PC Drivers & Software

[dbb]

Turned on computer this morning.  After a long and unnecessary battle to get this 6700xt working with ubuntu a few months ago, and all working OK until yesterday,  this 6700xt is now only outputting to one monitor, listed as unclaimed by OS.

Some issue with amdgpu-dkms . Tried to reinstall, failed with following log.
`amdpgu-uninstall > reboot > amdgpu-install -y --usecase=graphics
` produces similar logs.



I 100% regret this purchase.`   Almost as much as I regret trying to use backticks for code blocking on the worst forum text entry box I've ever seen.

Still, any advice on fixing the GPU driver issue appreciated.  It would be nice to be able to use the device I've purchased.


Got things working by rolling kernel back with boot menu, could set grub to use old kernel by default:
https://unix.stackexchange.com/questions/432393/downgrade-linux-kernel-without-grub

Attempting amdgpu-install still fails with the same error though.

```````

DKMS make.log for amdgpu-5.11.32-1337797 for kernel 5.13.0-28-generic (x86_64)
Fri 4 Feb 11:06:09 GMT 2022
make: Entering directory '/usr/src/linux-headers-5.13.0-28-generic'
/var/lib/dkms/amdgpu/5.11.32-1337797/build/Makefile:26: "Local GCC version 90303 does not match kernel compiler GCC version 90300"
/var/lib/dkms/amdgpu/5.11.32-1337797/build/Makefile:27: "This may cause unexpected and hard-to-isolate compiler-related issues"
CC [M] /var/lib/dkms/amdgpu/5.11.32-1337797/build/scheduler/sched_main.o
CC [M] /var/lib/dkms/amdgpu/5.11.32-1337797/build/scheduler/sched_fence.o
gcc: fatal error: cannot specify ‘-o’ with ‘-c’, ‘-S’ or ‘-E’ with multiple files
compilation terminated.
make[2]: *** [scripts/Makefile.build:281: /var/lib/dkms/amdgpu/5.11.32-1337797/build/scheduler/sched_main.o] Error 1
make[2]: *** Waiting for unfinished jobs....
CC [M] /var/lib/dkms/amdgpu/5.11.32-1337797/build/scheduler/sched_entity.o
gcc: fatal error: cannot specify ‘-o’ with ‘-c’, ‘-S’ or ‘-E’ with multiple files
compilation terminated.
make[2]: *** [scripts/Makefile.build:281: /var/lib/dkms/amdgpu/5.11.32-1337797/build/scheduler/sched_fence.o] Error 1
gcc: fatal error: cannot specify ‘-o’ with ‘-c’, ‘-S’ or ‘-E’ with multiple files
compilation terminated.
make[2]: *** [scripts/Makefile.build:281: /var/lib/dkms/amdgpu/5.11.32-1337797/build/scheduler/sched_entity.o] Error 1
make[1]: *** [scripts/Makefile.build:524: /var/lib/dkms/amdgpu/5.11.32-1337797/build/scheduler] Error 2
make[1]: *** Waiting for unfinished jobs....
CC [M] /var/lib/dkms/amdgpu/5.11.32-1337797/build/ttm/ttm_tt.o
CC [M] /var/lib/dkms/amdgpu/5.11.32-1337797/build/ttm/ttm_bo.o
gcc: fatal error: cannot specify ‘-o’ with ‘-c’, ‘-S’ or ‘-E’ with multiple files
compilation terminated.
CC [M] /var/lib/dkms/amdgpu/5.11.32-1337797/build/amd/amdkcl/main.o
make[2]: *** [scripts/Makefile.build:281: /var/lib/dkms/amdgpu/5.11.32-1337797/build/ttm/ttm_tt.o] Error 1
make[2]: *** Waiting for unfinished jobs....
gcc: fatal error: cannot specify ‘-o’ with ‘-c’, ‘-S’ or ‘-E’ with multiple files
compilation terminated.
make[2]: *** [scripts/Makefile.build:281: /var/lib/dkms/amdgpu/5.11.32-1337797/build/ttm/ttm_bo.o] Error 1
gcc: fatal error: cannot specify ‘-o’ with ‘-c’, ‘-S’ or ‘-E’ with multiple files
compilation terminated.
CC [M] /var/lib/dkms/amdgpu/5.11.32-1337797/build/amd/amdkcl/symbols.o
make[1]: *** [scripts/Makefile.build:524: /var/lib/dkms/amdgpu/5.11.32-1337797/build/ttm] Error 2
make[2]: *** [scripts/Makefile.build:281: /var/lib/dkms/amdgpu/5.11.32-1337797/build/amd/amdkcl/main.o] Error 1
make[2]: *** Waiting for unfinished jobs....
gcc: fatal error: cannot specify ‘-o’ with ‘-c’, ‘-S’ or ‘-E’ with multiple files
compilation terminated.
make[2]: *** [scripts/Makefile.build:281: /var/lib/dkms/amdgpu/5.11.32-1337797/build/amd/amdkcl/symbols.o] Error 1
make[1]: *** [scripts/Makefile.build:524: /var/lib/dkms/amdgpu/5.11.32-1337797/build/amd/amdkcl] Error 2
CC [M] /var/lib/dkms/amdgpu/5.11.32-1337797/build/amd/amdgpu/amdgpu_drv.o
gcc: fatal error: cannot specify ‘-o’ with ‘-c’, ‘-S’ or ‘-E’ with multiple files
compilation terminated.
make[2]: *** [scripts/Makefile.build:281: /var/lib/dkms/amdgpu/5.11.32-1337797/build/amd/amdgpu/amdgpu_drv.o] Error 1
make[1]: *** [scripts/Makefile.build:524: /var/lib/dkms/amdgpu/5.11.32-1337797/build/amd/amdgpu] Error 2
make: *** [Makefile:1879: /var/lib/dkms/amdgpu/5.11.32-1337797/build] Error 2
make: Leaving directory '/usr/src/linux-headers-5.13.0-28-generic'````

[binarydepth]

Which kernel is working for you? I tried the previous version but still problem persists.

[geraci]

Same issue here...

This is just ridiculous....

Is this a common thing? Should I be scared to run apt-get upgrade because AMDGPU drivers will just kill over?

[MarcelYT]

Hey if you update now to version 5.11.32-1350682 you can enjoy the same issues on ALL your kernels since it updates the broken stuff to all the initramfs Seriously, very amateur work. I don't expect very fast driver updates on linux, but borking people's systems is very not cool. What is up with recompiling the kernel modules on every upgrade anyway? I've had multiple times where I couldn't even shut off my PC.

What worked for me as a temp workaround to even make my PC usable instead of reverting kernels was going to a term not on your DM, exec sudo modprobe amdgpu and restart your DM.

[dbb]

...
I thought I had traced the whole "not being able to shut the ubuntu machine down without it immediately restarting" back to a Steam/Proton/new gpu+older system  issue.

But now you point this out, it's not an issue on my nvidia system.

ffs.

At least AMD's CPUs are good value for money.  Assuming they aren't plagued with the same issues.

[AntoineHe]

I have the exact same problem here. PC won't boot  because of GPU driver dead again.

I have to select the old linux header 5.13.0-27-generic on the GRUB menu. 5.13.0-28 won't work, and i have the amdgpu-dkms you copy pasted here.

[niclasborlin]

Same story here. AMD Ryzen 7 3800X with Radeon RX 5600 OEM/5600 XT / 5700/5700 XT GPU. Single-monitor output with 5.13.0-28, triple-monitor output with 5.13.0-27.

[bratwurstbratgeraet]

Hello everyone,

just registered here in this board, because I had the same problems like you. I have an AMD Radeon RX 6700 with an Ubuntu 20.04.3 LTS (bought in the mid of 2021). From time to time my computer didn't boot successfully (after doing some "apt update" and "apt dist-upgrade"), then I knew: I have to boot in "rescue mode", without dedicated graphics drivers, download newest drivers from amd.com, compile and install them. Then reboot and everything will be "fine". Because the kernel version changed.

Not this time. Today I only got errors:

ERROR: Cannot create report: [Errno 17] File exists: '/var/crash/amdgpu-dkms.0.crash'
Error! Bad return status for module build on kernel: 5.13.0-28-generic (x86_64)
Consult /var/lib/dkms/amdgpu/5.11.32-1350682/build/make.log for more information.
dpkg: error processing package amdgpu-dkms (--configure):
installed amdgpu-dkms package post-installation script subprocess returned error exit status 10

Okay, the "make.log" file didn't help either. To me it seems like the current AMD driver package was not build for 5.13.x.y kernel versions, although Ubuntu already has them in their own repositories (and users install them).

I had to remove all my installed 5.13.x.y kernels and fall back to 5.11:

1. find all installed kernels: dpkg -l | grep linux-image
2. remove everything > 5.11, in my case: sudo apt-get remove --purge linux-image-5.13.0-27-generic linux-image-5.13.0-28-generic linux-image-unsigned-5.13.0-27-generic linux-image-unsigned-5.13.0-28-generic
3. empty modules folder (just to be sure), in my case: sudo rm -rf /lib/modules/5.13.0-27-generic /lib/modules/5.13.0-28-generic
4. reboot or directly load fresh compiled AMD graphics driver: sudo modprobe amdgpu

On my machine that worked and is stable "again".

So from my point of view it seems like on AMD side there is a problem to reliably deliver working (means: compilable) drivers for current kernel versions for Ubuntu. That makes life harder and it doesn't make fun, especially if customers do not have a solution for this problem.

If AMD doesn't improve this process in general, I will sell my AMD card and buy an NVIDIA.

[xtbugzman]

Dissapointed to have spent well over an hour troubleshooting this. Lack of response from AMD pretty much guarantees my next computer will be Intel + NVIDIA. This is ridiculous - test your drivers and get your **bleep** together.

The fix was:

sudo mv /boot/*5.13* ~/trash

sudo update-grub2

Then reboot and the installation should work on the older kernel, the one that probably has security vulnerabilities and is obviously untested by AMD!

[dbb]

As per @AntoineHe , rolling back to 5.13.0-27-generic from the grub menu worked for me.

If that doesn't work for you either try reinstalling with amdgpu-install, try a different kernel, or set your gpu on fire and post the remains back to your local amd office.  I'm back at my other computer, and my nvidia3070 is working fine, just like it always has, without all the screwing around.

[ginodeis-dev]

Same issue for kernel 5.13.0-28-generic, please can you help us fixing this?

[jfoxwu]

Ever since the windows 11 upgrade requirement, I have been slowly trying out Linux and evaluating switching over. I have decided to sit down and work on the few remaining problems I had with ubuntu over the weekend. Oh dang, what a bad timing to do this. I couldn't install the driver/opencl properly (RX 6600XT), and it was driving me insane. Glad I find this post.

I have been testing different distro as well and found out the driver worked just fine with Linux Mint, which uses kernel v.5.4.0-91 currently.

The online forum always said AMD card is better for Linux. Oh the irony.

[niclasborlin]

I bought an AMD CPU to avoid the Inte CPU vulnerabilities, picked an AMD GPU because, well AMD+AMD should cause less compatiiblity problems, right? With linux, apparently wrong.

As of now, I have retreated to kernel 5.13.0-27 that works. Did a quick test of 20.04d (https://packages.ubuntu.com/focal/linux-oem-20.04d) that uses kernel 5.14.0.something, to no avail.

I must confess I have started to evaluate NVIDIA options.

[niclasborlin]

For what its worth, I did a reinstall of 20.04 and did not touch anything related to graphics. Now I have three healty screens with kernel 5.13.0-30.

My take is that I broke something while fiddling with amdgpu-install.

I will stick to the vanilla graphics install from now on.

[MarcelYT]

Hey I'd like to let you all know that the following is an (almost) full fix with broken Steam:

- Install the latest amdgpu driver package (21.50) from the AMD website.
- Run amdgpu-install with your wanted drivers (Pro stuff is broken for me, anyone else?)
- apt update/upgrade, go through the motions and let it build your new kernels, reboot, repeat
- Everything now works normally except Steam, which does not start at all anymore, not even in Wine. I've filed an issue at https://github.com/ValveSoftware/steam-for-linux/issues/8430
- Enjoy running kernel 5.13.0-30-generic like a normal person. Kind of.

While I have your attention, does anyone know how to prevent the AMD drivers from updating and messing everything up, only triggering manually? Thanks in advance.

[binarydepth]

I did : sudo apt remove linux-generic-hwe-20.04

I intend to update the kernels manually too, the generic package pulls the new kernels during updates so by removing it you stop this behavior.