[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 #8 from Peter Ujfalusi <peter.ujfalusi at gmail.com> ---
I don't think that the root cause is 339fd36238dd3494bc4617d181e7a37922c29ee9.
Yes, it triggers something for sure which looks to me a race or synchronization
issue in the code paths which happened to be hidden by the 10sec delay.

W/o the patch we had basically two scenarios:
1. When drm_kms_helper_poll_init() was called before
drm_helper_probe_single_connector_modes():

If there is a change in the connector state the output_poll_work would be
scheduled with 0 delay.

2. When drm_helper_probe_single_connector_modes() was called before
drm_kms_helper_poll_init()

If there is a change in the connector state, the code sets delayed_event and
will _not_ schedule right away the output_poll_work from
drm_helper_probe_single_connector_modes() as the work is not yet initialized.
When the drm_kms_helper_poll_init() is called, the output_poll_work will be
scheduled to run with 10sec delay.

In case 1 we run the work w/o delay while in case 2 we run it with 10sec delay.

The patch in question will run the work in both cases w/o delay when it is
possible to run it. iow when we have the work initialized and we have a
connector change noticed, we schedule the work to be executed w/o delay. This
connector change is basically the change from unknown to some known state
during the stack bring-up, if I'm not mistaken.

Probably shoprted delay in drm_kms_helper_poll_enable_locked() for the work
might hide the issue. It might surface later and the delay needs to be
increased.

If the decision is to revert the patch, I'm fine with it, but then no one will
ever going to take time and effort to figure out the why.

-- 
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/20161220/a0b3134c/attachment.html>