Merge tag 'amd-drm-next-6.10-2024-04-19' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-6.10-2024-04-19:

amdgpu:
- DC resource allocation logic updates
- DC IPS fixes
- DC YUV fixes
- DMCUB fixes
- DML2 fixes
- Devcoredump updates
- USB-C DSC fix
- Misc display code cleanups
- PSR fixes
- MES timeout fix
- RAS updates
- UAF fix in VA IOCTL
- Fix visible VRAM handling during faults
- Fix IP discovery handling during PCI rescans
- Misc code cleanups
- PSP 14 updates
- More runtime PM code rework
- SMU 14.0.2 support
- GPUVM page fault redirection to secondary IH rings for IH 6.x
- Suspend/resume fixes
- SR-IOV fixes

amdkfd:
- Fix eviction fence handling
- Fix leak in GPU memory allocation failure case
- DMABuf import handling fix

radeon:
- Silence UBSAN warnings related to flexible arrays

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240419224332.2938259-1-alexander.deucher@amd.com

Author: airlied

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -1409,6 +1409,7 @@ bool amdgpu_device_supports_px(struct drm_device *dev);
 bool amdgpu_device_supports_boco(struct drm_device *dev);
 bool amdgpu_device_supports_smart_shift(struct drm_device *dev);
 int amdgpu_device_supports_baco(struct drm_device *dev);
+void amdgpu_device_detect_runtime_pm_mode(struct amdgpu_device *adev);
 bool amdgpu_device_is_peer_accessible(struct amdgpu_device *adev,
 				      struct amdgpu_device *peer_adev);
 int amdgpu_device_baco_enter(struct drm_device *dev);

drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c

@@ -753,23 +753,13 @@ int aca_bank_info_decode(struct aca_bank *bank, struct aca_bank_info *info)
 
 static int aca_bank_get_error_code(struct amdgpu_device *adev, struct aca_bank *bank)
 {
-	int error_code;
-
-	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
-	case IP_VERSION(13, 0, 6):
-		if (!(adev->flags & AMD_IS_APU) && adev->pm.fw_version >= 0x00555600) {
-			error_code = ACA_REG__SYND__ERRORINFORMATION(bank->regs[ACA_REG_IDX_SYND]);
-			return error_code & 0xff;
-		}
-		break;
-	default:
-		break;
-	}
+	struct amdgpu_aca *aca = &adev->aca;
+	const struct aca_smu_funcs *smu_funcs = aca->smu_funcs;
 
-	/* NOTE: the true error code is encoded in status.errorcode[0:7] */
-	error_code = ACA_REG__STATUS__ERRORCODE(bank->regs[ACA_REG_IDX_STATUS]);
+	if (!smu_funcs || !smu_funcs->parse_error_code)
+		return -EOPNOTSUPP;
 
-	return error_code & 0xff;
+	return smu_funcs->parse_error_code(adev, bank);
 }
 
 int aca_bank_check_error_codes(struct amdgpu_device *adev, struct aca_bank *bank, int *err_codes, int size)
@@ -780,6 +770,9 @@ int aca_bank_check_error_codes(struct amdgpu_device *adev, struct aca_bank *bank
 		return -EINVAL;
 
 	error_code = aca_bank_get_error_code(adev, bank);
+	if (error_code < 0)
+		return error_code;
+
 	for (i = 0; i < size; i++) {
 		if (err_codes[i] == error_code)
 			return 0;

drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h

@@ -173,6 +173,7 @@ struct aca_smu_funcs {
 	int (*set_debug_mode)(struct amdgpu_device *adev, bool enable);
 	int (*get_valid_aca_count)(struct amdgpu_device *adev, enum aca_smu_type type, u32 *count);
 	int (*get_valid_aca_bank)(struct amdgpu_device *adev, enum aca_smu_type type, int idx, struct aca_bank *bank);
+	int (*parse_error_code)(struct amdgpu_device *adev, struct aca_bank *bank);
 };
 
 struct amdgpu_aca {

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c

@@ -1854,6 +1854,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 err_bo_create:
 	amdgpu_amdkfd_unreserve_mem_limit(adev, aligned_size, flags, xcp_id);
 err_reserve_limit:
+	amdgpu_sync_free(&(*mem)->sync);
 	mutex_destroy(&(*mem)->lock);
 	if (gobj)
 		drm_gem_object_put(gobj);
@@ -2900,13 +2901,12 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu *
 
 	amdgpu_sync_create(&sync_obj);
 
-	/* Validate BOs and map them to GPUVM (update VM page tables). */
+	/* Validate BOs managed by KFD */
 	list_for_each_entry(mem, &process_info->kfd_bo_list,
 			    validate_list) {
 
 		struct amdgpu_bo *bo = mem->bo;
 		uint32_t domain = mem->domain;
-		struct kfd_mem_attachment *attachment;
 		struct dma_resv_iter cursor;
 		struct dma_fence *fence;
 
@@ -2931,6 +2931,25 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu *
 				goto validate_map_fail;
 			}
 		}
+	}
+
+	if (failed_size)
+		pr_debug("0x%lx/0x%lx in system\n", failed_size, total_size);
+
+	/* Validate PDs, PTs and evicted DMABuf imports last. Otherwise BO
+	 * validations above would invalidate DMABuf imports again.
+	 */
+	ret = process_validate_vms(process_info, &exec.ticket);
+	if (ret) {
+		pr_debug("Validating VMs failed, ret: %d\n", ret);
+		goto validate_map_fail;
+	}
+
+	/* Update mappings managed by KFD. */
+	list_for_each_entry(mem, &process_info->kfd_bo_list,
+			    validate_list) {
+		struct kfd_mem_attachment *attachment;
+
 		list_for_each_entry(attachment, &mem->attachments, list) {
 			if (!attachment->is_mapped)
 				continue;
@@ -2947,18 +2966,6 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu *
 		}
 	}
 
-	if (failed_size)
-		pr_debug("0x%lx/0x%lx in system\n", failed_size, total_size);
-
-	/* Validate PDs, PTs and evicted DMABuf imports last. Otherwise BO
-	 * validations above would invalidate DMABuf imports again.
-	 */
-	ret = process_validate_vms(process_info, &exec.ticket);
-	if (ret) {
-		pr_debug("Validating VMs failed, ret: %d\n", ret);
-		goto validate_map_fail;
-	}
-
 	/* Update mappings not managed by KFD */
 	list_for_each_entry(peer_vm, &process_info->vm_list_head,
 			vm_list_node) {

drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c

@@ -819,7 +819,7 @@ static int amdgpu_cs_bo_validate(void *param, struct amdgpu_bo *bo)
 
 	p->bytes_moved += ctx.bytes_moved;
 	if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
-	    amdgpu_bo_in_cpu_visible_vram(bo))
+	    amdgpu_res_cpu_visible(adev, bo->tbo.resource))
 		p->bytes_moved_vis += ctx.bytes_moved;
 
 	if (unlikely(r == -ENOMEM) && domain != bo->allowed_domains) {

drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c

@@ -188,10 +188,11 @@ static void amdgpu_devcoredump_fw_info(struct amdgpu_device *adev,
 		   adev->vpe.feature_version, adev->vpe.fw_version);
 
 	drm_printf(p, "\nVBIOS Information\n");
-	drm_printf(p, "name: %s\n", ctx->name);
-	drm_printf(p, "pn %s\n", ctx->vbios_pn);
-	drm_printf(p, "version: %s\n", ctx->vbios_ver_str);
-	drm_printf(p, "date: %s\n", ctx->date);
+	drm_printf(p, "vbios name       : %s\n", ctx->name);
+	drm_printf(p, "vbios pn         : %s\n", ctx->vbios_pn);
+	drm_printf(p, "vbios version    : %d\n", ctx->version);
+	drm_printf(p, "vbios ver_str    : %s\n", ctx->vbios_ver_str);
+	drm_printf(p, "vbios date       : %s\n", ctx->date);
 }
 
 static ssize_t

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

@@ -350,6 +350,81 @@ int amdgpu_device_supports_baco(struct drm_device *dev)
 	return amdgpu_asic_supports_baco(adev);
 }
 
+void amdgpu_device_detect_runtime_pm_mode(struct amdgpu_device *adev)
+{
+	struct drm_device *dev;
+	int bamaco_support;
+
+	dev = adev_to_drm(adev);
+
+	adev->pm.rpm_mode = AMDGPU_RUNPM_NONE;
+	bamaco_support = amdgpu_device_supports_baco(dev);
+
+	switch (amdgpu_runtime_pm) {
+	case 2:
+		if (bamaco_support & MACO_SUPPORT) {
+			adev->pm.rpm_mode = AMDGPU_RUNPM_BAMACO;
+			dev_info(adev->dev, "Forcing BAMACO for runtime pm\n");
+		} else if (bamaco_support == BACO_SUPPORT) {
+			adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
+			dev_info(adev->dev, "Requested mode BAMACO not available,fallback to use BACO\n");
+		}
+		break;
+	case 1:
+		if (bamaco_support & BACO_SUPPORT) {
+			adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
+			dev_info(adev->dev, "Forcing BACO for runtime pm\n");
+		}
+		break;
+	case -1:
+	case -2:
+		if (amdgpu_device_supports_px(dev)) { /* enable PX as runtime mode */
+			adev->pm.rpm_mode = AMDGPU_RUNPM_PX;
+			dev_info(adev->dev, "Using ATPX for runtime pm\n");
+		} else if (amdgpu_device_supports_boco(dev)) { /* enable boco as runtime mode */
+			adev->pm.rpm_mode = AMDGPU_RUNPM_BOCO;
+			dev_info(adev->dev, "Using BOCO for runtime pm\n");
+		} else {
+			if (!bamaco_support)
+				goto no_runtime_pm;
+
+			switch (adev->asic_type) {
+			case CHIP_VEGA20:
+			case CHIP_ARCTURUS:
+				/* BACO are not supported on vega20 and arctrus */
+				break;
+			case CHIP_VEGA10:
+				/* enable BACO as runpm mode if noretry=0 */
+				if (!adev->gmc.noretry)
+					adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
+				break;
+			default:
+				/* enable BACO as runpm mode on CI+ */
+				adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
+				break;
+			}
+
+			if (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) {
+				if (bamaco_support & MACO_SUPPORT) {
+					adev->pm.rpm_mode = AMDGPU_RUNPM_BAMACO;
+					dev_info(adev->dev, "Using BAMACO for runtime pm\n");
+				} else {
+					dev_info(adev->dev, "Using BACO for runtime pm\n");
+				}
+			}
+		}
+		break;
+	case 0:
+		dev_info(adev->dev, "runtime pm is manually disabled\n");
+		break;
+	default:
+		break;
+	}
+
+no_runtime_pm:
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_NONE)
+		dev_info(adev->dev, "Runtime PM not available\n");
+}
 /**
  * amdgpu_device_supports_smart_shift - Is the device dGPU with
  * smart shift support
@@ -1460,7 +1535,7 @@ int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev)
 
 	/* PCI_EXT_CAP_ID_VNDR extended capability is located at 0x100 */
 	if (!pci_find_ext_capability(adev->pdev, PCI_EXT_CAP_ID_VNDR))
-		DRM_WARN("System can't access extended configuration space,please check!!\n");
+		DRM_WARN("System can't access extended configuration space, please check!!\n");
 
 	/* skip if the bios has already enabled large BAR */
 	if (adev->gmc.real_vram_size &&
@@ -5282,7 +5357,9 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 	/* Try reset handler method first */
 	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
 				    reset_list);
-	amdgpu_reset_reg_dumps(tmp_adev);
+
+	if (!test_bit(AMDGPU_SKIP_COREDUMP, &reset_context->flags))
+		amdgpu_reset_reg_dumps(tmp_adev);
 
 	reset_context->reset_device_list = device_list_handle;
 	r = amdgpu_reset_perform_reset(tmp_adev, reset_context);
@@ -5355,7 +5432,8 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 
 				vram_lost = amdgpu_device_check_vram_lost(tmp_adev);
 
-				amdgpu_coredump(tmp_adev, vram_lost, reset_context);
+				if (!test_bit(AMDGPU_SKIP_COREDUMP, &reset_context->flags))
+					amdgpu_coredump(tmp_adev, vram_lost, reset_context);
 
 				if (vram_lost) {
 					DRM_INFO("VRAM is lost due to GPU reset!\n");

drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c

@@ -255,7 +255,6 @@ static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev,
 	uint64_t vram_size;
 	u32 msg;
 	int i, ret = 0;
-	int ip_discovery_ver = 0;
 
 	/* It can take up to a second for IFWI init to complete on some dGPUs,
 	 * but generally it should be in the 60-100ms range.  Normally this starts
@@ -265,17 +264,13 @@ static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev,
 	 * continue.
 	 */
 
-	ip_discovery_ver = RREG32(mmIP_DISCOVERY_VERSION);
-	if ((dev_is_removable(&adev->pdev->dev)) ||
-	    (ip_discovery_ver == IP_DISCOVERY_V2) ||
-	    (ip_discovery_ver == IP_DISCOVERY_V4)) {
-		for (i = 0; i < 1000; i++) {
-			msg = RREG32(mmMP0_SMN_C2PMSG_33);
-			if (msg & 0x80000000)
-				break;
-			msleep(1);
-		}
+	for (i = 0; i < 1000; i++) {
+		msg = RREG32(mmMP0_SMN_C2PMSG_33);
+		if (msg & 0x80000000)
+			break;
+		usleep_range(1000, 1100);
 	}
+
 	vram_size = (uint64_t)RREG32(mmRCC_CONFIG_MEMSIZE) << 20;
 
 	if (vram_size) {
@@ -1906,6 +1901,8 @@ static int amdgpu_discovery_set_smu_ip_blocks(struct amdgpu_device *adev)
 		break;
 	case IP_VERSION(14, 0, 0):
 	case IP_VERSION(14, 0, 1):
+	case IP_VERSION(14, 0, 2):
+	case IP_VERSION(14, 0, 3):
 		amdgpu_device_ip_block_add(adev, &smu_v14_0_ip_block);
 		break;
 	default:

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

@@ -2481,6 +2481,7 @@ static void amdgpu_drv_delayed_reset_work_handler(struct work_struct *work)
 
 	/* Use a common context, just need to make sure full reset is done */
 	set_bit(AMDGPU_SKIP_HW_RESET, &reset_context.flags);
+	set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);
 	r = amdgpu_do_asic_reset(&device_list, &reset_context);
 
 	if (r) {

drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c

@@ -133,7 +133,6 @@ void amdgpu_register_gpu_instance(struct amdgpu_device *adev)
 int amdgpu_driver_load_kms(struct amdgpu_device *adev, unsigned long flags)
 {
 	struct drm_device *dev;
-	int bamaco_support = 0;
 	int r, acpi_status;
 
 	dev = adev_to_drm(adev);
@@ -150,52 +149,7 @@ int amdgpu_driver_load_kms(struct amdgpu_device *adev, unsigned long flags)
 		goto out;
 	}
 
-	adev->pm.rpm_mode = AMDGPU_RUNPM_NONE;
-	if (amdgpu_device_supports_px(dev) &&
-	    (amdgpu_runtime_pm != 0)) { /* enable PX as runtime mode */
-		adev->pm.rpm_mode = AMDGPU_RUNPM_PX;
-		dev_info(adev->dev, "Using ATPX for runtime pm\n");
-	} else if (amdgpu_device_supports_boco(dev) &&
-		   (amdgpu_runtime_pm != 0)) { /* enable boco as runtime mode */
-		adev->pm.rpm_mode = AMDGPU_RUNPM_BOCO;
-		dev_info(adev->dev, "Using BOCO for runtime pm\n");
-	} else if (amdgpu_runtime_pm != 0) {
-		bamaco_support = amdgpu_device_supports_baco(dev);
-
-		if (!bamaco_support)
-			goto no_runtime_pm;
-
-		switch (adev->asic_type) {
-		case CHIP_VEGA20:
-		case CHIP_ARCTURUS:
-			/* enable BACO as runpm mode if runpm=1 */
-			if (amdgpu_runtime_pm > 0)
-				adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
-			break;
-		case CHIP_VEGA10:
-			/* enable BACO as runpm mode if noretry=0 */
-			if (!adev->gmc.noretry)
-				adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
-			break;
-		default:
-			/* enable BACO as runpm mode on CI+ */
-			adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
-			break;
-		}
-
-		if (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) {
-			if (bamaco_support & MACO_SUPPORT) {
-				adev->pm.rpm_mode = AMDGPU_RUNPM_BAMACO;
-				dev_info(adev->dev, "Using BAMACO for runtime pm\n");
-			} else {
-				dev_info(adev->dev, "Using BACO for runtime pm\n");
-			}
-		}
-	}
-
-no_runtime_pm:
-	if (adev->pm.rpm_mode == AMDGPU_RUNPM_NONE)
-		dev_info(adev->dev, "NO pm mode for runtime pm\n");
+	amdgpu_device_detect_runtime_pm_mode(adev);
 
 	/* Call ACPI methods: require modeset init
 	 * but failure is not fatal