vmscan: implement swap token priority aging

While testing for memcg aware swap token, I observed a swap token was
often grabbed an intermittent running process (eg init, auditd) and they
never release a token.

Why?

Some processes (eg init, auditd, audispd) wake up when a process exiting.
And swap token can be get first page-in process when a process exiting
makes no swap token owner.  Thus such above intermittent running process
often get a token.

And currently, swap token priority is only decreased at page fault path.
Then, if the process sleep immediately after to grab swap token, the swap
token priority never be decreased.  That's obviously undesirable.

This patch implement very poor (and lightweight) priority aging.  It only
be affect to the above corner case and doesn't change swap tendency
workload performance (eg multi process qsbench load)

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

2 files changed, 23 insertions, 8 deletions

diff --git a/include/trace/events/vmscan.h b/include/trace/events/vmscan.h
index 1798e0cee2a9..b2c33bd955fa 100644
--- a/include/trace/events/vmscan.h
+++ b/include/trace/events/vmscan.h
@@ -366,9 +366,10 @@ DEFINE_EVENT_CONDITION(put_swap_token_template, disable_swap_token,
 
 TRACE_EVENT_CONDITION(update_swap_token_priority,
 	TP_PROTO(struct mm_struct *mm,
-		 unsigned int old_prio),
+		 unsigned int old_prio,
+		 struct mm_struct *swap_token_mm),
 
-	TP_ARGS(mm, old_prio),
+	TP_ARGS(mm, old_prio, swap_token_mm),
 
 	TP_CONDITION(mm->token_priority != old_prio),
 
@@ -376,16 +377,21 @@ TRACE_EVENT_CONDITION(update_swap_token_priority,
 		__field(struct mm_struct*, mm)
 		__field(unsigned int, old_prio)
 		__field(unsigned int, new_prio)
+		__field(struct mm_struct*, swap_token_mm)
+		__field(unsigned int, swap_token_prio)
 	),
 
 	TP_fast_assign(
-		__entry->mm = mm;
-		__entry->old_prio = old_prio;
-		__entry->new_prio = mm->token_priority;
+		__entry->mm		= mm;
+		__entry->old_prio	= old_prio;
+		__entry->new_prio	= mm->token_priority;
+		__entry->swap_token_mm	= swap_token_mm;
+		__entry->swap_token_prio = swap_token_mm ? swap_token_mm->token_priority : 0;
 	),
 
-	TP_printk("mm=%p old_prio=%u new_prio=%u",
-		  __entry->mm, __entry->old_prio, __entry->new_prio)
+	TP_printk("mm=%p old_prio=%u new_prio=%u swap_token_mm=%p token_prio=%u",
+		  __entry->mm, __entry->old_prio, __entry->new_prio,
+		  __entry->swap_token_mm, __entry->swap_token_prio)
 );
 
 #endif /* _TRACE_VMSCAN_H */
diff --git a/mm/thrash.c b/mm/thrash.c
index 17d9e29e4c9c..fabf2d0f5169 100644
--- a/mm/thrash.c
+++ b/mm/thrash.c
@@ -25,10 +25,13 @@
 
 #include <trace/events/vmscan.h>
 
+#define TOKEN_AGING_INTERVAL	(0xFF)
+
 static DEFINE_SPINLOCK(swap_token_lock);
 struct mm_struct *swap_token_mm;
 struct mem_cgroup *swap_token_memcg;
 static unsigned int global_faults;
+static unsigned int last_aging;
 
 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
 static struct mem_cgroup *swap_token_memcg_from_mm(struct mm_struct *mm)
@@ -64,6 +67,11 @@ void grab_swap_token(struct mm_struct *mm)
 	if (!swap_token_mm)
 		goto replace_token;
 
+	if ((global_faults - last_aging) > TOKEN_AGING_INTERVAL) {
+		swap_token_mm->token_priority /= 2;
+		last_aging = global_faults;
+	}
+
 	if (mm == swap_token_mm) {
 		mm->token_priority += 2;
 		goto update_priority;
@@ -81,7 +89,7 @@ void grab_swap_token(struct mm_struct *mm)
 		goto replace_token;
 
 update_priority:
-	trace_update_swap_token_priority(mm, old_prio);
+	trace_update_swap_token_priority(mm, old_prio, swap_token_mm);
 
 out:
 	mm->faultstamp = global_faults;
@@ -94,6 +102,7 @@ replace_token:
 	trace_replace_swap_token(swap_token_mm, mm);
 	swap_token_mm = mm;
 	swap_token_memcg = swap_token_memcg_from_mm(mm);
+	last_aging = global_faults;
 	goto out;
 }