svn commit: r288628 - head/sys/kern

[Mark Johnston]

On Sat, Oct 3, 2015 at 2:33 PM, Jilles Tjoelker <jilles at stack.nl> wrote:
> On Sat, Oct 03, 2015 at 07:37:42PM +0000, Mark Johnston wrote:
>> Author: markj
>> Date: Sat Oct  3 19:37:41 2015
>> New Revision: 288628
>> URL: https://svnweb.freebsd.org/changeset/base/288628
>
>> Log:
>>   The return value of posix_fadvise(2) is just an error status, so
>>   sys_posix_fadvise() should simply return the errno (or 0) to syscallenter()
>>   rather than setting a return value.
>
>>   MFC after:  1 week
>
>> Modified:
>>   head/sys/kern/vfs_syscalls.c
>>
>> Modified: head/sys/kern/vfs_syscalls.c
>> ==============================================================================
>> --- head/sys/kern/vfs_syscalls.c      Sat Oct  3 19:27:52 2015        (r288627)
>> +++ head/sys/kern/vfs_syscalls.c      Sat Oct  3 19:37:41 2015        (r288628)
>> @@ -4663,7 +4663,6 @@ int
>>  sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
>>  {
>>
>> -     td->td_retval[0] = kern_posix_fadvise(td, uap->fd, uap->offset,
>> -         uap->len, uap->advice);
>> -     return (0);
>> +     return (kern_posix_fadvise(td, uap->fd, uap->offset, uap->len,
>> +         uap->advice));
>>  }
>
> This change makes the code match the man page, but in fact, the code was
> right and the man page is wrong. Per POSIX, posix_fadvise() shall return
> 0 on success and an error number on failure, and need not modify errno.

Oops, thank you. This is fixed in r288640; I updated the posix_fadvise
man page to use the same wording as for posix_fallocate.

>
> Also, this kind of ABI change in general is almost always a bad idea
> when the function is already part of a stable branch.

Noted, thanks.