linux-stable/kernel/locking/rwbase_rt.c
Sebastian Andrzej Siewior 286deb7ec0 locking/rwbase: Mitigate indefinite writer starvation
On PREEMPT_RT, rw_semaphore and rwlock_t locks are unfair to writers.
Readers can indefinitely acquire the lock unless the writer fully acquired
the lock, which might never happen if there is always a reader in the
critical section owning the lock.

Mel Gorman reported that since LTP-20220121 the dio_truncate test case
went from having 1 reader to having 16 readers and that number of readers
is sufficient to prevent the down_write ever succeeding while readers
exist. Eventually the test is killed after 30 minutes as a failure.

Mel proposed a timeout to limit how long a writer can be blocked until
the reader is forced into the slowpath.

Thomas argued that there is no added value by providing this timeout.  From
a PREEMPT_RT point of view, there are no critical rw_semaphore or rwlock_t
locks left where the reader must be preferred.

Mitigate indefinite writer starvation by forcing the READER into the
slowpath once the WRITER attempts to acquire the lock.

Reported-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Link: https://lore.kernel.org/877cwbq4cq.ffs@tglx
Link: https://lore.kernel.org/r/20230321161140.HMcQEhHb@linutronix.de
Cc: Linus Torvalds <torvalds@linux-foundation.org>
2023-04-29 09:08:52 +02:00

289 lines
8 KiB
C

// SPDX-License-Identifier: GPL-2.0-only
/*
* RT-specific reader/writer semaphores and reader/writer locks
*
* down_write/write_lock()
* 1) Lock rtmutex
* 2) Remove the reader BIAS to force readers into the slow path
* 3) Wait until all readers have left the critical section
* 4) Mark it write locked
*
* up_write/write_unlock()
* 1) Remove the write locked marker
* 2) Set the reader BIAS, so readers can use the fast path again
* 3) Unlock rtmutex, to release blocked readers
*
* down_read/read_lock()
* 1) Try fast path acquisition (reader BIAS is set)
* 2) Take tmutex::wait_lock, which protects the writelocked flag
* 3) If !writelocked, acquire it for read
* 4) If writelocked, block on tmutex
* 5) unlock rtmutex, goto 1)
*
* up_read/read_unlock()
* 1) Try fast path release (reader count != 1)
* 2) Wake the writer waiting in down_write()/write_lock() #3
*
* down_read/read_lock()#3 has the consequence, that rw semaphores and rw
* locks 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 primitives 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.
*
* Fast-path orderings:
* The lock/unlock of readers can run in fast paths: lock and unlock are only
* atomic ops, and there is no inner lock to provide ACQUIRE and RELEASE
* semantics of rwbase_rt. Atomic ops should thus provide _acquire()
* and _release() (or stronger).
*
* Common code shared between RT rw_semaphore and rwlock
*/
static __always_inline int rwbase_read_trylock(struct rwbase_rt *rwb)
{
int r;
/*
* Increment reader count, if sem->readers < 0, i.e. READER_BIAS is
* set.
*/
for (r = atomic_read(&rwb->readers); r < 0;) {
if (likely(atomic_try_cmpxchg_acquire(&rwb->readers, &r, r + 1)))
return 1;
}
return 0;
}
static int __sched __rwbase_read_lock(struct rwbase_rt *rwb,
unsigned int state)
{
struct rt_mutex_base *rtm = &rwb->rtmutex;
int ret;
raw_spin_lock_irq(&rtm->wait_lock);
/*
* 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)
* wait()
* down_read()
* unlock(m->wait_lock)
* up_read()
* wake(Writer)
* 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)
* wait()
* rtmutex_lock(m)
*
* That would put Reader1 behind the writer waiting on
* Reader2 to call up_read(), which might be unbound.
*/
trace_contention_begin(rwb, LCB_F_RT | LCB_F_READ);
/*
* For rwlocks this returns 0 unconditionally, so the below
* !ret conditionals are optimized out.
*/
ret = rwbase_rtmutex_slowlock_locked(rtm, state);
/*
* On success the rtmutex is held, so there can't be a writer
* active. Increment the reader count and immediately drop the
* rtmutex again.
*
* rtmutex->wait_lock has to be unlocked in any case of course.
*/
if (!ret)
atomic_inc(&rwb->readers);
raw_spin_unlock_irq(&rtm->wait_lock);
if (!ret)
rwbase_rtmutex_unlock(rtm);
trace_contention_end(rwb, ret);
return ret;
}
static __always_inline int rwbase_read_lock(struct rwbase_rt *rwb,
unsigned int state)
{
if (rwbase_read_trylock(rwb))
return 0;
return __rwbase_read_lock(rwb, state);
}
static void __sched __rwbase_read_unlock(struct rwbase_rt *rwb,
unsigned int state)
{
struct rt_mutex_base *rtm = &rwb->rtmutex;
struct task_struct *owner;
DEFINE_RT_WAKE_Q(wqh);
raw_spin_lock_irq(&rtm->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 rwb->readers it needs to acquire rtm->wait_lock. The
* worst case which can happen is a spurious wakeup.
*/
owner = rt_mutex_owner(rtm);
if (owner)
rt_mutex_wake_q_add_task(&wqh, owner, state);
/* Pairs with the preempt_enable in rt_mutex_wake_up_q() */
preempt_disable();
raw_spin_unlock_irq(&rtm->wait_lock);
rt_mutex_wake_up_q(&wqh);
}
static __always_inline void rwbase_read_unlock(struct rwbase_rt *rwb,
unsigned int state)
{
/*
* rwb->readers can only hit 0 when a writer is waiting for the
* active readers to leave the critical section.
*
* dec_and_test() is fully ordered, provides RELEASE.
*/
if (unlikely(atomic_dec_and_test(&rwb->readers)))
__rwbase_read_unlock(rwb, state);
}
static inline void __rwbase_write_unlock(struct rwbase_rt *rwb, int bias,
unsigned long flags)
{
struct rt_mutex_base *rtm = &rwb->rtmutex;
/*
* _release() is needed in case that reader is in fast path, pairing
* with atomic_try_cmpxchg_acquire() in rwbase_read_trylock().
*/
(void)atomic_add_return_release(READER_BIAS - bias, &rwb->readers);
raw_spin_unlock_irqrestore(&rtm->wait_lock, flags);
rwbase_rtmutex_unlock(rtm);
}
static inline void rwbase_write_unlock(struct rwbase_rt *rwb)
{
struct rt_mutex_base *rtm = &rwb->rtmutex;
unsigned long flags;
raw_spin_lock_irqsave(&rtm->wait_lock, flags);
__rwbase_write_unlock(rwb, WRITER_BIAS, flags);
}
static inline void rwbase_write_downgrade(struct rwbase_rt *rwb)
{
struct rt_mutex_base *rtm = &rwb->rtmutex;
unsigned long flags;
raw_spin_lock_irqsave(&rtm->wait_lock, flags);
/* Release it and account current as reader */
__rwbase_write_unlock(rwb, WRITER_BIAS - 1, flags);
}
static inline bool __rwbase_write_trylock(struct rwbase_rt *rwb)
{
/* Can do without CAS because we're serialized by wait_lock. */
lockdep_assert_held(&rwb->rtmutex.wait_lock);
/*
* _acquire is needed in case the reader is in the fast path, pairing
* with rwbase_read_unlock(), provides ACQUIRE.
*/
if (!atomic_read_acquire(&rwb->readers)) {
atomic_set(&rwb->readers, WRITER_BIAS);
return 1;
}
return 0;
}
static int __sched rwbase_write_lock(struct rwbase_rt *rwb,
unsigned int state)
{
struct rt_mutex_base *rtm = &rwb->rtmutex;
unsigned long flags;
/* Take the rtmutex as a first step */
if (rwbase_rtmutex_lock_state(rtm, state))
return -EINTR;
/* Force readers into slow path */
atomic_sub(READER_BIAS, &rwb->readers);
raw_spin_lock_irqsave(&rtm->wait_lock, flags);
if (__rwbase_write_trylock(rwb))
goto out_unlock;
rwbase_set_and_save_current_state(state);
trace_contention_begin(rwb, LCB_F_RT | LCB_F_WRITE);
for (;;) {
/* Optimized out for rwlocks */
if (rwbase_signal_pending_state(state, current)) {
rwbase_restore_current_state();
__rwbase_write_unlock(rwb, 0, flags);
trace_contention_end(rwb, -EINTR);
return -EINTR;
}
if (__rwbase_write_trylock(rwb))
break;
raw_spin_unlock_irqrestore(&rtm->wait_lock, flags);
rwbase_schedule();
raw_spin_lock_irqsave(&rtm->wait_lock, flags);
set_current_state(state);
}
rwbase_restore_current_state();
trace_contention_end(rwb, 0);
out_unlock:
raw_spin_unlock_irqrestore(&rtm->wait_lock, flags);
return 0;
}
static inline int rwbase_write_trylock(struct rwbase_rt *rwb)
{
struct rt_mutex_base *rtm = &rwb->rtmutex;
unsigned long flags;
if (!rwbase_rtmutex_trylock(rtm))
return 0;
atomic_sub(READER_BIAS, &rwb->readers);
raw_spin_lock_irqsave(&rtm->wait_lock, flags);
if (__rwbase_write_trylock(rwb)) {
raw_spin_unlock_irqrestore(&rtm->wait_lock, flags);
return 1;
}
__rwbase_write_unlock(rwb, 0, flags);
return 0;
}