([#33684](https://github.com/kubernetes/kubernetes/pull/33684),
  [@fraenkel](https://github.com/fraenkel))
* Flag gate faster evictions based on node memory pressure using kernel memcg
  notifications - `--experimental-kernel-memcg-notification`.
  ([#38258](https://github.com/kubernetes/kubernetes/pull/38258),
  [@derekwaynecarr](https://github.com/derekwaynecarr))

  
  When the feature gate is enabled, kubelet ignores errors that happens during setting the following sysctl values: `vm.overcommit_memory`, `vm.panic_on_oom`, `kernel.panic`, `kernel.panic_on_oops`, `kernel.keys.root_maxkeys`, `kernel.keys.root_maxbytes`. (These sysctl values for the host, not for the containers)
  
  kubelet also ignores an error during opening `/dev/kmsg`.

- Faster mount detection for linux kernel 5.10+ using openat2 speeding up pod churn rates. On Kernel versions less 5.10, it will fallback to using the original way of detecting mount points i.e by parsing /proc/mounts. ([#109217](https://github.com/kubernetes/kubernetes/pull/109217), [@manugupt1](https://github.com/manugupt1))...

The RuntimeClass alpha feature enables a node to surface multiple runtime options to support a variety of workload types.  Examples include native linux containers, and “sandboxed” containers that isolate the container from the host kernel.

- SCTP support in API objects (Pod, Service, NetworkPolicy) is now GA.
  Note that this has no effect on whether SCTP is enabled on nodes at the kernel level,
  and note that some cloud platforms and network plugins do not support SCTP traffic. ([#95566](https://github.com/kubernetes/kubernetes/pull/95566), [@danwinship](https://github.com/danwinship)) [SIG Apps and Network]

- kubeadm: implement support for concurrent add/remove of stacked etcd members ([#79677](https://github.com/kubernetes/kubernetes/pull/79677), [@neolit123](https://github.com/neolit123))
- kubeadm: support any Linux kernel version newer than 3.10 ([#81623](https://github.com/kubernetes/kubernetes/pull/81623), [@neolit123](https://github.com/neolit123))

- Changes behaviour of kube-proxy start; does not attempt to set specific `sysctl` values (which does not work in recent Kernel versions anymore in non-init namespaces), when the current sysctl values are already set higher. ([#103174](https://github.com/kubernetes/kubernetes/pull/103174), [@Napsty](https://github.com/Napsty))

- Clarified the CFS quota as 100ms in the code comments and set the minimum `cpuCFSQuotaPeriod` to 1ms to match Linux kernel expectations. ([#112123](https://github.com/kubernetes/kubernetes/pull/112123), [@paskal](https://github.com/paskal))

- Publishing kube-proxy metrics for Windows kernel-mode ([#106581](https://github.com/kubernetes/kubernetes/pull/106581), [@knabben](https://github.com/knabben))

- Add `/sys/devices/virtual/powercap` to default masked paths. It avoids the potential security risk that the ability to read these files may offer a power-based sidechannel attack against any workloads running on the same kernel. ([#125970](https://github.com/kubernetes/kubernetes/pull/125970), [@carlory](https://github.com/carlory)) [SIG Node]
- Fix a bug that Pods could stuck in the unschedulable pod pool