Re: [discuss] Re: VIA x86-64 bootlogs needed

[Nicholas Miell <nmiell@comcast.net>]

On Wed, 2006-07-26 at 00:18 +0200, Andi Kleen wrote:
>  > MSI Master2-FAR with VIA K8T800, kernel-2.6.17-1.2157_FC5
> 
> Thanks.
> 
> > With any luck, this APIC rework will fix this board's habit of
> > spontaneously turning off interrupts at the IOAPIC level without the
> > kernel's knowledge.
> 
> Unlikely unfortunately. How does it look like when they get turned off?
> 

I stop getting any interrupts from the device in question (either the
on-board UHCI controllers or the on-board SATA controller) -- i.e. mouse
movement gets jerky because uhci-hcd is operating purely by polling or I
start getting libata timeouts in my logs.

I have an abomination of a systemtap script which calls
unmask_IO_APIC_irq directly which instantly "fixes" it when things go
wrong.

I also have another stap script which records stack traces when
mask_IO_APIC_irq and unmask_IO_APIC_irq called in order to check if the
IRQs are actually getting turned off without the kernel's knowledge, but
when I run that in the background, the problem doesn't ever manifest
itself. 

-- 
Nicholas Miell <nmiell@comcast.net>