[Bug 98690] [SKL bisected] System freeze when starting X using kernel 4.9-rc1 or later

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=98690

--- Comment #32 from Peter Ujfalusi <peter.ujfalusi at gmail.com> ---
(In reply to Gleb Nemshilov from comment #31)
> It seems that 4.8 is EOL now and 4.9 will be new LTS kernel. Can we expect
> that the fix (anything that will solve the issue) will land on 4.9.x? At
> least it could be reverted because for now kernel is unusable and reverting
> to old 4.4 is not a good idea.

The patch should be making it's way to 4.9 stable:
https://lkml.org/lkml/2017/1/9/372

> If there's something wrong with the logs, I can try to do some more stuff.

I'm still wondering what debug would be useful, but it is somewhere
nouveau_drm.c:nouveau_do_resume().

But out of curiosity if you could test the patch1.5 applied on top of patch1
and then all three, that would be great. The initial delayed work is disabled
several times according to the code, it might shed some lite to the events -
but certainly not going to get us closer to the race itself.

Since plain nouveau works w/o the delay it must be the Optimus/switcheroo calls
that fails. It might be that we have some work (delayed work?), tasklet run to
get the Optimus working and we have execution ordering issue? With delay the
theoretical work was running before the poll_work, which w/o the delay the
poll_work is executed first followed by the other work?

This is my best bet atm, but I need to study the code a bit more. Debugging DRM
just by reading the code is pita ;)

I'll attach the patch1.5

-- 
You are receiving this mail because:
You are the assignee for the bug.
You are the QA Contact for the bug.
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20170113/67204a64/attachment.html>