On Tue, 2006-06-27 at 11:39 +0100, Keir Fraser wrote:
> On 23 Jun 2006, at 22:19, Hollis Blanchard wrote:
>
> > These patches are similar to changeset 10747 (which added
> > vcpu_mark_events_pending), in that they allow PPC to work around some
> > unusually-sized atomic operations, this time in the grant table driver.
> >
> > Keir, do you think it's a good idea to make a wiki page to list where
> > the current ABI is causing problems? That way, when the ABI opens up
> > again (Xen 4?), we'll have a list of places to fix?
> >
> > Anyways, I expect you'll have alternate name/interface suggestions in
> > the patches, so they're just RFC for now.
>
> You should be able to hide all this behind the generic atomic
> operations without slowing down sufficiently aligned operations at all.
>
> e.g. something like:
> define clear_bit(p,i) ((alignof(p)>=alignof(long)) ? clear_bit(p,i) :
> clear_bit_unaligned(p,i))
In the couple cases so far, we know that even though the field is only
one or two bytes, it's actually safe to do a four-byte load/store to it
because the containing structure is large enough.
I'm not really comfortable with making that a blanket assumption. I'd
really like to know exactly what we're overwriting when performing these
hacks^Woperations.
--
Hollis Blanchard
IBM Linux Technology Center
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
Home