I’m attempting to use the latest Quadro driver on RHEL 7.6, kernel version 3.10.0-957.1.3.el7.x86_64. I tried versions 410.73, 410.93, and 415.25 on Quadro K2200, K620, and K600 cards. The error I’m seeing in the Xorg log is as follows:
[ 62026.912] (II) Initializing extension GLX
[ 62026.912] (II) AIGLX: Screen 0 is not DRI2 capable
[ 62026.912] (EE) AIGLX: reverting to software rendering
[ 62026.933] (II) IGLX: enabled GLX_MESA_copy_sub_buffer
[ 62026.934] (EE)
[ 62026.934] (EE) Backtrace:
[ 62026.934] (EE) 0: /usr/bin/X (xorg_backtrace+0x55) [0x55ce428f9185]
[ 62026.934] (EE) 1: /usr/bin/X (0x55ce42748000+0x1b4e09) [0x55ce428fce09]
[ 62026.934] (EE) 2: /lib64/libpthread.so.0 (0x7fdb24fae000+0xf5d0) [0x7fdb24fbd5d0]
[ 62026.934] (EE) 3: /usr/lib64/xorg/modules/extensions/libglx.so (0x7fdb222b2000+0x293a1) [0x7fdb222db3a1]
[ 62026.934] (EE) 4: /usr/lib64/xorg/modules/extensions/libglx.so (0x7fdb222b2000+0x287d2) [0x7fdb222da7d2]
[ 62026.934] (EE) 5: /usr/lib64/xorg/modules/extensions/libglx.so (0x7fdb222b2000+0x27bba) [0x7fdb222d9bba]
[ 62026.934] (EE) 6: /usr/bin/X (InitExtensions+0x5d) [0x55ce4281072d]
[ 62026.934] (EE) 7: /usr/bin/X (0x55ce42748000+0x601d6) [0x55ce427a81d6]
[ 62026.934] (EE) 8: /lib64/libc.so.6 (__libc_start_main+0xf5) [0x7fdb24c033d5]
[ 62026.935] (EE) 9: /usr/bin/X (0x55ce42748000+0x4a4ce) [0x55ce427924ce]
[ 62026.935] (EE)
[ 62026.935] (EE) Segmentation fault at address 0x7fdb20600948
[ 62026.935] (EE)
Fatal server error:
[ 62026.935] (EE) Caught signal 11 (Segmentation fault). Server aborting
I reverted back to version 390.87, which is working with no issues. With version 390.87 the DRI2 extension initializes without any problem.
Are there any known issues or bugs with latest NVIDIA drivers and RHEL 7.6? I need to determine if this is a problem with the NVIDIA driver or if this a Red Hat issue.
Thanks,
As documented in the “centos-7-geforce-gt740-unable-to-start-x-server” post:
The xorg-server version 1.20 with glx module version 1.19 was the issue was the problem. To fix the issue I did the following:
systemctl isolate runlevel3
./NVIDIA-Linux-x86_64-390.87.run --uninstall
yum -y reinstall rpm -qa | egrep -i "xorg|mesa"
./NVIDIA-Linux-x86_64-410.93.run
systemctl isolate runlevel5
Hi,
I’d like to add that these instructions helped me as well:
Fedora F29,
NVIDIA, quadro P5000
HP ZBook notebook.
thanks very much
&
best regards,
Hello Doug,
I noticed this tread was first stared in 01/2019.
But when I tried to install the latest Nvidia driver, ran into similar issues
Could you please help me to understand what you wrote in 01/10/2019?
The xorg-server version 1.20 with glx module version 1.19
What was the glx pkg/rpm you were looking for?
I was not able to locate a ‘glx’ with 1.19 or above?
These are the ‘glx’ pkgs that I have:
libglvnd-glx-1.0.1-0.8.git5baa1e5.el7.x86_64
glx-utils-8.3.0-10.el7.x86_64
libglvnd-glx-1.0.1-0.8.git5baa1e5.el7.i686
Appreciate your feedback
Hello Arfernando,
The glx module I’m loading is in the xorg-x11-server-Xorg package:
grep libglx.so /var/log/Xorg.0.log
[ 136.547] (II) Loading /usr/lib64/xorg/modules/extensions/libglx.so
rpm -qf /usr/lib64/xorg/modules/extensions/libglx.so
xorg-x11-server-Xorg-1.20.1-5.3.el7_6.x86_64
The original problem I was having appeared to be the result of one of the xorg or mesa packages installing an older version of /usr/lib64/xorg/modules/extensions/libglx.so. Removing and reinstalling all the xorg or mesa packages fixed the problem I has having.
Hello Arfernando,
The glx module I’m loading is in the xorg-x11-server-Xorg package:
grep libglx.so /var/log/Xorg.0.log
[ 136.547] (II) Loading /usr/lib64/xorg/modules/extensions/libglx.so
rpm -qf /usr/lib64/xorg/modules/extensions/libglx.so
xorg-x11-server-Xorg-1.20.1-5.3.el7_6.x86_64
The original problem I was having appeared to be the result of one of the xorg or mesa packages installing an older version of /usr/lib64/xorg/modules/extensions/libglx.so. Removing and reinstalling all the xorg or mesa packages fixed the problem I has having.
Hello Arfernando,
The glx module I’m loading is in the xorg-x11-server-Xorg package:
grep libglx.so /var/log/Xorg.0.log
[ 136.547] (II) Loading /usr/lib64/xorg/modules/extensions/libglx.so
rpm -qf /usr/lib64/xorg/modules/extensions/libglx.so
xorg-x11-server-Xorg-1.20.1-5.3.el7_6.x86_64
The original problem I was having appeared to be the result of one of the xorg or mesa packages installing an older version of /usr/lib64/xorg/modules/extensions/libglx.so. Removing and reinstalling all the xorg or mesa packages fixed the problem I was has having.
Hi Doug,
Thanks for the update.
(note: BTW you have three entries with the same content )
Yes, I have already verified the current “xorg-x11-server-Xorg-1.20.1-5.3.el7_6.x86_64” pkg
rpm -ql xorg-x11-server-Xorg-1.20.1-5.3.el7_6.x86_64 | grep glx
/usr/lib64/xorg/modules/extensions/libglx.so
Also noticed in the RHEL 7.6 workstation distro consists lf GLVND support:
rpm -ql libglvnd-glx-1.0.1-0.8.git5baa1e5.el7.x86_64
/usr/lib64/libGL.so.1
/usr/lib64/libGL.so.1.7.0
/usr/lib64/libGLX.so.0
/usr/lib64/libGLX.so.0.0.0
RPM for “libglx”, /usr/lib64/xorg/modules/extensions/libglx.so is the latest in RHEL7.6,
i.e. xorg-x11-server-Xorg-1.20.1-5.3.el7_6.x86_64 this is the latest pkg, is required to build the nvidia 430.14. I don’t find the older ver of ‘libglx.so’
Basically the fresh installation of RHEL 7.6 always installs 1.20.1.5.3 pkg (xorg-x11-server-Xorg-1.20.1-5.3.el7_6.x86_64). I don’t see a problem in the disro.
Also, Nvidia has stopped bundling libglx.so with the driver.
Could the problem is with new 430.14 driver doesn’t do a graceful uninstallation of the previous driver?
Or perhaps RH new pkg doesn’t overwrite the Nvidia libglx pkg?
Wish Nvidia provide further information.
Look forward to your comments.
With drivers <=390, the original glx module was backuped and replaced with the nvidia provided one.
With drivers >390, the driver layout was changed to prepare for glxvnd, not replcing the module anymore, instead providing a sub-module. This lead to problems with upgrading the xorg server depending on order:
1.have xserver 1.19 installed
2.install 390 ->glx 1.19 was backuped
3.upgrade to xserver 1.20
4.install driver 410-> calls uninstall 390->puts back glx 1.19
=crash
While this one would have worked without problems:
1.have xserver 1.19 installed
2.install 390 ->glx 1.19 was backuped
3.install driver 410-> calls uninstall 390->puts back glx 1.19
3.upgrade to xserver 1.20->replaces glx 1.19 with 1.20
=works
I have compared few drivers:
- then (early 2019) 418.56:
ll /usr/lib64/xorg/modules/extensions/glx
lrwxrwxrwx. 1 root root 29 Apr 29 12:45 /usr/lib64/xorg/modules/extensions/libglxserver_nvidia.so → libglxserver_nvidia.so.418.56
-rwxr-xr-x. 1 root root 13557744 Apr 29 12:45 /usr/lib64/xorg/modules/extensions/libglxserver_nvidia.so.418.56
-rwxr-xr-x. 1 root root 302336 Jun 26 2017 /usr/lib64/xorg/modules/extensions/libglx.so
- Then checked 430.14 driver:
ll /usr/lib64/xorg/modules/extensions/lib*
-rwxr-xr-x. 1 root root 297960 Feb 13 13:36 /usr/lib64/xorg/modules/extensions/libglx.so
lrwxrwxrwx. 1 root root 29 May 21 16:02 /usr/lib64/xorg/modules/extensions/libglxserver_nvidia.so → libglxserver_nvidia.so.430.14
-rwxr-xr-x. 1 root root 12597680 May 21 16:02 /usr/lib64/xorg/modules/extensions/libglxserver_nvidia.so.430.14
it seems Nvidia is still using their own lib:
/usr/lib64/xorg/modules/extensions/libglxserver_nvidia.so → libglxserver_nvidia.so.430.14
Then by extracting the drivers … I found what has provided by Nvidia:
older driver (mid 2018) 390.67:
ll -l NVIDIA-Linux-x86_64-390.67/libglx*
-rwxr-xr-x. 1 root root 14013096 Jun 1 2018 NVIDIA-Linux-x86_64-390.67/libglx.so.390.67
ls -ll NVIDIA-Linux-x86_64-418.56/glx
-rwxr-xr-x. 1 root root 13557744 Mar 15 13:36 NVIDIA-Linux-x86_64-418.56/libglxserver_nvidia.so.418.56
ll NVIDIA-Linux-x86_64-430.14/libglx*
-rwxr-xr-x. 1 root root 12597680 May 7 21:02 NVIDIA-Linux-x86_64-430.14/libglxserver_nvidia.so.430.14
So what do you think if NV is using libglxserver_nvidia.so why there is a dependency to libglx.so pkg?
Yes, I have tested fresh installation with nouveau and then installing 430.14 worked and also, uninstalling the previous driver works as well.
Fedora 29, nvidia driver 440.36 installed via rpmfusion. This machine was just upgraded from Fedora 27 (and the 387 driver) and I found that while turbovnc worked fine, and I could even run CUDA programs, X did not start and reboots just showed a black screen on console.
The above solution worked. Though on Fedora 29 the command line is a little different:
sudo dnf reinstall `rpm -qa | egrep -i “xorg|mesa”`
It took forever to track this problem down, thank you @generix !
And it only took my 8th nvidia account and four attempts to log in to mention that this works! I love technology. </sarcasm>