Re: [RFC][PATCH 1/2] hv: unify kicking off vCPU by INIT signal


Eddie Dong
 


-----Original Message-----
From: Dong, Eddie <eddie.dong@...>
Sent: Thursday, August 25, 2022 7:35 AM
To: acrn-dev@...
Cc: Cao, Minggui <minggui.cao@...>
Subject: RE: [acrn-dev] [RFC][PATCH 1/2] hv: unify kicking off vCPU by
INIT signal

Hi Minggui:
Did u see any latency difference with INIT signal?
[Cao, Minggui] No. the latency of INIT or IPI to kick off vCPU could be similar
(both are from LAPIC). If needed, I'll think how to calculate the latency.
Because both INIT and IPI are sent to another core, the beginning TSC
is from one core, and end TSC are from target core.
This is fine. As if the delta between them using INIT vs. that of using IPI is close to be same, it is fine.


BTW, do we test the guest issued normal INIT/SIPI usage? Make sure
it work well, though not directly related to this patch.
[Cao, Minggui] They shall work well. In ACRN HV, the guest VM INIT/SIPI goes
the path of VLAPIC (APIC-write or x2APIC MSR). They work for multi SMP cores.
I mean under the situation of LAPIC passthru, the VM can still issue INIT/SIPI signal to reboot. Can you have a test?

Join {acrn-dev@lists.projectacrn.org to automatically receive all group messages.