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 <net/netmsg2.h>
85 #include <netinet/sctp_peeloff.h>
93 static MALLOC_DEFINE(M_SENDFILE
, "sendfile", "sendfile sfbuf ref structures");
96 * System call interface to the socket abstraction.
99 extern struct fileops socketops
;
102 * socket_args(int domain, int type, int protocol)
105 kern_socket(int domain
, int type
, int protocol
, int *res
)
107 struct thread
*td
= curthread
;
108 struct proc
*p
= td
->td_proc
;
115 error
= falloc(p
, &fp
, &fd
);
118 error
= socreate(domain
, &so
, type
, protocol
, td
);
122 fp
->f_type
= DTYPE_SOCKET
;
123 fp
->f_flag
= FREAD
| FWRITE
;
124 fp
->f_ops
= &socketops
;
134 sys_socket(struct socket_args
*uap
)
138 error
= kern_socket(uap
->domain
, uap
->type
, uap
->protocol
,
139 &uap
->sysmsg_result
);
145 kern_bind(int s
, struct sockaddr
*sa
)
147 struct thread
*td
= curthread
;
148 struct proc
*p
= td
->td_proc
;
153 error
= holdsock(p
->p_fd
, s
, &fp
);
156 error
= sobind((struct socket
*)fp
->f_data
, sa
, td
);
162 * bind_args(int s, caddr_t name, int namelen)
165 sys_bind(struct bind_args
*uap
)
170 error
= getsockaddr(&sa
, uap
->name
, uap
->namelen
);
173 error
= kern_bind(uap
->s
, sa
);
180 kern_listen(int s
, int backlog
)
182 struct thread
*td
= curthread
;
183 struct proc
*p
= td
->td_proc
;
188 error
= holdsock(p
->p_fd
, s
, &fp
);
191 error
= solisten((struct socket
*)fp
->f_data
, backlog
, td
);
197 * listen_args(int s, int backlog)
200 sys_listen(struct listen_args
*uap
)
204 error
= kern_listen(uap
->s
, uap
->backlog
);
209 * Returns the accepted socket as well.
212 soaccept_predicate(struct netmsg
*msg0
)
214 struct netmsg_so_notify
*msg
= (struct netmsg_so_notify
*)msg0
;
215 struct socket
*head
= msg
->nm_so
;
217 if (head
->so_error
!= 0) {
218 msg
->nm_netmsg
.nm_lmsg
.ms_error
= head
->so_error
;
221 if (!TAILQ_EMPTY(&head
->so_comp
)) {
222 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
223 msg
->nm_so
= TAILQ_FIRST(&head
->so_comp
);
224 TAILQ_REMOVE(&head
->so_comp
, msg
->nm_so
, so_list
);
227 msg
->nm_netmsg
.nm_lmsg
.ms_error
= 0;
230 if (head
->so_state
& SS_CANTRCVMORE
) {
231 msg
->nm_netmsg
.nm_lmsg
.ms_error
= ECONNABORTED
;
234 if (msg
->nm_fflags
& FNONBLOCK
) {
235 msg
->nm_netmsg
.nm_lmsg
.ms_error
= EWOULDBLOCK
;
243 * The second argument to kern_accept() is a handle to a struct sockaddr.
244 * This allows kern_accept() to return a pointer to an allocated struct
245 * sockaddr which must be freed later with FREE(). The caller must
246 * initialize *name to NULL.
249 kern_accept(int s
, int fflags
, struct sockaddr
**name
, int *namelen
, int *res
)
251 struct thread
*td
= curthread
;
252 struct proc
*p
= td
->td_proc
;
253 struct file
*lfp
= NULL
;
254 struct file
*nfp
= NULL
;
256 struct socket
*head
, *so
;
257 struct netmsg_so_notify msg
;
260 u_int fflag
; /* type must match fp->f_flag */
264 if (name
&& namelen
&& *namelen
< 0)
267 error
= holdsock(p
->p_fd
, s
, &lfp
);
271 error
= falloc(p
, &nfp
, &fd
);
272 if (error
) { /* Probably ran out of file descriptors. */
276 head
= (struct socket
*)lfp
->f_data
;
277 if ((head
->so_options
& SO_ACCEPTCONN
) == 0) {
282 if (fflags
& O_FBLOCKING
)
283 fflags
|= lfp
->f_flag
& ~FNONBLOCK
;
284 else if (fflags
& O_FNONBLOCKING
)
285 fflags
|= lfp
->f_flag
| FNONBLOCK
;
287 fflags
= lfp
->f_flag
;
289 /* optimize for uniprocessor case later XXX JH */
290 port
= head
->so_proto
->pr_mport(head
, NULL
, NULL
, PRU_PRED
);
291 netmsg_init_abortable(&msg
.nm_netmsg
, &curthread
->td_msgport
,
294 netmsg_so_notify_doabort
);
295 msg
.nm_predicate
= soaccept_predicate
;
296 msg
.nm_fflags
= fflags
;
298 msg
.nm_etype
= NM_REVENT
;
299 error
= lwkt_domsg(port
, &msg
.nm_netmsg
.nm_lmsg
, PCATCH
);
304 * At this point we have the connection that's ready to be accepted.
310 /* connection has been removed from the listen queue */
311 KNOTE(&head
->so_rcv
.ssb_sel
.si_note
, 0);
313 so
->so_state
&= ~SS_COMP
;
315 if (head
->so_sigio
!= NULL
)
316 fsetown(fgetown(head
->so_sigio
), &so
->so_sigio
);
318 nfp
->f_type
= DTYPE_SOCKET
;
320 nfp
->f_ops
= &socketops
;
322 /* Sync socket nonblocking/async state with file flags */
323 tmp
= fflag
& FNONBLOCK
;
324 (void) fo_ioctl(nfp
, FIONBIO
, (caddr_t
)&tmp
, p
->p_ucred
);
325 tmp
= fflag
& FASYNC
;
326 (void) fo_ioctl(nfp
, FIOASYNC
, (caddr_t
)&tmp
, p
->p_ucred
);
329 error
= soaccept(so
, &sa
);
332 * Set the returned name and namelen as applicable. Set the returned
333 * namelen to 0 for older code which might ignore the return value
337 if (sa
&& name
&& namelen
) {
338 if (*namelen
> sa
->sa_len
)
339 *namelen
= sa
->sa_len
;
349 * If an error occured clear the reserved descriptor, else associate
352 * Note that *res is normally ignored if an error is returned but
353 * a syscall message will still have access to the result code.
367 * accept(int s, caddr_t name, int *anamelen)
370 sys_accept(struct accept_args
*uap
)
372 struct sockaddr
*sa
= NULL
;
377 error
= copyin(uap
->anamelen
, &sa_len
, sizeof(sa_len
));
381 error
= kern_accept(uap
->s
, 0, &sa
, &sa_len
, &uap
->sysmsg_result
);
384 error
= copyout(sa
, uap
->name
, sa_len
);
386 error
= copyout(&sa_len
, uap
->anamelen
,
387 sizeof(*uap
->anamelen
));
392 error
= kern_accept(uap
->s
, 0, NULL
, 0, &uap
->sysmsg_result
);
398 * extaccept(int s, int fflags, caddr_t name, int *anamelen)
401 sys_extaccept(struct extaccept_args
*uap
)
403 struct sockaddr
*sa
= NULL
;
406 int fflags
= uap
->flags
& O_FMASK
;
409 error
= copyin(uap
->anamelen
, &sa_len
, sizeof(sa_len
));
413 error
= kern_accept(uap
->s
, fflags
, &sa
, &sa_len
, &uap
->sysmsg_result
);
416 error
= copyout(sa
, uap
->name
, sa_len
);
418 error
= copyout(&sa_len
, uap
->anamelen
,
419 sizeof(*uap
->anamelen
));
424 error
= kern_accept(uap
->s
, fflags
, NULL
, 0, &uap
->sysmsg_result
);
431 * Returns TRUE if predicate satisfied.
434 soconnected_predicate(struct netmsg
*msg0
)
436 struct netmsg_so_notify
*msg
= (struct netmsg_so_notify
*)msg0
;
437 struct socket
*so
= msg
->nm_so
;
439 /* check predicate */
440 if (!(so
->so_state
& SS_ISCONNECTING
) || so
->so_error
!= 0) {
441 msg
->nm_netmsg
.nm_lmsg
.ms_error
= so
->so_error
;
449 kern_connect(int s
, int fflags
, struct sockaddr
*sa
)
451 struct thread
*td
= curthread
;
452 struct proc
*p
= td
->td_proc
;
455 int error
, interrupted
= 0;
457 error
= holdsock(p
->p_fd
, s
, &fp
);
460 so
= (struct socket
*)fp
->f_data
;
462 if (fflags
& O_FBLOCKING
)
463 /* fflags &= ~FNONBLOCK; */;
464 else if (fflags
& O_FNONBLOCKING
)
469 if (so
->so_state
& SS_ISCONNECTING
) {
473 error
= soconnect(so
, sa
, td
);
476 if ((fflags
& FNONBLOCK
) && (so
->so_state
& SS_ISCONNECTING
)) {
480 if ((so
->so_state
& SS_ISCONNECTING
) && so
->so_error
== 0) {
481 struct netmsg_so_notify msg
;
484 port
= so
->so_proto
->pr_mport(so
, sa
, NULL
, PRU_PRED
);
485 netmsg_init_abortable(&msg
.nm_netmsg
,
486 &curthread
->td_msgport
,
489 netmsg_so_notify_doabort
);
490 msg
.nm_predicate
= soconnected_predicate
;
492 msg
.nm_etype
= NM_REVENT
;
493 error
= lwkt_domsg(port
, &msg
.nm_netmsg
.nm_lmsg
, PCATCH
);
494 if (error
== EINTR
|| error
== ERESTART
)
498 error
= so
->so_error
;
503 so
->so_state
&= ~SS_ISCONNECTING
;
504 if (error
== ERESTART
)
512 * connect_args(int s, caddr_t name, int namelen)
515 sys_connect(struct connect_args
*uap
)
520 error
= getsockaddr(&sa
, uap
->name
, uap
->namelen
);
523 error
= kern_connect(uap
->s
, 0, sa
);
530 * connect_args(int s, int fflags, caddr_t name, int namelen)
533 sys_extconnect(struct extconnect_args
*uap
)
537 int fflags
= uap
->flags
& O_FMASK
;
539 error
= getsockaddr(&sa
, uap
->name
, uap
->namelen
);
542 error
= kern_connect(uap
->s
, fflags
, sa
);
549 kern_socketpair(int domain
, int type
, int protocol
, int *sv
)
551 struct thread
*td
= curthread
;
552 struct proc
*p
= td
->td_proc
;
553 struct file
*fp1
, *fp2
;
554 struct socket
*so1
, *so2
;
558 error
= socreate(domain
, &so1
, type
, protocol
, td
);
561 error
= socreate(domain
, &so2
, type
, protocol
, td
);
564 error
= falloc(p
, &fp1
, &fd1
);
569 error
= falloc(p
, &fp2
, &fd2
);
574 error
= soconnect2(so1
, so2
);
577 if (type
== SOCK_DGRAM
) {
579 * Datagram socket connection is asymmetric.
581 error
= soconnect2(so2
, so1
);
585 fp1
->f_type
= fp2
->f_type
= DTYPE_SOCKET
;
586 fp1
->f_flag
= fp2
->f_flag
= FREAD
|FWRITE
;
587 fp1
->f_ops
= fp2
->f_ops
= &socketops
;
594 fsetfd(p
, NULL
, fd2
);
597 fsetfd(p
, NULL
, fd1
);
600 (void)soclose(so2
, 0);
602 (void)soclose(so1
, 0);
607 * socketpair(int domain, int type, int protocol, int *rsv)
610 sys_socketpair(struct socketpair_args
*uap
)
614 error
= kern_socketpair(uap
->domain
, uap
->type
, uap
->protocol
, sockv
);
617 error
= copyout(sockv
, uap
->rsv
, sizeof(sockv
));
622 kern_sendmsg(int s
, struct sockaddr
*sa
, struct uio
*auio
,
623 struct mbuf
*control
, int flags
, int *res
)
625 struct thread
*td
= curthread
;
626 struct lwp
*lp
= td
->td_lwp
;
627 struct proc
*p
= td
->td_proc
;
632 struct iovec
*ktriov
= NULL
;
636 error
= holdsock(p
->p_fd
, s
, &fp
);
639 if (auio
->uio_resid
< 0) {
644 if (KTRPOINT(td
, KTR_GENIO
)) {
645 int iovlen
= auio
->uio_iovcnt
* sizeof (struct iovec
);
647 MALLOC(ktriov
, struct iovec
*, iovlen
, M_TEMP
, M_WAITOK
);
648 bcopy((caddr_t
)auio
->uio_iov
, (caddr_t
)ktriov
, iovlen
);
652 len
= auio
->uio_resid
;
653 so
= (struct socket
*)fp
->f_data
;
654 if ((flags
& (MSG_FNONBLOCKING
|MSG_FBLOCKING
)) == 0) {
655 if (fp
->f_flag
& FNONBLOCK
)
656 flags
|= MSG_FNONBLOCKING
;
658 error
= so_pru_sosend(so
, sa
, auio
, NULL
, control
, flags
, td
);
660 if (auio
->uio_resid
!= len
&& (error
== ERESTART
||
661 error
== EINTR
|| error
== EWOULDBLOCK
))
664 lwpsignal(p
, lp
, SIGPIPE
);
667 if (ktriov
!= NULL
) {
669 ktruio
.uio_iov
= ktriov
;
670 ktruio
.uio_resid
= len
- auio
->uio_resid
;
671 ktrgenio(lp
, s
, UIO_WRITE
, &ktruio
, error
);
673 FREE(ktriov
, M_TEMP
);
677 *res
= len
- auio
->uio_resid
;
684 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
687 sys_sendto(struct sendto_args
*uap
)
689 struct thread
*td
= curthread
;
692 struct sockaddr
*sa
= NULL
;
696 error
= getsockaddr(&sa
, uap
->to
, uap
->tolen
);
700 aiov
.iov_base
= uap
->buf
;
701 aiov
.iov_len
= uap
->len
;
702 auio
.uio_iov
= &aiov
;
705 auio
.uio_resid
= uap
->len
;
706 auio
.uio_segflg
= UIO_USERSPACE
;
707 auio
.uio_rw
= UIO_WRITE
;
710 error
= kern_sendmsg(uap
->s
, sa
, &auio
, NULL
, uap
->flags
,
711 &uap
->sysmsg_result
);
719 * sendmsg_args(int s, caddr_t msg, int flags)
722 sys_sendmsg(struct sendmsg_args
*uap
)
724 struct thread
*td
= curthread
;
727 struct iovec aiov
[UIO_SMALLIOV
], *iov
= NULL
;
728 struct sockaddr
*sa
= NULL
;
729 struct mbuf
*control
= NULL
;
732 error
= copyin(uap
->msg
, (caddr_t
)&msg
, sizeof(msg
));
737 * Conditionally copyin msg.msg_name.
740 error
= getsockaddr(&sa
, msg
.msg_name
, msg
.msg_namelen
);
748 error
= iovec_copyin(msg
.msg_iov
, &iov
, aiov
, msg
.msg_iovlen
,
753 auio
.uio_iovcnt
= msg
.msg_iovlen
;
755 auio
.uio_segflg
= UIO_USERSPACE
;
756 auio
.uio_rw
= UIO_WRITE
;
760 * Conditionally copyin msg.msg_control.
762 if (msg
.msg_control
) {
763 if (msg
.msg_controllen
< sizeof(struct cmsghdr
) ||
764 msg
.msg_controllen
> MLEN
) {
768 control
= m_get(MB_WAIT
, MT_CONTROL
);
769 if (control
== NULL
) {
773 control
->m_len
= msg
.msg_controllen
;
774 error
= copyin(msg
.msg_control
, mtod(control
, caddr_t
),
782 error
= kern_sendmsg(uap
->s
, sa
, &auio
, control
, uap
->flags
,
783 &uap
->sysmsg_result
);
786 iovec_free(&iov
, aiov
);
794 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
795 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
796 * Don't forget to FREE() and m_free() these if they are returned.
799 kern_recvmsg(int s
, struct sockaddr
**sa
, struct uio
*auio
,
800 struct mbuf
**control
, int *flags
, int *res
)
802 struct thread
*td
= curthread
;
803 struct proc
*p
= td
->td_proc
;
809 struct iovec
*ktriov
= NULL
;
813 error
= holdsock(p
->p_fd
, s
, &fp
);
816 if (auio
->uio_resid
< 0) {
821 if (KTRPOINT(td
, KTR_GENIO
)) {
822 int iovlen
= auio
->uio_iovcnt
* sizeof (struct iovec
);
824 MALLOC(ktriov
, struct iovec
*, iovlen
, M_TEMP
, M_WAITOK
);
825 bcopy(auio
->uio_iov
, ktriov
, iovlen
);
829 len
= auio
->uio_resid
;
830 so
= (struct socket
*)fp
->f_data
;
832 if (flags
== NULL
|| (*flags
& (MSG_FNONBLOCKING
|MSG_FBLOCKING
)) == 0) {
833 if (fp
->f_flag
& FNONBLOCK
) {
835 *flags
|= MSG_FNONBLOCKING
;
837 lflags
= MSG_FNONBLOCKING
;
843 error
= so_pru_soreceive(so
, sa
, auio
, NULL
, control
, flags
);
845 if (auio
->uio_resid
!= len
&& (error
== ERESTART
||
846 error
== EINTR
|| error
== EWOULDBLOCK
))
850 if (ktriov
!= NULL
) {
852 ktruio
.uio_iov
= ktriov
;
853 ktruio
.uio_resid
= len
- auio
->uio_resid
;
854 ktrgenio(td
->td_lwp
, s
, UIO_READ
, &ktruio
, error
);
856 FREE(ktriov
, M_TEMP
);
860 *res
= len
- auio
->uio_resid
;
867 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
868 * caddr_t from, int *fromlenaddr)
871 sys_recvfrom(struct recvfrom_args
*uap
)
873 struct thread
*td
= curthread
;
876 struct sockaddr
*sa
= NULL
;
879 if (uap
->from
&& uap
->fromlenaddr
) {
880 error
= copyin(uap
->fromlenaddr
, &fromlen
, sizeof(fromlen
));
888 aiov
.iov_base
= uap
->buf
;
889 aiov
.iov_len
= uap
->len
;
890 auio
.uio_iov
= &aiov
;
893 auio
.uio_resid
= uap
->len
;
894 auio
.uio_segflg
= UIO_USERSPACE
;
895 auio
.uio_rw
= UIO_READ
;
898 error
= kern_recvmsg(uap
->s
, uap
->from
? &sa
: NULL
, &auio
, NULL
,
899 &uap
->flags
, &uap
->sysmsg_result
);
901 if (error
== 0 && uap
->from
) {
902 /* note: sa may still be NULL */
904 fromlen
= MIN(fromlen
, sa
->sa_len
);
905 error
= copyout(sa
, uap
->from
, fromlen
);
910 error
= copyout(&fromlen
, uap
->fromlenaddr
,
921 * recvmsg_args(int s, struct msghdr *msg, int flags)
924 sys_recvmsg(struct recvmsg_args
*uap
)
926 struct thread
*td
= curthread
;
929 struct iovec aiov
[UIO_SMALLIOV
], *iov
= NULL
;
930 struct mbuf
*m
, *control
= NULL
;
931 struct sockaddr
*sa
= NULL
;
933 socklen_t
*ufromlenp
, *ucontrollenp
;
934 int error
, fromlen
, controllen
, len
, flags
, *uflagsp
;
937 * This copyin handles everything except the iovec.
939 error
= copyin(uap
->msg
, &msg
, sizeof(msg
));
943 if (msg
.msg_name
&& msg
.msg_namelen
< 0)
945 if (msg
.msg_control
&& msg
.msg_controllen
< 0)
948 ufromlenp
= (socklen_t
*)((caddr_t
)uap
->msg
+ offsetof(struct msghdr
,
950 ucontrollenp
= (socklen_t
*)((caddr_t
)uap
->msg
+ offsetof(struct msghdr
,
952 uflagsp
= (int *)((caddr_t
)uap
->msg
+ offsetof(struct msghdr
,
958 error
= iovec_copyin(msg
.msg_iov
, &iov
, aiov
, msg
.msg_iovlen
,
963 auio
.uio_iovcnt
= msg
.msg_iovlen
;
965 auio
.uio_segflg
= UIO_USERSPACE
;
966 auio
.uio_rw
= UIO_READ
;
971 error
= kern_recvmsg(uap
->s
, msg
.msg_name
? &sa
: NULL
, &auio
,
972 msg
.msg_control
? &control
: NULL
, &flags
, &uap
->sysmsg_result
);
975 * Conditionally copyout the name and populate the namelen field.
977 if (error
== 0 && msg
.msg_name
) {
978 /* note: sa may still be NULL */
980 fromlen
= MIN(msg
.msg_namelen
, sa
->sa_len
);
981 error
= copyout(sa
, msg
.msg_name
, fromlen
);
986 error
= copyout(&fromlen
, ufromlenp
,
991 * Copyout msg.msg_control and msg.msg_controllen.
993 if (error
== 0 && msg
.msg_control
) {
994 len
= msg
.msg_controllen
;
996 ctlbuf
= (caddr_t
)msg
.msg_control
;
998 while(m
&& len
> 0) {
1001 if (len
>= m
->m_len
) {
1004 msg
.msg_flags
|= MSG_CTRUNC
;
1008 error
= copyout(mtod(m
, caddr_t
), ctlbuf
, tocopy
);
1016 controllen
= ctlbuf
- (caddr_t
)msg
.msg_control
;
1017 error
= copyout(&controllen
, ucontrollenp
,
1018 sizeof(*ucontrollenp
));
1022 error
= copyout(&flags
, uflagsp
, sizeof(*uflagsp
));
1027 iovec_free(&iov
, aiov
);
1034 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1035 * in kernel pointer instead of a userland pointer. This allows us
1036 * to manipulate socket options in the emulation code.
1039 kern_setsockopt(int s
, struct sockopt
*sopt
)
1041 struct thread
*td
= curthread
;
1042 struct proc
*p
= td
->td_proc
;
1046 if (sopt
->sopt_val
== NULL
&& sopt
->sopt_valsize
!= 0)
1048 if (sopt
->sopt_valsize
< 0)
1051 error
= holdsock(p
->p_fd
, s
, &fp
);
1055 error
= sosetopt((struct socket
*)fp
->f_data
, sopt
);
1061 * setsockopt_args(int s, int level, int name, caddr_t val, int valsize)
1064 sys_setsockopt(struct setsockopt_args
*uap
)
1066 struct thread
*td
= curthread
;
1067 struct sockopt sopt
;
1070 sopt
.sopt_level
= uap
->level
;
1071 sopt
.sopt_name
= uap
->name
;
1072 sopt
.sopt_valsize
= uap
->valsize
;
1074 sopt
.sopt_val
= NULL
;
1076 if (sopt
.sopt_valsize
< 0 || sopt
.sopt_valsize
> SOMAXOPT_SIZE
)
1079 sopt
.sopt_val
= kmalloc(sopt
.sopt_valsize
, M_TEMP
, M_WAITOK
);
1080 error
= copyin(uap
->val
, sopt
.sopt_val
, sopt
.sopt_valsize
);
1085 error
= kern_setsockopt(uap
->s
, &sopt
);
1088 kfree(sopt
.sopt_val
, M_TEMP
);
1093 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1094 * in kernel pointer instead of a userland pointer. This allows us
1095 * to manipulate socket options in the emulation code.
1098 kern_getsockopt(int s
, struct sockopt
*sopt
)
1100 struct thread
*td
= curthread
;
1101 struct proc
*p
= td
->td_proc
;
1105 if (sopt
->sopt_val
== NULL
&& sopt
->sopt_valsize
!= 0)
1107 if (sopt
->sopt_valsize
< 0 || sopt
->sopt_valsize
> SOMAXOPT_SIZE
)
1110 error
= holdsock(p
->p_fd
, s
, &fp
);
1114 error
= sogetopt((struct socket
*)fp
->f_data
, sopt
);
1120 * getsockopt_Args(int s, int level, int name, caddr_t val, int *avalsize)
1123 sys_getsockopt(struct getsockopt_args
*uap
)
1125 struct thread
*td
= curthread
;
1126 struct sockopt sopt
;
1130 error
= copyin(uap
->avalsize
, &valsize
, sizeof(valsize
));
1137 sopt
.sopt_level
= uap
->level
;
1138 sopt
.sopt_name
= uap
->name
;
1139 sopt
.sopt_valsize
= valsize
;
1141 sopt
.sopt_val
= NULL
;
1143 if (sopt
.sopt_valsize
< 0 || sopt
.sopt_valsize
> SOMAXOPT_SIZE
)
1146 sopt
.sopt_val
= kmalloc(sopt
.sopt_valsize
, M_TEMP
, M_WAITOK
);
1147 error
= copyin(uap
->val
, sopt
.sopt_val
, sopt
.sopt_valsize
);
1152 error
= kern_getsockopt(uap
->s
, &sopt
);
1155 valsize
= sopt
.sopt_valsize
;
1156 error
= copyout(&valsize
, uap
->avalsize
, sizeof(valsize
));
1160 error
= copyout(sopt
.sopt_val
, uap
->val
, sopt
.sopt_valsize
);
1163 kfree(sopt
.sopt_val
, M_TEMP
);
1168 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1169 * This allows kern_getsockname() to return a pointer to an allocated struct
1170 * sockaddr which must be freed later with FREE(). The caller must
1171 * initialize *name to NULL.
1174 kern_getsockname(int s
, struct sockaddr
**name
, int *namelen
)
1176 struct thread
*td
= curthread
;
1177 struct proc
*p
= td
->td_proc
;
1180 struct sockaddr
*sa
= NULL
;
1183 error
= holdsock(p
->p_fd
, s
, &fp
);
1190 so
= (struct socket
*)fp
->f_data
;
1191 error
= so_pru_sockaddr(so
, &sa
);
1196 *namelen
= MIN(*namelen
, sa
->sa_len
);
1206 * getsockname_args(int fdes, caddr_t asa, int *alen)
1211 sys_getsockname(struct getsockname_args
*uap
)
1213 struct sockaddr
*sa
= NULL
;
1216 error
= copyin(uap
->alen
, &sa_len
, sizeof(sa_len
));
1220 error
= kern_getsockname(uap
->fdes
, &sa
, &sa_len
);
1223 error
= copyout(sa
, uap
->asa
, sa_len
);
1225 error
= copyout(&sa_len
, uap
->alen
, sizeof(*uap
->alen
));
1232 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1233 * This allows kern_getpeername() to return a pointer to an allocated struct
1234 * sockaddr which must be freed later with FREE(). The caller must
1235 * initialize *name to NULL.
1238 kern_getpeername(int s
, struct sockaddr
**name
, int *namelen
)
1240 struct thread
*td
= curthread
;
1241 struct proc
*p
= td
->td_proc
;
1244 struct sockaddr
*sa
= NULL
;
1247 error
= holdsock(p
->p_fd
, s
, &fp
);
1254 so
= (struct socket
*)fp
->f_data
;
1255 if ((so
->so_state
& (SS_ISCONNECTED
|SS_ISCONFIRMING
)) == 0) {
1259 error
= so_pru_peeraddr(so
, &sa
);
1264 *namelen
= MIN(*namelen
, sa
->sa_len
);
1274 * getpeername_args(int fdes, caddr_t asa, int *alen)
1276 * Get name of peer for connected socket.
1279 sys_getpeername(struct getpeername_args
*uap
)
1281 struct sockaddr
*sa
= NULL
;
1284 error
= copyin(uap
->alen
, &sa_len
, sizeof(sa_len
));
1288 error
= kern_getpeername(uap
->fdes
, &sa
, &sa_len
);
1291 error
= copyout(sa
, uap
->asa
, sa_len
);
1293 error
= copyout(&sa_len
, uap
->alen
, sizeof(*uap
->alen
));
1300 getsockaddr(struct sockaddr
**namp
, caddr_t uaddr
, size_t len
)
1302 struct sockaddr
*sa
;
1306 if (len
> SOCK_MAXADDRLEN
)
1307 return ENAMETOOLONG
;
1308 if (len
< offsetof(struct sockaddr
, sa_data
[0]))
1310 MALLOC(sa
, struct sockaddr
*, len
, M_SONAME
, M_WAITOK
);
1311 error
= copyin(uaddr
, sa
, len
);
1315 #if BYTE_ORDER != BIG_ENDIAN
1317 * The bind(), connect(), and sendto() syscalls were not
1318 * versioned for COMPAT_43. Thus, this check must stay.
1320 if (sa
->sa_family
== 0 && sa
->sa_len
< AF_MAX
)
1321 sa
->sa_family
= sa
->sa_len
;
1330 * Detach a mapped page and release resources back to the system.
1331 * We must release our wiring and if the object is ripped out
1332 * from under the vm_page we become responsible for freeing the
1333 * page. These routines must be MPSAFE.
1335 * XXX HACK XXX TEMPORARY UNTIL WE IMPLEMENT EXT MBUF REFERENCE COUNTING
1337 * XXX vm_page_*() routines are not MPSAFE yet, the MP lock is required.
1340 sf_buf_mref(void *arg
)
1342 struct sfbuf_mref
*sfm
= arg
;
1345 * We must already hold a ref so there is no race to 0, just
1346 * atomically increment the count.
1348 atomic_add_int(&sfm
->mref_count
, 1);
1352 sf_buf_mfree(void *arg
)
1354 struct sfbuf_mref
*sfm
= arg
;
1357 KKASSERT(sfm
->mref_count
> 0);
1358 if (atomic_fetchadd_int(&sfm
->mref_count
, -1) == 1) {
1360 * XXX vm_page_*() and SFBUF routines not MPSAFE yet.
1364 m
= sf_buf_page(sfm
->sf
);
1365 sf_buf_free(sfm
->sf
);
1366 vm_page_unwire(m
, 0);
1367 if (m
->wire_count
== 0 && m
->object
== NULL
)
1368 vm_page_try_to_free(m
);
1371 kfree(sfm
, M_SENDFILE
);
1377 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1378 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1380 * Send a file specified by 'fd' and starting at 'offset' to a socket
1381 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1382 * nbytes == 0. Optionally add a header and/or trailer to the socket
1383 * output. If specified, write the total number of bytes sent into *sbytes.
1385 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1386 * the headers to count against the remaining bytes to be sent from
1387 * the file descriptor. We may wish to implement a compatibility syscall
1391 sys_sendfile(struct sendfile_args
*uap
)
1393 struct thread
*td
= curthread
;
1394 struct proc
*p
= td
->td_proc
;
1396 struct vnode
*vp
= NULL
;
1397 struct sf_hdtr hdtr
;
1398 struct iovec aiov
[UIO_SMALLIOV
], *iov
= NULL
;
1400 struct mbuf
*mheader
= NULL
;
1401 off_t hdtr_size
= 0, sbytes
;
1402 int error
, hbytes
= 0, tbytes
;
1407 * Do argument checking. Must be a regular file in, stream
1408 * type and connected socket out, positive offset.
1410 fp
= holdfp(p
->p_fd
, uap
->fd
, FREAD
);
1414 if (fp
->f_type
!= DTYPE_VNODE
) {
1418 vp
= (struct vnode
*)fp
->f_data
;
1423 * If specified, get the pointer to the sf_hdtr struct for
1424 * any headers/trailers.
1427 error
= copyin(uap
->hdtr
, &hdtr
, sizeof(hdtr
));
1434 error
= iovec_copyin(hdtr
.headers
, &iov
, aiov
,
1435 hdtr
.hdr_cnt
, &hbytes
);
1439 auio
.uio_iovcnt
= hdtr
.hdr_cnt
;
1440 auio
.uio_offset
= 0;
1441 auio
.uio_segflg
= UIO_USERSPACE
;
1442 auio
.uio_rw
= UIO_WRITE
;
1444 auio
.uio_resid
= hbytes
;
1446 mheader
= m_uiomove(&auio
);
1448 iovec_free(&iov
, aiov
);
1449 if (mheader
== NULL
)
1454 error
= kern_sendfile(vp
, uap
->s
, uap
->offset
, uap
->nbytes
, mheader
,
1455 &sbytes
, uap
->flags
);
1460 * Send trailers. Wimp out and use writev(2).
1462 if (uap
->hdtr
!= NULL
&& hdtr
.trailers
!= NULL
) {
1463 error
= iovec_copyin(hdtr
.trailers
, &iov
, aiov
,
1464 hdtr
.trl_cnt
, &auio
.uio_resid
);
1468 auio
.uio_iovcnt
= hdtr
.trl_cnt
;
1469 auio
.uio_offset
= 0;
1470 auio
.uio_segflg
= UIO_USERSPACE
;
1471 auio
.uio_rw
= UIO_WRITE
;
1474 error
= kern_sendmsg(uap
->s
, NULL
, &auio
, NULL
, 0, &tbytes
);
1476 iovec_free(&iov
, aiov
);
1479 hdtr_size
+= tbytes
; /* trailer bytes successfully sent */
1483 if (uap
->sbytes
!= NULL
) {
1484 sbytes
+= hdtr_size
;
1485 copyout(&sbytes
, uap
->sbytes
, sizeof(off_t
));
1493 kern_sendfile(struct vnode
*vp
, int sfd
, off_t offset
, size_t nbytes
,
1494 struct mbuf
*mheader
, off_t
*sbytes
, int flags
)
1496 struct thread
*td
= curthread
;
1497 struct proc
*p
= td
->td_proc
;
1498 struct vm_object
*obj
;
1503 struct sfbuf_mref
*sfm
;
1509 if (vp
->v_type
!= VREG
) {
1513 if ((obj
= vp
->v_object
) == NULL
) {
1517 error
= holdsock(p
->p_fd
, sfd
, &fp
);
1520 so
= (struct socket
*)fp
->f_data
;
1521 if (so
->so_type
!= SOCK_STREAM
) {
1525 if ((so
->so_state
& SS_ISCONNECTED
) == 0) {
1536 * Protect against multiple writers to the socket.
1538 ssb_lock(&so
->so_snd
, M_WAITOK
);
1541 * Loop through the pages in the file, starting with the requested
1542 * offset. Get a file page (do I/O if necessary), map the file page
1543 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1546 for (off
= offset
; ; off
+= xfsize
, *sbytes
+= xfsize
+ hbytes
) {
1550 pindex
= OFF_TO_IDX(off
);
1553 * Calculate the amount to transfer. Not to exceed a page,
1554 * the EOF, or the passed in nbytes.
1556 xfsize
= vp
->v_filesize
- off
;
1557 if (xfsize
> PAGE_SIZE
)
1559 pgoff
= (vm_offset_t
)(off
& PAGE_MASK
);
1560 if (PAGE_SIZE
- pgoff
< xfsize
)
1561 xfsize
= PAGE_SIZE
- pgoff
;
1562 if (nbytes
&& xfsize
> (nbytes
- *sbytes
))
1563 xfsize
= nbytes
- *sbytes
;
1567 * Optimize the non-blocking case by looking at the socket space
1568 * before going to the extra work of constituting the sf_buf.
1570 if ((fp
->f_flag
& FNONBLOCK
) && ssb_space(&so
->so_snd
) <= 0) {
1571 if (so
->so_state
& SS_CANTSENDMORE
)
1575 ssb_unlock(&so
->so_snd
);
1579 * Attempt to look up the page.
1581 * Allocate if not found, wait and loop if busy, then
1582 * wire the page. critical section protection is
1583 * required to maintain the object association (an
1584 * interrupt can free the page) through to the
1585 * vm_page_wire() call.
1588 pg
= vm_page_lookup(obj
, pindex
);
1590 pg
= vm_page_alloc(obj
, pindex
, VM_ALLOC_NORMAL
);
1597 } else if (vm_page_sleep_busy(pg
, TRUE
, "sfpbsy")) {
1605 * If page is not valid for what we need, initiate I/O
1608 if (!pg
->valid
|| !vm_page_is_valid(pg
, pgoff
, xfsize
)) {
1614 * Ensure that our page is still around when the I/O
1617 vm_page_io_start(pg
);
1620 * Get the page from backing store.
1622 bsize
= vp
->v_mount
->mnt_stat
.f_iosize
;
1623 auio
.uio_iov
= &aiov
;
1624 auio
.uio_iovcnt
= 1;
1626 aiov
.iov_len
= MAXBSIZE
;
1627 auio
.uio_resid
= MAXBSIZE
;
1628 auio
.uio_offset
= trunc_page(off
);
1629 auio
.uio_segflg
= UIO_NOCOPY
;
1630 auio
.uio_rw
= UIO_READ
;
1632 vn_lock(vp
, LK_SHARED
| LK_RETRY
);
1633 error
= VOP_READ(vp
, &auio
,
1634 IO_VMIO
| ((MAXBSIZE
/ bsize
) << 16),
1637 vm_page_flag_clear(pg
, PG_ZERO
);
1638 vm_page_io_finish(pg
);
1641 vm_page_unwire(pg
, 0);
1642 vm_page_try_to_free(pg
);
1644 ssb_unlock(&so
->so_snd
);
1651 * Get a sendfile buf. We usually wait as long as necessary,
1652 * but this wait can be interrupted.
1654 if ((sf
= sf_buf_alloc(pg
, SFB_CATCH
)) == NULL
) {
1656 vm_page_unwire(pg
, 0);
1657 vm_page_try_to_free(pg
);
1659 ssb_unlock(&so
->so_snd
);
1665 * Get an mbuf header and set it up as having external storage.
1667 MGETHDR(m
, MB_WAIT
, MT_DATA
);
1671 ssb_unlock(&so
->so_snd
);
1676 * sfm is a temporary hack, use a per-cpu cache for this.
1678 sfm
= kmalloc(sizeof(struct sfbuf_mref
), M_SENDFILE
, M_WAITOK
);
1680 sfm
->mref_count
= 1;
1682 m
->m_ext
.ext_free
= sf_buf_mfree
;
1683 m
->m_ext
.ext_ref
= sf_buf_mref
;
1684 m
->m_ext
.ext_arg
= sfm
;
1685 m
->m_ext
.ext_buf
= (void *)sf
->kva
;
1686 m
->m_ext
.ext_size
= PAGE_SIZE
;
1687 m
->m_data
= (char *) sf
->kva
+ pgoff
;
1688 m
->m_flags
|= M_EXT
;
1689 m
->m_pkthdr
.len
= m
->m_len
= xfsize
;
1690 KKASSERT((m
->m_flags
& (M_EXT_CLUSTER
)) == 0);
1692 if (mheader
!= NULL
) {
1693 hbytes
= mheader
->m_pkthdr
.len
;
1694 mheader
->m_pkthdr
.len
+= m
->m_pkthdr
.len
;
1702 * Add the buffer to the socket buffer chain.
1707 * Make sure that the socket is still able to take more data.
1708 * CANTSENDMORE being true usually means that the connection
1709 * was closed. so_error is true when an error was sensed after
1711 * The state is checked after the page mapping and buffer
1712 * allocation above since those operations may block and make
1713 * any socket checks stale. From this point forward, nothing
1714 * blocks before the pru_send (or more accurately, any blocking
1715 * results in a loop back to here to re-check).
1717 if ((so
->so_state
& SS_CANTSENDMORE
) || so
->so_error
) {
1718 if (so
->so_state
& SS_CANTSENDMORE
) {
1721 error
= so
->so_error
;
1725 ssb_unlock(&so
->so_snd
);
1730 * Wait for socket space to become available. We do this just
1731 * after checking the connection state above in order to avoid
1732 * a race condition with ssb_wait().
1734 if (ssb_space(&so
->so_snd
) < so
->so_snd
.ssb_lowat
) {
1735 if (fp
->f_flag
& FNONBLOCK
) {
1737 ssb_unlock(&so
->so_snd
);
1742 error
= ssb_wait(&so
->so_snd
);
1744 * An error from ssb_wait usually indicates that we've
1745 * been interrupted by a signal. If we've sent anything
1746 * then return bytes sent, otherwise return the error.
1750 ssb_unlock(&so
->so_snd
);
1756 error
= so_pru_send(so
, 0, m
, NULL
, NULL
, td
);
1759 ssb_unlock(&so
->so_snd
);
1763 if (mheader
!= NULL
) {
1764 *sbytes
+= mheader
->m_pkthdr
.len
;
1765 error
= so_pru_send(so
, 0, mheader
, NULL
, NULL
, td
);
1768 ssb_unlock(&so
->so_snd
);
1773 if (mheader
!= NULL
)
1779 sys_sctp_peeloff(struct sctp_peeloff_args
*uap
)
1782 struct thread
*td
= curthread
;
1783 struct proc
*p
= td
->td_proc
;
1784 struct file
*lfp
= NULL
;
1785 struct file
*nfp
= NULL
;
1787 struct socket
*head
, *so
;
1790 short fflag
; /* type must match fp->f_flag */
1792 assoc_id
= uap
->name
;
1793 error
= holdsock(p
->p_fd
, uap
->sd
, &lfp
);
1798 head
= (struct socket
*)lfp
->f_data
;
1799 error
= sctp_can_peel_off(head
, assoc_id
);
1805 * At this point we know we do have a assoc to pull
1806 * we proceed to get the fd setup. This may block
1810 fflag
= lfp
->f_flag
;
1811 error
= falloc(p
, &nfp
, &fd
);
1814 * Probably ran out of file descriptors. Put the
1815 * unaccepted connection back onto the queue and
1816 * do another wakeup so some other process might
1817 * have a chance at it.
1822 uap
->sysmsg_result
= fd
;
1824 so
= sctp_get_peeloff(head
, assoc_id
, &error
);
1827 * Either someone else peeled it off OR
1828 * we can't get a socket.
1832 so
->so_state
&= ~SS_COMP
;
1833 so
->so_state
&= ~SS_NOFDREF
;
1835 if (head
->so_sigio
!= NULL
)
1836 fsetown(fgetown(head
->so_sigio
), &so
->so_sigio
);
1838 nfp
->f_type
= DTYPE_SOCKET
;
1839 nfp
->f_flag
= fflag
;
1840 nfp
->f_ops
= &socketops
;
1845 * Assign the file pointer to the reserved descriptor, or clear
1846 * the reserved descriptor if an error occured.
1849 fsetfd(p
, NULL
, fd
);
1854 * Release explicitly held references before returning.