1 files changed, 129 insertions, 4 deletions

diff --git a/kernel/locking/rwsem.c b/kernel/locking/rwsem.c
index 2d7cabcfca50..e1e0bac957c4 100644
--- a/kernel/locking/rwsem.c
+++ b/kernel/locking/rwsem.c
@@ -59,6 +59,42 @@
  * seems to hang on a reader owned rwsem especially if only one reader
  * is involved. Ideally we would like to track all the readers that own
  * a rwsem, but the overhead is simply too big.
+ *
+ * Reader optimistic spinning is helpful when the reader critical section
+ * is short and there aren't that many readers around. It makes readers
+ * relatively more preferred than writers. When a writer times out spinning
+ * on a reader-owned lock and set the nospinnable bits, there are two main
+ * reasons for that.
+ *
+ *  1) The reader critical section is long, perhaps the task sleeps after
+ *     acquiring the read lock.
+ *  2) There are just too many readers contending the lock causing it to
+ *     take a while to service all of them.
+ *
+ * In the former case, long reader critical section will impede the progress
+ * of writers which is usually more important for system performance. In
+ * the later case, reader optimistic spinning tends to make the reader
+ * groups that contain readers that acquire the lock together smaller
+ * leading to more of them. That may hurt performance in some cases. In
+ * other words, the setting of nonspinnable bits indicates that reader
+ * optimistic spinning may not be helpful for those workloads that cause
+ * it.
+ *
+ * Therefore, any writers that had observed the setting of the writer
+ * nonspinnable bit for a given rwsem after they fail to acquire the lock
+ * via optimistic spinning will set the reader nonspinnable bit once they
+ * acquire the write lock. Similarly, readers that observe the setting
+ * of reader nonspinnable bit at slowpath entry will set the reader
+ * nonspinnable bits when they acquire the read lock via the wakeup path.
+ *
+ * Once the reader nonspinnable bit is on, it will only be reset when
+ * a writer is able to acquire the rwsem in the fast path or somehow a
+ * reader or writer in the slowpath doesn't observe the nonspinable bit.
+ *
+ * This is to discourage reader optmistic spinning on that particular
+ * rwsem and make writers more preferred. This adaptive disabling of reader
+ * optimistic spinning will alleviate the negative side effect of this
+ * feature.
  */
 #define RWSEM_READER_OWNED	(1UL << 0)
 #define RWSEM_RD_NONSPINNABLE	(1UL << 1)
@@ -144,11 +180,14 @@ static inline bool rwsem_test_oflags(struct rw_semaphore *sem, long flags)
  * Note that the owner value just indicates the task has owned the rwsem
  * previously, it may not be the real owner or one of the real owners
  * anymore when that field is examined, so take it with a grain of salt.
+ *
+ * The reader non-spinnable bit is preserved.
  */
 static inline void __rwsem_set_reader_owned(struct rw_semaphore *sem,
 					    struct task_struct *owner)
 {
-	unsigned long val = (unsigned long)owner | RWSEM_READER_OWNED;
+	unsigned long val = (unsigned long)owner | RWSEM_READER_OWNED |
+		(atomic_long_read(&sem->owner) & RWSEM_RD_NONSPINNABLE);
 
 	atomic_long_set(&sem->owner, val);
 }
@@ -287,6 +326,7 @@ struct rwsem_waiter {
 	struct task_struct *task;
 	enum rwsem_waiter_type type;
 	unsigned long timeout;
+	unsigned long last_rowner;
 };
 #define rwsem_first_waiter(sem) \
 	list_first_entry(&sem->wait_list, struct rwsem_waiter, list)
@@ -368,6 +408,8 @@ static void rwsem_mark_wake(struct rw_semaphore *sem,
 	 * so we can bail out early if a writer stole the lock.
 	 */
 	if (wake_type != RWSEM_WAKE_READ_OWNED) {
+		struct task_struct *owner;
+
 		adjustment = RWSEM_READER_BIAS;
 		oldcount = atomic_long_fetch_add(adjustment, &sem->count);
 		if (unlikely(oldcount & RWSEM_WRITER_MASK)) {
@@ -388,8 +430,15 @@ static void rwsem_mark_wake(struct rw_semaphore *sem,
 		/*
 		 * Set it to reader-owned to give spinners an early
 		 * indication that readers now have the lock.
+		 * The reader nonspinnable bit seen at slowpath entry of
+		 * the reader is copied over.
 		 */
-		__rwsem_set_reader_owned(sem, waiter->task);
+		owner = waiter->task;
+		if (waiter->last_rowner & RWSEM_RD_NONSPINNABLE) {
+			owner = (void *)((unsigned long)owner | RWSEM_RD_NONSPINNABLE);
+			lockevent_inc(rwsem_opt_norspin);
+		}
+		__rwsem_set_reader_owned(sem, owner);
 	}
 
 	/*
@@ -836,6 +885,42 @@ static inline void clear_wr_nonspinnable(struct rw_semaphore *sem)
 	if (rwsem_test_oflags(sem, RWSEM_WR_NONSPINNABLE))
 		atomic_long_andnot(RWSEM_WR_NONSPINNABLE, &sem->owner);
 }
+
+/*
+ * This function is called when the reader fails to acquire the lock via
+ * optimistic spinning. In this case we will still attempt to do a trylock
+ * when comparing the rwsem state right now with the state when entering
+ * the slowpath indicates that the reader is still in a valid reader phase.
+ * This happens when the following conditions are true:
+ *
+ * 1) The lock is currently reader owned, and
+ * 2) The lock is previously not reader-owned or the last read owner changes.
+ *
+ * In the former case, we have transitioned from a writer phase to a
+ * reader-phase while spinning. In the latter case, it means the reader
+ * phase hasn't ended when we entered the optimistic spinning loop. In
+ * both cases, the reader is eligible to acquire the lock. This is the
+ * secondary path where a read lock is acquired optimistically.
+ *
+ * The reader non-spinnable bit wasn't set at time of entry or it will
+ * not be here at all.
+ */
+static inline bool rwsem_reader_phase_trylock(struct rw_semaphore *sem,
+					      unsigned long last_rowner)
+{
+	unsigned long owner = atomic_long_read(&sem->owner);
+
+	if (!(owner & RWSEM_READER_OWNED))
+		return false;
+
+	if (((owner ^ last_rowner) & ~RWSEM_OWNER_FLAGS_MASK) &&
+	    rwsem_try_read_lock_unqueued(sem)) {
+		lockevent_inc(rwsem_opt_rlock2);
+		lockevent_add(rwsem_opt_fail, -1);
+		return true;
+	}
+	return false;
+}
 #else
 static inline bool rwsem_can_spin_on_owner(struct rw_semaphore *sem,
 					   unsigned long nonspinnable)
@@ -849,6 +934,12 @@ static inline bool rwsem_optimistic_spin(struct rw_semaphore *sem, bool wlock)
 }
 
 static inline void clear_wr_nonspinnable(struct rw_semaphore *sem) { }
+
+static inline bool rwsem_reader_phase_trylock(struct rw_semaphore *sem,
+					      unsigned long last_rowner)
+{
+	return false;
+}
 #endif
 
 /*
@@ -862,6 +953,14 @@ rwsem_down_read_slowpath(struct rw_semaphore *sem, int state)
 	struct rwsem_waiter waiter;
 	DEFINE_WAKE_Q(wake_q);
 
+	/*
+	 * Save the current read-owner of rwsem, if available, and the
+	 * reader nonspinnable bit.
+	 */
+	waiter.last_rowner = atomic_long_read(&sem->owner);
+	if (!(waiter.last_rowner & RWSEM_READER_OWNED))
+		waiter.last_rowner &= RWSEM_RD_NONSPINNABLE;
+
 	if (!rwsem_can_spin_on_owner(sem, RWSEM_RD_NONSPINNABLE))
 		goto queue;
 
@@ -884,6 +983,8 @@ rwsem_down_read_slowpath(struct rw_semaphore *sem, int state)
 			wake_up_q(&wake_q);
 		}
 		return sem;
+	} else if (rwsem_reader_phase_trylock(sem, waiter.last_rowner)) {
+		return sem;
 	}
 
 queue:
@@ -965,12 +1066,26 @@ out_nolock:
 }
 
 /*
+ * This function is called by the a write lock owner. So the owner value
+ * won't get changed by others.
+ */
+static inline void rwsem_disable_reader_optspin(struct rw_semaphore *sem,
+						bool disable)
+{
+	if (unlikely(disable)) {
+		atomic_long_or(RWSEM_RD_NONSPINNABLE, &sem->owner);
+		lockevent_inc(rwsem_opt_norspin);
+	}
+}
+
+/*
  * Wait until we successfully acquire the write lock
  */
 static struct rw_semaphore *
 rwsem_down_write_slowpath(struct rw_semaphore *sem, int state)
 {
 	long count;
+	bool disable_rspin;
 	enum writer_wait_state wstate;
 	struct rwsem_waiter waiter;
 	struct rw_semaphore *ret = sem;
@@ -982,6 +1097,13 @@ rwsem_down_write_slowpath(struct rw_semaphore *sem, int state)
 		return sem;
 
 	/*
+	 * Disable reader optimistic spinning for this rwsem after
+	 * acquiring the write lock when the setting of the nonspinnable
+	 * bits are observed.
+	 */
+	disable_rspin = atomic_long_read(&sem->owner) & RWSEM_NONSPINNABLE;
+
+	/*
 	 * Optimistic spinning failed, proceed to the slowpath
 	 * and block until we can acquire the sem.
 	 */
@@ -1077,6 +1199,7 @@ wait:
 	}
 	__set_current_state(TASK_RUNNING);
 	list_del(&waiter.list);
+	rwsem_disable_reader_optspin(sem, disable_rspin);
 	raw_spin_unlock_irq(&sem->wait_lock);
 	lockevent_inc(rwsem_wlock);
 
@@ -1196,7 +1319,8 @@ static inline void __down_write(struct rw_semaphore *sem)
 	if (unlikely(!atomic_long_try_cmpxchg_acquire(&sem->count, &tmp,
 						      RWSEM_WRITER_LOCKED)))
 		rwsem_down_write_slowpath(sem, TASK_UNINTERRUPTIBLE);
-	rwsem_set_owner(sem);
+	else
+		rwsem_set_owner(sem);
 }
 
 static inline int __down_write_killable(struct rw_semaphore *sem)
@@ -1207,8 +1331,9 @@ static inline int __down_write_killable(struct rw_semaphore *sem)
 						      RWSEM_WRITER_LOCKED))) {
 		if (IS_ERR(rwsem_down_write_slowpath(sem, TASK_KILLABLE)))
 			return -EINTR;
+	} else {
+		rwsem_set_owner(sem);
 	}
-	rwsem_set_owner(sem);
 	return 0;
 }