Subject: Re: 2.6.22-rc4-mm1
From: Peter Zijlstra <a.p.zijlstra@chello.nl>
To: William Lee Irwin III <wli@holomorphy.com>
Cc: Mel Gorman <mel@skynet.ie>, Cedric Le Goater <clg@fr.ibm.com>,
       Michal Piotrowski <michal.k.k.piotrowski@gmail.com>,
       Andrew Morton <akpm@linux-foundation.org>, linux-kernel@vger.kernel.org,
       Nick Piggin <npiggin@suse.de>, aaw@google.com
In-Reply-To: <1181149580.5676.25.camel@lappy>
References: <20070606020737.4663d686.akpm@linux-foundation.org>
	 <4666ABD4.8050402@fr.ibm.com> <4666D050.7010603@googlemail.com>
	 <4666DA55.5090405@fr.ibm.com> <20070606162649.GA9932@skynet.ie>
	 <20070606163538.GN6909@holomorphy.com>  <1181149580.5676.25.camel@lappy>
Content-Type: text/plain
Date: Wed, 06 Jun 2007 21:27:55 +0200
Message-Id: <1181158075.5676.32.camel@lappy>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1729
Lines: 38

On Wed, 2007-06-06 at 19:06 +0200, Peter Zijlstra wrote:
> On Wed, 2007-06-06 at 09:35 -0700, William Lee Irwin III wrote:
> > On Wed, Jun 06, 2007 at 05:26:49PM +0100, Mel Gorman wrote:
> > > I do not believe this is Nick's problem. I encountered the same issue and
> > > the bisect ended up here;
> > > # BISECT HERE
> > > mm-variable-length-argument-support.patch
> > > mm-variable-length-argument-support-fix.patch
> > > # BISECT BAD
> > > Reverting those two patches boots ok on my standalone x86 laptop.
> > > Patch authors cc'd. I have not read the patches yet to see what might
> > > be the problem.
> > 
> > I found this a while ago and peterz already has a tentative fix for it at
> > http://programming.kicks-ass.net/kernel-patches/max_arg_pages/move_anon_vma.patch
> > I'm sure he himself will chime in with more/better code when he returns.
> 
> Ah, yes indeed. We were discussing wether it would be better to update
> page->index (as this patch does) or not change vma->vm_pgoff. Bill
> argued that the latter would be safe, it would just limit the merge
> capabilities of the anon vma (and the stack rarely, if ever, merges
> anyway).
> 
> I myself an not quite sure on the effects of ->vm_pgoff of anon vmas
> yet.

Ok, from what I can make of it, nobody relies on the value of
vma->vm_pgoff for anon vmas as long as the following invariant is kept:

  vaddr = vma->vm_start + (page->index - vma->vm_pgoff)*PAGE_SIZE

So the proper thing to do is not change vm_pgoff.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/