Completing CEL validation for Gateway

[robscott]

What type of PR is this?
/kind cleanup

What this PR does / why we need it:
This is a follow up to @gauravkghildiyal's PR that added CEL validation (#2226). It contains a few things:

• IPAddress validation (using https://regex101.com/r/Kzijkp/1 our custom CRD generation + oneOf)
• Validation for GatewayStatusAddress
• A few typo fixes

Does this PR introduce a user-facing change?:

NONE

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: robscott

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [robscott]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[robscott]

With all the escaping this gets to be pretty hard to parse - here's the original: https://regex101.com/r/Kzijkp/1

[shaneutt]

The escaping is making me wonder if we should hang out with the kubebuilder group a bit and see about options to get these rules from other files and locations.

[howardjohn]

@robscott thanks for the link, without the escaping it is super easy to understand!

[howardjohn]

Less sarcastic: can we use https://github.com/kubernetes/kubernetes/pull/85381/files? Seems we already use it in some other places (date-time - interestingly, the link I sent says it should be datetime.

Actually that may be tricky since we need this conditional... maybe there is some way to get these into the CEL... I can't help but think there must be a better way to do this (or if not, there should be)

[robscott]

Checked in with @jpbetz and sadly we're stuck with a regex for now. The good news is that this will get better with KEP 3937. Unfortunately in the best case this will be enabled by default in 1.31+, and with our 5 minor version support window, we'd need to wait for 1.36, or >2 years from now for this to be available.

The closest thing we have available now is isURL("https://" + ip + "/") which uses https://github.com/kubernetes/kubernetes/blob/97c7dbcd227ec253bb277a91575706fe30f519dc/staging/src/k8s.io/apiserver/pkg/cel/library/urls.go#L234, but of course that would still be too permissive for what we need here.

[Jefftree]

Oops sorry I didn't see I was tagged earlier. Yeah i think Joe's suggestion would be fine, though having

         type:
           not:
             enum: ["IPAddress"]

makes it look like any other value can be accepted if it doesn't fit the enum? If the fallback is string, why not keep that in the oneOf?

  oneOf:
     - properties:
         type:
           enum: ["IPAddress"]
         value:
           anyOf:
             - format: ipv4
             - format: ipv6
     - properties:
          type:
            type: string
          value:
            type: string

[jpbetz]

Eliminating the "not: enum: ..." part SGTM since the fallback is that you can have any value. Thanks @Jefftree

[robscott]

@Jefftree @jpbetz I think I'm misunderstanding the suggestion. It seems like that would be interpreted as

One Of:
A) type == IPAddress && (format == ipv4 || format == ipv6)
OR
B) type == any string && value == any string

Wouldn't that allow any value to pass through validation via B), including type==IPAddress?

What we really want is:

A) type == IPAddress && (format == ipv4 || format == ipv6)
OR
B) type == anything other than IPAddress

I think that's closer to what we actually have.

We already have the following x-kubernetes-validations value at the same level to cover the hostname type:

                  x-kubernetes-validations:
                  - message: Hostname value must only contain valid characters (matching
                      ^(\*\.)?[a-z0-9]([-a-z0-9]*[a-z0-9])?(\.[a-z0-9]([-a-z0-9]*[a-z0-9])?)*$)
                    rule: 'self.type == ''Hostname'' ? self.value.matches(''^(\\*\\.)?[a-z0-9]([-a-z0-9]*[a-z0-9])?(\\.[a-z0-9]([-a-z0-9]*[a-z0-9])?)*$''):
                      true'

Admittedly it would probably make sense to combine that all into the oneOf for a CRD that was a bit easier to understand. We're avoiding that because the Hostname validation can be expressed with a relatively simple regex that fits in a +kubebuilder annotation and doesn't require customizing the CRD generation like we're doing for IP Addresses here.

[Jefftree]

Ah right that will allow even invalid IPAddress types to pass validation via B. In that case would Joe's first suggestion should suffice then? It's already quite concise for what we want to express.

[robscott]

Yep, Joe's first suggestion seems to work well here, thanks for the help!

[gauravkghildiyal]

Looks good to me! Thanks a lot for the corrections and additions!

[youngnick]

I don't think the generator hack is that bad. 😁

/lgtm