[blfs-dev] qemu-kvm

Qrux qrux.qed at gmail.com
Tue Feb 21 04:17:39 PST 2012


On Feb 21, 2012, at 3:16 AM, Nathan Coulson wrote:

> On Wed, Feb 1, 2012 at 12:47 PM, Qrux <qrux.qed at gmail.com> wrote:
>> 
>> On Jan 31, 2012, at 1:26 PM, Bruce Dubbs wrote:
>> 
>>> You know I added kvm-qemu to BLFS a couple of days ago, right?
>> 
>> I added a BLFS-style markdown page:
>> 
>>        https://github.com/qrux/xlapp/blob/master/README-blfs.md
>> 
>> It tries to mimic the style of the BLFS book.  I'm hoping it can make it into the book if someone will take a stab at translating it into the appropriate SGML, etc.
>> 
>> I'm not sure about putting the config into this page, the way you did for KVM.  It's a pretty lengthy subject.  How about creating an additional page(s) (also in the Virtualization section) for each virt system?  I know it might be a bit unusual, and perhaps without precedent, but (as Nathan said earlier) VMs are sort of in this gray, meta, territory.
>> 
>>        Q
> 
> the patch gives a 404, but I installed dev86, and acpica (Based upon
> the cblfs book)
> https://github.com/qrux/xlapp/blob/master/src/xen-4.1.2-no_firmware-1.patch

Yeah--I removed the patch, and replaced it with a sed line (Bruce suggested it, and it seemed right).  You can see it at:

	https://github.com/qrux/xlapp/blob/master/b2.20-xen

I use it to disable building the firmware...

> Installed via the following:
> 
> make MAKEFLAGS="-j 1" xen tools &&
> make install-xen install-tools

...but, if you were able to compile dev86 and acpica then you probably don't have to disable it.  I'm on pure 64-bit, so some of the Xen tools don't build correctly for me, because dev86 needs a 32-bit glibc.  So, I disable xen/tools/firmware (only because I can, since I don't plan to run non-Linux guests).

> (Decided to keep things simple, BTW, is -j1 the default for make?  and
> is there a problem with the xorg check?).  I have a 32bit glibc
> userspace, although no clue what dependencies 32bit had.

The xorg check is a problem for me--Xen seems to want it, and it seemed to bail for me when I left it in.  I use -j1 explicitly because my script environment defaults to -j4.

> I wonder if some of the kernel options have changed since 3.0/3.1
> though,  I was not finding a few of them easy to find

They are a...total nightmare.  They're spread out *everywhere* in menuconfig.  Feel free to look at my host (dom0) and guest (domU) configs:

	dom0: https://github.com/qrux/xlapp/blob/master/zkernel-3.1-dom0.config
	domU: https://github.com/qrux/xlapp/blob/master/zkernel-3.1-domU.config

I have a debug-ready dom0 kernel, too (which can be useful if you run into issues).

> now to figure out how to actually "use" xen.  In a few weeks, I will
> be managing xen based virtual machines. (yes Bruce, the shoe is on the
> other foot now).  Seems a bit more "integrated" then qemu is.

You may want to take a look at my make-domU script:

	https://github.com/qrux/xlapp/blob/master/xen/make-domu

Which creates a disk-image for a PV guest, and might give some clues for how to get a guest running (assuming you have the hypervisor and host ready to go).

I'll try to help if you have any questions along the way.  I've just about finished an LFS-7.0 compatible complete Xen build "from scratch", which bootstraps an LFS-7.0 host, and then uses that host to build an LFS-7.0 guests.

	Q




More information about the blfs-dev mailing list