2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
38 #include "opt_inet6.h"
39 #include "opt_ipsec.h"
40 #include "opt_tcpdebug.h"
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/callout.h>
45 #include <sys/kernel.h>
46 #include <sys/sysctl.h>
47 #include <sys/malloc.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/syslog.h>
54 #include <sys/protosw.h>
55 #include <sys/random.h>
59 #include <net/route.h>
61 #include <net/if_var.h>
64 #include <netinet/in.h>
65 #include <netinet/in_kdtrace.h>
66 #include <netinet/in_pcb.h>
67 #include <netinet/in_systm.h>
68 #include <netinet/in_var.h>
69 #include <netinet/ip.h>
70 #include <netinet/ip_icmp.h>
71 #include <netinet/ip_var.h>
73 #include <netinet/ip6.h>
74 #include <netinet6/in6_pcb.h>
75 #include <netinet6/ip6_var.h>
76 #include <netinet6/scope6_var.h>
77 #include <netinet6/nd6.h>
79 #include <netinet/tcp.h>
80 #include <netinet/tcp_fsm.h>
81 #include <netinet/tcp_seq.h>
82 #include <netinet/tcp_timer.h>
83 #include <netinet/tcp_var.h>
84 #include <netinet/tcp_hpts.h>
85 #include <netinet/tcpip.h>
87 #include <netinet/tcp_debug.h>
90 #include <netinet/udp.h>
91 #include <netinet/udp_var.h>
93 #include <netipsec/ipsec_support.h>
95 #include <machine/in_cksum.h>
97 #include <security/mac/mac_framework.h>
99 VNET_DEFINE_STATIC(bool, nolocaltimewait) = true;
100 #define V_nolocaltimewait VNET(nolocaltimewait)
101 SYSCTL_BOOL(_net_inet_tcp, OID_AUTO, nolocaltimewait,
102 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nolocaltimewait), true,
103 "Do not create TCP TIME_WAIT state for local connections");
106 * Move a TCP connection into TIME_WAIT state.
107 * inp is locked, and is unlocked before returning.
109 * This function used to free tcpcb and allocate a compressed TCP time-wait
110 * structure tcptw. This served well for 20 years but is no longer relevant
111 * on modern machines in the modern internet. However, the function remains
112 * so that TCP stacks require less modification and we don't burn the bridge
113 * to go back to using compressed time-wait.
116 tcp_twstart(struct tcpcb *tp)
118 struct inpcb *inp = tp->t_inpcb;
120 bool isipv6 = inp->inp_inc.inc_flags & INC_ISIPV6;
124 INP_WLOCK_ASSERT(inp);
126 /* A dropped inp should never transition to TIME_WAIT state. */
127 KASSERT((inp->inp_flags & INP_DROPPED) == 0, ("tcp_twstart: "
128 "(inp->inp_flags & INP_DROPPED) != 0"));
130 tcp_state_change(tp, TCPS_TIME_WAIT);
131 soisdisconnected(inp->inp_socket);
133 if (tp->t_flags & TF_ACKNOW)
136 if (V_nolocaltimewait && (
138 isipv6 ? in6_localaddr(&inp->in6p_faddr) :
141 in_localip(inp->inp_faddr)
146 if ((tp = tcp_close(tp)) != NULL)
151 tcp_timer_activate(tp, TT_2MSL, 2 * V_tcp_msl);
156 * Returns true if the TIME_WAIT state was killed and we should start over,
157 * looking for a pcb in the listen state. Otherwise returns false and frees
160 * For pure SYN-segments the PCB shall be read-locked and the tcpopt pointer
161 * may be NULL. For the rest write-lock and valid tcpopt.
164 tcp_twcheck(struct inpcb *inp, struct tcpopt *to, struct tcphdr *th,
165 struct mbuf *m, int tlen)
167 struct tcpcb *tp = intotcpcb(inp);
173 INP_LOCK_ASSERT(inp);
175 thflags = tcp_get_flags(th);
177 if ((thflags & (TH_SYN | TH_ACK)) == TH_SYN)
178 INP_RLOCK_ASSERT(inp);
180 INP_WLOCK_ASSERT(inp);
182 ("%s: called without options on a non-SYN segment",
188 * NOTE: for FIN_WAIT_2 (to be added later),
189 * must validate sequence number before accepting RST
193 * If the segment contains RST:
194 * Drop the segment - see Stevens, vol. 2, p. 964 and
197 if (thflags & TH_RST)
201 /* PAWS not needed at the moment */
203 * RFC 1323 PAWS: If we have a timestamp reply on this segment
204 * and it's less than ts_recent, drop it.
206 if ((to.to_flags & TOF_TS) != 0 && tp->ts_recent &&
207 TSTMP_LT(to.to_tsval, tp->ts_recent)) {
208 if ((thflags & TH_ACK) == 0)
213 * ts_recent is never updated because we never accept new segments.
217 /* Honor the drop_synfin sysctl variable. */
218 if ((thflags & TH_SYN) && (thflags & TH_FIN) && V_drop_synfin) {
219 if ((s = tcp_log_addrs(&inp->inp_inc, th, NULL, NULL))) {
220 log(LOG_DEBUG, "%s; %s: "
221 "SYN|FIN segment ignored (based on "
222 "sysctl setting)\n", s, __func__);
229 * If a new connection request is received
230 * while in TIME_WAIT, drop the old connection
231 * and start over if the sequence numbers
232 * are above the previous ones.
233 * Allow UDP port number changes in this case.
235 if (((thflags & (TH_SYN | TH_ACK)) == TH_SYN) &&
236 SEQ_GT(th->th_seq, tp->rcv_nxt)) {
238 * In case we can't upgrade our lock just pretend we have
241 if (INP_TRY_UPGRADE(inp) == 0)
243 if ((tp = tcp_close(tp)) != NULL)
245 TCPSTAT_INC(tcps_tw_recycles);
250 * Send RST if UDP port numbers don't match
252 if (tp->t_port != m->m_pkthdr.tcp_tun_port) {
253 if (tcp_get_flags(th) & TH_ACK) {
254 tcp_respond(tp, mtod(m, void *), th, m,
255 (tcp_seq)0, th->th_ack, TH_RST);
257 if (tcp_get_flags(th) & TH_SYN)
259 if (tcp_get_flags(th) & TH_FIN)
261 tcp_respond(tp, mtod(m, void *), th, m,
262 th->th_seq+tlen, (tcp_seq)0, TH_RST|TH_ACK);
265 TCPSTAT_INC(tcps_tw_resets);
270 * Drop the segment if it does not contain an ACK.
272 if ((thflags & TH_ACK) == 0)
275 INP_WLOCK_ASSERT(inp);
278 * If timestamps were negotiated during SYN/ACK and a
279 * segment without a timestamp is received, silently drop
280 * the segment, unless the missing timestamps are tolerated.
281 * See section 3.2 of RFC 7323.
283 if (((to->to_flags & TOF_TS) == 0) && (tp->ts_recent != 0) &&
284 (V_tcp_tolerate_missing_ts == 0)) {
289 * Reset the 2MSL timer if this is a duplicate FIN.
291 if (thflags & TH_FIN) {
292 seq = th->th_seq + tlen + (thflags & TH_SYN ? 1 : 0);
293 if (seq + 1 == tp->rcv_nxt)
294 tcp_timer_activate(tp, TT_2MSL, 2 * V_tcp_msl);
298 * Acknowledge the segment if it has data or is not a duplicate ACK.
300 if (thflags != TH_ACK || tlen != 0 ||
301 th->th_seq != tp->rcv_nxt || th->th_ack != tp->snd_nxt) {
302 TCP_PROBE5(receive, NULL, NULL, m, NULL, th);
303 tcp_respond(tp, mtod(m, void *), th, m, tp->rcv_nxt,
304 tp->snd_nxt, TH_ACK);
306 TCPSTAT_INC(tcps_tw_responds);
310 TCP_PROBE5(receive, NULL, NULL, m, NULL, th);