[PATCH-next v2 08/10] cgroup/cpuset: Update isolated_cpus if necessary in compute_partition_effective_cpumask()

[Waiman Long <longman@redhat.com>]

It is found by sashiko that when a child isolated partition with parent
root partition is invalidated in compute_partition_effective_cpumask(),
the switching of the isolated exclusive CPUs from child to parent will
not be reflected in isolated_cpus leading to CPU leakage. Fix that by
calling isolated_cpus_update() when that happens.

Also simplify compute_partition_effective_cpumask() by doing
child partition invalidation only for the special case of no more
effective_cpus left in a populated parent. The other case doesn't
actually warrant invalidation at this stage and subsequent iteration
in update_cpumasks_hier() will handle that child later on anyway.

Also drop the make_partition_invalid() helper by inlining it into
compute_partition_effective_cpumask() as it is the only caller.

A new test case is add to test_cpuset_prs.sh to verify this.

Fixes: 11e5f407b64a ("cgroup/cpuset: Keep track of CPUs in isolated partitions")
Signed-off-by: Waiman Long <longman@redhat.com>
---
 kernel/cgroup/cpuset.c                        | 46 +++++++------------
 .../selftests/cgroup/test_cpuset_prs.sh       |  4 ++
 2 files changed, 21 insertions(+), 29 deletions(-)

diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index 2c3a148fa40b..03c666f658e4 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -254,15 +254,6 @@ static inline bool cs_is_member(const struct cpuset *cs)
 	return cs->partition_root_state == PRS_MEMBER;
 }
 
-/*
- * Callers should hold callback_lock to modify partition_root_state.
- */
-static inline void make_partition_invalid(struct cpuset *cs)
-{
-	if (cs->partition_root_state > 0)
-		cs->partition_root_state = -cs->partition_root_state;
-}
-
 /*
  * Send notification event of whenever partition_root_state changes.
  */
@@ -2106,8 +2097,6 @@ static int update_parent_effective_cpumask(struct cpuset *cs, int cmd,
  * or update_cpumasks_hier() where parent and children are modified
  * successively, we don't need to call update_parent_effective_cpumask()
  * and the child's effective_cpus will be updated in later iterations.
- *
- * Note that rcu_read_lock() is assumed to be held.
  */
 static void compute_partition_effective_cpumask(struct cpuset *cs,
 						struct cpumask *new_ecpus)
@@ -2119,15 +2108,16 @@ static void compute_partition_effective_cpumask(struct cpuset *cs,
 	WARN_ON_ONCE(!is_partition_valid(cs));
 
 	/*
-	 * Check child partition roots to see if they should be
-	 * invalidated when
-	 *  1) child effective_xcpus not a subset of new
-	 *     excluisve_cpus
-	 *  2) All the effective_cpus will be used up and cp
-	 *     has tasks
+	 * Check child partition roots to see if they should be invalidated
+	 * when all the effective_cpus will be used up and cs is populated.
+	 * It is better to sacrifice a child partition root than invalidating
+	 * the whole subtree.
+	 *
+	 * Return early if the effective_cpus is already empty.
 	 */
 	compute_excpus(cs, new_ecpus);
-	cpumask_and(new_ecpus, new_ecpus, cpu_active_mask);
+	if (!cpumask_and(new_ecpus, new_ecpus, cpu_active_mask))
+		return;
 
 	rcu_read_lock();
 	cpuset_for_each_child(child, css, cs) {
@@ -2139,22 +2129,20 @@ static void compute_partition_effective_cpumask(struct cpuset *cs,
 		 * partition root.
 		 */
 		WARN_ON_ONCE(is_remote_partition(child));
-		WRITE_ONCE(child->prs_err, 0);
-		if (!cpumask_subset(child->effective_xcpus,
-				    cs->effective_xcpus))
-			WRITE_ONCE(child->prs_err, PERR_INVCPUS);
-		else if (populated &&
-			 cpumask_subset(new_ecpus, child->effective_xcpus))
-			WRITE_ONCE(child->prs_err, PERR_NOCPUS);
-
-		if (child->prs_err) {
+		if (populated && cpumask_subset(new_ecpus, child->effective_xcpus)) {
 			int old_prs = child->partition_root_state;
+			int parent_prs = cs->partition_root_state;
 
 			/*
-			 * Invalidate child partition
+			 * Invalidate child partition and update isolated_cpus
+			 * if necessary.
 			 */
 			spin_lock_irq(&callback_lock);
-			make_partition_invalid(child);
+			WRITE_ONCE(child->prs_err, PERR_NOCPUS);
+			if ((old_prs == PRS_ISOLATED) && (parent_prs == PRS_ROOT))
+				isolated_cpus_update(old_prs, parent_prs,
+						     child->effective_xcpus);
+			child->partition_root_state = -old_prs;
 			spin_unlock_irq(&callback_lock);
 			notify_partition_change(child, old_prs);
 			continue;
diff --git a/tools/testing/selftests/cgroup/test_cpuset_prs.sh b/tools/testing/selftests/cgroup/test_cpuset_prs.sh
index f0d98da3cfd8..d7d0bef0a17a 100755
--- a/tools/testing/selftests/cgroup/test_cpuset_prs.sh
+++ b/tools/testing/selftests/cgroup/test_cpuset_prs.sh
@@ -518,6 +518,10 @@ TEST_MATRIX=(
 	" CX0-4:P2 CX2-4:P2 CX3-4:P2 .     .     X2     .      .     0 A1:0-1,3-4|A2:2|A3:2 \
 								       A1:P2|A2:P2|A3:P-2 0-4"
 
+	# When an isolated partition child with root partition parent is invalidated,
+	# its effective xcpus will be removed from isolated_cpus.
+	" C1-4:P1 C3-4:P2   .      .    T:C3-4   .      .      .     0 A1:3-4|A2:3-4 A1:P1|A2:P-2 ."
+
 	#
 	#  old-A1 old-A2 old-A3 old-B1 new-A1 new-A2 new-A3 new-B1 fail ECPUs Pstate ISOLCPUS
 	#  ------ ------ ------ ------ ------ ------ ------ ------ ---- ----- ------ --------
-- 
2.56.0