This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
Home Products Support Community News


[Xen-devel] Re: [rfc] [patch] "frame number" size in hypercall ABI

On Apr 19, 2006, at 2:24 AM, Keir Fraser wrote:

On 18 Apr 2006, at 20:17, Hollis Blanchard wrote:

I would like to avoid introducing new dom0_ops. If an invasive change
is only changing longs into an equivalent typedef'ed long (on x86/ia64)
then it's not hard to reason that it's correct. If it also causes
changes in users of libxc then well, that's a shame, but it's not
fixed by introducing a new dom0_op unless you duplicate every end user
for the ppc architecture too. :-)

OK, is this what you're looking for? Compile-tested only (on x86-32; I
haven't yet provided frameno_t in other arch-*.h).

I also haven't converted Xen itself to use frameno_t except where
absolutely necessary. It isn't required to solve my immediate problem...

I think the patch looks okay in principle, except I notice a couple of 'framno_t' misspellings. I think 'pfn_t' would be more in keeping with the Xen naming scheme for arbitrary page frame numbers, but perhaps that name is a bit unnecessarily cryptic. What is the immediate problem you're needing to solve?

Please see my recent mail titled "32/64-bit hypercall interface revisited"; I summarized the issues there. I have a 32-bit libxc communicating with 64-bit Xen. "unsigned long" is 32 bits in libxc, so I need to force that to 64 bits to match Xen's "unsigned long".

(If we require a 64-bit libxc in this case, that will also require a 64-bit python, and 64-bit versions of all the python modules. No distribution, other than maybe Gentoo, currently does this, and it would be a big PITA for them to do so.)

The earlier patch (in the mail mentioned above) deals with handles in the ABI. This patch deals with the PFN list. I recall seeing a few other unsigned longs in the hypercall ABI but haven't gotten to those yet.

Are you waiting to apply all these patches at once, or would it be ok to apply each individually?

Hollis Blanchard
IBM Linux Technology Center

Xen-devel mailing list