xfs
[Top] [All Lists]

Re: [PATCH 06/11] xfs: fully initialise temp leaf in xfs_attr3_leaf_unba

To: Ben Myers <bpm@xxxxxxx>
Subject: Re: [PATCH 06/11] xfs: fully initialise temp leaf in xfs_attr3_leaf_unbalance
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Thu, 23 May 2013 09:54:50 +1000
Cc: xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20130522205059.GP20028@xxxxxxx>
References: <1369123330-9579-1-git-send-email-david@xxxxxxxxxxxxx> <1369123330-9579-7-git-send-email-david@xxxxxxxxxxxxx> <20130522205059.GP20028@xxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Wed, May 22, 2013 at 03:50:59PM -0500, Ben Myers wrote:
> On Tue, May 21, 2013 at 06:02:05PM +1000, Dave Chinner wrote:
> > From: Dave Chinner <dchinner@xxxxxxxxxx>
> > 
> > xfs_attr3_leaf_unbalance() uses a temporary buffer for recombining
> > the entries in two leaves when the destination leaf requires
> > compaction. The temporary buffer ends up being copied back over the
> > original destination buffer, so the header in the temporary buffer
> > needs to contain all the information that is in the destination
> > buffer.
> > 
> > To make sure the temporary buffer is fully initialised, once we've
> > set up the temporary incore header appropriately, write is back to
> > the temporary buffer before starting to move entries around.
> > 
> > Signed-off-by: Dave Chinner <dchinner@xxxxxxxxxx>
> >
> > ---
> >  fs/xfs/xfs_attr_leaf.c |   16 +++++++++++++---
> >  1 file changed, 13 insertions(+), 3 deletions(-)
> > 
> > diff --git a/fs/xfs/xfs_attr_leaf.c b/fs/xfs/xfs_attr_leaf.c
> > index 0bce1b3..79ece72 100644
> > --- a/fs/xfs/xfs_attr_leaf.c
> > +++ b/fs/xfs/xfs_attr_leaf.c
> > @@ -2181,14 +2181,24 @@ xfs_attr3_leaf_unbalance(
> >             struct xfs_attr_leafblock *tmp_leaf;
> >             struct xfs_attr3_icleaf_hdr tmphdr;
> >  
> > -           tmp_leaf = kmem_alloc(state->blocksize, KM_SLEEP);
> > -           memset(tmp_leaf, 0, state->blocksize);
> > -           memset(&tmphdr, 0, sizeof(tmphdr));
> > +           tmp_leaf = kmem_zalloc(state->blocksize, KM_SLEEP);
> > +
> > +           /*
> > +            * Copy the header into the temp leaf so that all the stuff
> > +            * not in the incore header is present and gets copied back in
> > +            * once we've moved all the entries.
> > +            */
> > +           memcpy(tmp_leaf, save_leaf, xfs_attr3_leaf_hdr_size(save_leaf));
> >  
> > +           memset(&tmphdr, 0, sizeof(tmphdr));
> >             tmphdr.magic = savehdr.magic;
> >             tmphdr.forw = savehdr.forw;
> >             tmphdr.back = savehdr.back;
> >             tmphdr.firstused = state->blocksize;
> > +
> > +           /* write the header to the temp buffer to initialise it */
> > +           xfs_attr3_leaf_hdr_to_disk(tmp_leaf, &tmphdr);
> > +
> 
> Hrm.  It seemed like the memset 0 for the incore temp header should have been
> enough for the purposes of moveents.  Why did you have to add this additional
> hdr_to_disk?

Because we've modified the tmphdr so it is different to what is in
the tmp_leaf, and I'm being paranoid and not making any assumptions
that a callee doesn't reread the tmphdr from the tmp_leaf.

i.e. I'm initialising both the on-disk and the incore headers to
be identical before we do anything else. Once bitten, twice shy.

Cheers,

Dave.
-- 
Dave Chinner
david@xxxxxxxxxxxxx

<Prev in Thread] Current Thread [Next in Thread>