Bug 72443 - [snb] blorp hang
Summary: [snb] blorp hang
Status: RESOLVED WORKSFORME
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i965 (show other bugs)
Version: git
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: Intel 3D Bugs Mailing List
QA Contact: Intel 3D Bugs Mailing List
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-12-07 17:11 UTC by Mike Russo
Modified: 2014-02-02 03:42 UTC (History)
2 users (show)

See Also:
i915 platform:
i915 features:


Attachments
/sys/class/drm/card0/error file after GPU hang (2.15 MB, text/plain)
2013-12-07 17:11 UTC, Mike Russo
Details
GPU error file from update on 1/2/14 (2.15 MB, text/plain)
2014-01-02 19:27 UTC, Mike Russo
Details

Description Mike Russo 2013-12-07 17:11:48 UTC
Created attachment 90418 [details]
/sys/class/drm/card0/error file after GPU hang

Although they seemed to go away for a while, I'm getting these again in Google Maps on Chrome. Easily reproducible by zooming in on a complex area like Central Park and panning around.  I've been running psensor lately to verify that my CPU is not overheating (max is 49C through this process).   Things have improved since the last time I've gotten this issue - the host has not hung despite having caused this message about 10 times. 

Note that I'm running kernel 3.12.3 with oibaf's PPA so I've got the latest Intel drivers and Mesa 10.1 development binaries (just updated this morning before posting this). 

Here are the relevant lines from /var/log/syslog and I'll also attach the /sys/class/drm/card0/error file. Let me know if anything else would be helpful.
Comment 1 Mike Russo 2013-12-07 17:12:57 UTC
relevant lines from syslog:

Dec  7 12:00:48 ossy kernel: [  237.595053] [drm] stuck on render ring
Dec  7 12:00:48 ossy kernel: [  237.595057] [drm] capturing error event; look for more information in /sys/class/drm/card0/error
Dec  7 12:00:48 ossy kernel: [  237.603728] [drm:i915_set_reset_status] *ERROR* render ring hung inside bo (0xf701000 ctx 1) at 0xf703148
Dec  7 12:00:54 ossy kernel: [  243.614015] [drm] stuck on render ring
Dec  7 12:00:54 ossy kernel: [  243.614071] [drm:i915_set_reset_status] *ERROR* render ring hung inside bo (0x2caa1000 ctx 1) at 0x2caa19f0
Dec  7 12:01:05 ossy kernel: [  254.568848] [drm] stuck on render ring
Dec  7 12:01:05 ossy kernel: [  254.568909] [drm:i915_set_reset_status] *ERROR* render ring hung inside bo (0x112dd000 ctx 1) at 0x112e0b40
Dec  7 12:01:08 ossy charon: 11[IKE] sending keep alive
Dec  7 12:01:08 ossy charon: 11[NET] sending packet: from 10.4.2.10[4500] to 98.113.84.130[4500]
Dec  7 12:01:11 ossy kernel: [  260.587808] [drm] stuck on render ring
Dec  7 12:01:11 ossy kernel: [  260.587919] [drm:i915_set_reset_status] *ERROR* render ring hung inside bo (0x112cd000 ctx 1) at 0x112cd1e8

when I did it again a little later:

Dec  7 12:11:05 ossy kernel: [  854.079002] [drm] stuck on render ring
Dec  7 12:11:05 ossy kernel: [  854.079057] [drm:i915_set_reset_status] *ERROR* render rin
g hung inside bo (0x142cb000 ctx 1) at 0x142cbde0
Comment 2 Chris Wilson 2013-12-09 09:58:56 UTC
Believed fixed in current stable release of Mesa.
Comment 3 Mike Russo 2013-12-09 13:00:50 UTC
(In reply to comment #2)
> Believed fixed in current stable release of Mesa.

Is this not fixed in Mesa 10.1? That's what I'm running.
Comment 4 Daniel Vetter 2013-12-09 17:31:38 UTC
That should actually work ...
Comment 5 Mike Russo 2014-01-02 19:26:48 UTC
So I decided to try this in 3.13rc6 and I do still get the error. I see that the verbiage of the error reported in syslog has been updated so I now see:

Jan  2 14:20:12 ossy kernel: [  128.655128] [drm] stuck on render ring
Jan  2 14:20:12 ossy kernel: [  128.655136] [drm] GPU crash dump saved to /sys/class/drm/card0/error
Jan  2 14:20:12 ossy kernel: [  128.655137] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Jan  2 14:20:12 ossy kernel: [  128.655138] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
Jan  2 14:20:12 ossy kernel: [  128.655138] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Jan  2 14:20:12 ossy kernel: [  128.655139] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
Jan  2 14:20:12 ossy kernel: [  128.658048] [drm:i915_set_reset_status] *ERROR* render ring hung inside bo (0x1260000 ctx 3) at 0x1261c20


Mesa was updated this morning and is: 
OpenGL vendor string: Intel Open Source Technology Center
OpenGL renderer string: Mesa DRI Intel(R) Sandybridge Desktop 
OpenGL core profile version string: 3.1 (Core Profile) Mesa 10.1.0-devel (git-5a51c1b saucy-oibaf-ppa)

Intel driver:
[    32.143] (II) intel(0): SNA compiled: xserver-xorg-video-intel 2:2.99.907+git1401021930.4d8f78~gd~s (Oibaf <fmrummey@gmail.com>)

I reproduced the same way as my original bug report.  I also tried using the "ilo" driver and Gallium acceleration, but they both hung with similar messages.  I haven't tried it recently but only "NoAccel" made it impossible to hit the desktop lockups.  

I will attach the new error file after this update.
Comment 6 Mike Russo 2014-01-02 19:27:49 UTC
Created attachment 91433 [details]
GPU error file from update on 1/2/14
Comment 7 Kenneth Graunke 2014-01-03 03:21:19 UTC
This doesn't appear to be a BLORP hang.  Unfortunately, your Mesa build appears to include Chad's supposed fix for this issue...
Comment 8 Ian Romanick 2014-01-22 00:37:26 UTC
Since ilo produces the same hang (see comment #5), the issue cannot be blorp... that driver doesn't have blorp!

Does it still occur with Google Maps?  I'd like to update the bug description to something correct. :)
Comment 9 Mike Russo 2014-01-22 01:21:52 UTC
It does - however while using Mesa 10.1 I'd also experienced some other issues (like Xorg using an increasing amount of memory until it inevitably hangs and crashes, though I can get control again and restart X using SysRq combinations) that led me to remove it and go back to the Ubuntu 13.10 base (with the distribution-provided kernel and Mesa).  Google Maps still hangs with similar error messages but so far at least Xorg isn't leaking (as much) memory.  If any progress has been made on this I'll try a new kernel and Mesa to see what happens. 

Adding to the strangeness factor is that this does NOT happen in Firefox (also when using the "new" more 3D-ish Google Maps) even though the functionality appears to be the same.
Comment 10 Mike Russo 2014-01-22 01:22:45 UTC
As for what to do with the bug, I guess you can close and if I experience it again after re-basing on Ubuntu 14.10 I will open another one.


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.