[Devel] Re: [PATCH 1/5] Use existing macros for distinguishing mandatory locks

[Andrew Morton]

On Wed, 12 Sep 2007 15:17:58 +0400
Pavel Emelyanov <xemul at openvz.org> wrote:

> The combination of S_ISGID bit set and S_IXGRP bit unset is 
> used to mark the inode as "mandatory lockable" and there's a 
> macro for this check called MANDATORY_LOCK(inode). However, 
> fs/locks.c and some filesystems still perform the explicit 
> i_mode checking. 
> 
> Switch the fs/locks.c to macro making the code shorter and 
> more readable.
> 
> The __MANDATORY_LOCK() macro is to be used in places where
> the IS_MANDLOCK() for superblock is already known to be true.
> 

If we're going to churn this code then it would be better to switch from
ugly-upper-case-macro to nice-lower-case-C-function while we're doing it,
please.

> 
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index 291d40b..035ffda 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -1488,8 +1488,8 @@ extern int locks_mandatory_area(int, str
>   * Candidates for mandatory locking have the setgid bit set
>   * but no group execute bit -  an otherwise meaningless combination.
>   */
> -#define MANDATORY_LOCK(inode) \
> -	(IS_MANDLOCK(inode) && ((inode)->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID)
> +#define __MANDATORY_LOCK(ino) (((ino)->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID)
> +#define MANDATORY_LOCK(inode) (IS_MANDLOCK(inode) && __MANDATORY_LOCK(inode))

Especially as the macro is a buggy one which references its argument more
than once.