Centos-6 kernel soft lockup CPU#20 stuck for 67s! [smbd-notifyd] after upgrade form 4.2 to 4.4

[Alexander Bokovoy]

On Mon, 02 May 2016, Dr. Hansjoerg Maurer wrote:
> Hi
> 
> thank you for your fast reply.
> 
> Do you have any further information (e.g. is there already  a bug
> report at Redhat ? which kernel is affected? which subsystem causes
> that problem? ...)
> 
> I can open a bug at RedHat, but I would need more information.
> 
> Sould I open a samba bugzilla entry  in order to  get  it documented
> for other people who run into the sambe problem with samba > 4.2 and
> RHEL-6/Centos-6 (not an exotic plattform :-) )
Well, what you could do if you have RHEL subscription, is to make a bug
at Red Hat for kernel and try to get that fixed. Note that Samba > 4.2
most likely will not be backported to RHEL 6.x in a supported (by Red
Hat) way but if you could find out a way to reproduce it without Samba,
the bug would have good weight to fix -- perhaps it is already fixed
kernel upstream and the question is more of backporting that.

> > > I checked the release notes for 4.3.0  and found, that there have
> > > been changes in the  FileChangeNotify subsystem
> > 
> > Yes, we are using unix domain datagram sockets more excessively now.
> > The kernel does not like it, although it is a pretty standard API.
> > 
> > > Does anybody else has problems in this area.
> > > 
> > > We will try it again with
> > > 
> > >  kernel change notify = no
> > 
> > Won't help. Downgrade Samba to 4.2 and wait for the next CentOS kernel
> > to fix the issue.
Volker, do you know if this is actually a known bug kernel upstream?


-- 
/ Alexander Bokovoy