]>
Commit | Line | Data |
---|---|---|
f0cbd3ec FB |
1 | /* |
2 | * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994 | |
3 | * The Regents of the University of California. All rights reserved. | |
4 | * | |
5 | * Redistribution and use in source and binary forms, with or without | |
6 | * modification, are permitted provided that the following conditions | |
7 | * are met: | |
8 | * 1. Redistributions of source code must retain the above copyright | |
9 | * notice, this list of conditions and the following disclaimer. | |
10 | * 2. Redistributions in binary form must reproduce the above copyright | |
11 | * notice, this list of conditions and the following disclaimer in the | |
12 | * documentation and/or other materials provided with the distribution. | |
13 | * 3. All advertising materials mentioning features or use of this software | |
14 | * must display the following acknowledgement: | |
15 | * This product includes software developed by the University of | |
16 | * California, Berkeley and its contributors. | |
17 | * 4. Neither the name of the University nor the names of its contributors | |
18 | * may be used to endorse or promote products derived from this software | |
19 | * without specific prior written permission. | |
20 | * | |
21 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND | |
22 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | |
23 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | |
24 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | |
25 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | |
26 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | |
27 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | |
28 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | |
29 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | |
30 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | |
31 | * SUCH DAMAGE. | |
32 | * | |
33 | * @(#)tcp_input.c 8.5 (Berkeley) 4/10/94 | |
34 | * tcp_input.c,v 1.10 1994/10/13 18:36:32 wollman Exp | |
35 | */ | |
36 | ||
37 | /* | |
38 | * Changes and additions relating to SLiRP | |
39 | * Copyright (c) 1995 Danny Gasparovski. | |
5fafdf24 TS |
40 | * |
41 | * Please read the file COPYRIGHT for the | |
f0cbd3ec FB |
42 | * terms and conditions of the copyright. |
43 | */ | |
44 | ||
45 | #include <slirp.h> | |
46 | #include "ip_icmp.h" | |
47 | ||
48 | struct socket tcb; | |
49 | ||
9634d903 | 50 | #define TCPREXMTTHRESH 3 |
f0cbd3ec FB |
51 | struct socket *tcp_last_so = &tcb; |
52 | ||
53 | tcp_seq tcp_iss; /* tcp initial send seq # */ | |
54 | ||
55 | #define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ) | |
56 | ||
57 | /* for modulo comparisons of timestamps */ | |
58 | #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0) | |
59 | #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0) | |
60 | ||
61 | /* | |
62 | * Insert segment ti into reassembly queue of tcp with | |
63 | * control block tp. Return TH_FIN if reassembly now includes | |
64 | * a segment with FIN. The macro form does the common case inline | |
65 | * (segment is the next to be received on an established connection, | |
66 | * and the queue is empty), avoiding linkage into and removal | |
67 | * from the queue and repetition of various conversions. | |
68 | * Set DELACK for segments received in order, but ack immediately | |
69 | * when segments are out of order (so fast retransmit can work). | |
70 | */ | |
71 | #ifdef TCP_ACK_HACK | |
72 | #define TCP_REASS(tp, ti, m, so, flags) {\ | |
73 | if ((ti)->ti_seq == (tp)->rcv_nxt && \ | |
429d0a3d | 74 | tcpfrag_list_empty(tp) && \ |
f0cbd3ec FB |
75 | (tp)->t_state == TCPS_ESTABLISHED) {\ |
76 | if (ti->ti_flags & TH_PUSH) \ | |
77 | tp->t_flags |= TF_ACKNOW; \ | |
78 | else \ | |
79 | tp->t_flags |= TF_DELACK; \ | |
80 | (tp)->rcv_nxt += (ti)->ti_len; \ | |
81 | flags = (ti)->ti_flags & TH_FIN; \ | |
31a60e22 BS |
82 | STAT(tcpstat.tcps_rcvpack++); \ |
83 | STAT(tcpstat.tcps_rcvbyte += (ti)->ti_len); \ | |
f0cbd3ec FB |
84 | if (so->so_emu) { \ |
85 | if (tcp_emu((so),(m))) sbappend((so), (m)); \ | |
86 | } else \ | |
87 | sbappend((so), (m)); \ | |
88 | /* sorwakeup(so); */ \ | |
89 | } else {\ | |
90 | (flags) = tcp_reass((tp), (ti), (m)); \ | |
91 | tp->t_flags |= TF_ACKNOW; \ | |
92 | } \ | |
93 | } | |
94 | #else | |
95 | #define TCP_REASS(tp, ti, m, so, flags) { \ | |
96 | if ((ti)->ti_seq == (tp)->rcv_nxt && \ | |
429d0a3d | 97 | tcpfrag_list_empty(tp) && \ |
f0cbd3ec FB |
98 | (tp)->t_state == TCPS_ESTABLISHED) { \ |
99 | tp->t_flags |= TF_DELACK; \ | |
100 | (tp)->rcv_nxt += (ti)->ti_len; \ | |
101 | flags = (ti)->ti_flags & TH_FIN; \ | |
31a60e22 BS |
102 | STAT(tcpstat.tcps_rcvpack++); \ |
103 | STAT(tcpstat.tcps_rcvbyte += (ti)->ti_len); \ | |
f0cbd3ec FB |
104 | if (so->so_emu) { \ |
105 | if (tcp_emu((so),(m))) sbappend(so, (m)); \ | |
106 | } else \ | |
107 | sbappend((so), (m)); \ | |
108 | /* sorwakeup(so); */ \ | |
109 | } else { \ | |
110 | (flags) = tcp_reass((tp), (ti), (m)); \ | |
111 | tp->t_flags |= TF_ACKNOW; \ | |
112 | } \ | |
113 | } | |
114 | #endif | |
9634d903 BS |
115 | static void tcp_dooptions(struct tcpcb *tp, u_char *cp, int cnt, |
116 | struct tcpiphdr *ti); | |
117 | static void tcp_xmit_timer(register struct tcpcb *tp, int rtt); | |
f0cbd3ec | 118 | |
9634d903 BS |
119 | static int |
120 | tcp_reass(register struct tcpcb *tp, register struct tcpiphdr *ti, | |
121 | struct mbuf *m) | |
f0cbd3ec FB |
122 | { |
123 | register struct tcpiphdr *q; | |
124 | struct socket *so = tp->t_socket; | |
125 | int flags; | |
5fafdf24 | 126 | |
f0cbd3ec FB |
127 | /* |
128 | * Call with ti==0 after become established to | |
129 | * force pre-ESTABLISHED data up to user socket. | |
130 | */ | |
131 | if (ti == 0) | |
132 | goto present; | |
133 | ||
134 | /* | |
135 | * Find a segment which begins after this one does. | |
136 | */ | |
429d0a3d BS |
137 | for (q = tcpfrag_list_first(tp); !tcpfrag_list_end(q, tp); |
138 | q = tcpiphdr_next(q)) | |
f0cbd3ec FB |
139 | if (SEQ_GT(q->ti_seq, ti->ti_seq)) |
140 | break; | |
141 | ||
142 | /* | |
143 | * If there is a preceding segment, it may provide some of | |
144 | * our data already. If so, drop the data from the incoming | |
145 | * segment. If it provides all of our data, drop us. | |
146 | */ | |
429d0a3d | 147 | if (!tcpfrag_list_end(tcpiphdr_prev(q), tp)) { |
f0cbd3ec | 148 | register int i; |
429d0a3d | 149 | q = tcpiphdr_prev(q); |
f0cbd3ec FB |
150 | /* conversion to int (in i) handles seq wraparound */ |
151 | i = q->ti_seq + q->ti_len - ti->ti_seq; | |
152 | if (i > 0) { | |
153 | if (i >= ti->ti_len) { | |
31a60e22 BS |
154 | STAT(tcpstat.tcps_rcvduppack++); |
155 | STAT(tcpstat.tcps_rcvdupbyte += ti->ti_len); | |
f0cbd3ec FB |
156 | m_freem(m); |
157 | /* | |
158 | * Try to present any queued data | |
159 | * at the left window edge to the user. | |
160 | * This is needed after the 3-WHS | |
161 | * completes. | |
162 | */ | |
163 | goto present; /* ??? */ | |
164 | } | |
165 | m_adj(m, i); | |
166 | ti->ti_len -= i; | |
167 | ti->ti_seq += i; | |
168 | } | |
429d0a3d | 169 | q = tcpiphdr_next(q); |
f0cbd3ec | 170 | } |
31a60e22 BS |
171 | STAT(tcpstat.tcps_rcvoopack++); |
172 | STAT(tcpstat.tcps_rcvoobyte += ti->ti_len); | |
429d0a3d | 173 | ti->ti_mbuf = m; |
f0cbd3ec FB |
174 | |
175 | /* | |
176 | * While we overlap succeeding segments trim them or, | |
177 | * if they are completely covered, dequeue them. | |
178 | */ | |
429d0a3d | 179 | while (!tcpfrag_list_end(q, tp)) { |
f0cbd3ec FB |
180 | register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq; |
181 | if (i <= 0) | |
182 | break; | |
183 | if (i < q->ti_len) { | |
184 | q->ti_seq += i; | |
185 | q->ti_len -= i; | |
429d0a3d | 186 | m_adj(q->ti_mbuf, i); |
f0cbd3ec FB |
187 | break; |
188 | } | |
429d0a3d BS |
189 | q = tcpiphdr_next(q); |
190 | m = tcpiphdr_prev(q)->ti_mbuf; | |
191 | remque(tcpiphdr2qlink(tcpiphdr_prev(q))); | |
f0cbd3ec FB |
192 | m_freem(m); |
193 | } | |
194 | ||
195 | /* | |
196 | * Stick new segment in its place. | |
197 | */ | |
429d0a3d | 198 | insque(tcpiphdr2qlink(ti), tcpiphdr2qlink(tcpiphdr_prev(q))); |
f0cbd3ec FB |
199 | |
200 | present: | |
201 | /* | |
202 | * Present data to user, advancing rcv_nxt through | |
203 | * completed sequence space. | |
204 | */ | |
205 | if (!TCPS_HAVEESTABLISHED(tp->t_state)) | |
206 | return (0); | |
429d0a3d BS |
207 | ti = tcpfrag_list_first(tp); |
208 | if (tcpfrag_list_end(ti, tp) || ti->ti_seq != tp->rcv_nxt) | |
f0cbd3ec FB |
209 | return (0); |
210 | if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len) | |
211 | return (0); | |
212 | do { | |
213 | tp->rcv_nxt += ti->ti_len; | |
214 | flags = ti->ti_flags & TH_FIN; | |
429d0a3d BS |
215 | remque(tcpiphdr2qlink(ti)); |
216 | m = ti->ti_mbuf; | |
217 | ti = tcpiphdr_next(ti); | |
f0cbd3ec FB |
218 | /* if (so->so_state & SS_FCANTRCVMORE) */ |
219 | if (so->so_state & SS_FCANTSENDMORE) | |
220 | m_freem(m); | |
221 | else { | |
222 | if (so->so_emu) { | |
223 | if (tcp_emu(so,m)) sbappend(so, m); | |
224 | } else | |
225 | sbappend(so, m); | |
226 | } | |
227 | } while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt); | |
228 | /* sorwakeup(so); */ | |
229 | return (flags); | |
230 | } | |
231 | ||
232 | /* | |
233 | * TCP input routine, follows pages 65-76 of the | |
234 | * protocol specification dated September, 1981 very closely. | |
235 | */ | |
236 | void | |
237 | tcp_input(m, iphlen, inso) | |
238 | register struct mbuf *m; | |
239 | int iphlen; | |
240 | struct socket *inso; | |
241 | { | |
242 | struct ip save_ip, *ip; | |
243 | register struct tcpiphdr *ti; | |
244 | caddr_t optp = NULL; | |
245 | int optlen = 0; | |
246 | int len, tlen, off; | |
247 | register struct tcpcb *tp = 0; | |
248 | register int tiflags; | |
249 | struct socket *so = 0; | |
250 | int todrop, acked, ourfinisacked, needoutput = 0; | |
251 | /* int dropsocket = 0; */ | |
252 | int iss = 0; | |
253 | u_long tiwin; | |
254 | int ret; | |
255 | /* int ts_present = 0; */ | |
a9ba3a85 | 256 | struct ex_list *ex_ptr; |
f0cbd3ec FB |
257 | |
258 | DEBUG_CALL("tcp_input"); | |
5fafdf24 | 259 | DEBUG_ARGS((dfd," m = %8lx iphlen = %2d inso = %lx\n", |
f0cbd3ec | 260 | (long )m, iphlen, (long )inso )); |
5fafdf24 | 261 | |
f0cbd3ec FB |
262 | /* |
263 | * If called with m == 0, then we're continuing the connect | |
264 | */ | |
265 | if (m == NULL) { | |
266 | so = inso; | |
3b46e624 | 267 | |
f0cbd3ec FB |
268 | /* Re-set a few variables */ |
269 | tp = sototcpcb(so); | |
270 | m = so->so_m; | |
271 | so->so_m = 0; | |
272 | ti = so->so_ti; | |
273 | tiwin = ti->ti_win; | |
274 | tiflags = ti->ti_flags; | |
3b46e624 | 275 | |
f0cbd3ec FB |
276 | goto cont_conn; |
277 | } | |
5fafdf24 TS |
278 | |
279 | ||
31a60e22 | 280 | STAT(tcpstat.tcps_rcvtotal++); |
f0cbd3ec FB |
281 | /* |
282 | * Get IP and TCP header together in first mbuf. | |
283 | * Note: IP leaves IP header in first mbuf. | |
284 | */ | |
285 | ti = mtod(m, struct tcpiphdr *); | |
286 | if (iphlen > sizeof(struct ip )) { | |
287 | ip_stripoptions(m, (struct mbuf *)0); | |
288 | iphlen=sizeof(struct ip ); | |
289 | } | |
290 | /* XXX Check if too short */ | |
5fafdf24 | 291 | |
f0cbd3ec FB |
292 | |
293 | /* | |
294 | * Save a copy of the IP header in case we want restore it | |
295 | * for sending an ICMP error message in response. | |
296 | */ | |
297 | ip=mtod(m, struct ip *); | |
5fafdf24 | 298 | save_ip = *ip; |
f0cbd3ec FB |
299 | save_ip.ip_len+= iphlen; |
300 | ||
301 | /* | |
302 | * Checksum extended TCP header and data. | |
303 | */ | |
304 | tlen = ((struct ip *)ti)->ip_len; | |
429d0a3d BS |
305 | tcpiphdr2qlink(ti)->next = tcpiphdr2qlink(ti)->prev = 0; |
306 | memset(&ti->ti_i.ih_mbuf, 0 , sizeof(struct mbuf_ptr)); | |
f0cbd3ec FB |
307 | ti->ti_x1 = 0; |
308 | ti->ti_len = htons((u_int16_t)tlen); | |
309 | len = sizeof(struct ip ) + tlen; | |
310 | /* keep checksum for ICMP reply | |
5fafdf24 | 311 | * ti->ti_sum = cksum(m, len); |
f0cbd3ec FB |
312 | * if (ti->ti_sum) { */ |
313 | if(cksum(m, len)) { | |
31a60e22 | 314 | STAT(tcpstat.tcps_rcvbadsum++); |
f0cbd3ec FB |
315 | goto drop; |
316 | } | |
317 | ||
318 | /* | |
319 | * Check that TCP offset makes sense, | |
320 | * pull out TCP options and adjust length. XXX | |
321 | */ | |
322 | off = ti->ti_off << 2; | |
323 | if (off < sizeof (struct tcphdr) || off > tlen) { | |
31a60e22 | 324 | STAT(tcpstat.tcps_rcvbadoff++); |
f0cbd3ec FB |
325 | goto drop; |
326 | } | |
327 | tlen -= off; | |
328 | ti->ti_len = tlen; | |
329 | if (off > sizeof (struct tcphdr)) { | |
330 | optlen = off - sizeof (struct tcphdr); | |
331 | optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr); | |
332 | ||
5fafdf24 | 333 | /* |
f0cbd3ec FB |
334 | * Do quick retrieval of timestamp options ("options |
335 | * prediction?"). If timestamp is the only option and it's | |
336 | * formatted as recommended in RFC 1323 appendix A, we | |
337 | * quickly get the values now and not bother calling | |
338 | * tcp_dooptions(), etc. | |
339 | */ | |
340 | /* if ((optlen == TCPOLEN_TSTAMP_APPA || | |
341 | * (optlen > TCPOLEN_TSTAMP_APPA && | |
342 | * optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) && | |
343 | * *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) && | |
344 | * (ti->ti_flags & TH_SYN) == 0) { | |
345 | * ts_present = 1; | |
346 | * ts_val = ntohl(*(u_int32_t *)(optp + 4)); | |
347 | * ts_ecr = ntohl(*(u_int32_t *)(optp + 8)); | |
348 | * optp = NULL; / * we've parsed the options * / | |
349 | * } | |
350 | */ | |
351 | } | |
352 | tiflags = ti->ti_flags; | |
5fafdf24 | 353 | |
f0cbd3ec FB |
354 | /* |
355 | * Convert TCP protocol specific fields to host format. | |
356 | */ | |
357 | NTOHL(ti->ti_seq); | |
358 | NTOHL(ti->ti_ack); | |
359 | NTOHS(ti->ti_win); | |
360 | NTOHS(ti->ti_urp); | |
361 | ||
362 | /* | |
363 | * Drop TCP, IP headers and TCP options. | |
364 | */ | |
365 | m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr); | |
366 | m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr); | |
5fafdf24 | 367 | |
a9ba3a85 AL |
368 | if (slirp_restrict) { |
369 | for (ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) | |
370 | if (ex_ptr->ex_fport == ti->ti_dport && | |
371 | (ntohl(ti->ti_dst.s_addr) & 0xff) == ex_ptr->ex_addr) | |
372 | break; | |
373 | ||
374 | if (!ex_ptr) | |
375 | goto drop; | |
376 | } | |
f0cbd3ec FB |
377 | /* |
378 | * Locate pcb for segment. | |
379 | */ | |
380 | findso: | |
381 | so = tcp_last_so; | |
382 | if (so->so_fport != ti->ti_dport || | |
383 | so->so_lport != ti->ti_sport || | |
384 | so->so_laddr.s_addr != ti->ti_src.s_addr || | |
385 | so->so_faddr.s_addr != ti->ti_dst.s_addr) { | |
386 | so = solookup(&tcb, ti->ti_src, ti->ti_sport, | |
387 | ti->ti_dst, ti->ti_dport); | |
388 | if (so) | |
389 | tcp_last_so = so; | |
31a60e22 | 390 | STAT(tcpstat.tcps_socachemiss++); |
f0cbd3ec FB |
391 | } |
392 | ||
393 | /* | |
394 | * If the state is CLOSED (i.e., TCB does not exist) then | |
395 | * all data in the incoming segment is discarded. | |
396 | * If the TCB exists but is in CLOSED state, it is embryonic, | |
397 | * but should either do a listen or a connect soon. | |
398 | * | |
399 | * state == CLOSED means we've done socreate() but haven't | |
5fafdf24 TS |
400 | * attached it to a protocol yet... |
401 | * | |
f0cbd3ec FB |
402 | * XXX If a TCB does not exist, and the TH_SYN flag is |
403 | * the only flag set, then create a session, mark it | |
404 | * as if it was LISTENING, and continue... | |
405 | */ | |
406 | if (so == 0) { | |
407 | if ((tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) != TH_SYN) | |
408 | goto dropwithreset; | |
3b46e624 | 409 | |
f0cbd3ec FB |
410 | if ((so = socreate()) == NULL) |
411 | goto dropwithreset; | |
412 | if (tcp_attach(so) < 0) { | |
413 | free(so); /* Not sofree (if it failed, it's not insqued) */ | |
414 | goto dropwithreset; | |
415 | } | |
3b46e624 | 416 | |
9634d903 BS |
417 | sbreserve(&so->so_snd, TCP_SNDSPACE); |
418 | sbreserve(&so->so_rcv, TCP_RCVSPACE); | |
3b46e624 | 419 | |
f0cbd3ec FB |
420 | /* tcp_last_so = so; */ /* XXX ? */ |
421 | /* tp = sototcpcb(so); */ | |
3b46e624 | 422 | |
f0cbd3ec FB |
423 | so->so_laddr = ti->ti_src; |
424 | so->so_lport = ti->ti_sport; | |
425 | so->so_faddr = ti->ti_dst; | |
426 | so->so_fport = ti->ti_dport; | |
3b46e624 | 427 | |
f0cbd3ec FB |
428 | if ((so->so_iptos = tcp_tos(so)) == 0) |
429 | so->so_iptos = ((struct ip *)ti)->ip_tos; | |
3b46e624 | 430 | |
f0cbd3ec FB |
431 | tp = sototcpcb(so); |
432 | tp->t_state = TCPS_LISTEN; | |
433 | } | |
3b46e624 | 434 | |
f0cbd3ec FB |
435 | /* |
436 | * If this is a still-connecting socket, this probably | |
437 | * a retransmit of the SYN. Whether it's a retransmit SYN | |
438 | * or something else, we nuke it. | |
439 | */ | |
440 | if (so->so_state & SS_ISFCONNECTING) | |
441 | goto drop; | |
442 | ||
443 | tp = sototcpcb(so); | |
5fafdf24 | 444 | |
f0cbd3ec FB |
445 | /* XXX Should never fail */ |
446 | if (tp == 0) | |
447 | goto dropwithreset; | |
448 | if (tp->t_state == TCPS_CLOSED) | |
449 | goto drop; | |
5fafdf24 | 450 | |
f0cbd3ec FB |
451 | /* Unscale the window into a 32-bit value. */ |
452 | /* if ((tiflags & TH_SYN) == 0) | |
453 | * tiwin = ti->ti_win << tp->snd_scale; | |
454 | * else | |
455 | */ | |
456 | tiwin = ti->ti_win; | |
457 | ||
458 | /* | |
459 | * Segment received on connection. | |
460 | * Reset idle time and keep-alive timer. | |
461 | */ | |
462 | tp->t_idle = 0; | |
9634d903 BS |
463 | if (SO_OPTIONS) |
464 | tp->t_timer[TCPT_KEEP] = TCPTV_KEEPINTVL; | |
f0cbd3ec | 465 | else |
9634d903 | 466 | tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_IDLE; |
f0cbd3ec FB |
467 | |
468 | /* | |
469 | * Process options if not in LISTEN state, | |
470 | * else do it below (after getting remote address). | |
471 | */ | |
472 | if (optp && tp->t_state != TCPS_LISTEN) | |
5fafdf24 | 473 | tcp_dooptions(tp, (u_char *)optp, optlen, ti); |
f0cbd3ec FB |
474 | /* , */ |
475 | /* &ts_present, &ts_val, &ts_ecr); */ | |
476 | ||
5fafdf24 | 477 | /* |
f0cbd3ec FB |
478 | * Header prediction: check for the two common cases |
479 | * of a uni-directional data xfer. If the packet has | |
480 | * no control flags, is in-sequence, the window didn't | |
481 | * change and we're not retransmitting, it's a | |
482 | * candidate. If the length is zero and the ack moved | |
483 | * forward, we're the sender side of the xfer. Just | |
484 | * free the data acked & wake any higher level process | |
485 | * that was blocked waiting for space. If the length | |
486 | * is non-zero and the ack didn't move, we're the | |
487 | * receiver side. If we're getting packets in-order | |
488 | * (the reassembly queue is empty), add the data to | |
489 | * the socket buffer and note that we need a delayed ack. | |
490 | * | |
491 | * XXX Some of these tests are not needed | |
492 | * eg: the tiwin == tp->snd_wnd prevents many more | |
493 | * predictions.. with no *real* advantage.. | |
494 | */ | |
495 | if (tp->t_state == TCPS_ESTABLISHED && | |
496 | (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK && | |
497 | /* (!ts_present || TSTMP_GEQ(ts_val, tp->ts_recent)) && */ | |
498 | ti->ti_seq == tp->rcv_nxt && | |
499 | tiwin && tiwin == tp->snd_wnd && | |
500 | tp->snd_nxt == tp->snd_max) { | |
5fafdf24 | 501 | /* |
f0cbd3ec FB |
502 | * If last ACK falls within this segment's sequence numbers, |
503 | * record the timestamp. | |
504 | */ | |
505 | /* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) && | |
506 | * SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) { | |
507 | * tp->ts_recent_age = tcp_now; | |
508 | * tp->ts_recent = ts_val; | |
509 | * } | |
510 | */ | |
511 | if (ti->ti_len == 0) { | |
512 | if (SEQ_GT(ti->ti_ack, tp->snd_una) && | |
513 | SEQ_LEQ(ti->ti_ack, tp->snd_max) && | |
514 | tp->snd_cwnd >= tp->snd_wnd) { | |
515 | /* | |
516 | * this is a pure ack for outstanding data. | |
517 | */ | |
31a60e22 | 518 | STAT(tcpstat.tcps_predack++); |
f0cbd3ec FB |
519 | /* if (ts_present) |
520 | * tcp_xmit_timer(tp, tcp_now-ts_ecr+1); | |
5fafdf24 | 521 | * else |
f0cbd3ec FB |
522 | */ if (tp->t_rtt && |
523 | SEQ_GT(ti->ti_ack, tp->t_rtseq)) | |
524 | tcp_xmit_timer(tp, tp->t_rtt); | |
525 | acked = ti->ti_ack - tp->snd_una; | |
31a60e22 BS |
526 | STAT(tcpstat.tcps_rcvackpack++); |
527 | STAT(tcpstat.tcps_rcvackbyte += acked); | |
f0cbd3ec FB |
528 | sbdrop(&so->so_snd, acked); |
529 | tp->snd_una = ti->ti_ack; | |
530 | m_freem(m); | |
531 | ||
532 | /* | |
533 | * If all outstanding data are acked, stop | |
534 | * retransmit timer, otherwise restart timer | |
535 | * using current (possibly backed-off) value. | |
536 | * If process is waiting for space, | |
537 | * wakeup/selwakeup/signal. If data | |
538 | * are ready to send, let tcp_output | |
539 | * decide between more output or persist. | |
540 | */ | |
541 | if (tp->snd_una == tp->snd_max) | |
542 | tp->t_timer[TCPT_REXMT] = 0; | |
543 | else if (tp->t_timer[TCPT_PERSIST] == 0) | |
544 | tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; | |
545 | ||
5fafdf24 | 546 | /* |
f0cbd3ec FB |
547 | * There's room in so_snd, sowwakup will read() |
548 | * from the socket if we can | |
549 | */ | |
550 | /* if (so->so_snd.sb_flags & SB_NOTIFY) | |
551 | * sowwakeup(so); | |
552 | */ | |
5fafdf24 | 553 | /* |
f0cbd3ec FB |
554 | * This is called because sowwakeup might have |
555 | * put data into so_snd. Since we don't so sowwakeup, | |
556 | * we don't need this.. XXX??? | |
557 | */ | |
558 | if (so->so_snd.sb_cc) | |
559 | (void) tcp_output(tp); | |
560 | ||
561 | return; | |
562 | } | |
563 | } else if (ti->ti_ack == tp->snd_una && | |
429d0a3d | 564 | tcpfrag_list_empty(tp) && |
f0cbd3ec FB |
565 | ti->ti_len <= sbspace(&so->so_rcv)) { |
566 | /* | |
567 | * this is a pure, in-sequence data packet | |
568 | * with nothing on the reassembly queue and | |
569 | * we have enough buffer space to take it. | |
570 | */ | |
31a60e22 | 571 | STAT(tcpstat.tcps_preddat++); |
f0cbd3ec | 572 | tp->rcv_nxt += ti->ti_len; |
31a60e22 BS |
573 | STAT(tcpstat.tcps_rcvpack++); |
574 | STAT(tcpstat.tcps_rcvbyte += ti->ti_len); | |
f0cbd3ec FB |
575 | /* |
576 | * Add data to socket buffer. | |
577 | */ | |
578 | if (so->so_emu) { | |
579 | if (tcp_emu(so,m)) sbappend(so, m); | |
580 | } else | |
581 | sbappend(so, m); | |
3b46e624 | 582 | |
5fafdf24 | 583 | /* |
f0cbd3ec FB |
584 | * XXX This is called when data arrives. Later, check |
585 | * if we can actually write() to the socket | |
586 | * XXX Need to check? It's be NON_BLOCKING | |
587 | */ | |
588 | /* sorwakeup(so); */ | |
3b46e624 | 589 | |
f0cbd3ec FB |
590 | /* |
591 | * If this is a short packet, then ACK now - with Nagel | |
592 | * congestion avoidance sender won't send more until | |
593 | * he gets an ACK. | |
5fafdf24 | 594 | * |
4f552e3b FB |
595 | * It is better to not delay acks at all to maximize |
596 | * TCP throughput. See RFC 2581. | |
5fafdf24 | 597 | */ |
4f552e3b FB |
598 | tp->t_flags |= TF_ACKNOW; |
599 | tcp_output(tp); | |
f0cbd3ec FB |
600 | return; |
601 | } | |
602 | } /* header prediction */ | |
603 | /* | |
604 | * Calculate amount of space in receive window, | |
605 | * and then do TCP input processing. | |
606 | * Receive window is amount of space in rcv queue, | |
607 | * but not less than advertised window. | |
608 | */ | |
609 | { int win; | |
610 | win = sbspace(&so->so_rcv); | |
611 | if (win < 0) | |
612 | win = 0; | |
613 | tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt)); | |
614 | } | |
615 | ||
616 | switch (tp->t_state) { | |
617 | ||
618 | /* | |
619 | * If the state is LISTEN then ignore segment if it contains an RST. | |
620 | * If the segment contains an ACK then it is bad and send a RST. | |
621 | * If it does not contain a SYN then it is not interesting; drop it. | |
622 | * Don't bother responding if the destination was a broadcast. | |
623 | * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial | |
624 | * tp->iss, and send a segment: | |
625 | * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK> | |
626 | * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss. | |
627 | * Fill in remote peer address fields if not previously specified. | |
628 | * Enter SYN_RECEIVED state, and process any other fields of this | |
629 | * segment in this state. | |
630 | */ | |
631 | case TCPS_LISTEN: { | |
632 | ||
633 | if (tiflags & TH_RST) | |
634 | goto drop; | |
635 | if (tiflags & TH_ACK) | |
636 | goto dropwithreset; | |
637 | if ((tiflags & TH_SYN) == 0) | |
638 | goto drop; | |
3b46e624 | 639 | |
f0cbd3ec FB |
640 | /* |
641 | * This has way too many gotos... | |
642 | * But a bit of spaghetti code never hurt anybody :) | |
643 | */ | |
3b46e624 | 644 | |
f0cbd3ec FB |
645 | /* |
646 | * If this is destined for the control address, then flag to | |
647 | * tcp_ctl once connected, otherwise connect | |
648 | */ | |
649 | if ((so->so_faddr.s_addr&htonl(0xffffff00)) == special_addr.s_addr) { | |
650 | int lastbyte=ntohl(so->so_faddr.s_addr) & 0xff; | |
651 | if (lastbyte!=CTL_ALIAS && lastbyte!=CTL_DNS) { | |
652 | #if 0 | |
653 | if(lastbyte==CTL_CMD || lastbyte==CTL_EXEC) { | |
654 | /* Command or exec adress */ | |
655 | so->so_state |= SS_CTL; | |
5fafdf24 | 656 | } else |
a3d4af03 FB |
657 | #endif |
658 | { | |
f0cbd3ec | 659 | /* May be an add exec */ |
f0cbd3ec | 660 | for(ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) { |
5fafdf24 | 661 | if(ex_ptr->ex_fport == so->so_fport && |
f0cbd3ec FB |
662 | lastbyte == ex_ptr->ex_addr) { |
663 | so->so_state |= SS_CTL; | |
664 | break; | |
665 | } | |
666 | } | |
667 | } | |
668 | if(so->so_state & SS_CTL) goto cont_input; | |
f0cbd3ec FB |
669 | } |
670 | /* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */ | |
671 | } | |
3b46e624 | 672 | |
f0cbd3ec FB |
673 | if (so->so_emu & EMU_NOCONNECT) { |
674 | so->so_emu &= ~EMU_NOCONNECT; | |
675 | goto cont_input; | |
676 | } | |
3b46e624 | 677 | |
02d2c54c | 678 | if((tcp_fconnect(so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) { |
f0cbd3ec FB |
679 | u_char code=ICMP_UNREACH_NET; |
680 | DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n", | |
681 | errno,strerror(errno))); | |
682 | if(errno == ECONNREFUSED) { | |
683 | /* ACK the SYN, send RST to refuse the connection */ | |
684 | tcp_respond(tp, ti, m, ti->ti_seq+1, (tcp_seq)0, | |
5fafdf24 | 685 | TH_RST|TH_ACK); |
f0cbd3ec FB |
686 | } else { |
687 | if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST; | |
688 | HTONL(ti->ti_seq); /* restore tcp header */ | |
689 | HTONL(ti->ti_ack); | |
690 | HTONS(ti->ti_win); | |
691 | HTONS(ti->ti_urp); | |
692 | m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr); | |
693 | m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr); | |
694 | *ip=save_ip; | |
695 | icmp_error(m, ICMP_UNREACH,code, 0,strerror(errno)); | |
696 | } | |
697 | tp = tcp_close(tp); | |
698 | m_free(m); | |
699 | } else { | |
700 | /* | |
701 | * Haven't connected yet, save the current mbuf | |
702 | * and ti, and return | |
703 | * XXX Some OS's don't tell us whether the connect() | |
704 | * succeeded or not. So we must time it out. | |
705 | */ | |
706 | so->so_m = m; | |
707 | so->so_ti = ti; | |
708 | tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT; | |
709 | tp->t_state = TCPS_SYN_RECEIVED; | |
710 | } | |
711 | return; | |
712 | ||
3b46e624 | 713 | cont_conn: |
5fafdf24 | 714 | /* m==NULL |
f0cbd3ec FB |
715 | * Check if the connect succeeded |
716 | */ | |
717 | if (so->so_state & SS_NOFDREF) { | |
718 | tp = tcp_close(tp); | |
719 | goto dropwithreset; | |
720 | } | |
3b46e624 | 721 | cont_input: |
f0cbd3ec | 722 | tcp_template(tp); |
3b46e624 | 723 | |
f0cbd3ec FB |
724 | if (optp) |
725 | tcp_dooptions(tp, (u_char *)optp, optlen, ti); | |
726 | /* , */ | |
727 | /* &ts_present, &ts_val, &ts_ecr); */ | |
3b46e624 | 728 | |
f0cbd3ec FB |
729 | if (iss) |
730 | tp->iss = iss; | |
5fafdf24 | 731 | else |
f0cbd3ec FB |
732 | tp->iss = tcp_iss; |
733 | tcp_iss += TCP_ISSINCR/2; | |
734 | tp->irs = ti->ti_seq; | |
735 | tcp_sendseqinit(tp); | |
736 | tcp_rcvseqinit(tp); | |
737 | tp->t_flags |= TF_ACKNOW; | |
738 | tp->t_state = TCPS_SYN_RECEIVED; | |
739 | tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT; | |
31a60e22 | 740 | STAT(tcpstat.tcps_accepts++); |
f0cbd3ec FB |
741 | goto trimthenstep6; |
742 | } /* case TCPS_LISTEN */ | |
5fafdf24 | 743 | |
f0cbd3ec FB |
744 | /* |
745 | * If the state is SYN_SENT: | |
746 | * if seg contains an ACK, but not for our SYN, drop the input. | |
747 | * if seg contains a RST, then drop the connection. | |
748 | * if seg does not contain SYN, then drop it. | |
749 | * Otherwise this is an acceptable SYN segment | |
750 | * initialize tp->rcv_nxt and tp->irs | |
751 | * if seg contains ack then advance tp->snd_una | |
752 | * if SYN has been acked change to ESTABLISHED else SYN_RCVD state | |
753 | * arrange for segment to be acked (eventually) | |
754 | * continue processing rest of data/controls, beginning with URG | |
755 | */ | |
756 | case TCPS_SYN_SENT: | |
757 | if ((tiflags & TH_ACK) && | |
758 | (SEQ_LEQ(ti->ti_ack, tp->iss) || | |
759 | SEQ_GT(ti->ti_ack, tp->snd_max))) | |
760 | goto dropwithreset; | |
761 | ||
762 | if (tiflags & TH_RST) { | |
763 | if (tiflags & TH_ACK) | |
764 | tp = tcp_drop(tp,0); /* XXX Check t_softerror! */ | |
765 | goto drop; | |
766 | } | |
767 | ||
768 | if ((tiflags & TH_SYN) == 0) | |
769 | goto drop; | |
770 | if (tiflags & TH_ACK) { | |
771 | tp->snd_una = ti->ti_ack; | |
772 | if (SEQ_LT(tp->snd_nxt, tp->snd_una)) | |
773 | tp->snd_nxt = tp->snd_una; | |
774 | } | |
775 | ||
776 | tp->t_timer[TCPT_REXMT] = 0; | |
777 | tp->irs = ti->ti_seq; | |
778 | tcp_rcvseqinit(tp); | |
779 | tp->t_flags |= TF_ACKNOW; | |
780 | if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) { | |
31a60e22 | 781 | STAT(tcpstat.tcps_connects++); |
f0cbd3ec FB |
782 | soisfconnected(so); |
783 | tp->t_state = TCPS_ESTABLISHED; | |
3b46e624 | 784 | |
f0cbd3ec FB |
785 | /* Do window scaling on this connection? */ |
786 | /* if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == | |
787 | * (TF_RCVD_SCALE|TF_REQ_SCALE)) { | |
788 | * tp->snd_scale = tp->requested_s_scale; | |
789 | * tp->rcv_scale = tp->request_r_scale; | |
790 | * } | |
791 | */ | |
792 | (void) tcp_reass(tp, (struct tcpiphdr *)0, | |
793 | (struct mbuf *)0); | |
794 | /* | |
795 | * if we didn't have to retransmit the SYN, | |
796 | * use its rtt as our initial srtt & rtt var. | |
797 | */ | |
798 | if (tp->t_rtt) | |
799 | tcp_xmit_timer(tp, tp->t_rtt); | |
800 | } else | |
801 | tp->t_state = TCPS_SYN_RECEIVED; | |
802 | ||
803 | trimthenstep6: | |
804 | /* | |
805 | * Advance ti->ti_seq to correspond to first data byte. | |
806 | * If data, trim to stay within window, | |
807 | * dropping FIN if necessary. | |
808 | */ | |
809 | ti->ti_seq++; | |
810 | if (ti->ti_len > tp->rcv_wnd) { | |
811 | todrop = ti->ti_len - tp->rcv_wnd; | |
812 | m_adj(m, -todrop); | |
813 | ti->ti_len = tp->rcv_wnd; | |
814 | tiflags &= ~TH_FIN; | |
31a60e22 BS |
815 | STAT(tcpstat.tcps_rcvpackafterwin++); |
816 | STAT(tcpstat.tcps_rcvbyteafterwin += todrop); | |
f0cbd3ec FB |
817 | } |
818 | tp->snd_wl1 = ti->ti_seq - 1; | |
819 | tp->rcv_up = ti->ti_seq; | |
820 | goto step6; | |
821 | } /* switch tp->t_state */ | |
822 | /* | |
823 | * States other than LISTEN or SYN_SENT. | |
824 | * First check timestamp, if present. | |
5fafdf24 | 825 | * Then check that at least some bytes of segment are within |
f0cbd3ec FB |
826 | * receive window. If segment begins before rcv_nxt, |
827 | * drop leading data (and SYN); if nothing left, just ack. | |
5fafdf24 | 828 | * |
f0cbd3ec FB |
829 | * RFC 1323 PAWS: If we have a timestamp reply on this segment |
830 | * and it's less than ts_recent, drop it. | |
831 | */ | |
832 | /* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent && | |
833 | * TSTMP_LT(ts_val, tp->ts_recent)) { | |
834 | * | |
835 | */ /* Check to see if ts_recent is over 24 days old. */ | |
836 | /* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) { | |
837 | */ /* | |
838 | * * Invalidate ts_recent. If this segment updates | |
839 | * * ts_recent, the age will be reset later and ts_recent | |
840 | * * will get a valid value. If it does not, setting | |
841 | * * ts_recent to zero will at least satisfy the | |
842 | * * requirement that zero be placed in the timestamp | |
843 | * * echo reply when ts_recent isn't valid. The | |
844 | * * age isn't reset until we get a valid ts_recent | |
845 | * * because we don't want out-of-order segments to be | |
846 | * * dropped when ts_recent is old. | |
847 | * */ | |
848 | /* tp->ts_recent = 0; | |
849 | * } else { | |
850 | * tcpstat.tcps_rcvduppack++; | |
851 | * tcpstat.tcps_rcvdupbyte += ti->ti_len; | |
852 | * tcpstat.tcps_pawsdrop++; | |
853 | * goto dropafterack; | |
854 | * } | |
855 | * } | |
856 | */ | |
857 | ||
858 | todrop = tp->rcv_nxt - ti->ti_seq; | |
859 | if (todrop > 0) { | |
860 | if (tiflags & TH_SYN) { | |
861 | tiflags &= ~TH_SYN; | |
862 | ti->ti_seq++; | |
5fafdf24 | 863 | if (ti->ti_urp > 1) |
f0cbd3ec FB |
864 | ti->ti_urp--; |
865 | else | |
866 | tiflags &= ~TH_URG; | |
867 | todrop--; | |
868 | } | |
869 | /* | |
870 | * Following if statement from Stevens, vol. 2, p. 960. | |
871 | */ | |
872 | if (todrop > ti->ti_len | |
873 | || (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) { | |
874 | /* | |
875 | * Any valid FIN must be to the left of the window. | |
876 | * At this point the FIN must be a duplicate or out | |
877 | * of sequence; drop it. | |
878 | */ | |
879 | tiflags &= ~TH_FIN; | |
3b46e624 | 880 | |
f0cbd3ec FB |
881 | /* |
882 | * Send an ACK to resynchronize and drop any data. | |
883 | * But keep on processing for RST or ACK. | |
884 | */ | |
885 | tp->t_flags |= TF_ACKNOW; | |
886 | todrop = ti->ti_len; | |
31a60e22 BS |
887 | STAT(tcpstat.tcps_rcvduppack++); |
888 | STAT(tcpstat.tcps_rcvdupbyte += todrop); | |
f0cbd3ec | 889 | } else { |
31a60e22 BS |
890 | STAT(tcpstat.tcps_rcvpartduppack++); |
891 | STAT(tcpstat.tcps_rcvpartdupbyte += todrop); | |
f0cbd3ec FB |
892 | } |
893 | m_adj(m, todrop); | |
894 | ti->ti_seq += todrop; | |
895 | ti->ti_len -= todrop; | |
896 | if (ti->ti_urp > todrop) | |
897 | ti->ti_urp -= todrop; | |
898 | else { | |
899 | tiflags &= ~TH_URG; | |
900 | ti->ti_urp = 0; | |
901 | } | |
902 | } | |
903 | /* | |
904 | * If new data are received on a connection after the | |
905 | * user processes are gone, then RST the other end. | |
906 | */ | |
907 | if ((so->so_state & SS_NOFDREF) && | |
908 | tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) { | |
909 | tp = tcp_close(tp); | |
31a60e22 | 910 | STAT(tcpstat.tcps_rcvafterclose++); |
f0cbd3ec FB |
911 | goto dropwithreset; |
912 | } | |
913 | ||
914 | /* | |
915 | * If segment ends after window, drop trailing data | |
916 | * (and PUSH and FIN); if nothing left, just ACK. | |
917 | */ | |
918 | todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd); | |
919 | if (todrop > 0) { | |
31a60e22 | 920 | STAT(tcpstat.tcps_rcvpackafterwin++); |
f0cbd3ec | 921 | if (todrop >= ti->ti_len) { |
31a60e22 | 922 | STAT(tcpstat.tcps_rcvbyteafterwin += ti->ti_len); |
f0cbd3ec FB |
923 | /* |
924 | * If a new connection request is received | |
925 | * while in TIME_WAIT, drop the old connection | |
926 | * and start over if the sequence numbers | |
927 | * are above the previous ones. | |
928 | */ | |
929 | if (tiflags & TH_SYN && | |
930 | tp->t_state == TCPS_TIME_WAIT && | |
931 | SEQ_GT(ti->ti_seq, tp->rcv_nxt)) { | |
932 | iss = tp->rcv_nxt + TCP_ISSINCR; | |
933 | tp = tcp_close(tp); | |
934 | goto findso; | |
935 | } | |
936 | /* | |
937 | * If window is closed can only take segments at | |
938 | * window edge, and have to drop data and PUSH from | |
939 | * incoming segments. Continue processing, but | |
940 | * remember to ack. Otherwise, drop segment | |
941 | * and ack. | |
942 | */ | |
943 | if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) { | |
944 | tp->t_flags |= TF_ACKNOW; | |
31a60e22 | 945 | STAT(tcpstat.tcps_rcvwinprobe++); |
f0cbd3ec FB |
946 | } else |
947 | goto dropafterack; | |
948 | } else | |
31a60e22 | 949 | STAT(tcpstat.tcps_rcvbyteafterwin += todrop); |
f0cbd3ec FB |
950 | m_adj(m, -todrop); |
951 | ti->ti_len -= todrop; | |
952 | tiflags &= ~(TH_PUSH|TH_FIN); | |
953 | } | |
954 | ||
955 | /* | |
956 | * If last ACK falls within this segment's sequence numbers, | |
957 | * record its timestamp. | |
958 | */ | |
959 | /* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) && | |
960 | * SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len + | |
961 | * ((tiflags & (TH_SYN|TH_FIN)) != 0))) { | |
962 | * tp->ts_recent_age = tcp_now; | |
963 | * tp->ts_recent = ts_val; | |
964 | * } | |
965 | */ | |
966 | ||
967 | /* | |
968 | * If the RST bit is set examine the state: | |
969 | * SYN_RECEIVED STATE: | |
970 | * If passive open, return to LISTEN state. | |
971 | * If active open, inform user that connection was refused. | |
972 | * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES: | |
973 | * Inform user that connection was reset, and close tcb. | |
974 | * CLOSING, LAST_ACK, TIME_WAIT STATES | |
975 | * Close the tcb. | |
976 | */ | |
977 | if (tiflags&TH_RST) switch (tp->t_state) { | |
978 | ||
979 | case TCPS_SYN_RECEIVED: | |
980 | /* so->so_error = ECONNREFUSED; */ | |
981 | goto close; | |
982 | ||
983 | case TCPS_ESTABLISHED: | |
984 | case TCPS_FIN_WAIT_1: | |
985 | case TCPS_FIN_WAIT_2: | |
986 | case TCPS_CLOSE_WAIT: | |
987 | /* so->so_error = ECONNRESET; */ | |
988 | close: | |
989 | tp->t_state = TCPS_CLOSED; | |
31a60e22 | 990 | STAT(tcpstat.tcps_drops++); |
f0cbd3ec FB |
991 | tp = tcp_close(tp); |
992 | goto drop; | |
993 | ||
994 | case TCPS_CLOSING: | |
995 | case TCPS_LAST_ACK: | |
996 | case TCPS_TIME_WAIT: | |
997 | tp = tcp_close(tp); | |
998 | goto drop; | |
999 | } | |
1000 | ||
1001 | /* | |
1002 | * If a SYN is in the window, then this is an | |
1003 | * error and we send an RST and drop the connection. | |
1004 | */ | |
1005 | if (tiflags & TH_SYN) { | |
1006 | tp = tcp_drop(tp,0); | |
1007 | goto dropwithreset; | |
1008 | } | |
1009 | ||
1010 | /* | |
1011 | * If the ACK bit is off we drop the segment and return. | |
1012 | */ | |
1013 | if ((tiflags & TH_ACK) == 0) goto drop; | |
1014 | ||
1015 | /* | |
1016 | * Ack processing. | |
1017 | */ | |
1018 | switch (tp->t_state) { | |
1019 | /* | |
1020 | * In SYN_RECEIVED state if the ack ACKs our SYN then enter | |
1021 | * ESTABLISHED state and continue processing, otherwise | |
1022 | * send an RST. una<=ack<=max | |
1023 | */ | |
1024 | case TCPS_SYN_RECEIVED: | |
1025 | ||
1026 | if (SEQ_GT(tp->snd_una, ti->ti_ack) || | |
1027 | SEQ_GT(ti->ti_ack, tp->snd_max)) | |
1028 | goto dropwithreset; | |
31a60e22 | 1029 | STAT(tcpstat.tcps_connects++); |
f0cbd3ec | 1030 | tp->t_state = TCPS_ESTABLISHED; |
5fafdf24 TS |
1031 | /* |
1032 | * The sent SYN is ack'ed with our sequence number +1 | |
1033 | * The first data byte already in the buffer will get | |
f0cbd3ec FB |
1034 | * lost if no correction is made. This is only needed for |
1035 | * SS_CTL since the buffer is empty otherwise. | |
3b46e624 | 1036 | * tp->snd_una++; or: |
f0cbd3ec FB |
1037 | */ |
1038 | tp->snd_una=ti->ti_ack; | |
1039 | if (so->so_state & SS_CTL) { | |
1040 | /* So tcp_ctl reports the right state */ | |
1041 | ret = tcp_ctl(so); | |
1042 | if (ret == 1) { | |
1043 | soisfconnected(so); | |
1044 | so->so_state &= ~SS_CTL; /* success XXX */ | |
1045 | } else if (ret == 2) { | |
1046 | so->so_state = SS_NOFDREF; /* CTL_CMD */ | |
1047 | } else { | |
1048 | needoutput = 1; | |
1049 | tp->t_state = TCPS_FIN_WAIT_1; | |
1050 | } | |
1051 | } else { | |
1052 | soisfconnected(so); | |
1053 | } | |
3b46e624 | 1054 | |
f0cbd3ec FB |
1055 | /* Do window scaling? */ |
1056 | /* if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == | |
1057 | * (TF_RCVD_SCALE|TF_REQ_SCALE)) { | |
1058 | * tp->snd_scale = tp->requested_s_scale; | |
1059 | * tp->rcv_scale = tp->request_r_scale; | |
1060 | * } | |
1061 | */ | |
1062 | (void) tcp_reass(tp, (struct tcpiphdr *)0, (struct mbuf *)0); | |
1063 | tp->snd_wl1 = ti->ti_seq - 1; | |
1064 | /* Avoid ack processing; snd_una==ti_ack => dup ack */ | |
1065 | goto synrx_to_est; | |
1066 | /* fall into ... */ | |
1067 | ||
1068 | /* | |
1069 | * In ESTABLISHED state: drop duplicate ACKs; ACK out of range | |
1070 | * ACKs. If the ack is in the range | |
1071 | * tp->snd_una < ti->ti_ack <= tp->snd_max | |
1072 | * then advance tp->snd_una to ti->ti_ack and drop | |
1073 | * data from the retransmission queue. If this ACK reflects | |
1074 | * more up to date window information we update our window information. | |
1075 | */ | |
1076 | case TCPS_ESTABLISHED: | |
1077 | case TCPS_FIN_WAIT_1: | |
1078 | case TCPS_FIN_WAIT_2: | |
1079 | case TCPS_CLOSE_WAIT: | |
1080 | case TCPS_CLOSING: | |
1081 | case TCPS_LAST_ACK: | |
1082 | case TCPS_TIME_WAIT: | |
1083 | ||
1084 | if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) { | |
1085 | if (ti->ti_len == 0 && tiwin == tp->snd_wnd) { | |
31a60e22 | 1086 | STAT(tcpstat.tcps_rcvdupack++); |
f0cbd3ec FB |
1087 | DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n", |
1088 | (long )m, (long )so)); | |
1089 | /* | |
1090 | * If we have outstanding data (other than | |
1091 | * a window probe), this is a completely | |
1092 | * duplicate ack (ie, window info didn't | |
1093 | * change), the ack is the biggest we've | |
1094 | * seen and we've seen exactly our rexmt | |
1095 | * threshold of them, assume a packet | |
1096 | * has been dropped and retransmit it. | |
1097 | * Kludge snd_nxt & the congestion | |
1098 | * window so we send only this one | |
1099 | * packet. | |
1100 | * | |
1101 | * We know we're losing at the current | |
1102 | * window size so do congestion avoidance | |
1103 | * (set ssthresh to half the current window | |
1104 | * and pull our congestion window back to | |
1105 | * the new ssthresh). | |
1106 | * | |
1107 | * Dup acks mean that packets have left the | |
5fafdf24 | 1108 | * network (they're now cached at the receiver) |
f0cbd3ec FB |
1109 | * so bump cwnd by the amount in the receiver |
1110 | * to keep a constant cwnd packets in the | |
1111 | * network. | |
1112 | */ | |
1113 | if (tp->t_timer[TCPT_REXMT] == 0 || | |
1114 | ti->ti_ack != tp->snd_una) | |
1115 | tp->t_dupacks = 0; | |
9634d903 | 1116 | else if (++tp->t_dupacks == TCPREXMTTHRESH) { |
f0cbd3ec FB |
1117 | tcp_seq onxt = tp->snd_nxt; |
1118 | u_int win = | |
1119 | min(tp->snd_wnd, tp->snd_cwnd) / 2 / | |
1120 | tp->t_maxseg; | |
1121 | ||
1122 | if (win < 2) | |
1123 | win = 2; | |
1124 | tp->snd_ssthresh = win * tp->t_maxseg; | |
1125 | tp->t_timer[TCPT_REXMT] = 0; | |
1126 | tp->t_rtt = 0; | |
1127 | tp->snd_nxt = ti->ti_ack; | |
1128 | tp->snd_cwnd = tp->t_maxseg; | |
1129 | (void) tcp_output(tp); | |
1130 | tp->snd_cwnd = tp->snd_ssthresh + | |
1131 | tp->t_maxseg * tp->t_dupacks; | |
1132 | if (SEQ_GT(onxt, tp->snd_nxt)) | |
1133 | tp->snd_nxt = onxt; | |
1134 | goto drop; | |
9634d903 | 1135 | } else if (tp->t_dupacks > TCPREXMTTHRESH) { |
f0cbd3ec FB |
1136 | tp->snd_cwnd += tp->t_maxseg; |
1137 | (void) tcp_output(tp); | |
1138 | goto drop; | |
1139 | } | |
1140 | } else | |
1141 | tp->t_dupacks = 0; | |
1142 | break; | |
1143 | } | |
1144 | synrx_to_est: | |
1145 | /* | |
1146 | * If the congestion window was inflated to account | |
1147 | * for the other side's cached packets, retract it. | |
1148 | */ | |
9634d903 | 1149 | if (tp->t_dupacks > TCPREXMTTHRESH && |
f0cbd3ec FB |
1150 | tp->snd_cwnd > tp->snd_ssthresh) |
1151 | tp->snd_cwnd = tp->snd_ssthresh; | |
1152 | tp->t_dupacks = 0; | |
1153 | if (SEQ_GT(ti->ti_ack, tp->snd_max)) { | |
31a60e22 | 1154 | STAT(tcpstat.tcps_rcvacktoomuch++); |
f0cbd3ec FB |
1155 | goto dropafterack; |
1156 | } | |
1157 | acked = ti->ti_ack - tp->snd_una; | |
31a60e22 BS |
1158 | STAT(tcpstat.tcps_rcvackpack++); |
1159 | STAT(tcpstat.tcps_rcvackbyte += acked); | |
f0cbd3ec FB |
1160 | |
1161 | /* | |
1162 | * If we have a timestamp reply, update smoothed | |
1163 | * round trip time. If no timestamp is present but | |
1164 | * transmit timer is running and timed sequence | |
1165 | * number was acked, update smoothed round trip time. | |
1166 | * Since we now have an rtt measurement, cancel the | |
1167 | * timer backoff (cf., Phil Karn's retransmit alg.). | |
1168 | * Recompute the initial retransmit timer. | |
1169 | */ | |
1170 | /* if (ts_present) | |
1171 | * tcp_xmit_timer(tp, tcp_now-ts_ecr+1); | |
1172 | * else | |
3b46e624 | 1173 | */ |
f0cbd3ec FB |
1174 | if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq)) |
1175 | tcp_xmit_timer(tp,tp->t_rtt); | |
1176 | ||
1177 | /* | |
1178 | * If all outstanding data is acked, stop retransmit | |
1179 | * timer and remember to restart (more output or persist). | |
1180 | * If there is more data to be acked, restart retransmit | |
1181 | * timer, using current (possibly backed-off) value. | |
1182 | */ | |
1183 | if (ti->ti_ack == tp->snd_max) { | |
1184 | tp->t_timer[TCPT_REXMT] = 0; | |
1185 | needoutput = 1; | |
1186 | } else if (tp->t_timer[TCPT_PERSIST] == 0) | |
1187 | tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; | |
1188 | /* | |
1189 | * When new data is acked, open the congestion window. | |
1190 | * If the window gives us less than ssthresh packets | |
1191 | * in flight, open exponentially (maxseg per packet). | |
1192 | * Otherwise open linearly: maxseg per window | |
1193 | * (maxseg^2 / cwnd per packet). | |
1194 | */ | |
1195 | { | |
1196 | register u_int cw = tp->snd_cwnd; | |
1197 | register u_int incr = tp->t_maxseg; | |
1198 | ||
1199 | if (cw > tp->snd_ssthresh) | |
1200 | incr = incr * incr / cw; | |
1201 | tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale); | |
1202 | } | |
1203 | if (acked > so->so_snd.sb_cc) { | |
1204 | tp->snd_wnd -= so->so_snd.sb_cc; | |
1205 | sbdrop(&so->so_snd, (int )so->so_snd.sb_cc); | |
1206 | ourfinisacked = 1; | |
1207 | } else { | |
1208 | sbdrop(&so->so_snd, acked); | |
1209 | tp->snd_wnd -= acked; | |
1210 | ourfinisacked = 0; | |
1211 | } | |
1212 | /* | |
1213 | * XXX sowwakup is called when data is acked and there's room for | |
5fafdf24 | 1214 | * for more data... it should read() the socket |
f0cbd3ec FB |
1215 | */ |
1216 | /* if (so->so_snd.sb_flags & SB_NOTIFY) | |
1217 | * sowwakeup(so); | |
1218 | */ | |
1219 | tp->snd_una = ti->ti_ack; | |
1220 | if (SEQ_LT(tp->snd_nxt, tp->snd_una)) | |
1221 | tp->snd_nxt = tp->snd_una; | |
1222 | ||
1223 | switch (tp->t_state) { | |
1224 | ||
1225 | /* | |
1226 | * In FIN_WAIT_1 STATE in addition to the processing | |
1227 | * for the ESTABLISHED state if our FIN is now acknowledged | |
1228 | * then enter FIN_WAIT_2. | |
1229 | */ | |
1230 | case TCPS_FIN_WAIT_1: | |
1231 | if (ourfinisacked) { | |
1232 | /* | |
1233 | * If we can't receive any more | |
1234 | * data, then closing user can proceed. | |
1235 | * Starting the timer is contrary to the | |
1236 | * specification, but if we don't get a FIN | |
1237 | * we'll hang forever. | |
1238 | */ | |
1239 | if (so->so_state & SS_FCANTRCVMORE) { | |
1240 | soisfdisconnected(so); | |
9634d903 | 1241 | tp->t_timer[TCPT_2MSL] = TCP_MAXIDLE; |
f0cbd3ec FB |
1242 | } |
1243 | tp->t_state = TCPS_FIN_WAIT_2; | |
1244 | } | |
1245 | break; | |
1246 | ||
1247 | /* | |
1248 | * In CLOSING STATE in addition to the processing for | |
1249 | * the ESTABLISHED state if the ACK acknowledges our FIN | |
1250 | * then enter the TIME-WAIT state, otherwise ignore | |
1251 | * the segment. | |
1252 | */ | |
1253 | case TCPS_CLOSING: | |
1254 | if (ourfinisacked) { | |
1255 | tp->t_state = TCPS_TIME_WAIT; | |
1256 | tcp_canceltimers(tp); | |
1257 | tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; | |
1258 | soisfdisconnected(so); | |
1259 | } | |
1260 | break; | |
1261 | ||
1262 | /* | |
1263 | * In LAST_ACK, we may still be waiting for data to drain | |
1264 | * and/or to be acked, as well as for the ack of our FIN. | |
1265 | * If our FIN is now acknowledged, delete the TCB, | |
1266 | * enter the closed state and return. | |
1267 | */ | |
1268 | case TCPS_LAST_ACK: | |
1269 | if (ourfinisacked) { | |
1270 | tp = tcp_close(tp); | |
1271 | goto drop; | |
1272 | } | |
1273 | break; | |
1274 | ||
1275 | /* | |
1276 | * In TIME_WAIT state the only thing that should arrive | |
1277 | * is a retransmission of the remote FIN. Acknowledge | |
1278 | * it and restart the finack timer. | |
1279 | */ | |
1280 | case TCPS_TIME_WAIT: | |
1281 | tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; | |
1282 | goto dropafterack; | |
1283 | } | |
1284 | } /* switch(tp->t_state) */ | |
1285 | ||
1286 | step6: | |
1287 | /* | |
1288 | * Update window information. | |
1289 | * Don't look at window if no ACK: TAC's send garbage on first SYN. | |
1290 | */ | |
1291 | if ((tiflags & TH_ACK) && | |
5fafdf24 | 1292 | (SEQ_LT(tp->snd_wl1, ti->ti_seq) || |
f0cbd3ec FB |
1293 | (tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) || |
1294 | (tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) { | |
1295 | /* keep track of pure window updates */ | |
1296 | if (ti->ti_len == 0 && | |
1297 | tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd) | |
31a60e22 | 1298 | STAT(tcpstat.tcps_rcvwinupd++); |
f0cbd3ec FB |
1299 | tp->snd_wnd = tiwin; |
1300 | tp->snd_wl1 = ti->ti_seq; | |
1301 | tp->snd_wl2 = ti->ti_ack; | |
1302 | if (tp->snd_wnd > tp->max_sndwnd) | |
1303 | tp->max_sndwnd = tp->snd_wnd; | |
1304 | needoutput = 1; | |
1305 | } | |
1306 | ||
1307 | /* | |
1308 | * Process segments with URG. | |
1309 | */ | |
1310 | if ((tiflags & TH_URG) && ti->ti_urp && | |
1311 | TCPS_HAVERCVDFIN(tp->t_state) == 0) { | |
1312 | /* | |
1313 | * This is a kludge, but if we receive and accept | |
1314 | * random urgent pointers, we'll crash in | |
1315 | * soreceive. It's hard to imagine someone | |
1316 | * actually wanting to send this much urgent data. | |
1317 | */ | |
1318 | if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) { | |
1319 | ti->ti_urp = 0; | |
1320 | tiflags &= ~TH_URG; | |
1321 | goto dodata; | |
1322 | } | |
1323 | /* | |
1324 | * If this segment advances the known urgent pointer, | |
1325 | * then mark the data stream. This should not happen | |
1326 | * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since | |
5fafdf24 | 1327 | * a FIN has been received from the remote side. |
f0cbd3ec FB |
1328 | * In these states we ignore the URG. |
1329 | * | |
1330 | * According to RFC961 (Assigned Protocols), | |
1331 | * the urgent pointer points to the last octet | |
1332 | * of urgent data. We continue, however, | |
1333 | * to consider it to indicate the first octet | |
5fafdf24 | 1334 | * of data past the urgent section as the original |
f0cbd3ec FB |
1335 | * spec states (in one of two places). |
1336 | */ | |
1337 | if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) { | |
1338 | tp->rcv_up = ti->ti_seq + ti->ti_urp; | |
1339 | so->so_urgc = so->so_rcv.sb_cc + | |
1340 | (tp->rcv_up - tp->rcv_nxt); /* -1; */ | |
1341 | tp->rcv_up = ti->ti_seq + ti->ti_urp; | |
3b46e624 | 1342 | |
f0cbd3ec FB |
1343 | } |
1344 | } else | |
1345 | /* | |
1346 | * If no out of band data is expected, | |
1347 | * pull receive urgent pointer along | |
1348 | * with the receive window. | |
1349 | */ | |
1350 | if (SEQ_GT(tp->rcv_nxt, tp->rcv_up)) | |
1351 | tp->rcv_up = tp->rcv_nxt; | |
1352 | dodata: | |
1353 | ||
1354 | /* | |
1355 | * Process the segment text, merging it into the TCP sequencing queue, | |
1356 | * and arranging for acknowledgment of receipt if necessary. | |
1357 | * This process logically involves adjusting tp->rcv_wnd as data | |
1358 | * is presented to the user (this happens in tcp_usrreq.c, | |
1359 | * case PRU_RCVD). If a FIN has already been received on this | |
1360 | * connection then we just ignore the text. | |
1361 | */ | |
1362 | if ((ti->ti_len || (tiflags&TH_FIN)) && | |
1363 | TCPS_HAVERCVDFIN(tp->t_state) == 0) { | |
1364 | TCP_REASS(tp, ti, m, so, tiflags); | |
1365 | /* | |
1366 | * Note the amount of data that peer has sent into | |
1367 | * our window, in order to estimate the sender's | |
1368 | * buffer size. | |
1369 | */ | |
1370 | len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt); | |
1371 | } else { | |
1372 | m_free(m); | |
1373 | tiflags &= ~TH_FIN; | |
1374 | } | |
1375 | ||
1376 | /* | |
1377 | * If FIN is received ACK the FIN and let the user know | |
1378 | * that the connection is closing. | |
1379 | */ | |
1380 | if (tiflags & TH_FIN) { | |
1381 | if (TCPS_HAVERCVDFIN(tp->t_state) == 0) { | |
1382 | /* | |
1383 | * If we receive a FIN we can't send more data, | |
1384 | * set it SS_FDRAIN | |
1385 | * Shutdown the socket if there is no rx data in the | |
1386 | * buffer. | |
1387 | * soread() is called on completion of shutdown() and | |
1388 | * will got to TCPS_LAST_ACK, and use tcp_output() | |
1389 | * to send the FIN. | |
1390 | */ | |
1391 | /* sofcantrcvmore(so); */ | |
1392 | sofwdrain(so); | |
3b46e624 | 1393 | |
f0cbd3ec FB |
1394 | tp->t_flags |= TF_ACKNOW; |
1395 | tp->rcv_nxt++; | |
1396 | } | |
1397 | switch (tp->t_state) { | |
1398 | ||
1399 | /* | |
1400 | * In SYN_RECEIVED and ESTABLISHED STATES | |
1401 | * enter the CLOSE_WAIT state. | |
1402 | */ | |
1403 | case TCPS_SYN_RECEIVED: | |
1404 | case TCPS_ESTABLISHED: | |
1405 | if(so->so_emu == EMU_CTL) /* no shutdown on socket */ | |
1406 | tp->t_state = TCPS_LAST_ACK; | |
5fafdf24 | 1407 | else |
f0cbd3ec FB |
1408 | tp->t_state = TCPS_CLOSE_WAIT; |
1409 | break; | |
1410 | ||
1411 | /* | |
1412 | * If still in FIN_WAIT_1 STATE FIN has not been acked so | |
1413 | * enter the CLOSING state. | |
1414 | */ | |
1415 | case TCPS_FIN_WAIT_1: | |
1416 | tp->t_state = TCPS_CLOSING; | |
1417 | break; | |
1418 | ||
1419 | /* | |
1420 | * In FIN_WAIT_2 state enter the TIME_WAIT state, | |
5fafdf24 | 1421 | * starting the time-wait timer, turning off the other |
f0cbd3ec FB |
1422 | * standard timers. |
1423 | */ | |
1424 | case TCPS_FIN_WAIT_2: | |
1425 | tp->t_state = TCPS_TIME_WAIT; | |
1426 | tcp_canceltimers(tp); | |
1427 | tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; | |
1428 | soisfdisconnected(so); | |
1429 | break; | |
1430 | ||
1431 | /* | |
1432 | * In TIME_WAIT state restart the 2 MSL time_wait timer. | |
1433 | */ | |
1434 | case TCPS_TIME_WAIT: | |
1435 | tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; | |
1436 | break; | |
1437 | } | |
1438 | } | |
1439 | ||
1440 | /* | |
1441 | * If this is a small packet, then ACK now - with Nagel | |
1442 | * congestion avoidance sender won't send more until | |
1443 | * he gets an ACK. | |
5fafdf24 | 1444 | * |
f0cbd3ec FB |
1445 | * See above. |
1446 | */ | |
1447 | /* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) { | |
1448 | */ | |
1449 | /* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg && | |
1450 | * (so->so_iptos & IPTOS_LOWDELAY) == 0) || | |
1451 | * ((so->so_iptos & IPTOS_LOWDELAY) && | |
1452 | * ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) { | |
1453 | */ | |
1454 | if (ti->ti_len && (unsigned)ti->ti_len <= 5 && | |
1455 | ((struct tcpiphdr_2 *)ti)->first_char == (char)27) { | |
1456 | tp->t_flags |= TF_ACKNOW; | |
1457 | } | |
1458 | ||
1459 | /* | |
1460 | * Return any desired output. | |
1461 | */ | |
1462 | if (needoutput || (tp->t_flags & TF_ACKNOW)) { | |
1463 | (void) tcp_output(tp); | |
1464 | } | |
1465 | return; | |
1466 | ||
1467 | dropafterack: | |
1468 | /* | |
1469 | * Generate an ACK dropping incoming segment if it occupies | |
1470 | * sequence space, where the ACK reflects our state. | |
1471 | */ | |
1472 | if (tiflags & TH_RST) | |
1473 | goto drop; | |
1474 | m_freem(m); | |
1475 | tp->t_flags |= TF_ACKNOW; | |
1476 | (void) tcp_output(tp); | |
1477 | return; | |
1478 | ||
1479 | dropwithreset: | |
1480 | /* reuses m if m!=NULL, m_free() unnecessary */ | |
1481 | if (tiflags & TH_ACK) | |
1482 | tcp_respond(tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST); | |
1483 | else { | |
1484 | if (tiflags & TH_SYN) ti->ti_len++; | |
1485 | tcp_respond(tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0, | |
1486 | TH_RST|TH_ACK); | |
1487 | } | |
1488 | ||
1489 | return; | |
1490 | ||
1491 | drop: | |
1492 | /* | |
1493 | * Drop space held by incoming segment and return. | |
1494 | */ | |
1495 | m_free(m); | |
1496 | ||
1497 | return; | |
1498 | } | |
1499 | ||
1500 | /* , ts_present, ts_val, ts_ecr) */ | |
1501 | /* int *ts_present; | |
1502 | * u_int32_t *ts_val, *ts_ecr; | |
1503 | */ | |
9634d903 BS |
1504 | static void |
1505 | tcp_dooptions(struct tcpcb *tp, u_char *cp, int cnt, struct tcpiphdr *ti) | |
f0cbd3ec FB |
1506 | { |
1507 | u_int16_t mss; | |
1508 | int opt, optlen; | |
1509 | ||
1510 | DEBUG_CALL("tcp_dooptions"); | |
1511 | DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt)); | |
1512 | ||
1513 | for (; cnt > 0; cnt -= optlen, cp += optlen) { | |
1514 | opt = cp[0]; | |
1515 | if (opt == TCPOPT_EOL) | |
1516 | break; | |
1517 | if (opt == TCPOPT_NOP) | |
1518 | optlen = 1; | |
1519 | else { | |
1520 | optlen = cp[1]; | |
1521 | if (optlen <= 0) | |
1522 | break; | |
1523 | } | |
1524 | switch (opt) { | |
1525 | ||
1526 | default: | |
1527 | continue; | |
1528 | ||
1529 | case TCPOPT_MAXSEG: | |
1530 | if (optlen != TCPOLEN_MAXSEG) | |
1531 | continue; | |
1532 | if (!(ti->ti_flags & TH_SYN)) | |
1533 | continue; | |
1534 | memcpy((char *) &mss, (char *) cp + 2, sizeof(mss)); | |
1535 | NTOHS(mss); | |
1536 | (void) tcp_mss(tp, mss); /* sets t_maxseg */ | |
1537 | break; | |
1538 | ||
1539 | /* case TCPOPT_WINDOW: | |
1540 | * if (optlen != TCPOLEN_WINDOW) | |
1541 | * continue; | |
1542 | * if (!(ti->ti_flags & TH_SYN)) | |
1543 | * continue; | |
1544 | * tp->t_flags |= TF_RCVD_SCALE; | |
1545 | * tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT); | |
1546 | * break; | |
1547 | */ | |
1548 | /* case TCPOPT_TIMESTAMP: | |
1549 | * if (optlen != TCPOLEN_TIMESTAMP) | |
1550 | * continue; | |
1551 | * *ts_present = 1; | |
1552 | * memcpy((char *) ts_val, (char *)cp + 2, sizeof(*ts_val)); | |
1553 | * NTOHL(*ts_val); | |
1554 | * memcpy((char *) ts_ecr, (char *)cp + 6, sizeof(*ts_ecr)); | |
1555 | * NTOHL(*ts_ecr); | |
1556 | * | |
5fafdf24 | 1557 | */ /* |
f0cbd3ec FB |
1558 | * * A timestamp received in a SYN makes |
1559 | * * it ok to send timestamp requests and replies. | |
1560 | * */ | |
1561 | /* if (ti->ti_flags & TH_SYN) { | |
1562 | * tp->t_flags |= TF_RCVD_TSTMP; | |
1563 | * tp->ts_recent = *ts_val; | |
1564 | * tp->ts_recent_age = tcp_now; | |
1565 | * } | |
1566 | */ break; | |
1567 | } | |
1568 | } | |
1569 | } | |
1570 | ||
1571 | ||
1572 | /* | |
1573 | * Pull out of band byte out of a segment so | |
1574 | * it doesn't appear in the user's data queue. | |
1575 | * It is still reflected in the segment length for | |
1576 | * sequencing purposes. | |
1577 | */ | |
1578 | ||
1579 | #ifdef notdef | |
1580 | ||
1581 | void | |
1582 | tcp_pulloutofband(so, ti, m) | |
1583 | struct socket *so; | |
1584 | struct tcpiphdr *ti; | |
1585 | register struct mbuf *m; | |
1586 | { | |
1587 | int cnt = ti->ti_urp - 1; | |
5fafdf24 | 1588 | |
f0cbd3ec FB |
1589 | while (cnt >= 0) { |
1590 | if (m->m_len > cnt) { | |
1591 | char *cp = mtod(m, caddr_t) + cnt; | |
1592 | struct tcpcb *tp = sototcpcb(so); | |
1593 | ||
1594 | tp->t_iobc = *cp; | |
1595 | tp->t_oobflags |= TCPOOB_HAVEDATA; | |
1596 | memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1)); | |
1597 | m->m_len--; | |
1598 | return; | |
1599 | } | |
1600 | cnt -= m->m_len; | |
1601 | m = m->m_next; /* XXX WRONG! Fix it! */ | |
1602 | if (m == 0) | |
1603 | break; | |
1604 | } | |
1605 | panic("tcp_pulloutofband"); | |
1606 | } | |
1607 | ||
1608 | #endif /* notdef */ | |
1609 | ||
1610 | /* | |
1611 | * Collect new round-trip time estimate | |
1612 | * and update averages and current timeout. | |
1613 | */ | |
1614 | ||
9634d903 BS |
1615 | static void |
1616 | tcp_xmit_timer(register struct tcpcb *tp, int rtt) | |
f0cbd3ec FB |
1617 | { |
1618 | register short delta; | |
1619 | ||
1620 | DEBUG_CALL("tcp_xmit_timer"); | |
1621 | DEBUG_ARG("tp = %lx", (long)tp); | |
1622 | DEBUG_ARG("rtt = %d", rtt); | |
5fafdf24 | 1623 | |
31a60e22 | 1624 | STAT(tcpstat.tcps_rttupdated++); |
f0cbd3ec FB |
1625 | if (tp->t_srtt != 0) { |
1626 | /* | |
1627 | * srtt is stored as fixed point with 3 bits after the | |
1628 | * binary point (i.e., scaled by 8). The following magic | |
1629 | * is equivalent to the smoothing algorithm in rfc793 with | |
1630 | * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed | |
1631 | * point). Adjust rtt to origin 0. | |
1632 | */ | |
1633 | delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT); | |
1634 | if ((tp->t_srtt += delta) <= 0) | |
1635 | tp->t_srtt = 1; | |
1636 | /* | |
1637 | * We accumulate a smoothed rtt variance (actually, a | |
1638 | * smoothed mean difference), then set the retransmit | |
1639 | * timer to smoothed rtt + 4 times the smoothed variance. | |
1640 | * rttvar is stored as fixed point with 2 bits after the | |
1641 | * binary point (scaled by 4). The following is | |
1642 | * equivalent to rfc793 smoothing with an alpha of .75 | |
1643 | * (rttvar = rttvar*3/4 + |delta| / 4). This replaces | |
1644 | * rfc793's wired-in beta. | |
1645 | */ | |
1646 | if (delta < 0) | |
1647 | delta = -delta; | |
1648 | delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT); | |
1649 | if ((tp->t_rttvar += delta) <= 0) | |
1650 | tp->t_rttvar = 1; | |
1651 | } else { | |
5fafdf24 | 1652 | /* |
f0cbd3ec FB |
1653 | * No rtt measurement yet - use the unsmoothed rtt. |
1654 | * Set the variance to half the rtt (so our first | |
1655 | * retransmit happens at 3*rtt). | |
1656 | */ | |
1657 | tp->t_srtt = rtt << TCP_RTT_SHIFT; | |
1658 | tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1); | |
1659 | } | |
1660 | tp->t_rtt = 0; | |
1661 | tp->t_rxtshift = 0; | |
1662 | ||
1663 | /* | |
1664 | * the retransmit should happen at rtt + 4 * rttvar. | |
1665 | * Because of the way we do the smoothing, srtt and rttvar | |
1666 | * will each average +1/2 tick of bias. When we compute | |
1667 | * the retransmit timer, we want 1/2 tick of rounding and | |
1668 | * 1 extra tick because of +-1/2 tick uncertainty in the | |
1669 | * firing of the timer. The bias will give us exactly the | |
1670 | * 1.5 tick we need. But, because the bias is | |
1671 | * statistical, we have to test that we don't drop below | |
1672 | * the minimum feasible timer (which is 2 ticks). | |
1673 | */ | |
1674 | TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), | |
1675 | (short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */ | |
5fafdf24 | 1676 | |
f0cbd3ec FB |
1677 | /* |
1678 | * We received an ack for a packet that wasn't retransmitted; | |
1679 | * it is probably safe to discard any error indications we've | |
1680 | * received recently. This isn't quite right, but close enough | |
1681 | * for now (a route might have failed after we sent a segment, | |
1682 | * and the return path might not be symmetrical). | |
1683 | */ | |
1684 | tp->t_softerror = 0; | |
1685 | } | |
1686 | ||
1687 | /* | |
1688 | * Determine a reasonable value for maxseg size. | |
1689 | * If the route is known, check route for mtu. | |
1690 | * If none, use an mss that can be handled on the outgoing | |
1691 | * interface without forcing IP to fragment; if bigger than | |
1692 | * an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES | |
1693 | * to utilize large mbufs. If no route is found, route has no mtu, | |
1694 | * or the destination isn't local, use a default, hopefully conservative | |
1695 | * size (usually 512 or the default IP max size, but no more than the mtu | |
1696 | * of the interface), as we can't discover anything about intervening | |
1697 | * gateways or networks. We also initialize the congestion/slow start | |
1698 | * window to be a single segment if the destination isn't local. | |
1699 | * While looking at the routing entry, we also initialize other path-dependent | |
1700 | * parameters from pre-set or cached values in the routing entry. | |
1701 | */ | |
1702 | ||
1703 | int | |
1704 | tcp_mss(tp, offer) | |
1705 | register struct tcpcb *tp; | |
1706 | u_int offer; | |
1707 | { | |
1708 | struct socket *so = tp->t_socket; | |
1709 | int mss; | |
5fafdf24 | 1710 | |
f0cbd3ec FB |
1711 | DEBUG_CALL("tcp_mss"); |
1712 | DEBUG_ARG("tp = %lx", (long)tp); | |
1713 | DEBUG_ARG("offer = %d", offer); | |
5fafdf24 | 1714 | |
9634d903 | 1715 | mss = min(IF_MTU, IF_MRU) - sizeof(struct tcpiphdr); |
f0cbd3ec FB |
1716 | if (offer) |
1717 | mss = min(mss, offer); | |
1718 | mss = max(mss, 32); | |
1719 | if (mss < tp->t_maxseg || offer != 0) | |
1720 | tp->t_maxseg = mss; | |
5fafdf24 | 1721 | |
f0cbd3ec | 1722 | tp->snd_cwnd = mss; |
5fafdf24 | 1723 | |
9634d903 BS |
1724 | sbreserve(&so->so_snd, TCP_SNDSPACE + ((TCP_SNDSPACE % mss) ? |
1725 | (mss - (TCP_SNDSPACE % mss)) : | |
1726 | 0)); | |
1727 | sbreserve(&so->so_rcv, TCP_RCVSPACE + ((TCP_RCVSPACE % mss) ? | |
1728 | (mss - (TCP_RCVSPACE % mss)) : | |
1729 | 0)); | |
5fafdf24 | 1730 | |
f0cbd3ec | 1731 | DEBUG_MISC((dfd, " returning mss = %d\n", mss)); |
5fafdf24 | 1732 | |
f0cbd3ec FB |
1733 | return mss; |
1734 | } |