RE: [Xen-users] amd processors + virtualisation

["Petersson, Mats" <Mats.Petersson@xxxxxxx>]

> -----Original Message-----
> From: Thomas Harold [mailto:tgh@xxxxxxxxxxxx] 
> Sent: 29 May 2007 17:09
> To: Petersson, Mats
> Cc: xen-users@xxxxxxxxxxxxxxxxxxx
> Subject: Re: [Xen-users] amd processors + virtualisation
> 
> Petersson, Mats wrote:
> >> On the newer CentOS5 boxes using AM2 Athlon64 X2 and 
> SocketF Opteron 
> >> CPUs, the default CentOS5 kernel does allow the "svm" flag to peek 
> >> through.  And I have not installed virtualization support on 
> >> this box (yet).
> > 
> > Presumably, you have a recent enough kernel now (it's been 
> in there for
> > quite some time, I spent a few minutes trying to locate the version
> > where it went in, but I (using google) wasn't able to find it...)
> 
> Ah, that makes sense.  (Goes poking through his session logs...)
> 
> Here's the original install, probably using Gentoo 2006.0 w/ 
> the 2.6.15 
> kernel (but Gentoo's version).  As you can see, the grep 
> didn't find the 
> "svm" flag.
> 
> livecd ~ # cat /proc/cpuinfo | grep 'flags'
> flags           : fpu vme de pse tsc msr pae mce cx8 apic sep 
> mtrr pge 
> mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mm
> xext fxsr_opt lm 3dnowext 3dnow pni cx16 lahf_lm cmp_legacy
> 
> livecd ~ # cat /proc/cpuinfo | grep 'svm'
> livecd ~ # cat /proc/version
> Linux version 2.6.15-gentoo-r5 (root@poseidon) (gcc version 3.4.4 
> (Gentoo 3.4.4-r1, ssp-3.4.4-1.0, pie-8.7.8)) #1 SMP Tue Feb 21 17:
> 19:47 UTC 2006
> livecd ~ #
> 
> Here's an entry from the next day after upgrading to 2.6.17.  
> This one 
> shows the "svm" flag.
> 
> azure ~ # cat /proc/version
> Linux version 2.6.17-gentoo-r4 (root@livecd) (gcc version 
> 3.4.4 (Gentoo 
> 3.4.4-r1, ssp-3.4.4-1.0, pie-8.7.8)) #1 SMP Sat Aug 26 00:34
> :30 EDT 2006
> 
> livecd ~ # cat /proc/cpuinfo | grep 'flags'
> flags           : fpu vme de pse tsc msr pae mce cx8 apic sep 
> mtrr pge 
> mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mm
> xext fxsr_opt lm 3dnowext 3dnow pni cx16 lahf_lm cmp_legacy 
> svm cr8_legacy
> 
> So it was probably either 2.6.16 or 2.6.17 where the code for 
> the "svm" 
> flag was added.

Probably 2.6.16 then, as we've had that in Xen for quite some time, and
I know it worked there... 
> 
> ...
> 
> I wasn't sure how the kernel detected those flags.  I didn't know 
> whether it was special detection code (which had to be 
> updated for new 
> features) or whether the processor spit back a string listing.

It's not produced by the processor itself - the processor has a set of
bits reported via the CPUID instruction that are either 1 or 0 based on
what the processor can or can't do (e.g. it's a 1 in a particular bit to
say SVM is possible, another to say "supports Long Mode"[64-bit],
another for MMX, SSE[1, 2, 3], and so on). 

Some CPUID information is more complex, such as the number of bytes of
cache or number of bytes per cache-line, etc, but in summary, the
processor has "bits" to tell you what it can and can't do. The kernel
contains code to "decode" those bits into more easily readable symbols. 

--
Mats
> 
> Thanks for the responses.
> 
> 
> 



_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users