Re: VMI Interface Proposal Documentation for I386, Part 5

[Zachary Amsden]

Pavel Machek wrote:
> > from the hypervisor perspective - if the guest enables interrupts, and 
> > you have something pending to deliver, for correctness, you have to 
> > deliver it, right now.  But does the kernel truly require that interrupt 
> > deliver immediately - in most cases, no.  In particular, on the fast 
> >     
>
> I'd say PCI hardware can delay interrupts for any arbitrary
> delay... so if driver expects to get them "immediately", I'd say it is
> broken. It should be enough to deliver them "soon enough", like not
> more than 1msec late...
>   

I agree.  One case we hit that did cause us a bug was local APIC 
delivery of self-IPIs.  I didn't dig too deep into why Linux was unhappy 
without immediate delivery (we deferred delivery here unnecessarily, but 
did not stop it).  I believe this was in SMP specific code that was 
using self-IPIs to regenerate IRQs .

Zach
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/