diff options
| author | Peter Zijlstra <peterz@infradead.org> | 2021-03-03 16:45:41 +0100 |
|---|---|---|
| committer | Peter Zijlstra <peterz@infradead.org> | 2021-05-12 11:43:27 +0200 |
| commit | 9ef7e7e33bcdb57be1afb28884053c28b5f05240 (patch) | |
| tree | 40e43fa4c6d82adf7cd39fbc1f5dfb868701165b /kernel/sched/core.c | |
| parent | 9edeaea1bc452372718837ed2ba775811baf1ba1 (diff) | |
| download | linux-9ef7e7e33bcdb57be1afb28884053c28b5f05240.tar.bz2 | |
sched: Optimize rq_lockp() usage
rq_lockp() includes a static_branch(), which is asm-goto, which is
asm volatile which defeats regular CSE. This means that:
if (!static_branch(&foo))
return simple;
if (static_branch(&foo) && cond)
return complex;
Doesn't fold and we get horrible code. Introduce __rq_lockp() without
the static_branch() on.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Don Hiatt <dhiatt@digitalocean.com>
Tested-by: Hongyu Ning <hongyu.ning@linux.intel.com>
Tested-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/20210422123308.316696988@infradead.org
Diffstat (limited to 'kernel/sched/core.c')
| -rw-r--r-- | kernel/sched/core.c | 16 |
1 files changed, 8 insertions, 8 deletions
diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 384b79363a39..42c1c88741c0 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -300,9 +300,9 @@ void raw_spin_rq_lock_nested(struct rq *rq, int subclass) } for (;;) { - lock = rq_lockp(rq); + lock = __rq_lockp(rq); raw_spin_lock_nested(lock, subclass); - if (likely(lock == rq_lockp(rq))) { + if (likely(lock == __rq_lockp(rq))) { /* preempt_count *MUST* be > 1 */ preempt_enable_no_resched(); return; @@ -325,9 +325,9 @@ bool raw_spin_rq_trylock(struct rq *rq) } for (;;) { - lock = rq_lockp(rq); + lock = __rq_lockp(rq); ret = raw_spin_trylock(lock); - if (!ret || (likely(lock == rq_lockp(rq)))) { + if (!ret || (likely(lock == __rq_lockp(rq)))) { preempt_enable(); return ret; } @@ -352,7 +352,7 @@ void double_rq_lock(struct rq *rq1, struct rq *rq2) swap(rq1, rq2); raw_spin_rq_lock(rq1); - if (rq_lockp(rq1) == rq_lockp(rq2)) + if (__rq_lockp(rq1) == __rq_lockp(rq2)) return; raw_spin_rq_lock_nested(rq2, SINGLE_DEPTH_NESTING); @@ -2622,7 +2622,7 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu) * task_rq_lock(). */ WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) || - lockdep_is_held(rq_lockp(task_rq(p))))); + lockdep_is_held(__rq_lockp(task_rq(p))))); #endif /* * Clearly, migrating tasks to offline CPUs is a fairly daft thing. @@ -4248,7 +4248,7 @@ prepare_lock_switch(struct rq *rq, struct task_struct *next, struct rq_flags *rf * do an early lockdep release here: */ rq_unpin_lock(rq, rf); - spin_release(&rq_lockp(rq)->dep_map, _THIS_IP_); + spin_release(&__rq_lockp(rq)->dep_map, _THIS_IP_); #ifdef CONFIG_DEBUG_SPINLOCK /* this is a valid case when another task releases the spinlock */ rq_lockp(rq)->owner = next; @@ -4262,7 +4262,7 @@ static inline void finish_lock_switch(struct rq *rq) * fix up the runqueue lock - which gets 'carried over' from * prev into current: */ - spin_acquire(&rq_lockp(rq)->dep_map, 0, 0, _THIS_IP_); + spin_acquire(&__rq_lockp(rq)->dep_map, 0, 0, _THIS_IP_); __balance_callbacks(rq); raw_spin_rq_unlock_irq(rq); } |