2019-05-19 20:08:55 +08:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-only
|
2012-12-18 08:01:36 +08:00
|
|
|
#include <linux/atomic.h>
|
2012-12-18 08:01:32 +08:00
|
|
|
#include <linux/percpu.h>
|
2019-10-31 03:30:41 +08:00
|
|
|
#include <linux/wait.h>
|
2012-12-18 08:01:38 +08:00
|
|
|
#include <linux/lockdep.h>
|
2012-12-18 08:01:32 +08:00
|
|
|
#include <linux/percpu-rwsem.h>
|
|
|
|
#include <linux/rcupdate.h>
|
|
|
|
#include <linux/sched.h>
|
2019-10-31 03:30:41 +08:00
|
|
|
#include <linux/sched/task.h>
|
2012-12-18 08:01:32 +08:00
|
|
|
#include <linux/errno.h>
|
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
int __percpu_init_rwsem(struct percpu_rw_semaphore *sem,
|
2019-10-31 03:01:26 +08:00
|
|
|
const char *name, struct lock_class_key *key)
|
2012-12-18 08:01:32 +08:00
|
|
|
{
|
2016-07-15 02:08:46 +08:00
|
|
|
sem->read_count = alloc_percpu(int);
|
|
|
|
if (unlikely(!sem->read_count))
|
2012-12-18 08:01:32 +08:00
|
|
|
return -ENOMEM;
|
|
|
|
|
2019-04-23 20:07:24 +08:00
|
|
|
rcu_sync_init(&sem->rss);
|
2017-01-11 23:22:26 +08:00
|
|
|
rcuwait_init(&sem->writer);
|
2019-10-31 03:30:41 +08:00
|
|
|
init_waitqueue_head(&sem->waiters);
|
|
|
|
atomic_set(&sem->block, 0);
|
2019-10-31 03:01:26 +08:00
|
|
|
#ifdef CONFIG_DEBUG_LOCK_ALLOC
|
|
|
|
debug_check_no_locks_freed((void *)sem, sizeof(*sem));
|
|
|
|
lockdep_init_map(&sem->dep_map, name, key, 0);
|
|
|
|
#endif
|
2012-12-18 08:01:32 +08:00
|
|
|
return 0;
|
|
|
|
}
|
2015-09-01 11:21:59 +08:00
|
|
|
EXPORT_SYMBOL_GPL(__percpu_init_rwsem);
|
2012-12-18 08:01:32 +08:00
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
void percpu_free_rwsem(struct percpu_rw_semaphore *sem)
|
2012-12-18 08:01:32 +08:00
|
|
|
{
|
2015-08-22 01:42:55 +08:00
|
|
|
/*
|
|
|
|
* XXX: temporary kludge. The error path in alloc_super()
|
|
|
|
* assumes that percpu_free_rwsem() is safe after kzalloc().
|
|
|
|
*/
|
2016-07-15 02:08:46 +08:00
|
|
|
if (!sem->read_count)
|
2015-08-22 01:42:55 +08:00
|
|
|
return;
|
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
rcu_sync_dtor(&sem->rss);
|
|
|
|
free_percpu(sem->read_count);
|
|
|
|
sem->read_count = NULL; /* catch use after free bugs */
|
2012-12-18 08:01:32 +08:00
|
|
|
}
|
2016-04-26 11:22:35 +08:00
|
|
|
EXPORT_SYMBOL_GPL(percpu_free_rwsem);
|
2012-12-18 08:01:32 +08:00
|
|
|
|
2019-10-31 19:34:23 +08:00
|
|
|
static bool __percpu_down_read_trylock(struct percpu_rw_semaphore *sem)
|
2012-12-18 08:01:32 +08:00
|
|
|
{
|
2020-09-15 22:07:50 +08:00
|
|
|
this_cpu_inc(*sem->read_count);
|
2019-10-31 03:17:51 +08:00
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
/*
|
|
|
|
* Due to having preemption disabled the decrement happens on
|
|
|
|
* the same CPU as the increment, avoiding the
|
|
|
|
* increment-on-one-CPU-and-decrement-on-another problem.
|
|
|
|
*
|
2019-10-31 03:30:41 +08:00
|
|
|
* If the reader misses the writer's assignment of sem->block, then the
|
|
|
|
* writer is guaranteed to see the reader's increment.
|
2016-07-15 02:08:46 +08:00
|
|
|
*
|
|
|
|
* Conversely, any readers that increment their sem->read_count after
|
2019-10-31 03:30:41 +08:00
|
|
|
* the writer looks are guaranteed to see the sem->block value, which
|
|
|
|
* in turn means that they are guaranteed to immediately decrement
|
|
|
|
* their sem->read_count, so that it doesn't matter that the writer
|
|
|
|
* missed them.
|
2016-07-15 02:08:46 +08:00
|
|
|
*/
|
2012-12-18 08:01:32 +08:00
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
smp_mb(); /* A matches D */
|
2012-12-18 08:01:32 +08:00
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
/*
|
2019-10-31 03:30:41 +08:00
|
|
|
* If !sem->block the critical section starts here, matched by the
|
2016-07-15 02:08:46 +08:00
|
|
|
* release in percpu_up_write().
|
|
|
|
*/
|
2019-10-31 03:30:41 +08:00
|
|
|
if (likely(!atomic_read_acquire(&sem->block)))
|
2019-10-31 03:12:37 +08:00
|
|
|
return true;
|
2012-12-18 08:01:32 +08:00
|
|
|
|
2020-09-15 22:07:50 +08:00
|
|
|
this_cpu_dec(*sem->read_count);
|
2019-10-31 19:34:23 +08:00
|
|
|
|
|
|
|
/* Prod writer to re-evaluate readers_active_check() */
|
|
|
|
rcuwait_wake_up(&sem->writer);
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2019-10-31 03:30:41 +08:00
|
|
|
static inline bool __percpu_down_write_trylock(struct percpu_rw_semaphore *sem)
|
|
|
|
{
|
|
|
|
if (atomic_read(&sem->block))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
return atomic_xchg(&sem->block, 1) == 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static bool __percpu_rwsem_trylock(struct percpu_rw_semaphore *sem, bool reader)
|
|
|
|
{
|
|
|
|
if (reader) {
|
|
|
|
bool ret;
|
|
|
|
|
|
|
|
preempt_disable();
|
|
|
|
ret = __percpu_down_read_trylock(sem);
|
|
|
|
preempt_enable();
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
return __percpu_down_write_trylock(sem);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The return value of wait_queue_entry::func means:
|
|
|
|
*
|
|
|
|
* <0 - error, wakeup is terminated and the error is returned
|
|
|
|
* 0 - no wakeup, a next waiter is tried
|
|
|
|
* >0 - woken, if EXCLUSIVE, counted towards @nr_exclusive.
|
|
|
|
*
|
|
|
|
* We use EXCLUSIVE for both readers and writers to preserve FIFO order,
|
|
|
|
* and play games with the return value to allow waking multiple readers.
|
|
|
|
*
|
|
|
|
* Specifically, we wake readers until we've woken a single writer, or until a
|
|
|
|
* trylock fails.
|
|
|
|
*/
|
|
|
|
static int percpu_rwsem_wake_function(struct wait_queue_entry *wq_entry,
|
|
|
|
unsigned int mode, int wake_flags,
|
|
|
|
void *key)
|
|
|
|
{
|
|
|
|
bool reader = wq_entry->flags & WQ_FLAG_CUSTOM;
|
|
|
|
struct percpu_rw_semaphore *sem = key;
|
2020-03-31 05:30:02 +08:00
|
|
|
struct task_struct *p;
|
2019-10-31 03:30:41 +08:00
|
|
|
|
|
|
|
/* concurrent against percpu_down_write(), can get stolen */
|
|
|
|
if (!__percpu_rwsem_trylock(sem, reader))
|
|
|
|
return 1;
|
|
|
|
|
2020-03-31 05:30:02 +08:00
|
|
|
p = get_task_struct(wq_entry->private);
|
2019-10-31 03:30:41 +08:00
|
|
|
list_del_init(&wq_entry->entry);
|
|
|
|
smp_store_release(&wq_entry->private, NULL);
|
|
|
|
|
|
|
|
wake_up_process(p);
|
|
|
|
put_task_struct(p);
|
|
|
|
|
|
|
|
return !reader; /* wake (readers until) 1 writer */
|
|
|
|
}
|
|
|
|
|
|
|
|
static void percpu_rwsem_wait(struct percpu_rw_semaphore *sem, bool reader)
|
|
|
|
{
|
|
|
|
DEFINE_WAIT_FUNC(wq_entry, percpu_rwsem_wake_function);
|
|
|
|
bool wait;
|
|
|
|
|
|
|
|
spin_lock_irq(&sem->waiters.lock);
|
|
|
|
/*
|
|
|
|
* Serialize against the wakeup in percpu_up_write(), if we fail
|
|
|
|
* the trylock, the wakeup must see us on the list.
|
|
|
|
*/
|
|
|
|
wait = !__percpu_rwsem_trylock(sem, reader);
|
|
|
|
if (wait) {
|
|
|
|
wq_entry.flags |= WQ_FLAG_EXCLUSIVE | reader * WQ_FLAG_CUSTOM;
|
|
|
|
__add_wait_queue_entry_tail(&sem->waiters, &wq_entry);
|
|
|
|
}
|
|
|
|
spin_unlock_irq(&sem->waiters.lock);
|
|
|
|
|
|
|
|
while (wait) {
|
|
|
|
set_current_state(TASK_UNINTERRUPTIBLE);
|
|
|
|
if (!smp_load_acquire(&wq_entry.private))
|
|
|
|
break;
|
|
|
|
schedule();
|
|
|
|
}
|
|
|
|
__set_current_state(TASK_RUNNING);
|
|
|
|
}
|
|
|
|
|
2019-10-31 19:34:23 +08:00
|
|
|
bool __percpu_down_read(struct percpu_rw_semaphore *sem, bool try)
|
|
|
|
{
|
|
|
|
if (__percpu_down_read_trylock(sem))
|
|
|
|
return true;
|
2015-08-22 01:43:03 +08:00
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
if (try)
|
2019-10-31 03:12:37 +08:00
|
|
|
return false;
|
2012-12-18 08:01:32 +08:00
|
|
|
|
2019-10-31 03:30:41 +08:00
|
|
|
preempt_enable();
|
|
|
|
percpu_rwsem_wait(sem, /* .reader = */ true);
|
2016-07-15 02:08:46 +08:00
|
|
|
preempt_disable();
|
2019-10-31 03:30:41 +08:00
|
|
|
|
2019-10-31 03:12:37 +08:00
|
|
|
return true;
|
2015-07-21 23:45:57 +08:00
|
|
|
}
|
2016-07-15 02:08:46 +08:00
|
|
|
EXPORT_SYMBOL_GPL(__percpu_down_read);
|
2015-07-21 23:45:57 +08:00
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
#define per_cpu_sum(var) \
|
|
|
|
({ \
|
|
|
|
typeof(var) __sum = 0; \
|
|
|
|
int cpu; \
|
|
|
|
compiletime_assert_atomic_type(__sum); \
|
|
|
|
for_each_possible_cpu(cpu) \
|
|
|
|
__sum += per_cpu(var, cpu); \
|
|
|
|
__sum; \
|
|
|
|
})
|
2012-12-18 08:01:32 +08:00
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
/*
|
|
|
|
* Return true if the modular sum of the sem->read_count per-CPU variable is
|
|
|
|
* zero. If this sum is zero, then it is stable due to the fact that if any
|
|
|
|
* newly arriving readers increment a given counter, they will immediately
|
|
|
|
* decrement that same counter.
|
2019-10-31 03:30:41 +08:00
|
|
|
*
|
|
|
|
* Assumes sem->block is set.
|
2016-07-15 02:08:46 +08:00
|
|
|
*/
|
|
|
|
static bool readers_active_check(struct percpu_rw_semaphore *sem)
|
2012-12-18 08:01:32 +08:00
|
|
|
{
|
2016-07-15 02:08:46 +08:00
|
|
|
if (per_cpu_sum(*sem->read_count) != 0)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we observed the decrement; ensure we see the entire critical
|
|
|
|
* section.
|
|
|
|
*/
|
2012-12-18 08:01:32 +08:00
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
smp_mb(); /* C matches B */
|
2012-12-18 08:01:32 +08:00
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
return true;
|
2012-12-18 08:01:32 +08:00
|
|
|
}
|
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
void percpu_down_write(struct percpu_rw_semaphore *sem)
|
2012-12-18 08:01:32 +08:00
|
|
|
{
|
2020-01-08 09:33:04 +08:00
|
|
|
might_sleep();
|
2019-10-31 03:01:26 +08:00
|
|
|
rwsem_acquire(&sem->dep_map, 0, 0, _RET_IP_);
|
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
/* Notify readers to take the slow path. */
|
|
|
|
rcu_sync_enter(&sem->rss);
|
|
|
|
|
2012-12-18 08:01:32 +08:00
|
|
|
/*
|
2019-10-31 03:30:41 +08:00
|
|
|
* Try set sem->block; this provides writer-writer exclusion.
|
|
|
|
* Having sem->block set makes new readers block.
|
2012-12-18 08:01:32 +08:00
|
|
|
*/
|
2019-10-31 03:30:41 +08:00
|
|
|
if (!__percpu_down_write_trylock(sem))
|
|
|
|
percpu_rwsem_wait(sem, /* .reader = */ false);
|
2012-12-18 08:01:32 +08:00
|
|
|
|
2019-10-31 03:30:41 +08:00
|
|
|
/* smp_mb() implied by __percpu_down_write_trylock() on success -- D matches A */
|
2012-12-18 08:01:36 +08:00
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
/*
|
2019-10-31 03:30:41 +08:00
|
|
|
* If they don't see our store of sem->block, then we are guaranteed to
|
|
|
|
* see their sem->read_count increment, and therefore will wait for
|
|
|
|
* them.
|
2016-07-15 02:08:46 +08:00
|
|
|
*/
|
2012-12-18 08:01:32 +08:00
|
|
|
|
2019-10-31 03:30:41 +08:00
|
|
|
/* Wait for all active readers to complete. */
|
2020-03-21 19:25:55 +08:00
|
|
|
rcuwait_wait_event(&sem->writer, readers_active_check(sem), TASK_UNINTERRUPTIBLE);
|
2012-12-18 08:01:32 +08:00
|
|
|
}
|
2015-09-01 11:21:59 +08:00
|
|
|
EXPORT_SYMBOL_GPL(percpu_down_write);
|
2012-12-18 08:01:32 +08:00
|
|
|
|
2016-07-15 02:08:46 +08:00
|
|
|
void percpu_up_write(struct percpu_rw_semaphore *sem)
|
2012-12-18 08:01:32 +08:00
|
|
|
{
|
2019-10-31 03:01:26 +08:00
|
|
|
rwsem_release(&sem->dep_map, _RET_IP_);
|
|
|
|
|
2012-12-18 08:01:32 +08:00
|
|
|
/*
|
2016-07-15 02:08:46 +08:00
|
|
|
* Signal the writer is done, no fast path yet.
|
|
|
|
*
|
|
|
|
* One reason that we cannot just immediately flip to readers_fast is
|
|
|
|
* that new readers might fail to see the results of this writer's
|
|
|
|
* critical section.
|
|
|
|
*
|
|
|
|
* Therefore we force it through the slow path which guarantees an
|
|
|
|
* acquire and thereby guarantees the critical section's consistency.
|
|
|
|
*/
|
2019-10-31 03:30:41 +08:00
|
|
|
atomic_set_release(&sem->block, 0);
|
2016-07-15 02:08:46 +08:00
|
|
|
|
|
|
|
/*
|
2019-10-31 03:30:41 +08:00
|
|
|
* Prod any pending reader/writer to make progress.
|
2016-07-15 02:08:46 +08:00
|
|
|
*/
|
2019-10-31 03:30:41 +08:00
|
|
|
__wake_up(&sem->waiters, TASK_NORMAL, 1, sem);
|
2016-07-15 02:08:46 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Once this completes (at least one RCU-sched grace period hence) the
|
|
|
|
* reader fast path will be available again. Safe to use outside the
|
|
|
|
* exclusive write lock because its counting.
|
2012-12-18 08:01:32 +08:00
|
|
|
*/
|
2016-07-15 02:08:46 +08:00
|
|
|
rcu_sync_exit(&sem->rss);
|
2012-12-18 08:01:32 +08:00
|
|
|
}
|
2015-09-01 11:21:59 +08:00
|
|
|
EXPORT_SYMBOL_GPL(percpu_up_write);
|