[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ]

FreeBSD/Linux Kernel Cross Reference
sys/netipx/spx_usrreq.c

Version: -  FREEBSD  -  FREEBSD7  -  FREEBSD70  -  FREEBSD6  -  FREEBSD63  -  FREEBSD62  -  FREEBSD61  -  FREEBSD60  -  FREEBSD5  -  FREEBSD55  -  FREEBSD54  -  FREEBSD53  -  FREEBSD52  -  FREEBSD51  -  FREEBSD50  -  FREEBSD4  -  FREEBSD3  -  FREEBSD22  -  linux-2.6  -  linux-2.4.22  -  MK83  -  MK84  -  PLAN9  -  DFBSD  -  NETBSD  -  NETBSD4  -  NETBSD3  -  NETBSD20  -  OPENBSD  -  xnu-517  -  xnu-792  -  xnu-792.6.70  -  xnu-1228  -  OPENSOLARIS  -  minix-3-1-1  -  TRUSTEDBSD-SEBSD  -  FREEBSD-LIBC  -  FREEBSD7-LIBC  -  FREEBSD6-LIBC  -  GLIBC27 
SearchContext: -  none  -  excerpts  -  bigexcerpts 

  1 /*-
  2  * Copyright (c) 1984, 1985, 1986, 1987, 1993
  3  *      The Regents of the University of California.
  4  * Copyright (c) 2004-2006 Robert N. M. Watson
  5  * 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  * 4. Neither the name of the University nor the names of its contributors
 16  *    may be used to endorse or promote products derived from this software
 17  *    without specific prior written permission.
 18  *
 19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 29  * SUCH DAMAGE.
 30  *
 31  * Copyright (c) 1995, Mike Mitchell
 32  * All rights reserved.
 33  *
 34  * Redistribution and use in source and binary forms, with or without
 35  * modification, are permitted provided that the following conditions
 36  * are met:
 37  * 1. Redistributions of source code must retain the above copyright
 38  *    notice, this list of conditions and the following disclaimer.
 39  * 2. Redistributions in binary form must reproduce the above copyright
 40  *    notice, this list of conditions and the following disclaimer in the
 41  *    documentation and/or other materials provided with the distribution.
 42  * 3. All advertising materials mentioning features or use of this software
 43  *    must display the following acknowledgement:
 44  *      This product includes software developed by the University of
 45  *      California, Berkeley and its contributors.
 46  * 4. Neither the name of the University nor the names of its contributors
 47  *    may be used to endorse or promote products derived from this software
 48  *    without specific prior written permission.
 49  *
 50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 60  * SUCH DAMAGE.
 61  *
 62  *      @(#)spx_usrreq.h
 63  */
 64 
 65 #include <sys/cdefs.h>
 66 __FBSDID("$FreeBSD: src/sys/netipx/spx_usrreq.c,v 1.90 2008/10/23 15:53:51 des Exp $");
 67 
 68 #include <sys/param.h>
 69 #include <sys/lock.h>
 70 #include <sys/malloc.h>
 71 #include <sys/mbuf.h>
 72 #include <sys/mutex.h>
 73 #include <sys/proc.h>
 74 #include <sys/protosw.h>
 75 #include <sys/signalvar.h>
 76 #include <sys/socket.h>
 77 #include <sys/socketvar.h>
 78 #include <sys/sx.h>
 79 #include <sys/systm.h>
 80 
 81 #include <net/route.h>
 82 #include <netinet/tcp_fsm.h>
 83 
 84 #include <netipx/ipx.h>
 85 #include <netipx/ipx_pcb.h>
 86 #include <netipx/ipx_var.h>
 87 #include <netipx/spx.h>
 88 #include <netipx/spx_debug.h>
 89 #include <netipx/spx_timer.h>
 90 #include <netipx/spx_var.h>
 91 
 92 /*
 93  * SPX protocol implementation.
 94  */
 95 static struct   mtx spx_mtx;                    /* Protects only spx_iss. */
 96 static u_short  spx_iss;
 97 static u_short  spx_newchecks[50];
 98 static int      spx_hardnosed;
 99 static int      spx_use_delack = 0;
100 static int      traceallspxs = 0;
101 static struct   spx_istat spx_istat;
102 static int      spxrexmtthresh = 3;
103 
104 #define SPX_LOCK_INIT() mtx_init(&spx_mtx, "spx_mtx", NULL, MTX_DEF)
105 #define SPX_LOCK()      mtx_lock(&spx_mtx)
106 #define SPX_UNLOCK()    mtx_unlock(&spx_mtx)
107 
108 /* Following was struct spxstat spxstat; */
109 #ifndef spxstat
110 #define spxstat spx_istat.newstats
111 #endif
112 
113 static const int spx_backoff[SPX_MAXRXTSHIFT+1] =
114     { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
115 
116 static  void spx_close(struct spxpcb *cb);
117 static  void spx_disconnect(struct spxpcb *cb);
118 static  void spx_drop(struct spxpcb *cb, int errno);
119 static  int spx_output(struct spxpcb *cb, struct mbuf *m0);
120 static  int spx_reass(struct spxpcb *cb, struct spx *si);
121 static  void spx_setpersist(struct spxpcb *cb);
122 static  void spx_template(struct spxpcb *cb);
123 static  void spx_timers(struct spxpcb *cb, int timer);
124 static  void spx_usrclosed(struct spxpcb *cb);
125 
126 static  void spx_usr_abort(struct socket *so);
127 static  int spx_accept(struct socket *so, struct sockaddr **nam);
128 static  int spx_attach(struct socket *so, int proto, struct thread *td);
129 static  int spx_bind(struct socket *so, struct sockaddr *nam, struct thread *td);
130 static  void spx_usr_close(struct socket *so);
131 static  int spx_connect(struct socket *so, struct sockaddr *nam,
132                         struct thread *td);
133 static  void spx_detach(struct socket *so);
134 static  void spx_pcbdetach(struct ipxpcb *ipxp);
135 static  int spx_usr_disconnect(struct socket *so);
136 static  int spx_listen(struct socket *so, int backlog, struct thread *td);
137 static  int spx_rcvd(struct socket *so, int flags);
138 static  int spx_rcvoob(struct socket *so, struct mbuf *m, int flags);
139 static  int spx_send(struct socket *so, int flags, struct mbuf *m,
140                      struct sockaddr *addr, struct mbuf *control,
141                      struct thread *td);
142 static  int spx_shutdown(struct socket *so);
143 static  int spx_sp_attach(struct socket *so, int proto, struct thread *td);
144 
145 struct  pr_usrreqs spx_usrreqs = {
146         .pru_abort =            spx_usr_abort,
147         .pru_accept =           spx_accept,
148         .pru_attach =           spx_attach,
149         .pru_bind =             spx_bind,
150         .pru_connect =          spx_connect,
151         .pru_control =          ipx_control,
152         .pru_detach =           spx_detach,
153         .pru_disconnect =       spx_usr_disconnect,
154         .pru_listen =           spx_listen,
155         .pru_peeraddr =         ipx_peeraddr,
156         .pru_rcvd =             spx_rcvd,
157         .pru_rcvoob =           spx_rcvoob,
158         .pru_send =             spx_send,
159         .pru_shutdown =         spx_shutdown,
160         .pru_sockaddr =         ipx_sockaddr,
161         .pru_close =            spx_usr_close,
162 };
163 
164 struct  pr_usrreqs spx_usrreq_sps = {
165         .pru_abort =            spx_usr_abort,
166         .pru_accept =           spx_accept,
167         .pru_attach =           spx_sp_attach,
168         .pru_bind =             spx_bind,
169         .pru_connect =          spx_connect,
170         .pru_control =          ipx_control,
171         .pru_detach =           spx_detach,
172         .pru_disconnect =       spx_usr_disconnect,
173         .pru_listen =           spx_listen,
174         .pru_peeraddr =         ipx_peeraddr,
175         .pru_rcvd =             spx_rcvd,
176         .pru_rcvoob =           spx_rcvoob,
177         .pru_send =             spx_send,
178         .pru_shutdown =         spx_shutdown,
179         .pru_sockaddr =         ipx_sockaddr,
180         .pru_close =            spx_usr_close,
181 };
182 
183 void
184 spx_init(void)
185 {
186 
187         SPX_LOCK_INIT();
188         spx_iss = 1; /* WRONG !! should fish it out of TODR */
189 }
190 
191 void
192 spx_input(struct mbuf *m, struct ipxpcb *ipxp)
193 {
194         struct spxpcb *cb;
195         struct spx *si = mtod(m, struct spx *);
196         struct socket *so;
197         struct spx spx_savesi;
198         int dropsocket = 0;
199         short ostate = 0;
200 
201         spxstat.spxs_rcvtotal++;
202         KASSERT(ipxp != NULL, ("spx_input: ipxpcb == NULL"));
203 
204         /*
205          * spx_input() assumes that the caller will hold both the pcb list
206          * lock and also the ipxp lock.  spx_input() will release both before
207          * returning, and may in fact trade in the ipxp lock for another pcb
208          * lock following sonewconn().
209          */
210         IPX_LIST_LOCK_ASSERT();
211         IPX_LOCK_ASSERT(ipxp);
212 
213         cb = ipxtospxpcb(ipxp);
214         KASSERT(cb != NULL, ("spx_input: cb == NULL"));
215 
216         if (ipxp->ipxp_flags & IPXP_DROPPED)
217                 goto drop;
218 
219         if (m->m_len < sizeof(*si)) {
220                 if ((m = m_pullup(m, sizeof(*si))) == NULL) {
221                         IPX_UNLOCK(ipxp);
222                         IPX_LIST_UNLOCK();
223                         spxstat.spxs_rcvshort++;
224                         return;
225                 }
226                 si = mtod(m, struct spx *);
227         }
228         si->si_seq = ntohs(si->si_seq);
229         si->si_ack = ntohs(si->si_ack);
230         si->si_alo = ntohs(si->si_alo);
231 
232         so = ipxp->ipxp_socket;
233         KASSERT(so != NULL, ("spx_input: so == NULL"));
234 
235         if (so->so_options & SO_DEBUG || traceallspxs) {
236                 ostate = cb->s_state;
237                 spx_savesi = *si;
238         }
239         if (so->so_options & SO_ACCEPTCONN) {
240                 struct spxpcb *ocb = cb;
241 
242                 so = sonewconn(so, 0);
243                 if (so == NULL)
244                         goto drop;
245 
246                 /*
247                  * This is ugly, but ....
248                  *
249                  * Mark socket as temporary until we're committed to keeping
250                  * it.  The code at ``drop'' and ``dropwithreset'' check the
251                  * flag dropsocket to see if the temporary socket created
252                  * here should be discarded.  We mark the socket as
253                  * discardable until we're committed to it below in
254                  * TCPS_LISTEN.
255                  *
256                  * XXXRW: In the new world order of real kernel parallelism,
257                  * temporarily allocating the socket when we're "not sure"
258                  * seems like a bad idea, as we might race to remove it if
259                  * the listen socket is closed...?
260                  *
261                  * We drop the lock of the listen socket ipxp, and acquire
262                  * the lock of the new socket ippx.
263                  */
264                 dropsocket++;
265                 IPX_UNLOCK(ipxp);
266                 ipxp = (struct ipxpcb *)so->so_pcb;
267                 IPX_LOCK(ipxp);
268                 ipxp->ipxp_laddr = si->si_dna;
269                 cb = ipxtospxpcb(ipxp);
270                 cb->s_mtu = ocb->s_mtu;         /* preserve sockopts */
271                 cb->s_flags = ocb->s_flags;     /* preserve sockopts */
272                 cb->s_flags2 = ocb->s_flags2;   /* preserve sockopts */
273                 cb->s_state = TCPS_LISTEN;
274         }
275         IPX_LOCK_ASSERT(ipxp);
276 
277         /*
278          * Packet received on connection.  Reset idle time and keep-alive
279          * timer.
280          */
281         cb->s_idle = 0;
282         cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
283 
284         switch (cb->s_state) {
285         case TCPS_LISTEN:{
286                 struct sockaddr_ipx *sipx, ssipx;
287                 struct ipx_addr laddr;
288 
289                 /*
290                  * If somebody here was carying on a conversation and went
291                  * away, and his pen pal thinks he can still talk, we get the
292                  * misdirected packet.
293                  */
294                 if (spx_hardnosed && (si->si_did != 0 || si->si_seq != 0)) {
295                         spx_istat.gonawy++;
296                         goto dropwithreset;
297                 }
298                 sipx = &ssipx;
299                 bzero(sipx, sizeof *sipx);
300                 sipx->sipx_len = sizeof(*sipx);
301                 sipx->sipx_family = AF_IPX;
302                 sipx->sipx_addr = si->si_sna;
303                 laddr = ipxp->ipxp_laddr;
304                 if (ipx_nullhost(laddr))
305                         ipxp->ipxp_laddr = si->si_dna;
306                 if (ipx_pcbconnect(ipxp, (struct sockaddr *)sipx, &thread0)) {
307                         ipxp->ipxp_laddr = laddr;
308                         spx_istat.noconn++;
309                         goto drop;
310                 }
311                 spx_template(cb);
312                 dropsocket = 0;         /* committed to socket */
313                 cb->s_did = si->si_sid;
314                 cb->s_rack = si->si_ack;
315                 cb->s_ralo = si->si_alo;
316 #define THREEWAYSHAKE
317 #ifdef THREEWAYSHAKE
318                 cb->s_state = TCPS_SYN_RECEIVED;
319                 cb->s_force = 1 + SPXT_KEEP;
320                 spxstat.spxs_accepts++;
321                 cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
322                 }
323                 break;
324 
325          case TCPS_SYN_RECEIVED: {
326                 /*
327                  * This state means that we have heard a response to our
328                  * acceptance of their connection.  It is probably logically
329                  * unnecessary in this implementation.
330                  */
331                 if (si->si_did != cb->s_sid) {
332                         spx_istat.wrncon++;
333                         goto drop;
334                 }
335 #endif
336                 ipxp->ipxp_fport =  si->si_sport;
337                 cb->s_timer[SPXT_REXMT] = 0;
338                 cb->s_timer[SPXT_KEEP] = SPXTV_KEEP;
339                 soisconnected(so);
340                 cb->s_state = TCPS_ESTABLISHED;
341                 spxstat.spxs_accepts++;
342                 }
343                 break;
344 
345         case TCPS_SYN_SENT:
346                 /*
347                  * This state means that we have gotten a response to our
348                  * attempt to establish a connection.  We fill in the data
349                  * from the other side, telling us which port to respond to,
350                  * instead of the well-known one we might have sent to in the
351                  * first place.  We also require that this is a response to
352                  * our connection id.
353                  */
354                 if (si->si_did != cb->s_sid) {
355                         spx_istat.notme++;
356                         goto drop;
357                 }
358                 spxstat.spxs_connects++;
359                 cb->s_did = si->si_sid;
360                 cb->s_rack = si->si_ack;
361                 cb->s_ralo = si->si_alo;
362                 cb->s_dport = ipxp->ipxp_fport =  si->si_sport;
363                 cb->s_timer[SPXT_REXMT] = 0;
364                 cb->s_flags |= SF_ACKNOW;
365                 soisconnected(so);
366                 cb->s_state = TCPS_ESTABLISHED;
367 
368                 /*
369                  * Use roundtrip time of connection request for initial rtt.
370                  */
371                 if (cb->s_rtt) {
372                         cb->s_srtt = cb->s_rtt << 3;
373                         cb->s_rttvar = cb->s_rtt << 1;
374                         SPXT_RANGESET(cb->s_rxtcur,
375                             ((cb->s_srtt >> 2) + cb->s_rttvar) >> 1,
376                             SPXTV_MIN, SPXTV_REXMTMAX);
377                             cb->s_rtt = 0;
378                 }
379         }
380 
381         if (so->so_options & SO_DEBUG || traceallspxs)
382                 spx_trace(SA_INPUT, (u_char)ostate, cb, &spx_savesi, 0);
383 
384         m->m_len -= sizeof(struct ipx);
385         m->m_pkthdr.len -= sizeof(struct ipx);
386         m->m_data += sizeof(struct ipx);
387 
388         if (spx_reass(cb, si))
389                 m_freem(m);
390         if (cb->s_force || (cb->s_flags & (SF_ACKNOW|SF_WIN|SF_RXT)))
391                 spx_output(cb, NULL);
392         cb->s_flags &= ~(SF_WIN|SF_RXT);
393         IPX_UNLOCK(ipxp);
394         IPX_LIST_UNLOCK();
395         return;
396 
397 dropwithreset:
398         IPX_LOCK_ASSERT(ipxp);
399         if (cb == NULL || (cb->s_ipxpcb->ipxp_socket->so_options & SO_DEBUG ||
400             traceallspxs))
401                 spx_trace(SA_DROP, (u_char)ostate, cb, &spx_savesi, 0);
402         IPX_UNLOCK(ipxp);
403         if (dropsocket) {
404                 struct socket *head;
405                 ACCEPT_LOCK();
406                 KASSERT((so->so_qstate & SQ_INCOMP) != 0,
407                     ("spx_input: nascent socket not SQ_INCOMP on soabort()"));
408                 head = so->so_head;
409                 TAILQ_REMOVE(&head->so_incomp, so, so_list);
410                 head->so_incqlen--;
411                 so->so_qstate &= ~SQ_INCOMP;
412                 so->so_head = NULL;
413                 ACCEPT_UNLOCK();
414                 soabort(so);
415         }
416         IPX_LIST_UNLOCK();
417         m_freem(m);
418         return;
419 
420 drop:
421         IPX_LOCK_ASSERT(ipxp);
422         if (cb->s_ipxpcb->ipxp_socket->so_options & SO_DEBUG || traceallspxs)
423                 spx_trace(SA_DROP, (u_char)ostate, cb, &spx_savesi, 0);
424         IPX_UNLOCK(ipxp);
425         IPX_LIST_UNLOCK();
426         m_freem(m);
427 }
428 
429 /*
430  * This is structurally similar to the tcp reassembly routine but its
431  * function is somewhat different: it merely queues packets up, and
432  * suppresses duplicates.
433  */
434 static int
435 spx_reass(struct spxpcb *cb, struct spx *si)
436 {
437         struct spx_q *q;
438         struct mbuf *m;
439         struct socket *so = cb->s_ipxpcb->ipxp_socket;
440         char packetp = cb->s_flags & SF_HI;
441         int incr;
442         char wakeup = 0;
443 
444         IPX_LOCK_ASSERT(cb->s_ipxpcb);
445 
446         if (si == SI(0))
447                 goto present;
448 
449         /*
450          * Update our news from them.
451          */
452         if (si->si_cc & SPX_SA)
453                 cb->s_flags |= (spx_use_delack ? SF_DELACK : SF_ACKNOW);
454         if (SSEQ_GT(si->si_alo, cb->s_ralo))
455                 cb->s_flags |= SF_WIN;
456         if (SSEQ_LEQ(si->si_ack, cb->s_rack)) {
457                 if ((si->si_cc & SPX_SP) && cb->s_rack != (cb->s_smax + 1)) {
458                         spxstat.spxs_rcvdupack++;
459 
460                         /*
461                          * If this is a completely duplicate ack and other
462                          * conditions hold, we assume a packet has been
463                          * dropped and retransmit it exactly as in
464                          * tcp_input().
465                          */
466                         if (si->si_ack != cb->s_rack ||
467                             si->si_alo != cb->s_ralo)
468                                 cb->s_dupacks = 0;
469                         else if (++cb->s_dupacks == spxrexmtthresh) {
470                                 u_short onxt = cb->s_snxt;
471                                 int cwnd = cb->s_cwnd;
472 
473                                 cb->s_snxt = si->si_ack;
474                                 cb->s_cwnd = CUNIT;
475                                 cb->s_force = 1 + SPXT_REXMT;
476                                 spx_output(cb, NULL);
477                                 cb->s_timer[SPXT_REXMT] = cb->s_rxtcur;
478                                 cb->s_rtt = 0;
479                                 if (cwnd >= 4 * CUNIT)
480                                         cb->s_cwnd = cwnd / 2;
481                                 if (SSEQ_GT(onxt, cb->s_snxt))
482                                         cb->s_snxt = onxt;
483                                 return (1);
484                         }
485                 } else
486                         cb->s_dupacks = 0;
487                 goto update_window;
488         }
489         cb->s_dupacks = 0;
490 
491         /*
492          * If our correspondent acknowledges data we haven't sent TCP would
493          * drop the packet after acking.  We'll be a little more permissive.
494          */
495         if (SSEQ_GT(si->si_ack, (cb->s_smax + 1))) {
496                 spxstat.spxs_rcvacktoomuch++;
497                 si->si_ack = cb->s_smax + 1;
498         }
499         spxstat.spxs_rcvackpack++;
500 
501         /*
502          * If transmit timer is running and timed sequence number was acked,
503          * update smoothed round trip time.  See discussion of algorithm in
504          * tcp_input.c
505          */
506         if (cb->s_rtt && SSEQ_GT(si->si_ack, cb->s_rtseq)) {
507                 spxstat.spxs_rttupdated++;
508                 if (cb->s_srtt != 0) {
509                         short delta;
510                         delta = cb->s_rtt - (cb->s_srtt >> 3);
511                         if ((cb->s_srtt += delta) <= 0)
512                                 cb->s_srtt = 1;
513                         if (delta < 0)
514                                 delta = -delta;
515                         delta -= (cb->s_rttvar >> 2);
516                         if ((cb->s_rttvar += delta) <= 0)
517                                 cb->s_rttvar = 1;
518                 } else {
519                         /*
520                          * No rtt measurement yet.
521                          */
522                         cb->s_srtt = cb->s_rtt << 3;
523                         cb->s_rttvar = cb->s_rtt << 1;
524                 }
525                 cb->s_rtt = 0;
526                 cb->s_rxtshift = 0;
527                 SPXT_RANGESET(cb->s_rxtcur,
528                         ((cb->s_srtt >> 2) + cb->s_rttvar) >> 1,
529                         SPXTV_MIN, SPXTV_REXMTMAX);
530         }
531 
532         /*
533          * If all outstanding data is acked, stop retransmit timer and
534          * remember to restart (more output or persist).  If there is more
535          * data to be acked, restart retransmit timer, using current
536          * (possibly backed-off) value;
537          */
538         if (si->si_ack == cb->s_smax + 1) {
539                 cb->s_timer[SPXT_REXMT] = 0;
540                 cb->s_flags |= SF_RXT;
541         } else if (cb->s_timer[SPXT_PERSIST] == 0)
542                 cb->s_timer[SPXT_REXMT] = cb->s_rxtcur;
543 
544         /*
545          * When new data is acked, open the congestion window.  If the window
546          * gives us less than ssthresh packets in flight, open exponentially
547          * (maxseg at a time).  Otherwise open linearly (maxseg^2 / cwnd at a
548          * time).
549          */
550         incr = CUNIT;
551         if (cb->s_cwnd > cb->s_ssthresh)
552                 incr = max(incr * incr / cb->s_cwnd, 1);
553         cb->s_cwnd = min(cb->s_cwnd + incr, cb->s_cwmx);
554 
555         /*
556          * Trim Acked data from output queue.
557          */
558         SOCKBUF_LOCK(&so->so_snd);
559         while ((m = so->so_snd.sb_mb) != NULL) {
560                 if (SSEQ_LT((mtod(m, struct spx *))->si_seq, si->si_ack))
561                         sbdroprecord_locked(&so->so_snd);
562                 else
563                         break;
564         }
565         sowwakeup_locked(so);
566         cb->s_rack = si->si_ack;
567 update_window:
568         if (SSEQ_LT(cb->s_snxt, cb->s_rack))
569                 cb->s_snxt = cb->s_rack;
570         if (SSEQ_LT(cb->s_swl1, si->si_seq) || ((cb->s_swl1 == si->si_seq &&
571             (SSEQ_LT(cb->s_swl2, si->si_ack))) ||
572              (cb->s_swl2 == si->si_ack && SSEQ_LT(cb->s_ralo, si->si_alo)))) {
573                 /* keep track of pure window updates */
574                 if ((si->si_cc & SPX_SP) && cb->s_swl2 == si->si_ack
575                     && SSEQ_LT(cb->s_ralo, si->si_alo)) {
576                         spxstat.spxs_rcvwinupd++;
577                         spxstat.spxs_rcvdupack--;
578                 }
579                 cb->s_ralo = si->si_alo;
580                 cb->s_swl1 = si->si_seq;
581                 cb->s_swl2 = si->si_ack;
582                 cb->s_swnd = (1 + si->si_alo - si->si_ack);
583                 if (cb->s_swnd > cb->s_smxw)
584                         cb->s_smxw = cb->s_swnd;
585                 cb->s_flags |= SF_WIN;
586         }
587 
588         /*
589          * If this packet number is higher than that which we have allocated
590          * refuse it, unless urgent.
591          */
592         if (SSEQ_GT(si->si_seq, cb->s_alo)) {
593                 if (si->si_cc & SPX_SP) {
594                         spxstat.spxs_rcvwinprobe++;
595                         return (1);
596                 } else
597                         spxstat.spxs_rcvpackafterwin++;
598                 if (si->si_cc & SPX_OB) {
599                         if (SSEQ_GT(si->si_seq, cb->s_alo + 60))
600                                 return (1); /* else queue this packet; */
601                 } else {
602 #ifdef BROKEN
603                         /*
604                          * XXXRW: This is broken on at least one count:
605                          * spx_close() will free the ipxp and related parts,
606                          * which are then touched by spx_input() after the
607                          * return from spx_reass().
608                          */
609                         /*struct socket *so = cb->s_ipxpcb->ipxp_socket;
610                         if (so->so_state && SS_NOFDREF) {
611                                 spx_close(cb);
612                         } else
613                                        would crash system*/
614 #endif
615                         spx_istat.notyet++;
616                         return (1);
617                 }
618         }
619 
620         /*
621          * If this is a system packet, we don't need to queue it up, and
622          * won't update acknowledge #.
623          */
624         if (si->si_cc & SPX_SP)
625                 return (1);
626 
627         /*
628          * We have already seen this packet, so drop.
629          */
630         if (SSEQ_LT(si->si_seq, cb->s_ack)) {
631                 spx_istat.bdreas++;
632                 spxstat.spxs_rcvduppack++;
633                 if (si->si_seq == cb->s_ack - 1)
634                         spx_istat.lstdup++;
635                 return (1);
636         }
637 
638         /*
639          * Loop through all packets queued up to insert in appropriate
640          * sequence.
641          */
642         for (q = cb->s_q.si_next; q != &cb->s_q; q = q->si_next) {
643                 if (si->si_seq == SI(q)->si_seq) {
644                         spxstat.spxs_rcvduppack++;
645                         return (1);
646                 }
647                 if (SSEQ_LT(si->si_seq, SI(q)->si_seq)) {
648                         spxstat.spxs_rcvoopack++;
649                         break;
650                 }
651         }
652         insque(si, q->si_prev);
653 
654         /*
655          * If this packet is urgent, inform process
656          */
657         if (si->si_cc & SPX_OB) {
658                 cb->s_iobc = ((char *)si)[1 + sizeof(*si)];
659                 sohasoutofband(so);
660                 cb->s_oobflags |= SF_IOOB;
661         }
662 present:
663 #define SPINC sizeof(struct spxhdr)
664         SOCKBUF_LOCK(&so->so_rcv);
665 
666         /*
667          * Loop through all packets queued up to update acknowledge number,
668          * and present all acknowledged data to user; if in packet interface
669          * mode, show packet headers.
670          */
671         for (q = cb->s_q.si_next; q != &cb->s_q; q = q->si_next) {
672                   if (SI(q)->si_seq == cb->s_ack) {
673                         cb->s_ack++;
674                         m = dtom(q);
675                         if (SI(q)->si_cc & SPX_OB) {
676                                 cb->s_oobflags &= ~SF_IOOB;
677                                 if (so->so_rcv.sb_cc)
678                                         so->so_oobmark = so->so_rcv.sb_cc;
679                                 else
680                                         so->so_rcv.sb_state |= SBS_RCVATMARK;
681                         }
682                         q = q->si_prev;
683                         remque(q->si_next);
684                         wakeup = 1;
685                         spxstat.spxs_rcvpack++;
686 #ifdef SF_NEWCALL
687                         if (cb->s_flags2 & SF_NEWCALL) {
688                                 struct spxhdr *sp = mtod(m, struct spxhdr *);
689                                 u_char dt = sp->spx_dt;
690                                 spx_newchecks[4]++;
691                                 if (dt != cb->s_rhdr.spx_dt) {
692                                         struct mbuf *mm =
693                                            m_getclr(M_DONTWAIT, MT_CONTROL);
694                                         spx_newchecks[0]++;
695                                         if (mm != NULL) {
696                                                 u_short *s =
697                                                         mtod(mm, u_short *);
698                                                 cb->s_rhdr.spx_dt = dt;
699                                                 mm->m_len = 5; /*XXX*/
700                                                 s[0] = 5;
701                                                 s[1] = 1;
702                                                 *(u_char *)(&s[2]) = dt;
703                                                 sbappend_locked(&so->so_rcv, mm);
704                                         }
705                                 }
706                                 if (sp->spx_cc & SPX_OB) {
707                                         MCHTYPE(m, MT_OOBDATA);
708                                         spx_newchecks[1]++;
709                                         so->so_oobmark = 0;
710                                         so->so_rcv.sb_state &= ~SBS_RCVATMARK;
711                                 }
712                                 if (packetp == 0) {
713                                         m->m_data += SPINC;
714                                         m->m_len -= SPINC;
715                                         m->m_pkthdr.len -= SPINC;
716                                 }
717                                 if ((sp->spx_cc & SPX_EM) || packetp) {
718                                         sbappendrecord_locked(&so->so_rcv, m);
719                                         spx_newchecks[9]++;
720                                 } else
721                                         sbappend_locked(&so->so_rcv, m);
722                         } else
723 #endif
724                         if (packetp)
725                                 sbappendrecord_locked(&so->so_rcv, m);
726                         else {
727                                 cb->s_rhdr = *mtod(m, struct spxhdr *);
728                                 m->m_data += SPINC;
729                                 m->m_len -= SPINC;
730                                 m->m_pkthdr.len -= SPINC;
731                                 sbappend_locked(&so->so_rcv, m);
732                         }
733                   } else
734                         break;
735         }
736         if (wakeup)
737                 sorwakeup_locked(so);
738         else
739                 SOCKBUF_UNLOCK(&so->so_rcv);
740         return (0);
741 }
742 
743 void
744 spx_ctlinput(int cmd, struct sockaddr *arg_as_sa, void *dummy)
745 {
746 
747         /* Currently, nothing. */
748 }
749 
750 static int
751 spx_output(struct spxpcb *cb, struct mbuf *m0)
752 {
753         struct socket *so = cb->s_ipxpcb->ipxp_socket;
754         struct mbuf *m;
755         struct spx *si = NULL;
756         struct sockbuf *sb = &so->so_snd;
757         int len = 0, win, rcv_win;
758         short span, off, recordp = 0;
759         u_short alo;
760         int error = 0, sendalot;
761 #ifdef notdef
762         int idle;
763 #endif
764         struct mbuf *mprev;
765 
766         IPX_LOCK_ASSERT(cb->s_ipxpcb);
767 
768         if (m0 != NULL) {
769                 int mtu = cb->s_mtu;
770                 int datalen;
771 
772                 /*
773                  * Make sure that packet isn't too big.
774                  */
775                 for (m = m0; m != NULL; m = m->m_next) {
776                         mprev = m;
777                         len += m->m_len;
778                         if (m->m_flags & M_EOR)
779                                 recordp = 1;
780                 }
781                 datalen = (cb->s_flags & SF_HO) ?
782                                 len - sizeof(struct spxhdr) : len;
783                 if (datalen > mtu) {
784                         if (cb->s_flags & SF_PI) {
785                                 m_freem(m0);
786                                 return (EMSGSIZE);
787                         } else {
788                                 int oldEM = cb->s_cc & SPX_EM;
789 
790                                 cb->s_cc &= ~SPX_EM;
791                                 while (len > mtu) {
792                                         m = m_copym(m0, 0, mtu, M_DONTWAIT);
793                                         if (m == NULL) {
794                                             cb->s_cc |= oldEM;
795                                             m_freem(m0);
796                                             return (ENOBUFS);
797                                         }
798                                         if (cb->s_flags & SF_NEWCALL) {
799                                             struct mbuf *mm = m;
800                                             spx_newchecks[7]++;
801                                             while (mm != NULL) {
802                                                 mm->m_flags &= ~M_EOR;
803                                                 mm = mm->m_next;
804                                             }
805                                         }
806                                         error = spx_output(cb, m);
807                                         if (error) {
808                                                 cb->s_cc |= oldEM;
809                                                 m_freem(m0);
810                                                 return (error);
811                                         }
812                                         m_adj(m0, mtu);
813                                         len -= mtu;
814                                 }
815                                 cb->s_cc |= oldEM;
816                         }
817                 }
818 
819                 /*
820                  * Force length even, by adding a "garbage byte" if
821                  * necessary.
822                  */
823                 if (len & 1) {
824                         m = mprev;
825                         if (M_TRAILINGSPACE(m) >= 1)
826                                 m->m_len++;
827                         else {
828                                 struct mbuf *m1 = m_get(M_DONTWAIT, MT_DATA);
829 
830                                 if (m1 == NULL) {
831                                         m_freem(m0);
832                                         return (ENOBUFS);
833                                 }
834                                 m1->m_len = 1;
835                                 *(mtod(m1, u_char *)) = 0;
836                                 m->m_next = m1;
837                         }
838                 }
839                 m = m_gethdr(M_DONTWAIT, MT_DATA);
840                 if (m == NULL) {
841                         m_freem(m0);
842                         return (ENOBUFS);
843                 }
844 
845                 /*
846                  * Fill in mbuf with extended SP header and addresses and
847                  * length put into network format.
848                  */
849                 MH_ALIGN(m, sizeof(struct spx));
850                 m->m_len = sizeof(struct spx);
851                 m->m_next = m0;
852                 si = mtod(m, struct spx *);
853                 si->si_i = *cb->s_ipx;
854                 si->si_s = cb->s_shdr;
855                 if ((cb->s_flags & SF_PI) && (cb->s_flags & SF_HO)) {
856                         struct spxhdr *sh;
857                         if (m0->m_len < sizeof(*sh)) {
858                                 if((m0 = m_pullup(m0, sizeof(*sh))) == NULL) {
859                                         m_free(m);
860                                         m_freem(m0);
861                                         return (EINVAL);
862                                 }
863                                 m->m_next = m0;
864                         }
865                         sh = mtod(m0, struct spxhdr *);
866                         si->si_dt = sh->spx_dt;
867                         si->si_cc |= sh->spx_cc & SPX_EM;
868                         m0->m_len -= sizeof(*sh);
869                         m0->m_data += sizeof(*sh);
870                         len -= sizeof(*sh);
871                 }
872                 len += sizeof(*si);
873                 if ((cb->s_flags2 & SF_NEWCALL) && recordp) {
874                         si->si_cc |= SPX_EM;
875                         spx_newchecks[8]++;
876                 }
877                 if (cb->s_oobflags & SF_SOOB) {
878                         /*
879                          * Per jqj@cornell: Make sure OB packets convey
880                          * exactly 1 byte.  If the packet is 1 byte or
881                          * larger, we have already guaranted there to be at
882                          * least one garbage byte for the checksum, and extra
883                          * bytes shouldn't hurt!
884                          */
885                         if (len > sizeof(*si)) {
886                                 si->si_cc |= SPX_OB;
887                                 len = (1 + sizeof(*si));
888                         }
889                 }
890                 si->si_len = htons((u_short)len);
891                 m->m_pkthdr.len = ((len - 1) | 1) + 1;
892 
893                 /*
894                  * Queue stuff up for output.
895                  */
896                 sbappendrecord(sb, m);
897                 cb->s_seq++;
898         }
899 #ifdef notdef
900         idle = (cb->s_smax == (cb->s_rack - 1));
901 #endif
902 again:
903         sendalot = 0;
904         off = cb->s_snxt - cb->s_rack;
905         win = min(cb->s_swnd, (cb->s_cwnd / CUNIT));
906 
907         /*
908          * If in persist timeout with window of 0, send a probe.  Otherwise,
909          * if window is small but non-zero and timer expired, send what we
910          * can and go into transmit state.
911          */
912         if (cb->s_force == 1 + SPXT_PERSIST) {
913                 if (win != 0) {
914                         cb