| View previous topic :: View next topic |
| Author |
Message |
logrusx
Advocate
Joined: 22 Feb 2018
Posts: 2473
|
 Posted: Tue Sep 17, 2024 5:24 pm Post subject: Broken resume on battery. Help to bisect it needed. |
 |
|
Hello,
Recently I found my laptop unable to resume on battery when suspended plugged in. As I keep my laptop mostly plugged in, I don't know when did that happen. What I know is it works with 6.1.90 and it doesn't with 6.1.106, 6.6, 6.10.
Using SysRq commands I sometimes can switch to a FB console and reboot, but I don't see anything.
What I need is help to pinpoint which commit broke it. I think it might be related to AMDGPU unable to resume, but I can't say for sure. I have hybrid graphics with NVIDIA and my CPU's integrated graphic card which I think is Vega. I tried disabling the NVIDIA chip with udev rule but it didn't make a difference.
How should I go about it?
Best Regards,
Georgi
Last edited by logrusx on Thu Sep 19, 2024 5:14 pm; edited 1 time in total |
|
| Back to top |
|
 |
sublogic
Apprentice
Joined: 21 Mar 2022
Posts: 279
Location: Pennsylvania, USA
|
| Posted: Wed Sep 18, 2024 2:44 am Post subject: |
|
|
https://wiki.gentoo.org/wiki/Kernel_git-bisect
This is tedious work because you can't automate the test in a script. Instead, you have to boot the test kernel, suspend and attempt to resume.
When I had a similar problem, I saved some time by- running git bisect and building the test kernels on a (faster) desktop computer;
- scp-ing the test kernels to the laptop;
- rebooting the test kernels to single-user mode;
Maybe you can find similar tricks. Good luck. |
|
| Back to top |
|
|
logrusx
Advocate
Joined: 22 Feb 2018
Posts: 2473
|
| Posted: Wed Sep 18, 2024 4:02 pm Post subject: |
|
|
Thank you! I had forgotten there was a wiki page on that.
| sublogic wrote: | | This is tedious work because you can't automate the test in a script. Instead, you have to boot the test kernel, suspend and attempt to resume. |
I will try to find which kernel versions are unaffected first and then I'll dig into history related to AMDGPU to narrow it further and try to verify my suspicion it's AMDGPU driver that causes it indeed. I think it is because on one of the unsuccessful resumes I actually managed to login to a fb console and blindly issue a restart of GDM, which flooded the log with Xorg entries. This means GDM was unable to start a Wayland session.
| sublogic wrote: | When I had a similar problem, I saved some time by- running git bisect and building the test kernels on a (faster) desktop computer;
- scp-ing the test kernels to the laptop;
- rebooting the test kernels to single-user mode;
Maybe you can find similar tricks. Good luck. |
Luckily I bough a gaming laptop so it can compile Gentoo fast. And it does :)
Best Regards,
Georgi |
|
| Back to top |
|
|
logrusx
Advocate
Joined: 22 Feb 2018
Posts: 2473
|
| Posted: Thu Sep 19, 2024 1:58 pm Post subject: |
|
|
So far I was able to identify it was indeed AMDGPU. It actually turned out to be fairly simple - just disable the iGPU. With nvidia it worked. Also I found a bunch of AMDGPU related suspend/resume bugs. That'll simplify the job a lot once I identify the minimal version of the kernel it doesn't work. I'll go oved AMDGPU history only.
Best Regards,
Georgi |
|
| Back to top |
|
|
logrusx
Advocate
Joined: 22 Feb 2018
Posts: 2473
|
|
| Back to top |
|
|
|
Kernel & Hardware
