gc: collect entries where the refcounter could underflow and wrap

[tixxdz]

What happened?

Inside gc we discard process entries with a refcounter that could underflow and wrap, such entries won't be on the deletequeue again and will stay forever.

I have identified what could be the real cause but need to assert that our current tests and logic could handle this or write more tests for it, and also document the refcounting logic.

Tetragon Version

all

Kernel Version

doesn't matter

Kubernetes Version

No response

Bugtool

No response

Relevant log output

No response

Anything else?

No response

[Jay-JHe]

@tixxdz Hi, may I ask if there has been any progress regarding this issue?

[tixxdz]

@tixxdz Hi, may I ask if there has been any progress regarding this issue?

Hi @Jay-JHe for this to happen another bug must trigger it, as we have fixed that one #1485, I think we are good for now. Nevertheless we have to make gc robust regardless of that, so I'm planning this fix shortly after we tag 1.0 , as changing the gc logic now can introduce new bugs ;-) . let's see what @jrfastab thinks too.