From 7a8045dfd3b91c65e0b5beb967871559839b7a77 Mon Sep 17 00:00:00 2001
From: Tengfei Fan <tengfeif@codeaurora.org>
Date: Tue, 21 Apr 2020 09:01:51 +0800
Subject: [PATCH] sched/walt: Improve the scheduler

This change is for general scheduler improvement.

Change-Id: If03bf5f104740adf09d854837ae9b141dc97cbc6
Signed-off-by: Tengfei Fan <tengfeif@codeaurora.org>
[pkondeti@codeaurora: Added tunable based on priority]
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
---
 include/linux/sched/sysctl.h |  1 +
 kernel/sched/fair.c          |  6 ++++++
 kernel/sched/sched.h         |  6 ++++++
 kernel/sysctl.c              | 18 ++++++++++++++++++
 4 files changed, 31 insertions(+)

diff --git a/include/linux/sched/sysctl.h b/include/linux/sched/sysctl.h
index cd531a27bdf1..d9a15434b2db 100644
--- a/include/linux/sched/sysctl.h
+++ b/include/linux/sched/sysctl.h
@@ -55,6 +55,7 @@ extern unsigned int __weak sysctl_sched_ravg_window_nr_ticks;
 extern unsigned int __weak sysctl_sched_many_wakeup_threshold;
 extern unsigned int __weak sysctl_sched_dynamic_ravg_window_enable;
 extern unsigned int sysctl_sched_prefer_spread;
+extern unsigned int sysctl_walt_rtg_cfs_boost_prio;
 
 extern int
 walt_proc_group_thresholds_handler(struct ctl_table *table, int write,
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 1f5911b12a65..3ad74eafc0c9 100755
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -133,6 +133,7 @@ unsigned int sched_capacity_margin_down[NR_CPUS] = {
 
 #ifdef CONFIG_SCHED_WALT
 __read_mostly unsigned int sysctl_sched_prefer_spread;
+unsigned int sysctl_walt_rtg_cfs_boost_prio = 99; /* disabled by default */
 #endif
 unsigned int sched_small_task_threshold = 102;
 
@@ -4066,6 +4067,11 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
 				se->vruntime = min_vruntime(vruntime,
 							se->vruntime);
 				return;
+			} else if (task_rtg_high_prio(task_of(se))) {
+				vruntime -= thresh;
+				se->vruntime = min_vruntime(vruntime,
+							se->vruntime);
+				return;
 			}
 #endif
 		}
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 1e7f4813e5d7..60dd172bc943 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2991,6 +2991,12 @@ static inline bool task_in_related_thread_group(struct task_struct *p)
 	return (rcu_access_pointer(p->wts.grp) != NULL);
 }
 
+static inline bool task_rtg_high_prio(struct task_struct *p)
+{
+	return task_in_related_thread_group(p) &&
+		(p->prio <= sysctl_walt_rtg_cfs_boost_prio);
+}
+
 static inline struct walt_related_thread_group
 *task_related_thread_group(struct task_struct *p)
 {
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index ea162a9eb555..559147b4e88d 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -150,6 +150,15 @@ const int sched_user_hint_max = 1000;
 static unsigned int ns_per_sec = NSEC_PER_SEC;
 static unsigned int one_hundred_thousand = 100000;
 static unsigned int two_hundred_million = 200000000;
+/*
+ * CFS task prio range is [100 ... 139]
+ * 120 is the default prio.
+ * RTG boost range is [100 ... 119] because giving
+ * boost for [120 .. 139] does not make sense.
+ * 99 means disabled and it is the default value.
+ */
+static unsigned int min_cfs_boost_prio = 99;
+static unsigned int max_cfs_boost_prio = 119;
 #endif
 
 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
@@ -545,6 +554,15 @@ static struct ctl_table kern_table[] = {
 		.extra1		= SYSCTL_ZERO,
 		.extra2		= &two,
 	},
+	{
+		.procname	= "walt_rtg_cfs_boost_prio",
+		.data		= &sysctl_walt_rtg_cfs_boost_prio,
+		.maxlen		= sizeof(unsigned int),
+		.mode		= 0644,
+		.proc_handler   = proc_dointvec_minmax,
+		.extra1		= &min_cfs_boost_prio,
+		.extra2		= &max_cfs_boost_prio,
+	},
 #endif
 #ifdef CONFIG_SCHED_DEBUG
 	{