Laptops and hardware virtualization

[Jerry Feldman]

On 02/15/2010 12:38 PM, Jarod Wilson wrote:
> On Mon, Feb 15, 2010 at 12:35 PM, Jarod Wilson <jarod-ajLrJawYSntWk0Htik3J/w at public.gmane.org> wro=
te:
>  =20
>> On Mon, Feb 15, 2010 at 11:06 AM, Richard Pieri <richard.pieri at gmail.c=
om> wrote:
>>    =20
>>> On Feb 14, 2010, at 5:32 PM, Scott Ehrlich wrote:
>>>      =20
>>>> What is the desired end result?   As I learned from this list, among=

>>>> others, the VT support in BIOS only appears to affect the ability of=

>>>> the system to run a 64-bit VM.  Am I missing something else?    Was
>>>> there something in the original question of the OP I lost track of?
>>>>        =20
>>> I'm going to simplify things a bit...
>>>
>>> Usually, the hypervisor (whether type 1 or type 2) runs in privileged=
 mode and guest operating systems run in non-privileged mode.  When a gue=
st attempts to execute a privileged instruction the CPU throws an excepti=
on.  The hypervisor traps that exception and emulates the instruction in =
some manner, passing the result back to the guest.  The technique is call=
ed "trap and emulate".
>>>
>>> The problem -- I should say one of the problems -- with the x86 archi=
tecture is that there are a slew of privileged instructions that don't ge=
nerate exceptions in non-privileged mode.  The way that VMware and others=
 got around that was by emulating an entire x86 CPU in the hypervisor.  I=
f you ever used older versions of VMware Desktop then you're familiar wit=
h the CPU load that the virtual machine incurs.
>>>
>>> Along come AMD-V and VT-x.  These extensions to the x86 architecture =
provide wrappers around some (but not all) of those privileged instructio=
ns.  These wrappers generate exceptions in non-privileged mode.  The hype=
rvisor can trap the exceptions and emulate them.  If you've used VMware o=
r VirtualBox or Parallels on a recent system with the virtualization exte=
nsions then you've seen how low the VM overhead is.  If you cannot enable=
 the extensions then the hypervisor has to fall back to CPU emulation.
>>>
>>> Which brings me to the answer.  You can't run 64-bit guests without A=
MD-V/VT-x because the CPU emulators do not emulate 64-bit instructions.
>>>      =20
>> Only, yes, you can. As I said elsewhere in this thread yesterday, the
>> first-gen AMD Opteron processors, which have no AMD-V, *do* in fact
>> support 64-bit guests under VMware.
>>    =20
> First-gen Opteron Rev. E and Athlon 64 Rev. D and newer, that is.
>
> http://kb.vmware.com/selfservice/microsites/search.do?language=3Den_US&=
cmd=3DdisplayKC&externalId=3D1003945
>
>
>  =20
This I think is one of the things that differentiates VMware in that it
can run without the need for virtualization support.

--=20
Jerry Feldman <gaf-mNDKBlG2WHs at public.gmane.org>
Boston Linux and Unix
PGP key id: 537C5846
PGP Key fingerprint: 3D1B 8377 A3C0 A5F2 ECBB  CA3B 4607 4319 537C 5846