Re: [patch 0/3] [RFC] kernel/glibc mismatch of "readlink" syscall?

["Andries E. Brouwer" <Andries.Brouwer@cwi.nl>]

On Fri, Oct 24, 2008 at 05:53:25PM -0500, Michael Kerrisk wrote:

> Hi Daniel,
> 
> On Thu, Oct 23, 2008 at 9:50 AM, Daniel Gollub <dgollub@suse.de> wrote:
>>
>> found something which looks for me like a kernel/glibc syscall mismatch.
>>
>> The standard and the manpages and glibc have
>> ssize_t readlink(..., size_t bufsize);
>>
>> But the kernel has
>>
>> linux-2.6/include/linux/syscalls.h:
>> asmlinkage long sys_readlink(..., int bufsiz);
>>
>> All readlink prototypes, expect the one in the kernel, have an unsigned
>> buffer size.
> 
> I agree; the inconsistency is strange.

Hmm. I am inclined not to agree. There is no reason to expect
any particular relation between the kernel prototype of sys_foo
(if such a function exists)
and the user space prototype of the foo() C-library function.

The POSIX standard, man-pages, libc include files document the
userspace / libc interface, not the system call interface.
The system call interface is mostly undocumented.

The kernel prototype is chosen for kernel-internal or historical reasons.

Andries


sys_ppc32.c:
/* Note: it is necessary to treat bufsiz as an unsigned int ... */
asmlinkage long compat_sys_readlink(..., u32 bufsiz)