Page 1 of 2
Linux 3.5.7 vs Radeon HD 6570
Posted: Sun Apr 07, 2013 11:29 pm
by marcus
When running certain GL applications (such as the Android system emulator), I get system lockups related to the RadeonHD drm driver:
Code: Select all
[1821887.366661] radeon 0000:01:00.0: couldn't schedule ib
[1821887.366667] [drm:radeon_cs_ib_chunk] *ERROR* Failed to schedule IB !
[1821904.616133] radeon 0000:01:00.0: couldn't schedule ib
[1821904.616139] [drm:radeon_cs_ib_chunk] *ERROR* Failed to schedule IB !
When this happens, basically all X output is blocked (at most the mouse pointer can move), and I have to reboot the system.
Upgrading mesa from 9.0.3 to 9.1.1 made no difference at all, but this looks like a problem in the kernel driver anyway.
So, since the 3.5 kernel series is EOL:ed (so is 3.6 and 3.7), is someone working on a 3.8 kernel, which might fix this problem? It would be nice to be able to run any GL program without having to fear for a mandatory reboot...
Re: Linux 3.5.7 vs Radeon HD 6570
Posted: Thu Apr 11, 2013 3:21 pm
by xeno74
I have the same problem
Re: Linux 3.5.7 vs Radeon HD 6570
Posted: Thu Apr 11, 2013 9:54 pm
by zappa2009
I think this Problem is the KMS and DRM driver for the RadeonHD cards.
Or missing Kernel-Modules ? I dont know.
I have the same Problem with Supertuxkart on some Maps/Overworld.
Re: Linux 3.5.7 vs Radeon HD 6570
Posted: Sat Apr 13, 2013 12:54 pm
by xeno74
zappa2009 wrote:I think this Problem is the KMS and DRM driver for the RadeonHD cards.
Or missing Kernel-Modules ? I dont know.
I have the same Problem with Supertuxkart on some Maps/Overworld.
What do you think about to build a new kernel like 3.8?
Re: Linux 3.5.7 vs Radeon HD 6570
Posted: Sat Apr 13, 2013 2:34 pm
by marcus
What do you think about to build a new kernel like 3.8?
That's my plan. Hence the question in the OP if somebody is already working on it, to avoid duplicate efforts...
Re: Linux 3.5.7 vs Radeon HD 6570
Posted: Sat Apr 13, 2013 3:54 pm
by xeno74
marcus wrote:What do you think about to build a new kernel like 3.8?
That's my plan. Hence the question in the OP if somebody is already working on it, to avoid duplicate efforts...
That's very kind of you
Re: Linux 3.5.7 vs Radeon HD 6570
Posted: Sat Apr 13, 2013 4:22 pm
by marcus
Well, my kindness only allowed me to wait for so long before going ahead regardless...
I've now made a 3.8.7 build. Darren's 3.5.7 patch could be applied with only one small modification.
So far, it's looking promising. I've seen no kernel lockups yet. At one point (where I was expecting a kernel lockup), the mouse pointer froze for a few seconds, but after that things were able to proceed. Then the user process crashed, but that's still preferable to a kernel lockup, and probably not related anyway.
If someone else want's to give it a try and see if it fixes their lockups too, I've put the updated kernel patch up as
http://mc.pp.se/nemo_3.8.7.patch.
Re: Linux 3.5.7 vs Radeon HD 6570
Posted: Sat Apr 13, 2013 6:47 pm
by xeno74
marcus wrote:Well, my kindness only allowed me to wait for so long before going ahead regardless...
I've now made a 3.8.7 build. Darren's 3.5.7 patch could be applied with only one small modification.
So far, it's looking promising. I've seen no kernel lockups yet. At one point (where I was expecting a kernel lockup), the mouse pointer froze for a few seconds, but after that things were able to proceed. Then the user process crashed, but that's still preferable to a kernel lockup, and probably not related anyway.
If someone else want's to give it a try and see if it fixes their lockups too, I've put the updated kernel patch up as
http://mc.pp.se/nemo_3.8.7.patch.
Thank you! Could you post your kernel config, please?
Re: Linux 3.5.7 vs Radeon HD 6570
Posted: Sat Apr 13, 2013 6:57 pm
by xeno74
OK, I have found it
Re: Linux 3.5.7 vs Radeon HD 6570
Posted: Sat Apr 13, 2013 10:47 pm
by daz
marcus wrote:Well, my kindness only allowed me to wait for so long before going ahead regardless...
I've now made a 3.8.7 build. Darren's 3.5.7 patch could be applied with only one small modification.
Hmm, hand't thought of trying my patch directly against a newer kernel version, they have a tendency to change whole functions...
The modification you refer to will be the pci device secondary bus i assume (caused me a bit of trouble)
I have an experimental 3.8.6 running here, but is still under test.I'm trying (without much success) to add UVD.
Perhaps you can point me towards what doesn't work for you as I haven't seen problems, and am not sure it isn't a problem with the programs themselves, not the kernel driver (as many other things work fine).
So far, it's looking promising. I've seen no kernel lockups yet. At one point (where I was expecting a kernel lockup), the mouse pointer froze for a few seconds, but after that things were able to proceed. Then the user process crashed, but that's still preferable to a kernel lockup, and probably not related anyway.
If someone else want's to give it a try and see if it fixes their lockups too, I've put the updated kernel patch up as
http://mc.pp.se/nemo_3.8.7.patch.
3.8.7..hmm kernels been updated again I see (more work)
If I workout UVD I'll see about getting a patch for that here too.
Regards
Darren