The backport of upstream patch 5dccdc5a1916 ("ext4: do not iput inode under running transaction in ext4_rename()") introduced a regression on the stable kernels 4.14 and older. One of the end_rename error label was forgetting to change to release_bh, which may trigger below bug.
------------[ cut here ]------------ kernel BUG at /home/zhangyi/hulk-4.4/fs/ext4/ext4_jbd2.c:30! ... Call Trace: [<ffffffff8b4207b2>] ext4_rename+0x9e2/0x10c0 [<ffffffff8b331324>] ? unlazy_walk+0x124/0x2a0 [<ffffffff8b420eb5>] ext4_rename2+0x25/0x60 [<ffffffff8b335104>] vfs_rename+0x3a4/0xed0 [<ffffffff8b33a7ad>] SYSC_renameat2+0x57d/0x7f0 [<ffffffff8b33c119>] SyS_renameat+0x19/0x30 [<ffffffff8bc57bb8>] entry_SYSCALL_64_fastpath+0x18/0x78 ... ---[ end trace 75346ce7c76b9f06 ]---
Fixes: d962f1b4ef54 ("ext4: do not iput inode under running transaction in ext4_rename()") Signed-off-by: Zhang Yi yi.zhang@huawei.com --- fs/ext4/namei.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c index b4ec5a41797b..83bdae81721d 100644 --- a/fs/ext4/namei.c +++ b/fs/ext4/namei.c @@ -3641,7 +3641,7 @@ static int ext4_rename(struct inode *old_dir, struct dentry *old_dentry, ext4_encrypted_inode(new.dir) && !fscrypt_has_permitted_context(new.dir, old.inode)) { retval = -EXDEV; - goto end_rename; + goto release_bh; }
new.bh = ext4_find_entry(new.dir, &new.dentry->d_name,