Re: [Xen-devel] [PATCH v2 06/10] arm: compile libxenguest

[Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>]

On Mon, 13 Feb 2012, Ian Campbell wrote:
> On Fri, 2012-02-10 at 12:02 +0000, Stefano Stabellini wrote:
> > Introduce an empty implementation of the arch specific ARM functions in
> > xc_dom_arm.c.
> 
> This bit:
> Acked-by: Ian Campbell <ian.campbell@xxxxxxxxxx>
> 
> > Also provide empty implementations of xc_domain_save, xc_domain_restore
> > and xc_hvm_build_target_mem when CONFIG_HVM or CONFIG_MIGRATE are not
> > set.
> 
> Less sure about this bit.
> 
> Is the content of xc_hvm_build x86 specific? ia64 seems to have it's own
> version in tools/libxc/ia64/xc_ia64_hvm_build.c. (I guess ia64 does not
> define CONFG_HVM, despite apparently supporting HVM? Otherwise I don't
> see how the tools/libxc/Makefile stuff could work)
> 
> Perhaps xc_hvm_build should go to xc_x86_hvm_build.c and
> xc_x86_arm_build.c should have stubs?

Yes, good idea.

> Also xc_hvm_build.c has public functions other than the one you have
> stubbed out in it. Should stub them too?

Yes, they should.


> The migrate case seems more like a temporary band-aid since we will
> eventually want to support that on ARM -- I guess there's no quick hack
> to make it compile since we'd need to define all sorts of things we
> aren't ready to define yet?

I don't think quick hacks are possible here: give a look at
vcpu_guest_context_any_t: it contains two additional fields (x32 and
x64) that are x86_32/64 specific and they are referenced through all the
save/restore code without any ifdefs.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel