Commit 71fedc41 authored by Peter Zijlstra's avatar Peter Zijlstra Committed by Ingo Molnar
Browse files

sched/fair: Switch to rcu_dereference_all() 



With the {rcu,sched,bh} RCU flavours being unified, it doesn't really
make sense to check for just the rcu one. Switch to the _all family of
verification which includes all 3 of the listed flavours.

Notably, this will enable us to remove some superfluous
rcu_read_lock() regions when we know they are inside preempt/IRQ
disabled regions.

Signed-off-by: default avatarPeter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: default avatarIngo Molnar <mingo@kernel.org>
parent f24165bf

kernel/sched/fair.c

+24 −24
Original line number Diff line number Diff line
@@ -1513,7 +1513,7 @@ static unsigned int task_scan_start(struct task_struct *p) 


	/* Scale the maximum scan period with the amount of shared memory. */

	rcu_read_lock();

	ng = rcu_dereference(p->numa_group);

	ng = rcu_dereference_all(p->numa_group);

	if (ng) {

		unsigned long shared = group_faults_shared(ng);

		unsigned long private = group_faults_priv(ng);
@@ -1580,7 +1580,7 @@ pid_t task_numa_group_id(struct task_struct *p) 
	pid_t gid = 0;



	rcu_read_lock();

	ng = rcu_dereference(p->numa_group);

	ng = rcu_dereference_all(p->numa_group);

	if (ng)

		gid = ng->gid;

	rcu_read_unlock();
@@ -2239,7 +2239,7 @@ static bool task_numa_compare(struct task_numa_env *env, 
		return false;



	rcu_read_lock();

	cur = rcu_dereference(dst_rq->curr);

	cur = rcu_dereference_all(dst_rq->curr);

	if (cur && ((cur->flags & (PF_EXITING | PF_KTHREAD)) ||

		    !cur->mm))

		cur = NULL;
@@ -2284,7 +2284,7 @@ static bool task_numa_compare(struct task_numa_env *env, 
	 * If dst and source tasks are in the same NUMA group, or not

	 * in any group then look only at task weights.

	 */

	cur_ng = rcu_dereference(cur->numa_group);

	cur_ng = rcu_dereference_all(cur->numa_group);

	if (cur_ng == p_ng) {

		/*

		 * Do not swap within a group or between tasks that have
@@ -2499,7 +2499,7 @@ static int task_numa_migrate(struct task_struct *p) 
	 * to satisfy here.

	 */

	rcu_read_lock();

	sd = rcu_dereference(per_cpu(sd_numa, env.src_cpu));

	sd = rcu_dereference_all(per_cpu(sd_numa, env.src_cpu));

	if (sd) {

		env.imbalance_pct = 100 + (sd->imbalance_pct - 100) / 2;

		env.imb_numa_nr = sd->imb_numa_nr;
@@ -3022,7 +3022,7 @@ static void task_numa_group(struct task_struct *p, int cpupid, int flags, 
	if (!cpupid_match_pid(tsk, cpupid))

		goto no_join;



	grp = rcu_dereference(tsk->numa_group);

	grp = rcu_dereference_all(tsk->numa_group);

	if (!grp)

		goto no_join;
@@ -4435,7 +4435,7 @@ static inline void migrate_se_pelt_lag(struct sched_entity *se) 
	rq = rq_of(cfs_rq);



	rcu_read_lock();

	is_idle = is_idle_task(rcu_dereference(rq->curr));

	is_idle = is_idle_task(rcu_dereference_all(rq->curr));

	rcu_read_unlock();



	/*
@@ -7462,7 +7462,7 @@ static inline void set_idle_cores(int cpu, int val) 
{

	struct sched_domain_shared *sds;



	sds = rcu_dereference(per_cpu(sd_llc_shared, cpu));

	sds = rcu_dereference_all(per_cpu(sd_llc_shared, cpu));

	if (sds)

		WRITE_ONCE(sds->has_idle_cores, val);

}
@@ -7471,7 +7471,7 @@ static inline bool test_idle_cores(int cpu) 
{

	struct sched_domain_shared *sds;



	sds = rcu_dereference(per_cpu(sd_llc_shared, cpu));

	sds = rcu_dereference_all(per_cpu(sd_llc_shared, cpu));

	if (sds)

		return READ_ONCE(sds->has_idle_cores);
@@ -7600,7 +7600,7 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, bool 
	cpumask_and(cpus, sched_domain_span(sd), p->cpus_ptr);



	if (sched_feat(SIS_UTIL)) {

		sd_share = rcu_dereference(per_cpu(sd_llc_shared, target));

		sd_share = rcu_dereference_all(per_cpu(sd_llc_shared, target));

		if (sd_share) {

			/* because !--nr is the condition to stop scan */

			nr = READ_ONCE(sd_share->nr_idle_scan) + 1;
@@ -7806,7 +7806,7 @@ static int select_idle_sibling(struct task_struct *p, int prev, int target) 
	 * sd_asym_cpucapacity rather than sd_llc.

	 */

	if (sched_asym_cpucap_active()) {

		sd = rcu_dereference(per_cpu(sd_asym_cpucapacity, target));

		sd = rcu_dereference_all(per_cpu(sd_asym_cpucapacity, target));

		/*

		 * On an asymmetric CPU capacity system where an exclusive

		 * cpuset defines a symmetric island (i.e. one unique
@@ -7821,7 +7821,7 @@ static int select_idle_sibling(struct task_struct *p, int prev, int target) 
		}

	}



	sd = rcu_dereference(per_cpu(sd_llc, target));

	sd = rcu_dereference_all(per_cpu(sd_llc, target));

	if (!sd)

		return target;
@@ -8290,7 +8290,7 @@ static int find_energy_efficient_cpu(struct task_struct *p, int prev_cpu) 
	struct energy_env eenv;



	rcu_read_lock();

	pd = rcu_dereference(rd->pd);

	pd = rcu_dereference_all(rd->pd);

	if (!pd)

		goto unlock;
@@ -8298,7 +8298,7 @@ static int find_energy_efficient_cpu(struct task_struct *p, int prev_cpu) 
	 * Energy-aware wake-up happens on the lowest sched_domain starting

	 * from sd_asym_cpucapacity spanning over this_cpu and prev_cpu.

	 */

	sd = rcu_dereference(*this_cpu_ptr(&sd_asym_cpucapacity));

	sd = rcu_dereference_all(*this_cpu_ptr(&sd_asym_cpucapacity));

	while (sd && !cpumask_test_cpu(prev_cpu, sched_domain_span(sd)))

		sd = sd->parent;

	if (!sd)
@@ -9305,7 +9305,7 @@ static int task_hot(struct task_struct *p, struct lb_env *env) 
 */

static long migrate_degrades_locality(struct task_struct *p, struct lb_env *env)

{

	struct numa_group *numa_group = rcu_dereference(p->numa_group);

	struct numa_group *numa_group = rcu_dereference_all(p->numa_group);

	unsigned long src_weight, dst_weight;

	int src_nid, dst_nid, dist;
@@ -10985,7 +10985,7 @@ static void update_idle_cpu_scan(struct lb_env *env, 
	if (env->sd->span_weight != llc_weight)

		return;



	sd_share = rcu_dereference(per_cpu(sd_llc_shared, env->dst_cpu));

	sd_share = rcu_dereference_all(per_cpu(sd_llc_shared, env->dst_cpu));

	if (!sd_share)

		return;
@@ -11335,7 +11335,7 @@ static struct sched_group *sched_balance_find_src_group(struct lb_env *env) 
		goto force_balance;



	if (!is_rd_overutilized(env->dst_rq->rd) &&

	    rcu_dereference(env->dst_rq->rd->pd))

	    rcu_dereference_all(env->dst_rq->rd->pd))

		goto out_balanced;



	/* ASYM feature bypasses nice load balance check */
@@ -12424,7 +12424,7 @@ static void nohz_balancer_kick(struct rq *rq) 


	rcu_read_lock();



	sd = rcu_dereference(rq->sd);

	sd = rcu_dereference_all(rq->sd);

	if (sd) {

		/*

		 * If there's a runnable CFS task and the current CPU has reduced
@@ -12436,7 +12436,7 @@ static void nohz_balancer_kick(struct rq *rq) 
		}

	}



	sd = rcu_dereference(per_cpu(sd_asym_packing, cpu));

	sd = rcu_dereference_all(per_cpu(sd_asym_packing, cpu));

	if (sd) {

		/*

		 * When ASYM_PACKING; see if there's a more preferred CPU
@@ -12454,7 +12454,7 @@ static void nohz_balancer_kick(struct rq *rq) 
		}

	}



	sd = rcu_dereference(per_cpu(sd_asym_cpucapacity, cpu));

	sd = rcu_dereference_all(per_cpu(sd_asym_cpucapacity, cpu));

	if (sd) {

		/*

		 * When ASYM_CPUCAPACITY; see if there's a higher capacity CPU
@@ -12475,7 +12475,7 @@ static void nohz_balancer_kick(struct rq *rq) 
		goto unlock;

	}



	sds = rcu_dereference(per_cpu(sd_llc_shared, cpu));

	sds = rcu_dereference_all(per_cpu(sd_llc_shared, cpu));

	if (sds) {

		/*

		 * If there is an imbalance between LLC domains (IOW we could
@@ -12507,7 +12507,7 @@ static void set_cpu_sd_state_busy(int cpu) 
	struct sched_domain *sd;



	rcu_read_lock();

	sd = rcu_dereference(per_cpu(sd_llc, cpu));

	sd = rcu_dereference_all(per_cpu(sd_llc, cpu));



	if (!sd || !sd->nohz_idle)

		goto unlock;
@@ -12537,7 +12537,7 @@ static void set_cpu_sd_state_idle(int cpu) 
	struct sched_domain *sd;



	rcu_read_lock();

	sd = rcu_dereference(per_cpu(sd_llc, cpu));

	sd = rcu_dereference_all(per_cpu(sd_llc, cpu));



	if (!sd || sd->nohz_idle)

		goto unlock;
@@ -13915,7 +13915,7 @@ void show_numa_stats(struct task_struct *p, struct seq_file *m) 
	struct numa_group *ng;



	rcu_read_lock();

	ng = rcu_dereference(p->numa_group);

	ng = rcu_dereference_all(p->numa_group);

	for_each_online_node(node) {

		if (p->numa_faults) {

			tsf = p->numa_faults[task_faults_idx(NUMA_MEM, node, 0)];

kernel/sched/sched.h

+1 −1
Original line number Diff line number Diff line
@@ -2011,7 +2011,7 @@ queue_balance_callback(struct rq *rq, 
}



#define rcu_dereference_sched_domain(p) \

	rcu_dereference_check((p), lockdep_is_held(&sched_domains_mutex))

	rcu_dereference_all_check((p), lockdep_is_held(&sched_domains_mutex))



/*

 * The domain tree (rq->sd) is protected by RCU's quiescent state transition.