Steam for Linux

Steam for Linux

@R+5 Feb 17, 2019 @ 7:51am
How can I fix this error "OpenGL GLX extension not supported by display"?
When steam client is starting, the error happens after login and prevents steam to continue loading, and stops.

It started after an attempt to use a different nvidia driver, chosen in the "additional drivers" menu in "software and updates". the previous driver was working ok, but I forgot which method I used to install it (place or manually using a run file). the attempted upgrade was from driver 396.54 to 410.78 (this was already available in additional drivers, from previous attempt, which I also can't remember when I installed it, probably from ppa).

I tried to switch back to driver 396, but steam won't open, and the error remains.

I also tried to manually install a run file for nvidia driver 410.93. the install failed from be unable to find something related to headers of kernel.

I ended with a black screen. Then, I reinstalled 410.78 from place, and Im able to use again the desktop, but the problem with steam, among others, remain.

when I use "glxinfo" this happens:
name of display: 0.0
Error: couldn't find RGB GLX visual or fbconfig

if I use "grep -i Nvidia /var/log/Xorg.0.log" I get this
[23.251](ee) failed to load /usr/lib/x86_64-linux-gnu/sort/extra-modules/libglx.so:
libnvidia-tls.so.396.24.02: cannot open shared object file: no such file or directory

I tried also running apt --reinstall and --fix-missing but nothing changed.

the nvidia-settings panel won't open.

What should I do to repair, or properly install the new driver (410.78 or 410.93 ), or reinstall the old (396.54) to fix the problem? the driver currently selected is 410.78 (so if it's possible to fix it, without the run file, it would be great).

Thanks.



Last edited by @R+5; Feb 17, 2019 @ 8:04am
< >
Showing 1-15 of 86 comments
AndyMcDandyCDN Feb 17, 2019 @ 8:08am 
You don't mention which distribution you are using.

Every time I install a new Ubuntu based distro, I always add the Ubuntu graphics ppa to get the latest Nvidia drivers instead of trying to do it manually.
x_wing Feb 17, 2019 @ 8:10am 
Any chance you have mixed packages? Xorg complains about an object that it's definitely related with 396.54, but you said that you installed 410.93.

If you installed from a ppa (probably this one https://launchpad.net/~graphics-drivers/+archive/ubuntu/ppa) check that each built packages is installed on your system and there isn't any "forgotten" packages of the old version ("apt-cache search nvidia | grep 396" may show them all).

EDIT:

I think it is better to run: dpkg -l | grep nvidia | grep 396
Last edited by x_wing; Feb 17, 2019 @ 8:15am
Aoi Blue Feb 17, 2019 @ 8:59am 
You're nVidia drivers are improperly installed.

Purge them completely with configuration files then reinstall them.
ack0329 Feb 17, 2019 @ 9:13am 
Just FYI - Assuming you are using Ubuntu or a Debian Alternative (as it has the "Additional Driver" Option)
Never install NVidia from the .sh or run file - it's often / always problematic in Ubuntu ... resulting in black screen at startup.

TO get the NVidia driver reinstalled it is best / often necessary to uninstall and purge ANY existing driver ... and this is the easiest method in my experience:
sudo apt remove --purge nvidia*

sudo is super user, (and may ask for password), apt is Ubuntu's Installer/Uninstaller, and the * is for ANYTHING with this name.

Then go to the trusted "Additional Driver" tab and install the Driver of choice.
For Ubuntu LTR 390 is still the only shown option at this point as it is re-written to work in Ubuntu.

For a more recent Driver (often problematic in Ubuntu LTR in the past) ... it is now easily and problem freely installed in Ubuntu.

add this ppa first, restart and the cutting edge drivers will be available in Ubuntu's "Additional Driver Option" (415 as of this writing)
sudo add-apt-repository ppa:graphics-drivers

One may install it in Terminal also after the ppa is installed

$ sudo apt install nvidia-drivers-415

∘ or Auto-Update Driver Command
$ sudo ubuntu-drivers autoinstall

Always trying to help, Mark :steamhappy:
Last edited by ack0329; Feb 17, 2019 @ 9:14am
Aoi Blue Feb 17, 2019 @ 9:31am 
Originally posted by Rogue:
Originally posted by @R+5:
It started after an attempt to use a different nvidia driver, chosen in the "additional drivers" menu in "software and updates". the previous driver was working ok, but I forgot which method I used to install it (place or manually using a run file). the attempted upgrade was from driver 396.54 to 410.78 (this was already available in additional drivers, from previous attempt, which I also can't remember when I installed it, probably from ppa).

For future reference, don't install Nvidia drivers from the .run file. Install them from your distro's repos instead (if on an Ubuntu based distro, from the graphics drivers PPA).

I recommend installing from the CLI, but the additional drivers GUI should work too. Just remember to reboot after upgrading or changing drivers.

If you install from the nvidia.run file then you'll have to upgrade using that same method. Else you'll need to completely uninstall before installing the version from your distro's repos or PPA.
Yes, Last I checked, despite many requests from the community, nVidia's ".run" package STILL does not have a "build package" option.

Needless to say, this is yet another area where AMD is friendlier to the Linux community, as their proprietary drivers are available as such.
@R+5 Feb 17, 2019 @ 3:22pm 
sorry, I forgot to include which version of Ubuntu I'm using. it's 16.04.4.

after "purge", the last lines were
update-initramfs: generating boot initrd.img-4.16.18-041618-generic
w: possible missing firmware lib/firmware/i915/
skl_dmc_ver1_27.bin for module i915
kbl_dmc_ver1_04
kbl_guc_ver9_39
bxt_guc_ver9_29
skl_guc_ver9_33.bin for module i915

same lines were included after sudo install.
I've read they could be fixed with Intel driver update tool, but it's not available, so where I should start?
Last edited by @R+5; Feb 17, 2019 @ 3:56pm
x_wing Feb 17, 2019 @ 5:13pm 
Originally posted by @R+5:
sorry, I forgot to include which version of Ubuntu I'm using. it's 16.04.4.

after "purge", the last lines were
update-initramfs: generating boot initrd.img-4.16.18-041618-generic
w: possible missing firmware lib/firmware/i915/
skl_dmc_ver1_27.bin for module i915
kbl_dmc_ver1_04
kbl_guc_ver9_39
bxt_guc_ver9_29
skl_guc_ver9_33.bin for module i915

same lines were included after sudo install.
I've read they could be fixed with Intel driver update tool, but it's not available, so where I should start?

$ sudo apt-get install linux-firmware

Worth mention that the lack of this firmware is completely unrelated with the problem you have with your Nvidia GPU (but install them anyway).
@R+5 Feb 17, 2019 @ 7:18pm 
thanks. i tried that, but after running the command it says it's using the newest version of linux-firmware (1.157.21), and that is "set to manually installed".

also I'm unable to start steam, use Nvidia settings, and glxinfo still reports the same error.

I'm able to use the desktop, watch videos and browse.

before doing purge I used a tty and "service sddm stop". not sure why after installing back nvidia isn't working properly. what can i try?

update:
I used "lshw -c display" and I got this:
*-display UNCLAIMED
description: 3d controller
product: geforce GT 740m
etc

*-dispkay
description: VGA compatible controller
product: 3rd gen core processor graphics controller
Last edited by @R+5; Feb 17, 2019 @ 7:34pm
@R+5 Feb 18, 2019 @ 8:09am 
from bad to worse

I found that nvidia-prime wasn't reintalled and was deleted after purging Nvidia files. I also can't remember if I used it or bumblebee, so I tried it first. I reinstalled it, and after failing to enable Nvidia card from the indicator (only the icon changed), I used "sudo prime-select nvidia" and free trying glxinfo the information change a little:
name of display: :0.0
Xlib: extension "GTX" missing on display ":0.0".
* previous line is repeated 6 other times
Error: couldn't find RGB GLX visual or fbconfig
Xlib: extension "GTX" missing on display ":0.0".
Xlib: extension "GTX" missing on display ":0.0".
* this line is blank
* Xlib line is repeated agtain 8 times

I tried a new command I found under a different answer also about drivers:
sudo dkms status
your kernel headers for etc cannot be found
install or use --kernelsourcedir to tell dkms where is located

so, I checked which headers I had installed, and they weren't the right ones (they should have been "generic" but the generic part was absent in the name). I tried to install the right header file, but it wouldn't because it was lacking a dependency (the name was something like libssl 1.1 and only v 1.0 existed for my distro)

I decided to reinstall a previous kernel (4.15.0-45) using synaptic pm, because only the current one (lacking right headers) was the only one that I could use to reach desktop UI and use anything non-nvidia.

then I tried to reboot with the kernel that was working better, but at the begining my screen showed two times the name of the kernel before grub, and when autostarting the first kernel (the reinstalled one) I end in a black screen.

I'm now also unable to use "advanced option" in grub: when selecting it, a sequence of things related to initram happen, and ignore if or how could I use that.

I have a live usb: how can I use it to restore properly access?
it should be possible without resetting the whole system.

I used "sudo fdisk -l" and after the list with most of the drives, there was a red message: "partition 1 doesn't start on physical sector boundary"

then, I used boot-info to generate this
paste.ubuntu.com/p/bnpHZKFr52/

thanks for your patience.

Aoi Blue Feb 18, 2019 @ 11:08am 
You shouldn't need the live-usb. Just boot to recovery mode from the grub menu
Marlock Feb 18, 2019 @ 1:06pm 
You can use the liveusb to make a backup of your user home folder before delving any deeper (if it isn't done already, which it definitely should!)...

If you get tired from trying and need to start from scratch, just reinstall the OS and programs then copy the backed home folder over to the new install. It should restore most if not all your user configs for programs and the desktop as they were before.

If you still have time to spare, fixing things is a good way to learn your way around deeper parts of the OS though.
@R+5 Feb 18, 2019 @ 6:27pm 
sorry for the tldr

I'm unable to access recovery mode, as I commented earlier, because trying to open the menu in grub where the alternative kernels are listed will open a terminal screen that lists things related to initram, which I ignore how I could use to fix the problem.

how could I access and properly edit grub from a live usb, so I can include as first option a kernel with recovery mode?
or how could I use a terminal from grub, if I'm unable to start recovery mode?

my grub start screen isn't the most common: I have on top a kernel, followed by a low latency one, and then the submenu with other "backup" kernels I tried when updating the system. I didn't care to add a recovery mode option, since I didn't expected the submenu to stop working. after the submenu I still have the option to boot windows.

I really don't want to reinstall the os, since I took a long time to install many things I'm unsure how to back up, or if I would be able to. I tried a lot of programs from different repos (is there a way to find he list using live usb), because I like to discover what things can be done with "floss" software, and learn bits from experimenting. and yes, I could forget most of them, but I rather not since some tools that I rarely use can be useful, and "hunting them" would consume extra time.

my home folder is encrypted, and Im currently checking again how to use my login from a liveusb (since I'm unable to remember how), to recover theme and config data. there are other encrypted partitions I would need to check and figure out where to move some files: im unsure if I have enough space to move large photos and some videos, because my best option right now would be to make a new partition in unused space in the same disk to move them out. I don't want to start doing that, because I don't want to risk changing something that could make more complicated to recover the os. I would like to access the partitions, and then just check the files.

I think finding and fixing the kernel should be possible from the liveusb with a little guidance.
Aoi Blue Feb 19, 2019 @ 4:25am 
I often find it harder to recover theme and config data than just reconfigure everything.

It's also more fun to just do something new, and it allows me to refine what I had done before.

The only data you should be really concerned about is any essential data.

For disk encryption, many hard drives offer transparent whole disk encryption where you can wipe the whole hard drive by changing the key register on the drive. This is usually a better option in most environments.

@R+5 Feb 19, 2019 @ 6:36am 
i still think it would be far more useful to learn how to clean the bad update and replace the kernel, so if anything similar happened later for any new mistakes i made, i would know a bit better how to handle things on my own, and faster, and maybe make a guide for future reference.

im also searching ideas or answers elsewhere. i know it is possible to fix this, and should not be very complicated after finding the right process, from a few things that i read which explain a bit about similar problems ie askubuntuDOTcom/questions/28099/how-to-restore-a-system-after-accidentally-removing-all-kernels

But if anyone could offer me anything related to that, it could be very useful to fix this sooner. ths is how i learned a few things i more or less know about using linux as a "hobby" user.

Last edited by @R+5; Feb 19, 2019 @ 6:37am
Marlock Feb 20, 2019 @ 3:42am 
Originally posted by @R+5:
sorry for the tldr

I'm unable to access recovery mode, as I commented earlier, because trying to open the menu in grub where the alternative kernels are listed will open a terminal screen that lists things related to initram, which I ignore how I could use to fix the problem.

how could I access and properly edit grub from a live usb, so I can include as first option a kernel with recovery mode?

I think finding and fixing the kernel should be possible from the liveusb with a little guidance.

Boot to the liveboot
Mount the root partition of your damaged system (figure out how to mount it unencrypted for this to work...)
Open a terminal window
Run chroot /the/mounted/root
https://wiki.archlinux.org/index.php/Chroot
Work with the terminal window as if you were running the damaged system directly

Probably you should try rebuilding initramfs
https://www.systutorials.com/docs/linux/man/8-update-initramfs/

And then grub
< >
Showing 1-15 of 86 comments
Per page: 1530 50

Date Posted: Feb 17, 2019 @ 7:51am
Posts: 86