ext4: using nofail preallocation in ext4_es_remove_extent()

If __es_remove_extent() returns an error it means that when splitting
extent, allocating an extent that must be kept failed, where returning
an error directly would cause the extent tree to be inconsistent. So we
use GFP_NOFAIL to pre-allocate an extent_status and pass it to
__es_remove_extent() to avoid this problem.

In addition, since the allocated memory is outside the i_es_lock, the
extent_status tree may change and the pre-allocated extent_status is
no longer needed, so we release the pre-allocated extent_status when
es->es_len is not initialized.

Suggested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230424033846.4732-7-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
This commit is contained in:
Baokun Li 2023-04-24 11:38:40 +08:00 committed by Theodore Ts'o
parent bda3efaf77
commit e9fe2b882b

View File

@ -1454,6 +1454,7 @@ int ext4_es_remove_extent(struct inode *inode, ext4_lblk_t lblk,
ext4_lblk_t end;
int err = 0;
int reserved = 0;
struct extent_status *es = NULL;
if (EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
return 0;
@ -1468,17 +1469,25 @@ int ext4_es_remove_extent(struct inode *inode, ext4_lblk_t lblk,
end = lblk + len - 1;
BUG_ON(end < lblk);
retry:
if (err && !es)
es = __es_alloc_extent(true);
/*
* ext4_clear_inode() depends on us taking i_es_lock unconditionally
* so that we are sure __es_shrink() is done with the inode before it
* is reclaimed.
*/
write_lock(&EXT4_I(inode)->i_es_lock);
err = __es_remove_extent(inode, lblk, end, &reserved, NULL);
err = __es_remove_extent(inode, lblk, end, &reserved, es);
if (es && !es->es_len)
__es_free_extent(es);
write_unlock(&EXT4_I(inode)->i_es_lock);
if (err)
goto retry;
ext4_es_print_tree(inode);
ext4_da_release_space(inode, reserved);
return err;
return 0;
}
static int __es_shrink(struct ext4_sb_info *sbi, int nr_to_scan,