A crafted ext2 image can provide a directory entry pointing to an inode with i_links_count == 0 on disk. Calling unlink() on such an entry triggers WARN_ON inside drop_nlink(): WARNING: CPU: 3 PID: 609 at fs/inode.c:336 drop_nlink+0xad/0xd0 fs/inode.c:336 CPU: 3 UID: 0 PID: 609 Comm: syz-executor Not tainted 6.12.77+ #1 Call Trace: inode_dec_link_count include/linux/fs.h:2518 [inline] ext2_unlink+0x26c/0x300 fs/ext2/namei.c:295 vfs_unlink+0x2fc/0x9b0 fs/namei.c:4477 do_unlinkat+0x53e/0x730 fs/namei.c:4541 __do_sys_unlink fs/namei.c:4589 [inline] __se_sys_unlink fs/namei.c:4587 [inline] __x64_sys_unlink+0xc6/0x110 fs/namei.c:4587 do_syscall_x64 arch/x86/entry/common.c:47 [inline] do_syscall_64+0xf5/0x220 arch/x86/entry/common.c:78 entry_SYSCALL_64_after_hwframe+0x77/0x7f At the point of the crash, ext2_delete_entry() has already committed the removal of the directory entry to disk, so returning an error is not an option. Instead, skip the decrement and report the corruption via ext2_error(), which marks the superblock as having errors. The inode will be reclaimed when its last reference is dropped. Found by Linux Verification Center (linuxtesting.org) with Syzkaller. Cc: stable@vger.kernel.org Fixes: a513b035eadf ("[PATCH] ext2: switch to inode_inc_count, inode_dec_count") Signed-off-by: Vasiliy Kovalev --- fs/ext2/namei.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/fs/ext2/namei.c b/fs/ext2/namei.c index bde617a66cec..ea49e8f2b292 100644 --- a/fs/ext2/namei.c +++ b/fs/ext2/namei.c @@ -293,7 +293,12 @@ static int ext2_unlink(struct inode *dir, struct dentry *dentry) goto out; inode_set_ctime_to_ts(inode, inode_get_ctime(dir)); - inode_dec_link_count(inode); + if (!inode->i_nlink) + ext2_error(inode->i_sb, __func__, + "inode %lu has zero i_nlink on unlink, fs may be corrupt", + inode->i_ino); + else + inode_dec_link_count(inode); err = 0; out: return err; -- 2.50.1