1/*
2 * Copyright (c) 2004, Robert N. M. Watson
3 * Copyright (c) 1995, Mike Mitchell
4 * Copyright (c) 1984, 1985, 1986, 1987, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. All advertising materials mentioning features or use of this software
16 * must display the following acknowledgement:
17 * This product includes software developed by the University of
18 * California, Berkeley and its contributors.
19 * 4. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 *
35 * @(#)spx_usrreq.h
36 */
37
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD: head/sys/netipx/spx_usrreq.c 139444 2004-12-30 17:49:40Z rwatson $");
40
41#include <sys/param.h>
42#include <sys/lock.h>
43#include <sys/malloc.h>
44#include <sys/mbuf.h>
45#include <sys/mutex.h>
46#include <sys/proc.h>
47#include <sys/protosw.h>
48#include <sys/signalvar.h>
49#include <sys/socket.h>
50#include <sys/socketvar.h>
51#include <sys/sx.h>
52#include <sys/systm.h>
53
54#include <net/route.h>
55#include <netinet/tcp_fsm.h>
56
57#include <netipx/ipx.h>
58#include <netipx/ipx_pcb.h>
59#include <netipx/ipx_var.h>
60#include <netipx/spx.h>
61#include <netipx/spx_debug.h>
62#include <netipx/spx_timer.h>
63#include <netipx/spx_var.h>
64
65/*
66 * SPX protocol implementation.
67 */
68static u_short spx_iss;
69static u_short spx_newchecks[50];
70static int spx_hardnosed;
71static int spx_use_delack = 0;
72static int traceallspxs = 0;
73static struct spx spx_savesi;
74static struct spx_istat spx_istat;
75
76/* Following was struct spxstat spxstat; */
77#ifndef spxstat
78#define spxstat spx_istat.newstats
79#endif
80
81static const int spx_backoff[SPX_MAXRXTSHIFT+1] =
82 { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
83
84static struct spxpcb *spx_close(struct spxpcb *cb);
85static struct spxpcb *spx_disconnect(struct spxpcb *cb);
86static struct spxpcb *spx_drop(struct spxpcb *cb, int errno);
87static int spx_output(struct spxpcb *cb, struct mbuf *m0);
88static int spx_reass(struct spxpcb *cb, struct spx *si);
89static void spx_setpersist(struct spxpcb *cb);
90static void spx_template(struct spxpcb *cb);
91static struct spxpcb *spx_timers(struct spxpcb *cb, int timer);
92static struct spxpcb *spx_usrclosed(struct spxpcb *cb);
93
94static int spx_usr_abort(struct socket *so);
95static int spx_accept(struct socket *so, struct sockaddr **nam);
96static int spx_attach(struct socket *so, int proto, struct thread *td);
97static int spx_bind(struct socket *so, struct sockaddr *nam, struct thread *td);
98static int spx_connect(struct socket *so, struct sockaddr *nam,
99 struct thread *td);
100static int spx_detach(struct socket *so);
101static int spx_usr_disconnect(struct socket *so);
102static int spx_listen(struct socket *so, struct thread *td);
103static int spx_rcvd(struct socket *so, int flags);
104static int spx_rcvoob(struct socket *so, struct mbuf *m, int flags);
105static int spx_send(struct socket *so, int flags, struct mbuf *m,
106 struct sockaddr *addr, struct mbuf *control,
107 struct thread *td);
108static int spx_shutdown(struct socket *so);
109static int spx_sp_attach(struct socket *so, int proto, struct thread *td);
110
111struct pr_usrreqs spx_usrreqs = {
112 .pru_abort = spx_usr_abort,
113 .pru_accept = spx_accept,
114 .pru_attach = spx_attach,
115 .pru_bind = spx_bind,
116 .pru_connect = spx_connect,
117 .pru_control = ipx_control,
118 .pru_detach = spx_detach,
119 .pru_disconnect = spx_usr_disconnect,
120 .pru_listen = spx_listen,
121 .pru_peeraddr = ipx_peeraddr,
122 .pru_rcvd = spx_rcvd,
123 .pru_rcvoob = spx_rcvoob,
124 .pru_send = spx_send,
125 .pru_shutdown = spx_shutdown,
126 .pru_sockaddr = ipx_sockaddr,
127};
128
129struct pr_usrreqs spx_usrreq_sps = {
130 .pru_abort = spx_usr_abort,
131 .pru_accept = spx_accept,
132 .pru_attach = spx_sp_attach,
133 .pru_bind = spx_bind,
134 .pru_connect = spx_connect,
135 .pru_control = ipx_control,
136 .pru_detach = spx_detach,
137 .pru_disconnect = spx_usr_disconnect,
138 .pru_listen = spx_listen,
139 .pru_peeraddr = ipx_peeraddr,
140 .pru_rcvd = spx_rcvd,
141 .pru_rcvoob = spx_rcvoob,
142 .pru_send = spx_send,
143 .pru_shutdown = spx_shutdown,
144 .pru_sockaddr = ipx_sockaddr,
145};
146
147void
148spx_init()
149{
150
151 spx_iss = 1; /* WRONG !! should fish it out of TODR */
152}
153
154void
155spx_input(m, ipxp)
156 register struct mbuf *m;
157 register struct ipxpcb *ipxp;
158{
159 register struct spxpcb *cb;
160 register struct spx *si = mtod(m, struct spx *);
161 register struct socket *so;
162 int dropsocket = 0;
163 short ostate = 0;
164
165 spxstat.spxs_rcvtotal++;
166 if (ipxp == NULL) {
167 panic("No ipxpcb in spx_input\n");
168 return;
169 }
170
171 cb = ipxtospxpcb(ipxp);
172 if (cb == NULL)
173 goto bad;
174
175 if (m->m_len < sizeof(*si)) {
176 if ((m = m_pullup(m, sizeof(*si))) == NULL) {
177 spxstat.spxs_rcvshort++;
178 return;
179 }
180 si = mtod(m, struct spx *);
181 }
182 si->si_seq = ntohs(si->si_seq);
183 si->si_ack = ntohs(si->si_ack);
184 si->si_alo = ntohs(si->si_alo);
185
186 so = ipxp->ipxp_socket;
187
188 if (so->so_options & SO_DEBUG || traceallspxs) {
189 ostate = cb->s_state;
190 spx_savesi = *si;
191 }
192 if (so->so_options & SO_ACCEPTCONN) {
193 struct spxpcb *ocb = cb;
194
195 so = sonewconn(so, 0);
196 if (so == NULL) {
197 goto drop;
198 }
199 /*
200 * This is ugly, but ....
201 *
202 * Mark socket as temporary until we're
203 * committed to keeping it. The code at
204 * ``drop'' and ``dropwithreset'' check the
205 * flag dropsocket to see if the temporary
206 * socket created here should be discarded.
207 * We mark the socket as discardable until
208 * we're committed to it below in TCPS_LISTEN.
209 */
210 dropsocket++;
211 ipxp = (struct ipxpcb *)so->so_pcb;
212 ipxp->ipxp_laddr = si->si_dna;
213 cb = ipxtospxpcb(ipxp);
214 cb->s_mtu = ocb->s_mtu; /* preserve sockopts */
215 cb->s_flags = ocb->s_flags; /* preserve sockopts */
216 cb->s_flags2 = ocb->s_flags2; /* preserve sockopts */
217 cb->s_state = TCPS_LISTEN;
218 }
219
220 /*
221 * Packet received on connection.
222 * reset idle time and keep-alive timer;
223 */
224 cb->s_idle = 0;
225 cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
226
227 switch (cb->s_state) {
228
229 case TCPS_LISTEN:{
230 struct sockaddr_ipx *sipx, ssipx;
231 struct ipx_addr laddr;
232
233 /*
234 * If somebody here was carying on a conversation
235 * and went away, and his pen pal thinks he can
236 * still talk, we get the misdirected packet.
237 */
238 if (spx_hardnosed && (si->si_did != 0 || si->si_seq != 0)) {
239 spx_istat.gonawy++;
240 goto dropwithreset;
241 }
242 sipx = &ssipx;
243 bzero(sipx, sizeof *sipx);
244 sipx->sipx_len = sizeof(*sipx);
245 sipx->sipx_family = AF_IPX;
246 sipx->sipx_addr = si->si_sna;
247 laddr = ipxp->ipxp_laddr;
248 if (ipx_nullhost(laddr))
249 ipxp->ipxp_laddr = si->si_dna;
250 if (ipx_pcbconnect(ipxp, (struct sockaddr *)sipx, &thread0)) {
251 ipxp->ipxp_laddr = laddr;
252 spx_istat.noconn++;
253 goto drop;
254 }
255 spx_template(cb);
256 dropsocket = 0; /* committed to socket */
257 cb->s_did = si->si_sid;
258 cb->s_rack = si->si_ack;
259 cb->s_ralo = si->si_alo;
260#define THREEWAYSHAKE
261#ifdef THREEWAYSHAKE
262 cb->s_state = TCPS_SYN_RECEIVED;
263 cb->s_force = 1 + SPXT_KEEP;
264 spxstat.spxs_accepts++;
265 cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
266 }
267 break;
268 /*
269 * This state means that we have heard a response
270 * to our acceptance of their connection
271 * It is probably logically unnecessary in this
272 * implementation.
273 */
274 case TCPS_SYN_RECEIVED: {
275 if (si->si_did != cb->s_sid) {
276 spx_istat.wrncon++;
277 goto drop;
278 }
279#endif
280 ipxp->ipxp_fport = si->si_sport;
281 cb->s_timer[SPXT_REXMT] = 0;
282 cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
283 soisconnected(so);
284 cb->s_state = TCPS_ESTABLISHED;
285 spxstat.spxs_accepts++;
286 }
287 break;
288
289 /*
290 * This state means that we have gotten a response
291 * to our attempt to establish a connection.
292 * We fill in the data from the other side,
293 * telling us which port to respond to, instead of the well-
294 * known one we might have sent to in the first place.
295 * We also require that this is a response to our
296 * connection id.
297 */
298 case TCPS_SYN_SENT:
299 if (si->si_did != cb->s_sid) {
300 spx_istat.notme++;
301 goto drop;
302 }
303 spxstat.spxs_connects++;
304 cb->s_did = si->si_sid;
305 cb->s_rack = si->si_ack;
306 cb->s_ralo = si->si_alo;
307 cb->s_dport = ipxp->ipxp_fport = si->si_sport;
308 cb->s_timer[SPXT_REXMT] = 0;
309 cb->s_flags |= SF_ACKNOW;
310 soisconnected(so);
311 cb->s_state = TCPS_ESTABLISHED;
312 /* Use roundtrip time of connection request for initial rtt */
313 if (cb->s_rtt) {
314 cb->s_srtt = cb->s_rtt << 3;
315 cb->s_rttvar = cb->s_rtt << 1;
316 SPXT_RANGESET(cb->s_rxtcur,
317 ((cb->s_srtt >> 2) + cb->s_rttvar) >> 1,
318 SPXTV_MIN, SPXTV_REXMTMAX);
319 cb->s_rtt = 0;
320 }
321 }
322 if (so->so_options & SO_DEBUG || traceallspxs)
323 spx_trace(SA_INPUT, (u_char)ostate, cb, &spx_savesi, 0);
324
325 m->m_len -= sizeof(struct ipx);
326 m->m_pkthdr.len -= sizeof(struct ipx);
327 m->m_data += sizeof(struct ipx);
328
329 if (spx_reass(cb, si)) {
330 m_freem(m);
331 }
332 if (cb->s_force || (cb->s_flags & (SF_ACKNOW|SF_WIN|SF_RXT)))
333 spx_output(cb, (struct mbuf *)NULL);
334 cb->s_flags &= ~(SF_WIN|SF_RXT);
335 return;
336
337dropwithreset:
338 if (dropsocket) {
339 struct socket *head;
340 ACCEPT_LOCK();
341 KASSERT((so->so_qstate & SQ_INCOMP) != 0,
342 ("spx_input: nascent socket not SQ_INCOMP on soabort()"));
343 head = so->so_head;
344 TAILQ_REMOVE(&head->so_incomp, so, so_list);
345 head->so_incqlen--;
346 so->so_qstate &= ~SQ_INCOMP;
347 so->so_head = NULL;
348 ACCEPT_UNLOCK();
349 soabort(so);
350 }
351 si->si_seq = ntohs(si->si_seq);
352 si->si_ack = ntohs(si->si_ack);
353 si->si_alo = ntohs(si->si_alo);
354 m_freem(dtom(si));
355 if (cb->s_ipxpcb->ipxp_socket->so_options & SO_DEBUG || traceallspxs)
356 spx_trace(SA_DROP, (u_char)ostate, cb, &spx_savesi, 0);
357 return;
358
359drop:
360bad:
361 if (cb == 0 || cb->s_ipxpcb->ipxp_socket->so_options & SO_DEBUG ||
362 traceallspxs)
363 spx_trace(SA_DROP, (u_char)ostate, cb, &spx_savesi, 0);
364 m_freem(m);
365}
366
367static int spxrexmtthresh = 3;
368
369/*
370 * This is structurally similar to the tcp reassembly routine
371 * but its function is somewhat different: It merely queues
372 * packets up, and suppresses duplicates.
373 */
374static int
375spx_reass(cb, si)
376register struct spxpcb *cb;
377register struct spx *si;
378{
379 register struct spx_q *q;
380 register struct mbuf *m;
381 register struct socket *so = cb->s_ipxpcb->ipxp_socket;
382 char packetp = cb->s_flags & SF_HI;
383 int incr;
384 char wakeup = 0;
385
386 if (si == SI(0))
387 goto present;
388 /*
389 * Update our news from them.
390 */
391 if (si->si_cc & SPX_SA)
392 cb->s_flags |= (spx_use_delack ? SF_DELACK : SF_ACKNOW);
393 if (SSEQ_GT(si->si_alo, cb->s_ralo))
394 cb->s_flags |= SF_WIN;
395 if (SSEQ_LEQ(si->si_ack, cb->s_rack)) {
396 if ((si->si_cc & SPX_SP) && cb->s_rack != (cb->s_smax + 1)) {
397 spxstat.spxs_rcvdupack++;
398 /*
399 * If this is a completely duplicate ack
400 * and other conditions hold, we assume
401 * a packet has been dropped and retransmit
402 * it exactly as in tcp_input().
403 */
404 if (si->si_ack != cb->s_rack ||
405 si->si_alo != cb->s_ralo)
406 cb->s_dupacks = 0;
407 else if (++cb->s_dupacks == spxrexmtthresh) {
408 u_short onxt = cb->s_snxt;
409 int cwnd = cb->s_cwnd;
410
411 cb->s_snxt = si->si_ack;
412 cb->s_cwnd = CUNIT;
413 cb->s_force = 1 + SPXT_REXMT;
414 spx_output(cb, (struct mbuf *)NULL);
415 cb->s_timer[SPXT_REXMT] = cb->s_rxtcur;
416 cb->s_rtt = 0;
417 if (cwnd >= 4 * CUNIT)
418 cb->s_cwnd = cwnd / 2;
419 if (SSEQ_GT(onxt, cb->s_snxt))
420 cb->s_snxt = onxt;
421 return (1);
422 }
423 } else
424 cb->s_dupacks = 0;
425 goto update_window;
426 }
427 cb->s_dupacks = 0;
428 /*
429 * If our correspondent acknowledges data we haven't sent
430 * TCP would drop the packet after acking. We'll be a little
431 * more permissive
432 */
433 if (SSEQ_GT(si->si_ack, (cb->s_smax + 1))) {
434 spxstat.spxs_rcvacktoomuch++;
435 si->si_ack = cb->s_smax + 1;
436 }
437 spxstat.spxs_rcvackpack++;
438 /*
439 * If transmit timer is running and timed sequence
440 * number was acked, update smoothed round trip time.
441 * See discussion of algorithm in tcp_input.c
442 */
443 if (cb->s_rtt && SSEQ_GT(si->si_ack, cb->s_rtseq)) {
444 spxstat.spxs_rttupdated++;
445 if (cb->s_srtt != 0) {
446 register short delta;
447 delta = cb->s_rtt - (cb->s_srtt >> 3);
448 if ((cb->s_srtt += delta) <= 0)
449 cb->s_srtt = 1;
450 if (delta < 0)
451 delta = -delta;
452 delta -= (cb->s_rttvar >> 2);
453 if ((cb->s_rttvar += delta) <= 0)
454 cb->s_rttvar = 1;
455 } else {
456 /*
457 * No rtt measurement yet
458 */
459 cb->s_srtt = cb->s_rtt << 3;
460 cb->s_rttvar = cb->s_rtt << 1;
461 }
462 cb->s_rtt = 0;
463 cb->s_rxtshift = 0;
464 SPXT_RANGESET(cb->s_rxtcur,
465 ((cb->s_srtt >> 2) + cb->s_rttvar) >> 1,
466 SPXTV_MIN, SPXTV_REXMTMAX);
467 }
468 /*
469 * If all outstanding data is acked, stop retransmit
470 * timer and remember to restart (more output or persist).
471 * If there is more data to be acked, restart retransmit
472 * timer, using current (possibly backed-off) value;
473 */
474 if (si->si_ack == cb->s_smax + 1) {
475 cb->s_timer[SPXT_REXMT] = 0;
476 cb->s_flags |= SF_RXT;
477 } else if (cb->s_timer[SPXT_PERSIST] == 0)
478 cb->s_timer[SPXT_REXMT] = cb->s_rxtcur;
479 /*
480 * When new data is acked, open the congestion window.
481 * If the window gives us less than ssthresh packets
482 * in flight, open exponentially (maxseg at a time).
483 * Otherwise open linearly (maxseg^2 / cwnd at a time).
484 */
485 incr = CUNIT;
486 if (cb->s_cwnd > cb->s_ssthresh)
487 incr = max(incr * incr / cb->s_cwnd, 1);
488 cb->s_cwnd = min(cb->s_cwnd + incr, cb->s_cwmx);
489 /*
490 * Trim Acked data from output queue.
491 */
492 while ((m = so->so_snd.sb_mb) != NULL) {
493 if (SSEQ_LT((mtod(m, struct spx *))->si_seq, si->si_ack))
494 sbdroprecord(&so->so_snd);
495 else
496 break;
497 }
498 sowwakeup(so);
499 cb->s_rack = si->si_ack;
500update_window:
501 if (SSEQ_LT(cb->s_snxt, cb->s_rack))
502 cb->s_snxt = cb->s_rack;
503 if (SSEQ_LT(cb->s_swl1, si->si_seq) || ((cb->s_swl1 == si->si_seq &&
504 (SSEQ_LT(cb->s_swl2, si->si_ack))) ||
505 (cb->s_swl2 == si->si_ack && SSEQ_LT(cb->s_ralo, si->si_alo)))) {
506 /* keep track of pure window updates */
507 if ((si->si_cc & SPX_SP) && cb->s_swl2 == si->si_ack
508 && SSEQ_LT(cb->s_ralo, si->si_alo)) {
509 spxstat.spxs_rcvwinupd++;
510 spxstat.spxs_rcvdupack--;
511 }
512 cb->s_ralo = si->si_alo;
513 cb->s_swl1 = si->si_seq;
514 cb->s_swl2 = si->si_ack;
515 cb->s_swnd = (1 + si->si_alo - si->si_ack);
516 if (cb->s_swnd > cb->s_smxw)
517 cb->s_smxw = cb->s_swnd;
518 cb->s_flags |= SF_WIN;
519 }
520 /*
521 * If this packet number is higher than that which
522 * we have allocated refuse it, unless urgent
523 */
524 if (SSEQ_GT(si->si_seq, cb->s_alo)) {
525 if (si->si_cc & SPX_SP) {
526 spxstat.spxs_rcvwinprobe++;
527 return (1);
528 } else
529 spxstat.spxs_rcvpackafterwin++;
530 if (si->si_cc & SPX_OB) {
531 if (SSEQ_GT(si->si_seq, cb->s_alo + 60)) {
532 m_freem(dtom(si));
533 return (0);
534 } /* else queue this packet; */
535 } else {
536 /*register struct socket *so = cb->s_ipxpcb->ipxp_socket;
537 if (so->so_state && SS_NOFDREF) {
538 spx_close(cb);
539 } else
540 would crash system*/
541 spx_istat.notyet++;
542 m_freem(dtom(si));
543 return (0);
544 }
545 }
546 /*
547 * If this is a system packet, we don't need to
548 * queue it up, and won't update acknowledge #
549 */
550 if (si->si_cc & SPX_SP) {
551 return (1);
552 }
553 /*
554 * We have already seen this packet, so drop.
555 */
556 if (SSEQ_LT(si->si_seq, cb->s_ack)) {
557 spx_istat.bdreas++;
558 spxstat.spxs_rcvduppack++;
559 if (si->si_seq == cb->s_ack - 1)
560 spx_istat.lstdup++;
561 return (1);
562 }
563 /*
564 * Loop through all packets queued up to insert in
565 * appropriate sequence.
566 */
567 for (q = cb->s_q.si_next; q != &cb->s_q; q = q->si_next) {
568 if (si->si_seq == SI(q)->si_seq) {
569 spxstat.spxs_rcvduppack++;
570 return (1);
571 }
572 if (SSEQ_LT(si->si_seq, SI(q)->si_seq)) {
573 spxstat.spxs_rcvoopack++;
574 break;
575 }
576 }
577 insque(si, q->si_prev);
578 /*
579 * If this packet is urgent, inform process
580 */
581 if (si->si_cc & SPX_OB) {
582 cb->s_iobc = ((char *)si)[1 + sizeof(*si)];
583 sohasoutofband(so);
584 cb->s_oobflags |= SF_IOOB;
585 }
586present:
587#define SPINC sizeof(struct spxhdr)
588 /*
589 * Loop through all packets queued up to update acknowledge
590 * number, and present all acknowledged data to user;
591 * If in packet interface mode, show packet headers.
592 */
593 for (q = cb->s_q.si_next; q != &cb->s_q; q = q->si_next) {
594 if (SI(q)->si_seq == cb->s_ack) {
595 cb->s_ack++;
596 m = dtom(q);
597 if (SI(q)->si_cc & SPX_OB) {
598 cb->s_oobflags &= ~SF_IOOB;
599 SOCKBUF_LOCK(&so->so_rcv);
600 if (so->so_rcv.sb_cc)
601 so->so_oobmark = so->so_rcv.sb_cc;
602 else
603 so->so_rcv.sb_state |= SBS_RCVATMARK;
604 SOCKBUF_UNLOCK(&so->so_rcv);
605 }
606 q = q->si_prev;
607 remque(q->si_next);
608 wakeup = 1;
609 spxstat.spxs_rcvpack++;
610#ifdef SF_NEWCALL
611 if (cb->s_flags2 & SF_NEWCALL) {
612 struct spxhdr *sp = mtod(m, struct spxhdr *);
613 u_char dt = sp->spx_dt;
614 spx_newchecks[4]++;
615 if (dt != cb->s_rhdr.spx_dt) {
616 struct mbuf *mm =
617 m_getclr(M_DONTWAIT, MT_CONTROL);
618 spx_newchecks[0]++;
619 if (mm != NULL) {
620 u_short *s =
621 mtod(mm, u_short *);
622 cb->s_rhdr.spx_dt = dt;
623 mm->m_len = 5; /*XXX*/
624 s[0] = 5;
625 s[1] = 1;
626 *(u_char *)(&s[2]) = dt;
627 sbappend(&so->so_rcv, mm);
628 }
629 }
630 if (sp->spx_cc & SPX_OB) {
631 MCHTYPE(m, MT_OOBDATA);
632 spx_newchecks[1]++;
633 SOCKBUF_LOCK(&so->so_rcv);
634 so->so_oobmark = 0;
635 so->so_rcv.sb_state &= ~SBS_RCVATMARK;
636 SOCKBUF_UNLOCK(&so->so_rcv);
637 }
638 if (packetp == 0) {
639 m->m_data += SPINC;
640 m->m_len -= SPINC;
641 m->m_pkthdr.len -= SPINC;
642 }
643 if ((sp->spx_cc & SPX_EM) || packetp) {
644 sbappendrecord(&so->so_rcv, m);
645 spx_newchecks[9]++;
646 } else
647 sbappend(&so->so_rcv, m);
648 } else
649#endif
650 if (packetp) {
651 sbappendrecord(&so->so_rcv, m);
652 } else {
653 cb->s_rhdr = *mtod(m, struct spxhdr *);
654 m->m_data += SPINC;
655 m->m_len -= SPINC;
656 m->m_pkthdr.len -= SPINC;
657 sbappend(&so->so_rcv, m);
658 }
659 } else
660 break;
661 }
662 if (wakeup)
663 sorwakeup(so);
664 return (0);
665}
666
667void
668spx_ctlinput(cmd, arg_as_sa, dummy)
669 int cmd;
670 struct sockaddr *arg_as_sa; /* XXX should be swapped with dummy */
671 void *dummy;
672{
673 caddr_t arg = (/* XXX */ caddr_t)arg_as_sa;
674 struct ipx_addr *na;
675 struct sockaddr_ipx *sipx;
676
677 if (cmd < 0 || cmd >= PRC_NCMDS)
678 return;
679
680 switch (cmd) {
681
682 case PRC_ROUTEDEAD:
683 return;
684
685 case PRC_IFDOWN:
686 case PRC_HOSTDEAD:
687 case PRC_HOSTUNREACH:
688 sipx = (struct sockaddr_ipx *)arg;
689 if (sipx->sipx_family != AF_IPX)
690 return;
691 na = &sipx->sipx_addr;
692 break;
693
694 default:
695 break;
696 }
697}
698
699static int
700spx_output(cb, m0)
701 register struct spxpcb *cb;
702 struct mbuf *m0;
703{
704 struct socket *so = cb->s_ipxpcb->ipxp_socket;
705 register struct mbuf *m;
706 register struct spx *si = (struct spx *)NULL;
707 register struct sockbuf *sb = &so->so_snd;
708 int len = 0, win, rcv_win;
709 short span, off, recordp = 0;
710 u_short alo;
711 int error = 0, sendalot;
712#ifdef notdef
713 int idle;
714#endif
715 struct mbuf *mprev;
716
717 if (m0 != NULL) {
718 int mtu = cb->s_mtu;
719 int datalen;
720 /*
721 * Make sure that packet isn't too big.
722 */
723 for (m = m0; m != NULL; m = m->m_next) {
724 mprev = m;
725 len += m->m_len;
726 if (m->m_flags & M_EOR)
727 recordp = 1;
728 }
729 datalen = (cb->s_flags & SF_HO) ?
730 len - sizeof(struct spxhdr) : len;
731 if (datalen > mtu) {
732 if (cb->s_flags & SF_PI) {
733 m_freem(m0);
734 return (EMSGSIZE);
735 } else {
736 int oldEM = cb->s_cc & SPX_EM;
737
738 cb->s_cc &= ~SPX_EM;
739 while (len > mtu) {
740 /*
741 * Here we are only being called
742 * from usrreq(), so it is OK to
743 * block.
744 */
745 m = m_copym(m0, 0, mtu, M_TRYWAIT);
746 if (cb->s_flags & SF_NEWCALL) {
747 struct mbuf *mm = m;
748 spx_newchecks[7]++;
749 while (mm != NULL) {
750 mm->m_flags &= ~M_EOR;
751 mm = mm->m_next;
752 }
753 }
754 error = spx_output(cb, m);
755 if (error) {
756 cb->s_cc |= oldEM;
757 m_freem(m0);
758 return (error);
759 }
760 m_adj(m0, mtu);
761 len -= mtu;
762 }
763 cb->s_cc |= oldEM;
764 }
765 }
766 /*
767 * Force length even, by adding a "garbage byte" if
768 * necessary.
769 */
770 if (len & 1) {
771 m = mprev;
772 if (M_TRAILINGSPACE(m) >= 1)
773 m->m_len++;
774 else {
775 struct mbuf *m1 = m_get(M_DONTWAIT, MT_DATA);
776
777 if (m1 == NULL) {
778 m_freem(m0);
779 return (ENOBUFS);
780 }
781 m1->m_len = 1;
782 *(mtod(m1, u_char *)) = 0;
783 m->m_next = m1;
784 }
785 }
786 m = m_gethdr(M_DONTWAIT, MT_HEADER);
787 if (m == NULL) {
788 m_freem(m0);
789 return (ENOBUFS);
790 }
791 /*
792 * Fill in mbuf with extended SP header
793 * and addresses and length put into network format.
794 */
795 MH_ALIGN(m, sizeof(struct spx));
796 m->m_len = sizeof(struct spx);
797 m->m_next = m0;
798 si = mtod(m, struct spx *);
799 si->si_i = *cb->s_ipx;
800 si->si_s = cb->s_shdr;
801 if ((cb->s_flags & SF_PI) && (cb->s_flags & SF_HO)) {
802 register struct spxhdr *sh;
803 if (m0->m_len < sizeof(*sh)) {
804 if((m0 = m_pullup(m0, sizeof(*sh))) == NULL) {
805 m_free(m);
806 m_freem(m0);
807 return (EINVAL);
808 }
809 m->m_next = m0;
810 }
811 sh = mtod(m0, struct spxhdr *);
812 si->si_dt = sh->spx_dt;
813 si->si_cc |= sh->spx_cc & SPX_EM;
814 m0->m_len -= sizeof(*sh);
815 m0->m_data += sizeof(*sh);
816 len -= sizeof(*sh);
817 }
818 len += sizeof(*si);
819 if ((cb->s_flags2 & SF_NEWCALL) && recordp) {
820 si->si_cc |= SPX_EM;
821 spx_newchecks[8]++;
822 }
823 if (cb->s_oobflags & SF_SOOB) {
824 /*
825 * Per jqj@cornell:
826 * make sure OB packets convey exactly 1 byte.
827 * If the packet is 1 byte or larger, we
828 * have already guaranted there to be at least
829 * one garbage byte for the checksum, and
830 * extra bytes shouldn't hurt!
831 */
832 if (len > sizeof(*si)) {
833 si->si_cc |= SPX_OB;
834 len = (1 + sizeof(*si));
835 }
836 }
837 si->si_len = htons((u_short)len);
838 m->m_pkthdr.len = ((len - 1) | 1) + 1;
839 /*
840 * queue stuff up for output
841 */
842 sbappendrecord(sb, m);
843 cb->s_seq++;
844 }
845#ifdef notdef
846 idle = (cb->s_smax == (cb->s_rack - 1));
847#endif
848again:
849 sendalot = 0;
850 off = cb->s_snxt - cb->s_rack;
851 win = min(cb->s_swnd, (cb->s_cwnd / CUNIT));
852
853 /*
854 * If in persist timeout with window of 0, send a probe.
855 * Otherwise, if window is small but nonzero
856 * and timer expired, send what we can and go into
857 * transmit state.
858 */
859 if (cb->s_force == 1 + SPXT_PERSIST) {
860 if (win != 0) {
861 cb->s_timer[SPXT_PERSIST] = 0;
862 cb->s_rxtshift = 0;
863 }
864 }
865 span = cb->s_seq - cb->s_rack;
866 len = min(span, win) - off;
867
868 if (len < 0) {
869 /*
870 * Window shrank after we went into it.
871 * If window shrank to 0, cancel pending
872 * restransmission and pull s_snxt back
873 * to (closed) window. We will enter persist
874 * state below. If the widndow didn't close completely,
875 * just wait for an ACK.
876 */
877 len = 0;
878 if (win == 0) {
879 cb->s_timer[SPXT_REXMT] = 0;
880 cb->s_snxt = cb->s_rack;
881 }
882 }
883 if (len > 1)
884 sendalot = 1;
885 rcv_win = sbspace(&so->so_rcv);
886
887 /*
888 * Send if we owe peer an ACK.
889 */
890 if (cb->s_oobflags & SF_SOOB) {
891 /*
892 * must transmit this out of band packet
893 */
894 cb->s_oobflags &= ~ SF_SOOB;
895 sendalot = 1;
896 spxstat.spxs_sndurg++;
897 goto found;
898 }
899 if (cb->s_flags & SF_ACKNOW)
900 goto send;
901 if (cb->s_state < TCPS_ESTABLISHED)
902 goto send;
903 /*
904 * Silly window can't happen in spx.
905 * Code from tcp deleted.
906 */
907 if (len)
908 goto send;
909 /*
910 * Compare available window to amount of window
911 * known to peer (as advertised window less
912 * next expected input.) If the difference is at least two
913 * packets or at least 35% of the mximum possible window,
914 * then want to send a window update to peer.
915 */
916 if (rcv_win > 0) {
917 u_short delta = 1 + cb->s_alo - cb->s_ack;
918 int adv = rcv_win - (delta * cb->s_mtu);
919
920 if ((so->so_rcv.sb_cc == 0 && adv >= (2 * cb->s_mtu)) ||
921 (100 * adv / so->so_rcv.sb_hiwat >= 35)) {
922 spxstat.spxs_sndwinup++;
923 cb->s_flags |= SF_ACKNOW;
924 goto send;
925 }
926
927 }
928 /*
929 * Many comments from tcp_output.c are appropriate here
930 * including . . .
931 * If send window is too small, there is data to transmit, and no
932 * retransmit or persist is pending, then go to persist state.
933 * If nothing happens soon, send when timer expires:
934 * if window is nonzero, transmit what we can,
935 * otherwise send a probe.
936 */
937 if (so->so_snd.sb_cc && cb->s_timer[SPXT_REXMT] == 0 &&
938 cb->s_timer[SPXT_PERSIST] == 0) {
939 cb->s_rxtshift = 0;
940 spx_setpersist(cb);
941 }
942 /*
943 * No reason to send a packet, just return.
944 */
945 cb->s_outx = 1;
946 return (0);
947
948send:
949 /*
950 * Find requested packet.
951 */
952 si = 0;
953 if (len > 0) {
954 cb->s_want = cb->s_snxt;
955 for (m = sb->sb_mb; m != NULL; m = m->m_act) {
956 si = mtod(m, struct spx *);
957 if (SSEQ_LEQ(cb->s_snxt, si->si_seq))
958 break;
959 }
960 found:
961 if (si != NULL) {
962 if (si->si_seq == cb->s_snxt)
963 cb->s_snxt++;
964 else
965 spxstat.spxs_sndvoid++, si = 0;
966 }
967 }
968 /*
969 * update window
970 */
971 if (rcv_win < 0)
972 rcv_win = 0;
973 alo = cb->s_ack - 1 + (rcv_win / ((short)cb->s_mtu));
974 if (SSEQ_LT(alo, cb->s_alo))
975 alo = cb->s_alo;
976
977 if (si != NULL) {
978 /*
979 * must make a copy of this packet for
980 * ipx_output to monkey with
981 */
982 m = m_copy(dtom(si), 0, (int)M_COPYALL);
983 if (m == NULL) {
984 return (ENOBUFS);
985 }
986 si = mtod(m, struct spx *);
987 if (SSEQ_LT(si->si_seq, cb->s_smax))
988 spxstat.spxs_sndrexmitpack++;
989 else
990 spxstat.spxs_sndpack++;
991 } else if (cb->s_force || cb->s_flags & SF_ACKNOW) {
992 /*
993 * Must send an acknowledgement or a probe
994 */
995 if (cb->s_force)
996 spxstat.spxs_sndprobe++;
997 if (cb->s_flags & SF_ACKNOW)
998 spxstat.spxs_sndacks++;
999 m = m_gethdr(M_DONTWAIT, MT_HEADER);
1000 if (m == NULL)
1001 return (ENOBUFS);
1002 /*
1003 * Fill in mbuf with extended SP header
1004 * and addresses and length put into network format.
1005 */
1006 MH_ALIGN(m, sizeof(struct spx));
1007 m->m_len = sizeof(*si);
1008 m->m_pkthdr.len = sizeof(*si);
1009 si = mtod(m, struct spx *);
1010 si->si_i = *cb->s_ipx;
1011 si->si_s = cb->s_shdr;
1012 si->si_seq = cb->s_smax + 1;
1013 si->si_len = htons(sizeof(*si));
1014 si->si_cc |= SPX_SP;
1015 } else {
1016 cb->s_outx = 3;
1017 if (so->so_options & SO_DEBUG || traceallspxs)
1018 spx_trace(SA_OUTPUT, cb->s_state, cb, si, 0);
1019 return (0);
1020 }
1021 /*
1022 * Stuff checksum and output datagram.
1023 */
1024 if ((si->si_cc & SPX_SP) == 0) {
1025 if (cb->s_force != (1 + SPXT_PERSIST) ||
1026 cb->s_timer[SPXT_PERSIST] == 0) {
1027 /*
1028 * If this is a new packet and we are not currently
1029 * timing anything, time this one.
1030 */
1031 if (SSEQ_LT(cb->s_smax, si->si_seq)) {
1032 cb->s_smax = si->si_seq;
1033 if (cb->s_rtt == 0) {
1034 spxstat.spxs_segstimed++;
1035 cb->s_rtseq = si->si_seq;
1036 cb->s_rtt = 1;
1037 }
1038 }
1039 /*
1040 * Set rexmt timer if not currently set,
1041 * Initial value for retransmit timer is smoothed
1042 * round-trip time + 2 * round-trip time variance.
1043 * Initialize shift counter which is used for backoff
1044 * of retransmit time.
1045 */
1046 if (cb->s_timer[SPXT_REXMT] == 0 &&
1047 cb->s_snxt != cb->s_rack) {
1048 cb->s_timer[SPXT_REXMT] = cb->s_rxtcur;
1049 if (cb->s_timer[SPXT_PERSIST]) {
1050 cb->s_timer[SPXT_PERSIST] = 0;
1051 cb->s_rxtshift = 0;
1052 }
1053 }
1054 } else if (SSEQ_LT(cb->s_smax, si->si_seq)) {
1055 cb->s_smax = si->si_seq;
1056 }
1057 } else if (cb->s_state < TCPS_ESTABLISHED) {
1058 if (cb->s_rtt == 0)
1059 cb->s_rtt = 1; /* Time initial handshake */
1060 if (cb->s_timer[SPXT_REXMT] == 0)
1061 cb->s_timer[SPXT_REXMT] = cb->s_rxtcur;
1062 }
1063 {
1064 /*
1065 * Do not request acks when we ack their data packets or
1066 * when we do a gratuitous window update.
1067 */
1068 if (((si->si_cc & SPX_SP) == 0) || cb->s_force)
1069 si->si_cc |= SPX_SA;
1070 si->si_seq = htons(si->si_seq);
1071 si->si_alo = htons(alo);
1072 si->si_ack = htons(cb->s_ack);
1073
1074 if (ipxcksum) {
1075 si->si_sum = ipx_cksum(m, ntohs(si->si_len));
1076 } else
1077 si->si_sum = 0xffff;
1078
1079 cb->s_outx = 4;
1080 if (so->so_options & SO_DEBUG || traceallspxs)
1081 spx_trace(SA_OUTPUT, cb->s_state, cb, si, 0);
1082
1083 if (so->so_options & SO_DONTROUTE)
1084 error = ipx_outputfl(m, (struct route *)NULL, IPX_ROUTETOIF);
1085 else
1086 error = ipx_outputfl(m, &cb->s_ipxpcb->ipxp_route, 0);
1087 }
1088 if (error) {
1089 return (error);
1090 }
1091 spxstat.spxs_sndtotal++;
1092 /*
1093 * Data sent (as far as we can tell).
1094 * If this advertises a larger window than any other segment,
1095 * then remember the size of the advertized window.
1096 * Any pending ACK has now been sent.
1097 */
1098 cb->s_force = 0;
1099 cb->s_flags &= ~(SF_ACKNOW|SF_DELACK);
1100 if (SSEQ_GT(alo, cb->s_alo))
1101 cb->s_alo = alo;
1102 if (sendalot)
1103 goto again;
1104 cb->s_outx = 5;
1105 return (0);
1106}
1107
1108static int spx_do_persist_panics = 0;
1109
1110static void
1111spx_setpersist(cb)
1112 register struct spxpcb *cb;
1113{
1114 register int t = ((cb->s_srtt >> 2) + cb->s_rttvar) >> 1;
1115
1116 if (cb->s_timer[SPXT_REXMT] && spx_do_persist_panics)
1117 panic("spx_output REXMT");
1118 /*
1119 * Start/restart persistance timer.
1120 */
1121 SPXT_RANGESET(cb->s_timer[SPXT_PERSIST],
1122 t*spx_backoff[cb->s_rxtshift],
1123 SPXTV_PERSMIN, SPXTV_PERSMAX);
1124 if (cb->s_rxtshift < SPX_MAXRXTSHIFT)
1125 cb->s_rxtshift++;
1126}
1127
1128int
1129spx_ctloutput(so, sopt)
1130 struct socket *so;
1131 struct sockopt *sopt;
1132{
1133 struct ipxpcb *ipxp = sotoipxpcb(so);
1134 register struct spxpcb *cb;
1135 int mask, error;
1136 short soptval;
1137 u_short usoptval;
1138 int optval;
1139
1140 error = 0;
1141
1142 if (sopt->sopt_level != IPXPROTO_SPX) {
1143 /* This will have to be changed when we do more general
1144 stacking of protocols */
1145 return (ipx_ctloutput(so, sopt));
1146 }
1147 if (ipxp == NULL)
1148 return (EINVAL);
1149 else
1150 cb = ipxtospxpcb(ipxp);
1151
1152 switch (sopt->sopt_dir) {
1153 case SOPT_GET:
1154 switch (sopt->sopt_name) {
1155 case SO_HEADERS_ON_INPUT:
1156 mask = SF_HI;
1157 goto get_flags;
1158
1159 case SO_HEADERS_ON_OUTPUT:
1160 mask = SF_HO;
1161 get_flags:
1162 soptval = cb->s_flags & mask;
1163 error = sooptcopyout(sopt, &soptval, sizeof soptval);
1164 break;
1165
1166 case SO_MTU:
1167 usoptval = cb->s_mtu;
1168 error = sooptcopyout(sopt, &usoptval, sizeof usoptval);
1169 break;
1170
1171 case SO_LAST_HEADER:
1172 error = sooptcopyout(sopt, &cb->s_rhdr,
1173 sizeof cb->s_rhdr);
1174 break;
1175
1176 case SO_DEFAULT_HEADERS:
1177 error = sooptcopyout(sopt, &cb->s_shdr,
1178 sizeof cb->s_shdr);
1179 break;
1180
1181 default:
1182 error = ENOPROTOOPT;
1183 }
1184 break;
1185
1186 case SOPT_SET:
1187 switch (sopt->sopt_name) {
1188 /* XXX why are these shorts on get and ints on set?
1189 that doesn't make any sense... */
1190 case SO_HEADERS_ON_INPUT:
1191 mask = SF_HI;
1192 goto set_head;
1193
1194 case SO_HEADERS_ON_OUTPUT:
1195 mask = SF_HO;
1196 set_head:
1197 error = sooptcopyin(sopt, &optval, sizeof optval,
1198 sizeof optval);
1199 if (error)
1200 break;
1201
1202 if (cb->s_flags & SF_PI) {
1203 if (optval)
1204 cb->s_flags |= mask;
1205 else
1206 cb->s_flags &= ~mask;
1207 } else error = EINVAL;
1208 break;
1209
1210 case SO_MTU:
1211 error = sooptcopyin(sopt, &usoptval, sizeof usoptval,
1212 sizeof usoptval);
1213 if (error)
1214 break;
1215 cb->s_mtu = usoptval;
1216 break;
1217
1218#ifdef SF_NEWCALL
1219 case SO_NEWCALL:
1220 error = sooptcopyin(sopt, &optval, sizeof optval,
1221 sizeof optval);
1222 if (error)
1223 break;
1224 if (optval) {
1225 cb->s_flags2 |= SF_NEWCALL;
1226 spx_newchecks[5]++;
1227 } else {
1228 cb->s_flags2 &= ~SF_NEWCALL;
1229 spx_newchecks[6]++;
1230 }
1231 break;
1232#endif
1233
1234 case SO_DEFAULT_HEADERS:
1235 {
1236 struct spxhdr sp;
1237
1238 error = sooptcopyin(sopt, &sp, sizeof sp,
1239 sizeof sp);
1240 if (error)
1241 break;
1242 cb->s_dt = sp.spx_dt;
1243 cb->s_cc = sp.spx_cc & SPX_EM;
1244 }
1245 break;
1246
1247 default:
1248 error = ENOPROTOOPT;
1249 }
1250 break;
1251 }
1252 return (error);
1253}
1254
1255static int
1256spx_usr_abort(so)
1257 struct socket *so;
1258{
1259 int s;
1260 struct ipxpcb *ipxp;
1261 struct spxpcb *cb;
1262
1263 ipxp = sotoipxpcb(so);
1264 cb = ipxtospxpcb(ipxp);
1265
1266 s = splnet();
1267 spx_drop(cb, ECONNABORTED);
1268 splx(s);
1269 return (0);
1270}
1271
1272/*
1273 * Accept a connection. Essentially all the work is
1274 * done at higher levels; just return the address
1275 * of the peer, storing through addr.
1276 */
1277static int
1278spx_accept(so, nam)
1279 struct socket *so;
1280 struct sockaddr **nam;
1281{
1282 struct ipxpcb *ipxp;
1283 struct sockaddr_ipx *sipx, ssipx;
1284
1285 ipxp = sotoipxpcb(so);
1286 sipx = &ssipx;
1287 bzero(sipx, sizeof *sipx);
1288 sipx->sipx_len = sizeof *sipx;
1289 sipx->sipx_family = AF_IPX;
1290 sipx->sipx_addr = ipxp->ipxp_faddr;
1291 *nam = sodupsockaddr((struct sockaddr *)sipx, M_NOWAIT);
1292 return (0);
1293}
1294
1295static int
1296spx_attach(so, proto, td)
1297 struct socket *so;
1298 int proto;
1299 struct thread *td;
1300{
1301 int error;
1302 int s;
1303 struct ipxpcb *ipxp;
1304 struct spxpcb *cb;
1305 struct mbuf *mm;
1306 struct sockbuf *sb;
1307
1308 ipxp = sotoipxpcb(so);
1309 cb = ipxtospxpcb(ipxp);
1310
1311 if (ipxp != NULL)
1312 return (EISCONN);
1313 s = splnet();
1314 error = ipx_pcballoc(so, &ipxpcb_list, td);
1315 if (error)
1316 goto spx_attach_end;
1317 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1318 error = soreserve(so, (u_long) 3072, (u_long) 3072);
1319 if (error)
1320 goto spx_attach_end;
1321 }
1322 ipxp = sotoipxpcb(so);
1323
1324 MALLOC(cb, struct spxpcb *, sizeof *cb, M_PCB, M_NOWAIT | M_ZERO);
1325
1326 if (cb == NULL) {
1327 error = ENOBUFS;
1328 goto spx_attach_end;
1329 }
1330 sb = &so->so_snd;
1331
1332 mm = m_getclr(M_DONTWAIT, MT_HEADER);
1333 if (mm == NULL) {
1334 FREE(cb, M_PCB);
1335 error = ENOBUFS;
1336 goto spx_attach_end;
1337 }
1338 cb->s_ipx = mtod(mm, struct ipx *);
1339 cb->s_state = TCPS_LISTEN;
1340 cb->s_smax = -1;
1341 cb->s_swl1 = -1;
1342 cb->s_q.si_next = cb->s_q.si_prev = &cb->s_q;
1343 cb->s_ipxpcb = ipxp;
1344 cb->s_mtu = 576 - sizeof(struct spx);
1345 cb->s_cwnd = sbspace(sb) * CUNIT / cb->s_mtu;
1346 cb->s_ssthresh = cb->s_cwnd;
1347 cb->s_cwmx = sbspace(sb) * CUNIT / (2 * sizeof(struct spx));
1348 /* Above is recomputed when connecting to account
1349 for changed buffering or mtu's */
1350 cb->s_rtt = SPXTV_SRTTBASE;
1351 cb->s_rttvar = SPXTV_SRTTDFLT << 2;
1352 SPXT_RANGESET(cb->s_rxtcur,
1353 ((SPXTV_SRTTBASE >> 2) + (SPXTV_SRTTDFLT << 2)) >> 1,
1354 SPXTV_MIN, SPXTV_REXMTMAX);
1355 ipxp->ipxp_pcb = (caddr_t)cb;
1356spx_attach_end:
1357 splx(s);
1358 return (error);
1359}
1360
1361static int
1362spx_bind(so, nam, td)
1363 struct socket *so;
1364 struct sockaddr *nam;
1365 struct thread *td;
1366{
1367 struct ipxpcb *ipxp;
1368
1369 ipxp = sotoipxpcb(so);
1370
1371 return (ipx_pcbbind(ipxp, nam, td));
1372}
1373
1374/*
1375 * Initiate connection to peer.
1376 * Enter SYN_SENT state, and mark socket as connecting.
1377 * Start keep-alive timer, setup prototype header,
1378 * Send initial system packet requesting connection.
1379 */
1380static int
1381spx_connect(so, nam, td)
1382 struct socket *so;
1383 struct sockaddr *nam;
1384 struct thread *td;
1385{
1386 int error;
1387 int s;
1388 struct ipxpcb *ipxp;
1389 struct spxpcb *cb;
1390
1391 ipxp = sotoipxpcb(so);
1392 cb = ipxtospxpcb(ipxp);
1393
1394 s = splnet();
1395 if (ipxp->ipxp_lport == 0) {
1396 error = ipx_pcbbind(ipxp, (struct sockaddr *)NULL, td);
1397 if (error)
1398 goto spx_connect_end;
1399 }
1400 error = ipx_pcbconnect(ipxp, nam, td);
1401 if (error)
1402 goto spx_connect_end;
1403 soisconnecting(so);
1404 spxstat.spxs_connattempt++;
1405 cb->s_state = TCPS_SYN_SENT;
1406 cb->s_did = 0;
1407 spx_template(cb);
1408 cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
1409 cb->s_force = 1 + SPXTV_KEEP;
1410 /*
1411 * Other party is required to respond to
1412 * the port I send from, but he is not
1413 * required to answer from where I am sending to,
1414 * so allow wildcarding.
1415 * original port I am sending to is still saved in
1416 * cb->s_dport.
1417 */
1418 ipxp->ipxp_fport = 0;
1419 error = spx_output(cb, (struct mbuf *)NULL);
1420spx_connect_end:
1421 splx(s);
1422 return (error);
1423}
1424
1425static int
1426spx_detach(so)
1427 struct socket *so;
1428{
1429 int s;
1430 struct ipxpcb *ipxp;
1431 struct spxpcb *cb;
1432
1433 ipxp = sotoipxpcb(so);
1434 cb = ipxtospxpcb(ipxp);
1435
1436 if (ipxp == NULL)
1437 return (ENOTCONN);
1438 s = splnet();
1439 if (cb->s_state > TCPS_LISTEN)
1440 spx_disconnect(cb);
1441 else
1442 spx_close(cb);
1443 splx(s);
1444 return (0);
1445}
1446
1447/*
1448 * We may decide later to implement connection closing
1449 * handshaking at the spx level optionally.
1450 * here is the hook to do it:
1451 */
1452static int
1453spx_usr_disconnect(so)
1454 struct socket *so;
1455{
1456 int s;
1457 struct ipxpcb *ipxp;
1458 struct spxpcb *cb;
1459
1460 ipxp = sotoipxpcb(so);
1461 cb = ipxtospxpcb(ipxp);
1462
1463 s = splnet();
1464 spx_disconnect(cb);
1465 splx(s);
1466 return (0);
1467}
1468
1469static int
1470spx_listen(so, td)
1471 struct socket *so;
1472 struct thread *td;
1473{
1474 int error;
1475 struct ipxpcb *ipxp;
1476 struct spxpcb *cb;
1477
1478 error = 0;
1479 ipxp = sotoipxpcb(so);
1480 cb = ipxtospxpcb(ipxp);
1481
1482 if (ipxp->ipxp_lport == 0)
1483 error = ipx_pcbbind(ipxp, (struct sockaddr *)NULL, td);
1484 if (error == 0)
1485 cb->s_state = TCPS_LISTEN;
1486 return (error);
1487}
1488
1489/*
1490 * After a receive, possibly send acknowledgment
1491 * updating allocation.
1492 */
1493static int
1494spx_rcvd(so, flags)
1495 struct socket *so;
1496 int flags;
1497{
1498 int s;
1499 struct ipxpcb *ipxp;
1500 struct spxpcb *cb;
1501
1502 ipxp = sotoipxpcb(so);
1503 cb = ipxtospxpcb(ipxp);
1504
1505 s = splnet();
1506 cb->s_flags |= SF_RVD;
1507 spx_output(cb, (struct mbuf *)NULL);
1508 cb->s_flags &= ~SF_RVD;
1509 splx(s);
1510 return (0);
1511}
1512
1513static int
1514spx_rcvoob(so, m, flags)
1515 struct socket *so;
1516 struct mbuf *m;
1517 int flags;
1518{
1519 struct ipxpcb *ipxp;
1520 struct spxpcb *cb;
1521
1522 ipxp = sotoipxpcb(so);
1523 cb = ipxtospxpcb(ipxp);
1524
1525 if ((cb->s_oobflags & SF_IOOB) || so->so_oobmark ||
1526 (so->so_rcv.sb_state & SBS_RCVATMARK)) {
1527 m->m_len = 1;
1528 *mtod(m, caddr_t) = cb->s_iobc;
1529 return (0);
1530 }
1531 return (EINVAL);
1532}
1533
1534static int
1535spx_send(so, flags, m, addr, controlp, td)
1536 struct socket *so;
1537 int flags;
1538 struct mbuf *m;
1539 struct sockaddr *addr;
1540 struct mbuf *controlp;
1541 struct thread *td;
1542{
1543 int error;
1544 int s;
1545 struct ipxpcb *ipxp;
1546 struct spxpcb *cb;
1547
1548 error = 0;
1549 ipxp = sotoipxpcb(so);
1550 cb = ipxtospxpcb(ipxp);
1551
1552 s = splnet();
1553 if (flags & PRUS_OOB) {
1554 if (sbspace(&so->so_snd) < -512) {
1555 error = ENOBUFS;
1556 goto spx_send_end;
1557 }
1558 cb->s_oobflags |= SF_SOOB;
1559 }
1560 if (controlp != NULL) {
1561 u_short *p = mtod(controlp, u_short *);
1562 spx_newchecks[2]++;
1563 if ((p[0] == 5) && (p[1] == 1)) { /* XXXX, for testing */
1564 cb->s_shdr.spx_dt = *(u_char *)(&p[2]);
1565 spx_newchecks[3]++;
1566 }
1567 m_freem(controlp);
1568 }
1569 controlp = NULL;
1570 error = spx_output(cb, m);
1571 m = NULL;
1572spx_send_end:
1573 if (controlp != NULL)
1574 m_freem(controlp);
1575 if (m != NULL)
1576 m_freem(m);
1577 splx(s);
1578 return (error);
1579}
1580
1581static int
1582spx_shutdown(so)
1583 struct socket *so;
1584{
1585 int error;
1586 int s;
1587 struct ipxpcb *ipxp;
1588 struct spxpcb *cb;
1589
1590 error = 0;
1591 ipxp = sotoipxpcb(so);
1592 cb = ipxtospxpcb(ipxp);
1593
1594 s = splnet();
1595 socantsendmore(so);
1596 cb = spx_usrclosed(cb);
1597 if (cb != NULL)
1598 error = spx_output(cb, (struct mbuf *)NULL);
1599 splx(s);
1600 return (error);
1601}
1602
1603static int
1604spx_sp_attach(so, proto, td)
1605 struct socket *so;
1606 int proto;
1607 struct thread *td;
1608{
1609 int error;
1610 struct ipxpcb *ipxp;
1611
1612 error = spx_attach(so, proto, td);
1613 if (error == 0) {
1614 ipxp = sotoipxpcb(so);
1615 ((struct spxpcb *)ipxp->ipxp_pcb)->s_flags |=
1616 (SF_HI | SF_HO | SF_PI);
1617 }
1618 return (error);
1619}
1620
1621/*
1622 * Create template to be used to send spx packets on a connection.
1623 * Called after host entry created, fills
1624 * in a skeletal spx header (choosing connection id),
1625 * minimizing the amount of work necessary when the connection is used.
1626 */
1627static void
1628spx_template(cb)
1629 register struct spxpcb *cb;
1630{
1631 register struct ipxpcb *ipxp = cb->s_ipxpcb;
1632 register struct ipx *ipx = cb->s_ipx;
1633 register struct sockbuf *sb = &(ipxp->ipxp_socket->so_snd);
1634
1635 ipx->ipx_pt = IPXPROTO_SPX;
1636 ipx->ipx_sna = ipxp->ipxp_laddr;
1637 ipx->ipx_dna = ipxp->ipxp_faddr;
1638 cb->s_sid = htons(spx_iss);
1639 spx_iss += SPX_ISSINCR/2;
1640 cb->s_alo = 1;
1641 cb->s_cwnd = (sbspace(sb) * CUNIT) / cb->s_mtu;
1642 cb->s_ssthresh = cb->s_cwnd; /* Try to expand fast to full complement
1643 of large packets */
1644 cb->s_cwmx = (sbspace(sb) * CUNIT) / (2 * sizeof(struct spx));
1645 cb->s_cwmx = max(cb->s_cwmx, cb->s_cwnd);
1646 /* But allow for lots of little packets as well */
1647}
1648
1649/*
1650 * Close a SPIP control block:
1651 * discard spx control block itself
1652 * discard ipx protocol control block
1653 * wake up any sleepers
1654 */
1655static struct spxpcb *
1656spx_close(cb)
1657 register struct spxpcb *cb;
1658{
1659 register struct spx_q *s;
1660 struct ipxpcb *ipxp = cb->s_ipxpcb;
1661 struct socket *so = ipxp->ipxp_socket;
1662 register struct mbuf *m;
1663
1664 s = cb->s_q.si_next;
1665 while (s != &(cb->s_q)) {
1666 s = s->si_next;
1667 m = dtom(s->si_prev);
1668 remque(s->si_prev);
1669 m_freem(m);
1670 }
1671 m_free(dtom(cb->s_ipx));
1672 FREE(cb, M_PCB);
1673 ipxp->ipxp_pcb = 0;
1674 soisdisconnected(so);
1675 ipx_pcbdetach(ipxp);
1676 spxstat.spxs_closed++;
1677 return ((struct spxpcb *)NULL);
1678}
1679
1680/*
1681 * Someday we may do level 3 handshaking
1682 * to close a connection or send a xerox style error.
1683 * For now, just close.
1684 */
1685static struct spxpcb *
1686spx_usrclosed(cb)
1687 register struct spxpcb *cb;
1688{
1689 return (spx_close(cb));
1690}
1691
1692static struct spxpcb *
1693spx_disconnect(cb)
1694 register struct spxpcb *cb;
1695{
1696 return (spx_close(cb));
1697}
1698
1699/*
1700 * Drop connection, reporting
1701 * the specified error.
1702 */
1703static struct spxpcb *
1704spx_drop(cb, errno)
1705 register struct spxpcb *cb;
1706 int errno;
1707{
1708 struct socket *so = cb->s_ipxpcb->ipxp_socket;
1709
1710 /*
1711 * someday, in the xerox world
1712 * we will generate error protocol packets
1713 * announcing that the socket has gone away.
1714 */
1715 if (TCPS_HAVERCVDSYN(cb->s_state)) {
1716 spxstat.spxs_drops++;
1717 cb->s_state = TCPS_CLOSED;
1718 /*tcp_output(cb);*/
1719 } else
1720 spxstat.spxs_conndrops++;
1721 so->so_error = errno;
1722 return (spx_close(cb));
1723}
1724
1725/*
1726 * Fast timeout routine for processing delayed acks
1727 */
1728void
1729spx_fasttimo()
1730{
1731 register struct ipxpcb *ipxp;
1732 register struct spxpcb *cb;
1733 int s = splnet();
1734
1735 LIST_FOREACH(ipxp, &ipxpcb_list, ipxp_list) {
1736 if ((cb = (struct spxpcb *)ipxp->ipxp_pcb) != NULL &&
1737 (cb->s_flags & SF_DELACK)) {
1738 cb->s_flags &= ~SF_DELACK;
1739 cb->s_flags |= SF_ACKNOW;
1740 spxstat.spxs_delack++;
1741 spx_output(cb, (struct mbuf *)NULL);
1742 }
1743 }
1744
1745 splx(s);
1746}
1747
1748/*
1749 * spx protocol timeout routine called every 500 ms.
1750 * Updates the timers in all active pcb's and
1751 * causes finite state machine actions if timers expire.
1752 */
1753void
1754spx_slowtimo()
1755{
1756 register struct ipxpcb *ip, *ip_temp;
1757 register struct spxpcb *cb;
1758 int s = splnet();
1759 register int i;
1760
1761 /*
1762 * Search through tcb's and update active timers. Note that timers
1763 * may free the ipxpcb, so be sure to handle that case.
1764 */
1765 LIST_FOREACH_SAFE(ip, &ipxpcb_list, ipxp_list, ip_temp) {
1766 cb = ipxtospxpcb(ip);
1767 if (cb == NULL)
1768 continue;
1769 for (i = 0; i < SPXT_NTIMERS; i++) {
1770 if (cb->s_timer[i] && --cb->s_timer[i] == 0) {
1771 /*
1772 * spx_timers() returns (NULL) if it free'd
1773 * the pcb.
1774 */
1775 if (spx_timers(cb, i) == NULL)
1776 continue;
1777 }
1778 }
1779 cb->s_idle++;
1780 if (cb->s_rtt)
1781 cb->s_rtt++;
1782 }
1783 spx_iss += SPX_ISSINCR/PR_SLOWHZ; /* increment iss */
1784 splx(s);
1785}
1786
1787/*
1788 * SPX timer processing.
1789 */
1790static struct spxpcb *
1791spx_timers(cb, timer)
1792 register struct spxpcb *cb;
1793 int timer;
1794{
1795 long rexmt;
1796 int win;
1797
1798 cb->s_force = 1 + timer;
1799 switch (timer) {
1800
1801 /*
1802 * 2 MSL timeout in shutdown went off. TCP deletes connection
1803 * control block.
1804 */
1805 case SPXT_2MSL:
1806 printf("spx: SPXT_2MSL went off for no reason\n");
1807 cb->s_timer[timer] = 0;
1808 break;
1809
1810 /*
1811 * Retransmission timer went off. Message has not
1812 * been acked within retransmit interval. Back off
1813 * to a longer retransmit interval and retransmit one packet.
1814 */
1815 case SPXT_REXMT:
1816 if (++cb->s_rxtshift > SPX_MAXRXTSHIFT) {
1817 cb->s_rxtshift = SPX_MAXRXTSHIFT;
1818 spxstat.spxs_timeoutdrop++;
1819 cb = spx_drop(cb, ETIMEDOUT);
1820 break;
1821 }
1822 spxstat.spxs_rexmttimeo++;
1823 rexmt = ((cb->s_srtt >> 2) + cb->s_rttvar) >> 1;
1824 rexmt *= spx_backoff[cb->s_rxtshift];
1825 SPXT_RANGESET(cb->s_rxtcur, rexmt, SPXTV_MIN, SPXTV_REXMTMAX);
1826 cb->s_timer[SPXT_REXMT] = cb->s_rxtcur;
1827 /*
1828 * If we have backed off fairly far, our srtt
1829 * estimate is probably bogus. Clobber it
1830 * so we'll take the next rtt measurement as our srtt;
1831 * move the current srtt into rttvar to keep the current
1832 * retransmit times until then.
1833 */
1834 if (cb->s_rxtshift > SPX_MAXRXTSHIFT / 4 ) {
1835 cb->s_rttvar += (cb->s_srtt >> 2);
1836 cb->s_srtt = 0;
1837 }
1838 cb->s_snxt = cb->s_rack;
1839 /*
1840 * If timing a packet, stop the timer.
1841 */
1842 cb->s_rtt = 0;
1843 /*
1844 * See very long discussion in tcp_timer.c about congestion
1845 * window and sstrhesh
1846 */
1847 win = min(cb->s_swnd, (cb->s_cwnd/CUNIT)) / 2;
1848 if (win < 2)
1849 win = 2;
1850 cb->s_cwnd = CUNIT;
1851 cb->s_ssthresh = win * CUNIT;
1852 spx_output(cb, (struct mbuf *)NULL);
1853 break;
1854
1855 /*
1856 * Persistance timer into zero window.
1857 * Force a probe to be sent.
1858 */
1859 case SPXT_PERSIST:
1860 spxstat.spxs_persisttimeo++;
1861 spx_setpersist(cb);
1862 spx_output(cb, (struct mbuf *)NULL);
1863 break;
1864
1865 /*
1866 * Keep-alive timer went off; send something
1867 * or drop connection if idle for too long.
1868 */
1869 case SPXT_KEEP:
1870 spxstat.spxs_keeptimeo++;
1871 if (cb->s_state < TCPS_ESTABLISHED)
1872 goto dropit;
1873 if (cb->s_ipxpcb->ipxp_socket->so_options & SO_KEEPALIVE) {
1874 if (cb->s_idle >= SPXTV_MAXIDLE)
1875 goto dropit;
1876 spxstat.spxs_keepprobe++;
1877 spx_output(cb, (struct mbuf *)NULL);
1878 } else
1879 cb->s_idle = 0;
1880 cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
1881 break;
1882 dropit:
1883 spxstat.spxs_keepdrops++;
1884 cb = spx_drop(cb, ETIMEDOUT);
1885 break;
1886 }
1887 return (cb);
1888}
2 * Copyright (c) 2004, Robert N. M. Watson
3 * Copyright (c) 1995, Mike Mitchell
4 * Copyright (c) 1984, 1985, 1986, 1987, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. All advertising materials mentioning features or use of this software
16 * must display the following acknowledgement:
17 * This product includes software developed by the University of
18 * California, Berkeley and its contributors.
19 * 4. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 *
35 * @(#)spx_usrreq.h
36 */
37
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD: head/sys/netipx/spx_usrreq.c 139444 2004-12-30 17:49:40Z rwatson $");
40
41#include <sys/param.h>
42#include <sys/lock.h>
43#include <sys/malloc.h>
44#include <sys/mbuf.h>
45#include <sys/mutex.h>
46#include <sys/proc.h>
47#include <sys/protosw.h>
48#include <sys/signalvar.h>
49#include <sys/socket.h>
50#include <sys/socketvar.h>
51#include <sys/sx.h>
52#include <sys/systm.h>
53
54#include <net/route.h>
55#include <netinet/tcp_fsm.h>
56
57#include <netipx/ipx.h>
58#include <netipx/ipx_pcb.h>
59#include <netipx/ipx_var.h>
60#include <netipx/spx.h>
61#include <netipx/spx_debug.h>
62#include <netipx/spx_timer.h>
63#include <netipx/spx_var.h>
64
65/*
66 * SPX protocol implementation.
67 */
68static u_short spx_iss;
69static u_short spx_newchecks[50];
70static int spx_hardnosed;
71static int spx_use_delack = 0;
72static int traceallspxs = 0;
73static struct spx spx_savesi;
74static struct spx_istat spx_istat;
75
76/* Following was struct spxstat spxstat; */
77#ifndef spxstat
78#define spxstat spx_istat.newstats
79#endif
80
81static const int spx_backoff[SPX_MAXRXTSHIFT+1] =
82 { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
83
84static struct spxpcb *spx_close(struct spxpcb *cb);
85static struct spxpcb *spx_disconnect(struct spxpcb *cb);
86static struct spxpcb *spx_drop(struct spxpcb *cb, int errno);
87static int spx_output(struct spxpcb *cb, struct mbuf *m0);
88static int spx_reass(struct spxpcb *cb, struct spx *si);
89static void spx_setpersist(struct spxpcb *cb);
90static void spx_template(struct spxpcb *cb);
91static struct spxpcb *spx_timers(struct spxpcb *cb, int timer);
92static struct spxpcb *spx_usrclosed(struct spxpcb *cb);
93
94static int spx_usr_abort(struct socket *so);
95static int spx_accept(struct socket *so, struct sockaddr **nam);
96static int spx_attach(struct socket *so, int proto, struct thread *td);
97static int spx_bind(struct socket *so, struct sockaddr *nam, struct thread *td);
98static int spx_connect(struct socket *so, struct sockaddr *nam,
99 struct thread *td);
100static int spx_detach(struct socket *so);
101static int spx_usr_disconnect(struct socket *so);
102static int spx_listen(struct socket *so, struct thread *td);
103static int spx_rcvd(struct socket *so, int flags);
104static int spx_rcvoob(struct socket *so, struct mbuf *m, int flags);
105static int spx_send(struct socket *so, int flags, struct mbuf *m,
106 struct sockaddr *addr, struct mbuf *control,
107 struct thread *td);
108static int spx_shutdown(struct socket *so);
109static int spx_sp_attach(struct socket *so, int proto, struct thread *td);
110
111struct pr_usrreqs spx_usrreqs = {
112 .pru_abort = spx_usr_abort,
113 .pru_accept = spx_accept,
114 .pru_attach = spx_attach,
115 .pru_bind = spx_bind,
116 .pru_connect = spx_connect,
117 .pru_control = ipx_control,
118 .pru_detach = spx_detach,
119 .pru_disconnect = spx_usr_disconnect,
120 .pru_listen = spx_listen,
121 .pru_peeraddr = ipx_peeraddr,
122 .pru_rcvd = spx_rcvd,
123 .pru_rcvoob = spx_rcvoob,
124 .pru_send = spx_send,
125 .pru_shutdown = spx_shutdown,
126 .pru_sockaddr = ipx_sockaddr,
127};
128
129struct pr_usrreqs spx_usrreq_sps = {
130 .pru_abort = spx_usr_abort,
131 .pru_accept = spx_accept,
132 .pru_attach = spx_sp_attach,
133 .pru_bind = spx_bind,
134 .pru_connect = spx_connect,
135 .pru_control = ipx_control,
136 .pru_detach = spx_detach,
137 .pru_disconnect = spx_usr_disconnect,
138 .pru_listen = spx_listen,
139 .pru_peeraddr = ipx_peeraddr,
140 .pru_rcvd = spx_rcvd,
141 .pru_rcvoob = spx_rcvoob,
142 .pru_send = spx_send,
143 .pru_shutdown = spx_shutdown,
144 .pru_sockaddr = ipx_sockaddr,
145};
146
147void
148spx_init()
149{
150
151 spx_iss = 1; /* WRONG !! should fish it out of TODR */
152}
153
154void
155spx_input(m, ipxp)
156 register struct mbuf *m;
157 register struct ipxpcb *ipxp;
158{
159 register struct spxpcb *cb;
160 register struct spx *si = mtod(m, struct spx *);
161 register struct socket *so;
162 int dropsocket = 0;
163 short ostate = 0;
164
165 spxstat.spxs_rcvtotal++;
166 if (ipxp == NULL) {
167 panic("No ipxpcb in spx_input\n");
168 return;
169 }
170
171 cb = ipxtospxpcb(ipxp);
172 if (cb == NULL)
173 goto bad;
174
175 if (m->m_len < sizeof(*si)) {
176 if ((m = m_pullup(m, sizeof(*si))) == NULL) {
177 spxstat.spxs_rcvshort++;
178 return;
179 }
180 si = mtod(m, struct spx *);
181 }
182 si->si_seq = ntohs(si->si_seq);
183 si->si_ack = ntohs(si->si_ack);
184 si->si_alo = ntohs(si->si_alo);
185
186 so = ipxp->ipxp_socket;
187
188 if (so->so_options & SO_DEBUG || traceallspxs) {
189 ostate = cb->s_state;
190 spx_savesi = *si;
191 }
192 if (so->so_options & SO_ACCEPTCONN) {
193 struct spxpcb *ocb = cb;
194
195 so = sonewconn(so, 0);
196 if (so == NULL) {
197 goto drop;
198 }
199 /*
200 * This is ugly, but ....
201 *
202 * Mark socket as temporary until we're
203 * committed to keeping it. The code at
204 * ``drop'' and ``dropwithreset'' check the
205 * flag dropsocket to see if the temporary
206 * socket created here should be discarded.
207 * We mark the socket as discardable until
208 * we're committed to it below in TCPS_LISTEN.
209 */
210 dropsocket++;
211 ipxp = (struct ipxpcb *)so->so_pcb;
212 ipxp->ipxp_laddr = si->si_dna;
213 cb = ipxtospxpcb(ipxp);
214 cb->s_mtu = ocb->s_mtu; /* preserve sockopts */
215 cb->s_flags = ocb->s_flags; /* preserve sockopts */
216 cb->s_flags2 = ocb->s_flags2; /* preserve sockopts */
217 cb->s_state = TCPS_LISTEN;
218 }
219
220 /*
221 * Packet received on connection.
222 * reset idle time and keep-alive timer;
223 */
224 cb->s_idle = 0;
225 cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
226
227 switch (cb->s_state) {
228
229 case TCPS_LISTEN:{
230 struct sockaddr_ipx *sipx, ssipx;
231 struct ipx_addr laddr;
232
233 /*
234 * If somebody here was carying on a conversation
235 * and went away, and his pen pal thinks he can
236 * still talk, we get the misdirected packet.
237 */
238 if (spx_hardnosed && (si->si_did != 0 || si->si_seq != 0)) {
239 spx_istat.gonawy++;
240 goto dropwithreset;
241 }
242 sipx = &ssipx;
243 bzero(sipx, sizeof *sipx);
244 sipx->sipx_len = sizeof(*sipx);
245 sipx->sipx_family = AF_IPX;
246 sipx->sipx_addr = si->si_sna;
247 laddr = ipxp->ipxp_laddr;
248 if (ipx_nullhost(laddr))
249 ipxp->ipxp_laddr = si->si_dna;
250 if (ipx_pcbconnect(ipxp, (struct sockaddr *)sipx, &thread0)) {
251 ipxp->ipxp_laddr = laddr;
252 spx_istat.noconn++;
253 goto drop;
254 }
255 spx_template(cb);
256 dropsocket = 0; /* committed to socket */
257 cb->s_did = si->si_sid;
258 cb->s_rack = si->si_ack;
259 cb->s_ralo = si->si_alo;
260#define THREEWAYSHAKE
261#ifdef THREEWAYSHAKE
262 cb->s_state = TCPS_SYN_RECEIVED;
263 cb->s_force = 1 + SPXT_KEEP;
264 spxstat.spxs_accepts++;
265 cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
266 }
267 break;
268 /*
269 * This state means that we have heard a response
270 * to our acceptance of their connection
271 * It is probably logically unnecessary in this
272 * implementation.
273 */
274 case TCPS_SYN_RECEIVED: {
275 if (si->si_did != cb->s_sid) {
276 spx_istat.wrncon++;
277 goto drop;
278 }
279#endif
280 ipxp->ipxp_fport = si->si_sport;
281 cb->s_timer[SPXT_REXMT] = 0;
282 cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
283 soisconnected(so);
284 cb->s_state = TCPS_ESTABLISHED;
285 spxstat.spxs_accepts++;
286 }
287 break;
288
289 /*
290 * This state means that we have gotten a response
291 * to our attempt to establish a connection.
292 * We fill in the data from the other side,
293 * telling us which port to respond to, instead of the well-
294 * known one we might have sent to in the first place.
295 * We also require that this is a response to our
296 * connection id.
297 */
298 case TCPS_SYN_SENT:
299 if (si->si_did != cb->s_sid) {
300 spx_istat.notme++;
301 goto drop;
302 }
303 spxstat.spxs_connects++;
304 cb->s_did = si->si_sid;
305 cb->s_rack = si->si_ack;
306 cb->s_ralo = si->si_alo;
307 cb->s_dport = ipxp->ipxp_fport = si->si_sport;
308 cb->s_timer[SPXT_REXMT] = 0;
309 cb->s_flags |= SF_ACKNOW;
310 soisconnected(so);
311 cb->s_state = TCPS_ESTABLISHED;
312 /* Use roundtrip time of connection request for initial rtt */
313 if (cb->s_rtt) {
314 cb->s_srtt = cb->s_rtt << 3;
315 cb->s_rttvar = cb->s_rtt << 1;
316 SPXT_RANGESET(cb->s_rxtcur,
317 ((cb->s_srtt >> 2) + cb->s_rttvar) >> 1,
318 SPXTV_MIN, SPXTV_REXMTMAX);
319 cb->s_rtt = 0;
320 }
321 }
322 if (so->so_options & SO_DEBUG || traceallspxs)
323 spx_trace(SA_INPUT, (u_char)ostate, cb, &spx_savesi, 0);
324
325 m->m_len -= sizeof(struct ipx);
326 m->m_pkthdr.len -= sizeof(struct ipx);
327 m->m_data += sizeof(struct ipx);
328
329 if (spx_reass(cb, si)) {
330 m_freem(m);
331 }
332 if (cb->s_force || (cb->s_flags & (SF_ACKNOW|SF_WIN|SF_RXT)))
333 spx_output(cb, (struct mbuf *)NULL);
334 cb->s_flags &= ~(SF_WIN|SF_RXT);
335 return;
336
337dropwithreset:
338 if (dropsocket) {
339 struct socket *head;
340 ACCEPT_LOCK();
341 KASSERT((so->so_qstate & SQ_INCOMP) != 0,
342 ("spx_input: nascent socket not SQ_INCOMP on soabort()"));
343 head = so->so_head;
344 TAILQ_REMOVE(&head->so_incomp, so, so_list);
345 head->so_incqlen--;
346 so->so_qstate &= ~SQ_INCOMP;
347 so->so_head = NULL;
348 ACCEPT_UNLOCK();
349 soabort(so);
350 }
351 si->si_seq = ntohs(si->si_seq);
352 si->si_ack = ntohs(si->si_ack);
353 si->si_alo = ntohs(si->si_alo);
354 m_freem(dtom(si));
355 if (cb->s_ipxpcb->ipxp_socket->so_options & SO_DEBUG || traceallspxs)
356 spx_trace(SA_DROP, (u_char)ostate, cb, &spx_savesi, 0);
357 return;
358
359drop:
360bad:
361 if (cb == 0 || cb->s_ipxpcb->ipxp_socket->so_options & SO_DEBUG ||
362 traceallspxs)
363 spx_trace(SA_DROP, (u_char)ostate, cb, &spx_savesi, 0);
364 m_freem(m);
365}
366
367static int spxrexmtthresh = 3;
368
369/*
370 * This is structurally similar to the tcp reassembly routine
371 * but its function is somewhat different: It merely queues
372 * packets up, and suppresses duplicates.
373 */
374static int
375spx_reass(cb, si)
376register struct spxpcb *cb;
377register struct spx *si;
378{
379 register struct spx_q *q;
380 register struct mbuf *m;
381 register struct socket *so = cb->s_ipxpcb->ipxp_socket;
382 char packetp = cb->s_flags & SF_HI;
383 int incr;
384 char wakeup = 0;
385
386 if (si == SI(0))
387 goto present;
388 /*
389 * Update our news from them.
390 */
391 if (si->si_cc & SPX_SA)
392 cb->s_flags |= (spx_use_delack ? SF_DELACK : SF_ACKNOW);
393 if (SSEQ_GT(si->si_alo, cb->s_ralo))
394 cb->s_flags |= SF_WIN;
395 if (SSEQ_LEQ(si->si_ack, cb->s_rack)) {
396 if ((si->si_cc & SPX_SP) && cb->s_rack != (cb->s_smax + 1)) {
397 spxstat.spxs_rcvdupack++;
398 /*
399 * If this is a completely duplicate ack
400 * and other conditions hold, we assume
401 * a packet has been dropped and retransmit
402 * it exactly as in tcp_input().
403 */
404 if (si->si_ack != cb->s_rack ||
405 si->si_alo != cb->s_ralo)
406 cb->s_dupacks = 0;
407 else if (++cb->s_dupacks == spxrexmtthresh) {
408 u_short onxt = cb->s_snxt;
409 int cwnd = cb->s_cwnd;
410
411 cb->s_snxt = si->si_ack;
412 cb->s_cwnd = CUNIT;
413 cb->s_force = 1 + SPXT_REXMT;
414 spx_output(cb, (struct mbuf *)NULL);
415 cb->s_timer[SPXT_REXMT] = cb->s_rxtcur;
416 cb->s_rtt = 0;
417 if (cwnd >= 4 * CUNIT)
418 cb->s_cwnd = cwnd / 2;
419 if (SSEQ_GT(onxt, cb->s_snxt))
420 cb->s_snxt = onxt;
421 return (1);
422 }
423 } else
424 cb->s_dupacks = 0;
425 goto update_window;
426 }
427 cb->s_dupacks = 0;
428 /*
429 * If our correspondent acknowledges data we haven't sent
430 * TCP would drop the packet after acking. We'll be a little
431 * more permissive
432 */
433 if (SSEQ_GT(si->si_ack, (cb->s_smax + 1))) {
434 spxstat.spxs_rcvacktoomuch++;
435 si->si_ack = cb->s_smax + 1;
436 }
437 spxstat.spxs_rcvackpack++;
438 /*
439 * If transmit timer is running and timed sequence
440 * number was acked, update smoothed round trip time.
441 * See discussion of algorithm in tcp_input.c
442 */
443 if (cb->s_rtt && SSEQ_GT(si->si_ack, cb->s_rtseq)) {
444 spxstat.spxs_rttupdated++;
445 if (cb->s_srtt != 0) {
446 register short delta;
447 delta = cb->s_rtt - (cb->s_srtt >> 3);
448 if ((cb->s_srtt += delta) <= 0)
449 cb->s_srtt = 1;
450 if (delta < 0)
451 delta = -delta;
452 delta -= (cb->s_rttvar >> 2);
453 if ((cb->s_rttvar += delta) <= 0)
454 cb->s_rttvar = 1;
455 } else {
456 /*
457 * No rtt measurement yet
458 */
459 cb->s_srtt = cb->s_rtt << 3;
460 cb->s_rttvar = cb->s_rtt << 1;
461 }
462 cb->s_rtt = 0;
463 cb->s_rxtshift = 0;
464 SPXT_RANGESET(cb->s_rxtcur,
465 ((cb->s_srtt >> 2) + cb->s_rttvar) >> 1,
466 SPXTV_MIN, SPXTV_REXMTMAX);
467 }
468 /*
469 * If all outstanding data is acked, stop retransmit
470 * timer and remember to restart (more output or persist).
471 * If there is more data to be acked, restart retransmit
472 * timer, using current (possibly backed-off) value;
473 */
474 if (si->si_ack == cb->s_smax + 1) {
475 cb->s_timer[SPXT_REXMT] = 0;
476 cb->s_flags |= SF_RXT;
477 } else if (cb->s_timer[SPXT_PERSIST] == 0)
478 cb->s_timer[SPXT_REXMT] = cb->s_rxtcur;
479 /*
480 * When new data is acked, open the congestion window.
481 * If the window gives us less than ssthresh packets
482 * in flight, open exponentially (maxseg at a time).
483 * Otherwise open linearly (maxseg^2 / cwnd at a time).
484 */
485 incr = CUNIT;
486 if (cb->s_cwnd > cb->s_ssthresh)
487 incr = max(incr * incr / cb->s_cwnd, 1);
488 cb->s_cwnd = min(cb->s_cwnd + incr, cb->s_cwmx);
489 /*
490 * Trim Acked data from output queue.
491 */
492 while ((m = so->so_snd.sb_mb) != NULL) {
493 if (SSEQ_LT((mtod(m, struct spx *))->si_seq, si->si_ack))
494 sbdroprecord(&so->so_snd);
495 else
496 break;
497 }
498 sowwakeup(so);
499 cb->s_rack = si->si_ack;
500update_window:
501 if (SSEQ_LT(cb->s_snxt, cb->s_rack))
502 cb->s_snxt = cb->s_rack;
503 if (SSEQ_LT(cb->s_swl1, si->si_seq) || ((cb->s_swl1 == si->si_seq &&
504 (SSEQ_LT(cb->s_swl2, si->si_ack))) ||
505 (cb->s_swl2 == si->si_ack && SSEQ_LT(cb->s_ralo, si->si_alo)))) {
506 /* keep track of pure window updates */
507 if ((si->si_cc & SPX_SP) && cb->s_swl2 == si->si_ack
508 && SSEQ_LT(cb->s_ralo, si->si_alo)) {
509 spxstat.spxs_rcvwinupd++;
510 spxstat.spxs_rcvdupack--;
511 }
512 cb->s_ralo = si->si_alo;
513 cb->s_swl1 = si->si_seq;
514 cb->s_swl2 = si->si_ack;
515 cb->s_swnd = (1 + si->si_alo - si->si_ack);
516 if (cb->s_swnd > cb->s_smxw)
517 cb->s_smxw = cb->s_swnd;
518 cb->s_flags |= SF_WIN;
519 }
520 /*
521 * If this packet number is higher than that which
522 * we have allocated refuse it, unless urgent
523 */
524 if (SSEQ_GT(si->si_seq, cb->s_alo)) {
525 if (si->si_cc & SPX_SP) {
526 spxstat.spxs_rcvwinprobe++;
527 return (1);
528 } else
529 spxstat.spxs_rcvpackafterwin++;
530 if (si->si_cc & SPX_OB) {
531 if (SSEQ_GT(si->si_seq, cb->s_alo + 60)) {
532 m_freem(dtom(si));
533 return (0);
534 } /* else queue this packet; */
535 } else {
536 /*register struct socket *so = cb->s_ipxpcb->ipxp_socket;
537 if (so->so_state && SS_NOFDREF) {
538 spx_close(cb);
539 } else
540 would crash system*/
541 spx_istat.notyet++;
542 m_freem(dtom(si));
543 return (0);
544 }
545 }
546 /*
547 * If this is a system packet, we don't need to
548 * queue it up, and won't update acknowledge #
549 */
550 if (si->si_cc & SPX_SP) {
551 return (1);
552 }
553 /*
554 * We have already seen this packet, so drop.
555 */
556 if (SSEQ_LT(si->si_seq, cb->s_ack)) {
557 spx_istat.bdreas++;
558 spxstat.spxs_rcvduppack++;
559 if (si->si_seq == cb->s_ack - 1)
560 spx_istat.lstdup++;
561 return (1);
562 }
563 /*
564 * Loop through all packets queued up to insert in
565 * appropriate sequence.
566 */
567 for (q = cb->s_q.si_next; q != &cb->s_q; q = q->si_next) {
568 if (si->si_seq == SI(q)->si_seq) {
569 spxstat.spxs_rcvduppack++;
570 return (1);
571 }
572 if (SSEQ_LT(si->si_seq, SI(q)->si_seq)) {
573 spxstat.spxs_rcvoopack++;
574 break;
575 }
576 }
577 insque(si, q->si_prev);
578 /*
579 * If this packet is urgent, inform process
580 */
581 if (si->si_cc & SPX_OB) {
582 cb->s_iobc = ((char *)si)[1 + sizeof(*si)];
583 sohasoutofband(so);
584 cb->s_oobflags |= SF_IOOB;
585 }
586present:
587#define SPINC sizeof(struct spxhdr)
588 /*
589 * Loop through all packets queued up to update acknowledge
590 * number, and present all acknowledged data to user;
591 * If in packet interface mode, show packet headers.
592 */
593 for (q = cb->s_q.si_next; q != &cb->s_q; q = q->si_next) {
594 if (SI(q)->si_seq == cb->s_ack) {
595 cb->s_ack++;
596 m = dtom(q);
597 if (SI(q)->si_cc & SPX_OB) {
598 cb->s_oobflags &= ~SF_IOOB;
599 SOCKBUF_LOCK(&so->so_rcv);
600 if (so->so_rcv.sb_cc)
601 so->so_oobmark = so->so_rcv.sb_cc;
602 else
603 so->so_rcv.sb_state |= SBS_RCVATMARK;
604 SOCKBUF_UNLOCK(&so->so_rcv);
605 }
606 q = q->si_prev;
607 remque(q->si_next);
608 wakeup = 1;
609 spxstat.spxs_rcvpack++;
610#ifdef SF_NEWCALL
611 if (cb->s_flags2 & SF_NEWCALL) {
612 struct spxhdr *sp = mtod(m, struct spxhdr *);
613 u_char dt = sp->spx_dt;
614 spx_newchecks[4]++;
615 if (dt != cb->s_rhdr.spx_dt) {
616 struct mbuf *mm =
617 m_getclr(M_DONTWAIT, MT_CONTROL);
618 spx_newchecks[0]++;
619 if (mm != NULL) {
620 u_short *s =
621 mtod(mm, u_short *);
622 cb->s_rhdr.spx_dt = dt;
623 mm->m_len = 5; /*XXX*/
624 s[0] = 5;
625 s[1] = 1;
626 *(u_char *)(&s[2]) = dt;
627 sbappend(&so->so_rcv, mm);
628 }
629 }
630 if (sp->spx_cc & SPX_OB) {
631 MCHTYPE(m, MT_OOBDATA);
632 spx_newchecks[1]++;
633 SOCKBUF_LOCK(&so->so_rcv);
634 so->so_oobmark = 0;
635 so->so_rcv.sb_state &= ~SBS_RCVATMARK;
636 SOCKBUF_UNLOCK(&so->so_rcv);
637 }
638 if (packetp == 0) {
639 m->m_data += SPINC;
640 m->m_len -= SPINC;
641 m->m_pkthdr.len -= SPINC;
642 }
643 if ((sp->spx_cc & SPX_EM) || packetp) {
644 sbappendrecord(&so->so_rcv, m);
645 spx_newchecks[9]++;
646 } else
647 sbappend(&so->so_rcv, m);
648 } else
649#endif
650 if (packetp) {
651 sbappendrecord(&so->so_rcv, m);
652 } else {
653 cb->s_rhdr = *mtod(m, struct spxhdr *);
654 m->m_data += SPINC;
655 m->m_len -= SPINC;
656 m->m_pkthdr.len -= SPINC;
657 sbappend(&so->so_rcv, m);
658 }
659 } else
660 break;
661 }
662 if (wakeup)
663 sorwakeup(so);
664 return (0);
665}
666
667void
668spx_ctlinput(cmd, arg_as_sa, dummy)
669 int cmd;
670 struct sockaddr *arg_as_sa; /* XXX should be swapped with dummy */
671 void *dummy;
672{
673 caddr_t arg = (/* XXX */ caddr_t)arg_as_sa;
674 struct ipx_addr *na;
675 struct sockaddr_ipx *sipx;
676
677 if (cmd < 0 || cmd >= PRC_NCMDS)
678 return;
679
680 switch (cmd) {
681
682 case PRC_ROUTEDEAD:
683 return;
684
685 case PRC_IFDOWN:
686 case PRC_HOSTDEAD:
687 case PRC_HOSTUNREACH:
688 sipx = (struct sockaddr_ipx *)arg;
689 if (sipx->sipx_family != AF_IPX)
690 return;
691 na = &sipx->sipx_addr;
692 break;
693
694 default:
695 break;
696 }
697}
698
699static int
700spx_output(cb, m0)
701 register struct spxpcb *cb;
702 struct mbuf *m0;
703{
704 struct socket *so = cb->s_ipxpcb->ipxp_socket;
705 register struct mbuf *m;
706 register struct spx *si = (struct spx *)NULL;
707 register struct sockbuf *sb = &so->so_snd;
708 int len = 0, win, rcv_win;
709 short span, off, recordp = 0;
710 u_short alo;
711 int error = 0, sendalot;
712#ifdef notdef
713 int idle;
714#endif
715 struct mbuf *mprev;
716
717 if (m0 != NULL) {
718 int mtu = cb->s_mtu;
719 int datalen;
720 /*
721 * Make sure that packet isn't too big.
722 */
723 for (m = m0; m != NULL; m = m->m_next) {
724 mprev = m;
725 len += m->m_len;
726 if (m->m_flags & M_EOR)
727 recordp = 1;
728 }
729 datalen = (cb->s_flags & SF_HO) ?
730 len - sizeof(struct spxhdr) : len;
731 if (datalen > mtu) {
732 if (cb->s_flags & SF_PI) {
733 m_freem(m0);
734 return (EMSGSIZE);
735 } else {
736 int oldEM = cb->s_cc & SPX_EM;
737
738 cb->s_cc &= ~SPX_EM;
739 while (len > mtu) {
740 /*
741 * Here we are only being called
742 * from usrreq(), so it is OK to
743 * block.
744 */
745 m = m_copym(m0, 0, mtu, M_TRYWAIT);
746 if (cb->s_flags & SF_NEWCALL) {
747 struct mbuf *mm = m;
748 spx_newchecks[7]++;
749 while (mm != NULL) {
750 mm->m_flags &= ~M_EOR;
751 mm = mm->m_next;
752 }
753 }
754 error = spx_output(cb, m);
755 if (error) {
756 cb->s_cc |= oldEM;
757 m_freem(m0);
758 return (error);
759 }
760 m_adj(m0, mtu);
761 len -= mtu;
762 }
763 cb->s_cc |= oldEM;
764 }
765 }
766 /*
767 * Force length even, by adding a "garbage byte" if
768 * necessary.
769 */
770 if (len & 1) {
771 m = mprev;
772 if (M_TRAILINGSPACE(m) >= 1)
773 m->m_len++;
774 else {
775 struct mbuf *m1 = m_get(M_DONTWAIT, MT_DATA);
776
777 if (m1 == NULL) {
778 m_freem(m0);
779 return (ENOBUFS);
780 }
781 m1->m_len = 1;
782 *(mtod(m1, u_char *)) = 0;
783 m->m_next = m1;
784 }
785 }
786 m = m_gethdr(M_DONTWAIT, MT_HEADER);
787 if (m == NULL) {
788 m_freem(m0);
789 return (ENOBUFS);
790 }
791 /*
792 * Fill in mbuf with extended SP header
793 * and addresses and length put into network format.
794 */
795 MH_ALIGN(m, sizeof(struct spx));
796 m->m_len = sizeof(struct spx);
797 m->m_next = m0;
798 si = mtod(m, struct spx *);
799 si->si_i = *cb->s_ipx;
800 si->si_s = cb->s_shdr;
801 if ((cb->s_flags & SF_PI) && (cb->s_flags & SF_HO)) {
802 register struct spxhdr *sh;
803 if (m0->m_len < sizeof(*sh)) {
804 if((m0 = m_pullup(m0, sizeof(*sh))) == NULL) {
805 m_free(m);
806 m_freem(m0);
807 return (EINVAL);
808 }
809 m->m_next = m0;
810 }
811 sh = mtod(m0, struct spxhdr *);
812 si->si_dt = sh->spx_dt;
813 si->si_cc |= sh->spx_cc & SPX_EM;
814 m0->m_len -= sizeof(*sh);
815 m0->m_data += sizeof(*sh);
816 len -= sizeof(*sh);
817 }
818 len += sizeof(*si);
819 if ((cb->s_flags2 & SF_NEWCALL) && recordp) {
820 si->si_cc |= SPX_EM;
821 spx_newchecks[8]++;
822 }
823 if (cb->s_oobflags & SF_SOOB) {
824 /*
825 * Per jqj@cornell:
826 * make sure OB packets convey exactly 1 byte.
827 * If the packet is 1 byte or larger, we
828 * have already guaranted there to be at least
829 * one garbage byte for the checksum, and
830 * extra bytes shouldn't hurt!
831 */
832 if (len > sizeof(*si)) {
833 si->si_cc |= SPX_OB;
834 len = (1 + sizeof(*si));
835 }
836 }
837 si->si_len = htons((u_short)len);
838 m->m_pkthdr.len = ((len - 1) | 1) + 1;
839 /*
840 * queue stuff up for output
841 */
842 sbappendrecord(sb, m);
843 cb->s_seq++;
844 }
845#ifdef notdef
846 idle = (cb->s_smax == (cb->s_rack - 1));
847#endif
848again:
849 sendalot = 0;
850 off = cb->s_snxt - cb->s_rack;
851 win = min(cb->s_swnd, (cb->s_cwnd / CUNIT));
852
853 /*
854 * If in persist timeout with window of 0, send a probe.
855 * Otherwise, if window is small but nonzero
856 * and timer expired, send what we can and go into
857 * transmit state.
858 */
859 if (cb->s_force == 1 + SPXT_PERSIST) {
860 if (win != 0) {
861 cb->s_timer[SPXT_PERSIST] = 0;
862 cb->s_rxtshift = 0;
863 }
864 }
865 span = cb->s_seq - cb->s_rack;
866 len = min(span, win) - off;
867
868 if (len < 0) {
869 /*
870 * Window shrank after we went into it.
871 * If window shrank to 0, cancel pending
872 * restransmission and pull s_snxt back
873 * to (closed) window. We will enter persist
874 * state below. If the widndow didn't close completely,
875 * just wait for an ACK.
876 */
877 len = 0;
878 if (win == 0) {
879 cb->s_timer[SPXT_REXMT] = 0;
880 cb->s_snxt = cb->s_rack;
881 }
882 }
883 if (len > 1)
884 sendalot = 1;
885 rcv_win = sbspace(&so->so_rcv);
886
887 /*
888 * Send if we owe peer an ACK.
889 */
890 if (cb->s_oobflags & SF_SOOB) {
891 /*
892 * must transmit this out of band packet
893 */
894 cb->s_oobflags &= ~ SF_SOOB;
895 sendalot = 1;
896 spxstat.spxs_sndurg++;
897 goto found;
898 }
899 if (cb->s_flags & SF_ACKNOW)
900 goto send;
901 if (cb->s_state < TCPS_ESTABLISHED)
902 goto send;
903 /*
904 * Silly window can't happen in spx.
905 * Code from tcp deleted.
906 */
907 if (len)
908 goto send;
909 /*
910 * Compare available window to amount of window
911 * known to peer (as advertised window less
912 * next expected input.) If the difference is at least two
913 * packets or at least 35% of the mximum possible window,
914 * then want to send a window update to peer.
915 */
916 if (rcv_win > 0) {
917 u_short delta = 1 + cb->s_alo - cb->s_ack;
918 int adv = rcv_win - (delta * cb->s_mtu);
919
920 if ((so->so_rcv.sb_cc == 0 && adv >= (2 * cb->s_mtu)) ||
921 (100 * adv / so->so_rcv.sb_hiwat >= 35)) {
922 spxstat.spxs_sndwinup++;
923 cb->s_flags |= SF_ACKNOW;
924 goto send;
925 }
926
927 }
928 /*
929 * Many comments from tcp_output.c are appropriate here
930 * including . . .
931 * If send window is too small, there is data to transmit, and no
932 * retransmit or persist is pending, then go to persist state.
933 * If nothing happens soon, send when timer expires:
934 * if window is nonzero, transmit what we can,
935 * otherwise send a probe.
936 */
937 if (so->so_snd.sb_cc && cb->s_timer[SPXT_REXMT] == 0 &&
938 cb->s_timer[SPXT_PERSIST] == 0) {
939 cb->s_rxtshift = 0;
940 spx_setpersist(cb);
941 }
942 /*
943 * No reason to send a packet, just return.
944 */
945 cb->s_outx = 1;
946 return (0);
947
948send:
949 /*
950 * Find requested packet.
951 */
952 si = 0;
953 if (len > 0) {
954 cb->s_want = cb->s_snxt;
955 for (m = sb->sb_mb; m != NULL; m = m->m_act) {
956 si = mtod(m, struct spx *);
957 if (SSEQ_LEQ(cb->s_snxt, si->si_seq))
958 break;
959 }
960 found:
961 if (si != NULL) {
962 if (si->si_seq == cb->s_snxt)
963 cb->s_snxt++;
964 else
965 spxstat.spxs_sndvoid++, si = 0;
966 }
967 }
968 /*
969 * update window
970 */
971 if (rcv_win < 0)
972 rcv_win = 0;
973 alo = cb->s_ack - 1 + (rcv_win / ((short)cb->s_mtu));
974 if (SSEQ_LT(alo, cb->s_alo))
975 alo = cb->s_alo;
976
977 if (si != NULL) {
978 /*
979 * must make a copy of this packet for
980 * ipx_output to monkey with
981 */
982 m = m_copy(dtom(si), 0, (int)M_COPYALL);
983 if (m == NULL) {
984 return (ENOBUFS);
985 }
986 si = mtod(m, struct spx *);
987 if (SSEQ_LT(si->si_seq, cb->s_smax))
988 spxstat.spxs_sndrexmitpack++;
989 else
990 spxstat.spxs_sndpack++;
991 } else if (cb->s_force || cb->s_flags & SF_ACKNOW) {
992 /*
993 * Must send an acknowledgement or a probe
994 */
995 if (cb->s_force)
996 spxstat.spxs_sndprobe++;
997 if (cb->s_flags & SF_ACKNOW)
998 spxstat.spxs_sndacks++;
999 m = m_gethdr(M_DONTWAIT, MT_HEADER);
1000 if (m == NULL)
1001 return (ENOBUFS);
1002 /*
1003 * Fill in mbuf with extended SP header
1004 * and addresses and length put into network format.
1005 */
1006 MH_ALIGN(m, sizeof(struct spx));
1007 m->m_len = sizeof(*si);
1008 m->m_pkthdr.len = sizeof(*si);
1009 si = mtod(m, struct spx *);
1010 si->si_i = *cb->s_ipx;
1011 si->si_s = cb->s_shdr;
1012 si->si_seq = cb->s_smax + 1;
1013 si->si_len = htons(sizeof(*si));
1014 si->si_cc |= SPX_SP;
1015 } else {
1016 cb->s_outx = 3;
1017 if (so->so_options & SO_DEBUG || traceallspxs)
1018 spx_trace(SA_OUTPUT, cb->s_state, cb, si, 0);
1019 return (0);
1020 }
1021 /*
1022 * Stuff checksum and output datagram.
1023 */
1024 if ((si->si_cc & SPX_SP) == 0) {
1025 if (cb->s_force != (1 + SPXT_PERSIST) ||
1026 cb->s_timer[SPXT_PERSIST] == 0) {
1027 /*
1028 * If this is a new packet and we are not currently
1029 * timing anything, time this one.
1030 */
1031 if (SSEQ_LT(cb->s_smax, si->si_seq)) {
1032 cb->s_smax = si->si_seq;
1033 if (cb->s_rtt == 0) {
1034 spxstat.spxs_segstimed++;
1035 cb->s_rtseq = si->si_seq;
1036 cb->s_rtt = 1;
1037 }
1038 }
1039 /*
1040 * Set rexmt timer if not currently set,
1041 * Initial value for retransmit timer is smoothed
1042 * round-trip time + 2 * round-trip time variance.
1043 * Initialize shift counter which is used for backoff
1044 * of retransmit time.
1045 */
1046 if (cb->s_timer[SPXT_REXMT] == 0 &&
1047 cb->s_snxt != cb->s_rack) {
1048 cb->s_timer[SPXT_REXMT] = cb->s_rxtcur;
1049 if (cb->s_timer[SPXT_PERSIST]) {
1050 cb->s_timer[SPXT_PERSIST] = 0;
1051 cb->s_rxtshift = 0;
1052 }
1053 }
1054 } else if (SSEQ_LT(cb->s_smax, si->si_seq)) {
1055 cb->s_smax = si->si_seq;
1056 }
1057 } else if (cb->s_state < TCPS_ESTABLISHED) {
1058 if (cb->s_rtt == 0)
1059 cb->s_rtt = 1; /* Time initial handshake */
1060 if (cb->s_timer[SPXT_REXMT] == 0)
1061 cb->s_timer[SPXT_REXMT] = cb->s_rxtcur;
1062 }
1063 {
1064 /*
1065 * Do not request acks when we ack their data packets or
1066 * when we do a gratuitous window update.
1067 */
1068 if (((si->si_cc & SPX_SP) == 0) || cb->s_force)
1069 si->si_cc |= SPX_SA;
1070 si->si_seq = htons(si->si_seq);
1071 si->si_alo = htons(alo);
1072 si->si_ack = htons(cb->s_ack);
1073
1074 if (ipxcksum) {
1075 si->si_sum = ipx_cksum(m, ntohs(si->si_len));
1076 } else
1077 si->si_sum = 0xffff;
1078
1079 cb->s_outx = 4;
1080 if (so->so_options & SO_DEBUG || traceallspxs)
1081 spx_trace(SA_OUTPUT, cb->s_state, cb, si, 0);
1082
1083 if (so->so_options & SO_DONTROUTE)
1084 error = ipx_outputfl(m, (struct route *)NULL, IPX_ROUTETOIF);
1085 else
1086 error = ipx_outputfl(m, &cb->s_ipxpcb->ipxp_route, 0);
1087 }
1088 if (error) {
1089 return (error);
1090 }
1091 spxstat.spxs_sndtotal++;
1092 /*
1093 * Data sent (as far as we can tell).
1094 * If this advertises a larger window than any other segment,
1095 * then remember the size of the advertized window.
1096 * Any pending ACK has now been sent.
1097 */
1098 cb->s_force = 0;
1099 cb->s_flags &= ~(SF_ACKNOW|SF_DELACK);
1100 if (SSEQ_GT(alo, cb->s_alo))
1101 cb->s_alo = alo;
1102 if (sendalot)
1103 goto again;
1104 cb->s_outx = 5;
1105 return (0);
1106}
1107
1108static int spx_do_persist_panics = 0;
1109
1110static void
1111spx_setpersist(cb)
1112 register struct spxpcb *cb;
1113{
1114 register int t = ((cb->s_srtt >> 2) + cb->s_rttvar) >> 1;
1115
1116 if (cb->s_timer[SPXT_REXMT] && spx_do_persist_panics)
1117 panic("spx_output REXMT");
1118 /*
1119 * Start/restart persistance timer.
1120 */
1121 SPXT_RANGESET(cb->s_timer[SPXT_PERSIST],
1122 t*spx_backoff[cb->s_rxtshift],
1123 SPXTV_PERSMIN, SPXTV_PERSMAX);
1124 if (cb->s_rxtshift < SPX_MAXRXTSHIFT)
1125 cb->s_rxtshift++;
1126}
1127
1128int
1129spx_ctloutput(so, sopt)
1130 struct socket *so;
1131 struct sockopt *sopt;
1132{
1133 struct ipxpcb *ipxp = sotoipxpcb(so);
1134 register struct spxpcb *cb;
1135 int mask, error;
1136 short soptval;
1137 u_short usoptval;
1138 int optval;
1139
1140 error = 0;
1141
1142 if (sopt->sopt_level != IPXPROTO_SPX) {
1143 /* This will have to be changed when we do more general
1144 stacking of protocols */
1145 return (ipx_ctloutput(so, sopt));
1146 }
1147 if (ipxp == NULL)
1148 return (EINVAL);
1149 else
1150 cb = ipxtospxpcb(ipxp);
1151
1152 switch (sopt->sopt_dir) {
1153 case SOPT_GET:
1154 switch (sopt->sopt_name) {
1155 case SO_HEADERS_ON_INPUT:
1156 mask = SF_HI;
1157 goto get_flags;
1158
1159 case SO_HEADERS_ON_OUTPUT:
1160 mask = SF_HO;
1161 get_flags:
1162 soptval = cb->s_flags & mask;
1163 error = sooptcopyout(sopt, &soptval, sizeof soptval);
1164 break;
1165
1166 case SO_MTU:
1167 usoptval = cb->s_mtu;
1168 error = sooptcopyout(sopt, &usoptval, sizeof usoptval);
1169 break;
1170
1171 case SO_LAST_HEADER:
1172 error = sooptcopyout(sopt, &cb->s_rhdr,
1173 sizeof cb->s_rhdr);
1174 break;
1175
1176 case SO_DEFAULT_HEADERS:
1177 error = sooptcopyout(sopt, &cb->s_shdr,
1178 sizeof cb->s_shdr);
1179 break;
1180
1181 default:
1182 error = ENOPROTOOPT;
1183 }
1184 break;
1185
1186 case SOPT_SET:
1187 switch (sopt->sopt_name) {
1188 /* XXX why are these shorts on get and ints on set?
1189 that doesn't make any sense... */
1190 case SO_HEADERS_ON_INPUT:
1191 mask = SF_HI;
1192 goto set_head;
1193
1194 case SO_HEADERS_ON_OUTPUT:
1195 mask = SF_HO;
1196 set_head:
1197 error = sooptcopyin(sopt, &optval, sizeof optval,
1198 sizeof optval);
1199 if (error)
1200 break;
1201
1202 if (cb->s_flags & SF_PI) {
1203 if (optval)
1204 cb->s_flags |= mask;
1205 else
1206 cb->s_flags &= ~mask;
1207 } else error = EINVAL;
1208 break;
1209
1210 case SO_MTU:
1211 error = sooptcopyin(sopt, &usoptval, sizeof usoptval,
1212 sizeof usoptval);
1213 if (error)
1214 break;
1215 cb->s_mtu = usoptval;
1216 break;
1217
1218#ifdef SF_NEWCALL
1219 case SO_NEWCALL:
1220 error = sooptcopyin(sopt, &optval, sizeof optval,
1221 sizeof optval);
1222 if (error)
1223 break;
1224 if (optval) {
1225 cb->s_flags2 |= SF_NEWCALL;
1226 spx_newchecks[5]++;
1227 } else {
1228 cb->s_flags2 &= ~SF_NEWCALL;
1229 spx_newchecks[6]++;
1230 }
1231 break;
1232#endif
1233
1234 case SO_DEFAULT_HEADERS:
1235 {
1236 struct spxhdr sp;
1237
1238 error = sooptcopyin(sopt, &sp, sizeof sp,
1239 sizeof sp);
1240 if (error)
1241 break;
1242 cb->s_dt = sp.spx_dt;
1243 cb->s_cc = sp.spx_cc & SPX_EM;
1244 }
1245 break;
1246
1247 default:
1248 error = ENOPROTOOPT;
1249 }
1250 break;
1251 }
1252 return (error);
1253}
1254
1255static int
1256spx_usr_abort(so)
1257 struct socket *so;
1258{
1259 int s;
1260 struct ipxpcb *ipxp;
1261 struct spxpcb *cb;
1262
1263 ipxp = sotoipxpcb(so);
1264 cb = ipxtospxpcb(ipxp);
1265
1266 s = splnet();
1267 spx_drop(cb, ECONNABORTED);
1268 splx(s);
1269 return (0);
1270}
1271
1272/*
1273 * Accept a connection. Essentially all the work is
1274 * done at higher levels; just return the address
1275 * of the peer, storing through addr.
1276 */
1277static int
1278spx_accept(so, nam)
1279 struct socket *so;
1280 struct sockaddr **nam;
1281{
1282 struct ipxpcb *ipxp;
1283 struct sockaddr_ipx *sipx, ssipx;
1284
1285 ipxp = sotoipxpcb(so);
1286 sipx = &ssipx;
1287 bzero(sipx, sizeof *sipx);
1288 sipx->sipx_len = sizeof *sipx;
1289 sipx->sipx_family = AF_IPX;
1290 sipx->sipx_addr = ipxp->ipxp_faddr;
1291 *nam = sodupsockaddr((struct sockaddr *)sipx, M_NOWAIT);
1292 return (0);
1293}
1294
1295static int
1296spx_attach(so, proto, td)
1297 struct socket *so;
1298 int proto;
1299 struct thread *td;
1300{
1301 int error;
1302 int s;
1303 struct ipxpcb *ipxp;
1304 struct spxpcb *cb;
1305 struct mbuf *mm;
1306 struct sockbuf *sb;
1307
1308 ipxp = sotoipxpcb(so);
1309 cb = ipxtospxpcb(ipxp);
1310
1311 if (ipxp != NULL)
1312 return (EISCONN);
1313 s = splnet();
1314 error = ipx_pcballoc(so, &ipxpcb_list, td);
1315 if (error)
1316 goto spx_attach_end;
1317 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1318 error = soreserve(so, (u_long) 3072, (u_long) 3072);
1319 if (error)
1320 goto spx_attach_end;
1321 }
1322 ipxp = sotoipxpcb(so);
1323
1324 MALLOC(cb, struct spxpcb *, sizeof *cb, M_PCB, M_NOWAIT | M_ZERO);
1325
1326 if (cb == NULL) {
1327 error = ENOBUFS;
1328 goto spx_attach_end;
1329 }
1330 sb = &so->so_snd;
1331
1332 mm = m_getclr(M_DONTWAIT, MT_HEADER);
1333 if (mm == NULL) {
1334 FREE(cb, M_PCB);
1335 error = ENOBUFS;
1336 goto spx_attach_end;
1337 }
1338 cb->s_ipx = mtod(mm, struct ipx *);
1339 cb->s_state = TCPS_LISTEN;
1340 cb->s_smax = -1;
1341 cb->s_swl1 = -1;
1342 cb->s_q.si_next = cb->s_q.si_prev = &cb->s_q;
1343 cb->s_ipxpcb = ipxp;
1344 cb->s_mtu = 576 - sizeof(struct spx);
1345 cb->s_cwnd = sbspace(sb) * CUNIT / cb->s_mtu;
1346 cb->s_ssthresh = cb->s_cwnd;
1347 cb->s_cwmx = sbspace(sb) * CUNIT / (2 * sizeof(struct spx));
1348 /* Above is recomputed when connecting to account
1349 for changed buffering or mtu's */
1350 cb->s_rtt = SPXTV_SRTTBASE;
1351 cb->s_rttvar = SPXTV_SRTTDFLT << 2;
1352 SPXT_RANGESET(cb->s_rxtcur,
1353 ((SPXTV_SRTTBASE >> 2) + (SPXTV_SRTTDFLT << 2)) >> 1,
1354 SPXTV_MIN, SPXTV_REXMTMAX);
1355 ipxp->ipxp_pcb = (caddr_t)cb;
1356spx_attach_end:
1357 splx(s);
1358 return (error);
1359}
1360
1361static int
1362spx_bind(so, nam, td)
1363 struct socket *so;
1364 struct sockaddr *nam;
1365 struct thread *td;
1366{
1367 struct ipxpcb *ipxp;
1368
1369 ipxp = sotoipxpcb(so);
1370
1371 return (ipx_pcbbind(ipxp, nam, td));
1372}
1373
1374/*
1375 * Initiate connection to peer.
1376 * Enter SYN_SENT state, and mark socket as connecting.
1377 * Start keep-alive timer, setup prototype header,
1378 * Send initial system packet requesting connection.
1379 */
1380static int
1381spx_connect(so, nam, td)
1382 struct socket *so;
1383 struct sockaddr *nam;
1384 struct thread *td;
1385{
1386 int error;
1387 int s;
1388 struct ipxpcb *ipxp;
1389 struct spxpcb *cb;
1390
1391 ipxp = sotoipxpcb(so);
1392 cb = ipxtospxpcb(ipxp);
1393
1394 s = splnet();
1395 if (ipxp->ipxp_lport == 0) {
1396 error = ipx_pcbbind(ipxp, (struct sockaddr *)NULL, td);
1397 if (error)
1398 goto spx_connect_end;
1399 }
1400 error = ipx_pcbconnect(ipxp, nam, td);
1401 if (error)
1402 goto spx_connect_end;
1403 soisconnecting(so);
1404 spxstat.spxs_connattempt++;
1405 cb->s_state = TCPS_SYN_SENT;
1406 cb->s_did = 0;
1407 spx_template(cb);
1408 cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
1409 cb->s_force = 1 + SPXTV_KEEP;
1410 /*
1411 * Other party is required to respond to
1412 * the port I send from, but he is not
1413 * required to answer from where I am sending to,
1414 * so allow wildcarding.
1415 * original port I am sending to is still saved in
1416 * cb->s_dport.
1417 */
1418 ipxp->ipxp_fport = 0;
1419 error = spx_output(cb, (struct mbuf *)NULL);
1420spx_connect_end:
1421 splx(s);
1422 return (error);
1423}
1424
1425static int
1426spx_detach(so)
1427 struct socket *so;
1428{
1429 int s;
1430 struct ipxpcb *ipxp;
1431 struct spxpcb *cb;
1432
1433 ipxp = sotoipxpcb(so);
1434 cb = ipxtospxpcb(ipxp);
1435
1436 if (ipxp == NULL)
1437 return (ENOTCONN);
1438 s = splnet();
1439 if (cb->s_state > TCPS_LISTEN)
1440 spx_disconnect(cb);
1441 else
1442 spx_close(cb);
1443 splx(s);
1444 return (0);
1445}
1446
1447/*
1448 * We may decide later to implement connection closing
1449 * handshaking at the spx level optionally.
1450 * here is the hook to do it:
1451 */
1452static int
1453spx_usr_disconnect(so)
1454 struct socket *so;
1455{
1456 int s;
1457 struct ipxpcb *ipxp;
1458 struct spxpcb *cb;
1459
1460 ipxp = sotoipxpcb(so);
1461 cb = ipxtospxpcb(ipxp);
1462
1463 s = splnet();
1464 spx_disconnect(cb);
1465 splx(s);
1466 return (0);
1467}
1468
1469static int
1470spx_listen(so, td)
1471 struct socket *so;
1472 struct thread *td;
1473{
1474 int error;
1475 struct ipxpcb *ipxp;
1476 struct spxpcb *cb;
1477
1478 error = 0;
1479 ipxp = sotoipxpcb(so);
1480 cb = ipxtospxpcb(ipxp);
1481
1482 if (ipxp->ipxp_lport == 0)
1483 error = ipx_pcbbind(ipxp, (struct sockaddr *)NULL, td);
1484 if (error == 0)
1485 cb->s_state = TCPS_LISTEN;
1486 return (error);
1487}
1488
1489/*
1490 * After a receive, possibly send acknowledgment
1491 * updating allocation.
1492 */
1493static int
1494spx_rcvd(so, flags)
1495 struct socket *so;
1496 int flags;
1497{
1498 int s;
1499 struct ipxpcb *ipxp;
1500 struct spxpcb *cb;
1501
1502 ipxp = sotoipxpcb(so);
1503 cb = ipxtospxpcb(ipxp);
1504
1505 s = splnet();
1506 cb->s_flags |= SF_RVD;
1507 spx_output(cb, (struct mbuf *)NULL);
1508 cb->s_flags &= ~SF_RVD;
1509 splx(s);
1510 return (0);
1511}
1512
1513static int
1514spx_rcvoob(so, m, flags)
1515 struct socket *so;
1516 struct mbuf *m;
1517 int flags;
1518{
1519 struct ipxpcb *ipxp;
1520 struct spxpcb *cb;
1521
1522 ipxp = sotoipxpcb(so);
1523 cb = ipxtospxpcb(ipxp);
1524
1525 if ((cb->s_oobflags & SF_IOOB) || so->so_oobmark ||
1526 (so->so_rcv.sb_state & SBS_RCVATMARK)) {
1527 m->m_len = 1;
1528 *mtod(m, caddr_t) = cb->s_iobc;
1529 return (0);
1530 }
1531 return (EINVAL);
1532}
1533
1534static int
1535spx_send(so, flags, m, addr, controlp, td)
1536 struct socket *so;
1537 int flags;
1538 struct mbuf *m;
1539 struct sockaddr *addr;
1540 struct mbuf *controlp;
1541 struct thread *td;
1542{
1543 int error;
1544 int s;
1545 struct ipxpcb *ipxp;
1546 struct spxpcb *cb;
1547
1548 error = 0;
1549 ipxp = sotoipxpcb(so);
1550 cb = ipxtospxpcb(ipxp);
1551
1552 s = splnet();
1553 if (flags & PRUS_OOB) {
1554 if (sbspace(&so->so_snd) < -512) {
1555 error = ENOBUFS;
1556 goto spx_send_end;
1557 }
1558 cb->s_oobflags |= SF_SOOB;
1559 }
1560 if (controlp != NULL) {
1561 u_short *p = mtod(controlp, u_short *);
1562 spx_newchecks[2]++;
1563 if ((p[0] == 5) && (p[1] == 1)) { /* XXXX, for testing */
1564 cb->s_shdr.spx_dt = *(u_char *)(&p[2]);
1565 spx_newchecks[3]++;
1566 }
1567 m_freem(controlp);
1568 }
1569 controlp = NULL;
1570 error = spx_output(cb, m);
1571 m = NULL;
1572spx_send_end:
1573 if (controlp != NULL)
1574 m_freem(controlp);
1575 if (m != NULL)
1576 m_freem(m);
1577 splx(s);
1578 return (error);
1579}
1580
1581static int
1582spx_shutdown(so)
1583 struct socket *so;
1584{
1585 int error;
1586 int s;
1587 struct ipxpcb *ipxp;
1588 struct spxpcb *cb;
1589
1590 error = 0;
1591 ipxp = sotoipxpcb(so);
1592 cb = ipxtospxpcb(ipxp);
1593
1594 s = splnet();
1595 socantsendmore(so);
1596 cb = spx_usrclosed(cb);
1597 if (cb != NULL)
1598 error = spx_output(cb, (struct mbuf *)NULL);
1599 splx(s);
1600 return (error);
1601}
1602
1603static int
1604spx_sp_attach(so, proto, td)
1605 struct socket *so;
1606 int proto;
1607 struct thread *td;
1608{
1609 int error;
1610 struct ipxpcb *ipxp;
1611
1612 error = spx_attach(so, proto, td);
1613 if (error == 0) {
1614 ipxp = sotoipxpcb(so);
1615 ((struct spxpcb *)ipxp->ipxp_pcb)->s_flags |=
1616 (SF_HI | SF_HO | SF_PI);
1617 }
1618 return (error);
1619}
1620
1621/*
1622 * Create template to be used to send spx packets on a connection.
1623 * Called after host entry created, fills
1624 * in a skeletal spx header (choosing connection id),
1625 * minimizing the amount of work necessary when the connection is used.
1626 */
1627static void
1628spx_template(cb)
1629 register struct spxpcb *cb;
1630{
1631 register struct ipxpcb *ipxp = cb->s_ipxpcb;
1632 register struct ipx *ipx = cb->s_ipx;
1633 register struct sockbuf *sb = &(ipxp->ipxp_socket->so_snd);
1634
1635 ipx->ipx_pt = IPXPROTO_SPX;
1636 ipx->ipx_sna = ipxp->ipxp_laddr;
1637 ipx->ipx_dna = ipxp->ipxp_faddr;
1638 cb->s_sid = htons(spx_iss);
1639 spx_iss += SPX_ISSINCR/2;
1640 cb->s_alo = 1;
1641 cb->s_cwnd = (sbspace(sb) * CUNIT) / cb->s_mtu;
1642 cb->s_ssthresh = cb->s_cwnd; /* Try to expand fast to full complement
1643 of large packets */
1644 cb->s_cwmx = (sbspace(sb) * CUNIT) / (2 * sizeof(struct spx));
1645 cb->s_cwmx = max(cb->s_cwmx, cb->s_cwnd);
1646 /* But allow for lots of little packets as well */
1647}
1648
1649/*
1650 * Close a SPIP control block:
1651 * discard spx control block itself
1652 * discard ipx protocol control block
1653 * wake up any sleepers
1654 */
1655static struct spxpcb *
1656spx_close(cb)
1657 register struct spxpcb *cb;
1658{
1659 register struct spx_q *s;
1660 struct ipxpcb *ipxp = cb->s_ipxpcb;
1661 struct socket *so = ipxp->ipxp_socket;
1662 register struct mbuf *m;
1663
1664 s = cb->s_q.si_next;
1665 while (s != &(cb->s_q)) {
1666 s = s->si_next;
1667 m = dtom(s->si_prev);
1668 remque(s->si_prev);
1669 m_freem(m);
1670 }
1671 m_free(dtom(cb->s_ipx));
1672 FREE(cb, M_PCB);
1673 ipxp->ipxp_pcb = 0;
1674 soisdisconnected(so);
1675 ipx_pcbdetach(ipxp);
1676 spxstat.spxs_closed++;
1677 return ((struct spxpcb *)NULL);
1678}
1679
1680/*
1681 * Someday we may do level 3 handshaking
1682 * to close a connection or send a xerox style error.
1683 * For now, just close.
1684 */
1685static struct spxpcb *
1686spx_usrclosed(cb)
1687 register struct spxpcb *cb;
1688{
1689 return (spx_close(cb));
1690}
1691
1692static struct spxpcb *
1693spx_disconnect(cb)
1694 register struct spxpcb *cb;
1695{
1696 return (spx_close(cb));
1697}
1698
1699/*
1700 * Drop connection, reporting
1701 * the specified error.
1702 */
1703static struct spxpcb *
1704spx_drop(cb, errno)
1705 register struct spxpcb *cb;
1706 int errno;
1707{
1708 struct socket *so = cb->s_ipxpcb->ipxp_socket;
1709
1710 /*
1711 * someday, in the xerox world
1712 * we will generate error protocol packets
1713 * announcing that the socket has gone away.
1714 */
1715 if (TCPS_HAVERCVDSYN(cb->s_state)) {
1716 spxstat.spxs_drops++;
1717 cb->s_state = TCPS_CLOSED;
1718 /*tcp_output(cb);*/
1719 } else
1720 spxstat.spxs_conndrops++;
1721 so->so_error = errno;
1722 return (spx_close(cb));
1723}
1724
1725/*
1726 * Fast timeout routine for processing delayed acks
1727 */
1728void
1729spx_fasttimo()
1730{
1731 register struct ipxpcb *ipxp;
1732 register struct spxpcb *cb;
1733 int s = splnet();
1734
1735 LIST_FOREACH(ipxp, &ipxpcb_list, ipxp_list) {
1736 if ((cb = (struct spxpcb *)ipxp->ipxp_pcb) != NULL &&
1737 (cb->s_flags & SF_DELACK)) {
1738 cb->s_flags &= ~SF_DELACK;
1739 cb->s_flags |= SF_ACKNOW;
1740 spxstat.spxs_delack++;
1741 spx_output(cb, (struct mbuf *)NULL);
1742 }
1743 }
1744
1745 splx(s);
1746}
1747
1748/*
1749 * spx protocol timeout routine called every 500 ms.
1750 * Updates the timers in all active pcb's and
1751 * causes finite state machine actions if timers expire.
1752 */
1753void
1754spx_slowtimo()
1755{
1756 register struct ipxpcb *ip, *ip_temp;
1757 register struct spxpcb *cb;
1758 int s = splnet();
1759 register int i;
1760
1761 /*
1762 * Search through tcb's and update active timers. Note that timers
1763 * may free the ipxpcb, so be sure to handle that case.
1764 */
1765 LIST_FOREACH_SAFE(ip, &ipxpcb_list, ipxp_list, ip_temp) {
1766 cb = ipxtospxpcb(ip);
1767 if (cb == NULL)
1768 continue;
1769 for (i = 0; i < SPXT_NTIMERS; i++) {
1770 if (cb->s_timer[i] && --cb->s_timer[i] == 0) {
1771 /*
1772 * spx_timers() returns (NULL) if it free'd
1773 * the pcb.
1774 */
1775 if (spx_timers(cb, i) == NULL)
1776 continue;
1777 }
1778 }
1779 cb->s_idle++;
1780 if (cb->s_rtt)
1781 cb->s_rtt++;
1782 }
1783 spx_iss += SPX_ISSINCR/PR_SLOWHZ; /* increment iss */
1784 splx(s);
1785}
1786
1787/*
1788 * SPX timer processing.
1789 */
1790static struct spxpcb *
1791spx_timers(cb, timer)
1792 register struct spxpcb *cb;
1793 int timer;
1794{
1795 long rexmt;
1796 int win;
1797
1798 cb->s_force = 1 + timer;
1799 switch (timer) {
1800
1801 /*
1802 * 2 MSL timeout in shutdown went off. TCP deletes connection
1803 * control block.
1804 */
1805 case SPXT_2MSL:
1806 printf("spx: SPXT_2MSL went off for no reason\n");
1807 cb->s_timer[timer] = 0;
1808 break;
1809
1810 /*
1811 * Retransmission timer went off. Message has not
1812 * been acked within retransmit interval. Back off
1813 * to a longer retransmit interval and retransmit one packet.
1814 */
1815 case SPXT_REXMT:
1816 if (++cb->s_rxtshift > SPX_MAXRXTSHIFT) {
1817 cb->s_rxtshift = SPX_MAXRXTSHIFT;
1818 spxstat.spxs_timeoutdrop++;
1819 cb = spx_drop(cb, ETIMEDOUT);
1820 break;
1821 }
1822 spxstat.spxs_rexmttimeo++;
1823 rexmt = ((cb->s_srtt >> 2) + cb->s_rttvar) >> 1;
1824 rexmt *= spx_backoff[cb->s_rxtshift];
1825 SPXT_RANGESET(cb->s_rxtcur, rexmt, SPXTV_MIN, SPXTV_REXMTMAX);
1826 cb->s_timer[SPXT_REXMT] = cb->s_rxtcur;
1827 /*
1828 * If we have backed off fairly far, our srtt
1829 * estimate is probably bogus. Clobber it
1830 * so we'll take the next rtt measurement as our srtt;
1831 * move the current srtt into rttvar to keep the current
1832 * retransmit times until then.
1833 */
1834 if (cb->s_rxtshift > SPX_MAXRXTSHIFT / 4 ) {
1835 cb->s_rttvar += (cb->s_srtt >> 2);
1836 cb->s_srtt = 0;
1837 }
1838 cb->s_snxt = cb->s_rack;
1839 /*
1840 * If timing a packet, stop the timer.
1841 */
1842 cb->s_rtt = 0;
1843 /*
1844 * See very long discussion in tcp_timer.c about congestion
1845 * window and sstrhesh
1846 */
1847 win = min(cb->s_swnd, (cb->s_cwnd/CUNIT)) / 2;
1848 if (win < 2)
1849 win = 2;
1850 cb->s_cwnd = CUNIT;
1851 cb->s_ssthresh = win * CUNIT;
1852 spx_output(cb, (struct mbuf *)NULL);
1853 break;
1854
1855 /*
1856 * Persistance timer into zero window.
1857 * Force a probe to be sent.
1858 */
1859 case SPXT_PERSIST:
1860 spxstat.spxs_persisttimeo++;
1861 spx_setpersist(cb);
1862 spx_output(cb, (struct mbuf *)NULL);
1863 break;
1864
1865 /*
1866 * Keep-alive timer went off; send something
1867 * or drop connection if idle for too long.
1868 */
1869 case SPXT_KEEP:
1870 spxstat.spxs_keeptimeo++;
1871 if (cb->s_state < TCPS_ESTABLISHED)
1872 goto dropit;
1873 if (cb->s_ipxpcb->ipxp_socket->so_options & SO_KEEPALIVE) {
1874 if (cb->s_idle >= SPXTV_MAXIDLE)
1875 goto dropit;
1876 spxstat.spxs_keepprobe++;
1877 spx_output(cb, (struct mbuf *)NULL);
1878 } else
1879 cb->s_idle = 0;
1880 cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
1881 break;
1882 dropit:
1883 spxstat.spxs_keepdrops++;
1884 cb = spx_drop(cb, ETIMEDOUT);
1885 break;
1886 }
1887 return (cb);
1888}