On Monday 21 August 2006 00:00, Martin Husemann wrote:

>  This changes fixes the problem, at the expense of suboptimal
> interrupt routing (ath and auich now also go to IRQ 11, where nearly
> all other devices already are).

FWIW, I've been having this exact same problem recently on my Dell 8600 
after updating to -current a few weeks ago. Except s/ath0/iwi0/. I 
"fixed" it in acpi_pci_link.c in much the same way as Martin. As in, put 
everything on irq11.

Forcing either irq5 (Windows XP's preferred irq for these two devices) or 
irq9 for iwi0 and auich0 works for a short time as long as the interrupt 
load is low. Under moderate interrupt load (playing some audio through 
auich0 while accessing a network over iwi0) things go downhill fast. 
Both audio and wireless stop; seemingly no more interrupts are 
dispatched on that particular interrupt line.

When using irq9 (shared with the SCI), acpilid0 open/close events are no 
longer reported when auich0/iwi0 interrupts are stalled.

Cheers, Steve