From: "Kirill A. Shutemov" <kirill@shutemov.name>
To: Linus Torvalds <torvalds@linux-foundation.org>,
Peter Zijlstra <peterz@infradead.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>,
Dave Hansen <dave.hansen@linux.intel.com>,
Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>,
"Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
Andrew Morton <akpm@linux-foundation.org>,
Johannes Weiner <hannes@cmpxchg.org>,
Joonsoo Kim <iamjoonsoo.kim@lge.com>,
Mel Gorman <mgorman@techsingularity.net>,
Tony Luck <tony.luck@intel.com>, Vlastimil Babka <vbabka@suse.cz>,
Michal Hocko <mhocko@kernel.org>,
"hillf.zj" <hillf.zj@alibaba-inc.com>,
Hugh Dickins <hughd@google.com>, Oleg Nesterov <oleg@redhat.com>,
Rik van Riel <riel@redhat.com>,
Srikar Dronamraju <srikar@linux.vnet.ibm.com>,
Vladimir Davydov <vdavydov.dev@gmail.com>,
Ingo Molnar <mingo@kernel.org>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
linux-mm <linux-mm@kvack.org>,
the arch/x86 maintainers <x86@kernel.org>
Subject: Re: [mm 4.15-rc8] Random oopses under memory pressure.
Date: Fri, 19 Jan 2018 02:49:55 +0300 [thread overview]
Message-ID: <20180118234955.nlo55rw2qsfnavfm@node.shutemov.name> (raw)
In-Reply-To: <CA+55aFy43ypm0QvA5SqNR4O0ZJETbkR3NDR=dnSdvejc_nmSJQ@mail.gmail.com>
On Thu, Jan 18, 2018 at 09:26:25AM -0800, Linus Torvalds wrote:
> On Thu, Jan 18, 2018 at 8:56 AM, Kirill A. Shutemov
> <kirill@shutemov.name> wrote:
> >
> > I can't say I fully grasp how 'diff' got this value and how it leads to both
> > checks being false.
>
> I think the problem is that page difference when they are in different sections.
>
> When you do
>
> pte_page(*pvmw->pte) - pvmw->page
>
> then the compiler takes the pointer difference, and then divides by
> the size of "struct page" to get an index.
>
> But - and this is important - it does so knowing that the division it
> does will have no modulus: the two 'struct page *' pointers are really
> in the same array, and they really are 'n*sizeof(struct page)' apart
> for some 'n'.
>
> That means that the compiler can optimize the division. In fact, for
> this case, gcc will generate
>
> subl %ebx, %eax
> sarl $3, %eax
> imull $-858993459, %eax, %eax
>
> because 'struct page' is 40 bytes in size, and that magic sequence
> happens to divide by 40 (first divide by 8, then that magical "imull"
> will divide by 5 *IFF* the thing is evenly divisible by 5 (and not too
> big - but the shift guarantees that).
>
> Basically, it's a magic trick, because real divides are very
> expensive, but you can fake them more quickly if you can limit the
> input domain.
>
> But what does it mean if the two "struct page *" are not in the same
> array, and the two arrays were allocated not aligned exactly 40 bytes
> away, but some random number of pages away?
>
> You get *COMPLETE*GARBAGE* when you do the above optimized divide.
> Suddenly the divide had a modulus (because the base of the two arrays
> weren't 40-byte aligned), and the "trick" doesn't work.
>
> So that's why you can't do pointer diffs between two arrays. Not
> because you can't subtract the two pointers, but because the
> *division* part of the C pointer diff rules leads to issues.
Thanks a lot for the explanation!
I wounder if this may be a problem in other places?
For instance, perf uses address of a mutex to determinate the lock
ordering. See mutex_lock_double(). The mutex is embedded into struct
perf_event_context, which is allocated with kzalloc() so I don't see how
we can presume that alignment is consistent between them.
I don't think it's the only example in kernel. Are we just lucky?
--
Kirill A. Shutemov
next prev parent reply other threads:[~2018-01-18 23:50 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <201801112311.EHI90152.FLJMQOStVHFOFO@I-love.SAKURA.ne.jp>
[not found] ` <20180111142148.GD1732@dhcp22.suse.cz>
[not found] ` <201801120131.w0C1VJUN034283@www262.sakura.ne.jp>
[not found] ` <CA+55aFx4pH4odYDfuGemm5TK-CS4E8pL_ipHCVzVBmsQkyWp1Q@mail.gmail.com>
[not found] ` <201801122022.IDI35401.VOQOFOMLFSFtHJ@I-love.SAKURA.ne.jp>
2018-01-14 11:54 ` [mm 4.15-rc7] Random oopses under memory pressure Tetsuo Handa
2018-01-15 23:05 ` Linus Torvalds
2018-01-16 1:15 ` [mm 4.15-rc8] " Tetsuo Handa
2018-01-16 2:14 ` Linus Torvalds
2018-01-16 8:06 ` Dave Hansen
2018-01-16 8:37 ` Ingo Molnar
2018-01-16 19:30 ` Linus Torvalds
2018-01-16 17:33 ` Tetsuo Handa
2018-01-16 19:34 ` Linus Torvalds
2018-01-17 11:08 ` Tetsuo Handa
2018-01-17 21:39 ` Linus Torvalds
2018-01-17 21:51 ` Linus Torvalds
2018-01-17 22:04 ` Dave Hansen
2018-01-17 22:00 ` Dave Hansen
2018-01-17 22:15 ` Linus Torvalds
2018-01-18 8:12 ` Tetsuo Handa
2018-01-18 12:25 ` Kirill A. Shutemov
2018-01-18 13:12 ` Kirill A. Shutemov
2018-01-18 14:34 ` Kirill A. Shutemov
2018-01-18 14:38 ` Dave Hansen
2018-01-18 14:45 ` Kirill A. Shutemov
2018-01-18 14:51 ` Dave Hansen
2018-01-18 16:58 ` Linus Torvalds
2018-01-18 14:45 ` Dave Hansen
2018-01-18 14:58 ` Andrea Arcangeli
2018-01-18 16:56 ` Kirill A. Shutemov
2018-01-18 17:26 ` Luck, Tony
2018-01-18 17:28 ` Linus Torvalds
2018-01-18 17:26 ` Linus Torvalds
2018-01-18 23:49 ` Kirill A. Shutemov [this message]
2018-01-19 12:55 ` Matthew Wilcox
2018-01-19 18:42 ` Linus Torvalds
2018-01-19 22:12 ` Al Viro
2018-01-19 22:53 ` Linus Torvalds
2018-01-20 2:02 ` Al Viro
2018-01-20 5:24 ` Al Viro
2018-01-20 9:38 ` Luc Van Oostenryck
2018-01-22 13:26 ` Rasmus Villemoes
2018-01-22 19:58 ` Linus Torvalds
2018-01-18 15:40 ` Kirill A. Shutemov
2018-01-18 17:22 ` Michal Hocko
2018-01-19 10:02 ` Kirill A. Shutemov
2018-01-19 10:33 ` Michal Hocko
2018-01-19 11:49 ` Kirill A. Shutemov
2018-01-19 12:07 ` Michal Hocko
2018-01-19 12:30 ` Kirill A. Shutemov
2018-01-19 2:01 ` Tetsuo Handa
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=20180118234955.nlo55rw2qsfnavfm@node.shutemov.name \
--to=kirill@shutemov.name \
--cc=aarcange@redhat.com \
--cc=akpm@linux-foundation.org \
--cc=dave.hansen@linux.intel.com \
--cc=hannes@cmpxchg.org \
--cc=hillf.zj@alibaba-inc.com \
--cc=hughd@google.com \
--cc=iamjoonsoo.kim@lge.com \
--cc=kirill.shutemov@linux.intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mgorman@techsingularity.net \
--cc=mhocko@kernel.org \
--cc=mingo@kernel.org \
--cc=oleg@redhat.com \
--cc=penguin-kernel@i-love.sakura.ne.jp \
--cc=peterz@infradead.org \
--cc=riel@redhat.com \
--cc=srikar@linux.vnet.ibm.com \
--cc=tony.luck@intel.com \
--cc=torvalds@linux-foundation.org \
--cc=vbabka@suse.cz \
--cc=vdavydov.dev@gmail.com \
--cc=x86@kernel.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;
as well as URLs for NNTP newsgroup(s).