cgroup: wait for cgroup destruction to complete when umount

hulk inclusion
category: bugfix
bugzilla: https://gitee.com/openeuler/kernel/issues/I8QLND


CVE: N/A

-------------------------------------------------

Since commit 3c606d35 ("cgroup: prevent mount hang due to memory
controller lifetime"), a cgroup root won't be destroyed if there are any
child cgroups, dead or alive.

This introduced a small regression.

    # cat test.sh
    mount -t cgroup -o cpuset xxx /cgroup
    mkdir /cgroup/tmp
    rmdir /cgroup/tmp
    umount /cgroup

After running this script, you'll probably find the cgroup hierarchy
is still active.

    # cat /proc/cgroups | grep cpuset
    #subsys_name    hierarchy       num_cgroups     enabled
    cpuset  1       1       1
    ...

Fix this by waiting for a while when umount. Now run the script again
and you'll see:

    # cat /proc/cgroups | grep cpuset
    #subsys_name    hierarchy       num_cgroups     enabled
    cpuset  0       1       1
    ...

Cc: stable@vger.kernel.org # 3.19+
Signed-off-by: Zefan Li <lizefan@huawei.com>
Signed-off-by: chenridong <chenridong@huawei.com>