From: Jens Axboe <jens.axboe@oracle.com>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: linux-kernel@vger.kernel.org
Subject: Re: splice: move balance_dirty_pages_ratelimited() outside of splice actor
Date: Tue, 12 Jun 2007 13:20:59 +0200 [thread overview]
Message-ID: <20070612112059.GY18832@kernel.dk> (raw)
In-Reply-To: <20070612063510.GO18832@kernel.dk>
On Tue, Jun 12 2007, Jens Axboe wrote:
> On Mon, Jun 11 2007, Andrew Morton wrote:
> > On Mon, 11 Jun 2007 21:59:15 GMT
> > Linux Kernel Mailing List <linux-kernel@vger.kernel.org> wrote:
> >
> > > Gitweb: http://git.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=20d698db67059a63d217030dfd02872cb5f88dfb
> > > Commit: 20d698db67059a63d217030dfd02872cb5f88dfb
> > > Parent: 17374ff1aa9ce2a0597416a16729474b538af443
> > > Author: Jens Axboe <jens.axboe@oracle.com>
> > > AuthorDate: Tue Jun 5 11:05:11 2007 +0200
> > > Committer: Jens Axboe <jens.axboe@oracle.com>
> > > CommitDate: Fri Jun 8 08:33:59 2007 +0200
> > >
> > > splice: move balance_dirty_pages_ratelimited() outside of splice actor
> > >
> > > I've seen inode related deadlocks, so move this call outside of the
> > > actor itself, which may hold the inode lock.
> > >
> >
> > eh? If the pipe_to_file() caller holds inode_lock, our problems are large.
> >
> > I doubt if that's true, so what problem is this patch really fixing??
>
> I can repeatedly lock up balance_dirty_pages() if it's called inside the
> double lock of splice_from_pipe(). I'll fire up the test box and
> reproduce, then post the backtraces.
OK, here it is. Running
$ ./splice-fromnet 9999 | ./splice-out xxx
on one box (receive from network port 9999, store in file xxx) and a
simple
$ cat SUSE-Linux-10.1-GM-DVD-i386.iso | netcat centera 9999
on another box. After ~180mb has been transferred, splice-out gets
stuck. With the balancing moved outside of the splice actor, everything
works perfectly regardless of memory pressure.
splice-fromne S 00000000 0 9829 9778 (NOTLB)
efe63d4c 00000046 f524cb20 00000000 efe63cfc c0144ae3 00000000 f524cb20
c03c33fa f1b9ee0c 00000009 f524cb20 02ad51ce 00000024 00002406 f524cc30
c1cf3980 00000001 efe63d44 00000246 f6692040 00000282 efe63d54 f2f83600
Call Trace:
[<c01809bd>] pipe_wait+0x6d/0x90
[<c0198006>] splice_to_pipe+0x196/0x240
[<c035eb4d>] skb_splice_bits+0xcd/0xe0
[<c038157b>] tcp_splice_data_recv+0x2b/0x40
[<c0381501>] tcp_read_sock+0x131/0x180
[<c0381f55>] tcp_splice_read+0x85/0x1e0
[<c03572f5>] sock_splice_read+0x25/0x30
[<c019743e>] do_splice_to+0x5e/0x80
[<c0198beb>] sys_splice+0x1eb/0x200
[<c010511e>] sysenter_past_esp+0x5f/0x99
=======================
splice-out D EF423C98 0 9830 9778 (NOTLB)
ef423cac 00000046 00000002 ef423c98 ef423c94 00000000 00000046 f760ba98
f760ba98 f760b9a0 00000006 f7cba0f0 02ca49b4 00000024 001cf7e6 f7cba200
c1cf3980 00000001 c0144c73 c1cdaf78 f55c64c0 00000046 c1cdaf78 fffd5d5b
Call Trace:
[<c03c272d>] io_schedule+0x1d/0x30
[<c0154979>] sync_page+0x39/0x50
[<c03c2d20>] __wait_on_bit_lock+0x40/0x70
[<c0154929>] __lock_page+0x69/0x70
[<c015a9c5>] write_cache_pages+0x105/0x2f0
[<c015abd3>] generic_writepages+0x23/0x30
[<c015ac29>] do_writepages+0x49/0x50
[<c01960a4>] __writeback_single_inode+0x94/0x370
[<c0196783>] sync_sb_inodes+0x1b3/0x270
[<c0196aa8>] writeback_inodes+0xb8/0xf0
[<c015afed>] balance_dirty_pages_ratelimited_nr+0x9d/0x1b0
[<c01978b3>] pipe_to_file+0x163/0x260
[<c0197a0f>] __splice_from_pipe+0x5f/0x250
[<c0197d18>] splice_from_pipe+0x58/0x80
[<c0197dc4>] generic_file_splice_write+0x54/0x100
[<c01974bf>] do_splice_from+0x5f/0x80
[<c0198bce>] sys_splice+0x1ce/0x200
[<c010511e>] sysenter_past_esp+0x5f/0x99
There are two pdflushes, doing nothing:
=======================
pdflush S F7D8FF90 0 171 2 (L-TLB)
f7d8ffa4 00000046 00000002 f7d8ff90 f7d8ff8c 00000000 f7c50b60 c055ebac
f7c51090 00000001 0000000a f7c50b60 6fc07726 00000026 0002552e f7c50c70
c1ce9980 00000000 c015b750 f7d8ff98 f71d8b80 c0498c8c 00000001 fffd86d8
Call Trace:
[<c015b7ec>] pdflush+0x9c/0x1c0
[<c013aa42>] kthread+0x42/0x70
[<c0105e0b>] kernel_thread_helper+0x7/0x1c
=======================
pdflush S F7D91F90 0 172 2 (L-TLB)
f7d91fa4 00000046 00000002 f7d91f90 f7d91f8c 00000000 f7ca6aa0 c055ebac
f7ca6fd0 00000001 00000009 f7ca6aa0 01d76a76 00000024 003c4c0a f7ca6bb0
c1ce9980 00000000 c015b750 f7d91f98 f7255dc0 c0498c8c 00000001 fffd5df9
Call Trace:
[<c015b7ec>] pdflush+0x9c/0x1c0
[<c013aa42>] kthread+0x42/0x70
[<c0105e0b>] kernel_thread_helper+0x7/0x1c
--
Jens Axboe
next prev parent reply other threads:[~2007-06-12 11:23 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <200706112159.l5BLxF5x004043@hera.kernel.org>
2007-06-11 23:34 ` splice: move balance_dirty_pages_ratelimited() outside of splice actor Andrew Morton
2007-06-12 6:35 ` Jens Axboe
2007-06-12 11:20 ` Jens Axboe [this message]
2007-06-12 11:31 ` Jens Axboe
2007-06-12 12:06 ` Peter Zijlstra
2007-06-12 12:10 ` Jens Axboe
2007-06-12 12:16 ` Peter Zijlstra
2007-06-12 12:44 ` Jens Axboe
2007-06-12 16:02 ` Andrew Morton
2007-06-12 16:46 ` Andrew Morton
2007-06-12 17:32 ` Jens Axboe
2007-06-12 18:15 ` Jens Axboe
2007-06-12 18:43 ` Andrew Morton
2007-06-12 18:48 ` Jens Axboe
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20070612112059.GY18832@kernel.dk \
--to=jens.axboe@oracle.com \
--cc=akpm@linux-foundation.org \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.