cancel
Showing results for 
Search instead for 
Did you mean: 

Archives Discussions

ThomasI
Journeyman III

Linux: Segfault calInit() as user

Hi,

 

we are running an AMD FireStream 9170 in a SLES10 SP 1 x64 system.

Unfortunately all Cal (and brook) programs segfault in calInit when they are ran as user while they execute correctly as root.

The user does have an authenticated connection to the local X server. The user is connected with ssh but DISPLAY=:0 and xauth added. LD_LIBRARY_PATH is the same for user and root-

Here is a debug output:

(gdb) run

Starting program: /usr/local/amdcal/bin/lnx64/FindNumDevices

(no debugging symbols found) x7

[Thread debugging using libthread_db enabled]

[New Thread 47708956238512 (LWP 6299)]

(no debugging symbols found) x5


Program received signal SIGSEGV, Segmentation fault.

[Switching to Thread 47708956238512 (LWP 6299)]

0x00002b641a1c7d9f in calclImageGetSize () from /usr/local/amdcal/lib64/libamdcalrt.so

(gdb) bt

#0 0x00002b641a1c7d9f in calclImageGetSize () from /usr/local/amdcal/lib64/libamdcalrt.so

#1 0x00002b641a1c4689 in calclImageGetSize () from /usr/local/amdcal/lib64/libamdcalrt.so

#2 0x00002b641a1c3e70 in calclImageGetSize () from /usr/local/amdcal/lib64/libamdcalrt.so

#3 0x00002b641a1c2b2e in calclImageGetSize () from /usr/local/amdcal/lib64/libamdcalrt.so

#4 0x00002b641a1b839a in calclImageGetSize () from /usr/local/amdcal/lib64/libamdcalrt.so

#5 0x00002b641a0ffb38 in calclImageGetSize () from /usr/local/amdcal/lib64/libamdcalrt.so

#6 0x00002b641a134139 in calclImageGetSize () from /usr/local/amdcal/lib64/libamdcalrt.so

#7 0x00002b641a1cfef8 in calclImageGetSize () from /usr/local/amdcal/lib64/libamdcalrt.so

#8 0x00002b641a1cce8d in calclImageGetSize () from /usr/local/amdcal/lib64/libamdcalrt.so

#9 0x00002b641a1cc429 in calclImageGetSize () from /usr/local/amdcal/lib64/libamdcalrt.so

#10 0x00002b641a1d3a45 in calInit () from /usr/local/amdcal/lib64/libamdcalrt.so

#11 0x0000000000402a4e in calclImageGetSize ()

#12 0x00002b641b0ba154 in __libc_start_main () from /lib64/libc.so.6

#13 0x0000000000401c2a in calclImageGetSize ()

#14 0x00007fff90b29898 in ?? ()

#15 0x0000000000000000 in ?? ()

 

Any kind of hints where to find the problem would be very appreciated.

0 Likes
3 Replies
dholt
Journeyman III

Having an identical problem here (FireStream 9170/SLES10sp1 64-bit).

0 Likes
marcr
Staff


Hi,

I was able to replicate this in our lab on RHEL 5.1, and filed a bug against it with engineering. I'll post any workaround/update here. -- marcr
0 Likes

I solved the problem on my end by adding users to the group 'video'. /dev/dri/card0 seems to be the culprit, the user needs to have rw permissions.

The FireStream card is on the gold5 system, this method allowed me to work from my machine (aurora).  I guess an X session has to be running locally at all times:

[dholt@aurora ~]$ ssh gold5
[dholt@gold5 ~]$ export DISPLAY=gold5:0.0
[dholt@gold5 ~]$ startx >/dev/null 2>&1 &
[dholt@gold5 ~]$ xhost +
[dholt@gold5 ~]$ /usr/local/amdcal/bin/lnx64/FindNumDevices
CAL initialized.

Finding out number of devices :-
Device Count = 1

CAL shutdown successful.

 

0 Likes