uipc_socket2.c revision 1.22 1 /* $NetBSD: uipc_socket2.c,v 1.22 1998/01/07 23:47:09 thorpej Exp $ */
2
3 /*
4 * Copyright (c) 1982, 1986, 1988, 1990, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. All advertising materials mentioning features or use of this software
16 * must display the following acknowledgement:
17 * This product includes software developed by the University of
18 * California, Berkeley and its contributors.
19 * 4. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 *
35 * @(#)uipc_socket2.c 8.1 (Berkeley) 6/10/93
36 */
37
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/proc.h>
41 #include <sys/file.h>
42 #include <sys/buf.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/protosw.h>
46 #include <sys/socket.h>
47 #include <sys/socketvar.h>
48 #include <sys/signalvar.h>
49
50 /*
51 * Primitive routines for operating on sockets and socket buffers
52 */
53
54 /* strings for sleep message: */
55 const char netio[] = "netio";
56 const char netcon[] = "netcon";
57 const char netcls[] = "netcls";
58
59 u_long sb_max = SB_MAX; /* patchable */
60
61 /*
62 * Procedures to manipulate state flags of socket
63 * and do appropriate wakeups. Normal sequence from the
64 * active (originating) side is that soisconnecting() is
65 * called during processing of connect() call,
66 * resulting in an eventual call to soisconnected() if/when the
67 * connection is established. When the connection is torn down
68 * soisdisconnecting() is called during processing of disconnect() call,
69 * and soisdisconnected() is called when the connection to the peer
70 * is totally severed. The semantics of these routines are such that
71 * connectionless protocols can call soisconnected() and soisdisconnected()
72 * only, bypassing the in-progress calls when setting up a ``connection''
73 * takes no time.
74 *
75 * From the passive side, a socket is created with
76 * two queues of sockets: so_q0 for connections in progress
77 * and so_q for connections already made and awaiting user acceptance.
78 * As a protocol is preparing incoming connections, it creates a socket
79 * structure queued on so_q0 by calling sonewconn(). When the connection
80 * is established, soisconnected() is called, and transfers the
81 * socket structure to so_q, making it available to accept().
82 *
83 * If a socket is closed with sockets on either
84 * so_q0 or so_q, these sockets are dropped.
85 *
86 * If higher level protocols are implemented in
87 * the kernel, the wakeups done here will sometimes
88 * cause software-interrupt process scheduling.
89 */
90
91 void
92 soisconnecting(so)
93 register struct socket *so;
94 {
95
96 so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
97 so->so_state |= SS_ISCONNECTING;
98 }
99
100 void
101 soisconnected(so)
102 register struct socket *so;
103 {
104 register struct socket *head = so->so_head;
105
106 so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
107 so->so_state |= SS_ISCONNECTED;
108 if (head && soqremque(so, 0)) {
109 soqinsque(head, so, 1);
110 sorwakeup(head);
111 wakeup((caddr_t)&head->so_timeo);
112 } else {
113 wakeup((caddr_t)&so->so_timeo);
114 sorwakeup(so);
115 sowwakeup(so);
116 }
117 }
118
119 void
120 soisdisconnecting(so)
121 register struct socket *so;
122 {
123
124 so->so_state &= ~SS_ISCONNECTING;
125 so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
126 wakeup((caddr_t)&so->so_timeo);
127 sowwakeup(so);
128 sorwakeup(so);
129 }
130
131 void
132 soisdisconnected(so)
133 register struct socket *so;
134 {
135
136 so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
137 so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE);
138 wakeup((caddr_t)&so->so_timeo);
139 sowwakeup(so);
140 sorwakeup(so);
141 }
142
143 /*
144 * When an attempt at a new connection is noted on a socket
145 * which accepts connections, sonewconn is called. If the
146 * connection is possible (subject to space constraints, etc.)
147 * then we allocate a new structure, propoerly linked into the
148 * data structure of the original socket, and return this.
149 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
150 *
151 * Currently, sonewconn() is defined as sonewconn1() in socketvar.h
152 * to catch calls that are missing the (new) second parameter.
153 */
154 struct socket *
155 sonewconn1(head, connstatus)
156 register struct socket *head;
157 int connstatus;
158 {
159 register struct socket *so;
160 int soqueue = connstatus ? 1 : 0;
161
162 if (head->so_qlen + head->so_q0len > 3 * head->so_qlimit / 2)
163 return ((struct socket *)0);
164 MALLOC(so, struct socket *, sizeof(*so), M_SOCKET, M_DONTWAIT);
165 if (so == NULL)
166 return ((struct socket *)0);
167 bzero((caddr_t)so, sizeof(*so));
168 so->so_type = head->so_type;
169 so->so_options = head->so_options &~ SO_ACCEPTCONN;
170 so->so_linger = head->so_linger;
171 so->so_state = head->so_state | SS_NOFDREF;
172 so->so_proto = head->so_proto;
173 so->so_timeo = head->so_timeo;
174 so->so_pgid = head->so_pgid;
175 (void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat);
176 soqinsque(head, so, soqueue);
177 if ((*so->so_proto->pr_usrreq)(so, PRU_ATTACH,
178 (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
179 (struct proc *)0)) {
180 (void) soqremque(so, soqueue);
181 (void) free((caddr_t)so, M_SOCKET);
182 return ((struct socket *)0);
183 }
184 if (connstatus) {
185 sorwakeup(head);
186 wakeup((caddr_t)&head->so_timeo);
187 so->so_state |= connstatus;
188 }
189 return (so);
190 }
191
192 void
193 soqinsque(head, so, q)
194 register struct socket *head, *so;
195 int q;
196 {
197
198 #ifdef DIAGNOSTIC
199 if (so->so_onq != NULL)
200 panic("soqinsque");
201 #endif
202
203 so->so_head = head;
204 if (q == 0) {
205 head->so_q0len++;
206 so->so_onq = &head->so_q0;
207 } else {
208 head->so_qlen++;
209 so->so_onq = &head->so_q;
210 }
211 TAILQ_INSERT_TAIL(so->so_onq, so, so_qe);
212 }
213
214 int
215 soqremque(so, q)
216 register struct socket *so;
217 int q;
218 {
219 struct socket *head = so->so_head;
220
221 if (q == 0) {
222 if (so->so_onq != &head->so_q0)
223 return (0);
224 head->so_q0len--;
225 } else {
226 if (so->so_onq != &head->so_q)
227 return (0);
228 head->so_qlen--;
229 }
230 TAILQ_REMOVE(so->so_onq, so, so_qe);
231 so->so_onq = NULL;
232 so->so_head = NULL;
233 return (1);
234 }
235
236 /*
237 * Socantsendmore indicates that no more data will be sent on the
238 * socket; it would normally be applied to a socket when the user
239 * informs the system that no more data is to be sent, by the protocol
240 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data
241 * will be received, and will normally be applied to the socket by a
242 * protocol when it detects that the peer will send no more data.
243 * Data queued for reading in the socket may yet be read.
244 */
245
246 void
247 socantsendmore(so)
248 struct socket *so;
249 {
250
251 so->so_state |= SS_CANTSENDMORE;
252 sowwakeup(so);
253 }
254
255 void
256 socantrcvmore(so)
257 struct socket *so;
258 {
259
260 so->so_state |= SS_CANTRCVMORE;
261 sorwakeup(so);
262 }
263
264 /*
265 * Wait for data to arrive at/drain from a socket buffer.
266 */
267 int
268 sbwait(sb)
269 struct sockbuf *sb;
270 {
271
272 sb->sb_flags |= SB_WAIT;
273 return (tsleep((caddr_t)&sb->sb_cc,
274 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, netio,
275 sb->sb_timeo));
276 }
277
278 /*
279 * Lock a sockbuf already known to be locked;
280 * return any error returned from sleep (EINTR).
281 */
282 int
283 sb_lock(sb)
284 register struct sockbuf *sb;
285 {
286 int error;
287
288 while (sb->sb_flags & SB_LOCK) {
289 sb->sb_flags |= SB_WANT;
290 error = tsleep((caddr_t)&sb->sb_flags,
291 (sb->sb_flags & SB_NOINTR) ?
292 PSOCK : PSOCK|PCATCH, netio, 0);
293 if (error)
294 return (error);
295 }
296 sb->sb_flags |= SB_LOCK;
297 return (0);
298 }
299
300 /*
301 * Wakeup processes waiting on a socket buffer.
302 * Do asynchronous notification via SIGIO
303 * if the socket has the SS_ASYNC flag set.
304 */
305 void
306 sowakeup(so, sb)
307 register struct socket *so;
308 register struct sockbuf *sb;
309 {
310 struct proc *p;
311
312 selwakeup(&sb->sb_sel);
313 sb->sb_flags &= ~SB_SEL;
314 if (sb->sb_flags & SB_WAIT) {
315 sb->sb_flags &= ~SB_WAIT;
316 wakeup((caddr_t)&sb->sb_cc);
317 }
318 if (so->so_state & SS_ASYNC) {
319 if (so->so_pgid < 0)
320 gsignal(-so->so_pgid, SIGIO);
321 else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
322 psignal(p, SIGIO);
323 }
324 }
325
326 /*
327 * Socket buffer (struct sockbuf) utility routines.
328 *
329 * Each socket contains two socket buffers: one for sending data and
330 * one for receiving data. Each buffer contains a queue of mbufs,
331 * information about the number of mbufs and amount of data in the
332 * queue, and other fields allowing poll() statements and notification
333 * on data availability to be implemented.
334 *
335 * Data stored in a socket buffer is maintained as a list of records.
336 * Each record is a list of mbufs chained together with the m_next
337 * field. Records are chained together with the m_nextpkt field. The upper
338 * level routine soreceive() expects the following conventions to be
339 * observed when placing information in the receive buffer:
340 *
341 * 1. If the protocol requires each message be preceded by the sender's
342 * name, then a record containing that name must be present before
343 * any associated data (mbuf's must be of type MT_SONAME).
344 * 2. If the protocol supports the exchange of ``access rights'' (really
345 * just additional data associated with the message), and there are
346 * ``rights'' to be received, then a record containing this data
347 * should be present (mbuf's must be of type MT_CONTROL).
348 * 3. If a name or rights record exists, then it must be followed by
349 * a data record, perhaps of zero length.
350 *
351 * Before using a new socket structure it is first necessary to reserve
352 * buffer space to the socket, by calling sbreserve(). This should commit
353 * some of the available buffer space in the system buffer pool for the
354 * socket (currently, it does nothing but enforce limits). The space
355 * should be released by calling sbrelease() when the socket is destroyed.
356 */
357
358 int
359 soreserve(so, sndcc, rcvcc)
360 register struct socket *so;
361 u_long sndcc, rcvcc;
362 {
363
364 if (sbreserve(&so->so_snd, sndcc) == 0)
365 goto bad;
366 if (sbreserve(&so->so_rcv, rcvcc) == 0)
367 goto bad2;
368 if (so->so_rcv.sb_lowat == 0)
369 so->so_rcv.sb_lowat = 1;
370 if (so->so_snd.sb_lowat == 0)
371 so->so_snd.sb_lowat = MCLBYTES;
372 if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
373 so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
374 return (0);
375 bad2:
376 sbrelease(&so->so_snd);
377 bad:
378 return (ENOBUFS);
379 }
380
381 /*
382 * Allot mbufs to a sockbuf.
383 * Attempt to scale mbmax so that mbcnt doesn't become limiting
384 * if buffering efficiency is near the normal case.
385 */
386 int
387 sbreserve(sb, cc)
388 struct sockbuf *sb;
389 u_long cc;
390 {
391
392 if (cc == 0 || cc > sb_max * MCLBYTES / (MSIZE + MCLBYTES))
393 return (0);
394 sb->sb_hiwat = cc;
395 sb->sb_mbmax = min(cc * 2, 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 */
404 void
405 sbrelease(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 */
444 void
445 sbappend(sb, m)
446 struct sockbuf *sb;
447 struct mbuf *m;
448 {
449 register struct mbuf *n;
450
451 if (m == 0)
452 return;
453 if ((n = sb->sb_mb) != NULL) {
454 while (n->m_nextpkt)
455 n = n->m_nextpkt;
456 do {
457 if (n->m_flags & M_EOR) {
458 sbappendrecord(sb, m); /* XXXXXX!!!! */
459 return;
460 }
461 } while (n->m_next && (n = n->m_next));
462 }
463 sbcompress(sb, m, n);
464 }
465
466 #ifdef SOCKBUF_DEBUG
467 void
468 sbcheck(sb)
469 register struct sockbuf *sb;
470 {
471 register struct mbuf *m;
472 register int len = 0, mbcnt = 0;
473
474 for (m = sb->sb_mb; m; m = m->m_next) {
475 len += m->m_len;
476 mbcnt += MSIZE;
477 if (m->m_flags & M_EXT)
478 mbcnt += m->m_ext.ext_size;
479 if (m->m_nextpkt)
480 panic("sbcheck nextpkt");
481 }
482 if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
483 printf("cc %d != %d || mbcnt %d != %d\n", len, sb->sb_cc,
484 mbcnt, sb->sb_mbcnt);
485 panic("sbcheck");
486 }
487 }
488 #endif
489
490 /*
491 * As above, except the mbuf chain
492 * begins a new record.
493 */
494 void
495 sbappendrecord(sb, m0)
496 register struct sockbuf *sb;
497 register struct mbuf *m0;
498 {
499 register struct mbuf *m;
500
501 if (m0 == 0)
502 return;
503 if ((m = sb->sb_mb) != NULL)
504 while (m->m_nextpkt)
505 m = m->m_nextpkt;
506 /*
507 * Put the first mbuf on the queue.
508 * Note this permits zero length records.
509 */
510 sballoc(sb, m0);
511 if (m)
512 m->m_nextpkt = m0;
513 else
514 sb->sb_mb = m0;
515 m = m0->m_next;
516 m0->m_next = 0;
517 if (m && (m0->m_flags & M_EOR)) {
518 m0->m_flags &= ~M_EOR;
519 m->m_flags |= M_EOR;
520 }
521 sbcompress(sb, m, m0);
522 }
523
524 /*
525 * As above except that OOB data
526 * is inserted at the beginning of the sockbuf,
527 * but after any other OOB data.
528 */
529 void
530 sbinsertoob(sb, m0)
531 register struct sockbuf *sb;
532 register struct mbuf *m0;
533 {
534 register struct mbuf *m;
535 register struct mbuf **mp;
536
537 if (m0 == 0)
538 return;
539 for (mp = &sb->sb_mb; (m = *mp) != NULL; mp = &((*mp)->m_nextpkt)) {
540 again:
541 switch (m->m_type) {
542
543 case MT_OOBDATA:
544 continue; /* WANT next train */
545
546 case MT_CONTROL:
547 if ((m = m->m_next) != NULL)
548 goto again; /* inspect THIS train further */
549 }
550 break;
551 }
552 /*
553 * Put the first mbuf on the queue.
554 * Note this permits zero length records.
555 */
556 sballoc(sb, m0);
557 m0->m_nextpkt = *mp;
558 *mp = m0;
559 m = m0->m_next;
560 m0->m_next = 0;
561 if (m && (m0->m_flags & M_EOR)) {
562 m0->m_flags &= ~M_EOR;
563 m->m_flags |= M_EOR;
564 }
565 sbcompress(sb, m, m0);
566 }
567
568 /*
569 * Append address and data, and optionally, control (ancillary) data
570 * to the receive queue of a socket. If present,
571 * m0 must include a packet header with total length.
572 * Returns 0 if no space in sockbuf or insufficient mbufs.
573 */
574 int
575 sbappendaddr(sb, asa, m0, control)
576 register struct sockbuf *sb;
577 struct sockaddr *asa;
578 struct mbuf *m0, *control;
579 {
580 register struct mbuf *m, *n;
581 int space = asa->sa_len;
582
583 if (m0 && (m0->m_flags & M_PKTHDR) == 0)
584 panic("sbappendaddr");
585 if (m0)
586 space += m0->m_pkthdr.len;
587 for (n = control; n; n = n->m_next) {
588 space += n->m_len;
589 if (n->m_next == 0) /* keep pointer to last control buf */
590 break;
591 }
592 if (space > sbspace(sb))
593 return (0);
594 MGET(m, M_DONTWAIT, MT_SONAME);
595 if (m == 0)
596 return (0);
597 if (asa->sa_len > MLEN) {
598 MEXTMALLOC(m, asa->sa_len, M_NOWAIT);
599 if ((m->m_flags & M_EXT) == 0) {
600 m_free(m);
601 return (0);
602 }
603 }
604 m->m_len = asa->sa_len;
605 bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
606 if (n)
607 n->m_next = m0; /* concatenate data to control */
608 else
609 control = m0;
610 m->m_next = control;
611 for (n = m; n; n = n->m_next)
612 sballoc(sb, n);
613 if ((n = sb->sb_mb) != NULL) {
614 while (n->m_nextpkt)
615 n = n->m_nextpkt;
616 n->m_nextpkt = m;
617 } else
618 sb->sb_mb = m;
619 return (1);
620 }
621
622 int
623 sbappendcontrol(sb, m0, control)
624 struct sockbuf *sb;
625 struct mbuf *m0, *control;
626 {
627 register struct mbuf *m, *n;
628 int space = 0;
629
630 if (control == 0)
631 panic("sbappendcontrol");
632 for (m = control; ; m = m->m_next) {
633 space += m->m_len;
634 if (m->m_next == 0)
635 break;
636 }
637 n = m; /* save pointer to last control buffer */
638 for (m = m0; m; m = m->m_next)
639 space += m->m_len;
640 if (space > sbspace(sb))
641 return (0);
642 n->m_next = m0; /* concatenate data to control */
643 for (m = control; m; m = m->m_next)
644 sballoc(sb, m);
645 if ((n = sb->sb_mb) != NULL) {
646 while (n->m_nextpkt)
647 n = n->m_nextpkt;
648 n->m_nextpkt = control;
649 } else
650 sb->sb_mb = control;
651 return (1);
652 }
653
654 /*
655 * Compress mbuf chain m into the socket
656 * buffer sb following mbuf n. If n
657 * is null, the buffer is presumed empty.
658 */
659 void
660 sbcompress(sb, m, n)
661 register struct sockbuf *sb;
662 register struct mbuf *m, *n;
663 {
664 register int eor = 0;
665 register struct mbuf *o;
666
667 while (m) {
668 eor |= m->m_flags & M_EOR;
669 if (m->m_len == 0 &&
670 (eor == 0 ||
671 (((o = m->m_next) || (o = n)) &&
672 o->m_type == m->m_type))) {
673 m = m_free(m);
674 continue;
675 }
676 if (n && (n->m_flags & (M_EXT | M_EOR)) == 0 &&
677 (n->m_data + n->m_len + m->m_len) < &n->m_dat[MLEN] &&
678 n->m_type == m->m_type) {
679 bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
680 (unsigned)m->m_len);
681 n->m_len += m->m_len;
682 sb->sb_cc += m->m_len;
683 m = m_free(m);
684 continue;
685 }
686 if (n)
687 n->m_next = m;
688 else
689 sb->sb_mb = m;
690 sballoc(sb, m);
691 n = m;
692 m->m_flags &= ~M_EOR;
693 m = m->m_next;
694 n->m_next = 0;
695 }
696 if (eor) {
697 if (n)
698 n->m_flags |= eor;
699 else
700 printf("semi-panic: sbcompress\n");
701 }
702 }
703
704 /*
705 * Free all mbufs in a sockbuf.
706 * Check that all resources are reclaimed.
707 */
708 void
709 sbflush(sb)
710 register struct sockbuf *sb;
711 {
712
713 if (sb->sb_flags & SB_LOCK)
714 panic("sbflush");
715 while (sb->sb_mbcnt)
716 sbdrop(sb, (int)sb->sb_cc);
717 if (sb->sb_cc || sb->sb_mb)
718 panic("sbflush 2");
719 }
720
721 /*
722 * Drop data from (the front of) a sockbuf.
723 */
724 void
725 sbdrop(sb, len)
726 register struct sockbuf *sb;
727 register int len;
728 {
729 register struct mbuf *m, *mn;
730 struct mbuf *next;
731
732 next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
733 while (len > 0) {
734 if (m == 0) {
735 if (next == 0)
736 panic("sbdrop");
737 m = next;
738 next = m->m_nextpkt;
739 continue;
740 }
741 if (m->m_len > len) {
742 m->m_len -= len;
743 m->m_data += len;
744 sb->sb_cc -= len;
745 break;
746 }
747 len -= m->m_len;
748 sbfree(sb, m);
749 MFREE(m, mn);
750 m = mn;
751 }
752 while (m && m->m_len == 0) {
753 sbfree(sb, m);
754 MFREE(m, mn);
755 m = mn;
756 }
757 if (m) {
758 sb->sb_mb = m;
759 m->m_nextpkt = next;
760 } else
761 sb->sb_mb = next;
762 }
763
764 /*
765 * Drop a record off the front of a sockbuf
766 * and move the next record to the front.
767 */
768 void
769 sbdroprecord(sb)
770 register struct sockbuf *sb;
771 {
772 register struct mbuf *m, *mn;
773
774 m = sb->sb_mb;
775 if (m) {
776 sb->sb_mb = m->m_nextpkt;
777 do {
778 sbfree(sb, m);
779 MFREE(m, mn);
780 } while ((m = mn) != NULL);
781 }
782 }
783
784 /*
785 * Create a "control" mbuf containing the specified data
786 * with the specified type for presentation on a socket buffer.
787 */
788 struct mbuf *
789 sbcreatecontrol(p, size, type, level)
790 caddr_t p;
791 register int size;
792 int type, level;
793 {
794 register struct cmsghdr *cp;
795 struct mbuf *m;
796
797 if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL)
798 return ((struct mbuf *) NULL);
799 cp = mtod(m, struct cmsghdr *);
800 bcopy(p, CMSG_DATA(cp), size);
801 size += sizeof(*cp);
802 m->m_len = size;
803 cp->cmsg_len = size;
804 cp->cmsg_level = level;
805 cp->cmsg_type = type;
806 return (m);
807 }
808