Re: does i386 still work?

[Brad Spencer <brad%anduin.eldar.org@localhost>]

Greg Troxel <gdt%lexort.com@localhost> writes:

> Brad Spencer <brad%anduin.eldar.org@localhost> writes:
>
>> Greg Troxel <gdt%lexort.com@localhost> writes:
>>
>>> I am trying to use xen again after a long absence.
>>>
>>> dom0: netbsd-10 amd64, xen 4.15.4, netbsd/pkgsrc both up to date (WORKS FINE)
>>>
>>> domU: PV, netbsd-9 amd64, 2 cpus, vif/bridge, disk/zvol (WORKS)
>>> domU: PV, netbsd-9 i386 PAE, 1 or 2 cpus, vif/bridge, disk/zvol (DOES NOT BOOT)
>>>
>>>
>>> Do people think i386 guests still work?
>>
>> Works fine for me....  I run several 9.x i386 DOMU using the
>> XEN3PAE_DOMU config or a derivation of that.

Good...  glad to hear that...

> Thanks.  After disabling autoballoon (for dom0), checking things over,
> rebooting the dom0, and trying again, it worked normally.
>
>> All of my i386 DOMUs are PVSHIM style guests.  In theory, a full PVH
>> guest will boot (using something newer then 9.x), but the i386 guest
>> (unlike the amd64 guest) in PVH mode will hang itself after a certain
>> amount of time if there is any amount of disk activity (unpacking the
>> sets did it for me).
>
> Is there a PR?

kern/57199

> Is there a pointer to how to run PVSHIM?   Is this the same as "stub
> domain", or different?  I'd like to get this into the HOWTO.   I'm open
> to you explaining in private mail and me trying to figure it out and
> write it up.

This gets put into your config file, everything else should be allowed
to remain the same as a PV guest config (BTW - your processor must be
able to support PVH guests for this to work):

type="pvh"
pvshim=1

I think that this uses the stub domain concept, but as far as I can
tell, the stub is actually another instance of Xen running in PVH mode
which runs the PV guest under that (Xen kernel running as a guest to
Xen).  This grants a degree of isolation to the PV guest which is a good
thing to be doing, but doesn't appear to impact performance much or at
all.

>> Double check your config..  I have hung a DOMU pretty much exactly in
>> that location when 1) The back stored disk or file image was in use on
>> the DOM0 (like I vnconfig'ed the backstore file and then mounted it on
>> the DOM0 and then tried to boot the DOMU using the same file) 2) I used
>> the character device from a LVM when the block device was required.
>
> I might have mounted it and unmounted; I wonder if that's trouble.
> Using vnconfig failed on me -- zvols seem pesky -- but maybe it left bad
> state.

I can't say that is impossible.  A zvol is suppose to be, more or less
just a device, so vnconfig should not be involved.  I don't ZFS with
Xen, but use it in DOMUs all of the time.

> I did add to the HOWTO that you need the block device.

That is probaby a good idea to mention.




-- 
Brad Spencer - brad%anduin.eldar.org@localhost - KC8VKS - http://anduin.eldar.org