Re: [PATCH] lockdep: make lockdep_assert_held() not have a return value

[Andrew Morton <akpm@linux-foundation.org>]

On Fri, 18 Jan 2013 23:31:43 +0100
Johannes Berg <johannes@sipsolutions.net> wrote:

> From: Johannes Berg <johannes.berg@intel.com>
> 
> I recently made the mistake of writing:
> 
> foo = lockdep_dereference_protected(..., lockdep_assert_held(...));
> 
> which is clearly bogus. If lockdep is disabled in the
> config this would cause a compile failure, if it is
> enabled then it compiles and causes a puzzling warning
> about dereferencing without the correct protection.
> 
> Wrap the macro in "do { ... } while (0)" to also fail
> compile for this when lockdep is enabled.
> 
> ...
>
> --- a/include/linux/lockdep.h
> +++ b/include/linux/lockdep.h
> @@ -359,7 +359,9 @@ extern void lockdep_trace_alloc(gfp_t mask);
>  
>  #define lockdep_depth(tsk)	(debug_locks ? (tsk)->lockdep_depth : 0)
>  
> -#define lockdep_assert_held(l)	WARN_ON(debug_locks && !lockdep_is_held(l))
> +#define lockdep_assert_held(l)	do {				\
> +		WARN_ON(debug_locks && !lockdep_is_held(l));	\
> +	} while (0)
>  
>  #define lockdep_recursing(tsk)	((tsk)->lockdep_recursion)

A sane fix would be to convert lockdep_assert_held() into a static
inline void C function.  But, alas, much of the lockdep API is designed
to work on "any type which has a field called dep_map", which was
rather a silly stunt IMO.

lockdep_depth() and lockdep_assert_held() may or may not evaluate their
argument, and this is runtime controllable.  whee.