kernel - use new td_ucred in numerous places
[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
3919ced0
MD
133/*
134 * MPALMOSTSAFE
135 */
02844a31 136int
753fd850 137sys_socket(struct socket_args *uap)
75a872f8
DRJ
138{
139 int error;
140
3919ced0 141 get_mplock();
75a872f8 142 error = kern_socket(uap->domain, uap->type, uap->protocol,
e54488bb 143 &uap->sysmsg_iresult);
3919ced0 144 rel_mplock();
75a872f8
DRJ
145
146 return (error);
147}
6b6e0885 148
75a872f8 149int
5969a6f1 150kern_bind(int s, struct sockaddr *sa)
984263bc 151{
dadab5e9
MD
152 struct thread *td = curthread;
153 struct proc *p = td->td_proc;
984263bc 154 struct file *fp;
984263bc
MD
155 int error;
156
dadab5e9 157 KKASSERT(p);
d83b97b9 158 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
159 if (error)
160 return (error);
d83b97b9 161 error = sobind((struct socket *)fp->f_data, sa, td);
9f87144f 162 fdrop(fp);
d83b97b9
MD
163 return (error);
164}
165
166/*
167 * bind_args(int s, caddr_t name, int namelen)
3919ced0
MD
168 *
169 * MPALMOSTSAFE
d83b97b9
MD
170 */
171int
753fd850 172sys_bind(struct bind_args *uap)
d83b97b9
MD
173{
174 struct sockaddr *sa;
175 int error;
176
984263bc 177 error = getsockaddr(&sa, uap->name, uap->namelen);
d83b97b9 178 if (error)
984263bc 179 return (error);
3919ced0 180 get_mplock();
5969a6f1 181 error = kern_bind(uap->s, sa);
3919ced0 182 rel_mplock();
984263bc 183 FREE(sa, M_SONAME);
d83b97b9 184
984263bc
MD
185 return (error);
186}
187
984263bc 188int
5969a6f1 189kern_listen(int s, int backlog)
984263bc 190{
dadab5e9
MD
191 struct thread *td = curthread;
192 struct proc *p = td->td_proc;
984263bc
MD
193 struct file *fp;
194 int error;
195
dadab5e9 196 KKASSERT(p);
5969a6f1 197 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
198 if (error)
199 return (error);
5969a6f1 200 error = solisten((struct socket *)fp->f_data, backlog, td);
9f87144f 201 fdrop(fp);
984263bc
MD
202 return(error);
203}
204
41c20dac 205/*
5969a6f1 206 * listen_args(int s, int backlog)
3919ced0
MD
207 *
208 * MPALMOSTSAFE
5969a6f1
DRJ
209 */
210int
753fd850 211sys_listen(struct listen_args *uap)
5969a6f1
DRJ
212{
213 int error;
214
3919ced0 215 get_mplock();
5969a6f1 216 error = kern_listen(uap->s, uap->backlog);
3919ced0 217 rel_mplock();
5969a6f1
DRJ
218 return (error);
219}
220
221/*
f172717f
JH
222 * Returns the accepted socket as well.
223 */
224static boolean_t
225soaccept_predicate(struct netmsg *msg0)
226{
227 struct netmsg_so_notify *msg = (struct netmsg_so_notify *)msg0;
228 struct socket *head = msg->nm_so;
229
230 if (head->so_error != 0) {
4599cf19 231 msg->nm_netmsg.nm_lmsg.ms_error = head->so_error;
f172717f
JH
232 return (TRUE);
233 }
234 if (!TAILQ_EMPTY(&head->so_comp)) {
235 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
236 msg->nm_so = TAILQ_FIRST(&head->so_comp);
237 TAILQ_REMOVE(&head->so_comp, msg->nm_so, so_list);
238 head->so_qlen--;
239
4599cf19 240 msg->nm_netmsg.nm_lmsg.ms_error = 0;
f172717f
JH
241 return (TRUE);
242 }
243 if (head->so_state & SS_CANTRCVMORE) {
4599cf19 244 msg->nm_netmsg.nm_lmsg.ms_error = ECONNABORTED;
f172717f
JH
245 return (TRUE);
246 }
9ba76b73 247 if (msg->nm_fflags & FNONBLOCK) {
4599cf19 248 msg->nm_netmsg.nm_lmsg.ms_error = EWOULDBLOCK;
f172717f
JH
249 return (TRUE);
250 }
251
252 return (FALSE);
253}
254
255/*
5969a6f1
DRJ
256 * The second argument to kern_accept() is a handle to a struct sockaddr.
257 * This allows kern_accept() to return a pointer to an allocated struct
d83b97b9
MD
258 * sockaddr which must be freed later with FREE(). The caller must
259 * initialize *name to NULL.
41c20dac 260 */
02844a31 261int
358e1f78 262kern_accept(int s, int fflags, struct sockaddr **name, int *namelen, int *res)
984263bc 263{
dadab5e9
MD
264 struct thread *td = curthread;
265 struct proc *p = td->td_proc;
984263bc
MD
266 struct file *lfp = NULL;
267 struct file *nfp = NULL;
268 struct sockaddr *sa;
984263bc 269 struct socket *head, *so;
f172717f 270 struct netmsg_so_notify msg;
984263bc
MD
271 int fd;
272 u_int fflag; /* type must match fp->f_flag */
f172717f 273 int error, tmp;
984263bc 274
259b8ea0 275 *res = -1;
d83b97b9
MD
276 if (name && namelen && *namelen < 0)
277 return (EINVAL);
278
fa541be6 279 error = holdsock(p->p_fd, s, &lfp);
984263bc
MD
280 if (error)
281 return (error);
f172717f
JH
282
283 error = falloc(p, &nfp, &fd);
284 if (error) { /* Probably ran out of file descriptors. */
9f87144f 285 fdrop(lfp);
f172717f
JH
286 return (error);
287 }
984263bc
MD
288 head = (struct socket *)lfp->f_data;
289 if ((head->so_options & SO_ACCEPTCONN) == 0) {
984263bc
MD
290 error = EINVAL;
291 goto done;
292 }
f172717f 293
358e1f78
MD
294 if (fflags & O_FBLOCKING)
295 fflags |= lfp->f_flag & ~FNONBLOCK;
296 else if (fflags & O_FNONBLOCKING)
297 fflags |= lfp->f_flag | FNONBLOCK;
298 else
299 fflags = lfp->f_flag;
300
f172717f 301 /* optimize for uniprocessor case later XXX JH */
48e7b118
MD
302 netmsg_init_abortable(&msg.nm_netmsg, head, &curthread->td_msgport,
303 0, netmsg_so_notify, netmsg_so_notify_doabort);
f172717f 304 msg.nm_predicate = soaccept_predicate;
358e1f78 305 msg.nm_fflags = fflags;
f172717f
JH
306 msg.nm_so = head;
307 msg.nm_etype = NM_REVENT;
48e7b118 308 error = lwkt_domsg(head->so_port, &msg.nm_netmsg.nm_lmsg, PCATCH);
f172717f 309 if (error)
984263bc 310 goto done;
984263bc
MD
311
312 /*
f172717f 313 * At this point we have the connection that's ready to be accepted.
984263bc 314 */
f172717f 315 so = msg.nm_so;
984263bc
MD
316
317 fflag = lfp->f_flag;
984263bc
MD
318
319 /* connection has been removed from the listen queue */
6d49aa6f 320 KNOTE(&head->so_rcv.ssb_sel.si_note, 0);
984263bc
MD
321
322 so->so_state &= ~SS_COMP;
323 so->so_head = NULL;
324 if (head->so_sigio != NULL)
325 fsetown(fgetown(head->so_sigio), &so->so_sigio);
326
fbb4eeab 327 nfp->f_type = DTYPE_SOCKET;
984263bc
MD
328 nfp->f_flag = fflag;
329 nfp->f_ops = &socketops;
fbb4eeab 330 nfp->f_data = so;
984263bc
MD
331 /* Sync socket nonblocking/async state with file flags */
332 tmp = fflag & FNONBLOCK;
9910d07b 333 fo_ioctl(nfp, FIONBIO, (caddr_t)&tmp, td->td_ucred, NULL);
984263bc 334 tmp = fflag & FASYNC;
9910d07b 335 fo_ioctl(nfp, FIOASYNC, (caddr_t)&tmp, td->td_ucred, NULL);
d83b97b9
MD
336
337 sa = NULL;
984263bc 338 error = soaccept(so, &sa);
d83b97b9
MD
339
340 /*
341 * Set the returned name and namelen as applicable. Set the returned
342 * namelen to 0 for older code which might ignore the return value
343 * from accept.
344 */
345 if (error == 0) {
346 if (sa && name && namelen) {
347 if (*namelen > sa->sa_len)
348 *namelen = sa->sa_len;
349 *name = sa;
350 } else {
351 if (sa)
352 FREE(sa, M_SONAME);
984263bc 353 }
984263bc 354 }
984263bc 355
f172717f 356done:
984263bc 357 /*
259b8ea0
MD
358 * If an error occured clear the reserved descriptor, else associate
359 * nfp with it.
360 *
361 * Note that *res is normally ignored if an error is returned but
362 * a syscall message will still have access to the result code.
984263bc
MD
363 */
364 if (error) {
259b8ea0
MD
365 fsetfd(p, NULL, fd);
366 } else {
367 *res = fd;
368 fsetfd(p, nfp, fd);
984263bc 369 }
259b8ea0 370 fdrop(nfp);
9f87144f 371 fdrop(lfp);
984263bc
MD
372 return (error);
373}
374
d83b97b9 375/*
358e1f78 376 * accept(int s, caddr_t name, int *anamelen)
3919ced0
MD
377 *
378 * MPALMOSTSAFE
d83b97b9 379 */
984263bc 380int
753fd850 381sys_accept(struct accept_args *uap)
984263bc 382{
d83b97b9
MD
383 struct sockaddr *sa = NULL;
384 int sa_len;
385 int error;
386
387 if (uap->name) {
388 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
389 if (error)
390 return (error);
391
3919ced0 392 get_mplock();
e54488bb
MD
393 error = kern_accept(uap->s, 0, &sa, &sa_len,
394 &uap->sysmsg_iresult);
3919ced0 395 rel_mplock();
358e1f78
MD
396
397 if (error == 0)
398 error = copyout(sa, uap->name, sa_len);
399 if (error == 0) {
400 error = copyout(&sa_len, uap->anamelen,
401 sizeof(*uap->anamelen));
402 }
403 if (sa)
404 FREE(sa, M_SONAME);
405 } else {
3919ced0 406 get_mplock();
e54488bb
MD
407 error = kern_accept(uap->s, 0, NULL, 0,
408 &uap->sysmsg_iresult);
3919ced0 409 rel_mplock();
358e1f78
MD
410 }
411 return (error);
412}
413
414/*
b09fd398 415 * extaccept(int s, int fflags, caddr_t name, int *anamelen)
3919ced0
MD
416 *
417 * MPALMOSTSAFE
358e1f78
MD
418 */
419int
b09fd398 420sys_extaccept(struct extaccept_args *uap)
358e1f78
MD
421{
422 struct sockaddr *sa = NULL;
423 int sa_len;
424 int error;
425 int fflags = uap->flags & O_FMASK;
426
427 if (uap->name) {
428 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
429 if (error)
430 return (error);
431
3919ced0 432 get_mplock();
e54488bb
MD
433 error = kern_accept(uap->s, fflags, &sa, &sa_len,
434 &uap->sysmsg_iresult);
3919ced0 435 rel_mplock();
d83b97b9
MD
436
437 if (error == 0)
438 error = copyout(sa, uap->name, sa_len);
439 if (error == 0) {
440 error = copyout(&sa_len, uap->anamelen,
441 sizeof(*uap->anamelen));
442 }
443 if (sa)
444 FREE(sa, M_SONAME);
445 } else {
3919ced0 446 get_mplock();
e54488bb
MD
447 error = kern_accept(uap->s, fflags, NULL, 0,
448 &uap->sysmsg_iresult);
3919ced0 449 rel_mplock();
d83b97b9
MD
450 }
451 return (error);
984263bc
MD
452}
453
358e1f78 454
b44419cb
MD
455/*
456 * Returns TRUE if predicate satisfied.
457 */
458static boolean_t
459soconnected_predicate(struct netmsg *msg0)
460{
461 struct netmsg_so_notify *msg = (struct netmsg_so_notify *)msg0;
462 struct socket *so = msg->nm_so;
463
464 /* check predicate */
465 if (!(so->so_state & SS_ISCONNECTING) || so->so_error != 0) {
4599cf19 466 msg->nm_netmsg.nm_lmsg.ms_error = so->so_error;
b44419cb
MD
467 return (TRUE);
468 }
469
470 return (FALSE);
471}
472
02844a31 473int
358e1f78 474kern_connect(int s, int fflags, struct sockaddr *sa)
984263bc 475{
dadab5e9
MD
476 struct thread *td = curthread;
477 struct proc *p = td->td_proc;
984263bc 478 struct file *fp;
dadab5e9 479 struct socket *so;
8765eadc 480 int error, interrupted = 0;
984263bc 481
d83b97b9 482 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
483 if (error)
484 return (error);
485 so = (struct socket *)fp->f_data;
358e1f78
MD
486
487 if (fflags & O_FBLOCKING)
488 /* fflags &= ~FNONBLOCK; */;
489 else if (fflags & O_FNONBLOCKING)
490 fflags |= FNONBLOCK;
491 else
492 fflags = fp->f_flag;
493
8765eadc 494 if (so->so_state & SS_ISCONNECTING) {
984263bc
MD
495 error = EALREADY;
496 goto done;
497 }
dadab5e9 498 error = soconnect(so, sa, td);
984263bc
MD
499 if (error)
500 goto bad;
358e1f78 501 if ((fflags & FNONBLOCK) && (so->so_state & SS_ISCONNECTING)) {
984263bc
MD
502 error = EINPROGRESS;
503 goto done;
504 }
b44419cb
MD
505 if ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
506 struct netmsg_so_notify msg;
b44419cb 507
48e7b118 508 netmsg_init_abortable(&msg.nm_netmsg, so,
4599cf19 509 &curthread->td_msgport,
a22c590e 510 0,
4599cf19
MD
511 netmsg_so_notify,
512 netmsg_so_notify_doabort);
b44419cb
MD
513 msg.nm_predicate = soconnected_predicate;
514 msg.nm_so = so;
515 msg.nm_etype = NM_REVENT;
48e7b118 516 error = lwkt_domsg(so->so_port, &msg.nm_netmsg.nm_lmsg, PCATCH);
8765eadc
SZ
517 if (error == EINTR || error == ERESTART)
518 interrupted = 1;
984263bc
MD
519 }
520 if (error == 0) {
521 error = so->so_error;
522 so->so_error = 0;
523 }
984263bc 524bad:
8765eadc
SZ
525 if (!interrupted)
526 so->so_state &= ~SS_ISCONNECTING;
984263bc
MD
527 if (error == ERESTART)
528 error = EINTR;
529done:
9f87144f 530 fdrop(fp);
984263bc
MD
531 return (error);
532}
533
41c20dac 534/*
d83b97b9 535 * connect_args(int s, caddr_t name, int namelen)
3919ced0
MD
536 *
537 * MPALMOSTSAFE
d83b97b9
MD
538 */
539int
753fd850 540sys_connect(struct connect_args *uap)
d83b97b9
MD
541{
542 struct sockaddr *sa;
543 int error;
544
545 error = getsockaddr(&sa, uap->name, uap->namelen);
546 if (error)
547 return (error);
3919ced0 548 get_mplock();
358e1f78 549 error = kern_connect(uap->s, 0, sa);
3919ced0 550 rel_mplock();
358e1f78
MD
551 FREE(sa, M_SONAME);
552
553 return (error);
554}
555
556/*
557 * connect_args(int s, int fflags, caddr_t name, int namelen)
3919ced0
MD
558 *
559 * MPALMOSTSAFE
358e1f78
MD
560 */
561int
b09fd398 562sys_extconnect(struct extconnect_args *uap)
358e1f78
MD
563{
564 struct sockaddr *sa;
565 int error;
566 int fflags = uap->flags & O_FMASK;
567
568 error = getsockaddr(&sa, uap->name, uap->namelen);
569 if (error)
570 return (error);
3919ced0 571 get_mplock();
358e1f78 572 error = kern_connect(uap->s, fflags, sa);
3919ced0 573 rel_mplock();
d83b97b9
MD
574 FREE(sa, M_SONAME);
575
576 return (error);
577}
578
984263bc 579int
5969a6f1 580kern_socketpair(int domain, int type, int protocol, int *sv)
984263bc 581{
dadab5e9
MD
582 struct thread *td = curthread;
583 struct proc *p = td->td_proc;
984263bc
MD
584 struct file *fp1, *fp2;
585 struct socket *so1, *so2;
259b8ea0 586 int fd1, fd2, error;
984263bc 587
dadab5e9 588 KKASSERT(p);
5969a6f1 589 error = socreate(domain, &so1, type, protocol, td);
984263bc
MD
590 if (error)
591 return (error);
5969a6f1 592 error = socreate(domain, &so2, type, protocol, td);
984263bc
MD
593 if (error)
594 goto free1;
259b8ea0 595 error = falloc(p, &fp1, &fd1);
984263bc
MD
596 if (error)
597 goto free2;
259b8ea0 598 sv[0] = fd1;
fbb4eeab 599 fp1->f_data = so1;
259b8ea0 600 error = falloc(p, &fp2, &fd2);
984263bc
MD
601 if (error)
602 goto free3;
fbb4eeab 603 fp2->f_data = so2;
259b8ea0 604 sv[1] = fd2;
984263bc
MD
605 error = soconnect2(so1, so2);
606 if (error)
607 goto free4;
5969a6f1 608 if (type == SOCK_DGRAM) {
984263bc
MD
609 /*
610 * Datagram socket connection is asymmetric.
611 */
612 error = soconnect2(so2, so1);
613 if (error)
614 goto free4;
615 }
fbb4eeab 616 fp1->f_type = fp2->f_type = DTYPE_SOCKET;
984263bc
MD
617 fp1->f_flag = fp2->f_flag = FREAD|FWRITE;
618 fp1->f_ops = fp2->f_ops = &socketops;
259b8ea0
MD
619 fsetfd(p, fp1, fd1);
620 fsetfd(p, fp2, fd2);
9f87144f
MD
621 fdrop(fp1);
622 fdrop(fp2);
984263bc
MD
623 return (error);
624free4:
259b8ea0 625 fsetfd(p, NULL, fd2);
9f87144f 626 fdrop(fp2);
984263bc 627free3:
259b8ea0 628 fsetfd(p, NULL, fd1);
9f87144f 629 fdrop(fp1);
984263bc 630free2:
9ba76b73 631 (void)soclose(so2, 0);
984263bc 632free1:
9ba76b73 633 (void)soclose(so1, 0);
984263bc
MD
634 return (error);
635}
636
5969a6f1
DRJ
637/*
638 * socketpair(int domain, int type, int protocol, int *rsv)
3919ced0
MD
639 *
640 * MPALMOSTSAFE
5969a6f1
DRJ
641 */
642int
753fd850 643sys_socketpair(struct socketpair_args *uap)
5969a6f1
DRJ
644{
645 int error, sockv[2];
646
3919ced0 647 get_mplock();
5969a6f1 648 error = kern_socketpair(uap->domain, uap->type, uap->protocol, sockv);
3919ced0 649 rel_mplock();
5969a6f1
DRJ
650
651 if (error == 0)
652 error = copyout(sockv, uap->rsv, sizeof(sockv));
653 return (error);
654}
655
35fbb1d9 656int
3e1837ce 657kern_sendmsg(int s, struct sockaddr *sa, struct uio *auio,
e54488bb 658 struct mbuf *control, int flags, size_t *res)
984263bc 659{
dadab5e9 660 struct thread *td = curthread;
7278a846 661 struct lwp *lp = td->td_lwp;
dadab5e9 662 struct proc *p = td->td_proc;
984263bc 663 struct file *fp;
e54488bb
MD
664 size_t len;
665 int error;
984263bc
MD
666 struct socket *so;
667#ifdef KTRACE
668 struct iovec *ktriov = NULL;
669 struct uio ktruio;
670#endif
671
672 error = holdsock(p->p_fd, s, &fp);
673 if (error)
674 return (error);
984263bc 675#ifdef KTRACE
dadab5e9 676 if (KTRPOINT(td, KTR_GENIO)) {
3e1837ce 677 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
984263bc
MD
678
679 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
3e1837ce
DRJ
680 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
681 ktruio = *auio;
984263bc
MD
682 }
683#endif
3e1837ce 684 len = auio->uio_resid;
984263bc 685 so = (struct socket *)fp->f_data;
9ba76b73
MD
686 if ((flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
687 if (fp->f_flag & FNONBLOCK)
688 flags |= MSG_FNONBLOCKING;
689 }
6b6e0885 690 error = so_pru_sosend(so, sa, auio, NULL, control, flags, td);
984263bc 691 if (error) {
3e1837ce 692 if (auio->uio_resid != len && (error == ERESTART ||
984263bc
MD
693 error == EINTR || error == EWOULDBLOCK))
694 error = 0;
695 if (error == EPIPE)
7278a846 696 lwpsignal(p, lp, SIGPIPE);
984263bc 697 }
984263bc
MD
698#ifdef KTRACE
699 if (ktriov != NULL) {
700 if (error == 0) {
701 ktruio.uio_iov = ktriov;
3e1837ce 702 ktruio.uio_resid = len - auio->uio_resid;
9fb04d14 703 ktrgenio(lp, s, UIO_WRITE, &ktruio, error);
984263bc
MD
704 }
705 FREE(ktriov, M_TEMP);
706 }
707#endif
35fbb1d9 708 if (error == 0)
3e1837ce 709 *res = len - auio->uio_resid;
9f87144f 710 fdrop(fp);
984263bc
MD
711 return (error);
712}
713
41c20dac
MD
714/*
715 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
3919ced0
MD
716 *
717 * MPALMOSTSAFE
41c20dac 718 */
984263bc 719int
753fd850 720sys_sendto(struct sendto_args *uap)
984263bc 721{
3e1837ce
DRJ
722 struct thread *td = curthread;
723 struct uio auio;
984263bc 724 struct iovec aiov;
35fbb1d9
DRJ
725 struct sockaddr *sa = NULL;
726 int error;
984263bc 727
35fbb1d9
DRJ
728 if (uap->to) {
729 error = getsockaddr(&sa, uap->to, uap->tolen);
730 if (error)
731 return (error);
35fbb1d9 732 }
984263bc
MD
733 aiov.iov_base = uap->buf;
734 aiov.iov_len = uap->len;
3e1837ce
DRJ
735 auio.uio_iov = &aiov;
736 auio.uio_iovcnt = 1;
737 auio.uio_offset = 0;
738 auio.uio_resid = uap->len;
739 auio.uio_segflg = UIO_USERSPACE;
740 auio.uio_rw = UIO_WRITE;
741 auio.uio_td = td;
984263bc 742
3919ced0 743 get_mplock();
3e1837ce 744 error = kern_sendmsg(uap->s, sa, &auio, NULL, uap->flags,
e54488bb 745 &uap->sysmsg_szresult);
3919ced0 746 rel_mplock();
984263bc 747
35fbb1d9
DRJ
748 if (sa)
749 FREE(sa, M_SONAME);
750 return (error);
984263bc
MD
751}
752
41c20dac 753/*
35fbb1d9 754 * sendmsg_args(int s, caddr_t msg, int flags)
3919ced0
MD
755 *
756 * MPALMOSTSAFE
41c20dac 757 */
984263bc 758int
753fd850 759sys_sendmsg(struct sendmsg_args *uap)
984263bc 760{
3e1837ce 761 struct thread *td = curthread;
984263bc 762 struct msghdr msg;
3e1837ce 763 struct uio auio;
75a872f8 764 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
35fbb1d9
DRJ
765 struct sockaddr *sa = NULL;
766 struct mbuf *control = NULL;
75a872f8 767 int error;
984263bc 768
35fbb1d9 769 error = copyin(uap->msg, (caddr_t)&msg, sizeof(msg));
984263bc
MD
770 if (error)
771 return (error);
35fbb1d9
DRJ
772
773 /*
774 * Conditionally copyin msg.msg_name.
775 */
776 if (msg.msg_name) {
777 error = getsockaddr(&sa, msg.msg_name, msg.msg_namelen);
778 if (error)
779 return (error);
35fbb1d9
DRJ
780 }
781
782 /*
3e1837ce 783 * Populate auio.
35fbb1d9 784 */
75a872f8 785 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
ef5c76d7 786 &auio.uio_resid);
984263bc 787 if (error)
8130f673 788 goto cleanup2;
3e1837ce
DRJ
789 auio.uio_iov = iov;
790 auio.uio_iovcnt = msg.msg_iovlen;
791 auio.uio_offset = 0;
3e1837ce
DRJ
792 auio.uio_segflg = UIO_USERSPACE;
793 auio.uio_rw = UIO_WRITE;
794 auio.uio_td = td;
35fbb1d9
DRJ
795
796 /*
797 * Conditionally copyin msg.msg_control.
798 */
799 if (msg.msg_control) {
3e1837ce
DRJ
800 if (msg.msg_controllen < sizeof(struct cmsghdr) ||
801 msg.msg_controllen > MLEN) {
35fbb1d9
DRJ
802 error = EINVAL;
803 goto cleanup;
804 }
74f1caca 805 control = m_get(MB_WAIT, MT_CONTROL);
3e1837ce
DRJ
806 if (control == NULL) {
807 error = ENOBUFS;
35fbb1d9 808 goto cleanup;
3e1837ce
DRJ
809 }
810 control->m_len = msg.msg_controllen;
811 error = copyin(msg.msg_control, mtod(control, caddr_t),
3919ced0 812 msg.msg_controllen);
3e1837ce
DRJ
813 if (error) {
814 m_free(control);
815 goto cleanup;
816 }
35fbb1d9
DRJ
817 }
818
3919ced0 819 get_mplock();
3e1837ce 820 error = kern_sendmsg(uap->s, sa, &auio, control, uap->flags,
e54488bb 821 &uap->sysmsg_szresult);
3919ced0 822 rel_mplock();
35fbb1d9
DRJ
823
824cleanup:
8130f673
MD
825 iovec_free(&iov, aiov);
826cleanup2:
35fbb1d9
DRJ
827 if (sa)
828 FREE(sa, M_SONAME);
984263bc
MD
829 return (error);
830}
984263bc 831
41c20dac 832/*
3e1837ce
DRJ
833 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
834 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
835 * Don't forget to FREE() and m_free() these if they are returned.
41c20dac 836 */
984263bc 837int
3e1837ce 838kern_recvmsg(int s, struct sockaddr **sa, struct uio *auio,
e54488bb 839 struct mbuf **control, int *flags, size_t *res)
984263bc 840{
dadab5e9
MD
841 struct thread *td = curthread;
842 struct proc *p = td->td_proc;
984263bc 843 struct file *fp;
e54488bb
MD
844 size_t len;
845 int error;
9ba76b73 846 int lflags;
984263bc 847 struct socket *so;
984263bc
MD
848#ifdef KTRACE
849 struct iovec *ktriov = NULL;
850 struct uio ktruio;
851#endif
852
853 error = holdsock(p->p_fd, s, &fp);
854 if (error)
855 return (error);
984263bc 856#ifdef KTRACE
dadab5e9 857 if (KTRPOINT(td, KTR_GENIO)) {
3e1837ce 858 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
984263bc
MD
859
860 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
3e1837ce
DRJ
861 bcopy(auio->uio_iov, ktriov, iovlen);
862 ktruio = *auio;
984263bc
MD
863 }
864#endif
3e1837ce 865 len = auio->uio_resid;
984263bc 866 so = (struct socket *)fp->f_data;
9ba76b73
MD
867
868 if (flags == NULL || (*flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
869 if (fp->f_flag & FNONBLOCK) {
870 if (flags) {
871 *flags |= MSG_FNONBLOCKING;
872 } else {
873 lflags = MSG_FNONBLOCKING;
874 flags = &lflags;
875 }
876 }
877 }
878
6b6e0885 879 error = so_pru_soreceive(so, sa, auio, NULL, control, flags);
984263bc 880 if (error) {
3e1837ce 881 if (auio->uio_resid != len && (error == ERESTART ||
984263bc
MD
882 error == EINTR || error == EWOULDBLOCK))
883 error = 0;
884 }
885#ifdef KTRACE
886 if (ktriov != NULL) {
887 if (error == 0) {
888 ktruio.uio_iov = ktriov;
3e1837ce 889 ktruio.uio_resid = len - auio->uio_resid;
9fb04d14 890 ktrgenio(td->td_lwp, s, UIO_READ, &ktruio, error);
984263bc
MD
891 }
892 FREE(ktriov, M_TEMP);
893 }
894#endif
35fbb1d9 895 if (error == 0)
3e1837ce 896 *res = len - auio->uio_resid;
9f87144f 897 fdrop(fp);
984263bc
MD
898 return (error);
899}
900
41c20dac
MD
901/*
902 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
903 * caddr_t from, int *fromlenaddr)
3919ced0
MD
904 *
905 * MPALMOSTSAFE
41c20dac 906 */
984263bc 907int
753fd850 908sys_recvfrom(struct recvfrom_args *uap)
984263bc 909{
3e1837ce
DRJ
910 struct thread *td = curthread;
911 struct uio auio;
984263bc 912 struct iovec aiov;
3e1837ce 913 struct sockaddr *sa = NULL;
35fbb1d9 914 int error, fromlen;
984263bc 915
3e1837ce 916 if (uap->from && uap->fromlenaddr) {
35fbb1d9 917 error = copyin(uap->fromlenaddr, &fromlen, sizeof(fromlen));
984263bc
MD
918 if (error)
919 return (error);
3e1837ce
DRJ
920 if (fromlen < 0)
921 return (EINVAL);
35fbb1d9
DRJ
922 } else {
923 fromlen = 0;
924 }
984263bc
MD
925 aiov.iov_base = uap->buf;
926 aiov.iov_len = uap->len;
3e1837ce
DRJ
927 auio.uio_iov = &aiov;
928 auio.uio_iovcnt = 1;
929 auio.uio_offset = 0;
930 auio.uio_resid = uap->len;
931 auio.uio_segflg = UIO_USERSPACE;
932 auio.uio_rw = UIO_READ;
933 auio.uio_td = td;
984263bc 934
3919ced0 935 get_mplock();
3e1837ce 936 error = kern_recvmsg(uap->s, uap->from ? &sa : NULL, &auio, NULL,
e54488bb 937 &uap->flags, &uap->sysmsg_szresult);
3919ced0 938 rel_mplock();
984263bc 939
3e1837ce 940 if (error == 0 && uap->from) {
c3996757
MD
941 /* note: sa may still be NULL */
942 if (sa) {
943 fromlen = MIN(fromlen, sa->sa_len);
944 error = copyout(sa, uap->from, fromlen);
945 } else {
946 fromlen = 0;
947 }
948 if (error == 0) {
35fbb1d9 949 error = copyout(&fromlen, uap->fromlenaddr,
c3996757
MD
950 sizeof(fromlen));
951 }
35fbb1d9 952 }
3e1837ce
DRJ
953 if (sa)
954 FREE(sa, M_SONAME);
984263bc 955
984263bc
MD
956 return (error);
957}
984263bc 958
41c20dac
MD
959/*
960 * recvmsg_args(int s, struct msghdr *msg, int flags)
3919ced0
MD
961 *
962 * MPALMOSTSAFE
41c20dac 963 */
984263bc 964int
753fd850 965sys_recvmsg(struct recvmsg_args *uap)
984263bc 966{
3e1837ce 967 struct thread *td = curthread;
984263bc 968 struct msghdr msg;
3e1837ce 969 struct uio auio;
75a872f8
DRJ
970 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
971 struct mbuf *m, *control = NULL;
3e1837ce 972 struct sockaddr *sa = NULL;
35fbb1d9 973 caddr_t ctlbuf;
3e1837ce 974 socklen_t *ufromlenp, *ucontrollenp;
75a872f8 975 int error, fromlen, controllen, len, flags, *uflagsp;
984263bc 976
35fbb1d9
DRJ
977 /*
978 * This copyin handles everything except the iovec.
979 */
980 error = copyin(uap->msg, &msg, sizeof(msg));
984263bc
MD
981 if (error)
982 return (error);
35fbb1d9 983
3e1837ce
DRJ
984 if (msg.msg_name && msg.msg_namelen < 0)
985 return (EINVAL);
986 if (msg.msg_control && msg.msg_controllen < 0)
987 return (EINVAL);
988
989 ufromlenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
3919ced0 990 msg_namelen));
35fbb1d9 991 ucontrollenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
3919ced0 992 msg_controllen));
3e1837ce 993 uflagsp = (int *)((caddr_t)uap->msg + offsetof(struct msghdr,
3919ced0 994 msg_flags));
35fbb1d9
DRJ
995
996 /*
3e1837ce 997 * Populate auio.
35fbb1d9 998 */
75a872f8 999 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
ef5c76d7 1000 &auio.uio_resid);
984263bc 1001 if (error)
75a872f8 1002 return (error);
3e1837ce
DRJ
1003 auio.uio_iov = iov;
1004 auio.uio_iovcnt = msg.msg_iovlen;
1005 auio.uio_offset = 0;
3e1837ce
DRJ
1006 auio.uio_segflg = UIO_USERSPACE;
1007 auio.uio_rw = UIO_READ;
1008 auio.uio_td = td;
35fbb1d9 1009
b7ccd728 1010 flags = uap->flags;
35fbb1d9 1011
3919ced0 1012 get_mplock();
e54488bb
MD
1013 error = kern_recvmsg(uap->s,
1014 (msg.msg_name ? &sa : NULL), &auio,
1015 (msg.msg_control ? &control : NULL), &flags,
1016 &uap->sysmsg_szresult);
3919ced0 1017 rel_mplock();
35fbb1d9
DRJ
1018
1019 /*
3e1837ce 1020 * Conditionally copyout the name and populate the namelen field.
35fbb1d9 1021 */
3e1837ce 1022 if (error == 0 && msg.msg_name) {
b9cd15b9
YT
1023 /* note: sa may still be NULL */
1024 if (sa != NULL) {
1025 fromlen = MIN(msg.msg_namelen, sa->sa_len);
1026 error = copyout(sa, msg.msg_name, fromlen);
b4354d10 1027 } else {
b9cd15b9 1028 fromlen = 0;
b4354d10 1029 }
35fbb1d9 1030 if (error == 0)
3e1837ce
DRJ
1031 error = copyout(&fromlen, ufromlenp,
1032 sizeof(*ufromlenp));
984263bc 1033 }
35fbb1d9
DRJ
1034
1035 /*
1036 * Copyout msg.msg_control and msg.msg_controllen.
1037 */
3e1837ce 1038 if (error == 0 && msg.msg_control) {
35fbb1d9 1039 len = msg.msg_controllen;
3e1837ce
DRJ
1040 m = control;
1041 ctlbuf = (caddr_t)msg.msg_control;
35fbb1d9
DRJ
1042
1043 while(m && len > 0) {
1044 unsigned int tocopy;
1045
1046 if (len >= m->m_len) {
1047 tocopy = m->m_len;
1048 } else {
1049 msg.msg_flags |= MSG_CTRUNC;
1050 tocopy = len;
1051 }
1052
1053 error = copyout(mtod(m, caddr_t), ctlbuf, tocopy);
1054 if (error)
1055 goto cleanup;
1056
1057 ctlbuf += tocopy;
1058 len -= tocopy;
1059 m = m->m_next;
1060 }
3e1837ce
DRJ
1061 controllen = ctlbuf - (caddr_t)msg.msg_control;
1062 error = copyout(&controllen, ucontrollenp,
35fbb1d9
DRJ
1063 sizeof(*ucontrollenp));
1064 }
1065
3e1837ce
DRJ
1066 if (error == 0)
1067 error = copyout(&flags, uflagsp, sizeof(*uflagsp));
1068
35fbb1d9 1069cleanup:
3e1837ce
DRJ
1070 if (sa)
1071 FREE(sa, M_SONAME);
75a872f8 1072 iovec_free(&iov, aiov);
3e1837ce
DRJ
1073 if (control)
1074 m_freem(control);
984263bc
MD
1075 return (error);
1076}
1077
41c20dac 1078/*
201305ad
DRJ
1079 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1080 * in kernel pointer instead of a userland pointer. This allows us
1081 * to manipulate socket options in the emulation code.
41c20dac 1082 */
984263bc 1083int
201305ad 1084kern_setsockopt(int s, struct sockopt *sopt)
984263bc 1085{
dadab5e9
MD
1086 struct thread *td = curthread;
1087 struct proc *p = td->td_proc;
984263bc 1088 struct file *fp;
984263bc
MD
1089 int error;
1090
b4354d10 1091 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
984263bc 1092 return (EFAULT);
201305ad 1093 if (sopt->sopt_valsize < 0)
984263bc
MD
1094 return (EINVAL);
1095
201305ad 1096 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
1097 if (error)
1098 return (error);
1099
201305ad 1100 error = sosetopt((struct socket *)fp->f_data, sopt);
9f87144f 1101 fdrop(fp);
201305ad
DRJ
1102 return (error);
1103}
1104
1105/*
1106 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
3919ced0
MD
1107 *
1108 * MPALMOSTSAFE
201305ad
DRJ
1109 */
1110int
753fd850 1111sys_setsockopt(struct setsockopt_args *uap)
201305ad
DRJ
1112{
1113 struct thread *td = curthread;
1114 struct sockopt sopt;
1115 int error;
1116
984263bc
MD
1117 sopt.sopt_level = uap->level;
1118 sopt.sopt_name = uap->name;
984263bc 1119 sopt.sopt_valsize = uap->valsize;
dadab5e9 1120 sopt.sopt_td = td;
aca22a94 1121 sopt.sopt_val = NULL;
201305ad 1122
b4354d10
MD
1123 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1124 return (EINVAL);
792239df
AE
1125 if (uap->val) {
1126 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1127 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1128 if (error)
1129 goto out;
792239df 1130 }
aca22a94 1131
3919ced0 1132 get_mplock();
201305ad 1133 error = kern_setsockopt(uap->s, &sopt);
3919ced0 1134 rel_mplock();
de0003fe 1135out:
792239df
AE
1136 if (uap->val)
1137 kfree(sopt.sopt_val, M_TEMP);
984263bc
MD
1138 return(error);
1139}
1140
41c20dac 1141/*
201305ad
DRJ
1142 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1143 * in kernel pointer instead of a userland pointer. This allows us
1144 * to manipulate socket options in the emulation code.
41c20dac 1145 */
984263bc 1146int
201305ad 1147kern_getsockopt(int s, struct sockopt *sopt)
984263bc 1148{
dadab5e9
MD
1149 struct thread *td = curthread;
1150 struct proc *p = td->td_proc;
201305ad
DRJ
1151 struct file *fp;
1152 int error;
984263bc 1153
b4354d10 1154 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
201305ad 1155 return (EFAULT);
b4354d10 1156 if (sopt->sopt_valsize < 0 || sopt->sopt_valsize > SOMAXOPT_SIZE)
201305ad
DRJ
1157 return (EINVAL);
1158
1159 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
1160 if (error)
1161 return (error);
201305ad
DRJ
1162
1163 error = sogetopt((struct socket *)fp->f_data, sopt);
9f87144f 1164 fdrop(fp);
201305ad
DRJ
1165 return (error);
1166}
1167
1168/*
3919ced0
MD
1169 * getsockopt_args(int s, int level, int name, caddr_t val, int *avalsize)
1170 *
1171 * MPALMOSTSAFE
201305ad
DRJ
1172 */
1173int
753fd850 1174sys_getsockopt(struct getsockopt_args *uap)
201305ad
DRJ
1175{
1176 struct thread *td = curthread;
1177 struct sockopt sopt;
1178 int error, valsize;
1179
984263bc 1180 if (uap->val) {
201305ad
DRJ
1181 error = copyin(uap->avalsize, &valsize, sizeof(valsize));
1182 if (error)
984263bc 1183 return (error);
984263bc
MD
1184 } else {
1185 valsize = 0;
1186 }
1187
984263bc
MD
1188 sopt.sopt_level = uap->level;
1189 sopt.sopt_name = uap->name;
201305ad 1190 sopt.sopt_valsize = valsize;
dadab5e9 1191 sopt.sopt_td = td;
aca22a94 1192 sopt.sopt_val = NULL;
984263bc 1193
aca22a94 1194 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
b4354d10 1195 return (EINVAL);
792239df
AE
1196 if (uap->val) {
1197 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1198 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1199 if (error)
1200 goto out;
792239df 1201 }
aca22a94 1202
3919ced0 1203 get_mplock();
201305ad 1204 error = kern_getsockopt(uap->s, &sopt);
3919ced0 1205 rel_mplock();
de0003fe
AE
1206 if (error)
1207 goto out;
1208 valsize = sopt.sopt_valsize;
1209 error = copyout(&valsize, uap->avalsize, sizeof(valsize));
1210 if (error)
1211 goto out;
792239df
AE
1212 if (uap->val)
1213 error = copyout(sopt.sopt_val, uap->val, sopt.sopt_valsize);
de0003fe 1214out:
792239df
AE
1215 if (uap->val)
1216 kfree(sopt.sopt_val, M_TEMP);
984263bc
MD
1217 return (error);
1218}
1219
1220/*
5969a6f1
DRJ
1221 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1222 * This allows kern_getsockname() to return a pointer to an allocated struct
1223 * sockaddr which must be freed later with FREE(). The caller must
1224 * initialize *name to NULL.
984263bc 1225 */
5969a6f1
DRJ
1226int
1227kern_getsockname(int s, struct sockaddr **name, int *namelen)
984263bc 1228{
dadab5e9
MD
1229 struct thread *td = curthread;
1230 struct proc *p = td->td_proc;
984263bc 1231 struct file *fp;
dadab5e9 1232 struct socket *so;
5969a6f1
DRJ
1233 struct sockaddr *sa = NULL;
1234 int error;
984263bc 1235
5969a6f1 1236 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
1237 if (error)
1238 return (error);
5969a6f1 1239 if (*namelen < 0) {
9f87144f 1240 fdrop(fp);
984263bc
MD
1241 return (EINVAL);
1242 }
1243 so = (struct socket *)fp->f_data;
6b6e0885 1244 error = so_pru_sockaddr(so, &sa);
5969a6f1 1245 if (error == 0) {
b4354d10 1246 if (sa == NULL) {
5969a6f1
DRJ
1247 *namelen = 0;
1248 } else {
1249 *namelen = MIN(*namelen, sa->sa_len);
1250 *name = sa;
1251 }
984263bc
MD
1252 }
1253
9f87144f 1254 fdrop(fp);
984263bc
MD
1255 return (error);
1256}
1257
5969a6f1
DRJ
1258/*
1259 * getsockname_args(int fdes, caddr_t asa, int *alen)
1260 *
1261 * Get socket name.
3919ced0
MD
1262 *
1263 * MPALMOSTSAFE
5969a6f1 1264 */
984263bc 1265int
753fd850 1266sys_getsockname(struct getsockname_args *uap)
984263bc 1267{
5969a6f1
DRJ
1268 struct sockaddr *sa = NULL;
1269 int error, sa_len;
1270
1271 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1272 if (error)
1273 return (error);
1274
3919ced0 1275 get_mplock();
5969a6f1 1276 error = kern_getsockname(uap->fdes, &sa, &sa_len);
3919ced0 1277 rel_mplock();
984263bc 1278
5969a6f1
DRJ
1279 if (error == 0)
1280 error = copyout(sa, uap->asa, sa_len);
1281 if (error == 0)
1282 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1283 if (sa)
1284 FREE(sa, M_SONAME);
1285 return (error);
984263bc
MD
1286}
1287
984263bc 1288/*
5969a6f1
DRJ
1289 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1290 * This allows kern_getpeername() to return a pointer to an allocated struct
1291 * sockaddr which must be freed later with FREE(). The caller must
1292 * initialize *name to NULL.
984263bc 1293 */
5969a6f1
DRJ
1294int
1295kern_getpeername(int s, struct sockaddr **name, int *namelen)
984263bc 1296{
dadab5e9
MD
1297 struct thread *td = curthread;
1298 struct proc *p = td->td_proc;
984263bc 1299 struct file *fp;
dadab5e9 1300 struct socket *so;
5969a6f1
DRJ
1301 struct sockaddr *sa = NULL;
1302 int error;
984263bc 1303
5969a6f1 1304 error = holdsock(p->p_fd, s, &fp);
984263bc
MD
1305 if (error)
1306 return (error);
5969a6f1 1307 if (*namelen < 0) {
9f87144f 1308 fdrop(fp);
5969a6f1
DRJ
1309 return (EINVAL);
1310 }
984263bc
MD
1311 so = (struct socket *)fp->f_data;
1312 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
9f87144f 1313 fdrop(fp);
984263bc
MD
1314 return (ENOTCONN);
1315 }
6b6e0885 1316 error = so_pru_peeraddr(so, &sa);
5969a6f1 1317 if (error == 0) {
b4354d10 1318 if (sa == NULL) {
5969a6f1
DRJ
1319 *namelen = 0;
1320 } else {
1321 *namelen = MIN(*namelen, sa->sa_len);
1322 *name = sa;
1323 }
984263bc 1324 }
5969a6f1 1325
9f87144f 1326 fdrop(fp);
984263bc
MD
1327 return (error);
1328}
1329
5969a6f1
DRJ
1330/*
1331 * getpeername_args(int fdes, caddr_t asa, int *alen)
1332 *
1333 * Get name of peer for connected socket.
3919ced0
MD
1334 *
1335 * MPALMOSTSAFE
5969a6f1 1336 */
984263bc 1337int
753fd850 1338sys_getpeername(struct getpeername_args *uap)
984263bc 1339{
5969a6f1
DRJ
1340 struct sockaddr *sa = NULL;
1341 int error, sa_len;
1342
1343 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1344 if (error)
1345 return (error);
1346
3919ced0 1347 get_mplock();
5969a6f1 1348 error = kern_getpeername(uap->fdes, &sa, &sa_len);
3919ced0 1349 rel_mplock();
5969a6f1
DRJ
1350
1351 if (error == 0)
1352 error = copyout(sa, uap->asa, sa_len);
1353 if (error == 0)
1354 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1355 if (sa)
1356 FREE(sa, M_SONAME);
1357 return (error);
984263bc
MD
1358}
1359
984263bc 1360int
02844a31 1361getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len)
984263bc
MD
1362{
1363 struct sockaddr *sa;
1364 int error;
1365
02844a31 1366 *namp = NULL;
984263bc
MD
1367 if (len > SOCK_MAXADDRLEN)
1368 return ENAMETOOLONG;
02844a31
MD
1369 if (len < offsetof(struct sockaddr, sa_data[0]))
1370 return EDOM;
984263bc
MD
1371 MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1372 error = copyin(uaddr, sa, len);
1373 if (error) {
1374 FREE(sa, M_SONAME);
1375 } else {
75a872f8
DRJ
1376#if BYTE_ORDER != BIG_ENDIAN
1377 /*
1378 * The bind(), connect(), and sendto() syscalls were not
1379 * versioned for COMPAT_43. Thus, this check must stay.
1380 */
984263bc
MD
1381 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1382 sa->sa_family = sa->sa_len;
1383#endif
1384 sa->sa_len = len;
1385 *namp = sa;
1386 }
1387 return error;
1388}
1389
1390/*
b4caac98
MD
1391 * Detach a mapped page and release resources back to the system.
1392 * We must release our wiring and if the object is ripped out
1393 * from under the vm_page we become responsible for freeing the
df8d1020 1394 * page. These routines must be MPSAFE.
b4caac98
MD
1395 *
1396 * XXX HACK XXX TEMPORARY UNTIL WE IMPLEMENT EXT MBUF REFERENCE COUNTING
df8d1020
MD
1397 *
1398 * XXX vm_page_*() routines are not MPSAFE yet, the MP lock is required.
b4caac98
MD
1399 */
1400static void
013a4c0e 1401sf_buf_mref(void *arg)
b4caac98 1402{
e66bab2b
MD
1403 struct sfbuf_mref *sfm = arg;
1404
df8d1020
MD
1405 /*
1406 * We must already hold a ref so there is no race to 0, just
1407 * atomically increment the count.
1408 */
1409 atomic_add_int(&sfm->mref_count, 1);
b4caac98
MD
1410}
1411
1412static void
013a4c0e 1413sf_buf_mfree(void *arg)
b4caac98 1414{
e66bab2b 1415 struct sfbuf_mref *sfm = arg;
b4caac98 1416 vm_page_t m;
e66bab2b
MD
1417
1418 KKASSERT(sfm->mref_count > 0);
321e057f 1419 if (atomic_fetchadd_int(&sfm->mref_count, -1) == 1) {
df8d1020 1420 /*
df8d1020
MD
1421 * XXX vm_page_*() and SFBUF routines not MPSAFE yet.
1422 */
321e057f
SZ
1423 get_mplock();
1424 crit_enter();
1425 m = sf_buf_page(sfm->sf);
1426 sf_buf_free(sfm->sf);
1427 vm_page_unwire(m, 0);
1428 if (m->wire_count == 0 && m->object == NULL)
1429 vm_page_try_to_free(m);
1430 crit_exit();
1431 rel_mplock();
1432 kfree(sfm, M_SENDFILE);
b4caac98
MD
1433 }
1434}
1435
1436/*
984263bc
MD
1437 * sendfile(2).
1438 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1439 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1440 *
1441 * Send a file specified by 'fd' and starting at 'offset' to a socket
1442 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1443 * nbytes == 0. Optionally add a header and/or trailer to the socket
1444 * output. If specified, write the total number of bytes sent into *sbytes.
75a872f8
DRJ
1445 *
1446 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1447 * the headers to count against the remaining bytes to be sent from
1448 * the file descriptor. We may wish to implement a compatibility syscall
1449 * in the future.
3919ced0
MD
1450 *
1451 * MPALMOSTSAFE
984263bc
MD
1452 */
1453int
753fd850 1454sys_sendfile(struct sendfile_args *uap)
984263bc 1455{
dadab5e9
MD
1456 struct thread *td = curthread;
1457 struct proc *p = td->td_proc;
984263bc 1458 struct file *fp;
75a872f8 1459 struct vnode *vp = NULL;
984263bc 1460 struct sf_hdtr hdtr;
75a872f8
DRJ
1461 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
1462 struct uio auio;
30eeba44 1463 struct mbuf *mheader = NULL;
e54488bb
MD
1464 size_t hbytes = 0;
1465 size_t tbytes;
1466 off_t hdtr_size = 0;
1467 off_t sbytes;
1468 int error;
984263bc 1469
dadab5e9 1470 KKASSERT(p);
dadab5e9 1471
984263bc
MD
1472 /*
1473 * Do argument checking. Must be a regular file in, stream
1474 * type and connected socket out, positive offset.
1475 */
fa541be6 1476 fp = holdfp(p->p_fd, uap->fd, FREAD);
984263bc 1477 if (fp == NULL) {
f0846490 1478 return (EBADF);
984263bc
MD
1479 }
1480 if (fp->f_type != DTYPE_VNODE) {
9f87144f 1481 fdrop(fp);
f0846490 1482 return (EINVAL);
984263bc 1483 }
3919ced0 1484 get_mplock();
984263bc
MD
1485 vp = (struct vnode *)fp->f_data;
1486 vref(vp);
9f87144f 1487 fdrop(fp);
75a872f8
DRJ
1488
1489 /*
1490 * If specified, get the pointer to the sf_hdtr struct for
1491 * any headers/trailers.
1492 */
1493 if (uap->hdtr) {
1494 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1495 if (error)
1496 goto done;
1497 /*
1498 * Send any headers.
1499 */
1500 if (hdtr.headers) {
1501 error = iovec_copyin(hdtr.headers, &iov, aiov,
ef5c76d7 1502 hdtr.hdr_cnt, &hbytes);
75a872f8
DRJ
1503 if (error)
1504 goto done;
1505 auio.uio_iov = iov;
1506 auio.uio_iovcnt = hdtr.hdr_cnt;
1507 auio.uio_offset = 0;
1508 auio.uio_segflg = UIO_USERSPACE;
1509 auio.uio_rw = UIO_WRITE;
1510 auio.uio_td = td;
30eeba44 1511 auio.uio_resid = hbytes;
75a872f8 1512
e12241e1 1513 mheader = m_uiomove(&auio);
75a872f8
DRJ
1514
1515 iovec_free(&iov, aiov);
30eeba44 1516 if (mheader == NULL)
75a872f8 1517 goto done;
75a872f8
DRJ
1518 }
1519 }
1520
30eeba44 1521 error = kern_sendfile(vp, uap->s, uap->offset, uap->nbytes, mheader,
3919ced0 1522 &sbytes, uap->flags);
75a872f8
DRJ
1523 if (error)
1524 goto done;
1525
1526 /*
1527 * Send trailers. Wimp out and use writev(2).
1528 */
1529 if (uap->hdtr != NULL && hdtr.trailers != NULL) {
1530 error = iovec_copyin(hdtr.trailers, &iov, aiov,
ef5c76d7 1531 hdtr.trl_cnt, &auio.uio_resid);
75a872f8
DRJ
1532 if (error)
1533 goto done;
1534 auio.uio_iov = iov;
1535 auio.uio_iovcnt = hdtr.trl_cnt;
1536 auio.uio_offset = 0;
1537 auio.uio_segflg = UIO_USERSPACE;
1538 auio.uio_rw = UIO_WRITE;
1539 auio.uio_td = td;
1540
30eeba44 1541 error = kern_sendmsg(uap->s, NULL, &auio, NULL, 0, &tbytes);
75a872f8
DRJ
1542
1543 iovec_free(&iov, aiov);
1544 if (error)
1545 goto done;
30eeba44 1546 hdtr_size += tbytes; /* trailer bytes successfully sent */
75a872f8
DRJ
1547 }
1548
1549done:
3919ced0
MD
1550 if (vp)
1551 vrele(vp);
1552 rel_mplock();
75a872f8
DRJ
1553 if (uap->sbytes != NULL) {
1554 sbytes += hdtr_size;
1555 copyout(&sbytes, uap->sbytes, sizeof(off_t));
1556 }
75a872f8
DRJ
1557 return (error);
1558}
1559
1560int
06ecca5a 1561kern_sendfile(struct vnode *vp, int sfd, off_t offset, size_t nbytes,
30eeba44 1562 struct mbuf *mheader, off_t *sbytes, int flags)
75a872f8
DRJ
1563{
1564 struct thread *td = curthread;
1565 struct proc *p = td->td_proc;
1566 struct vm_object *obj;
1567 struct socket *so;
285332f0 1568 struct file *fp;
75a872f8
DRJ
1569 struct mbuf *m;
1570 struct sf_buf *sf;
e66bab2b 1571 struct sfbuf_mref *sfm;
75a872f8
DRJ
1572 struct vm_page *pg;
1573 off_t off, xfsize;
39b3370f 1574 off_t hbytes = 0;
75a872f8
DRJ
1575 int error = 0;
1576
7540ab49
MD
1577 if (vp->v_type != VREG) {
1578 error = EINVAL;
1579 goto done0;
1580 }
1581 if ((obj = vp->v_object) == NULL) {
984263bc 1582 error = EINVAL;
285332f0 1583 goto done0;
984263bc 1584 }
06ecca5a 1585 error = holdsock(p->p_fd, sfd, &fp);
984263bc 1586 if (error)
285332f0 1587 goto done0;
984263bc
MD
1588 so = (struct socket *)fp->f_data;
1589 if (so->so_type != SOCK_STREAM) {
1590 error = EINVAL;
1591 goto done;
1592 }
1593 if ((so->so_state & SS_ISCONNECTED) == 0) {
1594 error = ENOTCONN;
1595 goto done;
1596 }
75a872f8 1597 if (offset < 0) {
984263bc
MD
1598 error = EINVAL;
1599 goto done;
1600 }
1601
75a872f8 1602 *sbytes = 0;
984263bc
MD
1603 /*
1604 * Protect against multiple writers to the socket.
1605 */
6d49aa6f 1606 ssb_lock(&so->so_snd, M_WAITOK);
984263bc
MD
1607
1608 /*
1609 * Loop through the pages in the file, starting with the requested
1610 * offset. Get a file page (do I/O if necessary), map the file page
1611 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1612 * it on the socket.
1613 */
39b3370f 1614 for (off = offset; ; off += xfsize, *sbytes += xfsize + hbytes) {
984263bc
MD
1615 vm_pindex_t pindex;
1616 vm_offset_t pgoff;
1617
1618 pindex = OFF_TO_IDX(off);
1619retry_lookup:
1620 /*
1621 * Calculate the amount to transfer. Not to exceed a page,
1622 * the EOF, or the passed in nbytes.
1623 */
57f7b636 1624 xfsize = vp->v_filesize - off;
984263bc
MD
1625 if (xfsize > PAGE_SIZE)
1626 xfsize = PAGE_SIZE;
1627 pgoff = (vm_offset_t)(off & PAGE_MASK);
1628 if (PAGE_SIZE - pgoff < xfsize)
1629 xfsize = PAGE_SIZE - pgoff;
75a872f8
DRJ
1630 if (nbytes && xfsize > (nbytes - *sbytes))
1631 xfsize = nbytes - *sbytes;
984263bc
MD
1632 if (xfsize <= 0)
1633 break;
1634 /*
1635 * Optimize the non-blocking case by looking at the socket space
1636 * before going to the extra work of constituting the sf_buf.
1637 */
6d49aa6f 1638 if ((fp->f_flag & FNONBLOCK) && ssb_space(&so->so_snd) <= 0) {
984263bc
MD
1639 if (so->so_state & SS_CANTSENDMORE)
1640 error = EPIPE;
1641 else
1642 error = EAGAIN;
6d49aa6f 1643 ssb_unlock(&so->so_snd);
984263bc
MD
1644 goto done;
1645 }
1646 /*
1647 * Attempt to look up the page.
1648 *
06ecca5a 1649 * Allocate if not found, wait and loop if busy, then
5fd012e0
MD
1650 * wire the page. critical section protection is
1651 * required to maintain the object association (an
1652 * interrupt can free the page) through to the
1653 * vm_page_wire() call.
984263bc 1654 */
5fd012e0 1655 crit_enter();
984263bc 1656 pg = vm_page_lookup(obj, pindex);
984263bc
MD
1657 if (pg == NULL) {
1658 pg = vm_page_alloc(obj, pindex, VM_ALLOC_NORMAL);
1659 if (pg == NULL) {
4ecf7cc9 1660 vm_wait(0);
5fd012e0 1661 crit_exit();
984263bc
MD
1662 goto retry_lookup;
1663 }
1664 vm_page_wakeup(pg);
1665 } else if (vm_page_sleep_busy(pg, TRUE, "sfpbsy")) {
5fd012e0 1666 crit_exit();
984263bc
MD
1667 goto retry_lookup;
1668 }
984263bc 1669 vm_page_wire(pg);
5fd012e0 1670 crit_exit();
984263bc
MD
1671
1672 /*
1673 * If page is not valid for what we need, initiate I/O
1674 */
1675
1676 if (!pg->valid || !vm_page_is_valid(pg, pgoff, xfsize)) {
1677 struct uio auio;
1678 struct iovec aiov;
1679 int bsize;
1680
1681 /*
1682 * Ensure that our page is still around when the I/O
1683 * completes.
1684 */
1685 vm_page_io_start(pg);
1686
1687 /*
1688 * Get the page from backing store.
1689 */
1690 bsize = vp->v_mount->mnt_stat.f_iosize;
1691 auio.uio_iov = &aiov;
1692 auio.uio_iovcnt = 1;
1693 aiov.iov_base = 0;
1694 aiov.iov_len = MAXBSIZE;
1695 auio.uio_resid = MAXBSIZE;
1696 auio.uio_offset = trunc_page(off);
1697 auio.uio_segflg = UIO_NOCOPY;
1698 auio.uio_rw = UIO_READ;
dadab5e9 1699 auio.uio_td = td;
ab6f251b 1700 vn_lock(vp, LK_SHARED | LK_RETRY);
dadab5e9
MD
1701 error = VOP_READ(vp, &auio,
1702 IO_VMIO | ((MAXBSIZE / bsize) << 16),
9910d07b 1703 td->td_ucred);
a11aaa81 1704 vn_unlock(vp);
984263bc
MD
1705 vm_page_flag_clear(pg, PG_ZERO);
1706 vm_page_io_finish(pg);
1707 if (error) {
f2555cdd 1708 crit_enter();
984263bc 1709 vm_page_unwire(pg, 0);
f2555cdd
MD
1710 vm_page_try_to_free(pg);
1711 crit_exit();
6d49aa6f 1712 ssb_unlock(&so->so_snd);
984263bc
MD
1713 goto done;
1714 }
1715 }
1716
1717
1718 /*
1719 * Get a sendfile buf. We usually wait as long as necessary,
1720 * but this wait can be interrupted.
1721 */
4f1640d6 1722 if ((sf = sf_buf_alloc(pg, SFB_CATCH)) == NULL) {
5fd012e0 1723 crit_enter();
984263bc 1724 vm_page_unwire(pg, 0);
f2555cdd 1725 vm_page_try_to_free(pg);
5fd012e0 1726 crit_exit();
6d49aa6f 1727 ssb_unlock(&so->so_snd);
984263bc
MD
1728 error = EINTR;
1729 goto done;
1730 }
1731
984263bc
MD
1732 /*
1733 * Get an mbuf header and set it up as having external storage.
1734 */
74f1caca 1735 MGETHDR(m, MB_WAIT, MT_DATA);
984263bc
MD
1736 if (m == NULL) {
1737 error = ENOBUFS;
b4caac98 1738 sf_buf_free(sf);
6d49aa6f 1739 ssb_unlock(&so->so_snd);
984263bc
MD
1740 goto done;
1741 }
e66bab2b
MD
1742
1743 /*
1744 * sfm is a temporary hack, use a per-cpu cache for this.
1745 */
efda3bd0 1746 sfm = kmalloc(sizeof(struct sfbuf_mref), M_SENDFILE, M_WAITOK);
e66bab2b
MD
1747 sfm->sf = sf;
1748 sfm->mref_count = 1;
1749
b542cd49
JS
1750 m->m_ext.ext_free = sf_buf_mfree;
1751 m->m_ext.ext_ref = sf_buf_mref;
e66bab2b 1752 m->m_ext.ext_arg = sfm;
984263bc
MD
1753 m->m_ext.ext_buf = (void *)sf->kva;
1754 m->m_ext.ext_size = PAGE_SIZE;
1755 m->m_data = (char *) sf->kva + pgoff;
013a4c0e 1756 m->m_flags |= M_EXT;
984263bc 1757 m->m_pkthdr.len = m->m_len = xfsize;
b542cd49 1758 KKASSERT((m->m_flags & (M_EXT_CLUSTER)) == 0);
30eeba44 1759
39b3370f
JH
1760 if (mheader != NULL) {
1761 hbytes = mheader->m_pkthdr.len;
30eeba44
JH
1762 mheader->m_pkthdr.len += m->m_pkthdr.len;
1763 m_cat(mheader, m);
1764 m = mheader;
1765 mheader = NULL;
39b3370f
JH
1766 } else
1767 hbytes = 0;
30eeba44 1768
984263bc
MD
1769 /*
1770 * Add the buffer to the socket buffer chain.
1771 */
5fd012e0 1772 crit_enter();
984263bc
MD
1773retry_space:
1774 /*
1775 * Make sure that the socket is still able to take more data.
1776 * CANTSENDMORE being true usually means that the connection
1777 * was closed. so_error is true when an error was sensed after
1778 * a previous send.
1779 * The state is checked after the page mapping and buffer
1780 * allocation above since those operations may block and make
1781 * any socket checks stale. From this point forward, nothing
1782 * blocks before the pru_send (or more accurately, any blocking
1783 * results in a loop back to here to re-check).
1784 */
1785 if ((so->so_state & SS_CANTSENDMORE) || so->so_error) {
1786 if (so->so_state & SS_CANTSENDMORE) {
1787 error = EPIPE;
1788 } else {
1789 error = so->so_error;
1790 so->so_error = 0;
1791 }
1792 m_freem(m);
6d49aa6f 1793 ssb_unlock(&so->so_snd);
5fd012e0 1794 crit_exit();
984263bc
MD
1795 goto done;
1796 }
1797 /*
1798 * Wait for socket space to become available. We do this just
1799 * after checking the connection state above in order to avoid
6d49aa6f 1800 * a race condition with ssb_wait().
984263bc 1801 */
6d49aa6f 1802 if (ssb_space(&so->so_snd) < so->so_snd.ssb_lowat) {
9ba76b73 1803 if (fp->f_flag & FNONBLOCK) {
984263bc 1804 m_freem(m);
6d49aa6f 1805 ssb_unlock(&so->so_snd);
5fd012e0 1806 crit_exit();
984263bc
MD
1807 error = EAGAIN;
1808 goto done;
1809 }
6d49aa6f 1810 error = ssb_wait(&so->so_snd);
984263bc 1811 /*
6d49aa6f 1812 * An error from ssb_wait usually indicates that we've
984263bc
MD
1813 * been interrupted by a signal. If we've sent anything
1814 * then return bytes sent, otherwise return the error.
1815 */
1816 if (error) {
1817 m_freem(m);
6d49aa6f 1818 ssb_unlock(&so->so_snd);
5fd012e0 1819 crit_exit();
984263bc
MD
1820 goto done;
1821 }
1822 goto retry_space;
1823 }
3c6b2883 1824 error = so_pru_send(so, 0, m, NULL, NULL, td);
5fd012e0 1825 crit_exit();
984263bc 1826 if (error) {
6d49aa6f 1827 ssb_unlock(&so->so_snd);
984263bc
MD
1828 goto done;
1829 }
1830 }
d785f69d 1831 if (mheader != NULL) {
39b3370f 1832 *sbytes += mheader->m_pkthdr.len;
d785f69d
JH
1833 error = so_pru_send(so, 0, mheader, NULL, NULL, td);
1834 mheader = NULL;
1835 }
6d49aa6f 1836 ssb_unlock(&so->so_snd);
984263bc 1837
984263bc 1838done:
9f87144f 1839 fdrop(fp);
285332f0 1840done0:
30eeba44
JH
1841 if (mheader != NULL)
1842 m_freem(mheader);
984263bc
MD
1843 return (error);
1844}
78812139 1845
3919ced0
MD
1846/*
1847 * MPALMOSTSAFE
1848 */
78812139 1849int
753fd850 1850sys_sctp_peeloff(struct sctp_peeloff_args *uap)
78812139
EN
1851{
1852#ifdef SCTP
1853 struct thread *td = curthread;
1854 struct proc *p = td->td_proc;
78812139
EN
1855 struct file *lfp = NULL;
1856 struct file *nfp = NULL;
1857 int error;
1858 struct socket *head, *so;
1859 caddr_t assoc_id;
1860 int fd;
1861 short fflag; /* type must match fp->f_flag */
1862
1863 assoc_id = uap->name;
fa541be6 1864 error = holdsock(p->p_fd, uap->sd, &lfp);
3919ced0 1865 if (error)
78812139 1866 return (error);
3919ced0
MD
1867
1868 get_mplock();
78812139
EN
1869 crit_enter();
1870 head = (struct socket *)lfp->f_data;
1871 error = sctp_can_peel_off(head, assoc_id);
1872 if (error) {
1873 crit_exit();
1874 goto done;
1875 }
1876 /*
1877 * At this point we know we do have a assoc to pull
1878 * we proceed to get the fd setup. This may block
1879 * but that is ok.
1880 */
1881
1882 fflag = lfp->f_flag;
1883 error = falloc(p, &nfp, &fd);
1884 if (error) {
1885 /*
1886 * Probably ran out of file descriptors. Put the
1887 * unaccepted connection back onto the queue and
1888 * do another wakeup so some other process might
1889 * have a chance at it.
1890 */
1891 crit_exit();
1892 goto done;
1893 }
e54488bb 1894 uap->sysmsg_iresult = fd;
78812139
EN
1895
1896 so = sctp_get_peeloff(head, assoc_id, &error);
1897 if (so == NULL) {
1898 /*
1899 * Either someone else peeled it off OR
1900 * we can't get a socket.
1901 */
1902 goto noconnection;
1903 }
1904 so->so_state &= ~SS_COMP;
1905 so->so_state &= ~SS_NOFDREF;
1906 so->so_head = NULL;
1907 if (head->so_sigio != NULL)
1908 fsetown(fgetown(head->so_sigio), &so->so_sigio);
1909
fbb4eeab 1910 nfp->f_type = DTYPE_SOCKET;
78812139
EN
1911 nfp->f_flag = fflag;
1912 nfp->f_ops = &socketops;
fbb4eeab 1913 nfp->f_data = so;
78812139
EN
1914
1915noconnection:
1916 /*
259b8ea0
MD
1917 * Assign the file pointer to the reserved descriptor, or clear
1918 * the reserved descriptor if an error occured.
78812139 1919 */
fa541be6 1920 if (error)
259b8ea0
MD
1921 fsetfd(p, NULL, fd);
1922 else
1923 fsetfd(p, nfp, fd);
78812139
EN
1924 crit_exit();
1925 /*
1926 * Release explicitly held references before returning.
1927 */
1928done:
3919ced0 1929 rel_mplock();
78812139 1930 if (nfp != NULL)
9f87144f
MD
1931 fdrop(nfp);
1932 fdrop(lfp);
78812139
EN
1933 return (error);
1934#else /* SCTP */
1935 return(EOPNOTSUPP);
1936#endif /* SCTP */
1937}