Bug 83826

Summary: [HSW regression] full-ppgtt hang
Product: DRI Reporter: Thilo-Alexander Ginkel <thilo>
Component: DRM/IntelAssignee: Intel GFX Bugs mailing list <intel-gfx-bugs>
Status: CLOSED FIXED QA Contact: Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity: normal    
Priority: highest CC: intel-gfx-bugs, przanoni
Version: XOrg git   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
/sys/class/drm/card0/error (bzipped) none

Description Thilo-Alexander Ginkel 2014-09-13 21:45:06 UTC 
System environment:
-- chipset: QM87
-- system architecture: x86_64
-- xf86-video-intel/xserver/mesa/libdrm version:
   - xf86-video-intel: 2.99.914 (git commit f36b7a4aac86f5874c32d8f516ac1f00a6c8d8b3)
   - xserver: 1.15.1
   - mesa: 10.1.3
   - libdrm: 2.4.52
-- kernel version: 3.17.0-rc4 (built from -nightly @ 99f444f [same happens with current nightly])
-- Linux distribution: Kubuntu 14.04
-- Machine or mobo model: Lenovo ThinkPad T440p 20AN
-- Display connector: DVI, DisplayPort
3) Reproduce steps:

When logging into KDE the login progress comes to a halt for ~ 15-30s and the following is logged to dmesg:

[   69.722297] [drm] no progress on render ring
[   69.723011] [drm] GPU HANG: ecode -1:0x00000000, reason: Ring hung, action: reset
[   69.723014] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
[   69.723015] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
[   69.723016] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
[   69.723017] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
[   69.723018] [drm] GPU crash dump saved to /sys/class/drm/card0/error

I tried bisecting the problem, but due to bug #83166 this does not work too well...

Earlier nightly builds did not suffer from this problem, AFAICS 52addcf is ok.
Comment 1 Thilo-Alexander Ginkel 2014-09-13 21:46:51 UTC 
Created attachment 106235 [details]
/sys/class/drm/card0/error (bzipped)
Comment 2 Chris Wilson 2014-09-14 06:56:50 UTC 
The most relevant recent commit is

commit fbc69204b0d5911e623eb700f1b6fd637dcaa538
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Sep 10 12:18:27 2014 +0100

    drm/i915: HSW always use GGTT selector for secure batches

I can't tell whether you have that one or not, so could quickly check?
Comment 3 Thilo-Alexander Ginkel 2014-09-14 20:29:50 UTC 
AFAICS 99f444f (dated Sep 9) does not include fbc6920 (dated Sep 10).
Comment 4 Paulo Zanoni 2014-10-08 19:00:20 UTC 
(In reply to Thilo-Alexander Ginkel from comment #3)
> AFAICS 99f444f (dated Sep 9) does not include fbc6920 (dated Sep 10).

Current drm-intel-nightly now contains this commit. Can you please retest?

Thanks,
Paulo
Comment 5 Daniel Vetter 2014-11-18 09:36:33 UTC 
(In reply to Paulo Zanoni from comment #4)
> (In reply to Thilo-Alexander Ginkel from comment #3)
> > AFAICS 99f444f (dated Sep 9) does not include fbc6920 (dated Sep 10).
> 
> Current drm-intel-nightly now contains this commit. Can you please retest?

Presuming fixed, please reopen if that's not the case.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.