[PATCH 3/7] barrier: a scalable synchonisation barrier

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

This barrier thing is constructed so that it will not write in the sync()
condition (the hot path) when there are no active lock sections; thus avoiding
cacheline bouncing. -- I'm just not sure how this will work out in relation to
PI. We might track those in the barrier scope and boost those by the max prio
of the blockers.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 include/linux/barrier.h |   65 ++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 65 insertions(+)

Index: linux/include/linux/barrier.h
===================================================================
--- /dev/null
+++ linux/include/linux/barrier.h
@@ -0,0 +1,65 @@
+/*
+ * Copyright (C) 2006, Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
+ * Licenced under the GPLv2.
+ *
+ * simple synchonisation barrier
+ *
+ * The sync() operation will wait for completion of all lock sections if any.
+ *
+ * The lock sections are intended to be rare and the sync operation frequent.
+ * This construct is created to be scalable and does only 1 read in the fast
+ * path (sync), hence avoiding cacheline bounces.
+ *
+ * NOTE: it _synchronisation_ only, so if there are serialisation requirements
+ * those must be met by something external to this construct.
+ */
+#ifndef _LINUX_BARRIER_H
+#define _LINUX_BARRIER_H
+
+#ifdef __KERNEL__
+
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <asm/atomic.h>
+
+struct barrier {
+	atomic_t count;
+	wait_queue_head_t wait;
+};
+
+static inline void init_barrier(struct barrier *b)
+{
+	atomic_set(&b->count, 0);
+	init_waitqueue_head(&b->wait);
+	__acquire(b);
+}
+
+static inline void barrier_lock(struct barrier *b)
+{
+	__release(b);
+	atomic_inc(&b->count);
+	smp_wmb();
+}
+
+static inline void barrier_unlock(struct barrier *b)
+{
+	smp_wmb();
+	if (atomic_dec_and_test(&b->count))
+		__wake_up(&b->wait, TASK_INTERRUPTIBLE|TASK_UNINTERRUPTIBLE, 0, b);
+}
+
+static inline void barrier_sync(struct barrier *b)
+{
+	might_sleep();
+
+	if (unlikely(atomic_read(&b->count))) {
+		DEFINE_WAIT(wait);
+		prepare_to_wait(&b->wait, &wait, TASK_UNINTERRUPTIBLE);
+		while (atomic_read(&b->count))
+			schedule();
+		finish_wait(&b->wait, &wait);
+	}
+}
+
+#endif /* __KERNEL__ */
+#endif /* _LINUX_BARRIER_H */

--

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/