Add listen to BSDs

[tuxillo]

Copy over Darwin's code and reuse it for the most common BSDs. It is untested in all BSDs added but in DragonFly BSD but at least it'll allow the build to succeed.

Might want to merge darwin and BSD file together but for now they are separate because we might need to add something BSD-specific.

Build might fail in DragonFly BSD because it depends on insomniacslk/dhcp#510 , which is still not upstream. So a bump will also be needed for this specific OS.

This is for the ongoing issue lima-vm/lima#892

[openshift-ci]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: tuxillo
Once this PR has been reviewed and has the lgtm label, please assign baude for approval. For more information see the Kubernetes Code Review Process.

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

Needs approval from an approver in each of these files:

• OWNERS

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

[AkihiroSuda]

darwin.go can be merged here too?

[tuxillo]

I left it in different files just in case there was something BSD-y required that's not present in Darwin. I can't try all BSDs tho.

[cfergeau]

Seems like it would be easy enough to duplicate this file when BSD-specific support would be needed? I'd prefer to only have //go:build darwin || dragonfly || freebsd || netbsd || openbsd for now, instead of adding more duplicate listen code.
If you duplicate the code, I'd remove the case "vsock" block which is hyperkit specific, and ListenUnixgram is also not expected to be used as it's vfkit-specific, so this could error out.

[afbjorklund]

Seems like a good reason to have two, one for darwin and one for bsd

[cfergeau]

If the BSD files are different from the darwin files, yes, this is a good reason for having 2 files. As the PR is right now, the files are the same.

[cfergeau]

The commit log is missing a Signed-off-by (git commit --amend -s)

[cfergeau]

My take on this is https://github.com/cfergeau/gvisor-tap-vsock/tree/bsd - It's more involved than this PR as this refactors the existing code so that the platform files only contain platform specific code. Then this adds a generic Listen method which is used on platforms which are not darwin, not linux, not windows.
I've only tested that this builds, I haven't tried to run it yet.

[tuxillo]

My take on this is https://github.com/cfergeau/gvisor-tap-vsock/tree/bsd - It's more involved than this PR as this refactors the existing code so that the platform files only contain platform specific code. Then this adds a generic Listen method which is used on platforms which are not darwin, not linux, not windows. I've only tested that this builds, I haven't tried to run it yet.

Have some questions.

• If the vsock case is Hyperkit specific (MacOS), why is it still present in Linux code?
• Can the DefaultURL be defined per platforms so that all the unix stuff can be merged into a single file?

[cfergeau]

If the vsock case is Hyperkit specific (MacOS), why is it still present in Linux code?

The vsock implementation in listen_darwin.go is hyperkit-specific, the use of unix sockets, and the naming of the corresponding file (cid.port) is something only hyperkit uses.
On linux, vsock can also be used, but in a different, more generic way through mdlayher/vsock (which wraps AF_VSOCK sockets)

Can the DefaultURL be defined per platforms so that all the unix stuff can be merged into a single file?

I'm not sure which "unix stuff" you are referring to?

[tuxillo]

Can the DefaultURL be defined per platforms so that all the unix stuff can be merged into a single file?

I'm not sure which "unix stuff" you are referring to?

I was referring to all unix-like systems (linux, bsds, probably solaris) but not Darwin. But now that vsock is still needed in Linux for the vsock implementation you mentioned this makes less sense.

[tuxillo]

@cfergeau what are the next steps then?

[cfergeau]

@cfergeau what are the next steps then?

I filed #262 , though I have only tested it on linux.

I'm also fine with just adding the BSD build tags to the darwin code for now, and with a Signed-off-by added to the commit log.

[cfergeau]

With #262 merged, *bsd builds should be fixed, closing this PR.