Commit 01117332 authored by Kent Overstreet's avatar Kent Overstreet

bcachefs: six locks: Simplify optimistic spinning

osq lock maintainers don't want it to be used outside of kernel/locking/
- but, we can do better.

Since we have lock handoff signalled via waitlist entries, there's no
reason for optimistic spinning to have to look at the lock at all -
aside from checking lock-owner; we can just spin looking at our waitlist
entry.
Signed-off-by: default avatarKent Overstreet <kent.overstreet@linux.dev>
parent ee841b77
......@@ -85,6 +85,16 @@ config BCACHEFS_NO_LATENCY_ACCT
help
This disables device latency tracking and time stats, only for performance testing
config BCACHEFS_SIX_OPTIMISTIC_SPIN
bool "Optimistic spinning for six locks"
depends on BCACHEFS_FS
depends on SMP
default y
help
Instead of immediately sleeping when attempting to take a six lock that
is held by another thread, spin for a short while, as long as the
thread owning the lock is running.
config MEAN_AND_VARIANCE_UNIT_TEST
tristate "mean_and_variance unit tests" if !KUNIT_ALL_TESTS
depends on KUNIT
......
......@@ -324,101 +324,57 @@ bool six_relock_ip(struct six_lock *lock, enum six_lock_type type,
}
EXPORT_SYMBOL_GPL(six_relock_ip);
#ifdef CONFIG_SIX_LOCK_SPIN_ON_OWNER
#ifdef CONFIG_BCACHEFS_SIX_OPTIMISTIC_SPIN
static inline bool six_can_spin_on_owner(struct six_lock *lock)
static inline bool six_owner_running(struct six_lock *lock)
{
struct task_struct *owner;
bool ret;
if (need_resched())
return false;
/*
* When there's no owner, we might have preempted between the owner
* acquiring the lock and setting the owner field. If we're an RT task
* that will live-lock because we won't let the owner complete.
*/
rcu_read_lock();
owner = READ_ONCE(lock->owner);
ret = !owner || owner_on_cpu(owner);
struct task_struct *owner = READ_ONCE(lock->owner);
bool ret = owner ? owner_on_cpu(owner) : !rt_task(current);
rcu_read_unlock();
return ret;
}
static inline bool six_spin_on_owner(struct six_lock *lock,
struct task_struct *owner,
u64 end_time)
static inline bool six_optimistic_spin(struct six_lock *lock,
struct six_lock_waiter *wait,
enum six_lock_type type)
{
bool ret = true;
unsigned loop = 0;
rcu_read_lock();
while (lock->owner == owner) {
/*
* Ensure we emit the owner->on_cpu, dereference _after_
* checking lock->owner still matches owner. If that fails,
* owner might point to freed memory. If it still matches,
* the rcu_read_lock() ensures the memory stays valid.
*/
barrier();
if (!owner_on_cpu(owner) || need_resched()) {
ret = false;
break;
}
if (!(++loop & 0xf) && (time_after64(sched_clock(), end_time))) {
six_set_bitmask(lock, SIX_LOCK_NOSPIN);
ret = false;
break;
}
cpu_relax();
}
rcu_read_unlock();
return ret;
}
static inline bool six_optimistic_spin(struct six_lock *lock, enum six_lock_type type)
{
struct task_struct *task = current;
u64 end_time;
if (type == SIX_LOCK_write)
return false;
preempt_disable();
if (!six_can_spin_on_owner(lock))
goto fail;
if (lock->wait_list.next != &wait->list)
return false;
if (!osq_lock(&lock->osq))
goto fail;
if (atomic_read(&lock->state) & SIX_LOCK_NOSPIN)
return false;
preempt_disable();
end_time = sched_clock() + 10 * NSEC_PER_USEC;
while (1) {
struct task_struct *owner;
while (!need_resched() && six_owner_running(lock)) {
/*
* If there's an owner, wait for it to either
* release the lock or go to sleep.
* Ensures that writes to the waitlist entry happen after we see
* wait->lock_acquired: pairs with the smp_store_release in
* __six_lock_wakeup
*/
owner = READ_ONCE(lock->owner);
if (owner && !six_spin_on_owner(lock, owner, end_time))
break;
if (do_six_trylock(lock, type, false)) {
osq_unlock(&lock->osq);
if (smp_load_acquire(&wait->lock_acquired)) {
preempt_enable();
return true;
}
/*
* When there's no owner, we might have preempted between the
* owner acquiring the lock and setting the owner field. If
* we're an RT task that will live-lock because we won't let
* the owner complete.
*/
if (!owner && (need_resched() || rt_task(task)))
if (!(++loop & 0xf) && (time_after64(sched_clock(), end_time))) {
six_set_bitmask(lock, SIX_LOCK_NOSPIN);
break;
}
/*
* The cpu_relax() call is a compiler barrier which forces
......@@ -429,24 +385,15 @@ static inline bool six_optimistic_spin(struct six_lock *lock, enum six_lock_type
cpu_relax();
}
osq_unlock(&lock->osq);
fail:
preempt_enable();
/*
* If we fell out of the spin path because of need_resched(),
* reschedule now, before we try-lock again. This avoids getting
* scheduled out right after we obtained the lock.
*/
if (need_resched())
schedule();
return false;
}
#else /* CONFIG_SIX_LOCK_SPIN_ON_OWNER */
#else /* CONFIG_LOCK_SPIN_ON_OWNER */
static inline bool six_optimistic_spin(struct six_lock *lock, enum six_lock_type type)
static inline bool six_optimistic_spin(struct six_lock *lock,
struct six_lock_waiter *wait,
enum six_lock_type type)
{
return false;
}
......@@ -470,9 +417,6 @@ static int six_lock_slowpath(struct six_lock *lock, enum six_lock_type type,
trace_contention_begin(lock, 0);
lock_contended(&lock->dep_map, ip);
if (six_optimistic_spin(lock, type))
goto out;
wait->task = current;
wait->lock_want = type;
wait->lock_acquired = false;
......@@ -510,6 +454,9 @@ static int six_lock_slowpath(struct six_lock *lock, enum six_lock_type type,
ret = 0;
}
if (six_optimistic_spin(lock, wait, type))
goto out;
while (1) {
set_current_state(TASK_UNINTERRUPTIBLE);
......
......@@ -127,10 +127,6 @@
#include <linux/sched.h>
#include <linux/types.h>
#ifdef CONFIG_SIX_LOCK_SPIN_ON_OWNER
#include <linux/osq_lock.h>
#endif
enum six_lock_type {
SIX_LOCK_read,
SIX_LOCK_intent,
......@@ -143,9 +139,6 @@ struct six_lock {
unsigned intent_lock_recurse;
struct task_struct *owner;
unsigned __percpu *readers;
#ifdef CONFIG_SIX_LOCK_SPIN_ON_OWNER
struct optimistic_spin_queue osq;
#endif
raw_spinlock_t wait_lock;
struct list_head wait_list;
#ifdef CONFIG_DEBUG_LOCK_ALLOC
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment