diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
42 files changed, 960 insertions, 266 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index 14f68c028126..5ffb07b02810 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -234,14 +234,10 @@ uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *s }) /* GPUVM API */ -int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, u32 pasid, - void **vm, void **process_info, - struct dma_fence **ef); int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, struct file *filp, u32 pasid, void **vm, void **process_info, struct dma_fence **ef); -void amdgpu_amdkfd_gpuvm_destroy_process_vm(struct kgd_dev *kgd, void *vm); void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm); uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm); int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index e93850f2f3b1..7d4118c8128a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -445,22 +445,19 @@ static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem) mapping_flags |= AMDGPU_VM_MTYPE_UC; } else if (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) { if (bo_adev == adev) { - mapping_flags |= AMDGPU_VM_MTYPE_RW; + mapping_flags |= coherent ? + AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW; if (adev->gmc.xgmi.connected_to_cpu) snoop = true; } else { - mapping_flags |= AMDGPU_VM_MTYPE_NC; + mapping_flags |= AMDGPU_VM_MTYPE_UC; if (amdgpu_xgmi_same_hive(adev, bo_adev)) snoop = true; } } else { snoop = true; - if (adev->gmc.xgmi.connected_to_cpu) - /* system memory uses NC on A+A */ - mapping_flags |= AMDGPU_VM_MTYPE_NC; - else - mapping_flags |= coherent ? - AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; + mapping_flags |= coherent ? + AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC; } break; default: @@ -1037,41 +1034,6 @@ create_evict_fence_fail: return ret; } -int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, u32 pasid, - void **vm, void **process_info, - struct dma_fence **ef) -{ - struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct amdgpu_vm *new_vm; - int ret; - - new_vm = kzalloc(sizeof(*new_vm), GFP_KERNEL); - if (!new_vm) - return -ENOMEM; - - /* Initialize AMDGPU part of the VM */ - ret = amdgpu_vm_init(adev, new_vm, AMDGPU_VM_CONTEXT_COMPUTE, pasid); - if (ret) { - pr_err("Failed init vm ret %d\n", ret); - goto amdgpu_vm_init_fail; - } - - /* Initialize KFD part of the VM and process info */ - ret = init_kfd_vm(new_vm, process_info, ef); - if (ret) - goto init_kfd_vm_fail; - - *vm = (void *) new_vm; - - return 0; - -init_kfd_vm_fail: - amdgpu_vm_fini(adev, new_vm); -amdgpu_vm_init_fail: - kfree(new_vm); - return ret; -} - int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, struct file *filp, u32 pasid, void **vm, void **process_info, @@ -1138,21 +1100,6 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, } } -void amdgpu_amdkfd_gpuvm_destroy_process_vm(struct kgd_dev *kgd, void *vm) -{ - struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; - - if (WARN_ON(!kgd || !vm)) - return; - - pr_debug("Destroying process vm %p\n", vm); - - /* Release the VM context */ - amdgpu_vm_fini(adev, avm); - kfree(vm); -} - void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm) { struct amdgpu_device *adev = get_amdgpu_device(kgd); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index b4ad1c055c70..7d3b54615147 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3410,19 +3410,6 @@ int amdgpu_device_init(struct amdgpu_device *adev, /* doorbell bar mapping and doorbell index init*/ amdgpu_device_doorbell_init(adev); - /* if we have > 1 VGA cards, then disable the amdgpu VGA resources */ - /* this will fail for cards that aren't VGA class devices, just - * ignore it */ - if ((adev->pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA) - vga_client_register(adev->pdev, adev, NULL, amdgpu_device_vga_set_decode); - - if (amdgpu_device_supports_px(ddev)) { - px = true; - vga_switcheroo_register_client(adev->pdev, - &amdgpu_switcheroo_ops, px); - vga_switcheroo_init_domain_pm_ops(adev->dev, &adev->vga_pm_domain); - } - if (amdgpu_emu_mode == 1) { /* post the asic on emulation mode */ emu_soc_asic_init(adev); @@ -3619,6 +3606,19 @@ fence_driver_init: if (amdgpu_device_cache_pci_state(adev->pdev)) pci_restore_state(pdev); + /* if we have > 1 VGA cards, then disable the amdgpu VGA resources */ + /* this will fail for cards that aren't VGA class devices, just + * ignore it */ + if ((adev->pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA) + vga_client_register(adev->pdev, adev, NULL, amdgpu_device_vga_set_decode); + + if (amdgpu_device_supports_px(ddev)) { + px = true; + vga_switcheroo_register_client(adev->pdev, + &amdgpu_switcheroo_ops, px); + vga_switcheroo_init_domain_pm_ops(adev->dev, &adev->vga_pm_domain); + } + if (adev->gmc.xgmi.pending_reset) queue_delayed_work(system_wq, &mgpu_info.delayed_reset_work, msecs_to_jiffies(AMDGPU_RESUME_MS)); @@ -3630,8 +3630,6 @@ release_ras_con: failed: amdgpu_vf_error_trans_all(adev); - if (px) - vga_switcheroo_fini_domain_pm_ops(adev->dev); failed_unmap: iounmap(adev->rmmio); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index 9a2f811450ed..8a1fb8b6606e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -837,6 +837,174 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb) return 0; } +static void get_block_dimensions(unsigned int block_log2, unsigned int cpp, + unsigned int *width, unsigned int *height) +{ + unsigned int cpp_log2 = ilog2(cpp); + unsigned int pixel_log2 = block_log2 - cpp_log2; + unsigned int width_log2 = (pixel_log2 + 1) / 2; + unsigned int height_log2 = pixel_log2 - width_log2; + + *width = 1 << width_log2; + *height = 1 << height_log2; +} + +static unsigned int get_dcc_block_size(uint64_t modifier, bool rb_aligned, + bool pipe_aligned) +{ + unsigned int ver = AMD_FMT_MOD_GET(TILE_VERSION, modifier); + + switch (ver) { + case AMD_FMT_MOD_TILE_VER_GFX9: { + /* + * TODO: for pipe aligned we may need to check the alignment of the + * total size of the surface, which may need to be bigger than the + * natural alignment due to some HW workarounds + */ + return max(10 + (rb_aligned ? (int)AMD_FMT_MOD_GET(RB, modifier) : 0), 12); + } + case AMD_FMT_MOD_TILE_VER_GFX10: + case AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS: { + int pipes_log2 = AMD_FMT_MOD_GET(PIPE_XOR_BITS, modifier); + + if (ver == AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS && pipes_log2 > 1 && + AMD_FMT_MOD_GET(PACKERS, modifier) == pipes_log2) + ++pipes_log2; + + return max(8 + (pipe_aligned ? pipes_log2 : 0), 12); + } + default: + return 0; + } +} + +static int amdgpu_display_verify_plane(struct amdgpu_framebuffer *rfb, int plane, + const struct drm_format_info *format, + unsigned int block_width, unsigned int block_height, + unsigned int block_size_log2) +{ + unsigned int width = rfb->base.width / + ((plane && plane < format->num_planes) ? format->hsub : 1); + unsigned int height = rfb->base.height / + ((plane && plane < format->num_planes) ? format->vsub : 1); + unsigned int cpp = plane < format->num_planes ? format->cpp[plane] : 1; + unsigned int block_pitch = block_width * cpp; + unsigned int min_pitch = ALIGN(width * cpp, block_pitch); + unsigned int block_size = 1 << block_size_log2; + uint64_t size; + + if (rfb->base.pitches[plane] % block_pitch) { + drm_dbg_kms(rfb->base.dev, + "pitch %d for plane %d is not a multiple of block pitch %d\n", + rfb->base.pitches[plane], plane, block_pitch); + return -EINVAL; + } + if (rfb->base.pitches[plane] < min_pitch) { + drm_dbg_kms(rfb->base.dev, + "pitch %d for plane %d is less than minimum pitch %d\n", + rfb->base.pitches[plane], plane, min_pitch); + return -EINVAL; + } + + /* Force at least natural alignment. */ + if (rfb->base.offsets[plane] % block_size) { + drm_dbg_kms(rfb->base.dev, + "offset 0x%x for plane %d is not a multiple of block pitch 0x%x\n", + rfb->base.offsets[plane], plane, block_size); + return -EINVAL; + } + + size = rfb->base.offsets[plane] + + (uint64_t)rfb->base.pitches[plane] / block_pitch * + block_size * DIV_ROUND_UP(height, block_height); + + if (rfb->base.obj[0]->size < size) { + drm_dbg_kms(rfb->base.dev, + "BO size 0x%zx is less than 0x%llx required for plane %d\n", + rfb->base.obj[0]->size, size, plane); + return -EINVAL; + } + + return 0; +} + + +static int amdgpu_display_verify_sizes(struct amdgpu_framebuffer *rfb) +{ + const struct drm_format_info *format_info = drm_format_info(rfb->base.format->format); + uint64_t modifier = rfb->base.modifier; + int ret; + unsigned int i, block_width, block_height, block_size_log2; + + if (!rfb->base.dev->mode_config.allow_fb_modifiers) + return 0; + + for (i = 0; i < format_info->num_planes; ++i) { + if (modifier == DRM_FORMAT_MOD_LINEAR) { + block_width = 256 / format_info->cpp[i]; + block_height = 1; + block_size_log2 = 8; + } else { + int swizzle = AMD_FMT_MOD_GET(TILE, modifier); + + switch ((swizzle & ~3) + 1) { + case DC_SW_256B_S: + block_size_log2 = 8; + break; + case DC_SW_4KB_S: + case DC_SW_4KB_S_X: + block_size_log2 = 12; + break; + case DC_SW_64KB_S: + case DC_SW_64KB_S_T: + case DC_SW_64KB_S_X: + block_size_log2 = 16; + break; + default: + drm_dbg_kms(rfb->base.dev, + "Swizzle mode with unknown block size: %d\n", swizzle); + return -EINVAL; + } + + get_block_dimensions(block_size_log2, format_info->cpp[i], + &block_width, &block_height); + } + + ret = amdgpu_display_verify_plane(rfb, i, format_info, + block_width, block_height, block_size_log2); + if (ret) + return ret; + } + + if (AMD_FMT_MOD_GET(DCC, modifier)) { + if (AMD_FMT_MOD_GET(DCC_RETILE, modifier)) { + block_size_log2 = get_dcc_block_size(modifier, false, false); + get_block_dimensions(block_size_log2 + 8, format_info->cpp[0], + &block_width, &block_height); + ret = amdgpu_display_verify_plane(rfb, i, format_info, + block_width, block_height, + block_size_log2); + if (ret) + return ret; + + ++i; + block_size_log2 = get_dcc_block_size(modifier, true, true); + } else { + bool pipe_aligned = AMD_FMT_MOD_GET(DCC_PIPE_ALIGN, modifier); + + block_size_log2 = get_dcc_block_size(modifier, true, pipe_aligned); + } + get_block_dimensions(block_size_log2 + 8, format_info->cpp[0], + &block_width, &block_height); + ret = amdgpu_display_verify_plane(rfb, i, format_info, + block_width, block_height, block_size_log2); + if (ret) + return ret; + } + + return 0; +} + static int amdgpu_display_get_fb_info(const struct amdgpu_framebuffer *amdgpu_fb, uint64_t *tiling_flags, bool *tmz_surface) { @@ -902,12 +1070,23 @@ int amdgpu_display_gem_fb_verify_and_init( int ret; rfb->base.obj[0] = obj; - - /* Verify that bo size can fit the fb size. */ - ret = drm_gem_fb_init_with_funcs(dev, &rfb->base, file_priv, mode_cmd, - &amdgpu_fb_funcs); + drm_helper_mode_fill_fb_struct(dev, &rfb->base, mode_cmd); + ret = drm_framebuffer_init(dev, &rfb->base, &amdgpu_fb_funcs); if (ret) goto err; + /* Verify that the modifier is supported. */ + if (!drm_any_plane_has_format(dev, mode_cmd->pixel_format, + mode_cmd->modifier[0])) { + struct drm_format_name_buf format_name; + drm_dbg_kms(dev, + "unsupported pixel format %s / modifier 0x%llx\n", + drm_get_format_name(mode_cmd->pixel_format, + &format_name), + mode_cmd->modifier[0]); + + ret = -EINVAL; + goto err; + } ret = amdgpu_display_framebuffer_init(dev, rfb, mode_cmd, obj); if (ret) @@ -954,9 +1133,12 @@ int amdgpu_display_framebuffer_init(struct drm_device *dev, } } - for (i = 1; i < rfb->base.format->num_planes; ++i) { + ret = amdgpu_display_verify_sizes(rfb); + if (ret) + return ret; + + for (i = 0; i < rfb->base.format->num_planes; ++i) { drm_gem_object_get(rfb->base.obj[0]); - drm_gem_object_put(rfb->base.obj[i]); rfb->base.obj[i] = rfb->base.obj[0]; } @@ -986,6 +1168,7 @@ amdgpu_display_user_framebuffer_create(struct drm_device *dev, domains = amdgpu_display_supported_domains(drm_to_adev(dev), bo->flags); if (obj->import_attach && !(domains & AMDGPU_GEM_DOMAIN_GTT)) { drm_dbg_kms(dev, "Cannot create framebuffer from imported dma_buf\n"); + drm_gem_object_put(obj); return ERR_PTR(-EINVAL); } @@ -1399,7 +1582,7 @@ int amdgpu_display_suspend_helper(struct amdgpu_device *adev) } } } - return r; + return 0; } int amdgpu_display_resume_helper(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c index e0c4f7c7f1b9..baa980a477d9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c @@ -291,8 +291,8 @@ static struct sg_table *amdgpu_dma_buf_map(struct dma_buf_attachment *attach, break; case TTM_PL_VRAM: - r = amdgpu_vram_mgr_alloc_sgt(adev, &bo->tbo.mem, attach->dev, - dir, &sgt); + r = amdgpu_vram_mgr_alloc_sgt(adev, &bo->tbo.mem, 0, + bo->tbo.base.size, attach->dev, dir, &sgt); if (r) return ERR_PTR(r); break; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 6cf573293823..6fd20ea2935b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -186,7 +186,7 @@ uint amdgpu_ras_mask = 0xffffffff; int amdgpu_bad_page_threshold = -1; struct amdgpu_watchdog_timer amdgpu_watchdog_timer = { .timeout_fatal_disable = false, - .period = 0x23, /* default to max. timeout = 1 << 0x23 cycles */ + .period = 0x0, /* default to 0x0 (timeout disable) */ }; /** @@ -554,7 +554,7 @@ module_param_named(timeout_fatal_disable, amdgpu_watchdog_timer.timeout_fatal_di * DOC: timeout_period (uint) * Modify the watchdog timeout max_cycles as (1 << period) */ -MODULE_PARM_DESC(timeout_period, "watchdog timeout period (1 to 0x23(default), timeout maxCycles = (1 << period)"); +MODULE_PARM_DESC(timeout_period, "watchdog timeout period (0 = timeout disabled, 1 ~ 0x23 = timeout maxcycles = (1 << period)"); module_param_named(timeout_period, amdgpu_watchdog_timer.period, uint, 0644); /** @@ -1574,6 +1574,9 @@ static int amdgpu_pmops_runtime_resume(struct device *dev) amdgpu_device_baco_exit(drm_dev); } ret = amdgpu_device_resume(drm_dev, false); + if (ret) + return ret; + if (amdgpu_device_supports_px(drm_dev)) drm_dev->switch_power_state = DRM_SWITCH_POWER_ON; adev->in_runpm = false; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c index 4c5c19820d37..4f10c4529840 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c @@ -205,7 +205,6 @@ static int amdgpufb_create(struct drm_fb_helper *helper, struct drm_gem_object *gobj = NULL; struct amdgpu_bo *abo = NULL; int ret; - unsigned long tmp; memset(&mode_cmd, 0, sizeof(mode_cmd)); mode_cmd.width = sizes->surface_width; @@ -246,8 +245,7 @@ static int amdgpufb_create(struct drm_fb_helper *helper, info->fbops = &amdgpufb_ops; - tmp = amdgpu_bo_gpu_offset(abo) - adev->gmc.vram_start; - info->fix.smem_start = adev->gmc.aper_base + tmp; + info->fix.smem_start = amdgpu_gmc_vram_cpu_pa(adev, abo); info->fix.smem_len = amdgpu_bo_size(abo); info->screen_base = amdgpu_bo_kptr(abo); info->screen_size = amdgpu_bo_size(abo); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 4d32233cde92..c39ed9eb0987 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -487,6 +487,7 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev) { switch (adev->asic_type) { case CHIP_RAVEN: + case CHIP_RENOIR: if (amdgpu_tmz == 0) { adev->gmc.tmz_enabled = false; dev_info(adev->dev, @@ -497,7 +498,6 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev) "Trusted Memory Zone (TMZ) feature enabled\n"); } break; - case CHIP_RENOIR: case CHIP_NAVI10: case CHIP_NAVI14: case CHIP_NAVI12: @@ -661,8 +661,7 @@ void amdgpu_gmc_init_pdb0(struct amdgpu_device *adev) u64 vram_addr = adev->vm_manager.vram_base_offset - adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size; u64 vram_end = vram_addr + vram_size; - u64 gart_ptb_gpu_pa = amdgpu_bo_gpu_offset(adev->gart.bo) + - adev->vm_manager.vram_base_offset - adev->gmc.vram_start; + u64 gart_ptb_gpu_pa = amdgpu_gmc_vram_pa(adev, adev->gart.bo); flags |= AMDGPU_PTE_VALID | AMDGPU_PTE_READABLE; flags |= AMDGPU_PTE_WRITEABLE; @@ -685,3 +684,39 @@ void amdgpu_gmc_init_pdb0(struct amdgpu_device *adev) /* Requires gart_ptb_gpu_pa to be 4K aligned */ amdgpu_gmc_set_pte_pde(adev, adev->gmc.ptr_pdb0, i, gart_ptb_gpu_pa, flags); } + +/** + * amdgpu_gmc_vram_mc2pa - calculate vram buffer's physical address from MC + * address + * + * @adev: amdgpu_device pointer + * @mc_addr: MC address of buffer + */ +uint64_t amdgpu_gmc_vram_mc2pa(struct amdgpu_device *adev, uint64_t mc_addr) +{ + return mc_addr - adev->gmc.vram_start + adev->vm_manager.vram_base_offset; +} + +/** + * amdgpu_gmc_vram_pa - calculate vram buffer object's physical address from + * GPU's view + * + * @adev: amdgpu_device pointer + * @bo: amdgpu buffer object + */ +uint64_t amdgpu_gmc_vram_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo) +{ + return amdgpu_gmc_vram_mc2pa(adev, amdgpu_bo_gpu_offset(bo)); +} + +/** + * amdgpu_gmc_vram_cpu_pa - calculate vram buffer object's physical address + * from CPU's view + * + * @adev: amdgpu_device pointer + * @bo: amdgpu buffer object + */ +uint64_t amdgpu_gmc_vram_cpu_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo) +{ + return amdgpu_bo_gpu_offset(bo) - adev->gmc.vram_start + adev->gmc.aper_base; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index cbb7735c6988..9d11c02a3938 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -218,15 +218,6 @@ struct amdgpu_gmc { */ u64 fb_start; u64 fb_end; - /* In the case of use GART table for vmid0 FB access, [fb_start, fb_end] - * will be squeezed to GART aperture. But we have a PSP FW issue to fix - * for now. To temporarily workaround the PSP FW issue, added below two - * variables to remember the original fb_start/end to re-enable FB - * aperture to workaround the PSP FW issue. Will delete it after we - * get a proper PSP FW fix. - */ - u64 fb_start_original; - u64 fb_end_original; unsigned vram_width; u64 real_vram_size; int vram_mtrr; @@ -341,4 +332,7 @@ amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type, void amdgpu_gmc_get_vbios_allocations(struct amdgpu_device *adev); void amdgpu_gmc_init_pdb0(struct amdgpu_device *adev); +uint64_t amdgpu_gmc_vram_mc2pa(struct amdgpu_device *adev, uint64_t mc_addr); +uint64_t amdgpu_gmc_vram_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo); +uint64_t amdgpu_gmc_vram_cpu_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 148a3b481b12..a2fe2dac32c1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -76,6 +76,8 @@ int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, } ib->ptr = amdgpu_sa_bo_cpu_addr(ib->sa_bo); + /* flush the cache before commit the IB */ + ib->flags = AMDGPU_IB_FLAG_EMIT_MEM_SYNC; if (!vm) ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c index 94b069630db3..b4971e90b98c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c @@ -215,7 +215,11 @@ static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm, /* Check if we have an idle VMID */ i = 0; list_for_each_entry((*idle), &id_mgr->ids_lru, list) { - fences[i] = amdgpu_sync_peek_fence(&(*idle)->active, ring); + /* Don't use per engine and per process VMID at the same time */ + struct amdgpu_ring *r = adev->vm_manager.concurrent_flush ? + NULL : ring; + + fences[i] = amdgpu_sync_peek_fence(&(*idle)->active, r); if (!fences[i]) break; ++i; @@ -281,7 +285,7 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm, if (updates && (*id)->flushed_updates && updates->context == (*id)->flushed_updates->context && !dma_fence_is_later(updates, (*id)->flushed_updates)) - updates = NULL; + updates = NULL; if ((*id)->owner != vm->immediate.fence_context || job->vm_pd_addr != (*id)->pd_gpu_addr || @@ -290,6 +294,10 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm, !dma_fence_is_signaled((*id)->last_flush))) { struct dma_fence *tmp; + /* Don't use per engine and per process VMID at the same time */ + if (adev->vm_manager.concurrent_flush) + ring = NULL; + /* to prevent one context starved by another context */ (*id)->pd_gpu_addr = 0; tmp = amdgpu_sync_peek_fence(&(*id)->active, ring); @@ -365,12 +373,7 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm, if (updates && (!flushed || dma_fence_is_later(updates, flushed))) needs_flush = true; - /* Concurrent flushes are only possible starting with Vega10 and - * are broken on Navi10 and Navi14. - */ - if (needs_flush && (adev->asic_type < CHIP_VEGA10 || - adev->asic_type == CHIP_NAVI10 || - adev->asic_type == CHIP_NAVI14)) + if (needs_flush && !adev->vm_manager.concurrent_flush) continue; /* Good, we can use this VMID. Remember this submission as diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c index 19c0a3655228..82e9ecf84352 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c @@ -519,8 +519,10 @@ static int init_pmu_entry_by_type_and_add(struct amdgpu_pmu_entry *pmu_entry, pmu_entry->pmu.attr_groups = kmemdup(attr_groups, sizeof(attr_groups), GFP_KERNEL); - if (!pmu_entry->pmu.attr_groups) + if (!pmu_entry->pmu.attr_groups) { + ret = -ENOMEM; goto err_attr_group; + } snprintf(pmu_name, PMU_NAME_SIZE, "%s_%d", pmu_entry->pmu_file_prefix, adev_to_drm(pmu_entry->adev)->primary->index); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index 9e769cf6095b..a09483beb968 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -328,8 +328,12 @@ psp_cmd_submit_buf(struct psp_context *psp, static void psp_prep_tmr_cmd_buf(struct psp_context *psp, struct psp_gfx_cmd_resp *cmd, - uint64_t tmr_mc, uint32_t size) + uint64_t tmr_mc, struct amdgpu_bo *tmr_bo) { + struct amdgpu_device *adev = psp->adev; + uint32_t size = amdgpu_bo_size(tmr_bo); + uint64_t tmr_pa = amdgpu_gmc_vram_pa(adev, tmr_bo); + if (amdgpu_sriov_vf(psp->adev)) cmd->cmd_id = GFX_CMD_ID_SETUP_VMR; else @@ -337,6 +341,9 @@ static void psp_prep_tmr_cmd_buf(struct psp_context *psp, cmd->cmd.cmd_setup_tmr.buf_phy_addr_lo = lower_32_bits(tmr_mc); cmd->cmd.cmd_setup_tmr.buf_phy_addr_hi = upper_32_bits(tmr_mc); cmd->cmd.cmd_setup_tmr.buf_size = size; + cmd->cmd.cmd_setup_tmr.bitfield.virt_phy_addr = 1; + cmd->cmd.cmd_setup_tmr.system_phy_addr_lo = lower_32_bits(tmr_pa); + cmd->cmd.cmd_setup_tmr.system_phy_addr_hi = upper_32_bits(tmr_pa); } static void psp_prep_load_toc_cmd_buf(struct psp_gfx_cmd_resp *cmd, @@ -407,16 +414,6 @@ static int psp_tmr_init(struct psp_context *psp) AMDGPU_GEM_DOMAIN_VRAM, &psp->tmr_bo, &psp->tmr_mc_addr, pptr); - /* workaround the tmr_mc_addr: - * PSP requires an address in FB aperture. Right now driver produce - * tmr_mc_addr in the GART aperture. Convert it back to FB aperture - * for PSP. Will revert it after we get a fix from PSP FW. - */ - if (psp->adev->asic_type == CHIP_ALDEBARAN) { - psp->tmr_mc_addr -= psp->adev->gmc.fb_start; - psp->tmr_mc_addr += psp->adev->gmc.fb_start_original; - } - return ret; } @@ -466,8 +463,7 @@ static int psp_tmr_load(struct psp_context *psp) if (!cmd) return -ENOMEM; - psp_prep_tmr_cmd_buf(psp, cmd, psp->tmr_mc_addr, - amdgpu_bo_size(psp->tmr_bo)); + psp_prep_tmr_cmd_buf(psp, cmd, psp->tmr_mc_addr, psp->tmr_bo); DRM_INFO("reserve 0x%lx from 0x%llx for PSP TMR\n", amdgpu_bo_size(psp->tmr_bo), psp->tmr_mc_addr); @@ -561,7 +557,7 @@ static int psp_boot_config_set(struct amdgpu_device *adev) struct psp_context *psp = &adev->psp; struct psp_gfx_cmd_resp *cmd = psp->cmd; - if (adev->asic_type != CHIP_SIENNA_CICHLID) + if (adev->asic_type != CHIP_SIENNA_CICHLID || amdgpu_sriov_vf(adev)) return 0; memset(cmd, 0, sizeof(struct psp_gfx_cmd_resp)); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 0541196ae1ed..b0d2fc9454ca 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -114,7 +114,7 @@ static int amdgpu_reserve_page_direct(struct amdgpu_device *adev, uint64_t addre if (amdgpu_ras_check_bad_page(adev, address)) { dev_warn(adev->dev, - "RAS WARN: 0x%llx has been marked as bad page!\n", + "RAS WARN: 0x%llx has already been marked as bad page!\n", address); return 0; } @@ -221,18 +221,17 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f, op = 1; else if (sscanf(str, "inject %32s %8s", block_name, err) == 2) op = 2; - else if (sscanf(str, "retire_page") == 0) + else if (strstr(str, "retire_page") != NULL) op = 3; else if (str[0] && str[1] && str[2] && str[3]) /* ascii string, but commands are not matched. */ return -EINVAL; if (op != -1) { - if (op == 3) { - if (sscanf(str, "%*s %llu", &address) != 1) - if (sscanf(str, "%*s 0x%llx", &address) != 1) - return -EINVAL; + if (sscanf(str, "%*s 0x%llx", &address) != 1 && + sscanf(str, "%*s %llu", &address) != 1) + return -EINVAL; data->op = op; data->inject.address = address; @@ -255,11 +254,11 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f, data->op = op; if (op == 2) { - if (sscanf(str, "%*s %*s %*s %u %llu %llu", - &sub_block, &address, &value) != 3) - if (sscanf(str, "%*s %*s %*s 0x%x 0x%llx 0x%llx", - &sub_block, &address, &value) != 3) - return -EINVAL; + if (sscanf(str, "%*s %*s %*s 0x%x 0x%llx 0x%llx", + &sub_block, &address, &value) != 3 && + sscanf(str, "%*s %*s %*s %u %llu %llu", + &sub_block, &address, &value) != 3) + return -EINVAL; data->head.sub_block_index = sub_block; data->inject.address = address; data->inject.value = value; @@ -278,7 +277,7 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f, /** * DOC: AMDGPU RAS debugfs control interface * - * It accepts struct ras_debug_if who has two members. + * The control interface accepts struct ras_debug_if which has two members. * * First member: ras_debug_if::head or ras_debug_if::inject. * @@ -303,32 +302,33 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f, * * How to use the interface? * - * Programs + * In a program * - * Copy the struct ras_debug_if in your codes and initialize it. - * Write the struct to the control node. + * Copy the struct ras_debug_if in your code and initialize it. + * Write the struct to the control interface. * - * Shells + * From shell * * .. code-block:: bash * - * echo op block [error [sub_block address value]] > .../ras/ras_ctrl + * echo "disable <block>" > /sys/kernel/debug/dri/<N>/ras/ras_ctrl + * echo "enable <block> <error>" > /sys/kernel/debug/dri/<N>/ras/ras_ctrl + * echo "inject <block> <error> <sub-block> <address> <value> > /sys/kernel/debug/dri/<N>/ras/ras_ctrl * - * Parameters: + * Where N, is the card which you want to affect. * - * op: disable, enable, inject - * disable: only block is needed - * enable: block and error are needed - * inject: error, address, value are needed - * block: umc, sdma, gfx, ......... + * "disable" requires only the block. + * "enable" requires the block and error type. + * "inject" requires the block, error type, address, and value. + * The block is one of: umc, sdma, gfx, etc. * see ras_block_string[] for details - * error: ue, ce - * ue: multi_uncorrectable - * ce: single_correctable - * sub_block: - * sub block index, pass 0 if there is no sub block + * The error type is one of: ue, ce, where, + * ue is multi-uncorrectable + * ce is single-correctable + * The sub-block is a the sub-block index, pass 0 if there is no sub-block. + * The address and value are hexadecimal numbers, leading 0x is optional. * - * here are some examples for bash commands: + * For instance, * * .. code-block:: bash * @@ -336,17 +336,17 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f, * echo inject umc ce 0 0 0 > /sys/kernel/debug/dri/0/ras/ras_ctrl * echo disable umc > /sys/kernel/debug/dri/0/ras/ras_ctrl * - * How to check the result? + * How to check the result of the operation? * - * For disable/enable, please check ras features at + * To check disable/enable, see "ras" features at, * /sys/class/drm/card[0/1/2...]/device/ras/features * - * For inject, please check corresponding err count at - * /sys/class/drm/card[0/1/2...]/device/ras/[gfx/sdma/...]_err_count + * To check inject, see the corresponding error count at, + * /sys/class/drm/card[0/1/2...]/device/ras/[gfx|sdma|umc|...]_err_count * * .. note:: * Operations are only allowed on blocks which are supported. - * Please check ras mask at /sys/module/amdgpu/parameters/ras_mask + * Check the "ras" mask at /sys/module/amdgpu/parameters/ras_mask * to see which blocks support RAS on a particular asic. * */ @@ -367,11 +367,9 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f, const char __user * if (ret) return -EINVAL; - if (data.op == 3) - { + if (data.op == 3) { ret = amdgpu_reserve_page_direct(adev, data.inject.address); - - if (ret) + if (!ret) return size; else return ret; @@ -503,6 +501,12 @@ static ssize_t amdgpu_ras_sysfs_read(struct device *dev, if (amdgpu_ras_query_error_status(obj->adev, &info)) return -EINVAL; + + if (obj->adev->asic_type == CHIP_ALDEBARAN) { + if (amdgpu_ras_reset_error_status(obj->adev, info.head.block)) + DRM_WARN("Failed to reset error counter and error status"); + } + return sysfs_emit(buf, "%s: %lu\n%s: %lu\n", "ue", info.ue_count, "ce", info.ce_count); } @@ -1269,6 +1273,8 @@ static struct dentry *amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device * &amdgpu_ras_debugfs_ctrl_ops); debugfs_create_file("ras_eeprom_reset", S_IWUGO | S_IRUGO, dir, adev, &amdgpu_ras_debugfs_eeprom_ops); + debugfs_create_u32("bad_page_cnt_threshold", 0444, dir, + &con->bad_page_cnt_threshold); /* * After one uncorrectable error happens, usually GPU recovery will diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index dec0db8b0b13..9e38475e0f8d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -112,6 +112,7 @@ int amdgpu_gtt_mgr_recover(struct ttm_resource_manager *man); u64 amdgpu_vram_mgr_bo_visible_size(struct amdgpu_bo *bo); int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev, struct ttm_resource *mem, + u64 offset, u64 size, struct device *dev, enum dma_data_direction dir, struct sg_table **sgt); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 9f868cf3b832..4a3e3f72e127 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -3193,6 +3193,12 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev) { unsigned i; + /* Concurrent flushes are only possible starting with Vega10 and + * are broken on Navi10 and Navi14. + */ + adev->vm_manager.concurrent_flush = !(adev->asic_type < CHIP_VEGA10 || + adev->asic_type == CHIP_NAVI10 || + adev->asic_type == CHIP_NAVI14); amdgpu_vmid_mgr_init(adev); adev->vm_manager.fence_context = diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 6823fa589370..6fd7dad0540a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -331,6 +331,7 @@ struct amdgpu_vm_manager { /* Handling of VMIDs */ struct amdgpu_vmid_mgr id_mgr[AMDGPU_MAX_VMHUBS]; unsigned int first_kfd_vmid; + bool concurrent_flush; /* Handling of VM fences */ u64 fence_context; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c index 2344aba9dca3..f7235438535f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c @@ -25,6 +25,7 @@ #include <linux/dma-mapping.h> #include "amdgpu.h" #include "amdgpu_vm.h" +#include "amdgpu_res_cursor.h" #include "amdgpu_atomfirmware.h" #include "atom.h" @@ -566,6 +567,8 @@ static void amdgpu_vram_mgr_del(struct ttm_resource_manager *man, * * @adev: amdgpu device pointer * @mem: TTM memory object + * @offset: byte offset from the base of VRAM BO + * @length: number of bytes to export in sg_table * @dev: the other device * @dir: dma direction * @sgt: resulting sg table @@ -574,39 +577,47 @@ static void amdgpu_vram_mgr_del(struct ttm_resource_manager *man, */ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev, struct ttm_resource *mem, + u64 offset, u64 length, struct device *dev, enum dma_data_direction dir, struct sg_table **sgt) { - struct drm_mm_node *node; + struct amdgpu_res_cursor cursor; struct scatterlist *sg; int num_entries = 0; - unsigned int pages; int i, r; *sgt = kmalloc(sizeof(**sgt), GFP_KERNEL); if (!*sgt) return -ENOMEM; - for (pages = mem->num_pages, node = mem->mm_node; - pages; pages -= node->size, ++node) - ++num_entries; + /* Determine the number of DRM_MM nodes to export */ + amdgpu_res_first(mem, offset, length, &cursor); + while (cursor.remaining) { + num_entries++; + amdgpu_res_next(&cursor, cursor.size); + } r = sg_alloc_table(*sgt, num_entries, GFP_KERNEL); if (r) goto error_free; + /* Initialize scatterlist nodes of sg_table */ for_each_sgtable_sg((*sgt), sg, i) sg->length = 0; - node = mem->mm_node; + /* + * Walk down DRM_MM nodes to populate scatterlist nodes + * @note: Use iterator api to get first the DRM_MM node + * and the number of bytes from it. Access the following + * DRM_MM node(s) if more buffer needs to exported + */ + amdgpu_res_first(mem, offset, length, &cursor); for_each_sgtable_sg((*sgt), sg, i) { - phys_addr_t phys = (node->start << PAGE_SHIFT) + - adev->gmc.aper_base; - size_t size = node->size << PAGE_SHIFT; + phys_addr_t phys = cursor.start + adev->gmc.aper_base; + size_t size = cursor.size; dma_addr_t addr; - ++node; addr = dma_map_resource(dev, phys, size, dir, DMA_ATTR_SKIP_CPU_SYNC); r = dma_mapping_error(dev, addr); @@ -616,7 +627,10 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev, sg_set_page(sg, NULL, size, 0); sg_dma_address(sg) = addr; sg_dma_len(sg) = size; + + amdgpu_res_next(&cursor, cursor.size); } + return 0; error_unmap: diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c index 196d9d2a2e47..2408ed4c7d84 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c @@ -3373,7 +3373,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_3_4[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPF_GCR_CNTL, 0x0007ffff, 0x0000c000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG3, 0x00000280, 0x00000280), SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG4, 0x07800000, 0x00800000), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCR_GENERAL_CNTL, 0x00001d00, 0x00000500), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCR_GENERAL_CNTL_Sienna_Cichlid, 0x00001d00, 0x00000500), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGE_PC_CNTL, 0x003c0000, 0x00280400), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2A_ADDR_MATCH_MASK, 0xffffffff, 0xffffffcf), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGL2C_ADDR_MATCH_MASK, 0xffffffff, 0xffffffcf), diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 06811a1f4625..a078a38c2cee 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1587,6 +1587,9 @@ static int gfx_v9_0_init_cp_compute_microcode(struct amdgpu_device *adev, err = 0; adev->gfx.mec2_fw = NULL; } + } else { + adev->gfx.mec2_fw_version = adev->gfx.mec_fw_version; + adev->gfx.mec2_feature_version = adev->gfx.mec_feature_version; } if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c index 830080ff90d8..b4789dfc2bb9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c @@ -994,7 +994,7 @@ static int gfx_v9_4_ras_error_inject(struct amdgpu_device *adev, return ret; } -static const struct soc15_reg_entry gfx_v9_4_rdrsp_status_regs = +static const struct soc15_reg_entry gfx_v9_4_ea_err_status_regs = { SOC15_REG_ENTRY(GC, 0, mmGCEA_ERR_STATUS), 0, 1, 32 }; static void gfx_v9_4_query_ras_error_status(struct amdgpu_device *adev) @@ -1007,15 +1007,21 @@ static void gfx_v9_4_query_ras_error_status(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); - for (i = 0; i < gfx_v9_4_rdrsp_status_regs.se_num; i++) { - for (j = 0; j < gfx_v9_4_rdrsp_status_regs.instance; + for (i = 0; i < gfx_v9_4_ea_err_status_regs.se_num; i++) { + for (j = 0; j < gfx_v9_4_ea_err_status_regs.instance; j++) { gfx_v9_4_select_se_sh(adev, i, 0, j); reg_value = RREG32(SOC15_REG_ENTRY_OFFSET( - gfx_v9_4_rdrsp_status_regs)); - if (reg_value) + gfx_v9_4_ea_err_status_regs)); + if (REG_GET_FIELD(reg_value, GCEA_ERR_STATUS, SDP_RDRSP_STATUS) || + REG_GET_FIELD(reg_value, GCEA_ERR_STATUS, SDP_WRRSP_STATUS) || + REG_GET_FIELD(reg_value, GCEA_ERR_STATUS, SDP_RDRSP_DATAPARITY_ERROR)) { + /* SDP read/write error/parity error in FUE_IS_FATAL mode + * can cause system fatal error in arcturas. Harvest the error + * status before GPU reset */ dev_warn(adev->dev, "GCEA err detected at instance: %d, status: 0x%x!\n", j, reg_value); + } } } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c index 9ca76a3ac38c..a30c7c10cd9a 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c @@ -808,7 +808,7 @@ static struct gfx_v9_4_2_utc_block gfx_v9_4_2_utc_blocks[] = { REG_SET_FIELD(0, ATC_L2_CACHE_4K_DSM_CNTL, WRITE_COUNTERS, 1) }, }; -static const struct soc15_reg_entry gfx_v9_4_2_rdrsp_status_regs = +static const struct soc15_reg_entry gfx_v9_4_2_ea_err_status_regs = { SOC15_REG_ENTRY(GC, 0, regGCEA_ERR_STATUS), 0, 1, 16 }; static int gfx_v9_4_2_get_reg_error_count(struct amdgpu_device *adev, @@ -997,8 +997,9 @@ static int gfx_v9_4_2_query_utc_edc_count(struct amdgpu_device *adev, blk->clear); /* print the edc count */ - gfx_v9_4_2_log_utc_edc_count(adev, blk, j, sec_cnt, - ded_cnt); + if (sec_cnt || ded_cnt) + gfx_v9_4_2_log_utc_edc_count(adev, blk, j, sec_cnt, + ded_cnt); } } @@ -1040,11 +1041,11 @@ static void gfx_v9_4_2_reset_ea_err_status(struct amdgpu_device *adev) uint32_t i, j; mutex_lock(&adev->grbm_idx_mutex); - for (i = 0; i < gfx_v9_4_2_rdrsp_status_regs.se_num; i++) { - for (j = 0; j < gfx_v9_4_2_rdrsp_status_regs.instance; + for (i = 0; i < gfx_v9_4_2_ea_err_status_regs.se_num; i++) { + for (j = 0; j < gfx_v9_4_2_ea_err_status_regs.instance; j++) { gfx_v9_4_2_select_se_sh(adev, i, 0, j); - WREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_4_2_rdrsp_status_regs), 0x10); + WREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_4_2_ea_err_status_regs), 0x10); } } gfx_v9_4_2_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); @@ -1089,17 +1090,20 @@ static void gfx_v9_4_2_query_ea_err_status(struct amdgpu_device *adev) mutex_lock(&adev->grbm_idx_mutex); - for (i = 0; i < gfx_v9_4_2_rdrsp_status_regs.se_num; i++) { - for (j = 0; j < gfx_v9_4_2_rdrsp_status_regs.instance; + for (i = 0; i < gfx_v9_4_2_ea_err_status_regs.se_num; i++) { + for (j = 0; j < gfx_v9_4_2_ea_err_status_regs.instance; j++) { gfx_v9_4_2_select_se_sh(adev, i, 0, j); reg_value = RREG32(SOC15_REG_ENTRY_OFFSET( - gfx_v9_4_2_rdrsp_status_regs)); - if (reg_value) + gfx_v9_4_2_ea_err_status_regs)); + if (REG_GET_FIELD(reg_value, GCEA_ERR_STATUS, SDP_RDRSP_STATUS) || + REG_GET_FIELD(reg_value, GCEA_ERR_STATUS, SDP_WRRSP_STATUS) || + REG_GET_FIELD(reg_value, GCEA_ERR_STATUS, SDP_RDRSP_DATAPARITY_ERROR)) { dev_warn(adev->dev, "GCEA err detected at instance: %d, status: 0x%x!\n", j, reg_value); + } /* clear after read */ - WREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_4_2_rdrsp_status_regs), 0x10); + WREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_4_2_ea_err_status_regs), 0x10); } } @@ -1112,19 +1116,19 @@ static void gfx_v9_4_2_query_utc_err_status(struct amdgpu_device *adev) uint32_t data; data = RREG32_SOC15(GC, 0, regUTCL2_MEM_ECC_STATUS); - if (!data) { + if (data) { dev_warn(adev->dev, "GFX UTCL2 Mem Ecc Status: 0x%x!\n", data); WREG32_SOC15(GC, 0, regUTCL2_MEM_ECC_STATUS, 0x3); } data = RREG32_SOC15(GC, 0, regVML2_MEM_ECC_STATUS); - if (!data) { + if (data) { dev_warn(adev->dev, "GFX VML2 Mem Ecc Status: 0x%x!\n", data); WREG32_SOC15(GC, 0, regVML2_MEM_ECC_STATUS, 0x3); } data = RREG32_SOC15(GC, 0, regVML2_WALKER_MEM_ECC_STATUS); - if (!data) { + if (data) { dev_warn(adev->dev, "GFX VML2 Walker Mem Ecc Status: 0x%x!\n", data); WREG32_SOC15(GC, 0, regVML2_WALKER_MEM_ECC_STATUS, 0x3); } diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c index d189507dcef0..1e4678cb98f0 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c @@ -120,8 +120,7 @@ static void gfxhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev) max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18); /* Set default page address. */ - value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + - adev->vm_manager.vram_base_offset; + value = amdgpu_gmc_vram_mc2pa(adev, adev->vram_scratch.gpu_addr); WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, (u32)(value >> 12)); WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, @@ -141,21 +140,12 @@ static void gfxhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev) * FB aperture and AGP aperture. Disable them. */ if (adev->gmc.pdb0_bo) { - if (adev->asic_type == CHIP_ALDEBARAN) { - WREG32_SOC15(GC, 0, mmMC_VM_FB_LOCATION_TOP, adev->gmc.fb_end_original >> 24); - WREG32_SOC15(GC, 0, mmMC_VM_FB_LOCATION_BASE, adev->gmc.fb_start_original >> 24); - WREG32_SOC15(GC, 0, mmMC_VM_AGP_TOP, 0); - WREG32_SOC15(GC, 0, mmMC_VM_AGP_BOT, 0xFFFFFF); - WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_LOW_ADDR, adev->gmc.fb_start_original >> 18); - WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, adev->gmc.fb_end_original >> 18); - } else { - WREG32_SOC15(GC, 0, mmMC_VM_FB_LOCATION_TOP, 0); - WREG32_SOC15(GC, 0, mmMC_VM_FB_LOCATION_BASE, 0x00FFFFFF); - WREG32_SOC15(GC, 0, mmMC_VM_AGP_TOP, 0); - WREG32_SOC15(GC, 0, mmMC_VM_AGP_BOT, 0xFFFFFF); - WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x3FFFFFFF); - WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0); - } + WREG32_SOC15(GC, 0, mmMC_VM_FB_LOCATION_TOP, 0); + WREG32_SOC15(GC, 0, mmMC_VM_FB_LOCATION_BASE, 0x00FFFFFF); + WREG32_SOC15(GC, 0, mmMC_VM_AGP_TOP, 0); + WREG32_SOC15(GC, 0, mmMC_VM_AGP_BOT, 0xFFFFFF); + WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x3FFFFFFF); + WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0); } } diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c index 2aecc6a243e8..14c1c1a297dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c @@ -165,8 +165,7 @@ static void gfxhub_v2_0_init_system_aperture_regs(struct amdgpu_device *adev) max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18); /* Set default page address. */ - value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start - + adev->vm_manager.vram_base_offset; + value = amdgpu_gmc_vram_mc2pa(adev, adev->vram_scratch.gpu_addr); WREG32_SOC15(GC, 0, mmGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, (u32)(value >> 12)); WREG32_SOC15(GC, 0, mmGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c index 410fd3a1a388..41807817de7d 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c @@ -164,8 +164,7 @@ static void gfxhub_v2_1_init_system_aperture_regs(struct amdgpu_device *adev) max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18); /* Set default page address. */ - value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start - + adev->vm_manager.vram_base_offset; + value = amdgpu_gmc_vram_mc2pa(adev, adev->vram_scratch.gpu_addr); WREG32_SOC15(GC, 0, mmGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, (u32)(value >> 12)); WREG32_SOC15(GC, 0, mmGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c index 2bfd620576f2..498b28a35f5b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c @@ -568,8 +568,7 @@ static void gmc_v10_0_get_vm_pde(struct amdgpu_device *adev, int level, uint64_t *addr, uint64_t *flags) { if (!(*flags & AMDGPU_PDE_PTE) && !(*flags & AMDGPU_PTE_SYSTEM)) - *addr = adev->vm_manager.vram_base_offset + *addr - - adev->gmc.vram_start; + *addr = amdgpu_gmc_vram_mc2pa(adev, *addr); BUG_ON(*addr & 0xFFFF00000000003FULL); if (!adev->gmc.translate_further) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index c1bd190841f8..e4f27b3f28fb 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -59,6 +59,7 @@ MODULE_FIRMWARE("amdgpu/tonga_mc.bin"); MODULE_FIRMWARE("amdgpu/polaris11_mc.bin"); MODULE_FIRMWARE("amdgpu/polaris10_mc.bin"); MODULE_FIRMWARE("amdgpu/polaris12_mc.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_32_mc.bin"); MODULE_FIRMWARE("amdgpu/polaris11_k_mc.bin"); MODULE_FIRMWARE("amdgpu/polaris10_k_mc.bin"); MODULE_FIRMWARE("amdgpu/polaris12_k_mc.bin"); @@ -243,10 +244,16 @@ static int gmc_v8_0_init_microcode(struct amdgpu_device *adev) chip_name = "polaris10"; break; case CHIP_POLARIS12: - if (ASICID_IS_P23(adev->pdev->device, adev->pdev->revision)) + if (ASICID_IS_P23(adev->pdev->device, adev->pdev->revision)) { chip_name = "polaris12_k"; - else - chip_name = "polaris12"; + } else { + WREG32(mmMC_SEQ_IO_DEBUG_INDEX, ixMC_IO_DEBUG_UP_159); + /* Polaris12 32bit ASIC needs a special MC firmware */ + if (RREG32(mmMC_SEQ_IO_DEBUG_DATA) == 0x05b4dc40) + chip_name = "polaris12_32"; + else + chip_name = "polaris12"; + } break; case CHIP_FIJI: case CHIP_CARRIZO: diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index c82d82da2c73..455bb91060d0 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -574,7 +574,8 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev, * be updated to avoid reading an incorrect value due to * the new fast GRBM interface. */ - if (entry->vmid_src == AMDGPU_GFXHUB_0) + if ((entry->vmid_src == AMDGPU_GFXHUB_0) && + (adev->asic_type < CHIP_ALDEBARAN)) RREG32(hub->vm_l2_pro_fault_status); status = RREG32(hub->vm_l2_pro_fault_status); @@ -802,7 +803,8 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid, * be cleared to avoid a false ACK due to the new fast * GRBM interface. */ - if (vmhub == AMDGPU_GFXHUB_0) + if ((vmhub == AMDGPU_GFXHUB_0) && + (adev->asic_type < CHIP_ALDEBARAN)) RREG32_NO_KIQ(hub->vm_inv_eng0_req + hub->eng_distance * eng); @@ -1048,8 +1050,7 @@ static void gmc_v9_0_get_vm_pde(struct amdgpu_device *adev, int level, uint64_t *addr, uint64_t *flags) { if (!(*flags & AMDGPU_PDE_PTE) && !(*flags & AMDGPU_PTE_SYSTEM)) - *addr = adev->vm_manager.vram_base_offset + *addr - - adev->gmc.vram_start; + *addr = amdgpu_gmc_vram_mc2pa(adev, *addr); BUG_ON(*addr & 0xFFFF00000000003FULL); if (!adev->gmc.translate_further) diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c index aa9be5612c89..a99953833820 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -114,8 +114,7 @@ static void mmhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev) return; /* Set default page address. */ - value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + - adev->vm_manager.vram_base_offset; + value = amdgpu_gmc_vram_mc2pa(adev, adev->vram_scratch.gpu_addr); WREG32_SOC15(MMHUB, 0, mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, (u32)(value >> 12)); WREG32_SOC15(MMHUB, 0, mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c index 7977a7879b32..0103a5ab28e6 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c @@ -47,8 +47,6 @@ static u64 mmhub_v1_7_get_fb_location(struct amdgpu_device *adev) adev->gmc.fb_start = base; adev->gmc.fb_end = top; - adev->gmc.fb_start_original = base; - adev->gmc.fb_end_original = top; return base; } @@ -126,17 +124,16 @@ static void mmhub_v1_7_init_system_aperture_regs(struct amdgpu_device *adev) if (adev->gmc.pdb0_bo) { WREG32_SOC15(MMHUB, 0, regMC_VM_AGP_BOT, 0xFFFFFF); WREG32_SOC15(MMHUB, 0, regMC_VM_AGP_TOP, 0); - WREG32_SOC15(MMHUB, 0, regMC_VM_FB_LOCATION_TOP, adev->gmc.fb_end_original >> 24); - WREG32_SOC15(MMHUB, 0, regMC_VM_FB_LOCATION_BASE, adev->gmc.fb_start_original >> 24); - WREG32_SOC15(MMHUB, 0, regMC_VM_SYSTEM_APERTURE_LOW_ADDR, adev->gmc.fb_start_original >> 18); - WREG32_SOC15(MMHUB, 0, regMC_VM_SYSTEM_APERTURE_HIGH_ADDR, adev->gmc.fb_end_original >> 18); + WREG32_SOC15(MMHUB, 0, regMC_VM_FB_LOCATION_TOP, 0); + WREG32_SOC15(MMHUB, 0, regMC_VM_FB_LOCATION_BASE, 0x00FFFFFF); + WREG32_SOC15(MMHUB, 0, regMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x3FFFFFFF); + WREG32_SOC15(MMHUB, 0, regMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0); } if (amdgpu_sriov_vf(adev)) return; /* Set default page address. */ - value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + - adev->vm_manager.vram_base_offset; + value = amdgpu_gmc_vram_mc2pa(adev, adev->vram_scratch.gpu_addr); WREG32_SOC15(MMHUB, 0, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, (u32)(value >> 12)); WREG32_SOC15(MMHUB, 0, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, @@ -1287,7 +1284,7 @@ static void mmhub_v1_7_reset_ras_error_count(struct amdgpu_device *adev) } } -static const struct soc15_reg_entry mmhub_v1_7_err_status_regs[] = { +static const struct soc15_reg_entry mmhub_v1_7_ea_err_status_regs[] = { { SOC15_REG_ENTRY(MMHUB, 0, regMMEA0_ERR_STATUS), 0, 0, 0 }, { SOC15_REG_ENTRY(MMHUB, 0, regMMEA1_ERR_STATUS), 0, 0, 0 }, { SOC15_REG_ENTRY(MMHUB, 0, regMMEA2_ERR_STATUS), 0, 0, 0 }, @@ -1304,12 +1301,15 @@ static void mmhub_v1_7_query_ras_error_status(struct amdgpu_device *adev) if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__MMHUB)) return; - for (i = 0; i < ARRAY_SIZE(mmhub_v1_7_err_status_regs); i++) { + for (i = 0; i < ARRAY_SIZE(mmhub_v1_7_ea_err_status_regs); i++) { reg_value = - RREG32(SOC15_REG_ENTRY_OFFSET(mmhub_v1_7_err_status_regs[i])); - if (reg_value) + RREG32(SOC15_REG_ENTRY_OFFSET(mmhub_v1_7_ea_err_status_regs[i])); + if (REG_GET_FIELD(reg_value, MMEA0_ERR_STATUS, SDP_RDRSP_STATUS) || + REG_GET_FIELD(reg_value, MMEA0_ERR_STATUS, SDP_WRRSP_STATUS) || + REG_GET_FIELD(reg_value, MMEA0_ERR_STATUS, SDP_RDRSP_DATAPARITY_ERROR)) { dev_warn(adev->dev, "MMHUB EA err detected at instance: %d, status: 0x%x!\n", i, reg_value); + } } } diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c index da7edd1ed6b2..ac76081b91d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c @@ -210,8 +210,7 @@ static void mmhub_v2_0_init_system_aperture_regs(struct amdgpu_device *adev) } /* Set default page address. */ - value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + - adev->vm_manager.vram_base_offset; + value = amdgpu_gmc_vram_mc2pa(adev, adev->vram_scratch.gpu_addr); WREG32_SOC15(MMHUB, 0, mmMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, (u32)(value >> 12)); WREG32_SOC15(MMHUB, 0, mmMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c index 1141c37432f0..a9899335d0b1 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c @@ -162,8 +162,7 @@ static void mmhub_v2_3_init_system_aperture_regs(struct amdgpu_device *adev) max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18); /* Set default page address. */ - value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + - adev->vm_manager.vram_base_offset; + value = amdgpu_gmc_vram_mc2pa(adev, adev->vram_scratch.gpu_addr); WREG32_SOC15(MMHUB, 0, mmMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, (u32)(value >> 12)); WREG32_SOC15(MMHUB, 0, mmMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c index 0cffa820ea6e..47c8dd9d1c78 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c @@ -136,8 +136,7 @@ static void mmhub_v9_4_init_system_aperture_regs(struct amdgpu_device *adev, max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18); /* Set default page address. */ - value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + - adev->vm_manager.vram_base_offset; + value = amdgpu_gmc_vram_mc2pa(adev, adev->vram_scratch.gpu_addr); WREG32_SOC15_OFFSET( MMHUB, 0, mmVMSHAREDPF0_MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, @@ -1646,9 +1645,15 @@ static void mmhub_v9_4_query_ras_error_status(struct amdgpu_device *adev) for (i = 0; i < ARRAY_SIZE(mmhub_v9_4_err_status_regs); i++) { reg_value = RREG32(SOC15_REG_ENTRY_OFFSET(mmhub_v9_4_err_status_regs[i])); - if (reg_value) + if (REG_GET_FIELD(reg_value, MMEA0_ERR_STATUS, SDP_RDRSP_STATUS) || + REG_GET_FIELD(reg_value, MMEA0_ERR_STATUS, SDP_WRRSP_STATUS) || + REG_GET_FIELD(reg_value, MMEA0_ERR_STATUS, SDP_RDRSP_DATAPARITY_ERROR)) { + /* SDP read/write error/parity error in FUE_IS_FATAL mode + * can cause system fatal error in arcturas. Harvest the error + * status before GPU reset */ dev_warn(adev->dev, "MMHUB EA err detected at instance: %d, status: 0x%x!\n", i, reg_value); + } } } diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c index 83ea063388fd..0d2d629e2d6a 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c @@ -31,6 +31,28 @@ #include "vega10_enum.h" #include <uapi/linux/kfd_ioctl.h> +#define smnPCIE_LC_CNTL 0x11140280 +#define smnPCIE_LC_CNTL3 0x111402d4 +#define smnPCIE_LC_CNTL6 0x111402ec +#define smnPCIE_LC_CNTL7 0x111402f0 +#define smnNBIF_MGCG_CTRL_LCLK 0x1013a05c +#define NBIF_MGCG_CTRL_LCLK__NBIF_MGCG_REG_DIS_LCLK_MASK 0x00001000L +#define RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER_MASK 0x0000FFFFL +#define RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER_MASK 0xFFFF0000L +#define smnRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL 0x10123530 +#define smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2 0x1014008c +#define smnBIF_CFG_DEV0_EPF0_PCIE_LTR_CAP 0x10140324 +#define smnPSWUSP0_PCIE_LC_CNTL2 0x111402c4 +#define smnRCC_BIF_STRAP2 0x10123488 +#define smnRCC_BIF_STRAP3 0x1012348c +#define smnRCC_BIF_STRAP5 0x10123494 +#define BIF_CFG_DEV0_EPF0_DEVICE_CNTL2__LTR_EN_MASK 0x0400L +#define RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER_MASK 0x0000FFFFL +#define RCC_BIF_STRAP2__STRAP_LTR_IN_ASPML1_DIS_MASK 0x00004000L +#define RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER__SHIFT 0x0 +#define RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER__SHIFT 0x10 +#define RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER__SHIFT 0x0 + static void nbio_v6_1_remap_hdp_registers(struct amdgpu_device *adev) { WREG32_SOC15(NBIO, 0, mmREMAP_HDP_MEM_FLUSH_CNTL, @@ -256,6 +278,111 @@ static void nbio_v6_1_init_registers(struct amdgpu_device *adev) WREG32_PCIE(smnPCIE_CI_CNTL, data); } +static void nbio_v6_1_program_ltr(struct amdgpu_device *adev) +{ + uint32_t def, data; + + WREG32_PCIE(smnRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL, 0x75EB); + + def = data = RREG32_PCIE(smnRCC_BIF_STRAP2); + data &= ~RCC_BIF_STRAP2__STRAP_LTR_IN_ASPML1_DIS_MASK; + if (def != data) + WREG32_PCIE(smnRCC_BIF_STRAP2, data); + + def = data = RREG32_PCIE(smnRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL); + data &= ~EP_PCIE_TX_LTR_CNTL__LTR_PRIV_MSG_DIS_IN_PM_NON_D0_MASK; + if (def != data) + WREG32_PCIE(smnRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL, data); + + def = data = RREG32_PCIE(smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2); + data |= BIF_CFG_DEV0_EPF0_DEVICE_CNTL2__LTR_EN_MASK; + if (def != data) + WREG32_PCIE(smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2, data); +} + +static void nbio_v6_1_program_aspm(struct amdgpu_device *adev) +{ + uint32_t def, data; + + def = data = RREG32_PCIE(smnPCIE_LC_CNTL); + data &= ~PCIE_LC_CNTL__LC_L1_INACTIVITY_MASK; + data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK; + data |= PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK; + if (def != data) + WREG32_PCIE(smnPCIE_LC_CNTL, data); + + def = data = RREG32_PCIE(smnPCIE_LC_CNTL7); + data |= PCIE_LC_CNTL7__LC_NBIF_ASPM_INPUT_EN_MASK; + if (def != data) + WREG32_PCIE(smnPCIE_LC_CNTL7, data); + + def = data = RREG32_PCIE(smnNBIF_MGCG_CTRL_LCLK); + data |= NBIF_MGCG_CTRL_LCLK__NBIF_MGCG_REG_DIS_LCLK_MASK; + if (def != data) + WREG32_PCIE(smnNBIF_MGCG_CTRL_LCLK, data); + + def = data = RREG32_PCIE(smnPCIE_LC_CNTL3); + data |= PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK; + if (def != data) + WREG32_PCIE(smnPCIE_LC_CNTL3, data); + + def = data = RREG32_PCIE(smnRCC_BIF_STRAP3); + data &= ~RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER_MASK; + data &= ~RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER_MASK; + if (def != data) + WREG32_PCIE(smnRCC_BIF_STRAP3, data); + + def = data = RREG32_PCIE(smnRCC_BIF_STRAP5); + data &= ~RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER_MASK; + if (def != data) + WREG32_PCIE(smnRCC_BIF_STRAP5, data); + + def = data = RREG32_PCIE(smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2); + data &= ~BIF_CFG_DEV0_EPF0_DEVICE_CNTL2__LTR_EN_MASK; + if (def != data) + WREG32_PCIE(smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2, data); + + WREG32_PCIE(smnBIF_CFG_DEV0_EPF0_PCIE_LTR_CAP, 0x10011001); + + def = data = RREG32_PCIE(smnPSWUSP0_PCIE_LC_CNTL2); + data |= PSWUSP0_PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L1_MASK | + PSWUSP0_PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L23_MASK; + data &= ~PSWUSP0_PCIE_LC_CNTL2__LC_RCV_L0_TO_RCV_L0S_DIS_MASK; + if (def != data) + WREG32_PCIE(smnPSWUSP0_PCIE_LC_CNTL2, data); + + def = data = RREG32_PCIE(smnPCIE_LC_CNTL6); + data |= PCIE_LC_CNTL6__LC_L1_POWERDOWN_MASK | + PCIE_LC_CNTL6__LC_RX_L0S_STANDBY_EN_MASK; + if (def != data) + WREG32_PCIE(smnPCIE_LC_CNTL6, data); + + nbio_v6_1_program_ltr(adev); + + def = data = RREG32_PCIE(smnRCC_BIF_STRAP3); + data |= 0x5DE0 << RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER__SHIFT; + data |= 0x0010 << RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER__SHIFT; + if (def != data) + WREG32_PCIE(smnRCC_BIF_STRAP3, data); + + def = data = RREG32_PCIE(smnRCC_BIF_STRAP5); + data |= 0x0010 << RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER__SHIFT; + if (def != data) + WREG32_PCIE(smnRCC_BIF_STRAP5, data); + + def = data = RREG32_PCIE(smnPCIE_LC_CNTL); + data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK; + data |= 0x9 << PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT; + data |= 0x1 << PCIE_LC_CNTL__LC_PMI_TO_L1_DIS__SHIFT; + if (def != data) + WREG32_PCIE(smnPCIE_LC_CNTL, data); + + def = data = RREG32_PCIE(smnPCIE_LC_CNTL3); + data &= ~PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK; + if (def != data) + WREG32_PCIE(smnPCIE_LC_CNTL3, data); +} + const struct amdgpu_nbio_funcs nbio_v6_1_funcs = { .get_hdp_flush_req_offset = nbio_v6_1_get_hdp_flush_req_offset, .get_hdp_flush_done_offset = nbio_v6_1_get_hdp_flush_done_offset, @@ -274,4 +401,5 @@ const struct amdgpu_nbio_funcs nbio_v6_1_funcs = { .ih_control = nbio_v6_1_ih_control, .init_registers = nbio_v6_1_init_registers, .remap_hdp_registers = nbio_v6_1_remap_hdp_registers, + .program_aspm = nbio_v6_1_program_aspm, }; diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c index af44aad78171..cef929746739 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c @@ -31,7 +31,26 @@ #include "ivsrcid/nbio/irqsrcs_nbif_7_4.h" #include <uapi/linux/kfd_ioctl.h> +#define smnPCIE_LC_CNTL 0x11140280 +#define smnPCIE_LC_CNTL3 0x111402d4 +#define smnPCIE_LC_CNTL6 0x111402ec +#define smnPCIE_LC_CNTL7 0x111402f0 #define smnNBIF_MGCG_CTRL_LCLK 0x1013a21c +#define smnRCC_BIF_STRAP3 0x1012348c +#define RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER_MASK 0x0000FFFFL +#define RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER_MASK 0xFFFF0000L +#define smnRCC_BIF_STRAP5 0x10123494 +#define RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER_MASK 0x0000FFFFL +#define smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2 0x1014008c +#define BIF_CFG_DEV0_EPF0_DEVICE_CNTL2__LTR_EN_MASK 0x0400L +#define smnBIF_CFG_DEV0_EPF0_PCIE_LTR_CAP 0x10140324 +#define smnPSWUSP0_PCIE_LC_CNTL2 0x111402c4 +#define smnRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL 0x10123538 +#define smnRCC_BIF_STRAP2 0x10123488 +#define RCC_BIF_STRAP2__STRAP_LTR_IN_ASPML1_DIS_MASK 0x00004000L +#define RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER__SHIFT 0x0 +#define RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER__SHIFT 0x10 +#define RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER__SHIFT 0x0 /* * These are nbio v7_4_1 registers mask. Temporarily define these here since @@ -567,6 +586,111 @@ const struct amdgpu_nbio_ras_funcs nbio_v7_4_ras_funcs = { .ras_fini = amdgpu_nbio_ras_fini, }; +static void nbio_v7_4_program_ltr(struct amdgpu_device *adev) +{ + uint32_t def, data; + + WREG32_PCIE(smnRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL, 0x75EB); + + def = data = RREG32_PCIE(smnRCC_BIF_STRAP2); + data &= ~RCC_BIF_STRAP2__STRAP_LTR_IN_ASPML1_DIS_MASK; + if (def != data) + WREG32_PCIE(smnRCC_BIF_STRAP2, data); + + def = data = RREG32_PCIE(smnRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL); + data &= ~EP_PCIE_TX_LTR_CNTL__LTR_PRIV_MSG_DIS_IN_PM_NON_D0_MASK; + if (def != data) + WREG32_PCIE(smnRCC_EP_DEV0_0_EP_PCIE_TX_LTR_CNTL, data); + + def = data = RREG32_PCIE(smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2); + data |= BIF_CFG_DEV0_EPF0_DEVICE_CNTL2__LTR_EN_MASK; + if (def != data) + WREG32_PCIE(smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2, data); +} + +static void nbio_v7_4_program_aspm(struct amdgpu_device *adev) +{ + uint32_t def, data; + + def = data = RREG32_PCIE(smnPCIE_LC_CNTL); + data &= ~PCIE_LC_CNTL__LC_L1_INACTIVITY_MASK; + data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK; + data |= PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK; + if (def != data) + WREG32_PCIE(smnPCIE_LC_CNTL, data); + + def = data = RREG32_PCIE(smnPCIE_LC_CNTL7); + data |= PCIE_LC_CNTL7__LC_NBIF_ASPM_INPUT_EN_MASK; + if (def != data) + WREG32_PCIE(smnPCIE_LC_CNTL7, data); + + def = data = RREG32_PCIE(smnNBIF_MGCG_CTRL_LCLK); + data |= NBIF_MGCG_CTRL_LCLK__NBIF_MGCG_REG_DIS_LCLK_MASK; + if (def != data) + WREG32_PCIE(smnNBIF_MGCG_CTRL_LCLK, data); + + def = data = RREG32_PCIE(smnPCIE_LC_CNTL3); + data |= PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK; + if (def != data) + WREG32_PCIE(smnPCIE_LC_CNTL3, data); + + def = data = RREG32_PCIE(smnRCC_BIF_STRAP3); + data &= ~RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER_MASK; + data &= ~RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER_MASK; + if (def != data) + WREG32_PCIE(smnRCC_BIF_STRAP3, data); + + def = data = RREG32_PCIE(smnRCC_BIF_STRAP5); + data &= ~RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER_MASK; + if (def != data) + WREG32_PCIE(smnRCC_BIF_STRAP5, data); + + def = data = RREG32_PCIE(smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2); + data &= ~BIF_CFG_DEV0_EPF0_DEVICE_CNTL2__LTR_EN_MASK; + if (def != data) + WREG32_PCIE(smnBIF_CFG_DEV0_EPF0_DEVICE_CNTL2, data); + + WREG32_PCIE(smnBIF_CFG_DEV0_EPF0_PCIE_LTR_CAP, 0x10011001); + + def = data = RREG32_PCIE(smnPSWUSP0_PCIE_LC_CNTL2); + data |= PSWUSP0_PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L1_MASK | + PSWUSP0_PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L23_MASK; + data &= ~PSWUSP0_PCIE_LC_CNTL2__LC_RCV_L0_TO_RCV_L0S_DIS_MASK; + if (def != data) + WREG32_PCIE(smnPSWUSP0_PCIE_LC_CNTL2, data); + + def = data = RREG32_PCIE(smnPCIE_LC_CNTL6); + data |= PCIE_LC_CNTL6__LC_L1_POWERDOWN_MASK | + PCIE_LC_CNTL6__LC_RX_L0S_STANDBY_EN_MASK; + if (def != data) + WREG32_PCIE(smnPCIE_LC_CNTL6, data); + + nbio_v7_4_program_ltr(adev); + + def = data = RREG32_PCIE(smnRCC_BIF_STRAP3); + data |= 0x5DE0 << RCC_BIF_STRAP3__STRAP_VLINK_ASPM_IDLE_TIMER__SHIFT; + data |= 0x0010 << RCC_BIF_STRAP3__STRAP_VLINK_PM_L1_ENTRY_TIMER__SHIFT; + if (def != data) + WREG32_PCIE(smnRCC_BIF_STRAP3, data); + + def = data = RREG32_PCIE(smnRCC_BIF_STRAP5); + data |= 0x0010 << RCC_BIF_STRAP5__STRAP_VLINK_LDN_ENTRY_TIMER__SHIFT; + if (def != data) + WREG32_PCIE(smnRCC_BIF_STRAP5, data); + + def = data = RREG32_PCIE(smnPCIE_LC_CNTL); + data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK; + data |= 0x9 << PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT; + data |= 0x1 << PCIE_LC_CNTL__LC_PMI_TO_L1_DIS__SHIFT; + if (def != data) + WREG32_PCIE(smnPCIE_LC_CNTL, data); + + def = data = RREG32_PCIE(smnPCIE_LC_CNTL3); + data &= ~PCIE_LC_CNTL3__LC_DSC_DONT_ENTER_L23_AFTER_PME_ACK_MASK; + if (def != data) + WREG32_PCIE(smnPCIE_LC_CNTL3, data); +} + const struct amdgpu_nbio_funcs nbio_v7_4_funcs = { .get_hdp_flush_req_offset = nbio_v7_4_get_hdp_flush_req_offset, .get_hdp_flush_done_offset = nbio_v7_4_get_hdp_flush_done_offset, @@ -587,4 +711,5 @@ const struct amdgpu_nbio_funcs nbio_v7_4_funcs = { .ih_control = nbio_v7_4_ih_control, .init_registers = nbio_v7_4_init_registers, .remap_hdp_registers = nbio_v7_4_remap_hdp_registers, + .program_aspm = nbio_v7_4_program_aspm, }; diff --git a/drivers/gpu/drm/amd/amdgpu/nv.c b/drivers/gpu/drm/amd/amdgpu/nv.c index 46d4bbabce75..d54af7f8801b 100644 --- a/drivers/gpu/drm/amd/amdgpu/nv.c +++ b/drivers/gpu/drm/amd/amdgpu/nv.c @@ -601,8 +601,7 @@ static void nv_program_aspm(struct amdgpu_device *adev) if (amdgpu_aspm != 1) return; - if ((adev->asic_type >= CHIP_SIENNA_CICHLID) && - !(adev->flags & AMD_IS_APU) && + if (!(adev->flags & AMD_IS_APU) && (adev->nbio.funcs->program_aspm)) adev->nbio.funcs->program_aspm(adev); @@ -934,12 +933,7 @@ static int nv_update_umd_stable_pstate(struct amdgpu_device *adev, if (adev->gfx.funcs->update_perfmon_mgcg) adev->gfx.funcs->update_perfmon_mgcg(adev, !enter); - /* - * The ASPM function is not fully enabled and verified on - * Navi yet. Temporarily skip this until ASPM enabled. - */ - if ((adev->asic_type >= CHIP_SIENNA_CICHLID) && - !(adev->flags & AMD_IS_APU) && + if (!(adev->flags & AMD_IS_APU) && (adev->nbio.funcs->enable_aspm)) adev->nbio.funcs->enable_aspm(adev, !enter); diff --git a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h index dd4d65f7e0f0..96064c343163 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h @@ -185,10 +185,19 @@ struct psp_gfx_cmd_setup_tmr uint32_t buf_phy_addr_lo; /* bits [31:0] of GPU Virtual address of TMR buffer (must be 4 KB aligned) */ uint32_t buf_phy_addr_hi; /* bits [63:32] of GPU Virtual address of TMR buffer */ uint32_t buf_size; /* buffer size in bytes (must be multiple of 4 KB) */ + union { + struct { + uint32_t sriov_enabled:1; /* whether the device runs under SR-IOV*/ + uint32_t virt_phy_addr:1; /* driver passes both virtual and physical address to PSP*/ + uint32_t reserved:30; + } bitfield; + uint32_t tmr_flags; + }; + uint32_t system_phy_addr_lo; /* bits [31:0] of system physical address of TMR buffer (must be 4 KB aligned) */ + uint32_t system_phy_addr_hi; /* bits [63:32] of system physical address of TMR buffer */ }; - /* FW types for GFX_CMD_ID_LOAD_IP_FW command. Limit 31. */ enum psp_gfx_fw_type { GFX_FW_TYPE_NONE = 0, /* */ diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c index 6fcb95c89999..bf95007f0843 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c @@ -160,6 +160,7 @@ static const struct soc15_ras_field_entry sdma_v4_4_ras_fields[] = { }; static void sdma_v4_4_get_ras_error_count(struct amdgpu_device *adev, + uint32_t reg_offset, uint32_t value, uint32_t instance, uint32_t *sec_count) @@ -169,6 +170,9 @@ static void sdma_v4_4_get_ras_error_count(struct amdgpu_device *adev, /* double bits error (multiple bits) error detection is not supported */ for (i = 0; i < ARRAY_SIZE(sdma_v4_4_ras_fields); i++) { + if (sdma_v4_4_ras_fields[i].reg_offset != reg_offset) + continue; + /* the SDMA_EDC_COUNTER register in each sdma instance * shares the same sed shift_mask * */ @@ -197,13 +201,30 @@ static int sdma_v4_4_query_ras_error_count(struct amdgpu_device *adev, reg_value = RREG32(reg_offset); /* double bit error is not supported */ if (reg_value) - sdma_v4_4_get_ras_error_count(adev, reg_value, instance, &sec_count); - /* err_data->ce_count should be initialized to 0 - * before calling into this function */ - err_data->ce_count += sec_count; - /* double bit error is not supported - * set ue count to 0 */ - err_data->ue_count = 0; + sdma_v4_4_get_ras_error_count(adev, regSDMA0_EDC_COUNTER, reg_value, + instance, &sec_count); + + reg_offset = sdma_v4_4_get_reg_offset(adev, instance, regSDMA0_EDC_COUNTER2); + reg_value = RREG32(reg_offset); + /* double bit error is not supported */ + if (reg_value) + sdma_v4_4_get_ras_error_count(adev, regSDMA0_EDC_COUNTER2, reg_value, + instance, &sec_count); + + /* + * err_data->ue_count should be initialized to 0 + * before calling into this function + * + * SDMA RAS supports single bit uncorrectable error detection. + * So, increment uncorrectable error count. + */ + err_data->ue_count += sec_count; + + /* + * SDMA RAS does not support correctable errors. + * Set ce count to 0. + */ + err_data->ce_count = 0; return 0; }; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c index 93f826a7d3f0..b1ad9e52b234 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c @@ -370,6 +370,33 @@ static void sdma_v5_2_ring_emit_ib(struct amdgpu_ring *ring, } /** + * sdma_v5_2_ring_emit_mem_sync - flush the IB by graphics cache rinse + * + * @ring: amdgpu ring pointer + * @job: job to retrieve vmid from + * @ib: IB object to schedule + * + * flush the IB by graphics cache rinse. + */ +static void sdma_v5_2_ring_emit_mem_sync(struct amdgpu_ring *ring) +{ + uint32_t gcr_cntl = + SDMA_GCR_GL2_INV | SDMA_GCR_GL2_WB | SDMA_GCR_GLM_INV | + SDMA_GCR_GL1_INV | SDMA_GCR_GLV_INV | SDMA_GCR_GLK_INV | + SDMA_GCR_GLI_INV(1); + + /* flush entire cache L0/L1/L2, this can be optimized by performance requirement */ + amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_GCR_REQ)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD1_BASE_VA_31_7(0)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD2_GCR_CONTROL_15_0(gcr_cntl) | + SDMA_PKT_GCR_REQ_PAYLOAD2_BASE_VA_47_32(0)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD3_LIMIT_VA_31_7(0) | + SDMA_PKT_GCR_REQ_PAYLOAD3_GCR_CONTROL_18_16(gcr_cntl >> 16)); + amdgpu_ring_write(ring, SDMA_PKT_GCR_REQ_PAYLOAD4_LIMIT_VA_47_32(0) | + SDMA_PKT_GCR_REQ_PAYLOAD4_VMID(0)); +} + +/** * sdma_v5_2_ring_emit_hdp_flush - emit an hdp flush on the DMA ring * * @ring: amdgpu ring pointer @@ -1663,6 +1690,7 @@ static const struct amdgpu_ring_funcs sdma_v5_2_ring_funcs = { 10 + 10 + 10, /* sdma_v5_2_ring_emit_fence x3 for user fence, vm fence */ .emit_ib_size = 7 + 6, /* sdma_v5_2_ring_emit_ib */ .emit_ib = sdma_v5_2_ring_emit_ib, + .emit_mem_sync = sdma_v5_2_ring_emit_mem_sync, .emit_fence = sdma_v5_2_ring_emit_fence, .emit_pipeline_sync = sdma_v5_2_ring_emit_pipeline_sync, .emit_vm_flush = sdma_v5_2_ring_emit_vm_flush, diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 5c5eb3aed1b3..d80e12b80c7e 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -816,11 +816,12 @@ static void soc15_pcie_gen3_enable(struct amdgpu_device *adev) static void soc15_program_aspm(struct amdgpu_device *adev) { - - if (amdgpu_aspm == 0) + if (amdgpu_aspm != 1) return; - /* todo */ + if (!(adev->flags & AMD_IS_APU) && + (adev->nbio.funcs->program_aspm)) + adev->nbio.funcs->program_aspm(adev); } static void soc15_enable_doorbell_aperture(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c index 3f15bf34123a..cf165ab5dd26 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c @@ -589,6 +589,10 @@ static void vcn_v3_0_mc_resume_dpg_mode(struct amdgpu_device *adev, int inst_idx WREG32_SOC15_DPG_MODE(inst_idx, SOC15_DPG_MODE_OFFSET( VCN, inst_idx, mmUVD_VCPU_NONCACHE_SIZE0), AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_fw_shared)), 0, indirect); + + /* VCN global tiling registers */ + WREG32_SOC15_DPG_MODE(0, SOC15_DPG_MODE_OFFSET( + UVD, 0, mmUVD_GFX10_ADDR_CONFIG), adev->gfx.config.gb_addr_config, 0, indirect); } static void vcn_v3_0_disable_static_power_gating(struct amdgpu_device *adev, int inst) diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index ea338de5818a..735ebbd1148f 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -81,6 +81,30 @@ #include "mxgpu_vi.h" #include "amdgpu_dm.h" +#define ixPCIE_LC_L1_PM_SUBSTATE 0x100100C6 +#define PCIE_LC_L1_PM_SUBSTATE__LC_L1_SUBSTATES_OVERRIDE_EN_MASK 0x00000001L +#define PCIE_LC_L1_PM_SUBSTATE__LC_PCI_PM_L1_2_OVERRIDE_MASK 0x00000002L +#define PCIE_LC_L1_PM_SUBSTATE__LC_PCI_PM_L1_1_OVERRIDE_MASK 0x00000004L +#define PCIE_LC_L1_PM_SUBSTATE__LC_ASPM_L1_2_OVERRIDE_MASK 0x00000008L +#define PCIE_LC_L1_PM_SUBSTATE__LC_ASPM_L1_1_OVERRIDE_MASK 0x00000010L +#define ixPCIE_L1_PM_SUB_CNTL 0x378 +#define PCIE_L1_PM_SUB_CNTL__ASPM_L1_2_EN_MASK 0x00000004L +#define PCIE_L1_PM_SUB_CNTL__ASPM_L1_1_EN_MASK 0x00000008L +#define PCIE_L1_PM_SUB_CNTL__PCI_PM_L1_2_EN_MASK 0x00000001L +#define PCIE_L1_PM_SUB_CNTL__PCI_PM_L1_1_EN_MASK 0x00000002L +#define PCIE_LC_CNTL6__LC_L1_POWERDOWN_MASK 0x00200000L +#define LINK_CAP 0x64 +#define PCIE_LINK_CAP__CLOCK_POWER_MANAGEMENT_MASK 0x00040000L +#define ixCPM_CONTROL 0x1400118 +#define ixPCIE_LC_CNTL7 0x100100BC +#define PCIE_LC_CNTL7__LC_L1_SIDEBAND_CLKREQ_PDWN_EN_MASK 0x00000400L +#define PCIE_LC_CNTL__LC_L0S_INACTIVITY_DEFAULT 0x00000007 +#define PCIE_LC_CNTL__LC_L1_INACTIVITY_DEFAULT 0x00000009 +#define CPM_CONTROL__CLKREQb_UNGATE_TXCLK_ENABLE_MASK 0x01000000L +#define PCIE_L1_PM_SUB_CNTL 0x378 +#define ASIC_IS_P22(asic_type, rid) ((asic_type >= CHIP_POLARIS10) && \ + (asic_type <= CHIP_POLARIS12) && \ + (rid >= 0x6E)) /* Topaz */ static const struct amdgpu_video_codecs topaz_video_codecs_encode = { @@ -1091,13 +1115,178 @@ static void vi_pcie_gen3_enable(struct amdgpu_device *adev) /* todo */ } +static void vi_enable_aspm(struct amdgpu_device *adev) +{ + u32 data, orig; + + orig = data = RREG32_PCIE(ixPCIE_LC_CNTL); + data |= PCIE_LC_CNTL__LC_L0S_INACTIVITY_DEFAULT << + PCIE_LC_CNTL__LC_L0S_INACTIVITY__SHIFT; + data |= PCIE_LC_CNTL__LC_L1_INACTIVITY_DEFAULT << + PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT; + data &= ~PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK; + data |= PCIE_LC_CNTL__LC_DELAY_L1_EXIT_MASK; + if (orig != data) + WREG32_PCIE(ixPCIE_LC_CNTL, data); +} + static void vi_program_aspm(struct amdgpu_device *adev) { + u32 data, data1, orig; + bool bL1SS = false; + bool bClkReqSupport = true; - if (amdgpu_aspm == 0) + if (amdgpu_aspm != 1) return; - /* todo */ + if (adev->flags & AMD_IS_APU || + adev->asic_type < CHIP_POLARIS10) + return; + + orig = data = RREG32_PCIE(ixPCIE_LC_CNTL); + data &= ~PCIE_LC_CNTL__LC_L1_INACTIVITY_MASK; + data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK; + data |= PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK; + if (orig != data) + WREG32_PCIE(ixPCIE_LC_CNTL, data); + + orig = data = RREG32_PCIE(ixPCIE_LC_N_FTS_CNTL); + data &= ~PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS_MASK; + data |= 0x0024 << PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS__SHIFT; + data |= PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS_OVERRIDE_EN_MASK; + if (orig != data) + WREG32_PCIE(ixPCIE_LC_N_FTS_CNTL, data); + + orig = data = RREG32_PCIE(ixPCIE_LC_CNTL3); + data |= PCIE_LC_CNTL3__LC_GO_TO_RECOVERY_MASK; + if (orig != data) + WREG32_PCIE(ixPCIE_LC_CNTL3, data); + + orig = data = RREG32_PCIE(ixPCIE_P_CNTL); + data |= PCIE_P_CNTL__P_IGNORE_EDB_ERR_MASK; + if (orig != data) + WREG32_PCIE(ixPCIE_P_CNTL, data); + + data = RREG32_PCIE(ixPCIE_LC_L1_PM_SUBSTATE); + pci_read_config_dword(adev->pdev, PCIE_L1_PM_SUB_CNTL, &data1); + if (data & PCIE_LC_L1_PM_SUBSTATE__LC_L1_SUBSTATES_OVERRIDE_EN_MASK && + (data & (PCIE_LC_L1_PM_SUBSTATE__LC_PCI_PM_L1_2_OVERRIDE_MASK | + PCIE_LC_L1_PM_SUBSTATE__LC_PCI_PM_L1_1_OVERRIDE_MASK | + PCIE_LC_L1_PM_SUBSTATE__LC_ASPM_L1_2_OVERRIDE_MASK | + PCIE_LC_L1_PM_SUBSTATE__LC_ASPM_L1_1_OVERRIDE_MASK))) { + bL1SS = true; + } else if (data1 & (PCIE_L1_PM_SUB_CNTL__ASPM_L1_2_EN_MASK | + PCIE_L1_PM_SUB_CNTL__ASPM_L1_1_EN_MASK | + PCIE_L1_PM_SUB_CNTL__PCI_PM_L1_2_EN_MASK | + PCIE_L1_PM_SUB_CNTL__PCI_PM_L1_1_EN_MASK)) { + bL1SS = true; + } + + orig = data = RREG32_PCIE(ixPCIE_LC_CNTL6); + data |= PCIE_LC_CNTL6__LC_L1_POWERDOWN_MASK; + if (orig != data) + WREG32_PCIE(ixPCIE_LC_CNTL6, data); + + orig = data = RREG32_PCIE(ixPCIE_LC_LINK_WIDTH_CNTL); + data |= PCIE_LC_LINK_WIDTH_CNTL__LC_DYN_LANES_PWR_STATE_MASK; + if (orig != data) + WREG32_PCIE(ixPCIE_LC_LINK_WIDTH_CNTL, data); + + pci_read_config_dword(adev->pdev, LINK_CAP, &data); + if (!(data & PCIE_LINK_CAP__CLOCK_POWER_MANAGEMENT_MASK)) + bClkReqSupport = false; + + if (bClkReqSupport) { + orig = data = RREG32_SMC(ixTHM_CLK_CNTL); + data &= ~(THM_CLK_CNTL__CMON_CLK_SEL_MASK | THM_CLK_CNTL__TMON_CLK_SEL_MASK); + data |= (1 << THM_CLK_CNTL__CMON_CLK_SEL__SHIFT) | + (1 << THM_CLK_CNTL__TMON_CLK_SEL__SHIFT); + if (orig != data) + WREG32_SMC(ixTHM_CLK_CNTL, data); + + orig = data = RREG32_SMC(ixMISC_CLK_CTRL); + data &= ~(MISC_CLK_CTRL__DEEP_SLEEP_CLK_SEL_MASK | + MISC_CLK_CTRL__ZCLK_SEL_MASK | MISC_CLK_CTRL__DFT_SMS_PG_CLK_SEL_MASK); + data |= (1 << MISC_CLK_CTRL__DEEP_SLEEP_CLK_SEL__SHIFT) | + (1 << MISC_CLK_CTRL__ZCLK_SEL__SHIFT); + data |= (0x20 << MISC_CLK_CTRL__DFT_SMS_PG_CLK_SEL__SHIFT); + if (orig != data) + WREG32_SMC(ixMISC_CLK_CTRL, data); + + orig = data = RREG32_SMC(ixCG_CLKPIN_CNTL); + data |= CG_CLKPIN_CNTL__XTALIN_DIVIDE_MASK; + if (orig != data) + WREG32_SMC(ixCG_CLKPIN_CNTL, data); + + orig = data = RREG32_SMC(ixCG_CLKPIN_CNTL_2); + data |= CG_CLKPIN_CNTL_2__ENABLE_XCLK_MASK; + if (orig != data) + WREG32_SMC(ixCG_CLKPIN_CNTL, data); + + orig = data = RREG32_SMC(ixMPLL_BYPASSCLK_SEL); + data &= ~MPLL_BYPASSCLK_SEL__MPLL_CLKOUT_SEL_MASK; + data |= (4 << MPLL_BYPASSCLK_SEL__MPLL_CLKOUT_SEL__SHIFT); + if (orig != data) + WREG32_SMC(ixMPLL_BYPASSCLK_SEL, data); + + orig = data = RREG32_PCIE(ixCPM_CONTROL); + data |= (CPM_CONTROL__REFCLK_XSTCLK_ENABLE_MASK | + CPM_CONTROL__CLKREQb_UNGATE_TXCLK_ENABLE_MASK); + if (orig != data) + WREG32_PCIE(ixCPM_CONTROL, data); + + orig = data = RREG32_PCIE(ixPCIE_CONFIG_CNTL); + data &= ~PCIE_CONFIG_CNTL__DYN_CLK_LATENCY_MASK; + data |= (0xE << PCIE_CONFIG_CNTL__DYN_CLK_LATENCY__SHIFT); + if (orig != data) + WREG32_PCIE(ixPCIE_CONFIG_CNTL, data); + + orig = data = RREG32(mmBIF_CLK_CTRL); + data |= BIF_CLK_CTRL__BIF_XSTCLK_READY_MASK; + if (orig != data) + WREG32(mmBIF_CLK_CTRL, data); + + orig = data = RREG32_PCIE(ixPCIE_LC_CNTL7); + data |= PCIE_LC_CNTL7__LC_L1_SIDEBAND_CLKREQ_PDWN_EN_MASK; + if (orig != data) + WREG32_PCIE(ixPCIE_LC_CNTL7, data); + + orig = data = RREG32_PCIE(ixPCIE_HW_DEBUG); + data |= PCIE_HW_DEBUG__HW_01_DEBUG_MASK; + if (orig != data) + WREG32_PCIE(ixPCIE_HW_DEBUG, data); + + orig = data = RREG32_PCIE(ixPCIE_LC_CNTL2); + data |= PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L23_MASK; + data |= PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L1_MASK; + if (bL1SS) + data &= ~PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L1_MASK; + if (orig != data) + WREG32_PCIE(ixPCIE_LC_CNTL2, data); + + } + + vi_enable_aspm(adev); + + data = RREG32_PCIE(ixPCIE_LC_N_FTS_CNTL); + data1 = RREG32_PCIE(ixPCIE_LC_STATUS1); + if (((data & PCIE_LC_N_FTS_CNTL__LC_N_FTS_MASK) == PCIE_LC_N_FTS_CNTL__LC_N_FTS_MASK) && + data1 & PCIE_LC_STATUS1__LC_REVERSE_XMIT_MASK && + data1 & PCIE_LC_STATUS1__LC_REVERSE_RCVR_MASK) { + orig = data = RREG32_PCIE(ixPCIE_LC_CNTL); + data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK; + if (orig != data) + WREG32_PCIE(ixPCIE_LC_CNTL, data); + } + + if ((adev->asic_type == CHIP_POLARIS12 && + !(ASICID_IS_P23(adev->pdev->device, adev->pdev->revision))) || + ASIC_IS_P22(adev->asic_type, adev->external_rev_id)) { + orig = data = RREG32_PCIE(ixPCIE_LC_TRAINING_CNTL); + data &= ~PCIE_LC_TRAINING_CNTL__LC_DISABLE_TRAINING_BIT_ARCH_MASK; + if (orig != data) + WREG32_PCIE(ixPCIE_LC_TRAINING_CNTL, data); + } } static void vi_enable_doorbell_aperture(struct amdgpu_device *adev, |