fix: handling of cgroup driver setting

[z63d]

fix: #821

Fixes an issue where cgroupfs was always used as the cgroup driver.
Receives SystemdCgroup from containerd and allows you to configure the cgroup driver. (There is a deserialize issue, so we'll use a regular expression to work around it.)
Uses systemd as the default cgroup driver.

[Mossaka]

It looks like one of the test failed with the following error:

Error: failed to stop unit youki-test.scope: dbus error: dbus function call error: Unit youki-test.scope not loaded.

Caused by:
    failed to stop unit youki-test.scope: dbus error: dbus function call error: Unit youki-test.scope not loaded.

Stack backtrace:
   0: anyhow::error::<impl core::convert::From<E> for anyhow::Error>::from
   1: <core::result::Result<T,F> as core::ops::try_trait::FromResidual<core::result::Result<core::convert::Infallible,E>>>::from_residual
   2: containerd_shim_wasm::testing::WasiTest<WasiInstance>::wait
   3: containerd_shim_wasmedge::wasmedge_tests::test_seccomp::{{closure}}
   4: core::ops::function::FnOnce::call_once
   5: serial_test::serial_code_lock::local_serial_core_with_return
   6: containerd_shim_wasmedge::wasmedge_tests::test_seccomp
   7: containerd_shim_wasmedge::wasmedge_tests::test_seccomp::{{closure}}
   8: core::ops::function::FnOnce::call_once
   9: test::__rust_begin_short_backtrace
  10: test::run_test::{{closure}}
  11: std::sys::backtrace::__rust_begin_short_backtrace
  12: core::ops::function::FnOnce::call_once{{vtable.shim}}
  13: std::sys::pal::unix::thread::Thread::new::thread_start
test wasmedge_tests::test_seccomp ... FAILED
test wasmedge_tests::test_unreachable ... ok

failures:

failures:
    wasmedge_tests::test_seccomp

[z63d]

hmm...
This seems to have been a temporary issue. I ran it again and it passed without any issues.

But other problems are occurring...
This did not happen when I was testing with K3s.

  Warning  FailedCreatePodSandBox  6s (x21 over 4m47s)  kubelet            Failed to create pod sandbox: rpc error: code = Unknown desc = failed to create containerd task: failed to create shim task: failed to create container: intermediate process error cgroup error: systemd error: dbus error: dbus connection error: ENOENT: No such file or directory. error during cleanup: systemd error: dbus error: dbus connection error: ENOENT: No such file or directory: unknown

[jprendes]

I didn't fully follow the discussion in Slack about this, but I would expect runtime_path to be a path, while this is not a path.

[z63d]

I don't understand why the kind test fails.
This is what I wanted to do. Set runtime_type to path.
Is this difference okay?

runwasi/test/k3s/bootstrap.sh

Line 8 in 472f431

runtime_type = "$PWD/dist/bin/containerd-shim-$1-v1"

runwasi/test/k8s/Dockerfile

Line 20 in 472f431

runtime_type = "io.containerd.${RUNTIME}.v1"

[jprendes]

Would you mind trying this?
It needs 2 new dependencies

prost = "0.13"
toml = "0.8"

        #[derive(Deserialize, Default, Clone, PartialEq)]
        struct Config {
            #[serde(alias = "SystemdCgroup")]
            systemd_cgroup: bool,
        }
        
        fn get_config(options: Option<&Any>) -> anyhow::Result<Config> {
            let Some(opts) = options else {
                return Ok(Default::default());
            };

            ensure!(
                opts.type_url == "runtimeoptions.v1.Options",
                "Invalid options type {}",
                opts.type_url
            );

            #[derive(Message, Clone, PartialEq)]
            struct Options {
                #[prost(string)]
                type_url: String,
                #[prost(string)]
                config_path: String,
                #[prost(string)]
                config_body: String,
            }

            let opts = Options::decode(opts.value.as_slice())?;

            let config = toml::from_str(opts.config_body.as_str()).map_err(|err| {
                Error::InvalidArgument(format!("invalid shim options: {err}"))
            })?;

            Ok(config)
        }

        let config = get_config(req.options.as_ref()).map_err(|err| {
            Error::InvalidArgument(format!("invalid shim options: {err}"))
        })?;
        let systemd_cgroup = config.systemd_cgroup;

[z63d]

This is great. Works perfectly!
Would it be better to separate theOptions and Config logic into different files?

[z63d]

This error occurs randomly.
It doesn't seem to be a specific test.
Although the probability is low, there are cases where it does not occur.

Error: failed to stop unit youki-test.scope: dbus error: dbus function call error: Unit youki-test.scope not loaded.

Caused by:
    failed to stop unit youki-test.scope: dbus error: dbus function call error: Unit youki-test.scope not loaded.

Probably this happens too.
I followed the Makefile but it doesn't work locally so I haven't tried it yet.
https://github.com/containerd/runwasi/actions/runs/13409133082/job/37455738328

hmm... This seems to have been a temporary issue. I ran it again and it passed without any issues.

But other problems are occurring... This did not happen when I was testing with K3s.

  Warning  FailedCreatePodSandBox  6s (x21 over 4m47s)  kubelet            Failed to create pod sandbox: rpc error: code = Unknown desc = failed to create containerd task: failed to create shim task: failed to create container: intermediate process error cgroup error: systemd error: dbus error: dbus connection error: ENOENT: No such file or directory. error during cleanup: systemd error: dbus error: dbus connection error: ENOENT: No such file or directory: unknown