linux-ext4 - Re: Delayed allocation and journal locking order inversion.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Wed, 28 May 2008 12:08:33 +0200
From:	Jan Kara <jack@...e.cz>
To:	"Aneesh Kumar K.V" <aneesh.kumar@...ux.vnet.ibm.com>
Cc:	Mingming Cao <cmm@...ibm.com>,
	ext4 development <linux-ext4@...r.kernel.org>
Subject: Re: Delayed allocation and journal locking order inversion.

  Hi Aneesh,

  Thanks for testing!

On Wed 28-05-08 14:46:48, Aneesh Kumar K.V wrote:
> I am observing hangs with the delalloc with locking order inversion
> patches. I guess we can't start journal and call write_cache_pages.
  This should be fine after the lock inversion...

> The process get stuck as below
> 
> fsstress      D 00000008     0  2520      1
>        c69c9d70 00000046 c69c9d28 00000008 c6a300a0 c69c50e0 c69c5244 c1210d80 
>        00000000 c7a102a0 c69c50e0 c043c960 c69c9da8 c69c9d6c c0246fe8 00000000 
>        00000000 00000000 c69c9da8 c1210d80 c69c9da8 c11c0998 c69c9d7c c043a8cb 
> Call Trace:
>  [<c043c960>] ? _spin_unlock_irqrestore+0x36/0x58
>  [<c0246fe8>] ? blk_unplug+0x63/0x6b
>  [<c043a8cb>] io_schedule+0x1e/0x28
>  [<c014aac1>] sync_page+0x36/0x3a
>  [<c043aa17>] __wait_on_bit_lock+0x30/0x59
>  [<c014aa8b>] ? sync_page+0x0/0x3a
>  [<c014aa77>] __lock_page+0x4e/0x56
>  [<c01325a4>] ? wake_bit_function+0x0/0x43
>  [<c014ffca>] write_cache_pages+0x120/0x296
>  [<c018c516>] ? __mpage_da_writepage+0x0/0x105
>  [<c043c89d>] ? _spin_unlock+0x27/0x3c
>  [<c018bde8>] mpage_da_writepages+0x5c/0x7e
>  [<c01faa8f>] ? jbd2_journal_start+0xce/0xf0
>  [<c01faaa4>] ? jbd2_journal_start+0xe3/0xf0
>  [<c01d893b>] ? ext4_da_get_block_write+0x0/0x151
>  [<c01d8cc6>] ext4_da_writepages+0xbe/0x116
>  [<c01d8c08>] ? ext4_da_writepages+0x0/0x116
>  [<c015018a>] do_writepages+0x23/0x34
>  [<c0180ffa>] __writeback_single_inode+0x12a/0x260
>  [<c0181480>] sync_sb_inodes+0x18d/0x25b
>  [<c01815d0>] sync_inodes_sb+0x82/0x94
>  [<c0181638>] __sync_inodes+0x56/0x9c
>  [<c0181692>] sync_inodes+0x14/0x2c
>  [<c0183bc1>] do_sync+0x14/0x50
>  [<c0183c0a>] sys_sync+0xd/0x13
>  [<c0103931>] sysenter_past_esp+0x6a/0xb1
  The question here is, who is holding the lock from the page we wait
for here. The two processes you show below don't seem to hold it. I'll
check the full log ... searching ... I see!
  The problem is in generic_write_end()! It calls mark_inode_dirty() under
page lock. That can possibly start a new transaction (which happened in
your case) and that violates lock ordering (mark_inode_dirty() got stuck
waiting for journal commit which is stuck waiting for other user to do
journal_stop which waits for the page lock). Actually, there is no real
need to call mark_inode_dirty() from under page lock - we just need to
update i_size there. Something like the patch attached (untested)?

<snip>
> The full dmesg log is at 
> http://www.radian.org/~kvaneesh/ext4/delalloc-lockinversion/dmesg-1.log
> 
> Also starting journal in writepages make unmount throw lockdep errors.
> 
> unlink does journal_start and lock_super.
> umount does lock_super and later it need to sync_inodes does writepages
> which does a journal_start.
  Well, but isn't there this problem even without the lock inversion patch?
This is inversion between lock_super and journal_start. It hasn't been
changed by the lock inversion patch as far as I can tell. If you send me
lockdep traces I can have a look what we could do...

> I guess we will have to rework the delalloc related changes.

									Honza
-- 
Jan Kara <jack@...e.cz>
SUSE Labs, CR

View attachment "vfs-2.6.25-generic_write_end.diff" of type "text/x-patch" (1521 bytes)