On Mon, 2007-03-19 at 18:00 -0800, Zachary Amsden wrote:
> Rusty Russell wrote:
> > *This* was the reason that the current hand-coded calls only clobber %
> > eax. It was a compromise between native (no clobbers) and others (might
> > need a reg).
>
> I still don't think this was a good trade.
...
> Xen no longer needs such a register
Hmm, well, if VMI is happy, Xen is happy, and lguest is happy, then
perhaps we're better off with a cc-only clobber rule? Certainly makes
life simpler.
> > Now, since we decided to allow paravirt_ops operations to be normal C
> > (ie. the patching is optional and done late), we actually push and pop %
> > ecx and %edx. This makes the call site 10 bytes long, which is a nice
> > size for patching anyway (enough for a movl $0, <addr>, a-la lguest's
> > cli, or movw $0, %gs:<addr> if we supported SMP).
>
> You can do it in 11 bytes with no clobbers and normal C semantics by
> linking to a direct address instead of calling to an indirect, but then
> you need some gross fixup technology in paravirt_patch:
>
> if (call_addr == (void*)native_sti) {
> ...
> }
Well, I don't think we need such hacks: since we have to use handcoded
asm and mark the callsites anyway, marking what they're calling is
trivial.
The other idea from "btfixup" is that we can do the patching *much*
earlier, so we don't need the initial code to be valid at all if we
wanted to: we just need room to patch in a call insn. We could then
generate trampolines which do the necessary pushes & pops automatically
for backends which want to use C calling conventions.
Perhaps it's time for code and benchmarks?
Rusty.
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
Home