[ On Sunday, October 14, 2001 at 12:28:13 (-0400), Bill Sommerfeld wrote: ]
> Subject: Re: wrap up of pipe(2) 
>
> One possible way to fix this would involve changing pipe(2) so it's a
> "normal" syscall which *may* return EFAULT; other two-return-value
> syscalls would also need to be changed.

I think this would be a good thing to do, even if it didn't save a few
instructions -- consistency in the kernel interface has its own merits.

In terms of performance though presumably pipe(2) is faster because of
the way it's implemented now.  However I doubt it's so frequently called
in most systems that implementing it in a consistent way would result in
more overhead than would be saved in all the other syscalls.  I don't
know how this would mix with other two-return-value syscalls.  I'm not
even sure I know which ones they are....  [syscall(2) only mentions pipe]

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>     <woods@robohack.ca>
Planix, Inc. <woods@planix.com>;   Secrets of the Weird <woods@weird.com>