You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
With this in mind, it looks like we need to change how we're doing protocol parsing. For newer kernel (>4.6) we need to upload a BPF filter to do most of the work for us. If you're on older kernel we need to provide a fallback.
This may be done in two ways:
Make (use) a BPF interpreter in the userspace
Create a higher level abstraction/DSL, which either compiles to BPF and sent to kernel or interpreted directly
Either way, it looks like we want to take advantage of the KCM by default, instead of having two implementations for each protocol. We are already abstracting simple things here, like "wait for delimiter" and "wait for N bytes". The kernel facility is much more rich, and we should get advantage of the feature.
I'm not sure how well does it play with TLS. Presumably, we will be able to dissect TCP protocol packets by KCM and everything else only in userspace.
Update: It's likely a future of what rotor-stream should be, but probably we need to start another crate for this. Because it looks like semantics will be very different from what we have now.
The text was updated successfully, but these errors were encountered:
KCM is a kernel facility that provides a message based interface over TCP. Links:
With this in mind, it looks like we need to change how we're doing protocol parsing. For newer kernel (>4.6) we need to upload a BPF filter to do most of the work for us. If you're on older kernel we need to provide a fallback.
This may be done in two ways:
Either way, it looks like we want to take advantage of the KCM by default, instead of having two implementations for each protocol. We are already abstracting simple things here, like "wait for delimiter" and "wait for N bytes". The kernel facility is much more rich, and we should get advantage of the feature.
I'm not sure how well does it play with TLS. Presumably, we will be able to dissect TCP protocol packets by KCM and everything else only in userspace.
Update: It's likely a future of what rotor-stream should be, but probably we need to start another crate for this. Because it looks like semantics will be very different from what we have now.
The text was updated successfully, but these errors were encountered: