Merge branch 'next' into for-linus

Prepare input updates for 6.2 merge window.

1 files changed, 46 insertions, 28 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
index b059a77b6081..2797029bd500 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
@@ -223,7 +223,7 @@ svm_migrate_get_vram_page(struct svm_range *prange, unsigned long pfn)
 	page = pfn_to_page(pfn);
 	svm_range_bo_ref(prange->svm_bo);
 	page->zone_device_data = prange->svm_bo;
-	lock_page(page);
+	zone_device_page_init(page);
 }
 
 static void
@@ -322,12 +322,13 @@ svm_migrate_copy_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
 	for (i = j = 0; i < npages; i++) {
 		struct page *spage;
 
+		dst[i] = cursor.start + (j << PAGE_SHIFT);
+		migrate->dst[i] = svm_migrate_addr_to_pfn(adev, dst[i]);
+		svm_migrate_get_vram_page(prange, migrate->dst[i]);
+		migrate->dst[i] = migrate_pfn(migrate->dst[i]);
+
 		spage = migrate_pfn_to_page(migrate->src[i]);
 		if (spage && !is_zone_device_page(spage)) {
-			dst[i] = cursor.start + (j << PAGE_SHIFT);
-			migrate->dst[i] = svm_migrate_addr_to_pfn(adev, dst[i]);
-			svm_migrate_get_vram_page(prange, migrate->dst[i]);
-			migrate->dst[i] = migrate_pfn(migrate->dst[i]);
 			src[i] = dma_map_page(dev, spage, 0, PAGE_SIZE,
 					      DMA_TO_DEVICE);
 			r = dma_mapping_error(dev, src[i]);
@@ -409,7 +410,7 @@ svm_migrate_vma_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
 	uint64_t npages = (end - start) >> PAGE_SHIFT;
 	struct kfd_process_device *pdd;
 	struct dma_fence *mfence = NULL;
-	struct migrate_vma migrate;
+	struct migrate_vma migrate = { 0 };
 	unsigned long cpages = 0;
 	dma_addr_t *scratch;
 	void *buf;
@@ -522,9 +523,6 @@ svm_migrate_ram_to_vram(struct svm_range *prange, uint32_t best_loc,
 	pr_debug("svms 0x%p [0x%lx 0x%lx] to gpu 0x%x\n", prange->svms,
 		 prange->start, prange->last, best_loc);
 
-	/* FIXME: workaround for page locking bug with invalid pages */
-	svm_range_prefault(prange, mm, SVM_ADEV_PGMAP_OWNER(adev));
-
 	start = prange->start << PAGE_SHIFT;
 	end = (prange->last + 1) << PAGE_SHIFT;
 
@@ -668,7 +666,7 @@ out_oom:
 static long
 svm_migrate_vma_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
 		       struct vm_area_struct *vma, uint64_t start, uint64_t end,
-		       uint32_t trigger)
+		       uint32_t trigger, struct page *fault_page)
 {
 	struct kfd_process *p = container_of(prange->svms, struct kfd_process, svms);
 	uint64_t npages = (end - start) >> PAGE_SHIFT;
@@ -676,7 +674,7 @@ svm_migrate_vma_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
 	unsigned long cpages = 0;
 	struct kfd_process_device *pdd;
 	struct dma_fence *mfence = NULL;
-	struct migrate_vma migrate;
+	struct migrate_vma migrate = { 0 };
 	dma_addr_t *scratch;
 	void *buf;
 	int r = -ENOMEM;
@@ -699,6 +697,7 @@ svm_migrate_vma_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
 
 	migrate.src = buf;
 	migrate.dst = migrate.src + npages;
+	migrate.fault_page = fault_page;
 	scratch = (dma_addr_t *)(migrate.dst + npages);
 
 	kfd_smi_event_migration_start(adev->kfd.dev, p->lead_thread->pid,
@@ -766,7 +765,7 @@ out:
  * 0 - OK, otherwise error code
  */
 int svm_migrate_vram_to_ram(struct svm_range *prange, struct mm_struct *mm,
-			    uint32_t trigger)
+			    uint32_t trigger, struct page *fault_page)
 {
 	struct amdgpu_device *adev;
 	struct vm_area_struct *vma;
@@ -807,7 +806,8 @@ int svm_migrate_vram_to_ram(struct svm_range *prange, struct mm_struct *mm,
 		}
 
 		next = min(vma->vm_end, end);
-		r = svm_migrate_vma_to_ram(adev, prange, vma, addr, next, trigger);
+		r = svm_migrate_vma_to_ram(adev, prange, vma, addr, next, trigger,
+			fault_page);
 		if (r < 0) {
 			pr_debug("failed %ld to migrate prange %p\n", r, prange);
 			break;
@@ -851,7 +851,7 @@ svm_migrate_vram_to_vram(struct svm_range *prange, uint32_t best_loc,
 	pr_debug("from gpu 0x%x to gpu 0x%x\n", prange->actual_loc, best_loc);
 
 	do {
-		r = svm_migrate_vram_to_ram(prange, mm, trigger);
+		r = svm_migrate_vram_to_ram(prange, mm, trigger, NULL);
 		if (r)
 			return r;
 	} while (prange->actual_loc && --retries);
@@ -886,7 +886,7 @@ svm_migrate_to_vram(struct svm_range *prange, uint32_t best_loc,
 static vm_fault_t svm_migrate_to_ram(struct vm_fault *vmf)
 {
 	unsigned long addr = vmf->address;
-	struct vm_area_struct *vma;
+	struct svm_range_bo *svm_bo;
 	enum svm_work_list_ops op;
 	struct svm_range *parent;
 	struct svm_range *prange;
@@ -894,29 +894,42 @@ static vm_fault_t svm_migrate_to_ram(struct vm_fault *vmf)
 	struct mm_struct *mm;
 	int r = 0;
 
-	vma = vmf->vma;
-	mm = vma->vm_mm;
+	svm_bo = vmf->page->zone_device_data;
+	if (!svm_bo) {
+		pr_debug("failed get device page at addr 0x%lx\n", addr);
+		return VM_FAULT_SIGBUS;
+	}
+	if (!mmget_not_zero(svm_bo->eviction_fence->mm)) {
+		pr_debug("addr 0x%lx of process mm is destroyed\n", addr);
+		return VM_FAULT_SIGBUS;
+	}
+
+	mm = svm_bo->eviction_fence->mm;
+	if (mm != vmf->vma->vm_mm)
+		pr_debug("addr 0x%lx is COW mapping in child process\n", addr);
 
-	p = kfd_lookup_process_by_mm(vma->vm_mm);
+	p = kfd_lookup_process_by_mm(mm);
 	if (!p) {
 		pr_debug("failed find process at fault address 0x%lx\n", addr);
-		return VM_FAULT_SIGBUS;
+		r = VM_FAULT_SIGBUS;
+		goto out_mmput;
 	}
 	if (READ_ONCE(p->svms.faulting_task) == current) {
 		pr_debug("skipping ram migration\n");
-		kfd_unref_process(p);
-		return 0;
+		r = 0;
+		goto out_unref_process;
 	}
-	addr >>= PAGE_SHIFT;
+
 	pr_debug("CPU page fault svms 0x%p address 0x%lx\n", &p->svms, addr);
+	addr >>= PAGE_SHIFT;
 
 	mutex_lock(&p->svms.lock);
 
 	prange = svm_range_from_addr(&p->svms, addr, &parent);
 	if (!prange) {
-		pr_debug("cannot find svm range at 0x%lx\n", addr);
+		pr_debug("failed get range svms 0x%p addr 0x%lx\n", &p->svms, addr);
 		r = -EFAULT;
-		goto out;
+		goto out_unlock_svms;
 	}
 
 	mutex_lock(&parent->migrate_mutex);
@@ -938,10 +951,12 @@ static vm_fault_t svm_migrate_to_ram(struct vm_fault *vmf)
 		goto out_unlock_prange;
 	}
 
-	r = svm_migrate_vram_to_ram(prange, mm, KFD_MIGRATE_TRIGGER_PAGEFAULT_CPU);
+	r = svm_migrate_vram_to_ram(prange, vmf->vma->vm_mm,
+				    KFD_MIGRATE_TRIGGER_PAGEFAULT_CPU,
+				    vmf->page);
 	if (r)
-		pr_debug("failed %d migrate 0x%p [0x%lx 0x%lx] to ram\n", r,
-			 prange, prange->start, prange->last);
+		pr_debug("failed %d migrate svms 0x%p range 0x%p [0x%lx 0x%lx]\n",
+			 r, prange->svms, prange, prange->start, prange->last);
 
 	/* xnack on, update mapping on GPUs with ACCESS_IN_PLACE */
 	if (p->xnack_enabled && parent == prange)
@@ -955,9 +970,12 @@ out_unlock_prange:
 	if (prange != parent)
 		mutex_unlock(&prange->migrate_mutex);
 	mutex_unlock(&parent->migrate_mutex);
-out:
+out_unlock_svms:
 	mutex_unlock(&p->svms.lock);
+out_unref_process:
 	kfd_unref_process(p);
+out_mmput:
+	mmput(mm);
 
 	pr_debug("CPU fault svms 0x%p address 0x%lx done\n", &p->svms, addr);