cond_syscall() and new ABI.

[David Woodhouse]

On Tue, 2004-10-12 at 20:43 +0200, Arnd Bergmann wrote:
> A better solution IMHO would be to include the right headers from sys.c
> and have
> 
> #define cond_syscall(x) typeof(x) (x) __attribute__((weak,alias("sys_ni_syscall")));

That's true in theory, yes -- not that I can see any way that having the
'correct' prototype will actually make a difference in practice.

> Also, someone should try to find out which toolchains don't support this
> and if anybody is still using those. One issue seems to be the one from
> http://seclists.org/lists/linux-kernel/2004/Jan/2474.html, but I'm not
> sure if that is the problem that the comment refers to.

That happens with both the current inline asm method, and with the
'alias' method which translates to basically the same asm output from
gcc, but without the ifdefs.

-- 
dwmw2