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 * $FreeBSD$
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
55 u_long sb_max = SB_MAX; /* XXX should be static */
56
57 static 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
89 void
90 soisconnecting(so)
91 register struct socket *so;
92 {
93
94 so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
95 so->so_state |= SS_ISCONNECTING;
96 }
97
98 void
99 soisconnected(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
121 void
122 soisdisconnecting(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
133 void
134 soisdisconnected(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 */
155 struct socket *
156 sodropablereq(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 */
197 struct socket *
198 sonewconn(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_cred = head->so_cred;
217 if (so->so_cred)
218 so->so_cred->p_refcnt++;
219 (void) soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat);
220
221 if ((*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) {
222 sodealloc(so);
223 return ((struct socket *)0);
224 }
225
226 if (connstatus) {
227 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
228 so->so_state |= SS_COMP;
229 } else {
230 TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
231 so->so_state |= SS_INCOMP;
232 head->so_incqlen++;
233 }
234 head->so_qlen++;
235 if (connstatus) {
236 sorwakeup(head);
237 wakeup((caddr_t)&head->so_timeo);
238 so->so_state |= connstatus;
239 }
240 return (so);
241 }
242
243 /*
244 * Socantsendmore indicates that no more data will be sent on the
245 * socket; it would normally be applied to a socket when the user
246 * informs the system that no more data is to be sent, by the protocol
247 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data
248 * will be received, and will normally be applied to the socket by a
249 * protocol when it detects that the peer will send no more data.
250 * Data queued for reading in the socket may yet be read.
251 */
252
253 void
254 socantsendmore(so)
255 struct socket *so;
256 {
257
258 so->so_state |= SS_CANTSENDMORE;
259 sowwakeup(so);
260 }
261
262 void
263 socantrcvmore(so)
264 struct socket *so;
265 {
266
267 so->so_state |= SS_CANTRCVMORE;
268 sorwakeup(so);
269 }
270
271 /*
272 * Wait for data to arrive at/drain from a socket buffer.
273 */
274 int
275 sbwait(sb)
276 struct sockbuf *sb;
277 {
278
279 sb->sb_flags |= SB_WAIT;
280 return (tsleep((caddr_t)&sb->sb_cc,
281 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
282 sb->sb_timeo));
283 }
284
285 /*
286 * Lock a sockbuf already known to be locked;
287 * return any error returned from sleep (EINTR).
288 */
289 int
290 sb_lock(sb)
291 register struct sockbuf *sb;
292 {
293 int error;
294
295 while (sb->sb_flags & SB_LOCK) {
296 sb->sb_flags |= SB_WANT;
297 error = tsleep((caddr_t)&sb->sb_flags,
298 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH,
299 "sblock", 0);
300 if (error)
301 return (error);
302 }
303 sb->sb_flags |= SB_LOCK;
304 return (0);
305 }
306
307 /*
308 * Wakeup processes waiting on a socket buffer.
309 * Do asynchronous notification via SIGIO
310 * if the socket has the SS_ASYNC flag set.
311 */
312 void
313 sowakeup(so, sb)
314 register struct socket *so;
315 register struct sockbuf *sb;
316 {
317 selwakeup(&sb->sb_sel);
318 sb->sb_flags &= ~SB_SEL;
319 if (sb->sb_flags & SB_WAIT) {
320 sb->sb_flags &= ~SB_WAIT;
321 wakeup((caddr_t)&sb->sb_cc);
322 }
323 if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL)
324 pgsigio(so->so_sigio, SIGIO, 0);
325 if (sb->sb_flags & SB_UPCALL)
326 (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT);
327 }
328
329 /*
330 * Socket buffer (struct sockbuf) utility routines.
331 *
332 * Each socket contains two socket buffers: one for sending data and
333 * one for receiving data. Each buffer contains a queue of mbufs,
334 * information about the number of mbufs and amount of data in the
335 * queue, and other fields allowing select() statements and notification
336 * on data availability to be implemented.
337 *
338 * Data stored in a socket buffer is maintained as a list of records.
339 * Each record is a list of mbufs chained together with the m_next
340 * field. Records are chained together with the m_nextpkt field. The upper
341 * level routine soreceive() expects the following conventions to be
342 * observed when placing information in the receive buffer:
343 *
344 * 1. If the protocol requires each message be preceded by the sender's
345 * name, then a record containing that name must be present before
346 * any associated data (mbuf's must be of type MT_SONAME).
347 * 2. If the protocol supports the exchange of ``access rights'' (really
348 * just additional data associated with the message), and there are
349 * ``rights'' to be received, then a record containing this data
350 * should be present (mbuf's must be of type MT_RIGHTS).
351 * 3. If a name or rights record exists, then it must be followed by
352 * a data record, perhaps of zero length.
353 *
354 * Before using a new socket structure it is first necessary to reserve
355 * buffer space to the socket, by calling sbreserve(). This should commit
356 * some of the available buffer space in the system buffer pool for the
357 * socket (currently, it does nothing but enforce limits). The space
358 * should be released by calling sbrelease() when the socket is destroyed.
359 */
360
361 int
362 soreserve(so, sndcc, rcvcc)
363 register struct socket *so;
364 u_long sndcc, rcvcc;
365 {
366
367 if (sbreserve(&so->so_snd, sndcc) == 0)
368 goto bad;
369 if (sbreserve(&so->so_rcv, rcvcc) == 0)
370 goto bad2;
371 if (so->so_rcv.sb_lowat == 0)
372 so->so_rcv.sb_lowat = 1;
373 if (so->so_snd.sb_lowat == 0)
374 so->so_snd.sb_lowat = MCLBYTES;
375 if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
376 so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
377 return (0);
378 bad2:
379 sbrelease(&so->so_snd);
380 bad:
381 return (ENOBUFS);
382 }
383
384 /*
385 * Allot mbufs to a sockbuf.
386 * Attempt to scale mbmax so that mbcnt doesn't become limiting
387 * if buffering efficiency is near the normal case.
388 */
389 int
390 sbreserve(sb, cc)
391 struct sockbuf *sb;
392 u_long cc;
393 {
394 if ((u_quad_t)cc > (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES))
395 return (0);
396 sb->sb_hiwat = cc;
397 sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
398 if (sb->sb_lowat > sb->sb_hiwat)
399 sb->sb_lowat = sb->sb_hiwat;
400 return (1);
401 }
402
403 /*
404 * Free mbufs held by a socket, and reserved mbuf space.
405 */
406 void
407 sbrelease(sb)
408 struct sockbuf *sb;
409 {
410
411 sbflush(sb);
412 sb->sb_hiwat = sb->sb_mbmax = 0;
413 }
414
415 /*
416 * Routines to add and remove
417 * data from an mbuf queue.
418 *
419 * The routines sbappend() or sbappendrecord() are normally called to
420 * append new mbufs to a socket buffer, after checking that adequate
421 * space is available, comparing the function sbspace() with the amount
422 * of data to be added. sbappendrecord() differs from sbappend() in
423 * that data supplied is treated as the beginning of a new record.
424 * To place a sender's address, optional access rights, and data in a
425 * socket receive buffer, sbappendaddr() should be used. To place
426 * access rights and data in a socket receive buffer, sbappendrights()
427 * should be used. In either case, the new data begins a new record.
428 * Note that unlike sbappend() and sbappendrecord(), these routines check
429 * for the caller that there will be enough space to store the data.
430 * Each fails if there is not enough space, or if it cannot find mbufs
431 * to store additional information in.
432 *
433 * Reliable protocols may use the socket send buffer to hold data
434 * awaiting acknowledgement. Data is normally copied from a socket
435 * send buffer in a protocol with m_copy for output to a peer,
436 * and then removing the data from the socket buffer with sbdrop()
437 * or sbdroprecord() when the data is acknowledged by the peer.
438 */
439
440 /*
441 * Append mbuf chain m to the last record in the
442 * socket buffer sb. The additional space associated
443 * the mbuf chain is recorded in sb. Empty mbufs are
444 * discarded and mbufs are compacted where possible.
445 */
446 void
447 sbappend(sb, m)
448 struct sockbuf *sb;
449 struct mbuf *m;
450 {
451 register struct mbuf *n;
452
453 if (m == 0)
454 return;
455 n = sb->sb_mb;
456 if (n) {
457 while (n->m_nextpkt)
458 n = n->m_nextpkt;
459 do {
460 if (n->m_flags & M_EOR) {
461 sbappendrecord(sb, m); /* XXXXXX!!!! */
462 return;
463 }
464 } while (n->m_next && (n = n->m_next));
465 }
466 sbcompress(sb, m, n);
467 }
468
469 #ifdef SOCKBUF_DEBUG
470 void
471 sbcheck(sb)
472 register struct sockbuf *sb;
473 {
474 register struct mbuf *m;
475 register struct mbuf *n = 0;
476 register u_long len = 0, mbcnt = 0;
477
478 for (m = sb->sb_mb; m; m = n) {
479 n = m->m_nextpkt;
480 for (; m; m = m->m_next) {
481 len += m->m_len;
482 mbcnt += MSIZE;
483 if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */
484 mbcnt += m->m_ext.ext_size;
485 }
486 }
487 if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
488 printf("cc %ld != %ld || mbcnt %ld != %ld\n", len, sb->sb_cc,
489 mbcnt, sb->sb_mbcnt);
490 panic("sbcheck");
491 }
492 }
493 #endif
494
495 /*
496 * As above, except the mbuf chain
497 * begins a new record.
498 */
499 void
500 sbappendrecord(sb, m0)
501 register struct sockbuf *sb;
502 register struct mbuf *m0;
503 {
504 register struct mbuf *m;
505
506 if (m0 == 0)
507 return;
508 m = sb->sb_mb;
509 if (m)
510 while (m->m_nextpkt)
511 m = m->m_nextpkt;
512 /*
513 * Put the first mbuf on the queue.
514 * Note this permits zero length records.
515 */
516 sballoc(sb, m0);
517 if (m)
518 m->m_nextpkt = m0;
519 else
520 sb->sb_mb = m0;
521 m = m0->m_next;
522 m0->m_next = 0;
523 if (m && (m0->m_flags & M_EOR)) {
524 m0->m_flags &= ~M_EOR;
525 m->m_flags |= M_EOR;
526 }
527 sbcompress(sb, m, m0);
528 }
529
530 /*
531 * As above except that OOB data
532 * is inserted at the beginning of the sockbuf,
533 * but after any other OOB data.
534 */
535 void
536 sbinsertoob(sb, m0)
537 register struct sockbuf *sb;
538 register struct mbuf *m0;
539 {
540 register struct mbuf *m;
541 register struct mbuf **mp;
542
543 if (m0 == 0)
544 return;
545 for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) {
546 m = *mp;
547 again:
548 switch (m->m_type) {
549
550 case MT_OOBDATA:
551 continue; /* WANT next train */
552
553 case MT_CONTROL:
554 m = m->m_next;
555 if (m)
556 goto again; /* inspect THIS train further */
557 }
558 break;
559 }
560 /*
561 * Put the first mbuf on the queue.
562 * Note this permits zero length records.
563 */
564 sballoc(sb, m0);
565 m0->m_nextpkt = *mp;
566 *mp = m0;
567 m = m0->m_next;
568 m0->m_next = 0;
569 if (m && (m0->m_flags & M_EOR)) {
570 m0->m_flags &= ~M_EOR;
571 m->m_flags |= M_EOR;
572 }
573 sbcompress(sb, m, m0);
574 }
575
576 /*
577 * Append address and data, and optionally, control (ancillary) data
578 * to the receive queue of a socket. If present,
579 * m0 must include a packet header with total length.
580 * Returns 0 if no space in sockbuf or insufficient mbufs.
581 */
582 int
583 sbappendaddr(sb, asa, m0, control)
584 register struct sockbuf *sb;
585 struct sockaddr *asa;
586 struct mbuf *m0, *control;
587 {
588 register struct mbuf *m, *n;
589 int space = asa->sa_len;
590
591 if (m0 && (m0->m_flags & M_PKTHDR) == 0)
592 panic("sbappendaddr");
593 if (m0)
594 space += m0->m_pkthdr.len;
595 for (n = control; n; n = n->m_next) {
596 space += n->m_len;
597 if (n->m_next == 0) /* keep pointer to last control buf */
598 break;
599 }
600 if (space > sbspace(sb))
601 return (0);
602 if (asa->sa_len > MLEN)
603 return (0);
604 MGET(m, M_DONTWAIT, MT_SONAME);
605 if (m == 0)
606 return (0);
607 m->m_len = asa->sa_len;
608 bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
609 if (n)
610 n->m_next = m0; /* concatenate data to control */
611 else
612 control = m0;
613 m->m_next = control;
614 for (n = m; n; n = n->m_next)
615 sballoc(sb, n);
616 n = sb->sb_mb;
617 if (n) {
618 while (n->m_nextpkt)
619 n = n->m_nextpkt;
620 n->m_nextpkt = m;
621 } else
622 sb->sb_mb = m;
623 return (1);
624 }
625
626 int
627 sbappendcontrol(sb, m0, control)
628 struct sockbuf *sb;
629 struct mbuf *control, *m0;
630 {
631 register struct mbuf *m, *n;
632 int space = 0;
633
634 if (control == 0)
635 panic("sbappendcontrol");
636 for (m = control; ; m = m->m_next) {
637 space += m->m_len;
638 if (m->m_next == 0)
639 break;
640 }
641 n = m; /* save pointer to last control buffer */
642 for (m = m0; m; m = m->m_next)
643 space += m->m_len;
644 if (space > sbspace(sb))
645 return (0);
646 n->m_next = m0; /* concatenate data to control */
647 for (m = control; m; m = m->m_next)
648 sballoc(sb, m);
649 n = sb->sb_mb;
650 if (n) {
651 while (n->m_nextpkt)
652 n = n->m_nextpkt;
653 n->m_nextpkt = control;
654 } else
655 sb->sb_mb = control;
656 return (1);
657 }
658
659 /*
660 * Compress mbuf chain m into the socket
661 * buffer sb following mbuf n. If n
662 * is null, the buffer is presumed empty.
663 */
664 void
665 sbcompress(sb, m, n)
666 register struct sockbuf *sb;
667 register struct mbuf *m, *n;
668 {
669 register int eor = 0;
670 register struct mbuf *o;
671
672 while (m) {
673 eor |= m->m_flags & M_EOR;
674 if (m->m_len == 0 &&
675 (eor == 0 ||
676 (((o = m->m_next) || (o = n)) &&
677 o->m_type == m->m_type))) {
678 m = m_free(m);
679 continue;
680 }
681 if (n && (n->m_flags & (M_EXT | M_EOR)) == 0 &&
682 (n->m_data + n->m_len + m->m_len) < &n->m_dat[MLEN] &&
683 n->m_type == m->m_type) {
684 bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
685 (unsigned)m->m_len);
686 n->m_len += m->m_len;
687 sb->sb_cc += m->m_len;
688 m = m_free(m);
689 continue;
690 }
691 if (n)
692 n->m_next = m;
693 else
694 sb->sb_mb = m;
695 sballoc(sb, m);
696 n = m;
697 m->m_flags &= ~M_EOR;
698 m = m->m_next;
699 n->m_next = 0;
700 }
701 if (eor) {
702 if (n)
703 n->m_flags |= eor;
704 else
705 printf("semi-panic: sbcompress\n");
706 }
707 }
708
709 /*
710 * Free all mbufs in a sockbuf.
711 * Check that all resources are reclaimed.
712 */
713 void
714 sbflush(sb)
715 register struct sockbuf *sb;
716 {
717
718 if (sb->sb_flags & SB_LOCK)
719 panic("sbflush: locked");
720 while (sb->sb_mbcnt) {
721 /*
722 * Don't call sbdrop(sb, 0) if the leading mbuf is non-empty:
723 * we would loop forever. Panic instead.
724 */
725 if (!sb->sb_cc && (sb->sb_mb == NULL || sb->sb_mb->m_len))
726 break;
727 sbdrop(sb, (int)sb->sb_cc);
728 }
729 if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt)
730 panic("sbflush: cc %ld || mb %p || mbcnt %ld", sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt);
731 }
732
733 /*
734 * Drop data from (the front of) a sockbuf.
735 */
736 void
737 sbdrop(sb, len)
738 register struct sockbuf *sb;
739 register int len;
740 {
741 register struct mbuf *m, *mn;
742 struct mbuf *next;
743
744 next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
745 while (len > 0) {
746 if (m == 0) {
747 if (next == 0)
748 panic("sbdrop");
749 m = next;
750 next = m->m_nextpkt;
751 continue;
752 }
753 if (m->m_len > len) {
754 m->m_len -= len;
755 m->m_data += len;
756 sb->sb_cc -= len;
757 break;
758 }
759 len -= m->m_len;
760 sbfree(sb, m);
761 MFREE(m, mn);
762 m = mn;
763 }
764 while (m && m->m_len == 0) {
765 sbfree(sb, m);
766 MFREE(m, mn);
767 m = mn;
768 }
769 if (m) {
770 sb->sb_mb = m;
771 m->m_nextpkt = next;
772 } else
773 sb->sb_mb = next;
774 }
775
776 /*
777 * Drop a record off the front of a sockbuf
778 * and move the next record to the front.
779 */
780 void
781 sbdroprecord(sb)
782 register struct sockbuf *sb;
783 {
784 register struct mbuf *m, *mn;
785
786 m = sb->sb_mb;
787 if (m) {
788 sb->sb_mb = m->m_nextpkt;
789 do {
790 sbfree(sb, m);
791 MFREE(m, mn);
792 m = mn;
793 } while (m);
794 }
795 }
796
797 /*
798 * Create a "control" mbuf containing the specified data
799 * with the specified type for presentation on a socket buffer.
800 */
801 struct mbuf *
802 sbcreatecontrol(p, size, type, level)
803 caddr_t p;
804 register int size;
805 int type, level;
806 {
807 register struct cmsghdr *cp;
808 struct mbuf *m;
809
810 if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL)
811 return ((struct mbuf *) NULL);
812 cp = mtod(m, struct cmsghdr *);
813 /* XXX check size? */
814 (void)memcpy(CMSG_DATA(cp), p, size);
815 size += sizeof(*cp);
816 m->m_len = size;
817 cp->cmsg_len = size;
818 cp->cmsg_level = level;
819 cp->cmsg_type = type;
820 return (m);
821 }
822
823 /*
824 * Some routines that return EOPNOTSUPP for entry points that are not
825 * supported by a protocol. Fill in as needed.
826 */
827 int
828 pru_accept_notsupp(struct socket *so, struct sockaddr **nam)
829 {
830 return EOPNOTSUPP;
831 }
832
833 int
834 pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct proc *p)
835 {
836 return EOPNOTSUPP;
837 }
838
839 int
840 pru_connect2_notsupp(struct socket *so1, struct socket *so2)
841 {
842 return EOPNOTSUPP;
843 }
844
845 int
846 pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data,
847 struct ifnet *ifp, struct proc *p)
848 {
849 return EOPNOTSUPP;
850 }
851
852 int
853 pru_listen_notsupp(struct socket *so, struct proc *p)
854 {
855 return EOPNOTSUPP;
856 }
857
858 int
859 pru_rcvd_notsupp(struct socket *so, int flags)
860 {
861 return EOPNOTSUPP;
862 }
863
864 int
865 pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags)
866 {
867 return EOPNOTSUPP;
868 }
869
870 /*
871 * This isn't really a ``null'' operation, but it's the default one
872 * and doesn't do anything destructive.
873 */
874 int
875 pru_sense_null(struct socket *so, struct stat *sb)
876 {
877 sb->st_blksize = so->so_snd.sb_hiwat;
878 return 0;
879 }
880
881 /*
882 * Make a copy of a sockaddr in a malloced buffer of type M_SONAME.
883 */
884 struct sockaddr *
885 dup_sockaddr(sa, canwait)
886 struct sockaddr *sa;
887 int canwait;
888 {
889 struct sockaddr *sa2;
890
891 MALLOC(sa2, struct sockaddr *, sa->sa_len, M_SONAME,
892 canwait ? M_WAITOK : M_NOWAIT);
893 if (sa2)
894 bcopy(sa, sa2, sa->sa_len);
895 return sa2;
896 }
897
898 /*
899 * Create an external-format (``xsocket'') structure using the information
900 * in the kernel-format socket structure pointed to by so. This is done
901 * to reduce the spew of irrelevant information over this interface,
902 * to isolate user code from changes in the kernel structure, and
903 * potentially to provide information-hiding if we decide that
904 * some of this information should be hidden from users.
905 */
906 void
907 sotoxsocket(struct socket *so, struct xsocket *xso)
908 {
909 xso->xso_len = sizeof *xso;
910 xso->xso_so = so;
911 xso->so_type = so->so_type;
912 xso->so_options = so->so_options;
913 xso->so_linger = so->so_linger;
914 xso->so_state = so->so_state;
915 xso->so_pcb = so->so_pcb;
916 xso->xso_protocol = so->so_proto->pr_protocol;
917 xso->xso_family = so->so_proto->pr_domain->dom_family;
918 xso->so_qlen = so->so_qlen;
919 xso->so_incqlen = so->so_incqlen;
920 xso->so_qlimit = so->so_qlimit;
921 xso->so_timeo = so->so_timeo;
922 xso->so_error = so->so_error;
923 xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0;
924 xso->so_oobmark = so->so_oobmark;
925 sbtoxsockbuf(&so->so_snd, &xso->so_snd);
926 sbtoxsockbuf(&so->so_rcv, &xso->so_rcv);
927 xso->so_uid = so->so_cred ? so->so_cred->pc_ucred->cr_uid : -1;
928 }
929
930 /*
931 * This does the same for sockbufs. Note that the xsockbuf structure,
932 * since it is always embedded in a socket, does not include a self
933 * pointer nor a length. We make this entry point public in case
934 * some other mechanism needs it.
935 */
936 void
937 sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb)
938 {
939 xsb->sb_cc = sb->sb_cc;
940 xsb->sb_hiwat = sb->sb_hiwat;
941 xsb->sb_mbcnt = sb->sb_mbcnt;
942 xsb->sb_mbmax = sb->sb_mbmax;
943 xsb->sb_lowat = sb->sb_lowat;
944 xsb->sb_flags = sb->sb_flags;
945 xsb->sb_timeo = sb->sb_timeo;
946 }
947
948 /*
949 * Here is the definition of some of the basic objects in the kern.ipc
950 * branch of the MIB.
951 */
952 SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC");
953
954 /* This takes the place of kern.maxsockbuf, which moved to kern.ipc. */
955 static int dummy;
956 SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, "");
957
958 SYSCTL_INT(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLFLAG_RW, &sb_max, 0, "");
959 SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD, &maxsockets, 0, "");
960 SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW,
961 &sb_efficiency, 0, "");
962 SYSCTL_INT(_kern_ipc, KIPC_NMBCLUSTERS, nmbclusters, CTLFLAG_RD, &nmbclusters, 0, "");
963
Cache object: bb8d42d97b9c09f1933ec0db4a511c76
|