Re: [PATCH] x86-64: re-use kernel/syscall_table_32.S in ia32/ia32entry.S

[Arnd Bergmann <arnd@arndb.de>]

On Wednesday 27 April 2011 18:02:55 Jan Beulich wrote:
> >>> On 27.04.11 at 17:41, "H. Peter Anvin" <hpa@zytor.com> wrote:
> > On 04/27/2011 08:33 AM, Jan Beulich wrote:
> >> This requires a little bit of renaming, and a (much shorter than the
> >> original full table) set of #define-s for those table entries where
> >> native and compat mode entries differ.
> >> 
> >> Signed-off-by: Jan Beulich <jbeulich@novell.com>
> > 
> > Hm... not 100% sure I think this is an improvement, partly because it
> > creates "action at a distance" -- you now have to look in two separate
> > places to find out what actually happens with a system call -- and
> > because the i386 table is already a bit of an "odd man out".  I would
> > much rather like to see a setup where you have __SYSCALL() macros in
> > unistd_64.h-style and perhaps augment it with a 3-operand __SYSCALL()
> > macro for the case where compat and non-compat are different.
> > 
> > What do you think?
> 
> Yes, that's certainly an alternative.

How about following the same style as asm-generic/unistd.h? It's
originally derived from the x86 version, but has added support for
compat syscalls for the tile architecture.

	Arnd