linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Marcelo Tosatti <marcelo.tosatti@cyclades.com>
To: Jamie Clark <jclark@metaparadigm.com>
Cc: Andrea Arcangeli <andrea@suse.de>, <linux-kernel@vger.kernel.org>
Subject: Re: 2.4.23aa1 ext3 oops
Date: Tue, 16 Dec 2003 13:15:21 -0200 (BRST)	[thread overview]
Message-ID: <Pine.LNX.4.44.0312161310570.1533-100000@logos.cnet> (raw)
In-Reply-To: <3FD7D78A.4080409@metaparadigm.com>


Jamie, 

Did you try the patch I suggested for you to revert  ?

On Thu, 11 Dec 2003, Jamie Clark wrote:

> OK, no deadlock yet with 2.4.23aa1 however it oopsed under 
> ext3_file_write() in __mark_inode_dirty().
> 
> Just to recap: this test is dual PIII, running several bonnie++ loads on 
> an ext3+noatime+quota filesystem
> mounted off
> 
>  From the oops the fault happens on the last instruction of:
> 
>         movl $0,8(%ebx)
>         movl $0,4(%edx)
>         movl 100(%edi),%eax 
>         movl %edx,4(%eax)    <-- here
> 
> which appears to be this code in inode.c  [line 221+]
> 
>                 if (!(inode->i_state & (I_LOCK|I_FREEING|I_CLEAR)) &&
>                     !list_empty(&inode->i_hash)) {
>                         list_del(&inode->i_list);
>                         list_add(&inode->i_list, &sb->s_dirty);
> 
> After a quick browse of the assembler output the zeroing would appear to 
> be part of the list_del inline, and edi seems to equate to &sb. If I 
> have read that correctly then the 
> oops happens at the beginning of
> the list_add() inline and eax is the head of the s_dirty list - pointing 
> into oblivion.
> 
> __mark_inode_dirty() does not appear to take sb_lock before adding to 
> the s_dirty list. Could that
> be the culprit?   I'm completely unfamiliar with linux kernel so I might 
> be way off here.
> 
> -Jamie
> 
> Andrea Arcangeli wrote:
> 
> >On Tue, Nov 04, 2003 at 07:52:40PM +0800, Jamie Clark wrote:
> >  
> >
> >>I made the quick fix (disabling rq_mergeable) and started the load test.
> >>Will let it run for a week or so.
> >>    
> >>
> >
> >does your later recent email means it deadlocked again even with this
> >disabled?
> >
> >Could you try again with 2.4.23aa1 again just in case?
> >
> >  
> >
> >>FYI an observation from my last test: the read latency seems to be much
> >>improved and more consistent under this kernel (2.4.23pre6aa3, before
> >>the oops and before this fix).  The maximum latency seemed steady over
> >>the whole test without any of the longish pauses that showed up under
> >>2.4.19. Quite a difference.
> >>    
> >>
> >
> >nice to hear! thanks.
> >  
> >
> 


  reply	other threads:[~2003-12-16 15:34 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2003-11-04  2:49 2.4.23pre6aa3 scsi oops Jamie Clark
2003-11-04 10:28 ` Andrea Arcangeli
2003-11-04 11:52   ` Jamie Clark
2003-11-04 12:48     ` Andrea Arcangeli
2003-11-27  3:21     ` 2.4.23pre6aa3 - possible ext3 deadlock Jamie Clark
2003-11-27 12:56       ` Jan Kara
2003-12-06  1:05     ` 2.4.23pre6aa3 scsi oops Andrea Arcangeli
2003-12-06  1:39       ` Jamie Clark
2003-12-11  2:33       ` 2.4.23aa1 ext3 oops Jamie Clark
2003-12-16 15:15         ` Marcelo Tosatti [this message]
2003-12-16 15:55           ` Jamie Clark
2003-12-17 11:42         ` David Woodhouse
2003-12-18  6:19           ` Jamie Clark
2003-12-18  6:25             ` David Woodhouse
2003-12-18  7:33               ` Jamie Clark

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=Pine.LNX.4.44.0312161310570.1533-100000@logos.cnet \
    --to=marcelo.tosatti@cyclades.com \
    --cc=andrea@suse.de \
    --cc=jclark@metaparadigm.com \
    --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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).