ZEN: Tune mgLRU to protect cache used in the last second

damentz · intersectRaven · commit d70d2e7ec23f · 2023-08-31T09:11:58.000+08:00
Although not identical to the le9 patches that protect a byte-amount of cache through tunables, multigenerational LRU now supports protecting cache accessed in the last X milliseconds. In torvalds#218, Yu recommends starting with 1000ms and tuning as needed. This looks like a safe default and turning on this feature should help users that don't know they need it.
diff --git a/mm/vmscan.c b/mm/vmscan.c
@@ -4594,7 +4594,7 @@ static bool lruvec_is_reclaimable(struct lruvec *lruvec, struct scan_control *sc
 }
 
 /* to protect the working set of the last N jiffies */
-static unsigned long lru_gen_min_ttl __read_mostly;
+static unsigned long lru_gen_min_ttl __read_mostly = HZ;
 
 static void lru_gen_age_node(struct pglist_data *pgdat, struct scan_control *sc)
 {

Original file line number	Diff line number	Diff line change
`@@ -4594,7 +4594,7 @@ static bool lruvec_is_reclaimable(struct lruvec lruvec, struct scan_control sc`
`4594`	`4594`	`}`
`4595`	`4595`
`4596`	`4596`	`/* to protect the working set of the last N jiffies */`
`4597`		`-static unsigned long lru_gen_min_ttl __read_mostly;`
	`4597`	`+static unsigned long lru_gen_min_ttl __read_mostly = HZ;`
`4598`	`4598`
`4599`	`4599`	`static void lru_gen_age_node(struct pglist_data pgdat, struct scan_control sc)`
`4600`	`4600`	`{`