sched: Add cgroup-based criteria for upmigration

It may be desirable to discourage upmigration of tasks belonging to
some cgroups. Add a per-cgroup flag (upmigrate_discourage) that
discourages upmigration of tasks of a cgroup. Tasks of the cgroup are
allowed to upmigrate only under overcommitted scenario.

Change-Id: I1780e420af1b6865c5332fb55ee1ee408b74d8ce
Signed-off-by: Srivatsa Vaddagiri <vatsa@codeaurora.org>
[rameezmustafa@codeaurora.org: Use new cgroup APIs]
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>

3 files changed, 69 insertions, 6 deletions

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index eb6480b51d93..b4bb96da4efb 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -10006,6 +10006,45 @@ static int cpu_notify_on_migrate_write_u64(struct cgroup_subsys_state *css,
 	return 0;
 }
 
+#ifdef CONFIG_SCHED_HMP
+
+static u64 cpu_upmigrate_discourage_read_u64(struct cgroup_subsys_state *css,
+					  struct cftype *cft)
+{
+	struct task_group *tg = css_tg(css);
+
+	return tg->upmigrate_discouraged;
+}
+
+static int cpu_upmigrate_discourage_write_u64(struct cgroup_subsys_state *css,
+				struct cftype *cft, u64 upmigrate_discourage)
+{
+	struct task_group *tg = css_tg(css);
+	int discourage = upmigrate_discourage > 0;
+
+	if (tg->upmigrate_discouraged == discourage)
+		return 0;
+
+	/*
+	 * Revisit big-task classification for tasks of this cgroup. It would
+	 * have been efficient to walk tasks of just this cgroup in running
+	 * state, but we don't have easy means to do that. Walk all tasks in
+	 * running state on all cpus instead and re-visit their big task
+	 * classification.
+	 */
+	get_online_cpus();
+	pre_big_small_task_count_change(cpu_online_mask);
+
+	tg->upmigrate_discouraged = discourage;
+
+	post_big_small_task_count_change(cpu_online_mask);
+	put_online_cpus();
+
+	return 0;
+}
+
+#endif	/* CONFIG_SCHED_HMP */
+
 #ifdef CONFIG_FAIR_GROUP_SCHED
 static int cpu_shares_write_u64(struct cgroup_subsys_state *css,
 				struct cftype *cftype, u64 shareval)
@@ -10296,6 +10335,13 @@ static struct cftype cpu_files[] = {
 		.read_u64 = cpu_notify_on_migrate_read_u64,
 		.write_u64 = cpu_notify_on_migrate_write_u64,
 	},
+#ifdef CONFIG_SCHED_HMP
+	{
+		.name = "upmigrate_discourage",
+		.read_u64 = cpu_upmigrate_discourage_read_u64,
+		.write_u64 = cpu_upmigrate_discourage_write_u64,
+	},
+#endif
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	{
 		.name = "shares",
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index cc16d24352b2..da18be6d7bbe 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2808,14 +2808,29 @@ u64 scale_load_to_cpu(u64 task_load, int cpu)
 	return task_load;
 }
 
+#ifdef CONFIG_CGROUP_SCHED
+
+static inline int upmigrate_discouraged(struct task_struct *p)
+{
+	return task_group(p)->upmigrate_discouraged;
+}
+
+#else
+
+static inline int upmigrate_discouraged(struct task_struct *p)
+{
+	return 0;
+}
+
+#endif
+
 /* Is a task "big" on its current cpu */
 static inline int is_big_task(struct task_struct *p)
 {
 	u64 load = task_load(p);
 	int nice = task_nice(p);
 
-	/* Todo: Provide cgroup-based control as well? */
-	if (nice > sched_upmigrate_min_nice)
+	if (nice > sched_upmigrate_min_nice || upmigrate_discouraged(p))
 		return 0;
 
 	load = scale_load_to_cpu(load, task_cpu(p));
@@ -3002,8 +3017,7 @@ static int task_will_fit(struct task_struct *p, int cpu)
 		if (rq->capacity > prev_rq->capacity)
 			return 1;
 	} else {
-		/* Todo: Provide cgroup-based control as well? */
-		if (nice > sched_upmigrate_min_nice)
+		if (nice > sched_upmigrate_min_nice || upmigrate_discouraged(p))
 			return 1;
 
 		load = scale_load_to_cpu(task_load(p), cpu);
@@ -3981,8 +3995,8 @@ static inline int migration_needed(struct rq *rq, struct task_struct *p)
 	if (is_small_task(p))
 		return 0;
 
-	/* Todo: cgroup-based control? */
-	if (nice > sched_upmigrate_min_nice && rq->capacity > min_capacity)
+	if ((nice > sched_upmigrate_min_nice || upmigrate_discouraged(p)) &&
+			 rq->capacity > min_capacity)
 		return MOVE_TO_LITTLE_CPU;
 
 	if (!task_will_fit(p, cpu_of(rq)))
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index eae7973b37b9..d897c967bb87 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -241,6 +241,9 @@ struct task_group {
 	struct cgroup_subsys_state css;
 
 	bool notify_on_migrate;
+#ifdef CONFIG_SCHED_HMP
+	bool upmigrate_discouraged;
+#endif
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	/* schedulable entities of this group on each cpu */