@@ -33,6 +33,7 @@
struct btrfs_trans_handle;
struct btrfs_transaction;
+struct btrfs_pending_snapshot;
extern struct kmem_cache *btrfs_trans_handle_cachep;
extern struct kmem_cache *btrfs_transaction_cachep;
extern struct kmem_cache *btrfs_bit_radix_cachep;
@@ -2073,8 +2074,6 @@ u64 btrfs_reduce_alloc_profile(struct bt
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_reserve_metadata_space(struct btrfs_root *root, int num_items);
-int btrfs_unreserve_metadata_space(struct btrfs_root *root, int num_items);
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,
@@ -2087,6 +2086,13 @@ void btrfs_delalloc_reserve_space(struct
u64 bytes);
void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
u64 bytes);
+int btrfs_trans_reserve_metadata(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ int num_items, int *retries);
+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);
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_block_rsv *rsv);
@@ -1531,7 +1531,7 @@ static int transaction_kthread(void *arg
goto sleep;
}
mutex_unlock(&root->fs_info->trans_mutex);
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_join_transaction(root, 1);
ret = btrfs_commit_transaction(trans, root);
sleep:
@@ -2418,11 +2418,11 @@ int btrfs_commit_super(struct btrfs_root
down_write(&root->fs_info->cleanup_work_sem);
up_write(&root->fs_info->cleanup_work_sem);
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_join_transaction(root, 1);
ret = btrfs_commit_transaction(trans, root);
BUG_ON(ret);
/* run commit again to drop the original snapshot */
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_join_transaction(root, 1);
btrfs_commit_transaction(trans, root);
ret = btrfs_write_and_wait_transaction(NULL, root);
BUG_ON(ret);
@@ -2948,113 +2948,6 @@ again:
}
/*
- * unreserve num_items number of items worth of metadata space. This needs to
- * be paired with btrfs_reserve_metadata_space.
- *
- * NOTE: if you have the option, run this _AFTER_ you do a
- * btrfs_end_transaction, since btrfs_end_transaction will run delayed ref
- * oprations which will result in more used metadata, so we want to make sure we
- * can do that without issue.
- */
-int btrfs_unreserve_metadata_space(struct btrfs_root *root, int num_items)
-{
- struct btrfs_fs_info *info = root->fs_info;
- struct btrfs_space_info *meta_sinfo;
- u64 num_bytes;
- u64 alloc_target;
- bool bug = false;
-
- /* get the space info for where the metadata will live */
- alloc_target = btrfs_get_alloc_profile(root, 0);
- meta_sinfo = __find_space_info(info, alloc_target);
-
- num_bytes = calculate_bytes_needed(root, num_items);
-
- spin_lock(&meta_sinfo->lock);
- if (meta_sinfo->bytes_may_use < num_bytes) {
- bug = true;
- meta_sinfo->bytes_may_use = 0;
- } else {
- meta_sinfo->bytes_may_use -= num_bytes;
- }
- spin_unlock(&meta_sinfo->lock);
-
- BUG_ON(bug);
-
- return 0;
-}
-
-/*
- * Reserve some metadata space for use. We'll calculate the worste case number
- * of bytes that would be needed to modify num_items number of items. If we
- * have space, fantastic, if not, you get -ENOSPC. Please call
- * btrfs_unreserve_metadata_space when you are done for the _SAME_ number of
- * items you reserved, since whatever metadata you needed should have already
- * been allocated.
- *
- * This will commit the transaction to make more space if we don't have enough
- * metadata space. THe only time we don't do this is if we're reserving space
- * inside of a transaction, then we will just return -ENOSPC and it is the
- * callers responsibility to handle it properly.
- */
-int btrfs_reserve_metadata_space(struct btrfs_root *root, int num_items)
-{
- struct btrfs_fs_info *info = root->fs_info;
- struct btrfs_space_info *meta_sinfo;
- u64 num_bytes;
- u64 used;
- u64 alloc_target;
- int retries = 0;
-
- /* get the space info for where the metadata will live */
- alloc_target = btrfs_get_alloc_profile(root, 0);
- meta_sinfo = __find_space_info(info, alloc_target);
-
- num_bytes = calculate_bytes_needed(root, num_items);
-again:
- spin_lock(&meta_sinfo->lock);
-
- if (unlikely(!meta_sinfo->bytes_root))
- meta_sinfo->bytes_root = calculate_bytes_needed(root, 6);
-
- if (!retries)
- meta_sinfo->bytes_may_use += num_bytes;
-
- used = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
- meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
- meta_sinfo->bytes_super + meta_sinfo->bytes_root +
- meta_sinfo->bytes_may_use + meta_sinfo->bytes_delalloc;
-
- if (used > meta_sinfo->total_bytes) {
- retries++;
- if (retries == 1) {
- if (maybe_allocate_chunk(NULL, root, meta_sinfo,
- num_bytes))
- goto again;
- retries++;
- } else {
- spin_unlock(&meta_sinfo->lock);
- }
-
- if (retries == 2) {
- shrink_delalloc(NULL, root, meta_sinfo, num_bytes);
- goto again;
- }
- spin_lock(&meta_sinfo->lock);
- meta_sinfo->bytes_may_use -= num_bytes;
- spin_unlock(&meta_sinfo->lock);
-
- dump_space_info(meta_sinfo, 0, 0);
- return -ENOSPC;
- }
-
- check_force_delalloc(meta_sinfo);
- spin_unlock(&meta_sinfo->lock);
-
- return 0;
-}
-
-/*
* This will check the space that the inode allocates from to make sure we have
* enough space for bytes.
*/
@@ -3094,9 +2987,9 @@ again:
spin_unlock(&data_sinfo->lock);
alloc:
alloc_target = btrfs_get_alloc_profile(root, 1);
- trans = btrfs_start_transaction(root, 1);
- if (!trans)
- return -ENOMEM;
+ trans = btrfs_join_transaction(root, 1);
+ if (IS_ERR(trans))
+ return PTR_ERR(trans);
ret = do_chunk_alloc(trans, root->fs_info->extent_root,
bytes + 2 * 1024 * 1024,
@@ -3117,8 +3010,8 @@ alloc:
if (!committed && !root->fs_info->open_ioctl_trans) {
committed = 1;
trans = btrfs_join_transaction(root, 1);
- if (!trans)
- return -ENOMEM;
+ if (IS_ERR(trans))
+ return PTR_ERR(trans);
ret = btrfs_commit_transaction(trans, root);
if (ret)
return ret;
@@ -3680,6 +3573,58 @@ static void init_global_block_rsv(struct
fs_info->chunk_root->block_rsv = &fs_info->chunk_block_rsv;
}
+static u64 calc_trans_metadata_size(struct btrfs_root *root, int num_items)
+{
+ return (root->leafsize + root->nodesize * (BTRFS_MAX_LEVEL - 1)) *
+ 3 * num_items;
+}
+
+int btrfs_trans_reserve_metadata(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ int num_items, int *retries)
+{
+ u64 num_bytes;
+ int ret;
+
+ if (num_items == 0 || root->fs_info->chunk_root == root)
+ return 0;
+
+ num_bytes = calc_trans_metadata_size(root, num_items);
+ ret = btrfs_block_rsv_add(trans, root, &root->fs_info->trans_block_rsv,
+ num_bytes, retries);
+ if (!ret) {
+ trans->bytes_reserved += num_bytes;
+ trans->block_rsv = &root->fs_info->trans_block_rsv;
+ }
+ return ret;
+}
+
+void btrfs_trans_release_metadata(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root)
+{
+ if (!trans->bytes_reserved)
+ return;
+
+ BUG_ON(trans->block_rsv != &root->fs_info->trans_block_rsv);
+ block_rsv_release_bytes(trans->block_rsv, trans->bytes_reserved);
+ trans->bytes_reserved = 0;
+}
+
+int btrfs_snap_reserve_metadata(struct btrfs_trans_handle *trans,
+ struct btrfs_pending_snapshot *pending)
+{
+ struct btrfs_root *root = pending->root;
+ struct btrfs_block_rsv *src_rsv = get_block_rsv(trans, root);
+ struct btrfs_block_rsv *dst_rsv = &pending->block_rsv;
+ /*
+ * two for root back/forward refs, two for directory entries
+ * and one for root of the snapshot.
+ */
+ u64 num_bytes = calc_trans_metadata_size(root, 5);
+ dst_rsv->space_info = src_rsv->space_info;
+ return block_rsv_migrate_bytes(src_rsv, dst_rsv, num_bytes);
+}
+
static int update_block_group(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u64 num_bytes, int alloc)
@@ -5804,7 +5749,7 @@ int btrfs_drop_snapshot(struct btrfs_roo
wc = kzalloc(sizeof(*wc), GFP_NOFS);
BUG_ON(!wc);
- trans = btrfs_start_transaction(tree_root, 1);
+ trans = btrfs_start_transaction(tree_root, 0);
if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
level = btrfs_header_level(root->node);
@@ -5900,7 +5845,9 @@ int btrfs_drop_snapshot(struct btrfs_roo
BUG_ON(ret);
btrfs_end_transaction(trans, tree_root);
- trans = btrfs_start_transaction(tree_root, 1);
+ trans = btrfs_start_transaction(tree_root, 0);
+ if (IS_ERR(trans))
+ return PTR_ERR(trans);
} else {
unsigned long update;
update = trans->delayed_ref_updates;
@@ -125,8 +125,7 @@ static noinline int dirty_and_release_pa
end_of_last_block = start_pos + num_bytes - 1;
err = btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block,
NULL);
- if (err)
- return err;
+ BUG_ON(err);
for (i = 0; i < num_pages; i++) {
struct page *p = pages[i];
@@ -141,7 +140,7 @@ static noinline int dirty_and_release_pa
* at this time.
*/
}
- return err;
+ return 0;
}
/*
@@ -1007,7 +1006,7 @@ out_nolock:
num_written = err;
if ((file->f_flags & O_SYNC) || IS_SYNC(inode)) {
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_start_transaction(root, 0);
ret = btrfs_log_dentry_safe(trans, root,
file->f_dentry);
if (ret == 0) {
@@ -1103,7 +1102,7 @@ int btrfs_sync_file(struct file *file, s
if (file && file->private_data)
btrfs_ioctl_trans_end(file);
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_start_transaction(root, 0);
if (!trans) {
ret = -ENOMEM;
goto out;
@@ -2489,17 +2489,11 @@ static int btrfs_unlink(struct inode *di
/*
* 5 items for unlink inode
- * 1 for orphan
+ * 5 for orphan
*/
- ret = btrfs_reserve_metadata_space(root, 6);
- if (ret)
- return ret;
-
- trans = btrfs_start_transaction(root, 1);
- if (IS_ERR(trans)) {
- btrfs_unreserve_metadata_space(root, 6);
+ trans = btrfs_start_transaction(root, 10);
+ if (IS_ERR(trans))
return PTR_ERR(trans);
- }
btrfs_set_trans_block_group(trans, dir);
@@ -2507,14 +2501,16 @@ static int btrfs_unlink(struct inode *di
ret = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
dentry->d_name.name, dentry->d_name.len);
+ BUG_ON(ret);
- if (inode->i_nlink == 0)
+ if (inode->i_nlink == 0) {
ret = btrfs_orphan_add(trans, inode);
+ BUG_ON(ret);
+ }
nr = trans->blocks_used;
btrfs_end_transaction_throttle(trans, root);
- btrfs_unreserve_metadata_space(root, 6);
btrfs_btree_balance_dirty(root, nr);
return ret;
}
@@ -2595,15 +2591,9 @@ static int btrfs_rmdir(struct inode *dir
inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
return -ENOTEMPTY;
- ret = btrfs_reserve_metadata_space(root, 5);
- if (ret)
- return ret;
-
- trans = btrfs_start_transaction(root, 1);
- if (IS_ERR(trans)) {
- btrfs_unreserve_metadata_space(root, 5);
+ trans = btrfs_start_transaction(root, 10);
+ if (IS_ERR(trans))
return PTR_ERR(trans);
- }
btrfs_set_trans_block_group(trans, dir);
@@ -2627,7 +2617,6 @@ static int btrfs_rmdir(struct inode *dir
out:
nr = trans->blocks_used;
ret = btrfs_end_transaction_throttle(trans, root);
- btrfs_unreserve_metadata_space(root, 5);
btrfs_btree_balance_dirty(root, nr);
if (ret && !err)
@@ -3144,7 +3133,7 @@ int btrfs_cont_expand(struct inode *inod
struct btrfs_trans_handle *trans;
struct btrfs_root *root = BTRFS_I(inode)->root;
struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
- struct extent_map *em;
+ struct extent_map *em = NULL;
struct extent_state *cached_state = NULL;
u64 mask = root->sectorsize - 1;
u64 hole_start = (inode->i_size + mask) & ~mask;
@@ -3182,11 +3171,11 @@ int btrfs_cont_expand(struct inode *inod
u64 hint_byte = 0;
hole_size = last_byte - cur_offset;
- err = btrfs_reserve_metadata_space(root, 2);
- if (err)
+ trans = btrfs_start_transaction(root, 2);
+ if (IS_ERR(trans)) {
+ err = PTR_ERR(trans);
break;
-
- trans = btrfs_start_transaction(root, 1);
+ }
btrfs_set_trans_block_group(trans, inode);
err = btrfs_drop_extents(trans, inode, cur_offset,
@@ -3204,14 +3193,15 @@ int btrfs_cont_expand(struct inode *inod
last_byte - 1, 0);
btrfs_end_transaction(trans, root);
- btrfs_unreserve_metadata_space(root, 2);
}
free_extent_map(em);
+ em = NULL;
cur_offset = last_byte;
if (cur_offset >= block_end)
break;
}
+ free_extent_map(em);
unlock_extent_cached(io_tree, hole_start, block_end - 1, &cached_state,
GFP_NOFS);
return err;
@@ -3238,10 +3228,6 @@ static int btrfs_setattr_size(struct ino
}
}
- ret = btrfs_reserve_metadata_space(root, 1);
- if (ret)
- return ret;
-
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
@@ -3250,7 +3236,6 @@ static int btrfs_setattr_size(struct ino
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
- btrfs_unreserve_metadata_space(root, 1);
btrfs_btree_balance_dirty(root, nr);
if (attr->ia_size > inode->i_size) {
@@ -4222,26 +4207,21 @@ static int btrfs_mknod(struct inode *dir
if (!new_valid_dev(rdev))
return -EINVAL;
+ err = btrfs_find_free_objectid(NULL, root, dir->i_ino, &objectid);
+ if (err)
+ return err;
+
/*
* 2 for inode item and ref
* 2 for dir items
* 1 for xattr if selinux is on
*/
- err = btrfs_reserve_metadata_space(root, 5);
- if (err)
- return err;
+ trans = btrfs_start_transaction(root, 5);
+ if (IS_ERR(trans))
+ return PTR_ERR(trans);
- trans = btrfs_start_transaction(root, 1);
- if (!trans)
- goto fail;
btrfs_set_trans_block_group(trans, dir);
- err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
- if (err) {
- err = -ENOSPC;
- goto out_unlock;
- }
-
inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
dentry->d_name.len,
dentry->d_parent->d_inode->i_ino, objectid,
@@ -4270,13 +4250,11 @@ static int btrfs_mknod(struct inode *dir
out_unlock:
nr = trans->blocks_used;
btrfs_end_transaction_throttle(trans, root);
-fail:
- btrfs_unreserve_metadata_space(root, 5);
+ btrfs_btree_balance_dirty(root, nr);
if (drop_inode) {
inode_dec_link_count(inode);
iput(inode);
}
- btrfs_btree_balance_dirty(root, nr);
return err;
}
@@ -4286,32 +4264,26 @@ static int btrfs_create(struct inode *di
struct btrfs_trans_handle *trans;
struct btrfs_root *root = BTRFS_I(dir)->root;
struct inode *inode = NULL;
- int err;
int drop_inode = 0;
+ int err;
unsigned long nr = 0;
u64 objectid;
u64 index = 0;
+ err = btrfs_find_free_objectid(NULL, root, dir->i_ino, &objectid);
+ if (err)
+ return err;
/*
* 2 for inode item and ref
* 2 for dir items
* 1 for xattr if selinux is on
*/
- err = btrfs_reserve_metadata_space(root, 5);
- if (err)
- return err;
+ trans = btrfs_start_transaction(root, 5);
+ if (IS_ERR(trans))
+ return PTR_ERR(trans);
- trans = btrfs_start_transaction(root, 1);
- if (!trans)
- goto fail;
btrfs_set_trans_block_group(trans, dir);
- err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
- if (err) {
- err = -ENOSPC;
- goto out_unlock;
- }
-
inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
dentry->d_name.len,
dentry->d_parent->d_inode->i_ino,
@@ -4343,8 +4315,6 @@ static int btrfs_create(struct inode *di
out_unlock:
nr = trans->blocks_used;
btrfs_end_transaction_throttle(trans, root);
-fail:
- btrfs_unreserve_metadata_space(root, 5);
if (drop_inode) {
inode_dec_link_count(inode);
iput(inode);
@@ -4371,21 +4341,21 @@ static int btrfs_link(struct dentry *old
if (root->objectid != BTRFS_I(inode)->root->objectid)
return -EPERM;
- /*
- * 1 item for inode ref
- * 2 items for dir items
- */
- err = btrfs_reserve_metadata_space(root, 3);
- if (err)
- return err;
-
btrfs_inc_nlink(inode);
err = btrfs_set_inode_index(dir, &index);
if (err)
goto fail;
- trans = btrfs_start_transaction(root, 1);
+ /*
+ * 1 item for inode ref
+ * 2 items for dir items
+ */
+ trans = btrfs_start_transaction(root, 3);
+ if (IS_ERR(trans)) {
+ err = PTR_ERR(trans);
+ goto fail;
+ }
btrfs_set_trans_block_group(trans, dir);
atomic_inc(&inode->i_count);
@@ -4404,7 +4374,6 @@ static int btrfs_link(struct dentry *old
nr = trans->blocks_used;
btrfs_end_transaction_throttle(trans, root);
fail:
- btrfs_unreserve_metadata_space(root, 3);
if (drop_inode) {
inode_dec_link_count(inode);
iput(inode);
@@ -4424,28 +4393,20 @@ static int btrfs_mkdir(struct inode *dir
u64 index = 0;
unsigned long nr = 1;
+ err = btrfs_find_free_objectid(NULL, root, dir->i_ino, &objectid);
+ if (err)
+ return err;
+
/*
* 2 items for inode and ref
* 2 items for dir items
* 1 for xattr if selinux is on
*/
- err = btrfs_reserve_metadata_space(root, 5);
- if (err)
- return err;
-
- trans = btrfs_start_transaction(root, 1);
- if (!trans) {
- err = -ENOMEM;
- goto out_unlock;
- }
+ trans = btrfs_start_transaction(root, 5);
+ if (IS_ERR(trans))
+ return PTR_ERR(trans);
btrfs_set_trans_block_group(trans, dir);
- err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
- if (err) {
- err = -ENOSPC;
- goto out_fail;
- }
-
inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
dentry->d_name.len,
dentry->d_parent->d_inode->i_ino, objectid,
@@ -4485,9 +4446,6 @@ static int btrfs_mkdir(struct inode *dir
out_fail:
nr = trans->blocks_used;
btrfs_end_transaction_throttle(trans, root);
-
-out_unlock:
- btrfs_unreserve_metadata_space(root, 5);
if (drop_on_err)
iput(inode);
btrfs_btree_balance_dirty(root, nr);
@@ -5425,19 +5383,6 @@ static int btrfs_rename(struct inode *ol
if (S_ISDIR(old_inode->i_mode) && new_inode &&
new_inode->i_size > BTRFS_EMPTY_DIR_SIZE)
return -ENOTEMPTY;
-
- /*
- * We want to reserve the absolute worst case amount of items. So if
- * both inodes are subvols and we need to unlink them then that would
- * require 4 item modifications, but if they are both normal inodes it
- * would require 5 item modifications, so we'll assume their normal
- * inodes. So 5 * 2 is 10, plus 1 for the new link, so 11 total items
- * should cover the worst case number of items we'll modify.
- */
- ret = btrfs_reserve_metadata_space(root, 11);
- if (ret)
- return ret;
-
/*
* we're using rename to replace one file with another.
* and the replacement file is large. Start IO on it now so
@@ -5450,8 +5395,18 @@ static int btrfs_rename(struct inode *ol
/* close the racy window with snapshot create/destroy ioctl */
if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
down_read(&root->fs_info->subvol_sem);
+ /*
+ * We want to reserve the absolute worst case amount of items. So if
+ * both inodes are subvols and we need to unlink them then that would
+ * require 4 item modifications, but if they are both normal inodes it
+ * would require 5 item modifications, so we'll assume their normal
+ * inodes. So 5 * 2 is 10, plus 1 for the new link, so 11 total items
+ * should cover the worst case number of items we'll modify.
+ */
+ trans = btrfs_start_transaction(root, 20);
+ if (IS_ERR(trans))
+ return PTR_ERR(trans);
- trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, new_dir);
if (dest != root)
@@ -5550,7 +5505,6 @@ out_fail:
if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
up_read(&root->fs_info->subvol_sem);
- btrfs_unreserve_metadata_space(root, 11);
return ret;
}
@@ -5657,26 +5611,20 @@ static int btrfs_symlink(struct inode *d
if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
return -ENAMETOOLONG;
+ err = btrfs_find_free_objectid(NULL, root, dir->i_ino, &objectid);
+ if (err)
+ return err;
/*
* 2 items for inode item and ref
* 2 items for dir items
* 1 item for xattr if selinux is on
*/
- err = btrfs_reserve_metadata_space(root, 5);
- if (err)
- return err;
+ trans = btrfs_start_transaction(root, 5);
+ if (IS_ERR(trans))
+ return PTR_ERR(trans);
- trans = btrfs_start_transaction(root, 1);
- if (!trans)
- goto out_fail;
btrfs_set_trans_block_group(trans, dir);
- err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
- if (err) {
- err = -ENOSPC;
- goto out_unlock;
- }
-
inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
dentry->d_name.len,
dentry->d_parent->d_inode->i_ino, objectid,
@@ -5748,8 +5696,6 @@ static int btrfs_symlink(struct inode *d
out_unlock:
nr = trans->blocks_used;
btrfs_end_transaction_throttle(trans, root);
-out_fail:
- btrfs_unreserve_metadata_space(root, 5);
if (drop_inode) {
inode_dec_link_count(inode);
iput(inode);
@@ -5770,21 +5716,18 @@ static int prealloc_file_range(struct in
u64 i_size;
while (num_bytes > 0) {
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_start_transaction(root, 3);
+ if (IS_ERR(trans)) {
+ ret = PTR_ERR(trans);
+ break;
+ }
ret = btrfs_reserve_extent(trans, root, num_bytes,
root->sectorsize, 0, alloc_hint,
(u64)-1, &ins, 1);
if (ret) {
- WARN_ON(1);
- goto stop_trans;
- }
-
- ret = btrfs_reserve_metadata_space(root, 3);
- if (ret) {
- btrfs_free_reserved_extent(root, ins.objectid,
- ins.offset);
- goto stop_trans;
+ btrfs_end_transaction(trans, root);
+ break;
}
ret = insert_reserved_file_extent(trans, inode,
@@ -5818,14 +5761,8 @@ static int prealloc_file_range(struct in
BUG_ON(ret);
btrfs_end_transaction(trans, root);
- btrfs_unreserve_metadata_space(root, 3);
}
return ret;
-
-stop_trans:
- btrfs_end_transaction(trans, root);
- return ret;
-
}
static long btrfs_fallocate(struct inode *inode, int mode,
@@ -238,23 +238,19 @@ static noinline int create_subvol(struct
u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
u64 index = 0;
+ ret = btrfs_find_free_objectid(NULL, root->fs_info->tree_root,
+ 0, &objectid);
+ if (ret)
+ return ret;
/*
* 1 - inode item
* 2 - refs
* 1 - root item
* 2 - dir items
*/
- ret = btrfs_reserve_metadata_space(root, 6);
- if (ret)
- return ret;
-
- trans = btrfs_start_transaction(root, 1);
- BUG_ON(!trans);
-
- ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
- 0, &objectid);
- if (ret)
- goto fail;
+ trans = btrfs_start_transaction(root, 6);
+ if (IS_ERR(trans))
+ return PTR_ERR(trans);
leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
0, objectid, NULL, 0, 0, 0);
@@ -344,13 +340,10 @@ fail:
err = btrfs_commit_transaction(trans, root);
if (err && !ret)
ret = err;
-
- btrfs_unreserve_metadata_space(root, 6);
return ret;
}
-static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
- char *name, int namelen)
+static int create_snapshot(struct btrfs_root *root, struct dentry *dentry)
{
struct inode *inode;
struct btrfs_pending_snapshot *pending_snapshot;
@@ -360,40 +353,33 @@ static int create_snapshot(struct btrfs_
if (!root->ref_cows)
return -EINVAL;
- /*
- * 1 - inode item
- * 2 - refs
- * 1 - root item
- * 2 - dir items
- */
- ret = btrfs_reserve_metadata_space(root, 6);
- if (ret)
- goto fail;
-
pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
- if (!pending_snapshot) {
- ret = -ENOMEM;
- btrfs_unreserve_metadata_space(root, 6);
- goto fail;
- }
- pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
- if (!pending_snapshot->name) {
- ret = -ENOMEM;
- kfree(pending_snapshot);
- btrfs_unreserve_metadata_space(root, 6);
- goto fail;
- }
- memcpy(pending_snapshot->name, name, namelen);
- pending_snapshot->name[namelen] = '\0';
+ if (!pending_snapshot)
+ return -ENOMEM;
+
+ btrfs_init_block_rsv(&pending_snapshot->block_rsv);
pending_snapshot->dentry = dentry;
- trans = btrfs_start_transaction(root, 1);
- BUG_ON(!trans);
pending_snapshot->root = root;
+
+ trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
+ if (IS_ERR(trans)) {
+ ret = PTR_ERR(inode);
+ goto fail;
+ }
+
+ ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
+ BUG_ON(ret);
+
list_add(&pending_snapshot->list,
&trans->transaction->pending_snapshots);
- ret = btrfs_commit_transaction(trans, root);
+ ret = btrfs_commit_transaction(trans, root->fs_info->extent_root);
BUG_ON(ret);
- btrfs_unreserve_metadata_space(root, 6);
+
+ ret = pending_snapshot->error;
+ if (ret)
+ goto fail;
+
+ btrfs_orphan_cleanup(pending_snapshot->snap);
inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
if (IS_ERR(inode)) {
@@ -404,6 +390,7 @@ static int create_snapshot(struct btrfs_
d_instantiate(dentry, inode);
ret = 0;
fail:
+ kfree(pending_snapshot);
return ret;
}
@@ -455,8 +442,7 @@ static noinline int btrfs_mksubvol(struc
goto out_up_read;
if (snap_src) {
- error = create_snapshot(snap_src, dentry,
- name, namelen);
+ error = create_snapshot(snap_src, dentry);
} else {
error = create_subvol(BTRFS_I(dir)->root, dentry,
name, namelen);
@@ -810,7 +796,7 @@ static noinline int btrfs_ioctl_resize(s
device->name, (unsigned long long)new_size);
if (new_size > old_size) {
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_start_transaction(root, 0);
ret = btrfs_grow_device(trans, device, new_size);
btrfs_commit_transaction(trans, root);
} else {
@@ -1648,7 +1648,7 @@ static noinline_for_stack int merge_relo
}
if (level == 0 && rc->stage == UPDATE_DATA_PTRS) {
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_start_transaction(root, 0);
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &key, 0);
@@ -1674,7 +1674,7 @@ static noinline_for_stack int merge_relo
while (1) {
leaf = NULL;
replaced = 0;
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_start_transaction(root, 0);
max_level = level;
ret = walk_down_reloc_tree(reloc_root, path, &level);
@@ -1802,7 +1802,7 @@ static void merge_func(struct btrfs_work
merge_reloc_root(async->rc, root);
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_start_transaction(root, 0);
btrfs_update_reloc_root(trans, root);
btrfs_end_transaction(trans, root);
}
@@ -3296,11 +3296,11 @@ static noinline_for_stack int relocate_b
rc->create_reloc_root = 1;
set_reloc_control(rc);
- trans = btrfs_start_transaction(rc->extent_root, 1);
+ trans = btrfs_join_transaction(rc->extent_root, 1);
btrfs_commit_transaction(trans, rc->extent_root);
while (1) {
- trans = btrfs_start_transaction(rc->extent_root, 1);
+ trans = btrfs_start_transaction(rc->extent_root, 0);
ret = find_next_extent(trans, rc, path);
if (ret < 0)
@@ -3410,7 +3410,7 @@ static noinline_for_stack int relocate_b
smp_mb();
if (rc->extents_found > 0) {
- trans = btrfs_start_transaction(rc->extent_root, 1);
+ trans = btrfs_join_transaction(rc->extent_root, 1);
btrfs_commit_transaction(trans, rc->extent_root);
}
@@ -3419,7 +3419,7 @@ static noinline_for_stack int relocate_b
unset_reloc_control(rc);
/* get rid of pinned extents */
- trans = btrfs_start_transaction(rc->extent_root, 1);
+ trans = btrfs_join_transaction(rc->extent_root, 1);
btrfs_commit_transaction(trans, rc->extent_root);
return err;
@@ -3474,7 +3474,7 @@ static struct inode *create_reloc_inode(
if (IS_ERR(root))
return ERR_CAST(root);
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_start_transaction(root, 6);
BUG_ON(!trans);
err = btrfs_find_free_objectid(trans, root, objectid, &objectid);
@@ -3618,7 +3618,7 @@ static noinline_for_stack int mark_garba
struct btrfs_trans_handle *trans;
int ret;
- trans = btrfs_start_transaction(root->fs_info->tree_root, 1);
+ trans = btrfs_start_transaction(root->fs_info->tree_root, 0);
memset(&root->root_item.drop_progress, 0,
sizeof(root->root_item.drop_progress));
@@ -497,7 +497,7 @@ int btrfs_sync_fs(struct super_block *sb
btrfs_start_delalloc_inodes(root, 0);
btrfs_wait_ordered_extents(root, 0, 0);
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_start_transaction(root, 0);
ret = btrfs_commit_transaction(trans, root);
return ret;
}
@@ -164,53 +164,89 @@ enum btrfs_trans_type {
TRANS_USERSPACE,
};
+static int may_wait_transaction(struct btrfs_root *root, int type)
+{
+ if (!root->fs_info->log_root_recovering &&
+ ((type == TRANS_START && !root->fs_info->open_ioctl_trans) ||
+ type == TRANS_USERSPACE))
+ return 1;
+ return 0;
+}
+
static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
- int num_blocks, int type)
+ u64 num_items, int type)
{
- struct btrfs_trans_handle *h =
- kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
+ struct btrfs_trans_handle *h;
+ struct btrfs_transaction *cur_trans;
+ int retries = 0;
int ret;
+again:
+ h = kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
+ if (!h)
+ return ERR_PTR(-ENOMEM);
mutex_lock(&root->fs_info->trans_mutex);
- if (!root->fs_info->log_root_recovering &&
- ((type == TRANS_START && !root->fs_info->open_ioctl_trans) ||
- type == TRANS_USERSPACE))
+ if (may_wait_transaction(root, type))
wait_current_trans(root);
+
ret = join_transaction(root);
BUG_ON(ret);
- h->transid = root->fs_info->running_transaction->transid;
- h->transaction = root->fs_info->running_transaction;
- h->blocks_reserved = num_blocks;
+ cur_trans = root->fs_info->running_transaction;
+ cur_trans->use_count++;
+ mutex_unlock(&root->fs_info->trans_mutex);
+
+ h->transid = cur_trans->transid;
+ h->transaction = cur_trans;
h->blocks_used = 0;
h->block_group = 0;
+ h->bytes_reserved = 0;
h->delayed_ref_updates = 0;
h->block_rsv = NULL;
- if (!current->journal_info && type != TRANS_USERSPACE)
- current->journal_info = h;
+ smp_mb();
+ if (cur_trans->blocked && may_wait_transaction(root, type)) {
+ btrfs_commit_transaction(h, root);
+ goto again;
+ }
+
+ if (num_items > 0) {
+ ret = btrfs_trans_reserve_metadata(h, root, num_items,
+ &retries);
+ if (ret == -EAGAIN) {
+ btrfs_commit_transaction(h, root);
+ goto again;
+ }
+ if (ret < 0) {
+ btrfs_end_transaction(h, root);
+ return ERR_PTR(ret);
+ }
+ }
- root->fs_info->running_transaction->use_count++;
+ mutex_lock(&root->fs_info->trans_mutex);
record_root_in_trans(h, root);
mutex_unlock(&root->fs_info->trans_mutex);
+
+ if (!current->journal_info && type != TRANS_USERSPACE)
+ current->journal_info = h;
return h;
}
struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
- int num_blocks)
+ int num_items)
{
- return start_transaction(root, num_blocks, TRANS_START);
+ return start_transaction(root, num_items, TRANS_START);
}
struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root,
int num_blocks)
{
- return start_transaction(root, num_blocks, TRANS_JOIN);
+ return start_transaction(root, 0, TRANS_JOIN);
}
struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *r,
int num_blocks)
{
- return start_transaction(r, num_blocks, TRANS_USERSPACE);
+ return start_transaction(r, 0, TRANS_USERSPACE);
}
/* wait for a transaction commit to be fully complete */
@@ -311,6 +347,8 @@ static int __btrfs_end_transaction(struc
count++;
}
+ btrfs_trans_release_metadata(trans, root);
+
mutex_lock(&info->trans_mutex);
cur_trans = info->running_transaction;
WARN_ON(cur_trans != trans->transaction);
@@ -756,47 +794,49 @@ static noinline int create_pending_snaps
struct btrfs_root *root = pending->root;
struct btrfs_root *parent_root;
struct inode *parent_inode;
+ struct dentry *dentry;
struct extent_buffer *tmp;
struct extent_buffer *old;
int ret;
- u64 objectid;
- int namelen;
u64 index = 0;
-
- parent_inode = pending->dentry->d_parent->d_inode;
- parent_root = BTRFS_I(parent_inode)->root;
+ u64 objectid;
new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS);
if (!new_root_item) {
- ret = -ENOMEM;
+ pending->error = -ENOMEM;
goto fail;
}
+
ret = btrfs_find_free_objectid(trans, tree_root, 0, &objectid);
- if (ret)
+ if (ret) {
+ pending->error = ret;
goto fail;
+ }
key.objectid = objectid;
- /* record when the snapshot was created in key.offset */
- key.offset = trans->transid;
- btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
+ key.offset = (u64)-1;
+ key.type = BTRFS_ROOT_ITEM_KEY;
- memcpy(&pending->root_key, &key, sizeof(key));
- pending->root_key.offset = (u64)-1;
+ trans->block_rsv = &pending->block_rsv;
+ dentry = pending->dentry;
+ parent_inode = dentry->d_parent->d_inode;
+ parent_root = BTRFS_I(parent_inode)->root;
record_root_in_trans(trans, parent_root);
+
/*
* insert the directory item
*/
- namelen = strlen(pending->name);
ret = btrfs_set_inode_index(parent_inode, &index);
BUG_ON(ret);
ret = btrfs_insert_dir_item(trans, parent_root,
- pending->name, namelen,
- parent_inode->i_ino,
- &pending->root_key, BTRFS_FT_DIR, index);
+ dentry->d_name.name, dentry->d_name.len,
+ parent_inode->i_ino, &key,
+ BTRFS_FT_DIR, index);
BUG_ON(ret);
- btrfs_i_size_write(parent_inode, parent_inode->i_size + namelen * 2);
+ btrfs_i_size_write(parent_inode, parent_inode->i_size +
+ dentry->d_name.len * 2);
ret = btrfs_update_inode(trans, parent_root, parent_inode);
BUG_ON(ret);
@@ -813,22 +853,29 @@ static noinline int create_pending_snaps
free_extent_buffer(old);
btrfs_set_root_node(new_root_item, tmp);
- ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
- new_root_item);
- BUG_ON(ret);
+ /* record when the snapshot was created in key.offset */
+ key.offset = trans->transid;
+ ret = btrfs_insert_root(trans, tree_root, &key, new_root_item);
btrfs_tree_unlock(tmp);
free_extent_buffer(tmp);
+ BUG_ON(ret);
- ret = btrfs_add_root_ref(trans, parent_root->fs_info->tree_root,
- pending->root_key.objectid,
+ /*
+ * insert root back/forward references
+ */
+ ret = btrfs_add_root_ref(trans, tree_root, objectid,
parent_root->root_key.objectid,
- parent_inode->i_ino, index, pending->name,
- namelen);
+ parent_inode->i_ino, index,
+ dentry->d_name.name, dentry->d_name.len);
BUG_ON(ret);
+ key.offset = (u64)-1;
+ pending->snap = btrfs_read_fs_root_no_name(root->fs_info, &key);
+ BUG_ON(IS_ERR(pending->snap));
fail:
kfree(new_root_item);
- return ret;
+ btrfs_block_rsv_release(&pending->block_rsv, (u64)-1);
+ return 0;
}
/*
@@ -897,6 +944,8 @@ int btrfs_commit_transaction(struct btrf
ret = btrfs_run_delayed_refs(trans, root, 0);
BUG_ON(ret);
+ btrfs_trans_release_metadata(trans, root);
+
cur_trans = trans->transaction;
/*
* set the flushing flag so procs in this transaction have to
@@ -57,8 +57,11 @@ struct btrfs_trans_handle {
struct btrfs_pending_snapshot {
struct dentry *dentry;
struct btrfs_root *root;
- char *name;
- struct btrfs_key root_key;
+ struct btrfs_root *snap;
+ /* block reservation for the operation */
+ struct btrfs_block_rsv block_rsv;
+ /* extra metadata reseration for relocation */
+ int error;
struct list_head list;
};
@@ -85,11 +88,11 @@ static inline void btrfs_set_inode_last_
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_blocks);
struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *r,
- int num_blocks);
+ int num_blocks);
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,
@@ -1096,7 +1096,7 @@ static int btrfs_rm_dev_item(struct btrf
if (!path)
return -ENOMEM;
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_start_transaction(root, 0);
key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
key.type = BTRFS_DEV_ITEM_KEY;
key.offset = device->devid;
@@ -1485,7 +1485,7 @@ int btrfs_init_new_device(struct btrfs_r
goto error;
}
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_start_transaction(root, 0);
lock_chunks(root);
device->barriers = 1;
@@ -1750,9 +1750,10 @@ static int btrfs_relocate_chunk(struct b
/* step one, relocate all the extents inside this chunk */
ret = btrfs_relocate_block_group(extent_root, chunk_offset);
- BUG_ON(ret);
+ if (ret)
+ return ret;
- trans = btrfs_start_transaction(root, 1);
+ trans = btrfs_start_transaction(root, 0);
BUG_ON(!trans);
lock_chunks(root);
@@ -1924,7 +1925,7 @@ int btrfs_balance(struct btrfs_root *dev
break;
BUG_ON(ret);
- trans = btrfs_start_transaction(dev_root, 1);
+ trans = btrfs_start_transaction(dev_root, 0);
BUG_ON(!trans);
ret = btrfs_grow_device(trans, device, old_size);
@@ -2093,11 +2094,7 @@ again:
}
/* Shrinking succeeded, else we would be at "done". */
- trans = btrfs_start_transaction(root, 1);
- if (!trans) {
- ret = -ENOMEM;
- goto done;
- }
+ trans = btrfs_start_transaction(root, 0);
lock_chunks(root);
device->disk_total_bytes = new_size;
@@ -154,15 +154,10 @@ int __btrfs_setxattr(struct btrfs_trans_
if (trans)
return do_setxattr(trans, inode, name, value, size, flags);
- ret = btrfs_reserve_metadata_space(root, 2);
- if (ret)
- return ret;
+ trans = btrfs_start_transaction(root, 2);
+ if (IS_ERR(trans))
+ return PTR_ERR(trans);
- trans = btrfs_start_transaction(root, 1);
- if (!trans) {
- ret = -ENOMEM;
- goto out;
- }
btrfs_set_trans_block_group(trans, inode);
ret = do_setxattr(trans, inode, name, value, size, flags);
@@ -174,7 +169,6 @@ int __btrfs_setxattr(struct btrfs_trans_
BUG_ON(ret);
out:
btrfs_end_transaction_throttle(trans, root);
- btrfs_unreserve_metadata_space(root, 2);
return ret;
}