Add to a few comment of design decision of swap token aging.
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx>
---
mm/thrash.c | 11 +++++++++++
1 files changed, 11 insertions(+), 0 deletions(-)
diff --git a/mm/thrash.c b/mm/thrash.c
index 0504e8a..8832edb 100644
--- a/mm/thrash.c
+++ b/mm/thrash.c
@@ -50,6 +50,17 @@ void grab_swap_token(struct mm_struct *mm)
if (!swap_token_mm)
goto replace_token;
+ /*
+ * Usually, we don't need priority aging because long interval faults
+ * makes priority decrease quickly. But there is one exception. If the
+ * token owner task is sleeping, it never make long interval faults.
+ * Thus, we need a priority aging mechanism instead. The requirements
+ * of priority aging are
+ * 1) An aging interval is reasonable enough long. Too short aging
+ * interval makes quick swap token lost and decrease performance.
+ * 2) The swap token owner task have to get priority aging even if
+ * it's under sleep.
+ */
if ((global_faults - last_aging) > TOKEN_AGING_INTERVAL) {
swap_token_mm->token_priority /= 2;
last_aging = global_faults;
--
1.7.3.1
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxxx For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>