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 *msg0)
221 struct netmsg_so_notify *msg = (struct netmsg_so_notify *)msg0;
222 struct socket *head = msg->nm_so;
224 if (head->so_error != 0) {
225 msg->nm_netmsg.nm_lmsg.ms_error = head->so_error;
228 lwkt_gettoken(&head->so_rcv.ssb_token);
229 if (!TAILQ_EMPTY(&head->so_comp)) {
230 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
231 msg->nm_so = TAILQ_FIRST(&head->so_comp);
232 TAILQ_REMOVE(&head->so_comp, msg->nm_so, so_list);
235 msg->nm_netmsg.nm_lmsg.ms_error = 0;
236 lwkt_reltoken(&head->so_rcv.ssb_token);
239 lwkt_reltoken(&head->so_rcv.ssb_token);
240 if (head->so_state & SS_CANTRCVMORE) {
241 msg->nm_netmsg.nm_lmsg.ms_error = ECONNABORTED;
244 if (msg->nm_fflags & FNONBLOCK) {
245 msg->nm_netmsg.nm_lmsg.ms_error = EWOULDBLOCK;
253 * The second argument to kern_accept() is a handle to a struct sockaddr.
254 * This allows kern_accept() to return a pointer to an allocated struct
255 * sockaddr which must be freed later with FREE(). The caller must
256 * initialize *name to NULL.
259 kern_accept(int s, int fflags, struct sockaddr **name, int *namelen, int *res)
261 struct thread *td = curthread;
262 struct filedesc *fdp = td->td_proc->p_fd;
263 struct file *lfp = NULL;
264 struct file *nfp = NULL;
266 struct socket *head, *so;
267 struct netmsg_so_notify msg;
269 u_int fflag; /* type must match fp->f_flag */
273 if (name && namelen && *namelen < 0)
276 error = holdsock(td->td_proc->p_fd, s, &lfp);
280 error = falloc(td->td_lwp, &nfp, &fd);
281 if (error) { /* Probably ran out of file descriptors. */
285 head = (struct socket *)lfp->f_data;
286 if ((head->so_options & SO_ACCEPTCONN) == 0) {
291 if (fflags & O_FBLOCKING)
292 fflags |= lfp->f_flag & ~FNONBLOCK;
293 else if (fflags & O_FNONBLOCKING)
294 fflags |= lfp->f_flag | FNONBLOCK;
296 fflags = lfp->f_flag;
298 /* optimize for uniprocessor case later XXX JH */
299 netmsg_init_abortable(&msg.nm_netmsg, head, &curthread->td_msgport,
300 0, netmsg_so_notify, netmsg_so_notify_doabort);
301 msg.nm_predicate = soaccept_predicate;
302 msg.nm_fflags = fflags;
304 msg.nm_etype = NM_REVENT;
305 error = lwkt_domsg(head->so_port, &msg.nm_netmsg.nm_lmsg, PCATCH);
310 * At this point we have the connection that's ready to be accepted.
316 /* connection has been removed from the listen queue */
317 KNOTE(&head->so_rcv.ssb_kq.ki_note, 0);
319 soclrstate(so, SS_COMP);
321 if (head->so_sigio != NULL)
322 fsetown(fgetown(head->so_sigio), &so->so_sigio);
324 nfp->f_type = DTYPE_SOCKET;
326 nfp->f_ops = &socketops;
328 /* Sync socket nonblocking/async state with file flags */
329 tmp = fflag & FNONBLOCK;
330 fo_ioctl(nfp, FIONBIO, (caddr_t)&tmp, td->td_ucred, NULL);
331 tmp = fflag & FASYNC;
332 fo_ioctl(nfp, FIOASYNC, (caddr_t)&tmp, td->td_ucred, NULL);
335 error = soaccept(so, &sa);
338 * Set the returned name and namelen as applicable. Set the returned
339 * namelen to 0 for older code which might ignore the return value
343 if (sa && name && namelen) {
344 if (*namelen > sa->sa_len)
345 *namelen = sa->sa_len;
355 * If an error occured clear the reserved descriptor, else associate
358 * Note that *res is normally ignored if an error is returned but
359 * a syscall message will still have access to the result code.
362 fsetfd(fdp, NULL, fd);
365 fsetfd(fdp, nfp, fd);
373 * accept(int s, caddr_t name, int *anamelen)
378 sys_accept(struct accept_args *uap)
380 struct sockaddr *sa = NULL;
385 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
390 error = kern_accept(uap->s, 0, &sa, &sa_len,
391 &uap->sysmsg_iresult);
395 error = copyout(sa, uap->name, sa_len);
397 error = copyout(&sa_len, uap->anamelen,
398 sizeof(*uap->anamelen));
404 error = kern_accept(uap->s, 0, NULL, 0,
405 &uap->sysmsg_iresult);
412 * extaccept(int s, int fflags, caddr_t name, int *anamelen)
417 sys_extaccept(struct extaccept_args *uap)
419 struct sockaddr *sa = NULL;
422 int fflags = uap->flags & O_FMASK;
425 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
430 error = kern_accept(uap->s, fflags, &sa, &sa_len,
431 &uap->sysmsg_iresult);
435 error = copyout(sa, uap->name, sa_len);
437 error = copyout(&sa_len, uap->anamelen,
438 sizeof(*uap->anamelen));
444 error = kern_accept(uap->s, fflags, NULL, 0,
445 &uap->sysmsg_iresult);
453 * Returns TRUE if predicate satisfied.
456 soconnected_predicate(struct netmsg *msg0)
458 struct netmsg_so_notify *msg = (struct netmsg_so_notify *)msg0;
459 struct socket *so = msg->nm_so;
461 /* check predicate */
462 if (!(so->so_state & SS_ISCONNECTING) || so->so_error != 0) {
463 msg->nm_netmsg.nm_lmsg.ms_error = so->so_error;
471 kern_connect(int s, int fflags, struct sockaddr *sa)
473 struct thread *td = curthread;
474 struct proc *p = td->td_proc;
477 int error, interrupted = 0;
479 error = holdsock(p->p_fd, s, &fp);
482 so = (struct socket *)fp->f_data;
484 if (fflags & O_FBLOCKING)
485 /* fflags &= ~FNONBLOCK; */;
486 else if (fflags & O_FNONBLOCKING)
491 if (so->so_state & SS_ISCONNECTING) {
495 error = soconnect(so, sa, td);
498 if ((fflags & FNONBLOCK) && (so->so_state & SS_ISCONNECTING)) {
502 if ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
503 struct netmsg_so_notify msg;
505 netmsg_init_abortable(&msg.nm_netmsg, so,
506 &curthread->td_msgport,
509 netmsg_so_notify_doabort);
510 msg.nm_predicate = soconnected_predicate;
512 msg.nm_etype = NM_REVENT;
513 error = lwkt_domsg(so->so_port, &msg.nm_netmsg.nm_lmsg, PCATCH);
514 if (error == EINTR || error == ERESTART)
518 error = so->so_error;
523 soclrstate(so, SS_ISCONNECTING);
524 if (error == ERESTART)
532 * connect_args(int s, caddr_t name, int namelen)
537 sys_connect(struct connect_args *uap)
542 error = getsockaddr(&sa, uap->name, uap->namelen);
546 error = kern_connect(uap->s, 0, sa);
554 * connect_args(int s, int fflags, caddr_t name, int namelen)
559 sys_extconnect(struct extconnect_args *uap)
563 int fflags = uap->flags & O_FMASK;
565 error = getsockaddr(&sa, uap->name, uap->namelen);
569 error = kern_connect(uap->s, fflags, sa);
577 kern_socketpair(int domain, int type, int protocol, int *sv)
579 struct thread *td = curthread;
580 struct filedesc *fdp;
581 struct file *fp1, *fp2;
582 struct socket *so1, *so2;
585 fdp = td->td_proc->p_fd;
586 error = socreate(domain, &so1, type, protocol, td);
589 error = socreate(domain, &so2, type, protocol, td);
592 error = falloc(td->td_lwp, &fp1, &fd1);
597 error = falloc(td->td_lwp, &fp2, &fd2);
602 error = soconnect2(so1, so2);
605 if (type == SOCK_DGRAM) {
607 * Datagram socket connection is asymmetric.
609 error = soconnect2(so2, so1);
613 fp1->f_type = fp2->f_type = DTYPE_SOCKET;
614 fp1->f_flag = fp2->f_flag = FREAD|FWRITE;
615 fp1->f_ops = fp2->f_ops = &socketops;
616 fsetfd(fdp, fp1, fd1);
617 fsetfd(fdp, fp2, fd2);
622 fsetfd(fdp, NULL, fd2);
625 fsetfd(fdp, NULL, fd1);
628 (void)soclose(so2, 0);
630 (void)soclose(so1, 0);
635 * socketpair(int domain, int type, int protocol, int *rsv)
640 sys_socketpair(struct socketpair_args *uap)
645 error = kern_socketpair(uap->domain, uap->type, uap->protocol, sockv);
649 error = copyout(sockv, uap->rsv, sizeof(sockv));
654 kern_sendmsg(int s, struct sockaddr *sa, struct uio *auio,
655 struct mbuf *control, int flags, size_t *res)
657 struct thread *td = curthread;
658 struct lwp *lp = td->td_lwp;
659 struct proc *p = td->td_proc;
665 struct iovec *ktriov = NULL;
669 error = holdsock(p->p_fd, s, &fp);
673 if (KTRPOINT(td, KTR_GENIO)) {
674 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
676 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
677 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
681 len = auio->uio_resid;
682 so = (struct socket *)fp->f_data;
683 if ((flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
684 if (fp->f_flag & FNONBLOCK)
685 flags |= MSG_FNONBLOCKING;
687 error = so_pru_sosend(so, sa, auio, NULL, control, flags, td);
689 if (auio->uio_resid != len && (error == ERESTART ||
690 error == EINTR || error == EWOULDBLOCK))
693 lwpsignal(p, lp, SIGPIPE);
696 if (ktriov != NULL) {
698 ktruio.uio_iov = ktriov;
699 ktruio.uio_resid = len - auio->uio_resid;
700 ktrgenio(lp, s, UIO_WRITE, &ktruio, error);
702 FREE(ktriov, M_TEMP);
706 *res = len - auio->uio_resid;
712 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
717 sys_sendto(struct sendto_args *uap)
719 struct thread *td = curthread;
722 struct sockaddr *sa = NULL;
726 error = getsockaddr(&sa, uap->to, uap->tolen);
730 aiov.iov_base = uap->buf;
731 aiov.iov_len = uap->len;
732 auio.uio_iov = &aiov;
735 auio.uio_resid = uap->len;
736 auio.uio_segflg = UIO_USERSPACE;
737 auio.uio_rw = UIO_WRITE;
741 error = kern_sendmsg(uap->s, sa, &auio, NULL, uap->flags,
742 &uap->sysmsg_szresult);
751 * sendmsg_args(int s, caddr_t msg, int flags)
756 sys_sendmsg(struct sendmsg_args *uap)
758 struct thread *td = curthread;
761 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
762 struct sockaddr *sa = NULL;
763 struct mbuf *control = NULL;
766 error = copyin(uap->msg, (caddr_t)&msg, sizeof(msg));
771 * Conditionally copyin msg.msg_name.
774 error = getsockaddr(&sa, msg.msg_name, msg.msg_namelen);
782 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
787 auio.uio_iovcnt = msg.msg_iovlen;
789 auio.uio_segflg = UIO_USERSPACE;
790 auio.uio_rw = UIO_WRITE;
794 * Conditionally copyin msg.msg_control.
796 if (msg.msg_control) {
797 if (msg.msg_controllen < sizeof(struct cmsghdr) ||
798 msg.msg_controllen > MLEN) {
802 control = m_get(MB_WAIT, MT_CONTROL);
803 if (control == NULL) {
807 control->m_len = msg.msg_controllen;
808 error = copyin(msg.msg_control, mtod(control, caddr_t),
817 error = kern_sendmsg(uap->s, sa, &auio, control, uap->flags,
818 &uap->sysmsg_szresult);
822 iovec_free(&iov, aiov);
830 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
831 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
832 * Don't forget to FREE() and m_free() these if they are returned.
835 kern_recvmsg(int s, struct sockaddr **sa, struct uio *auio,
836 struct mbuf **control, int *flags, size_t *res)
838 struct thread *td = curthread;
839 struct proc *p = td->td_proc;
846 struct iovec *ktriov = NULL;
850 error = holdsock(p->p_fd, s, &fp);
854 if (KTRPOINT(td, KTR_GENIO)) {
855 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
857 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
858 bcopy(auio->uio_iov, ktriov, iovlen);
862 len = auio->uio_resid;
863 so = (struct socket *)fp->f_data;
865 if (flags == NULL || (*flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
866 if (fp->f_flag & FNONBLOCK) {
868 *flags |= MSG_FNONBLOCKING;
870 lflags = MSG_FNONBLOCKING;
876 error = so_pru_soreceive(so, sa, auio, NULL, control, flags);
878 if (auio->uio_resid != len && (error == ERESTART ||
879 error == EINTR || error == EWOULDBLOCK))
883 if (ktriov != NULL) {
885 ktruio.uio_iov = ktriov;
886 ktruio.uio_resid = len - auio->uio_resid;
887 ktrgenio(td->td_lwp, s, UIO_READ, &ktruio, error);
889 FREE(ktriov, M_TEMP);
893 *res = len - auio->uio_resid;
899 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
900 * caddr_t from, int *fromlenaddr)
905 sys_recvfrom(struct recvfrom_args *uap)
907 struct thread *td = curthread;
910 struct sockaddr *sa = NULL;
913 if (uap->from && uap->fromlenaddr) {
914 error = copyin(uap->fromlenaddr, &fromlen, sizeof(fromlen));
922 aiov.iov_base = uap->buf;
923 aiov.iov_len = uap->len;
924 auio.uio_iov = &aiov;
927 auio.uio_resid = uap->len;
928 auio.uio_segflg = UIO_USERSPACE;
929 auio.uio_rw = UIO_READ;
933 error = kern_recvmsg(uap->s, uap->from ? &sa : NULL, &auio, NULL,
934 &uap->flags, &uap->sysmsg_szresult);
937 if (error == 0 && uap->from) {
938 /* note: sa may still be NULL */
940 fromlen = MIN(fromlen, sa->sa_len);
941 error = copyout(sa, uap->from, fromlen);
946 error = copyout(&fromlen, uap->fromlenaddr,
957 * recvmsg_args(int s, struct msghdr *msg, int flags)
962 sys_recvmsg(struct recvmsg_args *uap)
964 struct thread *td = curthread;
967 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
968 struct mbuf *m, *control = NULL;
969 struct sockaddr *sa = NULL;
971 socklen_t *ufromlenp, *ucontrollenp;
972 int error, fromlen, controllen, len, flags, *uflagsp;
975 * This copyin handles everything except the iovec.
977 error = copyin(uap->msg, &msg, sizeof(msg));
981 if (msg.msg_name && msg.msg_namelen < 0)
983 if (msg.msg_control && msg.msg_controllen < 0)
986 ufromlenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
988 ucontrollenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
990 uflagsp = (int *)((caddr_t)uap->msg + offsetof(struct msghdr,
996 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
1001 auio.uio_iovcnt = msg.msg_iovlen;
1002 auio.uio_offset = 0;
1003 auio.uio_segflg = UIO_USERSPACE;
1004 auio.uio_rw = UIO_READ;
1010 error = kern_recvmsg(uap->s,
1011 (msg.msg_name ? &sa : NULL), &auio,
1012 (msg.msg_control ? &control : NULL), &flags,
1013 &uap->sysmsg_szresult);
1017 * Conditionally copyout the name and populate the namelen field.
1019 if (error == 0 && msg.msg_name) {
1020 /* note: sa may still be NULL */
1022 fromlen = MIN(msg.msg_namelen, sa->sa_len);
1023 error = copyout(sa, msg.msg_name, fromlen);
1028 error = copyout(&fromlen, ufromlenp,
1029 sizeof(*ufromlenp));
1033 * Copyout msg.msg_control and msg.msg_controllen.
1035 if (error == 0 && msg.msg_control) {
1036 len = msg.msg_controllen;
1038 ctlbuf = (caddr_t)msg.msg_control;
1040 while(m && len > 0) {
1041 unsigned int tocopy;
1043 if (len >= m->m_len) {
1046 msg.msg_flags |= MSG_CTRUNC;
1050 error = copyout(mtod(m, caddr_t), ctlbuf, tocopy);
1058 controllen = ctlbuf - (caddr_t)msg.msg_control;
1059 error = copyout(&controllen, ucontrollenp,
1060 sizeof(*ucontrollenp));
1064 error = copyout(&flags, uflagsp, sizeof(*uflagsp));
1069 iovec_free(&iov, aiov);
1076 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1077 * in kernel pointer instead of a userland pointer. This allows us
1078 * to manipulate socket options in the emulation code.
1081 kern_setsockopt(int s, struct sockopt *sopt)
1083 struct thread *td = curthread;
1084 struct proc *p = td->td_proc;
1088 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1090 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1092 if (sopt->sopt_valsize < 0)
1095 error = holdsock(p->p_fd, s, &fp);
1099 error = sosetopt((struct socket *)fp->f_data, sopt);
1105 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
1110 sys_setsockopt(struct setsockopt_args *uap)
1112 struct thread *td = curthread;
1113 struct sockopt sopt;
1116 sopt.sopt_level = uap->level;
1117 sopt.sopt_name = uap->name;
1118 sopt.sopt_valsize = uap->valsize;
1120 sopt.sopt_val = NULL;
1122 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1125 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1126 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1132 error = kern_setsockopt(uap->s, &sopt);
1136 kfree(sopt.sopt_val, M_TEMP);
1141 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1142 * in kernel pointer instead of a userland pointer. This allows us
1143 * to manipulate socket options in the emulation code.
1146 kern_getsockopt(int s, struct sockopt *sopt)
1148 struct thread *td = curthread;
1149 struct proc *p = td->td_proc;
1153 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1155 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1157 if (sopt->sopt_valsize < 0 || sopt->sopt_valsize > SOMAXOPT_SIZE)
1160 error = holdsock(p->p_fd, s, &fp);
1164 error = sogetopt((struct socket *)fp->f_data, sopt);
1170 * getsockopt_args(int s, int level, int name, caddr_t val, int *avalsize)
1175 sys_getsockopt(struct getsockopt_args *uap)
1177 struct thread *td = curthread;
1178 struct sockopt sopt;
1182 error = copyin(uap->avalsize, &valsize, sizeof(valsize));
1189 sopt.sopt_level = uap->level;
1190 sopt.sopt_name = uap->name;
1191 sopt.sopt_valsize = valsize;
1193 sopt.sopt_val = NULL;
1195 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1198 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1199 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1205 error = kern_getsockopt(uap->s, &sopt);
1209 valsize = sopt.sopt_valsize;
1210 error = copyout(&valsize, uap->avalsize, sizeof(valsize));
1214 error = copyout(sopt.sopt_val, uap->val, sopt.sopt_valsize);
1217 kfree(sopt.sopt_val, M_TEMP);
1222 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1223 * This allows kern_getsockname() to return a pointer to an allocated struct
1224 * sockaddr which must be freed later with FREE(). The caller must
1225 * initialize *name to NULL.
1228 kern_getsockname(int s, struct sockaddr **name, int *namelen)
1230 struct thread *td = curthread;
1231 struct proc *p = td->td_proc;
1234 struct sockaddr *sa = NULL;
1237 error = holdsock(p->p_fd, s, &fp);
1244 so = (struct socket *)fp->f_data;
1245 error = so_pru_sockaddr(so, &sa);
1250 *namelen = MIN(*namelen, sa->sa_len);
1260 * getsockname_args(int fdes, caddr_t asa, int *alen)
1267 sys_getsockname(struct getsockname_args *uap)
1269 struct sockaddr *sa = NULL;
1272 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1277 error = kern_getsockname(uap->fdes, &sa, &sa_len);
1281 error = copyout(sa, uap->asa, sa_len);
1283 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1290 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1291 * This allows kern_getpeername() to return a pointer to an allocated struct
1292 * sockaddr which must be freed later with FREE(). The caller must
1293 * initialize *name to NULL.
1296 kern_getpeername(int s, struct sockaddr **name, int *namelen)
1298 struct thread *td = curthread;
1299 struct proc *p = td->td_proc;
1302 struct sockaddr *sa = NULL;
1305 error = holdsock(p->p_fd, s, &fp);
1312 so = (struct socket *)fp->f_data;
1313 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1317 error = so_pru_peeraddr(so, &sa);
1322 *namelen = MIN(*namelen, sa->sa_len);
1332 * getpeername_args(int fdes, caddr_t asa, int *alen)
1334 * Get name of peer for connected socket.
1339 sys_getpeername(struct getpeername_args *uap)
1341 struct sockaddr *sa = NULL;
1344 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1349 error = kern_getpeername(uap->fdes, &sa, &sa_len);
1353 error = copyout(sa, uap->asa, sa_len);
1355 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1362 getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len)
1364 struct sockaddr *sa;
1368 if (len > SOCK_MAXADDRLEN)
1369 return ENAMETOOLONG;
1370 if (len < offsetof(struct sockaddr, sa_data[0]))
1372 MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1373 error = copyin(uaddr, sa, len);
1377 #if BYTE_ORDER != BIG_ENDIAN
1379 * The bind(), connect(), and sendto() syscalls were not
1380 * versioned for COMPAT_43. Thus, this check must stay.
1382 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1383 sa->sa_family = sa->sa_len;
1392 * Detach a mapped page and release resources back to the system.
1393 * We must release our wiring and if the object is ripped out
1394 * from under the vm_page we become responsible for freeing the
1395 * page. These routines must be MPSAFE.
1397 * XXX HACK XXX TEMPORARY UNTIL WE IMPLEMENT EXT MBUF REFERENCE COUNTING
1399 * XXX vm_page_*() routines are not MPSAFE yet, the MP lock is required.
1402 sf_buf_mfree(void *arg)
1404 struct sf_buf *sf = arg;
1408 * XXX vm_page_*() and SFBUF routines not MPSAFE yet.
1412 m = sf_buf_page(sf);
1413 if (sf_buf_free(sf) == 0) {
1414 vm_page_unwire(m, 0);
1415 if (m->wire_count == 0 && m->object == NULL)
1416 vm_page_try_to_free(m);
1424 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1425 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1427 * Send a file specified by 'fd' and starting at 'offset' to a socket
1428 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1429 * nbytes == 0. Optionally add a header and/or trailer to the socket
1430 * output. If specified, write the total number of bytes sent into *sbytes.
1432 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1433 * the headers to count against the remaining bytes to be sent from
1434 * the file descriptor. We may wish to implement a compatibility syscall
1440 sys_sendfile(struct sendfile_args *uap)
1442 struct thread *td = curthread;
1443 struct proc *p = td->td_proc;
1445 struct vnode *vp = NULL;
1446 struct sf_hdtr hdtr;
1447 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
1449 struct mbuf *mheader = NULL;
1452 off_t hdtr_size = 0;
1459 * Do argument checking. Must be a regular file in, stream
1460 * type and connected socket out, positive offset.
1462 fp = holdfp(p->p_fd, uap->fd, FREAD);
1466 if (fp->f_type != DTYPE_VNODE) {
1471 vp = (struct vnode *)fp->f_data;
1476 * If specified, get the pointer to the sf_hdtr struct for
1477 * any headers/trailers.
1480 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1487 error = iovec_copyin(hdtr.headers, &iov, aiov,
1488 hdtr.hdr_cnt, &hbytes);
1492 auio.uio_iovcnt = hdtr.hdr_cnt;
1493 auio.uio_offset = 0;
1494 auio.uio_segflg = UIO_USERSPACE;
1495 auio.uio_rw = UIO_WRITE;
1497 auio.uio_resid = hbytes;
1499 mheader = m_uiomove(&auio);
1501 iovec_free(&iov, aiov);
1502 if (mheader == NULL)
1507 error = kern_sendfile(vp, uap->s, uap->offset, uap->nbytes, mheader,
1508 &sbytes, uap->flags);
1513 * Send trailers. Wimp out and use writev(2).
1515 if (uap->hdtr != NULL && hdtr.trailers != NULL) {
1516 error = iovec_copyin(hdtr.trailers, &iov, aiov,
1517 hdtr.trl_cnt, &auio.uio_resid);
1521 auio.uio_iovcnt = hdtr.trl_cnt;
1522 auio.uio_offset = 0;
1523 auio.uio_segflg = UIO_USERSPACE;
1524 auio.uio_rw = UIO_WRITE;
1527 error = kern_sendmsg(uap->s, NULL, &auio, NULL, 0, &tbytes);
1529 iovec_free(&iov, aiov);
1532 hdtr_size += tbytes; /* trailer bytes successfully sent */
1539 if (uap->sbytes != NULL) {
1540 sbytes += hdtr_size;
1541 copyout(&sbytes, uap->sbytes, sizeof(off_t));
1547 kern_sendfile(struct vnode *vp, int sfd, off_t offset, size_t nbytes,
1548 struct mbuf *mheader, off_t *sbytes, int flags)
1550 struct thread *td = curthread;
1551 struct proc *p = td->td_proc;
1552 struct vm_object *obj;
1562 if (vp->v_type != VREG) {
1566 if ((obj = vp->v_object) == NULL) {
1570 error = holdsock(p->p_fd, sfd, &fp);
1573 so = (struct socket *)fp->f_data;
1574 if (so->so_type != SOCK_STREAM) {
1578 if ((so->so_state & SS_ISCONNECTED) == 0) {
1589 * Protect against multiple writers to the socket.
1591 ssb_lock(&so->so_snd, M_WAITOK);
1594 * Loop through the pages in the file, starting with the requested
1595 * offset. Get a file page (do I/O if necessary), map the file page
1596 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1599 for (off = offset; ; off += xfsize, *sbytes += xfsize + hbytes) {
1603 pindex = OFF_TO_IDX(off);
1606 * Calculate the amount to transfer. Not to exceed a page,
1607 * the EOF, or the passed in nbytes.
1609 xfsize = vp->v_filesize - off;
1610 if (xfsize > PAGE_SIZE)
1612 pgoff = (vm_offset_t)(off & PAGE_MASK);
1613 if (PAGE_SIZE - pgoff < xfsize)
1614 xfsize = PAGE_SIZE - pgoff;
1615 if (nbytes && xfsize > (nbytes - *sbytes))
1616 xfsize = nbytes - *sbytes;
1620 * Optimize the non-blocking case by looking at the socket space
1621 * before going to the extra work of constituting the sf_buf.
1623 if ((fp->f_flag & FNONBLOCK) && ssb_space(&so->so_snd) <= 0) {
1624 if (so->so_state & SS_CANTSENDMORE)
1628 ssb_unlock(&so->so_snd);
1632 * Attempt to look up the page.
1634 * Allocate if not found, wait and loop if busy, then
1635 * wire the page. critical section protection is
1636 * required to maintain the object association (an
1637 * interrupt can free the page) through to the
1638 * vm_page_wire() call.
1641 lwkt_gettoken(&vm_token);
1642 pg = vm_page_lookup(obj, pindex);
1644 pg = vm_page_alloc(obj, pindex, VM_ALLOC_NORMAL);
1647 lwkt_reltoken(&vm_token);
1652 } else if (vm_page_sleep_busy(pg, TRUE, "sfpbsy")) {
1653 lwkt_reltoken(&vm_token);
1658 lwkt_reltoken(&vm_token);
1662 * If page is not valid for what we need, initiate I/O
1665 if (!pg->valid || !vm_page_is_valid(pg, pgoff, xfsize)) {
1671 * Ensure that our page is still around when the I/O
1674 vm_page_io_start(pg);
1677 * Get the page from backing store.
1679 bsize = vp->v_mount->mnt_stat.f_iosize;
1680 auio.uio_iov = &aiov;
1681 auio.uio_iovcnt = 1;
1683 aiov.iov_len = MAXBSIZE;
1684 auio.uio_resid = MAXBSIZE;
1685 auio.uio_offset = trunc_page(off);
1686 auio.uio_segflg = UIO_NOCOPY;
1687 auio.uio_rw = UIO_READ;
1689 vn_lock(vp, LK_SHARED | LK_RETRY);
1690 error = VOP_READ(vp, &auio,
1691 IO_VMIO | ((MAXBSIZE / bsize) << 16),
1694 vm_page_flag_clear(pg, PG_ZERO);
1695 vm_page_io_finish(pg);
1698 vm_page_unwire(pg, 0);
1699 vm_page_try_to_free(pg);
1701 ssb_unlock(&so->so_snd);
1708 * Get a sendfile buf. We usually wait as long as necessary,
1709 * but this wait can be interrupted.
1711 if ((sf = sf_buf_alloc(pg)) == NULL) {
1713 vm_page_unwire(pg, 0);
1714 vm_page_try_to_free(pg);
1716 ssb_unlock(&so->so_snd);
1722 * Get an mbuf header and set it up as having external storage.
1724 MGETHDR(m, MB_WAIT, MT_DATA);
1728 ssb_unlock(&so->so_snd);
1732 m->m_ext.ext_free = sf_buf_mfree;
1733 m->m_ext.ext_ref = sf_buf_ref;
1734 m->m_ext.ext_arg = sf;
1735 m->m_ext.ext_buf = (void *)sf_buf_kva(sf);
1736 m->m_ext.ext_size = PAGE_SIZE;
1737 m->m_data = (char *)sf_buf_kva(sf) + pgoff;
1738 m->m_flags |= M_EXT;
1739 m->m_pkthdr.len = m->m_len = xfsize;
1740 KKASSERT((m->m_flags & (M_EXT_CLUSTER)) == 0);
1742 if (mheader != NULL) {
1743 hbytes = mheader->m_pkthdr.len;
1744 mheader->m_pkthdr.len += m->m_pkthdr.len;
1752 * Add the buffer to the socket buffer chain.
1757 * Make sure that the socket is still able to take more data.
1758 * CANTSENDMORE being true usually means that the connection
1759 * was closed. so_error is true when an error was sensed after
1761 * The state is checked after the page mapping and buffer
1762 * allocation above since those operations may block and make
1763 * any socket checks stale. From this point forward, nothing
1764 * blocks before the pru_send (or more accurately, any blocking
1765 * results in a loop back to here to re-check).
1767 if ((so->so_state & SS_CANTSENDMORE) || so->so_error) {
1768 if (so->so_state & SS_CANTSENDMORE) {
1771 error = so->so_error;
1775 ssb_unlock(&so->so_snd);
1780 * Wait for socket space to become available. We do this just
1781 * after checking the connection state above in order to avoid
1782 * a race condition with ssb_wait().
1784 if (ssb_space(&so->so_snd) < so->so_snd.ssb_lowat) {
1785 if (fp->f_flag & FNONBLOCK) {
1787 ssb_unlock(&so->so_snd);
1792 error = ssb_wait(&so->so_snd);
1794 * An error from ssb_wait usually indicates that we've
1795 * been interrupted by a signal. If we've sent anything
1796 * then return bytes sent, otherwise return the error.
1800 ssb_unlock(&so->so_snd);
1806 error = so_pru_send(so, 0, m, NULL, NULL, td);
1809 ssb_unlock(&so->so_snd);
1813 if (mheader != NULL) {
1814 *sbytes += mheader->m_pkthdr.len;
1815 error = so_pru_send(so, 0, mheader, NULL, NULL, td);
1818 ssb_unlock(&so->so_snd);
1823 if (mheader != NULL)
1832 sys_sctp_peeloff(struct sctp_peeloff_args *uap)
1835 struct thread *td = curthread;
1836 struct filedesc *fdp = td->td_proc->p_fd;
1837 struct file *lfp = NULL;
1838 struct file *nfp = NULL;
1840 struct socket *head, *so;
1843 short fflag; /* type must match fp->f_flag */
1845 assoc_id = uap->name;
1846 error = holdsock(td->td_proc->p_fd, uap->sd, &lfp);
1852 head = (struct socket *)lfp->f_data;
1853 error = sctp_can_peel_off(head, assoc_id);
1859 * At this point we know we do have a assoc to pull
1860 * we proceed to get the fd setup. This may block
1864 fflag = lfp->f_flag;
1865 error = falloc(td->td_lwp, &nfp, &fd);
1868 * Probably ran out of file descriptors. Put the
1869 * unaccepted connection back onto the queue and
1870 * do another wakeup so some other process might
1871 * have a chance at it.
1876 uap->sysmsg_iresult = fd;
1878 so = sctp_get_peeloff(head, assoc_id, &error);
1881 * Either someone else peeled it off OR
1882 * we can't get a socket.
1886 soreference(so); /* reference needed */
1887 soclrstate(so, SS_NOFDREF | SS_COMP); /* when clearing NOFDREF */
1889 if (head->so_sigio != NULL)
1890 fsetown(fgetown(head->so_sigio), &so->so_sigio);
1892 nfp->f_type = DTYPE_SOCKET;
1893 nfp->f_flag = fflag;
1894 nfp->f_ops = &socketops;
1899 * Assign the file pointer to the reserved descriptor, or clear
1900 * the reserved descriptor if an error occured.
1903 fsetfd(fdp, NULL, fd);
1905 fsetfd(fdp, nfp, fd);
1908 * Release explicitly held references before returning.