On Thu 07-06-12 23:48:03, Ted Tso wrote:
> I found this while testing 3.5-rc1 with a two ext4 bug fix patches
> (b22b1f178f from the ext4 tree) that will be shortly pushed to Linus.
>
> The two bug fixes are unrelated to the quota system; and I don't recall
> seeing anything like this when I was testing 3.4-rcX based kernels. I
> could try bisecting it, but I thought I'd ask if you could see something
> obvious.
Thanks for report Ted. It is caused by a80b12c3 which removed
I_MUTEX_QUOTA locking subclass and also reordered quota locks. The deadlock
is not actually possible because we never hit delalloc path with quota
files but still it would be nice to avoid the lockdep warning (BTW, using
ext4 hidden quota files will avoid the warning because we don't have to
do fancy stuff synchronizing kernel and userspace idea of file contents).
Anyway, probably the easiest way to silence the warning will be if we use
trylock of s_umount semaphore in writeback_sb_inodes_if_idle() - that
avoids creating i_mutex->s_umount dependency for ext4 which is part of the
lock chain. Btrfs people need this change as well for other purposes and
should hopefully push it out soon so after that the warning should be
silenced...
Honza
> [ 938.213365] ======================================================
> [ 938.214690] [ INFO: possible circular locking dependency detected ]
> [ 938.215853] 3.5.0-rc1-00002-gb22b1f1 #124 Not tainted
> [ 938.216005] -------------------------------------------------------
> [ 938.216005] quotaon/27560 is trying to acquire lock:
> [ 938.216005] (&sb->s_type->i_mutex_key#7){+.+.+.}, at: [<c0249bbd>] vfs_load_quota_inode+0x15b/0x3f2
> [ 938.216005]
> [ 938.216005] but task is already holding lock:
> [ 938.216005] (&s->s_dquot.dqonoff_mutex){+.+...}, at: [<c0249b80>] vfs_load_quota_inode+0x11e/0x3f2
> [ 938.216005]
> [ 938.216005] which lock already depends on the new lock.
> [ 938.216005]
> [ 938.216005]
> [ 938.216005] the existing dependency chain (in reverse order) is:
> [ 938.216005]
> [ 938.216005] -> #2 (&s->s_dquot.dqonoff_mutex){+.+...}:
> [ 938.216005] [<c019b4ff>] lock_acquire+0x99/0xbd
> [ 938.216005] [<c06ea154>] __mutex_lock_common+0x30/0x316
> [ 938.216005] [<c06ea4e8>] mutex_lock_nested+0x26/0x2f
> [ 938.216005] [<c024a03e>] dquot_resume+0x39/0xe8
> [ 938.216005] [<c028d631>] ext4_remount+0x358/0x429
> [ 938.216005] [<c0210aef>] do_remount_sb+0xae/0x120
> [ 938.216005] [<c0225af0>] do_mount+0x236/0x691
> [ 938.216005] [<c0225fb1>] sys_mount+0x66/0x95
> [ 938.216005] [<c06ec7d5>] syscall_call+0x7/0xb
> [ 938.216005]
> [ 938.216005] -> #1 (&type->s_umount_key#18){+++++.}:
> [ 938.216005] [<c019b4ff>] lock_acquire+0x99/0xbd
> [ 938.216005] [<c06ea99f>] down_read+0x28/0x65
> [ 938.216005] [<c022bf3a>] writeback_inodes_sb_if_idle+0x2c/0x47
> [ 938.216005] [<c02757bc>] ext4_nonda_switch+0xf4/0x105
> [ 938.216005] [<c027abfa>] ext4_da_write_begin+0x3c/0x194
> [ 938.216005] [<c01df8e6>] generic_file_buffered_write+0xd1/0x1db
> [ 938.216005] [<c01e0a29>] __generic_file_aio_write+0x25e/0x28f
> [ 938.216005] [<c01e0ac4>] generic_file_aio_write+0x6a/0xba
> [ 938.216005] [<c02734ba>] ext4_file_write+0x1dc/0x22a
> [ 938.216005] [<c020edb3>] do_sync_write+0x92/0xcd
> [ 938.216005] [<c020f702>] vfs_write+0x85/0xe3
> [ 938.216005] [<c020f801>] sys_write+0x40/0x65
> [ 938.216005] [<c06ec7d5>] syscall_call+0x7/0xb
> [ 938.216005]
> [ 938.216005] -> #0 (&sb->s_type->i_mutex_key#7){+.+.+.}:
> [ 938.216005] [<c019b1fa>] __lock_acquire+0x982/0xbee
> [ 938.216005] [<c019b4ff>] lock_acquire+0x99/0xbd
> [ 938.216005] [<c06ea154>] __mutex_lock_common+0x30/0x316
> [ 938.216005] [<c06ea4e8>] mutex_lock_nested+0x26/0x2f
> [ 938.216005] [<c0249bbd>] vfs_load_quota_inode+0x15b/0x3f2
> [ 938.216005] [<c0249ffd>] dquot_quota_on+0x41/0x49
> [ 938.216005] [<c028bfed>] ext4_quota_on+0x12e/0x13b
> [ 938.216005] [<c024cbd1>] do_quotactl+0x1b8/0x3a0
> [ 938.216005] [<c024cefb>] sys_quotactl+0x142/0x17b
> [ 938.216005] [<c06ec7d5>] syscall_call+0x7/0xb
> [ 938.216005]
> [ 938.216005] other info that might help us debug this:
> [ 938.216005]
> [ 938.216005] Chain exists of:
> [ 938.216005] &sb->s_type->i_mutex_key#7 --> &type->s_umount_key#18 --> &s->s_dquot.dqonoff_mutex
> [ 938.216005]
> [ 938.216005] Possible unsafe locking scenario:
> [ 938.216005]
> [ 938.216005] CPU0 CPU1
> [ 938.216005] ---- ----
> [ 938.216005] lock(&s->s_dquot.dqonoff_mutex);
> [ 938.216005] lock(&type->s_umount_key#18);
> [ 938.216005] lock(&s->s_dquot.dqonoff_mutex);
> [ 938.216005] lock(&sb->s_type->i_mutex_key#7);
> [ 938.216005]
> [ 938.216005] *** DEADLOCK ***
> [ 938.216005]
> [ 938.216005] 2 locks held by quotaon/27560:
> [ 938.216005] #0: (&type->s_umount_key#18){+++++.}, at: [<c0210d0a>] get_super+0x50/0x93
> [ 938.216005] #1: (&s->s_dquot.dqonoff_mutex){+.+...}, at: [<c0249b80>] vfs_load_quota_inode+0x11e/0x3f2
> [ 938.216005]
> [ 938.216005] stack backtrace:
> [ 938.216005] Pid: 27560, comm: quotaon Not tainted 3.5.0-rc1-00002-gb22b1f1 #124
> [ 938.216005] Call Trace:
> [ 938.216005] [<c0199164>] print_circular_bug+0x194/0x1a1
> [ 938.216005] [<c019b1fa>] __lock_acquire+0x982/0xbee
> [ 938.216005] [<c01dfe69>] ? find_get_pages+0xee/0xf8
> [ 938.216005] [<c019b4ff>] lock_acquire+0x99/0xbd
> [ 938.216005] [<c0249bbd>] ? vfs_load_quota_inode+0x15b/0x3f2
> [ 938.216005] [<c06ea154>] __mutex_lock_common+0x30/0x316
> [ 938.216005] [<c0249bbd>] ? vfs_load_quota_inode+0x15b/0x3f2
> [ 938.216005] [<c06ea4e8>] mutex_lock_nested+0x26/0x2f
> [ 938.216005] [<c0249bbd>] ? vfs_load_quota_inode+0x15b/0x3f2
> [ 938.216005] [<c0249bbd>] vfs_load_quota_inode+0x15b/0x3f2
> [ 938.216005] [<c0249ffd>] dquot_quota_on+0x41/0x49
> [ 938.216005] [<c028bfed>] ext4_quota_on+0x12e/0x13b
> [ 938.216005] [<c024cbd1>] do_quotactl+0x1b8/0x3a0
> [ 938.216005] [<c028bebf>] ? ext4_msg+0x3c/0x3c
> [ 938.216005] [<c06ea9d5>] ? down_read+0x5e/0x65
> [ 938.216005] [<c0210d0a>] ? get_super+0x50/0x93
> [ 938.216005] [<c0210f82>] ? get_super_thawed+0x25/0xa9
> [ 938.216005] [<c024cefb>] sys_quotactl+0x142/0x17b
> [ 938.216005] [<c06ec7d5>] syscall_call+0x7/0xb
--
Jan Kara <jack@xxxxxxx>
SUSE Labs, CR
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
