ext4: make sure to reset inode lockdep class when quota enabling fails
authorJan Kara <jack@suse.cz>
Thu, 7 Oct 2021 15:53:36 +0000 (17:53 +0200)
committerTheodore Ts'o <tytso@mit.edu>
Mon, 10 Jan 2022 18:25:55 +0000 (13:25 -0500)
When we succeed in enabling some quota type but fail to enable another
one with quota feature, we correctly disable all enabled quota types.
However we forget to reset i_data_sem lockdep class. When the inode gets
freed and reused, it will inherit this lockdep class (i_data_sem is
initialized only when a slab is created) and thus eventually lockdep
barfs about possible deadlocks.

Reported-and-tested-by: syzbot+3b6f9218b1301ddda3e2@syzkaller.appspotmail.com
Signed-off-by: Jan Kara <jack@suse.cz>
Cc: stable@kernel.org
Link: https://lore.kernel.org/r/20211007155336.12493-3-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
fs/ext4/super.c

index 863a3ea..1b55f23 100644 (file)
@@ -6837,8 +6837,19 @@ int ext4_enable_quotas(struct super_block *sb)
                                        "Failed to enable quota tracking "
                                        "(type=%d, err=%d). Please run "
                                        "e2fsck to fix.", type, err);
-                               for (type--; type >= 0; type--)
+                               for (type--; type >= 0; type--) {
+                                       struct inode *inode;
+
+                                       inode = sb_dqopt(sb)->files[type];
+                                       if (inode)
+                                               inode = igrab(inode);
                                        dquot_quota_off(sb, type);
+                                       if (inode) {
+                                               lockdep_set_quota_inode(inode,
+                                                       I_DATA_SEM_NORMAL);
+                                               iput(inode);
+                                       }
+                               }
 
                                return err;
                        }