1 files changed, 87 insertions, 70 deletions

diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index a4dd285cdf39..f321ed515f3a 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -84,9 +84,19 @@ static bool		have_boot_isolcpus;
 static struct list_head remote_children;
 
 /*
- * A flag to force sched domain rebuild at the end of an operation while
- * inhibiting it in the intermediate stages when set. Currently it is only
- * set in hotplug code.
+ * A flag to force sched domain rebuild at the end of an operation.
+ * It can be set in
+ *  - update_partition_sd_lb()
+ *  - remote_partition_check()
+ *  - update_cpumasks_hier()
+ *  - cpuset_update_flag()
+ *  - cpuset_hotplug_update_tasks()
+ *  - cpuset_handle_hotplug()
+ *
+ * Protected by cpuset_mutex (with cpus_read_lock held) or cpus_write_lock.
+ *
+ * Note that update_relax_domain_level() in cpuset-v1.c can still call
+ * rebuild_sched_domains_locked() directly without using this flag.
  */
 static bool force_sd_rebuild;
 
@@ -283,6 +293,12 @@ static inline void dec_attach_in_progress(struct cpuset *cs)
 	mutex_unlock(&cpuset_mutex);
 }
 
+static inline bool cpuset_v2(void)
+{
+	return !IS_ENABLED(CONFIG_CPUSETS_V1) ||
+		cgroup_subsys_on_dfl(cpuset_cgrp_subsys);
+}
+
 /*
  * Cgroup v2 behavior is used on the "cpus" and "mems" control files when
  * on default hierarchy or when the cpuset_v2_mode flag is set by mounting
@@ -293,7 +309,7 @@ static inline void dec_attach_in_progress(struct cpuset *cs)
  */
 static inline bool is_in_v2_mode(void)
 {
-	return cgroup_subsys_on_dfl(cpuset_cgrp_subsys) ||
+	return cpuset_v2() ||
 	      (cpuset_cgrp_subsys.root->flags & CGRP_ROOT_CPUSET_V2_MODE);
 }
 
@@ -565,12 +581,24 @@ static int validate_change(struct cpuset *cur, struct cpuset *trial)
 
 	/*
 	 * We can't shrink if we won't have enough room for SCHED_DEADLINE
-	 * tasks.
+	 * tasks. This check is not done when scheduling is disabled as the
+	 * users should know what they are doing.
+	 *
+	 * For v1, effective_cpus == cpus_allowed & user_xcpus() returns
+	 * cpus_allowed.
+	 *
+	 * For v2, is_cpu_exclusive() & is_sched_load_balance() are true only
+	 * for non-isolated partition root. At this point, the target
+	 * effective_cpus isn't computed yet. user_xcpus() is the best
+	 * approximation.
+	 *
+	 * TBD: May need to precompute the real effective_cpus here in case
+	 * incorrect scheduling of SCHED_DEADLINE tasks in a partition
+	 * becomes an issue.
 	 */
 	ret = -EBUSY;
-	if (is_cpu_exclusive(cur) &&
-	    !cpuset_cpumask_can_shrink(cur->cpus_allowed,
-				       trial->cpus_allowed))
+	if (is_cpu_exclusive(cur) && is_sched_load_balance(cur) &&
+	    !cpuset_cpumask_can_shrink(cur->effective_cpus, user_xcpus(trial)))
 		goto out;
 
 	/*
@@ -728,7 +756,7 @@ static int generate_sched_domains(cpumask_var_t **domains,
 	int nslot;		/* next empty doms[] struct cpumask slot */
 	struct cgroup_subsys_state *pos_css;
 	bool root_load_balance = is_sched_load_balance(&top_cpuset);
-	bool cgrpv2 = cgroup_subsys_on_dfl(cpuset_cgrp_subsys);
+	bool cgrpv2 = cpuset_v2();
 	int nslot_update;
 
 	doms = NULL;
@@ -990,6 +1018,7 @@ void rebuild_sched_domains_locked(void)
 
 	lockdep_assert_cpus_held();
 	lockdep_assert_held(&cpuset_mutex);
+	force_sd_rebuild = false;
 
 	/*
 	 * If we have raced with CPU hotplug, return early to avoid
@@ -1164,8 +1193,8 @@ static void update_partition_sd_lb(struct cpuset *cs, int old_prs)
 			clear_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
 	}
 
-	if (rebuild_domains && !force_sd_rebuild)
-		rebuild_sched_domains_locked();
+	if (rebuild_domains)
+		cpuset_force_rebuild();
 }
 
 /*
@@ -1187,7 +1216,7 @@ static void reset_partition_data(struct cpuset *cs)
 {
 	struct cpuset *parent = parent_cs(cs);
 
-	if (!cgroup_subsys_on_dfl(cpuset_cgrp_subsys))
+	if (!cpuset_v2())
 		return;
 
 	lockdep_assert_held(&callback_lock);
@@ -1339,7 +1368,7 @@ static inline bool is_local_partition(struct cpuset *cs)
  * remote_partition_enable - Enable current cpuset as a remote partition root
  * @cs: the cpuset to update
  * @new_prs: new partition_root_state
- * @tmp: temparary masks
+ * @tmp: temporary masks
  * Return: 0 if successful, errcode if error
  *
  * Enable the current cpuset to become a remote partition root taking CPUs
@@ -1377,7 +1406,7 @@ static int remote_partition_enable(struct cpuset *cs, int new_prs,
 	update_unbound_workqueue_cpumask(isolcpus_updated);
 
 	/*
-	 * Proprogate changes in top_cpuset's effective_cpus down the hierarchy.
+	 * Propagate changes in top_cpuset's effective_cpus down the hierarchy.
 	 */
 	cpuset_update_tasks_cpumask(&top_cpuset, tmp->new_cpus);
 	update_sibling_cpumasks(&top_cpuset, NULL, tmp);
@@ -1387,7 +1416,7 @@ static int remote_partition_enable(struct cpuset *cs, int new_prs,
 /*
  * remote_partition_disable - Remove current cpuset from remote partition list
  * @cs: the cpuset to update
- * @tmp: temparary masks
+ * @tmp: temporary masks
  *
  * The effective_cpus is also updated.
  *
@@ -1413,7 +1442,7 @@ static void remote_partition_disable(struct cpuset *cs, struct tmpmasks *tmp)
 	update_unbound_workqueue_cpumask(isolcpus_updated);
 
 	/*
-	 * Proprogate changes in top_cpuset's effective_cpus down the hierarchy.
+	 * Propagate changes in top_cpuset's effective_cpus down the hierarchy.
 	 */
 	cpuset_update_tasks_cpumask(&top_cpuset, tmp->new_cpus);
 	update_sibling_cpumasks(&top_cpuset, NULL, tmp);
@@ -1423,7 +1452,7 @@ static void remote_partition_disable(struct cpuset *cs, struct tmpmasks *tmp)
  * remote_cpus_update - cpus_exclusive change of remote partition
  * @cs: the cpuset to be updated
  * @newmask: the new effective_xcpus mask
- * @tmp: temparary masks
+ * @tmp: temporary masks
  *
  * top_cpuset and subpartitions_cpus will be updated or partition can be
  * invalidated.
@@ -1465,7 +1494,7 @@ static void remote_cpus_update(struct cpuset *cs, struct cpumask *newmask,
 	update_unbound_workqueue_cpumask(isolcpus_updated);
 
 	/*
-	 * Proprogate changes in top_cpuset's effective_cpus down the hierarchy.
+	 * Propagate changes in top_cpuset's effective_cpus down the hierarchy.
 	 */
 	cpuset_update_tasks_cpumask(&top_cpuset, tmp->new_cpus);
 	update_sibling_cpumasks(&top_cpuset, NULL, tmp);
@@ -1480,7 +1509,7 @@ invalidate:
  * @cs: the cpuset to be updated
  * @newmask: the new effective_xcpus mask
  * @delmask: temporary mask for deletion (not in tmp)
- * @tmp: temparary masks
+ * @tmp: temporary masks
  *
  * This should be called before the given cs has updated its cpus_allowed
  * and/or effective_xcpus.
@@ -1512,8 +1541,8 @@ static void remote_partition_check(struct cpuset *cs, struct cpumask *newmask,
 			remote_partition_disable(child, tmp);
 			disable_cnt++;
 		}
-	if (disable_cnt && !force_sd_rebuild)
-		rebuild_sched_domains_locked();
+	if (disable_cnt)
+		cpuset_force_rebuild();
 }
 
 /*
@@ -1923,12 +1952,6 @@ static void compute_partition_effective_cpumask(struct cpuset *cs,
 }
 
 /*
- * update_cpumasks_hier() flags
- */
-#define HIER_CHECKALL		0x01	/* Check all cpusets with no skipping */
-#define HIER_NO_SD_REBUILD	0x02	/* Don't rebuild sched domains */
-
-/*
  * update_cpumasks_hier - Update effective cpumasks and tasks in the subtree
  * @cs:  the cpuset to consider
  * @tmp: temp variables for calculating effective_cpus & partition setup
@@ -1942,7 +1965,7 @@ static void compute_partition_effective_cpumask(struct cpuset *cs,
  * Called with cpuset_mutex held
  */
 static void update_cpumasks_hier(struct cpuset *cs, struct tmpmasks *tmp,
-				 int flags)
+				 bool force)
 {
 	struct cpuset *cp;
 	struct cgroup_subsys_state *pos_css;
@@ -2007,12 +2030,12 @@ static void update_cpumasks_hier(struct cpuset *cs, struct tmpmasks *tmp,
 		 * Skip the whole subtree if
 		 * 1) the cpumask remains the same,
 		 * 2) has no partition root state,
-		 * 3) HIER_CHECKALL flag not set, and
+		 * 3) force flag not set, and
 		 * 4) for v2 load balance state same as its parent.
 		 */
-		if (!cp->partition_root_state && !(flags & HIER_CHECKALL) &&
+		if (!cp->partition_root_state && !force &&
 		    cpumask_equal(tmp->new_cpus, cp->effective_cpus) &&
-		    (!cgroup_subsys_on_dfl(cpuset_cgrp_subsys) ||
+		    (!cpuset_v2() ||
 		    (is_sched_load_balance(parent) == is_sched_load_balance(cp)))) {
 			pos_css = css_rightmost_descendant(pos_css);
 			continue;
@@ -2086,8 +2109,7 @@ get_css:
 		 * from parent if current cpuset isn't a valid partition root
 		 * and their load balance states differ.
 		 */
-		if (cgroup_subsys_on_dfl(cpuset_cgrp_subsys) &&
-		    !is_partition_valid(cp) &&
+		if (cpuset_v2() && !is_partition_valid(cp) &&
 		    (is_sched_load_balance(parent) != is_sched_load_balance(cp))) {
 			if (is_sched_load_balance(parent))
 				set_bit(CS_SCHED_LOAD_BALANCE, &cp->flags);
@@ -2103,8 +2125,7 @@ get_css:
 		 */
 		if (!cpumask_empty(cp->cpus_allowed) &&
 		    is_sched_load_balance(cp) &&
-		   (!cgroup_subsys_on_dfl(cpuset_cgrp_subsys) ||
-		    is_partition_valid(cp)))
+		   (!cpuset_v2() || is_partition_valid(cp)))
 			need_rebuild_sched_domains = true;
 
 		rcu_read_lock();
@@ -2112,9 +2133,8 @@ get_css:
 	}
 	rcu_read_unlock();
 
-	if (need_rebuild_sched_domains && !(flags & HIER_NO_SD_REBUILD) &&
-	    !force_sd_rebuild)
-		rebuild_sched_domains_locked();
+	if (need_rebuild_sched_domains)
+		cpuset_force_rebuild();
 }
 
 /**
@@ -2141,9 +2161,7 @@ static void update_sibling_cpumasks(struct cpuset *parent, struct cpuset *cs,
 	 * directly.
 	 *
 	 * The update_cpumasks_hier() function may sleep. So we have to
-	 * release the RCU read lock before calling it. HIER_NO_SD_REBUILD
-	 * flag is used to suppress rebuild of sched domains as the callers
-	 * will take care of that.
+	 * release the RCU read lock before calling it.
 	 */
 	rcu_read_lock();
 	cpuset_for_each_child(sibling, pos_css, parent) {
@@ -2159,7 +2177,7 @@ static void update_sibling_cpumasks(struct cpuset *parent, struct cpuset *cs,
 			continue;
 
 		rcu_read_unlock();
-		update_cpumasks_hier(sibling, tmp, HIER_NO_SD_REBUILD);
+		update_cpumasks_hier(sibling, tmp, false);
 		rcu_read_lock();
 		css_put(&sibling->css);
 	}
@@ -2179,7 +2197,7 @@ static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs,
 	struct tmpmasks tmp;
 	struct cpuset *parent = parent_cs(cs);
 	bool invalidate = false;
-	int hier_flags = 0;
+	bool force = false;
 	int old_prs = cs->partition_root_state;
 
 	/* top_cpuset.cpus_allowed tracks cpu_online_mask; it's read-only */
@@ -2206,7 +2224,7 @@ static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs,
 			return -EINVAL;
 
 		/*
-		 * When exclusive_cpus isn't explicitly set, it is constrainted
+		 * When exclusive_cpus isn't explicitly set, it is constrained
 		 * by cpus_allowed and parent's effective_xcpus. Otherwise,
 		 * trialcs->effective_xcpus is used as a temporary cpumask
 		 * for checking validity of the partition root.
@@ -2240,12 +2258,11 @@ static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs,
 	 * Check all the descendants in update_cpumasks_hier() if
 	 * effective_xcpus is to be changed.
 	 */
-	if (!cpumask_equal(cs->effective_xcpus, trialcs->effective_xcpus))
-		hier_flags = HIER_CHECKALL;
+	force = !cpumask_equal(cs->effective_xcpus, trialcs->effective_xcpus);
 
 	retval = validate_change(cs, trialcs);
 
-	if ((retval == -EINVAL) && cgroup_subsys_on_dfl(cpuset_cgrp_subsys)) {
+	if ((retval == -EINVAL) && cpuset_v2()) {
 		struct cgroup_subsys_state *css;
 		struct cpuset *cp;
 
@@ -2309,7 +2326,7 @@ static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs,
 	spin_unlock_irq(&callback_lock);
 
 	/* effective_cpus/effective_xcpus will be updated here */
-	update_cpumasks_hier(cs, &tmp, hier_flags);
+	update_cpumasks_hier(cs, &tmp, force);
 
 	/* Update CS_SCHED_LOAD_BALANCE and/or sched_domains, if necessary */
 	if (cs->partition_root_state)
@@ -2334,7 +2351,7 @@ static int update_exclusive_cpumask(struct cpuset *cs, struct cpuset *trialcs,
 	struct tmpmasks tmp;
 	struct cpuset *parent = parent_cs(cs);
 	bool invalidate = false;
-	int hier_flags = 0;
+	bool force = false;
 	int old_prs = cs->partition_root_state;
 
 	if (!*buf) {
@@ -2357,8 +2374,7 @@ static int update_exclusive_cpumask(struct cpuset *cs, struct cpuset *trialcs,
 	 * Check all the descendants in update_cpumasks_hier() if
 	 * effective_xcpus is to be changed.
 	 */
-	if (!cpumask_equal(cs->effective_xcpus, trialcs->effective_xcpus))
-		hier_flags = HIER_CHECKALL;
+	force = !cpumask_equal(cs->effective_xcpus, trialcs->effective_xcpus);
 
 	retval = validate_change(cs, trialcs);
 	if (retval)
@@ -2411,8 +2427,8 @@ static int update_exclusive_cpumask(struct cpuset *cs, struct cpuset *trialcs,
 	 * of the subtree when it is a valid partition root or effective_xcpus
 	 * is updated.
 	 */
-	if (is_partition_valid(cs) || hier_flags)
-		update_cpumasks_hier(cs, &tmp, hier_flags);
+	if (is_partition_valid(cs) || force)
+		update_cpumasks_hier(cs, &tmp, force);
 
 	/* Update CS_SCHED_LOAD_BALANCE and/or sched_domains, if necessary */
 	if (cs->partition_root_state)
@@ -2737,9 +2753,12 @@ int cpuset_update_flag(cpuset_flagbits_t bit, struct cpuset *cs,
 	cs->flags = trialcs->flags;
 	spin_unlock_irq(&callback_lock);
 
-	if (!cpumask_empty(trialcs->cpus_allowed) && balance_flag_changed &&
-	    !force_sd_rebuild)
-		rebuild_sched_domains_locked();
+	if (!cpumask_empty(trialcs->cpus_allowed) && balance_flag_changed) {
+		if (cpuset_v2())
+			cpuset_force_rebuild();
+		else
+			rebuild_sched_domains_locked();
+	}
 
 	if (spread_flag_changed)
 		cpuset1_update_tasks_flags(cs);
@@ -2853,12 +2872,14 @@ out:
 	update_unbound_workqueue_cpumask(new_xcpus_state);
 
 	/* Force update if switching back to member */
-	update_cpumasks_hier(cs, &tmpmask, !new_prs ? HIER_CHECKALL : 0);
+	update_cpumasks_hier(cs, &tmpmask, !new_prs);
 
 	/* Update sched domains and load balance flag */
 	update_partition_sd_lb(cs, old_prs);
 
 	notify_partition_change(cs, old_prs);
+	if (force_sd_rebuild)
+		rebuild_sched_domains_locked();
 	free_cpumasks(NULL, &tmpmask);
 	return 0;
 }
@@ -2919,8 +2940,7 @@ static int cpuset_can_attach(struct cgroup_taskset *tset)
 		 * migration permission derives from hierarchy ownership in
 		 * cgroup_procs_write_permission()).
 		 */
-		if (!cgroup_subsys_on_dfl(cpuset_cgrp_subsys) ||
-		    (cpus_updated || mems_updated)) {
+		if (!cpuset_v2() || (cpus_updated || mems_updated)) {
 			ret = security_task_setscheduler(task);
 			if (ret)
 				goto out_unlock;
@@ -3034,8 +3054,7 @@ static void cpuset_attach(struct cgroup_taskset *tset)
 	 * in effective cpus and mems. In that case, we can optimize out
 	 * by skipping the task iteration and update.
 	 */
-	if (cgroup_subsys_on_dfl(cpuset_cgrp_subsys) &&
-	    !cpus_updated && !mems_updated) {
+	if (cpuset_v2() && !cpus_updated && !mems_updated) {
 		cpuset_attach_nodemask_to = cs->effective_mems;
 		goto out;
 	}
@@ -3152,6 +3171,8 @@ ssize_t cpuset_write_resmask(struct kernfs_open_file *of,
 	}
 
 	free_cpuset(trialcs);
+	if (force_sd_rebuild)
+		rebuild_sched_domains_locked();
 out_unlock:
 	mutex_unlock(&cpuset_mutex);
 	cpus_read_unlock();
@@ -3383,7 +3404,7 @@ cpuset_css_alloc(struct cgroup_subsys_state *parent_css)
 	INIT_LIST_HEAD(&cs->remote_sibling);
 
 	/* Set CS_MEMORY_MIGRATE for default hierarchy */
-	if (cgroup_subsys_on_dfl(cpuset_cgrp_subsys))
+	if (cpuset_v2())
 		__set_bit(CS_MEMORY_MIGRATE, &cs->flags);
 
 	return &cs->css;
@@ -3410,8 +3431,7 @@ static int cpuset_css_online(struct cgroup_subsys_state *css)
 	/*
 	 * For v2, clear CS_SCHED_LOAD_BALANCE if parent is isolated
 	 */
-	if (cgroup_subsys_on_dfl(cpuset_cgrp_subsys) &&
-	    !is_sched_load_balance(parent))
+	if (cpuset_v2() && !is_sched_load_balance(parent))
 		clear_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
 
 	cpuset_inc();
@@ -3481,8 +3501,7 @@ static void cpuset_css_offline(struct cgroup_subsys_state *css)
 	if (is_partition_valid(cs))
 		update_prstate(cs, 0);
 
-	if (!cgroup_subsys_on_dfl(cpuset_cgrp_subsys) &&
-	    is_sched_load_balance(cs))
+	if (!cpuset_v2() && is_sched_load_balance(cs))
 		cpuset_update_flag(CS_SCHED_LOAD_BALANCE, cs, 0);
 
 	cpuset_dec();
@@ -3896,11 +3915,9 @@ static void cpuset_handle_hotplug(void)
 		rcu_read_unlock();
 	}
 
-	/* rebuild sched domains if cpus_allowed has changed */
-	if (force_sd_rebuild) {
-		force_sd_rebuild = false;
+	/* rebuild sched domains if necessary */
+	if (force_sd_rebuild)
 		rebuild_sched_domains_cpuslocked();
-	}
 
 	free_cpumasks(NULL, ptmp);
 }