Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Waiman Long <longman@redhat.com>
To:     Tejun Heo <tj@kernel.org>, Zefan Li <lizefan.x@bytedance.com>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Christian Brauner <brauner@kernel.org>
Cc:     cgroups@vger.kernel.org, linux-kernel@vger.kernel.org,
        Juri Lelli <juri.lelli@redhat.com>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        =?UTF-8?q?Michal=20Koutn=C3=BD?= <mkoutny@suse.com>,
        Giuseppe Scrivano <gscrivan@redhat.com>,
        Waiman Long <longman@redhat.com>
Subject: [PATCH v4 4/5] cgroup/cpuset: Make cpuset_attach_task() skip subpartitions CPUs for top_cpuset
Date:   Tue, 11 Apr 2023 09:36:00 -0400
Message-Id: <20230411133601.2969636-5-longman@redhat.com>
In-Reply-To: <20230411133601.2969636-1-longman@redhat.com>
References: <20230411133601.2969636-1-longman@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Precedence: bulk

It is found that attaching a task to the top_cpuset does not currently
ignore CPUs allocated to subpartitions in cpuset_attach_task(). So the
code is changed to fix that.

Signed-off-by: Waiman Long <longman@redhat.com>
Reviewed-by: Michal Koutný <mkoutny@suse.com>
---
 kernel/cgroup/cpuset.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index 132cdae4f03c..e4ca2dd2b764 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -2540,7 +2540,8 @@ static void cpuset_attach_task(struct cpuset *cs, struct task_struct *task)
 	if (cs != &top_cpuset)
 		guarantee_online_cpus(task, cpus_attach);
 	else
-		cpumask_copy(cpus_attach, task_cpu_possible_mask(task));
+		cpumask_andnot(cpus_attach, task_cpu_possible_mask(task),
+			       cs->subparts_cpus);
 	/*
 	 * can_attach beforehand should guarantee that this doesn't
 	 * fail.  TODO: have a better way to handle failure here
-- 
2.31.1