Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Nicholas Piggin <npiggin@gmail.com>
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     Nicholas Piggin <npiggin@gmail.com>,
        Ingo Molnar <mingo@redhat.com>, Will Deacon <will@kernel.org>,
        Waiman Long <longman@redhat.com>,
        Boqun Feng <boqun.feng@gmail.com>, linux-kernel@vger.kernel.org
Subject: [PATCH 13/13] locking/qspinlock: simplify pv_wait_head_or_lock calling scheme
Date:   Tue,  5 Jul 2022 00:38:20 +1000
Message-Id: <20220704143820.3071004-14-npiggin@gmail.com>
In-Reply-To: <20220704143820.3071004-1-npiggin@gmail.com>
References: <20220704143820.3071004-1-npiggin@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

pv_wait_head_or_lock returns the lock word value ORed with a constant,
which was done to achieve a constant folding compiler optimisation
when the code was generated for both pv and !pv cases. This is no
longer necessary with the explicit paravirt test, so make the calling
convention simpler.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
 kernel/locking/qspinlock.c | 37 +++++++++++++++----------------------
 1 file changed, 15 insertions(+), 22 deletions(-)

diff --git a/kernel/locking/qspinlock.c b/kernel/locking/qspinlock.c
index 19e2f286be0a..97f95bedfa66 100644
--- a/kernel/locking/qspinlock.c
+++ b/kernel/locking/qspinlock.c
@@ -609,8 +609,7 @@ static void pv_kick_node(struct qspinlock *lock, struct qnode *node)
  *
  * The current value of the lock will be returned for additional processing.
  */
-static u32
-pv_wait_head_or_lock(struct qspinlock *lock, struct qnode *node)
+static void pv_wait_head_or_lock(struct qspinlock *lock, struct qnode *node)
 {
 	struct qspinlock **lp = NULL;
 	int waitcnt = 0;
@@ -642,7 +641,7 @@ pv_wait_head_or_lock(struct qspinlock *lock, struct qnode *node)
 		set_pending(lock);
 		for (loop = SPIN_THRESHOLD; loop; loop--) {
 			if (trylock_clear_pending(lock))
-				goto gotlock;
+				return; /* got lock */
 			cpu_relax();
 		}
 		clear_pending(lock);
@@ -670,7 +669,7 @@ pv_wait_head_or_lock(struct qspinlock *lock, struct qnode *node)
 				 */
 				WRITE_ONCE(lock->locked, _Q_LOCKED_VAL);
 				WRITE_ONCE(*lp, NULL);
-				goto gotlock;
+				return; /* got lock */
 			}
 		}
 		WRITE_ONCE(node->state, vcpu_hashed);
@@ -686,12 +685,8 @@ pv_wait_head_or_lock(struct qspinlock *lock, struct qnode *node)
 
 	/*
 	 * The cmpxchg() or xchg() call before coming here provides the
-	 * acquire semantics for locking. The dummy ORing of _Q_LOCKED_VAL
-	 * here is to indicate to the compiler that the value will always
-	 * be nozero to enable better code optimization.
+	 * acquire semantics for locking.
 	 */
-gotlock:
-	return (u32)(atomic_read(&lock->val) | _Q_LOCKED_VAL);
 }
 
 /*
@@ -767,9 +762,8 @@ static __always_inline void pv_wait_node(struct qnode *node,
 					 struct qnode *prev) { }
 static __always_inline void pv_kick_node(struct qspinlock *lock,
 					 struct qnode *node) { }
-static __always_inline u32  pv_wait_head_or_lock(struct qspinlock *lock,
-						 struct qnode *node)
-						   { return 0; }
+static __always_inline void pv_wait_head_or_lock(struct qspinlock *lock,
+						 struct qnode *node) { }
 static __always_inline bool pv_hybrid_queued_unfair_trylock(struct qspinlock *lock) { BUILD_BUG(); }
 #endif /* CONFIG_PARAVIRT_SPINLOCKS */
 
@@ -890,24 +884,23 @@ static inline void queued_spin_lock_mcs_queue(struct qspinlock *lock, bool parav
 	 * sequentiality; this is because the set_locked() function below
 	 * does not imply a full barrier.
 	 *
-	 * The PV pv_wait_head_or_lock function, if active, will acquire
-	 * the lock and return a non-zero value. So we have to skip the
-	 * atomic_cond_read_acquire() call. As the next PV queue head hasn't
-	 * been designated yet, there is no way for the locked value to become
-	 * _Q_SLOW_VAL. So both the set_locked() and the
+	 * The PV pv_wait_head_or_lock function will acquire the lock, so
+	 * skip the atomic_cond_read_acquire() call. As the next PV queue head
+	 * hasn't been designated yet, there is no way for the locked value to
+	 * become _Q_SLOW_VAL. So both the set_locked() and the
 	 * atomic_cmpxchg_relaxed() calls will be safe.
 	 *
 	 * If PV isn't active, 0 will be returned instead.
 	 *
 	 */
 	if (paravirt) {
-		if ((val = pv_wait_head_or_lock(lock, node)))
-			goto locked;
+		pv_wait_head_or_lock(lock, node);
+		val = atomic_read(&lock->val);
+	} else {
+		val = atomic_cond_read_acquire(&lock->val,
+				!(VAL & _Q_LOCKED_PENDING_MASK));
 	}
 
-	val = atomic_cond_read_acquire(&lock->val, !(VAL & _Q_LOCKED_PENDING_MASK));
-
-locked:
 	/*
 	 * claim the lock:
 	 *
-- 
2.35.1