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>
87 extern int use_soaccept_pred_fast;
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)
135 error = kern_socket(uap->domain, uap->type, uap->protocol,
136 &uap->sysmsg_iresult);
142 kern_bind(int s, struct sockaddr *sa)
144 struct thread *td = curthread;
145 struct proc *p = td->td_proc;
150 error = holdsock(p->p_fd, s, &fp);
153 error = sobind((struct socket *)fp->f_data, sa, td);
159 * bind_args(int s, caddr_t name, int namelen)
164 sys_bind(struct bind_args *uap)
169 error = getsockaddr(&sa, uap->name, uap->namelen);
172 error = kern_bind(uap->s, sa);
179 kern_listen(int s, int backlog)
181 struct thread *td = curthread;
182 struct proc *p = td->td_proc;
187 error = holdsock(p->p_fd, s, &fp);
190 error = solisten((struct socket *)fp->f_data, backlog, td);
196 * listen_args(int s, int backlog)
201 sys_listen(struct listen_args *uap)
205 error = kern_listen(uap->s, uap->backlog);
210 * Returns the accepted socket as well.
212 * NOTE! The sockets sitting on so_comp/so_incomp might have 0 refs, the
213 * pool token is absolutely required to avoid a sofree() race,
214 * as well as to avoid tailq handling races.
217 soaccept_predicate(struct netmsg_so_notify *msg)
219 struct socket *head = msg->base.nm_so;
222 if (head->so_error != 0) {
223 msg->base.lmsg.ms_error = head->so_error;
226 lwkt_getpooltoken(head);
227 if (!TAILQ_EMPTY(&head->so_comp)) {
228 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
229 so = TAILQ_FIRST(&head->so_comp);
230 TAILQ_REMOVE(&head->so_comp, so, so_list);
232 soclrstate(so, SS_COMP);
236 lwkt_relpooltoken(head);
238 msg->base.lmsg.ms_error = 0;
239 msg->base.nm_so = so;
242 lwkt_relpooltoken(head);
243 if (head->so_state & SS_CANTRCVMORE) {
244 msg->base.lmsg.ms_error = ECONNABORTED;
247 if (msg->nm_fflags & FNONBLOCK) {
248 msg->base.lmsg.ms_error = EWOULDBLOCK;
256 * The second argument to kern_accept() is a handle to a struct sockaddr.
257 * This allows kern_accept() to return a pointer to an allocated struct
258 * sockaddr which must be freed later with FREE(). The caller must
259 * initialize *name to NULL.
262 kern_accept(int s, int fflags, struct sockaddr **name, int *namelen, int *res)
264 struct thread *td = curthread;
265 struct filedesc *fdp = td->td_proc->p_fd;
266 struct file *lfp = NULL;
267 struct file *nfp = NULL;
269 struct socket *head, *so;
270 struct netmsg_so_notify msg;
272 u_int fflag; /* type must match fp->f_flag */
276 if (name && namelen && *namelen < 0)
279 error = holdsock(td->td_proc->p_fd, s, &lfp);
283 error = falloc(td->td_lwp, &nfp, &fd);
284 if (error) { /* Probably ran out of file descriptors. */
288 head = (struct socket *)lfp->f_data;
289 if ((head->so_options & SO_ACCEPTCONN) == 0) {
294 if (fflags & O_FBLOCKING)
295 fflags |= lfp->f_flag & ~FNONBLOCK;
296 else if (fflags & O_FNONBLOCKING)
297 fflags |= lfp->f_flag | FNONBLOCK;
299 fflags = lfp->f_flag;
301 if (use_soaccept_pred_fast) {
304 /* Initialize necessary parts for soaccept_predicate() */
305 netmsg_init(&msg.base, head, &netisr_apanic_rport, 0, NULL);
306 msg.nm_fflags = fflags;
308 lwkt_getpooltoken(head);
309 pred = soaccept_predicate(&msg);
310 lwkt_relpooltoken(head);
313 error = msg.base.lmsg.ms_error;
321 /* optimize for uniprocessor case later XXX JH */
322 netmsg_init_abortable(&msg.base, head, &curthread->td_msgport,
323 0, netmsg_so_notify, netmsg_so_notify_doabort);
324 msg.nm_predicate = soaccept_predicate;
325 msg.nm_fflags = fflags;
326 msg.nm_etype = NM_REVENT;
327 error = lwkt_domsg(head->so_port, &msg.base.lmsg, PCATCH);
333 * At this point we have the connection that's ready to be accepted.
335 * NOTE! soaccept_predicate() ref'd so for us, and soaccept() expects
336 * to eat the ref and turn it into a descriptor.
342 /* connection has been removed from the listen queue */
343 KNOTE(&head->so_rcv.ssb_kq.ki_note, 0);
345 if (head->so_sigio != NULL)
346 fsetown(fgetown(&head->so_sigio), &so->so_sigio);
348 nfp->f_type = DTYPE_SOCKET;
350 nfp->f_ops = &socketops;
352 /* Sync socket nonblocking/async state with file flags */
353 tmp = fflag & FNONBLOCK;
354 fo_ioctl(nfp, FIONBIO, (caddr_t)&tmp, td->td_ucred, NULL);
355 tmp = fflag & FASYNC;
356 fo_ioctl(nfp, FIOASYNC, (caddr_t)&tmp, td->td_ucred, NULL);
359 error = soaccept(so, &sa);
362 * Set the returned name and namelen as applicable. Set the returned
363 * namelen to 0 for older code which might ignore the return value
367 if (sa && name && namelen) {
368 if (*namelen > sa->sa_len)
369 *namelen = sa->sa_len;
379 * If an error occured clear the reserved descriptor, else associate
382 * Note that *res is normally ignored if an error is returned but
383 * a syscall message will still have access to the result code.
386 fsetfd(fdp, NULL, fd);
389 fsetfd(fdp, nfp, fd);
397 * accept(int s, caddr_t name, int *anamelen)
402 sys_accept(struct accept_args *uap)
404 struct sockaddr *sa = NULL;
409 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
413 error = kern_accept(uap->s, 0, &sa, &sa_len,
414 &uap->sysmsg_iresult);
417 error = copyout(sa, uap->name, sa_len);
419 error = copyout(&sa_len, uap->anamelen,
420 sizeof(*uap->anamelen));
425 error = kern_accept(uap->s, 0, NULL, 0,
426 &uap->sysmsg_iresult);
432 * extaccept(int s, int fflags, caddr_t name, int *anamelen)
437 sys_extaccept(struct extaccept_args *uap)
439 struct sockaddr *sa = NULL;
442 int fflags = uap->flags & O_FMASK;
445 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
449 error = kern_accept(uap->s, fflags, &sa, &sa_len,
450 &uap->sysmsg_iresult);
453 error = copyout(sa, uap->name, sa_len);
455 error = copyout(&sa_len, uap->anamelen,
456 sizeof(*uap->anamelen));
461 error = kern_accept(uap->s, fflags, NULL, 0,
462 &uap->sysmsg_iresult);
469 * Returns TRUE if predicate satisfied.
472 soconnected_predicate(struct netmsg_so_notify *msg)
474 struct socket *so = msg->base.nm_so;
476 /* check predicate */
477 if (!(so->so_state & SS_ISCONNECTING) || so->so_error != 0) {
478 msg->base.lmsg.ms_error = so->so_error;
486 kern_connect(int s, int fflags, struct sockaddr *sa)
488 struct thread *td = curthread;
489 struct proc *p = td->td_proc;
492 int error, interrupted = 0;
494 error = holdsock(p->p_fd, s, &fp);
497 so = (struct socket *)fp->f_data;
499 if (fflags & O_FBLOCKING)
500 /* fflags &= ~FNONBLOCK; */;
501 else if (fflags & O_FNONBLOCKING)
506 if (so->so_state & SS_ISCONNECTING) {
510 error = soconnect(so, sa, td);
513 if ((fflags & FNONBLOCK) && (so->so_state & SS_ISCONNECTING)) {
517 if ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
518 struct netmsg_so_notify msg;
520 netmsg_init_abortable(&msg.base, so,
521 &curthread->td_msgport,
524 netmsg_so_notify_doabort);
525 msg.nm_predicate = soconnected_predicate;
526 msg.nm_etype = NM_REVENT;
527 error = lwkt_domsg(so->so_port, &msg.base.lmsg, PCATCH);
528 if (error == EINTR || error == ERESTART)
532 error = so->so_error;
537 soclrstate(so, SS_ISCONNECTING);
538 if (error == ERESTART)
546 * connect_args(int s, caddr_t name, int namelen)
551 sys_connect(struct connect_args *uap)
556 error = getsockaddr(&sa, uap->name, uap->namelen);
559 error = kern_connect(uap->s, 0, sa);
566 * connect_args(int s, int fflags, caddr_t name, int namelen)
571 sys_extconnect(struct extconnect_args *uap)
575 int fflags = uap->flags & O_FMASK;
577 error = getsockaddr(&sa, uap->name, uap->namelen);
580 error = kern_connect(uap->s, fflags, sa);
587 kern_socketpair(int domain, int type, int protocol, int *sv)
589 struct thread *td = curthread;
590 struct filedesc *fdp;
591 struct file *fp1, *fp2;
592 struct socket *so1, *so2;
595 fdp = td->td_proc->p_fd;
596 error = socreate(domain, &so1, type, protocol, td);
599 error = socreate(domain, &so2, type, protocol, td);
602 error = falloc(td->td_lwp, &fp1, &fd1);
607 error = falloc(td->td_lwp, &fp2, &fd2);
612 error = soconnect2(so1, so2);
615 if (type == SOCK_DGRAM) {
617 * Datagram socket connection is asymmetric.
619 error = soconnect2(so2, so1);
623 fp1->f_type = fp2->f_type = DTYPE_SOCKET;
624 fp1->f_flag = fp2->f_flag = FREAD|FWRITE;
625 fp1->f_ops = fp2->f_ops = &socketops;
626 fsetfd(fdp, fp1, fd1);
627 fsetfd(fdp, fp2, fd2);
632 fsetfd(fdp, NULL, fd2);
635 fsetfd(fdp, NULL, fd1);
638 (void)soclose(so2, 0);
640 (void)soclose(so1, 0);
645 * socketpair(int domain, int type, int protocol, int *rsv)
648 sys_socketpair(struct socketpair_args *uap)
652 error = kern_socketpair(uap->domain, uap->type, uap->protocol, sockv);
655 error = copyout(sockv, uap->rsv, sizeof(sockv));
658 kern_close(sockv[0]);
659 kern_close(sockv[1]);
667 kern_sendmsg(int s, struct sockaddr *sa, struct uio *auio,
668 struct mbuf *control, int flags, size_t *res)
670 struct thread *td = curthread;
671 struct lwp *lp = td->td_lwp;
672 struct proc *p = td->td_proc;
678 struct iovec *ktriov = NULL;
682 error = holdsock(p->p_fd, s, &fp);
686 if (KTRPOINT(td, KTR_GENIO)) {
687 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
689 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
690 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
694 len = auio->uio_resid;
695 so = (struct socket *)fp->f_data;
696 if ((flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
697 if (fp->f_flag & FNONBLOCK)
698 flags |= MSG_FNONBLOCKING;
700 error = so_pru_sosend(so, sa, auio, NULL, control, flags, td);
702 if (auio->uio_resid != len && (error == ERESTART ||
703 error == EINTR || error == EWOULDBLOCK))
705 if (error == EPIPE && !(flags & MSG_NOSIGNAL))
706 lwpsignal(p, lp, SIGPIPE);
709 if (ktriov != NULL) {
711 ktruio.uio_iov = ktriov;
712 ktruio.uio_resid = len - auio->uio_resid;
713 ktrgenio(lp, s, UIO_WRITE, &ktruio, error);
715 FREE(ktriov, M_TEMP);
719 *res = len - auio->uio_resid;
725 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
730 sys_sendto(struct sendto_args *uap)
732 struct thread *td = curthread;
735 struct sockaddr *sa = NULL;
739 error = getsockaddr(&sa, uap->to, uap->tolen);
743 aiov.iov_base = uap->buf;
744 aiov.iov_len = uap->len;
745 auio.uio_iov = &aiov;
748 auio.uio_resid = uap->len;
749 auio.uio_segflg = UIO_USERSPACE;
750 auio.uio_rw = UIO_WRITE;
753 error = kern_sendmsg(uap->s, sa, &auio, NULL, uap->flags,
754 &uap->sysmsg_szresult);
762 * sendmsg_args(int s, caddr_t msg, int flags)
767 sys_sendmsg(struct sendmsg_args *uap)
769 struct thread *td = curthread;
772 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
773 struct sockaddr *sa = NULL;
774 struct mbuf *control = NULL;
777 error = copyin(uap->msg, (caddr_t)&msg, sizeof(msg));
782 * Conditionally copyin msg.msg_name.
785 error = getsockaddr(&sa, msg.msg_name, msg.msg_namelen);
793 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
798 auio.uio_iovcnt = msg.msg_iovlen;
800 auio.uio_segflg = UIO_USERSPACE;
801 auio.uio_rw = UIO_WRITE;
805 * Conditionally copyin msg.msg_control.
807 if (msg.msg_control) {
808 if (msg.msg_controllen < sizeof(struct cmsghdr) ||
809 msg.msg_controllen > MLEN) {
813 control = m_get(MB_WAIT, MT_CONTROL);
814 if (control == NULL) {
818 control->m_len = msg.msg_controllen;
819 error = copyin(msg.msg_control, mtod(control, caddr_t),
827 error = kern_sendmsg(uap->s, sa, &auio, control, uap->flags,
828 &uap->sysmsg_szresult);
831 iovec_free(&iov, aiov);
839 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
840 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
841 * Don't forget to FREE() and m_free() these if they are returned.
844 kern_recvmsg(int s, struct sockaddr **sa, struct uio *auio,
845 struct mbuf **control, int *flags, size_t *res)
847 struct thread *td = curthread;
848 struct proc *p = td->td_proc;
855 struct iovec *ktriov = NULL;
859 error = holdsock(p->p_fd, s, &fp);
863 if (KTRPOINT(td, KTR_GENIO)) {
864 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
866 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
867 bcopy(auio->uio_iov, ktriov, iovlen);
871 len = auio->uio_resid;
872 so = (struct socket *)fp->f_data;
874 if (flags == NULL || (*flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
875 if (fp->f_flag & FNONBLOCK) {
877 *flags |= MSG_FNONBLOCKING;
879 lflags = MSG_FNONBLOCKING;
885 error = so_pru_soreceive(so, sa, auio, NULL, control, flags);
887 if (auio->uio_resid != len && (error == ERESTART ||
888 error == EINTR || error == EWOULDBLOCK))
892 if (ktriov != NULL) {
894 ktruio.uio_iov = ktriov;
895 ktruio.uio_resid = len - auio->uio_resid;
896 ktrgenio(td->td_lwp, s, UIO_READ, &ktruio, error);
898 FREE(ktriov, M_TEMP);
902 *res = len - auio->uio_resid;
908 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
909 * caddr_t from, int *fromlenaddr)
914 sys_recvfrom(struct recvfrom_args *uap)
916 struct thread *td = curthread;
919 struct sockaddr *sa = NULL;
922 if (uap->from && uap->fromlenaddr) {
923 error = copyin(uap->fromlenaddr, &fromlen, sizeof(fromlen));
931 aiov.iov_base = uap->buf;
932 aiov.iov_len = uap->len;
933 auio.uio_iov = &aiov;
936 auio.uio_resid = uap->len;
937 auio.uio_segflg = UIO_USERSPACE;
938 auio.uio_rw = UIO_READ;
941 error = kern_recvmsg(uap->s, uap->from ? &sa : NULL, &auio, NULL,
942 &uap->flags, &uap->sysmsg_szresult);
944 if (error == 0 && uap->from) {
945 /* note: sa may still be NULL */
947 fromlen = MIN(fromlen, sa->sa_len);
948 error = copyout(sa, uap->from, fromlen);
953 error = copyout(&fromlen, uap->fromlenaddr,
964 * recvmsg_args(int s, struct msghdr *msg, int flags)
969 sys_recvmsg(struct recvmsg_args *uap)
971 struct thread *td = curthread;
974 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
975 struct mbuf *m, *control = NULL;
976 struct sockaddr *sa = NULL;
978 socklen_t *ufromlenp, *ucontrollenp;
979 int error, fromlen, controllen, len, flags, *uflagsp;
982 * This copyin handles everything except the iovec.
984 error = copyin(uap->msg, &msg, sizeof(msg));
988 if (msg.msg_name && msg.msg_namelen < 0)
990 if (msg.msg_control && msg.msg_controllen < 0)
993 ufromlenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
995 ucontrollenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
997 uflagsp = (int *)((caddr_t)uap->msg + offsetof(struct msghdr,
1003 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
1008 auio.uio_iovcnt = msg.msg_iovlen;
1009 auio.uio_offset = 0;
1010 auio.uio_segflg = UIO_USERSPACE;
1011 auio.uio_rw = UIO_READ;
1016 error = kern_recvmsg(uap->s,
1017 (msg.msg_name ? &sa : NULL), &auio,
1018 (msg.msg_control ? &control : NULL), &flags,
1019 &uap->sysmsg_szresult);
1022 * Conditionally copyout the name and populate the namelen field.
1024 if (error == 0 && msg.msg_name) {
1025 /* note: sa may still be NULL */
1027 fromlen = MIN(msg.msg_namelen, sa->sa_len);
1028 error = copyout(sa, msg.msg_name, fromlen);
1033 error = copyout(&fromlen, ufromlenp,
1034 sizeof(*ufromlenp));
1038 * Copyout msg.msg_control and msg.msg_controllen.
1040 if (error == 0 && msg.msg_control) {
1041 len = msg.msg_controllen;
1043 ctlbuf = (caddr_t)msg.msg_control;
1045 while(m && len > 0) {
1046 unsigned int tocopy;
1048 if (len >= m->m_len) {
1051 msg.msg_flags |= MSG_CTRUNC;
1055 error = copyout(mtod(m, caddr_t), ctlbuf, tocopy);
1063 controllen = ctlbuf - (caddr_t)msg.msg_control;
1064 error = copyout(&controllen, ucontrollenp,
1065 sizeof(*ucontrollenp));
1069 error = copyout(&flags, uflagsp, sizeof(*uflagsp));
1074 iovec_free(&iov, aiov);
1081 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1082 * in kernel pointer instead of a userland pointer. This allows us
1083 * to manipulate socket options in the emulation code.
1086 kern_setsockopt(int s, struct sockopt *sopt)
1088 struct thread *td = curthread;
1089 struct proc *p = td->td_proc;
1093 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1095 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1097 if (sopt->sopt_valsize < 0)
1100 error = holdsock(p->p_fd, s, &fp);
1104 error = sosetopt((struct socket *)fp->f_data, sopt);
1110 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
1115 sys_setsockopt(struct setsockopt_args *uap)
1117 struct thread *td = curthread;
1118 struct sockopt sopt;
1121 sopt.sopt_level = uap->level;
1122 sopt.sopt_name = uap->name;
1123 sopt.sopt_valsize = uap->valsize;
1125 sopt.sopt_val = NULL;
1127 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1130 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1131 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1136 error = kern_setsockopt(uap->s, &sopt);
1139 kfree(sopt.sopt_val, M_TEMP);
1144 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1145 * in kernel pointer instead of a userland pointer. This allows us
1146 * to manipulate socket options in the emulation code.
1149 kern_getsockopt(int s, struct sockopt *sopt)
1151 struct thread *td = curthread;
1152 struct proc *p = td->td_proc;
1156 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1158 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1160 if (sopt->sopt_valsize < 0 || sopt->sopt_valsize > SOMAXOPT_SIZE)
1163 error = holdsock(p->p_fd, s, &fp);
1167 error = sogetopt((struct socket *)fp->f_data, sopt);
1173 * getsockopt_args(int s, int level, int name, caddr_t val, int *avalsize)
1178 sys_getsockopt(struct getsockopt_args *uap)
1180 struct thread *td = curthread;
1181 struct sockopt sopt;
1185 error = copyin(uap->avalsize, &valsize, sizeof(valsize));
1192 sopt.sopt_level = uap->level;
1193 sopt.sopt_name = uap->name;
1194 sopt.sopt_valsize = valsize;
1196 sopt.sopt_val = NULL;
1198 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1201 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1202 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1207 error = kern_getsockopt(uap->s, &sopt);
1210 valsize = sopt.sopt_valsize;
1211 error = copyout(&valsize, uap->avalsize, sizeof(valsize));
1215 error = copyout(sopt.sopt_val, uap->val, sopt.sopt_valsize);
1218 kfree(sopt.sopt_val, M_TEMP);
1223 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1224 * This allows kern_getsockname() to return a pointer to an allocated struct
1225 * sockaddr which must be freed later with FREE(). The caller must
1226 * initialize *name to NULL.
1229 kern_getsockname(int s, struct sockaddr **name, int *namelen)
1231 struct thread *td = curthread;
1232 struct proc *p = td->td_proc;
1235 struct sockaddr *sa = NULL;
1238 error = holdsock(p->p_fd, s, &fp);
1245 so = (struct socket *)fp->f_data;
1246 error = so_pru_sockaddr(so, &sa);
1251 *namelen = MIN(*namelen, sa->sa_len);
1261 * getsockname_args(int fdes, caddr_t asa, int *alen)
1268 sys_getsockname(struct getsockname_args *uap)
1270 struct sockaddr *sa = NULL;
1273 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1277 error = kern_getsockname(uap->fdes, &sa, &sa_len);
1280 error = copyout(sa, uap->asa, sa_len);
1282 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1289 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1290 * This allows kern_getpeername() to return a pointer to an allocated struct
1291 * sockaddr which must be freed later with FREE(). The caller must
1292 * initialize *name to NULL.
1295 kern_getpeername(int s, struct sockaddr **name, int *namelen)
1297 struct thread *td = curthread;
1298 struct proc *p = td->td_proc;
1301 struct sockaddr *sa = NULL;
1304 error = holdsock(p->p_fd, s, &fp);
1311 so = (struct socket *)fp->f_data;
1312 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1316 error = so_pru_peeraddr(so, &sa);
1321 *namelen = MIN(*namelen, sa->sa_len);
1331 * getpeername_args(int fdes, caddr_t asa, int *alen)
1333 * Get name of peer for connected socket.
1338 sys_getpeername(struct getpeername_args *uap)
1340 struct sockaddr *sa = NULL;
1343 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1347 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
1397 sf_buf_mfree(void *arg)
1399 struct sf_buf *sf = arg;
1402 m = sf_buf_page(sf);
1403 if (sf_buf_free(sf)) {
1404 /* sf invalid now */
1405 vm_page_busy_wait(m, FALSE, "sockpgf");
1406 vm_page_unwire(m, 0);
1408 if (m->wire_count == 0 && m->object == NULL)
1409 vm_page_try_to_free(m);
1415 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1416 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1418 * Send a file specified by 'fd' and starting at 'offset' to a socket
1419 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1420 * nbytes == 0. Optionally add a header and/or trailer to the socket
1421 * output. If specified, write the total number of bytes sent into *sbytes.
1423 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1424 * the headers to count against the remaining bytes to be sent from
1425 * the file descriptor. We may wish to implement a compatibility syscall
1431 sys_sendfile(struct sendfile_args *uap)
1433 struct thread *td = curthread;
1434 struct proc *p = td->td_proc;
1436 struct vnode *vp = NULL;
1437 struct sf_hdtr hdtr;
1438 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
1440 struct mbuf *mheader = NULL;
1443 off_t hdtr_size = 0;
1450 * Do argument checking. Must be a regular file in, stream
1451 * type and connected socket out, positive offset.
1453 fp = holdfp(p->p_fd, uap->fd, FREAD);
1457 if (fp->f_type != DTYPE_VNODE) {
1461 vp = (struct vnode *)fp->f_data;
1466 * If specified, get the pointer to the sf_hdtr struct for
1467 * any headers/trailers.
1470 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1477 error = iovec_copyin(hdtr.headers, &iov, aiov,
1478 hdtr.hdr_cnt, &hbytes);
1482 auio.uio_iovcnt = hdtr.hdr_cnt;
1483 auio.uio_offset = 0;
1484 auio.uio_segflg = UIO_USERSPACE;
1485 auio.uio_rw = UIO_WRITE;
1487 auio.uio_resid = hbytes;
1489 mheader = m_uiomove(&auio);
1491 iovec_free(&iov, aiov);
1492 if (mheader == NULL)
1497 error = kern_sendfile(vp, uap->s, uap->offset, uap->nbytes, mheader,
1498 &sbytes, uap->flags);
1503 * Send trailers. Wimp out and use writev(2).
1505 if (uap->hdtr != NULL && hdtr.trailers != NULL) {
1506 error = iovec_copyin(hdtr.trailers, &iov, aiov,
1507 hdtr.trl_cnt, &auio.uio_resid);
1511 auio.uio_iovcnt = hdtr.trl_cnt;
1512 auio.uio_offset = 0;
1513 auio.uio_segflg = UIO_USERSPACE;
1514 auio.uio_rw = UIO_WRITE;
1517 error = kern_sendmsg(uap->s, NULL, &auio, NULL, 0, &tbytes);
1519 iovec_free(&iov, aiov);
1522 hdtr_size += tbytes; /* trailer bytes successfully sent */
1528 if (uap->sbytes != NULL) {
1529 sbytes += hdtr_size;
1530 copyout(&sbytes, uap->sbytes, sizeof(off_t));
1536 kern_sendfile(struct vnode *vp, int sfd, off_t offset, size_t nbytes,
1537 struct mbuf *mheader, off_t *sbytes, int flags)
1539 struct thread *td = curthread;
1540 struct proc *p = td->td_proc;
1541 struct vm_object *obj;
1551 if (vp->v_type != VREG) {
1555 if ((obj = vp->v_object) == NULL) {
1559 error = holdsock(p->p_fd, sfd, &fp);
1562 so = (struct socket *)fp->f_data;
1563 if (so->so_type != SOCK_STREAM) {
1567 if ((so->so_state & SS_ISCONNECTED) == 0) {
1578 * Protect against multiple writers to the socket.
1580 ssb_lock(&so->so_snd, M_WAITOK);
1583 * Loop through the pages in the file, starting with the requested
1584 * offset. Get a file page (do I/O if necessary), map the file page
1585 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1588 for (off = offset; ; off += xfsize, *sbytes += xfsize + hbytes) {
1592 pindex = OFF_TO_IDX(off);
1595 * Calculate the amount to transfer. Not to exceed a page,
1596 * the EOF, or the passed in nbytes.
1598 xfsize = vp->v_filesize - off;
1599 if (xfsize > PAGE_SIZE)
1601 pgoff = (vm_offset_t)(off & PAGE_MASK);
1602 if (PAGE_SIZE - pgoff < xfsize)
1603 xfsize = PAGE_SIZE - pgoff;
1604 if (nbytes && xfsize > (nbytes - *sbytes))
1605 xfsize = nbytes - *sbytes;
1609 * Optimize the non-blocking case by looking at the socket space
1610 * before going to the extra work of constituting the sf_buf.
1612 if ((fp->f_flag & FNONBLOCK) && ssb_space(&so->so_snd) <= 0) {
1613 if (so->so_state & SS_CANTSENDMORE)
1617 ssb_unlock(&so->so_snd);
1621 * Attempt to look up the page.
1623 * Allocate if not found, wait and loop if busy, then
1624 * wire the page. critical section protection is
1625 * required to maintain the object association (an
1626 * interrupt can free the page) through to the
1627 * vm_page_wire() call.
1629 vm_object_hold(obj);
1630 pg = vm_page_lookup_busy_try(obj, pindex, TRUE, &error);
1632 vm_page_sleep_busy(pg, TRUE, "sfpbsy");
1633 vm_object_drop(obj);
1637 pg = vm_page_alloc(obj, pindex, VM_ALLOC_NORMAL |
1641 vm_object_drop(obj);
1646 vm_object_drop(obj);
1649 * If page is not valid for what we need, initiate I/O
1652 if (!pg->valid || !vm_page_is_valid(pg, pgoff, xfsize)) {
1658 * Ensure that our page is still around when the I/O
1661 vm_page_io_start(pg);
1665 * Get the page from backing store.
1667 bsize = vp->v_mount->mnt_stat.f_iosize;
1668 auio.uio_iov = &aiov;
1669 auio.uio_iovcnt = 1;
1671 aiov.iov_len = MAXBSIZE;
1672 auio.uio_resid = MAXBSIZE;
1673 auio.uio_offset = trunc_page(off);
1674 auio.uio_segflg = UIO_NOCOPY;
1675 auio.uio_rw = UIO_READ;
1677 vn_lock(vp, LK_SHARED | LK_RETRY);
1678 error = VOP_READ(vp, &auio,
1679 IO_VMIO | ((MAXBSIZE / bsize) << 16),
1682 vm_page_flag_clear(pg, PG_ZERO);
1683 vm_page_busy_wait(pg, FALSE, "sockpg");
1684 vm_page_io_finish(pg);
1686 vm_page_unwire(pg, 0);
1688 vm_page_try_to_free(pg);
1689 ssb_unlock(&so->so_snd);
1696 * Get a sendfile buf. We usually wait as long as necessary,
1697 * but this wait can be interrupted.
1699 if ((sf = sf_buf_alloc(pg)) == NULL) {
1700 vm_page_unwire(pg, 0);
1702 vm_page_try_to_free(pg);
1703 ssb_unlock(&so->so_snd);
1710 * Get an mbuf header and set it up as having external storage.
1712 MGETHDR(m, MB_WAIT, MT_DATA);
1716 ssb_unlock(&so->so_snd);
1720 m->m_ext.ext_free = sf_buf_mfree;
1721 m->m_ext.ext_ref = sf_buf_ref;
1722 m->m_ext.ext_arg = sf;
1723 m->m_ext.ext_buf = (void *)sf_buf_kva(sf);
1724 m->m_ext.ext_size = PAGE_SIZE;
1725 m->m_data = (char *)sf_buf_kva(sf) + pgoff;
1726 m->m_flags |= M_EXT;
1727 m->m_pkthdr.len = m->m_len = xfsize;
1728 KKASSERT((m->m_flags & (M_EXT_CLUSTER)) == 0);
1730 if (mheader != NULL) {
1731 hbytes = mheader->m_pkthdr.len;
1732 mheader->m_pkthdr.len += m->m_pkthdr.len;
1740 * Add the buffer to the socket buffer chain.
1745 * Make sure that the socket is still able to take more data.
1746 * CANTSENDMORE being true usually means that the connection
1747 * was closed. so_error is true when an error was sensed after
1749 * The state is checked after the page mapping and buffer
1750 * allocation above since those operations may block and make
1751 * any socket checks stale. From this point forward, nothing
1752 * blocks before the pru_send (or more accurately, any blocking
1753 * results in a loop back to here to re-check).
1755 if ((so->so_state & SS_CANTSENDMORE) || so->so_error) {
1756 if (so->so_state & SS_CANTSENDMORE) {
1759 error = so->so_error;
1763 ssb_unlock(&so->so_snd);
1768 * Wait for socket space to become available. We do this just
1769 * after checking the connection state above in order to avoid
1770 * a race condition with ssb_wait().
1772 if (ssb_space(&so->so_snd) < so->so_snd.ssb_lowat) {
1773 if (fp->f_flag & FNONBLOCK) {
1775 ssb_unlock(&so->so_snd);
1780 error = ssb_wait(&so->so_snd);
1782 * An error from ssb_wait usually indicates that we've
1783 * been interrupted by a signal. If we've sent anything
1784 * then return bytes sent, otherwise return the error.
1788 ssb_unlock(&so->so_snd);
1794 error = so_pru_senda(so, 0, m, NULL, NULL, td);
1797 ssb_unlock(&so->so_snd);
1801 if (mheader != NULL) {
1802 *sbytes += mheader->m_pkthdr.len;
1803 error = so_pru_senda(so, 0, mheader, NULL, NULL, td);
1806 ssb_unlock(&so->so_snd);
1811 if (mheader != NULL)
1820 sys_sctp_peeloff(struct sctp_peeloff_args *uap)
1823 struct thread *td = curthread;
1824 struct filedesc *fdp = td->td_proc->p_fd;
1825 struct file *lfp = NULL;
1826 struct file *nfp = NULL;
1828 struct socket *head, *so;
1831 short fflag; /* type must match fp->f_flag */
1833 assoc_id = uap->name;
1834 error = holdsock(td->td_proc->p_fd, uap->sd, &lfp);
1839 head = (struct socket *)lfp->f_data;
1840 error = sctp_can_peel_off(head, assoc_id);
1846 * At this point we know we do have a assoc to pull
1847 * we proceed to get the fd setup. This may block
1851 fflag = lfp->f_flag;
1852 error = falloc(td->td_lwp, &nfp, &fd);
1855 * Probably ran out of file descriptors. Put the
1856 * unaccepted connection back onto the queue and
1857 * do another wakeup so some other process might
1858 * have a chance at it.
1863 uap->sysmsg_iresult = fd;
1865 so = sctp_get_peeloff(head, assoc_id, &error);
1868 * Either someone else peeled it off OR
1869 * we can't get a socket.
1873 soreference(so); /* reference needed */
1874 soclrstate(so, SS_NOFDREF | SS_COMP); /* when clearing NOFDREF */
1876 if (head->so_sigio != NULL)
1877 fsetown(fgetown(&head->so_sigio), &so->so_sigio);
1879 nfp->f_type = DTYPE_SOCKET;
1880 nfp->f_flag = fflag;
1881 nfp->f_ops = &socketops;
1886 * Assign the file pointer to the reserved descriptor, or clear
1887 * the reserved descriptor if an error occured.
1890 fsetfd(fdp, NULL, fd);
1892 fsetfd(fdp, nfp, fd);
1895 * Release explicitly held references before returning.