2 * Copyright (c) 1982, 1986, 1989, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
5 * sendfile(2) and related extensions:
6 * Copyright (c) 1998, David Greenman. All rights reserved.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * @(#)uipc_syscalls.c 8.4 (Berkeley) 2/21/94
37 * $FreeBSD: src/sys/kern/uipc_syscalls.c,v 1.65.2.17 2003/04/04 17:11:16 tegge Exp $
38 * $DragonFly: src/sys/kern/uipc_syscalls.c,v 1.40 2004/07/30 21:56:14 dillon Exp $
41 #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>
77 #include <sys/thread2.h>
78 #include <sys/msgport2.h>
81 * System call interface to the socket abstraction.
84 extern struct fileops socketops;
87 * socket_args(int domain, int type, int protocol)
90 kern_socket(int domain, int type, int protocol, int *res)
92 struct thread *td = curthread;
93 struct proc *p = td->td_proc;
102 error = falloc(p, &fp, &fd);
106 error = socreate(domain, &so, type, protocol, td);
108 if (fdp->fd_ofiles[fd] == fp) {
109 fdp->fd_ofiles[fd] = NULL;
113 fp->f_data = (caddr_t)so;
114 fp->f_flag = FREAD|FWRITE;
115 fp->f_ops = &socketops;
116 fp->f_type = DTYPE_SOCKET;
124 socket(struct socket_args *uap)
128 error = kern_socket(uap->domain, uap->type, uap->protocol,
129 &uap->sysmsg_result);
135 kern_bind(int s, struct sockaddr *sa)
137 struct thread *td = curthread;
138 struct proc *p = td->td_proc;
143 error = holdsock(p->p_fd, s, &fp);
146 error = sobind((struct socket *)fp->f_data, sa, td);
152 * bind_args(int s, caddr_t name, int namelen)
155 bind(struct bind_args *uap)
160 error = getsockaddr(&sa, uap->name, uap->namelen);
163 error = kern_bind(uap->s, sa);
170 kern_listen(int s, int backlog)
172 struct thread *td = curthread;
173 struct proc *p = td->td_proc;
178 error = holdsock(p->p_fd, s, &fp);
181 error = solisten((struct socket *)fp->f_data, backlog, td);
187 * listen_args(int s, int backlog)
190 listen(struct listen_args *uap)
194 error = kern_listen(uap->s, uap->backlog);
199 * Returns the accepted socket as well.
202 soaccept_predicate(struct netmsg *msg0)
204 struct netmsg_so_notify *msg = (struct netmsg_so_notify *)msg0;
205 struct socket *head = msg->nm_so;
207 if (head->so_error != 0) {
208 msg->nm_lmsg.ms_error = head->so_error;
211 if (!TAILQ_EMPTY(&head->so_comp)) {
212 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
213 msg->nm_so = TAILQ_FIRST(&head->so_comp);
214 TAILQ_REMOVE(&head->so_comp, msg->nm_so, so_list);
217 msg->nm_lmsg.ms_error = 0;
220 if (head->so_state & SS_CANTRCVMORE) {
221 msg->nm_lmsg.ms_error = ECONNABORTED;
224 if (head->so_state & SS_NBIO) {
225 msg->nm_lmsg.ms_error = EWOULDBLOCK;
233 * The second argument to kern_accept() is a handle to a struct sockaddr.
234 * This allows kern_accept() to return a pointer to an allocated struct
235 * sockaddr which must be freed later with FREE(). The caller must
236 * initialize *name to NULL.
239 kern_accept(int s, struct sockaddr **name, int *namelen, int *res)
241 struct thread *td = curthread;
242 struct proc *p = td->td_proc;
243 struct filedesc *fdp = p->p_fd;
244 struct file *lfp = NULL;
245 struct file *nfp = NULL;
247 struct socket *head, *so;
248 struct netmsg_so_notify msg;
251 u_int fflag; /* type must match fp->f_flag */
254 if (name && namelen && *namelen < 0)
257 error = holdsock(fdp, s, &lfp);
261 error = falloc(p, &nfp, &fd);
262 if (error) { /* Probably ran out of file descriptors. */
270 head = (struct socket *)lfp->f_data;
271 if ((head->so_options & SO_ACCEPTCONN) == 0) {
276 /* optimize for uniprocessor case later XXX JH */
277 port = head->so_proto->pr_mport(head, NULL, PRU_PRED);
278 lwkt_initmsg(&msg.nm_lmsg, &curthread->td_msgport,
279 MSGF_PCATCH | MSGF_ABORTABLE,
280 lwkt_cmd_func(netmsg_so_notify),
281 lwkt_cmd_func(netmsg_so_notify_abort));
282 msg.nm_predicate = soaccept_predicate;
284 msg.nm_etype = NM_REVENT;
285 error = lwkt_domsg(port, &msg.nm_lmsg);
290 * At this point we have the connection that's ready to be accepted.
296 /* connection has been removed from the listen queue */
297 KNOTE(&head->so_rcv.sb_sel.si_note, 0);
299 so->so_state &= ~SS_COMP;
301 if (head->so_sigio != NULL)
302 fsetown(fgetown(head->so_sigio), &so->so_sigio);
304 nfp->f_data = (caddr_t)so;
306 nfp->f_ops = &socketops;
307 nfp->f_type = DTYPE_SOCKET;
308 /* Sync socket nonblocking/async state with file flags */
309 tmp = fflag & FNONBLOCK;
310 (void) fo_ioctl(nfp, FIONBIO, (caddr_t)&tmp, td);
311 tmp = fflag & FASYNC;
312 (void) fo_ioctl(nfp, FIOASYNC, (caddr_t)&tmp, td);
315 error = soaccept(so, &sa);
318 * Set the returned name and namelen as applicable. Set the returned
319 * namelen to 0 for older code which might ignore the return value
323 if (sa && name && namelen) {
324 if (*namelen > sa->sa_len)
325 *namelen = sa->sa_len;
335 * close the new descriptor, assuming someone hasn't ripped it
336 * out from under us. Note that *res is normally ignored if an
337 * error is returned but a syscall message will still have access
338 * to the result code.
342 if (fdp->fd_ofiles[fd] == nfp) {
343 fdp->fd_ofiles[fd] = NULL;
349 * Release explicitly held references before returning.
358 * accept_args(int s, caddr_t name, int *anamelen)
361 accept(struct accept_args *uap)
363 struct sockaddr *sa = NULL;
368 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
372 error = kern_accept(uap->s, &sa, &sa_len, &uap->sysmsg_result);
375 error = copyout(sa, uap->name, sa_len);
377 error = copyout(&sa_len, uap->anamelen,
378 sizeof(*uap->anamelen));
383 error = kern_accept(uap->s, NULL, 0, &uap->sysmsg_result);
389 * Returns TRUE if predicate satisfied.
392 soconnected_predicate(struct netmsg *msg0)
394 struct netmsg_so_notify *msg = (struct netmsg_so_notify *)msg0;
395 struct socket *so = msg->nm_so;
397 /* check predicate */
398 if (!(so->so_state & SS_ISCONNECTING) || so->so_error != 0) {
399 msg->nm_lmsg.ms_error = so->so_error;
407 kern_connect(int s, struct sockaddr *sa)
409 struct thread *td = curthread;
410 struct proc *p = td->td_proc;
415 error = holdsock(p->p_fd, s, &fp);
418 so = (struct socket *)fp->f_data;
419 if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
423 error = soconnect(so, sa, td);
426 if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
430 if ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
431 struct netmsg_so_notify msg;
434 port = so->so_proto->pr_mport(so, sa, PRU_PRED);
435 lwkt_initmsg(&msg.nm_lmsg,
436 &curthread->td_msgport,
437 MSGF_PCATCH | MSGF_ABORTABLE,
438 lwkt_cmd_func(netmsg_so_notify),
439 lwkt_cmd_func(netmsg_so_notify_abort));
440 msg.nm_predicate = soconnected_predicate;
442 msg.nm_etype = NM_REVENT;
443 error = lwkt_domsg(port, &msg.nm_lmsg);
446 error = so->so_error;
450 so->so_state &= ~SS_ISCONNECTING;
451 if (error == ERESTART)
459 * connect_args(int s, caddr_t name, int namelen)
462 connect(struct connect_args *uap)
467 error = getsockaddr(&sa, uap->name, uap->namelen);
470 error = kern_connect(uap->s, sa);
477 kern_socketpair(int domain, int type, int protocol, int *sv)
479 struct thread *td = curthread;
480 struct proc *p = td->td_proc;
481 struct filedesc *fdp;
482 struct file *fp1, *fp2;
483 struct socket *so1, *so2;
488 error = socreate(domain, &so1, type, protocol, td);
491 error = socreate(domain, &so2, type, protocol, td);
494 error = falloc(p, &fp1, &fd);
499 fp1->f_data = (caddr_t)so1;
500 error = falloc(p, &fp2, &fd);
504 fp2->f_data = (caddr_t)so2;
506 error = soconnect2(so1, so2);
509 if (type == SOCK_DGRAM) {
511 * Datagram socket connection is asymmetric.
513 error = soconnect2(so2, so1);
517 fp1->f_flag = fp2->f_flag = FREAD|FWRITE;
518 fp1->f_ops = fp2->f_ops = &socketops;
519 fp1->f_type = fp2->f_type = DTYPE_SOCKET;
524 if (fdp->fd_ofiles[sv[1]] == fp2) {
525 fdp->fd_ofiles[sv[1]] = NULL;
530 if (fdp->fd_ofiles[sv[0]] == fp1) {
531 fdp->fd_ofiles[sv[0]] = NULL;
543 * socketpair(int domain, int type, int protocol, int *rsv)
546 socketpair(struct socketpair_args *uap)
550 error = kern_socketpair(uap->domain, uap->type, uap->protocol, sockv);
553 error = copyout(sockv, uap->rsv, sizeof(sockv));
558 kern_sendmsg(int s, struct sockaddr *sa, struct uio *auio,
559 struct mbuf *control, int flags, int *res)
561 struct thread *td = curthread;
562 struct proc *p = td->td_proc;
567 struct iovec *ktriov = NULL;
571 error = holdsock(p->p_fd, s, &fp);
574 if (auio->uio_resid < 0) {
579 if (KTRPOINT(td, KTR_GENIO)) {
580 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
582 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
583 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
587 len = auio->uio_resid;
588 so = (struct socket *)fp->f_data;
589 error = so_pru_sosend(so, sa, auio, NULL, control, flags, td);
591 if (auio->uio_resid != len && (error == ERESTART ||
592 error == EINTR || error == EWOULDBLOCK))
598 if (ktriov != NULL) {
600 ktruio.uio_iov = ktriov;
601 ktruio.uio_resid = len - auio->uio_resid;
602 ktrgenio(p->p_tracep, s, UIO_WRITE, &ktruio, error);
604 FREE(ktriov, M_TEMP);
608 *res = len - auio->uio_resid;
615 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
618 sendto(struct sendto_args *uap)
620 struct thread *td = curthread;
623 struct sockaddr *sa = NULL;
627 error = getsockaddr(&sa, uap->to, uap->tolen);
631 aiov.iov_base = uap->buf;
632 aiov.iov_len = uap->len;
633 auio.uio_iov = &aiov;
636 auio.uio_resid = uap->len;
637 auio.uio_segflg = UIO_USERSPACE;
638 auio.uio_rw = UIO_WRITE;
641 error = kern_sendmsg(uap->s, sa, &auio, NULL, uap->flags,
642 &uap->sysmsg_result);
650 * sendmsg_args(int s, caddr_t msg, int flags)
653 sendmsg(struct sendmsg_args *uap)
655 struct thread *td = curthread;
658 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
659 struct sockaddr *sa = NULL;
660 struct mbuf *control = NULL;
663 error = copyin(uap->msg, (caddr_t)&msg, sizeof(msg));
668 * Conditionally copyin msg.msg_name.
671 error = getsockaddr(&sa, msg.msg_name, msg.msg_namelen);
679 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
684 auio.uio_iovcnt = msg.msg_iovlen;
686 auio.uio_segflg = UIO_USERSPACE;
687 auio.uio_rw = UIO_WRITE;
691 * Conditionally copyin msg.msg_control.
693 if (msg.msg_control) {
694 if (msg.msg_controllen < sizeof(struct cmsghdr) ||
695 msg.msg_controllen > MLEN) {
699 control = m_get(MB_WAIT, MT_CONTROL);
700 if (control == NULL) {
704 control->m_len = msg.msg_controllen;
705 error = copyin(msg.msg_control, mtod(control, caddr_t),
713 error = kern_sendmsg(uap->s, sa, &auio, control, uap->flags,
714 &uap->sysmsg_result);
719 iovec_free(&iov, aiov);
724 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
725 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
726 * Don't forget to FREE() and m_free() these if they are returned.
729 kern_recvmsg(int s, struct sockaddr **sa, struct uio *auio,
730 struct mbuf **control, int *flags, int *res)
732 struct thread *td = curthread;
733 struct proc *p = td->td_proc;
738 struct iovec *ktriov = NULL;
742 error = holdsock(p->p_fd, s, &fp);
745 if (auio->uio_resid < 0) {
750 if (KTRPOINT(td, KTR_GENIO)) {
751 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
753 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
754 bcopy(auio->uio_iov, ktriov, iovlen);
758 len = auio->uio_resid;
759 so = (struct socket *)fp->f_data;
760 error = so_pru_soreceive(so, sa, auio, NULL, control, flags);
762 if (auio->uio_resid != len && (error == ERESTART ||
763 error == EINTR || error == EWOULDBLOCK))
767 if (ktriov != NULL) {
769 ktruio.uio_iov = ktriov;
770 ktruio.uio_resid = len - auio->uio_resid;
771 ktrgenio(p->p_tracep, s, UIO_READ, &ktruio, error);
773 FREE(ktriov, M_TEMP);
777 *res = len - auio->uio_resid;
784 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
785 * caddr_t from, int *fromlenaddr)
788 recvfrom(struct recvfrom_args *uap)
790 struct thread *td = curthread;
793 struct sockaddr *sa = NULL;
796 if (uap->from && uap->fromlenaddr) {
797 error = copyin(uap->fromlenaddr, &fromlen, sizeof(fromlen));
805 aiov.iov_base = uap->buf;
806 aiov.iov_len = uap->len;
807 auio.uio_iov = &aiov;
810 auio.uio_resid = uap->len;
811 auio.uio_segflg = UIO_USERSPACE;
812 auio.uio_rw = UIO_READ;
815 error = kern_recvmsg(uap->s, uap->from ? &sa : NULL, &auio, NULL,
816 &uap->flags, &uap->sysmsg_result);
818 if (error == 0 && uap->from) {
819 /* note: sa may still be NULL */
821 fromlen = MIN(fromlen, sa->sa_len);
822 error = copyout(sa, uap->from, fromlen);
827 error = copyout(&fromlen, uap->fromlenaddr,
838 * recvmsg_args(int s, struct msghdr *msg, int flags)
841 recvmsg(struct recvmsg_args *uap)
843 struct thread *td = curthread;
846 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
847 struct mbuf *m, *control = NULL;
848 struct sockaddr *sa = NULL;
850 socklen_t *ufromlenp, *ucontrollenp;
851 int error, fromlen, controllen, len, flags, *uflagsp;
854 * This copyin handles everything except the iovec.
856 error = copyin(uap->msg, &msg, sizeof(msg));
860 if (msg.msg_name && msg.msg_namelen < 0)
862 if (msg.msg_control && msg.msg_controllen < 0)
865 ufromlenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
867 ucontrollenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
869 uflagsp = (int *)((caddr_t)uap->msg + offsetof(struct msghdr,
875 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
880 auio.uio_iovcnt = msg.msg_iovlen;
882 auio.uio_segflg = UIO_USERSPACE;
883 auio.uio_rw = UIO_READ;
886 flags = msg.msg_flags;
888 error = kern_recvmsg(uap->s, msg.msg_name ? &sa : NULL, &auio,
889 msg.msg_control ? &control : NULL, &flags, &uap->sysmsg_result);
892 * Conditionally copyout the name and populate the namelen field.
894 if (error == 0 && msg.msg_name) {
895 fromlen = MIN(msg.msg_namelen, sa->sa_len);
896 error = copyout(sa, msg.msg_name, fromlen);
898 error = copyout(&fromlen, ufromlenp,
903 * Copyout msg.msg_control and msg.msg_controllen.
905 if (error == 0 && msg.msg_control) {
906 len = msg.msg_controllen;
908 ctlbuf = (caddr_t)msg.msg_control;
910 while(m && len > 0) {
913 if (len >= m->m_len) {
916 msg.msg_flags |= MSG_CTRUNC;
920 error = copyout(mtod(m, caddr_t), ctlbuf, tocopy);
928 controllen = ctlbuf - (caddr_t)msg.msg_control;
929 error = copyout(&controllen, ucontrollenp,
930 sizeof(*ucontrollenp));
934 error = copyout(&flags, uflagsp, sizeof(*uflagsp));
939 iovec_free(&iov, aiov);
946 * shutdown_args(int s, int how)
949 kern_shutdown(int s, int how)
951 struct thread *td = curthread;
952 struct proc *p = td->td_proc;
957 error = holdsock(p->p_fd, s, &fp);
960 error = soshutdown((struct socket *)fp->f_data, how);
966 shutdown(struct shutdown_args *uap)
970 error = kern_shutdown(uap->s, uap->how);
976 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
977 * in kernel pointer instead of a userland pointer. This allows us
978 * to manipulate socket options in the emulation code.
981 kern_setsockopt(int s, struct sockopt *sopt)
983 struct thread *td = curthread;
984 struct proc *p = td->td_proc;
988 if (sopt->sopt_val == 0 && sopt->sopt_valsize != 0)
990 if (sopt->sopt_valsize < 0)
993 error = holdsock(p->p_fd, s, &fp);
997 error = sosetopt((struct socket *)fp->f_data, sopt);
1003 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
1006 setsockopt(struct setsockopt_args *uap)
1008 struct thread *td = curthread;
1009 struct sockopt sopt;
1012 sopt.sopt_dir = SOPT_SET;
1013 sopt.sopt_level = uap->level;
1014 sopt.sopt_name = uap->name;
1015 sopt.sopt_val = uap->val;
1016 sopt.sopt_valsize = uap->valsize;
1019 error = kern_setsockopt(uap->s, &sopt);
1024 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1025 * in kernel pointer instead of a userland pointer. This allows us
1026 * to manipulate socket options in the emulation code.
1029 kern_getsockopt(int s, struct sockopt *sopt)
1031 struct thread *td = curthread;
1032 struct proc *p = td->td_proc;
1036 if (sopt->sopt_val == 0 && sopt->sopt_valsize != 0)
1038 if (sopt->sopt_valsize < 0)
1041 error = holdsock(p->p_fd, s, &fp);
1045 error = sogetopt((struct socket *)fp->f_data, sopt);
1051 * getsockopt_Args(int s, int level, int name, caddr_t val, int *avalsize)
1054 getsockopt(struct getsockopt_args *uap)
1056 struct thread *td = curthread;
1057 struct sockopt sopt;
1061 error = copyin(uap->avalsize, &valsize, sizeof(valsize));
1070 sopt.sopt_dir = SOPT_GET;
1071 sopt.sopt_level = uap->level;
1072 sopt.sopt_name = uap->name;
1073 sopt.sopt_val = uap->val;
1074 sopt.sopt_valsize = valsize;
1077 error = kern_getsockopt(uap->s, &sopt);
1079 valsize = sopt.sopt_valsize;
1080 error = copyout(&valsize, uap->avalsize, sizeof(valsize));
1086 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1087 * This allows kern_getsockname() to return a pointer to an allocated struct
1088 * sockaddr which must be freed later with FREE(). The caller must
1089 * initialize *name to NULL.
1092 kern_getsockname(int s, struct sockaddr **name, int *namelen)
1094 struct thread *td = curthread;
1095 struct proc *p = td->td_proc;
1098 struct sockaddr *sa = NULL;
1101 error = holdsock(p->p_fd, s, &fp);
1108 so = (struct socket *)fp->f_data;
1109 error = so_pru_sockaddr(so, &sa);
1114 *namelen = MIN(*namelen, sa->sa_len);
1124 * getsockname_args(int fdes, caddr_t asa, int *alen)
1129 getsockname(struct getsockname_args *uap)
1131 struct sockaddr *sa = NULL;
1134 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1138 error = kern_getsockname(uap->fdes, &sa, &sa_len);
1141 error = copyout(sa, uap->asa, sa_len);
1143 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1150 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1151 * This allows kern_getpeername() to return a pointer to an allocated struct
1152 * sockaddr which must be freed later with FREE(). The caller must
1153 * initialize *name to NULL.
1156 kern_getpeername(int s, struct sockaddr **name, int *namelen)
1158 struct thread *td = curthread;
1159 struct proc *p = td->td_proc;
1162 struct sockaddr *sa = NULL;
1165 error = holdsock(p->p_fd, s, &fp);
1172 so = (struct socket *)fp->f_data;
1173 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1177 error = so_pru_peeraddr(so, &sa);
1182 *namelen = MIN(*namelen, sa->sa_len);
1192 * getpeername_args(int fdes, caddr_t asa, int *alen)
1194 * Get name of peer for connected socket.
1197 getpeername(struct getpeername_args *uap)
1199 struct sockaddr *sa = NULL;
1202 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1206 error = kern_getpeername(uap->fdes, &sa, &sa_len);
1209 error = copyout(sa, uap->asa, sa_len);
1211 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1218 getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len)
1220 struct sockaddr *sa;
1224 if (len > SOCK_MAXADDRLEN)
1225 return ENAMETOOLONG;
1226 if (len < offsetof(struct sockaddr, sa_data[0]))
1228 MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1229 error = copyin(uaddr, sa, len);
1233 #if BYTE_ORDER != BIG_ENDIAN
1235 * The bind(), connect(), and sendto() syscalls were not
1236 * versioned for COMPAT_43. Thus, this check must stay.
1238 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1239 sa->sa_family = sa->sa_len;
1248 * holdsock() - load the struct file pointer associated
1249 * with a socket into *fpp. If an error occurs, non-zero
1250 * will be returned and *fpp will be set to NULL.
1253 holdsock(fdp, fdes, fpp)
1254 struct filedesc *fdp;
1258 struct file *fp = NULL;
1261 if ((unsigned)fdes >= fdp->fd_nfiles ||
1262 (fp = fdp->fd_ofiles[fdes]) == NULL) {
1264 } else if (fp->f_type != DTYPE_SOCKET) {
1275 * Detach a mapped page and release resources back to the system.
1276 * We must release our wiring and if the object is ripped out
1277 * from under the vm_page we become responsible for freeing the
1280 * XXX HACK XXX TEMPORARY UNTIL WE IMPLEMENT EXT MBUF REFERENCE COUNTING
1283 sf_buf_mref(void *arg)
1285 struct sf_buf *sf = arg;
1290 sf_buf_mfree(void *arg)
1292 struct sf_buf *sf = arg;
1297 KKASSERT(sf->aux2 > 0);
1298 if (--sf->aux2 == 0) {
1299 m = sf_buf_page(sf);
1306 vm_page_unwire(m, 0);
1308 if (m->wire_count == 0 && m->object == NULL)
1316 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1317 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1319 * Send a file specified by 'fd' and starting at 'offset' to a socket
1320 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1321 * nbytes == 0. Optionally add a header and/or trailer to the socket
1322 * output. If specified, write the total number of bytes sent into *sbytes.
1324 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1325 * the headers to count against the remaining bytes to be sent from
1326 * the file descriptor. We may wish to implement a compatibility syscall
1330 sendfile(struct sendfile_args *uap)
1332 struct thread *td = curthread;
1333 struct proc *p = td->td_proc;
1335 struct filedesc *fdp;
1336 struct vnode *vp = NULL;
1337 struct sf_hdtr hdtr;
1338 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
1340 struct mbuf *mheader = NULL;
1341 off_t hdtr_size = 0, sbytes;
1342 int error, hbytes = 0, tbytes;
1348 * Do argument checking. Must be a regular file in, stream
1349 * type and connected socket out, positive offset.
1351 fp = holdfp(fdp, uap->fd, FREAD);
1355 if (fp->f_type != DTYPE_VNODE) {
1359 vp = (struct vnode *)fp->f_data;
1364 * If specified, get the pointer to the sf_hdtr struct for
1365 * any headers/trailers.
1368 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1375 error = iovec_copyin(hdtr.headers, &iov, aiov,
1376 hdtr.hdr_cnt, &hbytes);
1380 auio.uio_iovcnt = hdtr.hdr_cnt;
1381 auio.uio_offset = 0;
1382 auio.uio_segflg = UIO_USERSPACE;
1383 auio.uio_rw = UIO_WRITE;
1385 auio.uio_resid = hbytes;
1387 mheader = m_uiomove(&auio, MB_WAIT, 0);
1389 iovec_free(&iov, aiov);
1390 if (mheader == NULL)
1395 error = kern_sendfile(vp, uap->s, uap->offset, uap->nbytes, mheader,
1396 &sbytes, uap->flags);
1401 * Send trailers. Wimp out and use writev(2).
1403 if (uap->hdtr != NULL && hdtr.trailers != NULL) {
1404 error = iovec_copyin(hdtr.trailers, &iov, aiov,
1405 hdtr.trl_cnt, &auio.uio_resid);
1409 auio.uio_iovcnt = hdtr.trl_cnt;
1410 auio.uio_offset = 0;
1411 auio.uio_segflg = UIO_USERSPACE;
1412 auio.uio_rw = UIO_WRITE;
1415 error = kern_sendmsg(uap->s, NULL, &auio, NULL, 0, &tbytes);
1417 iovec_free(&iov, aiov);
1420 hdtr_size += tbytes; /* trailer bytes successfully sent */
1424 if (uap->sbytes != NULL) {
1425 sbytes += hdtr_size;
1426 copyout(&sbytes, uap->sbytes, sizeof(off_t));
1434 kern_sendfile(struct vnode *vp, int sfd, off_t offset, size_t nbytes,
1435 struct mbuf *mheader, off_t *sbytes, int flags)
1437 struct thread *td = curthread;
1438 struct proc *p = td->td_proc;
1439 struct vm_object *obj;
1450 if (vp->v_type != VREG || VOP_GETVOBJECT(vp, &obj) != 0) {
1454 error = holdsock(p->p_fd, sfd, &fp);
1457 so = (struct socket *)fp->f_data;
1458 if (so->so_type != SOCK_STREAM) {
1462 if ((so->so_state & SS_ISCONNECTED) == 0) {
1473 * Protect against multiple writers to the socket.
1475 (void) sblock(&so->so_snd, M_WAITOK);
1478 * Loop through the pages in the file, starting with the requested
1479 * offset. Get a file page (do I/O if necessary), map the file page
1480 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1483 for (off = offset; ; off += xfsize, *sbytes += xfsize + hbytes) {
1487 pindex = OFF_TO_IDX(off);
1490 * Calculate the amount to transfer. Not to exceed a page,
1491 * the EOF, or the passed in nbytes.
1493 xfsize = obj->un_pager.vnp.vnp_size - off;
1494 if (xfsize > PAGE_SIZE)
1496 pgoff = (vm_offset_t)(off & PAGE_MASK);
1497 if (PAGE_SIZE - pgoff < xfsize)
1498 xfsize = PAGE_SIZE - pgoff;
1499 if (nbytes && xfsize > (nbytes - *sbytes))
1500 xfsize = nbytes - *sbytes;
1504 * Optimize the non-blocking case by looking at the socket space
1505 * before going to the extra work of constituting the sf_buf.
1507 if ((so->so_state & SS_NBIO) && sbspace(&so->so_snd) <= 0) {
1508 if (so->so_state & SS_CANTSENDMORE)
1512 sbunlock(&so->so_snd);
1516 * Attempt to look up the page.
1518 * Allocate if not found, wait and loop if busy, then
1519 * wire the page. splvm() protection is required to
1520 * maintain the object association (an interrupt can
1521 * free the page) through to the vm_page_wire() call.
1524 pg = vm_page_lookup(obj, pindex);
1526 pg = vm_page_alloc(obj, pindex, VM_ALLOC_NORMAL);
1533 } else if (vm_page_sleep_busy(pg, TRUE, "sfpbsy")) {
1541 * If page is not valid for what we need, initiate I/O
1544 if (!pg->valid || !vm_page_is_valid(pg, pgoff, xfsize)) {
1550 * Ensure that our page is still around when the I/O
1553 vm_page_io_start(pg);
1556 * Get the page from backing store.
1558 bsize = vp->v_mount->mnt_stat.f_iosize;
1559 auio.uio_iov = &aiov;
1560 auio.uio_iovcnt = 1;
1562 aiov.iov_len = MAXBSIZE;
1563 auio.uio_resid = MAXBSIZE;
1564 auio.uio_offset = trunc_page(off);
1565 auio.uio_segflg = UIO_NOCOPY;
1566 auio.uio_rw = UIO_READ;
1568 vn_lock(vp, NULL, LK_SHARED | LK_NOPAUSE | LK_RETRY, td);
1569 error = VOP_READ(vp, &auio,
1570 IO_VMIO | ((MAXBSIZE / bsize) << 16),
1572 VOP_UNLOCK(vp, NULL, 0, td);
1573 vm_page_flag_clear(pg, PG_ZERO);
1574 vm_page_io_finish(pg);
1576 vm_page_unwire(pg, 0);
1578 * See if anyone else might know about this page.
1579 * If not and it is not valid, then free it.
1581 if (pg->wire_count == 0 && pg->valid == 0 &&
1582 pg->busy == 0 && !(pg->flags & PG_BUSY) &&
1583 pg->hold_count == 0) {
1587 sbunlock(&so->so_snd);
1594 * Get a sendfile buf. We usually wait as long as necessary,
1595 * but this wait can be interrupted.
1597 if ((sf = sf_buf_alloc(pg, SFBA_PCATCH)) == NULL) {
1599 vm_page_unwire(pg, 0);
1600 if (pg->wire_count == 0 && pg->object == NULL)
1603 sbunlock(&so->so_snd);
1609 * Get an mbuf header and set it up as having external storage.
1611 MGETHDR(m, MB_WAIT, MT_DATA);
1615 sbunlock(&so->so_snd);
1618 ++sf->aux1; /* wiring count */
1619 ++sf->aux2; /* initial reference */
1620 m->m_ext.ext_nfree.new = sf_buf_mfree;
1621 m->m_ext.ext_nref.new = sf_buf_mref;
1622 m->m_ext.ext_arg = sf;
1623 m->m_ext.ext_buf = (void *)sf->kva;
1624 m->m_ext.ext_size = PAGE_SIZE;
1625 m->m_data = (char *) sf->kva + pgoff;
1626 m->m_flags |= M_EXT;
1627 m->m_pkthdr.len = m->m_len = xfsize;
1629 if (mheader != NULL) {
1630 hbytes = mheader->m_pkthdr.len;
1631 mheader->m_pkthdr.len += m->m_pkthdr.len;
1639 * Add the buffer to the socket buffer chain.
1644 * Make sure that the socket is still able to take more data.
1645 * CANTSENDMORE being true usually means that the connection
1646 * was closed. so_error is true when an error was sensed after
1648 * The state is checked after the page mapping and buffer
1649 * allocation above since those operations may block and make
1650 * any socket checks stale. From this point forward, nothing
1651 * blocks before the pru_send (or more accurately, any blocking
1652 * results in a loop back to here to re-check).
1654 if ((so->so_state & SS_CANTSENDMORE) || so->so_error) {
1655 if (so->so_state & SS_CANTSENDMORE) {
1658 error = so->so_error;
1662 sbunlock(&so->so_snd);
1667 * Wait for socket space to become available. We do this just
1668 * after checking the connection state above in order to avoid
1669 * a race condition with sbwait().
1671 if (sbspace(&so->so_snd) < so->so_snd.sb_lowat) {
1672 if (so->so_state & SS_NBIO) {
1674 sbunlock(&so->so_snd);
1679 error = sbwait(&so->so_snd);
1681 * An error from sbwait usually indicates that we've
1682 * been interrupted by a signal. If we've sent anything
1683 * then return bytes sent, otherwise return the error.
1687 sbunlock(&so->so_snd);
1693 error = so_pru_send(so, 0, m, NULL, NULL, td);
1696 sbunlock(&so->so_snd);
1700 if (mheader != NULL) {
1701 *sbytes += mheader->m_pkthdr.len;
1702 error = so_pru_send(so, 0, mheader, NULL, NULL, td);
1705 sbunlock(&so->so_snd);
1710 if (mheader != NULL)