1 /* $OpenBSD: tcp_subr.c,v 1.206 2025/01/22 09:37:06 bluhm Exp $ */
2 /* $NetBSD: tcp_subr.c,v 1.22 1996/02/13 23:44:00 christos Exp $ */
3
4 /*
5 * Copyright (c) 1982, 1986, 1988, 1990, 1993
6 * The Regents of the University of California. All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 *
32 * @(#)COPYRIGHT 1.1 (NRL) 17 January 1995
33 *
34 * NRL grants permission for redistribution and use in source and binary
35 * forms, with or without modification, of the software and documentation
36 * created at NRL provided that the following conditions are met:
37 *
38 * 1. Redistributions of source code must retain the above copyright
39 * notice, this list of conditions and the following disclaimer.
40 * 2. Redistributions in binary form must reproduce the above copyright
41 * notice, this list of conditions and the following disclaimer in the
42 * documentation and/or other materials provided with the distribution.
43 * 3. All advertising materials mentioning features or use of this software
44 * must display the following acknowledgements:
45 * This product includes software developed by the University of
46 * California, Berkeley and its contributors.
47 * This product includes software developed at the Information
48 * Technology Division, US Naval Research Laboratory.
49 * 4. Neither the name of the NRL nor the names of its contributors
50 * may be used to endorse or promote products derived from this software
51 * without specific prior written permission.
52 *
53 * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
54 * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
55 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
56 * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR
57 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
58 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
59 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
60 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
61 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
62 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
63 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
64 *
65 * The views and conclusions contained in the software and documentation
66 * are those of the authors and should not be interpreted as representing
67 * official policies, either expressed or implied, of the US Naval
68 * Research Laboratory (NRL).
69 */
70
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/mbuf.h>
74 #include <sys/mutex.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/timeout.h>
78 #include <sys/protosw.h>
79 #include <sys/kernel.h>
80 #include <sys/pool.h>
81
82 #include <net/route.h>
83
84 #include <netinet/in.h>
85 #include <netinet/ip.h>
86 #include <netinet/in_pcb.h>
87 #include <netinet/ip_var.h>
88 #include <netinet6/ip6_var.h>
89 #include <netinet/ip_icmp.h>
90 #include <netinet/tcp.h>
91 #include <netinet/tcp_fsm.h>
92 #include <netinet/tcp_seq.h>
93 #include <netinet/tcp_timer.h>
94 #include <netinet/tcp_var.h>
95
96 #ifdef INET6
97 #include <netinet6/ip6protosw.h>
98 #endif /* INET6 */
99
100 #include <crypto/md5.h>
101 #include <crypto/sha2.h>
102
103 /*
104 * Locks used to protect struct members in this file:
105 * I immutable after creation
106 * T tcp_timer_mtx global tcp timer data structures
107 */
108
109 struct mutex tcp_timer_mtx = MUTEX_INITIALIZER(IPL_SOFTNET);
110
111 /* patchable/settable parameters for tcp */
112 int tcp_mssdflt = TCP_MSS;
113 int tcp_rttdflt = TCPTV_SRTTDFLT;
114
115 /* values controllable via sysctl */
116 int tcp_do_rfc1323 = 1;
117 int tcp_do_sack = 1; /* RFC 2018 selective ACKs */
118 int tcp_ack_on_push = 0; /* set to enable immediate ACK-on-PUSH */
119 #ifdef TCP_ECN
120 int tcp_do_ecn = 0; /* RFC3168 ECN enabled/disabled? */
121 #endif
122 int tcp_do_rfc3390 = 2; /* Increase TCP's Initial Window to 10*mss */
123 int tcp_do_tso = 1; /* TCP segmentation offload for output */
124
125 #ifndef TCB_INITIAL_HASH_SIZE
126 #define TCB_INITIAL_HASH_SIZE 128
127 #endif
128
129 int tcp_reass_limit = NMBCLUSTERS / 8; /* hardlimit for tcpqe_pool */
130 int tcp_sackhole_limit = 32*1024; /* hardlimit for sackhl_pool */
131
132 struct pool tcpcb_pool;
133 struct pool tcpqe_pool;
134 struct pool sackhl_pool;
135
136 struct cpumem *tcpcounters; /* tcp statistics */
137
138 u_char tcp_secret[16]; /* [I] */
139 SHA2_CTX tcp_secret_ctx; /* [I] */
140 tcp_seq tcp_iss; /* [T] updated by timer and connection */
141 uint64_t tcp_starttime; /* [I] random offset for tcp_now() */
142
143 /*
144 * Tcp initialization
145 */
146 void
tcp_init(void)147 tcp_init(void)
148 {
149 tcp_iss = 1; /* wrong */
150 /* 0 is treated special so add 1, 63 bits to count is enough */
151 arc4random_buf(&tcp_starttime, sizeof(tcp_starttime));
152 tcp_starttime = 1ULL + (tcp_starttime / 2);
153 pool_init(&tcpcb_pool, sizeof(struct tcpcb), 0, IPL_SOFTNET, 0,
154 "tcpcb", NULL);
155 pool_init(&tcpqe_pool, sizeof(struct tcpqent), 0, IPL_SOFTNET, 0,
156 "tcpqe", NULL);
157 pool_sethardlimit(&tcpqe_pool, tcp_reass_limit, NULL, 0);
158 pool_init(&sackhl_pool, sizeof(struct sackhole), 0, IPL_SOFTNET, 0,
159 "sackhl", NULL);
160 pool_sethardlimit(&sackhl_pool, tcp_sackhole_limit, NULL, 0);
161 in_pcbinit(&tcbtable, TCB_INITIAL_HASH_SIZE);
162 #ifdef INET6
163 in_pcbinit(&tcb6table, TCB_INITIAL_HASH_SIZE);
164 #endif
165 tcpcounters = counters_alloc(tcps_ncounters);
166
167 arc4random_buf(tcp_secret, sizeof(tcp_secret));
168 SHA512Init(&tcp_secret_ctx);
169 SHA512Update(&tcp_secret_ctx, tcp_secret, sizeof(tcp_secret));
170
171 #ifdef INET6
172 /*
173 * Since sizeof(struct ip6_hdr) > sizeof(struct ip), we
174 * do max length checks/computations only on the former.
175 */
176 if (max_protohdr < (sizeof(struct ip6_hdr) + sizeof(struct tcphdr)))
177 max_protohdr = (sizeof(struct ip6_hdr) + sizeof(struct tcphdr));
178 if ((max_linkhdr + sizeof(struct ip6_hdr) + sizeof(struct tcphdr)) >
179 MHLEN)
180 panic("tcp_init");
181
182 icmp6_mtudisc_callback_register(tcp6_mtudisc_callback);
183 #endif /* INET6 */
184
185 /* Initialize the compressed state engine. */
186 syn_cache_init();
187 }
188
189 /*
190 * Create template to be used to send tcp packets on a connection.
191 * Call after host entry created, allocates an mbuf and fills
192 * in a skeletal tcp/ip header, minimizing the amount of work
193 * necessary when the connection is used.
194 *
195 * To support IPv6 in addition to IPv4 and considering that the sizes of
196 * the IPv4 and IPv6 headers are not the same, we now use a separate pointer
197 * for the TCP header. Also, we made the former tcpiphdr header pointer
198 * into just an IP overlay pointer, with casting as appropriate for v6. rja
199 */
200 struct mbuf *
tcp_template(struct tcpcb * tp)201 tcp_template(struct tcpcb *tp)
202 {
203 struct inpcb *inp = tp->t_inpcb;
204 struct mbuf *m;
205 struct tcphdr *th;
206
207 CTASSERT(sizeof(struct ip) + sizeof(struct tcphdr) <= MHLEN);
208 CTASSERT(sizeof(struct ip6_hdr) + sizeof(struct tcphdr) <= MHLEN);
209
210 if ((m = tp->t_template) == 0) {
211 m = m_get(M_DONTWAIT, MT_HEADER);
212 if (m == NULL)
213 return (0);
214
215 switch (tp->pf) {
216 case 0: /*default to PF_INET*/
217 case AF_INET:
218 m->m_len = sizeof(struct ip);
219 break;
220 #ifdef INET6
221 case AF_INET6:
222 m->m_len = sizeof(struct ip6_hdr);
223 break;
224 #endif /* INET6 */
225 }
226 m->m_len += sizeof (struct tcphdr);
227 }
228
229 switch(tp->pf) {
230 case AF_INET:
231 {
232 struct ipovly *ipovly;
233
234 ipovly = mtod(m, struct ipovly *);
235
236 bzero(ipovly->ih_x1, sizeof ipovly->ih_x1);
237 ipovly->ih_pr = IPPROTO_TCP;
238 ipovly->ih_len = htons(sizeof (struct tcphdr));
239 ipovly->ih_src = inp->inp_laddr;
240 ipovly->ih_dst = inp->inp_faddr;
241
242 th = (struct tcphdr *)(mtod(m, caddr_t) +
243 sizeof(struct ip));
244 }
245 break;
246 #ifdef INET6
247 case AF_INET6:
248 {
249 struct ip6_hdr *ip6;
250
251 ip6 = mtod(m, struct ip6_hdr *);
252
253 ip6->ip6_src = inp->inp_laddr6;
254 ip6->ip6_dst = inp->inp_faddr6;
255 ip6->ip6_flow = htonl(0x60000000) |
256 (inp->inp_flowinfo & IPV6_FLOWLABEL_MASK);
257
258 ip6->ip6_nxt = IPPROTO_TCP;
259 ip6->ip6_plen = htons(sizeof(struct tcphdr)); /*XXX*/
260 ip6->ip6_hlim = in6_selecthlim(inp); /*XXX*/
261
262 th = (struct tcphdr *)(mtod(m, caddr_t) +
263 sizeof(struct ip6_hdr));
264 }
265 break;
266 #endif /* INET6 */
267 }
268
269 th->th_sport = inp->inp_lport;
270 th->th_dport = inp->inp_fport;
271 th->th_seq = 0;
272 th->th_ack = 0;
273 th->th_x2 = 0;
274 th->th_off = 5;
275 th->th_flags = 0;
276 th->th_win = 0;
277 th->th_urp = 0;
278 th->th_sum = 0;
279 return (m);
280 }
281
282 /*
283 * Send a single message to the TCP at address specified by
284 * the given TCP/IP header. If m == 0, then we make a copy
285 * of the tcpiphdr at ti and send directly to the addressed host.
286 * This is used to force keep alive messages out using the TCP
287 * template for a connection tp->t_template. If flags are given
288 * then we send a message back to the TCP which originated the
289 * segment ti, and discard the mbuf containing it and any other
290 * attached mbufs.
291 *
292 * In any case the ack and sequence number of the transmitted
293 * segment are as specified by the parameters.
294 */
295 void
tcp_respond(struct tcpcb * tp,caddr_t template,struct tcphdr * th0,tcp_seq ack,tcp_seq seq,int flags,u_int rtableid,uint64_t now)296 tcp_respond(struct tcpcb *tp, caddr_t template, struct tcphdr *th0,
297 tcp_seq ack, tcp_seq seq, int flags, u_int rtableid, uint64_t now)
298 {
299 int tlen;
300 int win = 0;
301 struct mbuf *m = NULL;
302 struct tcphdr *th;
303 struct ip *ip;
304 #ifdef INET6
305 struct ip6_hdr *ip6;
306 #endif
307 int af; /* af on wire */
308
309 if (tp) {
310 struct socket *so = tp->t_inpcb->inp_socket;
311 win = sbspace(so, &so->so_rcv);
312 /*
313 * If this is called with an unconnected
314 * socket/tp/pcb (tp->pf is 0), we lose.
315 */
316 af = tp->pf;
317 } else
318 af = (((struct ip *)template)->ip_v == 6) ? AF_INET6 : AF_INET;
319
320 m = m_gethdr(M_DONTWAIT, MT_HEADER);
321 if (m == NULL)
322 return;
323 m->m_data += max_linkhdr;
324 tlen = 0;
325
326 #define xchg(a,b,type) do { type t; t=a; a=b; b=t; } while (0)
327 switch (af) {
328 #ifdef INET6
329 case AF_INET6:
330 ip6 = mtod(m, struct ip6_hdr *);
331 th = (struct tcphdr *)(ip6 + 1);
332 tlen = sizeof(*ip6) + sizeof(*th);
333 if (th0) {
334 memcpy(ip6, template, sizeof(*ip6));
335 memcpy(th, th0, sizeof(*th));
336 xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr);
337 } else {
338 memcpy(ip6, template, tlen);
339 }
340 break;
341 #endif /* INET6 */
342 case AF_INET:
343 ip = mtod(m, struct ip *);
344 th = (struct tcphdr *)(ip + 1);
345 tlen = sizeof(*ip) + sizeof(*th);
346 if (th0) {
347 memcpy(ip, template, sizeof(*ip));
348 memcpy(th, th0, sizeof(*th));
349 xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, u_int32_t);
350 } else {
351 memcpy(ip, template, tlen);
352 }
353 break;
354 }
355 if (th0)
356 xchg(th->th_dport, th->th_sport, u_int16_t);
357 else
358 flags = TH_ACK;
359 #undef xchg
360
361 th->th_seq = htonl(seq);
362 th->th_ack = htonl(ack);
363 th->th_x2 = 0;
364 th->th_off = sizeof (struct tcphdr) >> 2;
365 th->th_flags = flags;
366 if (tp)
367 win >>= tp->rcv_scale;
368 if (win > TCP_MAXWIN)
369 win = TCP_MAXWIN;
370 th->th_win = htons((u_int16_t)win);
371 th->th_urp = 0;
372
373 if (tp && (tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP &&
374 (flags & TH_RST) == 0 && (tp->t_flags & TF_RCVD_TSTMP)) {
375 u_int32_t *lp = (u_int32_t *)(th + 1);
376 /* Form timestamp option as shown in appendix A of RFC 1323. */
377 *lp++ = htonl(TCPOPT_TSTAMP_HDR);
378 *lp++ = htonl(now + tp->ts_modulate);
379 *lp = htonl(tp->ts_recent);
380 tlen += TCPOLEN_TSTAMP_APPA;
381 th->th_off = (sizeof(struct tcphdr) + TCPOLEN_TSTAMP_APPA) >> 2;
382 }
383
384 m->m_len = tlen;
385 m->m_pkthdr.len = tlen;
386 m->m_pkthdr.ph_ifidx = 0;
387 m->m_pkthdr.csum_flags |= M_TCP_CSUM_OUT;
388
389 /* force routing table */
390 if (tp)
391 m->m_pkthdr.ph_rtableid = tp->t_inpcb->inp_rtableid;
392 else
393 m->m_pkthdr.ph_rtableid = rtableid;
394
395 switch (af) {
396 #ifdef INET6
397 case AF_INET6:
398 ip6->ip6_flow = htonl(0x60000000);
399 ip6->ip6_nxt = IPPROTO_TCP;
400 ip6->ip6_hlim = in6_selecthlim(tp ? tp->t_inpcb : NULL); /*XXX*/
401 ip6->ip6_plen = tlen - sizeof(struct ip6_hdr);
402 ip6->ip6_plen = htons(ip6->ip6_plen);
403 ip6_output(m, tp ? tp->t_inpcb->inp_outputopts6 : NULL,
404 tp ? &tp->t_inpcb->inp_route : NULL,
405 0, NULL,
406 tp ? &tp->t_inpcb->inp_seclevel : NULL);
407 break;
408 #endif /* INET6 */
409 case AF_INET:
410 ip->ip_len = htons(tlen);
411 ip->ip_ttl = ip_defttl;
412 ip->ip_tos = 0;
413 ip_output(m, NULL,
414 tp ? &tp->t_inpcb->inp_route : NULL,
415 ip_mtudisc ? IP_MTUDISC : 0, NULL,
416 tp ? &tp->t_inpcb->inp_seclevel : NULL, 0);
417 break;
418 }
419 }
420
421 /*
422 * Create a new TCP control block, making an
423 * empty reassembly queue and hooking it to the argument
424 * protocol control block.
425 */
426 struct tcpcb *
tcp_newtcpcb(struct inpcb * inp,int wait)427 tcp_newtcpcb(struct inpcb *inp, int wait)
428 {
429 struct tcpcb *tp;
430 int i;
431
432 tp = pool_get(&tcpcb_pool, (wait == M_WAIT ? PR_WAITOK : PR_NOWAIT) |
433 PR_ZERO);
434 if (tp == NULL)
435 return (NULL);
436 TAILQ_INIT(&tp->t_segq);
437 tp->t_maxseg = atomic_load_int(&tcp_mssdflt);
438 tp->t_maxopd = 0;
439
440 tp->t_inpcb = inp;
441 for (i = 0; i < TCPT_NTIMERS; i++)
442 TCP_TIMER_INIT(tp, i);
443 timeout_set_flags(&tp->t_timer_reaper, tcp_timer_reaper, tp,
444 KCLOCK_NONE, TIMEOUT_PROC | TIMEOUT_MPSAFE);
445
446 tp->sack_enable = atomic_load_int(&tcp_do_sack);
447 tp->t_flags = atomic_load_int(&tcp_do_rfc1323) ?
448 (TF_REQ_SCALE|TF_REQ_TSTMP) : 0;
449 /*
450 * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no
451 * rtt estimate. Set rttvar so that srtt + 2 * rttvar gives
452 * reasonable initial retransmit time.
453 */
454 tp->t_srtt = TCPTV_SRTTBASE;
455 tp->t_rttvar = tcp_rttdflt <<
456 (TCP_RTTVAR_SHIFT + TCP_RTT_BASE_SHIFT - 1);
457 tp->t_rttmin = TCPTV_MIN;
458 TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
459 TCPTV_MIN, TCPTV_REXMTMAX);
460 tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT;
461 tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT;
462
463 tp->t_pmtud_mtu_sent = 0;
464 tp->t_pmtud_mss_acked = 0;
465
466 #ifdef INET6
467 if (ISSET(inp->inp_flags, INP_IPV6)) {
468 tp->pf = PF_INET6;
469 inp->inp_ipv6.ip6_hlim = ip6_defhlim;
470 } else
471 #endif
472 {
473 tp->pf = PF_INET;
474 inp->inp_ip.ip_ttl = ip_defttl;
475 }
476
477 inp->inp_ppcb = (caddr_t)tp;
478 return (tp);
479 }
480
481 /*
482 * Drop a TCP connection, reporting
483 * the specified error. If connection is synchronized,
484 * then send a RST to peer.
485 */
486 struct tcpcb *
tcp_drop(struct tcpcb * tp,int errno)487 tcp_drop(struct tcpcb *tp, int errno)
488 {
489 struct socket *so = tp->t_inpcb->inp_socket;
490
491 if (TCPS_HAVERCVDSYN(tp->t_state)) {
492 tp->t_state = TCPS_CLOSED;
493 (void) tcp_output(tp);
494 tcpstat_inc(tcps_drops);
495 } else
496 tcpstat_inc(tcps_conndrops);
497 if (errno == ETIMEDOUT && tp->t_softerror)
498 errno = tp->t_softerror;
499 so->so_error = errno;
500 return (tcp_close(tp));
501 }
502
503 /*
504 * Close a TCP control block:
505 * discard all space held by the tcp
506 * discard internet protocol block
507 * wake up any sleepers
508 */
509 struct tcpcb *
tcp_close(struct tcpcb * tp)510 tcp_close(struct tcpcb *tp)
511 {
512 struct inpcb *inp = tp->t_inpcb;
513 struct socket *so = inp->inp_socket;
514 struct sackhole *p, *q;
515
516 /* free the reassembly queue, if any */
517 tcp_freeq(tp);
518
519 tcp_canceltimers(tp);
520 syn_cache_cleanup(tp);
521
522 /* Free SACK holes. */
523 q = p = tp->snd_holes;
524 while (p != 0) {
525 q = p->next;
526 pool_put(&sackhl_pool, p);
527 p = q;
528 }
529
530 m_free(tp->t_template);
531 /* Free tcpcb after all pending timers have been run. */
532 timeout_add(&tp->t_timer_reaper, 0);
533 inp->inp_ppcb = NULL;
534 soisdisconnected(so);
535 in_pcbdetach(inp);
536 tcpstat_inc(tcps_closed);
537 return (NULL);
538 }
539
540 int
tcp_freeq(struct tcpcb * tp)541 tcp_freeq(struct tcpcb *tp)
542 {
543 struct tcpqent *qe;
544 int rv = 0;
545
546 while ((qe = TAILQ_FIRST(&tp->t_segq)) != NULL) {
547 TAILQ_REMOVE(&tp->t_segq, qe, tcpqe_q);
548 m_freem(qe->tcpqe_m);
549 pool_put(&tcpqe_pool, qe);
550 rv = 1;
551 }
552 return (rv);
553 }
554
555 /*
556 * Compute proper scaling value for receiver window from buffer space
557 */
558
559 void
tcp_rscale(struct tcpcb * tp,u_long hiwat)560 tcp_rscale(struct tcpcb *tp, u_long hiwat)
561 {
562 tp->request_r_scale = 0;
563 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
564 TCP_MAXWIN << tp->request_r_scale < hiwat)
565 tp->request_r_scale++;
566 }
567
568 /*
569 * Notify a tcp user of an asynchronous error;
570 * store error as soft error, but wake up user
571 * (for now, won't do anything until can select for soft error).
572 */
573 void
tcp_notify(struct inpcb * inp,int error)574 tcp_notify(struct inpcb *inp, int error)
575 {
576 struct tcpcb *tp = intotcpcb(inp);
577 struct socket *so = inp->inp_socket;
578
579 /*
580 * Ignore some errors if we are hooked up.
581 * If connection hasn't completed, has retransmitted several times,
582 * and receives a second error, give up now. This is better
583 * than waiting a long time to establish a connection that
584 * can never complete.
585 */
586 if (tp->t_state == TCPS_ESTABLISHED &&
587 (error == EHOSTUNREACH || error == ENETUNREACH ||
588 error == EHOSTDOWN)) {
589 return;
590 } else if (TCPS_HAVEESTABLISHED(tp->t_state) == 0 &&
591 tp->t_rxtshift > 3 && tp->t_softerror)
592 so->so_error = error;
593 else
594 tp->t_softerror = error;
595 wakeup((caddr_t) &so->so_timeo);
596 sorwakeup(so);
597 sowwakeup(so);
598 }
599
600 #ifdef INET6
601 void
tcp6_ctlinput(int cmd,struct sockaddr * sa,u_int rdomain,void * d)602 tcp6_ctlinput(int cmd, struct sockaddr *sa, u_int rdomain, void *d)
603 {
604 struct tcphdr th;
605 struct tcpcb *tp;
606 void (*notify)(struct inpcb *, int) = tcp_notify;
607 struct ip6_hdr *ip6;
608 const struct sockaddr_in6 *sa6_src = NULL;
609 struct sockaddr_in6 *sa6 = satosin6(sa);
610 struct inpcb *inp;
611 struct mbuf *m;
612 tcp_seq seq;
613 int off;
614 struct {
615 u_int16_t th_sport;
616 u_int16_t th_dport;
617 u_int32_t th_seq;
618 } *thp;
619
620 CTASSERT(sizeof(*thp) <= sizeof(th));
621 if (sa->sa_family != AF_INET6 ||
622 sa->sa_len != sizeof(struct sockaddr_in6) ||
623 IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
624 IN6_IS_ADDR_V4MAPPED(&sa6->sin6_addr))
625 return;
626 if ((unsigned)cmd >= PRC_NCMDS)
627 return;
628 else if (cmd == PRC_QUENCH) {
629 /*
630 * Don't honor ICMP Source Quench messages meant for
631 * TCP connections.
632 */
633 /* XXX there's no PRC_QUENCH in IPv6 */
634 return;
635 } else if (PRC_IS_REDIRECT(cmd))
636 notify = in_rtchange, d = NULL;
637 else if (cmd == PRC_MSGSIZE)
638 ; /* special code is present, see below */
639 else if (cmd == PRC_HOSTDEAD)
640 d = NULL;
641 else if (inet6ctlerrmap[cmd] == 0)
642 return;
643
644 /* if the parameter is from icmp6, decode it. */
645 if (d != NULL) {
646 struct ip6ctlparam *ip6cp = (struct ip6ctlparam *)d;
647 m = ip6cp->ip6c_m;
648 ip6 = ip6cp->ip6c_ip6;
649 off = ip6cp->ip6c_off;
650 sa6_src = ip6cp->ip6c_src;
651 } else {
652 m = NULL;
653 ip6 = NULL;
654 sa6_src = &sa6_any;
655 }
656
657 if (ip6) {
658 /*
659 * XXX: We assume that when ip6 is non NULL,
660 * M and OFF are valid.
661 */
662
663 /* check if we can safely examine src and dst ports */
664 if (m->m_pkthdr.len < off + sizeof(*thp))
665 return;
666
667 bzero(&th, sizeof(th));
668 m_copydata(m, off, sizeof(*thp), &th);
669
670 /*
671 * Check to see if we have a valid TCP connection
672 * corresponding to the address in the ICMPv6 message
673 * payload.
674 */
675 inp = in6_pcblookup(&tcb6table, &sa6->sin6_addr,
676 th.th_dport, &sa6_src->sin6_addr, th.th_sport, rdomain);
677 if (cmd == PRC_MSGSIZE) {
678 /*
679 * Depending on the value of "valid" and routing table
680 * size (mtudisc_{hi,lo}wat), we will:
681 * - recalculate the new MTU and create the
682 * corresponding routing entry, or
683 * - ignore the MTU change notification.
684 */
685 icmp6_mtudisc_update((struct ip6ctlparam *)d,
686 inp != NULL);
687 in_pcbunref(inp);
688 return;
689 }
690 if (inp) {
691 seq = ntohl(th.th_seq);
692 if ((tp = intotcpcb(inp)) &&
693 SEQ_GEQ(seq, tp->snd_una) &&
694 SEQ_LT(seq, tp->snd_max))
695 notify(inp, inet6ctlerrmap[cmd]);
696 } else if (inet6ctlerrmap[cmd] == EHOSTUNREACH ||
697 inet6ctlerrmap[cmd] == ENETUNREACH ||
698 inet6ctlerrmap[cmd] == EHOSTDOWN)
699 syn_cache_unreach(sin6tosa_const(sa6_src), sa, &th,
700 rdomain);
701 in_pcbunref(inp);
702 } else {
703 in6_pcbnotify(&tcb6table, sa6, 0,
704 sa6_src, 0, rdomain, cmd, NULL, notify);
705 }
706 }
707 #endif
708
709 void
tcp_ctlinput(int cmd,struct sockaddr * sa,u_int rdomain,void * v)710 tcp_ctlinput(int cmd, struct sockaddr *sa, u_int rdomain, void *v)
711 {
712 struct ip *ip = v;
713 struct tcphdr *th;
714 struct tcpcb *tp;
715 struct inpcb *inp;
716 struct in_addr faddr;
717 tcp_seq seq;
718 u_int mtu;
719 void (*notify)(struct inpcb *, int) = tcp_notify;
720 int errno;
721
722 if (sa->sa_family != AF_INET)
723 return;
724 faddr = satosin(sa)->sin_addr;
725 if (faddr.s_addr == INADDR_ANY)
726 return;
727
728 if ((unsigned)cmd >= PRC_NCMDS)
729 return;
730 errno = inetctlerrmap[cmd];
731 if (cmd == PRC_QUENCH)
732 /*
733 * Don't honor ICMP Source Quench messages meant for
734 * TCP connections.
735 */
736 return;
737 else if (PRC_IS_REDIRECT(cmd))
738 notify = in_rtchange, ip = 0;
739 else if (cmd == PRC_MSGSIZE && ip_mtudisc && ip) {
740 /*
741 * Verify that the packet in the icmp payload refers
742 * to an existing TCP connection.
743 */
744 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
745 seq = ntohl(th->th_seq);
746 inp = in_pcblookup(&tcbtable,
747 ip->ip_dst, th->th_dport, ip->ip_src, th->th_sport,
748 rdomain);
749 if (inp && (tp = intotcpcb(inp)) &&
750 SEQ_GEQ(seq, tp->snd_una) &&
751 SEQ_LT(seq, tp->snd_max)) {
752 struct icmp *icp;
753 icp = (struct icmp *)((caddr_t)ip -
754 offsetof(struct icmp, icmp_ip));
755
756 /*
757 * If the ICMP message advertises a Next-Hop MTU
758 * equal or larger than the maximum packet size we have
759 * ever sent, drop the message.
760 */
761 mtu = (u_int)ntohs(icp->icmp_nextmtu);
762 if (mtu >= tp->t_pmtud_mtu_sent) {
763 in_pcbunref(inp);
764 return;
765 }
766 if (mtu >= tcp_hdrsz(tp) + tp->t_pmtud_mss_acked) {
767 /*
768 * Calculate new MTU, and create corresponding
769 * route (traditional PMTUD).
770 */
771 tp->t_flags &= ~TF_PMTUD_PEND;
772 icmp_mtudisc(icp, inp->inp_rtableid);
773 } else {
774 /*
775 * Record the information got in the ICMP
776 * message; act on it later.
777 * If we had already recorded an ICMP message,
778 * replace the old one only if the new message
779 * refers to an older TCP segment
780 */
781 if (tp->t_flags & TF_PMTUD_PEND) {
782 if (SEQ_LT(tp->t_pmtud_th_seq, seq)) {
783 in_pcbunref(inp);
784 return;
785 }
786 } else
787 tp->t_flags |= TF_PMTUD_PEND;
788 tp->t_pmtud_th_seq = seq;
789 tp->t_pmtud_nextmtu = icp->icmp_nextmtu;
790 tp->t_pmtud_ip_len = icp->icmp_ip.ip_len;
791 tp->t_pmtud_ip_hl = icp->icmp_ip.ip_hl;
792 in_pcbunref(inp);
793 return;
794 }
795 } else {
796 /* ignore if we don't have a matching connection */
797 in_pcbunref(inp);
798 return;
799 }
800 in_pcbunref(inp);
801 notify = tcp_mtudisc, ip = 0;
802 } else if (cmd == PRC_MTUINC)
803 notify = tcp_mtudisc_increase, ip = 0;
804 else if (cmd == PRC_HOSTDEAD)
805 ip = 0;
806 else if (errno == 0)
807 return;
808
809 if (ip) {
810 th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
811 inp = in_pcblookup(&tcbtable,
812 ip->ip_dst, th->th_dport, ip->ip_src, th->th_sport,
813 rdomain);
814 if (inp) {
815 seq = ntohl(th->th_seq);
816 if ((tp = intotcpcb(inp)) &&
817 SEQ_GEQ(seq, tp->snd_una) &&
818 SEQ_LT(seq, tp->snd_max))
819 notify(inp, errno);
820 } else if (inetctlerrmap[cmd] == EHOSTUNREACH ||
821 inetctlerrmap[cmd] == ENETUNREACH ||
822 inetctlerrmap[cmd] == EHOSTDOWN) {
823 struct sockaddr_in sin;
824
825 bzero(&sin, sizeof(sin));
826 sin.sin_len = sizeof(sin);
827 sin.sin_family = AF_INET;
828 sin.sin_port = th->th_sport;
829 sin.sin_addr = ip->ip_src;
830 syn_cache_unreach(sintosa(&sin), sa, th, rdomain);
831 }
832 in_pcbunref(inp);
833 } else
834 in_pcbnotifyall(&tcbtable, satosin(sa), rdomain, errno, notify);
835 }
836
837
838 #ifdef INET6
839 /*
840 * Path MTU Discovery handlers.
841 */
842 void
tcp6_mtudisc_callback(struct sockaddr_in6 * sin6,u_int rdomain)843 tcp6_mtudisc_callback(struct sockaddr_in6 *sin6, u_int rdomain)
844 {
845 in6_pcbnotify(&tcb6table, sin6, 0,
846 &sa6_any, 0, rdomain, PRC_MSGSIZE, NULL, tcp_mtudisc);
847 }
848 #endif /* INET6 */
849
850 /*
851 * On receipt of path MTU corrections, flush old route and replace it
852 * with the new one. Retransmit all unacknowledged packets, to ensure
853 * that all packets will be received.
854 */
855 void
tcp_mtudisc(struct inpcb * inp,int errno)856 tcp_mtudisc(struct inpcb *inp, int errno)
857 {
858 struct tcpcb *tp = intotcpcb(inp);
859 struct rtentry *rt;
860 int orig_maxseg, change = 0;
861
862 if (tp == NULL)
863 return;
864 orig_maxseg = tp->t_maxseg;
865
866 rt = in_pcbrtentry(inp);
867 if (rt != NULL) {
868 unsigned int orig_mtulock = (rt->rt_locks & RTV_MTU);
869
870 /*
871 * If this was not a host route, remove and realloc.
872 */
873 if ((rt->rt_flags & RTF_HOST) == 0) {
874 in_rtchange(inp, errno);
875 if ((rt = in_pcbrtentry(inp)) == NULL)
876 return;
877 }
878 if (orig_mtulock < (rt->rt_locks & RTV_MTU))
879 change = 1;
880 }
881 tcp_mss(tp, -1);
882 if (orig_maxseg > tp->t_maxseg)
883 change = 1;
884
885 /*
886 * Resend unacknowledged packets
887 */
888 tp->snd_nxt = tp->snd_una;
889 if (change || errno > 0)
890 tcp_output(tp);
891 }
892
893 void
tcp_mtudisc_increase(struct inpcb * inp,int errno)894 tcp_mtudisc_increase(struct inpcb *inp, int errno)
895 {
896 struct tcpcb *tp = intotcpcb(inp);
897 struct rtentry *rt = in_pcbrtentry(inp);
898
899 if (tp != 0 && rt != 0) {
900 /*
901 * If this was a host route, remove and realloc.
902 */
903 if (rt->rt_flags & RTF_HOST)
904 in_rtchange(inp, errno);
905
906 /* also takes care of congestion window */
907 tcp_mss(tp, -1);
908 }
909 }
910
911 /*
912 * Generate new ISNs with a method based on RFC1948
913 */
914 #define TCP_ISS_CONN_INC 4096
915
916 void
tcp_set_iss_tsm(struct tcpcb * tp)917 tcp_set_iss_tsm(struct tcpcb *tp)
918 {
919 SHA2_CTX ctx;
920 union {
921 uint8_t bytes[SHA512_DIGEST_LENGTH];
922 uint32_t words[2];
923 } digest;
924 u_int rdomain = rtable_l2(tp->t_inpcb->inp_rtableid);
925 tcp_seq iss;
926
927 mtx_enter(&tcp_timer_mtx);
928 tcp_iss += TCP_ISS_CONN_INC;
929 iss = tcp_iss;
930 mtx_leave(&tcp_timer_mtx);
931
932 ctx = tcp_secret_ctx;
933 SHA512Update(&ctx, &rdomain, sizeof(rdomain));
934 SHA512Update(&ctx, &tp->t_inpcb->inp_lport, sizeof(u_short));
935 SHA512Update(&ctx, &tp->t_inpcb->inp_fport, sizeof(u_short));
936 if (tp->pf == AF_INET6) {
937 SHA512Update(&ctx, &tp->t_inpcb->inp_laddr6,
938 sizeof(struct in6_addr));
939 SHA512Update(&ctx, &tp->t_inpcb->inp_faddr6,
940 sizeof(struct in6_addr));
941 } else {
942 SHA512Update(&ctx, &tp->t_inpcb->inp_laddr,
943 sizeof(struct in_addr));
944 SHA512Update(&ctx, &tp->t_inpcb->inp_faddr,
945 sizeof(struct in_addr));
946 }
947 SHA512Final(digest.bytes, &ctx);
948 tp->iss = digest.words[0] + iss;
949 tp->ts_modulate = digest.words[1];
950 }
951
952 #ifdef TCP_SIGNATURE
953 int
tcp_signature_tdb_attach(void)954 tcp_signature_tdb_attach(void)
955 {
956 return (0);
957 }
958
959 int
tcp_signature_tdb_init(struct tdb * tdbp,const struct xformsw * xsp,struct ipsecinit * ii)960 tcp_signature_tdb_init(struct tdb *tdbp, const struct xformsw *xsp,
961 struct ipsecinit *ii)
962 {
963 if ((ii->ii_authkeylen < 1) || (ii->ii_authkeylen > 80))
964 return (EINVAL);
965
966 tdbp->tdb_amxkey = malloc(ii->ii_authkeylen, M_XDATA, M_NOWAIT);
967 if (tdbp->tdb_amxkey == NULL)
968 return (ENOMEM);
969 memcpy(tdbp->tdb_amxkey, ii->ii_authkey, ii->ii_authkeylen);
970 tdbp->tdb_amxkeylen = ii->ii_authkeylen;
971
972 return (0);
973 }
974
975 int
tcp_signature_tdb_zeroize(struct tdb * tdbp)976 tcp_signature_tdb_zeroize(struct tdb *tdbp)
977 {
978 if (tdbp->tdb_amxkey) {
979 explicit_bzero(tdbp->tdb_amxkey, tdbp->tdb_amxkeylen);
980 free(tdbp->tdb_amxkey, M_XDATA, tdbp->tdb_amxkeylen);
981 tdbp->tdb_amxkey = NULL;
982 }
983
984 return (0);
985 }
986
987 int
tcp_signature_tdb_input(struct mbuf ** mp,struct tdb * tdbp,int skip,int protoff)988 tcp_signature_tdb_input(struct mbuf **mp, struct tdb *tdbp, int skip,
989 int protoff)
990 {
991 m_freemp(mp);
992 return (IPPROTO_DONE);
993 }
994
995 int
tcp_signature_tdb_output(struct mbuf * m,struct tdb * tdbp,int skip,int protoff)996 tcp_signature_tdb_output(struct mbuf *m, struct tdb *tdbp, int skip,
997 int protoff)
998 {
999 m_freem(m);
1000 return (EINVAL);
1001 }
1002
1003 int
tcp_signature_apply(caddr_t fstate,caddr_t data,unsigned int len)1004 tcp_signature_apply(caddr_t fstate, caddr_t data, unsigned int len)
1005 {
1006 MD5Update((MD5_CTX *)fstate, (char *)data, len);
1007 return 0;
1008 }
1009
1010 int
tcp_signature(struct tdb * tdb,int af,struct mbuf * m,struct tcphdr * th,int iphlen,int doswap,char * sig)1011 tcp_signature(struct tdb *tdb, int af, struct mbuf *m, struct tcphdr *th,
1012 int iphlen, int doswap, char *sig)
1013 {
1014 MD5_CTX ctx;
1015 int len;
1016 struct tcphdr th0;
1017
1018 MD5Init(&ctx);
1019
1020 switch(af) {
1021 case 0:
1022 case AF_INET: {
1023 struct ippseudo ippseudo;
1024 struct ip *ip;
1025
1026 ip = mtod(m, struct ip *);
1027
1028 ippseudo.ippseudo_src = ip->ip_src;
1029 ippseudo.ippseudo_dst = ip->ip_dst;
1030 ippseudo.ippseudo_pad = 0;
1031 ippseudo.ippseudo_p = IPPROTO_TCP;
1032 ippseudo.ippseudo_len = htons(m->m_pkthdr.len - iphlen);
1033
1034 MD5Update(&ctx, (char *)&ippseudo,
1035 sizeof(struct ippseudo));
1036 break;
1037 }
1038 #ifdef INET6
1039 case AF_INET6: {
1040 struct ip6_hdr_pseudo ip6pseudo;
1041 struct ip6_hdr *ip6;
1042
1043 ip6 = mtod(m, struct ip6_hdr *);
1044 bzero(&ip6pseudo, sizeof(ip6pseudo));
1045 ip6pseudo.ip6ph_src = ip6->ip6_src;
1046 ip6pseudo.ip6ph_dst = ip6->ip6_dst;
1047 in6_clearscope(&ip6pseudo.ip6ph_src);
1048 in6_clearscope(&ip6pseudo.ip6ph_dst);
1049 ip6pseudo.ip6ph_nxt = IPPROTO_TCP;
1050 ip6pseudo.ip6ph_len = htonl(m->m_pkthdr.len - iphlen);
1051
1052 MD5Update(&ctx, (char *)&ip6pseudo,
1053 sizeof(ip6pseudo));
1054 break;
1055 }
1056 #endif
1057 }
1058
1059 th0 = *th;
1060 th0.th_sum = 0;
1061
1062 if (doswap) {
1063 th0.th_seq = htonl(th0.th_seq);
1064 th0.th_ack = htonl(th0.th_ack);
1065 th0.th_win = htons(th0.th_win);
1066 th0.th_urp = htons(th0.th_urp);
1067 }
1068 MD5Update(&ctx, (char *)&th0, sizeof(th0));
1069
1070 len = m->m_pkthdr.len - iphlen - th->th_off * sizeof(uint32_t);
1071
1072 if (len > 0 &&
1073 m_apply(m, iphlen + th->th_off * sizeof(uint32_t), len,
1074 tcp_signature_apply, (caddr_t)&ctx))
1075 return (-1);
1076
1077 MD5Update(&ctx, tdb->tdb_amxkey, tdb->tdb_amxkeylen);
1078 MD5Final(sig, &ctx);
1079
1080 return (0);
1081 }
1082 #endif /* TCP_SIGNATURE */
1083