Re: [PATCH] compat: fix compat_sys_openat and friends

[Ralf Baechle <ralf@linux-mips.org>]

On Wed, Feb 01, 2006 at 09:36:40PM -0800, Linus Torvalds wrote:

> > Most of the 64 bit architectures will zero extend the first argument to
> > compat_sys_{openat,newfstatat,futimesat} which will fail if the 32 bit
> > syscall was passed AT_FDCWD (which is a small negative number).  Declare
> > the first argument to be an unsigned int which will force the correct
> > sign extension when the internal functions are called in each case.
> 
> Umm.
> 
> Wouldn't it be _much_ better to declare the argument as a "long", since 
> some architectures (alpha, for example) may assume that 32-bit arguments 
> have been _sign_extended, not zero-extended.

> Then, when the "compat_sys_xxxx()" function passes the "long" down to the 
> _real_ function (which takes an "int"), those architectures (and only 
> those architectures) that actually have assumptions about high bits will 
> have the compiler automatically do the right zero- or sign-extensions at 
> that call-site.

MIPS is one of those architectures where variables are always held
sign-extended to the full 32-bit or 64-bit size of the register.  The
signed-ness of the C data type doesn't matter at all here so Stephen's
change is a nop for MIPS.

  Ralf