Re: [XenPPC] Xencomm on ia64 - The patches

[Hollis Blanchard <hollisb@xxxxxxxxxx>]

On Fri, 2006-08-18 at 18:02 +0200, Tristan Gingold wrote:
> these are the patches to port xencomm on ia64.  This is just an RFC.  I have 
> to clean-up the code so that it could be shared between ia64 and ppc.

Great! Your patches were a little noisy though; they had some unrelated
and whitespace changes.

> I have added the xencomm_inline feature which makes (IMHO) hcall.c more 
> readable and safer (particularly at boot time).  Wether or not xencomm_mini 
> should be deprecated is a question for ppc people (xencomm_mini is buggy: the 
> area should be reserved as an array of struct xencomm_mini for alignment).

Good point.

> I have also fixed a few bugs in guest_access.h

Do you think you could split those out as a diff to
xen/include/asm-powerpc/guest_access.h, to be applied before we move it?

> Many files should be shared between ia64 and ppc.
> In xen:
> * guest_access.h should be the same.  I will propose to put as 
> include/xen/xencomm_access.h

And then xen/include/asm-{powerpc,ia64}/guest_access.h would just
#include <xen/xencomm_access.h>? Sounds good to me.

> * usercopy.c sould be shared too.  The paddr_to_maddr() function should be 
> defined in arch files as xencomm_paddr_to_maddr.  I will propose to put this 
> file into xen/common/xencomm.c

Agreed.

> In linux:
> * drivers/xen/core/xencomm.c should be shared.  We just need to agree on
> xen_pa().  On linux-ia64, __pa() doesn't work for this use.  asm/xencomm.h 
> should define this macro/function.

We got rid of __kern_paddr entirely, so __pa() shouldn't be a problem.
If you check the current PPC tree (specifically changeset 0b82e0cba7d3),
you'll find we always use the follow_vma() routine now. Please make sure
you pull down that changeset, as I'd hate to lose changes because you
were working from something older.

> * hcall.c should be shared because it is the biggest and the most boring 
> file.  
> This won't be that easy, but it will try to made a proposal soon.

Sure.

-- 
Hollis Blanchard
IBM Linux Technology Center


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