2 * Copyright (c) 1982, 1986, 1989, 1990, 1993
3 * The Regents of the University of California. All rights reserved.
5 * sendfile(2) and related extensions:
6 * Copyright (c) 1998, David Greenman. All rights reserved.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * @(#)uipc_syscalls.c 8.4 (Berkeley) 2/21/94
37 * $FreeBSD: src/sys/kern/uipc_syscalls.c,v 1.65.2.17 2003/04/04 17:11:16 tegge Exp $
38 * $DragonFly: src/sys/kern/uipc_syscalls.c,v 1.92 2008/11/26 13:10:56 sephe Exp $
41 #include "opt_ktrace.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/sysproto.h>
48 #include <sys/malloc.h>
49 #include <sys/filedesc.h>
50 #include <sys/event.h>
52 #include <sys/fcntl.h>
54 #include <sys/filio.h>
55 #include <sys/kern_syscall.h>
57 #include <sys/protosw.h>
58 #include <sys/sfbuf.h>
59 #include <sys/socket.h>
60 #include <sys/socketvar.h>
61 #include <sys/socketops.h>
63 #include <sys/vnode.h>
65 #include <sys/mount.h>
67 #include <sys/ktrace.h>
70 #include <vm/vm_object.h>
71 #include <vm/vm_page.h>
72 #include <vm/vm_pageout.h>
73 #include <vm/vm_kern.h>
74 #include <vm/vm_extern.h>
75 #include <sys/file2.h>
76 #include <sys/signalvar.h>
77 #include <sys/serialize.h>
79 #include <sys/thread2.h>
80 #include <sys/msgport2.h>
81 #include <sys/socketvar2.h>
82 #include <sys/mplock2.h>
83 #include <net/netmsg2.h>
86 #include <netinet/sctp_peeloff.h>
90 * System call interface to the socket abstraction.
93 extern struct fileops socketops;
96 * socket_args(int domain, int type, int protocol)
99 kern_socket(int domain, int type, int protocol, int *res)
101 struct thread *td = curthread;
102 struct filedesc *fdp = td->td_proc->p_fd;
107 KKASSERT(td->td_lwp);
109 error = falloc(td->td_lwp, &fp, &fd);
112 error = socreate(domain, &so, type, protocol, td);
114 fsetfd(fdp, NULL, fd);
116 fp->f_type = DTYPE_SOCKET;
117 fp->f_flag = FREAD | FWRITE;
118 fp->f_ops = &socketops;
131 sys_socket(struct socket_args *uap)
136 error = kern_socket(uap->domain, uap->type, uap->protocol,
137 &uap->sysmsg_iresult);
144 kern_bind(int s, struct sockaddr *sa)
146 struct thread *td = curthread;
147 struct proc *p = td->td_proc;
152 error = holdsock(p->p_fd, s, &fp);
155 error = sobind((struct socket *)fp->f_data, sa, td);
161 * bind_args(int s, caddr_t name, int namelen)
166 sys_bind(struct bind_args *uap)
171 error = getsockaddr(&sa, uap->name, uap->namelen);
175 error = kern_bind(uap->s, sa);
183 kern_listen(int s, int backlog)
185 struct thread *td = curthread;
186 struct proc *p = td->td_proc;
191 error = holdsock(p->p_fd, s, &fp);
194 error = solisten((struct socket *)fp->f_data, backlog, td);
200 * listen_args(int s, int backlog)
205 sys_listen(struct listen_args *uap)
210 error = kern_listen(uap->s, uap->backlog);
216 * Returns the accepted socket as well.
218 * NOTE! The sockets sitting on so_comp/so_incomp might have 0 refs, the
219 * pool token is absolutely required to avoid a sofree() race,
220 * as well as to avoid tailq handling races.
223 soaccept_predicate(struct netmsg_so_notify *msg)
225 struct socket *head = msg->base.nm_so;
228 if (head->so_error != 0) {
229 msg->base.lmsg.ms_error = head->so_error;
232 lwkt_getpooltoken(head);
233 if (!TAILQ_EMPTY(&head->so_comp)) {
234 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
235 so = TAILQ_FIRST(&head->so_comp);
236 TAILQ_REMOVE(&head->so_comp, so, so_list);
238 soclrstate(so, SS_COMP);
242 lwkt_relpooltoken(head);
244 msg->base.lmsg.ms_error = 0;
245 msg->base.nm_so = so;
248 lwkt_relpooltoken(head);
249 if (head->so_state & SS_CANTRCVMORE) {
250 msg->base.lmsg.ms_error = ECONNABORTED;
253 if (msg->nm_fflags & FNONBLOCK) {
254 msg->base.lmsg.ms_error = EWOULDBLOCK;
262 * The second argument to kern_accept() is a handle to a struct sockaddr.
263 * This allows kern_accept() to return a pointer to an allocated struct
264 * sockaddr which must be freed later with FREE(). The caller must
265 * initialize *name to NULL.
268 kern_accept(int s, int fflags, struct sockaddr **name, int *namelen, int *res)
270 struct thread *td = curthread;
271 struct filedesc *fdp = td->td_proc->p_fd;
272 struct file *lfp = NULL;
273 struct file *nfp = NULL;
275 struct socket *head, *so;
276 struct netmsg_so_notify msg;
278 u_int fflag; /* type must match fp->f_flag */
282 if (name && namelen && *namelen < 0)
285 error = holdsock(td->td_proc->p_fd, s, &lfp);
289 error = falloc(td->td_lwp, &nfp, &fd);
290 if (error) { /* Probably ran out of file descriptors. */
294 head = (struct socket *)lfp->f_data;
295 if ((head->so_options & SO_ACCEPTCONN) == 0) {
300 if (fflags & O_FBLOCKING)
301 fflags |= lfp->f_flag & ~FNONBLOCK;
302 else if (fflags & O_FNONBLOCKING)
303 fflags |= lfp->f_flag | FNONBLOCK;
305 fflags = lfp->f_flag;
307 /* optimize for uniprocessor case later XXX JH */
308 netmsg_init_abortable(&msg.base, head, &curthread->td_msgport,
309 0, netmsg_so_notify, netmsg_so_notify_doabort);
310 msg.nm_predicate = soaccept_predicate;
311 msg.nm_fflags = fflags;
312 msg.nm_etype = NM_REVENT;
313 error = lwkt_domsg(head->so_port, &msg.base.lmsg, PCATCH);
318 * At this point we have the connection that's ready to be accepted.
320 * NOTE! soaccept_predicate() ref'd so for us, and soaccept() expects
321 * to eat the ref and turn it into a descriptor.
327 /* connection has been removed from the listen queue */
328 KNOTE(&head->so_rcv.ssb_kq.ki_note, 0);
330 if (head->so_sigio != NULL)
331 fsetown(fgetown(head->so_sigio), &so->so_sigio);
333 nfp->f_type = DTYPE_SOCKET;
335 nfp->f_ops = &socketops;
337 /* Sync socket nonblocking/async state with file flags */
338 tmp = fflag & FNONBLOCK;
339 fo_ioctl(nfp, FIONBIO, (caddr_t)&tmp, td->td_ucred, NULL);
340 tmp = fflag & FASYNC;
341 fo_ioctl(nfp, FIOASYNC, (caddr_t)&tmp, td->td_ucred, NULL);
344 error = soaccept(so, &sa);
347 * Set the returned name and namelen as applicable. Set the returned
348 * namelen to 0 for older code which might ignore the return value
352 if (sa && name && namelen) {
353 if (*namelen > sa->sa_len)
354 *namelen = sa->sa_len;
364 * If an error occured clear the reserved descriptor, else associate
367 * Note that *res is normally ignored if an error is returned but
368 * a syscall message will still have access to the result code.
371 fsetfd(fdp, NULL, fd);
374 fsetfd(fdp, nfp, fd);
382 * accept(int s, caddr_t name, int *anamelen)
387 sys_accept(struct accept_args *uap)
389 struct sockaddr *sa = NULL;
394 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
399 error = kern_accept(uap->s, 0, &sa, &sa_len,
400 &uap->sysmsg_iresult);
404 error = copyout(sa, uap->name, sa_len);
406 error = copyout(&sa_len, uap->anamelen,
407 sizeof(*uap->anamelen));
413 error = kern_accept(uap->s, 0, NULL, 0,
414 &uap->sysmsg_iresult);
421 * extaccept(int s, int fflags, caddr_t name, int *anamelen)
426 sys_extaccept(struct extaccept_args *uap)
428 struct sockaddr *sa = NULL;
431 int fflags = uap->flags & O_FMASK;
434 error = copyin(uap->anamelen, &sa_len, sizeof(sa_len));
439 error = kern_accept(uap->s, fflags, &sa, &sa_len,
440 &uap->sysmsg_iresult);
444 error = copyout(sa, uap->name, sa_len);
446 error = copyout(&sa_len, uap->anamelen,
447 sizeof(*uap->anamelen));
453 error = kern_accept(uap->s, fflags, NULL, 0,
454 &uap->sysmsg_iresult);
462 * Returns TRUE if predicate satisfied.
465 soconnected_predicate(struct netmsg_so_notify *msg)
467 struct socket *so = msg->base.nm_so;
469 /* check predicate */
470 if (!(so->so_state & SS_ISCONNECTING) || so->so_error != 0) {
471 msg->base.lmsg.ms_error = so->so_error;
479 kern_connect(int s, int fflags, struct sockaddr *sa)
481 struct thread *td = curthread;
482 struct proc *p = td->td_proc;
485 int error, interrupted = 0;
487 error = holdsock(p->p_fd, s, &fp);
490 so = (struct socket *)fp->f_data;
492 if (fflags & O_FBLOCKING)
493 /* fflags &= ~FNONBLOCK; */;
494 else if (fflags & O_FNONBLOCKING)
499 if (so->so_state & SS_ISCONNECTING) {
503 error = soconnect(so, sa, td);
506 if ((fflags & FNONBLOCK) && (so->so_state & SS_ISCONNECTING)) {
510 if ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
511 struct netmsg_so_notify msg;
513 netmsg_init_abortable(&msg.base, so,
514 &curthread->td_msgport,
517 netmsg_so_notify_doabort);
518 msg.nm_predicate = soconnected_predicate;
519 msg.nm_etype = NM_REVENT;
520 error = lwkt_domsg(so->so_port, &msg.base.lmsg, PCATCH);
521 if (error == EINTR || error == ERESTART)
525 error = so->so_error;
530 soclrstate(so, SS_ISCONNECTING);
531 if (error == ERESTART)
539 * connect_args(int s, caddr_t name, int namelen)
544 sys_connect(struct connect_args *uap)
549 error = getsockaddr(&sa, uap->name, uap->namelen);
553 error = kern_connect(uap->s, 0, sa);
561 * connect_args(int s, int fflags, caddr_t name, int namelen)
566 sys_extconnect(struct extconnect_args *uap)
570 int fflags = uap->flags & O_FMASK;
572 error = getsockaddr(&sa, uap->name, uap->namelen);
576 error = kern_connect(uap->s, fflags, sa);
584 kern_socketpair(int domain, int type, int protocol, int *sv)
586 struct thread *td = curthread;
587 struct filedesc *fdp;
588 struct file *fp1, *fp2;
589 struct socket *so1, *so2;
592 fdp = td->td_proc->p_fd;
593 error = socreate(domain, &so1, type, protocol, td);
596 error = socreate(domain, &so2, type, protocol, td);
599 error = falloc(td->td_lwp, &fp1, &fd1);
604 error = falloc(td->td_lwp, &fp2, &fd2);
609 error = soconnect2(so1, so2);
612 if (type == SOCK_DGRAM) {
614 * Datagram socket connection is asymmetric.
616 error = soconnect2(so2, so1);
620 fp1->f_type = fp2->f_type = DTYPE_SOCKET;
621 fp1->f_flag = fp2->f_flag = FREAD|FWRITE;
622 fp1->f_ops = fp2->f_ops = &socketops;
623 fsetfd(fdp, fp1, fd1);
624 fsetfd(fdp, fp2, fd2);
629 fsetfd(fdp, NULL, fd2);
632 fsetfd(fdp, NULL, fd1);
635 (void)soclose(so2, 0);
637 (void)soclose(so1, 0);
642 * socketpair(int domain, int type, int protocol, int *rsv)
647 sys_socketpair(struct socketpair_args *uap)
652 error = kern_socketpair(uap->domain, uap->type, uap->protocol, sockv);
656 error = copyout(sockv, uap->rsv, sizeof(sockv));
661 kern_sendmsg(int s, struct sockaddr *sa, struct uio *auio,
662 struct mbuf *control, int flags, size_t *res)
664 struct thread *td = curthread;
665 struct lwp *lp = td->td_lwp;
666 struct proc *p = td->td_proc;
672 struct iovec *ktriov = NULL;
676 error = holdsock(p->p_fd, s, &fp);
680 if (KTRPOINT(td, KTR_GENIO)) {
681 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
683 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
684 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen);
688 len = auio->uio_resid;
689 so = (struct socket *)fp->f_data;
690 if ((flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
691 if (fp->f_flag & FNONBLOCK)
692 flags |= MSG_FNONBLOCKING;
694 error = so_pru_sosend(so, sa, auio, NULL, control, flags, td);
696 if (auio->uio_resid != len && (error == ERESTART ||
697 error == EINTR || error == EWOULDBLOCK))
700 lwpsignal(p, lp, SIGPIPE);
703 if (ktriov != NULL) {
705 ktruio.uio_iov = ktriov;
706 ktruio.uio_resid = len - auio->uio_resid;
707 ktrgenio(lp, s, UIO_WRITE, &ktruio, error);
709 FREE(ktriov, M_TEMP);
713 *res = len - auio->uio_resid;
719 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
724 sys_sendto(struct sendto_args *uap)
726 struct thread *td = curthread;
729 struct sockaddr *sa = NULL;
733 error = getsockaddr(&sa, uap->to, uap->tolen);
737 aiov.iov_base = uap->buf;
738 aiov.iov_len = uap->len;
739 auio.uio_iov = &aiov;
742 auio.uio_resid = uap->len;
743 auio.uio_segflg = UIO_USERSPACE;
744 auio.uio_rw = UIO_WRITE;
748 error = kern_sendmsg(uap->s, sa, &auio, NULL, uap->flags,
749 &uap->sysmsg_szresult);
758 * sendmsg_args(int s, caddr_t msg, int flags)
763 sys_sendmsg(struct sendmsg_args *uap)
765 struct thread *td = curthread;
768 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
769 struct sockaddr *sa = NULL;
770 struct mbuf *control = NULL;
773 error = copyin(uap->msg, (caddr_t)&msg, sizeof(msg));
778 * Conditionally copyin msg.msg_name.
781 error = getsockaddr(&sa, msg.msg_name, msg.msg_namelen);
789 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
794 auio.uio_iovcnt = msg.msg_iovlen;
796 auio.uio_segflg = UIO_USERSPACE;
797 auio.uio_rw = UIO_WRITE;
801 * Conditionally copyin msg.msg_control.
803 if (msg.msg_control) {
804 if (msg.msg_controllen < sizeof(struct cmsghdr) ||
805 msg.msg_controllen > MLEN) {
809 control = m_get(MB_WAIT, MT_CONTROL);
810 if (control == NULL) {
814 control->m_len = msg.msg_controllen;
815 error = copyin(msg.msg_control, mtod(control, caddr_t),
824 error = kern_sendmsg(uap->s, sa, &auio, control, uap->flags,
825 &uap->sysmsg_szresult);
829 iovec_free(&iov, aiov);
837 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
838 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
839 * Don't forget to FREE() and m_free() these if they are returned.
842 kern_recvmsg(int s, struct sockaddr **sa, struct uio *auio,
843 struct mbuf **control, int *flags, size_t *res)
845 struct thread *td = curthread;
846 struct proc *p = td->td_proc;
853 struct iovec *ktriov = NULL;
857 error = holdsock(p->p_fd, s, &fp);
861 if (KTRPOINT(td, KTR_GENIO)) {
862 int iovlen = auio->uio_iovcnt * sizeof (struct iovec);
864 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
865 bcopy(auio->uio_iov, ktriov, iovlen);
869 len = auio->uio_resid;
870 so = (struct socket *)fp->f_data;
872 if (flags == NULL || (*flags & (MSG_FNONBLOCKING|MSG_FBLOCKING)) == 0) {
873 if (fp->f_flag & FNONBLOCK) {
875 *flags |= MSG_FNONBLOCKING;
877 lflags = MSG_FNONBLOCKING;
883 error = so_pru_soreceive(so, sa, auio, NULL, control, flags);
885 if (auio->uio_resid != len && (error == ERESTART ||
886 error == EINTR || error == EWOULDBLOCK))
890 if (ktriov != NULL) {
892 ktruio.uio_iov = ktriov;
893 ktruio.uio_resid = len - auio->uio_resid;
894 ktrgenio(td->td_lwp, s, UIO_READ, &ktruio, error);
896 FREE(ktriov, M_TEMP);
900 *res = len - auio->uio_resid;
906 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
907 * caddr_t from, int *fromlenaddr)
912 sys_recvfrom(struct recvfrom_args *uap)
914 struct thread *td = curthread;
917 struct sockaddr *sa = NULL;
920 if (uap->from && uap->fromlenaddr) {
921 error = copyin(uap->fromlenaddr, &fromlen, sizeof(fromlen));
929 aiov.iov_base = uap->buf;
930 aiov.iov_len = uap->len;
931 auio.uio_iov = &aiov;
934 auio.uio_resid = uap->len;
935 auio.uio_segflg = UIO_USERSPACE;
936 auio.uio_rw = UIO_READ;
940 error = kern_recvmsg(uap->s, uap->from ? &sa : NULL, &auio, NULL,
941 &uap->flags, &uap->sysmsg_szresult);
944 if (error == 0 && uap->from) {
945 /* note: sa may still be NULL */
947 fromlen = MIN(fromlen, sa->sa_len);
948 error = copyout(sa, uap->from, fromlen);
953 error = copyout(&fromlen, uap->fromlenaddr,
964 * recvmsg_args(int s, struct msghdr *msg, int flags)
969 sys_recvmsg(struct recvmsg_args *uap)
971 struct thread *td = curthread;
974 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
975 struct mbuf *m, *control = NULL;
976 struct sockaddr *sa = NULL;
978 socklen_t *ufromlenp, *ucontrollenp;
979 int error, fromlen, controllen, len, flags, *uflagsp;
982 * This copyin handles everything except the iovec.
984 error = copyin(uap->msg, &msg, sizeof(msg));
988 if (msg.msg_name && msg.msg_namelen < 0)
990 if (msg.msg_control && msg.msg_controllen < 0)
993 ufromlenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
995 ucontrollenp = (socklen_t *)((caddr_t)uap->msg + offsetof(struct msghdr,
997 uflagsp = (int *)((caddr_t)uap->msg + offsetof(struct msghdr,
1003 error = iovec_copyin(msg.msg_iov, &iov, aiov, msg.msg_iovlen,
1008 auio.uio_iovcnt = msg.msg_iovlen;
1009 auio.uio_offset = 0;
1010 auio.uio_segflg = UIO_USERSPACE;
1011 auio.uio_rw = UIO_READ;
1017 error = kern_recvmsg(uap->s,
1018 (msg.msg_name ? &sa : NULL), &auio,
1019 (msg.msg_control ? &control : NULL), &flags,
1020 &uap->sysmsg_szresult);
1024 * Conditionally copyout the name and populate the namelen field.
1026 if (error == 0 && msg.msg_name) {
1027 /* note: sa may still be NULL */
1029 fromlen = MIN(msg.msg_namelen, sa->sa_len);
1030 error = copyout(sa, msg.msg_name, fromlen);
1035 error = copyout(&fromlen, ufromlenp,
1036 sizeof(*ufromlenp));
1040 * Copyout msg.msg_control and msg.msg_controllen.
1042 if (error == 0 && msg.msg_control) {
1043 len = msg.msg_controllen;
1045 ctlbuf = (caddr_t)msg.msg_control;
1047 while(m && len > 0) {
1048 unsigned int tocopy;
1050 if (len >= m->m_len) {
1053 msg.msg_flags |= MSG_CTRUNC;
1057 error = copyout(mtod(m, caddr_t), ctlbuf, tocopy);
1065 controllen = ctlbuf - (caddr_t)msg.msg_control;
1066 error = copyout(&controllen, ucontrollenp,
1067 sizeof(*ucontrollenp));
1071 error = copyout(&flags, uflagsp, sizeof(*uflagsp));
1076 iovec_free(&iov, aiov);
1083 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1084 * in kernel pointer instead of a userland pointer. This allows us
1085 * to manipulate socket options in the emulation code.
1088 kern_setsockopt(int s, struct sockopt *sopt)
1090 struct thread *td = curthread;
1091 struct proc *p = td->td_proc;
1095 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1097 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1099 if (sopt->sopt_valsize < 0)
1102 error = holdsock(p->p_fd, s, &fp);
1106 error = sosetopt((struct socket *)fp->f_data, sopt);
1112 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
1117 sys_setsockopt(struct setsockopt_args *uap)
1119 struct thread *td = curthread;
1120 struct sockopt sopt;
1123 sopt.sopt_level = uap->level;
1124 sopt.sopt_name = uap->name;
1125 sopt.sopt_valsize = uap->valsize;
1127 sopt.sopt_val = NULL;
1129 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1132 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1133 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1139 error = kern_setsockopt(uap->s, &sopt);
1143 kfree(sopt.sopt_val, M_TEMP);
1148 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1149 * in kernel pointer instead of a userland pointer. This allows us
1150 * to manipulate socket options in the emulation code.
1153 kern_getsockopt(int s, struct sockopt *sopt)
1155 struct thread *td = curthread;
1156 struct proc *p = td->td_proc;
1160 if (sopt->sopt_val == NULL && sopt->sopt_valsize != 0)
1162 if (sopt->sopt_val != NULL && sopt->sopt_valsize == 0)
1164 if (sopt->sopt_valsize < 0 || sopt->sopt_valsize > SOMAXOPT_SIZE)
1167 error = holdsock(p->p_fd, s, &fp);
1171 error = sogetopt((struct socket *)fp->f_data, sopt);
1177 * getsockopt_args(int s, int level, int name, caddr_t val, int *avalsize)
1182 sys_getsockopt(struct getsockopt_args *uap)
1184 struct thread *td = curthread;
1185 struct sockopt sopt;
1189 error = copyin(uap->avalsize, &valsize, sizeof(valsize));
1196 sopt.sopt_level = uap->level;
1197 sopt.sopt_name = uap->name;
1198 sopt.sopt_valsize = valsize;
1200 sopt.sopt_val = NULL;
1202 if (sopt.sopt_valsize < 0 || sopt.sopt_valsize > SOMAXOPT_SIZE)
1205 sopt.sopt_val = kmalloc(sopt.sopt_valsize, M_TEMP, M_WAITOK);
1206 error = copyin(uap->val, sopt.sopt_val, sopt.sopt_valsize);
1212 error = kern_getsockopt(uap->s, &sopt);
1216 valsize = sopt.sopt_valsize;
1217 error = copyout(&valsize, uap->avalsize, sizeof(valsize));
1221 error = copyout(sopt.sopt_val, uap->val, sopt.sopt_valsize);
1224 kfree(sopt.sopt_val, M_TEMP);
1229 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1230 * This allows kern_getsockname() to return a pointer to an allocated struct
1231 * sockaddr which must be freed later with FREE(). The caller must
1232 * initialize *name to NULL.
1235 kern_getsockname(int s, struct sockaddr **name, int *namelen)
1237 struct thread *td = curthread;
1238 struct proc *p = td->td_proc;
1241 struct sockaddr *sa = NULL;
1244 error = holdsock(p->p_fd, s, &fp);
1251 so = (struct socket *)fp->f_data;
1252 error = so_pru_sockaddr(so, &sa);
1257 *namelen = MIN(*namelen, sa->sa_len);
1267 * getsockname_args(int fdes, caddr_t asa, int *alen)
1274 sys_getsockname(struct getsockname_args *uap)
1276 struct sockaddr *sa = NULL;
1279 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1284 error = kern_getsockname(uap->fdes, &sa, &sa_len);
1288 error = copyout(sa, uap->asa, sa_len);
1290 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1297 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1298 * This allows kern_getpeername() to return a pointer to an allocated struct
1299 * sockaddr which must be freed later with FREE(). The caller must
1300 * initialize *name to NULL.
1303 kern_getpeername(int s, struct sockaddr **name, int *namelen)
1305 struct thread *td = curthread;
1306 struct proc *p = td->td_proc;
1309 struct sockaddr *sa = NULL;
1312 error = holdsock(p->p_fd, s, &fp);
1319 so = (struct socket *)fp->f_data;
1320 if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1324 error = so_pru_peeraddr(so, &sa);
1329 *namelen = MIN(*namelen, sa->sa_len);
1339 * getpeername_args(int fdes, caddr_t asa, int *alen)
1341 * Get name of peer for connected socket.
1346 sys_getpeername(struct getpeername_args *uap)
1348 struct sockaddr *sa = NULL;
1351 error = copyin(uap->alen, &sa_len, sizeof(sa_len));
1356 error = kern_getpeername(uap->fdes, &sa, &sa_len);
1360 error = copyout(sa, uap->asa, sa_len);
1362 error = copyout(&sa_len, uap->alen, sizeof(*uap->alen));
1369 getsockaddr(struct sockaddr **namp, caddr_t uaddr, size_t len)
1371 struct sockaddr *sa;
1375 if (len > SOCK_MAXADDRLEN)
1376 return ENAMETOOLONG;
1377 if (len < offsetof(struct sockaddr, sa_data[0]))
1379 MALLOC(sa, struct sockaddr *, len, M_SONAME, M_WAITOK);
1380 error = copyin(uaddr, sa, len);
1384 #if BYTE_ORDER != BIG_ENDIAN
1386 * The bind(), connect(), and sendto() syscalls were not
1387 * versioned for COMPAT_43. Thus, this check must stay.
1389 if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1390 sa->sa_family = sa->sa_len;
1399 * Detach a mapped page and release resources back to the system.
1400 * We must release our wiring and if the object is ripped out
1401 * from under the vm_page we become responsible for freeing the
1402 * page. These routines must be MPSAFE.
1404 * XXX HACK XXX TEMPORARY UNTIL WE IMPLEMENT EXT MBUF REFERENCE COUNTING
1406 * XXX vm_page_*() routines are not MPSAFE yet, the MP lock is required.
1409 sf_buf_mfree(void *arg)
1411 struct sf_buf *sf = arg;
1415 * XXX vm_page_*() and SFBUF routines not MPSAFE yet.
1419 m = sf_buf_page(sf);
1420 if (sf_buf_free(sf) == 0) {
1421 vm_page_unwire(m, 0);
1422 if (m->wire_count == 0 && m->object == NULL)
1423 vm_page_try_to_free(m);
1431 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1432 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1434 * Send a file specified by 'fd' and starting at 'offset' to a socket
1435 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1436 * nbytes == 0. Optionally add a header and/or trailer to the socket
1437 * output. If specified, write the total number of bytes sent into *sbytes.
1439 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1440 * the headers to count against the remaining bytes to be sent from
1441 * the file descriptor. We may wish to implement a compatibility syscall
1447 sys_sendfile(struct sendfile_args *uap)
1449 struct thread *td = curthread;
1450 struct proc *p = td->td_proc;
1452 struct vnode *vp = NULL;
1453 struct sf_hdtr hdtr;
1454 struct iovec aiov[UIO_SMALLIOV], *iov = NULL;
1456 struct mbuf *mheader = NULL;
1459 off_t hdtr_size = 0;
1466 * Do argument checking. Must be a regular file in, stream
1467 * type and connected socket out, positive offset.
1469 fp = holdfp(p->p_fd, uap->fd, FREAD);
1473 if (fp->f_type != DTYPE_VNODE) {
1478 vp = (struct vnode *)fp->f_data;
1483 * If specified, get the pointer to the sf_hdtr struct for
1484 * any headers/trailers.
1487 error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
1494 error = iovec_copyin(hdtr.headers, &iov, aiov,
1495 hdtr.hdr_cnt, &hbytes);
1499 auio.uio_iovcnt = hdtr.hdr_cnt;
1500 auio.uio_offset = 0;
1501 auio.uio_segflg = UIO_USERSPACE;
1502 auio.uio_rw = UIO_WRITE;
1504 auio.uio_resid = hbytes;
1506 mheader = m_uiomove(&auio);
1508 iovec_free(&iov, aiov);
1509 if (mheader == NULL)
1514 error = kern_sendfile(vp, uap->s, uap->offset, uap->nbytes, mheader,
1515 &sbytes, uap->flags);
1520 * Send trailers. Wimp out and use writev(2).
1522 if (uap->hdtr != NULL && hdtr.trailers != NULL) {
1523 error = iovec_copyin(hdtr.trailers, &iov, aiov,
1524 hdtr.trl_cnt, &auio.uio_resid);
1528 auio.uio_iovcnt = hdtr.trl_cnt;
1529 auio.uio_offset = 0;
1530 auio.uio_segflg = UIO_USERSPACE;
1531 auio.uio_rw = UIO_WRITE;
1534 error = kern_sendmsg(uap->s, NULL, &auio, NULL, 0, &tbytes);
1536 iovec_free(&iov, aiov);
1539 hdtr_size += tbytes; /* trailer bytes successfully sent */
1546 if (uap->sbytes != NULL) {
1547 sbytes += hdtr_size;
1548 copyout(&sbytes, uap->sbytes, sizeof(off_t));
1554 kern_sendfile(struct vnode *vp, int sfd, off_t offset, size_t nbytes,
1555 struct mbuf *mheader, off_t *sbytes, int flags)
1557 struct thread *td = curthread;
1558 struct proc *p = td->td_proc;
1559 struct vm_object *obj;
1569 if (vp->v_type != VREG) {
1573 if ((obj = vp->v_object) == NULL) {
1577 error = holdsock(p->p_fd, sfd, &fp);
1580 so = (struct socket *)fp->f_data;
1581 if (so->so_type != SOCK_STREAM) {
1585 if ((so->so_state & SS_ISCONNECTED) == 0) {
1596 * Protect against multiple writers to the socket.
1598 ssb_lock(&so->so_snd, M_WAITOK);
1601 * Loop through the pages in the file, starting with the requested
1602 * offset. Get a file page (do I/O if necessary), map the file page
1603 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1606 for (off = offset; ; off += xfsize, *sbytes += xfsize + hbytes) {
1610 pindex = OFF_TO_IDX(off);
1613 * Calculate the amount to transfer. Not to exceed a page,
1614 * the EOF, or the passed in nbytes.
1616 xfsize = vp->v_filesize - off;
1617 if (xfsize > PAGE_SIZE)
1619 pgoff = (vm_offset_t)(off & PAGE_MASK);
1620 if (PAGE_SIZE - pgoff < xfsize)
1621 xfsize = PAGE_SIZE - pgoff;
1622 if (nbytes && xfsize > (nbytes - *sbytes))
1623 xfsize = nbytes - *sbytes;
1627 * Optimize the non-blocking case by looking at the socket space
1628 * before going to the extra work of constituting the sf_buf.
1630 if ((fp->f_flag & FNONBLOCK) && ssb_space(&so->so_snd) <= 0) {
1631 if (so->so_state & SS_CANTSENDMORE)
1635 ssb_unlock(&so->so_snd);
1639 * Attempt to look up the page.
1641 * Allocate if not found, wait and loop if busy, then
1642 * wire the page. critical section protection is
1643 * required to maintain the object association (an
1644 * interrupt can free the page) through to the
1645 * vm_page_wire() call.
1648 lwkt_gettoken(&vm_token);
1649 pg = vm_page_lookup(obj, pindex);
1651 pg = vm_page_alloc(obj, pindex, VM_ALLOC_NORMAL);
1654 lwkt_reltoken(&vm_token);
1659 } else if (vm_page_sleep_busy(pg, TRUE, "sfpbsy")) {
1660 lwkt_reltoken(&vm_token);
1665 lwkt_reltoken(&vm_token);
1669 * If page is not valid for what we need, initiate I/O
1672 if (!pg->valid || !vm_page_is_valid(pg, pgoff, xfsize)) {
1678 * Ensure that our page is still around when the I/O
1681 vm_page_io_start(pg);
1684 * Get the page from backing store.
1686 bsize = vp->v_mount->mnt_stat.f_iosize;
1687 auio.uio_iov = &aiov;
1688 auio.uio_iovcnt = 1;
1690 aiov.iov_len = MAXBSIZE;
1691 auio.uio_resid = MAXBSIZE;
1692 auio.uio_offset = trunc_page(off);
1693 auio.uio_segflg = UIO_NOCOPY;
1694 auio.uio_rw = UIO_READ;
1696 vn_lock(vp, LK_SHARED | LK_RETRY);
1697 error = VOP_READ(vp, &auio,
1698 IO_VMIO | ((MAXBSIZE / bsize) << 16),
1701 vm_page_flag_clear(pg, PG_ZERO);
1702 vm_page_io_finish(pg);
1705 vm_page_unwire(pg, 0);
1706 vm_page_try_to_free(pg);
1708 ssb_unlock(&so->so_snd);
1715 * Get a sendfile buf. We usually wait as long as necessary,
1716 * but this wait can be interrupted.
1718 if ((sf = sf_buf_alloc(pg)) == NULL) {
1720 vm_page_unwire(pg, 0);
1721 vm_page_try_to_free(pg);
1723 ssb_unlock(&so->so_snd);
1729 * Get an mbuf header and set it up as having external storage.
1731 MGETHDR(m, MB_WAIT, MT_DATA);
1735 ssb_unlock(&so->so_snd);
1739 m->m_ext.ext_free = sf_buf_mfree;
1740 m->m_ext.ext_ref = sf_buf_ref;
1741 m->m_ext.ext_arg = sf;
1742 m->m_ext.ext_buf = (void *)sf_buf_kva(sf);
1743 m->m_ext.ext_size = PAGE_SIZE;
1744 m->m_data = (char *)sf_buf_kva(sf) + pgoff;
1745 m->m_flags |= M_EXT;
1746 m->m_pkthdr.len = m->m_len = xfsize;
1747 KKASSERT((m->m_flags & (M_EXT_CLUSTER)) == 0);
1749 if (mheader != NULL) {
1750 hbytes = mheader->m_pkthdr.len;
1751 mheader->m_pkthdr.len += m->m_pkthdr.len;
1759 * Add the buffer to the socket buffer chain.
1764 * Make sure that the socket is still able to take more data.
1765 * CANTSENDMORE being true usually means that the connection
1766 * was closed. so_error is true when an error was sensed after
1768 * The state is checked after the page mapping and buffer
1769 * allocation above since those operations may block and make
1770 * any socket checks stale. From this point forward, nothing
1771 * blocks before the pru_send (or more accurately, any blocking
1772 * results in a loop back to here to re-check).
1774 if ((so->so_state & SS_CANTSENDMORE) || so->so_error) {
1775 if (so->so_state & SS_CANTSENDMORE) {
1778 error = so->so_error;
1782 ssb_unlock(&so->so_snd);
1787 * Wait for socket space to become available. We do this just
1788 * after checking the connection state above in order to avoid
1789 * a race condition with ssb_wait().
1791 if (ssb_space(&so->so_snd) < so->so_snd.ssb_lowat) {
1792 if (fp->f_flag & FNONBLOCK) {
1794 ssb_unlock(&so->so_snd);
1799 error = ssb_wait(&so->so_snd);
1801 * An error from ssb_wait usually indicates that we've
1802 * been interrupted by a signal. If we've sent anything
1803 * then return bytes sent, otherwise return the error.
1807 ssb_unlock(&so->so_snd);
1813 error = so_pru_send(so, 0, m, NULL, NULL, td);
1816 ssb_unlock(&so->so_snd);
1820 if (mheader != NULL) {
1821 *sbytes += mheader->m_pkthdr.len;
1822 error = so_pru_send(so, 0, mheader, NULL, NULL, td);
1825 ssb_unlock(&so->so_snd);
1830 if (mheader != NULL)
1839 sys_sctp_peeloff(struct sctp_peeloff_args *uap)
1842 struct thread *td = curthread;
1843 struct filedesc *fdp = td->td_proc->p_fd;
1844 struct file *lfp = NULL;
1845 struct file *nfp = NULL;
1847 struct socket *head, *so;
1850 short fflag; /* type must match fp->f_flag */
1852 assoc_id = uap->name;
1853 error = holdsock(td->td_proc->p_fd, uap->sd, &lfp);
1859 head = (struct socket *)lfp->f_data;
1860 error = sctp_can_peel_off(head, assoc_id);
1866 * At this point we know we do have a assoc to pull
1867 * we proceed to get the fd setup. This may block
1871 fflag = lfp->f_flag;
1872 error = falloc(td->td_lwp, &nfp, &fd);
1875 * Probably ran out of file descriptors. Put the
1876 * unaccepted connection back onto the queue and
1877 * do another wakeup so some other process might
1878 * have a chance at it.
1883 uap->sysmsg_iresult = fd;
1885 so = sctp_get_peeloff(head, assoc_id, &error);
1888 * Either someone else peeled it off OR
1889 * we can't get a socket.
1893 soreference(so); /* reference needed */
1894 soclrstate(so, SS_NOFDREF | SS_COMP); /* when clearing NOFDREF */
1896 if (head->so_sigio != NULL)
1897 fsetown(fgetown(head->so_sigio), &so->so_sigio);
1899 nfp->f_type = DTYPE_SOCKET;
1900 nfp->f_flag = fflag;
1901 nfp->f_ops = &socketops;
1906 * Assign the file pointer to the reserved descriptor, or clear
1907 * the reserved descriptor if an error occured.
1910 fsetfd(fdp, NULL, fd);
1912 fsetfd(fdp, nfp, fd);
1915 * Release explicitly held references before returning.