Merge tag 'amd-drm-fixes-6.8-2024-01-18' of https://gitlab.freedeskto…

…p.org/agd5f/linux into drm-next amd-drm-fixes-6.8-2024-01-18: amdgpu: - DSC fixes - DC resource pool fixes - OTG fix - DML2 fixes - Aux fix - GFX10 RLC firmware handling fix - Revert a broken workaround for SMU 13.0.2 - DC writeback fix - Enable gfxoff when ROCm apps are active on gfx11 with the proper FW version amdkfd: - Fix dma-buf exports using GEM handles Signed-off-by: Dave Airlie <airlied@redhat.com> From: Alex Deucher <alexander.deucher@amd.com> Link: https://patchwork.freedesktop.org/patch/msgid/20240118223127.4904-1-alexander.deucher@amd.com
embeddedTS · Jan 19, 2024 · 0a1123c · 0a1123c
2 parents 205e18c + aa0901a
commit 0a1123c
Show file tree

Hide file tree

Showing 67 changed files with 554 additions and 373 deletions.
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -254,8 +254,6 @@ extern int amdgpu_agp;
 
 extern int amdgpu_wbrf;
 
-extern int fw_bo_location;
-
 #define AMDGPU_VM_MAX_NUM_CTX			4096
 #define AMDGPU_SG_THRESHOLD			(256*1024*1024)
 #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS	        3000
@@ -1146,6 +1144,7 @@ struct amdgpu_device {
 	bool                            debug_vm;
 	bool                            debug_largebar;
 	bool                            debug_disable_soft_recovery;
+	bool                            debug_use_vram_fw_buf;
 };
 
 static inline uint32_t amdgpu_ip_version(const struct amdgpu_device *adev,

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -138,6 +138,9 @@ static void amdgpu_amdkfd_reset_work(struct work_struct *work)
 	amdgpu_device_gpu_recover(adev, NULL, &reset_context);
 }
 
+static const struct drm_client_funcs kfd_client_funcs = {
+	.unregister	= drm_client_release,
+};
 void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
 {
 	int i;
@@ -161,7 +164,7 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
 			.enable_mes = adev->enable_mes,
 		};
 
-		ret = drm_client_init(&adev->ddev, &adev->kfd.client, "kfd", NULL);
+		ret = drm_client_init(&adev->ddev, &adev->kfd.client, "kfd", &kfd_client_funcs);
 		if (ret) {
 			dev_err(adev->dev, "Failed to init DRM client: %d\n", ret);
 			return;
@@ -695,10 +698,8 @@ int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
 void amdgpu_amdkfd_set_compute_idle(struct amdgpu_device *adev, bool idle)
 {
 	enum amd_powergating_state state = idle ? AMD_PG_STATE_GATE : AMD_PG_STATE_UNGATE;
-	/* Temporary workaround to fix issues observed in some
-	 * compute applications when GFXOFF is enabled on GFX11.
-	 */
-	if (IP_VERSION_MAJ(amdgpu_ip_version(adev, GC_HWIP, 0)) == 11) {
+	if (IP_VERSION_MAJ(amdgpu_ip_version(adev, GC_HWIP, 0)) == 11 &&
+	    ((adev->mes.kiq_version & AMDGPU_MES_VERSION_MASK) <= 64)) {
 		pr_debug("GFXOFF is %s\n", idle ? "enabled" : "disabled");
 		amdgpu_gfx_off_ctrl(adev, idle);
 	} else if ((IP_VERSION_MAJ(amdgpu_ip_version(adev, GC_HWIP, 0)) == 9) &&

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -311,7 +311,7 @@ void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_mem *mem);
 int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device *adev, struct amdgpu_bo *bo);
 
 int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info,
-					    struct dma_fence **ef);
+					    struct dma_fence __rcu **ef);
 int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev,
 					      struct kfd_vm_fault_info *info);
 int amdgpu_amdkfd_gpuvm_import_dmabuf_fd(struct amdgpu_device *adev, int fd,

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -2802,7 +2802,7 @@ static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work)
 	put_task_struct(usertask);
 }
 
-static void replace_eviction_fence(struct dma_fence **ef,
+static void replace_eviction_fence(struct dma_fence __rcu **ef,
 				   struct dma_fence *new_ef)
 {
 	struct dma_fence *old_ef = rcu_replace_pointer(*ef, new_ef, true
@@ -2837,7 +2837,7 @@ static void replace_eviction_fence(struct dma_fence **ef,
  * 7.  Add fence to all PD and PT BOs.
  * 8.  Unreserve all BOs
  */
-int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
+int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu **ef)
 {
 	struct amdkfd_process_info *process_info = info;
 	struct amdgpu_vm *peer_vm;

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -1544,6 +1544,7 @@ bool amdgpu_device_need_post(struct amdgpu_device *adev)
 				return true;
 
 			fw_ver = *((uint32_t *)adev->pm.fw->data + 69);
+			release_firmware(adev->pm.fw);
 			if (fw_ver < 0x00160e00)
 				return true;
 		}
@@ -5245,7 +5246,6 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 	struct amdgpu_device *tmp_adev = NULL;
 	bool need_full_reset, skip_hw_reset, vram_lost = false;
 	int r = 0;
-	bool gpu_reset_for_dev_remove = 0;
 
 	/* Try reset handler method first */
 	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
@@ -5265,10 +5265,6 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 		test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
 	skip_hw_reset = test_bit(AMDGPU_SKIP_HW_RESET, &reset_context->flags);
 
-	gpu_reset_for_dev_remove =
-		test_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context->flags) &&
-			test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
-
 	/*
 	 * ASIC reset has to be done on all XGMI hive nodes ASAP
 	 * to allow proper links negotiation in FW (within 1 sec)
@@ -5311,18 +5307,6 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 		amdgpu_ras_intr_cleared();
 	}
 
-	/* Since the mode1 reset affects base ip blocks, the
-	 * phase1 ip blocks need to be resumed. Otherwise there
-	 * will be a BIOS signature error and the psp bootloader
-	 * can't load kdb on the next amdgpu install.
-	 */
-	if (gpu_reset_for_dev_remove) {
-		list_for_each_entry(tmp_adev, device_list_handle, reset_list)
-			amdgpu_device_ip_resume_phase1(tmp_adev);
-
-		goto end;
-	}
-
 	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
 		if (need_full_reset) {
 			/* post card */
@@ -5559,11 +5543,6 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	int i, r = 0;
 	bool need_emergency_restart = false;
 	bool audio_suspended = false;
-	bool gpu_reset_for_dev_remove = false;
-
-	gpu_reset_for_dev_remove =
-			test_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context->flags) &&
-				test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
 
 	/*
 	 * Special case: RAS triggered and full reset isn't supported
@@ -5601,7 +5580,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1)) {
 		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
 			list_add_tail(&tmp_adev->reset_list, &device_list);
-			if (gpu_reset_for_dev_remove && adev->shutdown)
+			if (adev->shutdown)
 				tmp_adev->shutdown = true;
 		}
 		if (!list_is_first(&adev->reset_list, &device_list))
@@ -5686,10 +5665,6 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 
 retry:	/* Rest of adevs pre asic reset from XGMI hive. */
 	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
-		if (gpu_reset_for_dev_remove) {
-			/* Workaroud for ASICs need to disable SMC first */
-			amdgpu_device_smu_fini_early(tmp_adev);
-		}
 		r = amdgpu_device_pre_asic_reset(tmp_adev, reset_context);
 		/*TODO Should we stop ?*/
 		if (r) {
@@ -5721,9 +5696,6 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 		r = amdgpu_do_asic_reset(device_list_handle, reset_context);
 		if (r && r == -EAGAIN)
 			goto retry;
-
-		if (!r && gpu_reset_for_dev_remove)
-			goto recover_end;
 	}
 
 skip_hw_reset:
@@ -5779,7 +5751,6 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 		amdgpu_ras_set_error_query_ready(tmp_adev, true);
 	}
 
-recover_end:
 	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
 					    reset_list);
 	amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain);

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c
@@ -1963,8 +1963,6 @@ static int amdgpu_discovery_set_gc_ip_blocks(struct amdgpu_device *adev)
 		amdgpu_device_ip_block_add(adev, &gfx_v9_0_ip_block);
 		break;
 	case IP_VERSION(9, 4, 3):
-		if (!amdgpu_exp_hw_support)
-			return -EINVAL;
 		amdgpu_device_ip_block_add(adev, &gfx_v9_4_3_ip_block);
 		break;
 	case IP_VERSION(10, 1, 10):

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
@@ -128,6 +128,7 @@ enum AMDGPU_DEBUG_MASK {
 	AMDGPU_DEBUG_VM = BIT(0),
 	AMDGPU_DEBUG_LARGEBAR = BIT(1),
 	AMDGPU_DEBUG_DISABLE_GPU_SOFT_RECOVERY = BIT(2),
+	AMDGPU_DEBUG_USE_VRAM_FW_BUF = BIT(3),
 };
 
 unsigned int amdgpu_vram_limit = UINT_MAX;
@@ -210,7 +211,6 @@ int amdgpu_seamless = -1; /* auto */
 uint amdgpu_debug_mask;
 int amdgpu_agp = -1; /* auto */
 int amdgpu_wbrf = -1;
-int fw_bo_location = -1;
 
 static void amdgpu_drv_delayed_reset_work_handler(struct work_struct *work);
 
@@ -990,10 +990,6 @@ MODULE_PARM_DESC(wbrf,
 	"Enable Wifi RFI interference mitigation (0 = disabled, 1 = enabled, -1 = auto(default)");
 module_param_named(wbrf, amdgpu_wbrf, int, 0444);
 
-MODULE_PARM_DESC(fw_bo_location,
-	"location to put firmware bo for frontdoor loading (-1 = auto (default), 0 = on ram, 1 = on vram");
-module_param(fw_bo_location, int, 0644);
-
 /* These devices are not supported by amdgpu.
  * They are supported by the mach64, r128, radeon drivers
  */
@@ -2122,6 +2118,11 @@ static void amdgpu_init_debug_options(struct amdgpu_device *adev)
 		pr_info("debug: soft reset for GPU recovery disabled\n");
 		adev->debug_disable_soft_recovery = true;
 	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_USE_VRAM_FW_BUF) {
+		pr_info("debug: place fw in vram for frontdoor loading\n");
+		adev->debug_use_vram_fw_buf = true;
+	}
 }
 
 static unsigned long amdgpu_fix_asic_type(struct pci_dev *pdev, unsigned long flags)
@@ -2233,6 +2234,8 @@ static int amdgpu_pci_probe(struct pci_dev *pdev,
 
 	pci_set_drvdata(pdev, ddev);
 
+	amdgpu_init_debug_options(adev);
+
 	ret = amdgpu_driver_load_kms(adev, flags);
 	if (ret)
 		goto err_pci;
@@ -2313,8 +2316,6 @@ static int amdgpu_pci_probe(struct pci_dev *pdev,
 			amdgpu_get_secondary_funcs(adev);
 	}
 
-	amdgpu_init_debug_options(adev);
-
 	return 0;
 
 err_pci:
@@ -2336,38 +2337,6 @@ amdgpu_pci_remove(struct pci_dev *pdev)
 		pm_runtime_forbid(dev->dev);
 	}
 
-	if (amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 2) &&
-	    !amdgpu_sriov_vf(adev)) {
-		bool need_to_reset_gpu = false;
-
-		if (adev->gmc.xgmi.num_physical_nodes > 1) {
-			struct amdgpu_hive_info *hive;
-
-			hive = amdgpu_get_xgmi_hive(adev);
-			if (hive->device_remove_count == 0)
-				need_to_reset_gpu = true;
-			hive->device_remove_count++;
-			amdgpu_put_xgmi_hive(hive);
-		} else {
-			need_to_reset_gpu = true;
-		}
-
-		/* Workaround for ASICs need to reset SMU.
-		 * Called only when the first device is removed.
-		 */
-		if (need_to_reset_gpu) {
-			struct amdgpu_reset_context reset_context;
-
-			adev->shutdown = true;
-			memset(&reset_context, 0, sizeof(reset_context));
-			reset_context.method = AMD_RESET_METHOD_NONE;
-			reset_context.reset_req_dev = adev;
-			set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
-			set_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context.flags);
-			amdgpu_device_gpu_recover(adev, NULL, &reset_context);
-		}
-	}
-
 	amdgpu_driver_unload_kms(dev);
 
 	/*

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
@@ -1045,21 +1045,28 @@ int amdgpu_gmc_vram_checking(struct amdgpu_device *adev)
 	 * seconds, so here, we just pick up three parts for emulation.
 	 */
 	ret = memcmp(vram_ptr, cptr, 10);
-	if (ret)
-		return ret;
+	if (ret) {
+		ret = -EIO;
+		goto release_buffer;
+	}
 
 	ret = memcmp(vram_ptr + (size / 2), cptr, 10);
-	if (ret)
-		return ret;
+	if (ret) {
+		ret = -EIO;
+		goto release_buffer;
+	}
 
 	ret = memcmp(vram_ptr + size - 10, cptr, 10);
-	if (ret)
-		return ret;
+	if (ret) {
+		ret = -EIO;
+		goto release_buffer;
+	}
 
+release_buffer:
 	amdgpu_bo_free_kernel(&vram_bo, &vram_gpu,
 			&vram_ptr);
 
-	return 0;
+	return ret;
 }
 
 static ssize_t current_memory_partition_show(

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
@@ -1105,7 +1105,12 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 			if (amdgpu_dpm_read_sensor(adev,
 						   AMDGPU_PP_SENSOR_GPU_AVG_POWER,
 						   (void *)&ui32, &ui32_size)) {
-				return -EINVAL;
+				/* fall back to input power for backwards compat */
+				if (amdgpu_dpm_read_sensor(adev,
+							   AMDGPU_PP_SENSOR_GPU_INPUT_POWER,
+							   (void *)&ui32, &ui32_size)) {
+					return -EINVAL;
+				}
 			}
 			ui32 >>= 8;
 			break;

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
@@ -466,7 +466,7 @@ static int psp_sw_init(void *handle)
 	}
 
 	ret = amdgpu_bo_create_kernel(adev, PSP_1_MEG, PSP_1_MEG,
-				      (amdgpu_sriov_vf(adev) || fw_bo_location == 1) ?
+				      (amdgpu_sriov_vf(adev) || adev->debug_use_vram_fw_buf) ?
 				      AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
 				      &psp->fw_pri_bo,
 				      &psp->fw_pri_mc_addr,