Bug 73785

Summary:	[HyperZ] Team Fortress 2 causes random GPU stalls on radeonsi
Product:	Mesa	Reporter:	Itzamna <xamaniqinqu>
Component:	Drivers/Gallium/radeonsi	Assignee:	Default DRI bug account <dri-devel>
Status:	RESOLVED FIXED	QA Contact:
Severity:	normal
Priority:	medium
Version:	git
Hardware:	x86-64 (AMD64)
OS:	Linux (All)
Whiteboard:
i915 platform:		i915 features:
Bug Depends on:
Bug Blocks:	75112
Attachments:	dmesg output related to the GPU stall.

Description Itzamna 2014-01-19 05:52:48 UTC

Created attachment 92376 [details]
dmesg output related to the GPU stall.

Overview:
    The game "Team Fortress 2" causes the GPU to stall randomly. The stalls sometimes resolve by themselves after ~1 min, but will occur again rapidly.

Steps to reproduce:
    1) Install "Team Fortress 2" through Steam ( http://store.steampowered.com/app/440/ ).
    2) Start the game, join a map.
    3) Play until the GPU hangs.
    4) Optionally: wait until GPU recovers.

Actual results:
    Frequent GPU stalls while playing Team Fortress 2. Sometimes video output is restored after ~1 min.

Expected results:
    No stalls.

Build date and platform:
    Build date of all components: 01/18/2014
    Linux kernel version: 3.13-rc8 x86_64
    Mesa: git (566e0ddfd01dbadd75462fed5a3f141b9f494028)
    LLVM: 3.5-svn (commit 199570)

Additional information:
    I am running on a Radeon HD 7970 with DVI output at 2560x1440. Dynamic power management is enabled. Disabling Hi-Z does not alleviate the problem. The output of "dmesg 'drm\|radeon'" has been added as an attachment. I could not reproduce the issue (yet) in another Source engine title, "Left 4 Dead 2".

Comment 1 Alex Deucher 2014-01-19 20:13:14 UTC

Is this a regression?  If so when was the last time it was working?  Does disabling dpm help?  Boot with radeon.dpm=0 on the kernel command line in grub.

Comment 2 Itzamna 2014-01-22 14:29:40 UTC

Hello Alex,

Thanks for replying to my bug report.

I am not certain exactly when this bug was introduced or even which component is involved, but since TF2 was stable a month ago it must be a regression.

I can report that having played well over 2 h with dynamic power management disabled, that this bug is related to dpm.

However, playing with radeon.dpm=1 and setting /sys/class/drm/card0/device/power_dpm_state to 'balanced' (down from 'performance'), the stalls become much less frequent, more in the range of once or twice every 2-3 hours.

Comment 3 Alex Deucher 2014-01-22 14:37:08 UTC

(In reply to comment #2)
> Hello Alex,
> 
> Thanks for replying to my bug report.
> 
> I am not certain exactly when this bug was introduced or even which
> component is involved, but since TF2 was stable a month ago it must be a
> regression.
> 
> I can report that having played well over 2 h with dynamic power management
> disabled, that this bug is related to dpm.
> 
> However, playing with radeon.dpm=1 and setting
> /sys/class/drm/card0/device/power_dpm_state to 'balanced' (down from
> 'performance'), the stalls become much less frequent, more in the range of
> once or twice every 2-3 hours.

I think it's coincidence.  balanced and performance are the same on your chip.

Comment 4 Lukas Kahnert 2014-02-09 14:18:56 UTC

I have the same Problem but it doesn't matter if DPM is enabled or not.

I can reproduce this Bug when i play on the map "Ghost Town"(Wave 666).
Always when the Tanks are comming and I going to them the GPU hangs.
On other maps its more or less random.

Every component was build yesterday
   Linux 3.14-rc1(drm-fixes branch)
   LLVM 3.5-svn
   Mesa 10.2-devel git
   xorg 1.15.0
   glamor 0.6


PS: Sorry for my bad English >_<

Comment 5 Alex Deucher 2014-02-10 22:30:10 UTC

Does disabling hyperz help?

Comment 6 Lukas Kahnert 2014-02-14 17:20:29 UTC

Yeah it helps

With R600_DEBUG=nohyperz there were no hangs in TF2 or Rust(both had the same Problem)

I saw HyperZ in now disabled by default so R600_DEBUG is not longer needed.

Comment 7 Marek Olšák 2014-09-01 21:26:34 UTC

Is this fixed with current Mesa git?

Comment 8 Marek Olšák 2014-09-03 10:33:10 UTC

I don't see any hang here.

Comment 9 bahamut2k456 2014-09-13 20:08:08 UTC

Currently running revision git-d13d2fd, hyperz enabled, on a 280x. I'm seeing no issues.

Comment 10 Marek Olšák 2015-02-24 22:51:58 UTC

(In reply to bahamut2k456 from comment #9)
> Currently running revision git-d13d2fd, hyperz enabled, on a 280x. I'm
> seeing no issues.

Thanks. Closing.

Comment 11 Matías Locatti 2016-05-03 03:15:24 UTC

I'm having this issue with an HD7750, running mesa-git didn't fix it.

Comment 12 Nicolai Hähnle 2016-05-05 15:51:13 UTC

Hi Matías, to clarify: Does disabling HyperZ help for you as well?

Comment 13 Matías Locatti 2016-05-09 05:37:17 UTC

R600_DEBUG=nohyperz right?

Comment 14 Matías Locatti 2016-05-09 05:48:54 UTC

Nope, still happens.

Comment 15 Marek Olšák 2016-05-24 13:52:14 UTC

(In reply to Matías Locatti from comment #14)
> Nope, still happens.

which means it's not a HyperZ issue. It must be something else. Closing as this bug report is against HyperZ.

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.