I got oibaf ppa. They started to use LLVM 3.6 compiler around that time I started to have unstable dota. I cant finnish any dota 2 games. The computer hangs before 10min mark while in dota. But I have played around 2hour episode of game of thrones from tell tale. So this sounds only dota 2 issue for me. I could see console a message something like this: "Still active bo inside vm" The computer freezes, most of the times I could only force boot. Gpu r9 290 3.19.0-031900rc7 OpenGL renderer string: Gallium 0.4 on AMD HAWAII OpenGL core profile version string: 3.3 (Core Profile) Mesa 10.6.0-devel (git-08a06b6 2015-02-10 trusty-oibaf-ppa)
I upgraded into 3.9 final and the crash happened under 1min in dota 2. 2 times in row. It's even worse now, but the message it gives is more clear now. It has an error that it couldnt resume dpm *something*. I am not sure, but it feels like there is error changing dpm modes.
I set the kernel parameter radeon.dpm =0 and the fps dropped from 60--> 20fps it still hangs the same way
(In reply to Jarkko K from comment #0) > I got oibaf ppa. They started to use LLVM 3.6 compiler around that time I > started to have unstable dota. Was the LLVM version the only thing which changed in between, or also the version of the kernel and/or Mesa? If the latter, can you try isolating which component caused the problem? Please attach the actual dmesg output corresponding to the problem.
Mesa have to be somehow the cause. I booted pc with lower kernel version and I got the exact hang. I would not keep a kernel version that hangs with dota in my kernel list. It doesnt matter what version I try to boot now. All hang. I also noticed that some of my DPM files were outdated. I downloaded few of them and placed into /lib/firmware/radeon after that the screen become distorted with grey colors (after dota hang). But what should I do with .bin files? How do I get the exact dmesg when pc hangs? I cant proof it but it feels like that power saving feature is trying to change the speed of the gpu or trying to resume hanged gpu like in windows some years back it had a feature that it tries to recover itself when it hangs. It felt exact the same. screen hangs and it's like nothing happens and after sometime it could resume rarely.
I am currently playing csgo with no issues. So this issue is only dota 2 related.
(In reply to Jarkko K from comment #4) > Mesa have to be somehow the cause. I booted pc with lower kernel version and > I got the exact hang. I would not keep a kernel version that hangs with dota > in my kernel list. It doesnt matter what version I try to boot now. All hang. Can you find a good snapshot of Mesa, and if so, can you bisect?
I have been playing more csgo with no issues. As far as I see this is only dota 2 related. I havent build myself mesa, so cant do bisect...It has been like a month like this.
This might not be related at all, but are you using "xf86-video-ati" for mode setting or the "modesetting" driver that comes with Xorg? I switched to the modesetting one because I saw that you could gain a couple fps when doing so. I got a few more fps (1-8fps) in the games I tried. But when I try to play Dota2 it hard locks after a while. The computer doesn't freeze but I have to manually kill dota2. When switching back to video-ati, I haven't gotten any of these hard locks yet. Perhaps they are unrelated, or the same thing that managed to hardlock Dota2 with the "modesetting" driver is doing the same thing to you.
How can I swap between those 2? I don't know what linux mint 17.1 uses as default.
(In reply to Jarkko K from comment #9) > How can I swap between those 2? > > I don't know what linux mint 17.1 uses as default. You can see it in your Xorg log. If it manages to load "radeon_drv.so" and/or "ati_drv.so", and them prints: (II) RADEON: Driver for ATI Radeon chipsets: ATI Radeon Mobility X600 (M24) 3150 (PCIE), ATI FireMV 2400 (PCI), With a long list of cards, then you are using the video-ati modesetting driver.
Now with more up to date kernel 3.19.4 and latest oipaf ppa git I got a hang and konsole said something like ring 0/3 stalled for something like for 2000-3000msec.
Since it started with LLVM 3.6 marking as duplicate of 89685 *** This bug has been marked as a duplicate of bug 89685 ***
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.