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 if (!TAILQ_EMPTY(&head->so_comp)) {
229 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
230 msg->nm_so = TAILQ_FIRST(&head->so_comp);
231 TAILQ_REMOVE(&head->so_comp, msg->nm_so, so_list);
234 msg->nm_netmsg.nm_lmsg.ms_error = 0;
237 if (head->so_state & SS_CANTRCVMORE) {
238 msg->nm_netmsg.nm_lmsg.ms_error = ECONNABORTED;
241 if (msg->nm_fflags & FNONBLOCK) {
242 msg->nm_netmsg.nm_lmsg.ms_error = EWOULDBLOCK;
250 * The second argument to kern_accept() is a handle to a struct sockaddr.
251 * This allows kern_accept() to return a pointer to an allocated struct
252 * sockaddr which must be freed later with FREE(). The caller must
253 * initialize *name to NULL.
256 kern_accept(int s, int fflags, struct sockaddr **name, int *namelen, int *res)
258 struct thread *td = curthread;
259 struct filedesc *fdp = td->td_proc->p_fd;
260 struct file *lfp = NULL;
261 struct file *nfp = NULL;
263 struct socket *head, *so;
264 struct netmsg_so_notify msg;
266 u_int fflag; /* type must match fp->f_flag */
270 if (name && namelen && *namelen < 0)
273 error = holdsock(td->td_proc->p_fd, s, &lfp);
277 error = falloc(td->td_lwp, &nfp, &fd);
278 if (error) { /* Probably ran out of file descriptors. */
282 head = (struct socket *)lfp->f_data;
283 if ((head->so_options & SO_ACCEPTCONN) == 0) {
288 if (fflags & O_FBLOCKING)
289 fflags |= lfp->f_flag & ~FNONBLOCK;
290 else if (fflags & O_FNONBLOCKING)
291 fflags |= lfp->f_flag | FNONBLOCK;
293 fflags = lfp->f_flag;
295 /* optimize for uniprocessor case later XXX JH */
296 netmsg_init_abortable(&msg.nm_netmsg, head, &curthread->td_msgport,
297 0, netmsg_so_notify, netmsg_so_notify_doabort);
298 msg.nm_predicate = soaccept_predicate;
299 msg.nm_fflags = fflags;
301 msg.nm_etype = NM_REVENT;
302 error = lwkt_domsg(head->so_port, &msg.nm_netmsg.nm_lmsg, PCATCH);
307 * At this point we have the connection that's ready to be accepted.
313 /* connection has been removed from the listen queue */
314 KNOTE(&head->so_rcv.ssb_sel.si_note, 0);
316 so->so_state &= ~SS_COMP;
318 if (head->so_sigio != NULL)
319 fsetown(fgetown(head->so_sigio), &so->so_sigio);
321 nfp->f_type = DTYPE_SOCKET;
323 nfp->f_ops = &socketops;
325 /* Sync socket nonblocking/async state with file flags */
326 tmp = fflag & FNONBLOCK;
327 fo_ioctl(nfp, FIONBIO, (caddr_t)&tmp, td->td_ucred, NULL);
328 tmp = fflag & FASYNC;
329 fo_ioctl(nfp, FIOASYNC, (caddr_t)&tmp, td->td_ucred, NULL);
332 error = soaccept(so, &sa);
335 * Set the returned name and namelen as applicable. Set the returned
336 * namelen to 0 for older code which might ignore the return value
340 if (sa && name && namelen) {
341 if (*namelen > sa->sa_len)
342 *namelen = sa->sa_len;
352 * If an error occured clear the reserved descriptor, else associate
355 * Note that *res is normally ignored if an error is returned but
356 * a syscall message will still have access to the result code.
359 fsetfd(fdp, NULL, fd);
362 fsetfd(fdp, nfp, fd);
370 * accept(int s, caddr_t name, int *anamelen)
375 sys_accept(struct accept_args *uap)
377 struct sockaddr *sa = NULL;
382 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
387 error = kern_accept(uap->s, 0, &sa, &sa_len,
388 &uap->sysmsg_iresult);
392 error = copyout(sa, uap->name, sa_len);
394 error = copyout(&sa_len, uap->anamelen,
395 sizeof(*uap->anamelen));
401 error = kern_accept(uap->s, 0, NULL, 0,
402 &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));
427 error = kern_accept(uap->s, fflags, &sa, &sa_len,
428 &uap->sysmsg_iresult);
432 error = copyout(sa, uap->name, sa_len);
434 error = copyout(&sa_len, uap->anamelen,
435 sizeof(*uap->anamelen));
441 error = kern_accept(uap->s, fflags, NULL, 0,
442 &uap->sysmsg_iresult);
450 * Returns TRUE if predicate satisfied.
453 soconnected_predicate(struct netmsg *msg0)
455 struct netmsg_so_notify *msg = (struct netmsg_so_notify *)msg0;
456 struct socket *so = msg->nm_so;
458 /* check predicate */
459 if (!(so->so_state & SS_ISCONNECTING) || so->so_error != 0) {
460 msg->nm_netmsg.nm_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.nm_netmsg, so,
503 &curthread->td_msgport,
506 netmsg_so_notify_doabort);
507 msg.nm_predicate = soconnected_predicate;
509 msg.nm_etype = NM_REVENT;
510 error = lwkt_domsg(so->so_port, &msg.nm_netmsg.nm_lmsg, PCATCH);
511 if (error == EINTR || error == ERESTART)
515 error = so->so_error;
520 so->so_state &= ~SS_ISCONNECTING;
521 if (error == ERESTART)
529 * connect_args(int s, caddr_t name, int namelen)
534 sys_connect(struct connect_args *uap)
539 error = getsockaddr(&sa, uap->name, uap->namelen);
543 error = kern_connect(uap->s, 0, sa);
551 * connect_args(int s, int fflags, caddr_t name, int namelen)
556 sys_extconnect(struct extconnect_args *uap)
560 int fflags = uap->flags & O_FMASK;
562 error = getsockaddr(&sa, uap->name, uap->namelen);
566 error = kern_connect(uap->s, fflags, sa);
574 kern_socketpair(int domain, int type, int protocol, int *sv)
576 struct thread *td = curthread;
577 struct filedesc *fdp;
578 struct file *fp1, *fp2;
579 struct socket *so1, *so2;
582 fdp = td->td_proc->p_fd;
583 error = socreate(domain, &so1, type, protocol, td);
586 error = socreate(domain, &so2, type, protocol, td);
589 error = falloc(td->td_lwp, &fp1, &fd1);
594 error = falloc(td->td_lwp, &fp2, &fd2);
599 error = soconnect2(so1, so2);
602 if (type == SOCK_DGRAM) {
604 * Datagram socket connection is asymmetric.
606 error = soconnect2(so2, so1);
610 fp1->f_type = fp2->f_type = DTYPE_SOCKET;
611 fp1->f_flag = fp2->f_flag = FREAD|FWRITE;
612 fp1->f_ops = fp2->f_ops = &socketops;
613 fsetfd(fdp, fp1, fd1);
614 fsetfd(fdp, fp2, fd2);
619 fsetfd(fdp, NULL, fd2);
622 fsetfd(fdp, NULL, fd1);
625 (void)soclose(so2, 0);
627 (void)soclose(so1, 0);
632 * socketpair(int domain, int type, int protocol, int *rsv)
637 sys_socketpair(struct socketpair_args *uap)
642 error = kern_socketpair(uap->domain, uap->type, uap->protocol, sockv);
646 error = copyout(sockv, uap->rsv, sizeof(sockv));
651 kern_sendmsg(int s, struct sockaddr *sa, struct uio *auio,
652 struct mbuf *control, int flags, size_t *res)
654 struct thread *td = curthread;
655 struct lwp *lp = td->td_lwp;
656 struct proc *p = td->td_proc;
662 struct iovec *ktriov = NULL;
666 error = holdsock(p->p_fd, s, &fp);
670 if (KTRPOINT(td, KTR_GENIO)) {
671 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
673 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
674 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
678 len = auio->uio_resid;
679 so = (struct socket *)fp->f_data;
680 if ((flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
681 if (fp->f_flag & FNONBLOCK)
682 flags |= MSG_FNONBLOCKING;
684 error = so_pru_sosend(so, sa, auio, NULL, control, flags, td);
686 if (auio->uio_resid != len && (error == ERESTART ||
687 error == EINTR || error == EWOULDBLOCK))
690 lwpsignal(p, lp, SIGPIPE);
693 if (ktriov != NULL) {
695 ktruio.uio_iov = ktriov;
696 ktruio.uio_resid = len - auio->uio_resid;
697 ktrgenio(lp, s, UIO_WRITE, &ktruio, error);
699 FREE(ktriov, M_TEMP);
703 *res = len - auio->uio_resid;
709 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
714 sys_sendto(struct sendto_args *uap)
716 struct thread *td = curthread;
719 struct sockaddr *sa = NULL;
723 error = getsockaddr(&sa, uap->to, uap->tolen);
727 aiov.iov_base = uap->buf;
728 aiov.iov_len = uap->len;
729 auio.uio_iov = &aiov;
732 auio.uio_resid = uap->len;
733 auio.uio_segflg = UIO_USERSPACE;
734 auio.uio_rw = UIO_WRITE;
738 error = kern_sendmsg(uap->s, sa, &auio, NULL, uap->flags,
739 &uap->sysmsg_szresult);
748 * sendmsg_args(int s, caddr_t msg, int flags)
753 sys_sendmsg(struct sendmsg_args *uap)
755 struct thread *td = curthread;
758 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
759 struct sockaddr *sa = NULL;
760 struct mbuf *control = NULL;
763 error = copyin(uap->msg, (caddr_t)&msg, sizeof(msg));
768 * Conditionally copyin msg.msg_name.
771 error = getsockaddr(&sa, msg.msg_name, msg.msg_namelen);
779 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
784 auio.uio_iovcnt = msg.msg_iovlen;
786 auio.uio_segflg = UIO_USERSPACE;
787 auio.uio_rw = UIO_WRITE;
791 * Conditionally copyin msg.msg_control.
793 if (msg.msg_control) {
794 if (msg.msg_controllen < sizeof(struct cmsghdr) ||
795 msg.msg_controllen > MLEN) {
799 control = m_get(MB_WAIT, MT_CONTROL);
800 if (control == NULL) {
804 control->m_len = msg.msg_controllen;
805 error = copyin(msg.msg_control, mtod(control, caddr_t),
814 error = kern_sendmsg(uap->s, sa, &auio, control, uap->flags,
815 &uap->sysmsg_szresult);
819 iovec_free(&iov, aiov);
827 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
828 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
829 * Don't forget to FREE() and m_free() these if they are returned.
832 kern_recvmsg(int s, struct sockaddr **sa, struct uio *auio,
833 struct mbuf **control, int *flags, size_t *res)
835 struct thread *td = curthread;
836 struct proc *p = td->td_proc;
843 struct iovec *ktriov = NULL;
847 error = holdsock(p->p_fd, s, &fp);
851 if (KTRPOINT(td, KTR_GENIO)) {
852 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
854 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
855 bcopy(auio->uio_iov, ktriov, iovlen);
859 len = auio->uio_resid;
860 so = (struct socket *)fp->f_data;
862 if (flags == NULL || (*flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
863 if (fp->f_flag & FNONBLOCK) {
865 *flags |= MSG_FNONBLOCKING;
867 lflags = MSG_FNONBLOCKING;
873 error = so_pru_soreceive(so, sa, auio, NULL, control, flags);
875 if (auio->uio_resid != len && (error == ERESTART ||
876 error == EINTR || error == EWOULDBLOCK))
880 if (ktriov != NULL) {
882 ktruio.uio_iov = ktriov;
883 ktruio.uio_resid = len - auio->uio_resid;
884 ktrgenio(td->td_lwp, s, UIO_READ, &ktruio, error);
886 FREE(ktriov, M_TEMP);
890 *res = len - auio->uio_resid;
896 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
897 * caddr_t from, int *fromlenaddr)
902 sys_recvfrom(struct recvfrom_args *uap)
904 struct thread *td = curthread;
907 struct sockaddr *sa = NULL;
910 if (uap->from && uap->fromlenaddr) {
911 error = copyin(uap->fromlenaddr, &fromlen, sizeof(fromlen));
919 aiov.iov_base = uap->buf;
920 aiov.iov_len = uap->len;
921 auio.uio_iov = &aiov;
924 auio.uio_resid = uap->len;
925 auio.uio_segflg = UIO_USERSPACE;
926 auio.uio_rw = UIO_READ;
930 error = kern_recvmsg(uap->s, uap->from ? &sa : NULL, &auio, NULL,
931 &uap->flags, &uap->sysmsg_szresult);
934 if (error == 0 && uap->from) {
935 /* note: sa may still be NULL */
937 fromlen = MIN(fromlen, sa->sa_len);
938 error = copyout(sa, uap->from, fromlen);
943 error = copyout(&fromlen, uap->fromlenaddr,
954 * recvmsg_args(int s, struct msghdr *msg, int flags)
959 sys_recvmsg(struct recvmsg_args *uap)
961 struct thread *td = curthread;
964 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
965 struct mbuf *m, *control = NULL;
966 struct sockaddr *sa = NULL;
968 socklen_t *ufromlenp, *ucontrollenp;
969 int error, fromlen, controllen, len, flags, *uflagsp;
972 * This copyin handles everything except the iovec.
974 error = copyin(uap->msg, &msg, sizeof(msg));
978 if (msg.msg_name && msg.msg_namelen < 0)
980 if (msg.msg_control && msg.msg_controllen < 0)
983 ufromlenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
985 ucontrollenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
987 uflagsp = (int *)((caddr_t)uap->msg + offsetof(struct msghdr,
993 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
998 auio.uio_iovcnt = msg.msg_iovlen;
1000 auio.uio_segflg = UIO_USERSPACE;
1001 auio.uio_rw = UIO_READ;
1007 error = kern_recvmsg(uap->s,
1008 (msg.msg_name ? &sa : NULL), &auio,
1009 (msg.msg_control ? &control : NULL), &flags,
1010 &uap->sysmsg_szresult);
1014 * Conditionally copyout the name and populate the namelen field.
1016 if (error == 0 && msg.msg_name) {
1017 /* note: sa may still be NULL */
1019 fromlen = MIN(msg.msg_namelen, sa->sa_len);
1020 error = copyout(sa, msg.msg_name, fromlen);
1025 error = copyout(&fromlen, ufromlenp,
1026 sizeof(*ufromlenp));
1030 * Copyout msg.msg_control and msg.msg_controllen.
1032 if (error == 0 && msg.msg_control) {
1033 len = msg.msg_controllen;
1035 ctlbuf = (caddr_t)msg.msg_control;
1037 while(m && len > 0) {
1038 unsigned int tocopy;
1040 if (len >= m->m_len) {
1043 msg.msg_flags |= MSG_CTRUNC;
1047 error = copyout(mtod(m, caddr_t), ctlbuf, tocopy);
1055 controllen = ctlbuf - (caddr_t)msg.msg_control;
1056 error = copyout(&controllen, ucontrollenp,
1057 sizeof(*ucontrollenp));
1061 error = copyout(&flags, uflagsp, sizeof(*uflagsp));
1066 iovec_free(&iov, aiov);
1073 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1074 * in kernel pointer instead of a userland pointer. This allows us
1075 * to manipulate socket options in the emulation code.
1078 kern_setsockopt(int s, struct sockopt *sopt)
1080 struct thread *td = curthread;
1081 struct proc *p = td->td_proc;
1085 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1087 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1089 if (sopt->sopt_valsize < 0)
1092 error = holdsock(p->p_fd, s, &fp);
1096 error = sosetopt((struct socket *)fp->f_data, sopt);
1102 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
1107 sys_setsockopt(struct setsockopt_args *uap)
1109 struct thread *td = curthread;
1110 struct sockopt sopt;
1113 sopt.sopt_level = uap->level;
1114 sopt.sopt_name = uap->name;
1115 sopt.sopt_valsize = uap->valsize;
1117 sopt.sopt_val = NULL;
1119 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1122 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1123 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1129 error = kern_setsockopt(uap->s, &sopt);
1133 kfree(sopt.sopt_val, M_TEMP);
1138 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1139 * in kernel pointer instead of a userland pointer. This allows us
1140 * to manipulate socket options in the emulation code.
1143 kern_getsockopt(int s, struct sockopt *sopt)
1145 struct thread *td = curthread;
1146 struct proc *p = td->td_proc;
1150 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1152 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1154 if (sopt->sopt_valsize < 0 || sopt->sopt_valsize > SOMAXOPT_SIZE)
1157 error = holdsock(p->p_fd, s, &fp);
1161 error = sogetopt((struct socket *)fp->f_data, sopt);
1167 * getsockopt_args(int s, int level, int name, caddr_t val, int *avalsize)
1172 sys_getsockopt(struct getsockopt_args *uap)
1174 struct thread *td = curthread;
1175 struct sockopt sopt;
1179 error = copyin(uap->avalsize, &valsize, sizeof(valsize));
1186 sopt.sopt_level = uap->level;
1187 sopt.sopt_name = uap->name;
1188 sopt.sopt_valsize = valsize;
1190 sopt.sopt_val = NULL;
1192 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1195 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1196 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1202 error = kern_getsockopt(uap->s, &sopt);
1206 valsize = sopt.sopt_valsize;
1207 error = copyout(&valsize, uap->avalsize, sizeof(valsize));
1211 error = copyout(sopt.sopt_val, uap->val, sopt.sopt_valsize);
1214 kfree(sopt.sopt_val, M_TEMP);
1219 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1220 * This allows kern_getsockname() to return a pointer to an allocated struct
1221 * sockaddr which must be freed later with FREE(). The caller must
1222 * initialize *name to NULL.
1225 kern_getsockname(int s, struct sockaddr **name, int *namelen)
1227 struct thread *td = curthread;
1228 struct proc *p = td->td_proc;
1231 struct sockaddr *sa = NULL;
1234 error = holdsock(p->p_fd, s, &fp);
1241 so = (struct socket *)fp->f_data;
1242 error = so_pru_sockaddr(so, &sa);
1247 *namelen = MIN(*namelen, sa->sa_len);
1257 * getsockname_args(int fdes, caddr_t asa, int *alen)
1264 sys_getsockname(struct getsockname_args *uap)
1266 struct sockaddr *sa = NULL;
1269 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1274 error = kern_getsockname(uap->fdes, &sa, &sa_len);
1278 error = copyout(sa, uap->asa, sa_len);
1280 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1287 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1288 * This allows kern_getpeername() to return a pointer to an allocated struct
1289 * sockaddr which must be freed later with FREE(). The caller must
1290 * initialize *name to NULL.
1293 kern_getpeername(int s, struct sockaddr **name, int *namelen)
1295 struct thread *td = curthread;
1296 struct proc *p = td->td_proc;
1299 struct sockaddr *sa = NULL;
1302 error = holdsock(p->p_fd, s, &fp);
1309 so = (struct socket *)fp->f_data;
1310 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1314 error = so_pru_peeraddr(so, &sa);
1319 *namelen = MIN(*namelen, sa->sa_len);
1329 * getpeername_args(int fdes, caddr_t asa, int *alen)
1331 * Get name of peer for connected socket.
1336 sys_getpeername(struct getpeername_args *uap)
1338 struct sockaddr *sa = NULL;
1341 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1346 error = kern_getpeername(uap->fdes, &sa, &sa_len);
1350 error = copyout(sa, uap->asa, sa_len);
1352 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1359 getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len)
1361 struct sockaddr *sa;
1365 if (len > SOCK_MAXADDRLEN)
1366 return ENAMETOOLONG;
1367 if (len < offsetof(struct sockaddr, sa_data[0]))
1369 MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1370 error = copyin(uaddr, sa, len);
1374 #if BYTE_ORDER != BIG_ENDIAN
1376 * The bind(), connect(), and sendto() syscalls were not
1377 * versioned for COMPAT_43. Thus, this check must stay.
1379 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1380 sa->sa_family = sa->sa_len;
1389 * Detach a mapped page and release resources back to the system.
1390 * We must release our wiring and if the object is ripped out
1391 * from under the vm_page we become responsible for freeing the
1392 * page. These routines must be MPSAFE.
1394 * XXX HACK XXX TEMPORARY UNTIL WE IMPLEMENT EXT MBUF REFERENCE COUNTING
1396 * XXX vm_page_*() routines are not MPSAFE yet, the MP lock is required.
1399 sf_buf_mfree(void *arg)
1401 struct sf_buf *sf = arg;
1405 * XXX vm_page_*() and SFBUF routines not MPSAFE yet.
1409 m = sf_buf_page(sf);
1410 if (sf_buf_free(sf) == 0) {
1411 vm_page_unwire(m, 0);
1412 if (m->wire_count == 0 && m->object == NULL)
1413 vm_page_try_to_free(m);
1421 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1422 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1424 * Send a file specified by 'fd' and starting at 'offset' to a socket
1425 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1426 * nbytes == 0. Optionally add a header and/or trailer to the socket
1427 * output. If specified, write the total number of bytes sent into *sbytes.
1429 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1430 * the headers to count against the remaining bytes to be sent from
1431 * the file descriptor. We may wish to implement a compatibility syscall
1437 sys_sendfile(struct sendfile_args *uap)
1439 struct thread *td = curthread;
1440 struct proc *p = td->td_proc;
1442 struct vnode *vp = NULL;
1443 struct sf_hdtr hdtr;
1444 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
1446 struct mbuf *mheader = NULL;
1449 off_t hdtr_size = 0;
1456 * Do argument checking. Must be a regular file in, stream
1457 * type and connected socket out, positive offset.
1459 fp = holdfp(p->p_fd, uap->fd, FREAD);
1463 if (fp->f_type != DTYPE_VNODE) {
1468 vp = (struct vnode *)fp->f_data;
1473 * If specified, get the pointer to the sf_hdtr struct for
1474 * any headers/trailers.
1477 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1484 error = iovec_copyin(hdtr.headers, &iov, aiov,
1485 hdtr.hdr_cnt, &hbytes);
1489 auio.uio_iovcnt = hdtr.hdr_cnt;
1490 auio.uio_offset = 0;
1491 auio.uio_segflg = UIO_USERSPACE;
1492 auio.uio_rw = UIO_WRITE;
1494 auio.uio_resid = hbytes;
1496 mheader = m_uiomove(&auio);
1498 iovec_free(&iov, aiov);
1499 if (mheader == NULL)
1504 error = kern_sendfile(vp, uap->s, uap->offset, uap->nbytes, mheader,
1505 &sbytes, uap->flags);
1510 * Send trailers. Wimp out and use writev(2).
1512 if (uap->hdtr != NULL && hdtr.trailers != NULL) {
1513 error = iovec_copyin(hdtr.trailers, &iov, aiov,
1514 hdtr.trl_cnt, &auio.uio_resid);
1518 auio.uio_iovcnt = hdtr.trl_cnt;
1519 auio.uio_offset = 0;
1520 auio.uio_segflg = UIO_USERSPACE;
1521 auio.uio_rw = UIO_WRITE;
1524 error = kern_sendmsg(uap->s, NULL, &auio, NULL, 0, &tbytes);
1526 iovec_free(&iov, aiov);
1529 hdtr_size += tbytes; /* trailer bytes successfully sent */
1536 if (uap->sbytes != NULL) {
1537 sbytes += hdtr_size;
1538 copyout(&sbytes, uap->sbytes, sizeof(off_t));
1544 kern_sendfile(struct vnode *vp, int sfd, off_t offset, size_t nbytes,
1545 struct mbuf *mheader, off_t *sbytes, int flags)
1547 struct thread *td = curthread;
1548 struct proc *p = td->td_proc;
1549 struct vm_object *obj;
1559 if (vp->v_type != VREG) {
1563 if ((obj = vp->v_object) == NULL) {
1567 error = holdsock(p->p_fd, sfd, &fp);
1570 so = (struct socket *)fp->f_data;
1571 if (so->so_type != SOCK_STREAM) {
1575 if ((so->so_state & SS_ISCONNECTED) == 0) {
1586 * Protect against multiple writers to the socket.
1588 ssb_lock(&so->so_snd, M_WAITOK);
1591 * Loop through the pages in the file, starting with the requested
1592 * offset. Get a file page (do I/O if necessary), map the file page
1593 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1596 for (off = offset; ; off += xfsize, *sbytes += xfsize + hbytes) {
1600 pindex = OFF_TO_IDX(off);
1603 * Calculate the amount to transfer. Not to exceed a page,
1604 * the EOF, or the passed in nbytes.
1606 xfsize = vp->v_filesize - off;
1607 if (xfsize > PAGE_SIZE)
1609 pgoff = (vm_offset_t)(off & PAGE_MASK);
1610 if (PAGE_SIZE - pgoff < xfsize)
1611 xfsize = PAGE_SIZE - pgoff;
1612 if (nbytes && xfsize > (nbytes - *sbytes))
1613 xfsize = nbytes - *sbytes;
1617 * Optimize the non-blocking case by looking at the socket space
1618 * before going to the extra work of constituting the sf_buf.
1620 if ((fp->f_flag & FNONBLOCK) && ssb_space(&so->so_snd) <= 0) {
1621 if (so->so_state & SS_CANTSENDMORE)
1625 ssb_unlock(&so->so_snd);
1629 * Attempt to look up the page.
1631 * Allocate if not found, wait and loop if busy, then
1632 * wire the page. critical section protection is
1633 * required to maintain the object association (an
1634 * interrupt can free the page) through to the
1635 * vm_page_wire() call.
1638 pg = vm_page_lookup(obj, pindex);
1640 pg = vm_page_alloc(obj, pindex, VM_ALLOC_NORMAL);
1647 } else if (vm_page_sleep_busy(pg, TRUE, "sfpbsy")) {
1655 * If page is not valid for what we need, initiate I/O
1658 if (!pg->valid || !vm_page_is_valid(pg, pgoff, xfsize)) {
1664 * Ensure that our page is still around when the I/O
1667 vm_page_io_start(pg);
1670 * Get the page from backing store.
1672 bsize = vp->v_mount->mnt_stat.f_iosize;
1673 auio.uio_iov = &aiov;
1674 auio.uio_iovcnt = 1;
1676 aiov.iov_len = MAXBSIZE;
1677 auio.uio_resid = MAXBSIZE;
1678 auio.uio_offset = trunc_page(off);
1679 auio.uio_segflg = UIO_NOCOPY;
1680 auio.uio_rw = UIO_READ;
1682 vn_lock(vp, LK_SHARED | LK_RETRY);
1683 error = VOP_READ(vp, &auio,
1684 IO_VMIO | ((MAXBSIZE / bsize) << 16),
1687 vm_page_flag_clear(pg, PG_ZERO);
1688 vm_page_io_finish(pg);
1691 vm_page_unwire(pg, 0);
1692 vm_page_try_to_free(pg);
1694 ssb_unlock(&so->so_snd);
1701 * Get a sendfile buf. We usually wait as long as necessary,
1702 * but this wait can be interrupted.
1704 if ((sf = sf_buf_alloc(pg)) == NULL) {
1706 vm_page_unwire(pg, 0);
1707 vm_page_try_to_free(pg);
1709 ssb_unlock(&so->so_snd);
1715 * Get an mbuf header and set it up as having external storage.
1717 MGETHDR(m, MB_WAIT, MT_DATA);
1721 ssb_unlock(&so->so_snd);
1725 m->m_ext.ext_free = sf_buf_mfree;
1726 m->m_ext.ext_ref = sf_buf_ref;
1727 m->m_ext.ext_arg = sf;
1728 m->m_ext.ext_buf = (void *)sf_buf_kva(sf);
1729 m->m_ext.ext_size = PAGE_SIZE;
1730 m->m_data = (char *)sf_buf_kva(sf) + pgoff;
1731 m->m_flags |= M_EXT;
1732 m->m_pkthdr.len = m->m_len = xfsize;
1733 KKASSERT((m->m_flags & (M_EXT_CLUSTER)) == 0);
1735 if (mheader != NULL) {
1736 hbytes = mheader->m_pkthdr.len;
1737 mheader->m_pkthdr.len += m->m_pkthdr.len;
1745 * Add the buffer to the socket buffer chain.
1750 * Make sure that the socket is still able to take more data.
1751 * CANTSENDMORE being true usually means that the connection
1752 * was closed. so_error is true when an error was sensed after
1754 * The state is checked after the page mapping and buffer
1755 * allocation above since those operations may block and make
1756 * any socket checks stale. From this point forward, nothing
1757 * blocks before the pru_send (or more accurately, any blocking
1758 * results in a loop back to here to re-check).
1760 if ((so->so_state & SS_CANTSENDMORE) || so->so_error) {
1761 if (so->so_state & SS_CANTSENDMORE) {
1764 error = so->so_error;
1768 ssb_unlock(&so->so_snd);
1773 * Wait for socket space to become available. We do this just
1774 * after checking the connection state above in order to avoid
1775 * a race condition with ssb_wait().
1777 if (ssb_space(&so->so_snd) < so->so_snd.ssb_lowat) {
1778 if (fp->f_flag & FNONBLOCK) {
1780 ssb_unlock(&so->so_snd);
1785 error = ssb_wait(&so->so_snd);
1787 * An error from ssb_wait usually indicates that we've
1788 * been interrupted by a signal. If we've sent anything
1789 * then return bytes sent, otherwise return the error.
1793 ssb_unlock(&so->so_snd);
1799 error = so_pru_send(so, 0, m, NULL, NULL, td);
1802 ssb_unlock(&so->so_snd);
1806 if (mheader != NULL) {
1807 *sbytes += mheader->m_pkthdr.len;
1808 error = so_pru_send(so, 0, mheader, NULL, NULL, td);
1811 ssb_unlock(&so->so_snd);
1816 if (mheader != NULL)
1825 sys_sctp_peeloff(struct sctp_peeloff_args *uap)
1828 struct thread *td = curthread;
1829 struct filedesc *fdp = td->td_proc->p_fd;
1830 struct file *lfp = NULL;
1831 struct file *nfp = NULL;
1833 struct socket *head, *so;
1836 short fflag; /* type must match fp->f_flag */
1838 assoc_id = uap->name;
1839 error = holdsock(td->td_proc->p_fd, uap->sd, &lfp);
1845 head = (struct socket *)lfp->f_data;
1846 error = sctp_can_peel_off(head, assoc_id);
1852 * At this point we know we do have a assoc to pull
1853 * we proceed to get the fd setup. This may block
1857 fflag = lfp->f_flag;
1858 error = falloc(td->td_lwp, &nfp, &fd);
1861 * Probably ran out of file descriptors. Put the
1862 * unaccepted connection back onto the queue and
1863 * do another wakeup so some other process might
1864 * have a chance at it.
1869 uap->sysmsg_iresult = fd;
1871 so = sctp_get_peeloff(head, assoc_id, &error);
1874 * Either someone else peeled it off OR
1875 * we can't get a socket.
1879 so->so_state &= ~SS_COMP;
1880 so->so_state &= ~SS_NOFDREF;
1882 if (head->so_sigio != NULL)
1883 fsetown(fgetown(head->so_sigio), &so->so_sigio);
1885 nfp->f_type = DTYPE_SOCKET;
1886 nfp->f_flag = fflag;
1887 nfp->f_ops = &socketops;
1892 * Assign the file pointer to the reserved descriptor, or clear
1893 * the reserved descriptor if an error occured.
1896 fsetfd(fdp, NULL, fd);
1898 fsetfd(fdp, nfp, fd);
1901 * Release explicitly held references before returning.