Tcp timestamp negotiation

[danielinux]

Only send ACKs with TS option if negotiated.

Based on previous PR: #35

[Copilot]

Pull request overview

This PR updates wolfIP’s TCP behavior to only include the TCP Timestamp (TS) option in ACKs and data segments when TS has been negotiated, and expands TCP reliability logic around RTO/persist handling.

Changes:

• Gate TS option emission (ACK options, SYN/SYN-ACK TS inclusion, and data segment TS inclusion) on per-socket TS negotiation state (ts_enabled/ts_offer).
• Introduce RFC 6298-style RTO computation (SRTT/RTTVAR), Karn’s algorithm (skip RTT samples for retransmitted segments), and control-segment RTO handling for SYN/FIN states.
• Add TCP persist timer + zero-window probing and corresponding unit tests; update README to mention RTO computation support.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.

File	Description
src/wolfip.c	Implements TS negotiation gating; adds RTO/persist/control-RTO logic and new TCP socket fields/flags.
src/test/unit/unit.c	Adds/updates unit tests for control RTO, RFC6298 RTO updates, TS omission when not negotiated, and persist probing.
README.md	Documents retransmission timeout (RTO) computation support.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

tcp_send_zero_wnd_probe() chooses probe_seq = snd_una when snd_una falls inside a queued segment, but still sends payload[0] as the probe byte. If snd_una is not equal to seg_seq, the probe payload no longer matches the probe sequence number, so the receiver will see an inconsistent byte/sequence pairing. Compute the payload offset as snd_una - seg_seq (when snd_una is within the segment) and use that byte for the probe.

Suggested change

	probe_byte = payload[0];
	if (tcp_seq_leq(seg_seq, t->sock.tcp.snd_una) &&
	tcp_seq_lt(t->sock.tcp.snd_una, tcp_seq_inc(seg_seq, seg_len))) {
	probe_seq = t->sock.tcp.snd_una;
	break;
	}
	probe_seq = seg_seq;
	if (tcp_seq_leq(seg_seq, t->sock.tcp.snd_una) &&
	tcp_seq_lt(t->sock.tcp.snd_una, tcp_seq_inc(seg_seq, seg_len))) {
	uint32_t offset = t->sock.tcp.snd_una - seg_seq;
	if (offset < seg_len) {
	probe_byte = payload[offset];
	} else {
	/* Fallback: should not happen if the sequence checks above are correct. */
	probe_byte = payload[0];
	}
	probe_seq = t->sock.tcp.snd_una;
	break;
	}
	probe_seq = seg_seq;
	probe_byte = payload[0];

[Copilot]

The new FIN-ack handling in tcp_ack() (e.g., TCP_LAST_ACK/TCP_FIN_WAIT_1 transitions + tcp_ctrl_rto_stop()) won’t run for TCP_LAST_ACK in the normal receive path: tcp_input() currently does not call tcp_ack() for TCP_LAST_ACK state, and instead closes the socket elsewhere. This means the control RTO timer may not be stopped when the peer ACKs our FIN, and the socket may be closed without validating the ACK actually covers the FIN. Consider updating tcp_input() to route ACK processing for TCP_LAST_ACK through tcp_ack() (and only close when fin_acked is acknowledged), so the control-RTO state is cleaned up correctly.

[Copilot]

close_socket() now stops the persist timer but still leaves any active tmr_rto entry in the global timer heap. Because sockets are reused from a static array, a stale RTO timer can later fire on a different connection that reuses the same tsocket memory. Cancel ts->sock.tcp.tmr_rto (and set it to NO_TIMER) as part of close_socket() for TCP sockets (and similarly ensure control/data RTO timers are always canceled on close/reset paths).

Suggested change

	if (ts->proto == WI_IPPROTO_TCP)
	tcp_persist_stop(ts);
	if (ts->proto == WI_IPPROTO_TCP) {
	tcp_persist_stop(ts);
	if (ts->sock.tcp.tmr_rto != NO_TIMER) {
	wolfIP_timer_cancel(ts->sock.tcp.tmr_rto);
	ts->sock.tcp.tmr_rto = NO_TIMER;
	}
	}

[Copilot]

PR description focuses on TCP timestamp negotiation (only sending TS on ACKs when negotiated), but this PR also introduces substantial additional behavior (RFC6298 RTO computation, persist timer/zero-window probes, control-segment RTO, Karn RTT sampling, congestion-control tweaks). Please either update the PR title/description to reflect the broader scope or split the additional transport/RTO/persist changes into separate PRs to keep review and regression risk manageable.