i915: occasional hangs on Baytrail with deeper C-states

[Daniel Bilik]

On Wed, 6 Apr 2016 09:22:19 +0200
Francois Tigeot <ftigeot at wolfpond.org> wrote:

>> And now good news. Mika Kuoppala has come with tentative patches...
>> ... that solve stability issues on my system. Diff against recent
> Many thanks for this patch, I have pushed it to master.

Unfortunately, my "success story" was reported too soon, as I've
experienced another couple of hangs in last three weeks. Mika's
"tentative" patches do influence the behaviour notably, but don't prevent
from hangs completely. It probably just reduces "the hang condition
window" enough, so that it takes much much longer to hit it. Or maybe
there is another source of hangs (see below).

Nevertheless, disscussion on bugzilla.kernel.org continues, and there is
one very interesting new comment...

https://bugzilla.kernel.org/show_bug.cgi?id=109051#c284

... that first presents very comprehensive summary of the problem, and
second points to this comment from Chris Wilson on "intel-gfx" list at
freedesktop.org:

"Why those vlv_punit_read() result in a machine hang was never understood."

This explains why it still hasn't been "properly" fixed. Suggested "&& 0"
patch seems more like a quick hack than a real solution, however I've
patched the kernel and I'm currently testing it. It sort-of reverts one
change introduced by this commit...

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=31685c258e0b0ad6aa486c5ec001382cf8a64212

... that is mentioned on bugzilla.kernel.org in comment #c287 as another
potential source of hangs.

Stay tuned, I'll report when my "test" kernel hangs. ;-)

--
						Dan