From: "Kirill A. Shutemov" <kirill@shutemov.name>
To: Martin Tippmann <martin.tippmann@gmail.com>
Cc: linux-mm@kvack.org
Subject: Re: Bug Report: BUG: Bad rss-counter state mm:ffff88101705f800 idx:1 val:512 / application segfaults / thp
Date: Tue, 15 Dec 2015 02:29:46 +0200 [thread overview]
Message-ID: <20151215002946.GA8551@node.shutemov.name> (raw)
In-Reply-To: <CABL_Pd8xJny4h01TZ05Cd0qeWHfeAz1Eaqoz2ceWCbr2wFTdUA@mail.gmail.com>
On Mon, Dec 14, 2015 at 09:19:09PM +0100, Martin Tippmann wrote:
> Hi,
>
> I'm seeing random application crashes (SIGSEV) and after a few minutes
> this appears in the logfiles:
>
> [133933.729199]
> /build/linux-lts-wily-4x6IId/linux-lts-wily-4.2.0/mm/pgtable-generic.c:33:
> bad pmd ffff880fd06d6200(000000018da009e2)
Coould you put dump_stack() into pmd_clear_bad() after pmd_ERROR(). It can
give some clue.
> [133933.763015] BUG: Bad rss-counter state mm:ffff88101705f800 idx:1 val:512
> [133933.763039] BUG: non-zero nr_ptes on freeing mm: 1
>
> I'm quite certain that it's not a hardware error. The problems appears
> regularly on random machines of a 100+ machine cluster of Dell
> PowerEdge R720 servers with 2xXeon E5 (NUMA) and 64GB ECC Memory.
>
> The workload is mostly Hadoop YARN with MapReduce and Spark, the JVM
> (mostly from the DataNodes) crashes randomly under load with SIGSEV.
>
> The problems appears with Kernel 4.3.0 and 4.2.7 from Ubuntu Kernel
> Mainline PPA[1] and with the current 4.2 Ubuntu Wily Kernel - all of
> these kernels already have a related patch[2].
>
> However I'm still seeing the problem. The bug disappears when I
> disable transparent hugepages and reboot the machines!
>
> Before disabling transparent hugepages completely I ran this config:
>
> echo always > /sys/kernel/mm/transparent_hugepage/enabled
> echo never > /sys/kernel/mm/transparent_hugepage/defrag
>
> Unfortunately I can't provide any more data at the moment. Maybe I'm
> able to compile a kernel with debug options turned on over the
> holidays - if you have any hints where I can help to pin this down
> please tell me. On IRC
> CONFIG_DEBUG_VM was recommend.
>
> regards and thanks
> Martin
>
> 1: http://kernel.ubuntu.com/~kernel-ppa/mainline/?C=M;O=D
> 2: https://kernel.googlesource.com/pub/scm/linux/kernel/git/stable/linux-stable.git/+/47aee4d8e314384807e98b67ade07f6da476aa75
>
> --
> 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>
--
Kirill A. Shutemov
--
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>
prev parent reply other threads:[~2015-12-15 0:29 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-12-14 20:19 Bug Report: BUG: Bad rss-counter state mm:ffff88101705f800 idx:1 val:512 / application segfaults / thp Martin Tippmann
2015-12-15 0:29 ` Kirill A. Shutemov [this message]
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=20151215002946.GA8551@node.shutemov.name \
--to=kirill@shutemov.name \
--cc=linux-mm@kvack.org \
--cc=martin.tippmann@gmail.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;
as well as URLs for NNTP newsgroup(s).