GitHub

We currently don't have a lot of confidence in the simulation accuracy of TCP. We need to develop some tests, some of which might not initially pass. We expect that we'll ultimately end up wanting to rewrite the TCP simulation code (in Rust).

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Validate and improve TCP simulation accuracy

TCP sequence numbers should count bytes, not packets

Accumulating memory usage in tcp_new and payload_new

Incorrect packets retransmissions after process is killed and sockets have been closed?

Redesign packet and packet header classes for clarity

Incoming TCP connections should not be added to the descriptor table until they're accept()ed

Non-accepted sockets and their parent sockets are never disassociated/freed

Shadow doesn't send an RST response if server has no associated socket for incoming TCP packet

Eliminate host prioritization for packet events

Change or document the 100% reliability of control (0-length) packets

Closing a client socket shouldn't remove the connection from the server's accept queue

Token buckets can be applied to localhost packets

Localhost TCP sockets use delayed ACKs

Manually bound sockets should only receive packets from the peer

Some sockets are never disassociated from the network interface

Implement tcp_slow_start_after_idle

Replace the C `Packet` with a rust version

Support nanosecond precision in pcap files

Cannot read more bytes from TCP socket after `SHUT_RD`

TCP socket is incorrectly sending a FIN

Packet captures on loopback should only record packet once

TCP retransmissions cause socket to ignore FIN

FIN is never ACKed, leading to retransmitted FIN forever

closing a listener socket doesn't reset pending connections that haven't yet been accepted

Support `SO_REUSEADDR` (and lack of `SO_REUSEADDR`)

Validate and improve TCP simulation accuracy

List view