From: Mike Yuan <me@yhndnzj.com>
Date: Sat, 12 Jul 2025 15:39:14 +0000 (+0200)
Subject: core/cgroup: always submit unit to realize queue if all controllers are being invalidated
X-Git-Tag: v258-rc1~63^2
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=e0d3213e09a97ec2bbf0dd2a6dfe58f08f6cb5d5;p=thirdparty%2Fsystemd.git

core/cgroup: always submit unit to realize queue if all controllers are being invalidated

Alternative to #38190
Fixes #38112
---

diff --git a/src/core/cgroup.c b/src/core/cgroup.c
index c2f6f47e6e0..980d9b0f0ba 100644
--- a/src/core/cgroup.c
+++ b/src/core/cgroup.c
@@ -3950,7 +3950,11 @@ bool unit_invalidate_cgroup(Unit *u, CGroupMask m) {
         if (!crt)
                 return false;
 
-        if (FLAGS_SET(crt->cgroup_invalidated_mask, m)) /* NOP? */
+        /* If all controllers shall be invalidated, let's unconditionally submit the unit to realize queue.
+         * We initialize the field to _CGROUP_MASK_ALL after all, and semantically it makes sense to use
+         * it as a special signal to forcibly re-realize cgroup. */
+        if (m != _CGROUP_MASK_ALL &&
+            FLAGS_SET(crt->cgroup_invalidated_mask, m)) /* NOP? */
                 return false;
 
         crt->cgroup_invalidated_mask |= m;