Re: [Xen-ia64-devel] [PATCH] initial cleanup of ivt.S

[Tristan Gingold <Tristan.Gingold@xxxxxxxx>]

Le Vendredi 11 Août 2006 16:27, Al Stone a écrit :
> On Fri, 2006-08-11 at 11:51 +0200, Tristan Gingold wrote:
> > Le Vendredi 11 Août 2006 01:12, Al Stone a écrit :
[...]
> > IMHO, ivt.S is too far from linux ivt.S
> > A clean-up pass should be done: remove all #ifndef XEN code.
> > But you reformatting work is also a good thing.
>
> Just so I understand clearly -- the Xen ivt.S is for the
> hypervisor only, correct?  I assume so, so I agree that
> the Xen version and Linux are -- and should be -- quite
> different.
Yes.

>  What I want to do is make sure we don't lose
> any improvements made in upstream; e.g., the syscall
> code had changed (and improved) quite a bit.
But we don't use the syscall code, do we ?
However I think we should synchronize with entry.S and minstate.h.

> Which reminds me: there were some #if 0 blocks; do you
> think we still need or want any of those?  Most of those
> looked like they could go away.
Yes, most of them could be removed.

> > >    4) Do all of the TODOs in the file.
> > >
> > >    5) Handle the FIXMEs.
> > >
> > >    6) Go through an optimization pass.
> > >
> > > In all of this patch, I only changed one line of assembly; there
> > > was a 'mov r31=pr;' in entry 23 (daccess_rights) that I changed
> > > to 'mov r31=pr;;' which is most likely what was meant.  No other
> > > instructions were changed.  Hopefully, I didn't harm any other
> > > white space formatting.
> >
> > From what I read you could remove the ';;' stop bit.
>
> Probably true.  As I step through the instructions, I'll
> get rid of as many of the stop bits as I can (and I'll keep
> a special eye on this one :).
Ok.
You may also look on objdump output to try to compact the code.  That's a game 
I like!

Tristan.

_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel