Bug#793661: crashes the server
Package: nfs-kernel-server
Version: 1:1.2.8-9
Severity: serious
Ever since upgrading the server from wheezy to jessie, I've been having
crashes every couple of weeks. Running wheezy, it had been stable and
running continuously for over a year. NFS configuration was not changed
when upgrading.
Typically, the crash would occur when the screen was blank and I haven't
previously been able to see any console output from the crashes.
Yesterday, I had the crash twice within 10 minutes and I observed the
messages:
INFO: rcu_sched detected stalls on CPUs/tasks: {} (detected by 1, t=5252
jiffies, g=14894, c=14893, q=583)
INFO: Stall ended before state dump start
BUG: soft lockup - CPU#0 stuck for 23s! [nfsd:1781]
BUG: soft lockup - CPU#1 stuck for 23s! [nfsd:1785]
I found that this crash was repeatable, every time a particular NFS
client ran "git checkout master" in a particular project. The project
in question is quite big, it is on an NFS v3 mount and the filesystem on
the NFS server is btrfs. The client is also running jessie. This crash
happened about five times yesterday.
I logged into the server using ssh and went to the same directory and
ran "git checkout master" and there was no crash. Then I tried again on
the client, checkout out a branch and then checking out master again and
it seems to work now from the client without a crash. I had been in the
same directory a few days ago, checking out different branches, without
any crashes.
I have edited /etc/kbd/config setting
BLANK_TIME=0
so that I can see any messages on the screen when it crashes again.
I've also added kdump-tools on this machine and will try to obtain
further details. Please let me know if there is any specific data I
should try to obtain next time it crashes.
The kernel I am running:
$ uname -a
Linux srv1 3.16.0-4-amd64 #1 SMP Debian 3.16.7-ckt11-1 (2015-05-24)
x86_64 GNU/Linux
It was also crashing with previous jessie kernels.
Reply to: