Re: [syzbot] possible deadlock in jbd2_journal_lock_updates

[Jan Kara <jack@suse.cz>]

On Mon 08-08-22 09:38:26, syzbot wrote:
> syzbot has found a reproducer for the following issue on:
> 
> HEAD commit:    ca688bff68bc Add linux-next specific files for 20220808
> git tree:       linux-next
> console+strace: https://syzkaller.appspot.com/x/log.txt?x=1143ed46080000
> kernel config:  https://syzkaller.appspot.com/x/.config?x=4c20e006003cdecb
> dashboard link: https://syzkaller.appspot.com/bug?extid=79e6bbabf3ce17357969
> compiler:       gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2
> syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=1783be71080000
> C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=12bcdbfa080000
> 
> IMPORTANT: if you fix the issue, please add the following tag to the commit:
> Reported-by: syzbot+79e6bbabf3ce17357969@syzkaller.appspotmail.com
> 
> warning: checkpointing journal with EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT can be slow
> ======================================================
> WARNING: possible circular locking dependency detected
> 5.19.0-next-20220808-syzkaller #0 Not tainted
> ------------------------------------------------------
> syz-executor677/3614 is trying to acquire lock:
> ffff88814b8a4170 (&journal->j_barrier){+.+.}-{3:3}, at: jbd2_journal_lock_updates+0x15e/0x310 fs/jbd2/transaction.c:904
> 
> but task is already holding lock:
> ffff88814b8a2bd8 (&sbi->s_writepages_rwsem){++++}-{0:0}, at: ext4_change_inode_journal_flag+0x178/0x530 fs/ext4/inode.c:6013

So this seems like a real issue. Essentially, the problem is that
ext4_bmap() acquires inode->i_rwsem while its caller jbd2_journal_flush()
is holding journal->j_checkpoint_mutex. This looks like a real deadlock
possibility. This lock inversion was introduced by commit 51ae846cff5
("ext4: fix warning in ext4_iomap_begin as race between bmap and write")
which was fixing up a race in inline data handling.

Thinking about it some more, it does not seem locking i_rwsem in
ext4_bmap() is really workable and as I've noted in one of my replies to
this patch [1] it is not a complete solution to the problem anyway. So I
would be for reverting 51ae846cff5 and thinking more about how we can make
inline data locking suck less...

[1] https://lore.kernel.org/all/20220615172609.oydxhjw7exas23hd@quack3.lan/

								Honza

> which lock already depends on the new lock.
> 
> 
> the existing dependency chain (in reverse order) is:
> 
> -> #4 (&sbi->s_writepages_rwsem){++++}-{0:0}:
>        percpu_down_write+0x4d/0x440 kernel/locking/percpu-rwsem.c:221
>        ext4_change_inode_journal_flag+0x178/0x530 fs/ext4/inode.c:6013
>        ext4_ioctl_setflags fs/ext4/ioctl.c:685 [inline]
>        ext4_fileattr_set+0xddf/0x1930 fs/ext4/ioctl.c:1001
>        vfs_fileattr_set+0x7f5/0xbe0 fs/ioctl.c:696
>        ioctl_fssetxattr fs/ioctl.c:758 [inline]
>        do_vfs_ioctl+0xe62/0x15c0 fs/ioctl.c:845
>        __do_sys_ioctl fs/ioctl.c:868 [inline]
>        __se_sys_ioctl fs/ioctl.c:856 [inline]
>        __x64_sys_ioctl+0x108/0x200 fs/ioctl.c:856
>        do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>        do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
>        entry_SYSCALL_64_after_hwframe+0x63/0xcd
> 
> -> #3 (mapping.invalidate_lock){++++}-{3:3}:
>        down_write+0x90/0x150 kernel/locking/rwsem.c:1552
>        filemap_invalidate_lock include/linux/fs.h:806 [inline]
>        ext4_setattr+0xd9e/0x2c50 fs/ext4/inode.c:5445
>        notify_change+0xcd0/0x1440 fs/attr.c:418
>        do_truncate+0x13c/0x200 fs/open.c:65
>        do_sys_ftruncate+0x536/0x730 fs/open.c:193
>        do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>        do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
>        entry_SYSCALL_64_after_hwframe+0x63/0xcd
> 
> -> #2 (&sb->s_type->i_mutex_key#8){++++}-{3:3}:
>        down_read+0x98/0x450 kernel/locking/rwsem.c:1499
>        inode_lock_shared include/linux/fs.h:771 [inline]
>        ext4_bmap+0x4e/0x460 fs/ext4/inode.c:3157
>        bmap+0xaa/0x120 fs/inode.c:1799
>        jbd2_journal_bmap+0xa8/0x180 fs/jbd2/journal.c:971
>        __jbd2_journal_erase fs/jbd2/journal.c:1784 [inline]
>        jbd2_journal_flush+0x84f/0xc00 fs/jbd2/journal.c:2490
>        ext4_ioctl_checkpoint fs/ext4/ioctl.c:1082 [inline]
>        __ext4_ioctl+0x28fd/0x4ab0 fs/ext4/ioctl.c:1586
>        vfs_ioctl fs/ioctl.c:51 [inline]
>        __do_sys_ioctl fs/ioctl.c:870 [inline]
>        __se_sys_ioctl fs/ioctl.c:856 [inline]
>        __x64_sys_ioctl+0x193/0x200 fs/ioctl.c:856
>        do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>        do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
>        entry_SYSCALL_64_after_hwframe+0x63/0xcd
> 
> -> #1 (&journal->j_checkpoint_mutex){+.+.}-{3:3}:
>        __mutex_lock_common kernel/locking/mutex.c:603 [inline]
>        mutex_lock_io_nested+0x13f/0x1190 kernel/locking/mutex.c:833
>        jbd2_journal_flush+0x19a/0xc00 fs/jbd2/journal.c:2462
>        ext4_ioctl_checkpoint fs/ext4/ioctl.c:1082 [inline]
>        __ext4_ioctl+0x28fd/0x4ab0 fs/ext4/ioctl.c:1586
>        vfs_ioctl fs/ioctl.c:51 [inline]
>        __do_sys_ioctl fs/ioctl.c:870 [inline]
>        __se_sys_ioctl fs/ioctl.c:856 [inline]
>        __x64_sys_ioctl+0x193/0x200 fs/ioctl.c:856
>        do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>        do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
>        entry_SYSCALL_64_after_hwframe+0x63/0xcd
> 
> -> #0 (&journal->j_barrier){+.+.}-{3:3}:
>        check_prev_add kernel/locking/lockdep.c:3095 [inline]
>        check_prevs_add kernel/locking/lockdep.c:3214 [inline]
>        validate_chain kernel/locking/lockdep.c:3829 [inline]
>        __lock_acquire+0x2a43/0x56d0 kernel/locking/lockdep.c:5053
>        lock_acquire kernel/locking/lockdep.c:5666 [inline]
>        lock_acquire+0x1ab/0x570 kernel/locking/lockdep.c:5631
>        __mutex_lock_common kernel/locking/mutex.c:603 [inline]
>        __mutex_lock+0x12f/0x1350 kernel/locking/mutex.c:747
>        jbd2_journal_lock_updates+0x15e/0x310 fs/jbd2/transaction.c:904
>        ext4_change_inode_journal_flag+0x180/0x530 fs/ext4/inode.c:6014
>        ext4_ioctl_setflags fs/ext4/ioctl.c:685 [inline]
>        ext4_fileattr_set+0xddf/0x1930 fs/ext4/ioctl.c:1001
>        vfs_fileattr_set+0x7f5/0xbe0 fs/ioctl.c:696
>        ioctl_fssetxattr fs/ioctl.c:758 [inline]
>        do_vfs_ioctl+0xe62/0x15c0 fs/ioctl.c:845
>        __do_sys_ioctl fs/ioctl.c:868 [inline]
>        __se_sys_ioctl fs/ioctl.c:856 [inline]
>        __x64_sys_ioctl+0x108/0x200 fs/ioctl.c:856
>        do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>        do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
>        entry_SYSCALL_64_after_hwframe+0x63/0xcd
> 
> other info that might help us debug this:
> 
> Chain exists of:
>   &journal->j_barrier --> mapping.invalidate_lock --> &sbi->s_writepages_rwsem
> 
>  Possible unsafe locking scenario:
> 
>        CPU0                    CPU1
>        ----                    ----
>   lock(&sbi->s_writepages_rwsem);
>                                lock(mapping.invalidate_lock);
>                                lock(&sbi->s_writepages_rwsem);
>   lock(&journal->j_barrier);
> 
>  *** DEADLOCK ***
> 
> 4 locks held by syz-executor677/3614:
>  #0: ffff88814b8a0460 (sb_writers#4){.+.+}-{0:0}, at: ioctl_setflags fs/ioctl.c:725 [inline]
>  #0: ffff88814b8a0460 (sb_writers#4){.+.+}-{0:0}, at: do_vfs_ioctl+0xe27/0x15c0 fs/ioctl.c:839
>  #1: ffff88807599c030 (&sb->s_type->i_mutex_key#8){++++}-{3:3}, at: inode_lock include/linux/fs.h:761 [inline]
>  #1: ffff88807599c030 (&sb->s_type->i_mutex_key#8){++++}-{3:3}, at: vfs_fileattr_set+0x148/0xbe0 fs/ioctl.c:681
>  #2: ffff88807599c1d0 (mapping.invalidate_lock){++++}-{3:3}, at: filemap_invalidate_lock include/linux/fs.h:806 [inline]
>  #2: ffff88807599c1d0 (mapping.invalidate_lock){++++}-{3:3}, at: ext4_change_inode_journal_flag+0x11f/0x530 fs/ext4/inode.c:6005
>  #3: ffff88814b8a2bd8 (&sbi->s_writepages_rwsem){++++}-{0:0}, at: ext4_change_inode_journal_flag+0x178/0x530 fs/ext4/inode.c:6013
> 
> stack backtrace:
> CPU: 0 PID: 3614 Comm: syz-executor677 Not tainted 5.19.0-next-20220808-syzkaller #0
> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/22/2022
> Call Trace:
>  <TASK>
>  __dump_stack lib/dump_stack.c:88 [inline]
>  dump_stack_lvl+0xcd/0x134 lib/dump_stack.c:106
>  check_noncircular+0x25f/0x2e0 kernel/locking/lockdep.c:2175
>  check_prev_add kernel/locking/lockdep.c:3095 [inline]
>  check_prevs_add kernel/locking/lockdep.c:3214 [inline]
>  validate_chain kernel/locking/lockdep.c:3829 [inline]
>  __lock_acquire+0x2a43/0x56d0 kernel/locking/lockdep.c:5053
>  lock_acquire kernel/locking/lockdep.c:5666 [inline]
>  lock_acquire+0x1ab/0x570 kernel/locking/lockdep.c:5631
>  __mutex_lock_common kernel/locking/mutex.c:603 [inline]
>  __mutex_lock+0x12f/0x1350 kernel/locking/mutex.c:747
>  jbd2_journal_lock_updates+0x15e/0x310 fs/jbd2/transaction.c:904
>  ext4_change_inode_journal_flag+0x180/0x530 fs/ext4/inode.c:6014
>  ext4_ioctl_setflags fs/ext4/ioctl.c:685 [inline]
>  ext4_fileattr_set+0xddf/0x1930 fs/ext4/ioctl.c:1001
>  vfs_fileattr_set+0x7f5/0xbe0 fs/ioctl.c:696
>  ioctl_fssetxattr fs/ioctl.c:758 [inline]
>  do_vfs_ioctl+0xe62/0x15c0 fs/ioctl.c:845
>  __do_sys_ioctl fs/ioctl.c:868 [inline]
>  __se_sys_ioctl fs/ioctl.c:856 [inline]
>  __x64_sys_ioctl+0x108/0x200 fs/ioctl.c:856
>  do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>  do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
>  entry_SYSCALL_64_after_hwframe+0x63/0xcd
> RIP: 0033:0x7fb9f377eb39
> Code: 28 00 00 00 75 05 48 83 c4 28 c3 e8 e1 14 00 00 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 c0 ff ff ff f7 d8 64 89 01 48
> RSP: 002b:00007fffd29d1e18 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
> RAX: ffffffffffffffda RBX: 00007fffd29d1e58 RCX: 00007fb9f377eb39
> RDX: 0000000020000040 RSI: 0000000040086602 RDI: 0000000000000005
> RBP: 0000000000000000 R08: 431bde82d7b634db R09: 431bde82d7b634db
> R10: 431bde82d7b634db R11: 0000000000000246 R12: 00007fb9f3741e40
> R13: 431bde82d7b634db R14: 00007fffd29d1e50 R15: 0000000000000000
>  </TASK>
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR