Hi,
On Wed, 9 Mar 2022, Takashi Iwai wrote:
Takashi Iwai wrote:
The question is how often this problem hits. Basically it's a very corner case, and I even think we may leave as is; that's a matter of configuration, and lowering such a bar should expect some side-effect. OTOH, if the problem happens in many cases, it's beneficial to fix in the core part, indeed.
I'm basicly helping out the intel-gfx folks here. This is now happening systematically in the intel-gfx CI. The hung-task timeout is configured to 30sec (in intel-gfx CI), and there's some new hw configs where this happens every time (I have a separate patch in progress [1] that tries to detect this case and skip the init, but this will require more time as there is risk of breaking existing configurations).
[1] https://lists.freedesktop.org/archives/intel-gfx-trybot/2022-February/128278...
Tvrtko Ursulin wrote:
Takashi Iwai wrote:
Complete behaviour change how? Isn't this something ran on probe so likelihood of anyone sending SIGKILL to the modprobe process is only the init process? And in that case what is the fundamental difference
[...]
The point is that it does change the actual behavior, and changing the actual behavior just for working around the corner case like the above wouldn't be justified without the proper evaluation.
That said, if this behavior change is intentional and even desired, that's a way to go.
Let me try this out and test on a few configs (with and without the timeout occurring) and send a V3 if this seems ok.
Br, Kai