Bug 725268

Summary:	-smp 1,cores=2 has weird semantics
Product:	Red Hat Enterprise Linux 6	Reporter:	Cole Robinson <crobinso>
Component:	qemu-kvm	Assignee:	Virtualization Maintenance <virt-maint>
Status:	CLOSED NOTABUG	QA Contact:	Virtualization Bugs <virt-bugs>
Severity:	unspecified	Docs Contact:
Priority:	unspecified
Version:	6.2	CC:	juzhang, mkenneth, tburke, virt-maint
Target Milestone:	rc
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2011-08-11 21:44:14 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Cole Robinson 2011-07-24 22:32:49 UTC

I'd expect -smp 1,cores=2 to give me a guest with a dual core socket == 2 logical CPUs exposed to the guest. Instead I see just 1 logical CPU in /proc/cpuinfo: a socket claiming 2 cores but only 1 core is listed.

Ideally I'd think that if any topology info is specified, old style -smp X is ignored, and the logical CPU number is just calculated, defaulting unspecified topology values to 1. Or have old style -smp X just be the old way of specifying sockets=X, and if sockets= is specified, -smp X is ignored.

And even if nothing is changed the docs should mention these semantics (which isn't really relevant for RHEL but still worth mentioning).

(see https://bugzilla.redhat.com/show_bug.cgi?id=708892#c2 for some more background)

Comment 2 Dor Laor 2011-08-11 21:44:14 UTC

(In reply to comment #0)
> I'd expect -smp 1,cores=2 to give me a guest with a dual core socket == 2
> logical CPUs exposed to the guest. Instead I see just 1 logical CPU in
> /proc/cpuinfo: a socket claiming 2 cores but only 1 core is listed.

I think the current syntax requires -smp 2, cores=2.
Agree it is not the easiest. I doubt we'll be able to change it now.
If the above does not work, please reopen the bug

> 
> Ideally I'd think that if any topology info is specified, old style -smp X is
> ignored, and the logical CPU number is just calculated, defaulting unspecified
> topology values to 1. Or have old style -smp X just be the old way of
> specifying sockets=X, and if sockets= is specified, -smp X is ignored.
> 
> And even if nothing is changed the docs should mention these semantics (which
> isn't really relevant for RHEL but still worth mentioning).
> 
> (see https://bugzilla.redhat.com/show_bug.cgi?id=708892#c2 for some more
> background)