linux-kernel - [PATCH] sched,numa: fix "Examine a task move when examining a task swap"

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Date:	Tue, 5 Aug 2014 17:55:30 -0400
From:	Rik van Riel <riel@...hat.com>
To:	peterz@...radead.org
Cc:	linux-kernel@...r.kernel.org, mgorman@...e.de, mingo@...nel.org
Subject: [PATCH] sched,numa: fix "Examine a task move when examining a task
 swap"

The code in changeset 83230590 can occasionally get tripped up
by the "src_rq->nr_running == 1 && dst_rq->nr_running == 1"
check, and lead to undesirable and/or useless task moves, as
well as idle CPUs.

Luckily this bug can be avoided in a way that also simplifies
the code.

Signed-off-by: Rik van Riel <riel@...hat.com>
---
 kernel/sched/fair.c | 33 +++++++++++++--------------------
 1 file changed, 13 insertions(+), 20 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 376bc07c..32edd2c56 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -1159,7 +1159,7 @@ static void task_numa_compare(struct task_numa_env *env,
 	long src_load, dst_load;
 	long load;
 	long imp = env->p->numa_group ? groupimp : taskimp;
-	long moveimp = imp;
+	long moveimp = imp - 1;
 
 	rcu_read_lock();
 	cur = ACCESS_ONCE(dst_rq->curr);
@@ -1206,8 +1206,18 @@ static void task_numa_compare(struct task_numa_env *env,
 		}
 	}
 
-	if (imp <= env->best_imp && moveimp <= env->best_imp)
-		goto unlock;
+	if (imp <= env->best_imp) {
+		if (moveimp <= env->best_imp)
+			goto unlock;
+
+		/*
+		 * A task swap is not going to work; a task move may be
+		 * required to consolidate this workload, especially if
+		 * both nodes are overloaded and there are no idle CPUs.
+		 */
+		imp = moveimp;
+		cur = NULL;
+	}
 
 	if (!cur) {
 		/* Is there capacity at our destination? */
@@ -1231,23 +1241,6 @@ static void task_numa_compare(struct task_numa_env *env,
 	dst_load = env->dst_stats.load + load;
 	src_load = env->src_stats.load - load;
 
-	if (moveimp > imp && moveimp > env->best_imp) {
-		/*
-		 * If the improvement from just moving env->p direction is
-		 * better than swapping tasks around, check if a move is
-		 * possible. Store a slightly smaller score than moveimp,
-		 * so an actually idle CPU will win.
-		 */
-		if (!load_too_imbalanced(src_load, dst_load, env)) {
-			imp = moveimp - 1;
-			cur = NULL;
-			goto assign;
-		}
-	}
-
-	if (imp <= env->best_imp)
-		goto unlock;
-
 	if (cur) {
 		load = task_h_load(cur);
 		dst_load -= load;

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