Re: 4.15.0-rc3 can't create domain when dom0 uses newer kernel

[Tamas K Lengyel <tamas@xxxxxxxxxxxxx>]

On Mon, Mar 15, 2021 at 7:26 PM Marek Marczykowski-Górecki
<marmarek@xxxxxxxxxxxxxxxxxxxxxx> wrote:
>
> On Mon, Mar 15, 2021 at 04:18:42PM -0400, Tamas K Lengyel wrote:
> > Hi all,
> > I've ran into a pretty strange issue. I have a freshly installed 4.15.0-rc3
> > system and everything works fine with the stock Ubuntu 20.04
> > vmlinuz-5.4.0-66-generic as the dom0 kernel - except vmtrace, which needs a
> > newer kernel, that's expected. Now if I boot dom0 with
> > vmlinuz-5.8.0-44-generic or vmlinuz-5.10.0-1016-oem I can't start any
> > domains because of some strange disk error:
> >
> > root@t2:/shared/cfg# xl create ubuntu-20.04.cfg
> > Parsing config from ubuntu-20.04.cfg
> > libxl: error: libxl_device.c:1109:device_backend_callback: Domain 1:unable
> > to add device with path /local/domain/0/backend/vbd/1/51712
> > libxl: error: libxl_create.c:1613:domcreate_launch_dm: Domain 1:unable to
> > add disk devices
> > libxl: error: libxl_device.c:1109:device_backend_callback: Domain 1:unable
> > to remove device with path /local/domain/0/backend/vbd/1/51712
> > libxl: error: libxl_domain.c:1529:devices_destroy_cb: Domain
> > 1:libxl__devices_destroy failed
> > libxl: error: libxl_domain.c:1182:libxl__destroy_domid: Domain
> > 1:Non-existant domain
> > libxl: error: libxl_domain.c:1136:domain_destroy_callback: Domain 1:Unable
> > to destroy guest
> > libxl: error: libxl_domain.c:1063:domain_destroy_cb: Domain 1:Destruction
> > of domain failed
> >
> > This is the disk config:
> >
> > disk = ['phy:/dev/t2vg/ubuntu-20.04,xvda,w']
> >
> > It's a standard LVM volume that is active, lvscan shows:
> >
> >   ACTIVE            '/dev/t2vg/ubuntu-20.04' [40.00 GiB] inherit
> >
> > I checked all the logs and I don't see anything relevant anywhere. If I
> > reboot the system with the older kernel everything works as expected as far
> > as creating the same domain with the same config.
> >
> > Anyone has any idea what might be going on here?
>
> Make sure your kernel has this patch:
> https://lore.kernel.org/xen-devel/4c9af052a6e0f6485d1de43f2c38b1461996db99.camel@xxxxxxxxxxxxx/
> I'm not sure about the 5.8.x, but for 5.10 it is in >= 5.10.13.

Thanks, after bumping the version to
https://kernel.ubuntu.com/~kernel-ppa/mainline/v5.11.6 it all works
now. There should really be some better checks in the toolstack to
verify that the underlying kernel will work or not and give the user
an informative warning if not. The libxl errors are really not
indicative at all for what the underlying issue really is.

Tamas