1 files changed, 40 insertions, 79 deletions

diff --git a/lib/kernel_lock.c b/lib/kernel_lock.c
index b135d04aa48a..bc62ed84f71f 100644
--- a/lib/kernel_lock.c
+++ b/lib/kernel_lock.c
@@ -7,114 +7,64 @@
  */
 #include <linux/module.h>
 #include <linux/kallsyms.h>
-#include <linux/semaphore.h>
+#include <linux/mutex.h>
 #include <linux/smp_lock.h>
 
 #define CREATE_TRACE_POINTS
 #include <trace/events/bkl.h>
 
 /*
- * The 'big kernel lock'
+ * The 'big kernel semaphore'
  *
- * This spinlock is taken and released recursively by lock_kernel()
+ * This mutex 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  __cacheline_aligned_in_smp DEFINE_RAW_SPINLOCK(kernel_flag);
-
+DEFINE_MUTEX(kernel_sem);
 
 /*
- * Acquire/release the underlying lock from the scheduler.
+ * Re-acquire the kernel semaphore.
  *
- * This is called with preemption disabled, and should
- * return an error value if it cannot get the lock and
- * TIF_NEED_RESCHED gets set.
+ * This function is called with preemption off.
  *
- * If it successfully gets the lock, it should increment
- * the preemption count like any spinlock does.
+ * 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.
  *
- * (This works on UP too - do_raw_spin_trylock will never
- * return false in that case)
+ * Called with interrupts disabled.
  */
 int __lockfunc __reacquire_kernel_lock(void)
 {
-        while (!do_raw_spin_trylock(&kernel_flag)) {
-                if (need_resched())
-                        return -EAGAIN;
-                cpu_relax();
-        }
-        preempt_disable();
-        return 0;
-}
+        int saved_lock_depth = current->lock_depth;
 
-void __lockfunc __release_kernel_lock(void)
-{
-        do_raw_spin_unlock(&kernel_flag);
-        preempt_enable_no_resched();
-}
+        BUG_ON(saved_lock_depth < 0);
 
-/*
- * 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
- * do_raw_spin_trylock() will always succeed.
- */
-#ifdef CONFIG_PREEMPT
-static inline void __lock_kernel(void)
-{
-        preempt_disable();
-        if (unlikely(!do_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) {
-                        do_raw_spin_lock(&kernel_flag);
-                        return;
-                }
+        current->lock_depth = -1;
+        local_irq_enable();
 
-                /*
-                 * Otherwise, let's wait for the kernel lock
-                 * with preemption enabled..
-                 */
-                do {
-                        preempt_enable();
-                        while (raw_spin_is_locked(&kernel_flag))
-                                cpu_relax();
-                        preempt_disable();
-                } while (!do_raw_spin_trylock(&kernel_flag));
-        }
-}
+        mutex_lock(&kernel_sem);
 
-#else
+        local_irq_disable();
+        current->lock_depth = saved_lock_depth;
 
-/*
- * Non-preemption case - just get the spinlock
- */
-static inline void __lock_kernel(void)
-{
-        do_raw_spin_lock(&kernel_flag);
+        return 0;
 }
-#endif
 
-static inline void __unlock_kernel(void)
+void __lockfunc __release_kernel_lock(void)
 {
-        /*
-         * the BKL is not covered by lockdep, so we open-code the
-         * unlocking sequence (and thus avoid the dep-chain ops):
-         */
-        do_raw_spin_unlock(&kernel_flag);
-        preempt_enable();
+        mutex_unlock(&kernel_sem);
 }
 
 /*
- * Getting the big kernel lock.
- *
- * This cannot happen asynchronously, so we only need to
- * worry about other CPU's.
+ * Getting the big kernel semaphore.
  */
 void __lockfunc _lock_kernel(const char *func, const char *file, int line)
 {
@@ -124,17 +74,28 @@ void __lockfunc _lock_kernel(const char *func, const char *file, int line)
 
         if (likely(!depth)) {
                 might_sleep();
-                __lock_kernel();
+                /*
+                 * No recursion worries - we set up lock_depth _after_
+                 */
+                mutex_lock(&kernel_sem);
+#ifdef CONFIG_DEBUG_RT_MUTEXES
+                current->last_kernel_lock = __builtin_return_address(0);
+#endif
         }
+
         current->lock_depth = depth;
 }
 
 void __lockfunc _unlock_kernel(const char *func, const char *file, int line)
 {
         BUG_ON(current->lock_depth < 0);
-        if (likely(--current->lock_depth < 0))
-                __unlock_kernel();
 
+        if (likely(--current->lock_depth < 0)) {
+#ifdef CONFIG_DEBUG_RT_MUTEXES
+                current->last_kernel_lock = NULL;
+#endif
+                mutex_unlock(&kernel_sem);
+        }
         trace_unlock_kernel(func, file, line);
 }