public inbox for linux-xfs@vger.kernel.org
 help / color / mirror / Atom feed
From: "Darrick J. Wong" <darrick.wong@oracle.com>
To: Dave Chinner <david@fromorbit.com>
Cc: linux-xfs@vger.kernel.org
Subject: Re: [PATCH 5/7] repair: don't duplicate names in phase 6
Date: Thu, 22 Oct 2020 08:53:24 -0700	[thread overview]
Message-ID: <20201022155324.GS9832@magnolia> (raw)
In-Reply-To: <20201022082354.GU7391@dread.disaster.area>

On Thu, Oct 22, 2020 at 07:23:54PM +1100, Dave Chinner wrote:
> On Wed, Oct 21, 2020 at 11:21:52PM -0700, Darrick J. Wong wrote:
> > On Thu, Oct 22, 2020 at 04:15:35PM +1100, Dave Chinner wrote:
> ....
> > > @@ -1387,6 +1371,7 @@ dir2_kill_block(
> > >  		res_failed(error);
> > >  	libxfs_trans_ijoin(tp, ip, 0);
> > >  	libxfs_trans_bjoin(tp, bp);
> > > +	libxfs_trans_bhold(tp, bp);
> > 
> > Why hold on to the buffer?  We killed the block, why keep the reference
> > around so that someone else has to remember to drop it later?
> 
> Consistency in buffer handling. This "kill block" path nulled out
> the buffer in the bplist array (the reason it's passed as a **bpp
> to longform_dir2_entry_check_data(). This path releases the buffer
> through the trans_commit(), the alternate path here:
> 
> > > @@ -1558,10 +1541,8 @@ longform_dir2_entry_check_data(
> > >  			dir2_kill_block(mp, ip, da_bno, bp);
> > >  		} else {
> > >  			do_warn(_("would junk block\n"));
> > > -			libxfs_buf_relse(bp);
> > >  		}
> > >  		freetab->ents[db].v = NULLDATAOFF;
> > > -		*bpp = NULL;
> > >  		return;
> > >  	}
> 
> does an explicit release, and all other paths end up with the
> buffers being released way back where the bplist is defined.
> 
> If the directory is in block form, nulling out the buffer in the
> bplist here will result in dereferencing a null pointer later when
> the buffer is pulled from bplist[0] without checking.
> 
> So I changed longform_dir2_entry_check_data() to never release the
> buffer so that the caller knows that it has always got a valid
> buffer reference and the isblock path will always work correctly....

Hmmm, looking through the directory code, I see that it calls binval to
stale the buffer, so the kill_block caller won't be able to do much with
its incore reference.

Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>

--D

> 
> Cheers,
> 
> Dave.
> 
> 
> -- 
> Dave Chinner
> david@fromorbit.com

  reply	other threads:[~2020-10-22 15:53 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-10-22  5:15 [PATCH 0/7] repair: Phase 6 performance improvements Dave Chinner
2020-10-22  5:15 ` [PATCH 1/7] workqueue: bound maximum queue depth Dave Chinner
2020-10-22  5:54   ` Darrick J. Wong
2020-10-22  8:11     ` Dave Chinner
2020-10-25  4:41   ` Darrick J. Wong
2020-10-26 22:29     ` Dave Chinner
2020-10-26 22:40       ` Darrick J. Wong
2020-10-26 22:57         ` Dave Chinner
2020-10-22  5:15 ` [PATCH 2/7] repair: Protect bad inode list with mutex Dave Chinner
2020-10-22  5:45   ` Darrick J. Wong
2020-10-29  9:35   ` Christoph Hellwig
2020-10-22  5:15 ` [PATCH 3/7] repair: protect inode chunk tree records with a mutex Dave Chinner
2020-10-22  6:02   ` Darrick J. Wong
2020-10-22  8:15     ` Dave Chinner
2020-10-29 16:45       ` Darrick J. Wong
2020-10-22  5:15 ` [PATCH 4/7] repair: parallelise phase 6 Dave Chinner
2020-10-22  6:11   ` Darrick J. Wong
2020-10-27  5:10     ` Dave Chinner
2020-10-29 17:20       ` Darrick J. Wong
2020-10-22  5:15 ` [PATCH 5/7] repair: don't duplicate names in " Dave Chinner
2020-10-22  6:21   ` Darrick J. Wong
2020-10-22  8:23     ` Dave Chinner
2020-10-22 15:53       ` Darrick J. Wong [this message]
2020-10-29  9:39   ` Christoph Hellwig
2020-10-22  5:15 ` [PATCH 6/7] repair: convert the dir byaddr hash to a radix tree Dave Chinner
2020-10-29 16:41   ` Darrick J. Wong
2020-10-22  5:15 ` [PATCH 7/7] repair: scale duplicate name checking in phase 6 Dave Chinner
2020-10-29 16:29   ` Darrick J. Wong
  -- strict thread matches above, loose matches on Subject: below --
2021-03-19  1:33 [PATCH 0/7] repair: Phase 6 performance improvements Dave Chinner
2021-03-19  1:33 ` [PATCH 5/7] repair: don't duplicate names in phase 6 Dave Chinner

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=20201022155324.GS9832@magnolia \
    --to=darrick.wong@oracle.com \
    --cc=david@fromorbit.com \
    --cc=linux-xfs@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