From: Mel Gorman <mgorman@suse.de>
To: Oleg Nesterov <oleg@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>,
Andrea Arcangeli <aarcange@redhat.com>,
Thomas Gleixner <tglx@linutronix.de>,
Linus Torvalds <torvalds@linux-foundation.org>,
Dave Jones <davej@redhat.com>,
Darren Hart <dvhart@linux.intel.com>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
Peter Zijlstra <peterz@infradead.org>,
Martin Schwidefsky <schwidefsky@de.ibm.com>,
Heiko Carstens <heiko.carstens@de.ibm.com>
Subject: Re: [PATCH v2 1/1] mm: fix the theoretical compound_lock() vs prep_new_page() race
Date: Wed, 8 Jan 2014 18:02:02 +0000 [thread overview]
Message-ID: <20140108180202.GL27046@suse.de> (raw)
In-Reply-To: <20140108161338.GA10434@redhat.com>
On Wed, Jan 08, 2014 at 05:13:38PM +0100, Oleg Nesterov wrote:
> On 01/08, Mel Gorman wrote:
> >
> > On Sat, Jan 04, 2014 at 05:43:47PM +0100, Oleg Nesterov wrote:
> > >
> > > get/put_page(thp_tail) paths do get_page_unless_zero(page_head) +
> > > compound_lock(). In theory this page_head can be already freed and
> > > reallocated as alloc_pages(__GFP_COMP, smaller_order). In this case
> > > get_page_unless_zero() can succeed right after set_page_refcounted(),
> > > and compound_lock() can race with the non-atomic __SetPageHead() in
> > > prep_compound_page().
> > >
> > This patch is putting a write barrier in the page allocator fast path and
> > that is going to be a leading cause of Sad Face. We already have seen
> > large regressions before when write barriers were introduced to the page
> > allocator paths for cpusets. Sticking it under CONFIG_TRANSPARENT_HUGEPAGE
> > does not really address the issue.
>
> As you already mentioned in another email, smp_wmb() is mostly nop. On
> x86_64 at least.
Which sometimes means that it'll just take longer for someone to find it
and bitch about it.
> Although perhaps it would be nice to have
>
> static inline void atomic_store_release(atomic_t *v, int i)
> {
> smp_store_release(&v->counter, i);
> }
>
> > > Yes, but thp can access this page_head via stale pointer, tail->first_page,
> > > if it races with split_huge_page_refcount().
> >
> > To justify the introduction of a performance regression we need to be 100%
> > sure this race actually exists
>
> See below. But let me remind that I never looked at this code before,
> I can be easily wrong.
>
> > and not just theoretical.
>
> It is theoretical anyway, I guess.
>
> > For futex, the THP page (and the tail) must have been discovered via
> > the page tables in which case the page tables are temporarily preventing
> > the page being freed to the allocator.
>
> Yes. But, for example, get_futex_key() does
>
> if (unlikely(PageTail(page))) {
> put_page(page);
>
> why this put_page() can't race with _split? If nothing else, another thread
> can unmap the part of this vma.
>
The race is not prevented but that does not mean it matters. Basic
scenario where a split starts after the PageTail check but before the
put_page in get_futex_key
CPU A
get_futex_key
-> fast gup, page table removing prevents parallel unmap and free
-> gup_huge_pmd (arch/x86/mm/gup.c at least)
-> get_huge_page_tail (increment page tail _map_count)
-> get_huge_page_multiple (increment ref on head page)
-> Check PageTail
CPU B
split_huge_page_to_list
-> split_huge_page_refcount
spin_lock_irq(lru_lock)
compound_lock
-> put_page(tail_page)
->put_compound_page
looks up head page
takes reference unless zero
compound_lock (block)
complete split
compound_unlock
check PageTail
This put_page blocks on the compound lock, finds the page is no longer a
PageTail as the split barriers correctly and backs out gracefully. So sure,
splits can race but the case is cared for.
The parallel unmap is prevented by get_huge_page_multiple in the gup path
and held in place until put_page_compound frees it later.
I still don't see the case where a page gets freed to the page allocator
that causes weird problems here. Unfortunately, I also recognise I have
tunnel vision because subconsciously I don't *want* to see a problem here
that justifies adding a write barrier. Andrea may stomp all over this
reasoning in which case we'll get a good comment for the smp_wmb :/
--
Mel Gorman
SUSE Labs
next prev parent reply other threads:[~2014-01-08 18:02 UTC|newest]
Thread overview: 114+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-12-10 15:47 process 'stuck' at exit Dave Jones
2013-12-10 18:40 ` Linus Torvalds
2013-12-10 19:18 ` Thomas Gleixner
2013-12-10 19:55 ` Linus Torvalds
2013-12-10 20:27 ` Dave Jones
2013-12-10 20:34 ` Thomas Gleixner
2013-12-10 20:55 ` Dave Jones
2013-12-10 21:25 ` Darren Hart
2013-12-10 21:28 ` Thomas Gleixner
2013-12-10 21:39 ` Steven Rostedt
2013-12-10 20:33 ` Thomas Gleixner
2013-12-10 20:43 ` Linus Torvalds
2013-12-10 21:17 ` Thomas Gleixner
2013-12-10 20:35 ` Oleg Nesterov
2013-12-10 20:49 ` Dave Jones
2013-12-10 21:06 ` Darren Hart
2013-12-10 21:12 ` Dave Jones
2013-12-10 21:18 ` Linus Torvalds
2013-12-10 21:24 ` Linus Torvalds
2013-12-10 21:32 ` Dave Jones
2013-12-10 21:49 ` Linus Torvalds
2013-12-10 21:56 ` Dave Jones
2013-12-10 21:59 ` Linus Torvalds
2013-12-10 22:07 ` Dave Jones
2013-12-11 12:45 ` Ingo Molnar
2013-12-10 21:34 ` Oleg Nesterov
2013-12-10 21:41 ` Dave Jones
2013-12-10 21:57 ` Linus Torvalds
2013-12-10 22:02 ` Dave Jones
2013-12-10 22:09 ` Oleg Nesterov
2013-12-10 22:19 ` Linus Torvalds
2013-12-10 22:33 ` Linus Torvalds
2013-12-10 22:38 ` Darren Hart
2013-12-10 22:57 ` Thomas Gleixner
2013-12-10 23:05 ` Linus Torvalds
2013-12-10 23:28 ` Thomas Gleixner
2013-12-10 23:31 ` Al Viro
2013-12-11 17:08 ` Oleg Nesterov
2013-12-11 17:18 ` Thomas Gleixner
2013-12-11 17:56 ` Oleg Nesterov
2013-12-11 19:18 ` PATCH? introduce get_compound_page (Was: process 'stuck' at exit) Oleg Nesterov
2013-12-13 15:10 ` Andrea Arcangeli
2013-12-13 16:22 ` Oleg Nesterov
2013-12-13 17:34 ` Andrea Arcangeli
2013-12-16 18:36 ` Oleg Nesterov
2013-12-16 20:19 ` Andrea Arcangeli
2013-12-16 20:46 ` Oleg Nesterov
2013-12-17 16:53 ` Oleg Nesterov
2013-12-17 18:06 ` Andrea Arcangeli
2013-12-18 19:19 ` [PATCH -mm 0/7] (Was: introduce get_compound_page) Oleg Nesterov
2013-12-18 19:19 ` [PATCH -mm 1/7] mm: thp: introduce __put_nontail_page() Oleg Nesterov
2013-12-18 19:19 ` [PATCH -mm 2/7] mm: thp: change __get_page_tail() to use __put_nontail_page() Oleg Nesterov
2013-12-18 19:19 ` [PATCH -mm 3/7] mm: change release_pages() to use put_page() rather than put_compound_page() Oleg Nesterov
2013-12-18 19:19 ` [PATCH -mm 4/7] mm: thp: turn put_compound_page() into __put_page_tail() Oleg Nesterov
2013-12-18 19:36 ` Peter Zijlstra
2013-12-18 19:50 ` Oleg Nesterov
2013-12-18 19:20 ` [PATCH -mm 5/7] mm: thp: reorganize out_put_single code in __put_page_tail() Oleg Nesterov
2013-12-18 19:20 ` [PATCH -mm 6/7] mm: thp: introduce get_lock_thp_head() Oleg Nesterov
2013-12-18 21:37 ` Linus Torvalds
2013-12-19 16:29 ` Oleg Nesterov
2013-12-18 19:20 ` [PATCH -mm 7/7] mm: thp: introduce compound_head_put_tail(), change get_futex_key() to use it Oleg Nesterov
2013-12-18 19:28 ` Peter Zijlstra
2013-12-18 19:40 ` Oleg Nesterov
2013-12-19 19:08 ` [PATCH 0/1] mm: fix the theoretical compound_lock() vs prep_new_page() race Oleg Nesterov
2013-12-19 19:09 ` [PATCH 1/1] " Oleg Nesterov
2013-12-23 11:43 ` Andrea Arcangeli
2014-01-03 19:55 ` [PATCH v2 0/1] " Oleg Nesterov
2014-01-03 19:55 ` [PATCH v2 1/1] " Oleg Nesterov
2014-01-03 21:00 ` Andrew Morton
2014-01-04 16:43 ` Oleg Nesterov
2014-01-08 11:54 ` Mel Gorman
2014-01-08 13:14 ` Mel Gorman
2014-01-08 16:13 ` Oleg Nesterov
2014-01-08 18:02 ` Mel Gorman [this message]
2014-01-08 19:04 ` Oleg Nesterov
2014-01-09 11:27 ` Mel Gorman
2014-01-09 14:04 ` Oleg Nesterov
2014-01-09 18:52 ` Andrea Arcangeli
2014-01-09 19:43 ` Oleg Nesterov
2014-01-09 21:36 ` Andrea Arcangeli
2014-01-10 16:12 ` Oleg Nesterov
2014-01-10 16:50 ` Peter Zijlstra
2014-01-10 16:12 ` Mel Gorman
2013-12-20 14:19 ` [PATCH 0/1] " Martin Schwidefsky
2013-12-16 20:19 ` [PATCH 0/2] mm: thp: get_huge_page_tail() cleanups Oleg Nesterov
2013-12-16 20:19 ` [PATCH -mm 1/2] mm: thp: __get_page_tail_foll() can use get_huge_page_tail() Oleg Nesterov
2013-12-16 20:19 ` [PATCH -mm 2/2] mm: thp: turn compound_head() into BUG_ON(!PageTail) in get_huge_page_tail() Oleg Nesterov
2013-12-16 20:27 ` [PATCH 0/2] mm: thp: get_huge_page_tail() cleanups Andrea Arcangeli
2013-12-10 22:42 ` process 'stuck' at exit Thomas Gleixner
2013-12-10 22:48 ` Linus Torvalds
2013-12-10 22:58 ` Darren Hart
2013-12-10 23:01 ` Dave Jones
2013-12-10 23:00 ` Dave Jones
2013-12-11 0:05 ` Steven Rostedt
2013-12-11 0:23 ` Dave Jones
2013-12-11 0:55 ` Dave Jones
2013-12-14 20:17 ` Oleg Nesterov
2013-12-11 4:09 ` Dave Jones
2013-12-12 4:26 ` Dave Jones
2013-12-12 5:29 ` Darren Hart
2013-12-10 22:51 ` Darren Hart
2013-12-10 23:24 ` Al Viro
2013-12-11 16:31 ` Mel Gorman
2013-12-11 16:38 ` Thomas Gleixner
2013-12-11 17:57 ` Mel Gorman
2013-12-12 19:00 ` Andrea Arcangeli
2013-12-12 19:03 ` Linus Torvalds
2013-12-10 22:09 ` Steven Rostedt
2013-12-10 22:16 ` Dave Jones
2013-12-10 22:21 ` Steven Rostedt
2013-12-10 22:27 ` Dave Jones
2013-12-11 1:02 ` Mel Gorman
2013-12-10 20:57 ` Darren Hart
2013-12-10 21:09 ` Dave Jones
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=20140108180202.GL27046@suse.de \
--to=mgorman@suse.de \
--cc=aarcange@redhat.com \
--cc=akpm@linux-foundation.org \
--cc=davej@redhat.com \
--cc=dvhart@linux.intel.com \
--cc=heiko.carstens@de.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=oleg@redhat.com \
--cc=peterz@infradead.org \
--cc=schwidefsky@de.ibm.com \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.org \
/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
Powered by JetHome