vglrun gives a libc segfault

24 views
Skip to first unread message

Jan-Albert van Ree

unread,
Jul 25, 2018, 2:56:21 AM7/25/18
to VirtualGL User Discussion/Support
Hello,

We've been running VirtualGL 2.5.2 on several HPC login nodes for a year, without any issues. Machines run Scientific Linux 7.3 , kernel 3.10.0-327.10.1
Inside the systems is an nVidia Quadro K2200 with the 367.48 nVidia driver.

Now we are deploying a new cluster with newer hardware thus we have to use a newer kernel and newer nVidia drivers, and that's were all falls apart...
Kernel is now 3.10.0.862.6.3 , nVidia Quadro P4000 cards with the nVidia 390.46 driver

No matter what program we try to start with vglrun, we get a segfault such as
[78457.806124] glxinfo[163647]: segfault at 2aaaafb0e000 ip 00002aaaab8b8521 sp 00007fffffffcc88 error 6 in libc-2.17.so[2aaaab75d000+1c3000]

Different runs with different programs all produce a segfault, were the ip address location is always the same, only the first address and the sp location change, as well as the address in brackets at the end of the line

We tried upgrading to the latest beta build of VirtualGL but that did not solve the issue , it just slightly changes all the addresses
[77279.757070] glxinfo[155176]: segfault at 2aaaafb23000 ip 00002aaaab8cd521 sp 00007fffffffc298 error 6 in libc-2.17.so[2aaaab772000+1c3000]

Can anybody suggest a solution?
Regards,
--
Jan-Albert van Ree

DRC

unread,
Jul 25, 2018, 12:00:26 PM7/25/18
to virtual...@googlegroups.com
Does the segfault happen without VirtualGL? Can you run glxinfo on the
3D X server successfully without vglrun?

Can you verify whether downgrading the nVidia driver fixes the problem?

Jan-Albert van Ree

unread,
Jul 26, 2018, 5:18:09 AM7/26/18
to VirtualGL User Discussion/Support
Good morning,

To answer the first question : without VirtualGL everything was working fine, glxgears, glxinfo etc.

After some serious testing I've finally found a working combination.

RedHat/Scientific Linux glibc 2.17 build 222 appears to give compile problems with any nVidia driver older dan release 384 (I settled on 384.130)
VirtualGL won't work with any nVidia driver newer than release 384 (at least on these systems...)
So it appears the nVidia driver 390 introduced something which breaks the VirtualGL interface.

If the VirtualGL developers want more info or to have more scenario's tested to further debug the issue contact me, I'd be happy to help as far as I can.

Again thanks for the help!
--
Jan-Albert

Op woensdag 25 juli 2018 18:00:26 UTC+2 schreef DRC:

DRC

unread,
Aug 9, 2018, 9:19:19 PM8/9/18
to virtual...@googlegroups.com
I am using 390.xx on RHEL 6.9 and 7.5 with no issues, so unfortunately I
can't reproduce the problem.
> > libc-2.17.so <http://libc-2.17.so>[2aaaab75d000+1c3000]
> >
> > Different runs with different programs all produce a segfault,
> were the
> > ip address location is always the same, only the first address and
> the
> > sp location change, as well as the address in brackets at the end
> of the
> > line
> >
> > We tried upgrading to the latest beta build of VirtualGL but that did
> > not solve the issue , it just slightly changes all the addresses
> > [77279.757070] glxinfo[155176]: segfault at 2aaaafb23000 ip
> > 00002aaaab8cd521 sp 00007fffffffc298 error 6 in
> > libc-2.17.so <http://libc-2.17.so>[2aaaab772000+1c3000]
> >
> > Can anybody suggest a solution?
> > Regards,
> > --
> > Jan-Albert van Ree
>
> --
> You received this message because you are subscribed to the Google
> Groups "VirtualGL User Discussion/Support" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to virtualgl-use...@googlegroups.com
> <mailto:virtualgl-use...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/virtualgl-users/849991df-ce8b-4f68-9699-ea9898aab38f%40googlegroups.com
> <https://groups.google.com/d/msgid/virtualgl-users/849991df-ce8b-4f68-9699-ea9898aab38f%40googlegroups.com?utm_medium=email&utm_source=footer>.
> For more options, visit https://groups.google.com/d/optout.
Reply all
Reply to author
Forward
0 new messages