From 2685027fca387b602ae565bff17895188b803988 Mon Sep 17 00:00:00 2001 From: Waiman Long Date: Wed, 27 Apr 2022 10:54:28 -0400 Subject: cgroup/cpuset: Remove cpus_allowed/mems_allowed setup in cpuset_init_smp() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit There are 3 places where the cpu and node masks of the top cpuset can be initialized in the order they are executed: 1) start_kernel -> cpuset_init() 2) start_kernel -> cgroup_init() -> cpuset_bind() 3) kernel_init_freeable() -> do_basic_setup() -> cpuset_init_smp() The first cpuset_init() call just sets all the bits in the masks. The second cpuset_bind() call sets cpus_allowed and mems_allowed to the default v2 values. The third cpuset_init_smp() call sets them back to v1 values. For systems with cgroup v2 setup, cpuset_bind() is called once. As a result, cpu and memory node hot add may fail to update the cpu and node masks of the top cpuset to include the newly added cpu or node in a cgroup v2 environment. For systems with cgroup v1 setup, cpuset_bind() is called again by rebind_subsystem() when the v1 cpuset filesystem is mounted as shown in the dmesg log below with an instrumented kernel. [ 2.609781] cpuset_bind() called - v2 = 1 [ 3.079473] cpuset_init_smp() called [ 7.103710] cpuset_bind() called - v2 = 0 smp_init() is called after the first two init functions. So we don't have a complete list of active cpus and memory nodes until later in cpuset_init_smp() which is the right time to set up effective_cpus and effective_mems. To fix this cgroup v2 mask setup problem, the potentially incorrect cpus_allowed & mems_allowed setting in cpuset_init_smp() are removed. For cgroup v2 systems, the initial cpuset_bind() call will set the masks correctly. For cgroup v1 systems, the second call to cpuset_bind() will do the right setup. cc: stable@vger.kernel.org Signed-off-by: Waiman Long Tested-by: Feng Tang Reviewed-by: Michal Koutný Signed-off-by: Tejun Heo --- kernel/cgroup/cpuset.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'kernel/cgroup') diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c index 9390bfd9f1cd..71a418858a5e 100644 --- a/kernel/cgroup/cpuset.c +++ b/kernel/cgroup/cpuset.c @@ -3390,8 +3390,11 @@ static struct notifier_block cpuset_track_online_nodes_nb = { */ void __init cpuset_init_smp(void) { - cpumask_copy(top_cpuset.cpus_allowed, cpu_active_mask); - top_cpuset.mems_allowed = node_states[N_MEMORY]; + /* + * cpus_allowd/mems_allowed set to v2 values in the initial + * cpuset_bind() call will be reset to v1 values in another + * cpuset_bind() call when v1 cpuset is mounted. + */ top_cpuset.old_mems_allowed = top_cpuset.mems_allowed; cpumask_copy(top_cpuset.effective_cpus, cpu_active_mask); -- cgit v1.2.3