[systemd-devel] [PATCH] systemd-detect-s390-virt: add virtualization detection on s390x

[Lennart Poettering]

On Mon, 07.07.14 17:32, Thomas Blume (Thomas.Blume at suse.com) wrote:

> 
> On Mon, 7 Jul 2014, Lennart Poettering wrote:
> 
> >>IMHO the main difference is the level of maturity.
> >>z/VM is about 30 years old and has a huge amount of tools for everything you
> >>could imagine. KVM is relatively new and under heavy development.
> >>Furthermore, KVM is bound to the linux kernel, while z/VM is not.
> >>Finally, KVM could theoretically run inside z/VM (thought it doesn't make sense
> >>running KVM on an already virtualized CPU) but not vice-versa.
> >>
> >>Kay, my colleague Ihno told me that you were working in the s390 department at SUSE.
> >>Any opinion about the use of distinguishing z/VM from KVM under s390?
> >
> >Well, obviously, we should distuingish kvm from some s390-specific
> >virtualization. I was mostly referring to your original's patch
> >distinction between "PR/SM" and "z/VM". What is that about?
> >
> 
> Ah, sorry for the misunderstanding.
> PR/SM is the primary hypervisor that runs on the physical hardware, whereas
> z/VM can only run on top of PR/SM, but not below.
> Under PR/SM, the system resources can be partitioned first via LPAR and then
> by running a z/VM on each LPAR.
> However, apart from determining the level of virtualization, I don't see much
> practical relevance in making a distinction.

Ah! OK!

Then I think a patch that simply returns a generic "s390" id for all
s390 + s390x systems would be a good idea. i.e. when we compile for
s390/s390x we should just return that string unconditionally, without
checking anything else.

Please prepare a patch for that, and add the new type to the
systemd.unit(5) man page, where ConditionVirtualization= is documented. 

Thanks,

Lennart

-- 
Lennart Poettering, Red Hat