Kevin Krakauer

Agreed that those are window probes sent with exponential backoff. And thanks for looking into the conntrack/firewall stuff. - TCP checksums in the gVisor -> Linux direction are all zero....

Thanks for all the help Bhasker. A few thoughts: > ACKing a packet not seen seems to indicate Out of Order delivery of packets My understanding is that ACKing unseen...

Oh, here's a thought: our zero window probes are broken? IIUC they should be sending the next byte in the probe. From [RFC 9293 3.8.6.1](https://datatracker.ietf.org/doc/html/rfc9293#name-zero-window-probing): > The sending TCP peer...

#10252 has the change, although I'm still wrestling with tests.

@djs55 could you try #10252? @hbhasker looks like [Linux doesn't follow the RFC either](https://github.com/zephyrproject-rtos/zephyr/issues/47966): it basically sends a keepalive, which has the same effect. I.e. it sends a byte with...

Glad that worked! We're going to go with #10258 (don't need to segment split, more Linux-like, fewer test changes). Feel free to try it or wait for it to get...

Took a look and, while I don't have anything definitive, I wonder whether it could be related to the use of goroutines that starts in `firestack/intra/netstack/udp.go:udpForwarder`. For example, the function...

Looking now, I may have been wrong in #8458. It should probably be up to the caller of `NewForwarder` to `IncRef` the packet. I think I looked at the TCP...

I believe tcpdump was working before -- I'll take a look. I can prioritize outbound packets, but as @hbhasker mentioned it would be helpful to know more in order to...

There's sendmmsg/recvmmsg and PACKET_MMAP. io_uring would be interesting -- we haven't explored that and I'm not sure how it would perform.