Bug 100995

Summary:	System hang using OpenGL 4+ features
Product:	Mesa	Reporter:	Paul <Paul.Hancock.17041993>
Component:	Drivers/Gallium/radeonsi	Assignee:	Default DRI bug account <dri-devel>
Status:	RESOLVED FIXED	QA Contact:	Default DRI bug account <dri-devel>
Severity:	normal
Priority:	medium
Version:	17.0
Hardware:	x86-64 (AMD64)
OS:	Linux (All)
Whiteboard:
i915 platform:		i915 features:
Attachments:	errors that appear to be related to the hang glxinfo, should contain all the hardware and package info needed dmesg post-hang xorg log, may not be relevant

Description Paul 2017-05-11 02:52:45 UTC

This is easily reproducible via unigine's heaven benchmark, simply running it with normal tessellation settings results in the screen flashing black a few times and then an eventual system hang before anything even starts to render. A power cycle is the only way out.

Simply creating a 4.5 context will not trigger the bug, haven't been able to fully test what causes the hang specifically as accessible tools I've found are not compatible with mesa due to only creating a default context (which is currently 3.0), thus crashing on init.

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1689408

Comment 1 Michel Dänzer 2017-05-11 02:54:48 UTC

Please attach the output of glxinfo and dmesg and the Xorg log file.

Comment 2 Paul 2017-05-11 03:01:24 UTC

Created attachment 131303 [details]
errors that appear to be related to the hang

errors that appear to be related to the hang

Comment 3 Paul 2017-05-11 03:03:47 UTC

Created attachment 131304 [details]
glxinfo, should contain all the hardware and package info needed

glxinfo, should contain all the hardware and package info needed

Comment 4 Paul 2017-05-11 03:06:45 UTC

Created attachment 131305 [details]
dmesg

dmesg

(In reply to Michel Dänzer from comment #1)
> Please attach the output of glxinfo and dmesg and the Xorg log file.

Already in the process of dumping files...

Comment 5 Michel Dänzer 2017-05-11 03:09:17 UTC

Is this a regression from an older version of Mesa? If so, can you bisect?

Comment 6 Paul 2017-05-11 03:11:49 UTC

Created attachment 131306 [details]
post-hang xorg log, may not be relevant

post-hang xorg log, may not be relevant

Xorg doesn't appear to keep old log files, or the log may have been lost as a result of the hang, not exactly willing to trigger the hang again at this particular time...

Comment 7 Paul 2017-05-11 03:16:05 UTC

(In reply to Michel Dänzer from comment #5)
> Is this a regression from an older version of Mesa? If so, can you bisect?

Don't know, never had any working previous versions due to outdated kernel and drivers (kubuntu 16.10). This is as close as I've gotten > 3.0 to work on this machine by upgrading to 17.04 (4.10 kernel).

Comment 8 Timothy Arceri 2018-08-31 06:35:41 UTC

Is this working for you now?

Unigine heaven is used regularly for benchmarking and there are no other reports of this type as far as I'm aware.

Comment 9 Paul 2018-08-31 23:09:19 UTC

(In reply to Timothy Arceri from comment #8)
> Is this working for you now?

I've since removed linux from the system and handed it down to a family member, so it's not possible for me to test anymore. However I would expect the bug would have been fixed by now, according to the current RadeonFeature support table.

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.