Change to correct break statements to prevent infinite recursion causing OOM

[misuto]

Problem description and Solution

I discovered that an image pull could end up in infinite recursion if the context is cancelled while pulling. The reason is that the break statement only breaks the select instead of the for.
I found a few other places in pkg/bindings that also seemed to have the same issue, which I included a fix for. The solution is to put a label at the for loop and reference that when breaking in the cases for pull.go and push.go. In manifests.go there were no return statement outside of the for loop so I opted to return an empty string along with the error context.Canceled.

Does this PR introduce a user-facing change?

Prevent OOM when context is cancelled in a few situations 

[NO NEW TESTS NEEDED]

[misuto]

There does not seem to be any fitting test I can modify or extend. Am I missing something or is it not implemented yet?
https://github.com/containers/podman/blob/main/test/system/TODO.md seems to indicate that tests pull is not implemented yet, but I don't know if the bindings package are included in this.

Should I skip writing tests?

[edsantiago]

Thanks for your PR and for asking about tests. https://github.com/containers/podman/blob/24b586e7d6b7b4940aef8a5929283519020d2806/test/system/330-corrupt-images.bats might be a candidate for a test; but from my quick reading, is there even a reliable reproducer for this?

[misuto]

Thanks for your PR and for asking about tests. https://github.com/containers/podman/blob/24b586e7d6b7b4940aef8a5929283519020d2806/test/system/330-corrupt-images.bats might be a candidate for a test; but from my quick reading, is there even a reliable reproducer for this?

Thanks for your reply!
When I discovered this I did it purely by accident in the beginning and it took a few tries to figure out what the problem was.
The very hacky reproduction steps are something like this:

1. Create new context with cancel
2. Create podman connection with NewConnection, with the ctx created in previous step
3. Start image pull in go routine
4. Wait one second and call cancel in the context created in step 1

That seemed to reproduce the issue alright, but it's not a pretty solution. The faults with selects in push.go and manifest.go was found with the staticcheck tool so I guess that is one way to test that those types of issues won't make it back into the code base again.

[rhatdan]

@mheon PTAL

[mheon]

Code LGTM over here.

/approve

[openshift-ci]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: mheon, misuto

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 [mheon]

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

[rhatdan]

@misuto Could you rebase and repush, so the tests will pass.

[rhatdan]

Awesome work @misuto Thanks.
/lgtm

[misuto]

Awesome work @misuto Thanks. /lgtm

Thank you, hope I can contribute with something more in the future!