Radeon KMS bug causing wireless crashes, devs please help

[Porter]

Attn: radeon devs, if you're here... thanks for all your hard work.

There's an ongoing bug report on Launchpad, initially directed toward iwl3945, but it appears that the problem is actually with the radeon driver... wireless hangs/crashes during high GPU utilization (like video playback), but only occurs when kernel modesetting is turned on. With KMS off, the hardware runs normally. This seems to be some sort of regression, either in the kernel or in the driver.

Here's the bug report: https://bugs.launchpad.net/ubuntu/+s...ux/+bug/564376

Please help... there are lots of folks out there with systems that are unusable unless they manually disable kernel modesetting. I'm one of the affected users, in my case on a Thinkpad T60 with X1400 and Intel wireless. There are also users reporting audio corruption under the same conditions as mentioned above, which disappears when KMS is disabled.

Thanks for any help you can provide!


Best regards,
Jason Porter

[Porter]

I forgot to mention (and the edit button is missing)... this is a new problem on Ubuntu 10.04, the bug was present on launch of Lucid final.

Thanks!

[oyvind]

Yes, it would be great if this was looked into. I've got affected hardware as well. Problems with audio (crackling not caused by Pulseaudio) and wireless connection dropping when running with radeon KMS, typically triggered by fullscreen movie playback or in general “GPU load” (for lack of better way to describe it). Both problems disappear when disabling kernel mode setting. Thinkpad Z61m with ATI X1400.

These are the bugs I am currently tracking wrt. this issue:
https://bugs.launchpad.net/ubuntu/+s...ux/+bug/564376
https://bugs.launchpad.net/ubuntu/+s...er/+bug/578342
https://bugzilla.kernel.org/show_bug.cgi?id=15912
https://bugs.launchpad.net/ubuntu/+s...ux/+bug/571770

The three last bugs are related to the audio problems I'm seeing when using KMS, two of which I've filed myself.

I've been reluctant to file a bug at freedesktop.org so far, because I don't know if radeon KMS is directly to blame, but what I am sure of is that it triggers the problems. Whether there is some general kernel problem or if KMS alters conditions in a way that shake out bugs in other kernel drivers. So far it looks like only people with ATI IGP or Xnnnn-series cards are affected.

I'm on Ubuntu Lucid x86. Have tested up to mainline kernel 2.6.34-rc7, and the problem persists, so it's not Ubuntu-kernel specific.

[nanonyme]

Out of interest: got some easy way I could see if I can replicate this on Fedora 13? I've never had this kind of problems myself so just wondering. (as in, the audio part, I have no wireless)

[Porter]

I'm searching at freedesktop.org for any similar bugs... if I don't find any, I'll file a bug so that it's at least on their radar and they know of the problem.

If a bug report already exists, or if I create one (either way), I'll link the upstream in the relevant launchpad bugs.

[Porter]

OK, I filed a bug report on freedesktop.org and linked to the other reports on Launchpad and Bugzilla.

https://bugs.freedesktop.org/show_bug.cgi?id=28106


Apparently it may be due to a difference in PCI configuration between KMS and UMS?

Øyvind Stegard on Bugzilla had this to say on that...

Comment #29 From Øyvind Stegard 2010-05-10 17:31:29 (-) [reply] -------

(In reply to comment #28)
> Thank you for testing. Appearently, there might be a clash in the PCI device or
> bus configuration. Unfortunately, I'm not expert in this area. It might be a
> good idea to check the PCI device setup using lspci (compare good and bad
> behaviour).

Indeed, the PCI configuration for radeon is different under KMS (bad) compared
to UMS (good):

--- lspci-radeon-good.txt    2010-05-10 19:13:21.003538070 +0200
+++ lspci-radeon-bad.txt    2010-05-10 19:19:45.254230183 +0200
[...]
-    Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR+ FastB2B- DisINTx-
+    Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR+ FastB2B- DisINTx+
[...]
-    Interrupt: pin A routed to IRQ 16
+    Interrupt: pin A routed to IRQ 30
[...]
-    Capabilities: [80] Message Signalled Interrupts: Mask- 64bit+ Queue=0/0 Enable-
-        Address: 0000000000000000  Data: 0000
+    Capabilities: [80] Message Signalled Interrupts: Mask- 64bit+ Queue=0/0 Enable+
+        Address: 00000000fee0100c  Data: 41b9
+    Kernel driver in use: radeon

There is no difference for the HDA Intel PCI configuration.

Also, KMS causes problems for my wireless PCI network controller/driver iwl3945 (drops connection during Flash fullscreen video streaming, when snd-hda-intel crackling is at its worst . So, my system probably has issues out of scope for this bug report, wrt. to KMS. Thank you for all help.

[Porter]

Ha, I just realized Øyvind is the one who posted earlier in this thread. Hi Øyvind! Thanks for the links and the good information, hopefully it will get in front of the right eyeballs with all the linking.

[tomm]

Yes this is a very annoying bug. I can't even get on some wireless networks with radeon KMS enabled. Disable it and I get full bars, 80%+ connection quality.

Thinkpad T60, Radeon X1300, iwl3945.

[Porter]

Apparently it's slightly more serious than previously thought... users are reporting full kernel crashes when switching the wifi off using the hardware on/off switch. Disabling radeon KMS causes the behavior to disappear.

https://bugs.launchpad.net/xserver-x...ti/+bug/555286


Can any of the radeon devs chime in if they see this, to let me know if they're aware of this at all? Thanks!

[agd5f]

Please try my suggestions on the bug report.