[PATCH] mm: Write down design and intentions in English for proportial scan

[Minchan Kim]

Quote from Andrew
"
That code is absolutely awful :( It's terribly difficult to work out
what the design is - what the code is actually setting out to achieve.
One is reduced to trying to reverse-engineer the intent from the
implementation and that becomes near impossible when the
implementation has bugs!

<snip>

The only way we're going to fix all this up is to stop looking at the
code altogether.  Write down the design and the intentions in English.
Review that.  Then implement that design in C
"

One thing I know it might not be English Andrew expected but other
thing I know is there are lots of native people in here so one of them
will spend his time to make this horrible English smooth.

I alreday spent my time to try to fix this situation so it's your
turn. It's good time balance between non-native and native people
so open source community is fair for all of us.

Signed-off-by: Minchan Kim <minchan@xxxxxxxxxx>
---
 mm/vmscan.c | 37 +++++++++++++++++++++++++++++++------
 1 file changed, 31 insertions(+), 6 deletions(-)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index 521f7eab1798..3a9862895a64 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -2071,6 +2071,22 @@ static void shrink_lruvec(struct lruvec *lruvec, struct scan_control *sc)
 
 	get_scan_count(lruvec, sc, nr);
 
+	/*
+	 * Basically, VM scans file/anon LRU list proportionally depending
+	 * on the value of vm.swappiness but doesn't want to reclaim
+	 * excessive pages. So, it might be better to stop scan as soon as
+	 * we meet nr_to_reclaim but it breaks aging balance between LRU lists
+	 * To keep the balance, what we do is as follows.
+	 *
+	 * If we reclaim pages requested, we stop scanning first and
+	 * investigate which LRU should be scanned more depending on
+	 * remained lru size(ie, nr[xxx]). We will scan bigger one more
+	 * so, final goal is
+	 *
+	 * scanned[xxx] = targets[xxx] - nr[xxx]
+	 * targets[anon] : targets[file] = scanned[anon] : scanned[file]
+	 */
+
 	/* Record the original scan target for proportional adjustments later */
 	memcpy(targets, nr, sizeof(nr));
 
@@ -2091,8 +2107,14 @@ static void shrink_lruvec(struct lruvec *lruvec, struct scan_control *sc)
 	blk_start_plug(&plug);
 	while (nr[LRU_INACTIVE_ANON] || nr[LRU_ACTIVE_FILE] ||
 					nr[LRU_INACTIVE_FILE]) {
-		unsigned long nr_anon, nr_file, percentage;
+		unsigned long nr_anon, nr_file;
 		unsigned long nr_scanned;
+		/*
+		 * How many pages we should scan to meet target
+		 * calculated by get_scan_count. It means that
+		 * (100 - percentage) = already scanned ratio
+		 */
+		unsigned percentage;
 
 		for_each_evictable_lru(lru) {
 			if (nr[lru]) {
@@ -2108,11 +2130,10 @@ static void shrink_lruvec(struct lruvec *lruvec, struct scan_control *sc)
 			continue;
 
 		/*
-		 * For kswapd and memcg, reclaim at least the number of pages
-		 * requested. Ensure that the anon and file LRUs are scanned
-		 * proportionally what was requested by get_scan_count(). We
-		 * stop reclaiming one LRU and reduce the amount scanning
-		 * proportional to the original scan target.
+		 * Here, we reclaimed at least the number of pages requested.
+		 * Then, what we should do is the ensure that the anon and
+		 * file LRUs are scanned proportionally what was requested
+		 * by get_scan_count().
 		 */
 		nr_file = nr[LRU_INACTIVE_FILE] + nr[LRU_ACTIVE_FILE];
 		nr_anon = nr[LRU_INACTIVE_ANON] + nr[LRU_ACTIVE_ANON];
@@ -2126,6 +2147,10 @@ static void shrink_lruvec(struct lruvec *lruvec, struct scan_control *sc)
 		if (!nr_file || !nr_anon)
 			break;
 
+		/*
+		 * Scan the bigger of the LRU more while stop scanning
+		 * the smaller of the LRU to keep aging balance between LRUs
+		 */
 		if (nr_file > nr_anon) {
 			unsigned long scan_target = targets[LRU_INACTIVE_ANON] +
 						targets[LRU_ACTIVE_ANON] + 1;
-- 
2.0.0

-- 
Kind regards,
Minchan Kim
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/