xfs
[Top] [All Lists]

Re: [PATCH 6/7] xfs: replace xfs_mod_incore_sb_batched

To: Brian Foster <bfoster@xxxxxxxxxx>
Subject: Re: [PATCH 6/7] xfs: replace xfs_mod_incore_sb_batched
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Thu, 5 Feb 2015 06:19:54 -0800
Cc: Dave Chinner <david@xxxxxxxxxxxxx>, xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20150205141044.GH31625@xxxxxxxxxxxxxx>
References: <1423083249-27493-1-git-send-email-david@xxxxxxxxxxxxx> <1423083249-27493-7-git-send-email-david@xxxxxxxxxxxxx> <20150205141044.GH31625@xxxxxxxxxxxxxx>
User-agent: Mutt/1.5.23 (2014-03-12)
On Thu, Feb 05, 2015 at 09:10:44AM -0500, Brian Foster wrote:
> >     /* apply remaining deltas */
> > +   spin_lock(&mp->m_sb_lock);
> >     if (rtxdelta) {
> > -           error = xfs_mod_frextents(mp, rtxdelta);
> > +           error = xfs_sb_mod64(&mp->m_sb.sb_frextents, rtxdelta);
> 
> Any reason why we don't continue to use the xfs_mod_frextents() function
> introduced in the previous patch? Seems like we should be consistent one
> way or the other.

We're already under the sb_lock and would need another lock roundtrip to
avoid a deadlock in that case.  But mit might be worth to simply make
the real time extent counter a percpu one to behave similar to the other
counters that can be manipulated outside of the transaction commit code.

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