kexec of Xen hypervisor from a Linux payload

[Trammell Hudson]

I've successfully built a 4.6.4 Linux kernel payload for CoreBoot and
flashed it onto the top 4 MB of the boot ROM on my Thinkpad x230.
The runtime is a dynamic linked busybox with glibc, as well as a copy
of the kexec binary from my Ubuntu laptop.

kexec of a xen kernel (multiboot-x86) results in an immediate reboot
into the coreboot payload:

https://www.flickr.com/photos/osr/27913702804/lightbox

kexec of a Linux kernel produces the expected result (panic in this
case since I didn't pass any additional parameters to find a root fs):

https://www.flickr.com/photos/osr/27914585663/lightbox

This might be a xen problem (triple fault leading to reset vector?)
rather than a coreboot issue, however the same Xen hypervisor config
works with grub2 on the drive loaded by seabios payload.

There is this old document that describes a "--vmm" option to kexec,
but that doesn't seem to exist in the current kexec tree:

https://xenbits.xen.org/docs/4.3-testing/misc/kexec_and_kdump.txt

[Zoran Stojsavljevic]

Hello Trammell,

Reading what you are trying. Have no idea what Qubes OS is... I apologize
for my ignorance, but I am (fortunately) willing to learn. New OSes, new
tricks, and this old dog (me) is still willing to learn new tricks (in
other words, also learning experience for me)?! ;-)

So, since you are coming to dracut after all, I am suspecting that
something with your initramfs is not quite right... It often happens to me
with Linux distros I am using for HW/SoC testing (95% Fedora distro).

In order to determine what is really wrong with your setup, I asked my best
friend for dozens of years: www.google.com, searching for pattern ->
*initramfs*-4.1.13-9.pvops.qubes.x86_64.img
https://www.google.de/search?espv=2&biw=1348&bih=614&q=initramfs-4.1.13-9.pvops.qubes.x86_64.img&spell=1&sa=X&ved=0ahUKEwjejvTtzY7OAhXI1SwKHd8TCekQvwUIGSgA

The following appeared:
https://groups.google.com/forum/#!msg/qubes-users/DYw6f2LpDA8/z4fDsy9CBwAJ

*And this cot my eye (from Marek Marczykowski-Górecki):*

*Probably some module required for your baremetal fedora is missing
in initramfs. First of all, remove "quiet" option to receive more
details. If that's about missing module, you can regenerate initramfs: 0.
Make sure you've copied modules from
Qubes (/lib/modules/4.1.13-9.pvops.qubes.x86_64) 1. Call "dracut -f
/initramfs-4.1.13-9.pvops.qubes.x86_64.img 4.1.13-9.pvops.qubes.x86_64 *

Could help (you might search other options with the provided search
pattern) ... :-)

Please, let us know.

Zoran

On Mon, Jul 25, 2016 at 6:05 AM, Trammell Hudson  wrote:
> I've successfully built a 4.6.4 Linux kernel payload for CoreBoot and
> flashed it onto the top 4 MB of the boot ROM on my Thinkpad x230.
> The runtime is a dynamic linked busybox with glibc, as well as a copy
> of the kexec binary from my Ubuntu laptop.
>
> kexec of a xen kernel (multiboot-x86) results in an immediate reboot
> into the coreboot payload:
>
> https://www.flickr.com/photos/osr/27913702804/lightbox
>
> kexec of a Linux kernel produces the expected result (panic in this
> case since I didn't pass any additional parameters to find a root fs):
>
> https://www.flickr.com/photos/osr/27914585663/lightbox
>
> This might be a xen problem (triple fault leading to reset vector?)
> rather than a coreboot issue, however the same Xen hypervisor config
> works with grub2 on the drive loaded by seabios payload.
>
> There is this old document that describes a "--vmm" option to kexec,
> but that doesn't seem to exist in the current kexec tree:
>
> https://xenbits.xen.org/docs/4.3-testing/misc/kexec_and_kdump.txt