[syzbot] possible deadlock in jbd2_journal_lock_updates

[syzbot]

Hello,

syzbot found the following issue on:

HEAD commit: ca688bff68bc Add linux-next specific files for 20220808
git tree: linux-next
console output: https://syzkaller.appspot.com/x/log.txt?x=1228affe080000
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

Unfortunately, I don't have any reproducer for this issue yet.

IMPORTANT: if you fix the issue, please add the following tag to the commit:
Reported-by: syzbot+79e6bb...@syzkaller.appspotmail.com

======================================================
WARNING: possible circular locking dependency detected
5.19.0-next-20220808-syzkaller #0 Not tainted
------------------------------------------------------
syz-executor.2/10446 is trying to acquire lock:
ffff88807feb0170 (&journal->j_barrier){+.+.}-{3:3}, at: jbd2_journal_lock_updates+0x15e/0x310 fs/jbd2/transaction.c:904

but task is already holding lock:
ffff88807ff5ebd8 (&sbi->s_writepages_rwsem){++++}-{0:0}, at: ext4_change_inode_journal_flag+0x178/0x530 fs/ext4/inode.c:6013

which lock already depends on the new lock.


the existing dependency chain (in reverse order) is:

-> #3 (&sbi->s_writepages_rwsem){++++}-{0:0}:
percpu_down_write+0x4d/0x440 kernel/locking/percpu-rwsem.c:221
ext4_ind_migrate+0x237/0x840 fs/ext4/migrate.c:624
ext4_ioctl_setflags fs/ext4/ioctl.c:694 [inline]
ext4_fileattr_set+0x14b8/0x1930 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

-> #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 --> &sb->s_type->i_mutex_key#8 --> &sbi->s_writepages_rwsem

Possible unsafe locking scenario:

CPU0 CPU1
---- ----
lock(&sbi->s_writepages_rwsem);
lock(&sb->s_type->i_mutex_key#8);
lock(&sbi->s_writepages_rwsem);
lock(&journal->j_barrier);

*** DEADLOCK ***

4 locks held by syz-executor.2/10446:
#0: ffff88807ff5c460 (sb_writers#4){.+.+}-{0:0}, at: ioctl_setflags fs/ioctl.c:725 [inline]
#0: ffff88807ff5c460 (sb_writers#4){.+.+}-{0:0}, at: do_vfs_ioctl+0xe27/0x15c0 fs/ioctl.c:839
#1: ffff88802b44de48 (&sb->s_type->i_mutex_key#8){++++}-{3:3}, at: inode_lock include/linux/fs.h:761 [inline]
#1: ffff88802b44de48 (&sb->s_type->i_mutex_key#8){++++}-{3:3}, at: vfs_fileattr_set+0x148/0xbe0 fs/ioctl.c:681
#2: ffff88802b44dfe8 (mapping.invalidate_lock){++++}-{3:3}, at: filemap_invalidate_lock include/linux/fs.h:806 [inline]
#2: ffff88802b44dfe8 (mapping.invalidate_lock){++++}-{3:3}, at: ext4_change_inode_journal_flag+0x11f/0x530 fs/ext4/inode.c:6005
#3: ffff88807ff5ebd8 (&sbi->s_writepages_rwsem){++++}-{0:0}, at: ext4_change_inode_journal_flag+0x178/0x530 fs/ext4/inode.c:6013

stack backtrace:
CPU: 0 PID: 10446 Comm: syz-executor.2 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: 0033:0x7f2195489279
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 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 b8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f219651b168 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
RAX: ffffffffffffffda RBX: 00007f219559bf80 RCX: 00007f2195489279
RDX: 0000000020000380 RSI: 0000000040086602 RDI: 0000000000000003
RBP: 00007f21954e3189 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 00007ffd08022a2f R14: 00007f219651b300 R15: 0000000000022000
</TASK>


---
This report is generated by a bot. It may contain errors.
See https://goo.gl/tpsmEJ for more information about syzbot.
syzbot engineers can be reached at syzk...@googlegroups.com.

syzbot will keep track of this issue. See:
https://goo.gl/tpsmEJ#status for how to communicate with syzbot.

[syzbot]

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+79e6bb...@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

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

&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

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>

[Hillf Danton]

On Mon, 08 Aug 2022 09:38:26 -0700

> 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

In ext4_bmap() replace inode lock with i_bmap_sem to serialize bmap path
with write_iter path.

#syz test https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git ca688bff68bc

diff -pur e/fs/ext4/ext4.h f/fs/ext4/ext4.h
--- e/fs/ext4/ext4.h 2022-08-19 19:46:26.317620400 +0800
+++ f/fs/ext4/ext4.h 2022-08-19 19:50:41.005830300 +0800
@@ -1081,6 +1081,7 @@ struct ext4_inode_info {
/* Protect concurrent accesses on i_fc_lblk_start, i_fc_lblk_len */
struct mutex i_fc_lock;

+ struct rw_semaphore i_bmap_sem;
/*
* i_disksize keeps track of what the inode size is ON DISK, not
* in memory. During truncate, i_size is set to the new size by
diff -pur e/fs/ext4/file.c f/fs/ext4/file.c
--- e/fs/ext4/file.c 2022-08-19 19:45:25.253573600 +0800
+++ f/fs/ext4/file.c 2022-08-19 20:04:13.481483600 +0800
@@ -265,11 +265,13 @@ static ssize_t ext4_buffered_write_iter(
ret = ext4_write_checks(iocb, from);
if (ret <= 0)
goto out;
+ down_write(&EXT4_I(inode)->i_bmap_sem);

current->backing_dev_info = inode_to_bdi(inode);
ret = generic_perform_write(iocb, from);
current->backing_dev_info = NULL;

+ up_write(&EXT4_I(inode)->i_bmap_sem);
out:
inode_unlock(inode);
if (likely(ret > 0)) {
diff -pur e/fs/ext4/inode.c f/fs/ext4/inode.c
--- e/fs/ext4/inode.c 2022-08-19 19:44:33.471445500 +0800
+++ f/fs/ext4/inode.c 2022-08-19 20:00:25.319110800 +0800
@@ -3154,7 +3154,7 @@ static sector_t ext4_bmap(struct address
sector_t ret = 0;
int err;

- inode_lock_shared(inode);
+ down_read(&EXT4_I(inode)->i_bmap_sem);
/*
* We can get here for an inline file via the FIBMAP ioctl
*/
@@ -3204,7 +3204,7 @@ static sector_t ext4_bmap(struct address
ret = iomap_bmap(mapping, block, &ext4_iomap_ops);

out:
- inode_unlock_shared(inode);
+ up_read(&EXT4_I(inode)->i_bmap_sem);
return ret;
}

diff -pur e/fs/ext4/super.c f/fs/ext4/super.c
--- e/fs/ext4/super.c 2022-08-19 19:53:38.208033600 +0800
+++ f/fs/ext4/super.c 2022-08-19 19:55:01.503030300 +0800
@@ -1405,6 +1405,7 @@ static void init_once(void *foo)
INIT_LIST_HEAD(&ei->i_orphan);
init_rwsem(&ei->xattr_sem);
init_rwsem(&ei->i_data_sem);
+ init_rwsem(&ei->i_bmap_sem);
inode_init_once(&ei->vfs_inode);
ext4_fc_init_inode(&ei->vfs_inode);
}
--

[syzbot]

Hello,

syzbot has tested the proposed patch and the reproducer did not trigger any issue:

Reported-and-tested-by: syzbot+79e6bb...@syzkaller.appspotmail.com

Tested on:

commit: ca688bff Add linux-next specific files for 20220808
git tree: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git
console output: https://syzkaller.appspot.com/x/log.txt?x=151d45e3080000

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

patch: https://syzkaller.appspot.com/x/patch.diff?x=1267b2f3080000

Note: testing is done by a robot and is best-effort only.

[Hillf Danton]

On Mon, 08 Aug 2022 09:38:26 -0700

> 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

Serialize bmap and inline data pathes with xattr_sem.

#syz test https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git ca688bff68bc

--- e/fs/ext4/inode.c
+++ f/fs/ext4/inode.c

@@ -3154,7 +3154,7 @@ static sector_t ext4_bmap(struct address
sector_t ret = 0;
int err;

- inode_lock_shared(inode);

+ down_read(&EXT4_I(inode)->xattr_sem);

/*
* We can get here for an inline file via the FIBMAP ioctl
*/
@@ -3204,7 +3204,7 @@ static sector_t ext4_bmap(struct address
ret = iomap_bmap(mapping, block, &ext4_iomap_ops);

out:
- inode_unlock_shared(inode);

+ up_read(&EXT4_I(inode)->xattr_sem);
return ret;
}

--

[syzbot]

Hello,

syzbot has tested the proposed patch and the reproducer did not trigger any issue:

Reported-and-tested-by: syzbot+79e6bb...@syzkaller.appspotmail.com

Tested on:

commit: ca688bff Add linux-next specific files for 20220808
git tree: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git

console output: https://syzkaller.appspot.com/x/log.txt?x=16cbacad080000

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

patch: https://syzkaller.appspot.com/x/patch.diff?x=145dc1f3080000

[Jan Kara]

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+79e6bb...@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....@quack3.lan/

Honza

--
Jan Kara <ja...@suse.com>
SUSE Labs, CR

[Theodore Ts'o]

#syz fix: ext4, jbd2: add an optimized bmap for the journal inode