Port-xen archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Installing FreeBSD 8 as domU

On 05/30/11 18:29, Thor Lancelot Simon wrote:
> I have an old FreeBSD 8.0-BETA2 installation I built under HVM.  It
> is difficult (in the extreme) to upgrade to FreeBSD 8.2 since their
> "mergemaster" tool has trouble with the source versions of various
> config files that were shipped in the beta versions.
> So I am trying to do a fresh domU install of FreeBSD 8.2.  I
> believe it should be possible to do an installation into a HVM
> using the normal FreeBSD install tools, then switch to the FreeBSD HVMPV
> kernel and get improved disk and network performance.
> However, this presently isn't working.  The boot unpredictably hangs
> either partway through autoconf, or immediately after printing out
> that it's starting sysinstall as init.  If I turn on "verbose boot
> logging (option 5 from the FreeBSD boot menu)" the timing of the hang
> gets more unpredictable, but always seems to come at some point during
> autoconf.
> Booting FreeBSD with ACPI disabled (not necessary for my old 8.0 image)
> reliably gets me to "/stand/sysinstall running as init on vty0" at
> which point things seem to lock up hard.
> Help!  What's the right way to get this to go?
> The dom0 is currently NetBSD-5-stable just prior to the 5.1 release
> point.  Xen kernel is 3.3, xentools is xentools33 of the same vintage
> as the NetBSD kernel.  I'm certainly glad to change this around if
> anyone thinks it'll help.
> The FreeBSD image is i386 as I need to use it to run old FreeBSD 4
> binaries in a chroot, and that doesn't work well with a 64 bit FreeBSD
> kernel these days.
> Xen virtual's config file follows.
> Thor

Hello Thor,

Could this be the issue recently discussed in the list?


If it's the same thing, basically you'll want to update your xentools*
and use raw devices (or block, one of them will work).

Related to the above, I can reliably crash the dom0 [Manuel, I cc'd you
because I suspect it's a side effect of the patch that allows one to
boot FreeBSD HVM+PV] after starting/stopping FreeBSD several times.
Meant to write about this the other day, but got sidetracked.

One sees:
$ gzcat /var/log/messages.* | grep ref
May 29 12:01:39 minskytron /netbsd: xbdback backend/vbd/3/832: can't map
grant ref: 0/-1
May 29 12:01:39 minskytron /netbsd: xbdback backend/vbd/3/832: can't map
grant ref: 0/-1

It seems (if this info helps troubleshooting) that the last vm boot
before the dom0 spontaneously reboot will hang (like it would without
the patch). I would then destroy the VM forcibly, start it again, and
the moment I did that, the dom0 reboots.

I will be happy to crash my dom0 a few more times to help fix it!

But back to your problem -- that should be it. When you get it to boot,
you should run into another problem: xn0 on FreeBSD HVM+PV (tested i386
full pv, same thing) will probably give you trouble. I recently posted
in the freebsd-xen list about it, but so far nothing promising; see

There may be a patch in the thread that fixes the problem, but the text
in the link that I found in google is mangled. I tried to fix it
manually, but alas. No luck.

Let me know if you (don't) run into the same problem, and I hope the
above fixes the boot hang for you.



Home | Main Index | Thread Index | Old Index