proc->thread stage 3.5: Add an IO_CORE flag so coda doesn't have to dig
[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 $
1de703da 35 * $DragonFly: src/sys/kern/uipc_socket2.c,v 1.2 2003/06/17 04:28:41 dillon 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)
101 register struct socket *so;
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)
142 register struct socket *so;
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)
154 register struct socket *so;
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 *
174sonewconn(head, connstatus)
175 register struct socket *head;
176 int connstatus;
177{
178
179 return (sonewconn3(head, connstatus, NULL));
180}
181
182struct socket *
183sonewconn3(head, connstatus, p)
184 register struct socket *head;
185 int connstatus;
186 struct proc *p;
187{
188 register struct socket *so;
189
190 if (head->so_qlen > 3 * head->so_qlimit / 2)
191 return ((struct socket *)0);
192 so = soalloc(0);
193 if (so == NULL)
194 return ((struct socket *)0);
195 if ((head->so_options & SO_ACCEPTFILTER) != 0)
196 connstatus = 0;
197 so->so_head = head;
198 so->so_type = head->so_type;
199 so->so_options = head->so_options &~ SO_ACCEPTCONN;
200 so->so_linger = head->so_linger;
201 so->so_state = head->so_state | SS_NOFDREF;
202 so->so_proto = head->so_proto;
203 so->so_timeo = head->so_timeo;
204 so->so_cred = p ? p->p_ucred : head->so_cred;
205 crhold(so->so_cred);
206 if (soreserve(so, head->so_snd.sb_hiwat, head->so_rcv.sb_hiwat) ||
207 (*so->so_proto->pr_usrreqs->pru_attach)(so, 0, NULL)) {
208 sodealloc(so);
209 return ((struct socket *)0);
210 }
211
212 if (connstatus) {
213 TAILQ_INSERT_TAIL(&head->so_comp, so, so_list);
214 so->so_state |= SS_COMP;
215 head->so_qlen++;
216 } else {
217 if (head->so_incqlen > head->so_qlimit) {
218 struct socket *sp;
219 sp = TAILQ_FIRST(&head->so_incomp);
220 (void) soabort(sp);
221 }
222 TAILQ_INSERT_TAIL(&head->so_incomp, so, so_list);
223 so->so_state |= SS_INCOMP;
224 head->so_incqlen++;
225 }
226 if (connstatus) {
227 sorwakeup(head);
228 wakeup((caddr_t)&head->so_timeo);
229 so->so_state |= connstatus;
230 }
231 return (so);
232}
233
234/*
235 * Socantsendmore indicates that no more data will be sent on the
236 * socket; it would normally be applied to a socket when the user
237 * informs the system that no more data is to be sent, by the protocol
238 * code (in case PRU_SHUTDOWN). Socantrcvmore indicates that no more data
239 * will be received, and will normally be applied to the socket by a
240 * protocol when it detects that the peer will send no more data.
241 * Data queued for reading in the socket may yet be read.
242 */
243
244void
245socantsendmore(so)
246 struct socket *so;
247{
248
249 so->so_state |= SS_CANTSENDMORE;
250 sowwakeup(so);
251}
252
253void
254socantrcvmore(so)
255 struct socket *so;
256{
257
258 so->so_state |= SS_CANTRCVMORE;
259 sorwakeup(so);
260}
261
262/*
263 * Wait for data to arrive at/drain from a socket buffer.
264 */
265int
266sbwait(sb)
267 struct sockbuf *sb;
268{
269
270 sb->sb_flags |= SB_WAIT;
271 return (tsleep((caddr_t)&sb->sb_cc,
272 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK | PCATCH, "sbwait",
273 sb->sb_timeo));
274}
275
276/*
277 * Lock a sockbuf already known to be locked;
278 * return any error returned from sleep (EINTR).
279 */
280int
281sb_lock(sb)
282 register struct sockbuf *sb;
283{
284 int error;
285
286 while (sb->sb_flags & SB_LOCK) {
287 sb->sb_flags |= SB_WANT;
288 error = tsleep((caddr_t)&sb->sb_flags,
289 (sb->sb_flags & SB_NOINTR) ? PSOCK : PSOCK|PCATCH,
290 "sblock", 0);
291 if (error)
292 return (error);
293 }
294 sb->sb_flags |= SB_LOCK;
295 return (0);
296}
297
298/*
299 * Wakeup processes waiting on a socket buffer.
300 * Do asynchronous notification via SIGIO
301 * if the socket has the SS_ASYNC flag set.
302 */
303void
304sowakeup(so, sb)
305 register struct socket *so;
306 register struct sockbuf *sb;
307{
308 selwakeup(&sb->sb_sel);
309 sb->sb_flags &= ~SB_SEL;
310 if (sb->sb_flags & SB_WAIT) {
311 sb->sb_flags &= ~SB_WAIT;
312 wakeup((caddr_t)&sb->sb_cc);
313 }
314 if ((so->so_state & SS_ASYNC) && so->so_sigio != NULL)
315 pgsigio(so->so_sigio, SIGIO, 0);
316 if (sb->sb_flags & SB_UPCALL)
317 (*so->so_upcall)(so, so->so_upcallarg, M_DONTWAIT);
318 if (sb->sb_flags & SB_AIO)
319 aio_swake(so, sb);
320 KNOTE(&sb->sb_sel.si_note, 0);
321}
322
323/*
324 * Socket buffer (struct sockbuf) utility routines.
325 *
326 * Each socket contains two socket buffers: one for sending data and
327 * one for receiving data. Each buffer contains a queue of mbufs,
328 * information about the number of mbufs and amount of data in the
329 * queue, and other fields allowing select() statements and notification
330 * on data availability to be implemented.
331 *
332 * Data stored in a socket buffer is maintained as a list of records.
333 * Each record is a list of mbufs chained together with the m_next
334 * field. Records are chained together with the m_nextpkt field. The upper
335 * level routine soreceive() expects the following conventions to be
336 * observed when placing information in the receive buffer:
337 *
338 * 1. If the protocol requires each message be preceded by the sender's
339 * name, then a record containing that name must be present before
340 * any associated data (mbuf's must be of type MT_SONAME).
341 * 2. If the protocol supports the exchange of ``access rights'' (really
342 * just additional data associated with the message), and there are
343 * ``rights'' to be received, then a record containing this data
344 * should be present (mbuf's must be of type MT_RIGHTS).
345 * 3. If a name or rights record exists, then it must be followed by
346 * a data record, perhaps of zero length.
347 *
348 * Before using a new socket structure it is first necessary to reserve
349 * buffer space to the socket, by calling sbreserve(). This should commit
350 * some of the available buffer space in the system buffer pool for the
351 * socket (currently, it does nothing but enforce limits). The space
352 * should be released by calling sbrelease() when the socket is destroyed.
353 */
354
355int
356soreserve(so, sndcc, rcvcc)
357 register struct socket *so;
358 u_long sndcc, rcvcc;
359{
360 struct proc *p = curproc;
361
362 if (sbreserve(&so->so_snd, sndcc, so, p) == 0)
363 goto bad;
364 if (sbreserve(&so->so_rcv, rcvcc, so, p) == 0)
365 goto bad2;
366 if (so->so_rcv.sb_lowat == 0)
367 so->so_rcv.sb_lowat = 1;
368 if (so->so_snd.sb_lowat == 0)
369 so->so_snd.sb_lowat = MCLBYTES;
370 if (so->so_snd.sb_lowat > so->so_snd.sb_hiwat)
371 so->so_snd.sb_lowat = so->so_snd.sb_hiwat;
372 return (0);
373bad2:
374 sbrelease(&so->so_snd, so);
375bad:
376 return (ENOBUFS);
377}
378
379static int
380sysctl_handle_sb_max(SYSCTL_HANDLER_ARGS)
381{
382 int error = 0;
383 u_long old_sb_max = sb_max;
384
385 error = SYSCTL_OUT(req, arg1, sizeof(int));
386 if (error || !req->newptr)
387 return (error);
388 error = SYSCTL_IN(req, arg1, sizeof(int));
389 if (error)
390 return (error);
391 if (sb_max < MSIZE + MCLBYTES) {
392 sb_max = old_sb_max;
393 return (EINVAL);
394 }
395 sb_max_adj = (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES);
396 return (0);
397}
398
399/*
400 * Allot mbufs to a sockbuf.
401 * Attempt to scale mbmax so that mbcnt doesn't become limiting
402 * if buffering efficiency is near the normal case.
403 */
404int
405sbreserve(sb, cc, so, p)
406 struct sockbuf *sb;
407 u_long cc;
408 struct socket *so;
409 struct proc *p;
410{
411
412 /*
413 * p will only be NULL when we're in an interrupt
414 * (e.g. in tcp_input())
415 */
416 if (cc > sb_max_adj)
417 return (0);
418 if (!chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, cc,
419 p ? p->p_rlimit[RLIMIT_SBSIZE].rlim_cur : RLIM_INFINITY)) {
420 return (0);
421 }
422 sb->sb_mbmax = min(cc * sb_efficiency, sb_max);
423 if (sb->sb_lowat > sb->sb_hiwat)
424 sb->sb_lowat = sb->sb_hiwat;
425 return (1);
426}
427
428/*
429 * Free mbufs held by a socket, and reserved mbuf space.
430 */
431void
432sbrelease(sb, so)
433 struct sockbuf *sb;
434 struct socket *so;
435{
436
437 sbflush(sb);
438 (void)chgsbsize(so->so_cred->cr_uidinfo, &sb->sb_hiwat, 0,
439 RLIM_INFINITY);
440 sb->sb_mbmax = 0;
441}
442
443/*
444 * Routines to add and remove
445 * data from an mbuf queue.
446 *
447 * The routines sbappend() or sbappendrecord() are normally called to
448 * append new mbufs to a socket buffer, after checking that adequate
449 * space is available, comparing the function sbspace() with the amount
450 * of data to be added. sbappendrecord() differs from sbappend() in
451 * that data supplied is treated as the beginning of a new record.
452 * To place a sender's address, optional access rights, and data in a
453 * socket receive buffer, sbappendaddr() should be used. To place
454 * access rights and data in a socket receive buffer, sbappendrights()
455 * should be used. In either case, the new data begins a new record.
456 * Note that unlike sbappend() and sbappendrecord(), these routines check
457 * for the caller that there will be enough space to store the data.
458 * Each fails if there is not enough space, or if it cannot find mbufs
459 * to store additional information in.
460 *
461 * Reliable protocols may use the socket send buffer to hold data
462 * awaiting acknowledgement. Data is normally copied from a socket
463 * send buffer in a protocol with m_copy for output to a peer,
464 * and then removing the data from the socket buffer with sbdrop()
465 * or sbdroprecord() when the data is acknowledged by the peer.
466 */
467
468/*
469 * Append mbuf chain m to the last record in the
470 * socket buffer sb. The additional space associated
471 * the mbuf chain is recorded in sb. Empty mbufs are
472 * discarded and mbufs are compacted where possible.
473 */
474void
475sbappend(sb, m)
476 struct sockbuf *sb;
477 struct mbuf *m;
478{
479 register struct mbuf *n;
480
481 if (m == 0)
482 return;
483 n = sb->sb_mb;
484 if (n) {
485 while (n->m_nextpkt)
486 n = n->m_nextpkt;
487 do {
488 if (n->m_flags & M_EOR) {
489 sbappendrecord(sb, m); /* XXXXXX!!!! */
490 return;
491 }
492 } while (n->m_next && (n = n->m_next));
493 }
494 sbcompress(sb, m, n);
495}
496
497#ifdef SOCKBUF_DEBUG
498void
499sbcheck(sb)
500 register struct sockbuf *sb;
501{
502 register struct mbuf *m;
503 register struct mbuf *n = 0;
504 register u_long len = 0, mbcnt = 0;
505
506 for (m = sb->sb_mb; m; m = n) {
507 n = m->m_nextpkt;
508 for (; m; m = m->m_next) {
509 len += m->m_len;
510 mbcnt += MSIZE;
511 if (m->m_flags & M_EXT) /*XXX*/ /* pretty sure this is bogus */
512 mbcnt += m->m_ext.ext_size;
513 }
514 }
515 if (len != sb->sb_cc || mbcnt != sb->sb_mbcnt) {
516 printf("cc %ld != %ld || mbcnt %ld != %ld\n", len, sb->sb_cc,
517 mbcnt, sb->sb_mbcnt);
518 panic("sbcheck");
519 }
520}
521#endif
522
523/*
524 * As above, except the mbuf chain
525 * begins a new record.
526 */
527void
528sbappendrecord(sb, m0)
529 register struct sockbuf *sb;
530 register struct mbuf *m0;
531{
532 register struct mbuf *m;
533
534 if (m0 == 0)
535 return;
536 m = sb->sb_mb;
537 if (m)
538 while (m->m_nextpkt)
539 m = m->m_nextpkt;
540 /*
541 * Put the first mbuf on the queue.
542 * Note this permits zero length records.
543 */
544 sballoc(sb, m0);
545 if (m)
546 m->m_nextpkt = m0;
547 else
548 sb->sb_mb = m0;
549 m = m0->m_next;
550 m0->m_next = 0;
551 if (m && (m0->m_flags & M_EOR)) {
552 m0->m_flags &= ~M_EOR;
553 m->m_flags |= M_EOR;
554 }
555 sbcompress(sb, m, m0);
556}
557
558/*
559 * As above except that OOB data
560 * is inserted at the beginning of the sockbuf,
561 * but after any other OOB data.
562 */
563void
564sbinsertoob(sb, m0)
565 register struct sockbuf *sb;
566 register struct mbuf *m0;
567{
568 register struct mbuf *m;
569 register struct mbuf **mp;
570
571 if (m0 == 0)
572 return;
573 for (mp = &sb->sb_mb; *mp ; mp = &((*mp)->m_nextpkt)) {
574 m = *mp;
575 again:
576 switch (m->m_type) {
577
578 case MT_OOBDATA:
579 continue; /* WANT next train */
580
581 case MT_CONTROL:
582 m = m->m_next;
583 if (m)
584 goto again; /* inspect THIS train further */
585 }
586 break;
587 }
588 /*
589 * Put the first mbuf on the queue.
590 * Note this permits zero length records.
591 */
592 sballoc(sb, m0);
593 m0->m_nextpkt = *mp;
594 *mp = m0;
595 m = m0->m_next;
596 m0->m_next = 0;
597 if (m && (m0->m_flags & M_EOR)) {
598 m0->m_flags &= ~M_EOR;
599 m->m_flags |= M_EOR;
600 }
601 sbcompress(sb, m, m0);
602}
603
604/*
605 * Append address and data, and optionally, control (ancillary) data
606 * to the receive queue of a socket. If present,
607 * m0 must include a packet header with total length.
608 * Returns 0 if no space in sockbuf or insufficient mbufs.
609 */
610int
611sbappendaddr(sb, asa, m0, control)
612 register struct sockbuf *sb;
613 struct sockaddr *asa;
614 struct mbuf *m0, *control;
615{
616 register struct mbuf *m, *n;
617 int space = asa->sa_len;
618
619if (m0 && (m0->m_flags & M_PKTHDR) == 0)
620panic("sbappendaddr");
621 if (m0)
622 space += m0->m_pkthdr.len;
623 for (n = control; n; n = n->m_next) {
624 space += n->m_len;
625 if (n->m_next == 0) /* keep pointer to last control buf */
626 break;
627 }
628 if (space > sbspace(sb))
629 return (0);
630 if (asa->sa_len > MLEN)
631 return (0);
632 MGET(m, M_DONTWAIT, MT_SONAME);
633 if (m == 0)
634 return (0);
635 m->m_len = asa->sa_len;
636 bcopy((caddr_t)asa, mtod(m, caddr_t), asa->sa_len);
637 if (n)
638 n->m_next = m0; /* concatenate data to control */
639 else
640 control = m0;
641 m->m_next = control;
642 for (n = m; n; n = n->m_next)
643 sballoc(sb, n);
644 n = sb->sb_mb;
645 if (n) {
646 while (n->m_nextpkt)
647 n = n->m_nextpkt;
648 n->m_nextpkt = m;
649 } else
650 sb->sb_mb = m;
651 return (1);
652}
653
654int
655sbappendcontrol(sb, m0, control)
656 struct sockbuf *sb;
657 struct mbuf *control, *m0;
658{
659 register struct mbuf *m, *n;
660 int space = 0;
661
662 if (control == 0)
663 panic("sbappendcontrol");
664 for (m = control; ; m = m->m_next) {
665 space += m->m_len;
666 if (m->m_next == 0)
667 break;
668 }
669 n = m; /* save pointer to last control buffer */
670 for (m = m0; m; m = m->m_next)
671 space += m->m_len;
672 if (space > sbspace(sb))
673 return (0);
674 n->m_next = m0; /* concatenate data to control */
675 for (m = control; m; m = m->m_next)
676 sballoc(sb, m);
677 n = sb->sb_mb;
678 if (n) {
679 while (n->m_nextpkt)
680 n = n->m_nextpkt;
681 n->m_nextpkt = control;
682 } else
683 sb->sb_mb = control;
684 return (1);
685}
686
687/*
688 * Compress mbuf chain m into the socket
689 * buffer sb following mbuf n. If n
690 * is null, the buffer is presumed empty.
691 */
692void
693sbcompress(sb, m, n)
694 register struct sockbuf *sb;
695 register struct mbuf *m, *n;
696{
697 register int eor = 0;
698 register struct mbuf *o;
699
700 while (m) {
701 eor |= m->m_flags & M_EOR;
702 if (m->m_len == 0 &&
703 (eor == 0 ||
704 (((o = m->m_next) || (o = n)) &&
705 o->m_type == m->m_type))) {
706 m = m_free(m);
707 continue;
708 }
709 if (n && (n->m_flags & M_EOR) == 0 &&
710 M_WRITABLE(n) &&
711 m->m_len <= MCLBYTES / 4 && /* XXX: Don't copy too much */
712 m->m_len <= M_TRAILINGSPACE(n) &&
713 n->m_type == m->m_type) {
714 bcopy(mtod(m, caddr_t), mtod(n, caddr_t) + n->m_len,
715 (unsigned)m->m_len);
716 n->m_len += m->m_len;
717 sb->sb_cc += m->m_len;
718 m = m_free(m);
719 continue;
720 }
721 if (n)
722 n->m_next = m;
723 else
724 sb->sb_mb = m;
725 sballoc(sb, m);
726 n = m;
727 m->m_flags &= ~M_EOR;
728 m = m->m_next;
729 n->m_next = 0;
730 }
731 if (eor) {
732 if (n)
733 n->m_flags |= eor;
734 else
735 printf("semi-panic: sbcompress\n");
736 }
737}
738
739/*
740 * Free all mbufs in a sockbuf.
741 * Check that all resources are reclaimed.
742 */
743void
744sbflush(sb)
745 register struct sockbuf *sb;
746{
747
748 if (sb->sb_flags & SB_LOCK)
749 panic("sbflush: locked");
750 while (sb->sb_mbcnt) {
751 /*
752 * Don't call sbdrop(sb, 0) if the leading mbuf is non-empty:
753 * we would loop forever. Panic instead.
754 */
755 if (!sb->sb_cc && (sb->sb_mb == NULL || sb->sb_mb->m_len))
756 break;
757 sbdrop(sb, (int)sb->sb_cc);
758 }
759 if (sb->sb_cc || sb->sb_mb || sb->sb_mbcnt)
760 panic("sbflush: cc %ld || mb %p || mbcnt %ld", sb->sb_cc, (void *)sb->sb_mb, sb->sb_mbcnt);
761}
762
763/*
764 * Drop data from (the front of) a sockbuf.
765 */
766void
767sbdrop(sb, len)
768 register struct sockbuf *sb;
769 register int len;
770{
771 register struct mbuf *m;
772 struct mbuf *next;
773
774 next = (m = sb->sb_mb) ? m->m_nextpkt : 0;
775 while (len > 0) {
776 if (m == 0) {
777 if (next == 0)
778 panic("sbdrop");
779 m = next;
780 next = m->m_nextpkt;
781 continue;
782 }
783 if (m->m_len > len) {
784 m->m_len -= len;
785 m->m_data += len;
786 sb->sb_cc -= len;
787 break;
788 }
789 len -= m->m_len;
790 sbfree(sb, m);
791 m = m_free(m);
792 }
793 while (m && m->m_len == 0) {
794 sbfree(sb, m);
795 m = m_free(m);
796 }
797 if (m) {
798 sb->sb_mb = m;
799 m->m_nextpkt = next;
800 } else
801 sb->sb_mb = next;
802}
803
804/*
805 * Drop a record off the front of a sockbuf
806 * and move the next record to the front.
807 */
808void
809sbdroprecord(sb)
810 register struct sockbuf *sb;
811{
812 register struct mbuf *m;
813
814 m = sb->sb_mb;
815 if (m) {
816 sb->sb_mb = m->m_nextpkt;
817 do {
818 sbfree(sb, m);
819 m = m_free(m);
820 } while (m);
821 }
822}
823
824/*
825 * Create a "control" mbuf containing the specified data
826 * with the specified type for presentation on a socket buffer.
827 */
828struct mbuf *
829sbcreatecontrol(p, size, type, level)
830 caddr_t p;
831 register int size;
832 int type, level;
833{
834 register struct cmsghdr *cp;
835 struct mbuf *m;
836
837 if (CMSG_SPACE((u_int)size) > MCLBYTES)
838 return ((struct mbuf *) NULL);
839 if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL)
840 return ((struct mbuf *) NULL);
841 if (CMSG_SPACE((u_int)size) > MLEN) {
842 MCLGET(m, M_DONTWAIT);
843 if ((m->m_flags & M_EXT) == 0) {
844 m_free(m);
845 return ((struct mbuf *) NULL);
846 }
847 }
848 cp = mtod(m, struct cmsghdr *);
849 m->m_len = 0;
850 KASSERT(CMSG_SPACE((u_int)size) <= M_TRAILINGSPACE(m),
851 ("sbcreatecontrol: short mbuf"));
852 if (p != NULL)
853 (void)memcpy(CMSG_DATA(cp), p, size);
854 m->m_len = CMSG_SPACE(size);
855 cp->cmsg_len = CMSG_LEN(size);
856 cp->cmsg_level = level;
857 cp->cmsg_type = type;
858 return (m);
859}
860
861/*
862 * Some routines that return EOPNOTSUPP for entry points that are not
863 * supported by a protocol. Fill in as needed.
864 */
865int
866pru_accept_notsupp(struct socket *so, struct sockaddr **nam)
867{
868 return EOPNOTSUPP;
869}
870
871int
872pru_connect_notsupp(struct socket *so, struct sockaddr *nam, struct proc *p)
873{
874 return EOPNOTSUPP;
875}
876
877int
878pru_connect2_notsupp(struct socket *so1, struct socket *so2)
879{
880 return EOPNOTSUPP;
881}
882
883int
884pru_control_notsupp(struct socket *so, u_long cmd, caddr_t data,
885 struct ifnet *ifp, struct proc *p)
886{
887 return EOPNOTSUPP;
888}
889
890int
891pru_listen_notsupp(struct socket *so, struct proc *p)
892{
893 return EOPNOTSUPP;
894}
895
896int
897pru_rcvd_notsupp(struct socket *so, int flags)
898{
899 return EOPNOTSUPP;
900}
901
902int
903pru_rcvoob_notsupp(struct socket *so, struct mbuf *m, int flags)
904{
905 return EOPNOTSUPP;
906}
907
908/*
909 * This isn't really a ``null'' operation, but it's the default one
910 * and doesn't do anything destructive.
911 */
912int
913pru_sense_null(struct socket *so, struct stat *sb)
914{
915 sb->st_blksize = so->so_snd.sb_hiwat;
916 return 0;
917}
918
919/*
920 * Make a copy of a sockaddr in a malloced buffer of type M_SONAME.
921 */
922struct sockaddr *
923dup_sockaddr(sa, canwait)
924 struct sockaddr *sa;
925 int canwait;
926{
927 struct sockaddr *sa2;
928
929 MALLOC(sa2, struct sockaddr *, sa->sa_len, M_SONAME,
930 canwait ? M_WAITOK : M_NOWAIT);
931 if (sa2)
932 bcopy(sa, sa2, sa->sa_len);
933 return sa2;
934}
935
936/*
937 * Create an external-format (``xsocket'') structure using the information
938 * in the kernel-format socket structure pointed to by so. This is done
939 * to reduce the spew of irrelevant information over this interface,
940 * to isolate user code from changes in the kernel structure, and
941 * potentially to provide information-hiding if we decide that
942 * some of this information should be hidden from users.
943 */
944void
945sotoxsocket(struct socket *so, struct xsocket *xso)
946{
947 xso->xso_len = sizeof *xso;
948 xso->xso_so = so;
949 xso->so_type = so->so_type;
950 xso->so_options = so->so_options;
951 xso->so_linger = so->so_linger;
952 xso->so_state = so->so_state;
953 xso->so_pcb = so->so_pcb;
954 xso->xso_protocol = so->so_proto->pr_protocol;
955 xso->xso_family = so->so_proto->pr_domain->dom_family;
956 xso->so_qlen = so->so_qlen;
957 xso->so_incqlen = so->so_incqlen;
958 xso->so_qlimit = so->so_qlimit;
959 xso->so_timeo = so->so_timeo;
960 xso->so_error = so->so_error;
961 xso->so_pgid = so->so_sigio ? so->so_sigio->sio_pgid : 0;
962 xso->so_oobmark = so->so_oobmark;
963 sbtoxsockbuf(&so->so_snd, &xso->so_snd);
964 sbtoxsockbuf(&so->so_rcv, &xso->so_rcv);
965 xso->so_uid = so->so_cred->cr_uid;
966}
967
968/*
969 * This does the same for sockbufs. Note that the xsockbuf structure,
970 * since it is always embedded in a socket, does not include a self
971 * pointer nor a length. We make this entry point public in case
972 * some other mechanism needs it.
973 */
974void
975sbtoxsockbuf(struct sockbuf *sb, struct xsockbuf *xsb)
976{
977 xsb->sb_cc = sb->sb_cc;
978 xsb->sb_hiwat = sb->sb_hiwat;
979 xsb->sb_mbcnt = sb->sb_mbcnt;
980 xsb->sb_mbmax = sb->sb_mbmax;
981 xsb->sb_lowat = sb->sb_lowat;
982 xsb->sb_flags = sb->sb_flags;
983 xsb->sb_timeo = sb->sb_timeo;
984}
985
986/*
987 * Here is the definition of some of the basic objects in the kern.ipc
988 * branch of the MIB.
989 */
990SYSCTL_NODE(_kern, KERN_IPC, ipc, CTLFLAG_RW, 0, "IPC");
991
992/* This takes the place of kern.maxsockbuf, which moved to kern.ipc. */
993static int dummy;
994SYSCTL_INT(_kern, KERN_DUMMY, dummy, CTLFLAG_RW, &dummy, 0, "");
995SYSCTL_OID(_kern_ipc, KIPC_MAXSOCKBUF, maxsockbuf, CTLTYPE_INT|CTLFLAG_RW,
996 &sb_max, 0, sysctl_handle_sb_max, "I", "Maximum socket buffer size");
997SYSCTL_INT(_kern_ipc, OID_AUTO, maxsockets, CTLFLAG_RD,
998 &maxsockets, 0, "Maximum number of sockets avaliable");
999SYSCTL_INT(_kern_ipc, KIPC_SOCKBUF_WASTE, sockbuf_waste_factor, CTLFLAG_RW,
1000 &sb_efficiency, 0, "");
1001
1002/*
1003 * Initialise maxsockets
1004 */
1005static void init_maxsockets(void *ignored)
1006{
1007 TUNABLE_INT_FETCH("kern.ipc.maxsockets", &maxsockets);
1008 maxsockets = imax(maxsockets, imax(maxfiles, nmbclusters));
1009}
1010SYSINIT(param, SI_SUB_TUNABLES, SI_ORDER_ANY, init_maxsockets, NULL);