[PATCH] Fix memory barrier docs wrt atomic ops

Fix the memory barrier documentation to attempt to describe atomic ops
correctly.

atomic_t ops that return a value _do_ imply smp_mb() either side, and so
don't actually require smp_mb__*_atomic_*() special barriers.

Also explains why special barriers exist in addition to normal barriers.

Further fix the memory barrier documents to portray bitwise operation
memory barrier effects correctly following Nick Piggin's comments.

It makes the point that any atomic op that both modifies some state in
memory and returns information on that state implies memory barriers on
both sides.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

diff --git a/Documentation/memory-barriers.txt b/Documentation/memory-barriers.txt
index f855031..528d52f 100644 (file)
--- a/Documentation/memory-barriers.txt
+++ b/Documentation/memory-barriers.txt
@@ -829,8 +829,8 @@ There are some more advanced barrier functions:
  (*) smp_mb__after_atomic_inc();
 
      These are for use with atomic add, subtract, increment and decrement
-     functions, especially when used for reference counting.  These functions
-     do not imply memory barriers.
+     functions that don't return a value, especially when used for reference
+     counting.  These functions do not imply memory barriers.
 
      As an example, consider a piece of code that marks an object as being dead
      and then decrements the object's reference count:
@@ -1263,15 +1263,17 @@ else.
 ATOMIC OPERATIONS
 -----------------
 
-Though they are technically interprocessor interaction considerations, atomic
-operations are noted specially as they do _not_ generally imply memory
-barriers.  The possible offenders include:
+Whilst they are technically interprocessor interaction considerations, atomic
+operations are noted specially as some of them imply full memory barriers and
+some don't, but they're very heavily relied on as a group throughout the
+kernel.
+
+Any atomic operation that modifies some state in memory and returns information
+about the state (old or new) implies an SMP-conditional general memory barrier
+(smp_mb()) on each side of the actual operation.  These include:
 
        xchg();
        cmpxchg();
-       test_and_set_bit();
-       test_and_clear_bit();
-       test_and_change_bit();
        atomic_cmpxchg();
        atomic_inc_return();
        atomic_dec_return();
@@ -1282,21 +1284,31 @@ barriers.  The possible offenders include:
        atomic_sub_and_test();
        atomic_add_negative();
        atomic_add_unless();
+       test_and_set_bit();
+       test_and_clear_bit();
+       test_and_change_bit();
+
+These are used for such things as implementing LOCK-class and UNLOCK-class
+operations and adjusting reference counters towards object destruction, and as
+such the implicit memory barrier effects are necessary.
 
-These may be used for such things as implementing LOCK operations or controlling
-the lifetime of objects by decreasing their reference counts.  In such cases
-they need preceding memory barriers.
 
-The following may also be possible offenders as they may be used as UNLOCK
-operations.
+The following operation are potential problems as they do _not_ imply memory
+barriers, but might be used for implementing such things as UNLOCK-class
+operations:
 
+       atomic_set();
        set_bit();
        clear_bit();
        change_bit();
-       atomic_set();
+
+With these the appropriate explicit memory barrier should be used if necessary
+(smp_mb__before_clear_bit() for instance).
 
 
-The following are a little tricky:
+The following also do _not_ imply memory barriers, and so may require explicit
+memory barriers under some circumstances (smp_mb__before_atomic_dec() for
+instance)):
 
        atomic_add();
        atomic_sub();
@@ -1317,10 +1329,12 @@ specific order.
 
 
 Basically, each usage case has to be carefully considered as to whether memory
-barriers are needed or not.  The simplest rule is probably: if the atomic
-operation is protected by a lock, then it does not require a barrier unless
-there's another operation within the critical section with respect to which an
-ordering must be maintained.
+barriers are needed or not.
+
+[!] Note that special memory barrier primitives are available for these
+situations because on some CPUs the atomic instructions used imply full memory
+barriers, and so barrier instructions are superfluous in conjunction with them,
+and in such cases the special barrier primitives will be no-ops.
 
 See Documentation/atomic_ops.txt for more information.