Re: [PATCH v2] locking/rwsem: Add reader-owned state to the owner field

[Peter Zijlstra <peterz@infradead.org>]

On Thu, May 19, 2016 at 11:37:53AM +1000, Dave Chinner wrote:
> On Fri, May 06, 2016 at 08:20:24PM -0400, Waiman Long wrote:
> > Currently, it is not possible to determine for sure if a reader
> > owns a rwsem by looking at the content of the rwsem data structure.
> > This patch adds a new state RWSEM_READER_OWNED to the owner field
> > to indicate that readers currently own the lock.
> 
> Oh, yes please. This will enable us to get rid of the remaining
> mrlock rwsem abstraction we've carried since the days of Irix in
> XFS. The only reason the abstraction still exists is that we track
> write locks for the purposes of checking for correct inode locking
> contexts via ASSERT(xfs_isilocked()) calls....

That all seems to live under a DEBUG knob; lockdep could also easily
tell you these things.

---
 include/linux/lockdep.h  | 36 +++++++++++++++++++++++++++++-------
 kernel/locking/lockdep.c | 19 ++++++++++++-------
 2 files changed, 41 insertions(+), 14 deletions(-)

diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
index eabe0138eb06..82d6453c4660 100644
--- a/include/linux/lockdep.h
+++ b/include/linux/lockdep.h
@@ -338,9 +338,21 @@ extern void lock_acquire(struct lockdep_map *lock, unsigned int subclass,
 extern void lock_release(struct lockdep_map *lock, int nested,
 			 unsigned long ip);
 
-#define lockdep_is_held(lock)	lock_is_held(&(lock)->dep_map)
+#define LOCKDEP_HELD_ANY	(-1)
+#define LOCKDEP_HELD_EXCLUSIVE	(0)
+#define LOCKDEP_HELD_SHARED	(1)
+#define LOCKDEP_HELD_RECURSIVE	(2)
 
-extern int lock_is_held(struct lockdep_map *lock);
+extern int _lock_is_held(struct lockdep_map *lock, int read);
+
+static inline int lock_is_held(struct lockdep_map *lock)
+{
+	return _lock_is_held(lock, LOCKDEP_HELD_ANY);
+}
+
+#define lockdep_is_held(lock)		_lock_is_held(&(lock)->dep_map, LOCKDEP_HELD_ANY)
+#define lockdep_is_held_exclusive(lock)	_lock_is_held(&(lock)->dep_map, LOCKDEP_HELD_EXCLUSIVE)
+#define lockdep_is_held_shared(lock)	_lock_is_held(&(lock)->dep_map, LOCKDEP_HELD_SHARED)
 
 extern void lock_set_class(struct lockdep_map *lock, const char *name,
 			   struct lock_class_key *key, unsigned int subclass,
@@ -369,12 +381,20 @@ extern void lock_unpin_lock(struct lockdep_map *lock, struct pin_cookie);
 #define lockdep_depth(tsk)	(debug_locks ? (tsk)->lockdep_depth : 0)
 
 #define lockdep_assert_held(l)	do {				\
-		WARN_ON(debug_locks && !lockdep_is_held(l));	\
-	} while (0)
+	WARN_ON(debug_locks && !lockdep_is_held(l));		\
+} while (0)
 
-#define lockdep_assert_held_once(l)	do {				\
-		WARN_ON_ONCE(debug_locks && !lockdep_is_held(l));	\
-	} while (0)
+#define lockdep_assert_held_once(l)	do {			\
+	WARN_ON_ONCE(debug_locks && !lockdep_is_held(l));	\
+} while (0)
+
+#define lockdep_assert_held_exclusive(l) do {			\
+	WARN_ON(debug_locks && !lockdep_is_held_exclusive(l));	\
+} while (0)
+
+#define lockdep_assert_held_shared(l) do {			\
+	WARN_ON(debug_locks && !lockdep_is_held_shared(l));	\
+} while (0)
 
 #define lockdep_recursing(tsk)	((tsk)->lockdep_recursion)
 
@@ -430,6 +450,8 @@ struct lock_class_key { };
 
 #define lockdep_assert_held(l)			do { (void)(l); } while (0)
 #define lockdep_assert_held_once(l)		do { (void)(l); } while (0)
+#define lockdep_assert_held_exclusive(l)	do { (void)(l); } while (0)
+#define lockdep_assert_held_shared(l)		do { (void)(l); } while (0)
 
 #define lockdep_recursing(tsk)			(0)
 
diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 81f1a7107c0e..dca5d982b315 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -3183,7 +3183,7 @@ print_lock_nested_lock_not_held(struct task_struct *curr,
 	return 0;
 }
 
-static int __lock_is_held(struct lockdep_map *lock);
+static int __lock_is_held(struct lockdep_map *lock, int read);
 
 /*
  * This gets called for every mutex_lock*()/spin_lock*() operation.
@@ -3324,7 +3324,7 @@ static int __lock_acquire(struct lockdep_map *lock, unsigned int subclass,
 	}
 	chain_key = iterate_chain_key(chain_key, class_idx);
 
-	if (nest_lock && !__lock_is_held(nest_lock))
+	if (nest_lock && !__lock_is_held(nest_lock, LOCKDEP_HELD_ANY))
 		return print_lock_nested_lock_not_held(curr, hlock, ip);
 
 	if (!validate_chain(curr, lock, hlock, chain_head, chain_key))
@@ -3571,7 +3571,7 @@ __lock_release(struct lockdep_map *lock, int nested, unsigned long ip)
 	return 1;
 }
 
-static int __lock_is_held(struct lockdep_map *lock)
+static int __lock_is_held(struct lockdep_map *lock, int read)
 {
 	struct task_struct *curr = current;
 	int i;
@@ -3579,8 +3579,13 @@ static int __lock_is_held(struct lockdep_map *lock)
 	for (i = 0; i < curr->lockdep_depth; i++) {
 		struct held_lock *hlock = curr->held_locks + i;
 
-		if (match_held_lock(hlock, lock))
+		if (!match_held_lock(hlock, lock))
+			continue;
+
+		if (read == LOCKDEP_HELD_ANY)
 			return 1;
+
+		return read == hlock->read;
 	}
 
 	return 0;
@@ -3764,7 +3769,7 @@ void lock_release(struct lockdep_map *lock, int nested,
 }
 EXPORT_SYMBOL_GPL(lock_release);
 
-int lock_is_held(struct lockdep_map *lock)
+int _lock_is_held(struct lockdep_map *lock, int read)
 {
 	unsigned long flags;
 	int ret = 0;
@@ -3776,13 +3781,13 @@ int lock_is_held(struct lockdep_map *lock)
 	check_flags(flags);
 
 	current->lockdep_recursion = 1;
-	ret = __lock_is_held(lock);
+	ret = __lock_is_held(lock, read);
 	current->lockdep_recursion = 0;
 	raw_local_irq_restore(flags);
 
 	return ret;
 }
-EXPORT_SYMBOL_GPL(lock_is_held);
+EXPORT_SYMBOL_GPL(_lock_is_held);
 
 struct pin_cookie lock_pin_lock(struct lockdep_map *lock)
 {