[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 #71 from Peter Ujfalusi <peter.ujfalusi at gmail.com> ---
Gleb,

(In reply to Gleb Nemshilov from comment #70)
> I wonder if Peter is still active on this bug? 2 months passed since I've
> attached the logs with latest test patches on 4.10-rc5 and one of the
> patches actually helped (the one that failed to compile first time).

I can try to forward port my patches if you can test them again on top on
current mainline.

I have rechecked the state and I still maintain that the poll_work is not
running at all after:
cae9ff036eea drm/nouveau: Don't enabling polling twice on runtime resume

The commit checks that
+       if (!drm_dev->mode_config.poll_enabled)
+               drm_kms_helper_poll_enable(drm_dev);

which is wrong in two ways.
First it should not peek into mode_config and especially it should not check
poll_enabled as it is the wrong boolean to check. poll_enabled indicates that
the poll_init has been called and nothing more.

The nouveau driver does this:
poll_init() followed by poll_disable() during the driver probe. poll_init()
will set the poll_enabled to true, poll_disable() will kill the poll_work. Now
since the mentioned patch is checking !drm_dev->mode_config.poll_enabled to
enable the poll again, it will never going to enable the work again.

This might be OK as HDMI/DP have hpd (hotplug detection) and the LCD is always
connected, but VGA might not have hpd and it's state change need to be polled.
Or not. I'm not sure about it.

This means that the poll_work can not cause anymore any issue on nouveau ;)

But, the patches I have attached to this bug are fixing the issue correctly
(khm, I believe) and keeps the poll_work functional.

Give me couple of days to forward port and test the patches.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20170416/517b13d2/attachment.html>