Send connects to the right processor.
[dragonfly.git] / sys / kern / uipc_socket2.c
CommitLineData
984263bc
MD
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: src/sys/kern/uipc_socket2.c,v 1.55.2.17 2002/08/31 19:04:55 dwmalone Exp $
9eeaa8a9 35 * $DragonFly: src/sys/kern/uipc_socket2.c,v 1.8 2004/03/06 01:58:54 hsu Exp $
984263bc
MD
36 */
37
38#include "opt_param.h"
39#include <sys/param.h>
40#include <sys/systm.h>
41#include <sys/domain.h>
42#include <sys/file.h> /* for maxfiles */
43#include <sys/kernel.h>
44#include <sys/proc.h>
45#include <sys/malloc.h>
46#include <sys/mbuf.h>
47#include <sys/protosw.h>
48#include <sys/resourcevar.h>
49#include <sys/stat.h>
50#include <sys/socket.h>
51#include <sys/socketvar.h>
52#include <sys/signalvar.h>
53#include <sys/sysctl.h>
54#include <sys/aio.h> /* for aio_swake proto */
55#include <sys/event.h>
56
57int maxsockets;
58
59/*
60 * Primitive routines for operating on sockets and socket buffers
61 */
62
63u_long sb_max = SB_MAX;
64u_long sb_max_adj =
65 SB_MAX * MCLBYTES / (MSIZE + MCLBYTES); /* adjusted sb_max */
66
67static u_long sb_efficiency = 8; /* parameter for sbreserve() */
68
69/*
70 * Procedures to manipulate state flags of socket
71 * and do appropriate wakeups. Normal sequence from the
72 * active (originating) side is that soisconnecting() is
73 * called during processing of connect() call,
74 * resulting in an eventual call to soisconnected() if/when the
75 * connection is established. When the connection is torn down
76 * soisdisconnecting() is called during processing of disconnect() call,
77 * and soisdisconnected() is called when the connection to the peer
78 * is totally severed. The semantics of these routines are such that
79 * connectionless protocols can call soisconnected() and soisdisconnected()
80 * only, bypassing the in-progress calls when setting up a ``connection''
81 * takes no time.
82 *
83 * From the passive side, a socket is created with
84 * two queues of sockets: so_incomp for connections in progress
85 * and so_comp for connections already made and awaiting user acceptance.
86 * As a protocol is preparing incoming connections, it creates a socket
87 * structure queued on so_incomp by calling sonewconn(). When the connection
88 * is established, soisconnected() is called, and transfers the
89 * socket structure to so_comp, making it available to accept().
90 *
91 * If a socket is closed with sockets on either
92 * so_incomp or so_comp, these sockets are dropped.
93 *
94 * If higher level protocols are implemented in
95 * the kernel, the wakeups done here will sometimes
96 * cause software-interrupt process scheduling.
97 */
98
99void
100soisconnecting(so)
1fd87d54 101 struct socket *so;
984263bc
MD
102{
103
104 so->so_state &= ~(SS_ISCONNECTED|SS_ISDISCONNECTING);
105 so->so_state |= SS_ISCONNECTING;
106}
107
108void
109soisconnected(so)
110 struct socket *so;
111{
112 struct socket *head = so->so_head;
113
114 so->so_state &= ~(SS_ISCONNECTING|SS_ISDISCONNECTING|SS_ISCONFIRMING);
115 so->so_state |= SS_ISCONNECTED;
116 if (head && (so->so_state & SS_INCOMP)) {
117 if ((so->so_options & SO_ACCEPTFILTER) != 0) {
118 so->so_upcall = head->so_accf->so_accept_filter->accf_callback;
119 so->so_upcallarg = head->so_accf->so_accept_filter_arg;
120 so->so_rcv.sb_flags |= SB_UPCALL;
121 so->so_options &= ~SO_ACCEPTFILTER;
122 so->so_upcall(so, so->so_upcallarg, 0);
123 return;
124 }
125 TAILQ_REMOVE(&head->so_incomp, so, so_list);
126 head->so_incqlen--;
127 so->so_state &= ~SS_INCOMP;
128 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
129 head->so_qlen++;
130 so->so_state |= SS_COMP;
131 sorwakeup(head);
132 wakeup_one(&head->so_timeo);
133 } else {
134 wakeup(&so->so_timeo);
135 sorwakeup(so);
136 sowwakeup(so);
137 }
138}
139
140void
141soisdisconnecting(so)
1fd87d54 142 struct socket *so;
984263bc
MD
143{
144
145 so->so_state &= ~SS_ISCONNECTING;
146 so->so_state |= (SS_ISDISCONNECTING|SS_CANTRCVMORE|SS_CANTSENDMORE);
147 wakeup((caddr_t)&so->so_timeo);
148 sowwakeup(so);
149 sorwakeup(so);
150}
151
152void
153soisdisconnected(so)
1fd87d54 154 struct socket *so;
984263bc
MD
155{
156
157 so->so_state &= ~(SS_ISCONNECTING|SS_ISCONNECTED|SS_ISDISCONNECTING);
158 so->so_state |= (SS_CANTRCVMORE|SS_CANTSENDMORE|SS_ISDISCONNECTED);
159 wakeup((caddr_t)&so->so_timeo);
160 sbdrop(&so->so_snd, so->so_snd.sb_cc);
161 sowwakeup(so);
162 sorwakeup(so);
163}
164
165/*
166 * When an attempt at a new connection is noted on a socket
167 * which accepts connections, sonewconn is called. If the
168 * connection is possible (subject to space constraints, etc.)
169 * then we allocate a new structure, propoerly linked into the
170 * data structure of the original socket, and return this.
171 * Connstatus may be 0, or SO_ISCONFIRMING, or SO_ISCONNECTED.
172 */
173struct socket *
dadab5e9 174sonewconn(struct socket *head, int connstatus)
984263bc 175{
1fd87d54 176 struct socket *so;
e4700d00 177 struct pru_attach_info ai;
984263bc
MD
178
179 if (head->so_qlen > 3 * head->so_qlimit / 2)
180 return ((struct socket *)0);
181 so = soalloc(0);
182 if (so == NULL)
183 return ((struct socket *)0);
184 if ((head->so_options & SO_ACCEPTFILTER) != 0)
185 connstatus = 0;
186 so->so_head = head;
187 so->so_type = head->so_type;
188 so->so_options = head->so_options &~ SO_ACCEPTCONN;
189 so->so_linger = head->so_linger;
190 so->so_state = head->so_state | SS_NOFDREF;
191 so->so_proto = head->so_proto;
192 so->so_timeo = head->so_timeo;
dadab5e9 193 so->so_cred = crhold(head->so_cred);
e4700d00
JH
194 ai.sb_rlimit = NULL;
195 ai.p_ucred = NULL;
196 ai.fd_rdir = NULL; /* jail code cruft XXX JH */
197 if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat, NULL) ||
9eeaa8a9 198 /* Directly call function since we're already at protocol level. */
e4700d00 199 (*so->so_proto->pr_usrreqs->pru_attach)(so, 0, &ai)) {
984263bc
MD
200 sodealloc(so);
201 return ((struct socket *)0);
202 }
203
204 if (connstatus) {
205 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
206 so->so_state |= SS_COMP;
207 head->so_qlen++;
208 } else {
209 if (head->so_incqlen > head->so_qlimit) {
210 struct socket *sp;
211 sp = TAILQ_FIRST(&head->so_incomp);
212 (void) soabort(sp);
213 }
214 TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
215 so->so_state |= SS_INCOMP;
216 head->so_incqlen++;
217 }
218 if (connstatus) {
219 sorwakeup(head);
220 wakeup((caddr_t)&head->so_timeo);
221 so->so_state |= connstatus;
222 }
223 return (so);
224}
225
226/*
227 * Socantsendmore indicates that no more data will be sent on the
228 * socket; it would normally be applied to a socket when the user
229 * informs the system that no more data is to be sent, by the protocol
230 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data
231 * will be received, and will normally be applied to the socket by a
232 * protocol when it detects that the peer will send no more data.
233 * Data queued for reading in the socket may yet be read.
234 */
235
236void
237socantsendmore(so)
238 struct socket *so;
239{
240
241 so->so_state |= SS_CANTSENDMORE;
242 sowwakeup(so);
243}
244
245void
246socantrcvmore(so)
247 struct socket *so;
248{
249
250 so->so_state |= SS_CANTRCVMORE;
251 sorwakeup(so);
252}
253
254/*
255 * Wait for data to arrive at/drain from a socket buffer.
256 */
257int
258sbwait(sb)
259 struct sockbuf *sb;
260{
261
262 sb->sb_flags |= SB_WAIT;
263 return (tsleep((caddr_t)&sb->sb_cc,
377d4740
MD
264 ((sb->sb_flags & SB_NOINTR) ? 0 : PCATCH),
265 "sbwait",
266 sb->sb_timeo));
984263bc
MD
267}
268
269/*
270 * Lock a sockbuf already known to be locked;
271 * return any error returned from sleep (EINTR).
272 */
273int
274sb_lock(sb)
1fd87d54 275 struct sockbuf *sb;
984263bc
MD
276{
277 int error;
278
279 while (sb->sb_flags & SB_LOCK) {
280 sb->sb_flags |= SB_WANT;
281 error = tsleep((caddr_t)&sb->sb_flags,
377d4740
MD
282 ((sb->sb_flags & SB_NOINTR) ? 0 : PCATCH),
283 "sblock", 0);
984263bc
MD
284 if (error)
285 return (error);
286 }
287 sb->sb_flags |= SB_LOCK;
288 return (0);
289}
290
291/*
292 * Wakeup processes waiting on a socket buffer.
293 * Do asynchronous notification via SIGIO
294 * if the socket has the SS_ASYNC flag set.
295 */
296void
297sowakeup(so, sb)
1fd87d54
RG
298 struct socket *so;
299 struct sockbuf *sb;
984263bc
MD
300{
301 selwakeup(&sb->sb_sel);
302 sb->sb_flags &= ~SB_SEL;
303 if (sb->sb_flags & SB_WAIT) {
304 sb->sb_flags &= ~SB_WAIT;
305 wakeup((caddr_t)&sb->sb_cc);
306 }
307 if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL)
308 pgsigio(so->so_sigio, SIGIO, 0);
309 if (sb->sb_flags & SB_UPCALL)
310 (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT);
311 if (sb->sb_flags & SB_AIO)
312 aio_swake(so, sb);
313 KNOTE(&sb->sb_sel.si_note, 0);
314}
315
316/*
317 * Socket buffer (struct sockbuf) utility routines.
318 *
319 * Each socket contains two socket buffers: one for sending data and
320 * one for receiving data. Each buffer contains a queue of mbufs,
321 * information about the number of mbufs and amount of data in the
322 * queue, and other fields allowing select() statements and notification
323 * on data availability to be implemented.
324 *
325 * Data stored in a socket buffer is maintained as a list of records.
326 * Each record is a list of mbufs chained together with the m_next
327 * field. Records are chained together with the m_nextpkt field. The upper
328 * level routine soreceive() expects the following conventions to be
329 * observed when placing information in the receive buffer:
330 *
331 * 1. If the protocol requires each message be preceded by the sender's
332 * name, then a record containing that name must be present before
333 * any associated data (mbuf's must be of type MT_SONAME).
334 * 2. If the protocol supports the exchange of ``access rights'' (really
335 * just additional data associated with the message), and there are
336 * ``rights'' to be received, then a record containing this data
337 * should be present (mbuf's must be of type MT_RIGHTS).
338 * 3. If a name or rights record exists, then it must be followed by
339 * a data record, perhaps of zero length.
340 *
341 * Before using a new socket structure it is first necessary to reserve
342 * buffer space to the socket, by calling sbreserve(). This should commit
343 * some of the available buffer space in the system buffer pool for the
344 * socket (currently, it does nothing but enforce limits). The space
345 * should be released by calling sbrelease() when the socket is destroyed.
346 */
347
348int
e4700d00 349soreserve(struct socket *so, u_long sndcc, u_long rcvcc, struct rlimit *rl)
984263bc 350{
e4700d00 351 if (sbreserve(&so->so_snd, sndcc, so, rl) == 0)
984263bc 352 goto bad;
e4700d00 353 if (sbreserve(&so->so_rcv, rcvcc, so, rl) == 0)
984263bc
MD
354 goto bad2;
355 if (so->so_rcv.sb_lowat == 0)
356 so->so_rcv.sb_lowat = 1;
357 if (so->so_snd.sb_lowat == 0)
358 so->so_snd.sb_lowat = MCLBYTES;
359 if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
360 so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
361 return (0);
362bad2:
363 sbrelease(&so->so_snd, so);
364bad:
365 return (ENOBUFS);
366}
367
368static int
369sysctl_handle_sb_max(SYSCTL_HANDLER_ARGS)
370{
371 int error = 0;
372 u_long old_sb_max = sb_max;
373
374 error = SYSCTL_OUT(req, arg1, sizeof(int));
375 if (error || !req->newptr)
376 return (error);
377 error = SYSCTL_IN(req, arg1, sizeof(int));
378 if (error)
379 return (error);
380 if (sb_max < MSIZE + MCLBYTES) {
381 sb_max = old_sb_max;
382 return (EINVAL);
383 }
384 sb_max_adj = (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES);
385 return (0);
386}
387
388/*
389 * Allot mbufs to a sockbuf.
390 * Attempt to scale mbmax so that mbcnt doesn't become limiting
391 * if buffering efficiency is near the normal case.
392 */
393int
e4700d00 394sbreserve(struct sockbuf *sb, u_long cc, struct socket *so, struct rlimit *rl)
984263bc
MD
395{
396
397 /*
e4700d00
JH
398 * rl will only be NULL when we're in an interrupt (eg, in tcp_input)
399 * or when called from netgraph (ie, ngd_attach)
984263bc
MD
400 */
401 if (cc > sb_max_adj)
402 return (0);
403 if (!chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, cc,
e4700d00 404 rl ? rl->rlim_cur : RLIM_INFINITY)) {
984263bc
MD
405 return (0);
406 }
407 sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
408 if (sb->sb_lowat > sb->sb_hiwat)
409 sb->sb_lowat = sb->sb_hiwat;
410 return (1);
411}
412
413/*
414 * Free mbufs held by a socket, and reserved mbuf space.
415 */
416void
417sbrelease(sb, so)
418 struct sockbuf *sb;
419 struct socket *so;
420{
421
422 sbflush(sb);
423 (void)chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, 0,
424 RLIM_INFINITY);
425 sb->sb_mbmax = 0;
426}
427
428/*
429 * Routines to add and remove
430 * data from an mbuf queue.
431 *
432 * The routines sbappend() or sbappendrecord() are normally called to
433 * append new mbufs to a socket buffer, after checking that adequate
434 * space is available, comparing the function sbspace() with the amount
435 * of data to be added. sbappendrecord() differs from sbappend() in
436 * that data supplied is treated as the beginning of a new record.
437 * To place a sender's address, optional access rights, and data in a
438 * socket receive buffer, sbappendaddr() should be used. To place
439 * access rights and data in a socket receive buffer, sbappendrights()
440 * should be used. In either case, the new data begins a new record.
441 * Note that unlike sbappend() and sbappendrecord(), these routines check
442 * for the caller that there will be enough space to store the data.
443 * Each fails if there is not enough space, or if it cannot find mbufs
444 * to store additional information in.
445 *
446 * Reliable protocols may use the socket send buffer to hold data
447 * awaiting acknowledgement. Data is normally copied from a socket
448 * send buffer in a protocol with m_copy for output to a peer,
449 * and then removing the data from the socket buffer with sbdrop()
450 * or sbdroprecord() when the data is acknowledged by the peer.
451 */
452
453/*
454 * Append mbuf chain m to the last record in the
455 * socket buffer sb. The additional space associated
456 * the mbuf chain is recorded in sb. Empty mbufs are
457 * discarded and mbufs are compacted where possible.
458 */
459void
460sbappend(sb, m)
461 struct sockbuf *sb;
462 struct mbuf *m;
463{
1fd87d54 464 struct mbuf *n;
984263bc
MD
465
466 if (m == 0)
467 return;
468 n = sb->sb_mb;
469 if (n) {
470 while (n->m_nextpkt)
471 n = n->m_nextpkt;
472 do {
473 if (n->m_flags & M_EOR) {
474 sbappendrecord(sb, m); /* XXXXXX!!!! */
475 return;
476 }
477 } while (n->m_next && (n = n->m_next));
478 }
479 sbcompress(sb, m, n);
480}
481
482#ifdef SOCKBUF_DEBUG
483void
484sbcheck(sb)
1fd87d54 485 struct sockbuf *sb;
984263bc 486{
1fd87d54
RG
487 struct mbuf *m;
488 struct mbuf *n = 0;
489 u_long len = 0, mbcnt = 0;
984263bc
MD
490
491 for (m = sb->sb_mb; m; m = n) {
492 n = m->m_nextpkt;
493 for (; m; m = m->m_next) {
494 len += m->m_len;
495 mbcnt += MSIZE;
496 if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */
497 mbcnt += m->m_ext.ext_size;
498 }
499 }
500 if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
501 printf("cc %ld != %ld || mbcnt %ld != %ld\n", len, sb->sb_cc,
502 mbcnt, sb->sb_mbcnt);
503 panic("sbcheck");
504 }
505}
506#endif
507
508/*
509 * As above, except the mbuf chain
510 * begins a new record.
511 */
512void
513sbappendrecord(sb, m0)
1fd87d54
RG
514 struct sockbuf *sb;
515 struct mbuf *m0;
984263bc 516{
1fd87d54 517 struct mbuf *m;
984263bc
MD
518
519 if (m0 == 0)
520 return;
521 m = sb->sb_mb;
522 if (m)
523 while (m->m_nextpkt)
524 m = m->m_nextpkt;
525 /*
526 * Put the first mbuf on the queue.
527 * Note this permits zero length records.
528 */
529 sballoc(sb, m0);
530 if (m)
531 m->m_nextpkt = m0;
532 else
533 sb->sb_mb = m0;
534 m = m0->m_next;
535 m0->m_next = 0;
536 if (m && (m0->m_flags & M_EOR)) {
537 m0->m_flags &= ~M_EOR;
538 m->m_flags |= M_EOR;
539 }
540 sbcompress(sb, m, m0);
541}
542
543/*
544 * As above except that OOB data
545 * is inserted at the beginning of the sockbuf,
546 * but after any other OOB data.
547 */
548void
549sbinsertoob(sb, m0)
1fd87d54
RG
550 struct sockbuf *sb;
551 struct mbuf *m0;
984263bc 552{
1fd87d54
RG
553 struct mbuf *m;
554 struct mbuf **mp;
984263bc
MD
555
556 if (m0 == 0)
557 return;
558 for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) {
559 m = *mp;
560 again:
561 switch (m->m_type) {
562
563 case MT_OOBDATA:
564 continue; /* WANT next train */
565
566 case MT_CONTROL:
567 m = m->m_next;
568 if (m)
569 goto again; /* inspect THIS train further */
570 }
571 break;
572 }
573 /*
574 * Put the first mbuf on the queue.
575 * Note this permits zero length records.
576 */
577 sballoc(sb, m0);
578 m0->m_nextpkt = *mp;
579 *mp = m0;
580 m = m0->m_next;
581 m0->m_next = 0;
582 if (m && (m0->m_flags & M_EOR)) {
583 m0->m_flags &= ~M_EOR;
584 m->m_flags |= M_EOR;
585 }
586 sbcompress(sb, m, m0);
587}
588
589/*
590 * Append address and data, and optionally, control (ancillary) data
591 * to the receive queue of a socket. If present,
592 * m0 must include a packet header with total length.
593 * Returns 0 if no space in sockbuf or insufficient mbufs.
594 */
595int
596sbappendaddr(sb, asa, m0, control)
1fd87d54 597 struct sockbuf *sb;
984263bc
MD
598 struct sockaddr *asa;
599 struct mbuf *m0, *control;
600{
1fd87d54 601 struct mbuf *m, *n;
984263bc
MD
602 int space = asa->sa_len;
603
a8a35116
HP
604 if (m0 && (m0->m_flags & M_PKTHDR) == 0)
605 panic("sbappendaddr");
606
984263bc
MD
607 if (m0)
608 space += m0->m_pkthdr.len;
609 for (n = control; n; n = n->m_next) {
610 space += n->m_len;
611 if (n->m_next == 0) /* keep pointer to last control buf */
612 break;
613 }
614 if (space > sbspace(sb))
615 return (0);
616 if (asa->sa_len > MLEN)
617 return (0);
618 MGET(m, M_DONTWAIT, MT_SONAME);
619 if (m == 0)
620 return (0);
621 m->m_len = asa->sa_len;
622 bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
623 if (n)
624 n->m_next = m0; /* concatenate data to control */
625 else
626 control = m0;
627 m->m_next = control;
628 for (n = m; n; n = n->m_next)
629 sballoc(sb, n);
630 n = sb->sb_mb;
631 if (n) {
632 while (n->m_nextpkt)
633 n = n->m_nextpkt;
634 n->m_nextpkt = m;
635 } else
636 sb->sb_mb = m;
637 return (1);
638}
639
640int
641sbappendcontrol(sb, m0, control)
642 struct sockbuf *sb;
643 struct mbuf *control, *m0;
644{
1fd87d54 645 struct mbuf *m, *n;
984263bc
MD
646 int space = 0;
647
648 if (control == 0)
649 panic("sbappendcontrol");
650 for (m = control; ; m = m->m_next) {
651 space += m->m_len;
652 if (m->m_next == 0)
653 break;
654 }
655 n = m; /* save pointer to last control buffer */
656 for (m = m0; m; m = m->m_next)
657 space += m->m_len;
658 if (space > sbspace(sb))
659 return (0);
660 n->m_next = m0; /* concatenate data to control */
661 for (m = control; m; m = m->m_next)
662 sballoc(sb, m);
663 n = sb->sb_mb;
664 if (n) {
665 while (n->m_nextpkt)
666 n = n->m_nextpkt;
667 n->m_nextpkt = control;
668 } else
669 sb->sb_mb = control;
670 return (1);
671}
672
673/*
674 * Compress mbuf chain m into the socket
675 * buffer sb following mbuf n. If n
676 * is null, the buffer is presumed empty.
677 */
678void
679sbcompress(sb, m, n)
1fd87d54
RG
680 struct sockbuf *sb;
681 struct mbuf *m, *n;
984263bc 682{
1fd87d54
RG
683 int eor = 0;
684 struct mbuf *o;
984263bc
MD
685
686 while (m) {
687 eor |= m->m_flags & M_EOR;
688 if (m->m_len == 0 &&
689 (eor == 0 ||
690 (((o = m->m_next) || (o = n)) &&
691 o->m_type == m->m_type))) {
692 m = m_free(m);
693 continue;
694 }
695 if (n && (n->m_flags & M_EOR) == 0 &&
696 M_WRITABLE(n) &&
697 m->m_len <= MCLBYTES / 4 && /* XXX: Don't copy too much */
698 m->m_len <= M_TRAILINGSPACE(n) &&
699 n->m_type == m->m_type) {
700 bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
701 (unsigned)m->m_len);
702 n->m_len += m->m_len;
703 sb->sb_cc += m->m_len;
704 m = m_free(m);
705 continue;
706 }
707 if (n)
708 n->m_next = m;
709 else
710 sb->sb_mb = m;
711 sballoc(sb, m);
712 n = m;
713 m->m_flags &= ~M_EOR;
714 m = m->m_next;
715 n->m_next = 0;
716 }
717 if (eor) {
718 if (n)
719 n->m_flags |= eor;
720 else
721 printf("semi-panic: sbcompress\n");
722 }
723}
724
725/*
726 * Free all mbufs in a sockbuf.
727 * Check that all resources are reclaimed.
728 */
729void
730sbflush(sb)
1fd87d54 731 struct sockbuf *sb;
984263bc
MD
732{
733
734 if (sb->sb_flags & SB_LOCK)
735 panic("sbflush: locked");
736 while (sb->sb_mbcnt) {
737 /*
738 * Don't call sbdrop(sb, 0) if the leading mbuf is non-empty:
739 * we would loop forever. Panic instead.
740 */
741 if (!sb->sb_cc && (sb->sb_mb == NULL || sb->sb_mb->m_len))
742 break;
743 sbdrop(sb, (int)sb->sb_cc);
744 }
745 if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt)
746 panic("sbflush: cc %ld || mb %p || mbcnt %ld", sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt);
747}
748
749/*
750 * Drop data from (the front of) a sockbuf.
751 */
752void
753sbdrop(sb, len)
1fd87d54
RG
754 struct sockbuf *sb;
755 int len;
984263bc 756{
1fd87d54 757 struct mbuf *m;
984263bc
MD
758 struct mbuf *next;
759
760 next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
761 while (len > 0) {
762 if (m == 0) {
763 if (next == 0)
764 panic("sbdrop");
765 m = next;
766 next = m->m_nextpkt;
767 continue;
768 }
769 if (m->m_len > len) {
770 m->m_len -= len;
771 m->m_data += len;
772 sb->sb_cc -= len;
773 break;
774 }
775 len -= m->m_len;
776 sbfree(sb, m);
777 m = m_free(m);
778 }
779 while (m && m->m_len == 0) {
780 sbfree(sb, m);
781 m = m_free(m);
782 }
783 if (m) {
784 sb->sb_mb = m;
785 m->m_nextpkt = next;
786 } else
787 sb->sb_mb = next;
788}
789
790/*
791 * Drop a record off the front of a sockbuf
792 * and move the next record to the front.
793 */
794void
795sbdroprecord(sb)
1fd87d54 796 struct sockbuf *sb;
984263bc 797{
1fd87d54 798 struct mbuf *m;
984263bc
MD
799
800 m = sb->sb_mb;
801 if (m) {
802 sb->sb_mb = m->m_nextpkt;
803 do {
804 sbfree(sb, m);
805 m = m_free(m);
806 } while (m);
807 }
808}
809
810/*
811 * Create a "control" mbuf containing the specified data
812 * with the specified type for presentation on a socket buffer.
813 */
814struct mbuf *
815sbcreatecontrol(p, size, type, level)
816 caddr_t p;
1fd87d54 817 int size;
984263bc
MD
818 int type, level;
819{
1fd87d54 820 struct cmsghdr *cp;
984263bc
MD
821 struct mbuf *m;
822
823 if (CMSG_SPACE((u_int)size) > MCLBYTES)
824 return ((struct mbuf *) NULL);
825 if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL)
826 return ((struct mbuf *) NULL);
827 if (CMSG_SPACE((u_int)size) > MLEN) {
828 MCLGET(m, M_DONTWAIT);
829 if ((m->m_flags & M_EXT) == 0) {
830 m_free(m);
831 return ((struct mbuf *) NULL);
832 }
833 }
834 cp = mtod(m, struct cmsghdr *);
835 m->m_len = 0;
836 KASSERT(CMSG_SPACE((u_int)size) <= M_TRAILINGSPACE(m),
837 ("sbcreatecontrol: short mbuf"));
838 if (p != NULL)
839 (void)memcpy(CMSG_DATA(cp), p, size);
840 m->m_len = CMSG_SPACE(size);
841 cp->cmsg_len = CMSG_LEN(size);
842 cp->cmsg_level = level;
843 cp->cmsg_type = type;
844 return (m);
845}
846
847/*
848 * Some routines that return EOPNOTSUPP for entry points that are not
849 * supported by a protocol. Fill in as needed.
850 */
851int
852pru_accept_notsupp(struct socket *so, struct sockaddr **nam)
853{
854 return EOPNOTSUPP;
855}
856
857int
dadab5e9 858pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct thread *td)
984263bc
MD
859{
860 return EOPNOTSUPP;
861}
862
863int
864pru_connect2_notsupp(struct socket *so1, struct socket *so2)
865{
866 return EOPNOTSUPP;
867}
868
869int
870pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data,
dadab5e9 871 struct ifnet *ifp, struct thread *td)
984263bc
MD
872{
873 return EOPNOTSUPP;
874}
875
876int
dadab5e9 877pru_listen_notsupp(struct socket *so, struct thread *td)
984263bc
MD
878{
879 return EOPNOTSUPP;
880}
881
882int
883pru_rcvd_notsupp(struct socket *so, int flags)
884{
885 return EOPNOTSUPP;
886}
887
888int
889pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags)
890{
891 return EOPNOTSUPP;
892}
893
894/*
895 * This isn't really a ``null'' operation, but it's the default one
896 * and doesn't do anything destructive.
897 */
898int
899pru_sense_null(struct socket *so, struct stat *sb)
900{
901 sb->st_blksize = so->so_snd.sb_hiwat;
902 return 0;
903}
904
905/*
906 * Make a copy of a sockaddr in a malloced buffer of type M_SONAME.
907 */
908struct sockaddr *
909dup_sockaddr(sa, canwait)
910 struct sockaddr *sa;
911 int canwait;
912{
913 struct sockaddr *sa2;
914
915 MALLOC(sa2, struct sockaddr *, sa->sa_len, M_SONAME,
916 canwait ? M_WAITOK : M_NOWAIT);
917 if (sa2)
918 bcopy(sa, sa2, sa->sa_len);
919 return sa2;
920}
921
922/*
923 * Create an external-format (``xsocket'') structure using the information
924 * in the kernel-format socket structure pointed to by so. This is done
925 * to reduce the spew of irrelevant information over this interface,
926 * to isolate user code from changes in the kernel structure, and
927 * potentially to provide information-hiding if we decide that
928 * some of this information should be hidden from users.
929 */
930void
931sotoxsocket(struct socket *so, struct xsocket *xso)
932{
933 xso->xso_len = sizeof *xso;
934 xso->xso_so = so;
935 xso->so_type = so->so_type;
936 xso->so_options = so->so_options;
937 xso->so_linger = so->so_linger;
938 xso->so_state = so->so_state;
939 xso->so_pcb = so->so_pcb;
940 xso->xso_protocol = so->so_proto->pr_protocol;
941 xso->xso_family = so->so_proto->pr_domain->dom_family;
942 xso->so_qlen = so->so_qlen;
943 xso->so_incqlen = so->so_incqlen;
944 xso->so_qlimit = so->so_qlimit;
945 xso->so_timeo = so->so_timeo;
946 xso->so_error = so->so_error;
947 xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0;
948 xso->so_oobmark = so->so_oobmark;
949 sbtoxsockbuf(&so->so_snd, &xso->so_snd);
950 sbtoxsockbuf(&so->so_rcv, &xso->so_rcv);
951 xso->so_uid = so->so_cred->cr_uid;
952}
953
954/*
955 * This does the same for sockbufs. Note that the xsockbuf structure,
956 * since it is always embedded in a socket, does not include a self
957 * pointer nor a length. We make this entry point public in case
958 * some other mechanism needs it.
959 */
960void
961sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb)
962{
963 xsb->sb_cc = sb->sb_cc;
964 xsb->sb_hiwat = sb->sb_hiwat;
965 xsb->sb_mbcnt = sb->sb_mbcnt;
966 xsb->sb_mbmax = sb->sb_mbmax;
967 xsb->sb_lowat = sb->sb_lowat;
968 xsb->sb_flags = sb->sb_flags;
969 xsb->sb_timeo = sb->sb_timeo;
970}
971
972/*
973 * Here is the definition of some of the basic objects in the kern.ipc
974 * branch of the MIB.
975 */
976SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC");
977
978/* This takes the place of kern.maxsockbuf, which moved to kern.ipc. */
979static int dummy;
980SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, "");
981SYSCTL_OID(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLTYPE_INT|CTLFLAG_RW,
982 &sb_max, 0, sysctl_handle_sb_max, "I", "Maximum socket buffer size");
983SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD,
984 &maxsockets, 0, "Maximum number of sockets avaliable");
985SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW,
986 &sb_efficiency, 0, "");
987
988/*
989 * Initialise maxsockets
990 */
991static void init_maxsockets(void *ignored)
992{
993 TUNABLE_INT_FETCH("kern.ipc.maxsockets", &maxsockets);
994 maxsockets = imax(maxsockets, imax(maxfiles, nmbclusters));
995}
996SYSINIT(param, SI_SUB_TUNABLES, SI_ORDER_ANY, init_maxsockets, NULL);