GeForce 7300LE with 304.132: glxinfo fails now with "X Error of failed request: BadValue"

GPU Unix Graphics Linux
#1

With the recently released legacy driver package 304.132, glxinfo reports the following error now:

$ glxinfo
name of display: :0
X Error of failed request: BadValue (integer parameter out of range for operation)
Major opcode of failed request: 156 (GLX)
Minor opcode of failed request: 24 (X_GLXCreateNewContext)
Value in failed request: 0x0
Serial number of failed request: 87
Current serial number in output stream: 88

nvidia-settings crashes when I try to open OpenGL settings, probably for the same reason.

Hardware: GeForce 7300LE
OS: ROSA Linux R8 x86_64
Kernel: I tried both 4.1.33 and 4.7.5 - the error is the same.
X11 server: 1.17.4.

ldd /usr/bin/glxinfo

    linux-vdso.so.1 =>  (0x00007ffea3630000)
    libGL.so.1 => /usr/lib64/nvidia304/libGL.so.1 (0x00007ff4ea45e000)
    libX11.so.6 => /usr/lib64/libX11.so.6 (0x00007ff4ea11e000)
    libc.so.6 => /lib64/libc.so.6 (0x00007ff4e9d59000)
    libnvidia-tls.so.304.132 => /usr/lib64/nvidia304/tls/libnvidia-tls.so.304.132 (0x00007ff4e9b56000)
    libnvidia-glcore.so.304.132 => /usr/lib64/nvidia304/libnvidia-glcore.so.304.132 (0x00007ff4e776c000)
    libXext.so.6 => /usr/lib64/libXext.so.6 (0x00007ff4e7559000)
    libdl.so.2 => /lib64/libdl.so.2 (0x00007ff4e7355000)
    libxcb.so.1 => /usr/lib64/libxcb.so.1 (0x00007ff4e7135000)
    /lib64/ld-linux-x86-64.so.2 (0x0000556bab309000)
    libm.so.6 => /lib64/libm.so.6 (0x00007ff4e6e2d000)
    libXau.so.6 => /usr/lib64/libXau.so.6 (0x00007ff4e6c29000)
    libXdmcp.so.6 => /usr/lib64/libXdmcp.so.6 (0x00007ff4e6a23000)

If I remove version 304.132 and install 304.131 back (kernel 4.1.33), glxinfo and other GL apps are working fine on that system. So it looks like a regression.

I will add the logs collected by nvidia-bugreport below.
nvidia-bug-report-304.132.log.gz (52.2 KB)
nvidia-bug-report-304.131.log.gz (52.2 KB)

#2

Added nvidia-bug-report logs for 304.132 where the problem happens and, for comparison, for 304.131, where everything runs OK.

#3

I’m the same trouble using fedora24.
$ lspci | grep VGA
01:00.0 VGA compatible controller: NVIDIA Corporation G71 [GeForce 7900 GS] (rev a1)

$ glxinfo
name of display: :0
X Error of failed request: BadValue (integer parameter out of range for operation)
Major opcode of failed request: 154 (GLX)
Minor opcode of failed request: 24 (X_GLXCreateNewContext)
Value in failed request: 0x0
Serial number of failed request: 87
Current serial number in output stream: 88

Isn’t not the way to return it to 304.131, but 304.132 reconsidered?

nvidia-bug-report.log.gz (51.1 KB)

#4

Well, for now, I am going to try the previous version, 304.131, with both mtrr-related patches and other patches to support kernel 4.6+. Let us see if it still works…

#5

304.131 is release in the past.Even if it’s tested with that, it’s insignificant.

Why doesn’t OpenGL function in 304.132? We should look for the cause.

When needing material besides nvidia-bug-report.log.gz, please tell me.

Regards!

#6

Well, I maintain the packages with Nvidia drivers in ROSA Linux and some of our users still need version 304.x working for different kernels. That is why I am going to experiment with 304.131 more.

That being said, I would also prefer to have 304.132 fixed, of course. However, it seems, the cause of the problem is somewhere in the binary-only part of the X11 driver, difficult to analyze without the sources.

#7

Hi,
When 304.132 isn’t corrected, we’re in trouble, too, but it’s a ubuntu(Unity) distribution that that would be a problem more.
I make a mistake in the corrected order of priority.
304.132 is raised by nvidia ubuntu ppa already.

Could y pls answer to the above point.

p.s.
Who is managing a 304.132 source?

#8

Hello,everyone.

$ lspci | grep VGA
01:00.0 VGA compatible controller: NVIDIA Corporation G71 [GeForce 7900 GS] (rev a1)

$ glxinfo
name of display: :0
X Error of failed request: BadValue (integer parameter out of range for operation)
Major opcode of failed request: 154 (GLX)
Minor opcode of failed request: 24 (X_GLXCreateNewContext)
Value in failed request: 0x0
Serial number of failed request: 87
Current serial number in output stream: 88

Please advice me how to diagnose this problem.
or I would like to ask for your assistance in suggesting any possible solution.
Any suggestion will be appreciated.

#9

fedora rpmfusion have unpushed the 304.132 version back into updates-testing.
We expect to be fixed 304.132.

#10

Any update on this issue? We are receiving many bug reports in Debian with this exact problem.

#11

Hello.I have not yet confirmed a correction of 304.132.

#12

I am running Ubuntu 16.04 and recently updated the nvidia driver to 304.132 and started experiencing this issue.

As a normal user, the output of glxinfo is identical to what is stated above:
http://paste.ubuntu.com/23428983/

However, when I execute as root, the output is successful:
http://paste.ubuntu.com/23428982/

Copy of my /var/log/Xorg.0.log (to show what drivers are loading):
http://paste.ubuntu.com/23428987/

Output of getfacl /dev/nvid*:
http://paste.ubuntu.com/23429026/

Is there a permission issue going on somewhere, a conflict with another driver module or something else preventing nvidia driver from creating a new opengl context as a normal user? (i also created a new non-root user and i get the same issue)

Hope these details help. Thanks in advance

#13

Hello,Thanks a lot for the information!
I tested by the root user, too in fedora environment. OpenGL function worked as design.oh my god!!
I think there is a problem about a right of your comment street normal user permission.And a driver’s permission isn’t answering to a right.
nice information!.
Best Regards!

#14

Yea, 304.132 has issues, 304.131 doesn’t. I’ve confirmed this by downgrading to 304.131 on ubuntu 16.04 using the following command and everyhing is fine now:

sudo apt install nvidia-304=304.131-0ubuntu3

I’m going to stay on 304.131 since it works perfectly. The upgrade to 304.132 broke everything. I mean, the bug prevented me from being able to log into unity and prevented plasmashell from loading any menus in kde. on top of that, the crash reporting utility was unable to record a backtrace and thus was not successfully reported through the standard bug tracking system. I bet millions of people are affected by this on ubuntu distros and the devs are unaware because of this fluke in the reporting system.

There is some discussion related to this going on here related to ubuntu 14.04 but applies to 16.04 as well:
http://unix.stackexchange.com/questions/321185/after-update-glx-works-only-for-root-nvidia

The bug is so bad that I have put all further apt updates to the nvidia-304 package on hold by running:

echo “nvidia-304 hold” | dpkg —set-selextions

This way I don’t accidentally upgrade to the version that renders my system pretty much useless.

Hooe this helps any ubuntu user experiencing this issue.

#15

Sorry typo, the command is:

echo “nvidia-304 hold” | dpkg —set-selections

#16

We reproduced the problem and it’s being tracked in bug number 1835736.

#17

And how do we get updates on this? Is there a link or url to the tracking system?

#18

me too!

#19

Aaron Plattner, if you get this. Please send us a link to the case. Thanks

#20

I am highly interested in this as well, so I wanted to give this thread a bump.

The 304.132 driver seems to exhibit different problems depending on the distro. In Ubunutu, it becomes impossible to login. In Linux Mint, Cinnamon desktop crashes.

Unfortunately, downgrading to nvidia-304=304.131-0ubuntu3 does not work for me; I get corrupted graphics upon reaching the desktop in Ubuntu 16.04. Functionality is restored only by purging nvidia* and using the open source driver, but performance is poor as one might assume.

My laptop has a GeForce Go 7300.