diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 576 |
1 files changed, 320 insertions, 256 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 3a0b69b09ed6..da43813d67a4 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -20,6 +20,7 @@ * OTHER DEALINGS IN THE SOFTWARE. * */ +#include <linux/kernel.h> #include <linux/firmware.h> #include <drm/drmP.h> #include "amdgpu.h" @@ -66,38 +67,70 @@ MODULE_FIRMWARE("amdgpu/raven_rlc.bin"); static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] = { - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID0), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID0)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID1), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID1)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID2), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID2)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID3), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID3)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID4), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID4)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID5), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID5)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID6), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID6)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID7), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID7)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID8), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID8)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID9), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID9)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID10), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID10)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID11), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID11)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID12), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID12)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID13), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID13)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID14), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID14)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID15), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID15)} + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID0), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID0) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID1), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID1) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID2), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID2) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID3), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID3) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID4), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID4) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID5), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID5) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID6), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID6) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID7), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID7) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID8), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID8) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID9), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID9) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID10), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID10) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID11), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID11) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID12), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID12)}, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID13), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID13) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID14), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID14) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID15), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID15) } }; static const u32 golden_settings_gc_9_0[] = @@ -116,7 +149,9 @@ static const u32 golden_settings_gc_9_0[] = SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_UTCL1_CNTL_2), 0x08000000, 0x08000080, SOC15_REG_OFFSET(GC, 0, mmRLC_PREWALKER_UTCL1_CNTL), 0x08000000, 0x08000080, SOC15_REG_OFFSET(GC, 0, mmRLC_SPM_UTCL1_CNTL), 0x08000000, 0x08000080, + SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG), 0x00001000, 0x00001000, SOC15_REG_OFFSET(GC, 0, mmSPI_CONFIG_CNTL_1), 0x0000000f, 0x01000107, + SOC15_REG_OFFSET(GC, 0, mmSQC_CONFIG), 0x03000000, 0x020a2000, SOC15_REG_OFFSET(GC, 0, mmTA_CNTL_AUX), 0xfffffeef, 0x010b0000, SOC15_REG_OFFSET(GC, 0, mmTCP_CHAN_STEER_HI), 0xffffffff, 0x4a2c0e68, SOC15_REG_OFFSET(GC, 0, mmTCP_CHAN_STEER_LO), 0xffffffff, 0xb5d3f197, @@ -172,6 +207,12 @@ static const u32 golden_settings_gc_9_1_rv1[] = SOC15_REG_OFFSET(GC, 0, mmTD_CNTL), 0x01bd9f33, 0x00000800 }; +static const u32 golden_settings_gc_9_x_common[] = +{ + SOC15_REG_OFFSET(GC, 0, mmGRBM_CAM_INDEX), 0xffffffff, 0x00000000, + SOC15_REG_OFFSET(GC, 0, mmGRBM_CAM_DATA), 0xffffffff, 0x2544c382 +}; + #define VEGA10_GB_ADDR_CONFIG_GOLDEN 0x2a114042 #define RAVEN_GB_ADDR_CONFIG_GOLDEN 0x24000042 @@ -207,11 +248,14 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) default: break; } + + amdgpu_program_register_sequence(adev, golden_settings_gc_9_x_common, + (const u32)ARRAY_SIZE(golden_settings_gc_9_x_common)); } static void gfx_v9_0_scratch_init(struct amdgpu_device *adev) { - adev->gfx.scratch.num_reg = 7; + adev->gfx.scratch.num_reg = 8; adev->gfx.scratch.reg_base = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0); adev->gfx.scratch.free_mask = (1u << adev->gfx.scratch.num_reg) - 1; } @@ -350,6 +394,25 @@ err1: return r; } + +static void gfx_v9_0_free_microcode(struct amdgpu_device *adev) +{ + release_firmware(adev->gfx.pfp_fw); + adev->gfx.pfp_fw = NULL; + release_firmware(adev->gfx.me_fw); + adev->gfx.me_fw = NULL; + release_firmware(adev->gfx.ce_fw); + adev->gfx.ce_fw = NULL; + release_firmware(adev->gfx.rlc_fw); + adev->gfx.rlc_fw = NULL; + release_firmware(adev->gfx.mec_fw); + adev->gfx.mec_fw = NULL; + release_firmware(adev->gfx.mec2_fw); + adev->gfx.mec2_fw = NULL; + + kfree(adev->gfx.rlc.register_list_format); +} + static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) { const char *chip_name; @@ -772,18 +835,16 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) if (cs_data) { /* clear state block */ adev->gfx.rlc.clear_state_size = dws = gfx_v9_0_get_csb_size(adev); - if (adev->gfx.rlc.clear_state_obj == NULL) { - r = amdgpu_bo_create_kernel(adev, dws * 4, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.clear_state_obj, - &adev->gfx.rlc.clear_state_gpu_addr, - (void **)&adev->gfx.rlc.cs_ptr); - if (r) { - dev_err(adev->dev, - "(%d) failed to create rlc csb bo\n", r); - gfx_v9_0_rlc_fini(adev); - return r; - } + r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); + if (r) { + dev_err(adev->dev, "(%d) failed to create rlc csb bo\n", + r); + gfx_v9_0_rlc_fini(adev); + return r; } /* set up the cs buffer */ dst_ptr = adev->gfx.rlc.cs_ptr; @@ -795,18 +856,16 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) if (adev->asic_type == CHIP_RAVEN) { /* TODO: double check the cp_table_size for RV */ adev->gfx.rlc.cp_table_size = ALIGN(96 * 5 * 4, 2048) + (64 * 1024); /* JT + GDS */ - if (adev->gfx.rlc.cp_table_obj == NULL) { - r = amdgpu_bo_create_kernel(adev, adev->gfx.rlc.cp_table_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, - &adev->gfx.rlc.cp_table_obj, - &adev->gfx.rlc.cp_table_gpu_addr, - (void **)&adev->gfx.rlc.cp_table_ptr); - if (r) { - dev_err(adev->dev, - "(%d) failed to create cp table bo\n", r); - gfx_v9_0_rlc_fini(adev); - return r; - } + r = amdgpu_bo_create_reserved(adev, adev->gfx.rlc.cp_table_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, + &adev->gfx.rlc.cp_table_obj, + &adev->gfx.rlc.cp_table_gpu_addr, + (void **)&adev->gfx.rlc.cp_table_ptr); + if (r) { + dev_err(adev->dev, + "(%d) failed to create cp table bo\n", r); + gfx_v9_0_rlc_fini(adev); + return r; } rv_init_cp_jump_table(adev); @@ -821,28 +880,8 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) static void gfx_v9_0_mec_fini(struct amdgpu_device *adev) { - int r; - - if (adev->gfx.mec.hpd_eop_obj) { - r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve HPD EOP bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.mec.hpd_eop_obj); - amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj); - - amdgpu_bo_unref(&adev->gfx.mec.hpd_eop_obj); - adev->gfx.mec.hpd_eop_obj = NULL; - } - if (adev->gfx.mec.mec_fw_obj) { - r = amdgpu_bo_reserve(adev->gfx.mec.mec_fw_obj, true); - if (unlikely(r != 0)) - dev_warn(adev->dev, "(%d) reserve mec firmware bo failed\n", r); - amdgpu_bo_unpin(adev->gfx.mec.mec_fw_obj); - amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj); - - amdgpu_bo_unref(&adev->gfx.mec.mec_fw_obj); - adev->gfx.mec.mec_fw_obj = NULL; - } + amdgpu_bo_free_kernel(&adev->gfx.mec.hpd_eop_obj, NULL, NULL); + amdgpu_bo_free_kernel(&adev->gfx.mec.mec_fw_obj, NULL, NULL); } static int gfx_v9_0_mec_init(struct amdgpu_device *adev) @@ -862,33 +901,13 @@ static int gfx_v9_0_mec_init(struct amdgpu_device *adev) amdgpu_gfx_compute_queue_acquire(adev); mec_hpd_size = adev->gfx.num_compute_rings * GFX9_MEC_HPD_SIZE; - if (adev->gfx.mec.hpd_eop_obj == NULL) { - r = amdgpu_bo_create(adev, - mec_hpd_size, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, - &adev->gfx.mec.hpd_eop_obj); - if (r) { - dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r); - return r; - } - } - - r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, false); - if (unlikely(r != 0)) { - gfx_v9_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.mec.hpd_eop_obj, AMDGPU_GEM_DOMAIN_GTT, - &adev->gfx.mec.hpd_eop_gpu_addr); - if (r) { - dev_warn(adev->dev, "(%d) pin HDP EOP bo failed\n", r); - gfx_v9_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_kmap(adev->gfx.mec.hpd_eop_obj, (void **)&hpd); + r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &adev->gfx.mec.hpd_eop_obj, + &adev->gfx.mec.hpd_eop_gpu_addr, + (void **)&hpd); if (r) { - dev_warn(adev->dev, "(%d) map HDP EOP bo failed\n", r); + dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r); gfx_v9_0_mec_fini(adev); return r; } @@ -905,42 +924,22 @@ static int gfx_v9_0_mec_init(struct amdgpu_device *adev) le32_to_cpu(mec_hdr->header.ucode_array_offset_bytes)); fw_size = le32_to_cpu(mec_hdr->header.ucode_size_bytes) / 4; - if (adev->gfx.mec.mec_fw_obj == NULL) { - r = amdgpu_bo_create(adev, - mec_hdr->header.ucode_size_bytes, - PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, - &adev->gfx.mec.mec_fw_obj); - if (r) { - dev_warn(adev->dev, "(%d) create mec firmware bo failed\n", r); - return r; - } - } - - r = amdgpu_bo_reserve(adev->gfx.mec.mec_fw_obj, false); - if (unlikely(r != 0)) { - gfx_v9_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_pin(adev->gfx.mec.mec_fw_obj, AMDGPU_GEM_DOMAIN_GTT, - &adev->gfx.mec.mec_fw_gpu_addr); + r = amdgpu_bo_create_reserved(adev, mec_hdr->header.ucode_size_bytes, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, + &adev->gfx.mec.mec_fw_obj, + &adev->gfx.mec.mec_fw_gpu_addr, + (void **)&fw); if (r) { - dev_warn(adev->dev, "(%d) pin mec firmware bo failed\n", r); - gfx_v9_0_mec_fini(adev); - return r; - } - r = amdgpu_bo_kmap(adev->gfx.mec.mec_fw_obj, (void **)&fw); - if (r) { - dev_warn(adev->dev, "(%d) map firmware bo failed\n", r); + dev_warn(adev->dev, "(%d) create mec firmware bo failed\n", r); gfx_v9_0_mec_fini(adev); return r; } + memcpy(fw, fw_data, fw_size); amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_obj); amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj); - return 0; } @@ -998,12 +997,22 @@ static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd, start + SQIND_WAVE_SGPRS_OFFSET, size, dst); } +static void gfx_v9_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd, + uint32_t wave, uint32_t thread, + uint32_t start, uint32_t size, + uint32_t *dst) +{ + wave_read_regs( + adev, simd, wave, thread, + start + SQIND_WAVE_VGPRS_OFFSET, size, dst); +} static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = { .get_gpu_clock_counter = &gfx_v9_0_get_gpu_clock_counter, .select_se_sh = &gfx_v9_0_select_se_sh, .read_wave_data = &gfx_v9_0_read_wave_data, .read_wave_sgprs = &gfx_v9_0_read_wave_sgprs, + .read_wave_vgprs = &gfx_v9_0_read_wave_vgprs, }; static void gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) @@ -1182,30 +1191,22 @@ static int gfx_v9_0_ngg_en(struct amdgpu_device *adev) { struct amdgpu_ring *ring = &adev->gfx.gfx_ring[0]; int r; - u32 data; - u32 size; - u32 base; + u32 data, base; if (!amdgpu_ngg) return 0; /* Program buffer size */ - data = 0; - size = adev->gfx.ngg.buf[NGG_PRIM].size / 256; - data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, INDEX_BUF_SIZE, size); - - size = adev->gfx.ngg.buf[NGG_POS].size / 256; - data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, POS_BUF_SIZE, size); - + data = REG_SET_FIELD(0, WD_BUF_RESOURCE_1, INDEX_BUF_SIZE, + adev->gfx.ngg.buf[NGG_PRIM].size >> 8); + data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, POS_BUF_SIZE, + adev->gfx.ngg.buf[NGG_POS].size >> 8); WREG32_SOC15(GC, 0, mmWD_BUF_RESOURCE_1, data); - data = 0; - size = adev->gfx.ngg.buf[NGG_CNTL].size / 256; - data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, CNTL_SB_BUF_SIZE, size); - - size = adev->gfx.ngg.buf[NGG_PARAM].size / 1024; - data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, PARAM_BUF_SIZE, size); - + data = REG_SET_FIELD(0, WD_BUF_RESOURCE_2, CNTL_SB_BUF_SIZE, + adev->gfx.ngg.buf[NGG_CNTL].size >> 8); + data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, PARAM_BUF_SIZE, + adev->gfx.ngg.buf[NGG_PARAM].size >> 10); WREG32_SOC15(GC, 0, mmWD_BUF_RESOURCE_2, data); /* Program buffer base address */ @@ -1368,7 +1369,10 @@ static int gfx_v9_0_sw_init(void *handle) for (i = 0; i < adev->gfx.num_gfx_rings; i++) { ring = &adev->gfx.gfx_ring[i]; ring->ring_obj = NULL; - sprintf(ring->name, "gfx"); + if (!i) + sprintf(ring->name, "gfx"); + else + sprintf(ring->name, "gfx_%d", i); ring->use_doorbell = true; ring->doorbell_index = AMDGPU_DOORBELL64_GFX_RING0 << 1; r = amdgpu_ring_init(adev, ring, 1024, @@ -1408,7 +1412,7 @@ static int gfx_v9_0_sw_init(void *handle) return r; /* create MQD for all compute queues as wel as KIQ for SRIOV case */ - r = amdgpu_gfx_compute_mqd_sw_init(adev, sizeof(struct v9_mqd)); + r = amdgpu_gfx_compute_mqd_sw_init(adev, sizeof(struct v9_mqd_allocation)); if (r) return r; @@ -1460,9 +1464,19 @@ static int gfx_v9_0_sw_fini(void *handle) amdgpu_gfx_compute_mqd_sw_fini(adev); amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring, &adev->gfx.kiq.irq); amdgpu_gfx_kiq_fini(adev); + amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL); gfx_v9_0_mec_fini(adev); gfx_v9_0_ngg_fini(adev); + amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, + &adev->gfx.rlc.clear_state_gpu_addr, + (void **)&adev->gfx.rlc.cs_ptr); + if (adev->asic_type == CHIP_RAVEN) { + amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, + &adev->gfx.rlc.cp_table_gpu_addr, + (void **)&adev->gfx.rlc.cp_table_ptr); + } + gfx_v9_0_free_microcode(adev); return 0; } @@ -1475,21 +1489,23 @@ static void gfx_v9_0_tiling_mode_table_init(struct amdgpu_device *adev) static void gfx_v9_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num, u32 instance) { - u32 data = REG_SET_FIELD(0, GRBM_GFX_INDEX, INSTANCE_BROADCAST_WRITES, 1); + u32 data; - if ((se_num == 0xffffffff) && (sh_num == 0xffffffff)) { - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_BROADCAST_WRITES, 1); - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_BROADCAST_WRITES, 1); - } else if (se_num == 0xffffffff) { - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_INDEX, sh_num); + if (instance == 0xffffffff) + data = REG_SET_FIELD(0, GRBM_GFX_INDEX, INSTANCE_BROADCAST_WRITES, 1); + else + data = REG_SET_FIELD(0, GRBM_GFX_INDEX, INSTANCE_INDEX, instance); + + if (se_num == 0xffffffff) data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_BROADCAST_WRITES, 1); - } else if (sh_num == 0xffffffff) { - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_BROADCAST_WRITES, 1); + else data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX, se_num); - } else { + + if (sh_num == 0xffffffff) + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_BROADCAST_WRITES, 1); + else data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_INDEX, sh_num); - data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX, se_num); - } + WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, data); } @@ -1742,10 +1758,10 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev) adev->gfx.rlc.reg_list_format_size_bytes >> 2, unique_indirect_regs, &unique_indirect_reg_count, - sizeof(unique_indirect_regs)/sizeof(int), + ARRAY_SIZE(unique_indirect_regs), indirect_start_offsets, &indirect_start_offsets_count, - sizeof(indirect_start_offsets)/sizeof(int)); + ARRAY_SIZE(indirect_start_offsets)); /* enable auto inc in case it is disabled */ tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL)); @@ -1782,12 +1798,12 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev) /* write the starting offsets to RLC scratch ram */ WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_SCRATCH_ADDR), adev->gfx.rlc.starting_offsets_start); - for (i = 0; i < sizeof(indirect_start_offsets)/sizeof(int); i++) + for (i = 0; i < ARRAY_SIZE(indirect_start_offsets); i++) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_SCRATCH_DATA), indirect_start_offsets[i]); /* load unique indirect regs*/ - for (i = 0; i < sizeof(unique_indirect_regs)/sizeof(int); i++) { + for (i = 0; i < ARRAY_SIZE(unique_indirect_regs); i++) { WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_ADDR_0) + i, unique_indirect_regs[i] & 0x3FFFF); WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_DATA_0) + i, @@ -1800,11 +1816,7 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev) static void gfx_v9_0_enable_save_restore_machine(struct amdgpu_device *adev) { - u32 tmp = 0; - - tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL)); - tmp |= RLC_SRM_CNTL__SRM_ENABLE_MASK; - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL), tmp); + WREG32_FIELD15(GC, 0, RLC_SRM_CNTL, SRM_ENABLE, 1); } static void pwr_10_0_gfxip_control_over_cgpg(struct amdgpu_device *adev, @@ -1882,16 +1894,11 @@ static void gfx_v9_0_enable_sck_slow_down_on_power_up(struct amdgpu_device *adev uint32_t default_data = 0; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - - if (enable == true) { - data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK; - if (default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } else { - data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } + data = REG_SET_FIELD(data, RLC_PG_CNTL, + SMU_CLK_SLOWDOWN_ON_PU_ENABLE, + enable ? 1 : 0); + if (default_data != data) + WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } static void gfx_v9_0_enable_sck_slow_down_on_power_down(struct amdgpu_device *adev, @@ -1901,16 +1908,11 @@ static void gfx_v9_0_enable_sck_slow_down_on_power_down(struct amdgpu_device *ad uint32_t default_data = 0; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - - if (enable == true) { - data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } else { - data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } + data = REG_SET_FIELD(data, RLC_PG_CNTL, + SMU_CLK_SLOWDOWN_ON_PD_ENABLE, + enable ? 1 : 0); + if(default_data != data) + WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } static void gfx_v9_0_enable_cp_power_gating(struct amdgpu_device *adev, @@ -1920,16 +1922,11 @@ static void gfx_v9_0_enable_cp_power_gating(struct amdgpu_device *adev, uint32_t default_data = 0; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - - if (enable == true) { - data &= ~RLC_PG_CNTL__CP_PG_DISABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } else { - data |= RLC_PG_CNTL__CP_PG_DISABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } + data = REG_SET_FIELD(data, RLC_PG_CNTL, + CP_PG_DISABLE, + enable ? 0 : 1); + if(default_data != data) + WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } static void gfx_v9_0_enable_gfx_cg_power_gating(struct amdgpu_device *adev, @@ -1938,10 +1935,9 @@ static void gfx_v9_0_enable_gfx_cg_power_gating(struct amdgpu_device *adev, uint32_t data, default_data; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - if (enable == true) - data |= RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK; + data = REG_SET_FIELD(data, RLC_PG_CNTL, + GFX_POWER_GATING_ENABLE, + enable ? 1 : 0); if(default_data != data) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } @@ -1952,10 +1948,9 @@ static void gfx_v9_0_enable_gfx_pipeline_powergating(struct amdgpu_device *adev, uint32_t data, default_data; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - if (enable == true) - data |= RLC_PG_CNTL__GFX_PIPELINE_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__GFX_PIPELINE_PG_ENABLE_MASK; + data = REG_SET_FIELD(data, RLC_PG_CNTL, + GFX_PIPELINE_PG_ENABLE, + enable ? 1 : 0); if(default_data != data) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); @@ -1970,10 +1965,9 @@ static void gfx_v9_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *ade uint32_t data, default_data; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - if (enable == true) - data |= RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK; + data = REG_SET_FIELD(data, RLC_PG_CNTL, + STATIC_PER_CU_PG_ENABLE, + enable ? 1 : 0); if(default_data != data) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } @@ -1984,10 +1978,9 @@ static void gfx_v9_0_enable_gfx_dynamic_mg_power_gating(struct amdgpu_device *ad uint32_t data, default_data; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - if (enable == true) - data |= RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK; + data = REG_SET_FIELD(data, RLC_PG_CNTL, + DYN_PER_CU_PG_ENABLE, + enable ? 1 : 0); if(default_data != data) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } @@ -2027,13 +2020,8 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev) void gfx_v9_0_rlc_stop(struct amdgpu_device *adev) { - u32 tmp = RREG32_SOC15(GC, 0, mmRLC_CNTL); - - tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 0); - WREG32_SOC15(GC, 0, mmRLC_CNTL, tmp); - + WREG32_FIELD15(GC, 0, RLC_CNTL, RLC_ENABLE_F32, 0); gfx_v9_0_enable_gui_idle_interrupt(adev, false); - gfx_v9_0_wait_for_rlc_serdes(adev); } @@ -2105,8 +2093,10 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev) { int r; - if (amdgpu_sriov_vf(adev)) + if (amdgpu_sriov_vf(adev)) { + gfx_v9_0_init_csb(adev); return 0; + } gfx_v9_0_rlc_stop(adev); @@ -2217,7 +2207,7 @@ static int gfx_v9_0_cp_gfx_start(struct amdgpu_device *adev) struct amdgpu_ring *ring = &adev->gfx.gfx_ring[0]; const struct cs_section_def *sect = NULL; const struct cs_extent_def *ext = NULL; - int r, i; + int r, i, tmp; /* init the CP */ WREG32_SOC15(GC, 0, mmCP_MAX_CONTEXT, adev->gfx.config.max_hw_contexts - 1); @@ -2225,7 +2215,7 @@ static int gfx_v9_0_cp_gfx_start(struct amdgpu_device *adev) gfx_v9_0_cp_gfx_enable(adev, true); - r = amdgpu_ring_alloc(ring, gfx_v9_0_get_csb_size(adev) + 4); + r = amdgpu_ring_alloc(ring, gfx_v9_0_get_csb_size(adev) + 4 + 3); if (r) { DRM_ERROR("amdgpu: cp failed to lock ring (%d).\n", r); return r; @@ -2263,6 +2253,12 @@ static int gfx_v9_0_cp_gfx_start(struct amdgpu_device *adev) amdgpu_ring_write(ring, 0x8000); amdgpu_ring_write(ring, 0x8000); + amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG,1)); + tmp = (PACKET3_SET_UCONFIG_REG_INDEX_TYPE | + (SOC15_REG_OFFSET(GC, 0, mmVGT_INDEX_TYPE) - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(ring, tmp); + amdgpu_ring_write(ring, 0); + amdgpu_ring_commit(ring); return 0; @@ -2425,7 +2421,7 @@ static int gfx_v9_0_kiq_kcq_enable(struct amdgpu_device *adev) /* This situation may be hit in the future if a new HW * generation exposes more than 64 queues. If so, the * definition of queue_mask needs updating */ - if (WARN_ON(i > (sizeof(queue_mask)*8))) { + if (WARN_ON(i >= (sizeof(queue_mask)*8))) { DRM_ERROR("Invalid KCQ enabled: %d\n", i); break; } @@ -2517,6 +2513,13 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) mqd->compute_static_thread_mgmt_se3 = 0xffffffff; mqd->compute_misc_reserved = 0x00000003; + mqd->dynamic_cu_mask_addr_lo = + lower_32_bits(ring->mqd_gpu_addr + + offsetof(struct v9_mqd_allocation, dynamic_cu_mask)); + mqd->dynamic_cu_mask_addr_hi = + upper_32_bits(ring->mqd_gpu_addr + + offsetof(struct v9_mqd_allocation, dynamic_cu_mask)); + eop_base_addr = ring->eop_gpu_addr >> 8; mqd->cp_hqd_eop_base_addr_lo = eop_base_addr; mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); @@ -2540,10 +2543,10 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) DOORBELL_SOURCE, 0); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_HIT, 0); - } - else + } else { tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 0); + } mqd->cp_hqd_pq_doorbell_control = tmp; @@ -2746,10 +2749,10 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) gfx_v9_0_kiq_setting(ring); - if (adev->gfx.in_reset) { /* for GPU_RESET case */ + if (adev->in_sriov_reset) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd)); + memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation)); /* reset ring buffer */ ring->wptr = 0; @@ -2761,7 +2764,9 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) soc15_grbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); } else { - memset((void *)mqd, 0, sizeof(*mqd)); + memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation)); + ((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; + ((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; mutex_lock(&adev->srbm_mutex); soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v9_0_mqd_init(ring); @@ -2770,7 +2775,7 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) mutex_unlock(&adev->srbm_mutex); if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd)); + memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation)); } return 0; @@ -2782,8 +2787,10 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) struct v9_mqd *mqd = ring->mqd_ptr; int mqd_idx = ring - &adev->gfx.compute_ring[0]; - if (!adev->gfx.in_reset && !adev->gfx.in_suspend) { - memset((void *)mqd, 0, sizeof(*mqd)); + if (!adev->in_sriov_reset && !adev->gfx.in_suspend) { + memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation)); + ((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; + ((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; mutex_lock(&adev->srbm_mutex); soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v9_0_mqd_init(ring); @@ -2791,11 +2798,11 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) mutex_unlock(&adev->srbm_mutex); if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd)); - } else if (adev->gfx.in_reset) { /* for GPU_RESET case */ + memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation)); + } else if (adev->in_sriov_reset) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd)); + memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation)); /* reset ring buffer */ ring->wptr = 0; @@ -2936,12 +2943,70 @@ static int gfx_v9_0_hw_init(void *handle) return r; } +static int gfx_v9_0_kcq_disable(struct amdgpu_ring *kiq_ring,struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = kiq_ring->adev; + uint32_t scratch, tmp = 0; + int r, i; + + r = amdgpu_gfx_scratch_get(adev, &scratch); + if (r) { + DRM_ERROR("Failed to get scratch reg (%d).\n", r); + return r; + } + WREG32(scratch, 0xCAFEDEAD); + + r = amdgpu_ring_alloc(kiq_ring, 10); + if (r) { + DRM_ERROR("Failed to lock KIQ (%d).\n", r); + amdgpu_gfx_scratch_free(adev, scratch); + return r; + } + + /* unmap queues */ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); + amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + PACKET3_UNMAP_QUEUES_ACTION(1) | /* RESET_QUEUES */ + PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) | + PACKET3_UNMAP_QUEUES_ENGINE_SEL(0) | + PACKET3_UNMAP_QUEUES_NUM_QUEUES(1)); + amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + /* write to scratch for completion */ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); + amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(kiq_ring, 0xDEADBEEF); + amdgpu_ring_commit(kiq_ring); + + for (i = 0; i < adev->usec_timeout; i++) { + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + if (i >= adev->usec_timeout) { + DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp); + r = -EINVAL; + } + amdgpu_gfx_scratch_free(adev, scratch); + return r; +} + + static int gfx_v9_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i; amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); + + /* disable KCQ to avoid CPC touch memory not valid anymore */ + for (i = 0; i < adev->gfx.num_compute_rings; i++) + gfx_v9_0_kcq_disable(&adev->gfx.kiq.ring, &adev->gfx.compute_ring[i]); + if (amdgpu_sriov_vf(adev)) { pr_debug("For SRIOV client, shouldn't do anything.\n"); return 0; @@ -2984,15 +3049,10 @@ static bool gfx_v9_0_is_idle(void *handle) static int gfx_v9_0_wait_for_idle(void *handle) { unsigned i; - u32 tmp; struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->usec_timeout; i++) { - /* read MC_STATUS */ - tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS) & - GRBM_STATUS__GUI_ACTIVE_MASK; - - if (!REG_GET_FIELD(tmp, GRBM_STATUS, GUI_ACTIVE)) + if (gfx_v9_0_is_idle(handle)) return 0; udelay(1); } @@ -3551,9 +3611,11 @@ static void gfx_v9_0_ring_set_wptr_gfx(struct amdgpu_ring *ring) static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) { u32 ref_and_mask, reg_mem_engine; - struct nbio_hdp_flush_reg *nbio_hf_reg; + const struct nbio_hdp_flush_reg *nbio_hf_reg; - if (ring->adev->asic_type == CHIP_VEGA10) + if (ring->adev->flags & AMD_IS_APU) + nbio_hf_reg = &nbio_v7_0_hdp_flush_reg; + else nbio_hf_reg = &nbio_v6_1_hdp_flush_reg; if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) { @@ -3582,7 +3644,7 @@ static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) static void gfx_v9_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) { gfx_v9_0_write_data_to_reg(ring, 0, true, - SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0), 1); + SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1); } static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, @@ -3772,7 +3834,7 @@ static void gfx_v9_ring_emit_sb(struct amdgpu_ring *ring) static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring) { - static struct v9_ce_ib_state ce_payload = {0}; + struct v9_ce_ib_state ce_payload = {0}; uint64_t csa_addr; int cnt; @@ -3791,7 +3853,7 @@ static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring) static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring) { - static struct v9_de_ib_state de_payload = {0}; + struct v9_de_ib_state de_payload = {0}; uint64_t csa_addr, gds_addr; int cnt; @@ -3811,6 +3873,12 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring) amdgpu_ring_write_multiple(ring, (void *)&de_payload, sizeof(de_payload) >> 2); } +static void gfx_v9_0_ring_emit_tmz(struct amdgpu_ring *ring, bool start) +{ + amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0)); + amdgpu_ring_write(ring, FRAME_CMD(start ? 0 : 1)); /* frame_end */ +} + static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) { uint32_t dw2 = 0; @@ -3818,6 +3886,8 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) if (amdgpu_sriov_vf(ring->adev)) gfx_v9_0_ring_emit_ce_meta(ring); + gfx_v9_0_ring_emit_tmz(ring, true); + dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */ if (flags & AMDGPU_HAVE_CTX_SWITCH) { /* set load_global_config & load_global_uconfig */ @@ -3868,12 +3938,6 @@ static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size>>2) - offset + cur; } -static void gfx_v9_0_ring_emit_tmz(struct amdgpu_ring *ring, bool start) -{ - amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0)); - amdgpu_ring_write(ring, FRAME_CMD(start ? 0 : 1)); /* frame_end */ -} - static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) { struct amdgpu_device *adev = ring->adev; @@ -4156,7 +4220,7 @@ static int gfx_v9_0_kiq_irq(struct amdgpu_device *adev, return 0; } -const struct amd_ip_funcs gfx_v9_0_ip_funcs = { +static const struct amd_ip_funcs gfx_v9_0_ip_funcs = { .name = "gfx_v9_0", .early_init = gfx_v9_0_early_init, .late_init = gfx_v9_0_late_init, |