Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6

1 files changed, 81 insertions, 39 deletions

diff --git a/lib/kernel_lock.c b/lib/kernel_lock.c
index cd3e82530b0..01a3c22c1b5 100644
--- a/lib/kernel_lock.c
+++ b/lib/kernel_lock.c
@@ -11,79 +11,121 @@
 #include <linux/semaphore.h>
 
 /*
- * The 'big kernel semaphore'
+ * The 'big kernel lock'
  *
- * This mutex is taken and released recursively by lock_kernel()
+ * This spinlock is taken and released recursively by lock_kernel()
  * and unlock_kernel().  It is transparently dropped and reacquired
  * over schedule().  It is used to protect legacy code that hasn't
  * been migrated to a proper locking design yet.
  *
- * Note: code locked by this semaphore will only be serialized against
- * other code using the same locking facility. The code guarantees that
- * the task remains on the same CPU.
- *
  * Don't use in new code.
  */
-static DECLARE_MUTEX(kernel_sem);
+static  __cacheline_aligned_in_smp DEFINE_SPINLOCK(kernel_flag);
+
 
 /*
- * Re-acquire the kernel semaphore.
+ * Acquire/release the underlying lock from the scheduler.
  *
- * This function is called with preemption off.
+ * This is called with preemption disabled, and should
+ * return an error value if it cannot get the lock and
+ * TIF_NEED_RESCHED gets set.
  *
- * We are executing in schedule() so the code must be extremely careful
- * about recursion, both due to the down() and due to the enabling of
- * preemption. schedule() will re-check the preemption flag after
- * reacquiring the semaphore.
+ * If it successfully gets the lock, it should increment
+ * the preemption count like any spinlock does.
+ *
+ * (This works on UP too - _raw_spin_trylock will never
+ * return false in that case)
  */
 int __lockfunc __reacquire_kernel_lock(void)
 {
-	struct task_struct *task = current;
-	int saved_lock_depth = task->lock_depth;
-
-	BUG_ON(saved_lock_depth < 0);
-
-	task->lock_depth = -1;
-	preempt_enable_no_resched();
-
-	down(&kernel_sem);
-
+	while (!_raw_spin_trylock(&kernel_flag)) {
+		if (test_thread_flag(TIF_NEED_RESCHED))
+			return -EAGAIN;
+		cpu_relax();
+	}
 	preempt_disable();
-	task->lock_depth = saved_lock_depth;
-
 	return 0;
 }
 
 void __lockfunc __release_kernel_lock(void)
 {
-	up(&kernel_sem);
+	_raw_spin_unlock(&kernel_flag);
+	preempt_enable_no_resched();
 }
 
 /*
- * Getting the big kernel semaphore.
+ * These are the BKL spinlocks - we try to be polite about preemption.
+ * If SMP is not on (ie UP preemption), this all goes away because the
+ * _raw_spin_trylock() will always succeed.
  */
-void __lockfunc lock_kernel(void)
+#ifdef CONFIG_PREEMPT
+static inline void __lock_kernel(void)
 {
-	struct task_struct *task = current;
-	int depth = task->lock_depth + 1;
+	preempt_disable();
+	if (unlikely(!_raw_spin_trylock(&kernel_flag))) {
+		/*
+		 * If preemption was disabled even before this
+		 * was called, there's nothing we can be polite
+		 * about - just spin.
+		 */
+		if (preempt_count() > 1) {
+			_raw_spin_lock(&kernel_flag);
+			return;
+		}
 
-	if (likely(!depth))
 		/*
-		 * No recursion worries - we set up lock_depth _after_
+		 * Otherwise, let's wait for the kernel lock
+		 * with preemption enabled..
 		 */
-		down(&kernel_sem);
+		do {
+			preempt_enable();
+			while (spin_is_locked(&kernel_flag))
+				cpu_relax();
+			preempt_disable();
+		} while (!_raw_spin_trylock(&kernel_flag));
+	}
+}
 
-	task->lock_depth = depth;
+#else
+
+/*
+ * Non-preemption case - just get the spinlock
+ */
+static inline void __lock_kernel(void)
+{
+	_raw_spin_lock(&kernel_flag);
 }
+#endif
 
-void __lockfunc unlock_kernel(void)
+static inline void __unlock_kernel(void)
 {
-	struct task_struct *task = current;
+	/*
+	 * the BKL is not covered by lockdep, so we open-code the
+	 * unlocking sequence (and thus avoid the dep-chain ops):
+	 */
+	_raw_spin_unlock(&kernel_flag);
+	preempt_enable();
+}
 
-	BUG_ON(task->lock_depth < 0);
+/*
+ * Getting the big kernel lock.
+ *
+ * This cannot happen asynchronously, so we only need to
+ * worry about other CPU's.
+ */
+void __lockfunc lock_kernel(void)
+{
+	int depth = current->lock_depth+1;
+	if (likely(!depth))
+		__lock_kernel();
+	current->lock_depth = depth;
+}
 
-	if (likely(--task->lock_depth < 0))
-		up(&kernel_sem);
+void __lockfunc unlock_kernel(void)
+{
+	BUG_ON(current->lock_depth < 0);
+	if (likely(--current->lock_depth < 0))
+		__unlock_kernel();
 }
 
 EXPORT_SYMBOL(lock_kernel);