]> git.karo-electronics.de Git - linux-beck.git/commitdiff
locking/barriers: Suppress sparse warnings in lockless_dereference()
authorJohannes Berg <johannes.berg@intel.com>
Thu, 11 Aug 2016 09:50:22 +0000 (11:50 +0200)
committerIngo Molnar <mingo@kernel.org>
Thu, 18 Aug 2016 13:36:13 +0000 (15:36 +0200)
After Peter's commit:

  331b6d8c7afc ("locking/barriers: Validate lockless_dereference() is used on a pointer type")

... we get a lot of sparse warnings (one for every rcu_dereference, and more)
since the expression here is assigning to the wrong address space.

Instead of validating that 'p' is a pointer this way, instead make
it fail compilation when it's not by using sizeof(*(p)). This will
not cause any sparse warnings (tested, likely since the address
space is irrelevant for sizeof), and will fail compilation when
'p' isn't a pointer type.

Tested-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 331b6d8c7afc ("locking/barriers: Validate lockless_dereference() is used on a pointer type")
Link: http://lkml.kernel.org/r/1470909022-687-2-git-send-email-johannes@sipsolutions.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
include/linux/compiler.h

index 1bb95484272501bbc8d0603da489f56b4f87714a..436aa4e42221beb16edebd1d5e1bfb3a4fa6af26 100644 (file)
@@ -527,13 +527,13 @@ static __always_inline void __write_once_size(volatile void *p, void *res, int s
  * object's lifetime is managed by something other than RCU.  That
  * "something other" might be reference counting or simple immortality.
  *
- * The seemingly unused void * variable is to validate @p is indeed a pointer
- * type. All pointer types silently cast to void *.
+ * The seemingly unused size_t variable is to validate @p is indeed a pointer
+ * type by making sure it can be dereferenced.
  */
 #define lockless_dereference(p) \
 ({ \
        typeof(p) _________p1 = READ_ONCE(p); \
-       __maybe_unused const void * const _________p2 = _________p1; \
+       size_t __maybe_unused __size_of_ptr = sizeof(*(p)); \
        smp_read_barrier_depends(); /* Dependency order vs. p above. */ \
        (_________p1); \
 })