[PATCH 13/14] lockdep: Implement lock pinning

[Peter Zijlstra]

Add a lockdep annotation that WARNs if you 'accidentially' unlock a
lock.

This is especially helpful for code with callbacks, where the upper
layer assumes a lock remains taken but a lower layer thinks it maybe
can drop and reacquire the lock.

By unwittingly breaking up the lock, races can be introduced.

Lock pinning is a lockdep annotation that helps with this, when you
lockdep_pin_lock() a held lock, any unlock without a
lockdep_unpin_lock() will produce a WARN. Think of this as a relative
of lockdep_assert_held(), except you don't only assert its held now,
but ensure it stays held until you release your assertion.

RFC: a possible alternative API would be something like:

  int cookie = lockdep_pin_lock(&foo);
  ...
  lockdep_unpin_lock(&foo, cookie);

Where we pick a random number for the pin_count; this makes it
impossible to sneak a lock break in without also passing the right
cookie along.

I've not done this because it ends up generating code for !LOCKDEP,
esp. if you need to pass the cookie around for some reason.

Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
---
 include/linux/lockdep.h  |   10 +++++
 kernel/locking/lockdep.c |   80 +++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 90 insertions(+)

--- a/include/linux/lockdep.h
+++ b/include/linux/lockdep.h
@@ -255,6 +255,7 @@ struct held_lock {
 	unsigned int check:1;       /* see lock_acquire() comment */
 	unsigned int hardirqs_off:1;
 	unsigned int references:12;					/* 32 bits */
+	unsigned int pin_count;
 };
 
 /*
@@ -354,6 +355,9 @@ extern void lockdep_set_current_reclaim_
 extern void lockdep_clear_current_reclaim_state(void);
 extern void lockdep_trace_alloc(gfp_t mask);
 
+extern void lock_pin_lock(struct lockdep_map *lock);
+extern void lock_unpin_lock(struct lockdep_map *lock);
+
 # define INIT_LOCKDEP				.lockdep_recursion = 0, .lockdep_reclaim_gfp = 0,
 
 #define lockdep_depth(tsk)	(debug_locks ? (tsk)->lockdep_depth : 0)
@@ -368,6 +372,9 @@ extern void lockdep_trace_alloc(gfp_t ma
 
 #define lockdep_recursing(tsk)	((tsk)->lockdep_recursion)
 
+#define lockdep_pin_lock(l)		lock_pin_lock(&(l)->dep_map)
+#define lockdep_unpin_lock(l)	lock_unpin_lock(&(l)->dep_map)
+
 #else /* !CONFIG_LOCKDEP */
 
 static inline void lockdep_off(void)
@@ -420,6 +427,9 @@ struct lock_class_key { };
 
 #define lockdep_recursing(tsk)			(0)
 
+#define lockdep_pin_lock(l)				do { (void)(l); } while (0)
+#define lockdep_unpin_lock(l)			do { (void)(l); } while (0)
+
 #endif /* !LOCKDEP */
 
 #ifdef CONFIG_LOCK_STAT
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -3157,6 +3157,7 @@ static int __lock_acquire(struct lockdep
 	hlock->waittime_stamp = 0;
 	hlock->holdtime_stamp = lockstat_clock();
 #endif
+	hlock->pin_count = 0;
 
 	if (check && !mark_irqflags(curr, hlock))
 		return 0;
@@ -3403,6 +3404,8 @@ __lock_release(struct lockdep_map *lock,
 	if (hlock->instance == lock)
 		lock_release_holdtime(hlock);
 
+	WARN(hlock->pin_count, "releasing a stayed lock\n");
+
 	if (hlock->references) {
 		hlock->references--;
 		if (hlock->references) {
@@ -3459,6 +3462,49 @@ static int __lock_is_held(struct lockdep
 	return 0;
 }
 
+static void __lock_pin_lock(struct lockdep_map *lock)
+{
+	struct task_struct *curr = current;
+	int i;
+
+	if (unlikely(!debug_locks))
+		return;
+
+	for (i = 0; i < curr->lockdep_depth; i++) {
+		struct held_lock *hlock = curr->held_locks + i;
+
+		if (match_held_lock(hlock, lock)) {
+			hlock->pin_count++;
+			return;
+		}
+	}
+
+	WARN(1, "staying unheld lock\n");
+}
+
+static void __lock_unpin_lock(struct lockdep_map *lock)
+{
+	struct task_struct *curr = current;
+	int i;
+
+	if (unlikely(!debug_locks))
+		return;
+
+	for (i = 0; i < curr->lockdep_depth; i++) {
+		struct held_lock *hlock = curr->held_locks + i;
+
+		if (match_held_lock(hlock, lock)) {
+			if (WARN(!hlock->pin_count, "unstaying unstayed lock\n"))
+				return;
+
+			hlock->pin_count--;
+			return;
+		}
+	}
+
+	WARN(1, "unstaying unheld lock\n");
+}
+
 /*
  * Check whether we follow the irq-flags state precisely:
  */
@@ -3582,6 +3628,40 @@ int lock_is_held(struct lockdep_map *loc
 }
 EXPORT_SYMBOL_GPL(lock_is_held);
 
+void lock_pin_lock(struct lockdep_map *lock)
+{
+	unsigned long flags;
+
+	if (unlikely(current->lockdep_recursion))
+		return;
+
+	raw_local_irq_save(flags);
+	check_flags(flags);
+
+	current->lockdep_recursion = 1;
+	__lock_pin_lock(lock);
+	current->lockdep_recursion = 0;
+	raw_local_irq_restore(flags);
+}
+EXPORT_SYMBOL_GPL(lock_pin_lock);
+
+void lock_unpin_lock(struct lockdep_map *lock)
+{
+	unsigned long flags;
+
+	if (unlikely(current->lockdep_recursion))
+		return;
+
+	raw_local_irq_save(flags);
+	check_flags(flags);
+
+	current->lockdep_recursion = 1;
+	__lock_unpin_lock(lock);
+	current->lockdep_recursion = 0;
+	raw_local_irq_restore(flags);
+}
+EXPORT_SYMBOL_GPL(lock_unpin_lock);
+
 void lockdep_set_current_reclaim_state(gfp_t gfp_mask)
 {
 	current->lockdep_reclaim_gfp = gfp_mask;


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/