linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Robin Holt <holt@sgi.com>
To: Hugh Dickins <hugh@veritas.com>
Cc: Robin Holt <holt@sgi.com>, Nick Piggin <nickpiggin@yahoo.com.au>,
	Ingo Molnar <mingo@elte.hu>, Christoph Lameter <clameter@sgi.com>,
	Jack Steiner <steiner@sgi.com>,
	linux-mm@kvack.org
Subject: Re: Can get_user_pages( ,write=1, force=1, ) result in a read-only pte and _count=2?
Date: Mon, 23 Jun 2008 10:54:00 -0500	[thread overview]
Message-ID: <20080623155400.GH10123@sgi.com> (raw)
In-Reply-To: <Pine.LNX.4.64.0806191441040.25832@blonde.site>

On Thu, Jun 19, 2008 at 02:49:50PM +0100, Hugh Dickins wrote:
> On Thu, 19 Jun 2008, Robin Holt wrote:
> > On Thu, Jun 19, 2008 at 12:09:15PM +0100, Hugh Dickins wrote:
> > > 
> > > (I assume Robin is not forking, we do know that causes this kind
> > > of problem, but he didn't mention any forking so I assume not.)
> > 
> > There has been a fork long before this mapping was created.  There was a
> > hole at this location and the mapping gets established and pages populated
> > following all ranks of the MPI job getting initialized.
> 
> There's usually been a fork somewhen in the past!  That's no problem.
> 
> The fork problem comes when someone has done a get_user_pages to break
> all the COWs, then another thread does a fork which writeprotects and
> raises page_mapcount, so the next write from userspace breaks COW again
> and writes to a different page from that which the kernel is holding.
> 
> That one kept on coming up, but I've not heard of it again since we
> added madvise MADV_DONTFORK so apps could exclude such parts of the
> address space from copy_page_range.

I finally tracked this down.  I think it is a problem specific to XPMEM
on the SLES10 kernel and will not be a problem once Andrea's mmu_notifier
is in the kernel.  It is a problem, as far as I can tell, specific to
the way XPMEM works.

I will open a SuSE bugzilla to work the issue directly with them.

Prior to the transition event, we have a page of memory that was
pre-faulted by a process.  The process has exported (via XPMEM) a
window of its own address space.  A remote process has attached and
touched the page of memory.  The fault will call into XPMEM which does
the get_user_pages.

At this point, both processes have a writable PTE entry to the same
page and XPMEM has one additional reference count (_count) on the page
acquired via get_user_pages().

Memory pressure causes swap_page to get called.  It clears the two
process's page table entries, returns the _count values, etc.  The only
thing that remains different from normal at this point is XPMEM retains
a reference.

Both processes then read-fault the page which results in readable PTEs
being installed.

The failure point comes when either process write faults the page.
At that point, a COW is initiated and now the two processes are looking
at seperate pages.

The scenario would be different in the case of mmu_notifiers.
The notifier callout when the readable PTE was being replaced with a
writable PTE would result in the remote page table getting cleared and
XPMEM releasing the _count.


All that said, I think the race we discussed earlier in the thread is
a legitimate one and believe Hugh's fix is correct.

Thank you for all your patience,
Robin

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  reply	other threads:[~2008-06-23 15:54 UTC|newest]

Thread overview: 30+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-06-18 16:41 Robin Holt
2008-06-18 17:29 ` Nick Piggin
2008-06-18 19:01   ` Hugh Dickins
2008-06-18 20:33     ` Robin Holt
2008-06-18 21:46       ` Hugh Dickins
2008-06-19  3:31         ` Nick Piggin
2008-06-19  3:34           ` Nick Piggin
2008-06-19 11:39           ` Hugh Dickins
2008-06-19 12:07             ` Nick Piggin
2008-06-19 12:21               ` Nick Piggin
2008-06-19 17:48                 ` Christoph Lameter
2008-06-19 12:34               ` Hugh Dickins
2008-06-19 12:53                 ` Nick Piggin
2008-06-19 13:25                   ` Hugh Dickins
2008-06-19 13:35                     ` Robin Holt
2008-06-19 16:32         ` Robin Holt
2008-06-20  9:23           ` Nick Piggin
2008-06-19  3:07     ` Nick Piggin
2008-06-19 11:09       ` Hugh Dickins
2008-06-19 13:38         ` Robin Holt
2008-06-19 13:49           ` Hugh Dickins
2008-06-23 15:54             ` Robin Holt [this message]
2008-06-23 16:48               ` Hugh Dickins
2008-06-23 17:52                 ` Robin Holt
2008-06-23 20:58                   ` Hugh Dickins
2008-06-24 11:56                     ` Robin Holt
2008-06-24 15:19                     ` Robin Holt
2008-06-24 20:19                       ` Hugh Dickins
2008-06-23 19:11             ` Robin Holt
2008-06-23 19:12               ` Robin Holt

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20080623155400.GH10123@sgi.com \
    --to=holt@sgi.com \
    --cc=clameter@sgi.com \
    --cc=hugh@veritas.com \
    --cc=linux-mm@kvack.org \
    --cc=mingo@elte.hu \
    --cc=nickpiggin@yahoo.com.au \
    --cc=steiner@sgi.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox