* Implement send buffer
* Window scaling
* Handle retransmission
+ - Proper timeout handling
TODO v2.0:
Does this need special care or can we rely on higher level MACs?
+RFCs
+----
+
+793 Transmission Control Protocol (Functional Specification)
+2581 TCP Congestion Control
+2988 Computing TCP's Retransmission Timer
+
+
+
INVARIANTS
----------
- SYN and FIN each count as one byte for the sequence numbering, but no actual byte is transferred in the payload.
+CONNECTION TIMEOUT
+------------------
+
+This timer is intended to catch the case when we are waiting very long for a response but nothing happens.
+The timeout is in the order of minutes.
+
+- The conn timeout is set whenever there is unacknowledged data, or when we are in the TIME_WAIT status.
+- If snd.una is advanced while the timeout is set, we re-set the timeout.
+- If the conn timeout expires, close the connection immediately.
+
+RETRANSMIT TIMEOUT
+------------------
+
+(See RFC 2988, 3366)
+
+This timer is intended to catch the case where we didn't get an ACK from the peer.
+In principle, the timeout should be slightly longer than the maximum latency along the path.
+
+
+- The rtrx timeout is set whenever snd.nxt is advanced.
+- If the rtrx timeout expires, retransmit at least one packet, and re-set the timeout.
+
STATES
------
-CLOSED: this connection is cloed, all packets received will result in RST.
+CLOSED: this connection is closed, all packets received will result in RST.
+ RX: RST
+ TX: return error
+ RT: clear timers
+ RST: ignore
LISTEN: (= no connection yet): only allow SYN packets, it application does not accept, return RST|ACK, else SYN|ACK.
+ RX: on accept, send SYNACK, go to SYN_RECEIVED
+ TX: cannot happen
+ RT: cannot happen
+ RST: ignore
SYN_SENT: we sent a SYN, now expecting SYN|ACK
+ RX: must be valid SYNACK, send ACK, go to ESTABLISHED
+ TX: put in send buffer (TODO: send SYN again with data?)
+ RT: send SYN again
SYN_RECEIVED: we received a SYN, sent back a SYN|ACK, now expecting an ACK
+ RX: must be valid ACK, go to ESTABLISHED
+ TX: put in send buffer (TODO: send SYNACK again with data?)
+ RT: send SYNACK again
ESTABLISHED: SYN is acked, we can now send/receive normal data.
+ RX: process data, return ACK. If FIN set, go to CLOSE_WAIT
+ TX: put in send buffer, segmentize and send
+ RT: send unACKed data again
FIN_WAIT_1: we want to close the connection, and just sent a FIN, waiting for it to be ACKed.
+ RX: process data, return ACK. If our FIN is acked, go to FIN_WAIT_2, if a FIN was also received, go to CLOSING
+ TX: return error
+ RT: send unACKed data or else FIN again
-FIN_WAIT_2: FIXME
+FIN_WAIT_2: our FIN is ACKed, just waiting for more data or FIN from the peer.
+ RX: process data, return ACK. If a FIN was also received, go to CLOSING
+ TX: return error
+ RT: should not happen, clear timeouts
CLOSE_WAIT: we received a FIN, we sent back an ACK
+ RX: only return an ACK.
+ TX: put in send buffer, segmentize and send
+ RT: send unACKed data again
CLOSING: we had already sent a FIN, and we received a FIN back, now waiting for it to be ACKed.
+ RX: if it's ACKed, set conn timeout, go to TIME_WAIT
+ TX: return an error
+ RT: send unACKed data or else FIN again
LAST_ACK: we are waiting for the last ACK before we can CLOSE
+ RX: if it's ACKed, go to CLOSED
+ TX: return an error
+ RT: send FIN again
TIME_WAIT: connection is in princple closed, but our last ACK might not have been received, so just wait a while to see if a FIN gets retransmitted so we can resend the ACK.
+ RX: if we receive anything, reset conn timeout.
+ TX: return an error
+ RT: should not happen, clear rtrx timeout
SEND PACKET
-----------
- Drop invalid packets:
- Invalid flags or state
+ - ACK always set
- hdr.seq not within our receive window
- hdr.ack ahead of snd.nxt
- Handle RST packets
uint32_t una;
uint32_t nxt;
uint32_t wnd;
- uint32_t up;
- uint32_t wl1;
- uint32_t wl2;
uint32_t iss;
} snd;
struct {
uint32_t nxt;
uint32_t wnd;
- uint32_t up;
uint32_t irs;
} rcv;
print_packet(utcp, "recv", data, len);
+ // Drop packets smaller than the header
+
struct hdr hdr;
if(len < sizeof hdr) {
errno = EBADMSG;
return -1;
}
+ // Make a copy from the potentially unaligned data to a struct hdr
+
memcpy(&hdr, data, sizeof hdr);
data += sizeof hdr;
len -= sizeof hdr;
+ // Drop packets with an unknown CTL flag
+
if(hdr.ctl & ~(SYN | ACK | RST | FIN)) {
errno = EBADMSG;
return -1;
}
- //list_connections(utcp);
+ // Try to match the packet to an existing connection
struct utcp_connection *c = find_connection(utcp, hdr.dst, hdr.src);
// Is it for a new connection?
if(!c) {
+ // Ignore RST packets
+
if(hdr.ctl & RST)
return 0;
- if(hdr.ctl & SYN && !(hdr.ctl & ACK) && utcp->accept && (!utcp->pre_accept || utcp->pre_accept(utcp, hdr.dst)) && (c = allocate_connection(utcp, hdr.dst, hdr.src))) { // LISTEN
- // Return SYN+ACK
+ // Is it a SYN packet and are we LISTENing?
+
+ if(hdr.ctl & SYN && !(hdr.ctl & ACK) && utcp->accept) {
+ // If we don't want to accept it, send a RST back
+ if((utcp->pre_accept && !utcp->pre_accept(utcp, hdr.dst))) {
+ len = 1;
+ goto reset;
+ }
+
+ // Try to allocate memory, otherwise send a RST back
+ c = allocate_connection(utcp, hdr.dst, hdr.src);
+ if(!c) {
+ len = 1;
+ goto reset;
+ }
+
+ // Return SYN+ACK, go to SYN_RECEIVED state
c->snd.wnd = hdr.wnd;
c->rcv.irs = hdr.seq;
c->rcv.nxt = c->rcv.irs + 1;
hdr.ctl = SYN | ACK;
print_packet(c->utcp, "send", &hdr, sizeof hdr);
utcp->send(utcp, &hdr, sizeof hdr);
- return 0;
- } else { // CLOSED
+ } else {
+ // No, we don't want your packets, send a RST back
len = 1;
goto reset;
}
+
+ return 0;
}
fprintf(stderr, "%p state %s\n", c->utcp, strstate[c->state]);
- if(c->state == CLOSED) {
- fprintf(stderr, "Error: packet recv()d on closed connection %p\n", c);
- errno = EBADF;
- return -1;
- }
+ // In case this is for a CLOSED connection, ignore the packet.
+ // TODO: make it so incoming packets can never match a CLOSED connection.
+
+ if(c->state == CLOSED)
+ return 0;
// It is for an existing connection.
- if(c->state == SYN_SENT) {
- if(hdr.ctl & ACK) {
- if(seqdiff(hdr.ack, c->snd.iss) <= 0 || seqdiff(hdr.ack, c->snd.nxt) > 0) {
- fprintf(stderr, "Invalid ACK, %u %u %u\n", hdr.ack, c->snd.iss, c->snd.nxt);
- goto reset;
- }
- }
- if(hdr.ctl & RST) {
- if(!(hdr.ctl & ACK))
- return 0;
- set_state(c, CLOSED);
- errno = ECONNREFUSED;
- if(c->recv)
- c->recv(c, NULL, 0);
- return 0;
- }
- if(hdr.ctl & SYN) {
- c->dst = hdr.src;
- c->rcv.nxt = hdr.seq + 1;
- c->rcv.irs = hdr.seq;
- c->snd.wnd = hdr.wnd;
+ // 1. Drop invalid packets.
- if(hdr.ctl & ACK)
- c->snd.una = hdr.ack;
- if(seqdiff(c->snd.una, c->snd.iss) > 0) {
- set_state(c, ESTABLISHED);
- // TODO: signal app?
- swap_ports(&hdr);
- hdr.seq = c->snd.nxt;
- hdr.ack = c->rcv.nxt;
- hdr.ctl = ACK;
- } else {
- set_state(c, SYN_RECEIVED);
- swap_ports(&hdr);
- hdr.seq = c->snd.iss;
- hdr.ack = c->rcv.nxt;
- hdr.ctl = SYN | ACK;
- }
- print_packet(c->utcp, "send", &hdr, sizeof hdr);
- utcp->send(utcp, &hdr, sizeof hdr);
- // TODO: queue any data?
- }
+ // 1a. Drop packets that should not happen in our current state.
- return 0;
+ switch(c->state) {
+ case SYN_SENT:
+ case SYN_RECEIVED:
+ case ESTABLISHED:
+ case FIN_WAIT_1:
+ case FIN_WAIT_2:
+ case CLOSE_WAIT:
+ case CLOSING:
+ case LAST_ACK:
+ case TIME_WAIT:
+ break;
+ default:
+ abort();
}
+ // 1b. Drop packets with a sequence number not in our receive window.
+
bool acceptable;
- if(len == 0)
+ if(c->state == SYN_SENT)
+ acceptable = true;
+
+ // TODO: handle packets overlapping c->rcv.nxt.
+#if 0
+ // Only use this when accepting out-of-order packets.
+ else if(len == 0)
if(c->rcv.wnd == 0)
acceptable = hdr.seq == c->rcv.nxt;
else
- acceptable = (hdr.seq >= c->rcv.nxt && hdr.seq < c->rcv.nxt + c->rcv.wnd);
+ acceptable = (seqdiff(hdr.seq, c->rcv.nxt) >= 0 && seqdiff(hdr.seq, c->rcv.nxt + c->rcv.wnd) < 0);
else
if(c->rcv.wnd == 0)
+ // We don't accept data when the receive window is zero.
acceptable = false;
else
- acceptable = (hdr.seq >= c->rcv.nxt && hdr.seq < c->rcv.nxt + c->rcv.wnd)
- || (hdr.seq + len - 1 >= c->rcv.nxt && hdr.seq + len - 1 < c->rcv.nxt + c->rcv.wnd);
+ // Both start and end of packet must be within the receive window
+ acceptable = (seqdiff(hdr.seq, c->rcv.nxt) >= 0 && seqdiff(hdr.seq, c->rcv.nxt + c->rcv.wnd) < 0)
+ || (seqdiff(hdr.seq + len + 1, c->rcv.nxt) >= 0 && seqdiff(hdr.seq + len - 1, c->rcv.nxt + c->rcv.wnd) < 0);
+#else
+ if(c->state != SYN_SENT)
+ acceptable = hdr.seq == c->rcv.nxt;
+#endif
if(!acceptable) {
- fprintf(stderr, "Packet not acceptable, %u %u %u %zu\n", hdr.seq, c->rcv.nxt, c->rcv.wnd, len);
+ fprintf(stderr, "Packet not acceptable, %u <= %u + %zu < %u\n", c->rcv.nxt, hdr.seq, len, c->rcv.nxt + c->rcv.wnd);
+ // Ignore unacceptable RST packets.
if(hdr.ctl & RST)
return 0;
- goto ack_and_drop;
+ // Otherwise, send an ACK back in the hope things improve.
+ goto ack;
}
- c->snd.wnd = hdr.wnd;
+ c->snd.wnd = hdr.wnd; // TODO: move below
- // TODO: check whether segment really starts at rcv.nxt, otherwise trim it.
+ // 1c. Drop packets with an invalid ACK.
+ // ackno should not roll back, and it should also not be bigger than snd.nxt.
+
+ if(hdr.ctl & ACK && (seqdiff(hdr.ack, c->snd.nxt) > 0 || seqdiff(hdr.ack, c->snd.una) < 0)) {
+ fprintf(stderr, "Packet ack seqno out of range, %u %u %u\n", hdr.ack, c->snd.una, c->snd.nxt);
+ // Ignore unacceptable RST packets.
+ if(hdr.ctl & RST)
+ return 0;
+ goto reset;
+ }
+
+ // 2. Handle RST packets
if(hdr.ctl & RST) {
switch(c->state) {
+ case SYN_SENT:
+ if(!(hdr.ctl & ACK))
+ return 0;
+ // The peer has refused our connection.
+ set_state(c, CLOSED);
+ errno = ECONNREFUSED;
+ if(c->recv)
+ c->recv(c, NULL, 0);
+ return 0;
case SYN_RECEIVED:
- // TODO: delete connection?
- break;
+ if(hdr.ctl & ACK)
+ return 0;
+ // We haven't told the application about this connection yet. Silently delete.
+ free_connection(c);
+ return 0;
case ESTABLISHED:
case FIN_WAIT_1:
case FIN_WAIT_2:
case CLOSE_WAIT:
+ if(hdr.ctl & ACK)
+ return 0;
+ // The peer has aborted our connection.
set_state(c, CLOSED);
errno = ECONNRESET;
if(c->recv)
c->recv(c, NULL, 0);
- break;
+ return 0;
case CLOSING:
case LAST_ACK:
case TIME_WAIT:
- // TODO: delete connection?
- break;
- default:
- // TODO: wtf?
+ if(hdr.ctl & ACK)
+ return 0;
+ // As far as the application is concerned, the connection has already been closed.
+ // If it has called utcp_close() already, we can immediately free this connection.
+ if(c->reapable) {
+ free_connection(c);
+ return 0;
+ }
+ // Otherwise, immediately move to the CLOSED state.
+ set_state(c, CLOSED);
return 0;
+ default:
+ abort();
}
- set_state(c, CLOSED);
- return 0;
}
+ // 3. Advance snd.una
+
+ uint32_t advanced = seqdiff(hdr.ack, c->snd.una);
+ c->snd.una = hdr.ack;
+
+ if(advanced) {
+ fprintf(stderr, "%p advanced %u\n", utcp, advanced);
+ // Make room in the send buffer.
+ // TODO: try to avoid memmoving too much. Circular buffer?
+ uint32_t left = seqdiff(c->snd.nxt, hdr.ack);
+ if(left)
+ memmove(c->sndbuf, c->sndbuf + advanced, left);
+ }
+
+ // 4. Update timers
+
+ if(advanced) {
+ timerclear(&c->conn_timeout); // It should be set anew in utcp_timeout() if c->snd.una != c->snd.nxt.
+ if(c->snd.una == c->snd.nxt)
+ timerclear(&c->rtrx_timeout);
+ }
+
+ // 5. Process SYN stuff
+
if(hdr.ctl & SYN) {
switch(c->state) {
+ case SYN_SENT:
+ // This is a SYNACK. It should always have ACKed the SYN.
+ if(!advanced)
+ goto reset;
+ c->rcv.irs = hdr.seq;
+ c->rcv.nxt = hdr.seq;
+ set_state(c, ESTABLISHED);
+ // TODO: notify application of this somehow.
+ break;
case SYN_RECEIVED:
case ESTABLISHED:
case FIN_WAIT_1:
case CLOSING:
case LAST_ACK:
case TIME_WAIT:
- set_state(c, CLOSED);
- errno = ECONNRESET;
- if(c->recv)
- c->recv(c, NULL, 0);
+ // Ehm, no. We should never receive a second SYN.
goto reset;
- break;
default:
- // TODO: wtf?
- return 0;
+ abort();
}
+
+ // SYN counts as one sequence number
+ c->rcv.nxt++;
}
- if(!(hdr.ctl & ACK))
- return 0;
+ // 6. Process new data
- switch(c->state) {
- case SYN_RECEIVED:
- if(seqdiff(hdr.ack, c->snd.una) >= 0 && seqdiff(hdr.ack, c->snd.nxt) <= 0)
- c->utcp->accept(c, hdr.dst);
-
- if(c->state != ESTABLISHED)
+ if(c->state == SYN_RECEIVED) {
+ // This is the ACK after the SYNACK. It should always have ACKed the SYNACK.
+ if(!advanced)
goto reset;
- c->snd.una = hdr.ack;
- break;
- case ESTABLISHED:
- case CLOSE_WAIT:
- if(seqdiff(hdr.ack, c->snd.una) < 0)
- return 0;
- if(seqdiff(hdr.ack, c->snd.nxt) > 0)
- goto ack_and_drop;
- if(seqdiff(hdr.ack, c->snd.una) > 0 && seqdiff(hdr.ack, c->snd.nxt) <= 0) {
- c->snd.una = hdr.ack;
- if(seqdiff(c->snd.wl1, hdr.seq) < 0 || (c->snd.wl1 == hdr.seq && seqdiff(c->snd.wl2, hdr.ack) <= 0)) {
- c->snd.wnd = hdr.wnd;
- c->snd.wl1 = hdr.seq;
- c->snd.wl2 = hdr.ack;
- }
- }
- break;
- case FIN_WAIT_1:
- if(hdr.ack == c->snd.nxt)
- set_state(c, FIN_WAIT_2);
- break;
- case FIN_WAIT_2:
- // TODO: If nothing left to send, close.
- break;
- case CLOSING:
- if(hdr.ack == c->snd.nxt) {
- set_state(c, TIME_WAIT);
- }
- break;
- case LAST_ACK:
- if(hdr.ack == c->snd.nxt) {
+ // Are we still LISTENing?
+ if(utcp->accept)
+ utcp->accept(c, c->src);
+
+ if(c->state != ESTABLISHED) {
set_state(c, CLOSED);
+ c->reapable = true;
+ goto reset;
}
- return 0;
- case TIME_WAIT:
- // TODO: retransmission of remote FIN, ACK and restart 2 MSL timeout
- break;
- default:
- goto reset;
}
- // Process data
+ if(len) {
+ switch(c->state) {
+ case SYN_SENT:
+ case SYN_RECEIVED:
+ // This should never happen.
+ abort();
+ case ESTABLISHED:
+ case FIN_WAIT_1:
+ case FIN_WAIT_2:
+ break;
+ case CLOSE_WAIT:
+ case CLOSING:
+ case LAST_ACK:
+ case TIME_WAIT:
+ // Ehm no, We should never receive more data after a FIN.
+ goto reset;
+ default:
+ abort();
+ }
- switch(c->state) {
- case ESTABLISHED:
- case FIN_WAIT_1:
- case FIN_WAIT_2:
- // TODO: process the data, see page 74
- break;
- case CLOSE_WAIT:
- case CLOSING:
- case LAST_ACK:
- case TIME_WAIT:
- break;
- default:
- abort();
+ int rxd;
+
+ if(c->recv) {
+ rxd = c->recv(c, data, len);
+ if(rxd < 0)
+ rxd = 0;
+ else if(rxd > len)
+ rxd = len; // Bad application, bad!
+ } else {
+ rxd = len;
+ }
+
+ c->rcv.nxt += len;
}
+ // 7. Process FIN stuff
+
if(hdr.ctl & FIN) {
switch(c->state) {
- case CLOSED:
- case LISTEN:
case SYN_SENT:
- return 0;
case SYN_RECEIVED:
+ // This should never happen.
+ abort();
case ESTABLISHED:
set_state(c, CLOSE_WAIT);
- c->rcv.nxt++;
- goto ack_and_drop;
+ break;
case FIN_WAIT_1:
set_state(c, CLOSING);
- c->rcv.nxt++;
- goto ack_and_drop;
+ break;
case FIN_WAIT_2:
set_state(c, TIME_WAIT);
- c->rcv.nxt++;
- goto ack_and_drop;
+ break;
case CLOSE_WAIT:
case CLOSING:
case LAST_ACK:
case TIME_WAIT:
- break;
+ // Ehm, no. We should never receive a second FIN.
+ goto reset;
default:
abort();
}
- }
- // Process the data
+ // FIN counts as one sequence number
+ c->rcv.nxt++;
- if(len && c->recv) {
- c->recv(c, data, len);
- c->rcv.nxt += len;
- goto ack_and_drop;
+ // Inform the application that the peer closed the connection.
+ if(c->recv) {
+ errno = 0;
+ c->recv(c, NULL, 0);
+ }
}
+ if(!len && !advanced)
+ return 0;
+
+ if(!len && !(hdr.ctl & SYN) && !(hdr.ctl & FIN))
+ return 0;
+
+ack:
+ hdr.src = c->src;
+ hdr.dst = c->dst;
+ hdr.seq = c->snd.nxt;
+ hdr.ack = c->rcv.nxt;
+ hdr.ctl = ACK;
+ print_packet(c->utcp, "send", &hdr, sizeof hdr);
+ utcp->send(utcp, &hdr, sizeof hdr);
return 0;
reset:
hdr.seq = 0;
hdr.ctl = RST | ACK;
}
- print_packet(c->utcp, "send", &hdr, sizeof hdr);
+ print_packet(utcp, "send", &hdr, sizeof hdr);
utcp->send(utcp, &hdr, sizeof hdr);
return 0;
-ack_and_drop:
- swap_ports(&hdr);
- hdr.seq = c->snd.nxt;
- hdr.ack = c->rcv.nxt;
- hdr.ctl = ACK;
- print_packet(c->utcp, "send", &hdr, sizeof hdr);
- utcp->send(utcp, &hdr, sizeof hdr);
- if(c->state == CLOSE_WAIT || c->state == TIME_WAIT) {
- errno = 0;
- if(c->recv)
- c->recv(c, NULL, 0);
- }
- return 0;
}
int utcp_shutdown(struct utcp_connection *c, int dir) {
+ fprintf(stderr, "%p shutdown %d\n", c->utcp, dir);
if(!c) {
errno = EFAULT;
return -1;
case FIN_WAIT_2:
return 0;
case CLOSE_WAIT:
- set_state(c, LAST_ACK);
+ set_state(c, CLOSING);
break;
case CLOSING:
projects / utcp / commitdiff