kernel-packages team mailing list archive

Thread
Date
[Bug 1483343] Comment bridged from LTC Bugzilla

To: kernel-packages@xxxxxxxxxxxxxxxxxxx
From: bugproxy <bugproxy@xxxxxxxxxx>
Date: Sat, 16 Jan 2016 10:11:30 -0000
Reply-to: Bug 1483343 <1483343@xxxxxxxxxxxxxxxxxx>
Sender: bounces@xxxxxxxxxxxxx
------- Comment From pavsubra@xxxxxxxxxx 2016-01-16 05:05 EDT-------
I have copied lock_torture script on to the P8 server with the 3.13 kernel installed.
Then executed the test script again in-order to re-create the scenario.

root@powerkvmpok002:/opt/ltp/testcases/bin# ./lock_torture.sh
lock_torture 1 TCONF : Test requires locktorture module
lock_torture 1 TINFO : cleanup

root@powerkvmpok002:/opt/ltp/testcases/bin# modprobe locktorture
modprobe: FATAL: Module locktorture not found.

As can be seen above the locktorture module is required in 3.13 kernel.
I think, this module is not supported in the 3.13 kernel.

-- 
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/1483343

Title:
  NMI watchdog: BUG: soft lockup errors when we execute lock_torture_wr
  tests

Status in linux package in Ubuntu:
  Triaged
Status in linux source package in Vivid:
  In Progress

Bug description:
  ---Problem Description---
  NMI watchdog: BUG: soft lockup errors when we execute lock_torture_wr tests
    
  ---uname output---
  Linux alp15 3.19.0-18-generic #18~14.04.1-Ubuntu SMP Wed May 20 09:40:36 UTC 2015 ppc64le ppc64le ppc64le GNU/Linux
   
  Machine Type = P8 
    
  ---Steps to Reproduce---
  Install a P8 Power VM LPAR with Ubuntu 14.04.2 ISO.
  Then install the Ubuntu 14.04.3 kernel on the same and reboot.
  Then compile and build the LTP latest test suites on the same.

  root@alp15:~# tar -xvf ltp-full-20150420.tar.bz2
  root@alp15:~# cd ltp-full-20150420/
  root@alp15:~/ltp-full-20150420# ls
  aclocal.m4      configure     execltp.in  install-sh  Makefile          README                runltplite.sh    testcases    utils
  autom4te.cache  configure.ac  IDcheck.sh  lib         Makefile.release  README.kernel_config  runtest          testscripts  ver_linux
  config.guess    COPYING       include     ltpmenu     missing           runalltests.sh        scenario_groups  TODO         VERSION
  config.sub      doc           INSTALL     m4          pan               runltp                scripts          tools
  root@alp15:~/ltp-full-20150420# ./configure
  root@alp15:~/ltp-full-20150420# make
  root@alp15:~/ltp-full-20150420# make install

  root@alp15:/opt/ltp/testcases/bin# ./lock_torture.sh
  lock_torture 1 TINFO : estimate time 6.00 min
  lock_torture 1 TINFO : spin_lock: running 60 sec...

  Message from syslogd@alp15 at Thu Jun 18 01:23:32 2015 ...
  alp15 vmunix: [  308.034386] NMI watchdog: BUG: soft lockup - CPU#10 stuck for 21s! [lock_torture_wr:2337]

  Message from syslogd@alp15 at Thu Jun 18 01:23:32 2015 ...
  alp15 vmunix: [  308.034389] NMI watchdog: BUG: soft lockup - CPU#6 stuck for 22s! [lock_torture_wr:2331]

  Message from syslogd@alp15 at Thu Jun 18 01:23:32 2015 ...
  alp15 vmunix: [  308.034394] NMI watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [lock_torture_wr:2339]

  Message from syslogd@alp15 at Thu Jun 18 01:23:32 2015 ...
  alp15 vmunix: [  308.034396] NMI watchdog: BUG: soft lockup - CPU#3 stuck for 22s! [lock_torture_wr:2346]

  Message from syslogd@alp15 at Thu Jun 18 01:23:32 2015 ...
  alp15 vmunix: [  308.034398] NMI watchdog: BUG: soft lockup - CPU#7 stuck for 21s! [lock_torture_wr:2334]

  Message from syslogd@alp15 at Thu Jun 18 01:23:32 2015 ...
  alp15 vmunix: [  308.034410] NMI watchdog: BUG: soft lockup - CPU#11 stuck for 22s! [lock_torture_wr:2321]

  Message from syslogd@alp15 at Thu Jun 18 01:23:32 2015 ...
  alp15 vmunix: [  308.034412] NMI watchdog: BUG: soft lockup - CPU#9 stuck for 22s! [lock_torture_wr:2333]

  Message from syslogd@alp15 at Thu Jun 18 01:23:32 2015 ...
  alp15 vmunix: [  308.038386] NMI watchdog: BUG: soft lockup - CPU#14 stuck for 22s! [lock_torture_wr:2327]

   
  Stack trace output:
   root@alp15:~# dmesg | more
  [ 1717.146881] lock_torture_wr R  running task
  [ 1717.146881]
  [ 1717.146885]     0  2555      2 0x00000804
  [ 1717.146887] Call Trace:
  [ 1717.146894] [c000000c7551b820] [c000000c7551b860] 0xc000000c7551b860 (unreliable)
  [ 1717.146899] [c000000c7551b860] [c0000000000b4fb0] __do_softirq+0x220/0x3b0
  [ 1717.146904] [c000000c7551b960] [c0000000000b5478] irq_exit+0x98/0x100
  [ 1717.146909] [c000000c7551b980] [c00000000001fa54] timer_interrupt+0xa4/0xe0
  [ 1717.146913] [c000000c7551b9b0] [c000000000002758] decrementer_common+0x158/0x180
  [ 1717.146922] --- interrupt: 901 at _raw_write_lock+0x68/0xc0
  [ 1717.146922]     LR = torture_rwlock_write_lock+0x28/0x40 [locktorture]
  [ 1717.146927] [c000000c7551bca0] [c000000c7551bcd0] 0xc000000c7551bcd0 (unreliable)
  [ 1717.146934] [c000000c7551bcd0] [d00000000d4810b8] torture_rwlock_write_lock+0x28/0x40 [locktorture]
  [ 1717.146939] [c000000c7551bcf0] [d00000000d480578] lock_torture_writer+0x98/0x210 [locktorture]
  [ 1717.146944] [c000000c7551bd80] [c0000000000da4d4] kthread+0x114/0x140
  [ 1717.146948] [c000000c7551be30] [c00000000000956c] ret_from_kernel_thread+0x5c/0x70
  [ 1717.146951] Task dump for CPU 10:
  [ 1717.146953] lock_torture_wr R  running task        0  2537      2 0x00000804
  [ 1717.146957] Call Trace:
  [ 1717.146961] [c000000c7557b820] [c000000c7557b860] 0xc000000c7557b860 (unreliable)
  [ 1717.146966] [c000000c7557b860] [c0000000000b4fb0] __do_softirq+0x220/0x3b0
  [ 1717.146970] [c000000c7557b960] [c0000000000b5478] irq_exit+0x98/0x100
  [ 1717.146975] [c000000c7557b980] [c00000000001fa54] timer_interrupt+0xa4/0xe0
  [ 1717.146979] [c000000c7557b9b0] [c000000000002758] decrementer_common+0x158/0x180
  [ 1717.146988] --- interrupt: 901 at _raw_write_lock+0x68/0xc0
  [ 1717.146988]     LR = torture_rwlock_write_lock+0x28/0x40 [locktorture]
  [ 1717.146993] [c000000c7557bca0] [c000000c7557bcd0] 0xc000000c7557bcd0 (unreliable)
  [ 1717.147000] [c000000c7557bcd0] [d00000000d4810b8] torture_rwlock_write_lock+0x28/0x40 [locktorture]
  [ 1717.147006] [c000000c7557bcf0] [d00000000d480578] lock_torture_writer+0x98/0x210 [locktorture]
  [ 1717.147013] [c000000c7557bd80] [c0000000000da4d4] kthread+0x114/0x140
  [ 1717.147017] [c000000c7557be30] [c00000000000956c] ret_from_kernel_thread+0x5c/0x70
  [ 1717.147020] Task dump for CPU 17:
  [ 1717.147021] Task dump for CPU 2:
  [ 1717.147028] lock_torture_wr R
  [ 1717.147028] lock_torture_wr R  running task
  [ 1717.147033]   running task        0  2547      2 0x00000804
  [ 1717.147042]     0  2533      2 0x00000804
  [ 1717.147044] Call Trace:
  [ 1717.147045] Call Trace:
  [ 1717.147053] [c000000c732a3820] [c000000c7f688448] 0xc000000c7f688448
  [ 1717.147056] [c000000c7555f820] [c000000c7fa48448] 0xc000000c7fa48448
  [ 1717.147059]  (unreliable)
  [ 1717.147063]  (unreliable)
  [ 1717.147063]
  [ 1717.147067]
  [ 1717.147072] Task dump for CPU 18:
  [ 1717.147073] Task dump for CPU 7:
  [ 1717.147077] lock_torture_wr R  running task
  [ 1717.147082] lock_torture_wr R    0  2555      2 0x00000804
  [ 1717.147088]   running task
  [ 1717.147088] Call Trace:
  [ 1717.147096] [c000000c7551b820] [c000000c7551b860] 0xc000000c7551b860
  [ 1717.147096]     0  2559      2 0x00000804
  [ 1717.147102] Call Trace:
  [ 1717.147105]  (unreliable)

  It is possible that we are missing this commit that fixes a deadlock
  during these tests:

  https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit?id=f548d99ef4f5ec8f7080e88ad07c44d16d058ddc

  will check the Ubuntu source shortly as see if this is the case and we
  can suggest building a kernel to see if it helps.

  The apt-get source linux-image- on the test system didn't pull down
  the sources but the kernel being used is close to the one used for
  vivid (3.19.0-25.26) so I pulled down the git source tree for it with
  git clone git://kernel.ubuntu.com/ubuntu/ubuntu-vivid.git and the
  resulting source shows that the patch for the commit mentioned is not
  applied.

  As I basically understand it, the problem that was fixed is that while
  torture_rwlock_read_lock_irq() acquires a read lock on the lock
  called:

  torture_rwlock

  anything that calls the counterpart torture_rwlock_read_unlock_irq()
  to relinquish the read lock instead ends doing a
  write_unlock_irqrestore() on the torture_rwlock() in essence leaving
  the read lock. So when the locktorture module calls something like
  torture_rwlock_write_lock() as we see in the bug description, it will
  block indefinitely as there is at least one lock reader.

  I'll go ahead and mirror this since I pretty confident this is the
  issue (also should affect Vivid).

  We'll have to figure out how to get the sources for the LTS kernel to
  build a test kernel as well.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1483343/+subscriptions