[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH] Remove a set operation for VCPU_KICK_SOFTIRQ when post interrupt to vm.



Liuqiming (John) wrote on 2015-10-10:
> 
> On 2015/10/10 14:32, Zhang, Yang Z wrote:
>> Zhang, Yang Z wrote on 2015-09-24:
>>> Liuqiming (John) wrote on 2015-09-24:
>>>> 
>>>> On 2015/9/24 11:25, Zhang, Yang Z wrote:
>>>> it completed, the following vmentry will pick up the pending interrupt.
>>>> If you send the ipi unconditionally, actually it is received by
>>>> hypervisor and the interrupt handler dose nothing. You still rely
>>>> on the vmentry to pick up the interrupt. My confusion is:  when
>>>> vmentry, does physical CPU inject all interrupts on PIR into VM
>>>> and clear POSTED_INTR_ON bit? Or just inject the highest index of
>>>> PIR and may leave POSTED_INTR_ON bit being set to 1?
>>> I need to say your understanding is wrong. What we need to do is
>>> sync the pir to irr and let hardware or hypervisor to handle left
>>> part. So our focus is how to sync pir to irr correctly and timely.
>>> We don't care how the interrupt in PIR is injected to guest.
>> Hi Qiming, is there any other question with it? Also, do you have any
>> chance to test it?
>> 
>> Best regards,
>> Yang
>    No other question.
>    I have run some vm start-shutdown test, and haven't gotten a chance
> to test the performance.

Hi Qiming

Have your tested the performance? If it fixes your performance issue, I will 
send out a formal patch. 

btw, is it ok to add you in signed-by since the origin one is from you?

Best regards,
Yang


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


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.