Re: [PATCH v2 03/11] x86/vlapic: introduce an EOI callback mechanism

[Jan Beulich <jbeulich@xxxxxxxx>]

On 30.09.2020 13:49, Paul Durrant wrote:
>> From: Xen-devel <xen-devel-bounces@xxxxxxxxxxxxxxxxxxxx> On Behalf Of Roger 
>> Pau Monne
>> Sent: 30 September 2020 11:41
>>
>> @@ -159,8 +184,12 @@ void vlapic_set_irq(struct vlapic *vlapic, uint8_t vec, 
>> uint8_t trig)
>>      else
>>          vlapic_clear_vector(vec, &vlapic->regs->data[APIC_TMR]);
>>
>> +    if ( callback )
>> +        vlapic_set_callback(vlapic, vec, callback, data);
>> +
> 
> Can this not happen several times before an EOI? I.e. the vector could
> already be set in IRR, right?

Yes, but I take it the assumption is that it'll always be the same
callback that ought to get set here. Hence the warning printk() in
that function in case it isn't.

What I wonder while looking at this function is whether the TMR
handling is correct. The SDM says "Upon acceptance of an interrupt
into the IRR, ..." which I read as "when the IRR bit transitions
from 0 to 1" (but I can see room for reading this differently).

Jan