[PlanetCCRMA] IRQ problems on P8Z68-V ProGen3

[Fernando Lopez-Lezcano]

On 09/12/2012 12:28 PM, Stephan Neuhaus wrote:
>
> On Sep 11, 2012, at 07:13, Fernando Lopez-Lezcano wrote:
>
>> On 09/09/2012 02:02 AM, Stephan Neuhaus wrote:
>>> My sound card is a M-Audio Delta 1010, and it sits on IRQ 18.  Also on IRQ 18 sits something that might be a USB controller (more detailed information available on request).  When I try to do anything at all with the sound card (such as running jackd and connecting Hydrogen to it), I get "IRQ 18 resetting" messages in /var/log/messages, accompanied by massive xruns.
>>
>> Could you copy the exact error messages and anything else you see around them? What is it being shared with? ("grep 18 /proc/interrupts" to find out).
>
> Thanks, Fernando.
>
> Before I go into the technical details, let me express my huge admiration that you maintain this project so consistently for so long.  It can't always be easy, but without your good work, I'd never have started using Linux for audio.

I'm glad to hear that... yes, it's been a long time... there's now a 
very active group in Fedora itself doing what I used to do every day 
myself, the kernel is for now an exception that remains in my repository :-)

> Some updates: I misspelled the graphics card; it's a Radeon 7970, not a 7990 (if that even exists).  The CPU is an Intel Core I7-2600K.  I also rearranged the cards (graphics + sound) on the motherboard, but to no effect, other than that the sound card now shares a different interrupt with something else:
>
> [neuhaus at achilles ~]$ grep 19: /proc/interrupts
>   19:        147          0          0          0          0          0          0          0   IO-APIC-fasteoi   ahci, snd_ice1712
>
> The actual error messages start when i start qjackctl and try to start jack:
>
> Sep 12 20:13:19 achilles kernel: [  613.759528] IRQ 19 might be stuck.  Polling
> Sep 12 20:13:19 achilles kernel: [  614.158536] IRQ 19 might be stuck.  Polling
> Sep 12 20:13:20 achilles kernel: [  614.658201] IRQ 19 might be stuck.  Polling
> Sep 12 20:13:20 achilles kernel: [  615.157896] IRQ 19 might be stuck.  Polling
> Sep 12 20:13:21 achilles kernel: [  615.657565] IRQ 19 might be stuck.  Polling
...
> and at some point (3 minutes later), IRQ 19 seems to become unstuck again.
>
> Researching this a bit, I am now wondering whether this might not be an instance of this issue: https://bugzilla.redhat.com/show_bug.cgi?id=755956 (with more enlightening info to be had here https://bugzilla.redhat.com/show_bug.cgi?id=845211) since the motherboard seems to have the misbehaving PCIe/PCI bridge controller, called ASM108x.  Do you think that it might be a good idea to try booting with noirqdebug and irqpoll?

Sure, I don't think it could do harm (but don't take my word for it :-)

Looks like it is a current issue (resurfaced in the second link you sent 
- good detective work![*]). I have to read the second bugzilla report in 
detail to see if there might be a patch I need to drop from my kernel 
builds (I try to include all Fedora patches that can be made to work 
with the rt patch). I don't know when I'll have time - very busy these 
days... I could at least see if I can synch with the latest 3.4.10 koji 
build.

-- Fernando

[*] I vaguely remember this happening before and that it was fixed when 
I dropped some old(ish) patch I still had in my tree.