From: Thomas Gleixner <tglx@linutronix.de>
Date: Thu, 12 Oct 2017 17:28:34 +0200
Subject: rtmutex: add rwsem implementation based on rtmutex

The RT specific R/W semaphore implementation restricts the number of readers
to one because a writer cannot block on multiple readers and inherit its
priority or budget.

The single reader restricting is painful in various ways:

 - Performance bottleneck for multi-threaded applications in the page fault
   path (mmap sem)

 - Progress blocker for drivers which are carefully crafted to avoid the
   potential reader/writer deadlock in mainline.

The analysis of the writer code pathes shows, that properly written RT tasks
should not take them. Syscalls like mmap(), file access which take mmap sem
write locked have unbound latencies which are completely unrelated to mmap
sem. Other R/W sem users like graphics drivers are not suitable for RT tasks
either.

So there is little risk to hurt RT tasks when the RT rwsem implementation is
changed in the following way:

 - Allow concurrent readers

 - Make writers block until the last reader left the critical section. This
   blocking is not subject to priority/budget inheritance.

 - Readers blocked on a writer inherit their priority/budget in the normal
   way.

There is a drawback with this scheme. R/W semaphores become writer unfair
though the applications which have triggered writer starvation (mostly on
mmap_sem) in the past are not really the typical workloads running on a RT
system. So while it's unlikely to hit writer starvation, it's possible. If
there are unexpected workloads on RT systems triggering it, we need to rethink
the approach.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
 include/linux/rwsem_rt.h  |   68 ++++++++++
 kernel/locking/rwsem-rt.c |  293 ++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 361 insertions(+)
 create mode 100644 include/linux/rwsem_rt.h
 create mode 100644 kernel/locking/rwsem-rt.c

--- /dev/null
+++ b/include/linux/rwsem_rt.h
@@ -0,0 +1,68 @@
+#ifndef _LINUX_RWSEM_RT_H
+#define _LINUX_RWSEM_RT_H
+
+#ifndef _LINUX_RWSEM_H
+#error "Include rwsem.h"
+#endif
+
+#include <linux/rtmutex.h>
+#include <linux/swait.h>
+
+#define READER_BIAS		(1U << 31)
+#define WRITER_BIAS		(1U << 30)
+
+struct rw_semaphore {
+	atomic_t		readers;
+	struct rt_mutex		rtmutex;
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+	struct lockdep_map	dep_map;
+#endif
+};
+
+#define __RWSEM_INITIALIZER(name)				\
+{								\
+	.readers = ATOMIC_INIT(READER_BIAS),			\
+	.rtmutex = __RT_MUTEX_INITIALIZER(name.rtmutex),	\
+	RW_DEP_MAP_INIT(name)					\
+}
+
+#define DECLARE_RWSEM(lockname) \
+	struct rw_semaphore lockname = __RWSEM_INITIALIZER(lockname)
+
+extern void  __rwsem_init(struct rw_semaphore *rwsem, const char *name,
+			  struct lock_class_key *key);
+
+#define __init_rwsem(sem, name, key)			\
+do {							\
+		rt_mutex_init(&(sem)->rtmutex);		\
+		__rwsem_init((sem), (name), (key));	\
+} while (0)
+
+#define init_rwsem(sem)					\
+do {							\
+	static struct lock_class_key __key;		\
+							\
+	__init_rwsem((sem), #sem, &__key);		\
+} while (0)
+
+static inline int rwsem_is_locked(struct rw_semaphore *sem)
+{
+	return atomic_read(&sem->readers) != READER_BIAS;
+}
+
+static inline int rwsem_is_contended(struct rw_semaphore *sem)
+{
+	return atomic_read(&sem->readers) > 0;
+}
+
+extern void __down_read(struct rw_semaphore *sem);
+extern int __down_read_killable(struct rw_semaphore *sem);
+extern int __down_read_trylock(struct rw_semaphore *sem);
+extern void __down_write(struct rw_semaphore *sem);
+extern int __must_check __down_write_killable(struct rw_semaphore *sem);
+extern int __down_write_trylock(struct rw_semaphore *sem);
+extern void __up_read(struct rw_semaphore *sem);
+extern void __up_write(struct rw_semaphore *sem);
+extern void __downgrade_write(struct rw_semaphore *sem);
+
+#endif
--- /dev/null
+++ b/kernel/locking/rwsem-rt.c
@@ -0,0 +1,293 @@
+/*
+ */
+#include <linux/rwsem.h>
+#include <linux/sched/debug.h>
+#include <linux/sched/signal.h>
+#include <linux/export.h>
+
+#include "rtmutex_common.h"
+
+/*
+ * RT-specific reader/writer semaphores
+ *
+ * down_write()
+ *  1) Lock sem->rtmutex
+ *  2) Remove the reader BIAS to force readers into the slow path
+ *  3) Wait until all readers have left the critical region
+ *  4) Mark it write locked
+ *
+ * up_write()
+ *  1) Remove the write locked marker
+ *  2) Set the reader BIAS so readers can use the fast path again
+ *  3) Unlock sem->rtmutex to release blocked readers
+ *
+ * down_read()
+ *  1) Try fast path acquisition (reader BIAS is set)
+ *  2) Take sem->rtmutex.wait_lock which protects the writelocked flag
+ *  3) If !writelocked, acquire it for read
+ *  4) If writelocked, block on sem->rtmutex
+ *  5) unlock sem->rtmutex, goto 1)
+ *
+ * up_read()
+ *  1) Try fast path release (reader count != 1)
+ *  2) Wake the writer waiting in down_write()#3
+ *
+ * down_read()#3 has the consequence, that rw semaphores on RT are not writer
+ * fair, but writers, which should be avoided in RT tasks (think mmap_sem),
+ * are subject to the rtmutex priority/DL inheritance mechanism.
+ *
+ * It's possible to make the rw semaphores writer fair by keeping a list of
+ * active readers. A blocked writer would force all newly incoming readers to
+ * block on the rtmutex, but the rtmutex would have to be proxy locked for one
+ * reader after the other. We can't use multi-reader inheritance because there
+ * is no way to support that with SCHED_DEADLINE. Implementing the one by one
+ * reader boosting/handover mechanism is a major surgery for a very dubious
+ * value.
+ *
+ * The risk of writer starvation is there, but the pathological use cases
+ * which trigger it are not necessarily the typical RT workloads.
+ */
+
+void __rwsem_init(struct rw_semaphore *sem, const char *name,
+		  struct lock_class_key *key)
+{
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+	/*
+	 * Make sure we are not reinitializing a held semaphore:
+	 */
+	debug_check_no_locks_freed((void *)sem, sizeof(*sem));
+	lockdep_init_map(&sem->dep_map, name, key, 0);
+#endif
+	atomic_set(&sem->readers, READER_BIAS);
+}
+EXPORT_SYMBOL(__rwsem_init);
+
+int __down_read_trylock(struct rw_semaphore *sem)
+{
+	int r, old;
+
+	/*
+	 * Increment reader count, if sem->readers < 0, i.e. READER_BIAS is
+	 * set.
+	 */
+	for (r = atomic_read(&sem->readers); r < 0;) {
+		old = atomic_cmpxchg(&sem->readers, r, r + 1);
+		if (likely(old == r))
+			return 1;
+		r = old;
+	}
+	return 0;
+}
+
+static int __sched __down_read_common(struct rw_semaphore *sem, int state)
+{
+	struct rt_mutex *m = &sem->rtmutex;
+	struct rt_mutex_waiter waiter;
+	int ret;
+
+	if (__down_read_trylock(sem))
+		return 0;
+
+	might_sleep();
+	raw_spin_lock_irq(&m->wait_lock);
+	/*
+	 * Allow readers as long as the writer has not completely
+	 * acquired the semaphore for write.
+	 */
+	if (atomic_read(&sem->readers) != WRITER_BIAS) {
+		atomic_inc(&sem->readers);
+		raw_spin_unlock_irq(&m->wait_lock);
+		return 0;
+	}
+
+	/*
+	 * Call into the slow lock path with the rtmutex->wait_lock
+	 * held, so this can't result in the following race:
+	 *
+	 * Reader1		Reader2		Writer
+	 *			down_read()
+	 *					down_write()
+	 *					rtmutex_lock(m)
+	 *					swait()
+	 * down_read()
+	 * unlock(m->wait_lock)
+	 *			up_read()
+	 *			swake()
+	 *					lock(m->wait_lock)
+	 *					sem->writelocked=true
+	 *					unlock(m->wait_lock)
+	 *
+	 *					up_write()
+	 *					sem->writelocked=false
+	 *					rtmutex_unlock(m)
+	 *			down_read()
+	 *					down_write()
+	 *					rtmutex_lock(m)
+	 *					swait()
+	 * rtmutex_lock(m)
+	 *
+	 * That would put Reader1 behind the writer waiting on
+	 * Reader2 to call up_read() which might be unbound.
+	 */
+	rt_mutex_init_waiter(&waiter, false);
+	ret = rt_mutex_slowlock_locked(m, state, NULL, RT_MUTEX_MIN_CHAINWALK,
+				       &waiter);
+	/*
+	 * The slowlock() above is guaranteed to return with the rtmutex (for
+	 * ret = 0) is now held, so there can't be a writer active. Increment
+	 * the reader count and immediately drop the rtmutex again.
+	 * For ret != 0 we don't hold the rtmutex and need unlock the wait_lock.
+	 * We don't own the lock then.
+	 */
+	if (!ret)
+		atomic_inc(&sem->readers);
+	raw_spin_unlock_irq(&m->wait_lock);
+	if (!ret)
+		__rt_mutex_unlock(m);
+
+	debug_rt_mutex_free_waiter(&waiter);
+	return ret;
+}
+
+void __down_read(struct rw_semaphore *sem)
+{
+	int ret;
+
+	ret = __down_read_common(sem, TASK_UNINTERRUPTIBLE);
+	WARN_ON_ONCE(ret);
+}
+
+int __down_read_killable(struct rw_semaphore *sem)
+{
+	int ret;
+
+	ret = __down_read_common(sem, TASK_KILLABLE);
+	if (likely(!ret))
+		return ret;
+	WARN_ONCE(ret != -EINTR, "Unexpected state: %d\n", ret);
+	return -EINTR;
+}
+
+void __up_read(struct rw_semaphore *sem)
+{
+	struct rt_mutex *m = &sem->rtmutex;
+	struct task_struct *tsk;
+
+	/*
+	 * sem->readers can only hit 0 when a writer is waiting for the
+	 * active readers to leave the critical region.
+	 */
+	if (!atomic_dec_and_test(&sem->readers))
+		return;
+
+	might_sleep();
+	raw_spin_lock_irq(&m->wait_lock);
+	/*
+	 * Wake the writer, i.e. the rtmutex owner. It might release the
+	 * rtmutex concurrently in the fast path (due to a signal), but to
+	 * clean up the rwsem it needs to acquire m->wait_lock. The worst
+	 * case which can happen is a spurious wakeup.
+	 */
+	tsk = rt_mutex_owner(m);
+	if (tsk)
+		wake_up_process(tsk);
+
+	raw_spin_unlock_irq(&m->wait_lock);
+}
+
+static void __up_write_unlock(struct rw_semaphore *sem, int bias,
+			      unsigned long flags)
+{
+	struct rt_mutex *m = &sem->rtmutex;
+
+	atomic_add(READER_BIAS - bias, &sem->readers);
+	raw_spin_unlock_irqrestore(&m->wait_lock, flags);
+	__rt_mutex_unlock(m);
+}
+
+static int __sched __down_write_common(struct rw_semaphore *sem, int state)
+{
+	struct rt_mutex *m = &sem->rtmutex;
+	unsigned long flags;
+
+	/* Take the rtmutex as a first step */
+	if (__rt_mutex_lock_state(m, state))
+		return -EINTR;
+
+	/* Force readers into slow path */
+	atomic_sub(READER_BIAS, &sem->readers);
+	might_sleep();
+
+	set_current_state(state);
+	for (;;) {
+		raw_spin_lock_irqsave(&m->wait_lock, flags);
+		/* Have all readers left the critical region? */
+		if (!atomic_read(&sem->readers)) {
+			atomic_set(&sem->readers, WRITER_BIAS);
+			__set_current_state(TASK_RUNNING);
+			raw_spin_unlock_irqrestore(&m->wait_lock, flags);
+			return 0;
+		}
+
+		if (signal_pending_state(state, current)) {
+			__set_current_state(TASK_RUNNING);
+			__up_write_unlock(sem, 0, flags);
+			return -EINTR;
+		}
+		raw_spin_unlock_irqrestore(&m->wait_lock, flags);
+
+		if (atomic_read(&sem->readers) != 0) {
+			schedule();
+			set_current_state(state);
+		}
+	}
+}
+
+void __sched __down_write(struct rw_semaphore *sem)
+{
+	__down_write_common(sem, TASK_UNINTERRUPTIBLE);
+}
+
+int __sched __down_write_killable(struct rw_semaphore *sem)
+{
+	return __down_write_common(sem, TASK_KILLABLE);
+}
+
+int __down_write_trylock(struct rw_semaphore *sem)
+{
+	struct rt_mutex *m = &sem->rtmutex;
+	unsigned long flags;
+
+	if (!__rt_mutex_trylock(m))
+		return 0;
+
+	atomic_sub(READER_BIAS, &sem->readers);
+
+	raw_spin_lock_irqsave(&m->wait_lock, flags);
+	if (!atomic_read(&sem->readers)) {
+		atomic_set(&sem->readers, WRITER_BIAS);
+		raw_spin_unlock_irqrestore(&m->wait_lock, flags);
+		return 1;
+	}
+	__up_write_unlock(sem, 0, flags);
+	return 0;
+}
+
+void __up_write(struct rw_semaphore *sem)
+{
+	struct rt_mutex *m = &sem->rtmutex;
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&m->wait_lock, flags);
+	__up_write_unlock(sem, WRITER_BIAS, flags);
+}
+
+void __downgrade_write(struct rw_semaphore *sem)
+{
+	struct rt_mutex *m = &sem->rtmutex;
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&m->wait_lock, flags);
+	/* Release it and account current as reader */
+	__up_write_unlock(sem, WRITER_BIAS - 1, flags);
+}