AMD HD7770 crashes after a while

Faintful · 2014-06-15 17:14:45

Hello, I am on Arch 3.14.6 using the xf86-video-ati drivers. My monitor randomly goes to sleep and won't return until I rebooted the system. I am already booting with the radeon.dpm=0 kernel parameter, due to this it takes a bit longer to crash. This suddenly started happening and I have no idea why.. I hope someone has an idea on how to solve this!

My journal returns the following while my dmesg doesn't output any errors.

Jun 14 20:42:53 arch kernel: radeon 0000:01:00.0: GPU lockup CP stall for more than 4589433msec
Jun 14 20:42:53 arch kernel: radeon 0000:01:00.0: failed to get a new IB (-35)
Jun 14 20:42:53 arch kernel: [drm:radeon_cs_ib_fill] *ERROR* Failed to get ib !
Jun 14 20:42:53 arch kernel: radeon 0000:01:00.0: sa_manager is not empty, clearing anyway

Last edited by Faintful (2014-06-15 17:17:51)

svp · 2014-06-17 15:10:04

Hi! Same here - ArchLinux 3.14.6-1, GFX Card Radeon HD 7700 Series, Driver: "radeon"
Driver Modules: "drm"

radeon 0000:01:00.0: sa_manager is not empty, clearing anyway
radeon 0000:01:00.0: GPU lockup (waiting for 0x000000000005f288 last fence id 0x00000000000
radeon 0000:01:00.0: GPU lockup CP stall for more than 10023msec
radeon 0000:01:00.0: failed to get a new IB (-35)

Faintful · 2014-06-18 13:48:12

It hasn't made me reboot in the last days, the screen still completely shuts but after 10 - 20 seconds it works again.. I have the feeling it also happens less often now. The error message hasn't changed though and it's fairly annoying.

andre.ramaciotti · 2014-06-21 16:54:43

I'm having the same problem with a HD7870.

It seems to be worse since the update to 3.15.1. Now the monitor goes to sleep, then after 10-20s it comes back again, but everything freezes (including audio, it keeps playing the same buffer over and over again). The strangest thing is that it's not generating any log about it. There's nothing in /var/log/Xorg.0.old or journalctl.

Edit: this comment on the kernel bugzilla suggests appending radeon.audio=0 to the kernel parameters. As I don't need HDMI output, I'll try and see what happens.

Edit 2: It happened again, even with 'radeon.audio=0'. However, this time it came back again without a hard reboot.

Last edited by andre.ramaciotti (2014-06-21 22:31:55)

Faintful · 2014-06-25 12:32:26

It definitely seems to be worse with the 3.15.1 kernel! It requires a hard reboot now. I'm at work right now, I will see what happens when I append radeon.audio=0 when I'm back.

Edit: So okay, I added it, and it seems to only sometimes require a hard reboot now. The crash still occurs often though.

Last edited by Faintful (2014-06-26 14:22:49)

svp · 2014-06-27 06:39:47

Hi! Seems that 3.15 kernel and xf86-video-ati recent update fix the trouble
No more lockups

Faintful · 2014-06-27 13:29:31

They keep occuring for me. I have also created a bug report at the kernel bugzilla.

Edit:

A patch has been proposed but I have no idea how to apply it...
https://bugzilla.kernel.org/show_bug.cgi?id=79011

Last edited by Faintful (2014-06-28 07:48:23)

andre.ramaciotti · 2014-08-16 18:58:09

Sorry for the bump, but has anyone found a way to circumvent this?

Since 3.16 it has got worse for me. Now it freezes even when I'm not listening to anything, and the screen won't turn itself on again without a hard reset.

frumble · 2014-08-16 20:01:39

Hey, I found in this bug report that Mesa is to blame. 10.1.4 seems to work but I struggle with downgrading to it: https://bbs.archlinux.org/viewtopic.php … 2#p1447552

andre.ramaciotti · 2014-08-22 21:20:09

I've turned off hardware DPM and the problem seems to be gone. It's been a week since the last time a had to hard reset.

I don't know if it's related or if I just got lucky, but if anyone else wants to try, just put 'radeon.dpm=0' into your kernel command line.

frumble · 2014-08-23 00:38:05

Hey, found out that Linux 3.17 RC1 fixed it for my system which is running nearly crash free for three days now. The crashes were related to VLC and VDPAU and something got stuck but my system feels pretty stable now. For anyone interested: Compile linux-mainline from AUR and try for yourself.
(When you are at it, add "-march=native" to /etc/makepkg.conf for optimized compilation and you should tell it your real core number.)
Didn't wanted to turn off DPM because I guess it would be rather loud with my fan but Linux 3.17 seems to fix the issue. You can follow this official bug tracker.

andre.ramaciotti · 2014-08-23 01:02:14

That's what I thought, too, and that's why I've only tried turning DPM off recently. Actually, running without DPM means you won't get dynamic (by hardware) power management, but you still can get dynamic (by software) power management. There are a few drawbacks, but a loud fan is not one of them.

I've been running with radeon.dpm=0 and radeon-tray, and my GPU temperature hasn't gone over 30°C and it's as silent as ever.

However, it's good to know that it's been corrected. I'll try 3.17 when I get the chance.

frumble · 2014-08-27 15:37:29

OK, thank you! It's true, the fan is silent with DPM off. But it happened again, even with 3.17 RC2. The threads on bugs.freedesktop.org are still wondering about what makes this happen...

Arch Linux

#1 2014-06-15 17:14:45

AMD HD7770 crashes after a while

#2 2014-06-17 15:10:04

Re: AMD HD7770 crashes after a while

#3 2014-06-18 13:48:12

Re: AMD HD7770 crashes after a while

#4 2014-06-21 16:54:43

Re: AMD HD7770 crashes after a while

#5 2014-06-25 12:32:26

Re: AMD HD7770 crashes after a while

#6 2014-06-27 06:39:47

Re: AMD HD7770 crashes after a while

#7 2014-06-27 13:29:31

Re: AMD HD7770 crashes after a while

#8 2014-08-16 18:58:09

Re: AMD HD7770 crashes after a while

#9 2014-08-16 20:01:39

Re: AMD HD7770 crashes after a while

#10 2014-08-22 21:20:09

Re: AMD HD7770 crashes after a while

#11 2014-08-23 00:38:05

Re: AMD HD7770 crashes after a while

#12 2014-08-23 01:02:14

Re: AMD HD7770 crashes after a while

#13 2014-08-27 15:37:29

Re: AMD HD7770 crashes after a while

Board footer