xfs
[Top] [All Lists]

Re: 3.1-rc4: spectacular kernel errors / filesystem crash

To: Jesse Brandeburg <jesse.brandeburg@xxxxxxxxx>
Subject: Re: 3.1-rc4: spectacular kernel errors / filesystem crash
From: Eric Dumazet <eric.dumazet@xxxxxxxxx>
Date: Tue, 13 Sep 2011 06:05:06 +0200
Cc: Justin Piszcz <jpiszcz@xxxxxxxxxxxxxxx>, linux-kernel@xxxxxxxxxxxxxxx, xfs@xxxxxxxxxxx, Alan Piszcz <ap@xxxxxxxxxxxxx>, NetDEV list <netdev@xxxxxxxxxxxxxxx>
Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=subject:from:to:cc:in-reply-to:references:content-type:date :message-id:mime-version:x-mailer:content-transfer-encoding; bh=o5eBBV0dg78y2F8/JCgmG9MgkRxVqlkRpyaxo9JlpCI=; b=NMWZGQRelRB/dNOka5EkHHmbgYiO9SL5rzE48mwrMj42HR6SPw55z5X3ebJXgRmg/k aqIhKgWmlCTMd2dUinQTuh02AiMYDf2Tugf13CM57QXJca7EJ26xU1pwgHFbYL75XTPg //Qqq46y6TElX7qgksRVtTOPDpjJrTH2eYJiU=
In-reply-to: <CAEuXFEzs1f7n5taYzupux3AtKmRcY4P0m7yjkUQA8aLyS8eujw@xxxxxxxxxxxxxx>
References: <alpine.DEB.2.02.1109110511250.8626@xxxxxxxxxxxxxxxx> <CAEuXFEzs1f7n5taYzupux3AtKmRcY4P0m7yjkUQA8aLyS8eujw@xxxxxxxxxxxxxx>
Le lundi 12 septembre 2011 à 20:59 -0700, Jesse Brandeburg a écrit :
> added netdev because it appears to start with an igb tx hang
> 
> On Sun, Sep 11, 2011 at 2:40 AM, Justin Piszcz <jpiszcz@xxxxxxxxxxxxxxx> 
> wrote:
> > Hi,
> >
> > Over the past 24-48 hours I was running some CPU-intenstive jobs and there
> > was heavy I/O on the RAID (9750-24i4e + a RAID6)..
> >
> > I believe most of the problem started when I included many kernel options as
> > modules (before I only compiled in [*] the drivers I used), there appears to
> > have something to gone awry in the kernel and then afterwards, disks started
> > going in and out, XFS shut down, etcera.
> >
> > I'm opening a case with LSI to see what happened with the 3ware card;
> > however, after a power cycle, everything came back OK (the drives and HW) is
> > physically OK, it is rebuilding onto those two drives with CFG-OP-FAIL but
> > other than that, everything 'seems' OK, still need to do an fsck.
> >
> > Something went wrong in the kernel and caused a cascading effect of errors,
> > this occurred (I believe) when I started to run a lot of encoding jobs;
> > however, I was doing a lot of data transfer for the past 24-48 hours on the
> > RAID array, the system (separate SSD/EXT4) remained unaffected but other
> > weird stuff happened as well..
> >
> > I still see these in the logs as well after the reboot (not often; but e.g.,
> > the RAID controller is rebuilding from the two drives with CFG-OPT-FAIL (the
> > physical drives are 100% healthy):
> >
> > [ 1062.925904] 3w-sas 0000:83:00.0: vpd r/w failed.  This is likely a
> > firmware bug on this device.  Contact the card vendor for a firmware update.
> >
> > So, my plan:
> >
> > 1. Report this error to LKML+XFS mailing lists.
> > 2. Open case with LSI support.
> > 3. Recompile the kernel how I used for many years [only compile in options
> >   that you need [*] and do not compile drivers as modules]
> > 4. Reboot Linux systems and see if this recurs again under the same
> >   workload, after the RAID is done rebuilding.
> >
> > --
> >
> > So these errors are quite long, will upload to HTTP and paste the relevant
> > bits below.
> >
> > --
> >
> > URLs for FULL logs:
> >
> > 1. tw_cli /cX show diag:
> >   http://home.comcast.net/~jpiszcz/20110911/show_diag.txt
> >
> > 2. Full kernel log (and previous morning of kernel crash)
> >   http://home.comcast.net/~jpiszcz/20110911/kern.log.txt
> >
> > 3. tw_cli /cX show all
> >   http://home.comcast.net/~jpiszcz/20110911/cfg-fail.txt
> >
> > --
> >
> > Summary (what seems to have occurred, have not done a full analysis yet)
> >
> > 1. 3ware card freaked out due to kernel/RCU/APIC(?) errors
> >
> > 2. Then, the time source went unstable (this happens with weird kernel bugs
> >   on many different hosts, I have seen this over time).
> >
> > 3. Then, on the 3ward carde, drives started leaving and being re-inserted
> >   by themsevles, XFS went off-line to protect the filesystem due to the
> >   3ware issues
> >
> > --
> >
> > 3ware/RAID-- Interesting errors:
> >
> > I've never seen this before on a 3ware RAID controller, at least from what
> > I can remember and I've been using 3ware cards for many years..
> >
> > p2    CFG-OP-FAIL    -    2.73 TB   SATA  2   -            Hitachi
> > HDS723030AL p3    CFG-OP-FAIL    -    2.73 TB   SATA  3   -
> >  Hitachi HDS723030AL
> >
> > --
> >
> > Kernel/ERRORS:
> >
> > FWIW it all seem to start during an encoding job around 21:00:
> >
> > Sep 10 18:00:00 p34 kernel: [520427.143054] ixgbe 0000:03:00.0: eth6: NIC
> > Link is Down
> > Sep 10 19:20:04 p34 kernel: [525223.256098] 3w-sas: scsi1: AEN: INFO
> > (0x04:0x002B): Verify completed:unit=0.
> > Sep 10 20:59:39 p34 kernel: [531189.671361] ------------[ cut here
> > ]------------
> > Sep 10 20:59:39 p34 kernel: [531189.671376] WARNING: at
> > net/sched/sch_generic.c:255 dev_watchdog+0x23f/0x250()
> > Sep 10 20:59:39 p34 kernel: [531189.671378] Hardware name: X8DTH-i/6/iF/6F
> > Sep 10 20:59:39 p34 kernel: [531189.671380] NETDEV WATCHDOG: eth1 (igb):
> > transmit queue 5 timed out
> > Sep 10 20:59:39 p34 kernel: [531189.671382] Modules linked in: dm_mod
> > tcp_diag parport_pc ppdev lp parport inet_diag pl2303 ftdi_sio snd_usb_audio
> > snd_pcm_oss snd_mixer_oss snd_pcm snd_page_alloc snd_hwdep snd_usbmidi_lib
> > snd_seq_dummy snd_seq_oss snd_seq_midi snd_rawmidi snd_seq_midi_event
> > snd_seq snd_timer snd_seq_device snd soundcore ub cdc_acm usbserial joydev
> > serio_raw nouveau ttm drm_kms_helper drm agpgart i2c_algo_bit mxm_wmi wmi
> > i7core_edac edac_core video
> > Sep 10 20:59:39 p34 kernel: [531189.671414] Pid: 83, comm: ksoftirqd/19 Not
> > tainted 3.1.0-rc4 #1
> > Sep 10 20:59:39 p34 kernel: [531189.671415] Call Trace:
> > Sep 10 20:59:39 p34 kernel: [531189.671424]  [<ffffffff810379ba>]
> > warn_slowpath_common+0x7a/0xb0
> > Sep 10 20:59:39 p34 kernel: [531189.671427]  [<ffffffff81037a91>]
> > warn_slowpath_fmt+0x41/0x50
> > Sep 10 20:59:39 p34 kernel: [531189.671433]  [<ffffffff815d7874>] ?
> > schedule+0x2e4/0x950
> > Sep 10 20:59:39 p34 kernel: [531189.671436]  [<ffffffff814e5aff>]
> > dev_watchdog+0x23f/0x250
> > Sep 10 20:59:39 p34 kernel: [531189.671440]  [<ffffffff81043872>]
> > run_timer_softirq+0xf2/0x220
> > Sep 10 20:59:39 p34 kernel: [531189.671443]  [<ffffffff814e58c0>] ?
> > qdisc_reset+0x50/0x50
> > Sep 10 20:59:39 p34 kernel: [531189.671446]  [<ffffffff8103d208>]
> > __do_softirq+0x98/0x120
> > Sep 10 20:59:39 p34 kernel: [531189.671448]  [<ffffffff8103d345>]
> > run_ksoftirqd+0xb5/0x160
> > Sep 10 20:59:39 p34 kernel: [531189.671454]  [<ffffffff8103d290>] ?
> > __do_softirq+0x120/0x120
> > Sep 10 20:59:39 p34 kernel: [531189.671458]  [<ffffffff810523b7>]
> > kthread+0x87/0x90
> > Sep 10 20:59:39 p34 kernel: [531189.671462]  [<ffffffff815dbdb4>]
> > kernel_thread_helper+0x4/0x10
> > Sep 10 20:59:39 p34 kernel: [531189.671465]  [<ffffffff81052330>] ?
> > kthread_worker_fn+0x130/0x130
> > Sep 10 20:59:39 p34 kernel: [531189.671467]  [<ffffffff815dbdb0>] ?
> > gs_change+0xb/0xb
> > Sep 10 20:59:39 p34 kernel: [531189.671468] ---[ end trace 553dfe731fce91ba
> > ]---
> > Sep 10 20:59:39 p34 kernel: [531189.671478] igb 0000:01:00.1: eth1: Reset
> > adapter
> > Sep 10 20:59:42 p34 kernel: [531192.826058] igb: eth1 NIC Link is Up 1000
> > Mbps Full Duplex, Flow Control: RX/TX
> > Sep 10 21:00:00 p34 kernel: [531210.034506] BUG: soft lockup - CPU#0 stuck
> > for 22s! [kswapd0:947]
> >
> > --
> >
> > URLs for FULL logs:
> >
> > 1. tw_cli /cX show diag:
> >   http://home.comcast.net/~jpiszcz/20110911/show_diag.txt
> >
> > 2. Full kernel log (and previous morning of kernel crash)
> >   http://home.comcast.net/~jpiszcz/20110911/kern.log.txt
> >
> > 3. tw_cli /cX show all
> >   http://home.comcast.net/~jpiszcz/20110911/cfg-fail.txt
> >
> > --
> >
> > Currently...
> >
> > After all of this happened, I stopped all I/O on the system/all processes,
> > etc
> > I shutdown the host, removed the power, powered it back up, now the drives
> > that showed CFG-OP-FAIL before now show as REBUILDING, I am waiting for them
> > to rebuild before doing anything else.
> >
> > Justin.
> >
> >

Please Justin make sure you pulled commit 

commit ed2888e906b56769b4ffabb9c577190438aa68b8
Author: Jon Mason <mason@xxxxxxxx>
Date:   Thu Sep 8 16:41:18 2011 -0500

    PCI: Remove MRRS modification from MPS setting code
    
    Modifying the Maximum Read Request Size to 0 (value of 128Bytes) has
    massive negative ramifications on some devices.  Without knowing which
    devices have this issue, do not modify from the default value when
    walking the PCI-E bus in pcie_bus_safe mode.  Also, make pcie_bus_safe
    the default procedure.
    
    Tested-by: Sven Schnelle <svens@xxxxxxxxxxxxxx>
    Tested-by: Simon Kirby <sim@xxxxxxxxxx>
    Tested-by: Stephen M. Cameron <scameron@xxxxxxxxxxxxxxxxxx>
    Reported-and-tested-by: Eric Dumazet <eric.dumazet@xxxxxxxxx>
    Reported-and-tested-by: Niels Ole Salscheider <niels_ole@salscheider-online.
    References: https://bugzilla.kernel.org/show_bug.cgi?id=42162
    Signed-off-by: Jon Mason <mason@xxxxxxxx>
    Acked-by: Jesse Barnes <jbarnes@xxxxxxxxxxxxxxxx>
    Signed-off-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>


<Prev in Thread] Current Thread [Next in Thread>