Kevin Krakauer

IIUC that's specific to V3. I am thinking that PACKET_MMAP improvements would be for the existing V2 interface we use.

I've updated the title and description to make it clear that this issue is not related to use of V3, but rather is about improving the dispatcher using V1 or...

There's no specific support for Infiniband. Can you help me understand what support would be needed? gVisor containers typically communicate through a virtual device (often veth). On a machine with...

Having looked (maybe too) quickly at verbs, it should be possible to support if my understanding is correct. Thoughts: - Infiniband verbs are probably a bunch of `ioctl`s for their...

Thanks for the report. Let me see whether we can repro this. Also: I don't think I knew that gVisor's netstack was used by Docker on Mac. That's awesome.

I can repro this on Mac. On Linux it runs for the full hour without issue. I'll look at whether there's anything obviously platform-dependent, but it would be really helpful...

A couple first observations about the pcap: - There are a bunch of `[TCP ACKed unseen segment]` that I can't explain given that they occur when the connection looks otherwise...

I misunderstood the setup. gVisor _is_ sending packets that are originally from the client, and the server is running in a Linux VM. So it's not weird that messing with...

And another question: is the pcap between the client and proxy, or the proxy and iperf server?

Can you try a build with [this patch](https://github.com/google/gvisor/pull/10245/files)? My current theory is that the proxy is blocked waiting to write out some data, but the send buffer shrinks and we...