Re: [PATCH] new ioctl type checking causes gcc warning

[Arnd Bergmann <arnd@arndb.de>]

On Saturday 13 September 2003 02:31, Kevin P. Fleming wrote:
> Does that mean that
> these two lines:
>
> #define BLKGETSIZE64    _IOR(0x12,114,sizeof(__uint64_t))
> #define BLKGETSIZE64    _IOR(0x12,114,__uint64_t)
>
> actually produce different ioctl numbers?

Exactly. On 32 bit systems, the former is 0x80041272, the
latter is 0x80081272. On 64 bit systems, they are both
0x80081272.

>                                            If so, then I don't
> understand how the kernel can continue to offer the old/invalid
> interface when the new _IOR macro won't accept the first version any
> longer.

Inside the kernel, the first definition has to be changed to
something like:

#define BLKGETSIZE64    _IOR(0x12,114,size_t) /* broken: actually __u64 */
or
#define BLKGETSIZE64    _IOR_BAD(0x12,114,sizeof(__uint64_t)) /* broken */

in order to get a definition that will pass the check and
generate the well-known number.