Skip to contentSkip to conversation

sched remove lb_stopbalance counter

PATCH In review Chen, Kenneth W 2 participants 5 messages Dec 5, 2006 Scheduler

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 sched remove lb_stopbalance counter

Regarding to sched-decrease-number-of-load-balances.patch currently in -mm tree: I would like to revert the change on adding lb_stopbalance counter. This count can be calculated by: lb_balanced - lb_nobusyg - lb_nobusyq. There is no need to create gazillion counters while we can derive the value. I'm more of against changing sched-stat format unless it is absolutely necessary as all user land tool parsing /proc/schedstat needs to be updated and it's a real pain trying to keep multiple versions of it.

include/linux/sched.h renamed from include/linux/sched.h.orig +0 -1
@@ -684,7 +684,6 @@ struct sched_domain {
 	unsigned long lb_hot_gained[MAX_IDLE_TYPES];
 	unsigned long lb_nobusyg[MAX_IDLE_TYPES];
 	unsigned long lb_nobusyq[MAX_IDLE_TYPES];
-	unsigned long lb_stopbalance[MAX_IDLE_TYPES];
 
 	/* Active load balancing */
 	unsigned long alb_cnt;
kernel/sched.c renamed from kernel/sched.c.orig +4 -7
@@ -428,7 +428,7 @@ static inline void task_rq_unlock(struct
  * bump this up when changing the output format or the meaning of an existing
  * format, so that tools can adapt (or abort)
  */
-#define SCHEDSTAT_VERSION 13
+#define SCHEDSTAT_VERSION 12
 
 static int show_schedstat(struct seq_file *seq, void *v)
 {
@@ -466,7 +466,7 @@ static int show_schedstat(struct seq_fil
 			seq_printf(seq, "domain%d %s", dcnt++, mask_str);
 			for (itype = SCHED_IDLE; itype < MAX_IDLE_TYPES;
 					itype++) {
-				seq_printf(seq, " %lu %lu %lu %lu %lu %lu %lu %lu %lu",
+				seq_printf(seq, " %lu %lu %lu %lu %lu %lu %lu %lu",
 				    sd->lb_cnt[itype],
 				    sd->lb_balanced[itype],
 				    sd->lb_failed[itype],
@@ -474,8 +474,7 @@ static int show_schedstat(struct seq_fil
 				    sd->lb_gained[itype],
 				    sd->lb_hot_gained[itype],
 				    sd->lb_nobusyq[itype],
-				    sd->lb_nobusyg[itype],
-				    sd->lb_stopbalance[itype]);
+				    sd->lb_nobusyg[itype]);
 			}
 			seq_printf(seq, " %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu\n",
 			    sd->alb_cnt, sd->alb_failed, sd->alb_pushed,
@@ -2579,10 +2578,8 @@ redo:
 	group = find_busiest_group(sd, this_cpu, &imbalance, idle, &sd_idle,
 				   &cpus, balance);
 
-	if (*balance == 0) {
-		schedstat_inc(sd, lb_stopbalance[idle]);
+	if (*balance == 0)
 		goto out_balanced;
-	}
 
 	if (!group) {
 		schedstat_inc(sd, lb_nobusyg[idle]);
Acked-by on v1 1/1
#1 lore raw
Quoted 9 lines · Chen, Kenneth W

in -mm tree: I would like to revert the change on adding lb_stopbalance counter. This count can be calculated by: lb_balanced
- lb_nobusyg - lb_nobusyq. There is no need to create gazillion counters while we can derive the value. I'm more of against changing sched-stat format unless it is absolutely necessary as all user land tool parsing /proc/schedstat needs to be updated and it's a real pain trying to keep multiple versions of it.

but, please:

kernel/sched.c static inline void task_rq_unlock(struct · line 431

Quoted 2 lines
kernel/sched.c renamed from kernel/sched.c.orig +1 -1
-#define SCHEDSTAT_VERSION 13
+#define SCHEDSTAT_VERSION 12

change this to 14 instead. Versions should only go upwards, even if we revert to an earlier output format.

Ingo

#2 lore raw

Ingo Molnar wrote on Tuesday, December 05, 2006 7:32 AM

Quoted 13 lines · Chen, Kenneth W
Quoted 9 lines

in -mm tree: I would like to revert the change on adding lb_stopbalance counter. This count can be calculated by: lb_balanced
- lb_nobusyg - lb_nobusyq. There is no need to create gazillion counters while we can derive the value. I'm more of against changing sched-stat format unless it is absolutely necessary as all user land tool parsing /proc/schedstat needs to be updated and it's a real pain trying to keep multiple versions of it.

Quoted 7 lines

but, please:

kernel/sched.c static inline void task_rq_unlock(struct · line 431

Quoted 2 lines
kernel/sched.c renamed from kernel/sched.c.orig +1 -1
-#define SCHEDSTAT_VERSION 13
+#define SCHEDSTAT_VERSION 12

change this to 14 instead. Versions should only go upwards, even if we revert to an earlier output format.

Really? sched-decrease-number-of-load-balances.patch has not yet hit the mainline and I think it's in -mm for only a couple of weeks. I'm trying to back out the change after brief reviewing the patch.

- Ken

#3 lore raw
Quoted 11 lines · Chen, Kenneth W
Quoted 7 lines

but, please:

kernel/sched.c static inline void task_rq_unlock(struct · line 431

Quoted 2 lines
kernel/sched.c renamed from kernel/sched.c.orig +1 -1
-#define SCHEDSTAT_VERSION 13
+#define SCHEDSTAT_VERSION 12

change this to 14 instead. Versions should only go upwards, even if we revert to an earlier output format.

Really? sched-decrease-number-of-load-balances.patch has not yet hit the mainline and I think it's in -mm for only a couple of weeks. I'm trying to back out the change after brief reviewing the patch.

not a big issue but it costs nothing to go to version 14 - OTOH if any utility has been updated to version 13 and is forgotten about, it might break spuriously if we again go to 13 in the future.

Ingo

author Acked-by Ingo Molnar
#4 lore raw

Ingo Molnar wrote on Tuesday, December 05, 2006 7:42 AM

Quoted 9 lines · Chen, Kenneth W
Quoted 8 lines
Quoted 7 lines

but, please:

Review

Quoted 2 lines
(quoted diff) +1 -1
-#define SCHEDSTAT_VERSION 13
+#define SCHEDSTAT_VERSION 12

change this to 14 instead. Versions should only go upwards, even if we revert to an earlier output format.

Quoted 7 lines
Quoted 3 lines

Really? sched-decrease-number-of-load-balances.patch has not yet hit the mainline and I think it's in -mm for only a couple of weeks. I'm trying to back out the change after brief reviewing the patch.

not a big issue but it costs nothing to go to version 14 - OTOH if any utility has been updated to version 13 and is forgotten about, it might break spuriously if we again go to 13 in the future.

OK, with a reluctant agreement, I've changed the version to 14.

[patch] sched: remove lb_stopbalance counter

Remove scheduler stats lb_stopbalance counter. This counter can be calculated by: lb_balanced - lb_nobusyg - lb_nobusyq. There is no need to create gazillion counters while we can derive the value.

include/linux/sched.h renamed from include/linux/sched.h.orig +0 -1
@@ -684,7 +684,6 @@ struct sched_domain {
 	unsigned long lb_hot_gained[MAX_IDLE_TYPES];
 	unsigned long lb_nobusyg[MAX_IDLE_TYPES];
 	unsigned long lb_nobusyq[MAX_IDLE_TYPES];
-	unsigned long lb_stopbalance[MAX_IDLE_TYPES];
 
 	/* Active load balancing */
 	unsigned long alb_cnt;
kernel/sched.c renamed from kernel/sched.c.orig +4 -7
@@ -428,7 +428,7 @@ static inline void task_rq_unlock(struct
  * bump this up when changing the output format or the meaning of an existing
  * format, so that tools can adapt (or abort)
  */
-#define SCHEDSTAT_VERSION 13
+#define SCHEDSTAT_VERSION 14
 
 static int show_schedstat(struct seq_file *seq, void *v)
 {
@@ -466,7 +466,7 @@ static int show_schedstat(struct seq_fil
 			seq_printf(seq, "domain%d %s", dcnt++, mask_str);
 			for (itype = SCHED_IDLE; itype < MAX_IDLE_TYPES;
 					itype++) {
-				seq_printf(seq, " %lu %lu %lu %lu %lu %lu %lu %lu %lu",
+				seq_printf(seq, " %lu %lu %lu %lu %lu %lu %lu %lu",
 				    sd->lb_cnt[itype],
 				    sd->lb_balanced[itype],
 				    sd->lb_failed[itype],
@@ -474,8 +474,7 @@ static int show_schedstat(struct seq_fil
 				    sd->lb_gained[itype],
 				    sd->lb_hot_gained[itype],
 				    sd->lb_nobusyq[itype],
-				    sd->lb_nobusyg[itype],
-				    sd->lb_stopbalance[itype]);
+				    sd->lb_nobusyg[itype]);
 			}
 			seq_printf(seq, " %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu\n",
 			    sd->alb_cnt, sd->alb_failed, sd->alb_pushed,
@@ -2579,10 +2578,8 @@ redo:
 	group = find_busiest_group(sd, this_cpu, &imbalance, idle, &sd_idle,
 				   &cpus, balance);
 
-	if (*balance == 0) {
-		schedstat_inc(sd, lb_stopbalance[idle]);
+	if (*balance == 0)
 		goto out_balanced;
-	}
 
 	if (!group) {
 		schedstat_inc(sd, lb_nobusyg[idle]);