Re: [Xen-devel] guest being crashed from viridian_start_apic_assist()

[Paul Durrant <Paul.Durrant@xxxxxxxxxx>]

> -----Original Message-----
> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> Sent: 03 June 2016 10:59
> To: Paul Durrant
> Cc: xen-devel
> Subject: RE: guest being crashed from viridian_start_apic_assist()
> 
> >>> On 03.06.16 at 11:39, <Paul.Durrant@xxxxxxxxxx> wrote:
> >> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> >> Sent: 03 June 2016 10:04
> >>
> >> to test a guest ACPI change I went through all Windows versions that
> >> I have installed anywhere, to find 32-bit Win7 dying (the other
> >> variants I tried worked fine), albeit retrying a couple of times I then
> >> didn't see this a second time. Is this something you've observed
> >> elsewhere? The APIC assist logic isn't that complicated, and I can't
> >> really spot any race in there...
> >
> > How did it die?
> 
> From the single domain_crash() in that function. No other information
> was available (apart from the register dump, which doesn't look like
> it would be very useful for analysis here).
> 

There is an inconsistency between the test to abort a previous assist, which is 
based on:

    irr = vlapic_find_highest_irr(vlapic);

    isr = vlapic_find_highest_isr(vlapic);
    if ( (isr & 0xf0) >= (irr & 0xf0) )

and the test to start a new one:

    isr = vlapic_find_lowest_vector(&vlapic->regs->data[APIC_ISR]);
    if ( isr >= 0 && isr < vector )

I suspect that may be the problem you hit.

  Paul

> Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel