On Wed, Aug 24, 2022 at 7:09 PM Alexei Starovoitov
[off-list ref] wrote:

On Wed, Aug 24, 2022 at 4:31 PM Hao Luo [off-list ref] wrote:

quoted

+
+       for (i = 0; i < N_CGROUPS; i++) {
+               fd = create_and_get_cgroup(cgroups[i].path);
+               if (!ASSERT_GE(fd, 0, "create_and_get_cgroup"))
+                       return fd;
+
+               cgroups[i].fd = fd;
+               cgroups[i].id = get_cgroup_id(cgroups[i].path);
+
+               /*
+                * Enable memcg controller for the entire hierarchy.
+                * Note that stats are collected for all cgroups in a hierarchy
+                * with memcg enabled anyway, but are only exposed for cgroups
+                * that have memcg enabled.
+                */
+               if (i < N_NON_LEAF_CGROUPS) {
+                       err = enable_controllers(cgroups[i].path, "memory");
+                       if (!ASSERT_OK(err, "enable_controllers"))
+                               return err;
+               }
+       }

It passes BPF CI, but fails in my setup with:

# ./test_progs -t cgroup_hier -vv
bpf_testmod.ko is already unloaded.
Loading bpf_testmod.ko...
Successfully loaded bpf_testmod.ko.
setup_bpffs:PASS:mount 0 nsec
setup_cgroups:PASS:setup_cgroup_environment 0 nsec
setup_cgroups:PASS:get_root_cgroup 0 nsec
setup_cgroups:PASS:create_and_get_cgroup 0 nsec
(cgroup_helpers.c:92: errno: No such file or directory) Enabling
controller memory:
/mnt/cgroup-test-work-dir6526//test/cgroup.subtree_control
setup_cgroups:FAIL:enable_controllers unexpected error: 1 (errno 2)
cleanup_bpffs:FAIL:rmdir /sys/fs/bpf/vmscan/ unexpected error: -1 (errno 2)
#36      cgroup_hierarchical_stats:FAIL
Summary: 0/0 PASSED, 0 SKIPPED, 1 FAILED

How do I debug it?

The failure with ENOENT happens when we try to write "+memory" to
/mnt/cgroup-test-work-dir6526//test/cgroup.subtree_control, not when
we try to open it. So the file is there. AFAICT, ENOENT can be
returned from this write if the memory controller is not enabled on
this cgroup.

In setup_cgroup_environment(), we should be enabling all available
controllers on /mnt and /mnt/cgroup-test-work-dir6526 by this line:

        if (__enable_controllers(CGROUP_MOUNT_PATH, NULL) ||
              __enable_controllers(cgroup_workdir, NULL))
                  return 1;

The first thing that comes to mind is that maybe the memory controller
is not enabled on your setup at all? Can you check
/sys/fs/cgroup/cgroup.controllers (or wherever your global cgroup
mount is)?

I don't know much about namespaces, so I am not sure if the privately
mounted /mnt directory here would be the same as the cgroup root or
not. Maybe we can add a pause() somewhere and check
/mnt/cgroup.controllers as well?