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 $
40 #include "opt_ktrace.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
46 #include <sys/sysproto.h>
47 #include <sys/malloc.h>
48 #include <sys/filedesc.h>
49 #include <sys/event.h>
51 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/kern_syscall.h>
56 #include <sys/protosw.h>
57 #include <sys/sfbuf.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
60 #include <sys/socketops.h>
62 #include <sys/vnode.h>
64 #include <sys/mount.h>
66 #include <sys/ktrace.h>
69 #include <vm/vm_object.h>
70 #include <vm/vm_page.h>
71 #include <vm/vm_pageout.h>
72 #include <vm/vm_kern.h>
73 #include <vm/vm_extern.h>
74 #include <sys/file2.h>
75 #include <sys/signalvar.h>
76 #include <sys/serialize.h>
78 #include <sys/thread2.h>
79 #include <sys/msgport2.h>
80 #include <sys/socketvar2.h>
81 #include <net/netmsg2.h>
84 #include <netinet/sctp_peeloff.h>
88 * System call interface to the socket abstraction.
91 extern struct fileops socketops;
94 * socket_args(int domain, int type, int protocol)
97 kern_socket(int domain, int type, int protocol, int *res)
99 struct thread *td = curthread;
100 struct filedesc *fdp = td->td_proc->p_fd;
105 KKASSERT(td->td_lwp);
107 error = falloc(td->td_lwp, &fp, &fd);
110 error = socreate(domain, &so, type, protocol, td);
112 fsetfd(fdp, NULL, fd);
114 fp->f_type = DTYPE_SOCKET;
115 fp->f_flag = FREAD | FWRITE;
116 fp->f_ops = &socketops;
129 sys_socket(struct socket_args *uap)
133 error = kern_socket(uap->domain, uap->type, uap->protocol,
134 &uap->sysmsg_iresult);
140 kern_bind(int s, struct sockaddr *sa)
142 struct thread *td = curthread;
143 struct proc *p = td->td_proc;
148 error = holdsock(p->p_fd, s, &fp);
151 error = sobind((struct socket *)fp->f_data, sa, td);
157 * bind_args(int s, caddr_t name, int namelen)
162 sys_bind(struct bind_args *uap)
167 error = getsockaddr(&sa, uap->name, uap->namelen);
170 error = kern_bind(uap->s, sa);
177 kern_listen(int s, int backlog)
179 struct thread *td = curthread;
180 struct proc *p = td->td_proc;
185 error = holdsock(p->p_fd, s, &fp);
188 error = solisten((struct socket *)fp->f_data, backlog, td);
194 * listen_args(int s, int backlog)
199 sys_listen(struct listen_args *uap)
203 error = kern_listen(uap->s, uap->backlog);
208 * Returns the accepted socket as well.
210 * NOTE! The sockets sitting on so_comp/so_incomp might have 0 refs, the
211 * pool token is absolutely required to avoid a sofree() race,
212 * as well as to avoid tailq handling races.
215 soaccept_predicate(struct netmsg_so_notify *msg)
217 struct socket *head = msg->base.nm_so;
220 if (head->so_error != 0) {
221 msg->base.lmsg.ms_error = head->so_error;
224 lwkt_getpooltoken(head);
225 if (!TAILQ_EMPTY(&head->so_comp)) {
226 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
227 so = TAILQ_FIRST(&head->so_comp);
228 TAILQ_REMOVE(&head->so_comp, so, so_list);
230 soclrstate(so, SS_COMP);
234 lwkt_relpooltoken(head);
236 msg->base.lmsg.ms_error = 0;
237 msg->base.nm_so = so;
240 lwkt_relpooltoken(head);
241 if (head->so_state & SS_CANTRCVMORE) {
242 msg->base.lmsg.ms_error = ECONNABORTED;
245 if (msg->nm_fflags & FNONBLOCK) {
246 msg->base.lmsg.ms_error = EWOULDBLOCK;
254 * The second argument to kern_accept() is a handle to a struct sockaddr.
255 * This allows kern_accept() to return a pointer to an allocated struct
256 * sockaddr which must be freed later with FREE(). The caller must
257 * initialize *name to NULL.
260 kern_accept(int s, int fflags, struct sockaddr **name, int *namelen, int *res)
262 struct thread *td = curthread;
263 struct filedesc *fdp = td->td_proc->p_fd;
264 struct file *lfp = NULL;
265 struct file *nfp = NULL;
267 struct socket *head, *so;
268 struct netmsg_so_notify msg;
270 u_int fflag; /* type must match fp->f_flag */
274 if (name && namelen && *namelen < 0)
277 error = holdsock(td->td_proc->p_fd, s, &lfp);
281 error = falloc(td->td_lwp, &nfp, &fd);
282 if (error) { /* Probably ran out of file descriptors. */
286 head = (struct socket *)lfp->f_data;
287 if ((head->so_options & SO_ACCEPTCONN) == 0) {
292 if (fflags & O_FBLOCKING)
293 fflags |= lfp->f_flag & ~FNONBLOCK;
294 else if (fflags & O_FNONBLOCKING)
295 fflags |= lfp->f_flag | FNONBLOCK;
297 fflags = lfp->f_flag;
299 /* optimize for uniprocessor case later XXX JH */
300 netmsg_init_abortable(&msg.base, head, &curthread->td_msgport,
301 0, netmsg_so_notify, netmsg_so_notify_doabort);
302 msg.nm_predicate = soaccept_predicate;
303 msg.nm_fflags = fflags;
304 msg.nm_etype = NM_REVENT;
305 error = lwkt_domsg(head->so_port, &msg.base.lmsg, PCATCH);
310 * At this point we have the connection that's ready to be accepted.
312 * NOTE! soaccept_predicate() ref'd so for us, and soaccept() expects
313 * to eat the ref and turn it into a descriptor.
319 /* connection has been removed from the listen queue */
320 KNOTE(&head->so_rcv.ssb_kq.ki_note, 0);
322 if (head->so_sigio != NULL)
323 fsetown(fgetown(&head->so_sigio), &so->so_sigio);
325 nfp->f_type = DTYPE_SOCKET;
327 nfp->f_ops = &socketops;
329 /* Sync socket nonblocking/async state with file flags */
330 tmp = fflag & FNONBLOCK;
331 fo_ioctl(nfp, FIONBIO, (caddr_t)&tmp, td->td_ucred, NULL);
332 tmp = fflag & FASYNC;
333 fo_ioctl(nfp, FIOASYNC, (caddr_t)&tmp, td->td_ucred, NULL);
336 error = soaccept(so, &sa);
339 * Set the returned name and namelen as applicable. Set the returned
340 * namelen to 0 for older code which might ignore the return value
344 if (sa && name && namelen) {
345 if (*namelen > sa->sa_len)
346 *namelen = sa->sa_len;
356 * If an error occured clear the reserved descriptor, else associate
359 * Note that *res is normally ignored if an error is returned but
360 * a syscall message will still have access to the result code.
363 fsetfd(fdp, NULL, fd);
366 fsetfd(fdp, nfp, fd);
374 * accept(int s, caddr_t name, int *anamelen)
379 sys_accept(struct accept_args *uap)
381 struct sockaddr *sa = NULL;
386 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
390 error = kern_accept(uap->s, 0, &sa, &sa_len,
391 &uap->sysmsg_iresult);
394 error = copyout(sa, uap->name, sa_len);
396 error = copyout(&sa_len, uap->anamelen,
397 sizeof(*uap->anamelen));
402 error = kern_accept(uap->s, 0, NULL, 0,
403 &uap->sysmsg_iresult);
409 * extaccept(int s, int fflags, caddr_t name, int *anamelen)
414 sys_extaccept(struct extaccept_args *uap)
416 struct sockaddr *sa = NULL;
419 int fflags = uap->flags & O_FMASK;
422 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
426 error = kern_accept(uap->s, fflags, &sa, &sa_len,
427 &uap->sysmsg_iresult);
430 error = copyout(sa, uap->name, sa_len);
432 error = copyout(&sa_len, uap->anamelen,
433 sizeof(*uap->anamelen));
438 error = kern_accept(uap->s, fflags, NULL, 0,
439 &uap->sysmsg_iresult);
446 * Returns TRUE if predicate satisfied.
449 soconnected_predicate(struct netmsg_so_notify *msg)
451 struct socket *so = msg->base.nm_so;
453 /* check predicate */
454 if (!(so->so_state & SS_ISCONNECTING) || so->so_error != 0) {
455 msg->base.lmsg.ms_error = so->so_error;
463 kern_connect(int s, int fflags, struct sockaddr *sa)
465 struct thread *td = curthread;
466 struct proc *p = td->td_proc;
469 int error, interrupted = 0;
471 error = holdsock(p->p_fd, s, &fp);
474 so = (struct socket *)fp->f_data;
476 if (fflags & O_FBLOCKING)
477 /* fflags &= ~FNONBLOCK; */;
478 else if (fflags & O_FNONBLOCKING)
483 if (so->so_state & SS_ISCONNECTING) {
487 error = soconnect(so, sa, td);
490 if ((fflags & FNONBLOCK) && (so->so_state & SS_ISCONNECTING)) {
494 if ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
495 struct netmsg_so_notify msg;
497 netmsg_init_abortable(&msg.base, so,
498 &curthread->td_msgport,
501 netmsg_so_notify_doabort);
502 msg.nm_predicate = soconnected_predicate;
503 msg.nm_etype = NM_REVENT;
504 error = lwkt_domsg(so->so_port, &msg.base.lmsg, PCATCH);
505 if (error == EINTR || error == ERESTART)
509 error = so->so_error;
514 soclrstate(so, SS_ISCONNECTING);
515 if (error == ERESTART)
523 * connect_args(int s, caddr_t name, int namelen)
528 sys_connect(struct connect_args *uap)
533 error = getsockaddr(&sa, uap->name, uap->namelen);
536 error = kern_connect(uap->s, 0, sa);
543 * connect_args(int s, int fflags, caddr_t name, int namelen)
548 sys_extconnect(struct extconnect_args *uap)
552 int fflags = uap->flags & O_FMASK;
554 error = getsockaddr(&sa, uap->name, uap->namelen);
557 error = kern_connect(uap->s, fflags, sa);
564 kern_socketpair(int domain, int type, int protocol, int *sv)
566 struct thread *td = curthread;
567 struct filedesc *fdp;
568 struct file *fp1, *fp2;
569 struct socket *so1, *so2;
572 fdp = td->td_proc->p_fd;
573 error = socreate(domain, &so1, type, protocol, td);
576 error = socreate(domain, &so2, type, protocol, td);
579 error = falloc(td->td_lwp, &fp1, &fd1);
584 error = falloc(td->td_lwp, &fp2, &fd2);
589 error = soconnect2(so1, so2);
592 if (type == SOCK_DGRAM) {
594 * Datagram socket connection is asymmetric.
596 error = soconnect2(so2, so1);
600 fp1->f_type = fp2->f_type = DTYPE_SOCKET;
601 fp1->f_flag = fp2->f_flag = FREAD|FWRITE;
602 fp1->f_ops = fp2->f_ops = &socketops;
603 fsetfd(fdp, fp1, fd1);
604 fsetfd(fdp, fp2, fd2);
609 fsetfd(fdp, NULL, fd2);
612 fsetfd(fdp, NULL, fd1);
615 (void)soclose(so2, 0);
617 (void)soclose(so1, 0);
622 * socketpair(int domain, int type, int protocol, int *rsv)
627 sys_socketpair(struct socketpair_args *uap)
631 error = kern_socketpair(uap->domain, uap->type, uap->protocol, sockv);
634 error = copyout(sockv, uap->rsv, sizeof(sockv));
639 kern_sendmsg(int s, struct sockaddr *sa, struct uio *auio,
640 struct mbuf *control, int flags, size_t *res)
642 struct thread *td = curthread;
643 struct lwp *lp = td->td_lwp;
644 struct proc *p = td->td_proc;
650 struct iovec *ktriov = NULL;
654 error = holdsock(p->p_fd, s, &fp);
658 if (KTRPOINT(td, KTR_GENIO)) {
659 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
661 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
662 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
666 len = auio->uio_resid;
667 so = (struct socket *)fp->f_data;
668 if ((flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
669 if (fp->f_flag & FNONBLOCK)
670 flags |= MSG_FNONBLOCKING;
672 error = so_pru_sosend(so, sa, auio, NULL, control, flags, td);
674 if (auio->uio_resid != len && (error == ERESTART ||
675 error == EINTR || error == EWOULDBLOCK))
677 if (error == EPIPE && !(flags & MSG_NOSIGNAL))
678 lwpsignal(p, lp, SIGPIPE);
681 if (ktriov != NULL) {
683 ktruio.uio_iov = ktriov;
684 ktruio.uio_resid = len - auio->uio_resid;
685 ktrgenio(lp, s, UIO_WRITE, &ktruio, error);
687 FREE(ktriov, M_TEMP);
691 *res = len - auio->uio_resid;
697 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
702 sys_sendto(struct sendto_args *uap)
704 struct thread *td = curthread;
707 struct sockaddr *sa = NULL;
711 error = getsockaddr(&sa, uap->to, uap->tolen);
715 aiov.iov_base = uap->buf;
716 aiov.iov_len = uap->len;
717 auio.uio_iov = &aiov;
720 auio.uio_resid = uap->len;
721 auio.uio_segflg = UIO_USERSPACE;
722 auio.uio_rw = UIO_WRITE;
725 error = kern_sendmsg(uap->s, sa, &auio, NULL, uap->flags,
726 &uap->sysmsg_szresult);
734 * sendmsg_args(int s, caddr_t msg, int flags)
739 sys_sendmsg(struct sendmsg_args *uap)
741 struct thread *td = curthread;
744 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
745 struct sockaddr *sa = NULL;
746 struct mbuf *control = NULL;
749 error = copyin(uap->msg, (caddr_t)&msg, sizeof(msg));
754 * Conditionally copyin msg.msg_name.
757 error = getsockaddr(&sa, msg.msg_name, msg.msg_namelen);
765 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
770 auio.uio_iovcnt = msg.msg_iovlen;
772 auio.uio_segflg = UIO_USERSPACE;
773 auio.uio_rw = UIO_WRITE;
777 * Conditionally copyin msg.msg_control.
779 if (msg.msg_control) {
780 if (msg.msg_controllen < sizeof(struct cmsghdr) ||
781 msg.msg_controllen > MLEN) {
785 control = m_get(MB_WAIT, MT_CONTROL);
786 if (control == NULL) {
790 control->m_len = msg.msg_controllen;
791 error = copyin(msg.msg_control, mtod(control, caddr_t),
799 error = kern_sendmsg(uap->s, sa, &auio, control, uap->flags,
800 &uap->sysmsg_szresult);
803 iovec_free(&iov, aiov);
811 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
812 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
813 * Don't forget to FREE() and m_free() these if they are returned.
816 kern_recvmsg(int s, struct sockaddr **sa, struct uio *auio,
817 struct mbuf **control, int *flags, size_t *res)
819 struct thread *td = curthread;
820 struct proc *p = td->td_proc;
827 struct iovec *ktriov = NULL;
831 error = holdsock(p->p_fd, s, &fp);
835 if (KTRPOINT(td, KTR_GENIO)) {
836 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
838 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
839 bcopy(auio->uio_iov, ktriov, iovlen);
843 len = auio->uio_resid;
844 so = (struct socket *)fp->f_data;
846 if (flags == NULL || (*flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
847 if (fp->f_flag & FNONBLOCK) {
849 *flags |= MSG_FNONBLOCKING;
851 lflags = MSG_FNONBLOCKING;
857 error = so_pru_soreceive(so, sa, auio, NULL, control, flags);
859 if (auio->uio_resid != len && (error == ERESTART ||
860 error == EINTR || error == EWOULDBLOCK))
864 if (ktriov != NULL) {
866 ktruio.uio_iov = ktriov;
867 ktruio.uio_resid = len - auio->uio_resid;
868 ktrgenio(td->td_lwp, s, UIO_READ, &ktruio, error);
870 FREE(ktriov, M_TEMP);
874 *res = len - auio->uio_resid;
880 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
881 * caddr_t from, int *fromlenaddr)
886 sys_recvfrom(struct recvfrom_args *uap)
888 struct thread *td = curthread;
891 struct sockaddr *sa = NULL;
894 if (uap->from && uap->fromlenaddr) {
895 error = copyin(uap->fromlenaddr, &fromlen, sizeof(fromlen));
903 aiov.iov_base = uap->buf;
904 aiov.iov_len = uap->len;
905 auio.uio_iov = &aiov;
908 auio.uio_resid = uap->len;
909 auio.uio_segflg = UIO_USERSPACE;
910 auio.uio_rw = UIO_READ;
913 error = kern_recvmsg(uap->s, uap->from ? &sa : NULL, &auio, NULL,
914 &uap->flags, &uap->sysmsg_szresult);
916 if (error == 0 && uap->from) {
917 /* note: sa may still be NULL */
919 fromlen = MIN(fromlen, sa->sa_len);
920 error = copyout(sa, uap->from, fromlen);
925 error = copyout(&fromlen, uap->fromlenaddr,
936 * recvmsg_args(int s, struct msghdr *msg, int flags)
941 sys_recvmsg(struct recvmsg_args *uap)
943 struct thread *td = curthread;
946 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
947 struct mbuf *m, *control = NULL;
948 struct sockaddr *sa = NULL;
950 socklen_t *ufromlenp, *ucontrollenp;
951 int error, fromlen, controllen, len, flags, *uflagsp;
954 * This copyin handles everything except the iovec.
956 error = copyin(uap->msg, &msg, sizeof(msg));
960 if (msg.msg_name && msg.msg_namelen < 0)
962 if (msg.msg_control && msg.msg_controllen < 0)
965 ufromlenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
967 ucontrollenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
969 uflagsp = (int *)((caddr_t)uap->msg + offsetof(struct msghdr,
975 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
980 auio.uio_iovcnt = msg.msg_iovlen;
982 auio.uio_segflg = UIO_USERSPACE;
983 auio.uio_rw = UIO_READ;
988 error = kern_recvmsg(uap->s,
989 (msg.msg_name ? &sa : NULL), &auio,
990 (msg.msg_control ? &control : NULL), &flags,
991 &uap->sysmsg_szresult);
994 * Conditionally copyout the name and populate the namelen field.
996 if (error == 0 && msg.msg_name) {
997 /* note: sa may still be NULL */
999 fromlen = MIN(msg.msg_namelen, sa->sa_len);
1000 error = copyout(sa, msg.msg_name, fromlen);
1005 error = copyout(&fromlen, ufromlenp,
1006 sizeof(*ufromlenp));
1010 * Copyout msg.msg_control and msg.msg_controllen.
1012 if (error == 0 && msg.msg_control) {
1013 len = msg.msg_controllen;
1015 ctlbuf = (caddr_t)msg.msg_control;
1017 while(m && len > 0) {
1018 unsigned int tocopy;
1020 if (len >= m->m_len) {
1023 msg.msg_flags |= MSG_CTRUNC;
1027 error = copyout(mtod(m, caddr_t), ctlbuf, tocopy);
1035 controllen = ctlbuf - (caddr_t)msg.msg_control;
1036 error = copyout(&controllen, ucontrollenp,
1037 sizeof(*ucontrollenp));
1041 error = copyout(&flags, uflagsp, sizeof(*uflagsp));
1046 iovec_free(&iov, aiov);
1053 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1054 * in kernel pointer instead of a userland pointer. This allows us
1055 * to manipulate socket options in the emulation code.
1058 kern_setsockopt(int s, struct sockopt *sopt)
1060 struct thread *td = curthread;
1061 struct proc *p = td->td_proc;
1065 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1067 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1069 if (sopt->sopt_valsize < 0)
1072 error = holdsock(p->p_fd, s, &fp);
1076 error = sosetopt((struct socket *)fp->f_data, sopt);
1082 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
1087 sys_setsockopt(struct setsockopt_args *uap)
1089 struct thread *td = curthread;
1090 struct sockopt sopt;
1093 sopt.sopt_level = uap->level;
1094 sopt.sopt_name = uap->name;
1095 sopt.sopt_valsize = uap->valsize;
1097 sopt.sopt_val = NULL;
1099 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1102 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1103 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1108 error = kern_setsockopt(uap->s, &sopt);
1111 kfree(sopt.sopt_val, M_TEMP);
1116 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1117 * in kernel pointer instead of a userland pointer. This allows us
1118 * to manipulate socket options in the emulation code.
1121 kern_getsockopt(int s, struct sockopt *sopt)
1123 struct thread *td = curthread;
1124 struct proc *p = td->td_proc;
1128 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1130 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1132 if (sopt->sopt_valsize < 0 || sopt->sopt_valsize > SOMAXOPT_SIZE)
1135 error = holdsock(p->p_fd, s, &fp);
1139 error = sogetopt((struct socket *)fp->f_data, sopt);
1145 * getsockopt_args(int s, int level, int name, caddr_t val, int *avalsize)
1150 sys_getsockopt(struct getsockopt_args *uap)
1152 struct thread *td = curthread;
1153 struct sockopt sopt;
1157 error = copyin(uap->avalsize, &valsize, sizeof(valsize));
1164 sopt.sopt_level = uap->level;
1165 sopt.sopt_name = uap->name;
1166 sopt.sopt_valsize = valsize;
1168 sopt.sopt_val = NULL;
1170 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1173 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1174 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1179 error = kern_getsockopt(uap->s, &sopt);
1182 valsize = sopt.sopt_valsize;
1183 error = copyout(&valsize, uap->avalsize, sizeof(valsize));
1187 error = copyout(sopt.sopt_val, uap->val, sopt.sopt_valsize);
1190 kfree(sopt.sopt_val, M_TEMP);
1195 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1196 * This allows kern_getsockname() to return a pointer to an allocated struct
1197 * sockaddr which must be freed later with FREE(). The caller must
1198 * initialize *name to NULL.
1201 kern_getsockname(int s, struct sockaddr **name, int *namelen)
1203 struct thread *td = curthread;
1204 struct proc *p = td->td_proc;
1207 struct sockaddr *sa = NULL;
1210 error = holdsock(p->p_fd, s, &fp);
1217 so = (struct socket *)fp->f_data;
1218 error = so_pru_sockaddr(so, &sa);
1223 *namelen = MIN(*namelen, sa->sa_len);
1233 * getsockname_args(int fdes, caddr_t asa, int *alen)
1240 sys_getsockname(struct getsockname_args *uap)
1242 struct sockaddr *sa = NULL;
1245 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1249 error = kern_getsockname(uap->fdes, &sa, &sa_len);
1252 error = copyout(sa, uap->asa, sa_len);
1254 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1261 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1262 * This allows kern_getpeername() to return a pointer to an allocated struct
1263 * sockaddr which must be freed later with FREE(). The caller must
1264 * initialize *name to NULL.
1267 kern_getpeername(int s, struct sockaddr **name, int *namelen)
1269 struct thread *td = curthread;
1270 struct proc *p = td->td_proc;
1273 struct sockaddr *sa = NULL;
1276 error = holdsock(p->p_fd, s, &fp);
1283 so = (struct socket *)fp->f_data;
1284 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1288 error = so_pru_peeraddr(so, &sa);
1293 *namelen = MIN(*namelen, sa->sa_len);
1303 * getpeername_args(int fdes, caddr_t asa, int *alen)
1305 * Get name of peer for connected socket.
1310 sys_getpeername(struct getpeername_args *uap)
1312 struct sockaddr *sa = NULL;
1315 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1319 error = kern_getpeername(uap->fdes, &sa, &sa_len);
1322 error = copyout(sa, uap->asa, sa_len);
1324 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1331 getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len)
1333 struct sockaddr *sa;
1337 if (len > SOCK_MAXADDRLEN)
1338 return ENAMETOOLONG;
1339 if (len < offsetof(struct sockaddr, sa_data[0]))
1341 MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1342 error = copyin(uaddr, sa, len);
1346 #if BYTE_ORDER != BIG_ENDIAN
1348 * The bind(), connect(), and sendto() syscalls were not
1349 * versioned for COMPAT_43. Thus, this check must stay.
1351 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1352 sa->sa_family = sa->sa_len;
1361 * Detach a mapped page and release resources back to the system.
1362 * We must release our wiring and if the object is ripped out
1363 * from under the vm_page we become responsible for freeing the
1369 sf_buf_mfree(void *arg)
1371 struct sf_buf *sf = arg;
1374 m = sf_buf_page(sf);
1375 if (sf_buf_free(sf)) {
1376 /* sf invalid now */
1377 vm_page_unwire(m, 0);
1378 if (m->wire_count == 0 && m->object == NULL)
1379 vm_page_try_to_free(m);
1385 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1386 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1388 * Send a file specified by 'fd' and starting at 'offset' to a socket
1389 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1390 * nbytes == 0. Optionally add a header and/or trailer to the socket
1391 * output. If specified, write the total number of bytes sent into *sbytes.
1393 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1394 * the headers to count against the remaining bytes to be sent from
1395 * the file descriptor. We may wish to implement a compatibility syscall
1401 sys_sendfile(struct sendfile_args *uap)
1403 struct thread *td = curthread;
1404 struct proc *p = td->td_proc;
1406 struct vnode *vp = NULL;
1407 struct sf_hdtr hdtr;
1408 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
1410 struct mbuf *mheader = NULL;
1413 off_t hdtr_size = 0;
1420 * Do argument checking. Must be a regular file in, stream
1421 * type and connected socket out, positive offset.
1423 fp = holdfp(p->p_fd, uap->fd, FREAD);
1427 if (fp->f_type != DTYPE_VNODE) {
1431 vp = (struct vnode *)fp->f_data;
1436 * If specified, get the pointer to the sf_hdtr struct for
1437 * any headers/trailers.
1440 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1447 error = iovec_copyin(hdtr.headers, &iov, aiov,
1448 hdtr.hdr_cnt, &hbytes);
1452 auio.uio_iovcnt = hdtr.hdr_cnt;
1453 auio.uio_offset = 0;
1454 auio.uio_segflg = UIO_USERSPACE;
1455 auio.uio_rw = UIO_WRITE;
1457 auio.uio_resid = hbytes;
1459 mheader = m_uiomove(&auio);
1461 iovec_free(&iov, aiov);
1462 if (mheader == NULL)
1467 error = kern_sendfile(vp, uap->s, uap->offset, uap->nbytes, mheader,
1468 &sbytes, uap->flags);
1473 * Send trailers. Wimp out and use writev(2).
1475 if (uap->hdtr != NULL && hdtr.trailers != NULL) {
1476 error = iovec_copyin(hdtr.trailers, &iov, aiov,
1477 hdtr.trl_cnt, &auio.uio_resid);
1481 auio.uio_iovcnt = hdtr.trl_cnt;
1482 auio.uio_offset = 0;
1483 auio.uio_segflg = UIO_USERSPACE;
1484 auio.uio_rw = UIO_WRITE;
1487 error = kern_sendmsg(uap->s, NULL, &auio, NULL, 0, &tbytes);
1489 iovec_free(&iov, aiov);
1492 hdtr_size += tbytes; /* trailer bytes successfully sent */
1498 if (uap->sbytes != NULL) {
1499 sbytes += hdtr_size;
1500 copyout(&sbytes, uap->sbytes, sizeof(off_t));
1506 kern_sendfile(struct vnode *vp, int sfd, off_t offset, size_t nbytes,
1507 struct mbuf *mheader, off_t *sbytes, int flags)
1509 struct thread *td = curthread;
1510 struct proc *p = td->td_proc;
1511 struct vm_object *obj;
1521 if (vp->v_type != VREG) {
1525 if ((obj = vp->v_object) == NULL) {
1529 error = holdsock(p->p_fd, sfd, &fp);
1532 so = (struct socket *)fp->f_data;
1533 if (so->so_type != SOCK_STREAM) {
1537 if ((so->so_state & SS_ISCONNECTED) == 0) {
1548 * Protect against multiple writers to the socket.
1550 ssb_lock(&so->so_snd, M_WAITOK);
1553 * Loop through the pages in the file, starting with the requested
1554 * offset. Get a file page (do I/O if necessary), map the file page
1555 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1558 for (off = offset; ; off += xfsize, *sbytes += xfsize + hbytes) {
1562 pindex = OFF_TO_IDX(off);
1565 * Calculate the amount to transfer. Not to exceed a page,
1566 * the EOF, or the passed in nbytes.
1568 xfsize = vp->v_filesize - off;
1569 if (xfsize > PAGE_SIZE)
1571 pgoff = (vm_offset_t)(off & PAGE_MASK);
1572 if (PAGE_SIZE - pgoff < xfsize)
1573 xfsize = PAGE_SIZE - pgoff;
1574 if (nbytes && xfsize > (nbytes - *sbytes))
1575 xfsize = nbytes - *sbytes;
1579 * Optimize the non-blocking case by looking at the socket space
1580 * before going to the extra work of constituting the sf_buf.
1582 if ((fp->f_flag & FNONBLOCK) && ssb_space(&so->so_snd) <= 0) {
1583 if (so->so_state & SS_CANTSENDMORE)
1587 ssb_unlock(&so->so_snd);
1591 * Attempt to look up the page.
1593 * Allocate if not found, wait and loop if busy, then
1594 * wire the page. critical section protection is
1595 * required to maintain the object association (an
1596 * interrupt can free the page) through to the
1597 * vm_page_wire() call.
1599 lwkt_gettoken(&vm_token);
1600 pg = vm_page_lookup(obj, pindex);
1602 pg = vm_page_alloc(obj, pindex, VM_ALLOC_NORMAL);
1605 lwkt_reltoken(&vm_token);
1610 } else if (vm_page_sleep_busy(pg, TRUE, "sfpbsy")) {
1611 lwkt_reltoken(&vm_token);
1616 lwkt_reltoken(&vm_token);
1619 * If page is not valid for what we need, initiate I/O
1622 if (!pg->valid || !vm_page_is_valid(pg, pgoff, xfsize)) {
1628 * Ensure that our page is still around when the I/O
1631 vm_page_io_start(pg);
1634 * Get the page from backing store.
1636 bsize = vp->v_mount->mnt_stat.f_iosize;
1637 auio.uio_iov = &aiov;
1638 auio.uio_iovcnt = 1;
1640 aiov.iov_len = MAXBSIZE;
1641 auio.uio_resid = MAXBSIZE;
1642 auio.uio_offset = trunc_page(off);
1643 auio.uio_segflg = UIO_NOCOPY;
1644 auio.uio_rw = UIO_READ;
1646 vn_lock(vp, LK_SHARED | LK_RETRY);
1647 error = VOP_READ(vp, &auio,
1648 IO_VMIO | ((MAXBSIZE / bsize) << 16),
1651 vm_page_flag_clear(pg, PG_ZERO);
1652 vm_page_io_finish(pg);
1655 vm_page_unwire(pg, 0);
1656 vm_page_try_to_free(pg);
1658 ssb_unlock(&so->so_snd);
1665 * Get a sendfile buf. We usually wait as long as necessary,
1666 * but this wait can be interrupted.
1668 if ((sf = sf_buf_alloc(pg)) == NULL) {
1670 vm_page_unwire(pg, 0);
1671 vm_page_try_to_free(pg);
1673 ssb_unlock(&so->so_snd);
1679 * Get an mbuf header and set it up as having external storage.
1681 MGETHDR(m, MB_WAIT, MT_DATA);
1685 ssb_unlock(&so->so_snd);
1689 m->m_ext.ext_free = sf_buf_mfree;
1690 m->m_ext.ext_ref = sf_buf_ref;
1691 m->m_ext.ext_arg = sf;
1692 m->m_ext.ext_buf = (void *)sf_buf_kva(sf);
1693 m->m_ext.ext_size = PAGE_SIZE;
1694 m->m_data = (char *)sf_buf_kva(sf) + pgoff;
1695 m->m_flags |= M_EXT;
1696 m->m_pkthdr.len = m->m_len = xfsize;
1697 KKASSERT((m->m_flags & (M_EXT_CLUSTER)) == 0);
1699 if (mheader != NULL) {
1700 hbytes = mheader->m_pkthdr.len;
1701 mheader->m_pkthdr.len += m->m_pkthdr.len;
1709 * Add the buffer to the socket buffer chain.
1714 * Make sure that the socket is still able to take more data.
1715 * CANTSENDMORE being true usually means that the connection
1716 * was closed. so_error is true when an error was sensed after
1718 * The state is checked after the page mapping and buffer
1719 * allocation above since those operations may block and make
1720 * any socket checks stale. From this point forward, nothing
1721 * blocks before the pru_send (or more accurately, any blocking
1722 * results in a loop back to here to re-check).
1724 if ((so->so_state & SS_CANTSENDMORE) || so->so_error) {
1725 if (so->so_state & SS_CANTSENDMORE) {
1728 error = so->so_error;
1732 ssb_unlock(&so->so_snd);
1737 * Wait for socket space to become available. We do this just
1738 * after checking the connection state above in order to avoid
1739 * a race condition with ssb_wait().
1741 if (ssb_space(&so->so_snd) < so->so_snd.ssb_lowat) {
1742 if (fp->f_flag & FNONBLOCK) {
1744 ssb_unlock(&so->so_snd);
1749 error = ssb_wait(&so->so_snd);
1751 * An error from ssb_wait usually indicates that we've
1752 * been interrupted by a signal. If we've sent anything
1753 * then return bytes sent, otherwise return the error.
1757 ssb_unlock(&so->so_snd);
1763 error = so_pru_send(so, 0, m, NULL, NULL, td);
1766 ssb_unlock(&so->so_snd);
1770 if (mheader != NULL) {
1771 *sbytes += mheader->m_pkthdr.len;
1772 error = so_pru_send(so, 0, mheader, NULL, NULL, td);
1775 ssb_unlock(&so->so_snd);
1780 if (mheader != NULL)
1789 sys_sctp_peeloff(struct sctp_peeloff_args *uap)
1792 struct thread *td = curthread;
1793 struct filedesc *fdp = td->td_proc->p_fd;
1794 struct file *lfp = NULL;
1795 struct file *nfp = NULL;
1797 struct socket *head, *so;
1800 short fflag; /* type must match fp->f_flag */
1802 assoc_id = uap->name;
1803 error = holdsock(td->td_proc->p_fd, uap->sd, &lfp);
1808 head = (struct socket *)lfp->f_data;
1809 error = sctp_can_peel_off(head, assoc_id);
1815 * At this point we know we do have a assoc to pull
1816 * we proceed to get the fd setup. This may block
1820 fflag = lfp->f_flag;
1821 error = falloc(td->td_lwp, &nfp, &fd);
1824 * Probably ran out of file descriptors. Put the
1825 * unaccepted connection back onto the queue and
1826 * do another wakeup so some other process might
1827 * have a chance at it.
1832 uap->sysmsg_iresult = fd;
1834 so = sctp_get_peeloff(head, assoc_id, &error);
1837 * Either someone else peeled it off OR
1838 * we can't get a socket.
1842 soreference(so); /* reference needed */
1843 soclrstate(so, SS_NOFDREF | SS_COMP); /* when clearing NOFDREF */
1845 if (head->so_sigio != NULL)
1846 fsetown(fgetown(&head->so_sigio), &so->so_sigio);
1848 nfp->f_type = DTYPE_SOCKET;
1849 nfp->f_flag = fflag;
1850 nfp->f_ops = &socketops;
1855 * Assign the file pointer to the reserved descriptor, or clear
1856 * the reserved descriptor if an error occured.
1859 fsetfd(fdp, NULL, fd);
1861 fsetfd(fdp, nfp, fd);
1864 * Release explicitly held references before returning.