Merge tag 'amd-drm-next-6.8-2024-01-05' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-6.8-2024-01-05:

amdgpu:
- VRR fixes
- PSR-SU fixes
- SubVP fixes
- DCN 3.5 fixes
- Documentation updates
- DMCUB fixes
- DML2 fixes
- UMC 12.0 updates
- GPUVM fix
- Misc code cleanups and whitespace cleanups
- DP MST fix
- Let KFD sync with GPUVM fences
- GFX11 reset fix
- SMU 13.0.6 fixes
- VSC fix for DP/eDP
- Navi12 display fix
- RN/CZN system aperture fix
- DCN 2.1 bandwidth validation fix
- DCN INIT cleanup

amdkfd:
- SVM fixes
- Revert TBA/TMA location change

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

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240105220522.4976-1-alexander.deucher@amd.com

5 files changed, 87 insertions, 100 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c
index 62b205dac63a..6604a3f99c5e 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c
@@ -330,12 +330,6 @@ static void kfd_init_apertures_vi(struct kfd_process_device *pdd, uint8_t id)
 	pdd->gpuvm_limit =
 		pdd->dev->kfd->shared_resources.gpuvm_size - 1;
 
-	/* dGPUs: the reserved space for kernel
-	 * before SVM
-	 */
-	pdd->qpd.cwsr_base = SVM_CWSR_BASE;
-	pdd->qpd.ib_base = SVM_IB_BASE;
-
 	pdd->scratch_base = MAKE_SCRATCH_APP_BASE_VI();
 	pdd->scratch_limit = MAKE_SCRATCH_APP_LIMIT(pdd->scratch_base);
 }
@@ -345,18 +339,18 @@ static void kfd_init_apertures_v9(struct kfd_process_device *pdd, uint8_t id)
 	pdd->lds_base = MAKE_LDS_APP_BASE_V9();
 	pdd->lds_limit = MAKE_LDS_APP_LIMIT(pdd->lds_base);
 
-	pdd->gpuvm_base = PAGE_SIZE;
+        /* Raven needs SVM to support graphic handle, etc. Leave the small
+         * reserved space before SVM on Raven as well, even though we don't
+         * have to.
+         * Set gpuvm_base and gpuvm_limit to CANONICAL addresses so that they
+         * are used in Thunk to reserve SVM.
+         */
+        pdd->gpuvm_base = SVM_USER_BASE;
 	pdd->gpuvm_limit =
 		pdd->dev->kfd->shared_resources.gpuvm_size - 1;
 
 	pdd->scratch_base = MAKE_SCRATCH_APP_BASE_V9();
 	pdd->scratch_limit = MAKE_SCRATCH_APP_LIMIT(pdd->scratch_base);
-
-	/*
-	 * Place TBA/TMA on opposite side of VM hole to prevent
-	 * stray faults from triggering SVM on these pages.
-	 */
-	pdd->qpd.cwsr_base = pdd->dev->kfd->shared_resources.gpuvm_size;
 }
 
 int kfd_init_apertures(struct kfd_process *process)
@@ -413,6 +407,12 @@ int kfd_init_apertures(struct kfd_process *process)
 					return -EINVAL;
 				}
 			}
+
+                        /* dGPUs: the reserved space for kernel
+                         * before SVM
+                         */
+                        pdd->qpd.cwsr_base = SVM_CWSR_BASE;
+                        pdd->qpd.ib_base = SVM_IB_BASE;
 		}
 
 		dev_dbg(kfd_device, "node id %u\n", id);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
index b854cbf06dce..d630100b9e91 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
@@ -260,19 +260,6 @@ static void svm_migrate_put_sys_page(unsigned long addr)
 	put_page(page);
 }
 
-static unsigned long svm_migrate_successful_pages(struct migrate_vma *migrate)
-{
-	unsigned long cpages = 0;
-	unsigned long i;
-
-	for (i = 0; i < migrate->npages; i++) {
-		if (migrate->src[i] & MIGRATE_PFN_VALID &&
-		    migrate->src[i] & MIGRATE_PFN_MIGRATE)
-			cpages++;
-	}
-	return cpages;
-}
-
 static unsigned long svm_migrate_unsuccessful_pages(struct migrate_vma *migrate)
 {
 	unsigned long upages = 0;
@@ -402,6 +389,7 @@ svm_migrate_vma_to_vram(struct kfd_node *node, struct svm_range *prange,
 	struct dma_fence *mfence = NULL;
 	struct migrate_vma migrate = { 0 };
 	unsigned long cpages = 0;
+	unsigned long mpages = 0;
 	dma_addr_t *scratch;
 	void *buf;
 	int r = -ENOMEM;
@@ -450,12 +438,13 @@ svm_migrate_vma_to_vram(struct kfd_node *node, struct svm_range *prange,
 	r = svm_migrate_copy_to_vram(node, prange, &migrate, &mfence, scratch, ttm_res_offset);
 	migrate_vma_pages(&migrate);
 
-	pr_debug("successful/cpages/npages 0x%lx/0x%lx/0x%lx\n",
-		svm_migrate_successful_pages(&migrate), cpages, migrate.npages);
-
 	svm_migrate_copy_done(adev, mfence);
 	migrate_vma_finalize(&migrate);
 
+	mpages = cpages - svm_migrate_unsuccessful_pages(&migrate);
+	pr_debug("successful/cpages/npages 0x%lx/0x%lx/0x%lx\n",
+			 mpages, cpages, migrate.npages);
+
 	kfd_smi_event_migration_end(node, p->lead_thread->pid,
 				    start >> PAGE_SHIFT, end >> PAGE_SHIFT,
 				    0, node->id, trigger);
@@ -465,12 +454,12 @@ svm_migrate_vma_to_vram(struct kfd_node *node, struct svm_range *prange,
 out_free:
 	kvfree(buf);
 out:
-	if (!r && cpages) {
+	if (!r && mpages) {
 		pdd = svm_range_get_pdd_by_node(prange, node);
 		if (pdd)
-			WRITE_ONCE(pdd->page_in, pdd->page_in + cpages);
+			WRITE_ONCE(pdd->page_in, pdd->page_in + mpages);
 
-		return cpages;
+		return mpages;
 	}
 	return r;
 }
@@ -498,7 +487,7 @@ svm_migrate_ram_to_vram(struct svm_range *prange, uint32_t best_loc,
 	struct vm_area_struct *vma;
 	uint64_t ttm_res_offset;
 	struct kfd_node *node;
-	unsigned long cpages = 0;
+	unsigned long mpages = 0;
 	long r = 0;
 
 	if (start_mgr < prange->start || last_mgr > prange->last) {
@@ -540,15 +529,15 @@ svm_migrate_ram_to_vram(struct svm_range *prange, uint32_t best_loc,
 			pr_debug("failed %ld to migrate\n", r);
 			break;
 		} else {
-			cpages += r;
+			mpages += r;
 		}
 		ttm_res_offset += next - addr;
 		addr = next;
 	}
 
-	if (cpages) {
+	if (mpages) {
 		prange->actual_loc = best_loc;
-		prange->vram_pages = prange->vram_pages + cpages;
+		prange->vram_pages += mpages;
 	} else if (!prange->actual_loc) {
 		/* if no page migrated and all pages from prange are at
 		 * sys ram drop svm_bo got from svm_range_vram_node_new
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 45366b4ca976..745024b31340 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -970,7 +970,7 @@ struct kfd_process {
 	struct work_struct debug_event_workarea;
 
 	/* Tracks debug per-vmid request for debug flags */
-	bool dbg_flags;
+	u32 dbg_flags;
 
 	atomic_t poison;
 	/* Queues are in paused stated because we are in the process of doing a CRIU checkpoint */
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index 70ca761bd60b..ac84c4a2ca07 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -158,13 +158,12 @@ svm_is_valid_dma_mapping_addr(struct device *dev, dma_addr_t dma_addr)
 static int
 svm_range_dma_map_dev(struct amdgpu_device *adev, struct svm_range *prange,
 		      unsigned long offset, unsigned long npages,
-		      unsigned long *hmm_pfns, uint32_t gpuidx, uint64_t *vram_pages)
+		      unsigned long *hmm_pfns, uint32_t gpuidx)
 {
 	enum dma_data_direction dir = DMA_BIDIRECTIONAL;
 	dma_addr_t *addr = prange->dma_addr[gpuidx];
 	struct device *dev = adev->dev;
 	struct page *page;
-	uint64_t vram_pages_dev;
 	int i, r;
 
 	if (!addr) {
@@ -174,7 +173,6 @@ svm_range_dma_map_dev(struct amdgpu_device *adev, struct svm_range *prange,
 		prange->dma_addr[gpuidx] = addr;
 	}
 
-	vram_pages_dev = 0;
 	addr += offset;
 	for (i = 0; i < npages; i++) {
 		if (svm_is_valid_dma_mapping_addr(dev, addr[i]))
@@ -184,7 +182,6 @@ svm_range_dma_map_dev(struct amdgpu_device *adev, struct svm_range *prange,
 		if (is_zone_device_page(page)) {
 			struct amdgpu_device *bo_adev = prange->svm_bo->node->adev;
 
-			vram_pages_dev++;
 			addr[i] = (hmm_pfns[i] << PAGE_SHIFT) +
 				   bo_adev->vm_manager.vram_base_offset -
 				   bo_adev->kfd.pgmap.range.start;
@@ -201,14 +198,14 @@ svm_range_dma_map_dev(struct amdgpu_device *adev, struct svm_range *prange,
 		pr_debug_ratelimited("dma mapping 0x%llx for page addr 0x%lx\n",
 				     addr[i] >> PAGE_SHIFT, page_to_pfn(page));
 	}
-	*vram_pages = vram_pages_dev;
+
 	return 0;
 }
 
 static int
 svm_range_dma_map(struct svm_range *prange, unsigned long *bitmap,
 		  unsigned long offset, unsigned long npages,
-		  unsigned long *hmm_pfns, uint64_t *vram_pages)
+		  unsigned long *hmm_pfns)
 {
 	struct kfd_process *p;
 	uint32_t gpuidx;
@@ -227,7 +224,7 @@ svm_range_dma_map(struct svm_range *prange, unsigned long *bitmap,
 		}
 
 		r = svm_range_dma_map_dev(pdd->dev->adev, prange, offset, npages,
-					  hmm_pfns, gpuidx, vram_pages);
+					  hmm_pfns, gpuidx);
 		if (r)
 			break;
 	}
@@ -885,14 +882,29 @@ static void svm_range_debug_dump(struct svm_range_list *svms)
 
 static void *
 svm_range_copy_array(void *psrc, size_t size, uint64_t num_elements,
-		     uint64_t offset)
+		     uint64_t offset, uint64_t *vram_pages)
 {
+	unsigned char *src = (unsigned char *)psrc + offset;
 	unsigned char *dst;
+	uint64_t i;
 
 	dst = kvmalloc_array(num_elements, size, GFP_KERNEL);
 	if (!dst)
 		return NULL;
-	memcpy(dst, (unsigned char *)psrc + offset, num_elements * size);
+
+	if (!vram_pages) {
+		memcpy(dst, src, num_elements * size);
+		return (void *)dst;
+	}
+
+	*vram_pages = 0;
+	for (i = 0; i < num_elements; i++) {
+		dma_addr_t *temp;
+		temp = (dma_addr_t *)dst + i;
+		*temp = *((dma_addr_t *)src + i);
+		if (*temp&SVM_RANGE_VRAM_DOMAIN)
+			(*vram_pages)++;
+	}
 
 	return (void *)dst;
 }
@@ -906,7 +918,7 @@ svm_range_copy_dma_addrs(struct svm_range *dst, struct svm_range *src)
 		if (!src->dma_addr[i])
 			continue;
 		dst->dma_addr[i] = svm_range_copy_array(src->dma_addr[i],
-					sizeof(*src->dma_addr[i]), src->npages, 0);
+					sizeof(*src->dma_addr[i]), src->npages, 0, NULL);
 		if (!dst->dma_addr[i])
 			return -ENOMEM;
 	}
@@ -917,7 +929,7 @@ svm_range_copy_dma_addrs(struct svm_range *dst, struct svm_range *src)
 static int
 svm_range_split_array(void *ppnew, void *ppold, size_t size,
 		      uint64_t old_start, uint64_t old_n,
-		      uint64_t new_start, uint64_t new_n)
+		      uint64_t new_start, uint64_t new_n, uint64_t *new_vram_pages)
 {
 	unsigned char *new, *old, *pold;
 	uint64_t d;
@@ -929,11 +941,12 @@ svm_range_split_array(void *ppnew, void *ppold, size_t size,
 		return 0;
 
 	d = (new_start - old_start) * size;
-	new = svm_range_copy_array(pold, size, new_n, d);
+	/* get dma addr array for new range and calculte its vram page number */
+	new = svm_range_copy_array(pold, size, new_n, d, new_vram_pages);
 	if (!new)
 		return -ENOMEM;
 	d = (new_start == old_start) ? new_n * size : 0;
-	old = svm_range_copy_array(pold, size, old_n, d);
+	old = svm_range_copy_array(pold, size, old_n, d, NULL);
 	if (!old) {
 		kvfree(new);
 		return -ENOMEM;
@@ -955,10 +968,13 @@ svm_range_split_pages(struct svm_range *new, struct svm_range *old,
 	for (i = 0; i < MAX_GPU_INSTANCE; i++) {
 		r = svm_range_split_array(&new->dma_addr[i], &old->dma_addr[i],
 					  sizeof(*old->dma_addr[i]), old->start,
-					  npages, new->start, new->npages);
+					  npages, new->start, new->npages,
+					  old->actual_loc ? &new->vram_pages : NULL);
 		if (r)
 			return r;
 	}
+	if (old->actual_loc)
+		old->vram_pages -= new->vram_pages;
 
 	return 0;
 }
@@ -982,11 +998,6 @@ svm_range_split_nodes(struct svm_range *new, struct svm_range *old,
 	new->svm_bo = svm_range_bo_ref(old->svm_bo);
 	new->ttm_res = old->ttm_res;
 
-	/* set new's vram_pages as old range's now, the acurate vram_pages
-	 * will be updated during mapping
-	 */
-	new->vram_pages = min(old->vram_pages, new->npages);
-
 	spin_lock(&new->svm_bo->list_lock);
 	list_add(&new->svm_bo_list, &new->svm_bo->range_list);
 	spin_unlock(&new->svm_bo->list_lock);
@@ -1109,7 +1120,7 @@ static int
 svm_range_split_tail(struct svm_range *prange, uint64_t new_last,
 		     struct list_head *insert_list, struct list_head *remap_list)
 {
-	struct svm_range *tail;
+	struct svm_range *tail = NULL;
 	int r = svm_range_split(prange, prange->start, new_last, &tail);
 
 	if (!r) {
@@ -1124,7 +1135,7 @@ static int
 svm_range_split_head(struct svm_range *prange, uint64_t new_start,
 		     struct list_head *insert_list, struct list_head *remap_list)
 {
-	struct svm_range *head;
+	struct svm_range *head = NULL;
 	int r = svm_range_split(prange, new_start, prange->last, &head);
 
 	if (!r) {
@@ -1573,7 +1584,6 @@ static int svm_range_validate_and_map(struct mm_struct *mm,
 	struct svm_validate_context *ctx;
 	unsigned long start, end, addr;
 	struct kfd_process *p;
-	uint64_t vram_pages;
 	void *owner;
 	int32_t idx;
 	int r = 0;
@@ -1648,15 +1658,13 @@ static int svm_range_validate_and_map(struct mm_struct *mm,
 		}
 	}
 
-	vram_pages = 0;
-	start = prange->start << PAGE_SHIFT;
-	end = (prange->last + 1) << PAGE_SHIFT;
+	start = map_start << PAGE_SHIFT;
+	end = (map_last + 1) << PAGE_SHIFT;
 	for (addr = start; !r && addr < end; ) {
 		struct hmm_range *hmm_range;
 		unsigned long map_start_vma;
 		unsigned long map_last_vma;
 		struct vm_area_struct *vma;
-		uint64_t vram_pages_vma;
 		unsigned long next = 0;
 		unsigned long offset;
 		unsigned long npages;
@@ -1683,13 +1691,11 @@ static int svm_range_validate_and_map(struct mm_struct *mm,
 		}
 
 		if (!r) {
-			offset = (addr - start) >> PAGE_SHIFT;
+			offset = (addr >> PAGE_SHIFT) - prange->start;
 			r = svm_range_dma_map(prange, ctx->bitmap, offset, npages,
-					      hmm_range->hmm_pfns, &vram_pages_vma);
+					      hmm_range->hmm_pfns);
 			if (r)
 				pr_debug("failed %d to dma map range\n", r);
-			else
-				vram_pages += vram_pages_vma;
 		}
 
 		svm_range_lock(prange);
@@ -1722,19 +1728,6 @@ static int svm_range_validate_and_map(struct mm_struct *mm,
 		addr = next;
 	}
 
-	if (addr == end) {
-		prange->vram_pages = vram_pages;
-
-		/* if prange does not include any vram page and it
-		 * has not released svm_bo drop its svm_bo reference
-		 * and set its actaul_loc to sys ram
-		 */
-		if (!vram_pages && prange->ttm_res) {
-			prange->actual_loc = 0;
-			svm_range_vram_node_free(prange);
-		}
-	}
-
 	svm_range_unreserve_bos(ctx);
 	if (!r)
 		prange->validate_timestamp = ktime_get_boottime();
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index 057284bf50bb..e5f7c92eebcb 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -1342,10 +1342,11 @@ static int kfd_create_indirect_link_prop(struct kfd_topology_device *kdev, int g
 		num_cpu++;
 	}
 
+	if (list_empty(&kdev->io_link_props))
+		return -ENODATA;
+
 	gpu_link = list_first_entry(&kdev->io_link_props,
-					struct kfd_iolink_properties, list);
-	if (!gpu_link)
-		return -ENOMEM;
+				    struct kfd_iolink_properties, list);
 
 	for (i = 0; i < num_cpu; i++) {
 		/* CPU <--> GPU */
@@ -1423,15 +1424,17 @@ static int kfd_add_peer_prop(struct kfd_topology_device *kdev,
 				peer->gpu->adev))
 		return ret;
 
+	if (list_empty(&kdev->io_link_props))
+		return -ENODATA;
+
 	iolink1 = list_first_entry(&kdev->io_link_props,
-							struct kfd_iolink_properties, list);
-	if (!iolink1)
-		return -ENOMEM;
+				   struct kfd_iolink_properties, list);
+
+	if (list_empty(&peer->io_link_props))
+		return -ENODATA;
 
 	iolink2 = list_first_entry(&peer->io_link_props,
-							struct kfd_iolink_properties, list);
-	if (!iolink2)
-		return -ENOMEM;
+				   struct kfd_iolink_properties, list);
 
 	props = kfd_alloc_struct(props);
 	if (!props)
@@ -1449,17 +1452,19 @@ static int kfd_add_peer_prop(struct kfd_topology_device *kdev,
 		/* CPU->CPU  link*/
 		cpu_dev = kfd_topology_device_by_proximity_domain(iolink1->node_to);
 		if (cpu_dev) {
-			list_for_each_entry(iolink3, &cpu_dev->io_link_props, list)
-				if (iolink3->node_to == iolink2->node_to)
-					break;
-
-			props->weight += iolink3->weight;
-			props->min_latency += iolink3->min_latency;
-			props->max_latency += iolink3->max_latency;
-			props->min_bandwidth = min(props->min_bandwidth,
-							iolink3->min_bandwidth);
-			props->max_bandwidth = min(props->max_bandwidth,
-							iolink3->max_bandwidth);
+			list_for_each_entry(iolink3, &cpu_dev->io_link_props, list) {
+				if (iolink3->node_to != iolink2->node_to)
+					continue;
+
+				props->weight += iolink3->weight;
+				props->min_latency += iolink3->min_latency;
+				props->max_latency += iolink3->max_latency;
+				props->min_bandwidth = min(props->min_bandwidth,
+							   iolink3->min_bandwidth);
+				props->max_bandwidth = min(props->max_bandwidth,
+							   iolink3->max_bandwidth);
+				break;
+			}
 		} else {
 			WARN(1, "CPU node not found");
 		}