Question about HLT VMExit reason

Question about HLT VMExit reason

[zhangtrump]

[-- Attachment #1.1: Type: text/plain, Size: 483 bytes --]

Hi, all:
     I was testing networking performance from DomU to DomU running on different hosts. I used xentrace and xenalyze tools to get the trace info of receive side, and found that there was a time when vcpus (which handle the receive work) were in "Blocked" state.     From the information of "VMExit" reason gotten by xenalyze tool, the "HLT" is the main exit reason. I am wondering when DomU will be exited because of "HLT"?      Thanks.
------Best RegardsJames 		 	   		  

[-- Attachment #1.2: Type: text/html, Size: 843 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

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

Re: Question about HLT VMExit reason

[Andrew Cooper]

[-- Attachment #1.1: Type: text/plain, Size: 665 bytes --]

On 12/02/15 09:24, zhangtrump wrote:
> Hi, all:
>
>      I was testing networking performance from DomU to DomU running on
> different hosts. I used xentrace and xenalyze tools to get the trace
> info of receive side, and found that there was a time when vcpus
> (which handle the receive work) were in "Blocked" state.
>      From the information of "VMExit" reason gotten by xenalyze tool,
> the "HLT" is the main exit reason. I am wondering when DomU will be
> exited because of "HLT"? 
>      Thanks.

Yes.

When one vcpu halts, Xen will schedule another vcpu in its place.  There
is nothing to be gained at all by letting a pcpu stay halted in guest.

~Andrew

[-- Attachment #1.2: Type: text/html, Size: 1462 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

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

Re: Question about HLT VMExit reason

[Paul Durrant]

[-- Attachment #1.1: Type: text/plain, Size: 1270 bytes --]

I think that’s an answer to why hlt is a vmexit as opposed to when it is likely to occur…

This simple answer, of course, is the OS idle process will most likely the issuer of a hlt instruction and this is why you are seeing them so frequently. Give your guest some compute-bound task and you’ll  mostly likely see a big fall-off in the number of hlts.

  Paul

From: xen-devel-bounces@lists.xen.org [mailto:xen-devel-bounces@lists.xen.org] On Behalf Of Andrew Cooper
Sent: 12 February 2015 10:07
To: zhangtrump; xen-devel@lists.xen.org
Subject: Re: [Xen-devel] Question about HLT VMExit reason

On 12/02/15 09:24, zhangtrump wrote:
Hi, all:

     I was testing networking performance from DomU to DomU running on different hosts. I used xentrace and xenalyze tools to get the trace info of receive side, and found that there was a time when vcpus (which handle the receive work) were in "Blocked" state.
     From the information of "VMExit" reason gotten by xenalyze tool, the "HLT" is the main exit reason. I am wondering when DomU will be exited because of "HLT"?
     Thanks.

Yes.

When one vcpu halts, Xen will schedule another vcpu in its place.  There is nothing to be gained at all by letting a pcpu stay halted in guest.

~Andrew

[-- Attachment #1.2: Type: text/html, Size: 25788 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

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