diff --git a/fs/proc/base.c b/fs/proc/base.c
index cc32c775a5cf..3deb937f8901 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -1488,6 +1488,58 @@ static const struct file_operations proc_pid_sched_group_id_operations = {
 	.release	= single_release,
 };
 
+static int sched_low_latency_show(struct seq_file *m, void *v)
+{
+	struct inode *inode = m->private;
+	struct task_struct *p;
+	bool low_latency;
+
+	p = get_proc_task(inode);
+	if (!p)
+		return -ESRCH;
+
+	low_latency = p->wts.low_latency;
+	seq_printf(m, "%d\n", low_latency);
+
+	put_task_struct(p);
+
+	return 0;
+}
+
+static ssize_t
+sched_low_latency_write(struct file *file, const char __user *buf,
+	    size_t count, loff_t *offset)
+{
+	struct task_struct *p = get_proc_task(file_inode(file));
+	bool low_latency;
+	int err;
+
+	if (!p)
+		return -ESRCH;
+
+	err =  kstrtobool_from_user(buf, count, &low_latency);
+	if (err)
+		goto out;
+
+	p->wts.low_latency = low_latency;
+out:
+	put_task_struct(p);
+	return err < 0 ? err : count;
+}
+
+static int sched_low_latency_open(struct inode *inode, struct file *filp)
+{
+	return single_open(filp, sched_low_latency_show, inode);
+}
+
+static const struct file_operations proc_pid_sched_low_latency_operations = {
+	.open		= sched_low_latency_open,
+	.read		= seq_read,
+	.write		= sched_low_latency_write,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
 #endif	/* CONFIG_SCHED_WALT */
 
 #ifdef CONFIG_SCHED_AUTOGROUP
@@ -3212,6 +3264,7 @@ static const struct pid_entry tgid_base_stuff[] = {
 	REG("sched_group_id", 00666, proc_pid_sched_group_id_operations),
 	REG("sched_boost", 0666,  proc_task_boost_enabled_operations),
 	REG("sched_boost_period_ms", 0666, proc_task_boost_period_operations),
+	REG("sched_low_latency", 00666, proc_pid_sched_low_latency_operations),
 #endif
 #ifdef CONFIG_SCHED_DEBUG
 	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 35cb8f4f6006..96ba8aaaff47 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -618,6 +618,7 @@ struct walt_task_struct {
 	int				boost;
 	bool				wake_up_idle;
 	bool				misfit;
+	bool				low_latency;
 	u64				boost_period;
 	u64				boost_expires;
 	u64				last_sleep_ts;
diff --git a/include/trace/events/sched.h b/include/trace/events/sched.h
index 73a5fa7a6368..dd64373b2d37 100644
--- a/include/trace/events/sched.h
+++ b/include/trace/events/sched.h
@@ -451,48 +451,6 @@ TRACE_EVENT(sched_load_balance_stats,
 #endif /* NR_CPUS > BITS_PER_LONG */
 #endif /* CONFIG_SMP */
 
-#ifdef CONFIG_SCHED_WALT
-TRACE_EVENT(sched_load_balance_skip_tasks,
-
-	TP_PROTO(int scpu, int dcpu, int grp_type, int pid,
-		unsigned long h_load, unsigned long task_util,
-		unsigned long affinity),
-
-	TP_ARGS(scpu, dcpu, grp_type, pid, h_load, task_util, affinity),
-
-	TP_STRUCT__entry(
-		__field(int,            scpu)
-		__field(unsigned long,  src_util_cum)
-		__field(int,            grp_type)
-		__field(int,            dcpu)
-		__field(unsigned long,  dst_util_cum)
-		__field(int,            pid)
-		__field(unsigned long,  affinity)
-		__field(unsigned long,  task_util)
-		__field(unsigned long,  h_load)
-	),
-
-	TP_fast_assign(
-		__entry->scpu           = scpu;
-		__entry->src_util_cum   =
-				cpu_rq(scpu)->wrq.cum_window_demand_scaled;
-		__entry->grp_type       = grp_type;
-		__entry->dcpu           = dcpu;
-		__entry->dst_util_cum   =
-				cpu_rq(dcpu)->wrq.cum_window_demand_scaled;
-		__entry->pid            = pid;
-		__entry->affinity       = affinity;
-		__entry->task_util      = task_util;
-		__entry->h_load         = h_load;
-	),
-
-	TP_printk("source_cpu=%d util_cum=%lu group_type=%d dest_cpu=%d util_cum=%lu pid=%d affinity=%#lx task_util=%lu task_h_load=%lu",
-		__entry->scpu, __entry->src_util_cum, __entry->grp_type,
-		__entry->dcpu, __entry->dst_util_cum, __entry->pid,
-		__entry->affinity, __entry->task_util, __entry->h_load)
-);
-#endif
-
 DECLARE_EVENT_CLASS(sched_process_template,
 
 	TP_PROTO(struct task_struct *p),
@@ -905,7 +863,47 @@ DECLARE_TRACE(sched_overutilized_tp,
 	TP_PROTO(struct root_domain *rd, bool overutilized),
 	TP_ARGS(rd, overutilized));
 
-#ifdef CONFIG_SMP
+#ifdef CONFIG_SCHED_WALT
+TRACE_EVENT(sched_load_balance_skip_tasks,
+
+	TP_PROTO(int scpu, int dcpu, int grp_type, int pid,
+		unsigned long h_load, unsigned long task_util,
+		unsigned long affinity),
+
+	TP_ARGS(scpu, dcpu, grp_type, pid, h_load, task_util, affinity),
+
+	TP_STRUCT__entry(
+		__field(int,            scpu)
+		__field(unsigned long,  src_util_cum)
+		__field(int,            grp_type)
+		__field(int,            dcpu)
+		__field(unsigned long,  dst_util_cum)
+		__field(int,            pid)
+		__field(unsigned long,  affinity)
+		__field(unsigned long,  task_util)
+		__field(unsigned long,  h_load)
+	),
+
+	TP_fast_assign(
+		__entry->scpu           = scpu;
+		__entry->src_util_cum   =
+				cpu_rq(scpu)->wrq.cum_window_demand_scaled;
+		__entry->grp_type       = grp_type;
+		__entry->dcpu           = dcpu;
+		__entry->dst_util_cum   =
+				cpu_rq(dcpu)->wrq.cum_window_demand_scaled;
+		__entry->pid            = pid;
+		__entry->affinity       = affinity;
+		__entry->task_util      = task_util;
+		__entry->h_load         = h_load;
+	),
+
+	TP_printk("source_cpu=%d util_cum=%lu group_type=%d dest_cpu=%d util_cum=%lu pid=%d affinity=%#lx task_util=%lu task_h_load=%lu",
+		__entry->scpu, __entry->src_util_cum, __entry->grp_type,
+		__entry->dcpu, __entry->dst_util_cum, __entry->pid,
+		__entry->affinity, __entry->task_util, __entry->h_load)
+);
+
 TRACE_EVENT(sched_cpu_util,
 
 	TP_PROTO(int cpu),
@@ -1025,6 +1023,7 @@ TRACE_EVENT(sched_task_util,
 		__field(u32,            unfilter)
 		__field(unsigned long,	cpus_allowed)
 		__field(int,            task_boost)
+		__field(bool,		low_latency)
 	),
 
 	TP_fast_assign(
@@ -1043,28 +1042,22 @@ TRACE_EVENT(sched_task_util,
 		__entry->is_rtg                 = is_rtg;
 		__entry->rtg_skip_min           = rtg_skip_min;
 		__entry->start_cpu              = start_cpu;
-#ifdef CONFIG_SCHED_WALT
 		__entry->unfilter               = p->wts.unfilter;
-#else
-		__entry->unfilter               = 0;
-#endif
 		__entry->cpus_allowed		=
 					cpumask_bits(&p->cpus_mask)[0];
-#ifdef CONFIG_SCHED_WALT
 		__entry->task_boost		= per_task_boost(p);
-#else
-		__entry->task_boost		= 0;
-#endif
+		__entry->low_latency		= p->wts.low_latency;
 	),
 
-	TP_printk("pid=%d comm=%s util=%lu prev_cpu=%d candidates=%#lx best_energy_cpu=%d sync=%d need_idle=%d fastpath=%d placement_boost=%d latency=%llu stune_boosted=%d is_rtg=%d rtg_skip_min=%d start_cpu=%d unfilter=%u affinity=%lx task_boost=%d",
+	TP_printk("pid=%d comm=%s util=%lu prev_cpu=%d candidates=%#lx best_energy_cpu=%d sync=%d need_idle=%d fastpath=%d placement_boost=%d latency=%llu stune_boosted=%d is_rtg=%d rtg_skip_min=%d start_cpu=%d unfilter=%u affinity=%lx task_boost=%d low_latency=%d",
 		__entry->pid, __entry->comm, __entry->util, __entry->prev_cpu,
 		__entry->candidates, __entry->best_energy_cpu, __entry->sync,
 		__entry->need_idle, __entry->fastpath, __entry->placement_boost,
 		__entry->latency, __entry->uclamp_boosted,
 		__entry->is_rtg, __entry->rtg_skip_min, __entry->start_cpu,
-		__entry->unfilter, __entry->cpus_allowed, __entry->task_boost)
-)
+		__entry->unfilter, __entry->cpus_allowed, __entry->task_boost,
+		__entry->low_latency)
+);
 
 /*
  * Tracepoint for find_best_target
@@ -1120,12 +1113,10 @@ TRACE_EVENT(sched_find_best_target,
 		  __entry->skip,
 		  __entry->running)
 );
-#endif /* CONFIG_SMP */
 
 /*
  * Tracepoint for system overutilized flag
  */
-#ifdef CONFIG_SCHED_WALT
 struct sched_domain;
 TRACE_EVENT_CONDITION(sched_overutilized,
 
@@ -1148,7 +1139,7 @@ TRACE_EVENT_CONDITION(sched_overutilized,
 	TP_printk("overutilized=%d sd_span=%s",
 		__entry->overutilized ? 1 : 0, __entry->cpulist)
 );
-#endif
+#endif /* CONFIG_SCHED_WALT */
 
 #endif /* _TRACE_SCHED_H */
 
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 4cd0c831da1e..8178439f2d9f 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2800,10 +2800,11 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p)
 	p->se.vruntime			= 0;
 #ifdef CONFIG_SCHED_WALT
 	p->wts.last_sleep_ts		= 0;
-	p->wts.wake_up_idle			= false;
+	p->wts.wake_up_idle		= false;
 	p->wts.boost			= 0;
 	p->wts.boost_expires		= 0;
-	p->wts.boost_period			= 0;
+	p->wts.boost_period		= 0;
+	p->wts.low_latency		= 0;
 #endif
 	INIT_LIST_HEAD(&p->se.group_node);
 
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index d2aec043fe62..8cb1cb2906d6 100755
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4055,9 +4055,20 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
 			thresh >>= 1;
 
 		vruntime -= thresh;
-		if (entity_is_task(se) && per_task_boost(task_of(se)) ==
-				TASK_BOOST_STRICT_MAX)
-			vruntime -= sysctl_sched_latency;
+		if (entity_is_task(se)) {
+			if (per_task_boost(task_of(se)) ==
+						TASK_BOOST_STRICT_MAX)
+				vruntime -= sysctl_sched_latency;
+#ifdef CONFIG_SCHED_WALT
+			else if (unlikely(task_of(se)->wts.low_latency)) {
+				vruntime -= sysctl_sched_latency;
+				vruntime -= thresh;
+				se->vruntime = min_vruntime(vruntime,
+							se->vruntime);
+				return;
+			}
+#endif
+		}
 	}
 
 	/* ensure we never gain time by being placed backwards. */