linux-kernel - [PATCH v1 -next 2/3] cgroup/freezer: Reduce redundant propagation for cgroup_propagate

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

Message-ID: <20240905134130.1176443-3-chenridong@huawei.com>
Date: Thu, 5 Sep 2024 13:41:29 +0000
From: Chen Ridong <chenridong@...wei.com>
To: <tj@...nel.org>, <lizefan.x@...edance.com>, <hannes@...xchg.org>,
	<longman@...hat.com>, <adityakali@...gle.com>, <sergeh@...nel.org>,
	<mkoutny@...e.com>
CC: <cgroups@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
	<chenridong@...weicloud.com>
Subject: [PATCH v1 -next 2/3] cgroup/freezer: Reduce redundant propagation for cgroup_propagate_frozen

When a cgroup is frozen/unfrozen, it will always propagate down to up.
However it is unnecessary to propagate to the top every time. This patch
aims to reduce redundant propagation for cgroup_propagate_frozen.

For example, subtree like:
	a
	|
	b
      / | \
     c  d  e
If c is frozen, and d and e are not frozen now, it doesn't have to
propagate to a; Only when c, d and e are all frozen, b and a could be set
to frozen. Therefore, if nr_frozen_descendants is not equal to
nr_descendants, just stop propagate. If a descendant is frozen, the
parent's nr_frozen_descendants add child->nr_descendants + 1. This can
reduce redundant propagation.

Signed-off-by: Chen Ridong <chenridong@...wei.com>
---
 kernel/cgroup/freezer.c | 31 +++++++++++++++++++++++--------
 1 file changed, 23 insertions(+), 8 deletions(-)

diff --git a/kernel/cgroup/freezer.c b/kernel/cgroup/freezer.c
index 02af6c1fa957..e4bcc41b6a30 100644
--- a/kernel/cgroup/freezer.c
+++ b/kernel/cgroup/freezer.c
@@ -13,7 +13,7 @@
  */
 static void cgroup_propagate_frozen(struct cgroup *cgrp, bool frozen)
 {
-	int desc = 1;
+	struct cgroup *child = cgrp;
 
 	/*
 	 * If the new state is frozen, some freezing ancestor cgroups may change
@@ -23,23 +23,38 @@ static void cgroup_propagate_frozen(struct cgroup *cgrp, bool frozen)
 	 */
 	while ((cgrp = cgroup_parent(cgrp))) {
 		if (frozen) {
-			cgrp->freezer.nr_frozen_descendants += desc;
+			/*
+			 * A cgroup is frozen, parent nr frozen descendants should add
+			 * nr cgroups of the entire subtree , including child itself.
+			 */
+			cgrp->freezer.nr_frozen_descendants += child->nr_descendants + 1;
+
+			/*
+			 * If there is other descendant is not frozen,
+			 * cgrp and its parent couldn't be frozen, just break
+			 */
+			if (cgrp->freezer.nr_frozen_descendants !=
+			    cgrp->nr_descendants)
+				break;
+
+			child = cgrp;
 			if (!test_bit(CGRP_FROZEN, &cgrp->flags) &&
-			    test_bit(CGRP_FREEZE, &cgrp->flags) &&
-			    cgrp->freezer.nr_frozen_descendants ==
-			    cgrp->nr_descendants) {
+			    test_bit(CGRP_FREEZE, &cgrp->flags)) {
 				set_bit(CGRP_FROZEN, &cgrp->flags);
 				cgroup_file_notify(&cgrp->events_file);
 				TRACE_CGROUP_PATH(notify_frozen, cgrp, 1);
-				desc++;
 			}
 		} else {
-			cgrp->freezer.nr_frozen_descendants -= desc;
+			cgrp->freezer.nr_frozen_descendants -= (child->nr_descendants + 1);
+
+			child = cgrp;
 			if (test_bit(CGRP_FROZEN, &cgrp->flags)) {
 				clear_bit(CGRP_FROZEN, &cgrp->flags);
 				cgroup_file_notify(&cgrp->events_file);
 				TRACE_CGROUP_PATH(notify_frozen, cgrp, 0);
-				desc++;
+			} else {
+				/* If parent is unfrozen, don't have to propagate more */
+				break;
 			}
 		}
 	}
-- 
2.34.1