[Bug] memory controller missing on ubuntu

[HusterWan]

Ⅰ. Issue Description

We have a test case in: https://github.com/alibaba/pouch/blob/master/test/cli_run_cgroup_test.go#L36

This case is to test CgroupParent path, but we got an error on ubuntu:

FAIL: cli_run_cgroup_test.go:36: PouchRunCgroupSuite.TestRunWithCgroupParent
cli_run_cgroup_test.go:38:
    // cgroup-parent relative path
    testRunWithCgroupParent(c, "pouch",
        "TestRunWithRelativePathOfCgroupParent")
cli_run_cgroup_test.go:67:
    c.Fatalf("failed to Stat container %s cgroup mountpoint %s: %v", name, file, err)
... Error: failed to Stat container TestRunWithRelativePathOfCgroupParent cgroup mountpoint /sys/fs/cgroup/memory/pouch/eca63851d772b647efb8491a189e7268196c571e7faefef381ce2fa2215ceb5a/memory.limit_in_bytes: stat /sys/fs/cgroup/memory/pouch/eca63851d772b647efb8491a189e7268196c571e7faefef381ce2fa2215ceb5a/memory.limit_in_bytes: no such file or directory

This bug may relate to: https://stackoverflow.com/questions/15604055/lxc-cgroup-memory-controller-missing

Ⅱ. Describe what happened

Ⅲ. Describe what you expected to happen

Ⅳ. How to reproduce it (as minimally and precisely as possible)

Ⅴ. Anything else we need to know?

Ⅵ. Environment:

• pouch version (use pouch version):
• OS (e.g. from /etc/os-release):
• Kernel (e.g. uname -a):
• Install tools:
• Others:

[Ace-Tang]

This bug may relate to: https://stackoverflow.com/questions/15604055/lxc-cgroup-memory-controller-missing

From checking runc code, this may not related to the linked url

// GetOwnCgroup returns the relative path to the cgroup docker is running in.
func GetOwnCgroup(subsystem string) (string, error) {
    cgroups, err := ParseCgroupFile("/proc/self/cgroup")
    if err != nil {
        return "", err 
    }   

    return getControllerPath(subsystem, cgroups)
}

in ubuntu 16.04, the /proc/self/cgroup shows like

12:freezer:/
11:pids:/user.slice/user-1000.slice/user@1000.service
10:perf_event:/
9:devices:/user.slice
8:cpuset:/
7:hugetlb:/
6:rdma:/
5:net_cls,net_prio:/
4:blkio:/user.slice
3:memory:/user.slice
2:cpu,cpuacct:/user.slice
1:name=systemd:/user.slice/user-1000.slice/user@1000.service/gnome-terminal-server.service
0::/user.slice/user-1000.slice/user@1000.service/gnome-terminal-server.service

this means we can not get cgroup path like /sys/fs/cgroup + subsystem

[Ace-Tang]

I make a mistake to find container cgroup path, since when container use cgroup namespace, we can not get cgroup path from container use exec cat proc/self/cgroups. I think the correct way to get container host cgroup path is get from runc