/**************************************************************************** * net/tcp/tcp_input.c * Handling incoming TCP input * * Copyright (C) 2007-2014 Gregory Nutt. All rights reserved. * Author: Gregory Nutt * * Adapted for NuttX from logic in uIP which also has a BSD-like license: * * Original author Adam Dunkels * Copyright () 2001-2003, Adam Dunkels. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the author may not be used to endorse or promote * products derived from this software without specific prior * written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. * ****************************************************************************/ /**************************************************************************** * Included Files ****************************************************************************/ #include #if defined(CONFIG_NET) && defined(CONFIG_NET_TCP) #include #include #include #include #include #include #include #include #include "devif/devif.h" #include "utils/utils.h" #include "tcp/tcp.h" /**************************************************************************** * Private Functions ****************************************************************************/ /**************************************************************************** * Name: tcp_input * * Description: * Handle incoming TCP input * * Parameters: * dev - The device driver structure containing the received TCP packet. * tcp - A pointer to the TCP header in the packet * iplen - Combined length of the IP and TCP headers * * Return: * None * * Assumptions: * Called from the interrupt level or with interrupts disabled. * ****************************************************************************/ static void tcp_input(FAR struct net_driver_s *dev, unsigned int iplen) { FAR struct tcp_hdr_s *tcp; FAR struct tcp_conn_s *conn = NULL; unsigned int tcpiplen; unsigned int hdrlen; uint16_t tmp16; uint16_t flags; uint16_t result; uint8_t opt; int len; int i; #ifdef CONFIG_NET_STATISTICS /* Bump up the count of TCP packets received */ g_netstats.tcp.recv++; #endif /* Get a pointer to the TCP header. The TCP header lies just after the * the link layer header and the IP header. */ tcp = (FAR struct tcp_hdr_s *)&dev->d_buf[iplen + NET_LL_HDRLEN(dev)]; /* Get the size of the IP header and the TCP header */ tcpiplen = iplen + TCP_HDRLEN; /* Get the size of the link layer header, the IP header, and the TCP header */ hdrlen = tcpiplen + NET_LL_HDRLEN(dev); /* Start of TCP input header processing code. */ if (tcp_chksum(dev) != 0xffff) { /* Compute and check the TCP checksum. */ #ifdef CONFIG_NET_STATISTICS g_netstats.tcp.drop++; g_netstats.tcp.chkerr++; #endif nwarn("WARNING: Bad TCP checksum\n"); goto drop; } /* Demultiplex this segment. First check any active connections. */ conn = tcp_active(dev, tcp); if (conn) { /* We found an active connection.. Check for the subsequent SYN * arriving in TCP_SYN_RCVD state after the SYNACK packet was * lost. To avoid other issues, reset any active connection * where a SYN arrives in a state != TCP_SYN_RCVD. */ if ((conn->tcpstateflags & TCP_STATE_MASK) != TCP_SYN_RCVD && (tcp->flags & TCP_CTL) == TCP_SYN) { goto reset; } else { goto found; } } /* If we didn't find an active connection that expected the packet, * either (1) this packet is an old duplicate, or (2) this is a SYN packet * destined for a connection in LISTEN. If the SYN flag isn't set, * it is an old packet and we send a RST. */ if ((tcp->flags & TCP_CTL) == TCP_SYN) { /* This is a SYN packet for a connection. Find the connection * listening on this port. */ tmp16 = tcp->destport; if (tcp_islistener(tmp16)) { /* We matched the incoming packet with a connection in LISTEN. * We now need to create a new connection and send a SYNACK in * response. */ /* First allocate a new connection structure and see if there is any * user application to accept it. */ conn = tcp_alloc_accept(dev, tcp); if (conn) { /* The connection structure was successfully allocated. Now see if * there is an application waiting to accept the connection (or at * least queue it it for acceptance). */ conn->crefs = 1; if (tcp_accept_connection(dev, conn, tmp16) != OK) { /* No, then we have to give the connection back and drop the packet */ conn->crefs = 0; tcp_free(conn); conn = NULL; } } if (!conn) { /* Either (1) all available connections are in use, or (2) there is no * application in place to accept the connection. We drop packet and hope that * the remote end will retransmit the packet at a time when we * have more spare connections or someone waiting to accept the connection. */ #ifdef CONFIG_NET_STATISTICS g_netstats.tcp.syndrop++; #endif nerr("ERROR: No free TCP connections\n"); goto drop; } net_incr32(conn->rcvseq, 1); /* Parse the TCP MSS option, if present. */ if ((tcp->tcpoffset & 0xf0) > 0x50) { for (i = 0; i < ((tcp->tcpoffset >> 4) - 5) << 2 ; ) { opt = dev->d_buf[hdrlen + i]; if (opt == TCP_OPT_END) { /* End of options. */ break; } else if (opt == TCP_OPT_NOOP) { /* NOP option. */ ++i; } else if (opt == TCP_OPT_MSS && dev->d_buf[hdrlen + 1 + i] == TCP_OPT_MSS_LEN) { uint16_t tcp_mss = TCP_MSS(dev, iplen); /* An MSS option with the right option length. */ tmp16 = ((uint16_t)dev->d_buf[hdrlen + 2 + i] << 8) | (uint16_t)dev->d_buf[hdrlen + 3 + i]; conn->mss = tmp16 > tcp_mss ? tcp_mss : tmp16; /* And we are done processing options. */ break; } else { /* All other options have a length field, so that we easily * can skip past them. */ if (dev->d_buf[hdrlen + 1 + i] == 0) { /* If the length field is zero, the options are malformed * and we don't process them further. */ break; } i += dev->d_buf[hdrlen + 1 + i]; } } } /* Our response will be a SYNACK. */ tcp_ack(dev, conn, TCP_ACK | TCP_SYN); return; } } /* This is (1) an old duplicate packet or (2) a SYN packet but with * no matching listener found. Send RST packet in either case. */ reset: /* We do not send resets in response to resets. */ if ((tcp->flags & TCP_RST) != 0) { goto drop; } #ifdef CONFIG_NET_STATISTICS g_netstats.tcp.synrst++; #endif tcp_reset(dev); return; found: /* Update the connection's window size */ conn->winsize = ((uint16_t)tcp->wnd[0] << 8) + (uint16_t)tcp->wnd[1]; flags = 0; /* We do a very naive form of TCP reset processing; we just accept * any RST and kill our connection. We should in fact check if the * sequence number of this reset is within our advertised window * before we accept the reset. */ if ((tcp->flags & TCP_RST) != 0) { conn->tcpstateflags = TCP_CLOSED; nwarn("WARNING: RESET - TCP state: TCP_CLOSED\n"); (void)tcp_callback(dev, conn, TCP_ABORT); goto drop; } /* Calculated the length of the data, if the application has sent * any data to us. */ len = (tcp->tcpoffset >> 4) << 2; /* d_len will contain the length of the actual TCP data. This is * calculated by subtracting the length of the TCP header (in * len) and the length of the IP header. */ dev->d_len -= (len + iplen); /* First, check if the sequence number of the incoming packet is * what we're expecting next. If not, we send out an ACK with the * correct numbers in, unless we are in the SYN_RCVD state and * receive a SYN, in which case we should retransmit our SYNACK * (which is done further down). */ if (!((((conn->tcpstateflags & TCP_STATE_MASK) == TCP_SYN_SENT) && ((tcp->flags & TCP_CTL) == (TCP_SYN | TCP_ACK))) || (((conn->tcpstateflags & TCP_STATE_MASK) == TCP_SYN_RCVD) && ((tcp->flags & TCP_CTL) == TCP_SYN)))) { if ((dev->d_len > 0 || ((tcp->flags & (TCP_SYN | TCP_FIN)) != 0)) && memcmp(tcp->seqno, conn->rcvseq, 4) != 0) { tcp_send(dev, conn, TCP_ACK, tcpiplen); return; } } /* Next, check if the incoming segment acknowledges any outstanding * data. If so, we update the sequence number, reset the length of * the outstanding data, calculate RTT estimations, and reset the * retransmission timer. */ if ((tcp->flags & TCP_ACK) != 0 && conn->unacked > 0) { uint32_t unackseq; uint32_t ackseq; /* The next sequence number is equal to the current sequence * number (sndseq) plus the size of the outstanding, unacknowledged * data (unacked). */ #ifdef CONFIG_NET_TCP_WRITE_BUFFERS unackseq = conn->sndseq_max; #else unackseq = tcp_addsequence(conn->sndseq, conn->unacked); #endif /* Get the sequence number of that has just been acknowledged by this * incoming packet. */ ackseq = tcp_getsequence(tcp->ackno); /* Check how many of the outstanding bytes have been acknowledged. For * most send operations, this should always be true. However, * the send() API sends data ahead when it can without waiting for * the ACK. In this case, the 'ackseq' could be less than then the * new sequence number. */ if (ackseq <= unackseq) { /* Calculate the new number of outstanding, unacknowledged bytes */ conn->unacked = unackseq - ackseq; } else { /* What would it mean if ackseq > unackseq? The peer has ACKed * more bytes than we think we have sent? Someone has lost it. * Complain and reset the number of outstanding, unacknowledged * bytes */ if ((conn->tcpstateflags & TCP_STATE_MASK) == TCP_ESTABLISHED) { nwarn("WARNING: conn->sndseq %d, conn->unacked %d\n", tcp_getsequence(conn->sndseq), conn->unacked); goto reset; } } /* Update sequence number to the unacknowledge sequence number. If * there is still outstanding, unacknowledged data, then this will * be beyond ackseq. */ ninfo("sndseq: %08x->%08x unackseq: %08x new unacked: %d\n", conn->sndseq, ackseq, unackseq, conn->unacked); tcp_setsequence(conn->sndseq, ackseq); /* Do RTT estimation, unless we have done retransmissions. */ if (conn->nrtx == 0) { signed char m; m = conn->rto - conn->timer; /* This is taken directly from VJs original code in his paper */ m = m - (conn->sa >> 3); conn->sa += m; if (m < 0) { m = -m; } m = m - (conn->sv >> 2); conn->sv += m; conn->rto = (conn->sa >> 3) + conn->sv; } /* Set the acknowledged flag. */ flags |= TCP_ACKDATA; /* Reset the retransmission timer. */ conn->timer = conn->rto; } /* Do different things depending on in what state the connection is. */ switch (conn->tcpstateflags & TCP_STATE_MASK) { /* CLOSED and LISTEN are not handled here. CLOSE_WAIT is not * implemented, since we force the application to close when the * peer sends a FIN (hence the application goes directly from * ESTABLISHED to LAST_ACK). */ case TCP_SYN_RCVD: /* In SYN_RCVD we have sent out a SYNACK in response to a SYN, and * we are waiting for an ACK that acknowledges the data we sent * out the last time. Therefore, we want to have the TCP_ACKDATA * flag set. If so, we enter the ESTABLISHED state. */ if ((flags & TCP_ACKDATA) != 0) { conn->tcpstateflags = TCP_ESTABLISHED; #ifdef CONFIG_NET_TCP_WRITE_BUFFERS conn->isn = tcp_getsequence(tcp->ackno); tcp_setsequence(conn->sndseq, conn->isn); conn->sent = 0; conn->sndseq_max = 0; #endif conn->unacked = 0; flags = TCP_CONNECTED; ninfo("TCP state: TCP_ESTABLISHED\n"); if (dev->d_len > 0) { flags |= TCP_NEWDATA; net_incr32(conn->rcvseq, dev->d_len); } dev->d_sndlen = 0; result = tcp_callback(dev, conn, flags); tcp_appsend(dev, conn, result); return; } /* We need to retransmit the SYNACK */ if ((tcp->flags & TCP_CTL) == TCP_SYN) { tcp_ack(dev, conn, TCP_ACK | TCP_SYN); return; } goto drop; case TCP_SYN_SENT: /* In SYN_SENT, we wait for a SYNACK that is sent in response to * our SYN. The rcvseq is set to sequence number in the SYNACK * plus one, and we send an ACK. We move into the ESTABLISHED * state. */ if ((flags & TCP_ACKDATA) != 0 && (tcp->flags & TCP_CTL) == (TCP_SYN | TCP_ACK)) { /* Parse the TCP MSS option, if present. */ if ((tcp->tcpoffset & 0xf0) > 0x50) { for (i = 0; i < ((tcp->tcpoffset >> 4) - 5) << 2 ; ) { opt = dev->d_buf[hdrlen + i]; if (opt == TCP_OPT_END) { /* End of options. */ break; } else if (opt == TCP_OPT_NOOP) { /* NOP option. */ ++i; } else if (opt == TCP_OPT_MSS && dev->d_buf[hdrlen + 1 + i] == TCP_OPT_MSS_LEN) { uint16_t tcp_mss = TCP_MSS(dev, iplen); /* An MSS option with the right option length. */ tmp16 = (dev->d_buf[hdrlen + 2 + i] << 8) | dev->d_buf[hdrlen + 3 + i]; conn->mss = tmp16 > tcp_mss ? tcp_mss : tmp16; /* And we are done processing options. */ break; } else { /* All other options have a length field, so that we * easily can skip past them. */ if (dev->d_buf[hdrlen + 1 + i] == 0) { /* If the length field is zero, the options are * malformed and we don't process them further. */ break; } i += dev->d_buf[hdrlen + 1 + i]; } } } conn->tcpstateflags = TCP_ESTABLISHED; memcpy(conn->rcvseq, tcp->seqno, 4); net_incr32(conn->rcvseq, 1); conn->unacked = 0; #ifdef CONFIG_NET_TCP_WRITE_BUFFERS conn->isn = tcp_getsequence(tcp->ackno); tcp_setsequence(conn->sndseq, conn->isn); #endif dev->d_len = 0; dev->d_sndlen = 0; ninfo("TCP state: TCP_ESTABLISHED\n"); result = tcp_callback(dev, conn, TCP_CONNECTED | TCP_NEWDATA); tcp_appsend(dev, conn, result); return; } /* Inform the application that the connection failed */ (void)tcp_callback(dev, conn, TCP_ABORT); /* The connection is closed after we send the RST */ conn->tcpstateflags = TCP_CLOSED; ninfo("Connection failed - TCP state: TCP_CLOSED\n"); /* We do not send resets in response to resets. */ if ((tcp->flags & TCP_RST) != 0) { goto drop; } tcp_reset(dev); return; case TCP_ESTABLISHED: /* In the ESTABLISHED state, we call upon the application to feed * data into the d_buf. If the TCP_ACKDATA flag is set, the * application should put new data into the buffer, otherwise we are * retransmitting an old segment, and the application should put that * data into the buffer. * * If the incoming packet is a FIN, we should close the connection on * this side as well, and we send out a FIN and enter the LAST_ACK * state. We require that there is no outstanding data; otherwise the * sequence numbers will be screwed up. */ if ((tcp->flags & TCP_FIN) != 0 && (conn->tcpstateflags & TCP_STOPPED) == 0) { /* Needs to be investigated further. * Windows often sends FIN packets together with the last ACK for * the received data. So the socket layer has to get this ACK even * if the connection is going to be closed. */ #if 0 if (conn->unacked > 0) { goto drop; } #endif /* Update the sequence number and indicate that the connection has * been closed. */ net_incr32(conn->rcvseq, dev->d_len + 1); flags |= TCP_CLOSE; if (dev->d_len > 0) { flags |= TCP_NEWDATA; } (void)tcp_callback(dev, conn, flags); conn->tcpstateflags = TCP_LAST_ACK; conn->unacked = 1; conn->nrtx = 0; ninfo("TCP state: TCP_LAST_ACK\n"); tcp_send(dev, conn, TCP_FIN | TCP_ACK, tcpiplen); return; } /* Check the URG flag. If this is set, the segment carries urgent * data that we must pass to the application. */ if ((tcp->flags & TCP_URG) != 0) { #ifdef CONFIG_NET_TCPURGDATA dev->d_urglen = (tcp->urgp[0] << 8) | tcp->urgp[1]; if (dev->d_urglen > dev->d_len) { /* There is more urgent data in the next segment to come. */ dev->d_urglen = dev->d_len; } net_incr32(conn->rcvseq, dev->d_urglen); dev->d_len -= dev->d_urglen; dev->d_urgdata = dev->d_appdata; dev->d_appdata += dev->d_urglen; } else { dev->d_urglen = 0; #else /* CONFIG_NET_TCPURGDATA */ dev->d_appdata = ((FAR uint8_t *)dev->d_appdata) + ((tcp->urgp[0] << 8) | tcp->urgp[1]); dev->d_len -= (tcp->urgp[0] << 8) | tcp->urgp[1]; #endif /* CONFIG_NET_TCPURGDATA */ } /* If d_len > 0 we have TCP data in the packet, and we flag this * by setting the TCP_NEWDATA flag. If the application has stopped * the data flow using TCP_STOPPED, we must not accept any data * packets from the remote host. */ if (dev->d_len > 0 && (conn->tcpstateflags & TCP_STOPPED) == 0) { flags |= TCP_NEWDATA; } /* If this packet constitutes an ACK for outstanding data (flagged * by the TCP_ACKDATA flag), we should call the application since it * might want to send more data. If the incoming packet had data * from the peer (as flagged by the TCP_NEWDATA flag), the * application must also be notified. * * When the application is called, the d_len field * contains the length of the incoming data. The application can * access the incoming data through the global pointer * d_appdata, which usually points hdrlen bytes into the d_buf * array. * * If the application wishes to send any data, this data should be * put into the d_appdata and the length of the data should be * put into d_len. If the application don't have any data to * send, d_len must be set to 0. */ if ((flags & (TCP_NEWDATA | TCP_ACKDATA)) != 0) { /* Clear sndlen and remember the size in d_len. The application * may modify d_len and we will need this value later when we * update the sequence number. */ dev->d_sndlen = 0; len = dev->d_len; /* Provide the packet to the application */ result = tcp_callback(dev, conn, flags); /* If the application successfully handled the incoming data, * then TCP_SNDACK will be set in the result. In this case, * we need to update the sequence number. The ACK will be * send by tcp_appsend(). */ if ((result & TCP_SNDACK) != 0) { /* Update the sequence number using the saved length */ net_incr32(conn->rcvseq, len); } /* Send the response, ACKing the data or not, as appropriate */ tcp_appsend(dev, conn, result); return; } goto drop; case TCP_LAST_ACK: /* We can close this connection if the peer has acknowledged our * FIN. This is indicated by the TCP_ACKDATA flag. */ if ((flags & TCP_ACKDATA) != 0) { conn->tcpstateflags = TCP_CLOSED; ninfo("TCP_LAST_ACK TCP state: TCP_CLOSED\n"); (void)tcp_callback(dev, conn, TCP_CLOSE); } break; case TCP_FIN_WAIT_1: /* The application has closed the connection, but the remote host * hasn't closed its end yet. Thus we stay in the FIN_WAIT_1 state * until we receive a FIN from the remote. */ if (dev->d_len > 0) { net_incr32(conn->rcvseq, dev->d_len); } if ((tcp->flags & TCP_FIN) != 0) { if ((flags & TCP_ACKDATA) != 0) { conn->tcpstateflags = TCP_TIME_WAIT; conn->timer = 0; conn->unacked = 0; ninfo("TCP state: TCP_TIME_WAIT\n"); } else { conn->tcpstateflags = TCP_CLOSING; ninfo("TCP state: TCP_CLOSING\n"); } net_incr32(conn->rcvseq, 1); (void)tcp_callback(dev, conn, TCP_CLOSE); tcp_send(dev, conn, TCP_ACK, tcpiplen); return; } else if ((flags & TCP_ACKDATA) != 0) { conn->tcpstateflags = TCP_FIN_WAIT_2; conn->unacked = 0; ninfo("TCP state: TCP_FIN_WAIT_2\n"); goto drop; } if (dev->d_len > 0) { tcp_send(dev, conn, TCP_ACK, tcpiplen); return; } goto drop; case TCP_FIN_WAIT_2: if (dev->d_len > 0) { net_incr32(conn->rcvseq, dev->d_len); } if ((tcp->flags & TCP_FIN) != 0) { conn->tcpstateflags = TCP_TIME_WAIT; conn->timer = 0; ninfo("TCP state: TCP_TIME_WAIT\n"); net_incr32(conn->rcvseq, 1); (void)tcp_callback(dev, conn, TCP_CLOSE); tcp_send(dev, conn, TCP_ACK, tcpiplen); return; } if (dev->d_len > 0) { tcp_send(dev, conn, TCP_ACK, tcpiplen); return; } goto drop; case TCP_TIME_WAIT: tcp_send(dev, conn, TCP_ACK, tcpiplen); return; case TCP_CLOSING: if ((flags & TCP_ACKDATA) != 0) { conn->tcpstateflags = TCP_TIME_WAIT; conn->timer = 0; ninfo("TCP state: TCP_TIME_WAIT\n"); } default: break; } drop: dev->d_len = 0; } /**************************************************************************** * Public Functions ****************************************************************************/ /**************************************************************************** * Name: tcp_ipv4_input * * Description: * Handle incoming TCP input with IPv4 header * * Parameters: * dev - The device driver structure containing the received TCP packet. * * Return: * None * * Assumptions: * Called from the Ethernet driver with the network stack locked * ****************************************************************************/ #ifdef CONFIG_NET_IPv4 void tcp_ipv4_input(FAR struct net_driver_s *dev) { /* Configure to receive an TCP IPv4 packet */ tcp_ipv4_select(dev); /* Then process in the TCP IPv4 input */ tcp_input(dev, IPv4_HDRLEN); } #endif /**************************************************************************** * Name: tcp_ipv6_input * * Description: * Handle incoming TCP input with IPv4 header * * Parameters: * dev - The device driver structure containing the received TCP packet. * * Return: * None * * Assumptions: * Called from the Ethernet driver with the network stack locked * ****************************************************************************/ #ifdef CONFIG_NET_IPv6 void tcp_ipv6_input(FAR struct net_driver_s *dev) { /* Configure to receive an TCP IPv6 packet */ tcp_ipv6_select(dev); /* Then process in the TCP IPv6 input */ tcp_input(dev, IPv6_HDRLEN); } #endif #endif /* CONFIG_NET && CONFIG_NET_TCP */