Bug 101587

Summary: [BISECTED][REGRESSION] Nouveau kernel module (>4.10) on NVA8 renders the system unusable
Product: xorg Reporter: Daniel Otero <otero.o.daniel>
Component: Driver/nouveauAssignee: Nouveau Project <nouveau>
Status: RESOLVED FIXED QA Contact: Xorg Project Team <xorg-team>
Severity: normal    
Priority: medium CC: fdsfgs
Version: unspecified   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
See Also: https://bugs.freedesktop.org/show_bug.cgi?id=101368
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
Kernel dmesg without bug
none
Kernel dmesg with bug none

Description Daniel Otero 2017-06-25 17:09:12 UTC 
Created attachment 132233 [details]
Kernel dmesg without bug

It happens every single boot. The system becomes unresponsive seconds after boot and the fan go to full speed until I force the system to shutdown.

Sometimes the screen stays completely black, other times it shows the console login prompt (no X involved), but the system it's barely usable.

Bisected the offending commit, and resulted this one:

> commit 1e2115d8c0c0da62405400316f5499d909e479bc (refs/bisect/bad)
> Author: Alexandre Courbot <acourbot@nvidia.com>
> Date:   Tue Dec 13 17:11:20 2016 +0900
>  
>     drm/nouveau/pmu: instanciate the falcon in PMU device

I attached the kernel logs with "nouveau.debug=debug" for both the previous and the failing commit.

System details:
  > Arch Linux
  > CPU Intel Core i5 480M with Intel Integrated Graphics
  > GPU Nvidia GeForce 310M (with Optimus technology)

Thank you for your time.
Comment 1 Daniel Otero 2017-06-25 17:10:04 UTC 
Created attachment 132234 [details]
Kernel dmesg with bug
Comment 2 Daniel Otero 2017-06-26 11:21:09 UTC 
I just find out what it seems to be a previous report of this bug 
(https://bugs.freedesktop.org/show_bug.cgi?id=101368).

But because it doesn't have a clear fix, I'm not sure is the same.
Comment 3 Ilia Mirkin 2017-06-26 13:24:21 UTC 
Yeah, possibly the same bug. But now with a bisect, so much more likely to get fixed.

The issue is [probably] that the newly called function makes calls to nvkm_falcon_rd32, which won't work at the ctor stage in optimus setups.

So that logic needs to be reworked. I believe that Ben Skeggs is on it (at least we discussed it last night).
Comment 4 Daniel Otero 2017-10-07 12:18:51 UTC 
The bug has been fixed some time ago and no longer occurs.

Thanks to all people involved.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.