[Bug 79311] Call trace originating from i915 referencing intel_wait_for_vblank()

[bugzilla-daemon at bugzilla.kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=79311

--- Comment #36 from Alan Bartlett <ajb at elrepo.org> ---
(In reply to Ville Syrjala from comment #35)
> (In reply to Alan Bartlett from comment #34)
> > (1) Do you want a fresh bug report to be opened for the new "vblank wait
> > timed out" originating from "drivers/gpu/drm/i915/intel_display.c:902
> > g4x_wait_for_vblank+0xd5/0xf0 [i915]()" ?
> 
> Are you saying you're not getting this other vblank wait timeout with an
> unpatched kernel?

No. There has been evidence of the other vblank wait timeout
("g4x_wait_for_vblank") with vanilla (unpatched) kernels. Examples can be found
in some of the traces, earlier in this report (shown above).

I have performed a whole series of tests with the 3.17-rc4 kernel which
includes the patch. The results are inconsistent. At first I thought it was the
difference between a warm boot and a cold boot but further testing shows it not
to be the case.

Sometimes there can be a completely successful boot and no traces are produced,
yet on other times either one or two call traces are produced. (For
completeness, I will mention that sometimes there is a mention of a "*ERROR*
pipe A underrun", yet on other occasions there is not.)

The only tentative suggestion I can make to account for these varying results
is that it is a timing effect.

There is only one definite conclusion I can currently make and that is the i915
code is broken for *all* kernels > 3.15.10

I am still prepared to assist with testing -- as long as it does not consume
all of my time to the detriment of other tasks.

-- 
You are receiving this mail because:
You are on the CC list for the bug.