This idea is based on discussion with Michal and Johannes from LSF.
1. If soft_limit are all set to MAX, it wastes first three priority iterations
without scanning anything.
2. By default every memcg is eligible for softlimit reclaim, and we can also
set the value to MAX for special memcg which is immune to soft limit reclaim.
There is a behavior change after this patch: (N == DEF_PRIORITY - 2)
A: usage > softlimit B: usage <= softlimit U: softlimit unset
old: reclaim at each priority reclaim when priority < N reclaim when priority < N
new: reclaim at each priority reclaim when priority < N reclaim at each priority
Note: I can leave the counter->soft_limit uninitialized, at least all the
caller of res_counter_init() have the memcg as pre-zeroed structure. However, I
might be better not rely on that.
Signed-off-by: Ying Han <yinghan@xxxxxxxxxx>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
---
kernel/res_counter.c | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/kernel/res_counter.c b/kernel/res_counter.c
index d9ea45e..9cbffce 100644
--- a/kernel/res_counter.c
+++ b/kernel/res_counter.c
@@ -18,7 +18,7 @@ void res_counter_init(struct res_counter *counter, struct res_counter *parent)
{
spin_lock_init(&counter->lock);
counter->limit = RESOURCE_MAX;
- counter->soft_limit = RESOURCE_MAX;
+ counter->soft_limit = 0;
counter->parent = parent;
}
--
1.7.7.3
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>