Fix too restrictive network interface name checking

[thediveo]

Fix too restrictive network interface name checking

parsePodNetworkObjectName() had too restrictive network interface name checking, trying to enforce RFC 1123 DNS label syntax. However, https://github.com/K8sNetworkPlumbingWG/multi-net-spec/blob/master/%5Bv1%5D%20Kubernetes%20Network%20Custom%20Resource%20Definition%20De-facto%20Standard.md states:

4.1.2.1.5 "interface"
This optional key with value of type string requires the implementation to use the given name for the pod interface resulting from this network attachment. This keys value must be a valid Linux kernel interface name

Fixes #213 and adds tests for correct parsing behavior.

[coveralls]

Pull Request Test Coverage Report for Build 685

• 22 of 22 (100.0%) changed or added relevant lines in 1 file are covered.
• 171 unchanged lines in 3 files lost coverage.
• Overall coverage increased (+1.9%) to 54.72%

---

Files with Coverage Reduction	New Missed Lines	%
k8sclient/k8sclient.go	34	62.5%
types/conf.go	36	32.29%
multus/multus.go	101	56.75%

Totals
Change from base Build 569:	1.9%
Covered Lines:	626
Relevant Lines:	1144

---

💛 - Coveralls

[dougbtv]

Thanks for the PR @thediveo -- just a quick note to say that there will probably be delay before the maintainers get together to review this PR / other PRs during the upcoming holidays.

[dcbw]

The code comment about allowed characters is not actually correct. The kernel validates the ifname like so:

bool dev_valid_name(const char *name)
{
	if (*name == '\0')
		return false;
	if (strnlen(name, IFNAMSIZ) == IFNAMSIZ)
		return false;
	if (!strcmp(name, ".") || !strcmp(name, ".."))
		return false;

	while (*name) {
		if (*name == '/' || *name == ':' || isspace(*name))
			return false;
		name++;
	}
	return true;
}

[dougbtv]

Also some examples in practice: https://paste.fedoraproject.org/paste/Cm2556XZYIO9IPbZaIWDoQ

[thediveo]

In the light of dev_valid_name I will reconsider my code to adhere to the nif name restrictions implemented in the Linux kernel.

[rkamudhan]

@thediveo As a follow up, you planning to take the @dcbw and @dougbtv suggestion in PR ? if so, we can move this PR as WIP

[thediveo]

Yes, I do, so please move to WIP, as I want to bring my PR in sync with the better knowledge about what valid interfaces names in Linux are.

[thediveo]

@rkamudhan I've updated my PR to now take dev_valid_name rules into account; I've reverted to the original separation sequence to first split on "/" and then "@", with the latter separation process now accepting multiple "@"s and separating only at the first "@". I've also updated tests too.

For the tests I noticed that it seems to be acceptable to have an empty namespace and network name, and only a network interface name, such as in "@eth42": is this correct?

If it is, then I would suspect my PR to be ready.

[thediveo]

@rkamudhan @dcbw @dougbtv any interest in merging this after I've implemented the suggestions raised before?

[s1061123]

let me close it because RFC1123 is not same as linux kernel interface name convention.