AMD64 - Refactor uio_resid and size_t assumptions.
[dragonfly.git] / sys / kern / uipc_syscalls.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1982, 1986, 1989, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * sendfile(2) and related extensions:
6 * Copyright (c) 1998, David Greenman. All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)uipc_syscalls.c 8.4 (Berkeley) 2/21/94
37 * $FreeBSD: src/sys/kern/uipc_syscalls.c,v 1.65.2.17 2003/04/04 17:11:16 tegge Exp $
aca22a94 38 * $DragonFly: src/sys/kern/uipc_syscalls.c,v 1.92 2008/11/26 13:10:56 sephe Exp $
984263bc
MD
39 */
40
984263bc 41#include "opt_ktrace.h"
78812139 42#include "opt_sctp.h"
984263bc
MD
43
44#include <sys/param.h>
45#include <sys/systm.h>
46#include <sys/kernel.h>
47#include <sys/sysproto.h>
48#include <sys/malloc.h>
49#include <sys/filedesc.h>
50#include <sys/event.h>
51#include <sys/proc.h>
52#include <sys/fcntl.h>
53#include <sys/file.h>
54#include <sys/filio.h>
5969a6f1 55#include <sys/kern_syscall.h>
984263bc
MD
56#include <sys/mbuf.h>
57#include <sys/protosw.h>
4860553a 58#include <sys/sfbuf.h>
984263bc
MD
59#include <sys/socket.h>
60#include <sys/socketvar.h>
6b6e0885 61#include <sys/socketops.h>
984263bc
MD
62#include <sys/uio.h>
63#include <sys/vnode.h>
64#include <sys/lock.h>
65#include <sys/mount.h>
66#ifdef KTRACE
67#include <sys/ktrace.h>
68#endif
69#include <vm/vm.h>
70#include <vm/vm_object.h>
71#include <vm/vm_page.h>
72#include <vm/vm_pageout.h>
73#include <vm/vm_kern.h>
74#include <vm/vm_extern.h>
dadab5e9 75#include <sys/file2.h>
770d4c4d 76#include <sys/signalvar.h>
df8d1020 77#include <sys/serialize.h>
984263bc 78
b44419cb
MD
79#include <sys/thread2.h>
80#include <sys/msgport2.h>
d6cb521d 81#include <sys/socketvar2.h>
4599cf19 82#include <net/netmsg2.h>
b44419cb 83
78812139
EN
84#ifdef SCTP
85#include <netinet/sctp_peeloff.h>
86#endif /* SCTP */
87
e66bab2b
MD
88struct sfbuf_mref {
89 struct sf_buf *sf;
90 int mref_count;
91};
92
93static MALLOC_DEFINE(M_SENDFILE, "sendfile", "sendfile sfbuf ref structures");
94
984263bc
MD
95/*
96 * System call interface to the socket abstraction.
97 */
984263bc
MD
98
99extern struct fileops socketops;
100
41c20dac
MD
101/*
102 * socket_args(int domain, int type, int protocol)
103 */
984263bc 104int
75a872f8 105kern_socket(int domain, int type, int protocol, int *res)
984263bc 106{
dadab5e9
MD
107 struct thread *td = curthread;
108 struct proc *p = td->td_proc;
984263bc
MD
109 struct socket *so;
110 struct file *fp;
111 int fd, error;
112
dadab5e9 113 KKASSERT(p);
dadab5e9 114
984263bc
MD
115 error = falloc(p, &fp, &fd);
116 if (error)
117 return (error);
75a872f8 118 error = socreate(domain, &so, type, protocol, td);
984263bc 119 if (error) {
259b8ea0 120 fsetfd(p, NULL, fd);
984263bc 121 } else {
984263bc 122 fp->f_type = DTYPE_SOCKET;
fbb4eeab
JH
123 fp->f_flag = FREAD | FWRITE;
124 fp->f_ops = &socketops;
125 fp->f_data = so;
75a872f8 126 *res = fd;
259b8ea0 127 fsetfd(p, fp, fd);
984263bc 128 }
9f87144f 129 fdrop(fp);
984263bc
MD
130 return (error);
131}
132
02844a31 133int
753fd850 134sys_socket(struct socket_args *uap)
75a872f8
DRJ
135{
136 int error;
137
138 error = kern_socket(uap->domain, uap->type, uap->protocol,
e54488bb 139 &uap->sysmsg_iresult);
75a872f8
DRJ
140
141 return (error);
142}
6b6e0885 143
75a872f8 144int
5969a6f1 145kern_bind(int s, struct sockaddr *sa)
984263bc 146{
dadab5e9
MD
147 struct thread *td = curthread;
148 struct proc *p = td->td_proc;
984263bc 149 struct file *fp;
984263bc
MD
150 int error;
151
dadab5e9 152 KKASSERT(p);
d83b97b9 153 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
154 if (error)
155 return (error);
d83b97b9 156 error = sobind((struct socket *)fp->f_data, sa, td);
9f87144f 157 fdrop(fp);
d83b97b9
MD
158 return (error);
159}
160
161/*
162 * bind_args(int s, caddr_t name, int namelen)
163 */
164int
753fd850 165sys_bind(struct bind_args *uap)
d83b97b9
MD
166{
167 struct sockaddr *sa;
168 int error;
169
984263bc 170 error = getsockaddr(&sa, uap->name, uap->namelen);
d83b97b9 171 if (error)
984263bc 172 return (error);
5969a6f1 173 error = kern_bind(uap->s, sa);
984263bc 174 FREE(sa, M_SONAME);
d83b97b9 175
984263bc
MD
176 return (error);
177}
178
984263bc 179int
5969a6f1 180kern_listen(int s, int backlog)
984263bc 181{
dadab5e9
MD
182 struct thread *td = curthread;
183 struct proc *p = td->td_proc;
984263bc
MD
184 struct file *fp;
185 int error;
186
dadab5e9 187 KKASSERT(p);
5969a6f1 188 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
189 if (error)
190 return (error);
5969a6f1 191 error = solisten((struct socket *)fp->f_data, backlog, td);
9f87144f 192 fdrop(fp);
984263bc
MD
193 return(error);
194}
195
41c20dac 196/*
5969a6f1
DRJ
197 * listen_args(int s, int backlog)
198 */
199int
753fd850 200sys_listen(struct listen_args *uap)
5969a6f1
DRJ
201{
202 int error;
203
204 error = kern_listen(uap->s, uap->backlog);
205 return (error);
206}
207
208/*
f172717f
JH
209 * Returns the accepted socket as well.
210 */
211static boolean_t
212soaccept_predicate(struct netmsg *msg0)
213{
214 struct netmsg_so_notify *msg = (struct netmsg_so_notify *)msg0;
215 struct socket *head = msg->nm_so;
216
217 if (head->so_error != 0) {
4599cf19 218 msg->nm_netmsg.nm_lmsg.ms_error = head->so_error;
f172717f
JH
219 return (TRUE);
220 }
221 if (!TAILQ_EMPTY(&head->so_comp)) {
222 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
223 msg->nm_so = TAILQ_FIRST(&head->so_comp);
224 TAILQ_REMOVE(&head->so_comp, msg->nm_so, so_list);
225 head->so_qlen--;
226
4599cf19 227 msg->nm_netmsg.nm_lmsg.ms_error = 0;
f172717f
JH
228 return (TRUE);
229 }
230 if (head->so_state & SS_CANTRCVMORE) {
4599cf19 231 msg->nm_netmsg.nm_lmsg.ms_error = ECONNABORTED;
f172717f
JH
232 return (TRUE);
233 }
9ba76b73 234 if (msg->nm_fflags & FNONBLOCK) {
4599cf19 235 msg->nm_netmsg.nm_lmsg.ms_error = EWOULDBLOCK;
f172717f
JH
236 return (TRUE);
237 }
238
239 return (FALSE);
240}
241
242/*
5969a6f1
DRJ
243 * The second argument to kern_accept() is a handle to a struct sockaddr.
244 * This allows kern_accept() to return a pointer to an allocated struct
d83b97b9
MD
245 * sockaddr which must be freed later with FREE(). The caller must
246 * initialize *name to NULL.
41c20dac 247 */
02844a31 248int
358e1f78 249kern_accept(int s, int fflags, struct sockaddr **name, int *namelen, int *res)
984263bc 250{
dadab5e9
MD
251 struct thread *td = curthread;
252 struct proc *p = td->td_proc;
984263bc
MD
253 struct file *lfp = NULL;
254 struct file *nfp = NULL;
255 struct sockaddr *sa;
984263bc 256 struct socket *head, *so;
f172717f
JH
257 struct netmsg_so_notify msg;
258 lwkt_port_t port;
984263bc
MD
259 int fd;
260 u_int fflag; /* type must match fp->f_flag */
f172717f 261 int error, tmp;
984263bc 262
259b8ea0 263 *res = -1;
d83b97b9
MD
264 if (name && namelen && *namelen < 0)
265 return (EINVAL);
266
fa541be6 267 error = holdsock(p->p_fd, s, &lfp);
984263bc
MD
268 if (error)
269 return (error);
f172717f
JH
270
271 error = falloc(p, &nfp, &fd);
272 if (error) { /* Probably ran out of file descriptors. */
9f87144f 273 fdrop(lfp);
f172717f
JH
274 return (error);
275 }
984263bc
MD
276 head = (struct socket *)lfp->f_data;
277 if ((head->so_options & SO_ACCEPTCONN) == 0) {
984263bc
MD
278 error = EINVAL;
279 goto done;
280 }
f172717f 281
358e1f78
MD
282 if (fflags & O_FBLOCKING)
283 fflags |= lfp->f_flag & ~FNONBLOCK;
284 else if (fflags & O_FNONBLOCKING)
285 fflags |= lfp->f_flag | FNONBLOCK;
286 else
287 fflags = lfp->f_flag;
288
f172717f 289 /* optimize for uniprocessor case later XXX JH */
934c6849 290 port = head->so_proto->pr_mport(head, NULL, NULL, PRU_PRED);
4599cf19 291 netmsg_init_abortable(&msg.nm_netmsg, &curthread->td_msgport,
a22c590e 292 0,
4599cf19
MD
293 netmsg_so_notify,
294 netmsg_so_notify_doabort);
f172717f 295 msg.nm_predicate = soaccept_predicate;
358e1f78 296 msg.nm_fflags = fflags;
f172717f
JH
297 msg.nm_so = head;
298 msg.nm_etype = NM_REVENT;
a22c590e 299 error = lwkt_domsg(port, &msg.nm_netmsg.nm_lmsg, PCATCH);
f172717f 300 if (error)
984263bc 301 goto done;
984263bc
MD
302
303 /*
f172717f 304 * At this point we have the connection that's ready to be accepted.
984263bc 305 */
f172717f 306 so = msg.nm_so;
984263bc
MD
307
308 fflag = lfp->f_flag;
984263bc
MD
309
310 /* connection has been removed from the listen queue */
6d49aa6f 311 KNOTE(&head->so_rcv.ssb_sel.si_note, 0);
984263bc
MD
312
313 so->so_state &= ~SS_COMP;
314 so->so_head = NULL;
315 if (head->so_sigio != NULL)
316 fsetown(fgetown(head->so_sigio), &so->so_sigio);
317
fbb4eeab 318 nfp->f_type = DTYPE_SOCKET;
984263bc
MD
319 nfp->f_flag = fflag;
320 nfp->f_ops = &socketops;
fbb4eeab 321 nfp->f_data = so;
984263bc
MD
322 /* Sync socket nonblocking/async state with file flags */
323 tmp = fflag & FNONBLOCK;
87de5057 324 (void) fo_ioctl(nfp, FIONBIO, (caddr_t)&tmp, p->p_ucred);
984263bc 325 tmp = fflag & FASYNC;
87de5057 326 (void) fo_ioctl(nfp, FIOASYNC, (caddr_t)&tmp, p->p_ucred);
d83b97b9
MD
327
328 sa = NULL;
984263bc 329 error = soaccept(so, &sa);
d83b97b9
MD
330
331 /*
332 * Set the returned name and namelen as applicable. Set the returned
333 * namelen to 0 for older code which might ignore the return value
334 * from accept.
335 */
336 if (error == 0) {
337 if (sa && name && namelen) {
338 if (*namelen > sa->sa_len)
339 *namelen = sa->sa_len;
340 *name = sa;
341 } else {
342 if (sa)
343 FREE(sa, M_SONAME);
984263bc 344 }
984263bc 345 }
984263bc 346
f172717f 347done:
984263bc 348 /*
259b8ea0
MD
349 * If an error occured clear the reserved descriptor, else associate
350 * nfp with it.
351 *
352 * Note that *res is normally ignored if an error is returned but
353 * a syscall message will still have access to the result code.
984263bc
MD
354 */
355 if (error) {
259b8ea0
MD
356 fsetfd(p, NULL, fd);
357 } else {
358 *res = fd;
359 fsetfd(p, nfp, fd);
984263bc 360 }
259b8ea0 361 fdrop(nfp);
9f87144f 362 fdrop(lfp);
984263bc
MD
363 return (error);
364}
365
d83b97b9 366/*
358e1f78 367 * accept(int s, caddr_t name, int *anamelen)
d83b97b9 368 */
984263bc 369int
753fd850 370sys_accept(struct accept_args *uap)
984263bc 371{
d83b97b9
MD
372 struct sockaddr *sa = NULL;
373 int sa_len;
374 int error;
375
376 if (uap->name) {
377 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
378 if (error)
379 return (error);
380
e54488bb
MD
381 error = kern_accept(uap->s, 0, &sa, &sa_len,
382 &uap->sysmsg_iresult);
358e1f78
MD
383
384 if (error == 0)
385 error = copyout(sa, uap->name, sa_len);
386 if (error == 0) {
387 error = copyout(&sa_len, uap->anamelen,
388 sizeof(*uap->anamelen));
389 }
390 if (sa)
391 FREE(sa, M_SONAME);
392 } else {
e54488bb
MD
393 error = kern_accept(uap->s, 0, NULL, 0,
394 &uap->sysmsg_iresult);
358e1f78
MD
395 }
396 return (error);
397}
398
399/*
b09fd398 400 * extaccept(int s, int fflags, caddr_t name, int *anamelen)
358e1f78
MD
401 */
402int
b09fd398 403sys_extaccept(struct extaccept_args *uap)
358e1f78
MD
404{
405 struct sockaddr *sa = NULL;
406 int sa_len;
407 int error;
408 int fflags = uap->flags & O_FMASK;
409
410 if (uap->name) {
411 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
412 if (error)
413 return (error);
414
e54488bb
MD
415 error = kern_accept(uap->s, fflags, &sa, &sa_len,
416 &uap->sysmsg_iresult);
d83b97b9
MD
417
418 if (error == 0)
419 error = copyout(sa, uap->name, sa_len);
420 if (error == 0) {
421 error = copyout(&sa_len, uap->anamelen,
422 sizeof(*uap->anamelen));
423 }
424 if (sa)
425 FREE(sa, M_SONAME);
426 } else {
e54488bb
MD
427 error = kern_accept(uap->s, fflags, NULL, 0,
428 &uap->sysmsg_iresult);
d83b97b9
MD
429 }
430 return (error);
984263bc
MD
431}
432
358e1f78 433
b44419cb
MD
434/*
435 * Returns TRUE if predicate satisfied.
436 */
437static boolean_t
438soconnected_predicate(struct netmsg *msg0)
439{
440 struct netmsg_so_notify *msg = (struct netmsg_so_notify *)msg0;
441 struct socket *so = msg->nm_so;
442
443 /* check predicate */
444 if (!(so->so_state & SS_ISCONNECTING) || so->so_error != 0) {
4599cf19 445 msg->nm_netmsg.nm_lmsg.ms_error = so->so_error;
b44419cb
MD
446 return (TRUE);
447 }
448
449 return (FALSE);
450}
451
02844a31 452int
358e1f78 453kern_connect(int s, int fflags, struct sockaddr *sa)
984263bc 454{
dadab5e9
MD
455 struct thread *td = curthread;
456 struct proc *p = td->td_proc;
984263bc 457 struct file *fp;
dadab5e9 458 struct socket *so;
8765eadc 459 int error, interrupted = 0;
984263bc 460
d83b97b9 461 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
462 if (error)
463 return (error);
464 so = (struct socket *)fp->f_data;
358e1f78
MD
465
466 if (fflags & O_FBLOCKING)
467 /* fflags &= ~FNONBLOCK; */;
468 else if (fflags & O_FNONBLOCKING)
469 fflags |= FNONBLOCK;
470 else
471 fflags = fp->f_flag;
472
8765eadc 473 if (so->so_state & SS_ISCONNECTING) {
984263bc
MD
474 error = EALREADY;
475 goto done;
476 }
dadab5e9 477 error = soconnect(so, sa, td);
984263bc
MD
478 if (error)
479 goto bad;
358e1f78 480 if ((fflags & FNONBLOCK) && (so->so_state & SS_ISCONNECTING)) {
984263bc
MD
481 error = EINPROGRESS;
482 goto done;
483 }
b44419cb
MD
484 if ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
485 struct netmsg_so_notify msg;
486 lwkt_port_t port;
487
934c6849 488 port = so->so_proto->pr_mport(so, sa, NULL, PRU_PRED);
4599cf19
MD
489 netmsg_init_abortable(&msg.nm_netmsg,
490 &curthread->td_msgport,
a22c590e 491 0,
4599cf19
MD
492 netmsg_so_notify,
493 netmsg_so_notify_doabort);
b44419cb
MD
494 msg.nm_predicate = soconnected_predicate;
495 msg.nm_so = so;
496 msg.nm_etype = NM_REVENT;
a22c590e 497 error = lwkt_domsg(port, &msg.nm_netmsg.nm_lmsg, PCATCH);
8765eadc
SZ
498 if (error == EINTR || error == ERESTART)
499 interrupted = 1;
984263bc
MD
500 }
501 if (error == 0) {
502 error = so->so_error;
503 so->so_error = 0;
504 }
984263bc 505bad:
8765eadc
SZ
506 if (!interrupted)
507 so->so_state &= ~SS_ISCONNECTING;
984263bc
MD
508 if (error == ERESTART)
509 error = EINTR;
510done:
9f87144f 511 fdrop(fp);
984263bc
MD
512 return (error);
513}
514
41c20dac 515/*
d83b97b9
MD
516 * connect_args(int s, caddr_t name, int namelen)
517 */
518int
753fd850 519sys_connect(struct connect_args *uap)
d83b97b9
MD
520{
521 struct sockaddr *sa;
522 int error;
523
524 error = getsockaddr(&sa, uap->name, uap->namelen);
525 if (error)
526 return (error);
358e1f78
MD
527 error = kern_connect(uap->s, 0, sa);
528 FREE(sa, M_SONAME);
529
530 return (error);
531}
532
533/*
534 * connect_args(int s, int fflags, caddr_t name, int namelen)
535 */
536int
b09fd398 537sys_extconnect(struct extconnect_args *uap)
358e1f78
MD
538{
539 struct sockaddr *sa;
540 int error;
541 int fflags = uap->flags & O_FMASK;
542
543 error = getsockaddr(&sa, uap->name, uap->namelen);
544 if (error)
545 return (error);
546 error = kern_connect(uap->s, fflags, sa);
d83b97b9
MD
547 FREE(sa, M_SONAME);
548
549 return (error);
550}
551
984263bc 552int
5969a6f1 553kern_socketpair(int domain, int type, int protocol, int *sv)
984263bc 554{
dadab5e9
MD
555 struct thread *td = curthread;
556 struct proc *p = td->td_proc;
984263bc
MD
557 struct file *fp1, *fp2;
558 struct socket *so1, *so2;
259b8ea0 559 int fd1, fd2, error;
984263bc 560
dadab5e9 561 KKASSERT(p);
5969a6f1 562 error = socreate(domain, &so1, type, protocol, td);
984263bc
MD
563 if (error)
564 return (error);
5969a6f1 565 error = socreate(domain, &so2, type, protocol, td);
984263bc
MD
566 if (error)
567 goto free1;
259b8ea0 568 error = falloc(p, &fp1, &fd1);
984263bc
MD
569 if (error)
570 goto free2;
259b8ea0 571 sv[0] = fd1;
fbb4eeab 572 fp1->f_data = so1;
259b8ea0 573 error = falloc(p, &fp2, &fd2);
984263bc
MD
574 if (error)
575 goto free3;
fbb4eeab 576 fp2->f_data = so2;
259b8ea0 577 sv[1] = fd2;
984263bc
MD
578 error = soconnect2(so1, so2);
579 if (error)
580 goto free4;
5969a6f1 581 if (type == SOCK_DGRAM) {
984263bc
MD
582 /*
583 * Datagram socket connection is asymmetric.
584 */
585 error = soconnect2(so2, so1);
586 if (error)
587 goto free4;
588 }
fbb4eeab 589 fp1->f_type = fp2->f_type = DTYPE_SOCKET;
984263bc
MD
590 fp1->f_flag = fp2->f_flag = FREAD|FWRITE;
591 fp1->f_ops = fp2->f_ops = &socketops;
259b8ea0
MD
592 fsetfd(p, fp1, fd1);
593 fsetfd(p, fp2, fd2);
9f87144f
MD
594 fdrop(fp1);
595 fdrop(fp2);
984263bc
MD
596 return (error);
597free4:
259b8ea0 598 fsetfd(p, NULL, fd2);
9f87144f 599 fdrop(fp2);
984263bc 600free3:
259b8ea0 601 fsetfd(p, NULL, fd1);
9f87144f 602 fdrop(fp1);
984263bc 603free2:
9ba76b73 604 (void)soclose(so2, 0);
984263bc 605free1:
9ba76b73 606 (void)soclose(so1, 0);
984263bc
MD
607 return (error);
608}
609
5969a6f1
DRJ
610/*
611 * socketpair(int domain, int type, int protocol, int *rsv)
612 */
613int
753fd850 614sys_socketpair(struct socketpair_args *uap)
5969a6f1
DRJ
615{
616 int error, sockv[2];
617
618 error = kern_socketpair(uap->domain, uap->type, uap->protocol, sockv);
619
620 if (error == 0)
621 error = copyout(sockv, uap->rsv, sizeof(sockv));
622 return (error);
623}
624
35fbb1d9 625int
3e1837ce 626kern_sendmsg(int s, struct sockaddr *sa, struct uio *auio,
e54488bb 627 struct mbuf *control, int flags, size_t *res)
984263bc 628{
dadab5e9 629 struct thread *td = curthread;
7278a846 630 struct lwp *lp = td->td_lwp;
dadab5e9 631 struct proc *p = td->td_proc;
984263bc 632 struct file *fp;
e54488bb
MD
633 size_t len;
634 int error;
984263bc
MD
635 struct socket *so;
636#ifdef KTRACE
637 struct iovec *ktriov = NULL;
638 struct uio ktruio;
639#endif
640
641 error = holdsock(p->p_fd, s, &fp);
642 if (error)
643 return (error);
984263bc 644#ifdef KTRACE
dadab5e9 645 if (KTRPOINT(td, KTR_GENIO)) {
3e1837ce 646 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
984263bc
MD
647
648 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
3e1837ce
DRJ
649 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
650 ktruio = *auio;
984263bc
MD
651 }
652#endif
3e1837ce 653 len = auio->uio_resid;
984263bc 654 so = (struct socket *)fp->f_data;
9ba76b73
MD
655 if ((flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
656 if (fp->f_flag & FNONBLOCK)
657 flags |= MSG_FNONBLOCKING;
658 }
6b6e0885 659 error = so_pru_sosend(so, sa, auio, NULL, control, flags, td);
984263bc 660 if (error) {
3e1837ce 661 if (auio->uio_resid != len && (error == ERESTART ||
984263bc
MD
662 error == EINTR || error == EWOULDBLOCK))
663 error = 0;
664 if (error == EPIPE)
7278a846 665 lwpsignal(p, lp, SIGPIPE);
984263bc 666 }
984263bc
MD
667#ifdef KTRACE
668 if (ktriov != NULL) {
669 if (error == 0) {
670 ktruio.uio_iov = ktriov;
3e1837ce 671 ktruio.uio_resid = len - auio->uio_resid;
9fb04d14 672 ktrgenio(lp, s, UIO_WRITE, &ktruio, error);
984263bc
MD
673 }
674 FREE(ktriov, M_TEMP);
675 }
676#endif
35fbb1d9 677 if (error == 0)
3e1837ce 678 *res = len - auio->uio_resid;
0bbfa91b 679done:
9f87144f 680 fdrop(fp);
984263bc
MD
681 return (error);
682}
683
41c20dac
MD
684/*
685 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
686 */
984263bc 687int
753fd850 688sys_sendto(struct sendto_args *uap)
984263bc 689{
3e1837ce
DRJ
690 struct thread *td = curthread;
691 struct uio auio;
984263bc 692 struct iovec aiov;
35fbb1d9
DRJ
693 struct sockaddr *sa = NULL;
694 int error;
984263bc 695
35fbb1d9
DRJ
696 if (uap->to) {
697 error = getsockaddr(&sa, uap->to, uap->tolen);
698 if (error)
699 return (error);
35fbb1d9 700 }
984263bc
MD
701 aiov.iov_base = uap->buf;
702 aiov.iov_len = uap->len;
3e1837ce
DRJ
703 auio.uio_iov = &aiov;
704 auio.uio_iovcnt = 1;
705 auio.uio_offset = 0;
706 auio.uio_resid = uap->len;
707 auio.uio_segflg = UIO_USERSPACE;
708 auio.uio_rw = UIO_WRITE;
709 auio.uio_td = td;
984263bc 710
3e1837ce 711 error = kern_sendmsg(uap->s, sa, &auio, NULL, uap->flags,
e54488bb 712 &uap->sysmsg_szresult);
984263bc 713
35fbb1d9
DRJ
714 if (sa)
715 FREE(sa, M_SONAME);
716 return (error);
984263bc
MD
717}
718
41c20dac 719/*
35fbb1d9 720 * sendmsg_args(int s, caddr_t msg, int flags)
41c20dac 721 */
984263bc 722int
753fd850 723sys_sendmsg(struct sendmsg_args *uap)
984263bc 724{
3e1837ce 725 struct thread *td = curthread;
984263bc 726 struct msghdr msg;
3e1837ce 727 struct uio auio;
75a872f8 728 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
35fbb1d9
DRJ
729 struct sockaddr *sa = NULL;
730 struct mbuf *control = NULL;
75a872f8 731 int error;
984263bc 732
35fbb1d9 733 error = copyin(uap->msg, (caddr_t)&msg, sizeof(msg));
984263bc
MD
734 if (error)
735 return (error);
35fbb1d9
DRJ
736
737 /*
738 * Conditionally copyin msg.msg_name.
739 */
740 if (msg.msg_name) {
741 error = getsockaddr(&sa, msg.msg_name, msg.msg_namelen);
742 if (error)
743 return (error);
35fbb1d9
DRJ
744 }
745
746 /*
3e1837ce 747 * Populate auio.
35fbb1d9 748 */
75a872f8 749 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
ef5c76d7 750 &auio.uio_resid);
984263bc 751 if (error)
8130f673 752 goto cleanup2;
3e1837ce
DRJ
753 auio.uio_iov = iov;
754 auio.uio_iovcnt = msg.msg_iovlen;
755 auio.uio_offset = 0;
3e1837ce
DRJ
756 auio.uio_segflg = UIO_USERSPACE;
757 auio.uio_rw = UIO_WRITE;
758 auio.uio_td = td;
35fbb1d9
DRJ
759
760 /*
761 * Conditionally copyin msg.msg_control.
762 */
763 if (msg.msg_control) {
3e1837ce
DRJ
764 if (msg.msg_controllen < sizeof(struct cmsghdr) ||
765 msg.msg_controllen > MLEN) {
35fbb1d9
DRJ
766 error = EINVAL;
767 goto cleanup;
768 }
74f1caca 769 control = m_get(MB_WAIT, MT_CONTROL);
3e1837ce
DRJ
770 if (control == NULL) {
771 error = ENOBUFS;
35fbb1d9 772 goto cleanup;
3e1837ce
DRJ
773 }
774 control->m_len = msg.msg_controllen;
775 error = copyin(msg.msg_control, mtod(control, caddr_t),
776 msg.msg_controllen);
777 if (error) {
778 m_free(control);
779 goto cleanup;
780 }
35fbb1d9
DRJ
781 }
782
3e1837ce 783 error = kern_sendmsg(uap->s, sa, &auio, control, uap->flags,
e54488bb 784 &uap->sysmsg_szresult);
35fbb1d9
DRJ
785
786cleanup:
8130f673
MD
787 iovec_free(&iov, aiov);
788cleanup2:
35fbb1d9
DRJ
789 if (sa)
790 FREE(sa, M_SONAME);
984263bc
MD
791 return (error);
792}
984263bc 793
41c20dac 794/*
3e1837ce
DRJ
795 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
796 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
797 * Don't forget to FREE() and m_free() these if they are returned.
41c20dac 798 */
984263bc 799int
3e1837ce 800kern_recvmsg(int s, struct sockaddr **sa, struct uio *auio,
e54488bb 801 struct mbuf **control, int *flags, size_t *res)
984263bc 802{
dadab5e9
MD
803 struct thread *td = curthread;
804 struct proc *p = td->td_proc;
984263bc 805 struct file *fp;
e54488bb
MD
806 size_t len;
807 int error;
9ba76b73 808 int lflags;
984263bc 809 struct socket *so;
984263bc
MD
810#ifdef KTRACE
811 struct iovec *ktriov = NULL;
812 struct uio ktruio;
813#endif
814
815 error = holdsock(p->p_fd, s, &fp);
816 if (error)
817 return (error);
984263bc 818#ifdef KTRACE
dadab5e9 819 if (KTRPOINT(td, KTR_GENIO)) {
3e1837ce 820 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
984263bc
MD
821
822 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
3e1837ce
DRJ
823 bcopy(auio->uio_iov, ktriov, iovlen);
824 ktruio = *auio;
984263bc
MD
825 }
826#endif
3e1837ce 827 len = auio->uio_resid;
984263bc 828 so = (struct socket *)fp->f_data;
9ba76b73
MD
829
830 if (flags == NULL || (*flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
831 if (fp->f_flag & FNONBLOCK) {
832 if (flags) {
833 *flags |= MSG_FNONBLOCKING;
834 } else {
835 lflags = MSG_FNONBLOCKING;
836 flags = &lflags;
837 }
838 }
839 }
840
6b6e0885 841 error = so_pru_soreceive(so, sa, auio, NULL, control, flags);
984263bc 842 if (error) {
3e1837ce 843 if (auio->uio_resid != len && (error == ERESTART ||
984263bc
MD
844 error == EINTR || error == EWOULDBLOCK))
845 error = 0;
846 }
847#ifdef KTRACE
848 if (ktriov != NULL) {
849 if (error == 0) {
850 ktruio.uio_iov = ktriov;
3e1837ce 851 ktruio.uio_resid = len - auio->uio_resid;
9fb04d14 852 ktrgenio(td->td_lwp, s, UIO_READ, &ktruio, error);
984263bc
MD
853 }
854 FREE(ktriov, M_TEMP);
855 }
856#endif
35fbb1d9 857 if (error == 0)
3e1837ce 858 *res = len - auio->uio_resid;
0bbfa91b 859done:
9f87144f 860 fdrop(fp);
984263bc
MD
861 return (error);
862}
863
41c20dac
MD
864/*
865 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
866 * caddr_t from, int *fromlenaddr)
867 */
984263bc 868int
753fd850 869sys_recvfrom(struct recvfrom_args *uap)
984263bc 870{
3e1837ce
DRJ
871 struct thread *td = curthread;
872 struct uio auio;
984263bc 873 struct iovec aiov;
3e1837ce 874 struct sockaddr *sa = NULL;
35fbb1d9 875 int error, fromlen;
984263bc 876
3e1837ce 877 if (uap->from && uap->fromlenaddr) {
35fbb1d9 878 error = copyin(uap->fromlenaddr, &fromlen, sizeof(fromlen));
984263bc
MD
879 if (error)
880 return (error);
3e1837ce
DRJ
881 if (fromlen < 0)
882 return (EINVAL);
35fbb1d9
DRJ
883 } else {
884 fromlen = 0;
885 }
984263bc
MD
886 aiov.iov_base = uap->buf;
887 aiov.iov_len = uap->len;
3e1837ce
DRJ
888 auio.uio_iov = &aiov;
889 auio.uio_iovcnt = 1;
890 auio.uio_offset = 0;
891 auio.uio_resid = uap->len;
892 auio.uio_segflg = UIO_USERSPACE;
893 auio.uio_rw = UIO_READ;
894 auio.uio_td = td;
984263bc 895
3e1837ce 896 error = kern_recvmsg(uap->s, uap->from ? &sa : NULL, &auio, NULL,
e54488bb 897 &uap->flags, &uap->sysmsg_szresult);
984263bc 898
3e1837ce 899 if (error == 0 && uap->from) {
c3996757
MD
900 /* note: sa may still be NULL */
901 if (sa) {
902 fromlen = MIN(fromlen, sa->sa_len);
903 error = copyout(sa, uap->from, fromlen);
904 } else {
905 fromlen = 0;
906 }
907 if (error == 0) {
35fbb1d9 908 error = copyout(&fromlen, uap->fromlenaddr,
c3996757
MD
909 sizeof(fromlen));
910 }
35fbb1d9 911 }
3e1837ce
DRJ
912 if (sa)
913 FREE(sa, M_SONAME);
984263bc 914
984263bc
MD
915 return (error);
916}
984263bc 917
41c20dac
MD
918/*
919 * recvmsg_args(int s, struct msghdr *msg, int flags)
920 */
984263bc 921int
753fd850 922sys_recvmsg(struct recvmsg_args *uap)
984263bc 923{
3e1837ce 924 struct thread *td = curthread;
984263bc 925 struct msghdr msg;
3e1837ce 926 struct uio auio;
75a872f8
DRJ
927 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
928 struct mbuf *m, *control = NULL;
3e1837ce 929 struct sockaddr *sa = NULL;
35fbb1d9 930 caddr_t ctlbuf;
3e1837ce 931 socklen_t *ufromlenp, *ucontrollenp;
75a872f8 932 int error, fromlen, controllen, len, flags, *uflagsp;
984263bc 933
35fbb1d9
DRJ
934 /*
935 * This copyin handles everything except the iovec.
936 */
937 error = copyin(uap->msg, &msg, sizeof(msg));
984263bc
MD
938 if (error)
939 return (error);
35fbb1d9 940
3e1837ce
DRJ
941 if (msg.msg_name && msg.msg_namelen < 0)
942 return (EINVAL);
943 if (msg.msg_control && msg.msg_controllen < 0)
944 return (EINVAL);
945
946 ufromlenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
35fbb1d9 947 msg_namelen));
35fbb1d9
DRJ
948 ucontrollenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
949 msg_controllen));
3e1837ce
DRJ
950 uflagsp = (int *)((caddr_t)uap->msg + offsetof(struct msghdr,
951 msg_flags));
35fbb1d9
DRJ
952
953 /*
3e1837ce 954 * Populate auio.
35fbb1d9 955 */
75a872f8 956 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
ef5c76d7 957 &auio.uio_resid);
984263bc 958 if (error)
75a872f8 959 return (error);
3e1837ce
DRJ
960 auio.uio_iov = iov;
961 auio.uio_iovcnt = msg.msg_iovlen;
962 auio.uio_offset = 0;
3e1837ce
DRJ
963 auio.uio_segflg = UIO_USERSPACE;
964 auio.uio_rw = UIO_READ;
965 auio.uio_td = td;
35fbb1d9 966
b7ccd728 967 flags = uap->flags;
35fbb1d9 968
e54488bb
MD
969 error = kern_recvmsg(uap->s,
970 (msg.msg_name ? &sa : NULL), &auio,
971 (msg.msg_control ? &control : NULL), &flags,
972 &uap->sysmsg_szresult);
35fbb1d9
DRJ
973
974 /*
3e1837ce 975 * Conditionally copyout the name and populate the namelen field.
35fbb1d9 976 */
3e1837ce 977 if (error == 0 && msg.msg_name) {
b9cd15b9
YT
978 /* note: sa may still be NULL */
979 if (sa != NULL) {
980 fromlen = MIN(msg.msg_namelen, sa->sa_len);
981 error = copyout(sa, msg.msg_name, fromlen);
b4354d10 982 } else {
b9cd15b9 983 fromlen = 0;
b4354d10 984 }
35fbb1d9 985 if (error == 0)
3e1837ce
DRJ
986 error = copyout(&fromlen, ufromlenp,
987 sizeof(*ufromlenp));
984263bc 988 }
35fbb1d9
DRJ
989
990 /*
991 * Copyout msg.msg_control and msg.msg_controllen.
992 */
3e1837ce 993 if (error == 0 && msg.msg_control) {
35fbb1d9 994 len = msg.msg_controllen;
3e1837ce
DRJ
995 m = control;
996 ctlbuf = (caddr_t)msg.msg_control;
35fbb1d9
DRJ
997
998 while(m && len > 0) {
999 unsigned int tocopy;
1000
1001 if (len >= m->m_len) {
1002 tocopy = m->m_len;
1003 } else {
1004 msg.msg_flags |= MSG_CTRUNC;
1005 tocopy = len;
1006 }
1007
1008 error = copyout(mtod(m, caddr_t), ctlbuf, tocopy);
1009 if (error)
1010 goto cleanup;
1011
1012 ctlbuf += tocopy;
1013 len -= tocopy;
1014 m = m->m_next;
1015 }
3e1837ce
DRJ
1016 controllen = ctlbuf - (caddr_t)msg.msg_control;
1017 error = copyout(&controllen, ucontrollenp,
35fbb1d9
DRJ
1018 sizeof(*ucontrollenp));
1019 }
1020
3e1837ce
DRJ
1021 if (error == 0)
1022 error = copyout(&flags, uflagsp, sizeof(*uflagsp));
1023
35fbb1d9 1024cleanup:
3e1837ce
DRJ
1025 if (sa)
1026 FREE(sa, M_SONAME);
75a872f8 1027 iovec_free(&iov, aiov);
3e1837ce
DRJ
1028 if (control)
1029 m_freem(control);
984263bc
MD
1030 return (error);
1031}
1032
41c20dac 1033/*
201305ad
DRJ
1034 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1035 * in kernel pointer instead of a userland pointer. This allows us
1036 * to manipulate socket options in the emulation code.
41c20dac 1037 */
984263bc 1038int
201305ad 1039kern_setsockopt(int s, struct sockopt *sopt)
984263bc 1040{
dadab5e9
MD
1041 struct thread *td = curthread;
1042 struct proc *p = td->td_proc;
984263bc 1043 struct file *fp;
984263bc
MD
1044 int error;
1045
b4354d10 1046 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
984263bc 1047 return (EFAULT);
201305ad 1048 if (sopt->sopt_valsize < 0)
984263bc
MD
1049 return (EINVAL);
1050
201305ad 1051 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
1052 if (error)
1053 return (error);
1054
201305ad 1055 error = sosetopt((struct socket *)fp->f_data, sopt);
9f87144f 1056 fdrop(fp);
201305ad
DRJ
1057 return (error);
1058}
1059
1060/*
1061 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
1062 */
1063int
753fd850 1064sys_setsockopt(struct setsockopt_args *uap)
201305ad
DRJ
1065{
1066 struct thread *td = curthread;
1067 struct sockopt sopt;
1068 int error;
1069
984263bc
MD
1070 sopt.sopt_level = uap->level;
1071 sopt.sopt_name = uap->name;
984263bc 1072 sopt.sopt_valsize = uap->valsize;
dadab5e9 1073 sopt.sopt_td = td;
aca22a94 1074 sopt.sopt_val = NULL;
201305ad 1075
b4354d10
MD
1076 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1077 return (EINVAL);
792239df
AE
1078 if (uap->val) {
1079 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1080 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1081 if (error)
1082 goto out;
792239df 1083 }
aca22a94 1084
201305ad 1085 error = kern_setsockopt(uap->s, &sopt);
de0003fe 1086out:
792239df
AE
1087 if (uap->val)
1088 kfree(sopt.sopt_val, M_TEMP);
984263bc
MD
1089 return(error);
1090}
1091
41c20dac 1092/*
201305ad
DRJ
1093 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1094 * in kernel pointer instead of a userland pointer. This allows us
1095 * to manipulate socket options in the emulation code.
41c20dac 1096 */
984263bc 1097int
201305ad 1098kern_getsockopt(int s, struct sockopt *sopt)
984263bc 1099{
dadab5e9
MD
1100 struct thread *td = curthread;
1101 struct proc *p = td->td_proc;
201305ad
DRJ
1102 struct file *fp;
1103 int error;
984263bc 1104
b4354d10 1105 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
201305ad 1106 return (EFAULT);
b4354d10 1107 if (sopt->sopt_valsize < 0 || sopt->sopt_valsize > SOMAXOPT_SIZE)
201305ad
DRJ
1108 return (EINVAL);
1109
1110 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
1111 if (error)
1112 return (error);
201305ad
DRJ
1113
1114 error = sogetopt((struct socket *)fp->f_data, sopt);
9f87144f 1115 fdrop(fp);
201305ad
DRJ
1116 return (error);
1117}
1118
1119/*
1120 * getsockopt_Args(int s, int level, int name, caddr_t val, int *avalsize)
1121 */
1122int
753fd850 1123sys_getsockopt(struct getsockopt_args *uap)
201305ad
DRJ
1124{
1125 struct thread *td = curthread;
1126 struct sockopt sopt;
1127 int error, valsize;
1128
984263bc 1129 if (uap->val) {
201305ad
DRJ
1130 error = copyin(uap->avalsize, &valsize, sizeof(valsize));
1131 if (error)
984263bc 1132 return (error);
984263bc
MD
1133 } else {
1134 valsize = 0;
1135 }
1136
984263bc
MD
1137 sopt.sopt_level = uap->level;
1138 sopt.sopt_name = uap->name;
201305ad 1139 sopt.sopt_valsize = valsize;
dadab5e9 1140 sopt.sopt_td = td;
aca22a94 1141 sopt.sopt_val = NULL;
984263bc 1142
aca22a94 1143 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
b4354d10 1144 return (EINVAL);
792239df
AE
1145 if (uap->val) {
1146 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1147 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1148 if (error)
1149 goto out;
792239df 1150 }
aca22a94 1151
201305ad 1152 error = kern_getsockopt(uap->s, &sopt);
de0003fe
AE
1153 if (error)
1154 goto out;
1155 valsize = sopt.sopt_valsize;
1156 error = copyout(&valsize, uap->avalsize, sizeof(valsize));
1157 if (error)
1158 goto out;
792239df
AE
1159 if (uap->val)
1160 error = copyout(sopt.sopt_val, uap->val, sopt.sopt_valsize);
de0003fe 1161out:
792239df
AE
1162 if (uap->val)
1163 kfree(sopt.sopt_val, M_TEMP);
984263bc
MD
1164 return (error);
1165}
1166
1167/*
5969a6f1
DRJ
1168 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1169 * This allows kern_getsockname() to return a pointer to an allocated struct
1170 * sockaddr which must be freed later with FREE(). The caller must
1171 * initialize *name to NULL.
984263bc 1172 */
5969a6f1
DRJ
1173int
1174kern_getsockname(int s, struct sockaddr **name, int *namelen)
984263bc 1175{
dadab5e9
MD
1176 struct thread *td = curthread;
1177 struct proc *p = td->td_proc;
984263bc 1178 struct file *fp;
dadab5e9 1179 struct socket *so;
5969a6f1
DRJ
1180 struct sockaddr *sa = NULL;
1181 int error;
984263bc 1182
5969a6f1 1183 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
1184 if (error)
1185 return (error);
5969a6f1 1186 if (*namelen < 0) {
9f87144f 1187 fdrop(fp);
984263bc
MD
1188 return (EINVAL);
1189 }
1190 so = (struct socket *)fp->f_data;
6b6e0885 1191 error = so_pru_sockaddr(so, &sa);
5969a6f1 1192 if (error == 0) {
b4354d10 1193 if (sa == NULL) {
5969a6f1
DRJ
1194 *namelen = 0;
1195 } else {
1196 *namelen = MIN(*namelen, sa->sa_len);
1197 *name = sa;
1198 }
984263bc
MD
1199 }
1200
9f87144f 1201 fdrop(fp);
984263bc
MD
1202 return (error);
1203}
1204
5969a6f1
DRJ
1205/*
1206 * getsockname_args(int fdes, caddr_t asa, int *alen)
1207 *
1208 * Get socket name.
1209 */
984263bc 1210int
753fd850 1211sys_getsockname(struct getsockname_args *uap)
984263bc 1212{
5969a6f1
DRJ
1213 struct sockaddr *sa = NULL;
1214 int error, sa_len;
1215
1216 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1217 if (error)
1218 return (error);
1219
1220 error = kern_getsockname(uap->fdes, &sa, &sa_len);
984263bc 1221
5969a6f1
DRJ
1222 if (error == 0)
1223 error = copyout(sa, uap->asa, sa_len);
1224 if (error == 0)
1225 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1226 if (sa)
1227 FREE(sa, M_SONAME);
1228 return (error);
984263bc
MD
1229}
1230
984263bc 1231/*
5969a6f1
DRJ
1232 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1233 * This allows kern_getpeername() to return a pointer to an allocated struct
1234 * sockaddr which must be freed later with FREE(). The caller must
1235 * initialize *name to NULL.
984263bc 1236 */
5969a6f1
DRJ
1237int
1238kern_getpeername(int s, struct sockaddr **name, int *namelen)
984263bc 1239{
dadab5e9
MD
1240 struct thread *td = curthread;
1241 struct proc *p = td->td_proc;
984263bc 1242 struct file *fp;
dadab5e9 1243 struct socket *so;
5969a6f1
DRJ
1244 struct sockaddr *sa = NULL;
1245 int error;
984263bc 1246
5969a6f1 1247 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
1248 if (error)
1249 return (error);
5969a6f1 1250 if (*namelen < 0) {
9f87144f 1251 fdrop(fp);
5969a6f1
DRJ
1252 return (EINVAL);
1253 }
984263bc
MD
1254 so = (struct socket *)fp->f_data;
1255 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
9f87144f 1256 fdrop(fp);
984263bc
MD
1257 return (ENOTCONN);
1258 }
6b6e0885 1259 error = so_pru_peeraddr(so, &sa);
5969a6f1 1260 if (error == 0) {
b4354d10 1261 if (sa == NULL) {
5969a6f1
DRJ
1262 *namelen = 0;
1263 } else {
1264 *namelen = MIN(*namelen, sa->sa_len);
1265 *name = sa;
1266 }
984263bc 1267 }
5969a6f1 1268
9f87144f 1269 fdrop(fp);
984263bc
MD
1270 return (error);
1271}
1272
5969a6f1
DRJ
1273/*
1274 * getpeername_args(int fdes, caddr_t asa, int *alen)
1275 *
1276 * Get name of peer for connected socket.
1277 */
984263bc 1278int
753fd850 1279sys_getpeername(struct getpeername_args *uap)
984263bc 1280{
5969a6f1
DRJ
1281 struct sockaddr *sa = NULL;
1282 int error, sa_len;
1283
1284 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1285 if (error)
1286 return (error);
1287
1288 error = kern_getpeername(uap->fdes, &sa, &sa_len);
1289
1290 if (error == 0)
1291 error = copyout(sa, uap->asa, sa_len);
1292 if (error == 0)
1293 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1294 if (sa)
1295 FREE(sa, M_SONAME);
1296 return (error);
984263bc
MD
1297}
1298
984263bc 1299int
02844a31 1300getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len)
984263bc
MD
1301{
1302 struct sockaddr *sa;
1303 int error;
1304
02844a31 1305 *namp = NULL;
984263bc
MD
1306 if (len > SOCK_MAXADDRLEN)
1307 return ENAMETOOLONG;
02844a31
MD
1308 if (len < offsetof(struct sockaddr, sa_data[0]))
1309 return EDOM;
984263bc
MD
1310 MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1311 error = copyin(uaddr, sa, len);
1312 if (error) {
1313 FREE(sa, M_SONAME);
1314 } else {
75a872f8
DRJ
1315#if BYTE_ORDER != BIG_ENDIAN
1316 /*
1317 * The bind(), connect(), and sendto() syscalls were not
1318 * versioned for COMPAT_43. Thus, this check must stay.
1319 */
984263bc
MD
1320 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1321 sa->sa_family = sa->sa_len;
1322#endif
1323 sa->sa_len = len;
1324 *namp = sa;
1325 }
1326 return error;
1327}
1328
1329/*
b4caac98
MD
1330 * Detach a mapped page and release resources back to the system.
1331 * We must release our wiring and if the object is ripped out
1332 * from under the vm_page we become responsible for freeing the
df8d1020 1333 * page. These routines must be MPSAFE.
b4caac98
MD
1334 *
1335 * XXX HACK XXX TEMPORARY UNTIL WE IMPLEMENT EXT MBUF REFERENCE COUNTING
df8d1020
MD
1336 *
1337 * XXX vm_page_*() routines are not MPSAFE yet, the MP lock is required.
b4caac98
MD
1338 */
1339static void
013a4c0e 1340sf_buf_mref(void *arg)
b4caac98 1341{
e66bab2b
MD
1342 struct sfbuf_mref *sfm = arg;
1343
df8d1020
MD
1344 /*
1345 * We must already hold a ref so there is no race to 0, just
1346 * atomically increment the count.
1347 */
1348 atomic_add_int(&sfm->mref_count, 1);
b4caac98
MD
1349}
1350
1351static void
013a4c0e 1352sf_buf_mfree(void *arg)
b4caac98 1353{
e66bab2b 1354 struct sfbuf_mref *sfm = arg;
b4caac98 1355 vm_page_t m;
e66bab2b
MD
1356
1357 KKASSERT(sfm->mref_count > 0);
321e057f 1358 if (atomic_fetchadd_int(&sfm->mref_count, -1) == 1) {
df8d1020 1359 /*
df8d1020
MD
1360 * XXX vm_page_*() and SFBUF routines not MPSAFE yet.
1361 */
321e057f
SZ
1362 get_mplock();
1363 crit_enter();
1364 m = sf_buf_page(sfm->sf);
1365 sf_buf_free(sfm->sf);
1366 vm_page_unwire(m, 0);
1367 if (m->wire_count == 0 && m->object == NULL)
1368 vm_page_try_to_free(m);
1369 crit_exit();
1370 rel_mplock();
1371 kfree(sfm, M_SENDFILE);
b4caac98
MD
1372 }
1373}
1374
1375/*
984263bc
MD
1376 * sendfile(2).
1377 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1378 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1379 *
1380 * Send a file specified by 'fd' and starting at 'offset' to a socket
1381 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1382 * nbytes == 0. Optionally add a header and/or trailer to the socket
1383 * output. If specified, write the total number of bytes sent into *sbytes.
75a872f8
DRJ
1384 *
1385 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1386 * the headers to count against the remaining bytes to be sent from
1387 * the file descriptor. We may wish to implement a compatibility syscall
1388 * in the future.
984263bc
MD
1389 */
1390int
753fd850 1391sys_sendfile(struct sendfile_args *uap)
984263bc 1392{
dadab5e9
MD
1393 struct thread *td = curthread;
1394 struct proc *p = td->td_proc;
984263bc 1395 struct file *fp;
75a872f8 1396 struct vnode *vp = NULL;
984263bc 1397 struct sf_hdtr hdtr;
75a872f8
DRJ
1398 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
1399 struct uio auio;
30eeba44 1400 struct mbuf *mheader = NULL;
e54488bb
MD
1401 size_t hbytes = 0;
1402 size_t tbytes;
1403 off_t hdtr_size = 0;
1404 off_t sbytes;
1405 int error;
984263bc 1406
dadab5e9 1407 KKASSERT(p);
dadab5e9 1408
984263bc
MD
1409 /*
1410 * Do argument checking. Must be a regular file in, stream
1411 * type and connected socket out, positive offset.
1412 */
fa541be6 1413 fp = holdfp(p->p_fd, uap->fd, FREAD);
984263bc 1414 if (fp == NULL) {
f0846490 1415 return (EBADF);
984263bc
MD
1416 }
1417 if (fp->f_type != DTYPE_VNODE) {
9f87144f 1418 fdrop(fp);
f0846490 1419 return (EINVAL);
984263bc
MD
1420 }
1421 vp = (struct vnode *)fp->f_data;
1422 vref(vp);
9f87144f 1423 fdrop(fp);
75a872f8
DRJ
1424
1425 /*
1426 * If specified, get the pointer to the sf_hdtr struct for
1427 * any headers/trailers.
1428 */
1429 if (uap->hdtr) {
1430 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1431 if (error)
1432 goto done;
1433 /*
1434 * Send any headers.
1435 */
1436 if (hdtr.headers) {
1437 error = iovec_copyin(hdtr.headers, &iov, aiov,
ef5c76d7 1438 hdtr.hdr_cnt, &hbytes);
75a872f8
DRJ
1439 if (error)
1440 goto done;
1441 auio.uio_iov = iov;
1442 auio.uio_iovcnt = hdtr.hdr_cnt;
1443 auio.uio_offset = 0;
1444 auio.uio_segflg = UIO_USERSPACE;
1445 auio.uio_rw = UIO_WRITE;
1446 auio.uio_td = td;
30eeba44 1447 auio.uio_resid = hbytes;
75a872f8 1448
e12241e1 1449 mheader = m_uiomove(&auio);
75a872f8
DRJ
1450
1451 iovec_free(&iov, aiov);
30eeba44 1452 if (mheader == NULL)
75a872f8 1453 goto done;
75a872f8
DRJ
1454 }
1455 }
1456
30eeba44 1457 error = kern_sendfile(vp, uap->s, uap->offset, uap->nbytes, mheader,
75a872f8
DRJ
1458 &sbytes, uap->flags);
1459 if (error)
1460 goto done;
1461
1462 /*
1463 * Send trailers. Wimp out and use writev(2).
1464 */
1465 if (uap->hdtr != NULL && hdtr.trailers != NULL) {
1466 error = iovec_copyin(hdtr.trailers, &iov, aiov,
ef5c76d7 1467 hdtr.trl_cnt, &auio.uio_resid);
75a872f8
DRJ
1468 if (error)
1469 goto done;
1470 auio.uio_iov = iov;
1471 auio.uio_iovcnt = hdtr.trl_cnt;
1472 auio.uio_offset = 0;
1473 auio.uio_segflg = UIO_USERSPACE;
1474 auio.uio_rw = UIO_WRITE;
1475 auio.uio_td = td;
1476
30eeba44 1477 error = kern_sendmsg(uap->s, NULL, &auio, NULL, 0, &tbytes);
75a872f8
DRJ
1478
1479 iovec_free(&iov, aiov);
1480 if (error)
1481 goto done;
30eeba44 1482 hdtr_size += tbytes; /* trailer bytes successfully sent */
75a872f8
DRJ
1483 }
1484
1485done:
1486 if (uap->sbytes != NULL) {
1487 sbytes += hdtr_size;
1488 copyout(&sbytes, uap->sbytes, sizeof(off_t));
1489 }
1490 if (vp)
1491 vrele(vp);
1492 return (error);
1493}
1494
1495int
06ecca5a 1496kern_sendfile(struct vnode *vp, int sfd, off_t offset, size_t nbytes,
30eeba44 1497 struct mbuf *mheader, off_t *sbytes, int flags)
75a872f8
DRJ
1498{
1499 struct thread *td = curthread;
1500 struct proc *p = td->td_proc;
1501 struct vm_object *obj;
1502 struct socket *so;
285332f0 1503 struct file *fp;
75a872f8
DRJ
1504 struct mbuf *m;
1505 struct sf_buf *sf;
e66bab2b 1506 struct sfbuf_mref *sfm;
75a872f8
DRJ
1507 struct vm_page *pg;
1508 off_t off, xfsize;
39b3370f 1509 off_t hbytes = 0;
75a872f8
DRJ
1510 int error = 0;
1511
7540ab49
MD
1512 if (vp->v_type != VREG) {
1513 error = EINVAL;
1514 goto done0;
1515 }
1516 if ((obj = vp->v_object) == NULL) {
984263bc 1517 error = EINVAL;
285332f0 1518 goto done0;
984263bc 1519 }
06ecca5a 1520 error = holdsock(p->p_fd, sfd, &fp);
984263bc 1521 if (error)
285332f0 1522 goto done0;
984263bc
MD
1523 so = (struct socket *)fp->f_data;
1524 if (so->so_type != SOCK_STREAM) {
1525 error = EINVAL;
1526 goto done;
1527 }
1528 if ((so->so_state & SS_ISCONNECTED) == 0) {
1529 error = ENOTCONN;
1530 goto done;
1531 }
75a872f8 1532 if (offset < 0) {
984263bc
MD
1533 error = EINVAL;
1534 goto done;
1535 }
1536
75a872f8 1537 *sbytes = 0;
984263bc
MD
1538 /*
1539 * Protect against multiple writers to the socket.
1540 */
6d49aa6f 1541 ssb_lock(&so->so_snd, M_WAITOK);
984263bc
MD
1542
1543 /*
1544 * Loop through the pages in the file, starting with the requested
1545 * offset. Get a file page (do I/O if necessary), map the file page
1546 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1547 * it on the socket.
1548 */
39b3370f 1549 for (off = offset; ; off += xfsize, *sbytes += xfsize + hbytes) {
984263bc
MD
1550 vm_pindex_t pindex;
1551 vm_offset_t pgoff;
1552
1553 pindex = OFF_TO_IDX(off);
1554retry_lookup:
1555 /*
1556 * Calculate the amount to transfer. Not to exceed a page,
1557 * the EOF, or the passed in nbytes.
1558 */
57f7b636 1559 xfsize = vp->v_filesize - off;
984263bc
MD
1560 if (xfsize > PAGE_SIZE)
1561 xfsize = PAGE_SIZE;
1562 pgoff = (vm_offset_t)(off & PAGE_MASK);
1563 if (PAGE_SIZE - pgoff < xfsize)
1564 xfsize = PAGE_SIZE - pgoff;
75a872f8
DRJ
1565 if (nbytes && xfsize > (nbytes - *sbytes))
1566 xfsize = nbytes - *sbytes;
984263bc
MD
1567 if (xfsize <= 0)
1568 break;
1569 /*
1570 * Optimize the non-blocking case by looking at the socket space
1571 * before going to the extra work of constituting the sf_buf.
1572 */
6d49aa6f 1573 if ((fp->f_flag & FNONBLOCK) && ssb_space(&so->so_snd) <= 0) {
984263bc
MD
1574 if (so->so_state & SS_CANTSENDMORE)
1575 error = EPIPE;
1576 else
1577 error = EAGAIN;
6d49aa6f 1578 ssb_unlock(&so->so_snd);
984263bc
MD
1579 goto done;
1580 }
1581 /*
1582 * Attempt to look up the page.
1583 *
06ecca5a 1584 * Allocate if not found, wait and loop if busy, then
5fd012e0
MD
1585 * wire the page. critical section protection is
1586 * required to maintain the object association (an
1587 * interrupt can free the page) through to the
1588 * vm_page_wire() call.
984263bc 1589 */
5fd012e0 1590 crit_enter();
984263bc 1591 pg = vm_page_lookup(obj, pindex);
984263bc
MD
1592 if (pg == NULL) {
1593 pg = vm_page_alloc(obj, pindex, VM_ALLOC_NORMAL);
1594 if (pg == NULL) {
4ecf7cc9 1595 vm_wait(0);
5fd012e0 1596 crit_exit();
984263bc
MD
1597 goto retry_lookup;
1598 }
1599 vm_page_wakeup(pg);
1600 } else if (vm_page_sleep_busy(pg, TRUE, "sfpbsy")) {
5fd012e0 1601 crit_exit();
984263bc
MD
1602 goto retry_lookup;
1603 }
984263bc 1604 vm_page_wire(pg);
5fd012e0 1605 crit_exit();
984263bc
MD
1606
1607 /*
1608 * If page is not valid for what we need, initiate I/O
1609 */
1610
1611 if (!pg->valid || !vm_page_is_valid(pg, pgoff, xfsize)) {
1612 struct uio auio;
1613 struct iovec aiov;
1614 int bsize;
1615
1616 /*
1617 * Ensure that our page is still around when the I/O
1618 * completes.
1619 */
1620 vm_page_io_start(pg);
1621
1622 /*
1623 * Get the page from backing store.
1624 */
1625 bsize = vp->v_mount->mnt_stat.f_iosize;
1626 auio.uio_iov = &aiov;
1627 auio.uio_iovcnt = 1;
1628 aiov.iov_base = 0;
1629 aiov.iov_len = MAXBSIZE;
1630 auio.uio_resid = MAXBSIZE;
1631 auio.uio_offset = trunc_page(off);
1632 auio.uio_segflg = UIO_NOCOPY;
1633 auio.uio_rw = UIO_READ;
dadab5e9 1634 auio.uio_td = td;
ab6f251b 1635 vn_lock(vp, LK_SHARED | LK_RETRY);
dadab5e9
MD
1636 error = VOP_READ(vp, &auio,
1637 IO_VMIO | ((MAXBSIZE / bsize) << 16),
1638 p->p_ucred);
a11aaa81 1639 vn_unlock(vp);
984263bc
MD
1640 vm_page_flag_clear(pg, PG_ZERO);
1641 vm_page_io_finish(pg);
1642 if (error) {
f2555cdd 1643 crit_enter();
984263bc 1644 vm_page_unwire(pg, 0);
f2555cdd
MD
1645 vm_page_try_to_free(pg);
1646 crit_exit();
6d49aa6f 1647 ssb_unlock(&so->so_snd);
984263bc
MD
1648 goto done;
1649 }
1650 }
1651
1652
1653 /*
1654 * Get a sendfile buf. We usually wait as long as necessary,
1655 * but this wait can be interrupted.
1656 */
4f1640d6 1657 if ((sf = sf_buf_alloc(pg, SFB_CATCH)) == NULL) {
5fd012e0 1658 crit_enter();
984263bc 1659 vm_page_unwire(pg, 0);
f2555cdd 1660 vm_page_try_to_free(pg);
5fd012e0 1661 crit_exit();
6d49aa6f 1662 ssb_unlock(&so->so_snd);
984263bc
MD
1663 error = EINTR;
1664 goto done;
1665 }
1666
984263bc
MD
1667 /*
1668 * Get an mbuf header and set it up as having external storage.
1669 */
74f1caca 1670 MGETHDR(m, MB_WAIT, MT_DATA);
984263bc
MD
1671 if (m == NULL) {
1672 error = ENOBUFS;
b4caac98 1673 sf_buf_free(sf);
6d49aa6f 1674 ssb_unlock(&so->so_snd);
984263bc
MD
1675 goto done;
1676 }
e66bab2b
MD
1677
1678 /*
1679 * sfm is a temporary hack, use a per-cpu cache for this.
1680 */
efda3bd0 1681 sfm = kmalloc(sizeof(struct sfbuf_mref), M_SENDFILE, M_WAITOK);
e66bab2b
MD
1682 sfm->sf = sf;
1683 sfm->mref_count = 1;
1684
b542cd49
JS
1685 m->m_ext.ext_free = sf_buf_mfree;
1686 m->m_ext.ext_ref = sf_buf_mref;
e66bab2b 1687 m->m_ext.ext_arg = sfm;
984263bc
MD
1688 m->m_ext.ext_buf = (void *)sf->kva;
1689 m->m_ext.ext_size = PAGE_SIZE;
1690 m->m_data = (char *) sf->kva + pgoff;
013a4c0e 1691 m->m_flags |= M_EXT;
984263bc 1692 m->m_pkthdr.len = m->m_len = xfsize;
b542cd49 1693 KKASSERT((m->m_flags & (M_EXT_CLUSTER)) == 0);
30eeba44 1694
39b3370f
JH
1695 if (mheader != NULL) {
1696 hbytes = mheader->m_pkthdr.len;
30eeba44
JH
1697 mheader->m_pkthdr.len += m->m_pkthdr.len;
1698 m_cat(mheader, m);
1699 m = mheader;
1700 mheader = NULL;
39b3370f
JH
1701 } else
1702 hbytes = 0;
30eeba44 1703
984263bc
MD
1704 /*
1705 * Add the buffer to the socket buffer chain.
1706 */
5fd012e0 1707 crit_enter();
984263bc
MD
1708retry_space:
1709 /*
1710 * Make sure that the socket is still able to take more data.
1711 * CANTSENDMORE being true usually means that the connection
1712 * was closed. so_error is true when an error was sensed after
1713 * a previous send.
1714 * The state is checked after the page mapping and buffer
1715 * allocation above since those operations may block and make
1716 * any socket checks stale. From this point forward, nothing
1717 * blocks before the pru_send (or more accurately, any blocking
1718 * results in a loop back to here to re-check).
1719 */
1720 if ((so->so_state & SS_CANTSENDMORE) || so->so_error) {
1721 if (so->so_state & SS_CANTSENDMORE) {
1722 error = EPIPE;
1723 } else {
1724 error = so->so_error;
1725 so->so_error = 0;
1726 }
1727 m_freem(m);
6d49aa6f 1728 ssb_unlock(&so->so_snd);
5fd012e0 1729 crit_exit();
984263bc
MD
1730 goto done;
1731 }
1732 /*
1733 * Wait for socket space to become available. We do this just
1734 * after checking the connection state above in order to avoid
6d49aa6f 1735 * a race condition with ssb_wait().
984263bc 1736 */
6d49aa6f 1737 if (ssb_space(&so->so_snd) < so->so_snd.ssb_lowat) {
9ba76b73 1738 if (fp->f_flag & FNONBLOCK) {
984263bc 1739 m_freem(m);
6d49aa6f 1740 ssb_unlock(&so->so_snd);
5fd012e0 1741 crit_exit();
984263bc
MD
1742 error = EAGAIN;
1743 goto done;
1744 }
6d49aa6f 1745 error = ssb_wait(&so->so_snd);
984263bc 1746 /*
6d49aa6f 1747 * An error from ssb_wait usually indicates that we've
984263bc
MD
1748 * been interrupted by a signal. If we've sent anything
1749 * then return bytes sent, otherwise return the error.
1750 */
1751 if (error) {
1752 m_freem(m);
6d49aa6f 1753 ssb_unlock(&so->so_snd);
5fd012e0 1754 crit_exit();
984263bc
MD
1755 goto done;
1756 }
1757 goto retry_space;
1758 }
3c6b2883 1759 error = so_pru_send(so, 0, m, NULL, NULL, td);
5fd012e0 1760 crit_exit();
984263bc 1761 if (error) {
6d49aa6f 1762 ssb_unlock(&so->so_snd);
984263bc
MD
1763 goto done;
1764 }
1765 }
d785f69d 1766 if (mheader != NULL) {
39b3370f 1767 *sbytes += mheader->m_pkthdr.len;
d785f69d
JH
1768 error = so_pru_send(so, 0, mheader, NULL, NULL, td);
1769 mheader = NULL;
1770 }
6d49aa6f 1771 ssb_unlock(&so->so_snd);
984263bc 1772
984263bc 1773done:
9f87144f 1774 fdrop(fp);
285332f0 1775done0:
30eeba44
JH
1776 if (mheader != NULL)
1777 m_freem(mheader);
984263bc
MD
1778 return (error);
1779}
78812139
EN
1780
1781int
753fd850 1782sys_sctp_peeloff(struct sctp_peeloff_args *uap)
78812139
EN
1783{
1784#ifdef SCTP
1785 struct thread *td = curthread;
1786 struct proc *p = td->td_proc;
78812139
EN
1787 struct file *lfp = NULL;
1788 struct file *nfp = NULL;
1789 int error;
1790 struct socket *head, *so;
1791 caddr_t assoc_id;
1792 int fd;
1793 short fflag; /* type must match fp->f_flag */
1794
1795 assoc_id = uap->name;
fa541be6 1796 error = holdsock(p->p_fd, uap->sd, &lfp);
78812139
EN
1797 if (error) {
1798 return (error);
1799 }
1800 crit_enter();
1801 head = (struct socket *)lfp->f_data;
1802 error = sctp_can_peel_off(head, assoc_id);
1803 if (error) {
1804 crit_exit();
1805 goto done;
1806 }
1807 /*
1808 * At this point we know we do have a assoc to pull
1809 * we proceed to get the fd setup. This may block
1810 * but that is ok.
1811 */
1812
1813 fflag = lfp->f_flag;
1814 error = falloc(p, &nfp, &fd);
1815 if (error) {
1816 /*
1817 * Probably ran out of file descriptors. Put the
1818 * unaccepted connection back onto the queue and
1819 * do another wakeup so some other process might
1820 * have a chance at it.
1821 */
1822 crit_exit();
1823 goto done;
1824 }
e54488bb 1825 uap->sysmsg_iresult = fd;
78812139
EN
1826
1827 so = sctp_get_peeloff(head, assoc_id, &error);
1828 if (so == NULL) {
1829 /*
1830 * Either someone else peeled it off OR
1831 * we can't get a socket.
1832 */
1833 goto noconnection;
1834 }
1835 so->so_state &= ~SS_COMP;
1836 so->so_state &= ~SS_NOFDREF;
1837 so->so_head = NULL;
1838 if (head->so_sigio != NULL)
1839 fsetown(fgetown(head->so_sigio), &so->so_sigio);
1840
fbb4eeab 1841 nfp->f_type = DTYPE_SOCKET;
78812139
EN
1842 nfp->f_flag = fflag;
1843 nfp->f_ops = &socketops;
fbb4eeab 1844 nfp->f_data = so;
78812139
EN
1845
1846noconnection:
1847 /*
259b8ea0
MD
1848 * Assign the file pointer to the reserved descriptor, or clear
1849 * the reserved descriptor if an error occured.
78812139 1850 */
fa541be6 1851 if (error)
259b8ea0
MD
1852 fsetfd(p, NULL, fd);
1853 else
1854 fsetfd(p, nfp, fd);
78812139
EN
1855 crit_exit();
1856 /*
1857 * Release explicitly held references before returning.
1858 */
1859done:
1860 if (nfp != NULL)
9f87144f
MD
1861 fdrop(nfp);
1862 fdrop(lfp);
78812139
EN
1863 return (error);
1864#else /* SCTP */
1865 return(EOPNOTSUPP);
1866#endif /* SCTP */
1867}