proc->thread stage 4: rework the VFS and DEVICE subsystems to take thread
[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 $
dadab5e9 38 * $DragonFly: src/sys/kern/uipc_syscalls.c,v 1.4 2003/06/25 03:55:57 dillon Exp $
984263bc
MD
39 */
40
41#include "opt_compat.h"
42#include "opt_ktrace.h"
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>
55#include <sys/mbuf.h>
56#include <sys/protosw.h>
57#include <sys/socket.h>
58#include <sys/socketvar.h>
59#include <sys/signalvar.h>
60#include <sys/uio.h>
61#include <sys/vnode.h>
62#include <sys/lock.h>
63#include <sys/mount.h>
64#ifdef KTRACE
65#include <sys/ktrace.h>
66#endif
67#include <vm/vm.h>
68#include <vm/vm_object.h>
69#include <vm/vm_page.h>
70#include <vm/vm_pageout.h>
71#include <vm/vm_kern.h>
72#include <vm/vm_extern.h>
dadab5e9 73#include <sys/file2.h>
984263bc
MD
74
75static void sf_buf_init(void *arg);
76SYSINIT(sock_sf, SI_SUB_MBUF, SI_ORDER_ANY, sf_buf_init, NULL)
77
41c20dac
MD
78static int sendit __P((int s, struct msghdr *mp, int flags));
79static int recvit __P((int s, struct msghdr *mp,
984263bc
MD
80 caddr_t namelenp));
81
41c20dac
MD
82static int accept1 __P((struct accept_args *uap, int compat));
83static int do_sendfile __P((struct sendfile_args *uap, int compat));
84static int getsockname1 __P((struct getsockname_args *uap, int compat));
85static int getpeername1 __P((struct getpeername_args *uap, int compat));
984263bc
MD
86
87static SLIST_HEAD(, sf_buf) sf_freelist;
88static vm_offset_t sf_base;
89static struct sf_buf *sf_bufs;
90static int sf_buf_alloc_want;
91
92/*
93 * System call interface to the socket abstraction.
94 */
95#if defined(COMPAT_43) || defined(COMPAT_SUNOS)
96#define COMPAT_OLDSOCK
97#endif
98
99extern struct fileops socketops;
100
41c20dac
MD
101/*
102 * socket_args(int domain, int type, int protocol)
103 */
984263bc 104int
41c20dac 105socket(struct socket_args *uap)
984263bc 106{
dadab5e9
MD
107 struct thread *td = curthread;
108 struct proc *p = td->td_proc;
109 struct filedesc *fdp;
984263bc
MD
110 struct socket *so;
111 struct file *fp;
112 int fd, error;
113
dadab5e9
MD
114 KKASSERT(p);
115 fdp = p->p_fd;
116
984263bc
MD
117 error = falloc(p, &fp, &fd);
118 if (error)
119 return (error);
120 fhold(fp);
dadab5e9 121 error = socreate(uap->domain, &so, uap->type, uap->protocol, td);
984263bc
MD
122 if (error) {
123 if (fdp->fd_ofiles[fd] == fp) {
124 fdp->fd_ofiles[fd] = NULL;
dadab5e9 125 fdrop(fp, td);
984263bc
MD
126 }
127 } else {
128 fp->f_data = (caddr_t)so;
129 fp->f_flag = FREAD|FWRITE;
130 fp->f_ops = &socketops;
131 fp->f_type = DTYPE_SOCKET;
132 p->p_retval[0] = fd;
133 }
dadab5e9 134 fdrop(fp, td);
984263bc
MD
135 return (error);
136}
137
41c20dac
MD
138/*
139 * bind_args(int s, caddr_t name, int namelen)
140 *
141 */
984263bc
MD
142/* ARGSUSED */
143int
41c20dac 144bind(struct bind_args *uap)
984263bc 145{
dadab5e9
MD
146 struct thread *td = curthread;
147 struct proc *p = td->td_proc;
984263bc
MD
148 struct file *fp;
149 struct sockaddr *sa;
150 int error;
151
dadab5e9 152 KKASSERT(p);
984263bc
MD
153 error = holdsock(p->p_fd, uap->s, &fp);
154 if (error)
155 return (error);
156 error = getsockaddr(&sa, uap->name, uap->namelen);
157 if (error) {
dadab5e9 158 fdrop(fp, td);
984263bc
MD
159 return (error);
160 }
dadab5e9 161 error = sobind((struct socket *)fp->f_data, sa, td);
984263bc 162 FREE(sa, M_SONAME);
dadab5e9 163 fdrop(fp, td);
984263bc
MD
164 return (error);
165}
166
41c20dac
MD
167/*
168 * listen_args(int s, int backlog)
169 */
984263bc
MD
170/* ARGSUSED */
171int
41c20dac 172listen(struct listen_args *uap)
984263bc 173{
dadab5e9
MD
174 struct thread *td = curthread;
175 struct proc *p = td->td_proc;
984263bc
MD
176 struct file *fp;
177 int error;
178
dadab5e9 179 KKASSERT(p);
984263bc
MD
180 error = holdsock(p->p_fd, uap->s, &fp);
181 if (error)
182 return (error);
dadab5e9
MD
183 error = solisten((struct socket *)fp->f_data, uap->backlog, td);
184 fdrop(fp, td);
984263bc
MD
185 return(error);
186}
187
41c20dac
MD
188/*
189 * accept_args(int s, caddr_t name, int *anamelen)
190 */
984263bc 191static int
41c20dac 192accept1(struct accept_args *uap, int compat)
984263bc 193{
dadab5e9
MD
194 struct thread *td = curthread;
195 struct proc *p = td->td_proc;
984263bc
MD
196 struct filedesc *fdp = p->p_fd;
197 struct file *lfp = NULL;
198 struct file *nfp = NULL;
199 struct sockaddr *sa;
200 int namelen, error, s;
201 struct socket *head, *so;
202 int fd;
203 u_int fflag; /* type must match fp->f_flag */
204 int tmp;
205
206 if (uap->name) {
207 error = copyin((caddr_t)uap->anamelen, (caddr_t)&namelen,
208 sizeof (namelen));
209 if(error)
210 return (error);
211 if (namelen < 0)
212 return (EINVAL);
213 }
214 error = holdsock(fdp, uap->s, &lfp);
215 if (error)
216 return (error);
217 s = splnet();
218 head = (struct socket *)lfp->f_data;
219 if ((head->so_options & SO_ACCEPTCONN) == 0) {
220 splx(s);
221 error = EINVAL;
222 goto done;
223 }
224 while (TAILQ_EMPTY(&head->so_comp) && head->so_error == 0) {
225 if (head->so_state & SS_CANTRCVMORE) {
226 head->so_error = ECONNABORTED;
227 break;
228 }
229 if ((head->so_state & SS_NBIO) != 0) {
230 head->so_error = EWOULDBLOCK;
231 break;
232 }
233 error = tsleep((caddr_t)&head->so_timeo, PSOCK | PCATCH,
234 "accept", 0);
235 if (error) {
236 splx(s);
237 goto done;
238 }
239 }
240 if (head->so_error) {
241 error = head->so_error;
242 head->so_error = 0;
243 splx(s);
244 goto done;
245 }
246
247 /*
248 * At this point we know that there is at least one connection
249 * ready to be accepted. Remove it from the queue prior to
250 * allocating the file descriptor for it since falloc() may
251 * block allowing another process to accept the connection
252 * instead.
253 */
254 so = TAILQ_FIRST(&head->so_comp);
255 TAILQ_REMOVE(&head->so_comp, so, so_list);
256 head->so_qlen--;
257
258 fflag = lfp->f_flag;
259 error = falloc(p, &nfp, &fd);
260 if (error) {
261 /*
262 * Probably ran out of file descriptors. Put the
263 * unaccepted connection back onto the queue and
264 * do another wakeup so some other process might
265 * have a chance at it.
266 */
267 TAILQ_INSERT_HEAD(&head->so_comp, so, so_list);
268 head->so_qlen++;
269 wakeup_one(&head->so_timeo);
270 splx(s);
271 goto done;
272 }
273 fhold(nfp);
274 p->p_retval[0] = fd;
275
276 /* connection has been removed from the listen queue */
277 KNOTE(&head->so_rcv.sb_sel.si_note, 0);
278
279 so->so_state &= ~SS_COMP;
280 so->so_head = NULL;
281 if (head->so_sigio != NULL)
282 fsetown(fgetown(head->so_sigio), &so->so_sigio);
283
284 nfp->f_data = (caddr_t)so;
285 nfp->f_flag = fflag;
286 nfp->f_ops = &socketops;
287 nfp->f_type = DTYPE_SOCKET;
288 /* Sync socket nonblocking/async state with file flags */
289 tmp = fflag & FNONBLOCK;
dadab5e9 290 (void) fo_ioctl(nfp, FIONBIO, (caddr_t)&tmp, td);
984263bc 291 tmp = fflag & FASYNC;
dadab5e9 292 (void) fo_ioctl(nfp, FIOASYNC, (caddr_t)&tmp, td);
984263bc
MD
293 sa = 0;
294 error = soaccept(so, &sa);
295 if (error) {
296 /*
297 * return a namelen of zero for older code which might
298 * ignore the return value from accept.
299 */
300 if (uap->name != NULL) {
301 namelen = 0;
302 (void) copyout((caddr_t)&namelen,
303 (caddr_t)uap->anamelen, sizeof(*uap->anamelen));
304 }
305 goto noconnection;
306 }
307 if (sa == NULL) {
308 namelen = 0;
309 if (uap->name)
310 goto gotnoname;
311 splx(s);
312 error = 0;
313 goto done;
314 }
315 if (uap->name) {
316 /* check sa_len before it is destroyed */
317 if (namelen > sa->sa_len)
318 namelen = sa->sa_len;
319#ifdef COMPAT_OLDSOCK
320 if (compat)
321 ((struct osockaddr *)sa)->sa_family =
322 sa->sa_family;
323#endif
324 error = copyout(sa, (caddr_t)uap->name, (u_int)namelen);
325 if (!error)
326gotnoname:
327 error = copyout((caddr_t)&namelen,
328 (caddr_t)uap->anamelen, sizeof (*uap->anamelen));
329 }
330noconnection:
331 if (sa)
332 FREE(sa, M_SONAME);
333
334 /*
335 * close the new descriptor, assuming someone hasn't ripped it
336 * out from under us.
337 */
338 if (error) {
339 if (fdp->fd_ofiles[fd] == nfp) {
340 fdp->fd_ofiles[fd] = NULL;
dadab5e9 341 fdrop(nfp, td);
984263bc
MD
342 }
343 }
344 splx(s);
345
346 /*
347 * Release explicitly held references before returning.
348 */
349done:
350 if (nfp != NULL)
dadab5e9
MD
351 fdrop(nfp, td);
352 fdrop(lfp, td);
984263bc
MD
353 return (error);
354}
355
356int
41c20dac 357accept(struct accept_args *uap)
984263bc 358{
41c20dac 359 return (accept1(uap, 0));
984263bc
MD
360}
361
362#ifdef COMPAT_OLDSOCK
363int
41c20dac 364oaccept(struct accept_args *uap)
984263bc
MD
365{
366
41c20dac 367 return (accept1(uap, 1));
984263bc
MD
368}
369#endif /* COMPAT_OLDSOCK */
370
41c20dac
MD
371/*
372 * connect_args(int s, caddr_t name, int namelen)
373 */
984263bc
MD
374/* ARGSUSED */
375int
41c20dac 376connect(struct connect_args *uap)
984263bc 377{
dadab5e9
MD
378 struct thread *td = curthread;
379 struct proc *p = td->td_proc;
984263bc 380 struct file *fp;
dadab5e9 381 struct socket *so;
984263bc
MD
382 struct sockaddr *sa;
383 int error, s;
384
385 error = holdsock(p->p_fd, uap->s, &fp);
386 if (error)
387 return (error);
388 so = (struct socket *)fp->f_data;
389 if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
390 error = EALREADY;
391 goto done;
392 }
393 error = getsockaddr(&sa, uap->name, uap->namelen);
394 if (error)
395 goto done;
dadab5e9 396 error = soconnect(so, sa, td);
984263bc
MD
397 if (error)
398 goto bad;
399 if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
400 FREE(sa, M_SONAME);
401 error = EINPROGRESS;
402 goto done;
403 }
404 s = splnet();
405 while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
406 error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
407 "connec", 0);
408 if (error)
409 break;
410 }
411 if (error == 0) {
412 error = so->so_error;
413 so->so_error = 0;
414 }
415 splx(s);
416bad:
417 so->so_state &= ~SS_ISCONNECTING;
418 FREE(sa, M_SONAME);
419 if (error == ERESTART)
420 error = EINTR;
421done:
dadab5e9 422 fdrop(fp, td);
984263bc
MD
423 return (error);
424}
425
41c20dac
MD
426/*
427 * socketpair(int domain, int type, int protocol, int *rsv)
428 */
984263bc 429int
41c20dac 430socketpair(struct socketpair_args *uap)
984263bc 431{
dadab5e9
MD
432 struct thread *td = curthread;
433 struct proc *p = td->td_proc;
434 struct filedesc *fdp;
984263bc
MD
435 struct file *fp1, *fp2;
436 struct socket *so1, *so2;
437 int fd, error, sv[2];
438
dadab5e9
MD
439 KKASSERT(p);
440 fdp = p->p_fd;
441 error = socreate(uap->domain, &so1, uap->type, uap->protocol, td);
984263bc
MD
442 if (error)
443 return (error);
dadab5e9 444 error = socreate(uap->domain, &so2, uap->type, uap->protocol, td);
984263bc
MD
445 if (error)
446 goto free1;
447 error = falloc(p, &fp1, &fd);
448 if (error)
449 goto free2;
450 fhold(fp1);
451 sv[0] = fd;
452 fp1->f_data = (caddr_t)so1;
453 error = falloc(p, &fp2, &fd);
454 if (error)
455 goto free3;
456 fhold(fp2);
457 fp2->f_data = (caddr_t)so2;
458 sv[1] = fd;
459 error = soconnect2(so1, so2);
460 if (error)
461 goto free4;
462 if (uap->type == SOCK_DGRAM) {
463 /*
464 * Datagram socket connection is asymmetric.
465 */
466 error = soconnect2(so2, so1);
467 if (error)
468 goto free4;
469 }
470 fp1->f_flag = fp2->f_flag = FREAD|FWRITE;
471 fp1->f_ops = fp2->f_ops = &socketops;
472 fp1->f_type = fp2->f_type = DTYPE_SOCKET;
473 error = copyout((caddr_t)sv, (caddr_t)uap->rsv, 2 * sizeof (int));
dadab5e9
MD
474 fdrop(fp1, td);
475 fdrop(fp2, td);
984263bc
MD
476 return (error);
477free4:
478 if (fdp->fd_ofiles[sv[1]] == fp2) {
479 fdp->fd_ofiles[sv[1]] = NULL;
dadab5e9 480 fdrop(fp2, td);
984263bc 481 }
dadab5e9 482 fdrop(fp2, td);
984263bc
MD
483free3:
484 if (fdp->fd_ofiles[sv[0]] == fp1) {
485 fdp->fd_ofiles[sv[0]] = NULL;
dadab5e9 486 fdrop(fp1, td);
984263bc 487 }
dadab5e9 488 fdrop(fp1, td);
984263bc
MD
489free2:
490 (void)soclose(so2);
491free1:
492 (void)soclose(so1);
493 return (error);
494}
495
496static int
dadab5e9 497sendit(int s, struct msghdr *mp, int flags)
984263bc 498{
dadab5e9
MD
499 struct thread *td = curthread;
500 struct proc *p = td->td_proc;
984263bc
MD
501 struct file *fp;
502 struct uio auio;
503 register struct iovec *iov;
504 register int i;
505 struct mbuf *control;
506 struct sockaddr *to;
507 int len, error;
508 struct socket *so;
509#ifdef KTRACE
510 struct iovec *ktriov = NULL;
511 struct uio ktruio;
512#endif
513
514 error = holdsock(p->p_fd, s, &fp);
515 if (error)
516 return (error);
517 auio.uio_iov = mp->msg_iov;
518 auio.uio_iovcnt = mp->msg_iovlen;
519 auio.uio_segflg = UIO_USERSPACE;
520 auio.uio_rw = UIO_WRITE;
dadab5e9 521 auio.uio_td = td;
984263bc
MD
522 auio.uio_offset = 0; /* XXX */
523 auio.uio_resid = 0;
524 iov = mp->msg_iov;
525 for (i = 0; i < mp->msg_iovlen; i++, iov++) {
526 if ((auio.uio_resid += iov->iov_len) < 0) {
dadab5e9 527 fdrop(fp, td);
984263bc
MD
528 return (EINVAL);
529 }
530 }
531 if (mp->msg_name) {
532 error = getsockaddr(&to, mp->msg_name, mp->msg_namelen);
533 if (error) {
dadab5e9 534 fdrop(fp, td);
984263bc
MD
535 return (error);
536 }
537 } else {
538 to = 0;
539 }
540 if (mp->msg_control) {
541 if (mp->msg_controllen < sizeof(struct cmsghdr)
542#ifdef COMPAT_OLDSOCK
543 && mp->msg_flags != MSG_COMPAT
544#endif
545 ) {
546 error = EINVAL;
547 goto bad;
548 }
549 error = sockargs(&control, mp->msg_control,
550 mp->msg_controllen, MT_CONTROL);
551 if (error)
552 goto bad;
553#ifdef COMPAT_OLDSOCK
554 if (mp->msg_flags == MSG_COMPAT) {
555 register struct cmsghdr *cm;
556
557 M_PREPEND(control, sizeof(*cm), M_WAIT);
558 if (control == 0) {
559 error = ENOBUFS;
560 goto bad;
561 } else {
562 cm = mtod(control, struct cmsghdr *);
563 cm->cmsg_len = control->m_len;
564 cm->cmsg_level = SOL_SOCKET;
565 cm->cmsg_type = SCM_RIGHTS;
566 }
567 }
568#endif
569 } else {
570 control = 0;
571 }
572#ifdef KTRACE
dadab5e9 573 if (KTRPOINT(td, KTR_GENIO)) {
984263bc
MD
574 int iovlen = auio.uio_iovcnt * sizeof (struct iovec);
575
576 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
577 bcopy((caddr_t)auio.uio_iov, (caddr_t)ktriov, iovlen);
578 ktruio = auio;
579 }
580#endif
581 len = auio.uio_resid;
582 so = (struct socket *)fp->f_data;
583 error = so->so_proto->pr_usrreqs->pru_sosend(so, to, &auio, 0, control,
dadab5e9 584 flags, td);
984263bc
MD
585 if (error) {
586 if (auio.uio_resid != len && (error == ERESTART ||
587 error == EINTR || error == EWOULDBLOCK))
588 error = 0;
589 if (error == EPIPE)
590 psignal(p, SIGPIPE);
591 }
592 if (error == 0)
593 p->p_retval[0] = len - auio.uio_resid;
594#ifdef KTRACE
595 if (ktriov != NULL) {
596 if (error == 0) {
597 ktruio.uio_iov = ktriov;
598 ktruio.uio_resid = p->p_retval[0];
599 ktrgenio(p->p_tracep, s, UIO_WRITE, &ktruio, error);
600 }
601 FREE(ktriov, M_TEMP);
602 }
603#endif
604bad:
dadab5e9 605 fdrop(fp, td);
984263bc
MD
606 if (to)
607 FREE(to, M_SONAME);
608 return (error);
609}
610
41c20dac
MD
611/*
612 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
613 */
984263bc 614int
41c20dac 615sendto(struct sendto_args *uap)
984263bc
MD
616{
617 struct msghdr msg;
618 struct iovec aiov;
619
620 msg.msg_name = uap->to;
621 msg.msg_namelen = uap->tolen;
622 msg.msg_iov = &aiov;
623 msg.msg_iovlen = 1;
624 msg.msg_control = 0;
625#ifdef COMPAT_OLDSOCK
626 msg.msg_flags = 0;
627#endif
628 aiov.iov_base = uap->buf;
629 aiov.iov_len = uap->len;
41c20dac 630 return (sendit(uap->s, &msg, uap->flags));
984263bc
MD
631}
632
633#ifdef COMPAT_OLDSOCK
41c20dac
MD
634/*
635 * osend_args(int s, caddr_t buf, int len, int flags)
636 */
984263bc 637int
41c20dac 638osend(struct osend_args *uap)
984263bc
MD
639{
640 struct msghdr msg;
641 struct iovec aiov;
642
643 msg.msg_name = 0;
644 msg.msg_namelen = 0;
645 msg.msg_iov = &aiov;
646 msg.msg_iovlen = 1;
647 aiov.iov_base = uap->buf;
648 aiov.iov_len = uap->len;
649 msg.msg_control = 0;
650 msg.msg_flags = 0;
41c20dac 651 return (sendit(uap->s, &msg, uap->flags));
984263bc
MD
652}
653
41c20dac
MD
654/*
655 * osendmsg_args(int s, caddr_t msg, int flags)
656 */
984263bc 657int
41c20dac 658osendmsg(struct osendmsg_args *uap)
984263bc
MD
659{
660 struct msghdr msg;
661 struct iovec aiov[UIO_SMALLIOV], *iov;
662 int error;
663
664 error = copyin(uap->msg, (caddr_t)&msg, sizeof (struct omsghdr));
665 if (error)
666 return (error);
667 if ((u_int)msg.msg_iovlen >= UIO_SMALLIOV) {
668 if ((u_int)msg.msg_iovlen >= UIO_MAXIOV)
669 return (EMSGSIZE);
670 MALLOC(iov, struct iovec *,
671 sizeof(struct iovec) * (u_int)msg.msg_iovlen, M_IOV,
672 M_WAITOK);
673 } else
674 iov = aiov;
675 error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
676 (unsigned)(msg.msg_iovlen * sizeof (struct iovec)));
677 if (error)
678 goto done;
679 msg.msg_flags = MSG_COMPAT;
680 msg.msg_iov = iov;
41c20dac 681 error = sendit(uap->s, &msg, uap->flags);
984263bc
MD
682done:
683 if (iov != aiov)
684 FREE(iov, M_IOV);
685 return (error);
686}
687#endif
688
41c20dac
MD
689/*
690 * sendmsg_args(int s, caddr_t msg, int flags)
691 */
984263bc 692int
41c20dac 693sendmsg(struct sendmsg_args *uap)
984263bc
MD
694{
695 struct msghdr msg;
696 struct iovec aiov[UIO_SMALLIOV], *iov;
697 int error;
698
699 error = copyin(uap->msg, (caddr_t)&msg, sizeof (msg));
700 if (error)
701 return (error);
702 if ((u_int)msg.msg_iovlen >= UIO_SMALLIOV) {
703 if ((u_int)msg.msg_iovlen >= UIO_MAXIOV)
704 return (EMSGSIZE);
705 MALLOC(iov, struct iovec *,
706 sizeof(struct iovec) * (u_int)msg.msg_iovlen, M_IOV,
707 M_WAITOK);
708 } else
709 iov = aiov;
710 if (msg.msg_iovlen &&
711 (error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
712 (unsigned)(msg.msg_iovlen * sizeof (struct iovec)))))
713 goto done;
714 msg.msg_iov = iov;
715#ifdef COMPAT_OLDSOCK
716 msg.msg_flags = 0;
717#endif
41c20dac 718 error = sendit(uap->s, &msg, uap->flags);
984263bc
MD
719done:
720 if (iov != aiov)
721 FREE(iov, M_IOV);
722 return (error);
723}
724
725static int
dadab5e9 726recvit(int s, struct msghdr *mp, caddr_t namelenp)
984263bc 727{
dadab5e9
MD
728 struct thread *td = curthread;
729 struct proc *p = td->td_proc;
984263bc
MD
730 struct file *fp;
731 struct uio auio;
732 register struct iovec *iov;
733 register int i;
734 int len, error;
735 struct mbuf *m, *control = 0;
736 caddr_t ctlbuf;
737 struct socket *so;
738 struct sockaddr *fromsa = 0;
739#ifdef KTRACE
740 struct iovec *ktriov = NULL;
741 struct uio ktruio;
742#endif
743
744 error = holdsock(p->p_fd, s, &fp);
745 if (error)
746 return (error);
747 auio.uio_iov = mp->msg_iov;
748 auio.uio_iovcnt = mp->msg_iovlen;
749 auio.uio_segflg = UIO_USERSPACE;
750 auio.uio_rw = UIO_READ;
dadab5e9 751 auio.uio_td = td;
984263bc
MD
752 auio.uio_offset = 0; /* XXX */
753 auio.uio_resid = 0;
754 iov = mp->msg_iov;
755 for (i = 0; i < mp->msg_iovlen; i++, iov++) {
756 if ((auio.uio_resid += iov->iov_len) < 0) {
dadab5e9 757 fdrop(fp, td);
984263bc
MD
758 return (EINVAL);
759 }
760 }
761#ifdef KTRACE
dadab5e9 762 if (KTRPOINT(td, KTR_GENIO)) {
984263bc
MD
763 int iovlen = auio.uio_iovcnt * sizeof (struct iovec);
764
765 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
766 bcopy((caddr_t)auio.uio_iov, (caddr_t)ktriov, iovlen);
767 ktruio = auio;
768 }
769#endif
770 len = auio.uio_resid;
771 so = (struct socket *)fp->f_data;
772 error = so->so_proto->pr_usrreqs->pru_soreceive(so, &fromsa, &auio,
773 (struct mbuf **)0, mp->msg_control ? &control : (struct mbuf **)0,
774 &mp->msg_flags);
775 if (error) {
776 if (auio.uio_resid != len && (error == ERESTART ||
777 error == EINTR || error == EWOULDBLOCK))
778 error = 0;
779 }
780#ifdef KTRACE
781 if (ktriov != NULL) {
782 if (error == 0) {
783 ktruio.uio_iov = ktriov;
784 ktruio.uio_resid = len - auio.uio_resid;
785 ktrgenio(p->p_tracep, s, UIO_READ, &ktruio, error);
786 }
787 FREE(ktriov, M_TEMP);
788 }
789#endif
790 if (error)
791 goto out;
792 p->p_retval[0] = len - auio.uio_resid;
793 if (mp->msg_name) {
794 len = mp->msg_namelen;
795 if (len <= 0 || fromsa == 0)
796 len = 0;
797 else {
798#ifndef MIN
799#define MIN(a,b) ((a)>(b)?(b):(a))
800#endif
801 /* save sa_len before it is destroyed by MSG_COMPAT */
802 len = MIN(len, fromsa->sa_len);
803#ifdef COMPAT_OLDSOCK
804 if (mp->msg_flags & MSG_COMPAT)
805 ((struct osockaddr *)fromsa)->sa_family =
806 fromsa->sa_family;
807#endif
808 error = copyout(fromsa,
809 (caddr_t)mp->msg_name, (unsigned)len);
810 if (error)
811 goto out;
812 }
813 mp->msg_namelen = len;
814 if (namelenp &&
815 (error = copyout((caddr_t)&len, namelenp, sizeof (int)))) {
816#ifdef COMPAT_OLDSOCK
817 if (mp->msg_flags & MSG_COMPAT)
818 error = 0; /* old recvfrom didn't check */
819 else
820#endif
821 goto out;
822 }
823 }
824 if (mp->msg_control) {
825#ifdef COMPAT_OLDSOCK
826 /*
827 * We assume that old recvmsg calls won't receive access
828 * rights and other control info, esp. as control info
829 * is always optional and those options didn't exist in 4.3.
830 * If we receive rights, trim the cmsghdr; anything else
831 * is tossed.
832 */
833 if (control && mp->msg_flags & MSG_COMPAT) {
834 if (mtod(control, struct cmsghdr *)->cmsg_level !=
835 SOL_SOCKET ||
836 mtod(control, struct cmsghdr *)->cmsg_type !=
837 SCM_RIGHTS) {
838 mp->msg_controllen = 0;
839 goto out;
840 }
841 control->m_len -= sizeof (struct cmsghdr);
842 control->m_data += sizeof (struct cmsghdr);
843 }
844#endif
845 len = mp->msg_controllen;
846 m = control;
847 mp->msg_controllen = 0;
848 ctlbuf = (caddr_t) mp->msg_control;
849
850 while (m && len > 0) {
851 unsigned int tocopy;
852
853 if (len >= m->m_len)
854 tocopy = m->m_len;
855 else {
856 mp->msg_flags |= MSG_CTRUNC;
857 tocopy = len;
858 }
859
860 if ((error = copyout((caddr_t)mtod(m, caddr_t),
861 ctlbuf, tocopy)) != 0)
862 goto out;
863
864 ctlbuf += tocopy;
865 len -= tocopy;
866 m = m->m_next;
867 }
868 mp->msg_controllen = ctlbuf - (caddr_t)mp->msg_control;
869 }
870out:
dadab5e9 871 fdrop(fp, td);
984263bc
MD
872 if (fromsa)
873 FREE(fromsa, M_SONAME);
874 if (control)
875 m_freem(control);
876 return (error);
877}
878
41c20dac
MD
879/*
880 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
881 * caddr_t from, int *fromlenaddr)
882 */
984263bc 883int
41c20dac 884recvfrom(struct recvfrom_args *uap)
984263bc
MD
885{
886 struct msghdr msg;
887 struct iovec aiov;
888 int error;
889
890 if (uap->fromlenaddr) {
891 error = copyin((caddr_t)uap->fromlenaddr,
892 (caddr_t)&msg.msg_namelen, sizeof (msg.msg_namelen));
893 if (error)
894 return (error);
895 } else
896 msg.msg_namelen = 0;
897 msg.msg_name = uap->from;
898 msg.msg_iov = &aiov;
899 msg.msg_iovlen = 1;
900 aiov.iov_base = uap->buf;
901 aiov.iov_len = uap->len;
902 msg.msg_control = 0;
903 msg.msg_flags = uap->flags;
41c20dac 904 return (recvit(uap->s, &msg, (caddr_t)uap->fromlenaddr));
984263bc
MD
905}
906
907#ifdef COMPAT_OLDSOCK
908int
41c20dac 909orecvfrom(struct recvfrom_args *uap)
984263bc 910{
984263bc 911 uap->flags |= MSG_COMPAT;
41c20dac 912 return (recvfrom(uap));
984263bc
MD
913}
914#endif
915
984263bc 916#ifdef COMPAT_OLDSOCK
41c20dac
MD
917/*
918 * struct orecv_args(int s, caddr_t buf, int len, int flags)
919 */
984263bc 920int
41c20dac 921orecv(struct orecv_args *uap)
984263bc
MD
922{
923 struct msghdr msg;
924 struct iovec aiov;
925
926 msg.msg_name = 0;
927 msg.msg_namelen = 0;
928 msg.msg_iov = &aiov;
929 msg.msg_iovlen = 1;
930 aiov.iov_base = uap->buf;
931 aiov.iov_len = uap->len;
932 msg.msg_control = 0;
933 msg.msg_flags = uap->flags;
41c20dac 934 return (recvit(uap->s, &msg, (caddr_t)0));
984263bc
MD
935}
936
937/*
938 * Old recvmsg. This code takes advantage of the fact that the old msghdr
939 * overlays the new one, missing only the flags, and with the (old) access
940 * rights where the control fields are now.
41c20dac
MD
941 *
942 * orecvmsg_args(int s, struct omsghdr *msg, int flags)
984263bc
MD
943 */
944int
41c20dac 945orecvmsg(struct orecvmsg_args *uap)
984263bc
MD
946{
947 struct msghdr msg;
948 struct iovec aiov[UIO_SMALLIOV], *iov;
949 int error;
950
951 error = copyin((caddr_t)uap->msg, (caddr_t)&msg,
952 sizeof (struct omsghdr));
953 if (error)
954 return (error);
955 if ((u_int)msg.msg_iovlen >= UIO_SMALLIOV) {
956 if ((u_int)msg.msg_iovlen >= UIO_MAXIOV)
957 return (EMSGSIZE);
958 MALLOC(iov, struct iovec *,
959 sizeof(struct iovec) * (u_int)msg.msg_iovlen, M_IOV,
960 M_WAITOK);
961 } else
962 iov = aiov;
963 msg.msg_flags = uap->flags | MSG_COMPAT;
964 error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
965 (unsigned)(msg.msg_iovlen * sizeof (struct iovec)));
966 if (error)
967 goto done;
968 msg.msg_iov = iov;
41c20dac 969 error = recvit(uap->s, &msg, (caddr_t)&uap->msg->msg_namelen);
984263bc
MD
970
971 if (msg.msg_controllen && error == 0)
972 error = copyout((caddr_t)&msg.msg_controllen,
973 (caddr_t)&uap->msg->msg_accrightslen, sizeof (int));
974done:
975 if (iov != aiov)
976 FREE(iov, M_IOV);
977 return (error);
978}
979#endif
980
41c20dac
MD
981/*
982 * recvmsg_args(int s, struct msghdr *msg, int flags)
983 */
984263bc 984int
41c20dac 985recvmsg(struct recvmsg_args *uap)
984263bc
MD
986{
987 struct msghdr msg;
988 struct iovec aiov[UIO_SMALLIOV], *uiov, *iov;
41c20dac 989 int error;
984263bc
MD
990
991 error = copyin((caddr_t)uap->msg, (caddr_t)&msg, sizeof (msg));
992 if (error)
993 return (error);
994 if ((u_int)msg.msg_iovlen >= UIO_SMALLIOV) {
995 if ((u_int)msg.msg_iovlen >= UIO_MAXIOV)
996 return (EMSGSIZE);
997 MALLOC(iov, struct iovec *,
998 sizeof(struct iovec) * (u_int)msg.msg_iovlen, M_IOV,
999 M_WAITOK);
1000 } else
1001 iov = aiov;
1002#ifdef COMPAT_OLDSOCK
1003 msg.msg_flags = uap->flags &~ MSG_COMPAT;
1004#else
1005 msg.msg_flags = uap->flags;
1006#endif
1007 uiov = msg.msg_iov;
1008 msg.msg_iov = iov;
1009 error = copyin((caddr_t)uiov, (caddr_t)iov,
1010 (unsigned)(msg.msg_iovlen * sizeof (struct iovec)));
1011 if (error)
1012 goto done;
41c20dac 1013 error = recvit(uap->s, &msg, (caddr_t)0);
984263bc
MD
1014 if (!error) {
1015 msg.msg_iov = uiov;
1016 error = copyout((caddr_t)&msg, (caddr_t)uap->msg, sizeof(msg));
1017 }
1018done:
1019 if (iov != aiov)
1020 FREE(iov, M_IOV);
1021 return (error);
1022}
1023
41c20dac
MD
1024/*
1025 * shutdown_args(int s, int how)
1026 */
984263bc
MD
1027/* ARGSUSED */
1028int
41c20dac 1029shutdown(struct shutdown_args *uap)
984263bc 1030{
dadab5e9
MD
1031 struct thread *td = curthread;
1032 struct proc *p = td->td_proc;
984263bc
MD
1033 struct file *fp;
1034 int error;
1035
dadab5e9 1036 KKASSERT(p);
984263bc
MD
1037 error = holdsock(p->p_fd, uap->s, &fp);
1038 if (error)
1039 return (error);
1040 error = soshutdown((struct socket *)fp->f_data, uap->how);
dadab5e9 1041 fdrop(fp, td);
984263bc
MD
1042 return(error);
1043}
1044
41c20dac
MD
1045/*
1046 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
1047 */
984263bc
MD
1048/* ARGSUSED */
1049int
41c20dac 1050setsockopt(struct setsockopt_args *uap)
984263bc 1051{
dadab5e9
MD
1052 struct thread *td = curthread;
1053 struct proc *p = td->td_proc;
984263bc
MD
1054 struct file *fp;
1055 struct sockopt sopt;
1056 int error;
1057
1058 if (uap->val == 0 && uap->valsize != 0)
1059 return (EFAULT);
1060 if (uap->valsize < 0)
1061 return (EINVAL);
1062
1063 error = holdsock(p->p_fd, uap->s, &fp);
1064 if (error)
1065 return (error);
1066
1067 sopt.sopt_dir = SOPT_SET;
1068 sopt.sopt_level = uap->level;
1069 sopt.sopt_name = uap->name;
1070 sopt.sopt_val = uap->val;
1071 sopt.sopt_valsize = uap->valsize;
dadab5e9 1072 sopt.sopt_td = td;
984263bc 1073 error = sosetopt((struct socket *)fp->f_data, &sopt);
dadab5e9 1074 fdrop(fp, td);
984263bc
MD
1075 return(error);
1076}
1077
41c20dac
MD
1078/*
1079 * getsockopt_Args(int s, int level, int name, caddr_t val, int *avalsize)
1080 */
984263bc
MD
1081/* ARGSUSED */
1082int
41c20dac 1083getsockopt(struct getsockopt_args *uap)
984263bc 1084{
dadab5e9
MD
1085 struct thread *td = curthread;
1086 struct proc *p = td->td_proc;
984263bc
MD
1087 int valsize, error;
1088 struct file *fp;
1089 struct sockopt sopt;
1090
1091 error = holdsock(p->p_fd, uap->s, &fp);
1092 if (error)
1093 return (error);
1094 if (uap->val) {
1095 error = copyin((caddr_t)uap->avalsize, (caddr_t)&valsize,
1096 sizeof (valsize));
1097 if (error) {
dadab5e9 1098 fdrop(fp, td);
984263bc
MD
1099 return (error);
1100 }
1101 if (valsize < 0) {
dadab5e9 1102 fdrop(fp, td);
984263bc
MD
1103 return (EINVAL);
1104 }
1105 } else {
1106 valsize = 0;
1107 }
1108
1109 sopt.sopt_dir = SOPT_GET;
1110 sopt.sopt_level = uap->level;
1111 sopt.sopt_name = uap->name;
1112 sopt.sopt_val = uap->val;
1113 sopt.sopt_valsize = (size_t)valsize; /* checked non-negative above */
dadab5e9 1114 sopt.sopt_td = td;
984263bc
MD
1115
1116 error = sogetopt((struct socket *)fp->f_data, &sopt);
1117 if (error == 0) {
1118 valsize = sopt.sopt_valsize;
1119 error = copyout((caddr_t)&valsize,
1120 (caddr_t)uap->avalsize, sizeof (valsize));
1121 }
dadab5e9 1122 fdrop(fp, td);
984263bc
MD
1123 return (error);
1124}
1125
1126/*
41c20dac
MD
1127 * getsockname_args(int fdes, caddr_t asa, int *alen)
1128 *
984263bc
MD
1129 * Get socket name.
1130 */
1131/* ARGSUSED */
1132static int
41c20dac 1133getsockname1(struct getsockname_args *uap, int compat)
984263bc 1134{
dadab5e9
MD
1135 struct thread *td = curthread;
1136 struct proc *p = td->td_proc;
984263bc 1137 struct file *fp;
dadab5e9 1138 struct socket *so;
984263bc
MD
1139 struct sockaddr *sa;
1140 int len, error;
1141
1142 error = holdsock(p->p_fd, uap->fdes, &fp);
1143 if (error)
1144 return (error);
1145 error = copyin((caddr_t)uap->alen, (caddr_t)&len, sizeof (len));
1146 if (error) {
dadab5e9 1147 fdrop(fp, td);
984263bc
MD
1148 return (error);
1149 }
1150 if (len < 0) {
dadab5e9 1151 fdrop(fp, td);
984263bc
MD
1152 return (EINVAL);
1153 }
1154 so = (struct socket *)fp->f_data;
1155 sa = 0;
1156 error = (*so->so_proto->pr_usrreqs->pru_sockaddr)(so, &sa);
1157 if (error)
1158 goto bad;
1159 if (sa == 0) {
1160 len = 0;
1161 goto gotnothing;
1162 }
1163
1164 len = MIN(len, sa->sa_len);
1165#ifdef COMPAT_OLDSOCK
1166 if (compat)
1167 ((struct osockaddr *)sa)->sa_family = sa->sa_family;
1168#endif
1169 error = copyout(sa, (caddr_t)uap->asa, (u_int)len);
1170 if (error == 0)
1171gotnothing:
1172 error = copyout((caddr_t)&len, (caddr_t)uap->alen,
1173 sizeof (len));
1174bad:
1175 if (sa)
1176 FREE(sa, M_SONAME);
dadab5e9 1177 fdrop(fp, td);
984263bc
MD
1178 return (error);
1179}
1180
1181int
41c20dac 1182getsockname(struct getsockname_args *uap)
984263bc
MD
1183{
1184
41c20dac 1185 return (getsockname1(uap, 0));
984263bc
MD
1186}
1187
1188#ifdef COMPAT_OLDSOCK
1189int
41c20dac 1190ogetsockname(struct getsockname_args *uap)
984263bc
MD
1191{
1192
41c20dac 1193 return (getsockname1(uap, 1));
984263bc
MD
1194}
1195#endif /* COMPAT_OLDSOCK */
1196
1197/*
41c20dac
MD
1198 * getpeername_args(int fdes, caddr_t asa, int *alen)
1199 *
984263bc
MD
1200 * Get name of peer for connected socket.
1201 */
1202/* ARGSUSED */
1203static int
41c20dac 1204getpeername1(struct getpeername_args *uap, int compat)
984263bc 1205{
dadab5e9
MD
1206 struct thread *td = curthread;
1207 struct proc *p = td->td_proc;
984263bc 1208 struct file *fp;
dadab5e9 1209 struct socket *so;
984263bc
MD
1210 struct sockaddr *sa;
1211 int len, error;
1212
1213 error = holdsock(p->p_fd, uap->fdes, &fp);
1214 if (error)
1215 return (error);
1216 so = (struct socket *)fp->f_data;
1217 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
dadab5e9 1218 fdrop(fp, td);
984263bc
MD
1219 return (ENOTCONN);
1220 }
1221 error = copyin((caddr_t)uap->alen, (caddr_t)&len, sizeof (len));
1222 if (error) {
dadab5e9 1223 fdrop(fp, td);
984263bc
MD
1224 return (error);
1225 }
1226 if (len < 0) {
dadab5e9 1227 fdrop(fp, td);
984263bc
MD
1228 return (EINVAL);
1229 }
1230 sa = 0;
1231 error = (*so->so_proto->pr_usrreqs->pru_peeraddr)(so, &sa);
1232 if (error)
1233 goto bad;
1234 if (sa == 0) {
1235 len = 0;
1236 goto gotnothing;
1237 }
1238 len = MIN(len, sa->sa_len);
1239#ifdef COMPAT_OLDSOCK
1240 if (compat)
1241 ((struct osockaddr *)sa)->sa_family =
1242 sa->sa_family;
1243#endif
1244 error = copyout(sa, (caddr_t)uap->asa, (u_int)len);
1245 if (error)
1246 goto bad;
1247gotnothing:
1248 error = copyout((caddr_t)&len, (caddr_t)uap->alen, sizeof (len));
1249bad:
1250 if (sa)
1251 FREE(sa, M_SONAME);
dadab5e9 1252 fdrop(fp, td);
984263bc
MD
1253 return (error);
1254}
1255
1256int
41c20dac 1257getpeername(struct getpeername_args *uap)
984263bc 1258{
41c20dac 1259 return (getpeername1(uap, 0));
984263bc
MD
1260}
1261
1262#ifdef COMPAT_OLDSOCK
1263int
41c20dac 1264ogetpeername(struct ogetpeername_args *uap)
984263bc 1265{
984263bc 1266 /* XXX uap should have type `getpeername_args *' to begin with. */
41c20dac 1267 return (getpeername1((struct getpeername_args *)uap, 1));
984263bc
MD
1268}
1269#endif /* COMPAT_OLDSOCK */
1270
1271int
1272sockargs(mp, buf, buflen, type)
1273 struct mbuf **mp;
1274 caddr_t buf;
1275 int buflen, type;
1276{
1277 register struct sockaddr *sa;
1278 register struct mbuf *m;
1279 int error;
1280
1281 if ((u_int)buflen > MLEN) {
1282#ifdef COMPAT_OLDSOCK
1283 if (type == MT_SONAME && (u_int)buflen <= 112)
1284 buflen = MLEN; /* unix domain compat. hack */
1285 else
1286#endif
1287 return (EINVAL);
1288 }
1289 m = m_get(M_WAIT, type);
1290 if (m == NULL)
1291 return (ENOBUFS);
1292 m->m_len = buflen;
1293 error = copyin(buf, mtod(m, caddr_t), (u_int)buflen);
1294 if (error)
1295 (void) m_free(m);
1296 else {
1297 *mp = m;
1298 if (type == MT_SONAME) {
1299 sa = mtod(m, struct sockaddr *);
1300
1301#if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
1302 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1303 sa->sa_family = sa->sa_len;
1304#endif
1305 sa->sa_len = buflen;
1306 }
1307 }
1308 return (error);
1309}
1310
1311int
1312getsockaddr(namp, uaddr, len)
1313 struct sockaddr **namp;
1314 caddr_t uaddr;
1315 size_t len;
1316{
1317 struct sockaddr *sa;
1318 int error;
1319
1320 if (len > SOCK_MAXADDRLEN)
1321 return ENAMETOOLONG;
1322 MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1323 error = copyin(uaddr, sa, len);
1324 if (error) {
1325 FREE(sa, M_SONAME);
1326 } else {
1327#if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
1328 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1329 sa->sa_family = sa->sa_len;
1330#endif
1331 sa->sa_len = len;
1332 *namp = sa;
1333 }
1334 return error;
1335}
1336
1337/*
1338 * holdsock() - load the struct file pointer associated
1339 * with a socket into *fpp. If an error occurs, non-zero
1340 * will be returned and *fpp will be set to NULL.
1341 */
1342int
1343holdsock(fdp, fdes, fpp)
1344 struct filedesc *fdp;
1345 int fdes;
1346 struct file **fpp;
1347{
1348 register struct file *fp = NULL;
1349 int error = 0;
1350
1351 if ((unsigned)fdes >= fdp->fd_nfiles ||
1352 (fp = fdp->fd_ofiles[fdes]) == NULL) {
1353 error = EBADF;
1354 } else if (fp->f_type != DTYPE_SOCKET) {
1355 error = ENOTSOCK;
1356 fp = NULL;
1357 } else {
1358 fhold(fp);
1359 }
1360 *fpp = fp;
1361 return(error);
1362}
1363
1364/*
1365 * Allocate a pool of sf_bufs (sendfile(2) or "super-fast" if you prefer. :-))
1366 */
1367static void
1368sf_buf_init(void *arg)
1369{
1370 int i;
1371
1372 SLIST_INIT(&sf_freelist);
1373 sf_base = kmem_alloc_pageable(kernel_map, nsfbufs * PAGE_SIZE);
1374 sf_bufs = malloc(nsfbufs * sizeof(struct sf_buf), M_TEMP, M_NOWAIT);
1375 bzero(sf_bufs, nsfbufs * sizeof(struct sf_buf));
1376 for (i = 0; i < nsfbufs; i++) {
1377 sf_bufs[i].kva = sf_base + i * PAGE_SIZE;
1378 SLIST_INSERT_HEAD(&sf_freelist, &sf_bufs[i], free_list);
1379 }
1380}
1381
1382/*
1383 * Get an sf_buf from the freelist. Will block if none are available.
1384 */
1385struct sf_buf *
1386sf_buf_alloc()
1387{
1388 struct sf_buf *sf;
1389 int s;
1390 int error;
1391
1392 s = splimp();
1393 while ((sf = SLIST_FIRST(&sf_freelist)) == NULL) {
1394 sf_buf_alloc_want = 1;
1395 error = tsleep(&sf_freelist, PVM|PCATCH, "sfbufa", 0);
1396 if (error)
1397 break;
1398 }
1399 if (sf != NULL) {
1400 SLIST_REMOVE_HEAD(&sf_freelist, free_list);
1401 sf->refcnt = 1;
1402 }
1403 splx(s);
1404 return (sf);
1405}
1406
1407#define dtosf(x) (&sf_bufs[((uintptr_t)(x) - (uintptr_t)sf_base) >> PAGE_SHIFT])
1408void
1409sf_buf_ref(caddr_t addr, u_int size)
1410{
1411 struct sf_buf *sf;
1412
1413 sf = dtosf(addr);
1414 if (sf->refcnt == 0)
1415 panic("sf_buf_ref: referencing a free sf_buf");
1416 sf->refcnt++;
1417}
1418
1419/*
1420 * Lose a reference to an sf_buf. When none left, detach mapped page
1421 * and release resources back to the system.
1422 *
1423 * Must be called at splimp.
1424 */
1425void
1426sf_buf_free(caddr_t addr, u_int size)
1427{
1428 struct sf_buf *sf;
1429 struct vm_page *m;
1430 int s;
1431
1432 sf = dtosf(addr);
1433 if (sf->refcnt == 0)
1434 panic("sf_buf_free: freeing free sf_buf");
1435 sf->refcnt--;
1436 if (sf->refcnt == 0) {
1437 pmap_qremove((vm_offset_t)addr, 1);
1438 m = sf->m;
1439 s = splvm();
1440 vm_page_unwire(m, 0);
1441 /*
1442 * Check for the object going away on us. This can
1443 * happen since we don't hold a reference to it.
1444 * If so, we're responsible for freeing the page.
1445 */
1446 if (m->wire_count == 0 && m->object == NULL)
1447 vm_page_free(m);
1448 splx(s);
1449 sf->m = NULL;
1450 SLIST_INSERT_HEAD(&sf_freelist, sf, free_list);
1451 if (sf_buf_alloc_want) {
1452 sf_buf_alloc_want = 0;
1453 wakeup(&sf_freelist);
1454 }
1455 }
1456}
1457
1458/*
1459 * sendfile(2).
1460 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1461 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1462 *
1463 * Send a file specified by 'fd' and starting at 'offset' to a socket
1464 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1465 * nbytes == 0. Optionally add a header and/or trailer to the socket
1466 * output. If specified, write the total number of bytes sent into *sbytes.
1467 */
1468int
41c20dac 1469sendfile(struct sendfile_args *uap)
984263bc 1470{
41c20dac 1471 return (do_sendfile(uap, 0));
984263bc
MD
1472}
1473
1474#ifdef COMPAT_43
1475int
41c20dac 1476osendfile(struct osendfile_args *uap)
984263bc
MD
1477{
1478 struct sendfile_args args;
1479
1480 args.fd = uap->fd;
1481 args.s = uap->s;
1482 args.offset = uap->offset;
1483 args.nbytes = uap->nbytes;
1484 args.hdtr = uap->hdtr;
1485 args.sbytes = uap->sbytes;
1486 args.flags = uap->flags;
1487
41c20dac 1488 return (do_sendfile(&args, 1));
984263bc
MD
1489}
1490#endif
1491
1492int
41c20dac 1493do_sendfile(struct sendfile_args *uap, int compat)
984263bc 1494{
dadab5e9
MD
1495 struct thread *td = curthread;
1496 struct proc *p = td->td_proc;
984263bc 1497 struct file *fp;
dadab5e9 1498 struct filedesc *fdp;
984263bc
MD
1499 struct vnode *vp;
1500 struct vm_object *obj;
1501 struct socket *so;
1502 struct mbuf *m;
1503 struct sf_buf *sf;
1504 struct vm_page *pg;
1505 struct writev_args nuap;
1506 struct sf_hdtr hdtr;
1507 off_t off, xfsize, hdtr_size, sbytes = 0;
1508 int error = 0, s;
1509
dadab5e9
MD
1510 KKASSERT(p);
1511 fdp = p->p_fd;
1512
984263bc
MD
1513 vp = NULL;
1514 hdtr_size = 0;
1515 /*
1516 * Do argument checking. Must be a regular file in, stream
1517 * type and connected socket out, positive offset.
1518 */
1519 fp = holdfp(fdp, uap->fd, FREAD);
1520 if (fp == NULL) {
1521 error = EBADF;
1522 goto done;
1523 }
1524 if (fp->f_type != DTYPE_VNODE) {
1525 error = EINVAL;
1526 goto done;
1527 }
1528 vp = (struct vnode *)fp->f_data;
1529 vref(vp);
1530 if (vp->v_type != VREG || VOP_GETVOBJECT(vp, &obj) != 0) {
1531 error = EINVAL;
1532 goto done;
1533 }
dadab5e9 1534 fdrop(fp, td);
984263bc
MD
1535 error = holdsock(p->p_fd, uap->s, &fp);
1536 if (error)
1537 goto done;
1538 so = (struct socket *)fp->f_data;
1539 if (so->so_type != SOCK_STREAM) {
1540 error = EINVAL;
1541 goto done;
1542 }
1543 if ((so->so_state & SS_ISCONNECTED) == 0) {
1544 error = ENOTCONN;
1545 goto done;
1546 }
1547 if (uap->offset < 0) {
1548 error = EINVAL;
1549 goto done;
1550 }
1551
1552 /*
1553 * If specified, get the pointer to the sf_hdtr struct for
1554 * any headers/trailers.
1555 */
1556 if (uap->hdtr != NULL) {
1557 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1558 if (error)
1559 goto done;
1560 /*
1561 * Send any headers. Wimp out and use writev(2).
1562 */
1563 if (hdtr.headers != NULL) {
1564 nuap.fd = uap->s;
1565 nuap.iovp = hdtr.headers;
1566 nuap.iovcnt = hdtr.hdr_cnt;
41c20dac 1567 error = writev(&nuap);
984263bc
MD
1568 if (error)
1569 goto done;
1570 if (compat)
1571 sbytes += p->p_retval[0];
1572 else
1573 hdtr_size += p->p_retval[0];
1574 }
1575 }
1576
1577 /*
1578 * Protect against multiple writers to the socket.
1579 */
1580 (void) sblock(&so->so_snd, M_WAITOK);
1581
1582 /*
1583 * Loop through the pages in the file, starting with the requested
1584 * offset. Get a file page (do I/O if necessary), map the file page
1585 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1586 * it on the socket.
1587 */
1588 for (off = uap->offset; ; off += xfsize, sbytes += xfsize) {
1589 vm_pindex_t pindex;
1590 vm_offset_t pgoff;
1591
1592 pindex = OFF_TO_IDX(off);
1593retry_lookup:
1594 /*
1595 * Calculate the amount to transfer. Not to exceed a page,
1596 * the EOF, or the passed in nbytes.
1597 */
1598 xfsize = obj->un_pager.vnp.vnp_size - off;
1599 if (xfsize > PAGE_SIZE)
1600 xfsize = PAGE_SIZE;
1601 pgoff = (vm_offset_t)(off & PAGE_MASK);
1602 if (PAGE_SIZE - pgoff < xfsize)
1603 xfsize = PAGE_SIZE - pgoff;
1604 if (uap->nbytes && xfsize > (uap->nbytes - sbytes))
1605 xfsize = uap->nbytes - sbytes;
1606 if (xfsize <= 0)
1607 break;
1608 /*
1609 * Optimize the non-blocking case by looking at the socket space
1610 * before going to the extra work of constituting the sf_buf.
1611 */
1612 if ((so->so_state & SS_NBIO) && sbspace(&so->so_snd) <= 0) {
1613 if (so->so_state & SS_CANTSENDMORE)
1614 error = EPIPE;
1615 else
1616 error = EAGAIN;
1617 sbunlock(&so->so_snd);
1618 goto done;
1619 }
1620 /*
1621 * Attempt to look up the page.
1622 *
1623 * Allocate if not found
1624 *
1625 * Wait and loop if busy.
1626 */
1627 pg = vm_page_lookup(obj, pindex);
1628
1629 if (pg == NULL) {
1630 pg = vm_page_alloc(obj, pindex, VM_ALLOC_NORMAL);
1631 if (pg == NULL) {
1632 VM_WAIT;
1633 goto retry_lookup;
1634 }
1635 vm_page_wakeup(pg);
1636 } else if (vm_page_sleep_busy(pg, TRUE, "sfpbsy")) {
1637 goto retry_lookup;
1638 }
1639
1640 /*
1641 * Wire the page so it does not get ripped out from under
1642 * us.
1643 */
1644
1645 vm_page_wire(pg);
1646
1647 /*
1648 * If page is not valid for what we need, initiate I/O
1649 */
1650
1651 if (!pg->valid || !vm_page_is_valid(pg, pgoff, xfsize)) {
1652 struct uio auio;
1653 struct iovec aiov;
1654 int bsize;
1655
1656 /*
1657 * Ensure that our page is still around when the I/O
1658 * completes.
1659 */
1660 vm_page_io_start(pg);
1661
1662 /*
1663 * Get the page from backing store.
1664 */
1665 bsize = vp->v_mount->mnt_stat.f_iosize;
1666 auio.uio_iov = &aiov;
1667 auio.uio_iovcnt = 1;
1668 aiov.iov_base = 0;
1669 aiov.iov_len = MAXBSIZE;
1670 auio.uio_resid = MAXBSIZE;
1671 auio.uio_offset = trunc_page(off);
1672 auio.uio_segflg = UIO_NOCOPY;
1673 auio.uio_rw = UIO_READ;
dadab5e9
MD
1674 auio.uio_td = td;
1675 vn_lock(vp, LK_SHARED | LK_NOPAUSE | LK_RETRY, td);
1676 error = VOP_READ(vp, &auio,
1677 IO_VMIO | ((MAXBSIZE / bsize) << 16),
1678 p->p_ucred);
1679 VOP_UNLOCK(vp, 0, td);
984263bc
MD
1680 vm_page_flag_clear(pg, PG_ZERO);
1681 vm_page_io_finish(pg);
1682 if (error) {
1683 vm_page_unwire(pg, 0);
1684 /*
1685 * See if anyone else might know about this page.
1686 * If not and it is not valid, then free it.
1687 */
1688 if (pg->wire_count == 0 && pg->valid == 0 &&
1689 pg->busy == 0 && !(pg->flags & PG_BUSY) &&
1690 pg->hold_count == 0) {
1691 vm_page_busy(pg);
1692 vm_page_free(pg);
1693 }
1694 sbunlock(&so->so_snd);
1695 goto done;
1696 }
1697 }
1698
1699
1700 /*
1701 * Get a sendfile buf. We usually wait as long as necessary,
1702 * but this wait can be interrupted.
1703 */
1704 if ((sf = sf_buf_alloc()) == NULL) {
1705 s = splvm();
1706 vm_page_unwire(pg, 0);
1707 if (pg->wire_count == 0 && pg->object == NULL)
1708 vm_page_free(pg);
1709 splx(s);
1710 sbunlock(&so->so_snd);
1711 error = EINTR;
1712 goto done;
1713 }
1714
1715
1716 /*
1717 * Allocate a kernel virtual page and insert the physical page
1718 * into it.
1719 */
1720
1721 sf->m = pg;
1722 pmap_qenter(sf->kva, &pg, 1);
1723 /*
1724 * Get an mbuf header and set it up as having external storage.
1725 */
1726 MGETHDR(m, M_WAIT, MT_DATA);
1727 if (m == NULL) {
1728 error = ENOBUFS;
1729 sf_buf_free((void *)sf->kva, PAGE_SIZE);
1730 sbunlock(&so->so_snd);
1731 goto done;
1732 }
1733 m->m_ext.ext_free = sf_buf_free;
1734 m->m_ext.ext_ref = sf_buf_ref;
1735 m->m_ext.ext_buf = (void *)sf->kva;
1736 m->m_ext.ext_size = PAGE_SIZE;
1737 m->m_data = (char *) sf->kva + pgoff;
1738 m->m_flags |= M_EXT;
1739 m->m_pkthdr.len = m->m_len = xfsize;
1740 /*
1741 * Add the buffer to the socket buffer chain.
1742 */
1743 s = splnet();
1744retry_space:
1745 /*
1746 * Make sure that the socket is still able to take more data.
1747 * CANTSENDMORE being true usually means that the connection
1748 * was closed. so_error is true when an error was sensed after
1749 * a previous send.
1750 * The state is checked after the page mapping and buffer
1751 * allocation above since those operations may block and make
1752 * any socket checks stale. From this point forward, nothing
1753 * blocks before the pru_send (or more accurately, any blocking
1754 * results in a loop back to here to re-check).
1755 */
1756 if ((so->so_state & SS_CANTSENDMORE) || so->so_error) {
1757 if (so->so_state & SS_CANTSENDMORE) {
1758 error = EPIPE;
1759 } else {
1760 error = so->so_error;
1761 so->so_error = 0;
1762 }
1763 m_freem(m);
1764 sbunlock(&so->so_snd);
1765 splx(s);
1766 goto done;
1767 }
1768 /*
1769 * Wait for socket space to become available. We do this just
1770 * after checking the connection state above in order to avoid
1771 * a race condition with sbwait().
1772 */
1773 if (sbspace(&so->so_snd) < so->so_snd.sb_lowat) {
1774 if (so->so_state & SS_NBIO) {
1775 m_freem(m);
1776 sbunlock(&so->so_snd);
1777 splx(s);
1778 error = EAGAIN;
1779 goto done;
1780 }
1781 error = sbwait(&so->so_snd);
1782 /*
1783 * An error from sbwait usually indicates that we've
1784 * been interrupted by a signal. If we've sent anything
1785 * then return bytes sent, otherwise return the error.
1786 */
1787 if (error) {
1788 m_freem(m);
1789 sbunlock(&so->so_snd);
1790 splx(s);
1791 goto done;
1792 }
1793 goto retry_space;
1794 }
dadab5e9
MD
1795 error =
1796 (*so->so_proto->pr_usrreqs->pru_send)(so, 0, m, 0, 0, td);
984263bc
MD
1797 splx(s);
1798 if (error) {
1799 sbunlock(&so->so_snd);
1800 goto done;
1801 }
1802 }
1803 sbunlock(&so->so_snd);
1804
1805 /*
1806 * Send trailers. Wimp out and use writev(2).
1807 */
1808 if (uap->hdtr != NULL && hdtr.trailers != NULL) {
1809 nuap.fd = uap->s;
1810 nuap.iovp = hdtr.trailers;
1811 nuap.iovcnt = hdtr.trl_cnt;
41c20dac 1812 error = writev(&nuap);
984263bc
MD
1813 if (error)
1814 goto done;
1815 if (compat)
1816 sbytes += p->p_retval[0];
1817 else
1818 hdtr_size += p->p_retval[0];
1819 }
1820
1821done:
1822 if (uap->sbytes != NULL) {
1823 if (compat == 0)
1824 sbytes += hdtr_size;
1825 copyout(&sbytes, uap->sbytes, sizeof(off_t));
1826 }
1827 if (vp)
1828 vrele(vp);
1829 if (fp)
dadab5e9 1830 fdrop(fp, td);
984263bc
MD
1831 return (error);
1832}