locking/rwsem: Support optimistic spinning

We have reached the point where our mutexes are quite fine tuned
for a number of situations. This includes the use of heuristics
and optimistic spinning, based on MCS locking techniques.

Exclusive ownership of read-write semaphores are, conceptually,
just about the same as mutexes, making them close cousins. To
this end we need to make them both perform similarly, and
right now, rwsems are simply not up to it. This was discovered
by both reverting commit 4fc3f1d6 (mm/rmap, migration: Make
rmap_walk_anon() and try_to_unmap_anon() more scalable) and
similarly, converting some other mutexes (ie: i_mmap_mutex) to
rwsems. This creates a situation where users have to choose
between a rwsem and mutex taking into account this important
performance difference. Specifically, biggest difference between
both locks is when we fail to acquire a mutex in the fastpath,
optimistic spinning comes in to play and we can avoid a large
amount of unnecessary sleeping and overhead of moving tasks in
and out of wait queue. Rwsems do not have such logic.

This patch, based on the work from Tim Chen and I, adds support
for write-side optimistic spinning when the lock is contended.
It also includes support for the recently added cancelable MCS
locking for adaptive spinning. Note that is is only applicable
to the xadd method, and the spinlock rwsem variant remains intact.

Allowing optimistic spinning before putting the writer on the wait
queue reduces wait queue contention and provided greater chance
for the rwsem to get acquired. With these changes, rwsem is on par
with mutex. The performance benefits can be seen on a number of
workloads. For instance, on a 8 socket, 80 core 64bit Westmere box,
aim7 shows the following improvements in throughput:

 +--------------+---------------------+-----------------+
 |   Workload   | throughput-increase | number of users |
 +--------------+---------------------+-----------------+
 | alltests     | 20%                 | >1000           |
 | custom       | 27%, 60%            | 10-100, >1000   |
 | high_systime | 36%, 30%            | >100, >1000     |
 | shared       | 58%, 29%            | 10-100, >1000   |
 +--------------+---------------------+-----------------+

There was also improvement on smaller systems, such as a quad-core
x86-64 laptop running a 30Gb PostgreSQL (pgbench) workload for up
to +60% in throughput for over 50 clients. Additionally, benefits
were also noticed in exim (mail server) workloads. Furthermore, no
performance regression have been seen at all.

Based-on-work-from: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Davidlohr Bueso <davidlohr@hp.com>
[peterz: rej fixup due to comment patches, sched/rt.h header]
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Alex Shi <alex.shi@linaro.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Michel Lespinasse <walken@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Peter Hurley <peter@hurleysoftware.com>
Cc: "Paul E.McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Aswin Chandramouleeswaran <aswin@hp.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: "Scott J Norton" <scott.norton@hp.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Chris Mason <clm@fb.com>
Cc: Josef Bacik <jbacik@fusionio.com>
Link: http://lkml.kernel.org/r/1399055055.6275.15.camel@buesod1.americas.hpqcorp.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>

2 files changed, 226 insertions, 30 deletions

diff --git a/kernel/locking/rwsem-xadd.c b/kernel/locking/rwsem-xadd.c
index b4219ff..4a75278 100644
--- a/kernel/locking/rwsem-xadd.c
+++ b/kernel/locking/rwsem-xadd.c
@@ -5,11 +5,17 @@
  *
  * Writer lock-stealing by Alex Shi <alex.shi@intel.com>
  * and Michel Lespinasse <walken@google.com>
+ *
+ * Optimistic spinning by Tim Chen <tim.c.chen@intel.com>
+ * and Davidlohr Bueso <davidlohr@hp.com>. Based on mutexes.
  */
 #include <linux/rwsem.h>
 #include <linux/sched.h>
 #include <linux/init.h>
 #include <linux/export.h>
+#include <linux/sched/rt.h>
+
+#include "mcs_spinlock.h"
 
 /*
  * Guide to the rw_semaphore's count field for common values.
@@ -76,6 +82,10 @@ void __init_rwsem(struct rw_semaphore *sem, const char *name,
 	sem->count = RWSEM_UNLOCKED_VALUE;
 	raw_spin_lock_init(&sem->wait_lock);
 	INIT_LIST_HEAD(&sem->wait_list);
+#ifdef CONFIG_SMP
+	sem->owner = NULL;
+	sem->osq = NULL;
+#endif
 }
 
 EXPORT_SYMBOL(__init_rwsem);
@@ -190,7 +200,7 @@ __rwsem_do_wake(struct rw_semaphore *sem, enum rwsem_wake_type wake_type)
 }
 
 /*
- * wait for the read lock to be granted
+ * Wait for the read lock to be granted
  */
 __visible
 struct rw_semaphore __sched *rwsem_down_read_failed(struct rw_semaphore *sem)
@@ -237,64 +247,221 @@ struct rw_semaphore __sched *rwsem_down_read_failed(struct rw_semaphore *sem)
 	return sem;
 }
 
+static inline bool rwsem_try_write_lock(long count, struct rw_semaphore *sem)
+{
+	if (!(count & RWSEM_ACTIVE_MASK)) {
+		/* try acquiring the write lock */
+		if (sem->count == RWSEM_WAITING_BIAS &&
+		    cmpxchg(&sem->count, RWSEM_WAITING_BIAS,
+			    RWSEM_ACTIVE_WRITE_BIAS) == RWSEM_WAITING_BIAS) {
+			if (!list_is_singular(&sem->wait_list))
+				rwsem_atomic_update(RWSEM_WAITING_BIAS, sem);
+			return true;
+		}
+	}
+	return false;
+}
+
+#ifdef CONFIG_SMP
 /*
- * wait until we successfully acquire the write lock
+ * Try to acquire write lock before the writer has been put on wait queue.
+ */
+static inline bool rwsem_try_write_lock_unqueued(struct rw_semaphore *sem)
+{
+	long old, count = ACCESS_ONCE(sem->count);
+
+	while (true) {
+		if (!(count == 0 || count == RWSEM_WAITING_BIAS))
+			return false;
+
+		old = cmpxchg(&sem->count, count, count + RWSEM_ACTIVE_WRITE_BIAS);
+		if (old == count)
+			return true;
+
+		count = old;
+	}
+}
+
+static inline bool rwsem_can_spin_on_owner(struct rw_semaphore *sem)
+{
+	struct task_struct *owner;
+	bool on_cpu = true;
+
+	if (need_resched())
+		return 0;
+
+	rcu_read_lock();
+	owner = ACCESS_ONCE(sem->owner);
+	if (owner)
+		on_cpu = owner->on_cpu;
+	rcu_read_unlock();
+
+	/*
+	 * If sem->owner is not set, the rwsem owner may have
+	 * just acquired it and not set the owner yet or the rwsem
+	 * has been released.
+	 */
+	return on_cpu;
+}
+
+static inline bool owner_running(struct rw_semaphore *sem,
+				 struct task_struct *owner)
+{
+	if (sem->owner != owner)
+		return false;
+
+	/*
+	 * Ensure we emit the owner->on_cpu, dereference _after_ checking
+	 * sem->owner still matches owner, if that fails, owner might
+	 * point to free()d memory, if it still matches, the rcu_read_lock()
+	 * ensures the memory stays valid.
+	 */
+	barrier();
+
+	return owner->on_cpu;
+}
+
+static noinline
+bool rwsem_spin_on_owner(struct rw_semaphore *sem, struct task_struct *owner)
+{
+	rcu_read_lock();
+	while (owner_running(sem, owner)) {
+		if (need_resched())
+			break;
+
+		arch_mutex_cpu_relax();
+	}
+	rcu_read_unlock();
+
+	/*
+	 * We break out the loop above on need_resched() or when the
+	 * owner changed, which is a sign for heavy contention. Return
+	 * success only when sem->owner is NULL.
+	 */
+	return sem->owner == NULL;
+}
+
+static bool rwsem_optimistic_spin(struct rw_semaphore *sem)
+{
+	struct task_struct *owner;
+	bool taken = false;
+
+	preempt_disable();
+
+	/* sem->wait_lock should not be held when doing optimistic spinning */
+	if (!rwsem_can_spin_on_owner(sem))
+		goto done;
+
+	if (!osq_lock(&sem->osq))
+		goto done;
+
+	while (true) {
+		owner = ACCESS_ONCE(sem->owner);
+		if (owner && !rwsem_spin_on_owner(sem, owner))
+			break;
+
+		/* wait_lock will be acquired if write_lock is obtained */
+		if (rwsem_try_write_lock_unqueued(sem)) {
+			taken = true;
+			break;
+		}
+
+		/*
+		 * When there's no owner, we might have preempted between the
+		 * owner acquiring the lock and setting the owner field. If
+		 * we're an RT task that will live-lock because we won't let
+		 * the owner complete.
+		 */
+		if (!owner && (need_resched() || rt_task(current)))
+			break;
+
+		/*
+		 * The cpu_relax() call is a compiler barrier which forces
+		 * everything in this loop to be re-loaded. We don't need
+		 * memory barriers as we'll eventually observe the right
+		 * values at the cost of a few extra spins.
+		 */
+		arch_mutex_cpu_relax();
+	}
+	osq_unlock(&sem->osq);
+done:
+	preempt_enable();
+	return taken;
+}
+
+#else
+static bool rwsem_optimistic_spin(struct rw_semaphore *sem)
+{
+	return false;
+}
+#endif
+
+/*
+ * Wait until we successfully acquire the write lock
  */
 __visible
 struct rw_semaphore __sched *rwsem_down_write_failed(struct rw_semaphore *sem)
 {
-	long count, adjustment = -RWSEM_ACTIVE_WRITE_BIAS;
+	long count;
+	bool waiting = true; /* any queued threads before us */
 	struct rwsem_waiter waiter;
-	struct task_struct *tsk = current;
 
-	/* set up my own style of waitqueue */
-	waiter.task = tsk;
+	/* undo write bias from down_write operation, stop active locking */
+	count = rwsem_atomic_update(-RWSEM_ACTIVE_WRITE_BIAS, sem);
+
+	/* do optimistic spinning and steal lock if possible */
+	if (rwsem_optimistic_spin(sem))
+		return sem;
+
+	/*
+	 * Optimistic spinning failed, proceed to the slowpath
+	 * and block until we can acquire the sem.
+	 */
+	waiter.task = current;
 	waiter.type = RWSEM_WAITING_FOR_WRITE;
 
 	raw_spin_lock_irq(&sem->wait_lock);
+
+	/* account for this before adding a new element to the list */
 	if (list_empty(&sem->wait_list))
-		adjustment += RWSEM_WAITING_BIAS;
+		waiting = false;
+
 	list_add_tail(&waiter.list, &sem->wait_list);
 
 	/* we're now waiting on the lock, but no longer actively locking */
-	count = rwsem_atomic_update(adjustment, sem);
+	if (waiting) {
+		count = ACCESS_ONCE(sem->count);
 
-	/* If there were already threads queued before us and there are no
-	 * active writers, the lock must be read owned; so we try to wake
-	 * any read locks that were queued ahead of us. */
-	if (count > RWSEM_WAITING_BIAS &&
-	    adjustment == -RWSEM_ACTIVE_WRITE_BIAS)
-		sem = __rwsem_do_wake(sem, RWSEM_WAKE_READERS);
+		/*
+		 * If there were already threads queued before us and there are no
+		 * active writers, the lock must be read owned; so we try to wake
+		 * any read locks that were queued ahead of us.
+		 */
+		if (count > RWSEM_WAITING_BIAS)
+			sem = __rwsem_do_wake(sem, RWSEM_WAKE_READERS);
+
+	} else
+		count = rwsem_atomic_update(RWSEM_WAITING_BIAS, sem);
 
 	/* wait until we successfully acquire the lock */
-	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
+	set_current_state(TASK_UNINTERRUPTIBLE);
 	while (true) {
-		if (!(count & RWSEM_ACTIVE_MASK)) {
-			/* Try acquiring the write lock. */
-			count = RWSEM_ACTIVE_WRITE_BIAS;
-			if (!list_is_singular(&sem->wait_list))
-				count += RWSEM_WAITING_BIAS;
-
-			if (sem->count == RWSEM_WAITING_BIAS &&
-			    cmpxchg(&sem->count, RWSEM_WAITING_BIAS, count) ==
-							RWSEM_WAITING_BIAS)
-				break;
-		}
-
+		if (rwsem_try_write_lock(count, sem))
+			break;
 		raw_spin_unlock_irq(&sem->wait_lock);
 
 		/* Block until there are no active lockers. */
 		do {
 			schedule();
-			set_task_state(tsk, TASK_UNINTERRUPTIBLE);
+			set_current_state(TASK_UNINTERRUPTIBLE);
 		} while ((count = sem->count) & RWSEM_ACTIVE_MASK);
 
 		raw_spin_lock_irq(&sem->wait_lock);
 	}
+	__set_current_state(TASK_RUNNING);
 
 	list_del(&waiter.list);
 	raw_spin_unlock_irq(&sem->wait_lock);
-	tsk->state = TASK_RUNNING;
 
 	return sem;
 }
diff --git a/kernel/locking/rwsem.c b/kernel/locking/rwsem.c
index cfff143..42f806d 100644
--- a/kernel/locking/rwsem.c
+++ b/kernel/locking/rwsem.c
@@ -12,6 +12,27 @@
 
 #include <linux/atomic.h>
 
+#if defined(CONFIG_SMP) && defined(CONFIG_RWSEM_XCHGADD_ALGORITHM)
+static inline void rwsem_set_owner(struct rw_semaphore *sem)
+{
+	sem->owner = current;
+}
+
+static inline void rwsem_clear_owner(struct rw_semaphore *sem)
+{
+	sem->owner = NULL;
+}
+
+#else
+static inline void rwsem_set_owner(struct rw_semaphore *sem)
+{
+}
+
+static inline void rwsem_clear_owner(struct rw_semaphore *sem)
+{
+}
+#endif
+
 /*
  * lock for reading
  */
@@ -48,6 +69,7 @@ void __sched down_write(struct rw_semaphore *sem)
 	rwsem_acquire(&sem->dep_map, 0, 0, _RET_IP_);
 
 	LOCK_CONTENDED(sem, __down_write_trylock, __down_write);
+	rwsem_set_owner(sem);
 }
 
 EXPORT_SYMBOL(down_write);
@@ -59,8 +81,11 @@ int down_write_trylock(struct rw_semaphore *sem)
 {
 	int ret = __down_write_trylock(sem);
 
-	if (ret == 1)
+	if (ret == 1) {
 		rwsem_acquire(&sem->dep_map, 0, 1, _RET_IP_);
+		rwsem_set_owner(sem);
+	}
+
 	return ret;
 }
 
@@ -85,6 +110,7 @@ void up_write(struct rw_semaphore *sem)
 {
 	rwsem_release(&sem->dep_map, 1, _RET_IP_);
 
+	rwsem_clear_owner(sem);
 	__up_write(sem);
 }
 
@@ -99,6 +125,7 @@ void downgrade_write(struct rw_semaphore *sem)
 	 * lockdep: a downgraded write will live on as a write
 	 * dependency.
 	 */
+	rwsem_clear_owner(sem);
 	__downgrade_write(sem);
 }
 
@@ -122,6 +149,7 @@ void _down_write_nest_lock(struct rw_semaphore *sem, struct lockdep_map *nest)
 	rwsem_acquire_nest(&sem->dep_map, 0, 0, nest, _RET_IP_);
 
 	LOCK_CONTENDED(sem, __down_write_trylock, __down_write);
+	rwsem_set_owner(sem);
 }
 
 EXPORT_SYMBOL(_down_write_nest_lock);
@@ -141,6 +169,7 @@ void down_write_nested(struct rw_semaphore *sem, int subclass)
 	rwsem_acquire(&sem->dep_map, subclass, 0, _RET_IP_);
 
 	LOCK_CONTENDED(sem, __down_write_trylock, __down_write);
+	rwsem_set_owner(sem);
 }
 
 EXPORT_SYMBOL(down_write_nested);