Edgar Fuß wrote:
>> Let me know what you see.
>>     
> OK, it may be a HARDWARE bug.
>
> What I saw was that the "receipt" bit was set for pin1 in ioapi1,
> but Xintr_ioapic_level10 was not being called. Looked like a missing EOI.
> All the vectors and idt entries looked reasonable.
>
> Googling revealed that linux had a strange workaround
> (arch/i386/kernel/io_apic.c):
>   
Thanks for the data, Edgar, that's very useful. Just out of curiosity, 
what is the reported version of the ioapic in question?

- Frank