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 <sys/mplock2.h>
82 #include <net/netmsg2.h>
85 #include <netinet/sctp_peeloff.h>
89 * System call interface to the socket abstraction.
92 extern struct fileops socketops;
95 * socket_args(int domain, int type, int protocol)
98 kern_socket(int domain, int type, int protocol, int *res)
100 struct thread *td = curthread;
101 struct filedesc *fdp = td->td_proc->p_fd;
106 KKASSERT(td->td_lwp);
108 error = falloc(td->td_lwp, &fp, &fd);
111 error = socreate(domain, &so, type, protocol, td);
113 fsetfd(fdp, NULL, fd);
115 fp->f_type = DTYPE_SOCKET;
116 fp->f_flag = FREAD | FWRITE;
117 fp->f_ops = &socketops;
130 sys_socket(struct socket_args *uap)
135 error = kern_socket(uap->domain, uap->type, uap->protocol,
136 &uap->sysmsg_iresult);
143 kern_bind(int s, struct sockaddr *sa)
145 struct thread *td = curthread;
146 struct proc *p = td->td_proc;
151 error = holdsock(p->p_fd, s, &fp);
154 error = sobind((struct socket *)fp->f_data, sa, td);
160 * bind_args(int s, caddr_t name, int namelen)
165 sys_bind(struct bind_args *uap)
170 error = getsockaddr(&sa, uap->name, uap->namelen);
174 error = kern_bind(uap->s, sa);
182 kern_listen(int s, int backlog)
184 struct thread *td = curthread;
185 struct proc *p = td->td_proc;
190 error = holdsock(p->p_fd, s, &fp);
193 error = solisten((struct socket *)fp->f_data, backlog, td);
199 * listen_args(int s, int backlog)
204 sys_listen(struct listen_args *uap)
209 error = kern_listen(uap->s, uap->backlog);
215 * Returns the accepted socket as well.
217 * NOTE! The sockets sitting on so_comp/so_incomp might have 0 refs, the
218 * pool token is absolutely required to avoid a sofree() race,
219 * as well as to avoid tailq handling races.
222 soaccept_predicate(struct netmsg_so_notify *msg)
224 struct socket *head = msg->base.nm_so;
227 if (head->so_error != 0) {
228 msg->base.lmsg.ms_error = head->so_error;
231 lwkt_getpooltoken(head);
232 if (!TAILQ_EMPTY(&head->so_comp)) {
233 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
234 so = TAILQ_FIRST(&head->so_comp);
235 TAILQ_REMOVE(&head->so_comp, so, so_list);
237 soclrstate(so, SS_COMP);
241 lwkt_relpooltoken(head);
243 msg->base.lmsg.ms_error = 0;
244 msg->base.nm_so = so;
247 lwkt_relpooltoken(head);
248 if (head->so_state & SS_CANTRCVMORE) {
249 msg->base.lmsg.ms_error = ECONNABORTED;
252 if (msg->nm_fflags & FNONBLOCK) {
253 msg->base.lmsg.ms_error = EWOULDBLOCK;
261 * The second argument to kern_accept() is a handle to a struct sockaddr.
262 * This allows kern_accept() to return a pointer to an allocated struct
263 * sockaddr which must be freed later with FREE(). The caller must
264 * initialize *name to NULL.
267 kern_accept(int s, int fflags, struct sockaddr **name, int *namelen, int *res)
269 struct thread *td = curthread;
270 struct filedesc *fdp = td->td_proc->p_fd;
271 struct file *lfp = NULL;
272 struct file *nfp = NULL;
274 struct socket *head, *so;
275 struct netmsg_so_notify msg;
277 u_int fflag; /* type must match fp->f_flag */
281 if (name && namelen && *namelen < 0)
284 error = holdsock(td->td_proc->p_fd, s, &lfp);
288 error = falloc(td->td_lwp, &nfp, &fd);
289 if (error) { /* Probably ran out of file descriptors. */
293 head = (struct socket *)lfp->f_data;
294 if ((head->so_options & SO_ACCEPTCONN) == 0) {
299 if (fflags & O_FBLOCKING)
300 fflags |= lfp->f_flag & ~FNONBLOCK;
301 else if (fflags & O_FNONBLOCKING)
302 fflags |= lfp->f_flag | FNONBLOCK;
304 fflags = lfp->f_flag;
306 /* optimize for uniprocessor case later XXX JH */
307 netmsg_init_abortable(&msg.base, head, &curthread->td_msgport,
308 0, netmsg_so_notify, netmsg_so_notify_doabort);
309 msg.nm_predicate = soaccept_predicate;
310 msg.nm_fflags = fflags;
311 msg.nm_etype = NM_REVENT;
312 error = lwkt_domsg(head->so_port, &msg.base.lmsg, PCATCH);
317 * At this point we have the connection that's ready to be accepted.
319 * NOTE! soaccept_predicate() ref'd so for us, and soaccept() expects
320 * to eat the ref and turn it into a descriptor.
326 /* connection has been removed from the listen queue */
327 KNOTE(&head->so_rcv.ssb_kq.ki_note, 0);
329 if (head->so_sigio != NULL)
330 fsetown(fgetown(head->so_sigio), &so->so_sigio);
332 nfp->f_type = DTYPE_SOCKET;
334 nfp->f_ops = &socketops;
336 /* Sync socket nonblocking/async state with file flags */
337 tmp = fflag & FNONBLOCK;
338 fo_ioctl(nfp, FIONBIO, (caddr_t)&tmp, td->td_ucred, NULL);
339 tmp = fflag & FASYNC;
340 fo_ioctl(nfp, FIOASYNC, (caddr_t)&tmp, td->td_ucred, NULL);
343 error = soaccept(so, &sa);
346 * Set the returned name and namelen as applicable. Set the returned
347 * namelen to 0 for older code which might ignore the return value
351 if (sa && name && namelen) {
352 if (*namelen > sa->sa_len)
353 *namelen = sa->sa_len;
363 * If an error occured clear the reserved descriptor, else associate
366 * Note that *res is normally ignored if an error is returned but
367 * a syscall message will still have access to the result code.
370 fsetfd(fdp, NULL, fd);
373 fsetfd(fdp, nfp, fd);
381 * accept(int s, caddr_t name, int *anamelen)
386 sys_accept(struct accept_args *uap)
388 struct sockaddr *sa = NULL;
393 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
398 error = kern_accept(uap->s, 0, &sa, &sa_len,
399 &uap->sysmsg_iresult);
403 error = copyout(sa, uap->name, sa_len);
405 error = copyout(&sa_len, uap->anamelen,
406 sizeof(*uap->anamelen));
412 error = kern_accept(uap->s, 0, NULL, 0,
413 &uap->sysmsg_iresult);
420 * extaccept(int s, int fflags, caddr_t name, int *anamelen)
425 sys_extaccept(struct extaccept_args *uap)
427 struct sockaddr *sa = NULL;
430 int fflags = uap->flags & O_FMASK;
433 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
438 error = kern_accept(uap->s, fflags, &sa, &sa_len,
439 &uap->sysmsg_iresult);
443 error = copyout(sa, uap->name, sa_len);
445 error = copyout(&sa_len, uap->anamelen,
446 sizeof(*uap->anamelen));
452 error = kern_accept(uap->s, fflags, NULL, 0,
453 &uap->sysmsg_iresult);
461 * Returns TRUE if predicate satisfied.
464 soconnected_predicate(struct netmsg_so_notify *msg)
466 struct socket *so = msg->base.nm_so;
468 /* check predicate */
469 if (!(so->so_state & SS_ISCONNECTING) || so->so_error != 0) {
470 msg->base.lmsg.ms_error = so->so_error;
478 kern_connect(int s, int fflags, struct sockaddr *sa)
480 struct thread *td = curthread;
481 struct proc *p = td->td_proc;
484 int error, interrupted = 0;
486 error = holdsock(p->p_fd, s, &fp);
489 so = (struct socket *)fp->f_data;
491 if (fflags & O_FBLOCKING)
492 /* fflags &= ~FNONBLOCK; */;
493 else if (fflags & O_FNONBLOCKING)
498 if (so->so_state & SS_ISCONNECTING) {
502 error = soconnect(so, sa, td);
505 if ((fflags & FNONBLOCK) && (so->so_state & SS_ISCONNECTING)) {
509 if ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
510 struct netmsg_so_notify msg;
512 netmsg_init_abortable(&msg.base, so,
513 &curthread->td_msgport,
516 netmsg_so_notify_doabort);
517 msg.nm_predicate = soconnected_predicate;
518 msg.nm_etype = NM_REVENT;
519 error = lwkt_domsg(so->so_port, &msg.base.lmsg, PCATCH);
520 if (error == EINTR || error == ERESTART)
524 error = so->so_error;
529 soclrstate(so, SS_ISCONNECTING);
530 if (error == ERESTART)
538 * connect_args(int s, caddr_t name, int namelen)
543 sys_connect(struct connect_args *uap)
548 error = getsockaddr(&sa, uap->name, uap->namelen);
552 error = kern_connect(uap->s, 0, sa);
560 * connect_args(int s, int fflags, caddr_t name, int namelen)
565 sys_extconnect(struct extconnect_args *uap)
569 int fflags = uap->flags & O_FMASK;
571 error = getsockaddr(&sa, uap->name, uap->namelen);
575 error = kern_connect(uap->s, fflags, sa);
583 kern_socketpair(int domain, int type, int protocol, int *sv)
585 struct thread *td = curthread;
586 struct filedesc *fdp;
587 struct file *fp1, *fp2;
588 struct socket *so1, *so2;
591 fdp = td->td_proc->p_fd;
592 error = socreate(domain, &so1, type, protocol, td);
595 error = socreate(domain, &so2, type, protocol, td);
598 error = falloc(td->td_lwp, &fp1, &fd1);
603 error = falloc(td->td_lwp, &fp2, &fd2);
608 error = soconnect2(so1, so2);
611 if (type == SOCK_DGRAM) {
613 * Datagram socket connection is asymmetric.
615 error = soconnect2(so2, so1);
619 fp1->f_type = fp2->f_type = DTYPE_SOCKET;
620 fp1->f_flag = fp2->f_flag = FREAD|FWRITE;
621 fp1->f_ops = fp2->f_ops = &socketops;
622 fsetfd(fdp, fp1, fd1);
623 fsetfd(fdp, fp2, fd2);
628 fsetfd(fdp, NULL, fd2);
631 fsetfd(fdp, NULL, fd1);
634 (void)soclose(so2, 0);
636 (void)soclose(so1, 0);
641 * socketpair(int domain, int type, int protocol, int *rsv)
646 sys_socketpair(struct socketpair_args *uap)
651 error = kern_socketpair(uap->domain, uap->type, uap->protocol, sockv);
655 error = copyout(sockv, uap->rsv, sizeof(sockv));
660 kern_sendmsg(int s, struct sockaddr *sa, struct uio *auio,
661 struct mbuf *control, int flags, size_t *res)
663 struct thread *td = curthread;
664 struct lwp *lp = td->td_lwp;
665 struct proc *p = td->td_proc;
671 struct iovec *ktriov = NULL;
675 error = holdsock(p->p_fd, s, &fp);
679 if (KTRPOINT(td, KTR_GENIO)) {
680 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
682 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
683 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
687 len = auio->uio_resid;
688 so = (struct socket *)fp->f_data;
689 if ((flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
690 if (fp->f_flag & FNONBLOCK)
691 flags |= MSG_FNONBLOCKING;
693 error = so_pru_sosend(so, sa, auio, NULL, control, flags, td);
695 if (auio->uio_resid != len && (error == ERESTART ||
696 error == EINTR || error == EWOULDBLOCK))
698 if (error == EPIPE && !(flags & MSG_NOSIGNAL))
699 lwpsignal(p, lp, SIGPIPE);
702 if (ktriov != NULL) {
704 ktruio.uio_iov = ktriov;
705 ktruio.uio_resid = len - auio->uio_resid;
706 ktrgenio(lp, s, UIO_WRITE, &ktruio, error);
708 FREE(ktriov, M_TEMP);
712 *res = len - auio->uio_resid;
718 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
723 sys_sendto(struct sendto_args *uap)
725 struct thread *td = curthread;
728 struct sockaddr *sa = NULL;
732 error = getsockaddr(&sa, uap->to, uap->tolen);
736 aiov.iov_base = uap->buf;
737 aiov.iov_len = uap->len;
738 auio.uio_iov = &aiov;
741 auio.uio_resid = uap->len;
742 auio.uio_segflg = UIO_USERSPACE;
743 auio.uio_rw = UIO_WRITE;
747 error = kern_sendmsg(uap->s, sa, &auio, NULL, uap->flags,
748 &uap->sysmsg_szresult);
757 * sendmsg_args(int s, caddr_t msg, int flags)
762 sys_sendmsg(struct sendmsg_args *uap)
764 struct thread *td = curthread;
767 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
768 struct sockaddr *sa = NULL;
769 struct mbuf *control = NULL;
772 error = copyin(uap->msg, (caddr_t)&msg, sizeof(msg));
777 * Conditionally copyin msg.msg_name.
780 error = getsockaddr(&sa, msg.msg_name, msg.msg_namelen);
788 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
793 auio.uio_iovcnt = msg.msg_iovlen;
795 auio.uio_segflg = UIO_USERSPACE;
796 auio.uio_rw = UIO_WRITE;
800 * Conditionally copyin msg.msg_control.
802 if (msg.msg_control) {
803 if (msg.msg_controllen < sizeof(struct cmsghdr) ||
804 msg.msg_controllen > MLEN) {
808 control = m_get(MB_WAIT, MT_CONTROL);
809 if (control == NULL) {
813 control->m_len = msg.msg_controllen;
814 error = copyin(msg.msg_control, mtod(control, caddr_t),
823 error = kern_sendmsg(uap->s, sa, &auio, control, uap->flags,
824 &uap->sysmsg_szresult);
828 iovec_free(&iov, aiov);
836 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
837 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
838 * Don't forget to FREE() and m_free() these if they are returned.
841 kern_recvmsg(int s, struct sockaddr **sa, struct uio *auio,
842 struct mbuf **control, int *flags, size_t *res)
844 struct thread *td = curthread;
845 struct proc *p = td->td_proc;
852 struct iovec *ktriov = NULL;
856 error = holdsock(p->p_fd, s, &fp);
860 if (KTRPOINT(td, KTR_GENIO)) {
861 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
863 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
864 bcopy(auio->uio_iov, ktriov, iovlen);
868 len = auio->uio_resid;
869 so = (struct socket *)fp->f_data;
871 if (flags == NULL || (*flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
872 if (fp->f_flag & FNONBLOCK) {
874 *flags |= MSG_FNONBLOCKING;
876 lflags = MSG_FNONBLOCKING;
882 error = so_pru_soreceive(so, sa, auio, NULL, control, flags);
884 if (auio->uio_resid != len && (error == ERESTART ||
885 error == EINTR || error == EWOULDBLOCK))
889 if (ktriov != NULL) {
891 ktruio.uio_iov = ktriov;
892 ktruio.uio_resid = len - auio->uio_resid;
893 ktrgenio(td->td_lwp, s, UIO_READ, &ktruio, error);
895 FREE(ktriov, M_TEMP);
899 *res = len - auio->uio_resid;
905 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
906 * caddr_t from, int *fromlenaddr)
911 sys_recvfrom(struct recvfrom_args *uap)
913 struct thread *td = curthread;
916 struct sockaddr *sa = NULL;
919 if (uap->from && uap->fromlenaddr) {
920 error = copyin(uap->fromlenaddr, &fromlen, sizeof(fromlen));
928 aiov.iov_base = uap->buf;
929 aiov.iov_len = uap->len;
930 auio.uio_iov = &aiov;
933 auio.uio_resid = uap->len;
934 auio.uio_segflg = UIO_USERSPACE;
935 auio.uio_rw = UIO_READ;
939 error = kern_recvmsg(uap->s, uap->from ? &sa : NULL, &auio, NULL,
940 &uap->flags, &uap->sysmsg_szresult);
943 if (error == 0 && uap->from) {
944 /* note: sa may still be NULL */
946 fromlen = MIN(fromlen, sa->sa_len);
947 error = copyout(sa, uap->from, fromlen);
952 error = copyout(&fromlen, uap->fromlenaddr,
963 * recvmsg_args(int s, struct msghdr *msg, int flags)
968 sys_recvmsg(struct recvmsg_args *uap)
970 struct thread *td = curthread;
973 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
974 struct mbuf *m, *control = NULL;
975 struct sockaddr *sa = NULL;
977 socklen_t *ufromlenp, *ucontrollenp;
978 int error, fromlen, controllen, len, flags, *uflagsp;
981 * This copyin handles everything except the iovec.
983 error = copyin(uap->msg, &msg, sizeof(msg));
987 if (msg.msg_name && msg.msg_namelen < 0)
989 if (msg.msg_control && msg.msg_controllen < 0)
992 ufromlenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
994 ucontrollenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
996 uflagsp = (int *)((caddr_t)uap->msg + offsetof(struct msghdr,
1002 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
1007 auio.uio_iovcnt = msg.msg_iovlen;
1008 auio.uio_offset = 0;
1009 auio.uio_segflg = UIO_USERSPACE;
1010 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);
1023 * Conditionally copyout the name and populate the namelen field.
1025 if (error == 0 && msg.msg_name) {
1026 /* note: sa may still be NULL */
1028 fromlen = MIN(msg.msg_namelen, sa->sa_len);
1029 error = copyout(sa, msg.msg_name, fromlen);
1034 error = copyout(&fromlen, ufromlenp,
1035 sizeof(*ufromlenp));
1039 * Copyout msg.msg_control and msg.msg_controllen.
1041 if (error == 0 && msg.msg_control) {
1042 len = msg.msg_controllen;
1044 ctlbuf = (caddr_t)msg.msg_control;
1046 while(m && len > 0) {
1047 unsigned int tocopy;
1049 if (len >= m->m_len) {
1052 msg.msg_flags |= MSG_CTRUNC;
1056 error = copyout(mtod(m, caddr_t), ctlbuf, tocopy);
1064 controllen = ctlbuf - (caddr_t)msg.msg_control;
1065 error = copyout(&controllen, ucontrollenp,
1066 sizeof(*ucontrollenp));
1070 error = copyout(&flags, uflagsp, sizeof(*uflagsp));
1075 iovec_free(&iov, aiov);
1082 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1083 * in kernel pointer instead of a userland pointer. This allows us
1084 * to manipulate socket options in the emulation code.
1087 kern_setsockopt(int s, struct sockopt *sopt)
1089 struct thread *td = curthread;
1090 struct proc *p = td->td_proc;
1094 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1096 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1098 if (sopt->sopt_valsize < 0)
1101 error = holdsock(p->p_fd, s, &fp);
1105 error = sosetopt((struct socket *)fp->f_data, sopt);
1111 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
1116 sys_setsockopt(struct setsockopt_args *uap)
1118 struct thread *td = curthread;
1119 struct sockopt sopt;
1122 sopt.sopt_level = uap->level;
1123 sopt.sopt_name = uap->name;
1124 sopt.sopt_valsize = uap->valsize;
1126 sopt.sopt_val = NULL;
1128 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1131 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1132 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1138 error = kern_setsockopt(uap->s, &sopt);
1142 kfree(sopt.sopt_val, M_TEMP);
1147 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1148 * in kernel pointer instead of a userland pointer. This allows us
1149 * to manipulate socket options in the emulation code.
1152 kern_getsockopt(int s, struct sockopt *sopt)
1154 struct thread *td = curthread;
1155 struct proc *p = td->td_proc;
1159 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1161 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1163 if (sopt->sopt_valsize < 0 || sopt->sopt_valsize > SOMAXOPT_SIZE)
1166 error = holdsock(p->p_fd, s, &fp);
1170 error = sogetopt((struct socket *)fp->f_data, sopt);
1176 * getsockopt_args(int s, int level, int name, caddr_t val, int *avalsize)
1181 sys_getsockopt(struct getsockopt_args *uap)
1183 struct thread *td = curthread;
1184 struct sockopt sopt;
1188 error = copyin(uap->avalsize, &valsize, sizeof(valsize));
1195 sopt.sopt_level = uap->level;
1196 sopt.sopt_name = uap->name;
1197 sopt.sopt_valsize = valsize;
1199 sopt.sopt_val = NULL;
1201 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1204 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1205 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1211 error = kern_getsockopt(uap->s, &sopt);
1215 valsize = sopt.sopt_valsize;
1216 error = copyout(&valsize, uap->avalsize, sizeof(valsize));
1220 error = copyout(sopt.sopt_val, uap->val, sopt.sopt_valsize);
1223 kfree(sopt.sopt_val, M_TEMP);
1228 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1229 * This allows kern_getsockname() to return a pointer to an allocated struct
1230 * sockaddr which must be freed later with FREE(). The caller must
1231 * initialize *name to NULL.
1234 kern_getsockname(int s, struct sockaddr **name, int *namelen)
1236 struct thread *td = curthread;
1237 struct proc *p = td->td_proc;
1240 struct sockaddr *sa = NULL;
1243 error = holdsock(p->p_fd, s, &fp);
1250 so = (struct socket *)fp->f_data;
1251 error = so_pru_sockaddr(so, &sa);
1256 *namelen = MIN(*namelen, sa->sa_len);
1266 * getsockname_args(int fdes, caddr_t asa, int *alen)
1273 sys_getsockname(struct getsockname_args *uap)
1275 struct sockaddr *sa = NULL;
1278 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1283 error = kern_getsockname(uap->fdes, &sa, &sa_len);
1287 error = copyout(sa, uap->asa, sa_len);
1289 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1296 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1297 * This allows kern_getpeername() to return a pointer to an allocated struct
1298 * sockaddr which must be freed later with FREE(). The caller must
1299 * initialize *name to NULL.
1302 kern_getpeername(int s, struct sockaddr **name, int *namelen)
1304 struct thread *td = curthread;
1305 struct proc *p = td->td_proc;
1308 struct sockaddr *sa = NULL;
1311 error = holdsock(p->p_fd, s, &fp);
1318 so = (struct socket *)fp->f_data;
1319 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1323 error = so_pru_peeraddr(so, &sa);
1328 *namelen = MIN(*namelen, sa->sa_len);
1338 * getpeername_args(int fdes, caddr_t asa, int *alen)
1340 * Get name of peer for connected socket.
1345 sys_getpeername(struct getpeername_args *uap)
1347 struct sockaddr *sa = NULL;
1350 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1355 error = kern_getpeername(uap->fdes, &sa, &sa_len);
1359 error = copyout(sa, uap->asa, sa_len);
1361 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1368 getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len)
1370 struct sockaddr *sa;
1374 if (len > SOCK_MAXADDRLEN)
1375 return ENAMETOOLONG;
1376 if (len < offsetof(struct sockaddr, sa_data[0]))
1378 MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1379 error = copyin(uaddr, sa, len);
1383 #if BYTE_ORDER != BIG_ENDIAN
1385 * The bind(), connect(), and sendto() syscalls were not
1386 * versioned for COMPAT_43. Thus, this check must stay.
1388 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1389 sa->sa_family = sa->sa_len;
1398 * Detach a mapped page and release resources back to the system.
1399 * We must release our wiring and if the object is ripped out
1400 * from under the vm_page we become responsible for freeing the
1401 * page. These routines must be MPSAFE.
1403 * XXX HACK XXX TEMPORARY UNTIL WE IMPLEMENT EXT MBUF REFERENCE COUNTING
1405 * XXX vm_page_*() routines are not MPSAFE yet, the MP lock is required.
1408 sf_buf_mfree(void *arg)
1410 struct sf_buf *sf = arg;
1414 * XXX vm_page_*() and SFBUF routines not MPSAFE yet.
1418 m = sf_buf_page(sf);
1419 if (sf_buf_free(sf) == 0) {
1420 vm_page_unwire(m, 0);
1421 if (m->wire_count == 0 && m->object == NULL)
1422 vm_page_try_to_free(m);
1430 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1431 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1433 * Send a file specified by 'fd' and starting at 'offset' to a socket
1434 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1435 * nbytes == 0. Optionally add a header and/or trailer to the socket
1436 * output. If specified, write the total number of bytes sent into *sbytes.
1438 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1439 * the headers to count against the remaining bytes to be sent from
1440 * the file descriptor. We may wish to implement a compatibility syscall
1446 sys_sendfile(struct sendfile_args *uap)
1448 struct thread *td = curthread;
1449 struct proc *p = td->td_proc;
1451 struct vnode *vp = NULL;
1452 struct sf_hdtr hdtr;
1453 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
1455 struct mbuf *mheader = NULL;
1458 off_t hdtr_size = 0;
1465 * Do argument checking. Must be a regular file in, stream
1466 * type and connected socket out, positive offset.
1468 fp = holdfp(p->p_fd, uap->fd, FREAD);
1472 if (fp->f_type != DTYPE_VNODE) {
1477 vp = (struct vnode *)fp->f_data;
1482 * If specified, get the pointer to the sf_hdtr struct for
1483 * any headers/trailers.
1486 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1493 error = iovec_copyin(hdtr.headers, &iov, aiov,
1494 hdtr.hdr_cnt, &hbytes);
1498 auio.uio_iovcnt = hdtr.hdr_cnt;
1499 auio.uio_offset = 0;
1500 auio.uio_segflg = UIO_USERSPACE;
1501 auio.uio_rw = UIO_WRITE;
1503 auio.uio_resid = hbytes;
1505 mheader = m_uiomove(&auio);
1507 iovec_free(&iov, aiov);
1508 if (mheader == NULL)
1513 error = kern_sendfile(vp, uap->s, uap->offset, uap->nbytes, mheader,
1514 &sbytes, uap->flags);
1519 * Send trailers. Wimp out and use writev(2).
1521 if (uap->hdtr != NULL && hdtr.trailers != NULL) {
1522 error = iovec_copyin(hdtr.trailers, &iov, aiov,
1523 hdtr.trl_cnt, &auio.uio_resid);
1527 auio.uio_iovcnt = hdtr.trl_cnt;
1528 auio.uio_offset = 0;
1529 auio.uio_segflg = UIO_USERSPACE;
1530 auio.uio_rw = UIO_WRITE;
1533 error = kern_sendmsg(uap->s, NULL, &auio, NULL, 0, &tbytes);
1535 iovec_free(&iov, aiov);
1538 hdtr_size += tbytes; /* trailer bytes successfully sent */
1545 if (uap->sbytes != NULL) {
1546 sbytes += hdtr_size;
1547 copyout(&sbytes, uap->sbytes, sizeof(off_t));
1553 kern_sendfile(struct vnode *vp, int sfd, off_t offset, size_t nbytes,
1554 struct mbuf *mheader, off_t *sbytes, int flags)
1556 struct thread *td = curthread;
1557 struct proc *p = td->td_proc;
1558 struct vm_object *obj;
1568 if (vp->v_type != VREG) {
1572 if ((obj = vp->v_object) == NULL) {
1576 error = holdsock(p->p_fd, sfd, &fp);
1579 so = (struct socket *)fp->f_data;
1580 if (so->so_type != SOCK_STREAM) {
1584 if ((so->so_state & SS_ISCONNECTED) == 0) {
1595 * Protect against multiple writers to the socket.
1597 ssb_lock(&so->so_snd, M_WAITOK);
1600 * Loop through the pages in the file, starting with the requested
1601 * offset. Get a file page (do I/O if necessary), map the file page
1602 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1605 for (off = offset; ; off += xfsize, *sbytes += xfsize + hbytes) {
1609 pindex = OFF_TO_IDX(off);
1612 * Calculate the amount to transfer. Not to exceed a page,
1613 * the EOF, or the passed in nbytes.
1615 xfsize = vp->v_filesize - off;
1616 if (xfsize > PAGE_SIZE)
1618 pgoff = (vm_offset_t)(off & PAGE_MASK);
1619 if (PAGE_SIZE - pgoff < xfsize)
1620 xfsize = PAGE_SIZE - pgoff;
1621 if (nbytes && xfsize > (nbytes - *sbytes))
1622 xfsize = nbytes - *sbytes;
1626 * Optimize the non-blocking case by looking at the socket space
1627 * before going to the extra work of constituting the sf_buf.
1629 if ((fp->f_flag & FNONBLOCK) && ssb_space(&so->so_snd) <= 0) {
1630 if (so->so_state & SS_CANTSENDMORE)
1634 ssb_unlock(&so->so_snd);
1638 * Attempt to look up the page.
1640 * Allocate if not found, wait and loop if busy, then
1641 * wire the page. critical section protection is
1642 * required to maintain the object association (an
1643 * interrupt can free the page) through to the
1644 * vm_page_wire() call.
1646 lwkt_gettoken(&vm_token);
1647 pg = vm_page_lookup(obj, pindex);
1649 pg = vm_page_alloc(obj, pindex, VM_ALLOC_NORMAL);
1652 lwkt_reltoken(&vm_token);
1657 } else if (vm_page_sleep_busy(pg, TRUE, "sfpbsy")) {
1658 lwkt_reltoken(&vm_token);
1663 lwkt_reltoken(&vm_token);
1666 * If page is not valid for what we need, initiate I/O
1669 if (!pg->valid || !vm_page_is_valid(pg, pgoff, xfsize)) {
1675 * Ensure that our page is still around when the I/O
1678 vm_page_io_start(pg);
1681 * Get the page from backing store.
1683 bsize = vp->v_mount->mnt_stat.f_iosize;
1684 auio.uio_iov = &aiov;
1685 auio.uio_iovcnt = 1;
1687 aiov.iov_len = MAXBSIZE;
1688 auio.uio_resid = MAXBSIZE;
1689 auio.uio_offset = trunc_page(off);
1690 auio.uio_segflg = UIO_NOCOPY;
1691 auio.uio_rw = UIO_READ;
1693 vn_lock(vp, LK_SHARED | LK_RETRY);
1694 error = VOP_READ(vp, &auio,
1695 IO_VMIO | ((MAXBSIZE / bsize) << 16),
1698 vm_page_flag_clear(pg, PG_ZERO);
1699 vm_page_io_finish(pg);
1702 vm_page_unwire(pg, 0);
1703 vm_page_try_to_free(pg);
1705 ssb_unlock(&so->so_snd);
1712 * Get a sendfile buf. We usually wait as long as necessary,
1713 * but this wait can be interrupted.
1715 if ((sf = sf_buf_alloc(pg)) == NULL) {
1717 vm_page_unwire(pg, 0);
1718 vm_page_try_to_free(pg);
1720 ssb_unlock(&so->so_snd);
1726 * Get an mbuf header and set it up as having external storage.
1728 MGETHDR(m, MB_WAIT, MT_DATA);
1732 ssb_unlock(&so->so_snd);
1736 m->m_ext.ext_free = sf_buf_mfree;
1737 m->m_ext.ext_ref = sf_buf_ref;
1738 m->m_ext.ext_arg = sf;
1739 m->m_ext.ext_buf = (void *)sf_buf_kva(sf);
1740 m->m_ext.ext_size = PAGE_SIZE;
1741 m->m_data = (char *)sf_buf_kva(sf) + pgoff;
1742 m->m_flags |= M_EXT;
1743 m->m_pkthdr.len = m->m_len = xfsize;
1744 KKASSERT((m->m_flags & (M_EXT_CLUSTER)) == 0);
1746 if (mheader != NULL) {
1747 hbytes = mheader->m_pkthdr.len;
1748 mheader->m_pkthdr.len += m->m_pkthdr.len;
1756 * Add the buffer to the socket buffer chain.
1761 * Make sure that the socket is still able to take more data.
1762 * CANTSENDMORE being true usually means that the connection
1763 * was closed. so_error is true when an error was sensed after
1765 * The state is checked after the page mapping and buffer
1766 * allocation above since those operations may block and make
1767 * any socket checks stale. From this point forward, nothing
1768 * blocks before the pru_send (or more accurately, any blocking
1769 * results in a loop back to here to re-check).
1771 if ((so->so_state & SS_CANTSENDMORE) || so->so_error) {
1772 if (so->so_state & SS_CANTSENDMORE) {
1775 error = so->so_error;
1779 ssb_unlock(&so->so_snd);
1784 * Wait for socket space to become available. We do this just
1785 * after checking the connection state above in order to avoid
1786 * a race condition with ssb_wait().
1788 if (ssb_space(&so->so_snd) < so->so_snd.ssb_lowat) {
1789 if (fp->f_flag & FNONBLOCK) {
1791 ssb_unlock(&so->so_snd);
1796 error = ssb_wait(&so->so_snd);
1798 * An error from ssb_wait usually indicates that we've
1799 * been interrupted by a signal. If we've sent anything
1800 * then return bytes sent, otherwise return the error.
1804 ssb_unlock(&so->so_snd);
1810 error = so_pru_send(so, 0, m, NULL, NULL, td);
1813 ssb_unlock(&so->so_snd);
1817 if (mheader != NULL) {
1818 *sbytes += mheader->m_pkthdr.len;
1819 error = so_pru_send(so, 0, mheader, NULL, NULL, td);
1822 ssb_unlock(&so->so_snd);
1827 if (mheader != NULL)
1836 sys_sctp_peeloff(struct sctp_peeloff_args *uap)
1839 struct thread *td = curthread;
1840 struct filedesc *fdp = td->td_proc->p_fd;
1841 struct file *lfp = NULL;
1842 struct file *nfp = NULL;
1844 struct socket *head, *so;
1847 short fflag; /* type must match fp->f_flag */
1849 assoc_id = uap->name;
1850 error = holdsock(td->td_proc->p_fd, uap->sd, &lfp);
1856 head = (struct socket *)lfp->f_data;
1857 error = sctp_can_peel_off(head, assoc_id);
1863 * At this point we know we do have a assoc to pull
1864 * we proceed to get the fd setup. This may block
1868 fflag = lfp->f_flag;
1869 error = falloc(td->td_lwp, &nfp, &fd);
1872 * Probably ran out of file descriptors. Put the
1873 * unaccepted connection back onto the queue and
1874 * do another wakeup so some other process might
1875 * have a chance at it.
1880 uap->sysmsg_iresult = fd;
1882 so = sctp_get_peeloff(head, assoc_id, &error);
1885 * Either someone else peeled it off OR
1886 * we can't get a socket.
1890 soreference(so); /* reference needed */
1891 soclrstate(so, SS_NOFDREF | SS_COMP); /* when clearing NOFDREF */
1893 if (head->so_sigio != NULL)
1894 fsetown(fgetown(head->so_sigio), &so->so_sigio);
1896 nfp->f_type = DTYPE_SOCKET;
1897 nfp->f_flag = fflag;
1898 nfp->f_ops = &socketops;
1903 * Assign the file pointer to the reserved descriptor, or clear
1904 * the reserved descriptor if an error occured.
1907 fsetfd(fdp, NULL, fd);
1909 fsetfd(fdp, nfp, fd);
1912 * Release explicitly held references before returning.