On Thu, May 17, 2012 at 5:14 PM, H. Peter Anvin <hpa@xxxxxxxxx> wrote:
>
> The "logical" thing to do here seems to just use __s64, but I have no
> idea if that would suddenly break bigendian architectures...
Well, it would break the case of a 32-bit kernel and 32-bit user land.
So the thing is, that the __BITS_PER_LONG games there currently work
for two cases:
- when the kernel and user land agree on either 32-bit or 64-bit
(this is probably the common MIPS/PPC case)
- when the kernel is 64-bit, user-land is 32-bit, and we're little-endian
Using __s64 would actually break old 32-bit big-endian cases, because
the 32-bit binaries would continue to read the first 32 bits, while a
32-bit kernel would consider the first 32 bits to be the *high* bits
of the __s64 value.
That's why I think it's unfixable. It started out broken, and I
presume that 32-bit user land on a 64-bit MIPS/PPC thing either do not
work, or there's some compat crap (like special user-land headers)
fixing things up. Or they just don't use that buggered msqid64_ds
thing at all.
Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-arch" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
