drm/amdgpu: implement AMDGPU_VA_OP_CLEAR v2

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>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

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

@@ -507,14 +507,16 @@ 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)
@@ -529,16 +531,16 @@ 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;
 
@@ -592,6 +594,7 @@ 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",
@@ -600,7 +603,8 @@ 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;
@@ -625,8 +629,10 @@ 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) {
@@ -644,11 +650,18 @@ 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);

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

@@ -188,7 +188,7 @@ TRACE_EVENT(amdgpu_vm_bo_map,
 			     ),
 
 	    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;

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

@@ -1612,6 +1612,105 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 	return 0;
 }
 
+/**
+ * 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
  *

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);
 

include/uapi/drm/amdgpu_drm.h

@@ -350,6 +350,7 @@ 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)