Btrfs: Update metadata reservation for delayed allocation

Introduce metadata reservation context for delayed allocation
and update various related functions.

This patch also introduces EXTENT_FIRST_DELALLOC control bit for
set/clear_extent_bit. It tells set/clear_bit_hook whether they
are processing the first extent_state with EXTENT_DELALLOC bit
set. This change is important if set/clear_extent_bit involves
multiple extent_state.
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

fs/btrfs/btrfs_inode.h

@@ -137,8 +137,8 @@ struct btrfs_inode {
 	 * of extent items we've reserved metadata for.
 	 */
 	spinlock_t accounting_lock;
+	atomic_t outstanding_extents;
 	int reserved_extents;
-	int outstanding_extents;
 
 	/*
 	 * ordered_data_close is set by truncate when a file that used

fs/btrfs/ctree.h

@@ -2079,19 +2079,8 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags);
 void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *ionde);
 void btrfs_clear_space_info_full(struct btrfs_fs_info *info);
-
-int btrfs_unreserve_metadata_for_delalloc(struct btrfs_root *root,
-					  struct inode *inode, int num_items);
-int btrfs_reserve_metadata_for_delalloc(struct btrfs_root *root,
-					struct inode *inode, int num_items);
-int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode,
-				u64 bytes);
-void btrfs_free_reserved_data_space(struct btrfs_root *root,
-				    struct inode *inode, u64 bytes);
-void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode,
-				 u64 bytes);
-void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
-			      u64 bytes);
+int btrfs_check_data_free_space(struct inode *inode, u64 bytes);
+void btrfs_free_reserved_data_space(struct inode *inode, u64 bytes);
 int btrfs_trans_reserve_metadata(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
 				int num_items, int *retries);
@@ -2099,6 +2088,10 @@ void btrfs_trans_release_metadata(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root);
 int btrfs_snap_reserve_metadata(struct btrfs_trans_handle *trans,
 				struct btrfs_pending_snapshot *pending);
+int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes);
+void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes);
+int btrfs_delalloc_reserve_space(struct inode *inode, u64 num_bytes);
+void btrfs_delalloc_release_space(struct inode *inode, u64 num_bytes);
 void btrfs_init_block_rsv(struct btrfs_block_rsv *rsv);
 struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_root *root);
 void btrfs_free_block_rsv(struct btrfs_root *root,

fs/btrfs/extent_io.c

@@ -335,21 +335,18 @@ static int merge_state(struct extent_io_tree *tree,
 }
 
 static int set_state_cb(struct extent_io_tree *tree,
-			 struct extent_state *state,
-			 unsigned long bits)
+			 struct extent_state *state, int *bits)
 {
 	if (tree->ops && tree->ops->set_bit_hook) {
 		return tree->ops->set_bit_hook(tree->mapping->host,
-					       state->start, state->end,
-					       state->state, bits);
+					       state, bits);
 	}
 
 	return 0;
 }
 
 static void clear_state_cb(struct extent_io_tree *tree,
-			   struct extent_state *state,
-			   unsigned long bits)
+			   struct extent_state *state, int *bits)
 {
 	if (tree->ops && tree->ops->clear_bit_hook)
 		tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
@@ -367,9 +364,10 @@ static void clear_state_cb(struct extent_io_tree *tree,
  */
 static int insert_state(struct extent_io_tree *tree,
 			struct extent_state *state, u64 start, u64 end,
-			int bits)
+			int *bits)
 {
 	struct rb_node *node;
+	int bits_to_set = *bits & ~EXTENT_CTLBITS;
 	int ret;
 
 	if (end < start) {
@@ -384,9 +382,9 @@ static int insert_state(struct extent_io_tree *tree,
 	if (ret)
 		return ret;
 
-	if (bits & EXTENT_DIRTY)
+	if (bits_to_set & EXTENT_DIRTY)
 		tree->dirty_bytes += end - start + 1;
-	state->state |= bits;
+	state->state |= bits_to_set;
 	node = tree_insert(&tree->state, end, &state->rb_node);
 	if (node) {
 		struct extent_state *found;
@@ -456,13 +454,13 @@ static int split_state(struct extent_io_tree *tree, struct extent_state *orig,
  * struct is freed and removed from the tree
  */
 static int clear_state_bit(struct extent_io_tree *tree,
-			    struct extent_state *state, int bits, int wake,
-			    int delete)
+			    struct extent_state *state,
+			    int *bits, int wake)
 {
-	int bits_to_clear = bits & ~EXTENT_DO_ACCOUNTING;
+	int bits_to_clear = *bits & ~EXTENT_CTLBITS;
 	int ret = state->state & bits_to_clear;
 
-	if ((bits & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
+	if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
 		u64 range = state->end - state->start + 1;
 		WARN_ON(range > tree->dirty_bytes);
 		tree->dirty_bytes -= range;
@@ -471,9 +469,8 @@ static int clear_state_bit(struct extent_io_tree *tree,
 	state->state &= ~bits_to_clear;
 	if (wake)
 		wake_up(&state->wq);
-	if (delete || state->state == 0) {
+	if (state->state == 0) {
 		if (state->tree) {
-			clear_state_cb(tree, state, state->state);
 			rb_erase(&state->rb_node, &tree->state);
 			state->tree = NULL;
 			free_extent_state(state);
@@ -514,6 +511,10 @@ int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
 	int set = 0;
 	int clear = 0;
 
+	if (delete)
+		bits |= ~EXTENT_CTLBITS;
+	bits |= EXTENT_FIRST_DELALLOC;
+
 	if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
 		clear = 1;
 again:
@@ -580,8 +581,7 @@ int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
 		if (err)
 			goto out;
 		if (state->end <= end) {
-			set |= clear_state_bit(tree, state, bits, wake,
-					       delete);
+			set |= clear_state_bit(tree, state, &bits, wake);
 			if (last_end == (u64)-1)
 				goto out;
 			start = last_end + 1;
@@ -602,7 +602,7 @@ int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
 		if (wake)
 			wake_up(&state->wq);
 
-		set |= clear_state_bit(tree, prealloc, bits, wake, delete);
+		set |= clear_state_bit(tree, prealloc, &bits, wake);
 
 		prealloc = NULL;
 		goto out;
@@ -613,7 +613,7 @@ int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
 	else
 		next_node = NULL;
 
-	set |= clear_state_bit(tree, state, bits, wake, delete);
+	set |= clear_state_bit(tree, state, &bits, wake);
 	if (last_end == (u64)-1)
 		goto out;
 	start = last_end + 1;
@@ -706,19 +706,19 @@ int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits)
 
 static int set_state_bits(struct extent_io_tree *tree,
 			   struct extent_state *state,
-			   int bits)
+			   int *bits)
 {
 	int ret;
+	int bits_to_set = *bits & ~EXTENT_CTLBITS;
 
 	ret = set_state_cb(tree, state, bits);
 	if (ret)
 		return ret;
-
-	if ((bits & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
+	if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
 		u64 range = state->end - state->start + 1;
 		tree->dirty_bytes += range;
 	}
-	state->state |= bits;
+	state->state |= bits_to_set;
 
 	return 0;
 }
@@ -757,6 +757,7 @@ static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
 	u64 last_start;
 	u64 last_end;
 
+	bits |= EXTENT_FIRST_DELALLOC;
 again:
 	if (!prealloc && (mask & __GFP_WAIT)) {
 		prealloc = alloc_extent_state(mask);
@@ -778,7 +779,7 @@ static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
 	 */
 	node = tree_search(tree, start);
 	if (!node) {
-		err = insert_state(tree, prealloc, start, end, bits);
+		err = insert_state(tree, prealloc, start, end, &bits);
 		prealloc = NULL;
 		BUG_ON(err == -EEXIST);
 		goto out;
@@ -802,7 +803,7 @@ static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
 			goto out;
 		}
 
-		err = set_state_bits(tree, state, bits);
+		err = set_state_bits(tree, state, &bits);
 		if (err)
 			goto out;
 
@@ -852,7 +853,7 @@ static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
 		if (err)
 			goto out;
 		if (state->end <= end) {
-			err = set_state_bits(tree, state, bits);
+			err = set_state_bits(tree, state, &bits);
 			if (err)
 				goto out;
 			cache_state(state, cached_state);
@@ -877,7 +878,7 @@ static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
 		else
 			this_end = last_start - 1;
 		err = insert_state(tree, prealloc, start, this_end,
-				   bits);
+				   &bits);
 		BUG_ON(err == -EEXIST);
 		if (err) {
 			prealloc = NULL;
@@ -903,7 +904,7 @@ static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
 		err = split_state(tree, state, prealloc, end + 1);
 		BUG_ON(err == -EEXIST);
 
-		err = set_state_bits(tree, prealloc, bits);
+		err = set_state_bits(tree, prealloc, &bits);
 		if (err) {
 			prealloc = NULL;
 			goto out;
@@ -966,8 +967,7 @@ int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
 {
 	return clear_extent_bit(tree, start, end,
 				EXTENT_DIRTY | EXTENT_DELALLOC |
-				EXTENT_DO_ACCOUNTING, 0, 0,
-				NULL, mask);
+				EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask);
 }
 
 int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
@@ -1435,9 +1435,6 @@ int extent_clear_unlock_delalloc(struct inode *inode,
 	if (op & EXTENT_CLEAR_DELALLOC)
 		clear_bits |= EXTENT_DELALLOC;
 
-	if (op & EXTENT_CLEAR_ACCOUNTING)
-		clear_bits |= EXTENT_DO_ACCOUNTING;
-
 	clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
 	if (!(op & (EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
 		    EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK |

fs/btrfs/extent_io.h

@@ -16,7 +16,9 @@
 #define EXTENT_BOUNDARY (1 << 9)
 #define EXTENT_NODATASUM (1 << 10)
 #define EXTENT_DO_ACCOUNTING (1 << 11)
+#define EXTENT_FIRST_DELALLOC (1 << 12)
 #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
+#define EXTENT_CTLBITS (EXTENT_DO_ACCOUNTING | EXTENT_FIRST_DELALLOC)
 
 /* flags for bio submission */
 #define EXTENT_BIO_COMPRESSED 1
@@ -69,10 +71,10 @@ struct extent_io_ops {
 				    struct extent_state *state);
 	int (*writepage_end_io_hook)(struct page *page, u64 start, u64 end,
 				      struct extent_state *state, int uptodate);
-	int (*set_bit_hook)(struct inode *inode, u64 start, u64 end,
-			    unsigned long old, unsigned long bits);
+	int (*set_bit_hook)(struct inode *inode, struct extent_state *state,
+			    int *bits);
 	int (*clear_bit_hook)(struct inode *inode, struct extent_state *state,
-			      unsigned long bits);
+			      int *bits);
 	int (*merge_extent_hook)(struct inode *inode,
 				 struct extent_state *new,
 				 struct extent_state *other);

fs/btrfs/file.c

@@ -852,13 +852,6 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 
 	vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
 
-	/* do the reserve before the mutex lock in case we have to do some
-	 * flushing.  We wouldn't deadlock, but this is more polite.
-	 */
-	err = btrfs_reserve_metadata_for_delalloc(root, inode, 1);
-	if (err)
-		goto out_nolock;
-
 	mutex_lock(&inode->i_mutex);
 
 	current->backing_dev_info = inode->i_mapping->backing_dev_info;
@@ -921,7 +914,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 		WARN_ON(num_pages > nrptrs);
 		memset(pages, 0, sizeof(struct page *) * nrptrs);
 
-		ret = btrfs_check_data_free_space(root, inode, write_bytes);
+		ret = btrfs_delalloc_reserve_space(inode, write_bytes);
 		if (ret)
 			goto out;
 
@@ -929,26 +922,20 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 				    pos, first_index, last_index,
 				    write_bytes);
 		if (ret) {
-			btrfs_free_reserved_data_space(root, inode,
-						       write_bytes);
+			btrfs_delalloc_release_space(inode, write_bytes);
 			goto out;
 		}
 
 		ret = btrfs_copy_from_user(pos, num_pages,
 					   write_bytes, pages, buf);
-		if (ret) {
-			btrfs_free_reserved_data_space(root, inode,
-						       write_bytes);
-			btrfs_drop_pages(pages, num_pages);
-			goto out;
+		if (ret == 0) {
+			dirty_and_release_pages(NULL, root, file, pages,
+						num_pages, pos, write_bytes);
 		}
 
-		ret = dirty_and_release_pages(NULL, root, file, pages,
-					      num_pages, pos, write_bytes);
 		btrfs_drop_pages(pages, num_pages);
 		if (ret) {
-			btrfs_free_reserved_data_space(root, inode,
-						       write_bytes);
+			btrfs_delalloc_release_space(inode, write_bytes);
 			goto out;
 		}
 
@@ -975,9 +962,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 	mutex_unlock(&inode->i_mutex);
 	if (ret)
 		err = ret;
-	btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
 
-out_nolock:
 	kfree(pages);
 	if (pinned[0])
 		page_cache_release(pinned[0]);

fs/btrfs/ioctl.c

@@ -587,19 +587,9 @@ static int btrfs_defrag_file(struct file *file,
 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
 			BTRFS_I(inode)->force_compress = 1;
 
-		ret = btrfs_check_data_free_space(root, inode, PAGE_CACHE_SIZE);
-		if (ret) {
-			ret = -ENOSPC;
-			break;
-		}
-
-		ret = btrfs_reserve_metadata_for_delalloc(root, inode, 1);
-		if (ret) {
-			btrfs_free_reserved_data_space(root, inode,
-						       PAGE_CACHE_SIZE);
-			ret = -ENOSPC;
-			break;
-		}
+		ret  = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
+		if (ret)
+			goto err_unlock;
 again:
 		if (inode->i_size == 0 ||
 		    i > ((inode->i_size - 1) >> PAGE_CACHE_SHIFT)) {
@@ -608,8 +598,10 @@ static int btrfs_defrag_file(struct file *file,
 		}
 
 		page = grab_cache_page(inode->i_mapping, i);
-		if (!page)
+		if (!page) {
+			ret = -ENOMEM;
 			goto err_reservations;
+		}
 
 		if (!PageUptodate(page)) {
 			btrfs_readpage(NULL, page);
@@ -617,6 +609,7 @@ static int btrfs_defrag_file(struct file *file,
 			if (!PageUptodate(page)) {
 				unlock_page(page);
 				page_cache_release(page);
+				ret = -EIO;
 				goto err_reservations;
 			}
 		}
@@ -630,8 +623,7 @@ static int btrfs_defrag_file(struct file *file,
 		wait_on_page_writeback(page);
 
 		if (PageDirty(page)) {
-			btrfs_free_reserved_data_space(root, inode,
-						       PAGE_CACHE_SIZE);
+			btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
 			goto loop_unlock;
 		}
 
@@ -669,7 +661,6 @@ static int btrfs_defrag_file(struct file *file,
 		page_cache_release(page);
 		mutex_unlock(&inode->i_mutex);
 
-		btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
 		i++;
 	}
@@ -699,9 +690,9 @@ static int btrfs_defrag_file(struct file *file,
 	return 0;
 
 err_reservations:
+	btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
+err_unlock:
 	mutex_unlock(&inode->i_mutex);
-	btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE);
-	btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
 	return ret;
 }
 

fs/btrfs/ordered-data.c

@@ -311,13 +311,6 @@ static int __btrfs_remove_ordered_extent(struct inode *inode,
 	tree->last = NULL;
 	set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags);
 
-	spin_lock(&BTRFS_I(inode)->accounting_lock);
-	WARN_ON(!BTRFS_I(inode)->outstanding_extents);
-	BTRFS_I(inode)->outstanding_extents--;
-	spin_unlock(&BTRFS_I(inode)->accounting_lock);
-	btrfs_unreserve_metadata_for_delalloc(BTRFS_I(inode)->root,
-					      inode, 1);
-
 	spin_lock(&root->fs_info->ordered_extent_lock);
 	list_del_init(&entry->root_extent_list);