Btrfs: proper metadata -ENOSPC handling

index db02e26..003400a 100644
@@ -672,6 +672,7 @@ struct btrfs_space_info {
 	u64 bytes_reserved;	/* total bytes the allocator has reserved for
 				   current allocations */
 	u64 bytes_readonly;	/* total bytes that are read only */
+	u64 bytes_super;	/* total bytes reserved for the super blocks */
 
 	/* delalloc accounting */
 	u64 bytes_delalloc;	/* number of bytes reserved for allocation,

@@ -744,6 +745,7 @@ struct btrfs_block_group_cache {
 	spinlock_t lock;
 	u64 pinned;
 	u64 reserved;
+	u64 bytes_super;
 	u64 flags;
 	u64 sectorsize;
 	int extents_thresh;

@@ -2012,7 +2014,7 @@ 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_check_metadata_free_space(struct btrfs_root *root);
+u64 btrfs_metadata_free_space(struct btrfs_root *root);
 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,

index fafd6f2..dbd2f13 100644
@@ -1456,7 +1456,6 @@ static int cleaner_kthread(void *arg)
 static int transaction_kthread(void *arg)
 {
 	struct btrfs_root *root = arg;
-	struct btrfs_trans_handle *trans;
 	struct btrfs_transaction *cur;
 	unsigned long now;
 	unsigned long delay;

@@ -1485,8 +1484,7 @@ static int transaction_kthread(void *arg)
 			goto sleep;
 		}
 		mutex_unlock(&root->fs_info->trans_mutex);
-		trans = btrfs_start_transaction(root, 1);
-		ret = btrfs_commit_transaction(trans, root);
+		ret = btrfs_force_transaction_commit(root, 0);
 
 sleep:
 		wake_up_process(root->fs_info->cleaner_kthread);

@@ -1574,7 +1572,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	fs_info->btree_inode = new_inode(sb);
 	fs_info->btree_inode->i_ino = 1;
 	fs_info->btree_inode->i_nlink = 1;
-	fs_info->metadata_ratio = 8;
+	fs_info->metadata_ratio = 0;
 
 	fs_info->thread_pool_size = min_t(unsigned long,
 					  num_online_cpus() + 2, 8);

@@ -2285,18 +2283,15 @@ int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
 
 int btrfs_commit_super(struct btrfs_root *root)
 {
-	struct btrfs_trans_handle *trans;
 	int ret;
 
 	mutex_lock(&root->fs_info->cleaner_mutex);
 	btrfs_clean_old_snapshots(root);
 	mutex_unlock(&root->fs_info->cleaner_mutex);
-	trans = btrfs_start_transaction(root, 1);
-	ret = btrfs_commit_transaction(trans, root);
+	ret = btrfs_force_transaction_commit(root, 0);
 	BUG_ON(ret);
 	/* run commit again to drop the original snapshot */
-	trans = btrfs_start_transaction(root, 1);
-	btrfs_commit_transaction(trans, root);
+	btrfs_force_transaction_commit(root, 0);
 	ret = btrfs_write_and_wait_transaction(NULL, root);
 	BUG_ON(ret);
 

302 lines not on this page · raw message

index 7c3cd24..cabd0b2 100644
@@ -129,8 +129,8 @@ static noinline int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 
 	lock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS);
 	trans = btrfs_join_transaction(root, 1);
-	if (!trans) {
-		err = -ENOMEM;
+	if (IS_ERR(trans)) {
+		err = PTR_ERR(trans);
 		goto out_unlock;
 	}
 	btrfs_set_trans_block_group(trans, inode);

@@ -323,6 +323,10 @@ noinline int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 	while (1) {
 		recow = 0;
 		btrfs_release_path(root, path);
+		ret = btrfs_extend_transaction(trans, root, 1);
+		if (ret)
+			goto out;
+
 		ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
 					       search_start, -1);
 		if (ret < 0)

@@ -1080,6 +1084,10 @@ out_nolock:
 
 		if ((file->f_flags & O_SYNC) || IS_SYNC(inode)) {
 			trans = btrfs_start_transaction(root, 1);
+			if (IS_ERR(trans)) {
+				err = PTR_ERR(trans);
+				goto fail;
+			}
 			ret = btrfs_log_dentry_safe(trans, root,
 						    file->f_dentry);
 			if (ret == 0) {

@@ -1092,6 +1100,7 @@ out_nolock:
 				btrfs_commit_transaction(trans, root);
 			}
 		}
+fail:
 		if (file->f_flags & O_DIRECT) {
 			invalidate_mapping_pages(inode->i_mapping,
 			      start_pos >> PAGE_CACHE_SHIFT,

@@ -1169,8 +1178,8 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
 		btrfs_ioctl_trans_end(file);
 
 	trans = btrfs_start_transaction(root, 1);
-	if (!trans) {
-		ret = -ENOMEM;
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
 		goto out;
 	}
 

385 lines not on this page · raw message

index 9f4db84..e7acdd5 100644
@@ -240,12 +240,8 @@ static noinline int create_subvol(struct btrfs_root *root,
 	u64 index = 0;
 	unsigned long nr = 1;
 
-	ret = btrfs_check_metadata_free_space(root);
-	if (ret)
-		goto fail_commit;
-
-	trans = btrfs_start_transaction(root, 1);
-	BUG_ON(!trans);
+	trans = btrfs_start_transaction(root, 4);
+	BUG_ON(IS_ERR(trans));
 
 	ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
 				       0, &objectid);

@@ -376,10 +372,6 @@ static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
 	if (!root->ref_cows)
 		return -EINVAL;
 
-	ret = btrfs_check_metadata_free_space(root);
-	if (ret)
-		goto fail_unlock;
-
 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
 	if (!pending_snapshot) {
 		ret = -ENOMEM;

@@ -395,7 +387,7 @@ static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
 	pending_snapshot->name[namelen] = '\0';
 	pending_snapshot->dentry = dentry;
 	trans = btrfs_start_transaction(root, 1);
-	BUG_ON(!trans);
+	BUG_ON(IS_ERR(trans));
 	pending_snapshot->root = root;
 	list_add(&pending_snapshot->list,
 		 &trans->transaction->pending_snapshots);

118 lines not on this page · raw message

index 708ac06..cf21c1b 100644
@@ -407,6 +407,9 @@ int btrfs_sync_fs(struct super_block *sb, int wait)
 	btrfs_wait_ordered_extents(root, 0);
 
 	trans = btrfs_start_transaction(root, 1);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
+
 	ret = btrfs_commit_transaction(trans, root);
 	sb->s_dirt = 0;
 	return ret;

422 lines not on this page · raw message

index 663c674..06e9ba7 100644
@@ -30,6 +30,8 @@ struct btrfs_transaction {
 	unsigned long num_writers;
 
 	unsigned long num_joined;
+	u64 bytes_reserved;
+	u64 max_bytes;
 	int in_commit;
 	int use_count;
 	int commit_done;

@@ -45,13 +47,24 @@ struct btrfs_transaction {
 
 struct btrfs_trans_handle {
 	u64 transid;
-	unsigned long blocks_reserved;
+
+	/*
+	 * keep track of how many bytes are in use by this handle.  Ideally when
+	 * this all works perfectly bytes_reserved can take a hike, but until
+	 * then its usefull in making sure we are making the right guesses for
+	 * how many bytes we plan to use for different operations
+	 */
+	u64 bytes_reserved;
+	u64 bytes_used;
 	unsigned long blocks_used;
 	struct btrfs_transaction *transaction;
 	u64 block_group;
 	u64 alloc_exclude_start;
 	u64 alloc_exclude_nr;
 	unsigned long delayed_ref_updates;
+	int use_count;
+	bool is_relocate;	/* if true, this trans is being used to
+				 * relocate chunks */
 };
 
 struct btrfs_pending_snapshot {

@@ -84,11 +97,16 @@ static inline void btrfs_set_inode_last_trans(struct btrfs_trans_handle *trans,
 int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root);
 struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
-						   int num_blocks);
+						   int num_items);
 struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root,
-						   int num_blocks);
+						   int num_items);
 struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *r,
-						   int num_blocks);
+						   int num_items);
+int btrfs_try_extend_transaction(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root, int num_items);
+int btrfs_extend_transaction(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root, int num_items);
+void btrfs_trans_used_bytes(struct btrfs_trans_handle *trans, u64 num_bytes);
 int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root);
 int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,

@@ -98,6 +116,7 @@ int btrfs_add_dead_root(struct btrfs_root *root);
 int btrfs_drop_dead_root(struct btrfs_root *root);
 int btrfs_defrag_root(struct btrfs_root *root, int cacheonly);
 int btrfs_clean_old_snapshots(struct btrfs_root *root);
+int btrfs_force_transaction_commit(struct btrfs_root *root, int num_items);
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root);
 int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,

index 8d9c42e..b3430cc 100644
@@ -1743,10 +1743,18 @@ static int btrfs_relocate_chunk(struct btrfs_root *root,
 	ret = btrfs_relocate_block_group(extent_root, chunk_offset);
 	BUG_ON(ret);
 
-	trans = btrfs_start_transaction(root, 1);
-	BUG_ON(!trans);
+	/*
+	 * 2 - for one dev extent plus updating the device item, we'll have to
+	 * extend the transaction if there are more than one stripe
+	 * 1 - the chunk item
+	 * 1 - the block group item
+	 */
+	trans = btrfs_start_transaction(root, 4);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
 
 	lock_chunks(root);
+	trans->is_relocate = true;
 
 	/*
 	 * step two, delete the device extents and the

@@ -1761,6 +1769,11 @@ static int btrfs_relocate_chunk(struct btrfs_root *root,
 	map = (struct map_lookup *)em->bdev;
 
 	for (i = 0; i < map->num_stripes; i++) {
+		if (i > 0) {
+			/* 2 - 1 for the dev item plus the device item */
+			ret = btrfs_extend_transaction(trans, root, 2);
+			BUG_ON(ret);
+		}
 		ret = btrfs_free_dev_extent(trans, map->stripes[i].dev,
 					    map->stripes[i].physical);
 		BUG_ON(ret);