1/*
2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)uipc_socket2.c 8.1 (Berkeley) 6/10/93
34 * $Id: uipc_socket2.c,v 1.43 1998/12/07 21:58:29 archie Exp $
35 */
36
37#include <sys/param.h>
38#include <sys/systm.h>
39#include <sys/domain.h>
40#include <sys/kernel.h>
41#include <sys/proc.h>
42#include <sys/malloc.h>
43#include <sys/mbuf.h>
44#include <sys/protosw.h>
45#include <sys/stat.h>
46#include <sys/socket.h>
47#include <sys/socketvar.h>
48#include <sys/signalvar.h>
49#include <sys/sysctl.h>
50
51/*
52 * Primitive routines for operating on sockets and socket buffers
53 */
54
55u_long sb_max = SB_MAX; /* XXX should be static */
56
57static u_long sb_efficiency = 8; /* parameter for sbreserve() */
58
59/*
60 * Procedures to manipulate state flags of socket
61 * and do appropriate wakeups. Normal sequence from the
62 * active (originating) side is that soisconnecting() is
63 * called during processing of connect() call,
64 * resulting in an eventual call to soisconnected() if/when the
65 * connection is established. When the connection is torn down
66 * soisdisconnecting() is called during processing of disconnect() call,
67 * and soisdisconnected() is called when the connection to the peer
68 * is totally severed. The semantics of these routines are such that
69 * connectionless protocols can call soisconnected() and soisdisconnected()
70 * only, bypassing the in-progress calls when setting up a ``connection''
71 * takes no time.
72 *
73 * From the passive side, a socket is created with
74 * two queues of sockets: so_q0 for connections in progress
75 * and so_q for connections already made and awaiting user acceptance.
76 * As a protocol is preparing incoming connections, it creates a socket
77 * structure queued on so_q0 by calling sonewconn(). When the connection
78 * is established, soisconnected() is called, and transfers the
79 * socket structure to so_q, making it available to accept().
80 *
81 * If a socket is closed with sockets on either
82 * so_q0 or so_q, these sockets are dropped.
83 *
84 * If higher level protocols are implemented in
85 * the kernel, the wakeups done here will sometimes
86 * cause software-interrupt process scheduling.
87 */
88
89void
90soisconnecting(so)
91 register struct socket *so;
92{
93
94 so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
95 so->so_state |= SS_ISCONNECTING;
96}
97
98void
99soisconnected(so)
100 register struct socket *so;
101{
102 register struct socket *head = so->so_head;
103
104 so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
105 so->so_state |= SS_ISCONNECTED;
106 if (head && (so->so_state & SS_INCOMP)) {
107 TAILQ_REMOVE(&head->so_incomp, so, so_list);
108 head->so_incqlen--;
109 so->so_state &= ~SS_INCOMP;
110 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
111 so->so_state |= SS_COMP;
112 sorwakeup(head);
113 wakeup_one(&head->so_timeo);
114 } else {
115 wakeup(&so->so_timeo);
116 sorwakeup(so);
117 sowwakeup(so);
118 }
119}
120
121void
122soisdisconnecting(so)
123 register struct socket *so;
124{
125
126 so->so_state &= ~SS_ISCONNECTING;
127 so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
128 wakeup((caddr_t)&so->so_timeo);
129 sowwakeup(so);
130 sorwakeup(so);
131}
132
133void
134soisdisconnected(so)
135 register struct socket *so;
136{
137
138 so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
139 so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED);
140 wakeup((caddr_t)&so->so_timeo);
141 sowwakeup(so);
142 sorwakeup(so);
143}
144
145/*
146 * Return a random connection that hasn't been serviced yet and
147 * is eligible for discard. There is a one in qlen chance that
148 * we will return a null, saying that there are no dropable
149 * requests. In this case, the protocol specific code should drop
150 * the new request. This insures fairness.
151 *
152 * This may be used in conjunction with protocol specific queue
153 * congestion routines.
154 */
155struct socket *
156sodropablereq(head)
157 register struct socket *head;
158{
159 register struct socket *so;
160 unsigned int i, j, qlen;
161 static int rnd;
162 static struct timeval old_runtime;
163 static unsigned int cur_cnt, old_cnt;
164 struct timeval tv;
165
166 getmicrouptime(&tv);
167 if ((i = (tv.tv_sec - old_runtime.tv_sec)) != 0) {
168 old_runtime = tv;
169 old_cnt = cur_cnt / i;
170 cur_cnt = 0;
171 }
172
173 so = TAILQ_FIRST(&head->so_incomp);
174 if (!so)
175 return (so);
176
177 qlen = head->so_incqlen;
178 if (++cur_cnt > qlen || old_cnt > qlen) {
179 rnd = (314159 * rnd + 66329) & 0xffff;
180 j = ((qlen + 1) * rnd) >> 16;
181
182 while (j-- && so)
183 so = TAILQ_NEXT(so, so_list);
184 }
185
186 return (so);
187}
188
189/*
190 * When an attempt at a new connection is noted on a socket
191 * which accepts connections, sonewconn is called. If the
192 * connection is possible (subject to space constraints, etc.)
193 * then we allocate a new structure, propoerly linked into the
194 * data structure of the original socket, and return this.
195 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
196 */
197struct socket *
198sonewconn(head, connstatus)
199 register struct socket *head;
200 int connstatus;
201{
202 register struct socket *so;
203
204 if (head->so_qlen > 3 * head->so_qlimit / 2)
205 return ((struct socket *)0);
206 so = soalloc(0);
207 if (so == NULL)
208 return ((struct socket *)0);
209 so->so_head = head;
210 so->so_type = head->so_type;
211 so->so_options = head->so_options &~ SO_ACCEPTCONN;
212 so->so_linger = head->so_linger;
213 so->so_state = head->so_state | SS_NOFDREF;
214 so->so_proto = head->so_proto;
215 so->so_timeo = head->so_timeo;
216 so->so_uid = head->so_uid;
217 (void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat);
218
219 if ((*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) {
220 sodealloc(so);
221 return ((struct socket *)0);
222 }
223
224 if (connstatus) {
225 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
226 so->so_state |= SS_COMP;
227 } else {
228 TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
229 so->so_state |= SS_INCOMP;
230 head->so_incqlen++;
231 }
232 head->so_qlen++;
233 if (connstatus) {
234 sorwakeup(head);
235 wakeup((caddr_t)&head->so_timeo);
236 so->so_state |= connstatus;
237 }
238 return (so);
239}
240
241/*
242 * Socantsendmore indicates that no more data will be sent on the
243 * socket; it would normally be applied to a socket when the user
244 * informs the system that no more data is to be sent, by the protocol
245 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data
246 * will be received, and will normally be applied to the socket by a
247 * protocol when it detects that the peer will send no more data.
248 * Data queued for reading in the socket may yet be read.
249 */
250
251void
252socantsendmore(so)
253 struct socket *so;
254{
255
256 so->so_state |= SS_CANTSENDMORE;
257 sowwakeup(so);
258}
259
260void
261socantrcvmore(so)
262 struct socket *so;
263{
264
265 so->so_state |= SS_CANTRCVMORE;
266 sorwakeup(so);
267}
268
269/*
270 * Wait for data to arrive at/drain from a socket buffer.
271 */
272int
273sbwait(sb)
274 struct sockbuf *sb;
275{
276
277 sb->sb_flags |= SB_WAIT;
278 return (tsleep((caddr_t)&sb->sb_cc,
279 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
280 sb->sb_timeo));
281}
282
283/*
284 * Lock a sockbuf already known to be locked;
285 * return any error returned from sleep (EINTR).
286 */
287int
288sb_lock(sb)
289 register struct sockbuf *sb;
290{
291 int error;
292
293 while (sb->sb_flags & SB_LOCK) {
294 sb->sb_flags |= SB_WANT;
295 error = tsleep((caddr_t)&sb->sb_flags,
296 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH,
297 "sblock", 0);
298 if (error)
299 return (error);
300 }
301 sb->sb_flags |= SB_LOCK;
302 return (0);
303}
304
305/*
306 * Wakeup processes waiting on a socket buffer.
307 * Do asynchronous notification via SIGIO
308 * if the socket has the SS_ASYNC flag set.
309 */
310void
311sowakeup(so, sb)
312 register struct socket *so;
313 register struct sockbuf *sb;
314{
315 selwakeup(&sb->sb_sel);
316 sb->sb_flags &= ~SB_SEL;
317 if (sb->sb_flags & SB_WAIT) {
318 sb->sb_flags &= ~SB_WAIT;
319 wakeup((caddr_t)&sb->sb_cc);
320 }
321 if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL)
322 pgsigio(so->so_sigio, SIGIO, 0);
323 if (sb->sb_flags & SB_UPCALL)
324 (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT);
325}
326
327/*
328 * Socket buffer (struct sockbuf) utility routines.
329 *
330 * Each socket contains two socket buffers: one for sending data and
331 * one for receiving data. Each buffer contains a queue of mbufs,
332 * information about the number of mbufs and amount of data in the
333 * queue, and other fields allowing select() statements and notification
334 * on data availability to be implemented.
335 *
336 * Data stored in a socket buffer is maintained as a list of records.
337 * Each record is a list of mbufs chained together with the m_next
338 * field. Records are chained together with the m_nextpkt field. The upper
339 * level routine soreceive() expects the following conventions to be
340 * observed when placing information in the receive buffer:
341 *
342 * 1. If the protocol requires each message be preceded by the sender's
343 * name, then a record containing that name must be present before
344 * any associated data (mbuf's must be of type MT_SONAME).
345 * 2. If the protocol supports the exchange of ``access rights'' (really
346 * just additional data associated with the message), and there are
347 * ``rights'' to be received, then a record containing this data
348 * should be present (mbuf's must be of type MT_RIGHTS).
349 * 3. If a name or rights record exists, then it must be followed by
350 * a data record, perhaps of zero length.
351 *
352 * Before using a new socket structure it is first necessary to reserve
353 * buffer space to the socket, by calling sbreserve(). This should commit
354 * some of the available buffer space in the system buffer pool for the
355 * socket (currently, it does nothing but enforce limits). The space
356 * should be released by calling sbrelease() when the socket is destroyed.
357 */
358
359int
360soreserve(so, sndcc, rcvcc)
361 register struct socket *so;
362 u_long sndcc, rcvcc;
363{
364
365 if (sbreserve(&so->so_snd, sndcc) == 0)
366 goto bad;
367 if (sbreserve(&so->so_rcv, rcvcc) == 0)
368 goto bad2;
369 if (so->so_rcv.sb_lowat == 0)
370 so->so_rcv.sb_lowat = 1;
371 if (so->so_snd.sb_lowat == 0)
372 so->so_snd.sb_lowat = MCLBYTES;
373 if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
374 so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
375 return (0);
376bad2:
377 sbrelease(&so->so_snd);
378bad:
379 return (ENOBUFS);
380}
381
382/*
383 * Allot mbufs to a sockbuf.
384 * Attempt to scale mbmax so that mbcnt doesn't become limiting
385 * if buffering efficiency is near the normal case.
386 */
387int
388sbreserve(sb, cc)
389 struct sockbuf *sb;
390 u_long cc;
391{
392 if ((u_quad_t)cc > (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES))
393 return (0);
394 sb->sb_hiwat = cc;
395 sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
396 if (sb->sb_lowat > sb->sb_hiwat)
397 sb->sb_lowat = sb->sb_hiwat;
398 return (1);
399}
400
401/*
402 * Free mbufs held by a socket, and reserved mbuf space.
403 */
404void
405sbrelease(sb)
406 struct sockbuf *sb;
407{
408
409 sbflush(sb);
410 sb->sb_hiwat = sb->sb_mbmax = 0;
411}
412
413/*
414 * Routines to add and remove
415 * data from an mbuf queue.
416 *
417 * The routines sbappend() or sbappendrecord() are normally called to
418 * append new mbufs to a socket buffer, after checking that adequate
419 * space is available, comparing the function sbspace() with the amount
420 * of data to be added. sbappendrecord() differs from sbappend() in
421 * that data supplied is treated as the beginning of a new record.
422 * To place a sender's address, optional access rights, and data in a
423 * socket receive buffer, sbappendaddr() should be used. To place
424 * access rights and data in a socket receive buffer, sbappendrights()
425 * should be used. In either case, the new data begins a new record.
426 * Note that unlike sbappend() and sbappendrecord(), these routines check
427 * for the caller that there will be enough space to store the data.
428 * Each fails if there is not enough space, or if it cannot find mbufs
429 * to store additional information in.
430 *
431 * Reliable protocols may use the socket send buffer to hold data
432 * awaiting acknowledgement. Data is normally copied from a socket
433 * send buffer in a protocol with m_copy for output to a peer,
434 * and then removing the data from the socket buffer with sbdrop()
435 * or sbdroprecord() when the data is acknowledged by the peer.
436 */
437
438/*
439 * Append mbuf chain m to the last record in the
440 * socket buffer sb. The additional space associated
441 * the mbuf chain is recorded in sb. Empty mbufs are
442 * discarded and mbufs are compacted where possible.
443 */
444void
445sbappend(sb, m)
446 struct sockbuf *sb;
447 struct mbuf *m;
448{
449 register struct mbuf *n;
450
451 if (m == 0)
452 return;
453 n = sb->sb_mb;
454 if (n) {
455 while (n->m_nextpkt)
456 n = n->m_nextpkt;
457 do {
458 if (n->m_flags & M_EOR) {
459 sbappendrecord(sb, m); /* XXXXXX!!!! */
460 return;
461 }
462 } while (n->m_next && (n = n->m_next));
463 }
464 sbcompress(sb, m, n);
465}
466
467#ifdef SOCKBUF_DEBUG
468void
469sbcheck(sb)
470 register struct sockbuf *sb;
471{
472 register struct mbuf *m;
473 register struct mbuf *n = 0;
474 register u_long len = 0, mbcnt = 0;
475
476 for (m = sb->sb_mb; m; m = n) {
477 n = m->m_nextpkt;
478 for (; m; m = m->m_next) {
479 len += m->m_len;
480 mbcnt += MSIZE;
481 if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */
482 mbcnt += m->m_ext.ext_size;
483 }
484 }
485 if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
486 printf("cc %ld != %ld || mbcnt %ld != %ld\n", len, sb->sb_cc,
487 mbcnt, sb->sb_mbcnt);
488 panic("sbcheck");
489 }
490}
491#endif
492
493/*
494 * As above, except the mbuf chain
495 * begins a new record.
496 */
497void
498sbappendrecord(sb, m0)
499 register struct sockbuf *sb;
500 register struct mbuf *m0;
501{
502 register struct mbuf *m;
503
504 if (m0 == 0)
505 return;
506 m = sb->sb_mb;
507 if (m)
508 while (m->m_nextpkt)
509 m = m->m_nextpkt;
510 /*
511 * Put the first mbuf on the queue.
512 * Note this permits zero length records.
513 */
514 sballoc(sb, m0);
515 if (m)
516 m->m_nextpkt = m0;
517 else
518 sb->sb_mb = m0;
519 m = m0->m_next;
520 m0->m_next = 0;
521 if (m && (m0->m_flags & M_EOR)) {
522 m0->m_flags &= ~M_EOR;
523 m->m_flags |= M_EOR;
524 }
525 sbcompress(sb, m, m0);
526}
527
528/*
529 * As above except that OOB data
530 * is inserted at the beginning of the sockbuf,
531 * but after any other OOB data.
532 */
533void
534sbinsertoob(sb, m0)
535 register struct sockbuf *sb;
536 register struct mbuf *m0;
537{
538 register struct mbuf *m;
539 register struct mbuf **mp;
540
541 if (m0 == 0)
542 return;
543 for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) {
544 m = *mp;
545 again:
546 switch (m->m_type) {
547
548 case MT_OOBDATA:
549 continue; /* WANT next train */
550
551 case MT_CONTROL:
552 m = m->m_next;
553 if (m)
554 goto again; /* inspect THIS train further */
555 }
556 break;
557 }
558 /*
559 * Put the first mbuf on the queue.
560 * Note this permits zero length records.
561 */
562 sballoc(sb, m0);
563 m0->m_nextpkt = *mp;
564 *mp = m0;
565 m = m0->m_next;
566 m0->m_next = 0;
567 if (m && (m0->m_flags & M_EOR)) {
568 m0->m_flags &= ~M_EOR;
569 m->m_flags |= M_EOR;
570 }
571 sbcompress(sb, m, m0);
572}
573
574/*
575 * Append address and data, and optionally, control (ancillary) data
576 * to the receive queue of a socket. If present,
577 * m0 must include a packet header with total length.
578 * Returns 0 if no space in sockbuf or insufficient mbufs.
579 */
580int
581sbappendaddr(sb, asa, m0, control)
582 register struct sockbuf *sb;
583 struct sockaddr *asa;
584 struct mbuf *m0, *control;
585{
586 register struct mbuf *m, *n;
587 int space = asa->sa_len;
588
589if (m0 && (m0->m_flags & M_PKTHDR) == 0)
590panic("sbappendaddr");
591 if (m0)
592 space += m0->m_pkthdr.len;
593 for (n = control; n; n = n->m_next) {
594 space += n->m_len;
595 if (n->m_next == 0) /* keep pointer to last control buf */
596 break;
597 }
598 if (space > sbspace(sb))
599 return (0);
600 if (asa->sa_len > MLEN)
601 return (0);
602 MGET(m, M_DONTWAIT, MT_SONAME);
603 if (m == 0)
604 return (0);
605 m->m_len = asa->sa_len;
606 bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
607 if (n)
608 n->m_next = m0; /* concatenate data to control */
609 else
610 control = m0;
611 m->m_next = control;
612 for (n = m; n; n = n->m_next)
613 sballoc(sb, n);
614 n = sb->sb_mb;
615 if (n) {
616 while (n->m_nextpkt)
617 n = n->m_nextpkt;
618 n->m_nextpkt = m;
619 } else
620 sb->sb_mb = m;
621 return (1);
622}
623
624int
625sbappendcontrol(sb, m0, control)
626 struct sockbuf *sb;
627 struct mbuf *control, *m0;
628{
629 register struct mbuf *m, *n;
630 int space = 0;
631
632 if (control == 0)
633 panic("sbappendcontrol");
634 for (m = control; ; m = m->m_next) {
635 space += m->m_len;
636 if (m->m_next == 0)
637 break;
638 }
639 n = m; /* save pointer to last control buffer */
640 for (m = m0; m; m = m->m_next)
641 space += m->m_len;
642 if (space > sbspace(sb))
643 return (0);
644 n->m_next = m0; /* concatenate data to control */
645 for (m = control; m; m = m->m_next)
646 sballoc(sb, m);
647 n = sb->sb_mb;
648 if (n) {
649 while (n->m_nextpkt)
650 n = n->m_nextpkt;
651 n->m_nextpkt = control;
652 } else
653 sb->sb_mb = control;
654 return (1);
655}
656
657/*
658 * Compress mbuf chain m into the socket
659 * buffer sb following mbuf n. If n
660 * is null, the buffer is presumed empty.
661 */
662void
663sbcompress(sb, m, n)
664 register struct sockbuf *sb;
665 register struct mbuf *m, *n;
666{
667 register int eor = 0;
668 register struct mbuf *o;
669
670 while (m) {
671 eor |= m->m_flags & M_EOR;
672 if (m->m_len == 0 &&
673 (eor == 0 ||
674 (((o = m->m_next) || (o = n)) &&
675 o->m_type == m->m_type))) {
676 m = m_free(m);
677 continue;
678 }
679 if (n && (n->m_flags & (M_EXT | M_EOR)) == 0 &&
680 (n->m_data + n->m_len + m->m_len) < &n->m_dat[MLEN] &&
681 n->m_type == m->m_type) {
682 bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
683 (unsigned)m->m_len);
684 n->m_len += m->m_len;
685 sb->sb_cc += m->m_len;
686 m = m_free(m);
687 continue;
688 }
689 if (n)
690 n->m_next = m;
691 else
692 sb->sb_mb = m;
693 sballoc(sb, m);
694 n = m;
695 m->m_flags &= ~M_EOR;
696 m = m->m_next;
697 n->m_next = 0;
698 }
699 if (eor) {
700 if (n)
701 n->m_flags |= eor;
702 else
703 printf("semi-panic: sbcompress\n");
704 }
705}
706
707/*
708 * Free all mbufs in a sockbuf.
709 * Check that all resources are reclaimed.
710 */
711void
712sbflush(sb)
713 register struct sockbuf *sb;
714{
715
716 if (sb->sb_flags & SB_LOCK)
717 panic("sbflush: locked");
718 while (sb->sb_mbcnt && sb->sb_cc)
719 sbdrop(sb, (int)sb->sb_cc);
720 if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt)
721 panic("sbflush: cc %ld || mb %p || mbcnt %ld", sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt);
722}
723
724/*
725 * Drop data from (the front of) a sockbuf.
726 */
727void
728sbdrop(sb, len)
729 register struct sockbuf *sb;
730 register int len;
731{
732 register struct mbuf *m, *mn;
733 struct mbuf *next;
734
735 next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
736 while (len > 0) {
737 if (m == 0) {
738 if (next == 0)
739 panic("sbdrop");
740 m = next;
741 next = m->m_nextpkt;
742 continue;
743 }
744 if (m->m_len > len) {
745 m->m_len -= len;
746 m->m_data += len;
747 sb->sb_cc -= len;
748 break;
749 }
750 len -= m->m_len;
751 sbfree(sb, m);
752 MFREE(m, mn);
753 m = mn;
754 }
755 while (m && m->m_len == 0) {
756 sbfree(sb, m);
757 MFREE(m, mn);
758 m = mn;
759 }
760 if (m) {
761 sb->sb_mb = m;
762 m->m_nextpkt = next;
763 } else
764 sb->sb_mb = next;
765}
766
767/*
768 * Drop a record off the front of a sockbuf
769 * and move the next record to the front.
770 */
771void
772sbdroprecord(sb)
773 register struct sockbuf *sb;
774{
775 register struct mbuf *m, *mn;
776
777 m = sb->sb_mb;
778 if (m) {
779 sb->sb_mb = m->m_nextpkt;
780 do {
781 sbfree(sb, m);
782 MFREE(m, mn);
783 m = mn;
784 } while (m);
785 }
786}
787
788/*
789 * Create a "control" mbuf containing the specified data
790 * with the specified type for presentation on a socket buffer.
791 */
792struct mbuf *
793sbcreatecontrol(p, size, type, level)
794 caddr_t p;
795 register int size;
796 int type, level;
797{
798 register struct cmsghdr *cp;
799 struct mbuf *m;
800
801 if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL)
802 return ((struct mbuf *) NULL);
803 cp = mtod(m, struct cmsghdr *);
804 /* XXX check size? */
805 (void)memcpy(CMSG_DATA(cp), p, size);
806 size += sizeof(*cp);
807 m->m_len = size;
808 cp->cmsg_len = size;
809 cp->cmsg_level = level;
810 cp->cmsg_type = type;
811 return (m);
812}
813
814/*
815 * Some routines that return EOPNOTSUPP for entry points that are not
816 * supported by a protocol. Fill in as needed.
817 */
818int
819pru_accept_notsupp(struct socket *so, struct sockaddr **nam)
820{
821 return EOPNOTSUPP;
822}
823
824int
825pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct proc *p)
826{
827 return EOPNOTSUPP;
828}
829
830int
831pru_connect2_notsupp(struct socket *so1, struct socket *so2)
832{
833 return EOPNOTSUPP;
834}
835
836int
837pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data,
838 struct ifnet *ifp, struct proc *p)
839{
840 return EOPNOTSUPP;
841}
842
843int
844pru_listen_notsupp(struct socket *so, struct proc *p)
845{
846 return EOPNOTSUPP;
847}
848
849int
850pru_rcvd_notsupp(struct socket *so, int flags)
851{
852 return EOPNOTSUPP;
853}
854
855int
856pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags)
857{
858 return EOPNOTSUPP;
859}
860
861/*
862 * This isn't really a ``null'' operation, but it's the default one
863 * and doesn't do anything destructive.
864 */
865int
866pru_sense_null(struct socket *so, struct stat *sb)
867{
868 sb->st_blksize = so->so_snd.sb_hiwat;
869 return 0;
870}
871
872/*
873 * Make a copy of a sockaddr in a malloced buffer of type M_SONAME.
874 */
875struct sockaddr *
876dup_sockaddr(sa, canwait)
877 struct sockaddr *sa;
878 int canwait;
879{
880 struct sockaddr *sa2;
881
882 MALLOC(sa2, struct sockaddr *, sa->sa_len, M_SONAME,
883 canwait ? M_WAITOK : M_NOWAIT);
884 if (sa2)
885 bcopy(sa, sa2, sa->sa_len);
886 return sa2;
887}
888
889/*
890 * Create an external-format (``xsocket'') structure using the information
891 * in the kernel-format socket structure pointed to by so. This is done
892 * to reduce the spew of irrelevant information over this interface,
893 * to isolate user code from changes in the kernel structure, and
894 * potentially to provide information-hiding if we decide that
895 * some of this information should be hidden from users.
896 */
897void
898sotoxsocket(struct socket *so, struct xsocket *xso)
899{
900 xso->xso_len = sizeof *xso;
901 xso->xso_so = so;
902 xso->so_type = so->so_type;
903 xso->so_options = so->so_options;
904 xso->so_linger = so->so_linger;
905 xso->so_state = so->so_state;
906 xso->so_pcb = so->so_pcb;
907 xso->xso_protocol = so->so_proto->pr_protocol;
908 xso->xso_family = so->so_proto->pr_domain->dom_family;
909 xso->so_qlen = so->so_qlen;
910 xso->so_incqlen = so->so_incqlen;
911 xso->so_qlimit = so->so_qlimit;
912 xso->so_timeo = so->so_timeo;
913 xso->so_error = so->so_error;
914 xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0;
915 xso->so_oobmark = so->so_oobmark;
916 sbtoxsockbuf(&so->so_snd, &xso->so_snd);
917 sbtoxsockbuf(&so->so_rcv, &xso->so_rcv);
918 xso->so_uid = so->so_uid;
919}
920
921/*
922 * This does the same for sockbufs. Note that the xsockbuf structure,
923 * since it is always embedded in a socket, does not include a self
924 * pointer nor a length. We make this entry point public in case
925 * some other mechanism needs it.
926 */
927void
928sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb)
929{
930 xsb->sb_cc = sb->sb_cc;
931 xsb->sb_hiwat = sb->sb_hiwat;
932 xsb->sb_mbcnt = sb->sb_mbcnt;
933 xsb->sb_mbmax = sb->sb_mbmax;
934 xsb->sb_lowat = sb->sb_lowat;
935 xsb->sb_flags = sb->sb_flags;
936 xsb->sb_timeo = sb->sb_timeo;
937}
938
939/*
940 * Here is the definition of some of the basic objects in the kern.ipc
941 * branch of the MIB.
942 */
943SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC");
944
945/* This takes the place of kern.maxsockbuf, which moved to kern.ipc. */
946static int dummy;
947SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, "");
948
949SYSCTL_INT(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLFLAG_RW, &sb_max, 0, "");
950SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD, &maxsockets, 0, "");
951SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW,
952 &sb_efficiency, 0, "");
953SYSCTL_INT(_kern_ipc, KIPC_NMBCLUSTERS, nmbclusters, CTLFLAG_RD, &nmbclusters, 0, "");
954
2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)uipc_socket2.c 8.1 (Berkeley) 6/10/93
34 * $Id: uipc_socket2.c,v 1.43 1998/12/07 21:58:29 archie Exp $
35 */
36
37#include <sys/param.h>
38#include <sys/systm.h>
39#include <sys/domain.h>
40#include <sys/kernel.h>
41#include <sys/proc.h>
42#include <sys/malloc.h>
43#include <sys/mbuf.h>
44#include <sys/protosw.h>
45#include <sys/stat.h>
46#include <sys/socket.h>
47#include <sys/socketvar.h>
48#include <sys/signalvar.h>
49#include <sys/sysctl.h>
50
51/*
52 * Primitive routines for operating on sockets and socket buffers
53 */
54
55u_long sb_max = SB_MAX; /* XXX should be static */
56
57static u_long sb_efficiency = 8; /* parameter for sbreserve() */
58
59/*
60 * Procedures to manipulate state flags of socket
61 * and do appropriate wakeups. Normal sequence from the
62 * active (originating) side is that soisconnecting() is
63 * called during processing of connect() call,
64 * resulting in an eventual call to soisconnected() if/when the
65 * connection is established. When the connection is torn down
66 * soisdisconnecting() is called during processing of disconnect() call,
67 * and soisdisconnected() is called when the connection to the peer
68 * is totally severed. The semantics of these routines are such that
69 * connectionless protocols can call soisconnected() and soisdisconnected()
70 * only, bypassing the in-progress calls when setting up a ``connection''
71 * takes no time.
72 *
73 * From the passive side, a socket is created with
74 * two queues of sockets: so_q0 for connections in progress
75 * and so_q for connections already made and awaiting user acceptance.
76 * As a protocol is preparing incoming connections, it creates a socket
77 * structure queued on so_q0 by calling sonewconn(). When the connection
78 * is established, soisconnected() is called, and transfers the
79 * socket structure to so_q, making it available to accept().
80 *
81 * If a socket is closed with sockets on either
82 * so_q0 or so_q, these sockets are dropped.
83 *
84 * If higher level protocols are implemented in
85 * the kernel, the wakeups done here will sometimes
86 * cause software-interrupt process scheduling.
87 */
88
89void
90soisconnecting(so)
91 register struct socket *so;
92{
93
94 so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
95 so->so_state |= SS_ISCONNECTING;
96}
97
98void
99soisconnected(so)
100 register struct socket *so;
101{
102 register struct socket *head = so->so_head;
103
104 so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
105 so->so_state |= SS_ISCONNECTED;
106 if (head && (so->so_state & SS_INCOMP)) {
107 TAILQ_REMOVE(&head->so_incomp, so, so_list);
108 head->so_incqlen--;
109 so->so_state &= ~SS_INCOMP;
110 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
111 so->so_state |= SS_COMP;
112 sorwakeup(head);
113 wakeup_one(&head->so_timeo);
114 } else {
115 wakeup(&so->so_timeo);
116 sorwakeup(so);
117 sowwakeup(so);
118 }
119}
120
121void
122soisdisconnecting(so)
123 register struct socket *so;
124{
125
126 so->so_state &= ~SS_ISCONNECTING;
127 so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
128 wakeup((caddr_t)&so->so_timeo);
129 sowwakeup(so);
130 sorwakeup(so);
131}
132
133void
134soisdisconnected(so)
135 register struct socket *so;
136{
137
138 so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
139 so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED);
140 wakeup((caddr_t)&so->so_timeo);
141 sowwakeup(so);
142 sorwakeup(so);
143}
144
145/*
146 * Return a random connection that hasn't been serviced yet and
147 * is eligible for discard. There is a one in qlen chance that
148 * we will return a null, saying that there are no dropable
149 * requests. In this case, the protocol specific code should drop
150 * the new request. This insures fairness.
151 *
152 * This may be used in conjunction with protocol specific queue
153 * congestion routines.
154 */
155struct socket *
156sodropablereq(head)
157 register struct socket *head;
158{
159 register struct socket *so;
160 unsigned int i, j, qlen;
161 static int rnd;
162 static struct timeval old_runtime;
163 static unsigned int cur_cnt, old_cnt;
164 struct timeval tv;
165
166 getmicrouptime(&tv);
167 if ((i = (tv.tv_sec - old_runtime.tv_sec)) != 0) {
168 old_runtime = tv;
169 old_cnt = cur_cnt / i;
170 cur_cnt = 0;
171 }
172
173 so = TAILQ_FIRST(&head->so_incomp);
174 if (!so)
175 return (so);
176
177 qlen = head->so_incqlen;
178 if (++cur_cnt > qlen || old_cnt > qlen) {
179 rnd = (314159 * rnd + 66329) & 0xffff;
180 j = ((qlen + 1) * rnd) >> 16;
181
182 while (j-- && so)
183 so = TAILQ_NEXT(so, so_list);
184 }
185
186 return (so);
187}
188
189/*
190 * When an attempt at a new connection is noted on a socket
191 * which accepts connections, sonewconn is called. If the
192 * connection is possible (subject to space constraints, etc.)
193 * then we allocate a new structure, propoerly linked into the
194 * data structure of the original socket, and return this.
195 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
196 */
197struct socket *
198sonewconn(head, connstatus)
199 register struct socket *head;
200 int connstatus;
201{
202 register struct socket *so;
203
204 if (head->so_qlen > 3 * head->so_qlimit / 2)
205 return ((struct socket *)0);
206 so = soalloc(0);
207 if (so == NULL)
208 return ((struct socket *)0);
209 so->so_head = head;
210 so->so_type = head->so_type;
211 so->so_options = head->so_options &~ SO_ACCEPTCONN;
212 so->so_linger = head->so_linger;
213 so->so_state = head->so_state | SS_NOFDREF;
214 so->so_proto = head->so_proto;
215 so->so_timeo = head->so_timeo;
216 so->so_uid = head->so_uid;
217 (void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat);
218
219 if ((*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) {
220 sodealloc(so);
221 return ((struct socket *)0);
222 }
223
224 if (connstatus) {
225 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
226 so->so_state |= SS_COMP;
227 } else {
228 TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
229 so->so_state |= SS_INCOMP;
230 head->so_incqlen++;
231 }
232 head->so_qlen++;
233 if (connstatus) {
234 sorwakeup(head);
235 wakeup((caddr_t)&head->so_timeo);
236 so->so_state |= connstatus;
237 }
238 return (so);
239}
240
241/*
242 * Socantsendmore indicates that no more data will be sent on the
243 * socket; it would normally be applied to a socket when the user
244 * informs the system that no more data is to be sent, by the protocol
245 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data
246 * will be received, and will normally be applied to the socket by a
247 * protocol when it detects that the peer will send no more data.
248 * Data queued for reading in the socket may yet be read.
249 */
250
251void
252socantsendmore(so)
253 struct socket *so;
254{
255
256 so->so_state |= SS_CANTSENDMORE;
257 sowwakeup(so);
258}
259
260void
261socantrcvmore(so)
262 struct socket *so;
263{
264
265 so->so_state |= SS_CANTRCVMORE;
266 sorwakeup(so);
267}
268
269/*
270 * Wait for data to arrive at/drain from a socket buffer.
271 */
272int
273sbwait(sb)
274 struct sockbuf *sb;
275{
276
277 sb->sb_flags |= SB_WAIT;
278 return (tsleep((caddr_t)&sb->sb_cc,
279 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
280 sb->sb_timeo));
281}
282
283/*
284 * Lock a sockbuf already known to be locked;
285 * return any error returned from sleep (EINTR).
286 */
287int
288sb_lock(sb)
289 register struct sockbuf *sb;
290{
291 int error;
292
293 while (sb->sb_flags & SB_LOCK) {
294 sb->sb_flags |= SB_WANT;
295 error = tsleep((caddr_t)&sb->sb_flags,
296 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH,
297 "sblock", 0);
298 if (error)
299 return (error);
300 }
301 sb->sb_flags |= SB_LOCK;
302 return (0);
303}
304
305/*
306 * Wakeup processes waiting on a socket buffer.
307 * Do asynchronous notification via SIGIO
308 * if the socket has the SS_ASYNC flag set.
309 */
310void
311sowakeup(so, sb)
312 register struct socket *so;
313 register struct sockbuf *sb;
314{
315 selwakeup(&sb->sb_sel);
316 sb->sb_flags &= ~SB_SEL;
317 if (sb->sb_flags & SB_WAIT) {
318 sb->sb_flags &= ~SB_WAIT;
319 wakeup((caddr_t)&sb->sb_cc);
320 }
321 if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL)
322 pgsigio(so->so_sigio, SIGIO, 0);
323 if (sb->sb_flags & SB_UPCALL)
324 (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT);
325}
326
327/*
328 * Socket buffer (struct sockbuf) utility routines.
329 *
330 * Each socket contains two socket buffers: one for sending data and
331 * one for receiving data. Each buffer contains a queue of mbufs,
332 * information about the number of mbufs and amount of data in the
333 * queue, and other fields allowing select() statements and notification
334 * on data availability to be implemented.
335 *
336 * Data stored in a socket buffer is maintained as a list of records.
337 * Each record is a list of mbufs chained together with the m_next
338 * field. Records are chained together with the m_nextpkt field. The upper
339 * level routine soreceive() expects the following conventions to be
340 * observed when placing information in the receive buffer:
341 *
342 * 1. If the protocol requires each message be preceded by the sender's
343 * name, then a record containing that name must be present before
344 * any associated data (mbuf's must be of type MT_SONAME).
345 * 2. If the protocol supports the exchange of ``access rights'' (really
346 * just additional data associated with the message), and there are
347 * ``rights'' to be received, then a record containing this data
348 * should be present (mbuf's must be of type MT_RIGHTS).
349 * 3. If a name or rights record exists, then it must be followed by
350 * a data record, perhaps of zero length.
351 *
352 * Before using a new socket structure it is first necessary to reserve
353 * buffer space to the socket, by calling sbreserve(). This should commit
354 * some of the available buffer space in the system buffer pool for the
355 * socket (currently, it does nothing but enforce limits). The space
356 * should be released by calling sbrelease() when the socket is destroyed.
357 */
358
359int
360soreserve(so, sndcc, rcvcc)
361 register struct socket *so;
362 u_long sndcc, rcvcc;
363{
364
365 if (sbreserve(&so->so_snd, sndcc) == 0)
366 goto bad;
367 if (sbreserve(&so->so_rcv, rcvcc) == 0)
368 goto bad2;
369 if (so->so_rcv.sb_lowat == 0)
370 so->so_rcv.sb_lowat = 1;
371 if (so->so_snd.sb_lowat == 0)
372 so->so_snd.sb_lowat = MCLBYTES;
373 if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
374 so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
375 return (0);
376bad2:
377 sbrelease(&so->so_snd);
378bad:
379 return (ENOBUFS);
380}
381
382/*
383 * Allot mbufs to a sockbuf.
384 * Attempt to scale mbmax so that mbcnt doesn't become limiting
385 * if buffering efficiency is near the normal case.
386 */
387int
388sbreserve(sb, cc)
389 struct sockbuf *sb;
390 u_long cc;
391{
392 if ((u_quad_t)cc > (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES))
393 return (0);
394 sb->sb_hiwat = cc;
395 sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
396 if (sb->sb_lowat > sb->sb_hiwat)
397 sb->sb_lowat = sb->sb_hiwat;
398 return (1);
399}
400
401/*
402 * Free mbufs held by a socket, and reserved mbuf space.
403 */
404void
405sbrelease(sb)
406 struct sockbuf *sb;
407{
408
409 sbflush(sb);
410 sb->sb_hiwat = sb->sb_mbmax = 0;
411}
412
413/*
414 * Routines to add and remove
415 * data from an mbuf queue.
416 *
417 * The routines sbappend() or sbappendrecord() are normally called to
418 * append new mbufs to a socket buffer, after checking that adequate
419 * space is available, comparing the function sbspace() with the amount
420 * of data to be added. sbappendrecord() differs from sbappend() in
421 * that data supplied is treated as the beginning of a new record.
422 * To place a sender's address, optional access rights, and data in a
423 * socket receive buffer, sbappendaddr() should be used. To place
424 * access rights and data in a socket receive buffer, sbappendrights()
425 * should be used. In either case, the new data begins a new record.
426 * Note that unlike sbappend() and sbappendrecord(), these routines check
427 * for the caller that there will be enough space to store the data.
428 * Each fails if there is not enough space, or if it cannot find mbufs
429 * to store additional information in.
430 *
431 * Reliable protocols may use the socket send buffer to hold data
432 * awaiting acknowledgement. Data is normally copied from a socket
433 * send buffer in a protocol with m_copy for output to a peer,
434 * and then removing the data from the socket buffer with sbdrop()
435 * or sbdroprecord() when the data is acknowledged by the peer.
436 */
437
438/*
439 * Append mbuf chain m to the last record in the
440 * socket buffer sb. The additional space associated
441 * the mbuf chain is recorded in sb. Empty mbufs are
442 * discarded and mbufs are compacted where possible.
443 */
444void
445sbappend(sb, m)
446 struct sockbuf *sb;
447 struct mbuf *m;
448{
449 register struct mbuf *n;
450
451 if (m == 0)
452 return;
453 n = sb->sb_mb;
454 if (n) {
455 while (n->m_nextpkt)
456 n = n->m_nextpkt;
457 do {
458 if (n->m_flags & M_EOR) {
459 sbappendrecord(sb, m); /* XXXXXX!!!! */
460 return;
461 }
462 } while (n->m_next && (n = n->m_next));
463 }
464 sbcompress(sb, m, n);
465}
466
467#ifdef SOCKBUF_DEBUG
468void
469sbcheck(sb)
470 register struct sockbuf *sb;
471{
472 register struct mbuf *m;
473 register struct mbuf *n = 0;
474 register u_long len = 0, mbcnt = 0;
475
476 for (m = sb->sb_mb; m; m = n) {
477 n = m->m_nextpkt;
478 for (; m; m = m->m_next) {
479 len += m->m_len;
480 mbcnt += MSIZE;
481 if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */
482 mbcnt += m->m_ext.ext_size;
483 }
484 }
485 if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
486 printf("cc %ld != %ld || mbcnt %ld != %ld\n", len, sb->sb_cc,
487 mbcnt, sb->sb_mbcnt);
488 panic("sbcheck");
489 }
490}
491#endif
492
493/*
494 * As above, except the mbuf chain
495 * begins a new record.
496 */
497void
498sbappendrecord(sb, m0)
499 register struct sockbuf *sb;
500 register struct mbuf *m0;
501{
502 register struct mbuf *m;
503
504 if (m0 == 0)
505 return;
506 m = sb->sb_mb;
507 if (m)
508 while (m->m_nextpkt)
509 m = m->m_nextpkt;
510 /*
511 * Put the first mbuf on the queue.
512 * Note this permits zero length records.
513 */
514 sballoc(sb, m0);
515 if (m)
516 m->m_nextpkt = m0;
517 else
518 sb->sb_mb = m0;
519 m = m0->m_next;
520 m0->m_next = 0;
521 if (m && (m0->m_flags & M_EOR)) {
522 m0->m_flags &= ~M_EOR;
523 m->m_flags |= M_EOR;
524 }
525 sbcompress(sb, m, m0);
526}
527
528/*
529 * As above except that OOB data
530 * is inserted at the beginning of the sockbuf,
531 * but after any other OOB data.
532 */
533void
534sbinsertoob(sb, m0)
535 register struct sockbuf *sb;
536 register struct mbuf *m0;
537{
538 register struct mbuf *m;
539 register struct mbuf **mp;
540
541 if (m0 == 0)
542 return;
543 for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) {
544 m = *mp;
545 again:
546 switch (m->m_type) {
547
548 case MT_OOBDATA:
549 continue; /* WANT next train */
550
551 case MT_CONTROL:
552 m = m->m_next;
553 if (m)
554 goto again; /* inspect THIS train further */
555 }
556 break;
557 }
558 /*
559 * Put the first mbuf on the queue.
560 * Note this permits zero length records.
561 */
562 sballoc(sb, m0);
563 m0->m_nextpkt = *mp;
564 *mp = m0;
565 m = m0->m_next;
566 m0->m_next = 0;
567 if (m && (m0->m_flags & M_EOR)) {
568 m0->m_flags &= ~M_EOR;
569 m->m_flags |= M_EOR;
570 }
571 sbcompress(sb, m, m0);
572}
573
574/*
575 * Append address and data, and optionally, control (ancillary) data
576 * to the receive queue of a socket. If present,
577 * m0 must include a packet header with total length.
578 * Returns 0 if no space in sockbuf or insufficient mbufs.
579 */
580int
581sbappendaddr(sb, asa, m0, control)
582 register struct sockbuf *sb;
583 struct sockaddr *asa;
584 struct mbuf *m0, *control;
585{
586 register struct mbuf *m, *n;
587 int space = asa->sa_len;
588
589if (m0 && (m0->m_flags & M_PKTHDR) == 0)
590panic("sbappendaddr");
591 if (m0)
592 space += m0->m_pkthdr.len;
593 for (n = control; n; n = n->m_next) {
594 space += n->m_len;
595 if (n->m_next == 0) /* keep pointer to last control buf */
596 break;
597 }
598 if (space > sbspace(sb))
599 return (0);
600 if (asa->sa_len > MLEN)
601 return (0);
602 MGET(m, M_DONTWAIT, MT_SONAME);
603 if (m == 0)
604 return (0);
605 m->m_len = asa->sa_len;
606 bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
607 if (n)
608 n->m_next = m0; /* concatenate data to control */
609 else
610 control = m0;
611 m->m_next = control;
612 for (n = m; n; n = n->m_next)
613 sballoc(sb, n);
614 n = sb->sb_mb;
615 if (n) {
616 while (n->m_nextpkt)
617 n = n->m_nextpkt;
618 n->m_nextpkt = m;
619 } else
620 sb->sb_mb = m;
621 return (1);
622}
623
624int
625sbappendcontrol(sb, m0, control)
626 struct sockbuf *sb;
627 struct mbuf *control, *m0;
628{
629 register struct mbuf *m, *n;
630 int space = 0;
631
632 if (control == 0)
633 panic("sbappendcontrol");
634 for (m = control; ; m = m->m_next) {
635 space += m->m_len;
636 if (m->m_next == 0)
637 break;
638 }
639 n = m; /* save pointer to last control buffer */
640 for (m = m0; m; m = m->m_next)
641 space += m->m_len;
642 if (space > sbspace(sb))
643 return (0);
644 n->m_next = m0; /* concatenate data to control */
645 for (m = control; m; m = m->m_next)
646 sballoc(sb, m);
647 n = sb->sb_mb;
648 if (n) {
649 while (n->m_nextpkt)
650 n = n->m_nextpkt;
651 n->m_nextpkt = control;
652 } else
653 sb->sb_mb = control;
654 return (1);
655}
656
657/*
658 * Compress mbuf chain m into the socket
659 * buffer sb following mbuf n. If n
660 * is null, the buffer is presumed empty.
661 */
662void
663sbcompress(sb, m, n)
664 register struct sockbuf *sb;
665 register struct mbuf *m, *n;
666{
667 register int eor = 0;
668 register struct mbuf *o;
669
670 while (m) {
671 eor |= m->m_flags & M_EOR;
672 if (m->m_len == 0 &&
673 (eor == 0 ||
674 (((o = m->m_next) || (o = n)) &&
675 o->m_type == m->m_type))) {
676 m = m_free(m);
677 continue;
678 }
679 if (n && (n->m_flags & (M_EXT | M_EOR)) == 0 &&
680 (n->m_data + n->m_len + m->m_len) < &n->m_dat[MLEN] &&
681 n->m_type == m->m_type) {
682 bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
683 (unsigned)m->m_len);
684 n->m_len += m->m_len;
685 sb->sb_cc += m->m_len;
686 m = m_free(m);
687 continue;
688 }
689 if (n)
690 n->m_next = m;
691 else
692 sb->sb_mb = m;
693 sballoc(sb, m);
694 n = m;
695 m->m_flags &= ~M_EOR;
696 m = m->m_next;
697 n->m_next = 0;
698 }
699 if (eor) {
700 if (n)
701 n->m_flags |= eor;
702 else
703 printf("semi-panic: sbcompress\n");
704 }
705}
706
707/*
708 * Free all mbufs in a sockbuf.
709 * Check that all resources are reclaimed.
710 */
711void
712sbflush(sb)
713 register struct sockbuf *sb;
714{
715
716 if (sb->sb_flags & SB_LOCK)
717 panic("sbflush: locked");
718 while (sb->sb_mbcnt && sb->sb_cc)
719 sbdrop(sb, (int)sb->sb_cc);
720 if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt)
721 panic("sbflush: cc %ld || mb %p || mbcnt %ld", sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt);
722}
723
724/*
725 * Drop data from (the front of) a sockbuf.
726 */
727void
728sbdrop(sb, len)
729 register struct sockbuf *sb;
730 register int len;
731{
732 register struct mbuf *m, *mn;
733 struct mbuf *next;
734
735 next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
736 while (len > 0) {
737 if (m == 0) {
738 if (next == 0)
739 panic("sbdrop");
740 m = next;
741 next = m->m_nextpkt;
742 continue;
743 }
744 if (m->m_len > len) {
745 m->m_len -= len;
746 m->m_data += len;
747 sb->sb_cc -= len;
748 break;
749 }
750 len -= m->m_len;
751 sbfree(sb, m);
752 MFREE(m, mn);
753 m = mn;
754 }
755 while (m && m->m_len == 0) {
756 sbfree(sb, m);
757 MFREE(m, mn);
758 m = mn;
759 }
760 if (m) {
761 sb->sb_mb = m;
762 m->m_nextpkt = next;
763 } else
764 sb->sb_mb = next;
765}
766
767/*
768 * Drop a record off the front of a sockbuf
769 * and move the next record to the front.
770 */
771void
772sbdroprecord(sb)
773 register struct sockbuf *sb;
774{
775 register struct mbuf *m, *mn;
776
777 m = sb->sb_mb;
778 if (m) {
779 sb->sb_mb = m->m_nextpkt;
780 do {
781 sbfree(sb, m);
782 MFREE(m, mn);
783 m = mn;
784 } while (m);
785 }
786}
787
788/*
789 * Create a "control" mbuf containing the specified data
790 * with the specified type for presentation on a socket buffer.
791 */
792struct mbuf *
793sbcreatecontrol(p, size, type, level)
794 caddr_t p;
795 register int size;
796 int type, level;
797{
798 register struct cmsghdr *cp;
799 struct mbuf *m;
800
801 if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL)
802 return ((struct mbuf *) NULL);
803 cp = mtod(m, struct cmsghdr *);
804 /* XXX check size? */
805 (void)memcpy(CMSG_DATA(cp), p, size);
806 size += sizeof(*cp);
807 m->m_len = size;
808 cp->cmsg_len = size;
809 cp->cmsg_level = level;
810 cp->cmsg_type = type;
811 return (m);
812}
813
814/*
815 * Some routines that return EOPNOTSUPP for entry points that are not
816 * supported by a protocol. Fill in as needed.
817 */
818int
819pru_accept_notsupp(struct socket *so, struct sockaddr **nam)
820{
821 return EOPNOTSUPP;
822}
823
824int
825pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct proc *p)
826{
827 return EOPNOTSUPP;
828}
829
830int
831pru_connect2_notsupp(struct socket *so1, struct socket *so2)
832{
833 return EOPNOTSUPP;
834}
835
836int
837pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data,
838 struct ifnet *ifp, struct proc *p)
839{
840 return EOPNOTSUPP;
841}
842
843int
844pru_listen_notsupp(struct socket *so, struct proc *p)
845{
846 return EOPNOTSUPP;
847}
848
849int
850pru_rcvd_notsupp(struct socket *so, int flags)
851{
852 return EOPNOTSUPP;
853}
854
855int
856pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags)
857{
858 return EOPNOTSUPP;
859}
860
861/*
862 * This isn't really a ``null'' operation, but it's the default one
863 * and doesn't do anything destructive.
864 */
865int
866pru_sense_null(struct socket *so, struct stat *sb)
867{
868 sb->st_blksize = so->so_snd.sb_hiwat;
869 return 0;
870}
871
872/*
873 * Make a copy of a sockaddr in a malloced buffer of type M_SONAME.
874 */
875struct sockaddr *
876dup_sockaddr(sa, canwait)
877 struct sockaddr *sa;
878 int canwait;
879{
880 struct sockaddr *sa2;
881
882 MALLOC(sa2, struct sockaddr *, sa->sa_len, M_SONAME,
883 canwait ? M_WAITOK : M_NOWAIT);
884 if (sa2)
885 bcopy(sa, sa2, sa->sa_len);
886 return sa2;
887}
888
889/*
890 * Create an external-format (``xsocket'') structure using the information
891 * in the kernel-format socket structure pointed to by so. This is done
892 * to reduce the spew of irrelevant information over this interface,
893 * to isolate user code from changes in the kernel structure, and
894 * potentially to provide information-hiding if we decide that
895 * some of this information should be hidden from users.
896 */
897void
898sotoxsocket(struct socket *so, struct xsocket *xso)
899{
900 xso->xso_len = sizeof *xso;
901 xso->xso_so = so;
902 xso->so_type = so->so_type;
903 xso->so_options = so->so_options;
904 xso->so_linger = so->so_linger;
905 xso->so_state = so->so_state;
906 xso->so_pcb = so->so_pcb;
907 xso->xso_protocol = so->so_proto->pr_protocol;
908 xso->xso_family = so->so_proto->pr_domain->dom_family;
909 xso->so_qlen = so->so_qlen;
910 xso->so_incqlen = so->so_incqlen;
911 xso->so_qlimit = so->so_qlimit;
912 xso->so_timeo = so->so_timeo;
913 xso->so_error = so->so_error;
914 xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0;
915 xso->so_oobmark = so->so_oobmark;
916 sbtoxsockbuf(&so->so_snd, &xso->so_snd);
917 sbtoxsockbuf(&so->so_rcv, &xso->so_rcv);
918 xso->so_uid = so->so_uid;
919}
920
921/*
922 * This does the same for sockbufs. Note that the xsockbuf structure,
923 * since it is always embedded in a socket, does not include a self
924 * pointer nor a length. We make this entry point public in case
925 * some other mechanism needs it.
926 */
927void
928sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb)
929{
930 xsb->sb_cc = sb->sb_cc;
931 xsb->sb_hiwat = sb->sb_hiwat;
932 xsb->sb_mbcnt = sb->sb_mbcnt;
933 xsb->sb_mbmax = sb->sb_mbmax;
934 xsb->sb_lowat = sb->sb_lowat;
935 xsb->sb_flags = sb->sb_flags;
936 xsb->sb_timeo = sb->sb_timeo;
937}
938
939/*
940 * Here is the definition of some of the basic objects in the kern.ipc
941 * branch of the MIB.
942 */
943SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC");
944
945/* This takes the place of kern.maxsockbuf, which moved to kern.ipc. */
946static int dummy;
947SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, "");
948
949SYSCTL_INT(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLFLAG_RW, &sb_max, 0, "");
950SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD, &maxsockets, 0, "");
951SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW,
952 &sb_efficiency, 0, "");
953SYSCTL_INT(_kern_ipc, KIPC_NMBCLUSTERS, nmbclusters, CTLFLAG_RD, &nmbclusters, 0, "");
954