tools/memory-model: Finish the removal of rb-dep, smp_read_barrier_depends(), and...

Commit:

  bf28ae562744 ("tools/memory-model: Remove rb-dep, smp_read_barrier_depends, and lockless_dereference")

was merged too early, while it was still in RFC form.  This patch adds in
the missing pieces.

Akira pointed out some typos in the original patch, and he noted that
cheatsheet.txt should indicate that READ_ONCE() now implies an address
dependency.  Andrea suggested documenting the relationship betwwen
unsuccessful RMW operations and address dependencies.

Andrea pointed out that the macro for rcu_dereference() in linux.def
should now use the "once" annotation instead of "deref".  He also
suggested that the comments should mention commit:

  5a8897cc7631 ("locking/atomics/alpha: Add smp_read_barrier_depends() to _release()/_relaxed() atomics")

... as an important precursor, and he contributed commit:

  cb13b424e986 ("locking/xchg/alpha: Add unconditional memory barrier to cmpxchg()")

which is another prerequisite.

Suggested-by: Akira Yokosawa <akiyks@gmail.com>
Suggested-by: Andrea Parri <parri.andrea@gmail.com>
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
[ Fixed read_read_lock() typo reported by Akira. ]
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Andrea Parri <parri.andrea@gmail.com>
Acked-by: Akira Yokosawa <akiyks@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: boqun.feng@gmail.com
Cc: dhowells@redhat.com
Cc: j.alglave@ucl.ac.uk
Cc: linux-arch@vger.kernel.org
Cc: luc.maranget@inria.fr
Cc: npiggin@gmail.com
Cc: will.deacon@arm.com
Fixes: bf28ae562744 ("tools/memory-model: Remove rb-dep, smp_read_barrier_depends, and lockless_dereference")
Link: http://lkml.kernel.org/r/1520443660-16858-4-git-send-email-paulmck@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

diff --git a/tools/memory-model/Documentation/cheatsheet.txt b/tools/memory-model/Documentation/cheatsheet.txt
index 04e458acd6d4dfbd8aef24025ceb484fb0297a4d..956b1ae4aafbb2ea622d1aebfe94f0c7001a960e 100644 (file)
--- a/tools/memory-model/Documentation/cheatsheet.txt
+++ b/tools/memory-model/Documentation/cheatsheet.txt
@@ -1,11 +1,11 @@
                                   Prior Operation     Subsequent Operation
                                   ---------------  ---------------------------
                                C  Self  R  W  RWM  Self  R  W  DR  DW  RMW  SV
-                              __  ----  -  -  ---  ----  -  -  --  --  ---  --
+                              --  ----  -  -  ---  ----  -  -  --  --  ---  --
 
 Store, e.g., WRITE_ONCE()            Y                                       Y
-Load, e.g., READ_ONCE()              Y                              Y        Y
-Unsuccessful RMW operation           Y                              Y        Y
+Load, e.g., READ_ONCE()              Y                          Y   Y        Y
+Unsuccessful RMW operation           Y                          Y   Y        Y
 rcu_dereference()                    Y                          Y   Y        Y
 Successful *_acquire()               R                   Y  Y   Y   Y    Y   Y
 Successful *_release()         C        Y  Y    Y     W                      Y

diff --git a/tools/memory-model/Documentation/explanation.txt b/tools/memory-model/Documentation/explanation.txt
index dae8b8cb2ad36b269ba1af582215b946172b3156..a727c82bd434f51aded17198b9abf08d0de622a2 100644 (file)
--- a/tools/memory-model/Documentation/explanation.txt
+++ b/tools/memory-model/Documentation/explanation.txt
@@ -826,7 +826,7 @@ A-cumulative; they only affect the propagation of stores that are
 executed on C before the fence (i.e., those which precede the fence in
 program order).
 
-read_lock(), rcu_read_unlock(), and synchronize_rcu() fences have
+rcu_read_lock(), rcu_read_unlock(), and synchronize_rcu() fences have
 other properties which we discuss later.
 
 
@@ -1138,7 +1138,7 @@ final effect is that even though the two loads really are executed in
 program order, it appears that they aren't.
 
 This could not have happened if the local cache had processed the
-incoming stores in FIFO order.  In constrast, other architectures
+incoming stores in FIFO order.  By contrast, other architectures
 maintain at least the appearance of FIFO order.
 
 In practice, this difficulty is solved by inserting a special fence

diff --git a/tools/memory-model/linux-kernel.def b/tools/memory-model/linux-kernel.def
index 5dfb9c7f34621bd76db81d48151a542928b9d179..397e4e67e8c849e98726d5a4f3a716264dd02ed6 100644 (file)
--- a/tools/memory-model/linux-kernel.def
+++ b/tools/memory-model/linux-kernel.def
@@ -13,7 +13,7 @@ WRITE_ONCE(X,V) { __store{once}(X,V); }
 smp_store_release(X,V) { __store{release}(*X,V); }
 smp_load_acquire(X) __load{acquire}(*X)
 rcu_assign_pointer(X,V) { __store{release}(X,V); }
-rcu_dereference(X) __load{deref}(X)
+rcu_dereference(X) __load{once}(X)
 
 // Fences
 smp_mb() { __fence{mb} ; }