Re: [Xen-devel] [PATCH] enable port accesses with (almost) full register context

[Keir Fraser <Keir.Fraser@xxxxxxxxxxxx>]

On 13/9/06 10:46, "Jan Beulich" <jbeulich@xxxxxxxxxx> wrote:

> It would, provided the above assumption about the need to modify the
> output value would never become true.

I hope it doesn't. :-) We'll cross this bridge if we come to it.

> Additionally, for 64-bits, I'm
> concerned about the potential need for using indirect calls here (as well
> as in the syscall trampolines): there's nothing keeping a user from making
> the Xen heap 2Gb or more in size.

Not much of a concern. Perhaps we should clamp the heap_size parameter to
2GB as a short-term fix for this issue. As you say, it can also affect the
syscall trampolines so users would soon notice if this was broken!

When we merge Xen and domain heaps on x86/64, we'll probably require Xen
allocations to come from a zone <= 2GB. Xen doesn't allocate much memory, so
that's not going to be a particularly serious constraint.

> Percentages of full-context relative to simply emulated i/o, without having
> changed the assembly file approach to the stub building one, yet (as per
> above issues):
> 
> PentiumIII (32-bit) with locking 67%
> PentiumIII (32-bit) without locking 84%
> Pentium4 (64-bit) with locking  86%
> Pentium4 (64-bit) without locking 89%

A little bit higher overhead than I'd hoped, but not terrible. Let's see how
it looks with the stub-building method, and then decide whether to bother
with the sysctl interface. Perhaps highly-optimised assembly code
save/restore routines will be required after all. :-)

 Cheers,
 Keir

> Revised patch (domctl->sysctl, naming) attached.



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