2 * Copyright (c) 1982, 1986, 1989, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
5 * sendfile(2) and related extensions:
6 * Copyright (c) 1998, David Greenman. All rights reserved.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * @(#)uipc_syscalls.c 8.4 (Berkeley) 2/21/94
37 * $FreeBSD: src/sys/kern/uipc_syscalls.c,v 1.65.2.17 2003/04/04 17:11:16 tegge Exp $
40 #include "opt_ktrace.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
46 #include <sys/sysproto.h>
47 #include <sys/malloc.h>
48 #include <sys/filedesc.h>
49 #include <sys/event.h>
51 #include <sys/fcntl.h>
53 #include <sys/filio.h>
54 #include <sys/kern_syscall.h>
56 #include <sys/protosw.h>
57 #include <sys/sfbuf.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
60 #include <sys/socketops.h>
62 #include <sys/vnode.h>
64 #include <sys/mount.h>
66 #include <sys/ktrace.h>
69 #include <vm/vm_object.h>
70 #include <vm/vm_page.h>
71 #include <vm/vm_pageout.h>
72 #include <vm/vm_kern.h>
73 #include <vm/vm_extern.h>
74 #include <sys/file2.h>
75 #include <sys/signalvar.h>
76 #include <sys/serialize.h>
78 #include <sys/thread2.h>
79 #include <sys/msgport2.h>
80 #include <sys/socketvar2.h>
81 #include <net/netmsg2.h>
84 #include <netinet/sctp_peeloff.h>
88 * System call interface to the socket abstraction.
91 extern struct fileops socketops;
94 * socket_args(int domain, int type, int protocol)
97 kern_socket(int domain, int type, int protocol, int *res)
99 struct thread *td = curthread;
100 struct filedesc *fdp = td->td_proc->p_fd;
105 KKASSERT(td->td_lwp);
107 error = falloc(td->td_lwp, &fp, &fd);
110 error = socreate(domain, &so, type, protocol, td);
112 fsetfd(fdp, NULL, fd);
114 fp->f_type = DTYPE_SOCKET;
115 fp->f_flag = FREAD | FWRITE;
116 fp->f_ops = &socketops;
129 sys_socket(struct socket_args *uap)
133 error = kern_socket(uap->domain, uap->type, uap->protocol,
134 &uap->sysmsg_iresult);
140 kern_bind(int s, struct sockaddr *sa)
142 struct thread *td = curthread;
143 struct proc *p = td->td_proc;
148 error = holdsock(p->p_fd, s, &fp);
151 error = sobind((struct socket *)fp->f_data, sa, td);
157 * bind_args(int s, caddr_t name, int namelen)
162 sys_bind(struct bind_args *uap)
167 error = getsockaddr(&sa, uap->name, uap->namelen);
170 error = kern_bind(uap->s, sa);
177 kern_listen(int s, int backlog)
179 struct thread *td = curthread;
180 struct proc *p = td->td_proc;
185 error = holdsock(p->p_fd, s, &fp);
188 error = solisten((struct socket *)fp->f_data, backlog, td);
194 * listen_args(int s, int backlog)
199 sys_listen(struct listen_args *uap)
203 error = kern_listen(uap->s, uap->backlog);
208 * Returns the accepted socket as well.
210 * NOTE! The sockets sitting on so_comp/so_incomp might have 0 refs, the
211 * pool token is absolutely required to avoid a sofree() race,
212 * as well as to avoid tailq handling races.
215 soaccept_predicate(struct netmsg_so_notify *msg)
217 struct socket *head = msg->base.nm_so;
220 if (head->so_error != 0) {
221 msg->base.lmsg.ms_error = head->so_error;
224 lwkt_getpooltoken(head);
225 if (!TAILQ_EMPTY(&head->so_comp)) {
226 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
227 so = TAILQ_FIRST(&head->so_comp);
228 TAILQ_REMOVE(&head->so_comp, so, so_list);
230 soclrstate(so, SS_COMP);
234 lwkt_relpooltoken(head);
236 msg->base.lmsg.ms_error = 0;
237 msg->base.nm_so = so;
240 lwkt_relpooltoken(head);
241 if (head->so_state & SS_CANTRCVMORE) {
242 msg->base.lmsg.ms_error = ECONNABORTED;
245 if (msg->nm_fflags & FNONBLOCK) {
246 msg->base.lmsg.ms_error = EWOULDBLOCK;
254 * The second argument to kern_accept() is a handle to a struct sockaddr.
255 * This allows kern_accept() to return a pointer to an allocated struct
256 * sockaddr which must be freed later with FREE(). The caller must
257 * initialize *name to NULL.
260 kern_accept(int s, int fflags, struct sockaddr **name, int *namelen, int *res)
262 struct thread *td = curthread;
263 struct filedesc *fdp = td->td_proc->p_fd;
264 struct file *lfp = NULL;
265 struct file *nfp = NULL;
267 struct socket *head, *so;
268 struct netmsg_so_notify msg;
270 u_int fflag; /* type must match fp->f_flag */
274 if (name && namelen && *namelen < 0)
277 error = holdsock(td->td_proc->p_fd, s, &lfp);
281 error = falloc(td->td_lwp, &nfp, &fd);
282 if (error) { /* Probably ran out of file descriptors. */
286 head = (struct socket *)lfp->f_data;
287 if ((head->so_options & SO_ACCEPTCONN) == 0) {
292 if (fflags & O_FBLOCKING)
293 fflags |= lfp->f_flag & ~FNONBLOCK;
294 else if (fflags & O_FNONBLOCKING)
295 fflags |= lfp->f_flag | FNONBLOCK;
297 fflags = lfp->f_flag;
299 /* optimize for uniprocessor case later XXX JH */
300 netmsg_init_abortable(&msg.base, head, &curthread->td_msgport,
301 0, netmsg_so_notify, netmsg_so_notify_doabort);
302 msg.nm_predicate = soaccept_predicate;
303 msg.nm_fflags = fflags;
304 msg.nm_etype = NM_REVENT;
305 error = lwkt_domsg(head->so_port, &msg.base.lmsg, PCATCH);
310 * At this point we have the connection that's ready to be accepted.
312 * NOTE! soaccept_predicate() ref'd so for us, and soaccept() expects
313 * to eat the ref and turn it into a descriptor.
319 /* connection has been removed from the listen queue */
320 KNOTE(&head->so_rcv.ssb_kq.ki_note, 0);
322 if (head->so_sigio != NULL)
323 fsetown(fgetown(&head->so_sigio), &so->so_sigio);
325 nfp->f_type = DTYPE_SOCKET;
327 nfp->f_ops = &socketops;
329 /* Sync socket nonblocking/async state with file flags */
330 tmp = fflag & FNONBLOCK;
331 fo_ioctl(nfp, FIONBIO, (caddr_t)&tmp, td->td_ucred, NULL);
332 tmp = fflag & FASYNC;
333 fo_ioctl(nfp, FIOASYNC, (caddr_t)&tmp, td->td_ucred, NULL);
336 error = soaccept(so, &sa);
339 * Set the returned name and namelen as applicable. Set the returned
340 * namelen to 0 for older code which might ignore the return value
344 if (sa && name && namelen) {
345 if (*namelen > sa->sa_len)
346 *namelen = sa->sa_len;
356 * If an error occured clear the reserved descriptor, else associate
359 * Note that *res is normally ignored if an error is returned but
360 * a syscall message will still have access to the result code.
363 fsetfd(fdp, NULL, fd);
366 fsetfd(fdp, nfp, fd);
374 * accept(int s, caddr_t name, int *anamelen)
379 sys_accept(struct accept_args *uap)
381 struct sockaddr *sa = NULL;
386 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
390 error = kern_accept(uap->s, 0, &sa, &sa_len,
391 &uap->sysmsg_iresult);
394 error = copyout(sa, uap->name, sa_len);
396 error = copyout(&sa_len, uap->anamelen,
397 sizeof(*uap->anamelen));
402 error = kern_accept(uap->s, 0, NULL, 0,
403 &uap->sysmsg_iresult);
409 * extaccept(int s, int fflags, caddr_t name, int *anamelen)
414 sys_extaccept(struct extaccept_args *uap)
416 struct sockaddr *sa = NULL;
419 int fflags = uap->flags & O_FMASK;
422 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
426 error = kern_accept(uap->s, fflags, &sa, &sa_len,
427 &uap->sysmsg_iresult);
430 error = copyout(sa, uap->name, sa_len);
432 error = copyout(&sa_len, uap->anamelen,
433 sizeof(*uap->anamelen));
438 error = kern_accept(uap->s, fflags, NULL, 0,
439 &uap->sysmsg_iresult);
446 * Returns TRUE if predicate satisfied.
449 soconnected_predicate(struct netmsg_so_notify *msg)
451 struct socket *so = msg->base.nm_so;
453 /* check predicate */
454 if (!(so->so_state & SS_ISCONNECTING) || so->so_error != 0) {
455 msg->base.lmsg.ms_error = so->so_error;
463 kern_connect(int s, int fflags, struct sockaddr *sa)
465 struct thread *td = curthread;
466 struct proc *p = td->td_proc;
469 int error, interrupted = 0;
471 error = holdsock(p->p_fd, s, &fp);
474 so = (struct socket *)fp->f_data;
476 if (fflags & O_FBLOCKING)
477 /* fflags &= ~FNONBLOCK; */;
478 else if (fflags & O_FNONBLOCKING)
483 if (so->so_state & SS_ISCONNECTING) {
487 error = soconnect(so, sa, td);
490 if ((fflags & FNONBLOCK) && (so->so_state & SS_ISCONNECTING)) {
494 if ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
495 struct netmsg_so_notify msg;
497 netmsg_init_abortable(&msg.base, so,
498 &curthread->td_msgport,
501 netmsg_so_notify_doabort);
502 msg.nm_predicate = soconnected_predicate;
503 msg.nm_etype = NM_REVENT;
504 error = lwkt_domsg(so->so_port, &msg.base.lmsg, PCATCH);
505 if (error == EINTR || error == ERESTART)
509 error = so->so_error;
514 soclrstate(so, SS_ISCONNECTING);
515 if (error == ERESTART)
523 * connect_args(int s, caddr_t name, int namelen)
528 sys_connect(struct connect_args *uap)
533 error = getsockaddr(&sa, uap->name, uap->namelen);
536 error = kern_connect(uap->s, 0, sa);
543 * connect_args(int s, int fflags, caddr_t name, int namelen)
548 sys_extconnect(struct extconnect_args *uap)
552 int fflags = uap->flags & O_FMASK;
554 error = getsockaddr(&sa, uap->name, uap->namelen);
557 error = kern_connect(uap->s, fflags, sa);
564 kern_socketpair(int domain, int type, int protocol, int *sv)
566 struct thread *td = curthread;
567 struct filedesc *fdp;
568 struct file *fp1, *fp2;
569 struct socket *so1, *so2;
572 fdp = td->td_proc->p_fd;
573 error = socreate(domain, &so1, type, protocol, td);
576 error = socreate(domain, &so2, type, protocol, td);
579 error = falloc(td->td_lwp, &fp1, &fd1);
584 error = falloc(td->td_lwp, &fp2, &fd2);
589 error = soconnect2(so1, so2);
592 if (type == SOCK_DGRAM) {
594 * Datagram socket connection is asymmetric.
596 error = soconnect2(so2, so1);
600 fp1->f_type = fp2->f_type = DTYPE_SOCKET;
601 fp1->f_flag = fp2->f_flag = FREAD|FWRITE;
602 fp1->f_ops = fp2->f_ops = &socketops;
603 fsetfd(fdp, fp1, fd1);
604 fsetfd(fdp, fp2, fd2);
609 fsetfd(fdp, NULL, fd2);
612 fsetfd(fdp, NULL, fd1);
615 (void)soclose(so2, 0);
617 (void)soclose(so1, 0);
622 * socketpair(int domain, int type, int protocol, int *rsv)
625 sys_socketpair(struct socketpair_args *uap)
629 error = kern_socketpair(uap->domain, uap->type, uap->protocol, sockv);
632 error = copyout(sockv, uap->rsv, sizeof(sockv));
635 kern_close(sockv[0]);
636 kern_close(sockv[1]);
644 kern_sendmsg(int s, struct sockaddr *sa, struct uio *auio,
645 struct mbuf *control, int flags, size_t *res)
647 struct thread *td = curthread;
648 struct lwp *lp = td->td_lwp;
649 struct proc *p = td->td_proc;
655 struct iovec *ktriov = NULL;
659 error = holdsock(p->p_fd, s, &fp);
663 if (KTRPOINT(td, KTR_GENIO)) {
664 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
666 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
667 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
671 len = auio->uio_resid;
672 so = (struct socket *)fp->f_data;
673 if ((flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
674 if (fp->f_flag & FNONBLOCK)
675 flags |= MSG_FNONBLOCKING;
677 error = so_pru_sosend(so, sa, auio, NULL, control, flags, td);
679 if (auio->uio_resid != len && (error == ERESTART ||
680 error == EINTR || error == EWOULDBLOCK))
682 if (error == EPIPE && !(flags & MSG_NOSIGNAL))
683 lwpsignal(p, lp, SIGPIPE);
686 if (ktriov != NULL) {
688 ktruio.uio_iov = ktriov;
689 ktruio.uio_resid = len - auio->uio_resid;
690 ktrgenio(lp, s, UIO_WRITE, &ktruio, error);
692 FREE(ktriov, M_TEMP);
696 *res = len - auio->uio_resid;
702 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
707 sys_sendto(struct sendto_args *uap)
709 struct thread *td = curthread;
712 struct sockaddr *sa = NULL;
716 error = getsockaddr(&sa, uap->to, uap->tolen);
720 aiov.iov_base = uap->buf;
721 aiov.iov_len = uap->len;
722 auio.uio_iov = &aiov;
725 auio.uio_resid = uap->len;
726 auio.uio_segflg = UIO_USERSPACE;
727 auio.uio_rw = UIO_WRITE;
730 error = kern_sendmsg(uap->s, sa, &auio, NULL, uap->flags,
731 &uap->sysmsg_szresult);
739 * sendmsg_args(int s, caddr_t msg, int flags)
744 sys_sendmsg(struct sendmsg_args *uap)
746 struct thread *td = curthread;
749 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
750 struct sockaddr *sa = NULL;
751 struct mbuf *control = NULL;
754 error = copyin(uap->msg, (caddr_t)&msg, sizeof(msg));
759 * Conditionally copyin msg.msg_name.
762 error = getsockaddr(&sa, msg.msg_name, msg.msg_namelen);
770 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
775 auio.uio_iovcnt = msg.msg_iovlen;
777 auio.uio_segflg = UIO_USERSPACE;
778 auio.uio_rw = UIO_WRITE;
782 * Conditionally copyin msg.msg_control.
784 if (msg.msg_control) {
785 if (msg.msg_controllen < sizeof(struct cmsghdr) ||
786 msg.msg_controllen > MLEN) {
790 control = m_get(MB_WAIT, MT_CONTROL);
791 if (control == NULL) {
795 control->m_len = msg.msg_controllen;
796 error = copyin(msg.msg_control, mtod(control, caddr_t),
804 error = kern_sendmsg(uap->s, sa, &auio, control, uap->flags,
805 &uap->sysmsg_szresult);
808 iovec_free(&iov, aiov);
816 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
817 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
818 * Don't forget to FREE() and m_free() these if they are returned.
821 kern_recvmsg(int s, struct sockaddr **sa, struct uio *auio,
822 struct mbuf **control, int *flags, size_t *res)
824 struct thread *td = curthread;
825 struct proc *p = td->td_proc;
832 struct iovec *ktriov = NULL;
836 error = holdsock(p->p_fd, s, &fp);
840 if (KTRPOINT(td, KTR_GENIO)) {
841 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
843 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
844 bcopy(auio->uio_iov, ktriov, iovlen);
848 len = auio->uio_resid;
849 so = (struct socket *)fp->f_data;
851 if (flags == NULL || (*flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
852 if (fp->f_flag & FNONBLOCK) {
854 *flags |= MSG_FNONBLOCKING;
856 lflags = MSG_FNONBLOCKING;
862 error = so_pru_soreceive(so, sa, auio, NULL, control, flags);
864 if (auio->uio_resid != len && (error == ERESTART ||
865 error == EINTR || error == EWOULDBLOCK))
869 if (ktriov != NULL) {
871 ktruio.uio_iov = ktriov;
872 ktruio.uio_resid = len - auio->uio_resid;
873 ktrgenio(td->td_lwp, s, UIO_READ, &ktruio, error);
875 FREE(ktriov, M_TEMP);
879 *res = len - auio->uio_resid;
885 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
886 * caddr_t from, int *fromlenaddr)
891 sys_recvfrom(struct recvfrom_args *uap)
893 struct thread *td = curthread;
896 struct sockaddr *sa = NULL;
899 if (uap->from && uap->fromlenaddr) {
900 error = copyin(uap->fromlenaddr, &fromlen, sizeof(fromlen));
908 aiov.iov_base = uap->buf;
909 aiov.iov_len = uap->len;
910 auio.uio_iov = &aiov;
913 auio.uio_resid = uap->len;
914 auio.uio_segflg = UIO_USERSPACE;
915 auio.uio_rw = UIO_READ;
918 error = kern_recvmsg(uap->s, uap->from ? &sa : NULL, &auio, NULL,
919 &uap->flags, &uap->sysmsg_szresult);
921 if (error == 0 && uap->from) {
922 /* note: sa may still be NULL */
924 fromlen = MIN(fromlen, sa->sa_len);
925 error = copyout(sa, uap->from, fromlen);
930 error = copyout(&fromlen, uap->fromlenaddr,
941 * recvmsg_args(int s, struct msghdr *msg, int flags)
946 sys_recvmsg(struct recvmsg_args *uap)
948 struct thread *td = curthread;
951 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
952 struct mbuf *m, *control = NULL;
953 struct sockaddr *sa = NULL;
955 socklen_t *ufromlenp, *ucontrollenp;
956 int error, fromlen, controllen, len, flags, *uflagsp;
959 * This copyin handles everything except the iovec.
961 error = copyin(uap->msg, &msg, sizeof(msg));
965 if (msg.msg_name && msg.msg_namelen < 0)
967 if (msg.msg_control && msg.msg_controllen < 0)
970 ufromlenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
972 ucontrollenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
974 uflagsp = (int *)((caddr_t)uap->msg + offsetof(struct msghdr,
980 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
985 auio.uio_iovcnt = msg.msg_iovlen;
987 auio.uio_segflg = UIO_USERSPACE;
988 auio.uio_rw = UIO_READ;
993 error = kern_recvmsg(uap->s,
994 (msg.msg_name ? &sa : NULL), &auio,
995 (msg.msg_control ? &control : NULL), &flags,
996 &uap->sysmsg_szresult);
999 * Conditionally copyout the name and populate the namelen field.
1001 if (error == 0 && msg.msg_name) {
1002 /* note: sa may still be NULL */
1004 fromlen = MIN(msg.msg_namelen, sa->sa_len);
1005 error = copyout(sa, msg.msg_name, fromlen);
1010 error = copyout(&fromlen, ufromlenp,
1011 sizeof(*ufromlenp));
1015 * Copyout msg.msg_control and msg.msg_controllen.
1017 if (error == 0 && msg.msg_control) {
1018 len = msg.msg_controllen;
1020 ctlbuf = (caddr_t)msg.msg_control;
1022 while(m && len > 0) {
1023 unsigned int tocopy;
1025 if (len >= m->m_len) {
1028 msg.msg_flags |= MSG_CTRUNC;
1032 error = copyout(mtod(m, caddr_t), ctlbuf, tocopy);
1040 controllen = ctlbuf - (caddr_t)msg.msg_control;
1041 error = copyout(&controllen, ucontrollenp,
1042 sizeof(*ucontrollenp));
1046 error = copyout(&flags, uflagsp, sizeof(*uflagsp));
1051 iovec_free(&iov, aiov);
1058 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1059 * in kernel pointer instead of a userland pointer. This allows us
1060 * to manipulate socket options in the emulation code.
1063 kern_setsockopt(int s, struct sockopt *sopt)
1065 struct thread *td = curthread;
1066 struct proc *p = td->td_proc;
1070 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1072 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1074 if (sopt->sopt_valsize < 0)
1077 error = holdsock(p->p_fd, s, &fp);
1081 error = sosetopt((struct socket *)fp->f_data, sopt);
1087 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
1092 sys_setsockopt(struct setsockopt_args *uap)
1094 struct thread *td = curthread;
1095 struct sockopt sopt;
1098 sopt.sopt_level = uap->level;
1099 sopt.sopt_name = uap->name;
1100 sopt.sopt_valsize = uap->valsize;
1102 sopt.sopt_val = NULL;
1104 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1107 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1108 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1113 error = kern_setsockopt(uap->s, &sopt);
1116 kfree(sopt.sopt_val, M_TEMP);
1121 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1122 * in kernel pointer instead of a userland pointer. This allows us
1123 * to manipulate socket options in the emulation code.
1126 kern_getsockopt(int s, struct sockopt *sopt)
1128 struct thread *td = curthread;
1129 struct proc *p = td->td_proc;
1133 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1135 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1137 if (sopt->sopt_valsize < 0 || sopt->sopt_valsize > SOMAXOPT_SIZE)
1140 error = holdsock(p->p_fd, s, &fp);
1144 error = sogetopt((struct socket *)fp->f_data, sopt);
1150 * getsockopt_args(int s, int level, int name, caddr_t val, int *avalsize)
1155 sys_getsockopt(struct getsockopt_args *uap)
1157 struct thread *td = curthread;
1158 struct sockopt sopt;
1162 error = copyin(uap->avalsize, &valsize, sizeof(valsize));
1169 sopt.sopt_level = uap->level;
1170 sopt.sopt_name = uap->name;
1171 sopt.sopt_valsize = valsize;
1173 sopt.sopt_val = NULL;
1175 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1178 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1179 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1184 error = kern_getsockopt(uap->s, &sopt);
1187 valsize = sopt.sopt_valsize;
1188 error = copyout(&valsize, uap->avalsize, sizeof(valsize));
1192 error = copyout(sopt.sopt_val, uap->val, sopt.sopt_valsize);
1195 kfree(sopt.sopt_val, M_TEMP);
1200 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1201 * This allows kern_getsockname() to return a pointer to an allocated struct
1202 * sockaddr which must be freed later with FREE(). The caller must
1203 * initialize *name to NULL.
1206 kern_getsockname(int s, struct sockaddr **name, int *namelen)
1208 struct thread *td = curthread;
1209 struct proc *p = td->td_proc;
1212 struct sockaddr *sa = NULL;
1215 error = holdsock(p->p_fd, s, &fp);
1222 so = (struct socket *)fp->f_data;
1223 error = so_pru_sockaddr(so, &sa);
1228 *namelen = MIN(*namelen, sa->sa_len);
1238 * getsockname_args(int fdes, caddr_t asa, int *alen)
1245 sys_getsockname(struct getsockname_args *uap)
1247 struct sockaddr *sa = NULL;
1250 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1254 error = kern_getsockname(uap->fdes, &sa, &sa_len);
1257 error = copyout(sa, uap->asa, sa_len);
1259 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1266 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1267 * This allows kern_getpeername() to return a pointer to an allocated struct
1268 * sockaddr which must be freed later with FREE(). The caller must
1269 * initialize *name to NULL.
1272 kern_getpeername(int s, struct sockaddr **name, int *namelen)
1274 struct thread *td = curthread;
1275 struct proc *p = td->td_proc;
1278 struct sockaddr *sa = NULL;
1281 error = holdsock(p->p_fd, s, &fp);
1288 so = (struct socket *)fp->f_data;
1289 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1293 error = so_pru_peeraddr(so, &sa);
1298 *namelen = MIN(*namelen, sa->sa_len);
1308 * getpeername_args(int fdes, caddr_t asa, int *alen)
1310 * Get name of peer for connected socket.
1315 sys_getpeername(struct getpeername_args *uap)
1317 struct sockaddr *sa = NULL;
1320 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1324 error = kern_getpeername(uap->fdes, &sa, &sa_len);
1327 error = copyout(sa, uap->asa, sa_len);
1329 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1336 getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len)
1338 struct sockaddr *sa;
1342 if (len > SOCK_MAXADDRLEN)
1343 return ENAMETOOLONG;
1344 if (len < offsetof(struct sockaddr, sa_data[0]))
1346 MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1347 error = copyin(uaddr, sa, len);
1351 #if BYTE_ORDER != BIG_ENDIAN
1353 * The bind(), connect(), and sendto() syscalls were not
1354 * versioned for COMPAT_43. Thus, this check must stay.
1356 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1357 sa->sa_family = sa->sa_len;
1366 * Detach a mapped page and release resources back to the system.
1367 * We must release our wiring and if the object is ripped out
1368 * from under the vm_page we become responsible for freeing the
1374 sf_buf_mfree(void *arg)
1376 struct sf_buf *sf = arg;
1379 m = sf_buf_page(sf);
1380 if (sf_buf_free(sf)) {
1381 /* sf invalid now */
1382 vm_page_busy_wait(m, FALSE, "sockpgf");
1383 vm_page_unwire(m, 0);
1385 if (m->wire_count == 0 && m->object == NULL)
1386 vm_page_try_to_free(m);
1392 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1393 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1395 * Send a file specified by 'fd' and starting at 'offset' to a socket
1396 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1397 * nbytes == 0. Optionally add a header and/or trailer to the socket
1398 * output. If specified, write the total number of bytes sent into *sbytes.
1400 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1401 * the headers to count against the remaining bytes to be sent from
1402 * the file descriptor. We may wish to implement a compatibility syscall
1408 sys_sendfile(struct sendfile_args *uap)
1410 struct thread *td = curthread;
1411 struct proc *p = td->td_proc;
1413 struct vnode *vp = NULL;
1414 struct sf_hdtr hdtr;
1415 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
1417 struct mbuf *mheader = NULL;
1420 off_t hdtr_size = 0;
1427 * Do argument checking. Must be a regular file in, stream
1428 * type and connected socket out, positive offset.
1430 fp = holdfp(p->p_fd, uap->fd, FREAD);
1434 if (fp->f_type != DTYPE_VNODE) {
1438 vp = (struct vnode *)fp->f_data;
1443 * If specified, get the pointer to the sf_hdtr struct for
1444 * any headers/trailers.
1447 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1454 error = iovec_copyin(hdtr.headers, &iov, aiov,
1455 hdtr.hdr_cnt, &hbytes);
1459 auio.uio_iovcnt = hdtr.hdr_cnt;
1460 auio.uio_offset = 0;
1461 auio.uio_segflg = UIO_USERSPACE;
1462 auio.uio_rw = UIO_WRITE;
1464 auio.uio_resid = hbytes;
1466 mheader = m_uiomove(&auio);
1468 iovec_free(&iov, aiov);
1469 if (mheader == NULL)
1474 error = kern_sendfile(vp, uap->s, uap->offset, uap->nbytes, mheader,
1475 &sbytes, uap->flags);
1480 * Send trailers. Wimp out and use writev(2).
1482 if (uap->hdtr != NULL && hdtr.trailers != NULL) {
1483 error = iovec_copyin(hdtr.trailers, &iov, aiov,
1484 hdtr.trl_cnt, &auio.uio_resid);
1488 auio.uio_iovcnt = hdtr.trl_cnt;
1489 auio.uio_offset = 0;
1490 auio.uio_segflg = UIO_USERSPACE;
1491 auio.uio_rw = UIO_WRITE;
1494 error = kern_sendmsg(uap->s, NULL, &auio, NULL, 0, &tbytes);
1496 iovec_free(&iov, aiov);
1499 hdtr_size += tbytes; /* trailer bytes successfully sent */
1505 if (uap->sbytes != NULL) {
1506 sbytes += hdtr_size;
1507 copyout(&sbytes, uap->sbytes, sizeof(off_t));
1513 kern_sendfile(struct vnode *vp, int sfd, off_t offset, size_t nbytes,
1514 struct mbuf *mheader, off_t *sbytes, int flags)
1516 struct thread *td = curthread;
1517 struct proc *p = td->td_proc;
1518 struct vm_object *obj;
1528 if (vp->v_type != VREG) {
1532 if ((obj = vp->v_object) == NULL) {
1536 error = holdsock(p->p_fd, sfd, &fp);
1539 so = (struct socket *)fp->f_data;
1540 if (so->so_type != SOCK_STREAM) {
1544 if ((so->so_state & SS_ISCONNECTED) == 0) {
1555 * Protect against multiple writers to the socket.
1557 ssb_lock(&so->so_snd, M_WAITOK);
1560 * Loop through the pages in the file, starting with the requested
1561 * offset. Get a file page (do I/O if necessary), map the file page
1562 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1565 for (off = offset; ; off += xfsize, *sbytes += xfsize + hbytes) {
1569 pindex = OFF_TO_IDX(off);
1572 * Calculate the amount to transfer. Not to exceed a page,
1573 * the EOF, or the passed in nbytes.
1575 xfsize = vp->v_filesize - off;
1576 if (xfsize > PAGE_SIZE)
1578 pgoff = (vm_offset_t)(off & PAGE_MASK);
1579 if (PAGE_SIZE - pgoff < xfsize)
1580 xfsize = PAGE_SIZE - pgoff;
1581 if (nbytes && xfsize > (nbytes - *sbytes))
1582 xfsize = nbytes - *sbytes;
1586 * Optimize the non-blocking case by looking at the socket space
1587 * before going to the extra work of constituting the sf_buf.
1589 if ((fp->f_flag & FNONBLOCK) && ssb_space(&so->so_snd) <= 0) {
1590 if (so->so_state & SS_CANTSENDMORE)
1594 ssb_unlock(&so->so_snd);
1598 * Attempt to look up the page.
1600 * Allocate if not found, wait and loop if busy, then
1601 * wire the page. critical section protection is
1602 * required to maintain the object association (an
1603 * interrupt can free the page) through to the
1604 * vm_page_wire() call.
1606 vm_object_hold(obj);
1607 pg = vm_page_lookup_busy_try(obj, pindex, TRUE, &error);
1609 vm_page_sleep_busy(pg, TRUE, "sfpbsy");
1610 vm_object_drop(obj);
1614 pg = vm_page_alloc(obj, pindex, VM_ALLOC_NORMAL |
1618 vm_object_drop(obj);
1623 vm_object_drop(obj);
1626 * If page is not valid for what we need, initiate I/O
1629 if (!pg->valid || !vm_page_is_valid(pg, pgoff, xfsize)) {
1635 * Ensure that our page is still around when the I/O
1638 vm_page_io_start(pg);
1642 * Get the page from backing store.
1644 bsize = vp->v_mount->mnt_stat.f_iosize;
1645 auio.uio_iov = &aiov;
1646 auio.uio_iovcnt = 1;
1648 aiov.iov_len = MAXBSIZE;
1649 auio.uio_resid = MAXBSIZE;
1650 auio.uio_offset = trunc_page(off);
1651 auio.uio_segflg = UIO_NOCOPY;
1652 auio.uio_rw = UIO_READ;
1654 vn_lock(vp, LK_SHARED | LK_RETRY);
1655 error = VOP_READ(vp, &auio,
1656 IO_VMIO | ((MAXBSIZE / bsize) << 16),
1659 vm_page_flag_clear(pg, PG_ZERO);
1660 vm_page_busy_wait(pg, FALSE, "sockpg");
1661 vm_page_io_finish(pg);
1663 vm_page_unwire(pg, 0);
1665 vm_page_try_to_free(pg);
1666 ssb_unlock(&so->so_snd);
1673 * Get a sendfile buf. We usually wait as long as necessary,
1674 * but this wait can be interrupted.
1676 if ((sf = sf_buf_alloc(pg)) == NULL) {
1677 vm_page_unwire(pg, 0);
1679 vm_page_try_to_free(pg);
1680 ssb_unlock(&so->so_snd);
1687 * Get an mbuf header and set it up as having external storage.
1689 MGETHDR(m, MB_WAIT, MT_DATA);
1693 ssb_unlock(&so->so_snd);
1697 m->m_ext.ext_free = sf_buf_mfree;
1698 m->m_ext.ext_ref = sf_buf_ref;
1699 m->m_ext.ext_arg = sf;
1700 m->m_ext.ext_buf = (void *)sf_buf_kva(sf);
1701 m->m_ext.ext_size = PAGE_SIZE;
1702 m->m_data = (char *)sf_buf_kva(sf) + pgoff;
1703 m->m_flags |= M_EXT;
1704 m->m_pkthdr.len = m->m_len = xfsize;
1705 KKASSERT((m->m_flags & (M_EXT_CLUSTER)) == 0);
1707 if (mheader != NULL) {
1708 hbytes = mheader->m_pkthdr.len;
1709 mheader->m_pkthdr.len += m->m_pkthdr.len;
1717 * Add the buffer to the socket buffer chain.
1722 * Make sure that the socket is still able to take more data.
1723 * CANTSENDMORE being true usually means that the connection
1724 * was closed. so_error is true when an error was sensed after
1726 * The state is checked after the page mapping and buffer
1727 * allocation above since those operations may block and make
1728 * any socket checks stale. From this point forward, nothing
1729 * blocks before the pru_send (or more accurately, any blocking
1730 * results in a loop back to here to re-check).
1732 if ((so->so_state & SS_CANTSENDMORE) || so->so_error) {
1733 if (so->so_state & SS_CANTSENDMORE) {
1736 error = so->so_error;
1740 ssb_unlock(&so->so_snd);
1745 * Wait for socket space to become available. We do this just
1746 * after checking the connection state above in order to avoid
1747 * a race condition with ssb_wait().
1749 if (ssb_space(&so->so_snd) < so->so_snd.ssb_lowat) {
1750 if (fp->f_flag & FNONBLOCK) {
1752 ssb_unlock(&so->so_snd);
1757 error = ssb_wait(&so->so_snd);
1759 * An error from ssb_wait usually indicates that we've
1760 * been interrupted by a signal. If we've sent anything
1761 * then return bytes sent, otherwise return the error.
1765 ssb_unlock(&so->so_snd);
1771 error = so_pru_send(so, 0, m, NULL, NULL, td);
1774 ssb_unlock(&so->so_snd);
1778 if (mheader != NULL) {
1779 *sbytes += mheader->m_pkthdr.len;
1780 error = so_pru_send(so, 0, mheader, NULL, NULL, td);
1783 ssb_unlock(&so->so_snd);
1788 if (mheader != NULL)
1797 sys_sctp_peeloff(struct sctp_peeloff_args *uap)
1800 struct thread *td = curthread;
1801 struct filedesc *fdp = td->td_proc->p_fd;
1802 struct file *lfp = NULL;
1803 struct file *nfp = NULL;
1805 struct socket *head, *so;
1808 short fflag; /* type must match fp->f_flag */
1810 assoc_id = uap->name;
1811 error = holdsock(td->td_proc->p_fd, uap->sd, &lfp);
1816 head = (struct socket *)lfp->f_data;
1817 error = sctp_can_peel_off(head, assoc_id);
1823 * At this point we know we do have a assoc to pull
1824 * we proceed to get the fd setup. This may block
1828 fflag = lfp->f_flag;
1829 error = falloc(td->td_lwp, &nfp, &fd);
1832 * Probably ran out of file descriptors. Put the
1833 * unaccepted connection back onto the queue and
1834 * do another wakeup so some other process might
1835 * have a chance at it.
1840 uap->sysmsg_iresult = fd;
1842 so = sctp_get_peeloff(head, assoc_id, &error);
1845 * Either someone else peeled it off OR
1846 * we can't get a socket.
1850 soreference(so); /* reference needed */
1851 soclrstate(so, SS_NOFDREF | SS_COMP); /* when clearing NOFDREF */
1853 if (head->so_sigio != NULL)
1854 fsetown(fgetown(&head->so_sigio), &so->so_sigio);
1856 nfp->f_type = DTYPE_SOCKET;
1857 nfp->f_flag = fflag;
1858 nfp->f_ops = &socketops;
1863 * Assign the file pointer to the reserved descriptor, or clear
1864 * the reserved descriptor if an error occured.
1867 fsetfd(fdp, NULL, fd);
1869 fsetfd(fdp, nfp, fd);
1872 * Release explicitly held references before returning.