diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
87 files changed, 6784 insertions, 2980 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/Kconfig b/drivers/gpu/drm/amd/amdgpu/Kconfig index 7335c0420c70..f3cb69de0c44 100644 --- a/drivers/gpu/drm/amd/amdgpu/Kconfig +++ b/drivers/gpu/drm/amd/amdgpu/Kconfig @@ -25,3 +25,5 @@ config DRM_AMDGPU_GART_DEBUGFS Selecting this option creates a debugfs file to inspect the mapped pages. Uses more memory for housekeeping, enable only for debugging. +source "drivers/gpu/drm/amd/powerplay/Kconfig" +source "drivers/gpu/drm/amd/acp/Kconfig" diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index c7fcdcedaadb..21dd7c00da15 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -58,7 +58,8 @@ amdgpu-y += \ # add DCE block amdgpu-y += \ dce_v10_0.o \ - dce_v11_0.o + dce_v11_0.o \ + dce_virtual.o # add GFX block amdgpu-y += \ diff --git a/drivers/gpu/drm/amd/amdgpu/ObjectID.h b/drivers/gpu/drm/amd/amdgpu/ObjectID.h index 06192698bd96..b8d66670bb17 100644 --- a/drivers/gpu/drm/amd/amdgpu/ObjectID.h +++ b/drivers/gpu/drm/amd/amdgpu/ObjectID.h @@ -90,6 +90,7 @@ #define ENCODER_OBJECT_ID_INTERNAL_VCE 0x24 #define ENCODER_OBJECT_ID_INTERNAL_UNIPHY3 0x25 #define ENCODER_OBJECT_ID_INTERNAL_AMCLK 0x27 +#define ENCODER_OBJECT_ID_VIRTUAL 0x28 #define ENCODER_OBJECT_ID_GENERAL_EXTERNAL_DVO 0xFF @@ -119,6 +120,7 @@ #define CONNECTOR_OBJECT_ID_eDP 0x14 #define CONNECTOR_OBJECT_ID_MXM 0x15 #define CONNECTOR_OBJECT_ID_LVDS_eDP 0x16 +#define CONNECTOR_OBJECT_ID_VIRTUAL 0x17 /* deleted */ @@ -147,6 +149,7 @@ #define GRAPH_OBJECT_ENUM_ID5 0x05 #define GRAPH_OBJECT_ENUM_ID6 0x06 #define GRAPH_OBJECT_ENUM_ID7 0x07 +#define GRAPH_OBJECT_ENUM_VIRTUAL 0x08 /****************************************************/ /* Graphics Object ID Bit definition */ @@ -408,6 +411,10 @@ GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\ ENCODER_OBJECT_ID_HDMI_ANX9805 << OBJECT_ID_SHIFT) +#define ENCODER_VIRTUAL_ENUM_VIRTUAL ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_VIRTUAL << ENUM_ID_SHIFT |\ + ENCODER_OBJECT_ID_VIRTUAL << OBJECT_ID_SHIFT) + /****************************************************/ /* Connector Object ID definition - Shared with BIOS */ /****************************************************/ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index e055d5be1c3c..3cc2629eb158 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -51,6 +51,7 @@ #include "amdgpu_ih.h" #include "amdgpu_irq.h" #include "amdgpu_ucode.h" +#include "amdgpu_ttm.h" #include "amdgpu_gds.h" #include "amd_powerplay.h" #include "amdgpu_acp.h" @@ -85,8 +86,14 @@ extern int amdgpu_vm_debug; extern int amdgpu_sched_jobs; extern int amdgpu_sched_hw_submission; extern int amdgpu_powerplay; +extern int amdgpu_powercontainment; extern unsigned amdgpu_pcie_gen_cap; extern unsigned amdgpu_pcie_lane_cap; +extern unsigned amdgpu_cg_mask; +extern unsigned amdgpu_pg_mask; +extern char *amdgpu_disable_cu; +extern int amdgpu_sclk_deep_sleep_en; +extern char *amdgpu_virtual_display; #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS 3000 #define AMDGPU_MAX_USEC_TIMEOUT 100000 /* 100 ms */ @@ -183,6 +190,10 @@ int amdgpu_set_clockgating_state(struct amdgpu_device *adev, int amdgpu_set_powergating_state(struct amdgpu_device *adev, enum amd_ip_block_type block_type, enum amd_powergating_state state); +int amdgpu_wait_for_idle(struct amdgpu_device *adev, + enum amd_ip_block_type block_type); +bool amdgpu_is_idle(struct amdgpu_device *adev, + enum amd_ip_block_type block_type); struct amdgpu_ip_block_version { enum amd_ip_block_type type; @@ -240,10 +251,9 @@ struct amdgpu_vm_pte_funcs { 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, - const dma_addr_t *pages_addr, uint64_t pe, - uint64_t addr, unsigned count, - uint32_t incr, uint32_t flags); + void (*write_pte)(struct amdgpu_ib *ib, uint64_t pe, + uint64_t value, unsigned count, + uint32_t incr); /* for linear pte/pde updates without addr mapping */ void (*set_pte_pde)(struct amdgpu_ib *ib, uint64_t pe, @@ -298,13 +308,16 @@ struct amdgpu_ring_funcs { uint32_t oa_base, uint32_t oa_size); /* testing functions */ int (*test_ring)(struct amdgpu_ring *ring); - int (*test_ib)(struct amdgpu_ring *ring); + int (*test_ib)(struct amdgpu_ring *ring, long timeout); /* insert NOP packets */ void (*insert_nop)(struct amdgpu_ring *ring, uint32_t count); /* pad the indirect buffer to the necessary number of dw */ void (*pad_ib)(struct amdgpu_ring *ring, struct amdgpu_ib *ib); unsigned (*init_cond_exec)(struct amdgpu_ring *ring); void (*patch_cond_exec)(struct amdgpu_ring *ring, unsigned offset); + /* note usage for clock and power gating */ + void (*begin_use)(struct amdgpu_ring *ring); + void (*end_use)(struct amdgpu_ring *ring); }; /* @@ -385,46 +398,9 @@ int amdgpu_fence_wait_empty(struct amdgpu_ring *ring); unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring); /* - * TTM. + * BO. */ -#define AMDGPU_TTM_LRU_SIZE 20 - -struct amdgpu_mman_lru { - struct list_head *lru[TTM_NUM_MEM_TYPES]; - struct list_head *swap_lru; -}; - -struct amdgpu_mman { - struct ttm_bo_global_ref bo_global_ref; - struct drm_global_reference mem_global_ref; - struct ttm_bo_device bdev; - bool mem_global_referenced; - bool initialized; - -#if defined(CONFIG_DEBUG_FS) - struct dentry *vram; - struct dentry *gtt; -#endif - - /* buffer handling */ - const struct amdgpu_buffer_funcs *buffer_funcs; - struct amdgpu_ring *buffer_funcs_ring; - /* Scheduler entity for buffer moves */ - struct amd_sched_entity entity; - - /* custom LRU management */ - struct amdgpu_mman_lru log2_size[AMDGPU_TTM_LRU_SIZE]; -}; - -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 fence **fence); -int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma); - struct amdgpu_bo_list_entry { struct amdgpu_bo *robj; struct ttm_validate_buffer tv; @@ -487,10 +463,12 @@ struct amdgpu_bo { struct amdgpu_device *adev; 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) @@ -594,11 +572,9 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync, struct reservation_object *resv, void *owner); -bool amdgpu_sync_is_idle(struct amdgpu_sync *sync); -int amdgpu_sync_cycle_fences(struct amdgpu_sync *dst, struct amdgpu_sync *src, - struct fence *fence); +struct fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync, + struct amdgpu_ring *ring); struct fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync); -int amdgpu_sync_wait(struct amdgpu_sync *sync); void amdgpu_sync_free(struct amdgpu_sync *sync); int amdgpu_sync_init(void); void amdgpu_sync_fini(void); @@ -668,6 +644,8 @@ struct amdgpu_mc { 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; }; /* @@ -712,10 +690,11 @@ void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev, */ struct amdgpu_flip_work { - struct work_struct flip_work; + struct delayed_work flip_work; struct work_struct unpin_work; struct amdgpu_device *adev; int crtc_id; + u32 target_vblank; uint64_t base; struct drm_pending_vblank_event *event; struct amdgpu_bo *old_rbo; @@ -754,12 +733,11 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, struct amdgpu_job **job); +void amdgpu_job_free_resources(struct amdgpu_job *job); void amdgpu_job_free(struct amdgpu_job *job); -void amdgpu_job_free_func(struct kref *refcount); int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring, struct amd_sched_entity *entity, void *owner, struct fence **f); -void amdgpu_job_timeout_func(struct work_struct *work); struct amdgpu_ring { struct amdgpu_device *adev; @@ -767,12 +745,9 @@ struct amdgpu_ring { struct amdgpu_fence_driver fence_drv; struct amd_gpu_scheduler sched; - spinlock_t fence_lock; struct amdgpu_bo *ring_obj; volatile uint32_t *ring; unsigned rptr_offs; - u64 next_rptr_gpu_addr; - volatile u32 *next_rptr_cpu_addr; unsigned wptr; unsigned wptr_old; unsigned ring_size; @@ -791,14 +766,16 @@ struct amdgpu_ring { u32 doorbell_index; bool use_doorbell; unsigned wptr_offs; - unsigned next_rptr_offs; unsigned fence_offs; uint64_t current_ctx; enum amdgpu_ring_type type; char name[16]; unsigned cond_exe_offs; - u64 cond_exe_gpu_addr; - volatile u32 *cond_exe_cpu_addr; + u64 cond_exe_gpu_addr; + volatile u32 *cond_exe_cpu_addr; +#if defined(CONFIG_DEBUG_FS) + struct dentry *ent; +#endif }; /* @@ -808,13 +785,17 @@ struct amdgpu_ring { /* maximum number of VMIDs */ #define AMDGPU_NUM_VM 16 +/* Maximum number of PTEs the hardware can write with one command */ +#define AMDGPU_VM_MAX_UPDATE_SIZE 0x3FFFF + /* number of entries in page table */ #define AMDGPU_VM_PTE_COUNT (1 << amdgpu_vm_block_size) /* PTBs (Page Table Blocks) need to be aligned to 32K */ #define AMDGPU_VM_PTB_ALIGN_SIZE 32768 -#define AMDGPU_VM_PTB_ALIGN_MASK (AMDGPU_VM_PTB_ALIGN_SIZE - 1) -#define AMDGPU_VM_PTB_ALIGN(a) (((a) + AMDGPU_VM_PTB_ALIGN_MASK) & ~AMDGPU_VM_PTB_ALIGN_MASK) + +/* LOG2 number of continuous pages for the fragment field */ +#define AMDGPU_LOG2_PAGES_PER_FRAG 4 #define AMDGPU_PTE_VALID (1 << 0) #define AMDGPU_PTE_SYSTEM (1 << 1) @@ -826,10 +807,7 @@ struct amdgpu_ring { #define AMDGPU_PTE_READABLE (1 << 5) #define AMDGPU_PTE_WRITEABLE (1 << 6) -/* PTE (Page Table Entry) fragment field for different page sizes */ -#define AMDGPU_PTE_FRAG_4KB (0 << 7) -#define AMDGPU_PTE_FRAG_64KB (4 << 7) -#define AMDGPU_LOG2_PAGES_PER_FRAG 4 +#define AMDGPU_PTE_FRAG(x) ((x & 0x1f) << 7) /* How to programm VM fault handling */ #define AMDGPU_VM_FAULT_STOP_NEVER 0 @@ -839,6 +817,7 @@ struct amdgpu_ring { struct amdgpu_vm_pt { struct amdgpu_bo_list_entry entry; uint64_t addr; + uint64_t shadow_addr; }; struct amdgpu_vm { @@ -861,6 +840,7 @@ struct amdgpu_vm { struct amdgpu_bo *page_directory; unsigned max_pde_used; struct fence *page_directory_fence; + uint64_t last_eviction_counter; /* array of page tables, one for each page directory entry */ struct amdgpu_vm_pt *page_tables; @@ -883,13 +863,14 @@ struct amdgpu_vm_id { struct fence *first; struct amdgpu_sync active; struct fence *last_flush; - struct amdgpu_ring *last_user; atomic64_t owner; uint64_t pd_gpu_addr; /* last flushed PD/PT update */ struct fence *flushed_updates; + uint32_t current_gpu_reset_count; + uint32_t gds_base; uint32_t gds_size; uint32_t gws_base; @@ -905,6 +886,10 @@ struct amdgpu_vm_manager { struct list_head ids_lru; struct amdgpu_vm_id ids[AMDGPU_NUM_VM]; + /* Handling of VM fences */ + u64 fence_context; + unsigned seqno[AMDGPU_MAX_RINGS]; + uint32_t max_pfn; /* vram base address for page table entry */ u64 vram_base_offset; @@ -926,19 +911,15 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm); void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, struct list_head *validated, struct amdgpu_bo_list_entry *entry); -void amdgpu_vm_get_pt_bos(struct amdgpu_vm *vm, struct list_head *duplicates); +void amdgpu_vm_get_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct list_head *duplicates); void amdgpu_vm_move_pt_bos_in_lru(struct amdgpu_device *adev, struct amdgpu_vm *vm); int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring, struct amdgpu_sync *sync, struct fence *fence, - unsigned *vm_id, uint64_t *vm_pd_addr); -int amdgpu_vm_flush(struct amdgpu_ring *ring, - unsigned vm_id, uint64_t pd_addr, - uint32_t gds_base, uint32_t gds_size, - uint32_t gws_base, uint32_t gws_size, - uint32_t oa_base, uint32_t oa_size); + struct amdgpu_job *job); +int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job); void amdgpu_vm_reset_id(struct amdgpu_device *adev, unsigned vm_id); -uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr); int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, struct amdgpu_vm *vm); int amdgpu_vm_clear_freed(struct amdgpu_device *adev, @@ -947,7 +928,7 @@ int amdgpu_vm_clear_invalids(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct amdgpu_sync *sync); int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, - struct ttm_mem_reg *mem); + bool clear); void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, struct amdgpu_bo *bo); struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm, @@ -1142,6 +1123,12 @@ struct amdgpu_cu_info { uint32_t bitmap[4][4]; }; +struct amdgpu_gfx_funcs { + /* get the gpu clock counter */ + uint64_t (*get_gpu_clock_counter)(struct amdgpu_device *adev); + void (*select_se_sh)(struct amdgpu_device *adev, u32 se_num, u32 sh_num, u32 instance); +}; + struct amdgpu_gfx { struct mutex gpu_clock_mutex; struct amdgpu_gca_config config; @@ -1178,6 +1165,11 @@ struct amdgpu_gfx { /* ce ram size*/ unsigned ce_ram_size; struct amdgpu_cu_info cu_info; + const struct amdgpu_gfx_funcs *funcs; + + /* reset mask */ + uint32_t grbm_soft_reset; + uint32_t srbm_soft_reset; }; int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, @@ -1195,10 +1187,6 @@ 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); -unsigned amdgpu_ring_backup(struct amdgpu_ring *ring, - uint32_t **data); -int amdgpu_ring_restore(struct amdgpu_ring *ring, - unsigned size, uint32_t *data); int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned ring_size, u32 nop, u32 align_mask, struct amdgpu_irq_src *irq_src, unsigned irq_type, @@ -1250,6 +1238,7 @@ struct amdgpu_job { uint32_t num_ibs; void *owner; uint64_t ctx; + bool vm_needs_flush; unsigned vm_id; uint64_t vm_pd_addr; uint32_t gds_base, gds_size; @@ -1257,8 +1246,7 @@ struct amdgpu_job { uint32_t oa_base, oa_size; /* user fence handling */ - struct amdgpu_bo *uf_bo; - uint32_t uf_offset; + uint64_t uf_addr; uint64_t uf_sequence; }; @@ -1560,6 +1548,12 @@ struct amdgpu_dpm_funcs { 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); }; struct amdgpu_dpm { @@ -1662,7 +1656,9 @@ struct amdgpu_uvd { struct amdgpu_ring ring; struct amdgpu_irq_src irq; bool address_64_bit; + bool use_ctx_buf; struct amd_sched_entity entity; + uint32_t srbm_soft_reset; }; /* @@ -1683,11 +1679,13 @@ struct amdgpu_vce { struct drm_file *filp[AMDGPU_MAX_VCE_HANDLES]; uint32_t img_size[AMDGPU_MAX_VCE_HANDLES]; struct delayed_work idle_work; + struct mutex idle_mutex; const struct firmware *fw; /* VCE firmware */ struct amdgpu_ring ring[AMDGPU_MAX_VCE_RINGS]; struct amdgpu_irq_src irq; unsigned harvest_config; struct amd_sched_entity entity; + uint32_t srbm_soft_reset; }; /* @@ -1708,6 +1706,7 @@ struct amdgpu_sdma { struct amdgpu_irq_src trap_irq; struct amdgpu_irq_src illegal_inst_irq; int num_instances; + uint32_t srbm_soft_reset; }; /* @@ -1767,6 +1766,8 @@ int amdgpu_debugfs_init(struct drm_minor *minor); void amdgpu_debugfs_cleanup(struct drm_minor *minor); #endif +int amdgpu_debugfs_firmware_init(struct amdgpu_device *adev); + /* * amdgpu smumgr functions */ @@ -1811,12 +1812,8 @@ struct amdgpu_asic_funcs { u32 sh_num, u32 reg_offset, u32 *value); void (*set_vga_state)(struct amdgpu_device *adev, bool state); int (*reset)(struct amdgpu_device *adev); - /* wait for mc_idle */ - int (*wait_for_mc_idle)(struct amdgpu_device *adev); /* get the reference clock */ u32 (*get_xclk)(struct amdgpu_device *adev); - /* get the gpu clock counter */ - uint64_t (*get_gpu_clock_counter)(struct amdgpu_device *adev); /* MM block clocks */ int (*set_uvd_clocks)(struct amdgpu_device *adev, u32 vclk, u32 dclk); int (*set_vce_clocks)(struct amdgpu_device *adev, u32 evclk, u32 ecclk); @@ -1937,6 +1934,7 @@ struct amdgpu_ip_block_status { bool valid; bool sw; bool hw; + bool hang; }; struct amdgpu_device { @@ -2003,6 +2001,10 @@ struct amdgpu_device { spinlock_t didt_idx_lock; amdgpu_rreg_t didt_rreg; amdgpu_wreg_t didt_wreg; + /* protects concurrent gc_cac register access */ + spinlock_t gc_cac_idx_lock; + amdgpu_rreg_t gc_cac_rreg; + amdgpu_wreg_t gc_cac_wreg; /* protects concurrent ENDPOINT (audio) register access */ spinlock_t audio_endpt_idx_lock; amdgpu_block_rreg_t audio_endpt_rreg; @@ -2028,9 +2030,11 @@ struct amdgpu_device { atomic64_t vram_vis_usage; atomic64_t gtt_usage; atomic64_t num_bytes_moved; + atomic64_t num_evictions; atomic_t gpu_reset_counter; /* display */ + bool enable_virtual_display; struct amdgpu_mode_info mode_info; struct work_struct hotplug_work; struct amdgpu_irq_src crtc_irq; @@ -2038,7 +2042,7 @@ struct amdgpu_device { struct amdgpu_irq_src hpd_irq; /* rings */ - unsigned fence_context; + u64 fence_context; unsigned num_rings; struct amdgpu_ring *rings[AMDGPU_MAX_RINGS]; bool ib_pool_ready; @@ -2093,6 +2097,10 @@ struct amdgpu_device { struct kfd_dev *kfd; struct amdgpu_virtualization virtualization; + + /* link all shadow bo */ + struct list_head shadow_list; + struct mutex shadow_list_lock; }; bool amdgpu_device_is_px(struct drm_device *dev); @@ -2131,6 +2139,8 @@ void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v); #define WREG32_UVD_CTX(reg, v) adev->uvd_ctx_wreg(adev, (reg), (v)) #define RREG32_DIDT(reg) adev->didt_rreg(adev, (reg)) #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_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) \ @@ -2166,6 +2176,9 @@ void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v); #define REG_GET_FIELD(value, reg, field) \ (((value) & REG_FIELD_MASK(reg, field)) >> REG_FIELD_SHIFT(reg, field)) +#define WREG32_FIELD(reg, field, val) \ + WREG32(mm##reg, (RREG32(mm##reg) & ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field)) + /* * BIOS helpers. */ @@ -2206,23 +2219,21 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) */ #define amdgpu_asic_set_vga_state(adev, state) (adev)->asic_funcs->set_vga_state((adev), (state)) #define amdgpu_asic_reset(adev) (adev)->asic_funcs->reset((adev)) -#define amdgpu_asic_wait_for_mc_idle(adev) (adev)->asic_funcs->wait_for_mc_idle((adev)) #define amdgpu_asic_get_xclk(adev) (adev)->asic_funcs->get_xclk((adev)) #define amdgpu_asic_set_uvd_clocks(adev, v, d) (adev)->asic_funcs->set_uvd_clocks((adev), (v), (d)) #define amdgpu_asic_set_vce_clocks(adev, ev, ec) (adev)->asic_funcs->set_vce_clocks((adev), (ev), (ec)) #define amdgpu_asic_get_virtual_caps(adev) ((adev)->asic_funcs->get_virtual_caps((adev))) -#define amdgpu_asic_get_gpu_clock_counter(adev) (adev)->asic_funcs->get_gpu_clock_counter((adev)) #define amdgpu_asic_read_disabled_bios(adev) (adev)->asic_funcs->read_disabled_bios((adev)) #define amdgpu_asic_read_bios_from_rom(adev, b, l) (adev)->asic_funcs->read_bios_from_rom((adev), (b), (l)) #define amdgpu_asic_read_register(adev, se, sh, offset, v)((adev)->asic_funcs->read_register((adev), (se), (sh), (offset), (v))) #define amdgpu_gart_flush_gpu_tlb(adev, vmid) (adev)->gart.gart_funcs->flush_gpu_tlb((adev), (vmid)) #define amdgpu_gart_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gart.gart_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags)) #define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count))) -#define amdgpu_vm_write_pte(adev, ib, pa, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pa), (pe), (addr), (count), (incr), (flags))) +#define amdgpu_vm_write_pte(adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr))) #define amdgpu_vm_set_pte_pde(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags))) #define amdgpu_ring_parse_cs(r, p, ib) ((r)->funcs->parse_cs((p), (ib))) #define amdgpu_ring_test_ring(r) (r)->funcs->test_ring((r)) -#define amdgpu_ring_test_ib(r) (r)->funcs->test_ib((r)) +#define amdgpu_ring_test_ib(r, t) (r)->funcs->test_ib((r), (t)) #define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r)) #define amdgpu_ring_get_wptr(r) (r)->funcs->get_wptr((r)) #define amdgpu_ring_set_wptr(r) (r)->funcs->set_wptr((r)) @@ -2264,6 +2275,8 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #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_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev)) +#define amdgpu_gfx_select_se_sh(adev, se, sh, instance) (adev)->gfx.funcs->select_se_sh((adev), (se), (sh), (instance)) #define amdgpu_dpm_get_temperature(adev) \ ((adev)->pp_enabled ? \ @@ -2342,6 +2355,18 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_dpm_force_clock_level(adev, 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) + +#define amdgpu_dpm_set_sclk_od(adev, 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)) + +#define amdgpu_dpm_set_mclk_od(adev, 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)) @@ -2349,6 +2374,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) /* Common functions */ int amdgpu_gpu_reset(struct amdgpu_device *adev); +bool amdgpu_need_backup(struct amdgpu_device *adev); void amdgpu_pci_config_reset(struct amdgpu_device *adev); bool amdgpu_card_posted(struct amdgpu_device *adev); void amdgpu_update_display_priority(struct amdgpu_device *adev); @@ -2374,6 +2400,8 @@ uint32_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, 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_ttm_set_active_vram_size(struct amdgpu_device *adev, u64 size); +u64 amdgpu_ttm_get_gtt_mem_size(struct amdgpu_device *adev); +int amdgpu_ttm_global_init(struct amdgpu_device *adev); void amdgpu_program_register_sequence(struct amdgpu_device *adev, const u32 *registers, const u32 array_size); @@ -2383,9 +2411,13 @@ bool amdgpu_device_is_px(struct drm_device *dev); #if defined(CONFIG_VGA_SWITCHEROO) void amdgpu_register_atpx_handler(void); void amdgpu_unregister_atpx_handler(void); +bool amdgpu_has_atpx_dgpu_power_cntl(void); +bool amdgpu_is_atpx_hybrid(void); #else static inline void amdgpu_register_atpx_handler(void) {} static inline void amdgpu_unregister_atpx_handler(void) {} +static inline bool amdgpu_has_atpx_dgpu_power_cntl(void) { return false; } +static inline bool amdgpu_is_atpx_hybrid(void) { return false; } #endif /* diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c index 252edba16e36..892d60fb225b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c @@ -421,29 +421,6 @@ static int acp_suspend(void *handle) static int acp_resume(void *handle) { - int i, ret; - struct acp_pm_domain *apd; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - /* return early if no ACP */ - if (!adev->acp.acp_genpd) - return 0; - - /* SMU block will power on ACP irrespective of ACP runtime status. - * Power off explicitly based on genpd ACP runtime status so that ACP - * hw and ACP-genpd status are in sync. - * 'suspend_power_off' represents "Power status before system suspend" - */ - if (adev->acp.acp_genpd->gpd.suspend_power_off == true) { - apd = container_of(&adev->acp.acp_genpd->gpd, - struct acp_pm_domain, gpd); - - for (i = 4; i >= 0 ; i--) { - ret = acp_suspend_tile(apd->cgs_dev, ACP_TILE_P1 + i); - if (ret) - pr_err("ACP tile %d tile suspend failed\n", i); - } - } return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 32809f749903..d080d0807a5b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -240,8 +240,8 @@ uint64_t get_gpu_clock_counter(struct kgd_dev *kgd) { struct amdgpu_device *rdev = (struct amdgpu_device *)kgd; - if (rdev->asic_funcs->get_gpu_clock_counter) - return rdev->asic_funcs->get_gpu_clock_counter(rdev); + if (rdev->gfx.funcs->get_gpu_clock_counter) + return rdev->gfx.funcs->get_gpu_clock_counter(rdev); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c index 9df1bcb35bf0..1b621160b52e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c @@ -259,6 +259,33 @@ static const int object_connector_convert[] = { DRM_MODE_CONNECTOR_Unknown }; +bool amdgpu_atombios_has_dce_engine_info(struct amdgpu_device *adev) +{ + struct amdgpu_mode_info *mode_info = &adev->mode_info; + struct atom_context *ctx = mode_info->atom_context; + int index = GetIndexIntoMasterTable(DATA, Object_Header); + u16 size, data_offset; + u8 frev, crev; + ATOM_DISPLAY_OBJECT_PATH_TABLE *path_obj; + ATOM_OBJECT_HEADER *obj_header; + + if (!amdgpu_atom_parse_data_header(ctx, index, &size, &frev, &crev, &data_offset)) + return false; + + if (crev < 2) + return false; + + obj_header = (ATOM_OBJECT_HEADER *) (ctx->bios + data_offset); + path_obj = (ATOM_DISPLAY_OBJECT_PATH_TABLE *) + (ctx->bios + data_offset + + le16_to_cpu(obj_header->usDisplayPathTableOffset)); + + if (path_obj->ucNumOfDispPath) + return true; + else + return false; +} + bool amdgpu_atombios_get_connector_info_from_object_table(struct amdgpu_device *adev) { struct amdgpu_mode_info *mode_info = &adev->mode_info; @@ -551,28 +578,19 @@ int amdgpu_atombios_get_clock_info(struct amdgpu_device *adev) le16_to_cpu(firmware_info->info.usReferenceClock); ppll->reference_div = 0; - if (crev < 2) - ppll->pll_out_min = - le16_to_cpu(firmware_info->info.usMinPixelClockPLL_Output); - else - ppll->pll_out_min = - le32_to_cpu(firmware_info->info_12.ulMinPixelClockPLL_Output); + ppll->pll_out_min = + le32_to_cpu(firmware_info->info_12.ulMinPixelClockPLL_Output); ppll->pll_out_max = le32_to_cpu(firmware_info->info.ulMaxPixelClockPLL_Output); - if (crev >= 4) { - ppll->lcd_pll_out_min = - le16_to_cpu(firmware_info->info_14.usLcdMinPixelClockPLL_Output) * 100; - if (ppll->lcd_pll_out_min == 0) - ppll->lcd_pll_out_min = ppll->pll_out_min; - ppll->lcd_pll_out_max = - le16_to_cpu(firmware_info->info_14.usLcdMaxPixelClockPLL_Output) * 100; - if (ppll->lcd_pll_out_max == 0) - ppll->lcd_pll_out_max = ppll->pll_out_max; - } else { + ppll->lcd_pll_out_min = + le16_to_cpu(firmware_info->info_14.usLcdMinPixelClockPLL_Output) * 100; + if (ppll->lcd_pll_out_min == 0) ppll->lcd_pll_out_min = ppll->pll_out_min; + ppll->lcd_pll_out_max = + le16_to_cpu(firmware_info->info_14.usLcdMaxPixelClockPLL_Output) * 100; + if (ppll->lcd_pll_out_max == 0) ppll->lcd_pll_out_max = ppll->pll_out_max; - } if (ppll->pll_out_min == 0) ppll->pll_out_min = 64800; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h index 8c2e69661799..15dd43ec38bb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h @@ -140,6 +140,8 @@ struct amdgpu_i2c_bus_rec amdgpu_atombios_lookup_i2c_gpio(struct amdgpu_device * uint8_t id); void amdgpu_atombios_i2c_init(struct amdgpu_device *adev); +bool amdgpu_atombios_has_dce_engine_info(struct amdgpu_device *adev); + bool amdgpu_atombios_get_connector_info_from_object_table(struct amdgpu_device *adev); int amdgpu_atombios_get_clock_info(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c index 35a1248aaa77..49de92600074 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c @@ -10,6 +10,7 @@ #include <linux/slab.h> #include <linux/acpi.h> #include <linux/pci.h> +#include <linux/delay.h> #include "amd_acpi.h" @@ -27,6 +28,7 @@ struct amdgpu_atpx_functions { struct amdgpu_atpx { acpi_handle handle; struct amdgpu_atpx_functions functions; + bool is_hybrid; }; static struct amdgpu_atpx_priv { @@ -63,6 +65,14 @@ bool amdgpu_has_atpx(void) { return amdgpu_atpx_priv.atpx_detected; } +bool amdgpu_has_atpx_dgpu_power_cntl(void) { + return amdgpu_atpx_priv.atpx.functions.power_cntl; +} + +bool amdgpu_is_atpx_hybrid(void) { + return amdgpu_atpx_priv.atpx.is_hybrid; +} + /** * amdgpu_atpx_call - call an ATPX method * @@ -142,18 +152,12 @@ static void amdgpu_atpx_parse_functions(struct amdgpu_atpx_functions *f, u32 mas */ static int amdgpu_atpx_validate(struct amdgpu_atpx *atpx) { - /* make sure required functions are enabled */ - /* dGPU power control is required */ - if (atpx->functions.power_cntl == false) { - printk("ATPX dGPU power cntl not present, forcing\n"); - atpx->functions.power_cntl = true; - } + u32 valid_bits = 0; if (atpx->functions.px_params) { union acpi_object *info; struct atpx_px_params output; size_t size; - u32 valid_bits; info = amdgpu_atpx_call(atpx->handle, ATPX_FUNCTION_GET_PX_PARAMETERS, NULL); if (!info) @@ -172,19 +176,43 @@ static int amdgpu_atpx_validate(struct amdgpu_atpx *atpx) memcpy(&output, info->buffer.pointer, size); valid_bits = output.flags & output.valid_flags; - /* if separate mux flag is set, mux controls are required */ - if (valid_bits & ATPX_SEPARATE_MUX_FOR_I2C) { - atpx->functions.i2c_mux_cntl = true; - atpx->functions.disp_mux_cntl = true; - } - /* if any outputs are muxed, mux controls are required */ - if (valid_bits & (ATPX_CRT1_RGB_SIGNAL_MUXED | - ATPX_TV_SIGNAL_MUXED | - ATPX_DFP_SIGNAL_MUXED)) - atpx->functions.disp_mux_cntl = true; kfree(info); } + + /* if separate mux flag is set, mux controls are required */ + if (valid_bits & ATPX_SEPARATE_MUX_FOR_I2C) { + atpx->functions.i2c_mux_cntl = true; + atpx->functions.disp_mux_cntl = true; + } + /* if any outputs are muxed, mux controls are required */ + if (valid_bits & (ATPX_CRT1_RGB_SIGNAL_MUXED | + ATPX_TV_SIGNAL_MUXED | + ATPX_DFP_SIGNAL_MUXED)) + atpx->functions.disp_mux_cntl = true; + + + /* some bioses set these bits rather than flagging power_cntl as supported */ + if (valid_bits & (ATPX_DYNAMIC_PX_SUPPORTED | + ATPX_DYNAMIC_DGPU_POWER_OFF_SUPPORTED)) + atpx->functions.power_cntl = true; + + atpx->is_hybrid = false; + if (valid_bits & ATPX_MS_HYBRID_GFX_SUPPORTED) { + printk("ATPX Hybrid Graphics\n"); +#if 1 + /* This is a temporary hack until the D3 cold support + * makes it upstream. The ATPX power_control method seems + * to still work on even if the system should be using + * the new standardized hybrid D3 cold ACPI interface. + */ + atpx->functions.power_cntl = true; +#else + atpx->functions.power_cntl = false; +#endif + atpx->is_hybrid = true; + } + return 0; } @@ -259,6 +287,10 @@ static int amdgpu_atpx_set_discrete_state(struct amdgpu_atpx *atpx, u8 state) if (!info) return -EIO; kfree(info); + + /* 200ms delay is required after off */ + if (state == 0) + msleep(200); } return 0; } @@ -507,7 +539,6 @@ static int amdgpu_atpx_get_client_id(struct pci_dev *pdev) static const struct vga_switcheroo_handler amdgpu_atpx_handler = { .switchto = amdgpu_atpx_switchto, .power_state = amdgpu_atpx_power_state, - .init = amdgpu_atpx_init, .get_client_id = amdgpu_atpx_get_client_id, }; @@ -542,6 +573,7 @@ static bool amdgpu_atpx_detect(void) printk(KERN_INFO "vga_switcheroo: detected switching method %s handle\n", acpi_method_name); amdgpu_atpx_priv.atpx_detected = true; + amdgpu_atpx_init(); return true; } return false; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c index 33e47a43ae32..345305235349 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c @@ -39,7 +39,8 @@ static int amdgpu_benchmark_do_move(struct amdgpu_device *adev, unsigned size, start_jiffies = jiffies; 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); + r = amdgpu_copy_buffer(ring, saddr, daddr, size, NULL, &fence, + false); if (r) goto exit_do_move; r = fence_wait(fence, false); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c index 99ca75baa47d..2b6afe123f3d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c @@ -352,22 +352,22 @@ bool amdgpu_get_bios(struct amdgpu_device *adev) uint16_t tmp, bios_header_start; r = amdgpu_atrm_get_bios(adev); - if (r == false) + if (!r) r = amdgpu_acpi_vfct_bios(adev); - if (r == false) + if (!r) r = igp_read_bios_from_vram(adev); - if (r == false) + if (!r) r = amdgpu_read_bios(adev); - if (r == false) { + if (!r) { r = amdgpu_read_bios_from_rom(adev); } - if (r == false) { + if (!r) { r = amdgpu_read_disabled_bios(adev); } - if (r == false) { + if (!r) { r = amdgpu_read_platform_bios(adev); } - if (r == false || adev->bios == NULL) { + if (!r || adev->bios == NULL) { DRM_ERROR("Unable to locate a BIOS ROM\n"); adev->bios = NULL; return false; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c index 823bf5e0b0c8..651115dcce12 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c @@ -94,6 +94,7 @@ static int amdgpu_bo_list_set(struct amdgpu_device *adev, unsigned last_entry = 0, first_userptr = num_entries; unsigned i; int r; + unsigned long total_size = 0; array = drm_malloc_ab(num_entries, sizeof(struct amdgpu_bo_list_entry)); if (!array) @@ -140,6 +141,7 @@ static int amdgpu_bo_list_set(struct amdgpu_device *adev, if (entry->robj->prefered_domains == AMDGPU_GEM_DOMAIN_OA) oa_obj = entry->robj; + total_size += amdgpu_bo_size(entry->robj); trace_amdgpu_bo_list_set(list, entry->robj); } @@ -155,6 +157,7 @@ static int amdgpu_bo_list_set(struct amdgpu_device *adev, list->array = array; list->num_entries = num_entries; + trace_amdgpu_cs_bo_status(list->num_entries, total_size); return 0; error_free: diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c index cf6f49fc1c75..bc0440f7a31d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c @@ -312,6 +312,8 @@ static uint32_t amdgpu_cgs_read_ind_register(struct cgs_device *cgs_device, return RREG32_UVD_CTX(index); case CGS_IND_REG__DIDT: return RREG32_DIDT(index); + case CGS_IND_REG_GC_CAC: + return RREG32_GC_CAC(index); case CGS_IND_REG__AUDIO_ENDPT: DRM_ERROR("audio endpt register access not implemented.\n"); return 0; @@ -336,6 +338,8 @@ static void amdgpu_cgs_write_ind_register(struct cgs_device *cgs_device, return WREG32_UVD_CTX(index, value); case CGS_IND_REG__DIDT: return WREG32_DIDT(index, value); + case CGS_IND_REG_GC_CAC: + return WREG32_GC_CAC(index, value); case CGS_IND_REG__AUDIO_ENDPT: DRM_ERROR("audio endpt register access not implemented.\n"); return; @@ -748,6 +752,9 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, if (!adev->pm.fw) { switch (adev->asic_type) { + case CHIP_TOPAZ: + strcpy(fw_name, "amdgpu/topaz_smc.bin"); + break; case CHIP_TONGA: strcpy(fw_name, "amdgpu/tonga_smc.bin"); break; @@ -787,6 +794,7 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, } hdr = (const struct smc_firmware_header_v1_0 *) adev->pm.fw->data; + amdgpu_ucode_print_smc_hdr(&hdr->header); adev->pm.fw_version = le32_to_cpu(hdr->header.ucode_version); ucode_size = le32_to_cpu(hdr->header.ucode_size_bytes); ucode_start_address = le32_to_cpu(hdr->ucode_start_addr); @@ -795,13 +803,14 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, info->version = adev->pm.fw_version; info->image_size = ucode_size; + info->ucode_start_address = ucode_start_address; info->kptr = (void *)src; } return 0; } static int amdgpu_cgs_query_system_info(struct cgs_device *cgs_device, - struct cgs_system_info *sys_info) + struct cgs_system_info *sys_info) { CGS_FUNC_ADEV; @@ -821,6 +830,12 @@ static int amdgpu_cgs_query_system_info(struct cgs_device *cgs_device, case CGS_SYSTEM_INFO_PCIE_MLW: sys_info->value = adev->pm.pcie_mlw_mask; break; + case CGS_SYSTEM_INFO_PCIE_DEV: + sys_info->value = adev->pdev->device; + break; + case CGS_SYSTEM_INFO_PCIE_REV: + sys_info->value = adev->pdev->revision; + break; case CGS_SYSTEM_INFO_CG_FLAGS: sys_info->value = adev->cg_flags; break; @@ -830,6 +845,9 @@ static int amdgpu_cgs_query_system_info(struct cgs_device *cgs_device, case CGS_SYSTEM_INFO_GFX_CU_INFO: sys_info->value = adev->gfx.cu_info.number; break; + case CGS_SYSTEM_INFO_GFX_SE_INFO: + sys_info->value = adev->gfx.config.max_shader_engines; + break; default: return -ENODEV; } @@ -903,14 +921,12 @@ static int amdgpu_cgs_acpi_eval_object(struct cgs_device *cgs_device, acpi_handle handle; struct acpi_object_list input; struct acpi_buffer output = { ACPI_ALLOCATE_BUFFER, NULL }; - union acpi_object *params = NULL; - union acpi_object *obj = NULL; + union acpi_object *params, *obj; uint8_t name[5] = {'\0'}; - struct cgs_acpi_method_argument *argument = NULL; + struct cgs_acpi_method_argument *argument; uint32_t i, count; acpi_status status; - int result = 0; - uint32_t func_no = 0xFFFFFFFF; + int result; handle = ACPI_HANDLE(&adev->pdev->dev); if (!handle) @@ -927,7 +943,6 @@ static int amdgpu_cgs_acpi_eval_object(struct cgs_device *cgs_device, if (info->pinput_argument == NULL) return -EINVAL; argument = info->pinput_argument; - func_no = argument->value; for (i = 0; i < info->input_count; i++) { if (((argument->type == ACPI_TYPE_STRING) || (argument->type == ACPI_TYPE_BUFFER)) && @@ -972,11 +987,11 @@ static int amdgpu_cgs_acpi_eval_object(struct cgs_device *cgs_device, params->integer.value = argument->value; break; case ACPI_TYPE_STRING: - params->string.length = argument->method_length; + params->string.length = argument->data_length; params->string.pointer = argument->pointer; break; case ACPI_TYPE_BUFFER: - params->buffer.length = argument->method_length; + params->buffer.length = argument->data_length; params->buffer.pointer = argument->pointer; break; default: @@ -996,7 +1011,7 @@ static int amdgpu_cgs_acpi_eval_object(struct cgs_device *cgs_device, if (ACPI_FAILURE(status)) { result = -EIO; - goto error; + goto free_input; } /* return the output info */ @@ -1006,7 +1021,7 @@ static int amdgpu_cgs_acpi_eval_object(struct cgs_device *cgs_device, if ((obj->type != ACPI_TYPE_PACKAGE) || (obj->package.count != count)) { result = -EIO; - goto error; + goto free_obj; } params = obj->package.elements; } else @@ -1014,13 +1029,13 @@ static int amdgpu_cgs_acpi_eval_object(struct cgs_device *cgs_device, if (params == NULL) { result = -EIO; - goto error; + goto free_obj; } for (i = 0; i < count; i++) { if (argument->type != params->type) { result = -EIO; - goto error; + goto free_obj; } switch (params->type) { case ACPI_TYPE_INTEGER: @@ -1030,7 +1045,7 @@ static int amdgpu_cgs_acpi_eval_object(struct cgs_device *cgs_device, if ((params->string.length != argument->data_length) || (params->string.pointer == NULL)) { result = -EIO; - goto error; + goto free_obj; } strncpy(argument->pointer, params->string.pointer, @@ -1039,7 +1054,7 @@ static int amdgpu_cgs_acpi_eval_object(struct cgs_device *cgs_device, case ACPI_TYPE_BUFFER: if (params->buffer.pointer == NULL) { result = -EIO; - goto error; + goto free_obj; } memcpy(argument->pointer, params->buffer.pointer, @@ -1052,9 +1067,10 @@ static int amdgpu_cgs_acpi_eval_object(struct cgs_device *cgs_device, params++; } -error: - if (obj != NULL) - kfree(obj); + result = 0; +free_obj: + kfree(obj); +free_input: kfree((void *)input.pointer); return result; } @@ -1066,7 +1082,7 @@ static int amdgpu_cgs_acpi_eval_object(struct cgs_device *cgs_device, } #endif -int amdgpu_cgs_call_acpi_method(struct cgs_device *cgs_device, +static int amdgpu_cgs_call_acpi_method(struct cgs_device *cgs_device, uint32_t acpi_method, uint32_t acpi_function, void *pinput, void *poutput, @@ -1079,17 +1095,14 @@ int amdgpu_cgs_call_acpi_method(struct cgs_device *cgs_device, struct cgs_acpi_method_info info = {0}; acpi_input[0].type = CGS_ACPI_TYPE_INTEGER; - acpi_input[0].method_length = sizeof(uint32_t); acpi_input[0].data_length = sizeof(uint32_t); acpi_input[0].value = acpi_function; acpi_input[1].type = CGS_ACPI_TYPE_BUFFER; - acpi_input[1].method_length = CGS_ACPI_MAX_BUFFER_SIZE; acpi_input[1].data_length = input_size; acpi_input[1].pointer = pinput; acpi_output.type = CGS_ACPI_TYPE_BUFFER; - acpi_output.method_length = CGS_ACPI_MAX_BUFFER_SIZE; acpi_output.data_length = output_size; acpi_output.pointer = poutput; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index cb07da41152b..319a5e1d9389 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -1504,6 +1504,86 @@ static const struct drm_connector_funcs amdgpu_connector_edp_funcs = { .force = amdgpu_connector_dvi_force, }; +static struct drm_encoder * +amdgpu_connector_virtual_encoder(struct drm_connector *connector) +{ + int enc_id = connector->encoder_ids[0]; + struct drm_encoder *encoder; + int i; + for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) { + if (connector->encoder_ids[i] == 0) + break; + + encoder = drm_encoder_find(connector->dev, connector->encoder_ids[i]); + if (!encoder) + continue; + + if (encoder->encoder_type == DRM_MODE_ENCODER_VIRTUAL) + return encoder; + } + + /* pick the first one */ + if (enc_id) + return drm_encoder_find(connector->dev, enc_id); + return NULL; +} + +static int amdgpu_connector_virtual_get_modes(struct drm_connector *connector) +{ + struct drm_encoder *encoder = amdgpu_connector_best_single_encoder(connector); + + if (encoder) { + amdgpu_connector_add_common_modes(encoder, connector); + } + + return 0; +} + +static int amdgpu_connector_virtual_mode_valid(struct drm_connector *connector, + struct drm_display_mode *mode) +{ + return MODE_OK; +} + +int amdgpu_connector_virtual_dpms(struct drm_connector *connector, int mode) +{ + return 0; +} + +static enum drm_connector_status + +amdgpu_connector_virtual_detect(struct drm_connector *connector, bool force) +{ + return connector_status_connected; +} + +int amdgpu_connector_virtual_set_property(struct drm_connector *connector, + struct drm_property *property, + uint64_t val) +{ + return 0; +} + +static void amdgpu_connector_virtual_force(struct drm_connector *connector) +{ + return; +} + +static const struct drm_connector_helper_funcs amdgpu_connector_virtual_helper_funcs = { + .get_modes = amdgpu_connector_virtual_get_modes, + .mode_valid = amdgpu_connector_virtual_mode_valid, + .best_encoder = amdgpu_connector_virtual_encoder, +}; + +static const struct drm_connector_funcs amdgpu_connector_virtual_funcs = { + .dpms = amdgpu_connector_virtual_dpms, + .detect = amdgpu_connector_virtual_detect, + .fill_modes = drm_helper_probe_single_connector_modes, + .set_property = amdgpu_connector_virtual_set_property, + .destroy = amdgpu_connector_destroy, + .force = amdgpu_connector_virtual_force, +}; + void amdgpu_connector_add(struct amdgpu_device *adev, uint32_t connector_id, @@ -1690,7 +1770,6 @@ amdgpu_connector_add(struct amdgpu_device *adev, DRM_MODE_SCALE_NONE); /* no HPD on analog connectors */ amdgpu_connector->hpd.hpd = AMDGPU_HPD_NONE; - connector->polled = DRM_CONNECTOR_POLL_CONNECT; connector->interlace_allowed = true; connector->doublescan_allowed = true; break; @@ -1889,12 +1968,25 @@ amdgpu_connector_add(struct amdgpu_device *adev, connector->interlace_allowed = false; connector->doublescan_allowed = false; break; + case DRM_MODE_CONNECTOR_VIRTUAL: + amdgpu_dig_connector = kzalloc(sizeof(struct amdgpu_connector_atom_dig), GFP_KERNEL); + if (!amdgpu_dig_connector) + goto failed; + amdgpu_connector->con_priv = amdgpu_dig_connector; + drm_connector_init(dev, &amdgpu_connector->base, &amdgpu_connector_virtual_funcs, connector_type); + drm_connector_helper_add(&amdgpu_connector->base, &amdgpu_connector_virtual_helper_funcs); + subpixel_order = SubPixelHorizontalRGB; + connector->interlace_allowed = false; + connector->doublescan_allowed = false; + break; } } if (amdgpu_connector->hpd.hpd == AMDGPU_HPD_NONE) { - if (i2c_bus->valid) - connector->polled = DRM_CONNECTOR_POLL_CONNECT; + if (i2c_bus->valid) { + connector->polled = DRM_CONNECTOR_POLL_CONNECT | + DRM_CONNECTOR_POLL_DISCONNECT; + } } else connector->polled = DRM_CONNECTOR_POLL_HPD; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 9bc8f1d99733..d80e5d3a4add 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -216,11 +216,8 @@ int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data) if (ret) goto free_all_kdata; - if (p->uf_entry.robj) { - p->job->uf_bo = amdgpu_bo_ref(p->uf_entry.robj); - p->job->uf_offset = uf_offset; - } - + if (p->uf_entry.robj) + p->job->uf_addr = uf_offset; kfree(chunk_array); return 0; @@ -290,18 +287,56 @@ static u64 amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev) return max(bytes_moved_threshold, 1024*1024ull); } +static int amdgpu_cs_bo_validate(struct amdgpu_cs_parser *p, + struct amdgpu_bo *bo) +{ + u64 initial_bytes_moved; + uint32_t domain; + int r; + + if (bo->pin_count) + return 0; + + /* Avoid moving this one if we have moved too many buffers + * for this IB already. + * + * Note that this allows moving at least one buffer of + * any size, because it doesn't take the current "bo" + * into account. We don't want to disallow buffer moves + * completely. + */ + if (p->bytes_moved <= p->bytes_moved_threshold) + domain = bo->prefered_domains; + else + domain = bo->allowed_domains; + +retry: + amdgpu_ttm_placement_from_domain(bo, domain); + initial_bytes_moved = atomic64_read(&bo->adev->num_bytes_moved); + r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); + p->bytes_moved += atomic64_read(&bo->adev->num_bytes_moved) - + initial_bytes_moved; + + if (unlikely(r)) { + if (r != -ERESTARTSYS && domain != bo->allowed_domains) { + domain = bo->allowed_domains; + goto retry; + } + } + + return r; +} + int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p, struct list_head *validated) { struct amdgpu_bo_list_entry *lobj; - u64 initial_bytes_moved; int r; list_for_each_entry(lobj, validated, tv.head) { struct amdgpu_bo *bo = lobj->robj; bool binding_userptr = false; struct mm_struct *usermm; - uint32_t domain; usermm = amdgpu_ttm_tt_get_usermm(bo->tbo.ttm); if (usermm && usermm != current->mm) @@ -316,35 +351,13 @@ int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p, binding_userptr = true; } - if (bo->pin_count) - continue; - - /* Avoid moving this one if we have moved too many buffers - * for this IB already. - * - * Note that this allows moving at least one buffer of - * any size, because it doesn't take the current "bo" - * into account. We don't want to disallow buffer moves - * completely. - */ - if (p->bytes_moved <= p->bytes_moved_threshold) - domain = bo->prefered_domains; - else - domain = bo->allowed_domains; - - retry: - amdgpu_ttm_placement_from_domain(bo, domain); - initial_bytes_moved = atomic64_read(&bo->adev->num_bytes_moved); - r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); - p->bytes_moved += atomic64_read(&bo->adev->num_bytes_moved) - - initial_bytes_moved; - - if (unlikely(r)) { - if (r != -ERESTARTSYS && domain != bo->allowed_domains) { - domain = bo->allowed_domains; - goto retry; - } + r = amdgpu_cs_bo_validate(p, bo); + if (r) return r; + if (bo->shadow) { + r = amdgpu_cs_bo_validate(p, bo); + if (r) + return r; } if (binding_userptr) { @@ -389,8 +402,10 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, r = ttm_eu_reserve_buffers(&p->ticket, &p->validated, true, &duplicates); - if (unlikely(r != 0)) + if (unlikely(r != 0)) { + DRM_ERROR("ttm_eu_reserve_buffers failed.\n"); goto error_free_pages; + } /* Without a BO list we don't have userptr BOs */ if (!p->bo_list) @@ -430,9 +445,10 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, /* Unreserve everything again. */ ttm_eu_backoff_reservation(&p->ticket, &p->validated); - /* We tried to often, just abort */ + /* We tried too many times, just abort */ if (!--tries) { r = -EDEADLK; + DRM_ERROR("deadlock in %s\n", __func__); goto error_free_pages; } @@ -444,11 +460,13 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, sizeof(struct page*)); if (!e->user_pages) { r = -ENOMEM; + DRM_ERROR("calloc failure in %s\n", __func__); goto error_free_pages; } r = amdgpu_ttm_tt_get_user_pages(ttm, e->user_pages); if (r) { + DRM_ERROR("amdgpu_ttm_tt_get_user_pages failed.\n"); drm_free_large(e->user_pages); e->user_pages = NULL; goto error_free_pages; @@ -459,18 +477,25 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, list_splice(&need_pages, &p->validated); } - amdgpu_vm_get_pt_bos(&fpriv->vm, &duplicates); + amdgpu_vm_get_pt_bos(p->adev, &fpriv->vm, &duplicates); p->bytes_moved_threshold = amdgpu_cs_get_threshold_for_moves(p->adev); p->bytes_moved = 0; r = amdgpu_cs_list_validate(p, &duplicates); - if (r) + if (r) { + DRM_ERROR("amdgpu_cs_list_validate(duplicates) failed.\n"); goto error_validate; + } r = amdgpu_cs_list_validate(p, &p->validated); - if (r) + if (r) { + DRM_ERROR("amdgpu_cs_list_validate(validated) failed.\n"); goto error_validate; + } + + fpriv->vm.last_eviction_counter = + atomic64_read(&p->adev->num_evictions); if (p->bo_list) { struct amdgpu_bo *gds = p->bo_list->gds_obj; @@ -499,6 +524,9 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, } } + if (p->uf_entry.robj) + p->job->uf_addr += amdgpu_bo_gpu_offset(p->uf_entry.robj); + error_validate: if (r) { amdgpu_vm_move_pt_bos_in_lru(p->adev, &fpriv->vm); @@ -614,7 +642,7 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p, if (bo_va == NULL) continue; - r = amdgpu_vm_bo_update(adev, bo_va, &bo->tbo.mem); + r = amdgpu_vm_bo_update(adev, bo_va, false); if (r) return r; @@ -653,18 +681,21 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev, /* Only for UVD/VCE VM emulation */ if (ring->funcs->parse_cs) { + p->job->vm = NULL; for (i = 0; i < p->job->num_ibs; i++) { r = amdgpu_ring_parse_cs(ring, p, i); if (r) return r; } - } + } else { + p->job->vm_pd_addr = amdgpu_bo_gpu_offset(vm->page_directory); - r = amdgpu_bo_vm_update_pte(p, vm); - if (!r) - amdgpu_cs_sync_rings(p); + r = amdgpu_bo_vm_update_pte(p, vm); + if (r) + return r; + } - return r; + return amdgpu_cs_sync_rings(p); } static int amdgpu_cs_handle_lockup(struct amdgpu_device *adev, int r) @@ -761,7 +792,7 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev, } /* UVD & VCE fw doesn't support user fences */ - if (parser->job->uf_bo && ( + if (parser->job->uf_addr && ( parser->job->ring->type == AMDGPU_RING_TYPE_UVD || parser->job->ring->type == AMDGPU_RING_TYPE_VCE)) return -EINVAL; @@ -830,17 +861,13 @@ 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 fence *fence; struct amdgpu_job *job; int r; job = p->job; p->job = NULL; - r = amd_sched_job_init(&job->base, &ring->sched, - entity, amdgpu_job_timeout_func, - amdgpu_job_free_func, - p->filp, &fence); + r = amd_sched_job_init(&job->base, &ring->sched, entity, p->filp); if (r) { amdgpu_job_free(job); return r; @@ -848,9 +875,10 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, job->owner = p->filp; job->ctx = entity->fence_context; - p->fence = fence_get(fence); - cs->out.handle = amdgpu_ctx_add_fence(p->ctx, ring, fence); + p->fence = fence_get(&job->base.s_fence->finished); + cs->out.handle = amdgpu_ctx_add_fence(p->ctx, ring, p->fence); job->uf_sequence = cs->out.handle; + amdgpu_job_free_resources(job); trace_amdgpu_cs_ioctl(job); amd_sched_entity_push_job(&job->base); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 6e920086af46..c38dc47cd767 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -25,6 +25,7 @@ * Alex Deucher * Jerome Glisse */ +#include <linux/kthread.h> #include <linux/console.h> #include <linux/slab.h> #include <linux/debugfs.h> @@ -35,6 +36,7 @@ #include <linux/vga_switcheroo.h> #include <linux/efi.h> #include "amdgpu.h" +#include "amdgpu_trace.h" #include "amdgpu_i2c.h" #include "atom.h" #include "amdgpu_atombios.h" @@ -44,6 +46,7 @@ #endif #include "vi.h" #include "bif/bif_4_1_d.h" +#include <linux/pci.h> static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev); static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev); @@ -79,24 +82,27 @@ bool amdgpu_device_is_px(struct drm_device *dev) uint32_t amdgpu_mm_rreg(struct amdgpu_device *adev, uint32_t reg, bool always_indirect) { + uint32_t ret; + if ((reg * 4) < adev->rmmio_size && !always_indirect) - return readl(((void __iomem *)adev->rmmio) + (reg * 4)); + ret = readl(((void __iomem *)adev->rmmio) + (reg * 4)); else { unsigned long flags; - uint32_t ret; spin_lock_irqsave(&adev->mmio_idx_lock, flags); writel((reg * 4), ((void __iomem *)adev->rmmio) + (mmMM_INDEX * 4)); ret = readl(((void __iomem *)adev->rmmio) + (mmMM_DATA * 4)); spin_unlock_irqrestore(&adev->mmio_idx_lock, flags); - - return ret; } + trace_amdgpu_mm_rreg(adev->pdev->device, reg, ret); + return ret; } void amdgpu_mm_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, bool always_indirect) { + trace_amdgpu_mm_wreg(adev->pdev->device, reg, v); + if ((reg * 4) < adev->rmmio_size && !always_indirect) writel(v, ((void __iomem *)adev->rmmio) + (reg * 4)); else { @@ -1070,11 +1076,14 @@ int amdgpu_set_clockgating_state(struct amdgpu_device *adev, int i, r = 0; for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_block_status[i].valid) + continue; if (adev->ip_blocks[i].type == block_type) { r = adev->ip_blocks[i].funcs->set_clockgating_state((void *)adev, state); if (r) return r; + break; } } return r; @@ -1087,16 +1096,53 @@ int amdgpu_set_powergating_state(struct amdgpu_device *adev, int i, r = 0; for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_block_status[i].valid) + continue; if (adev->ip_blocks[i].type == block_type) { r = adev->ip_blocks[i].funcs->set_powergating_state((void *)adev, state); if (r) return r; + break; } } return r; } +int amdgpu_wait_for_idle(struct amdgpu_device *adev, + enum amd_ip_block_type block_type) +{ + int i, r; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_block_status[i].valid) + continue; + if (adev->ip_blocks[i].type == block_type) { + r = adev->ip_blocks[i].funcs->wait_for_idle((void *)adev); + if (r) + return r; + break; + } + } + return 0; + +} + +bool amdgpu_is_idle(struct amdgpu_device *adev, + enum amd_ip_block_type block_type) +{ + int i; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_block_status[i].valid) + continue; + if (adev->ip_blocks[i].type == block_type) + return adev->ip_blocks[i].funcs->is_idle((void *)adev); + } + return true; + +} + const struct amdgpu_ip_block_version * amdgpu_get_ip_block( struct amdgpu_device *adev, enum amd_ip_block_type type) @@ -1136,10 +1182,38 @@ int amdgpu_ip_block_version_cmp(struct amdgpu_device *adev, return 1; } +static void amdgpu_whether_enable_virtual_display(struct amdgpu_device *adev) +{ + adev->enable_virtual_display = false; + + if (amdgpu_virtual_display) { + struct drm_device *ddev = adev->ddev; + const char *pci_address_name = pci_name(ddev->pdev); + char *pciaddstr, *pciaddstr_tmp, *pciaddname; + + pciaddstr = kstrdup(amdgpu_virtual_display, GFP_KERNEL); + pciaddstr_tmp = pciaddstr; + while ((pciaddname = strsep(&pciaddstr_tmp, ";"))) { + if (!strcmp(pci_address_name, pciaddname)) { + adev->enable_virtual_display = true; + break; + } + } + + DRM_INFO("virtual display string:%s, %s:virtual_display:%d\n", + amdgpu_virtual_display, pci_address_name, + adev->enable_virtual_display); + + kfree(pciaddstr); + } +} + static int amdgpu_early_init(struct amdgpu_device *adev) { int i, r; + amdgpu_whether_enable_virtual_display(adev); + switch (adev->asic_type) { case CHIP_TOPAZ: case CHIP_TONGA: @@ -1209,6 +1283,9 @@ static int amdgpu_early_init(struct amdgpu_device *adev) } } + adev->cg_flags &= amdgpu_cg_mask; + adev->pg_flags &= amdgpu_pg_mask; + return 0; } @@ -1440,9 +1517,12 @@ int amdgpu_device_init(struct amdgpu_device *adev, adev->uvd_ctx_wreg = &amdgpu_invalid_wreg; adev->didt_rreg = &amdgpu_invalid_rreg; adev->didt_wreg = &amdgpu_invalid_wreg; + adev->gc_cac_rreg = &amdgpu_invalid_rreg; + adev->gc_cac_wreg = &amdgpu_invalid_wreg; 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); @@ -1467,8 +1547,12 @@ int amdgpu_device_init(struct amdgpu_device *adev, spin_lock_init(&adev->pcie_idx_lock); 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->audio_endpt_idx_lock); + INIT_LIST_HEAD(&adev->shadow_list); + mutex_init(&adev->shadow_list_lock); + adev->rmmio_base = pci_resource_start(adev->pdev, 5); adev->rmmio_size = pci_resource_len(adev->pdev, 5); adev->rmmio = ioremap(adev->rmmio_base, adev->rmmio_size); @@ -1511,17 +1595,20 @@ int amdgpu_device_init(struct amdgpu_device *adev, vga_switcheroo_init_domain_pm_ops(adev->dev, &adev->vga_pm_domain); /* Read BIOS */ - if (!amdgpu_get_bios(adev)) - return -EINVAL; + if (!amdgpu_get_bios(adev)) { + r = -EINVAL; + goto failed; + } /* Must be an ATOMBIOS */ if (!adev->is_atom_bios) { dev_err(adev->dev, "Expecting atombios for GPU\n"); - return -EINVAL; + r = -EINVAL; + goto failed; } r = amdgpu_atombios_init(adev); if (r) { dev_err(adev->dev, "amdgpu_atombios_init failed\n"); - return r; + goto failed; } /* See if the asic supports SR-IOV */ @@ -1538,7 +1625,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, !(adev->virtualization.caps & AMDGPU_VIRT_CAPS_SRIOV_EN))) { if (!adev->bios) { dev_err(adev->dev, "Card not posted and no BIOS - ignoring\n"); - return -EINVAL; + r = -EINVAL; + goto failed; } DRM_INFO("GPU not posted. posting now...\n"); amdgpu_atom_asic_init(adev->mode_info.atom_context); @@ -1548,7 +1636,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_atombios_get_clock_info(adev); if (r) { dev_err(adev->dev, "amdgpu_atombios_get_clock_info failed\n"); - return r; + goto failed; } /* init i2c buses */ amdgpu_atombios_i2c_init(adev); @@ -1557,7 +1645,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_fence_driver_init(adev); if (r) { dev_err(adev->dev, "amdgpu_fence_driver_init failed\n"); - return r; + goto failed; } /* init the mode config */ @@ -1567,7 +1655,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, if (r) { dev_err(adev->dev, "amdgpu_init failed\n"); amdgpu_fini(adev); - return r; + goto failed; } adev->accel_working = true; @@ -1577,7 +1665,7 @@ 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); - return r; + goto failed; } r = amdgpu_ib_ring_tests(adev); @@ -1594,6 +1682,12 @@ int amdgpu_device_init(struct amdgpu_device *adev, DRM_ERROR("registering register debugfs failed (%d).\n", r); } + r = amdgpu_debugfs_firmware_init(adev); + if (r) { + DRM_ERROR("registering firmware debugfs failed (%d).\n", r); + return r; + } + if ((amdgpu_testing & 1)) { if (adev->accel_working) amdgpu_test_moves(adev); @@ -1619,10 +1713,15 @@ int amdgpu_device_init(struct amdgpu_device *adev, r = amdgpu_late_init(adev); if (r) { dev_err(adev->dev, "amdgpu_late_init failed\n"); - return r; + goto failed; } return 0; + +failed: + if (runtime) + vga_switcheroo_fini_domain_pm_ops(adev->dev); + return r; } static void amdgpu_debugfs_remove_files(struct amdgpu_device *adev); @@ -1645,6 +1744,7 @@ void amdgpu_device_fini(struct amdgpu_device *adev) amdgpu_bo_evict_vram(adev); amdgpu_ib_pool_fini(adev); amdgpu_fence_driver_fini(adev); + drm_crtc_force_disable_all(adev->ddev); amdgpu_fbdev_fini(adev); r = amdgpu_fini(adev); kfree(adev->ip_block_status); @@ -1656,6 +1756,8 @@ void amdgpu_device_fini(struct amdgpu_device *adev) kfree(adev->bios); adev->bios = NULL; vga_switcheroo_unregister_client(adev->pdev); + if (adev->flags & AMD_IS_PX) + vga_switcheroo_fini_domain_pm_ops(adev->dev); vga_client_register(adev->pdev, NULL, NULL, NULL); if (adev->rio_mem) pci_iounmap(adev->pdev, adev->rio_mem); @@ -1841,7 +1943,23 @@ int amdgpu_resume_kms(struct drm_device *dev, bool resume, bool fbcon) } drm_kms_helper_poll_enable(dev); + + /* + * Most of the connector probing functions try to acquire runtime pm + * refs to ensure that the GPU is powered on when connector polling is + * performed. Since we're calling this from a runtime PM callback, + * trying to acquire rpm refs will cause us to deadlock. + * + * Since we're guaranteed to be holding the rpm lock, it's safe to + * temporarily disable the rpm helpers so this doesn't deadlock us. + */ +#ifdef CONFIG_PM + dev->dev->power.disable_depth++; +#endif drm_helper_hpd_irq_event(dev); +#ifdef CONFIG_PM + dev->dev->power.disable_depth--; +#endif if (fbcon) { amdgpu_fbdev_set_suspend(adev, 0); @@ -1851,6 +1969,126 @@ int amdgpu_resume_kms(struct drm_device *dev, bool resume, bool fbcon) return 0; } +static bool amdgpu_check_soft_reset(struct amdgpu_device *adev) +{ + int i; + bool asic_hang = false; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_block_status[i].valid) + continue; + if (adev->ip_blocks[i].funcs->check_soft_reset) + adev->ip_blocks[i].funcs->check_soft_reset(adev); + if (adev->ip_block_status[i].hang) { + DRM_INFO("IP block:%d is hang!\n", i); + asic_hang = true; + } + } + return asic_hang; +} + +int amdgpu_pre_soft_reset(struct amdgpu_device *adev) +{ + int i, r = 0; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_block_status[i].valid) + continue; + if (adev->ip_block_status[i].hang && + adev->ip_blocks[i].funcs->pre_soft_reset) { + r = adev->ip_blocks[i].funcs->pre_soft_reset(adev); + if (r) + return r; + } + } + + return 0; +} + +static bool amdgpu_need_full_reset(struct amdgpu_device *adev) +{ + if (adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang || + adev->ip_block_status[AMD_IP_BLOCK_TYPE_SMC].hang || + adev->ip_block_status[AMD_IP_BLOCK_TYPE_ACP].hang || + adev->ip_block_status[AMD_IP_BLOCK_TYPE_DCE].hang) { + DRM_INFO("Some block need full reset!\n"); + return true; + } + return false; +} + +static int amdgpu_soft_reset(struct amdgpu_device *adev) +{ + int i, r = 0; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_block_status[i].valid) + continue; + if (adev->ip_block_status[i].hang && + adev->ip_blocks[i].funcs->soft_reset) { + r = adev->ip_blocks[i].funcs->soft_reset(adev); + if (r) + return r; + } + } + + return 0; +} + +static int amdgpu_post_soft_reset(struct amdgpu_device *adev) +{ + int i, r = 0; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_block_status[i].valid) + continue; + if (adev->ip_block_status[i].hang && + adev->ip_blocks[i].funcs->post_soft_reset) + r = adev->ip_blocks[i].funcs->post_soft_reset(adev); + if (r) + return r; + } + + return 0; +} + +bool amdgpu_need_backup(struct amdgpu_device *adev) +{ + if (adev->flags & AMD_IS_APU) + return false; + + return amdgpu_lockup_timeout > 0 ? true : false; +} + +static int amdgpu_recover_vram_from_shadow(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_bo *bo, + struct fence **fence) +{ + uint32_t domain; + int r; + + if (!bo->shadow) + return 0; + + r = amdgpu_bo_reserve(bo, false); + if (r) + return r; + domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type); + /* if bo has been evicted, then no need to recover */ + if (domain == AMDGPU_GEM_DOMAIN_VRAM) { + r = amdgpu_bo_restore_from_shadow(adev, ring, bo, + NULL, fence, true); + if (r) { + DRM_ERROR("recover page table failed!\n"); + goto err; + } + } +err: + amdgpu_bo_unreserve(bo); + return r; +} + /** * amdgpu_gpu_reset - reset the asic * @@ -1861,69 +2099,123 @@ int amdgpu_resume_kms(struct drm_device *dev, bool resume, bool fbcon) */ int amdgpu_gpu_reset(struct amdgpu_device *adev) { - unsigned ring_sizes[AMDGPU_MAX_RINGS]; - uint32_t *ring_data[AMDGPU_MAX_RINGS]; - - bool saved = false; - int i, r; int resched; + bool need_full_reset; + + if (!amdgpu_check_soft_reset(adev)) { + DRM_INFO("No hardware hang detected. Did some blocks stall?\n"); + return 0; + } atomic_inc(&adev->gpu_reset_counter); /* block TTM */ resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev); - r = amdgpu_suspend(adev); - + /* block scheduler */ for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = adev->rings[i]; + if (!ring) continue; + kthread_park(ring->sched.thread); + amd_sched_hw_job_reset(&ring->sched); + } + /* after all hw jobs are reset, hw fence is meaningless, so force_completion */ + amdgpu_fence_driver_force_completion(adev); + + need_full_reset = amdgpu_need_full_reset(adev); - ring_sizes[i] = amdgpu_ring_backup(ring, &ring_data[i]); - if (ring_sizes[i]) { - saved = true; - dev_info(adev->dev, "Saved %d dwords of commands " - "on ring %d.\n", ring_sizes[i], i); + if (!need_full_reset) { + amdgpu_pre_soft_reset(adev); + r = amdgpu_soft_reset(adev); + amdgpu_post_soft_reset(adev); + if (r || amdgpu_check_soft_reset(adev)) { + DRM_INFO("soft reset failed, will fallback to full reset!\n"); + need_full_reset = true; } } + if (need_full_reset) { + /* save scratch */ + amdgpu_atombios_scratch_regs_save(adev); + r = amdgpu_suspend(adev); + retry: - r = amdgpu_asic_reset(adev); - /* post card */ - amdgpu_atom_asic_init(adev->mode_info.atom_context); + /* 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 (!r) { - dev_info(adev->dev, "GPU reset succeeded, trying to resume\n"); - r = amdgpu_resume(adev); - } + r = amdgpu_asic_reset(adev); + /* post card */ + amdgpu_atom_asic_init(adev->mode_info.atom_context); + if (!r) { + dev_info(adev->dev, "GPU reset succeeded, trying to resume\n"); + r = amdgpu_resume(adev); + } + /* restore scratch */ + amdgpu_atombios_scratch_regs_restore(adev); + } if (!r) { + amdgpu_irq_gpu_reset_resume_helper(adev); + r = amdgpu_ib_ring_tests(adev); + if (r) { + dev_err(adev->dev, "ib ring test failed (%d).\n", r); + r = amdgpu_suspend(adev); + need_full_reset = true; + goto retry; + } + /** + * recovery vm page tables, since we cannot depend on VRAM is + * consistent after gpu full reset. + */ + if (need_full_reset && amdgpu_need_backup(adev)) { + struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; + struct amdgpu_bo *bo, *tmp; + struct fence *fence = NULL, *next = NULL; + + DRM_INFO("recover vram bo from shadow\n"); + mutex_lock(&adev->shadow_list_lock); + list_for_each_entry_safe(bo, tmp, &adev->shadow_list, shadow_list) { + amdgpu_recover_vram_from_shadow(adev, ring, bo, &next); + if (fence) { + r = fence_wait(fence, false); + if (r) { + WARN(r, "recovery from shadow isn't comleted\n"); + break; + } + } + + fence_put(fence); + fence = next; + } + mutex_unlock(&adev->shadow_list_lock); + if (fence) { + r = fence_wait(fence, false); + if (r) + WARN(r, "recovery from shadow isn't comleted\n"); + } + fence_put(fence); + } for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = adev->rings[i]; if (!ring) continue; - amdgpu_ring_restore(ring, ring_sizes[i], ring_data[i]); - ring_sizes[i] = 0; - ring_data[i] = NULL; - } - - r = amdgpu_ib_ring_tests(adev); - if (r) { - dev_err(adev->dev, "ib ring test failed (%d).\n", r); - if (saved) { - saved = false; - r = amdgpu_suspend(adev); - goto retry; - } + amd_sched_job_recovery(&ring->sched); + kthread_unpark(ring->sched.thread); } } else { - amdgpu_fence_driver_force_completion(adev); + dev_err(adev->dev, "asic resume failed (%d).\n", r); for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { - if (adev->rings[i]) - kfree(ring_data[i]); + if (adev->rings[i]) { + kthread_unpark(adev->rings[i]->sched.thread); + } } } @@ -1938,9 +2230,6 @@ retry: return r; } -#define AMDGPU_DEFAULT_PCIE_GEN_MASK 0x30007 /* gen: chipset 1/2, asic 1/2/3 */ -#define AMDGPU_DEFAULT_PCIE_MLW_MASK 0x2f0000 /* 1/2/4/8/16 lanes */ - void amdgpu_get_pcie_info(struct amdgpu_device *adev) { u32 mask; @@ -2094,20 +2383,50 @@ static ssize_t amdgpu_debugfs_regs_read(struct file *f, char __user *buf, struct amdgpu_device *adev = f->f_inode->i_private; ssize_t result = 0; int r; + bool pm_pg_lock, use_bank; + unsigned instance_bank, sh_bank, se_bank; if (size & 0x3 || *pos & 0x3) return -EINVAL; + /* are we reading registers for which a PG lock is necessary? */ + 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; + use_bank = 1; + } else { + use_bank = 0; + } + + *pos &= 0x3FFFF; + + if (use_bank) { + if (sh_bank >= adev->gfx.config.max_sh_per_se || + se_bank >= adev->gfx.config.max_shader_engines) + return -EINVAL; + mutex_lock(&adev->grbm_idx_mutex); + amdgpu_gfx_select_se_sh(adev, se_bank, + sh_bank, instance_bank); + } + + if (pm_pg_lock) + mutex_lock(&adev->pm.mutex); + while (size) { uint32_t value; if (*pos > adev->rmmio_size) - return result; + goto end; value = RREG32(*pos >> 2); r = put_user(value, (uint32_t *)buf); - if (r) - return r; + if (r) { + result = r; + goto end; + } result += 4; buf += 4; @@ -2115,6 +2434,15 @@ static ssize_t amdgpu_debugfs_regs_read(struct file *f, char __user *buf, size -= 4; } +end: + if (use_bank) { + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + mutex_unlock(&adev->grbm_idx_mutex); + } + + if (pm_pg_lock) + mutex_unlock(&adev->pm.mutex); + return result; } @@ -2314,6 +2642,77 @@ static ssize_t amdgpu_debugfs_regs_smc_write(struct file *f, const char __user * return result; } +static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = f->f_inode->i_private; + ssize_t result = 0; + int r; + uint32_t *config, no_regs = 0; + + if (size & 0x3 || *pos & 0x3) + return -EINVAL; + + config = kmalloc(256 * sizeof(*config), GFP_KERNEL); + if (!config) + return -ENOMEM; + + /* version, increment each time something is added */ + config[no_regs++] = 2; + config[no_regs++] = adev->gfx.config.max_shader_engines; + config[no_regs++] = adev->gfx.config.max_tile_pipes; + config[no_regs++] = adev->gfx.config.max_cu_per_sh; + config[no_regs++] = adev->gfx.config.max_sh_per_se; + config[no_regs++] = adev->gfx.config.max_backends_per_se; + config[no_regs++] = adev->gfx.config.max_texture_channel_caches; + config[no_regs++] = adev->gfx.config.max_gprs; + config[no_regs++] = adev->gfx.config.max_gs_threads; + config[no_regs++] = adev->gfx.config.max_hw_contexts; + config[no_regs++] = adev->gfx.config.sc_prim_fifo_size_frontend; + config[no_regs++] = adev->gfx.config.sc_prim_fifo_size_backend; + config[no_regs++] = adev->gfx.config.sc_hiz_tile_fifo_size; + config[no_regs++] = adev->gfx.config.sc_earlyz_tile_fifo_size; + config[no_regs++] = adev->gfx.config.num_tile_pipes; + config[no_regs++] = adev->gfx.config.backend_enable_mask; + config[no_regs++] = adev->gfx.config.mem_max_burst_length_bytes; + config[no_regs++] = adev->gfx.config.mem_row_size_in_kb; + config[no_regs++] = adev->gfx.config.shader_engine_tile_size; + config[no_regs++] = adev->gfx.config.num_gpus; + config[no_regs++] = adev->gfx.config.multi_gpu_tile_size; + config[no_regs++] = adev->gfx.config.mc_arb_ramcfg; + config[no_regs++] = adev->gfx.config.gb_addr_config; + config[no_regs++] = adev->gfx.config.num_rbs; + + /* rev==1 */ + config[no_regs++] = adev->rev_id; + config[no_regs++] = adev->pg_flags; + config[no_regs++] = adev->cg_flags; + + /* rev==2 */ + config[no_regs++] = adev->family; + config[no_regs++] = adev->external_rev_id; + + while (size && (*pos < no_regs * 4)) { + uint32_t value; + + value = config[*pos >> 2]; + r = put_user(value, (uint32_t *)buf); + if (r) { + kfree(config); + return r; + } + + result += 4; + buf += 4; + *pos += 4; + size -= 4; + } + + kfree(config); + return result; +} + + static const struct file_operations amdgpu_debugfs_regs_fops = { .owner = THIS_MODULE, .read = amdgpu_debugfs_regs_read, @@ -2339,11 +2738,18 @@ static const struct file_operations amdgpu_debugfs_regs_smc_fops = { .llseek = default_llseek }; +static const struct file_operations amdgpu_debugfs_gca_config_fops = { + .owner = THIS_MODULE, + .read = amdgpu_debugfs_gca_config_read, + .llseek = default_llseek +}; + static const struct file_operations *debugfs_regs[] = { &amdgpu_debugfs_regs_fops, &amdgpu_debugfs_regs_didt_fops, &amdgpu_debugfs_regs_pcie_fops, &amdgpu_debugfs_regs_smc_fops, + &amdgpu_debugfs_gca_config_fops, }; static const char *debugfs_regs_names[] = { @@ -2351,6 +2757,7 @@ static const char *debugfs_regs_names[] = { "amdgpu_regs_didt", "amdgpu_regs_pcie", "amdgpu_regs_smc", + "amdgpu_gca_config", }; static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index b0832da2ef7e..9af8d3c7ae8b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -41,7 +41,7 @@ static void amdgpu_flip_callback(struct fence *f, struct fence_cb *cb) container_of(cb, struct amdgpu_flip_work, cb); fence_put(f); - schedule_work(&work->flip_work); + schedule_work(&work->flip_work.work); } static bool amdgpu_flip_handle_fence(struct amdgpu_flip_work *work, @@ -63,16 +63,17 @@ static bool amdgpu_flip_handle_fence(struct amdgpu_flip_work *work, static void amdgpu_flip_work_func(struct work_struct *__work) { + struct delayed_work *delayed_work = + container_of(__work, struct delayed_work, work); struct amdgpu_flip_work *work = - container_of(__work, struct amdgpu_flip_work, flip_work); + container_of(delayed_work, struct amdgpu_flip_work, flip_work); struct amdgpu_device *adev = work->adev; struct amdgpu_crtc *amdgpuCrtc = adev->mode_info.crtcs[work->crtc_id]; struct drm_crtc *crtc = &amdgpuCrtc->base; unsigned long flags; - unsigned i, repcnt = 4; - int vpos, hpos, stat, min_udelay = 0; - struct drm_vblank_crtc *vblank = &crtc->dev->vblank[work->crtc_id]; + unsigned i; + int vpos, hpos; if (amdgpu_flip_handle_fence(work, &work->excl)) return; @@ -81,55 +82,23 @@ static void amdgpu_flip_work_func(struct work_struct *__work) if (amdgpu_flip_handle_fence(work, &work->shared[i])) return; - /* We borrow the event spin lock for protecting flip_status */ - spin_lock_irqsave(&crtc->dev->event_lock, flags); - - /* If this happens to execute within the "virtually extended" vblank - * interval before the start of the real vblank interval then it needs - * to delay programming the mmio flip until the real vblank is entered. - * This prevents completing a flip too early due to the way we fudge - * our vblank counter and vblank timestamps in order to work around the - * problem that the hw fires vblank interrupts before actual start of - * vblank (when line buffer refilling is done for a frame). It - * complements the fudging logic in amdgpu_get_crtc_scanoutpos() for - * timestamping and amdgpu_get_vblank_counter_kms() for vblank counts. - * - * In practice this won't execute very often unless on very fast - * machines because the time window for this to happen is very small. + /* Wait until we're out of the vertical blank period before the one + * targeted by the flip */ - while (amdgpuCrtc->enabled && --repcnt) { - /* GET_DISTANCE_TO_VBLANKSTART returns distance to real vblank - * start in hpos, and to the "fudged earlier" vblank start in - * vpos. - */ - stat = amdgpu_get_crtc_scanoutpos(adev->ddev, work->crtc_id, - GET_DISTANCE_TO_VBLANKSTART, - &vpos, &hpos, NULL, NULL, - &crtc->hwmode); - - if ((stat & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_ACCURATE)) != - (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_ACCURATE) || - !(vpos >= 0 && hpos <= 0)) - break; - - /* Sleep at least until estimated real start of hw vblank */ - min_udelay = (-hpos + 1) * max(vblank->linedur_ns / 1000, 5); - if (min_udelay > vblank->framedur_ns / 2000) { - /* Don't wait ridiculously long - something is wrong */ - repcnt = 0; - break; - } - spin_unlock_irqrestore(&crtc->dev->event_lock, flags); - usleep_range(min_udelay, 2 * min_udelay); - spin_lock_irqsave(&crtc->dev->event_lock, flags); - }; + if (amdgpuCrtc->enabled && + (amdgpu_get_crtc_scanoutpos(adev->ddev, work->crtc_id, 0, + &vpos, &hpos, NULL, NULL, + &crtc->hwmode) + & (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK)) == + (DRM_SCANOUTPOS_VALID | DRM_SCANOUTPOS_IN_VBLANK) && + (int)(work->target_vblank - + amdgpu_get_vblank_counter_kms(adev->ddev, amdgpuCrtc->crtc_id)) > 0) { + schedule_delayed_work(&work->flip_work, usecs_to_jiffies(1000)); + return; + } - if (!repcnt) - DRM_DEBUG_DRIVER("Delay problem on crtc %d: min_udelay %d, " - "framedur %d, linedur %d, stat %d, vpos %d, " - "hpos %d\n", work->crtc_id, min_udelay, - vblank->framedur_ns / 1000, - vblank->linedur_ns / 1000, stat, vpos, hpos); + /* We borrow the event spin lock for protecting flip_status */ + spin_lock_irqsave(&crtc->dev->event_lock, flags); /* Do the flip (mmio) */ adev->mode_info.funcs->page_flip(adev, work->crtc_id, work->base, work->async); @@ -169,10 +138,10 @@ static void amdgpu_unpin_work_func(struct work_struct *__work) kfree(work); } -int amdgpu_crtc_page_flip(struct drm_crtc *crtc, - struct drm_framebuffer *fb, - struct drm_pending_vblank_event *event, - uint32_t page_flip_flags) +int amdgpu_crtc_page_flip_target(struct drm_crtc *crtc, + struct drm_framebuffer *fb, + struct drm_pending_vblank_event *event, + uint32_t page_flip_flags, uint32_t target) { struct drm_device *dev = crtc->dev; struct amdgpu_device *adev = dev->dev_private; @@ -191,7 +160,7 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc, if (work == NULL) return -ENOMEM; - INIT_WORK(&work->flip_work, amdgpu_flip_work_func); + INIT_DELAYED_WORK(&work->flip_work, amdgpu_flip_work_func); INIT_WORK(&work->unpin_work, amdgpu_unpin_work_func); work->event = event; @@ -220,31 +189,25 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc, r = amdgpu_bo_pin_restricted(new_rbo, AMDGPU_GEM_DOMAIN_VRAM, 0, 0, &base); if (unlikely(r != 0)) { - amdgpu_bo_unreserve(new_rbo); r = -EINVAL; DRM_ERROR("failed to pin new rbo buffer before flip\n"); - goto cleanup; + goto unreserve; } r = reservation_object_get_fences_rcu(new_rbo->tbo.resv, &work->excl, &work->shared_count, &work->shared); if (unlikely(r != 0)) { - amdgpu_bo_unreserve(new_rbo); DRM_ERROR("failed to get fences for buffer\n"); - goto cleanup; + goto unpin; } amdgpu_bo_get_tiling_flags(new_rbo, &tiling_flags); amdgpu_bo_unreserve(new_rbo); work->base = base; - - r = drm_vblank_get(crtc->dev, amdgpu_crtc->crtc_id); - if (r) { - DRM_ERROR("failed to get vblank before flip\n"); - goto pflip_cleanup; - } + work->target_vblank = target - drm_crtc_vblank_count(crtc) + + amdgpu_get_vblank_counter_kms(dev, work->crtc_id); /* we borrow the event spin lock for protecting flip_wrok */ spin_lock_irqsave(&crtc->dev->event_lock, flags); @@ -252,7 +215,7 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc, DRM_DEBUG_DRIVER("flip queue: crtc already busy\n"); spin_unlock_irqrestore(&crtc->dev->event_lock, flags); r = -EBUSY; - goto vblank_cleanup; + goto pflip_cleanup; } amdgpu_crtc->pflip_status = AMDGPU_FLIP_PENDING; @@ -264,20 +227,19 @@ int amdgpu_crtc_page_flip(struct drm_crtc *crtc, /* update crtc fb */ crtc->primary->fb = fb; spin_unlock_irqrestore(&crtc->dev->event_lock, flags); - amdgpu_flip_work_func(&work->flip_work); + amdgpu_flip_work_func(&work->flip_work.work); return 0; -vblank_cleanup: - drm_vblank_put(crtc->dev, amdgpu_crtc->crtc_id); - pflip_cleanup: if (unlikely(amdgpu_bo_reserve(new_rbo, false) != 0)) { DRM_ERROR("failed to reserve new rbo in error path\n"); goto cleanup; } +unpin: if (unlikely(amdgpu_bo_unpin(new_rbo) != 0)) { DRM_ERROR("failed to unpin new rbo in error path\n"); } +unreserve: amdgpu_bo_unreserve(new_rbo); cleanup: @@ -335,7 +297,7 @@ int amdgpu_crtc_set_config(struct drm_mode_set *set) return ret; } -static const char *encoder_names[38] = { +static const char *encoder_names[41] = { "NONE", "INTERNAL_LVDS", "INTERNAL_TMDS1", @@ -374,6 +336,9 @@ static const char *encoder_names[38] = { "TRAVIS", "INTERNAL_VCE", "INTERNAL_UNIPHY3", + "HDMI_ANX9805", + "INTERNAL_AMCLK", + "VIRTUAL", }; static const char *hpd_names[6] = { @@ -516,9 +481,7 @@ static void amdgpu_user_framebuffer_destroy(struct drm_framebuffer *fb) { struct amdgpu_framebuffer *amdgpu_fb = to_amdgpu_framebuffer(fb); - if (amdgpu_fb->obj) { - drm_gem_object_unreference_unlocked(amdgpu_fb->obj); - } + drm_gem_object_unreference_unlocked(amdgpu_fb->obj); drm_framebuffer_cleanup(fb); kfree(amdgpu_fb); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index f888c015f76c..7c911d0be2b3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -52,9 +52,12 @@ * - 3.1.0 - allow reading more status registers (GRBM, SRBM, SDMA, CP) * - 3.2.0 - GFX8: Uses EOP_TC_WB_ACTION_EN, so UMDs don't have to do the same * at the end of IBs. + * - 3.3.0 - Add VM support for UVD on supported hardware. + * - 3.4.0 - Add AMDGPU_INFO_NUM_EVICTIONS. + * - 3.5.0 - Add support for new UVD_NO_OP register. */ #define KMS_DRIVER_MAJOR 3 -#define KMS_DRIVER_MINOR 2 +#define KMS_DRIVER_MINOR 5 #define KMS_DRIVER_PATCHLEVEL 0 int amdgpu_vram_limit = 0; @@ -82,8 +85,14 @@ int amdgpu_exp_hw_support = 0; int amdgpu_sched_jobs = 32; int amdgpu_sched_hw_submission = 2; int amdgpu_powerplay = -1; +int amdgpu_powercontainment = 1; +int amdgpu_sclk_deep_sleep_en = 1; unsigned amdgpu_pcie_gen_cap = 0; unsigned amdgpu_pcie_lane_cap = 0; +unsigned amdgpu_cg_mask = 0xffffffff; +unsigned amdgpu_pg_mask = 0xffffffff; +char *amdgpu_disable_cu = NULL; +char *amdgpu_virtual_display = NULL; MODULE_PARM_DESC(vramlimit, "Restrict VRAM for testing, in megabytes"); module_param_named(vramlimit, amdgpu_vram_limit, int, 0600); @@ -160,14 +169,32 @@ module_param_named(sched_hw_submission, amdgpu_sched_hw_submission, int, 0444); #ifdef CONFIG_DRM_AMD_POWERPLAY MODULE_PARM_DESC(powerplay, "Powerplay component (1 = enable, 0 = disable, -1 = auto (default))"); module_param_named(powerplay, amdgpu_powerplay, int, 0444); + +MODULE_PARM_DESC(powercontainment, "Power Containment (1 = enable (default), 0 = disable)"); +module_param_named(powercontainment, amdgpu_powercontainment, int, 0444); #endif +MODULE_PARM_DESC(sclkdeepsleep, "SCLK Deep Sleep (1 = enable (default), 0 = disable)"); +module_param_named(sclkdeepsleep, amdgpu_sclk_deep_sleep_en, int, 0444); + MODULE_PARM_DESC(pcie_gen_cap, "PCIE Gen Caps (0: autodetect (default))"); module_param_named(pcie_gen_cap, amdgpu_pcie_gen_cap, uint, 0444); MODULE_PARM_DESC(pcie_lane_cap, "PCIE Lane Caps (0: autodetect (default))"); module_param_named(pcie_lane_cap, amdgpu_pcie_lane_cap, uint, 0444); +MODULE_PARM_DESC(cg_mask, "Clockgating flags mask (0 = disable clock gating)"); +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(disable_cu, "Disable CUs (se.sh.cu,...)"); +module_param_named(disable_cu, amdgpu_disable_cu, charp, 0444); + +MODULE_PARM_DESC(virtual_display, "Enable virtual display feature (the virtual_display will be set like xxxx:xx:xx.x;xxxx:xx:xx.x)"); +module_param_named(virtual_display, amdgpu_virtual_display, charp, 0444); + static const struct pci_device_id pciidlist[] = { #ifdef CONFIG_DRM_AMDGPU_CIK /* Kaveri */ @@ -324,7 +351,7 @@ static int amdgpu_kick_out_firmware_fb(struct pci_dev *pdev) #ifdef CONFIG_X86 primary = pdev->resource[PCI_ROM_RESOURCE].flags & IORESOURCE_ROM_SHADOW; #endif - remove_conflicting_framebuffers(ap, "amdgpudrmfb", primary); + drm_fb_helper_remove_conflicting_framebuffers(ap, "amdgpudrmfb", primary); kfree(ap); return 0; @@ -413,7 +440,10 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev) pci_save_state(pdev); pci_disable_device(pdev); pci_ignore_hotplug(pdev); - pci_set_power_state(pdev, PCI_D3cold); + if (amdgpu_is_atpx_hybrid()) + pci_set_power_state(pdev, PCI_D3cold); + else if (!amdgpu_has_atpx_dgpu_power_cntl()) + pci_set_power_state(pdev, PCI_D3hot); drm_dev->switch_power_state = DRM_SWITCH_POWER_DYNAMIC_OFF; return 0; @@ -430,7 +460,9 @@ static int amdgpu_pmops_runtime_resume(struct device *dev) drm_dev->switch_power_state = DRM_SWITCH_POWER_CHANGING; - pci_set_power_state(pdev, PCI_D0); + if (amdgpu_is_atpx_hybrid() || + !amdgpu_has_atpx_dgpu_power_cntl()) + pci_set_power_state(pdev, PCI_D0); pci_restore_state(pdev); ret = pci_enable_device(pdev); if (ret) @@ -515,7 +547,7 @@ static struct drm_driver kms_driver = { .driver_features = DRIVER_USE_AGP | DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED | DRIVER_GEM | - DRIVER_PRIME | DRIVER_RENDER, + DRIVER_PRIME | DRIVER_RENDER | DRIVER_MODESET, .dev_priv_size = 0, .load = amdgpu_driver_load_kms, .open = amdgpu_driver_open_kms, @@ -590,7 +622,6 @@ static int __init amdgpu_init(void) DRM_INFO("amdgpu kernel modesetting enabled.\n"); driver = &kms_driver; pdriver = &amdgpu_kms_pci_driver; - driver->driver_features |= DRIVER_MODESET; driver->num_ioctls = amdgpu_max_kms_ioctl; amdgpu_register_atpx_handler(); /* let modprobe override vga console setting */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c index 919146780a15..bf033b58056c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c @@ -25,7 +25,6 @@ */ #include <linux/module.h> #include <linux/slab.h> -#include <linux/fb.h> #include <drm/drmP.h> #include <drm/drm_crtc.h> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index d1558768cfb7..0b109aebfec6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -204,16 +204,25 @@ void amdgpu_fence_process(struct amdgpu_ring *ring) if (seq != ring->fence_drv.sync_seq) amdgpu_fence_schedule_fallback(ring); - while (last_seq != seq) { + if (unlikely(seq == last_seq)) + return; + + last_seq &= drv->num_fences_mask; + seq &= drv->num_fences_mask; + + do { struct fence *fence, **ptr; - ptr = &drv->fences[++last_seq & drv->num_fences_mask]; + ++last_seq; + last_seq &= drv->num_fences_mask; + ptr = &drv->fences[last_seq]; /* There is always exactly one thread signaling this fence slot */ fence = rcu_dereference_protected(*ptr, 1); RCU_INIT_POINTER(*ptr, NULL); - BUG_ON(!fence); + if (!fence) + continue; r = fence_signal(fence); if (!r) @@ -222,7 +231,7 @@ void amdgpu_fence_process(struct amdgpu_ring *ring) BUG(); fence_put(fence); - } + } while (last_seq != seq); } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h index 503d54098128..e73728d90388 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h @@ -31,14 +31,6 @@ #define AMDGPU_GWS_SHIFT PAGE_SHIFT #define AMDGPU_OA_SHIFT PAGE_SHIFT -#define AMDGPU_PL_GDS TTM_PL_PRIV0 -#define AMDGPU_PL_GWS TTM_PL_PRIV1 -#define AMDGPU_PL_OA TTM_PL_PRIV2 - -#define AMDGPU_PL_FLAG_GDS TTM_PL_FLAG_PRIV0 -#define AMDGPU_PL_FLAG_GWS TTM_PL_FLAG_PRIV1 -#define AMDGPU_PL_FLAG_OA TTM_PL_FLAG_PRIV2 - struct amdgpu_ring; struct amdgpu_bo; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 8fab6486064f..88fbed2389c0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -503,7 +503,7 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev, if (r) goto error_print; - amdgpu_vm_get_pt_bos(bo_va->vm, &duplicates); + amdgpu_vm_get_pt_bos(adev, bo_va->vm, &duplicates); list_for_each_entry(entry, &list, head) { domain = amdgpu_mem_type_to_domain(entry->bo->mem.mem_type); /* if anything is swapped out don't swap it in here, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index 9f95da4f0536..a074edd95c70 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -70,3 +70,47 @@ void amdgpu_gfx_scratch_free(struct amdgpu_device *adev, uint32_t reg) } } } + +/** + * amdgpu_gfx_parse_disable_cu - Parse the disable_cu module parameter + * + * @mask: array in which the per-shader array disable masks will be stored + * @max_se: number of SEs + * @max_sh: number of SHs + * + * The bitmask of CUs to be disabled in the shader array determined by se and + * sh is stored in mask[se * max_sh + sh]. + */ +void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, unsigned max_sh) +{ + unsigned se, sh, cu; + const char *p; + + memset(mask, 0, sizeof(*mask) * max_se * max_sh); + + if (!amdgpu_disable_cu || !*amdgpu_disable_cu) + return; + + p = amdgpu_disable_cu; + for (;;) { + char *next; + int ret = sscanf(p, "%u.%u.%u", &se, &sh, &cu); + if (ret < 3) { + DRM_ERROR("amdgpu: could not parse disable_cu\n"); + return; + } + + if (se < max_se && sh < max_sh && cu < 16) { + DRM_INFO("amdgpu: disabling CU %u.%u.%u\n", se, sh, cu); + mask[se * max_sh + sh] |= 1u << cu; + } else { + DRM_ERROR("amdgpu: disable_cu %u.%u.%u is out of range\n", + se, sh, cu); + } + + next = strchr(p, ','); + if (!next) + break; + p = next + 1; + } +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h index dc06cbda7be6..51321e154c09 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h @@ -27,4 +27,6 @@ int amdgpu_gfx_scratch_get(struct amdgpu_device *adev, uint32_t *reg); void amdgpu_gfx_scratch_free(struct amdgpu_device *adev, uint32_t reg); +unsigned amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, unsigned max_sh); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c index 31a676376d73..c93a92a840ea 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c @@ -186,10 +186,8 @@ struct amdgpu_i2c_chan *amdgpu_i2c_create(struct drm_device *dev, "AMDGPU i2c hw bus %s", name); i2c->adapter.algo = &amdgpu_atombios_i2c_algo; ret = i2c_add_adapter(&i2c->adapter); - if (ret) { - DRM_ERROR("Failed to register hw i2c %s\n", name); + if (ret) goto out_free; - } } else { /* set the amdgpu bit adapter */ snprintf(i2c->adapter.name, sizeof(i2c->adapter.name), diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 34e35423b78e..f5810f700668 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -33,6 +33,8 @@ #include "amdgpu.h" #include "atom.h" +#define AMDGPU_IB_TEST_TIMEOUT msecs_to_jiffies(1000) + /* * IB * IBs (Indirect Buffers) and areas of GPU accessible memory where @@ -122,7 +124,6 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, bool skip_preamble, need_ctx_switch; unsigned patch_offset = ~0; struct amdgpu_vm *vm; - struct fence *hwf; uint64_t ctx; unsigned i; @@ -141,7 +142,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, } if (!ring->ready) { - dev_err(adev->dev, "couldn't schedule ib\n"); + dev_err(adev->dev, "couldn't schedule ib on ring <%s>\n", ring->name); return -EINVAL; } @@ -160,10 +161,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, patch_offset = amdgpu_ring_init_cond_exec(ring); if (vm) { - r = amdgpu_vm_flush(ring, job->vm_id, job->vm_pd_addr, - job->gds_base, job->gds_size, - job->gws_base, job->gws_size, - job->oa_base, job->oa_size); + r = amdgpu_vm_flush(ring, job); if (r) { amdgpu_ring_undo(ring); return r; @@ -193,7 +191,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, if (ring->funcs->emit_hdp_invalidate) amdgpu_ring_emit_hdp_invalidate(ring); - r = amdgpu_fence_emit(ring, &hwf); + r = amdgpu_fence_emit(ring, f); if (r) { dev_err(adev->dev, "failed to emit fence (%d)\n", r); if (job && job->vm_id) @@ -203,17 +201,11 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, } /* wrap the last IB with fence */ - if (job && job->uf_bo) { - uint64_t addr = amdgpu_bo_gpu_offset(job->uf_bo); - - addr += job->uf_offset; - amdgpu_ring_emit_fence(ring, addr, job->uf_sequence, + if (job && job->uf_addr) { + amdgpu_ring_emit_fence(ring, job->uf_addr, job->uf_sequence, AMDGPU_FENCE_FLAG_64BIT); } - if (f) - *f = fence_get(hwf); - if (patch_offset != ~0 && ring->funcs->patch_cond_exec) amdgpu_ring_patch_cond_exec(ring, patch_offset); @@ -296,7 +288,7 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev) if (!ring || !ring->ready) continue; - r = amdgpu_ring_test_ib(ring); + r = amdgpu_ring_test_ib(ring, AMDGPU_IB_TEST_TIMEOUT); if (r) { ring->ready = false; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c index 534fc04e80fd..5ebb3f43feb6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c @@ -40,32 +40,15 @@ static int amdgpu_ih_ring_alloc(struct amdgpu_device *adev) /* Allocate ring buffer */ if (adev->irq.ih.ring_obj == NULL) { - r = amdgpu_bo_create(adev, adev->irq.ih.ring_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, - NULL, NULL, &adev->irq.ih.ring_obj); + r = amdgpu_bo_create_kernel(adev, adev->irq.ih.ring_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, + &adev->irq.ih.ring_obj, + &adev->irq.ih.gpu_addr, + (void **)&adev->irq.ih.ring); if (r) { DRM_ERROR("amdgpu: failed to create ih ring buffer (%d).\n", r); return r; } - r = amdgpu_bo_reserve(adev->irq.ih.ring_obj, false); - if (unlikely(r != 0)) - return r; - r = amdgpu_bo_pin(adev->irq.ih.ring_obj, - AMDGPU_GEM_DOMAIN_GTT, - &adev->irq.ih.gpu_addr); - if (r) { - amdgpu_bo_unreserve(adev->irq.ih.ring_obj); - DRM_ERROR("amdgpu: failed to pin ih ring buffer (%d).\n", r); - return r; - } - r = amdgpu_bo_kmap(adev->irq.ih.ring_obj, - (void **)&adev->irq.ih.ring); - amdgpu_bo_unreserve(adev->irq.ih.ring_obj); - if (r) { - DRM_ERROR("amdgpu: failed to map ih ring buffer (%d).\n", r); - return r; - } } return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c index 835a3fa8d8df..278708f5a744 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c @@ -383,6 +383,18 @@ int amdgpu_irq_update(struct amdgpu_device *adev, return r; } +void amdgpu_irq_gpu_reset_resume_helper(struct amdgpu_device *adev) +{ + int i, j; + for (i = 0; i < AMDGPU_MAX_IRQ_SRC_ID; i++) { + struct amdgpu_irq_src *src = adev->irq.sources[i]; + if (!src) + continue; + for (j = 0; j < src->num_types; j++) + amdgpu_irq_update(adev, src, j); + } +} + /** * amdgpu_irq_get - enable interrupt * diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h index e124b59f39c1..f016464035b8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h @@ -70,6 +70,7 @@ struct amdgpu_irq { /* gen irq stuff */ struct irq_domain *domain; /* GPU irq controller domain */ unsigned virq[AMDGPU_MAX_IRQ_SRC_ID]; + uint32_t srbm_soft_reset; }; void amdgpu_irq_preinstall(struct drm_device *dev); @@ -94,6 +95,7 @@ int amdgpu_irq_put(struct amdgpu_device *adev, struct amdgpu_irq_src *src, unsigned type); bool amdgpu_irq_enabled(struct amdgpu_device *adev, struct amdgpu_irq_src *src, unsigned type); +void amdgpu_irq_gpu_reset_resume_helper(struct amdgpu_device *adev); int amdgpu_irq_add_domain(struct amdgpu_device *adev); void amdgpu_irq_remove_domain(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index f0dafa514fe4..6674d40eb3ab 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -28,21 +28,15 @@ #include "amdgpu.h" #include "amdgpu_trace.h" -static void amdgpu_job_free_handler(struct work_struct *ws) +static void amdgpu_job_timedout(struct amd_sched_job *s_job) { - struct amdgpu_job *job = container_of(ws, struct amdgpu_job, base.work_free_job); - amd_sched_job_put(&job->base); -} + struct amdgpu_job *job = container_of(s_job, struct amdgpu_job, base); -void amdgpu_job_timeout_func(struct work_struct *work) -{ - struct amdgpu_job *job = container_of(work, struct amdgpu_job, base.work_tdr.work); DRM_ERROR("ring %s timeout, last signaled seq=%u, last emitted seq=%u\n", - job->base.sched->name, - (uint32_t)atomic_read(&job->ring->fence_drv.last_seq), - job->ring->fence_drv.sync_seq); - - amd_sched_job_put(&job->base); + job->base.sched->name, + atomic_read(&job->ring->fence_drv.last_seq), + job->ring->fence_drv.sync_seq); + amdgpu_gpu_reset(job->adev); } int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, @@ -63,7 +57,6 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, (*job)->vm = vm; (*job)->ibs = (void *)&(*job)[1]; (*job)->num_ibs = num_ibs; - INIT_WORK(&(*job)->base.work_free_job, amdgpu_job_free_handler); amdgpu_sync_create(&(*job)->sync); @@ -86,27 +79,33 @@ int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, return r; } -void amdgpu_job_free(struct amdgpu_job *job) +void amdgpu_job_free_resources(struct amdgpu_job *job) { - unsigned i; struct fence *f; + unsigned i; + /* use sched fence if available */ - f = (job->base.s_fence)? &job->base.s_fence->base : job->fence; + f = job->base.s_fence ? &job->base.s_fence->finished : job->fence; for (i = 0; i < job->num_ibs; ++i) - amdgpu_sa_bo_free(job->adev, &job->ibs[i].sa_bo, f); - fence_put(job->fence); + amdgpu_ib_free(job->adev, &job->ibs[i], f); +} - amdgpu_bo_unref(&job->uf_bo); - amdgpu_sync_free(&job->sync); +void amdgpu_job_free_cb(struct amd_sched_job *s_job) +{ + struct amdgpu_job *job = container_of(s_job, struct amdgpu_job, base); - if (!job->base.use_sched) - kfree(job); + fence_put(job->fence); + amdgpu_sync_free(&job->sync); + kfree(job); } -void amdgpu_job_free_func(struct kref *refcount) +void amdgpu_job_free(struct amdgpu_job *job) { - struct amdgpu_job *job = container_of(refcount, struct amdgpu_job, base.refcount); + amdgpu_job_free_resources(job); + + fence_put(job->fence); + amdgpu_sync_free(&job->sync); kfree(job); } @@ -114,22 +113,20 @@ int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring, struct amd_sched_entity *entity, void *owner, struct fence **f) { - struct fence *fence; int r; job->ring = ring; if (!f) return -EINVAL; - r = amd_sched_job_init(&job->base, &ring->sched, - entity, amdgpu_job_timeout_func, - amdgpu_job_free_func, owner, &fence); + r = amd_sched_job_init(&job->base, &ring->sched, entity, owner); if (r) return r; job->owner = owner; job->ctx = entity->fence_context; - *f = fence_get(fence); + *f = fence_get(&job->base.s_fence->finished); + amdgpu_job_free_resources(job); amd_sched_entity_push_job(&job->base); return 0; @@ -147,8 +144,8 @@ static struct fence *amdgpu_job_dependency(struct amd_sched_job *sched_job) int r; r = amdgpu_vm_grab_id(vm, ring, &job->sync, - &job->base.s_fence->base, - &job->vm_id, &job->vm_pd_addr); + &job->base.s_fence->finished, + job); if (r) DRM_ERROR("Error getting VM ID (%d)\n", r); @@ -170,29 +167,24 @@ static struct fence *amdgpu_job_run(struct amd_sched_job *sched_job) } job = to_amdgpu_job(sched_job); - r = amdgpu_sync_wait(&job->sync); - if (r) { - DRM_ERROR("failed to sync wait (%d)\n", r); - return NULL; - } + BUG_ON(amdgpu_sync_peek_fence(&job->sync, NULL)); trace_amdgpu_sched_run_job(job); r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs, job->sync.last_vm_update, job, &fence); - if (r) { + if (r) DRM_ERROR("Error scheduling IBs (%d)\n", r); - goto err; - } -err: - job->fence = fence; - amdgpu_job_free(job); + /* if gpu reset, hw fence will be replaced here */ + fence_put(job->fence); + job->fence = fence_get(fence); + amdgpu_job_free_resources(job); return fence; } const struct amd_sched_backend_ops amdgpu_sched_ops = { .dependency = amdgpu_job_dependency, .run_job = amdgpu_job_run, - .begin_job = amd_sched_job_begin, - .finish_job = amd_sched_job_finish, + .timedout_job = amdgpu_job_timedout, + .free_job = amdgpu_job_free_cb }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index d851ea15059f..b78e74048f3d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -60,7 +60,10 @@ int amdgpu_driver_unload_kms(struct drm_device *dev) if (adev->rmmio == NULL) goto done_free; - pm_runtime_get_sync(dev->dev); + if (amdgpu_device_is_px(dev)) { + pm_runtime_get_sync(dev->dev); + pm_runtime_forbid(dev->dev); + } amdgpu_amdkfd_device_fini(adev); @@ -135,13 +138,75 @@ int amdgpu_driver_load_kms(struct drm_device *dev, unsigned long flags) } out: - if (r) + if (r) { + /* balance pm_runtime_get_sync in amdgpu_driver_unload_kms */ + if (adev->rmmio && amdgpu_device_is_px(dev)) + pm_runtime_put_noidle(dev->dev); amdgpu_driver_unload_kms(dev); - + } return r; } +static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info, + struct drm_amdgpu_query_fw *query_fw, + struct amdgpu_device *adev) +{ + switch (query_fw->fw_type) { + case AMDGPU_INFO_FW_VCE: + fw_info->ver = adev->vce.fw_version; + fw_info->feature = adev->vce.fb_version; + break; + case AMDGPU_INFO_FW_UVD: + fw_info->ver = adev->uvd.fw_version; + fw_info->feature = 0; + break; + case AMDGPU_INFO_FW_GMC: + fw_info->ver = adev->mc.fw_version; + fw_info->feature = 0; + break; + case AMDGPU_INFO_FW_GFX_ME: + fw_info->ver = adev->gfx.me_fw_version; + fw_info->feature = adev->gfx.me_feature_version; + break; + case AMDGPU_INFO_FW_GFX_PFP: + fw_info->ver = adev->gfx.pfp_fw_version; + fw_info->feature = adev->gfx.pfp_feature_version; + break; + case AMDGPU_INFO_FW_GFX_CE: + fw_info->ver = adev->gfx.ce_fw_version; + fw_info->feature = adev->gfx.ce_feature_version; + break; + case AMDGPU_INFO_FW_GFX_RLC: + fw_info->ver = adev->gfx.rlc_fw_version; + fw_info->feature = adev->gfx.rlc_feature_version; + break; + case AMDGPU_INFO_FW_GFX_MEC: + if (query_fw->index == 0) { + fw_info->ver = adev->gfx.mec_fw_version; + fw_info->feature = adev->gfx.mec_feature_version; + } else if (query_fw->index == 1) { + fw_info->ver = adev->gfx.mec2_fw_version; + fw_info->feature = adev->gfx.mec2_feature_version; + } else + return -EINVAL; + break; + case AMDGPU_INFO_FW_SMC: + fw_info->ver = adev->pm.fw_version; + fw_info->feature = 0; + break; + case AMDGPU_INFO_FW_SDMA: + if (query_fw->index >= adev->sdma.num_instances) + return -EINVAL; + fw_info->ver = adev->sdma.instance[query_fw->index].fw_version; + fw_info->feature = adev->sdma.instance[query_fw->index].feature_version; + break; + default: + return -EINVAL; + } + return 0; +} + /* * Userspace get information ioctl */ @@ -227,14 +292,14 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file type = AMD_IP_BLOCK_TYPE_UVD; ring_mask = adev->uvd.ring.ready ? 1 : 0; ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; - ib_size_alignment = 8; + ib_size_alignment = 16; break; case AMDGPU_HW_IP_VCE: type = AMD_IP_BLOCK_TYPE_VCE; for (i = 0; i < AMDGPU_MAX_VCE_RINGS; i++) ring_mask |= ((adev->vce.ring[i].ready ? 1 : 0) << i); ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; - ib_size_alignment = 8; + ib_size_alignment = 1; break; default: return -EINVAL; @@ -288,73 +353,29 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file return copy_to_user(out, &count, min(size, 4u)) ? -EFAULT : 0; } case AMDGPU_INFO_TIMESTAMP: - ui64 = amdgpu_asic_get_gpu_clock_counter(adev); + ui64 = amdgpu_gfx_get_gpu_clock_counter(adev); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; case AMDGPU_INFO_FW_VERSION: { struct drm_amdgpu_info_firmware fw_info; + int ret; /* We only support one instance of each IP block right now. */ if (info->query_fw.ip_instance != 0) return -EINVAL; - switch (info->query_fw.fw_type) { - case AMDGPU_INFO_FW_VCE: - fw_info.ver = adev->vce.fw_version; - fw_info.feature = adev->vce.fb_version; - break; - case AMDGPU_INFO_FW_UVD: - fw_info.ver = adev->uvd.fw_version; - fw_info.feature = 0; - break; - case AMDGPU_INFO_FW_GMC: - fw_info.ver = adev->mc.fw_version; - fw_info.feature = 0; - break; - case AMDGPU_INFO_FW_GFX_ME: - fw_info.ver = adev->gfx.me_fw_version; - fw_info.feature = adev->gfx.me_feature_version; - break; - case AMDGPU_INFO_FW_GFX_PFP: - fw_info.ver = adev->gfx.pfp_fw_version; - fw_info.feature = adev->gfx.pfp_feature_version; - break; - case AMDGPU_INFO_FW_GFX_CE: - fw_info.ver = adev->gfx.ce_fw_version; - fw_info.feature = adev->gfx.ce_feature_version; - break; - case AMDGPU_INFO_FW_GFX_RLC: - fw_info.ver = adev->gfx.rlc_fw_version; - fw_info.feature = adev->gfx.rlc_feature_version; - break; - case AMDGPU_INFO_FW_GFX_MEC: - if (info->query_fw.index == 0) { - fw_info.ver = adev->gfx.mec_fw_version; - fw_info.feature = adev->gfx.mec_feature_version; - } else if (info->query_fw.index == 1) { - fw_info.ver = adev->gfx.mec2_fw_version; - fw_info.feature = adev->gfx.mec2_feature_version; - } else - return -EINVAL; - break; - case AMDGPU_INFO_FW_SMC: - fw_info.ver = adev->pm.fw_version; - fw_info.feature = 0; - break; - case AMDGPU_INFO_FW_SDMA: - if (info->query_fw.index >= adev->sdma.num_instances) - return -EINVAL; - fw_info.ver = adev->sdma.instance[info->query_fw.index].fw_version; - fw_info.feature = adev->sdma.instance[info->query_fw.index].feature_version; - break; - default: - return -EINVAL; - } + ret = amdgpu_firmware_info(&fw_info, &info->query_fw, adev); + if (ret) + return ret; + return copy_to_user(out, &fw_info, min((size_t)size, sizeof(fw_info))) ? -EFAULT : 0; } case AMDGPU_INFO_NUM_BYTES_MOVED: ui64 = atomic64_read(&adev->num_bytes_moved); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; + case AMDGPU_INFO_NUM_EVICTIONS: + ui64 = atomic64_read(&adev->num_evictions); + return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; case AMDGPU_INFO_VRAM_USAGE: ui64 = atomic64_read(&adev->vram_usage); return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0; @@ -566,6 +587,9 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, amdgpu_ctx_mgr_fini(&fpriv->ctx_mgr); + amdgpu_uvd_free_handles(adev, file_priv); + amdgpu_vce_free_handles(adev, file_priv); + amdgpu_vm_fini(adev, &fpriv->vm); idr_for_each_entry(&fpriv->bo_list_handles, list, handle) @@ -590,10 +614,6 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, void amdgpu_driver_preclose_kms(struct drm_device *dev, struct drm_file *file_priv) { - struct amdgpu_device *adev = dev->dev_private; - - amdgpu_uvd_free_handles(adev, file_priv); - amdgpu_vce_free_handles(adev, file_priv); } /* @@ -756,3 +776,130 @@ const struct drm_ioctl_desc amdgpu_ioctls_kms[] = { 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); + +/* + * Debugfs info + */ +#if defined(CONFIG_DEBUG_FS) + +static int amdgpu_debugfs_firmware_info(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; + struct drm_amdgpu_info_firmware fw_info; + struct drm_amdgpu_query_fw query_fw; + int ret, i; + + /* VCE */ + query_fw.fw_type = AMDGPU_INFO_FW_VCE; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "VCE feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* UVD */ + query_fw.fw_type = AMDGPU_INFO_FW_UVD; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "UVD feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* GMC */ + query_fw.fw_type = AMDGPU_INFO_FW_GMC; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "MC feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* ME */ + query_fw.fw_type = AMDGPU_INFO_FW_GFX_ME; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "ME feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* PFP */ + query_fw.fw_type = AMDGPU_INFO_FW_GFX_PFP; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "PFP feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* CE */ + query_fw.fw_type = AMDGPU_INFO_FW_GFX_CE; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "CE feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* RLC */ + query_fw.fw_type = AMDGPU_INFO_FW_GFX_RLC; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "RLC feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* MEC */ + query_fw.fw_type = AMDGPU_INFO_FW_GFX_MEC; + query_fw.index = 0; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "MEC feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* MEC2 */ + if (adev->asic_type == CHIP_KAVERI || + (adev->asic_type > CHIP_TOPAZ && adev->asic_type != CHIP_STONEY)) { + query_fw.index = 1; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "MEC2 feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + } + + /* SMC */ + query_fw.fw_type = AMDGPU_INFO_FW_SMC; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "SMC feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* SDMA */ + query_fw.fw_type = AMDGPU_INFO_FW_SDMA; + for (i = 0; i < adev->sdma.num_instances; i++) { + query_fw.index = i; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "SDMA%d feature version: %u, firmware version: 0x%08x\n", + i, fw_info.feature, fw_info.ver); + } + + return 0; +} + +static const struct drm_info_list amdgpu_firmware_info_list[] = { + {"amdgpu_firmware_info", amdgpu_debugfs_firmware_info, 0, NULL}, +}; +#endif + +int amdgpu_debugfs_firmware_init(struct amdgpu_device *adev) +{ +#if defined(CONFIG_DEBUG_FS) + return amdgpu_debugfs_add_files(adev, amdgpu_firmware_info_list, + ARRAY_SIZE(amdgpu_firmware_info_list)); +#else + return 0; +#endif +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index 6b1d7d306564..7b0eff7d060b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -39,6 +39,8 @@ #include <drm/drm_plane_helper.h> #include <linux/i2c.h> #include <linux/i2c-algo-bit.h> +#include <linux/hrtimer.h> +#include "amdgpu_irq.h" struct amdgpu_bo; struct amdgpu_device; @@ -339,6 +341,8 @@ struct amdgpu_mode_info { int num_dig; /* number of dig blocks */ int disp_priority; const struct amdgpu_display_funcs *funcs; + struct hrtimer vblank_timer; + enum amdgpu_interrupt_state vsync_timer_enabled; }; #define AMDGPU_MAX_BL_LEVEL 0xFF @@ -587,10 +591,10 @@ int amdgpu_align_pitch(struct amdgpu_device *adev, int width, int bpp, bool tile void amdgpu_print_display_setup(struct drm_device *dev); int amdgpu_modeset_create_props(struct amdgpu_device *adev); int amdgpu_crtc_set_config(struct drm_mode_set *set); -int amdgpu_crtc_page_flip(struct drm_crtc *crtc, - struct drm_framebuffer *fb, - struct drm_pending_vblank_event *event, - uint32_t page_flip_flags); +int amdgpu_crtc_page_flip_target(struct drm_crtc *crtc, + struct drm_framebuffer *fb, + struct drm_pending_vblank_event *event, + uint32_t page_flip_flags, uint32_t target); extern const struct drm_mode_config_funcs amdgpu_mode_funcs; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 7ecea83ce453..b17734e0ecc8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -44,14 +44,13 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev); static u64 amdgpu_get_vis_part_size(struct amdgpu_device *adev, struct ttm_mem_reg *mem) { - u64 ret = 0; - if (mem->start << PAGE_SHIFT < adev->mc.visible_vram_size) { - ret = (u64)((mem->start << PAGE_SHIFT) + mem->size) > - adev->mc.visible_vram_size ? - adev->mc.visible_vram_size - (mem->start << PAGE_SHIFT) : - mem->size; - } - return ret; + 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, @@ -99,6 +98,11 @@ static void amdgpu_ttm_bo_destroy(struct ttm_buffer_object *tbo) drm_gem_object_release(&bo->gem_base); amdgpu_bo_unref(&bo->parent); + if (!list_empty(&bo->shadow_list)) { + mutex_lock(&bo->adev->shadow_list_lock); + list_del_init(&bo->shadow_list); + mutex_unlock(&bo->adev->shadow_list_lock); + } kfree(bo->metadata); kfree(bo); } @@ -112,84 +116,93 @@ bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo) static void amdgpu_ttm_placement_init(struct amdgpu_device *adev, struct ttm_placement *placement, - struct ttm_place *placements, + struct ttm_place *places, u32 domain, u64 flags) { - u32 c = 0, i; - - placement->placement = placements; - placement->busy_placement = placements; + u32 c = 0; if (domain & AMDGPU_GEM_DOMAIN_VRAM) { + unsigned visible_pfn = adev->mc.visible_vram_size >> PAGE_SHIFT; + if (flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS && - adev->mc.visible_vram_size < adev->mc.real_vram_size) { - placements[c].fpfn = - adev->mc.visible_vram_size >> PAGE_SHIFT; - placements[c++].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | - TTM_PL_FLAG_VRAM | TTM_PL_FLAG_TOPDOWN; + !(flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) && + adev->mc.visible_vram_size < adev->mc.real_vram_size) { + places[c].fpfn = visible_pfn; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_WC | + TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_VRAM | + TTM_PL_FLAG_TOPDOWN; + c++; } - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | + + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_VRAM; - if (!(flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)) - placements[c - 1].flags |= TTM_PL_FLAG_TOPDOWN; + if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) + places[c].lpfn = visible_pfn; + else + places[c].flags |= TTM_PL_FLAG_TOPDOWN; + c++; } if (domain & AMDGPU_GEM_DOMAIN_GTT) { - if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_TT | + places[c].fpfn = 0; + 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 | TTM_PL_FLAG_UNCACHED; - } else { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_TT; - } + else + places[c].flags |= TTM_PL_FLAG_CACHED; + c++; } if (domain & AMDGPU_GEM_DOMAIN_CPU) { - if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_SYSTEM | + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_SYSTEM; + if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) + places[c].flags |= TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED; - } else { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_SYSTEM; - } + else + places[c].flags |= TTM_PL_FLAG_CACHED; + c++; } if (domain & AMDGPU_GEM_DOMAIN_GDS) { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_UNCACHED | - AMDGPU_PL_FLAG_GDS; + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_UNCACHED | AMDGPU_PL_FLAG_GDS; + c++; } + if (domain & AMDGPU_GEM_DOMAIN_GWS) { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_UNCACHED | - AMDGPU_PL_FLAG_GWS; + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_UNCACHED | AMDGPU_PL_FLAG_GWS; + c++; } + if (domain & AMDGPU_GEM_DOMAIN_OA) { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_FLAG_UNCACHED | - AMDGPU_PL_FLAG_OA; + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_FLAG_UNCACHED | AMDGPU_PL_FLAG_OA; + c++; } if (!c) { - placements[c].fpfn = 0; - placements[c++].flags = TTM_PL_MASK_CACHING | - TTM_PL_FLAG_SYSTEM; + places[c].fpfn = 0; + places[c].lpfn = 0; + places[c].flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM; + c++; } + placement->num_placement = c; - placement->num_busy_placement = c; + placement->placement = places; - for (i = 0; i < c; i++) { - if ((flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) && - (placements[i].flags & TTM_PL_FLAG_VRAM) && - !placements[i].fpfn) - placements[i].lpfn = - adev->mc.visible_vram_size >> PAGE_SHIFT; - else - placements[i].lpfn = 0; - } + placement->num_busy_placement = c; + placement->busy_placement = places; } void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *rbo, u32 domain) @@ -211,6 +224,69 @@ static void amdgpu_fill_placement_to_bo(struct amdgpu_bo *bo, bo->placement.busy_placement = bo->placements; } +/** + * 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(adev, size, align, true, domain, + AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED, + NULL, NULL, bo_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to allocate kernel bo\n", r); + return r; + } + + r = amdgpu_bo_reserve(*bo_ptr, false); + if (r) { + dev_err(adev->dev, "(%d) failed to reserve kernel bo\n", r); + goto error_free; + } + + r = amdgpu_bo_pin(*bo_ptr, domain, gpu_addr); + if (r) { + dev_err(adev->dev, "(%d) kernel bo pin failed\n", r); + goto error_unreserve; + } + + if (cpu_addr) { + r = amdgpu_bo_kmap(*bo_ptr, cpu_addr); + if (r) { + dev_err(adev->dev, "(%d) kernel bo map failed\n", r); + goto error_unreserve; + } + } + + amdgpu_bo_unreserve(*bo_ptr); + + return 0; + +error_unreserve: + amdgpu_bo_unreserve(*bo_ptr); + +error_free: + amdgpu_bo_unref(bo_ptr); + + return r; +} + int amdgpu_bo_create_restricted(struct amdgpu_device *adev, unsigned long size, int byte_align, bool kernel, u32 domain, u64 flags, @@ -250,6 +326,7 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev, } bo->adev = adev; INIT_LIST_HEAD(&bo->list); + INIT_LIST_HEAD(&bo->shadow_list); INIT_LIST_HEAD(&bo->va); bo->prefered_domains = domain & (AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT | @@ -277,11 +354,79 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev, if (unlikely(r != 0)) { return r; } + + if (flags & AMDGPU_GEM_CREATE_VRAM_CLEARED && + bo->tbo.mem.placement & TTM_PL_FLAG_VRAM) { + struct fence *fence; + + if (adev->mman.buffer_funcs_ring == NULL || + !adev->mman.buffer_funcs_ring->ready) { + r = -EBUSY; + goto fail_free; + } + + r = amdgpu_bo_reserve(bo, false); + if (unlikely(r != 0)) + goto fail_free; + + amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_VRAM); + r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); + if (unlikely(r != 0)) + goto fail_unreserve; + + amdgpu_fill_buffer(bo, 0, bo->tbo.resv, &fence); + amdgpu_bo_fence(bo, fence, false); + amdgpu_bo_unreserve(bo); + fence_put(bo->tbo.moving); + bo->tbo.moving = fence_get(fence); + fence_put(fence); + } *bo_ptr = bo; trace_amdgpu_bo_create(bo); return 0; + +fail_unreserve: + amdgpu_bo_unreserve(bo); +fail_free: + amdgpu_bo_unref(&bo); + return r; +} + +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); + if (!r) { + bo->shadow->parent = amdgpu_bo_ref(bo); + mutex_lock(&adev->shadow_list_lock); + list_add_tail(&bo->shadow_list, &adev->shadow_list); + mutex_unlock(&adev->shadow_list_lock); + } + + return r; } int amdgpu_bo_create(struct amdgpu_device *adev, @@ -293,6 +438,7 @@ int amdgpu_bo_create(struct amdgpu_device *adev, { struct ttm_placement placement = {0}; struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1]; + int r; memset(&placements, 0, (AMDGPU_GEM_DOMAIN_MAX + 1) * sizeof(struct ttm_place)); @@ -300,9 +446,83 @@ int amdgpu_bo_create(struct amdgpu_device *adev, amdgpu_ttm_placement_init(adev, &placement, placements, domain, flags); - return amdgpu_bo_create_restricted(adev, size, byte_align, kernel, - domain, flags, sg, &placement, - resv, bo_ptr); + r = amdgpu_bo_create_restricted(adev, size, byte_align, kernel, + domain, flags, sg, &placement, + resv, bo_ptr); + if (r) + return r; + + if (amdgpu_need_backup(adev) && (flags & AMDGPU_GEM_CREATE_SHADOW)) { + r = amdgpu_bo_create_shadow(adev, size, byte_align, (*bo_ptr)); + if (r) + amdgpu_bo_unref(bo_ptr); + } + + return r; +} + +int amdgpu_bo_backup_to_shadow(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_bo *bo, + struct reservation_object *resv, + struct fence **fence, + bool direct) + +{ + struct amdgpu_bo *shadow = bo->shadow; + uint64_t bo_addr, shadow_addr; + int r; + + if (!shadow) + return -EINVAL; + + bo_addr = amdgpu_bo_gpu_offset(bo); + shadow_addr = amdgpu_bo_gpu_offset(bo->shadow); + + r = reservation_object_reserve_shared(bo->tbo.resv); + if (r) + goto err; + + r = amdgpu_copy_buffer(ring, bo_addr, shadow_addr, + amdgpu_bo_size(bo), resv, fence, + direct); + if (!r) + amdgpu_bo_fence(bo, *fence, true); + +err: + return r; +} + +int amdgpu_bo_restore_from_shadow(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_bo *bo, + struct reservation_object *resv, + struct fence **fence, + bool direct) + +{ + struct amdgpu_bo *shadow = bo->shadow; + uint64_t bo_addr, shadow_addr; + int r; + + if (!shadow) + return -EINVAL; + + bo_addr = amdgpu_bo_gpu_offset(bo); + shadow_addr = amdgpu_bo_gpu_offset(bo->shadow); + + r = reservation_object_reserve_shared(bo->tbo.resv); + if (r) + goto err; + + r = amdgpu_copy_buffer(ring, shadow_addr, bo_addr, + amdgpu_bo_size(bo), resv, fence, + direct); + if (!r) + amdgpu_bo_fence(bo, *fence, true); + +err: + return r; } int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr) @@ -380,16 +600,17 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, return -EINVAL; if (bo->pin_count) { + uint32_t mem_type = bo->tbo.mem.mem_type; + + if (domain != amdgpu_mem_type_to_domain(mem_type)) + return -EINVAL; + bo->pin_count++; if (gpu_addr) *gpu_addr = amdgpu_bo_gpu_offset(bo); if (max_offset != 0) { - u64 domain_start; - if (domain == AMDGPU_GEM_DOMAIN_VRAM) - domain_start = bo->adev->mc.vram_start; - else - domain_start = bo->adev->mc.gtt_start; + u64 domain_start = bo->tbo.bdev->man[mem_type].gpu_offset; WARN_ON_ONCE(max_offset < (amdgpu_bo_gpu_offset(bo) - domain_start)); } @@ -401,7 +622,8 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, /* 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 > bo->adev->mc.visible_vram_size)) { + (!max_offset || max_offset > + bo->adev->mc.visible_vram_size)) { if (WARN_ON_ONCE(min_offset > bo->adev->mc.visible_vram_size)) return -EINVAL; @@ -420,19 +642,23 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, } r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); - if (likely(r == 0)) { - bo->pin_count = 1; - if (gpu_addr != NULL) - *gpu_addr = amdgpu_bo_gpu_offset(bo); - if (domain == AMDGPU_GEM_DOMAIN_VRAM) { - bo->adev->vram_pin_size += amdgpu_bo_size(bo); - if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) - bo->adev->invisible_pin_size += amdgpu_bo_size(bo); - } else - bo->adev->gart_pin_size += amdgpu_bo_size(bo); - } else { + if (unlikely(r)) { dev_err(bo->adev->dev, "%p pin failed\n", bo); + goto error; + } + + bo->pin_count = 1; + if (gpu_addr != NULL) + *gpu_addr = amdgpu_bo_gpu_offset(bo); + if (domain == AMDGPU_GEM_DOMAIN_VRAM) { + bo->adev->vram_pin_size += amdgpu_bo_size(bo); + if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) + bo->adev->invisible_pin_size += amdgpu_bo_size(bo); + } else if (domain == AMDGPU_GEM_DOMAIN_GTT) { + bo->adev->gart_pin_size += amdgpu_bo_size(bo); } + +error: return r; } @@ -457,16 +683,20 @@ int amdgpu_bo_unpin(struct amdgpu_bo *bo) bo->placements[i].flags &= ~TTM_PL_FLAG_NO_EVICT; } r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); - if (likely(r == 0)) { - if (bo->tbo.mem.mem_type == TTM_PL_VRAM) { - bo->adev->vram_pin_size -= amdgpu_bo_size(bo); - if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) - bo->adev->invisible_pin_size -= amdgpu_bo_size(bo); - } else - bo->adev->gart_pin_size -= amdgpu_bo_size(bo); - } else { + if (unlikely(r)) { dev_err(bo->adev->dev, "%p validate failed for unpin\n", bo); + goto error; + } + + if (bo->tbo.mem.mem_type == TTM_PL_VRAM) { + bo->adev->vram_pin_size -= amdgpu_bo_size(bo); + if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) + bo->adev->invisible_pin_size -= amdgpu_bo_size(bo); + } else { + bo->adev->gart_pin_size -= amdgpu_bo_size(bo); } + +error: return r; } @@ -589,6 +819,7 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, struct ttm_mem_reg *new_mem) { struct amdgpu_bo *rbo; + struct ttm_mem_reg *old_mem = &bo->mem; if (!amdgpu_ttm_bo_is_amdgpu_bo(bo)) return; @@ -602,6 +833,8 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, /* move_notify is called before move happens */ amdgpu_update_memory_usage(rbo->adev, &bo->mem, new_mem); + + trace_amdgpu_ttm_bo_move(rbo, new_mem->mem_type, old_mem->mem_type); } int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) @@ -634,7 +867,8 @@ int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) 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 || + abo->placements[i].lpfn > lpfn)) abo->placements[i].lpfn = lpfn; } r = ttm_bo_validate(bo, &abo->placement, false, false); @@ -671,3 +905,21 @@ void amdgpu_bo_fence(struct amdgpu_bo *bo, struct fence *fence, else reservation_object_add_excl_fence(resv, fence); } + +/** + * amdgpu_bo_gpu_offset - return GPU offset of bo + * @bo: amdgpu object for which we query the offset + * + * Returns current GPU offset of the object. + * + * Note: object should either be pinned or reserved when calling this + * function, it might be useful to add check for this for debugging. + */ +u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo) +{ + WARN_ON_ONCE(bo->tbo.mem.mem_type == TTM_PL_SYSTEM); + WARN_ON_ONCE(!ww_mutex_is_locked(&bo->tbo.resv->lock) && + !bo->pin_count); + + return bo->tbo.offset; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index bdb01d932548..b6a27390ef88 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -85,21 +85,6 @@ static inline void amdgpu_bo_unreserve(struct amdgpu_bo *bo) ttm_bo_unreserve(&bo->tbo); } -/** - * amdgpu_bo_gpu_offset - return GPU offset of bo - * @bo: amdgpu object for which we query the offset - * - * Returns current GPU offset of the object. - * - * Note: object should either be pinned or reserved when calling this - * function, it might be useful to add check for this for debugging. - */ -static inline u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo) -{ - WARN_ON_ONCE(bo->tbo.mem.mem_type == TTM_PL_SYSTEM); - return bo->tbo.offset; -} - static inline unsigned long amdgpu_bo_size(struct amdgpu_bo *bo) { return bo->tbo.num_pages << PAGE_SHIFT; @@ -139,6 +124,10 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev, struct ttm_placement *placement, struct reservation_object *resv, struct amdgpu_bo **bo_ptr); +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_kmap(struct amdgpu_bo *bo, void **ptr); void amdgpu_bo_kunmap(struct amdgpu_bo *bo); struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo); @@ -165,6 +154,19 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo); void amdgpu_bo_fence(struct amdgpu_bo *bo, struct fence *fence, bool shared); +u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo); +int amdgpu_bo_backup_to_shadow(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_bo *bo, + struct reservation_object *resv, + struct fence **fence, bool direct); +int amdgpu_bo_restore_from_shadow(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_bo *bo, + struct reservation_object *resv, + struct fence **fence, + bool direct); + /* * sub allocation diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c index 0e13d80d2a95..d4ec3cb187a5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c @@ -305,7 +305,7 @@ static ssize_t amdgpu_get_pp_table(struct device *dev, struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; char *table = NULL; - int size, i; + int size; if (adev->pp_enabled) size = amdgpu_dpm_get_pp_table(adev, &table); @@ -315,10 +315,7 @@ static ssize_t amdgpu_get_pp_table(struct device *dev, if (size >= PAGE_SIZE) size = PAGE_SIZE - 1; - for (i = 0; i < size; i++) { - sprintf(buf + i, "%02x", table[i]); - } - sprintf(buf + i, "\n"); + memcpy(buf, table, size); return size; } @@ -347,6 +344,8 @@ static ssize_t amdgpu_get_pp_dpm_sclk(struct device *dev, 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; } @@ -363,7 +362,9 @@ static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev, uint32_t i, mask = 0; char sub_str[2]; - for (i = 0; i < strlen(buf) - 1; i++) { + for (i = 0; i < strlen(buf); i++) { + if (*(buf + i) == '\n') + continue; sub_str[0] = *(buf + i); sub_str[1] = '\0'; ret = kstrtol(sub_str, 0, &level); @@ -377,6 +378,8 @@ static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev, if (adev->pp_enabled) 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; } @@ -391,6 +394,8 @@ static ssize_t amdgpu_get_pp_dpm_mclk(struct device *dev, 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; } @@ -407,7 +412,9 @@ static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev, uint32_t i, mask = 0; char sub_str[2]; - for (i = 0; i < strlen(buf) - 1; i++) { + for (i = 0; i < strlen(buf); i++) { + if (*(buf + i) == '\n') + continue; sub_str[0] = *(buf + i); sub_str[1] = '\0'; ret = kstrtol(sub_str, 0, &level); @@ -421,6 +428,8 @@ static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev, if (adev->pp_enabled) 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; } @@ -435,6 +444,8 @@ static ssize_t amdgpu_get_pp_dpm_pcie(struct device *dev, 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; } @@ -451,7 +462,9 @@ static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev, uint32_t i, mask = 0; char sub_str[2]; - for (i = 0; i < strlen(buf) - 1; i++) { + for (i = 0; i < strlen(buf); i++) { + if (*(buf + i) == '\n') + continue; sub_str[0] = *(buf + i); sub_str[1] = '\0'; ret = kstrtol(sub_str, 0, &level); @@ -465,6 +478,100 @@ static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev, if (adev->pp_enabled) 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; +} + +static ssize_t amdgpu_get_pp_sclk_od(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + uint32_t value = 0; + + if (adev->pp_enabled) + 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); +} + +static ssize_t amdgpu_set_pp_sclk_od(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + int ret; + long int value; + + ret = kstrtol(buf, 0, &value); + + if (ret) { + count = -EINVAL; + goto fail; + } + + if (adev->pp_enabled) { + 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); + adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps; + amdgpu_pm_compute_clocks(adev); + } + +fail: + return count; +} + +static ssize_t amdgpu_get_pp_mclk_od(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + uint32_t value = 0; + + if (adev->pp_enabled) + 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); +} + +static ssize_t amdgpu_set_pp_mclk_od(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + int ret; + long int value; + + ret = kstrtol(buf, 0, &value); + + if (ret) { + count = -EINVAL; + goto fail; + } + + if (adev->pp_enabled) { + 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); + adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps; + amdgpu_pm_compute_clocks(adev); + } + fail: return count; } @@ -490,6 +597,12 @@ static DEVICE_ATTR(pp_dpm_mclk, S_IRUGO | S_IWUSR, static DEVICE_ATTR(pp_dpm_pcie, S_IRUGO | S_IWUSR, amdgpu_get_pp_dpm_pcie, amdgpu_set_pp_dpm_pcie); +static DEVICE_ATTR(pp_sclk_od, S_IRUGO | S_IWUSR, + amdgpu_get_pp_sclk_od, + amdgpu_set_pp_sclk_od); +static DEVICE_ATTR(pp_mclk_od, S_IRUGO | S_IWUSR, + amdgpu_get_pp_mclk_od, + amdgpu_set_pp_mclk_od); static ssize_t amdgpu_hwmon_show_temp(struct device *dev, struct device_attribute *attr, @@ -990,54 +1103,46 @@ force: void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable) { - if (adev->pp_enabled) + if (adev->pp_enabled || adev->pm.funcs->powergate_uvd) { + /* enable/disable UVD */ + mutex_lock(&adev->pm.mutex); amdgpu_dpm_powergate_uvd(adev, !enable); - else { - if (adev->pm.funcs->powergate_uvd) { + mutex_unlock(&adev->pm.mutex); + } else { + if (enable) { mutex_lock(&adev->pm.mutex); - /* enable/disable UVD */ - amdgpu_dpm_powergate_uvd(adev, !enable); + adev->pm.dpm.uvd_active = true; + adev->pm.dpm.state = POWER_STATE_TYPE_INTERNAL_UVD; mutex_unlock(&adev->pm.mutex); } else { - if (enable) { - mutex_lock(&adev->pm.mutex); - adev->pm.dpm.uvd_active = true; - adev->pm.dpm.state = POWER_STATE_TYPE_INTERNAL_UVD; - mutex_unlock(&adev->pm.mutex); - } else { - mutex_lock(&adev->pm.mutex); - adev->pm.dpm.uvd_active = false; - mutex_unlock(&adev->pm.mutex); - } - amdgpu_pm_compute_clocks(adev); + mutex_lock(&adev->pm.mutex); + adev->pm.dpm.uvd_active = false; + mutex_unlock(&adev->pm.mutex); } - + amdgpu_pm_compute_clocks(adev); } } void amdgpu_dpm_enable_vce(struct amdgpu_device *adev, bool enable) { - if (adev->pp_enabled) + if (adev->pp_enabled || adev->pm.funcs->powergate_vce) { + /* enable/disable VCE */ + mutex_lock(&adev->pm.mutex); amdgpu_dpm_powergate_vce(adev, !enable); - else { - if (adev->pm.funcs->powergate_vce) { + mutex_unlock(&adev->pm.mutex); + } else { + if (enable) { mutex_lock(&adev->pm.mutex); - amdgpu_dpm_powergate_vce(adev, !enable); + adev->pm.dpm.vce_active = true; + /* XXX select vce level based on ring/task */ + adev->pm.dpm.vce_level = AMDGPU_VCE_LEVEL_AC_ALL; mutex_unlock(&adev->pm.mutex); } else { - if (enable) { - mutex_lock(&adev->pm.mutex); - adev->pm.dpm.vce_active = true; - /* XXX select vce level based on ring/task */ - adev->pm.dpm.vce_level = AMDGPU_VCE_LEVEL_AC_ALL; - mutex_unlock(&adev->pm.mutex); - } else { - mutex_lock(&adev->pm.mutex); - adev->pm.dpm.vce_active = false; - mutex_unlock(&adev->pm.mutex); - } - amdgpu_pm_compute_clocks(adev); + mutex_lock(&adev->pm.mutex); + adev->pm.dpm.vce_active = false; + mutex_unlock(&adev->pm.mutex); } + amdgpu_pm_compute_clocks(adev); } } @@ -1108,22 +1213,34 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) DRM_ERROR("failed to create device file pp_table\n"); return ret; } - ret = device_create_file(adev->dev, &dev_attr_pp_dpm_sclk); - if (ret) { - DRM_ERROR("failed to create device file pp_dpm_sclk\n"); - return ret; - } - ret = device_create_file(adev->dev, &dev_attr_pp_dpm_mclk); - if (ret) { - DRM_ERROR("failed to create device file pp_dpm_mclk\n"); - return ret; - } - ret = device_create_file(adev->dev, &dev_attr_pp_dpm_pcie); - if (ret) { - DRM_ERROR("failed to create device file pp_dpm_pcie\n"); - return ret; - } } + + ret = device_create_file(adev->dev, &dev_attr_pp_dpm_sclk); + if (ret) { + DRM_ERROR("failed to create device file pp_dpm_sclk\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_pp_dpm_mclk); + if (ret) { + DRM_ERROR("failed to create device file pp_dpm_mclk\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_pp_dpm_pcie); + if (ret) { + DRM_ERROR("failed to create device file pp_dpm_pcie\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_pp_sclk_od); + if (ret) { + DRM_ERROR("failed to create device file pp_sclk_od\n"); + return ret; + } + ret = device_create_file(adev->dev, &dev_attr_pp_mclk_od); + if (ret) { + DRM_ERROR("failed to create device file pp_mclk_od\n"); + return ret; + } + ret = amdgpu_debugfs_pm_init(adev); if (ret) { DRM_ERROR("Failed to register debugfs file for dpm!\n"); @@ -1146,10 +1263,12 @@ void amdgpu_pm_sysfs_fini(struct amdgpu_device *adev) 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); } + 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); + device_remove_file(adev->dev, &dev_attr_pp_sclk_od); + device_remove_file(adev->dev, &dev_attr_pp_mclk_od); } void amdgpu_pm_compute_clocks(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 82256558e0f5..545074479e1f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c @@ -52,6 +52,9 @@ static int amdgpu_powerplay_init(struct amdgpu_device *adev) pp_init->chip_family = adev->family; pp_init->chip_id = adev->asic_type; pp_init->device = amdgpu_cgs_create_device(adev); + pp_init->rev_id = adev->pdev->revision; + pp_init->sub_sys_id = adev->pdev->subsystem_device; + pp_init->sub_vendor_id = adev->pdev->subsystem_vendor; ret = amd_powerplay_init(pp_init, amd_pp); kfree(pp_init); @@ -105,11 +108,10 @@ static int amdgpu_pp_early_init(void *handle) break; case CHIP_TONGA: case CHIP_FIJI: - adev->pp_enabled = (amdgpu_powerplay == 0) ? false : true; - break; + case CHIP_TOPAZ: case CHIP_CARRIZO: case CHIP_STONEY: - adev->pp_enabled = (amdgpu_powerplay > 0) ? true : false; + adev->pp_enabled = (amdgpu_powerplay == 0) ? false : true; break; /* These chips don't have powerplay implemenations */ case CHIP_BONAIRE: @@ -117,7 +119,6 @@ static int amdgpu_pp_early_init(void *handle) case CHIP_KABINI: case CHIP_MULLINS: case CHIP_KAVERI: - case CHIP_TOPAZ: default: adev->pp_enabled = false; break; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index 870f9494252c..242ba04bfde6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -28,6 +28,7 @@ */ #include <linux/seq_file.h> #include <linux/slab.h> +#include <linux/debugfs.h> #include <drm/drmP.h> #include <drm/amdgpu_drm.h> #include "amdgpu.h" @@ -48,6 +49,7 @@ */ static int amdgpu_debugfs_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring); +static void amdgpu_debugfs_ring_fini(struct amdgpu_ring *ring); /** * amdgpu_ring_alloc - allocate space on the ring buffer @@ -73,6 +75,10 @@ int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw) ring->count_dw = ndw; ring->wptr_old = ring->wptr; + + if (ring->funcs->begin_use) + ring->funcs->begin_use(ring); + return 0; } @@ -125,6 +131,9 @@ void amdgpu_ring_commit(struct amdgpu_ring *ring) mb(); amdgpu_ring_set_wptr(ring); + + if (ring->funcs->end_use) + ring->funcs->end_use(ring); } /** @@ -137,78 +146,9 @@ void amdgpu_ring_commit(struct amdgpu_ring *ring) void amdgpu_ring_undo(struct amdgpu_ring *ring) { ring->wptr = ring->wptr_old; -} - -/** - * amdgpu_ring_backup - Back up the content of a ring - * - * @ring: the ring we want to back up - * - * Saves all unprocessed commits from a ring, returns the number of dwords saved. - */ -unsigned amdgpu_ring_backup(struct amdgpu_ring *ring, - uint32_t **data) -{ - unsigned size, ptr, i; - - *data = NULL; - - if (ring->ring_obj == NULL) - return 0; - - /* it doesn't make sense to save anything if all fences are signaled */ - if (!amdgpu_fence_count_emitted(ring)) - return 0; - - ptr = le32_to_cpu(*ring->next_rptr_cpu_addr); - - size = ring->wptr + (ring->ring_size / 4); - size -= ptr; - size &= ring->ptr_mask; - if (size == 0) - return 0; - - /* and then save the content of the ring */ - *data = kmalloc_array(size, sizeof(uint32_t), GFP_KERNEL); - if (!*data) - return 0; - for (i = 0; i < size; ++i) { - (*data)[i] = ring->ring[ptr++]; - ptr &= ring->ptr_mask; - } - return size; -} - -/** - * amdgpu_ring_restore - append saved commands to the ring again - * - * @ring: ring to append commands to - * @size: number of dwords we want to write - * @data: saved commands - * - * Allocates space on the ring and restore the previously saved commands. - */ -int amdgpu_ring_restore(struct amdgpu_ring *ring, - unsigned size, uint32_t *data) -{ - int i, r; - - if (!size || !data) - return 0; - - /* restore the saved ring content */ - r = amdgpu_ring_alloc(ring, size); - if (r) - return r; - - for (i = 0; i < size; ++i) { - amdgpu_ring_write(ring, data[i]); - } - - amdgpu_ring_commit(ring); - kfree(data); - return 0; + if (ring->funcs->end_use) + ring->funcs->end_use(ring); } /** @@ -260,14 +200,6 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, return r; } - r = amdgpu_wb_get(adev, &ring->next_rptr_offs); - if (r) { - dev_err(adev->dev, "(%d) ring next_rptr wb alloc failed\n", r); - return r; - } - ring->next_rptr_gpu_addr = adev->wb.gpu_addr + ring->next_rptr_offs * 4; - ring->next_rptr_cpu_addr = &adev->wb.wb[ring->next_rptr_offs]; - r = amdgpu_wb_get(adev, &ring->cond_exe_offs); if (r) { dev_err(adev->dev, "(%d) ring cond_exec_polling wb alloc failed\n", r); @@ -276,7 +208,6 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, ring->cond_exe_gpu_addr = adev->wb.gpu_addr + (ring->cond_exe_offs * 4); ring->cond_exe_cpu_addr = &adev->wb.wb[ring->cond_exe_offs]; - spin_lock_init(&ring->fence_lock); r = amdgpu_fence_driver_start_ring(ring, irq_src, irq_type); if (r) { dev_err(adev->dev, "failed initializing fences (%d).\n", r); @@ -291,30 +222,16 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, /* Allocate ring buffer */ if (ring->ring_obj == NULL) { - r = amdgpu_bo_create(adev, ring->ring_size, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, - NULL, NULL, &ring->ring_obj); + r = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &ring->ring_obj, + &ring->gpu_addr, + (void **)&ring->ring); if (r) { dev_err(adev->dev, "(%d) ring create failed\n", r); return r; } - r = amdgpu_bo_reserve(ring->ring_obj, false); - if (unlikely(r != 0)) - return r; - r = amdgpu_bo_pin(ring->ring_obj, AMDGPU_GEM_DOMAIN_GTT, - &ring->gpu_addr); - if (r) { - amdgpu_bo_unreserve(ring->ring_obj); - dev_err(adev->dev, "(%d) ring pin failed\n", r); - return r; - } - r = amdgpu_bo_kmap(ring->ring_obj, - (void **)&ring->ring); - amdgpu_bo_unreserve(ring->ring_obj); - if (r) { - dev_err(adev->dev, "(%d) ring map failed\n", r); - return r; - } + memset((void *)ring->ring, 0, ring->ring_size); } ring->ptr_mask = (ring->ring_size / 4) - 1; ring->max_dw = max_dw; @@ -347,7 +264,6 @@ void amdgpu_ring_fini(struct amdgpu_ring *ring) amdgpu_wb_free(ring->adev, ring->fence_offs); amdgpu_wb_free(ring->adev, ring->rptr_offs); amdgpu_wb_free(ring->adev, ring->wptr_offs); - amdgpu_wb_free(ring->adev, ring->next_rptr_offs); if (ring_obj) { r = amdgpu_bo_reserve(ring_obj, false); @@ -358,6 +274,7 @@ void amdgpu_ring_fini(struct amdgpu_ring *ring) } amdgpu_bo_unref(&ring_obj); } + amdgpu_debugfs_ring_fini(ring); } /* @@ -365,57 +282,62 @@ void amdgpu_ring_fini(struct amdgpu_ring *ring) */ #if defined(CONFIG_DEBUG_FS) -static int amdgpu_debugfs_ring_info(struct seq_file *m, void *data) +/* Layout of file is 12 bytes consisting of + * - rptr + * - wptr + * - driver's copy of wptr + * + * followed by n-words of ring data + */ +static ssize_t amdgpu_debugfs_ring_read(struct file *f, char __user *buf, + size_t size, loff_t *pos) { - 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; - int roffset = (unsigned long)node->info_ent->data; - struct amdgpu_ring *ring = (void *)(((uint8_t*)adev) + roffset); - uint32_t rptr, wptr, rptr_next; - unsigned i; - - wptr = amdgpu_ring_get_wptr(ring); - seq_printf(m, "wptr: 0x%08x [%5d]\n", wptr, wptr); - - rptr = amdgpu_ring_get_rptr(ring); - rptr_next = le32_to_cpu(*ring->next_rptr_cpu_addr); - - seq_printf(m, "rptr: 0x%08x [%5d]\n", rptr, rptr); - - seq_printf(m, "driver's copy of the wptr: 0x%08x [%5d]\n", - ring->wptr, ring->wptr); - - if (!ring->ready) - return 0; - - /* print 8 dw before current rptr as often it's the last executed - * packet that is the root issue - */ - i = (rptr + ring->ptr_mask + 1 - 32) & ring->ptr_mask; - while (i != rptr) { - seq_printf(m, "r[%5d]=0x%08x", i, ring->ring[i]); - if (i == rptr) - seq_puts(m, " *"); - if (i == rptr_next) - seq_puts(m, " #"); - seq_puts(m, "\n"); - i = (i + 1) & ring->ptr_mask; + struct amdgpu_ring *ring = (struct amdgpu_ring*)f->f_inode->i_private; + int r, i; + uint32_t value, result, early[3]; + + if (*pos & 3 || size & 3) + return -EINVAL; + + result = 0; + + if (*pos < 12) { + early[0] = amdgpu_ring_get_rptr(ring); + early[1] = amdgpu_ring_get_wptr(ring); + early[2] = ring->wptr; + for (i = *pos / 4; i < 3 && size; i++) { + r = put_user(early[i], (uint32_t *)buf); + if (r) + return r; + buf += 4; + result += 4; + size -= 4; + *pos += 4; + } } - while (i != wptr) { - seq_printf(m, "r[%5d]=0x%08x", i, ring->ring[i]); - if (i == rptr) - seq_puts(m, " *"); - if (i == rptr_next) - seq_puts(m, " #"); - seq_puts(m, "\n"); - i = (i + 1) & ring->ptr_mask; + + while (size) { + if (*pos >= (ring->ring_size + 12)) + return result; + + value = ring->ring[(*pos - 12)/4]; + r = put_user(value, (uint32_t*)buf); + if (r) + return r; + buf += 4; + result += 4; + size -= 4; + *pos += 4; } - return 0; + + return result; } -static struct drm_info_list amdgpu_debugfs_ring_info_list[AMDGPU_MAX_RINGS]; -static char amdgpu_debugfs_ring_names[AMDGPU_MAX_RINGS][32]; +static const struct file_operations amdgpu_debugfs_ring_fops = { + .owner = THIS_MODULE, + .read = amdgpu_debugfs_ring_read, + .llseek = default_llseek +}; #endif @@ -423,28 +345,27 @@ static int amdgpu_debugfs_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring) { #if defined(CONFIG_DEBUG_FS) - unsigned offset = (uint8_t*)ring - (uint8_t*)adev; - unsigned i; - struct drm_info_list *info; - char *name; - - for (i = 0; i < ARRAY_SIZE(amdgpu_debugfs_ring_info_list); ++i) { - info = &amdgpu_debugfs_ring_info_list[i]; - if (!info->data) - break; - } - - if (i == ARRAY_SIZE(amdgpu_debugfs_ring_info_list)) - return -ENOSPC; + struct drm_minor *minor = adev->ddev->primary; + struct dentry *ent, *root = minor->debugfs_root; + char name[32]; - name = &amdgpu_debugfs_ring_names[i][0]; sprintf(name, "amdgpu_ring_%s", ring->name); - info->name = name; - info->show = amdgpu_debugfs_ring_info; - info->driver_features = 0; - info->data = (void*)(uintptr_t)offset; - return amdgpu_debugfs_add_files(adev, info, 1); + ent = debugfs_create_file(name, + S_IFREG | S_IRUGO, root, + ring, &amdgpu_debugfs_ring_fops); + if (IS_ERR(ent)) + return PTR_ERR(ent); + + i_size_write(ent->d_inode, ring->ring_size + 12); + ring->ent = ent; #endif return 0; } + +static void amdgpu_debugfs_ring_fini(struct amdgpu_ring *ring) +{ +#if defined(CONFIG_DEBUG_FS) + debugfs_remove(ring->ent); +#endif +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c index 48618ee324eb..d8af37a845f4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c @@ -428,7 +428,7 @@ void amdgpu_sa_bo_dump_debug_info(struct amdgpu_sa_manager *sa_manager, soffset, eoffset, eoffset - soffset); if (i->fence) - seq_printf(m, " protected by 0x%08x on context %d", + seq_printf(m, " protected by 0x%08x on context %llu", i->fence->seqno, i->fence->context); seq_printf(m, "\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c index 34a92808bbd4..5c8d3022fb87 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c @@ -223,13 +223,16 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, } /** - * amdgpu_sync_is_idle - test if all fences are signaled + * amdgpu_sync_peek_fence - get the next fence not signaled yet * * @sync: the sync object + * @ring: optional ring to use for test * - * Returns true if all fences in the sync object are signaled. + * Returns the next fence not signaled yet without removing it from the sync + * object. */ -bool amdgpu_sync_is_idle(struct amdgpu_sync *sync) +struct fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync, + struct amdgpu_ring *ring) { struct amdgpu_sync_entry *e; struct hlist_node *tmp; @@ -237,6 +240,19 @@ bool amdgpu_sync_is_idle(struct amdgpu_sync *sync) hash_for_each_safe(sync->fences, i, tmp, e, node) { struct fence *f = e->fence; + struct amd_sched_fence *s_fence = to_amd_sched_fence(f); + + if (ring && s_fence) { + /* For fences from the same ring it is sufficient + * when they are scheduled. + */ + if (s_fence->sched == &ring->sched) { + if (fence_is_signaled(&s_fence->scheduled)) + continue; + + return &s_fence->scheduled; + } + } if (fence_is_signaled(f)) { hash_del(&e->node); @@ -245,58 +261,19 @@ bool amdgpu_sync_is_idle(struct amdgpu_sync *sync) continue; } - return false; + return f; } - return true; + return NULL; } /** - * amdgpu_sync_cycle_fences - move fences from one sync object into another + * amdgpu_sync_get_fence - get the next fence from the sync object * - * @dst: the destination sync object - * @src: the source sync object - * @fence: fence to add to source + * @sync: sync object to use * - * Remove all fences from source and put them into destination and add - * fence as new one into source. + * Get and removes the next fence from the sync object not signaled yet. */ -int amdgpu_sync_cycle_fences(struct amdgpu_sync *dst, struct amdgpu_sync *src, - struct fence *fence) -{ - struct amdgpu_sync_entry *e, *newone; - struct hlist_node *tmp; - int i; - - /* Allocate the new entry before moving the old ones */ - newone = kmem_cache_alloc(amdgpu_sync_slab, GFP_KERNEL); - if (!newone) - return -ENOMEM; - - hash_for_each_safe(src->fences, i, tmp, e, node) { - struct fence *f = e->fence; - - hash_del(&e->node); - if (fence_is_signaled(f)) { - fence_put(f); - kmem_cache_free(amdgpu_sync_slab, e); - continue; - } - - if (amdgpu_sync_add_later(dst, f)) { - kmem_cache_free(amdgpu_sync_slab, e); - continue; - } - - hash_add(dst->fences, &e->node, f->context); - } - - hash_add(src->fences, &newone->node, fence->context); - newone->fence = fence_get(fence); - - return 0; -} - struct fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync) { struct amdgpu_sync_entry *e; @@ -319,25 +296,6 @@ struct fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync) return NULL; } -int amdgpu_sync_wait(struct amdgpu_sync *sync) -{ - struct amdgpu_sync_entry *e; - struct hlist_node *tmp; - int i, r; - - hash_for_each_safe(sync->fences, i, tmp, e, node) { - r = fence_wait(e->fence, false); - if (r) - return r; - - hash_del(&e->node); - fence_put(e->fence); - kmem_cache_free(amdgpu_sync_slab, e); - } - - return 0; -} - /** * amdgpu_sync_free - free the sync object * diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c index 05a53f4fc334..b827c75e95de 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c @@ -111,7 +111,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) amdgpu_bo_kunmap(gtt_obj[i]); r = amdgpu_copy_buffer(ring, gtt_addr, vram_addr, - size, NULL, &fence); + size, NULL, &fence, false); if (r) { DRM_ERROR("Failed GTT->VRAM copy %d\n", i); @@ -156,7 +156,7 @@ 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); + size, NULL, &fence, false); if (r) { DRM_ERROR("Failed VRAM->GTT copy %d\n", i); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h index 26a5f4acf584..0d8d65eb46cd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h @@ -11,19 +11,68 @@ #define TRACE_SYSTEM amdgpu #define TRACE_INCLUDE_FILE amdgpu_trace +TRACE_EVENT(amdgpu_mm_rreg, + TP_PROTO(unsigned did, uint32_t reg, uint32_t value), + TP_ARGS(did, reg, value), + TP_STRUCT__entry( + __field(unsigned, did) + __field(uint32_t, reg) + __field(uint32_t, value) + ), + TP_fast_assign( + __entry->did = did; + __entry->reg = reg; + __entry->value = value; + ), + TP_printk("0x%04lx, 0x%04lx, 0x%08lx", + (unsigned long)__entry->did, + (unsigned long)__entry->reg, + (unsigned long)__entry->value) +); + +TRACE_EVENT(amdgpu_mm_wreg, + TP_PROTO(unsigned did, uint32_t reg, uint32_t value), + TP_ARGS(did, reg, value), + TP_STRUCT__entry( + __field(unsigned, did) + __field(uint32_t, reg) + __field(uint32_t, value) + ), + TP_fast_assign( + __entry->did = did; + __entry->reg = reg; + __entry->value = value; + ), + TP_printk("0x%04lx, 0x%04lx, 0x%08lx", + (unsigned long)__entry->did, + (unsigned long)__entry->reg, + (unsigned long)__entry->value) +); + TRACE_EVENT(amdgpu_bo_create, TP_PROTO(struct amdgpu_bo *bo), TP_ARGS(bo), TP_STRUCT__entry( __field(struct amdgpu_bo *, bo) __field(u32, pages) + __field(u32, type) + __field(u32, prefer) + __field(u32, allow) + __field(u32, visible) ), TP_fast_assign( __entry->bo = bo; __entry->pages = bo->tbo.num_pages; + __entry->type = bo->tbo.mem.mem_type; + __entry->prefer = bo->prefered_domains; + __entry->allow = bo->allowed_domains; + __entry->visible = bo->flags; ), - TP_printk("bo=%p, pages=%u", __entry->bo, __entry->pages) + + TP_printk("bo=%p,pages=%u,type=%d,prefered=%d,allowed=%d,visible=%d", + __entry->bo, __entry->pages, __entry->type, + __entry->prefer, __entry->allow, __entry->visible) ); TRACE_EVENT(amdgpu_cs, @@ -64,7 +113,7 @@ TRACE_EVENT(amdgpu_cs_ioctl, __entry->adev = job->adev; __entry->sched_job = &job->base; __entry->ib = job->ibs; - __entry->fence = &job->base.s_fence->base; + __entry->fence = &job->base.s_fence->finished; __entry->ring_name = job->ring->name; __entry->num_ibs = job->num_ibs; ), @@ -89,7 +138,7 @@ TRACE_EVENT(amdgpu_sched_run_job, __entry->adev = job->adev; __entry->sched_job = &job->base; __entry->ib = job->ibs; - __entry->fence = &job->base.s_fence->base; + __entry->fence = &job->base.s_fence->finished; __entry->ring_name = job->ring->name; __entry->num_ibs = job->num_ibs; ), @@ -100,24 +149,26 @@ TRACE_EVENT(amdgpu_sched_run_job, TRACE_EVENT(amdgpu_vm_grab_id, - TP_PROTO(struct amdgpu_vm *vm, int ring, unsigned vmid, - uint64_t pd_addr), - TP_ARGS(vm, ring, vmid, pd_addr), + TP_PROTO(struct amdgpu_vm *vm, int ring, struct amdgpu_job *job), + TP_ARGS(vm, ring, job), TP_STRUCT__entry( __field(struct amdgpu_vm *, vm) __field(u32, ring) __field(u32, vmid) __field(u64, pd_addr) + __field(u32, needs_flush) ), TP_fast_assign( __entry->vm = vm; __entry->ring = ring; - __entry->vmid = vmid; - __entry->pd_addr = pd_addr; + __entry->vmid = job->vm_id; + __entry->pd_addr = job->vm_pd_addr; + __entry->needs_flush = job->vm_needs_flush; ), - TP_printk("vm=%p, ring=%u, id=%u, pd_addr=%010Lx", __entry->vm, - __entry->ring, __entry->vmid, __entry->pd_addr) + TP_printk("vm=%p, ring=%u, id=%u, pd_addr=%010Lx needs_flush=%u", + __entry->vm, __entry->ring, __entry->vmid, + __entry->pd_addr, __entry->needs_flush) ); TRACE_EVENT(amdgpu_vm_bo_map, @@ -244,13 +295,55 @@ TRACE_EVENT(amdgpu_bo_list_set, TP_STRUCT__entry( __field(struct amdgpu_bo_list *, list) __field(struct amdgpu_bo *, bo) + __field(u64, bo_size) ), TP_fast_assign( __entry->list = list; __entry->bo = bo; + __entry->bo_size = amdgpu_bo_size(bo); ), - TP_printk("list=%p, bo=%p", __entry->list, __entry->bo) + TP_printk("list=%p, bo=%p, bo_size = %Ld", + __entry->list, + __entry->bo, + __entry->bo_size) +); + +TRACE_EVENT(amdgpu_cs_bo_status, + TP_PROTO(uint64_t total_bo, uint64_t total_size), + TP_ARGS(total_bo, total_size), + TP_STRUCT__entry( + __field(u64, total_bo) + __field(u64, total_size) + ), + + TP_fast_assign( + __entry->total_bo = total_bo; + __entry->total_size = total_size; + ), + TP_printk("total bo size = %Ld, total bo count = %Ld", + __entry->total_bo, __entry->total_size) +); + +TRACE_EVENT(amdgpu_ttm_bo_move, + TP_PROTO(struct amdgpu_bo* bo, uint32_t new_placement, uint32_t old_placement), + TP_ARGS(bo, new_placement, old_placement), + TP_STRUCT__entry( + __field(struct amdgpu_bo *, bo) + __field(u64, bo_size) + __field(u32, new_placement) + __field(u32, old_placement) + ), + + TP_fast_assign( + __entry->bo = bo; + __entry->bo_size = amdgpu_bo_size(bo); + __entry->new_placement = new_placement; + __entry->old_placement = old_placement; + ), + TP_printk("bo=%p from:%d to %d with size = %Ld", + __entry->bo, __entry->old_placement, + __entry->new_placement, __entry->bo_size) ); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 3b9053af4762..5447973483ec 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -34,6 +34,7 @@ #include <ttm/ttm_placement.h> #include <ttm/ttm_module.h> #include <ttm/ttm_page_alloc.h> +#include <ttm/ttm_memory.h> #include <drm/drmP.h> #include <drm/amdgpu_drm.h> #include <linux/seq_file.h> @@ -74,7 +75,7 @@ static void amdgpu_ttm_mem_global_release(struct drm_global_reference *ref) ttm_mem_global_release(ref->object); } -static int amdgpu_ttm_global_init(struct amdgpu_device *adev) +int amdgpu_ttm_global_init(struct amdgpu_device *adev) { struct drm_global_reference *global_ref; struct amdgpu_ring *ring; @@ -256,10 +257,8 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo, switch (old_mem->mem_type) { case TTM_PL_VRAM: - old_start += adev->mc.vram_start; - break; case TTM_PL_TT: - old_start += adev->mc.gtt_start; + old_start += bo->bdev->man[old_mem->mem_type].gpu_offset; break; default: DRM_ERROR("Unknown placement %d\n", old_mem->mem_type); @@ -267,10 +266,8 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo, } switch (new_mem->mem_type) { case TTM_PL_VRAM: - new_start += adev->mc.vram_start; - break; case TTM_PL_TT: - new_start += adev->mc.gtt_start; + new_start += bo->bdev->man[new_mem->mem_type].gpu_offset; break; default: DRM_ERROR("Unknown placement %d\n", old_mem->mem_type); @@ -285,10 +282,11 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo, r = amdgpu_copy_buffer(ring, old_start, new_start, new_mem->num_pages * PAGE_SIZE, /* bytes */ - bo->resv, &fence); - /* FIXME: handle copy error */ - r = ttm_bo_move_accel_cleanup(bo, fence, - evict, no_wait_gpu, new_mem); + bo->resv, &fence, false); + if (r) + return r; + + r = ttm_bo_pipeline_move(bo, fence, evict, new_mem); fence_put(fence); return r; } @@ -334,7 +332,7 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, if (unlikely(r)) { goto out_cleanup; } - r = ttm_bo_move_ttm(bo, true, no_wait_gpu, new_mem); + r = ttm_bo_move_ttm(bo, interruptible, no_wait_gpu, new_mem); out_cleanup: ttm_bo_mem_put(bo, &tmp_mem); return r; @@ -367,7 +365,7 @@ static int amdgpu_move_ram_vram(struct ttm_buffer_object *bo, if (unlikely(r)) { return r; } - r = ttm_bo_move_ttm(bo, true, no_wait_gpu, &tmp_mem); + r = ttm_bo_move_ttm(bo, interruptible, no_wait_gpu, &tmp_mem); if (unlikely(r)) { goto out_cleanup; } @@ -396,6 +394,11 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, return -EINVAL; adev = amdgpu_get_adev(bo->bdev); + + /* remember the eviction */ + if (evict) + atomic64_inc(&adev->num_evictions); + if (old_mem->mem_type == TTM_PL_SYSTEM && bo->ttm == NULL) { amdgpu_move_null(bo, new_mem); return 0; @@ -429,7 +432,7 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, if (r) { memcpy: - r = ttm_bo_move_memcpy(bo, evict, no_wait_gpu, new_mem); + r = ttm_bo_move_memcpy(bo, interruptible, no_wait_gpu, new_mem); if (r) { return r; } @@ -943,6 +946,8 @@ static struct list_head *amdgpu_ttm_lru_tail(struct ttm_buffer_object *tbo) struct list_head *res = lru->lru[tbo->mem.mem_type]; lru->lru[tbo->mem.mem_type] = &tbo->lru; + while ((++lru)->lru[tbo->mem.mem_type] == res) + lru->lru[tbo->mem.mem_type] = &tbo->lru; return res; } @@ -953,6 +958,8 @@ static struct list_head *amdgpu_ttm_swap_lru_tail(struct ttm_buffer_object *tbo) struct list_head *res = lru->swap_lru; lru->swap_lru = &tbo->swap; + while ((++lru)->swap_lru == res) + lru->swap_lru = &tbo->swap; return res; } @@ -980,10 +987,6 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) unsigned i, j; int r; - r = amdgpu_ttm_global_init(adev); - if (r) { - return r; - } /* No others user of address space so set it to 0 */ r = ttm_bo_device_init(&adev->mman.bdev, adev->mman.bo_global_ref.ref.object, @@ -1004,6 +1007,10 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) lru->swap_lru = &adev->mman.bdev.glob->swap_lru; } + for (j = 0; j < TTM_NUM_MEM_TYPES; ++j) + adev->mman.guard.lru[j] = NULL; + adev->mman.guard.swap_lru = NULL; + adev->mman.initialized = true; r = ttm_bo_init_mm(&adev->mman.bdev, TTM_PL_VRAM, adev->mc.real_vram_size >> PAGE_SHIFT); @@ -1144,7 +1151,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t dst_offset, uint32_t byte_count, struct reservation_object *resv, - struct fence **fence) + struct fence **fence, bool direct_submit) { struct amdgpu_device *adev = ring->adev; struct amdgpu_job *job; @@ -1188,8 +1195,79 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, amdgpu_ring_pad_ib(ring, &job->ibs[0]); WARN_ON(job->ibs[0].length_dw > num_dw); + if (direct_submit) { + r = amdgpu_ib_schedule(ring, job->num_ibs, job->ibs, + NULL, NULL, fence); + job->fence = fence_get(*fence); + if (r) + DRM_ERROR("Error scheduling IBs (%d)\n", r); + amdgpu_job_free(job); + } else { + r = amdgpu_job_submit(job, ring, &adev->mman.entity, + AMDGPU_FENCE_OWNER_UNDEFINED, fence); + if (r) + goto error_free; + } + + return r; + +error_free: + amdgpu_job_free(job); + return r; +} + +int amdgpu_fill_buffer(struct amdgpu_bo *bo, + uint32_t src_data, + struct reservation_object *resv, + struct fence **fence) +{ + struct amdgpu_device *adev = bo->adev; + struct amdgpu_job *job; + struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; + + uint32_t max_bytes, byte_count; + uint64_t dst_offset; + unsigned int num_loops, num_dw; + unsigned int i; + int r; + + byte_count = bo->tbo.num_pages << PAGE_SHIFT; + max_bytes = adev->mman.buffer_funcs->fill_max_bytes; + num_loops = DIV_ROUND_UP(byte_count, max_bytes); + num_dw = num_loops * adev->mman.buffer_funcs->fill_num_dw; + + /* for IB padding */ + while (num_dw & 0x7) + num_dw++; + + r = amdgpu_job_alloc_with_ib(adev, num_dw * 4, &job); + if (r) + return r; + + if (resv) { + r = amdgpu_sync_resv(adev, &job->sync, resv, + AMDGPU_FENCE_OWNER_UNDEFINED); + if (r) { + DRM_ERROR("sync failed (%d).\n", r); + goto error_free; + } + } + + dst_offset = bo->tbo.mem.start << PAGE_SHIFT; + for (i = 0; i < num_loops; i++) { + uint32_t cur_size_in_bytes = min(byte_count, max_bytes); + + amdgpu_emit_fill_buffer(adev, &job->ibs[0], src_data, + dst_offset, cur_size_in_bytes); + + dst_offset += cur_size_in_bytes; + byte_count -= cur_size_in_bytes; + } + + amdgpu_ring_pad_ib(ring, &job->ibs[0]); + WARN_ON(job->ibs[0].length_dw > num_dw); r = amdgpu_job_submit(job, ring, &adev->mman.entity, - AMDGPU_FENCE_OWNER_UNDEFINED, fence); + AMDGPU_FENCE_OWNER_UNDEFINED, fence); if (r) goto error_free; @@ -1380,3 +1458,8 @@ static void amdgpu_ttm_debugfs_fini(struct amdgpu_device *adev) #endif } + +u64 amdgpu_ttm_get_gtt_mem_size(struct amdgpu_device *adev) +{ + return ttm_get_kernel_zone_memory_size(adev->mman.mem_global_ref.object); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h new file mode 100644 index 000000000000..72f6bfc15d8f --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -0,0 +1,80 @@ +/* + * Copyright 2016 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_TTM_H__ +#define __AMDGPU_TTM_H__ + +#include "gpu_scheduler.h" + +#define AMDGPU_PL_GDS TTM_PL_PRIV0 +#define AMDGPU_PL_GWS TTM_PL_PRIV1 +#define AMDGPU_PL_OA TTM_PL_PRIV2 + +#define AMDGPU_PL_FLAG_GDS TTM_PL_FLAG_PRIV0 +#define AMDGPU_PL_FLAG_GWS TTM_PL_FLAG_PRIV1 +#define AMDGPU_PL_FLAG_OA TTM_PL_FLAG_PRIV2 + +#define AMDGPU_TTM_LRU_SIZE 20 + +struct amdgpu_mman_lru { + struct list_head *lru[TTM_NUM_MEM_TYPES]; + struct list_head *swap_lru; +}; + +struct amdgpu_mman { + struct ttm_bo_global_ref bo_global_ref; + struct drm_global_reference mem_global_ref; + struct ttm_bo_device bdev; + bool mem_global_referenced; + bool initialized; + +#if defined(CONFIG_DEBUG_FS) + struct dentry *vram; + struct dentry *gtt; +#endif + + /* buffer handling */ + const struct amdgpu_buffer_funcs *buffer_funcs; + struct amdgpu_ring *buffer_funcs_ring; + /* Scheduler entity for buffer moves */ + struct amd_sched_entity entity; + + /* custom LRU management */ + struct amdgpu_mman_lru log2_size[AMDGPU_TTM_LRU_SIZE]; + /* guard for log2_size array, don't add anything in between */ + struct amdgpu_mman_lru guard; +}; + +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 fence **fence, bool direct_submit); +int amdgpu_fill_buffer(struct amdgpu_bo *bo, + uint32_t src_data, + struct reservation_object *resv, + struct fence **fence); + +int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma); +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index d9c88d13f8db..cc766cc53a87 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -40,9 +40,16 @@ #include "uvd/uvd_4_2_d.h" /* 1 second timeout */ -#define UVD_IDLE_TIMEOUT_MS 1000 +#define UVD_IDLE_TIMEOUT msecs_to_jiffies(1000) + +/* Firmware versions for VI */ +#define FW_1_65_10 ((1 << 24) | (65 << 16) | (10 << 8)) +#define FW_1_87_11 ((1 << 24) | (87 << 16) | (11 << 8)) +#define FW_1_87_12 ((1 << 24) | (87 << 16) | (12 << 8)) +#define FW_1_37_15 ((1 << 24) | (37 << 16) | (15 << 8)) + /* Polaris10/11 firmware version */ -#define FW_1_66_16 ((1 << 24) | (66 << 16) | (16 << 8)) +#define FW_1_66_16 ((1 << 24) | (66 << 16) | (16 << 8)) /* Firmware Names */ #ifdef CONFIG_DRM_AMDGPU_CIK @@ -92,7 +99,6 @@ MODULE_FIRMWARE(FIRMWARE_STONEY); MODULE_FIRMWARE(FIRMWARE_POLARIS10); MODULE_FIRMWARE(FIRMWARE_POLARIS11); -static void amdgpu_uvd_note_usage(struct amdgpu_device *adev); static void amdgpu_uvd_idle_work_handler(struct work_struct *work); int amdgpu_uvd_sw_init(struct amdgpu_device *adev) @@ -195,39 +201,14 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) bo_size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8) + AMDGPU_UVD_STACK_SIZE + AMDGPU_UVD_HEAP_SIZE + AMDGPU_UVD_SESSION_SIZE * adev->uvd.max_handles; - r = amdgpu_bo_create(adev, bo_size, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED, - NULL, NULL, &adev->uvd.vcpu_bo); + r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, &adev->uvd.vcpu_bo, + &adev->uvd.gpu_addr, &adev->uvd.cpu_addr); if (r) { dev_err(adev->dev, "(%d) failed to allocate UVD bo\n", r); return r; } - r = amdgpu_bo_reserve(adev->uvd.vcpu_bo, false); - if (r) { - amdgpu_bo_unref(&adev->uvd.vcpu_bo); - dev_err(adev->dev, "(%d) failed to reserve UVD bo\n", r); - return r; - } - - r = amdgpu_bo_pin(adev->uvd.vcpu_bo, AMDGPU_GEM_DOMAIN_VRAM, - &adev->uvd.gpu_addr); - if (r) { - amdgpu_bo_unreserve(adev->uvd.vcpu_bo); - amdgpu_bo_unref(&adev->uvd.vcpu_bo); - dev_err(adev->dev, "(%d) UVD bo pin failed\n", r); - return r; - } - - r = amdgpu_bo_kmap(adev->uvd.vcpu_bo, &adev->uvd.cpu_addr); - if (r) { - dev_err(adev->dev, "(%d) UVD map failed\n", r); - return r; - } - - amdgpu_bo_unreserve(adev->uvd.vcpu_bo); - ring = &adev->uvd.ring; rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL]; r = amd_sched_entity_init(&ring->sched, &adev->uvd.entity, @@ -246,6 +227,23 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) if (!amdgpu_ip_block_version_cmp(adev, AMD_IP_BLOCK_TYPE_UVD, 5, 0)) adev->uvd.address_64_bit = true; + switch (adev->asic_type) { + case CHIP_TONGA: + adev->uvd.use_ctx_buf = adev->uvd.fw_version >= FW_1_65_10; + break; + case CHIP_CARRIZO: + adev->uvd.use_ctx_buf = adev->uvd.fw_version >= FW_1_87_11; + break; + case CHIP_FIJI: + adev->uvd.use_ctx_buf = adev->uvd.fw_version >= FW_1_87_12; + break; + case CHIP_STONEY: + adev->uvd.use_ctx_buf = adev->uvd.fw_version >= FW_1_37_15; + break; + default: + adev->uvd.use_ctx_buf = adev->asic_type >= CHIP_POLARIS10; + } + return 0; } @@ -300,7 +298,7 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev) if (!adev->uvd.saved_bo) return -ENOMEM; - memcpy(adev->uvd.saved_bo, ptr, size); + memcpy_fromio(adev->uvd.saved_bo, ptr, size); return 0; } @@ -317,7 +315,7 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev) ptr = adev->uvd.cpu_addr; if (adev->uvd.saved_bo != NULL) { - memcpy(ptr, adev->uvd.saved_bo, size); + memcpy_toio(ptr, adev->uvd.saved_bo, size); kfree(adev->uvd.saved_bo); adev->uvd.saved_bo = NULL; } else { @@ -326,11 +324,11 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev) hdr = (const struct common_firmware_header *)adev->uvd.fw->data; offset = le32_to_cpu(hdr->ucode_array_offset_bytes); - memcpy(adev->uvd.cpu_addr, (adev->uvd.fw->data) + offset, - (adev->uvd.fw->size) - offset); + memcpy_toio(adev->uvd.cpu_addr, adev->uvd.fw->data + offset, + le32_to_cpu(hdr->ucode_size_bytes)); size -= le32_to_cpu(hdr->ucode_size_bytes); ptr += le32_to_cpu(hdr->ucode_size_bytes); - memset(ptr, 0, size); + memset_io(ptr, 0, size); } return 0; @@ -346,8 +344,6 @@ void amdgpu_uvd_free_handles(struct amdgpu_device *adev, struct drm_file *filp) if (handle != 0 && adev->uvd.filp[i] == filp) { struct fence *fence; - amdgpu_uvd_note_usage(adev); - r = amdgpu_uvd_get_destroy_msg(ring, handle, false, &fence); if (r) { @@ -438,7 +434,7 @@ static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg, unsigned fs_in_mb = width_in_mb * height_in_mb; unsigned image_size, tmp, min_dpb_size, num_dpb_buffer; - unsigned min_ctx_size = 0; + unsigned min_ctx_size = ~0; image_size = width * height; image_size += image_size / 2; @@ -557,7 +553,7 @@ static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg, /* reference picture buffer */ min_dpb_size = image_size * num_dpb_buffer; - if (adev->asic_type < CHIP_POLARIS10){ + if (!adev->uvd.use_ctx_buf){ /* macroblock context buffer */ min_dpb_size += width_in_mb * height_in_mb * num_dpb_buffer * 192; @@ -662,7 +658,7 @@ static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx, } DRM_ERROR("No more free UVD handles!\n"); - return -EINVAL; + return -ENOSPC; case 1: /* it's a decode msg, calc buffer sizes */ @@ -822,6 +818,7 @@ static int amdgpu_uvd_cs_reg(struct amdgpu_uvd_cs_ctx *ctx, return r; break; case mmUVD_ENGINE_CNTL: + case mmUVD_NO_OP: break; default: DRM_ERROR("Invalid reg 0x%X!\n", reg); @@ -913,8 +910,6 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx) return -EINVAL; } - amdgpu_uvd_note_usage(ctx.parser->adev); - return 0; } @@ -962,13 +957,15 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, ib->ptr[3] = addr >> 32; ib->ptr[4] = PACKET0(mmUVD_GPCOM_VCPU_CMD, 0); ib->ptr[5] = 0; - for (i = 6; i < 16; ++i) - ib->ptr[i] = PACKET2(0); + for (i = 6; i < 16; i += 2) { + ib->ptr[i] = PACKET0(mmUVD_NO_OP, 0); + ib->ptr[i+1] = 0; + } ib->length_dw = 16; if (direct) { r = amdgpu_ib_schedule(ring, 1, ib, NULL, NULL, &f); - job->fence = f; + job->fence = fence_get(f); if (r) goto err_free; @@ -1095,35 +1092,23 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work) { struct amdgpu_device *adev = container_of(work, struct amdgpu_device, uvd.idle_work.work); - unsigned i, fences, handles = 0; - - fences = amdgpu_fence_count_emitted(&adev->uvd.ring); - - for (i = 0; i < adev->uvd.max_handles; ++i) - if (atomic_read(&adev->uvd.handles[i])) - ++handles; + unsigned fences = amdgpu_fence_count_emitted(&adev->uvd.ring); - if (fences == 0 && handles == 0) { + if (fences == 0) { if (adev->pm.dpm_enabled) { amdgpu_dpm_enable_uvd(adev, false); - /* just work around for uvd clock remain high even - * when uvd dpm disabled on Polaris10 */ - if (adev->asic_type == CHIP_POLARIS10) - amdgpu_asic_set_uvd_clocks(adev, 0, 0); } else { amdgpu_asic_set_uvd_clocks(adev, 0, 0); } } else { - schedule_delayed_work(&adev->uvd.idle_work, - msecs_to_jiffies(UVD_IDLE_TIMEOUT_MS)); + schedule_delayed_work(&adev->uvd.idle_work, UVD_IDLE_TIMEOUT); } } -static void amdgpu_uvd_note_usage(struct amdgpu_device *adev) +void amdgpu_uvd_ring_begin_use(struct amdgpu_ring *ring) { + struct amdgpu_device *adev = ring->adev; bool set_clocks = !cancel_delayed_work_sync(&adev->uvd.idle_work); - set_clocks &= schedule_delayed_work(&adev->uvd.idle_work, - msecs_to_jiffies(UVD_IDLE_TIMEOUT_MS)); if (set_clocks) { if (adev->pm.dpm_enabled) { @@ -1133,3 +1118,48 @@ static void amdgpu_uvd_note_usage(struct amdgpu_device *adev) } } } + +void amdgpu_uvd_ring_end_use(struct amdgpu_ring *ring) +{ + schedule_delayed_work(&ring->adev->uvd.idle_work, UVD_IDLE_TIMEOUT); +} + +/** + * amdgpu_uvd_ring_test_ib - test ib execution + * + * @ring: amdgpu_ring pointer + * + * Test if we can successfully execute an IB + */ +int amdgpu_uvd_ring_test_ib(struct amdgpu_ring *ring, long timeout) +{ + struct fence *fence; + long r; + + r = amdgpu_uvd_get_create_msg(ring, 1, NULL); + if (r) { + DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r); + goto error; + } + + r = amdgpu_uvd_get_destroy_msg(ring, 1, true, &fence); + if (r) { + DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r); + goto error; + } + + r = 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: + fence_put(fence); + return r; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h index 9a3b449081a7..c850009602d1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h @@ -35,5 +35,8 @@ int amdgpu_uvd_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, void amdgpu_uvd_free_handles(struct amdgpu_device *adev, struct drm_file *filp); int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx); +void amdgpu_uvd_ring_begin_use(struct amdgpu_ring *ring); +void amdgpu_uvd_ring_end_use(struct amdgpu_ring *ring); +int amdgpu_uvd_ring_test_ib(struct amdgpu_ring *ring, long timeout); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index 875626a2eccb..da52af2a935a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -36,7 +36,7 @@ #include "cikd.h" /* 1 second timeout */ -#define VCE_IDLE_TIMEOUT_MS 1000 +#define VCE_IDLE_TIMEOUT msecs_to_jiffies(1000) /* Firmware Names */ #ifdef CONFIG_DRM_AMDGPU_CIK @@ -85,8 +85,6 @@ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size) unsigned ucode_version, version_major, version_minor, binary_id; int i, r; - INIT_DELAYED_WORK(&adev->vce.idle_work, amdgpu_vce_idle_work_handler); - switch (adev->asic_type) { #ifdef CONFIG_DRM_AMDGPU_CIK case CHIP_BONAIRE: @@ -197,6 +195,9 @@ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size) adev->vce.filp[i] = NULL; } + INIT_DELAYED_WORK(&adev->vce.idle_work, amdgpu_vce_idle_work_handler); + mutex_init(&adev->vce.idle_mutex); + return 0; } @@ -220,6 +221,7 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev) amdgpu_ring_fini(&adev->vce.ring[1]); release_firmware(adev->vce.fw); + mutex_destroy(&adev->vce.idle_mutex); return 0; } @@ -280,8 +282,8 @@ int amdgpu_vce_resume(struct amdgpu_device *adev) hdr = (const struct common_firmware_header *)adev->vce.fw->data; offset = le32_to_cpu(hdr->ucode_array_offset_bytes); - memcpy(cpu_addr, (adev->vce.fw->data) + offset, - (adev->vce.fw->size) - offset); + memcpy_toio(cpu_addr, adev->vce.fw->data + offset, + adev->vce.fw->size - offset); amdgpu_bo_kunmap(adev->vce.vcpu_bo); @@ -310,37 +312,44 @@ static void amdgpu_vce_idle_work_handler(struct work_struct *work) amdgpu_asic_set_vce_clocks(adev, 0, 0); } } else { - schedule_delayed_work(&adev->vce.idle_work, - msecs_to_jiffies(VCE_IDLE_TIMEOUT_MS)); + schedule_delayed_work(&adev->vce.idle_work, VCE_IDLE_TIMEOUT); } } /** - * amdgpu_vce_note_usage - power up VCE + * amdgpu_vce_ring_begin_use - power up VCE * - * @adev: amdgpu_device pointer + * @ring: amdgpu ring * * Make sure VCE is powerd up when we want to use it */ -static void amdgpu_vce_note_usage(struct amdgpu_device *adev) +void amdgpu_vce_ring_begin_use(struct amdgpu_ring *ring) { - bool streams_changed = false; - bool set_clocks = !cancel_delayed_work_sync(&adev->vce.idle_work); - set_clocks &= schedule_delayed_work(&adev->vce.idle_work, - msecs_to_jiffies(VCE_IDLE_TIMEOUT_MS)); - - if (adev->pm.dpm_enabled) { - /* XXX figure out if the streams changed */ - streams_changed = false; - } + struct amdgpu_device *adev = ring->adev; + bool set_clocks; - if (set_clocks || streams_changed) { + mutex_lock(&adev->vce.idle_mutex); + set_clocks = !cancel_delayed_work_sync(&adev->vce.idle_work); + if (set_clocks) { if (adev->pm.dpm_enabled) { amdgpu_dpm_enable_vce(adev, true); } else { amdgpu_asic_set_vce_clocks(adev, 53300, 40000); } } + mutex_unlock(&adev->vce.idle_mutex); +} + +/** + * amdgpu_vce_ring_end_use - power VCE down + * + * @ring: amdgpu ring + * + * Schedule work to power VCE down again + */ +void amdgpu_vce_ring_end_use(struct amdgpu_ring *ring) +{ + schedule_delayed_work(&ring->adev->vce.idle_work, VCE_IDLE_TIMEOUT); } /** @@ -357,11 +366,10 @@ void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp) int i, r; for (i = 0; i < AMDGPU_MAX_VCE_HANDLES; ++i) { uint32_t handle = atomic_read(&adev->vce.handles[i]); + if (!handle || adev->vce.filp[i] != filp) continue; - amdgpu_vce_note_usage(adev); - r = amdgpu_vce_get_destroy_msg(ring, handle, false, NULL); if (r) DRM_ERROR("Error destroying VCE handle (%d)!\n", r); @@ -437,7 +445,7 @@ int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, ib->ptr[i] = 0x0; r = amdgpu_ib_schedule(ring, 1, ib, NULL, NULL, &f); - job->fence = f; + job->fence = fence_get(f); if (r) goto err; @@ -469,7 +477,6 @@ int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, struct amdgpu_job *job; struct amdgpu_ib *ib; struct fence *f = NULL; - uint64_t dummy; int i, r; r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, &job); @@ -477,7 +484,6 @@ int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, return r; ib = &job->ibs[0]; - dummy = ib->gpu_addr + 1024; /* stitch together an VCE destroy msg */ ib->length_dw = 0; @@ -485,11 +491,14 @@ int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, ib->ptr[ib->length_dw++] = 0x00000001; /* session cmd */ ib->ptr[ib->length_dw++] = handle; - ib->ptr[ib->length_dw++] = 0x00000014; /* len */ - ib->ptr[ib->length_dw++] = 0x05000005; /* feedback buffer */ - ib->ptr[ib->length_dw++] = upper_32_bits(dummy); - ib->ptr[ib->length_dw++] = dummy; - ib->ptr[ib->length_dw++] = 0x00000001; + ib->ptr[ib->length_dw++] = 0x00000020; /* len */ + ib->ptr[ib->length_dw++] = 0x00000002; /* task info */ + ib->ptr[ib->length_dw++] = 0xffffffff; /* next task info, set to 0xffffffff if no */ + ib->ptr[ib->length_dw++] = 0x00000001; /* destroy session */ + ib->ptr[ib->length_dw++] = 0x00000000; + ib->ptr[ib->length_dw++] = 0x00000000; + ib->ptr[ib->length_dw++] = 0xffffffff; /* feedback is not needed, set to 0xffffffff and firmware will not output feedback */ + ib->ptr[ib->length_dw++] = 0x00000000; ib->ptr[ib->length_dw++] = 0x00000008; /* len */ ib->ptr[ib->length_dw++] = 0x02000001; /* destroy cmd */ @@ -499,7 +508,7 @@ int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, if (direct) { r = amdgpu_ib_schedule(ring, 1, ib, NULL, NULL, &f); - job->fence = f; + job->fence = fence_get(f); if (r) goto err; @@ -580,12 +589,10 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, uint32_t ib_idx, * we we don't have another free session index. */ static int amdgpu_vce_validate_handle(struct amdgpu_cs_parser *p, - uint32_t handle, bool *allocated) + uint32_t handle, uint32_t *allocated) { unsigned i; - *allocated = false; - /* validate the handle */ for (i = 0; i < AMDGPU_MAX_VCE_HANDLES; ++i) { if (atomic_read(&p->adev->vce.handles[i]) == handle) { @@ -602,7 +609,7 @@ static int amdgpu_vce_validate_handle(struct amdgpu_cs_parser *p, if (!atomic_cmpxchg(&p->adev->vce.handles[i], 0, handle)) { p->adev->vce.filp[i] = p->filp; p->adev->vce.img_size[i] = 0; - *allocated = true; + *allocated |= 1 << i; return i; } } @@ -622,15 +629,13 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) struct amdgpu_ib *ib = &p->job->ibs[ib_idx]; unsigned fb_idx = 0, bs_idx = 0; int session_idx = -1; - bool destroyed = false; - bool created = false; - bool allocated = false; + uint32_t destroyed = 0; + uint32_t created = 0; + uint32_t allocated = 0; uint32_t tmp, handle = 0; uint32_t *size = &tmp; int i, r = 0, idx = 0; - amdgpu_vce_note_usage(p->adev); - 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); @@ -641,30 +646,30 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) goto out; } - if (destroyed) { - DRM_ERROR("No other command allowed after destroy!\n"); - r = -EINVAL; - goto out; - } - switch (cmd) { - case 0x00000001: // session + case 0x00000001: /* session */ handle = amdgpu_get_ib_value(p, ib_idx, idx + 2); session_idx = amdgpu_vce_validate_handle(p, handle, &allocated); - if (session_idx < 0) - return session_idx; + if (session_idx < 0) { + r = session_idx; + goto out; + } size = &p->adev->vce.img_size[session_idx]; break; - case 0x00000002: // task info + case 0x00000002: /* task info */ fb_idx = amdgpu_get_ib_value(p, ib_idx, idx + 6); bs_idx = amdgpu_get_ib_value(p, ib_idx, idx + 7); break; - case 0x01000001: // create - created = true; - if (!allocated) { + case 0x01000001: /* create */ + created |= 1 << session_idx; + if (destroyed & (1 << session_idx)) { + destroyed &= ~(1 << session_idx); + allocated |= 1 << session_idx; + + } else if (!(allocated & (1 << session_idx))) { DRM_ERROR("Handle already in use!\n"); r = -EINVAL; goto out; @@ -675,16 +680,16 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) 8 * 3 / 2; break; - case 0x04000001: // config extension - case 0x04000002: // pic control - case 0x04000005: // rate control - case 0x04000007: // motion estimation - case 0x04000008: // rdo - case 0x04000009: // vui - case 0x05000002: // auxiliary buffer + case 0x04000001: /* config extension */ + case 0x04000002: /* pic control */ + case 0x04000005: /* rate control */ + case 0x04000007: /* motion estimation */ + case 0x04000008: /* rdo */ + case 0x04000009: /* vui */ + case 0x05000002: /* auxiliary buffer */ break; - case 0x03000001: // encode + case 0x03000001: /* encode */ r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 10, idx + 9, *size, 0); if (r) @@ -696,18 +701,18 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) goto out; break; - case 0x02000001: // destroy - destroyed = true; + case 0x02000001: /* destroy */ + destroyed |= 1 << session_idx; break; - case 0x05000001: // context buffer + case 0x05000001: /* context buffer */ r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 3, idx + 2, *size * 2, 0); if (r) goto out; break; - case 0x05000004: // video bitstream buffer + case 0x05000004: /* video bitstream buffer */ tmp = amdgpu_get_ib_value(p, ib_idx, idx + 4); r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 3, idx + 2, tmp, bs_idx); @@ -715,7 +720,7 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) goto out; break; - case 0x05000005: // feedback buffer + case 0x05000005: /* feedback buffer */ r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 3, idx + 2, 4096, fb_idx); if (r) @@ -737,21 +742,24 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) idx += len / 4; } - if (allocated && !created) { + if (allocated & ~created) { DRM_ERROR("New session without create command!\n"); r = -ENOENT; } out: - if ((!r && destroyed) || (r && allocated)) { - /* - * IB contains a destroy msg or we have allocated an - * handle and got an error, anyway free the handle - */ - for (i = 0; i < AMDGPU_MAX_VCE_HANDLES; ++i) - atomic_cmpxchg(&p->adev->vce.handles[i], handle, 0); + if (!r) { + /* No error, free all destroyed handle slots */ + tmp = destroyed; + } else { + /* Error during parsing, free all allocated handle slots */ + tmp = allocated; } + for (i = 0; i < AMDGPU_MAX_VCE_HANDLES; ++i) + if (tmp & (1 << i)) + atomic_set(&p->adev->vce.handles[i], 0); + return r; } @@ -837,10 +845,10 @@ int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring) * @ring: the engine to test on * */ -int amdgpu_vce_ring_test_ib(struct amdgpu_ring *ring) +int amdgpu_vce_ring_test_ib(struct amdgpu_ring *ring, long timeout) { struct fence *fence = NULL; - int r; + long r; /* skip vce ring1 ib test for now, since it's not reliable */ if (ring == &ring->adev->vce.ring[1]) @@ -848,21 +856,25 @@ int amdgpu_vce_ring_test_ib(struct amdgpu_ring *ring) r = amdgpu_vce_get_create_msg(ring, 1, NULL); if (r) { - DRM_ERROR("amdgpu: failed to get create msg (%d).\n", r); + DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r); goto error; } r = amdgpu_vce_get_destroy_msg(ring, 1, true, &fence); if (r) { - DRM_ERROR("amdgpu: failed to get destroy ib (%d).\n", r); + DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r); goto error; } - r = fence_wait(fence, false); - if (r) { - DRM_ERROR("amdgpu: fence wait failed (%d).\n", r); + r = 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: fence_put(fence); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h index f40cf761c66f..63f83d0d985c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h @@ -39,6 +39,8 @@ void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib, void amdgpu_vce_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq, unsigned flags); int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring); -int amdgpu_vce_ring_test_ib(struct amdgpu_ring *ring); +int amdgpu_vce_ring_test_ib(struct amdgpu_ring *ring, long timeout); +void amdgpu_vce_ring_begin_use(struct amdgpu_ring *ring); +void amdgpu_vce_ring_end_use(struct amdgpu_ring *ring); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 9f36ed30ba11..bf56f1814437 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -25,6 +25,7 @@ * Alex Deucher * Jerome Glisse */ +#include <linux/fence-array.h> #include <drm/drmP.h> #include <drm/amdgpu_drm.h> #include "amdgpu.h" @@ -50,19 +51,22 @@ * SI supports 16. */ -/* Special value that no flush is necessary */ -#define AMDGPU_VM_NO_FLUSH (~0ll) - /* Local structure. Encapsulate some VM table update parameters to reduce * the number of function parameters */ -struct amdgpu_vm_update_params { +struct amdgpu_pte_update_params { + /* amdgpu device we do this update for */ + struct amdgpu_device *adev; /* address where to copy page table entries from */ uint64_t src; - /* DMA addresses to use for mapping */ - dma_addr_t *pages_addr; /* indirect buffer to fill with commands */ struct amdgpu_ib *ib; + /* Function which actually does the update */ + void (*func)(struct amdgpu_pte_update_params *params, uint64_t pe, + uint64_t addr, unsigned count, uint32_t incr, + uint32_t flags); + /* indicate update pt or its shadow */ + bool shadow; }; /** @@ -114,16 +118,26 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, /** * amdgpu_vm_get_bos - add the vm BOs to a duplicates list * + * @adev: amdgpu device pointer * @vm: vm providing the BOs * @duplicates: head of duplicates list * * Add the page directory to the BO duplicates list * for command submission. */ -void amdgpu_vm_get_pt_bos(struct amdgpu_vm *vm, struct list_head *duplicates) +void amdgpu_vm_get_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, + struct list_head *duplicates) { + uint64_t num_evictions; unsigned i; + /* 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; + /* add the vm page table to the list */ for (i = 0; i <= vm->max_pde_used; ++i) { struct amdgpu_bo_list_entry *entry = &vm->page_tables[i].entry; @@ -162,6 +176,13 @@ void amdgpu_vm_move_pt_bos_in_lru(struct amdgpu_device *adev, spin_unlock(&glob->lru_lock); } +static bool amdgpu_vm_is_gpu_reset(struct amdgpu_device *adev, + struct amdgpu_vm_id *id) +{ + return id->current_gpu_reset_count != + atomic_read(&adev->gpu_reset_counter) ? true : false; +} + /** * amdgpu_vm_grab_id - allocate the next free VMID * @@ -174,18 +195,67 @@ void amdgpu_vm_move_pt_bos_in_lru(struct amdgpu_device *adev, */ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring, struct amdgpu_sync *sync, struct fence *fence, - unsigned *vm_id, uint64_t *vm_pd_addr) + struct amdgpu_job *job) { - uint64_t pd_addr = amdgpu_bo_gpu_offset(vm->page_directory); struct amdgpu_device *adev = ring->adev; + uint64_t fence_context = adev->fence_context + ring->idx; struct fence *updates = sync->last_vm_update; - struct amdgpu_vm_id *id; - unsigned i = ring->idx; - int r; + struct amdgpu_vm_id *id, *idle; + struct fence **fences; + unsigned i; + int r = 0; + + fences = kmalloc_array(sizeof(void *), adev->vm_manager.num_ids, + GFP_KERNEL); + if (!fences) + return -ENOMEM; mutex_lock(&adev->vm_manager.lock); + /* Check if we have an idle VMID */ + i = 0; + list_for_each_entry(idle, &adev->vm_manager.ids_lru, list) { + fences[i] = amdgpu_sync_peek_fence(&idle->active, ring); + if (!fences[i]) + break; + ++i; + } + + /* If we can't find a idle VMID to use, wait till one becomes available */ + if (&idle->list == &adev->vm_manager.ids_lru) { + u64 fence_context = adev->vm_manager.fence_context + ring->idx; + unsigned seqno = ++adev->vm_manager.seqno[ring->idx]; + struct fence_array *array; + unsigned j; + + for (j = 0; j < i; ++j) + fence_get(fences[j]); + + array = fence_array_create(i, fences, fence_context, + seqno, true); + if (!array) { + for (j = 0; j < i; ++j) + fence_put(fences[j]); + kfree(fences); + r = -ENOMEM; + goto error; + } + + + r = amdgpu_sync_fence(ring->adev, sync, &array->base); + fence_put(&array->base); + if (r) + goto error; + + mutex_unlock(&adev->vm_manager.lock); + return 0; + + } + kfree(fences); + + job->vm_needs_flush = true; /* Check if we can use a VMID already assigned to this VM */ + i = ring->idx; do { struct fence *flushed; @@ -196,67 +266,52 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring, /* Check all the prerequisites to using this VMID */ if (!id) continue; + if (amdgpu_vm_is_gpu_reset(adev, id)) + continue; if (atomic64_read(&id->owner) != vm->client_id) continue; - if (pd_addr != id->pd_gpu_addr) + if (job->vm_pd_addr != id->pd_gpu_addr) continue; - if (id->last_user != ring && - (!id->last_flush || !fence_is_signaled(id->last_flush))) + if (!id->last_flush) continue; - flushed = id->flushed_updates; - if (updates && (!flushed || fence_is_later(updates, flushed))) + if (id->last_flush->context != fence_context && + !fence_is_signaled(id->last_flush)) continue; - /* Good we can use this VMID */ - if (id->last_user == ring) { - r = amdgpu_sync_fence(ring->adev, sync, - id->first); - if (r) - goto error; - } + flushed = id->flushed_updates; + if (updates && + (!flushed || fence_is_later(updates, flushed))) + continue; - /* And remember this submission as user of the VMID */ + /* Good we can use this VMID. Remember this submission as + * user of the VMID. + */ r = amdgpu_sync_fence(ring->adev, &id->active, fence); if (r) goto error; + id->current_gpu_reset_count = atomic_read(&adev->gpu_reset_counter); list_move_tail(&id->list, &adev->vm_manager.ids_lru); vm->ids[ring->idx] = id; - *vm_id = id - adev->vm_manager.ids; - *vm_pd_addr = AMDGPU_VM_NO_FLUSH; - trace_amdgpu_vm_grab_id(vm, ring->idx, *vm_id, *vm_pd_addr); + job->vm_id = id - adev->vm_manager.ids; + job->vm_needs_flush = false; + trace_amdgpu_vm_grab_id(vm, ring->idx, job); mutex_unlock(&adev->vm_manager.lock); return 0; } while (i != ring->idx); - id = list_first_entry(&adev->vm_manager.ids_lru, - struct amdgpu_vm_id, - list); - - if (!amdgpu_sync_is_idle(&id->active)) { - struct list_head *head = &adev->vm_manager.ids_lru; - struct amdgpu_vm_id *tmp; - - list_for_each_entry_safe(id, tmp, &adev->vm_manager.ids_lru, - list) { - if (amdgpu_sync_is_idle(&id->active)) { - list_move(&id->list, head); - head = &id->list; - } - } - id = list_first_entry(&adev->vm_manager.ids_lru, - struct amdgpu_vm_id, - list); - } + /* Still no ID to use? Then use the idle one found earlier */ + id = idle; - r = amdgpu_sync_cycle_fences(sync, &id->active, fence); + /* Remember this submission as user of the VMID */ + r = amdgpu_sync_fence(ring->adev, &id->active, fence); if (r) goto error; @@ -269,22 +324,46 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring, fence_put(id->flushed_updates); id->flushed_updates = fence_get(updates); - id->pd_gpu_addr = pd_addr; - + id->pd_gpu_addr = job->vm_pd_addr; + id->current_gpu_reset_count = atomic_read(&adev->gpu_reset_counter); list_move_tail(&id->list, &adev->vm_manager.ids_lru); - id->last_user = ring; atomic64_set(&id->owner, vm->client_id); vm->ids[ring->idx] = id; - *vm_id = id - adev->vm_manager.ids; - *vm_pd_addr = pd_addr; - trace_amdgpu_vm_grab_id(vm, ring->idx, *vm_id, *vm_pd_addr); + job->vm_id = id - adev->vm_manager.ids; + trace_amdgpu_vm_grab_id(vm, ring->idx, job); error: mutex_unlock(&adev->vm_manager.lock); return r; } +static bool amdgpu_vm_ring_has_compute_vm_bug(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + const struct amdgpu_ip_block_version *ip_block; + + if (ring->type != AMDGPU_RING_TYPE_COMPUTE) + /* only compute rings */ + return false; + + ip_block = amdgpu_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX); + if (!ip_block) + return false; + + if (ip_block->major <= 7) { + /* gfx7 has no workaround */ + return true; + } else if (ip_block->major == 8) { + if (adev->gfx.mec_fw_version >= 673) + /* gfx8 is fixed in MEC firmware 673 */ + return false; + else + return true; + } + return false; +} + /** * amdgpu_vm_flush - hardware flush the vm * @@ -294,59 +373,52 @@ error: * * Emit a VM flush when it is necessary. */ -int amdgpu_vm_flush(struct amdgpu_ring *ring, - unsigned vm_id, uint64_t pd_addr, - uint32_t gds_base, uint32_t gds_size, - uint32_t gws_base, uint32_t gws_size, - uint32_t oa_base, uint32_t oa_size) +int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job) { struct amdgpu_device *adev = ring->adev; - struct amdgpu_vm_id *id = &adev->vm_manager.ids[vm_id]; + struct amdgpu_vm_id *id = &adev->vm_manager.ids[job->vm_id]; bool gds_switch_needed = ring->funcs->emit_gds_switch && ( - id->gds_base != gds_base || - id->gds_size != gds_size || - id->gws_base != gws_base || - id->gws_size != gws_size || - id->oa_base != oa_base || - id->oa_size != oa_size); + id->gds_base != job->gds_base || + id->gds_size != job->gds_size || + id->gws_base != job->gws_base || + id->gws_size != job->gws_size || + id->oa_base != job->oa_base || + id->oa_size != job->oa_size); int r; if (ring->funcs->emit_pipeline_sync && ( - pd_addr != AMDGPU_VM_NO_FLUSH || gds_switch_needed || - ring->type == AMDGPU_RING_TYPE_COMPUTE)) + job->vm_needs_flush || gds_switch_needed || + amdgpu_vm_ring_has_compute_vm_bug(ring))) amdgpu_ring_emit_pipeline_sync(ring); - if (ring->funcs->emit_vm_flush && - pd_addr != AMDGPU_VM_NO_FLUSH) { + if (ring->funcs->emit_vm_flush && (job->vm_needs_flush || + amdgpu_vm_is_gpu_reset(adev, id))) { struct fence *fence; - trace_amdgpu_vm_flush(pd_addr, ring->idx, vm_id); - amdgpu_ring_emit_vm_flush(ring, vm_id, pd_addr); + trace_amdgpu_vm_flush(job->vm_pd_addr, ring->idx, job->vm_id); + amdgpu_ring_emit_vm_flush(ring, job->vm_id, job->vm_pd_addr); + + r = amdgpu_fence_emit(ring, &fence); + if (r) + return r; mutex_lock(&adev->vm_manager.lock); - if ((id->pd_gpu_addr == pd_addr) && (id->last_user == ring)) { - r = amdgpu_fence_emit(ring, &fence); - if (r) { - mutex_unlock(&adev->vm_manager.lock); - return r; - } - fence_put(id->last_flush); - id->last_flush = fence; - } + fence_put(id->last_flush); + id->last_flush = fence; mutex_unlock(&adev->vm_manager.lock); } if (gds_switch_needed) { - id->gds_base = gds_base; - id->gds_size = gds_size; - id->gws_base = gws_base; - id->gws_size = gws_size; - id->oa_base = oa_base; - id->oa_size = oa_size; - amdgpu_ring_emit_gds_switch(ring, vm_id, - gds_base, gds_size, - gws_base, gws_size, - oa_base, oa_size); + id->gds_base = job->gds_base; + id->gds_size = job->gds_size; + id->gws_base = job->gws_base; + id->gws_size = job->gws_size; + id->oa_base = job->oa_base; + id->oa_size = job->oa_size; + amdgpu_ring_emit_gds_switch(ring, job->vm_id, + job->gds_base, job->gds_size, + job->gws_base, job->gws_size, + job->oa_base, job->oa_size); } return 0; @@ -398,10 +470,9 @@ struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm, } /** - * amdgpu_vm_update_pages - helper to call the right asic function + * amdgpu_vm_do_set_ptes - helper to call the right asic function * - * @adev: amdgpu_device pointer - * @vm_update_params: see amdgpu_vm_update_params definition + * @params: see amdgpu_pte_update_params definition * @pe: addr of the page entry * @addr: dst addr to write into pe * @count: number of page entries to update @@ -411,35 +482,47 @@ struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm, * Traces the parameters and calls the right asic functions * to setup the page table using the DMA. */ -static void amdgpu_vm_update_pages(struct amdgpu_device *adev, - struct amdgpu_vm_update_params - *vm_update_params, - uint64_t pe, uint64_t addr, - unsigned count, uint32_t incr, - uint32_t flags) +static void amdgpu_vm_do_set_ptes(struct amdgpu_pte_update_params *params, + uint64_t pe, uint64_t addr, + unsigned count, uint32_t incr, + uint32_t flags) { trace_amdgpu_vm_set_page(pe, addr, count, incr, flags); - if (vm_update_params->src) { - amdgpu_vm_copy_pte(adev, vm_update_params->ib, - pe, (vm_update_params->src + (addr >> 12) * 8), count); - - } else if (vm_update_params->pages_addr) { - amdgpu_vm_write_pte(adev, vm_update_params->ib, - vm_update_params->pages_addr, - pe, addr, count, incr, flags); - - } else if (count < 3) { - amdgpu_vm_write_pte(adev, vm_update_params->ib, NULL, pe, addr, - count, incr, flags); + if (count < 3) { + amdgpu_vm_write_pte(params->adev, params->ib, pe, + addr | flags, count, incr); } else { - amdgpu_vm_set_pte_pde(adev, vm_update_params->ib, pe, addr, + amdgpu_vm_set_pte_pde(params->adev, params->ib, pe, addr, count, incr, flags); } } /** + * amdgpu_vm_do_copy_ptes - copy the PTEs from the GART + * + * @params: see amdgpu_pte_update_params definition + * @pe: addr of the page entry + * @addr: dst addr to write into pe + * @count: number of page entries to update + * @incr: increase next addr by incr bytes + * @flags: hw access flags + * + * Traces the parameters and calls the DMA function to copy the PTEs. + */ +static void amdgpu_vm_do_copy_ptes(struct amdgpu_pte_update_params *params, + uint64_t pe, uint64_t addr, + unsigned count, uint32_t incr, + uint32_t flags) +{ + trace_amdgpu_vm_set_page(pe, addr, count, incr, flags); + + amdgpu_vm_copy_pte(params->adev, params->ib, pe, + (params->src + (addr >> 12) * 8), count); +} + +/** * amdgpu_vm_clear_bo - initially clear the page dir/table * * @adev: amdgpu_device pointer @@ -454,12 +537,11 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, struct amdgpu_ring *ring; struct fence *fence = NULL; struct amdgpu_job *job; - struct amdgpu_vm_update_params vm_update_params; + struct amdgpu_pte_update_params params; unsigned entries; uint64_t addr; int r; - memset(&vm_update_params, 0, sizeof(vm_update_params)); ring = container_of(vm->entity.sched, struct amdgpu_ring, sched); r = reservation_object_reserve_shared(bo->tbo.resv); @@ -477,9 +559,10 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, if (r) goto error; - vm_update_params.ib = &job->ibs[0]; - amdgpu_vm_update_pages(adev, &vm_update_params, addr, 0, entries, - 0, 0); + memset(¶ms, 0, sizeof(params)); + params.adev = adev; + params.ib = &job->ibs[0]; + amdgpu_vm_do_set_ptes(¶ms, addr, 0, entries, 0, 0); amdgpu_ring_pad_ib(ring, &job->ibs[0]); WARN_ON(job->ibs[0].length_dw > 64); @@ -508,55 +591,41 @@ error: * Look up the physical address of the page that the pte resolves * to and return the pointer for the page table entry. */ -uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr) +static uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr) { uint64_t result; - if (pages_addr) { - /* page table offset */ - result = pages_addr[addr >> PAGE_SHIFT]; - - /* in case cpu page size != gpu page size*/ - result |= addr & (~PAGE_MASK); + /* page table offset */ + result = pages_addr[addr >> PAGE_SHIFT]; - } else { - /* No mapping required */ - result = addr; - } + /* in case cpu page size != gpu page size*/ + result |= addr & (~PAGE_MASK); result &= 0xFFFFFFFFFFFFF000ULL; return result; } -/** - * amdgpu_vm_update_pdes - make sure that page directory is valid - * - * @adev: amdgpu_device pointer - * @vm: requested vm - * @start: start of GPU address range - * @end: end of GPU address range - * - * Allocates new page tables if necessary - * and updates the page directory. - * Returns 0 for success, error for failure. - */ -int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, - struct amdgpu_vm *vm) +static int amdgpu_vm_update_pd_or_shadow(struct amdgpu_device *adev, + struct amdgpu_vm *vm, + bool shadow) { struct amdgpu_ring *ring; - struct amdgpu_bo *pd = vm->page_directory; - uint64_t pd_addr = amdgpu_bo_gpu_offset(pd); + struct amdgpu_bo *pd = shadow ? vm->page_directory->shadow : + vm->page_directory; + uint64_t pd_addr; uint32_t incr = AMDGPU_VM_PTE_COUNT * 8; uint64_t last_pde = ~0, last_pt = ~0; unsigned count = 0, pt_idx, ndw; struct amdgpu_job *job; - struct amdgpu_vm_update_params vm_update_params; + struct amdgpu_pte_update_params params; struct fence *fence = NULL; int r; - memset(&vm_update_params, 0, sizeof(vm_update_params)); + if (!pd) + return 0; + pd_addr = amdgpu_bo_gpu_offset(pd); ring = container_of(vm->entity.sched, struct amdgpu_ring, sched); /* padding, etc. */ @@ -569,7 +638,9 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, if (r) return r; - vm_update_params.ib = &job->ibs[0]; + memset(¶ms, 0, sizeof(params)); + params.adev = adev; + params.ib = &job->ibs[0]; /* walk over the address space and update the page directory */ for (pt_idx = 0; pt_idx <= vm->max_pde_used; ++pt_idx) { @@ -580,19 +651,25 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, continue; pt = amdgpu_bo_gpu_offset(bo); - if (vm->page_tables[pt_idx].addr == pt) - continue; - vm->page_tables[pt_idx].addr = pt; + if (!shadow) { + if (vm->page_tables[pt_idx].addr == pt) + continue; + vm->page_tables[pt_idx].addr = pt; + } else { + if (vm->page_tables[pt_idx].shadow_addr == pt) + continue; + vm->page_tables[pt_idx].shadow_addr = pt; + } pde = pd_addr + pt_idx * 8; if (((last_pde + 8 * count) != pde) || - ((last_pt + incr * count) != pt)) { + ((last_pt + incr * count) != pt) || + (count == AMDGPU_VM_MAX_UPDATE_SIZE)) { if (count) { - amdgpu_vm_update_pages(adev, &vm_update_params, - last_pde, last_pt, - count, incr, - AMDGPU_PTE_VALID); + amdgpu_vm_do_set_ptes(¶ms, last_pde, + last_pt, count, incr, + AMDGPU_PTE_VALID); } count = 1; @@ -604,15 +681,14 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, } if (count) - amdgpu_vm_update_pages(adev, &vm_update_params, - last_pde, last_pt, - count, incr, AMDGPU_PTE_VALID); + amdgpu_vm_do_set_ptes(¶ms, last_pde, last_pt, + count, incr, AMDGPU_PTE_VALID); - if (vm_update_params.ib->length_dw != 0) { - amdgpu_ring_pad_ib(ring, vm_update_params.ib); + if (params.ib->length_dw != 0) { + amdgpu_ring_pad_ib(ring, params.ib); amdgpu_sync_resv(adev, &job->sync, pd->tbo.resv, AMDGPU_FENCE_OWNER_VM); - WARN_ON(vm_update_params.ib->length_dw > ndw); + WARN_ON(params.ib->length_dw > ndw); r = amdgpu_job_submit(job, ring, &vm->entity, AMDGPU_FENCE_OWNER_VM, &fence); if (r) @@ -634,21 +710,135 @@ error_free: return r; } +/* + * amdgpu_vm_update_pdes - make sure that page directory is valid + * + * @adev: amdgpu_device pointer + * @vm: requested vm + * @start: start of GPU address range + * @end: end of GPU address range + * + * Allocates new page tables if necessary + * and updates the page directory. + * Returns 0 for success, error for failure. + */ +int amdgpu_vm_update_page_directory(struct amdgpu_device *adev, + struct amdgpu_vm *vm) +{ + int r; + + r = amdgpu_vm_update_pd_or_shadow(adev, vm, true); + if (r) + return r; + return amdgpu_vm_update_pd_or_shadow(adev, vm, false); +} + /** + * amdgpu_vm_update_ptes - make sure that page tables are valid + * + * @params: see amdgpu_pte_update_params definition + * @vm: requested vm + * @start: start of GPU address range + * @end: end of GPU address range + * @dst: destination address to map to, the next dst inside the function + * @flags: mapping flags + * + * Update the page tables in the range @start - @end. + */ +static void amdgpu_vm_update_ptes(struct amdgpu_pte_update_params *params, + struct amdgpu_vm *vm, + uint64_t start, uint64_t end, + uint64_t dst, uint32_t flags) +{ + const uint64_t mask = AMDGPU_VM_PTE_COUNT - 1; + + uint64_t cur_pe_start, cur_nptes, cur_dst; + uint64_t addr; /* next GPU address to be updated */ + uint64_t pt_idx; + struct amdgpu_bo *pt; + unsigned nptes; /* next number of ptes to be updated */ + uint64_t next_pe_start; + + /* initialize the variables */ + addr = start; + pt_idx = addr >> amdgpu_vm_block_size; + pt = vm->page_tables[pt_idx].entry.robj; + if (params->shadow) { + if (!pt->shadow) + return; + pt = vm->page_tables[pt_idx].entry.robj->shadow; + } + if ((addr & ~mask) == (end & ~mask)) + nptes = end - addr; + else + nptes = AMDGPU_VM_PTE_COUNT - (addr & mask); + + cur_pe_start = amdgpu_bo_gpu_offset(pt); + cur_pe_start += (addr & mask) * 8; + cur_nptes = nptes; + cur_dst = dst; + + /* for next ptb*/ + addr += nptes; + dst += nptes * AMDGPU_GPU_PAGE_SIZE; + + /* walk over the address space and update the page tables */ + while (addr < end) { + pt_idx = addr >> amdgpu_vm_block_size; + pt = vm->page_tables[pt_idx].entry.robj; + if (params->shadow) { + if (!pt->shadow) + return; + pt = vm->page_tables[pt_idx].entry.robj->shadow; + } + + if ((addr & ~mask) == (end & ~mask)) + nptes = end - addr; + else + nptes = AMDGPU_VM_PTE_COUNT - (addr & mask); + + next_pe_start = amdgpu_bo_gpu_offset(pt); + next_pe_start += (addr & mask) * 8; + + if ((cur_pe_start + 8 * cur_nptes) == next_pe_start && + ((cur_nptes + nptes) <= AMDGPU_VM_MAX_UPDATE_SIZE)) { + /* The next ptb is consecutive to current ptb. + * Don't call the update function now. + * Will update two ptbs together in future. + */ + cur_nptes += nptes; + } else { + params->func(params, cur_pe_start, cur_dst, cur_nptes, + AMDGPU_GPU_PAGE_SIZE, flags); + + cur_pe_start = next_pe_start; + cur_nptes = nptes; + cur_dst = dst; + } + + /* for next ptb*/ + addr += nptes; + dst += nptes * AMDGPU_GPU_PAGE_SIZE; + } + + params->func(params, cur_pe_start, cur_dst, cur_nptes, + AMDGPU_GPU_PAGE_SIZE, flags); +} + +/* * amdgpu_vm_frag_ptes - add fragment information to PTEs * - * @adev: amdgpu_device pointer - * @vm_update_params: see amdgpu_vm_update_params definition - * @pe_start: first PTE to handle - * @pe_end: last PTE to handle - * @addr: addr those PTEs should point to + * @params: see amdgpu_pte_update_params definition + * @vm: requested vm + * @start: first PTE to handle + * @end: last PTE to handle + * @dst: addr those PTEs should point to * @flags: hw mapping flags */ -static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev, - struct amdgpu_vm_update_params - *vm_update_params, - uint64_t pe_start, uint64_t pe_end, - uint64_t addr, uint32_t flags) +static void amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params *params, + struct amdgpu_vm *vm, + uint64_t start, uint64_t end, + uint64_t dst, uint32_t flags) { /** * The MC L1 TLB supports variable sized pages, based on a fragment @@ -669,117 +859,48 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev, * allocation size to the fragment size. */ - /* SI and newer are optimized for 64KB */ - uint64_t frag_flags = AMDGPU_PTE_FRAG_64KB; - uint64_t frag_align = 0x80; + const uint64_t frag_align = 1 << AMDGPU_LOG2_PAGES_PER_FRAG; - uint64_t frag_start = ALIGN(pe_start, frag_align); - uint64_t frag_end = pe_end & ~(frag_align - 1); + uint64_t frag_start = ALIGN(start, frag_align); + uint64_t frag_end = end & ~(frag_align - 1); - unsigned count; - - /* Abort early if there isn't anything to do */ - if (pe_start == pe_end) - return; + uint32_t frag; /* system pages are non continuously */ - if (vm_update_params->src || vm_update_params->pages_addr || - !(flags & AMDGPU_PTE_VALID) || (frag_start >= frag_end)) { + if (params->src || !(flags & AMDGPU_PTE_VALID) || + (frag_start >= frag_end)) { - count = (pe_end - pe_start) / 8; - amdgpu_vm_update_pages(adev, vm_update_params, pe_start, - addr, count, AMDGPU_GPU_PAGE_SIZE, - flags); + amdgpu_vm_update_ptes(params, vm, start, end, dst, flags); return; } + /* use more than 64KB fragment size if possible */ + frag = lower_32_bits(frag_start | frag_end); + frag = likely(frag) ? __ffs(frag) : 31; + /* handle the 4K area at the beginning */ - if (pe_start != frag_start) { - count = (frag_start - pe_start) / 8; - amdgpu_vm_update_pages(adev, vm_update_params, pe_start, addr, - count, AMDGPU_GPU_PAGE_SIZE, flags); - addr += AMDGPU_GPU_PAGE_SIZE * count; + if (start != frag_start) { + amdgpu_vm_update_ptes(params, vm, start, frag_start, + dst, flags); + dst += (frag_start - start) * AMDGPU_GPU_PAGE_SIZE; } /* handle the area in the middle */ - count = (frag_end - frag_start) / 8; - amdgpu_vm_update_pages(adev, vm_update_params, frag_start, addr, count, - AMDGPU_GPU_PAGE_SIZE, flags | frag_flags); + amdgpu_vm_update_ptes(params, vm, frag_start, frag_end, dst, + flags | AMDGPU_PTE_FRAG(frag)); /* handle the 4K area at the end */ - if (frag_end != pe_end) { - addr += AMDGPU_GPU_PAGE_SIZE * count; - count = (pe_end - frag_end) / 8; - amdgpu_vm_update_pages(adev, vm_update_params, frag_end, addr, - count, AMDGPU_GPU_PAGE_SIZE, flags); - } -} - -/** - * amdgpu_vm_update_ptes - make sure that page tables are valid - * - * @adev: amdgpu_device pointer - * @vm_update_params: see amdgpu_vm_update_params definition - * @vm: requested vm - * @start: start of GPU address range - * @end: end of GPU address range - * @dst: destination address to map to - * @flags: mapping flags - * - * Update the page tables in the range @start - @end. - */ -static void amdgpu_vm_update_ptes(struct amdgpu_device *adev, - struct amdgpu_vm_update_params - *vm_update_params, - struct amdgpu_vm *vm, - uint64_t start, uint64_t end, - uint64_t dst, uint32_t flags) -{ - const uint64_t mask = AMDGPU_VM_PTE_COUNT - 1; - - uint64_t last_pe_start = ~0, last_pe_end = ~0, last_dst = ~0; - uint64_t addr; - - /* walk over the address space and update the page tables */ - for (addr = start; addr < end; ) { - uint64_t pt_idx = addr >> amdgpu_vm_block_size; - struct amdgpu_bo *pt = vm->page_tables[pt_idx].entry.robj; - unsigned nptes; - uint64_t pe_start; - - if ((addr & ~mask) == (end & ~mask)) - nptes = end - addr; - else - nptes = AMDGPU_VM_PTE_COUNT - (addr & mask); - - pe_start = amdgpu_bo_gpu_offset(pt); - pe_start += (addr & mask) * 8; - - if (last_pe_end != pe_start) { - - amdgpu_vm_frag_ptes(adev, vm_update_params, - last_pe_start, last_pe_end, - last_dst, flags); - - last_pe_start = pe_start; - last_pe_end = pe_start + 8 * nptes; - last_dst = dst; - } else { - last_pe_end += 8 * nptes; - } - - addr += nptes; - dst += nptes * AMDGPU_GPU_PAGE_SIZE; + if (frag_end != end) { + dst += (frag_end - frag_start) * AMDGPU_GPU_PAGE_SIZE; + amdgpu_vm_update_ptes(params, vm, frag_end, end, dst, flags); } - - amdgpu_vm_frag_ptes(adev, vm_update_params, last_pe_start, - last_pe_end, last_dst, flags); } /** * amdgpu_vm_bo_update_mapping - update a mapping in the vm page table * * @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 @@ -793,6 +914,7 @@ static void amdgpu_vm_update_ptes(struct amdgpu_device *adev, * Returns 0 for success, -EINVAL for failure. */ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, + struct fence *exclusive, uint64_t src, dma_addr_t *pages_addr, struct amdgpu_vm *vm, @@ -804,14 +926,19 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, void *owner = AMDGPU_FENCE_OWNER_VM; unsigned nptes, ncmds, ndw; struct amdgpu_job *job; - struct amdgpu_vm_update_params vm_update_params; + struct amdgpu_pte_update_params params; struct fence *f = NULL; int r; + memset(¶ms, 0, sizeof(params)); + params.adev = adev; + params.src = src; + ring = container_of(vm->entity.sched, struct amdgpu_ring, sched); - memset(&vm_update_params, 0, sizeof(vm_update_params)); - vm_update_params.src = src; - vm_update_params.pages_addr = pages_addr; + + memset(¶ms, 0, sizeof(params)); + params.adev = adev; + params.src = src; /* sync to everything on unmapping */ if (!(flags & AMDGPU_PTE_VALID)) @@ -828,30 +955,56 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, /* padding, etc. */ ndw = 64; - if (vm_update_params.src) { + if (src) { /* only copy commands needed */ ndw += ncmds * 7; - } else if (vm_update_params.pages_addr) { - /* header for write data commands */ - ndw += ncmds * 4; + params.func = amdgpu_vm_do_copy_ptes; + + } else if (pages_addr) { + /* copy commands needed */ + ndw += ncmds * 7; - /* body of write data command */ + /* and also PTEs */ ndw += nptes * 2; + params.func = amdgpu_vm_do_copy_ptes; + } else { /* set page commands needed */ ndw += ncmds * 10; /* two extra commands for begin/end of fragment */ ndw += 2 * 10; + + params.func = amdgpu_vm_do_set_ptes; } r = amdgpu_job_alloc_with_ib(adev, ndw * 4, &job); if (r) return r; - vm_update_params.ib = &job->ibs[0]; + params.ib = &job->ibs[0]; + + if (!src && pages_addr) { + uint64_t *pte; + unsigned i; + + /* Put the PTEs at the end of the IB. */ + i = ndw - nptes * 2; + pte= (uint64_t *)&(job->ibs->ptr[i]); + params.src = job->ibs->gpu_addr + i * 4; + + for (i = 0; i < nptes; ++i) { + pte[i] = amdgpu_vm_map_gart(pages_addr, addr + i * + AMDGPU_GPU_PAGE_SIZE); + pte[i] |= flags; + } + } + + r = amdgpu_sync_fence(adev, &job->sync, exclusive); + if (r) + goto error_free; r = amdgpu_sync_resv(adev, &job->sync, vm->page_directory->tbo.resv, owner); @@ -862,11 +1015,13 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (r) goto error_free; - amdgpu_vm_update_ptes(adev, &vm_update_params, vm, start, - last + 1, addr, flags); + params.shadow = true; + amdgpu_vm_frag_ptes(¶ms, vm, start, last + 1, addr, flags); + params.shadow = false; + amdgpu_vm_frag_ptes(¶ms, vm, start, last + 1, addr, flags); - amdgpu_ring_pad_ib(ring, vm_update_params.ib); - WARN_ON(vm_update_params.ib->length_dw > ndw); + amdgpu_ring_pad_ib(ring, params.ib); + WARN_ON(params.ib->length_dw > ndw); r = amdgpu_job_submit(job, ring, &vm->entity, AMDGPU_FENCE_OWNER_VM, &f); if (r) @@ -889,6 +1044,7 @@ error_free: * amdgpu_vm_bo_split_mapping - split a mapping into smaller chunks * * @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 @@ -902,6 +1058,7 @@ error_free: * Returns 0 for success, -EINVAL for failure. */ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, + struct fence *exclusive, uint32_t gtt_flags, dma_addr_t *pages_addr, struct amdgpu_vm *vm, @@ -932,7 +1089,8 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, addr += mapping->offset; if (!pages_addr || src) - return amdgpu_vm_bo_update_mapping(adev, src, pages_addr, vm, + return amdgpu_vm_bo_update_mapping(adev, exclusive, + src, pages_addr, vm, start, mapping->it.last, flags, addr, fence); @@ -940,7 +1098,8 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, uint64_t last; last = min((uint64_t)mapping->it.last, start + max_size - 1); - r = amdgpu_vm_bo_update_mapping(adev, src, pages_addr, vm, + r = amdgpu_vm_bo_update_mapping(adev, exclusive, + src, pages_addr, vm, start, last, flags, addr, fence); if (r) @@ -958,27 +1117,32 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, * * @adev: amdgpu_device pointer * @bo_va: requested BO and VM object - * @mem: ttm mem + * @clear: if true clear the entries * * Fill in the page table entries for @bo_va. * Returns 0 for success, -EINVAL for failure. - * - * Object have to be reserved and mutex must be locked! */ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, - struct ttm_mem_reg *mem) + bool clear) { struct amdgpu_vm *vm = bo_va->vm; struct amdgpu_bo_va_mapping *mapping; dma_addr_t *pages_addr = NULL; uint32_t gtt_flags, flags; + struct ttm_mem_reg *mem; + struct fence *exclusive; uint64_t addr; int r; - if (mem) { + if (clear) { + mem = NULL; + addr = 0; + exclusive = NULL; + } else { struct ttm_dma_tt *ttm; + mem = &bo_va->bo->tbo.mem; addr = (u64)mem->start << PAGE_SHIFT; switch (mem->mem_type) { case TTM_PL_TT: @@ -994,8 +1158,8 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, default: break; } - } else { - addr = 0; + + exclusive = reservation_object_get_excl(bo_va->bo->tbo.resv); } flags = amdgpu_ttm_tt_pte_flags(adev, bo_va->bo->tbo.ttm, mem); @@ -1007,7 +1171,8 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, spin_unlock(&vm->status_lock); list_for_each_entry(mapping, &bo_va->invalids, list) { - r = amdgpu_vm_bo_split_mapping(adev, gtt_flags, pages_addr, vm, + r = amdgpu_vm_bo_split_mapping(adev, exclusive, + gtt_flags, pages_addr, vm, mapping, flags, addr, &bo_va->last_pt_update); if (r) @@ -1025,7 +1190,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, spin_lock(&vm->status_lock); list_splice_init(&bo_va->invalids, &bo_va->valids); list_del_init(&bo_va->vm_status); - if (!mem) + if (clear) list_add(&bo_va->vm_status, &vm->cleared); spin_unlock(&vm->status_lock); @@ -1054,7 +1219,7 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, struct amdgpu_bo_va_mapping, list); list_del(&mapping->list); - r = amdgpu_vm_bo_split_mapping(adev, 0, NULL, vm, mapping, + r = amdgpu_vm_bo_split_mapping(adev, NULL, 0, NULL, vm, mapping, 0, 0, NULL); kfree(mapping); if (r) @@ -1088,7 +1253,7 @@ int amdgpu_vm_clear_invalids(struct amdgpu_device *adev, struct amdgpu_bo_va, vm_status); spin_unlock(&vm->status_lock); - r = amdgpu_vm_bo_update(adev, bo_va, NULL); + r = amdgpu_vm_bo_update(adev, bo_va, true); if (r) return r; @@ -1233,7 +1398,8 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev, r = amdgpu_bo_create(adev, AMDGPU_VM_PTE_COUNT * 8, AMDGPU_GPU_PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_NO_CPU_ACCESS, + AMDGPU_GEM_CREATE_NO_CPU_ACCESS | + AMDGPU_GEM_CREATE_SHADOW, NULL, resv, &pt); if (r) goto error_free; @@ -1432,7 +1598,8 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm) r = amdgpu_bo_create(adev, pd_size, align, true, AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_NO_CPU_ACCESS, + AMDGPU_GEM_CREATE_NO_CPU_ACCESS | + AMDGPU_GEM_CREATE_SHADOW, NULL, NULL, &vm->page_directory); if (r) goto error_free_sched_entity; @@ -1445,6 +1612,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm) amdgpu_bo_unreserve(vm->page_directory); if (r) goto error_free_page_directory; + vm->last_eviction_counter = atomic64_read(&adev->num_evictions); return 0; @@ -1487,10 +1655,16 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) kfree(mapping); } - for (i = 0; i < amdgpu_vm_num_pdes(adev); i++) + for (i = 0; i < amdgpu_vm_num_pdes(adev); i++) { + if (vm->page_tables[i].entry.robj && + vm->page_tables[i].entry.robj->shadow) + amdgpu_bo_unref(&vm->page_tables[i].entry.robj->shadow); amdgpu_bo_unref(&vm->page_tables[i].entry.robj); + } drm_free_large(vm->page_tables); + if (vm->page_directory->shadow) + amdgpu_bo_unref(&vm->page_directory->shadow); amdgpu_bo_unref(&vm->page_directory); fence_put(vm->page_directory_fence); } @@ -1516,6 +1690,10 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev) &adev->vm_manager.ids_lru); } + adev->vm_manager.fence_context = fence_context_alloc(AMDGPU_MAX_RINGS); + for (i = 0; i < AMDGPU_MAX_RINGS; ++i) + adev->vm_manager.seqno[i] = 0; + atomic_set(&adev->vm_manager.vm_pte_next_ring, 0); atomic64_set(&adev->vm_manager.client_counter, 0); } diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_dp.c b/drivers/gpu/drm/amd/amdgpu/atombios_dp.c index 7f85c2c1d681..f81068ba4cc6 100644 --- a/drivers/gpu/drm/amd/amdgpu/atombios_dp.c +++ b/drivers/gpu/drm/amd/amdgpu/atombios_dp.c @@ -88,7 +88,6 @@ static int amdgpu_atombios_dp_process_aux_ch(struct amdgpu_i2c_chan *chan, /* timeout */ if (args.v2.ucReplyStatus == 1) { - DRM_DEBUG_KMS("dp_aux_ch timeout\n"); r = -ETIMEDOUT; goto done; } @@ -339,22 +338,21 @@ int amdgpu_atombios_dp_get_dpcd(struct amdgpu_connector *amdgpu_connector) { struct amdgpu_connector_atom_dig *dig_connector = amdgpu_connector->con_priv; u8 msg[DP_DPCD_SIZE]; - int ret, i; + int ret; - for (i = 0; i < 7; i++) { - ret = drm_dp_dpcd_read(&amdgpu_connector->ddc_bus->aux, DP_DPCD_REV, msg, - DP_DPCD_SIZE); - if (ret == DP_DPCD_SIZE) { - memcpy(dig_connector->dpcd, msg, DP_DPCD_SIZE); + ret = drm_dp_dpcd_read(&amdgpu_connector->ddc_bus->aux, DP_DPCD_REV, + msg, DP_DPCD_SIZE); + if (ret == DP_DPCD_SIZE) { + memcpy(dig_connector->dpcd, msg, DP_DPCD_SIZE); - DRM_DEBUG_KMS("DPCD: %*ph\n", (int)sizeof(dig_connector->dpcd), - dig_connector->dpcd); + DRM_DEBUG_KMS("DPCD: %*ph\n", (int)sizeof(dig_connector->dpcd), + dig_connector->dpcd); - amdgpu_atombios_dp_probe_oui(amdgpu_connector); + amdgpu_atombios_dp_probe_oui(amdgpu_connector); - return 0; - } + return 0; } + dig_connector->dpcd[0] = 0; return -EINVAL; } diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c index 48b6bd671cda..c32eca26155c 100644 --- a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c +++ b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c @@ -98,6 +98,7 @@ amdgpu_atombios_encoder_set_backlight_level(struct amdgpu_encoder *amdgpu_encode case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2: + case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3: if (dig->backlight_level == 0) amdgpu_atombios_encoder_setup_dig_transmitter(encoder, ATOM_TRANSMITTER_ACTION_LCD_BLOFF, 0, 0); diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c index 5ec1f1e9c983..a0d63a293bb0 100644 --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c @@ -50,7 +50,9 @@ #include "gmc/gmc_7_1_sh_mask.h" MODULE_FIRMWARE("radeon/bonaire_smc.bin"); +MODULE_FIRMWARE("radeon/bonaire_k_smc.bin"); MODULE_FIRMWARE("radeon/hawaii_smc.bin"); +MODULE_FIRMWARE("radeon/hawaii_k_smc.bin"); #define MC_CG_ARB_FREQ_F0 0x0a #define MC_CG_ARB_FREQ_F1 0x0b @@ -84,12 +86,14 @@ static const struct ci_pt_defaults defaults_bonaire_xt = { 0x17C, 0x172, 0x180, 0x1BC, 0x1B3, 0x1BD, 0x206, 0x200, 0x203, 0x25D, 0x25A, 0x255, 0x2C3, 0x2C5, 0x2B4 } }; +#if 0 static const struct ci_pt_defaults defaults_bonaire_pro = { 1, 0xF, 0xFD, 0x19, 5, 45, 0, 0x65062, { 0x8C, 0x23F, 0x244, 0xA6, 0x83, 0x85, 0x86, 0x86, 0x83, 0xDB, 0xDB, 0xDA, 0x67, 0x60, 0x5F }, { 0x187, 0x193, 0x193, 0x1C7, 0x1D1, 0x1D1, 0x210, 0x219, 0x219, 0x266, 0x26C, 0x26C, 0x2C9, 0x2CB, 0x2CB } }; +#endif static const struct ci_pt_defaults defaults_saturn_xt = { @@ -98,12 +102,14 @@ static const struct ci_pt_defaults defaults_saturn_xt = { 0x187, 0x187, 0x187, 0x1C7, 0x1C7, 0x1C7, 0x210, 0x210, 0x210, 0x266, 0x266, 0x266, 0x2C9, 0x2C9, 0x2C9 } }; +#if 0 static const struct ci_pt_defaults defaults_saturn_pro = { 1, 0xF, 0xFD, 0x19, 5, 55, 0, 0x30000, { 0x96, 0x21D, 0x23B, 0xA1, 0x85, 0x87, 0x83, 0x84, 0x81, 0xE6, 0xE6, 0xE6, 0x71, 0x6A, 0x6A }, { 0x193, 0x19E, 0x19E, 0x1D2, 0x1DC, 0x1DC, 0x21A, 0x223, 0x223, 0x26E, 0x27E, 0x274, 0x2CF, 0x2D2, 0x2D2 } }; +#endif static const struct ci_pt_config_reg didt_config_ci[] = { @@ -736,19 +742,19 @@ static int ci_enable_didt(struct amdgpu_device *adev, bool enable) if (pi->caps_sq_ramping || pi->caps_db_ramping || pi->caps_td_ramping || pi->caps_tcp_ramping) { - gfx_v7_0_enter_rlc_safe_mode(adev); + adev->gfx.rlc.funcs->enter_safe_mode(adev); if (enable) { ret = ci_program_pt_config_registers(adev, didt_config_ci); if (ret) { - gfx_v7_0_exit_rlc_safe_mode(adev); + adev->gfx.rlc.funcs->exit_safe_mode(adev); return ret; } } ci_do_enable_didt(adev, enable); - gfx_v7_0_exit_rlc_safe_mode(adev); + adev->gfx.rlc.funcs->exit_safe_mode(adev); } return 0; @@ -3030,7 +3036,7 @@ static int ci_populate_single_memory_level(struct amdgpu_device *adev, if (pi->mclk_stutter_mode_threshold && (memory_clock <= pi->mclk_stutter_mode_threshold) && - (pi->uvd_enabled == false) && + (!pi->uvd_enabled) && (RREG32(mmDPG_PIPE_STUTTER_CONTROL) & DPG_PIPE_STUTTER_CONTROL__STUTTER_ENABLE_MASK) && (adev->pm.dpm.new_active_crtc_count <= 2)) memory_level->StutterEnable = true; @@ -3636,6 +3642,10 @@ static int ci_setup_default_dpm_tables(struct amdgpu_device *adev) ci_setup_default_pcie_tables(adev); + /* save a copy of the default DPM table */ + memcpy(&(pi->golden_dpm_table), &(pi->dpm_table), + sizeof(struct ci_dpm_table)); + return 0; } @@ -5754,13 +5764,22 @@ static int ci_dpm_init_microcode(struct amdgpu_device *adev) switch (adev->asic_type) { case CHIP_BONAIRE: - chip_name = "bonaire"; + if ((adev->pdev->revision == 0x80) || + (adev->pdev->revision == 0x81) || + (adev->pdev->device == 0x665f)) + chip_name = "bonaire_k"; + else + chip_name = "bonaire"; break; case CHIP_HAWAII: - chip_name = "hawaii"; + if (adev->pdev->revision == 0x80) + chip_name = "hawaii_k"; + else + chip_name = "hawaii"; break; case CHIP_KAVERI: case CHIP_KABINI: + case CHIP_MULLINS: default: BUG(); } @@ -5855,7 +5874,10 @@ static int ci_dpm_init(struct amdgpu_device *adev) pi->pcie_dpm_key_disabled = 0; pi->thermal_sclk_dpm_enabled = 0; - pi->caps_sclk_ds = true; + if (amdgpu_sclk_deep_sleep_en) + pi->caps_sclk_ds = true; + else + pi->caps_sclk_ds = false; pi->mclk_strobe_mode_threshold = 40000; pi->mclk_stutter_mode_threshold = 40000; @@ -6404,6 +6426,186 @@ static int ci_dpm_set_powergating_state(void *handle, return 0; } +static int ci_dpm_print_clock_levels(struct amdgpu_device *adev, + enum pp_clock_type type, char *buf) +{ + 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; + struct ci_single_dpm_table *pcie_table = &pi->dpm_table.pcie_speed_table; + + int i, now, size = 0; + uint32_t clock, pcie_speed; + + switch (type) { + case PP_SCLK: + amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_API_GetSclkFrequency); + clock = RREG32(mmSMC_MSG_ARG_0); + + for (i = 0; i < sclk_table->count; i++) { + if (clock > sclk_table->dpm_levels[i].value) + continue; + break; + } + now = i; + + for (i = 0; i < sclk_table->count; i++) + size += sprintf(buf + size, "%d: %uMhz %s\n", + i, sclk_table->dpm_levels[i].value / 100, + (i == now) ? "*" : ""); + break; + case PP_MCLK: + amdgpu_ci_send_msg_to_smc(adev, PPSMC_MSG_API_GetMclkFrequency); + clock = RREG32(mmSMC_MSG_ARG_0); + + for (i = 0; i < mclk_table->count; i++) { + if (clock > mclk_table->dpm_levels[i].value) + continue; + break; + } + now = i; + + for (i = 0; i < mclk_table->count; i++) + size += sprintf(buf + size, "%d: %uMhz %s\n", + i, mclk_table->dpm_levels[i].value / 100, + (i == now) ? "*" : ""); + break; + case PP_PCIE: + pcie_speed = ci_get_current_pcie_speed(adev); + for (i = 0; i < pcie_table->count; i++) { + if (pcie_speed != pcie_table->dpm_levels[i].value) + continue; + break; + } + now = i; + + for (i = 0; i < pcie_table->count; i++) + size += sprintf(buf + size, "%d: %s %s\n", i, + (pcie_table->dpm_levels[i].value == 0) ? "2.5GB, x1" : + (pcie_table->dpm_levels[i].value == 1) ? "5.0GB, x16" : + (pcie_table->dpm_levels[i].value == 2) ? "8.0GB, x16" : "", + (i == now) ? "*" : ""); + break; + default: + break; + } + + return size; +} + +static int ci_dpm_force_clock_level(struct amdgpu_device *adev, + enum pp_clock_type type, uint32_t mask) +{ + struct ci_power_info *pi = ci_get_pi(adev); + + if (adev->pm.dpm.forced_level + != AMDGPU_DPM_FORCED_LEVEL_MANUAL) + return -EINVAL; + + switch (type) { + case PP_SCLK: + if (!pi->sclk_dpm_key_disabled) + amdgpu_ci_send_msg_to_smc_with_parameter(adev, + PPSMC_MSG_SCLKDPM_SetEnabledMask, + pi->dpm_level_enable_mask.sclk_dpm_enable_mask & mask); + break; + + case PP_MCLK: + if (!pi->mclk_dpm_key_disabled) + amdgpu_ci_send_msg_to_smc_with_parameter(adev, + PPSMC_MSG_MCLKDPM_SetEnabledMask, + pi->dpm_level_enable_mask.mclk_dpm_enable_mask & mask); + break; + + case PP_PCIE: + { + uint32_t tmp = mask & pi->dpm_level_enable_mask.pcie_dpm_enable_mask; + uint32_t level = 0; + + while (tmp >>= 1) + level++; + + if (!pi->pcie_dpm_key_disabled) + amdgpu_ci_send_msg_to_smc_with_parameter(adev, + PPSMC_MSG_PCIeDPM_ForceLevel, + level); + break; + } + default: + break; + } + + return 0; +} + +static int ci_dpm_get_sclk_od(struct amdgpu_device *adev) +{ + 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 = + &(pi->golden_dpm_table.sclk_table); + int value; + + value = (sclk_table->dpm_levels[sclk_table->count - 1].value - + golden_sclk_table->dpm_levels[golden_sclk_table->count - 1].value) * + 100 / + golden_sclk_table->dpm_levels[golden_sclk_table->count - 1].value; + + return value; +} + +static int ci_dpm_set_sclk_od(struct amdgpu_device *adev, uint32_t value) +{ + 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 = + &(pi->golden_dpm_table.sclk_table); + + if (value > 20) + value = 20; + + ps->performance_levels[ps->performance_level_count - 1].sclk = + golden_sclk_table->dpm_levels[golden_sclk_table->count - 1].value * + value / 100 + + golden_sclk_table->dpm_levels[golden_sclk_table->count - 1].value; + + return 0; +} + +static int ci_dpm_get_mclk_od(struct amdgpu_device *adev) +{ + 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 = + &(pi->golden_dpm_table.mclk_table); + int value; + + value = (mclk_table->dpm_levels[mclk_table->count - 1].value - + golden_mclk_table->dpm_levels[golden_mclk_table->count - 1].value) * + 100 / + golden_mclk_table->dpm_levels[golden_mclk_table->count - 1].value; + + return value; +} + +static int ci_dpm_set_mclk_od(struct amdgpu_device *adev, uint32_t value) +{ + 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 = + &(pi->golden_dpm_table.mclk_table); + + if (value > 20) + value = 20; + + ps->performance_levels[ps->performance_level_count - 1].mclk = + golden_mclk_table->dpm_levels[golden_mclk_table->count - 1].value * + value / 100 + + golden_mclk_table->dpm_levels[golden_mclk_table->count - 1].value; + + return 0; +} + const struct amd_ip_funcs ci_dpm_ip_funcs = { .name = "ci_dpm", .early_init = ci_dpm_early_init, @@ -6438,6 +6640,12 @@ static const struct amdgpu_dpm_funcs ci_dpm_funcs = { .get_fan_control_mode = &ci_dpm_get_fan_control_mode, .set_fan_speed_percent = &ci_dpm_set_fan_speed_percent, .get_fan_speed_percent = &ci_dpm_get_fan_speed_percent, + .print_clock_levels = ci_dpm_print_clock_levels, + .force_clock_level = ci_dpm_force_clock_level, + .get_sclk_od = ci_dpm_get_sclk_od, + .set_sclk_od = ci_dpm_set_sclk_od, + .get_mclk_od = ci_dpm_get_mclk_od, + .set_mclk_od = ci_dpm_set_mclk_od, }; static void ci_dpm_set_dpm_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.h b/drivers/gpu/drm/amd/amdgpu/ci_dpm.h index faccc30c93bf..91be2996ae7c 100644 --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.h @@ -193,6 +193,7 @@ struct ci_pt_defaults { struct ci_power_info { struct ci_dpm_table dpm_table; + struct ci_dpm_table golden_dpm_table; u32 voltage_control; u32 mvdd_control; u32 vddci_control; diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c index 910431808542..825de800b798 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik.c +++ b/drivers/gpu/drm/amd/amdgpu/cik.c @@ -67,6 +67,7 @@ #include "amdgpu_amdkfd.h" #include "amdgpu_powerplay.h" +#include "dce_virtual.h" /* * Indirect registers accessor @@ -879,7 +880,7 @@ static void cik_vga_set_state(struct amdgpu_device *adev, bool state) uint32_t tmp; tmp = RREG32(mmCONFIG_CNTL); - if (state == false) + if (!state) tmp |= CONFIG_CNTL__VGA_DIS_MASK; else tmp &= ~CONFIG_CNTL__VGA_DIS_MASK; @@ -1035,12 +1036,12 @@ static uint32_t cik_read_indexed_register(struct amdgpu_device *adev, mutex_lock(&adev->grbm_idx_mutex); if (se_num != 0xffffffff || sh_num != 0xffffffff) - gfx_v7_0_select_se_sh(adev, se_num, sh_num); + amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff); val = RREG32(reg_offset); if (se_num != 0xffffffff || sh_num != 0xffffffff) - gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); return val; } @@ -1158,10 +1159,11 @@ static void kv_restore_regs_for_reset(struct amdgpu_device *adev, WREG32(mmGMCON_RENG_EXECUTE, save->gmcon_reng_execute); } -static void cik_gpu_pci_config_reset(struct amdgpu_device *adev) +static int cik_gpu_pci_config_reset(struct amdgpu_device *adev) { struct kv_reset_save_regs kv_save = { 0 }; u32 i; + int r = -EINVAL; dev_info(adev->dev, "GPU pci config reset\n"); @@ -1177,14 +1179,20 @@ static void cik_gpu_pci_config_reset(struct amdgpu_device *adev) /* wait for asic to come out of reset */ for (i = 0; i < adev->usec_timeout; i++) { - if (RREG32(mmCONFIG_MEMSIZE) != 0xffffffff) + if (RREG32(mmCONFIG_MEMSIZE) != 0xffffffff) { + /* enable BM */ + pci_set_master(adev->pdev); + r = 0; break; + } udelay(1); } /* does asic init need to be run first??? */ if (adev->flags & AMD_IS_APU) kv_restore_regs_for_reset(adev, &kv_save); + + return r; } static void cik_set_bios_scratch_engine_hung(struct amdgpu_device *adev, bool hung) @@ -1210,13 +1218,14 @@ static void cik_set_bios_scratch_engine_hung(struct amdgpu_device *adev, bool hu */ static int cik_asic_reset(struct amdgpu_device *adev) { + int r; cik_set_bios_scratch_engine_hung(adev, true); - cik_gpu_pci_config_reset(adev); + r = cik_gpu_pci_config_reset(adev); cik_set_bios_scratch_engine_hung(adev, false); - return 0; + return r; } static int cik_set_uvd_clock(struct amdgpu_device *adev, u32 clock, @@ -1700,6 +1709,74 @@ static const struct amdgpu_ip_block_version bonaire_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version bonaire_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &cik_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &gmc_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 8, + .minor = 2, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 7, + .minor = 2, + .rev = 0, + .funcs = &gfx_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_sdma_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 4, + .minor = 2, + .rev = 0, + .funcs = &uvd_v4_2_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vce_v2_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version hawaii_ip_blocks[] = { /* ORDER MATTERS! */ @@ -1768,6 +1845,74 @@ static const struct amdgpu_ip_block_version hawaii_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version hawaii_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &cik_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &gmc_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 8, + .minor = 5, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 7, + .minor = 3, + .rev = 0, + .funcs = &gfx_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_sdma_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 4, + .minor = 2, + .rev = 0, + .funcs = &uvd_v4_2_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vce_v2_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version kabini_ip_blocks[] = { /* ORDER MATTERS! */ @@ -1836,6 +1981,74 @@ static const struct amdgpu_ip_block_version kabini_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version kabini_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &cik_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &gmc_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 8, + .minor = 3, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 7, + .minor = 2, + .rev = 0, + .funcs = &gfx_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_sdma_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 4, + .minor = 2, + .rev = 0, + .funcs = &uvd_v4_2_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vce_v2_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version mullins_ip_blocks[] = { /* ORDER MATTERS! */ @@ -1904,6 +2117,74 @@ static const struct amdgpu_ip_block_version mullins_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version mullins_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &cik_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &gmc_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 8, + .minor = 3, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 7, + .minor = 2, + .rev = 0, + .funcs = &gfx_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_sdma_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 4, + .minor = 2, + .rev = 0, + .funcs = &uvd_v4_2_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vce_v2_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version kaveri_ip_blocks[] = { /* ORDER MATTERS! */ @@ -1972,32 +2253,128 @@ static const struct amdgpu_ip_block_version kaveri_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version kaveri_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &cik_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &gmc_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 0, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 8, + .minor = 1, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 7, + .minor = 1, + .rev = 0, + .funcs = &gfx_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &cik_sdma_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 4, + .minor = 2, + .rev = 0, + .funcs = &uvd_v4_2_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vce_v2_0_ip_funcs, + }, +}; + int cik_set_ip_blocks(struct amdgpu_device *adev) { - switch (adev->asic_type) { - case CHIP_BONAIRE: - adev->ip_blocks = bonaire_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(bonaire_ip_blocks); - break; - case CHIP_HAWAII: - adev->ip_blocks = hawaii_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(hawaii_ip_blocks); - break; - case CHIP_KAVERI: - adev->ip_blocks = kaveri_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(kaveri_ip_blocks); - break; - case CHIP_KABINI: - adev->ip_blocks = kabini_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(kabini_ip_blocks); - break; - case CHIP_MULLINS: - adev->ip_blocks = mullins_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(mullins_ip_blocks); - break; - default: - /* FIXME: not supported yet */ - return -EINVAL; + if (adev->enable_virtual_display) { + switch (adev->asic_type) { + case CHIP_BONAIRE: + adev->ip_blocks = bonaire_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(bonaire_ip_blocks_vd); + break; + case CHIP_HAWAII: + adev->ip_blocks = hawaii_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(hawaii_ip_blocks_vd); + break; + case CHIP_KAVERI: + adev->ip_blocks = kaveri_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(kaveri_ip_blocks_vd); + break; + case CHIP_KABINI: + adev->ip_blocks = kabini_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(kabini_ip_blocks_vd); + break; + case CHIP_MULLINS: + adev->ip_blocks = mullins_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(mullins_ip_blocks_vd); + break; + default: + /* FIXME: not supported yet */ + return -EINVAL; + } + } else { + switch (adev->asic_type) { + case CHIP_BONAIRE: + adev->ip_blocks = bonaire_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(bonaire_ip_blocks); + break; + case CHIP_HAWAII: + adev->ip_blocks = hawaii_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(hawaii_ip_blocks); + break; + case CHIP_KAVERI: + adev->ip_blocks = kaveri_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(kaveri_ip_blocks); + break; + case CHIP_KABINI: + adev->ip_blocks = kabini_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(kabini_ip_blocks); + break; + case CHIP_MULLINS: + adev->ip_blocks = mullins_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(mullins_ip_blocks); + break; + default: + /* FIXME: not supported yet */ + return -EINVAL; + } } return 0; @@ -2014,9 +2391,6 @@ static const struct amdgpu_asic_funcs cik_asic_funcs = .set_uvd_clocks = &cik_set_uvd_clocks, .set_vce_clocks = &cik_set_vce_clocks, .get_virtual_caps = &cik_get_virtual_caps, - /* these should be moved to their own ip modules */ - .get_gpu_clock_counter = &gfx_v7_0_get_gpu_clock_counter, - .wait_for_mc_idle = &gmc_v7_0_mc_wait_for_idle, }; static int cik_common_early_init(void *handle) diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index 9dc4e24e31e7..e71cd12104b3 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -224,17 +224,6 @@ static void cik_sdma_ring_emit_ib(struct amdgpu_ring *ring, unsigned vm_id, bool ctx_switch) { u32 extra_bits = vm_id & 0xf; - u32 next_rptr = ring->wptr + 5; - - while ((next_rptr & 7) != 4) - next_rptr++; - - next_rptr += 4; - amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0)); - amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); - amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); - amdgpu_ring_write(ring, 1); /* number of DWs to follow */ - amdgpu_ring_write(ring, next_rptr); /* IB packet must end on a 8 DW boundary */ cik_sdma_ring_insert_nop(ring, (12 - (ring->wptr & 7)) % 8); @@ -365,7 +354,7 @@ static void cik_sdma_enable(struct amdgpu_device *adev, bool enable) u32 me_cntl; int i; - if (enable == false) { + if (!enable) { cik_sdma_gfx_stop(adev); cik_sdma_rlc_stop(adev); } @@ -628,20 +617,19 @@ static int cik_sdma_ring_test_ring(struct amdgpu_ring *ring) * Test a simple IB in the DMA ring (CIK). * Returns 0 on success, error on failure. */ -static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring) +static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring, long timeout) { struct amdgpu_device *adev = ring->adev; struct amdgpu_ib ib; struct fence *f = NULL; - unsigned i; unsigned index; - int r; u32 tmp = 0; u64 gpu_addr; + long r; r = amdgpu_wb_get(adev, &index); if (r) { - dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); + dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); return r; } @@ -651,11 +639,12 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring) memset(&ib, 0, sizeof(ib)); r = amdgpu_ib_get(adev, NULL, 256, &ib); if (r) { - DRM_ERROR("amdgpu: failed to get ib (%d).\n", r); + DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); goto err0; } - ib.ptr[0] = SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0); + ib.ptr[0] = SDMA_PACKET(SDMA_OPCODE_WRITE, + SDMA_WRITE_SUB_OPCODE_LINEAR, 0); ib.ptr[1] = lower_32_bits(gpu_addr); ib.ptr[2] = upper_32_bits(gpu_addr); ib.ptr[3] = 1; @@ -665,28 +654,25 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring) if (r) goto err1; - r = fence_wait(f, false); - if (r) { - DRM_ERROR("amdgpu: fence wait failed (%d).\n", r); + r = fence_wait_timeout(f, false, timeout); + if (r == 0) { + DRM_ERROR("amdgpu: IB test timed out\n"); + r = -ETIMEDOUT; goto err1; - } - for (i = 0; i < adev->usec_timeout; i++) { - tmp = le32_to_cpu(adev->wb.wb[index]); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - if (i < adev->usec_timeout) { - DRM_INFO("ib test on ring %d succeeded in %u usecs\n", - ring->idx, i); + } else if (r < 0) { + DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto err1; + } + tmp = le32_to_cpu(adev->wb.wb[index]); + if (tmp == 0xDEADBEEF) { + DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + r = 0; } else { DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp); r = -EINVAL; } err1: - fence_put(f); amdgpu_ib_free(adev, &ib, NULL); fence_put(f); err0: @@ -708,24 +694,16 @@ static void cik_sdma_vm_copy_pte(struct amdgpu_ib *ib, uint64_t pe, uint64_t src, unsigned count) { - while (count) { - unsigned bytes = count * 8; - if (bytes > 0x1FFFF8) - bytes = 0x1FFFF8; - - ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_COPY, - SDMA_WRITE_SUB_OPCODE_LINEAR, 0); - ib->ptr[ib->length_dw++] = bytes; - ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ - ib->ptr[ib->length_dw++] = lower_32_bits(src); - ib->ptr[ib->length_dw++] = upper_32_bits(src); - ib->ptr[ib->length_dw++] = lower_32_bits(pe); - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - - pe += bytes; - src += bytes; - count -= bytes / 8; - } + unsigned bytes = count * 8; + + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_COPY, + SDMA_WRITE_SUB_OPCODE_LINEAR, 0); + ib->ptr[ib->length_dw++] = bytes; + ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ + ib->ptr[ib->length_dw++] = lower_32_bits(src); + ib->ptr[ib->length_dw++] = upper_32_bits(src); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); } /** @@ -733,39 +711,27 @@ static void cik_sdma_vm_copy_pte(struct amdgpu_ib *ib, * * @ib: indirect buffer to fill with commands * @pe: addr of the page entry - * @addr: dst addr to write into pe + * @value: dst addr to write into pe * @count: number of page entries to update * @incr: increase next addr by incr bytes - * @flags: access flags * * Update PTEs by writing them manually using sDMA (CIK). */ -static void cik_sdma_vm_write_pte(struct amdgpu_ib *ib, - const dma_addr_t *pages_addr, uint64_t pe, - uint64_t addr, unsigned count, - uint32_t incr, uint32_t flags) +static void cik_sdma_vm_write_pte(struct amdgpu_ib *ib, uint64_t pe, + uint64_t value, unsigned count, + uint32_t incr) { - uint64_t value; - unsigned ndw; - - while (count) { - ndw = count * 2; - if (ndw > 0xFFFFE) - ndw = 0xFFFFE; - - /* for non-physically contiguous pages (system) */ - ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_WRITE, - SDMA_WRITE_SUB_OPCODE_LINEAR, 0); - ib->ptr[ib->length_dw++] = pe; - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = ndw; - for (; ndw > 0; ndw -= 2, --count, pe += 8) { - value = amdgpu_vm_map_gart(pages_addr, addr); - addr += incr; - value |= flags; - ib->ptr[ib->length_dw++] = value; - ib->ptr[ib->length_dw++] = upper_32_bits(value); - } + unsigned ndw = count * 2; + + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_WRITE, + SDMA_WRITE_SUB_OPCODE_LINEAR, 0); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = ndw; + for (; ndw > 0; ndw -= 2) { + ib->ptr[ib->length_dw++] = lower_32_bits(value); + ib->ptr[ib->length_dw++] = upper_32_bits(value); + value += incr; } } @@ -781,40 +747,21 @@ static void cik_sdma_vm_write_pte(struct amdgpu_ib *ib, * * Update the page tables using sDMA (CIK). */ -static void cik_sdma_vm_set_pte_pde(struct amdgpu_ib *ib, - uint64_t pe, +static void cik_sdma_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe, uint64_t addr, unsigned count, uint32_t incr, uint32_t flags) { - uint64_t value; - unsigned ndw; - - while (count) { - ndw = count; - if (ndw > 0x7FFFF) - ndw = 0x7FFFF; - - if (flags & AMDGPU_PTE_VALID) - value = addr; - else - value = 0; - - /* for physically contiguous pages (vram) */ - ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_GENERATE_PTE_PDE, 0, 0); - ib->ptr[ib->length_dw++] = pe; /* dst addr */ - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = flags; /* mask */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = value; /* value */ - ib->ptr[ib->length_dw++] = upper_32_bits(value); - ib->ptr[ib->length_dw++] = incr; /* increment size */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = ndw; /* number of entries */ - - pe += ndw * 8; - addr += ndw * incr; - count -= ndw; - } + /* for physically contiguous pages (vram) */ + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_GENERATE_PTE_PDE, 0, 0); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); /* dst addr */ + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = flags; /* mask */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = lower_32_bits(addr); /* value */ + ib->ptr[ib->length_dw++] = upper_32_bits(addr); + ib->ptr[ib->length_dw++] = incr; /* increment size */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = count; /* number of entries */ } /** diff --git a/drivers/gpu/drm/amd/amdgpu/cz_dpm.c b/drivers/gpu/drm/amd/amdgpu/cz_dpm.c index 933e425a8154..794c5f36ca68 100644 --- a/drivers/gpu/drm/amd/amdgpu/cz_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/cz_dpm.c @@ -425,7 +425,7 @@ static int cz_dpm_init(struct amdgpu_device *adev) pi->mgcg_cgtt_local1 = 0x0; pi->clock_slow_down_step = 25000; pi->skip_clock_slow_down = 1; - pi->enable_nb_ps_policy = 0; + pi->enable_nb_ps_policy = false; pi->caps_power_containment = true; pi->caps_cac = true; pi->didt_enabled = false; @@ -435,7 +435,11 @@ static int cz_dpm_init(struct amdgpu_device *adev) pi->caps_td_ramping = true; pi->caps_tcp_ramping = true; } - pi->caps_sclk_ds = true; + if (amdgpu_sclk_deep_sleep_en) + pi->caps_sclk_ds = true; + else + pi->caps_sclk_ds = false; + pi->voting_clients = 0x00c00033; pi->auto_thermal_throttling_enabled = true; pi->bapm_enabled = false; @@ -2108,29 +2112,58 @@ static void cz_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate) /* disable clockgating so we can properly shut down the block */ ret = amdgpu_set_clockgating_state(adev, AMD_IP_BLOCK_TYPE_UVD, AMD_CG_STATE_UNGATE); + if (ret) { + DRM_ERROR("UVD DPM Power Gating failed to set clockgating state\n"); + return; + } + /* shutdown the UVD block */ ret = amdgpu_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_UVD, AMD_PG_STATE_GATE); - /* XXX: check for errors */ + + if (ret) { + DRM_ERROR("UVD DPM Power Gating failed to set powergating state\n"); + return; + } } cz_update_uvd_dpm(adev, gate); - if (pi->caps_uvd_pg) + if (pi->caps_uvd_pg) { /* power off the UVD block */ - cz_send_msg_to_smc(adev, PPSMC_MSG_UVDPowerOFF); + ret = cz_send_msg_to_smc(adev, PPSMC_MSG_UVDPowerOFF); + if (ret) { + DRM_ERROR("UVD DPM Power Gating failed to send SMU PowerOFF message\n"); + return; + } + } } else { if (pi->caps_uvd_pg) { /* power on the UVD block */ if (pi->uvd_dynamic_pg) - cz_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_UVDPowerON, 1); + ret = cz_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_UVDPowerON, 1); else - cz_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_UVDPowerON, 0); + ret = cz_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_UVDPowerON, 0); + + if (ret) { + DRM_ERROR("UVD DPM Power Gating Failed to send SMU PowerON message\n"); + return; + } + /* re-init the UVD block */ ret = amdgpu_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_UVD, AMD_PG_STATE_UNGATE); + + if (ret) { + DRM_ERROR("UVD DPM Power Gating Failed to set powergating state\n"); + return; + } + /* enable clockgating. hw will dynamically gate/ungate clocks on the fly */ ret = amdgpu_set_clockgating_state(adev, AMD_IP_BLOCK_TYPE_UVD, AMD_CG_STATE_GATE); - /* XXX: check for errors */ + if (ret) { + DRM_ERROR("UVD DPM Power Gating Failed to set clockgating state\n"); + return; + } } cz_update_uvd_dpm(adev, gate); } @@ -2219,6 +2252,7 @@ static void cz_dpm_powergate_vce(struct amdgpu_device *adev, bool gate) } } } else { /*pi->caps_vce_pg*/ + pi->vce_power_gated = gate; cz_update_vce_dpm(adev); cz_enable_vce_dpm(adev, !gate); } diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c index 8227344d2ff6..bc5bb4eb9625 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c @@ -646,8 +646,8 @@ static void dce_v10_0_resume_mc_access(struct amdgpu_device *adev, if (save->crtc_enabled[i]) { tmp = RREG32(mmMASTER_UPDATE_MODE + crtc_offsets[i]); - if (REG_GET_FIELD(tmp, MASTER_UPDATE_MODE, MASTER_UPDATE_MODE) != 3) { - tmp = REG_SET_FIELD(tmp, MASTER_UPDATE_MODE, MASTER_UPDATE_MODE, 3); + 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]); @@ -712,6 +712,45 @@ static void dce_v10_0_set_vga_render_state(struct amdgpu_device *adev, WREG32(mmVGA_RENDER_CONTROL, tmp); } +static int dce_v10_0_get_num_crtc(struct amdgpu_device *adev) +{ + int num_crtc = 0; + + switch (adev->asic_type) { + case CHIP_FIJI: + case CHIP_TONGA: + num_crtc = 6; + break; + default: + num_crtc = 0; + } + return num_crtc; +} + +void dce_v10_0_disable_dce(struct amdgpu_device *adev) +{ + /*Disable VGA render and enabled crtc, if has DCE engine*/ + if (amdgpu_atombios_has_dce_engine_info(adev)) { + u32 tmp; + int crtc_enabled, i; + + dce_v10_0_set_vga_render_state(adev, false); + + /*Disable crtc*/ + for (i = 0; i < dce_v10_0_get_num_crtc(adev); i++) { + crtc_enabled = REG_GET_FIELD(RREG32(mmCRTC_CONTROL + crtc_offsets[i]), + CRTC_CONTROL, CRTC_MASTER_EN); + if (crtc_enabled) { + 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); + } + } + } +} + static void dce_v10_0_program_fmt(struct drm_encoder *encoder) { struct drm_device *dev = encoder->dev; @@ -2071,6 +2110,7 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc, u32 tmp, viewport_w, viewport_h; int r; bool bypass_lut = false; + char *format_name; /* no fb bound */ if (!atomic && !crtc->primary->fb) { @@ -2182,8 +2222,9 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc, bypass_lut = true; break; default: - DRM_ERROR("Unsupported screen format %s\n", - drm_get_format_name(target_fb->pixel_format)); + format_name = drm_get_format_name(target_fb->pixel_format); + DRM_ERROR("Unsupported screen format %s\n", format_name); + kfree(format_name); return -EINVAL; } @@ -2275,8 +2316,8 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc, WREG32(mmVIEWPORT_SIZE + amdgpu_crtc->crtc_offset, (viewport_w << 16) | viewport_h); - /* set pageflip to happen only at start of vblank interval (front porch) */ - WREG32(mmMASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 3); + /* set pageflip to happen anywhere in vblank interval */ + WREG32(mmMASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 0); if (!atomic && fb && fb != crtc->primary->fb) { amdgpu_fb = to_amdgpu_framebuffer(fb); @@ -2667,19 +2708,21 @@ static void dce_v10_0_cursor_reset(struct drm_crtc *crtc) } } -static void dce_v10_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, - u16 *blue, uint32_t start, uint32_t size) +static int dce_v10_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, + u16 *blue, uint32_t size) { struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - int end = (start + size > 256) ? 256 : start + size, i; + int i; /* userspace palettes are always correct as is */ - for (i = start; i < end; i++) { + 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; } static void dce_v10_0_crtc_destroy(struct drm_crtc *crtc) @@ -2696,7 +2739,7 @@ static const struct drm_crtc_funcs dce_v10_0_crtc_funcs = { .gamma_set = dce_v10_0_crtc_gamma_set, .set_config = amdgpu_crtc_set_config, .destroy = dce_v10_0_crtc_destroy, - .page_flip = amdgpu_crtc_page_flip, + .page_flip_target = amdgpu_crtc_page_flip_target, }; static void dce_v10_0_crtc_dpms(struct drm_crtc *crtc, int mode) @@ -2717,13 +2760,13 @@ static void dce_v10_0_crtc_dpms(struct drm_crtc *crtc, int mode) type = amdgpu_crtc_idx_to_irq_type(adev, amdgpu_crtc->crtc_id); amdgpu_irq_update(adev, &adev->crtc_irq, type); amdgpu_irq_update(adev, &adev->pageflip_irq, type); - drm_vblank_on(dev, amdgpu_crtc->crtc_id); + drm_crtc_vblank_on(crtc); dce_v10_0_crtc_load_lut(crtc); break; case DRM_MODE_DPMS_STANDBY: case DRM_MODE_DPMS_SUSPEND: case DRM_MODE_DPMS_OFF: - drm_vblank_off(dev, amdgpu_crtc->crtc_id); + drm_crtc_vblank_off(crtc); if (amdgpu_crtc->enabled) { dce_v10_0_vga_enable(crtc, true); amdgpu_atombios_crtc_blank(crtc, ATOM_ENABLE); @@ -2960,10 +3003,11 @@ static int dce_v10_0_early_init(void *handle) dce_v10_0_set_display_funcs(adev); dce_v10_0_set_irq_funcs(adev); + adev->mode_info.num_crtc = dce_v10_0_get_num_crtc(adev); + switch (adev->asic_type) { case CHIP_FIJI: case CHIP_TONGA: - adev->mode_info.num_crtc = 6; /* XXX 7??? */ adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 7; break; @@ -3139,11 +3183,26 @@ static int dce_v10_0_wait_for_idle(void *handle) return 0; } +static int dce_v10_0_check_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (dce_v10_0_is_display_hung(adev)) + adev->ip_block_status[AMD_IP_BLOCK_TYPE_DCE].hang = true; + else + adev->ip_block_status[AMD_IP_BLOCK_TYPE_DCE].hang = false; + + return 0; +} + static int dce_v10_0_soft_reset(void *handle) { u32 srbm_soft_reset = 0, tmp; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_DCE].hang) + return 0; + if (dce_v10_0_is_display_hung(adev)) srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_DC_MASK; @@ -3372,7 +3431,7 @@ static int dce_v10_0_pageflip_irq(struct amdgpu_device *adev, spin_unlock_irqrestore(&adev->ddev->event_lock, flags); - drm_vblank_put(adev->ddev, amdgpu_crtc->crtc_id); + drm_crtc_vblank_put(&amdgpu_crtc->base); schedule_work(&works->unpin_work); return 0; @@ -3510,6 +3569,7 @@ const struct amd_ip_funcs dce_v10_0_ip_funcs = { .resume = dce_v10_0_resume, .is_idle = dce_v10_0_is_idle, .wait_for_idle = dce_v10_0_wait_for_idle, + .check_soft_reset = dce_v10_0_check_soft_reset, .soft_reset = dce_v10_0_soft_reset, .set_clockgating_state = dce_v10_0_set_clockgating_state, .set_powergating_state = dce_v10_0_set_powergating_state, diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.h b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.h index 1bfa48ddd8a6..e3dc04d293e4 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.h +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.h @@ -26,4 +26,6 @@ extern const struct amd_ip_funcs dce_v10_0_ip_funcs; +void dce_v10_0_disable_dce(struct amdgpu_device *adev); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c index af26ec0bc59d..b93eba077950 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c @@ -307,11 +307,10 @@ static void dce_v11_0_page_flip(struct amdgpu_device *adev, struct amdgpu_crtc *amdgpu_crtc = adev->mode_info.crtcs[crtc_id]; u32 tmp; - /* flip at hsync for async, default is vsync */ - /* use UPDATE_IMMEDIATE_EN instead for async? */ + /* flip immediate for async, default is vsync */ tmp = RREG32(mmGRPH_FLIP_CONTROL + amdgpu_crtc->crtc_offset); tmp = REG_SET_FIELD(tmp, GRPH_FLIP_CONTROL, - GRPH_SURFACE_UPDATE_H_RETRACE_EN, async ? 1 : 0); + GRPH_SURFACE_UPDATE_IMMEDIATE_EN, async ? 1 : 0); WREG32(mmGRPH_FLIP_CONTROL + amdgpu_crtc->crtc_offset, tmp); /* update the scanout addresses */ WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset, @@ -674,6 +673,53 @@ static void dce_v11_0_set_vga_render_state(struct amdgpu_device *adev, WREG32(mmVGA_RENDER_CONTROL, tmp); } +static int dce_v11_0_get_num_crtc (struct amdgpu_device *adev) +{ + int num_crtc = 0; + + switch (adev->asic_type) { + case CHIP_CARRIZO: + num_crtc = 3; + break; + case CHIP_STONEY: + num_crtc = 2; + break; + case CHIP_POLARIS10: + num_crtc = 6; + break; + case CHIP_POLARIS11: + num_crtc = 5; + break; + default: + num_crtc = 0; + } + return num_crtc; +} + +void dce_v11_0_disable_dce(struct amdgpu_device *adev) +{ + /*Disable VGA render and enabled crtc, if has DCE engine*/ + if (amdgpu_atombios_has_dce_engine_info(adev)) { + u32 tmp; + int crtc_enabled, i; + + dce_v11_0_set_vga_render_state(adev, false); + + /*Disable crtc*/ + for (i = 0; i < dce_v11_0_get_num_crtc(adev); i++) { + crtc_enabled = REG_GET_FIELD(RREG32(mmCRTC_CONTROL + crtc_offsets[i]), + CRTC_CONTROL, CRTC_MASTER_EN); + if (crtc_enabled) { + 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); + } + } + } +} + static void dce_v11_0_program_fmt(struct drm_encoder *encoder) { struct drm_device *dev = encoder->dev; @@ -2047,6 +2093,7 @@ static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc, u32 tmp, viewport_w, viewport_h; int r; bool bypass_lut = false; + char *format_name; /* no fb bound */ if (!atomic && !crtc->primary->fb) { @@ -2158,8 +2205,9 @@ static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc, bypass_lut = true; break; default: - DRM_ERROR("Unsupported screen format %s\n", - drm_get_format_name(target_fb->pixel_format)); + format_name = drm_get_format_name(target_fb->pixel_format); + DRM_ERROR("Unsupported screen format %s\n", format_name); + kfree(format_name); return -EINVAL; } @@ -2251,8 +2299,8 @@ static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc, WREG32(mmVIEWPORT_SIZE + amdgpu_crtc->crtc_offset, (viewport_w << 16) | viewport_h); - /* set pageflip to happen only at start of vblank interval (front porch) */ - WREG32(mmCRTC_MASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 3); + /* set pageflip to happen anywhere in vblank interval */ + WREG32(mmCRTC_MASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 0); if (!atomic && fb && fb != crtc->primary->fb) { amdgpu_fb = to_amdgpu_framebuffer(fb); @@ -2678,19 +2726,21 @@ static void dce_v11_0_cursor_reset(struct drm_crtc *crtc) } } -static void dce_v11_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, - u16 *blue, uint32_t start, uint32_t size) +static int dce_v11_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, + u16 *blue, uint32_t size) { struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - int end = (start + size > 256) ? 256 : start + size, i; + int i; /* userspace palettes are always correct as is */ - for (i = start; i < end; i++) { + 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; } static void dce_v11_0_crtc_destroy(struct drm_crtc *crtc) @@ -2707,7 +2757,7 @@ static const struct drm_crtc_funcs dce_v11_0_crtc_funcs = { .gamma_set = dce_v11_0_crtc_gamma_set, .set_config = amdgpu_crtc_set_config, .destroy = dce_v11_0_crtc_destroy, - .page_flip = amdgpu_crtc_page_flip, + .page_flip_target = amdgpu_crtc_page_flip_target, }; static void dce_v11_0_crtc_dpms(struct drm_crtc *crtc, int mode) @@ -2728,13 +2778,13 @@ static void dce_v11_0_crtc_dpms(struct drm_crtc *crtc, int mode) type = amdgpu_crtc_idx_to_irq_type(adev, amdgpu_crtc->crtc_id); amdgpu_irq_update(adev, &adev->crtc_irq, type); amdgpu_irq_update(adev, &adev->pageflip_irq, type); - drm_vblank_on(dev, amdgpu_crtc->crtc_id); + drm_crtc_vblank_on(crtc); dce_v11_0_crtc_load_lut(crtc); break; case DRM_MODE_DPMS_STANDBY: case DRM_MODE_DPMS_SUSPEND: case DRM_MODE_DPMS_OFF: - drm_vblank_off(dev, amdgpu_crtc->crtc_id); + drm_crtc_vblank_off(crtc); if (amdgpu_crtc->enabled) { dce_v11_0_vga_enable(crtc, true); amdgpu_atombios_crtc_blank(crtc, ATOM_ENABLE); @@ -2998,24 +3048,22 @@ static int dce_v11_0_early_init(void *handle) dce_v11_0_set_display_funcs(adev); dce_v11_0_set_irq_funcs(adev); + adev->mode_info.num_crtc = dce_v11_0_get_num_crtc(adev); + switch (adev->asic_type) { case CHIP_CARRIZO: - adev->mode_info.num_crtc = 3; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 9; break; case CHIP_STONEY: - adev->mode_info.num_crtc = 2; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 9; break; case CHIP_POLARIS10: - adev->mode_info.num_crtc = 6; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 6; break; case CHIP_POLARIS11: - adev->mode_info.num_crtc = 5; adev->mode_info.num_hpd = 5; adev->mode_info.num_dig = 5; break; @@ -3433,7 +3481,7 @@ static int dce_v11_0_pageflip_irq(struct amdgpu_device *adev, spin_unlock_irqrestore(&adev->ddev->event_lock, flags); - drm_vblank_put(adev->ddev, amdgpu_crtc->crtc_id); + drm_crtc_vblank_put(&amdgpu_crtc->base); schedule_work(&works->unpin_work); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.h b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.h index 84e4618f5253..1f58a65ba2ef 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.h +++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.h @@ -26,4 +26,6 @@ extern const struct amd_ip_funcs dce_v11_0_ip_funcs; +void dce_v11_0_disable_dce(struct amdgpu_device *adev); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c index 3fb65e41a6ef..abd5213dfe18 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c @@ -526,36 +526,16 @@ static void dce_v8_0_stop_mc_access(struct amdgpu_device *adev, 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; - +#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) { - amdgpu_display_vblank_wait(adev, i); - WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1); + /*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); - 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); } + 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); @@ -575,55 +555,22 @@ static void dce_v8_0_stop_mc_access(struct amdgpu_device *adev, static void dce_v8_0_resume_mc_access(struct amdgpu_device *adev, struct amdgpu_mode_mc_save *save) { - u32 tmp, frame_count; - int i, j; + 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_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) != 3) { - tmp = REG_SET_FIELD(tmp, MASTER_UPDATE_MODE, MASTER_UPDATE_MODE, 3); - 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); - } } + mdelay(20); } WREG32(mmVGA_MEMORY_BASE_ADDRESS_HIGH, upper_32_bits(adev->mc.vram_start)); @@ -657,6 +604,52 @@ static void dce_v8_0_set_vga_render_state(struct amdgpu_device *adev, WREG32(mmVGA_RENDER_CONTROL, tmp); } +static int dce_v8_0_get_num_crtc(struct amdgpu_device *adev) +{ + int num_crtc = 0; + + switch (adev->asic_type) { + case CHIP_BONAIRE: + case CHIP_HAWAII: + num_crtc = 6; + break; + case CHIP_KAVERI: + num_crtc = 4; + break; + case CHIP_KABINI: + case CHIP_MULLINS: + num_crtc = 2; + break; + default: + num_crtc = 0; + } + return num_crtc; +} + +void dce_v8_0_disable_dce(struct amdgpu_device *adev) +{ + /*Disable VGA render and enabled crtc, if has DCE engine*/ + if (amdgpu_atombios_has_dce_engine_info(adev)) { + u32 tmp; + int crtc_enabled, i; + + dce_v8_0_set_vga_render_state(adev, false); + + /*Disable crtc*/ + for (i = 0; i < dce_v8_0_get_num_crtc(adev); i++) { + crtc_enabled = REG_GET_FIELD(RREG32(mmCRTC_CONTROL + crtc_offsets[i]), + CRTC_CONTROL, CRTC_MASTER_EN); + if (crtc_enabled) { + 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); + } + } + } +} + static void dce_v8_0_program_fmt(struct drm_encoder *encoder) { struct drm_device *dev = encoder->dev; @@ -1554,13 +1547,13 @@ static void dce_v8_0_audio_write_sad_regs(struct drm_encoder *encoder) if (sad->format == eld_reg_to_type[i][1]) { if (sad->channels > max_channels) { - value = (sad->channels << - AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__MAX_CHANNELS__SHIFT) | - (sad->byte2 << - AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__DESCRIPTOR_BYTE_2__SHIFT) | - (sad->freq << - AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__SUPPORTED_FREQUENCIES__SHIFT); - max_channels = sad->channels; + value = (sad->channels << + AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__MAX_CHANNELS__SHIFT) | + (sad->byte2 << + AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__DESCRIPTOR_BYTE_2__SHIFT) | + (sad->freq << + AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__SUPPORTED_FREQUENCIES__SHIFT); + max_channels = sad->channels; } if (sad->format == HDMI_AUDIO_CODING_TYPE_PCM) @@ -1666,7 +1659,7 @@ static void dce_v8_0_afmt_update_ACR(struct drm_encoder *encoder, uint32_t clock struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv; uint32_t offset = dig->afmt->offset; - WREG32(mmHDMI_ACR_32_0 + offset, (acr.cts_32khz << HDMI_ACR_44_0__HDMI_ACR_CTS_44__SHIFT)); + WREG32(mmHDMI_ACR_32_0 + offset, (acr.cts_32khz << HDMI_ACR_32_0__HDMI_ACR_CTS_32__SHIFT)); WREG32(mmHDMI_ACR_32_1 + offset, acr.n_32khz); WREG32(mmHDMI_ACR_44_0 + offset, (acr.cts_44_1khz << HDMI_ACR_44_0__HDMI_ACR_CTS_44__SHIFT)); @@ -1746,6 +1739,7 @@ static void dce_v8_0_afmt_setmode(struct drm_encoder *encoder, /* Silent, r600_hdmi_enable will raise WARN for us */ if (!dig->afmt->enabled) return; + offset = dig->afmt->offset; /* hdmi deep color mode general control packets setup, if bpc > 8 */ @@ -1870,7 +1864,7 @@ static void dce_v8_0_afmt_setmode(struct drm_encoder *encoder, WREG32_OR(mmHDMI_INFOFRAME_CONTROL0 + offset, HDMI_INFOFRAME_CONTROL0__HDMI_AVI_INFO_SEND_MASK | /* enable AVI info frames */ - HDMI_INFOFRAME_CONTROL0__HDMI_AVI_INFO_SEND_MASK); /* required for audio info values to be updated */ + HDMI_INFOFRAME_CONTROL0__HDMI_AVI_INFO_CONT_MASK); /* required for audio info values to be updated */ WREG32_P(mmHDMI_INFOFRAME_CONTROL1 + offset, (2 << HDMI_INFOFRAME_CONTROL1__HDMI_AVI_INFO_LINE__SHIFT), /* anything other than 0 */ @@ -1879,13 +1873,12 @@ static void dce_v8_0_afmt_setmode(struct drm_encoder *encoder, WREG32_OR(mmAFMT_AUDIO_PACKET_CONTROL + offset, AFMT_AUDIO_PACKET_CONTROL__AFMT_AUDIO_SAMPLE_SEND_MASK); /* send audio packets */ - /* it's unknown what these bits do excatly, but it's indeed quite useful for debugging */ WREG32(mmAFMT_RAMP_CONTROL0 + offset, 0x00FFFFFF); WREG32(mmAFMT_RAMP_CONTROL1 + offset, 0x007FFFFF); WREG32(mmAFMT_RAMP_CONTROL2 + offset, 0x00000001); WREG32(mmAFMT_RAMP_CONTROL3 + offset, 0x00000001); - /* enable audio after to setting up hw */ + /* enable audio after setting up hw */ dce_v8_0_audio_enable(adev, dig->afmt->pin, true); } @@ -2005,6 +1998,7 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc, u32 viewport_w, viewport_h; int r; bool bypass_lut = false; + char *format_name; /* no fb bound */ if (!atomic && !crtc->primary->fb) { @@ -2052,7 +2046,7 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc, case DRM_FORMAT_XRGB4444: case DRM_FORMAT_ARGB4444: fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) | - (GRPH_FORMAT_ARGB1555 << GRPH_CONTROL__GRPH_FORMAT__SHIFT)); + (GRPH_FORMAT_ARGB4444 << GRPH_CONTROL__GRPH_FORMAT__SHIFT)); #ifdef __BIG_ENDIAN fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT); #endif @@ -2109,8 +2103,9 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc, bypass_lut = true; break; default: - DRM_ERROR("Unsupported screen format %s\n", - drm_get_format_name(target_fb->pixel_format)); + format_name = drm_get_format_name(target_fb->pixel_format); + DRM_ERROR("Unsupported screen format %s\n", format_name); + kfree(format_name); return -EINVAL; } @@ -2190,8 +2185,8 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc, WREG32(mmVIEWPORT_SIZE + amdgpu_crtc->crtc_offset, (viewport_w << 16) | viewport_h); - /* set pageflip to happen only at start of vblank interval (front porch) */ - WREG32(mmMASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 3); + /* set pageflip to happen anywhere in vblank interval */ + WREG32(mmMASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 0); if (!atomic && fb && fb != crtc->primary->fb) { amdgpu_fb = to_amdgpu_framebuffer(fb); @@ -2574,19 +2569,21 @@ static void dce_v8_0_cursor_reset(struct drm_crtc *crtc) } } -static void dce_v8_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, - u16 *blue, uint32_t start, uint32_t size) +static int dce_v8_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green, + u16 *blue, uint32_t size) { struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - int end = (start + size > 256) ? 256 : start + size, i; + int i; /* userspace palettes are always correct as is */ - for (i = start; i < end; i++) { + 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; } static void dce_v8_0_crtc_destroy(struct drm_crtc *crtc) @@ -2603,7 +2600,7 @@ static const struct drm_crtc_funcs dce_v8_0_crtc_funcs = { .gamma_set = dce_v8_0_crtc_gamma_set, .set_config = amdgpu_crtc_set_config, .destroy = dce_v8_0_crtc_destroy, - .page_flip = amdgpu_crtc_page_flip, + .page_flip_target = amdgpu_crtc_page_flip_target, }; static void dce_v8_0_crtc_dpms(struct drm_crtc *crtc, int mode) @@ -2624,13 +2621,13 @@ static void dce_v8_0_crtc_dpms(struct drm_crtc *crtc, int mode) type = amdgpu_crtc_idx_to_irq_type(adev, amdgpu_crtc->crtc_id); amdgpu_irq_update(adev, &adev->crtc_irq, type); amdgpu_irq_update(adev, &adev->pageflip_irq, type); - drm_vblank_on(dev, amdgpu_crtc->crtc_id); + drm_crtc_vblank_on(crtc); dce_v8_0_crtc_load_lut(crtc); break; case DRM_MODE_DPMS_STANDBY: case DRM_MODE_DPMS_SUSPEND: case DRM_MODE_DPMS_OFF: - drm_vblank_off(dev, amdgpu_crtc->crtc_id); + drm_crtc_vblank_off(crtc); if (amdgpu_crtc->enabled) { dce_v8_0_vga_enable(crtc, true); amdgpu_atombios_crtc_blank(crtc, ATOM_ENABLE); @@ -2704,7 +2701,7 @@ static void dce_v8_0_crtc_disable(struct drm_crtc *crtc) case ATOM_PPLL2: /* disable the ppll */ amdgpu_atombios_crtc_program_pll(crtc, amdgpu_crtc->crtc_id, amdgpu_crtc->pll_id, - 0, 0, ATOM_DISABLE, 0, 0, 0, 0, 0, false, &ss); + 0, 0, ATOM_DISABLE, 0, 0, 0, 0, 0, false, &ss); break; case ATOM_PPLL0: /* disable the ppll */ @@ -2854,21 +2851,20 @@ static int dce_v8_0_early_init(void *handle) dce_v8_0_set_display_funcs(adev); dce_v8_0_set_irq_funcs(adev); + adev->mode_info.num_crtc = dce_v8_0_get_num_crtc(adev); + switch (adev->asic_type) { case CHIP_BONAIRE: case CHIP_HAWAII: - adev->mode_info.num_crtc = 6; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 6; break; case CHIP_KAVERI: - adev->mode_info.num_crtc = 4; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 7; break; case CHIP_KABINI: case CHIP_MULLINS: - adev->mode_info.num_crtc = 2; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 6; /* ? */ break; @@ -3287,7 +3283,6 @@ static int dce_v8_0_crtc_irq(struct amdgpu_device *adev, drm_handle_vblank(adev->ddev, crtc); } DRM_DEBUG("IH: D%d vblank\n", crtc + 1); - break; case 1: /* vline */ if (disp_int & interrupt_status_offsets[crtc].vline) @@ -3296,7 +3291,6 @@ static int dce_v8_0_crtc_irq(struct amdgpu_device *adev, DRM_DEBUG("IH: IH event w/o asserted irq bit?\n"); DRM_DEBUG("IH: D%d vline\n", crtc + 1); - break; default: DRM_DEBUG("Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data); @@ -3376,7 +3370,7 @@ static int dce_v8_0_pageflip_irq(struct amdgpu_device *adev, spin_unlock_irqrestore(&adev->ddev->event_lock, flags); - drm_vblank_put(adev->ddev, amdgpu_crtc->crtc_id); + drm_crtc_vblank_put(&amdgpu_crtc->base); schedule_work(&works->unpin_work); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.h b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.h index 77016852b252..7d0770c3a49b 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.h +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.h @@ -26,4 +26,6 @@ extern const struct amd_ip_funcs dce_v8_0_ip_funcs; +void dce_v8_0_disable_dce(struct amdgpu_device *adev); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c new file mode 100644 index 000000000000..00663a7b4053 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c @@ -0,0 +1,806 @@ +/* + * Copyright 2014 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 "drmP.h" +#include "amdgpu.h" +#include "amdgpu_pm.h" +#include "amdgpu_i2c.h" +#include "atom.h" +#include "amdgpu_pll.h" +#include "amdgpu_connectors.h" +#ifdef CONFIG_DRM_AMDGPU_CIK +#include "dce_v8_0.h" +#endif +#include "dce_v10_0.h" +#include "dce_v11_0.h" +#include "dce_virtual.h" + +static void dce_virtual_set_display_funcs(struct amdgpu_device *adev); +static void dce_virtual_set_irq_funcs(struct amdgpu_device *adev); +static int dce_virtual_pageflip_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry); + +/** + * dce_virtual_vblank_wait - vblank wait asic callback. + * + * @adev: amdgpu_device pointer + * @crtc: crtc to wait for vblank on + * + * Wait for vblank on the requested crtc (evergreen+). + */ +static void dce_virtual_vblank_wait(struct amdgpu_device *adev, int crtc) +{ + return; +} + +static u32 dce_virtual_vblank_get_counter(struct amdgpu_device *adev, int crtc) +{ + return 0; +} + +static void dce_virtual_page_flip(struct amdgpu_device *adev, + int crtc_id, u64 crtc_base, bool async) +{ + return; +} + +static int dce_virtual_crtc_get_scanoutpos(struct amdgpu_device *adev, int crtc, + u32 *vbl, u32 *position) +{ + *vbl = 0; + *position = 0; + + return -EINVAL; +} + +static bool dce_virtual_hpd_sense(struct amdgpu_device *adev, + enum amdgpu_hpd_id hpd) +{ + return true; +} + +static void dce_virtual_hpd_set_polarity(struct amdgpu_device *adev, + enum amdgpu_hpd_id hpd) +{ + return; +} + +static u32 dce_virtual_hpd_get_gpio_reg(struct amdgpu_device *adev) +{ + return 0; +} + +static bool dce_virtual_is_display_hung(struct amdgpu_device *adev) +{ + return false; +} + +void dce_virtual_stop_mc_access(struct amdgpu_device *adev, + struct amdgpu_mode_mc_save *save) +{ + switch (adev->asic_type) { + case CHIP_BONAIRE: + case CHIP_HAWAII: + case CHIP_KAVERI: + case CHIP_KABINI: + case CHIP_MULLINS: +#ifdef CONFIG_DRM_AMDGPU_CIK + dce_v8_0_disable_dce(adev); +#endif + break; + 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: + /* no DCE */ + return; + default: + DRM_ERROR("Virtual display unsupported ASIC type: 0x%X\n", adev->asic_type); + } + + return; +} +void dce_virtual_resume_mc_access(struct amdgpu_device *adev, + struct amdgpu_mode_mc_save *save) +{ + return; +} + +void dce_virtual_set_vga_render_state(struct amdgpu_device *adev, + bool render) +{ + return; +} + +/** + * dce_virtual_bandwidth_update - program display watermarks + * + * @adev: amdgpu_device pointer + * + * Calculate and program the display watermarks and line + * buffer allocation (CIK). + */ +static void dce_virtual_bandwidth_update(struct amdgpu_device *adev) +{ + return; +} + +static int dce_virtual_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, + u16 *green, u16 *blue, uint32_t size) +{ + 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; +} + +static void dce_virtual_crtc_destroy(struct drm_crtc *crtc) +{ + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + + drm_crtc_cleanup(crtc); + kfree(amdgpu_crtc); +} + +static const struct drm_crtc_funcs dce_virtual_crtc_funcs = { + .cursor_set2 = NULL, + .cursor_move = NULL, + .gamma_set = dce_virtual_crtc_gamma_set, + .set_config = amdgpu_crtc_set_config, + .destroy = dce_virtual_crtc_destroy, + .page_flip_target = amdgpu_crtc_page_flip_target, +}; + +static void dce_virtual_crtc_dpms(struct drm_crtc *crtc, int mode) +{ + struct drm_device *dev = crtc->dev; + struct amdgpu_device *adev = dev->dev_private; + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + unsigned type; + + switch (mode) { + case DRM_MODE_DPMS_ON: + amdgpu_crtc->enabled = true; + /* Make sure VBLANK and PFLIP interrupts are still enabled */ + type = amdgpu_crtc_idx_to_irq_type(adev, amdgpu_crtc->crtc_id); + amdgpu_irq_update(adev, &adev->crtc_irq, type); + amdgpu_irq_update(adev, &adev->pageflip_irq, type); + drm_vblank_on(dev, amdgpu_crtc->crtc_id); + break; + case DRM_MODE_DPMS_STANDBY: + case DRM_MODE_DPMS_SUSPEND: + case DRM_MODE_DPMS_OFF: + drm_vblank_off(dev, amdgpu_crtc->crtc_id); + amdgpu_crtc->enabled = false; + break; + } +} + + +static void dce_virtual_crtc_prepare(struct drm_crtc *crtc) +{ + dce_virtual_crtc_dpms(crtc, DRM_MODE_DPMS_OFF); +} + +static void dce_virtual_crtc_commit(struct drm_crtc *crtc) +{ + dce_virtual_crtc_dpms(crtc, DRM_MODE_DPMS_ON); +} + +static void dce_virtual_crtc_disable(struct drm_crtc *crtc) +{ + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + + dce_virtual_crtc_dpms(crtc, DRM_MODE_DPMS_OFF); + if (crtc->primary->fb) { + int r; + struct amdgpu_framebuffer *amdgpu_fb; + struct amdgpu_bo *rbo; + + amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb); + rbo = gem_to_amdgpu_bo(amdgpu_fb->obj); + r = amdgpu_bo_reserve(rbo, false); + if (unlikely(r)) + DRM_ERROR("failed to reserve rbo before unpin\n"); + else { + amdgpu_bo_unpin(rbo); + amdgpu_bo_unreserve(rbo); + } + } + + amdgpu_crtc->pll_id = ATOM_PPLL_INVALID; + amdgpu_crtc->encoder = NULL; + amdgpu_crtc->connector = NULL; +} + +static int dce_virtual_crtc_mode_set(struct drm_crtc *crtc, + struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode, + int x, int y, struct drm_framebuffer *old_fb) +{ + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + + /* update the hw version fpr dpm */ + amdgpu_crtc->hw_mode = *adjusted_mode; + + return 0; +} + +static bool dce_virtual_crtc_mode_fixup(struct drm_crtc *crtc, + const struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode) +{ + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + struct drm_device *dev = crtc->dev; + struct drm_encoder *encoder; + + /* assign the encoder to the amdgpu crtc to avoid repeated lookups later */ + list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) { + if (encoder->crtc == crtc) { + amdgpu_crtc->encoder = encoder; + amdgpu_crtc->connector = amdgpu_get_connector_for_encoder(encoder); + break; + } + } + if ((amdgpu_crtc->encoder == NULL) || (amdgpu_crtc->connector == NULL)) { + amdgpu_crtc->encoder = NULL; + amdgpu_crtc->connector = NULL; + return false; + } + + return true; +} + + +static int dce_virtual_crtc_set_base(struct drm_crtc *crtc, int x, int y, + struct drm_framebuffer *old_fb) +{ + 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) +{ + return 0; +} + +static const struct drm_crtc_helper_funcs dce_virtual_crtc_helper_funcs = { + .dpms = dce_virtual_crtc_dpms, + .mode_fixup = dce_virtual_crtc_mode_fixup, + .mode_set = dce_virtual_crtc_mode_set, + .mode_set_base = dce_virtual_crtc_set_base, + .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); + if (amdgpu_crtc == NULL) + return -ENOMEM; + + drm_crtc_init(adev->ddev, &amdgpu_crtc->base, &dce_virtual_crtc_funcs); + + drm_mode_crtc_set_gamma_size(&amdgpu_crtc->base, 256); + 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; + drm_crtc_helper_add(&amdgpu_crtc->base, &dce_virtual_crtc_helper_funcs); + + return 0; +} + +static int dce_virtual_early_init(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + adev->mode_info.vsync_timer_enabled = AMDGPU_IRQ_STATE_DISABLE; + dce_virtual_set_display_funcs(adev); + dce_virtual_set_irq_funcs(adev); + + adev->mode_info.num_crtc = 1; + adev->mode_info.num_hpd = 1; + adev->mode_info.num_dig = 1; + return 0; +} + +static bool dce_virtual_get_connector_info(struct amdgpu_device *adev) +{ + struct amdgpu_i2c_bus_rec ddc_bus; + struct amdgpu_router router; + struct amdgpu_hpd hpd; + + /* look up gpio for ddc, hpd */ + ddc_bus.valid = false; + hpd.hpd = AMDGPU_HPD_NONE; + /* needed for aux chan transactions */ + ddc_bus.hpd = hpd.hpd; + + memset(&router, 0, sizeof(router)); + router.ddc_valid = false; + router.cd_valid = false; + amdgpu_display_add_connector(adev, + 0, + ATOM_DEVICE_CRT1_SUPPORT, + DRM_MODE_CONNECTOR_VIRTUAL, &ddc_bus, + CONNECTOR_OBJECT_ID_VIRTUAL, + &hpd, + &router); + + amdgpu_display_add_encoder(adev, ENCODER_VIRTUAL_ENUM_VIRTUAL, + ATOM_DEVICE_CRT1_SUPPORT, + 0); + + amdgpu_link_encoder_connector(adev->ddev); + + return true; +} + +static int dce_virtual_sw_init(void *handle) +{ + int r, i; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = amdgpu_irq_add_id(adev, 229, &adev->crtc_irq); + if (r) + return r; + + adev->ddev->max_vblank_count = 0; + + adev->ddev->mode_config.funcs = &amdgpu_mode_funcs; + + adev->ddev->mode_config.max_width = 16384; + adev->ddev->mode_config.max_height = 16384; + + adev->ddev->mode_config.preferred_depth = 24; + adev->ddev->mode_config.prefer_shadow = 1; + + adev->ddev->mode_config.fb_base = adev->mc.aper_base; + + r = amdgpu_modeset_create_props(adev); + if (r) + return r; + + adev->ddev->mode_config.max_width = 16384; + adev->ddev->mode_config.max_height = 16384; + + /* allocate crtcs */ + for (i = 0; i < adev->mode_info.num_crtc; i++) { + r = dce_virtual_crtc_init(adev, i); + if (r) + return r; + } + + dce_virtual_get_connector_info(adev); + amdgpu_print_display_setup(adev->ddev); + + drm_kms_helper_poll_init(adev->ddev); + + adev->mode_info.mode_config_initialized = true; + return 0; +} + +static int dce_virtual_sw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + kfree(adev->mode_info.bios_hardcoded_edid); + + drm_kms_helper_poll_fini(adev->ddev); + + drm_mode_config_cleanup(adev->ddev); + adev->mode_info.mode_config_initialized = false; + return 0; +} + +static int dce_virtual_hw_init(void *handle) +{ + return 0; +} + +static int dce_virtual_hw_fini(void *handle) +{ + return 0; +} + +static int dce_virtual_suspend(void *handle) +{ + return dce_virtual_hw_fini(handle); +} + +static int dce_virtual_resume(void *handle) +{ + int ret; + + ret = dce_virtual_hw_init(handle); + + return ret; +} + +static bool dce_virtual_is_idle(void *handle) +{ + return true; +} + +static int dce_virtual_wait_for_idle(void *handle) +{ + return 0; +} + +static int dce_virtual_soft_reset(void *handle) +{ + return 0; +} + +static int dce_virtual_set_clockgating_state(void *handle, + enum amd_clockgating_state state) +{ + return 0; +} + +static int dce_virtual_set_powergating_state(void *handle, + enum amd_powergating_state state) +{ + return 0; +} + +const struct amd_ip_funcs dce_virtual_ip_funcs = { + .name = "dce_virtual", + .early_init = dce_virtual_early_init, + .late_init = NULL, + .sw_init = dce_virtual_sw_init, + .sw_fini = dce_virtual_sw_fini, + .hw_init = dce_virtual_hw_init, + .hw_fini = dce_virtual_hw_fini, + .suspend = dce_virtual_suspend, + .resume = dce_virtual_resume, + .is_idle = dce_virtual_is_idle, + .wait_for_idle = dce_virtual_wait_for_idle, + .soft_reset = dce_virtual_soft_reset, + .set_clockgating_state = dce_virtual_set_clockgating_state, + .set_powergating_state = dce_virtual_set_powergating_state, +}; + +/* these are handled by the primary encoders */ +static void dce_virtual_encoder_prepare(struct drm_encoder *encoder) +{ + return; +} + +static void dce_virtual_encoder_commit(struct drm_encoder *encoder) +{ + return; +} + +static void +dce_virtual_encoder_mode_set(struct drm_encoder *encoder, + struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode) +{ + return; +} + +static void dce_virtual_encoder_disable(struct drm_encoder *encoder) +{ + return; +} + +static void +dce_virtual_encoder_dpms(struct drm_encoder *encoder, int mode) +{ + return; +} + +static bool dce_virtual_encoder_mode_fixup(struct drm_encoder *encoder, + const struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode) +{ + + /* set the active encoder to connector routing */ + amdgpu_encoder_set_active_device(encoder); + + return true; +} + +static const struct drm_encoder_helper_funcs dce_virtual_encoder_helper_funcs = { + .dpms = dce_virtual_encoder_dpms, + .mode_fixup = dce_virtual_encoder_mode_fixup, + .prepare = dce_virtual_encoder_prepare, + .mode_set = dce_virtual_encoder_mode_set, + .commit = dce_virtual_encoder_commit, + .disable = dce_virtual_encoder_disable, +}; + +static void dce_virtual_encoder_destroy(struct drm_encoder *encoder) +{ + struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder); + + kfree(amdgpu_encoder->enc_priv); + drm_encoder_cleanup(encoder); + kfree(amdgpu_encoder); +} + +static const struct drm_encoder_funcs dce_virtual_encoder_funcs = { + .destroy = dce_virtual_encoder_destroy, +}; + +static void dce_virtual_encoder_add(struct amdgpu_device *adev, + uint32_t encoder_enum, + uint32_t supported_device, + u16 caps) +{ + struct drm_device *dev = adev->ddev; + struct drm_encoder *encoder; + struct amdgpu_encoder *amdgpu_encoder; + + /* see if we already added it */ + list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) { + amdgpu_encoder = to_amdgpu_encoder(encoder); + if (amdgpu_encoder->encoder_enum == encoder_enum) { + amdgpu_encoder->devices |= supported_device; + return; + } + + } + + /* add a new one */ + amdgpu_encoder = kzalloc(sizeof(struct amdgpu_encoder), GFP_KERNEL); + if (!amdgpu_encoder) + return; + + encoder = &amdgpu_encoder->base; + encoder->possible_crtcs = 0x1; + amdgpu_encoder->enc_priv = NULL; + amdgpu_encoder->encoder_enum = encoder_enum; + amdgpu_encoder->encoder_id = (encoder_enum & OBJECT_ID_MASK) >> OBJECT_ID_SHIFT; + amdgpu_encoder->devices = supported_device; + amdgpu_encoder->rmx_type = RMX_OFF; + amdgpu_encoder->underscan_type = UNDERSCAN_OFF; + amdgpu_encoder->is_ext_encoder = false; + amdgpu_encoder->caps = caps; + + drm_encoder_init(dev, encoder, &dce_virtual_encoder_funcs, + DRM_MODE_ENCODER_VIRTUAL, NULL); + drm_encoder_helper_add(encoder, &dce_virtual_encoder_helper_funcs); + DRM_INFO("[FM]encoder: %d is VIRTUAL\n", amdgpu_encoder->encoder_id); +} + +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, + .is_display_hung = &dce_virtual_is_display_hung, + .backlight_set_level = NULL, + .backlight_get_level = NULL, + .hpd_sense = &dce_virtual_hpd_sense, + .hpd_set_polarity = &dce_virtual_hpd_set_polarity, + .hpd_get_gpio_reg = &dce_virtual_hpd_get_gpio_reg, + .page_flip = &dce_virtual_page_flip, + .page_flip_get_scanoutpos = &dce_virtual_crtc_get_scanoutpos, + .add_encoder = &dce_virtual_encoder_add, + .add_connector = &amdgpu_connector_add, + .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) +{ + if (adev->mode_info.funcs == NULL) + adev->mode_info.funcs = &dce_virtual_display_funcs; +} + +static enum hrtimer_restart dce_virtual_vblank_timer_handle(struct hrtimer *vblank_timer) +{ + struct amdgpu_mode_info *mode_info = container_of(vblank_timer, struct amdgpu_mode_info ,vblank_timer); + struct amdgpu_device *adev = container_of(mode_info, struct amdgpu_device ,mode_info); + unsigned crtc = 0; + drm_handle_vblank(adev->ddev, crtc); + dce_virtual_pageflip_irq(adev, NULL, NULL); + hrtimer_start(vblank_timer, ktime_set(0, DCE_VIRTUAL_VBLANK_PERIOD), HRTIMER_MODE_REL); + return HRTIMER_NORESTART; +} + +static void dce_virtual_set_crtc_vblank_interrupt_state(struct amdgpu_device *adev, + int crtc, + enum amdgpu_interrupt_state state) +{ + if (crtc >= adev->mode_info.num_crtc) { + DRM_DEBUG("invalid crtc %d\n", crtc); + return; + } + + if (state && !adev->mode_info.vsync_timer_enabled) { + DRM_DEBUG("Enable software vsync timer\n"); + hrtimer_init(&adev->mode_info.vblank_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); + hrtimer_set_expires(&adev->mode_info.vblank_timer, ktime_set(0, DCE_VIRTUAL_VBLANK_PERIOD)); + adev->mode_info.vblank_timer.function = dce_virtual_vblank_timer_handle; + hrtimer_start(&adev->mode_info.vblank_timer, ktime_set(0, DCE_VIRTUAL_VBLANK_PERIOD), HRTIMER_MODE_REL); + } else if (!state && adev->mode_info.vsync_timer_enabled) { + DRM_DEBUG("Disable software vsync timer\n"); + hrtimer_cancel(&adev->mode_info.vblank_timer); + } + + adev->mode_info.vsync_timer_enabled = state; + DRM_DEBUG("[FM]set crtc %d vblank interrupt state %d\n", crtc, state); +} + + +static int dce_virtual_set_crtc_irq_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned type, + enum amdgpu_interrupt_state state) +{ + switch (type) { + case AMDGPU_CRTC_IRQ_VBLANK1: + dce_virtual_set_crtc_vblank_interrupt_state(adev, 0, state); + break; + default: + break; + } + return 0; +} + +static void dce_virtual_crtc_vblank_int_ack(struct amdgpu_device *adev, + int crtc) +{ + if (crtc >= adev->mode_info.num_crtc) { + DRM_DEBUG("invalid crtc %d\n", crtc); + return; + } +} + +static int dce_virtual_crtc_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + unsigned crtc = 0; + unsigned irq_type = AMDGPU_CRTC_IRQ_VBLANK1; + + dce_virtual_crtc_vblank_int_ack(adev, crtc); + + if (amdgpu_irq_enabled(adev, source, irq_type)) { + drm_handle_vblank(adev->ddev, crtc); + } + dce_virtual_pageflip_irq(adev, NULL, NULL); + DRM_DEBUG("IH: D%d vblank\n", crtc + 1); + return 0; +} + +static int dce_virtual_set_pageflip_irq_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *src, + unsigned type, + enum amdgpu_interrupt_state state) +{ + if (type >= adev->mode_info.num_crtc) { + DRM_ERROR("invalid pageflip crtc %d\n", type); + return -EINVAL; + } + DRM_DEBUG("[FM]set pageflip irq type %d state %d\n", type, state); + + return 0; +} + +static int dce_virtual_pageflip_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + unsigned long flags; + unsigned crtc_id = 0; + struct amdgpu_crtc *amdgpu_crtc; + struct amdgpu_flip_work *works; + + crtc_id = 0; + amdgpu_crtc = adev->mode_info.crtcs[crtc_id]; + + if (crtc_id >= adev->mode_info.num_crtc) { + DRM_ERROR("invalid pageflip crtc %d\n", crtc_id); + return -EINVAL; + } + + /* IRQ could occur when in initial stage */ + if (amdgpu_crtc == NULL) + return 0; + + spin_lock_irqsave(&adev->ddev->event_lock, flags); + works = amdgpu_crtc->pflip_works; + if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED) { + DRM_DEBUG_DRIVER("amdgpu_crtc->pflip_status = %d != " + "AMDGPU_FLIP_SUBMITTED(%d)\n", + amdgpu_crtc->pflip_status, + AMDGPU_FLIP_SUBMITTED); + spin_unlock_irqrestore(&adev->ddev->event_lock, flags); + return 0; + } + + /* page flip completed. clean up */ + amdgpu_crtc->pflip_status = AMDGPU_FLIP_NONE; + amdgpu_crtc->pflip_works = NULL; + + /* wakeup usersapce */ + if (works->event) + drm_crtc_send_vblank_event(&amdgpu_crtc->base, works->event); + + spin_unlock_irqrestore(&adev->ddev->event_lock, flags); + + drm_crtc_vblank_put(&amdgpu_crtc->base); + schedule_work(&works->unpin_work); + + return 0; +} + +static const struct amdgpu_irq_src_funcs dce_virtual_crtc_irq_funcs = { + .set = dce_virtual_set_crtc_irq_state, + .process = dce_virtual_crtc_irq, +}; + +static const struct amdgpu_irq_src_funcs dce_virtual_pageflip_irq_funcs = { + .set = dce_virtual_set_pageflip_irq_state, + .process = dce_virtual_pageflip_irq, +}; + +static void dce_virtual_set_irq_funcs(struct amdgpu_device *adev) +{ + adev->crtc_irq.num_types = AMDGPU_CRTC_IRQ_LAST; + adev->crtc_irq.funcs = &dce_virtual_crtc_irq_funcs; + + adev->pageflip_irq.num_types = AMDGPU_PAGEFLIP_IRQ_LAST; + adev->pageflip_irq.funcs = &dce_virtual_pageflip_irq_funcs; +} + diff --git a/drivers/gpu/drm/amd/amdgpu/dce_virtual.h b/drivers/gpu/drm/amd/amdgpu/dce_virtual.h new file mode 100644 index 000000000000..e239243f6ebc --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/dce_virtual.h @@ -0,0 +1,31 @@ +/* + * Copyright 2014 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 __DCE_VIRTUAL_H__ +#define __DCE_VIRTUAL_H__ + +extern const struct amd_ip_funcs dce_virtual_ip_funcs; +#define DCE_VIRTUAL_VBLANK_PERIOD 16666666 + +#endif + diff --git a/drivers/gpu/drm/amd/amdgpu/fiji_smc.c b/drivers/gpu/drm/amd/amdgpu/fiji_smc.c index b336c918d6a7..b3e19ba4c57f 100644 --- a/drivers/gpu/drm/amd/amdgpu/fiji_smc.c +++ b/drivers/gpu/drm/amd/amdgpu/fiji_smc.c @@ -173,7 +173,7 @@ static int fiji_send_msg_to_smc(struct amdgpu_device *adev, PPSMC_Msg msg) { if (!fiji_is_smc_ram_running(adev)) { - return -EINVAL;; + return -EINVAL; } if (wait_smu_response(adev)) { diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index fc8ff4d3ccf8..f4fbec3e224e 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -1583,9 +1583,15 @@ static void gfx_v7_0_tiling_mode_table_init(struct amdgpu_device *adev) * registers are instanced per SE or SH. 0xffffffff means * broadcast to all SEs or SHs (CIK). */ -void gfx_v7_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num) +static void gfx_v7_0_select_se_sh(struct amdgpu_device *adev, + u32 se_num, u32 sh_num, u32 instance) { - u32 data = GRBM_GFX_INDEX__INSTANCE_BROADCAST_WRITES_MASK; + u32 data; + + 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) && (sh_num == 0xffffffff)) data |= GRBM_GFX_INDEX__SH_BROADCAST_WRITES_MASK | @@ -1659,13 +1665,13 @@ static void gfx_v7_0_setup_rb(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { - gfx_v7_0_select_se_sh(adev, i, j); + gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff); data = gfx_v7_0_get_rb_active_bitmap(adev); active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) * rb_bitmap_width_per_sh); } } - gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); adev->gfx.config.backend_enable_mask = active_rbs; @@ -1746,7 +1752,7 @@ static void gfx_v7_0_gpu_init(struct amdgpu_device *adev) * making sure that the following register writes will be broadcasted * to all the shaders */ - gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); /* XXX SH_MEM regs */ /* where to put LDS, scratch, GPUVM in FSA64 space */ @@ -2050,17 +2056,6 @@ static void gfx_v7_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, unsigned vm_id, bool ctx_switch) { u32 header, control = 0; - u32 next_rptr = ring->wptr + 5; - - if (ctx_switch) - next_rptr += 2; - - next_rptr += 4; - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM); - amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); - amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); - amdgpu_ring_write(ring, next_rptr); /* insert SWITCH_BUFFER packet before first IB in the ring frame */ if (ctx_switch) { @@ -2089,22 +2084,9 @@ static void gfx_v7_0_ring_emit_ib_compute(struct amdgpu_ring *ring, struct amdgpu_ib *ib, unsigned vm_id, bool ctx_switch) { - u32 header, control = 0; - u32 next_rptr = ring->wptr + 5; - - control |= INDIRECT_BUFFER_VALID; - next_rptr += 4; - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM); - amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); - amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); - amdgpu_ring_write(ring, next_rptr); - - header = PACKET3(PACKET3_INDIRECT_BUFFER, 2); + u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vm_id << 24); - control |= ib->length_dw | (vm_id << 24); - - amdgpu_ring_write(ring, header); + amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); amdgpu_ring_write(ring, #ifdef __BIG_ENDIAN (2 << 0) | @@ -2123,26 +2105,25 @@ static void gfx_v7_0_ring_emit_ib_compute(struct amdgpu_ring *ring, * Provides a basic gfx ring test to verify that IBs are working. * Returns 0 on success, error on failure. */ -static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring) +static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) { struct amdgpu_device *adev = ring->adev; struct amdgpu_ib ib; struct fence *f = NULL; uint32_t scratch; uint32_t tmp = 0; - unsigned i; - int r; + long r; r = amdgpu_gfx_scratch_get(adev, &scratch); if (r) { - DRM_ERROR("amdgpu: failed to get scratch reg (%d).\n", r); + DRM_ERROR("amdgpu: failed to get scratch reg (%ld).\n", r); return r; } WREG32(scratch, 0xCAFEDEAD); memset(&ib, 0, sizeof(ib)); r = amdgpu_ib_get(adev, NULL, 256, &ib); if (r) { - DRM_ERROR("amdgpu: failed to get ib (%d).\n", r); + DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); goto err1; } ib.ptr[0] = PACKET3(PACKET3_SET_UCONFIG_REG, 1); @@ -2154,21 +2135,19 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring) if (r) goto err2; - r = fence_wait(f, false); - if (r) { - DRM_ERROR("amdgpu: fence wait failed (%d).\n", r); + r = fence_wait_timeout(f, false, timeout); + if (r == 0) { + DRM_ERROR("amdgpu: IB test timed out\n"); + r = -ETIMEDOUT; goto err2; - } - for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(scratch); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - if (i < adev->usec_timeout) { - DRM_INFO("ib test on ring %d succeeded in %u usecs\n", - ring->idx, i); + } else if (r < 0) { + DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto err2; + } + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) { + DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + r = 0; } else { DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp); @@ -2176,7 +2155,6 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring) } err2: - fence_put(f); amdgpu_ib_free(adev, &ib, NULL); fence_put(f); err1: @@ -3221,7 +3199,8 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev) } } adev->gfx.rlc.cs_data = ci_cs_data; - adev->gfx.rlc.cp_table_size = CP_ME_TABLE_SIZE * 5 * 4; + adev->gfx.rlc.cp_table_size = ALIGN(CP_ME_TABLE_SIZE * 5 * 4, 2048); /* CP JT */ + adev->gfx.rlc.cp_table_size += 64 * 1024; /* GDS */ src_ptr = adev->gfx.rlc.reg_list; dws = adev->gfx.rlc.reg_list_size; @@ -3379,7 +3358,7 @@ static void gfx_v7_0_wait_for_rlc_serdes(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { - gfx_v7_0_select_se_sh(adev, i, j); + gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff); for (k = 0; k < adev->usec_timeout; k++) { if (RREG32(mmRLC_SERDES_CU_MASTER_BUSY) == 0) break; @@ -3387,7 +3366,7 @@ static void gfx_v7_0_wait_for_rlc_serdes(struct amdgpu_device *adev) } } } - gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK | @@ -3434,7 +3413,7 @@ static u32 gfx_v7_0_halt_rlc(struct amdgpu_device *adev) return orig; } -void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev) +static void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev) { u32 tmp, i, mask; @@ -3456,7 +3435,7 @@ void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev) } } -void gfx_v7_0_exit_rlc_safe_mode(struct amdgpu_device *adev) +static void gfx_v7_0_exit_rlc_safe_mode(struct amdgpu_device *adev) { u32 tmp; @@ -3471,7 +3450,7 @@ void gfx_v7_0_exit_rlc_safe_mode(struct amdgpu_device *adev) * * Halt the RLC ME (MicroEngine) (CIK). */ -void gfx_v7_0_rlc_stop(struct amdgpu_device *adev) +static void gfx_v7_0_rlc_stop(struct amdgpu_device *adev) { WREG32(mmRLC_CNTL, 0); @@ -3547,7 +3526,7 @@ static int gfx_v7_0_rlc_resume(struct amdgpu_device *adev) WREG32(mmRLC_LB_CNTR_MAX, 0x00008000); mutex_lock(&adev->grbm_idx_mutex); - gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); WREG32(mmRLC_LB_INIT_CU_MASK, 0xffffffff); WREG32(mmRLC_LB_PARAMS, 0x00600408); WREG32(mmRLC_LB_CNTL, 0x80000004); @@ -3587,7 +3566,7 @@ static void gfx_v7_0_enable_cgcg(struct amdgpu_device *adev, bool enable) tmp = gfx_v7_0_halt_rlc(adev); mutex_lock(&adev->grbm_idx_mutex); - gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff); WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff); tmp2 = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK | @@ -3638,7 +3617,7 @@ static void gfx_v7_0_enable_mgcg(struct amdgpu_device *adev, bool enable) tmp = gfx_v7_0_halt_rlc(adev); mutex_lock(&adev->grbm_idx_mutex); - gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff); WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff); data = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK | @@ -3689,7 +3668,7 @@ static void gfx_v7_0_enable_mgcg(struct amdgpu_device *adev, bool enable) tmp = gfx_v7_0_halt_rlc(adev); mutex_lock(&adev->grbm_idx_mutex); - gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff); WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff); data = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK | RLC_SERDES_WR_CTRL__MGCG_OVERRIDE_1_MASK; @@ -3867,6 +3846,20 @@ static void gfx_v7_0_enable_gfx_cgpg(struct amdgpu_device *adev, } } +static void gfx_v7_0_set_user_cu_inactive_bitmap(struct amdgpu_device *adev, + u32 bitmap) +{ + u32 data; + + if (!bitmap) + return; + + data = bitmap << GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT; + data &= GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK; + + WREG32(mmGC_USER_SHADER_ARRAY_CONFIG, data); +} + static u32 gfx_v7_0_get_cu_active_bitmap(struct amdgpu_device *adev) { u32 data, mask; @@ -4123,7 +4116,7 @@ static void gfx_v7_0_fini_pg(struct amdgpu_device *adev) * Fetches a GPU clock counter snapshot (SI). * Returns the 64 bit clock counter snapshot. */ -uint64_t gfx_v7_0_get_gpu_clock_counter(struct amdgpu_device *adev) +static uint64_t gfx_v7_0_get_gpu_clock_counter(struct amdgpu_device *adev) { uint64_t clock; @@ -4183,12 +4176,24 @@ static void gfx_v7_0_ring_emit_gds_switch(struct amdgpu_ring *ring, amdgpu_ring_write(ring, (1 << (oa_size + oa_base)) - (1 << oa_base)); } +static const struct amdgpu_gfx_funcs gfx_v7_0_gfx_funcs = { + .get_gpu_clock_counter = &gfx_v7_0_get_gpu_clock_counter, + .select_se_sh = &gfx_v7_0_select_se_sh, +}; + +static const struct amdgpu_rlc_funcs gfx_v7_0_rlc_funcs = { + .enter_safe_mode = gfx_v7_0_enter_rlc_safe_mode, + .exit_safe_mode = gfx_v7_0_exit_rlc_safe_mode +}; + static int gfx_v7_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; adev->gfx.num_gfx_rings = GFX7_NUM_GFX_RINGS; adev->gfx.num_compute_rings = GFX7_NUM_COMPUTE_RINGS; + adev->gfx.funcs = &gfx_v7_0_gfx_funcs; + adev->gfx.rlc.funcs = &gfx_v7_0_rlc_funcs; gfx_v7_0_set_ring_funcs(adev); gfx_v7_0_set_irq_funcs(adev); gfx_v7_0_set_gds_init(adev); @@ -4460,24 +4465,21 @@ static int gfx_v7_0_sw_init(void *handle) } /* reserve GDS, GWS and OA resource for gfx */ - r = amdgpu_bo_create(adev, adev->gds.mem.gfx_partition_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GDS, 0, - NULL, NULL, &adev->gds.gds_gfx_bo); + r = amdgpu_bo_create_kernel(adev, adev->gds.mem.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GDS, + &adev->gds.gds_gfx_bo, NULL, NULL); if (r) return r; - r = amdgpu_bo_create(adev, adev->gds.gws.gfx_partition_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GWS, 0, - NULL, NULL, &adev->gds.gws_gfx_bo); + r = amdgpu_bo_create_kernel(adev, adev->gds.gws.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GWS, + &adev->gds.gws_gfx_bo, NULL, NULL); if (r) return r; - r = amdgpu_bo_create(adev, adev->gds.oa.gfx_partition_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_OA, 0, - NULL, NULL, &adev->gds.oa_gfx_bo); + r = amdgpu_bo_create_kernel(adev, adev->gds.oa.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_OA, + &adev->gds.oa_gfx_bo, NULL, NULL); if (r) return r; @@ -5032,16 +5034,22 @@ static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev) int i, j, k, counter, active_cu_number = 0; u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0; struct amdgpu_cu_info *cu_info = &adev->gfx.cu_info; + unsigned disable_masks[4 * 2]; memset(cu_info, 0, sizeof(*cu_info)); + amdgpu_gfx_parse_disable_cu(disable_masks, 4, 2); + mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { mask = 1; ao_bitmap = 0; counter = 0; - gfx_v7_0_select_se_sh(adev, i, j); + gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff); + if (i < 4 && j < 2) + gfx_v7_0_set_user_cu_inactive_bitmap( + adev, disable_masks[i * 2 + j]); bitmap = gfx_v7_0_get_cu_active_bitmap(adev); cu_info->bitmap[i][j] = bitmap; @@ -5057,7 +5065,7 @@ static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev) ao_cu_mask |= (ao_bitmap << (i * 16 + j * 8)); } } - gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); cu_info->number = active_cu_number; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.h index e747aa935c88..94e3ea147c26 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.h @@ -26,11 +26,4 @@ extern const struct amd_ip_funcs gfx_v7_0_ip_funcs; -/* XXX these shouldn't be exported */ -void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev); -void gfx_v7_0_exit_rlc_safe_mode(struct amdgpu_device *adev); -void gfx_v7_0_rlc_stop(struct amdgpu_device *adev); -uint64_t gfx_v7_0_get_gpu_clock_counter(struct amdgpu_device *adev); -void gfx_v7_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num); - #endif diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index c2ef94511f70..c6a63c2f91e3 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -270,7 +270,8 @@ static const u32 tonga_mgcg_cgcg_init[] = static const u32 golden_settings_polaris11_a11[] = { - mmCB_HW_CONTROL, 0xfffdf3cf, 0x00006208, + mmCB_HW_CONTROL, 0x0000f3cf, 0x00007208, + mmCB_HW_CONTROL_2, 0x0f000000, 0x0f000000, mmCB_HW_CONTROL_3, 0x000001ff, 0x00000040, mmDB_DEBUG2, 0xf00fffff, 0x00000400, mmPA_SC_ENHANCE, 0xffffffff, 0x20000001, @@ -279,7 +280,7 @@ static const u32 golden_settings_polaris11_a11[] = mmPA_SC_RASTER_CONFIG_1, 0x0000003f, 0x00000000, mmRLC_CGCG_CGLS_CTRL, 0x00000003, 0x0001003c, mmRLC_CGCG_CGLS_CTRL_3D, 0xffffffff, 0x0001003c, - mmSQ_CONFIG, 0x07f80000, 0x07180000, + mmSQ_CONFIG, 0x07f80000, 0x01180000, mmTA_CNTL_AUX, 0x000f000f, 0x000b0000, mmTCC_CTRL, 0x00100000, 0xf31fff7f, mmTCP_ADDR_CONFIG, 0x000003ff, 0x000000f3, @@ -301,8 +302,8 @@ static const u32 polaris11_golden_common_all[] = static const u32 golden_settings_polaris10_a11[] = { mmATC_MISC_CG, 0x000c0fc0, 0x000c0200, - mmCB_HW_CONTROL, 0xfffdf3cf, 0x00007208, - mmCB_HW_CONTROL_2, 0, 0x0f000000, + mmCB_HW_CONTROL, 0x0001f3cf, 0x00007208, + mmCB_HW_CONTROL_2, 0x0f000000, 0x0f000000, mmCB_HW_CONTROL_3, 0x000001ff, 0x00000040, mmDB_DEBUG2, 0xf00fffff, 0x00000400, mmPA_SC_ENHANCE, 0xffffffff, 0x20000001, @@ -409,6 +410,7 @@ static const u32 golden_settings_iceland_a11[] = mmPA_SC_LINE_STIPPLE_STATE, 0x0000ff0f, 0x00000000, mmPA_SC_RASTER_CONFIG, 0x3f3fffff, 0x00000002, mmPA_SC_RASTER_CONFIG_1, 0x0000003f, 0x00000000, + mmRLC_CGCG_CGLS_CTRL, 0x00000003, 0x0000003c, mmSQ_RANDOM_WAVE_PRI, 0x001fffff, 0x000006fd, mmTA_CNTL_AUX, 0x000f000f, 0x000b0000, mmTCC_CTRL, 0x00100000, 0xf31fff7f, @@ -505,8 +507,10 @@ static const u32 cz_golden_settings_a11[] = mmGB_GPU_ID, 0x0000000f, 0x00000000, mmPA_SC_ENHANCE, 0xffffffff, 0x00000001, mmPA_SC_LINE_STIPPLE_STATE, 0x0000ff0f, 0x00000000, + mmRLC_CGCG_CGLS_CTRL, 0x00000003, 0x0000003c, mmSQ_RANDOM_WAVE_PRI, 0x001fffff, 0x000006fd, mmTA_CNTL_AUX, 0x000f000f, 0x00010000, + mmTCC_CTRL, 0x00100000, 0xf31fff7f, mmTCC_EXE_DISABLE, 0x00000002, 0x00000002, mmTCP_ADDR_CONFIG, 0x0000000f, 0x000000f3, mmTCP_CHAN_STEER_LO, 0xffffffff, 0x00001302 @@ -699,7 +703,10 @@ static void gfx_v8_0_init_golden_registers(struct amdgpu_device *adev) polaris10_golden_common_all, (const u32)ARRAY_SIZE(polaris10_golden_common_all)); WREG32_SMC(ixCG_ACLK_CNTL, 0x0000001C); - if (adev->pdev->revision == 0xc7) { + if (adev->pdev->revision == 0xc7 && + ((adev->pdev->subsystem_device == 0xb37 && adev->pdev->subsystem_vendor == 0x1002) || + (adev->pdev->subsystem_device == 0x4a8 && adev->pdev->subsystem_vendor == 0x1043) || + (adev->pdev->subsystem_device == 0x9480 && adev->pdev->subsystem_vendor == 0x1682))) { amdgpu_atombios_i2c_channel_trans(adev, 0x10, 0x96, 0x1E, 0xDD); amdgpu_atombios_i2c_channel_trans(adev, 0x10, 0x96, 0x1F, 0xD0); } @@ -787,26 +794,25 @@ static int gfx_v8_0_ring_test_ring(struct amdgpu_ring *ring) return r; } -static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring) +static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) { struct amdgpu_device *adev = ring->adev; struct amdgpu_ib ib; struct fence *f = NULL; uint32_t scratch; uint32_t tmp = 0; - unsigned i; - int r; + long r; r = amdgpu_gfx_scratch_get(adev, &scratch); if (r) { - DRM_ERROR("amdgpu: failed to get scratch reg (%d).\n", r); + DRM_ERROR("amdgpu: failed to get scratch reg (%ld).\n", r); return r; } WREG32(scratch, 0xCAFEDEAD); memset(&ib, 0, sizeof(ib)); r = amdgpu_ib_get(adev, NULL, 256, &ib); if (r) { - DRM_ERROR("amdgpu: failed to get ib (%d).\n", r); + DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); goto err1; } ib.ptr[0] = PACKET3(PACKET3_SET_UCONFIG_REG, 1); @@ -818,28 +824,25 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring) if (r) goto err2; - r = fence_wait(f, false); - if (r) { - DRM_ERROR("amdgpu: fence wait failed (%d).\n", r); + r = fence_wait_timeout(f, false, timeout); + if (r == 0) { + DRM_ERROR("amdgpu: IB test timed out.\n"); + r = -ETIMEDOUT; goto err2; - } - for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(scratch); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - if (i < adev->usec_timeout) { - DRM_INFO("ib test on ring %d succeeded in %u usecs\n", - ring->idx, i); + } else if (r < 0) { + DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto err2; + } + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) { + DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + r = 0; } else { DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp); r = -EINVAL; } err2: - fence_put(f); amdgpu_ib_free(adev, &ib, NULL); fence_put(f); err1: @@ -1160,6 +1163,71 @@ static void gfx_v8_0_get_csb_buffer(struct amdgpu_device *adev, buffer[count++] = cpu_to_le32(0); } +static void cz_init_cp_jump_table(struct amdgpu_device *adev) +{ + const __le32 *fw_data; + volatile u32 *dst_ptr; + int me, i, max_me = 4; + u32 bo_offset = 0; + u32 table_offset, table_size; + + if (adev->asic_type == CHIP_CARRIZO) + max_me = 5; + + /* write the cp table buffer */ + dst_ptr = adev->gfx.rlc.cp_table_ptr; + for (me = 0; me < max_me; me++) { + if (me == 0) { + const struct gfx_firmware_header_v1_0 *hdr = + (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data; + fw_data = (const __le32 *) + (adev->gfx.ce_fw->data + + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + table_offset = le32_to_cpu(hdr->jt_offset); + table_size = le32_to_cpu(hdr->jt_size); + } else if (me == 1) { + const struct gfx_firmware_header_v1_0 *hdr = + (const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data; + fw_data = (const __le32 *) + (adev->gfx.pfp_fw->data + + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + table_offset = le32_to_cpu(hdr->jt_offset); + table_size = le32_to_cpu(hdr->jt_size); + } else if (me == 2) { + const struct gfx_firmware_header_v1_0 *hdr = + (const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data; + fw_data = (const __le32 *) + (adev->gfx.me_fw->data + + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + table_offset = le32_to_cpu(hdr->jt_offset); + table_size = le32_to_cpu(hdr->jt_size); + } else if (me == 3) { + const struct gfx_firmware_header_v1_0 *hdr = + (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; + fw_data = (const __le32 *) + (adev->gfx.mec_fw->data + + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + table_offset = le32_to_cpu(hdr->jt_offset); + table_size = le32_to_cpu(hdr->jt_size); + } else if (me == 4) { + const struct gfx_firmware_header_v1_0 *hdr = + (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data; + fw_data = (const __le32 *) + (adev->gfx.mec2_fw->data + + le32_to_cpu(hdr->header.ucode_array_offset_bytes)); + table_offset = le32_to_cpu(hdr->jt_offset); + table_size = le32_to_cpu(hdr->jt_size); + } + + for (i = 0; i < table_size; i ++) { + dst_ptr[bo_offset + i] = + cpu_to_le32(le32_to_cpu(fw_data[table_offset + i])); + } + + bo_offset += table_size; + } +} + static void gfx_v8_0_rlc_fini(struct amdgpu_device *adev) { int r; @@ -1168,13 +1236,23 @@ static void gfx_v8_0_rlc_fini(struct amdgpu_device *adev) if (adev->gfx.rlc.clear_state_obj) { r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, false); if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve RLC c bo failed\n", r); + 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, false); + 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; + } } static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) @@ -1213,14 +1291,14 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) &adev->gfx.rlc.clear_state_gpu_addr); 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) pin RLC cbs 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 c bo failed\n", r); + dev_warn(adev->dev, "(%d) map RLC cbs bo failed\n", r); gfx_v8_0_rlc_fini(adev); return r; } @@ -1231,6 +1309,45 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev) amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); } + 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, + 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); + 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); + return r; + } + + cz_init_cp_jump_table(adev); + + amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj); + amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); + } + return 0; } @@ -1244,7 +1361,6 @@ static void gfx_v8_0_mec_fini(struct amdgpu_device *adev) 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; } @@ -1612,7 +1728,6 @@ static int gfx_v8_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) RREG32(sec_ded_counter_registers[i]); fail: - fence_put(f); amdgpu_ib_free(adev, &ib, NULL); fence_put(f); @@ -1966,24 +2081,21 @@ static int gfx_v8_0_sw_init(void *handle) } /* reserve GDS, GWS and OA resource for gfx */ - r = amdgpu_bo_create(adev, adev->gds.mem.gfx_partition_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GDS, 0, NULL, - NULL, &adev->gds.gds_gfx_bo); + r = amdgpu_bo_create_kernel(adev, adev->gds.mem.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GDS, + &adev->gds.gds_gfx_bo, NULL, NULL); if (r) return r; - r = amdgpu_bo_create(adev, adev->gds.gws.gfx_partition_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GWS, 0, NULL, - NULL, &adev->gds.gws_gfx_bo); + r = amdgpu_bo_create_kernel(adev, adev->gds.gws.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GWS, + &adev->gds.gws_gfx_bo, NULL, NULL); if (r) return r; - r = amdgpu_bo_create(adev, adev->gds.oa.gfx_partition_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_OA, 0, NULL, - NULL, &adev->gds.oa_gfx_bo); + r = amdgpu_bo_create_kernel(adev, adev->gds.oa.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_OA, + &adev->gds.oa_gfx_bo, NULL, NULL); if (r) return r; @@ -2011,9 +2123,7 @@ static int gfx_v8_0_sw_fini(void *handle) amdgpu_ring_fini(&adev->gfx.compute_ring[i]); gfx_v8_0_mec_fini(adev); - gfx_v8_0_rlc_fini(adev); - gfx_v8_0_free_microcode(adev); return 0; @@ -3339,23 +3449,26 @@ static void gfx_v8_0_tiling_mode_table_init(struct amdgpu_device *adev) } } -void gfx_v8_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num) +static void gfx_v8_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(mmGRBM_GFX_INDEX, data); } @@ -3368,11 +3481,10 @@ static u32 gfx_v8_0_get_rb_active_bitmap(struct amdgpu_device *adev) { u32 data, mask; - data = RREG32(mmCC_RB_BACKEND_DISABLE); - data |= RREG32(mmGC_USER_RB_BACKEND_DISABLE); + data = RREG32(mmCC_RB_BACKEND_DISABLE) | + RREG32(mmGC_USER_RB_BACKEND_DISABLE); - data &= CC_RB_BACKEND_DISABLE__BACKEND_DISABLE_MASK; - data >>= GC_USER_RB_BACKEND_DISABLE__BACKEND_DISABLE__SHIFT; + data = REG_GET_FIELD(data, GC_USER_RB_BACKEND_DISABLE, BACKEND_DISABLE); mask = gfx_v8_0_create_bitmask(adev->gfx.config.max_backends_per_se / adev->gfx.config.max_sh_per_se); @@ -3391,13 +3503,13 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { - gfx_v8_0_select_se_sh(adev, i, j); + gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff); data = gfx_v8_0_get_rb_active_bitmap(adev); active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) * rb_bitmap_width_per_sh); } } - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); adev->gfx.config.backend_enable_mask = active_rbs; @@ -3454,16 +3566,12 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) u32 tmp; int i; - tmp = RREG32(mmGRBM_CNTL); - tmp = REG_SET_FIELD(tmp, GRBM_CNTL, READ_TIMEOUT, 0xff); - WREG32(mmGRBM_CNTL, tmp); - + WREG32_FIELD(GRBM_CNTL, READ_TIMEOUT, 0xFF); WREG32(mmGB_ADDR_CONFIG, adev->gfx.config.gb_addr_config); WREG32(mmHDP_ADDR_CONFIG, adev->gfx.config.gb_addr_config); WREG32(mmDMIF_ADDR_CALC, adev->gfx.config.gb_addr_config); gfx_v8_0_tiling_mode_table_init(adev); - gfx_v8_0_setup_rb(adev); gfx_v8_0_get_cu_info(adev); @@ -3501,7 +3609,7 @@ static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) * making sure that the following register writes will be broadcasted * to all the shaders */ - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); WREG32(mmPA_SC_FIFO_SIZE, (adev->gfx.config.sc_prim_fifo_size_frontend << @@ -3524,7 +3632,7 @@ static void gfx_v8_0_wait_for_rlc_serdes(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { - gfx_v8_0_select_se_sh(adev, i, j); + gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff); for (k = 0; k < adev->usec_timeout; k++) { if (RREG32(mmRLC_SERDES_CU_MASTER_BUSY) == 0) break; @@ -3532,7 +3640,7 @@ static void gfx_v8_0_wait_for_rlc_serdes(struct amdgpu_device *adev) } } } - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK | @@ -3647,9 +3755,7 @@ static int gfx_v8_0_init_save_restore_list(struct amdgpu_device *adev) sizeof(indirect_start_offsets)/sizeof(int)); /* save and restore list */ - temp = RREG32(mmRLC_SRM_CNTL); - temp |= RLC_SRM_CNTL__AUTO_INCR_ADDR_MASK; - WREG32(mmRLC_SRM_CNTL, temp); + WREG32_FIELD(RLC_SRM_CNTL, AUTO_INCR_ADDR, 1); WREG32(mmRLC_SRM_ARAM_ADDR, 0); for (i = 0; i < adev->gfx.rlc.reg_list_size_bytes >> 2; i++) @@ -3686,44 +3792,46 @@ static int gfx_v8_0_init_save_restore_list(struct amdgpu_device *adev) static void gfx_v8_0_enable_save_restore_machine(struct amdgpu_device *adev) { - uint32_t data; - - data = RREG32(mmRLC_SRM_CNTL); - data |= RLC_SRM_CNTL__SRM_ENABLE_MASK; - WREG32(mmRLC_SRM_CNTL, data); + WREG32_FIELD(RLC_SRM_CNTL, SRM_ENABLE, 1); } -static void polaris11_init_power_gating(struct amdgpu_device *adev) +static void gfx_v8_0_init_power_gating(struct amdgpu_device *adev) { uint32_t data; if (adev->pg_flags & (AMD_PG_SUPPORT_GFX_PG | - AMD_PG_SUPPORT_GFX_SMG | - AMD_PG_SUPPORT_GFX_DMG)) { - data = RREG32(mmCP_RB_WPTR_POLL_CNTL); - data &= ~CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT_MASK; - data |= (0x60 << CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT__SHIFT); - WREG32(mmCP_RB_WPTR_POLL_CNTL, data); - - data = 0; - data |= (0x10 << RLC_PG_DELAY__POWER_UP_DELAY__SHIFT); - data |= (0x10 << RLC_PG_DELAY__POWER_DOWN_DELAY__SHIFT); - data |= (0x10 << RLC_PG_DELAY__CMD_PROPAGATE_DELAY__SHIFT); - data |= (0x10 << RLC_PG_DELAY__MEM_SLEEP_DELAY__SHIFT); - WREG32(mmRLC_PG_DELAY, data); + AMD_PG_SUPPORT_GFX_SMG | + AMD_PG_SUPPORT_GFX_DMG)) { + WREG32_FIELD(CP_RB_WPTR_POLL_CNTL, IDLE_POLL_COUNT, 0x60); - data = RREG32(mmRLC_PG_DELAY_2); - data &= ~RLC_PG_DELAY_2__SERDES_CMD_DELAY_MASK; - data |= (0x3 << RLC_PG_DELAY_2__SERDES_CMD_DELAY__SHIFT); - WREG32(mmRLC_PG_DELAY_2, data); + data = REG_SET_FIELD(0, RLC_PG_DELAY, POWER_UP_DELAY, 0x10); + data = REG_SET_FIELD(data, RLC_PG_DELAY, POWER_DOWN_DELAY, 0x10); + data = REG_SET_FIELD(data, RLC_PG_DELAY, CMD_PROPAGATE_DELAY, 0x10); + data = REG_SET_FIELD(data, RLC_PG_DELAY, MEM_SLEEP_DELAY, 0x10); + WREG32(mmRLC_PG_DELAY, data); - data = RREG32(mmRLC_AUTO_PG_CTRL); - data &= ~RLC_AUTO_PG_CTRL__GRBM_REG_SAVE_GFX_IDLE_THRESHOLD_MASK; - data |= (0x55f0 << RLC_AUTO_PG_CTRL__GRBM_REG_SAVE_GFX_IDLE_THRESHOLD__SHIFT); - WREG32(mmRLC_AUTO_PG_CTRL, data); + WREG32_FIELD(RLC_PG_DELAY_2, SERDES_CMD_DELAY, 0x3); + WREG32_FIELD(RLC_AUTO_PG_CTRL, GRBM_REG_SAVE_GFX_IDLE_THRESHOLD, 0x55f0); } } +static void cz_enable_sck_slow_down_on_power_up(struct amdgpu_device *adev, + bool enable) +{ + WREG32_FIELD(RLC_PG_CNTL, SMU_CLK_SLOWDOWN_ON_PU_ENABLE, enable ? 1 : 0); +} + +static void cz_enable_sck_slow_down_on_power_down(struct amdgpu_device *adev, + bool enable) +{ + WREG32_FIELD(RLC_PG_CNTL, SMU_CLK_SLOWDOWN_ON_PD_ENABLE, enable ? 1 : 0); +} + +static void cz_enable_cp_power_gating(struct amdgpu_device *adev, bool enable) +{ + WREG32_FIELD(RLC_PG_CNTL, CP_PG_DISABLE, enable ? 1 : 0); +} + static void gfx_v8_0_init_pg(struct amdgpu_device *adev) { if (adev->pg_flags & (AMD_PG_SUPPORT_GFX_PG | @@ -3736,41 +3844,48 @@ static void gfx_v8_0_init_pg(struct amdgpu_device *adev) gfx_v8_0_init_save_restore_list(adev); gfx_v8_0_enable_save_restore_machine(adev); - if (adev->asic_type == CHIP_POLARIS11) - polaris11_init_power_gating(adev); + if ((adev->asic_type == CHIP_CARRIZO) || + (adev->asic_type == CHIP_STONEY)) { + WREG32(mmRLC_JUMP_TABLE_RESTORE, adev->gfx.rlc.cp_table_gpu_addr >> 8); + gfx_v8_0_init_power_gating(adev); + WREG32(mmRLC_PG_ALWAYS_ON_CU_MASK, adev->gfx.cu_info.ao_cu_mask); + if (adev->pg_flags & AMD_PG_SUPPORT_RLC_SMU_HS) { + cz_enable_sck_slow_down_on_power_up(adev, true); + cz_enable_sck_slow_down_on_power_down(adev, true); + } else { + cz_enable_sck_slow_down_on_power_up(adev, false); + cz_enable_sck_slow_down_on_power_down(adev, false); + } + if (adev->pg_flags & AMD_PG_SUPPORT_CP) + cz_enable_cp_power_gating(adev, true); + else + cz_enable_cp_power_gating(adev, false); + } else if (adev->asic_type == CHIP_POLARIS11) { + gfx_v8_0_init_power_gating(adev); + } } } void gfx_v8_0_rlc_stop(struct amdgpu_device *adev) { - u32 tmp = RREG32(mmRLC_CNTL); - - tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 0); - WREG32(mmRLC_CNTL, tmp); + WREG32_FIELD(RLC_CNTL, RLC_ENABLE_F32, 0); gfx_v8_0_enable_gui_idle_interrupt(adev, false); - gfx_v8_0_wait_for_rlc_serdes(adev); } static void gfx_v8_0_rlc_reset(struct amdgpu_device *adev) { - u32 tmp = RREG32(mmGRBM_SOFT_RESET); - - tmp = REG_SET_FIELD(tmp, GRBM_SOFT_RESET, SOFT_RESET_RLC, 1); - WREG32(mmGRBM_SOFT_RESET, tmp); + WREG32_FIELD(GRBM_SOFT_RESET, SOFT_RESET_RLC, 1); udelay(50); - tmp = REG_SET_FIELD(tmp, GRBM_SOFT_RESET, SOFT_RESET_RLC, 0); - WREG32(mmGRBM_SOFT_RESET, tmp); + + WREG32_FIELD(GRBM_SOFT_RESET, SOFT_RESET_RLC, 0); udelay(50); } static void gfx_v8_0_rlc_start(struct amdgpu_device *adev) { - u32 tmp = RREG32(mmRLC_CNTL); - - tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 1); - WREG32(mmRLC_CNTL, tmp); + WREG32_FIELD(RLC_CNTL, RLC_ENABLE_F32, 1); /* carrizo do enable cp interrupt after cp inited */ if (!(adev->flags & AMD_IS_APU)) @@ -3812,14 +3927,13 @@ static int gfx_v8_0_rlc_resume(struct amdgpu_device *adev) /* disable CG */ WREG32(mmRLC_CGCG_CGLS_CTRL, 0); if (adev->asic_type == CHIP_POLARIS11 || - adev->asic_type == CHIP_POLARIS10) + adev->asic_type == CHIP_POLARIS10) WREG32(mmRLC_CGCG_CGLS_CTRL_3D, 0); /* disable PG */ WREG32(mmRLC_PG_CNTL, 0); gfx_v8_0_rlc_reset(adev); - gfx_v8_0_init_pg(adev); if (!adev->pp_enabled) { @@ -4114,12 +4228,10 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev) gfx_v8_0_cp_gfx_start(adev); ring->ready = true; r = amdgpu_ring_test_ring(ring); - if (r) { + if (r) ring->ready = false; - return r; - } - return 0; + return r; } static void gfx_v8_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) @@ -4794,7 +4906,6 @@ static int gfx_v8_0_hw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; gfx_v8_0_init_golden_registers(adev); - gfx_v8_0_gpu_init(adev); r = gfx_v8_0_rlc_resume(adev); @@ -4802,8 +4913,6 @@ static int gfx_v8_0_hw_init(void *handle) return r; r = gfx_v8_0_cp_resume(adev); - if (r) - return r; return r; } @@ -4851,25 +4960,22 @@ static bool gfx_v8_0_is_idle(void *handle) static int gfx_v8_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(mmGRBM_STATUS) & GRBM_STATUS__GUI_ACTIVE_MASK; - - if (!REG_GET_FIELD(tmp, GRBM_STATUS, GUI_ACTIVE)) + if (gfx_v8_0_is_idle(handle)) return 0; + udelay(1); } return -ETIMEDOUT; } -static int gfx_v8_0_soft_reset(void *handle) +static int gfx_v8_0_check_soft_reset(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; u32 grbm_soft_reset = 0, srbm_soft_reset = 0; u32 tmp; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* GRBM_STATUS */ tmp = RREG32(mmGRBM_STATUS); @@ -4878,16 +4984,12 @@ static int gfx_v8_0_soft_reset(void *handle) GRBM_STATUS__TA_BUSY_MASK | GRBM_STATUS__VGT_BUSY_MASK | GRBM_STATUS__DB_BUSY_MASK | GRBM_STATUS__CB_BUSY_MASK | GRBM_STATUS__GDS_BUSY_MASK | GRBM_STATUS__SPI_BUSY_MASK | - GRBM_STATUS__IA_BUSY_MASK | GRBM_STATUS__IA_BUSY_NO_DMA_MASK)) { + GRBM_STATUS__IA_BUSY_MASK | GRBM_STATUS__IA_BUSY_NO_DMA_MASK | + GRBM_STATUS__CP_BUSY_MASK | GRBM_STATUS__CP_COHERENCY_BUSY_MASK)) { grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP, 1); grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX, 1); - } - - if (tmp & (GRBM_STATUS__CP_BUSY_MASK | GRBM_STATUS__CP_COHERENCY_BUSY_MASK)) { - grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, - GRBM_SOFT_RESET, SOFT_RESET_CP, 1); srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_GRBM, 1); } @@ -4898,73 +5000,199 @@ static int gfx_v8_0_soft_reset(void *handle) grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_RLC, 1); + if (REG_GET_FIELD(tmp, GRBM_STATUS2, CPF_BUSY) || + REG_GET_FIELD(tmp, GRBM_STATUS2, CPC_BUSY) || + REG_GET_FIELD(tmp, GRBM_STATUS2, CPG_BUSY)) { + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CPF, 1); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CPC, 1); + grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, + SOFT_RESET_CPG, 1); + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, + SOFT_RESET_GRBM, 1); + } + /* SRBM_STATUS */ tmp = RREG32(mmSRBM_STATUS); if (REG_GET_FIELD(tmp, SRBM_STATUS, GRBM_RQ_PENDING)) srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_GRBM, 1); + if (REG_GET_FIELD(tmp, SRBM_STATUS, SEM_BUSY)) + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, + SRBM_SOFT_RESET, SOFT_RESET_SEM, 1); if (grbm_soft_reset || srbm_soft_reset) { - /* stop the rlc */ - gfx_v8_0_rlc_stop(adev); + adev->ip_block_status[AMD_IP_BLOCK_TYPE_GFX].hang = true; + adev->gfx.grbm_soft_reset = grbm_soft_reset; + adev->gfx.srbm_soft_reset = srbm_soft_reset; + } else { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_GFX].hang = false; + adev->gfx.grbm_soft_reset = 0; + adev->gfx.srbm_soft_reset = 0; + } + + return 0; +} + +static void gfx_v8_0_inactive_hqd(struct amdgpu_device *adev, + struct amdgpu_ring *ring) +{ + int i; + + vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + if (RREG32(mmCP_HQD_ACTIVE) & CP_HQD_ACTIVE__ACTIVE_MASK) { + u32 tmp; + tmp = RREG32(mmCP_HQD_DEQUEUE_REQUEST); + tmp = REG_SET_FIELD(tmp, CP_HQD_DEQUEUE_REQUEST, + DEQUEUE_REQ, 2); + WREG32(mmCP_HQD_DEQUEUE_REQUEST, tmp); + for (i = 0; i < adev->usec_timeout; i++) { + if (!(RREG32(mmCP_HQD_ACTIVE) & CP_HQD_ACTIVE__ACTIVE_MASK)) + break; + udelay(1); + } + } +} + +static int gfx_v8_0_pre_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_GFX].hang) + return 0; + + grbm_soft_reset = adev->gfx.grbm_soft_reset; + srbm_soft_reset = adev->gfx.srbm_soft_reset; + /* stop the rlc */ + gfx_v8_0_rlc_stop(adev); + + if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX)) /* Disable GFX parsing/prefetching */ gfx_v8_0_cp_gfx_enable(adev, false); + if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPF) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPC) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPG)) { + int i; + + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + struct amdgpu_ring *ring = &adev->gfx.compute_ring[i]; + + gfx_v8_0_inactive_hqd(adev, ring); + } /* Disable MEC parsing/prefetching */ gfx_v8_0_cp_compute_enable(adev, false); + } - if (grbm_soft_reset || srbm_soft_reset) { - tmp = RREG32(mmGMCON_DEBUG); - tmp = REG_SET_FIELD(tmp, - GMCON_DEBUG, GFX_STALL, 1); - tmp = REG_SET_FIELD(tmp, - GMCON_DEBUG, GFX_CLEAR, 1); - WREG32(mmGMCON_DEBUG, tmp); + return 0; +} - udelay(50); - } +static int gfx_v8_0_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + u32 tmp; - if (grbm_soft_reset) { - tmp = RREG32(mmGRBM_SOFT_RESET); - tmp |= grbm_soft_reset; - dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp); - WREG32(mmGRBM_SOFT_RESET, tmp); - tmp = RREG32(mmGRBM_SOFT_RESET); + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_GFX].hang) + return 0; - udelay(50); + grbm_soft_reset = adev->gfx.grbm_soft_reset; + srbm_soft_reset = adev->gfx.srbm_soft_reset; - tmp &= ~grbm_soft_reset; - WREG32(mmGRBM_SOFT_RESET, tmp); - tmp = RREG32(mmGRBM_SOFT_RESET); - } + if (grbm_soft_reset || srbm_soft_reset) { + tmp = RREG32(mmGMCON_DEBUG); + tmp = REG_SET_FIELD(tmp, GMCON_DEBUG, GFX_STALL, 1); + tmp = REG_SET_FIELD(tmp, GMCON_DEBUG, GFX_CLEAR, 1); + WREG32(mmGMCON_DEBUG, tmp); + udelay(50); + } - if (srbm_soft_reset) { - tmp = RREG32(mmSRBM_SOFT_RESET); - tmp |= srbm_soft_reset; - dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); - WREG32(mmSRBM_SOFT_RESET, tmp); - tmp = RREG32(mmSRBM_SOFT_RESET); + if (grbm_soft_reset) { + tmp = RREG32(mmGRBM_SOFT_RESET); + tmp |= grbm_soft_reset; + dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp); + WREG32(mmGRBM_SOFT_RESET, tmp); + tmp = RREG32(mmGRBM_SOFT_RESET); - udelay(50); + udelay(50); - tmp &= ~srbm_soft_reset; - WREG32(mmSRBM_SOFT_RESET, tmp); - tmp = RREG32(mmSRBM_SOFT_RESET); - } + tmp &= ~grbm_soft_reset; + WREG32(mmGRBM_SOFT_RESET, tmp); + tmp = RREG32(mmGRBM_SOFT_RESET); + } - if (grbm_soft_reset || srbm_soft_reset) { - tmp = RREG32(mmGMCON_DEBUG); - tmp = REG_SET_FIELD(tmp, - GMCON_DEBUG, GFX_STALL, 0); - tmp = REG_SET_FIELD(tmp, - GMCON_DEBUG, GFX_CLEAR, 0); - WREG32(mmGMCON_DEBUG, tmp); - } + if (srbm_soft_reset) { + tmp = RREG32(mmSRBM_SOFT_RESET); + tmp |= srbm_soft_reset; + dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); + WREG32(mmSRBM_SOFT_RESET, tmp); + tmp = RREG32(mmSRBM_SOFT_RESET); - /* Wait a little for things to settle down */ udelay(50); + + tmp &= ~srbm_soft_reset; + WREG32(mmSRBM_SOFT_RESET, tmp); + tmp = RREG32(mmSRBM_SOFT_RESET); + } + + if (grbm_soft_reset || srbm_soft_reset) { + tmp = RREG32(mmGMCON_DEBUG); + tmp = REG_SET_FIELD(tmp, GMCON_DEBUG, GFX_STALL, 0); + tmp = REG_SET_FIELD(tmp, GMCON_DEBUG, GFX_CLEAR, 0); + WREG32(mmGMCON_DEBUG, tmp); + } + + /* Wait a little for things to settle down */ + udelay(50); + + return 0; +} + +static void gfx_v8_0_init_hqd(struct amdgpu_device *adev, + struct amdgpu_ring *ring) +{ + vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0); + WREG32(mmCP_HQD_DEQUEUE_REQUEST, 0); + WREG32(mmCP_HQD_PQ_RPTR, 0); + WREG32(mmCP_HQD_PQ_WPTR, 0); + vi_srbm_select(adev, 0, 0, 0, 0); +} + +static int gfx_v8_0_post_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_GFX].hang) + return 0; + + grbm_soft_reset = adev->gfx.grbm_soft_reset; + srbm_soft_reset = adev->gfx.srbm_soft_reset; + + if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX)) + gfx_v8_0_cp_gfx_resume(adev); + + if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPF) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPC) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPG)) { + int i; + + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + struct amdgpu_ring *ring = &adev->gfx.compute_ring[i]; + + gfx_v8_0_init_hqd(adev, ring); + } + gfx_v8_0_cp_compute_resume(adev); } + gfx_v8_0_rlc_start(adev); + return 0; } @@ -4976,7 +5204,7 @@ static int gfx_v8_0_soft_reset(void *handle) * Fetches a GPU clock counter snapshot. * Returns the 64 bit clock counter snapshot. */ -uint64_t gfx_v8_0_get_gpu_clock_counter(struct amdgpu_device *adev) +static uint64_t gfx_v8_0_get_gpu_clock_counter(struct amdgpu_device *adev) { uint64_t clock; @@ -5036,12 +5264,18 @@ static void gfx_v8_0_ring_emit_gds_switch(struct amdgpu_ring *ring, amdgpu_ring_write(ring, (1 << (oa_size + oa_base)) - (1 << oa_base)); } +static const struct amdgpu_gfx_funcs gfx_v8_0_gfx_funcs = { + .get_gpu_clock_counter = &gfx_v8_0_get_gpu_clock_counter, + .select_se_sh = &gfx_v8_0_select_se_sh, +}; + static int gfx_v8_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; adev->gfx.num_gfx_rings = GFX8_NUM_GFX_RINGS; adev->gfx.num_compute_rings = GFX8_NUM_COMPUTE_RINGS; + adev->gfx.funcs = &gfx_v8_0_gfx_funcs; gfx_v8_0_set_ring_funcs(adev); gfx_v8_0_set_irq_funcs(adev); gfx_v8_0_set_gds_init(adev); @@ -5074,71 +5308,57 @@ static int gfx_v8_0_late_init(void *handle) return 0; } -static void polaris11_enable_gfx_static_mg_power_gating(struct amdgpu_device *adev, - bool enable) +static void gfx_v8_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *adev, + bool enable) { - uint32_t data, temp; - - /* Send msg to SMU via Powerplay */ - amdgpu_set_powergating_state(adev, - AMD_IP_BLOCK_TYPE_SMC, - enable ? AMD_PG_STATE_GATE : AMD_PG_STATE_UNGATE); - - if (enable) { - /* Enable static MGPG */ - temp = data = RREG32(mmRLC_PG_CNTL); - data |= RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK; - - if (temp != data) - WREG32(mmRLC_PG_CNTL, data); - } else { - temp = data = RREG32(mmRLC_PG_CNTL); - data &= ~RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK; + if (adev->asic_type == CHIP_POLARIS11) + /* Send msg to SMU via Powerplay */ + amdgpu_set_powergating_state(adev, + AMD_IP_BLOCK_TYPE_SMC, + enable ? + AMD_PG_STATE_GATE : AMD_PG_STATE_UNGATE); + + WREG32_FIELD(RLC_PG_CNTL, STATIC_PER_CU_PG_ENABLE, enable ? 1 : 0); +} - if (temp != data) - WREG32(mmRLC_PG_CNTL, data); - } +static void gfx_v8_0_enable_gfx_dynamic_mg_power_gating(struct amdgpu_device *adev, + bool enable) +{ + WREG32_FIELD(RLC_PG_CNTL, DYN_PER_CU_PG_ENABLE, enable ? 1 : 0); } -static void polaris11_enable_gfx_dynamic_mg_power_gating(struct amdgpu_device *adev, +static void polaris11_enable_gfx_quick_mg_power_gating(struct amdgpu_device *adev, bool enable) { - uint32_t data, temp; - - if (enable) { - /* Enable dynamic MGPG */ - temp = data = RREG32(mmRLC_PG_CNTL); - data |= RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK; - - if (temp != data) - WREG32(mmRLC_PG_CNTL, data); - } else { - temp = data = RREG32(mmRLC_PG_CNTL); - data &= ~RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK; + WREG32_FIELD(RLC_PG_CNTL, QUICK_PG_ENABLE, enable ? 1 : 0); +} - if (temp != data) - WREG32(mmRLC_PG_CNTL, data); - } +static void cz_enable_gfx_cg_power_gating(struct amdgpu_device *adev, + bool enable) +{ + WREG32_FIELD(RLC_PG_CNTL, GFX_POWER_GATING_ENABLE, enable ? 1 : 0); } -static void polaris11_enable_gfx_quick_mg_power_gating(struct amdgpu_device *adev, - bool enable) +static void cz_enable_gfx_pipeline_power_gating(struct amdgpu_device *adev, + bool enable) { - uint32_t data, temp; + WREG32_FIELD(RLC_PG_CNTL, GFX_PIPELINE_PG_ENABLE, enable ? 1 : 0); - if (enable) { - /* Enable quick PG */ - temp = data = RREG32(mmRLC_PG_CNTL); - data |= 0x100000; + /* Read any GFX register to wake up GFX. */ + if (!enable) + RREG32(mmDB_RENDER_CONTROL); +} - if (temp != data) - WREG32(mmRLC_PG_CNTL, data); +static void cz_update_gfx_cg_power_gating(struct amdgpu_device *adev, + bool enable) +{ + if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_PG) && enable) { + cz_enable_gfx_cg_power_gating(adev, true); + if (adev->pg_flags & AMD_PG_SUPPORT_GFX_PIPELINE) + cz_enable_gfx_pipeline_power_gating(adev, true); } else { - temp = data = RREG32(mmRLC_PG_CNTL); - data &= ~0x100000; - - if (temp != data) - WREG32(mmRLC_PG_CNTL, data); + cz_enable_gfx_cg_power_gating(adev, false); + cz_enable_gfx_pipeline_power_gating(adev, false); } } @@ -5146,21 +5366,42 @@ static int gfx_v8_0_set_powergating_state(void *handle, enum amd_powergating_state state) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + bool enable = (state == AMD_PG_STATE_GATE) ? true : false; if (!(adev->pg_flags & AMD_PG_SUPPORT_GFX_PG)) return 0; switch (adev->asic_type) { + case CHIP_CARRIZO: + case CHIP_STONEY: + if (adev->pg_flags & AMD_PG_SUPPORT_GFX_PG) + cz_update_gfx_cg_power_gating(adev, enable); + + if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_SMG) && enable) + gfx_v8_0_enable_gfx_static_mg_power_gating(adev, true); + else + gfx_v8_0_enable_gfx_static_mg_power_gating(adev, false); + + if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_DMG) && enable) + gfx_v8_0_enable_gfx_dynamic_mg_power_gating(adev, true); + else + gfx_v8_0_enable_gfx_dynamic_mg_power_gating(adev, false); + break; case CHIP_POLARIS11: - if (adev->pg_flags & AMD_PG_SUPPORT_GFX_SMG) - polaris11_enable_gfx_static_mg_power_gating(adev, - state == AMD_PG_STATE_GATE ? true : false); - else if (adev->pg_flags & AMD_PG_SUPPORT_GFX_DMG) - polaris11_enable_gfx_dynamic_mg_power_gating(adev, - state == AMD_PG_STATE_GATE ? true : false); + if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_SMG) && enable) + gfx_v8_0_enable_gfx_static_mg_power_gating(adev, true); + else + gfx_v8_0_enable_gfx_static_mg_power_gating(adev, false); + + if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_DMG) && enable) + gfx_v8_0_enable_gfx_dynamic_mg_power_gating(adev, true); + else + gfx_v8_0_enable_gfx_dynamic_mg_power_gating(adev, false); + + if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_QUICK_MG) && enable) + polaris11_enable_gfx_quick_mg_power_gating(adev, true); else - polaris11_enable_gfx_quick_mg_power_gating(adev, - state == AMD_PG_STATE_GATE ? true : false); + polaris11_enable_gfx_quick_mg_power_gating(adev, false); break; default: break; @@ -5174,22 +5415,22 @@ static void gfx_v8_0_send_serdes_cmd(struct amdgpu_device *adev, { uint32_t data; - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff); WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff); data = RREG32(mmRLC_SERDES_WR_CTRL); if (adev->asic_type == CHIP_STONEY) - data &= ~(RLC_SERDES_WR_CTRL__WRITE_COMMAND_MASK | - RLC_SERDES_WR_CTRL__READ_COMMAND_MASK | - RLC_SERDES_WR_CTRL__P1_SELECT_MASK | - RLC_SERDES_WR_CTRL__P2_SELECT_MASK | - RLC_SERDES_WR_CTRL__RDDATA_RESET_MASK | - RLC_SERDES_WR_CTRL__POWER_DOWN_MASK | - RLC_SERDES_WR_CTRL__POWER_UP_MASK | - RLC_SERDES_WR_CTRL__SHORT_FORMAT_MASK | - RLC_SERDES_WR_CTRL__SRBM_OVERRIDE_MASK); + data &= ~(RLC_SERDES_WR_CTRL__WRITE_COMMAND_MASK | + RLC_SERDES_WR_CTRL__READ_COMMAND_MASK | + RLC_SERDES_WR_CTRL__P1_SELECT_MASK | + RLC_SERDES_WR_CTRL__P2_SELECT_MASK | + RLC_SERDES_WR_CTRL__RDDATA_RESET_MASK | + RLC_SERDES_WR_CTRL__POWER_DOWN_MASK | + RLC_SERDES_WR_CTRL__POWER_UP_MASK | + RLC_SERDES_WR_CTRL__SHORT_FORMAT_MASK | + RLC_SERDES_WR_CTRL__SRBM_OVERRIDE_MASK); else data &= ~(RLC_SERDES_WR_CTRL__WRITE_COMMAND_MASK | RLC_SERDES_WR_CTRL__READ_COMMAND_MASK | @@ -5212,10 +5453,10 @@ static void gfx_v8_0_send_serdes_cmd(struct amdgpu_device *adev, #define MSG_ENTER_RLC_SAFE_MODE 1 #define MSG_EXIT_RLC_SAFE_MODE 0 - -#define RLC_GPR_REG2__REQ_MASK 0x00000001 -#define RLC_GPR_REG2__MESSAGE__SHIFT 0x00000001 -#define RLC_GPR_REG2__MESSAGE_MASK 0x0000001e +#define RLC_GPR_REG2__REQ_MASK 0x00000001 +#define RLC_GPR_REG2__REQ__SHIFT 0 +#define RLC_GPR_REG2__MESSAGE__SHIFT 0x00000001 +#define RLC_GPR_REG2__MESSAGE_MASK 0x0000001e static void cz_enter_rlc_safe_mode(struct amdgpu_device *adev) { @@ -5245,7 +5486,7 @@ static void cz_enter_rlc_safe_mode(struct amdgpu_device *adev) } for (i = 0; i < adev->usec_timeout; i++) { - if ((RREG32(mmRLC_GPR_REG2) & RLC_GPR_REG2__REQ_MASK) == 0) + if (!REG_GET_FIELD(RREG32(mmRLC_GPR_REG2), RLC_GPR_REG2, REQ)) break; udelay(1); } @@ -5273,7 +5514,7 @@ static void cz_exit_rlc_safe_mode(struct amdgpu_device *adev) } for (i = 0; i < adev->usec_timeout; i++) { - if ((RREG32(mmRLC_GPR_REG2) & RLC_GPR_REG2__REQ_MASK) == 0) + if (!REG_GET_FIELD(RREG32(mmRLC_GPR_REG2), RLC_GPR_REG2, REQ)) break; udelay(1); } @@ -5305,7 +5546,7 @@ static void iceland_enter_rlc_safe_mode(struct amdgpu_device *adev) } for (i = 0; i < adev->usec_timeout; i++) { - if ((RREG32(mmRLC_SAFE_MODE) & RLC_SAFE_MODE__CMD_MASK) == 0) + if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) break; udelay(1); } @@ -5332,7 +5573,7 @@ static void iceland_exit_rlc_safe_mode(struct amdgpu_device *adev) } for (i = 0; i < adev->usec_timeout; i++) { - if ((RREG32(mmRLC_SAFE_MODE) & RLC_SAFE_MODE__CMD_MASK) == 0) + if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) break; udelay(1); } @@ -5373,21 +5614,12 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev /* It is disabled by HW by default */ if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) { if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGLS) { - if (adev->cg_flags & AMD_CG_SUPPORT_GFX_RLC_LS) { + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_RLC_LS) /* 1 - RLC memory Light sleep */ - temp = data = RREG32(mmRLC_MEM_SLP_CNTL); - data |= RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK; - if (temp != data) - WREG32(mmRLC_MEM_SLP_CNTL, data); - } + WREG32_FIELD(RLC_MEM_SLP_CNTL, RLC_MEM_LS_EN, 1); - if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CP_LS) { - /* 2 - CP memory Light sleep */ - temp = data = RREG32(mmCP_MEM_SLP_CNTL); - data |= CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK; - if (temp != data) - WREG32(mmCP_MEM_SLP_CNTL, data); - } + if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CP_LS) + WREG32_FIELD(CP_MEM_SLP_CNTL, CP_MEM_LS_EN, 1); } /* 3 - RLC_CGTT_MGCG_OVERRIDE */ @@ -5562,6 +5794,8 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev WREG32(mmRLC_CGCG_CGLS_CTRL, data); } + gfx_v8_0_wait_for_rlc_serdes(adev); + adev->gfx.rlc.funcs->exit_safe_mode(adev); } static int gfx_v8_0_update_gfx_clock_gating(struct amdgpu_device *adev, @@ -5603,25 +5837,18 @@ static int gfx_v8_0_set_clockgating_state(void *handle, static u32 gfx_v8_0_ring_get_rptr_gfx(struct amdgpu_ring *ring) { - u32 rptr; - - rptr = ring->adev->wb.wb[ring->rptr_offs]; - - return rptr; + return ring->adev->wb.wb[ring->rptr_offs]; } static u32 gfx_v8_0_ring_get_wptr_gfx(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - u32 wptr; if (ring->use_doorbell) /* XXX check if swapping is necessary on BE */ - wptr = ring->adev->wb.wb[ring->wptr_offs]; + return ring->adev->wb.wb[ring->wptr_offs]; else - wptr = RREG32(mmCP_RB0_WPTR); - - return wptr; + return RREG32(mmCP_RB0_WPTR); } static void gfx_v8_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) @@ -5687,17 +5914,6 @@ static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, unsigned vm_id, bool ctx_switch) { u32 header, control = 0; - u32 next_rptr = ring->wptr + 5; - - if (ctx_switch) - next_rptr += 2; - - next_rptr += 4; - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM); - amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); - amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); - amdgpu_ring_write(ring, next_rptr); /* insert SWITCH_BUFFER packet before first IB in the ring frame */ if (ctx_switch) { @@ -5726,28 +5942,14 @@ static void gfx_v8_0_ring_emit_ib_compute(struct amdgpu_ring *ring, struct amdgpu_ib *ib, unsigned vm_id, bool ctx_switch) { - u32 header, control = 0; - u32 next_rptr = ring->wptr + 5; - - control |= INDIRECT_BUFFER_VALID; - - next_rptr += 4; - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM); - amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); - amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); - amdgpu_ring_write(ring, next_rptr); - - header = PACKET3(PACKET3_INDIRECT_BUFFER, 2); - - control |= ib->length_dw | (vm_id << 24); + u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vm_id << 24); - amdgpu_ring_write(ring, header); + amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); amdgpu_ring_write(ring, #ifdef __BIG_ENDIAN - (2 << 0) | + (2 << 0) | #endif - (ib->gpu_addr & 0xFFFFFFFC)); + (ib->gpu_addr & 0xFFFFFFFC)); amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF); amdgpu_ring_write(ring, control); } @@ -5892,33 +6094,14 @@ static void gfx_v8_0_ring_emit_fence_compute(struct amdgpu_ring *ring, static void gfx_v8_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev, enum amdgpu_interrupt_state state) { - u32 cp_int_cntl; - - switch (state) { - case AMDGPU_IRQ_STATE_DISABLE: - cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0); - cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, - TIME_STAMP_INT_ENABLE, 0); - WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl); - break; - case AMDGPU_IRQ_STATE_ENABLE: - cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0); - cp_int_cntl = - REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, - TIME_STAMP_INT_ENABLE, 1); - WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl); - break; - default: - break; - } + WREG32_FIELD(CP_INT_CNTL_RING0, TIME_STAMP_INT_ENABLE, + state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); } static void gfx_v8_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev, int me, int pipe, enum amdgpu_interrupt_state state) { - u32 mec_int_cntl, mec_int_cntl_reg; - /* * amdgpu controls only pipe 0 of MEC1. That's why this function only * handles the setting of interrupts for this specific pipe. All other @@ -5928,7 +6111,6 @@ static void gfx_v8_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev, if (me == 1) { switch (pipe) { case 0: - mec_int_cntl_reg = mmCP_ME1_PIPE0_INT_CNTL; break; default: DRM_DEBUG("invalid pipe %d\n", pipe); @@ -5939,22 +6121,8 @@ static void gfx_v8_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev, return; } - switch (state) { - case AMDGPU_IRQ_STATE_DISABLE: - mec_int_cntl = RREG32(mec_int_cntl_reg); - mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL, - TIME_STAMP_INT_ENABLE, 0); - WREG32(mec_int_cntl_reg, mec_int_cntl); - break; - case AMDGPU_IRQ_STATE_ENABLE: - mec_int_cntl = RREG32(mec_int_cntl_reg); - mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL, - TIME_STAMP_INT_ENABLE, 1); - WREG32(mec_int_cntl_reg, mec_int_cntl); - break; - default: - break; - } + WREG32_FIELD(CP_ME1_PIPE0_INT_CNTL, TIME_STAMP_INT_ENABLE, + state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); } static int gfx_v8_0_set_priv_reg_fault_state(struct amdgpu_device *adev, @@ -5962,24 +6130,8 @@ static int gfx_v8_0_set_priv_reg_fault_state(struct amdgpu_device *adev, unsigned type, enum amdgpu_interrupt_state state) { - u32 cp_int_cntl; - - switch (state) { - case AMDGPU_IRQ_STATE_DISABLE: - cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0); - cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, - PRIV_REG_INT_ENABLE, 0); - WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl); - break; - case AMDGPU_IRQ_STATE_ENABLE: - cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0); - cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, - PRIV_REG_INT_ENABLE, 1); - WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl); - break; - default: - break; - } + WREG32_FIELD(CP_INT_CNTL_RING0, PRIV_REG_INT_ENABLE, + state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); return 0; } @@ -5989,24 +6141,8 @@ static int gfx_v8_0_set_priv_inst_fault_state(struct amdgpu_device *adev, unsigned type, enum amdgpu_interrupt_state state) { - u32 cp_int_cntl; - - switch (state) { - case AMDGPU_IRQ_STATE_DISABLE: - cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0); - cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, - PRIV_INSTR_INT_ENABLE, 0); - WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl); - break; - case AMDGPU_IRQ_STATE_ENABLE: - cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0); - cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0, - PRIV_INSTR_INT_ENABLE, 1); - WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl); - break; - default: - break; - } + WREG32_FIELD(CP_INT_CNTL_RING0, PRIV_INSTR_INT_ENABLE, + state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); return 0; } @@ -6112,7 +6248,10 @@ const struct amd_ip_funcs gfx_v8_0_ip_funcs = { .resume = gfx_v8_0_resume, .is_idle = gfx_v8_0_is_idle, .wait_for_idle = gfx_v8_0_wait_for_idle, + .check_soft_reset = gfx_v8_0_check_soft_reset, + .pre_soft_reset = gfx_v8_0_pre_soft_reset, .soft_reset = gfx_v8_0_soft_reset, + .post_soft_reset = gfx_v8_0_post_soft_reset, .set_clockgating_state = gfx_v8_0_set_clockgating_state, .set_powergating_state = gfx_v8_0_set_powergating_state, }; @@ -6195,9 +6334,9 @@ static void gfx_v8_0_set_rlc_funcs(struct amdgpu_device *adev) { switch (adev->asic_type) { case CHIP_TOPAZ: - case CHIP_STONEY: adev->gfx.rlc.funcs = &iceland_rlc_funcs; break; + case CHIP_STONEY: case CHIP_CARRIZO: adev->gfx.rlc.funcs = &cz_rlc_funcs; break; @@ -6235,19 +6374,30 @@ static void gfx_v8_0_set_gds_init(struct amdgpu_device *adev) } } +static void gfx_v8_0_set_user_cu_inactive_bitmap(struct amdgpu_device *adev, + u32 bitmap) +{ + u32 data; + + if (!bitmap) + return; + + data = bitmap << GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT; + data &= GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK; + + WREG32(mmGC_USER_SHADER_ARRAY_CONFIG, data); +} + static u32 gfx_v8_0_get_cu_active_bitmap(struct amdgpu_device *adev) { u32 data, mask; - data = RREG32(mmCC_GC_SHADER_ARRAY_CONFIG); - data |= RREG32(mmGC_USER_SHADER_ARRAY_CONFIG); - - data &= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK; - data >>= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT; + data = RREG32(mmCC_GC_SHADER_ARRAY_CONFIG) | + RREG32(mmGC_USER_SHADER_ARRAY_CONFIG); mask = gfx_v8_0_create_bitmask(adev->gfx.config.max_cu_per_sh); - return (~data) & mask; + return ~REG_GET_FIELD(data, CC_GC_SHADER_ARRAY_CONFIG, INACTIVE_CUS) & mask; } static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev) @@ -6255,16 +6405,22 @@ static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev) int i, j, k, counter, active_cu_number = 0; u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0; struct amdgpu_cu_info *cu_info = &adev->gfx.cu_info; + unsigned disable_masks[4 * 2]; memset(cu_info, 0, sizeof(*cu_info)); + amdgpu_gfx_parse_disable_cu(disable_masks, 4, 2); + mutex_lock(&adev->grbm_idx_mutex); for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { mask = 1; ao_bitmap = 0; counter = 0; - gfx_v8_0_select_se_sh(adev, i, j); + gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff); + if (i < 4 && j < 2) + gfx_v8_0_set_user_cu_inactive_bitmap( + adev, disable_masks[i * 2 + j]); bitmap = gfx_v8_0_get_cu_active_bitmap(adev); cu_info->bitmap[i][j] = bitmap; @@ -6280,7 +6436,7 @@ static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev) ao_cu_mask |= (ao_bitmap << (i * 16 + j * 8)); } } - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); cu_info->number = active_cu_number; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h index 16a49f53a2fa..ebed1f829297 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.h @@ -26,7 +26,4 @@ extern const struct amd_ip_funcs gfx_v8_0_ip_funcs; -uint64_t gfx_v8_0_get_gpu_clock_counter(struct amdgpu_device *adev); -void gfx_v8_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num); - #endif diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index 1feb6439cb0b..aa0c4b964621 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -39,6 +39,7 @@ static void gmc_v7_0_set_gart_funcs(struct amdgpu_device *adev); static void gmc_v7_0_set_irq_funcs(struct amdgpu_device *adev); +static int gmc_v7_0_wait_for_idle(void *handle); MODULE_FIRMWARE("radeon/bonaire_mc.bin"); MODULE_FIRMWARE("radeon/hawaii_mc.bin"); @@ -73,39 +74,15 @@ static void gmc_v7_0_init_golden_registers(struct amdgpu_device *adev) } } -/** - * gmc7_mc_wait_for_idle - wait for MC idle callback. - * - * @adev: amdgpu_device pointer - * - * Wait for the MC (memory controller) to be idle. - * (evergreen+). - * Returns 0 if the MC is idle, -1 if not. - */ -int gmc_v7_0_mc_wait_for_idle(struct amdgpu_device *adev) -{ - unsigned i; - u32 tmp; - - for (i = 0; i < adev->usec_timeout; i++) { - /* read MC_STATUS */ - tmp = RREG32(mmSRBM_STATUS) & 0x1F00; - if (!tmp) - return 0; - udelay(1); - } - return -1; -} - -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, + struct amdgpu_mode_mc_save *save) { u32 blackout; if (adev->mode_info.num_crtc) amdgpu_display_stop_mc_access(adev, save); - amdgpu_asic_wait_for_mc_idle(adev); + gmc_v7_0_wait_for_idle((void *)adev); blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL); if (REG_GET_FIELD(blackout, MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE) != 1) { @@ -120,8 +97,8 @@ void gmc_v7_0_mc_stop(struct amdgpu_device *adev, udelay(100); } -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, + struct amdgpu_mode_mc_save *save) { u32 tmp; @@ -167,6 +144,7 @@ static int gmc_v7_0_init_microcode(struct amdgpu_device *adev) break; case CHIP_KAVERI: case CHIP_KABINI: + case CHIP_MULLINS: return 0; default: BUG(); } @@ -205,7 +183,7 @@ static int gmc_v7_0_mc_load_microcode(struct amdgpu_device *adev) const struct mc_firmware_header_v1_0 *hdr; const __le32 *fw_data = NULL; const __le32 *io_mc_regs = NULL; - u32 running, blackout = 0; + u32 running; int i, ucode_size, regs_size; if (!adev->mc.fw) @@ -225,11 +203,6 @@ static int gmc_v7_0_mc_load_microcode(struct amdgpu_device *adev) running = REG_GET_FIELD(RREG32(mmMC_SEQ_SUP_CNTL), MC_SEQ_SUP_CNTL, RUN); if (running == 0) { - if (running) { - blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL); - WREG32(mmMC_SHARED_BLACKOUT_CNTL, blackout | 1); - } - /* reset the engine and set to writable */ WREG32(mmMC_SEQ_SUP_CNTL, 0x00000008); WREG32(mmMC_SEQ_SUP_CNTL, 0x00000010); @@ -261,9 +234,6 @@ static int gmc_v7_0_mc_load_microcode(struct amdgpu_device *adev) break; udelay(1); } - - if (running) - WREG32(mmMC_SHARED_BLACKOUT_CNTL, blackout); } return 0; @@ -311,7 +281,7 @@ static void gmc_v7_0_mc_program(struct amdgpu_device *adev) amdgpu_display_set_vga_render_state(adev, false); gmc_v7_0_mc_stop(adev, &save); - if (amdgpu_asic_wait_for_mc_idle(adev)) { + if (gmc_v7_0_wait_for_idle((void *)adev)) { dev_warn(adev->dev, "Wait for MC idle timedout !\n"); } /* Update configuration */ @@ -331,7 +301,7 @@ static void gmc_v7_0_mc_program(struct amdgpu_device *adev) WREG32(mmMC_VM_AGP_BASE, 0); WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF); WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF); - if (amdgpu_asic_wait_for_mc_idle(adev)) { + 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); @@ -415,7 +385,7 @@ static int gmc_v7_0_mc_init(struct amdgpu_device *adev) * size equal to the 1024 or vram, whichever is larger. */ if (amdgpu_gart_size == -1) - adev->mc.gtt_size = max((1024ULL << 20), adev->mc.mc_vram_size); + adev->mc.gtt_size = amdgpu_ttm_get_gtt_mem_size(adev); else adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20; @@ -975,6 +945,11 @@ static int gmc_v7_0_sw_init(void *handle) return r; } + r = amdgpu_ttm_global_init(adev); + if (r) { + return r; + } + r = gmc_v7_0_mc_init(adev); if (r) return r; @@ -1137,7 +1112,7 @@ static int gmc_v7_0_soft_reset(void *handle) if (srbm_soft_reset) { gmc_v7_0_mc_stop(adev, &save); - if (gmc_v7_0_wait_for_idle(adev)) { + if (gmc_v7_0_wait_for_idle((void *)adev)) { dev_warn(adev->dev, "Wait for GMC idle timed out !\n"); } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.h b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.h index 36fcbbc46ada..0b386b5d2f7a 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.h +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.h @@ -26,11 +26,4 @@ extern const struct amd_ip_funcs gmc_v7_0_ip_funcs; -/* XXX these shouldn't be exported */ -void gmc_v7_0_mc_stop(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save); -void gmc_v7_0_mc_resume(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save); -int gmc_v7_0_mc_wait_for_idle(struct amdgpu_device *adev); - #endif diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index 9945d5bbf1fe..84c10d5117a9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -41,6 +41,7 @@ static void gmc_v8_0_set_gart_funcs(struct amdgpu_device *adev); static void gmc_v8_0_set_irq_funcs(struct amdgpu_device *adev); +static int gmc_v8_0_wait_for_idle(void *handle); MODULE_FIRMWARE("amdgpu/tonga_mc.bin"); MODULE_FIRMWARE("amdgpu/polaris11_mc.bin"); @@ -102,6 +103,11 @@ static const u32 stoney_mgcg_cgcg_init[] = mmMC_MEM_POWER_LS, 0xffffffff, 0x00000104 }; +static const u32 golden_settings_stoney_common[] = +{ + mmMC_HUB_RDREQ_UVD, MC_HUB_RDREQ_UVD__PRESCALE_MASK, 0x00000004, + mmMC_RD_GRP_OTH, MC_RD_GRP_OTH__UVD_MASK, 0x00600000 +}; static void gmc_v8_0_init_golden_registers(struct amdgpu_device *adev) { @@ -141,50 +147,24 @@ static void gmc_v8_0_init_golden_registers(struct amdgpu_device *adev) amdgpu_program_register_sequence(adev, stoney_mgcg_cgcg_init, (const u32)ARRAY_SIZE(stoney_mgcg_cgcg_init)); + amdgpu_program_register_sequence(adev, + golden_settings_stoney_common, + (const u32)ARRAY_SIZE(golden_settings_stoney_common)); break; default: break; } } -/** - * gmc8_mc_wait_for_idle - wait for MC idle callback. - * - * @adev: amdgpu_device pointer - * - * Wait for the MC (memory controller) to be idle. - * (evergreen+). - * Returns 0 if the MC is idle, -1 if not. - */ -int gmc_v8_0_mc_wait_for_idle(struct amdgpu_device *adev) -{ - unsigned i; - u32 tmp; - - for (i = 0; i < adev->usec_timeout; i++) { - /* read MC_STATUS */ - tmp = RREG32(mmSRBM_STATUS) & (SRBM_STATUS__VMC_BUSY_MASK | - SRBM_STATUS__MCB_BUSY_MASK | - SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK | - SRBM_STATUS__MCC_BUSY_MASK | - SRBM_STATUS__MCD_BUSY_MASK | - SRBM_STATUS__VMC1_BUSY_MASK); - if (!tmp) - return 0; - udelay(1); - } - return -1; -} - -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, + struct amdgpu_mode_mc_save *save) { u32 blackout; if (adev->mode_info.num_crtc) amdgpu_display_stop_mc_access(adev, save); - amdgpu_asic_wait_for_mc_idle(adev); + gmc_v8_0_wait_for_idle(adev); blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL); if (REG_GET_FIELD(blackout, MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE) != 1) { @@ -199,8 +179,8 @@ void gmc_v8_0_mc_stop(struct amdgpu_device *adev, udelay(100); } -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, + struct amdgpu_mode_mc_save *save) { u32 tmp; @@ -281,7 +261,7 @@ static int gmc_v8_0_mc_load_microcode(struct amdgpu_device *adev) const struct mc_firmware_header_v1_0 *hdr; const __le32 *fw_data = NULL; const __le32 *io_mc_regs = NULL; - u32 running, blackout = 0; + u32 running; int i, ucode_size, regs_size; if (!adev->mc.fw) @@ -307,11 +287,6 @@ static int gmc_v8_0_mc_load_microcode(struct amdgpu_device *adev) running = REG_GET_FIELD(RREG32(mmMC_SEQ_SUP_CNTL), MC_SEQ_SUP_CNTL, RUN); if (running == 0) { - if (running) { - blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL); - WREG32(mmMC_SHARED_BLACKOUT_CNTL, blackout | 1); - } - /* reset the engine and set to writable */ WREG32(mmMC_SEQ_SUP_CNTL, 0x00000008); WREG32(mmMC_SEQ_SUP_CNTL, 0x00000010); @@ -343,9 +318,6 @@ static int gmc_v8_0_mc_load_microcode(struct amdgpu_device *adev) break; udelay(1); } - - if (running) - WREG32(mmMC_SHARED_BLACKOUT_CNTL, blackout); } return 0; @@ -393,7 +365,7 @@ static void gmc_v8_0_mc_program(struct amdgpu_device *adev) amdgpu_display_set_vga_render_state(adev, false); gmc_v8_0_mc_stop(adev, &save); - if (amdgpu_asic_wait_for_mc_idle(adev)) { + if (gmc_v8_0_wait_for_idle((void *)adev)) { dev_warn(adev->dev, "Wait for MC idle timedout !\n"); } /* Update configuration */ @@ -413,7 +385,7 @@ static void gmc_v8_0_mc_program(struct amdgpu_device *adev) WREG32(mmMC_VM_AGP_BASE, 0); WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF); WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF); - if (amdgpu_asic_wait_for_mc_idle(adev)) { + 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); @@ -497,7 +469,7 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev) * size equal to the 1024 or vram, whichever is larger. */ if (amdgpu_gart_size == -1) - adev->mc.gtt_size = max((1024ULL << 20), adev->mc.mc_vram_size); + adev->mc.gtt_size = amdgpu_ttm_get_gtt_mem_size(adev); else adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20; @@ -977,6 +949,11 @@ static int gmc_v8_0_sw_init(void *handle) return r; } + r = amdgpu_ttm_global_init(adev); + if (r) { + return r; + } + r = gmc_v8_0_mc_init(adev); if (r) return r; @@ -1120,9 +1097,8 @@ static int gmc_v8_0_wait_for_idle(void *handle) } -static int gmc_v8_0_soft_reset(void *handle) +static int gmc_v8_0_check_soft_reset(void *handle) { - struct amdgpu_mode_mc_save save; u32 srbm_soft_reset = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; u32 tmp = RREG32(mmSRBM_STATUS); @@ -1137,13 +1113,42 @@ static int gmc_v8_0_soft_reset(void *handle) srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_MC, 1); } - if (srbm_soft_reset) { - gmc_v8_0_mc_stop(adev, &save); - if (gmc_v8_0_wait_for_idle(adev)) { - dev_warn(adev->dev, "Wait for GMC idle timed out !\n"); - } + adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang = true; + adev->mc.srbm_soft_reset = srbm_soft_reset; + } else { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang = false; + adev->mc.srbm_soft_reset = 0; + } + return 0; +} +static int gmc_v8_0_pre_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang) + return 0; + + gmc_v8_0_mc_stop(adev, &adev->mc.save); + if (gmc_v8_0_wait_for_idle(adev)) { + dev_warn(adev->dev, "Wait for GMC idle timed out !\n"); + } + + return 0; +} + +static int gmc_v8_0_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang) + return 0; + srbm_soft_reset = adev->mc.srbm_soft_reset; + + if (srbm_soft_reset) { + u32 tmp; tmp = RREG32(mmSRBM_SOFT_RESET); tmp |= srbm_soft_reset; @@ -1159,14 +1164,22 @@ static int gmc_v8_0_soft_reset(void *handle) /* Wait a little for things to settle down */ udelay(50); - - gmc_v8_0_mc_resume(adev, &save); - udelay(50); } return 0; } +static int gmc_v8_0_post_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_GMC].hang) + return 0; + + gmc_v8_0_mc_resume(adev, &adev->mc.save); + return 0; +} + static int gmc_v8_0_vm_fault_interrupt_state(struct amdgpu_device *adev, struct amdgpu_irq_src *src, unsigned type, @@ -1434,7 +1447,10 @@ const struct amd_ip_funcs gmc_v8_0_ip_funcs = { .resume = gmc_v8_0_resume, .is_idle = gmc_v8_0_is_idle, .wait_for_idle = gmc_v8_0_wait_for_idle, + .check_soft_reset = gmc_v8_0_check_soft_reset, + .pre_soft_reset = gmc_v8_0_pre_soft_reset, .soft_reset = gmc_v8_0_soft_reset, + .post_soft_reset = gmc_v8_0_post_soft_reset, .set_clockgating_state = gmc_v8_0_set_clockgating_state, .set_powergating_state = gmc_v8_0_set_powergating_state, }; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.h b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.h index 973436086b38..fc5001a8119d 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.h +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.h @@ -26,11 +26,4 @@ extern const struct amd_ip_funcs gmc_v8_0_ip_funcs; -/* XXX these shouldn't be exported */ -void gmc_v8_0_mc_stop(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save); -void gmc_v8_0_mc_resume(struct amdgpu_device *adev, - struct amdgpu_mode_mc_save *save); -int gmc_v8_0_mc_wait_for_idle(struct amdgpu_device *adev); - #endif diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_dpm.c b/drivers/gpu/drm/amd/amdgpu/iceland_dpm.c index 825ccd63f2dc..2f078ad6095c 100644 --- a/drivers/gpu/drm/amd/amdgpu/iceland_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/iceland_dpm.c @@ -24,7 +24,7 @@ #include <linux/firmware.h> #include "drmP.h" #include "amdgpu.h" -#include "iceland_smumgr.h" +#include "iceland_smum.h" MODULE_FIRMWARE("amdgpu/topaz_smc.bin"); diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_smc.c b/drivers/gpu/drm/amd/amdgpu/iceland_smc.c index 52ee08193295..211839913728 100644 --- a/drivers/gpu/drm/amd/amdgpu/iceland_smc.c +++ b/drivers/gpu/drm/amd/amdgpu/iceland_smc.c @@ -25,7 +25,7 @@ #include "drmP.h" #include "amdgpu.h" #include "ppsmc.h" -#include "iceland_smumgr.h" +#include "iceland_smum.h" #include "smu_ucode_xfer_vi.h" #include "amdgpu_ucode.h" @@ -211,7 +211,7 @@ static int iceland_send_msg_to_smc_without_waiting(struct amdgpu_device *adev, PPSMC_Msg msg) { if (!iceland_is_smc_ram_running(adev)) - return -EINVAL;; + return -EINVAL; if (wait_smu_response(adev)) { DRM_ERROR("Failed to send previous message\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_smumgr.h b/drivers/gpu/drm/amd/amdgpu/iceland_smum.h index 1e0769e110fa..5983e3150cc5 100644 --- a/drivers/gpu/drm/amd/amdgpu/iceland_smumgr.h +++ b/drivers/gpu/drm/amd/amdgpu/iceland_smum.h @@ -21,8 +21,8 @@ * */ -#ifndef ICELAND_SMUMGR_H -#define ICELAND_SMUMGR_H +#ifndef ICELAND_SMUM_H +#define ICELAND_SMUM_H #include "ppsmc.h" diff --git a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c index a789a863d677..f8618a3881a8 100644 --- a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c @@ -191,6 +191,7 @@ static void sumo_construct_vid_mapping_table(struct amdgpu_device *adev, vid_mapping_table->num_entries = i; } +#if 0 static const struct kv_lcac_config_values sx_local_cac_cfg_kv[] = { { 0, 4, 1 }, @@ -289,6 +290,7 @@ static const struct kv_lcac_config_reg cpl_cac_config_reg[] = { { 0xc0400d80, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 } }; +#endif static const struct kv_pt_config_reg didt_config_kv[] = { @@ -507,19 +509,19 @@ static int kv_enable_didt(struct amdgpu_device *adev, bool enable) pi->caps_db_ramping || pi->caps_td_ramping || pi->caps_tcp_ramping) { - gfx_v7_0_enter_rlc_safe_mode(adev); + adev->gfx.rlc.funcs->enter_safe_mode(adev); if (enable) { ret = kv_program_pt_config_registers(adev, didt_config_kv); if (ret) { - gfx_v7_0_exit_rlc_safe_mode(adev); + adev->gfx.rlc.funcs->exit_safe_mode(adev); return ret; } } kv_do_enable_didt(adev, enable); - gfx_v7_0_exit_rlc_safe_mode(adev); + adev->gfx.rlc.funcs->exit_safe_mode(adev); } return 0; @@ -2843,7 +2845,11 @@ static int kv_dpm_init(struct amdgpu_device *adev) pi->caps_tcp_ramping = true; } - pi->caps_sclk_ds = true; + if (amdgpu_sclk_deep_sleep_en) + pi->caps_sclk_ds = true; + else + pi->caps_sclk_ds = false; + pi->enable_auto_thermal_throttling = true; pi->disable_nb_ps3_in_battery = false; if (amdgpu_bapm == 0) diff --git a/drivers/gpu/drm/amd/amdgpu/ppsmc.h b/drivers/gpu/drm/amd/amdgpu/ppsmc.h index 7837f2ecc357..8463245f424f 100644 --- a/drivers/gpu/drm/amd/amdgpu/ppsmc.h +++ b/drivers/gpu/drm/amd/amdgpu/ppsmc.h @@ -90,7 +90,9 @@ typedef uint8_t PPSMC_Result; #define PPSMC_StartFanControl ((uint8_t)0x5B) #define PPSMC_StopFanControl ((uint8_t)0x5C) #define PPSMC_MSG_NoDisplay ((uint8_t)0x5D) +#define PPSMC_NoDisplay ((uint8_t)0x5D) #define PPSMC_MSG_HasDisplay ((uint8_t)0x5E) +#define PPSMC_HasDisplay ((uint8_t)0x5E) #define PPSMC_MSG_UVDPowerOFF ((uint8_t)0x60) #define PPSMC_MSG_UVDPowerON ((uint8_t)0x61) #define PPSMC_MSG_EnableULV ((uint8_t)0x62) @@ -108,6 +110,7 @@ typedef uint8_t PPSMC_Result; #define PPSMC_MSG_DisableDTE ((uint8_t)0x88) #define PPSMC_MSG_ThrottleOVRDSCLKDS ((uint8_t)0x96) #define PPSMC_MSG_CancelThrottleOVRDSCLKDS ((uint8_t)0x97) +#define PPSMC_MSG_EnableACDCGPIOInterrupt ((uint16_t) 0x149) /* CI/KV/KB */ #define PPSMC_MSG_UVDDPM_SetEnabledMask ((uint16_t) 0x12D) @@ -161,6 +164,7 @@ typedef uint8_t PPSMC_Result; #define PPSMC_MSG_MASTER_DeepSleep_OFF ((uint16_t) 0x190) #define PPSMC_MSG_Remove_DC_Clamp ((uint16_t) 0x191) #define PPSMC_MSG_SetFanPwmMax ((uint16_t) 0x19A) +#define PPSMC_MSG_SetFanRpmMax ((uint16_t) 0x205) #define PPSMC_MSG_ENABLE_THERMAL_DPM ((uint16_t) 0x19C) #define PPSMC_MSG_DISABLE_THERMAL_DPM ((uint16_t) 0x19D) diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index b556bd0a8797..e82229686783 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -255,19 +255,6 @@ static void sdma_v2_4_ring_emit_ib(struct amdgpu_ring *ring, unsigned vm_id, bool ctx_switch) { u32 vmid = vm_id & 0xf; - u32 next_rptr = ring->wptr + 5; - - while ((next_rptr & 7) != 2) - next_rptr++; - - next_rptr += 6; - - amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | - SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR)); - amdgpu_ring_write(ring, lower_32_bits(ring->next_rptr_gpu_addr) & 0xfffffffc); - amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr)); - amdgpu_ring_write(ring, SDMA_PKT_WRITE_UNTILED_DW_3_COUNT(1)); - amdgpu_ring_write(ring, next_rptr); /* IB packet must end on a 8 DW boundary */ sdma_v2_4_ring_insert_nop(ring, (10 - (ring->wptr & 7)) % 8); @@ -406,7 +393,7 @@ static void sdma_v2_4_enable(struct amdgpu_device *adev, bool enable) u32 f32_cntl; int i; - if (enable == false) { + if (!enable) { sdma_v2_4_gfx_stop(adev); sdma_v2_4_rlc_stop(adev); } @@ -580,19 +567,21 @@ static int sdma_v2_4_start(struct amdgpu_device *adev) { int r; - if (!adev->firmware.smu_load) { - 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->pp_enabled) { + if (!adev->firmware.smu_load) { + 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; + } } /* halt the engine before programing */ @@ -679,20 +668,19 @@ static int sdma_v2_4_ring_test_ring(struct amdgpu_ring *ring) * Test a simple IB in the DMA ring (VI). * Returns 0 on success, error on failure. */ -static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring) +static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring, long timeout) { struct amdgpu_device *adev = ring->adev; struct amdgpu_ib ib; struct fence *f = NULL; - unsigned i; unsigned index; - int r; u32 tmp = 0; u64 gpu_addr; + long r; r = amdgpu_wb_get(adev, &index); if (r) { - dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); + dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); return r; } @@ -702,7 +690,7 @@ static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring) memset(&ib, 0, sizeof(ib)); r = amdgpu_ib_get(adev, NULL, 256, &ib); if (r) { - DRM_ERROR("amdgpu: failed to get ib (%d).\n", r); + DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); goto err0; } @@ -721,28 +709,25 @@ static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring) if (r) goto err1; - r = fence_wait(f, false); - if (r) { - DRM_ERROR("amdgpu: fence wait failed (%d).\n", r); + r = fence_wait_timeout(f, false, timeout); + if (r == 0) { + DRM_ERROR("amdgpu: IB test timed out\n"); + r = -ETIMEDOUT; goto err1; - } - for (i = 0; i < adev->usec_timeout; i++) { - tmp = le32_to_cpu(adev->wb.wb[index]); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - if (i < adev->usec_timeout) { - DRM_INFO("ib test on ring %d succeeded in %u usecs\n", - ring->idx, i); + } else if (r) { + DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto err1; + } + tmp = le32_to_cpu(adev->wb.wb[index]); + if (tmp == 0xDEADBEEF) { + DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + r = 0; } else { DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp); r = -EINVAL; } err1: - fence_put(f); amdgpu_ib_free(adev, &ib, NULL); fence_put(f); err0: @@ -764,24 +749,16 @@ static void sdma_v2_4_vm_copy_pte(struct amdgpu_ib *ib, uint64_t pe, uint64_t src, unsigned count) { - while (count) { - unsigned bytes = count * 8; - if (bytes > 0x1FFFF8) - bytes = 0x1FFFF8; - - ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) | - SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); - ib->ptr[ib->length_dw++] = bytes; - ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ - ib->ptr[ib->length_dw++] = lower_32_bits(src); - ib->ptr[ib->length_dw++] = upper_32_bits(src); - ib->ptr[ib->length_dw++] = lower_32_bits(pe); - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - - pe += bytes; - src += bytes; - count -= bytes / 8; - } + unsigned bytes = count * 8; + + ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) | + SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); + ib->ptr[ib->length_dw++] = bytes; + ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ + ib->ptr[ib->length_dw++] = lower_32_bits(src); + ib->ptr[ib->length_dw++] = upper_32_bits(src); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); } /** @@ -789,39 +766,27 @@ static void sdma_v2_4_vm_copy_pte(struct amdgpu_ib *ib, * * @ib: indirect buffer to fill with commands * @pe: addr of the page entry - * @addr: dst addr to write into pe + * @value: dst addr to write into pe * @count: number of page entries to update * @incr: increase next addr by incr bytes - * @flags: access flags * * Update PTEs by writing them manually using sDMA (CIK). */ -static void sdma_v2_4_vm_write_pte(struct amdgpu_ib *ib, - const dma_addr_t *pages_addr, uint64_t pe, - uint64_t addr, unsigned count, - uint32_t incr, uint32_t flags) +static void sdma_v2_4_vm_write_pte(struct amdgpu_ib *ib, uint64_t pe, + uint64_t value, unsigned count, + uint32_t incr) { - uint64_t value; - unsigned ndw; - - while (count) { - ndw = count * 2; - if (ndw > 0xFFFFE) - ndw = 0xFFFFE; - - /* for non-physically contiguous pages (system) */ - ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | - SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); - ib->ptr[ib->length_dw++] = pe; - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = ndw; - for (; ndw > 0; ndw -= 2, --count, pe += 8) { - value = amdgpu_vm_map_gart(pages_addr, addr); - addr += incr; - value |= flags; - ib->ptr[ib->length_dw++] = value; - ib->ptr[ib->length_dw++] = upper_32_bits(value); - } + unsigned ndw = count * 2; + + ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | + SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); + ib->ptr[ib->length_dw++] = pe; + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = ndw; + for (; ndw > 0; ndw -= 2, --count, pe += 8) { + ib->ptr[ib->length_dw++] = lower_32_bits(value); + ib->ptr[ib->length_dw++] = upper_32_bits(value); + value += incr; } } @@ -837,40 +802,21 @@ static void sdma_v2_4_vm_write_pte(struct amdgpu_ib *ib, * * Update the page tables using sDMA (CIK). */ -static void sdma_v2_4_vm_set_pte_pde(struct amdgpu_ib *ib, - uint64_t pe, +static void sdma_v2_4_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe, uint64_t addr, unsigned count, uint32_t incr, uint32_t flags) { - uint64_t value; - unsigned ndw; - - while (count) { - ndw = count; - if (ndw > 0x7FFFF) - ndw = 0x7FFFF; - - if (flags & AMDGPU_PTE_VALID) - value = addr; - else - value = 0; - - /* for physically contiguous pages (vram) */ - ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_GEN_PTEPDE); - ib->ptr[ib->length_dw++] = pe; /* dst addr */ - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = flags; /* mask */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = value; /* value */ - ib->ptr[ib->length_dw++] = upper_32_bits(value); - ib->ptr[ib->length_dw++] = incr; /* increment size */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = ndw; /* number of entries */ - - pe += ndw * 8; - addr += ndw * incr; - count -= ndw; - } + /* for physically contiguous pages (vram) */ + ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_GEN_PTEPDE); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); /* dst addr */ + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = flags; /* mask */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = lower_32_bits(addr); /* value */ + ib->ptr[ib->length_dw++] = upper_32_bits(addr); + ib->ptr[ib->length_dw++] = incr; /* increment size */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = count; /* number of entries */ } /** diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 532ea88da66a..bee4978bec73 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -415,18 +415,6 @@ static void sdma_v3_0_ring_emit_ib(struct amdgpu_ring *ring, unsigned vm_id, bool ctx_switch) { u32 vmid = vm_id & 0xf; - u32 next_rptr = ring->wptr + 5; - - while ((next_rptr & 7) != 2) - next_rptr++; - next_rptr += 6; - - amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | - SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR)); - amdgpu_ring_write(ring, lower_32_bits(ring->next_rptr_gpu_addr) & 0xfffffffc); - amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr)); - amdgpu_ring_write(ring, SDMA_PKT_WRITE_UNTILED_DW_3_COUNT(1)); - amdgpu_ring_write(ring, next_rptr); /* IB packet must end on a 8 DW boundary */ sdma_v3_0_ring_insert_nop(ring, (10 - (ring->wptr & 7)) % 8); @@ -616,7 +604,7 @@ static void sdma_v3_0_enable(struct amdgpu_device *adev, bool enable) u32 f32_cntl; int i; - if (enable == false) { + if (!enable) { sdma_v3_0_gfx_stop(adev); sdma_v3_0_rlc_stop(adev); } @@ -908,20 +896,19 @@ static int sdma_v3_0_ring_test_ring(struct amdgpu_ring *ring) * Test a simple IB in the DMA ring (VI). * Returns 0 on success, error on failure. */ -static int sdma_v3_0_ring_test_ib(struct amdgpu_ring *ring) +static int sdma_v3_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) { struct amdgpu_device *adev = ring->adev; struct amdgpu_ib ib; struct fence *f = NULL; - unsigned i; unsigned index; - int r; u32 tmp = 0; u64 gpu_addr; + long r; r = amdgpu_wb_get(adev, &index); if (r) { - dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r); + dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); return r; } @@ -931,7 +918,7 @@ static int sdma_v3_0_ring_test_ib(struct amdgpu_ring *ring) memset(&ib, 0, sizeof(ib)); r = amdgpu_ib_get(adev, NULL, 256, &ib); if (r) { - DRM_ERROR("amdgpu: failed to get ib (%d).\n", r); + DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); goto err0; } @@ -950,27 +937,24 @@ static int sdma_v3_0_ring_test_ib(struct amdgpu_ring *ring) if (r) goto err1; - r = fence_wait(f, false); - if (r) { - DRM_ERROR("amdgpu: fence wait failed (%d).\n", r); + r = fence_wait_timeout(f, false, timeout); + if (r == 0) { + DRM_ERROR("amdgpu: IB test timed out\n"); + r = -ETIMEDOUT; goto err1; - } - for (i = 0; i < adev->usec_timeout; i++) { - tmp = le32_to_cpu(adev->wb.wb[index]); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - if (i < adev->usec_timeout) { - DRM_INFO("ib test on ring %d succeeded in %u usecs\n", - ring->idx, i); + } else if (r < 0) { + DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto err1; + } + tmp = le32_to_cpu(adev->wb.wb[index]); + if (tmp == 0xDEADBEEF) { + DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + r = 0; } else { DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp); r = -EINVAL; } err1: - fence_put(f); amdgpu_ib_free(adev, &ib, NULL); fence_put(f); err0: @@ -992,24 +976,16 @@ static void sdma_v3_0_vm_copy_pte(struct amdgpu_ib *ib, uint64_t pe, uint64_t src, unsigned count) { - while (count) { - unsigned bytes = count * 8; - if (bytes > 0x1FFFF8) - bytes = 0x1FFFF8; - - ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) | - SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); - ib->ptr[ib->length_dw++] = bytes; - ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ - ib->ptr[ib->length_dw++] = lower_32_bits(src); - ib->ptr[ib->length_dw++] = upper_32_bits(src); - ib->ptr[ib->length_dw++] = lower_32_bits(pe); - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - - pe += bytes; - src += bytes; - count -= bytes / 8; - } + unsigned bytes = count * 8; + + ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) | + SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); + ib->ptr[ib->length_dw++] = bytes; + ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ + ib->ptr[ib->length_dw++] = lower_32_bits(src); + ib->ptr[ib->length_dw++] = upper_32_bits(src); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); } /** @@ -1017,39 +993,27 @@ static void sdma_v3_0_vm_copy_pte(struct amdgpu_ib *ib, * * @ib: indirect buffer to fill with commands * @pe: addr of the page entry - * @addr: dst addr to write into pe + * @value: dst addr to write into pe * @count: number of page entries to update * @incr: increase next addr by incr bytes - * @flags: access flags * * Update PTEs by writing them manually using sDMA (CIK). */ -static void sdma_v3_0_vm_write_pte(struct amdgpu_ib *ib, - const dma_addr_t *pages_addr, uint64_t pe, - uint64_t addr, unsigned count, - uint32_t incr, uint32_t flags) -{ - uint64_t value; - unsigned ndw; - - while (count) { - ndw = count * 2; - if (ndw > 0xFFFFE) - ndw = 0xFFFFE; - - /* for non-physically contiguous pages (system) */ - ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | - SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); - ib->ptr[ib->length_dw++] = pe; - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = ndw; - for (; ndw > 0; ndw -= 2, --count, pe += 8) { - value = amdgpu_vm_map_gart(pages_addr, addr); - addr += incr; - value |= flags; - ib->ptr[ib->length_dw++] = value; - ib->ptr[ib->length_dw++] = upper_32_bits(value); - } +static void sdma_v3_0_vm_write_pte(struct amdgpu_ib *ib, uint64_t pe, + uint64_t value, unsigned count, + uint32_t incr) +{ + unsigned ndw = count * 2; + + ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) | + SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = ndw; + for (; ndw > 0; ndw -= 2, --count, pe += 8) { + ib->ptr[ib->length_dw++] = lower_32_bits(value); + ib->ptr[ib->length_dw++] = upper_32_bits(value); + value += incr; } } @@ -1065,40 +1029,21 @@ static void sdma_v3_0_vm_write_pte(struct amdgpu_ib *ib, * * Update the page tables using sDMA (CIK). */ -static void sdma_v3_0_vm_set_pte_pde(struct amdgpu_ib *ib, - uint64_t pe, +static void sdma_v3_0_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe, uint64_t addr, unsigned count, uint32_t incr, uint32_t flags) { - uint64_t value; - unsigned ndw; - - while (count) { - ndw = count; - if (ndw > 0x7FFFF) - ndw = 0x7FFFF; - - if (flags & AMDGPU_PTE_VALID) - value = addr; - else - value = 0; - - /* for physically contiguous pages (vram) */ - ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_GEN_PTEPDE); - ib->ptr[ib->length_dw++] = pe; /* dst addr */ - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = flags; /* mask */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = value; /* value */ - ib->ptr[ib->length_dw++] = upper_32_bits(value); - ib->ptr[ib->length_dw++] = incr; /* increment size */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = ndw; /* number of entries */ - - pe += ndw * 8; - addr += ndw * incr; - count -= ndw; - } + /* for physically contiguous pages (vram) */ + ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_GEN_PTEPDE); + ib->ptr[ib->length_dw++] = lower_32_bits(pe); /* dst addr */ + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = flags; /* mask */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = lower_32_bits(addr); /* value */ + ib->ptr[ib->length_dw++] = upper_32_bits(addr); + ib->ptr[ib->length_dw++] = incr; /* increment size */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = count; /* number of entries */ } /** @@ -1336,28 +1281,79 @@ static int sdma_v3_0_wait_for_idle(void *handle) return -ETIMEDOUT; } -static int sdma_v3_0_soft_reset(void *handle) +static int sdma_v3_0_check_soft_reset(void *handle) { - u32 srbm_soft_reset = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; u32 tmp = RREG32(mmSRBM_STATUS2); - if (tmp & SRBM_STATUS2__SDMA_BUSY_MASK) { - /* sdma0 */ - tmp = RREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET); - tmp = REG_SET_FIELD(tmp, SDMA0_F32_CNTL, HALT, 0); - WREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET, tmp); + if ((tmp & SRBM_STATUS2__SDMA_BUSY_MASK) || + (tmp & SRBM_STATUS2__SDMA1_BUSY_MASK)) { srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA_MASK; - } - if (tmp & SRBM_STATUS2__SDMA1_BUSY_MASK) { - /* sdma1 */ - tmp = RREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET); - tmp = REG_SET_FIELD(tmp, SDMA0_F32_CNTL, HALT, 0); - WREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET, tmp); srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA1_MASK; } if (srbm_soft_reset) { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_SDMA].hang = true; + adev->sdma.srbm_soft_reset = srbm_soft_reset; + } else { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_SDMA].hang = false; + adev->sdma.srbm_soft_reset = 0; + } + + return 0; +} + +static int sdma_v3_0_pre_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_SDMA].hang) + return 0; + + srbm_soft_reset = adev->sdma.srbm_soft_reset; + + if (REG_GET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_SDMA) || + REG_GET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_SDMA1)) { + sdma_v3_0_ctx_switch_enable(adev, false); + sdma_v3_0_enable(adev, false); + } + + return 0; +} + +static int sdma_v3_0_post_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_SDMA].hang) + return 0; + + srbm_soft_reset = adev->sdma.srbm_soft_reset; + + if (REG_GET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_SDMA) || + REG_GET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_SDMA1)) { + sdma_v3_0_gfx_resume(adev); + sdma_v3_0_rlc_resume(adev); + } + + return 0; +} + +static int sdma_v3_0_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; + u32 tmp; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_SDMA].hang) + return 0; + + srbm_soft_reset = adev->sdma.srbm_soft_reset; + + if (srbm_soft_reset) { tmp = RREG32(mmSRBM_SOFT_RESET); tmp |= srbm_soft_reset; dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); @@ -1575,6 +1571,9 @@ const struct amd_ip_funcs sdma_v3_0_ip_funcs = { .resume = sdma_v3_0_resume, .is_idle = sdma_v3_0_is_idle, .wait_for_idle = sdma_v3_0_wait_for_idle, + .check_soft_reset = sdma_v3_0_check_soft_reset, + .pre_soft_reset = sdma_v3_0_pre_soft_reset, + .post_soft_reset = sdma_v3_0_post_soft_reset, .soft_reset = sdma_v3_0_soft_reset, .set_clockgating_state = sdma_v3_0_set_clockgating_state, .set_powergating_state = sdma_v3_0_set_powergating_state, diff --git a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c index c92055805a45..d127d59f953a 100644 --- a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c @@ -373,10 +373,10 @@ static int tonga_ih_wait_for_idle(void *handle) return -ETIMEDOUT; } -static int tonga_ih_soft_reset(void *handle) +static int tonga_ih_check_soft_reset(void *handle) { - u32 srbm_soft_reset = 0; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; u32 tmp = RREG32(mmSRBM_STATUS); if (tmp & SRBM_STATUS__IH_BUSY_MASK) @@ -384,6 +384,48 @@ static int tonga_ih_soft_reset(void *handle) SOFT_RESET_IH, 1); if (srbm_soft_reset) { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_IH].hang = true; + adev->irq.srbm_soft_reset = srbm_soft_reset; + } else { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_IH].hang = false; + adev->irq.srbm_soft_reset = 0; + } + + return 0; +} + +static int tonga_ih_pre_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_IH].hang) + return 0; + + return tonga_ih_hw_fini(adev); +} + +static int tonga_ih_post_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_IH].hang) + return 0; + + return tonga_ih_hw_init(adev); +} + +static int tonga_ih_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_IH].hang) + return 0; + srbm_soft_reset = adev->irq.srbm_soft_reset; + + if (srbm_soft_reset) { + u32 tmp; + tmp = RREG32(mmSRBM_SOFT_RESET); tmp |= srbm_soft_reset; dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); @@ -427,7 +469,10 @@ const struct amd_ip_funcs tonga_ih_ip_funcs = { .resume = tonga_ih_resume, .is_idle = tonga_ih_is_idle, .wait_for_idle = tonga_ih_wait_for_idle, + .check_soft_reset = tonga_ih_check_soft_reset, + .pre_soft_reset = tonga_ih_pre_soft_reset, .soft_reset = tonga_ih_soft_reset, + .post_soft_reset = tonga_ih_post_soft_reset, .set_clockgating_state = tonga_ih_set_clockgating_state, .set_powergating_state = tonga_ih_set_powergating_state, }; diff --git a/drivers/gpu/drm/amd/amdgpu/tonga_smc.c b/drivers/gpu/drm/amd/amdgpu/tonga_smc.c index 083893dd68c0..940de1836f8f 100644 --- a/drivers/gpu/drm/amd/amdgpu/tonga_smc.c +++ b/drivers/gpu/drm/amd/amdgpu/tonga_smc.c @@ -173,7 +173,7 @@ static int tonga_send_msg_to_smc(struct amdgpu_device *adev, PPSMC_Msg msg) { if (!tonga_is_smc_ram_running(adev)) { - return -EINVAL;; + return -EINVAL; } if (wait_smu_response(adev)) { diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c index f07551476a70..10c0407dcb6e 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c @@ -34,6 +34,8 @@ #include "oss/oss_2_0_d.h" #include "oss/oss_2_0_sh_mask.h" +#include "bif/bif_4_1_d.h" + static void uvd_v4_2_mc_resume(struct amdgpu_device *adev); static void uvd_v4_2_init_cg(struct amdgpu_device *adev); static void uvd_v4_2_set_ring_funcs(struct amdgpu_device *adev); @@ -114,7 +116,7 @@ static int uvd_v4_2_sw_init(void *handle) ring = &adev->uvd.ring; sprintf(ring->name, "uvd"); - r = amdgpu_ring_init(adev, ring, 512, CP_PACKET2, 0xf, + r = amdgpu_ring_init(adev, ring, 512, PACKET0(mmUVD_NO_OP, 0), 0xf, &adev->uvd.irq, 0, AMDGPU_RING_TYPE_UVD); return r; @@ -439,6 +441,32 @@ static void uvd_v4_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq } /** + * uvd_v4_2_ring_emit_hdp_flush - emit an hdp flush + * + * @ring: amdgpu_ring pointer + * + * Emits an hdp flush. + */ +static void uvd_v4_2_ring_emit_hdp_flush(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, PACKET0(mmHDP_MEM_COHERENCY_FLUSH_CNTL, 0)); + amdgpu_ring_write(ring, 0); +} + +/** + * uvd_v4_2_ring_hdp_invalidate - emit an hdp invalidate + * + * @ring: amdgpu_ring pointer + * + * Emits an hdp invalidate. + */ +static void uvd_v4_2_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, PACKET0(mmHDP_DEBUG0, 0)); + amdgpu_ring_write(ring, 1); +} + +/** * uvd_v4_2_ring_test_ring - register write test * * @ring: amdgpu_ring pointer @@ -499,49 +527,6 @@ static void uvd_v4_2_ring_emit_ib(struct amdgpu_ring *ring, } /** - * uvd_v4_2_ring_test_ib - test ib execution - * - * @ring: amdgpu_ring pointer - * - * Test if we can successfully execute an IB - */ -static int uvd_v4_2_ring_test_ib(struct amdgpu_ring *ring) -{ - struct amdgpu_device *adev = ring->adev; - struct fence *fence = NULL; - int r; - - r = amdgpu_asic_set_uvd_clocks(adev, 53300, 40000); - if (r) { - DRM_ERROR("amdgpu: failed to raise UVD clocks (%d).\n", r); - return r; - } - - r = amdgpu_uvd_get_create_msg(ring, 1, NULL); - if (r) { - DRM_ERROR("amdgpu: failed to get create msg (%d).\n", r); - goto error; - } - - r = amdgpu_uvd_get_destroy_msg(ring, 1, true, &fence); - if (r) { - DRM_ERROR("amdgpu: failed to get destroy ib (%d).\n", r); - goto error; - } - - r = fence_wait(fence, false); - if (r) { - DRM_ERROR("amdgpu: fence wait failed (%d).\n", r); - goto error; - } - DRM_INFO("ib test on ring %d succeeded\n", ring->idx); -error: - fence_put(fence); - amdgpu_asic_set_uvd_clocks(adev, 0, 0); - return r; -} - -/** * uvd_v4_2_mc_resume - memory controller programming * * @adev: amdgpu_device pointer @@ -763,10 +748,14 @@ static const struct amdgpu_ring_funcs uvd_v4_2_ring_funcs = { .parse_cs = amdgpu_uvd_ring_parse_cs, .emit_ib = uvd_v4_2_ring_emit_ib, .emit_fence = uvd_v4_2_ring_emit_fence, + .emit_hdp_flush = uvd_v4_2_ring_emit_hdp_flush, + .emit_hdp_invalidate = uvd_v4_2_ring_emit_hdp_invalidate, .test_ring = uvd_v4_2_ring_test_ring, - .test_ib = uvd_v4_2_ring_test_ib, + .test_ib = amdgpu_uvd_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, + .begin_use = amdgpu_uvd_ring_begin_use, + .end_use = amdgpu_uvd_ring_end_use, }; static void uvd_v4_2_set_ring_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c index e0a76a883d46..8513376062c1 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c @@ -31,6 +31,7 @@ #include "uvd/uvd_5_0_sh_mask.h" #include "oss/oss_2_0_d.h" #include "oss/oss_2_0_sh_mask.h" +#include "bif/bif_5_0_d.h" #include "vi.h" static void uvd_v5_0_set_ring_funcs(struct amdgpu_device *adev); @@ -111,7 +112,7 @@ static int uvd_v5_0_sw_init(void *handle) ring = &adev->uvd.ring; sprintf(ring->name, "uvd"); - r = amdgpu_ring_init(adev, ring, 512, CP_PACKET2, 0xf, + r = amdgpu_ring_init(adev, ring, 512, PACKET0(mmUVD_NO_OP, 0), 0xf, &adev->uvd.irq, 0, AMDGPU_RING_TYPE_UVD); return r; @@ -489,6 +490,32 @@ static void uvd_v5_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq } /** + * uvd_v5_0_ring_emit_hdp_flush - emit an hdp flush + * + * @ring: amdgpu_ring pointer + * + * Emits an hdp flush. + */ +static void uvd_v5_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, PACKET0(mmHDP_MEM_COHERENCY_FLUSH_CNTL, 0)); + amdgpu_ring_write(ring, 0); +} + +/** + * uvd_v5_0_ring_hdp_invalidate - emit an hdp invalidate + * + * @ring: amdgpu_ring pointer + * + * Emits an hdp invalidate. + */ +static void uvd_v5_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, PACKET0(mmHDP_DEBUG0, 0)); + amdgpu_ring_write(ring, 1); +} + +/** * uvd_v5_0_ring_test_ring - register write test * * @ring: amdgpu_ring pointer @@ -550,49 +577,6 @@ static void uvd_v5_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, ib->length_dw); } -/** - * uvd_v5_0_ring_test_ib - test ib execution - * - * @ring: amdgpu_ring pointer - * - * Test if we can successfully execute an IB - */ -static int uvd_v5_0_ring_test_ib(struct amdgpu_ring *ring) -{ - struct amdgpu_device *adev = ring->adev; - struct fence *fence = NULL; - int r; - - r = amdgpu_asic_set_uvd_clocks(adev, 53300, 40000); - if (r) { - DRM_ERROR("amdgpu: failed to raise UVD clocks (%d).\n", r); - return r; - } - - r = amdgpu_uvd_get_create_msg(ring, 1, NULL); - if (r) { - DRM_ERROR("amdgpu: failed to get create msg (%d).\n", r); - goto error; - } - - r = amdgpu_uvd_get_destroy_msg(ring, 1, true, &fence); - if (r) { - DRM_ERROR("amdgpu: failed to get destroy ib (%d).\n", r); - goto error; - } - - r = fence_wait(fence, false); - if (r) { - DRM_ERROR("amdgpu: fence wait failed (%d).\n", r); - goto error; - } - DRM_INFO("ib test on ring %d succeeded\n", ring->idx); -error: - fence_put(fence); - amdgpu_asic_set_uvd_clocks(adev, 0, 0); - return r; -} - static bool uvd_v5_0_is_idle(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -815,10 +799,14 @@ static const struct amdgpu_ring_funcs uvd_v5_0_ring_funcs = { .parse_cs = amdgpu_uvd_ring_parse_cs, .emit_ib = uvd_v5_0_ring_emit_ib, .emit_fence = uvd_v5_0_ring_emit_fence, + .emit_hdp_flush = uvd_v5_0_ring_emit_hdp_flush, + .emit_hdp_invalidate = uvd_v5_0_ring_emit_hdp_invalidate, .test_ring = uvd_v5_0_ring_test_ring, - .test_ib = uvd_v5_0_ring_test_ib, + .test_ib = amdgpu_uvd_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, + .begin_use = amdgpu_uvd_ring_begin_use, + .end_use = amdgpu_uvd_ring_end_use, }; static void uvd_v5_0_set_ring_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c index c9929d665c01..2abe8a93c99f 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c @@ -33,6 +33,8 @@ #include "oss/oss_2_0_sh_mask.h" #include "smu/smu_7_1_3_d.h" #include "smu/smu_7_1_3_sh_mask.h" +#include "bif/bif_5_1_d.h" +#include "gmc/gmc_8_1_d.h" #include "vi.h" static void uvd_v6_0_set_ring_funcs(struct amdgpu_device *adev); @@ -114,7 +116,7 @@ static int uvd_v6_0_sw_init(void *handle) ring = &adev->uvd.ring; sprintf(ring->name, "uvd"); - r = amdgpu_ring_init(adev, ring, 512, CP_PACKET2, 0xf, + r = amdgpu_ring_init(adev, ring, 512, PACKET0(mmUVD_NO_OP, 0), 0xf, &adev->uvd.irq, 0, AMDGPU_RING_TYPE_UVD); return r; @@ -385,8 +387,8 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) uint32_t mp_swap_cntl; int i, j, r; - /*disable DPG */ - WREG32_P(mmUVD_POWER_STATUS, 0, ~(1 << 2)); + /* disable DPG */ + WREG32_P(mmUVD_POWER_STATUS, 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); /* disable byte swapping */ lmi_swap_cntl = 0; @@ -394,38 +396,40 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) uvd_v6_0_mc_resume(adev); - /* Set dynamic clock gating in S/W control mode */ - if (adev->cg_flags & AMD_CG_SUPPORT_UVD_MGCG) { - uvd_v6_0_set_sw_clock_gating(adev); - } else { - /* disable clock gating */ - uint32_t data = RREG32(mmUVD_CGC_CTRL); - data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK; - WREG32(mmUVD_CGC_CTRL, data); - } + /* disable clock gating */ + WREG32_FIELD(UVD_CGC_CTRL, DYN_CLOCK_MODE, 0); /* disable interupt */ - WREG32_P(mmUVD_MASTINT_EN, 0, ~(1 << 1)); + WREG32_FIELD(UVD_MASTINT_EN, VCPU_EN, 0); /* stall UMC and register bus before resetting VCPU */ - WREG32_P(mmUVD_LMI_CTRL2, 1 << 8, ~(1 << 8)); + WREG32_FIELD(UVD_LMI_CTRL2, STALL_ARB_UMC, 1); mdelay(1); /* put LMI, VCPU, RBC etc... into reset */ - WREG32(mmUVD_SOFT_RESET, UVD_SOFT_RESET__LMI_SOFT_RESET_MASK | - UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK | UVD_SOFT_RESET__LBSI_SOFT_RESET_MASK | - UVD_SOFT_RESET__RBC_SOFT_RESET_MASK | UVD_SOFT_RESET__CSM_SOFT_RESET_MASK | - UVD_SOFT_RESET__CXW_SOFT_RESET_MASK | UVD_SOFT_RESET__TAP_SOFT_RESET_MASK | + WREG32(mmUVD_SOFT_RESET, + UVD_SOFT_RESET__LMI_SOFT_RESET_MASK | + UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK | + UVD_SOFT_RESET__LBSI_SOFT_RESET_MASK | + UVD_SOFT_RESET__RBC_SOFT_RESET_MASK | + UVD_SOFT_RESET__CSM_SOFT_RESET_MASK | + UVD_SOFT_RESET__CXW_SOFT_RESET_MASK | + UVD_SOFT_RESET__TAP_SOFT_RESET_MASK | UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK); mdelay(5); /* take UVD block out of reset */ - WREG32_P(mmSRBM_SOFT_RESET, 0, ~SRBM_SOFT_RESET__SOFT_RESET_UVD_MASK); + WREG32_FIELD(SRBM_SOFT_RESET, SOFT_RESET_UVD, 0); mdelay(5); /* initialize UVD memory controller */ - WREG32(mmUVD_LMI_CTRL, 0x40 | (1 << 8) | (1 << 13) | - (1 << 21) | (1 << 9) | (1 << 20)); + WREG32(mmUVD_LMI_CTRL, + (0x40 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) | + UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | + UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__REQ_MODE_MASK | + UVD_LMI_CTRL__DISABLE_ON_FWV_FAIL_MASK); #ifdef __BIG_ENDIAN /* swap (8 in 32) RB and IB */ @@ -447,10 +451,10 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) mdelay(5); /* enable VCPU clock */ - WREG32(mmUVD_VCPU_CNTL, 1 << 9); + WREG32(mmUVD_VCPU_CNTL, UVD_VCPU_CNTL__CLK_EN_MASK); /* enable UMC */ - WREG32_P(mmUVD_LMI_CTRL2, 0, ~(1 << 8)); + WREG32_FIELD(UVD_LMI_CTRL2, STALL_ARB_UMC, 0); /* boot up the VCPU */ WREG32(mmUVD_SOFT_RESET, 0); @@ -470,11 +474,9 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) break; DRM_ERROR("UVD not responding, trying to reset the VCPU!!!\n"); - WREG32_P(mmUVD_SOFT_RESET, UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK, - ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); + WREG32_FIELD(UVD_SOFT_RESET, VCPU_SOFT_RESET, 1); mdelay(10); - WREG32_P(mmUVD_SOFT_RESET, 0, - ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); + WREG32_FIELD(UVD_SOFT_RESET, VCPU_SOFT_RESET, 0); mdelay(10); r = -1; } @@ -484,20 +486,21 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) return r; } /* enable master interrupt */ - WREG32_P(mmUVD_MASTINT_EN, 3 << 1, ~(3 << 1)); + WREG32_P(mmUVD_MASTINT_EN, + (UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK), + ~(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK)); /* clear the bit 4 of UVD_STATUS */ - WREG32_P(mmUVD_STATUS, 0, ~(2 << 1)); + WREG32_P(mmUVD_STATUS, 0, ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT)); + /* force RBC into idle state */ rb_bufsz = order_base_2(ring->ring_size); - tmp = 0; - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz); + tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz); 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); - /* force RBC into idle state */ WREG32(mmUVD_RBC_RB_CNTL, tmp); /* set the write pointer delay */ @@ -518,7 +521,7 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) ring->wptr = RREG32(mmUVD_RBC_RB_RPTR); WREG32(mmUVD_RBC_RB_WPTR, ring->wptr); - WREG32_P(mmUVD_RBC_RB_CNTL, 0, ~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK); + WREG32_FIELD(UVD_RBC_RB_CNTL, RB_NO_FETCH, 0); return 0; } @@ -581,6 +584,32 @@ static void uvd_v6_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq } /** + * uvd_v6_0_ring_emit_hdp_flush - emit an hdp flush + * + * @ring: amdgpu_ring pointer + * + * Emits an hdp flush. + */ +static void uvd_v6_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, PACKET0(mmHDP_MEM_COHERENCY_FLUSH_CNTL, 0)); + amdgpu_ring_write(ring, 0); +} + +/** + * uvd_v6_0_ring_hdp_invalidate - emit an hdp invalidate + * + * @ring: amdgpu_ring pointer + * + * Emits an hdp invalidate. + */ +static void uvd_v6_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, PACKET0(mmHDP_DEBUG0, 0)); + amdgpu_ring_write(ring, 1); +} + +/** * uvd_v6_0_ring_test_ring - register write test * * @ring: amdgpu_ring pointer @@ -634,6 +663,9 @@ static void uvd_v6_0_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib, unsigned vm_id, bool ctx_switch) { + amdgpu_ring_write(ring, PACKET0(mmUVD_LMI_RBC_IB_VMID, 0)); + amdgpu_ring_write(ring, vm_id); + amdgpu_ring_write(ring, PACKET0(mmUVD_LMI_RBC_IB_64BIT_BAR_LOW, 0)); amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); amdgpu_ring_write(ring, PACKET0(mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH, 0)); @@ -642,39 +674,55 @@ static void uvd_v6_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, ib->length_dw); } -/** - * uvd_v6_0_ring_test_ib - test ib execution - * - * @ring: amdgpu_ring pointer - * - * Test if we can successfully execute an IB - */ -static int uvd_v6_0_ring_test_ib(struct amdgpu_ring *ring) +static void uvd_v6_0_ring_emit_vm_flush(struct amdgpu_ring *ring, + unsigned vm_id, uint64_t pd_addr) { - struct fence *fence = NULL; - int r; + uint32_t reg; - r = amdgpu_uvd_get_create_msg(ring, 1, NULL); - if (r) { - DRM_ERROR("amdgpu: failed to get create msg (%d).\n", r); - goto error; - } + if (vm_id < 8) + reg = mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + vm_id; + else + reg = mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vm_id - 8; - r = amdgpu_uvd_get_destroy_msg(ring, 1, true, &fence); - if (r) { - DRM_ERROR("amdgpu: failed to get destroy ib (%d).\n", r); - goto error; - } + amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA0, 0)); + amdgpu_ring_write(ring, reg << 2); + amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA1, 0)); + amdgpu_ring_write(ring, pd_addr >> 12); + amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_CMD, 0)); + amdgpu_ring_write(ring, 0x8); - r = fence_wait(fence, false); - if (r) { - DRM_ERROR("amdgpu: fence wait failed (%d).\n", r); - goto error; - } - DRM_INFO("ib test on ring %d succeeded\n", ring->idx); -error: - fence_put(fence); - return r; + amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA0, 0)); + amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST << 2); + amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA1, 0)); + amdgpu_ring_write(ring, 1 << vm_id); + amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_CMD, 0)); + amdgpu_ring_write(ring, 0x8); + + amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA0, 0)); + amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST << 2); + amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA1, 0)); + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, PACKET0(mmUVD_GP_SCRATCH8, 0)); + amdgpu_ring_write(ring, 1 << vm_id); /* mask */ + amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_CMD, 0)); + amdgpu_ring_write(ring, 0xC); +} + +static void uvd_v6_0_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, PACKET0(mmUVD_GPCOM_VCPU_DATA0, 0)); + amdgpu_ring_write(ring, lower_32_bits(addr)); + amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA1, 0)); + amdgpu_ring_write(ring, upper_32_bits(addr)); + amdgpu_ring_write(ring, PACKET0(mmUVD_GP_SCRATCH8, 0)); + amdgpu_ring_write(ring, 0xffffffff); /* mask */ + amdgpu_ring_write(ring, PACKET0(mmUVD_GP_SCRATCH9, 0)); + amdgpu_ring_write(ring, seq); + amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_CMD, 0)); + amdgpu_ring_write(ring, 0xE); } static bool uvd_v6_0_is_idle(void *handle) @@ -690,20 +738,82 @@ static int uvd_v6_0_wait_for_idle(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->usec_timeout; i++) { - if (!(RREG32(mmSRBM_STATUS) & SRBM_STATUS__UVD_BUSY_MASK)) + if (uvd_v6_0_is_idle(handle)) return 0; } return -ETIMEDOUT; } -static int uvd_v6_0_soft_reset(void *handle) +#define AMDGPU_UVD_STATUS_BUSY_MASK 0xfd +static int uvd_v6_0_check_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; + u32 tmp = RREG32(mmSRBM_STATUS); + + if (REG_GET_FIELD(tmp, SRBM_STATUS, UVD_RQ_PENDING) || + REG_GET_FIELD(tmp, SRBM_STATUS, UVD_BUSY) || + (RREG32(mmUVD_STATUS) & AMDGPU_UVD_STATUS_BUSY_MASK)) + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_UVD, 1); + + if (srbm_soft_reset) { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_UVD].hang = true; + adev->uvd.srbm_soft_reset = srbm_soft_reset; + } else { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_UVD].hang = false; + adev->uvd.srbm_soft_reset = 0; + } + return 0; +} +static int uvd_v6_0_pre_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_UVD].hang) + return 0; + uvd_v6_0_stop(adev); + return 0; +} + +static int uvd_v6_0_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_UVD].hang) + return 0; + srbm_soft_reset = adev->uvd.srbm_soft_reset; + + if (srbm_soft_reset) { + u32 tmp; + + tmp = RREG32(mmSRBM_SOFT_RESET); + tmp |= srbm_soft_reset; + dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); + WREG32(mmSRBM_SOFT_RESET, tmp); + tmp = RREG32(mmSRBM_SOFT_RESET); + + udelay(50); + + tmp &= ~srbm_soft_reset; + WREG32(mmSRBM_SOFT_RESET, tmp); + tmp = RREG32(mmSRBM_SOFT_RESET); + + /* Wait a little for things to settle down */ + udelay(50); + } + + return 0; +} + +static int uvd_v6_0_post_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_UVD].hang) + return 0; - WREG32_P(mmSRBM_SOFT_RESET, SRBM_SOFT_RESET__SOFT_RESET_UVD_MASK, - ~SRBM_SOFT_RESET__SOFT_RESET_UVD_MASK); mdelay(5); return uvd_v6_0_start(adev); @@ -844,20 +954,15 @@ static int uvd_v6_0_set_clockgating_state(void *handle, enum amd_clockgating_state state) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - bool enable = (state == AMD_CG_STATE_GATE) ? true : false; - static int curstate = -1; - if (adev->asic_type == CHIP_FIJI) - uvd_v6_set_bypass_mode(adev, enable); + if (adev->asic_type == CHIP_FIJI || + adev->asic_type == CHIP_POLARIS10) + uvd_v6_set_bypass_mode(adev, state == AMD_CG_STATE_GATE ? true : false); if (!(adev->cg_flags & AMD_CG_SUPPORT_UVD_MGCG)) return 0; - if (curstate == state) - return 0; - - curstate = state; - if (enable) { + if (state == AMD_CG_STATE_GATE) { /* disable HW gating and enable Sw gating */ uvd_v6_0_set_sw_clock_gating(adev); } else { @@ -887,6 +992,8 @@ static int uvd_v6_0_set_powergating_state(void *handle, if (!(adev->pg_flags & AMD_PG_SUPPORT_UVD)) return 0; + WREG32(mmUVD_POWER_STATUS, UVD_POWER_STATUS__UVD_PG_EN_MASK); + if (state == AMD_PG_STATE_GATE) { uvd_v6_0_stop(adev); return 0; @@ -907,27 +1014,59 @@ const struct amd_ip_funcs uvd_v6_0_ip_funcs = { .resume = uvd_v6_0_resume, .is_idle = uvd_v6_0_is_idle, .wait_for_idle = uvd_v6_0_wait_for_idle, + .check_soft_reset = uvd_v6_0_check_soft_reset, + .pre_soft_reset = uvd_v6_0_pre_soft_reset, .soft_reset = uvd_v6_0_soft_reset, + .post_soft_reset = uvd_v6_0_post_soft_reset, .set_clockgating_state = uvd_v6_0_set_clockgating_state, .set_powergating_state = uvd_v6_0_set_powergating_state, }; -static const struct amdgpu_ring_funcs uvd_v6_0_ring_funcs = { +static const struct amdgpu_ring_funcs uvd_v6_0_ring_phys_funcs = { .get_rptr = uvd_v6_0_ring_get_rptr, .get_wptr = uvd_v6_0_ring_get_wptr, .set_wptr = uvd_v6_0_ring_set_wptr, .parse_cs = amdgpu_uvd_ring_parse_cs, .emit_ib = uvd_v6_0_ring_emit_ib, .emit_fence = uvd_v6_0_ring_emit_fence, + .emit_hdp_flush = uvd_v6_0_ring_emit_hdp_flush, + .emit_hdp_invalidate = uvd_v6_0_ring_emit_hdp_invalidate, .test_ring = uvd_v6_0_ring_test_ring, - .test_ib = uvd_v6_0_ring_test_ib, + .test_ib = amdgpu_uvd_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, + .begin_use = amdgpu_uvd_ring_begin_use, + .end_use = amdgpu_uvd_ring_end_use, +}; + +static const struct amdgpu_ring_funcs uvd_v6_0_ring_vm_funcs = { + .get_rptr = uvd_v6_0_ring_get_rptr, + .get_wptr = uvd_v6_0_ring_get_wptr, + .set_wptr = uvd_v6_0_ring_set_wptr, + .parse_cs = NULL, + .emit_ib = uvd_v6_0_ring_emit_ib, + .emit_fence = uvd_v6_0_ring_emit_fence, + .emit_vm_flush = uvd_v6_0_ring_emit_vm_flush, + .emit_pipeline_sync = uvd_v6_0_ring_emit_pipeline_sync, + .emit_hdp_flush = uvd_v6_0_ring_emit_hdp_flush, + .emit_hdp_invalidate = uvd_v6_0_ring_emit_hdp_invalidate, + .test_ring = uvd_v6_0_ring_test_ring, + .test_ib = amdgpu_uvd_ring_test_ib, + .insert_nop = amdgpu_ring_insert_nop, + .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) { - adev->uvd.ring.funcs = &uvd_v6_0_ring_funcs; + if (adev->asic_type >= CHIP_POLARIS10) { + adev->uvd.ring.funcs = &uvd_v6_0_ring_vm_funcs; + DRM_INFO("UVD is enabled in VM mode\n"); + } else { + adev->uvd.ring.funcs = &uvd_v6_0_ring_phys_funcs; + DRM_INFO("UVD is enabled in physical mode\n"); + } } static const struct amdgpu_irq_src_funcs uvd_v6_0_irq_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c index 45d92aceb485..5fa55b52c00e 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c @@ -40,6 +40,7 @@ #define VCE_V2_0_FW_SIZE (256 * 1024) #define VCE_V2_0_STACK_SIZE (64 * 1024) #define VCE_V2_0_DATA_SIZE (23552 * AMDGPU_MAX_VCE_HANDLES) +#define VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK 0x02 static void vce_v2_0_mc_resume(struct amdgpu_device *adev); static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev); @@ -96,6 +97,49 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring) WREG32(mmVCE_RB_WPTR2, ring->wptr); } +static int vce_v2_0_lmi_clean(struct amdgpu_device *adev) +{ + int i, j; + + for (i = 0; i < 10; ++i) { + for (j = 0; j < 100; ++j) { + uint32_t status = RREG32(mmVCE_LMI_STATUS); + + if (status & 0x337f) + return 0; + mdelay(10); + } + } + + return -ETIMEDOUT; +} + +static int vce_v2_0_firmware_loaded(struct amdgpu_device *adev) +{ + int i, j; + + for (i = 0; i < 10; ++i) { + for (j = 0; j < 100; ++j) { + uint32_t status = RREG32(mmVCE_STATUS); + + if (status & VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK) + return 0; + mdelay(10); + } + + DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n"); + WREG32_P(mmVCE_SOFT_RESET, + VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, + ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + mdelay(10); + WREG32_P(mmVCE_SOFT_RESET, 0, + ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + mdelay(10); + } + + return -ETIMEDOUT; +} + /** * vce_v2_0_start - start VCE block * @@ -106,7 +150,7 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring) static int vce_v2_0_start(struct amdgpu_device *adev) { struct amdgpu_ring *ring; - int i, j, r; + int r; vce_v2_0_mc_resume(adev); @@ -127,36 +171,12 @@ static int vce_v2_0_start(struct amdgpu_device *adev) WREG32(mmVCE_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); WREG32(mmVCE_RB_SIZE2, ring->ring_size / 4); - WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, ~VCE_VCPU_CNTL__CLK_EN_MASK); - - WREG32_P(mmVCE_SOFT_RESET, - VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, - ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - + WREG32_FIELD(VCE_VCPU_CNTL, CLK_EN, 1); + WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 1); mdelay(100); + WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 0); - WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - - for (i = 0; i < 10; ++i) { - uint32_t status; - for (j = 0; j < 100; ++j) { - status = RREG32(mmVCE_STATUS); - if (status & 2) - break; - mdelay(10); - } - r = 0; - if (status & 2) - break; - - DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n"); - WREG32_P(mmVCE_SOFT_RESET, VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, - ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - mdelay(10); - WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - mdelay(10); - r = -1; - } + r = vce_v2_0_firmware_loaded(adev); /* clear BUSY flag */ WREG32_P(mmVCE_STATUS, 0, ~1); @@ -338,47 +358,50 @@ static void vce_v2_0_set_sw_cg(struct amdgpu_device *adev, bool gated) static void vce_v2_0_set_dyn_cg(struct amdgpu_device *adev, bool gated) { - u32 orig, tmp; + if (vce_v2_0_wait_for_idle(adev)) { + DRM_INFO("VCE is busy, Can't set clock gateing"); + return; + } - if (gated) { - if (vce_v2_0_wait_for_idle(adev)) { - DRM_INFO("VCE is busy, Can't set clock gateing"); - return; - } - WREG32_P(mmVCE_VCPU_CNTL, 0, ~VCE_VCPU_CNTL__CLK_EN_MASK); - WREG32_P(mmVCE_SOFT_RESET, VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - mdelay(100); - WREG32(mmVCE_STATUS, 0); - } else { - WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, ~VCE_VCPU_CNTL__CLK_EN_MASK); - WREG32_P(mmVCE_SOFT_RESET, VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - mdelay(100); + WREG32_P(mmVCE_LMI_CTRL2, 0x100, ~0x100); + + if (vce_v2_0_lmi_clean(adev)) { + DRM_INFO("LMI is busy, Can't set clock gateing"); + return; } - tmp = RREG32(mmVCE_CLOCK_GATING_B); - tmp &= ~0x00060006; + WREG32_P(mmVCE_VCPU_CNTL, 0, ~VCE_VCPU_CNTL__CLK_EN_MASK); + WREG32_P(mmVCE_SOFT_RESET, + VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, + ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + WREG32(mmVCE_STATUS, 0); + + if (gated) + WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0); + /* LMI_MC/LMI_UMC always set in dynamic, set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {0, 0} */ if (gated) { - tmp |= 0xe10000; + /* Force CLOCK OFF , set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {*, 1} */ + WREG32(mmVCE_CLOCK_GATING_B, 0xe90010); } else { - tmp |= 0xe1; - tmp &= ~0xe10000; + /* Force CLOCK ON, set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {1, 0} */ + WREG32(mmVCE_CLOCK_GATING_B, 0x800f1); } - WREG32(mmVCE_CLOCK_GATING_B, tmp); - orig = tmp = RREG32(mmVCE_UENC_CLOCK_GATING); - tmp &= ~0x1fe000; - tmp &= ~0xff000000; - if (tmp != orig) - WREG32(mmVCE_UENC_CLOCK_GATING, tmp); + /* Set VCE_UENC_CLOCK_GATING always in dynamic mode {*_FORCE_ON, *_FORCE_OFF} = {0, 0}*/; + WREG32(mmVCE_UENC_CLOCK_GATING, 0x40); - orig = tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING); - tmp &= ~0x3fc; - if (tmp != orig) - WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp); + /* set VCE_UENC_REG_CLOCK_GATING always in dynamic mode */ + WREG32(mmVCE_UENC_REG_CLOCK_GATING, 0x00); - if (gated) - WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0); - WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + WREG32_P(mmVCE_LMI_CTRL2, 0, ~0x100); + if(!gated) { + WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, ~VCE_VCPU_CNTL__CLK_EN_MASK); + mdelay(100); + WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + + vce_v2_0_firmware_loaded(adev); + WREG32_P(mmVCE_STATUS, 0, ~VCE_STATUS__JOB_BUSY_MASK); + } } static void vce_v2_0_disable_cg(struct amdgpu_device *adev) @@ -458,9 +481,7 @@ static void vce_v2_0_mc_resume(struct amdgpu_device *adev) WREG32(mmVCE_VCPU_CACHE_SIZE2, size); WREG32_P(mmVCE_LMI_CTRL2, 0x0, ~0x100); - - WREG32_P(mmVCE_SYS_INT_EN, VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK, - ~VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK); + WREG32_FIELD(VCE_SYS_INT_EN, VCE_SYS_INT_TRAP_INTERRUPT_EN, 1); vce_v2_0_init_cg(adev); } @@ -474,11 +495,11 @@ static bool vce_v2_0_is_idle(void *handle) static int vce_v2_0_wait_for_idle(void *handle) { - unsigned i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + unsigned i; for (i = 0; i < adev->usec_timeout; i++) { - if (!(RREG32(mmSRBM_STATUS2) & SRBM_STATUS2__VCE_BUSY_MASK)) + if (vce_v2_0_is_idle(handle)) return 0; } return -ETIMEDOUT; @@ -488,8 +509,7 @@ static int vce_v2_0_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - WREG32_P(mmSRBM_SOFT_RESET, SRBM_SOFT_RESET__SOFT_RESET_VCE_MASK, - ~SRBM_SOFT_RESET__SOFT_RESET_VCE_MASK); + WREG32_FIELD(SRBM_SOFT_RESET, SOFT_RESET_VCE, 1); mdelay(5); return vce_v2_0_start(adev); @@ -516,10 +536,8 @@ static int vce_v2_0_process_interrupt(struct amdgpu_device *adev, DRM_DEBUG("IH: VCE\n"); switch (entry->src_data) { case 0: - amdgpu_fence_process(&adev->vce.ring[0]); - break; case 1: - amdgpu_fence_process(&adev->vce.ring[1]); + amdgpu_fence_process(&adev->vce.ring[entry->src_data]); break; default: DRM_ERROR("Unhandled interrupt: %d %d\n", @@ -594,6 +612,8 @@ static const struct amdgpu_ring_funcs vce_v2_0_ring_funcs = { .test_ib = amdgpu_vce_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, + .begin_use = amdgpu_vce_ring_begin_use, + .end_use = amdgpu_vce_ring_end_use, }; static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c index 30e8099e94c5..615b8b16ad04 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c @@ -37,12 +37,16 @@ #include "gca/gfx_8_0_d.h" #include "smu/smu_7_1_2_d.h" #include "smu/smu_7_1_2_sh_mask.h" +#include "gca/gfx_8_0_d.h" +#include "gca/gfx_8_0_sh_mask.h" + #define GRBM_GFX_INDEX__VCE_INSTANCE__SHIFT 0x04 #define GRBM_GFX_INDEX__VCE_INSTANCE_MASK 0x10 #define mmVCE_LMI_VCPU_CACHE_40BIT_BAR0 0x8616 #define mmVCE_LMI_VCPU_CACHE_40BIT_BAR1 0x8617 #define mmVCE_LMI_VCPU_CACHE_40BIT_BAR2 0x8618 +#define VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK 0x02 #define VCE_V3_0_FW_SIZE (384 * 1024) #define VCE_V3_0_STACK_SIZE (64 * 1024) @@ -51,6 +55,7 @@ static void vce_v3_0_mc_resume(struct amdgpu_device *adev, int idx); static void vce_v3_0_set_ring_funcs(struct amdgpu_device *adev); static void vce_v3_0_set_irq_funcs(struct amdgpu_device *adev); +static int vce_v3_0_wait_for_idle(void *handle); /** * vce_v3_0_ring_get_rptr - get read pointer @@ -105,106 +110,99 @@ static void vce_v3_0_ring_set_wptr(struct amdgpu_ring *ring) static void vce_v3_0_override_vce_clock_gating(struct amdgpu_device *adev, bool override) { - u32 tmp, data; - - tmp = data = RREG32(mmVCE_RB_ARB_CTRL); - if (override) - data |= VCE_RB_ARB_CTRL__VCE_CGTT_OVERRIDE_MASK; - else - data &= ~VCE_RB_ARB_CTRL__VCE_CGTT_OVERRIDE_MASK; - - if (tmp != data) - WREG32(mmVCE_RB_ARB_CTRL, data); + WREG32_FIELD(VCE_RB_ARB_CTRL, VCE_CGTT_OVERRIDE, override ? 1 : 0); } static void vce_v3_0_set_vce_sw_clock_gating(struct amdgpu_device *adev, bool gated) { - u32 tmp, data; + u32 data; + /* Set Override to disable Clock Gating */ vce_v3_0_override_vce_clock_gating(adev, true); - if (!gated) { - /* Force CLOCK ON for VCE_CLOCK_GATING_B, - * {*_FORCE_ON, *_FORCE_OFF} = {1, 0} - * VREG can be FORCE ON or set to Dynamic, but can't be OFF - */ - tmp = data = RREG32(mmVCE_CLOCK_GATING_B); + /* This function enables MGCG which is controlled by firmware. + With the clocks in the gated state the core is still + accessible but the firmware will throttle the clocks on the + fly as necessary. + */ + if (gated) { + data = RREG32(mmVCE_CLOCK_GATING_B); data |= 0x1ff; data &= ~0xef0000; - if (tmp != data) - WREG32(mmVCE_CLOCK_GATING_B, data); + WREG32(mmVCE_CLOCK_GATING_B, data); - /* Force CLOCK ON for VCE_UENC_CLOCK_GATING, - * {*_FORCE_ON, *_FORCE_OFF} = {1, 0} - */ - tmp = data = RREG32(mmVCE_UENC_CLOCK_GATING); + data = RREG32(mmVCE_UENC_CLOCK_GATING); data |= 0x3ff000; data &= ~0xffc00000; - if (tmp != data) - WREG32(mmVCE_UENC_CLOCK_GATING, data); + WREG32(mmVCE_UENC_CLOCK_GATING, data); - /* set VCE_UENC_CLOCK_GATING_2 */ - tmp = data = RREG32(mmVCE_UENC_CLOCK_GATING_2); + data = RREG32(mmVCE_UENC_CLOCK_GATING_2); data |= 0x2; - data &= ~0x2; - if (tmp != data) - WREG32(mmVCE_UENC_CLOCK_GATING_2, data); + data &= ~0x00010000; + WREG32(mmVCE_UENC_CLOCK_GATING_2, data); - /* Force CLOCK ON for VCE_UENC_REG_CLOCK_GATING */ - tmp = data = RREG32(mmVCE_UENC_REG_CLOCK_GATING); + data = RREG32(mmVCE_UENC_REG_CLOCK_GATING); data |= 0x37f; - if (tmp != data) - WREG32(mmVCE_UENC_REG_CLOCK_GATING, data); + WREG32(mmVCE_UENC_REG_CLOCK_GATING, data); - /* Force VCE_UENC_DMA_DCLK_CTRL Clock ON */ - tmp = data = RREG32(mmVCE_UENC_DMA_DCLK_CTRL); + data = RREG32(mmVCE_UENC_DMA_DCLK_CTRL); data |= VCE_UENC_DMA_DCLK_CTRL__WRDMCLK_FORCEON_MASK | - VCE_UENC_DMA_DCLK_CTRL__RDDMCLK_FORCEON_MASK | - VCE_UENC_DMA_DCLK_CTRL__REGCLK_FORCEON_MASK | - 0x8; - if (tmp != data) - WREG32(mmVCE_UENC_DMA_DCLK_CTRL, data); + VCE_UENC_DMA_DCLK_CTRL__RDDMCLK_FORCEON_MASK | + VCE_UENC_DMA_DCLK_CTRL__REGCLK_FORCEON_MASK | + 0x8; + WREG32(mmVCE_UENC_DMA_DCLK_CTRL, data); } else { - /* Force CLOCK OFF for VCE_CLOCK_GATING_B, - * {*, *_FORCE_OFF} = {*, 1} - * set VREG to Dynamic, as it can't be OFF - */ - tmp = data = RREG32(mmVCE_CLOCK_GATING_B); + data = RREG32(mmVCE_CLOCK_GATING_B); data &= ~0x80010; data |= 0xe70008; - if (tmp != data) - WREG32(mmVCE_CLOCK_GATING_B, data); - /* Force CLOCK OFF for VCE_UENC_CLOCK_GATING, - * Force ClOCK OFF takes precedent over Force CLOCK ON setting. - * {*_FORCE_ON, *_FORCE_OFF} = {*, 1} - */ - tmp = data = RREG32(mmVCE_UENC_CLOCK_GATING); + WREG32(mmVCE_CLOCK_GATING_B, data); + + data = RREG32(mmVCE_UENC_CLOCK_GATING); data |= 0xffc00000; - if (tmp != data) - WREG32(mmVCE_UENC_CLOCK_GATING, data); - /* Set VCE_UENC_CLOCK_GATING_2 */ - tmp = data = RREG32(mmVCE_UENC_CLOCK_GATING_2); + WREG32(mmVCE_UENC_CLOCK_GATING, data); + + data = RREG32(mmVCE_UENC_CLOCK_GATING_2); data |= 0x10000; - if (tmp != data) - WREG32(mmVCE_UENC_CLOCK_GATING_2, data); - /* Set VCE_UENC_REG_CLOCK_GATING to dynamic */ - tmp = data = RREG32(mmVCE_UENC_REG_CLOCK_GATING); + WREG32(mmVCE_UENC_CLOCK_GATING_2, data); + + data = RREG32(mmVCE_UENC_REG_CLOCK_GATING); data &= ~0xffc00000; - if (tmp != data) - WREG32(mmVCE_UENC_REG_CLOCK_GATING, data); - /* Set VCE_UENC_DMA_DCLK_CTRL CG always in dynamic mode */ - tmp = data = RREG32(mmVCE_UENC_DMA_DCLK_CTRL); + WREG32(mmVCE_UENC_REG_CLOCK_GATING, data); + + data = RREG32(mmVCE_UENC_DMA_DCLK_CTRL); data &= ~(VCE_UENC_DMA_DCLK_CTRL__WRDMCLK_FORCEON_MASK | - VCE_UENC_DMA_DCLK_CTRL__RDDMCLK_FORCEON_MASK | - VCE_UENC_DMA_DCLK_CTRL__REGCLK_FORCEON_MASK | - 0x8); - if (tmp != data) - WREG32(mmVCE_UENC_DMA_DCLK_CTRL, data); + VCE_UENC_DMA_DCLK_CTRL__RDDMCLK_FORCEON_MASK | + VCE_UENC_DMA_DCLK_CTRL__REGCLK_FORCEON_MASK | + 0x8); + WREG32(mmVCE_UENC_DMA_DCLK_CTRL, data); } vce_v3_0_override_vce_clock_gating(adev, false); } +static int vce_v3_0_firmware_loaded(struct amdgpu_device *adev) +{ + int i, j; + + for (i = 0; i < 10; ++i) { + for (j = 0; j < 100; ++j) { + uint32_t status = RREG32(mmVCE_STATUS); + + if (status & VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK) + return 0; + mdelay(10); + } + + DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n"); + WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 1); + mdelay(10); + WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 0); + mdelay(10); + } + + return -ETIMEDOUT; +} + /** * vce_v3_0_start - start VCE block * @@ -215,70 +213,43 @@ static void vce_v3_0_set_vce_sw_clock_gating(struct amdgpu_device *adev, static int vce_v3_0_start(struct amdgpu_device *adev) { struct amdgpu_ring *ring; - int idx, i, j, r; + int idx, r; + + ring = &adev->vce.ring[0]; + WREG32(mmVCE_RB_RPTR, ring->wptr); + WREG32(mmVCE_RB_WPTR, ring->wptr); + WREG32(mmVCE_RB_BASE_LO, ring->gpu_addr); + WREG32(mmVCE_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); + WREG32(mmVCE_RB_SIZE, ring->ring_size / 4); + + ring = &adev->vce.ring[1]; + WREG32(mmVCE_RB_RPTR2, ring->wptr); + WREG32(mmVCE_RB_WPTR2, ring->wptr); + WREG32(mmVCE_RB_BASE_LO2, ring->gpu_addr); + WREG32(mmVCE_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); + WREG32(mmVCE_RB_SIZE2, ring->ring_size / 4); mutex_lock(&adev->grbm_idx_mutex); for (idx = 0; idx < 2; ++idx) { - if (adev->vce.harvest_config & (1 << idx)) continue; - if (idx == 0) - WREG32_P(mmGRBM_GFX_INDEX, 0, - ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); - else - WREG32_P(mmGRBM_GFX_INDEX, - GRBM_GFX_INDEX__VCE_INSTANCE_MASK, - ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); - + WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, idx); vce_v3_0_mc_resume(adev, idx); + WREG32_FIELD(VCE_STATUS, JOB_BUSY, 1); - /* set BUSY flag */ - WREG32_P(mmVCE_STATUS, 1, ~1); if (adev->asic_type >= CHIP_STONEY) WREG32_P(mmVCE_VCPU_CNTL, 1, ~0x200001); else - WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, - ~VCE_VCPU_CNTL__CLK_EN_MASK); - - WREG32_P(mmVCE_SOFT_RESET, - VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, - ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); + WREG32_FIELD(VCE_VCPU_CNTL, CLK_EN, 1); + WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 0); mdelay(100); - WREG32_P(mmVCE_SOFT_RESET, 0, - ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - - for (i = 0; i < 10; ++i) { - uint32_t status; - for (j = 0; j < 100; ++j) { - status = RREG32(mmVCE_STATUS); - if (status & 2) - break; - mdelay(10); - } - r = 0; - if (status & 2) - break; - - DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n"); - WREG32_P(mmVCE_SOFT_RESET, - VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, - ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - mdelay(10); - WREG32_P(mmVCE_SOFT_RESET, 0, - ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); - mdelay(10); - r = -1; - } + r = vce_v3_0_firmware_loaded(adev); /* clear BUSY flag */ - WREG32_P(mmVCE_STATUS, 0, ~1); - - /* Set Clock-Gating off */ - if (adev->cg_flags & AMD_CG_SUPPORT_VCE_MGCG) - vce_v3_0_set_vce_sw_clock_gating(adev, false); + WREG32_FIELD(VCE_STATUS, JOB_BUSY, 0); if (r) { DRM_ERROR("VCE not responding, giving up!!!\n"); @@ -287,22 +258,41 @@ static int vce_v3_0_start(struct amdgpu_device *adev) } } - WREG32_P(mmGRBM_GFX_INDEX, 0, ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); + WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, 0); mutex_unlock(&adev->grbm_idx_mutex); - ring = &adev->vce.ring[0]; - WREG32(mmVCE_RB_RPTR, ring->wptr); - WREG32(mmVCE_RB_WPTR, ring->wptr); - WREG32(mmVCE_RB_BASE_LO, ring->gpu_addr); - WREG32(mmVCE_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); - WREG32(mmVCE_RB_SIZE, ring->ring_size / 4); + return 0; +} - ring = &adev->vce.ring[1]; - WREG32(mmVCE_RB_RPTR2, ring->wptr); - WREG32(mmVCE_RB_WPTR2, ring->wptr); - WREG32(mmVCE_RB_BASE_LO2, ring->gpu_addr); - WREG32(mmVCE_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); - WREG32(mmVCE_RB_SIZE2, ring->ring_size / 4); +static int vce_v3_0_stop(struct amdgpu_device *adev) +{ + int idx; + + mutex_lock(&adev->grbm_idx_mutex); + for (idx = 0; idx < 2; ++idx) { + if (adev->vce.harvest_config & (1 << idx)) + continue; + + WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, idx); + + if (adev->asic_type >= CHIP_STONEY) + WREG32_P(mmVCE_VCPU_CNTL, 0, ~0x200001); + else + WREG32_FIELD(VCE_VCPU_CNTL, CLK_EN, 0); + + /* hold on ECPU */ + WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 1); + + /* clear BUSY flag */ + WREG32_FIELD(VCE_STATUS, JOB_BUSY, 0); + + /* Set Clock-Gating off */ + if (adev->cg_flags & AMD_CG_SUPPORT_VCE_MGCG) + vce_v3_0_set_vce_sw_clock_gating(adev, false); + } + + WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, 0); + mutex_unlock(&adev->grbm_idx_mutex); return 0; } @@ -441,7 +431,14 @@ static int vce_v3_0_hw_init(void *handle) static int vce_v3_0_hw_fini(void *handle) { - return 0; + int r; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + r = vce_v3_0_wait_for_idle(handle); + if (r) + return r; + + return vce_v3_0_stop(adev); } static int vce_v3_0_suspend(void *handle) @@ -483,7 +480,7 @@ static void vce_v3_0_mc_resume(struct amdgpu_device *adev, int idx) WREG32_P(mmVCE_CLOCK_GATING_A, 0, ~(1 << 16)); WREG32_P(mmVCE_UENC_CLOCK_GATING, 0x1FF000, ~0xFF9FF000); WREG32_P(mmVCE_UENC_REG_CLOCK_GATING, 0x3F, ~0x3F); - WREG32(mmVCE_CLOCK_GATING_B, 0xf7); + WREG32(mmVCE_CLOCK_GATING_B, 0x1FF); WREG32(mmVCE_LMI_CTRL, 0x00398000); WREG32_P(mmVCE_LMI_CACHE_CTRL, 0x0, ~0x1); @@ -522,9 +519,7 @@ static void vce_v3_0_mc_resume(struct amdgpu_device *adev, int idx) } WREG32_P(mmVCE_LMI_CTRL2, 0x0, ~0x100); - - WREG32_P(mmVCE_SYS_INT_EN, VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK, - ~VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK); + WREG32_FIELD(VCE_SYS_INT_EN, VCE_SYS_INT_TRAP_INTERRUPT_EN, 1); } static bool vce_v3_0_is_idle(void *handle) @@ -550,20 +545,108 @@ static int vce_v3_0_wait_for_idle(void *handle) return -ETIMEDOUT; } +#define VCE_STATUS_VCPU_REPORT_AUTO_BUSY_MASK 0x00000008L /* AUTO_BUSY */ +#define VCE_STATUS_VCPU_REPORT_RB0_BUSY_MASK 0x00000010L /* RB0_BUSY */ +#define VCE_STATUS_VCPU_REPORT_RB1_BUSY_MASK 0x00000020L /* RB1_BUSY */ +#define AMDGPU_VCE_STATUS_BUSY_MASK (VCE_STATUS_VCPU_REPORT_AUTO_BUSY_MASK | \ + VCE_STATUS_VCPU_REPORT_RB0_BUSY_MASK) + +static int vce_v3_0_check_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + u32 srbm_soft_reset = 0; + + /* According to VCE team , we should use VCE_STATUS instead + * SRBM_STATUS.VCE_BUSY bit for busy status checking. + * GRBM_GFX_INDEX.INSTANCE_INDEX is used to specify which VCE + * instance's registers are accessed + * (0 for 1st instance, 10 for 2nd instance). + * + *VCE_STATUS + *|UENC|ACPI|AUTO ACTIVE|RB1 |RB0 |RB2 | |FW_LOADED|JOB | + *|----+----+-----------+----+----+----+----------+---------+----| + *|bit8|bit7| bit6 |bit5|bit4|bit3| bit2 | bit1 |bit0| + * + * VCE team suggest use bit 3--bit 6 for busy status check + */ + mutex_lock(&adev->grbm_idx_mutex); + WREG32_FIELD(GRBM_GFX_INDEX, INSTANCE_INDEX, 0); + if (RREG32(mmVCE_STATUS) & AMDGPU_VCE_STATUS_BUSY_MASK) { + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_VCE0, 1); + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_VCE1, 1); + } + WREG32_FIELD(GRBM_GFX_INDEX, INSTANCE_INDEX, 0x10); + if (RREG32(mmVCE_STATUS) & AMDGPU_VCE_STATUS_BUSY_MASK) { + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_VCE0, 1); + srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_VCE1, 1); + } + WREG32_FIELD(GRBM_GFX_INDEX, INSTANCE_INDEX, 0); + + if (srbm_soft_reset) { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_VCE].hang = true; + adev->vce.srbm_soft_reset = srbm_soft_reset; + } else { + adev->ip_block_status[AMD_IP_BLOCK_TYPE_VCE].hang = false; + adev->vce.srbm_soft_reset = 0; + } + mutex_unlock(&adev->grbm_idx_mutex); + return 0; +} + static int vce_v3_0_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - u32 mask = 0; + u32 srbm_soft_reset; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_VCE].hang) + return 0; + srbm_soft_reset = adev->vce.srbm_soft_reset; + + if (srbm_soft_reset) { + u32 tmp; + + tmp = RREG32(mmSRBM_SOFT_RESET); + tmp |= srbm_soft_reset; + dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); + WREG32(mmSRBM_SOFT_RESET, tmp); + tmp = RREG32(mmSRBM_SOFT_RESET); + + udelay(50); + + tmp &= ~srbm_soft_reset; + WREG32(mmSRBM_SOFT_RESET, tmp); + tmp = RREG32(mmSRBM_SOFT_RESET); + + /* Wait a little for things to settle down */ + udelay(50); + } + + return 0; +} + +static int vce_v3_0_pre_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_VCE].hang) + return 0; - mask |= (adev->vce.harvest_config & AMDGPU_VCE_HARVEST_VCE0) ? 0 : SRBM_SOFT_RESET__SOFT_RESET_VCE0_MASK; - mask |= (adev->vce.harvest_config & AMDGPU_VCE_HARVEST_VCE1) ? 0 : SRBM_SOFT_RESET__SOFT_RESET_VCE1_MASK; + mdelay(5); + + return vce_v3_0_suspend(adev); +} + + +static int vce_v3_0_post_soft_reset(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (!adev->ip_block_status[AMD_IP_BLOCK_TYPE_VCE].hang) + return 0; - WREG32_P(mmSRBM_SOFT_RESET, mask, - ~(SRBM_SOFT_RESET__SOFT_RESET_VCE0_MASK | - SRBM_SOFT_RESET__SOFT_RESET_VCE1_MASK)); mdelay(5); - return vce_v3_0_start(adev); + return vce_v3_0_resume(adev); } static int vce_v3_0_set_interrupt_state(struct amdgpu_device *adev, @@ -586,9 +669,7 @@ static int vce_v3_0_process_interrupt(struct amdgpu_device *adev, { DRM_DEBUG("IH: VCE\n"); - WREG32_P(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); + WREG32_FIELD(VCE_SYS_INT_STATUS, VCE_SYS_INT_TRAP_INTERRUPT_INT, 1); switch (entry->src_data) { case 0: @@ -604,6 +685,18 @@ static int vce_v3_0_process_interrupt(struct amdgpu_device *adev, return 0; } +static void vce_v3_set_bypass_mode(struct amdgpu_device *adev, bool enable) +{ + u32 tmp = RREG32_SMC(ixGCK_DFS_BYPASS_CNTL); + + if (enable) + tmp |= GCK_DFS_BYPASS_CNTL__BYPASSECLK_MASK; + else + tmp &= ~GCK_DFS_BYPASS_CNTL__BYPASSECLK_MASK; + + WREG32_SMC(ixGCK_DFS_BYPASS_CNTL, tmp); +} + static int vce_v3_0_set_clockgating_state(void *handle, enum amd_clockgating_state state) { @@ -611,6 +704,9 @@ static int vce_v3_0_set_clockgating_state(void *handle, bool enable = (state == AMD_CG_STATE_GATE) ? true : false; int i; + if (adev->asic_type == CHIP_POLARIS10) + vce_v3_set_bypass_mode(adev, enable); + if (!(adev->cg_flags & AMD_CG_SUPPORT_VCE_MGCG)) return 0; @@ -620,13 +716,7 @@ static int vce_v3_0_set_clockgating_state(void *handle, if (adev->vce.harvest_config & (1 << i)) continue; - if (i == 0) - WREG32_P(mmGRBM_GFX_INDEX, 0, - ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); - else - WREG32_P(mmGRBM_GFX_INDEX, - GRBM_GFX_INDEX__VCE_INSTANCE_MASK, - ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); + WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, i); if (enable) { /* initialize VCE_CLOCK_GATING_A: Clock ON/OFF delay */ @@ -645,7 +735,7 @@ static int vce_v3_0_set_clockgating_state(void *handle, vce_v3_0_set_vce_sw_clock_gating(adev, enable); } - WREG32_P(mmGRBM_GFX_INDEX, 0, ~GRBM_GFX_INDEX__VCE_INSTANCE_MASK); + WREG32_FIELD(GRBM_GFX_INDEX, VCE_INSTANCE, 0); mutex_unlock(&adev->grbm_idx_mutex); return 0; @@ -685,7 +775,10 @@ const struct amd_ip_funcs vce_v3_0_ip_funcs = { .resume = vce_v3_0_resume, .is_idle = vce_v3_0_is_idle, .wait_for_idle = vce_v3_0_wait_for_idle, + .check_soft_reset = vce_v3_0_check_soft_reset, + .pre_soft_reset = vce_v3_0_pre_soft_reset, .soft_reset = vce_v3_0_soft_reset, + .post_soft_reset = vce_v3_0_post_soft_reset, .set_clockgating_state = vce_v3_0_set_clockgating_state, .set_powergating_state = vce_v3_0_set_powergating_state, }; @@ -701,6 +794,8 @@ static const struct amdgpu_ring_funcs vce_v3_0_ring_funcs = { .test_ib = amdgpu_vce_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, + .begin_use = amdgpu_vce_ring_begin_use, + .end_use = amdgpu_vce_ring_end_use, }; static void vce_v3_0_set_ring_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index a65c96029476..f2e8aa1a0dbd 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -77,6 +77,7 @@ #if defined(CONFIG_DRM_AMD_ACP) #include "amdgpu_acp.h" #endif +#include "dce_virtual.h" MODULE_FIRMWARE("amdgpu/polaris10_smc.bin"); MODULE_FIRMWARE("amdgpu/polaris10_smc_sk.bin"); @@ -203,6 +204,29 @@ static void vi_didt_wreg(struct amdgpu_device *adev, u32 reg, u32 v) spin_unlock_irqrestore(&adev->didt_idx_lock, flags); } +static u32 vi_gc_cac_rreg(struct amdgpu_device *adev, u32 reg) +{ + unsigned long flags; + u32 r; + + spin_lock_irqsave(&adev->gc_cac_idx_lock, flags); + WREG32(mmGC_CAC_IND_INDEX, (reg)); + r = RREG32(mmGC_CAC_IND_DATA); + spin_unlock_irqrestore(&adev->gc_cac_idx_lock, flags); + return r; +} + +static void vi_gc_cac_wreg(struct amdgpu_device *adev, u32 reg, u32 v) +{ + unsigned long flags; + + spin_lock_irqsave(&adev->gc_cac_idx_lock, flags); + WREG32(mmGC_CAC_IND_INDEX, (reg)); + WREG32(mmGC_CAC_IND_DATA, (v)); + spin_unlock_irqrestore(&adev->gc_cac_idx_lock, flags); +} + + static const u32 tonga_mgcg_cgcg_init[] = { mmCGTT_DRM_CLK_CTRL0, 0xffffffff, 0x00600100, @@ -533,12 +557,12 @@ static uint32_t vi_read_indexed_register(struct amdgpu_device *adev, u32 se_num, mutex_lock(&adev->grbm_idx_mutex); if (se_num != 0xffffffff || sh_num != 0xffffffff) - gfx_v8_0_select_se_sh(adev, se_num, sh_num); + amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff); val = RREG32(reg_offset); if (se_num != 0xffffffff || sh_num != 0xffffffff) - gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff); + amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); mutex_unlock(&adev->grbm_idx_mutex); return val; } @@ -597,7 +621,7 @@ static int vi_read_register(struct amdgpu_device *adev, u32 se_num, return -EINVAL; } -static void vi_gpu_pci_config_reset(struct amdgpu_device *adev) +static int vi_gpu_pci_config_reset(struct amdgpu_device *adev) { u32 i; @@ -612,11 +636,14 @@ static void vi_gpu_pci_config_reset(struct amdgpu_device *adev) /* wait for asic to come out of reset */ for (i = 0; i < adev->usec_timeout; i++) { - if (RREG32(mmCONFIG_MEMSIZE) != 0xffffffff) - break; + if (RREG32(mmCONFIG_MEMSIZE) != 0xffffffff) { + /* enable BM */ + pci_set_master(adev->pdev); + return 0; + } udelay(1); } - + return -EINVAL; } static void vi_set_bios_scratch_engine_hung(struct amdgpu_device *adev, bool hung) @@ -642,13 +669,15 @@ static void vi_set_bios_scratch_engine_hung(struct amdgpu_device *adev, bool hun */ static int vi_asic_reset(struct amdgpu_device *adev) { + int r; + vi_set_bios_scratch_engine_hung(adev, true); - vi_gpu_pci_config_reset(adev); + r = vi_gpu_pci_config_reset(adev); vi_set_bios_scratch_engine_hung(adev, false); - return 0; + return r; } static int vi_set_uvd_clock(struct amdgpu_device *adev, u32 clock, @@ -794,6 +823,60 @@ static const struct amdgpu_ip_block_version topaz_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version topaz_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vi_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 7, + .minor = 4, + .rev = 0, + .funcs = &gmc_v7_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 2, + .minor = 4, + .rev = 0, + .funcs = &iceland_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 1, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 1, + .minor = 0, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gfx_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 2, + .minor = 4, + .rev = 0, + .funcs = &sdma_v2_4_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version tonga_ip_blocks[] = { /* ORDER MATTERS! */ @@ -862,6 +945,74 @@ static const struct amdgpu_ip_block_version tonga_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version tonga_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vi_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gmc_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &tonga_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 1, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 10, + .minor = 0, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gfx_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &sdma_v3_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 5, + .minor = 0, + .rev = 0, + .funcs = &uvd_v5_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &vce_v3_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version fiji_ip_blocks[] = { /* ORDER MATTERS! */ @@ -930,6 +1081,74 @@ static const struct amdgpu_ip_block_version fiji_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version fiji_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vi_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 8, + .minor = 5, + .rev = 0, + .funcs = &gmc_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &tonga_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 1, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 10, + .minor = 1, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gfx_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &sdma_v3_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 6, + .minor = 0, + .rev = 0, + .funcs = &uvd_v6_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &vce_v3_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version polaris11_ip_blocks[] = { /* ORDER MATTERS! */ @@ -998,6 +1217,74 @@ static const struct amdgpu_ip_block_version polaris11_ip_blocks[] = }, }; +static const struct amdgpu_ip_block_version polaris11_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vi_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 8, + .minor = 1, + .rev = 0, + .funcs = &gmc_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 3, + .minor = 1, + .rev = 0, + .funcs = &tonga_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 7, + .minor = 2, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 11, + .minor = 2, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gfx_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 3, + .minor = 1, + .rev = 0, + .funcs = &sdma_v3_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 6, + .minor = 3, + .rev = 0, + .funcs = &uvd_v6_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 3, + .minor = 4, + .rev = 0, + .funcs = &vce_v3_0_ip_funcs, + }, +}; + static const struct amdgpu_ip_block_version cz_ip_blocks[] = { /* ORDER MATTERS! */ @@ -1075,34 +1362,142 @@ static const struct amdgpu_ip_block_version cz_ip_blocks[] = #endif }; +static const struct amdgpu_ip_block_version cz_ip_blocks_vd[] = +{ + /* ORDER MATTERS! */ + { + .type = AMD_IP_BLOCK_TYPE_COMMON, + .major = 2, + .minor = 0, + .rev = 0, + .funcs = &vi_common_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GMC, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gmc_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_IH, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &cz_ih_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SMC, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &amdgpu_pp_ip_funcs + }, + { + .type = AMD_IP_BLOCK_TYPE_DCE, + .major = 11, + .minor = 0, + .rev = 0, + .funcs = &dce_virtual_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_GFX, + .major = 8, + .minor = 0, + .rev = 0, + .funcs = &gfx_v8_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_SDMA, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &sdma_v3_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_UVD, + .major = 6, + .minor = 0, + .rev = 0, + .funcs = &uvd_v6_0_ip_funcs, + }, + { + .type = AMD_IP_BLOCK_TYPE_VCE, + .major = 3, + .minor = 0, + .rev = 0, + .funcs = &vce_v3_0_ip_funcs, + }, +#if defined(CONFIG_DRM_AMD_ACP) + { + .type = AMD_IP_BLOCK_TYPE_ACP, + .major = 2, + .minor = 2, + .rev = 0, + .funcs = &acp_ip_funcs, + }, +#endif +}; + int vi_set_ip_blocks(struct amdgpu_device *adev) { - switch (adev->asic_type) { - case CHIP_TOPAZ: - adev->ip_blocks = topaz_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(topaz_ip_blocks); - break; - case CHIP_FIJI: - adev->ip_blocks = fiji_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(fiji_ip_blocks); - break; - case CHIP_TONGA: - adev->ip_blocks = tonga_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(tonga_ip_blocks); - break; - case CHIP_POLARIS11: - case CHIP_POLARIS10: - adev->ip_blocks = polaris11_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(polaris11_ip_blocks); - break; - case CHIP_CARRIZO: - case CHIP_STONEY: - adev->ip_blocks = cz_ip_blocks; - adev->num_ip_blocks = ARRAY_SIZE(cz_ip_blocks); - break; - default: - /* FIXME: not supported yet */ - return -EINVAL; + if (adev->enable_virtual_display) { + switch (adev->asic_type) { + case CHIP_TOPAZ: + adev->ip_blocks = topaz_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(topaz_ip_blocks_vd); + break; + case CHIP_FIJI: + adev->ip_blocks = fiji_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(fiji_ip_blocks_vd); + break; + case CHIP_TONGA: + adev->ip_blocks = tonga_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(tonga_ip_blocks_vd); + break; + case CHIP_POLARIS11: + case CHIP_POLARIS10: + adev->ip_blocks = polaris11_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(polaris11_ip_blocks_vd); + break; + + case CHIP_CARRIZO: + case CHIP_STONEY: + adev->ip_blocks = cz_ip_blocks_vd; + adev->num_ip_blocks = ARRAY_SIZE(cz_ip_blocks_vd); + break; + default: + /* FIXME: not supported yet */ + return -EINVAL; + } + } else { + switch (adev->asic_type) { + case CHIP_TOPAZ: + adev->ip_blocks = topaz_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(topaz_ip_blocks); + break; + case CHIP_FIJI: + adev->ip_blocks = fiji_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(fiji_ip_blocks); + break; + case CHIP_TONGA: + adev->ip_blocks = tonga_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(tonga_ip_blocks); + break; + case CHIP_POLARIS11: + case CHIP_POLARIS10: + adev->ip_blocks = polaris11_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(polaris11_ip_blocks); + break; + case CHIP_CARRIZO: + case CHIP_STONEY: + adev->ip_blocks = cz_ip_blocks; + adev->num_ip_blocks = ARRAY_SIZE(cz_ip_blocks); + break; + default: + /* FIXME: not supported yet */ + return -EINVAL; + } } return 0; @@ -1133,9 +1528,6 @@ static const struct amdgpu_asic_funcs vi_asic_funcs = .set_uvd_clocks = &vi_set_uvd_clocks, .set_vce_clocks = &vi_set_vce_clocks, .get_virtual_caps = &vi_get_virtual_caps, - /* these should be moved to their own ip modules */ - .get_gpu_clock_counter = &gfx_v8_0_get_gpu_clock_counter, - .wait_for_mc_idle = &gmc_v8_0_mc_wait_for_idle, }; static int vi_common_early_init(void *handle) @@ -1156,6 +1548,8 @@ static int vi_common_early_init(void *handle) adev->uvd_ctx_wreg = &vi_uvd_ctx_wreg; adev->didt_rreg = &vi_didt_rreg; adev->didt_wreg = &vi_didt_wreg; + adev->gc_cac_rreg = &vi_gc_cac_rreg; + adev->gc_cac_wreg = &vi_gc_cac_wreg; adev->asic_funcs = &vi_asic_funcs; @@ -1221,20 +1615,41 @@ static int vi_common_early_init(void *handle) AMD_CG_SUPPORT_HDP_MGCG | AMD_CG_SUPPORT_HDP_LS | AMD_CG_SUPPORT_SDMA_MGCG | - AMD_CG_SUPPORT_SDMA_LS; + AMD_CG_SUPPORT_SDMA_LS | + AMD_CG_SUPPORT_VCE_MGCG; + /* rev0 hardware requires workarounds to support PG */ adev->pg_flags = 0; + if (adev->rev_id != 0x00) { + adev->pg_flags |= AMD_PG_SUPPORT_GFX_PG | + AMD_PG_SUPPORT_GFX_SMG | + AMD_PG_SUPPORT_GFX_PIPELINE | + AMD_PG_SUPPORT_UVD | + AMD_PG_SUPPORT_VCE; + } adev->external_rev_id = adev->rev_id + 0x1; break; case CHIP_STONEY: adev->cg_flags = AMD_CG_SUPPORT_UVD_MGCG | AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS | + AMD_CG_SUPPORT_GFX_RLC_LS | + AMD_CG_SUPPORT_GFX_CP_LS | + AMD_CG_SUPPORT_GFX_CGTS | + AMD_CG_SUPPORT_GFX_MGLS | + AMD_CG_SUPPORT_GFX_CGTS_LS | + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | AMD_CG_SUPPORT_BIF_LS | AMD_CG_SUPPORT_HDP_MGCG | AMD_CG_SUPPORT_HDP_LS | AMD_CG_SUPPORT_SDMA_MGCG | - AMD_CG_SUPPORT_SDMA_LS; - adev->pg_flags = 0; + AMD_CG_SUPPORT_SDMA_LS | + AMD_CG_SUPPORT_VCE_MGCG; + adev->pg_flags |= AMD_PG_SUPPORT_GFX_PG | + AMD_PG_SUPPORT_GFX_SMG | + AMD_PG_SUPPORT_GFX_PIPELINE | + AMD_PG_SUPPORT_UVD | + AMD_PG_SUPPORT_VCE; adev->external_rev_id = adev->rev_id + 0x1; break; default: |