2 * Copyright (c) 1982, 1986, 1989, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
5 * sendfile(2) and related extensions:
6 * Copyright (c) 1998, David Greenman. All rights reserved.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
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 * 4. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * @(#)uipc_syscalls.c 8.4 (Berkeley) 2/21/94
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
39 #include "opt_compat.h"
40 #include "opt_ktrace.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
47 #include <sys/mutex.h>
48 #include <sys/sysproto.h>
49 #include <sys/malloc.h>
50 #include <sys/filedesc.h>
51 #include <sys/event.h>
53 #include <sys/fcntl.h>
55 #include <sys/filio.h>
56 #include <sys/mount.h>
58 #include <sys/protosw.h>
59 #include <sys/sf_buf.h>
60 #include <sys/socket.h>
61 #include <sys/socketvar.h>
62 #include <sys/signalvar.h>
63 #include <sys/syscallsubr.h>
64 #include <sys/sysctl.h>
66 #include <sys/vnode.h>
68 #include <sys/ktrace.h>
71 #include <security/mac/mac_framework.h>
74 #include <vm/vm_object.h>
75 #include <vm/vm_page.h>
76 #include <vm/vm_pageout.h>
77 #include <vm/vm_kern.h>
78 #include <vm/vm_extern.h>
81 #include <netinet/sctp.h>
82 #include <netinet/sctp_peeloff.h>
85 static int sendit(struct thread *td, int s, struct msghdr *mp, int flags);
86 static int recvit(struct thread *td, int s, struct msghdr *mp, void *namelenp);
88 static int accept1(struct thread *td, struct accept_args *uap, int compat);
89 static int do_sendfile(struct thread *td, struct sendfile_args *uap, int compat);
90 static int getsockname1(struct thread *td, struct getsockname_args *uap,
92 static int getpeername1(struct thread *td, struct getpeername_args *uap,
96 * NSFBUFS-related variables and associated sysctls
102 SYSCTL_INT(_kern_ipc, OID_AUTO, nsfbufs, CTLFLAG_RDTUN, &nsfbufs, 0,
103 "Maximum number of sendfile(2) sf_bufs available");
104 SYSCTL_INT(_kern_ipc, OID_AUTO, nsfbufspeak, CTLFLAG_RD, &nsfbufspeak, 0,
105 "Number of sendfile(2) sf_bufs at peak usage");
106 SYSCTL_INT(_kern_ipc, OID_AUTO, nsfbufsused, CTLFLAG_RD, &nsfbufsused, 0,
107 "Number of sendfile(2) sf_bufs in use");
110 * Convert a user file descriptor to a kernel file entry. A reference on the
111 * file entry is held upon returning. This is lighter weight than
112 * fgetsock(), which bumps the socket reference drops the file reference
113 * count instead, as this approach avoids several additional mutex operations
114 * associated with the additional reference count. If requested, return the
118 getsock(struct filedesc *fdp, int fd, struct file **fpp, u_int *fflagp)
128 fp = fget_locked(fdp, fd);
131 else if (fp->f_type != DTYPE_SOCKET) {
137 *fflagp = fp->f_flag;
140 FILEDESC_SUNLOCK(fdp);
147 * System call interface to the socket abstraction.
149 #if defined(COMPAT_43)
150 #define COMPAT_OLDSOCK
156 struct socket_args /* {
162 struct filedesc *fdp;
168 error = mac_check_socket_create(td->td_ucred, uap->domain, uap->type,
173 fdp = td->td_proc->p_fd;
174 error = falloc(td, &fp, &fd);
177 /* An extra reference on `fp' has been held for us by falloc(). */
179 error = socreate(uap->domain, &so, uap->type, uap->protocol,
183 fdclose(fdp, fp, fd, td);
186 fp->f_data = so; /* already has ref count */
187 fp->f_flag = FREAD|FWRITE;
188 fp->f_type = DTYPE_SOCKET;
189 fp->f_ops = &socketops;
191 td->td_retval[0] = fd;
201 struct bind_args /* {
210 if ((error = getsockaddr(&sa, uap->name, uap->namelen)) != 0)
213 error = kern_bind(td, uap->s, sa);
219 kern_bind(td, fd, sa)
229 error = getsock(td->td_proc->p_fd, fd, &fp, NULL);
235 error = mac_check_socket_bind(td->td_ucred, so, sa);
240 error = sobind(so, sa, td);
254 struct listen_args /* {
264 error = getsock(td->td_proc->p_fd, uap->s, &fp, NULL);
269 error = mac_check_socket_listen(td->td_ucred, so);
274 error = solisten(so, uap->backlog, td);
288 accept1(td, uap, compat)
290 struct accept_args /* {
292 struct sockaddr * __restrict name;
293 socklen_t * __restrict anamelen;
297 struct sockaddr *name;
302 if (uap->name == NULL)
303 return (kern_accept(td, uap->s, NULL, NULL, NULL));
305 error = copyin(uap->anamelen, &namelen, sizeof (namelen));
309 error = kern_accept(td, uap->s, &name, &namelen, &fp);
312 * return a namelen of zero for older code which might
313 * ignore the return value from accept.
316 (void) copyout(&namelen,
317 uap->anamelen, sizeof(*uap->anamelen));
321 if (error == 0 && name != NULL) {
322 #ifdef COMPAT_OLDSOCK
324 ((struct osockaddr *)name)->sa_family =
327 error = copyout(name, uap->name, namelen);
330 error = copyout(&namelen, uap->anamelen,
333 fdclose(td->td_proc->p_fd, fp, td->td_retval[0], td);
335 free(name, M_SONAME);
340 kern_accept(struct thread *td, int s, struct sockaddr **name,
341 socklen_t *namelen, struct file **fp)
343 struct filedesc *fdp;
344 struct file *headfp, *nfp = NULL;
345 struct sockaddr *sa = NULL;
347 struct socket *head, *so;
359 fdp = td->td_proc->p_fd;
361 error = getsock(fdp, s, &headfp, &fflag);
364 head = headfp->f_data;
365 if ((head->so_options & SO_ACCEPTCONN) == 0) {
371 error = mac_check_socket_accept(td->td_ucred, head);
376 error = falloc(td, &nfp, &fd);
380 if ((head->so_state & SS_NBIO) && TAILQ_EMPTY(&head->so_comp)) {
385 while (TAILQ_EMPTY(&head->so_comp) && head->so_error == 0) {
386 if (head->so_rcv.sb_state & SBS_CANTRCVMORE) {
387 head->so_error = ECONNABORTED;
390 error = msleep(&head->so_timeo, &accept_mtx, PSOCK | PCATCH,
397 if (head->so_error) {
398 error = head->so_error;
403 so = TAILQ_FIRST(&head->so_comp);
404 KASSERT(!(so->so_qstate & SQ_INCOMP), ("accept1: so SQ_INCOMP"));
405 KASSERT(so->so_qstate & SQ_COMP, ("accept1: so not SQ_COMP"));
408 * Before changing the flags on the socket, we have to bump the
409 * reference count. Otherwise, if the protocol calls sofree(),
410 * the socket will be released due to a zero refcount.
412 SOCK_LOCK(so); /* soref() and so_state update */
413 soref(so); /* file descriptor reference */
415 TAILQ_REMOVE(&head->so_comp, so, so_list);
417 so->so_state |= (head->so_state & SS_NBIO);
418 so->so_qstate &= ~SQ_COMP;
424 /* An extra reference on `nfp' has been held for us by falloc(). */
425 td->td_retval[0] = fd;
427 /* connection has been removed from the listen queue */
428 KNOTE_UNLOCKED(&head->so_rcv.sb_sel.si_note, 0);
430 pgid = fgetown(&head->so_sigio);
432 fsetown(pgid, &so->so_sigio);
435 nfp->f_data = so; /* nfp has ref count from falloc */
437 nfp->f_type = DTYPE_SOCKET;
438 nfp->f_ops = &socketops;
440 /* Sync socket nonblocking/async state with file flags */
441 tmp = fflag & FNONBLOCK;
442 (void) fo_ioctl(nfp, FIONBIO, &tmp, td->td_ucred, td);
443 tmp = fflag & FASYNC;
444 (void) fo_ioctl(nfp, FIOASYNC, &tmp, td->td_ucred, td);
446 error = soaccept(so, &sa);
449 * return a namelen of zero for older code which might
450 * ignore the return value from accept.
462 /* check sa_len before it is destroyed */
463 if (*namelen > sa->sa_len)
464 *namelen = sa->sa_len;
473 * close the new descriptor, assuming someone hasn't ripped it
477 fdclose(fdp, nfp, fd, td);
480 * Release explicitly held references before returning. We return
481 * a reference on nfp to the caller on success if they request it.
502 struct accept_args *uap;
505 return (accept1(td, uap, 0));
508 #ifdef COMPAT_OLDSOCK
512 struct accept_args *uap;
515 return (accept1(td, uap, 1));
517 #endif /* COMPAT_OLDSOCK */
523 struct connect_args /* {
532 error = getsockaddr(&sa, uap->name, uap->namelen);
536 error = kern_connect(td, uap->s, sa);
543 kern_connect(td, fd, sa)
554 error = getsock(td->td_proc->p_fd, fd, &fp, NULL);
558 if (so->so_state & SS_ISCONNECTING) {
564 error = mac_check_socket_connect(td->td_ucred, so, sa);
569 error = soconnect(so, sa, td);
572 if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
577 while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
578 error = msleep(&so->so_timeo, SOCK_MTX(so), PSOCK | PCATCH,
581 if (error == EINTR || error == ERESTART)
587 error = so->so_error;
593 so->so_state &= ~SS_ISCONNECTING;
594 if (error == ERESTART)
606 struct socketpair_args /* {
613 struct filedesc *fdp = td->td_proc->p_fd;
614 struct file *fp1, *fp2;
615 struct socket *so1, *so2;
616 int fd, error, sv[2];
619 /* We might want to have a separate check for socket pairs. */
620 error = mac_check_socket_create(td->td_ucred, uap->domain, uap->type,
627 error = socreate(uap->domain, &so1, uap->type, uap->protocol,
631 error = socreate(uap->domain, &so2, uap->type, uap->protocol,
635 /* On success extra reference to `fp1' and 'fp2' is set by falloc. */
636 error = falloc(td, &fp1, &fd);
640 fp1->f_data = so1; /* so1 already has ref count */
641 error = falloc(td, &fp2, &fd);
644 fp2->f_data = so2; /* so2 already has ref count */
646 error = soconnect2(so1, so2);
649 if (uap->type == SOCK_DGRAM) {
651 * Datagram socket connection is asymmetric.
653 error = soconnect2(so2, so1);
658 fp1->f_flag = FREAD|FWRITE;
659 fp1->f_type = DTYPE_SOCKET;
660 fp1->f_ops = &socketops;
663 fp2->f_flag = FREAD|FWRITE;
664 fp2->f_type = DTYPE_SOCKET;
665 fp2->f_ops = &socketops;
668 error = copyout(sv, uap->rsv, 2 * sizeof (int));
675 fdclose(fdp, fp2, sv[1], td);
678 fdclose(fdp, fp1, sv[0], td);
692 sendit(td, s, mp, flags)
698 struct mbuf *control;
702 if (mp->msg_name != NULL) {
703 error = getsockaddr(&to, mp->msg_name, mp->msg_namelen);
713 if (mp->msg_control) {
714 if (mp->msg_controllen < sizeof(struct cmsghdr)
715 #ifdef COMPAT_OLDSOCK
716 && mp->msg_flags != MSG_COMPAT
722 error = sockargs(&control, mp->msg_control,
723 mp->msg_controllen, MT_CONTROL);
726 #ifdef COMPAT_OLDSOCK
727 if (mp->msg_flags == MSG_COMPAT) {
730 M_PREPEND(control, sizeof(*cm), M_TRYWAIT);
735 cm = mtod(control, struct cmsghdr *);
736 cm->cmsg_len = control->m_len;
737 cm->cmsg_level = SOL_SOCKET;
738 cm->cmsg_type = SCM_RIGHTS;
746 error = kern_sendit(td, s, mp, flags, control, UIO_USERSPACE);
755 kern_sendit(td, s, mp, flags, control, segflg)
760 struct mbuf *control;
770 struct uio *ktruio = NULL;
774 error = getsock(td->td_proc->p_fd, s, &fp, NULL);
777 so = (struct socket *)fp->f_data;
781 error = mac_check_socket_send(td->td_ucred, so);
787 auio.uio_iov = mp->msg_iov;
788 auio.uio_iovcnt = mp->msg_iovlen;
789 auio.uio_segflg = segflg;
790 auio.uio_rw = UIO_WRITE;
792 auio.uio_offset = 0; /* XXX */
795 for (i = 0; i < mp->msg_iovlen; i++, iov++) {
796 if ((auio.uio_resid += iov->iov_len) < 0) {
802 if (KTRPOINT(td, KTR_GENIO))
803 ktruio = cloneuio(&auio);
805 len = auio.uio_resid;
806 error = sosend(so, mp->msg_name, &auio, 0, control, flags, td);
808 if (auio.uio_resid != len && (error == ERESTART ||
809 error == EINTR || error == EWOULDBLOCK))
811 /* Generation of SIGPIPE can be controlled per socket */
812 if (error == EPIPE && !(so->so_options & SO_NOSIGPIPE) &&
813 !(flags & MSG_NOSIGNAL)) {
814 PROC_LOCK(td->td_proc);
815 psignal(td->td_proc, SIGPIPE);
816 PROC_UNLOCK(td->td_proc);
820 td->td_retval[0] = len - auio.uio_resid;
822 if (ktruio != NULL) {
823 ktruio->uio_resid = td->td_retval[0];
824 ktrgenio(s, UIO_WRITE, ktruio, error);
837 struct sendto_args /* {
850 msg.msg_name = uap->to;
851 msg.msg_namelen = uap->tolen;
855 #ifdef COMPAT_OLDSOCK
858 aiov.iov_base = uap->buf;
859 aiov.iov_len = uap->len;
860 error = sendit(td, uap->s, &msg, uap->flags);
864 #ifdef COMPAT_OLDSOCK
868 struct osend_args /* {
883 aiov.iov_base = uap->buf;
884 aiov.iov_len = uap->len;
887 error = sendit(td, uap->s, &msg, uap->flags);
894 struct osendmsg_args /* {
904 error = copyin(uap->msg, &msg, sizeof (struct omsghdr));
907 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
911 msg.msg_flags = MSG_COMPAT;
912 error = sendit(td, uap->s, &msg, uap->flags);
921 struct sendmsg_args /* {
931 error = copyin(uap->msg, &msg, sizeof (msg));
934 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
938 #ifdef COMPAT_OLDSOCK
941 error = sendit(td, uap->s, &msg, uap->flags);
947 kern_recvit(td, s, mp, fromseg, controlp)
951 enum uio_seg fromseg;
952 struct mbuf **controlp;
959 struct mbuf *m, *control = 0;
963 struct sockaddr *fromsa = 0;
965 struct uio *ktruio = NULL;
972 error = getsock(td->td_proc->p_fd, s, &fp, NULL);
981 error = mac_check_socket_receive(td->td_ucred, so);
990 auio.uio_iov = mp->msg_iov;
991 auio.uio_iovcnt = mp->msg_iovlen;
992 auio.uio_segflg = UIO_USERSPACE;
993 auio.uio_rw = UIO_READ;
995 auio.uio_offset = 0; /* XXX */
998 for (i = 0; i < mp->msg_iovlen; i++, iov++) {
999 if ((auio.uio_resid += iov->iov_len) < 0) {
1006 if (KTRPOINT(td, KTR_GENIO))
1007 ktruio = cloneuio(&auio);
1009 len = auio.uio_resid;
1010 error = soreceive(so, &fromsa, &auio, (struct mbuf **)0,
1011 (mp->msg_control || controlp) ? &control : (struct mbuf **)0,
1014 if (auio.uio_resid != (int)len && (error == ERESTART ||
1015 error == EINTR || error == EWOULDBLOCK))
1019 if (ktruio != NULL) {
1020 ktruio->uio_resid = (int)len - auio.uio_resid;
1021 ktrgenio(s, UIO_READ, ktruio, error);
1026 td->td_retval[0] = (int)len - auio.uio_resid;
1028 len = mp->msg_namelen;
1029 if (len <= 0 || fromsa == 0)
1032 /* save sa_len before it is destroyed by MSG_COMPAT */
1033 len = MIN(len, fromsa->sa_len);
1034 #ifdef COMPAT_OLDSOCK
1035 if (mp->msg_flags & MSG_COMPAT)
1036 ((struct osockaddr *)fromsa)->sa_family =
1039 if (fromseg == UIO_USERSPACE) {
1040 error = copyout(fromsa, mp->msg_name,
1045 bcopy(fromsa, mp->msg_name, len);
1047 mp->msg_namelen = len;
1049 if (mp->msg_control && controlp == NULL) {
1050 #ifdef COMPAT_OLDSOCK
1052 * We assume that old recvmsg calls won't receive access
1053 * rights and other control info, esp. as control info
1054 * is always optional and those options didn't exist in 4.3.
1055 * If we receive rights, trim the cmsghdr; anything else
1058 if (control && mp->msg_flags & MSG_COMPAT) {
1059 if (mtod(control, struct cmsghdr *)->cmsg_level !=
1061 mtod(control, struct cmsghdr *)->cmsg_type !=
1063 mp->msg_controllen = 0;
1066 control->m_len -= sizeof (struct cmsghdr);
1067 control->m_data += sizeof (struct cmsghdr);
1070 len = mp->msg_controllen;
1072 mp->msg_controllen = 0;
1073 ctlbuf = mp->msg_control;
1075 while (m && len > 0) {
1076 unsigned int tocopy;
1078 if (len >= m->m_len)
1081 mp->msg_flags |= MSG_CTRUNC;
1085 if ((error = copyout(mtod(m, caddr_t),
1086 ctlbuf, tocopy)) != 0)
1093 mp->msg_controllen = ctlbuf - (caddr_t)mp->msg_control;
1099 FREE(fromsa, M_SONAME);
1101 if (error == 0 && controlp != NULL)
1102 *controlp = control;
1110 recvit(td, s, mp, namelenp)
1118 error = kern_recvit(td, s, mp, UIO_USERSPACE, NULL);
1122 error = copyout(&mp->msg_namelen, namelenp, sizeof (socklen_t));
1123 #ifdef COMPAT_OLDSOCK
1124 if (mp->msg_flags & MSG_COMPAT)
1125 error = 0; /* old recvfrom didn't check */
1134 struct recvfrom_args /* {
1139 struct sockaddr * __restrict from;
1140 socklen_t * __restrict fromlenaddr;
1147 if (uap->fromlenaddr) {
1148 error = copyin(uap->fromlenaddr,
1149 &msg.msg_namelen, sizeof (msg.msg_namelen));
1153 msg.msg_namelen = 0;
1155 msg.msg_name = uap->from;
1156 msg.msg_iov = &aiov;
1158 aiov.iov_base = uap->buf;
1159 aiov.iov_len = uap->len;
1160 msg.msg_control = 0;
1161 msg.msg_flags = uap->flags;
1162 error = recvit(td, uap->s, &msg, uap->fromlenaddr);
1167 #ifdef COMPAT_OLDSOCK
1171 struct recvfrom_args *uap;
1174 uap->flags |= MSG_COMPAT;
1175 return (recvfrom(td, uap));
1179 #ifdef COMPAT_OLDSOCK
1183 struct orecv_args /* {
1195 msg.msg_namelen = 0;
1196 msg.msg_iov = &aiov;
1198 aiov.iov_base = uap->buf;
1199 aiov.iov_len = uap->len;
1200 msg.msg_control = 0;
1201 msg.msg_flags = uap->flags;
1202 error = recvit(td, uap->s, &msg, NULL);
1207 * Old recvmsg. This code takes advantage of the fact that the old msghdr
1208 * overlays the new one, missing only the flags, and with the (old) access
1209 * rights where the control fields are now.
1214 struct orecvmsg_args /* {
1216 struct omsghdr *msg;
1224 error = copyin(uap->msg, &msg, sizeof (struct omsghdr));
1227 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1230 msg.msg_flags = uap->flags | MSG_COMPAT;
1232 error = recvit(td, uap->s, &msg, &uap->msg->msg_namelen);
1233 if (msg.msg_controllen && error == 0)
1234 error = copyout(&msg.msg_controllen,
1235 &uap->msg->msg_accrightslen, sizeof (int));
1244 struct recvmsg_args /* {
1251 struct iovec *uiov, *iov;
1254 error = copyin(uap->msg, &msg, sizeof (msg));
1257 error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
1260 msg.msg_flags = uap->flags;
1261 #ifdef COMPAT_OLDSOCK
1262 msg.msg_flags &= ~MSG_COMPAT;
1266 error = recvit(td, uap->s, &msg, NULL);
1269 error = copyout(&msg, uap->msg, sizeof(msg));
1279 struct shutdown_args /* {
1289 error = getsock(td->td_proc->p_fd, uap->s, &fp, NULL);
1292 error = soshutdown(so, uap->how);
1303 struct setsockopt_args /* {
1312 return (kern_setsockopt(td, uap->s, uap->level, uap->name,
1313 uap->val, UIO_USERSPACE, uap->valsize));
1317 kern_setsockopt(td, s, level, name, val, valseg, valsize)
1323 enum uio_seg valseg;
1329 struct sockopt sopt;
1331 if (val == NULL && valsize != 0)
1333 if ((int)valsize < 0)
1336 sopt.sopt_dir = SOPT_SET;
1337 sopt.sopt_level = level;
1338 sopt.sopt_name = name;
1339 sopt.sopt_val = val;
1340 sopt.sopt_valsize = valsize;
1346 sopt.sopt_td = NULL;
1349 panic("kern_setsockopt called with bad valseg");
1353 error = getsock(td->td_proc->p_fd, s, &fp, NULL);
1356 error = sosetopt(so, &sopt);
1367 struct getsockopt_args /* {
1371 void * __restrict val;
1372 socklen_t * __restrict avalsize;
1379 error = copyin(uap->avalsize, &valsize, sizeof (valsize));
1384 error = kern_getsockopt(td, uap->s, uap->level, uap->name,
1385 uap->val, UIO_USERSPACE, &valsize);
1388 error = copyout(&valsize, uap->avalsize, sizeof (valsize));
1393 * Kernel version of getsockopt.
1394 * optval can be a userland or userspace. optlen is always a kernel pointer.
1397 kern_getsockopt(td, s, level, name, val, valseg, valsize)
1403 enum uio_seg valseg;
1409 struct sockopt sopt;
1413 if ((int)*valsize < 0)
1416 sopt.sopt_dir = SOPT_GET;
1417 sopt.sopt_level = level;
1418 sopt.sopt_name = name;
1419 sopt.sopt_val = val;
1420 sopt.sopt_valsize = (size_t)*valsize; /* checked non-negative above */
1426 sopt.sopt_td = NULL;
1429 panic("kern_getsockopt called with bad valseg");
1433 error = getsock(td->td_proc->p_fd, s, &fp, NULL);
1436 error = sogetopt(so, &sopt);
1437 *valsize = sopt.sopt_valsize;
1445 * getsockname1() - Get socket name.
1449 getsockname1(td, uap, compat)
1451 struct getsockname_args /* {
1453 struct sockaddr * __restrict asa;
1454 socklen_t * __restrict alen;
1458 struct sockaddr *sa;
1462 error = copyin(uap->alen, &len, sizeof(len));
1466 error = kern_getsockname(td, uap->fdes, &sa, &len);
1471 #ifdef COMPAT_OLDSOCK
1473 ((struct osockaddr *)sa)->sa_family = sa->sa_family;
1475 error = copyout(sa, uap->asa, (u_int)len);
1479 error = copyout(&len, uap->alen, sizeof(len));
1484 kern_getsockname(struct thread *td, int fd, struct sockaddr **sa,
1496 error = getsock(td->td_proc->p_fd, fd, &fp, NULL);
1501 error = (*so->so_proto->pr_usrreqs->pru_sockaddr)(so, sa);
1507 len = MIN(*alen, (*sa)->sa_len);
1512 free(*sa, M_SONAME);
1521 getsockname(td, uap)
1523 struct getsockname_args *uap;
1526 return (getsockname1(td, uap, 0));
1529 #ifdef COMPAT_OLDSOCK
1531 ogetsockname(td, uap)
1533 struct getsockname_args *uap;
1536 return (getsockname1(td, uap, 1));
1538 #endif /* COMPAT_OLDSOCK */
1541 * getpeername1() - Get name of peer for connected socket.
1545 getpeername1(td, uap, compat)
1547 struct getpeername_args /* {
1549 struct sockaddr * __restrict asa;
1550 socklen_t * __restrict alen;
1554 struct sockaddr *sa;
1558 error = copyin(uap->alen, &len, sizeof (len));
1562 error = kern_getpeername(td, uap->fdes, &sa, &len);
1567 #ifdef COMPAT_OLDSOCK
1569 ((struct osockaddr *)sa)->sa_family = sa->sa_family;
1571 error = copyout(sa, uap->asa, (u_int)len);
1575 error = copyout(&len, uap->alen, sizeof(len));
1580 kern_getpeername(struct thread *td, int fd, struct sockaddr **sa,
1592 error = getsock(td->td_proc->p_fd, fd, &fp, NULL);
1596 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1601 error = (*so->so_proto->pr_usrreqs->pru_peeraddr)(so, sa);
1607 len = MIN(*alen, (*sa)->sa_len);
1611 free(*sa, M_SONAME);
1622 getpeername(td, uap)
1624 struct getpeername_args *uap;
1627 return (getpeername1(td, uap, 0));
1630 #ifdef COMPAT_OLDSOCK
1632 ogetpeername(td, uap)
1634 struct ogetpeername_args *uap;
1637 /* XXX uap should have type `getpeername_args *' to begin with. */
1638 return (getpeername1(td, (struct getpeername_args *)uap, 1));
1640 #endif /* COMPAT_OLDSOCK */
1643 sockargs(mp, buf, buflen, type)
1648 struct sockaddr *sa;
1652 if ((u_int)buflen > MLEN) {
1653 #ifdef COMPAT_OLDSOCK
1654 if (type == MT_SONAME && (u_int)buflen <= 112)
1655 buflen = MLEN; /* unix domain compat. hack */
1658 if ((u_int)buflen > MCLBYTES)
1661 m = m_get(M_TRYWAIT, type);
1664 if ((u_int)buflen > MLEN) {
1665 MCLGET(m, M_TRYWAIT);
1666 if ((m->m_flags & M_EXT) == 0) {
1672 error = copyin(buf, mtod(m, caddr_t), (u_int)buflen);
1677 if (type == MT_SONAME) {
1678 sa = mtod(m, struct sockaddr *);
1680 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
1681 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1682 sa->sa_family = sa->sa_len;
1684 sa->sa_len = buflen;
1691 getsockaddr(namp, uaddr, len)
1692 struct sockaddr **namp;
1696 struct sockaddr *sa;
1699 if (len > SOCK_MAXADDRLEN)
1700 return (ENAMETOOLONG);
1701 if (len < offsetof(struct sockaddr, sa_data[0]))
1703 MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1704 error = copyin(uaddr, sa, len);
1708 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
1709 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1710 sa->sa_family = sa->sa_len;
1719 * Detach mapped page and release resources back to the system.
1722 sf_buf_mext(void *addr, void *args)
1726 m = sf_buf_page(args);
1728 vm_page_lock_queues();
1729 vm_page_unwire(m, 0);
1731 * Check for the object going away on us. This can
1732 * happen since we don't hold a reference to it.
1733 * If so, we're responsible for freeing the page.
1735 if (m->wire_count == 0 && m->object == NULL)
1737 vm_page_unlock_queues();
1743 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1744 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1746 * Send a file specified by 'fd' and starting at 'offset' to a socket
1747 * specified by 's'. Send only 'nbytes' of the file or until EOF if nbytes ==
1748 * 0. Optionally add a header and/or trailer to the socket output. If
1749 * specified, write the total number of bytes sent into *sbytes.
1752 sendfile(struct thread *td, struct sendfile_args *uap)
1755 return (do_sendfile(td, uap, 0));
1759 do_sendfile(struct thread *td, struct sendfile_args *uap, int compat)
1761 struct sf_hdtr hdtr;
1762 struct uio *hdr_uio, *trl_uio;
1765 hdr_uio = trl_uio = NULL;
1767 if (uap->hdtr != NULL) {
1768 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1771 if (hdtr.headers != NULL) {
1772 error = copyinuio(hdtr.headers, hdtr.hdr_cnt, &hdr_uio);
1776 if (hdtr.trailers != NULL) {
1777 error = copyinuio(hdtr.trailers, hdtr.trl_cnt, &trl_uio);
1784 error = kern_sendfile(td, uap, hdr_uio, trl_uio, compat);
1787 free(hdr_uio, M_IOV);
1789 free(trl_uio, M_IOV);
1793 #ifdef COMPAT_FREEBSD4
1795 freebsd4_sendfile(struct thread *td, struct freebsd4_sendfile_args *uap)
1797 struct sendfile_args args;
1801 args.offset = uap->offset;
1802 args.nbytes = uap->nbytes;
1803 args.hdtr = uap->hdtr;
1804 args.sbytes = uap->sbytes;
1805 args.flags = uap->flags;
1807 return (do_sendfile(td, &args, 1));
1809 #endif /* COMPAT_FREEBSD4 */
1812 kern_sendfile(struct thread *td, struct sendfile_args *uap,
1813 struct uio *hdr_uio, struct uio *trl_uio, int compat)
1815 struct file *sock_fp;
1817 struct vm_object *obj = NULL;
1818 struct socket *so = NULL;
1819 struct mbuf *m = NULL;
1822 off_t off, xfsize, sbytes = 0, rem = 0;
1829 * The file descriptor must be a regular file and have a
1830 * backing VM object.
1831 * File offset must be positive. If it goes beyond EOF
1832 * we send only the header/trailer and no payload data.
1834 if ((error = fgetvp_read(td, uap->fd, &vp)) != 0)
1836 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1837 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
1841 * Temporarily increase the backing VM object's reference
1842 * count so that a forced reclamation of its vnode does not
1843 * immediately destroy it.
1845 VM_OBJECT_LOCK(obj);
1846 if ((obj->flags & OBJ_DEAD) == 0) {
1847 vm_object_reference_locked(obj);
1848 VM_OBJECT_UNLOCK(obj);
1850 VM_OBJECT_UNLOCK(obj);
1854 VOP_UNLOCK(vp, 0, td);
1855 VFS_UNLOCK_GIANT(vfslocked);
1860 if (uap->offset < 0) {
1866 * The socket must be a stream socket and connected.
1867 * Remember if it a blocking or non-blocking socket.
1869 if ((error = getsock(td->td_proc->p_fd, uap->s, &sock_fp,
1872 so = sock_fp->f_data;
1873 if (so->so_type != SOCK_STREAM) {
1877 if ((so->so_state & SS_ISCONNECTED) == 0) {
1882 * Do not wait on memory allocations but return ENOMEM for
1883 * caller to retry later.
1884 * XXX: Experimental.
1886 if (uap->flags & SF_MNOWAIT)
1891 error = mac_check_socket_send(td->td_ucred, so);
1897 /* If headers are specified copy them into mbufs. */
1898 if (hdr_uio != NULL) {
1899 hdr_uio->uio_td = td;
1900 hdr_uio->uio_rw = UIO_WRITE;
1901 if (hdr_uio->uio_resid > 0) {
1903 * In FBSD < 5.0 the nbytes to send also included
1904 * the header. If compat is specified subtract the
1905 * header size from nbytes.
1908 if (uap->nbytes > hdr_uio->uio_resid)
1909 uap->nbytes -= hdr_uio->uio_resid;
1913 m = m_uiotombuf(hdr_uio, (mnw ? M_NOWAIT : M_WAITOK),
1916 error = mnw ? EAGAIN : ENOBUFS;
1922 /* Protect against multiple writers to the socket. */
1923 (void) sblock(&so->so_snd, M_WAITOK);
1926 * Loop through the pages of the file, starting with the requested
1927 * offset. Get a file page (do I/O if necessary), map the file page
1928 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1930 * This is done in two loops. The inner loop turns as many pages
1931 * as it can, up to available socket buffer space, without blocking
1932 * into mbufs to have it bulk delivered into the socket send buffer.
1933 * The outer loop checks the state and available space of the socket
1934 * and takes care of the overall progress.
1936 for (off = uap->offset; ; ) {
1942 * Check the socket state for ongoing connection,
1943 * no errors and space in socket buffer.
1944 * If space is low allow for the remainder of the
1945 * file to be processed if it fits the socket buffer.
1946 * Otherwise block in waiting for sufficient space
1947 * to proceed, or if the socket is nonblocking, return
1948 * to userland with EAGAIN while reporting how far
1950 * We wait until the socket buffer has significant free
1951 * space to do bulk sends. This makes good use of file
1952 * system read ahead and allows packet segmentation
1953 * offloading hardware to take over lots of work. If
1954 * we were not careful here we would send off only one
1957 SOCKBUF_LOCK(&so->so_snd);
1958 if (so->so_snd.sb_lowat < so->so_snd.sb_hiwat / 2)
1959 so->so_snd.sb_lowat = so->so_snd.sb_hiwat / 2;
1961 if (so->so_snd.sb_state & SBS_CANTSENDMORE) {
1963 SOCKBUF_UNLOCK(&so->so_snd);
1965 } else if (so->so_error) {
1966 error = so->so_error;
1968 SOCKBUF_UNLOCK(&so->so_snd);
1971 space = sbspace(&so->so_snd);
1974 space < so->so_snd.sb_lowat)) {
1975 if (so->so_state & SS_NBIO) {
1976 SOCKBUF_UNLOCK(&so->so_snd);
1981 * sbwait drops the lock while sleeping.
1982 * When we loop back to retry_space the
1983 * state may have changed and we retest
1986 error = sbwait(&so->so_snd);
1988 * An error from sbwait usually indicates that we've
1989 * been interrupted by a signal. If we've sent anything
1990 * then return bytes sent, otherwise return the error.
1993 SOCKBUF_UNLOCK(&so->so_snd);
1998 SOCKBUF_UNLOCK(&so->so_snd);
2001 * Loop and construct maximum sized mbuf chain to be bulk
2002 * dumped into socket buffer.
2004 while(space > loopbytes) {
2009 VM_OBJECT_LOCK(obj);
2011 * Calculate the amount to transfer.
2012 * Not to exceed a page, the EOF,
2013 * or the passed in nbytes.
2015 pgoff = (vm_offset_t)(off & PAGE_MASK);
2016 xfsize = omin(PAGE_SIZE - pgoff,
2017 obj->un_pager.vnp.vnp_size - uap->offset -
2018 sbytes - loopbytes);
2020 rem = (uap->nbytes - sbytes - loopbytes);
2022 rem = obj->un_pager.vnp.vnp_size - uap->offset -
2024 xfsize = omin(rem, xfsize);
2026 VM_OBJECT_UNLOCK(obj);
2027 done = 1; /* all data sent */
2031 * Don't overflow the send buffer.
2032 * Stop here and send out what we've
2035 if (space < loopbytes + xfsize) {
2036 VM_OBJECT_UNLOCK(obj);
2041 * Attempt to look up the page.
2042 * Allocate if not found or
2043 * wait and loop if busy.
2045 pindex = OFF_TO_IDX(off);
2046 pg = vm_page_grab(obj, pindex, VM_ALLOC_NOBUSY |
2047 VM_ALLOC_NORMAL | VM_ALLOC_WIRED | VM_ALLOC_RETRY);
2050 * Check if page is valid for what we need,
2051 * otherwise initiate I/O.
2052 * If we already turned some pages into mbufs,
2053 * send them off before we come here again and
2056 if (pg->valid && vm_page_is_valid(pg, pgoff, xfsize))
2057 VM_OBJECT_UNLOCK(obj);
2059 error = EAGAIN; /* send what we already got */
2060 else if (uap->flags & SF_NODISKIO)
2066 * Ensure that our page is still around
2067 * when the I/O completes.
2069 vm_page_io_start(pg);
2070 VM_OBJECT_UNLOCK(obj);
2073 * Get the page from backing store.
2075 bsize = vp->v_mount->mnt_stat.f_iosize;
2076 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
2077 vn_lock(vp, LK_SHARED | LK_RETRY, td);
2080 * XXXMAC: Because we don't have fp->f_cred
2081 * here, we pass in NOCRED. This is probably
2082 * wrong, but is consistent with our original
2085 error = vn_rdwr(UIO_READ, vp, NULL, MAXBSIZE,
2086 trunc_page(off), UIO_NOCOPY, IO_NODELOCKED |
2087 IO_VMIO | ((MAXBSIZE / bsize) << IO_SEQSHIFT),
2088 td->td_ucred, NOCRED, &resid, td);
2089 VOP_UNLOCK(vp, 0, td);
2090 VFS_UNLOCK_GIANT(vfslocked);
2091 VM_OBJECT_LOCK(obj);
2092 vm_page_io_finish(pg);
2094 VM_OBJECT_UNLOCK(obj);
2098 vm_page_lock_queues();
2099 vm_page_unwire(pg, 0);
2101 * See if anyone else might know about
2102 * this page. If not and it is not valid,
2105 if (pg->wire_count == 0 && pg->valid == 0 &&
2106 pg->busy == 0 && !(pg->oflags & VPO_BUSY) &&
2107 pg->hold_count == 0) {
2110 vm_page_unlock_queues();
2111 VM_OBJECT_UNLOCK(obj);
2112 if (error == EAGAIN)
2113 error = 0; /* not a real error */
2118 * Get a sendfile buf. We usually wait as long
2119 * as necessary, but this wait can be interrupted.
2121 if ((sf = sf_buf_alloc(pg,
2122 (mnw ? SFB_NOWAIT : SFB_CATCH))) == NULL) {
2123 mbstat.sf_allocfail++;
2124 vm_page_lock_queues();
2125 vm_page_unwire(pg, 0);
2127 * XXX: Not same check as above!?
2129 if (pg->wire_count == 0 && pg->object == NULL)
2131 vm_page_unlock_queues();
2132 error = (mnw ? EAGAIN : EINTR);
2137 * Get an mbuf and set it up as having
2140 m0 = m_get((mnw ? M_NOWAIT : M_WAITOK), MT_DATA);
2142 error = (mnw ? EAGAIN : ENOBUFS);
2143 sf_buf_mext((void *)sf_buf_kva(sf), sf);
2146 MEXTADD(m0, sf_buf_kva(sf), PAGE_SIZE, sf_buf_mext,
2147 sf, M_RDONLY, EXT_SFBUF);
2148 m0->m_data = (char *)sf_buf_kva(sf) + pgoff;
2151 /* Append to mbuf chain. */
2157 /* Keep track of bits processed. */
2158 loopbytes += xfsize;
2162 /* Add the buffer chain to the socket buffer. */
2166 mlen = m_length(m, NULL);
2167 SOCKBUF_LOCK(&so->so_snd);
2168 if (so->so_snd.sb_state & SBS_CANTSENDMORE) {
2170 SOCKBUF_UNLOCK(&so->so_snd);
2173 SOCKBUF_UNLOCK(&so->so_snd);
2174 error = (*so->so_proto->pr_usrreqs->pru_send)
2175 (so, 0, m, NULL, NULL, td);
2178 m = NULL; /* pru_send always consumes */
2181 /* Quit outer loop on error or when we're done. */
2187 * Send trailers. Wimp out and use writev(2).
2189 if (trl_uio != NULL) {
2190 error = kern_writev(td, uap->s, trl_uio);
2193 sbytes += td->td_retval[0];
2197 sbunlock(&so->so_snd);
2200 * If there was no error we have to clear td->td_retval[0]
2201 * because it may have been set by writev.
2204 td->td_retval[0] = 0;
2206 if (uap->sbytes != NULL) {
2207 copyout(&sbytes, uap->sbytes, sizeof(off_t));
2210 vm_object_deallocate(obj);
2212 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
2214 VFS_UNLOCK_GIANT(vfslocked);
2223 if (error == ERESTART)
2231 * Functionality only compiled in if SCTP is defined in the kernel Makefile,
2232 * otherwise all return EOPNOTSUPP.
2233 * XXX: We should make this loadable one day.
2236 sctp_peeloff(td, uap)
2238 struct sctp_peeloff_args /* {
2244 struct filedesc *fdp;
2245 struct file *nfp = NULL;
2247 struct socket *head, *so;
2251 fdp = td->td_proc->p_fd;
2252 error = fgetsock(td, uap->sd, &head, &fflag);
2255 error = sctp_can_peel_off(head, (sctp_assoc_t)uap->name);
2259 * At this point we know we do have a assoc to pull
2260 * we proceed to get the fd setup. This may block
2264 error = falloc(td, &nfp, &fd);
2267 td->td_retval[0] = fd;
2269 so = sonewconn(head, SS_ISCONNECTED);
2273 * Before changing the flags on the socket, we have to bump the
2274 * reference count. Otherwise, if the protocol calls sofree(),
2275 * the socket will be released due to a zero refcount.
2278 soref(so); /* file descriptor reference */
2283 TAILQ_REMOVE(&head->so_comp, so, so_list);
2285 so->so_state |= (head->so_state & SS_NBIO);
2286 so->so_state &= ~SS_NOFDREF;
2287 so->so_qstate &= ~SQ_COMP;
2292 error = sctp_do_peeloff(head, so, (sctp_assoc_t)uap->name);
2295 if (head->so_sigio != NULL)
2296 fsetown(fgetown(&head->so_sigio), &so->so_sigio);
2300 nfp->f_flag = fflag;
2301 nfp->f_type = DTYPE_SOCKET;
2302 nfp->f_ops = &socketops;
2307 * close the new descriptor, assuming someone hasn't ripped it
2308 * out from under us.
2311 fdclose(fdp, nfp, fd, td);
2314 * Release explicitly held references before returning.
2323 return (EOPNOTSUPP);
2328 sctp_generic_sendmsg (td, uap)
2330 struct sctp_generic_sendmsg_args /* {
2336 struct sctp_sndrcvinfo *sinfo,
2341 struct sctp_sndrcvinfo sinfo, *u_sinfo = NULL;
2344 int use_rcvinfo = 1;
2346 struct sockaddr *to = NULL;
2348 struct uio *ktruio = NULL;
2351 struct iovec iov[1];
2354 error = copyin(uap->sinfo, &sinfo, sizeof (sinfo));
2360 error = getsockaddr(&to, uap->to, uap->tolen);
2367 error = getsock(td->td_proc->p_fd, uap->sd, &fp, NULL);
2371 iov[0].iov_base = uap->msg;
2372 iov[0].iov_len = uap->mlen;
2374 so = (struct socket *)fp->f_data;
2377 error = mac_check_socket_send(td->td_ucred, so);
2384 auio.uio_iovcnt = 1;
2385 auio.uio_segflg = UIO_USERSPACE;
2386 auio.uio_rw = UIO_WRITE;
2388 auio.uio_offset = 0; /* XXX */
2390 len = auio.uio_resid = uap->mlen;
2391 error = sctp_lower_sosend(so, to, &auio,
2392 (struct mbuf *)NULL, (struct mbuf *)NULL,
2393 uap->flags, use_rcvinfo, u_sinfo, td);
2395 if (auio.uio_resid != len && (error == ERESTART ||
2396 error == EINTR || error == EWOULDBLOCK))
2398 /* Generation of SIGPIPE can be controlled per socket. */
2399 if (error == EPIPE && !(so->so_options & SO_NOSIGPIPE) &&
2400 !(uap->flags & MSG_NOSIGNAL)) {
2401 PROC_LOCK(td->td_proc);
2402 psignal(td->td_proc, SIGPIPE);
2403 PROC_UNLOCK(td->td_proc);
2407 td->td_retval[0] = len - auio.uio_resid;
2409 if (ktruio != NULL) {
2410 ktruio->uio_resid = td->td_retval[0];
2411 ktrgenio(uap->sd, UIO_WRITE, ktruio, error);
2421 return (EOPNOTSUPP);
2426 sctp_generic_sendmsg_iov(td, uap)
2428 struct sctp_generic_sendmsg_iov_args /* {
2434 struct sctp_sndrcvinfo *sinfo,
2439 struct sctp_sndrcvinfo sinfo, *u_sinfo = NULL;
2442 int use_rcvinfo = 1;
2443 int error=0, len, i;
2444 struct sockaddr *to = NULL;
2446 struct uio *ktruio = NULL;
2449 struct iovec *iov, *tiov;
2452 error = copyin(uap->sinfo, &sinfo, sizeof (sinfo));
2458 error = getsockaddr(&to, uap->to, uap->tolen);
2465 error = getsock(td->td_proc->p_fd, uap->sd, &fp, NULL);
2469 error = copyiniov(uap->iov, uap->iovlen, &iov, EMSGSIZE);
2473 so = (struct socket *)fp->f_data;
2476 error = mac_check_socket_send(td->td_ucred, so);
2483 auio.uio_iovcnt = uap->iovlen;
2484 auio.uio_segflg = UIO_USERSPACE;
2485 auio.uio_rw = UIO_WRITE;
2487 auio.uio_offset = 0; /* XXX */
2490 for (i = 0; i <uap->iovlen; i++, tiov++) {
2491 if ((auio.uio_resid += tiov->iov_len) < 0) {
2496 len = auio.uio_resid;
2497 error = sctp_lower_sosend(so, to, &auio,
2498 (struct mbuf *)NULL, (struct mbuf *)NULL,
2499 uap->flags, use_rcvinfo, u_sinfo, td);
2501 if (auio.uio_resid != len && (error == ERESTART ||
2502 error == EINTR || error == EWOULDBLOCK))
2504 /* Generation of SIGPIPE can be controlled per socket */
2505 if (error == EPIPE && !(so->so_options & SO_NOSIGPIPE) &&
2506 !(uap->flags & MSG_NOSIGNAL)) {
2507 PROC_LOCK(td->td_proc);
2508 psignal(td->td_proc, SIGPIPE);
2509 PROC_UNLOCK(td->td_proc);
2513 td->td_retval[0] = len - auio.uio_resid;
2515 if (ktruio != NULL) {
2516 ktruio->uio_resid = td->td_retval[0];
2517 ktrgenio(uap->sd, UIO_WRITE, ktruio, error);
2529 return (EOPNOTSUPP);
2534 sctp_generic_recvmsg(td, uap)
2536 struct sctp_generic_recvmsg_args /* {
2540 struct sockaddr *from,
2541 __socklen_t *fromlenaddr,
2542 struct sctp_sndrcvinfo *sinfo,
2547 u_int8_t sockbufstore[256];
2549 struct iovec *iov, *tiov;
2550 struct sctp_sndrcvinfo sinfo;
2553 struct sockaddr *fromsa;
2555 int len, i, msg_flags;
2558 struct uio *ktruio = NULL;
2560 error = getsock(td->td_proc->p_fd, uap->sd, &fp, NULL);
2564 error = copyiniov(uap->iov, uap->iovlen, &iov, EMSGSIZE);
2572 error = mac_check_socket_receive(td->td_ucred, so);
2580 if (uap->fromlenaddr) {
2581 error = copyin(uap->fromlenaddr,
2582 &fromlen, sizeof (fromlen));
2589 if(uap->msg_flags) {
2590 error = copyin(uap->msg_flags, &msg_flags, sizeof (int));
2598 auio.uio_iovcnt = uap->iovlen;
2599 auio.uio_segflg = UIO_USERSPACE;
2600 auio.uio_rw = UIO_READ;
2602 auio.uio_offset = 0; /* XXX */
2605 for (i = 0; i <uap->iovlen; i++, tiov++) {
2606 if ((auio.uio_resid += tiov->iov_len) < 0) {
2611 len = auio.uio_resid;
2612 fromsa = (struct sockaddr *)sockbufstore;
2615 if (KTRPOINT(td, KTR_GENIO))
2616 ktruio = cloneuio(&auio);
2618 error = sctp_sorecvmsg(so, &auio, (struct mbuf **)NULL,
2619 fromsa, fromlen, &msg_flags,
2620 (struct sctp_sndrcvinfo *)&sinfo, 1);
2622 if (auio.uio_resid != (int)len && (error == ERESTART ||
2623 error == EINTR || error == EWOULDBLOCK))
2627 error = copyout(&sinfo, uap->sinfo, sizeof (sinfo));
2630 if (ktruio != NULL) {
2631 ktruio->uio_resid = (int)len - auio.uio_resid;
2632 ktrgenio(uap->sd, UIO_READ, ktruio, error);
2637 td->td_retval[0] = (int)len - auio.uio_resid;
2639 if (fromlen && uap->from) {
2641 if (len <= 0 || fromsa == 0)
2644 len = MIN(len, fromsa->sa_len);
2645 error = copyout(fromsa, uap->from, (unsigned)len);
2649 error = copyout(&len, uap->fromlenaddr, sizeof (socklen_t));
2654 if (uap->msg_flags) {
2655 error = copyout(&msg_flags, uap->msg_flags, sizeof (int));
2666 return (EOPNOTSUPP);