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 * 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.
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
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 $
38 * $DragonFly: src/sys/kern/uipc_syscalls.c,v 1.92 2008/11/26 13:10:56 sephe Exp $
41 #include "opt_ktrace.h"
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>
52 #include <sys/fcntl.h>
54 #include <sys/filio.h>
55 #include <sys/kern_syscall.h>
57 #include <sys/protosw.h>
58 #include <sys/sfbuf.h>
59 #include <sys/socket.h>
60 #include <sys/socketvar.h>
61 #include <sys/socketops.h>
63 #include <sys/vnode.h>
65 #include <sys/mount.h>
67 #include <sys/ktrace.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>
75 #include <sys/file2.h>
76 #include <sys/signalvar.h>
77 #include <sys/serialize.h>
79 #include <sys/thread2.h>
80 #include <sys/msgport2.h>
81 #include <sys/socketvar2.h>
82 #include <sys/mplock2.h>
83 #include <net/netmsg2.h>
86 #include <netinet/sctp_peeloff.h>
90 * System call interface to the socket abstraction.
93 extern struct fileops socketops;
96 * socket_args(int domain, int type, int protocol)
99 kern_socket(int domain, int type, int protocol, int *res)
101 struct thread *td = curthread;
102 struct filedesc *fdp = td->td_proc->p_fd;
107 KKASSERT(td->td_lwp);
109 error = falloc(td->td_lwp, &fp, &fd);
112 error = socreate(domain, &so, type, protocol, td);
114 fsetfd(fdp, NULL, fd);
116 fp->f_type = DTYPE_SOCKET;
117 fp->f_flag = FREAD | FWRITE;
118 fp->f_ops = &socketops;
131 sys_socket(struct socket_args *uap)
136 error = kern_socket(uap->domain, uap->type, uap->protocol,
137 &uap->sysmsg_iresult);
144 kern_bind(int s, struct sockaddr *sa)
146 struct thread *td = curthread;
147 struct proc *p = td->td_proc;
152 error = holdsock(p->p_fd, s, &fp);
155 error = sobind((struct socket *)fp->f_data, sa, td);
161 * bind_args(int s, caddr_t name, int namelen)
166 sys_bind(struct bind_args *uap)
171 error = getsockaddr(&sa, uap->name, uap->namelen);
175 error = kern_bind(uap->s, sa);
183 kern_listen(int s, int backlog)
185 struct thread *td = curthread;
186 struct proc *p = td->td_proc;
191 error = holdsock(p->p_fd, s, &fp);
194 error = solisten((struct socket *)fp->f_data, backlog, td);
200 * listen_args(int s, int backlog)
205 sys_listen(struct listen_args *uap)
210 error = kern_listen(uap->s, uap->backlog);
216 * Returns the accepted socket as well.
219 soaccept_predicate(struct netmsg_so_notify *msg)
221 struct socket *head = msg->base.nm_so;
223 if (head->so_error != 0) {
224 msg->base.lmsg.ms_error = head->so_error;
227 lwkt_gettoken(&head->so_rcv.ssb_token);
228 if (!TAILQ_EMPTY(&head->so_comp)) {
229 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
230 msg->base.nm_so = TAILQ_FIRST(&head->so_comp);
231 TAILQ_REMOVE(&head->so_comp, msg->base.nm_so, so_list);
234 msg->base.lmsg.ms_error = 0;
235 lwkt_reltoken(&head->so_rcv.ssb_token);
238 lwkt_reltoken(&head->so_rcv.ssb_token);
239 if (head->so_state & SS_CANTRCVMORE) {
240 msg->base.lmsg.ms_error = ECONNABORTED;
243 if (msg->nm_fflags & FNONBLOCK) {
244 msg->base.lmsg.ms_error = EWOULDBLOCK;
252 * The second argument to kern_accept() is a handle to a struct sockaddr.
253 * This allows kern_accept() to return a pointer to an allocated struct
254 * sockaddr which must be freed later with FREE(). The caller must
255 * initialize *name to NULL.
258 kern_accept(int s, int fflags, struct sockaddr **name, int *namelen, int *res)
260 struct thread *td = curthread;
261 struct filedesc *fdp = td->td_proc->p_fd;
262 struct file *lfp = NULL;
263 struct file *nfp = NULL;
265 struct socket *head, *so;
266 struct netmsg_so_notify msg;
268 u_int fflag; /* type must match fp->f_flag */
272 if (name && namelen && *namelen < 0)
275 error = holdsock(td->td_proc->p_fd, s, &lfp);
279 error = falloc(td->td_lwp, &nfp, &fd);
280 if (error) { /* Probably ran out of file descriptors. */
284 head = (struct socket *)lfp->f_data;
285 if ((head->so_options & SO_ACCEPTCONN) == 0) {
290 if (fflags & O_FBLOCKING)
291 fflags |= lfp->f_flag & ~FNONBLOCK;
292 else if (fflags & O_FNONBLOCKING)
293 fflags |= lfp->f_flag | FNONBLOCK;
295 fflags = lfp->f_flag;
297 /* optimize for uniprocessor case later XXX JH */
298 netmsg_init_abortable(&msg.base, head, &curthread->td_msgport,
299 0, netmsg_so_notify, netmsg_so_notify_doabort);
300 msg.nm_predicate = soaccept_predicate;
301 msg.nm_fflags = fflags;
302 msg.nm_etype = NM_REVENT;
303 error = lwkt_domsg(head->so_port, &msg.base.lmsg, PCATCH);
308 * At this point we have the connection that's ready to be accepted.
314 /* connection has been removed from the listen queue */
315 KNOTE(&head->so_rcv.ssb_kq.ki_note, 0);
317 soclrstate(so, SS_COMP);
319 if (head->so_sigio != NULL)
320 fsetown(fgetown(head->so_sigio), &so->so_sigio);
322 nfp->f_type = DTYPE_SOCKET;
324 nfp->f_ops = &socketops;
326 /* Sync socket nonblocking/async state with file flags */
327 tmp = fflag & FNONBLOCK;
328 fo_ioctl(nfp, FIONBIO, (caddr_t)&tmp, td->td_ucred, NULL);
329 tmp = fflag & FASYNC;
330 fo_ioctl(nfp, FIOASYNC, (caddr_t)&tmp, td->td_ucred, NULL);
333 error = soaccept(so, &sa);
336 * Set the returned name and namelen as applicable. Set the returned
337 * namelen to 0 for older code which might ignore the return value
341 if (sa && name && namelen) {
342 if (*namelen > sa->sa_len)
343 *namelen = sa->sa_len;
353 * If an error occured clear the reserved descriptor, else associate
356 * Note that *res is normally ignored if an error is returned but
357 * a syscall message will still have access to the result code.
360 fsetfd(fdp, NULL, fd);
363 fsetfd(fdp, nfp, fd);
371 * accept(int s, caddr_t name, int *anamelen)
376 sys_accept(struct accept_args *uap)
378 struct sockaddr *sa = NULL;
383 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
388 error = kern_accept(uap->s, 0, &sa, &sa_len,
389 &uap->sysmsg_iresult);
393 error = copyout(sa, uap->name, sa_len);
395 error = copyout(&sa_len, uap->anamelen,
396 sizeof(*uap->anamelen));
402 error = kern_accept(uap->s, 0, NULL, 0,
403 &uap->sysmsg_iresult);
410 * extaccept(int s, int fflags, caddr_t name, int *anamelen)
415 sys_extaccept(struct extaccept_args *uap)
417 struct sockaddr *sa = NULL;
420 int fflags = uap->flags & O_FMASK;
423 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
428 error = kern_accept(uap->s, fflags, &sa, &sa_len,
429 &uap->sysmsg_iresult);
433 error = copyout(sa, uap->name, sa_len);
435 error = copyout(&sa_len, uap->anamelen,
436 sizeof(*uap->anamelen));
442 error = kern_accept(uap->s, fflags, NULL, 0,
443 &uap->sysmsg_iresult);
451 * Returns TRUE if predicate satisfied.
454 soconnected_predicate(struct netmsg_so_notify *msg)
456 struct socket *so = msg->base.nm_so;
458 /* check predicate */
459 if (!(so->so_state & SS_ISCONNECTING) || so->so_error != 0) {
460 msg->base.lmsg.ms_error = so->so_error;
468 kern_connect(int s, int fflags, struct sockaddr *sa)
470 struct thread *td = curthread;
471 struct proc *p = td->td_proc;
474 int error, interrupted = 0;
476 error = holdsock(p->p_fd, s, &fp);
479 so = (struct socket *)fp->f_data;
481 if (fflags & O_FBLOCKING)
482 /* fflags &= ~FNONBLOCK; */;
483 else if (fflags & O_FNONBLOCKING)
488 if (so->so_state & SS_ISCONNECTING) {
492 error = soconnect(so, sa, td);
495 if ((fflags & FNONBLOCK) && (so->so_state & SS_ISCONNECTING)) {
499 if ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
500 struct netmsg_so_notify msg;
502 netmsg_init_abortable(&msg.base, so,
503 &curthread->td_msgport,
506 netmsg_so_notify_doabort);
507 msg.nm_predicate = soconnected_predicate;
508 msg.nm_etype = NM_REVENT;
509 error = lwkt_domsg(so->so_port, &msg.base.lmsg, PCATCH);
510 if (error == EINTR || error == ERESTART)
514 error = so->so_error;
519 soclrstate(so, SS_ISCONNECTING);
520 if (error == ERESTART)
528 * connect_args(int s, caddr_t name, int namelen)
533 sys_connect(struct connect_args *uap)
538 error = getsockaddr(&sa, uap->name, uap->namelen);
542 error = kern_connect(uap->s, 0, sa);
550 * connect_args(int s, int fflags, caddr_t name, int namelen)
555 sys_extconnect(struct extconnect_args *uap)
559 int fflags = uap->flags & O_FMASK;
561 error = getsockaddr(&sa, uap->name, uap->namelen);
565 error = kern_connect(uap->s, fflags, sa);
573 kern_socketpair(int domain, int type, int protocol, int *sv)
575 struct thread *td = curthread;
576 struct filedesc *fdp;
577 struct file *fp1, *fp2;
578 struct socket *so1, *so2;
581 fdp = td->td_proc->p_fd;
582 error = socreate(domain, &so1, type, protocol, td);
585 error = socreate(domain, &so2, type, protocol, td);
588 error = falloc(td->td_lwp, &fp1, &fd1);
593 error = falloc(td->td_lwp, &fp2, &fd2);
598 error = soconnect2(so1, so2);
601 if (type == SOCK_DGRAM) {
603 * Datagram socket connection is asymmetric.
605 error = soconnect2(so2, so1);
609 fp1->f_type = fp2->f_type = DTYPE_SOCKET;
610 fp1->f_flag = fp2->f_flag = FREAD|FWRITE;
611 fp1->f_ops = fp2->f_ops = &socketops;
612 fsetfd(fdp, fp1, fd1);
613 fsetfd(fdp, fp2, fd2);
618 fsetfd(fdp, NULL, fd2);
621 fsetfd(fdp, NULL, fd1);
624 (void)soclose(so2, 0);
626 (void)soclose(so1, 0);
631 * socketpair(int domain, int type, int protocol, int *rsv)
636 sys_socketpair(struct socketpair_args *uap)
641 error = kern_socketpair(uap->domain, uap->type, uap->protocol, sockv);
645 error = copyout(sockv, uap->rsv, sizeof(sockv));
650 kern_sendmsg(int s, struct sockaddr *sa, struct uio *auio,
651 struct mbuf *control, int flags, size_t *res)
653 struct thread *td = curthread;
654 struct lwp *lp = td->td_lwp;
655 struct proc *p = td->td_proc;
661 struct iovec *ktriov = NULL;
665 error = holdsock(p->p_fd, s, &fp);
669 if (KTRPOINT(td, KTR_GENIO)) {
670 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
672 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
673 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
677 len = auio->uio_resid;
678 so = (struct socket *)fp->f_data;
679 if ((flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
680 if (fp->f_flag & FNONBLOCK)
681 flags |= MSG_FNONBLOCKING;
683 error = so_pru_sosend(so, sa, auio, NULL, control, flags, td);
685 if (auio->uio_resid != len && (error == ERESTART ||
686 error == EINTR || error == EWOULDBLOCK))
689 lwpsignal(p, lp, SIGPIPE);
692 if (ktriov != NULL) {
694 ktruio.uio_iov = ktriov;
695 ktruio.uio_resid = len - auio->uio_resid;
696 ktrgenio(lp, s, UIO_WRITE, &ktruio, error);
698 FREE(ktriov, M_TEMP);
702 *res = len - auio->uio_resid;
708 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
713 sys_sendto(struct sendto_args *uap)
715 struct thread *td = curthread;
718 struct sockaddr *sa = NULL;
722 error = getsockaddr(&sa, uap->to, uap->tolen);
726 aiov.iov_base = uap->buf;
727 aiov.iov_len = uap->len;
728 auio.uio_iov = &aiov;
731 auio.uio_resid = uap->len;
732 auio.uio_segflg = UIO_USERSPACE;
733 auio.uio_rw = UIO_WRITE;
737 error = kern_sendmsg(uap->s, sa, &auio, NULL, uap->flags,
738 &uap->sysmsg_szresult);
747 * sendmsg_args(int s, caddr_t msg, int flags)
752 sys_sendmsg(struct sendmsg_args *uap)
754 struct thread *td = curthread;
757 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
758 struct sockaddr *sa = NULL;
759 struct mbuf *control = NULL;
762 error = copyin(uap->msg, (caddr_t)&msg, sizeof(msg));
767 * Conditionally copyin msg.msg_name.
770 error = getsockaddr(&sa, msg.msg_name, msg.msg_namelen);
778 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
783 auio.uio_iovcnt = msg.msg_iovlen;
785 auio.uio_segflg = UIO_USERSPACE;
786 auio.uio_rw = UIO_WRITE;
790 * Conditionally copyin msg.msg_control.
792 if (msg.msg_control) {
793 if (msg.msg_controllen < sizeof(struct cmsghdr) ||
794 msg.msg_controllen > MLEN) {
798 control = m_get(MB_WAIT, MT_CONTROL);
799 if (control == NULL) {
803 control->m_len = msg.msg_controllen;
804 error = copyin(msg.msg_control, mtod(control, caddr_t),
813 error = kern_sendmsg(uap->s, sa, &auio, control, uap->flags,
814 &uap->sysmsg_szresult);
818 iovec_free(&iov, aiov);
826 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
827 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
828 * Don't forget to FREE() and m_free() these if they are returned.
831 kern_recvmsg(int s, struct sockaddr **sa, struct uio *auio,
832 struct mbuf **control, int *flags, size_t *res)
834 struct thread *td = curthread;
835 struct proc *p = td->td_proc;
842 struct iovec *ktriov = NULL;
846 error = holdsock(p->p_fd, s, &fp);
850 if (KTRPOINT(td, KTR_GENIO)) {
851 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
853 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
854 bcopy(auio->uio_iov, ktriov, iovlen);
858 len = auio->uio_resid;
859 so = (struct socket *)fp->f_data;
861 if (flags == NULL || (*flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
862 if (fp->f_flag & FNONBLOCK) {
864 *flags |= MSG_FNONBLOCKING;
866 lflags = MSG_FNONBLOCKING;
872 error = so_pru_soreceive(so, sa, auio, NULL, control, flags);
874 if (auio->uio_resid != len && (error == ERESTART ||
875 error == EINTR || error == EWOULDBLOCK))
879 if (ktriov != NULL) {
881 ktruio.uio_iov = ktriov;
882 ktruio.uio_resid = len - auio->uio_resid;
883 ktrgenio(td->td_lwp, s, UIO_READ, &ktruio, error);
885 FREE(ktriov, M_TEMP);
889 *res = len - auio->uio_resid;
895 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
896 * caddr_t from, int *fromlenaddr)
901 sys_recvfrom(struct recvfrom_args *uap)
903 struct thread *td = curthread;
906 struct sockaddr *sa = NULL;
909 if (uap->from && uap->fromlenaddr) {
910 error = copyin(uap->fromlenaddr, &fromlen, sizeof(fromlen));
918 aiov.iov_base = uap->buf;
919 aiov.iov_len = uap->len;
920 auio.uio_iov = &aiov;
923 auio.uio_resid = uap->len;
924 auio.uio_segflg = UIO_USERSPACE;
925 auio.uio_rw = UIO_READ;
929 error = kern_recvmsg(uap->s, uap->from ? &sa : NULL, &auio, NULL,
930 &uap->flags, &uap->sysmsg_szresult);
933 if (error == 0 && uap->from) {
934 /* note: sa may still be NULL */
936 fromlen = MIN(fromlen, sa->sa_len);
937 error = copyout(sa, uap->from, fromlen);
942 error = copyout(&fromlen, uap->fromlenaddr,
953 * recvmsg_args(int s, struct msghdr *msg, int flags)
958 sys_recvmsg(struct recvmsg_args *uap)
960 struct thread *td = curthread;
963 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
964 struct mbuf *m, *control = NULL;
965 struct sockaddr *sa = NULL;
967 socklen_t *ufromlenp, *ucontrollenp;
968 int error, fromlen, controllen, len, flags, *uflagsp;
971 * This copyin handles everything except the iovec.
973 error = copyin(uap->msg, &msg, sizeof(msg));
977 if (msg.msg_name && msg.msg_namelen < 0)
979 if (msg.msg_control && msg.msg_controllen < 0)
982 ufromlenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
984 ucontrollenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
986 uflagsp = (int *)((caddr_t)uap->msg + offsetof(struct msghdr,
992 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
997 auio.uio_iovcnt = msg.msg_iovlen;
999 auio.uio_segflg = UIO_USERSPACE;
1000 auio.uio_rw = UIO_READ;
1006 error = kern_recvmsg(uap->s,
1007 (msg.msg_name ? &sa : NULL), &auio,
1008 (msg.msg_control ? &control : NULL), &flags,
1009 &uap->sysmsg_szresult);
1013 * Conditionally copyout the name and populate the namelen field.
1015 if (error == 0 && msg.msg_name) {
1016 /* note: sa may still be NULL */
1018 fromlen = MIN(msg.msg_namelen, sa->sa_len);
1019 error = copyout(sa, msg.msg_name, fromlen);
1024 error = copyout(&fromlen, ufromlenp,
1025 sizeof(*ufromlenp));
1029 * Copyout msg.msg_control and msg.msg_controllen.
1031 if (error == 0 && msg.msg_control) {
1032 len = msg.msg_controllen;
1034 ctlbuf = (caddr_t)msg.msg_control;
1036 while(m && len > 0) {
1037 unsigned int tocopy;
1039 if (len >= m->m_len) {
1042 msg.msg_flags |= MSG_CTRUNC;
1046 error = copyout(mtod(m, caddr_t), ctlbuf, tocopy);
1054 controllen = ctlbuf - (caddr_t)msg.msg_control;
1055 error = copyout(&controllen, ucontrollenp,
1056 sizeof(*ucontrollenp));
1060 error = copyout(&flags, uflagsp, sizeof(*uflagsp));
1065 iovec_free(&iov, aiov);
1072 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1073 * in kernel pointer instead of a userland pointer. This allows us
1074 * to manipulate socket options in the emulation code.
1077 kern_setsockopt(int s, struct sockopt *sopt)
1079 struct thread *td = curthread;
1080 struct proc *p = td->td_proc;
1084 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1086 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1088 if (sopt->sopt_valsize < 0)
1091 error = holdsock(p->p_fd, s, &fp);
1095 error = sosetopt((struct socket *)fp->f_data, sopt);
1101 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
1106 sys_setsockopt(struct setsockopt_args *uap)
1108 struct thread *td = curthread;
1109 struct sockopt sopt;
1112 sopt.sopt_level = uap->level;
1113 sopt.sopt_name = uap->name;
1114 sopt.sopt_valsize = uap->valsize;
1116 sopt.sopt_val = NULL;
1118 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1121 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1122 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1128 error = kern_setsockopt(uap->s, &sopt);
1132 kfree(sopt.sopt_val, M_TEMP);
1137 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1138 * in kernel pointer instead of a userland pointer. This allows us
1139 * to manipulate socket options in the emulation code.
1142 kern_getsockopt(int s, struct sockopt *sopt)
1144 struct thread *td = curthread;
1145 struct proc *p = td->td_proc;
1149 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1151 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1153 if (sopt->sopt_valsize < 0 || sopt->sopt_valsize > SOMAXOPT_SIZE)
1156 error = holdsock(p->p_fd, s, &fp);
1160 error = sogetopt((struct socket *)fp->f_data, sopt);
1166 * getsockopt_args(int s, int level, int name, caddr_t val, int *avalsize)
1171 sys_getsockopt(struct getsockopt_args *uap)
1173 struct thread *td = curthread;
1174 struct sockopt sopt;
1178 error = copyin(uap->avalsize, &valsize, sizeof(valsize));
1185 sopt.sopt_level = uap->level;
1186 sopt.sopt_name = uap->name;
1187 sopt.sopt_valsize = valsize;
1189 sopt.sopt_val = NULL;
1191 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1194 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1195 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1201 error = kern_getsockopt(uap->s, &sopt);
1205 valsize = sopt.sopt_valsize;
1206 error = copyout(&valsize, uap->avalsize, sizeof(valsize));
1210 error = copyout(sopt.sopt_val, uap->val, sopt.sopt_valsize);
1213 kfree(sopt.sopt_val, M_TEMP);
1218 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1219 * This allows kern_getsockname() to return a pointer to an allocated struct
1220 * sockaddr which must be freed later with FREE(). The caller must
1221 * initialize *name to NULL.
1224 kern_getsockname(int s, struct sockaddr **name, int *namelen)
1226 struct thread *td = curthread;
1227 struct proc *p = td->td_proc;
1230 struct sockaddr *sa = NULL;
1233 error = holdsock(p->p_fd, s, &fp);
1240 so = (struct socket *)fp->f_data;
1241 error = so_pru_sockaddr(so, &sa);
1246 *namelen = MIN(*namelen, sa->sa_len);
1256 * getsockname_args(int fdes, caddr_t asa, int *alen)
1263 sys_getsockname(struct getsockname_args *uap)
1265 struct sockaddr *sa = NULL;
1268 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1273 error = kern_getsockname(uap->fdes, &sa, &sa_len);
1277 error = copyout(sa, uap->asa, sa_len);
1279 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1286 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1287 * This allows kern_getpeername() to return a pointer to an allocated struct
1288 * sockaddr which must be freed later with FREE(). The caller must
1289 * initialize *name to NULL.
1292 kern_getpeername(int s, struct sockaddr **name, int *namelen)
1294 struct thread *td = curthread;
1295 struct proc *p = td->td_proc;
1298 struct sockaddr *sa = NULL;
1301 error = holdsock(p->p_fd, s, &fp);
1308 so = (struct socket *)fp->f_data;
1309 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1313 error = so_pru_peeraddr(so, &sa);
1318 *namelen = MIN(*namelen, sa->sa_len);
1328 * getpeername_args(int fdes, caddr_t asa, int *alen)
1330 * Get name of peer for connected socket.
1335 sys_getpeername(struct getpeername_args *uap)
1337 struct sockaddr *sa = NULL;
1340 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1345 error = kern_getpeername(uap->fdes, &sa, &sa_len);
1349 error = copyout(sa, uap->asa, sa_len);
1351 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1358 getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len)
1360 struct sockaddr *sa;
1364 if (len > SOCK_MAXADDRLEN)
1365 return ENAMETOOLONG;
1366 if (len < offsetof(struct sockaddr, sa_data[0]))
1368 MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1369 error = copyin(uaddr, sa, len);
1373 #if BYTE_ORDER != BIG_ENDIAN
1375 * The bind(), connect(), and sendto() syscalls were not
1376 * versioned for COMPAT_43. Thus, this check must stay.
1378 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1379 sa->sa_family = sa->sa_len;
1388 * Detach a mapped page and release resources back to the system.
1389 * We must release our wiring and if the object is ripped out
1390 * from under the vm_page we become responsible for freeing the
1391 * page. These routines must be MPSAFE.
1393 * XXX HACK XXX TEMPORARY UNTIL WE IMPLEMENT EXT MBUF REFERENCE COUNTING
1395 * XXX vm_page_*() routines are not MPSAFE yet, the MP lock is required.
1398 sf_buf_mfree(void *arg)
1400 struct sf_buf *sf = arg;
1404 * XXX vm_page_*() and SFBUF routines not MPSAFE yet.
1408 m = sf_buf_page(sf);
1409 if (sf_buf_free(sf) == 0) {
1410 vm_page_unwire(m, 0);
1411 if (m->wire_count == 0 && m->object == NULL)
1412 vm_page_try_to_free(m);
1420 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1421 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1423 * Send a file specified by 'fd' and starting at 'offset' to a socket
1424 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1425 * nbytes == 0. Optionally add a header and/or trailer to the socket
1426 * output. If specified, write the total number of bytes sent into *sbytes.
1428 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1429 * the headers to count against the remaining bytes to be sent from
1430 * the file descriptor. We may wish to implement a compatibility syscall
1436 sys_sendfile(struct sendfile_args *uap)
1438 struct thread *td = curthread;
1439 struct proc *p = td->td_proc;
1441 struct vnode *vp = NULL;
1442 struct sf_hdtr hdtr;
1443 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
1445 struct mbuf *mheader = NULL;
1448 off_t hdtr_size = 0;
1455 * Do argument checking. Must be a regular file in, stream
1456 * type and connected socket out, positive offset.
1458 fp = holdfp(p->p_fd, uap->fd, FREAD);
1462 if (fp->f_type != DTYPE_VNODE) {
1467 vp = (struct vnode *)fp->f_data;
1472 * If specified, get the pointer to the sf_hdtr struct for
1473 * any headers/trailers.
1476 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1483 error = iovec_copyin(hdtr.headers, &iov, aiov,
1484 hdtr.hdr_cnt, &hbytes);
1488 auio.uio_iovcnt = hdtr.hdr_cnt;
1489 auio.uio_offset = 0;
1490 auio.uio_segflg = UIO_USERSPACE;
1491 auio.uio_rw = UIO_WRITE;
1493 auio.uio_resid = hbytes;
1495 mheader = m_uiomove(&auio);
1497 iovec_free(&iov, aiov);
1498 if (mheader == NULL)
1503 error = kern_sendfile(vp, uap->s, uap->offset, uap->nbytes, mheader,
1504 &sbytes, uap->flags);
1509 * Send trailers. Wimp out and use writev(2).
1511 if (uap->hdtr != NULL && hdtr.trailers != NULL) {
1512 error = iovec_copyin(hdtr.trailers, &iov, aiov,
1513 hdtr.trl_cnt, &auio.uio_resid);
1517 auio.uio_iovcnt = hdtr.trl_cnt;
1518 auio.uio_offset = 0;
1519 auio.uio_segflg = UIO_USERSPACE;
1520 auio.uio_rw = UIO_WRITE;
1523 error = kern_sendmsg(uap->s, NULL, &auio, NULL, 0, &tbytes);
1525 iovec_free(&iov, aiov);
1528 hdtr_size += tbytes; /* trailer bytes successfully sent */
1535 if (uap->sbytes != NULL) {
1536 sbytes += hdtr_size;
1537 copyout(&sbytes, uap->sbytes, sizeof(off_t));
1543 kern_sendfile(struct vnode *vp, int sfd, off_t offset, size_t nbytes,
1544 struct mbuf *mheader, off_t *sbytes, int flags)
1546 struct thread *td = curthread;
1547 struct proc *p = td->td_proc;
1548 struct vm_object *obj;
1558 if (vp->v_type != VREG) {
1562 if ((obj = vp->v_object) == NULL) {
1566 error = holdsock(p->p_fd, sfd, &fp);
1569 so = (struct socket *)fp->f_data;
1570 if (so->so_type != SOCK_STREAM) {
1574 if ((so->so_state & SS_ISCONNECTED) == 0) {
1585 * Protect against multiple writers to the socket.
1587 ssb_lock(&so->so_snd, M_WAITOK);
1590 * Loop through the pages in the file, starting with the requested
1591 * offset. Get a file page (do I/O if necessary), map the file page
1592 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1595 for (off = offset; ; off += xfsize, *sbytes += xfsize + hbytes) {
1599 pindex = OFF_TO_IDX(off);
1602 * Calculate the amount to transfer. Not to exceed a page,
1603 * the EOF, or the passed in nbytes.
1605 xfsize = vp->v_filesize - off;
1606 if (xfsize > PAGE_SIZE)
1608 pgoff = (vm_offset_t)(off & PAGE_MASK);
1609 if (PAGE_SIZE - pgoff < xfsize)
1610 xfsize = PAGE_SIZE - pgoff;
1611 if (nbytes && xfsize > (nbytes - *sbytes))
1612 xfsize = nbytes - *sbytes;
1616 * Optimize the non-blocking case by looking at the socket space
1617 * before going to the extra work of constituting the sf_buf.
1619 if ((fp->f_flag & FNONBLOCK) && ssb_space(&so->so_snd) <= 0) {
1620 if (so->so_state & SS_CANTSENDMORE)
1624 ssb_unlock(&so->so_snd);
1628 * Attempt to look up the page.
1630 * Allocate if not found, wait and loop if busy, then
1631 * wire the page. critical section protection is
1632 * required to maintain the object association (an
1633 * interrupt can free the page) through to the
1634 * vm_page_wire() call.
1637 lwkt_gettoken(&vm_token);
1638 pg = vm_page_lookup(obj, pindex);
1640 pg = vm_page_alloc(obj, pindex, VM_ALLOC_NORMAL);
1643 lwkt_reltoken(&vm_token);
1648 } else if (vm_page_sleep_busy(pg, TRUE, "sfpbsy")) {
1649 lwkt_reltoken(&vm_token);
1654 lwkt_reltoken(&vm_token);
1658 * If page is not valid for what we need, initiate I/O
1661 if (!pg->valid || !vm_page_is_valid(pg, pgoff, xfsize)) {
1667 * Ensure that our page is still around when the I/O
1670 vm_page_io_start(pg);
1673 * Get the page from backing store.
1675 bsize = vp->v_mount->mnt_stat.f_iosize;
1676 auio.uio_iov = &aiov;
1677 auio.uio_iovcnt = 1;
1679 aiov.iov_len = MAXBSIZE;
1680 auio.uio_resid = MAXBSIZE;
1681 auio.uio_offset = trunc_page(off);
1682 auio.uio_segflg = UIO_NOCOPY;
1683 auio.uio_rw = UIO_READ;
1685 vn_lock(vp, LK_SHARED | LK_RETRY);
1686 error = VOP_READ(vp, &auio,
1687 IO_VMIO | ((MAXBSIZE / bsize) << 16),
1690 vm_page_flag_clear(pg, PG_ZERO);
1691 vm_page_io_finish(pg);
1694 vm_page_unwire(pg, 0);
1695 vm_page_try_to_free(pg);
1697 ssb_unlock(&so->so_snd);
1704 * Get a sendfile buf. We usually wait as long as necessary,
1705 * but this wait can be interrupted.
1707 if ((sf = sf_buf_alloc(pg)) == NULL) {
1709 vm_page_unwire(pg, 0);
1710 vm_page_try_to_free(pg);
1712 ssb_unlock(&so->so_snd);
1718 * Get an mbuf header and set it up as having external storage.
1720 MGETHDR(m, MB_WAIT, MT_DATA);
1724 ssb_unlock(&so->so_snd);
1728 m->m_ext.ext_free = sf_buf_mfree;
1729 m->m_ext.ext_ref = sf_buf_ref;
1730 m->m_ext.ext_arg = sf;
1731 m->m_ext.ext_buf = (void *)sf_buf_kva(sf);
1732 m->m_ext.ext_size = PAGE_SIZE;
1733 m->m_data = (char *)sf_buf_kva(sf) + pgoff;
1734 m->m_flags |= M_EXT;
1735 m->m_pkthdr.len = m->m_len = xfsize;
1736 KKASSERT((m->m_flags & (M_EXT_CLUSTER)) == 0);
1738 if (mheader != NULL) {
1739 hbytes = mheader->m_pkthdr.len;
1740 mheader->m_pkthdr.len += m->m_pkthdr.len;
1748 * Add the buffer to the socket buffer chain.
1753 * Make sure that the socket is still able to take more data.
1754 * CANTSENDMORE being true usually means that the connection
1755 * was closed. so_error is true when an error was sensed after
1757 * The state is checked after the page mapping and buffer
1758 * allocation above since those operations may block and make
1759 * any socket checks stale. From this point forward, nothing
1760 * blocks before the pru_send (or more accurately, any blocking
1761 * results in a loop back to here to re-check).
1763 if ((so->so_state & SS_CANTSENDMORE) || so->so_error) {
1764 if (so->so_state & SS_CANTSENDMORE) {
1767 error = so->so_error;
1771 ssb_unlock(&so->so_snd);
1776 * Wait for socket space to become available. We do this just
1777 * after checking the connection state above in order to avoid
1778 * a race condition with ssb_wait().
1780 if (ssb_space(&so->so_snd) < so->so_snd.ssb_lowat) {
1781 if (fp->f_flag & FNONBLOCK) {
1783 ssb_unlock(&so->so_snd);
1788 error = ssb_wait(&so->so_snd);
1790 * An error from ssb_wait usually indicates that we've
1791 * been interrupted by a signal. If we've sent anything
1792 * then return bytes sent, otherwise return the error.
1796 ssb_unlock(&so->so_snd);
1802 error = so_pru_send(so, 0, m, NULL, NULL, td);
1805 ssb_unlock(&so->so_snd);
1809 if (mheader != NULL) {
1810 *sbytes += mheader->m_pkthdr.len;
1811 error = so_pru_send(so, 0, mheader, NULL, NULL, td);
1814 ssb_unlock(&so->so_snd);
1819 if (mheader != NULL)
1828 sys_sctp_peeloff(struct sctp_peeloff_args *uap)
1831 struct thread *td = curthread;
1832 struct filedesc *fdp = td->td_proc->p_fd;
1833 struct file *lfp = NULL;
1834 struct file *nfp = NULL;
1836 struct socket *head, *so;
1839 short fflag; /* type must match fp->f_flag */
1841 assoc_id = uap->name;
1842 error = holdsock(td->td_proc->p_fd, uap->sd, &lfp);
1848 head = (struct socket *)lfp->f_data;
1849 error = sctp_can_peel_off(head, assoc_id);
1855 * At this point we know we do have a assoc to pull
1856 * we proceed to get the fd setup. This may block
1860 fflag = lfp->f_flag;
1861 error = falloc(td->td_lwp, &nfp, &fd);
1864 * Probably ran out of file descriptors. Put the
1865 * unaccepted connection back onto the queue and
1866 * do another wakeup so some other process might
1867 * have a chance at it.
1872 uap->sysmsg_iresult = fd;
1874 so = sctp_get_peeloff(head, assoc_id, &error);
1877 * Either someone else peeled it off OR
1878 * we can't get a socket.
1882 soreference(so); /* reference needed */
1883 soclrstate(so, SS_NOFDREF | SS_COMP); /* when clearing NOFDREF */
1885 if (head->so_sigio != NULL)
1886 fsetown(fgetown(head->so_sigio), &so->so_sigio);
1888 nfp->f_type = DTYPE_SOCKET;
1889 nfp->f_flag = fflag;
1890 nfp->f_ops = &socketops;
1895 * Assign the file pointer to the reserved descriptor, or clear
1896 * the reserved descriptor if an error occured.
1899 fsetfd(fdp, NULL, fd);
1901 fsetfd(fdp, nfp, fd);
1904 * Release explicitly held references before returning.