locking/pvqspinlock: Only kick CPU at unlock time

For an over-committed guest with more vCPUs than physical CPUs
available, it is possible that a vCPU may be kicked twice before
getting the lock - once before it becomes queue head and once again
before it gets the lock. All these CPU kicking and halting (VMEXIT)
can be expensive and slow down system performance.

This patch adds a new vCPU state (vcpu_hashed) which enables the code
to delay CPU kicking until at unlock time. Once this state is set,
the new lock holder will set _Q_SLOW_VAL and fill in the hash table
on behalf of the halted queue head vCPU. The original vcpu_halted
state will be used by pv_wait_node() only to differentiate other
queue nodes from the qeue head.

Signed-off-by: Waiman Long <Waiman.Long@hp.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Douglas Hatch <doug.hatch@hp.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Scott J Norton <scott.norton@hp.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1436647018-49734-2-git-send-email-Waiman.Long@hp.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

2 files changed, 51 insertions, 21 deletions

diff --git a/kernel/locking/qspinlock.c b/kernel/locking/qspinlock.c
index 38c4920..337c881 100644
--- a/kernel/locking/qspinlock.c
+++ b/kernel/locking/qspinlock.c
@@ -239,8 +239,8 @@ static __always_inline void set_locked(struct qspinlock *lock)
 
 static __always_inline void __pv_init_node(struct mcs_spinlock *node) { }
 static __always_inline void __pv_wait_node(struct mcs_spinlock *node) { }
-static __always_inline void __pv_kick_node(struct mcs_spinlock *node) { }
-
+static __always_inline void __pv_kick_node(struct qspinlock *lock,
+					   struct mcs_spinlock *node) { }
 static __always_inline void __pv_wait_head(struct qspinlock *lock,
 					   struct mcs_spinlock *node) { }
 
@@ -440,7 +440,7 @@ queue:
 		cpu_relax();
 
 	arch_mcs_spin_unlock_contended(&next->locked);
-	pv_kick_node(next);
+	pv_kick_node(lock, next);
 
 release:
 	/*
diff --git a/kernel/locking/qspinlock_paravirt.h b/kernel/locking/qspinlock_paravirt.h
index ab8b1bb..c8e6e9a 100644
--- a/kernel/locking/qspinlock_paravirt.h
+++ b/kernel/locking/qspinlock_paravirt.h
@@ -22,9 +22,14 @@
 
 #define _Q_SLOW_VAL	(3U << _Q_LOCKED_OFFSET)
 
+/*
+ * Queue node uses: vcpu_running & vcpu_halted.
+ * Queue head uses: vcpu_running & vcpu_hashed.
+ */
 enum vcpu_state {
 	vcpu_running = 0,
-	vcpu_halted,
+	vcpu_halted,		/* Used only in pv_wait_node */
+	vcpu_hashed,		/* = pv_hash'ed + vcpu_halted */
 };
 
 struct pv_node {
@@ -153,7 +158,8 @@ static void pv_init_node(struct mcs_spinlock *node)
 
 /*
  * Wait for node->locked to become true, halt the vcpu after a short spin.
- * pv_kick_node() is used to wake the vcpu again.
+ * pv_kick_node() is used to set _Q_SLOW_VAL and fill in hash table on its
+ * behalf.
  */
 static void pv_wait_node(struct mcs_spinlock *node)
 {
@@ -172,9 +178,9 @@ static void pv_wait_node(struct mcs_spinlock *node)
 		 *
 		 * [S] pn->state = vcpu_halted	  [S] next->locked = 1
 		 *     MB			      MB
-		 * [L] pn->locked		[RmW] pn->state = vcpu_running
+		 * [L] pn->locked		[RmW] pn->state = vcpu_hashed
 		 *
-		 * Matches the xchg() from pv_kick_node().
+		 * Matches the cmpxchg() from pv_kick_node().
 		 */
 		smp_store_mb(pn->state, vcpu_halted);
 
@@ -182,9 +188,10 @@ static void pv_wait_node(struct mcs_spinlock *node)
 			pv_wait(&pn->state, vcpu_halted);
 
 		/*
-		 * Reset the vCPU state to avoid unncessary CPU kicking
+		 * If pv_kick_node() changed us to vcpu_hashed, retain that value
+		 * so that pv_wait_head() knows to not also try to hash this lock.
 		 */
-		WRITE_ONCE(pn->state, vcpu_running);
+		cmpxchg(&pn->state, vcpu_halted, vcpu_running);
 
 		/*
 		 * If the locked flag is still not set after wakeup, it is a
@@ -194,6 +201,7 @@ static void pv_wait_node(struct mcs_spinlock *node)
 		 * MCS lock will be released soon.
 		 */
 	}
+
 	/*
 	 * By now our node->locked should be 1 and our caller will not actually
 	 * spin-wait for it. We do however rely on our caller to do a
@@ -202,24 +210,35 @@ static void pv_wait_node(struct mcs_spinlock *node)
 }
 
 /*
- * Called after setting next->locked = 1, used to wake those stuck in
- * pv_wait_node().
+ * Called after setting next->locked = 1 when we're the lock owner.
+ *
+ * Instead of waking the waiters stuck in pv_wait_node() advance their state such
+ * that they're waiting in pv_wait_head(), this avoids a wake/sleep cycle.
  */
-static void pv_kick_node(struct mcs_spinlock *node)
+static void pv_kick_node(struct qspinlock *lock, struct mcs_spinlock *node)
 {
 	struct pv_node *pn = (struct pv_node *)node;
+	struct __qspinlock *l = (void *)lock;
 
 	/*
-	 * Note that because node->locked is already set, this actual
-	 * mcs_spinlock entry could be re-used already.
+	 * If the vCPU is indeed halted, advance its state to match that of
+	 * pv_wait_node(). If OTOH this fails, the vCPU was running and will
+	 * observe its next->locked value and advance itself.
 	 *
-	 * This should be fine however, kicking people for no reason is
-	 * harmless.
+	 * Matches with smp_store_mb() and cmpxchg() in pv_wait_node()
+	 */
+	if (cmpxchg(&pn->state, vcpu_halted, vcpu_hashed) != vcpu_halted)
+		return;
+
+	/*
+	 * Put the lock into the hash table and set the _Q_SLOW_VAL.
 	 *
-	 * See the comment in pv_wait_node().
+	 * As this is the same vCPU that will check the _Q_SLOW_VAL value and
+	 * the hash table later on at unlock time, no atomic instruction is
+	 * needed.
 	 */
-	if (xchg(&pn->state, vcpu_running) == vcpu_halted)
-		pv_kick(pn->cpu);
+	WRITE_ONCE(l->locked, _Q_SLOW_VAL);
+	(void)pv_hash(lock, pn);
 }
 
 /*
@@ -233,6 +252,13 @@ static void pv_wait_head(struct qspinlock *lock, struct mcs_spinlock *node)
 	struct qspinlock **lp = NULL;
 	int loop;
 
+	/*
+	 * If pv_kick_node() already advanced our state, we don't need to
+	 * insert ourselves into the hash table anymore.
+	 */
+	if (READ_ONCE(pn->state) == vcpu_hashed)
+		lp = (struct qspinlock **)1;
+
 	for (;;) {
 		for (loop = SPIN_THRESHOLD; loop; loop--) {
 			if (!READ_ONCE(l->locked))
@@ -240,9 +266,10 @@ static void pv_wait_head(struct qspinlock *lock, struct mcs_spinlock *node)
 			cpu_relax();
 		}
 
-		WRITE_ONCE(pn->state, vcpu_halted);
 		if (!lp) { /* ONCE */
+			WRITE_ONCE(pn->state, vcpu_hashed);
 			lp = pv_hash(lock, pn);
+
 			/*
 			 * We must hash before setting _Q_SLOW_VAL, such that
 			 * when we observe _Q_SLOW_VAL in __pv_queued_spin_unlock()
@@ -333,8 +360,11 @@ __visible void __pv_queued_spin_unlock(struct qspinlock *lock)
 	/*
 	 * At this point the memory pointed at by lock can be freed/reused,
 	 * however we can still use the pv_node to kick the CPU.
+	 * The other vCPU may not really be halted, but kicking an active
+	 * vCPU is harmless other than the additional latency in completing
+	 * the unlock.
 	 */
-	if (READ_ONCE(node->state) == vcpu_halted)
+	if (READ_ONCE(node->state) == vcpu_hashed)
 		pv_kick(node->cpu);
 }
 /*