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_socket.c 8.3 (Berkeley) 4/15/94
34 * $FreeBSD: head/sys/kern/uipc_socket.c 92654 2002-03-19 09:11:49Z jeff $
35 */
36
37#include "opt_inet.h"
38
39#include <sys/param.h>
40#include <sys/systm.h>
41#include <sys/fcntl.h>
42#include <sys/lock.h>
43#include <sys/malloc.h>
44#include <sys/mbuf.h>
45#include <sys/mutex.h>
46#include <sys/domain.h>
47#include <sys/file.h> /* for struct knote */
48#include <sys/kernel.h>
49#include <sys/malloc.h>
50#include <sys/event.h>
51#include <sys/poll.h>
52#include <sys/proc.h>
53#include <sys/protosw.h>
54#include <sys/socket.h>
55#include <sys/socketvar.h>
56#include <sys/resourcevar.h>
57#include <sys/signalvar.h>
58#include <sys/sysctl.h>
59#include <sys/uio.h>
60#include <sys/jail.h>
61
62#include <vm/vm_zone.h>
63
64#include <machine/limits.h>
65
66#ifdef INET
67static int do_setopt_accept_filter(struct socket *so, struct sockopt *sopt);
68#endif
69
70static void filt_sordetach(struct knote *kn);
71static int filt_soread(struct knote *kn, long hint);
72static void filt_sowdetach(struct knote *kn);
73static int filt_sowrite(struct knote *kn, long hint);
74static int filt_solisten(struct knote *kn, long hint);
75
76static struct filterops solisten_filtops =
77 { 1, NULL, filt_sordetach, filt_solisten };
78static struct filterops soread_filtops =
79 { 1, NULL, filt_sordetach, filt_soread };
80static struct filterops sowrite_filtops =
81 { 1, NULL, filt_sowdetach, filt_sowrite };
82
83vm_zone_t socket_zone;
84so_gen_t so_gencnt; /* generation count for sockets */
85
86MALLOC_DEFINE(M_SONAME, "soname", "socket name");
87MALLOC_DEFINE(M_PCB, "pcb", "protocol control block");
88
89SYSCTL_DECL(_kern_ipc);
90
91static int somaxconn = SOMAXCONN;
92SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW,
93 &somaxconn, 0, "Maximum pending socket connection queue size");
94static int numopensockets;
95SYSCTL_INT(_kern_ipc, OID_AUTO, numopensockets, CTLFLAG_RD,
96 &numopensockets, 0, "Number of open sockets");
97
98
99/*
100 * Socket operation routines.
101 * These routines are called by the routines in
102 * sys_socket.c or from a system process, and
103 * implement the semantics of socket operations by
104 * switching out to the protocol specific routines.
105 */
106
107/*
108 * Get a socket structure from our zone, and initialize it.
109 * We don't implement `waitok' yet (see comments in uipc_domain.c).
110 * Note that it would probably be better to allocate socket
111 * and PCB at the same time, but I'm not convinced that all
112 * the protocols can be easily modified to do this.
113 *
114 * soalloc() returns a socket with a ref count of 0.
115 */
116struct socket *
117soalloc(waitok)
118 int waitok;
119{
120 struct socket *so;
121
122 so = zalloc(socket_zone);
123 if (so) {
124 /* XXX race condition for reentrant kernel */
125 bzero(so, sizeof *so);
126 so->so_gencnt = ++so_gencnt;
127 so->so_zone = socket_zone;
128 /* sx_init(&so->so_sxlock, "socket sxlock"); */
129 TAILQ_INIT(&so->so_aiojobq);
130 ++numopensockets;
131 }
132 return so;
133}
134
135/*
136 * socreate returns a socket with a ref count of 1. The socket should be
137 * closed with soclose().
138 */
139int
140socreate(dom, aso, type, proto, cred, td)
141 int dom;
142 struct socket **aso;
143 register int type;
144 int proto;
145 struct ucred *cred;
146 struct thread *td;
147{
148 register struct protosw *prp;
149 register struct socket *so;
150 register int error;
151
152 if (proto)
153 prp = pffindproto(dom, proto, type);
154 else
155 prp = pffindtype(dom, type);
156
157 if (prp == 0 || prp->pr_usrreqs->pru_attach == 0)
158 return (EPROTONOSUPPORT);
159
160 if (jailed(td->td_ucred) && jail_socket_unixiproute_only &&
161 prp->pr_domain->dom_family != PF_LOCAL &&
162 prp->pr_domain->dom_family != PF_INET &&
163 prp->pr_domain->dom_family != PF_ROUTE) {
164 return (EPROTONOSUPPORT);
165 }
166
167 if (prp->pr_type != type)
168 return (EPROTOTYPE);
169 so = soalloc(td != 0);
170 if (so == 0)
171 return (ENOBUFS);
172
173 TAILQ_INIT(&so->so_incomp);
174 TAILQ_INIT(&so->so_comp);
175 so->so_type = type;
176 so->so_cred = crhold(cred);
177 so->so_proto = prp;
178 soref(so);
179 error = (*prp->pr_usrreqs->pru_attach)(so, proto, td);
180 if (error) {
181 so->so_state |= SS_NOFDREF;
182 sorele(so);
183 return (error);
184 }
185 *aso = so;
186 return (0);
187}
188
189int
190sobind(so, nam, td)
191 struct socket *so;
192 struct sockaddr *nam;
193 struct thread *td;
194{
195 int s = splnet();
196 int error;
197
198 error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam, td);
199 splx(s);
200 return (error);
201}
202
203static void
204sodealloc(struct socket *so)
205{
206
207 KASSERT(so->so_count == 0, ("sodealloc(): so_count %d", so->so_count));
208 so->so_gencnt = ++so_gencnt;
209 if (so->so_rcv.sb_hiwat)
210 (void)chgsbsize(so->so_cred->cr_uidinfo,
211 &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY);
212 if (so->so_snd.sb_hiwat)
213 (void)chgsbsize(so->so_cred->cr_uidinfo,
214 &so->so_snd.sb_hiwat, 0, RLIM_INFINITY);
215#ifdef INET
216 if (so->so_accf != NULL) {
217 if (so->so_accf->so_accept_filter != NULL &&
218 so->so_accf->so_accept_filter->accf_destroy != NULL) {
219 so->so_accf->so_accept_filter->accf_destroy(so);
220 }
221 if (so->so_accf->so_accept_filter_str != NULL)
222 FREE(so->so_accf->so_accept_filter_str, M_ACCF);
223 FREE(so->so_accf, M_ACCF);
224 }
225#endif
226 crfree(so->so_cred);
227 /* sx_destroy(&so->so_sxlock); */
228 zfree(so->so_zone, so);
229 --numopensockets;
230}
231
232int
233solisten(so, backlog, td)
234 register struct socket *so;
235 int backlog;
236 struct thread *td;
237{
238 int s, error;
239
240 s = splnet();
241 error = (*so->so_proto->pr_usrreqs->pru_listen)(so, td);
242 if (error) {
243 splx(s);
244 return (error);
245 }
246 if (TAILQ_EMPTY(&so->so_comp))
247 so->so_options |= SO_ACCEPTCONN;
248 if (backlog < 0 || backlog > somaxconn)
249 backlog = somaxconn;
250 so->so_qlimit = backlog;
251 splx(s);
252 return (0);
253}
254
255void
256sofree(so)
257 register struct socket *so;
258{
259 struct socket *head = so->so_head;
260
261 KASSERT(so->so_count == 0, ("socket %p so_count not 0", so));
262
263 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
264 return;
265 if (head != NULL) {
266 if (so->so_state & SS_INCOMP) {
267 TAILQ_REMOVE(&head->so_incomp, so, so_list);
268 head->so_incqlen--;
269 } else if (so->so_state & SS_COMP) {
270 /*
271 * We must not decommission a socket that's
272 * on the accept(2) queue. If we do, then
273 * accept(2) may hang after select(2) indicated
274 * that the listening socket was ready.
275 */
276 return;
277 } else {
278 panic("sofree: not queued");
279 }
280 head->so_qlen--;
281 so->so_state &= ~SS_INCOMP;
282 so->so_head = NULL;
283 }
284 sbrelease(&so->so_snd, so);
285 sorflush(so);
286 sodealloc(so);
287}
288
289/*
290 * Close a socket on last file table reference removal.
291 * Initiate disconnect if connected.
292 * Free socket when disconnect complete.
293 *
294 * This function will sorele() the socket. Note that soclose() may be
295 * called prior to the ref count reaching zero. The actual socket
296 * structure will not be freed until the ref count reaches zero.
297 */
298int
299soclose(so)
300 register struct socket *so;
301{
302 int s = splnet(); /* conservative */
303 int error = 0;
304
305 funsetown(so->so_sigio);
306 if (so->so_options & SO_ACCEPTCONN) {
307 struct socket *sp, *sonext;
308
309 sp = TAILQ_FIRST(&so->so_incomp);
310 for (; sp != NULL; sp = sonext) {
311 sonext = TAILQ_NEXT(sp, so_list);
312 (void) soabort(sp);
313 }
314 for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) {
315 sonext = TAILQ_NEXT(sp, so_list);
316 /* Dequeue from so_comp since sofree() won't do it */
317 TAILQ_REMOVE(&so->so_comp, sp, so_list);
318 so->so_qlen--;
319 sp->so_state &= ~SS_COMP;
320 sp->so_head = NULL;
321 (void) soabort(sp);
322 }
323 }
324 if (so->so_pcb == 0)
325 goto discard;
326 if (so->so_state & SS_ISCONNECTED) {
327 if ((so->so_state & SS_ISDISCONNECTING) == 0) {
328 error = sodisconnect(so);
329 if (error)
330 goto drop;
331 }
332 if (so->so_options & SO_LINGER) {
333 if ((so->so_state & SS_ISDISCONNECTING) &&
334 (so->so_state & SS_NBIO))
335 goto drop;
336 while (so->so_state & SS_ISCONNECTED) {
337 error = tsleep((caddr_t)&so->so_timeo,
338 PSOCK | PCATCH, "soclos", so->so_linger * hz);
339 if (error)
340 break;
341 }
342 }
343 }
344drop:
345 if (so->so_pcb) {
346 int error2 = (*so->so_proto->pr_usrreqs->pru_detach)(so);
347 if (error == 0)
348 error = error2;
349 }
350discard:
351 if (so->so_state & SS_NOFDREF)
352 panic("soclose: NOFDREF");
353 so->so_state |= SS_NOFDREF;
354 sorele(so);
355 splx(s);
356 return (error);
357}
358
359/*
360 * Must be called at splnet...
361 */
362int
363soabort(so)
364 struct socket *so;
365{
366 int error;
367
368 error = (*so->so_proto->pr_usrreqs->pru_abort)(so);
369 if (error) {
370 sotryfree(so); /* note: does not decrement the ref count */
371 return error;
372 }
373 return (0);
374}
375
376int
377soaccept(so, nam)
378 register struct socket *so;
379 struct sockaddr **nam;
380{
381 int s = splnet();
382 int error;
383
384 if ((so->so_state & SS_NOFDREF) == 0)
385 panic("soaccept: !NOFDREF");
386 so->so_state &= ~SS_NOFDREF;
387 error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam);
388 splx(s);
389 return (error);
390}
391
392int
393soconnect(so, nam, td)
394 register struct socket *so;
395 struct sockaddr *nam;
396 struct thread *td;
397{
398 int s;
399 int error;
400
401 if (so->so_options & SO_ACCEPTCONN)
402 return (EOPNOTSUPP);
403 s = splnet();
404 /*
405 * If protocol is connection-based, can only connect once.
406 * Otherwise, if connected, try to disconnect first.
407 * This allows user to disconnect by connecting to, e.g.,
408 * a null address.
409 */
410 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
411 ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
412 (error = sodisconnect(so))))
413 error = EISCONN;
414 else
415 error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, td);
416 splx(s);
417 return (error);
418}
419
420int
421soconnect2(so1, so2)
422 register struct socket *so1;
423 struct socket *so2;
424{
425 int s = splnet();
426 int error;
427
428 error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2);
429 splx(s);
430 return (error);
431}
432
433int
434sodisconnect(so)
435 register struct socket *so;
436{
437 int s = splnet();
438 int error;
439
440 if ((so->so_state & SS_ISCONNECTED) == 0) {
441 error = ENOTCONN;
442 goto bad;
443 }
444 if (so->so_state & SS_ISDISCONNECTING) {
445 error = EALREADY;
446 goto bad;
447 }
448 error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so);
449bad:
450 splx(s);
451 return (error);
452}
453
454#define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
455/*
456 * Send on a socket.
457 * If send must go all at once and message is larger than
458 * send buffering, then hard error.
459 * Lock against other senders.
460 * If must go all at once and not enough room now, then
461 * inform user that this would block and do nothing.
462 * Otherwise, if nonblocking, send as much as possible.
463 * The data to be sent is described by "uio" if nonzero,
464 * otherwise by the mbuf chain "top" (which must be null
465 * if uio is not). Data provided in mbuf chain must be small
466 * enough to send all at once.
467 *
468 * Returns nonzero on error, timeout or signal; callers
469 * must check for short counts if EINTR/ERESTART are returned.
470 * Data and control buffers are freed on return.
471 */
472int
473sosend(so, addr, uio, top, control, flags, td)
474 register struct socket *so;
475 struct sockaddr *addr;
476 struct uio *uio;
477 struct mbuf *top;
478 struct mbuf *control;
479 int flags;
480 struct thread *td;
481{
482 struct mbuf **mp;
483 register struct mbuf *m;
484 register long space, len, resid;
485 int clen = 0, error, s, dontroute, mlen;
486 int atomic = sosendallatonce(so) || top;
487
488 if (uio)
489 resid = uio->uio_resid;
490 else
491 resid = top->m_pkthdr.len;
492 /*
493 * In theory resid should be unsigned.
494 * However, space must be signed, as it might be less than 0
495 * if we over-committed, and we must use a signed comparison
496 * of space and resid. On the other hand, a negative resid
497 * causes us to loop sending 0-length segments to the protocol.
498 *
499 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
500 * type sockets since that's an error.
501 */
502 if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) {
503 error = EINVAL;
504 goto out;
505 }
506
507 dontroute =
508 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
509 (so->so_proto->pr_flags & PR_ATOMIC);
510 if (td)
511 td->td_proc->p_stats->p_ru.ru_msgsnd++;
512 if (control)
513 clen = control->m_len;
514#define snderr(errno) { error = errno; splx(s); goto release; }
515
516restart:
517 error = sblock(&so->so_snd, SBLOCKWAIT(flags));
518 if (error)
519 goto out;
520 do {
521 s = splnet();
522 if (so->so_state & SS_CANTSENDMORE)
523 snderr(EPIPE);
524 if (so->so_error) {
525 error = so->so_error;
526 so->so_error = 0;
527 splx(s);
528 goto release;
529 }
530 if ((so->so_state & SS_ISCONNECTED) == 0) {
531 /*
532 * `sendto' and `sendmsg' is allowed on a connection-
533 * based socket if it supports implied connect.
534 * Return ENOTCONN if not connected and no address is
535 * supplied.
536 */
537 if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
538 (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
539 if ((so->so_state & SS_ISCONFIRMING) == 0 &&
540 !(resid == 0 && clen != 0))
541 snderr(ENOTCONN);
542 } else if (addr == 0)
543 snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
544 ENOTCONN : EDESTADDRREQ);
545 }
546 space = sbspace(&so->so_snd);
547 if (flags & MSG_OOB)
548 space += 1024;
549 if ((atomic && resid > so->so_snd.sb_hiwat) ||
550 clen > so->so_snd.sb_hiwat)
551 snderr(EMSGSIZE);
552 if (space < resid + clen &&
553 (atomic || space < so->so_snd.sb_lowat || space < clen)) {
554 if (so->so_state & SS_NBIO)
555 snderr(EWOULDBLOCK);
556 sbunlock(&so->so_snd);
557 error = sbwait(&so->so_snd);
558 splx(s);
559 if (error)
560 goto out;
561 goto restart;
562 }
563 splx(s);
564 mp = ⊤
565 space -= clen;
566 do {
567 if (uio == NULL) {
568 /*
569 * Data is prepackaged in "top".
570 */
571 resid = 0;
572 if (flags & MSG_EOR)
573 top->m_flags |= M_EOR;
574 } else do {
575 if (top == 0) {
576 MGETHDR(m, M_TRYWAIT, MT_DATA);
577 if (m == NULL) {
578 error = ENOBUFS;
579 goto release;
580 }
581 mlen = MHLEN;
582 m->m_pkthdr.len = 0;
583 m->m_pkthdr.rcvif = (struct ifnet *)0;
584 } else {
585 MGET(m, M_TRYWAIT, MT_DATA);
586 if (m == NULL) {
587 error = ENOBUFS;
588 goto release;
589 }
590 mlen = MLEN;
591 }
592 if (resid >= MINCLSIZE) {
593 MCLGET(m, M_TRYWAIT);
594 if ((m->m_flags & M_EXT) == 0)
595 goto nopages;
596 mlen = MCLBYTES;
597 len = min(min(mlen, resid), space);
598 } else {
599nopages:
600 len = min(min(mlen, resid), space);
601 /*
602 * For datagram protocols, leave room
603 * for protocol headers in first mbuf.
604 */
605 if (atomic && top == 0 && len < mlen)
606 MH_ALIGN(m, len);
607 }
608 space -= len;
609 error = uiomove(mtod(m, caddr_t), (int)len, uio);
610 resid = uio->uio_resid;
611 m->m_len = len;
612 *mp = m;
613 top->m_pkthdr.len += len;
614 if (error)
615 goto release;
616 mp = &m->m_next;
617 if (resid <= 0) {
618 if (flags & MSG_EOR)
619 top->m_flags |= M_EOR;
620 break;
621 }
622 } while (space > 0 && atomic);
623 if (dontroute)
624 so->so_options |= SO_DONTROUTE;
625 s = splnet(); /* XXX */
626 /*
627 * XXX all the SS_CANTSENDMORE checks previously
628 * done could be out of date. We could have recieved
629 * a reset packet in an interrupt or maybe we slept
630 * while doing page faults in uiomove() etc. We could
631 * probably recheck again inside the splnet() protection
632 * here, but there are probably other places that this
633 * also happens. We must rethink this.
634 */
635 error = (*so->so_proto->pr_usrreqs->pru_send)(so,
636 (flags & MSG_OOB) ? PRUS_OOB :
637 /*
638 * If the user set MSG_EOF, the protocol
639 * understands this flag and nothing left to
640 * send then use PRU_SEND_EOF instead of PRU_SEND.
641 */
642 ((flags & MSG_EOF) &&
643 (so->so_proto->pr_flags & PR_IMPLOPCL) &&
644 (resid <= 0)) ?
645 PRUS_EOF :
646 /* If there is more to send set PRUS_MORETOCOME */
647 (resid > 0 && space > 0) ? PRUS_MORETOCOME : 0,
648 top, addr, control, td);
649 splx(s);
650 if (dontroute)
651 so->so_options &= ~SO_DONTROUTE;
652 clen = 0;
653 control = 0;
654 top = 0;
655 mp = ⊤
656 if (error)
657 goto release;
658 } while (resid && space > 0);
659 } while (resid);
660
661release:
662 sbunlock(&so->so_snd);
663out:
664 if (top)
665 m_freem(top);
666 if (control)
667 m_freem(control);
668 return (error);
669}
670
671/*
672 * Implement receive operations on a socket.
673 * We depend on the way that records are added to the sockbuf
674 * by sbappend*. In particular, each record (mbufs linked through m_next)
675 * must begin with an address if the protocol so specifies,
676 * followed by an optional mbuf or mbufs containing ancillary data,
677 * and then zero or more mbufs of data.
678 * In order to avoid blocking network interrupts for the entire time here,
679 * we splx() while doing the actual copy to user space.
680 * Although the sockbuf is locked, new data may still be appended,
681 * and thus we must maintain consistency of the sockbuf during that time.
682 *
683 * The caller may receive the data as a single mbuf chain by supplying
684 * an mbuf **mp0 for use in returning the chain. The uio is then used
685 * only for the count in uio_resid.
686 */
687int
688soreceive(so, psa, uio, mp0, controlp, flagsp)
689 register struct socket *so;
690 struct sockaddr **psa;
691 struct uio *uio;
692 struct mbuf **mp0;
693 struct mbuf **controlp;
694 int *flagsp;
695{
696 struct mbuf *m, **mp;
697 register int flags, len, error, s, offset;
698 struct protosw *pr = so->so_proto;
699 struct mbuf *nextrecord;
700 int moff, type = 0;
701 int orig_resid = uio->uio_resid;
702
703 mp = mp0;
704 if (psa)
705 *psa = 0;
706 if (controlp)
707 *controlp = 0;
708 if (flagsp)
709 flags = *flagsp &~ MSG_EOR;
710 else
711 flags = 0;
712 if (flags & MSG_OOB) {
713 m = m_get(M_TRYWAIT, MT_DATA);
714 if (m == NULL)
715 return (ENOBUFS);
716 error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK);
717 if (error)
718 goto bad;
719 do {
720 error = uiomove(mtod(m, caddr_t),
721 (int) min(uio->uio_resid, m->m_len), uio);
722 m = m_free(m);
723 } while (uio->uio_resid && error == 0 && m);
724bad:
725 if (m)
726 m_freem(m);
727 return (error);
728 }
729 if (mp)
730 *mp = (struct mbuf *)0;
731 if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
732 (*pr->pr_usrreqs->pru_rcvd)(so, 0);
733
734restart:
735 error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
736 if (error)
737 return (error);
738 s = splnet();
739
740 m = so->so_rcv.sb_mb;
741 /*
742 * If we have less data than requested, block awaiting more
743 * (subject to any timeout) if:
744 * 1. the current count is less than the low water mark, or
745 * 2. MSG_WAITALL is set, and it is possible to do the entire
746 * receive operation at once if we block (resid <= hiwat).
747 * 3. MSG_DONTWAIT is not set
748 * If MSG_WAITALL is set but resid is larger than the receive buffer,
749 * we have to do the receive in sections, and thus risk returning
750 * a short count if a timeout or signal occurs after we start.
751 */
752 if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
753 so->so_rcv.sb_cc < uio->uio_resid) &&
754 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
755 ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
756 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
757 KASSERT(m != 0 || !so->so_rcv.sb_cc,
758 ("receive: m == %p so->so_rcv.sb_cc == %lu",
759 m, so->so_rcv.sb_cc));
760 if (so->so_error) {
761 if (m)
762 goto dontblock;
763 error = so->so_error;
764 if ((flags & MSG_PEEK) == 0)
765 so->so_error = 0;
766 goto release;
767 }
768 if (so->so_state & SS_CANTRCVMORE) {
769 if (m)
770 goto dontblock;
771 else
772 goto release;
773 }
774 for (; m; m = m->m_next)
775 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
776 m = so->so_rcv.sb_mb;
777 goto dontblock;
778 }
779 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
780 (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
781 error = ENOTCONN;
782 goto release;
783 }
784 if (uio->uio_resid == 0)
785 goto release;
786 if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
787 error = EWOULDBLOCK;
788 goto release;
789 }
790 sbunlock(&so->so_rcv);
791 error = sbwait(&so->so_rcv);
792 splx(s);
793 if (error)
794 return (error);
795 goto restart;
796 }
797dontblock:
798 if (uio->uio_td)
799 uio->uio_td->td_proc->p_stats->p_ru.ru_msgrcv++;
800 nextrecord = m->m_nextpkt;
801 if (pr->pr_flags & PR_ADDR) {
802 KASSERT(m->m_type == MT_SONAME,
803 ("m->m_type == %d", m->m_type));
804 orig_resid = 0;
805 if (psa)
806 *psa = dup_sockaddr(mtod(m, struct sockaddr *),
807 mp0 == 0);
808 if (flags & MSG_PEEK) {
809 m = m->m_next;
810 } else {
811 sbfree(&so->so_rcv, m);
812 so->so_rcv.sb_mb = m_free(m);
813 m = so->so_rcv.sb_mb;
814 }
815 }
816 while (m && m->m_type == MT_CONTROL && error == 0) {
817 if (flags & MSG_PEEK) {
818 if (controlp)
819 *controlp = m_copy(m, 0, m->m_len);
820 m = m->m_next;
821 } else {
822 sbfree(&so->so_rcv, m);
823 so->so_rcv.sb_mb = m->m_next;
824 m->m_next = NULL;
825 if (pr->pr_domain->dom_externalize)
826 error =
827 (*pr->pr_domain->dom_externalize)(m, controlp);
828 else if (controlp)
829 *controlp = m;
830 else
831 m_freem(m);
832 m = so->so_rcv.sb_mb;
833 }
834 if (controlp) {
835 orig_resid = 0;
836 do
837 controlp = &(*controlp)->m_next;
838 while (*controlp != NULL);
839 }
840 }
841 if (m) {
842 if ((flags & MSG_PEEK) == 0)
843 m->m_nextpkt = nextrecord;
844 type = m->m_type;
845 if (type == MT_OOBDATA)
846 flags |= MSG_OOB;
847 }
848 moff = 0;
849 offset = 0;
850 while (m && uio->uio_resid > 0 && error == 0) {
851 if (m->m_type == MT_OOBDATA) {
852 if (type != MT_OOBDATA)
853 break;
854 } else if (type == MT_OOBDATA)
855 break;
856 else
857 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
858 ("m->m_type == %d", m->m_type));
859 so->so_state &= ~SS_RCVATMARK;
860 len = uio->uio_resid;
861 if (so->so_oobmark && len > so->so_oobmark - offset)
862 len = so->so_oobmark - offset;
863 if (len > m->m_len - moff)
864 len = m->m_len - moff;
865 /*
866 * If mp is set, just pass back the mbufs.
867 * Otherwise copy them out via the uio, then free.
868 * Sockbuf must be consistent here (points to current mbuf,
869 * it points to next record) when we drop priority;
870 * we must note any additions to the sockbuf when we
871 * block interrupts again.
872 */
873 if (mp == 0) {
874 splx(s);
875 error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
876 s = splnet();
877 if (error)
878 goto release;
879 } else
880 uio->uio_resid -= len;
881 if (len == m->m_len - moff) {
882 if (m->m_flags & M_EOR)
883 flags |= MSG_EOR;
884 if (flags & MSG_PEEK) {
885 m = m->m_next;
886 moff = 0;
887 } else {
888 nextrecord = m->m_nextpkt;
889 sbfree(&so->so_rcv, m);
890 if (mp) {
891 *mp = m;
892 mp = &m->m_next;
893 so->so_rcv.sb_mb = m = m->m_next;
894 *mp = (struct mbuf *)0;
895 } else {
896 so->so_rcv.sb_mb = m_free(m);
897 m = so->so_rcv.sb_mb;
898 }
899 if (m)
900 m->m_nextpkt = nextrecord;
901 }
902 } else {
903 if (flags & MSG_PEEK)
904 moff += len;
905 else {
906 if (mp)
907 *mp = m_copym(m, 0, len, M_TRYWAIT);
908 m->m_data += len;
909 m->m_len -= len;
910 so->so_rcv.sb_cc -= len;
911 }
912 }
913 if (so->so_oobmark) {
914 if ((flags & MSG_PEEK) == 0) {
915 so->so_oobmark -= len;
916 if (so->so_oobmark == 0) {
917 so->so_state |= SS_RCVATMARK;
918 break;
919 }
920 } else {
921 offset += len;
922 if (offset == so->so_oobmark)
923 break;
924 }
925 }
926 if (flags & MSG_EOR)
927 break;
928 /*
929 * If the MSG_WAITALL flag is set (for non-atomic socket),
930 * we must not quit until "uio->uio_resid == 0" or an error
931 * termination. If a signal/timeout occurs, return
932 * with a short count but without error.
933 * Keep sockbuf locked against other readers.
934 */
935 while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
936 !sosendallatonce(so) && !nextrecord) {
937 if (so->so_error || so->so_state & SS_CANTRCVMORE)
938 break;
939 /*
940 * Notify the protocol that some data has been
941 * drained before blocking.
942 */
943 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
944 (*pr->pr_usrreqs->pru_rcvd)(so, flags);
945 error = sbwait(&so->so_rcv);
946 if (error) {
947 sbunlock(&so->so_rcv);
948 splx(s);
949 return (0);
950 }
951 m = so->so_rcv.sb_mb;
952 if (m)
953 nextrecord = m->m_nextpkt;
954 }
955 }
956
957 if (m && pr->pr_flags & PR_ATOMIC) {
958 flags |= MSG_TRUNC;
959 if ((flags & MSG_PEEK) == 0)
960 (void) sbdroprecord(&so->so_rcv);
961 }
962 if ((flags & MSG_PEEK) == 0) {
963 if (m == 0)
964 so->so_rcv.sb_mb = nextrecord;
965 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
966 (*pr->pr_usrreqs->pru_rcvd)(so, flags);
967 }
968 if (orig_resid == uio->uio_resid && orig_resid &&
969 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
970 sbunlock(&so->so_rcv);
971 splx(s);
972 goto restart;
973 }
974
975 if (flagsp)
976 *flagsp |= flags;
977release:
978 sbunlock(&so->so_rcv);
979 splx(s);
980 return (error);
981}
982
983int
984soshutdown(so, how)
985 register struct socket *so;
986 register int how;
987{
988 register struct protosw *pr = so->so_proto;
989
990 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
991 return (EINVAL);
992
993 if (how != SHUT_WR)
994 sorflush(so);
995 if (how != SHUT_RD)
996 return ((*pr->pr_usrreqs->pru_shutdown)(so));
997 return (0);
998}
999
1000void
1001sorflush(so)
1002 register struct socket *so;
1003{
1004 register struct sockbuf *sb = &so->so_rcv;
1005 register struct protosw *pr = so->so_proto;
1006 register int s;
1007 struct sockbuf asb;
1008
1009 sb->sb_flags |= SB_NOINTR;
1010 (void) sblock(sb, M_WAITOK);
1011 s = splimp();
1012 socantrcvmore(so);
1013 sbunlock(sb);
1014 asb = *sb;
1015 bzero((caddr_t)sb, sizeof (*sb));
1016 splx(s);
1017 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1018 (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1019 sbrelease(&asb, so);
1020}
1021
1022#ifdef INET
1023static int
1024do_setopt_accept_filter(so, sopt)
1025 struct socket *so;
1026 struct sockopt *sopt;
1027{
1028 struct accept_filter_arg *afap = NULL;
1029 struct accept_filter *afp;
1030 struct so_accf *af = so->so_accf;
1031 int error = 0;
1032
1033 /* do not set/remove accept filters on non listen sockets */
1034 if ((so->so_options & SO_ACCEPTCONN) == 0) {
1035 error = EINVAL;
1036 goto out;
1037 }
1038
1039 /* removing the filter */
1040 if (sopt == NULL) {
1041 if (af != NULL) {
1042 if (af->so_accept_filter != NULL &&
1043 af->so_accept_filter->accf_destroy != NULL) {
1044 af->so_accept_filter->accf_destroy(so);
1045 }
1046 if (af->so_accept_filter_str != NULL) {
1047 FREE(af->so_accept_filter_str, M_ACCF);
1048 }
1049 FREE(af, M_ACCF);
1050 so->so_accf = NULL;
1051 }
1052 so->so_options &= ~SO_ACCEPTFILTER;
1053 return (0);
1054 }
1055 /* adding a filter */
1056 /* must remove previous filter first */
1057 if (af != NULL) {
1058 error = EINVAL;
1059 goto out;
1060 }
1061 /* don't put large objects on the kernel stack */
1062 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK);
1063 error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap);
1064 afap->af_name[sizeof(afap->af_name)-1] = '\0';
1065 afap->af_arg[sizeof(afap->af_arg)-1] = '\0';
1066 if (error)
1067 goto out;
1068 afp = accept_filt_get(afap->af_name);
1069 if (afp == NULL) {
1070 error = ENOENT;
1071 goto out;
1072 }
1073 MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK | M_ZERO);
1074 if (afp->accf_create != NULL) {
1075 if (afap->af_name[0] != '\0') {
1076 int len = strlen(afap->af_name) + 1;
1077
1078 MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK);
1079 strcpy(af->so_accept_filter_str, afap->af_name);
1080 }
1081 af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg);
1082 if (af->so_accept_filter_arg == NULL) {
1083 FREE(af->so_accept_filter_str, M_ACCF);
1084 FREE(af, M_ACCF);
1085 so->so_accf = NULL;
1086 error = EINVAL;
1087 goto out;
1088 }
1089 }
1090 af->so_accept_filter = afp;
1091 so->so_accf = af;
1092 so->so_options |= SO_ACCEPTFILTER;
1093out:
1094 if (afap != NULL)
1095 FREE(afap, M_TEMP);
1096 return (error);
1097}
1098#endif /* INET */
1099
1100/*
1101 * Perhaps this routine, and sooptcopyout(), below, ought to come in
1102 * an additional variant to handle the case where the option value needs
1103 * to be some kind of integer, but not a specific size.
1104 * In addition to their use here, these functions are also called by the
1105 * protocol-level pr_ctloutput() routines.
1106 */
1107int
1108sooptcopyin(sopt, buf, len, minlen)
1109 struct sockopt *sopt;
1110 void *buf;
1111 size_t len;
1112 size_t minlen;
1113{
1114 size_t valsize;
1115
1116 /*
1117 * If the user gives us more than we wanted, we ignore it,
1118 * but if we don't get the minimum length the caller
1119 * wants, we return EINVAL. On success, sopt->sopt_valsize
1120 * is set to however much we actually retrieved.
1121 */
1122 if ((valsize = sopt->sopt_valsize) < minlen)
1123 return EINVAL;
1124 if (valsize > len)
1125 sopt->sopt_valsize = valsize = len;
1126
1127 if (sopt->sopt_td != 0)
1128 return (copyin(sopt->sopt_val, buf, valsize));
1129
1130 bcopy(sopt->sopt_val, buf, valsize);
1131 return 0;
1132}
1133
1134int
1135sosetopt(so, sopt)
1136 struct socket *so;
1137 struct sockopt *sopt;
1138{
1139 int error, optval;
1140 struct linger l;
1141 struct timeval tv;
1142 u_long val;
1143
1144 error = 0;
1145 if (sopt->sopt_level != SOL_SOCKET) {
1146 if (so->so_proto && so->so_proto->pr_ctloutput)
1147 return ((*so->so_proto->pr_ctloutput)
1148 (so, sopt));
1149 error = ENOPROTOOPT;
1150 } else {
1151 switch (sopt->sopt_name) {
1152#ifdef INET
1153 case SO_ACCEPTFILTER:
1154 error = do_setopt_accept_filter(so, sopt);
1155 if (error)
1156 goto bad;
1157 break;
1158#endif
1159 case SO_LINGER:
1160 error = sooptcopyin(sopt, &l, sizeof l, sizeof l);
1161 if (error)
1162 goto bad;
1163
1164 so->so_linger = l.l_linger;
1165 if (l.l_onoff)
1166 so->so_options |= SO_LINGER;
1167 else
1168 so->so_options &= ~SO_LINGER;
1169 break;
1170
1171 case SO_DEBUG:
1172 case SO_KEEPALIVE:
1173 case SO_DONTROUTE:
1174 case SO_USELOOPBACK:
1175 case SO_BROADCAST:
1176 case SO_REUSEADDR:
1177 case SO_REUSEPORT:
1178 case SO_OOBINLINE:
1179 case SO_TIMESTAMP:
1180 error = sooptcopyin(sopt, &optval, sizeof optval,
1181 sizeof optval);
1182 if (error)
1183 goto bad;
1184 if (optval)
1185 so->so_options |= sopt->sopt_name;
1186 else
1187 so->so_options &= ~sopt->sopt_name;
1188 break;
1189
1190 case SO_SNDBUF:
1191 case SO_RCVBUF:
1192 case SO_SNDLOWAT:
1193 case SO_RCVLOWAT:
1194 error = sooptcopyin(sopt, &optval, sizeof optval,
1195 sizeof optval);
1196 if (error)
1197 goto bad;
1198
1199 /*
1200 * Values < 1 make no sense for any of these
1201 * options, so disallow them.
1202 */
1203 if (optval < 1) {
1204 error = EINVAL;
1205 goto bad;
1206 }
1207
1208 switch (sopt->sopt_name) {
1209 case SO_SNDBUF:
1210 case SO_RCVBUF:
1211 if (sbreserve(sopt->sopt_name == SO_SNDBUF ?
1212 &so->so_snd : &so->so_rcv, (u_long)optval,
1213 so, curthread) == 0) {
1214 error = ENOBUFS;
1215 goto bad;
1216 }
1217 break;
1218
1219 /*
1220 * Make sure the low-water is never greater than
1221 * the high-water.
1222 */
1223 case SO_SNDLOWAT:
1224 so->so_snd.sb_lowat =
1225 (optval > so->so_snd.sb_hiwat) ?
1226 so->so_snd.sb_hiwat : optval;
1227 break;
1228 case SO_RCVLOWAT:
1229 so->so_rcv.sb_lowat =
1230 (optval > so->so_rcv.sb_hiwat) ?
1231 so->so_rcv.sb_hiwat : optval;
1232 break;
1233 }
1234 break;
1235
1236 case SO_SNDTIMEO:
1237 case SO_RCVTIMEO:
1238 error = sooptcopyin(sopt, &tv, sizeof tv,
1239 sizeof tv);
1240 if (error)
1241 goto bad;
1242
1243 /* assert(hz > 0); */
1244 if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz ||
1245 tv.tv_usec < 0 || tv.tv_usec >= 1000000) {
1246 error = EDOM;
1247 goto bad;
1248 }
1249 /* assert(tick > 0); */
1250 /* assert(ULONG_MAX - SHRT_MAX >= 1000000); */
1251 val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick;
1252 if (val > SHRT_MAX) {
1253 error = EDOM;
1254 goto bad;
1255 }
1256
1257 switch (sopt->sopt_name) {
1258 case SO_SNDTIMEO:
1259 so->so_snd.sb_timeo = val;
1260 break;
1261 case SO_RCVTIMEO:
1262 so->so_rcv.sb_timeo = val;
1263 break;
1264 }
1265 break;
1266 default:
1267 error = ENOPROTOOPT;
1268 break;
1269 }
1270 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1271 (void) ((*so->so_proto->pr_ctloutput)
1272 (so, sopt));
1273 }
1274 }
1275bad:
1276 return (error);
1277}
1278
1279/* Helper routine for getsockopt */
1280int
1281sooptcopyout(sopt, buf, len)
1282 struct sockopt *sopt;
1283 void *buf;
1284 size_t len;
1285{
1286 int error;
1287 size_t valsize;
1288
1289 error = 0;
1290
1291 /*
1292 * Documented get behavior is that we always return a value,
1293 * possibly truncated to fit in the user's buffer.
1294 * Traditional behavior is that we always tell the user
1295 * precisely how much we copied, rather than something useful
1296 * like the total amount we had available for her.
1297 * Note that this interface is not idempotent; the entire answer must
1298 * generated ahead of time.
1299 */
1300 valsize = min(len, sopt->sopt_valsize);
1301 sopt->sopt_valsize = valsize;
1302 if (sopt->sopt_val != 0) {
1303 if (sopt->sopt_td != 0)
1304 error = copyout(buf, sopt->sopt_val, valsize);
1305 else
1306 bcopy(buf, sopt->sopt_val, valsize);
1307 }
1308 return error;
1309}
1310
1311int
1312sogetopt(so, sopt)
1313 struct socket *so;
1314 struct sockopt *sopt;
1315{
1316 int error, optval;
1317 struct linger l;
1318 struct timeval tv;
1319#ifdef INET
1320 struct accept_filter_arg *afap;
1321#endif
1322
1323 error = 0;
1324 if (sopt->sopt_level != SOL_SOCKET) {
1325 if (so->so_proto && so->so_proto->pr_ctloutput) {
1326 return ((*so->so_proto->pr_ctloutput)
1327 (so, sopt));
1328 } else
1329 return (ENOPROTOOPT);
1330 } else {
1331 switch (sopt->sopt_name) {
1332#ifdef INET
1333 case SO_ACCEPTFILTER:
1334 if ((so->so_options & SO_ACCEPTCONN) == 0)
1335 return (EINVAL);
1336 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap),
1337 M_TEMP, M_WAITOK | M_ZERO);
1338 if ((so->so_options & SO_ACCEPTFILTER) != 0) {
1339 strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name);
1340 if (so->so_accf->so_accept_filter_str != NULL)
1341 strcpy(afap->af_arg, so->so_accf->so_accept_filter_str);
1342 }
1343 error = sooptcopyout(sopt, afap, sizeof(*afap));
1344 FREE(afap, M_TEMP);
1345 break;
1346#endif
1347
1348 case SO_LINGER:
1349 l.l_onoff = so->so_options & SO_LINGER;
1350 l.l_linger = so->so_linger;
1351 error = sooptcopyout(sopt, &l, sizeof l);
1352 break;
1353
1354 case SO_USELOOPBACK:
1355 case SO_DONTROUTE:
1356 case SO_DEBUG:
1357 case SO_KEEPALIVE:
1358 case SO_REUSEADDR:
1359 case SO_REUSEPORT:
1360 case SO_BROADCAST:
1361 case SO_OOBINLINE:
1362 case SO_TIMESTAMP:
1363 optval = so->so_options & sopt->sopt_name;
1364integer:
1365 error = sooptcopyout(sopt, &optval, sizeof optval);
1366 break;
1367
1368 case SO_TYPE:
1369 optval = so->so_type;
1370 goto integer;
1371
1372 case SO_ERROR:
1373 optval = so->so_error;
1374 so->so_error = 0;
1375 goto integer;
1376
1377 case SO_SNDBUF:
1378 optval = so->so_snd.sb_hiwat;
1379 goto integer;
1380
1381 case SO_RCVBUF:
1382 optval = so->so_rcv.sb_hiwat;
1383 goto integer;
1384
1385 case SO_SNDLOWAT:
1386 optval = so->so_snd.sb_lowat;
1387 goto integer;
1388
1389 case SO_RCVLOWAT:
1390 optval = so->so_rcv.sb_lowat;
1391 goto integer;
1392
1393 case SO_SNDTIMEO:
1394 case SO_RCVTIMEO:
1395 optval = (sopt->sopt_name == SO_SNDTIMEO ?
1396 so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1397
1398 tv.tv_sec = optval / hz;
1399 tv.tv_usec = (optval % hz) * tick;
1400 error = sooptcopyout(sopt, &tv, sizeof tv);
1401 break;
1402
1403 default:
1404 error = ENOPROTOOPT;
1405 break;
1406 }
1407 return (error);
1408 }
1409}
1410
1411/* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */
1412int
1413soopt_getm(struct sockopt *sopt, struct mbuf **mp)
1414{
1415 struct mbuf *m, *m_prev;
1416 int sopt_size = sopt->sopt_valsize;
1417
1418 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA);
1419 if (m == 0)
1420 return ENOBUFS;
1421 if (sopt_size > MLEN) {
1422 MCLGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT);
1423 if ((m->m_flags & M_EXT) == 0) {
1424 m_free(m);
1425 return ENOBUFS;
1426 }
1427 m->m_len = min(MCLBYTES, sopt_size);
1428 } else {
1429 m->m_len = min(MLEN, sopt_size);
1430 }
1431 sopt_size -= m->m_len;
1432 *mp = m;
1433 m_prev = m;
1434
1435 while (sopt_size) {
1436 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA);
1437 if (m == 0) {
1438 m_freem(*mp);
1439 return ENOBUFS;
1440 }
1441 if (sopt_size > MLEN) {
1442 MCLGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT);
1443 if ((m->m_flags & M_EXT) == 0) {
1444 m_freem(*mp);
1445 return ENOBUFS;
1446 }
1447 m->m_len = min(MCLBYTES, sopt_size);
1448 } else {
1449 m->m_len = min(MLEN, sopt_size);
1450 }
1451 sopt_size -= m->m_len;
1452 m_prev->m_next = m;
1453 m_prev = m;
1454 }
1455 return 0;
1456}
1457
1458/* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
1459int
1460soopt_mcopyin(struct sockopt *sopt, struct mbuf *m)
1461{
1462 struct mbuf *m0 = m;
1463
1464 if (sopt->sopt_val == NULL)
1465 return 0;
1466 while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1467 if (sopt->sopt_td != NULL) {
1468 int error;
1469
1470 error = copyin(sopt->sopt_val, mtod(m, char *),
1471 m->m_len);
1472 if (error != 0) {
1473 m_freem(m0);
1474 return(error);
1475 }
1476 } else
1477 bcopy(sopt->sopt_val, mtod(m, char *), m->m_len);
1478 sopt->sopt_valsize -= m->m_len;
1479 (caddr_t)sopt->sopt_val += m->m_len;
1480 m = m->m_next;
1481 }
1482 if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */
1483 panic("ip6_sooptmcopyin");
1484 return 0;
1485}
1486
1487/* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
1488int
1489soopt_mcopyout(struct sockopt *sopt, struct mbuf *m)
1490{
1491 struct mbuf *m0 = m;
1492 size_t valsize = 0;
1493
1494 if (sopt->sopt_val == NULL)
1495 return 0;
1496 while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1497 if (sopt->sopt_td != NULL) {
1498 int error;
1499
1500 error = copyout(mtod(m, char *), sopt->sopt_val,
1501 m->m_len);
1502 if (error != 0) {
1503 m_freem(m0);
1504 return(error);
1505 }
1506 } else
1507 bcopy(mtod(m, char *), sopt->sopt_val, m->m_len);
1508 sopt->sopt_valsize -= m->m_len;
1509 (caddr_t)sopt->sopt_val += m->m_len;
1510 valsize += m->m_len;
1511 m = m->m_next;
1512 }
1513 if (m != NULL) {
1514 /* enough soopt buffer should be given from user-land */
1515 m_freem(m0);
1516 return(EINVAL);
1517 }
1518 sopt->sopt_valsize = valsize;
1519 return 0;
1520}
1521
1522void
1523sohasoutofband(so)
1524 register struct socket *so;
1525{
1526 if (so->so_sigio != NULL)
1527 pgsigio(so->so_sigio, SIGURG, 0);
1528 selwakeup(&so->so_rcv.sb_sel);
1529}
1530
1531int
1532sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td)
1533{
1534 int revents = 0;
1535 int s = splnet();
1536
1537 if (events & (POLLIN | POLLRDNORM))
1538 if (soreadable(so))
1539 revents |= events & (POLLIN | POLLRDNORM);
1540
1541 if (events & POLLINIGNEOF)
1542 if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat ||
1543 !TAILQ_EMPTY(&so->so_comp) || so->so_error)
1544 revents |= POLLINIGNEOF;
1545
1546 if (events & (POLLOUT | POLLWRNORM))
1547 if (sowriteable(so))
1548 revents |= events & (POLLOUT | POLLWRNORM);
1549
1550 if (events & (POLLPRI | POLLRDBAND))
1551 if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
1552 revents |= events & (POLLPRI | POLLRDBAND);
1553
1554 if (revents == 0) {
1555 if (events &
1556 (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM |
1557 POLLRDBAND)) {
1558 selrecord(td, &so->so_rcv.sb_sel);
1559 so->so_rcv.sb_flags |= SB_SEL;
1560 }
1561
1562 if (events & (POLLOUT | POLLWRNORM)) {
1563 selrecord(td, &so->so_snd.sb_sel);
1564 so->so_snd.sb_flags |= SB_SEL;
1565 }
1566 }
1567
1568 splx(s);
1569 return (revents);
1570}
1571
1572int
1573sokqfilter(struct file *fp, struct knote *kn)
1574{
1575 struct socket *so = (struct socket *)kn->kn_fp->f_data;
1576 struct sockbuf *sb;
1577 int s;
1578
1579 switch (kn->kn_filter) {
1580 case EVFILT_READ:
1581 if (so->so_options & SO_ACCEPTCONN)
1582 kn->kn_fop = &solisten_filtops;
1583 else
1584 kn->kn_fop = &soread_filtops;
1585 sb = &so->so_rcv;
1586 break;
1587 case EVFILT_WRITE:
1588 kn->kn_fop = &sowrite_filtops;
1589 sb = &so->so_snd;
1590 break;
1591 default:
1592 return (1);
1593 }
1594
1595 s = splnet();
1596 SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext);
1597 sb->sb_flags |= SB_KNOTE;
1598 splx(s);
1599 return (0);
1600}
1601
1602static void
1603filt_sordetach(struct knote *kn)
1604{
1605 struct socket *so = (struct socket *)kn->kn_fp->f_data;
1606 int s = splnet();
1607
1608 SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext);
1609 if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note))
1610 so->so_rcv.sb_flags &= ~SB_KNOTE;
1611 splx(s);
1612}
1613
1614/*ARGSUSED*/
1615static int
1616filt_soread(struct knote *kn, long hint)
1617{
1618 struct socket *so = (struct socket *)kn->kn_fp->f_data;
1619
1620 kn->kn_data = so->so_rcv.sb_cc;
1621 if (so->so_state & SS_CANTRCVMORE) {
1622 kn->kn_flags |= EV_EOF;
1623 kn->kn_fflags = so->so_error;
1624 return (1);
1625 }
1626 if (so->so_error) /* temporary udp error */
1627 return (1);
1628 if (kn->kn_sfflags & NOTE_LOWAT)
1629 return (kn->kn_data >= kn->kn_sdata);
1630 return (kn->kn_data >= so->so_rcv.sb_lowat);
1631}
1632
1633static void
1634filt_sowdetach(struct knote *kn)
1635{
1636 struct socket *so = (struct socket *)kn->kn_fp->f_data;
1637 int s = splnet();
1638
1639 SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext);
1640 if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note))
1641 so->so_snd.sb_flags &= ~SB_KNOTE;
1642 splx(s);
1643}
1644
1645/*ARGSUSED*/
1646static int
1647filt_sowrite(struct knote *kn, long hint)
1648{
1649 struct socket *so = (struct socket *)kn->kn_fp->f_data;
1650
1651 kn->kn_data = sbspace(&so->so_snd);
1652 if (so->so_state & SS_CANTSENDMORE) {
1653 kn->kn_flags |= EV_EOF;
1654 kn->kn_fflags = so->so_error;
1655 return (1);
1656 }
1657 if (so->so_error) /* temporary udp error */
1658 return (1);
1659 if (((so->so_state & SS_ISCONNECTED) == 0) &&
1660 (so->so_proto->pr_flags & PR_CONNREQUIRED))
1661 return (0);
1662 if (kn->kn_sfflags & NOTE_LOWAT)
1663 return (kn->kn_data >= kn->kn_sdata);
1664 return (kn->kn_data >= so->so_snd.sb_lowat);
1665}
1666
1667/*ARGSUSED*/
1668static int
1669filt_solisten(struct knote *kn, long hint)
1670{
1671 struct socket *so = (struct socket *)kn->kn_fp->f_data;
1672
1673 kn->kn_data = so->so_qlen - so->so_incqlen;
1674 return (! TAILQ_EMPTY(&so->so_comp));
1675}
1676
1677int
1678socheckuid(struct socket *so, uid_t uid)
1679{
1680
1681 if (so == NULL)
1682 return (EPERM);
1683 if (so->so_cred->cr_uid == uid)
1684 return (0);
1685 return (EPERM);
1686}
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_socket.c 8.3 (Berkeley) 4/15/94
34 * $FreeBSD: head/sys/kern/uipc_socket.c 92654 2002-03-19 09:11:49Z jeff $
35 */
36
37#include "opt_inet.h"
38
39#include <sys/param.h>
40#include <sys/systm.h>
41#include <sys/fcntl.h>
42#include <sys/lock.h>
43#include <sys/malloc.h>
44#include <sys/mbuf.h>
45#include <sys/mutex.h>
46#include <sys/domain.h>
47#include <sys/file.h> /* for struct knote */
48#include <sys/kernel.h>
49#include <sys/malloc.h>
50#include <sys/event.h>
51#include <sys/poll.h>
52#include <sys/proc.h>
53#include <sys/protosw.h>
54#include <sys/socket.h>
55#include <sys/socketvar.h>
56#include <sys/resourcevar.h>
57#include <sys/signalvar.h>
58#include <sys/sysctl.h>
59#include <sys/uio.h>
60#include <sys/jail.h>
61
62#include <vm/vm_zone.h>
63
64#include <machine/limits.h>
65
66#ifdef INET
67static int do_setopt_accept_filter(struct socket *so, struct sockopt *sopt);
68#endif
69
70static void filt_sordetach(struct knote *kn);
71static int filt_soread(struct knote *kn, long hint);
72static void filt_sowdetach(struct knote *kn);
73static int filt_sowrite(struct knote *kn, long hint);
74static int filt_solisten(struct knote *kn, long hint);
75
76static struct filterops solisten_filtops =
77 { 1, NULL, filt_sordetach, filt_solisten };
78static struct filterops soread_filtops =
79 { 1, NULL, filt_sordetach, filt_soread };
80static struct filterops sowrite_filtops =
81 { 1, NULL, filt_sowdetach, filt_sowrite };
82
83vm_zone_t socket_zone;
84so_gen_t so_gencnt; /* generation count for sockets */
85
86MALLOC_DEFINE(M_SONAME, "soname", "socket name");
87MALLOC_DEFINE(M_PCB, "pcb", "protocol control block");
88
89SYSCTL_DECL(_kern_ipc);
90
91static int somaxconn = SOMAXCONN;
92SYSCTL_INT(_kern_ipc, KIPC_SOMAXCONN, somaxconn, CTLFLAG_RW,
93 &somaxconn, 0, "Maximum pending socket connection queue size");
94static int numopensockets;
95SYSCTL_INT(_kern_ipc, OID_AUTO, numopensockets, CTLFLAG_RD,
96 &numopensockets, 0, "Number of open sockets");
97
98
99/*
100 * Socket operation routines.
101 * These routines are called by the routines in
102 * sys_socket.c or from a system process, and
103 * implement the semantics of socket operations by
104 * switching out to the protocol specific routines.
105 */
106
107/*
108 * Get a socket structure from our zone, and initialize it.
109 * We don't implement `waitok' yet (see comments in uipc_domain.c).
110 * Note that it would probably be better to allocate socket
111 * and PCB at the same time, but I'm not convinced that all
112 * the protocols can be easily modified to do this.
113 *
114 * soalloc() returns a socket with a ref count of 0.
115 */
116struct socket *
117soalloc(waitok)
118 int waitok;
119{
120 struct socket *so;
121
122 so = zalloc(socket_zone);
123 if (so) {
124 /* XXX race condition for reentrant kernel */
125 bzero(so, sizeof *so);
126 so->so_gencnt = ++so_gencnt;
127 so->so_zone = socket_zone;
128 /* sx_init(&so->so_sxlock, "socket sxlock"); */
129 TAILQ_INIT(&so->so_aiojobq);
130 ++numopensockets;
131 }
132 return so;
133}
134
135/*
136 * socreate returns a socket with a ref count of 1. The socket should be
137 * closed with soclose().
138 */
139int
140socreate(dom, aso, type, proto, cred, td)
141 int dom;
142 struct socket **aso;
143 register int type;
144 int proto;
145 struct ucred *cred;
146 struct thread *td;
147{
148 register struct protosw *prp;
149 register struct socket *so;
150 register int error;
151
152 if (proto)
153 prp = pffindproto(dom, proto, type);
154 else
155 prp = pffindtype(dom, type);
156
157 if (prp == 0 || prp->pr_usrreqs->pru_attach == 0)
158 return (EPROTONOSUPPORT);
159
160 if (jailed(td->td_ucred) && jail_socket_unixiproute_only &&
161 prp->pr_domain->dom_family != PF_LOCAL &&
162 prp->pr_domain->dom_family != PF_INET &&
163 prp->pr_domain->dom_family != PF_ROUTE) {
164 return (EPROTONOSUPPORT);
165 }
166
167 if (prp->pr_type != type)
168 return (EPROTOTYPE);
169 so = soalloc(td != 0);
170 if (so == 0)
171 return (ENOBUFS);
172
173 TAILQ_INIT(&so->so_incomp);
174 TAILQ_INIT(&so->so_comp);
175 so->so_type = type;
176 so->so_cred = crhold(cred);
177 so->so_proto = prp;
178 soref(so);
179 error = (*prp->pr_usrreqs->pru_attach)(so, proto, td);
180 if (error) {
181 so->so_state |= SS_NOFDREF;
182 sorele(so);
183 return (error);
184 }
185 *aso = so;
186 return (0);
187}
188
189int
190sobind(so, nam, td)
191 struct socket *so;
192 struct sockaddr *nam;
193 struct thread *td;
194{
195 int s = splnet();
196 int error;
197
198 error = (*so->so_proto->pr_usrreqs->pru_bind)(so, nam, td);
199 splx(s);
200 return (error);
201}
202
203static void
204sodealloc(struct socket *so)
205{
206
207 KASSERT(so->so_count == 0, ("sodealloc(): so_count %d", so->so_count));
208 so->so_gencnt = ++so_gencnt;
209 if (so->so_rcv.sb_hiwat)
210 (void)chgsbsize(so->so_cred->cr_uidinfo,
211 &so->so_rcv.sb_hiwat, 0, RLIM_INFINITY);
212 if (so->so_snd.sb_hiwat)
213 (void)chgsbsize(so->so_cred->cr_uidinfo,
214 &so->so_snd.sb_hiwat, 0, RLIM_INFINITY);
215#ifdef INET
216 if (so->so_accf != NULL) {
217 if (so->so_accf->so_accept_filter != NULL &&
218 so->so_accf->so_accept_filter->accf_destroy != NULL) {
219 so->so_accf->so_accept_filter->accf_destroy(so);
220 }
221 if (so->so_accf->so_accept_filter_str != NULL)
222 FREE(so->so_accf->so_accept_filter_str, M_ACCF);
223 FREE(so->so_accf, M_ACCF);
224 }
225#endif
226 crfree(so->so_cred);
227 /* sx_destroy(&so->so_sxlock); */
228 zfree(so->so_zone, so);
229 --numopensockets;
230}
231
232int
233solisten(so, backlog, td)
234 register struct socket *so;
235 int backlog;
236 struct thread *td;
237{
238 int s, error;
239
240 s = splnet();
241 error = (*so->so_proto->pr_usrreqs->pru_listen)(so, td);
242 if (error) {
243 splx(s);
244 return (error);
245 }
246 if (TAILQ_EMPTY(&so->so_comp))
247 so->so_options |= SO_ACCEPTCONN;
248 if (backlog < 0 || backlog > somaxconn)
249 backlog = somaxconn;
250 so->so_qlimit = backlog;
251 splx(s);
252 return (0);
253}
254
255void
256sofree(so)
257 register struct socket *so;
258{
259 struct socket *head = so->so_head;
260
261 KASSERT(so->so_count == 0, ("socket %p so_count not 0", so));
262
263 if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
264 return;
265 if (head != NULL) {
266 if (so->so_state & SS_INCOMP) {
267 TAILQ_REMOVE(&head->so_incomp, so, so_list);
268 head->so_incqlen--;
269 } else if (so->so_state & SS_COMP) {
270 /*
271 * We must not decommission a socket that's
272 * on the accept(2) queue. If we do, then
273 * accept(2) may hang after select(2) indicated
274 * that the listening socket was ready.
275 */
276 return;
277 } else {
278 panic("sofree: not queued");
279 }
280 head->so_qlen--;
281 so->so_state &= ~SS_INCOMP;
282 so->so_head = NULL;
283 }
284 sbrelease(&so->so_snd, so);
285 sorflush(so);
286 sodealloc(so);
287}
288
289/*
290 * Close a socket on last file table reference removal.
291 * Initiate disconnect if connected.
292 * Free socket when disconnect complete.
293 *
294 * This function will sorele() the socket. Note that soclose() may be
295 * called prior to the ref count reaching zero. The actual socket
296 * structure will not be freed until the ref count reaches zero.
297 */
298int
299soclose(so)
300 register struct socket *so;
301{
302 int s = splnet(); /* conservative */
303 int error = 0;
304
305 funsetown(so->so_sigio);
306 if (so->so_options & SO_ACCEPTCONN) {
307 struct socket *sp, *sonext;
308
309 sp = TAILQ_FIRST(&so->so_incomp);
310 for (; sp != NULL; sp = sonext) {
311 sonext = TAILQ_NEXT(sp, so_list);
312 (void) soabort(sp);
313 }
314 for (sp = TAILQ_FIRST(&so->so_comp); sp != NULL; sp = sonext) {
315 sonext = TAILQ_NEXT(sp, so_list);
316 /* Dequeue from so_comp since sofree() won't do it */
317 TAILQ_REMOVE(&so->so_comp, sp, so_list);
318 so->so_qlen--;
319 sp->so_state &= ~SS_COMP;
320 sp->so_head = NULL;
321 (void) soabort(sp);
322 }
323 }
324 if (so->so_pcb == 0)
325 goto discard;
326 if (so->so_state & SS_ISCONNECTED) {
327 if ((so->so_state & SS_ISDISCONNECTING) == 0) {
328 error = sodisconnect(so);
329 if (error)
330 goto drop;
331 }
332 if (so->so_options & SO_LINGER) {
333 if ((so->so_state & SS_ISDISCONNECTING) &&
334 (so->so_state & SS_NBIO))
335 goto drop;
336 while (so->so_state & SS_ISCONNECTED) {
337 error = tsleep((caddr_t)&so->so_timeo,
338 PSOCK | PCATCH, "soclos", so->so_linger * hz);
339 if (error)
340 break;
341 }
342 }
343 }
344drop:
345 if (so->so_pcb) {
346 int error2 = (*so->so_proto->pr_usrreqs->pru_detach)(so);
347 if (error == 0)
348 error = error2;
349 }
350discard:
351 if (so->so_state & SS_NOFDREF)
352 panic("soclose: NOFDREF");
353 so->so_state |= SS_NOFDREF;
354 sorele(so);
355 splx(s);
356 return (error);
357}
358
359/*
360 * Must be called at splnet...
361 */
362int
363soabort(so)
364 struct socket *so;
365{
366 int error;
367
368 error = (*so->so_proto->pr_usrreqs->pru_abort)(so);
369 if (error) {
370 sotryfree(so); /* note: does not decrement the ref count */
371 return error;
372 }
373 return (0);
374}
375
376int
377soaccept(so, nam)
378 register struct socket *so;
379 struct sockaddr **nam;
380{
381 int s = splnet();
382 int error;
383
384 if ((so->so_state & SS_NOFDREF) == 0)
385 panic("soaccept: !NOFDREF");
386 so->so_state &= ~SS_NOFDREF;
387 error = (*so->so_proto->pr_usrreqs->pru_accept)(so, nam);
388 splx(s);
389 return (error);
390}
391
392int
393soconnect(so, nam, td)
394 register struct socket *so;
395 struct sockaddr *nam;
396 struct thread *td;
397{
398 int s;
399 int error;
400
401 if (so->so_options & SO_ACCEPTCONN)
402 return (EOPNOTSUPP);
403 s = splnet();
404 /*
405 * If protocol is connection-based, can only connect once.
406 * Otherwise, if connected, try to disconnect first.
407 * This allows user to disconnect by connecting to, e.g.,
408 * a null address.
409 */
410 if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
411 ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
412 (error = sodisconnect(so))))
413 error = EISCONN;
414 else
415 error = (*so->so_proto->pr_usrreqs->pru_connect)(so, nam, td);
416 splx(s);
417 return (error);
418}
419
420int
421soconnect2(so1, so2)
422 register struct socket *so1;
423 struct socket *so2;
424{
425 int s = splnet();
426 int error;
427
428 error = (*so1->so_proto->pr_usrreqs->pru_connect2)(so1, so2);
429 splx(s);
430 return (error);
431}
432
433int
434sodisconnect(so)
435 register struct socket *so;
436{
437 int s = splnet();
438 int error;
439
440 if ((so->so_state & SS_ISCONNECTED) == 0) {
441 error = ENOTCONN;
442 goto bad;
443 }
444 if (so->so_state & SS_ISDISCONNECTING) {
445 error = EALREADY;
446 goto bad;
447 }
448 error = (*so->so_proto->pr_usrreqs->pru_disconnect)(so);
449bad:
450 splx(s);
451 return (error);
452}
453
454#define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
455/*
456 * Send on a socket.
457 * If send must go all at once and message is larger than
458 * send buffering, then hard error.
459 * Lock against other senders.
460 * If must go all at once and not enough room now, then
461 * inform user that this would block and do nothing.
462 * Otherwise, if nonblocking, send as much as possible.
463 * The data to be sent is described by "uio" if nonzero,
464 * otherwise by the mbuf chain "top" (which must be null
465 * if uio is not). Data provided in mbuf chain must be small
466 * enough to send all at once.
467 *
468 * Returns nonzero on error, timeout or signal; callers
469 * must check for short counts if EINTR/ERESTART are returned.
470 * Data and control buffers are freed on return.
471 */
472int
473sosend(so, addr, uio, top, control, flags, td)
474 register struct socket *so;
475 struct sockaddr *addr;
476 struct uio *uio;
477 struct mbuf *top;
478 struct mbuf *control;
479 int flags;
480 struct thread *td;
481{
482 struct mbuf **mp;
483 register struct mbuf *m;
484 register long space, len, resid;
485 int clen = 0, error, s, dontroute, mlen;
486 int atomic = sosendallatonce(so) || top;
487
488 if (uio)
489 resid = uio->uio_resid;
490 else
491 resid = top->m_pkthdr.len;
492 /*
493 * In theory resid should be unsigned.
494 * However, space must be signed, as it might be less than 0
495 * if we over-committed, and we must use a signed comparison
496 * of space and resid. On the other hand, a negative resid
497 * causes us to loop sending 0-length segments to the protocol.
498 *
499 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
500 * type sockets since that's an error.
501 */
502 if (resid < 0 || (so->so_type == SOCK_STREAM && (flags & MSG_EOR))) {
503 error = EINVAL;
504 goto out;
505 }
506
507 dontroute =
508 (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
509 (so->so_proto->pr_flags & PR_ATOMIC);
510 if (td)
511 td->td_proc->p_stats->p_ru.ru_msgsnd++;
512 if (control)
513 clen = control->m_len;
514#define snderr(errno) { error = errno; splx(s); goto release; }
515
516restart:
517 error = sblock(&so->so_snd, SBLOCKWAIT(flags));
518 if (error)
519 goto out;
520 do {
521 s = splnet();
522 if (so->so_state & SS_CANTSENDMORE)
523 snderr(EPIPE);
524 if (so->so_error) {
525 error = so->so_error;
526 so->so_error = 0;
527 splx(s);
528 goto release;
529 }
530 if ((so->so_state & SS_ISCONNECTED) == 0) {
531 /*
532 * `sendto' and `sendmsg' is allowed on a connection-
533 * based socket if it supports implied connect.
534 * Return ENOTCONN if not connected and no address is
535 * supplied.
536 */
537 if ((so->so_proto->pr_flags & PR_CONNREQUIRED) &&
538 (so->so_proto->pr_flags & PR_IMPLOPCL) == 0) {
539 if ((so->so_state & SS_ISCONFIRMING) == 0 &&
540 !(resid == 0 && clen != 0))
541 snderr(ENOTCONN);
542 } else if (addr == 0)
543 snderr(so->so_proto->pr_flags & PR_CONNREQUIRED ?
544 ENOTCONN : EDESTADDRREQ);
545 }
546 space = sbspace(&so->so_snd);
547 if (flags & MSG_OOB)
548 space += 1024;
549 if ((atomic && resid > so->so_snd.sb_hiwat) ||
550 clen > so->so_snd.sb_hiwat)
551 snderr(EMSGSIZE);
552 if (space < resid + clen &&
553 (atomic || space < so->so_snd.sb_lowat || space < clen)) {
554 if (so->so_state & SS_NBIO)
555 snderr(EWOULDBLOCK);
556 sbunlock(&so->so_snd);
557 error = sbwait(&so->so_snd);
558 splx(s);
559 if (error)
560 goto out;
561 goto restart;
562 }
563 splx(s);
564 mp = ⊤
565 space -= clen;
566 do {
567 if (uio == NULL) {
568 /*
569 * Data is prepackaged in "top".
570 */
571 resid = 0;
572 if (flags & MSG_EOR)
573 top->m_flags |= M_EOR;
574 } else do {
575 if (top == 0) {
576 MGETHDR(m, M_TRYWAIT, MT_DATA);
577 if (m == NULL) {
578 error = ENOBUFS;
579 goto release;
580 }
581 mlen = MHLEN;
582 m->m_pkthdr.len = 0;
583 m->m_pkthdr.rcvif = (struct ifnet *)0;
584 } else {
585 MGET(m, M_TRYWAIT, MT_DATA);
586 if (m == NULL) {
587 error = ENOBUFS;
588 goto release;
589 }
590 mlen = MLEN;
591 }
592 if (resid >= MINCLSIZE) {
593 MCLGET(m, M_TRYWAIT);
594 if ((m->m_flags & M_EXT) == 0)
595 goto nopages;
596 mlen = MCLBYTES;
597 len = min(min(mlen, resid), space);
598 } else {
599nopages:
600 len = min(min(mlen, resid), space);
601 /*
602 * For datagram protocols, leave room
603 * for protocol headers in first mbuf.
604 */
605 if (atomic && top == 0 && len < mlen)
606 MH_ALIGN(m, len);
607 }
608 space -= len;
609 error = uiomove(mtod(m, caddr_t), (int)len, uio);
610 resid = uio->uio_resid;
611 m->m_len = len;
612 *mp = m;
613 top->m_pkthdr.len += len;
614 if (error)
615 goto release;
616 mp = &m->m_next;
617 if (resid <= 0) {
618 if (flags & MSG_EOR)
619 top->m_flags |= M_EOR;
620 break;
621 }
622 } while (space > 0 && atomic);
623 if (dontroute)
624 so->so_options |= SO_DONTROUTE;
625 s = splnet(); /* XXX */
626 /*
627 * XXX all the SS_CANTSENDMORE checks previously
628 * done could be out of date. We could have recieved
629 * a reset packet in an interrupt or maybe we slept
630 * while doing page faults in uiomove() etc. We could
631 * probably recheck again inside the splnet() protection
632 * here, but there are probably other places that this
633 * also happens. We must rethink this.
634 */
635 error = (*so->so_proto->pr_usrreqs->pru_send)(so,
636 (flags & MSG_OOB) ? PRUS_OOB :
637 /*
638 * If the user set MSG_EOF, the protocol
639 * understands this flag and nothing left to
640 * send then use PRU_SEND_EOF instead of PRU_SEND.
641 */
642 ((flags & MSG_EOF) &&
643 (so->so_proto->pr_flags & PR_IMPLOPCL) &&
644 (resid <= 0)) ?
645 PRUS_EOF :
646 /* If there is more to send set PRUS_MORETOCOME */
647 (resid > 0 && space > 0) ? PRUS_MORETOCOME : 0,
648 top, addr, control, td);
649 splx(s);
650 if (dontroute)
651 so->so_options &= ~SO_DONTROUTE;
652 clen = 0;
653 control = 0;
654 top = 0;
655 mp = ⊤
656 if (error)
657 goto release;
658 } while (resid && space > 0);
659 } while (resid);
660
661release:
662 sbunlock(&so->so_snd);
663out:
664 if (top)
665 m_freem(top);
666 if (control)
667 m_freem(control);
668 return (error);
669}
670
671/*
672 * Implement receive operations on a socket.
673 * We depend on the way that records are added to the sockbuf
674 * by sbappend*. In particular, each record (mbufs linked through m_next)
675 * must begin with an address if the protocol so specifies,
676 * followed by an optional mbuf or mbufs containing ancillary data,
677 * and then zero or more mbufs of data.
678 * In order to avoid blocking network interrupts for the entire time here,
679 * we splx() while doing the actual copy to user space.
680 * Although the sockbuf is locked, new data may still be appended,
681 * and thus we must maintain consistency of the sockbuf during that time.
682 *
683 * The caller may receive the data as a single mbuf chain by supplying
684 * an mbuf **mp0 for use in returning the chain. The uio is then used
685 * only for the count in uio_resid.
686 */
687int
688soreceive(so, psa, uio, mp0, controlp, flagsp)
689 register struct socket *so;
690 struct sockaddr **psa;
691 struct uio *uio;
692 struct mbuf **mp0;
693 struct mbuf **controlp;
694 int *flagsp;
695{
696 struct mbuf *m, **mp;
697 register int flags, len, error, s, offset;
698 struct protosw *pr = so->so_proto;
699 struct mbuf *nextrecord;
700 int moff, type = 0;
701 int orig_resid = uio->uio_resid;
702
703 mp = mp0;
704 if (psa)
705 *psa = 0;
706 if (controlp)
707 *controlp = 0;
708 if (flagsp)
709 flags = *flagsp &~ MSG_EOR;
710 else
711 flags = 0;
712 if (flags & MSG_OOB) {
713 m = m_get(M_TRYWAIT, MT_DATA);
714 if (m == NULL)
715 return (ENOBUFS);
716 error = (*pr->pr_usrreqs->pru_rcvoob)(so, m, flags & MSG_PEEK);
717 if (error)
718 goto bad;
719 do {
720 error = uiomove(mtod(m, caddr_t),
721 (int) min(uio->uio_resid, m->m_len), uio);
722 m = m_free(m);
723 } while (uio->uio_resid && error == 0 && m);
724bad:
725 if (m)
726 m_freem(m);
727 return (error);
728 }
729 if (mp)
730 *mp = (struct mbuf *)0;
731 if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
732 (*pr->pr_usrreqs->pru_rcvd)(so, 0);
733
734restart:
735 error = sblock(&so->so_rcv, SBLOCKWAIT(flags));
736 if (error)
737 return (error);
738 s = splnet();
739
740 m = so->so_rcv.sb_mb;
741 /*
742 * If we have less data than requested, block awaiting more
743 * (subject to any timeout) if:
744 * 1. the current count is less than the low water mark, or
745 * 2. MSG_WAITALL is set, and it is possible to do the entire
746 * receive operation at once if we block (resid <= hiwat).
747 * 3. MSG_DONTWAIT is not set
748 * If MSG_WAITALL is set but resid is larger than the receive buffer,
749 * we have to do the receive in sections, and thus risk returning
750 * a short count if a timeout or signal occurs after we start.
751 */
752 if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
753 so->so_rcv.sb_cc < uio->uio_resid) &&
754 (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
755 ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
756 m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
757 KASSERT(m != 0 || !so->so_rcv.sb_cc,
758 ("receive: m == %p so->so_rcv.sb_cc == %lu",
759 m, so->so_rcv.sb_cc));
760 if (so->so_error) {
761 if (m)
762 goto dontblock;
763 error = so->so_error;
764 if ((flags & MSG_PEEK) == 0)
765 so->so_error = 0;
766 goto release;
767 }
768 if (so->so_state & SS_CANTRCVMORE) {
769 if (m)
770 goto dontblock;
771 else
772 goto release;
773 }
774 for (; m; m = m->m_next)
775 if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
776 m = so->so_rcv.sb_mb;
777 goto dontblock;
778 }
779 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
780 (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
781 error = ENOTCONN;
782 goto release;
783 }
784 if (uio->uio_resid == 0)
785 goto release;
786 if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
787 error = EWOULDBLOCK;
788 goto release;
789 }
790 sbunlock(&so->so_rcv);
791 error = sbwait(&so->so_rcv);
792 splx(s);
793 if (error)
794 return (error);
795 goto restart;
796 }
797dontblock:
798 if (uio->uio_td)
799 uio->uio_td->td_proc->p_stats->p_ru.ru_msgrcv++;
800 nextrecord = m->m_nextpkt;
801 if (pr->pr_flags & PR_ADDR) {
802 KASSERT(m->m_type == MT_SONAME,
803 ("m->m_type == %d", m->m_type));
804 orig_resid = 0;
805 if (psa)
806 *psa = dup_sockaddr(mtod(m, struct sockaddr *),
807 mp0 == 0);
808 if (flags & MSG_PEEK) {
809 m = m->m_next;
810 } else {
811 sbfree(&so->so_rcv, m);
812 so->so_rcv.sb_mb = m_free(m);
813 m = so->so_rcv.sb_mb;
814 }
815 }
816 while (m && m->m_type == MT_CONTROL && error == 0) {
817 if (flags & MSG_PEEK) {
818 if (controlp)
819 *controlp = m_copy(m, 0, m->m_len);
820 m = m->m_next;
821 } else {
822 sbfree(&so->so_rcv, m);
823 so->so_rcv.sb_mb = m->m_next;
824 m->m_next = NULL;
825 if (pr->pr_domain->dom_externalize)
826 error =
827 (*pr->pr_domain->dom_externalize)(m, controlp);
828 else if (controlp)
829 *controlp = m;
830 else
831 m_freem(m);
832 m = so->so_rcv.sb_mb;
833 }
834 if (controlp) {
835 orig_resid = 0;
836 do
837 controlp = &(*controlp)->m_next;
838 while (*controlp != NULL);
839 }
840 }
841 if (m) {
842 if ((flags & MSG_PEEK) == 0)
843 m->m_nextpkt = nextrecord;
844 type = m->m_type;
845 if (type == MT_OOBDATA)
846 flags |= MSG_OOB;
847 }
848 moff = 0;
849 offset = 0;
850 while (m && uio->uio_resid > 0 && error == 0) {
851 if (m->m_type == MT_OOBDATA) {
852 if (type != MT_OOBDATA)
853 break;
854 } else if (type == MT_OOBDATA)
855 break;
856 else
857 KASSERT(m->m_type == MT_DATA || m->m_type == MT_HEADER,
858 ("m->m_type == %d", m->m_type));
859 so->so_state &= ~SS_RCVATMARK;
860 len = uio->uio_resid;
861 if (so->so_oobmark && len > so->so_oobmark - offset)
862 len = so->so_oobmark - offset;
863 if (len > m->m_len - moff)
864 len = m->m_len - moff;
865 /*
866 * If mp is set, just pass back the mbufs.
867 * Otherwise copy them out via the uio, then free.
868 * Sockbuf must be consistent here (points to current mbuf,
869 * it points to next record) when we drop priority;
870 * we must note any additions to the sockbuf when we
871 * block interrupts again.
872 */
873 if (mp == 0) {
874 splx(s);
875 error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
876 s = splnet();
877 if (error)
878 goto release;
879 } else
880 uio->uio_resid -= len;
881 if (len == m->m_len - moff) {
882 if (m->m_flags & M_EOR)
883 flags |= MSG_EOR;
884 if (flags & MSG_PEEK) {
885 m = m->m_next;
886 moff = 0;
887 } else {
888 nextrecord = m->m_nextpkt;
889 sbfree(&so->so_rcv, m);
890 if (mp) {
891 *mp = m;
892 mp = &m->m_next;
893 so->so_rcv.sb_mb = m = m->m_next;
894 *mp = (struct mbuf *)0;
895 } else {
896 so->so_rcv.sb_mb = m_free(m);
897 m = so->so_rcv.sb_mb;
898 }
899 if (m)
900 m->m_nextpkt = nextrecord;
901 }
902 } else {
903 if (flags & MSG_PEEK)
904 moff += len;
905 else {
906 if (mp)
907 *mp = m_copym(m, 0, len, M_TRYWAIT);
908 m->m_data += len;
909 m->m_len -= len;
910 so->so_rcv.sb_cc -= len;
911 }
912 }
913 if (so->so_oobmark) {
914 if ((flags & MSG_PEEK) == 0) {
915 so->so_oobmark -= len;
916 if (so->so_oobmark == 0) {
917 so->so_state |= SS_RCVATMARK;
918 break;
919 }
920 } else {
921 offset += len;
922 if (offset == so->so_oobmark)
923 break;
924 }
925 }
926 if (flags & MSG_EOR)
927 break;
928 /*
929 * If the MSG_WAITALL flag is set (for non-atomic socket),
930 * we must not quit until "uio->uio_resid == 0" or an error
931 * termination. If a signal/timeout occurs, return
932 * with a short count but without error.
933 * Keep sockbuf locked against other readers.
934 */
935 while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
936 !sosendallatonce(so) && !nextrecord) {
937 if (so->so_error || so->so_state & SS_CANTRCVMORE)
938 break;
939 /*
940 * Notify the protocol that some data has been
941 * drained before blocking.
942 */
943 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
944 (*pr->pr_usrreqs->pru_rcvd)(so, flags);
945 error = sbwait(&so->so_rcv);
946 if (error) {
947 sbunlock(&so->so_rcv);
948 splx(s);
949 return (0);
950 }
951 m = so->so_rcv.sb_mb;
952 if (m)
953 nextrecord = m->m_nextpkt;
954 }
955 }
956
957 if (m && pr->pr_flags & PR_ATOMIC) {
958 flags |= MSG_TRUNC;
959 if ((flags & MSG_PEEK) == 0)
960 (void) sbdroprecord(&so->so_rcv);
961 }
962 if ((flags & MSG_PEEK) == 0) {
963 if (m == 0)
964 so->so_rcv.sb_mb = nextrecord;
965 if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
966 (*pr->pr_usrreqs->pru_rcvd)(so, flags);
967 }
968 if (orig_resid == uio->uio_resid && orig_resid &&
969 (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
970 sbunlock(&so->so_rcv);
971 splx(s);
972 goto restart;
973 }
974
975 if (flagsp)
976 *flagsp |= flags;
977release:
978 sbunlock(&so->so_rcv);
979 splx(s);
980 return (error);
981}
982
983int
984soshutdown(so, how)
985 register struct socket *so;
986 register int how;
987{
988 register struct protosw *pr = so->so_proto;
989
990 if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
991 return (EINVAL);
992
993 if (how != SHUT_WR)
994 sorflush(so);
995 if (how != SHUT_RD)
996 return ((*pr->pr_usrreqs->pru_shutdown)(so));
997 return (0);
998}
999
1000void
1001sorflush(so)
1002 register struct socket *so;
1003{
1004 register struct sockbuf *sb = &so->so_rcv;
1005 register struct protosw *pr = so->so_proto;
1006 register int s;
1007 struct sockbuf asb;
1008
1009 sb->sb_flags |= SB_NOINTR;
1010 (void) sblock(sb, M_WAITOK);
1011 s = splimp();
1012 socantrcvmore(so);
1013 sbunlock(sb);
1014 asb = *sb;
1015 bzero((caddr_t)sb, sizeof (*sb));
1016 splx(s);
1017 if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1018 (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1019 sbrelease(&asb, so);
1020}
1021
1022#ifdef INET
1023static int
1024do_setopt_accept_filter(so, sopt)
1025 struct socket *so;
1026 struct sockopt *sopt;
1027{
1028 struct accept_filter_arg *afap = NULL;
1029 struct accept_filter *afp;
1030 struct so_accf *af = so->so_accf;
1031 int error = 0;
1032
1033 /* do not set/remove accept filters on non listen sockets */
1034 if ((so->so_options & SO_ACCEPTCONN) == 0) {
1035 error = EINVAL;
1036 goto out;
1037 }
1038
1039 /* removing the filter */
1040 if (sopt == NULL) {
1041 if (af != NULL) {
1042 if (af->so_accept_filter != NULL &&
1043 af->so_accept_filter->accf_destroy != NULL) {
1044 af->so_accept_filter->accf_destroy(so);
1045 }
1046 if (af->so_accept_filter_str != NULL) {
1047 FREE(af->so_accept_filter_str, M_ACCF);
1048 }
1049 FREE(af, M_ACCF);
1050 so->so_accf = NULL;
1051 }
1052 so->so_options &= ~SO_ACCEPTFILTER;
1053 return (0);
1054 }
1055 /* adding a filter */
1056 /* must remove previous filter first */
1057 if (af != NULL) {
1058 error = EINVAL;
1059 goto out;
1060 }
1061 /* don't put large objects on the kernel stack */
1062 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap), M_TEMP, M_WAITOK);
1063 error = sooptcopyin(sopt, afap, sizeof *afap, sizeof *afap);
1064 afap->af_name[sizeof(afap->af_name)-1] = '\0';
1065 afap->af_arg[sizeof(afap->af_arg)-1] = '\0';
1066 if (error)
1067 goto out;
1068 afp = accept_filt_get(afap->af_name);
1069 if (afp == NULL) {
1070 error = ENOENT;
1071 goto out;
1072 }
1073 MALLOC(af, struct so_accf *, sizeof(*af), M_ACCF, M_WAITOK | M_ZERO);
1074 if (afp->accf_create != NULL) {
1075 if (afap->af_name[0] != '\0') {
1076 int len = strlen(afap->af_name) + 1;
1077
1078 MALLOC(af->so_accept_filter_str, char *, len, M_ACCF, M_WAITOK);
1079 strcpy(af->so_accept_filter_str, afap->af_name);
1080 }
1081 af->so_accept_filter_arg = afp->accf_create(so, afap->af_arg);
1082 if (af->so_accept_filter_arg == NULL) {
1083 FREE(af->so_accept_filter_str, M_ACCF);
1084 FREE(af, M_ACCF);
1085 so->so_accf = NULL;
1086 error = EINVAL;
1087 goto out;
1088 }
1089 }
1090 af->so_accept_filter = afp;
1091 so->so_accf = af;
1092 so->so_options |= SO_ACCEPTFILTER;
1093out:
1094 if (afap != NULL)
1095 FREE(afap, M_TEMP);
1096 return (error);
1097}
1098#endif /* INET */
1099
1100/*
1101 * Perhaps this routine, and sooptcopyout(), below, ought to come in
1102 * an additional variant to handle the case where the option value needs
1103 * to be some kind of integer, but not a specific size.
1104 * In addition to their use here, these functions are also called by the
1105 * protocol-level pr_ctloutput() routines.
1106 */
1107int
1108sooptcopyin(sopt, buf, len, minlen)
1109 struct sockopt *sopt;
1110 void *buf;
1111 size_t len;
1112 size_t minlen;
1113{
1114 size_t valsize;
1115
1116 /*
1117 * If the user gives us more than we wanted, we ignore it,
1118 * but if we don't get the minimum length the caller
1119 * wants, we return EINVAL. On success, sopt->sopt_valsize
1120 * is set to however much we actually retrieved.
1121 */
1122 if ((valsize = sopt->sopt_valsize) < minlen)
1123 return EINVAL;
1124 if (valsize > len)
1125 sopt->sopt_valsize = valsize = len;
1126
1127 if (sopt->sopt_td != 0)
1128 return (copyin(sopt->sopt_val, buf, valsize));
1129
1130 bcopy(sopt->sopt_val, buf, valsize);
1131 return 0;
1132}
1133
1134int
1135sosetopt(so, sopt)
1136 struct socket *so;
1137 struct sockopt *sopt;
1138{
1139 int error, optval;
1140 struct linger l;
1141 struct timeval tv;
1142 u_long val;
1143
1144 error = 0;
1145 if (sopt->sopt_level != SOL_SOCKET) {
1146 if (so->so_proto && so->so_proto->pr_ctloutput)
1147 return ((*so->so_proto->pr_ctloutput)
1148 (so, sopt));
1149 error = ENOPROTOOPT;
1150 } else {
1151 switch (sopt->sopt_name) {
1152#ifdef INET
1153 case SO_ACCEPTFILTER:
1154 error = do_setopt_accept_filter(so, sopt);
1155 if (error)
1156 goto bad;
1157 break;
1158#endif
1159 case SO_LINGER:
1160 error = sooptcopyin(sopt, &l, sizeof l, sizeof l);
1161 if (error)
1162 goto bad;
1163
1164 so->so_linger = l.l_linger;
1165 if (l.l_onoff)
1166 so->so_options |= SO_LINGER;
1167 else
1168 so->so_options &= ~SO_LINGER;
1169 break;
1170
1171 case SO_DEBUG:
1172 case SO_KEEPALIVE:
1173 case SO_DONTROUTE:
1174 case SO_USELOOPBACK:
1175 case SO_BROADCAST:
1176 case SO_REUSEADDR:
1177 case SO_REUSEPORT:
1178 case SO_OOBINLINE:
1179 case SO_TIMESTAMP:
1180 error = sooptcopyin(sopt, &optval, sizeof optval,
1181 sizeof optval);
1182 if (error)
1183 goto bad;
1184 if (optval)
1185 so->so_options |= sopt->sopt_name;
1186 else
1187 so->so_options &= ~sopt->sopt_name;
1188 break;
1189
1190 case SO_SNDBUF:
1191 case SO_RCVBUF:
1192 case SO_SNDLOWAT:
1193 case SO_RCVLOWAT:
1194 error = sooptcopyin(sopt, &optval, sizeof optval,
1195 sizeof optval);
1196 if (error)
1197 goto bad;
1198
1199 /*
1200 * Values < 1 make no sense for any of these
1201 * options, so disallow them.
1202 */
1203 if (optval < 1) {
1204 error = EINVAL;
1205 goto bad;
1206 }
1207
1208 switch (sopt->sopt_name) {
1209 case SO_SNDBUF:
1210 case SO_RCVBUF:
1211 if (sbreserve(sopt->sopt_name == SO_SNDBUF ?
1212 &so->so_snd : &so->so_rcv, (u_long)optval,
1213 so, curthread) == 0) {
1214 error = ENOBUFS;
1215 goto bad;
1216 }
1217 break;
1218
1219 /*
1220 * Make sure the low-water is never greater than
1221 * the high-water.
1222 */
1223 case SO_SNDLOWAT:
1224 so->so_snd.sb_lowat =
1225 (optval > so->so_snd.sb_hiwat) ?
1226 so->so_snd.sb_hiwat : optval;
1227 break;
1228 case SO_RCVLOWAT:
1229 so->so_rcv.sb_lowat =
1230 (optval > so->so_rcv.sb_hiwat) ?
1231 so->so_rcv.sb_hiwat : optval;
1232 break;
1233 }
1234 break;
1235
1236 case SO_SNDTIMEO:
1237 case SO_RCVTIMEO:
1238 error = sooptcopyin(sopt, &tv, sizeof tv,
1239 sizeof tv);
1240 if (error)
1241 goto bad;
1242
1243 /* assert(hz > 0); */
1244 if (tv.tv_sec < 0 || tv.tv_sec > SHRT_MAX / hz ||
1245 tv.tv_usec < 0 || tv.tv_usec >= 1000000) {
1246 error = EDOM;
1247 goto bad;
1248 }
1249 /* assert(tick > 0); */
1250 /* assert(ULONG_MAX - SHRT_MAX >= 1000000); */
1251 val = (u_long)(tv.tv_sec * hz) + tv.tv_usec / tick;
1252 if (val > SHRT_MAX) {
1253 error = EDOM;
1254 goto bad;
1255 }
1256
1257 switch (sopt->sopt_name) {
1258 case SO_SNDTIMEO:
1259 so->so_snd.sb_timeo = val;
1260 break;
1261 case SO_RCVTIMEO:
1262 so->so_rcv.sb_timeo = val;
1263 break;
1264 }
1265 break;
1266 default:
1267 error = ENOPROTOOPT;
1268 break;
1269 }
1270 if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1271 (void) ((*so->so_proto->pr_ctloutput)
1272 (so, sopt));
1273 }
1274 }
1275bad:
1276 return (error);
1277}
1278
1279/* Helper routine for getsockopt */
1280int
1281sooptcopyout(sopt, buf, len)
1282 struct sockopt *sopt;
1283 void *buf;
1284 size_t len;
1285{
1286 int error;
1287 size_t valsize;
1288
1289 error = 0;
1290
1291 /*
1292 * Documented get behavior is that we always return a value,
1293 * possibly truncated to fit in the user's buffer.
1294 * Traditional behavior is that we always tell the user
1295 * precisely how much we copied, rather than something useful
1296 * like the total amount we had available for her.
1297 * Note that this interface is not idempotent; the entire answer must
1298 * generated ahead of time.
1299 */
1300 valsize = min(len, sopt->sopt_valsize);
1301 sopt->sopt_valsize = valsize;
1302 if (sopt->sopt_val != 0) {
1303 if (sopt->sopt_td != 0)
1304 error = copyout(buf, sopt->sopt_val, valsize);
1305 else
1306 bcopy(buf, sopt->sopt_val, valsize);
1307 }
1308 return error;
1309}
1310
1311int
1312sogetopt(so, sopt)
1313 struct socket *so;
1314 struct sockopt *sopt;
1315{
1316 int error, optval;
1317 struct linger l;
1318 struct timeval tv;
1319#ifdef INET
1320 struct accept_filter_arg *afap;
1321#endif
1322
1323 error = 0;
1324 if (sopt->sopt_level != SOL_SOCKET) {
1325 if (so->so_proto && so->so_proto->pr_ctloutput) {
1326 return ((*so->so_proto->pr_ctloutput)
1327 (so, sopt));
1328 } else
1329 return (ENOPROTOOPT);
1330 } else {
1331 switch (sopt->sopt_name) {
1332#ifdef INET
1333 case SO_ACCEPTFILTER:
1334 if ((so->so_options & SO_ACCEPTCONN) == 0)
1335 return (EINVAL);
1336 MALLOC(afap, struct accept_filter_arg *, sizeof(*afap),
1337 M_TEMP, M_WAITOK | M_ZERO);
1338 if ((so->so_options & SO_ACCEPTFILTER) != 0) {
1339 strcpy(afap->af_name, so->so_accf->so_accept_filter->accf_name);
1340 if (so->so_accf->so_accept_filter_str != NULL)
1341 strcpy(afap->af_arg, so->so_accf->so_accept_filter_str);
1342 }
1343 error = sooptcopyout(sopt, afap, sizeof(*afap));
1344 FREE(afap, M_TEMP);
1345 break;
1346#endif
1347
1348 case SO_LINGER:
1349 l.l_onoff = so->so_options & SO_LINGER;
1350 l.l_linger = so->so_linger;
1351 error = sooptcopyout(sopt, &l, sizeof l);
1352 break;
1353
1354 case SO_USELOOPBACK:
1355 case SO_DONTROUTE:
1356 case SO_DEBUG:
1357 case SO_KEEPALIVE:
1358 case SO_REUSEADDR:
1359 case SO_REUSEPORT:
1360 case SO_BROADCAST:
1361 case SO_OOBINLINE:
1362 case SO_TIMESTAMP:
1363 optval = so->so_options & sopt->sopt_name;
1364integer:
1365 error = sooptcopyout(sopt, &optval, sizeof optval);
1366 break;
1367
1368 case SO_TYPE:
1369 optval = so->so_type;
1370 goto integer;
1371
1372 case SO_ERROR:
1373 optval = so->so_error;
1374 so->so_error = 0;
1375 goto integer;
1376
1377 case SO_SNDBUF:
1378 optval = so->so_snd.sb_hiwat;
1379 goto integer;
1380
1381 case SO_RCVBUF:
1382 optval = so->so_rcv.sb_hiwat;
1383 goto integer;
1384
1385 case SO_SNDLOWAT:
1386 optval = so->so_snd.sb_lowat;
1387 goto integer;
1388
1389 case SO_RCVLOWAT:
1390 optval = so->so_rcv.sb_lowat;
1391 goto integer;
1392
1393 case SO_SNDTIMEO:
1394 case SO_RCVTIMEO:
1395 optval = (sopt->sopt_name == SO_SNDTIMEO ?
1396 so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1397
1398 tv.tv_sec = optval / hz;
1399 tv.tv_usec = (optval % hz) * tick;
1400 error = sooptcopyout(sopt, &tv, sizeof tv);
1401 break;
1402
1403 default:
1404 error = ENOPROTOOPT;
1405 break;
1406 }
1407 return (error);
1408 }
1409}
1410
1411/* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */
1412int
1413soopt_getm(struct sockopt *sopt, struct mbuf **mp)
1414{
1415 struct mbuf *m, *m_prev;
1416 int sopt_size = sopt->sopt_valsize;
1417
1418 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA);
1419 if (m == 0)
1420 return ENOBUFS;
1421 if (sopt_size > MLEN) {
1422 MCLGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT);
1423 if ((m->m_flags & M_EXT) == 0) {
1424 m_free(m);
1425 return ENOBUFS;
1426 }
1427 m->m_len = min(MCLBYTES, sopt_size);
1428 } else {
1429 m->m_len = min(MLEN, sopt_size);
1430 }
1431 sopt_size -= m->m_len;
1432 *mp = m;
1433 m_prev = m;
1434
1435 while (sopt_size) {
1436 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA);
1437 if (m == 0) {
1438 m_freem(*mp);
1439 return ENOBUFS;
1440 }
1441 if (sopt_size > MLEN) {
1442 MCLGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT);
1443 if ((m->m_flags & M_EXT) == 0) {
1444 m_freem(*mp);
1445 return ENOBUFS;
1446 }
1447 m->m_len = min(MCLBYTES, sopt_size);
1448 } else {
1449 m->m_len = min(MLEN, sopt_size);
1450 }
1451 sopt_size -= m->m_len;
1452 m_prev->m_next = m;
1453 m_prev = m;
1454 }
1455 return 0;
1456}
1457
1458/* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
1459int
1460soopt_mcopyin(struct sockopt *sopt, struct mbuf *m)
1461{
1462 struct mbuf *m0 = m;
1463
1464 if (sopt->sopt_val == NULL)
1465 return 0;
1466 while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1467 if (sopt->sopt_td != NULL) {
1468 int error;
1469
1470 error = copyin(sopt->sopt_val, mtod(m, char *),
1471 m->m_len);
1472 if (error != 0) {
1473 m_freem(m0);
1474 return(error);
1475 }
1476 } else
1477 bcopy(sopt->sopt_val, mtod(m, char *), m->m_len);
1478 sopt->sopt_valsize -= m->m_len;
1479 (caddr_t)sopt->sopt_val += m->m_len;
1480 m = m->m_next;
1481 }
1482 if (m != NULL) /* should be allocated enoughly at ip6_sooptmcopyin() */
1483 panic("ip6_sooptmcopyin");
1484 return 0;
1485}
1486
1487/* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
1488int
1489soopt_mcopyout(struct sockopt *sopt, struct mbuf *m)
1490{
1491 struct mbuf *m0 = m;
1492 size_t valsize = 0;
1493
1494 if (sopt->sopt_val == NULL)
1495 return 0;
1496 while (m != NULL && sopt->sopt_valsize >= m->m_len) {
1497 if (sopt->sopt_td != NULL) {
1498 int error;
1499
1500 error = copyout(mtod(m, char *), sopt->sopt_val,
1501 m->m_len);
1502 if (error != 0) {
1503 m_freem(m0);
1504 return(error);
1505 }
1506 } else
1507 bcopy(mtod(m, char *), sopt->sopt_val, m->m_len);
1508 sopt->sopt_valsize -= m->m_len;
1509 (caddr_t)sopt->sopt_val += m->m_len;
1510 valsize += m->m_len;
1511 m = m->m_next;
1512 }
1513 if (m != NULL) {
1514 /* enough soopt buffer should be given from user-land */
1515 m_freem(m0);
1516 return(EINVAL);
1517 }
1518 sopt->sopt_valsize = valsize;
1519 return 0;
1520}
1521
1522void
1523sohasoutofband(so)
1524 register struct socket *so;
1525{
1526 if (so->so_sigio != NULL)
1527 pgsigio(so->so_sigio, SIGURG, 0);
1528 selwakeup(&so->so_rcv.sb_sel);
1529}
1530
1531int
1532sopoll(struct socket *so, int events, struct ucred *cred, struct thread *td)
1533{
1534 int revents = 0;
1535 int s = splnet();
1536
1537 if (events & (POLLIN | POLLRDNORM))
1538 if (soreadable(so))
1539 revents |= events & (POLLIN | POLLRDNORM);
1540
1541 if (events & POLLINIGNEOF)
1542 if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat ||
1543 !TAILQ_EMPTY(&so->so_comp) || so->so_error)
1544 revents |= POLLINIGNEOF;
1545
1546 if (events & (POLLOUT | POLLWRNORM))
1547 if (sowriteable(so))
1548 revents |= events & (POLLOUT | POLLWRNORM);
1549
1550 if (events & (POLLPRI | POLLRDBAND))
1551 if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
1552 revents |= events & (POLLPRI | POLLRDBAND);
1553
1554 if (revents == 0) {
1555 if (events &
1556 (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM |
1557 POLLRDBAND)) {
1558 selrecord(td, &so->so_rcv.sb_sel);
1559 so->so_rcv.sb_flags |= SB_SEL;
1560 }
1561
1562 if (events & (POLLOUT | POLLWRNORM)) {
1563 selrecord(td, &so->so_snd.sb_sel);
1564 so->so_snd.sb_flags |= SB_SEL;
1565 }
1566 }
1567
1568 splx(s);
1569 return (revents);
1570}
1571
1572int
1573sokqfilter(struct file *fp, struct knote *kn)
1574{
1575 struct socket *so = (struct socket *)kn->kn_fp->f_data;
1576 struct sockbuf *sb;
1577 int s;
1578
1579 switch (kn->kn_filter) {
1580 case EVFILT_READ:
1581 if (so->so_options & SO_ACCEPTCONN)
1582 kn->kn_fop = &solisten_filtops;
1583 else
1584 kn->kn_fop = &soread_filtops;
1585 sb = &so->so_rcv;
1586 break;
1587 case EVFILT_WRITE:
1588 kn->kn_fop = &sowrite_filtops;
1589 sb = &so->so_snd;
1590 break;
1591 default:
1592 return (1);
1593 }
1594
1595 s = splnet();
1596 SLIST_INSERT_HEAD(&sb->sb_sel.si_note, kn, kn_selnext);
1597 sb->sb_flags |= SB_KNOTE;
1598 splx(s);
1599 return (0);
1600}
1601
1602static void
1603filt_sordetach(struct knote *kn)
1604{
1605 struct socket *so = (struct socket *)kn->kn_fp->f_data;
1606 int s = splnet();
1607
1608 SLIST_REMOVE(&so->so_rcv.sb_sel.si_note, kn, knote, kn_selnext);
1609 if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_note))
1610 so->so_rcv.sb_flags &= ~SB_KNOTE;
1611 splx(s);
1612}
1613
1614/*ARGSUSED*/
1615static int
1616filt_soread(struct knote *kn, long hint)
1617{
1618 struct socket *so = (struct socket *)kn->kn_fp->f_data;
1619
1620 kn->kn_data = so->so_rcv.sb_cc;
1621 if (so->so_state & SS_CANTRCVMORE) {
1622 kn->kn_flags |= EV_EOF;
1623 kn->kn_fflags = so->so_error;
1624 return (1);
1625 }
1626 if (so->so_error) /* temporary udp error */
1627 return (1);
1628 if (kn->kn_sfflags & NOTE_LOWAT)
1629 return (kn->kn_data >= kn->kn_sdata);
1630 return (kn->kn_data >= so->so_rcv.sb_lowat);
1631}
1632
1633static void
1634filt_sowdetach(struct knote *kn)
1635{
1636 struct socket *so = (struct socket *)kn->kn_fp->f_data;
1637 int s = splnet();
1638
1639 SLIST_REMOVE(&so->so_snd.sb_sel.si_note, kn, knote, kn_selnext);
1640 if (SLIST_EMPTY(&so->so_snd.sb_sel.si_note))
1641 so->so_snd.sb_flags &= ~SB_KNOTE;
1642 splx(s);
1643}
1644
1645/*ARGSUSED*/
1646static int
1647filt_sowrite(struct knote *kn, long hint)
1648{
1649 struct socket *so = (struct socket *)kn->kn_fp->f_data;
1650
1651 kn->kn_data = sbspace(&so->so_snd);
1652 if (so->so_state & SS_CANTSENDMORE) {
1653 kn->kn_flags |= EV_EOF;
1654 kn->kn_fflags = so->so_error;
1655 return (1);
1656 }
1657 if (so->so_error) /* temporary udp error */
1658 return (1);
1659 if (((so->so_state & SS_ISCONNECTED) == 0) &&
1660 (so->so_proto->pr_flags & PR_CONNREQUIRED))
1661 return (0);
1662 if (kn->kn_sfflags & NOTE_LOWAT)
1663 return (kn->kn_data >= kn->kn_sdata);
1664 return (kn->kn_data >= so->so_snd.sb_lowat);
1665}
1666
1667/*ARGSUSED*/
1668static int
1669filt_solisten(struct knote *kn, long hint)
1670{
1671 struct socket *so = (struct socket *)kn->kn_fp->f_data;
1672
1673 kn->kn_data = so->so_qlen - so->so_incqlen;
1674 return (! TAILQ_EMPTY(&so->so_comp));
1675}
1676
1677int
1678socheckuid(struct socket *so, uid_t uid)
1679{
1680
1681 if (so == NULL)
1682 return (EPERM);
1683 if (so->so_cred->cr_uid == uid)
1684 return (0);
1685 return (EPERM);
1686}