Re: [PATCH] [6/16] HWPOISON: Add various poison checks in mm/memory.c

[Wu Fengguang <fengguang.wu@intel.com>]

On Thu, Jun 04, 2009 at 09:02:55PM +0800, Andi Kleen wrote:
> On Thu, Jun 04, 2009 at 08:50:26PM +0800, Wu Fengguang wrote:
> > On Thu, Jun 04, 2009 at 08:52:28PM +0800, Andi Kleen wrote:
> > > On Thu, Jun 04, 2009 at 07:55:33PM +0800, Wu Fengguang wrote:
> > > > On Thu, Jun 04, 2009 at 01:19:15PM +0800, Andi Kleen wrote:
> > > > > On Thu, Jun 04, 2009 at 12:26:03PM +0800, Wu Fengguang wrote:
> > > > > > On Thu, Jun 04, 2009 at 02:46:38AM +0800, Andi Kleen wrote:
> > > > > > >
> > > > > > > Bail out early when hardware poisoned pages are found in page fault handling.
> > > > > >
> > > > > > I suspect this patch is also not absolutely necessary: the poisoned
> > > > > > page will normally have been isolated already.
> > > > >
> > > > > It's needed to prevent new pages comming in when there is a parallel
> > > > > fault while the memory failure handling is in process.
> > > > > Otherwise the pages could get remapped in that small window.
> > > > 
> > > > This patch makes no difference at least for file pages, including tmpfs.
> > > 
> > > I was more thinking of anonymous pages with multiple mappers (e.g.
> > > COW after fork)
> > 
> > I guess they are handled by do_anonymous_page() or do_wp_page(),
> > instead of do_linear_fault()/do_nonlinear_fault()?
> 
> You're right. Sorry was a little confused in my earlier reply.
> 
> I think what I meant is: what happens during the window
> when the page has just the poison bit set, but is not isolated/unmapped yet.
> During that window I want new mappers to not come in.
> That is why that check is there.

As soon as the poisoned page is locked, it is in safe hand - the new
mappers will have to wait, and then find it either truncated (mapping
== NULL) for file pages, or its PTE updated through the pte_same()
checks in do_wp_page().  do_anonymous_page() is safe because it
allocates the good new page.

We lock the page immediately after setting PG_hwpoison, so the window
is small enough :)

Thanks,
Fengguang