xfs
[Top] [All Lists]

Re: [PATCH] xfs: fix mmap_sem vs iolock lock order inversion inxfs_free_

To: Alex Elder <aelder@xxxxxxx>
Subject: Re: [PATCH] xfs: fix mmap_sem vs iolock lock order inversion inxfs_free_eofblocks
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Tue, 3 Nov 2009 09:51:52 -0500
Cc: Christoph Hellwig <hch@xxxxxxxxxxxxx>, xfs@xxxxxxxxxxx
In-reply-to: <1AB9A794DBDDF54A8A81BE2296F7BDFE83AE5C@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>
References: <20091019040346.GB21115@xxxxxxxxxxxxx> <1AB9A794DBDDF54A8A81BE2296F7BDFE83AE5C@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>
User-agent: Mutt/1.5.19 (2009-01-05)
On Mon, Nov 02, 2009 at 03:10:33PM -0600, Alex Elder wrote:
> I have a minor suggestion below, but it looks correct to me.
> I tried to get a better idea of what the conditions were
> where mmap_sem would be held by VM when ->release gets
> called, but didn't get to the bottom of that.  If it is
> easily characterized you could mention it in comments.

It's the VMA merging code.  But IMHO that's too much of an
implementation detail to put into the comment here.  Commit message
might be fine.

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