RE: [Xen-devel] interrupt delivery to VCPU

["Agarwal, Lomesh" <lomesh.agarwal@xxxxxxxxx>]

Does it make sense to change interrupt delivery to VCPU which is going to be 
runnable next so that interrupt latency can be minimized for HVM guest?

-----Original Message-----
From: Shan, Haitao 
Sent: Saturday, October 20, 2007 6:05 PM
To: Agarwal, Lomesh; Keir Fraser; xen-devel@xxxxxxxxxxxxxxxxxxx
Subject: RE: [Xen-devel] interrupt delivery to VCPU

Maybe windows uses IPI for forwarding interrupt to another processor?
>From my observation, SMP Windows XP handles RTC timer interrupt this way.

-----Original Message-----
From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx 
[mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Agarwal, Lomesh
Sent: 2007年10月20日 5:51
To: Keir Fraser; xen-devel@xxxxxxxxxxxxxxxxxxx
Subject: RE: [Xen-devel] interrupt delivery to VCPU

Does anybody know how windows rebalances irqs among processors?

-----Original Message-----
From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
[mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Keir Fraser
Sent: Friday, October 19, 2007 1:49 PM
To: Agarwal, Lomesh; xen-devel@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-devel] interrupt delivery to VCPU

I don't know about the VT-d support. I doubt that's been optimised so
much.
I only know about physical IRQ delivery to PV guests. For Linux, a
daemon
(irqbalanced) usually periodically rebalances irqs by writing affinity
info
to /proc/irq/#/affinity.

 -- Keir


On 19/10/07 21:45, "Agarwal, Lomesh" <lomesh.agarwal@xxxxxxxxx> wrote:

> What happens for windows HVM guest?
> Also for PV Linux guest do you know how does Linux setup irq affinity?
> 
> -----Original Message-----
> From: Keir Fraser [mailto:Keir.Fraser@xxxxxxxxxxxx]
> Sent: Friday, October 19, 2007 1:42 PM
> To: Agarwal, Lomesh; xen-devel@xxxxxxxxxxxxxxxxxxx
> Subject: Re: [Xen-devel] interrupt delivery to VCPU
> 
> The guest can choose which VCPU gets any particular interrupt. This is
> tied
> into Linux's normal irq affinity logic, for example.
> 
>  -- Keir
> 
> On 19/10/07 21:39, "Agarwal, Lomesh" <lomesh.agarwal@xxxxxxxxx> wrote:
> 
>> How does Xen deliver an interrupt to VCPU? If there are 2 VCPUs for a
>> guest and there is no interrupt affinity set which VCPU will get the
>> interrupt? is it round robin?
>> 
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@xxxxxxxxxxxxxxxxxxx
>> http://lists.xensource.com/xen-devel



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel