[RFC] psi: introduce rt property for trigger

Commit Message

Zhaoyang Huang Dec. 15, 2021, 10:49 a.m. UTC

From: Zhaoyang Huang <[e-mail]>

There are several pratical reasons to introduce such a rt flag to let the
trigger wake up within the window,

1. WINDOW_MIN_US=500ms is too big for some scenarios where the trigger is
expected to launch some behavious on the resource under pressure.

2. Window size works as both of average factor and wakeup timing value.
However, user could expect seperate value on this two roles. eg, we expect
to watch the pressure as 'SOME 100/1000ms' and got triggered whenever it
reached.

Signed-off-by: Zhaoyang Huang <[e-mail]>
---
 include/linux/psi_types.h |  3 +++
 kernel/sched/psi.c        | 21 +++++++++++++++++++--
 2 files changed, 22 insertions(+), 2 deletions(-)

Patch

diff --git a/include/linux/psi_types.h b/include/linux/psi_types.h
index 0a23300..6c83c75 100644
--- a/include/linux/psi_types.h
+++ b/include/linux/psi_types.h
@@ -124,6 +124,9 @@  struct psi_trigger {
 	/* Tracking window */
 	struct psi_window win;
 
+	/*real time trigger flag*/
+	u32 rt;
+
 	/*
 	 * Time last event was generated. Used for rate-limiting
 	 * events to one per window
diff --git a/kernel/sched/psi.c b/kernel/sched/psi.c
index 1652f2b..e1f08aa 100644
--- a/kernel/sched/psi.c
+++ b/kernel/sched/psi.c
@@ -145,6 +145,7 @@ 
 #include <linux/psi.h>
 #include "sched.h"
 
+extern unsigned int sysctl_sched_latency;
 static int psi_bug __read_mostly;
 
 DEFINE_STATIC_KEY_FALSE(psi_disabled);
@@ -458,9 +459,12 @@  static void psi_avgs_work(struct work_struct *work)
 static void window_reset(struct psi_window *win, u64 now, u64 value,
 			 u64 prev_growth)
 {
+	struct psi_trigger *t = container_of(win, struct psi_trigger, win);
+
 	win->start_time = now;
 	win->start_value = value;
 	win->prev_growth = prev_growth;
+	t->rt = t->rt ? 5 : 0;
 }
 
 /*
@@ -542,6 +546,17 @@  static u64 update_triggers(struct psi_group *group, u64 now)
 		if (growth < t->threshold)
 			continue;
 
+		/* wakeup if trigger has rt and at least 5 sched_latency surpassed*/
+		if (t->rt && growth >= t->threshold) {
+			u64 rt_trigger_time = min(t->last_event_time + t->win.size,
+				sysctl_sched_latency * t->rt + t->last_event_time);
+			if (now < rt_trigger_time)
+				continue;
+			if (cmpxchg(&t->event, 0, 1) == 0)
+				wake_up_interruptible(&t->event_wait);
+			t->rt += 5;
+		}
+
 		/* Limit event signaling to once per window */
 		if (now < t->last_event_time + t->win.size)
 			continue;
@@ -1116,13 +1131,14 @@  struct psi_trigger *psi_trigger_create(struct psi_group *group,
 	enum psi_states state;
 	u32 threshold_us;
 	u32 window_us;
+	u32 rt = 0;
 
 	if (static_branch_likely(&psi_disabled))
 		return ERR_PTR(-EOPNOTSUPP);
 
-	if (sscanf(buf, "some %u %u", &threshold_us, &window_us) == 2)
+	if (sscanf(buf, "some %u %u %u", &threshold_us, &window_us, &rt) >= 2)
 		state = PSI_IO_SOME + res * 2;
-	else if (sscanf(buf, "full %u %u", &threshold_us, &window_us) == 2)
+	else if (sscanf(buf, "full %u %u %u", &threshold_us, &window_us, &rt) >= 2)
 		state = PSI_IO_FULL + res * 2;
 	else
 		return ERR_PTR(-EINVAL);
@@ -1152,6 +1168,7 @@  struct psi_trigger *psi_trigger_create(struct psi_group *group,
 	t->last_event_time = 0;
 	init_waitqueue_head(&t->event_wait);
 	kref_init(&t->refcount);
+	t->rt = rt ? 5 : 0;
 
 	mutex_lock(&group->trigger_lock);