FC4 T1 hangs on "initializing GLX"

[magneto]

well id suggest compiling your own kernel and try installing 7167 again - fact is being that its super buggy fc4 pre-release you are out there on a limb anyway

if those other rpms deadlock you(they really shouldnt) just boot into init S and remove the rpms via the rpm command - or use "nv" as your driver and go right into gnome and do whatever you have to

[duvvid]

Originally Posted by cdhouch

My xorg.conf works with the new NVIDIA 7167 drivers as long as the "load "glx"" command is commented out. Otherwise it hangs on the initializing glx part of startx.

I'm seeing the same thing on FC4test1.

I noticed the SELinux file contexts were not set correctly after installation. Fixing the file contexts didn't get it working though. Shouldn't the nvidia installer run restorecon on the libs it installs?

Originally Posted by magneto

...fact is being that its super buggy fc4 pre-release ...

The test releases can get pretty ugly, but it's surprisingly good at the moment - as long as you don't need the nvidia 3d drivers.

[Dragoran]

Originally Posted by cdhouch

Ok I upgraded kernels today to the 2.6.11-1.1191_FC4 kernel and re-ran the nvidia bug report script. Here is that output (much cleaner this time) along with /var/log/messages from the time of a boot till the time that X fails to start.

http://brainbug.dnsalias.com/nvidia-bug-report.log
http://brainbug.dnsalias.com/messages.txt

as duvvid said its a selinux issue:

Mar 21 11:30:44 speedybug kernel: audit(1111404619.873:0): avc: denied { execmod } for pid=1172 comm=X path=/usr/X11R6/lib/modules/drivers/nvidia_drv.so dev=dm-0 ino=7668683 scontext=user_u:system_r:initrc_t tcontext=rootbject_r:lib_t tclass=file

are you using the strict or targeted policy?

[cdhouch]

Originally Posted by Dragoran

as duvvid said its a selinux issue:

are you using the strict or targeted policy?

I would have never guessed selinux was the culprit. I'm using targeted.
I'll dig into that and see what I can come up with. To be honest I know next to nothing about selinux (yet).

[Dragoran]

try

restorecon /usr/X11R6/lib/modules/drivers/nvidia_drv.so

as root
or

chcon system_u:object_r:lib_t /usr/X11R6/lib/modules/drivers/nvidia_drv.so

post the output of
ls -Z /usr/X11R6/lib/modules/drivers/

[cdhouch]

haven't had a chance to troubleshoot this properly, however I did disable SElinux and everything works like a charm now. This is my dev box anyway and is safely behind my firewalls.

So if your having similar problems and need a quick fix, disable SElinux, reinstall the driver, modprobe it, and it all seems fine... tho less secure ;D

[duvvid]

Well I worked out why I couldn't get it working with SELinux. You need to disable SELinux enforcement ('setenforce 0') *before* running the nvidia installer. If you don't, /usr/lib/tls/libnvidia-tls.so.1.0.7167 doesn't get installed. (one of the temp files executed by the installer gets an AVC denial)

As previously mentioned, the file contexts need to be fixed after installation.

# restorecon -v /usr/lib/{tls/,}libnvidia-tls.so.1.0.7167 \
  /usr/lib/libGL*so.1.0.7167 \
  /usr/X11R6/lib/libXvMCNVIDIA.so.1.0.7167 \
  /usr/X11R6/lib/modules/drivers/nvidia_drv.so \
  /usr/X11R6/lib/modules/extensions/libglx.so.1.0.7167

The policy might need a small alteration. /usr/X11R6/lib/libGL* (ie. the system OpenGL libs) are labelled texrel_shlib_t, but /usr/lib/libGL* (ie. the nvidia libs) are only labelled shlib_t. I've seen no denials yet, but it's something to keep in mind.