blockdev: turn a rw semaphore into a percpu rw semaphore

This avoids cache line bouncing when many processes lock the semaphore
for read.

New percpu lock implementation

The lock consists of an array of percpu unsigned integers, a boolean
variable and a mutex.

When we take the lock for read, we enter rcu read section, check for a
"locked" variable. If it is false, we increase a percpu counter on the
current cpu and exit the rcu section. If "locked" is true, we exit the
rcu section, take the mutex and drop it (this waits until a writer
finished) and retry.

Unlocking for read just decreases percpu variable. Note that we can
unlock on a difference cpu than where we locked, in this case the
counter underflows. The sum of all percpu counters represents the number
of processes that hold the lock for read.

When we need to lock for write, we take the mutex, set "locked" variable
to true and synchronize rcu. Since RCU has been synchronized, no
processes can create new read locks. We wait until the sum of percpu
counters is zero - when it is, there are no readers in the critical
section.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 89 insertions, 0 deletions

diff --git a/include/linux/percpu-rwsem.h b/include/linux/percpu-rwsem.h
new file mode 100644
index 000000000000..cf80f7e5277f
--- /dev/null
+++ b/include/linux/percpu-rwsem.h
@@ -0,0 +1,89 @@
+#ifndef _LINUX_PERCPU_RWSEM_H
+#define _LINUX_PERCPU_RWSEM_H
+
+#include <linux/mutex.h>
+#include <linux/percpu.h>
+#include <linux/rcupdate.h>
+#include <linux/delay.h>
+
+struct percpu_rw_semaphore {
+	unsigned __percpu *counters;
+	bool locked;
+	struct mutex mtx;
+};
+
+static inline void percpu_down_read(struct percpu_rw_semaphore *p)
+{
+	rcu_read_lock();
+	if (unlikely(p->locked)) {
+		rcu_read_unlock();
+		mutex_lock(&p->mtx);
+		this_cpu_inc(*p->counters);
+		mutex_unlock(&p->mtx);
+		return;
+	}
+	this_cpu_inc(*p->counters);
+	rcu_read_unlock();
+}
+
+static inline void percpu_up_read(struct percpu_rw_semaphore *p)
+{
+	/*
+	 * On X86, write operation in this_cpu_dec serves as a memory unlock
+	 * barrier (i.e. memory accesses may be moved before the write, but
+	 * no memory accesses are moved past the write).
+	 * On other architectures this may not be the case, so we need smp_mb()
+	 * there.
+	 */
+#if defined(CONFIG_X86) && (!defined(CONFIG_X86_PPRO_FENCE) && !defined(CONFIG_X86_OOSTORE))
+	barrier();
+#else
+	smp_mb();
+#endif
+	this_cpu_dec(*p->counters);
+}
+
+static inline unsigned __percpu_count(unsigned __percpu *counters)
+{
+	unsigned total = 0;
+	int cpu;
+
+	for_each_possible_cpu(cpu)
+		total += ACCESS_ONCE(*per_cpu_ptr(counters, cpu));
+
+	return total;
+}
+
+static inline void percpu_down_write(struct percpu_rw_semaphore *p)
+{
+	mutex_lock(&p->mtx);
+	p->locked = true;
+	synchronize_rcu();
+	while (__percpu_count(p->counters))
+		msleep(1);
+	smp_rmb(); /* paired with smp_mb() in percpu_sem_up_read() */
+}
+
+static inline void percpu_up_write(struct percpu_rw_semaphore *p)
+{
+	p->locked = false;
+	mutex_unlock(&p->mtx);
+}
+
+static inline int percpu_init_rwsem(struct percpu_rw_semaphore *p)
+{
+	p->counters = alloc_percpu(unsigned);
+	if (unlikely(!p->counters))
+		return -ENOMEM;
+	p->locked = false;
+	mutex_init(&p->mtx);
+	return 0;
+}
+
+static inline void percpu_free_rwsem(struct percpu_rw_semaphore *p)
+{
+	free_percpu(p->counters);
+	p->counters = NULL; /* catch use after free bugs */
+}
+
+#endif