Re: Running NetBSD as PV under XCP-NG v.8.0...

[Stephen Borrill <netbsd%precedence.co.uk@localhost>]

On Sat, 17 Aug 2019, Chavdar Ivanov wrote:
> Under XCP-NG 7.6 and earlier, as well as under Citrx XenServer, I used
> to run a few NeetBSD systems in PV mode, just by setting PV-kernel
> parameter to a locally copied XEN3_DOMU kernel and clearing
> HVM-boot-policy parameter.
>
> I upgraded my XCP-NG host today to the recently released 8.0 version;
> this no longer works now. Perhaps I am missing some other parameter,
> now I am getting:
> ...
> "Failed","Starting VM 'n999'
> Internal error: xenopsd internal error: VM =
> 18bf7b9e-f23c-61ac-6456-14dc6a046f28; domid = 32; Bootloader.Bad_error
> No valid Btrfs found on
> /dev/sm/backend/e7367c91-e374-7f7b-cc05-f22b4137ef75/93d140ed-0d84-45c6-8d27-902a22463f11
> No valid Btrfs found on
> /dev/sm/backend/e7367c91-e374-7f7b-cc05-f22b4137ef75/93d140ed-0d84-45c6-8d27-902a22463f11

Looks like you have something in your filesystem that unintentially looks 
like btrfs. Can you check what your PV bootloader is set to? The following 
should tell you?

xe vm-param-get param-name=PV-bootloader uuid=18bf7b9e-f23c-61ac-6456-14dc6a046f28

Having said the next lines suggest that it is pygrub, which is correct.

You can probably read through pygrub to work out what signature it is 
looking for - I had similar problems when I had 0x55aa in a location that 
led the bootloader to believe it was a MBR disk.

> Superblock bytenr is larger than device size
> Traceback (most recent call last):
>  File "/usr/bin/pygrub", line 990, in <module>
>    fs = fsimage.open(file, part_offs[0], bootfsoptions)
> IOError: [Errno 95] Operation not supported
>
> Time: 00:00:14","xcp-ng-ci","Aug 17, 2019 11:38 PM"
>
> I am wondering whether to revert to 7.6 or just recreate by NetBSD
> hosts  as HVM guests...

For the record, I've had no issues at all with NetBSD PV on XenServer 8.0.

--
Stephen