if (p->flags & PF_EXITING)
return;
- /*
- * We do not care about task placement until a task runs on a node
- * other than the first one used by the address space. This is
- * largely because migrations are driven by what CPU the task
- * is running on. If it's never scheduled on another node, it'll
- * not migrate so why bother trapping the fault.
- */
- if (mm->first_nid == NUMA_PTE_SCAN_INIT)
- mm->first_nid = numa_node_id();
- if (mm->first_nid != NUMA_PTE_SCAN_ACTIVE) {
- /* Are we running on a new node yet? */
- if (numa_node_id() == mm->first_nid &&
- !sched_feat_numa(NUMA_FORCE))
- return;
-
- mm->first_nid = NUMA_PTE_SCAN_ACTIVE;
+ if (!mm->numa_next_reset || !mm->numa_next_scan) {
+ mm->numa_next_scan = now +
+ msecs_to_jiffies(sysctl_numa_balancing_scan_delay);
+ mm->numa_next_reset = now +
+ msecs_to_jiffies(sysctl_numa_balancing_scan_period_reset);
}
/*
return;
/*
- * Do not set pte_numa if the current running node is rate-limited.
- * This loses statistics on the fault but if we are unwilling to
- * migrate to this node, it is less likely we can do useful work
+ * Delay this task enough that another task of this mm will likely win
+ * the next time around.
*/
- if (migrate_ratelimited(numa_node_id()))
- return;
+ p->node_stamp += 2 * TICK_NSEC;
start = mm->numa_scan_offset;
pages = sysctl_numa_balancing_scan_size;
out:
/*
- * It is possible to reach the end of the VMA list but the last few VMAs are
- * not guaranteed to the vma_migratable. If they are not, we would find the
- * !migratable VMA on the next scan but not reset the scanner to the start
- * so check it now.
+ * It is possible to reach the end of the VMA list but the last few
+ * VMAs are not guaranteed to the vma_migratable. If they are not, we
+ * would find the !migratable VMA on the next scan but not reset the
+ * scanner to the start so check it now.
*/
if (vma)
mm->numa_scan_offset = start;
if (now - curr->node_stamp > period) {
if (!curr->node_stamp)
curr->numa_scan_period = sysctl_numa_balancing_scan_period_min;
- curr->node_stamp = now;
+ curr->node_stamp += period;
if (!time_before(jiffies, curr->mm->numa_next_scan)) {
init_task_work(work, task_numa_work); /* TODO: move this into sched_fork() */
#define LBF_ALL_PINNED 0x01
#define LBF_NEED_BREAK 0x02
-#define LBF_SOME_PINNED 0x04
+#define LBF_DST_PINNED 0x04
+#define LBF_SOME_PINNED 0x08
struct lb_env {
struct sched_domain *sd;
schedstat_inc(p, se.statistics.nr_failed_migrations_affine);
+ env->flags |= LBF_SOME_PINNED;
+
/*
* Remember if this task can be migrated to any other cpu in
* our sched_group. We may want to revisit it if we couldn't
* Also avoid computing new_dst_cpu if we have already computed
* one in current iteration.
*/
- if (!env->dst_grpmask || (env->flags & LBF_SOME_PINNED))
+ if (!env->dst_grpmask || (env->flags & LBF_DST_PINNED))
return 0;
/* Prevent to re-select dst_cpu via env's cpus */
for_each_cpu_and(cpu, env->dst_grpmask, env->cpus) {
if (cpumask_test_cpu(cpu, tsk_cpus_allowed(p))) {
- env->flags |= LBF_SOME_PINNED;
+ env->flags |= LBF_DST_PINNED;
env->new_dst_cpu = cpu;
break;
}
}
if (!se) {
- cfs_rq->h_load = rq->avg.load_avg_contrib;
+ cfs_rq->h_load = cfs_rq->runnable_load_avg;
cfs_rq->last_h_load_update = now;
}
{
struct sched_domain *child = sd->child;
struct sched_group *group, *sdg = sd->groups;
- unsigned long power;
+ unsigned long power, power_orig;
unsigned long interval;
interval = msecs_to_jiffies(sd->balance_interval);
return;
}
- power = 0;
+ power_orig = power = 0;
if (child->flags & SD_OVERLAP) {
/*
* span the current group.
*/
- for_each_cpu(cpu, sched_group_cpus(sdg))
- power += power_of(cpu);
+ for_each_cpu(cpu, sched_group_cpus(sdg)) {
+ struct sched_group *sg = cpu_rq(cpu)->sd->groups;
+
+ power_orig += sg->sgp->power_orig;
+ power += sg->sgp->power;
+ }
} else {
/*
* !SD_OVERLAP domains can assume that child groups
group = child->groups;
do {
+ power_orig += group->sgp->power_orig;
power += group->sgp->power;
group = group->next;
} while (group != child->groups);
}
- sdg->sgp->power_orig = sdg->sgp->power = power;
+ sdg->sgp->power_orig = power_orig;
+ sdg->sgp->power = power;
}
/*
* cpu 3 and leave one of the cpus in the second group unused.
*
* The current solution to this issue is detecting the skew in the first group
- * by noticing it has a cpu that is overloaded while the remaining cpus are
- * idle -- or rather, there's a distinct imbalance in the cpus; see
- * sg_imbalanced().
+ * by noticing the lower domain failed to reach balance and had difficulty
+ * moving tasks due to affinity constraints.
*
* When this is so detected; this group becomes a candidate for busiest; see
* update_sd_pick_busiest(). And calculcate_imbalance() and
- * find_busiest_group() avoid some of the usual balance conditional to allow it
+ * find_busiest_group() avoid some of the usual balance conditions to allow it
* to create an effective group imbalance.
*
* This is a somewhat tricky proposition since the next run might not find the
* subtle and fragile situation.
*/
-struct sg_imb_stats {
- unsigned long max_nr_running, min_nr_running;
- unsigned long max_cpu_load, min_cpu_load;
-};
-
-static inline void init_sg_imb_stats(struct sg_imb_stats *sgi)
+static inline int sg_imbalanced(struct sched_group *group)
{
- sgi->max_cpu_load = sgi->max_nr_running = 0UL;
- sgi->min_cpu_load = sgi->min_nr_running = ~0UL;
+ return group->sgp->imbalance;
}
-static inline void
-update_sg_imb_stats(struct sg_imb_stats *sgi,
- unsigned long load, unsigned long nr_running)
+/*
+ * Compute the group capacity.
+ *
+ * Avoid the issue where N*frac(smt_power) >= 1 creates 'phantom' cores by
+ * first dividing out the smt factor and computing the actual number of cores
+ * and limit power unit capacity with that.
+ */
+static inline int sg_capacity(struct lb_env *env, struct sched_group *group)
{
- if (load > sgi->max_cpu_load)
- sgi->max_cpu_load = load;
- if (sgi->min_cpu_load > load)
- sgi->min_cpu_load = load;
+ unsigned int capacity, smt, cpus;
+ unsigned int power, power_orig;
- if (nr_running > sgi->max_nr_running)
- sgi->max_nr_running = nr_running;
- if (sgi->min_nr_running > nr_running)
- sgi->min_nr_running = nr_running;
-}
+ power = group->sgp->power;
+ power_orig = group->sgp->power_orig;
+ cpus = group->group_weight;
-static inline int
-sg_imbalanced(struct sg_lb_stats *sgs, struct sg_imb_stats *sgi)
-{
- /*
- * Consider the group unbalanced when the imbalance is larger
- * than the average weight of a task.
- *
- * APZ: with cgroup the avg task weight can vary wildly and
- * might not be a suitable number - should we keep a
- * normalized nr_running number somewhere that negates
- * the hierarchy?
- */
- if ((sgi->max_cpu_load - sgi->min_cpu_load) >= sgs->load_per_task &&
- (sgi->max_nr_running - sgi->min_nr_running) > 1)
- return 1;
+ /* smt := ceil(cpus / power), assumes: 1 < smt_power < 2 */
+ smt = DIV_ROUND_UP(SCHED_POWER_SCALE * cpus, power_orig);
+ capacity = cpus / smt; /* cores */
- return 0;
+ capacity = min_t(unsigned, capacity, DIV_ROUND_CLOSEST(power, SCHED_POWER_SCALE));
+ if (!capacity)
+ capacity = fix_small_capacity(env->sd, group);
+
+ return capacity;
}
/**
struct sched_group *group, int load_idx,
int local_group, struct sg_lb_stats *sgs)
{
- struct sg_imb_stats sgi;
unsigned long nr_running;
unsigned long load;
int i;
- init_sg_imb_stats(&sgi);
+ memset(sgs, 0, sizeof(*sgs));
for_each_cpu_and(i, sched_group_cpus(group), env->cpus) {
struct rq *rq = cpu_rq(i);
nr_running = rq->nr_running;
/* Bias balancing toward cpus of our domain */
- if (local_group) {
+ if (local_group)
load = target_load(i, load_idx);
- } else {
+ else
load = source_load(i, load_idx);
- update_sg_imb_stats(&sgi, load, nr_running);
- }
sgs->group_load += load;
sgs->sum_nr_running += nr_running;
sgs->idle_cpus++;
}
- if (local_group && (env->idle != CPU_NEWLY_IDLE ||
- time_after_eq(jiffies, group->sgp->next_update)))
- update_group_power(env->sd, env->dst_cpu);
-
/* Adjust by relative CPU power of the group */
sgs->group_power = group->sgp->power;
sgs->avg_load = (sgs->group_load*SCHED_POWER_SCALE) / sgs->group_power;
if (sgs->sum_nr_running)
sgs->load_per_task = sgs->sum_weighted_load / sgs->sum_nr_running;
- sgs->group_imb = sg_imbalanced(sgs, &sgi);
-
- sgs->group_capacity =
- DIV_ROUND_CLOSEST(sgs->group_power, SCHED_POWER_SCALE);
-
- if (!sgs->group_capacity)
- sgs->group_capacity = fix_small_capacity(env->sd, group);
-
sgs->group_weight = group->group_weight;
+ sgs->group_imb = sg_imbalanced(group);
+ sgs->group_capacity = sg_capacity(env, group);
+
if (sgs->group_capacity > sgs->sum_nr_running)
sgs->group_has_capacity = 1;
}
if (local_group) {
sds->local = sg;
sgs = &sds->local_stat;
+
+ if (env->idle != CPU_NEWLY_IDLE ||
+ time_after_eq(jiffies, sg->sgp->next_update))
+ update_group_power(env->sd, env->dst_cpu);
}
- memset(sgs, 0, sizeof(*sgs));
update_sg_lb_stats(env, sg, load_idx, local_group, sgs);
+ if (local_group)
+ goto next_group;
+
/*
* In case the child domain prefers tasks go to siblings
* first, lower the sg capacity to one so that we'll try
* heaviest group when it is already under-utilized (possible
* with a large weight task outweighs the tasks on the system).
*/
- if (prefer_sibling && !local_group &&
- sds->local && sds->local_stat.group_has_capacity)
+ if (prefer_sibling && sds->local &&
+ sds->local_stat.group_has_capacity)
sgs->group_capacity = min(sgs->group_capacity, 1U);
- /* Now, start updating sd_lb_stats */
- sds->total_load += sgs->group_load;
- sds->total_pwr += sgs->group_power;
-
- if (!local_group && update_sd_pick_busiest(env, sds, sg, sgs)) {
+ if (update_sd_pick_busiest(env, sds, sg, sgs)) {
sds->busiest = sg;
sds->busiest_stat = *sgs;
}
+next_group:
+ /* Now, start updating sd_lb_stats */
+ sds->total_load += sgs->group_load;
+ sds->total_pwr += sgs->group_power;
+
sg = sg->next;
} while (sg != env->sd->groups);
}
(busiest->load_per_task * SCHED_POWER_SCALE) /
busiest->group_power;
- if (busiest->avg_load - local->avg_load + scaled_busy_load_per_task >=
- (scaled_busy_load_per_task * imbn)) {
+ if (busiest->avg_load + scaled_busy_load_per_task >=
+ local->avg_load + (scaled_busy_load_per_task * imbn)) {
env->imbalance = busiest->load_per_task;
return;
}
* max load less than avg load(as we skip the groups at or below
* its cpu_power, while calculating max_load..)
*/
- if (busiest->avg_load < sds->avg_load) {
+ if (busiest->avg_load <= sds->avg_load ||
+ local->avg_load >= sds->avg_load) {
env->imbalance = 0;
return fix_small_imbalance(env, sds);
}
int *continue_balancing)
{
int ld_moved, cur_ld_moved, active_balance = 0;
+ struct sched_domain *sd_parent = sd->parent;
struct sched_group *group;
struct rq *busiest;
unsigned long flags;
* moreover subsequent load balance cycles should correct the
* excess load moved.
*/
- if ((env.flags & LBF_SOME_PINNED) && env.imbalance > 0) {
+ if ((env.flags & LBF_DST_PINNED) && env.imbalance > 0) {
+
+ /* Prevent to re-select dst_cpu via env's cpus */
+ cpumask_clear_cpu(env.dst_cpu, env.cpus);
env.dst_rq = cpu_rq(env.new_dst_cpu);
env.dst_cpu = env.new_dst_cpu;
- env.flags &= ~LBF_SOME_PINNED;
+ env.flags &= ~LBF_DST_PINNED;
env.loop = 0;
env.loop_break = sched_nr_migrate_break;
- /* Prevent to re-select dst_cpu via env's cpus */
- cpumask_clear_cpu(env.dst_cpu, env.cpus);
-
/*
* Go back to "more_balance" rather than "redo" since we
* need to continue with same src_cpu.
goto more_balance;
}
+ /*
+ * We failed to reach balance because of affinity.
+ */
+ if (sd_parent) {
+ int *group_imbalance = &sd_parent->groups->sgp->imbalance;
+
+ if ((env.flags & LBF_SOME_PINNED) && env.imbalance > 0) {
+ *group_imbalance = 1;
+ } else if (*group_imbalance)
+ *group_imbalance = 0;
+ }
+
/* All tasks on this runqueue were pinned by CPU affinity */
if (unlikely(env.flags & LBF_ALL_PINNED)) {
cpumask_clear_cpu(cpu_of(busiest), cpus);
struct sched_domain *sd;
int pulled_task = 0;
unsigned long next_balance = jiffies + HZ;
+ u64 curr_cost = 0;
this_rq->idle_stamp = rq_clock(this_rq);
for_each_domain(this_cpu, sd) {
unsigned long interval;
int continue_balancing = 1;
+ u64 t0, domain_cost;
if (!(sd->flags & SD_LOAD_BALANCE))
continue;
+ if (this_rq->avg_idle < curr_cost + sd->max_newidle_lb_cost)
+ break;
+
if (sd->flags & SD_BALANCE_NEWIDLE) {
+ t0 = sched_clock_cpu(this_cpu);
+
/* If we've pulled tasks over stop searching: */
pulled_task = load_balance(this_cpu, this_rq,
sd, CPU_NEWLY_IDLE,
&continue_balancing);
+
+ domain_cost = sched_clock_cpu(this_cpu) - t0;
+ if (domain_cost > sd->max_newidle_lb_cost)
+ sd->max_newidle_lb_cost = domain_cost;
+
+ curr_cost += domain_cost;
}
interval = msecs_to_jiffies(sd->balance_interval);
*/
this_rq->next_balance = next_balance;
}
+
+ if (curr_cost > this_rq->max_idle_balance_cost)
+ this_rq->max_idle_balance_cost = curr_cost;
}
/*
/* Earliest time when we have to do rebalance again */
unsigned long next_balance = jiffies + 60*HZ;
int update_next_balance = 0;
- int need_serialize;
+ int need_serialize, need_decay = 0;
+ u64 max_cost = 0;
update_blocked_averages(cpu);
rcu_read_lock();
for_each_domain(cpu, sd) {
+ /*
+ * Decay the newidle max times here because this is a regular
+ * visit to all the domains. Decay ~1% per second.
+ */
+ if (time_after(jiffies, sd->next_decay_max_lb_cost)) {
+ sd->max_newidle_lb_cost =
+ (sd->max_newidle_lb_cost * 253) / 256;
+ sd->next_decay_max_lb_cost = jiffies + HZ;
+ need_decay = 1;
+ }
+ max_cost += sd->max_newidle_lb_cost;
+
if (!(sd->flags & SD_LOAD_BALANCE))
continue;
+ /*
+ * Stop the load balance at this level. There is another
+ * CPU in our sched group which is doing load balancing more
+ * actively.
+ */
+ if (!continue_balancing) {
+ if (need_decay)
+ continue;
+ break;
+ }
+
interval = sd->balance_interval;
if (idle != CPU_IDLE)
interval *= sd->busy_factor;
if (time_after_eq(jiffies, sd->last_balance + interval)) {
if (load_balance(cpu, rq, sd, idle, &continue_balancing)) {
/*
- * The LBF_SOME_PINNED logic could have changed
+ * The LBF_DST_PINNED logic could have changed
* env->dst_cpu, so we can't know our idle
* state even if we migrated tasks. Update it.
*/
next_balance = sd->last_balance + interval;
update_next_balance = 1;
}
-
+ }
+ if (need_decay) {
/*
- * Stop the load balance at this level. There is another
- * CPU in our sched group which is doing load balancing more
- * actively.
+ * Ensure the rq-wide value also decays but keep it at a
+ * reasonable floor to avoid funnies with rq->avg_idle.
*/
- if (!continue_balancing)
- break;
+ rq->max_idle_balance_cost =
+ max((u64)sysctl_sched_migration_cost, max_cost);
}
rcu_read_unlock();