root / slirp / tcp_input.c @ dc5d0b3d
History | View | Annotate | Download (48.5 kB)
| 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.
|
| 40 |
*
|
| 41 |
* Please read the file COPYRIGHT for the
|
| 42 |
* terms and conditions of the copyright.
|
| 43 |
*/
|
| 44 |
|
| 45 |
#include <slirp.h> |
| 46 |
#include "ip_icmp.h" |
| 47 |
|
| 48 |
struct socket tcb;
|
| 49 |
|
| 50 |
#define min(x,y) ((x) < (y) ? (x) : (y))
|
| 51 |
#define max(x,y) ((x) > (y) ? (x) : (y))
|
| 52 |
|
| 53 |
int tcprexmtthresh = 3; |
| 54 |
struct socket *tcp_last_so = &tcb;
|
| 55 |
|
| 56 |
tcp_seq tcp_iss; /* tcp initial send seq # */
|
| 57 |
|
| 58 |
#define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ) |
| 59 |
|
| 60 |
/* for modulo comparisons of timestamps */
|
| 61 |
#define TSTMP_LT(a,b) ((int)((a)-(b)) < 0) |
| 62 |
#define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0) |
| 63 |
|
| 64 |
/*
|
| 65 |
* Insert segment ti into reassembly queue of tcp with
|
| 66 |
* control block tp. Return TH_FIN if reassembly now includes
|
| 67 |
* a segment with FIN. The macro form does the common case inline
|
| 68 |
* (segment is the next to be received on an established connection,
|
| 69 |
* and the queue is empty), avoiding linkage into and removal
|
| 70 |
* from the queue and repetition of various conversions.
|
| 71 |
* Set DELACK for segments received in order, but ack immediately
|
| 72 |
* when segments are out of order (so fast retransmit can work).
|
| 73 |
*/
|
| 74 |
#ifdef TCP_ACK_HACK
|
| 75 |
#define TCP_REASS(tp, ti, m, so, flags) {\
|
| 76 |
if ((ti)->ti_seq == (tp)->rcv_nxt && \
|
| 77 |
(tp)->seg_next == (tcpiphdrp_32)(tp) && \ |
| 78 |
(tp)->t_state == TCPS_ESTABLISHED) {\
|
| 79 |
if (ti->ti_flags & TH_PUSH) \
|
| 80 |
tp->t_flags |= TF_ACKNOW; \ |
| 81 |
else \
|
| 82 |
tp->t_flags |= TF_DELACK; \ |
| 83 |
(tp)->rcv_nxt += (ti)->ti_len; \ |
| 84 |
flags = (ti)->ti_flags & TH_FIN; \ |
| 85 |
tcpstat.tcps_rcvpack++;\ |
| 86 |
tcpstat.tcps_rcvbyte += (ti)->ti_len;\ |
| 87 |
if (so->so_emu) { \
|
| 88 |
if (tcp_emu((so),(m))) sbappend((so), (m)); \
|
| 89 |
} else \
|
| 90 |
sbappend((so), (m)); \ |
| 91 |
/* sorwakeup(so); */ \
|
| 92 |
} else {\
|
| 93 |
(flags) = tcp_reass((tp), (ti), (m)); \ |
| 94 |
tp->t_flags |= TF_ACKNOW; \ |
| 95 |
} \ |
| 96 |
} |
| 97 |
#else
|
| 98 |
#define TCP_REASS(tp, ti, m, so, flags) { \
|
| 99 |
if ((ti)->ti_seq == (tp)->rcv_nxt && \
|
| 100 |
(tp)->seg_next == (tcpiphdrp_32)(tp) && \ |
| 101 |
(tp)->t_state == TCPS_ESTABLISHED) { \
|
| 102 |
tp->t_flags |= TF_DELACK; \ |
| 103 |
(tp)->rcv_nxt += (ti)->ti_len; \ |
| 104 |
flags = (ti)->ti_flags & TH_FIN; \ |
| 105 |
tcpstat.tcps_rcvpack++;\ |
| 106 |
tcpstat.tcps_rcvbyte += (ti)->ti_len;\ |
| 107 |
if (so->so_emu) { \
|
| 108 |
if (tcp_emu((so),(m))) sbappend(so, (m)); \
|
| 109 |
} else \
|
| 110 |
sbappend((so), (m)); \ |
| 111 |
/* sorwakeup(so); */ \
|
| 112 |
} else { \
|
| 113 |
(flags) = tcp_reass((tp), (ti), (m)); \ |
| 114 |
tp->t_flags |= TF_ACKNOW; \ |
| 115 |
} \ |
| 116 |
} |
| 117 |
#endif
|
| 118 |
|
| 119 |
int
|
| 120 |
tcp_reass(tp, ti, m) |
| 121 |
register struct tcpcb *tp; |
| 122 |
register struct tcpiphdr *ti; |
| 123 |
struct mbuf *m;
|
| 124 |
{
|
| 125 |
register struct tcpiphdr *q; |
| 126 |
struct socket *so = tp->t_socket;
|
| 127 |
int flags;
|
| 128 |
|
| 129 |
/*
|
| 130 |
* Call with ti==0 after become established to
|
| 131 |
* force pre-ESTABLISHED data up to user socket.
|
| 132 |
*/
|
| 133 |
if (ti == 0) |
| 134 |
goto present;
|
| 135 |
|
| 136 |
/*
|
| 137 |
* Find a segment which begins after this one does.
|
| 138 |
*/
|
| 139 |
for (q = (struct tcpiphdr *)tp->seg_next; q != (struct tcpiphdr *)tp; |
| 140 |
q = (struct tcpiphdr *)q->ti_next)
|
| 141 |
if (SEQ_GT(q->ti_seq, ti->ti_seq))
|
| 142 |
break;
|
| 143 |
|
| 144 |
/*
|
| 145 |
* If there is a preceding segment, it may provide some of
|
| 146 |
* our data already. If so, drop the data from the incoming
|
| 147 |
* segment. If it provides all of our data, drop us.
|
| 148 |
*/
|
| 149 |
if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) { |
| 150 |
register int i; |
| 151 |
q = (struct tcpiphdr *)q->ti_prev;
|
| 152 |
/* conversion to int (in i) handles seq wraparound */
|
| 153 |
i = q->ti_seq + q->ti_len - ti->ti_seq; |
| 154 |
if (i > 0) { |
| 155 |
if (i >= ti->ti_len) {
|
| 156 |
tcpstat.tcps_rcvduppack++; |
| 157 |
tcpstat.tcps_rcvdupbyte += ti->ti_len; |
| 158 |
m_freem(m); |
| 159 |
/*
|
| 160 |
* Try to present any queued data
|
| 161 |
* at the left window edge to the user.
|
| 162 |
* This is needed after the 3-WHS
|
| 163 |
* completes.
|
| 164 |
*/
|
| 165 |
goto present; /* ??? */ |
| 166 |
} |
| 167 |
m_adj(m, i); |
| 168 |
ti->ti_len -= i; |
| 169 |
ti->ti_seq += i; |
| 170 |
} |
| 171 |
q = (struct tcpiphdr *)(q->ti_next);
|
| 172 |
} |
| 173 |
tcpstat.tcps_rcvoopack++; |
| 174 |
tcpstat.tcps_rcvoobyte += ti->ti_len; |
| 175 |
REASS_MBUF(ti) = (mbufp_32) m; /* XXX */
|
| 176 |
|
| 177 |
/*
|
| 178 |
* While we overlap succeeding segments trim them or,
|
| 179 |
* if they are completely covered, dequeue them.
|
| 180 |
*/
|
| 181 |
while (q != (struct tcpiphdr *)tp) { |
| 182 |
register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq; |
| 183 |
if (i <= 0) |
| 184 |
break;
|
| 185 |
if (i < q->ti_len) {
|
| 186 |
q->ti_seq += i; |
| 187 |
q->ti_len -= i; |
| 188 |
m_adj((struct mbuf *) REASS_MBUF(q), i);
|
| 189 |
break;
|
| 190 |
} |
| 191 |
q = (struct tcpiphdr *)q->ti_next;
|
| 192 |
m = (struct mbuf *) REASS_MBUF((struct tcpiphdr *)q->ti_prev); |
| 193 |
remque_32((void *)(q->ti_prev));
|
| 194 |
m_freem(m); |
| 195 |
} |
| 196 |
|
| 197 |
/*
|
| 198 |
* Stick new segment in its place.
|
| 199 |
*/
|
| 200 |
insque_32(ti, (void *)(q->ti_prev));
|
| 201 |
|
| 202 |
present:
|
| 203 |
/*
|
| 204 |
* Present data to user, advancing rcv_nxt through
|
| 205 |
* completed sequence space.
|
| 206 |
*/
|
| 207 |
if (!TCPS_HAVEESTABLISHED(tp->t_state))
|
| 208 |
return (0); |
| 209 |
ti = (struct tcpiphdr *) tp->seg_next;
|
| 210 |
if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt) |
| 211 |
return (0); |
| 212 |
if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
|
| 213 |
return (0); |
| 214 |
do {
|
| 215 |
tp->rcv_nxt += ti->ti_len; |
| 216 |
flags = ti->ti_flags & TH_FIN; |
| 217 |
remque_32(ti); |
| 218 |
m = (struct mbuf *) REASS_MBUF(ti); /* XXX */ |
| 219 |
ti = (struct tcpiphdr *)ti->ti_next;
|
| 220 |
/* if (so->so_state & SS_FCANTRCVMORE) */
|
| 221 |
if (so->so_state & SS_FCANTSENDMORE)
|
| 222 |
m_freem(m); |
| 223 |
else {
|
| 224 |
if (so->so_emu) {
|
| 225 |
if (tcp_emu(so,m)) sbappend(so, m);
|
| 226 |
} else
|
| 227 |
sbappend(so, m); |
| 228 |
} |
| 229 |
} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt); |
| 230 |
/* sorwakeup(so); */
|
| 231 |
return (flags);
|
| 232 |
} |
| 233 |
|
| 234 |
/*
|
| 235 |
* TCP input routine, follows pages 65-76 of the
|
| 236 |
* protocol specification dated September, 1981 very closely.
|
| 237 |
*/
|
| 238 |
void
|
| 239 |
tcp_input(m, iphlen, inso) |
| 240 |
register struct mbuf *m; |
| 241 |
int iphlen;
|
| 242 |
struct socket *inso;
|
| 243 |
{
|
| 244 |
struct ip save_ip, *ip;
|
| 245 |
register struct tcpiphdr *ti; |
| 246 |
caddr_t optp = NULL;
|
| 247 |
int optlen = 0; |
| 248 |
int len, tlen, off;
|
| 249 |
register struct tcpcb *tp = 0; |
| 250 |
register int tiflags; |
| 251 |
struct socket *so = 0; |
| 252 |
int todrop, acked, ourfinisacked, needoutput = 0; |
| 253 |
/* int dropsocket = 0; */
|
| 254 |
int iss = 0; |
| 255 |
u_long tiwin; |
| 256 |
int ret;
|
| 257 |
/* int ts_present = 0; */
|
| 258 |
|
| 259 |
DEBUG_CALL("tcp_input");
|
| 260 |
DEBUG_ARGS((dfd," m = %8lx iphlen = %2d inso = %lx\n",
|
| 261 |
(long )m, iphlen, (long )inso )); |
| 262 |
|
| 263 |
/*
|
| 264 |
* If called with m == 0, then we're continuing the connect
|
| 265 |
*/
|
| 266 |
if (m == NULL) { |
| 267 |
so = inso; |
| 268 |
|
| 269 |
/* Re-set a few variables */
|
| 270 |
tp = sototcpcb(so); |
| 271 |
m = so->so_m; |
| 272 |
so->so_m = 0;
|
| 273 |
ti = so->so_ti; |
| 274 |
tiwin = ti->ti_win; |
| 275 |
tiflags = ti->ti_flags; |
| 276 |
|
| 277 |
goto cont_conn;
|
| 278 |
} |
| 279 |
|
| 280 |
|
| 281 |
tcpstat.tcps_rcvtotal++; |
| 282 |
/*
|
| 283 |
* Get IP and TCP header together in first mbuf.
|
| 284 |
* Note: IP leaves IP header in first mbuf.
|
| 285 |
*/
|
| 286 |
ti = mtod(m, struct tcpiphdr *);
|
| 287 |
if (iphlen > sizeof(struct ip )) { |
| 288 |
ip_stripoptions(m, (struct mbuf *)0); |
| 289 |
iphlen=sizeof(struct ip ); |
| 290 |
} |
| 291 |
/* XXX Check if too short */
|
| 292 |
|
| 293 |
|
| 294 |
/*
|
| 295 |
* Save a copy of the IP header in case we want restore it
|
| 296 |
* for sending an ICMP error message in response.
|
| 297 |
*/
|
| 298 |
ip=mtod(m, struct ip *);
|
| 299 |
save_ip = *ip; |
| 300 |
save_ip.ip_len+= iphlen; |
| 301 |
|
| 302 |
/*
|
| 303 |
* Checksum extended TCP header and data.
|
| 304 |
*/
|
| 305 |
tlen = ((struct ip *)ti)->ip_len;
|
| 306 |
ti->ti_next = ti->ti_prev = 0;
|
| 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
|
| 311 |
* ti->ti_sum = cksum(m, len);
|
| 312 |
* if (ti->ti_sum) { */
|
| 313 |
if(cksum(m, len)) {
|
| 314 |
tcpstat.tcps_rcvbadsum++; |
| 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) { |
| 324 |
tcpstat.tcps_rcvbadoff++; |
| 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 |
|
| 333 |
/*
|
| 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; |
| 353 |
|
| 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); |
| 367 |
|
| 368 |
/*
|
| 369 |
* Locate pcb for segment.
|
| 370 |
*/
|
| 371 |
findso:
|
| 372 |
so = tcp_last_so; |
| 373 |
if (so->so_fport != ti->ti_dport ||
|
| 374 |
so->so_lport != ti->ti_sport || |
| 375 |
so->so_laddr.s_addr != ti->ti_src.s_addr || |
| 376 |
so->so_faddr.s_addr != ti->ti_dst.s_addr) {
|
| 377 |
so = solookup(&tcb, ti->ti_src, ti->ti_sport, |
| 378 |
ti->ti_dst, ti->ti_dport); |
| 379 |
if (so)
|
| 380 |
tcp_last_so = so; |
| 381 |
++tcpstat.tcps_socachemiss; |
| 382 |
} |
| 383 |
|
| 384 |
/*
|
| 385 |
* If the state is CLOSED (i.e., TCB does not exist) then
|
| 386 |
* all data in the incoming segment is discarded.
|
| 387 |
* If the TCB exists but is in CLOSED state, it is embryonic,
|
| 388 |
* but should either do a listen or a connect soon.
|
| 389 |
*
|
| 390 |
* state == CLOSED means we've done socreate() but haven't
|
| 391 |
* attached it to a protocol yet...
|
| 392 |
*
|
| 393 |
* XXX If a TCB does not exist, and the TH_SYN flag is
|
| 394 |
* the only flag set, then create a session, mark it
|
| 395 |
* as if it was LISTENING, and continue...
|
| 396 |
*/
|
| 397 |
if (so == 0) { |
| 398 |
if ((tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) != TH_SYN)
|
| 399 |
goto dropwithreset;
|
| 400 |
|
| 401 |
if ((so = socreate()) == NULL) |
| 402 |
goto dropwithreset;
|
| 403 |
if (tcp_attach(so) < 0) { |
| 404 |
free(so); /* Not sofree (if it failed, it's not insqued) */
|
| 405 |
goto dropwithreset;
|
| 406 |
} |
| 407 |
|
| 408 |
sbreserve(&so->so_snd, tcp_sndspace); |
| 409 |
sbreserve(&so->so_rcv, tcp_rcvspace); |
| 410 |
|
| 411 |
/* tcp_last_so = so; */ /* XXX ? */ |
| 412 |
/* tp = sototcpcb(so); */
|
| 413 |
|
| 414 |
so->so_laddr = ti->ti_src; |
| 415 |
so->so_lport = ti->ti_sport; |
| 416 |
so->so_faddr = ti->ti_dst; |
| 417 |
so->so_fport = ti->ti_dport; |
| 418 |
|
| 419 |
if ((so->so_iptos = tcp_tos(so)) == 0) |
| 420 |
so->so_iptos = ((struct ip *)ti)->ip_tos;
|
| 421 |
|
| 422 |
tp = sototcpcb(so); |
| 423 |
tp->t_state = TCPS_LISTEN; |
| 424 |
} |
| 425 |
|
| 426 |
/*
|
| 427 |
* If this is a still-connecting socket, this probably
|
| 428 |
* a retransmit of the SYN. Whether it's a retransmit SYN
|
| 429 |
* or something else, we nuke it.
|
| 430 |
*/
|
| 431 |
if (so->so_state & SS_ISFCONNECTING)
|
| 432 |
goto drop;
|
| 433 |
|
| 434 |
tp = sototcpcb(so); |
| 435 |
|
| 436 |
/* XXX Should never fail */
|
| 437 |
if (tp == 0) |
| 438 |
goto dropwithreset;
|
| 439 |
if (tp->t_state == TCPS_CLOSED)
|
| 440 |
goto drop;
|
| 441 |
|
| 442 |
/* Unscale the window into a 32-bit value. */
|
| 443 |
/* if ((tiflags & TH_SYN) == 0)
|
| 444 |
* tiwin = ti->ti_win << tp->snd_scale;
|
| 445 |
* else
|
| 446 |
*/
|
| 447 |
tiwin = ti->ti_win; |
| 448 |
|
| 449 |
/*
|
| 450 |
* Segment received on connection.
|
| 451 |
* Reset idle time and keep-alive timer.
|
| 452 |
*/
|
| 453 |
tp->t_idle = 0;
|
| 454 |
if (so_options)
|
| 455 |
tp->t_timer[TCPT_KEEP] = tcp_keepintvl; |
| 456 |
else
|
| 457 |
tp->t_timer[TCPT_KEEP] = tcp_keepidle; |
| 458 |
|
| 459 |
/*
|
| 460 |
* Process options if not in LISTEN state,
|
| 461 |
* else do it below (after getting remote address).
|
| 462 |
*/
|
| 463 |
if (optp && tp->t_state != TCPS_LISTEN)
|
| 464 |
tcp_dooptions(tp, (u_char *)optp, optlen, ti); |
| 465 |
/* , */
|
| 466 |
/* &ts_present, &ts_val, &ts_ecr); */
|
| 467 |
|
| 468 |
/*
|
| 469 |
* Header prediction: check for the two common cases
|
| 470 |
* of a uni-directional data xfer. If the packet has
|
| 471 |
* no control flags, is in-sequence, the window didn't
|
| 472 |
* change and we're not retransmitting, it's a
|
| 473 |
* candidate. If the length is zero and the ack moved
|
| 474 |
* forward, we're the sender side of the xfer. Just
|
| 475 |
* free the data acked & wake any higher level process
|
| 476 |
* that was blocked waiting for space. If the length
|
| 477 |
* is non-zero and the ack didn't move, we're the
|
| 478 |
* receiver side. If we're getting packets in-order
|
| 479 |
* (the reassembly queue is empty), add the data to
|
| 480 |
* the socket buffer and note that we need a delayed ack.
|
| 481 |
*
|
| 482 |
* XXX Some of these tests are not needed
|
| 483 |
* eg: the tiwin == tp->snd_wnd prevents many more
|
| 484 |
* predictions.. with no *real* advantage..
|
| 485 |
*/
|
| 486 |
if (tp->t_state == TCPS_ESTABLISHED &&
|
| 487 |
(tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK && |
| 488 |
/* (!ts_present || TSTMP_GEQ(ts_val, tp->ts_recent)) && */
|
| 489 |
ti->ti_seq == tp->rcv_nxt && |
| 490 |
tiwin && tiwin == tp->snd_wnd && |
| 491 |
tp->snd_nxt == tp->snd_max) {
|
| 492 |
/*
|
| 493 |
* If last ACK falls within this segment's sequence numbers,
|
| 494 |
* record the timestamp.
|
| 495 |
*/
|
| 496 |
/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
|
| 497 |
* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
|
| 498 |
* tp->ts_recent_age = tcp_now;
|
| 499 |
* tp->ts_recent = ts_val;
|
| 500 |
* }
|
| 501 |
*/
|
| 502 |
if (ti->ti_len == 0) { |
| 503 |
if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
|
| 504 |
SEQ_LEQ(ti->ti_ack, tp->snd_max) && |
| 505 |
tp->snd_cwnd >= tp->snd_wnd) {
|
| 506 |
/*
|
| 507 |
* this is a pure ack for outstanding data.
|
| 508 |
*/
|
| 509 |
++tcpstat.tcps_predack; |
| 510 |
/* if (ts_present)
|
| 511 |
* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
|
| 512 |
* else
|
| 513 |
*/ if (tp->t_rtt && |
| 514 |
SEQ_GT(ti->ti_ack, tp->t_rtseq)) |
| 515 |
tcp_xmit_timer(tp, tp->t_rtt); |
| 516 |
acked = ti->ti_ack - tp->snd_una; |
| 517 |
tcpstat.tcps_rcvackpack++; |
| 518 |
tcpstat.tcps_rcvackbyte += acked; |
| 519 |
sbdrop(&so->so_snd, acked); |
| 520 |
tp->snd_una = ti->ti_ack; |
| 521 |
m_freem(m); |
| 522 |
|
| 523 |
/*
|
| 524 |
* If all outstanding data are acked, stop
|
| 525 |
* retransmit timer, otherwise restart timer
|
| 526 |
* using current (possibly backed-off) value.
|
| 527 |
* If process is waiting for space,
|
| 528 |
* wakeup/selwakeup/signal. If data
|
| 529 |
* are ready to send, let tcp_output
|
| 530 |
* decide between more output or persist.
|
| 531 |
*/
|
| 532 |
if (tp->snd_una == tp->snd_max)
|
| 533 |
tp->t_timer[TCPT_REXMT] = 0;
|
| 534 |
else if (tp->t_timer[TCPT_PERSIST] == 0) |
| 535 |
tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; |
| 536 |
|
| 537 |
/*
|
| 538 |
* There's room in so_snd, sowwakup will read()
|
| 539 |
* from the socket if we can
|
| 540 |
*/
|
| 541 |
/* if (so->so_snd.sb_flags & SB_NOTIFY)
|
| 542 |
* sowwakeup(so);
|
| 543 |
*/
|
| 544 |
/*
|
| 545 |
* This is called because sowwakeup might have
|
| 546 |
* put data into so_snd. Since we don't so sowwakeup,
|
| 547 |
* we don't need this.. XXX???
|
| 548 |
*/
|
| 549 |
if (so->so_snd.sb_cc)
|
| 550 |
(void) tcp_output(tp);
|
| 551 |
|
| 552 |
return;
|
| 553 |
} |
| 554 |
} else if (ti->ti_ack == tp->snd_una && |
| 555 |
tp->seg_next == (tcpiphdrp_32)tp && |
| 556 |
ti->ti_len <= sbspace(&so->so_rcv)) {
|
| 557 |
/*
|
| 558 |
* this is a pure, in-sequence data packet
|
| 559 |
* with nothing on the reassembly queue and
|
| 560 |
* we have enough buffer space to take it.
|
| 561 |
*/
|
| 562 |
++tcpstat.tcps_preddat; |
| 563 |
tp->rcv_nxt += ti->ti_len; |
| 564 |
tcpstat.tcps_rcvpack++; |
| 565 |
tcpstat.tcps_rcvbyte += ti->ti_len; |
| 566 |
/*
|
| 567 |
* Add data to socket buffer.
|
| 568 |
*/
|
| 569 |
if (so->so_emu) {
|
| 570 |
if (tcp_emu(so,m)) sbappend(so, m);
|
| 571 |
} else
|
| 572 |
sbappend(so, m); |
| 573 |
|
| 574 |
/*
|
| 575 |
* XXX This is called when data arrives. Later, check
|
| 576 |
* if we can actually write() to the socket
|
| 577 |
* XXX Need to check? It's be NON_BLOCKING
|
| 578 |
*/
|
| 579 |
/* sorwakeup(so); */
|
| 580 |
|
| 581 |
/*
|
| 582 |
* If this is a short packet, then ACK now - with Nagel
|
| 583 |
* congestion avoidance sender won't send more until
|
| 584 |
* he gets an ACK.
|
| 585 |
*
|
| 586 |
* Here are 3 interpretations of what should happen.
|
| 587 |
* The best (for me) is to delay-ack everything except
|
| 588 |
* if it's a one-byte packet containing an ESC
|
| 589 |
* (this means it's an arrow key (or similar) sent using
|
| 590 |
* Nagel, hence there will be no echo)
|
| 591 |
* The first of these is the original, the second is the
|
| 592 |
* middle ground between the other 2
|
| 593 |
*/
|
| 594 |
/* if (((unsigned)ti->ti_len < tp->t_maxseg)) {
|
| 595 |
*/
|
| 596 |
/* if (((unsigned)ti->ti_len < tp->t_maxseg &&
|
| 597 |
* (so->so_iptos & IPTOS_LOWDELAY) == 0) ||
|
| 598 |
* ((so->so_iptos & IPTOS_LOWDELAY) &&
|
| 599 |
* ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
|
| 600 |
*/
|
| 601 |
if ((unsigned)ti->ti_len == 1 && |
| 602 |
((struct tcpiphdr_2 *)ti)->first_char == (char)27) { |
| 603 |
tp->t_flags |= TF_ACKNOW; |
| 604 |
tcp_output(tp); |
| 605 |
} else {
|
| 606 |
tp->t_flags |= TF_DELACK; |
| 607 |
} |
| 608 |
return;
|
| 609 |
} |
| 610 |
} /* header prediction */
|
| 611 |
/*
|
| 612 |
* Calculate amount of space in receive window,
|
| 613 |
* and then do TCP input processing.
|
| 614 |
* Receive window is amount of space in rcv queue,
|
| 615 |
* but not less than advertised window.
|
| 616 |
*/
|
| 617 |
{ int win;
|
| 618 |
win = sbspace(&so->so_rcv); |
| 619 |
if (win < 0) |
| 620 |
win = 0;
|
| 621 |
tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
|
| 622 |
} |
| 623 |
|
| 624 |
switch (tp->t_state) {
|
| 625 |
|
| 626 |
/*
|
| 627 |
* If the state is LISTEN then ignore segment if it contains an RST.
|
| 628 |
* If the segment contains an ACK then it is bad and send a RST.
|
| 629 |
* If it does not contain a SYN then it is not interesting; drop it.
|
| 630 |
* Don't bother responding if the destination was a broadcast.
|
| 631 |
* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
|
| 632 |
* tp->iss, and send a segment:
|
| 633 |
* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
|
| 634 |
* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
|
| 635 |
* Fill in remote peer address fields if not previously specified.
|
| 636 |
* Enter SYN_RECEIVED state, and process any other fields of this
|
| 637 |
* segment in this state.
|
| 638 |
*/
|
| 639 |
case TCPS_LISTEN: {
|
| 640 |
|
| 641 |
if (tiflags & TH_RST)
|
| 642 |
goto drop;
|
| 643 |
if (tiflags & TH_ACK)
|
| 644 |
goto dropwithreset;
|
| 645 |
if ((tiflags & TH_SYN) == 0) |
| 646 |
goto drop;
|
| 647 |
|
| 648 |
/*
|
| 649 |
* This has way too many gotos...
|
| 650 |
* But a bit of spaghetti code never hurt anybody :)
|
| 651 |
*/
|
| 652 |
|
| 653 |
/*
|
| 654 |
* If this is destined for the control address, then flag to
|
| 655 |
* tcp_ctl once connected, otherwise connect
|
| 656 |
*/
|
| 657 |
if ((so->so_faddr.s_addr&htonl(0xffffff00)) == special_addr.s_addr) { |
| 658 |
int lastbyte=ntohl(so->so_faddr.s_addr) & 0xff; |
| 659 |
if (lastbyte!=CTL_ALIAS && lastbyte!=CTL_DNS) {
|
| 660 |
#if 0
|
| 661 |
if(lastbyte==CTL_CMD || lastbyte==CTL_EXEC) {
|
| 662 |
/* Command or exec adress */
|
| 663 |
so->so_state |= SS_CTL;
|
| 664 |
} else {
|
| 665 |
/* May be an add exec */
|
| 666 |
struct ex_list *ex_ptr;
|
| 667 |
|
| 668 |
for(ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
|
| 669 |
if(ex_ptr->ex_fport == so->so_fport &&
|
| 670 |
lastbyte == ex_ptr->ex_addr) {
|
| 671 |
so->so_state |= SS_CTL;
|
| 672 |
break;
|
| 673 |
}
|
| 674 |
}
|
| 675 |
}
|
| 676 |
if(so->so_state & SS_CTL) goto cont_input;
|
| 677 |
#endif
|
| 678 |
} |
| 679 |
/* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
|
| 680 |
} |
| 681 |
|
| 682 |
if (so->so_emu & EMU_NOCONNECT) {
|
| 683 |
so->so_emu &= ~EMU_NOCONNECT; |
| 684 |
goto cont_input;
|
| 685 |
} |
| 686 |
|
| 687 |
if(tcp_fconnect(so) == -1 && errno != EINPROGRESS) { |
| 688 |
u_char code=ICMP_UNREACH_NET; |
| 689 |
DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
|
| 690 |
errno,strerror(errno))); |
| 691 |
if(errno == ECONNREFUSED) {
|
| 692 |
/* ACK the SYN, send RST to refuse the connection */
|
| 693 |
tcp_respond(tp, ti, m, ti->ti_seq+1, (tcp_seq)0, |
| 694 |
TH_RST|TH_ACK); |
| 695 |
} else {
|
| 696 |
if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
|
| 697 |
HTONL(ti->ti_seq); /* restore tcp header */
|
| 698 |
HTONL(ti->ti_ack); |
| 699 |
HTONS(ti->ti_win); |
| 700 |
HTONS(ti->ti_urp); |
| 701 |
m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr); |
| 702 |
m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr); |
| 703 |
*ip=save_ip; |
| 704 |
icmp_error(m, ICMP_UNREACH,code, 0,strerror(errno));
|
| 705 |
} |
| 706 |
tp = tcp_close(tp); |
| 707 |
m_free(m); |
| 708 |
} else {
|
| 709 |
/*
|
| 710 |
* Haven't connected yet, save the current mbuf
|
| 711 |
* and ti, and return
|
| 712 |
* XXX Some OS's don't tell us whether the connect()
|
| 713 |
* succeeded or not. So we must time it out.
|
| 714 |
*/
|
| 715 |
so->so_m = m; |
| 716 |
so->so_ti = ti; |
| 717 |
tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT; |
| 718 |
tp->t_state = TCPS_SYN_RECEIVED; |
| 719 |
} |
| 720 |
return;
|
| 721 |
|
| 722 |
cont_conn:
|
| 723 |
/* m==NULL
|
| 724 |
* Check if the connect succeeded
|
| 725 |
*/
|
| 726 |
if (so->so_state & SS_NOFDREF) {
|
| 727 |
tp = tcp_close(tp); |
| 728 |
goto dropwithreset;
|
| 729 |
} |
| 730 |
cont_input:
|
| 731 |
tcp_template(tp); |
| 732 |
|
| 733 |
if (optp)
|
| 734 |
tcp_dooptions(tp, (u_char *)optp, optlen, ti); |
| 735 |
/* , */
|
| 736 |
/* &ts_present, &ts_val, &ts_ecr); */
|
| 737 |
|
| 738 |
if (iss)
|
| 739 |
tp->iss = iss; |
| 740 |
else
|
| 741 |
tp->iss = tcp_iss; |
| 742 |
tcp_iss += TCP_ISSINCR/2;
|
| 743 |
tp->irs = ti->ti_seq; |
| 744 |
tcp_sendseqinit(tp); |
| 745 |
tcp_rcvseqinit(tp); |
| 746 |
tp->t_flags |= TF_ACKNOW; |
| 747 |
tp->t_state = TCPS_SYN_RECEIVED; |
| 748 |
tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT; |
| 749 |
tcpstat.tcps_accepts++; |
| 750 |
goto trimthenstep6;
|
| 751 |
} /* case TCPS_LISTEN */
|
| 752 |
|
| 753 |
/*
|
| 754 |
* If the state is SYN_SENT:
|
| 755 |
* if seg contains an ACK, but not for our SYN, drop the input.
|
| 756 |
* if seg contains a RST, then drop the connection.
|
| 757 |
* if seg does not contain SYN, then drop it.
|
| 758 |
* Otherwise this is an acceptable SYN segment
|
| 759 |
* initialize tp->rcv_nxt and tp->irs
|
| 760 |
* if seg contains ack then advance tp->snd_una
|
| 761 |
* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
|
| 762 |
* arrange for segment to be acked (eventually)
|
| 763 |
* continue processing rest of data/controls, beginning with URG
|
| 764 |
*/
|
| 765 |
case TCPS_SYN_SENT:
|
| 766 |
if ((tiflags & TH_ACK) &&
|
| 767 |
(SEQ_LEQ(ti->ti_ack, tp->iss) || |
| 768 |
SEQ_GT(ti->ti_ack, tp->snd_max))) |
| 769 |
goto dropwithreset;
|
| 770 |
|
| 771 |
if (tiflags & TH_RST) {
|
| 772 |
if (tiflags & TH_ACK)
|
| 773 |
tp = tcp_drop(tp,0); /* XXX Check t_softerror! */ |
| 774 |
goto drop;
|
| 775 |
} |
| 776 |
|
| 777 |
if ((tiflags & TH_SYN) == 0) |
| 778 |
goto drop;
|
| 779 |
if (tiflags & TH_ACK) {
|
| 780 |
tp->snd_una = ti->ti_ack; |
| 781 |
if (SEQ_LT(tp->snd_nxt, tp->snd_una))
|
| 782 |
tp->snd_nxt = tp->snd_una; |
| 783 |
} |
| 784 |
|
| 785 |
tp->t_timer[TCPT_REXMT] = 0;
|
| 786 |
tp->irs = ti->ti_seq; |
| 787 |
tcp_rcvseqinit(tp); |
| 788 |
tp->t_flags |= TF_ACKNOW; |
| 789 |
if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
|
| 790 |
tcpstat.tcps_connects++; |
| 791 |
soisfconnected(so); |
| 792 |
tp->t_state = TCPS_ESTABLISHED; |
| 793 |
|
| 794 |
/* Do window scaling on this connection? */
|
| 795 |
/* if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
|
| 796 |
* (TF_RCVD_SCALE|TF_REQ_SCALE)) {
|
| 797 |
* tp->snd_scale = tp->requested_s_scale;
|
| 798 |
* tp->rcv_scale = tp->request_r_scale;
|
| 799 |
* }
|
| 800 |
*/
|
| 801 |
(void) tcp_reass(tp, (struct tcpiphdr *)0, |
| 802 |
(struct mbuf *)0); |
| 803 |
/*
|
| 804 |
* if we didn't have to retransmit the SYN,
|
| 805 |
* use its rtt as our initial srtt & rtt var.
|
| 806 |
*/
|
| 807 |
if (tp->t_rtt)
|
| 808 |
tcp_xmit_timer(tp, tp->t_rtt); |
| 809 |
} else
|
| 810 |
tp->t_state = TCPS_SYN_RECEIVED; |
| 811 |
|
| 812 |
trimthenstep6:
|
| 813 |
/*
|
| 814 |
* Advance ti->ti_seq to correspond to first data byte.
|
| 815 |
* If data, trim to stay within window,
|
| 816 |
* dropping FIN if necessary.
|
| 817 |
*/
|
| 818 |
ti->ti_seq++; |
| 819 |
if (ti->ti_len > tp->rcv_wnd) {
|
| 820 |
todrop = ti->ti_len - tp->rcv_wnd; |
| 821 |
m_adj(m, -todrop); |
| 822 |
ti->ti_len = tp->rcv_wnd; |
| 823 |
tiflags &= ~TH_FIN; |
| 824 |
tcpstat.tcps_rcvpackafterwin++; |
| 825 |
tcpstat.tcps_rcvbyteafterwin += todrop; |
| 826 |
} |
| 827 |
tp->snd_wl1 = ti->ti_seq - 1;
|
| 828 |
tp->rcv_up = ti->ti_seq; |
| 829 |
goto step6;
|
| 830 |
} /* switch tp->t_state */
|
| 831 |
/*
|
| 832 |
* States other than LISTEN or SYN_SENT.
|
| 833 |
* First check timestamp, if present.
|
| 834 |
* Then check that at least some bytes of segment are within
|
| 835 |
* receive window. If segment begins before rcv_nxt,
|
| 836 |
* drop leading data (and SYN); if nothing left, just ack.
|
| 837 |
*
|
| 838 |
* RFC 1323 PAWS: If we have a timestamp reply on this segment
|
| 839 |
* and it's less than ts_recent, drop it.
|
| 840 |
*/
|
| 841 |
/* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
|
| 842 |
* TSTMP_LT(ts_val, tp->ts_recent)) {
|
| 843 |
*
|
| 844 |
*/ /* Check to see if ts_recent is over 24 days old. */ |
| 845 |
/* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
|
| 846 |
*/ /* |
| 847 |
* * Invalidate ts_recent. If this segment updates
|
| 848 |
* * ts_recent, the age will be reset later and ts_recent
|
| 849 |
* * will get a valid value. If it does not, setting
|
| 850 |
* * ts_recent to zero will at least satisfy the
|
| 851 |
* * requirement that zero be placed in the timestamp
|
| 852 |
* * echo reply when ts_recent isn't valid. The
|
| 853 |
* * age isn't reset until we get a valid ts_recent
|
| 854 |
* * because we don't want out-of-order segments to be
|
| 855 |
* * dropped when ts_recent is old.
|
| 856 |
* */
|
| 857 |
/* tp->ts_recent = 0;
|
| 858 |
* } else {
|
| 859 |
* tcpstat.tcps_rcvduppack++;
|
| 860 |
* tcpstat.tcps_rcvdupbyte += ti->ti_len;
|
| 861 |
* tcpstat.tcps_pawsdrop++;
|
| 862 |
* goto dropafterack;
|
| 863 |
* }
|
| 864 |
* }
|
| 865 |
*/
|
| 866 |
|
| 867 |
todrop = tp->rcv_nxt - ti->ti_seq; |
| 868 |
if (todrop > 0) { |
| 869 |
if (tiflags & TH_SYN) {
|
| 870 |
tiflags &= ~TH_SYN; |
| 871 |
ti->ti_seq++; |
| 872 |
if (ti->ti_urp > 1) |
| 873 |
ti->ti_urp--; |
| 874 |
else
|
| 875 |
tiflags &= ~TH_URG; |
| 876 |
todrop--; |
| 877 |
} |
| 878 |
/*
|
| 879 |
* Following if statement from Stevens, vol. 2, p. 960.
|
| 880 |
*/
|
| 881 |
if (todrop > ti->ti_len
|
| 882 |
|| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
|
| 883 |
/*
|
| 884 |
* Any valid FIN must be to the left of the window.
|
| 885 |
* At this point the FIN must be a duplicate or out
|
| 886 |
* of sequence; drop it.
|
| 887 |
*/
|
| 888 |
tiflags &= ~TH_FIN; |
| 889 |
|
| 890 |
/*
|
| 891 |
* Send an ACK to resynchronize and drop any data.
|
| 892 |
* But keep on processing for RST or ACK.
|
| 893 |
*/
|
| 894 |
tp->t_flags |= TF_ACKNOW; |
| 895 |
todrop = ti->ti_len; |
| 896 |
tcpstat.tcps_rcvduppack++; |
| 897 |
tcpstat.tcps_rcvdupbyte += todrop; |
| 898 |
} else {
|
| 899 |
tcpstat.tcps_rcvpartduppack++; |
| 900 |
tcpstat.tcps_rcvpartdupbyte += todrop; |
| 901 |
} |
| 902 |
m_adj(m, todrop); |
| 903 |
ti->ti_seq += todrop; |
| 904 |
ti->ti_len -= todrop; |
| 905 |
if (ti->ti_urp > todrop)
|
| 906 |
ti->ti_urp -= todrop; |
| 907 |
else {
|
| 908 |
tiflags &= ~TH_URG; |
| 909 |
ti->ti_urp = 0;
|
| 910 |
} |
| 911 |
} |
| 912 |
/*
|
| 913 |
* If new data are received on a connection after the
|
| 914 |
* user processes are gone, then RST the other end.
|
| 915 |
*/
|
| 916 |
if ((so->so_state & SS_NOFDREF) &&
|
| 917 |
tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
|
| 918 |
tp = tcp_close(tp); |
| 919 |
tcpstat.tcps_rcvafterclose++; |
| 920 |
goto dropwithreset;
|
| 921 |
} |
| 922 |
|
| 923 |
/*
|
| 924 |
* If segment ends after window, drop trailing data
|
| 925 |
* (and PUSH and FIN); if nothing left, just ACK.
|
| 926 |
*/
|
| 927 |
todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd); |
| 928 |
if (todrop > 0) { |
| 929 |
tcpstat.tcps_rcvpackafterwin++; |
| 930 |
if (todrop >= ti->ti_len) {
|
| 931 |
tcpstat.tcps_rcvbyteafterwin += ti->ti_len; |
| 932 |
/*
|
| 933 |
* If a new connection request is received
|
| 934 |
* while in TIME_WAIT, drop the old connection
|
| 935 |
* and start over if the sequence numbers
|
| 936 |
* are above the previous ones.
|
| 937 |
*/
|
| 938 |
if (tiflags & TH_SYN &&
|
| 939 |
tp->t_state == TCPS_TIME_WAIT && |
| 940 |
SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
|
| 941 |
iss = tp->rcv_nxt + TCP_ISSINCR; |
| 942 |
tp = tcp_close(tp); |
| 943 |
goto findso;
|
| 944 |
} |
| 945 |
/*
|
| 946 |
* If window is closed can only take segments at
|
| 947 |
* window edge, and have to drop data and PUSH from
|
| 948 |
* incoming segments. Continue processing, but
|
| 949 |
* remember to ack. Otherwise, drop segment
|
| 950 |
* and ack.
|
| 951 |
*/
|
| 952 |
if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) { |
| 953 |
tp->t_flags |= TF_ACKNOW; |
| 954 |
tcpstat.tcps_rcvwinprobe++; |
| 955 |
} else
|
| 956 |
goto dropafterack;
|
| 957 |
} else
|
| 958 |
tcpstat.tcps_rcvbyteafterwin += todrop; |
| 959 |
m_adj(m, -todrop); |
| 960 |
ti->ti_len -= todrop; |
| 961 |
tiflags &= ~(TH_PUSH|TH_FIN); |
| 962 |
} |
| 963 |
|
| 964 |
/*
|
| 965 |
* If last ACK falls within this segment's sequence numbers,
|
| 966 |
* record its timestamp.
|
| 967 |
*/
|
| 968 |
/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
|
| 969 |
* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
|
| 970 |
* ((tiflags & (TH_SYN|TH_FIN)) != 0))) {
|
| 971 |
* tp->ts_recent_age = tcp_now;
|
| 972 |
* tp->ts_recent = ts_val;
|
| 973 |
* }
|
| 974 |
*/
|
| 975 |
|
| 976 |
/*
|
| 977 |
* If the RST bit is set examine the state:
|
| 978 |
* SYN_RECEIVED STATE:
|
| 979 |
* If passive open, return to LISTEN state.
|
| 980 |
* If active open, inform user that connection was refused.
|
| 981 |
* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
|
| 982 |
* Inform user that connection was reset, and close tcb.
|
| 983 |
* CLOSING, LAST_ACK, TIME_WAIT STATES
|
| 984 |
* Close the tcb.
|
| 985 |
*/
|
| 986 |
if (tiflags&TH_RST) switch (tp->t_state) { |
| 987 |
|
| 988 |
case TCPS_SYN_RECEIVED:
|
| 989 |
/* so->so_error = ECONNREFUSED; */
|
| 990 |
goto close;
|
| 991 |
|
| 992 |
case TCPS_ESTABLISHED:
|
| 993 |
case TCPS_FIN_WAIT_1:
|
| 994 |
case TCPS_FIN_WAIT_2:
|
| 995 |
case TCPS_CLOSE_WAIT:
|
| 996 |
/* so->so_error = ECONNRESET; */
|
| 997 |
close:
|
| 998 |
tp->t_state = TCPS_CLOSED; |
| 999 |
tcpstat.tcps_drops++; |
| 1000 |
tp = tcp_close(tp); |
| 1001 |
goto drop;
|
| 1002 |
|
| 1003 |
case TCPS_CLOSING:
|
| 1004 |
case TCPS_LAST_ACK:
|
| 1005 |
case TCPS_TIME_WAIT:
|
| 1006 |
tp = tcp_close(tp); |
| 1007 |
goto drop;
|
| 1008 |
} |
| 1009 |
|
| 1010 |
/*
|
| 1011 |
* If a SYN is in the window, then this is an
|
| 1012 |
* error and we send an RST and drop the connection.
|
| 1013 |
*/
|
| 1014 |
if (tiflags & TH_SYN) {
|
| 1015 |
tp = tcp_drop(tp,0);
|
| 1016 |
goto dropwithreset;
|
| 1017 |
} |
| 1018 |
|
| 1019 |
/*
|
| 1020 |
* If the ACK bit is off we drop the segment and return.
|
| 1021 |
*/
|
| 1022 |
if ((tiflags & TH_ACK) == 0) goto drop; |
| 1023 |
|
| 1024 |
/*
|
| 1025 |
* Ack processing.
|
| 1026 |
*/
|
| 1027 |
switch (tp->t_state) {
|
| 1028 |
/*
|
| 1029 |
* In SYN_RECEIVED state if the ack ACKs our SYN then enter
|
| 1030 |
* ESTABLISHED state and continue processing, otherwise
|
| 1031 |
* send an RST. una<=ack<=max
|
| 1032 |
*/
|
| 1033 |
case TCPS_SYN_RECEIVED:
|
| 1034 |
|
| 1035 |
if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
|
| 1036 |
SEQ_GT(ti->ti_ack, tp->snd_max)) |
| 1037 |
goto dropwithreset;
|
| 1038 |
tcpstat.tcps_connects++; |
| 1039 |
tp->t_state = TCPS_ESTABLISHED; |
| 1040 |
/*
|
| 1041 |
* The sent SYN is ack'ed with our sequence number +1
|
| 1042 |
* The first data byte already in the buffer will get
|
| 1043 |
* lost if no correction is made. This is only needed for
|
| 1044 |
* SS_CTL since the buffer is empty otherwise.
|
| 1045 |
* tp->snd_una++; or:
|
| 1046 |
*/
|
| 1047 |
tp->snd_una=ti->ti_ack; |
| 1048 |
if (so->so_state & SS_CTL) {
|
| 1049 |
/* So tcp_ctl reports the right state */
|
| 1050 |
ret = tcp_ctl(so); |
| 1051 |
if (ret == 1) { |
| 1052 |
soisfconnected(so); |
| 1053 |
so->so_state &= ~SS_CTL; /* success XXX */
|
| 1054 |
} else if (ret == 2) { |
| 1055 |
so->so_state = SS_NOFDREF; /* CTL_CMD */
|
| 1056 |
} else {
|
| 1057 |
needoutput = 1;
|
| 1058 |
tp->t_state = TCPS_FIN_WAIT_1; |
| 1059 |
} |
| 1060 |
} else {
|
| 1061 |
soisfconnected(so); |
| 1062 |
} |
| 1063 |
|
| 1064 |
/* Do window scaling? */
|
| 1065 |
/* if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
|
| 1066 |
* (TF_RCVD_SCALE|TF_REQ_SCALE)) {
|
| 1067 |
* tp->snd_scale = tp->requested_s_scale;
|
| 1068 |
* tp->rcv_scale = tp->request_r_scale;
|
| 1069 |
* }
|
| 1070 |
*/
|
| 1071 |
(void) tcp_reass(tp, (struct tcpiphdr *)0, (struct mbuf *)0); |
| 1072 |
tp->snd_wl1 = ti->ti_seq - 1;
|
| 1073 |
/* Avoid ack processing; snd_una==ti_ack => dup ack */
|
| 1074 |
goto synrx_to_est;
|
| 1075 |
/* fall into ... */
|
| 1076 |
|
| 1077 |
/*
|
| 1078 |
* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
|
| 1079 |
* ACKs. If the ack is in the range
|
| 1080 |
* tp->snd_una < ti->ti_ack <= tp->snd_max
|
| 1081 |
* then advance tp->snd_una to ti->ti_ack and drop
|
| 1082 |
* data from the retransmission queue. If this ACK reflects
|
| 1083 |
* more up to date window information we update our window information.
|
| 1084 |
*/
|
| 1085 |
case TCPS_ESTABLISHED:
|
| 1086 |
case TCPS_FIN_WAIT_1:
|
| 1087 |
case TCPS_FIN_WAIT_2:
|
| 1088 |
case TCPS_CLOSE_WAIT:
|
| 1089 |
case TCPS_CLOSING:
|
| 1090 |
case TCPS_LAST_ACK:
|
| 1091 |
case TCPS_TIME_WAIT:
|
| 1092 |
|
| 1093 |
if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
|
| 1094 |
if (ti->ti_len == 0 && tiwin == tp->snd_wnd) { |
| 1095 |
tcpstat.tcps_rcvdupack++; |
| 1096 |
DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
|
| 1097 |
(long )m, (long )so)); |
| 1098 |
/*
|
| 1099 |
* If we have outstanding data (other than
|
| 1100 |
* a window probe), this is a completely
|
| 1101 |
* duplicate ack (ie, window info didn't
|
| 1102 |
* change), the ack is the biggest we've
|
| 1103 |
* seen and we've seen exactly our rexmt
|
| 1104 |
* threshold of them, assume a packet
|
| 1105 |
* has been dropped and retransmit it.
|
| 1106 |
* Kludge snd_nxt & the congestion
|
| 1107 |
* window so we send only this one
|
| 1108 |
* packet.
|
| 1109 |
*
|
| 1110 |
* We know we're losing at the current
|
| 1111 |
* window size so do congestion avoidance
|
| 1112 |
* (set ssthresh to half the current window
|
| 1113 |
* and pull our congestion window back to
|
| 1114 |
* the new ssthresh).
|
| 1115 |
*
|
| 1116 |
* Dup acks mean that packets have left the
|
| 1117 |
* network (they're now cached at the receiver)
|
| 1118 |
* so bump cwnd by the amount in the receiver
|
| 1119 |
* to keep a constant cwnd packets in the
|
| 1120 |
* network.
|
| 1121 |
*/
|
| 1122 |
if (tp->t_timer[TCPT_REXMT] == 0 || |
| 1123 |
ti->ti_ack != tp->snd_una) |
| 1124 |
tp->t_dupacks = 0;
|
| 1125 |
else if (++tp->t_dupacks == tcprexmtthresh) { |
| 1126 |
tcp_seq onxt = tp->snd_nxt; |
| 1127 |
u_int win = |
| 1128 |
min(tp->snd_wnd, tp->snd_cwnd) / 2 /
|
| 1129 |
tp->t_maxseg; |
| 1130 |
|
| 1131 |
if (win < 2) |
| 1132 |
win = 2;
|
| 1133 |
tp->snd_ssthresh = win * tp->t_maxseg; |
| 1134 |
tp->t_timer[TCPT_REXMT] = 0;
|
| 1135 |
tp->t_rtt = 0;
|
| 1136 |
tp->snd_nxt = ti->ti_ack; |
| 1137 |
tp->snd_cwnd = tp->t_maxseg; |
| 1138 |
(void) tcp_output(tp);
|
| 1139 |
tp->snd_cwnd = tp->snd_ssthresh + |
| 1140 |
tp->t_maxseg * tp->t_dupacks; |
| 1141 |
if (SEQ_GT(onxt, tp->snd_nxt))
|
| 1142 |
tp->snd_nxt = onxt; |
| 1143 |
goto drop;
|
| 1144 |
} else if (tp->t_dupacks > tcprexmtthresh) { |
| 1145 |
tp->snd_cwnd += tp->t_maxseg; |
| 1146 |
(void) tcp_output(tp);
|
| 1147 |
goto drop;
|
| 1148 |
} |
| 1149 |
} else
|
| 1150 |
tp->t_dupacks = 0;
|
| 1151 |
break;
|
| 1152 |
} |
| 1153 |
synrx_to_est:
|
| 1154 |
/*
|
| 1155 |
* If the congestion window was inflated to account
|
| 1156 |
* for the other side's cached packets, retract it.
|
| 1157 |
*/
|
| 1158 |
if (tp->t_dupacks > tcprexmtthresh &&
|
| 1159 |
tp->snd_cwnd > tp->snd_ssthresh) |
| 1160 |
tp->snd_cwnd = tp->snd_ssthresh; |
| 1161 |
tp->t_dupacks = 0;
|
| 1162 |
if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
|
| 1163 |
tcpstat.tcps_rcvacktoomuch++; |
| 1164 |
goto dropafterack;
|
| 1165 |
} |
| 1166 |
acked = ti->ti_ack - tp->snd_una; |
| 1167 |
tcpstat.tcps_rcvackpack++; |
| 1168 |
tcpstat.tcps_rcvackbyte += acked; |
| 1169 |
|
| 1170 |
/*
|
| 1171 |
* If we have a timestamp reply, update smoothed
|
| 1172 |
* round trip time. If no timestamp is present but
|
| 1173 |
* transmit timer is running and timed sequence
|
| 1174 |
* number was acked, update smoothed round trip time.
|
| 1175 |
* Since we now have an rtt measurement, cancel the
|
| 1176 |
* timer backoff (cf., Phil Karn's retransmit alg.).
|
| 1177 |
* Recompute the initial retransmit timer.
|
| 1178 |
*/
|
| 1179 |
/* if (ts_present)
|
| 1180 |
* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
|
| 1181 |
* else
|
| 1182 |
*/
|
| 1183 |
if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
|
| 1184 |
tcp_xmit_timer(tp,tp->t_rtt); |
| 1185 |
|
| 1186 |
/*
|
| 1187 |
* If all outstanding data is acked, stop retransmit
|
| 1188 |
* timer and remember to restart (more output or persist).
|
| 1189 |
* If there is more data to be acked, restart retransmit
|
| 1190 |
* timer, using current (possibly backed-off) value.
|
| 1191 |
*/
|
| 1192 |
if (ti->ti_ack == tp->snd_max) {
|
| 1193 |
tp->t_timer[TCPT_REXMT] = 0;
|
| 1194 |
needoutput = 1;
|
| 1195 |
} else if (tp->t_timer[TCPT_PERSIST] == 0) |
| 1196 |
tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; |
| 1197 |
/*
|
| 1198 |
* When new data is acked, open the congestion window.
|
| 1199 |
* If the window gives us less than ssthresh packets
|
| 1200 |
* in flight, open exponentially (maxseg per packet).
|
| 1201 |
* Otherwise open linearly: maxseg per window
|
| 1202 |
* (maxseg^2 / cwnd per packet).
|
| 1203 |
*/
|
| 1204 |
{
|
| 1205 |
register u_int cw = tp->snd_cwnd;
|
| 1206 |
register u_int incr = tp->t_maxseg;
|
| 1207 |
|
| 1208 |
if (cw > tp->snd_ssthresh)
|
| 1209 |
incr = incr * incr / cw; |
| 1210 |
tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale); |
| 1211 |
} |
| 1212 |
if (acked > so->so_snd.sb_cc) {
|
| 1213 |
tp->snd_wnd -= so->so_snd.sb_cc; |
| 1214 |
sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
|
| 1215 |
ourfinisacked = 1;
|
| 1216 |
} else {
|
| 1217 |
sbdrop(&so->so_snd, acked); |
| 1218 |
tp->snd_wnd -= acked; |
| 1219 |
ourfinisacked = 0;
|
| 1220 |
} |
| 1221 |
/*
|
| 1222 |
* XXX sowwakup is called when data is acked and there's room for
|
| 1223 |
* for more data... it should read() the socket
|
| 1224 |
*/
|
| 1225 |
/* if (so->so_snd.sb_flags & SB_NOTIFY)
|
| 1226 |
* sowwakeup(so);
|
| 1227 |
*/
|
| 1228 |
tp->snd_una = ti->ti_ack; |
| 1229 |
if (SEQ_LT(tp->snd_nxt, tp->snd_una))
|
| 1230 |
tp->snd_nxt = tp->snd_una; |
| 1231 |
|
| 1232 |
switch (tp->t_state) {
|
| 1233 |
|
| 1234 |
/*
|
| 1235 |
* In FIN_WAIT_1 STATE in addition to the processing
|
| 1236 |
* for the ESTABLISHED state if our FIN is now acknowledged
|
| 1237 |
* then enter FIN_WAIT_2.
|
| 1238 |
*/
|
| 1239 |
case TCPS_FIN_WAIT_1:
|
| 1240 |
if (ourfinisacked) {
|
| 1241 |
/*
|
| 1242 |
* If we can't receive any more
|
| 1243 |
* data, then closing user can proceed.
|
| 1244 |
* Starting the timer is contrary to the
|
| 1245 |
* specification, but if we don't get a FIN
|
| 1246 |
* we'll hang forever.
|
| 1247 |
*/
|
| 1248 |
if (so->so_state & SS_FCANTRCVMORE) {
|
| 1249 |
soisfdisconnected(so); |
| 1250 |
tp->t_timer[TCPT_2MSL] = tcp_maxidle; |
| 1251 |
} |
| 1252 |
tp->t_state = TCPS_FIN_WAIT_2; |
| 1253 |
} |
| 1254 |
break;
|
| 1255 |
|
| 1256 |
/*
|
| 1257 |
* In CLOSING STATE in addition to the processing for
|
| 1258 |
* the ESTABLISHED state if the ACK acknowledges our FIN
|
| 1259 |
* then enter the TIME-WAIT state, otherwise ignore
|
| 1260 |
* the segment.
|
| 1261 |
*/
|
| 1262 |
case TCPS_CLOSING:
|
| 1263 |
if (ourfinisacked) {
|
| 1264 |
tp->t_state = TCPS_TIME_WAIT; |
| 1265 |
tcp_canceltimers(tp); |
| 1266 |
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
|
| 1267 |
soisfdisconnected(so); |
| 1268 |
} |
| 1269 |
break;
|
| 1270 |
|
| 1271 |
/*
|
| 1272 |
* In LAST_ACK, we may still be waiting for data to drain
|
| 1273 |
* and/or to be acked, as well as for the ack of our FIN.
|
| 1274 |
* If our FIN is now acknowledged, delete the TCB,
|
| 1275 |
* enter the closed state and return.
|
| 1276 |
*/
|
| 1277 |
case TCPS_LAST_ACK:
|
| 1278 |
if (ourfinisacked) {
|
| 1279 |
tp = tcp_close(tp); |
| 1280 |
goto drop;
|
| 1281 |
} |
| 1282 |
break;
|
| 1283 |
|
| 1284 |
/*
|
| 1285 |
* In TIME_WAIT state the only thing that should arrive
|
| 1286 |
* is a retransmission of the remote FIN. Acknowledge
|
| 1287 |
* it and restart the finack timer.
|
| 1288 |
*/
|
| 1289 |
case TCPS_TIME_WAIT:
|
| 1290 |
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
|
| 1291 |
goto dropafterack;
|
| 1292 |
} |
| 1293 |
} /* switch(tp->t_state) */
|
| 1294 |
|
| 1295 |
step6:
|
| 1296 |
/*
|
| 1297 |
* Update window information.
|
| 1298 |
* Don't look at window if no ACK: TAC's send garbage on first SYN.
|
| 1299 |
*/
|
| 1300 |
if ((tiflags & TH_ACK) &&
|
| 1301 |
(SEQ_LT(tp->snd_wl1, ti->ti_seq) || |
| 1302 |
(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) || |
| 1303 |
(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
|
| 1304 |
/* keep track of pure window updates */
|
| 1305 |
if (ti->ti_len == 0 && |
| 1306 |
tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd) |
| 1307 |
tcpstat.tcps_rcvwinupd++; |
| 1308 |
tp->snd_wnd = tiwin; |
| 1309 |
tp->snd_wl1 = ti->ti_seq; |
| 1310 |
tp->snd_wl2 = ti->ti_ack; |
| 1311 |
if (tp->snd_wnd > tp->max_sndwnd)
|
| 1312 |
tp->max_sndwnd = tp->snd_wnd; |
| 1313 |
needoutput = 1;
|
| 1314 |
} |
| 1315 |
|
| 1316 |
/*
|
| 1317 |
* Process segments with URG.
|
| 1318 |
*/
|
| 1319 |
if ((tiflags & TH_URG) && ti->ti_urp &&
|
| 1320 |
TCPS_HAVERCVDFIN(tp->t_state) == 0) {
|
| 1321 |
/*
|
| 1322 |
* This is a kludge, but if we receive and accept
|
| 1323 |
* random urgent pointers, we'll crash in
|
| 1324 |
* soreceive. It's hard to imagine someone
|
| 1325 |
* actually wanting to send this much urgent data.
|
| 1326 |
*/
|
| 1327 |
if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
|
| 1328 |
ti->ti_urp = 0;
|
| 1329 |
tiflags &= ~TH_URG; |
| 1330 |
goto dodata;
|
| 1331 |
} |
| 1332 |
/*
|
| 1333 |
* If this segment advances the known urgent pointer,
|
| 1334 |
* then mark the data stream. This should not happen
|
| 1335 |
* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
|
| 1336 |
* a FIN has been received from the remote side.
|
| 1337 |
* In these states we ignore the URG.
|
| 1338 |
*
|
| 1339 |
* According to RFC961 (Assigned Protocols),
|
| 1340 |
* the urgent pointer points to the last octet
|
| 1341 |
* of urgent data. We continue, however,
|
| 1342 |
* to consider it to indicate the first octet
|
| 1343 |
* of data past the urgent section as the original
|
| 1344 |
* spec states (in one of two places).
|
| 1345 |
*/
|
| 1346 |
if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
|
| 1347 |
tp->rcv_up = ti->ti_seq + ti->ti_urp; |
| 1348 |
so->so_urgc = so->so_rcv.sb_cc + |
| 1349 |
(tp->rcv_up - tp->rcv_nxt); /* -1; */
|
| 1350 |
tp->rcv_up = ti->ti_seq + ti->ti_urp; |
| 1351 |
|
| 1352 |
} |
| 1353 |
} else
|
| 1354 |
/*
|
| 1355 |
* If no out of band data is expected,
|
| 1356 |
* pull receive urgent pointer along
|
| 1357 |
* with the receive window.
|
| 1358 |
*/
|
| 1359 |
if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
|
| 1360 |
tp->rcv_up = tp->rcv_nxt; |
| 1361 |
dodata:
|
| 1362 |
|
| 1363 |
/*
|
| 1364 |
* Process the segment text, merging it into the TCP sequencing queue,
|
| 1365 |
* and arranging for acknowledgment of receipt if necessary.
|
| 1366 |
* This process logically involves adjusting tp->rcv_wnd as data
|
| 1367 |
* is presented to the user (this happens in tcp_usrreq.c,
|
| 1368 |
* case PRU_RCVD). If a FIN has already been received on this
|
| 1369 |
* connection then we just ignore the text.
|
| 1370 |
*/
|
| 1371 |
if ((ti->ti_len || (tiflags&TH_FIN)) &&
|
| 1372 |
TCPS_HAVERCVDFIN(tp->t_state) == 0) {
|
| 1373 |
TCP_REASS(tp, ti, m, so, tiflags); |
| 1374 |
/*
|
| 1375 |
* Note the amount of data that peer has sent into
|
| 1376 |
* our window, in order to estimate the sender's
|
| 1377 |
* buffer size.
|
| 1378 |
*/
|
| 1379 |
len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt); |
| 1380 |
} else {
|
| 1381 |
m_free(m); |
| 1382 |
tiflags &= ~TH_FIN; |
| 1383 |
} |
| 1384 |
|
| 1385 |
/*
|
| 1386 |
* If FIN is received ACK the FIN and let the user know
|
| 1387 |
* that the connection is closing.
|
| 1388 |
*/
|
| 1389 |
if (tiflags & TH_FIN) {
|
| 1390 |
if (TCPS_HAVERCVDFIN(tp->t_state) == 0) { |
| 1391 |
/*
|
| 1392 |
* If we receive a FIN we can't send more data,
|
| 1393 |
* set it SS_FDRAIN
|
| 1394 |
* Shutdown the socket if there is no rx data in the
|
| 1395 |
* buffer.
|
| 1396 |
* soread() is called on completion of shutdown() and
|
| 1397 |
* will got to TCPS_LAST_ACK, and use tcp_output()
|
| 1398 |
* to send the FIN.
|
| 1399 |
*/
|
| 1400 |
/* sofcantrcvmore(so); */
|
| 1401 |
sofwdrain(so); |
| 1402 |
|
| 1403 |
tp->t_flags |= TF_ACKNOW; |
| 1404 |
tp->rcv_nxt++; |
| 1405 |
} |
| 1406 |
switch (tp->t_state) {
|
| 1407 |
|
| 1408 |
/*
|
| 1409 |
* In SYN_RECEIVED and ESTABLISHED STATES
|
| 1410 |
* enter the CLOSE_WAIT state.
|
| 1411 |
*/
|
| 1412 |
case TCPS_SYN_RECEIVED:
|
| 1413 |
case TCPS_ESTABLISHED:
|
| 1414 |
if(so->so_emu == EMU_CTL) /* no shutdown on socket */ |
| 1415 |
tp->t_state = TCPS_LAST_ACK; |
| 1416 |
else
|
| 1417 |
tp->t_state = TCPS_CLOSE_WAIT; |
| 1418 |
break;
|
| 1419 |
|
| 1420 |
/*
|
| 1421 |
* If still in FIN_WAIT_1 STATE FIN has not been acked so
|
| 1422 |
* enter the CLOSING state.
|
| 1423 |
*/
|
| 1424 |
case TCPS_FIN_WAIT_1:
|
| 1425 |
tp->t_state = TCPS_CLOSING; |
| 1426 |
break;
|
| 1427 |
|
| 1428 |
/*
|
| 1429 |
* In FIN_WAIT_2 state enter the TIME_WAIT state,
|
| 1430 |
* starting the time-wait timer, turning off the other
|
| 1431 |
* standard timers.
|
| 1432 |
*/
|
| 1433 |
case TCPS_FIN_WAIT_2:
|
| 1434 |
tp->t_state = TCPS_TIME_WAIT; |
| 1435 |
tcp_canceltimers(tp); |
| 1436 |
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
|
| 1437 |
soisfdisconnected(so); |
| 1438 |
break;
|
| 1439 |
|
| 1440 |
/*
|
| 1441 |
* In TIME_WAIT state restart the 2 MSL time_wait timer.
|
| 1442 |
*/
|
| 1443 |
case TCPS_TIME_WAIT:
|
| 1444 |
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
|
| 1445 |
break;
|
| 1446 |
} |
| 1447 |
} |
| 1448 |
|
| 1449 |
/*
|
| 1450 |
* If this is a small packet, then ACK now - with Nagel
|
| 1451 |
* congestion avoidance sender won't send more until
|
| 1452 |
* he gets an ACK.
|
| 1453 |
*
|
| 1454 |
* See above.
|
| 1455 |
*/
|
| 1456 |
/* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) {
|
| 1457 |
*/
|
| 1458 |
/* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg &&
|
| 1459 |
* (so->so_iptos & IPTOS_LOWDELAY) == 0) ||
|
| 1460 |
* ((so->so_iptos & IPTOS_LOWDELAY) &&
|
| 1461 |
* ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
|
| 1462 |
*/
|
| 1463 |
if (ti->ti_len && (unsigned)ti->ti_len <= 5 && |
| 1464 |
((struct tcpiphdr_2 *)ti)->first_char == (char)27) { |
| 1465 |
tp->t_flags |= TF_ACKNOW; |
| 1466 |
} |
| 1467 |
|
| 1468 |
/*
|
| 1469 |
* Return any desired output.
|
| 1470 |
*/
|
| 1471 |
if (needoutput || (tp->t_flags & TF_ACKNOW)) {
|
| 1472 |
(void) tcp_output(tp);
|
| 1473 |
} |
| 1474 |
return;
|
| 1475 |
|
| 1476 |
dropafterack:
|
| 1477 |
/*
|
| 1478 |
* Generate an ACK dropping incoming segment if it occupies
|
| 1479 |
* sequence space, where the ACK reflects our state.
|
| 1480 |
*/
|
| 1481 |
if (tiflags & TH_RST)
|
| 1482 |
goto drop;
|
| 1483 |
m_freem(m); |
| 1484 |
tp->t_flags |= TF_ACKNOW; |
| 1485 |
(void) tcp_output(tp);
|
| 1486 |
return;
|
| 1487 |
|
| 1488 |
dropwithreset:
|
| 1489 |
/* reuses m if m!=NULL, m_free() unnecessary */
|
| 1490 |
if (tiflags & TH_ACK)
|
| 1491 |
tcp_respond(tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
|
| 1492 |
else {
|
| 1493 |
if (tiflags & TH_SYN) ti->ti_len++;
|
| 1494 |
tcp_respond(tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
|
| 1495 |
TH_RST|TH_ACK); |
| 1496 |
} |
| 1497 |
|
| 1498 |
return;
|
| 1499 |
|
| 1500 |
drop:
|
| 1501 |
/*
|
| 1502 |
* Drop space held by incoming segment and return.
|
| 1503 |
*/
|
| 1504 |
m_free(m); |
| 1505 |
|
| 1506 |
return;
|
| 1507 |
} |
| 1508 |
|
| 1509 |
/* , ts_present, ts_val, ts_ecr) */
|
| 1510 |
/* int *ts_present;
|
| 1511 |
* u_int32_t *ts_val, *ts_ecr;
|
| 1512 |
*/
|
| 1513 |
void
|
| 1514 |
tcp_dooptions(tp, cp, cnt, ti) |
| 1515 |
struct tcpcb *tp;
|
| 1516 |
u_char *cp; |
| 1517 |
int cnt;
|
| 1518 |
struct tcpiphdr *ti;
|
| 1519 |
{
|
| 1520 |
u_int16_t mss; |
| 1521 |
int opt, optlen;
|
| 1522 |
|
| 1523 |
DEBUG_CALL("tcp_dooptions");
|
| 1524 |
DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt)); |
| 1525 |
|
| 1526 |
for (; cnt > 0; cnt -= optlen, cp += optlen) { |
| 1527 |
opt = cp[0];
|
| 1528 |
if (opt == TCPOPT_EOL)
|
| 1529 |
break;
|
| 1530 |
if (opt == TCPOPT_NOP)
|
| 1531 |
optlen = 1;
|
| 1532 |
else {
|
| 1533 |
optlen = cp[1];
|
| 1534 |
if (optlen <= 0) |
| 1535 |
break;
|
| 1536 |
} |
| 1537 |
switch (opt) {
|
| 1538 |
|
| 1539 |
default:
|
| 1540 |
continue;
|
| 1541 |
|
| 1542 |
case TCPOPT_MAXSEG:
|
| 1543 |
if (optlen != TCPOLEN_MAXSEG)
|
| 1544 |
continue;
|
| 1545 |
if (!(ti->ti_flags & TH_SYN))
|
| 1546 |
continue;
|
| 1547 |
memcpy((char *) &mss, (char *) cp + 2, sizeof(mss)); |
| 1548 |
NTOHS(mss); |
| 1549 |
(void) tcp_mss(tp, mss); /* sets t_maxseg */ |
| 1550 |
break;
|
| 1551 |
|
| 1552 |
/* case TCPOPT_WINDOW:
|
| 1553 |
* if (optlen != TCPOLEN_WINDOW)
|
| 1554 |
* continue;
|
| 1555 |
* if (!(ti->ti_flags & TH_SYN))
|
| 1556 |
* continue;
|
| 1557 |
* tp->t_flags |= TF_RCVD_SCALE;
|
| 1558 |
* tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
|
| 1559 |
* break;
|
| 1560 |
*/
|
| 1561 |
/* case TCPOPT_TIMESTAMP:
|
| 1562 |
* if (optlen != TCPOLEN_TIMESTAMP)
|
| 1563 |
* continue;
|
| 1564 |
* *ts_present = 1;
|
| 1565 |
* memcpy((char *) ts_val, (char *)cp + 2, sizeof(*ts_val));
|
| 1566 |
* NTOHL(*ts_val);
|
| 1567 |
* memcpy((char *) ts_ecr, (char *)cp + 6, sizeof(*ts_ecr));
|
| 1568 |
* NTOHL(*ts_ecr);
|
| 1569 |
*
|
| 1570 |
*/ /* |
| 1571 |
* * A timestamp received in a SYN makes
|
| 1572 |
* * it ok to send timestamp requests and replies.
|
| 1573 |
* */
|
| 1574 |
/* if (ti->ti_flags & TH_SYN) {
|
| 1575 |
* tp->t_flags |= TF_RCVD_TSTMP;
|
| 1576 |
* tp->ts_recent = *ts_val;
|
| 1577 |
* tp->ts_recent_age = tcp_now;
|
| 1578 |
* }
|
| 1579 |
*/ break; |
| 1580 |
} |
| 1581 |
} |
| 1582 |
} |
| 1583 |
|
| 1584 |
|
| 1585 |
/*
|
| 1586 |
* Pull out of band byte out of a segment so
|
| 1587 |
* it doesn't appear in the user's data queue.
|
| 1588 |
* It is still reflected in the segment length for
|
| 1589 |
* sequencing purposes.
|
| 1590 |
*/
|
| 1591 |
|
| 1592 |
#ifdef notdef
|
| 1593 |
|
| 1594 |
void
|
| 1595 |
tcp_pulloutofband(so, ti, m) |
| 1596 |
struct socket *so;
|
| 1597 |
struct tcpiphdr *ti;
|
| 1598 |
register struct mbuf *m; |
| 1599 |
{
|
| 1600 |
int cnt = ti->ti_urp - 1; |
| 1601 |
|
| 1602 |
while (cnt >= 0) { |
| 1603 |
if (m->m_len > cnt) {
|
| 1604 |
char *cp = mtod(m, caddr_t) + cnt;
|
| 1605 |
struct tcpcb *tp = sototcpcb(so);
|
| 1606 |
|
| 1607 |
tp->t_iobc = *cp; |
| 1608 |
tp->t_oobflags |= TCPOOB_HAVEDATA; |
| 1609 |
memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1)); |
| 1610 |
m->m_len--; |
| 1611 |
return;
|
| 1612 |
} |
| 1613 |
cnt -= m->m_len; |
| 1614 |
m = m->m_next; /* XXX WRONG! Fix it! */
|
| 1615 |
if (m == 0) |
| 1616 |
break;
|
| 1617 |
} |
| 1618 |
panic("tcp_pulloutofband");
|
| 1619 |
} |
| 1620 |
|
| 1621 |
#endif /* notdef */ |
| 1622 |
|
| 1623 |
/*
|
| 1624 |
* Collect new round-trip time estimate
|
| 1625 |
* and update averages and current timeout.
|
| 1626 |
*/
|
| 1627 |
|
| 1628 |
void
|
| 1629 |
tcp_xmit_timer(tp, rtt) |
| 1630 |
register struct tcpcb *tp; |
| 1631 |
int rtt;
|
| 1632 |
{
|
| 1633 |
register short delta; |
| 1634 |
|
| 1635 |
DEBUG_CALL("tcp_xmit_timer");
|
| 1636 |
DEBUG_ARG("tp = %lx", (long)tp); |
| 1637 |
DEBUG_ARG("rtt = %d", rtt);
|
| 1638 |
|
| 1639 |
tcpstat.tcps_rttupdated++; |
| 1640 |
if (tp->t_srtt != 0) { |
| 1641 |
/*
|
| 1642 |
* srtt is stored as fixed point with 3 bits after the
|
| 1643 |
* binary point (i.e., scaled by 8). The following magic
|
| 1644 |
* is equivalent to the smoothing algorithm in rfc793 with
|
| 1645 |
* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
|
| 1646 |
* point). Adjust rtt to origin 0.
|
| 1647 |
*/
|
| 1648 |
delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
|
| 1649 |
if ((tp->t_srtt += delta) <= 0) |
| 1650 |
tp->t_srtt = 1;
|
| 1651 |
/*
|
| 1652 |
* We accumulate a smoothed rtt variance (actually, a
|
| 1653 |
* smoothed mean difference), then set the retransmit
|
| 1654 |
* timer to smoothed rtt + 4 times the smoothed variance.
|
| 1655 |
* rttvar is stored as fixed point with 2 bits after the
|
| 1656 |
* binary point (scaled by 4). The following is
|
| 1657 |
* equivalent to rfc793 smoothing with an alpha of .75
|
| 1658 |
* (rttvar = rttvar*3/4 + |delta| / 4). This replaces
|
| 1659 |
* rfc793's wired-in beta.
|
| 1660 |
*/
|
| 1661 |
if (delta < 0) |
| 1662 |
delta = -delta; |
| 1663 |
delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT); |
| 1664 |
if ((tp->t_rttvar += delta) <= 0) |
| 1665 |
tp->t_rttvar = 1;
|
| 1666 |
} else {
|
| 1667 |
/*
|
| 1668 |
* No rtt measurement yet - use the unsmoothed rtt.
|
| 1669 |
* Set the variance to half the rtt (so our first
|
| 1670 |
* retransmit happens at 3*rtt).
|
| 1671 |
*/
|
| 1672 |
tp->t_srtt = rtt << TCP_RTT_SHIFT; |
| 1673 |
tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
|
| 1674 |
} |
| 1675 |
tp->t_rtt = 0;
|
| 1676 |
tp->t_rxtshift = 0;
|
| 1677 |
|
| 1678 |
/*
|
| 1679 |
* the retransmit should happen at rtt + 4 * rttvar.
|
| 1680 |
* Because of the way we do the smoothing, srtt and rttvar
|
| 1681 |
* will each average +1/2 tick of bias. When we compute
|
| 1682 |
* the retransmit timer, we want 1/2 tick of rounding and
|
| 1683 |
* 1 extra tick because of +-1/2 tick uncertainty in the
|
| 1684 |
* firing of the timer. The bias will give us exactly the
|
| 1685 |
* 1.5 tick we need. But, because the bias is
|
| 1686 |
* statistical, we have to test that we don't drop below
|
| 1687 |
* the minimum feasible timer (which is 2 ticks).
|
| 1688 |
*/
|
| 1689 |
TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), |
| 1690 |
(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */ |
| 1691 |
|
| 1692 |
/*
|
| 1693 |
* We received an ack for a packet that wasn't retransmitted;
|
| 1694 |
* it is probably safe to discard any error indications we've
|
| 1695 |
* received recently. This isn't quite right, but close enough
|
| 1696 |
* for now (a route might have failed after we sent a segment,
|
| 1697 |
* and the return path might not be symmetrical).
|
| 1698 |
*/
|
| 1699 |
tp->t_softerror = 0;
|
| 1700 |
} |
| 1701 |
|
| 1702 |
/*
|
| 1703 |
* Determine a reasonable value for maxseg size.
|
| 1704 |
* If the route is known, check route for mtu.
|
| 1705 |
* If none, use an mss that can be handled on the outgoing
|
| 1706 |
* interface without forcing IP to fragment; if bigger than
|
| 1707 |
* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
|
| 1708 |
* to utilize large mbufs. If no route is found, route has no mtu,
|
| 1709 |
* or the destination isn't local, use a default, hopefully conservative
|
| 1710 |
* size (usually 512 or the default IP max size, but no more than the mtu
|
| 1711 |
* of the interface), as we can't discover anything about intervening
|
| 1712 |
* gateways or networks. We also initialize the congestion/slow start
|
| 1713 |
* window to be a single segment if the destination isn't local.
|
| 1714 |
* While looking at the routing entry, we also initialize other path-dependent
|
| 1715 |
* parameters from pre-set or cached values in the routing entry.
|
| 1716 |
*/
|
| 1717 |
|
| 1718 |
int
|
| 1719 |
tcp_mss(tp, offer) |
| 1720 |
register struct tcpcb *tp; |
| 1721 |
u_int offer; |
| 1722 |
{
|
| 1723 |
struct socket *so = tp->t_socket;
|
| 1724 |
int mss;
|
| 1725 |
|
| 1726 |
DEBUG_CALL("tcp_mss");
|
| 1727 |
DEBUG_ARG("tp = %lx", (long)tp); |
| 1728 |
DEBUG_ARG("offer = %d", offer);
|
| 1729 |
|
| 1730 |
mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr); |
| 1731 |
if (offer)
|
| 1732 |
mss = min(mss, offer); |
| 1733 |
mss = max(mss, 32);
|
| 1734 |
if (mss < tp->t_maxseg || offer != 0) |
| 1735 |
tp->t_maxseg = mss; |
| 1736 |
|
| 1737 |
tp->snd_cwnd = mss; |
| 1738 |
|
| 1739 |
sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
|
| 1740 |
sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
|
| 1741 |
|
| 1742 |
DEBUG_MISC((dfd, " returning mss = %d\n", mss));
|
| 1743 |
|
| 1744 |
return mss;
|
| 1745 |
} |