* GCE installs will now avoid IP masquerade for all RFC-1918 IP blocks, rather than just 10.0.0.0/8. This means that clusters can be created in 192.168.0.0./16 and 172.16.0.0/12 while preserving the container IPs (which would be lost before). ([#46473](https://github.com/kubernetes/kubernetes/pull/46473), [@thockin](https://github.com/thockin))

- Kubelet should now more reliably report the same primary node IP even if the set of node IPs reported by the CloudProvider changes. ([#79391](https://github.com/kubernetes/kubernetes/pull/79391), [@danwinship](https://github.com/danwinship))

* kubenet: SyncHostports for both running and ready to run pods. ([#31388](https://github.com/kubernetes/kubernetes/pull/31388), [@yifan-gu](https://github.com/yifan-gu))
* GC pod ips ([#35572](https://github.com/kubernetes/kubernetes/pull/35572), [@bprashanth](https://github.com/bprashanth))

- Added the latest GCE pinhole firewall feature, which introduces `destination-ranges` in the ingress `firewall-rules`. It restricts access to the backend IPs by allowing traffic through `ILB` or `NetLB` only. This change does **NOT** change the existing `ILB` or `NetLB` behavior. ([#109510](https://github.com/kubernetes/kubernetes/pull/109510), [@sugangli](https://github.com/sugangli))

* Fixed bug in AWS provider to handle multiple IPs when using more than 1 network interface per ec2 instance. ([#50112](https://github.com/kubernetes/kubernetes/pull/50112), [@jlz27](https://github.com/jlz27))

        for ( ;; ) {
            try {
                doConnect();
                return;
            } catch(SmbAuthException sae) {
                throw sae; // Prevents account lockout on servers with multiple IPs
            } catch(SmbException se) {
                if (getNextAddress() == null) 
                    throw se;
                if (log.level >= 3)
                    se.printStackTrace(log);
            }

- Added an e2e test to verify that the cluster is not vulnerable to CVE-2021-29923 when using Services with IPs with leading zeros, note that this test is a necessary but not sufficient condition, all the components in the clusters that consume IPs addresses from the APIs MUST interpret them as decimal or discard them. ([#107552](https://github.com/kubernetes/kubernetes/pull/107552), [@aojea](https://github.com/aojea))

* Fixed bug in AWS provider to handle multiple IPs when using more than 1 network interface per ec2 instance. ([#50112](https://github.com/kubernetes/kubernetes/pull/50112), [@jlz27](https://github.com/jlz27))

   ([#123599](https://github.com/kubernetes/kubernetes/pull/123599), [@bobbypage](https://github.com/bobbypage))
- Changed `--nodeport-addresses` behavior to default to "primary node IP(s) only" rather than "all node IPs".
   ([#122724](https://github.com/kubernetes/kubernetes/pull/122724), [@nayihz](https://github.com/nayihz))

- The `apiserver` exposes 4 new metrics that allow to track the status of the Service CIDRs allocations:
      - current number of available IPs per Service CIDR
      - current number of used IPs per Service CIDR
      - total number of allocation per Service CIDR