, u32 domain, u64 flags, + struct sg_table *sg, + struct reservation_object *resv, + uint64_t init_value, + struct amdgpu_bo **bo_ptr) { struct amdgpu_bo *bo; enum ttm_bo_type type; unsigned long page_align; - u64 initial_bytes_moved; + u64 initial_bytes_moved, bytes_moved; size_t acc_size; int r; @@ -351,13 +321,13 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev, } INIT_LIST_HEAD(&bo->shadow_list); INIT_LIST_HEAD(&bo->va); - bo->prefered_domains = domain & (AMDGPU_GEM_DOMAIN_VRAM | + bo->preferred_domains = domain & (AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT | AMDGPU_GEM_DOMAIN_CPU | AMDGPU_GEM_DOMAIN_GDS | AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA); - bo->allowed_domains = bo->prefered_domains; + bo->allowed_domains = bo->preferred_domains; if (!kernel && bo->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM) bo->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT; @@ -391,19 +361,26 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev, bo->flags &= ~AMDGPU_GEM_CREATE_CPU_GTT_USWC; #endif - amdgpu_fill_placement_to_bo(bo, placement); - /* Kernel allocation are uninterruptible */ + bo->tbo.bdev = &adev->mman.bdev; + amdgpu_ttm_placement_from_domain(bo, domain); initial_bytes_moved = atomic64_read(&adev->num_bytes_moved); + /* Kernel allocation are uninterruptible */ r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, type, &bo->placement, page_align, !kernel, NULL, acc_size, sg, resv, &amdgpu_ttm_bo_destroy); - amdgpu_cs_report_moved_bytes(adev, - atomic64_read(&adev->num_bytes_moved) - initial_bytes_moved); - if (unlikely(r != 0)) return r; + bytes_moved = atomic64_read(&adev->num_bytes_moved) - + initial_bytes_moved; + if (adev->mc.visible_vram_size < adev->mc.real_vram_size && + bo->tbo.mem.mem_type == TTM_PL_VRAM && + bo->tbo.mem.start < adev->mc.visible_vram_size >> PAGE_SHIFT) + amdgpu_cs_report_moved_bytes(adev, bytes_moved, bytes_moved); + else + amdgpu_cs_report_moved_bytes(adev, bytes_moved, 0); + if (kernel) bo->tbo.priority = 1; @@ -411,7 +388,7 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev, bo->tbo.mem.placement & TTM_PL_FLAG_VRAM) { struct dma_fence *fence; - r = amdgpu_fill_buffer(bo, 0, bo->tbo.resv, &fence); + r = amdgpu_fill_buffer(bo, init_value, bo->tbo.resv, &fence); if (unlikely(r)) goto fail_unreserve; @@ -426,6 +403,10 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev, trace_amdgpu_bo_create(bo); + /* Treat CPU_ACCESS_REQUIRED only as a hint if given by UMD */ + if (type == ttm_bo_type_device) + bo->flags &= ~AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; + return 0; fail_unreserve: @@ -439,27 +420,17 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, unsigned long size, int byte_align, struct amdgpu_bo *bo) { - struct ttm_placement placement = {0}; - struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1]; int r; if (bo->shadow) return 0; - bo->flags |= AMDGPU_GEM_CREATE_SHADOW; - memset(&placements, 0, - (AMDGPU_GEM_DOMAIN_MAX + 1) * sizeof(struct ttm_place)); - - amdgpu_ttm_placement_init(adev, &placement, - placements, AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_CPU_GTT_USWC); - - r = amdgpu_bo_create_restricted(adev, size, byte_align, true, - AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_CPU_GTT_USWC, - NULL, &placement, - bo->tbo.resv, - &bo->shadow); + r = amdgpu_bo_do_create(adev, size, byte_align, true, + AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_CREATE_CPU_GTT_USWC | + AMDGPU_GEM_CREATE_SHADOW, + NULL, bo->tbo.resv, 0, + &bo->shadow); if (!r) { bo->shadow->parent = amdgpu_bo_ref(bo); mutex_lock(&adev->shadow_list_lock); @@ -470,39 +441,34 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, return r; } +/* init_value will only take effect when flags contains + * AMDGPU_GEM_CREATE_VRAM_CLEARED. + */ int amdgpu_bo_create(struct amdgpu_device *adev, unsigned long size, int byte_align, bool kernel, u32 domain, u64 flags, struct sg_table *sg, struct reservation_object *resv, + uint64_t init_value, struct amdgpu_bo **bo_ptr) { - struct ttm_placement placement = {0}; - struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1]; + uint64_t parent_flags = flags & ~AMDGPU_GEM_CREATE_SHADOW; int r; - memset(&placements, 0, - (AMDGPU_GEM_DOMAIN_MAX + 1) * sizeof(struct ttm_place)); - - amdgpu_ttm_placement_init(adev, &placement, - placements, domain, flags); - - r = amdgpu_bo_create_restricted(adev, size, byte_align, kernel, - domain, flags, sg, &placement, - resv, bo_ptr); + r = amdgpu_bo_do_create(adev, size, byte_align, kernel, domain, + parent_flags, sg, resv, init_value, bo_ptr); if (r) return r; - if (amdgpu_need_backup(adev) && (flags & AMDGPU_GEM_CREATE_SHADOW)) { - if (!resv) { - r = ww_mutex_lock(&(*bo_ptr)->tbo.resv->lock, NULL); - WARN_ON(r != 0); - } + if ((flags & AMDGPU_GEM_CREATE_SHADOW) && amdgpu_need_backup(adev)) { + if (!resv) + WARN_ON(reservation_object_lock((*bo_ptr)->tbo.resv, + NULL)); r = amdgpu_bo_create_shadow(adev, size, byte_align, (*bo_ptr)); if (!resv) - ww_mutex_unlock(&(*bo_ptr)->tbo.resv->lock); + reservation_object_unlock((*bo_ptr)->tbo.resv); if (r) amdgpu_bo_unref(bo_ptr); @@ -535,7 +501,7 @@ int amdgpu_bo_backup_to_shadow(struct amdgpu_device *adev, r = amdgpu_copy_buffer(ring, bo_addr, shadow_addr, amdgpu_bo_size(bo), resv, fence, - direct); + direct, false); if (!r) amdgpu_bo_fence(bo, *fence, true); @@ -551,7 +517,7 @@ int amdgpu_bo_validate(struct amdgpu_bo *bo) if (bo->pin_count) return 0; - domain = bo->prefered_domains; + domain = bo->preferred_domains; retry: amdgpu_ttm_placement_from_domain(bo, domain); @@ -588,7 +554,7 @@ int amdgpu_bo_restore_from_shadow(struct amdgpu_device *adev, r = amdgpu_copy_buffer(ring, shadow_addr, bo_addr, amdgpu_bo_size(bo), resv, fence, - direct); + direct, false); if (!r) amdgpu_bo_fence(bo, *fence, true); @@ -598,16 +564,16 @@ err: int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr) { - bool is_iomem; + void *kptr; long r; if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) return -EPERM; - if (bo->kptr) { - if (ptr) { - *ptr = bo->kptr; - } + kptr = amdgpu_bo_kptr(bo); + if (kptr) { + if (ptr) + *ptr = kptr; return 0; } @@ -620,19 +586,23 @@ int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr) if (r) return r; - bo->kptr = ttm_kmap_obj_virtual(&bo->kmap, &is_iomem); if (ptr) - *ptr = bo->kptr; + *ptr = amdgpu_bo_kptr(bo); return 0; } +void *amdgpu_bo_kptr(struct amdgpu_bo *bo) +{ + bool is_iomem; + + return ttm_kmap_obj_virtual(&bo->kmap, &is_iomem); +} + void amdgpu_bo_kunmap(struct amdgpu_bo *bo) { - if (bo->kptr == NULL) - return; - bo->kptr = NULL; - ttm_bo_kunmap(&bo->kmap); + if (bo->kmap.bo) + ttm_bo_kunmap(&bo->kmap); } struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo) @@ -663,7 +633,6 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); int r, i; - unsigned fpfn, lpfn; if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) return -EPERM; @@ -695,22 +664,16 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, } bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; + /* force to pin into visible video ram */ + if (!(bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) + bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; amdgpu_ttm_placement_from_domain(bo, domain); for (i = 0; i < bo->placement.num_placement; i++) { - /* force to pin into visible video ram */ - if ((bo->placements[i].flags & TTM_PL_FLAG_VRAM) && - !(bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) && - (!max_offset || max_offset > - adev->mc.visible_vram_size)) { - if (WARN_ON_ONCE(min_offset > - adev->mc.visible_vram_size)) - return -EINVAL; - fpfn = min_offset >> PAGE_SHIFT; - lpfn = adev->mc.visible_vram_size >> PAGE_SHIFT; - } else { - fpfn = min_offset >> PAGE_SHIFT; - lpfn = max_offset >> PAGE_SHIFT; - } + unsigned fpfn, lpfn; + + fpfn = min_offset >> PAGE_SHIFT; + lpfn = max_offset >> PAGE_SHIFT; + if (fpfn > bo->placements[i].fpfn) bo->placements[i].fpfn = fpfn; if (!bo->placements[i].lpfn || @@ -724,15 +687,16 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, dev_err(adev->dev, "%p pin failed\n", bo); goto error; } - r = amdgpu_ttm_bind(&bo->tbo, &bo->tbo.mem); - if (unlikely(r)) { - dev_err(adev->dev, "%p bind failed\n", bo); - goto error; - } bo->pin_count = 1; - if (gpu_addr != NULL) + if (gpu_addr != NULL) { + r = amdgpu_ttm_bind(&bo->tbo, &bo->tbo.mem); + if (unlikely(r)) { + dev_err(adev->dev, "%p bind failed\n", bo); + goto error; + } *gpu_addr = amdgpu_bo_gpu_offset(bo); + } if (domain == AMDGPU_GEM_DOMAIN_VRAM) { adev->vram_pin_size += amdgpu_bo_size(bo); if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) @@ -918,8 +882,10 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, if (!amdgpu_ttm_bo_is_amdgpu_bo(bo)) return; - abo = container_of(bo, struct amdgpu_bo, tbo); - amdgpu_vm_bo_invalidate(adev, abo); + abo = ttm_to_amdgpu_bo(bo); + amdgpu_vm_bo_invalidate(adev, abo, evict); + + amdgpu_bo_kunmap(abo); /* remember the eviction */ if (evict) @@ -930,8 +896,6 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, return; /* move_notify is called before move happens */ - amdgpu_update_memory_usage(adev, &bo->mem, new_mem); - trace_amdgpu_ttm_bo_move(abo, new_mem->mem_type, old_mem->mem_type); } @@ -939,19 +903,22 @@ int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct amdgpu_bo *abo; - unsigned long offset, size, lpfn; - int i, r; + unsigned long offset, size; + int r; if (!amdgpu_ttm_bo_is_amdgpu_bo(bo)) return 0; - abo = container_of(bo, struct amdgpu_bo, tbo); + abo = ttm_to_amdgpu_bo(bo); + + /* Remember that this BO was accessed by the CPU */ + abo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; + if (bo->mem.mem_type != TTM_PL_VRAM) return 0; size = bo->mem.num_pages << PAGE_SHIFT; offset = bo->mem.start << PAGE_SHIFT; - /* TODO: figure out how to map scattered VRAM to the CPU */ if ((offset + size) <= adev->mc.visible_vram_size) return 0; @@ -961,26 +928,21 @@ int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) /* hurrah the memory is not visible ! */ atomic64_inc(&adev->num_vram_cpu_page_faults); - amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_VRAM); - lpfn = adev->mc.visible_vram_size >> PAGE_SHIFT; - for (i = 0; i < abo->placement.num_placement; i++) { - /* Force into visible VRAM */ - if ((abo->placements[i].flags & TTM_PL_FLAG_VRAM) && - (!abo->placements[i].lpfn || - abo->placements[i].lpfn > lpfn)) - abo->placements[i].lpfn = lpfn; - } + amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_VRAM | + AMDGPU_GEM_DOMAIN_GTT); + + /* Avoid costly evictions; only set GTT as a busy placement */ + abo->placement.num_busy_placement = 1; + abo->placement.busy_placement = &abo->placements[1]; + r = ttm_bo_validate(bo, &abo->placement, false, false); - if (unlikely(r == -ENOMEM)) { - amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT); - return ttm_bo_validate(bo, &abo->placement, false, false); - } else if (unlikely(r != 0)) { + if (unlikely(r != 0)) return r; - } offset = bo->mem.start << PAGE_SHIFT; /* this should never happen */ - if ((offset + size) > adev->mc.visible_vram_size) + if (bo->mem.mem_type == TTM_PL_VRAM && + (offset + size) > adev->mc.visible_vram_size) return -EINVAL; return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 382485115b06..428aae048f4b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -33,6 +33,72 @@ #define AMDGPU_BO_INVALID_OFFSET LONG_MAX +/* bo virtual addresses in a vm */ +struct amdgpu_bo_va_mapping { + struct amdgpu_bo_va *bo_va; + struct list_head list; + struct rb_node rb; + uint64_t start; + uint64_t last; + uint64_t __subtree_last; + uint64_t offset; + uint64_t flags; +}; + +/* User space allocated BO in a VM */ +struct amdgpu_bo_va { + struct amdgpu_vm_bo_base base; + + /* protected by bo being reserved */ + unsigned ref_count; + + /* all other members protected by the VM PD being reserved */ + struct dma_fence *last_pt_update; + + /* mappings for this bo_va */ + struct list_head invalids; + struct list_head valids; + + /* If the mappings are cleared or filled */ + bool cleared; +}; + +struct amdgpu_bo { + /* Protected by tbo.reserved */ + u32 preferred_domains; + u32 allowed_domains; + struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1]; + struct ttm_placement placement; + struct ttm_buffer_object tbo; + struct ttm_bo_kmap_obj kmap; + u64 flags; + unsigned pin_count; + u64 tiling_flags; + u64 metadata_flags; + void *metadata; + u32 metadata_size; + unsigned prime_shared_count; + /* list of all virtual address to which this bo is associated to */ + struct list_head va; + /* Constant after initialization */ + struct drm_gem_object gem_base; + struct amdgpu_bo *parent; + struct amdgpu_bo *shadow; + + struct ttm_bo_kmap_obj dma_buf_vmap; + struct amdgpu_mn *mn; + + union { + struct list_head mn_list; + struct list_head shadow_list; + }; +}; + +static inline struct amdgpu_bo *ttm_to_amdgpu_bo(struct ttm_buffer_object *tbo) +{ + return container_of(tbo, struct amdgpu_bo, tbo); +} + /** * amdgpu_mem_type_to_domain - return domain corresponding to mem_type * @mem_type: ttm memory type @@ -120,7 +186,19 @@ static inline u64 amdgpu_bo_mmap_offset(struct amdgpu_bo *bo) */ static inline bool amdgpu_bo_gpu_accessible(struct amdgpu_bo *bo) { - return bo->tbo.mem.mem_type != TTM_PL_SYSTEM; + switch (bo->tbo.mem.mem_type) { + case TTM_PL_TT: return amdgpu_ttm_is_bound(bo->tbo.ttm); + case TTM_PL_VRAM: return true; + default: return false; + } +} + +/** + * amdgpu_bo_explicit_sync - return whether the bo is explicitly synced + */ +static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo) +{ + return bo->flags & AMDGPU_GEM_CREATE_EXPLICIT_SYNC; } int amdgpu_bo_create(struct amdgpu_device *adev, @@ -128,14 +206,12 @@ int amdgpu_bo_create(struct amdgpu_device *adev, bool kernel, u32 domain, u64 flags, struct sg_table *sg, struct reservation_object *resv, + uint64_t init_value, struct amdgpu_bo **bo_ptr); -int amdgpu_bo_create_restricted(struct amdgpu_device *adev, - unsigned long size, int byte_align, - bool kernel, u32 domain, u64 flags, - struct sg_table *sg, - struct ttm_placement *placement, - struct reservation_object *resv, - struct amdgpu_bo **bo_ptr); +int amdgpu_bo_create_reserved(struct amdgpu_device *adev, + unsigned long size, int align, + u32 domain, struct amdgpu_bo **bo_ptr, + u64 *gpu_addr, void **cpu_addr); int amdgpu_bo_create_kernel(struct amdgpu_device *adev, unsigned long size, int align, u32 domain, struct amdgpu_bo **bo_ptr, @@ -143,6 +219,7 @@ int amdgpu_bo_create_kernel(struct amdgpu_device *adev, void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr, void **cpu_addr); int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr); +void *amdgpu_bo_kptr(struct amdgpu_bo *bo); void amdgpu_bo_kunmap(struct amdgpu_bo *bo); struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo); void amdgpu_bo_unref(struct amdgpu_bo **bo); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c index 7df503aedb69..6c570d4e4516 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c @@ -64,17 +64,13 @@ static const struct cg_flag_name clocks[] = { void amdgpu_pm_acpi_event_handler(struct amdgpu_device *adev) { - if (adev->pp_enabled) - /* TODO */ - return; - if (adev->pm.dpm_enabled) { mutex_lock(&adev->pm.mutex); if (power_supply_is_system_supplied() > 0) adev->pm.dpm.ac_power = true; else adev->pm.dpm.ac_power = false; - if (adev->pm.funcs->enable_bapm) + if (adev->powerplay.pp_funcs->enable_bapm) amdgpu_dpm_enable_bapm(adev, adev->pm.dpm.ac_power); mutex_unlock(&adev->pm.mutex); } @@ -88,9 +84,9 @@ static ssize_t amdgpu_get_dpm_state(struct device *dev, struct amdgpu_device *adev = ddev->dev_private; enum amd_pm_state_type pm; - if (adev->pp_enabled) { + if (adev->powerplay.pp_funcs->get_current_power_state) pm = amdgpu_dpm_get_current_power_state(adev); - } else + else pm = adev->pm.dpm.user_state; return snprintf(buf, PAGE_SIZE, "%s\n", @@ -118,8 +114,8 @@ static ssize_t amdgpu_set_dpm_state(struct device *dev, goto fail; } - if (adev->pp_enabled) { - amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_ENABLE_USER_STATE, &state, NULL); + if (adev->powerplay.pp_funcs->dispatch_tasks) { + amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_ENABLE_USER_STATE, &state, NULL); } else { mutex_lock(&adev->pm.mutex); adev->pm.dpm.user_state = state; @@ -140,13 +136,17 @@ static ssize_t amdgpu_get_dpm_forced_performance_level(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - enum amd_dpm_forced_level level; + enum amd_dpm_forced_level level = 0xff; if ((adev->flags & AMD_IS_PX) && (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) return snprintf(buf, PAGE_SIZE, "off\n"); - level = amdgpu_dpm_get_performance_level(adev); + if (adev->powerplay.pp_funcs->get_performance_level) + level = amdgpu_dpm_get_performance_level(adev); + else + level = adev->pm.dpm.forced_level; + return snprintf(buf, PAGE_SIZE, "%s\n", (level == AMD_DPM_FORCED_LEVEL_AUTO) ? "auto" : (level == AMD_DPM_FORCED_LEVEL_LOW) ? "low" : @@ -167,7 +167,7 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev, struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; enum amd_dpm_forced_level level; - enum amd_dpm_forced_level current_level; + enum amd_dpm_forced_level current_level = 0xff; int ret = 0; /* Can't force performance level when the card is off */ @@ -175,7 +175,8 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev, (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) return -EINVAL; - current_level = amdgpu_dpm_get_performance_level(adev); + if (adev->powerplay.pp_funcs->get_performance_level) + current_level = amdgpu_dpm_get_performance_level(adev); if (strncmp("low", buf, strlen("low")) == 0) { level = AMD_DPM_FORCED_LEVEL_LOW; @@ -203,9 +204,7 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev, if (current_level == level) return count; - if (adev->pp_enabled) - amdgpu_dpm_force_performance_level(adev, level); - else { + if (adev->powerplay.pp_funcs->force_performance_level) { mutex_lock(&adev->pm.mutex); if (adev->pm.dpm.thermal_active) { count = -EINVAL; @@ -233,7 +232,7 @@ static ssize_t amdgpu_get_pp_num_states(struct device *dev, struct pp_states_info data; int i, buf_len; - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->get_pp_num_states) amdgpu_dpm_get_pp_num_states(adev, &data); buf_len = snprintf(buf, PAGE_SIZE, "states: %d\n", data.nums); @@ -257,8 +256,8 @@ static ssize_t amdgpu_get_pp_cur_state(struct device *dev, enum amd_pm_state_type pm = 0; int i = 0; - if (adev->pp_enabled) { - + if (adev->powerplay.pp_funcs->get_current_power_state + && adev->powerplay.pp_funcs->get_pp_num_states) { pm = amdgpu_dpm_get_current_power_state(adev); amdgpu_dpm_get_pp_num_states(adev, &data); @@ -280,25 +279,10 @@ static ssize_t amdgpu_get_pp_force_state(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - struct pp_states_info data; - enum amd_pm_state_type pm = 0; - int i; - - if (adev->pp_force_state_enabled && adev->pp_enabled) { - pm = amdgpu_dpm_get_current_power_state(adev); - amdgpu_dpm_get_pp_num_states(adev, &data); - for (i = 0; i < data.nums; i++) { - if (pm == data.states[i]) - break; - } - - if (i == data.nums) - i = -EINVAL; - - return snprintf(buf, PAGE_SIZE, "%d\n", i); - - } else + if (adev->pp_force_state_enabled) + return amdgpu_get_pp_cur_state(dev, attr, buf); + else return snprintf(buf, PAGE_SIZE, "\n"); } @@ -315,7 +299,8 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev, if (strlen(buf) == 1) adev->pp_force_state_enabled = false; - else if (adev->pp_enabled) { + else if (adev->powerplay.pp_funcs->dispatch_tasks && + adev->powerplay.pp_funcs->get_pp_num_states) { struct pp_states_info data; ret = kstrtoul(buf, 0, &idx); @@ -330,7 +315,7 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev, if (state != POWER_STATE_TYPE_INTERNAL_BOOT && state != POWER_STATE_TYPE_DEFAULT) { amdgpu_dpm_dispatch_task(adev, - AMD_PP_EVENT_ENABLE_USER_STATE, &state, NULL); + AMD_PP_TASK_ENABLE_USER_STATE, &state, NULL); adev->pp_force_state_enabled = true; } } @@ -347,7 +332,7 @@ static ssize_t amdgpu_get_pp_table(struct device *dev, char *table = NULL; int size; - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->get_pp_table) size = amdgpu_dpm_get_pp_table(adev, &table); else return 0; @@ -368,7 +353,7 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->set_pp_table) amdgpu_dpm_set_pp_table(adev, buf, count); return count; @@ -380,14 +365,11 @@ static ssize_t amdgpu_get_pp_dpm_sclk(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - ssize_t size = 0; - - if (adev->pp_enabled) - size = amdgpu_dpm_print_clock_levels(adev, PP_SCLK, buf); - else if (adev->pm.funcs->print_clock_levels) - size = adev->pm.funcs->print_clock_levels(adev, PP_SCLK, buf); - return size; + if (adev->powerplay.pp_funcs->print_clock_levels) + return amdgpu_dpm_print_clock_levels(adev, PP_SCLK, buf); + else + return snprintf(buf, PAGE_SIZE, "\n"); } static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev, @@ -416,10 +398,9 @@ static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev, mask |= 1 << level; } - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->force_clock_level) amdgpu_dpm_force_clock_level(adev, PP_SCLK, mask); - else if (adev->pm.funcs->force_clock_level) - adev->pm.funcs->force_clock_level(adev, PP_SCLK, mask); + fail: return count; } @@ -430,14 +411,11 @@ static ssize_t amdgpu_get_pp_dpm_mclk(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - ssize_t size = 0; - - if (adev->pp_enabled) - size = amdgpu_dpm_print_clock_levels(adev, PP_MCLK, buf); - else if (adev->pm.funcs->print_clock_levels) - size = adev->pm.funcs->print_clock_levels(adev, PP_MCLK, buf); - return size; + if (adev->powerplay.pp_funcs->print_clock_levels) + return amdgpu_dpm_print_clock_levels(adev, PP_MCLK, buf); + else + return snprintf(buf, PAGE_SIZE, "\n"); } static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev, @@ -465,11 +443,9 @@ static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev, } mask |= 1 << level; } - - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->force_clock_level) amdgpu_dpm_force_clock_level(adev, PP_MCLK, mask); - else if (adev->pm.funcs->force_clock_level) - adev->pm.funcs->force_clock_level(adev, PP_MCLK, mask); + fail: return count; } @@ -480,14 +456,11 @@ static ssize_t amdgpu_get_pp_dpm_pcie(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - ssize_t size = 0; - - if (adev->pp_enabled) - size = amdgpu_dpm_print_clock_levels(adev, PP_PCIE, buf); - else if (adev->pm.funcs->print_clock_levels) - size = adev->pm.funcs->print_clock_levels(adev, PP_PCIE, buf); - return size; + if (adev->powerplay.pp_funcs->print_clock_levels) + return amdgpu_dpm_print_clock_levels(adev, PP_PCIE, buf); + else + return snprintf(buf, PAGE_SIZE, "\n"); } static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev, @@ -515,11 +488,9 @@ static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev, } mask |= 1 << level; } - - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->force_clock_level) amdgpu_dpm_force_clock_level(adev, PP_PCIE, mask); - else if (adev->pm.funcs->force_clock_level) - adev->pm.funcs->force_clock_level(adev, PP_PCIE, mask); + fail: return count; } @@ -532,10 +503,8 @@ static ssize_t amdgpu_get_pp_sclk_od(struct device *dev, struct amdgpu_device *adev = ddev->dev_private; uint32_t value = 0; - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->get_sclk_od) value = amdgpu_dpm_get_sclk_od(adev); - else if (adev->pm.funcs->get_sclk_od) - value = adev->pm.funcs->get_sclk_od(adev); return snprintf(buf, PAGE_SIZE, "%d\n", value); } @@ -556,12 +525,12 @@ static ssize_t amdgpu_set_pp_sclk_od(struct device *dev, count = -EINVAL; goto fail; } - - if (adev->pp_enabled) { + if (adev->powerplay.pp_funcs->set_sclk_od) amdgpu_dpm_set_sclk_od(adev, (uint32_t)value); - amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_READJUST_POWER_STATE, NULL, NULL); - } else if (adev->pm.funcs->set_sclk_od) { - adev->pm.funcs->set_sclk_od(adev, (uint32_t)value); + + if (adev->powerplay.pp_funcs->dispatch_tasks) { + amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL); + } else { adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps; amdgpu_pm_compute_clocks(adev); } @@ -578,10 +547,8 @@ static ssize_t amdgpu_get_pp_mclk_od(struct device *dev, struct amdgpu_device *adev = ddev->dev_private; uint32_t value = 0; - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->get_mclk_od) value = amdgpu_dpm_get_mclk_od(adev); - else if (adev->pm.funcs->get_mclk_od) - value = adev->pm.funcs->get_mclk_od(adev); return snprintf(buf, PAGE_SIZE, "%d\n", value); } @@ -602,12 +569,12 @@ static ssize_t amdgpu_set_pp_mclk_od(struct device *dev, count = -EINVAL; goto fail; } - - if (adev->pp_enabled) { + if (adev->powerplay.pp_funcs->set_mclk_od) amdgpu_dpm_set_mclk_od(adev, (uint32_t)value); - amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_READJUST_POWER_STATE, NULL, NULL); - } else if (adev->pm.funcs->set_mclk_od) { - adev->pm.funcs->set_mclk_od(adev, (uint32_t)value); + + if (adev->powerplay.pp_funcs->dispatch_tasks) { + amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL); + } else { adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps; amdgpu_pm_compute_clocks(adev); } @@ -621,14 +588,11 @@ static ssize_t amdgpu_get_pp_power_profile(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - int ret = 0; + int ret = 0xff; - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->get_power_profile_state) ret = amdgpu_dpm_get_power_profile_state( adev, query); - else if (adev->pm.funcs->get_power_profile_state) - ret = adev->pm.funcs->get_power_profile_state( - adev, query); if (ret) return ret; @@ -675,15 +639,12 @@ static ssize_t amdgpu_set_pp_power_profile(struct device *dev, char *sub_str, buf_cpy[128], *tmp_str; const char delimiter[3] = {' ', '\n', '\0'}; long int value; - int ret = 0; + int ret = 0xff; if (strncmp("reset", buf, strlen("reset")) == 0) { - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->reset_power_profile_state) ret = amdgpu_dpm_reset_power_profile_state( adev, request); - else if (adev->pm.funcs->reset_power_profile_state) - ret = adev->pm.funcs->reset_power_profile_state( - adev, request); if (ret) { count = -EINVAL; goto fail; @@ -692,12 +653,10 @@ static ssize_t amdgpu_set_pp_power_profile(struct device *dev, } if (strncmp("set", buf, strlen("set")) == 0) { - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->set_power_profile_state) ret = amdgpu_dpm_set_power_profile_state( adev, request); - else if (adev->pm.funcs->set_power_profile_state) - ret = adev->pm.funcs->set_power_profile_state( - adev, request); + if (ret) { count = -EINVAL; goto fail; @@ -745,13 +704,8 @@ static ssize_t amdgpu_set_pp_power_profile(struct device *dev, loop++; } - - if (adev->pp_enabled) - ret = amdgpu_dpm_set_power_profile_state( - adev, request); - else if (adev->pm.funcs->set_power_profile_state) - ret = adev->pm.funcs->set_power_profile_state( - adev, request); + if (adev->powerplay.pp_funcs->set_power_profile_state) + ret = amdgpu_dpm_set_power_profile_state(adev, request); if (ret) count = -EINVAL; @@ -831,7 +785,7 @@ static ssize_t amdgpu_hwmon_show_temp(struct device *dev, (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) return -EINVAL; - if (!adev->pp_enabled && !adev->pm.funcs->get_temperature) + if (!adev->powerplay.pp_funcs->get_temperature) temp = 0; else temp = amdgpu_dpm_get_temperature(adev); @@ -862,7 +816,7 @@ static ssize_t amdgpu_hwmon_get_pwm1_enable(struct device *dev, struct amdgpu_device *adev = dev_get_drvdata(dev); u32 pwm_mode = 0; - if (!adev->pp_enabled && !adev->pm.funcs->get_fan_control_mode) + if (!adev->powerplay.pp_funcs->get_fan_control_mode) return -EINVAL; pwm_mode = amdgpu_dpm_get_fan_control_mode(adev); @@ -879,7 +833,7 @@ static ssize_t amdgpu_hwmon_set_pwm1_enable(struct device *dev, int err; int value; - if (!adev->pp_enabled && !adev->pm.funcs->set_fan_control_mode) + if (!adev->powerplay.pp_funcs->set_fan_control_mode) return -EINVAL; err = kstrtoint(buf, 10, &value); @@ -919,9 +873,11 @@ static ssize_t amdgpu_hwmon_set_pwm1(struct device *dev, value = (value * 100) / 255; - err = amdgpu_dpm_set_fan_speed_percent(adev, value); - if (err) - return err; + if (adev->powerplay.pp_funcs->set_fan_speed_percent) { + err = amdgpu_dpm_set_fan_speed_percent(adev, value); + if (err) + return err; + } return count; } @@ -932,11 +888,13 @@ static ssize_t amdgpu_hwmon_get_pwm1(struct device *dev, { struct amdgpu_device *adev = dev_get_drvdata(dev); int err; - u32 speed; + u32 speed = 0; - err = amdgpu_dpm_get_fan_speed_percent(adev, &speed); - if (err) - return err; + if (adev->powerplay.pp_funcs->get_fan_speed_percent) { + err = amdgpu_dpm_get_fan_speed_percent(adev, &speed); + if (err) + return err; + } speed = (speed * 255) / 100; @@ -949,11 +907,13 @@ static ssize_t amdgpu_hwmon_get_fan1_input(struct device *dev, { struct amdgpu_device *adev = dev_get_drvdata(dev); int err; - u32 speed; + u32 speed = 0; - err = amdgpu_dpm_get_fan_speed_rpm(adev, &speed); - if (err) - return err; + if (adev->powerplay.pp_funcs->get_fan_speed_rpm) { + err = amdgpu_dpm_get_fan_speed_rpm(adev, &speed); + if (err) + return err; + } return sprintf(buf, "%i\n", speed); } @@ -986,6 +946,10 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, struct amdgpu_device *adev = dev_get_drvdata(dev); umode_t effective_mode = attr->mode; + /* no skipping for powerplay */ + if (adev->powerplay.cgs_device) + return effective_mode; + /* Skip limit attributes if DPM is not enabled */ if (!adev->pm.dpm_enabled && (attr == &sensor_dev_attr_temp1_crit.dev_attr.attr || @@ -996,9 +960,6 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, attr == &sensor_dev_attr_pwm1_min.dev_attr.attr)) return 0; - if (adev->pp_enabled) - return effective_mode; - /* Skip fan attributes if fan is not present */ if (adev->pm.no_fan && (attr == &sensor_dev_attr_pwm1.dev_attr.attr || @@ -1008,21 +969,21 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, return 0; /* mask fan attributes if we have no bindings for this asic to expose */ - if ((!adev->pm.funcs->get_fan_speed_percent && + if ((!adev->powerplay.pp_funcs->get_fan_speed_percent && attr == &sensor_dev_attr_pwm1.dev_attr.attr) || /* can't query fan */ - (!adev->pm.funcs->get_fan_control_mode && + (!adev->powerplay.pp_funcs->get_fan_control_mode && attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr)) /* can't query state */ effective_mode &= ~S_IRUGO; - if ((!adev->pm.funcs->set_fan_speed_percent && + if ((!adev->powerplay.pp_funcs->set_fan_speed_percent && attr == &sensor_dev_attr_pwm1.dev_attr.attr) || /* can't manage fan */ - (!adev->pm.funcs->set_fan_control_mode && + (!adev->powerplay.pp_funcs->set_fan_control_mode && attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr)) /* can't manage state */ effective_mode &= ~S_IWUSR; /* hide max/min values if we can't both query and manage the fan */ - if ((!adev->pm.funcs->set_fan_speed_percent && - !adev->pm.funcs->get_fan_speed_percent) && + if ((!adev->powerplay.pp_funcs->set_fan_speed_percent && + !adev->powerplay.pp_funcs->get_fan_speed_percent) && (attr == &sensor_dev_attr_pwm1_max.dev_attr.attr || attr == &sensor_dev_attr_pwm1_min.dev_attr.attr)) return 0; @@ -1055,7 +1016,7 @@ void amdgpu_dpm_thermal_work_handler(struct work_struct *work) if (!adev->pm.dpm_enabled) return; - if (adev->pm.funcs->get_temperature) { + if (adev->powerplay.pp_funcs->get_temperature) { int temp = amdgpu_dpm_get_temperature(adev); if (temp < adev->pm.dpm.thermal.min_temp) @@ -1087,7 +1048,7 @@ static struct amdgpu_ps *amdgpu_dpm_pick_power_state(struct amdgpu_device *adev, true : false; /* check if the vblank period is too short to adjust the mclk */ - if (single_display && adev->pm.funcs->vblank_too_short) { + if (single_display && adev->powerplay.pp_funcs->vblank_too_short) { if (amdgpu_dpm_vblank_too_short(adev)) single_display = false; } @@ -1216,7 +1177,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev) struct amdgpu_ps *ps; enum amd_pm_state_type dpm_state; int ret; - bool equal; + bool equal = false; /* if dpm init failed */ if (!adev->pm.dpm_enabled) @@ -1236,7 +1197,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev) else return; - if (amdgpu_dpm == 1) { + if (amdgpu_dpm == 1 && adev->powerplay.pp_funcs->print_power_state) { printk("switching from power state:\n"); amdgpu_dpm_print_power_state(adev, adev->pm.dpm.current_ps); printk("switching to power state:\n"); @@ -1245,15 +1206,17 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev) /* update whether vce is active */ ps->vce_active = adev->pm.dpm.vce_active; - - amdgpu_dpm_display_configuration_changed(adev); + if (adev->powerplay.pp_funcs->display_configuration_changed) + amdgpu_dpm_display_configuration_changed(adev); ret = amdgpu_dpm_pre_set_power_state(adev); if (ret) return; - if ((0 != amgdpu_dpm_check_state_equal(adev, adev->pm.dpm.current_ps, adev->pm.dpm.requested_ps, &equal))) - equal = false; + if (adev->powerplay.pp_funcs->check_state_equal) { + if (0 != amdgpu_dpm_check_state_equal(adev, adev->pm.dpm.current_ps, adev->pm.dpm.requested_ps, &equal)) + equal = false; + } if (equal) return; @@ -1264,7 +1227,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev) adev->pm.dpm.current_active_crtcs = adev->pm.dpm.new_active_crtcs; adev->pm.dpm.current_active_crtc_count = adev->pm.dpm.new_active_crtc_count; - if (adev->pm.funcs->force_performance_level) { + if (adev->powerplay.pp_funcs->force_performance_level) { if (adev->pm.dpm.thermal_active) { enum amd_dpm_forced_level level = adev->pm.dpm.forced_level; /* force low perf level for thermal */ @@ -1280,7 +1243,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev) void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable) { - if (adev->pp_enabled || adev->pm.funcs->powergate_uvd) { + if (adev->powerplay.pp_funcs->powergate_uvd) { /* enable/disable UVD */ mutex_lock(&adev->pm.mutex); amdgpu_dpm_powergate_uvd(adev, !enable); @@ -1302,7 +1265,7 @@ void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable) void amdgpu_dpm_enable_vce(struct amdgpu_device *adev, bool enable) { - if (adev->pp_enabled || adev->pm.funcs->powergate_vce) { + if (adev->powerplay.pp_funcs->powergate_vce) { /* enable/disable VCE */ mutex_lock(&adev->pm.mutex); amdgpu_dpm_powergate_vce(adev, !enable); @@ -1337,8 +1300,7 @@ void amdgpu_pm_print_power_states(struct amdgpu_device *adev) { int i; - if (adev->pp_enabled) - /* TO DO */ + if (adev->powerplay.pp_funcs->print_power_state == NULL) return; for (i = 0; i < adev->pm.dpm.num_ps; i++) @@ -1353,10 +1315,11 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) if (adev->pm.sysfs_initialized) return 0; - if (!adev->pp_enabled) { - if (adev->pm.funcs->get_temperature == NULL) - return 0; - } + if (adev->pm.dpm_enabled == 0) + return 0; + + if (adev->powerplay.pp_funcs->get_temperature == NULL) + return 0; adev->pm.int_hwmon_dev = hwmon_device_register_with_groups(adev->dev, DRIVER_NAME, adev, @@ -1379,27 +1342,26 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) return ret; } - if (adev->pp_enabled) { - ret = device_create_file(adev->dev, &dev_attr_pp_num_states); - if (ret) { - DRM_ERROR("failed to create device file pp_num_states\n"); - return ret; - } - ret = device_create_file(adev->dev, &dev_attr_pp_cur_state); - if (ret) { - DRM_ERROR("failed to create device file pp_cur_state\n"); - return ret; - } - ret = device_create_file(adev->dev, &dev_attr_pp_force_state); - if (ret) { - DRM_ERROR("failed to create device file pp_force_state\n"); - return ret; - } - ret = device_create_file(adev->dev, &dev_attr_pp_table); - if (ret) { - DRM_ERROR("failed to create device file pp_table\n"); - return ret; - } + + ret = device_create_file(adev->dev, &dev_attr_pp_num_states); + if (ret) { + DRM_ERROR("failed to create device file pp_num_states\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_pp_cur_state); + if (ret) { + DRM_ERROR("failed to create device file pp_cur_state\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_pp_force_state); + if (ret) { + DRM_ERROR("failed to create device file pp_force_state\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_pp_table); + if (ret) { + DRM_ERROR("failed to create device file pp_table\n"); + return ret; } ret = device_create_file(adev->dev, &dev_attr_pp_dpm_sclk); @@ -1455,16 +1417,19 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) void amdgpu_pm_sysfs_fini(struct amdgpu_device *adev) { + if (adev->pm.dpm_enabled == 0) + return; + if (adev->pm.int_hwmon_dev) hwmon_device_unregister(adev->pm.int_hwmon_dev); device_remove_file(adev->dev, &dev_attr_power_dpm_state); device_remove_file(adev->dev, &dev_attr_power_dpm_force_performance_level); - if (adev->pp_enabled) { - device_remove_file(adev->dev, &dev_attr_pp_num_states); - device_remove_file(adev->dev, &dev_attr_pp_cur_state); - device_remove_file(adev->dev, &dev_attr_pp_force_state); - device_remove_file(adev->dev, &dev_attr_pp_table); - } + + device_remove_file(adev->dev, &dev_attr_pp_num_states); + device_remove_file(adev->dev, &dev_attr_pp_cur_state); + device_remove_file(adev->dev, &dev_attr_pp_force_state); + device_remove_file(adev->dev, &dev_attr_pp_table); + device_remove_file(adev->dev, &dev_attr_pp_dpm_sclk); device_remove_file(adev->dev, &dev_attr_pp_dpm_mclk); device_remove_file(adev->dev, &dev_attr_pp_dpm_pcie); @@ -1495,8 +1460,8 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev) amdgpu_fence_wait_empty(ring); } - if (adev->pp_enabled) { - amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_DISPLAY_CONFIG_CHANGE, NULL, NULL); + if (adev->powerplay.pp_funcs->dispatch_tasks) { + amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_DISPLAY_CONFIG_CHANGE, NULL, NULL); } else { mutex_lock(&adev->pm.mutex); adev->pm.dpm.new_active_crtcs = 0; @@ -1505,7 +1470,7 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev) list_for_each_entry(crtc, &ddev->mode_config.crtc_list, head) { amdgpu_crtc = to_amdgpu_crtc(crtc); - if (crtc->enabled) { + if (amdgpu_crtc->enabled) { adev->pm.dpm.new_active_crtcs |= (1 << amdgpu_crtc->crtc_id); adev->pm.dpm.new_active_crtc_count++; } @@ -1630,15 +1595,15 @@ static int amdgpu_debugfs_pm_info(struct seq_file *m, void *data) if ((adev->flags & AMD_IS_PX) && (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) { seq_printf(m, "PX asic powered off\n"); - } else if (adev->pp_enabled) { - return amdgpu_debugfs_pm_info_pp(m, adev); - } else { + } else if (adev->powerplay.pp_funcs->debugfs_print_current_performance_level) { mutex_lock(&adev->pm.mutex); - if (adev->pm.funcs->debugfs_print_current_performance_level) - adev->pm.funcs->debugfs_print_current_performance_level(adev, m); + if (adev->powerplay.pp_funcs->debugfs_print_current_performance_level) + adev->powerplay.pp_funcs->debugfs_print_current_performance_level(adev, m); else seq_printf(m, "Debugfs support not implemented for this asic\n"); mutex_unlock(&adev->pm.mutex); + } else { + return amdgpu_debugfs_pm_info_pp(m, adev); } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.h index c19c4d138751..f21a7716b90e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.h @@ -30,6 +30,7 @@ struct cg_flag_name const char *name; }; +void amdgpu_pm_acpi_event_handler(struct amdgpu_device *adev); int amdgpu_pm_sysfs_init(struct amdgpu_device *adev); void amdgpu_pm_sysfs_fini(struct amdgpu_device *adev); void amdgpu_pm_print_power_states(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c index b7e1c026c0c8..033fba2def6f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c @@ -34,24 +34,6 @@ #include "cik_dpm.h" #include "vi_dpm.h" -static int amdgpu_create_pp_handle(struct amdgpu_device *adev) -{ - struct amd_pp_init pp_init; - struct amd_powerplay *amd_pp; - int ret; - - amd_pp = &(adev->powerplay); - pp_init.chip_family = adev->family; - pp_init.chip_id = adev->asic_type; - pp_init.pm_en = (amdgpu_dpm != 0 && !amdgpu_sriov_vf(adev)) ? true : false; - pp_init.feature_mask = amdgpu_pp_feature_mask; - pp_init.device = amdgpu_cgs_create_device(adev); - ret = amd_powerplay_create(&pp_init, &(amd_pp->pp_handle)); - if (ret) - return -EINVAL; - return 0; -} - static int amdgpu_pp_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -59,7 +41,6 @@ static int amdgpu_pp_early_init(void *handle) int ret = 0; amd_pp = &(adev->powerplay); - adev->pp_enabled = false; amd_pp->pp_handle = (void *)adev; switch (adev->asic_type) { @@ -73,9 +54,7 @@ static int amdgpu_pp_early_init(void *handle) case CHIP_STONEY: case CHIP_VEGA10: case CHIP_RAVEN: - adev->pp_enabled = true; - if (amdgpu_create_pp_handle(adev)) - return -EINVAL; + amd_pp->cgs_device = amdgpu_cgs_create_device(adev); amd_pp->ip_funcs = &pp_ip_funcs; amd_pp->pp_funcs = &pp_dpm_funcs; break; @@ -87,17 +66,26 @@ static int amdgpu_pp_early_init(void *handle) case CHIP_OLAND: case CHIP_HAINAN: amd_pp->ip_funcs = &si_dpm_ip_funcs; + amd_pp->pp_funcs = &si_dpm_funcs; break; #endif #ifdef CONFIG_DRM_AMDGPU_CIK case CHIP_BONAIRE: case CHIP_HAWAII: - amd_pp->ip_funcs = &ci_dpm_ip_funcs; + if (amdgpu_dpm == -1) { + amd_pp->ip_funcs = &ci_dpm_ip_funcs; + amd_pp->pp_funcs = &ci_dpm_funcs; + } else { + amd_pp->cgs_device = amdgpu_cgs_create_device(adev); + amd_pp->ip_funcs = &pp_ip_funcs; + amd_pp->pp_funcs = &pp_dpm_funcs; + } break; case CHIP_KABINI: case CHIP_MULLINS: case CHIP_KAVERI: amd_pp->ip_funcs = &kv_dpm_ip_funcs; + amd_pp->pp_funcs = &kv_dpm_funcs; break; #endif default: @@ -107,12 +95,9 @@ static int amdgpu_pp_early_init(void *handle) if (adev->powerplay.ip_funcs->early_init) ret = adev->powerplay.ip_funcs->early_init( - adev->powerplay.pp_handle); + amd_pp->cgs_device ? amd_pp->cgs_device : + amd_pp->pp_handle); - if (ret == PP_DPM_DISABLED) { - adev->pm.dpm_enabled = false; - return 0; - } return ret; } @@ -126,11 +111,6 @@ static int amdgpu_pp_late_init(void *handle) ret = adev->powerplay.ip_funcs->late_init( adev->powerplay.pp_handle); - if (adev->pp_enabled && adev->pm.dpm_enabled) { - amdgpu_pm_sysfs_init(adev); - amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_COMPLETE_INIT, NULL, NULL); - } - return ret; } @@ -165,21 +145,13 @@ static int amdgpu_pp_hw_init(void *handle) int ret = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->pp_enabled && adev->firmware.load_type == AMDGPU_FW_LOAD_SMU) + if (adev->firmware.load_type == AMDGPU_FW_LOAD_SMU) amdgpu_ucode_init_bo(adev); if (adev->powerplay.ip_funcs->hw_init) ret = adev->powerplay.ip_funcs->hw_init( adev->powerplay.pp_handle); - if (ret == PP_DPM_DISABLED) { - adev->pm.dpm_enabled = false; - return 0; - } - - if ((amdgpu_dpm != 0) && !amdgpu_sriov_vf(adev)) - adev->pm.dpm_enabled = true; - return ret; } @@ -188,16 +160,10 @@ static int amdgpu_pp_hw_fini(void *handle) int ret = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->pp_enabled && adev->pm.dpm_enabled) - amdgpu_pm_sysfs_fini(adev); - if (adev->powerplay.ip_funcs->hw_fini) ret = adev->powerplay.ip_funcs->hw_fini( adev->powerplay.pp_handle); - if (adev->pp_enabled && adev->firmware.load_type == AMDGPU_FW_LOAD_SMU) - amdgpu_ucode_fini_bo(adev); - return ret; } @@ -209,9 +175,8 @@ static void amdgpu_pp_late_fini(void *handle) adev->powerplay.ip_funcs->late_fini( adev->powerplay.pp_handle); - - if (adev->pp_enabled) - amd_powerplay_destroy(adev->powerplay.pp_handle); + if (adev->powerplay.cgs_device) + amdgpu_cgs_destroy_device(adev->powerplay.cgs_device); } static int amdgpu_pp_suspend(void *handle) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index 6bdc866570ab..ae9c106979d7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -57,6 +57,40 @@ void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr) ttm_bo_kunmap(&bo->dma_buf_vmap); } +int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) +{ + struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj); + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + unsigned asize = amdgpu_bo_size(bo); + int ret; + + if (!vma->vm_file) + return -ENODEV; + + if (adev == NULL) + return -ENODEV; + + /* Check for valid size. */ + if (asize < vma->vm_end - vma->vm_start) + return -EINVAL; + + if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) || + (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) { + return -EPERM; + } + vma->vm_pgoff += amdgpu_bo_mmap_offset(bo) >> PAGE_SHIFT; + + /* prime mmap does not need to check access, so allow here */ + ret = drm_vma_node_allow(&obj->vma_node, vma->vm_file->private_data); + if (ret) + return ret; + + ret = ttm_bo_mmap(vma->vm_file, vma, &adev->mman.bdev); + drm_vma_node_revoke(&obj->vma_node, vma->vm_file->private_data); + + return ret; +} + struct drm_gem_object * amdgpu_gem_prime_import_sg_table(struct drm_device *dev, struct dma_buf_attachment *attach, @@ -69,7 +103,7 @@ amdgpu_gem_prime_import_sg_table(struct drm_device *dev, ww_mutex_lock(&resv->lock, NULL); ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, false, - AMDGPU_GEM_DOMAIN_GTT, 0, sg, resv, &bo); + AMDGPU_GEM_DOMAIN_GTT, 0, sg, resv, 0, &bo); ww_mutex_unlock(&resv->lock); if (ret) return ERR_PTR(ret); @@ -135,9 +169,14 @@ struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev, int flags) { struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj); + struct dma_buf *buf; - if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) + if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) || + bo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) return ERR_PTR(-EPERM); - return drm_gem_prime_export(dev, gobj, flags); + buf = drm_gem_prime_export(dev, gobj, flags); + if (!IS_ERR(buf)) + buf->file->f_mapping = dev->anon_inode->i_mapping; + return buf; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index 4083be61b328..7714f4a6c8b0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -57,16 +57,23 @@ static int psp_sw_init(void *handle) psp->prep_cmd_buf = psp_v3_1_prep_cmd_buf; psp->ring_init = psp_v3_1_ring_init; psp->ring_create = psp_v3_1_ring_create; + psp->ring_stop = psp_v3_1_ring_stop; psp->ring_destroy = psp_v3_1_ring_destroy; psp->cmd_submit = psp_v3_1_cmd_submit; psp->compare_sram_data = psp_v3_1_compare_sram_data; psp->smu_reload_quirk = psp_v3_1_smu_reload_quirk; + psp->mode1_reset = psp_v3_1_mode1_reset; break; case CHIP_RAVEN: + psp->init_microcode = psp_v10_0_init_microcode; psp->prep_cmd_buf = psp_v10_0_prep_cmd_buf; psp->ring_init = psp_v10_0_ring_init; + psp->ring_create = psp_v10_0_ring_create; + psp->ring_stop = psp_v10_0_ring_stop; + psp->ring_destroy = psp_v10_0_ring_destroy; psp->cmd_submit = psp_v10_0_cmd_submit; psp->compare_sram_data = psp_v10_0_compare_sram_data; + psp->mode1_reset = psp_v10_0_mode1_reset; break; default: return -EINVAL; @@ -85,6 +92,12 @@ static int psp_sw_init(void *handle) static int psp_sw_fini(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + release_firmware(adev->psp.sos_fw); + adev->psp.sos_fw = NULL; + release_firmware(adev->psp.asd_fw); + adev->psp.asd_fw = NULL; return 0; } @@ -95,9 +108,8 @@ int psp_wait_for(struct psp_context *psp, uint32_t reg_index, int i; struct amdgpu_device *adev = psp->adev; - val = RREG32(reg_index); - for (i = 0; i < adev->usec_timeout; i++) { + val = RREG32(reg_index); if (check_changed) { if (val != reg_val) return 0; @@ -118,33 +130,18 @@ psp_cmd_submit_buf(struct psp_context *psp, int index) { int ret; - struct amdgpu_bo *cmd_buf_bo; - uint64_t cmd_buf_mc_addr; - struct psp_gfx_cmd_resp *cmd_buf_mem; - struct amdgpu_device *adev = psp->adev; - - ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &cmd_buf_bo, &cmd_buf_mc_addr, - (void **)&cmd_buf_mem); - if (ret) - return ret; - memset(cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE); + memset(psp->cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE); - memcpy(cmd_buf_mem, cmd, sizeof(struct psp_gfx_cmd_resp)); + memcpy(psp->cmd_buf_mem, cmd, sizeof(struct psp_gfx_cmd_resp)); - ret = psp_cmd_submit(psp, ucode, cmd_buf_mc_addr, + ret = psp_cmd_submit(psp, ucode, psp->cmd_buf_mc_addr, fence_mc_addr, index); while (*((unsigned int *)psp->fence_buf) != index) { msleep(1); } - amdgpu_bo_free_kernel(&cmd_buf_bo, - &cmd_buf_mc_addr, - (void **)&cmd_buf_mem); - return ret; } @@ -264,15 +261,18 @@ static int psp_asd_load(struct psp_context *psp) static int psp_hw_start(struct psp_context *psp) { + struct amdgpu_device *adev = psp->adev; int ret; - ret = psp_bootloader_load_sysdrv(psp); - if (ret) - return ret; + if (!amdgpu_sriov_vf(adev) || !adev->in_sriov_reset) { + ret = psp_bootloader_load_sysdrv(psp); + if (ret) + return ret; - ret = psp_bootloader_load_sos(psp); - if (ret) - return ret; + ret = psp_bootloader_load_sos(psp); + if (ret) + return ret; + } ret = psp_ring_create(psp, PSP_RING_TYPE__KM); if (ret) @@ -352,13 +352,20 @@ static int psp_load_fw(struct amdgpu_device *adev) &psp->fence_buf_mc_addr, &psp->fence_buf); if (ret) + goto failed_mem2; + + ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &psp->cmd_buf_bo, &psp->cmd_buf_mc_addr, + (void **)&psp->cmd_buf_mem); + if (ret) goto failed_mem1; memset(psp->fence_buf, 0, PSP_FENCE_BUFFER_SIZE); ret = psp_ring_init(psp, PSP_RING_TYPE__KM); if (ret) - goto failed_mem1; + goto failed_mem; ret = psp_tmr_init(psp); if (ret) @@ -379,9 +386,13 @@ static int psp_load_fw(struct amdgpu_device *adev) return 0; failed_mem: + amdgpu_bo_free_kernel(&psp->cmd_buf_bo, + &psp->cmd_buf_mc_addr, + (void **)&psp->cmd_buf_mem); +failed_mem1: amdgpu_bo_free_kernel(&psp->fence_buf_bo, &psp->fence_buf_mc_addr, &psp->fence_buf); -failed_mem1: +failed_mem2: amdgpu_bo_free_kernel(&psp->fw_pri_bo, &psp->fw_pri_mc_addr, &psp->fw_pri_buf); failed: @@ -431,20 +442,17 @@ static int psp_hw_fini(void *handle) if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) return 0; - amdgpu_ucode_fini_bo(adev); - psp_ring_destroy(psp, PSP_RING_TYPE__KM); - if (psp->tmr_buf) - amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf); - - if (psp->fw_pri_buf) - amdgpu_bo_free_kernel(&psp->fw_pri_bo, - &psp->fw_pri_mc_addr, &psp->fw_pri_buf); - - if (psp->fence_buf_bo) - amdgpu_bo_free_kernel(&psp->fence_buf_bo, - &psp->fence_buf_mc_addr, &psp->fence_buf); + amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf); + amdgpu_bo_free_kernel(&psp->fw_pri_bo, + &psp->fw_pri_mc_addr, &psp->fw_pri_buf); + amdgpu_bo_free_kernel(&psp->fence_buf_bo, + &psp->fence_buf_mc_addr, &psp->fence_buf); + amdgpu_bo_free_kernel(&psp->asd_shared_bo, &psp->asd_shared_mc_addr, + &psp->asd_shared_buf); + amdgpu_bo_free_kernel(&psp->cmd_buf_bo, &psp->cmd_buf_mc_addr, + (void **)&psp->cmd_buf_mem); kfree(psp->cmd); psp->cmd = NULL; @@ -454,6 +462,16 @@ static int psp_hw_fini(void *handle) static int psp_suspend(void *handle) { + int ret; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct psp_context *psp = &adev->psp; + + ret = psp_ring_stop(psp, PSP_RING_TYPE__KM); + if (ret) { + DRM_ERROR("PSP ring stop failed\n"); + return ret; + } + return 0; } @@ -488,6 +506,22 @@ failed: return ret; } +static bool psp_check_reset(void* handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (adev->flags & AMD_IS_APU) + return true; + + return false; +} + +static int psp_reset(void* handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + return psp_mode1_reset(&adev->psp); +} + static bool psp_check_fw_loading_status(struct amdgpu_device *adev, enum AMDGPU_UCODE_ID ucode_type) { @@ -531,8 +565,9 @@ const struct amd_ip_funcs psp_ip_funcs = { .suspend = psp_suspend, .resume = psp_resume, .is_idle = NULL, + .check_soft_reset = psp_check_reset, .wait_for_idle = NULL, - .soft_reset = NULL, + .soft_reset = psp_reset, .set_clockgating_state = psp_set_clockgating_state, .set_powergating_state = psp_set_powergating_state, }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h index 1a1c8b469f93..ce4654550416 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h @@ -66,6 +66,8 @@ struct psp_context struct psp_gfx_cmd_resp *cmd); int (*ring_init)(struct psp_context *psp, enum psp_ring_type ring_type); int (*ring_create)(struct psp_context *psp, enum psp_ring_type ring_type); + int (*ring_stop)(struct psp_context *psp, + enum psp_ring_type ring_type); int (*ring_destroy)(struct psp_context *psp, enum psp_ring_type ring_type); int (*cmd_submit)(struct psp_context *psp, struct amdgpu_firmware_info *ucode, @@ -74,6 +76,7 @@ struct psp_context struct amdgpu_firmware_info *ucode, enum AMDGPU_UCODE_ID ucode_type); bool (*smu_reload_quirk)(struct psp_context *psp); + int (*mode1_reset)(struct psp_context *psp); /* fence buffer */ struct amdgpu_bo *fw_pri_bo; @@ -108,6 +111,11 @@ struct psp_context struct amdgpu_bo *fence_buf_bo; uint64_t fence_buf_mc_addr; void *fence_buf; + + /* cmd buffer */ + struct amdgpu_bo *cmd_buf_bo; + uint64_t cmd_buf_mc_addr; + struct psp_gfx_cmd_resp *cmd_buf_mem; }; struct amdgpu_psp_funcs { @@ -118,6 +126,7 @@ struct amdgpu_psp_funcs { #define psp_prep_cmd_buf(ucode, type) (psp)->prep_cmd_buf((ucode), (type)) #define psp_ring_init(psp, type) (psp)->ring_init((psp), (type)) #define psp_ring_create(psp, type) (psp)->ring_create((psp), (type)) +#define psp_ring_stop(psp, type) (psp)->ring_stop((psp), (type)) #define psp_ring_destroy(psp, type) ((psp)->ring_destroy((psp), (type))) #define psp_cmd_submit(psp, ucode, cmd_mc, fence_mc, index) \ (psp)->cmd_submit((psp), (ucode), (cmd_mc), (fence_mc), (index)) @@ -131,6 +140,8 @@ struct amdgpu_psp_funcs { ((psp)->bootloader_load_sos ? (psp)->bootloader_load_sos((psp)) : 0) #define psp_smu_reload_quirk(psp) \ ((psp)->smu_reload_quirk ? (psp)->smu_reload_quirk((psp)) : false) +#define psp_mode1_reset(psp) \ + ((psp)->mode1_reset ? (psp)->mode1_reset((psp)) : false) extern const struct amd_ip_funcs psp_ip_funcs; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c index befc09b68543..190e28cb827e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c @@ -121,7 +121,7 @@ static enum amdgpu_ring_type amdgpu_hw_ip_to_ring_type(int hw_ip) static int amdgpu_lru_map(struct amdgpu_device *adev, struct amdgpu_queue_mapper *mapper, - int user_ring, + int user_ring, bool lru_pipe_order, struct amdgpu_ring **out_ring) { int r, i, j; @@ -139,7 +139,7 @@ static int amdgpu_lru_map(struct amdgpu_device *adev, } r = amdgpu_ring_lru_get(adev, ring_type, ring_blacklist, - j, out_ring); + j, lru_pipe_order, out_ring); if (r) return r; @@ -284,8 +284,10 @@ int amdgpu_queue_mgr_map(struct amdgpu_device *adev, r = amdgpu_identity_map(adev, mapper, ring, out_ring); break; case AMDGPU_HW_IP_DMA: + r = amdgpu_lru_map(adev, mapper, ring, false, out_ring); + break; case AMDGPU_HW_IP_COMPUTE: - r = amdgpu_lru_map(adev, mapper, ring, out_ring); + r = amdgpu_lru_map(adev, mapper, ring, true, out_ring); break; default: *out_ring = NULL; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index 75165e07b1cd..a98fbbb4739f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -136,7 +136,8 @@ void amdgpu_ring_commit(struct amdgpu_ring *ring) if (ring->funcs->end_use) ring->funcs->end_use(ring); - amdgpu_ring_lru_touch(ring->adev, ring); + if (ring->funcs->type != AMDGPU_RING_TYPE_KIQ) + amdgpu_ring_lru_touch(ring->adev, ring); } /** @@ -155,6 +156,75 @@ void amdgpu_ring_undo(struct amdgpu_ring *ring) } /** + * amdgpu_ring_priority_put - restore a ring's priority + * + * @ring: amdgpu_ring structure holding the information + * @priority: target priority + * + * Release a request for executing at @priority + */ +void amdgpu_ring_priority_put(struct amdgpu_ring *ring, + enum amd_sched_priority priority) +{ + int i; + + if (!ring->funcs->set_priority) + return; + + if (atomic_dec_return(&ring->num_jobs[priority]) > 0) + return; + + /* no need to restore if the job is already at the lowest priority */ + if (priority == AMD_SCHED_PRIORITY_NORMAL) + return; + + mutex_lock(&ring->priority_mutex); + /* something higher prio is executing, no need to decay */ + if (ring->priority > priority) + goto out_unlock; + + /* decay priority to the next level with a job available */ + for (i = priority; i >= AMD_SCHED_PRIORITY_MIN; i--) { + if (i == AMD_SCHED_PRIORITY_NORMAL + || atomic_read(&ring->num_jobs[i])) { + ring->priority = i; + ring->funcs->set_priority(ring, i); + break; + } + } + +out_unlock: + mutex_unlock(&ring->priority_mutex); +} + +/** + * amdgpu_ring_priority_get - change the ring's priority + * + * @ring: amdgpu_ring structure holding the information + * @priority: target priority + * + * Request a ring's priority to be raised to @priority (refcounted). + */ +void amdgpu_ring_priority_get(struct amdgpu_ring *ring, + enum amd_sched_priority priority) +{ + if (!ring->funcs->set_priority) + return; + + atomic_inc(&ring->num_jobs[priority]); + + mutex_lock(&ring->priority_mutex); + if (priority <= ring->priority) + goto out_unlock; + + ring->priority = priority; + ring->funcs->set_priority(ring, priority); + +out_unlock: + mutex_unlock(&ring->priority_mutex); +} + +/** * amdgpu_ring_init - init driver ring struct. * * @adev: amdgpu_device pointer @@ -169,7 +239,17 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned max_dw, struct amdgpu_irq_src *irq_src, unsigned irq_type) { - int r; + int r, i; + int sched_hw_submission = amdgpu_sched_hw_submission; + + /* Set the hw submission limit higher for KIQ because + * it's used for a number of gfx/compute tasks by both + * KFD and KGD which may have outstanding fences and + * it doesn't really use the gpu scheduler anyway; + * KIQ tasks get submitted directly to the ring. + */ + if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) + sched_hw_submission = max(sched_hw_submission, 256); if (ring->adev == NULL) { if (adev->num_rings >= AMDGPU_MAX_RINGS) @@ -178,38 +258,21 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, ring->adev = adev; ring->idx = adev->num_rings++; adev->rings[ring->idx] = ring; - r = amdgpu_fence_driver_init_ring(ring, - amdgpu_sched_hw_submission); + r = amdgpu_fence_driver_init_ring(ring, sched_hw_submission); if (r) return r; } - if (ring->funcs->support_64bit_ptrs) { - r = amdgpu_wb_get_64bit(adev, &ring->rptr_offs); - if (r) { - dev_err(adev->dev, "(%d) ring rptr_offs wb alloc failed\n", r); - return r; - } - - r = amdgpu_wb_get_64bit(adev, &ring->wptr_offs); - if (r) { - dev_err(adev->dev, "(%d) ring wptr_offs wb alloc failed\n", r); - return r; - } - - } else { - r = amdgpu_wb_get(adev, &ring->rptr_offs); - if (r) { - dev_err(adev->dev, "(%d) ring rptr_offs wb alloc failed\n", r); - return r; - } - - r = amdgpu_wb_get(adev, &ring->wptr_offs); - if (r) { - dev_err(adev->dev, "(%d) ring wptr_offs wb alloc failed\n", r); - return r; - } + r = amdgpu_wb_get(adev, &ring->rptr_offs); + if (r) { + dev_err(adev->dev, "(%d) ring rptr_offs wb alloc failed\n", r); + return r; + } + r = amdgpu_wb_get(adev, &ring->wptr_offs); + if (r) { + dev_err(adev->dev, "(%d) ring wptr_offs wb alloc failed\n", r); + return r; } r = amdgpu_wb_get(adev, &ring->fence_offs); @@ -234,8 +297,7 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, return r; } - ring->ring_size = roundup_pow_of_two(max_dw * 4 * - amdgpu_sched_hw_submission); + ring->ring_size = roundup_pow_of_two(max_dw * 4 * sched_hw_submission); ring->buf_mask = (ring->ring_size / 4) - 1; ring->ptr_mask = ring->funcs->support_64bit_ptrs ? @@ -255,9 +317,14 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, } ring->max_dw = max_dw; + ring->priority = AMD_SCHED_PRIORITY_NORMAL; + mutex_init(&ring->priority_mutex); INIT_LIST_HEAD(&ring->lru_list); amdgpu_ring_lru_touch(adev, ring); + for (i = 0; i < AMD_SCHED_PRIORITY_MAX; ++i) + atomic_set(&ring->num_jobs[i], 0); + if (amdgpu_debugfs_ring_init(adev, ring)) { DRM_ERROR("Failed to register debugfs file for rings !\n"); } @@ -277,18 +344,15 @@ void amdgpu_ring_fini(struct amdgpu_ring *ring) { ring->ready = false; - if (ring->funcs->support_64bit_ptrs) { - amdgpu_wb_free_64bit(ring->adev, ring->cond_exe_offs); - amdgpu_wb_free_64bit(ring->adev, ring->fence_offs); - amdgpu_wb_free_64bit(ring->adev, ring->rptr_offs); - amdgpu_wb_free_64bit(ring->adev, ring->wptr_offs); - } else { - amdgpu_wb_free(ring->adev, ring->cond_exe_offs); - amdgpu_wb_free(ring->adev, ring->fence_offs); - amdgpu_wb_free(ring->adev, ring->rptr_offs); - amdgpu_wb_free(ring->adev, ring->wptr_offs); - } + /* Not to finish a ring which is not initialized */ + if (!(ring->adev) || !(ring->adev->rings[ring->idx])) + return; + + amdgpu_wb_free(ring->adev, ring->rptr_offs); + amdgpu_wb_free(ring->adev, ring->wptr_offs); + amdgpu_wb_free(ring->adev, ring->cond_exe_offs); + amdgpu_wb_free(ring->adev, ring->fence_offs); amdgpu_bo_free_kernel(&ring->ring_obj, &ring->gpu_addr, @@ -326,14 +390,16 @@ static bool amdgpu_ring_is_blacklisted(struct amdgpu_ring *ring, * @type: amdgpu_ring_type enum * @blacklist: blacklisted ring ids array * @num_blacklist: number of entries in @blacklist + * @lru_pipe_order: find a ring from the least recently used pipe * @ring: output ring * * Retrieve the amdgpu_ring structure for the least recently used ring of * a specific IP block (all asics). * Returns 0 on success, error on failure. */ -int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist, - int num_blacklist, struct amdgpu_ring **ring) +int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, + int *blacklist, int num_blacklist, + bool lru_pipe_order, struct amdgpu_ring **ring) { struct amdgpu_ring *entry; @@ -348,10 +414,23 @@ int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist, if (amdgpu_ring_is_blacklisted(entry, blacklist, num_blacklist)) continue; - *ring = entry; - amdgpu_ring_lru_touch_locked(adev, *ring); - break; + if (!*ring) { + *ring = entry; + + /* We are done for ring LRU */ + if (!lru_pipe_order) + break; + } + + /* Move all rings on the same pipe to the end of the list */ + if (entry->pipe == (*ring)->pipe) + amdgpu_ring_lru_touch_locked(adev, entry); } + + /* Move the ring we found to the end of the list */ + if (*ring) + amdgpu_ring_lru_touch_locked(adev, *ring); + spin_unlock(&adev->ring_lru_list_lock); if (!*ring) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index bc8dec992f73..b18c2b96691f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -24,6 +24,7 @@ #ifndef __AMDGPU_RING_H__ #define __AMDGPU_RING_H__ +#include <drm/amdgpu_drm.h> #include "gpu_scheduler.h" /* max number of rings */ @@ -56,6 +57,7 @@ struct amdgpu_device; struct amdgpu_ring; struct amdgpu_ib; struct amdgpu_cs_parser; +struct amdgpu_job; /* * Fences. @@ -88,8 +90,12 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring, void amdgpu_fence_driver_suspend(struct amdgpu_device *adev); void amdgpu_fence_driver_resume(struct amdgpu_device *adev); int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **fence); +int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s); void amdgpu_fence_process(struct amdgpu_ring *ring); int amdgpu_fence_wait_empty(struct amdgpu_ring *ring); +signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring, + uint32_t wait_seq, + signed long timeout); unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring); /* @@ -147,6 +153,9 @@ struct amdgpu_ring_funcs { void (*emit_rreg)(struct amdgpu_ring *ring, uint32_t reg); void (*emit_wreg)(struct amdgpu_ring *ring, uint32_t reg, uint32_t val); void (*emit_tmz)(struct amdgpu_ring *ring, bool start); + /* priority functions */ + void (*set_priority) (struct amdgpu_ring *ring, + enum amd_sched_priority priority); }; struct amdgpu_ring { @@ -187,6 +196,12 @@ struct amdgpu_ring { volatile u32 *cond_exe_cpu_addr; unsigned vm_inv_eng; bool has_compute_vm_bug; + + atomic_t num_jobs[AMD_SCHED_PRIORITY_MAX]; + struct mutex priority_mutex; + /* protected by priority_mutex */ + int priority; + #if defined(CONFIG_DEBUG_FS) struct dentry *ent; #endif @@ -197,12 +212,17 @@ void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count); void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib); void amdgpu_ring_commit(struct amdgpu_ring *ring); void amdgpu_ring_undo(struct amdgpu_ring *ring); +void amdgpu_ring_priority_get(struct amdgpu_ring *ring, + enum amd_sched_priority priority); +void amdgpu_ring_priority_put(struct amdgpu_ring *ring, + enum amd_sched_priority priority); int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned ring_size, struct amdgpu_irq_src *irq_src, unsigned irq_type); void amdgpu_ring_fini(struct amdgpu_ring *ring); -int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist, - int num_blacklist, struct amdgpu_ring **ring); +int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, + int *blacklist, int num_blacklist, + bool lru_pipe_order, struct amdgpu_ring **ring); void amdgpu_ring_lru_touch(struct amdgpu_device *adev, struct amdgpu_ring *ring); static inline void amdgpu_ring_clear_ring(struct amdgpu_ring *ring) { @@ -212,4 +232,44 @@ static inline void amdgpu_ring_clear_ring(struct amdgpu_ring *ring) } +static inline void amdgpu_ring_write(struct amdgpu_ring *ring, uint32_t v) +{ + if (ring->count_dw <= 0) + DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n"); + ring->ring[ring->wptr++ & ring->buf_mask] = v; + ring->wptr &= ring->ptr_mask; + ring->count_dw--; +} + +static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring, + void *src, int count_dw) +{ + unsigned occupied, chunk1, chunk2; + void *dst; + + if (unlikely(ring->count_dw < count_dw)) + DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n"); + + occupied = ring->wptr & ring->buf_mask; + dst = (void *)&ring->ring[occupied]; + chunk1 = ring->buf_mask + 1 - occupied; + chunk1 = (chunk1 >= count_dw) ? count_dw: chunk1; + chunk2 = count_dw - chunk1; + chunk1 <<= 2; + chunk2 <<= 2; + + if (chunk1) + memcpy(dst, src, chunk1); + + if (chunk2) { + src += chunk1; + dst = (void *)ring->ring; + memcpy(dst, src, chunk2); + } + + ring->wptr += count_dw; + ring->wptr &= ring->ptr_mask; + ring->count_dw -= count_dw; +} + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c index 5ca75a456ad2..3144400435b7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c @@ -64,7 +64,7 @@ int amdgpu_sa_bo_manager_init(struct amdgpu_device *adev, INIT_LIST_HEAD(&sa_manager->flist[i]); r = amdgpu_bo_create(adev, size, align, true, domain, - 0, NULL, NULL, &sa_manager->bo); + 0, NULL, NULL, 0, &sa_manager->bo); if (r) { dev_err(adev->dev, "(%d) failed to allocate bo for manager\n", r); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c new file mode 100644 index 000000000000..290cc3f9c433 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c @@ -0,0 +1,109 @@ +/* + * Copyright 2017 Valve Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Andres Rodriguez <andresx7@gmail.com> + */ + +#include <linux/fdtable.h> +#include <linux/pid.h> +#include <drm/amdgpu_drm.h> +#include "amdgpu.h" + +#include "amdgpu_vm.h" + +enum amd_sched_priority amdgpu_to_sched_priority(int amdgpu_priority) +{ + switch (amdgpu_priority) { + case AMDGPU_CTX_PRIORITY_VERY_HIGH: + return AMD_SCHED_PRIORITY_HIGH_HW; + case AMDGPU_CTX_PRIORITY_HIGH: + return AMD_SCHED_PRIORITY_HIGH_SW; + case AMDGPU_CTX_PRIORITY_NORMAL: + return AMD_SCHED_PRIORITY_NORMAL; + case AMDGPU_CTX_PRIORITY_LOW: + case AMDGPU_CTX_PRIORITY_VERY_LOW: + return AMD_SCHED_PRIORITY_LOW; + case AMDGPU_CTX_PRIORITY_UNSET: + return AMD_SCHED_PRIORITY_UNSET; + default: + WARN(1, "Invalid context priority %d\n", amdgpu_priority); + return AMD_SCHED_PRIORITY_INVALID; + } +} + +static int amdgpu_sched_process_priority_override(struct amdgpu_device *adev, + int fd, + enum amd_sched_priority priority) +{ + struct file *filp = fcheck(fd); + struct drm_file *file; + struct pid *pid; + struct amdgpu_fpriv *fpriv; + struct amdgpu_ctx *ctx; + uint32_t id; + + if (!filp) + return -EINVAL; + + pid = get_pid(((struct drm_file *)filp->private_data)->pid); + + mutex_lock(&adev->ddev->filelist_mutex); + list_for_each_entry(file, &adev->ddev->filelist, lhead) { + if (file->pid != pid) + continue; + + fpriv = file->driver_priv; + idr_for_each_entry(&fpriv->ctx_mgr.ctx_handles, ctx, id) + amdgpu_ctx_priority_override(ctx, priority); + } + mutex_unlock(&adev->ddev->filelist_mutex); + + put_pid(pid); + + return 0; +} + +int amdgpu_sched_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp) +{ + union drm_amdgpu_sched *args = data; + struct amdgpu_device *adev = dev->dev_private; + enum amd_sched_priority priority; + int r; + + priority = amdgpu_to_sched_priority(args->in.priority); + if (args->in.flags || priority == AMD_SCHED_PRIORITY_INVALID) + return -EINVAL; + + switch (args->in.op) { + case AMDGPU_SCHED_OP_PROCESS_PRIORITY_OVERRIDE: + r = amdgpu_sched_process_priority_override(adev, + args->in.fd, + priority); + break; + default: + DRM_ERROR("Invalid sched op specified: %d\n", args->in.op); + r = -EINVAL; + break; + } + + return r; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.h new file mode 100644 index 000000000000..b28c067d3822 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.h @@ -0,0 +1,34 @@ +/* + * Copyright 2017 Valve Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Andres Rodriguez <andresx7@gmail.com> + */ + +#ifndef __AMDGPU_SCHED_H__ +#define __AMDGPU_SCHED_H__ + +#include <drm/drmP.h> + +enum amd_sched_priority amdgpu_to_sched_priority(int amdgpu_priority); +int amdgpu_sched_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); + +#endif // __AMDGPU_SCHED_H__ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c index a6899180b265..a4bf21f8f1c1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c @@ -169,14 +169,14 @@ int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync, * * @sync: sync object to add fences from reservation object to * @resv: reservation object with embedded fence - * @shared: true if we should only sync to the exclusive fence + * @explicit_sync: true if we should only sync to the exclusive fence * * Sync to the fence */ int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync, struct reservation_object *resv, - void *owner) + void *owner, bool explicit_sync) { struct reservation_object_list *flist; struct dma_fence *f; @@ -191,6 +191,9 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, f = reservation_object_get_excl(resv); r = amdgpu_sync_fence(adev, sync, f); + if (explicit_sync) + return r; + flist = reservation_object_get_list(resv); if (!flist || r) return r; @@ -244,6 +247,12 @@ struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync, struct dma_fence *f = e->fence; struct amd_sched_fence *s_fence = to_amd_sched_fence(f); + if (dma_fence_is_signaled(f)) { + hash_del(&e->node); + dma_fence_put(f); + kmem_cache_free(amdgpu_sync_slab, e); + continue; + } if (ring && s_fence) { /* For fences from the same ring it is sufficient * when they are scheduled. @@ -256,13 +265,6 @@ struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync, } } - if (dma_fence_is_signaled(f)) { - hash_del(&e->node); - dma_fence_put(f); - kmem_cache_free(amdgpu_sync_slab, e); - continue; - } - return f; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h index dc7687993317..70d7e3a279a0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h @@ -45,7 +45,8 @@ int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync, int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync, struct reservation_object *resv, - void *owner); + void *owner, + bool explicit_sync); struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync, struct amdgpu_ring *ring); struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c index 15510dadde01..ed8c3739015b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c @@ -33,7 +33,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; struct amdgpu_bo *vram_obj = NULL; struct amdgpu_bo **gtt_obj = NULL; - uint64_t gtt_addr, vram_addr; + uint64_t gart_addr, vram_addr; unsigned n, size; int i, r; @@ -42,7 +42,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) /* Number of tests = * (Total GTT - IB pool - writeback page - ring buffers) / test size */ - n = adev->mc.gtt_size - AMDGPU_IB_POOL_SIZE*64*1024; + n = adev->mc.gart_size - AMDGPU_IB_POOL_SIZE*64*1024; for (i = 0; i < AMDGPU_MAX_RINGS; ++i) if (adev->rings[i]) n -= adev->rings[i]->ring_size; @@ -61,7 +61,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM, 0, - NULL, NULL, &vram_obj); + NULL, NULL, 0, &vram_obj); if (r) { DRM_ERROR("Failed to create VRAM object\n"); goto out_cleanup; @@ -76,13 +76,13 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) } for (i = 0; i < n; i++) { void *gtt_map, *vram_map; - void **gtt_start, **gtt_end; + void **gart_start, **gart_end; void **vram_start, **vram_end; struct dma_fence *fence = NULL; r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_GTT, 0, NULL, - NULL, gtt_obj + i); + NULL, 0, gtt_obj + i); if (r) { DRM_ERROR("Failed to create GTT object %d\n", i); goto out_lclean; @@ -91,7 +91,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) r = amdgpu_bo_reserve(gtt_obj[i], false); if (unlikely(r != 0)) goto out_lclean_unref; - r = amdgpu_bo_pin(gtt_obj[i], AMDGPU_GEM_DOMAIN_GTT, >t_addr); + r = amdgpu_bo_pin(gtt_obj[i], AMDGPU_GEM_DOMAIN_GTT, &gart_addr); if (r) { DRM_ERROR("Failed to pin GTT object %d\n", i); goto out_lclean_unres; @@ -103,15 +103,15 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) goto out_lclean_unpin; } - for (gtt_start = gtt_map, gtt_end = gtt_map + size; - gtt_start < gtt_end; - gtt_start++) - *gtt_start = gtt_start; + for (gart_start = gtt_map, gart_end = gtt_map + size; + gart_start < gart_end; + gart_start++) + *gart_start = gart_start; amdgpu_bo_kunmap(gtt_obj[i]); - r = amdgpu_copy_buffer(ring, gtt_addr, vram_addr, - size, NULL, &fence, false); + r = amdgpu_copy_buffer(ring, gart_addr, vram_addr, + size, NULL, &fence, false, false); if (r) { DRM_ERROR("Failed GTT->VRAM copy %d\n", i); @@ -132,21 +132,21 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) goto out_lclean_unpin; } - for (gtt_start = gtt_map, gtt_end = gtt_map + size, + for (gart_start = gtt_map, gart_end = gtt_map + size, vram_start = vram_map, vram_end = vram_map + size; vram_start < vram_end; - gtt_start++, vram_start++) { - if (*vram_start != gtt_start) { + gart_start++, vram_start++) { + if (*vram_start != gart_start) { DRM_ERROR("Incorrect GTT->VRAM copy %d: Got 0x%p, " "expected 0x%p (GTT/VRAM offset " "0x%16llx/0x%16llx)\n", - i, *vram_start, gtt_start, + i, *vram_start, gart_start, (unsigned long long) - (gtt_addr - adev->mc.gtt_start + - (void*)gtt_start - gtt_map), + (gart_addr - adev->mc.gart_start + + (void*)gart_start - gtt_map), (unsigned long long) (vram_addr - adev->mc.vram_start + - (void*)gtt_start - gtt_map)); + (void*)gart_start - gtt_map)); amdgpu_bo_kunmap(vram_obj); goto out_lclean_unpin; } @@ -155,8 +155,8 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) amdgpu_bo_kunmap(vram_obj); - r = amdgpu_copy_buffer(ring, vram_addr, gtt_addr, - size, NULL, &fence, false); + r = amdgpu_copy_buffer(ring, vram_addr, gart_addr, + size, NULL, &fence, false, false); if (r) { DRM_ERROR("Failed VRAM->GTT copy %d\n", i); @@ -177,20 +177,20 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) goto out_lclean_unpin; } - for (gtt_start = gtt_map, gtt_end = gtt_map + size, + for (gart_start = gtt_map, gart_end = gtt_map + size, vram_start = vram_map, vram_end = vram_map + size; - gtt_start < gtt_end; - gtt_start++, vram_start++) { - if (*gtt_start != vram_start) { + gart_start < gart_end; + gart_start++, vram_start++) { + if (*gart_start != vram_start) { DRM_ERROR("Incorrect VRAM->GTT copy %d: Got 0x%p, " "expected 0x%p (VRAM/GTT offset " "0x%16llx/0x%16llx)\n", - i, *gtt_start, vram_start, + i, *gart_start, vram_start, (unsigned long long) (vram_addr - adev->mc.vram_start + (void*)vram_start - vram_map), (unsigned long long) - (gtt_addr - adev->mc.gtt_start + + (gart_addr - adev->mc.gart_start + (void*)vram_start - vram_map)); amdgpu_bo_kunmap(gtt_obj[i]); goto out_lclean_unpin; @@ -200,7 +200,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) amdgpu_bo_kunmap(gtt_obj[i]); DRM_INFO("Tested GTT->VRAM and VRAM->GTT copy for GTT offset 0x%llx\n", - gtt_addr - adev->mc.gtt_start); + gart_addr - adev->mc.gart_start); continue; out_lclean_unpin: diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h index 8601904e670a..f337c316ec2c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h @@ -1,3 +1,4 @@ +/* SPDX-License-Identifier: GPL-2.0 */ #if !defined(_AMDGPU_TRACE_H) || defined(TRACE_HEADER_MULTI_READ) #define _AMDGPU_TRACE_H_ @@ -105,12 +106,12 @@ TRACE_EVENT(amdgpu_bo_create, __entry->bo = bo; __entry->pages = bo->tbo.num_pages; __entry->type = bo->tbo.mem.mem_type; - __entry->prefer = bo->prefered_domains; + __entry->prefer = bo->preferred_domains; __entry->allow = bo->allowed_domains; __entry->visible = bo->flags; ), - TP_printk("bo=%p, pages=%u, type=%d, prefered=%d, allowed=%d, visible=%d", + TP_printk("bo=%p, pages=%u, type=%d, preferred=%d, allowed=%d, visible=%d", __entry->bo, __entry->pages, __entry->type, __entry->prefer, __entry->allow, __entry->visible) ); @@ -224,17 +225,17 @@ TRACE_EVENT(amdgpu_vm_bo_map, __field(long, start) __field(long, last) __field(u64, offset) - __field(u32, flags) + __field(u64, flags) ), TP_fast_assign( - __entry->bo = bo_va ? bo_va->bo : NULL; + __entry->bo = bo_va ? bo_va->base.bo : NULL; __entry->start = mapping->start; __entry->last = mapping->last; __entry->offset = mapping->offset; __entry->flags = mapping->flags; ), - TP_printk("bo=%p, start=%lx, last=%lx, offset=%010llx, flags=%08x", + TP_printk("bo=%p, start=%lx, last=%lx, offset=%010llx, flags=%llx", __entry->bo, __entry->start, __entry->last, __entry->offset, __entry->flags) ); @@ -248,17 +249,17 @@ TRACE_EVENT(amdgpu_vm_bo_unmap, __field(long, start) __field(long, last) __field(u64, offset) - __field(u32, flags) + __field(u64, flags) ), TP_fast_assign( - __entry->bo = bo_va->bo; + __entry->bo = bo_va->base.bo; __entry->start = mapping->start; __entry->last = mapping->last; __entry->offset = mapping->offset; __entry->flags = mapping->flags; ), - TP_printk("bo=%p, start=%lx, last=%lx, offset=%010llx, flags=%08x", + TP_printk("bo=%p, start=%lx, last=%lx, offset=%010llx, flags=%llx", __entry->bo, __entry->start, __entry->last, __entry->offset, __entry->flags) ); @@ -269,7 +270,7 @@ DECLARE_EVENT_CLASS(amdgpu_vm_mapping, TP_STRUCT__entry( __field(u64, soffset) __field(u64, eoffset) - __field(u32, flags) + __field(u64, flags) ), TP_fast_assign( @@ -277,7 +278,7 @@ DECLARE_EVENT_CLASS(amdgpu_vm_mapping, __entry->eoffset = mapping->last + 1; __entry->flags = mapping->flags; ), - TP_printk("soffs=%010llx, eoffs=%010llx, flags=%08x", + TP_printk("soffs=%010llx, eoffs=%010llx, flags=%llx", __entry->soffset, __entry->eoffset, __entry->flags) ); @@ -293,14 +294,14 @@ DEFINE_EVENT(amdgpu_vm_mapping, amdgpu_vm_bo_mapping, TRACE_EVENT(amdgpu_vm_set_ptes, TP_PROTO(uint64_t pe, uint64_t addr, unsigned count, - uint32_t incr, uint32_t flags), + uint32_t incr, uint64_t flags), TP_ARGS(pe, addr, count, incr, flags), TP_STRUCT__entry( __field(u64, pe) __field(u64, addr) __field(u32, count) __field(u32, incr) - __field(u32, flags) + __field(u64, flags) ), TP_fast_assign( @@ -310,7 +311,7 @@ TRACE_EVENT(amdgpu_vm_set_ptes, __entry->incr = incr; __entry->flags = flags; ), - TP_printk("pe=%010Lx, addr=%010Lx, incr=%u, flags=%08x, count=%u", + TP_printk("pe=%010Lx, addr=%010Lx, incr=%u, flags=%llx, count=%u", __entry->pe, __entry->addr, __entry->incr, __entry->flags, __entry->count) ); @@ -417,5 +418,5 @@ TRACE_EVENT(amdgpu_ttm_bo_move, /* This part must be outside protection */ #undef TRACE_INCLUDE_PATH -#define TRACE_INCLUDE_PATH . +#define TRACE_INCLUDE_PATH ../../drivers/gpu/drm/amd/amdgpu #include <trace/define_trace.h> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c index 385b7e1d72f9..b160b958e5fe 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c @@ -1,4 +1,24 @@ +// SPDX-License-Identifier: GPL-2.0 /* Copyright Red Hat Inc 2010. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * * Author : Dave Airlie <airlied@redhat.com> */ #include <drm/drmP.h> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index c9b131b13ef7..ad5bf86ee8a3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -42,15 +42,23 @@ #include <linux/swap.h> #include <linux/pagemap.h> #include <linux/debugfs.h> +#include <linux/iommu.h> #include "amdgpu.h" +#include "amdgpu_object.h" +#include "amdgpu_trace.h" #include "bif/bif_4_1_d.h" #define DRM_FILE_PAGE_OFFSET (0x100000000ULL >> PAGE_SHIFT) +static int amdgpu_map_buffer(struct ttm_buffer_object *bo, + struct ttm_mem_reg *mem, unsigned num_pages, + uint64_t offset, unsigned window, + struct amdgpu_ring *ring, + uint64_t *addr); + static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev); static void amdgpu_ttm_debugfs_fini(struct amdgpu_device *adev); - /* * Global memory. */ @@ -97,6 +105,8 @@ static int amdgpu_ttm_global_init(struct amdgpu_device *adev) goto error_bo; } + mutex_init(&adev->mman.gtt_window_lock); + ring = adev->mman.buffer_funcs_ring; rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_KERNEL]; r = amd_sched_entity_init(&ring->sched, &adev->mman.entity, @@ -123,6 +133,7 @@ static void amdgpu_ttm_global_fini(struct amdgpu_device *adev) if (adev->mman.mem_global_referenced) { amd_sched_entity_fini(adev->mman.entity.sched, &adev->mman.entity); + mutex_destroy(&adev->mman.gtt_window_lock); drm_global_item_unref(&adev->mman.bo_global_ref.ref); drm_global_item_unref(&adev->mman.mem_global_ref); adev->mman.mem_global_referenced = false; @@ -150,7 +161,7 @@ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type, break; case TTM_PL_TT: man->func = &amdgpu_gtt_mgr_func; - man->gpu_offset = adev->mc.gtt_start; + man->gpu_offset = adev->mc.gart_start; man->available_caching = TTM_PL_MASK_CACHING; man->default_caching = TTM_PL_FLAG_CACHED; man->flags = TTM_MEMTYPE_FLAG_MAPPABLE | TTM_MEMTYPE_FLAG_CMA; @@ -186,12 +197,11 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct amdgpu_bo *abo; - static struct ttm_place placements = { + static const struct ttm_place placements = { .fpfn = 0, .lpfn = 0, .flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM }; - unsigned i; if (!amdgpu_ttm_bo_is_amdgpu_bo(bo)) { placement->placement = &placements; @@ -200,29 +210,43 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, placement->num_busy_placement = 1; return; } - abo = container_of(bo, struct amdgpu_bo, tbo); + abo = ttm_to_amdgpu_bo(bo); switch (bo->mem.mem_type) { case TTM_PL_VRAM: if (adev->mman.buffer_funcs && adev->mman.buffer_funcs_ring && adev->mman.buffer_funcs_ring->ready == false) { amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU); + } else if (adev->mc.visible_vram_size < adev->mc.real_vram_size && + !(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)) { + unsigned fpfn = adev->mc.visible_vram_size >> PAGE_SHIFT; + struct drm_mm_node *node = bo->mem.mm_node; + unsigned long pages_left; + + for (pages_left = bo->mem.num_pages; + pages_left; + pages_left -= node->size, node++) { + if (node->start < fpfn) + break; + } + + if (!pages_left) + goto gtt; + + /* Try evicting to the CPU inaccessible part of VRAM + * first, but only set GTT as busy placement, so this + * BO will be evicted to GTT rather than causing other + * BOs to be evicted from VRAM + */ + amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_VRAM | + AMDGPU_GEM_DOMAIN_GTT); + abo->placements[0].fpfn = fpfn; + abo->placements[0].lpfn = 0; + abo->placement.busy_placement = &abo->placements[1]; + abo->placement.num_busy_placement = 1; } else { +gtt: amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT); - for (i = 0; i < abo->placement.num_placement; ++i) { - if (!(abo->placements[i].flags & - TTM_PL_FLAG_TT)) - continue; - - if (abo->placements[i].lpfn) - continue; - - /* set an upper limit to force directly - * allocating address space for the BO. - */ - abo->placements[i].lpfn = - adev->mc.gtt_size >> PAGE_SHIFT; - } } break; case TTM_PL_TT: @@ -234,7 +258,7 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, static int amdgpu_verify_access(struct ttm_buffer_object *bo, struct file *filp) { - struct amdgpu_bo *abo = container_of(bo, struct amdgpu_bo, tbo); + struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo); if (amdgpu_ttm_tt_get_usermm(bo->ttm)) return -EPERM; @@ -252,106 +276,185 @@ static void amdgpu_move_null(struct ttm_buffer_object *bo, new_mem->mm_node = NULL; } -static int amdgpu_mm_node_addr(struct ttm_buffer_object *bo, - struct drm_mm_node *mm_node, - struct ttm_mem_reg *mem, - uint64_t *addr) +static uint64_t amdgpu_mm_node_addr(struct ttm_buffer_object *bo, + struct drm_mm_node *mm_node, + struct ttm_mem_reg *mem) { - int r; - - switch (mem->mem_type) { - case TTM_PL_TT: - r = amdgpu_ttm_bind(bo, mem); - if (r) - return r; + uint64_t addr = 0; - case TTM_PL_VRAM: - *addr = mm_node->start << PAGE_SHIFT; - *addr += bo->bdev->man[mem->mem_type].gpu_offset; - break; - default: - DRM_ERROR("Unknown placement %d\n", mem->mem_type); - return -EINVAL; + if (mem->mem_type != TTM_PL_TT || + amdgpu_gtt_mgr_is_allocated(mem)) { + addr = mm_node->start << PAGE_SHIFT; + addr += bo->bdev->man[mem->mem_type].gpu_offset; } + return addr; +} - return 0; +/** + * amdgpu_find_mm_node - Helper function finds the drm_mm_node + * corresponding to @offset. It also modifies the offset to be + * within the drm_mm_node returned + */ +static struct drm_mm_node *amdgpu_find_mm_node(struct ttm_mem_reg *mem, + unsigned long *offset) +{ + struct drm_mm_node *mm_node = mem->mm_node; + + while (*offset >= (mm_node->size << PAGE_SHIFT)) { + *offset -= (mm_node->size << PAGE_SHIFT); + ++mm_node; + } + return mm_node; } -static int amdgpu_move_blit(struct ttm_buffer_object *bo, - bool evict, bool no_wait_gpu, - struct ttm_mem_reg *new_mem, - struct ttm_mem_reg *old_mem) +/** + * amdgpu_copy_ttm_mem_to_mem - Helper function for copy + * + * The function copies @size bytes from {src->mem + src->offset} to + * {dst->mem + dst->offset}. src->bo and dst->bo could be same BO for a + * move and different for a BO to BO copy. + * + * @f: Returns the last fence if multiple jobs are submitted. + */ +int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev, + struct amdgpu_copy_mem *src, + struct amdgpu_copy_mem *dst, + uint64_t size, + struct reservation_object *resv, + struct dma_fence **f) { - struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; - - struct drm_mm_node *old_mm, *new_mm; - uint64_t old_start, old_size, new_start, new_size; - unsigned long num_pages; + struct drm_mm_node *src_mm, *dst_mm; + uint64_t src_node_start, dst_node_start, src_node_size, + dst_node_size, src_page_offset, dst_page_offset; struct dma_fence *fence = NULL; - int r; - - BUILD_BUG_ON((PAGE_SIZE % AMDGPU_GPU_PAGE_SIZE) != 0); + int r = 0; + const uint64_t GTT_MAX_BYTES = (AMDGPU_GTT_MAX_TRANSFER_SIZE * + AMDGPU_GPU_PAGE_SIZE); if (!ring->ready) { DRM_ERROR("Trying to move memory with ring turned off.\n"); return -EINVAL; } - old_mm = old_mem->mm_node; - r = amdgpu_mm_node_addr(bo, old_mm, old_mem, &old_start); - if (r) - return r; - old_size = old_mm->size; + src_mm = amdgpu_find_mm_node(src->mem, &src->offset); + src_node_start = amdgpu_mm_node_addr(src->bo, src_mm, src->mem) + + src->offset; + src_node_size = (src_mm->size << PAGE_SHIFT) - src->offset; + src_page_offset = src_node_start & (PAGE_SIZE - 1); + dst_mm = amdgpu_find_mm_node(dst->mem, &dst->offset); + dst_node_start = amdgpu_mm_node_addr(dst->bo, dst_mm, dst->mem) + + dst->offset; + dst_node_size = (dst_mm->size << PAGE_SHIFT) - dst->offset; + dst_page_offset = dst_node_start & (PAGE_SIZE - 1); - new_mm = new_mem->mm_node; - r = amdgpu_mm_node_addr(bo, new_mm, new_mem, &new_start); - if (r) - return r; - new_size = new_mm->size; + mutex_lock(&adev->mman.gtt_window_lock); - num_pages = new_mem->num_pages; - while (num_pages) { - unsigned long cur_pages = min(old_size, new_size); + while (size) { + unsigned long cur_size; + uint64_t from = src_node_start, to = dst_node_start; struct dma_fence *next; - r = amdgpu_copy_buffer(ring, old_start, new_start, - cur_pages * PAGE_SIZE, - bo->resv, &next, false); + /* Copy size cannot exceed GTT_MAX_BYTES. So if src or dst + * begins at an offset, then adjust the size accordingly + */ + cur_size = min3(min(src_node_size, dst_node_size), size, + GTT_MAX_BYTES); + if (cur_size + src_page_offset > GTT_MAX_BYTES || + cur_size + dst_page_offset > GTT_MAX_BYTES) + cur_size -= max(src_page_offset, dst_page_offset); + + /* Map only what needs to be accessed. Map src to window 0 and + * dst to window 1 + */ + if (src->mem->mem_type == TTM_PL_TT && + !amdgpu_gtt_mgr_is_allocated(src->mem)) { + r = amdgpu_map_buffer(src->bo, src->mem, + PFN_UP(cur_size + src_page_offset), + src_node_start, 0, ring, + &from); + if (r) + goto error; + /* Adjust the offset because amdgpu_map_buffer returns + * start of mapped page + */ + from += src_page_offset; + } + + if (dst->mem->mem_type == TTM_PL_TT && + !amdgpu_gtt_mgr_is_allocated(dst->mem)) { + r = amdgpu_map_buffer(dst->bo, dst->mem, + PFN_UP(cur_size + dst_page_offset), + dst_node_start, 1, ring, + &to); + if (r) + goto error; + to += dst_page_offset; + } + + r = amdgpu_copy_buffer(ring, from, to, cur_size, + resv, &next, false, true); if (r) goto error; dma_fence_put(fence); fence = next; - num_pages -= cur_pages; - if (!num_pages) + size -= cur_size; + if (!size) break; - old_size -= cur_pages; - if (!old_size) { - r = amdgpu_mm_node_addr(bo, ++old_mm, old_mem, - &old_start); - if (r) - goto error; - old_size = old_mm->size; + src_node_size -= cur_size; + if (!src_node_size) { + src_node_start = amdgpu_mm_node_addr(src->bo, ++src_mm, + src->mem); + src_node_size = (src_mm->size << PAGE_SHIFT); } else { - old_start += cur_pages * PAGE_SIZE; + src_node_start += cur_size; + src_page_offset = src_node_start & (PAGE_SIZE - 1); } - - new_size -= cur_pages; - if (!new_size) { - r = amdgpu_mm_node_addr(bo, ++new_mm, new_mem, - &new_start); - if (r) - goto error; - - new_size = new_mm->size; + dst_node_size -= cur_size; + if (!dst_node_size) { + dst_node_start = amdgpu_mm_node_addr(dst->bo, ++dst_mm, + dst->mem); + dst_node_size = (dst_mm->size << PAGE_SHIFT); } else { - new_start += cur_pages * PAGE_SIZE; + dst_node_start += cur_size; + dst_page_offset = dst_node_start & (PAGE_SIZE - 1); } } +error: + mutex_unlock(&adev->mman.gtt_window_lock); + if (f) + *f = dma_fence_get(fence); + dma_fence_put(fence); + return r; +} + + +static int amdgpu_move_blit(struct ttm_buffer_object *bo, + bool evict, bool no_wait_gpu, + struct ttm_mem_reg *new_mem, + struct ttm_mem_reg *old_mem) +{ + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); + struct amdgpu_copy_mem src, dst; + struct dma_fence *fence = NULL; + int r; + + src.bo = bo; + dst.bo = bo; + src.mem = old_mem; + dst.mem = new_mem; + src.offset = 0; + dst.offset = 0; + + r = amdgpu_ttm_copy_mem_to_mem(adev, &src, &dst, + new_mem->num_pages << PAGE_SHIFT, + bo->resv, &fence); + if (r) + goto error; r = ttm_bo_pipeline_move(bo, fence, evict, new_mem); dma_fence_put(fence); @@ -384,7 +487,7 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, placement.num_busy_placement = 1; placement.busy_placement = &placements; placements.fpfn = 0; - placements.lpfn = adev->mc.gtt_size >> PAGE_SHIFT; + placements.lpfn = 0; placements.flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT; r = ttm_bo_mem_space(bo, &placement, &tmp_mem, interruptible, no_wait_gpu); @@ -431,7 +534,7 @@ static int amdgpu_move_ram_vram(struct ttm_buffer_object *bo, placement.num_busy_placement = 1; placement.busy_placement = &placements; placements.fpfn = 0; - placements.lpfn = adev->mc.gtt_size >> PAGE_SHIFT; + placements.lpfn = 0; placements.flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT; r = ttm_bo_mem_space(bo, &placement, &tmp_mem, interruptible, no_wait_gpu); @@ -462,7 +565,7 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, int r; /* Can't move a pinned BO */ - abo = container_of(bo, struct amdgpu_bo, tbo); + abo = ttm_to_amdgpu_bo(bo); if (WARN_ON_ONCE(abo->pin_count > 0)) return -EINVAL; @@ -507,6 +610,15 @@ memcpy: } } + if (bo->type == ttm_bo_type_device && + new_mem->mem_type == TTM_PL_VRAM && + old_mem->mem_type != TTM_PL_VRAM) { + /* amdgpu_bo_fault_reserve_notify will re-set this if the CPU + * accesses the BO after it's moved. + */ + abo->flags &= ~AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; + } + /* update statistics */ atomic64_add((u64)bo->num_pages << PAGE_SHIFT, &adev->num_bytes_moved); return 0; @@ -551,13 +663,12 @@ static void amdgpu_ttm_io_mem_free(struct ttm_bo_device *bdev, struct ttm_mem_re static unsigned long amdgpu_ttm_io_mem_pfn(struct ttm_buffer_object *bo, unsigned long page_offset) { - struct drm_mm_node *mm = bo->mem.mm_node; - uint64_t size = mm->size; - uint64_t offset = page_offset; + struct drm_mm_node *mm; + unsigned long offset = (page_offset << PAGE_SHIFT); - page_offset = do_div(offset, size); - mm += offset; - return (bo->mem.bus.base >> PAGE_SHIFT) + mm->start + page_offset; + mm = amdgpu_find_mm_node(&bo->mem, &offset); + return (bo->mem.bus.base >> PAGE_SHIFT) + mm->start + + (offset >> PAGE_SHIFT); } /* @@ -578,6 +689,7 @@ struct amdgpu_ttm_tt { spinlock_t guptasklock; struct list_head guptasks; atomic_t mmu_invalidations; + uint32_t last_set_pages; struct list_head list; }; @@ -591,6 +703,8 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages) if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY)) flags |= FOLL_WRITE; + down_read(¤t->mm->mmap_sem); + if (gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) { /* check that we only use anonymous memory to prevent problems with writeback */ @@ -598,8 +712,10 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages) struct vm_area_struct *vma; vma = find_vma(gtt->usermm, gtt->userptr); - if (!vma || vma->vm_file || vma->vm_end < end) + if (!vma || vma->vm_file || vma->vm_end < end) { + up_read(¤t->mm->mmap_sem); return -EPERM; + } } do { @@ -626,13 +742,47 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages) } while (pinned < ttm->num_pages); + up_read(¤t->mm->mmap_sem); return 0; release_pages: - release_pages(pages, pinned, 0); + release_pages(pages, pinned); + up_read(¤t->mm->mmap_sem); return r; } +void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages) +{ + struct amdgpu_ttm_tt *gtt = (void *)ttm; + unsigned i; + + gtt->last_set_pages = atomic_read(>t->mmu_invalidations); + for (i = 0; i < ttm->num_pages; ++i) { + if (ttm->pages[i]) + put_page(ttm->pages[i]); + + ttm->pages[i] = pages ? pages[i] : NULL; + } +} + +void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm) +{ + struct amdgpu_ttm_tt *gtt = (void *)ttm; + unsigned i; + + for (i = 0; i < ttm->num_pages; ++i) { + struct page *page = ttm->pages[i]; + + if (!page) + continue; + + if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY)) + set_page_dirty(page); + + mark_page_accessed(page); + } +} + /* prepare the sg table with the user pages */ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm) { @@ -670,7 +820,6 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); struct amdgpu_ttm_tt *gtt = (void *)ttm; - struct sg_page_iter sg_iter; int write = !(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY); enum dma_data_direction direction = write ? @@ -683,14 +832,7 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm) /* free the sg table and pages again */ dma_unmap_sg(adev->dev, ttm->sg->sgl, ttm->sg->nents, direction); - for_each_sg_page(ttm->sg->sgl, &sg_iter, ttm->sg->nents, 0) { - struct page *page = sg_page_iter_page(&sg_iter); - if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY)) - set_page_dirty(page); - - mark_page_accessed(page); - put_page(page); - } + amdgpu_ttm_tt_mark_user_pages(ttm); sg_free_table(ttm->sg); } @@ -699,7 +841,8 @@ static int amdgpu_ttm_backend_bind(struct ttm_tt *ttm, struct ttm_mem_reg *bo_mem) { struct amdgpu_ttm_tt *gtt = (void*)ttm; - int r; + uint64_t flags; + int r = 0; if (gtt->userptr) { r = amdgpu_ttm_tt_pin_userptr(ttm); @@ -718,7 +861,25 @@ static int amdgpu_ttm_backend_bind(struct ttm_tt *ttm, bo_mem->mem_type == AMDGPU_PL_OA) return -EINVAL; - return 0; + if (!amdgpu_gtt_mgr_is_allocated(bo_mem)) + return 0; + + spin_lock(>t->adev->gtt_list_lock); + flags = amdgpu_ttm_tt_pte_flags(gtt->adev, ttm, bo_mem); + gtt->offset = (u64)bo_mem->start << PAGE_SHIFT; + r = amdgpu_gart_bind(gtt->adev, gtt->offset, ttm->num_pages, + ttm->pages, gtt->ttm.dma_address, flags); + + if (r) { + DRM_ERROR("failed to bind %lu pages at 0x%08llX\n", + ttm->num_pages, gtt->offset); + goto error_gart_bind; + } + + list_add_tail(>t->list, >t->adev->gtt_list); +error_gart_bind: + spin_unlock(>t->adev->gtt_list_lock); + return r; } bool amdgpu_ttm_is_bound(struct ttm_tt *ttm) @@ -730,36 +891,38 @@ bool amdgpu_ttm_is_bound(struct ttm_tt *ttm) int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem) { + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct ttm_tt *ttm = bo->ttm; - struct amdgpu_ttm_tt *gtt = (void *)bo->ttm; - uint64_t flags; + struct ttm_mem_reg tmp; + struct ttm_placement placement; + struct ttm_place placements; int r; if (!ttm || amdgpu_ttm_is_bound(ttm)) return 0; - r = amdgpu_gtt_mgr_alloc(&bo->bdev->man[TTM_PL_TT], bo, - NULL, bo_mem); - if (r) { - DRM_ERROR("Failed to allocate GTT address space (%d)\n", r); - return r; - } + tmp = bo->mem; + tmp.mm_node = NULL; + placement.num_placement = 1; + placement.placement = &placements; + placement.num_busy_placement = 1; + placement.busy_placement = &placements; + placements.fpfn = 0; + placements.lpfn = adev->mc.gart_size >> PAGE_SHIFT; + placements.flags = (bo->mem.placement & ~TTM_PL_MASK_MEM) | + TTM_PL_FLAG_TT; - spin_lock(>t->adev->gtt_list_lock); - flags = amdgpu_ttm_tt_pte_flags(gtt->adev, ttm, bo_mem); - gtt->offset = (u64)bo_mem->start << PAGE_SHIFT; - r = amdgpu_gart_bind(gtt->adev, gtt->offset, ttm->num_pages, - ttm->pages, gtt->ttm.dma_address, flags); + r = ttm_bo_mem_space(bo, &placement, &tmp, true, false); + if (unlikely(r)) + return r; - if (r) { - DRM_ERROR("failed to bind %lu pages at 0x%08llX\n", - ttm->num_pages, gtt->offset); - goto error_gart_bind; - } + r = ttm_bo_move_ttm(bo, true, false, &tmp); + if (unlikely(r)) + ttm_bo_mem_put(bo, &tmp); + else + bo->offset = (bo->mem.start << PAGE_SHIFT) + + bo->bdev->man[bo->mem.mem_type].gpu_offset; - list_add_tail(>t->list, >t->adev->gtt_list); -error_gart_bind: - spin_unlock(>t->adev->gtt_list_lock); return r; } @@ -852,10 +1015,8 @@ static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_bo_device *bdev, static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm) { - struct amdgpu_device *adev; + struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); struct amdgpu_ttm_tt *gtt = (void *)ttm; - unsigned i; - int r; bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG); if (ttm->state != tt_unpopulated) @@ -878,44 +1039,23 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm) return 0; } - adev = amdgpu_ttm_adev(ttm->bdev); - #ifdef CONFIG_SWIOTLB if (swiotlb_nr_tbl()) { return ttm_dma_populate(>t->ttm, adev->dev); } #endif - r = ttm_pool_populate(ttm); - if (r) { - return r; - } - - for (i = 0; i < ttm->num_pages; i++) { - gtt->ttm.dma_address[i] = pci_map_page(adev->pdev, ttm->pages[i], - 0, PAGE_SIZE, - PCI_DMA_BIDIRECTIONAL); - if (pci_dma_mapping_error(adev->pdev, gtt->ttm.dma_address[i])) { - while (i--) { - pci_unmap_page(adev->pdev, gtt->ttm.dma_address[i], - PAGE_SIZE, PCI_DMA_BIDIRECTIONAL); - gtt->ttm.dma_address[i] = 0; - } - ttm_pool_unpopulate(ttm); - return -EFAULT; - } - } - return 0; + return ttm_populate_and_map_pages(adev->dev, >t->ttm); } static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm) { struct amdgpu_device *adev; struct amdgpu_ttm_tt *gtt = (void *)ttm; - unsigned i; bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG); if (gtt && gtt->userptr) { + amdgpu_ttm_tt_set_user_pages(ttm, NULL); kfree(ttm->sg); ttm->page_flags &= ~TTM_PAGE_FLAG_SG; return; @@ -933,14 +1073,7 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm) } #endif - for (i = 0; i < ttm->num_pages; i++) { - if (gtt->ttm.dma_address[i]) { - pci_unmap_page(adev->pdev, gtt->ttm.dma_address[i], - PAGE_SIZE, PCI_DMA_BIDIRECTIONAL); - } - } - - ttm_pool_unpopulate(ttm); + ttm_unmap_and_unpopulate_pages(adev->dev, >t->ttm); } int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, @@ -957,6 +1090,7 @@ int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, spin_lock_init(>t->guptasklock); INIT_LIST_HEAD(>t->guptasks); atomic_set(>t->mmu_invalidations, 0); + gtt->last_set_pages = 0; return 0; } @@ -1009,6 +1143,16 @@ bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, return prev_invalidated != *last_invalidated; } +bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm) +{ + struct amdgpu_ttm_tt *gtt = (void *)ttm; + + if (gtt == NULL || !gtt->userptr) + return false; + + return atomic_read(>t->mmu_invalidations) != gtt->last_set_pages; +} + bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm) { struct amdgpu_ttm_tt *gtt = (void *)ttm; @@ -1049,9 +1193,6 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo, unsigned long num_pages = bo->mem.num_pages; struct drm_mm_node *node = bo->mem.mm_node; - if (bo->mem.start != AMDGPU_BO_INVALID_OFFSET) - return ttm_bo_eviction_valuable(bo, place); - switch (bo->mem.mem_type) { case TTM_PL_TT: return true; @@ -1066,7 +1207,7 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo, num_pages -= node->size; ++node; } - break; + return false; default: break; @@ -1075,6 +1216,64 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo, return ttm_bo_eviction_valuable(bo, place); } +static int amdgpu_ttm_access_memory(struct ttm_buffer_object *bo, + unsigned long offset, + void *buf, int len, int write) +{ + struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo); + struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev); + struct drm_mm_node *nodes; + uint32_t value = 0; + int ret = 0; + uint64_t pos; + unsigned long flags; + + if (bo->mem.mem_type != TTM_PL_VRAM) + return -EIO; + + nodes = amdgpu_find_mm_node(&abo->tbo.mem, &offset); + pos = (nodes->start << PAGE_SHIFT) + offset; + + while (len && pos < adev->mc.mc_vram_size) { + uint64_t aligned_pos = pos & ~(uint64_t)3; + uint32_t bytes = 4 - (pos & 3); + uint32_t shift = (pos & 3) * 8; + uint32_t mask = 0xffffffff << shift; + + if (len < bytes) { + mask &= 0xffffffff >> (bytes - len) * 8; + bytes = len; + } + + spin_lock_irqsave(&adev->mmio_idx_lock, flags); + WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)aligned_pos) | 0x80000000); + WREG32_NO_KIQ(mmMM_INDEX_HI, aligned_pos >> 31); + if (!write || mask != 0xffffffff) + value = RREG32_NO_KIQ(mmMM_DATA); + if (write) { + value &= ~mask; + value |= (*(uint32_t *)buf << shift) & mask; + WREG32_NO_KIQ(mmMM_DATA, value); + } + spin_unlock_irqrestore(&adev->mmio_idx_lock, flags); + if (!write) { + value = (value & mask) >> shift; + memcpy(buf, &value, bytes); + } + + ret += bytes; + buf = (uint8_t *)buf + bytes; + pos += bytes; + len -= bytes; + if (pos >= (nodes->start + nodes->size) << PAGE_SHIFT) { + ++nodes; + pos = (nodes->start << PAGE_SHIFT); + } + } + + return ret; +} + static struct ttm_bo_driver amdgpu_bo_driver = { .ttm_tt_create = &amdgpu_ttm_tt_create, .ttm_tt_populate = &amdgpu_ttm_tt_populate, @@ -1090,11 +1289,14 @@ static struct ttm_bo_driver amdgpu_bo_driver = { .io_mem_reserve = &amdgpu_ttm_io_mem_reserve, .io_mem_free = &amdgpu_ttm_io_mem_free, .io_mem_pfn = amdgpu_ttm_io_mem_pfn, + .access_memory = &amdgpu_ttm_access_memory }; int amdgpu_ttm_init(struct amdgpu_device *adev) { + uint64_t gtt_size; int r; + u64 vis_vram_limit; r = amdgpu_ttm_global_init(adev); if (r) { @@ -1118,36 +1320,46 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) DRM_ERROR("Failed initializing VRAM heap.\n"); return r; } + + /* Reduce size of CPU-visible VRAM if requested */ + vis_vram_limit = (u64)amdgpu_vis_vram_limit * 1024 * 1024; + if (amdgpu_vis_vram_limit > 0 && + vis_vram_limit <= adev->mc.visible_vram_size) + adev->mc.visible_vram_size = vis_vram_limit; + /* Change the size here instead of the init above so only lpfn is affected */ amdgpu_ttm_set_active_vram_size(adev, adev->mc.visible_vram_size); - r = amdgpu_bo_create(adev, adev->mc.stolen_size, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, &adev->stollen_vga_memory); + /* + *The reserved vram for firmware must be pinned to the specified + *place on the VRAM, so reserve it early. + */ + r = amdgpu_fw_reserve_vram_init(adev); if (r) { return r; } - r = amdgpu_bo_reserve(adev->stollen_vga_memory, false); + + r = amdgpu_bo_create_kernel(adev, adev->mc.stolen_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->stolen_vga_memory, + NULL, NULL); if (r) return r; - r = amdgpu_bo_pin(adev->stollen_vga_memory, AMDGPU_GEM_DOMAIN_VRAM, NULL); - amdgpu_bo_unreserve(adev->stollen_vga_memory); - if (r) { - amdgpu_bo_unref(&adev->stollen_vga_memory); - return r; - } DRM_INFO("amdgpu: %uM of VRAM memory ready\n", (unsigned) (adev->mc.real_vram_size / (1024 * 1024))); - r = ttm_bo_init_mm(&adev->mman.bdev, TTM_PL_TT, - adev->mc.gtt_size >> PAGE_SHIFT); + + if (amdgpu_gtt_size == -1) + gtt_size = max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20), + adev->mc.mc_vram_size); + else + gtt_size = (uint64_t)amdgpu_gtt_size << 20; + r = ttm_bo_init_mm(&adev->mman.bdev, TTM_PL_TT, gtt_size >> PAGE_SHIFT); if (r) { DRM_ERROR("Failed initializing GTT heap.\n"); return r; } DRM_INFO("amdgpu: %uM of GTT memory ready.\n", - (unsigned)(adev->mc.gtt_size / (1024 * 1024))); + (unsigned)(gtt_size / (1024 * 1024))); adev->gds.mem.total_size = adev->gds.mem.total_size << AMDGPU_GDS_SHIFT; adev->gds.mem.gfx_partition_size = adev->gds.mem.gfx_partition_size << AMDGPU_GDS_SHIFT; @@ -1203,13 +1415,13 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev) if (!adev->mman.initialized) return; amdgpu_ttm_debugfs_fini(adev); - if (adev->stollen_vga_memory) { - r = amdgpu_bo_reserve(adev->stollen_vga_memory, true); + if (adev->stolen_vga_memory) { + r = amdgpu_bo_reserve(adev->stolen_vga_memory, true); if (r == 0) { - amdgpu_bo_unpin(adev->stollen_vga_memory); - amdgpu_bo_unreserve(adev->stollen_vga_memory); + amdgpu_bo_unpin(adev->stolen_vga_memory); + amdgpu_bo_unreserve(adev->stolen_vga_memory); } - amdgpu_bo_unref(&adev->stollen_vga_memory); + amdgpu_bo_unref(&adev->stolen_vga_memory); } ttm_bo_clean_mm(&adev->mman.bdev, TTM_PL_VRAM); ttm_bo_clean_mm(&adev->mman.bdev, TTM_PL_TT); @@ -1256,12 +1468,77 @@ int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma) return ttm_bo_mmap(filp, vma, &adev->mman.bdev); } -int amdgpu_copy_buffer(struct amdgpu_ring *ring, - uint64_t src_offset, - uint64_t dst_offset, - uint32_t byte_count, +static int amdgpu_map_buffer(struct ttm_buffer_object *bo, + struct ttm_mem_reg *mem, unsigned num_pages, + uint64_t offset, unsigned window, + struct amdgpu_ring *ring, + uint64_t *addr) +{ + struct amdgpu_ttm_tt *gtt = (void *)bo->ttm; + struct amdgpu_device *adev = ring->adev; + struct ttm_tt *ttm = bo->ttm; + struct amdgpu_job *job; + unsigned num_dw, num_bytes; + dma_addr_t *dma_address; + struct dma_fence *fence; + uint64_t src_addr, dst_addr; + uint64_t flags; + int r; + + BUG_ON(adev->mman.buffer_funcs->copy_max_bytes < + AMDGPU_GTT_MAX_TRANSFER_SIZE * 8); + + *addr = adev->mc.gart_start; + *addr += (u64)window * AMDGPU_GTT_MAX_TRANSFER_SIZE * + AMDGPU_GPU_PAGE_SIZE; + + num_dw = adev->mman.buffer_funcs->copy_num_dw; + while (num_dw & 0x7) + num_dw++; + + num_bytes = num_pages * 8; + + r = amdgpu_job_alloc_with_ib(adev, num_dw * 4 + num_bytes, &job); + if (r) + return r; + + src_addr = num_dw * 4; + src_addr += job->ibs[0].gpu_addr; + + dst_addr = adev->gart.table_addr; + dst_addr += window * AMDGPU_GTT_MAX_TRANSFER_SIZE * 8; + amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_addr, + dst_addr, num_bytes); + + amdgpu_ring_pad_ib(ring, &job->ibs[0]); + WARN_ON(job->ibs[0].length_dw > num_dw); + + dma_address = >t->ttm.dma_address[offset >> PAGE_SHIFT]; + flags = amdgpu_ttm_tt_pte_flags(adev, ttm, mem); + r = amdgpu_gart_map(adev, 0, num_pages, dma_address, flags, + &job->ibs[0].ptr[num_dw]); + if (r) + goto error_free; + + r = amdgpu_job_submit(job, ring, &adev->mman.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, &fence); + if (r) + goto error_free; + + dma_fence_put(fence); + + return r; + +error_free: + amdgpu_job_free(job); + return r; +} + +int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset, + uint64_t dst_offset, uint32_t byte_count, struct reservation_object *resv, - struct dma_fence **fence, bool direct_submit) + struct dma_fence **fence, bool direct_submit, + bool vm_needs_flush) { struct amdgpu_device *adev = ring->adev; struct amdgpu_job *job; @@ -1283,9 +1560,11 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, if (r) return r; + job->vm_needs_flush = vm_needs_flush; if (resv) { r = amdgpu_sync_resv(adev, &job->sync, resv, - AMDGPU_FENCE_OWNER_UNDEFINED); + AMDGPU_FENCE_OWNER_UNDEFINED, + false); if (r) { DRM_ERROR("sync failed (%d).\n", r); goto error_free; @@ -1327,12 +1606,13 @@ error_free: } int amdgpu_fill_buffer(struct amdgpu_bo *bo, - uint32_t src_data, + uint64_t src_data, struct reservation_object *resv, struct dma_fence **fence) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - uint32_t max_bytes = adev->mman.buffer_funcs->fill_max_bytes; + uint32_t max_bytes = 8 * + adev->vm_manager.vm_pte_funcs->set_max_nums_pte_pde; struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; struct drm_mm_node *mm_node; @@ -1347,6 +1627,12 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo, return -EINVAL; } + if (bo->tbo.mem.mem_type == TTM_PL_TT) { + r = amdgpu_ttm_bind(&bo->tbo, &bo->tbo.mem); + if (r) + return r; + } + num_pages = bo->tbo.num_pages; mm_node = bo->tbo.mem.mm_node; num_loops = 0; @@ -1357,7 +1643,9 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo, num_pages -= mm_node->size; ++mm_node; } - num_dw = num_loops * adev->mman.buffer_funcs->fill_num_dw; + + /* num of dwords for each SDMA_OP_PTEPDE cmd */ + num_dw = num_loops * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw; /* for IB padding */ num_dw += 64; @@ -1368,7 +1656,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo, if (resv) { r = amdgpu_sync_resv(adev, &job->sync, resv, - AMDGPU_FENCE_OWNER_UNDEFINED); + AMDGPU_FENCE_OWNER_UNDEFINED, false); if (r) { DRM_ERROR("sync failed (%d).\n", r); goto error_free; @@ -1382,16 +1670,16 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo, uint32_t byte_count = mm_node->size << PAGE_SHIFT; uint64_t dst_addr; - r = amdgpu_mm_node_addr(&bo->tbo, mm_node, - &bo->tbo.mem, &dst_addr); - if (r) - return r; + WARN_ONCE(byte_count & 0x7, "size should be a multiple of 8"); + dst_addr = amdgpu_mm_node_addr(&bo->tbo, mm_node, &bo->tbo.mem); while (byte_count) { uint32_t cur_size_in_bytes = min(byte_count, max_bytes); - amdgpu_emit_fill_buffer(adev, &job->ibs[0], src_data, - dst_addr, cur_size_in_bytes); + amdgpu_vm_set_pte_pde(adev, &job->ibs[0], + dst_addr, 0, + cur_size_in_bytes >> 3, 0, + src_data); dst_addr += cur_size_in_bytes; byte_count -= cur_size_in_bytes; @@ -1417,32 +1705,16 @@ error_free: #if defined(CONFIG_DEBUG_FS) -extern void amdgpu_gtt_mgr_print(struct seq_file *m, struct ttm_mem_type_manager - *man); static int amdgpu_mm_dump_table(struct seq_file *m, void *data) { struct drm_info_node *node = (struct drm_info_node *)m->private; unsigned ttm_pl = *(int *)node->info_ent->data; struct drm_device *dev = node->minor->dev; struct amdgpu_device *adev = dev->dev_private; - struct drm_mm *mm = (struct drm_mm *)adev->mman.bdev.man[ttm_pl].priv; - struct ttm_bo_global *glob = adev->mman.bdev.glob; + struct ttm_mem_type_manager *man = &adev->mman.bdev.man[ttm_pl]; struct drm_printer p = drm_seq_file_printer(m); - spin_lock(&glob->lru_lock); - drm_mm_print(mm, &p); - spin_unlock(&glob->lru_lock); - switch (ttm_pl) { - case TTM_PL_VRAM: - seq_printf(m, "man size:%llu pages, ram usage:%lluMB, vis usage:%lluMB\n", - adev->mman.bdev.man[ttm_pl].size, - (u64)atomic64_read(&adev->vram_usage) >> 20, - (u64)atomic64_read(&adev->vram_vis_usage) >> 20); - break; - case TTM_PL_TT: - amdgpu_gtt_mgr_print(m, &adev->mman.bdev.man[TTM_PL_TT]); - break; - } + man->func->debug(man, &p); return 0; } @@ -1479,9 +1751,9 @@ static ssize_t amdgpu_ttm_vram_read(struct file *f, char __user *buf, return result; spin_lock_irqsave(&adev->mmio_idx_lock, flags); - WREG32(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000); - WREG32(mmMM_INDEX_HI, *pos >> 31); - value = RREG32(mmMM_DATA); + WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000); + WREG32_NO_KIQ(mmMM_INDEX_HI, *pos >> 31); + value = RREG32_NO_KIQ(mmMM_DATA); spin_unlock_irqrestore(&adev->mmio_idx_lock, flags); r = put_user(value, (uint32_t *)buf); @@ -1497,10 +1769,50 @@ static ssize_t amdgpu_ttm_vram_read(struct file *f, char __user *buf, return result; } +static ssize_t amdgpu_ttm_vram_write(struct file *f, const char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = file_inode(f)->i_private; + ssize_t result = 0; + int r; + + if (size & 0x3 || *pos & 0x3) + return -EINVAL; + + if (*pos >= adev->mc.mc_vram_size) + return -ENXIO; + + while (size) { + unsigned long flags; + uint32_t value; + + if (*pos >= adev->mc.mc_vram_size) + return result; + + r = get_user(value, (uint32_t *)buf); + if (r) + return r; + + spin_lock_irqsave(&adev->mmio_idx_lock, flags); + WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000); + WREG32_NO_KIQ(mmMM_INDEX_HI, *pos >> 31); + WREG32_NO_KIQ(mmMM_DATA, value); + spin_unlock_irqrestore(&adev->mmio_idx_lock, flags); + + result += 4; + buf += 4; + *pos += 4; + size -= 4; + } + + return result; +} + static const struct file_operations amdgpu_ttm_vram_fops = { .owner = THIS_MODULE, .read = amdgpu_ttm_vram_read, - .llseek = default_llseek + .write = amdgpu_ttm_vram_write, + .llseek = default_llseek, }; #ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS @@ -1552,6 +1864,53 @@ static const struct file_operations amdgpu_ttm_gtt_fops = { #endif +static ssize_t amdgpu_iova_to_phys_read(struct file *f, char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = file_inode(f)->i_private; + int r; + uint64_t phys; + struct iommu_domain *dom; + + // always return 8 bytes + if (size != 8) + return -EINVAL; + + // only accept page addresses + if (*pos & 0xFFF) + return -EINVAL; + + dom = iommu_get_domain_for_dev(adev->dev); + if (dom) + phys = iommu_iova_to_phys(dom, *pos); + else + phys = *pos; + + r = copy_to_user(buf, &phys, 8); + if (r) + return -EFAULT; + + return 8; +} + +static const struct file_operations amdgpu_ttm_iova_fops = { + .owner = THIS_MODULE, + .read = amdgpu_iova_to_phys_read, + .llseek = default_llseek +}; + +static const struct { + char *name; + const struct file_operations *fops; + int domain; +} ttm_debugfs_entries[] = { + { "amdgpu_vram", &amdgpu_ttm_vram_fops, TTM_PL_VRAM }, +#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS + { "amdgpu_gtt", &amdgpu_ttm_gtt_fops, TTM_PL_TT }, +#endif + { "amdgpu_iova", &amdgpu_ttm_iova_fops, TTM_PL_SYSTEM }, +}; + #endif static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev) @@ -1562,22 +1921,21 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev) struct drm_minor *minor = adev->ddev->primary; struct dentry *ent, *root = minor->debugfs_root; - ent = debugfs_create_file("amdgpu_vram", S_IFREG | S_IRUGO, root, - adev, &amdgpu_ttm_vram_fops); - if (IS_ERR(ent)) - return PTR_ERR(ent); - i_size_write(ent->d_inode, adev->mc.mc_vram_size); - adev->mman.vram = ent; - -#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS - ent = debugfs_create_file("amdgpu_gtt", S_IFREG | S_IRUGO, root, - adev, &amdgpu_ttm_gtt_fops); - if (IS_ERR(ent)) - return PTR_ERR(ent); - i_size_write(ent->d_inode, adev->mc.gtt_size); - adev->mman.gtt = ent; + for (count = 0; count < ARRAY_SIZE(ttm_debugfs_entries); count++) { + ent = debugfs_create_file( + ttm_debugfs_entries[count].name, + S_IFREG | S_IRUGO, root, + adev, + ttm_debugfs_entries[count].fops); + if (IS_ERR(ent)) + return PTR_ERR(ent); + if (ttm_debugfs_entries[count].domain == TTM_PL_VRAM) + i_size_write(ent->d_inode, adev->mc.mc_vram_size); + else if (ttm_debugfs_entries[count].domain == TTM_PL_TT) + i_size_write(ent->d_inode, adev->mc.gart_size); + adev->mman.debugfs_entries[count] = ent; + } -#endif count = ARRAY_SIZE(amdgpu_ttm_debugfs_list); #ifdef CONFIG_SWIOTLB @@ -1587,7 +1945,6 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev) return amdgpu_debugfs_add_files(adev, amdgpu_ttm_debugfs_list, count); #else - return 0; #endif } @@ -1595,14 +1952,9 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev) static void amdgpu_ttm_debugfs_fini(struct amdgpu_device *adev) { #if defined(CONFIG_DEBUG_FS) + unsigned i; - debugfs_remove(adev->mman.vram); - adev->mman.vram = NULL; - -#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS - debugfs_remove(adev->mman.gtt); - adev->mman.gtt = NULL; -#endif - + for (i = 0; i < ARRAY_SIZE(ttm_debugfs_entries); i++) + debugfs_remove(adev->mman.debugfs_entries[i]); #endif } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index 6bdede8ff12b..abd4084982a3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -24,6 +24,7 @@ #ifndef __AMDGPU_TTM_H__ #define __AMDGPU_TTM_H__ +#include "amdgpu.h" #include "gpu_scheduler.h" #define AMDGPU_PL_GDS (TTM_PL_PRIV + 0) @@ -34,6 +35,9 @@ #define AMDGPU_PL_FLAG_GWS (TTM_PL_FLAG_PRIV << 1) #define AMDGPU_PL_FLAG_OA (TTM_PL_FLAG_PRIV << 2) +#define AMDGPU_GTT_MAX_TRANSFER_SIZE 512 +#define AMDGPU_GTT_NUM_TRANSFER_WINDOWS 2 + struct amdgpu_mman { struct ttm_bo_global_ref bo_global_ref; struct drm_global_reference mem_global_ref; @@ -42,38 +46,68 @@ struct amdgpu_mman { bool initialized; #if defined(CONFIG_DEBUG_FS) - struct dentry *vram; - struct dentry *gtt; + struct dentry *debugfs_entries[8]; #endif /* buffer handling */ const struct amdgpu_buffer_funcs *buffer_funcs; struct amdgpu_ring *buffer_funcs_ring; + + struct mutex gtt_window_lock; /* Scheduler entity for buffer moves */ struct amd_sched_entity entity; }; +struct amdgpu_copy_mem { + struct ttm_buffer_object *bo; + struct ttm_mem_reg *mem; + unsigned long offset; +}; + extern const struct ttm_mem_type_manager_func amdgpu_gtt_mgr_func; extern const struct ttm_mem_type_manager_func amdgpu_vram_mgr_func; -int amdgpu_gtt_mgr_alloc(struct ttm_mem_type_manager *man, - struct ttm_buffer_object *tbo, - const struct ttm_place *place, - struct ttm_mem_reg *mem); +bool amdgpu_gtt_mgr_is_allocated(struct ttm_mem_reg *mem); +uint64_t amdgpu_gtt_mgr_usage(struct ttm_mem_type_manager *man); + +uint64_t amdgpu_vram_mgr_usage(struct ttm_mem_type_manager *man); +uint64_t amdgpu_vram_mgr_vis_usage(struct ttm_mem_type_manager *man); -int amdgpu_copy_buffer(struct amdgpu_ring *ring, - uint64_t src_offset, - uint64_t dst_offset, - uint32_t byte_count, +int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset, + uint64_t dst_offset, uint32_t byte_count, struct reservation_object *resv, - struct dma_fence **fence, bool direct_submit); + struct dma_fence **fence, bool direct_submit, + bool vm_needs_flush); +int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev, + struct amdgpu_copy_mem *src, + struct amdgpu_copy_mem *dst, + uint64_t size, + struct reservation_object *resv, + struct dma_fence **f); int amdgpu_fill_buffer(struct amdgpu_bo *bo, - uint32_t src_data, + uint64_t src_data, struct reservation_object *resv, struct dma_fence **fence); int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma); bool amdgpu_ttm_is_bound(struct ttm_tt *ttm); int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem); +int amdgpu_ttm_recover_gart(struct amdgpu_device *adev); + +int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages); +void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages); +void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm); +int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, + uint32_t flags); +bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm); +struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm); +bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, + unsigned long end); +bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, + int *last_invalidated); +bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm); +bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm); +uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, + struct ttm_mem_reg *mem); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c index 4f50eeb65855..65649026b836 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c @@ -270,19 +270,11 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type) else return AMDGPU_FW_LOAD_SMU; case CHIP_VEGA10: - if (!load_type) - return AMDGPU_FW_LOAD_DIRECT; - else - return AMDGPU_FW_LOAD_PSP; case CHIP_RAVEN: -#if 0 if (!load_type) return AMDGPU_FW_LOAD_DIRECT; else return AMDGPU_FW_LOAD_PSP; -#else - return AMDGPU_FW_LOAD_DIRECT; -#endif default: DRM_ERROR("Unknow firmware load type\n"); } @@ -362,52 +354,55 @@ static int amdgpu_ucode_patch_jt(struct amdgpu_firmware_info *ucode, (le32_to_cpu(header->jt_offset) * 4); memcpy(dst_addr, src_addr, le32_to_cpu(header->jt_size) * 4); - ucode->ucode_size += le32_to_cpu(header->jt_size) * 4; - return 0; } int amdgpu_ucode_init_bo(struct amdgpu_device *adev) { struct amdgpu_bo **bo = &adev->firmware.fw_buf; - uint64_t fw_mc_addr; - void *fw_buf_ptr = NULL; uint64_t fw_offset = 0; int i, err; struct amdgpu_firmware_info *ucode = NULL; const struct common_firmware_header *header = NULL; - err = amdgpu_bo_create(adev, adev->firmware.fw_size, PAGE_SIZE, true, - amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, bo); - if (err) { - dev_err(adev->dev, "(%d) Firmware buffer allocate failed\n", err); - goto failed; + if (!adev->firmware.fw_size) { + dev_warn(adev->dev, "No ip firmware need to load\n"); + return 0; } - err = amdgpu_bo_reserve(*bo, false); - if (err) { - dev_err(adev->dev, "(%d) Firmware buffer reserve failed\n", err); - goto failed_reserve; - } + if (!amdgpu_sriov_vf(adev) || !adev->in_sriov_reset) { + err = amdgpu_bo_create(adev, adev->firmware.fw_size, PAGE_SIZE, true, + amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, + NULL, NULL, 0, bo); + if (err) { + dev_err(adev->dev, "(%d) Firmware buffer allocate failed\n", err); + goto failed; + } - err = amdgpu_bo_pin(*bo, amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, - &fw_mc_addr); - if (err) { - dev_err(adev->dev, "(%d) Firmware buffer pin failed\n", err); - goto failed_pin; - } + err = amdgpu_bo_reserve(*bo, false); + if (err) { + dev_err(adev->dev, "(%d) Firmware buffer reserve failed\n", err); + goto failed_reserve; + } - err = amdgpu_bo_kmap(*bo, &fw_buf_ptr); - if (err) { - dev_err(adev->dev, "(%d) Firmware buffer kmap failed\n", err); - goto failed_kmap; - } + err = amdgpu_bo_pin(*bo, amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, + &adev->firmware.fw_buf_mc); + if (err) { + dev_err(adev->dev, "(%d) Firmware buffer pin failed\n", err); + goto failed_pin; + } - amdgpu_bo_unreserve(*bo); + err = amdgpu_bo_kmap(*bo, &adev->firmware.fw_buf_ptr); + if (err) { + dev_err(adev->dev, "(%d) Firmware buffer kmap failed\n", err); + goto failed_kmap; + } + + amdgpu_bo_unreserve(*bo); + } - memset(fw_buf_ptr, 0, adev->firmware.fw_size); + memset(adev->firmware.fw_buf_ptr, 0, adev->firmware.fw_size); /* * if SMU loaded firmware, it needn't add SMC, UVD, and VCE @@ -426,14 +421,14 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev) ucode = &adev->firmware.ucode[i]; if (ucode->fw) { header = (const struct common_firmware_header *)ucode->fw->data; - amdgpu_ucode_init_single_fw(adev, ucode, fw_mc_addr + fw_offset, - (void *)((uint8_t *)fw_buf_ptr + fw_offset)); + amdgpu_ucode_init_single_fw(adev, ucode, adev->firmware.fw_buf_mc + fw_offset, + adev->firmware.fw_buf_ptr + fw_offset); if (i == AMDGPU_UCODE_ID_CP_MEC1 && adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { const struct gfx_firmware_header_v1_0 *cp_hdr; cp_hdr = (const struct gfx_firmware_header_v1_0 *)ucode->fw->data; - amdgpu_ucode_patch_jt(ucode, fw_mc_addr + fw_offset, - fw_buf_ptr + fw_offset); + amdgpu_ucode_patch_jt(ucode, adev->firmware.fw_buf_mc + fw_offset, + adev->firmware.fw_buf_ptr + fw_offset); fw_offset += ALIGN(le32_to_cpu(cp_hdr->jt_size) << 2, PAGE_SIZE); } fw_offset += ALIGN(ucode->ucode_size, PAGE_SIZE); @@ -459,6 +454,9 @@ int amdgpu_ucode_fini_bo(struct amdgpu_device *adev) int i; struct amdgpu_firmware_info *ucode = NULL; + if (!adev->firmware.fw_size) + return 0; + for (i = 0; i < adev->firmware.max_ucodes; i++) { ucode = &adev->firmware.ucode[i]; if (ucode->fw) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index 2ca09f111f08..e8bd50cf9785 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -269,6 +269,7 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) int amdgpu_uvd_sw_fini(struct amdgpu_device *adev) { + int i; kfree(adev->uvd.saved_bo); amd_sched_entity_fini(&adev->uvd.ring.sched, &adev->uvd.entity); @@ -279,6 +280,9 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev) amdgpu_ring_fini(&adev->uvd.ring); + for (i = 0; i < AMDGPU_MAX_UVD_ENC_RINGS; ++i) + amdgpu_ring_fini(&adev->uvd.ring_enc[i]); + release_firmware(adev->uvd.fw); return 0; @@ -410,10 +414,10 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx) uint64_t addr = amdgpu_uvd_get_addr_from_ctx(ctx); int r = 0; - mapping = amdgpu_cs_find_mapping(ctx->parser, addr, &bo); - if (mapping == NULL) { + r = amdgpu_cs_find_mapping(ctx->parser, addr, &bo, &mapping); + if (r) { DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr); - return -EINVAL; + return r; } if (!ctx->parser->adev->uvd.address_64_bit) { @@ -588,6 +592,10 @@ static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg, } break; + case 8: /* MJPEG */ + min_dpb_size = 0; + break; + case 16: /* H265 */ image_size = (ALIGN(width, 16) * ALIGN(height, 16) * 3) / 2; image_size = ALIGN(image_size, 256); @@ -733,10 +741,10 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx) uint64_t addr = amdgpu_uvd_get_addr_from_ctx(ctx); int r; - mapping = amdgpu_cs_find_mapping(ctx->parser, addr, &bo); - if (mapping == NULL) { + r = amdgpu_cs_find_mapping(ctx->parser, addr, &bo, &mapping); + if (r) { DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr); - return -EINVAL; + return r; } start = amdgpu_bo_gpu_offset(bo); @@ -913,10 +921,6 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx) return -EINVAL; } - r = amdgpu_cs_sysvm_access_required(parser); - if (r) - return r; - ctx.parser = parser; ctx.buf_sizes = buf_sizes; ctx.ib_idx = ib_idx; @@ -1051,7 +1055,7 @@ int amdgpu_uvd_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, &bo); + NULL, NULL, 0, &bo); if (r) return r; @@ -1101,7 +1105,7 @@ int amdgpu_uvd_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, &bo); + NULL, NULL, 0, &bo); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index b692ad402252..2918de2f39ec 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -559,6 +559,7 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, uint32_t ib_idx, struct amdgpu_bo_va_mapping *mapping; struct amdgpu_bo *bo; uint64_t addr; + int r; if (index == 0xffffffff) index = 0; @@ -567,11 +568,11 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, uint32_t ib_idx, ((uint64_t)amdgpu_get_ib_value(p, ib_idx, hi)) << 32; addr += ((uint64_t)size) * ((uint64_t)index); - mapping = amdgpu_cs_find_mapping(p, addr, &bo); - if (mapping == NULL) { + r = amdgpu_cs_find_mapping(p, addr, &bo, &mapping); + if (r) { DRM_ERROR("Can't find BO for addr 0x%010Lx %d %d %d %d\n", addr, lo, hi, size, index); - return -EINVAL; + return r; } if ((addr + (uint64_t)size) > @@ -647,15 +648,11 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) uint32_t allocated = 0; uint32_t tmp, handle = 0; uint32_t *size = &tmp; - int i, r, idx = 0; + int i, r = 0, idx = 0; p->job->vm = NULL; ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo); - r = amdgpu_cs_sysvm_access_required(p); - if (r) - return r; - while (idx < ib->length_dw) { uint32_t len = amdgpu_get_ib_value(p, ib_idx, idx); uint32_t cmd = amdgpu_get_ib_value(p, ib_idx, idx + 1); @@ -937,9 +934,9 @@ int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring) unsigned i; int r, timeout = adev->usec_timeout; - /* workaround VCE ring test slow issue for sriov*/ + /* skip ring test for sriov*/ if (amdgpu_sriov_vf(adev)) - timeout *= 10; + return 0; r = amdgpu_ring_alloc(ring, 16); if (r) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c index 09190fadd228..041e0121590c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c @@ -209,9 +209,9 @@ static void amdgpu_vcn_idle_work_handler(struct work_struct *work) if (fences == 0) { if (adev->pm.dpm_enabled) { + /* might be used when with pg/cg amdgpu_dpm_enable_uvd(adev, false); - } else { - amdgpu_asic_set_uvd_clocks(adev, 0, 0); + */ } } else { schedule_delayed_work(&adev->vcn.idle_work, VCN_IDLE_TIMEOUT); @@ -223,12 +223,10 @@ void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; bool set_clocks = !cancel_delayed_work_sync(&adev->vcn.idle_work); - if (set_clocks) { - if (adev->pm.dpm_enabled) { - amdgpu_dpm_enable_uvd(adev, true); - } else { - amdgpu_asic_set_uvd_clocks(adev, 53300, 40000); - } + if (set_clocks && adev->pm.dpm_enabled) { + /* might be used when with pg/cg + amdgpu_dpm_enable_uvd(adev, true); + */ } } @@ -361,7 +359,7 @@ static int amdgpu_vcn_dec_get_create_msg(struct amdgpu_ring *ring, uint32_t hand AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, &bo); + NULL, NULL, 0, &bo); if (r) return r; @@ -413,7 +411,7 @@ static int amdgpu_vcn_dec_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, &bo); + NULL, NULL, 0, &bo); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c new file mode 100644 index 000000000000..7f7097931c6f --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c @@ -0,0 +1,86 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "amdgpu.h" +#include "amdgpu_vf_error.h" +#include "mxgpu_ai.h" + +void amdgpu_vf_error_put(struct amdgpu_device *adev, + uint16_t sub_error_code, + uint16_t error_flags, + uint64_t error_data) +{ + int index; + uint16_t error_code; + + if (!amdgpu_sriov_vf(adev)) + return; + + error_code = AMDGIM_ERROR_CODE(AMDGIM_ERROR_CATEGORY_VF, sub_error_code); + + mutex_lock(&adev->virt.vf_errors.lock); + index = adev->virt.vf_errors.write_count % AMDGPU_VF_ERROR_ENTRY_SIZE; + adev->virt.vf_errors.code [index] = error_code; + adev->virt.vf_errors.flags [index] = error_flags; + adev->virt.vf_errors.data [index] = error_data; + adev->virt.vf_errors.write_count ++; + mutex_unlock(&adev->virt.vf_errors.lock); +} + + +void amdgpu_vf_error_trans_all(struct amdgpu_device *adev) +{ + /* u32 pf2vf_flags = 0; */ + u32 data1, data2, data3; + int index; + + if ((NULL == adev) || (!amdgpu_sriov_vf(adev)) || + (!adev->virt.ops) || (!adev->virt.ops->trans_msg)) { + return; + } +/* + TODO: Enable these code when pv2vf_info is merged + AMDGPU_FW_VRAM_PF2VF_READ (adev, feature_flags, &pf2vf_flags); + if (!(pf2vf_flags & AMDGIM_FEATURE_ERROR_LOG_COLLECT)) { + return; + } +*/ + + mutex_lock(&adev->virt.vf_errors.lock); + /* The errors are overlay of array, correct read_count as full. */ + if (adev->virt.vf_errors.write_count - adev->virt.vf_errors.read_count > AMDGPU_VF_ERROR_ENTRY_SIZE) { + adev->virt.vf_errors.read_count = adev->virt.vf_errors.write_count - AMDGPU_VF_ERROR_ENTRY_SIZE; + } + + while (adev->virt.vf_errors.read_count < adev->virt.vf_errors.write_count) { + index =adev->virt.vf_errors.read_count % AMDGPU_VF_ERROR_ENTRY_SIZE; + data1 = AMDGIM_ERROR_CODE_FLAGS_TO_MAILBOX(adev->virt.vf_errors.code[index], + adev->virt.vf_errors.flags[index]); + data2 = adev->virt.vf_errors.data[index] & 0xFFFFFFFF; + data3 = (adev->virt.vf_errors.data[index] >> 32) & 0xFFFFFFFF; + + adev->virt.ops->trans_msg(adev, IDH_LOG_VF_ERROR, data1, data2, data3); + adev->virt.vf_errors.read_count ++; + } + mutex_unlock(&adev->virt.vf_errors.lock); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h new file mode 100644 index 000000000000..6436bd053325 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h @@ -0,0 +1,65 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __VF_ERROR_H__ +#define __VF_ERROR_H__ + +#define AMDGIM_ERROR_CODE_FLAGS_TO_MAILBOX(c,f) (((c & 0xFFFF) << 16) | (f & 0xFFFF)) +#define AMDGIM_ERROR_CODE(t,c) (((t&0xF)<<12)|(c&0xFFF)) + +/* Please keep enum same as AMD GIM driver */ +enum AMDGIM_ERROR_VF { + AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL = 0, + AMDGIM_ERROR_VF_NO_VBIOS, + AMDGIM_ERROR_VF_GPU_POST_ERROR, + AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, + AMDGIM_ERROR_VF_FENCE_INIT_FAIL, + + AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, + AMDGIM_ERROR_VF_IB_INIT_FAIL, + AMDGIM_ERROR_VF_AMDGPU_LATE_INIT_FAIL, + AMDGIM_ERROR_VF_ASIC_RESUME_FAIL, + AMDGIM_ERROR_VF_GPU_RESET_FAIL, + + AMDGIM_ERROR_VF_TEST, + AMDGIM_ERROR_VF_MAX +}; + +enum AMDGIM_ERROR_CATEGORY { + AMDGIM_ERROR_CATEGORY_NON_USED = 0, + AMDGIM_ERROR_CATEGORY_GIM, + AMDGIM_ERROR_CATEGORY_PF, + AMDGIM_ERROR_CATEGORY_VF, + AMDGIM_ERROR_CATEGORY_VBIOS, + AMDGIM_ERROR_CATEGORY_MONITOR, + + AMDGIM_ERROR_CATEGORY_MAX +}; + +void amdgpu_vf_error_put(struct amdgpu_device *adev, + uint16_t sub_error_code, + uint16_t error_flags, + uint64_t error_data); +void amdgpu_vf_error_trans_all (struct amdgpu_device *adev); + +#endif /* __VF_ERROR_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c index 8a081e162d13..6738df836a70 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c @@ -22,7 +22,7 @@ */ #include "amdgpu.h" -#define MAX_KIQ_REG_WAIT 100000 +#define MAX_KIQ_REG_WAIT 100000000 /* in usecs */ int amdgpu_allocate_static_csa(struct amdgpu_device *adev) { @@ -46,14 +46,14 @@ int amdgpu_allocate_static_csa(struct amdgpu_device *adev) * address within META_DATA init package to support SRIOV gfx preemption. */ -int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm) +int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct amdgpu_bo_va **bo_va) { - int r; - struct amdgpu_bo_va *bo_va; struct ww_acquire_ctx ticket; struct list_head list; struct amdgpu_bo_list_entry pd; struct ttm_validate_buffer csa_tv; + int r; INIT_LIST_HEAD(&list); INIT_LIST_HEAD(&csa_tv.head); @@ -69,34 +69,33 @@ int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm) return r; } - bo_va = amdgpu_vm_bo_add(adev, vm, adev->virt.csa_obj); - if (!bo_va) { + *bo_va = amdgpu_vm_bo_add(adev, vm, adev->virt.csa_obj); + if (!*bo_va) { ttm_eu_backoff_reservation(&ticket, &list); DRM_ERROR("failed to create bo_va for static CSA\n"); return -ENOMEM; } - r = amdgpu_vm_alloc_pts(adev, bo_va->vm, AMDGPU_CSA_VADDR, - AMDGPU_CSA_SIZE); + r = amdgpu_vm_alloc_pts(adev, (*bo_va)->base.vm, AMDGPU_CSA_VADDR, + AMDGPU_CSA_SIZE); if (r) { DRM_ERROR("failed to allocate pts for static CSA, err=%d\n", r); - amdgpu_vm_bo_rmv(adev, bo_va); + amdgpu_vm_bo_rmv(adev, *bo_va); ttm_eu_backoff_reservation(&ticket, &list); return r; } - r = amdgpu_vm_bo_map(adev, bo_va, AMDGPU_CSA_VADDR, 0,AMDGPU_CSA_SIZE, - AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE | - AMDGPU_PTE_EXECUTABLE); + r = amdgpu_vm_bo_map(adev, *bo_va, AMDGPU_CSA_VADDR, 0, AMDGPU_CSA_SIZE, + AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE | + AMDGPU_PTE_EXECUTABLE); if (r) { DRM_ERROR("failed to do bo_map on static CSA, err=%d\n", r); - amdgpu_vm_bo_rmv(adev, bo_va); + amdgpu_vm_bo_rmv(adev, *bo_va); ttm_eu_backoff_reservation(&ticket, &list); return r; } - vm->csa_bo_va = bo_va; ttm_eu_backoff_reservation(&ticket, &list); return 0; } @@ -115,27 +114,25 @@ void amdgpu_virt_init_setting(struct amdgpu_device *adev) uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) { signed long r; - uint32_t val; - struct dma_fence *f; + unsigned long flags; + uint32_t val, seq; struct amdgpu_kiq *kiq = &adev->gfx.kiq; struct amdgpu_ring *ring = &kiq->ring; BUG_ON(!ring->funcs->emit_rreg); - mutex_lock(&kiq->ring_mutex); + spin_lock_irqsave(&kiq->ring_lock, flags); amdgpu_ring_alloc(ring, 32); amdgpu_ring_emit_rreg(ring, reg); - amdgpu_fence_emit(ring, &f); + amdgpu_fence_emit_polling(ring, &seq); amdgpu_ring_commit(ring); - mutex_unlock(&kiq->ring_mutex); + spin_unlock_irqrestore(&kiq->ring_lock, flags); - r = dma_fence_wait_timeout(f, false, msecs_to_jiffies(MAX_KIQ_REG_WAIT)); - dma_fence_put(f); + r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT); if (r < 1) { - DRM_ERROR("wait for kiq fence error: %ld.\n", r); + DRM_ERROR("wait for kiq fence error: %ld\n", r); return ~0; } - val = adev->wb.wb[adev->virt.reg_val_offs]; return val; @@ -144,23 +141,23 @@ uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg) void amdgpu_virt_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v) { signed long r; - struct dma_fence *f; + unsigned long flags; + uint32_t seq; struct amdgpu_kiq *kiq = &adev->gfx.kiq; struct amdgpu_ring *ring = &kiq->ring; BUG_ON(!ring->funcs->emit_wreg); - mutex_lock(&kiq->ring_mutex); + spin_lock_irqsave(&kiq->ring_lock, flags); amdgpu_ring_alloc(ring, 32); amdgpu_ring_emit_wreg(ring, reg, v); - amdgpu_fence_emit(ring, &f); + amdgpu_fence_emit_polling(ring, &seq); amdgpu_ring_commit(ring); - mutex_unlock(&kiq->ring_mutex); + spin_unlock_irqrestore(&kiq->ring_lock, flags); - r = dma_fence_wait_timeout(f, false, msecs_to_jiffies(MAX_KIQ_REG_WAIT)); + r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT); if (r < 1) - DRM_ERROR("wait for kiq fence error: %ld.\n", r); - dma_fence_put(f); + DRM_ERROR("wait for kiq fence error: %ld\n", r); } /** @@ -275,3 +272,80 @@ void amdgpu_virt_free_mm_table(struct amdgpu_device *adev) (void *)&adev->virt.mm_table.cpu_addr); adev->virt.mm_table.gpu_addr = 0; } + + +int amdgpu_virt_fw_reserve_get_checksum(void *obj, + unsigned long obj_size, + unsigned int key, + unsigned int chksum) +{ + unsigned int ret = key; + unsigned long i = 0; + unsigned char *pos; + + pos = (char *)obj; + /* calculate checksum */ + for (i = 0; i < obj_size; ++i) + ret += *(pos + i); + /* minus the chksum itself */ + pos = (char *)&chksum; + for (i = 0; i < sizeof(chksum); ++i) + ret -= *(pos + i); + return ret; +} + +void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev) +{ + uint32_t pf2vf_ver = 0; + uint32_t pf2vf_size = 0; + uint32_t checksum = 0; + uint32_t checkval; + char *str; + + adev->virt.fw_reserve.p_pf2vf = NULL; + adev->virt.fw_reserve.p_vf2pf = NULL; + + if (adev->fw_vram_usage.va != NULL) { + adev->virt.fw_reserve.p_pf2vf = + (struct amdgim_pf2vf_info_header *)( + adev->fw_vram_usage.va + AMDGIM_DATAEXCHANGE_OFFSET); + pf2vf_ver = adev->virt.fw_reserve.p_pf2vf->version; + AMDGPU_FW_VRAM_PF2VF_READ(adev, header.size, &pf2vf_size); + AMDGPU_FW_VRAM_PF2VF_READ(adev, checksum, &checksum); + + /* pf2vf message must be in 4K */ + if (pf2vf_size > 0 && pf2vf_size < 4096) { + checkval = amdgpu_virt_fw_reserve_get_checksum( + adev->virt.fw_reserve.p_pf2vf, pf2vf_size, + adev->virt.fw_reserve.checksum_key, checksum); + if (checkval == checksum) { + adev->virt.fw_reserve.p_vf2pf = + ((void *)adev->virt.fw_reserve.p_pf2vf + + pf2vf_size); + memset((void *)adev->virt.fw_reserve.p_vf2pf, 0, + sizeof(amdgim_vf2pf_info)); + AMDGPU_FW_VRAM_VF2PF_WRITE(adev, header.version, + AMDGPU_FW_VRAM_VF2PF_VER); + AMDGPU_FW_VRAM_VF2PF_WRITE(adev, header.size, + sizeof(amdgim_vf2pf_info)); + AMDGPU_FW_VRAM_VF2PF_READ(adev, driver_version, + &str); +#ifdef MODULE + if (THIS_MODULE->version != NULL) + strcpy(str, THIS_MODULE->version); + else +#endif + strcpy(str, "N/A"); + AMDGPU_FW_VRAM_VF2PF_WRITE(adev, driver_cert, + 0); + AMDGPU_FW_VRAM_VF2PF_WRITE(adev, checksum, + amdgpu_virt_fw_reserve_get_checksum( + adev->virt.fw_reserve.p_vf2pf, + pf2vf_size, + adev->virt.fw_reserve.checksum_key, 0)); + } + } + } +} + + diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h index 9e1062edb76e..b89d37fc406f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h @@ -36,6 +36,18 @@ struct amdgpu_mm_table { uint64_t gpu_addr; }; +#define AMDGPU_VF_ERROR_ENTRY_SIZE 16 + +/* struct error_entry - amdgpu VF error information. */ +struct amdgpu_vf_error_buffer { + struct mutex lock; + int read_count; + int write_count; + uint16_t code[AMDGPU_VF_ERROR_ENTRY_SIZE]; + uint16_t flags[AMDGPU_VF_ERROR_ENTRY_SIZE]; + uint64_t data[AMDGPU_VF_ERROR_ENTRY_SIZE]; +}; + /** * struct amdgpu_virt_ops - amdgpu device virt operations */ @@ -43,8 +55,182 @@ struct amdgpu_virt_ops { int (*req_full_gpu)(struct amdgpu_device *adev, bool init); int (*rel_full_gpu)(struct amdgpu_device *adev, bool init); int (*reset_gpu)(struct amdgpu_device *adev); + void (*trans_msg)(struct amdgpu_device *adev, u32 req, u32 data1, u32 data2, u32 data3); +}; + +/* + * Firmware Reserve Frame buffer + */ +struct amdgpu_virt_fw_reserve { + struct amdgim_pf2vf_info_header *p_pf2vf; + struct amdgim_vf2pf_info_header *p_vf2pf; + unsigned int checksum_key; +}; +/* + * Defination between PF and VF + * Structures forcibly aligned to 4 to keep the same style as PF. + */ +#define AMDGIM_DATAEXCHANGE_OFFSET (64 * 1024) + +#define AMDGIM_GET_STRUCTURE_RESERVED_SIZE(total, u8, u16, u32, u64) \ + (total - (((u8)+3) / 4 + ((u16)+1) / 2 + (u32) + (u64)*2)) + +enum AMDGIM_FEATURE_FLAG { + /* GIM supports feature of Error log collecting */ + AMDGIM_FEATURE_ERROR_LOG_COLLECT = 0x1, + /* GIM supports feature of loading uCodes */ + AMDGIM_FEATURE_GIM_LOAD_UCODES = 0x2, }; +struct amdgim_pf2vf_info_header { + /* the total structure size in byte. */ + uint32_t size; + /* version of this structure, written by the GIM */ + uint32_t version; +} __aligned(4); +struct amdgim_pf2vf_info_v1 { + /* header contains size and version */ + struct amdgim_pf2vf_info_header header; + /* max_width * max_height */ + unsigned int uvd_enc_max_pixels_count; + /* 16x16 pixels/sec, codec independent */ + unsigned int uvd_enc_max_bandwidth; + /* max_width * max_height */ + unsigned int vce_enc_max_pixels_count; + /* 16x16 pixels/sec, codec independent */ + unsigned int vce_enc_max_bandwidth; + /* MEC FW position in kb from the start of visible frame buffer */ + unsigned int mecfw_kboffset; + /* The features flags of the GIM driver supports. */ + unsigned int feature_flags; + /* use private key from mailbox 2 to create chueksum */ + unsigned int checksum; +} __aligned(4); + +struct amdgim_pf2vf_info_v2 { + /* header contains size and version */ + struct amdgim_pf2vf_info_header header; + /* use private key from mailbox 2 to create chueksum */ + uint32_t checksum; + /* The features flags of the GIM driver supports. */ + uint32_t feature_flags; + /* max_width * max_height */ + uint32_t uvd_enc_max_pixels_count; + /* 16x16 pixels/sec, codec independent */ + uint32_t uvd_enc_max_bandwidth; + /* max_width * max_height */ + uint32_t vce_enc_max_pixels_count; + /* 16x16 pixels/sec, codec independent */ + uint32_t vce_enc_max_bandwidth; + /* MEC FW position in kb from the start of VF visible frame buffer */ + uint64_t mecfw_kboffset; + /* MEC FW size in KB */ + uint32_t mecfw_ksize; + /* UVD FW position in kb from the start of VF visible frame buffer */ + uint64_t uvdfw_kboffset; + /* UVD FW size in KB */ + uint32_t uvdfw_ksize; + /* VCE FW position in kb from the start of VF visible frame buffer */ + uint64_t vcefw_kboffset; + /* VCE FW size in KB */ + uint32_t vcefw_ksize; + uint32_t reserved[AMDGIM_GET_STRUCTURE_RESERVED_SIZE(256, 0, 0, (9 + sizeof(struct amdgim_pf2vf_info_header)/sizeof(uint32_t)), 3)]; +} __aligned(4); + + +struct amdgim_vf2pf_info_header { + /* the total structure size in byte. */ + uint32_t size; + /*version of this structure, written by the guest */ + uint32_t version; +} __aligned(4); + +struct amdgim_vf2pf_info_v1 { + /* header contains size and version */ + struct amdgim_vf2pf_info_header header; + /* driver version */ + char driver_version[64]; + /* driver certification, 1=WHQL, 0=None */ + unsigned int driver_cert; + /* guest OS type and version: need a define */ + unsigned int os_info; + /* in the unit of 1M */ + unsigned int fb_usage; + /* guest gfx engine usage percentage */ + unsigned int gfx_usage; + /* guest gfx engine health percentage */ + unsigned int gfx_health; + /* guest compute engine usage percentage */ + unsigned int compute_usage; + /* guest compute engine health percentage */ + unsigned int compute_health; + /* guest vce engine usage percentage. 0xffff means N/A. */ + unsigned int vce_enc_usage; + /* guest vce engine health percentage. 0xffff means N/A. */ + unsigned int vce_enc_health; + /* guest uvd engine usage percentage. 0xffff means N/A. */ + unsigned int uvd_enc_usage; + /* guest uvd engine usage percentage. 0xffff means N/A. */ + unsigned int uvd_enc_health; + unsigned int checksum; +} __aligned(4); + +struct amdgim_vf2pf_info_v2 { + /* header contains size and version */ + struct amdgim_vf2pf_info_header header; + uint32_t checksum; + /* driver version */ + uint8_t driver_version[64]; + /* driver certification, 1=WHQL, 0=None */ + uint32_t driver_cert; + /* guest OS type and version: need a define */ + uint32_t os_info; + /* in the unit of 1M */ + uint32_t fb_usage; + /* guest gfx engine usage percentage */ + uint32_t gfx_usage; + /* guest gfx engine health percentage */ + uint32_t gfx_health; + /* guest compute engine usage percentage */ + uint32_t compute_usage; + /* guest compute engine health percentage */ + uint32_t compute_health; + /* guest vce engine usage percentage. 0xffff means N/A. */ + uint32_t vce_enc_usage; + /* guest vce engine health percentage. 0xffff means N/A. */ + uint32_t vce_enc_health; + /* guest uvd engine usage percentage. 0xffff means N/A. */ + uint32_t uvd_enc_usage; + /* guest uvd engine usage percentage. 0xffff means N/A. */ + uint32_t uvd_enc_health; + uint32_t reserved[AMDGIM_GET_STRUCTURE_RESERVED_SIZE(256, 64, 0, (12 + sizeof(struct amdgim_vf2pf_info_header)/sizeof(uint32_t)), 0)]; +} __aligned(4); + +#define AMDGPU_FW_VRAM_VF2PF_VER 2 +typedef struct amdgim_vf2pf_info_v2 amdgim_vf2pf_info ; + +#define AMDGPU_FW_VRAM_VF2PF_WRITE(adev, field, val) \ + do { \ + ((amdgim_vf2pf_info *)adev->virt.fw_reserve.p_vf2pf)->field = (val); \ + } while (0) + +#define AMDGPU_FW_VRAM_VF2PF_READ(adev, field, val) \ + do { \ + (*val) = ((amdgim_vf2pf_info *)adev->virt.fw_reserve.p_vf2pf)->field; \ + } while (0) + +#define AMDGPU_FW_VRAM_PF2VF_READ(adev, field, val) \ + do { \ + if (!adev->virt.fw_reserve.p_pf2vf) \ + *(val) = 0; \ + else { \ + if (adev->virt.fw_reserve.p_pf2vf->version == 1) \ + *(val) = ((struct amdgim_pf2vf_info_v1 *)adev->virt.fw_reserve.p_pf2vf)->field; \ + if (adev->virt.fw_reserve.p_pf2vf->version == 2) \ + *(val) = ((struct amdgim_pf2vf_info_v2 *)adev->virt.fw_reserve.p_pf2vf)->field; \ + } \ + } while (0) + /* GPU virtualization */ struct amdgpu_virt { uint32_t caps; @@ -58,6 +244,8 @@ struct amdgpu_virt { struct work_struct flr_work; struct amdgpu_mm_table mm_table; const struct amdgpu_virt_ops *ops; + struct amdgpu_vf_error_buffer vf_errors; + struct amdgpu_virt_fw_reserve fw_reserve; }; #define AMDGPU_CSA_SIZE (8 * 1024) @@ -89,7 +277,8 @@ static inline bool is_virtual_machine(void) struct amdgpu_vm; int amdgpu_allocate_static_csa(struct amdgpu_device *adev); -int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm); +int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct amdgpu_bo_va **bo_va); void amdgpu_virt_init_setting(struct amdgpu_device *adev); uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg); void amdgpu_virt_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v); @@ -99,5 +288,9 @@ int amdgpu_virt_reset_gpu(struct amdgpu_device *adev); int amdgpu_sriov_gpu_reset(struct amdgpu_device *adev, struct amdgpu_job *job); int amdgpu_virt_alloc_mm_table(struct amdgpu_device *adev); void amdgpu_virt_free_mm_table(struct amdgpu_device *adev); +int amdgpu_virt_fw_reserve_get_checksum(void *obj, unsigned long obj_size, + unsigned int key, + unsigned int chksum); +void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 5795f81369f0..c8c26f21993c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -27,12 +27,59 @@ */ #include <linux/dma-fence-array.h> #include <linux/interval_tree_generic.h> +#include <linux/idr.h> #include <drm/drmP.h> #include <drm/amdgpu_drm.h> #include "amdgpu.h" #include "amdgpu_trace.h" /* + * PASID manager + * + * PASIDs are global address space identifiers that can be shared + * between the GPU, an IOMMU and the driver. VMs on different devices + * may use the same PASID if they share the same address + * space. Therefore PASIDs are allocated using a global IDA. VMs are + * looked up from the PASID per amdgpu_device. + */ +static DEFINE_IDA(amdgpu_vm_pasid_ida); + +/** + * amdgpu_vm_alloc_pasid - Allocate a PASID + * @bits: Maximum width of the PASID in bits, must be at least 1 + * + * Allocates a PASID of the given width while keeping smaller PASIDs + * available if possible. + * + * Returns a positive integer on success. Returns %-EINVAL if bits==0. + * Returns %-ENOSPC if no PASID was available. Returns %-ENOMEM on + * memory allocation failure. + */ +int amdgpu_vm_alloc_pasid(unsigned int bits) +{ + int pasid = -EINVAL; + + for (bits = min(bits, 31U); bits > 0; bits--) { + pasid = ida_simple_get(&amdgpu_vm_pasid_ida, + 1U << (bits - 1), 1U << bits, + GFP_KERNEL); + if (pasid != -ENOSPC) + break; + } + + return pasid; +} + +/** + * amdgpu_vm_free_pasid - Free a PASID + * @pasid: PASID to free + */ +void amdgpu_vm_free_pasid(unsigned int pasid) +{ + ida_simple_remove(&amdgpu_vm_pasid_ida, pasid); +} + +/* * GPUVM * GPUVM is similar to the legacy gart on older asics, however * rather than there being a single global gart table @@ -77,8 +124,6 @@ struct amdgpu_pte_update_params { void (*func)(struct amdgpu_pte_update_params *params, uint64_t pe, uint64_t addr, unsigned count, uint32_t incr, uint64_t flags); - /* indicate update pt or its shadow */ - bool shadow; /* The next two are used during VM update by CPU * DMA addresses to use for mapping * Kernel pointer of PD/PT BO that needs to be updated @@ -142,7 +187,7 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, struct list_head *validated, struct amdgpu_bo_list_entry *entry) { - entry->robj = vm->root.bo; + entry->robj = vm->root.base.bo; entry->priority = 0; entry->tv.bo = &entry->robj->tbo; entry->tv.shared = true; @@ -151,47 +196,6 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, } /** - * amdgpu_vm_validate_layer - validate a single page table level - * - * @parent: parent page table level - * @validate: callback to do the validation - * @param: parameter for the validation callback - * - * Validate the page table BOs on command submission if neccessary. - */ -static int amdgpu_vm_validate_level(struct amdgpu_vm_pt *parent, - int (*validate)(void *, struct amdgpu_bo *), - void *param) -{ - unsigned i; - int r; - - if (!parent->entries) - return 0; - - for (i = 0; i <= parent->last_entry_used; ++i) { - struct amdgpu_vm_pt *entry = &parent->entries[i]; - - if (!entry->bo) - continue; - - r = validate(param, entry->bo); - if (r) - return r; - - /* - * Recurse into the sub directory. This is harmless because we - * have only a maximum of 5 layers. - */ - r = amdgpu_vm_validate_level(entry, validate, param); - if (r) - return r; - } - - return r; -} - -/** * amdgpu_vm_validate_pt_bos - validate the page table BOs * * @adev: amdgpu device pointer @@ -205,63 +209,70 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, int (*validate)(void *p, struct amdgpu_bo *bo), void *param) { - uint64_t num_evictions; - - /* We only need to validate the page tables - * if they aren't already valid. - */ - num_evictions = atomic64_read(&adev->num_evictions); - if (num_evictions == vm->last_eviction_counter) - return 0; + struct ttm_bo_global *glob = adev->mman.bdev.glob; + int r; - return amdgpu_vm_validate_level(&vm->root, validate, param); -} + spin_lock(&vm->status_lock); + while (!list_empty(&vm->evicted)) { + struct amdgpu_vm_bo_base *bo_base; + struct amdgpu_bo *bo; -/** - * amdgpu_vm_move_level_in_lru - move one level of PT BOs to the LRU tail - * - * @adev: amdgpu device instance - * @vm: vm providing the BOs - * - * Move the PT BOs to the tail of the LRU. - */ -static void amdgpu_vm_move_level_in_lru(struct amdgpu_vm_pt *parent) -{ - unsigned i; + bo_base = list_first_entry(&vm->evicted, + struct amdgpu_vm_bo_base, + vm_status); + spin_unlock(&vm->status_lock); - if (!parent->entries) - return; + bo = bo_base->bo; + BUG_ON(!bo); + if (bo->parent) { + r = validate(param, bo); + if (r) + return r; - for (i = 0; i <= parent->last_entry_used; ++i) { - struct amdgpu_vm_pt *entry = &parent->entries[i]; + spin_lock(&glob->lru_lock); + ttm_bo_move_to_lru_tail(&bo->tbo); + if (bo->shadow) + ttm_bo_move_to_lru_tail(&bo->shadow->tbo); + spin_unlock(&glob->lru_lock); + } - if (!entry->bo) - continue; + if (bo->tbo.type == ttm_bo_type_kernel && + vm->use_cpu_for_update) { + r = amdgpu_bo_kmap(bo, NULL); + if (r) + return r; + } - ttm_bo_move_to_lru_tail(&entry->bo->tbo); - amdgpu_vm_move_level_in_lru(entry); + spin_lock(&vm->status_lock); + if (bo->tbo.type != ttm_bo_type_kernel) + list_move(&bo_base->vm_status, &vm->moved); + else + list_move(&bo_base->vm_status, &vm->relocated); } + spin_unlock(&vm->status_lock); + + return 0; } /** - * amdgpu_vm_move_pt_bos_in_lru - move the PT BOs to the LRU tail + * amdgpu_vm_ready - check VM is ready for updates * - * @adev: amdgpu device instance - * @vm: vm providing the BOs + * @vm: VM to check * - * Move the PT BOs to the tail of the LRU. + * Check if all VM PDs/PTs are ready for updates */ -void amdgpu_vm_move_pt_bos_in_lru(struct amdgpu_device *adev, - struct amdgpu_vm *vm) +bool amdgpu_vm_ready(struct amdgpu_vm *vm) { - struct ttm_bo_global *glob = adev->mman.bdev.glob; + bool ready; + + spin_lock(&vm->status_lock); + ready = list_empty(&vm->evicted); + spin_unlock(&vm->status_lock); - spin_lock(&glob->lru_lock); - amdgpu_vm_move_level_in_lru(&vm->root); - spin_unlock(&glob->lru_lock); + return ready; } - /** +/** * amdgpu_vm_alloc_levels - allocate the PD/PT levels * * @adev: amdgpu_device pointer @@ -282,6 +293,7 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, unsigned pt_idx, from, to; int r; u64 flags; + uint64_t init_value = 0; if (!parent->entries) { unsigned num_entries = amdgpu_vm_num_entries(adev, level); @@ -315,28 +327,48 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, flags |= (AMDGPU_GEM_CREATE_NO_CPU_ACCESS | AMDGPU_GEM_CREATE_SHADOW); + if (vm->pte_support_ats) { + init_value = AMDGPU_PTE_DEFAULT_ATC; + if (level != adev->vm_manager.num_level - 1) + init_value |= AMDGPU_PDE_PTE; + + } + /* walk over the address space and allocate the page tables */ for (pt_idx = from; pt_idx <= to; ++pt_idx) { - struct reservation_object *resv = vm->root.bo->tbo.resv; + struct reservation_object *resv = vm->root.base.bo->tbo.resv; struct amdgpu_vm_pt *entry = &parent->entries[pt_idx]; struct amdgpu_bo *pt; - if (!entry->bo) { + if (!entry->base.bo) { r = amdgpu_bo_create(adev, amdgpu_vm_bo_size(adev, level), AMDGPU_GPU_PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM, flags, - NULL, resv, &pt); + NULL, resv, init_value, &pt); if (r) return r; + if (vm->use_cpu_for_update) { + r = amdgpu_bo_kmap(pt, NULL); + if (r) { + amdgpu_bo_unref(&pt); + return r; + } + } + /* Keep a reference to the root directory to avoid * freeing them up in the wrong order. */ - pt->parent = amdgpu_bo_ref(vm->root.bo); - - entry->bo = pt; + pt->parent = amdgpu_bo_ref(parent->base.bo); + + entry->base.vm = vm; + entry->base.bo = pt; + list_add_tail(&entry->base.bo_list, &pt->va); + spin_lock(&vm->status_lock); + list_add(&entry->base.vm_status, &vm->relocated); + spin_unlock(&vm->status_lock); entry->addr = 0; } @@ -424,7 +456,7 @@ static int amdgpu_vm_grab_reserved_vmid_locked(struct amdgpu_vm *vm, struct dma_fence *updates = sync->last_vm_update; int r = 0; struct dma_fence *flushed, *tmp; - bool needs_flush = false; + bool needs_flush = vm->use_cpu_for_update; flushed = id->flushed_updates; if ((amdgpu_vm_had_gpu_reset(adev, id)) || @@ -545,11 +577,11 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring, } kfree(fences); - job->vm_needs_flush = false; + job->vm_needs_flush = vm->use_cpu_for_update; /* Check if we can use a VMID already assigned to this VM */ list_for_each_entry_reverse(id, &id_mgr->ids_lru, list) { struct dma_fence *flushed; - bool needs_flush = false; + bool needs_flush = vm->use_cpu_for_update; /* Check all the prerequisites to using this VMID */ if (amdgpu_vm_had_gpu_reset(adev, id)) @@ -745,7 +777,7 @@ static bool amdgpu_vm_is_large_bar(struct amdgpu_device *adev) * * Emit a VM flush when it is necessary. */ -int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job) +int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_pipe_sync) { struct amdgpu_device *adev = ring->adev; unsigned vmhub = ring->funcs->vmhub; @@ -767,12 +799,15 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job) vm_flush_needed = true; } - if (!vm_flush_needed && !gds_switch_needed) + if (!vm_flush_needed && !gds_switch_needed && !need_pipe_sync) return 0; if (ring->funcs->init_cond_exec) patch_offset = amdgpu_ring_init_cond_exec(ring); + if (need_pipe_sync) + amdgpu_ring_emit_pipeline_sync(ring); + if (ring->funcs->emit_vm_flush && vm_flush_needed) { struct dma_fence *fence; @@ -874,8 +909,8 @@ struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm, { struct amdgpu_bo_va *bo_va; - list_for_each_entry(bo_va, &bo->va, bo_list) { - if (bo_va->vm == vm) { + list_for_each_entry(bo_va, &bo->va, base.bo_list) { + if (bo_va->base.vm == vm) { return bo_va; } } @@ -981,6 +1016,8 @@ static void amdgpu_vm_cpu_set_ptes(struct amdgpu_pte_update_params *params, unsigned int i; uint64_t value; + trace_amdgpu_vm_set_ptes(pe, addr, count, incr, flags); + for (i = 0; i < count; i++) { value = params->pages_addr ? amdgpu_vm_map_gart(params->pages_addr, addr) : @@ -989,19 +1026,16 @@ static void amdgpu_vm_cpu_set_ptes(struct amdgpu_pte_update_params *params, i, value, flags); addr += incr; } - - /* Flush HDP */ - mb(); - amdgpu_gart_flush_gpu_tlb(params->adev, 0); } -static int amdgpu_vm_bo_wait(struct amdgpu_device *adev, struct amdgpu_bo *bo) +static int amdgpu_vm_wait_pd(struct amdgpu_device *adev, struct amdgpu_vm *vm, + void *owner) { struct amdgpu_sync sync; int r; amdgpu_sync_create(&sync); - amdgpu_sync_resv(adev, &sync, bo->tbo.resv, AMDGPU_FENCE_OWNER_VM); + amdgpu_sync_resv(adev, &sync, vm->root.base.bo->tbo.resv, owner, false); r = amdgpu_sync_wait(&sync, true); amdgpu_sync_free(&sync); @@ -1020,18 +1054,17 @@ static int amdgpu_vm_bo_wait(struct amdgpu_device *adev, struct amdgpu_bo *bo) */ static int amdgpu_vm_update_level(struct amdgpu_device *adev, struct amdgpu_vm *vm, - struct amdgpu_vm_pt *parent, - unsigned level) + struct amdgpu_vm_pt *parent) { struct amdgpu_bo *shadow; struct amdgpu_ring *ring = NULL; uint64_t pd_addr, shadow_addr = 0; - uint32_t incr = amdgpu_vm_bo_size(adev, level + 1); uint64_t last_pde = ~0, last_pt = ~0, last_shadow = ~0; unsigned count = 0, pt_idx, ndw = 0; struct amdgpu_job *job; struct amdgpu_pte_update_params params; struct dma_fence *fence = NULL; + uint32_t incr; int r; @@ -1040,25 +1073,16 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, memset(¶ms, 0, sizeof(params)); params.adev = adev; - shadow = parent->bo->shadow; + shadow = parent->base.bo->shadow; - WARN_ON(vm->use_cpu_for_update && shadow); - if (vm->use_cpu_for_update && !shadow) { - r = amdgpu_bo_kmap(parent->bo, (void **)&pd_addr); - if (r) - return r; - r = amdgpu_vm_bo_wait(adev, parent->bo); - if (unlikely(r)) { - amdgpu_bo_kunmap(parent->bo); + if (vm->use_cpu_for_update) { + pd_addr = (unsigned long)amdgpu_bo_kptr(parent->base.bo); + r = amdgpu_vm_wait_pd(adev, vm, AMDGPU_FENCE_OWNER_VM); + if (unlikely(r)) return r; - } + params.func = amdgpu_vm_cpu_set_ptes; } else { - if (shadow) { - r = amdgpu_ttm_bind(&shadow->tbo, &shadow->tbo.mem); - if (r) - return r; - } ring = container_of(vm->entity.sched, struct amdgpu_ring, sched); @@ -1068,7 +1092,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, /* assume the worst case */ ndw += parent->last_entry_used * 6; - pd_addr = amdgpu_bo_gpu_offset(parent->bo); + pd_addr = amdgpu_bo_gpu_offset(parent->base.bo); if (shadow) { shadow_addr = amdgpu_bo_gpu_offset(shadow); @@ -1088,29 +1112,28 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, /* walk over the address space and update the directory */ for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) { - struct amdgpu_bo *bo = parent->entries[pt_idx].bo; + struct amdgpu_vm_pt *entry = &parent->entries[pt_idx]; + struct amdgpu_bo *bo = entry->base.bo; uint64_t pde, pt; if (bo == NULL) continue; - if (bo->shadow) { - struct amdgpu_bo *pt_shadow = bo->shadow; - - r = amdgpu_ttm_bind(&pt_shadow->tbo, - &pt_shadow->tbo.mem); - if (r) - return r; - } + spin_lock(&vm->status_lock); + list_del_init(&entry->base.vm_status); + spin_unlock(&vm->status_lock); pt = amdgpu_bo_gpu_offset(bo); pt = amdgpu_gart_get_vm_pde(adev, pt); - if (parent->entries[pt_idx].addr == pt) + /* Don't update huge pages here */ + if ((parent->entries[pt_idx].addr & AMDGPU_PDE_PTE) || + parent->entries[pt_idx].addr == (pt | AMDGPU_PTE_VALID)) continue; - parent->entries[pt_idx].addr = pt; + parent->entries[pt_idx].addr = pt | AMDGPU_PTE_VALID; pde = pd_addr + pt_idx * 8; + incr = amdgpu_bo_size(bo); if (((last_pde + 8 * count) != pde) || ((last_pt + incr * count) != pt) || (count == AMDGPU_VM_MAX_UPDATE_SIZE)) { @@ -1138,7 +1161,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, } if (count) { - if (vm->root.bo->shadow) + if (vm->root.base.bo->shadow) params.func(¶ms, last_shadow, last_pt, count, incr, AMDGPU_PTE_VALID); @@ -1146,42 +1169,29 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, count, incr, AMDGPU_PTE_VALID); } - if (params.func == amdgpu_vm_cpu_set_ptes) - amdgpu_bo_kunmap(parent->bo); - else if (params.ib->length_dw == 0) { - amdgpu_job_free(job); - } else { - amdgpu_ring_pad_ib(ring, params.ib); - amdgpu_sync_resv(adev, &job->sync, parent->bo->tbo.resv, - AMDGPU_FENCE_OWNER_VM); - if (shadow) - amdgpu_sync_resv(adev, &job->sync, shadow->tbo.resv, - AMDGPU_FENCE_OWNER_VM); - - WARN_ON(params.ib->length_dw > ndw); - r = amdgpu_job_submit(job, ring, &vm->entity, - AMDGPU_FENCE_OWNER_VM, &fence); - if (r) - goto error_free; - - amdgpu_bo_fence(parent->bo, fence, true); - dma_fence_put(vm->last_dir_update); - vm->last_dir_update = dma_fence_get(fence); - dma_fence_put(fence); - } - /* - * Recurse into the subdirectories. This recursion is harmless because - * we only have a maximum of 5 layers. - */ - for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) { - struct amdgpu_vm_pt *entry = &parent->entries[pt_idx]; - - if (!entry->bo) - continue; + if (!vm->use_cpu_for_update) { + if (params.ib->length_dw == 0) { + amdgpu_job_free(job); + } else { + amdgpu_ring_pad_ib(ring, params.ib); + amdgpu_sync_resv(adev, &job->sync, + parent->base.bo->tbo.resv, + AMDGPU_FENCE_OWNER_VM, false); + if (shadow) + amdgpu_sync_resv(adev, &job->sync, + shadow->tbo.resv, + AMDGPU_FENCE_OWNER_VM, false); + + WARN_ON(params.ib->length_dw > ndw); + r = amdgpu_job_submit(job, ring, &vm->entity, + AMDGPU_FENCE_OWNER_VM, &fence); + if (r) + goto error_free; - r = amdgpu_vm_update_level(adev, vm, entry, level + 1); - if (r) - return r; + amdgpu_bo_fence(parent->base.bo, fence, true); + dma_fence_put(vm->last_update); + vm->last_update = fence; + } } return 0; @@ -1198,7 +1208,8 @@ error_free: * * Mark all PD level as invalid after an error. */ -static void amdgpu_vm_invalidate_level(struct amdgpu_vm_pt *parent) +static void amdgpu_vm_invalidate_level(struct amdgpu_vm *vm, + struct amdgpu_vm_pt *parent) { unsigned pt_idx; @@ -1209,11 +1220,15 @@ static void amdgpu_vm_invalidate_level(struct amdgpu_vm_pt *parent) for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) { struct amdgpu_vm_pt *entry = &parent->entries[pt_idx]; - if (!entry->bo) + if (!entry->base.bo) continue; entry->addr = ~0ULL; - amdgpu_vm_invalidate_level(entry); + spin_lock(&vm->status_lock); + if (list_empty(&entry->base.vm_status)) + list_add(&entry->base.vm_status, &vm->relocated); + spin_unlock(&vm->status_lock); + amdgpu_vm_invalidate_level(vm, entry); } } @@ -1229,39 +1244,147 @@ static void amdgpu_vm_invalidate_level(struct amdgpu_vm_pt *parent) int amdgpu_vm_update_directories(struct amdgpu_device *adev, struct amdgpu_vm *vm) { - int r; + int r = 0; - r = amdgpu_vm_update_level(adev, vm, &vm->root, 0); - if (r) - amdgpu_vm_invalidate_level(&vm->root); + spin_lock(&vm->status_lock); + while (!list_empty(&vm->relocated)) { + struct amdgpu_vm_bo_base *bo_base; + struct amdgpu_bo *bo; + + bo_base = list_first_entry(&vm->relocated, + struct amdgpu_vm_bo_base, + vm_status); + spin_unlock(&vm->status_lock); + + bo = bo_base->bo->parent; + if (bo) { + struct amdgpu_vm_bo_base *parent; + struct amdgpu_vm_pt *pt; + + parent = list_first_entry(&bo->va, + struct amdgpu_vm_bo_base, + bo_list); + pt = container_of(parent, struct amdgpu_vm_pt, base); + + r = amdgpu_vm_update_level(adev, vm, pt); + if (r) { + amdgpu_vm_invalidate_level(vm, &vm->root); + return r; + } + spin_lock(&vm->status_lock); + } else { + spin_lock(&vm->status_lock); + list_del_init(&bo_base->vm_status); + } + } + spin_unlock(&vm->status_lock); + + if (vm->use_cpu_for_update) { + /* Flush HDP */ + mb(); + amdgpu_gart_flush_gpu_tlb(adev, 0); + } return r; } /** - * amdgpu_vm_find_pt - find the page table for an address + * amdgpu_vm_find_entry - find the entry for an address * * @p: see amdgpu_pte_update_params definition * @addr: virtual address in question + * @entry: resulting entry or NULL + * @parent: parent entry * - * Find the page table BO for a virtual address, return NULL when none found. + * Find the vm_pt entry and it's parent for the given address. */ -static struct amdgpu_bo *amdgpu_vm_get_pt(struct amdgpu_pte_update_params *p, - uint64_t addr) +void amdgpu_vm_get_entry(struct amdgpu_pte_update_params *p, uint64_t addr, + struct amdgpu_vm_pt **entry, + struct amdgpu_vm_pt **parent) { - struct amdgpu_vm_pt *entry = &p->vm->root; unsigned idx, level = p->adev->vm_manager.num_level; - while (entry->entries) { + *parent = NULL; + *entry = &p->vm->root; + while ((*entry)->entries) { idx = addr >> (p->adev->vm_manager.block_size * level--); - idx %= amdgpu_bo_size(entry->bo) / 8; - entry = &entry->entries[idx]; + idx %= amdgpu_bo_size((*entry)->base.bo) / 8; + *parent = *entry; + *entry = &(*entry)->entries[idx]; } if (level) - return NULL; + *entry = NULL; +} - return entry->bo; +/** + * amdgpu_vm_handle_huge_pages - handle updating the PD with huge pages + * + * @p: see amdgpu_pte_update_params definition + * @entry: vm_pt entry to check + * @parent: parent entry + * @nptes: number of PTEs updated with this operation + * @dst: destination address where the PTEs should point to + * @flags: access flags fro the PTEs + * + * Check if we can update the PD with a huge page. + */ +static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p, + struct amdgpu_vm_pt *entry, + struct amdgpu_vm_pt *parent, + unsigned nptes, uint64_t dst, + uint64_t flags) +{ + bool use_cpu_update = (p->func == amdgpu_vm_cpu_set_ptes); + uint64_t pd_addr, pde; + + /* In the case of a mixed PT the PDE must point to it*/ + if (p->adev->asic_type < CHIP_VEGA10 || + nptes != AMDGPU_VM_PTE_COUNT(p->adev) || + p->src || + !(flags & AMDGPU_PTE_VALID)) { + + dst = amdgpu_bo_gpu_offset(entry->base.bo); + dst = amdgpu_gart_get_vm_pde(p->adev, dst); + flags = AMDGPU_PTE_VALID; + } else { + /* Set the huge page flag to stop scanning at this PDE */ + flags |= AMDGPU_PDE_PTE; + } + + if (entry->addr == (dst | flags)) + return; + + entry->addr = (dst | flags); + + if (use_cpu_update) { + /* In case a huge page is replaced with a system + * memory mapping, p->pages_addr != NULL and + * amdgpu_vm_cpu_set_ptes would try to translate dst + * through amdgpu_vm_map_gart. But dst is already a + * GPU address (of the page table). Disable + * amdgpu_vm_map_gart temporarily. + */ + dma_addr_t *tmp; + + tmp = p->pages_addr; + p->pages_addr = NULL; + + pd_addr = (unsigned long)amdgpu_bo_kptr(parent->base.bo); + pde = pd_addr + (entry - parent->entries) * 8; + amdgpu_vm_cpu_set_ptes(p, pde, dst, 1, 0, flags); + + p->pages_addr = tmp; + } else { + if (parent->base.bo->shadow) { + pd_addr = amdgpu_bo_gpu_offset(parent->base.bo->shadow); + pde = pd_addr + (entry - parent->entries) * 8; + amdgpu_vm_do_set_ptes(p, pde, dst, 1, 0, flags); + } + pd_addr = amdgpu_bo_gpu_offset(parent->base.bo); + pde = pd_addr + (entry - parent->entries) * 8; + amdgpu_vm_do_set_ptes(p, pde, dst, 1, 0, flags); + } } /** @@ -1287,49 +1410,44 @@ static int amdgpu_vm_update_ptes(struct amdgpu_pte_update_params *params, uint64_t addr, pe_start; struct amdgpu_bo *pt; unsigned nptes; - int r; bool use_cpu_update = (params->func == amdgpu_vm_cpu_set_ptes); - /* walk over the address space and update the page tables */ - for (addr = start; addr < end; addr += nptes) { - pt = amdgpu_vm_get_pt(params, addr); - if (!pt) { - pr_err("PT not found, aborting update_ptes\n"); - return -EINVAL; - } - - if (params->shadow) { - if (WARN_ONCE(use_cpu_update, - "CPU VM update doesn't suuport shadow pages")) - return 0; + for (addr = start; addr < end; addr += nptes, + dst += nptes * AMDGPU_GPU_PAGE_SIZE) { + struct amdgpu_vm_pt *entry, *parent; - if (!pt->shadow) - return 0; - pt = pt->shadow; - } + amdgpu_vm_get_entry(params, addr, &entry, &parent); + if (!entry) + return -ENOENT; if ((addr & ~mask) == (end & ~mask)) nptes = end - addr; else nptes = AMDGPU_VM_PTE_COUNT(adev) - (addr & mask); + amdgpu_vm_handle_huge_pages(params, entry, parent, + nptes, dst, flags); + /* We don't need to update PTEs for huge pages */ + if (entry->addr & AMDGPU_PDE_PTE) + continue; + + pt = entry->base.bo; if (use_cpu_update) { - r = amdgpu_bo_kmap(pt, (void *)&pe_start); - if (r) - return r; - } else + pe_start = (unsigned long)amdgpu_bo_kptr(pt); + } else { + if (pt->shadow) { + pe_start = amdgpu_bo_gpu_offset(pt->shadow); + pe_start += (addr & mask) * 8; + params->func(params, pe_start, dst, nptes, + AMDGPU_GPU_PAGE_SIZE, flags); + } pe_start = amdgpu_bo_gpu_offset(pt); + } pe_start += (addr & mask) * 8; - params->func(params, pe_start, dst, nptes, AMDGPU_GPU_PAGE_SIZE, flags); - - dst += nptes * AMDGPU_GPU_PAGE_SIZE; - - if (use_cpu_update) - amdgpu_bo_kunmap(pt); } return 0; @@ -1350,8 +1468,6 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params *params, uint64_t start, uint64_t end, uint64_t dst, uint64_t flags) { - int r; - /** * The MC L1 TLB supports variable sized pages, based on a fragment * field in the PTE. When this field is set to a non-zero value, page @@ -1370,40 +1486,38 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params *params, * Userspace can support this by aligning virtual base address and * allocation size to the fragment size. */ - - /* SI and newer are optimized for 64KB */ - uint64_t frag_flags = AMDGPU_PTE_FRAG(AMDGPU_LOG2_PAGES_PER_FRAG); - uint64_t frag_align = 1 << AMDGPU_LOG2_PAGES_PER_FRAG; - - uint64_t frag_start = ALIGN(start, frag_align); - uint64_t frag_end = end & ~(frag_align - 1); + unsigned max_frag = params->adev->vm_manager.fragment_size; + int r; /* system pages are non continuously */ - if (params->src || !(flags & AMDGPU_PTE_VALID) || - (frag_start >= frag_end)) + if (params->src || !(flags & AMDGPU_PTE_VALID)) return amdgpu_vm_update_ptes(params, start, end, dst, flags); - /* handle the 4K area at the beginning */ - if (start != frag_start) { - r = amdgpu_vm_update_ptes(params, start, frag_start, - dst, flags); + while (start != end) { + uint64_t frag_flags, frag_end; + unsigned frag; + + /* This intentionally wraps around if no bit is set */ + frag = min((unsigned)ffs(start) - 1, + (unsigned)fls64(end - start) - 1); + if (frag >= max_frag) { + frag_flags = AMDGPU_PTE_FRAG(max_frag); + frag_end = end & ~((1ULL << max_frag) - 1); + } else { + frag_flags = AMDGPU_PTE_FRAG(frag); + frag_end = start + (1 << frag); + } + + r = amdgpu_vm_update_ptes(params, start, frag_end, dst, + flags | frag_flags); if (r) return r; - dst += (frag_start - start) * AMDGPU_GPU_PAGE_SIZE; - } - /* handle the area in the middle */ - r = amdgpu_vm_update_ptes(params, frag_start, frag_end, dst, - flags | frag_flags); - if (r) - return r; - - /* handle the 4K area at the end */ - if (frag_end != end) { - dst += (frag_end - frag_start) * AMDGPU_GPU_PAGE_SIZE; - r = amdgpu_vm_update_ptes(params, frag_end, end, dst, flags); + dst += (frag_end - start) * AMDGPU_GPU_PAGE_SIZE; + start = frag_end; } - return r; + + return 0; } /** @@ -1411,7 +1525,6 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params *params, * * @adev: amdgpu_device pointer * @exclusive: fence we need to sync to - * @src: address where to copy page table entries from * @pages_addr: DMA addresses to use for mapping * @vm: requested vm * @start: start of mapped range @@ -1425,7 +1538,6 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params *params, */ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, struct dma_fence *exclusive, - uint64_t src, dma_addr_t *pages_addr, struct amdgpu_vm *vm, uint64_t start, uint64_t last, @@ -1443,7 +1555,10 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, memset(¶ms, 0, sizeof(params)); params.adev = adev; params.vm = vm; - params.src = src; + + /* sync to everything on unmapping */ + if (!(flags & AMDGPU_PTE_VALID)) + owner = AMDGPU_FENCE_OWNER_UNDEFINED; if (vm->use_cpu_for_update) { /* params.src is used as flag to indicate system Memory */ @@ -1453,43 +1568,37 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, /* Wait for PT BOs to be free. PTs share the same resv. object * as the root PD BO */ - r = amdgpu_vm_bo_wait(adev, vm->root.bo); + r = amdgpu_vm_wait_pd(adev, vm, owner); if (unlikely(r)) return r; params.func = amdgpu_vm_cpu_set_ptes; params.pages_addr = pages_addr; - params.shadow = false; return amdgpu_vm_frag_ptes(¶ms, start, last + 1, addr, flags); } ring = container_of(vm->entity.sched, struct amdgpu_ring, sched); - /* sync to everything on unmapping */ - if (!(flags & AMDGPU_PTE_VALID)) - owner = AMDGPU_FENCE_OWNER_UNDEFINED; - nptes = last - start + 1; /* - * reserve space for one command every (1 << BLOCK_SIZE) + * reserve space for two commands every (1 << BLOCK_SIZE) * entries or 2k dwords (whatever is smaller) + * + * The second command is for the shadow pagetables. */ - ncmds = (nptes >> min(adev->vm_manager.block_size, 11u)) + 1; + ncmds = ((nptes >> min(adev->vm_manager.block_size, 11u)) + 1) * 2; /* padding, etc. */ ndw = 64; - if (src) { - /* only copy commands needed */ - ndw += ncmds * 7; + /* one PDE write for each huge page */ + ndw += ((nptes >> adev->vm_manager.block_size) + 1) * 6; - params.func = amdgpu_vm_do_copy_ptes; - - } else if (pages_addr) { + if (pages_addr) { /* copy commands needed */ - ndw += ncmds * 7; + ndw += ncmds * adev->vm_manager.vm_pte_funcs->copy_pte_num_dw; /* and also PTEs */ ndw += nptes * 2; @@ -1498,10 +1607,11 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, } else { /* set page commands needed */ - ndw += ncmds * 10; + ndw += ncmds * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw; - /* two extra commands for begin/end of fragment */ - ndw += 2 * 10; + /* extra commands for begin/end fragments */ + ndw += 2 * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw + * adev->vm_manager.fragment_size; params.func = amdgpu_vm_do_set_ptes; } @@ -1512,7 +1622,7 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, params.ib = &job->ibs[0]; - if (!src && pages_addr) { + if (pages_addr) { uint64_t *pte; unsigned i; @@ -1533,20 +1643,15 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (r) goto error_free; - r = amdgpu_sync_resv(adev, &job->sync, vm->root.bo->tbo.resv, - owner); + r = amdgpu_sync_resv(adev, &job->sync, vm->root.base.bo->tbo.resv, + owner, false); if (r) goto error_free; - r = reservation_object_reserve_shared(vm->root.bo->tbo.resv); + r = reservation_object_reserve_shared(vm->root.base.bo->tbo.resv); if (r) goto error_free; - params.shadow = true; - r = amdgpu_vm_frag_ptes(¶ms, start, last + 1, addr, flags); - if (r) - goto error_free; - params.shadow = false; r = amdgpu_vm_frag_ptes(¶ms, start, last + 1, addr, flags); if (r) goto error_free; @@ -1558,13 +1663,14 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (r) goto error_free; - amdgpu_bo_fence(vm->root.bo, f, true); + amdgpu_bo_fence(vm->root.base.bo, f, true); dma_fence_put(*fence); *fence = f; return 0; error_free: amdgpu_job_free(job); + amdgpu_vm_invalidate_level(vm, &vm->root); return r; } @@ -1573,7 +1679,6 @@ error_free: * * @adev: amdgpu_device pointer * @exclusive: fence we need to sync to - * @gtt_flags: flags as they are used for GTT * @pages_addr: DMA addresses to use for mapping * @vm: requested vm * @mapping: mapped range and flags to use for the update @@ -1587,7 +1692,6 @@ error_free: */ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, struct dma_fence *exclusive, - uint64_t gtt_flags, dma_addr_t *pages_addr, struct amdgpu_vm *vm, struct amdgpu_bo_va_mapping *mapping, @@ -1595,7 +1699,8 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, struct drm_mm_node *nodes, struct dma_fence **fence) { - uint64_t pfn, src = 0, start = mapping->start; + unsigned min_linear_pages = 1 << adev->vm_manager.fragment_size; + uint64_t pfn, start = mapping->start; int r; /* normally,bo_va->flags only contians READABLE and WIRTEABLE bit go here @@ -1629,6 +1734,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, } do { + dma_addr_t *dma_addr = NULL; uint64_t max_entries; uint64_t addr, last; @@ -1642,20 +1748,32 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, } if (pages_addr) { - if (flags == gtt_flags) - src = adev->gart.table_addr + - (addr >> AMDGPU_GPU_PAGE_SHIFT) * 8; - else - max_entries = min(max_entries, 16ull * 1024ull); - addr = 0; + uint64_t count; + + max_entries = min(max_entries, 16ull * 1024ull); + for (count = 1; count < max_entries; ++count) { + uint64_t idx = pfn + count; + + if (pages_addr[idx] != + (pages_addr[idx - 1] + PAGE_SIZE)) + break; + } + + if (count < min_linear_pages) { + addr = pfn << PAGE_SHIFT; + dma_addr = pages_addr; + } else { + addr = pages_addr[pfn]; + max_entries = count; + } + } else if (flags & AMDGPU_PTE_VALID) { addr += adev->vm_manager.vram_base_offset; + addr += pfn << PAGE_SHIFT; } - addr += pfn << PAGE_SHIFT; last = min((uint64_t)mapping->last, start + max_entries - 1); - r = amdgpu_vm_bo_update_mapping(adev, exclusive, - src, pages_addr, vm, + r = amdgpu_vm_bo_update_mapping(adev, exclusive, dma_addr, vm, start, last, flags, addr, fence); if (r) @@ -1687,71 +1805,77 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, bool clear) { - struct amdgpu_vm *vm = bo_va->vm; + struct amdgpu_bo *bo = bo_va->base.bo; + struct amdgpu_vm *vm = bo_va->base.vm; struct amdgpu_bo_va_mapping *mapping; dma_addr_t *pages_addr = NULL; - uint64_t gtt_flags, flags; struct ttm_mem_reg *mem; struct drm_mm_node *nodes; - struct dma_fence *exclusive; + struct dma_fence *exclusive, **last_update; + uint64_t flags; int r; - if (clear || !bo_va->bo) { + if (clear || !bo_va->base.bo) { mem = NULL; nodes = NULL; exclusive = NULL; } else { struct ttm_dma_tt *ttm; - mem = &bo_va->bo->tbo.mem; + mem = &bo_va->base.bo->tbo.mem; nodes = mem->mm_node; if (mem->mem_type == TTM_PL_TT) { - ttm = container_of(bo_va->bo->tbo.ttm, struct - ttm_dma_tt, ttm); + ttm = container_of(bo_va->base.bo->tbo.ttm, + struct ttm_dma_tt, ttm); pages_addr = ttm->dma_address; } - exclusive = reservation_object_get_excl(bo_va->bo->tbo.resv); + exclusive = reservation_object_get_excl(bo->tbo.resv); } - if (bo_va->bo) { - flags = amdgpu_ttm_tt_pte_flags(adev, bo_va->bo->tbo.ttm, mem); - gtt_flags = (amdgpu_ttm_is_bound(bo_va->bo->tbo.ttm) && - adev == amdgpu_ttm_adev(bo_va->bo->tbo.bdev)) ? - flags : 0; - } else { + if (bo) + flags = amdgpu_ttm_tt_pte_flags(adev, bo->tbo.ttm, mem); + else flags = 0x0; - gtt_flags = ~0x0; - } - spin_lock(&vm->status_lock); - if (!list_empty(&bo_va->vm_status)) + if (clear || (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv)) + last_update = &vm->last_update; + else + last_update = &bo_va->last_pt_update; + + if (!clear && bo_va->base.moved) { + bo_va->base.moved = false; list_splice_init(&bo_va->valids, &bo_va->invalids); - spin_unlock(&vm->status_lock); + + } else if (bo_va->cleared != clear) { + list_splice_init(&bo_va->valids, &bo_va->invalids); + } list_for_each_entry(mapping, &bo_va->invalids, list) { - r = amdgpu_vm_bo_split_mapping(adev, exclusive, - gtt_flags, pages_addr, vm, + r = amdgpu_vm_bo_split_mapping(adev, exclusive, pages_addr, vm, mapping, flags, nodes, - &bo_va->last_pt_update); + last_update); if (r) return r; } - if (trace_amdgpu_vm_bo_mapping_enabled()) { - list_for_each_entry(mapping, &bo_va->valids, list) - trace_amdgpu_vm_bo_mapping(mapping); - - list_for_each_entry(mapping, &bo_va->invalids, list) - trace_amdgpu_vm_bo_mapping(mapping); + if (vm->use_cpu_for_update) { + /* Flush HDP */ + mb(); + amdgpu_gart_flush_gpu_tlb(adev, 0); } spin_lock(&vm->status_lock); - list_splice_init(&bo_va->invalids, &bo_va->valids); - list_del_init(&bo_va->vm_status); - if (clear) - list_add(&bo_va->vm_status, &vm->cleared); + list_del_init(&bo_va->base.vm_status); spin_unlock(&vm->status_lock); + list_splice_init(&bo_va->invalids, &bo_va->valids); + bo_va->cleared = clear; + + if (trace_amdgpu_vm_bo_mapping_enabled()) { + list_for_each_entry(mapping, &bo_va->valids, list) + trace_amdgpu_vm_bo_mapping(mapping); + } + return 0; } @@ -1857,7 +1981,7 @@ static void amdgpu_vm_free_mapping(struct amdgpu_device *adev, */ static void amdgpu_vm_prt_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) { - struct reservation_object *resv = vm->root.bo->tbo.resv; + struct reservation_object *resv = vm->root.base.bo->tbo.resv; struct dma_fence *excl, **shared; unsigned i, shared_count; int r; @@ -1905,15 +2029,19 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, struct amdgpu_bo_va_mapping *mapping; struct dma_fence *f = NULL; int r; + uint64_t init_pte_value = 0; while (!list_empty(&vm->freed)) { mapping = list_first_entry(&vm->freed, struct amdgpu_bo_va_mapping, list); list_del(&mapping->list); - r = amdgpu_vm_bo_update_mapping(adev, NULL, 0, NULL, vm, + if (vm->pte_support_ats) + init_pte_value = AMDGPU_PTE_DEFAULT_ATC; + + r = amdgpu_vm_bo_update_mapping(adev, NULL, NULL, vm, mapping->start, mapping->last, - 0, 0, &f); + init_pte_value, 0, &f); amdgpu_vm_free_mapping(adev, vm, mapping, f); if (r) { dma_fence_put(f); @@ -1933,29 +2061,35 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, } /** - * amdgpu_vm_clear_invalids - clear invalidated BOs in the PT + * amdgpu_vm_handle_moved - handle moved BOs in the PT * * @adev: amdgpu_device pointer * @vm: requested vm + * @sync: sync object to add fences to * - * Make sure all invalidated BOs are cleared in the PT. + * Make sure all BOs which are moved are updated in the PTs. * Returns 0 for success. * - * PTs have to be reserved and mutex must be locked! + * PTs have to be reserved! */ -int amdgpu_vm_clear_invalids(struct amdgpu_device *adev, - struct amdgpu_vm *vm, struct amdgpu_sync *sync) +int amdgpu_vm_handle_moved(struct amdgpu_device *adev, + struct amdgpu_vm *vm) { - struct amdgpu_bo_va *bo_va = NULL; + bool clear; int r = 0; spin_lock(&vm->status_lock); - while (!list_empty(&vm->invalidated)) { - bo_va = list_first_entry(&vm->invalidated, - struct amdgpu_bo_va, vm_status); + while (!list_empty(&vm->moved)) { + struct amdgpu_bo_va *bo_va; + + bo_va = list_first_entry(&vm->moved, + struct amdgpu_bo_va, base.vm_status); spin_unlock(&vm->status_lock); - r = amdgpu_vm_bo_update(adev, bo_va, true); + /* Per VM BOs never need to bo cleared in the page tables */ + clear = bo_va->base.bo->tbo.resv != vm->root.base.bo->tbo.resv; + + r = amdgpu_vm_bo_update(adev, bo_va, clear); if (r) return r; @@ -1963,9 +2097,6 @@ int amdgpu_vm_clear_invalids(struct amdgpu_device *adev, } spin_unlock(&vm->status_lock); - if (bo_va) - r = amdgpu_sync_fence(adev, sync, bo_va->last_pt_update); - return r; } @@ -1992,20 +2123,54 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev, if (bo_va == NULL) { return NULL; } - bo_va->vm = vm; - bo_va->bo = bo; + bo_va->base.vm = vm; + bo_va->base.bo = bo; + INIT_LIST_HEAD(&bo_va->base.bo_list); + INIT_LIST_HEAD(&bo_va->base.vm_status); + bo_va->ref_count = 1; - INIT_LIST_HEAD(&bo_va->bo_list); INIT_LIST_HEAD(&bo_va->valids); INIT_LIST_HEAD(&bo_va->invalids); - INIT_LIST_HEAD(&bo_va->vm_status); if (bo) - list_add_tail(&bo_va->bo_list, &bo->va); + list_add_tail(&bo_va->base.bo_list, &bo->va); return bo_va; } + +/** + * amdgpu_vm_bo_insert_mapping - insert a new mapping + * + * @adev: amdgpu_device pointer + * @bo_va: bo_va to store the address + * @mapping: the mapping to insert + * + * Insert a new mapping into all structures. + */ +static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev, + struct amdgpu_bo_va *bo_va, + struct amdgpu_bo_va_mapping *mapping) +{ + struct amdgpu_vm *vm = bo_va->base.vm; + struct amdgpu_bo *bo = bo_va->base.bo; + + mapping->bo_va = bo_va; + list_add(&mapping->list, &bo_va->invalids); + amdgpu_vm_it_insert(mapping, &vm->va); + + if (mapping->flags & AMDGPU_PTE_PRT) + amdgpu_vm_prt_get(adev); + + if (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv) { + spin_lock(&vm->status_lock); + if (list_empty(&bo_va->base.vm_status)) + list_add(&bo_va->base.vm_status, &vm->moved); + spin_unlock(&vm->status_lock); + } + trace_amdgpu_vm_bo_map(bo_va, mapping); +} + /** * amdgpu_vm_bo_map - map bo inside a vm * @@ -2026,7 +2191,8 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev, uint64_t size, uint64_t flags) { struct amdgpu_bo_va_mapping *mapping, *tmp; - struct amdgpu_vm *vm = bo_va->vm; + struct amdgpu_bo *bo = bo_va->base.bo; + struct amdgpu_vm *vm = bo_va->base.vm; uint64_t eaddr; /* validate the parameters */ @@ -2037,7 +2203,7 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev, /* make sure object fit at this offset */ eaddr = saddr + size - 1; if (saddr >= eaddr || - (bo_va->bo && offset + size > amdgpu_bo_size(bo_va->bo))) + (bo && offset + size > amdgpu_bo_size(bo))) return -EINVAL; saddr /= AMDGPU_GPU_PAGE_SIZE; @@ -2047,7 +2213,7 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev, if (tmp) { /* bo and tmp overlap, invalid addr */ dev_err(adev->dev, "bo %p va 0x%010Lx-0x%010Lx conflict with " - "0x%010Lx-0x%010Lx\n", bo_va->bo, saddr, eaddr, + "0x%010Lx-0x%010Lx\n", bo, saddr, eaddr, tmp->start, tmp->last + 1); return -EINVAL; } @@ -2056,17 +2222,12 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev, if (!mapping) return -ENOMEM; - INIT_LIST_HEAD(&mapping->list); mapping->start = saddr; mapping->last = eaddr; mapping->offset = offset; mapping->flags = flags; - list_add(&mapping->list, &bo_va->invalids); - amdgpu_vm_it_insert(mapping, &vm->va); - - if (flags & AMDGPU_PTE_PRT) - amdgpu_vm_prt_get(adev); + amdgpu_vm_bo_insert_map(adev, bo_va, mapping); return 0; } @@ -2092,7 +2253,7 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev, uint64_t size, uint64_t flags) { struct amdgpu_bo_va_mapping *mapping; - struct amdgpu_vm *vm = bo_va->vm; + struct amdgpu_bo *bo = bo_va->base.bo; uint64_t eaddr; int r; @@ -2104,7 +2265,7 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev, /* make sure object fit at this offset */ eaddr = saddr + size - 1; if (saddr >= eaddr || - (bo_va->bo && offset + size > amdgpu_bo_size(bo_va->bo))) + (bo && offset + size > amdgpu_bo_size(bo))) return -EINVAL; /* Allocate all the needed memory */ @@ -2112,7 +2273,7 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev, if (!mapping) return -ENOMEM; - r = amdgpu_vm_bo_clear_mappings(adev, bo_va->vm, saddr, size); + r = amdgpu_vm_bo_clear_mappings(adev, bo_va->base.vm, saddr, size); if (r) { kfree(mapping); return r; @@ -2126,11 +2287,7 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev, mapping->offset = offset; mapping->flags = flags; - list_add(&mapping->list, &bo_va->invalids); - amdgpu_vm_it_insert(mapping, &vm->va); - - if (flags & AMDGPU_PTE_PRT) - amdgpu_vm_prt_get(adev); + amdgpu_vm_bo_insert_map(adev, bo_va, mapping); return 0; } @@ -2152,7 +2309,7 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev, uint64_t saddr) { struct amdgpu_bo_va_mapping *mapping; - struct amdgpu_vm *vm = bo_va->vm; + struct amdgpu_vm *vm = bo_va->base.vm; bool valid = true; saddr /= AMDGPU_GPU_PAGE_SIZE; @@ -2176,6 +2333,7 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev, list_del(&mapping->list); amdgpu_vm_it_remove(mapping, &vm->va); + mapping->bo_va = NULL; trace_amdgpu_vm_bo_unmap(bo_va, mapping); if (valid) @@ -2261,6 +2419,7 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev, if (tmp->last > eaddr) tmp->last = eaddr; + tmp->bo_va = NULL; list_add(&tmp->list, &vm->freed); trace_amdgpu_vm_bo_unmap(NULL, tmp); } @@ -2287,6 +2446,19 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev, } /** + * amdgpu_vm_bo_lookup_mapping - find mapping by address + * + * @vm: the requested VM + * + * Find a mapping by it's address. + */ +struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm, + uint64_t addr) +{ + return amdgpu_vm_it_iter_first(&vm->va, addr, addr); +} + +/** * amdgpu_vm_bo_rmv - remove a bo to a specific vm * * @adev: amdgpu_device pointer @@ -2300,17 +2472,18 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va) { struct amdgpu_bo_va_mapping *mapping, *next; - struct amdgpu_vm *vm = bo_va->vm; + struct amdgpu_vm *vm = bo_va->base.vm; - list_del(&bo_va->bo_list); + list_del(&bo_va->base.bo_list); spin_lock(&vm->status_lock); - list_del(&bo_va->vm_status); + list_del(&bo_va->base.vm_status); spin_unlock(&vm->status_lock); list_for_each_entry_safe(mapping, next, &bo_va->valids, list) { list_del(&mapping->list); amdgpu_vm_it_remove(mapping, &vm->va); + mapping->bo_va = NULL; trace_amdgpu_vm_bo_unmap(bo_va, mapping); list_add(&mapping->list, &vm->freed); } @@ -2335,15 +2508,37 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, * Mark @bo as invalid. */ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, - struct amdgpu_bo *bo) + struct amdgpu_bo *bo, bool evicted) { - struct amdgpu_bo_va *bo_va; + struct amdgpu_vm_bo_base *bo_base; + + list_for_each_entry(bo_base, &bo->va, bo_list) { + struct amdgpu_vm *vm = bo_base->vm; - list_for_each_entry(bo_va, &bo->va, bo_list) { - spin_lock(&bo_va->vm->status_lock); - if (list_empty(&bo_va->vm_status)) - list_add(&bo_va->vm_status, &bo_va->vm->invalidated); - spin_unlock(&bo_va->vm->status_lock); + bo_base->moved = true; + if (evicted && bo->tbo.resv == vm->root.base.bo->tbo.resv) { + spin_lock(&bo_base->vm->status_lock); + if (bo->tbo.type == ttm_bo_type_kernel) + list_move(&bo_base->vm_status, &vm->evicted); + else + list_move_tail(&bo_base->vm_status, + &vm->evicted); + spin_unlock(&bo_base->vm->status_lock); + continue; + } + + if (bo->tbo.type == ttm_bo_type_kernel) { + spin_lock(&bo_base->vm->status_lock); + if (list_empty(&bo_base->vm_status)) + list_add(&bo_base->vm_status, &vm->relocated); + spin_unlock(&bo_base->vm->status_lock); + continue; + } + + spin_lock(&bo_base->vm->status_lock); + if (list_empty(&bo_base->vm_status)) + list_add(&bo_base->vm_status, &vm->moved); + spin_unlock(&bo_base->vm->status_lock); } } @@ -2361,12 +2556,28 @@ static uint32_t amdgpu_vm_get_block_size(uint64_t vm_size) } /** - * amdgpu_vm_adjust_size - adjust vm size and block size + * amdgpu_vm_set_fragment_size - adjust fragment size in PTE + * + * @adev: amdgpu_device pointer + * @fragment_size_default: the default fragment size if it's set auto + */ +void amdgpu_vm_set_fragment_size(struct amdgpu_device *adev, + uint32_t fragment_size_default) +{ + if (amdgpu_vm_fragment_size == -1) + adev->vm_manager.fragment_size = fragment_size_default; + else + adev->vm_manager.fragment_size = amdgpu_vm_fragment_size; +} + +/** + * amdgpu_vm_adjust_size - adjust vm size, block size and fragment size * * @adev: amdgpu_device pointer * @vm_size: the default vm size if it's set auto */ -void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size) +void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size, + uint32_t fragment_size_default) { /* adjust vm size firstly */ if (amdgpu_vm_size == -1) @@ -2381,8 +2592,11 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size) else adev->vm_manager.block_size = amdgpu_vm_block_size; - DRM_INFO("vm size is %llu GB, block size is %u-bit\n", - adev->vm_manager.vm_size, adev->vm_manager.block_size); + amdgpu_vm_set_fragment_size(adev, fragment_size_default); + + DRM_INFO("vm size is %llu GB, block size is %u-bit, fragment size is %u-bit\n", + adev->vm_manager.vm_size, adev->vm_manager.block_size, + adev->vm_manager.fragment_size); } /** @@ -2395,7 +2609,7 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size) * Init @vm fields. */ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, - int vm_context) + int vm_context, unsigned int pasid) { const unsigned align = min(AMDGPU_VM_PTB_ALIGN_SIZE, AMDGPU_VM_PTE_COUNT(adev) * 8); @@ -2404,14 +2618,16 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct amd_sched_rq *rq; int r, i; u64 flags; + uint64_t init_pde_value = 0; - vm->va = RB_ROOT; + vm->va = RB_ROOT_CACHED; vm->client_id = atomic64_inc_return(&adev->vm_manager.client_counter); for (i = 0; i < AMDGPU_MAX_VMHUBS; i++) vm->reserved_vmid[i] = NULL; spin_lock_init(&vm->status_lock); - INIT_LIST_HEAD(&vm->invalidated); - INIT_LIST_HEAD(&vm->cleared); + INIT_LIST_HEAD(&vm->evicted); + INIT_LIST_HEAD(&vm->relocated); + INIT_LIST_HEAD(&vm->moved); INIT_LIST_HEAD(&vm->freed); /* create scheduler entity for page table updates */ @@ -2425,17 +2641,26 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, if (r) return r; - if (vm_context == AMDGPU_VM_CONTEXT_COMPUTE) + vm->pte_support_ats = false; + + if (vm_context == AMDGPU_VM_CONTEXT_COMPUTE) { vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode & AMDGPU_VM_USE_CPU_FOR_COMPUTE); - else + + if (adev->asic_type == CHIP_RAVEN) { + vm->pte_support_ats = true; + init_pde_value = AMDGPU_PTE_DEFAULT_ATC + | AMDGPU_PDE_PTE; + + } + } else vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode & AMDGPU_VM_USE_CPU_FOR_GFX); DRM_DEBUG_DRIVER("VM update mode is %s\n", vm->use_cpu_for_update ? "CPU" : "SDMA"); WARN_ONCE((vm->use_cpu_for_update & !amdgpu_vm_is_large_bar(adev)), "CPU update of VM recommended only for large BAR system\n"); - vm->last_dir_update = NULL; + vm->last_update = NULL; flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | AMDGPU_GEM_CREATE_VRAM_CLEARED; @@ -2448,23 +2673,47 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, r = amdgpu_bo_create(adev, amdgpu_vm_bo_size(adev, 0), align, true, AMDGPU_GEM_DOMAIN_VRAM, flags, - NULL, NULL, &vm->root.bo); + NULL, NULL, init_pde_value, &vm->root.base.bo); if (r) goto error_free_sched_entity; - r = amdgpu_bo_reserve(vm->root.bo, false); - if (r) - goto error_free_root; + vm->root.base.vm = vm; + list_add_tail(&vm->root.base.bo_list, &vm->root.base.bo->va); + INIT_LIST_HEAD(&vm->root.base.vm_status); + + if (vm->use_cpu_for_update) { + r = amdgpu_bo_reserve(vm->root.base.bo, false); + if (r) + goto error_free_root; + + r = amdgpu_bo_kmap(vm->root.base.bo, NULL); + amdgpu_bo_unreserve(vm->root.base.bo); + if (r) + goto error_free_root; + } + + if (pasid) { + unsigned long flags; - vm->last_eviction_counter = atomic64_read(&adev->num_evictions); - amdgpu_bo_unreserve(vm->root.bo); + spin_lock_irqsave(&adev->vm_manager.pasid_lock, flags); + r = idr_alloc(&adev->vm_manager.pasid_idr, vm, pasid, pasid + 1, + GFP_ATOMIC); + spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags); + if (r < 0) + goto error_free_root; + + vm->pasid = pasid; + } + + INIT_KFIFO(vm->faults); + vm->fault_credit = 16; return 0; error_free_root: - amdgpu_bo_unref(&vm->root.bo->shadow); - amdgpu_bo_unref(&vm->root.bo); - vm->root.bo = NULL; + amdgpu_bo_unref(&vm->root.base.bo->shadow); + amdgpu_bo_unref(&vm->root.base.bo); + vm->root.base.bo = NULL; error_free_sched_entity: amd_sched_entity_fini(&ring->sched, &vm->entity); @@ -2483,9 +2732,11 @@ static void amdgpu_vm_free_levels(struct amdgpu_vm_pt *level) { unsigned i; - if (level->bo) { - amdgpu_bo_unref(&level->bo->shadow); - amdgpu_bo_unref(&level->bo); + if (level->base.bo) { + list_del(&level->base.bo_list); + list_del(&level->base.vm_status); + amdgpu_bo_unref(&level->base.bo->shadow); + amdgpu_bo_unref(&level->base.bo); } if (level->entries) @@ -2508,14 +2759,29 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) { struct amdgpu_bo_va_mapping *mapping, *tmp; bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; - int i; + struct amdgpu_bo *root; + u64 fault; + int i, r; + + /* Clear pending page faults from IH when the VM is destroyed */ + while (kfifo_get(&vm->faults, &fault)) + amdgpu_ih_clear_fault(adev, fault); + + if (vm->pasid) { + unsigned long flags; + + spin_lock_irqsave(&adev->vm_manager.pasid_lock, flags); + idr_remove(&adev->vm_manager.pasid_idr, vm->pasid); + spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags); + } amd_sched_entity_fini(vm->entity.sched, &vm->entity); - if (!RB_EMPTY_ROOT(&vm->va)) { + if (!RB_EMPTY_ROOT(&vm->va.rb_root)) { dev_err(adev->dev, "still active bo inside vm\n"); } - rbtree_postorder_for_each_entry_safe(mapping, tmp, &vm->va, rb) { + rbtree_postorder_for_each_entry_safe(mapping, tmp, + &vm->va.rb_root, rb) { list_del(&mapping->list); amdgpu_vm_it_remove(mapping, &vm->va); kfree(mapping); @@ -2530,13 +2796,51 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) amdgpu_vm_free_mapping(adev, vm, mapping, NULL); } - amdgpu_vm_free_levels(&vm->root); - dma_fence_put(vm->last_dir_update); + root = amdgpu_bo_ref(vm->root.base.bo); + r = amdgpu_bo_reserve(root, true); + if (r) { + dev_err(adev->dev, "Leaking page tables because BO reservation failed\n"); + } else { + amdgpu_vm_free_levels(&vm->root); + amdgpu_bo_unreserve(root); + } + amdgpu_bo_unref(&root); + dma_fence_put(vm->last_update); for (i = 0; i < AMDGPU_MAX_VMHUBS; i++) amdgpu_vm_free_reserved_vmid(adev, vm, i); } /** + * amdgpu_vm_pasid_fault_credit - Check fault credit for given PASID + * + * @adev: amdgpu_device pointer + * @pasid: PASID do identify the VM + * + * This function is expected to be called in interrupt context. Returns + * true if there was fault credit, false otherwise + */ +bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev, + unsigned int pasid) +{ + struct amdgpu_vm *vm; + + spin_lock(&adev->vm_manager.pasid_lock); + vm = idr_find(&adev->vm_manager.pasid_idr, pasid); + spin_unlock(&adev->vm_manager.pasid_lock); + if (!vm) + /* VM not found, can't track fault credit */ + return true; + + /* No lock needed. only accessed by IRQ handler */ + if (!vm->fault_credit) + /* Too many faults in this VM */ + return false; + + vm->fault_credit--; + return true; +} + +/** * amdgpu_vm_manager_init - init the VM manager * * @adev: amdgpu_device pointer @@ -2589,6 +2893,8 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev) adev->vm_manager.vm_update_mode = 0; #endif + idr_init(&adev->vm_manager.pasid_idr); + spin_lock_init(&adev->vm_manager.pasid_lock); } /** @@ -2602,6 +2908,9 @@ void amdgpu_vm_manager_fini(struct amdgpu_device *adev) { unsigned i, j; + WARN_ON(!idr_is_empty(&adev->vm_manager.pasid_idr)); + idr_destroy(&adev->vm_manager.pasid_idr); + for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) { struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[i]; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 936f158bc5ec..bae77353447b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -25,6 +25,7 @@ #define __AMDGPU_VM_H__ #include <linux/rbtree.h> +#include <linux/idr.h> #include "gpu_scheduler.h" #include "amdgpu_sync.h" @@ -50,9 +51,6 @@ struct amdgpu_bo_list_entry; /* PTBs (Page Table Blocks) need to be aligned to 32K */ #define AMDGPU_VM_PTB_ALIGN_SIZE 32768 -/* LOG2 number of continuous pages for the fragment field */ -#define AMDGPU_LOG2_PAGES_PER_FRAG 4 - #define AMDGPU_PTE_VALID (1ULL << 0) #define AMDGPU_PTE_SYSTEM (1ULL << 1) #define AMDGPU_PTE_SNOOPED (1ULL << 2) @@ -68,10 +66,23 @@ struct amdgpu_bo_list_entry; /* TILED for VEGA10, reserved for older ASICs */ #define AMDGPU_PTE_PRT (1ULL << 51) +/* PDE is handled as PTE for VEGA10 */ +#define AMDGPU_PDE_PTE (1ULL << 54) + /* VEGA10 only */ #define AMDGPU_PTE_MTYPE(a) ((uint64_t)a << 57) #define AMDGPU_PTE_MTYPE_MASK AMDGPU_PTE_MTYPE(3ULL) +/* For Raven */ +#define AMDGPU_MTYPE_CC 2 + +#define AMDGPU_PTE_DEFAULT_ATC (AMDGPU_PTE_SYSTEM \ + | AMDGPU_PTE_SNOOPED \ + | AMDGPU_PTE_EXECUTABLE \ + | AMDGPU_PTE_READABLE \ + | AMDGPU_PTE_WRITEABLE \ + | AMDGPU_PTE_MTYPE(AMDGPU_MTYPE_CC)) + /* How to programm VM fault handling */ #define AMDGPU_VM_FAULT_STOP_NEVER 0 #define AMDGPU_VM_FAULT_STOP_FIRST 1 @@ -83,7 +94,8 @@ struct amdgpu_bo_list_entry; #define AMDGPU_MMHUB 1 /* hardcode that limit for now */ -#define AMDGPU_VA_RESERVED_SIZE (8 << 20) +#define AMDGPU_VA_RESERVED_SIZE (8ULL << 20) + /* max vmids dedicated for process */ #define AMDGPU_VM_MAX_RESERVED_VMID 1 @@ -94,36 +106,57 @@ struct amdgpu_bo_list_entry; #define AMDGPU_VM_USE_CPU_FOR_GFX (1 << 0) #define AMDGPU_VM_USE_CPU_FOR_COMPUTE (1 << 1) +/* base structure for tracking BO usage in a VM */ +struct amdgpu_vm_bo_base { + /* constant after initialization */ + struct amdgpu_vm *vm; + struct amdgpu_bo *bo; + + /* protected by bo being reserved */ + struct list_head bo_list; + + /* protected by spinlock */ + struct list_head vm_status; + + /* protected by the BO being reserved */ + bool moved; +}; struct amdgpu_vm_pt { - struct amdgpu_bo *bo; - uint64_t addr; + struct amdgpu_vm_bo_base base; + uint64_t addr; /* array of page tables, one for each directory entry */ - struct amdgpu_vm_pt *entries; - unsigned last_entry_used; + struct amdgpu_vm_pt *entries; + unsigned last_entry_used; }; +#define AMDGPU_VM_FAULT(pasid, addr) (((u64)(pasid) << 48) | (addr)) +#define AMDGPU_VM_FAULT_PASID(fault) ((u64)(fault) >> 48) +#define AMDGPU_VM_FAULT_ADDR(fault) ((u64)(fault) & 0xfffffffff000ULL) + struct amdgpu_vm { /* tree of virtual addresses mapped */ - struct rb_root va; + struct rb_root_cached va; /* protecting invalidated */ spinlock_t status_lock; - /* BOs moved, but not yet updated in the PT */ - struct list_head invalidated; + /* BOs who needs a validation */ + struct list_head evicted; - /* BOs cleared in the PT because of a move */ - struct list_head cleared; + /* PT BOs which relocated and their parent need an update */ + struct list_head relocated; + + /* BOs moved, but not yet updated in the PT */ + struct list_head moved; /* BO mappings freed, but not yet updated in the PT */ struct list_head freed; /* contains the page directory */ struct amdgpu_vm_pt root; - struct dma_fence *last_dir_update; - uint64_t last_eviction_counter; + struct dma_fence *last_update; /* protecting freed */ spinlock_t freed_lock; @@ -131,15 +164,23 @@ struct amdgpu_vm { /* Scheduler entity for page table updates */ struct amd_sched_entity entity; - /* client id */ + /* client id and PASID (TODO: replace client_id with PASID) */ u64 client_id; + unsigned int pasid; /* dedicated to vm */ struct amdgpu_vm_id *reserved_vmid[AMDGPU_MAX_VMHUBS]; - /* each VM will map on CSA */ - struct amdgpu_bo_va *csa_bo_va; /* Flag to indicate if VM tables are updated by CPU or GPU (SDMA) */ bool use_cpu_for_update; + + /* Flag to indicate ATS support from PTE for GFX9 */ + bool pte_support_ats; + + /* Up to 128 pending retry page faults */ + DECLARE_KFIFO(faults, u64, 128); + + /* Limit non-retry fault storms */ + unsigned int fault_credit; }; struct amdgpu_vm_id { @@ -182,6 +223,7 @@ struct amdgpu_vm_manager { uint32_t num_level; uint64_t vm_size; uint32_t block_size; + uint32_t fragment_size; /* vram base address for page table entry */ u64 vram_base_offset; /* vm pte handling */ @@ -201,28 +243,37 @@ struct amdgpu_vm_manager { * BIT1[= 0] Compute updated by SDMA [= 1] by CPU */ int vm_update_mode; + + /* PASID to VM mapping, will be used in interrupt context to + * look up VM of a page fault + */ + struct idr pasid_idr; + spinlock_t pasid_lock; }; +int amdgpu_vm_alloc_pasid(unsigned int bits); +void amdgpu_vm_free_pasid(unsigned int pasid); void amdgpu_vm_manager_init(struct amdgpu_device *adev); void amdgpu_vm_manager_fini(struct amdgpu_device *adev); int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, - int vm_context); + int vm_context, unsigned int pasid); void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm); +bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev, + unsigned int pasid); void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, struct list_head *validated, struct amdgpu_bo_list_entry *entry); +bool amdgpu_vm_ready(struct amdgpu_vm *vm); int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, int (*callback)(void *p, struct amdgpu_bo *bo), void *param); -void amdgpu_vm_move_pt_bos_in_lru(struct amdgpu_device *adev, - struct amdgpu_vm *vm); int amdgpu_vm_alloc_pts(struct amdgpu_device *adev, struct amdgpu_vm *vm, uint64_t saddr, uint64_t size); int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring, struct amdgpu_sync *sync, struct dma_fence *fence, struct amdgpu_job *job); -int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job); +int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_pipe_sync); void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vmhub, unsigned vmid); void amdgpu_vm_reset_all_ids(struct amdgpu_device *adev); @@ -231,13 +282,13 @@ int amdgpu_vm_update_directories(struct amdgpu_device *adev, int amdgpu_vm_clear_freed(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct dma_fence **fence); -int amdgpu_vm_clear_invalids(struct amdgpu_device *adev, struct amdgpu_vm *vm, - struct amdgpu_sync *sync); +int amdgpu_vm_handle_moved(struct amdgpu_device *adev, + struct amdgpu_vm *vm); int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, bool clear); void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, - struct amdgpu_bo *bo); + struct amdgpu_bo *bo, bool evicted); struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm, struct amdgpu_bo *bo); struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev, @@ -257,9 +308,14 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev, int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev, struct amdgpu_vm *vm, uint64_t saddr, uint64_t size); +struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm, + uint64_t addr); void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va); -void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size); +void amdgpu_vm_set_fragment_size(struct amdgpu_device *adev, + uint32_t fragment_size_default); +void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size, + uint32_t fragment_size_default); int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring, struct amdgpu_job *job); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c index a2c59a08b2bd..4acca92f6a52 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c @@ -28,6 +28,8 @@ struct amdgpu_vram_mgr { struct drm_mm mm; spinlock_t lock; + atomic64_t usage; + atomic64_t vis_usage; }; /** @@ -66,11 +68,6 @@ static int amdgpu_vram_mgr_fini(struct ttm_mem_type_manager *man) struct amdgpu_vram_mgr *mgr = man->priv; spin_lock(&mgr->lock); - if (!drm_mm_clean(&mgr->mm)) { - spin_unlock(&mgr->lock); - return -EBUSY; - } - drm_mm_takedown(&mgr->mm); spin_unlock(&mgr->lock); kfree(mgr); @@ -79,6 +76,27 @@ static int amdgpu_vram_mgr_fini(struct ttm_mem_type_manager *man) } /** + * amdgpu_vram_mgr_vis_size - Calculate visible node size + * + * @adev: amdgpu device structure + * @node: MM node structure + * + * Calculate how many bytes of the MM node are inside visible VRAM + */ +static u64 amdgpu_vram_mgr_vis_size(struct amdgpu_device *adev, + struct drm_mm_node *node) +{ + uint64_t start = node->start << PAGE_SHIFT; + uint64_t end = (node->size + node->start) << PAGE_SHIFT; + + if (start >= adev->mc.visible_vram_size) + return 0; + + return (end > adev->mc.visible_vram_size ? + adev->mc.visible_vram_size : end) - start; +} + +/** * amdgpu_vram_mgr_new - allocate new ranges * * @man: TTM memory type manager @@ -93,11 +111,13 @@ static int amdgpu_vram_mgr_new(struct ttm_mem_type_manager *man, const struct ttm_place *place, struct ttm_mem_reg *mem) { + struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev); struct amdgpu_vram_mgr *mgr = man->priv; struct drm_mm *mm = &mgr->mm; struct drm_mm_node *nodes; enum drm_mm_insert_mode mode; unsigned long lpfn, num_nodes, pages_per_node, pages_left; + uint64_t usage = 0, vis_usage = 0; unsigned i; int r; @@ -142,6 +162,9 @@ static int amdgpu_vram_mgr_new(struct ttm_mem_type_manager *man, if (unlikely(r)) goto error; + usage += nodes[i].size << PAGE_SHIFT; + vis_usage += amdgpu_vram_mgr_vis_size(adev, &nodes[i]); + /* Calculate a virtual BO start address to easily check if * everything is CPU accessible. */ @@ -155,6 +178,9 @@ static int amdgpu_vram_mgr_new(struct ttm_mem_type_manager *man, } spin_unlock(&mgr->lock); + atomic64_add(usage, &mgr->usage); + atomic64_add(vis_usage, &mgr->vis_usage); + mem->mm_node = nodes; return 0; @@ -181,8 +207,10 @@ error: static void amdgpu_vram_mgr_del(struct ttm_mem_type_manager *man, struct ttm_mem_reg *mem) { + struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev); struct amdgpu_vram_mgr *mgr = man->priv; struct drm_mm_node *nodes = mem->mm_node; + uint64_t usage = 0, vis_usage = 0; unsigned pages = mem->num_pages; if (!mem->mm_node) @@ -192,31 +220,67 @@ static void amdgpu_vram_mgr_del(struct ttm_mem_type_manager *man, while (pages) { pages -= nodes->size; drm_mm_remove_node(nodes); + usage += nodes->size << PAGE_SHIFT; + vis_usage += amdgpu_vram_mgr_vis_size(adev, nodes); ++nodes; } spin_unlock(&mgr->lock); + atomic64_sub(usage, &mgr->usage); + atomic64_sub(vis_usage, &mgr->vis_usage); + kfree(mem->mm_node); mem->mm_node = NULL; } /** + * amdgpu_vram_mgr_usage - how many bytes are used in this domain + * + * @man: TTM memory type manager + * + * Returns how many bytes are used in this domain. + */ +uint64_t amdgpu_vram_mgr_usage(struct ttm_mem_type_manager *man) +{ + struct amdgpu_vram_mgr *mgr = man->priv; + + return atomic64_read(&mgr->usage); +} + +/** + * amdgpu_vram_mgr_vis_usage - how many bytes are used in the visible part + * + * @man: TTM memory type manager + * + * Returns how many bytes are used in the visible part of VRAM + */ +uint64_t amdgpu_vram_mgr_vis_usage(struct ttm_mem_type_manager *man) +{ + struct amdgpu_vram_mgr *mgr = man->priv; + + return atomic64_read(&mgr->vis_usage); +} + +/** * amdgpu_vram_mgr_debug - dump VRAM table * * @man: TTM memory type manager - * @prefix: text prefix + * @printer: DRM printer to use * * Dump the table content using printk. */ static void amdgpu_vram_mgr_debug(struct ttm_mem_type_manager *man, - const char *prefix) + struct drm_printer *printer) { struct amdgpu_vram_mgr *mgr = man->priv; - struct drm_printer p = drm_debug_printer(prefix); spin_lock(&mgr->lock); - drm_mm_print(&mgr->mm, &p); + drm_mm_print(&mgr->mm, printer); spin_unlock(&mgr->lock); + + drm_printf(printer, "man size:%llu pages, ram usage:%lluMB, vis usage:%lluMB\n", + man->size, amdgpu_vram_mgr_usage(man) >> 20, + amdgpu_vram_mgr_vis_usage(man) >> 20); } const struct ttm_mem_type_manager_func amdgpu_vram_mgr_func = { diff --git a/drivers/gpu/drm/amd/amdgpu/atom.c b/drivers/gpu/drm/amd/amdgpu/atom.c index d69aa2e179bb..69500a8b4e2d 100644 --- a/drivers/gpu/drm/amd/amdgpu/atom.c +++ b/drivers/gpu/drm/amd/amdgpu/atom.c @@ -1343,8 +1343,11 @@ struct atom_context *amdgpu_atom_parse(struct card_info *card, void *bios) idx = 0x80; str = CSTR(idx); - if (*str != '\0') + if (*str != '\0') { pr_info("ATOM BIOS: %s\n", str); + strlcpy(ctx->vbios_version, str, sizeof(ctx->vbios_version)); + } + return ctx; } diff --git a/drivers/gpu/drm/amd/amdgpu/atom.h b/drivers/gpu/drm/amd/amdgpu/atom.h index ddd8045accf3..a39170991afe 100644 --- a/drivers/gpu/drm/amd/amdgpu/atom.h +++ b/drivers/gpu/drm/amd/amdgpu/atom.h @@ -140,6 +140,7 @@ struct atom_context { int io_mode; uint32_t *scratch; int scratch_size_bytes; + char vbios_version[20]; }; extern int amdgpu_atom_debug; diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c index cb508a211b2f..68b505c768ad 100644 --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c @@ -307,7 +307,6 @@ static int ci_set_power_limit(struct amdgpu_device *adev, u32 n); static int ci_set_overdrive_target_tdp(struct amdgpu_device *adev, u32 target_tdp); static int ci_update_uvd_dpm(struct amdgpu_device *adev, bool gate); -static void ci_dpm_set_dpm_funcs(struct amdgpu_device *adev); static void ci_dpm_set_irq_funcs(struct amdgpu_device *adev); static PPSMC_Result amdgpu_ci_send_msg_to_smc_with_parameter(struct amdgpu_device *adev, @@ -883,8 +882,9 @@ static int ci_power_control_set_level(struct amdgpu_device *adev) return ret; } -static void ci_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate) +static void ci_dpm_powergate_uvd(void *handle, bool gate) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); pi->uvd_power_gated = gate; @@ -901,8 +901,9 @@ static void ci_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate) } } -static bool ci_dpm_vblank_too_short(struct amdgpu_device *adev) +static bool ci_dpm_vblank_too_short(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; u32 vblank_time = amdgpu_dpm_get_vblank_time(adev); u32 switch_limit = adev->mc.vram_type == AMDGPU_VRAM_TYPE_GDDR5 ? 450 : 300; @@ -1210,11 +1211,12 @@ static int ci_fan_ctrl_stop_smc_fan_control(struct amdgpu_device *adev) } } -static int ci_dpm_get_fan_speed_percent(struct amdgpu_device *adev, +static int ci_dpm_get_fan_speed_percent(void *handle, u32 *speed) { u32 duty, duty100; u64 tmp64; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (adev->pm.no_fan) return -ENOENT; @@ -1237,12 +1239,13 @@ static int ci_dpm_get_fan_speed_percent(struct amdgpu_device *adev, return 0; } -static int ci_dpm_set_fan_speed_percent(struct amdgpu_device *adev, +static int ci_dpm_set_fan_speed_percent(void *handle, u32 speed) { u32 tmp; u32 duty, duty100; u64 tmp64; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); if (adev->pm.no_fan) @@ -1271,8 +1274,10 @@ static int ci_dpm_set_fan_speed_percent(struct amdgpu_device *adev, return 0; } -static void ci_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode) +static void ci_dpm_set_fan_control_mode(void *handle, u32 mode) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + switch (mode) { case AMD_FAN_CTRL_NONE: if (adev->pm.dpm.fan.ucode_fan_control) @@ -1292,8 +1297,9 @@ static void ci_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode) } } -static u32 ci_dpm_get_fan_control_mode(struct amdgpu_device *adev) +static u32 ci_dpm_get_fan_control_mode(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); if (pi->fan_is_controlled_by_smc) @@ -4378,9 +4384,10 @@ static u32 ci_get_lowest_enabled_level(struct amdgpu_device *adev, } -static int ci_dpm_force_performance_level(struct amdgpu_device *adev, +static int ci_dpm_force_performance_level(void *handle, enum amd_dpm_forced_level level) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); u32 tmp, levels, i; int ret; @@ -5291,8 +5298,9 @@ static void ci_update_requested_ps(struct amdgpu_device *adev, adev->pm.dpm.requested_ps = &pi->requested_rps; } -static int ci_dpm_pre_set_power_state(struct amdgpu_device *adev) +static int ci_dpm_pre_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps; struct amdgpu_ps *new_ps = &requested_ps; @@ -5304,8 +5312,9 @@ static int ci_dpm_pre_set_power_state(struct amdgpu_device *adev) return 0; } -static void ci_dpm_post_set_power_state(struct amdgpu_device *adev) +static void ci_dpm_post_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct amdgpu_ps *new_ps = &pi->requested_rps; @@ -5479,8 +5488,9 @@ static void ci_dpm_disable(struct amdgpu_device *adev) ci_update_current_ps(adev, boot_ps); } -static int ci_dpm_set_power_state(struct amdgpu_device *adev) +static int ci_dpm_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct amdgpu_ps *new_ps = &pi->requested_rps; struct amdgpu_ps *old_ps = &pi->current_rps; @@ -5551,8 +5561,10 @@ static void ci_dpm_reset_asic(struct amdgpu_device *adev) } #endif -static void ci_dpm_display_configuration_changed(struct amdgpu_device *adev) +static void ci_dpm_display_configuration_changed(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + ci_program_display_gap(adev); } @@ -6105,9 +6117,10 @@ static int ci_dpm_init(struct amdgpu_device *adev) } static void -ci_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev, +ci_dpm_debugfs_print_current_performance_level(void *handle, struct seq_file *m) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct amdgpu_ps *rps = &pi->current_rps; u32 sclk = ci_get_average_sclk_freq(adev); @@ -6131,12 +6144,13 @@ ci_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev, seq_printf(m, "GPU load: %u %%\n", activity_percent); } -static void ci_dpm_print_power_state(struct amdgpu_device *adev, - struct amdgpu_ps *rps) +static void ci_dpm_print_power_state(void *handle, void *current_ps) { + struct amdgpu_ps *rps = (struct amdgpu_ps *)current_ps; struct ci_ps *ps = ci_get_ps(rps); struct ci_pl *pl; int i; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_dpm_print_class_info(rps->class, rps->class2); amdgpu_dpm_print_cap_info(rps->caps); @@ -6158,20 +6172,23 @@ static inline bool ci_are_power_levels_equal(const struct ci_pl *ci_cpl1, (ci_cpl1->pcie_lane == ci_cpl2->pcie_lane)); } -static int ci_check_state_equal(struct amdgpu_device *adev, - struct amdgpu_ps *cps, - struct amdgpu_ps *rps, +static int ci_check_state_equal(void *handle, + void *current_ps, + void *request_ps, bool *equal) { struct ci_ps *ci_cps; struct ci_ps *ci_rps; int i; + struct amdgpu_ps *cps = (struct amdgpu_ps *)current_ps; + struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (adev == NULL || cps == NULL || rps == NULL || equal == NULL) return -EINVAL; - ci_cps = ci_get_ps(cps); - ci_rps = ci_get_ps(rps); + ci_cps = ci_get_ps((struct amdgpu_ps *)cps); + ci_rps = ci_get_ps((struct amdgpu_ps *)rps); if (ci_cps == NULL) { *equal = false; @@ -6199,8 +6216,9 @@ static int ci_check_state_equal(struct amdgpu_device *adev, return 0; } -static u32 ci_dpm_get_sclk(struct amdgpu_device *adev, bool low) +static u32 ci_dpm_get_sclk(void *handle, bool low) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_ps *requested_state = ci_get_ps(&pi->requested_rps); @@ -6210,8 +6228,9 @@ static u32 ci_dpm_get_sclk(struct amdgpu_device *adev, bool low) return requested_state->performance_levels[requested_state->performance_level_count - 1].sclk; } -static u32 ci_dpm_get_mclk(struct amdgpu_device *adev, bool low) +static u32 ci_dpm_get_mclk(void *handle, bool low) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_ps *requested_state = ci_get_ps(&pi->requested_rps); @@ -6222,10 +6241,11 @@ static u32 ci_dpm_get_mclk(struct amdgpu_device *adev, bool low) } /* get temperature in millidegrees */ -static int ci_dpm_get_temp(struct amdgpu_device *adev) +static int ci_dpm_get_temp(void *handle) { u32 temp; int actual_temp = 0; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; temp = (RREG32_SMC(ixCG_MULT_THERMAL_STATUS) & CG_MULT_THERMAL_STATUS__CTF_TEMP_MASK) >> CG_MULT_THERMAL_STATUS__CTF_TEMP__SHIFT; @@ -6261,7 +6281,6 @@ static int ci_dpm_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - ci_dpm_set_dpm_funcs(adev); ci_dpm_set_irq_funcs(adev); return 0; @@ -6346,7 +6365,6 @@ static int ci_dpm_sw_fini(void *handle) flush_work(&adev->pm.dpm.thermal.work); mutex_lock(&adev->pm.mutex); - amdgpu_pm_sysfs_fini(adev); ci_dpm_fini(adev); mutex_unlock(&adev->pm.mutex); @@ -6551,9 +6569,10 @@ static int ci_dpm_set_powergating_state(void *handle, return 0; } -static int ci_dpm_print_clock_levels(struct amdgpu_device *adev, +static int ci_dpm_print_clock_levels(void *handle, enum pp_clock_type type, char *buf) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_single_dpm_table *sclk_table = &pi->dpm_table.sclk_table; struct ci_single_dpm_table *mclk_table = &pi->dpm_table.mclk_table; @@ -6618,9 +6637,10 @@ static int ci_dpm_print_clock_levels(struct amdgpu_device *adev, return size; } -static int ci_dpm_force_clock_level(struct amdgpu_device *adev, +static int ci_dpm_force_clock_level(void *handle, enum pp_clock_type type, uint32_t mask) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); if (adev->pm.dpm.forced_level & (AMD_DPM_FORCED_LEVEL_AUTO | @@ -6664,8 +6684,9 @@ static int ci_dpm_force_clock_level(struct amdgpu_device *adev, return 0; } -static int ci_dpm_get_sclk_od(struct amdgpu_device *adev) +static int ci_dpm_get_sclk_od(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_single_dpm_table *sclk_table = &(pi->dpm_table.sclk_table); struct ci_single_dpm_table *golden_sclk_table = @@ -6680,8 +6701,9 @@ static int ci_dpm_get_sclk_od(struct amdgpu_device *adev) return value; } -static int ci_dpm_set_sclk_od(struct amdgpu_device *adev, uint32_t value) +static int ci_dpm_set_sclk_od(void *handle, uint32_t value) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_ps *ps = ci_get_ps(adev->pm.dpm.requested_ps); struct ci_single_dpm_table *golden_sclk_table = @@ -6698,8 +6720,9 @@ static int ci_dpm_set_sclk_od(struct amdgpu_device *adev, uint32_t value) return 0; } -static int ci_dpm_get_mclk_od(struct amdgpu_device *adev) +static int ci_dpm_get_mclk_od(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_single_dpm_table *mclk_table = &(pi->dpm_table.mclk_table); struct ci_single_dpm_table *golden_mclk_table = @@ -6714,8 +6737,9 @@ static int ci_dpm_get_mclk_od(struct amdgpu_device *adev) return value; } -static int ci_dpm_set_mclk_od(struct amdgpu_device *adev, uint32_t value) +static int ci_dpm_set_mclk_od(void *handle, uint32_t value) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_ps *ps = ci_get_ps(adev->pm.dpm.requested_ps); struct ci_single_dpm_table *golden_mclk_table = @@ -6732,9 +6756,10 @@ static int ci_dpm_set_mclk_od(struct amdgpu_device *adev, uint32_t value) return 0; } -static int ci_dpm_get_power_profile_state(struct amdgpu_device *adev, +static int ci_dpm_get_power_profile_state(void *handle, struct amd_pp_profile *query) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); if (!pi || !query) @@ -6851,9 +6876,10 @@ static int ci_set_power_profile_state(struct amdgpu_device *adev, return result; } -static int ci_dpm_set_power_profile_state(struct amdgpu_device *adev, +static int ci_dpm_set_power_profile_state(void *handle, struct amd_pp_profile *request) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); int ret = -1; @@ -6906,9 +6932,10 @@ static int ci_dpm_set_power_profile_state(struct amdgpu_device *adev, return 0; } -static int ci_dpm_reset_power_profile_state(struct amdgpu_device *adev, +static int ci_dpm_reset_power_profile_state(void *handle, struct amd_pp_profile *request) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); if (!pi || !request) @@ -6927,9 +6954,10 @@ static int ci_dpm_reset_power_profile_state(struct amdgpu_device *adev, return -EINVAL; } -static int ci_dpm_switch_power_profile(struct amdgpu_device *adev, +static int ci_dpm_switch_power_profile(void *handle, enum amd_pp_profile_type type) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct amd_pp_profile request = {0}; @@ -6944,11 +6972,12 @@ static int ci_dpm_switch_power_profile(struct amdgpu_device *adev, return 0; } -static int ci_dpm_read_sensor(struct amdgpu_device *adev, int idx, +static int ci_dpm_read_sensor(void *handle, int idx, void *value, int *size) { u32 activity_percent = 50; int ret; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* size must be at least 4 bytes for all sensors */ if (*size < 4) @@ -7003,7 +7032,7 @@ const struct amd_ip_funcs ci_dpm_ip_funcs = { .set_powergating_state = ci_dpm_set_powergating_state, }; -static const struct amdgpu_dpm_funcs ci_dpm_funcs = { +const struct amd_pm_funcs ci_dpm_funcs = { .get_temperature = &ci_dpm_get_temp, .pre_set_power_state = &ci_dpm_pre_set_power_state, .set_power_state = &ci_dpm_set_power_state, @@ -7035,12 +7064,6 @@ static const struct amdgpu_dpm_funcs ci_dpm_funcs = { .read_sensor = ci_dpm_read_sensor, }; -static void ci_dpm_set_dpm_funcs(struct amdgpu_device *adev) -{ - if (adev->pm.funcs == NULL) - adev->pm.funcs = &ci_dpm_funcs; -} - static const struct amdgpu_irq_src_funcs ci_dpm_irq_funcs = { .set = ci_dpm_set_interrupt_state, .process = ci_dpm_process_interrupt, diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c index 37a499ab30eb..793b1470284d 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik.c +++ b/drivers/gpu/drm/amd/amdgpu/cik.c @@ -65,6 +65,7 @@ #include "oss/oss_2_0_d.h" #include "oss/oss_2_0_sh_mask.h" +#include "amdgpu_dm.h" #include "amdgpu_amdkfd.h" #include "amdgpu_powerplay.h" #include "dce_virtual.h" @@ -1824,21 +1825,14 @@ static int cik_common_suspend(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - amdgpu_amdkfd_suspend(adev); - return cik_common_hw_fini(adev); } static int cik_common_resume(void *handle) { - int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - r = cik_common_hw_init(adev); - if (r) - return r; - - return amdgpu_amdkfd_resume(adev); + return cik_common_hw_init(adev); } static bool cik_common_is_idle(void *handle) @@ -1907,6 +1901,10 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) amdgpu_ip_block_add(adev, &amdgpu_pp_ip_block); if (adev->enable_virtual_display) amdgpu_ip_block_add(adev, &dce_virtual_ip_block); +#if defined(CONFIG_DRM_AMD_DC) + else if (amdgpu_device_has_dc_support(adev)) + amdgpu_ip_block_add(adev, &dm_ip_block); +#endif else amdgpu_ip_block_add(adev, &dce_v8_2_ip_block); amdgpu_ip_block_add(adev, &gfx_v7_2_ip_block); @@ -1921,6 +1919,10 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) amdgpu_ip_block_add(adev, &amdgpu_pp_ip_block); if (adev->enable_virtual_display) amdgpu_ip_block_add(adev, &dce_virtual_ip_block); +#if defined(CONFIG_DRM_AMD_DC) + else if (amdgpu_device_has_dc_support(adev)) + amdgpu_ip_block_add(adev, &dm_ip_block); +#endif else amdgpu_ip_block_add(adev, &dce_v8_5_ip_block); amdgpu_ip_block_add(adev, &gfx_v7_3_ip_block); @@ -1935,6 +1937,10 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) amdgpu_ip_block_add(adev, &amdgpu_pp_ip_block); if (adev->enable_virtual_display) amdgpu_ip_block_add(adev, &dce_virtual_ip_block); +#if defined(CONFIG_DRM_AMD_DC) + else if (amdgpu_device_has_dc_support(adev)) + amdgpu_ip_block_add(adev, &dm_ip_block); +#endif else amdgpu_ip_block_add(adev, &dce_v8_1_ip_block); amdgpu_ip_block_add(adev, &gfx_v7_1_ip_block); @@ -1950,6 +1956,10 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) amdgpu_ip_block_add(adev, &amdgpu_pp_ip_block); if (adev->enable_virtual_display) amdgpu_ip_block_add(adev, &dce_virtual_ip_block); +#if defined(CONFIG_DRM_AMD_DC) + else if (amdgpu_device_has_dc_support(adev)) + amdgpu_ip_block_add(adev, &dm_ip_block); +#endif else amdgpu_ip_block_add(adev, &dce_v8_3_ip_block); amdgpu_ip_block_add(adev, &gfx_v7_2_ip_block); diff --git a/drivers/gpu/drm/amd/amdgpu/cik_dpm.h b/drivers/gpu/drm/amd/amdgpu/cik_dpm.h index b1c8e7b446ea..c7b4349f6319 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/cik_dpm.h @@ -26,5 +26,6 @@ extern const struct amd_ip_funcs ci_dpm_ip_funcs; extern const struct amd_ip_funcs kv_dpm_ip_funcs; - +extern const struct amd_pm_funcs ci_dpm_funcs; +extern const struct amd_pm_funcs kv_dpm_funcs; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/cik_ih.c b/drivers/gpu/drm/amd/amdgpu/cik_ih.c index b8918432c572..a870b354e3f7 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_ih.c @@ -228,6 +228,34 @@ static u32 cik_ih_get_wptr(struct amdgpu_device *adev) * [127:96] - reserved */ +/** + * cik_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool cik_ih_prescreen_iv(struct amdgpu_device *adev) +{ + u32 ring_index = adev->irq.ih.rptr >> 2; + u16 pasid; + + switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { + case 146: + case 147: + pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; + if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) + return true; + break; + default: + /* Not a VM fault */ + return true; + } + + adev->irq.ih.rptr += 16; + return false; +} + /** * cik_ih_decode_iv - decode an interrupt vector * @@ -433,6 +461,7 @@ static const struct amd_ip_funcs cik_ih_ip_funcs = { static const struct amdgpu_ih_funcs cik_ih_funcs = { .get_wptr = cik_ih_get_wptr, + .prescreen_iv = cik_ih_prescreen_iv, .decode_iv = cik_ih_decode_iv, .set_rptr = cik_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index c216e16826c9..60cecd117705 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -342,6 +342,63 @@ static void cik_sdma_rlc_stop(struct amdgpu_device *adev) } /** + * cik_ctx_switch_enable - stop the async dma engines context switch + * + * @adev: amdgpu_device pointer + * @enable: enable/disable the DMA MEs context switch. + * + * Halt or unhalt the async dma engines context switch (VI). + */ +static void cik_ctx_switch_enable(struct amdgpu_device *adev, bool enable) +{ + u32 f32_cntl, phase_quantum = 0; + int i; + + if (amdgpu_sdma_phase_quantum) { + unsigned value = amdgpu_sdma_phase_quantum; + unsigned unit = 0; + + while (value > (SDMA0_PHASE0_QUANTUM__VALUE_MASK >> + SDMA0_PHASE0_QUANTUM__VALUE__SHIFT)) { + value = (value + 1) >> 1; + unit++; + } + if (unit > (SDMA0_PHASE0_QUANTUM__UNIT_MASK >> + SDMA0_PHASE0_QUANTUM__UNIT__SHIFT)) { + value = (SDMA0_PHASE0_QUANTUM__VALUE_MASK >> + SDMA0_PHASE0_QUANTUM__VALUE__SHIFT); + unit = (SDMA0_PHASE0_QUANTUM__UNIT_MASK >> + SDMA0_PHASE0_QUANTUM__UNIT__SHIFT); + WARN_ONCE(1, + "clamping sdma_phase_quantum to %uK clock cycles\n", + value << unit); + } + phase_quantum = + value << SDMA0_PHASE0_QUANTUM__VALUE__SHIFT | + unit << SDMA0_PHASE0_QUANTUM__UNIT__SHIFT; + } + + for (i = 0; i < adev->sdma.num_instances; i++) { + f32_cntl = RREG32(mmSDMA0_CNTL + sdma_offsets[i]); + if (enable) { + f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL, + AUTO_CTXSW_ENABLE, 1); + if (amdgpu_sdma_phase_quantum) { + WREG32(mmSDMA0_PHASE0_QUANTUM + sdma_offsets[i], + phase_quantum); + WREG32(mmSDMA0_PHASE1_QUANTUM + sdma_offsets[i], + phase_quantum); + } + } else { + f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL, + AUTO_CTXSW_ENABLE, 0); + } + + WREG32(mmSDMA0_CNTL + sdma_offsets[i], f32_cntl); + } +} + +/** * cik_sdma_enable - stop the async dma engines * * @adev: amdgpu_device pointer @@ -537,6 +594,8 @@ static int cik_sdma_start(struct amdgpu_device *adev) /* halt the engine before programing */ cik_sdma_enable(adev, false); + /* enable sdma ring preemption */ + cik_ctx_switch_enable(adev, true); /* start the gfx rings and rlc compute queues */ r = cik_sdma_gfx_resume(adev); @@ -984,6 +1043,7 @@ static int cik_sdma_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + cik_ctx_switch_enable(adev, false); cik_sdma_enable(adev, false); return 0; @@ -1327,8 +1387,13 @@ static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev) } static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = { + .copy_pte_num_dw = 7, .copy_pte = cik_sdma_vm_copy_pte, + .write_pte = cik_sdma_vm_write_pte, + + .set_max_nums_pte_pde = 0x1fffff >> 3, + .set_pte_pde_num_dw = 10, .set_pte_pde = cik_sdma_vm_set_pte_pde, }; diff --git a/drivers/gpu/drm/amd/amdgpu/clearstate_gfx9.h b/drivers/gpu/drm/amd/amdgpu/clearstate_gfx9.h index 18fd01f3e4b2..003a131bad47 100644 --- a/drivers/gpu/drm/amd/amdgpu/clearstate_gfx9.h +++ b/drivers/gpu/drm/amd/amdgpu/clearstate_gfx9.h @@ -1,24 +1,25 @@ - /* -*************************************************************************************************** -* -* Trade secret of Advanced Micro Devices, Inc. -* Copyright (c) 2010 Advanced Micro Devices, Inc. (unpublished) -* -* All rights reserved. This notice is intended as a precaution against inadvertent publication and -* does not imply publication or any waiver of confidentiality. The year included in the foregoing -* notice is the year of creation of the work. -* -*************************************************************************************************** -*/ -/** -*************************************************************************************************** -* @brief gfx9 Clearstate Definitions -*************************************************************************************************** -* -* Do not edit! This is a machine-generated file! -* -*/ + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ static const unsigned int gfx9_SECT_CONTEXT_def_1[] = { diff --git a/drivers/gpu/drm/amd/amdgpu/cz_ih.c b/drivers/gpu/drm/amd/amdgpu/cz_ih.c index 0c1209cdd1cb..fa61d649bb44 100644 --- a/drivers/gpu/drm/amd/amdgpu/cz_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cz_ih.c @@ -208,6 +208,34 @@ static u32 cz_ih_get_wptr(struct amdgpu_device *adev) } /** + * cz_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool cz_ih_prescreen_iv(struct amdgpu_device *adev) +{ + u32 ring_index = adev->irq.ih.rptr >> 2; + u16 pasid; + + switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { + case 146: + case 147: + pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; + if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) + return true; + break; + default: + /* Not a VM fault */ + return true; + } + + adev->irq.ih.rptr += 16; + return false; +} + +/** * cz_ih_decode_iv - decode an interrupt vector * * @adev: amdgpu_device pointer @@ -414,6 +442,7 @@ static const struct amd_ip_funcs cz_ih_ip_funcs = { static const struct amdgpu_ih_funcs cz_ih_funcs = { .get_wptr = cz_ih_get_wptr, + .prescreen_iv = cz_ih_prescreen_iv, .decode_iv = cz_ih_decode_iv, .set_rptr = cz_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c index 9f78c03a2e31..4e519dc42916 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c @@ -484,134 +484,6 @@ static bool dce_v10_0_is_display_hung(struct amdgpu_device *adev) return true; } -static void dce_v10_0_stop_mc_access(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) -{ - u32 crtc_enabled, tmp; - int i; - - save->vga_render_control = RREG32(mmVGA_RENDER_CONTROL); - save->vga_hdp_control = RREG32(mmVGA_HDP_CONTROL); - - /* disable VGA render */ - tmp = RREG32(mmVGA_RENDER_CONTROL); - tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0); - WREG32(mmVGA_RENDER_CONTROL, tmp); - - /* blank the display controllers */ - for (i = 0; i < adev->mode_info.num_crtc; i++) { - crtc_enabled = REG_GET_FIELD(RREG32(mmCRTC_CONTROL + crtc_offsets[i]), - CRTC_CONTROL, CRTC_MASTER_EN); - if (crtc_enabled) { -#if 0 - u32 frame_count; - int j; - - save->crtc_enabled[i] = true; - tmp = RREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i]); - if (REG_GET_FIELD(tmp, CRTC_BLANK_CONTROL, CRTC_BLANK_DATA_EN) == 0) { - amdgpu_display_vblank_wait(adev, i); - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1); - tmp = REG_SET_FIELD(tmp, CRTC_BLANK_CONTROL, CRTC_BLANK_DATA_EN, 1); - WREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i], tmp); - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0); - } - /* wait for the next frame */ - frame_count = amdgpu_display_vblank_get_counter(adev, i); - for (j = 0; j < adev->usec_timeout; j++) { - if (amdgpu_display_vblank_get_counter(adev, i) != frame_count) - break; - udelay(1); - } - tmp = RREG32(mmGRPH_UPDATE + crtc_offsets[i]); - if (REG_GET_FIELD(tmp, GRPH_UPDATE, GRPH_UPDATE_LOCK) == 0) { - tmp = REG_SET_FIELD(tmp, GRPH_UPDATE, GRPH_UPDATE_LOCK, 1); - WREG32(mmGRPH_UPDATE + crtc_offsets[i], tmp); - } - tmp = RREG32(mmMASTER_UPDATE_LOCK + crtc_offsets[i]); - if (REG_GET_FIELD(tmp, MASTER_UPDATE_LOCK, MASTER_UPDATE_LOCK) == 0) { - tmp = REG_SET_FIELD(tmp, MASTER_UPDATE_LOCK, MASTER_UPDATE_LOCK, 1); - WREG32(mmMASTER_UPDATE_LOCK + crtc_offsets[i], tmp); - } -#else - /* XXX this is a hack to avoid strange behavior with EFI on certain systems */ - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1); - tmp = RREG32(mmCRTC_CONTROL + crtc_offsets[i]); - tmp = REG_SET_FIELD(tmp, CRTC_CONTROL, CRTC_MASTER_EN, 0); - WREG32(mmCRTC_CONTROL + crtc_offsets[i], tmp); - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0); - save->crtc_enabled[i] = false; - /* ***** */ -#endif - } else { - save->crtc_enabled[i] = false; - } - } -} - -static void dce_v10_0_resume_mc_access(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) -{ - u32 tmp, frame_count; - int i, j; - - /* update crtc base addresses */ - for (i = 0; i < adev->mode_info.num_crtc; i++) { - WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + crtc_offsets[i], - upper_32_bits(adev->mc.vram_start)); - WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS_HIGH + crtc_offsets[i], - upper_32_bits(adev->mc.vram_start)); - WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + crtc_offsets[i], - (u32)adev->mc.vram_start); - WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS + crtc_offsets[i], - (u32)adev->mc.vram_start); - - if (save->crtc_enabled[i]) { - tmp = RREG32(mmMASTER_UPDATE_MODE + crtc_offsets[i]); - if (REG_GET_FIELD(tmp, MASTER_UPDATE_MODE, MASTER_UPDATE_MODE) != 0) { - tmp = REG_SET_FIELD(tmp, MASTER_UPDATE_MODE, MASTER_UPDATE_MODE, 0); - WREG32(mmMASTER_UPDATE_MODE + crtc_offsets[i], tmp); - } - tmp = RREG32(mmGRPH_UPDATE + crtc_offsets[i]); - if (REG_GET_FIELD(tmp, GRPH_UPDATE, GRPH_UPDATE_LOCK)) { - tmp = REG_SET_FIELD(tmp, GRPH_UPDATE, GRPH_UPDATE_LOCK, 0); - WREG32(mmGRPH_UPDATE + crtc_offsets[i], tmp); - } - tmp = RREG32(mmMASTER_UPDATE_LOCK + crtc_offsets[i]); - if (REG_GET_FIELD(tmp, MASTER_UPDATE_LOCK, MASTER_UPDATE_LOCK)) { - tmp = REG_SET_FIELD(tmp, MASTER_UPDATE_LOCK, MASTER_UPDATE_LOCK, 0); - WREG32(mmMASTER_UPDATE_LOCK + crtc_offsets[i], tmp); - } - for (j = 0; j < adev->usec_timeout; j++) { - tmp = RREG32(mmGRPH_UPDATE + crtc_offsets[i]); - if (REG_GET_FIELD(tmp, GRPH_UPDATE, GRPH_SURFACE_UPDATE_PENDING) == 0) - break; - udelay(1); - } - tmp = RREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i]); - tmp = REG_SET_FIELD(tmp, CRTC_BLANK_CONTROL, CRTC_BLANK_DATA_EN, 0); - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1); - WREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i], tmp); - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0); - /* wait for the next frame */ - frame_count = amdgpu_display_vblank_get_counter(adev, i); - for (j = 0; j < adev->usec_timeout; j++) { - if (amdgpu_display_vblank_get_counter(adev, i) != frame_count) - break; - udelay(1); - } - } - } - - WREG32(mmVGA_MEMORY_BASE_ADDRESS_HIGH, upper_32_bits(adev->mc.vram_start)); - WREG32(mmVGA_MEMORY_BASE_ADDRESS, lower_32_bits(adev->mc.vram_start)); - - /* Unlock vga access */ - WREG32(mmVGA_HDP_CONTROL, save->vga_hdp_control); - mdelay(1); - WREG32(mmVGA_RENDER_CONTROL, save->vga_render_control); -} - static void dce_v10_0_set_vga_render_state(struct amdgpu_device *adev, bool render) { @@ -1867,7 +1739,7 @@ static void dce_v10_0_afmt_setmode(struct drm_encoder *encoder, dce_v10_0_audio_write_sad_regs(encoder); dce_v10_0_audio_write_latency_fields(encoder, mode); - err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode); + err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode, false); if (err < 0) { DRM_ERROR("failed to setup AVI infoframe: %zd\n", err); return; @@ -2267,6 +2139,7 @@ static void dce_v10_0_crtc_load_lut(struct drm_crtc *crtc) struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); struct drm_device *dev = crtc->dev; struct amdgpu_device *adev = dev->dev_private; + u16 *r, *g, *b; int i; u32 tmp; @@ -2304,11 +2177,14 @@ static void dce_v10_0_crtc_load_lut(struct drm_crtc *crtc) WREG32(mmDC_LUT_WRITE_EN_MASK + amdgpu_crtc->crtc_offset, 0x00000007); WREG32(mmDC_LUT_RW_INDEX + amdgpu_crtc->crtc_offset, 0); + r = crtc->gamma_store; + g = r + crtc->gamma_size; + b = g + crtc->gamma_size; for (i = 0; i < 256; i++) { WREG32(mmDC_LUT_30_COLOR + amdgpu_crtc->crtc_offset, - (amdgpu_crtc->lut_r[i] << 20) | - (amdgpu_crtc->lut_g[i] << 10) | - (amdgpu_crtc->lut_b[i] << 0)); + ((*r++ & 0xffc0) << 14) | + ((*g++ & 0xffc0) << 4) | + (*b++ >> 6)); } tmp = RREG32(mmDEGAMMA_CONTROL + amdgpu_crtc->crtc_offset); @@ -2555,7 +2431,7 @@ static int dce_v10_0_crtc_cursor_set2(struct drm_crtc *crtc, aobj = gem_to_amdgpu_bo(obj); ret = amdgpu_bo_reserve(aobj, false); if (ret != 0) { - drm_gem_object_unreference_unlocked(obj); + drm_gem_object_put_unlocked(obj); return ret; } @@ -2563,7 +2439,7 @@ static int dce_v10_0_crtc_cursor_set2(struct drm_crtc *crtc, amdgpu_bo_unreserve(aobj); if (ret) { DRM_ERROR("Failed to pin new cursor BO (%d)\n", ret); - drm_gem_object_unreference_unlocked(obj); + drm_gem_object_put_unlocked(obj); return ret; } @@ -2597,7 +2473,7 @@ unpin: amdgpu_bo_unpin(aobj); amdgpu_bo_unreserve(aobj); } - drm_gem_object_unreference_unlocked(amdgpu_crtc->cursor_bo); + drm_gem_object_put_unlocked(amdgpu_crtc->cursor_bo); } amdgpu_crtc->cursor_bo = obj; @@ -2624,15 +2500,6 @@ static int dce_v10_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, u16 *blue, uint32_t size, struct drm_modeset_acquire_ctx *ctx) { - struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - int i; - - /* userspace palettes are always correct as is */ - for (i = 0; i < size; i++) { - amdgpu_crtc->lut_r[i] = red[i] >> 6; - amdgpu_crtc->lut_g[i] = green[i] >> 6; - amdgpu_crtc->lut_b[i] = blue[i] >> 6; - } dce_v10_0_crtc_load_lut(crtc); return 0; @@ -2844,14 +2711,12 @@ static const struct drm_crtc_helper_funcs dce_v10_0_crtc_helper_funcs = { .mode_set_base_atomic = dce_v10_0_crtc_set_base_atomic, .prepare = dce_v10_0_crtc_prepare, .commit = dce_v10_0_crtc_commit, - .load_lut = dce_v10_0_crtc_load_lut, .disable = dce_v10_0_crtc_disable, }; static int dce_v10_0_crtc_init(struct amdgpu_device *adev, int index) { struct amdgpu_crtc *amdgpu_crtc; - int i; amdgpu_crtc = kzalloc(sizeof(struct amdgpu_crtc) + (AMDGPUFB_CONN_LIMIT * sizeof(struct drm_connector *)), GFP_KERNEL); @@ -2869,12 +2734,6 @@ static int dce_v10_0_crtc_init(struct amdgpu_device *adev, int index) adev->ddev->mode_config.cursor_width = amdgpu_crtc->max_cursor_width; adev->ddev->mode_config.cursor_height = amdgpu_crtc->max_cursor_height; - for (i = 0; i < 256; i++) { - amdgpu_crtc->lut_r[i] = i << 2; - amdgpu_crtc->lut_g[i] = i << 2; - amdgpu_crtc->lut_b[i] = i << 2; - } - switch (amdgpu_crtc->crtc_id) { case 0: default: @@ -3025,6 +2884,8 @@ static int dce_v10_0_hw_init(void *handle) dce_v10_0_init_golden_registers(adev); + /* disable vga render */ + dce_v10_0_set_vga_render_state(adev, false); /* init dig PHYs, disp eng pll */ amdgpu_atombios_encoder_init_dig(adev); amdgpu_atombios_crtc_set_disp_eng_pll(adev, adev->clock.default_dispclk); @@ -3737,7 +3598,6 @@ static void dce_v10_0_encoder_add(struct amdgpu_device *adev, } static const struct amdgpu_display_funcs dce_v10_0_display_funcs = { - .set_vga_render_state = &dce_v10_0_set_vga_render_state, .bandwidth_update = &dce_v10_0_bandwidth_update, .vblank_get_counter = &dce_v10_0_vblank_get_counter, .vblank_wait = &dce_v10_0_vblank_wait, @@ -3750,8 +3610,6 @@ static const struct amdgpu_display_funcs dce_v10_0_display_funcs = { .page_flip_get_scanoutpos = &dce_v10_0_crtc_get_scanoutpos, .add_encoder = &dce_v10_0_encoder_add, .add_connector = &amdgpu_connector_add, - .stop_mc_access = &dce_v10_0_stop_mc_access, - .resume_mc_access = &dce_v10_0_resume_mc_access, }; static void dce_v10_0_set_display_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c index 4bcf01dc567a..11edc75edaa9 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c @@ -499,79 +499,6 @@ static bool dce_v11_0_is_display_hung(struct amdgpu_device *adev) return true; } -static void dce_v11_0_stop_mc_access(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) -{ - u32 crtc_enabled, tmp; - int i; - - save->vga_render_control = RREG32(mmVGA_RENDER_CONTROL); - save->vga_hdp_control = RREG32(mmVGA_HDP_CONTROL); - - /* disable VGA render */ - tmp = RREG32(mmVGA_RENDER_CONTROL); - tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0); - WREG32(mmVGA_RENDER_CONTROL, tmp); - - /* blank the display controllers */ - for (i = 0; i < adev->mode_info.num_crtc; i++) { - crtc_enabled = REG_GET_FIELD(RREG32(mmCRTC_CONTROL + crtc_offsets[i]), - CRTC_CONTROL, CRTC_MASTER_EN); - if (crtc_enabled) { -#if 1 - save->crtc_enabled[i] = true; - tmp = RREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i]); - if (REG_GET_FIELD(tmp, CRTC_BLANK_CONTROL, CRTC_BLANK_DATA_EN) == 0) { - /*it is correct only for RGB ; black is 0*/ - WREG32(mmCRTC_BLANK_DATA_COLOR + crtc_offsets[i], 0); - tmp = REG_SET_FIELD(tmp, CRTC_BLANK_CONTROL, CRTC_BLANK_DATA_EN, 1); - WREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i], tmp); - } -#else - /* XXX this is a hack to avoid strange behavior with EFI on certain systems */ - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1); - tmp = RREG32(mmCRTC_CONTROL + crtc_offsets[i]); - tmp = REG_SET_FIELD(tmp, CRTC_CONTROL, CRTC_MASTER_EN, 0); - WREG32(mmCRTC_CONTROL + crtc_offsets[i], tmp); - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0); - save->crtc_enabled[i] = false; - /* ***** */ -#endif - } else { - save->crtc_enabled[i] = false; - } - } -} - -static void dce_v11_0_resume_mc_access(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) -{ - u32 tmp; - int i; - - /* update crtc base addresses */ - for (i = 0; i < adev->mode_info.num_crtc; i++) { - WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + crtc_offsets[i], - upper_32_bits(adev->mc.vram_start)); - WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + crtc_offsets[i], - (u32)adev->mc.vram_start); - - if (save->crtc_enabled[i]) { - tmp = RREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i]); - tmp = REG_SET_FIELD(tmp, CRTC_BLANK_CONTROL, CRTC_BLANK_DATA_EN, 0); - WREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i], tmp); - } - } - - WREG32(mmVGA_MEMORY_BASE_ADDRESS_HIGH, upper_32_bits(adev->mc.vram_start)); - WREG32(mmVGA_MEMORY_BASE_ADDRESS, lower_32_bits(adev->mc.vram_start)); - - /* Unlock vga access */ - WREG32(mmVGA_HDP_CONTROL, save->vga_hdp_control); - mdelay(1); - WREG32(mmVGA_RENDER_CONTROL, save->vga_render_control); -} - static void dce_v11_0_set_vga_render_state(struct amdgpu_device *adev, bool render) { @@ -1851,7 +1778,7 @@ static void dce_v11_0_afmt_setmode(struct drm_encoder *encoder, dce_v11_0_audio_write_sad_regs(encoder); dce_v11_0_audio_write_latency_fields(encoder, mode); - err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode); + err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode, false); if (err < 0) { DRM_ERROR("failed to setup AVI infoframe: %zd\n", err); return; @@ -2251,6 +2178,7 @@ static void dce_v11_0_crtc_load_lut(struct drm_crtc *crtc) struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); struct drm_device *dev = crtc->dev; struct amdgpu_device *adev = dev->dev_private; + u16 *r, *g, *b; int i; u32 tmp; @@ -2282,11 +2210,14 @@ static void dce_v11_0_crtc_load_lut(struct drm_crtc *crtc) WREG32(mmDC_LUT_WRITE_EN_MASK + amdgpu_crtc->crtc_offset, 0x00000007); WREG32(mmDC_LUT_RW_INDEX + amdgpu_crtc->crtc_offset, 0); + r = crtc->gamma_store; + g = r + crtc->gamma_size; + b = g + crtc->gamma_size; for (i = 0; i < 256; i++) { WREG32(mmDC_LUT_30_COLOR + amdgpu_crtc->crtc_offset, - (amdgpu_crtc->lut_r[i] << 20) | - (amdgpu_crtc->lut_g[i] << 10) | - (amdgpu_crtc->lut_b[i] << 0)); + ((*r++ & 0xffc0) << 14) | + ((*g++ & 0xffc0) << 4) | + (*b++ >> 6)); } tmp = RREG32(mmDEGAMMA_CONTROL + amdgpu_crtc->crtc_offset); @@ -2575,7 +2506,7 @@ static int dce_v11_0_crtc_cursor_set2(struct drm_crtc *crtc, aobj = gem_to_amdgpu_bo(obj); ret = amdgpu_bo_reserve(aobj, false); if (ret != 0) { - drm_gem_object_unreference_unlocked(obj); + drm_gem_object_put_unlocked(obj); return ret; } @@ -2583,7 +2514,7 @@ static int dce_v11_0_crtc_cursor_set2(struct drm_crtc *crtc, amdgpu_bo_unreserve(aobj); if (ret) { DRM_ERROR("Failed to pin new cursor BO (%d)\n", ret); - drm_gem_object_unreference_unlocked(obj); + drm_gem_object_put_unlocked(obj); return ret; } @@ -2617,7 +2548,7 @@ unpin: amdgpu_bo_unpin(aobj); amdgpu_bo_unreserve(aobj); } - drm_gem_object_unreference_unlocked(amdgpu_crtc->cursor_bo); + drm_gem_object_put_unlocked(amdgpu_crtc->cursor_bo); } amdgpu_crtc->cursor_bo = obj; @@ -2644,15 +2575,6 @@ static int dce_v11_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, u16 *blue, uint32_t size, struct drm_modeset_acquire_ctx *ctx) { - struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - int i; - - /* userspace palettes are always correct as is */ - for (i = 0; i < size; i++) { - amdgpu_crtc->lut_r[i] = red[i] >> 6; - amdgpu_crtc->lut_g[i] = green[i] >> 6; - amdgpu_crtc->lut_b[i] = blue[i] >> 6; - } dce_v11_0_crtc_load_lut(crtc); return 0; @@ -2892,14 +2814,12 @@ static const struct drm_crtc_helper_funcs dce_v11_0_crtc_helper_funcs = { .mode_set_base_atomic = dce_v11_0_crtc_set_base_atomic, .prepare = dce_v11_0_crtc_prepare, .commit = dce_v11_0_crtc_commit, - .load_lut = dce_v11_0_crtc_load_lut, .disable = dce_v11_0_crtc_disable, }; static int dce_v11_0_crtc_init(struct amdgpu_device *adev, int index) { struct amdgpu_crtc *amdgpu_crtc; - int i; amdgpu_crtc = kzalloc(sizeof(struct amdgpu_crtc) + (AMDGPUFB_CONN_LIMIT * sizeof(struct drm_connector *)), GFP_KERNEL); @@ -2917,12 +2837,6 @@ static int dce_v11_0_crtc_init(struct amdgpu_device *adev, int index) adev->ddev->mode_config.cursor_width = amdgpu_crtc->max_cursor_width; adev->ddev->mode_config.cursor_height = amdgpu_crtc->max_cursor_height; - for (i = 0; i < 256; i++) { - amdgpu_crtc->lut_r[i] = i << 2; - amdgpu_crtc->lut_g[i] = i << 2; - amdgpu_crtc->lut_b[i] = i << 2; - } - switch (amdgpu_crtc->crtc_id) { case 0: default: @@ -3086,6 +3000,8 @@ static int dce_v11_0_hw_init(void *handle) dce_v11_0_init_golden_registers(adev); + /* disable vga render */ + dce_v11_0_set_vga_render_state(adev, false); /* init dig PHYs, disp eng pll */ amdgpu_atombios_crtc_powergate_init(adev); amdgpu_atombios_encoder_init_dig(adev); @@ -3806,7 +3722,6 @@ static void dce_v11_0_encoder_add(struct amdgpu_device *adev, } static const struct amdgpu_display_funcs dce_v11_0_display_funcs = { - .set_vga_render_state = &dce_v11_0_set_vga_render_state, .bandwidth_update = &dce_v11_0_bandwidth_update, .vblank_get_counter = &dce_v11_0_vblank_get_counter, .vblank_wait = &dce_v11_0_vblank_wait, @@ -3819,8 +3734,6 @@ static const struct amdgpu_display_funcs dce_v11_0_display_funcs = { .page_flip_get_scanoutpos = &dce_v11_0_crtc_get_scanoutpos, .add_encoder = &dce_v11_0_encoder_add, .add_connector = &amdgpu_connector_add, - .stop_mc_access = &dce_v11_0_stop_mc_access, - .resume_mc_access = &dce_v11_0_resume_mc_access, }; static void dce_v11_0_set_display_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c index fd134a4629d7..a51e35f824a1 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c @@ -42,6 +42,7 @@ #include "dce/dce_6_0_d.h" #include "dce/dce_6_0_sh_mask.h" #include "gca/gfx_7_2_enum.h" +#include "dce_v6_0.h" #include "si_enums.h" static void dce_v6_0_set_display_funcs(struct amdgpu_device *adev); @@ -392,117 +393,6 @@ static u32 dce_v6_0_hpd_get_gpio_reg(struct amdgpu_device *adev) return mmDC_GPIO_HPD_A; } -static u32 evergreen_get_vblank_counter(struct amdgpu_device* adev, int crtc) -{ - if (crtc >= adev->mode_info.num_crtc) - return 0; - else - return RREG32(mmCRTC_STATUS_FRAME_COUNT + crtc_offsets[crtc]); -} - -static void dce_v6_0_stop_mc_access(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) -{ - u32 crtc_enabled, tmp, frame_count; - int i, j; - - save->vga_render_control = RREG32(mmVGA_RENDER_CONTROL); - save->vga_hdp_control = RREG32(mmVGA_HDP_CONTROL); - - /* disable VGA render */ - WREG32(mmVGA_RENDER_CONTROL, 0); - - /* blank the display controllers */ - for (i = 0; i < adev->mode_info.num_crtc; i++) { - crtc_enabled = RREG32(mmCRTC_CONTROL + crtc_offsets[i]) & CRTC_CONTROL__CRTC_MASTER_EN_MASK; - if (crtc_enabled) { - save->crtc_enabled[i] = true; - tmp = RREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i]); - - if (!(tmp & CRTC_BLANK_CONTROL__CRTC_BLANK_DATA_EN_MASK)) { - dce_v6_0_vblank_wait(adev, i); - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1); - tmp |= CRTC_BLANK_CONTROL__CRTC_BLANK_DATA_EN_MASK; - WREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i], tmp); - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0); - } - /* wait for the next frame */ - frame_count = evergreen_get_vblank_counter(adev, i); - for (j = 0; j < adev->usec_timeout; j++) { - if (evergreen_get_vblank_counter(adev, i) != frame_count) - break; - udelay(1); - } - - /* XXX this is a hack to avoid strange behavior with EFI on certain systems */ - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1); - tmp = RREG32(mmCRTC_CONTROL + crtc_offsets[i]); - tmp &= ~CRTC_CONTROL__CRTC_MASTER_EN_MASK; - WREG32(mmCRTC_CONTROL + crtc_offsets[i], tmp); - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0); - save->crtc_enabled[i] = false; - /* ***** */ - } else { - save->crtc_enabled[i] = false; - } - } -} - -static void dce_v6_0_resume_mc_access(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) -{ - u32 tmp; - int i, j; - - /* update crtc base addresses */ - for (i = 0; i < adev->mode_info.num_crtc; i++) { - WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + crtc_offsets[i], - upper_32_bits(adev->mc.vram_start)); - WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS_HIGH + crtc_offsets[i], - upper_32_bits(adev->mc.vram_start)); - WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + crtc_offsets[i], - (u32)adev->mc.vram_start); - WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS + crtc_offsets[i], - (u32)adev->mc.vram_start); - } - - WREG32(mmVGA_MEMORY_BASE_ADDRESS_HIGH, upper_32_bits(adev->mc.vram_start)); - WREG32(mmVGA_MEMORY_BASE_ADDRESS, (u32)adev->mc.vram_start); - - /* unlock regs and wait for update */ - for (i = 0; i < adev->mode_info.num_crtc; i++) { - if (save->crtc_enabled[i]) { - tmp = RREG32(mmMASTER_UPDATE_MODE + crtc_offsets[i]); - if ((tmp & 0x7) != 0) { - tmp &= ~0x7; - WREG32(mmMASTER_UPDATE_MODE + crtc_offsets[i], tmp); - } - tmp = RREG32(mmGRPH_UPDATE + crtc_offsets[i]); - if (tmp & GRPH_UPDATE__GRPH_UPDATE_LOCK_MASK) { - tmp &= ~GRPH_UPDATE__GRPH_UPDATE_LOCK_MASK; - WREG32(mmGRPH_UPDATE + crtc_offsets[i], tmp); - } - tmp = RREG32(mmMASTER_UPDATE_LOCK + crtc_offsets[i]); - if (tmp & 1) { - tmp &= ~1; - WREG32(mmMASTER_UPDATE_LOCK + crtc_offsets[i], tmp); - } - for (j = 0; j < adev->usec_timeout; j++) { - tmp = RREG32(mmGRPH_UPDATE + crtc_offsets[i]); - if ((tmp & GRPH_UPDATE__GRPH_SURFACE_UPDATE_PENDING_MASK) == 0) - break; - udelay(1); - } - } - } - - /* Unlock vga access */ - WREG32(mmVGA_HDP_CONTROL, save->vga_hdp_control); - mdelay(1); - WREG32(mmVGA_RENDER_CONTROL, save->vga_render_control); - -} - static void dce_v6_0_set_vga_render_state(struct amdgpu_device *adev, bool render) { @@ -1597,7 +1487,7 @@ static void dce_v6_0_audio_set_avi_infoframe(struct drm_encoder *encoder, ssize_t err; u32 tmp; - err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode); + err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode, false); if (err < 0) { DRM_ERROR("failed to setup AVI infoframe: %zd\n", err); return; @@ -2182,6 +2072,7 @@ static void dce_v6_0_crtc_load_lut(struct drm_crtc *crtc) struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); struct drm_device *dev = crtc->dev; struct amdgpu_device *adev = dev->dev_private; + u16 *r, *g, *b; int i; DRM_DEBUG_KMS("%d\n", amdgpu_crtc->crtc_id); @@ -2211,11 +2102,14 @@ static void dce_v6_0_crtc_load_lut(struct drm_crtc *crtc) WREG32(mmDC_LUT_WRITE_EN_MASK + amdgpu_crtc->crtc_offset, 0x00000007); WREG32(mmDC_LUT_RW_INDEX + amdgpu_crtc->crtc_offset, 0); + r = crtc->gamma_store; + g = r + crtc->gamma_size; + b = g + crtc->gamma_size; for (i = 0; i < 256; i++) { WREG32(mmDC_LUT_30_COLOR + amdgpu_crtc->crtc_offset, - (amdgpu_crtc->lut_r[i] << 20) | - (amdgpu_crtc->lut_g[i] << 10) | - (amdgpu_crtc->lut_b[i] << 0)); + ((*r++ & 0xffc0) << 14) | + ((*g++ & 0xffc0) << 4) | + (*b++ >> 6)); } WREG32(mmDEGAMMA_CONTROL + amdgpu_crtc->crtc_offset, @@ -2428,7 +2322,7 @@ static int dce_v6_0_crtc_cursor_set2(struct drm_crtc *crtc, aobj = gem_to_amdgpu_bo(obj); ret = amdgpu_bo_reserve(aobj, false); if (ret != 0) { - drm_gem_object_unreference_unlocked(obj); + drm_gem_object_put_unlocked(obj); return ret; } @@ -2436,7 +2330,7 @@ static int dce_v6_0_crtc_cursor_set2(struct drm_crtc *crtc, amdgpu_bo_unreserve(aobj); if (ret) { DRM_ERROR("Failed to pin new cursor BO (%d)\n", ret); - drm_gem_object_unreference_unlocked(obj); + drm_gem_object_put_unlocked(obj); return ret; } @@ -2470,7 +2364,7 @@ unpin: amdgpu_bo_unpin(aobj); amdgpu_bo_unreserve(aobj); } - drm_gem_object_unreference_unlocked(amdgpu_crtc->cursor_bo); + drm_gem_object_put_unlocked(amdgpu_crtc->cursor_bo); } amdgpu_crtc->cursor_bo = obj; @@ -2496,15 +2390,6 @@ static int dce_v6_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, u16 *blue, uint32_t size, struct drm_modeset_acquire_ctx *ctx) { - struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - int i; - - /* userspace palettes are always correct as is */ - for (i = 0; i < size; i++) { - amdgpu_crtc->lut_r[i] = red[i] >> 6; - amdgpu_crtc->lut_g[i] = green[i] >> 6; - amdgpu_crtc->lut_b[i] = blue[i] >> 6; - } dce_v6_0_crtc_load_lut(crtc); return 0; @@ -2712,14 +2597,12 @@ static const struct drm_crtc_helper_funcs dce_v6_0_crtc_helper_funcs = { .mode_set_base_atomic = dce_v6_0_crtc_set_base_atomic, .prepare = dce_v6_0_crtc_prepare, .commit = dce_v6_0_crtc_commit, - .load_lut = dce_v6_0_crtc_load_lut, .disable = dce_v6_0_crtc_disable, }; static int dce_v6_0_crtc_init(struct amdgpu_device *adev, int index) { struct amdgpu_crtc *amdgpu_crtc; - int i; amdgpu_crtc = kzalloc(sizeof(struct amdgpu_crtc) + (AMDGPUFB_CONN_LIMIT * sizeof(struct drm_connector *)), GFP_KERNEL); @@ -2737,12 +2620,6 @@ static int dce_v6_0_crtc_init(struct amdgpu_device *adev, int index) adev->ddev->mode_config.cursor_width = amdgpu_crtc->max_cursor_width; adev->ddev->mode_config.cursor_height = amdgpu_crtc->max_cursor_height; - for (i = 0; i < 256; i++) { - amdgpu_crtc->lut_r[i] = i << 2; - amdgpu_crtc->lut_g[i] = i << 2; - amdgpu_crtc->lut_b[i] = i << 2; - } - amdgpu_crtc->crtc_offset = crtc_offsets[amdgpu_crtc->crtc_id]; amdgpu_crtc->pll_id = ATOM_PPLL_INVALID; @@ -2873,6 +2750,8 @@ static int dce_v6_0_hw_init(void *handle) int i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + /* disable vga render */ + dce_v6_0_set_vga_render_state(adev, false); /* init dig PHYs, disp eng pll */ amdgpu_atombios_encoder_init_dig(adev); amdgpu_atombios_crtc_set_disp_eng_pll(adev, adev->clock.default_dispclk); @@ -3525,7 +3404,6 @@ static void dce_v6_0_encoder_add(struct amdgpu_device *adev, } static const struct amdgpu_display_funcs dce_v6_0_display_funcs = { - .set_vga_render_state = &dce_v6_0_set_vga_render_state, .bandwidth_update = &dce_v6_0_bandwidth_update, .vblank_get_counter = &dce_v6_0_vblank_get_counter, .vblank_wait = &dce_v6_0_vblank_wait, @@ -3538,8 +3416,6 @@ static const struct amdgpu_display_funcs dce_v6_0_display_funcs = { .page_flip_get_scanoutpos = &dce_v6_0_crtc_get_scanoutpos, .add_encoder = &dce_v6_0_encoder_add, .add_connector = &amdgpu_connector_add, - .stop_mc_access = &dce_v6_0_stop_mc_access, - .resume_mc_access = &dce_v6_0_resume_mc_access, }; static void dce_v6_0_set_display_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c index a9e869554627..9cf14b8b2db9 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c @@ -419,81 +419,6 @@ static bool dce_v8_0_is_display_hung(struct amdgpu_device *adev) return true; } -static void dce_v8_0_stop_mc_access(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) -{ - u32 crtc_enabled, tmp; - int i; - - save->vga_render_control = RREG32(mmVGA_RENDER_CONTROL); - save->vga_hdp_control = RREG32(mmVGA_HDP_CONTROL); - - /* disable VGA render */ - tmp = RREG32(mmVGA_RENDER_CONTROL); - tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0); - WREG32(mmVGA_RENDER_CONTROL, tmp); - - /* blank the display controllers */ - for (i = 0; i < adev->mode_info.num_crtc; i++) { - crtc_enabled = REG_GET_FIELD(RREG32(mmCRTC_CONTROL + crtc_offsets[i]), - CRTC_CONTROL, CRTC_MASTER_EN); - if (crtc_enabled) { -#if 1 - save->crtc_enabled[i] = true; - tmp = RREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i]); - if (REG_GET_FIELD(tmp, CRTC_BLANK_CONTROL, CRTC_BLANK_DATA_EN) == 0) { - /*it is correct only for RGB ; black is 0*/ - WREG32(mmCRTC_BLANK_DATA_COLOR + crtc_offsets[i], 0); - tmp = REG_SET_FIELD(tmp, CRTC_BLANK_CONTROL, CRTC_BLANK_DATA_EN, 1); - WREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i], tmp); - } - mdelay(20); -#else - /* XXX this is a hack to avoid strange behavior with EFI on certain systems */ - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1); - tmp = RREG32(mmCRTC_CONTROL + crtc_offsets[i]); - tmp = REG_SET_FIELD(tmp, CRTC_CONTROL, CRTC_MASTER_EN, 0); - WREG32(mmCRTC_CONTROL + crtc_offsets[i], tmp); - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0); - save->crtc_enabled[i] = false; - /* ***** */ -#endif - } else { - save->crtc_enabled[i] = false; - } - } -} - -static void dce_v8_0_resume_mc_access(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) -{ - u32 tmp; - int i; - - /* update crtc base addresses */ - for (i = 0; i < adev->mode_info.num_crtc; i++) { - WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + crtc_offsets[i], - upper_32_bits(adev->mc.vram_start)); - WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + crtc_offsets[i], - (u32)adev->mc.vram_start); - - if (save->crtc_enabled[i]) { - tmp = RREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i]); - tmp = REG_SET_FIELD(tmp, CRTC_BLANK_CONTROL, CRTC_BLANK_DATA_EN, 0); - WREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i], tmp); - } - mdelay(20); - } - - WREG32(mmVGA_MEMORY_BASE_ADDRESS_HIGH, upper_32_bits(adev->mc.vram_start)); - WREG32(mmVGA_MEMORY_BASE_ADDRESS, lower_32_bits(adev->mc.vram_start)); - - /* Unlock vga access */ - WREG32(mmVGA_HDP_CONTROL, save->vga_hdp_control); - mdelay(1); - WREG32(mmVGA_RENDER_CONTROL, save->vga_render_control); -} - static void dce_v8_0_set_vga_render_state(struct amdgpu_device *adev, bool render) { @@ -1750,7 +1675,7 @@ static void dce_v8_0_afmt_setmode(struct drm_encoder *encoder, dce_v8_0_audio_write_sad_regs(encoder); dce_v8_0_audio_write_latency_fields(encoder, mode); - err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode); + err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode, false); if (err < 0) { DRM_ERROR("failed to setup AVI infoframe: %zd\n", err); return; @@ -2124,6 +2049,7 @@ static void dce_v8_0_crtc_load_lut(struct drm_crtc *crtc) struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); struct drm_device *dev = crtc->dev; struct amdgpu_device *adev = dev->dev_private; + u16 *r, *g, *b; int i; DRM_DEBUG_KMS("%d\n", amdgpu_crtc->crtc_id); @@ -2153,11 +2079,14 @@ static void dce_v8_0_crtc_load_lut(struct drm_crtc *crtc) WREG32(mmDC_LUT_WRITE_EN_MASK + amdgpu_crtc->crtc_offset, 0x00000007); WREG32(mmDC_LUT_RW_INDEX + amdgpu_crtc->crtc_offset, 0); + r = crtc->gamma_store; + g = r + crtc->gamma_size; + b = g + crtc->gamma_size; for (i = 0; i < 256; i++) { WREG32(mmDC_LUT_30_COLOR + amdgpu_crtc->crtc_offset, - (amdgpu_crtc->lut_r[i] << 20) | - (amdgpu_crtc->lut_g[i] << 10) | - (amdgpu_crtc->lut_b[i] << 0)); + ((*r++ & 0xffc0) << 14) | + ((*g++ & 0xffc0) << 4) | + (*b++ >> 6)); } WREG32(mmDEGAMMA_CONTROL + amdgpu_crtc->crtc_offset, @@ -2406,7 +2335,7 @@ static int dce_v8_0_crtc_cursor_set2(struct drm_crtc *crtc, aobj = gem_to_amdgpu_bo(obj); ret = amdgpu_bo_reserve(aobj, false); if (ret != 0) { - drm_gem_object_unreference_unlocked(obj); + drm_gem_object_put_unlocked(obj); return ret; } @@ -2414,7 +2343,7 @@ static int dce_v8_0_crtc_cursor_set2(struct drm_crtc *crtc, amdgpu_bo_unreserve(aobj); if (ret) { DRM_ERROR("Failed to pin new cursor BO (%d)\n", ret); - drm_gem_object_unreference_unlocked(obj); + drm_gem_object_put_unlocked(obj); return ret; } @@ -2448,7 +2377,7 @@ unpin: amdgpu_bo_unpin(aobj); amdgpu_bo_unreserve(aobj); } - drm_gem_object_unreference_unlocked(amdgpu_crtc->cursor_bo); + drm_gem_object_put_unlocked(amdgpu_crtc->cursor_bo); } amdgpu_crtc->cursor_bo = obj; @@ -2475,15 +2404,6 @@ static int dce_v8_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, u16 *blue, uint32_t size, struct drm_modeset_acquire_ctx *ctx) { - struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - int i; - - /* userspace palettes are always correct as is */ - for (i = 0; i < size; i++) { - amdgpu_crtc->lut_r[i] = red[i] >> 6; - amdgpu_crtc->lut_g[i] = green[i] >> 6; - amdgpu_crtc->lut_b[i] = blue[i] >> 6; - } dce_v8_0_crtc_load_lut(crtc); return 0; @@ -2702,14 +2622,12 @@ static const struct drm_crtc_helper_funcs dce_v8_0_crtc_helper_funcs = { .mode_set_base_atomic = dce_v8_0_crtc_set_base_atomic, .prepare = dce_v8_0_crtc_prepare, .commit = dce_v8_0_crtc_commit, - .load_lut = dce_v8_0_crtc_load_lut, .disable = dce_v8_0_crtc_disable, }; static int dce_v8_0_crtc_init(struct amdgpu_device *adev, int index) { struct amdgpu_crtc *amdgpu_crtc; - int i; amdgpu_crtc = kzalloc(sizeof(struct amdgpu_crtc) + (AMDGPUFB_CONN_LIMIT * sizeof(struct drm_connector *)), GFP_KERNEL); @@ -2727,12 +2645,6 @@ static int dce_v8_0_crtc_init(struct amdgpu_device *adev, int index) adev->ddev->mode_config.cursor_width = amdgpu_crtc->max_cursor_width; adev->ddev->mode_config.cursor_height = amdgpu_crtc->max_cursor_height; - for (i = 0; i < 256; i++) { - amdgpu_crtc->lut_r[i] = i << 2; - amdgpu_crtc->lut_g[i] = i << 2; - amdgpu_crtc->lut_b[i] = i << 2; - } - amdgpu_crtc->crtc_offset = crtc_offsets[amdgpu_crtc->crtc_id]; amdgpu_crtc->pll_id = ATOM_PPLL_INVALID; @@ -2870,6 +2782,8 @@ static int dce_v8_0_hw_init(void *handle) int i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + /* disable vga render */ + dce_v8_0_set_vga_render_state(adev, false); /* init dig PHYs, disp eng pll */ amdgpu_atombios_encoder_init_dig(adev); amdgpu_atombios_crtc_set_disp_eng_pll(adev, adev->clock.default_dispclk); @@ -3574,7 +3488,6 @@ static void dce_v8_0_encoder_add(struct amdgpu_device *adev, } static const struct amdgpu_display_funcs dce_v8_0_display_funcs = { - .set_vga_render_state = &dce_v8_0_set_vga_render_state, .bandwidth_update = &dce_v8_0_bandwidth_update, .vblank_get_counter = &dce_v8_0_vblank_get_counter, .vblank_wait = &dce_v8_0_vblank_wait, @@ -3587,8 +3500,6 @@ static const struct amdgpu_display_funcs dce_v8_0_display_funcs = { .page_flip_get_scanoutpos = &dce_v8_0_crtc_get_scanoutpos, .add_encoder = &dce_v8_0_encoder_add, .add_connector = &amdgpu_connector_add, - .stop_mc_access = &dce_v8_0_stop_mc_access, - .resume_mc_access = &dce_v8_0_resume_mc_access, }; static void dce_v8_0_set_display_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c index 90bb08309a53..a8829af120c1 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c @@ -95,62 +95,6 @@ static u32 dce_virtual_hpd_get_gpio_reg(struct amdgpu_device *adev) return 0; } -static void dce_virtual_stop_mc_access(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) -{ - switch (adev->asic_type) { -#ifdef CONFIG_DRM_AMDGPU_SI - case CHIP_TAHITI: - case CHIP_PITCAIRN: - case CHIP_VERDE: - case CHIP_OLAND: - dce_v6_0_disable_dce(adev); - break; -#endif -#ifdef CONFIG_DRM_AMDGPU_CIK - case CHIP_BONAIRE: - case CHIP_HAWAII: - case CHIP_KAVERI: - case CHIP_KABINI: - case CHIP_MULLINS: - dce_v8_0_disable_dce(adev); - break; -#endif - case CHIP_FIJI: - case CHIP_TONGA: - dce_v10_0_disable_dce(adev); - break; - case CHIP_CARRIZO: - case CHIP_STONEY: - case CHIP_POLARIS10: - case CHIP_POLARIS11: - case CHIP_POLARIS12: - dce_v11_0_disable_dce(adev); - break; - case CHIP_TOPAZ: -#ifdef CONFIG_DRM_AMDGPU_SI - case CHIP_HAINAN: -#endif - /* no DCE */ - return; - default: - DRM_ERROR("Virtual display unsupported ASIC type: 0x%X\n", adev->asic_type); - } - - return; -} -static void dce_virtual_resume_mc_access(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) -{ - return; -} - -static void dce_virtual_set_vga_render_state(struct amdgpu_device *adev, - bool render) -{ - return; -} - /** * dce_virtual_bandwidth_update - program display watermarks * @@ -168,16 +112,6 @@ static int dce_virtual_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, u16 *blue, uint32_t size, struct drm_modeset_acquire_ctx *ctx) { - struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - int i; - - /* userspace palettes are always correct as is */ - for (i = 0; i < size; i++) { - amdgpu_crtc->lut_r[i] = red[i] >> 6; - amdgpu_crtc->lut_g[i] = green[i] >> 6; - amdgpu_crtc->lut_b[i] = blue[i] >> 6; - } - return 0; } @@ -289,11 +223,6 @@ static int dce_virtual_crtc_set_base(struct drm_crtc *crtc, int x, int y, return 0; } -static void dce_virtual_crtc_load_lut(struct drm_crtc *crtc) -{ - return; -} - static int dce_virtual_crtc_set_base_atomic(struct drm_crtc *crtc, struct drm_framebuffer *fb, int x, int y, enum mode_set_atomic state) @@ -309,14 +238,12 @@ static const struct drm_crtc_helper_funcs dce_virtual_crtc_helper_funcs = { .mode_set_base_atomic = dce_virtual_crtc_set_base_atomic, .prepare = dce_virtual_crtc_prepare, .commit = dce_virtual_crtc_commit, - .load_lut = dce_virtual_crtc_load_lut, .disable = dce_virtual_crtc_disable, }; static int dce_virtual_crtc_init(struct amdgpu_device *adev, int index) { struct amdgpu_crtc *amdgpu_crtc; - int i; amdgpu_crtc = kzalloc(sizeof(struct amdgpu_crtc) + (AMDGPUFB_CONN_LIMIT * sizeof(struct drm_connector *)), GFP_KERNEL); @@ -329,12 +256,6 @@ static int dce_virtual_crtc_init(struct amdgpu_device *adev, int index) amdgpu_crtc->crtc_id = index; adev->mode_info.crtcs[index] = amdgpu_crtc; - for (i = 0; i < 256; i++) { - amdgpu_crtc->lut_r[i] = i << 2; - amdgpu_crtc->lut_g[i] = i << 2; - amdgpu_crtc->lut_b[i] = i << 2; - } - amdgpu_crtc->pll_id = ATOM_PPLL_INVALID; amdgpu_crtc->encoder = NULL; amdgpu_crtc->connector = NULL; @@ -367,7 +288,7 @@ dce_virtual_encoder(struct drm_connector *connector) if (connector->encoder_ids[i] == 0) break; - encoder = drm_encoder_find(connector->dev, connector->encoder_ids[i]); + encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]); if (!encoder) continue; @@ -377,7 +298,7 @@ dce_virtual_encoder(struct drm_connector *connector) /* pick the first one */ if (enc_id) - return drm_encoder_find(connector->dev, enc_id); + return drm_encoder_find(connector->dev, NULL, enc_id); return NULL; } @@ -522,6 +443,47 @@ static int dce_virtual_sw_fini(void *handle) static int dce_virtual_hw_init(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + switch (adev->asic_type) { +#ifdef CONFIG_DRM_AMDGPU_SI + case CHIP_TAHITI: + case CHIP_PITCAIRN: + case CHIP_VERDE: + case CHIP_OLAND: + dce_v6_0_disable_dce(adev); + break; +#endif +#ifdef CONFIG_DRM_AMDGPU_CIK + case CHIP_BONAIRE: + case CHIP_HAWAII: + case CHIP_KAVERI: + case CHIP_KABINI: + case CHIP_MULLINS: + dce_v8_0_disable_dce(adev); + break; +#endif + case CHIP_FIJI: + case CHIP_TONGA: + dce_v10_0_disable_dce(adev); + break; + case CHIP_CARRIZO: + case CHIP_STONEY: + case CHIP_POLARIS11: + case CHIP_POLARIS10: + dce_v11_0_disable_dce(adev); + break; + case CHIP_TOPAZ: +#ifdef CONFIG_DRM_AMDGPU_SI + case CHIP_HAINAN: +#endif + /* no DCE */ + break; + case CHIP_VEGA10: + break; + default: + DRM_ERROR("Virtual display unsupported ASIC type: 0x%X\n", adev->asic_type); + } return 0; } @@ -677,7 +639,6 @@ static int dce_virtual_connector_encoder_init(struct amdgpu_device *adev, } static const struct amdgpu_display_funcs dce_virtual_display_funcs = { - .set_vga_render_state = &dce_virtual_set_vga_render_state, .bandwidth_update = &dce_virtual_bandwidth_update, .vblank_get_counter = &dce_virtual_vblank_get_counter, .vblank_wait = &dce_virtual_vblank_wait, @@ -690,8 +651,6 @@ static const struct amdgpu_display_funcs dce_virtual_display_funcs = { .page_flip_get_scanoutpos = &dce_virtual_crtc_get_scanoutpos, .add_encoder = NULL, .add_connector = NULL, - .stop_mc_access = &dce_virtual_stop_mc_access, - .resume_mc_access = &dce_virtual_resume_mc_access, }; static void dce_virtual_set_display_funcs(struct amdgpu_device *adev) @@ -809,7 +768,7 @@ static const struct amdgpu_irq_src_funcs dce_virtual_crtc_irq_funcs = { static void dce_virtual_set_irq_funcs(struct amdgpu_device *adev) { - adev->crtc_irq.num_types = AMDGPU_CRTC_IRQ_LAST; + adev->crtc_irq.num_types = AMDGPU_CRTC_IRQ_VBLANK6 + 1; adev->crtc_irq.funcs = &dce_virtual_crtc_irq_funcs; } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c index 5173ca1fd159..dbbe986f90f2 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c @@ -636,7 +636,194 @@ static void gfx_v6_0_tiling_mode_table_init(struct amdgpu_device *adev) NUM_BANKS(ADDR_SURF_2_BANK); for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++) WREG32(mmGB_TILE_MODE0 + reg_offset, tilemode[reg_offset]); - } else if (adev->asic_type == CHIP_OLAND || adev->asic_type == CHIP_HAINAN) { + } else if (adev->asic_type == CHIP_OLAND) { + tilemode[0] = MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4); + tilemode[1] = MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4); + tilemode[2] = MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4); + tilemode[3] = MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4); + tilemode[4] = MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) | + ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[5] = MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(split_equal_to_row_size) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[6] = MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(split_equal_to_row_size) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[7] = MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(split_equal_to_row_size) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4); + tilemode[8] = MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) | + ARRAY_MODE(ARRAY_LINEAR_ALIGNED) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[9] = MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) | + ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[10] = MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4); + tilemode[11] = MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[12] = MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[13] = MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) | + ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[14] = MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[15] = MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[16] = MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[17] = MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(split_equal_to_row_size) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[21] = MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[22] = MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4); + tilemode[23] = MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[24] = MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B) | + NUM_BANKS(ADDR_SURF_16_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2); + tilemode[25] = MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) | + ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_1KB) | + NUM_BANKS(ADDR_SURF_8_BANK) | + BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1); + for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++) + WREG32(mmGB_TILE_MODE0 + reg_offset, tilemode[reg_offset]); + } else if (adev->asic_type == CHIP_HAINAN) { tilemode[0] = MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) | ARRAY_MODE(ARRAY_2D_TILED_THIN1) | PIPE_CONFIG(ADDR_SURF_P2) | @@ -1573,7 +1760,7 @@ static void gfx_v6_0_gpu_init(struct amdgpu_device *adev) static void gfx_v6_0_scratch_init(struct amdgpu_device *adev) { - adev->gfx.scratch.num_reg = 7; + adev->gfx.scratch.num_reg = 8; adev->gfx.scratch.reg_base = mmSCRATCH_REG0; adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1; } @@ -2217,40 +2404,9 @@ static void gfx_v6_0_ring_emit_vm_flush(struct amdgpu_ring *ring, static void gfx_v6_0_rlc_fini(struct amdgpu_device *adev) { - int r; - - if (adev->gfx.rlc.save_restore_obj) { - r = amdgpu_bo_reserve(adev->gfx.rlc.save_restore_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve RLC sr bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.rlc.save_restore_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj); - - amdgpu_bo_unref(&adev->gfx.rlc.save_restore_obj); - adev->gfx.rlc.save_restore_obj = NULL; - } - - if (adev->gfx.rlc.clear_state_obj) { - r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve RLC c bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); - - amdgpu_bo_unref(&adev->gfx.rlc.clear_state_obj); - adev->gfx.rlc.clear_state_obj = NULL; - } - - if (adev->gfx.rlc.cp_table_obj) { - r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.rlc.cp_table_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); - - amdgpu_bo_unref(&adev->gfx.rlc.cp_table_obj); - adev->gfx.rlc.cp_table_obj = NULL; - } + amdgpu_bo_free_kernel(&adev->gfx.rlc.save_restore_obj, NULL, NULL); + amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, NULL, NULL); + amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, NULL, NULL); } static int gfx_v6_0_rlc_init(struct amdgpu_device *adev) @@ -2273,43 +2429,23 @@ static int gfx_v6_0_rlc_init(struct amdgpu_device *adev) if (src_ptr) { /* save restore block */ - if (adev->gfx.rlc.save_restore_obj == NULL) { - r = amdgpu_bo_create(adev, dws * 4, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED, - NULL, NULL, - &adev->gfx.rlc.save_restore_obj); - - if (r) { - dev_warn(adev->dev, "(%d) create RLC sr bo failed\n", r); - return r; - } - } - - r = amdgpu_bo_reserve(adev->gfx.rlc.save_restore_obj, false); - if (unlikely(r != 0)) { - gfx_v6_0_rlc_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.rlc.save_restore_obj, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.save_restore_gpu_addr); + r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.save_restore_obj, + &adev->gfx.rlc.save_restore_gpu_addr, + (void **)&adev->gfx.rlc.sr_ptr); if (r) { - amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj); - dev_warn(adev->dev, "(%d) pin RLC sr bo failed\n", r); + dev_warn(adev->dev, "(%d) create RLC sr bo failed\n", + r); gfx_v6_0_rlc_fini(adev); return r; } - r = amdgpu_bo_kmap(adev->gfx.rlc.save_restore_obj, (void **)&adev->gfx.rlc.sr_ptr); - if (r) { - dev_warn(adev->dev, "(%d) map RLC sr bo failed\n", r); - gfx_v6_0_rlc_fini(adev); - return r; - } /* write the sr buffer */ dst_ptr = adev->gfx.rlc.sr_ptr; for (i = 0; i < adev->gfx.rlc.reg_list_size; i++) dst_ptr[i] = cpu_to_le32(src_ptr[i]); + amdgpu_bo_kunmap(adev->gfx.rlc.save_restore_obj); amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj); } @@ -2319,39 +2455,17 @@ static int gfx_v6_0_rlc_init(struct amdgpu_device *adev) adev->gfx.rlc.clear_state_size = gfx_v6_0_get_csb_size(adev); dws = adev->gfx.rlc.clear_state_size + (256 / 4); - if (adev->gfx.rlc.clear_state_obj == NULL) { - r = amdgpu_bo_create(adev, dws * 4, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED, - NULL, NULL, - &adev->gfx.rlc.clear_state_obj); - - if (r) { - dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r); - gfx_v6_0_rlc_fini(adev); - return r; - } - } - r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, false); - if (unlikely(r != 0)) { - gfx_v6_0_rlc_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.rlc.clear_state_obj, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.clear_state_gpu_addr); + r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); if (r) { - amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); - dev_warn(adev->dev, "(%d) pin RLC c bo failed\n", r); + dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r); gfx_v6_0_rlc_fini(adev); return r; } - r = amdgpu_bo_kmap(adev->gfx.rlc.clear_state_obj, (void **)&adev->gfx.rlc.cs_ptr); - if (r) { - dev_warn(adev->dev, "(%d) map RLC c bo failed\n", r); - gfx_v6_0_rlc_fini(adev); - return r; - } /* set up the cs buffer */ dst_ptr = adev->gfx.rlc.cs_ptr; reg_list_mc_addr = adev->gfx.rlc.clear_state_gpu_addr + 256; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index 37b45e4403d1..5c8a7a48a4ad 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -1823,7 +1823,7 @@ static void gfx_v7_0_setup_rb(struct amdgpu_device *adev) } /** - * gmc_v7_0_init_compute_vmid - gart enable + * gfx_v7_0_init_compute_vmid - gart enable * * @adev: amdgpu_device pointer * @@ -1833,7 +1833,7 @@ static void gfx_v7_0_setup_rb(struct amdgpu_device *adev) #define DEFAULT_SH_MEM_BASES (0x6000) #define FIRST_COMPUTE_VMID (8) #define LAST_COMPUTE_VMID (16) -static void gmc_v7_0_init_compute_vmid(struct amdgpu_device *adev) +static void gfx_v7_0_init_compute_vmid(struct amdgpu_device *adev) { int i; uint32_t sh_mem_config; @@ -1921,6 +1921,7 @@ static void gfx_v7_0_gpu_init(struct amdgpu_device *adev) ELEMENT_SIZE, 1); sh_static_mem_cfg = REG_SET_FIELD(sh_static_mem_cfg, SH_STATIC_MEM_CONFIG, INDEX_STRIDE, 3); + WREG32(mmSH_STATIC_MEM_CONFIG, sh_static_mem_cfg); mutex_lock(&adev->srbm_mutex); for (i = 0; i < adev->vm_manager.id_mgr[0].num_ids; i++) { @@ -1934,12 +1935,11 @@ static void gfx_v7_0_gpu_init(struct amdgpu_device *adev) WREG32(mmSH_MEM_APE1_BASE, 1); WREG32(mmSH_MEM_APE1_LIMIT, 0); WREG32(mmSH_MEM_BASES, sh_mem_base); - WREG32(mmSH_STATIC_MEM_CONFIG, sh_static_mem_cfg); } cik_srbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); - gmc_v7_0_init_compute_vmid(adev); + gfx_v7_0_init_compute_vmid(adev); WREG32(mmSX_DEBUG_1, 0x20); @@ -2021,7 +2021,7 @@ static void gfx_v7_0_gpu_init(struct amdgpu_device *adev) */ static void gfx_v7_0_scratch_init(struct amdgpu_device *adev) { - adev->gfx.scratch.num_reg = 7; + adev->gfx.scratch.num_reg = 8; adev->gfx.scratch.reg_base = mmSCRATCH_REG0; adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1; } @@ -2774,39 +2774,18 @@ static int gfx_v7_0_cp_compute_load_microcode(struct amdgpu_device *adev) */ static void gfx_v7_0_cp_compute_fini(struct amdgpu_device *adev) { - int i, r; + int i; for (i = 0; i < adev->gfx.num_compute_rings; i++) { struct amdgpu_ring *ring = &adev->gfx.compute_ring[i]; - if (ring->mqd_obj) { - r = amdgpu_bo_reserve(ring->mqd_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve MQD bo failed\n", r); - - amdgpu_bo_unpin(ring->mqd_obj); - amdgpu_bo_unreserve(ring->mqd_obj); - - amdgpu_bo_unref(&ring->mqd_obj); - ring->mqd_obj = NULL; - } + amdgpu_bo_free_kernel(&ring->mqd_obj, NULL, NULL); } } static void gfx_v7_0_mec_fini(struct amdgpu_device *adev) { - int r; - - if (adev->gfx.mec.hpd_eop_obj) { - r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve HPD EOP bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.mec.hpd_eop_obj); - amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj); - - amdgpu_bo_unref(&adev->gfx.mec.hpd_eop_obj); - adev->gfx.mec.hpd_eop_obj = NULL; - } + amdgpu_bo_free_kernel(&adev->gfx.mec.hpd_eop_obj, NULL, NULL); } static int gfx_v7_0_mec_init(struct amdgpu_device *adev) @@ -2823,33 +2802,14 @@ static int gfx_v7_0_mec_init(struct amdgpu_device *adev) /* allocate space for ALL pipes (even the ones we don't own) */ mec_hpd_size = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec * GFX7_MEC_HPD_SIZE * 2; - if (adev->gfx.mec.hpd_eop_obj == NULL) { - r = amdgpu_bo_create(adev, - mec_hpd_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, - &adev->gfx.mec.hpd_eop_obj); - if (r) { - dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r); - return r; - } - } - r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, false); - if (unlikely(r != 0)) { - gfx_v7_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.mec.hpd_eop_obj, AMDGPU_GEM_DOMAIN_GTT, - &adev->gfx.mec.hpd_eop_gpu_addr); - if (r) { - dev_warn(adev->dev, "(%d) pin HDP EOP bo failed\n", r); - gfx_v7_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_kmap(adev->gfx.mec.hpd_eop_obj, (void **)&hpd); + r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &adev->gfx.mec.hpd_eop_obj, + &adev->gfx.mec.hpd_eop_gpu_addr, + (void **)&hpd); if (r) { - dev_warn(adev->dev, "(%d) map HDP EOP bo failed\n", r); + dev_warn(adev->dev, "(%d) create, pin or map of HDP EOP bo failed\n", r); gfx_v7_0_mec_fini(adev); return r; } @@ -3108,32 +3068,12 @@ static int gfx_v7_0_compute_queue_init(struct amdgpu_device *adev, int ring_id) struct cik_mqd *mqd; struct amdgpu_ring *ring = &adev->gfx.compute_ring[ring_id]; - if (ring->mqd_obj == NULL) { - r = amdgpu_bo_create(adev, - sizeof(struct cik_mqd), - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, - &ring->mqd_obj); - if (r) { - dev_warn(adev->dev, "(%d) create MQD bo failed\n", r); - return r; - } - } - - r = amdgpu_bo_reserve(ring->mqd_obj, false); - if (unlikely(r != 0)) - goto out; - - r = amdgpu_bo_pin(ring->mqd_obj, AMDGPU_GEM_DOMAIN_GTT, - &mqd_gpu_addr); - if (r) { - dev_warn(adev->dev, "(%d) pin MQD bo failed\n", r); - goto out_unreserve; - } - r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&mqd); + r = amdgpu_bo_create_reserved(adev, sizeof(struct cik_mqd), PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj, + &mqd_gpu_addr, (void **)&mqd); if (r) { - dev_warn(adev->dev, "(%d) map MQD bo failed\n", r); - goto out_unreserve; + dev_warn(adev->dev, "(%d) create MQD bo failed\n", r); + return r; } mutex_lock(&adev->srbm_mutex); @@ -3147,9 +3087,7 @@ static int gfx_v7_0_compute_queue_init(struct amdgpu_device *adev, int ring_id) mutex_unlock(&adev->srbm_mutex); amdgpu_bo_kunmap(ring->mqd_obj); -out_unreserve: amdgpu_bo_unreserve(ring->mqd_obj); -out: return 0; } @@ -3361,43 +3299,9 @@ static void gfx_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring, */ static void gfx_v7_0_rlc_fini(struct amdgpu_device *adev) { - int r; - - /* save restore block */ - if (adev->gfx.rlc.save_restore_obj) { - r = amdgpu_bo_reserve(adev->gfx.rlc.save_restore_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve RLC sr bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.rlc.save_restore_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj); - - amdgpu_bo_unref(&adev->gfx.rlc.save_restore_obj); - adev->gfx.rlc.save_restore_obj = NULL; - } - - /* clear state block */ - if (adev->gfx.rlc.clear_state_obj) { - r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve RLC c bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); - - amdgpu_bo_unref(&adev->gfx.rlc.clear_state_obj); - adev->gfx.rlc.clear_state_obj = NULL; - } - - /* clear state block */ - if (adev->gfx.rlc.cp_table_obj) { - r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.rlc.cp_table_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); - - amdgpu_bo_unref(&adev->gfx.rlc.cp_table_obj); - adev->gfx.rlc.cp_table_obj = NULL; - } + amdgpu_bo_free_kernel(&adev->gfx.rlc.save_restore_obj, NULL, NULL); + amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, NULL, NULL); + amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, NULL, NULL); } static int gfx_v7_0_rlc_init(struct amdgpu_device *adev) @@ -3432,39 +3336,17 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev) if (src_ptr) { /* save restore block */ - if (adev->gfx.rlc.save_restore_obj == NULL) { - r = amdgpu_bo_create(adev, dws * 4, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, - &adev->gfx.rlc.save_restore_obj); - if (r) { - dev_warn(adev->dev, "(%d) create RLC sr bo failed\n", r); - return r; - } - } - - r = amdgpu_bo_reserve(adev->gfx.rlc.save_restore_obj, false); - if (unlikely(r != 0)) { - gfx_v7_0_rlc_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.rlc.save_restore_obj, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.save_restore_gpu_addr); + r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.save_restore_obj, + &adev->gfx.rlc.save_restore_gpu_addr, + (void **)&adev->gfx.rlc.sr_ptr); if (r) { - amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj); - dev_warn(adev->dev, "(%d) pin RLC sr bo failed\n", r); + dev_warn(adev->dev, "(%d) create, pin or map of RLC sr bo failed\n", r); gfx_v7_0_rlc_fini(adev); return r; } - r = amdgpu_bo_kmap(adev->gfx.rlc.save_restore_obj, (void **)&adev->gfx.rlc.sr_ptr); - if (r) { - dev_warn(adev->dev, "(%d) map RLC sr bo failed\n", r); - gfx_v7_0_rlc_fini(adev); - return r; - } /* write the sr buffer */ dst_ptr = adev->gfx.rlc.sr_ptr; for (i = 0; i < adev->gfx.rlc.reg_list_size; i++) @@ -3477,39 +3359,17 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev) /* clear state block */ adev->gfx.rlc.clear_state_size = dws = gfx_v7_0_get_csb_size(adev); - if (adev->gfx.rlc.clear_state_obj == NULL) { - r = amdgpu_bo_create(adev, dws * 4, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, - &adev->gfx.rlc.clear_state_obj); - if (r) { - dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r); - gfx_v7_0_rlc_fini(adev); - return r; - } - } - r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, false); - if (unlikely(r != 0)) { - gfx_v7_0_rlc_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.rlc.clear_state_obj, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.clear_state_gpu_addr); + r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); if (r) { - amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); - dev_warn(adev->dev, "(%d) pin RLC c bo failed\n", r); + dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r); gfx_v7_0_rlc_fini(adev); return r; } - r = amdgpu_bo_kmap(adev->gfx.rlc.clear_state_obj, (void **)&adev->gfx.rlc.cs_ptr); - if (r) { - dev_warn(adev->dev, "(%d) map RLC c bo failed\n", r); - gfx_v7_0_rlc_fini(adev); - return r; - } /* set up the cs buffer */ dst_ptr = adev->gfx.rlc.cs_ptr; gfx_v7_0_get_csb_buffer(adev, dst_ptr); @@ -3518,37 +3378,14 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev) } if (adev->gfx.rlc.cp_table_size) { - if (adev->gfx.rlc.cp_table_obj == NULL) { - r = amdgpu_bo_create(adev, adev->gfx.rlc.cp_table_size, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, - &adev->gfx.rlc.cp_table_obj); - if (r) { - dev_warn(adev->dev, "(%d) create RLC cp table bo failed\n", r); - gfx_v7_0_rlc_fini(adev); - return r; - } - } - r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, false); - if (unlikely(r != 0)) { - dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r); - gfx_v7_0_rlc_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.rlc.cp_table_obj, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.cp_table_gpu_addr); + r = amdgpu_bo_create_reserved(adev, adev->gfx.rlc.cp_table_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.cp_table_obj, + &adev->gfx.rlc.cp_table_gpu_addr, + (void **)&adev->gfx.rlc.cp_table_ptr); if (r) { - amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); - dev_warn(adev->dev, "(%d) pin RLC cp_table bo failed\n", r); - gfx_v7_0_rlc_fini(adev); - return r; - } - r = amdgpu_bo_kmap(adev->gfx.rlc.cp_table_obj, (void **)&adev->gfx.rlc.cp_table_ptr); - if (r) { - dev_warn(adev->dev, "(%d) map RLC cp table bo failed\n", r); + dev_warn(adev->dev, "(%d) create RLC cp table bo failed\n", r); gfx_v7_0_rlc_fini(adev); return r; } @@ -4833,6 +4670,14 @@ static int gfx_v7_0_sw_fini(void *handle) gfx_v7_0_cp_compute_fini(adev); gfx_v7_0_rlc_fini(adev); gfx_v7_0_mec_fini(adev); + amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); + if (adev->gfx.rlc.cp_table_size) { + amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, + &adev->gfx.rlc.cp_table_gpu_addr, + (void **)&adev->gfx.rlc.cp_table_ptr); + } gfx_v7_0_free_microcode(adev); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index aa5a50f5eac8..9ecdf621a74a 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -20,6 +20,7 @@ * OTHER DEALINGS IN THE SOFTWARE. * */ +#include <linux/kernel.h> #include <linux/firmware.h> #include <drm/drmP.h> #include "amdgpu.h" @@ -125,24 +126,39 @@ MODULE_FIRMWARE("amdgpu/fiji_mec2.bin"); MODULE_FIRMWARE("amdgpu/fiji_rlc.bin"); MODULE_FIRMWARE("amdgpu/polaris11_ce.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_ce_2.bin"); MODULE_FIRMWARE("amdgpu/polaris11_pfp.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_pfp_2.bin"); MODULE_FIRMWARE("amdgpu/polaris11_me.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_me_2.bin"); MODULE_FIRMWARE("amdgpu/polaris11_mec.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_mec_2.bin"); MODULE_FIRMWARE("amdgpu/polaris11_mec2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_mec2_2.bin"); MODULE_FIRMWARE("amdgpu/polaris11_rlc.bin"); MODULE_FIRMWARE("amdgpu/polaris10_ce.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_ce_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_pfp.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_pfp_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_me.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_me_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_mec.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_mec_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_mec2.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_mec2_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_rlc.bin"); MODULE_FIRMWARE("amdgpu/polaris12_ce.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_ce_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_pfp.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_pfp_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_me.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_me_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_mec.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_mec_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_mec2.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_mec2_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_rlc.bin"); static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] = @@ -193,8 +209,8 @@ static const u32 tonga_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x22011003, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF }; static const u32 tonga_mgcg_cgcg_init[] = @@ -303,8 +319,8 @@ static const u32 polaris11_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x22011002, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF, + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF, }; static const u32 golden_settings_polaris10_a11[] = @@ -336,8 +352,8 @@ static const u32 polaris10_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x22011003, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF, + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF, }; static const u32 fiji_golden_common_all[] = @@ -348,8 +364,8 @@ static const u32 fiji_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x22011003, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF, + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF, mmGRBM_GFX_INDEX, 0xffffffff, 0xe0000000, mmSPI_CONFIG_CNTL_1, 0x0000000f, 0x00000009, }; @@ -436,8 +452,8 @@ static const u32 iceland_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x22010001, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF }; static const u32 iceland_mgcg_cgcg_init[] = @@ -532,8 +548,8 @@ static const u32 cz_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x22010001, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF }; static const u32 cz_mgcg_cgcg_init[] = @@ -637,8 +653,8 @@ static const u32 stoney_golden_common_all[] = mmGB_ADDR_CONFIG, 0xffffffff, 0x12010001, mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, - mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00007FBF, - mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00007FAF, + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF, }; static const u32 stoney_mgcg_cgcg_init[] = @@ -750,7 +766,7 @@ static void gfx_v8_0_init_golden_registers(struct amdgpu_device *adev) static void gfx_v8_0_scratch_init(struct amdgpu_device *adev) { - adev->gfx.scratch.num_reg = 7; + adev->gfx.scratch.num_reg = 8; adev->gfx.scratch.reg_base = mmSCRATCH_REG0; adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1; } @@ -918,8 +934,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) BUG(); } - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name); - err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp_2.bin", chip_name); + err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); + if (err == -ENOENT) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name); + err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); + } + } else { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name); + err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); + } if (err) goto out; err = amdgpu_ucode_validate(adev->gfx.pfp_fw); @@ -929,8 +954,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) adev->gfx.pfp_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); - err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me_2.bin", chip_name); + err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); + if (err == -ENOENT) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); + err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); + } + } else { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); + err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); + } if (err) goto out; err = amdgpu_ucode_validate(adev->gfx.me_fw); @@ -941,8 +975,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); - err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce_2.bin", chip_name); + err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); + if (err == -ENOENT) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); + err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); + } + } else { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); + err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); + } if (err) goto out; err = amdgpu_ucode_validate(adev->gfx.ce_fw); @@ -1012,8 +1055,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) for (i = 0 ; i < (rlc_hdr->reg_list_size_bytes >> 2); i++) adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); - err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec_2.bin", chip_name); + err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); + if (err == -ENOENT) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); + err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); + } + } else { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); + err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); + } if (err) goto out; err = amdgpu_ucode_validate(adev->gfx.mec_fw); @@ -1025,8 +1077,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) if ((adev->asic_type != CHIP_STONEY) && (adev->asic_type != CHIP_TOPAZ)) { - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); - err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2_2.bin", chip_name); + err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); + if (err == -ENOENT) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); + err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); + } + } else { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); + err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); + } if (!err) { err = amdgpu_ucode_validate(adev->gfx.mec2_fw); if (err) @@ -1238,29 +1299,8 @@ static void cz_init_cp_jump_table(struct amdgpu_device *adev) static void gfx_v8_0_rlc_fini(struct amdgpu_device *adev) { - int r; - - /* clear state block */ - if (adev->gfx.rlc.clear_state_obj) { - r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve RLC cbs bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); - amdgpu_bo_unref(&adev->gfx.rlc.clear_state_obj); - adev->gfx.rlc.clear_state_obj = NULL; - } - - /* jump table block */ - if (adev->gfx.rlc.cp_table_obj) { - r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.rlc.cp_table_obj); - amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); - amdgpu_bo_unref(&adev->gfx.rlc.cp_table_obj); - adev->gfx.rlc.cp_table_obj = NULL; - } + amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, NULL, NULL); + amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, NULL, NULL); } static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) @@ -1278,39 +1318,17 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) /* clear state block */ adev->gfx.rlc.clear_state_size = dws = gfx_v8_0_get_csb_size(adev); - if (adev->gfx.rlc.clear_state_obj == NULL) { - r = amdgpu_bo_create(adev, dws * 4, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, - &adev->gfx.rlc.clear_state_obj); - if (r) { - dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r); - gfx_v8_0_rlc_fini(adev); - return r; - } - } - r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, false); - if (unlikely(r != 0)) { - gfx_v8_0_rlc_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.rlc.clear_state_obj, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.clear_state_gpu_addr); + r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); if (r) { - amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); - dev_warn(adev->dev, "(%d) pin RLC cbs bo failed\n", r); + dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r); gfx_v8_0_rlc_fini(adev); return r; } - r = amdgpu_bo_kmap(adev->gfx.rlc.clear_state_obj, (void **)&adev->gfx.rlc.cs_ptr); - if (r) { - dev_warn(adev->dev, "(%d) map RLC cbs bo failed\n", r); - gfx_v8_0_rlc_fini(adev); - return r; - } /* set up the cs buffer */ dst_ptr = adev->gfx.rlc.cs_ptr; gfx_v8_0_get_csb_buffer(adev, dst_ptr); @@ -1321,34 +1339,13 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) if ((adev->asic_type == CHIP_CARRIZO) || (adev->asic_type == CHIP_STONEY)) { adev->gfx.rlc.cp_table_size = ALIGN(96 * 5 * 4, 2048) + (64 * 1024); /* JT + GDS */ - if (adev->gfx.rlc.cp_table_obj == NULL) { - r = amdgpu_bo_create(adev, adev->gfx.rlc.cp_table_size, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, - &adev->gfx.rlc.cp_table_obj); - if (r) { - dev_warn(adev->dev, "(%d) create RLC cp table bo failed\n", r); - return r; - } - } - - r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, false); - if (unlikely(r != 0)) { - dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r); - return r; - } - r = amdgpu_bo_pin(adev->gfx.rlc.cp_table_obj, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.cp_table_gpu_addr); + r = amdgpu_bo_create_reserved(adev, adev->gfx.rlc.cp_table_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.cp_table_obj, + &adev->gfx.rlc.cp_table_gpu_addr, + (void **)&adev->gfx.rlc.cp_table_ptr); if (r) { - amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); - dev_warn(adev->dev, "(%d) pin RLC cp table bo failed\n", r); - return r; - } - r = amdgpu_bo_kmap(adev->gfx.rlc.cp_table_obj, (void **)&adev->gfx.rlc.cp_table_ptr); - if (r) { - dev_warn(adev->dev, "(%d) map RLC cp table bo failed\n", r); + dev_warn(adev->dev, "(%d) create RLC cp table bo failed\n", r); return r; } @@ -1363,17 +1360,7 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) static void gfx_v8_0_mec_fini(struct amdgpu_device *adev) { - int r; - - if (adev->gfx.mec.hpd_eop_obj) { - r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve HPD EOP bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.mec.hpd_eop_obj); - amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj); - amdgpu_bo_unref(&adev->gfx.mec.hpd_eop_obj); - adev->gfx.mec.hpd_eop_obj = NULL; - } + amdgpu_bo_free_kernel(&adev->gfx.mec.hpd_eop_obj, NULL, NULL); } static int gfx_v8_0_mec_init(struct amdgpu_device *adev) @@ -1389,34 +1376,13 @@ static int gfx_v8_0_mec_init(struct amdgpu_device *adev) mec_hpd_size = adev->gfx.num_compute_rings * GFX8_MEC_HPD_SIZE; - if (adev->gfx.mec.hpd_eop_obj == NULL) { - r = amdgpu_bo_create(adev, - mec_hpd_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, - &adev->gfx.mec.hpd_eop_obj); - if (r) { - dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r); - return r; - } - } - - r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, false); - if (unlikely(r != 0)) { - gfx_v8_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.mec.hpd_eop_obj, AMDGPU_GEM_DOMAIN_GTT, - &adev->gfx.mec.hpd_eop_gpu_addr); + r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &adev->gfx.mec.hpd_eop_obj, + &adev->gfx.mec.hpd_eop_gpu_addr, + (void **)&hpd); if (r) { - dev_warn(adev->dev, "(%d) pin HDP EOP bo failed\n", r); - gfx_v8_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_kmap(adev->gfx.mec.hpd_eop_obj, (void **)&hpd); - if (r) { - dev_warn(adev->dev, "(%d) map HDP EOP bo failed\n", r); - gfx_v8_0_mec_fini(adev); + dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r); return r; } @@ -2148,9 +2114,19 @@ static int gfx_v8_0_sw_fini(void *handle) amdgpu_gfx_compute_mqd_sw_fini(adev); amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring, &adev->gfx.kiq.irq); amdgpu_gfx_kiq_fini(adev); + amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL); gfx_v8_0_mec_fini(adev); gfx_v8_0_rlc_fini(adev); + amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); + if ((adev->asic_type == CHIP_CARRIZO) || + (adev->asic_type == CHIP_STONEY)) { + amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, + &adev->gfx.rlc.cp_table_gpu_addr, + (void **)&adev->gfx.rlc.cp_table_ptr); + } gfx_v8_0_free_microcode(adev); return 0; @@ -3802,6 +3778,8 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) ELEMENT_SIZE, 1); sh_static_mem_cfg = REG_SET_FIELD(sh_static_mem_cfg, SH_STATIC_MEM_CONFIG, INDEX_STRIDE, 3); + WREG32(mmSH_STATIC_MEM_CONFIG, sh_static_mem_cfg); + mutex_lock(&adev->srbm_mutex); for (i = 0; i < adev->vm_manager.id_mgr[0].num_ids; i++) { vi_srbm_select(adev, 0, 0, 0, i); @@ -3825,7 +3803,6 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) WREG32(mmSH_MEM_APE1_BASE, 1); WREG32(mmSH_MEM_APE1_LIMIT, 0); - WREG32(mmSH_STATIC_MEM_CONFIG, sh_static_mem_cfg); } vi_srbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); @@ -3985,10 +3962,10 @@ static int gfx_v8_0_init_save_restore_list(struct amdgpu_device *adev) adev->gfx.rlc.reg_list_format_size_bytes >> 2, unique_indices, &indices_count, - sizeof(unique_indices) / sizeof(int), + ARRAY_SIZE(unique_indices), indirect_start_offsets, &offset_count, - sizeof(indirect_start_offsets)/sizeof(int)); + ARRAY_SIZE(indirect_start_offsets)); /* save and restore list */ WREG32_FIELD(RLC_SRM_CNTL, AUTO_INCR_ADDR, 1); @@ -4010,14 +3987,14 @@ static int gfx_v8_0_init_save_restore_list(struct amdgpu_device *adev) /* starting offsets starts */ WREG32(mmRLC_GPM_SCRATCH_ADDR, adev->gfx.rlc.starting_offsets_start); - for (i = 0; i < sizeof(indirect_start_offsets)/sizeof(int); i++) + for (i = 0; i < ARRAY_SIZE(indirect_start_offsets); i++) WREG32(mmRLC_GPM_SCRATCH_DATA, indirect_start_offsets[i]); /* unique indices */ temp = mmRLC_SRM_INDEX_CNTL_ADDR_0; data = mmRLC_SRM_INDEX_CNTL_DATA_0; - for (i = 0; i < sizeof(unique_indices) / sizeof(int); i++) { + for (i = 0; i < ARRAY_SIZE(unique_indices); i++) { if (unique_indices[i] != 0) { WREG32(temp + i, unique_indices[i] & 0x3FFFF); WREG32(data + i, unique_indices[i] >> 20); @@ -4165,18 +4142,12 @@ static int gfx_v8_0_rlc_resume(struct amdgpu_device *adev) gfx_v8_0_rlc_reset(adev); gfx_v8_0_init_pg(adev); - if (!adev->pp_enabled) { - if (adev->firmware.load_type != AMDGPU_FW_LOAD_SMU) { - /* legacy rlc firmware loading */ - r = gfx_v8_0_rlc_load_microcode(adev); - if (r) - return r; - } else { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_RLC_G); - if (r) - return -EINVAL; - } + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + /* legacy rlc firmware loading */ + r = gfx_v8_0_rlc_load_microcode(adev); + if (r) + return r; } gfx_v8_0_rlc_start(adev); @@ -4564,7 +4535,7 @@ static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev) /* This situation may be hit in the future if a new HW * generation exposes more than 64 queues. If so, the * definition of queue_mask needs updating */ - if (WARN_ON(i > (sizeof(queue_mask)*8))) { + if (WARN_ON(i >= (sizeof(queue_mask)*8))) { DRM_ERROR("Invalid KCQ enabled: %d\n", i); break; } @@ -4671,12 +4642,10 @@ static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring) mqd->compute_static_thread_mgmt_se2 = 0xffffffff; mqd->compute_static_thread_mgmt_se3 = 0xffffffff; mqd->compute_misc_reserved = 0x00000003; - if (!(adev->flags & AMD_IS_APU)) { - mqd->dynamic_cu_mask_addr_lo = lower_32_bits(ring->mqd_gpu_addr - + offsetof(struct vi_mqd_allocation, dyamic_cu_mask)); - mqd->dynamic_cu_mask_addr_hi = upper_32_bits(ring->mqd_gpu_addr - + offsetof(struct vi_mqd_allocation, dyamic_cu_mask)); - } + mqd->dynamic_cu_mask_addr_lo = lower_32_bits(ring->mqd_gpu_addr + + offsetof(struct vi_mqd_allocation, dynamic_cu_mask)); + mqd->dynamic_cu_mask_addr_hi = upper_32_bits(ring->mqd_gpu_addr + + offsetof(struct vi_mqd_allocation, dynamic_cu_mask)); eop_base_addr = ring->eop_gpu_addr >> 8; mqd->cp_hqd_eop_base_addr_lo = eop_base_addr; mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); @@ -4847,7 +4816,7 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring) gfx_v8_0_kiq_setting(ring); - if (adev->gfx.in_reset) { /* for GPU_RESET case */ + if (adev->in_sriov_reset) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation)); @@ -4862,8 +4831,8 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring) mutex_unlock(&adev->srbm_mutex); } else { memset((void *)mqd, 0, sizeof(struct vi_mqd_allocation)); - ((struct vi_mqd_allocation *)mqd)->dyamic_cu_mask = 0xFFFFFFFF; - ((struct vi_mqd_allocation *)mqd)->dyamic_rb_mask = 0xFFFFFFFF; + ((struct vi_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; + ((struct vi_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; mutex_lock(&adev->srbm_mutex); vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v8_0_mqd_init(ring); @@ -4884,10 +4853,10 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring) struct vi_mqd *mqd = ring->mqd_ptr; int mqd_idx = ring - &adev->gfx.compute_ring[0]; - if (!adev->gfx.in_reset && !adev->gfx.in_suspend) { + if (!adev->in_sriov_reset && !adev->gfx.in_suspend) { memset((void *)mqd, 0, sizeof(struct vi_mqd_allocation)); - ((struct vi_mqd_allocation *)mqd)->dyamic_cu_mask = 0xFFFFFFFF; - ((struct vi_mqd_allocation *)mqd)->dyamic_rb_mask = 0xFFFFFFFF; + ((struct vi_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; + ((struct vi_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; mutex_lock(&adev->srbm_mutex); vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v8_0_mqd_init(ring); @@ -4896,7 +4865,7 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring) if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation)); - } else if (adev->gfx.in_reset) { /* for GPU_RESET case */ + } else if (adev->in_sriov_reset) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation)); @@ -4994,43 +4963,15 @@ static int gfx_v8_0_cp_resume(struct amdgpu_device *adev) if (!(adev->flags & AMD_IS_APU)) gfx_v8_0_enable_gui_idle_interrupt(adev, false); - if (!adev->pp_enabled) { - if (adev->firmware.load_type != AMDGPU_FW_LOAD_SMU) { + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { /* legacy firmware loading */ - r = gfx_v8_0_cp_gfx_load_microcode(adev); - if (r) - return r; + r = gfx_v8_0_cp_gfx_load_microcode(adev); + if (r) + return r; - r = gfx_v8_0_cp_compute_load_microcode(adev); - if (r) - return r; - } else { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_CP_CE); - if (r) - return -EINVAL; - - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_CP_PFP); - if (r) - return -EINVAL; - - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_CP_ME); - if (r) - return -EINVAL; - - if (adev->asic_type == CHIP_TOPAZ) { - r = gfx_v8_0_cp_compute_load_microcode(adev); - if (r) - return r; - } else { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_CP_MEC1); - if (r) - return -EINVAL; - } - } + r = gfx_v8_0_cp_compute_load_microcode(adev); + if (r) + return r; } r = gfx_v8_0_cp_gfx_resume(adev); @@ -5069,12 +5010,69 @@ static int gfx_v8_0_hw_init(void *handle) return r; } +static int gfx_v8_0_kcq_disable(struct amdgpu_ring *kiq_ring,struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = kiq_ring->adev; + uint32_t scratch, tmp = 0; + int r, i; + + r = amdgpu_gfx_scratch_get(adev, &scratch); + if (r) { + DRM_ERROR("Failed to get scratch reg (%d).\n", r); + return r; + } + WREG32(scratch, 0xCAFEDEAD); + + r = amdgpu_ring_alloc(kiq_ring, 10); + if (r) { + DRM_ERROR("Failed to lock KIQ (%d).\n", r); + amdgpu_gfx_scratch_free(adev, scratch); + return r; + } + + /* unmap queues */ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); + amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + PACKET3_UNMAP_QUEUES_ACTION(1) | /* RESET_QUEUES */ + PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) | + PACKET3_UNMAP_QUEUES_ENGINE_SEL(0) | + PACKET3_UNMAP_QUEUES_NUM_QUEUES(1)); + amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + /* write to scratch for completion */ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); + amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(kiq_ring, 0xDEADBEEF); + amdgpu_ring_commit(kiq_ring); + + for (i = 0; i < adev->usec_timeout; i++) { + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + if (i >= adev->usec_timeout) { + DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp); + r = -EINVAL; + } + amdgpu_gfx_scratch_free(adev, scratch); + return r; +} + static int gfx_v8_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i; amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); + + /* disable KCQ to avoid CPC touch memory not valid anymore */ + for (i = 0; i < adev->gfx.num_compute_rings; i++) + gfx_v8_0_kcq_disable(&adev->gfx.kiq.ring, &adev->gfx.compute_ring[i]); + if (amdgpu_sriov_vf(adev)) { pr_debug("For SRIOV client, shouldn't do anything.\n"); return 0; @@ -5996,7 +5994,6 @@ static int gfx_v8_0_tonga_update_gfx_clock_gating(struct amdgpu_device *adev, { uint32_t msg_id, pp_state = 0; uint32_t pp_support_state = 0; - void *pp_handle = adev->powerplay.pp_handle; if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_CGLS)) { if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) { @@ -6014,7 +6011,8 @@ static int gfx_v8_0_tonga_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_CG, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS)) { @@ -6035,7 +6033,8 @@ static int gfx_v8_0_tonga_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_MG, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } return 0; @@ -6047,7 +6046,6 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, uint32_t msg_id, pp_state = 0; uint32_t pp_support_state = 0; - void *pp_handle = adev->powerplay.pp_handle; if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_CGLS)) { if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) { @@ -6065,7 +6063,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_CG, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_3D_CGCG | AMD_CG_SUPPORT_GFX_3D_CGLS)) { @@ -6084,7 +6083,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_3D, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS)) { @@ -6105,7 +6105,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_MG, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & AMD_CG_SUPPORT_GFX_RLC_LS) { @@ -6120,7 +6121,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_RLC, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CP_LS) { @@ -6134,7 +6136,8 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev, PP_BLOCK_GFX_CP, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } return 0; @@ -6401,6 +6404,104 @@ static void gfx_v8_0_ring_set_wptr_compute(struct amdgpu_ring *ring) WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr)); } +static void gfx_v8_0_ring_set_pipe_percent(struct amdgpu_ring *ring, + bool acquire) +{ + struct amdgpu_device *adev = ring->adev; + int pipe_num, tmp, reg; + int pipe_percent = acquire ? SPI_WCL_PIPE_PERCENT_GFX__VALUE_MASK : 0x1; + + pipe_num = ring->me * adev->gfx.mec.num_pipe_per_mec + ring->pipe; + + /* first me only has 2 entries, GFX and HP3D */ + if (ring->me > 0) + pipe_num -= 2; + + reg = mmSPI_WCL_PIPE_PERCENT_GFX + pipe_num; + tmp = RREG32(reg); + tmp = REG_SET_FIELD(tmp, SPI_WCL_PIPE_PERCENT_GFX, VALUE, pipe_percent); + WREG32(reg, tmp); +} + +static void gfx_v8_0_pipe_reserve_resources(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + bool acquire) +{ + int i, pipe; + bool reserve; + struct amdgpu_ring *iring; + + mutex_lock(&adev->gfx.pipe_reserve_mutex); + pipe = amdgpu_gfx_queue_to_bit(adev, ring->me, ring->pipe, 0); + if (acquire) + set_bit(pipe, adev->gfx.pipe_reserve_bitmap); + else + clear_bit(pipe, adev->gfx.pipe_reserve_bitmap); + + if (!bitmap_weight(adev->gfx.pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES)) { + /* Clear all reservations - everyone reacquires all resources */ + for (i = 0; i < adev->gfx.num_gfx_rings; ++i) + gfx_v8_0_ring_set_pipe_percent(&adev->gfx.gfx_ring[i], + true); + + for (i = 0; i < adev->gfx.num_compute_rings; ++i) + gfx_v8_0_ring_set_pipe_percent(&adev->gfx.compute_ring[i], + true); + } else { + /* Lower all pipes without a current reservation */ + for (i = 0; i < adev->gfx.num_gfx_rings; ++i) { + iring = &adev->gfx.gfx_ring[i]; + pipe = amdgpu_gfx_queue_to_bit(adev, + iring->me, + iring->pipe, + 0); + reserve = test_bit(pipe, adev->gfx.pipe_reserve_bitmap); + gfx_v8_0_ring_set_pipe_percent(iring, reserve); + } + + for (i = 0; i < adev->gfx.num_compute_rings; ++i) { + iring = &adev->gfx.compute_ring[i]; + pipe = amdgpu_gfx_queue_to_bit(adev, + iring->me, + iring->pipe, + 0); + reserve = test_bit(pipe, adev->gfx.pipe_reserve_bitmap); + gfx_v8_0_ring_set_pipe_percent(iring, reserve); + } + } + + mutex_unlock(&adev->gfx.pipe_reserve_mutex); +} + +static void gfx_v8_0_hqd_set_priority(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + bool acquire) +{ + uint32_t pipe_priority = acquire ? 0x2 : 0x0; + uint32_t queue_priority = acquire ? 0xf : 0x0; + + mutex_lock(&adev->srbm_mutex); + vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + + WREG32(mmCP_HQD_PIPE_PRIORITY, pipe_priority); + WREG32(mmCP_HQD_QUEUE_PRIORITY, queue_priority); + + vi_srbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); +} +static void gfx_v8_0_ring_set_priority_compute(struct amdgpu_ring *ring, + enum amd_sched_priority priority) +{ + struct amdgpu_device *adev = ring->adev; + bool acquire = priority == AMD_SCHED_PRIORITY_HIGH_HW; + + if (ring->funcs->type != AMDGPU_RING_TYPE_COMPUTE) + return; + + gfx_v8_0_hqd_set_priority(adev, ring, acquire); + gfx_v8_0_pipe_reserve_resources(adev, ring, acquire); +} + static void gfx_v8_0_ring_emit_fence_compute(struct amdgpu_ring *ring, u64 addr, u64 seq, unsigned flags) @@ -6846,6 +6947,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = { .test_ib = gfx_v8_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, + .set_priority = gfx_v8_0_ring_set_priority_compute, }; static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_kiq = { @@ -7054,7 +7156,7 @@ static void gfx_v8_0_ring_emit_ce_meta(struct amdgpu_ring *ring) { uint64_t ce_payload_addr; int cnt_ce; - static union { + union { struct vi_ce_ib_state regular; struct vi_ce_ib_state_chained_ib chained; } ce_payload = {}; @@ -7083,7 +7185,7 @@ static void gfx_v8_0_ring_emit_de_meta(struct amdgpu_ring *ring) { uint64_t de_payload_addr, gds_addr, csa_addr; int cnt_de; - static union { + union { struct vi_de_ib_state regular; struct vi_de_ib_state_chained_ib chained; } de_payload = {}; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 3a0b69b09ed6..da43813d67a4 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -20,6 +20,7 @@ * OTHER DEALINGS IN THE SOFTWARE. * */ +#include <linux/kernel.h> #include <linux/firmware.h> #include <drm/drmP.h> #include "amdgpu.h" @@ -66,38 +67,70 @@ MODULE_FIRMWARE("amdgpu/raven_rlc.bin"); static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] = { - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID0), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID0)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID1), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID1)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID2), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID2)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID3), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID3)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID4), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID4)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID5), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID5)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID6), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID6)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID7), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID7)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID8), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID8)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID9), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID9)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID10), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID10)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID11), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID11)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID12), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID12)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID13), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID13)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID14), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID14)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID15), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID15)} + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID0), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID0) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID1), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID1) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID2), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID2) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID3), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID3) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID4), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID4) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID5), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID5) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID6), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID6) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID7), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID7) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID8), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID8) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID9), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID9) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID10), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID10) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID11), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID11) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID12), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID12)}, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID13), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID13) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID14), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID14) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID15), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID15) } }; static const u32 golden_settings_gc_9_0[] = @@ -116,7 +149,9 @@ static const u32 golden_settings_gc_9_0[] = SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_UTCL1_CNTL_2), 0x08000000, 0x08000080, SOC15_REG_OFFSET(GC, 0, mmRLC_PREWALKER_UTCL1_CNTL), 0x08000000, 0x08000080, SOC15_REG_OFFSET(GC, 0, mmRLC_SPM_UTCL1_CNTL), 0x08000000, 0x08000080, + SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG), 0x00001000, 0x00001000, SOC15_REG_OFFSET(GC, 0, mmSPI_CONFIG_CNTL_1), 0x0000000f, 0x01000107, + SOC15_REG_OFFSET(GC, 0, mmSQC_CONFIG), 0x03000000, 0x020a2000, SOC15_REG_OFFSET(GC, 0, mmTA_CNTL_AUX), 0xfffffeef, 0x010b0000, SOC15_REG_OFFSET(GC, 0, mmTCP_CHAN_STEER_HI), 0xffffffff, 0x4a2c0e68, SOC15_REG_OFFSET(GC, 0, mmTCP_CHAN_STEER_LO), 0xffffffff, 0xb5d3f197, @@ -172,6 +207,12 @@ static const u32 golden_settings_gc_9_1_rv1[] = SOC15_REG_OFFSET(GC, 0, mmTD_CNTL), 0x01bd9f33, 0x00000800 }; +static const u32 golden_settings_gc_9_x_common[] = +{ + SOC15_REG_OFFSET(GC, 0, mmGRBM_CAM_INDEX), 0xffffffff, 0x00000000, + SOC15_REG_OFFSET(GC, 0, mmGRBM_CAM_DATA), 0xffffffff, 0x2544c382 +}; + #define VEGA10_GB_ADDR_CONFIG_GOLDEN 0x2a114042 #define RAVEN_GB_ADDR_CONFIG_GOLDEN 0x24000042 @@ -207,11 +248,14 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) default: break; } + + amdgpu_program_register_sequence(adev, golden_settings_gc_9_x_common, + (const u32)ARRAY_SIZE(golden_settings_gc_9_x_common)); } static void gfx_v9_0_scratch_init(struct amdgpu_device *adev) { - adev->gfx.scratch.num_reg = 7; + adev->gfx.scratch.num_reg = 8; adev->gfx.scratch.reg_base = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0); adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1; } @@ -350,6 +394,25 @@ err1: return r; } + +static void gfx_v9_0_free_microcode(struct amdgpu_device *adev) +{ + release_firmware(adev->gfx.pfp_fw); + adev->gfx.pfp_fw = NULL; + release_firmware(adev->gfx.me_fw); + adev->gfx.me_fw = NULL; + release_firmware(adev->gfx.ce_fw); + adev->gfx.ce_fw = NULL; + release_firmware(adev->gfx.rlc_fw); + adev->gfx.rlc_fw = NULL; + release_firmware(adev->gfx.mec_fw); + adev->gfx.mec_fw = NULL; + release_firmware(adev->gfx.mec2_fw); + adev->gfx.mec2_fw = NULL; + + kfree(adev->gfx.rlc.register_list_format); +} + static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) { const char *chip_name; @@ -772,18 +835,16 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) if (cs_data) { /* clear state block */ adev->gfx.rlc.clear_state_size = dws = gfx_v9_0_get_csb_size(adev); - if (adev->gfx.rlc.clear_state_obj == NULL) { - r = amdgpu_bo_create_kernel(adev, dws * 4, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.clear_state_obj, - &adev->gfx.rlc.clear_state_gpu_addr, - (void **)&adev->gfx.rlc.cs_ptr); - if (r) { - dev_err(adev->dev, - "(%d) failed to create rlc csb bo\n", r); - gfx_v9_0_rlc_fini(adev); - return r; - } + r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to create rlc csb bo\n", + r); + gfx_v9_0_rlc_fini(adev); + return r; } /* set up the cs buffer */ dst_ptr = adev->gfx.rlc.cs_ptr; @@ -795,18 +856,16 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) if (adev->asic_type == CHIP_RAVEN) { /* TODO: double check the cp_table_size for RV */ adev->gfx.rlc.cp_table_size = ALIGN(96 * 5 * 4, 2048) + (64 * 1024); /* JT + GDS */ - if (adev->gfx.rlc.cp_table_obj == NULL) { - r = amdgpu_bo_create_kernel(adev, adev->gfx.rlc.cp_table_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.cp_table_obj, - &adev->gfx.rlc.cp_table_gpu_addr, - (void **)&adev->gfx.rlc.cp_table_ptr); - if (r) { - dev_err(adev->dev, - "(%d) failed to create cp table bo\n", r); - gfx_v9_0_rlc_fini(adev); - return r; - } + r = amdgpu_bo_create_reserved(adev, adev->gfx.rlc.cp_table_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.cp_table_obj, + &adev->gfx.rlc.cp_table_gpu_addr, + (void **)&adev->gfx.rlc.cp_table_ptr); + if (r) { + dev_err(adev->dev, + "(%d) failed to create cp table bo\n", r); + gfx_v9_0_rlc_fini(adev); + return r; } rv_init_cp_jump_table(adev); @@ -821,28 +880,8 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) static void gfx_v9_0_mec_fini(struct amdgpu_device *adev) { - int r; - - if (adev->gfx.mec.hpd_eop_obj) { - r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve HPD EOP bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.mec.hpd_eop_obj); - amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj); - - amdgpu_bo_unref(&adev->gfx.mec.hpd_eop_obj); - adev->gfx.mec.hpd_eop_obj = NULL; - } - if (adev->gfx.mec.mec_fw_obj) { - r = amdgpu_bo_reserve(adev->gfx.mec.mec_fw_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve mec firmware bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.mec.mec_fw_obj); - amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj); - - amdgpu_bo_unref(&adev->gfx.mec.mec_fw_obj); - adev->gfx.mec.mec_fw_obj = NULL; - } + amdgpu_bo_free_kernel(&adev->gfx.mec.hpd_eop_obj, NULL, NULL); + amdgpu_bo_free_kernel(&adev->gfx.mec.mec_fw_obj, NULL, NULL); } static int gfx_v9_0_mec_init(struct amdgpu_device *adev) @@ -862,33 +901,13 @@ static int gfx_v9_0_mec_init(struct amdgpu_device *adev) amdgpu_gfx_compute_queue_acquire(adev); mec_hpd_size = adev->gfx.num_compute_rings * GFX9_MEC_HPD_SIZE; - if (adev->gfx.mec.hpd_eop_obj == NULL) { - r = amdgpu_bo_create(adev, - mec_hpd_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, - &adev->gfx.mec.hpd_eop_obj); - if (r) { - dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r); - return r; - } - } - - r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, false); - if (unlikely(r != 0)) { - gfx_v9_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.mec.hpd_eop_obj, AMDGPU_GEM_DOMAIN_GTT, - &adev->gfx.mec.hpd_eop_gpu_addr); - if (r) { - dev_warn(adev->dev, "(%d) pin HDP EOP bo failed\n", r); - gfx_v9_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_kmap(adev->gfx.mec.hpd_eop_obj, (void **)&hpd); + r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &adev->gfx.mec.hpd_eop_obj, + &adev->gfx.mec.hpd_eop_gpu_addr, + (void **)&hpd); if (r) { - dev_warn(adev->dev, "(%d) map HDP EOP bo failed\n", r); + dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r); gfx_v9_0_mec_fini(adev); return r; } @@ -905,42 +924,22 @@ static int gfx_v9_0_mec_init(struct amdgpu_device *adev) le32_to_cpu(mec_hdr->header.ucode_array_offset_bytes)); fw_size = le32_to_cpu(mec_hdr->header.ucode_size_bytes) / 4; - if (adev->gfx.mec.mec_fw_obj == NULL) { - r = amdgpu_bo_create(adev, - mec_hdr->header.ucode_size_bytes, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, - &adev->gfx.mec.mec_fw_obj); - if (r) { - dev_warn(adev->dev, "(%d) create mec firmware bo failed\n", r); - return r; - } - } - - r = amdgpu_bo_reserve(adev->gfx.mec.mec_fw_obj, false); - if (unlikely(r != 0)) { - gfx_v9_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.mec.mec_fw_obj, AMDGPU_GEM_DOMAIN_GTT, - &adev->gfx.mec.mec_fw_gpu_addr); + r = amdgpu_bo_create_reserved(adev, mec_hdr->header.ucode_size_bytes, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, + &adev->gfx.mec.mec_fw_obj, + &adev->gfx.mec.mec_fw_gpu_addr, + (void **)&fw); if (r) { - dev_warn(adev->dev, "(%d) pin mec firmware bo failed\n", r); - gfx_v9_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_kmap(adev->gfx.mec.mec_fw_obj, (void **)&fw); - if (r) { - dev_warn(adev->dev, "(%d) map firmware bo failed\n", r); + dev_warn(adev->dev, "(%d) create mec firmware bo failed\n", r); gfx_v9_0_mec_fini(adev); return r; } + memcpy(fw, fw_data, fw_size); amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_obj); amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj); - return 0; } @@ -998,12 +997,22 @@ static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd, start + SQIND_WAVE_SGPRS_OFFSET, size, dst); } +static void gfx_v9_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd, + uint32_t wave, uint32_t thread, + uint32_t start, uint32_t size, + uint32_t *dst) +{ + wave_read_regs( + adev, simd, wave, thread, + start + SQIND_WAVE_VGPRS_OFFSET, size, dst); +} static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = { .get_gpu_clock_counter = &gfx_v9_0_get_gpu_clock_counter, .select_se_sh = &gfx_v9_0_select_se_sh, .read_wave_data = &gfx_v9_0_read_wave_data, .read_wave_sgprs = &gfx_v9_0_read_wave_sgprs, + .read_wave_vgprs = &gfx_v9_0_read_wave_vgprs, }; static void gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) @@ -1182,30 +1191,22 @@ static int gfx_v9_0_ngg_en(struct amdgpu_device *adev) { struct amdgpu_ring *ring = &adev->gfx.gfx_ring[0]; int r; - u32 data; - u32 size; - u32 base; + u32 data, base; if (!amdgpu_ngg) return 0; /* Program buffer size */ - data = 0; - size = adev->gfx.ngg.buf[NGG_PRIM].size / 256; - data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, INDEX_BUF_SIZE, size); - - size = adev->gfx.ngg.buf[NGG_POS].size / 256; - data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, POS_BUF_SIZE, size); - + data = REG_SET_FIELD(0, WD_BUF_RESOURCE_1, INDEX_BUF_SIZE, + adev->gfx.ngg.buf[NGG_PRIM].size >> 8); + data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, POS_BUF_SIZE, + adev->gfx.ngg.buf[NGG_POS].size >> 8); WREG32_SOC15(GC, 0, mmWD_BUF_RESOURCE_1, data); - data = 0; - size = adev->gfx.ngg.buf[NGG_CNTL].size / 256; - data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, CNTL_SB_BUF_SIZE, size); - - size = adev->gfx.ngg.buf[NGG_PARAM].size / 1024; - data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, PARAM_BUF_SIZE, size); - + data = REG_SET_FIELD(0, WD_BUF_RESOURCE_2, CNTL_SB_BUF_SIZE, + adev->gfx.ngg.buf[NGG_CNTL].size >> 8); + data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, PARAM_BUF_SIZE, + adev->gfx.ngg.buf[NGG_PARAM].size >> 10); WREG32_SOC15(GC, 0, mmWD_BUF_RESOURCE_2, data); /* Program buffer base address */ @@ -1368,7 +1369,10 @@ static int gfx_v9_0_sw_init(void *handle) for (i = 0; i < adev->gfx.num_gfx_rings; i++) { ring = &adev->gfx.gfx_ring[i]; ring->ring_obj = NULL; - sprintf(ring->name, "gfx"); + if (!i) + sprintf(ring->name, "gfx"); + else + sprintf(ring->name, "gfx_%d", i); ring->use_doorbell = true; ring->doorbell_index = AMDGPU_DOORBELL64_GFX_RING0 << 1; r = amdgpu_ring_init(adev, ring, 1024, @@ -1408,7 +1412,7 @@ static int gfx_v9_0_sw_init(void *handle) return r; /* create MQD for all compute queues as wel as KIQ for SRIOV case */ - r = amdgpu_gfx_compute_mqd_sw_init(adev, sizeof(struct v9_mqd)); + r = amdgpu_gfx_compute_mqd_sw_init(adev, sizeof(struct v9_mqd_allocation)); if (r) return r; @@ -1460,9 +1464,19 @@ static int gfx_v9_0_sw_fini(void *handle) amdgpu_gfx_compute_mqd_sw_fini(adev); amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring, &adev->gfx.kiq.irq); amdgpu_gfx_kiq_fini(adev); + amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL); gfx_v9_0_mec_fini(adev); gfx_v9_0_ngg_fini(adev); + amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); + if (adev->asic_type == CHIP_RAVEN) { + amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, + &adev->gfx.rlc.cp_table_gpu_addr, + (void **)&adev->gfx.rlc.cp_table_ptr); + } + gfx_v9_0_free_microcode(adev); return 0; } @@ -1475,21 +1489,23 @@ static void gfx_v9_0_tiling_mode_table_init(struct amdgpu_device *adev) static void gfx_v9_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num, u32 instance) { - u32 data = REG_SET_FIELD(0, GRBM_GFX_INDEX, INSTANCE_BROADCAST_WRITES, 1); + u32 data; - if ((se_num == 0xffffffff) && (sh_num == 0xffffffff)) { - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_BROADCAST_WRITES, 1); - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_BROADCAST_WRITES, 1); - } else if (se_num == 0xffffffff) { - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_INDEX, sh_num); + if (instance == 0xffffffff) + data = REG_SET_FIELD(0, GRBM_GFX_INDEX, INSTANCE_BROADCAST_WRITES, 1); + else + data = REG_SET_FIELD(0, GRBM_GFX_INDEX, INSTANCE_INDEX, instance); + + if (se_num == 0xffffffff) data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_BROADCAST_WRITES, 1); - } else if (sh_num == 0xffffffff) { - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_BROADCAST_WRITES, 1); + else data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX, se_num); - } else { + + if (sh_num == 0xffffffff) + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_BROADCAST_WRITES, 1); + else data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_INDEX, sh_num); - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX, se_num); - } + WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, data); } @@ -1742,10 +1758,10 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev) adev->gfx.rlc.reg_list_format_size_bytes >> 2, unique_indirect_regs, &unique_indirect_reg_count, - sizeof(unique_indirect_regs)/sizeof(int), + ARRAY_SIZE(unique_indirect_regs), indirect_start_offsets, &indirect_start_offsets_count, - sizeof(indirect_start_offsets)/sizeof(int)); + ARRAY_SIZE(indirect_start_offsets)); /* enable auto inc in case it is disabled */ tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL)); @@ -1782,12 +1798,12 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev) /* write the starting offsets to RLC scratch ram */ WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_SCRATCH_ADDR), adev->gfx.rlc.starting_offsets_start); - for (i = 0; i < sizeof(indirect_start_offsets)/sizeof(int); i++) + for (i = 0; i < ARRAY_SIZE(indirect_start_offsets); i++) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_SCRATCH_DATA), indirect_start_offsets[i]); /* load unique indirect regs*/ - for (i = 0; i < sizeof(unique_indirect_regs)/sizeof(int); i++) { + for (i = 0; i < ARRAY_SIZE(unique_indirect_regs); i++) { WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_ADDR_0) + i, unique_indirect_regs[i] & 0x3FFFF); WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_DATA_0) + i, @@ -1800,11 +1816,7 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev) static void gfx_v9_0_enable_save_restore_machine(struct amdgpu_device *adev) { - u32 tmp = 0; - - tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL)); - tmp |= RLC_SRM_CNTL__SRM_ENABLE_MASK; - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL), tmp); + WREG32_FIELD15(GC, 0, RLC_SRM_CNTL, SRM_ENABLE, 1); } static void pwr_10_0_gfxip_control_over_cgpg(struct amdgpu_device *adev, @@ -1882,16 +1894,11 @@ static void gfx_v9_0_enable_sck_slow_down_on_power_up(struct amdgpu_device *adev uint32_t default_data = 0; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - - if (enable == true) { - data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK; - if (default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } else { - data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } + data = REG_SET_FIELD(data, RLC_PG_CNTL, + SMU_CLK_SLOWDOWN_ON_PU_ENABLE, + enable ? 1 : 0); + if (default_data != data) + WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } static void gfx_v9_0_enable_sck_slow_down_on_power_down(struct amdgpu_device *adev, @@ -1901,16 +1908,11 @@ static void gfx_v9_0_enable_sck_slow_down_on_power_down(struct amdgpu_device *ad uint32_t default_data = 0; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - - if (enable == true) { - data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } else { - data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } + data = REG_SET_FIELD(data, RLC_PG_CNTL, + SMU_CLK_SLOWDOWN_ON_PD_ENABLE, + enable ? 1 : 0); + if(default_data != data) + WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } static void gfx_v9_0_enable_cp_power_gating(struct amdgpu_device *adev, @@ -1920,16 +1922,11 @@ static void gfx_v9_0_enable_cp_power_gating(struct amdgpu_device *adev, uint32_t default_data = 0; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - - if (enable == true) { - data &= ~RLC_PG_CNTL__CP_PG_DISABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } else { - data |= RLC_PG_CNTL__CP_PG_DISABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } + data = REG_SET_FIELD(data, RLC_PG_CNTL, + CP_PG_DISABLE, + enable ? 0 : 1); + if(default_data != data) + WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } static void gfx_v9_0_enable_gfx_cg_power_gating(struct amdgpu_device *adev, @@ -1938,10 +1935,9 @@ static void gfx_v9_0_enable_gfx_cg_power_gating(struct amdgpu_device *adev, uint32_t data, default_data; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - if (enable == true) - data |= RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK; + data = REG_SET_FIELD(data, RLC_PG_CNTL, + GFX_POWER_GATING_ENABLE, + enable ? 1 : 0); if(default_data != data) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } @@ -1952,10 +1948,9 @@ static void gfx_v9_0_enable_gfx_pipeline_powergating(struct amdgpu_device *adev, uint32_t data, default_data; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - if (enable == true) - data |= RLC_PG_CNTL__GFX_PIPELINE_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__GFX_PIPELINE_PG_ENABLE_MASK; + data = REG_SET_FIELD(data, RLC_PG_CNTL, + GFX_PIPELINE_PG_ENABLE, + enable ? 1 : 0); if(default_data != data) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); @@ -1970,10 +1965,9 @@ static void gfx_v9_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *ade uint32_t data, default_data; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - if (enable == true) - data |= RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK; + data = REG_SET_FIELD(data, RLC_PG_CNTL, + STATIC_PER_CU_PG_ENABLE, + enable ? 1 : 0); if(default_data != data) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } @@ -1984,10 +1978,9 @@ static void gfx_v9_0_enable_gfx_dynamic_mg_power_gating(struct amdgpu_device *ad uint32_t data, default_data; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - if (enable == true) - data |= RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK; + data = REG_SET_FIELD(data, RLC_PG_CNTL, + DYN_PER_CU_PG_ENABLE, + enable ? 1 : 0); if(default_data != data) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } @@ -2027,13 +2020,8 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev) void gfx_v9_0_rlc_stop(struct amdgpu_device *adev) { - u32 tmp = RREG32_SOC15(GC, 0, mmRLC_CNTL); - - tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 0); - WREG32_SOC15(GC, 0, mmRLC_CNTL, tmp); - + WREG32_FIELD15(GC, 0, RLC_CNTL, RLC_ENABLE_F32, 0); gfx_v9_0_enable_gui_idle_interrupt(adev, false); - gfx_v9_0_wait_for_rlc_serdes(adev); } @@ -2105,8 +2093,10 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev) { int r; - if (amdgpu_sriov_vf(adev)) + if (amdgpu_sriov_vf(adev)) { + gfx_v9_0_init_csb(adev); return 0; + } gfx_v9_0_rlc_stop(adev); @@ -2217,7 +2207,7 @@ static int gfx_v9_0_cp_gfx_start(struct amdgpu_device *adev) struct amdgpu_ring *ring = &adev->gfx.gfx_ring[0]; const struct cs_section_def *sect = NULL; const struct cs_extent_def *ext = NULL; - int r, i; + int r, i, tmp; /* init the CP */ WREG32_SOC15(GC, 0, mmCP_MAX_CONTEXT, adev->gfx.config.max_hw_contexts - 1); @@ -2225,7 +2215,7 @@ static int gfx_v9_0_cp_gfx_start(struct amdgpu_device *adev) gfx_v9_0_cp_gfx_enable(adev, true); - r = amdgpu_ring_alloc(ring, gfx_v9_0_get_csb_size(adev) + 4); + r = amdgpu_ring_alloc(ring, gfx_v9_0_get_csb_size(adev) + 4 + 3); if (r) { DRM_ERROR("amdgpu: cp failed to lock ring (%d).\n", r); return r; @@ -2263,6 +2253,12 @@ static int gfx_v9_0_cp_gfx_start(struct amdgpu_device *adev) amdgpu_ring_write(ring, 0x8000); amdgpu_ring_write(ring, 0x8000); + amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG,1)); + tmp = (PACKET3_SET_UCONFIG_REG_INDEX_TYPE | + (SOC15_REG_OFFSET(GC, 0, mmVGT_INDEX_TYPE) - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(ring, tmp); + amdgpu_ring_write(ring, 0); + amdgpu_ring_commit(ring); return 0; @@ -2425,7 +2421,7 @@ static int gfx_v9_0_kiq_kcq_enable(struct amdgpu_device *adev) /* This situation may be hit in the future if a new HW * generation exposes more than 64 queues. If so, the * definition of queue_mask needs updating */ - if (WARN_ON(i > (sizeof(queue_mask)*8))) { + if (WARN_ON(i >= (sizeof(queue_mask)*8))) { DRM_ERROR("Invalid KCQ enabled: %d\n", i); break; } @@ -2517,6 +2513,13 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) mqd->compute_static_thread_mgmt_se3 = 0xffffffff; mqd->compute_misc_reserved = 0x00000003; + mqd->dynamic_cu_mask_addr_lo = + lower_32_bits(ring->mqd_gpu_addr + + offsetof(struct v9_mqd_allocation, dynamic_cu_mask)); + mqd->dynamic_cu_mask_addr_hi = + upper_32_bits(ring->mqd_gpu_addr + + offsetof(struct v9_mqd_allocation, dynamic_cu_mask)); + eop_base_addr = ring->eop_gpu_addr >> 8; mqd->cp_hqd_eop_base_addr_lo = eop_base_addr; mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); @@ -2540,10 +2543,10 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) DOORBELL_SOURCE, 0); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_HIT, 0); - } - else + } else { tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 0); + } mqd->cp_hqd_pq_doorbell_control = tmp; @@ -2746,10 +2749,10 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) gfx_v9_0_kiq_setting(ring); - if (adev->gfx.in_reset) { /* for GPU_RESET case */ + if (adev->in_sriov_reset) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd)); + memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation)); /* reset ring buffer */ ring->wptr = 0; @@ -2761,7 +2764,9 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) soc15_grbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); } else { - memset((void *)mqd, 0, sizeof(*mqd)); + memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation)); + ((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; + ((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; mutex_lock(&adev->srbm_mutex); soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v9_0_mqd_init(ring); @@ -2770,7 +2775,7 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) mutex_unlock(&adev->srbm_mutex); if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd)); + memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation)); } return 0; @@ -2782,8 +2787,10 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) struct v9_mqd *mqd = ring->mqd_ptr; int mqd_idx = ring - &adev->gfx.compute_ring[0]; - if (!adev->gfx.in_reset && !adev->gfx.in_suspend) { - memset((void *)mqd, 0, sizeof(*mqd)); + if (!adev->in_sriov_reset && !adev->gfx.in_suspend) { + memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation)); + ((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; + ((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; mutex_lock(&adev->srbm_mutex); soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v9_0_mqd_init(ring); @@ -2791,11 +2798,11 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) mutex_unlock(&adev->srbm_mutex); if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd)); - } else if (adev->gfx.in_reset) { /* for GPU_RESET case */ + memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation)); + } else if (adev->in_sriov_reset) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd)); + memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation)); /* reset ring buffer */ ring->wptr = 0; @@ -2936,12 +2943,70 @@ static int gfx_v9_0_hw_init(void *handle) return r; } +static int gfx_v9_0_kcq_disable(struct amdgpu_ring *kiq_ring,struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = kiq_ring->adev; + uint32_t scratch, tmp = 0; + int r, i; + + r = amdgpu_gfx_scratch_get(adev, &scratch); + if (r) { + DRM_ERROR("Failed to get scratch reg (%d).\n", r); + return r; + } + WREG32(scratch, 0xCAFEDEAD); + + r = amdgpu_ring_alloc(kiq_ring, 10); + if (r) { + DRM_ERROR("Failed to lock KIQ (%d).\n", r); + amdgpu_gfx_scratch_free(adev, scratch); + return r; + } + + /* unmap queues */ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); + amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + PACKET3_UNMAP_QUEUES_ACTION(1) | /* RESET_QUEUES */ + PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) | + PACKET3_UNMAP_QUEUES_ENGINE_SEL(0) | + PACKET3_UNMAP_QUEUES_NUM_QUEUES(1)); + amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + /* write to scratch for completion */ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); + amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(kiq_ring, 0xDEADBEEF); + amdgpu_ring_commit(kiq_ring); + + for (i = 0; i < adev->usec_timeout; i++) { + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + if (i >= adev->usec_timeout) { + DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp); + r = -EINVAL; + } + amdgpu_gfx_scratch_free(adev, scratch); + return r; +} + + static int gfx_v9_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i; amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); + + /* disable KCQ to avoid CPC touch memory not valid anymore */ + for (i = 0; i < adev->gfx.num_compute_rings; i++) + gfx_v9_0_kcq_disable(&adev->gfx.kiq.ring, &adev->gfx.compute_ring[i]); + if (amdgpu_sriov_vf(adev)) { pr_debug("For SRIOV client, shouldn't do anything.\n"); return 0; @@ -2984,15 +3049,10 @@ static bool gfx_v9_0_is_idle(void *handle) static int gfx_v9_0_wait_for_idle(void *handle) { unsigned i; - u32 tmp; struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->usec_timeout; i++) { - /* read MC_STATUS */ - tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS) & - GRBM_STATUS__GUI_ACTIVE_MASK; - - if (!REG_GET_FIELD(tmp, GRBM_STATUS, GUI_ACTIVE)) + if (gfx_v9_0_is_idle(handle)) return 0; udelay(1); } @@ -3551,9 +3611,11 @@ static void gfx_v9_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) { u32 ref_and_mask, reg_mem_engine; - struct nbio_hdp_flush_reg *nbio_hf_reg; + const struct nbio_hdp_flush_reg *nbio_hf_reg; - if (ring->adev->asic_type == CHIP_VEGA10) + if (ring->adev->flags & AMD_IS_APU) + nbio_hf_reg = &nbio_v7_0_hdp_flush_reg; + else nbio_hf_reg = &nbio_v6_1_hdp_flush_reg; if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) { @@ -3582,7 +3644,7 @@ static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) static void gfx_v9_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) { gfx_v9_0_write_data_to_reg(ring, 0, true, - SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0), 1); + SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1); } static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, @@ -3772,7 +3834,7 @@ static void gfx_v9_ring_emit_sb(struct amdgpu_ring *ring) static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring) { - static struct v9_ce_ib_state ce_payload = {0}; + struct v9_ce_ib_state ce_payload = {0}; uint64_t csa_addr; int cnt; @@ -3791,7 +3853,7 @@ static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring) static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring) { - static struct v9_de_ib_state de_payload = {0}; + struct v9_de_ib_state de_payload = {0}; uint64_t csa_addr, gds_addr; int cnt; @@ -3811,6 +3873,12 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring) amdgpu_ring_write_multiple(ring, (void *)&de_payload, sizeof(de_payload) >> 2); } +static void gfx_v9_0_ring_emit_tmz(struct amdgpu_ring *ring, bool start) +{ + amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0)); + amdgpu_ring_write(ring, FRAME_CMD(start ? 0 : 1)); /* frame_end */ +} + static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) { uint32_t dw2 = 0; @@ -3818,6 +3886,8 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) if (amdgpu_sriov_vf(ring->adev)) gfx_v9_0_ring_emit_ce_meta(ring); + gfx_v9_0_ring_emit_tmz(ring, true); + dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */ if (flags & AMDGPU_HAVE_CTX_SWITCH) { /* set load_global_config & load_global_uconfig */ @@ -3868,12 +3938,6 @@ static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size>>2) - offset + cur; } -static void gfx_v9_0_ring_emit_tmz(struct amdgpu_ring *ring, bool start) -{ - amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0)); - amdgpu_ring_write(ring, FRAME_CMD(start ? 0 : 1)); /* frame_end */ -} - static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) { struct amdgpu_device *adev = ring->adev; @@ -4156,7 +4220,7 @@ static int gfx_v9_0_kiq_irq(struct amdgpu_device *adev, return 0; } -const struct amd_ip_funcs gfx_v9_0_ip_funcs = { +static const struct amd_ip_funcs gfx_v9_0_ip_funcs = { .name = "gfx_v9_0", .early_init = gfx_v9_0_early_init, .late_init = gfx_v9_0_late_init, diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.h index 56ef652a575d..fa5a3fbaf6ab 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.h @@ -24,7 +24,6 @@ #ifndef __GFX_V9_0_H__ #define __GFX_V9_0_H__ -extern const struct amd_ip_funcs gfx_v9_0_ip_funcs; extern const struct amdgpu_ip_block_version gfx_v9_0_ip_block; void gfx_v9_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num); diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c index a42f483767e7..c17996e18086 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c @@ -58,14 +58,14 @@ static void gfxhub_v1_0_init_gart_aperture_regs(struct amdgpu_device *adev) gfxhub_v1_0_init_gart_pt_regs(adev); WREG32_SOC15(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32, - (u32)(adev->mc.gtt_start >> 12)); + (u32)(adev->mc.gart_start >> 12)); WREG32_SOC15(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32, - (u32)(adev->mc.gtt_start >> 44)); + (u32)(adev->mc.gart_start >> 44)); WREG32_SOC15(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32, - (u32)(adev->mc.gtt_end >> 12)); + (u32)(adev->mc.gart_end >> 12)); WREG32_SOC15(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32, - (u32)(adev->mc.gtt_end >> 44)); + (u32)(adev->mc.gart_end >> 44)); } static void gfxhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev) @@ -129,7 +129,7 @@ static void gfxhub_v1_0_init_cache_regs(struct amdgpu_device *adev) /* Setup L2 cache */ tmp = RREG32_SOC15(GC, 0, mmVM_L2_CNTL); tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1); - tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 1); /* XXX for emulation, Refer to closed source code.*/ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE, 0); @@ -144,6 +144,8 @@ static void gfxhub_v1_0_init_cache_regs(struct amdgpu_device *adev) WREG32_SOC15(GC, 0, mmVM_L2_CNTL2, tmp); tmp = mmVM_L2_CNTL3_DEFAULT; + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 9); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_FRAGMENT_SIZE, 6); WREG32_SOC15(GC, 0, mmVM_L2_CNTL3, tmp); tmp = mmVM_L2_CNTL4_DEFAULT; @@ -206,6 +208,9 @@ static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_BLOCK_SIZE, adev->vm_manager.block_size - 9); + /* Send no-retry XNACK on fault to suppress VM fault storm. */ + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0); WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_CNTL, i, tmp); WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, i*2, 0); WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, i*2, 0); @@ -314,6 +319,12 @@ void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value); tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + if (!value) { + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_NO_RETRY_FAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_RETRY_FAULT, 1); + } WREG32_SOC15(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL, tmp); } diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h index d2dbb085f480..206e29cad753 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h @@ -30,7 +30,5 @@ void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, bool value); void gfxhub_v1_0_init(struct amdgpu_device *adev); u64 gfxhub_v1_0_get_mc_fb_offset(struct amdgpu_device *adev); -extern const struct amd_ip_funcs gfxhub_v1_0_ip_funcs; -extern const struct amdgpu_ip_block_version gfxhub_v1_0_ip_block; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c index d0214d942bfc..f4603a7c8ef3 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c @@ -66,14 +66,10 @@ static const u32 crtc_offsets[6] = SI_CRTC5_REGISTER_OFFSET }; -static void gmc_v6_0_mc_stop(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) +static void gmc_v6_0_mc_stop(struct amdgpu_device *adev) { u32 blackout; - if (adev->mode_info.num_crtc) - amdgpu_display_stop_mc_access(adev, save); - gmc_v6_0_wait_for_idle((void *)adev); blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL); @@ -90,8 +86,7 @@ static void gmc_v6_0_mc_stop(struct amdgpu_device *adev, } -static void gmc_v6_0_mc_resume(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) +static void gmc_v6_0_mc_resume(struct amdgpu_device *adev) { u32 tmp; @@ -103,10 +98,6 @@ static void gmc_v6_0_mc_resume(struct amdgpu_device *adev, tmp = REG_SET_FIELD(0, BIF_FB_EN, FB_READ_EN, 1); tmp = REG_SET_FIELD(tmp, BIF_FB_EN, FB_WRITE_EN, 1); WREG32(mmBIF_FB_EN, tmp); - - if (adev->mode_info.num_crtc) - amdgpu_display_resume_mc_access(adev, save); - } static int gmc_v6_0_init_microcode(struct amdgpu_device *adev) @@ -228,20 +219,20 @@ static int gmc_v6_0_mc_load_microcode(struct amdgpu_device *adev) static void gmc_v6_0_vram_gtt_location(struct amdgpu_device *adev, struct amdgpu_mc *mc) { + u64 base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF; + base <<= 24; + if (mc->mc_vram_size > 0xFFC0000000ULL) { dev_warn(adev->dev, "limiting VRAM\n"); mc->real_vram_size = 0xFFC0000000ULL; mc->mc_vram_size = 0xFFC0000000ULL; } - amdgpu_vram_location(adev, &adev->mc, 0); - adev->mc.gtt_base_align = 0; - amdgpu_gtt_location(adev, mc); + amdgpu_vram_location(adev, &adev->mc, base); + amdgpu_gart_location(adev, mc); } static void gmc_v6_0_mc_program(struct amdgpu_device *adev) { - struct amdgpu_mode_mc_save save; - u32 tmp; int i, j; /* Initialize HDP */ @@ -254,16 +245,23 @@ static void gmc_v6_0_mc_program(struct amdgpu_device *adev) } WREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL, 0); - if (adev->mode_info.num_crtc) - amdgpu_display_set_vga_render_state(adev, false); - - gmc_v6_0_mc_stop(adev, &save); - if (gmc_v6_0_wait_for_idle((void *)adev)) { dev_warn(adev->dev, "Wait for MC idle timedout !\n"); } - WREG32(mmVGA_HDP_CONTROL, VGA_HDP_CONTROL__VGA_MEMORY_DISABLE_MASK); + if (adev->mode_info.num_crtc) { + u32 tmp; + + /* Lockout access through VGA aperture*/ + tmp = RREG32(mmVGA_HDP_CONTROL); + tmp |= VGA_HDP_CONTROL__VGA_MEMORY_DISABLE_MASK; + WREG32(mmVGA_HDP_CONTROL, tmp); + + /* disable VGA render */ + tmp = RREG32(mmVGA_RENDER_CONTROL); + tmp &= ~VGA_VSTATUS_CNTL; + WREG32(mmVGA_RENDER_CONTROL, tmp); + } /* Update configuration */ WREG32(mmMC_VM_SYSTEM_APERTURE_LOW_ADDR, adev->mc.vram_start >> 12); @@ -271,13 +269,6 @@ static void gmc_v6_0_mc_program(struct amdgpu_device *adev) adev->mc.vram_end >> 12); WREG32(mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR, adev->vram_scratch.gpu_addr >> 12); - tmp = ((adev->mc.vram_end >> 24) & 0xFFFF) << 16; - tmp |= ((adev->mc.vram_start >> 24) & 0xFFFF); - WREG32(mmMC_VM_FB_LOCATION, tmp); - /* XXX double check these! */ - WREG32(mmHDP_NONSURFACE_BASE, (adev->mc.vram_start >> 8)); - WREG32(mmHDP_NONSURFACE_INFO, (2 << 7) | (1 << 30)); - WREG32(mmHDP_NONSURFACE_SIZE, 0x3FFFFFFF); WREG32(mmMC_VM_AGP_BASE, 0); WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF); WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF); @@ -285,7 +276,6 @@ static void gmc_v6_0_mc_program(struct amdgpu_device *adev) if (gmc_v6_0_wait_for_idle((void *)adev)) { dev_warn(adev->dev, "Wait for MC idle timedout !\n"); } - gmc_v6_0_mc_resume(adev, &save); } static int gmc_v6_0_mc_init(struct amdgpu_device *adev) @@ -342,14 +332,23 @@ static int gmc_v6_0_mc_init(struct amdgpu_device *adev) adev->mc.real_vram_size = RREG32(mmCONFIG_MEMSIZE) * 1024ULL * 1024ULL; adev->mc.visible_vram_size = adev->mc.aper_size; - /* unless the user had overridden it, set the gart - * size equal to the 1024 or vram, whichever is larger. - */ - if (amdgpu_gart_size == -1) - adev->mc.gtt_size = max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20), - adev->mc.mc_vram_size); - else - adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20; + /* set the gart size */ + if (amdgpu_gart_size == -1) { + switch (adev->asic_type) { + case CHIP_HAINAN: /* no MM engines */ + default: + adev->mc.gart_size = 256ULL << 20; + break; + case CHIP_VERDE: /* UVD, VCE do not support GPUVM */ + case CHIP_TAHITI: /* UVD, VCE do not support GPUVM */ + case CHIP_PITCAIRN: /* UVD, VCE do not support GPUVM */ + case CHIP_OLAND: /* UVD, VCE do not support GPUVM */ + adev->mc.gart_size = 1024ULL << 20; + break; + } + } else { + adev->mc.gart_size = (u64)amdgpu_gart_size << 20; + } gmc_v6_0_vram_gtt_location(adev, &adev->mc); @@ -479,6 +478,7 @@ static void gmc_v6_0_set_prt(struct amdgpu_device *adev, bool enable) static int gmc_v6_0_gart_enable(struct amdgpu_device *adev) { int r, i; + u32 field; if (adev->gart.robj == NULL) { dev_err(adev->dev, "No VRAM object for PCIE GART.\n"); @@ -506,13 +506,15 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev) WREG32(mmVM_L2_CNTL2, VM_L2_CNTL2__INVALIDATE_ALL_L1_TLBS_MASK | VM_L2_CNTL2__INVALIDATE_L2_CACHE_MASK); + + field = adev->vm_manager.fragment_size; WREG32(mmVM_L2_CNTL3, VM_L2_CNTL3__L2_CACHE_BIGK_ASSOCIATIVITY_MASK | - (4UL << VM_L2_CNTL3__BANK_SELECT__SHIFT) | - (4UL << VM_L2_CNTL3__L2_CACHE_BIGK_FRAGMENT_SIZE__SHIFT)); + (field << VM_L2_CNTL3__BANK_SELECT__SHIFT) | + (field << VM_L2_CNTL3__L2_CACHE_BIGK_FRAGMENT_SIZE__SHIFT)); /* setup context0 */ - WREG32(mmVM_CONTEXT0_PAGE_TABLE_START_ADDR, adev->mc.gtt_start >> 12); - WREG32(mmVM_CONTEXT0_PAGE_TABLE_END_ADDR, adev->mc.gtt_end >> 12); + WREG32(mmVM_CONTEXT0_PAGE_TABLE_START_ADDR, adev->mc.gart_start >> 12); + WREG32(mmVM_CONTEXT0_PAGE_TABLE_END_ADDR, adev->mc.gart_end >> 12); WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR, adev->gart.table_addr >> 12); WREG32(mmVM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR, (u32)(adev->dummy_page.addr >> 12)); @@ -559,7 +561,7 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev) gmc_v6_0_gart_flush_gpu_tlb(adev, 0); dev_info(adev->dev, "PCIE GART of %uM enabled (table at 0x%016llX).\n", - (unsigned)(adev->mc.gtt_size >> 20), + (unsigned)(adev->mc.gart_size >> 20), (unsigned long long)adev->gart.table_addr); adev->gart.ready = true; return 0; @@ -829,7 +831,7 @@ static int gmc_v6_0_sw_init(void *handle) if (r) return r; - amdgpu_vm_adjust_size(adev, 64); + amdgpu_vm_adjust_size(adev, 64, 9); adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18; adev->mc.mc_mask = 0xffffffffffULL; @@ -899,6 +901,8 @@ static int gmc_v6_0_sw_fini(void *handle) gmc_v6_0_gart_fini(adev); amdgpu_gem_force_release(adev); amdgpu_bo_fini(adev); + release_firmware(adev->mc.fw); + adev->mc.fw = NULL; return 0; } @@ -987,7 +991,6 @@ static int gmc_v6_0_wait_for_idle(void *handle) static int gmc_v6_0_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct amdgpu_mode_mc_save save; u32 srbm_soft_reset = 0; u32 tmp = RREG32(mmSRBM_STATUS); @@ -1003,7 +1006,7 @@ static int gmc_v6_0_soft_reset(void *handle) } if (srbm_soft_reset) { - gmc_v6_0_mc_stop(adev, &save); + gmc_v6_0_mc_stop(adev); if (gmc_v6_0_wait_for_idle(adev)) { dev_warn(adev->dev, "Wait for GMC idle timed out !\n"); } @@ -1023,7 +1026,7 @@ static int gmc_v6_0_soft_reset(void *handle) udelay(50); - gmc_v6_0_mc_resume(adev, &save); + gmc_v6_0_mc_resume(adev); udelay(50); } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index 7e9ea53edf8b..b0528ca9207b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -37,6 +37,9 @@ #include "oss/oss_2_0_d.h" #include "oss/oss_2_0_sh_mask.h" +#include "dce/dce_8_0_d.h" +#include "dce/dce_8_0_sh_mask.h" + #include "amdgpu_atombios.h" static void gmc_v7_0_set_gart_funcs(struct amdgpu_device *adev); @@ -76,14 +79,10 @@ static void gmc_v7_0_init_golden_registers(struct amdgpu_device *adev) } } -static void gmc_v7_0_mc_stop(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) +static void gmc_v7_0_mc_stop(struct amdgpu_device *adev) { u32 blackout; - if (adev->mode_info.num_crtc) - amdgpu_display_stop_mc_access(adev, save); - gmc_v7_0_wait_for_idle((void *)adev); blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL); @@ -99,8 +98,7 @@ static void gmc_v7_0_mc_stop(struct amdgpu_device *adev, udelay(100); } -static void gmc_v7_0_mc_resume(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) +static void gmc_v7_0_mc_resume(struct amdgpu_device *adev) { u32 tmp; @@ -112,9 +110,6 @@ static void gmc_v7_0_mc_resume(struct amdgpu_device *adev, tmp = REG_SET_FIELD(0, BIF_FB_EN, FB_READ_EN, 1); tmp = REG_SET_FIELD(tmp, BIF_FB_EN, FB_WRITE_EN, 1); WREG32(mmBIF_FB_EN, tmp); - - if (adev->mode_info.num_crtc) - amdgpu_display_resume_mc_access(adev, save); } /** @@ -242,15 +237,17 @@ static int gmc_v7_0_mc_load_microcode(struct amdgpu_device *adev) static void gmc_v7_0_vram_gtt_location(struct amdgpu_device *adev, struct amdgpu_mc *mc) { + u64 base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF; + base <<= 24; + if (mc->mc_vram_size > 0xFFC0000000ULL) { /* leave room for at least 1024M GTT */ dev_warn(adev->dev, "limiting VRAM\n"); mc->real_vram_size = 0xFFC0000000ULL; mc->mc_vram_size = 0xFFC0000000ULL; } - amdgpu_vram_location(adev, &adev->mc, 0); - adev->mc.gtt_base_align = 0; - amdgpu_gtt_location(adev, mc); + amdgpu_vram_location(adev, &adev->mc, base); + amdgpu_gart_location(adev, mc); } /** @@ -263,7 +260,6 @@ static void gmc_v7_0_vram_gtt_location(struct amdgpu_device *adev, */ static void gmc_v7_0_mc_program(struct amdgpu_device *adev) { - struct amdgpu_mode_mc_save save; u32 tmp; int i, j; @@ -277,13 +273,20 @@ static void gmc_v7_0_mc_program(struct amdgpu_device *adev) } WREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL, 0); - if (adev->mode_info.num_crtc) - amdgpu_display_set_vga_render_state(adev, false); - - gmc_v7_0_mc_stop(adev, &save); if (gmc_v7_0_wait_for_idle((void *)adev)) { dev_warn(adev->dev, "Wait for MC idle timedout !\n"); } + if (adev->mode_info.num_crtc) { + /* Lockout access through VGA aperture*/ + tmp = RREG32(mmVGA_HDP_CONTROL); + tmp = REG_SET_FIELD(tmp, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1); + WREG32(mmVGA_HDP_CONTROL, tmp); + + /* disable VGA render */ + tmp = RREG32(mmVGA_RENDER_CONTROL); + tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0); + WREG32(mmVGA_RENDER_CONTROL, tmp); + } /* Update configuration */ WREG32(mmMC_VM_SYSTEM_APERTURE_LOW_ADDR, adev->mc.vram_start >> 12); @@ -291,20 +294,12 @@ static void gmc_v7_0_mc_program(struct amdgpu_device *adev) adev->mc.vram_end >> 12); WREG32(mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR, adev->vram_scratch.gpu_addr >> 12); - tmp = ((adev->mc.vram_end >> 24) & 0xFFFF) << 16; - tmp |= ((adev->mc.vram_start >> 24) & 0xFFFF); - WREG32(mmMC_VM_FB_LOCATION, tmp); - /* XXX double check these! */ - WREG32(mmHDP_NONSURFACE_BASE, (adev->mc.vram_start >> 8)); - WREG32(mmHDP_NONSURFACE_INFO, (2 << 7) | (1 << 30)); - WREG32(mmHDP_NONSURFACE_SIZE, 0x3FFFFFFF); WREG32(mmMC_VM_AGP_BASE, 0); WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF); WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF); if (gmc_v7_0_wait_for_idle((void *)adev)) { dev_warn(adev->dev, "Wait for MC idle timedout !\n"); } - gmc_v7_0_mc_resume(adev, &save); WREG32(mmBIF_FB_EN, BIF_FB_EN__FB_READ_EN_MASK | BIF_FB_EN__FB_WRITE_EN_MASK); @@ -391,14 +386,26 @@ static int gmc_v7_0_mc_init(struct amdgpu_device *adev) if (adev->mc.visible_vram_size > adev->mc.real_vram_size) adev->mc.visible_vram_size = adev->mc.real_vram_size; - /* unless the user had overridden it, set the gart - * size equal to the 1024 or vram, whichever is larger. - */ - if (amdgpu_gart_size == -1) - adev->mc.gtt_size = max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20), - adev->mc.mc_vram_size); - else - adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20; + /* set the gart size */ + if (amdgpu_gart_size == -1) { + switch (adev->asic_type) { + case CHIP_TOPAZ: /* no MM engines */ + default: + adev->mc.gart_size = 256ULL << 20; + break; +#ifdef CONFIG_DRM_AMDGPU_CIK + case CHIP_BONAIRE: /* UVD, VCE do not support GPUVM */ + case CHIP_HAWAII: /* UVD, VCE do not support GPUVM */ + case CHIP_KAVERI: /* UVD, VCE do not support GPUVM */ + case CHIP_KABINI: /* UVD, VCE do not support GPUVM */ + case CHIP_MULLINS: /* UVD, VCE do not support GPUVM */ + adev->mc.gart_size = 1024ULL << 20; + break; +#endif + } + } else { + adev->mc.gart_size = (u64)amdgpu_gart_size << 20; + } gmc_v7_0_vram_gtt_location(adev, &adev->mc); @@ -575,7 +582,7 @@ static void gmc_v7_0_set_prt(struct amdgpu_device *adev, bool enable) static int gmc_v7_0_gart_enable(struct amdgpu_device *adev) { int r, i; - u32 tmp; + u32 tmp, field; if (adev->gart.robj == NULL) { dev_err(adev->dev, "No VRAM object for PCIE GART.\n"); @@ -605,14 +612,16 @@ static int gmc_v7_0_gart_enable(struct amdgpu_device *adev) tmp = REG_SET_FIELD(0, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1); tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1); WREG32(mmVM_L2_CNTL2, tmp); + + field = adev->vm_manager.fragment_size; tmp = RREG32(mmVM_L2_CNTL3); tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_ASSOCIATIVITY, 1); - tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 4); - tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_FRAGMENT_SIZE, 4); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, field); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_FRAGMENT_SIZE, field); WREG32(mmVM_L2_CNTL3, tmp); /* setup context0 */ - WREG32(mmVM_CONTEXT0_PAGE_TABLE_START_ADDR, adev->mc.gtt_start >> 12); - WREG32(mmVM_CONTEXT0_PAGE_TABLE_END_ADDR, adev->mc.gtt_end >> 12); + WREG32(mmVM_CONTEXT0_PAGE_TABLE_START_ADDR, adev->mc.gart_start >> 12); + WREG32(mmVM_CONTEXT0_PAGE_TABLE_END_ADDR, adev->mc.gart_end >> 12); WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR, adev->gart.table_addr >> 12); WREG32(mmVM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR, (u32)(adev->dummy_page.addr >> 12)); @@ -666,7 +675,7 @@ static int gmc_v7_0_gart_enable(struct amdgpu_device *adev) gmc_v7_0_gart_flush_gpu_tlb(adev, 0); DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", - (unsigned)(adev->mc.gtt_size >> 20), + (unsigned)(adev->mc.gart_size >> 20), (unsigned long long)adev->gart.table_addr); adev->gart.ready = true; return 0; @@ -961,7 +970,7 @@ static int gmc_v7_0_sw_init(void *handle) * Currently set to 4GB ((1 << 20) 4k pages). * Max GPUVM size for cayman and SI is 40 bits. */ - amdgpu_vm_adjust_size(adev, 64); + amdgpu_vm_adjust_size(adev, 64, 9); adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18; /* Set the internal MC address mask @@ -1041,6 +1050,8 @@ static int gmc_v7_0_sw_fini(void *handle) gmc_v7_0_gart_fini(adev); amdgpu_gem_force_release(adev); amdgpu_bo_fini(adev); + release_firmware(adev->mc.fw); + adev->mc.fw = NULL; return 0; } @@ -1138,7 +1149,6 @@ static int gmc_v7_0_wait_for_idle(void *handle) static int gmc_v7_0_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct amdgpu_mode_mc_save save; u32 srbm_soft_reset = 0; u32 tmp = RREG32(mmSRBM_STATUS); @@ -1154,7 +1164,7 @@ static int gmc_v7_0_soft_reset(void *handle) } if (srbm_soft_reset) { - gmc_v7_0_mc_stop(adev, &save); + gmc_v7_0_mc_stop(adev); if (gmc_v7_0_wait_for_idle((void *)adev)) { dev_warn(adev->dev, "Wait for GMC idle timed out !\n"); } @@ -1175,7 +1185,7 @@ static int gmc_v7_0_soft_reset(void *handle) /* Wait a little for things to settle down */ udelay(50); - gmc_v7_0_mc_resume(adev, &save); + gmc_v7_0_mc_resume(adev); udelay(50); } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index cc9f88057cd5..f368cfe2f585 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -35,6 +35,9 @@ #include "oss/oss_3_0_d.h" #include "oss/oss_3_0_sh_mask.h" +#include "dce/dce_10_0_d.h" +#include "dce/dce_10_0_sh_mask.h" + #include "vid.h" #include "vi.h" @@ -161,14 +164,10 @@ static void gmc_v8_0_init_golden_registers(struct amdgpu_device *adev) } } -static void gmc_v8_0_mc_stop(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) +static void gmc_v8_0_mc_stop(struct amdgpu_device *adev) { u32 blackout; - if (adev->mode_info.num_crtc) - amdgpu_display_stop_mc_access(adev, save); - gmc_v8_0_wait_for_idle(adev); blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL); @@ -184,8 +183,7 @@ static void gmc_v8_0_mc_stop(struct amdgpu_device *adev, udelay(100); } -static void gmc_v8_0_mc_resume(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save) +static void gmc_v8_0_mc_resume(struct amdgpu_device *adev) { u32 tmp; @@ -197,9 +195,6 @@ static void gmc_v8_0_mc_resume(struct amdgpu_device *adev, tmp = REG_SET_FIELD(0, BIF_FB_EN, FB_READ_EN, 1); tmp = REG_SET_FIELD(tmp, BIF_FB_EN, FB_WRITE_EN, 1); WREG32(mmBIF_FB_EN, tmp); - - if (adev->mode_info.num_crtc) - amdgpu_display_resume_mc_access(adev, save); } /** @@ -404,15 +399,20 @@ static int gmc_v8_0_polaris_mc_load_microcode(struct amdgpu_device *adev) static void gmc_v8_0_vram_gtt_location(struct amdgpu_device *adev, struct amdgpu_mc *mc) { + u64 base = 0; + + if (!amdgpu_sriov_vf(adev)) + base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF; + base <<= 24; + if (mc->mc_vram_size > 0xFFC0000000ULL) { /* leave room for at least 1024M GTT */ dev_warn(adev->dev, "limiting VRAM\n"); mc->real_vram_size = 0xFFC0000000ULL; mc->mc_vram_size = 0xFFC0000000ULL; } - amdgpu_vram_location(adev, &adev->mc, 0); - adev->mc.gtt_base_align = 0; - amdgpu_gtt_location(adev, mc); + amdgpu_vram_location(adev, &adev->mc, base); + amdgpu_gart_location(adev, mc); } /** @@ -425,7 +425,6 @@ static void gmc_v8_0_vram_gtt_location(struct amdgpu_device *adev, */ static void gmc_v8_0_mc_program(struct amdgpu_device *adev) { - struct amdgpu_mode_mc_save save; u32 tmp; int i, j; @@ -439,13 +438,20 @@ static void gmc_v8_0_mc_program(struct amdgpu_device *adev) } WREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL, 0); - if (adev->mode_info.num_crtc) - amdgpu_display_set_vga_render_state(adev, false); - - gmc_v8_0_mc_stop(adev, &save); if (gmc_v8_0_wait_for_idle((void *)adev)) { dev_warn(adev->dev, "Wait for MC idle timedout !\n"); } + if (adev->mode_info.num_crtc) { + /* Lockout access through VGA aperture*/ + tmp = RREG32(mmVGA_HDP_CONTROL); + tmp = REG_SET_FIELD(tmp, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1); + WREG32(mmVGA_HDP_CONTROL, tmp); + + /* disable VGA render */ + tmp = RREG32(mmVGA_RENDER_CONTROL); + tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0); + WREG32(mmVGA_RENDER_CONTROL, tmp); + } /* Update configuration */ WREG32(mmMC_VM_SYSTEM_APERTURE_LOW_ADDR, adev->mc.vram_start >> 12); @@ -453,20 +459,23 @@ static void gmc_v8_0_mc_program(struct amdgpu_device *adev) adev->mc.vram_end >> 12); WREG32(mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR, adev->vram_scratch.gpu_addr >> 12); - tmp = ((adev->mc.vram_end >> 24) & 0xFFFF) << 16; - tmp |= ((adev->mc.vram_start >> 24) & 0xFFFF); - WREG32(mmMC_VM_FB_LOCATION, tmp); - /* XXX double check these! */ - WREG32(mmHDP_NONSURFACE_BASE, (adev->mc.vram_start >> 8)); - WREG32(mmHDP_NONSURFACE_INFO, (2 << 7) | (1 << 30)); - WREG32(mmHDP_NONSURFACE_SIZE, 0x3FFFFFFF); + + if (amdgpu_sriov_vf(adev)) { + tmp = ((adev->mc.vram_end >> 24) & 0xFFFF) << 16; + tmp |= ((adev->mc.vram_start >> 24) & 0xFFFF); + WREG32(mmMC_VM_FB_LOCATION, tmp); + /* XXX double check these! */ + WREG32(mmHDP_NONSURFACE_BASE, (adev->mc.vram_start >> 8)); + WREG32(mmHDP_NONSURFACE_INFO, (2 << 7) | (1 << 30)); + WREG32(mmHDP_NONSURFACE_SIZE, 0x3FFFFFFF); + } + WREG32(mmMC_VM_AGP_BASE, 0); WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF); WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF); if (gmc_v8_0_wait_for_idle((void *)adev)) { dev_warn(adev->dev, "Wait for MC idle timedout !\n"); } - gmc_v8_0_mc_resume(adev, &save); WREG32(mmBIF_FB_EN, BIF_FB_EN__FB_READ_EN_MASK | BIF_FB_EN__FB_WRITE_EN_MASK); @@ -553,14 +562,25 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev) if (adev->mc.visible_vram_size > adev->mc.real_vram_size) adev->mc.visible_vram_size = adev->mc.real_vram_size; - /* unless the user had overridden it, set the gart - * size equal to the 1024 or vram, whichever is larger. - */ - if (amdgpu_gart_size == -1) - adev->mc.gtt_size = max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20), - adev->mc.mc_vram_size); - else - adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20; + /* set the gart size */ + if (amdgpu_gart_size == -1) { + switch (adev->asic_type) { + case CHIP_POLARIS11: /* all engines support GPUVM */ + case CHIP_POLARIS10: /* all engines support GPUVM */ + case CHIP_POLARIS12: /* all engines support GPUVM */ + default: + adev->mc.gart_size = 256ULL << 20; + break; + case CHIP_TONGA: /* UVD, VCE do not support GPUVM */ + case CHIP_FIJI: /* UVD, VCE do not support GPUVM */ + case CHIP_CARRIZO: /* UVD, VCE do not support GPUVM, DCE SG support */ + case CHIP_STONEY: /* UVD does not support GPUVM, DCE SG support */ + adev->mc.gart_size = 1024ULL << 20; + break; + } + } else { + adev->mc.gart_size = (u64)amdgpu_gart_size << 20; + } gmc_v8_0_vram_gtt_location(adev, &adev->mc); @@ -761,7 +781,7 @@ static void gmc_v8_0_set_prt(struct amdgpu_device *adev, bool enable) static int gmc_v8_0_gart_enable(struct amdgpu_device *adev) { int r, i; - u32 tmp; + u32 tmp, field; if (adev->gart.robj == NULL) { dev_err(adev->dev, "No VRAM object for PCIE GART.\n"); @@ -792,10 +812,12 @@ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1); tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1); WREG32(mmVM_L2_CNTL2, tmp); + + field = adev->vm_manager.fragment_size; tmp = RREG32(mmVM_L2_CNTL3); tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_ASSOCIATIVITY, 1); - tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 4); - tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_FRAGMENT_SIZE, 4); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, field); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_FRAGMENT_SIZE, field); WREG32(mmVM_L2_CNTL3, tmp); /* XXX: set to enable PTE/PDE in system memory */ tmp = RREG32(mmVM_L2_CNTL4); @@ -813,8 +835,8 @@ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_CONTEXT1_PTE_REQUEST_SNOOP, 0); WREG32(mmVM_L2_CNTL4, tmp); /* setup context0 */ - WREG32(mmVM_CONTEXT0_PAGE_TABLE_START_ADDR, adev->mc.gtt_start >> 12); - WREG32(mmVM_CONTEXT0_PAGE_TABLE_END_ADDR, adev->mc.gtt_end >> 12); + WREG32(mmVM_CONTEXT0_PAGE_TABLE_START_ADDR, adev->mc.gart_start >> 12); + WREG32(mmVM_CONTEXT0_PAGE_TABLE_END_ADDR, adev->mc.gart_end >> 12); WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR, adev->gart.table_addr >> 12); WREG32(mmVM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR, (u32)(adev->dummy_page.addr >> 12)); @@ -869,7 +891,7 @@ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev) gmc_v8_0_gart_flush_gpu_tlb(adev, 0); DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", - (unsigned)(adev->mc.gtt_size >> 20), + (unsigned)(adev->mc.gart_size >> 20), (unsigned long long)adev->gart.table_addr); adev->gart.ready = true; return 0; @@ -1045,7 +1067,7 @@ static int gmc_v8_0_sw_init(void *handle) * Currently set to 4GB ((1 << 20) 4k pages). * Max GPUVM size for cayman and SI is 40 bits. */ - amdgpu_vm_adjust_size(adev, 64); + amdgpu_vm_adjust_size(adev, 64, 9); adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18; /* Set the internal MC address mask @@ -1125,6 +1147,8 @@ static int gmc_v8_0_sw_fini(void *handle) gmc_v8_0_gart_fini(adev); amdgpu_gem_force_release(adev); amdgpu_bo_fini(adev); + release_firmware(adev->mc.fw); + adev->mc.fw = NULL; return 0; } @@ -1260,7 +1284,7 @@ static int gmc_v8_0_pre_soft_reset(void *handle) if (!adev->mc.srbm_soft_reset) return 0; - gmc_v8_0_mc_stop(adev, &adev->mc.save); + gmc_v8_0_mc_stop(adev); if (gmc_v8_0_wait_for_idle(adev)) { dev_warn(adev->dev, "Wait for GMC idle timed out !\n"); } @@ -1306,7 +1330,7 @@ static int gmc_v8_0_post_soft_reset(void *handle) if (!adev->mc.srbm_soft_reset) return 0; - gmc_v8_0_mc_resume(adev, &adev->mc.save); + gmc_v8_0_mc_resume(adev); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 175ba5f9691c..c8f1aebeac7a 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -23,12 +23,17 @@ #include <linux/firmware.h> #include "amdgpu.h" #include "gmc_v9_0.h" +#include "amdgpu_atomfirmware.h" #include "vega10/soc15ip.h" #include "vega10/HDP/hdp_4_0_offset.h" #include "vega10/HDP/hdp_4_0_sh_mask.h" #include "vega10/GC/gc_9_0_sh_mask.h" +#include "vega10/DC/dce_12_0_offset.h" +#include "vega10/DC/dce_12_0_sh_mask.h" #include "vega10/vega10_enum.h" +#include "vega10/MMHUB/mmhub_1_0_offset.h" +#include "vega10/ATHUB/athub_1_0_offset.h" #include "soc15_common.h" @@ -68,13 +73,25 @@ static const u32 golden_settings_vega10_hdp[] = 0xf6e, 0x0fffffff, 0x00000000, }; +static const u32 golden_settings_mmhub_1_0_0[] = +{ + SOC15_REG_OFFSET(MMHUB, 0, mmDAGB1_WRCLI2), 0x00000007, 0xfe5fe0fa, + SOC15_REG_OFFSET(MMHUB, 0, mmMMEA1_DRAM_WR_CLI2GRP_MAP0), 0x00000030, 0x55555565 +}; + +static const u32 golden_settings_athub_1_0_0[] = +{ + SOC15_REG_OFFSET(ATHUB, 0, mmRPB_ARB_CNTL), 0x0000ff00, 0x00000800, + SOC15_REG_OFFSET(ATHUB, 0, mmRPB_ARB_CNTL2), 0x00ff00ff, 0x00080008 +}; + static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev, struct amdgpu_irq_src *src, unsigned type, enum amdgpu_interrupt_state state) { struct amdgpu_vmhub *hub; - u32 tmp, reg, bits, i; + u32 tmp, reg, bits, i, j; bits = VM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | VM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | @@ -86,43 +103,26 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev, switch (state) { case AMDGPU_IRQ_STATE_DISABLE: - /* MM HUB */ - hub = &adev->vmhub[AMDGPU_MMHUB]; - for (i = 0; i< 16; i++) { - reg = hub->vm_context0_cntl + i; - tmp = RREG32(reg); - tmp &= ~bits; - WREG32(reg, tmp); - } - - /* GFX HUB */ - hub = &adev->vmhub[AMDGPU_GFXHUB]; - for (i = 0; i < 16; i++) { - reg = hub->vm_context0_cntl + i; - tmp = RREG32(reg); - tmp &= ~bits; - WREG32(reg, tmp); + for (j = 0; j < AMDGPU_MAX_VMHUBS; j++) { + hub = &adev->vmhub[j]; + for (i = 0; i < 16; i++) { + reg = hub->vm_context0_cntl + i; + tmp = RREG32(reg); + tmp &= ~bits; + WREG32(reg, tmp); + } } break; case AMDGPU_IRQ_STATE_ENABLE: - /* MM HUB */ - hub = &adev->vmhub[AMDGPU_MMHUB]; - for (i = 0; i< 16; i++) { - reg = hub->vm_context0_cntl + i; - tmp = RREG32(reg); - tmp |= bits; - WREG32(reg, tmp); + for (j = 0; j < AMDGPU_MAX_VMHUBS; j++) { + hub = &adev->vmhub[j]; + for (i = 0; i < 16; i++) { + reg = hub->vm_context0_cntl + i; + tmp = RREG32(reg); + tmp |= bits; + WREG32(reg, tmp); + } } - - /* GFX HUB */ - hub = &adev->vmhub[AMDGPU_GFXHUB]; - for (i = 0; i < 16; i++) { - reg = hub->vm_context0_cntl + i; - tmp = RREG32(reg); - tmp |= bits; - WREG32(reg, tmp); - } - break; default: break; } @@ -392,7 +392,16 @@ static int gmc_v9_0_early_init(void *handle) static int gmc_v9_0_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - unsigned vm_inv_eng[AMDGPU_MAX_VMHUBS] = { 3, 3 }; + /* + * The latest engine allocation on gfx9 is: + * Engine 0, 1: idle + * Engine 2, 3: firmware + * Engine 4~13: amdgpu ring, subject to change when ring number changes + * Engine 14~15: idle + * Engine 16: kfd tlb invalidation + * Engine 17: Gart flushes + */ + unsigned vm_inv_eng[AMDGPU_MAX_VMHUBS] = { 4, 4 }; unsigned i; for(i = 0; i < adev->num_rings; ++i) { @@ -405,9 +414,9 @@ static int gmc_v9_0_late_init(void *handle) ring->funcs->vmhub); } - /* Engine 17 is used for GART flushes */ + /* Engine 16 is used for KFD and 17 for GART flushes */ for(i = 0; i < AMDGPU_MAX_VMHUBS; ++i) - BUG_ON(vm_inv_eng[i] > 17); + BUG_ON(vm_inv_eng[i] > 16); return amdgpu_irq_get(adev, &adev->mc.vm_fault, 0); } @@ -419,8 +428,7 @@ static void gmc_v9_0_vram_gtt_location(struct amdgpu_device *adev, if (!amdgpu_sriov_vf(adev)) base = mmhub_v1_0_get_fb_location(adev); amdgpu_vram_location(adev, &adev->mc, base); - adev->mc.gtt_base_align = 0; - amdgpu_gtt_location(adev, mc); + amdgpu_gart_location(adev, mc); /* base offset of vram pages */ if (adev->flags & AMD_IS_APU) adev->vm_manager.vram_base_offset = gfxhub_v1_0_get_mc_fb_offset(adev); @@ -442,43 +450,46 @@ static int gmc_v9_0_mc_init(struct amdgpu_device *adev) u32 tmp; int chansize, numchan; - /* hbm memory channel size */ - chansize = 128; - - tmp = RREG32_SOC15(DF, 0, mmDF_CS_AON0_DramBaseAddress0); - tmp &= DF_CS_AON0_DramBaseAddress0__IntLvNumChan_MASK; - tmp >>= DF_CS_AON0_DramBaseAddress0__IntLvNumChan__SHIFT; - switch (tmp) { - case 0: - default: - numchan = 1; - break; - case 1: - numchan = 2; - break; - case 2: - numchan = 0; - break; - case 3: - numchan = 4; - break; - case 4: - numchan = 0; - break; - case 5: - numchan = 8; - break; - case 6: - numchan = 0; - break; - case 7: - numchan = 16; - break; - case 8: - numchan = 2; - break; + adev->mc.vram_width = amdgpu_atomfirmware_get_vram_width(adev); + if (!adev->mc.vram_width) { + /* hbm memory channel size */ + chansize = 128; + + tmp = RREG32_SOC15(DF, 0, mmDF_CS_AON0_DramBaseAddress0); + tmp &= DF_CS_AON0_DramBaseAddress0__IntLvNumChan_MASK; + tmp >>= DF_CS_AON0_DramBaseAddress0__IntLvNumChan__SHIFT; + switch (tmp) { + case 0: + default: + numchan = 1; + break; + case 1: + numchan = 2; + break; + case 2: + numchan = 0; + break; + case 3: + numchan = 4; + break; + case 4: + numchan = 0; + break; + case 5: + numchan = 8; + break; + case 6: + numchan = 0; + break; + case 7: + numchan = 16; + break; + case 8: + numchan = 2; + break; + } + adev->mc.vram_width = numchan * chansize; } - adev->mc.vram_width = numchan * chansize; /* Could aper size report 0 ? */ adev->mc.aper_base = pci_resource_start(adev->pdev, 0); @@ -494,14 +505,20 @@ static int gmc_v9_0_mc_init(struct amdgpu_device *adev) if (adev->mc.visible_vram_size > adev->mc.real_vram_size) adev->mc.visible_vram_size = adev->mc.real_vram_size; - /* unless the user had overridden it, set the gart - * size equal to the 1024 or vram, whichever is larger. - */ - if (amdgpu_gart_size == -1) - adev->mc.gtt_size = max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20), - adev->mc.mc_vram_size); - else - adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20; + /* set the gart size */ + if (amdgpu_gart_size == -1) { + switch (adev->asic_type) { + case CHIP_VEGA10: /* all engines support GPUVM */ + default: + adev->mc.gart_size = 256ULL << 20; + break; + case CHIP_RAVEN: /* DCE SG support */ + adev->mc.gart_size = 1024ULL << 20; + break; + } + } else { + adev->mc.gart_size = (u64)amdgpu_gart_size << 20; + } gmc_v9_0_vram_gtt_location(adev, &adev->mc); @@ -537,10 +554,21 @@ static int gmc_v9_0_sw_init(void *handle) spin_lock_init(&adev->mc.invalidate_lock); - if (adev->flags & AMD_IS_APU) { + switch (adev->asic_type) { + case CHIP_RAVEN: adev->mc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN; - amdgpu_vm_adjust_size(adev, 64); - } else { + if (adev->rev_id == 0x0 || adev->rev_id == 0x1) { + adev->vm_manager.vm_size = 1U << 18; + adev->vm_manager.block_size = 9; + adev->vm_manager.num_level = 3; + amdgpu_vm_set_fragment_size(adev, 9); + } else { + /* vm_size is 64GB for legacy 2-level page support */ + amdgpu_vm_adjust_size(adev, 64, 9); + adev->vm_manager.num_level = 1; + } + break; + case CHIP_VEGA10: /* XXX Don't know how to get VRAM type yet. */ adev->mc.vram_type = AMDGPU_VRAM_TYPE_HBM; /* @@ -550,11 +578,18 @@ static int gmc_v9_0_sw_init(void *handle) */ adev->vm_manager.vm_size = 1U << 18; adev->vm_manager.block_size = 9; - DRM_INFO("vm size is %llu GB, block size is %u-bit\n", - adev->vm_manager.vm_size, - adev->vm_manager.block_size); + adev->vm_manager.num_level = 3; + amdgpu_vm_set_fragment_size(adev, 9); + break; + default: + break; } + DRM_INFO("vm size is %llu GB, block size is %u-bit,fragment size is %u-bit\n", + adev->vm_manager.vm_size, + adev->vm_manager.block_size, + adev->vm_manager.fragment_size); + /* This interrupt is VMC page fault.*/ r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_VMC, 0, &adev->mc.vm_fault); @@ -619,11 +654,6 @@ static int gmc_v9_0_sw_init(void *handle) adev->vm_manager.id_mgr[AMDGPU_GFXHUB].num_ids = AMDGPU_NUM_OF_VMIDS; adev->vm_manager.id_mgr[AMDGPU_MMHUB].num_ids = AMDGPU_NUM_OF_VMIDS; - /* TODO: fix num_level for APU when updating vm size and block size */ - if (adev->flags & AMD_IS_APU) - adev->vm_manager.num_level = 1; - else - adev->vm_manager.num_level = 3; amdgpu_vm_manager_init(adev); return 0; @@ -658,8 +688,17 @@ static void gmc_v9_0_init_golden_registers(struct amdgpu_device *adev) { switch (adev->asic_type) { case CHIP_VEGA10: + amdgpu_program_register_sequence(adev, + golden_settings_mmhub_1_0_0, + (const u32)ARRAY_SIZE(golden_settings_mmhub_1_0_0)); + amdgpu_program_register_sequence(adev, + golden_settings_athub_1_0_0, + (const u32)ARRAY_SIZE(golden_settings_athub_1_0_0)); break; case CHIP_RAVEN: + amdgpu_program_register_sequence(adev, + golden_settings_athub_1_0_0, + (const u32)ARRAY_SIZE(golden_settings_athub_1_0_0)); break; default: break; @@ -689,12 +728,6 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) if (r) return r; - /* After HDP is initialized, flush HDP.*/ - if (adev->flags & AMD_IS_APU) - nbio_v7_0_hdp_flush(adev); - else - nbio_v6_1_hdp_flush(adev); - switch (adev->asic_type) { case CHIP_RAVEN: mmhub_v1_0_initialize_power_gating(adev); @@ -712,13 +745,16 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) if (r) return r; - tmp = RREG32_SOC15(HDP, 0, mmHDP_MISC_CNTL); - tmp |= HDP_MISC_CNTL__FLUSH_INVALIDATE_CACHE_MASK; - WREG32_SOC15(HDP, 0, mmHDP_MISC_CNTL, tmp); + WREG32_FIELD15(HDP, 0, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 1); tmp = RREG32_SOC15(HDP, 0, mmHDP_HOST_PATH_CNTL); WREG32_SOC15(HDP, 0, mmHDP_HOST_PATH_CNTL, tmp); + /* After HDP is initialized, flush HDP.*/ + if (adev->flags & AMD_IS_APU) + nbio_v7_0_hdp_flush(adev); + else + nbio_v6_1_hdp_flush(adev); if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) value = false; @@ -727,11 +763,10 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) gfxhub_v1_0_set_fault_enable_default(adev, value); mmhub_v1_0_set_fault_enable_default(adev, value); - gmc_v9_0_gart_flush_gpu_tlb(adev, 0); DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", - (unsigned)(adev->mc.gtt_size >> 20), + (unsigned)(adev->mc.gart_size >> 20), (unsigned long long)adev->gart.table_addr); adev->gart.ready = true; return 0; @@ -745,6 +780,14 @@ static int gmc_v9_0_hw_init(void *handle) /* The sequence of these two function calls matters.*/ gmc_v9_0_init_golden_registers(adev); + if (adev->mode_info.num_crtc) { + /* Lockout access through VGA aperture*/ + WREG32_FIELD15(DCE, 0, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1); + + /* disable VGA render */ + WREG32_FIELD15(DCE, 0, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0); + } + r = gmc_v9_0_gart_enable(adev); return r; @@ -784,9 +827,7 @@ static int gmc_v9_0_suspend(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - gmc_v9_0_hw_fini(adev); - - return 0; + return gmc_v9_0_hw_fini(adev); } static int gmc_v9_0_resume(void *handle) diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c index 7a0ea27ac429..bd592cb39f37 100644 --- a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c @@ -208,6 +208,34 @@ static u32 iceland_ih_get_wptr(struct amdgpu_device *adev) } /** + * iceland_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool iceland_ih_prescreen_iv(struct amdgpu_device *adev) +{ + u32 ring_index = adev->irq.ih.rptr >> 2; + u16 pasid; + + switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { + case 146: + case 147: + pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; + if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) + return true; + break; + default: + /* Not a VM fault */ + return true; + } + + adev->irq.ih.rptr += 16; + return false; +} + +/** * iceland_ih_decode_iv - decode an interrupt vector * * @adev: amdgpu_device pointer @@ -412,6 +440,7 @@ static const struct amd_ip_funcs iceland_ih_ip_funcs = { static const struct amdgpu_ih_funcs iceland_ih_funcs = { .get_wptr = iceland_ih_get_wptr, + .prescreen_iv = iceland_ih_prescreen_iv, .decode_iv = iceland_ih_decode_iv, .set_rptr = iceland_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c index 3bbf2ccfca89..f33d1ffdb20b 100644 --- a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c @@ -42,7 +42,6 @@ #define KV_MINIMUM_ENGINE_CLOCK 800 #define SMC_RAM_END 0x40000 -static void kv_dpm_set_dpm_funcs(struct amdgpu_device *adev); static void kv_dpm_set_irq_funcs(struct amdgpu_device *adev); static int kv_enable_nb_dpm(struct amdgpu_device *adev, bool enable); @@ -64,7 +63,7 @@ static int kv_set_thermal_temperature_range(struct amdgpu_device *adev, int min_temp, int max_temp); static int kv_init_fps_limits(struct amdgpu_device *adev); -static void kv_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate); +static void kv_dpm_powergate_uvd(void *handle, bool gate); static void kv_dpm_powergate_vce(struct amdgpu_device *adev, bool gate); static void kv_dpm_powergate_samu(struct amdgpu_device *adev, bool gate); static void kv_dpm_powergate_acp(struct amdgpu_device *adev, bool gate); @@ -1245,8 +1244,9 @@ static void kv_update_requested_ps(struct amdgpu_device *adev, adev->pm.dpm.requested_ps = &pi->requested_rps; } -static void kv_dpm_enable_bapm(struct amdgpu_device *adev, bool enable) +static void kv_dpm_enable_bapm(void *handle, bool enable) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); int ret; @@ -1672,8 +1672,9 @@ static int kv_update_acp_dpm(struct amdgpu_device *adev, bool gate) return kv_enable_acp_dpm(adev, !gate); } -static void kv_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate) +static void kv_dpm_powergate_uvd(void *handle, bool gate) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); int ret; @@ -1868,10 +1869,11 @@ static int kv_enable_nb_dpm(struct amdgpu_device *adev, return ret; } -static int kv_dpm_force_performance_level(struct amdgpu_device *adev, +static int kv_dpm_force_performance_level(void *handle, enum amd_dpm_forced_level level) { int ret; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (level == AMD_DPM_FORCED_LEVEL_HIGH) { ret = kv_force_dpm_highest(adev); @@ -1892,8 +1894,9 @@ static int kv_dpm_force_performance_level(struct amdgpu_device *adev, return 0; } -static int kv_dpm_pre_set_power_state(struct amdgpu_device *adev) +static int kv_dpm_pre_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps; struct amdgpu_ps *new_ps = &requested_ps; @@ -1907,8 +1910,9 @@ static int kv_dpm_pre_set_power_state(struct amdgpu_device *adev) return 0; } -static int kv_dpm_set_power_state(struct amdgpu_device *adev) +static int kv_dpm_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); struct amdgpu_ps *new_ps = &pi->requested_rps; struct amdgpu_ps *old_ps = &pi->current_rps; @@ -1981,8 +1985,9 @@ static int kv_dpm_set_power_state(struct amdgpu_device *adev) return 0; } -static void kv_dpm_post_set_power_state(struct amdgpu_device *adev) +static void kv_dpm_post_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); struct amdgpu_ps *new_ps = &pi->requested_rps; @@ -2848,9 +2853,10 @@ static int kv_dpm_init(struct amdgpu_device *adev) } static void -kv_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev, +kv_dpm_debugfs_print_current_performance_level(void *handle, struct seq_file *m) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); u32 current_index = (RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX) & @@ -2875,11 +2881,12 @@ kv_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev, } static void -kv_dpm_print_power_state(struct amdgpu_device *adev, - struct amdgpu_ps *rps) +kv_dpm_print_power_state(void *handle, void *request_ps) { int i; + struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps; struct kv_ps *ps = kv_get_ps(rps); + struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_dpm_print_class_info(rps->class, rps->class2); amdgpu_dpm_print_cap_info(rps->caps); @@ -2905,13 +2912,14 @@ static void kv_dpm_fini(struct amdgpu_device *adev) amdgpu_free_extended_power_table(adev); } -static void kv_dpm_display_configuration_changed(struct amdgpu_device *adev) +static void kv_dpm_display_configuration_changed(void *handle) { } -static u32 kv_dpm_get_sclk(struct amdgpu_device *adev, bool low) +static u32 kv_dpm_get_sclk(void *handle, bool low) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); struct kv_ps *requested_state = kv_get_ps(&pi->requested_rps); @@ -2921,18 +2929,20 @@ static u32 kv_dpm_get_sclk(struct amdgpu_device *adev, bool low) return requested_state->levels[requested_state->num_levels - 1].sclk; } -static u32 kv_dpm_get_mclk(struct amdgpu_device *adev, bool low) +static u32 kv_dpm_get_mclk(void *handle, bool low) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); return pi->sys_info.bootup_uma_clk; } /* get temperature in millidegrees */ -static int kv_dpm_get_temp(struct amdgpu_device *adev) +static int kv_dpm_get_temp(void *handle) { u32 temp; int actual_temp = 0; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; temp = RREG32_SMC(0xC0300E0C); @@ -2950,7 +2960,6 @@ static int kv_dpm_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - kv_dpm_set_dpm_funcs(adev); kv_dpm_set_irq_funcs(adev); return 0; @@ -2960,16 +2969,10 @@ static int kv_dpm_late_init(void *handle) { /* powerdown unused blocks for now */ struct amdgpu_device *adev = (struct amdgpu_device *)handle; - int ret; if (!amdgpu_dpm) return 0; - /* init the sysfs and debugfs files late */ - ret = amdgpu_pm_sysfs_init(adev); - if (ret) - return ret; - kv_dpm_powergate_acp(adev, true); kv_dpm_powergate_samu(adev, true); @@ -3031,7 +3034,6 @@ static int kv_dpm_sw_fini(void *handle) flush_work(&adev->pm.dpm.thermal.work); mutex_lock(&adev->pm.mutex); - amdgpu_pm_sysfs_fini(adev); kv_dpm_fini(adev); mutex_unlock(&adev->pm.mutex); @@ -3222,14 +3224,17 @@ static inline bool kv_are_power_levels_equal(const struct kv_pl *kv_cpl1, (kv_cpl1->force_nbp_state == kv_cpl2->force_nbp_state)); } -static int kv_check_state_equal(struct amdgpu_device *adev, - struct amdgpu_ps *cps, - struct amdgpu_ps *rps, +static int kv_check_state_equal(void *handle, + void *current_ps, + void *request_ps, bool *equal) { struct kv_ps *kv_cps; struct kv_ps *kv_rps; int i; + struct amdgpu_ps *cps = (struct amdgpu_ps *)current_ps; + struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (adev == NULL || cps == NULL || rps == NULL || equal == NULL) return -EINVAL; @@ -3262,9 +3267,10 @@ static int kv_check_state_equal(struct amdgpu_device *adev, return 0; } -static int kv_dpm_read_sensor(struct amdgpu_device *adev, int idx, +static int kv_dpm_read_sensor(void *handle, int idx, void *value, int *size) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); uint32_t sclk; u32 pl_index = @@ -3312,7 +3318,7 @@ const struct amd_ip_funcs kv_dpm_ip_funcs = { .set_powergating_state = kv_dpm_set_powergating_state, }; -static const struct amdgpu_dpm_funcs kv_dpm_funcs = { +const struct amd_pm_funcs kv_dpm_funcs = { .get_temperature = &kv_dpm_get_temp, .pre_set_power_state = &kv_dpm_pre_set_power_state, .set_power_state = &kv_dpm_set_power_state, @@ -3330,12 +3336,6 @@ static const struct amdgpu_dpm_funcs kv_dpm_funcs = { .read_sensor = &kv_dpm_read_sensor, }; -static void kv_dpm_set_dpm_funcs(struct amdgpu_device *adev) -{ - if (adev->pm.funcs == NULL) - adev->pm.funcs = &kv_dpm_funcs; -} - static const struct amdgpu_irq_src_funcs kv_dpm_irq_funcs = { .set = kv_dpm_set_interrupt_state, .process = kv_dpm_process_interrupt, diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c index 9804318f3488..cc21c4bdec27 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -69,14 +69,14 @@ static void mmhub_v1_0_init_gart_aperture_regs(struct amdgpu_device *adev) mmhub_v1_0_init_gart_pt_regs(adev); WREG32_SOC15(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32, - (u32)(adev->mc.gtt_start >> 12)); + (u32)(adev->mc.gart_start >> 12)); WREG32_SOC15(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32, - (u32)(adev->mc.gtt_start >> 44)); + (u32)(adev->mc.gart_start >> 44)); WREG32_SOC15(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32, - (u32)(adev->mc.gtt_end >> 12)); + (u32)(adev->mc.gart_end >> 12)); WREG32_SOC15(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32, - (u32)(adev->mc.gtt_end >> 44)); + (u32)(adev->mc.gart_end >> 44)); } static void mmhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev) @@ -143,7 +143,7 @@ static void mmhub_v1_0_init_cache_regs(struct amdgpu_device *adev) /* Setup L2 cache */ tmp = RREG32_SOC15(MMHUB, 0, mmVM_L2_CNTL); tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1); - tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 1); /* XXX for emulation, Refer to closed source code.*/ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE, 0); @@ -158,6 +158,8 @@ static void mmhub_v1_0_init_cache_regs(struct amdgpu_device *adev) WREG32_SOC15(MMHUB, 0, mmVM_L2_CNTL2, tmp); tmp = mmVM_L2_CNTL3_DEFAULT; + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 9); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_FRAGMENT_SIZE, 6); WREG32_SOC15(MMHUB, 0, mmVM_L2_CNTL3, tmp); tmp = mmVM_L2_CNTL4_DEFAULT; @@ -222,6 +224,9 @@ static void mmhub_v1_0_setup_vmid_config(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_BLOCK_SIZE, adev->vm_manager.block_size - 9); + /* Send no-retry XNACK on fault to suppress VM fault storm. */ + tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, + RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0); WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT1_CNTL, i, tmp); WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32, i*2, 0); WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32, i*2, 0); @@ -245,61 +250,68 @@ static void mmhub_v1_0_program_invalidation(struct amdgpu_device *adev) } struct pctl_data { - uint32_t index; - uint32_t data; + uint32_t index; + uint32_t data; }; -const struct pctl_data pctl0_data[] = { - {0x0, 0x7a640}, - {0x9, 0x2a64a}, - {0xd, 0x2a680}, - {0x11, 0x6a684}, - {0x19, 0xea68e}, - {0x29, 0xa69e}, - {0x2b, 0x34a6c0}, - {0x61, 0x83a707}, - {0xe6, 0x8a7a4}, - {0xf0, 0x1a7b8}, - {0xf3, 0xfa7cc}, - {0x104, 0x17a7dd}, - {0x11d, 0xa7dc}, - {0x11f, 0x12a7f5}, - {0x133, 0xa808}, - {0x135, 0x12a810}, - {0x149, 0x7a82c} +static const struct pctl_data pctl0_data[] = { + {0x0, 0x7a640}, + {0x9, 0x2a64a}, + {0xd, 0x2a680}, + {0x11, 0x6a684}, + {0x19, 0xea68e}, + {0x29, 0xa69e}, + {0x2b, 0x34a6c0}, + {0x61, 0x83a707}, + {0xe6, 0x8a7a4}, + {0xf0, 0x1a7b8}, + {0xf3, 0xfa7cc}, + {0x104, 0x17a7dd}, + {0x11d, 0xa7dc}, + {0x11f, 0x12a7f5}, + {0x133, 0xa808}, + {0x135, 0x12a810}, + {0x149, 0x7a82c} }; -#define PCTL0_DATA_LEN (sizeof(pctl0_data)/sizeof(pctl0_data[0])) +#define PCTL0_DATA_LEN (ARRAY_SIZE(pctl0_data)) #define PCTL0_RENG_EXEC_END_PTR 0x151 #define PCTL0_STCTRL_REG_SAVE_RANGE0_BASE 0xa640 #define PCTL0_STCTRL_REG_SAVE_RANGE0_LIMIT 0xa833 -const struct pctl_data pctl1_data[] = { - {0x0, 0x39a000}, - {0x3b, 0x44a040}, - {0x81, 0x2a08d}, - {0x85, 0x6ba094}, - {0xf2, 0x18a100}, - {0x10c, 0x4a132}, - {0x112, 0xca141}, - {0x120, 0x2fa158}, - {0x151, 0x17a1d0}, - {0x16a, 0x1a1e9}, - {0x16d, 0x13a1ec}, - {0x182, 0x7a201}, - {0x18b, 0x3a20a}, - {0x190, 0x7a580}, - {0x199, 0xa590}, - {0x19b, 0x4a594}, - {0x1a1, 0x1a59c}, - {0x1a4, 0x7a82c}, - {0x1ad, 0xfa7cc}, - {0x1be, 0x17a7dd}, - {0x1d7, 0x12a810} +static const struct pctl_data pctl1_data[] = { + {0x0, 0x39a000}, + {0x3b, 0x44a040}, + {0x81, 0x2a08d}, + {0x85, 0x6ba094}, + {0xf2, 0x18a100}, + {0x10c, 0x4a132}, + {0x112, 0xca141}, + {0x120, 0x2fa158}, + {0x151, 0x17a1d0}, + {0x16a, 0x1a1e9}, + {0x16d, 0x13a1ec}, + {0x182, 0x7a201}, + {0x18b, 0x3a20a}, + {0x190, 0x7a580}, + {0x199, 0xa590}, + {0x19b, 0x4a594}, + {0x1a1, 0x1a59c}, + {0x1a4, 0x7a82c}, + {0x1ad, 0xfa7cc}, + {0x1be, 0x17a7dd}, + {0x1d7, 0x12a810}, + {0x1eb, 0x4000a7e1}, + {0x1ec, 0x5000a7f5}, + {0x1ed, 0x4000a7e2}, + {0x1ee, 0x5000a7dc}, + {0x1ef, 0x4000a7e3}, + {0x1f0, 0x5000a7f6}, + {0x1f1, 0x5000a7e4} }; -#define PCTL1_DATA_LEN (sizeof(pctl1_data)/sizeof(pctl1_data[0])) +#define PCTL1_DATA_LEN (ARRAY_SIZE(pctl1_data)) -#define PCTL1_RENG_EXEC_END_PTR 0x1ea +#define PCTL1_RENG_EXEC_END_PTR 0x1f1 #define PCTL1_STCTRL_REG_SAVE_RANGE0_BASE 0xa000 #define PCTL1_STCTRL_REG_SAVE_RANGE0_LIMIT 0xa20d #define PCTL1_STCTRL_REG_SAVE_RANGE1_BASE 0xa580 @@ -549,6 +561,13 @@ void mmhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, bool value) WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value); tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + if (!value) { + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_NO_RETRY_FAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_RETRY_FAULT, 1); + } + WREG32_SOC15(MMHUB, 0, mmVM_L2_PROTECTION_FAULT_CNTL, tmp); } diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h index 57bb940c0ecd..5d38229baf69 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h @@ -36,7 +36,4 @@ void mmhub_v1_0_initialize_power_gating(struct amdgpu_device *adev); void mmhub_v1_0_update_power_gating(struct amdgpu_device *adev, bool enable); -extern const struct amd_ip_funcs mmhub_v1_0_ip_funcs; -extern const struct amdgpu_ip_block_version mmhub_v1_0_ip_block; - #endif diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c index bde3ca3c21c1..b4906d2f30d3 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c @@ -72,21 +72,6 @@ static void xgpu_ai_mailbox_set_valid(struct amdgpu_device *adev, bool val) reg); } -static void xgpu_ai_mailbox_trans_msg(struct amdgpu_device *adev, - enum idh_request req) -{ - u32 reg; - - reg = RREG32_NO_KIQ(SOC15_REG_OFFSET(NBIO, 0, - mmBIF_BX_PF0_MAILBOX_MSGBUF_TRN_DW0)); - reg = REG_SET_FIELD(reg, BIF_BX_PF0_MAILBOX_MSGBUF_TRN_DW0, - MSGBUF_DATA, req); - WREG32_NO_KIQ(SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_MAILBOX_MSGBUF_TRN_DW0), - reg); - - xgpu_ai_mailbox_set_valid(adev, true); -} - static int xgpu_ai_mailbox_rcv_msg(struct amdgpu_device *adev, enum idh_event event) { @@ -154,13 +139,25 @@ static int xgpu_ai_poll_msg(struct amdgpu_device *adev, enum idh_event event) return r; } - -static int xgpu_ai_send_access_requests(struct amdgpu_device *adev, - enum idh_request req) -{ +static void xgpu_ai_mailbox_trans_msg (struct amdgpu_device *adev, + enum idh_request req, u32 data1, u32 data2, u32 data3) { + u32 reg; int r; - xgpu_ai_mailbox_trans_msg(adev, req); + reg = RREG32_NO_KIQ(SOC15_REG_OFFSET(NBIO, 0, + mmBIF_BX_PF0_MAILBOX_MSGBUF_TRN_DW0)); + reg = REG_SET_FIELD(reg, BIF_BX_PF0_MAILBOX_MSGBUF_TRN_DW0, + MSGBUF_DATA, req); + WREG32_NO_KIQ(SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_MAILBOX_MSGBUF_TRN_DW0), + reg); + WREG32_NO_KIQ(SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_MAILBOX_MSGBUF_TRN_DW1), + data1); + WREG32_NO_KIQ(SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_MAILBOX_MSGBUF_TRN_DW2), + data2); + WREG32_NO_KIQ(SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_MAILBOX_MSGBUF_TRN_DW3), + data3); + + xgpu_ai_mailbox_set_valid(adev, true); /* start to poll ack */ r = xgpu_ai_poll_ack(adev); @@ -168,6 +165,14 @@ static int xgpu_ai_send_access_requests(struct amdgpu_device *adev, pr_err("Doesn't get ack from pf, continue\n"); xgpu_ai_mailbox_set_valid(adev, false); +} + +static int xgpu_ai_send_access_requests(struct amdgpu_device *adev, + enum idh_request req) +{ + int r; + + xgpu_ai_mailbox_trans_msg(adev, req, 0, 0, 0); /* start to check msg if request is idh_req_gpu_init_access */ if (req == IDH_REQ_GPU_INIT_ACCESS || @@ -178,6 +183,12 @@ static int xgpu_ai_send_access_requests(struct amdgpu_device *adev, pr_err("Doesn't get READY_TO_ACCESS_GPU from pf, give up\n"); return r; } + /* Retrieve checksum from mailbox2 */ + if (req == IDH_REQ_GPU_INIT_ACCESS) { + adev->virt.fw_reserve.checksum_key = + RREG32_NO_KIQ(SOC15_REG_OFFSET(NBIO, 0, + mmBIF_BX_PF0_MAILBOX_MSGBUF_RCV_DW2)); + } } return 0; @@ -342,4 +353,5 @@ const struct amdgpu_virt_ops xgpu_ai_virt_ops = { .req_full_gpu = xgpu_ai_request_full_gpu_access, .rel_full_gpu = xgpu_ai_release_full_gpu_access, .reset_gpu = xgpu_ai_request_reset, + .trans_msg = xgpu_ai_mailbox_trans_msg, }; diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h index 9aefc44d2c34..67e78576a9eb 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h @@ -24,14 +24,16 @@ #ifndef __MXGPU_AI_H__ #define __MXGPU_AI_H__ -#define AI_MAILBOX_TIMEDOUT 5000 +#define AI_MAILBOX_TIMEDOUT 12000 enum idh_request { IDH_REQ_GPU_INIT_ACCESS = 1, IDH_REL_GPU_INIT_ACCESS, IDH_REQ_GPU_FINI_ACCESS, IDH_REL_GPU_FINI_ACCESS, - IDH_REQ_GPU_RESET_ACCESS + IDH_REQ_GPU_RESET_ACCESS, + + IDH_LOG_VF_ERROR = 200, }; enum idh_event { diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c index 171a658135b5..c25a831f94ec 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c @@ -613,4 +613,5 @@ const struct amdgpu_virt_ops xgpu_vi_virt_ops = { .req_full_gpu = xgpu_vi_request_full_gpu_access, .rel_full_gpu = xgpu_vi_release_full_gpu_access, .reset_gpu = xgpu_vi_request_reset, + .trans_msg = NULL, /* Does not need to trans VF errors to host. */ }; diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.h b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.h index 2db741131bc6..f13dc6cc158f 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.h +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.h @@ -23,7 +23,7 @@ #ifndef __MXGPU_VI_H__ #define __MXGPU_VI_H__ -#define VI_MAILBOX_TIMEDOUT 5000 +#define VI_MAILBOX_TIMEDOUT 12000 #define VI_MAILBOX_RESET_TIME 12 /* VI mailbox messages request */ @@ -32,7 +32,9 @@ enum idh_request { IDH_REL_GPU_INIT_ACCESS, IDH_REQ_GPU_FINI_ACCESS, IDH_REL_GPU_FINI_ACCESS, - IDH_REQ_GPU_RESET_ACCESS + IDH_REQ_GPU_RESET_ACCESS, + + IDH_LOG_VF_ERROR = 200, }; /* VI mailbox messages data */ diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c index 1e272f785def..904a1bab9b9f 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c @@ -32,6 +32,7 @@ #define smnCPM_CONTROL 0x11180460 #define smnPCIE_CNTL2 0x11180070 +#define smnPCIE_CONFIG_CNTL 0x11180044 u32 nbio_v6_1_get_rev_id(struct amdgpu_device *adev) { @@ -67,7 +68,7 @@ void nbio_v6_1_mc_access_enable(struct amdgpu_device *adev, bool enable) void nbio_v6_1_hdp_flush(struct amdgpu_device *adev) { - WREG32_SOC15(NBIO, 0, mmBIF_BX_PF0_HDP_MEM_COHERENCY_FLUSH_CNTL, 0); + WREG32_SOC15_NO_KIQ(NBIO, 0, mmBIF_BX_PF0_HDP_MEM_COHERENCY_FLUSH_CNTL, 0); } u32 nbio_v6_1_get_memsize(struct amdgpu_device *adev) @@ -214,31 +215,27 @@ void nbio_v6_1_get_clockgating_state(struct amdgpu_device *adev, u32 *flags) *flags |= AMD_CG_SUPPORT_BIF_LS; } -struct nbio_hdp_flush_reg nbio_v6_1_hdp_flush_reg; -struct nbio_pcie_index_data nbio_v6_1_pcie_index_data; +const struct nbio_hdp_flush_reg nbio_v6_1_hdp_flush_reg = { + .hdp_flush_req_offset = SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_REQ), + .hdp_flush_done_offset = SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_DONE), + .ref_and_mask_cp0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP0_MASK, + .ref_and_mask_cp1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP1_MASK, + .ref_and_mask_cp2 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP2_MASK, + .ref_and_mask_cp3 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP3_MASK, + .ref_and_mask_cp4 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP4_MASK, + .ref_and_mask_cp5 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP5_MASK, + .ref_and_mask_cp6 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP6_MASK, + .ref_and_mask_cp7 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP7_MASK, + .ref_and_mask_cp8 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP8_MASK, + .ref_and_mask_cp9 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP9_MASK, + .ref_and_mask_sdma0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA0_MASK, + .ref_and_mask_sdma1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA1_MASK +}; -int nbio_v6_1_init(struct amdgpu_device *adev) -{ - nbio_v6_1_hdp_flush_reg.hdp_flush_req_offset = SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_REQ); - nbio_v6_1_hdp_flush_reg.hdp_flush_done_offset = SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_DONE); - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP0_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP1_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp2 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP2_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp3 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP3_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp4 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP4_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp5 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP5_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp6 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP6_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp7 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP7_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp8 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP8_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_cp9 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP9_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_sdma0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA0_MASK; - nbio_v6_1_hdp_flush_reg.ref_and_mask_sdma1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA1_MASK; - - nbio_v6_1_pcie_index_data.index_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX); - nbio_v6_1_pcie_index_data.data_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA); - - return 0; -} +const struct nbio_pcie_index_data nbio_v6_1_pcie_index_data = { + .index_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX), + .data_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA), +}; void nbio_v6_1_detect_hw_virt(struct amdgpu_device *adev) { @@ -256,3 +253,15 @@ void nbio_v6_1_detect_hw_virt(struct amdgpu_device *adev) adev->virt.caps |= AMDGPU_PASSTHROUGH_MODE; } } + +void nbio_v6_1_init_registers(struct amdgpu_device *adev) +{ + uint32_t def, data; + + def = data = RREG32_PCIE(smnPCIE_CONFIG_CNTL); + data = REG_SET_FIELD(data, PCIE_CONFIG_CNTL, CI_SWUS_MAX_READ_REQUEST_SIZE_MODE, 1); + data = REG_SET_FIELD(data, PCIE_CONFIG_CNTL, CI_SWUS_MAX_READ_REQUEST_SIZE_PRIV, 1); + + if (def != data) + WREG32_PCIE(smnPCIE_CONFIG_CNTL, data); +} diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.h b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.h index f6f8bc045518..14ca8d45a46c 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.h +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.h @@ -26,8 +26,8 @@ #include "soc15_common.h" -extern struct nbio_hdp_flush_reg nbio_v6_1_hdp_flush_reg; -extern struct nbio_pcie_index_data nbio_v6_1_pcie_index_data; +extern const struct nbio_hdp_flush_reg nbio_v6_1_hdp_flush_reg; +extern const struct nbio_pcie_index_data nbio_v6_1_pcie_index_data; int nbio_v6_1_init(struct amdgpu_device *adev); u32 nbio_v6_1_get_atombios_scratch_regs(struct amdgpu_device *adev, uint32_t idx); @@ -50,5 +50,6 @@ void nbio_v6_1_update_medium_grain_clock_gating(struct amdgpu_device *adev, bool void nbio_v6_1_update_medium_grain_light_sleep(struct amdgpu_device *adev, bool enable); void nbio_v6_1_get_clockgating_state(struct amdgpu_device *adev, u32 *flags); void nbio_v6_1_detect_hw_virt(struct amdgpu_device *adev); +void nbio_v6_1_init_registers(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c index aa04632523fa..f802b973410a 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c @@ -65,7 +65,7 @@ void nbio_v7_0_mc_access_enable(struct amdgpu_device *adev, bool enable) void nbio_v7_0_hdp_flush(struct amdgpu_device *adev) { - WREG32_SOC15(NBIO, 0, mmHDP_MEM_COHERENCY_FLUSH_CNTL, 0); + WREG32_SOC15_NO_KIQ(NBIO, 0, mmHDP_MEM_COHERENCY_FLUSH_CNTL, 0); } u32 nbio_v7_0_get_memsize(struct amdgpu_device *adev) @@ -185,28 +185,24 @@ void nbio_v7_0_ih_control(struct amdgpu_device *adev) WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL, interrupt_cntl); } -struct nbio_hdp_flush_reg nbio_v7_0_hdp_flush_reg; -struct nbio_pcie_index_data nbio_v7_0_pcie_index_data; +const struct nbio_hdp_flush_reg nbio_v7_0_hdp_flush_reg = { + .hdp_flush_req_offset = SOC15_REG_OFFSET(NBIO, 0, mmGPU_HDP_FLUSH_REQ), + .hdp_flush_done_offset = SOC15_REG_OFFSET(NBIO, 0, mmGPU_HDP_FLUSH_DONE), + .ref_and_mask_cp0 = GPU_HDP_FLUSH_DONE__CP0_MASK, + .ref_and_mask_cp1 = GPU_HDP_FLUSH_DONE__CP1_MASK, + .ref_and_mask_cp2 = GPU_HDP_FLUSH_DONE__CP2_MASK, + .ref_and_mask_cp3 = GPU_HDP_FLUSH_DONE__CP3_MASK, + .ref_and_mask_cp4 = GPU_HDP_FLUSH_DONE__CP4_MASK, + .ref_and_mask_cp5 = GPU_HDP_FLUSH_DONE__CP5_MASK, + .ref_and_mask_cp6 = GPU_HDP_FLUSH_DONE__CP6_MASK, + .ref_and_mask_cp7 = GPU_HDP_FLUSH_DONE__CP7_MASK, + .ref_and_mask_cp8 = GPU_HDP_FLUSH_DONE__CP8_MASK, + .ref_and_mask_cp9 = GPU_HDP_FLUSH_DONE__CP9_MASK, + .ref_and_mask_sdma0 = GPU_HDP_FLUSH_DONE__SDMA0_MASK, + .ref_and_mask_sdma1 = GPU_HDP_FLUSH_DONE__SDMA1_MASK, +}; -int nbio_v7_0_init(struct amdgpu_device *adev) -{ - nbio_v7_0_hdp_flush_reg.hdp_flush_req_offset = SOC15_REG_OFFSET(NBIO, 0, mmGPU_HDP_FLUSH_REQ); - nbio_v7_0_hdp_flush_reg.hdp_flush_done_offset = SOC15_REG_OFFSET(NBIO, 0, mmGPU_HDP_FLUSH_DONE); - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp0 = GPU_HDP_FLUSH_DONE__CP0_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp1 = GPU_HDP_FLUSH_DONE__CP1_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp2 = GPU_HDP_FLUSH_DONE__CP2_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp3 = GPU_HDP_FLUSH_DONE__CP3_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp4 = GPU_HDP_FLUSH_DONE__CP4_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp5 = GPU_HDP_FLUSH_DONE__CP5_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp6 = GPU_HDP_FLUSH_DONE__CP6_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp7 = GPU_HDP_FLUSH_DONE__CP7_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp8 = GPU_HDP_FLUSH_DONE__CP8_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_cp9 = GPU_HDP_FLUSH_DONE__CP9_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_sdma0 = GPU_HDP_FLUSH_DONE__SDMA0_MASK; - nbio_v7_0_hdp_flush_reg.ref_and_mask_sdma1 = GPU_HDP_FLUSH_DONE__SDMA1_MASK; - - nbio_v7_0_pcie_index_data.index_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX2); - nbio_v7_0_pcie_index_data.data_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA2); - - return 0; -} +const struct nbio_pcie_index_data nbio_v7_0_pcie_index_data = { + .index_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX2), + .data_offset = SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA2) +}; diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.h b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.h index 054ff49427e6..df8fa90f40d7 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.h +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.h @@ -26,8 +26,8 @@ #include "soc15_common.h" -extern struct nbio_hdp_flush_reg nbio_v7_0_hdp_flush_reg; -extern struct nbio_pcie_index_data nbio_v7_0_pcie_index_data; +extern const struct nbio_hdp_flush_reg nbio_v7_0_hdp_flush_reg; +extern const struct nbio_pcie_index_data nbio_v7_0_pcie_index_data; int nbio_v7_0_init(struct amdgpu_device *adev); u32 nbio_v7_0_get_atombios_scratch_regs(struct amdgpu_device *adev, uint32_t idx); diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c index 2258323a3c26..4e20d91d5d50 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c @@ -35,6 +35,8 @@ #include "raven1/GC/gc_9_1_offset.h" #include "raven1/SDMA0/sdma0_4_1_offset.h" +MODULE_FIRMWARE("amdgpu/raven_asd.bin"); + static int psp_v10_0_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type *type) { @@ -86,19 +88,63 @@ psp_v10_0_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type * return 0; } +int psp_v10_0_init_microcode(struct psp_context *psp) +{ + struct amdgpu_device *adev = psp->adev; + const char *chip_name; + char fw_name[30]; + int err = 0; + const struct psp_firmware_header_v1_0 *hdr; + + DRM_DEBUG("\n"); + + switch (adev->asic_type) { + case CHIP_RAVEN: + chip_name = "raven"; + break; + default: BUG(); + } + + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_asd.bin", chip_name); + err = request_firmware(&adev->psp.asd_fw, fw_name, adev->dev); + if (err) + goto out; + + err = amdgpu_ucode_validate(adev->psp.asd_fw); + if (err) + goto out; + + hdr = (const struct psp_firmware_header_v1_0 *)adev->psp.asd_fw->data; + adev->psp.asd_fw_version = le32_to_cpu(hdr->header.ucode_version); + adev->psp.asd_feature_version = le32_to_cpu(hdr->ucode_feature_version); + adev->psp.asd_ucode_size = le32_to_cpu(hdr->header.ucode_size_bytes); + adev->psp.asd_start_addr = (uint8_t *)hdr + + le32_to_cpu(hdr->header.ucode_array_offset_bytes); + + return 0; +out: + if (err) { + dev_err(adev->dev, + "psp v10.0: Failed to load firmware \"%s\"\n", + fw_name); + release_firmware(adev->psp.asd_fw); + adev->psp.asd_fw = NULL; + } + + return err; +} + int psp_v10_0_prep_cmd_buf(struct amdgpu_firmware_info *ucode, struct psp_gfx_cmd_resp *cmd) { int ret; uint64_t fw_mem_mc_addr = ucode->mc_addr; - struct common_firmware_header *header; memset(cmd, 0, sizeof(struct psp_gfx_cmd_resp)); - header = (struct common_firmware_header *)ucode->fw; cmd->cmd_id = GFX_CMD_ID_LOAD_IP_FW; cmd->cmd.cmd_load_ip_fw.fw_phy_addr_lo = lower_32_bits(fw_mem_mc_addr); cmd->cmd.cmd_load_ip_fw.fw_phy_addr_hi = upper_32_bits(fw_mem_mc_addr); - cmd->cmd.cmd_load_ip_fw.fw_size = le32_to_cpu(header->ucode_size_bytes); + cmd->cmd.cmd_load_ip_fw.fw_size = ucode->ucode_size; ret = psp_v10_0_get_fw_type(ucode, &cmd->cmd.cmd_load_ip_fw.fw_type); if (ret) @@ -110,7 +156,6 @@ int psp_v10_0_prep_cmd_buf(struct amdgpu_firmware_info *ucode, struct psp_gfx_cm int psp_v10_0_ring_init(struct psp_context *psp, enum psp_ring_type ring_type) { int ret = 0; - unsigned int psp_ring_reg = 0; struct psp_ring *ring; struct amdgpu_device *adev = psp->adev; @@ -130,6 +175,16 @@ int psp_v10_0_ring_init(struct psp_context *psp, enum psp_ring_type ring_type) return ret; } + return 0; +} + +int psp_v10_0_ring_create(struct psp_context *psp, enum psp_ring_type ring_type) +{ + int ret = 0; + unsigned int psp_ring_reg = 0; + struct psp_ring *ring = &psp->km_ring; + struct amdgpu_device *adev = psp->adev; + /* Write low address of the ring to C2PMSG_69 */ psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr); WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, psp_ring_reg); @@ -143,13 +198,55 @@ int psp_v10_0_ring_init(struct psp_context *psp, enum psp_ring_type ring_type) psp_ring_reg = ring_type; psp_ring_reg = psp_ring_reg << 16; WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, psp_ring_reg); + + /* There might be handshake issue with hardware which needs delay */ + mdelay(20); + /* Wait for response flag (bit 31) in C2PMSG_64 */ - psp_ring_reg = 0; - while ((psp_ring_reg & 0x80000000) == 0) { - psp_ring_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64); - } + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), + 0x80000000, 0x8000FFFF, false); - return 0; + return ret; +} + +int psp_v10_0_ring_stop(struct psp_context *psp, enum psp_ring_type ring_type) +{ + int ret = 0; + struct psp_ring *ring; + unsigned int psp_ring_reg = 0; + struct amdgpu_device *adev = psp->adev; + + ring = &psp->km_ring; + + /* Write the ring destroy command to C2PMSG_64 */ + psp_ring_reg = 3 << 16; + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, psp_ring_reg); + + /* There might be handshake issue with hardware which needs delay */ + mdelay(20); + + /* Wait for response flag (bit 31) in C2PMSG_64 */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), + 0x80000000, 0x80000000, false); + + return ret; +} + +int psp_v10_0_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type) +{ + int ret = 0; + struct psp_ring *ring = &psp->km_ring; + struct amdgpu_device *adev = psp->adev; + + ret = psp_v10_0_ring_stop(psp, ring_type); + if (ret) + DRM_ERROR("Fail to stop psp ring\n"); + + amdgpu_bo_free_kernel(&adev->firmware.rbuf, + &ring->ring_mem_mc_addr, + (void **)&ring->ring_mem); + + return ret; } int psp_v10_0_cmd_submit(struct psp_context *psp, @@ -160,16 +257,31 @@ int psp_v10_0_cmd_submit(struct psp_context *psp, unsigned int psp_write_ptr_reg = 0; struct psp_gfx_rb_frame * write_frame = psp->km_ring.ring_mem; struct psp_ring *ring = &psp->km_ring; + struct psp_gfx_rb_frame *ring_buffer_start = ring->ring_mem; + struct psp_gfx_rb_frame *ring_buffer_end = ring_buffer_start + + ring->ring_size / sizeof(struct psp_gfx_rb_frame) - 1; struct amdgpu_device *adev = psp->adev; + uint32_t ring_size_dw = ring->ring_size / 4; + uint32_t rb_frame_size_dw = sizeof(struct psp_gfx_rb_frame) / 4; /* KM (GPCOM) prepare write pointer */ psp_write_ptr_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67); /* Update KM RB frame pointer to new frame */ - if ((psp_write_ptr_reg % ring->ring_size) == 0) - write_frame = ring->ring_mem; + if ((psp_write_ptr_reg % ring_size_dw) == 0) + write_frame = ring_buffer_start; else - write_frame = ring->ring_mem + (psp_write_ptr_reg / (sizeof(struct psp_gfx_rb_frame) / 4)); + write_frame = ring_buffer_start + (psp_write_ptr_reg / rb_frame_size_dw); + /* Check invalid write_frame ptr address */ + if ((write_frame < ring_buffer_start) || (ring_buffer_end < write_frame)) { + DRM_ERROR("ring_buffer_start = %p; ring_buffer_end = %p; write_frame = %p\n", + ring_buffer_start, ring_buffer_end, write_frame); + DRM_ERROR("write_frame is pointing to address out of bounds\n"); + return -EINVAL; + } + + /* Initialize KM RB frame */ + memset(write_frame, 0, sizeof(struct psp_gfx_rb_frame)); /* Update KM RB frame */ write_frame->cmd_buf_addr_hi = upper_32_bits(cmd_buf_mc_addr); @@ -179,8 +291,7 @@ int psp_v10_0_cmd_submit(struct psp_context *psp, write_frame->fence_value = index; /* Update the write Pointer in DWORDs */ - psp_write_ptr_reg += sizeof(struct psp_gfx_rb_frame) / 4; - psp_write_ptr_reg = (psp_write_ptr_reg >= ring->ring_size) ? 0 : psp_write_ptr_reg; + psp_write_ptr_reg = (psp_write_ptr_reg + rb_frame_size_dw) % ring_size_dw; WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67, psp_write_ptr_reg); return 0; @@ -306,3 +417,10 @@ bool psp_v10_0_compare_sram_data(struct psp_context *psp, return true; } + + +int psp_v10_0_mode1_reset(struct psp_context *psp) +{ + DRM_INFO("psp mode 1 reset not supported now! \n"); + return -EINVAL; +} diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.h b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.h index 2022b7b7151e..451e8308303f 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.h @@ -27,10 +27,17 @@ #include "amdgpu_psp.h" +extern int psp_v10_0_init_microcode(struct psp_context *psp); extern int psp_v10_0_prep_cmd_buf(struct amdgpu_firmware_info *ucode, struct psp_gfx_cmd_resp *cmd); extern int psp_v10_0_ring_init(struct psp_context *psp, enum psp_ring_type ring_type); +extern int psp_v10_0_ring_create(struct psp_context *psp, + enum psp_ring_type ring_type); +extern int psp_v10_0_ring_stop(struct psp_context *psp, + enum psp_ring_type ring_type); +extern int psp_v10_0_ring_destroy(struct psp_context *psp, + enum psp_ring_type ring_type); extern int psp_v10_0_cmd_submit(struct psp_context *psp, struct amdgpu_firmware_info *ucode, uint64_t cmd_buf_mc_addr, uint64_t fence_mc_addr, @@ -38,4 +45,6 @@ extern int psp_v10_0_cmd_submit(struct psp_context *psp, extern bool psp_v10_0_compare_sram_data(struct psp_context *psp, struct amdgpu_firmware_info *ucode, enum AMDGPU_UCODE_ID ucode_type); + +extern int psp_v10_0_mode1_reset(struct psp_context *psp); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c index c98d77d0c8f8..c7bcfe8e286c 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c @@ -237,11 +237,9 @@ int psp_v3_1_bootloader_load_sos(struct psp_context *psp) /* there might be handshake issue with hardware which needs delay */ mdelay(20); -#if 0 ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_81), RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81), 0, true); -#endif return ret; } @@ -321,7 +319,7 @@ int psp_v3_1_ring_create(struct psp_context *psp, enum psp_ring_type ring_type) return ret; } -int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type) +int psp_v3_1_ring_stop(struct psp_context *psp, enum psp_ring_type ring_type) { int ret = 0; struct psp_ring *ring; @@ -341,10 +339,23 @@ int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type) ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), 0x80000000, 0x80000000, false); - if (ring->ring_mem) - amdgpu_bo_free_kernel(&adev->firmware.rbuf, - &ring->ring_mem_mc_addr, - (void **)&ring->ring_mem); + return ret; +} + +int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type) +{ + int ret = 0; + struct psp_ring *ring = &psp->km_ring; + struct amdgpu_device *adev = psp->adev; + + ret = psp_v3_1_ring_stop(psp, ring_type); + if (ret) + DRM_ERROR("Fail to stop psp ring\n"); + + amdgpu_bo_free_kernel(&adev->firmware.rbuf, + &ring->ring_mem_mc_addr, + (void **)&ring->ring_mem); + return ret; } @@ -356,6 +367,9 @@ int psp_v3_1_cmd_submit(struct psp_context *psp, unsigned int psp_write_ptr_reg = 0; struct psp_gfx_rb_frame * write_frame = psp->km_ring.ring_mem; struct psp_ring *ring = &psp->km_ring; + struct psp_gfx_rb_frame *ring_buffer_start = ring->ring_mem; + struct psp_gfx_rb_frame *ring_buffer_end = ring_buffer_start + + ring->ring_size / sizeof(struct psp_gfx_rb_frame) - 1; struct amdgpu_device *adev = psp->adev; uint32_t ring_size_dw = ring->ring_size / 4; uint32_t rb_frame_size_dw = sizeof(struct psp_gfx_rb_frame) / 4; @@ -367,9 +381,16 @@ int psp_v3_1_cmd_submit(struct psp_context *psp, /* write_frame ptr increments by size of rb_frame in bytes */ /* psp_write_ptr_reg increments by size of rb_frame in DWORDs */ if ((psp_write_ptr_reg % ring_size_dw) == 0) - write_frame = ring->ring_mem; + write_frame = ring_buffer_start; else - write_frame = ring->ring_mem + (psp_write_ptr_reg / rb_frame_size_dw); + write_frame = ring_buffer_start + (psp_write_ptr_reg / rb_frame_size_dw); + /* Check invalid write_frame ptr address */ + if ((write_frame < ring_buffer_start) || (ring_buffer_end < write_frame)) { + DRM_ERROR("ring_buffer_start = %p; ring_buffer_end = %p; write_frame = %p\n", + ring_buffer_start, ring_buffer_end, write_frame); + DRM_ERROR("write_frame is pointing to address out of bounds\n"); + return -EINVAL; + } /* Initialize KM RB frame */ memset(write_frame, 0, sizeof(struct psp_gfx_rb_frame)); @@ -519,3 +540,37 @@ bool psp_v3_1_smu_reload_quirk(struct psp_context *psp) reg = RREG32_SOC15(NBIO, 0, mmPCIE_DATA2); return (reg & MP1_FIRMWARE_FLAGS__INTERRUPTS_ENABLED_MASK) ? true : false; } + +int psp_v3_1_mode1_reset(struct psp_context *psp) +{ + int ret; + uint32_t offset; + struct amdgpu_device *adev = psp->adev; + + offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64); + + ret = psp_wait_for(psp, offset, 0x80000000, 0x8000FFFF, false); + + if (ret) { + DRM_INFO("psp is not working correctly before mode1 reset!\n"); + return -EINVAL; + } + + /*send the mode 1 reset command*/ + WREG32(offset, 0x70000); + + mdelay(1000); + + offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33); + + ret = psp_wait_for(psp, offset, 0x80000000, 0x80000000, false); + + if (ret) { + DRM_INFO("psp mode 1 reset failed!\n"); + return -EINVAL; + } + + DRM_INFO("psp mode1 reset succeed \n"); + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.h b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.h index 9dcd0b25c4c6..b05dbada7751 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.h @@ -41,6 +41,8 @@ extern int psp_v3_1_ring_init(struct psp_context *psp, enum psp_ring_type ring_type); extern int psp_v3_1_ring_create(struct psp_context *psp, enum psp_ring_type ring_type); +extern int psp_v3_1_ring_stop(struct psp_context *psp, + enum psp_ring_type ring_type); extern int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type); extern int psp_v3_1_cmd_submit(struct psp_context *psp, @@ -51,4 +53,5 @@ extern bool psp_v3_1_compare_sram_data(struct psp_context *psp, struct amdgpu_firmware_info *ucode, enum AMDGPU_UCODE_ID ucode_type); extern bool psp_v3_1_smu_reload_quirk(struct psp_context *psp); +extern int psp_v3_1_mode1_reset(struct psp_context *psp); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index f2d0710258cb..67f375bfe452 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -561,21 +561,11 @@ static int sdma_v2_4_start(struct amdgpu_device *adev) { int r; - if (!adev->pp_enabled) { - if (adev->firmware.load_type != AMDGPU_FW_LOAD_SMU) { - r = sdma_v2_4_load_microcode(adev); - if (r) - return r; - } else { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_SDMA0); - if (r) - return -EINVAL; - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - AMDGPU_UCODE_ID_SDMA1); - if (r) - return -EINVAL; - } + + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + r = sdma_v2_4_load_microcode(adev); + if (r) + return r; } /* halt the engine before programing */ @@ -1324,8 +1314,13 @@ static void sdma_v2_4_set_buffer_funcs(struct amdgpu_device *adev) } static const struct amdgpu_vm_pte_funcs sdma_v2_4_vm_pte_funcs = { + .copy_pte_num_dw = 7, .copy_pte = sdma_v2_4_vm_copy_pte, + .write_pte = sdma_v2_4_vm_write_pte, + + .set_max_nums_pte_pde = 0x1fffff >> 3, + .set_pte_pde_num_dw = 10, .set_pte_pde = sdma_v2_4_vm_set_pte_pde, }; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 1d766ae98dc8..6d06f8eb659f 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -379,8 +379,10 @@ static void sdma_v3_0_ring_set_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) { + u32 *wb = (u32 *)&adev->wb.wb[ring->wptr_offs]; + /* XXX check if swapping is necessary on BE */ - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr) << 2; + WRITE_ONCE(*wb, (lower_32_bits(ring->wptr) << 2)); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr) << 2); } else { int me = (ring == &ring->adev->sdma.instance[0].ring) ? 0 : 1; @@ -551,17 +553,53 @@ static void sdma_v3_0_rlc_stop(struct amdgpu_device *adev) */ static void sdma_v3_0_ctx_switch_enable(struct amdgpu_device *adev, bool enable) { - u32 f32_cntl; + u32 f32_cntl, phase_quantum = 0; int i; + if (amdgpu_sdma_phase_quantum) { + unsigned value = amdgpu_sdma_phase_quantum; + unsigned unit = 0; + + while (value > (SDMA0_PHASE0_QUANTUM__VALUE_MASK >> + SDMA0_PHASE0_QUANTUM__VALUE__SHIFT)) { + value = (value + 1) >> 1; + unit++; + } + if (unit > (SDMA0_PHASE0_QUANTUM__UNIT_MASK >> + SDMA0_PHASE0_QUANTUM__UNIT__SHIFT)) { + value = (SDMA0_PHASE0_QUANTUM__VALUE_MASK >> + SDMA0_PHASE0_QUANTUM__VALUE__SHIFT); + unit = (SDMA0_PHASE0_QUANTUM__UNIT_MASK >> + SDMA0_PHASE0_QUANTUM__UNIT__SHIFT); + WARN_ONCE(1, + "clamping sdma_phase_quantum to %uK clock cycles\n", + value << unit); + } + phase_quantum = + value << SDMA0_PHASE0_QUANTUM__VALUE__SHIFT | + unit << SDMA0_PHASE0_QUANTUM__UNIT__SHIFT; + } + for (i = 0; i < adev->sdma.num_instances; i++) { f32_cntl = RREG32(mmSDMA0_CNTL + sdma_offsets[i]); - if (enable) + if (enable) { f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL, AUTO_CTXSW_ENABLE, 1); - else + f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL, + ATC_L1_ENABLE, 1); + if (amdgpu_sdma_phase_quantum) { + WREG32(mmSDMA0_PHASE0_QUANTUM + sdma_offsets[i], + phase_quantum); + WREG32(mmSDMA0_PHASE1_QUANTUM + sdma_offsets[i], + phase_quantum); + } + } else { f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL, AUTO_CTXSW_ENABLE, 0); + f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL, + ATC_L1_ENABLE, 1); + } + WREG32(mmSDMA0_CNTL + sdma_offsets[i], f32_cntl); } } @@ -605,10 +643,11 @@ static void sdma_v3_0_enable(struct amdgpu_device *adev, bool enable) static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev) { struct amdgpu_ring *ring; - u32 rb_cntl, ib_cntl; + u32 rb_cntl, ib_cntl, wptr_poll_cntl; u32 rb_bufsz; u32 wb_offset; u32 doorbell; + u64 wptr_gpu_addr; int i, j, r; for (i = 0; i < adev->sdma.num_instances; i++) { @@ -671,6 +710,20 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev) } WREG32(mmSDMA0_GFX_DOORBELL + sdma_offsets[i], doorbell); + /* setup the wptr shadow polling */ + wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + + WREG32(mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO + sdma_offsets[i], + lower_32_bits(wptr_gpu_addr)); + WREG32(mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI + sdma_offsets[i], + upper_32_bits(wptr_gpu_addr)); + wptr_poll_cntl = RREG32(mmSDMA0_GFX_RB_WPTR_POLL_CNTL + sdma_offsets[i]); + if (amdgpu_sriov_vf(adev)) + wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, SDMA0_GFX_RB_WPTR_POLL_CNTL, F32_POLL_ENABLE, 1); + else + wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, SDMA0_GFX_RB_WPTR_POLL_CNTL, F32_POLL_ENABLE, 0); + WREG32(mmSDMA0_GFX_RB_WPTR_POLL_CNTL + sdma_offsets[i], wptr_poll_cntl); + /* enable DMA RB */ rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1); WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i], rb_cntl); @@ -766,23 +819,12 @@ static int sdma_v3_0_load_microcode(struct amdgpu_device *adev) */ static int sdma_v3_0_start(struct amdgpu_device *adev) { - int r, i; + int r; - if (!adev->pp_enabled) { - if (adev->firmware.load_type != AMDGPU_FW_LOAD_SMU) { - r = sdma_v3_0_load_microcode(adev); - if (r) - return r; - } else { - for (i = 0; i < adev->sdma.num_instances; i++) { - r = adev->smu.smumgr_funcs->check_fw_load_finish(adev, - (i == 0) ? - AMDGPU_UCODE_ID_SDMA0 : - AMDGPU_UCODE_ID_SDMA1); - if (r) - return -EINVAL; - } - } + if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { + r = sdma_v3_0_load_microcode(adev); + if (r) + return r; } /* disable sdma engine before programing it */ @@ -1677,11 +1719,11 @@ static void sdma_v3_0_emit_fill_buffer(struct amdgpu_ib *ib, } static const struct amdgpu_buffer_funcs sdma_v3_0_buffer_funcs = { - .copy_max_bytes = 0x1fffff, + .copy_max_bytes = 0x3fffe0, /* not 0x3fffff due to HW limitation */ .copy_num_dw = 7, .emit_copy_buffer = sdma_v3_0_emit_copy_buffer, - .fill_max_bytes = 0x1fffff, + .fill_max_bytes = 0x3fffe0, /* not 0x3fffff due to HW limitation */ .fill_num_dw = 5, .emit_fill_buffer = sdma_v3_0_emit_fill_buffer, }; @@ -1695,8 +1737,14 @@ static void sdma_v3_0_set_buffer_funcs(struct amdgpu_device *adev) } static const struct amdgpu_vm_pte_funcs sdma_v3_0_vm_pte_funcs = { + .copy_pte_num_dw = 7, .copy_pte = sdma_v3_0_vm_copy_pte, + .write_pte = sdma_v3_0_vm_write_pte, + + /* not 0x3fffff due to HW limitation */ + .set_max_nums_pte_pde = 0x3fffe0 >> 3, + .set_pte_pde_num_dw = 10, .set_pte_pde = sdma_v3_0_vm_set_pte_pde, }; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 4a65697ccc94..46009db3d195 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -54,7 +54,7 @@ static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev); static void sdma_v4_0_set_irq_funcs(struct amdgpu_device *adev); static const u32 golden_settings_sdma_4[] = { - SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831f07, + SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831d07, SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CLK_CTRL), 0xff000ff0, 0x3f000100, SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_IB_CNTL), 0x800f0100, 0x00000100, SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_RB_WPTR_POLL_CNTL), 0xfffffff7, 0x00403000, @@ -89,7 +89,7 @@ static const u32 golden_settings_sdma_vg10[] = { static const u32 golden_settings_sdma_4_1[] = { - SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831f07, + SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831d07, SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CLK_CTRL), 0xffffffff, 0x3f000100, SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_IB_CNTL), 0x800f0111, 0x00000100, SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_RB_WPTR_POLL_CNTL), 0xfffffff7, 0x00403000, @@ -291,6 +291,8 @@ static void sdma_v4_0_ring_set_wptr(struct amdgpu_ring *ring) DRM_DEBUG("Setting write pointer\n"); if (ring->use_doorbell) { + u64 *wb = (u64 *)&adev->wb.wb[ring->wptr_offs]; + DRM_DEBUG("Using doorbell -- " "wptr_offs == 0x%08x " "lower_32_bits(ring->wptr) << 2 == 0x%08x " @@ -299,8 +301,7 @@ static void sdma_v4_0_ring_set_wptr(struct amdgpu_ring *ring) lower_32_bits(ring->wptr << 2), upper_32_bits(ring->wptr << 2)); /* XXX check if swapping is necessary on BE */ - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr << 2); - adev->wb.wb[ring->wptr_offs + 1] = upper_32_bits(ring->wptr << 2); + WRITE_ONCE(*wb, (ring->wptr << 2)); DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n", ring->doorbell_index, ring->wptr << 2); WDOORBELL64(ring->doorbell_index, ring->wptr << 2); @@ -370,7 +371,7 @@ static void sdma_v4_0_ring_emit_ib(struct amdgpu_ring *ring, static void sdma_v4_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) { u32 ref_and_mask = 0; - struct nbio_hdp_flush_reg *nbio_hf_reg; + const struct nbio_hdp_flush_reg *nbio_hf_reg; if (ring->adev->flags & AMD_IS_APU) nbio_hf_reg = &nbio_v7_0_hdp_flush_reg; @@ -397,7 +398,7 @@ static void sdma_v4_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) { amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) | SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf)); - amdgpu_ring_write(ring, SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0)); + amdgpu_ring_write(ring, SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE)); amdgpu_ring_write(ring, 1); } @@ -493,13 +494,45 @@ static void sdma_v4_0_rlc_stop(struct amdgpu_device *adev) */ static void sdma_v4_0_ctx_switch_enable(struct amdgpu_device *adev, bool enable) { - u32 f32_cntl; + u32 f32_cntl, phase_quantum = 0; int i; + if (amdgpu_sdma_phase_quantum) { + unsigned value = amdgpu_sdma_phase_quantum; + unsigned unit = 0; + + while (value > (SDMA0_PHASE0_QUANTUM__VALUE_MASK >> + SDMA0_PHASE0_QUANTUM__VALUE__SHIFT)) { + value = (value + 1) >> 1; + unit++; + } + if (unit > (SDMA0_PHASE0_QUANTUM__UNIT_MASK >> + SDMA0_PHASE0_QUANTUM__UNIT__SHIFT)) { + value = (SDMA0_PHASE0_QUANTUM__VALUE_MASK >> + SDMA0_PHASE0_QUANTUM__VALUE__SHIFT); + unit = (SDMA0_PHASE0_QUANTUM__UNIT_MASK >> + SDMA0_PHASE0_QUANTUM__UNIT__SHIFT); + WARN_ONCE(1, + "clamping sdma_phase_quantum to %uK clock cycles\n", + value << unit); + } + phase_quantum = + value << SDMA0_PHASE0_QUANTUM__VALUE__SHIFT | + unit << SDMA0_PHASE0_QUANTUM__UNIT__SHIFT; + } + for (i = 0; i < adev->sdma.num_instances; i++) { f32_cntl = RREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_CNTL)); f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL, AUTO_CTXSW_ENABLE, enable ? 1 : 0); + if (enable && amdgpu_sdma_phase_quantum) { + WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_PHASE0_QUANTUM), + phase_quantum); + WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_PHASE1_QUANTUM), + phase_quantum); + WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_PHASE2_QUANTUM), + phase_quantum); + } WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_CNTL), f32_cntl); } @@ -541,12 +574,13 @@ static void sdma_v4_0_enable(struct amdgpu_device *adev, bool enable) static int sdma_v4_0_gfx_resume(struct amdgpu_device *adev) { struct amdgpu_ring *ring; - u32 rb_cntl, ib_cntl; + u32 rb_cntl, ib_cntl, wptr_poll_cntl; u32 rb_bufsz; u32 wb_offset; u32 doorbell; u32 doorbell_offset; u32 temp; + u64 wptr_gpu_addr; int i, r; for (i = 0; i < adev->sdma.num_instances; i++) { @@ -628,6 +662,19 @@ static int sdma_v4_0_gfx_resume(struct amdgpu_device *adev) WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_F32_CNTL), temp); } + /* setup the wptr shadow polling */ + wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO), + lower_32_bits(wptr_gpu_addr)); + WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI), + upper_32_bits(wptr_gpu_addr)); + wptr_poll_cntl = RREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL)); + if (amdgpu_sriov_vf(adev)) + wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, SDMA0_GFX_RB_WPTR_POLL_CNTL, F32_POLL_ENABLE, 1); + else + wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, SDMA0_GFX_RB_WPTR_POLL_CNTL, F32_POLL_ENABLE, 0); + WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL), wptr_poll_cntl); + /* enable DMA RB */ rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1); WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_GFX_RB_CNTL), rb_cntl); @@ -655,6 +702,7 @@ static int sdma_v4_0_gfx_resume(struct amdgpu_device *adev) if (adev->mman.buffer_funcs_ring == ring) amdgpu_ttm_set_active_vram_size(adev, adev->mc.real_vram_size); + } return 0; @@ -751,15 +799,12 @@ static int sdma_v4_0_load_microcode(struct amdgpu_device *adev) const struct sdma_firmware_header_v1_0 *hdr; const __le32 *fw_data; u32 fw_size; - u32 digest_size = 0; int i, j; /* halt the MEs */ sdma_v4_0_enable(adev, false); for (i = 0; i < adev->sdma.num_instances; i++) { - uint16_t version_major; - uint16_t version_minor; if (!adev->sdma.instance[i].fw) return -EINVAL; @@ -767,23 +812,12 @@ static int sdma_v4_0_load_microcode(struct amdgpu_device *adev) amdgpu_ucode_print_sdma_hdr(&hdr->header); fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4; - version_major = le16_to_cpu(hdr->header.header_version_major); - version_minor = le16_to_cpu(hdr->header.header_version_minor); - - if (version_major == 1 && version_minor >= 1) { - const struct sdma_firmware_header_v1_1 *sdma_v1_1_hdr = (const struct sdma_firmware_header_v1_1 *) hdr; - digest_size = le32_to_cpu(sdma_v1_1_hdr->digest_size); - } - - fw_size -= digest_size; - fw_data = (const __le32 *) (adev->sdma.instance[i].fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_UCODE_ADDR), 0); - for (j = 0; j < fw_size; j++) WREG32(sdma_v4_0_get_reg_offset(i, mmSDMA0_UCODE_DATA), le32_to_cpup(fw_data++)); @@ -1230,6 +1264,11 @@ static int sdma_v4_0_sw_fini(void *handle) for (i = 0; i < adev->sdma.num_instances; i++) amdgpu_ring_fini(&adev->sdma.instance[i].ring); + for (i = 0; i < adev->sdma.num_instances; i++) { + release_firmware(adev->sdma.instance[i].fw); + adev->sdma.instance[i].fw = NULL; + } + return 0; } @@ -1680,8 +1719,13 @@ static void sdma_v4_0_set_buffer_funcs(struct amdgpu_device *adev) } static const struct amdgpu_vm_pte_funcs sdma_v4_0_vm_pte_funcs = { + .copy_pte_num_dw = 7, .copy_pte = sdma_v4_0_vm_copy_pte, + .write_pte = sdma_v4_0_vm_write_pte, + + .set_max_nums_pte_pde = 0x400000 >> 3, + .set_pte_pde_num_dw = 10, .set_pte_pde = sdma_v4_0_vm_set_pte_pde, }; diff --git a/drivers/gpu/drm/amd/amdgpu/si.c b/drivers/gpu/drm/amd/amdgpu/si.c index f45fb0f022b3..8284d5dbfc30 100644 --- a/drivers/gpu/drm/amd/amdgpu/si.c +++ b/drivers/gpu/drm/amd/amdgpu/si.c @@ -1150,6 +1150,33 @@ static bool si_read_disabled_bios(struct amdgpu_device *adev) return r; } +#define mmROM_INDEX 0x2A +#define mmROM_DATA 0x2B + +static bool si_read_bios_from_rom(struct amdgpu_device *adev, + u8 *bios, u32 length_bytes) +{ + u32 *dw_ptr; + u32 i, length_dw; + + if (bios == NULL) + return false; + if (length_bytes == 0) + return false; + /* APU vbios image is part of sbios image */ + if (adev->flags & AMD_IS_APU) + return false; + + dw_ptr = (u32 *)bios; + length_dw = ALIGN(length_bytes, 4) / 4; + /* set rom index to 0 */ + WREG32(mmROM_INDEX, 0); + for (i = 0; i < length_dw; i++) + dw_ptr[i] = RREG32(mmROM_DATA); + + return true; +} + //xxx: not implemented static int si_asic_reset(struct amdgpu_device *adev) { @@ -1206,6 +1233,7 @@ static void si_detect_hw_virtualization(struct amdgpu_device *adev) static const struct amdgpu_asic_funcs si_asic_funcs = { .read_disabled_bios = &si_read_disabled_bios, + .read_bios_from_rom = &si_read_bios_from_rom, .read_register = &si_read_register, .reset = &si_asic_reset, .set_vga_state = &si_vga_set_state, @@ -1385,6 +1413,7 @@ static void si_init_golden_registers(struct amdgpu_device *adev) amdgpu_program_register_sequence(adev, pitcairn_mgcg_cgcg_init, (const u32)ARRAY_SIZE(pitcairn_mgcg_cgcg_init)); + break; case CHIP_VERDE: amdgpu_program_register_sequence(adev, verde_golden_registers, @@ -1409,6 +1438,7 @@ static void si_init_golden_registers(struct amdgpu_device *adev) amdgpu_program_register_sequence(adev, oland_mgcg_cgcg_init, (const u32)ARRAY_SIZE(oland_mgcg_cgcg_init)); + break; case CHIP_HAINAN: amdgpu_program_register_sequence(adev, hainan_golden_registers, diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c index 112969f3301a..3fa2fbf8c9a1 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dma.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c @@ -887,8 +887,13 @@ static void si_dma_set_buffer_funcs(struct amdgpu_device *adev) } static const struct amdgpu_vm_pte_funcs si_dma_vm_pte_funcs = { + .copy_pte_num_dw = 5, .copy_pte = si_dma_vm_copy_pte, + .write_pte = si_dma_vm_write_pte, + + .set_max_nums_pte_pde = 0xffff8 >> 3, + .set_pte_pde_num_dw = 9, .set_pte_pde = si_dma_vm_set_pte_pde, }; diff --git a/drivers/gpu/drm/amd/amdgpu/si_dpm.c b/drivers/gpu/drm/amd/amdgpu/si_dpm.c index a7ad8390981c..51fd0c9a20a5 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dpm.c @@ -1847,7 +1847,6 @@ static int si_calculate_sclk_params(struct amdgpu_device *adev, static void si_thermal_start_smc_fan_control(struct amdgpu_device *adev); static void si_fan_ctrl_set_default_mode(struct amdgpu_device *adev); -static void si_dpm_set_dpm_funcs(struct amdgpu_device *adev); static void si_dpm_set_irq_funcs(struct amdgpu_device *adev); static struct si_power_info *si_get_pi(struct amdgpu_device *adev) @@ -2055,6 +2054,7 @@ static void si_initialize_powertune_defaults(struct amdgpu_device *adev) case 0x682C: si_pi->cac_weights = cac_weights_cape_verde_pro; si_pi->dte_data = dte_data_sun_xt; + update_dte_from_pl2 = true; break; case 0x6825: case 0x6827: @@ -3059,9 +3059,9 @@ static int si_get_vce_clock_voltage(struct amdgpu_device *adev, return ret; } -static bool si_dpm_vblank_too_short(struct amdgpu_device *adev) +static bool si_dpm_vblank_too_short(void *handle) { - + struct amdgpu_device *adev = (struct amdgpu_device *)handle; u32 vblank_time = amdgpu_dpm_get_vblank_time(adev); /* we never hit the non-gddr5 limit so disable it */ u32 switch_limit = adev->mc.vram_type == AMDGPU_VRAM_TYPE_GDDR5 ? 450 : 0; @@ -3870,9 +3870,10 @@ static int si_restrict_performance_levels_before_switch(struct amdgpu_device *ad 0 : -EINVAL; } -static int si_dpm_force_performance_level(struct amdgpu_device *adev, +static int si_dpm_force_performance_level(void *handle, enum amd_dpm_forced_level level) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_ps *rps = adev->pm.dpm.current_ps; struct si_ps *ps = si_get_ps(rps); u32 levels = ps->performance_level_count; @@ -6574,11 +6575,12 @@ static int si_fan_ctrl_stop_smc_fan_control(struct amdgpu_device *adev) } } -static int si_dpm_get_fan_speed_percent(struct amdgpu_device *adev, +static int si_dpm_get_fan_speed_percent(void *handle, u32 *speed) { u32 duty, duty100; u64 tmp64; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (adev->pm.no_fan) return -ENOENT; @@ -6599,9 +6601,10 @@ static int si_dpm_get_fan_speed_percent(struct amdgpu_device *adev, return 0; } -static int si_dpm_set_fan_speed_percent(struct amdgpu_device *adev, +static int si_dpm_set_fan_speed_percent(void *handle, u32 speed) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct si_power_info *si_pi = si_get_pi(adev); u32 tmp; u32 duty, duty100; @@ -6632,8 +6635,10 @@ static int si_dpm_set_fan_speed_percent(struct amdgpu_device *adev, return 0; } -static void si_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode) +static void si_dpm_set_fan_control_mode(void *handle, u32 mode) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (mode) { /* stop auto-manage */ if (adev->pm.dpm.fan.ucode_fan_control) @@ -6648,8 +6653,9 @@ static void si_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode) } } -static u32 si_dpm_get_fan_control_mode(struct amdgpu_device *adev) +static u32 si_dpm_get_fan_control_mode(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct si_power_info *si_pi = si_get_pi(adev); u32 tmp; @@ -6945,8 +6951,9 @@ static void si_dpm_disable(struct amdgpu_device *adev) ni_update_current_ps(adev, boot_ps); } -static int si_dpm_pre_set_power_state(struct amdgpu_device *adev) +static int si_dpm_pre_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps; struct amdgpu_ps *new_ps = &requested_ps; @@ -6983,8 +6990,9 @@ static int si_power_control_set_level(struct amdgpu_device *adev) return 0; } -static int si_dpm_set_power_state(struct amdgpu_device *adev) +static int si_dpm_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct amdgpu_ps *new_ps = &eg_pi->requested_rps; struct amdgpu_ps *old_ps = &eg_pi->current_rps; @@ -7085,8 +7093,9 @@ static int si_dpm_set_power_state(struct amdgpu_device *adev) return 0; } -static void si_dpm_post_set_power_state(struct amdgpu_device *adev) +static void si_dpm_post_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct amdgpu_ps *new_ps = &eg_pi->requested_rps; @@ -7102,8 +7111,10 @@ void si_dpm_reset_asic(struct amdgpu_device *adev) } #endif -static void si_dpm_display_configuration_changed(struct amdgpu_device *adev) +static void si_dpm_display_configuration_changed(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + si_program_display_gap(adev); } @@ -7485,9 +7496,10 @@ static void si_dpm_fini(struct amdgpu_device *adev) amdgpu_free_extended_power_table(adev); } -static void si_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev, +static void si_dpm_debugfs_print_current_performance_level(void *handle, struct seq_file *m) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct amdgpu_ps *rps = &eg_pi->current_rps; struct si_ps *ps = si_get_ps(rps); @@ -7592,11 +7604,6 @@ static int si_dpm_late_init(void *handle) if (!amdgpu_dpm) return 0; - /* init the sysfs and debugfs files late */ - ret = amdgpu_pm_sysfs_init(adev); - if (ret) - return ret; - ret = si_set_temperature_range(adev); if (ret) return ret; @@ -7752,7 +7759,6 @@ static int si_dpm_sw_fini(void *handle) flush_work(&adev->pm.dpm.thermal.work); mutex_lock(&adev->pm.mutex); - amdgpu_pm_sysfs_fini(adev); si_dpm_fini(adev); mutex_unlock(&adev->pm.mutex); @@ -7859,10 +7865,11 @@ static int si_dpm_set_powergating_state(void *handle, } /* get temperature in millidegrees */ -static int si_dpm_get_temp(struct amdgpu_device *adev) +static int si_dpm_get_temp(void *handle) { u32 temp; int actual_temp = 0; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; temp = (RREG32(CG_MULT_THERMAL_STATUS) & CTF_TEMP_MASK) >> CTF_TEMP_SHIFT; @@ -7877,8 +7884,9 @@ static int si_dpm_get_temp(struct amdgpu_device *adev) return actual_temp; } -static u32 si_dpm_get_sclk(struct amdgpu_device *adev, bool low) +static u32 si_dpm_get_sclk(void *handle, bool low) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct si_ps *requested_state = si_get_ps(&eg_pi->requested_rps); @@ -7888,8 +7896,9 @@ static u32 si_dpm_get_sclk(struct amdgpu_device *adev, bool low) return requested_state->performance_levels[requested_state->performance_level_count - 1].sclk; } -static u32 si_dpm_get_mclk(struct amdgpu_device *adev, bool low) +static u32 si_dpm_get_mclk(void *handle, bool low) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct si_ps *requested_state = si_get_ps(&eg_pi->requested_rps); @@ -7899,9 +7908,11 @@ static u32 si_dpm_get_mclk(struct amdgpu_device *adev, bool low) return requested_state->performance_levels[requested_state->performance_level_count - 1].mclk; } -static void si_dpm_print_power_state(struct amdgpu_device *adev, - struct amdgpu_ps *rps) +static void si_dpm_print_power_state(void *handle, + void *current_ps) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_ps *rps = (struct amdgpu_ps *)current_ps; struct si_ps *ps = si_get_ps(rps); struct rv7xx_pl *pl; int i; @@ -7926,7 +7937,6 @@ static int si_dpm_early_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; - si_dpm_set_dpm_funcs(adev); si_dpm_set_irq_funcs(adev); return 0; } @@ -7941,20 +7951,23 @@ static inline bool si_are_power_levels_equal(const struct rv7xx_pl *si_cpl1, (si_cpl1->vddci == si_cpl2->vddci)); } -static int si_check_state_equal(struct amdgpu_device *adev, - struct amdgpu_ps *cps, - struct amdgpu_ps *rps, +static int si_check_state_equal(void *handle, + void *current_ps, + void *request_ps, bool *equal) { struct si_ps *si_cps; struct si_ps *si_rps; int i; + struct amdgpu_ps *cps = (struct amdgpu_ps *)current_ps; + struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (adev == NULL || cps == NULL || rps == NULL || equal == NULL) return -EINVAL; - si_cps = si_get_ps(cps); - si_rps = si_get_ps(rps); + si_cps = si_get_ps((struct amdgpu_ps *)cps); + si_rps = si_get_ps((struct amdgpu_ps *)rps); if (si_cps == NULL) { printk("si_cps is NULL\n"); @@ -7982,9 +7995,10 @@ static int si_check_state_equal(struct amdgpu_device *adev, return 0; } -static int si_dpm_read_sensor(struct amdgpu_device *adev, int idx, +static int si_dpm_read_sensor(void *handle, int idx, void *value, int *size) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct amdgpu_ps *rps = &eg_pi->current_rps; struct si_ps *ps = si_get_ps(rps); @@ -8040,7 +8054,7 @@ const struct amd_ip_funcs si_dpm_ip_funcs = { .set_powergating_state = si_dpm_set_powergating_state, }; -static const struct amdgpu_dpm_funcs si_dpm_funcs = { +const struct amd_pm_funcs si_dpm_funcs = { .get_temperature = &si_dpm_get_temp, .pre_set_power_state = &si_dpm_pre_set_power_state, .set_power_state = &si_dpm_set_power_state, @@ -8061,12 +8075,6 @@ static const struct amdgpu_dpm_funcs si_dpm_funcs = { .read_sensor = &si_dpm_read_sensor, }; -static void si_dpm_set_dpm_funcs(struct amdgpu_device *adev) -{ - if (adev->pm.funcs == NULL) - adev->pm.funcs = &si_dpm_funcs; -} - static const struct amdgpu_irq_src_funcs si_dpm_irq_funcs = { .set = si_dpm_set_interrupt_state, .process = si_dpm_process_interrupt, diff --git a/drivers/gpu/drm/amd/amdgpu/si_dpm.h b/drivers/gpu/drm/amd/amdgpu/si_dpm.h index 51ce21c5f4fb..9fe343de3477 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/si_dpm.h @@ -246,6 +246,7 @@ enum si_display_gap }; extern const struct amd_ip_funcs si_dpm_ip_funcs; +extern const struct amd_pm_funcs si_dpm_funcs; struct ni_leakage_coeffients { diff --git a/drivers/gpu/drm/amd/amdgpu/si_ih.c b/drivers/gpu/drm/amd/amdgpu/si_ih.c index ce25e03a077d..d2c6b80309c8 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/si_ih.c @@ -118,6 +118,19 @@ static u32 si_ih_get_wptr(struct amdgpu_device *adev) return (wptr & adev->irq.ih.ptr_mask); } +/** + * si_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool si_ih_prescreen_iv(struct amdgpu_device *adev) +{ + /* Process all interrupts */ + return true; +} + static void si_ih_decode_iv(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { @@ -288,6 +301,7 @@ static const struct amd_ip_funcs si_ih_ip_funcs = { static const struct amdgpu_ih_funcs si_ih_funcs = { .get_wptr = si_ih_get_wptr, + .prescreen_iv = si_ih_prescreen_iv, .decode_iv = si_ih_decode_iv, .set_rptr = si_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index a7341d88a320..4e67fe1e7955 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -25,7 +25,7 @@ #include <linux/module.h> #include <drm/drmP.h> #include "amdgpu.h" -#include "amdgpu_atomfirmware.h" +#include "amdgpu_atombios.h" #include "amdgpu_ih.h" #include "amdgpu_uvd.h" #include "amdgpu_vce.h" @@ -62,8 +62,6 @@ #include "dce_virtual.h" #include "mxgpu_ai.h" -MODULE_FIRMWARE("amdgpu/vega10_smc.bin"); - #define mmFabricConfigAccessControl 0x0410 #define mmFabricConfigAccessControl_BASE_IDX 0 #define mmFabricConfigAccessControl_DEFAULT 0x00000000 @@ -103,7 +101,7 @@ static u32 soc15_pcie_rreg(struct amdgpu_device *adev, u32 reg) { unsigned long flags, address, data; u32 r; - struct nbio_pcie_index_data *nbio_pcie_id; + const struct nbio_pcie_index_data *nbio_pcie_id; if (adev->flags & AMD_IS_APU) nbio_pcie_id = &nbio_v7_0_pcie_index_data; @@ -124,7 +122,7 @@ static u32 soc15_pcie_rreg(struct amdgpu_device *adev, u32 reg) static void soc15_pcie_wreg(struct amdgpu_device *adev, u32 reg, u32 v) { unsigned long flags, address, data; - struct nbio_pcie_index_data *nbio_pcie_id; + const struct nbio_pcie_index_data *nbio_pcie_id; if (adev->flags & AMD_IS_APU) nbio_pcie_id = &nbio_v7_0_pcie_index_data; @@ -198,6 +196,50 @@ static void soc15_didt_wreg(struct amdgpu_device *adev, u32 reg, u32 v) spin_unlock_irqrestore(&adev->didt_idx_lock, flags); } +static u32 soc15_gc_cac_rreg(struct amdgpu_device *adev, u32 reg) +{ + unsigned long flags; + u32 r; + + spin_lock_irqsave(&adev->gc_cac_idx_lock, flags); + WREG32_SOC15(GC, 0, mmGC_CAC_IND_INDEX, (reg)); + r = RREG32_SOC15(GC, 0, mmGC_CAC_IND_DATA); + spin_unlock_irqrestore(&adev->gc_cac_idx_lock, flags); + return r; +} + +static void soc15_gc_cac_wreg(struct amdgpu_device *adev, u32 reg, u32 v) +{ + unsigned long flags; + + spin_lock_irqsave(&adev->gc_cac_idx_lock, flags); + WREG32_SOC15(GC, 0, mmGC_CAC_IND_INDEX, (reg)); + WREG32_SOC15(GC, 0, mmGC_CAC_IND_DATA, (v)); + spin_unlock_irqrestore(&adev->gc_cac_idx_lock, flags); +} + +static u32 soc15_se_cac_rreg(struct amdgpu_device *adev, u32 reg) +{ + unsigned long flags; + u32 r; + + spin_lock_irqsave(&adev->se_cac_idx_lock, flags); + WREG32_SOC15(GC, 0, mmSE_CAC_IND_INDEX, (reg)); + r = RREG32_SOC15(GC, 0, mmSE_CAC_IND_DATA); + spin_unlock_irqrestore(&adev->se_cac_idx_lock, flags); + return r; +} + +static void soc15_se_cac_wreg(struct amdgpu_device *adev, u32 reg, u32 v) +{ + unsigned long flags; + + spin_lock_irqsave(&adev->se_cac_idx_lock, flags); + WREG32_SOC15(GC, 0, mmSE_CAC_IND_INDEX, (reg)); + WREG32_SOC15(GC, 0, mmSE_CAC_IND_DATA, (v)); + spin_unlock_irqrestore(&adev->se_cac_idx_lock, flags); +} + static u32 soc15_get_config_memsize(struct amdgpu_device *adev) { if (adev->flags & AMD_IS_APU) @@ -237,10 +279,7 @@ static void soc15_init_golden_registers(struct amdgpu_device *adev) } static u32 soc15_get_xclk(struct amdgpu_device *adev) { - if (adev->asic_type == CHIP_VEGA10) - return adev->clock.spll.reference_freq/4; - else - return adev->clock.spll.reference_freq; + return adev->clock.spll.reference_freq; } @@ -365,18 +404,27 @@ static int soc15_read_register(struct amdgpu_device *adev, u32 se_num, return -EINVAL; } -static void soc15_gpu_pci_config_reset(struct amdgpu_device *adev) +static int soc15_asic_reset(struct amdgpu_device *adev) { u32 i; - dev_info(adev->dev, "GPU pci config reset\n"); + amdgpu_atombios_scratch_regs_engine_hung(adev, true); + + dev_info(adev->dev, "GPU reset\n"); /* disable BM */ pci_clear_master(adev->pdev); - /* reset */ - amdgpu_pci_config_reset(adev); - udelay(100); + pci_save_state(adev->pdev); + + for (i = 0; i < AMDGPU_MAX_IP_NUM; i++) { + if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP){ + adev->ip_blocks[i].version->funcs->soft_reset((void *)adev); + break; + } + } + + pci_restore_state(adev->pdev); /* wait for asic to come out of reset */ for (i = 0; i < adev->usec_timeout; i++) { @@ -388,15 +436,7 @@ static void soc15_gpu_pci_config_reset(struct amdgpu_device *adev) udelay(1); } -} - -static int soc15_asic_reset(struct amdgpu_device *adev) -{ - amdgpu_atomfirmware_scratch_regs_engine_hung(adev, true); - - soc15_gpu_pci_config_reset(adev); - - amdgpu_atomfirmware_scratch_regs_engine_hung(adev, false); + amdgpu_atombios_scratch_regs_engine_hung(adev, false); return 0; } @@ -492,6 +532,12 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) amdgpu_ip_block_add(adev, &amdgpu_pp_ip_block); if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) amdgpu_ip_block_add(adev, &dce_virtual_ip_block); +#if defined(CONFIG_DRM_AMD_DC) + else if (amdgpu_device_has_dc_support(adev)) + amdgpu_ip_block_add(adev, &dm_ip_block); +#else +# warning "Enable CONFIG_DRM_AMD_DC for display support on SOC15." +#endif amdgpu_ip_block_add(adev, &gfx_v9_0_ip_block); amdgpu_ip_block_add(adev, &sdma_v4_0_ip_block); amdgpu_ip_block_add(adev, &uvd_v7_0_ip_block); @@ -505,6 +551,12 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) amdgpu_ip_block_add(adev, &amdgpu_pp_ip_block); if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) amdgpu_ip_block_add(adev, &dce_virtual_ip_block); +#if defined(CONFIG_DRM_AMD_DC) + else if (amdgpu_device_has_dc_support(adev)) + amdgpu_ip_block_add(adev, &dm_ip_block); +#else +# warning "Enable CONFIG_DRM_AMD_DC for display support on SOC15." +#endif amdgpu_ip_block_add(adev, &gfx_v9_0_ip_block); amdgpu_ip_block_add(adev, &sdma_v4_0_ip_block); amdgpu_ip_block_add(adev, &vcn_v1_0_ip_block); @@ -524,13 +576,6 @@ static uint32_t soc15_get_rev_id(struct amdgpu_device *adev) return nbio_v6_1_get_rev_id(adev); } - -int gmc_v9_0_mc_wait_for_idle(struct amdgpu_device *adev) -{ - /* to be implemented in MC IP*/ - return 0; -} - static const struct amdgpu_asic_funcs soc15_asic_funcs = { .read_disabled_bios = &soc15_read_disabled_bios, @@ -557,6 +602,10 @@ static int soc15_common_early_init(void *handle) adev->uvd_ctx_wreg = &soc15_uvd_ctx_wreg; adev->didt_rreg = &soc15_didt_rreg; adev->didt_wreg = &soc15_didt_wreg; + adev->gc_cac_rreg = &soc15_gc_cac_rreg; + adev->gc_cac_wreg = &soc15_gc_cac_wreg; + adev->se_cac_rreg = &soc15_se_cac_rreg; + adev->se_cac_wreg = &soc15_se_cac_wreg; adev->asic_funcs = &soc15_asic_funcs; @@ -564,21 +613,6 @@ static int soc15_common_early_init(void *handle) (amdgpu_ip_block_mask & (1 << AMD_IP_BLOCK_TYPE_PSP))) psp_enabled = true; - /* - * nbio need be used for both sdma and gfx9, but only - * initializes once - */ - switch(adev->asic_type) { - case CHIP_VEGA10: - nbio_v6_1_init(adev); - break; - case CHIP_RAVEN: - nbio_v7_0_init(adev); - break; - default: - return -EINVAL; - } - adev->rev_id = soc15_get_rev_id(adev); adev->external_rev_id = 0xFF; switch (adev->asic_type) { @@ -681,6 +715,9 @@ static int soc15_common_hw_init(void *handle) soc15_pcie_gen3_enable(adev); /* enable aspm */ soc15_program_aspm(adev); + /* setup nbio registers */ + if (!(adev->flags & AMD_IS_APU)) + nbio_v6_1_init_registers(adev); /* enable the doorbell aperture */ soc15_enable_doorbell_aperture(adev, true); diff --git a/drivers/gpu/drm/amd/amdgpu/soc15_common.h b/drivers/gpu/drm/amd/amdgpu/soc15_common.h index e2d330eed952..7a8e4e28abb2 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15_common.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15_common.h @@ -77,6 +77,13 @@ struct nbio_pcie_index_data { (3 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG3 + reg : \ (ip##_BASE__INST##inst##_SEG4 + reg))))), value) +#define WREG32_SOC15_NO_KIQ(ip, inst, reg, value) \ + WREG32_NO_KIQ( (0 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG0 + reg : \ + (1 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG1 + reg : \ + (2 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG2 + reg : \ + (3 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG3 + reg : \ + (ip##_BASE__INST##inst##_SEG4 + reg))))), value) + #define WREG32_SOC15_OFFSET(ip, inst, reg, offset, value) \ WREG32( (0 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG0 + reg : \ (1 == reg##_BASE_IDX ? ip##_BASE__INST##inst##_SEG1 + reg : \ diff --git a/drivers/gpu/drm/amd/amdgpu/soc15d.h b/drivers/gpu/drm/amd/amdgpu/soc15d.h index e79befd80eed..7f408f85fdb6 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15d.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15d.h @@ -250,6 +250,7 @@ #define PACKET3_SET_UCONFIG_REG 0x79 #define PACKET3_SET_UCONFIG_REG_START 0x0000c000 #define PACKET3_SET_UCONFIG_REG_END 0x0000c400 +#define PACKET3_SET_UCONFIG_REG_INDEX_TYPE (2 << 28) #define PACKET3_SCRATCH_RAM_WRITE 0x7D #define PACKET3_SCRATCH_RAM_READ 0x7E #define PACKET3_LOAD_CONST_RAM 0x80 diff --git a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c index 923df2c0e535..aa4e320e31f8 100644 --- a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c @@ -219,6 +219,34 @@ static u32 tonga_ih_get_wptr(struct amdgpu_device *adev) } /** + * tonga_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool tonga_ih_prescreen_iv(struct amdgpu_device *adev) +{ + u32 ring_index = adev->irq.ih.rptr >> 2; + u16 pasid; + + switch (le32_to_cpu(adev->irq.ih.ring[ring_index]) & 0xff) { + case 146: + case 147: + pasid = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]) >> 16; + if (!pasid || amdgpu_vm_pasid_fault_credit(adev, pasid)) + return true; + break; + default: + /* Not a VM fault */ + return true; + } + + adev->irq.ih.rptr += 16; + return false; +} + +/** * tonga_ih_decode_iv - decode an interrupt vector * * @adev: amdgpu_device pointer @@ -478,6 +506,7 @@ static const struct amd_ip_funcs tonga_ih_ip_funcs = { static const struct amdgpu_ih_funcs tonga_ih_funcs = { .get_wptr = tonga_ih_get_wptr, + .prescreen_iv = tonga_ih_prescreen_iv, .decode_iv = tonga_ih_decode_iv, .set_rptr = tonga_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c index 31db356476f8..920910ac8663 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c @@ -38,6 +38,8 @@ #include "vi.h" static void uvd_v6_0_set_ring_funcs(struct amdgpu_device *adev); +static void uvd_v6_0_set_enc_ring_funcs(struct amdgpu_device *adev); + static void uvd_v6_0_set_irq_funcs(struct amdgpu_device *adev); static int uvd_v6_0_start(struct amdgpu_device *adev); static void uvd_v6_0_stop(struct amdgpu_device *adev); @@ -48,6 +50,18 @@ static void uvd_v6_0_enable_mgcg(struct amdgpu_device *adev, bool enable); /** +* uvd_v6_0_enc_support - get encode support status +* +* @adev: amdgpu_device pointer +* +* Returns the current hardware encode support status +*/ +static inline bool uvd_v6_0_enc_support(struct amdgpu_device *adev) +{ + return ((adev->asic_type >= CHIP_POLARIS10) && (adev->asic_type <= CHIP_POLARIS12)); +} + +/** * uvd_v6_0_ring_get_rptr - get read pointer * * @ring: amdgpu_ring pointer @@ -62,6 +76,22 @@ static uint64_t uvd_v6_0_ring_get_rptr(struct amdgpu_ring *ring) } /** + * uvd_v6_0_enc_ring_get_rptr - get enc read pointer + * + * @ring: amdgpu_ring pointer + * + * Returns the current hardware enc read pointer + */ +static uint64_t uvd_v6_0_enc_ring_get_rptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + if (ring == &adev->uvd.ring_enc[0]) + return RREG32(mmUVD_RB_RPTR); + else + return RREG32(mmUVD_RB_RPTR2); +} +/** * uvd_v6_0_ring_get_wptr - get write pointer * * @ring: amdgpu_ring pointer @@ -76,6 +106,23 @@ static uint64_t uvd_v6_0_ring_get_wptr(struct amdgpu_ring *ring) } /** + * uvd_v6_0_enc_ring_get_wptr - get enc write pointer + * + * @ring: amdgpu_ring pointer + * + * Returns the current hardware enc write pointer + */ +static uint64_t uvd_v6_0_enc_ring_get_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + if (ring == &adev->uvd.ring_enc[0]) + return RREG32(mmUVD_RB_WPTR); + else + return RREG32(mmUVD_RB_WPTR2); +} + +/** * uvd_v6_0_ring_set_wptr - set write pointer * * @ring: amdgpu_ring pointer @@ -89,11 +136,252 @@ static void uvd_v6_0_ring_set_wptr(struct amdgpu_ring *ring) WREG32(mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr)); } +/** + * uvd_v6_0_enc_ring_set_wptr - set enc write pointer + * + * @ring: amdgpu_ring pointer + * + * Commits the enc write pointer to the hardware + */ +static void uvd_v6_0_enc_ring_set_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + if (ring == &adev->uvd.ring_enc[0]) + WREG32(mmUVD_RB_WPTR, + lower_32_bits(ring->wptr)); + else + WREG32(mmUVD_RB_WPTR2, + lower_32_bits(ring->wptr)); +} + +/** + * uvd_v6_0_enc_ring_test_ring - test if UVD ENC ring is working + * + * @ring: the engine to test on + * + */ +static int uvd_v6_0_enc_ring_test_ring(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t rptr = amdgpu_ring_get_rptr(ring); + unsigned i; + int r; + + r = amdgpu_ring_alloc(ring, 16); + if (r) { + DRM_ERROR("amdgpu: uvd enc failed to lock ring %d (%d).\n", + ring->idx, r); + return r; + } + amdgpu_ring_write(ring, HEVC_ENC_CMD_END); + amdgpu_ring_commit(ring); + + for (i = 0; i < adev->usec_timeout; i++) { + if (amdgpu_ring_get_rptr(ring) != rptr) + break; + DRM_UDELAY(1); + } + + if (i < adev->usec_timeout) { + DRM_INFO("ring test on %d succeeded in %d usecs\n", + ring->idx, i); + } else { + DRM_ERROR("amdgpu: ring %d test failed\n", + ring->idx); + r = -ETIMEDOUT; + } + + return r; +} + +/** + * uvd_v6_0_enc_get_create_msg - generate a UVD ENC create msg + * + * @adev: amdgpu_device pointer + * @ring: ring we should submit the msg to + * @handle: session handle to use + * @fence: optional fence to return + * + * Open up a stream for HW test + */ +static int uvd_v6_0_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, + struct dma_fence **fence) +{ + const unsigned ib_size_dw = 16; + struct amdgpu_job *job; + struct amdgpu_ib *ib; + struct dma_fence *f = NULL; + uint64_t dummy; + int i, r; + + r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, &job); + if (r) + return r; + + ib = &job->ibs[0]; + dummy = ib->gpu_addr + 1024; + + ib->length_dw = 0; + ib->ptr[ib->length_dw++] = 0x00000018; + ib->ptr[ib->length_dw++] = 0x00000001; /* session info */ + ib->ptr[ib->length_dw++] = handle; + ib->ptr[ib->length_dw++] = 0x00010000; + ib->ptr[ib->length_dw++] = upper_32_bits(dummy); + ib->ptr[ib->length_dw++] = dummy; + + ib->ptr[ib->length_dw++] = 0x00000014; + ib->ptr[ib->length_dw++] = 0x00000002; /* task info */ + ib->ptr[ib->length_dw++] = 0x0000001c; + ib->ptr[ib->length_dw++] = 0x00000001; + ib->ptr[ib->length_dw++] = 0x00000000; + + ib->ptr[ib->length_dw++] = 0x00000008; + ib->ptr[ib->length_dw++] = 0x08000001; /* op initialize */ + + for (i = ib->length_dw; i < ib_size_dw; ++i) + ib->ptr[i] = 0x0; + + r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); + job->fence = dma_fence_get(f); + if (r) + goto err; + + amdgpu_job_free(job); + if (fence) + *fence = dma_fence_get(f); + dma_fence_put(f); + return 0; + +err: + amdgpu_job_free(job); + return r; +} + +/** + * uvd_v6_0_enc_get_destroy_msg - generate a UVD ENC destroy msg + * + * @adev: amdgpu_device pointer + * @ring: ring we should submit the msg to + * @handle: session handle to use + * @fence: optional fence to return + * + * Close up a stream for HW test or if userspace failed to do so + */ +static int uvd_v6_0_enc_get_destroy_msg(struct amdgpu_ring *ring, + uint32_t handle, + bool direct, struct dma_fence **fence) +{ + const unsigned ib_size_dw = 16; + struct amdgpu_job *job; + struct amdgpu_ib *ib; + struct dma_fence *f = NULL; + uint64_t dummy; + int i, r; + + r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, &job); + if (r) + return r; + + ib = &job->ibs[0]; + dummy = ib->gpu_addr + 1024; + + ib->length_dw = 0; + ib->ptr[ib->length_dw++] = 0x00000018; + ib->ptr[ib->length_dw++] = 0x00000001; /* session info */ + ib->ptr[ib->length_dw++] = handle; + ib->ptr[ib->length_dw++] = 0x00010000; + ib->ptr[ib->length_dw++] = upper_32_bits(dummy); + ib->ptr[ib->length_dw++] = dummy; + + ib->ptr[ib->length_dw++] = 0x00000014; + ib->ptr[ib->length_dw++] = 0x00000002; /* task info */ + ib->ptr[ib->length_dw++] = 0x0000001c; + ib->ptr[ib->length_dw++] = 0x00000001; + ib->ptr[ib->length_dw++] = 0x00000000; + + ib->ptr[ib->length_dw++] = 0x00000008; + ib->ptr[ib->length_dw++] = 0x08000002; /* op close session */ + + for (i = ib->length_dw; i < ib_size_dw; ++i) + ib->ptr[i] = 0x0; + + if (direct) { + r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); + job->fence = dma_fence_get(f); + if (r) + goto err; + + amdgpu_job_free(job); + } else { + r = amdgpu_job_submit(job, ring, &ring->adev->vce.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, &f); + if (r) + goto err; + } + + if (fence) + *fence = dma_fence_get(f); + dma_fence_put(f); + return 0; + +err: + amdgpu_job_free(job); + return r; +} + +/** + * uvd_v6_0_enc_ring_test_ib - test if UVD ENC IBs are working + * + * @ring: the engine to test on + * + */ +static int uvd_v6_0_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout) +{ + struct dma_fence *fence = NULL; + long r; + + r = uvd_v6_0_enc_get_create_msg(ring, 1, NULL); + if (r) { + DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r); + goto error; + } + + r = uvd_v6_0_enc_get_destroy_msg(ring, 1, true, &fence); + if (r) { + DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r); + goto error; + } + + r = dma_fence_wait_timeout(fence, false, timeout); + if (r == 0) { + DRM_ERROR("amdgpu: IB test timed out.\n"); + r = -ETIMEDOUT; + } else if (r < 0) { + DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); + } else { + DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + r = 0; + } +error: + dma_fence_put(fence); + return r; +} static int uvd_v6_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (!(adev->flags & AMD_IS_APU) && + (RREG32_SMC(ixCC_HARVEST_FUSES) & CC_HARVEST_FUSES__UVD_DISABLE_MASK)) + return -ENOENT; + uvd_v6_0_set_ring_funcs(adev); + + if (uvd_v6_0_enc_support(adev)) { + adev->uvd.num_enc_rings = 2; + uvd_v6_0_set_enc_ring_funcs(adev); + } + uvd_v6_0_set_irq_funcs(adev); return 0; @@ -102,7 +390,7 @@ static int uvd_v6_0_early_init(void *handle) static int uvd_v6_0_sw_init(void *handle) { struct amdgpu_ring *ring; - int r; + int i, r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* UVD TRAP */ @@ -110,10 +398,31 @@ static int uvd_v6_0_sw_init(void *handle) if (r) return r; + /* UVD ENC TRAP */ + if (uvd_v6_0_enc_support(adev)) { + for (i = 0; i < adev->uvd.num_enc_rings; ++i) { + r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 119, &adev->uvd.irq); + if (r) + return r; + } + } + r = amdgpu_uvd_sw_init(adev); if (r) return r; + if (uvd_v6_0_enc_support(adev)) { + struct amd_sched_rq *rq; + ring = &adev->uvd.ring_enc[0]; + rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL]; + r = amd_sched_entity_init(&ring->sched, &adev->uvd.entity_enc, + rq, amdgpu_sched_jobs); + if (r) { + DRM_ERROR("Failed setting up UVD ENC run queue.\n"); + return r; + } + } + r = amdgpu_uvd_resume(adev); if (r) return r; @@ -121,19 +430,38 @@ static int uvd_v6_0_sw_init(void *handle) ring = &adev->uvd.ring; sprintf(ring->name, "uvd"); r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0); + if (r) + return r; + + if (uvd_v6_0_enc_support(adev)) { + for (i = 0; i < adev->uvd.num_enc_rings; ++i) { + ring = &adev->uvd.ring_enc[i]; + sprintf(ring->name, "uvd_enc%d", i); + r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0); + if (r) + return r; + } + } return r; } static int uvd_v6_0_sw_fini(void *handle) { - int r; + int i, r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; r = amdgpu_uvd_suspend(adev); if (r) return r; + if (uvd_v6_0_enc_support(adev)) { + amd_sched_entity_fini(&adev->uvd.ring_enc[0].sched, &adev->uvd.entity_enc); + + for (i = 0; i < adev->uvd.num_enc_rings; ++i) + amdgpu_ring_fini(&adev->uvd.ring_enc[i]); + } + return amdgpu_uvd_sw_fini(adev); } @@ -149,7 +477,7 @@ static int uvd_v6_0_hw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_ring *ring = &adev->uvd.ring; uint32_t tmp; - int r; + int i, r; amdgpu_asic_set_uvd_clocks(adev, 10000, 10000); uvd_v6_0_set_clockgating_state(adev, AMD_CG_STATE_UNGATE); @@ -189,9 +517,25 @@ static int uvd_v6_0_hw_init(void *handle) amdgpu_ring_commit(ring); + if (uvd_v6_0_enc_support(adev)) { + for (i = 0; i < adev->uvd.num_enc_rings; ++i) { + ring = &adev->uvd.ring_enc[i]; + ring->ready = true; + r = amdgpu_ring_test_ring(ring); + if (r) { + ring->ready = false; + goto done; + } + } + } + done: - if (!r) - DRM_INFO("UVD initialized successfully.\n"); + if (!r) { + if (uvd_v6_0_enc_support(adev)) + DRM_INFO("UVD and UVD ENC initialized successfully.\n"); + else + DRM_INFO("UVD initialized successfully.\n"); + } return r; } @@ -225,11 +569,7 @@ static int uvd_v6_0_suspend(void *handle) if (r) return r; - /* Skip this for APU for now */ - if (!(adev->flags & AMD_IS_APU)) - r = amdgpu_uvd_suspend(adev); - - return r; + return amdgpu_uvd_suspend(adev); } static int uvd_v6_0_resume(void *handle) @@ -237,12 +577,10 @@ static int uvd_v6_0_resume(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - /* Skip this for APU for now */ - if (!(adev->flags & AMD_IS_APU)) { - r = amdgpu_uvd_resume(adev); - if (r) - return r; - } + r = amdgpu_uvd_resume(adev); + if (r) + return r; + return uvd_v6_0_hw_init(adev); } @@ -514,6 +852,22 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) WREG32_FIELD(UVD_RBC_RB_CNTL, RB_NO_FETCH, 0); + if (uvd_v6_0_enc_support(adev)) { + ring = &adev->uvd.ring_enc[0]; + WREG32(mmUVD_RB_RPTR, lower_32_bits(ring->wptr)); + WREG32(mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); + WREG32(mmUVD_RB_BASE_LO, ring->gpu_addr); + WREG32(mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); + WREG32(mmUVD_RB_SIZE, ring->ring_size / 4); + + ring = &adev->uvd.ring_enc[1]; + WREG32(mmUVD_RB_RPTR2, lower_32_bits(ring->wptr)); + WREG32(mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); + WREG32(mmUVD_RB_BASE_LO2, ring->gpu_addr); + WREG32(mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); + WREG32(mmUVD_RB_SIZE2, ring->ring_size / 4); + } + return 0; } @@ -577,6 +931,26 @@ static void uvd_v6_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq } /** + * uvd_v6_0_enc_ring_emit_fence - emit an enc fence & trap command + * + * @ring: amdgpu_ring pointer + * @fence: fence to emit + * + * Write enc a fence and a trap command to the ring. + */ +static void uvd_v6_0_enc_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, + u64 seq, unsigned flags) +{ + WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT); + + amdgpu_ring_write(ring, HEVC_ENC_CMD_FENCE); + amdgpu_ring_write(ring, addr); + amdgpu_ring_write(ring, upper_32_bits(addr)); + amdgpu_ring_write(ring, seq); + amdgpu_ring_write(ring, HEVC_ENC_CMD_TRAP); +} + +/** * uvd_v6_0_ring_emit_hdp_flush - emit an hdp flush * * @ring: amdgpu_ring pointer @@ -667,6 +1041,24 @@ static void uvd_v6_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, ib->length_dw); } +/** + * uvd_v6_0_enc_ring_emit_ib - enc execute indirect buffer + * + * @ring: amdgpu_ring pointer + * @ib: indirect buffer to execute + * + * Write enc ring commands to execute the indirect buffer + */ +static void uvd_v6_0_enc_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_ib *ib, unsigned int vm_id, bool ctx_switch) +{ + amdgpu_ring_write(ring, HEVC_ENC_CMD_IB_VM); + amdgpu_ring_write(ring, vm_id); + amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); + amdgpu_ring_write(ring, ib->length_dw); +} + static void uvd_v6_0_ring_emit_vm_flush(struct amdgpu_ring *ring, unsigned vm_id, uint64_t pd_addr) { @@ -718,6 +1110,33 @@ static void uvd_v6_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) amdgpu_ring_write(ring, 0xE); } +static void uvd_v6_0_enc_ring_emit_pipeline_sync(struct amdgpu_ring *ring) +{ + uint32_t seq = ring->fence_drv.sync_seq; + uint64_t addr = ring->fence_drv.gpu_addr; + + amdgpu_ring_write(ring, HEVC_ENC_CMD_WAIT_GE); + amdgpu_ring_write(ring, lower_32_bits(addr)); + amdgpu_ring_write(ring, upper_32_bits(addr)); + amdgpu_ring_write(ring, seq); +} + +static void uvd_v6_0_enc_ring_insert_end(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, HEVC_ENC_CMD_END); +} + +static void uvd_v6_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring, + unsigned int vm_id, uint64_t pd_addr) +{ + amdgpu_ring_write(ring, HEVC_ENC_CMD_UPDATE_PTB); + amdgpu_ring_write(ring, vm_id); + amdgpu_ring_write(ring, pd_addr >> 12); + + amdgpu_ring_write(ring, HEVC_ENC_CMD_FLUSH_TLB); + amdgpu_ring_write(ring, vm_id); +} + static bool uvd_v6_0_is_idle(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -825,8 +1244,31 @@ static int uvd_v6_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { + bool int_handled = true; DRM_DEBUG("IH: UVD TRAP\n"); - amdgpu_fence_process(&adev->uvd.ring); + + switch (entry->src_id) { + case 124: + amdgpu_fence_process(&adev->uvd.ring); + break; + case 119: + if (likely(uvd_v6_0_enc_support(adev))) + amdgpu_fence_process(&adev->uvd.ring_enc[0]); + else + int_handled = false; + break; + case 120: + if (likely(uvd_v6_0_enc_support(adev))) + amdgpu_fence_process(&adev->uvd.ring_enc[1]); + else + int_handled = false; + break; + } + + if (false == int_handled) + DRM_ERROR("Unhandled interrupt: %d %d\n", + entry->src_id, entry->src_data[0]); + return 0; } @@ -1153,6 +1595,33 @@ static const struct amdgpu_ring_funcs uvd_v6_0_ring_vm_funcs = { .end_use = amdgpu_uvd_ring_end_use, }; +static const struct amdgpu_ring_funcs uvd_v6_0_enc_ring_vm_funcs = { + .type = AMDGPU_RING_TYPE_UVD_ENC, + .align_mask = 0x3f, + .nop = HEVC_ENC_CMD_NO_OP, + .support_64bit_ptrs = false, + .get_rptr = uvd_v6_0_enc_ring_get_rptr, + .get_wptr = uvd_v6_0_enc_ring_get_wptr, + .set_wptr = uvd_v6_0_enc_ring_set_wptr, + .emit_frame_size = + 4 + /* uvd_v6_0_enc_ring_emit_pipeline_sync */ + 6 + /* uvd_v6_0_enc_ring_emit_vm_flush */ + 5 + 5 + /* uvd_v6_0_enc_ring_emit_fence x2 vm fence */ + 1, /* uvd_v6_0_enc_ring_insert_end */ + .emit_ib_size = 5, /* uvd_v6_0_enc_ring_emit_ib */ + .emit_ib = uvd_v6_0_enc_ring_emit_ib, + .emit_fence = uvd_v6_0_enc_ring_emit_fence, + .emit_vm_flush = uvd_v6_0_enc_ring_emit_vm_flush, + .emit_pipeline_sync = uvd_v6_0_enc_ring_emit_pipeline_sync, + .test_ring = uvd_v6_0_enc_ring_test_ring, + .test_ib = uvd_v6_0_enc_ring_test_ib, + .insert_nop = amdgpu_ring_insert_nop, + .insert_end = uvd_v6_0_enc_ring_insert_end, + .pad_ib = amdgpu_ring_generic_pad_ib, + .begin_use = amdgpu_uvd_ring_begin_use, + .end_use = amdgpu_uvd_ring_end_use, +}; + static void uvd_v6_0_set_ring_funcs(struct amdgpu_device *adev) { if (adev->asic_type >= CHIP_POLARIS10) { @@ -1164,6 +1633,16 @@ static void uvd_v6_0_set_ring_funcs(struct amdgpu_device *adev) } } +static void uvd_v6_0_set_enc_ring_funcs(struct amdgpu_device *adev) +{ + int i; + + for (i = 0; i < adev->uvd.num_enc_rings; ++i) + adev->uvd.ring_enc[i].funcs = &uvd_v6_0_enc_ring_vm_funcs; + + DRM_INFO("UVD ENC is enabled in VM mode\n"); +} + static const struct amdgpu_irq_src_funcs uvd_v6_0_irq_funcs = { .set = uvd_v6_0_set_interrupt_state, .process = uvd_v6_0_process_interrupt, @@ -1171,7 +1650,11 @@ static const struct amdgpu_irq_src_funcs uvd_v6_0_irq_funcs = { static void uvd_v6_0_set_irq_funcs(struct amdgpu_device *adev) { - adev->uvd.irq.num_types = 1; + if (uvd_v6_0_enc_support(adev)) + adev->uvd.irq.num_types = adev->uvd.num_enc_rings + 1; + else + adev->uvd.irq.num_types = 1; + adev->uvd.irq.funcs = &uvd_v6_0_irq_funcs; } diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c index 987b958368ac..6634545060fd 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c @@ -165,6 +165,9 @@ static int uvd_v7_0_enc_ring_test_ring(struct amdgpu_ring *ring) unsigned i; int r; + if (amdgpu_sriov_vf(adev)) + return 0; + r = amdgpu_ring_alloc(ring, 16); if (r) { DRM_ERROR("amdgpu: uvd enc failed to lock ring %d (%d).\n", @@ -432,13 +435,19 @@ static int uvd_v7_0_sw_init(void *handle) return r; } - for (i = 0; i < adev->uvd.num_enc_rings; ++i) { ring = &adev->uvd.ring_enc[i]; sprintf(ring->name, "uvd_enc%d", i); if (amdgpu_sriov_vf(adev)) { ring->use_doorbell = true; - ring->doorbell_index = AMDGPU_DOORBELL64_UVD_RING0_1 * 2; + + /* currently only use the first enconding ring for + * sriov, so set unused location for other unused rings. + */ + if (i == 0) + ring->doorbell_index = AMDGPU_DOORBELL64_UVD_RING0_1 * 2; + else + ring->doorbell_index = AMDGPU_DOORBELL64_UVD_RING2_3 * 2 + 1; } r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0); if (r) @@ -583,11 +592,7 @@ static int uvd_v7_0_suspend(void *handle) if (r) return r; - /* Skip this for APU for now */ - if (!(adev->flags & AMD_IS_APU)) - r = amdgpu_uvd_suspend(adev); - - return r; + return amdgpu_uvd_suspend(adev); } static int uvd_v7_0_resume(void *handle) @@ -595,12 +600,10 @@ static int uvd_v7_0_resume(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - /* Skip this for APU for now */ - if (!(adev->flags & AMD_IS_APU)) { - r = amdgpu_uvd_resume(adev); - if (r) - return r; - } + r = amdgpu_uvd_resume(adev); + if (r) + return r; + return uvd_v7_0_hw_init(adev); } @@ -685,6 +688,11 @@ static int uvd_v7_0_mmsch_start(struct amdgpu_device *adev, /* 4, set resp to zero */ WREG32_SOC15(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_RESP, 0); + WDOORBELL32(adev->uvd.ring_enc[0].doorbell_index, 0); + adev->wb.wb[adev->uvd.ring_enc[0].wptr_offs] = 0; + adev->uvd.ring_enc[0].wptr = 0; + adev->uvd.ring_enc[0].wptr_old = 0; + /* 5, kick off the initialization and wait until VCE_MMSCH_VF_MAILBOX_RESP becomes non-zero */ WREG32_SOC15(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_HOST, 0x10000001); @@ -702,7 +710,6 @@ static int uvd_v7_0_mmsch_start(struct amdgpu_device *adev, dev_err(adev->dev, "failed to init MMSCH, mmVCE_MMSCH_VF_MAILBOX_RESP = %x\n", data); return -EBUSY; } - WDOORBELL32(adev->uvd.ring_enc[0].doorbell_index, 0); return 0; } @@ -736,11 +743,9 @@ static int uvd_v7_0_sriov_start(struct amdgpu_device *adev) init_table += header->uvd_table_offset; ring = &adev->uvd.ring; + ring->wptr = 0; size = AMDGPU_GPU_PAGE_ALIGN(adev->uvd.fw->size + 4); - /* disable clock gating */ - MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), - ~UVD_POWER_STATUS__UVD_PG_MODE_MASK, 0); MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_STATUS), 0xFFFFFFFF, 0x00000004); /* mc resume*/ @@ -777,12 +782,6 @@ static int uvd_v7_0_sriov_start(struct amdgpu_device *adev) MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_SIZE2), AMDGPU_UVD_STACK_SIZE + (AMDGPU_UVD_SESSION_SIZE * 40)); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_UDEC_ADDR_CONFIG), - adev->gfx.config.gb_addr_config); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_UDEC_DB_ADDR_CONFIG), - adev->gfx.config.gb_addr_config); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG), - adev->gfx.config.gb_addr_config); MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_GP_SCRATCH4), adev->uvd.max_handles); /* mc resume end*/ @@ -819,17 +818,6 @@ static int uvd_v7_0_sriov_start(struct amdgpu_device *adev) UVD_LMI_CTRL__REQ_MODE_MASK | 0x00100000L)); - /* disable byte swapping */ - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_SWAP_CNTL), 0); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MP_SWAP_CNTL), 0); - - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MPC_SET_MUXA0), 0x40c2040); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MPC_SET_MUXA1), 0x0); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MPC_SET_MUXB0), 0x40c2040); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MPC_SET_MUXB1), 0x0); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MPC_SET_ALU), 0); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MPC_SET_MUX), 0x88); - /* take all subblocks out of reset, except VCPU */ MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); @@ -838,15 +826,6 @@ static int uvd_v7_0_sriov_start(struct amdgpu_device *adev) MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CNTL), UVD_VCPU_CNTL__CLK_EN_MASK); - /* enable UMC */ - MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), - ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, 0); - - /* boot up the VCPU */ - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0); - - MMSCH_V1_0_INSERT_DIRECT_POLL(SOC15_REG_OFFSET(UVD, 0, mmUVD_STATUS), 0x02, 0x02); - /* enable master interrupt */ MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), ~(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK), @@ -859,40 +838,31 @@ static int uvd_v7_0_sriov_start(struct amdgpu_device *adev) /* force RBC into idle state */ size = order_base_2(ring->ring_size); tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, size); - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1); tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1); - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_WPTR_POLL_EN, 0); - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1); - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1); MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), tmp); - /* set the write pointer delay */ - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_WPTR_CNTL), 0); - - /* set the wb address */ - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_RPTR_ADDR), - (upper_32_bits(ring->gpu_addr) >> 2)); - - /* programm the RB_BASE for ring buffer */ - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW), - lower_32_bits(ring->gpu_addr)); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH), - upper_32_bits(ring->gpu_addr)); - - ring->wptr = 0; ring = &adev->uvd.ring_enc[0]; + ring->wptr = 0; MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RB_BASE_LO), ring->gpu_addr); MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RB_BASE_HI), upper_32_bits(ring->gpu_addr)); MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RB_SIZE), ring->ring_size / 4); + /* boot up the VCPU */ + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0); + + /* enable UMC */ + MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), + ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, 0); + + MMSCH_V1_0_INSERT_DIRECT_POLL(SOC15_REG_OFFSET(UVD, 0, mmUVD_STATUS), 0x02, 0x02); + /* add end packet */ memcpy((void *)init_table, &end, sizeof(struct mmsch_v1_0_cmd_end)); table_size += sizeof(struct mmsch_v1_0_cmd_end) / 4; header->uvd_table_size = table_size; - return uvd_v7_0_mmsch_start(adev, &adev->virt.mm_table); } - return -EINVAL; /* already initializaed ? */ + return uvd_v7_0_mmsch_start(adev, &adev->virt.mm_table); } /** @@ -1185,7 +1155,7 @@ static void uvd_v7_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) */ static void uvd_v7_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) { - amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0), 0)); + amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 0)); amdgpu_ring_write(ring, 1); } diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c index 90332f55cfba..cf81065e3c5a 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c @@ -365,15 +365,10 @@ static unsigned vce_v3_0_get_harvest_config(struct amdgpu_device *adev) { u32 tmp; - /* Fiji, Stoney, Polaris10, Polaris11, Polaris12 are single pipe */ if ((adev->asic_type == CHIP_FIJI) || - (adev->asic_type == CHIP_STONEY) || - (adev->asic_type == CHIP_POLARIS10) || - (adev->asic_type == CHIP_POLARIS11) || - (adev->asic_type == CHIP_POLARIS12)) + (adev->asic_type == CHIP_STONEY)) return AMDGPU_VCE_HARVEST_VCE1; - /* Tonga and CZ are dual or single pipe */ if (adev->flags & AMD_IS_APU) tmp = (RREG32_SMC(ixVCE_HARVEST_FUSE_MACRO__ADDRESS) & VCE_HARVEST_FUSE_MACRO__MASK) >> @@ -391,6 +386,11 @@ static unsigned vce_v3_0_get_harvest_config(struct amdgpu_device *adev) case 3: return AMDGPU_VCE_HARVEST_VCE0 | AMDGPU_VCE_HARVEST_VCE1; default: + if ((adev->asic_type == CHIP_POLARIS10) || + (adev->asic_type == CHIP_POLARIS11) || + (adev->asic_type == CHIP_POLARIS12)) + return AMDGPU_VCE_HARVEST_VCE1; + return 0; } } diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c index 1ecd6bb90c1f..75745544600a 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c @@ -173,6 +173,11 @@ static int vce_v4_0_mmsch_start(struct amdgpu_device *adev, /* 4, set resp to zero */ WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_RESP), 0); + WDOORBELL32(adev->vce.ring[0].doorbell_index, 0); + adev->wb.wb[adev->vce.ring[0].wptr_offs] = 0; + adev->vce.ring[0].wptr = 0; + adev->vce.ring[0].wptr_old = 0; + /* 5, kick off the initialization and wait until VCE_MMSCH_VF_MAILBOX_RESP becomes non-zero */ WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_HOST), 0x10000001); @@ -190,7 +195,6 @@ static int vce_v4_0_mmsch_start(struct amdgpu_device *adev, dev_err(adev->dev, "failed to init MMSCH, mmVCE_MMSCH_VF_MAILBOX_RESP = %x\n", data); return -EBUSY; } - WDOORBELL32(adev->vce.ring[0].doorbell_index, 0); return 0; } @@ -274,7 +278,8 @@ static int vce_v4_0_sriov_start(struct amdgpu_device *adev) MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_CTRL2), ~0x100, 0); MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_SYS_INT_EN), - 0xffffffff, VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK); + VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK, + VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK); /* end of MC_RESUME */ MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_STATUS), @@ -296,11 +301,9 @@ static int vce_v4_0_sriov_start(struct amdgpu_device *adev) memcpy((void *)init_table, &end, sizeof(struct mmsch_v1_0_cmd_end)); table_size += sizeof(struct mmsch_v1_0_cmd_end) / 4; header->vce_table_size = table_size; - - return vce_v4_0_mmsch_start(adev, &adev->virt.mm_table); } - return -EINVAL; /* already initializaed ? */ + return vce_v4_0_mmsch_start(adev, &adev->virt.mm_table); } /** @@ -443,12 +446,14 @@ static int vce_v4_0_sw_init(void *handle) if (amdgpu_sriov_vf(adev)) { /* DOORBELL only works under SRIOV */ ring->use_doorbell = true; + + /* currently only use the first encoding ring for sriov, + * so set unused location for other unused rings. + */ if (i == 0) - ring->doorbell_index = AMDGPU_DOORBELL64_RING0_1 * 2; - else if (i == 1) - ring->doorbell_index = AMDGPU_DOORBELL64_RING2_3 * 2; + ring->doorbell_index = AMDGPU_DOORBELL64_VCE_RING0_1 * 2; else - ring->doorbell_index = AMDGPU_DOORBELL64_RING2_3 * 2 + 1; + ring->doorbell_index = AMDGPU_DOORBELL64_VCE_RING2_3 * 2 + 1; } r = amdgpu_ring_init(adev, ring, 512, &adev->vce.irq, 0); if (r) @@ -990,11 +995,13 @@ static int vce_v4_0_set_interrupt_state(struct amdgpu_device *adev, { uint32_t val = 0; - if (state == AMDGPU_IRQ_STATE_ENABLE) - val |= VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK; + if (!amdgpu_sriov_vf(adev)) { + if (state == AMDGPU_IRQ_STATE_ENABLE) + val |= VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK; - WREG32_P(SOC15_REG_OFFSET(VCE, 0, mmVCE_SYS_INT_EN), val, - ~VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK); + WREG32_P(SOC15_REG_OFFSET(VCE, 0, mmVCE_SYS_INT_EN), val, + ~VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK); + } return 0; } @@ -1004,10 +1011,6 @@ static int vce_v4_0_process_interrupt(struct amdgpu_device *adev, { DRM_DEBUG("IH: VCE\n"); - WREG32_P(SOC15_REG_OFFSET(VCE, 0, mmVCE_SYS_INT_STATUS), - VCE_SYS_INT_STATUS__VCE_SYS_INT_TRAP_INTERRUPT_INT_MASK, - ~VCE_SYS_INT_STATUS__VCE_SYS_INT_TRAP_INTERRUPT_INT_MASK); - switch (entry->src_data[0]) { case 0: case 1: diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c index 21e7b88401e1..1eb4d79d6e30 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c @@ -812,7 +812,7 @@ static void vcn_v1_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 */ static void vcn_v1_0_dec_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) { - amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0), 0)); + amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 0)); amdgpu_ring_write(ring, 1); } diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c index 56150e8d1ed2..697325737ba8 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c @@ -219,14 +219,95 @@ static u32 vega10_ih_get_wptr(struct amdgpu_device *adev) wptr, adev->irq.ih.rptr, tmp); adev->irq.ih.rptr = tmp; - tmp = RREG32(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL)); + tmp = RREG32_NO_KIQ(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL)); tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1); - WREG32(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL), tmp); + WREG32_NO_KIQ(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL), tmp); } return (wptr & adev->irq.ih.ptr_mask); } /** + * vega10_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool vega10_ih_prescreen_iv(struct amdgpu_device *adev) +{ + u32 ring_index = adev->irq.ih.rptr >> 2; + u32 dw0, dw3, dw4, dw5; + u16 pasid; + u64 addr, key; + struct amdgpu_vm *vm; + int r; + + dw0 = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]); + dw3 = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); + dw4 = le32_to_cpu(adev->irq.ih.ring[ring_index + 4]); + dw5 = le32_to_cpu(adev->irq.ih.ring[ring_index + 5]); + + /* Filter retry page faults, let only the first one pass. If + * there are too many outstanding faults, ignore them until + * some faults get cleared. + */ + switch (dw0 & 0xff) { + case AMDGPU_IH_CLIENTID_VMC: + case AMDGPU_IH_CLIENTID_UTCL2: + break; + default: + /* Not a VM fault */ + return true; + } + + pasid = dw3 & 0xffff; + /* No PASID, can't identify faulting process */ + if (!pasid) + return true; + + /* Not a retry fault, check fault credit */ + if (!(dw5 & 0x80)) { + if (!amdgpu_vm_pasid_fault_credit(adev, pasid)) + goto ignore_iv; + return true; + } + + addr = ((u64)(dw5 & 0xf) << 44) | ((u64)dw4 << 12); + key = AMDGPU_VM_FAULT(pasid, addr); + r = amdgpu_ih_add_fault(adev, key); + + /* Hash table is full or the fault is already being processed, + * ignore further page faults + */ + if (r != 0) + goto ignore_iv; + + /* Track retry faults in per-VM fault FIFO. */ + spin_lock(&adev->vm_manager.pasid_lock); + vm = idr_find(&adev->vm_manager.pasid_idr, pasid); + spin_unlock(&adev->vm_manager.pasid_lock); + if (WARN_ON_ONCE(!vm)) { + /* VM not found, process it normally */ + amdgpu_ih_clear_fault(adev, key); + return true; + } + /* No locking required with single writer and single reader */ + r = kfifo_put(&vm->faults, key); + if (!r) { + /* FIFO is full. Ignore it until there is space */ + amdgpu_ih_clear_fault(adev, key); + goto ignore_iv; + } + + /* It's the first fault for this address, process it normally */ + return true; + +ignore_iv: + adev->irq.ih.rptr += 32; + return false; +} + +/** * vega10_ih_decode_iv - decode an interrupt vector * * @adev: amdgpu_device pointer @@ -310,6 +391,14 @@ static int vega10_ih_sw_init(void *handle) adev->irq.ih.use_doorbell = true; adev->irq.ih.doorbell_index = AMDGPU_DOORBELL64_IH << 1; + adev->irq.ih.faults = kmalloc(sizeof(*adev->irq.ih.faults), GFP_KERNEL); + if (!adev->irq.ih.faults) + return -ENOMEM; + INIT_CHASH_TABLE(adev->irq.ih.faults->hash, + AMDGPU_PAGEFAULT_HASH_BITS, 8, 0); + spin_lock_init(&adev->irq.ih.faults->lock); + adev->irq.ih.faults->count = 0; + r = amdgpu_irq_init(adev); return r; @@ -322,6 +411,9 @@ static int vega10_ih_sw_fini(void *handle) amdgpu_irq_fini(adev); amdgpu_ih_ring_fini(adev); + kfree(adev->irq.ih.faults); + adev->irq.ih.faults = NULL; + return 0; } @@ -410,6 +502,7 @@ const struct amd_ip_funcs vega10_ih_ip_funcs = { static const struct amdgpu_ih_funcs vega10_ih_funcs = { .get_wptr = vega10_ih_get_wptr, + .prescreen_iv = vega10_ih_prescreen_iv, .decode_iv = vega10_ih_decode_iv, .set_rptr = vega10_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index 6cac291c96da..3a4c2fa7e36d 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -77,6 +77,7 @@ #endif #include "dce_virtual.h" #include "mxgpu_vi.h" +#include "amdgpu_dm.h" /* * Indirect registers accessor @@ -1028,8 +1029,7 @@ static int vi_common_early_init(void *handle) /* rev0 hardware requires workarounds to support PG */ adev->pg_flags = 0; if (adev->rev_id != 0x00 || CZ_REV_BRISTOL(adev->pdev->revision)) { - adev->pg_flags |= AMD_PG_SUPPORT_GFX_PG | - AMD_PG_SUPPORT_GFX_SMG | + adev->pg_flags |= AMD_PG_SUPPORT_GFX_SMG | AMD_PG_SUPPORT_GFX_PIPELINE | AMD_PG_SUPPORT_CP | AMD_PG_SUPPORT_UVD | @@ -1255,7 +1255,6 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, uint32_t msg_id, pp_state = 0; uint32_t pp_support_state = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - void *pp_handle = adev->powerplay.pp_handle; if (adev->cg_flags & (AMD_CG_SUPPORT_MC_LS | AMD_CG_SUPPORT_MC_MGCG)) { if (adev->cg_flags & AMD_CG_SUPPORT_MC_LS) { @@ -1272,7 +1271,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_MC, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_SDMA_LS | AMD_CG_SUPPORT_SDMA_MGCG)) { @@ -1290,7 +1290,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_SDMA, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS | AMD_CG_SUPPORT_HDP_MGCG)) { @@ -1308,7 +1309,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_HDP, pp_support_state, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } @@ -1322,7 +1324,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_BIF, PP_STATE_SUPPORT_LS, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG) { if (state == AMD_CG_STATE_UNGATE) @@ -1334,7 +1337,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_BIF, PP_STATE_SUPPORT_CG, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & AMD_CG_SUPPORT_DRM_LS) { @@ -1348,7 +1352,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_DRM, PP_STATE_SUPPORT_LS, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } if (adev->cg_flags & AMD_CG_SUPPORT_ROM_MGCG) { @@ -1362,7 +1367,8 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, PP_BLOCK_SYS_ROM, PP_STATE_SUPPORT_CG, pp_state); - amd_set_clockgating_by_smu(pp_handle, msg_id); + if (adev->powerplay.pp_funcs->set_clockgating_by_smu) + amdgpu_dpm_set_clockgating_by_smu(adev, msg_id); } return 0; } @@ -1497,6 +1503,10 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_ip_block_add(adev, &amdgpu_pp_ip_block); if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) amdgpu_ip_block_add(adev, &dce_virtual_ip_block); +#if defined(CONFIG_DRM_AMD_DC) + else if (amdgpu_device_has_dc_support(adev)) + amdgpu_ip_block_add(adev, &dm_ip_block); +#endif else amdgpu_ip_block_add(adev, &dce_v10_1_ip_block); amdgpu_ip_block_add(adev, &gfx_v8_0_ip_block); @@ -1513,6 +1523,10 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_ip_block_add(adev, &amdgpu_pp_ip_block); if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) amdgpu_ip_block_add(adev, &dce_virtual_ip_block); +#if defined(CONFIG_DRM_AMD_DC) + else if (amdgpu_device_has_dc_support(adev)) + amdgpu_ip_block_add(adev, &dm_ip_block); +#endif else amdgpu_ip_block_add(adev, &dce_v10_0_ip_block); amdgpu_ip_block_add(adev, &gfx_v8_0_ip_block); @@ -1531,6 +1545,10 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_ip_block_add(adev, &amdgpu_pp_ip_block); if (adev->enable_virtual_display) amdgpu_ip_block_add(adev, &dce_virtual_ip_block); +#if defined(CONFIG_DRM_AMD_DC) + else if (amdgpu_device_has_dc_support(adev)) + amdgpu_ip_block_add(adev, &dm_ip_block); +#endif else amdgpu_ip_block_add(adev, &dce_v11_2_ip_block); amdgpu_ip_block_add(adev, &gfx_v8_0_ip_block); @@ -1545,6 +1563,10 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_ip_block_add(adev, &amdgpu_pp_ip_block); if (adev->enable_virtual_display) amdgpu_ip_block_add(adev, &dce_virtual_ip_block); +#if defined(CONFIG_DRM_AMD_DC) + else if (amdgpu_device_has_dc_support(adev)) + amdgpu_ip_block_add(adev, &dm_ip_block); +#endif else amdgpu_ip_block_add(adev, &dce_v11_0_ip_block); amdgpu_ip_block_add(adev, &gfx_v8_0_ip_block); @@ -1562,6 +1584,10 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_ip_block_add(adev, &amdgpu_pp_ip_block); if (adev->enable_virtual_display) amdgpu_ip_block_add(adev, &dce_virtual_ip_block); +#if defined(CONFIG_DRM_AMD_DC) + else if (amdgpu_device_has_dc_support(adev)) + amdgpu_ip_block_add(adev, &dm_ip_block); +#endif else amdgpu_ip_block_add(adev, &dce_v11_0_ip_block); amdgpu_ip_block_add(adev, &gfx_v8_1_ip_block); diff --git a/drivers/gpu/drm/amd/amdgpu/vid.h b/drivers/gpu/drm/amd/amdgpu/vid.h index a6485254a169..dbf3703cbd1b 100644 --- a/drivers/gpu/drm/amd/amdgpu/vid.h +++ b/drivers/gpu/drm/amd/amdgpu/vid.h @@ -465,6 +465,16 @@ #define VCE_CMD_UPDATE_PTB 0x00000107 #define VCE_CMD_FLUSH_TLB 0x00000108 +/* HEVC ENC */ +#define HEVC_ENC_CMD_NO_OP 0x00000000 +#define HEVC_ENC_CMD_END 0x00000001 +#define HEVC_ENC_CMD_FENCE 0x00000003 +#define HEVC_ENC_CMD_TRAP 0x00000004 +#define HEVC_ENC_CMD_IB_VM 0x00000102 +#define HEVC_ENC_CMD_WAIT_GE 0x00000106 +#define HEVC_ENC_CMD_UPDATE_PTB 0x00000107 +#define HEVC_ENC_CMD_FLUSH_TLB 0x00000108 + /* mmPA_SC_RASTER_CONFIG mask */ #define RB_MAP_PKR0(x) ((x) << 0) #define RB_MAP_PKR0_MASK (0x3 << 0) |