| [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
 Re: [Xen-users] Multiple VCPUs
 
To: "Betak, Travis" <travis.betak@xxxxxxx>From: "Jared Bellows" <xen@xxxxxxxxxxxxxxx>Date: Fri, 30 Jun 2006 13:42:47 -0600Cc: "Petersson, Mats" <Mats.Petersson@xxxxxxx>,	Xen Users <xen-users@xxxxxxxxxxxxxxxxxxx>Delivery-date: Fri, 30 Jun 2006 12:43:27 -0700Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com;	h=received:message-id:date:from:sender:to:subject:cc:in-reply-to:mime-version:content-type:references:x-google-sender-auth;	b=RqoyfMRVyi44xfJlwQTM3F69rw2PZLwoziwBX8vJE8ILWOwgG5jCqOUGB7/kuQFn+L9Hb19uNz9hGobjLYsp/72aA2lvwCviw83Ph/qJFiZyWJ0F7EnjVup2A/OL9v5gdSEu3OkaTeGq7d6WARR88cHo16N+IgffiYF6zVgHj+E=List-id: Xen user discussion <xen-users.lists.xensource.com> 
 On 6/29/06, Betak, Travis <travis.betak@xxxxxxx> wrote:
Petersson, Mats wrote:> Not sure if this will solve your problem or not, but if I
 > understand things right, the HVM guest needs to have
 > MP-tables (MP=multiprocessor) generated during startup (or
 > hard-coded in BIOS), which the default build didn't have
 > until my colleague Travis sent some patches in a few days
 > ago. So anything other than unstable is doesn't have this
 > change - it went in Wednesday (yesterday).
 >
 > Get the latest unstable and it should work, or
 > .../tools/firmware/rombios/Makefile to have
 > BIOS_BUILDS += BIOS-bochs-8-processors
 >
 > and change .../tools/firmware/hvmloader/Makefile to use
 >
 > sh ./mkhex rombios ../rombios/BIOS-bochs-8-processors > roms.h
 >
 > [And remove the other BIOS line, of course]
 >
 > > From: 
xen-users-bounces@xxxxxxxxxxxxxxxxxxx
 > > [mailto:xen-users-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Jared
 > > Bellows Sent: 29 June 2006 18:13
 > > To: Xen Users
 > > Subject: [Xen-users] Multiple VCPUs
 > >
 > >
 > > I have a self built system using an Intel D 920
 > > processor and a MB that supports VT. I'm able to run HVM
 > > domains fine, but have trouble getting multiple VCPUs for
 > > these domains. Here is an example of one of my configs.
 > >
 > > ...
 > >
 > >     I'm running Xen 3.0.2-2
 from the binary download. The
 > > HVM domain only sees 1 cpu and xm list displays on VCPU.
 > >
 
 This is the main reason I submitted this patch.  Before my
 change, the firmware had to be rebuilt specifically for
 the processor configuration.
 
 First, for SMP you'll need to enable the apic in your
 hvmbuilder config and maybe acpi also.  This may be the only
 thing getting in the way of a working SMP guest.
 
 Second, like Mats said, you may need firmware with MP
 tables built into it.  Some older OS's rely upon these
 tables.  I would suggest building the ROMBIOS that matches
 the number of VCPUs you want (e.g. BIOS-bochs-2-processors
 for 2 VCPUs).  If you do this, you will need the xen
 source, but you won't need to build everything, just the
 HVM firmware.  You'll need the dev86 package (or bcc if
 your a debian user) to build the firmware.  Also, if you
 perform this step, make sure the source is
 pre-changeset-10520 because the firmware created by my
 patch isn't compatible with older xen binaries.
 
 If you dare, you can download the get the latest xen-unstable
 (at least c/s 10520 with my dynamic mp table patch)
 and build xen from scratch.  This way, all you need to
 do is enable apic and maybe acpi in the hvmbuilder config;
 the firmware should take care of itself.  Of course this
 is the unstable tree so it may or may not work depending
 on the planets alignment ;)
 
 All these steps probably sound a little discouraging.
 I hope this helps you out a little.
 
 --travis
 
 
 
   I opted to use the 3.0.2-2 source code to recompile the HVM firmware. When the domain boots it now says 2 processors in the BIOS boot. xm list still only displays 1 VCPU for the HVM domain and I'm unable to get an OS to boot (both Linux and Windows) with the  new HVM firmware. I tried all iterations of acpi and apic. 
   I'm thinking of trying the 3.0-testing in hopes that it might fix the problem and have more stability than unstable. If you have any other ideas, I'd be happy to hear them. Thanks for your help thus far.
 _______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users 
 |