[Bug 112159] Machine hang with i915.enable_psr=1 (Kaby Lake R)

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=112159

Jose Roberto de Souza <jose.souza at intel.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #16 from Jose Roberto de Souza <jose.souza at intel.com> ---
(In reply to Raphael Kubo da Costa from comment #15)
> Alright, I've tested your psr-test branch at commit 241ccbd67399c ("wip:
> drm/i915/psr: Increase psr idle paterns") and I'm unable to reproduce the
> machine hang there.
> 
> ... However, drm-tip _without_ your patches (i.e. 6eb50a2e5d102, "drm-tip:
> 2019y-11m-04d-18h-07m-25s UTC integration manifest", i.e. the tree you based
> your patches on) also works!
> 
> I've spent a few hours trying to figure out what happened between last
> week's drm-tip tree (5efc50549, "drm-tip: 2019y-10m-30d-22h-53m-25s UTC
> integration manifest") and yours. There's been rc6, and a lot of merges, but
> bisection (*) points to commit 58d124ea2739e1440ddd743d46c470fe724aca9a
> ("drm/i915: Complete crtc hw/uapi split, v6"), which was merged in commit
> 74dfc380537aa49ad0b741cd0358bdb9e0248e4b ("Merge remote-tracking branch
> 'drm-intel/drm-intel-next-queued' into drm-tip").
> 
> I can reliably reproduce my machine hang when using commit 2225f3c6f1d793
> ("drm/i915: Perform automated conversions for crtc uapi/hw split, base ->
> uapi"), which is 58d124ea2739e1440ddd743d46c470fe724aca9a's parent, and I
> have not managed to cause any hangs since commit
> 58d124ea2739e1440ddd743d46c470fe724aca9a.
> 
> I just have no idea if unrelated changes ended up fixing the worst cases of
> my actual issue and the machine no longer hangs, or if those commits were
> really supposed to fix this, or even whether the other failures in older
> dmesgs were actually other problems that are not related to the hang I was
> having.
> 
> (*) Due to the merges it wasn't a straight bisect; I reset to commit
> 4b40699236ebc009cba7c690e96c5873e2a75d6f ("Merge remote-tracking branch
> 'drm-misc/drm-misc-next' into drm-tip"), which was the latest drm-tip merge
> before the drm-intel-next-queued series, and then used git merge to merge a
> different range of the interval
> [f05816cbbcd0aa0af1efdd888ea6964644197e13..
> dbb1a6fbbb809fe364f51293026616c58916f330].

58d124ea2739e1440ddd743d46c470fe724aca9a wasn't supposed to fix anything but
maybe some register was left without programming since of the beginning of the
conversion and when DMC was restoring the registers the missing part was
causing the GTT fault.

Anyways, thanks for testing and doing the bisect.
I will close this issue then, please reopen if it regress.

-- 
You are receiving this mail because:
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/20191105/4cc47854/attachment.html>