Vertualization of Unmodified Operating Systems

Vertualization of Unmodified Operating Systems

[Dave Feustel]

AMD Pacifica and Intel's VT make possible the virtualization of 
unmodified operating systems. Is it still necessary to add code 
to the hypervisor to support specific operating systems, or can 
Xen, as written, support any arbitrary OS that successfully boots
on a PC? (I'm thinking of the BSDs here).

Thanks,
Dave Feustel
-- 
Lose, v., experience a loss, get rid of, "lose the weight"
Loose, adj., not tight, let go, free, "loose clothing"

Re: Vertualization of Unmodified Operating Systems

[Anthony Liguori]

On Sat, 15 Apr 2006 17:39:10 -0500, Dave Feustel wrote:

> AMD Pacifica and Intel's VT make possible the virtualization of unmodified
> operating systems. Is it still necessary to add code to the hypervisor to
> support specific operating systems, or can Xen, as written, support any
> arbitrary OS that successfully boots on a PC? (I'm thinking of the BSDs
> here).

This sort of thing has been addressed here before.  While theoritically,
VT and SVM ought to allow any OS to run under Xen, in practice, if an OS
hasn't been tested as a guest under Xen, it is likely to turn up some bugs
or incompleteness.  Over time, this will certainly be a less of an issue.

The problem has to do with the fact that different OS's will use different
instructions when accessing things like page tables.  Right now, Xen only
emulates the instructions that we know are used by the systems we test
with (things like Linux and certain versions of Windows).

Regards,

Anthony Liguori

> Thanks,
> Dave Feustel

Re: Re: Vertualization of Unmodified Operating Systems

[Dave Feustel]

On Saturday 15 April 2006 17:53, Anthony Liguori wrote:
> On Sat, 15 Apr 2006 17:39:10 -0500, Dave Feustel wrote:
> 
> > AMD Pacifica and Intel's VT make possible the virtualization of unmodified
> > operating systems. Is it still necessary to add code to the hypervisor to
> > support specific operating systems, or can Xen, as written, support any
> > arbitrary OS that successfully boots on a PC? (I'm thinking of the BSDs
> > here).
> 
> This sort of thing has been addressed here before. 

I know this and I appreciate your patience. I definitely don't
pick things up or figure things out as quickly now as I did 
when I was younger.

> While theoritically, 
> VT and SVM ought to allow any OS to run under Xen, in practice, if an OS
> hasn't been tested as a guest under Xen, it is likely to turn up some bugs
> or incompleteness.  Over time, this will certainly be a less of an issue.
> 
> The problem has to do with the fact that different OS's will use different
> instructions when accessing things like page tables.  Right now, Xen only
> emulates the instructions that we know are used by the systems we test
> with (things like Linux and certain versions of Windows).

Xen and OpenBSD running under Xen are rapidly rising to the top of my list
of things to work with as general availability of AM2-socket motherboards and 
revision F AMD64 chips approaches. Xen and hardware virtualization have been
for a while now at the very top of the list of topics I follow in the news.

Dave Feustel

> Regards,
> 
> Anthony Liguori
> 
> > Thanks,
> > Dave Feustel
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
> 

-- 
Lose, v., experience a loss, get rid of, "lose the weight"
Loose, adj., not tight, let go, free, "loose clothing"

Re: Re: Vertualization of Unmodified Operating Systems

[Ronald G Minnich]

Is there a VT or pacifica laptop out there? Or is that technology just 
"assumed" for later CPUs? I'd like to get such a laptop.

It seems to me that I could boot Plan 9 unmodified, and from there, 
develop the xenfront drivers I need, thus making test/debug much easier.

thanks

ron

Re: Re: Vertualization of Unmodified Operating Systems

[Bastian Blank]

[-- Attachment #1.1: Type: text/plain, Size: 391 bytes --]

On Sun, Apr 16, 2006 at 02:31:50PM -0600, Ronald G Minnich wrote:
> Is there a VT or pacifica laptop out there? Or is that technology just 
> "assumed" for later CPUs? I'd like to get such a laptop.

Yes. The new ThinkPads includes Core Duo CPUs with VT enabled.

Bastian

-- 
Each kiss is as the first.
		-- Miramanee, Kirk's wife, "The Paradise Syndrome",
		   stardate 4842.6

[-- Attachment #1.2: Digital signature --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

RE: Re: Vertualization of Unmodified Operating Systems

[Rolf Neugebauer]

Someone recently confirmed the IBM thinkpad t60p has a BIOS option to enable
VT.

Mac Mini's (and presumably the Apple notebooks) also have VT enabled, at
least the linux live cd has shows the VMX CPU flag

Rolf

> -----Original Message-----
> From: xen-devel-bounces@lists.xensource.com [mailto:xen-devel-
> bounces@lists.xensource.com] On Behalf Of Bastian Blank
> Sent: 16 April 2006 22:02
> To: Ronald G Minnich
> Cc: xen-devel@lists.xensource.com; Dave Feustel
> Subject: Re: [Xen-devel] Re: Vertualization of Unmodified Operating
> Systems
> 
> On Sun, Apr 16, 2006 at 02:31:50PM -0600, Ronald G Minnich wrote:
> > Is there a VT or pacifica laptop out there? Or is that technology just
> > "assumed" for later CPUs? I'd like to get such a laptop.
> 
> Yes. The new ThinkPads includes Core Duo CPUs with VT enabled.
> 
> Bastian
> 
> --
> Each kiss is as the first.
> 		-- Miramanee, Kirk's wife, "The Paradise Syndrome",
> 		   stardate 4842.6

Re: Re: Vertualization of Unmodified Operating Systems

[Randy Thelen]

Ronald G Minnich wrote:

> Is there a VT or pacifica laptop out there? Or is that technology  
> just "assumed" for later CPUs? I'd like to get such a laptop.
>
> It seems to me that I could boot Plan 9 unmodified, and from there,  
> develop the xenfront drivers I need, thus making test/debug much  
> easier.

All Apple Intel based Macintosh computers have VT-x capability (as of  
April, 2006; I can't speak for future products).  However, on some  
Mac Mini's a procedure is needed to enable VT-x.  I have followed the  
procedure 'documented' (I use that word in the loosest of meanings, I  
assure you) with correct results:

http://forum.parallels.com/thread577.html

However, I do not know of anyone who has run Xen on top of EFI based  
Intel machines.  EFI is Intel's replacement for BIOS.  The acronym  
expands to Extended Firmware Interface.  It's a much more powerful  
boot mechanism than BIOS and the tools for it are open sourced with  
the BSD license:

https://edk.tianocore.org/

In addition, in order to get Xen to work with Mac OS X as dom0, much  
work would have to go into porting the device drivers used for  
pciback, console, and other event channels; as well as porting the  
user applications.   It's a non-trivial porting effort, but I'm quite  
sure there are those who have contemplated it.  (Maybe there's a  
brave soul -doing- it?!)

-- Randy

Re: Re: Vertualization of Unmodified Operating Systems

[Yinghai Lu]

You can get Laptop from HP with turion X2. It will be amd64 with SVM
support and dual core.

YH

On 4/16/06, Ronald G Minnich <rminnich@lanl.gov> wrote:
> Is there a VT or pacifica laptop out there? Or is that technology just
> "assumed" for later CPUs? I'd like to get such a laptop.
>
> It seems to me that I could boot Plan 9 unmodified, and from there,
> develop the xenfront drivers I need, thus making test/debug much easier.
>
> thanks
>
> ron
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
>

Re: Re: Vertualization of Unmodified Operating Systems

[George Dunlap]

I was pleasantly surprised when "cat /proc/cpuinfo" on new Dell laptop
w/ a Core Duo listed 'vmx' as one of the features.  Haven't tested it
yet, tho...
 -George

On 4/16/06, Ronald G Minnich <rminnich@lanl.gov> wrote:
> Is there a VT or pacifica laptop out there? Or is that technology just
> "assumed" for later CPUs? I'd like to get such a laptop.
>
> It seems to me that I could boot Plan 9 unmodified, and from there,
> develop the xenfront drivers I need, thus making test/debug much easier.
>
> thanks
>
> ron
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
>

Re: Re: Vertualization of Unmodified Operating Systems

[Stefan Kaltenbrunner]

Dave Feustel wrote:
> On Saturday 15 April 2006 17:53, Anthony Liguori wrote:
> 
>>On Sat, 15 Apr 2006 17:39:10 -0500, Dave Feustel wrote:
>>
>>
>>>AMD Pacifica and Intel's VT make possible the virtualization of unmodified
>>>operating systems. Is it still necessary to add code to the hypervisor to
>>>support specific operating systems, or can Xen, as written, support any
>>>arbitrary OS that successfully boots on a PC? (I'm thinking of the BSDs
>>>here).
>>
>>This sort of thing has been addressed here before. 
> 
> 
> I know this and I appreciate your patience. I definitely don't
> pick things up or figure things out as quickly now as I did 
> when I was younger.
> 
> 
>>While theoritically, 
>>VT and SVM ought to allow any OS to run under Xen, in practice, if an OS
>>hasn't been tested as a guest under Xen, it is likely to turn up some bugs
>>or incompleteness.  Over time, this will certainly be a less of an issue.
>>
>>The problem has to do with the fact that different OS's will use different
>>instructions when accessing things like page tables.  Right now, Xen only
>>emulates the instructions that we know are used by the systems we test
>>with (things like Linux and certain versions of Windows).
> 
> 
> Xen and OpenBSD running under Xen are rapidly rising to the top of my list
> of things to work with as general availability of AM2-socket motherboards and 
> revision F AMD64 chips approaches. Xen and hardware virtualization have been
> for a while now at the very top of the list of topics I follow in the news.

OpenBSD 3.9 works quite fine (installed using the native installer in
the virtualized environment!) as an unmodified guest on my Intel VT box,
with following caveats:

*) pcn(4) - aka AMD Pcnet does not seem to work well with the emulated
one (send works - receive does not)

*) ne(4) does work but is complaining about corrupted nic memory under
heavy traffic (does not seem to affect it much other than logging th errors)


Stefan

Re: Vertualization of Unmodified Operating Systems

[M.A. Williamson]

>AMD Pacifica and Intel's VT make possible the virtualization of 
>unmodified operating systems. Is it still necessary to add code 
>to the hypervisor to support specific operating systems, or can 
>Xen, as written, support any arbitrary OS that successfully boots
>on a PC? (I'm thinking of the BSDs here).

Shouldn't be necessary to add special support usually... However, if an OS 
makes use of particularly esoteric hardware features that other OSes do not 
use then it's possible those will not be emulated in Xen yet. In these 
cases, one would add the emulation to Xen and then that OS (and any others 
which used those features) would be able to run unmodified.

x86 is full of little "tricks", of the "whoa, people (can) do that?" 
variety ;-)

Cheers,
Mark

RE: Re: Vertualization of Unmodified Operating Systems

[Ian Pratt]

> > Xen and OpenBSD running under Xen are rapidly rising to the 
> top of my 
> > list of things to work with as general availability of AM2-socket 
> > motherboards and revision F AMD64 chips approaches. Xen and 
> hardware 
> > virtualization have been for a while now at the very top of 
> the list of topics I follow in the news.
> 
> OpenBSD 3.9 works quite fine (installed using the native 
> installer in the virtualized environment!) as an unmodified 
> guest on my Intel VT box, with following caveats:
> 
> *) pcn(4) - aka AMD Pcnet does not seem to work well with the 
> emulated one (send works - receive does not)
> 
> *) ne(4) does work but is complaining about corrupted nic 
> memory under heavy traffic (does not seem to affect it much 
> other than logging th errors)

Thanks for the success report. We haven't been testing OpenBSD so its
useful to hear it works OK. 

I'd be interested to hear whether pcnet32 works under the latest version
of qemu, in which case it should be easy to fix on xen/vt.

Also, are you using a 32, 32p, or 64 bit hypervisor? Have you tried
other combinations? (I'm guessing you're using a 32b guest, but do you
have pae enabled?)
Have you tried other combinations? Are there any stress tests you can
run to give it a good workout?

Thanks,
Ian

Re: Re: Vertualization of Unmodified Operating Systems

[Stefan Kaltenbrunner]

Ian Pratt wrote:
>>>Xen and OpenBSD running under Xen are rapidly rising to the 
>>
>>top of my 
>>
>>>list of things to work with as general availability of AM2-socket 
>>>motherboards and revision F AMD64 chips approaches. Xen and 
>>
>>hardware 
>>
>>>virtualization have been for a while now at the very top of 
>>
>>the list of topics I follow in the news.
>>
>>OpenBSD 3.9 works quite fine (installed using the native 
>>installer in the virtualized environment!) as an unmodified 
>>guest on my Intel VT box, with following caveats:
>>
>>*) pcn(4) - aka AMD Pcnet does not seem to work well with the 
>>emulated one (send works - receive does not)
>>
>>*) ne(4) does work but is complaining about corrupted nic 
>>memory under heavy traffic (does not seem to affect it much 
>>other than logging th errors)
> 
> 
> Thanks for the success report. We haven't been testing OpenBSD so its
> useful to hear it works OK. 

heh - I actually tried to install NetBSD and FreeBSD too but both crash
very early in their installer routines.

> 
> I'd be interested to hear whether pcnet32 works under the latest version
> of qemu, in which case it should be easy to fix on xen/vt.

will try that during the next few days

> 
> Also, are you using a 32, 32p, or 64 bit hypervisor? Have you tried
> other combinations? (I'm guessing you're using a 32b guest, but do you
> have pae enabled?)

32bit hypervisor currently - but I plan to change that and will test
with a 64b guest too.

> Have you tried other combinations? Are there any stress tests you can
> run to give it a good workout?

it successfully completes a full recompile of kernel & userspace which
is usually quite a good test for "working hardware" :-)


Stefan

Re: Re: Vertualization of Unmodified Operating Systems

[Stefan Kaltenbrunner]

Ian Pratt wrote:
>>>Xen and OpenBSD running under Xen are rapidly rising to the 
>>
>>top of my 
>>
>>>list of things to work with as general availability of AM2-socket 
>>>motherboards and revision F AMD64 chips approaches. Xen and 
>>
>>hardware 
>>
>>>virtualization have been for a while now at the very top of 
>>
>>the list of topics I follow in the news.
>>
>>OpenBSD 3.9 works quite fine (installed using the native 
>>installer in the virtualized environment!) as an unmodified 
>>guest on my Intel VT box, with following caveats:
>>
>>*) pcn(4) - aka AMD Pcnet does not seem to work well with the 
>>emulated one (send works - receive does not)
>>
>>*) ne(4) does work but is complaining about corrupted nic 
>>memory under heavy traffic (does not seem to affect it much 
>>other than logging th errors)
> 
> 
> Thanks for the success report. We haven't been testing OpenBSD so its
> useful to hear it works OK. 
> 
> I'd be interested to hear whether pcnet32 works under the latest version
> of qemu, in which case it should be easy to fix on xen/vt.
> 
> Also, are you using a 32, 32p, or 64 bit hypervisor? Have you tried
> other combinations? (I'm guessing you're using a 32b guest, but do you
> have pae enabled?)
> Have you tried other combinations? Are there any stress tests you can
> run to give it a good workout?

played a bit with various combinations (after upgrading to a 64 bit
hypervisor) and it looks like I'm hitting some issues (not too surprising):

*) a 32b guest seems to continue to work though starting more than one
(either another OpenBSD instance or WindowsXP) seems to completely
freeze the whole box (no logs or errors - just freezes).
Starting a single unmodified guest does not seem to trigger that issue ...

*) trying to install a 64bit guest segfaults qemu-dm when the installer
tries to boot the initial kernel:

qemu-dm[5117]: segfault at 00002ab60f896008 rip 00002ab5efa21040 rsp
00007fffffcb8f48 error 4

this on should be quite easy to reproduce by downloading
ftp://ftp.openbsd.org/pub/OpenBSD/snapshots/amd64/cd39.iso (4MB) and
using that as "cdrom" in the configfile.
Boot from there and watch the VNC connection dying a few seconds later :-)

all this testing  is with a checkout of the xen-unstable tree from a day
ago or so on a Pentium D920 box.


Stefan

RE: Re: Vertualization of Unmodified Operating Systems

[Ian Pratt]

> > Thanks for the success report. We haven't been testing 
> OpenBSD so its 
> > useful to hear it works OK.
> 
> heh - I actually tried to install NetBSD and FreeBSD too but 
> both crash very early in their installer routines.

I expect that the crash is in some grotty real mode code, as a result of
the 'vmxassist' code being incomplete. This code isn't used on AMD-V (it
has h/w support for real mode virtualziation), so it might work.

There are already plans to enhance vmxassist so that it's more complete.

Ian