[PATCH 1/4] drm/amdgpu: make set_prt callback optional and fix error handling

[PATCH 1/4] drm/amdgpu: make set_prt callback optional and fix error handling

[Christian König]

From: Christian König <christian.koenig@amd.com>

PRT support is completely implemented now and we left it
turned on accidentially in the error path.

Signed-off-by: Christian König <christian.koenig@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 29 +++++++++++++++--------------
 1 file changed, 15 insertions(+), 14 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 344b535..0d700fc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -1194,10 +1194,13 @@ static void amdgpu_vm_update_prt_state(struct amdgpu_device *adev)
 }
 
 /**
- * amdgpu_vm_prt_put - add a PRT user
+ * amdgpu_vm_prt_get - add a PRT user
  */
 static void amdgpu_vm_prt_get(struct amdgpu_device *adev)
 {
+	if (!adev->gart.gart_funcs->set_prt)
+		return;
+
 	if (atomic_inc_return(&adev->vm_manager.num_prt_users) == 1)
 		amdgpu_vm_update_prt_state(adev);
 }
@@ -1228,9 +1231,12 @@ static void amdgpu_vm_prt_cb(struct fence *fence, struct fence_cb *_cb)
 static void amdgpu_vm_add_prt_cb(struct amdgpu_device *adev,
 				 struct fence *fence)
 {
-	struct amdgpu_prt_cb *cb = kmalloc(sizeof(struct amdgpu_prt_cb),
-					   GFP_KERNEL);
+	struct amdgpu_prt_cb *cb;
 
+	if (!adev->gart.gart_funcs->set_prt)
+		return;
+
+	cb = kmalloc(sizeof(struct amdgpu_prt_cb), GFP_KERNEL);
 	if (!cb) {
 		/* Last resort when we are OOM */
 		if (fence)
@@ -1445,14 +1451,6 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 	    size == 0 || size & AMDGPU_GPU_PAGE_MASK)
 		return -EINVAL;
 
-	if (flags & AMDGPU_PTE_PRT) {
-		/* Check if we have PRT hardware support */
-		if (!adev->gart.gart_funcs->set_prt)
-			return -EINVAL;
-
-		amdgpu_vm_prt_get(adev);
-	}
-
 	/* make sure object fit at this offset */
 	eaddr = saddr + size - 1;
 	if (saddr >= eaddr ||
@@ -1533,6 +1531,9 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 		vm->page_tables[pt_idx].addr = 0;
 	}
 
+	if (flags & AMDGPU_PTE_PRT)
+		amdgpu_vm_prt_get(adev);
+
 	return 0;
 
 error_free:
@@ -1753,7 +1754,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 {
 	struct amdgpu_bo_va_mapping *mapping, *tmp;
-	bool prt_fini_called = false;
+	bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt;
 	int i;
 
 	amd_sched_entity_fini(vm->entity.sched, &vm->entity);
@@ -1767,9 +1768,9 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 		kfree(mapping);
 	}
 	list_for_each_entry_safe(mapping, tmp, &vm->freed, list) {
-		if (mapping->flags & AMDGPU_PTE_PRT && !prt_fini_called) {
+		if (mapping->flags & AMDGPU_PTE_PRT && prt_fini_needed) {
 			amdgpu_vm_prt_fini(adev, vm);
-			prt_fini_called = true;
+			prt_fini_needed = false;
 		}
 
 		list_del(&mapping->list);
-- 
2.7.4

_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

[PATCH 2/4] drm/amdgpu: separate page table allocation from mapping

[Christian König]

From: Christian König <christian.koenig@amd.com>

This makes it easier to implement a replace operation.

Signed-off-by: Christian König <christian.koenig@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c  |   6 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c |   9 ++
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c   | 136 +++++++++++++++++--------------
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h   |   3 +
 4 files changed, 92 insertions(+), 62 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
index ba9077b..f3bf864 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
@@ -633,8 +633,12 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 
 	switch (args->operation) {
 	case AMDGPU_VA_OP_MAP:
-		va_flags = amdgpu_vm_get_pte_flags(adev, args->flags);
+		r = amdgpu_vm_alloc_pts(adev, bo_va->vm, args->va_address,
+					args->map_size);
+		if (r)
+			goto error_backoff;
 
+		va_flags = amdgpu_vm_get_pte_flags(adev, args->flags);
 		r = amdgpu_vm_bo_map(adev, bo_va, args->va_address,
 				     args->offset_in_bo, args->map_size,
 				     va_flags);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
index 20b4fa6..31b30e9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
@@ -75,6 +75,15 @@ int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 		return -ENOMEM;
 	}
 
+	r = amdgpu_vm_alloc_pts(adev, bo_va->vm, AMDGPU_CSA_VADDR,
+				   AMDGPU_CSA_SIZE);
+	if (r) {
+		DRM_ERROR("failed to allocate pts for static CSA, err=%d\n", r);
+		amdgpu_vm_bo_rmv(adev, bo_va);
+		ttm_eu_backoff_reservation(&ticket, &list);
+		return r;
+	}
+
 	r = amdgpu_vm_bo_map(adev, bo_va, AMDGPU_CSA_VADDR, 0,AMDGPU_CSA_SIZE,
 						AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE |
 						AMDGPU_PTE_EXECUTABLE);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 0d700fc..7638271 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -187,6 +187,78 @@ void amdgpu_vm_move_pt_bos_in_lru(struct amdgpu_device *adev,
 	spin_unlock(&glob->lru_lock);
 }
 
+/**
+ * amdgpu_vm_alloc_pts - Allocate page tables.
+ *
+ * @adev: amdgpu_device pointer
+ * @vm: VM to allocate page tables for
+ * @saddr: Start address which needs to be allocated
+ * @size: Size from start address we need.
+ *
+ * Make sure the page tables are allocated.
+ */
+int amdgpu_vm_alloc_pts(struct amdgpu_device *adev,
+			struct amdgpu_vm *vm,
+			uint64_t saddr, uint64_t size)
+{
+	unsigned last_pfn, pt_idx;
+	uint64_t eaddr;
+	int r;
+
+	/* validate the parameters */
+	if (saddr & AMDGPU_GPU_PAGE_MASK || size & AMDGPU_GPU_PAGE_MASK)
+		return -EINVAL;
+
+	eaddr = saddr + size - 1;
+	last_pfn = eaddr / AMDGPU_GPU_PAGE_SIZE;
+	if (last_pfn >= adev->vm_manager.max_pfn) {
+		dev_err(adev->dev, "va above limit (0x%08X >= 0x%08X)\n",
+			last_pfn, adev->vm_manager.max_pfn);
+		return -EINVAL;
+	}
+
+	saddr /= AMDGPU_GPU_PAGE_SIZE;
+	eaddr /= AMDGPU_GPU_PAGE_SIZE;
+
+	saddr >>= amdgpu_vm_block_size;
+	eaddr >>= amdgpu_vm_block_size;
+
+	BUG_ON(eaddr >= amdgpu_vm_num_pdes(adev));
+
+	if (eaddr > vm->max_pde_used)
+		vm->max_pde_used = eaddr;
+
+	/* walk over the address space and allocate the page tables */
+	for (pt_idx = saddr; pt_idx <= eaddr; ++pt_idx) {
+		struct reservation_object *resv = vm->page_directory->tbo.resv;
+		struct amdgpu_bo *pt;
+
+		if (vm->page_tables[pt_idx].bo)
+			continue;
+
+		r = amdgpu_bo_create(adev, AMDGPU_VM_PTE_COUNT * 8,
+				     AMDGPU_GPU_PAGE_SIZE, true,
+				     AMDGPU_GEM_DOMAIN_VRAM,
+				     AMDGPU_GEM_CREATE_NO_CPU_ACCESS |
+				     AMDGPU_GEM_CREATE_SHADOW |
+				     AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
+				     AMDGPU_GEM_CREATE_VRAM_CLEARED,
+				     NULL, resv, &pt);
+		if (r)
+			return r;
+
+		/* Keep a reference to the page table to avoid freeing
+		 * them up in the wrong order.
+		 */
+		pt->parent = amdgpu_bo_ref(vm->page_directory);
+
+		vm->page_tables[pt_idx].bo = pt;
+		vm->page_tables[pt_idx].addr = 0;
+	}
+
+	return 0;
+}
+
 static bool amdgpu_vm_is_gpu_reset(struct amdgpu_device *adev,
 			      struct amdgpu_vm_id *id)
 {
@@ -1442,9 +1514,7 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 	struct amdgpu_bo_va_mapping *mapping;
 	struct amdgpu_vm *vm = bo_va->vm;
 	struct interval_tree_node *it;
-	unsigned last_pfn, pt_idx;
 	uint64_t eaddr;
-	int r;
 
 	/* validate the parameters */
 	if (saddr & AMDGPU_GPU_PAGE_MASK || offset & AMDGPU_GPU_PAGE_MASK ||
@@ -1457,13 +1527,6 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 	    (bo_va->bo && offset + size > amdgpu_bo_size(bo_va->bo)))
 		return -EINVAL;
 
-	last_pfn = eaddr / AMDGPU_GPU_PAGE_SIZE;
-	if (last_pfn >= adev->vm_manager.max_pfn) {
-		dev_err(adev->dev, "va above limit (0x%08X >= 0x%08X)\n",
-			last_pfn, adev->vm_manager.max_pfn);
-		return -EINVAL;
-	}
-
 	saddr /= AMDGPU_GPU_PAGE_SIZE;
 	eaddr /= AMDGPU_GPU_PAGE_SIZE;
 
@@ -1475,15 +1538,12 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 		dev_err(adev->dev, "bo %p va 0x%010Lx-0x%010Lx conflict with "
 			"0x%010lx-0x%010lx\n", bo_va->bo, saddr, eaddr,
 			tmp->it.start, tmp->it.last + 1);
-		r = -EINVAL;
-		goto error;
+		return -EINVAL;
 	}
 
 	mapping = kmalloc(sizeof(*mapping), GFP_KERNEL);
-	if (!mapping) {
-		r = -ENOMEM;
-		goto error;
-	}
+	if (!mapping)
+		return -ENOMEM;
 
 	INIT_LIST_HEAD(&mapping->list);
 	mapping->it.start = saddr;
@@ -1494,56 +1554,10 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 	list_add(&mapping->list, &bo_va->invalids);
 	interval_tree_insert(&mapping->it, &vm->va);
 
-	/* Make sure the page tables are allocated */
-	saddr >>= amdgpu_vm_block_size;
-	eaddr >>= amdgpu_vm_block_size;
-
-	BUG_ON(eaddr >= amdgpu_vm_num_pdes(adev));
-
-	if (eaddr > vm->max_pde_used)
-		vm->max_pde_used = eaddr;
-
-	/* walk over the address space and allocate the page tables */
-	for (pt_idx = saddr; pt_idx <= eaddr; ++pt_idx) {
-		struct reservation_object *resv = vm->page_directory->tbo.resv;
-		struct amdgpu_bo *pt;
-
-		if (vm->page_tables[pt_idx].bo)
-			continue;
-
-		r = amdgpu_bo_create(adev, AMDGPU_VM_PTE_COUNT * 8,
-				     AMDGPU_GPU_PAGE_SIZE, true,
-				     AMDGPU_GEM_DOMAIN_VRAM,
-				     AMDGPU_GEM_CREATE_NO_CPU_ACCESS |
-				     AMDGPU_GEM_CREATE_SHADOW |
-				     AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
-				     AMDGPU_GEM_CREATE_VRAM_CLEARED,
-				     NULL, resv, &pt);
-		if (r)
-			goto error_free;
-
-		/* Keep a reference to the page table to avoid freeing
-		 * them up in the wrong order.
-		 */
-		pt->parent = amdgpu_bo_ref(vm->page_directory);
-
-		vm->page_tables[pt_idx].bo = pt;
-		vm->page_tables[pt_idx].addr = 0;
-	}
-
 	if (flags & AMDGPU_PTE_PRT)
 		amdgpu_vm_prt_get(adev);
 
 	return 0;
-
-error_free:
-	list_del(&mapping->list);
-	interval_tree_remove(&mapping->it, &vm->va);
-	trace_amdgpu_vm_bo_unmap(bo_va, mapping);
-	amdgpu_vm_free_mapping(adev, vm, mapping, NULL);
-
-error:
-	return r;
 }
 
 /**
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
index 8e5abd2..36ca3b2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
@@ -179,6 +179,9 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			      void *param);
 void amdgpu_vm_move_pt_bos_in_lru(struct amdgpu_device *adev,
 				  struct amdgpu_vm *vm);
+int amdgpu_vm_alloc_pts(struct amdgpu_device *adev,
+			struct amdgpu_vm *vm,
+			uint64_t saddr, uint64_t size);
 int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 		      struct amdgpu_sync *sync, struct fence *fence,
 		      struct amdgpu_job *job);
-- 
2.7.4

_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

[PATCH 3/4] drm/amdgpu: implement AMDGPU_VA_OP_CLEAR v2

[Christian König]

From: Christian König <christian.koenig@amd.com>

A new VM operation to remove all mappings in a range.

v2: limit unmapped area as noted by Jerry

Signed-off-by: Christian König <christian.koenig@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c   | 27 ++++++---
 drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h |  2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c    | 99 +++++++++++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h    |  3 +
 include/uapi/drm/amdgpu_drm.h             |  1 +
 5 files changed, 124 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
index f3bf864..5c3580d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
@@ -509,14 +509,16 @@ static int amdgpu_gem_va_check(void *param, struct amdgpu_bo *bo)
  * amdgpu_gem_va_update_vm -update the bo_va in its VM
  *
  * @adev: amdgpu_device pointer
+ * @vm: vm to update
  * @bo_va: bo_va to update
  * @list: validation list
- * @operation: map or unmap
+ * @operation: map, unmap or clear
  *
  * Update the bo_va directly after setting its address. Errors are not
  * vital here, so they are not reported back to userspace.
  */
 static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
+				    struct amdgpu_vm *vm,
 				    struct amdgpu_bo_va *bo_va,
 				    struct list_head *list,
 				    uint32_t operation)
@@ -531,16 +533,16 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
 			goto error;
 	}
 
-	r = amdgpu_vm_validate_pt_bos(adev, bo_va->vm, amdgpu_gem_va_check,
+	r = amdgpu_vm_validate_pt_bos(adev, vm, amdgpu_gem_va_check,
 				      NULL);
 	if (r)
 		goto error;
 
-	r = amdgpu_vm_update_page_directory(adev, bo_va->vm);
+	r = amdgpu_vm_update_page_directory(adev, vm);
 	if (r)
 		goto error;
 
-	r = amdgpu_vm_clear_freed(adev, bo_va->vm);
+	r = amdgpu_vm_clear_freed(adev, vm);
 	if (r)
 		goto error;
 
@@ -594,6 +596,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	switch (args->operation) {
 	case AMDGPU_VA_OP_MAP:
 	case AMDGPU_VA_OP_UNMAP:
+	case AMDGPU_VA_OP_CLEAR:
 		break;
 	default:
 		dev_err(&dev->pdev->dev, "unsupported operation %d\n",
@@ -602,7 +605,8 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	}
 
 	INIT_LIST_HEAD(&list);
-	if (!(args->flags & AMDGPU_VM_PAGE_PRT)) {
+	if ((args->operation != AMDGPU_VA_OP_CLEAR) &&
+	    !(args->flags & AMDGPU_VM_PAGE_PRT)) {
 		gobj = drm_gem_object_lookup(filp, args->handle);
 		if (gobj == NULL)
 			return -ENOENT;
@@ -627,8 +631,10 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 			r = -ENOENT;
 			goto error_backoff;
 		}
-	} else {
+	} else if (args->operation != AMDGPU_VA_OP_CLEAR) {
 		bo_va = fpriv->prt_va;
+	} else {
+		bo_va = NULL;
 	}
 
 	switch (args->operation) {
@@ -646,11 +652,18 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	case AMDGPU_VA_OP_UNMAP:
 		r = amdgpu_vm_bo_unmap(adev, bo_va, args->va_address);
 		break;
+
+	case AMDGPU_VA_OP_CLEAR:
+		r = amdgpu_vm_bo_clear_mappings(adev, &fpriv->vm,
+						args->va_address,
+						args->map_size);
+		break;
 	default:
 		break;
 	}
 	if (!r && !(args->flags & AMDGPU_VM_DELAY_UPDATE) && !amdgpu_vm_debug)
-		amdgpu_gem_va_update_vm(adev, bo_va, &list, args->operation);
+		amdgpu_gem_va_update_vm(adev, &fpriv->vm, bo_va, &list,
+					args->operation);
 
 error_backoff:
 	ttm_eu_backoff_reservation(&ticket, &list);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
index 03f598e..747d7ca 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
@@ -215,7 +215,7 @@ TRACE_EVENT(amdgpu_vm_bo_unmap,
 			     ),
 
 	    TP_fast_assign(
-			   __entry->bo = bo_va->bo;
+			   __entry->bo = bo_va ? bo_va->bo : NULL;
 			   __entry->start = mapping->it.start;
 			   __entry->last = mapping->it.last;
 			   __entry->offset = mapping->offset;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 7638271..7bfe76d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -1613,6 +1613,105 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 }
 
 /**
+ * amdgpu_vm_bo_clear_mappings - remove all mappings in a specific range
+ *
+ * @adev: amdgpu_device pointer
+ * @vm: VM structure to use
+ * @saddr: start of the range
+ * @size: size of the range
+ *
+ * Remove all mappings in a range, split them as appropriate.
+ * Returns 0 for success, error for failure.
+ */
+int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
+				struct amdgpu_vm *vm,
+				uint64_t saddr, uint64_t size)
+{
+	struct amdgpu_bo_va_mapping *before, *after, *tmp, *next;
+	struct interval_tree_node *it;
+	LIST_HEAD(removed);
+	uint64_t eaddr;
+
+	eaddr = saddr + size - 1;
+	saddr /= AMDGPU_GPU_PAGE_SIZE;
+	eaddr /= AMDGPU_GPU_PAGE_SIZE;
+
+	/* Allocate all the needed memory */
+	before = kzalloc(sizeof(*before), GFP_KERNEL);
+	if (!before)
+		return -ENOMEM;
+
+	after = kzalloc(sizeof(*after), GFP_KERNEL);
+	if (!after) {
+		kfree(before);
+		return -ENOMEM;
+	}
+
+	/* Now gather all removed mappings */
+	it = interval_tree_iter_first(&vm->va, saddr, eaddr);
+	while (it) {
+		tmp = container_of(it, struct amdgpu_bo_va_mapping, it);
+		it = interval_tree_iter_next(it, saddr, eaddr);
+
+		/* Remember mapping split at the start */
+		if (tmp->it.start < saddr) {
+			before->it.start = tmp->it.start;;
+			before->it.last = saddr - 1;
+			before->offset = tmp->offset;
+			before->flags = tmp->flags;
+			list_add(&before->list, &tmp->list);
+		}
+
+		/* Remember mapping split at the end */
+		if (tmp->it.last > eaddr) {
+			after->it.start = eaddr + 1;
+			after->it.last = tmp->it.last;
+			after->offset = tmp->offset;
+			after->offset += after->it.start - tmp->it.start;
+			after->flags = tmp->flags;
+			list_add(&after->list, &tmp->list);
+		}
+
+		list_del(&tmp->list);
+		list_add(&tmp->list, &removed);
+	}
+
+	/* And free them up */
+	list_for_each_entry_safe(tmp, next, &removed, list) {
+		interval_tree_remove(&tmp->it, &vm->va);
+		list_del(&tmp->list);
+
+		if (tmp->it.start < saddr)
+		    tmp->it.start = saddr;
+		if (tmp->it.last > eaddr)
+		    tmp->it.last = eaddr;
+
+		list_add(&tmp->list, &vm->freed);
+		trace_amdgpu_vm_bo_unmap(NULL, tmp);
+	}
+
+	/* Insert partial mapping before the range*/
+	if (before->it.start != before->it.last) {
+		interval_tree_insert(&before->it, &vm->va);
+		if (before->flags & AMDGPU_PTE_PRT)
+			amdgpu_vm_prt_get(adev);
+	} else {
+		kfree(before);
+	}
+
+	/* Insert partial mapping after the range */
+	if (after->it.start != after->it.last) {
+		interval_tree_insert(&after->it, &vm->va);
+		if (after->flags & AMDGPU_PTE_PRT)
+			amdgpu_vm_prt_get(adev);
+	} else {
+		kfree(after);
+	}
+
+	return 0;
+}
+
+/**
  * amdgpu_vm_bo_rmv - remove a bo to a specific vm
  *
  * @adev: amdgpu_device pointer
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
index 36ca3b2..7a10b49 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
@@ -210,6 +210,9 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 		       struct amdgpu_bo_va *bo_va,
 		       uint64_t addr);
+int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
+				struct amdgpu_vm *vm,
+				uint64_t saddr, uint64_t size);
 void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
 		      struct amdgpu_bo_va *bo_va);
 
diff --git a/include/uapi/drm/amdgpu_drm.h b/include/uapi/drm/amdgpu_drm.h
index dd6c934..67173d7 100644
--- a/include/uapi/drm/amdgpu_drm.h
+++ b/include/uapi/drm/amdgpu_drm.h
@@ -352,6 +352,7 @@ struct drm_amdgpu_gem_op {
 
 #define AMDGPU_VA_OP_MAP			1
 #define AMDGPU_VA_OP_UNMAP			2
+#define AMDGPU_VA_OP_CLEAR			3
 
 /* Delay the page table update till the next CS */
 #define AMDGPU_VM_DELAY_UPDATE		(1 << 0)
-- 
2.7.4

_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

[PATCH 4/4] drm/amdgpu: add a VM mapping replace operation v2

[Christian König]

From: Christian König <christian.koenig@amd.com>

Add a new operation to replace mappings in a VM with a new one.

v2: Fix Jerry's comment, separate out clear operation.

Signed-off-by: Christian König <christian.koenig@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 15 +++++++-
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c  | 64 +++++++++++++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h  |  4 +++
 include/uapi/drm/amdgpu_drm.h           |  1 +
 4 files changed, 83 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
index 5c3580d..4aea0a4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
@@ -546,7 +546,8 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
 	if (r)
 		goto error;
 
-	if (operation == AMDGPU_VA_OP_MAP)
+	if (operation == AMDGPU_VA_OP_MAP ||
+	    operation == AMDGPU_VA_OP_REPLACE)
 		r = amdgpu_vm_bo_update(adev, bo_va, false);
 
 error:
@@ -597,6 +598,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	case AMDGPU_VA_OP_MAP:
 	case AMDGPU_VA_OP_UNMAP:
 	case AMDGPU_VA_OP_CLEAR:
+	case AMDGPU_VA_OP_REPLACE:
 		break;
 	default:
 		dev_err(&dev->pdev->dev, "unsupported operation %d\n",
@@ -658,6 +660,17 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 						args->va_address,
 						args->map_size);
 		break;
+	case AMDGPU_VA_OP_REPLACE:
+		r = amdgpu_vm_alloc_pts(adev, bo_va->vm, args->va_address,
+					args->map_size);
+		if (r)
+			goto error_backoff;
+
+		va_flags = amdgpu_vm_get_pte_flags(adev, args->flags);
+		r = amdgpu_vm_bo_replace_map(adev, bo_va, args->va_address,
+					     args->offset_in_bo, args->map_size,
+					     va_flags);
+		break;
 	default:
 		break;
 	}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 7bfe76d..a119c2b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -1561,6 +1561,70 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 }
 
 /**
+ * amdgpu_vm_bo_replace_map - map bo inside a vm, replacing existing mappings
+ *
+ * @adev: amdgpu_device pointer
+ * @bo_va: bo_va to store the address
+ * @saddr: where to map the BO
+ * @offset: requested offset in the BO
+ * @flags: attributes of pages (read/write/valid/etc.)
+ *
+ * Add a mapping of the BO at the specefied addr into the VM. Replace existing
+ * mappings as we do so.
+ * Returns 0 for success, error for failure.
+ *
+ * Object has to be reserved and unreserved outside!
+ */
+int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
+			     struct amdgpu_bo_va *bo_va,
+			     uint64_t saddr, uint64_t offset,
+			     uint64_t size, uint64_t flags)
+{
+	struct amdgpu_bo_va_mapping *mapping;
+	struct amdgpu_vm *vm = bo_va->vm;
+	uint64_t eaddr;
+	int r;
+
+	/* validate the parameters */
+	if (saddr & AMDGPU_GPU_PAGE_MASK || offset & AMDGPU_GPU_PAGE_MASK ||
+	    size == 0 || size & AMDGPU_GPU_PAGE_MASK)
+		return -EINVAL;
+
+	/* make sure object fit at this offset */
+	eaddr = saddr + size - 1;
+	if (saddr >= eaddr ||
+	    (bo_va->bo && offset + size > amdgpu_bo_size(bo_va->bo)))
+		return -EINVAL;
+
+	/* Allocate all the needed memory */
+	mapping = kmalloc(sizeof(*mapping), GFP_KERNEL);
+	if (!mapping)
+		return -ENOMEM;
+
+	r = amdgpu_vm_bo_clear_mappings(adev, bo_va->vm, saddr, size);
+	if (r) {
+		kfree(mapping);
+		return r;
+	}
+
+	saddr /= AMDGPU_GPU_PAGE_SIZE;
+	eaddr /= AMDGPU_GPU_PAGE_SIZE;
+
+	mapping->it.start = saddr;
+	mapping->it.last = eaddr;
+	mapping->offset = offset;
+	mapping->flags = flags;
+
+	list_add(&mapping->list, &bo_va->invalids);
+	interval_tree_insert(&mapping->it, &vm->va);
+
+	if (flags & AMDGPU_PTE_PRT)
+		amdgpu_vm_prt_get(adev);
+
+	return 0;
+}
+
+/**
  * amdgpu_vm_bo_unmap - remove bo mapping from vm
  *
  * @adev: amdgpu_device pointer
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
index 7a10b49..14f61fa 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
@@ -207,6 +207,10 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 		     struct amdgpu_bo_va *bo_va,
 		     uint64_t addr, uint64_t offset,
 		     uint64_t size, uint64_t flags);
+int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
+			     struct amdgpu_bo_va *bo_va,
+			     uint64_t addr, uint64_t offset,
+			     uint64_t size, uint64_t flags);
 int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 		       struct amdgpu_bo_va *bo_va,
 		       uint64_t addr);
diff --git a/include/uapi/drm/amdgpu_drm.h b/include/uapi/drm/amdgpu_drm.h
index 67173d7..d590364 100644
--- a/include/uapi/drm/amdgpu_drm.h
+++ b/include/uapi/drm/amdgpu_drm.h
@@ -353,6 +353,7 @@ struct drm_amdgpu_gem_op {
 #define AMDGPU_VA_OP_MAP			1
 #define AMDGPU_VA_OP_UNMAP			2
 #define AMDGPU_VA_OP_CLEAR			3
+#define AMDGPU_VA_OP_REPLACE			4
 
 /* Delay the page table update till the next CS */
 #define AMDGPU_VM_DELAY_UPDATE		(1 << 0)
-- 
2.7.4

_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

RE: [PATCH 1/4] drm/amdgpu: make set_prt callback optional and fix error handling

[Zhang, Jerry]

The series(1~4) are already Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>

Regards,
Jerry (Junwei Zhang)

Linux Base Graphics
SRDC Software Development
_____________________________________


> -----Original Message-----
> From: amd-gfx [mailto:amd-gfx-bounces@lists.freedesktop.org] On Behalf Of
> Christian K?nig
> Sent: Monday, March 13, 2017 17:14
> To: amd-gfx@lists.freedesktop.org
> Subject: [PATCH 1/4] drm/amdgpu: make set_prt callback optional and fix error
> handling
> 
> From: Christian König <christian.koenig@amd.com>
> 
> PRT support is completely implemented now and we left it turned on
> accidentially in the error path.
> 
> Signed-off-by: Christian König <christian.koenig@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 29 +++++++++++++++-----------
> ---
>  1 file changed, 15 insertions(+), 14 deletions(-)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> index 344b535..0d700fc 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> @@ -1194,10 +1194,13 @@ static void amdgpu_vm_update_prt_state(struct
> amdgpu_device *adev)  }
> 
>  /**
> - * amdgpu_vm_prt_put - add a PRT user
> + * amdgpu_vm_prt_get - add a PRT user
>   */
>  static void amdgpu_vm_prt_get(struct amdgpu_device *adev)  {
> +	if (!adev->gart.gart_funcs->set_prt)
> +		return;
> +
>  	if (atomic_inc_return(&adev->vm_manager.num_prt_users) == 1)
>  		amdgpu_vm_update_prt_state(adev);
>  }
> @@ -1228,9 +1231,12 @@ static void amdgpu_vm_prt_cb(struct fence *fence,
> struct fence_cb *_cb)  static void amdgpu_vm_add_prt_cb(struct
> amdgpu_device *adev,
>  				 struct fence *fence)
>  {
> -	struct amdgpu_prt_cb *cb = kmalloc(sizeof(struct amdgpu_prt_cb),
> -					   GFP_KERNEL);
> +	struct amdgpu_prt_cb *cb;
> 
> +	if (!adev->gart.gart_funcs->set_prt)
> +		return;
> +
> +	cb = kmalloc(sizeof(struct amdgpu_prt_cb), GFP_KERNEL);
>  	if (!cb) {
>  		/* Last resort when we are OOM */
>  		if (fence)
> @@ -1445,14 +1451,6 @@ int amdgpu_vm_bo_map(struct amdgpu_device
> *adev,
>  	    size == 0 || size & AMDGPU_GPU_PAGE_MASK)
>  		return -EINVAL;
> 
> -	if (flags & AMDGPU_PTE_PRT) {
> -		/* Check if we have PRT hardware support */
> -		if (!adev->gart.gart_funcs->set_prt)
> -			return -EINVAL;
> -
> -		amdgpu_vm_prt_get(adev);
> -	}
> -
>  	/* make sure object fit at this offset */
>  	eaddr = saddr + size - 1;
>  	if (saddr >= eaddr ||
> @@ -1533,6 +1531,9 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
>  		vm->page_tables[pt_idx].addr = 0;
>  	}
> 
> +	if (flags & AMDGPU_PTE_PRT)
> +		amdgpu_vm_prt_get(adev);
> +
>  	return 0;
> 
>  error_free:
> @@ -1753,7 +1754,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev,
> struct amdgpu_vm *vm)  void amdgpu_vm_fini(struct amdgpu_device *adev,
> struct amdgpu_vm *vm)  {
>  	struct amdgpu_bo_va_mapping *mapping, *tmp;
> -	bool prt_fini_called = false;
> +	bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt;
>  	int i;
> 
>  	amd_sched_entity_fini(vm->entity.sched, &vm->entity); @@ -1767,9
> +1768,9 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct
> amdgpu_vm *vm)
>  		kfree(mapping);
>  	}
>  	list_for_each_entry_safe(mapping, tmp, &vm->freed, list) {
> -		if (mapping->flags & AMDGPU_PTE_PRT && !prt_fini_called) {
> +		if (mapping->flags & AMDGPU_PTE_PRT && prt_fini_needed) {
>  			amdgpu_vm_prt_fini(adev, vm);
> -			prt_fini_called = true;
> +			prt_fini_needed = false;
>  		}
> 
>  		list_del(&mapping->list);
> --
> 2.7.4
> 
> _______________________________________________
> amd-gfx mailing list
> amd-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx