Re: [PATCH] hv: prevent VMs make pCPU enter C-state.


Eddie Dong
 

-----Original Message-----
From: Zhao, Yuanyuan <yuanyuan.zhao@...>
Sent: Sunday, July 24, 2022 8:44 PM
To: Dong, Eddie <eddie.dong@...>; acrn-dev@...;
Mao, Junjie <junjie.mao@...>; Li, Fei1 <fei1.li@...>
Subject: Re: [acrn-dev] [PATCH] hv: prevent VMs make pCPU enter C-state.

On 2022/7/22 4:17, Dong, Eddie wrote:
I thought what you did is to remove the MONITOR cap from guest, and make
sure VM-exit to MONITOR execution of guest.

This is fine. Please make the commit message clear.
Ok, i will change it to:
"Remove the MONITOR capability from VMs which share pCPU with others.
For these VMs, execution of MONITOR and MWAIT will cause a VM-exit."


Also, we need to make it conditionally: for dedicated core case, we do want
to let vCPU execute MONITOR and enter deep C state.

OK. I will check if the VM is Prelaunched VM or RTVM before hide it.
The offline tool can tell you if the VM uses dedicated core or not even for postlaunched VM.

But if we simplify it to be for pre-launched VM only (as dedicated CPUs), it might be ok too. But please get +1 from Junjie.

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