Hi Waiman,
Thanks for the patch!
On Sun, Apr 24, 2022 at 10:09:26PM -0400, Waiman Long wrote:
> 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() function just sets all the bits in the masks.
> The last one executed is cpuset_init_smp() which sets up cpu and node
> masks suitable for v1, but not v2. cpuset_bind() does the right setup
> for both v1 and v2 assuming that effective_mems and effective_cpus have
> been set up properly which is not strictly the case here. 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.
>
> To fix this problem, the redundant cpus_allowed and mems_allowed
> mask setup in cpuset_init_smp() are removed. The effective_cpus and
> effective_mems setup there are moved to cpuset_bind().
>
> cc: stable@xxxxxxxxxxxxxxx
> Signed-off-by: Waiman Long <longman@xxxxxxxxxx>
> ---
> kernel/cgroup/cpuset.c | 10 +++-------
> 1 file changed, 3 insertions(+), 7 deletions(-)
>
> diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
> index 9390bfd9f1cd..a2e15a43397e 100644
> --- a/kernel/cgroup/cpuset.c
> +++ b/kernel/cgroup/cpuset.c
> @@ -2961,6 +2961,9 @@ static void cpuset_bind(struct cgroup_subsys_state *root_css)
> percpu_down_write(&cpuset_rwsem);
> spin_lock_irq(&callback_lock);
>
> + cpumask_copy(top_cpuset.effective_cpus, cpu_active_mask);
> + top_cpuset.effective_mems = node_states[N_MEMORY];
> +
> if (is_in_v2_mode()) {
> cpumask_copy(top_cpuset.cpus_allowed, cpu_possible_mask);
> top_cpuset.mems_allowed = node_possible_map;
> @@ -3390,13 +3393,6 @@ 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];
> - top_cpuset.old_mems_allowed = top_cpuset.mems_allowed;
> -
> - cpumask_copy(top_cpuset.effective_cpus, cpu_active_mask);
> - top_cpuset.effective_mems = node_states[N_MEMORY];
IIUC, the init order is:
cpuset_bind()
smp_init()
cpuset_init_smp()
while all cpus except boot cpu is brought up in smp_init(), so I'm
thinking moving the cpus_allowed init from cpuset_init_smp() to
cpuset_bind() may cause some problem.
Thanks,
Feng
