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 1.88.2.1 2008/07/26 15:42:06 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>
91 struct lwkt_serialize serializer
;
94 static MALLOC_DEFINE(M_SENDFILE
, "sendfile", "sendfile sfbuf ref structures");
97 * System call interface to the socket abstraction.
100 extern struct fileops socketops
;
103 * socket_args(int domain, int type, int protocol)
106 kern_socket(int domain
, int type
, int protocol
, int *res
)
108 struct thread
*td
= curthread
;
109 struct proc
*p
= td
->td_proc
;
116 error
= falloc(p
, &fp
, &fd
);
119 error
= socreate(domain
, &so
, type
, protocol
, td
);
123 fp
->f_type
= DTYPE_SOCKET
;
124 fp
->f_flag
= FREAD
| FWRITE
;
125 fp
->f_ops
= &socketops
;
135 sys_socket(struct socket_args
*uap
)
139 error
= kern_socket(uap
->domain
, uap
->type
, uap
->protocol
,
140 &uap
->sysmsg_result
);
146 kern_bind(int s
, struct sockaddr
*sa
)
148 struct thread
*td
= curthread
;
149 struct proc
*p
= td
->td_proc
;
154 error
= holdsock(p
->p_fd
, s
, &fp
);
157 error
= sobind((struct socket
*)fp
->f_data
, sa
, td
);
163 * bind_args(int s, caddr_t name, int namelen)
166 sys_bind(struct bind_args
*uap
)
171 error
= getsockaddr(&sa
, uap
->name
, uap
->namelen
);
174 error
= kern_bind(uap
->s
, sa
);
181 kern_listen(int s
, int backlog
)
183 struct thread
*td
= curthread
;
184 struct proc
*p
= td
->td_proc
;
189 error
= holdsock(p
->p_fd
, s
, &fp
);
192 error
= solisten((struct socket
*)fp
->f_data
, backlog
, td
);
198 * listen_args(int s, int backlog)
201 sys_listen(struct listen_args
*uap
)
205 error
= kern_listen(uap
->s
, uap
->backlog
);
210 * Returns the accepted socket as well.
213 soaccept_predicate(struct netmsg
*msg0
)
215 struct netmsg_so_notify
*msg
= (struct netmsg_so_notify
*)msg0
;
216 struct socket
*head
= msg
->nm_so
;
218 if (head
->so_error
!= 0) {
219 msg
->nm_netmsg
.nm_lmsg
.ms_error
= head
->so_error
;
222 if (!TAILQ_EMPTY(&head
->so_comp
)) {
223 /* Abuse nm_so field as copy in/copy out parameter. XXX JH */
224 msg
->nm_so
= TAILQ_FIRST(&head
->so_comp
);
225 TAILQ_REMOVE(&head
->so_comp
, msg
->nm_so
, so_list
);
228 msg
->nm_netmsg
.nm_lmsg
.ms_error
= 0;
231 if (head
->so_state
& SS_CANTRCVMORE
) {
232 msg
->nm_netmsg
.nm_lmsg
.ms_error
= ECONNABORTED
;
235 if (msg
->nm_fflags
& FNONBLOCK
) {
236 msg
->nm_netmsg
.nm_lmsg
.ms_error
= EWOULDBLOCK
;
244 * The second argument to kern_accept() is a handle to a struct sockaddr.
245 * This allows kern_accept() to return a pointer to an allocated struct
246 * sockaddr which must be freed later with FREE(). The caller must
247 * initialize *name to NULL.
250 kern_accept(int s
, int fflags
, struct sockaddr
**name
, int *namelen
, int *res
)
252 struct thread
*td
= curthread
;
253 struct proc
*p
= td
->td_proc
;
254 struct file
*lfp
= NULL
;
255 struct file
*nfp
= NULL
;
257 struct socket
*head
, *so
;
258 struct netmsg_so_notify msg
;
261 u_int fflag
; /* type must match fp->f_flag */
265 if (name
&& namelen
&& *namelen
< 0)
268 error
= holdsock(p
->p_fd
, s
, &lfp
);
272 error
= falloc(p
, &nfp
, &fd
);
273 if (error
) { /* Probably ran out of file descriptors. */
277 head
= (struct socket
*)lfp
->f_data
;
278 if ((head
->so_options
& SO_ACCEPTCONN
) == 0) {
283 if (fflags
& O_FBLOCKING
)
284 fflags
|= lfp
->f_flag
& ~FNONBLOCK
;
285 else if (fflags
& O_FNONBLOCKING
)
286 fflags
|= lfp
->f_flag
| FNONBLOCK
;
288 fflags
= lfp
->f_flag
;
290 /* optimize for uniprocessor case later XXX JH */
291 port
= head
->so_proto
->pr_mport(head
, NULL
, NULL
, PRU_PRED
);
292 netmsg_init_abortable(&msg
.nm_netmsg
, &curthread
->td_msgport
,
295 netmsg_so_notify_doabort
);
296 msg
.nm_predicate
= soaccept_predicate
;
297 msg
.nm_fflags
= fflags
;
299 msg
.nm_etype
= NM_REVENT
;
300 error
= lwkt_domsg(port
, &msg
.nm_netmsg
.nm_lmsg
, PCATCH
);
305 * At this point we have the connection that's ready to be accepted.
311 /* connection has been removed from the listen queue */
312 KNOTE(&head
->so_rcv
.ssb_sel
.si_note
, 0);
314 so
->so_state
&= ~SS_COMP
;
316 if (head
->so_sigio
!= NULL
)
317 fsetown(fgetown(head
->so_sigio
), &so
->so_sigio
);
319 nfp
->f_type
= DTYPE_SOCKET
;
321 nfp
->f_ops
= &socketops
;
323 /* Sync socket nonblocking/async state with file flags */
324 tmp
= fflag
& FNONBLOCK
;
325 (void) fo_ioctl(nfp
, FIONBIO
, (caddr_t
)&tmp
, p
->p_ucred
);
326 tmp
= fflag
& FASYNC
;
327 (void) fo_ioctl(nfp
, FIOASYNC
, (caddr_t
)&tmp
, p
->p_ucred
);
330 error
= soaccept(so
, &sa
);
333 * Set the returned name and namelen as applicable. Set the returned
334 * namelen to 0 for older code which might ignore the return value
338 if (sa
&& name
&& namelen
) {
339 if (*namelen
> sa
->sa_len
)
340 *namelen
= sa
->sa_len
;
350 * If an error occured clear the reserved descriptor, else associate
353 * Note that *res is normally ignored if an error is returned but
354 * a syscall message will still have access to the result code.
368 * accept(int s, caddr_t name, int *anamelen)
371 sys_accept(struct accept_args
*uap
)
373 struct sockaddr
*sa
= NULL
;
378 error
= copyin(uap
->anamelen
, &sa_len
, sizeof(sa_len
));
382 error
= kern_accept(uap
->s
, 0, &sa
, &sa_len
, &uap
->sysmsg_result
);
385 error
= copyout(sa
, uap
->name
, sa_len
);
387 error
= copyout(&sa_len
, uap
->anamelen
,
388 sizeof(*uap
->anamelen
));
393 error
= kern_accept(uap
->s
, 0, NULL
, 0, &uap
->sysmsg_result
);
399 * extaccept(int s, int fflags, caddr_t name, int *anamelen)
402 sys_extaccept(struct extaccept_args
*uap
)
404 struct sockaddr
*sa
= NULL
;
407 int fflags
= uap
->flags
& O_FMASK
;
410 error
= copyin(uap
->anamelen
, &sa_len
, sizeof(sa_len
));
414 error
= kern_accept(uap
->s
, fflags
, &sa
, &sa_len
, &uap
->sysmsg_result
);
417 error
= copyout(sa
, uap
->name
, sa_len
);
419 error
= copyout(&sa_len
, uap
->anamelen
,
420 sizeof(*uap
->anamelen
));
425 error
= kern_accept(uap
->s
, fflags
, NULL
, 0, &uap
->sysmsg_result
);
432 * Returns TRUE if predicate satisfied.
435 soconnected_predicate(struct netmsg
*msg0
)
437 struct netmsg_so_notify
*msg
= (struct netmsg_so_notify
*)msg0
;
438 struct socket
*so
= msg
->nm_so
;
440 /* check predicate */
441 if (!(so
->so_state
& SS_ISCONNECTING
) || so
->so_error
!= 0) {
442 msg
->nm_netmsg
.nm_lmsg
.ms_error
= so
->so_error
;
450 kern_connect(int s
, int fflags
, struct sockaddr
*sa
)
452 struct thread
*td
= curthread
;
453 struct proc
*p
= td
->td_proc
;
456 int error
, interrupted
= 0;
458 error
= holdsock(p
->p_fd
, s
, &fp
);
461 so
= (struct socket
*)fp
->f_data
;
463 if (fflags
& O_FBLOCKING
)
464 /* fflags &= ~FNONBLOCK; */;
465 else if (fflags
& O_FNONBLOCKING
)
470 if (so
->so_state
& SS_ISCONNECTING
) {
474 error
= soconnect(so
, sa
, td
);
477 if ((fflags
& FNONBLOCK
) && (so
->so_state
& SS_ISCONNECTING
)) {
481 if ((so
->so_state
& SS_ISCONNECTING
) && so
->so_error
== 0) {
482 struct netmsg_so_notify msg
;
485 port
= so
->so_proto
->pr_mport(so
, sa
, NULL
, PRU_PRED
);
486 netmsg_init_abortable(&msg
.nm_netmsg
,
487 &curthread
->td_msgport
,
490 netmsg_so_notify_doabort
);
491 msg
.nm_predicate
= soconnected_predicate
;
493 msg
.nm_etype
= NM_REVENT
;
494 error
= lwkt_domsg(port
, &msg
.nm_netmsg
.nm_lmsg
, PCATCH
);
495 if (error
== EINTR
|| error
== ERESTART
)
499 error
= so
->so_error
;
504 so
->so_state
&= ~SS_ISCONNECTING
;
505 if (error
== ERESTART
)
513 * connect_args(int s, caddr_t name, int namelen)
516 sys_connect(struct connect_args
*uap
)
521 error
= getsockaddr(&sa
, uap
->name
, uap
->namelen
);
524 error
= kern_connect(uap
->s
, 0, sa
);
531 * connect_args(int s, int fflags, caddr_t name, int namelen)
534 sys_extconnect(struct extconnect_args
*uap
)
538 int fflags
= uap
->flags
& O_FMASK
;
540 error
= getsockaddr(&sa
, uap
->name
, uap
->namelen
);
543 error
= kern_connect(uap
->s
, fflags
, sa
);
550 kern_socketpair(int domain
, int type
, int protocol
, int *sv
)
552 struct thread
*td
= curthread
;
553 struct proc
*p
= td
->td_proc
;
554 struct file
*fp1
, *fp2
;
555 struct socket
*so1
, *so2
;
559 error
= socreate(domain
, &so1
, type
, protocol
, td
);
562 error
= socreate(domain
, &so2
, type
, protocol
, td
);
565 error
= falloc(p
, &fp1
, &fd1
);
570 error
= falloc(p
, &fp2
, &fd2
);
575 error
= soconnect2(so1
, so2
);
578 if (type
== SOCK_DGRAM
) {
580 * Datagram socket connection is asymmetric.
582 error
= soconnect2(so2
, so1
);
586 fp1
->f_type
= fp2
->f_type
= DTYPE_SOCKET
;
587 fp1
->f_flag
= fp2
->f_flag
= FREAD
|FWRITE
;
588 fp1
->f_ops
= fp2
->f_ops
= &socketops
;
595 fsetfd(p
, NULL
, fd2
);
598 fsetfd(p
, NULL
, fd1
);
601 (void)soclose(so2
, 0);
603 (void)soclose(so1
, 0);
608 * socketpair(int domain, int type, int protocol, int *rsv)
611 sys_socketpair(struct socketpair_args
*uap
)
615 error
= kern_socketpair(uap
->domain
, uap
->type
, uap
->protocol
, sockv
);
618 error
= copyout(sockv
, uap
->rsv
, sizeof(sockv
));
623 kern_sendmsg(int s
, struct sockaddr
*sa
, struct uio
*auio
,
624 struct mbuf
*control
, int flags
, int *res
)
626 struct thread
*td
= curthread
;
627 struct lwp
*lp
= td
->td_lwp
;
628 struct proc
*p
= td
->td_proc
;
633 struct iovec
*ktriov
= NULL
;
637 error
= holdsock(p
->p_fd
, s
, &fp
);
640 if (auio
->uio_resid
< 0) {
645 if (KTRPOINT(td
, KTR_GENIO
)) {
646 int iovlen
= auio
->uio_iovcnt
* sizeof (struct iovec
);
648 MALLOC(ktriov
, struct iovec
*, iovlen
, M_TEMP
, M_WAITOK
);
649 bcopy((caddr_t
)auio
->uio_iov
, (caddr_t
)ktriov
, iovlen
);
653 len
= auio
->uio_resid
;
654 so
= (struct socket
*)fp
->f_data
;
655 if ((flags
& (MSG_FNONBLOCKING
|MSG_FBLOCKING
)) == 0) {
656 if (fp
->f_flag
& FNONBLOCK
)
657 flags
|= MSG_FNONBLOCKING
;
659 error
= so_pru_sosend(so
, sa
, auio
, NULL
, control
, flags
, td
);
661 if (auio
->uio_resid
!= len
&& (error
== ERESTART
||
662 error
== EINTR
|| error
== EWOULDBLOCK
))
665 lwpsignal(p
, lp
, SIGPIPE
);
668 if (ktriov
!= NULL
) {
670 ktruio
.uio_iov
= ktriov
;
671 ktruio
.uio_resid
= len
- auio
->uio_resid
;
672 ktrgenio(lp
, s
, UIO_WRITE
, &ktruio
, error
);
674 FREE(ktriov
, M_TEMP
);
678 *res
= len
- auio
->uio_resid
;
685 * sendto_args(int s, caddr_t buf, size_t len, int flags, caddr_t to, int tolen)
688 sys_sendto(struct sendto_args
*uap
)
690 struct thread
*td
= curthread
;
693 struct sockaddr
*sa
= NULL
;
697 error
= getsockaddr(&sa
, uap
->to
, uap
->tolen
);
701 aiov
.iov_base
= uap
->buf
;
702 aiov
.iov_len
= uap
->len
;
703 auio
.uio_iov
= &aiov
;
706 auio
.uio_resid
= uap
->len
;
707 auio
.uio_segflg
= UIO_USERSPACE
;
708 auio
.uio_rw
= UIO_WRITE
;
711 error
= kern_sendmsg(uap
->s
, sa
, &auio
, NULL
, uap
->flags
,
712 &uap
->sysmsg_result
);
720 * sendmsg_args(int s, caddr_t msg, int flags)
723 sys_sendmsg(struct sendmsg_args
*uap
)
725 struct thread
*td
= curthread
;
728 struct iovec aiov
[UIO_SMALLIOV
], *iov
= NULL
;
729 struct sockaddr
*sa
= NULL
;
730 struct mbuf
*control
= NULL
;
733 error
= copyin(uap
->msg
, (caddr_t
)&msg
, sizeof(msg
));
738 * Conditionally copyin msg.msg_name.
741 error
= getsockaddr(&sa
, msg
.msg_name
, msg
.msg_namelen
);
749 error
= iovec_copyin(msg
.msg_iov
, &iov
, aiov
, msg
.msg_iovlen
,
754 auio
.uio_iovcnt
= msg
.msg_iovlen
;
756 auio
.uio_segflg
= UIO_USERSPACE
;
757 auio
.uio_rw
= UIO_WRITE
;
761 * Conditionally copyin msg.msg_control.
763 if (msg
.msg_control
) {
764 if (msg
.msg_controllen
< sizeof(struct cmsghdr
) ||
765 msg
.msg_controllen
> MLEN
) {
769 control
= m_get(MB_WAIT
, MT_CONTROL
);
770 if (control
== NULL
) {
774 control
->m_len
= msg
.msg_controllen
;
775 error
= copyin(msg
.msg_control
, mtod(control
, caddr_t
),
783 error
= kern_sendmsg(uap
->s
, sa
, &auio
, control
, uap
->flags
,
784 &uap
->sysmsg_result
);
787 iovec_free(&iov
, aiov
);
795 * kern_recvmsg() takes a handle to sa and control. If the handle is non-
796 * null, it returns a dynamically allocated struct sockaddr and an mbuf.
797 * Don't forget to FREE() and m_free() these if they are returned.
800 kern_recvmsg(int s
, struct sockaddr
**sa
, struct uio
*auio
,
801 struct mbuf
**control
, int *flags
, int *res
)
803 struct thread
*td
= curthread
;
804 struct proc
*p
= td
->td_proc
;
810 struct iovec
*ktriov
= NULL
;
814 error
= holdsock(p
->p_fd
, s
, &fp
);
817 if (auio
->uio_resid
< 0) {
822 if (KTRPOINT(td
, KTR_GENIO
)) {
823 int iovlen
= auio
->uio_iovcnt
* sizeof (struct iovec
);
825 MALLOC(ktriov
, struct iovec
*, iovlen
, M_TEMP
, M_WAITOK
);
826 bcopy(auio
->uio_iov
, ktriov
, iovlen
);
830 len
= auio
->uio_resid
;
831 so
= (struct socket
*)fp
->f_data
;
833 if (flags
== NULL
|| (*flags
& (MSG_FNONBLOCKING
|MSG_FBLOCKING
)) == 0) {
834 if (fp
->f_flag
& FNONBLOCK
) {
836 *flags
|= MSG_FNONBLOCKING
;
838 lflags
= MSG_FNONBLOCKING
;
844 error
= so_pru_soreceive(so
, sa
, auio
, NULL
, control
, flags
);
846 if (auio
->uio_resid
!= len
&& (error
== ERESTART
||
847 error
== EINTR
|| error
== EWOULDBLOCK
))
851 if (ktriov
!= NULL
) {
853 ktruio
.uio_iov
= ktriov
;
854 ktruio
.uio_resid
= len
- auio
->uio_resid
;
855 ktrgenio(td
->td_lwp
, s
, UIO_READ
, &ktruio
, error
);
857 FREE(ktriov
, M_TEMP
);
861 *res
= len
- auio
->uio_resid
;
868 * recvfrom_args(int s, caddr_t buf, size_t len, int flags,
869 * caddr_t from, int *fromlenaddr)
872 sys_recvfrom(struct recvfrom_args
*uap
)
874 struct thread
*td
= curthread
;
877 struct sockaddr
*sa
= NULL
;
880 if (uap
->from
&& uap
->fromlenaddr
) {
881 error
= copyin(uap
->fromlenaddr
, &fromlen
, sizeof(fromlen
));
889 aiov
.iov_base
= uap
->buf
;
890 aiov
.iov_len
= uap
->len
;
891 auio
.uio_iov
= &aiov
;
894 auio
.uio_resid
= uap
->len
;
895 auio
.uio_segflg
= UIO_USERSPACE
;
896 auio
.uio_rw
= UIO_READ
;
899 error
= kern_recvmsg(uap
->s
, uap
->from
? &sa
: NULL
, &auio
, NULL
,
900 &uap
->flags
, &uap
->sysmsg_result
);
902 if (error
== 0 && uap
->from
) {
903 /* note: sa may still be NULL */
905 fromlen
= MIN(fromlen
, sa
->sa_len
);
906 error
= copyout(sa
, uap
->from
, fromlen
);
911 error
= copyout(&fromlen
, uap
->fromlenaddr
,
922 * recvmsg_args(int s, struct msghdr *msg, int flags)
925 sys_recvmsg(struct recvmsg_args
*uap
)
927 struct thread
*td
= curthread
;
930 struct iovec aiov
[UIO_SMALLIOV
], *iov
= NULL
;
931 struct mbuf
*m
, *control
= NULL
;
932 struct sockaddr
*sa
= NULL
;
934 socklen_t
*ufromlenp
, *ucontrollenp
;
935 int error
, fromlen
, controllen
, len
, flags
, *uflagsp
;
938 * This copyin handles everything except the iovec.
940 error
= copyin(uap
->msg
, &msg
, sizeof(msg
));
944 if (msg
.msg_name
&& msg
.msg_namelen
< 0)
946 if (msg
.msg_control
&& msg
.msg_controllen
< 0)
949 ufromlenp
= (socklen_t
*)((caddr_t
)uap
->msg
+ offsetof(struct msghdr
,
951 ucontrollenp
= (socklen_t
*)((caddr_t
)uap
->msg
+ offsetof(struct msghdr
,
953 uflagsp
= (int *)((caddr_t
)uap
->msg
+ offsetof(struct msghdr
,
959 error
= iovec_copyin(msg
.msg_iov
, &iov
, aiov
, msg
.msg_iovlen
,
964 auio
.uio_iovcnt
= msg
.msg_iovlen
;
966 auio
.uio_segflg
= UIO_USERSPACE
;
967 auio
.uio_rw
= UIO_READ
;
972 error
= kern_recvmsg(uap
->s
, msg
.msg_name
? &sa
: NULL
, &auio
,
973 msg
.msg_control
? &control
: NULL
, &flags
, &uap
->sysmsg_result
);
976 * Conditionally copyout the name and populate the namelen field.
978 if (error
== 0 && msg
.msg_name
) {
979 /* note: sa may still be NULL */
981 fromlen
= MIN(msg
.msg_namelen
, sa
->sa_len
);
982 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
== 0 && 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
;
1076 sopt
.sopt_val
= kmalloc(sopt
.sopt_valsize
, M_TEMP
, M_WAITOK
);
1077 error
= copyin(uap
->val
, sopt
.sopt_val
, sopt
.sopt_valsize
);
1081 sopt
.sopt_val
= NULL
;
1083 error
= kern_setsockopt(uap
->s
, &sopt
);
1087 error
= copyout(sopt
.sopt_val
, uap
->val
, sopt
.sopt_valsize
);
1090 kfree(sopt
.sopt_val
, M_TEMP
);
1095 * If sopt->sopt_td == NULL, then sopt->sopt_val is treated as an
1096 * in kernel pointer instead of a userland pointer. This allows us
1097 * to manipulate socket options in the emulation code.
1100 kern_getsockopt(int s
, struct sockopt
*sopt
)
1102 struct thread
*td
= curthread
;
1103 struct proc
*p
= td
->td_proc
;
1107 if (sopt
->sopt_val
== 0 && sopt
->sopt_valsize
!= 0)
1109 if (sopt
->sopt_valsize
< 0)
1112 error
= holdsock(p
->p_fd
, s
, &fp
);
1116 error
= sogetopt((struct socket
*)fp
->f_data
, sopt
);
1122 * getsockopt_Args(int s, int level, int name, caddr_t val, int *avalsize)
1125 sys_getsockopt(struct getsockopt_args
*uap
)
1127 struct thread
*td
= curthread
;
1128 struct sockopt sopt
;
1132 error
= copyin(uap
->avalsize
, &valsize
, sizeof(valsize
));
1141 sopt
.sopt_level
= uap
->level
;
1142 sopt
.sopt_name
= uap
->name
;
1143 sopt
.sopt_valsize
= valsize
;
1147 sopt
.sopt_val
= kmalloc(sopt
.sopt_valsize
, M_TEMP
, M_WAITOK
);
1148 error
= copyin(uap
->val
, sopt
.sopt_val
, sopt
.sopt_valsize
);
1152 sopt
.sopt_val
= NULL
;
1154 error
= kern_getsockopt(uap
->s
, &sopt
);
1157 valsize
= sopt
.sopt_valsize
;
1158 error
= copyout(&valsize
, uap
->avalsize
, sizeof(valsize
));
1162 error
= copyout(sopt
.sopt_val
, uap
->val
, sopt
.sopt_valsize
);
1165 kfree(sopt
.sopt_val
, M_TEMP
);
1170 * The second argument to kern_getsockname() is a handle to a struct sockaddr.
1171 * This allows kern_getsockname() to return a pointer to an allocated struct
1172 * sockaddr which must be freed later with FREE(). The caller must
1173 * initialize *name to NULL.
1176 kern_getsockname(int s
, struct sockaddr
**name
, int *namelen
)
1178 struct thread
*td
= curthread
;
1179 struct proc
*p
= td
->td_proc
;
1182 struct sockaddr
*sa
= NULL
;
1185 error
= holdsock(p
->p_fd
, s
, &fp
);
1192 so
= (struct socket
*)fp
->f_data
;
1193 error
= so_pru_sockaddr(so
, &sa
);
1198 *namelen
= MIN(*namelen
, sa
->sa_len
);
1208 * getsockname_args(int fdes, caddr_t asa, int *alen)
1213 sys_getsockname(struct getsockname_args
*uap
)
1215 struct sockaddr
*sa
= NULL
;
1218 error
= copyin(uap
->alen
, &sa_len
, sizeof(sa_len
));
1222 error
= kern_getsockname(uap
->fdes
, &sa
, &sa_len
);
1225 error
= copyout(sa
, uap
->asa
, sa_len
);
1227 error
= copyout(&sa_len
, uap
->alen
, sizeof(*uap
->alen
));
1234 * The second argument to kern_getpeername() is a handle to a struct sockaddr.
1235 * This allows kern_getpeername() to return a pointer to an allocated struct
1236 * sockaddr which must be freed later with FREE(). The caller must
1237 * initialize *name to NULL.
1240 kern_getpeername(int s
, struct sockaddr
**name
, int *namelen
)
1242 struct thread
*td
= curthread
;
1243 struct proc
*p
= td
->td_proc
;
1246 struct sockaddr
*sa
= NULL
;
1249 error
= holdsock(p
->p_fd
, s
, &fp
);
1256 so
= (struct socket
*)fp
->f_data
;
1257 if ((so
->so_state
& (SS_ISCONNECTED
|SS_ISCONFIRMING
)) == 0) {
1261 error
= so_pru_peeraddr(so
, &sa
);
1266 *namelen
= MIN(*namelen
, sa
->sa_len
);
1276 * getpeername_args(int fdes, caddr_t asa, int *alen)
1278 * Get name of peer for connected socket.
1281 sys_getpeername(struct getpeername_args
*uap
)
1283 struct sockaddr
*sa
= NULL
;
1286 error
= copyin(uap
->alen
, &sa_len
, sizeof(sa_len
));
1290 error
= kern_getpeername(uap
->fdes
, &sa
, &sa_len
);
1293 error
= copyout(sa
, uap
->asa
, sa_len
);
1295 error
= copyout(&sa_len
, uap
->alen
, sizeof(*uap
->alen
));
1302 getsockaddr(struct sockaddr
**namp
, caddr_t uaddr
, size_t len
)
1304 struct sockaddr
*sa
;
1308 if (len
> SOCK_MAXADDRLEN
)
1309 return ENAMETOOLONG
;
1310 if (len
< offsetof(struct sockaddr
, sa_data
[0]))
1312 MALLOC(sa
, struct sockaddr
*, len
, M_SONAME
, M_WAITOK
);
1313 error
= copyin(uaddr
, sa
, len
);
1317 #if BYTE_ORDER != BIG_ENDIAN
1319 * The bind(), connect(), and sendto() syscalls were not
1320 * versioned for COMPAT_43. Thus, this check must stay.
1322 if (sa
->sa_family
== 0 && sa
->sa_len
< AF_MAX
)
1323 sa
->sa_family
= sa
->sa_len
;
1332 * Detach a mapped page and release resources back to the system.
1333 * We must release our wiring and if the object is ripped out
1334 * from under the vm_page we become responsible for freeing the
1335 * page. These routines must be MPSAFE.
1337 * XXX HACK XXX TEMPORARY UNTIL WE IMPLEMENT EXT MBUF REFERENCE COUNTING
1339 * XXX vm_page_*() routines are not MPSAFE yet, the MP lock is required.
1342 sf_buf_mref(void *arg
)
1344 struct sfbuf_mref
*sfm
= arg
;
1347 * We must already hold a ref so there is no race to 0, just
1348 * atomically increment the count.
1350 atomic_add_int(&sfm
->mref_count
, 1);
1354 sf_buf_mfree(void *arg
)
1356 struct sfbuf_mref
*sfm
= arg
;
1359 KKASSERT(sfm
->mref_count
> 0);
1360 if (sfm
->mref_count
== 1) {
1362 * We are the only holder so no further locking is required,
1363 * the sfbuf can simply be freed.
1365 sfm
->mref_count
= 0;
1369 * There may be other holders, we must obtain the serializer
1370 * to protect against a sf_buf_mfree() race to 0. An atomic
1371 * operation is still required for races against
1374 * XXX vm_page_*() and SFBUF routines not MPSAFE yet.
1376 lwkt_serialize_enter(&sfm
->serializer
);
1377 atomic_subtract_int(&sfm
->mref_count
, 1);
1378 if (sfm
->mref_count
== 0) {
1379 lwkt_serialize_exit(&sfm
->serializer
);
1383 m
= sf_buf_page(sfm
->sf
);
1384 sf_buf_free(sfm
->sf
);
1385 vm_page_unwire(m
, 0);
1386 if (m
->wire_count
== 0 && m
->object
== NULL
)
1387 vm_page_try_to_free(m
);
1390 kfree(sfm
, M_SENDFILE
);
1392 lwkt_serialize_exit(&sfm
->serializer
);
1399 * int sendfile(int fd, int s, off_t offset, size_t nbytes,
1400 * struct sf_hdtr *hdtr, off_t *sbytes, int flags)
1402 * Send a file specified by 'fd' and starting at 'offset' to a socket
1403 * specified by 's'. Send only 'nbytes' of the file or until EOF if
1404 * nbytes == 0. Optionally add a header and/or trailer to the socket
1405 * output. If specified, write the total number of bytes sent into *sbytes.
1407 * In FreeBSD kern/uipc_syscalls.c,v 1.103, a bug was fixed that caused
1408 * the headers to count against the remaining bytes to be sent from
1409 * the file descriptor. We may wish to implement a compatibility syscall
1413 sys_sendfile(struct sendfile_args
*uap
)
1415 struct thread
*td
= curthread
;
1416 struct proc
*p
= td
->td_proc
;
1418 struct vnode
*vp
= NULL
;
1419 struct sf_hdtr hdtr
;
1420 struct iovec aiov
[UIO_SMALLIOV
], *iov
= NULL
;
1422 struct mbuf
*mheader
= NULL
;
1423 off_t hdtr_size
= 0, sbytes
;
1424 int error
, hbytes
= 0, tbytes
;
1429 * Do argument checking. Must be a regular file in, stream
1430 * type and connected socket out, positive offset.
1432 fp
= holdfp(p
->p_fd
, uap
->fd
, FREAD
);
1436 if (fp
->f_type
!= DTYPE_VNODE
) {
1440 vp
= (struct vnode
*)fp
->f_data
;
1445 * If specified, get the pointer to the sf_hdtr struct for
1446 * any headers/trailers.
1449 error
= copyin(uap
->hdtr
, &hdtr
, sizeof(hdtr
));
1456 error
= iovec_copyin(hdtr
.headers
, &iov
, aiov
,
1457 hdtr
.hdr_cnt
, &hbytes
);
1461 auio
.uio_iovcnt
= hdtr
.hdr_cnt
;
1462 auio
.uio_offset
= 0;
1463 auio
.uio_segflg
= UIO_USERSPACE
;
1464 auio
.uio_rw
= UIO_WRITE
;
1466 auio
.uio_resid
= hbytes
;
1468 mheader
= m_uiomove(&auio
);
1470 iovec_free(&iov
, aiov
);
1471 if (mheader
== NULL
)
1476 error
= kern_sendfile(vp
, uap
->s
, uap
->offset
, uap
->nbytes
, mheader
,
1477 &sbytes
, uap
->flags
);
1482 * Send trailers. Wimp out and use writev(2).
1484 if (uap
->hdtr
!= NULL
&& hdtr
.trailers
!= NULL
) {
1485 error
= iovec_copyin(hdtr
.trailers
, &iov
, aiov
,
1486 hdtr
.trl_cnt
, &auio
.uio_resid
);
1490 auio
.uio_iovcnt
= hdtr
.trl_cnt
;
1491 auio
.uio_offset
= 0;
1492 auio
.uio_segflg
= UIO_USERSPACE
;
1493 auio
.uio_rw
= UIO_WRITE
;
1496 error
= kern_sendmsg(uap
->s
, NULL
, &auio
, NULL
, 0, &tbytes
);
1498 iovec_free(&iov
, aiov
);
1501 hdtr_size
+= tbytes
; /* trailer bytes successfully sent */
1505 if (uap
->sbytes
!= NULL
) {
1506 sbytes
+= hdtr_size
;
1507 copyout(&sbytes
, uap
->sbytes
, sizeof(off_t
));
1515 kern_sendfile(struct vnode
*vp
, int sfd
, off_t offset
, size_t nbytes
,
1516 struct mbuf
*mheader
, off_t
*sbytes
, int flags
)
1518 struct thread
*td
= curthread
;
1519 struct proc
*p
= td
->td_proc
;
1520 struct vm_object
*obj
;
1525 struct sfbuf_mref
*sfm
;
1531 if (vp
->v_type
!= VREG
) {
1535 if ((obj
= vp
->v_object
) == NULL
) {
1539 error
= holdsock(p
->p_fd
, sfd
, &fp
);
1542 so
= (struct socket
*)fp
->f_data
;
1543 if (so
->so_type
!= SOCK_STREAM
) {
1547 if ((so
->so_state
& SS_ISCONNECTED
) == 0) {
1558 * Protect against multiple writers to the socket.
1560 ssb_lock(&so
->so_snd
, M_WAITOK
);
1563 * Loop through the pages in the file, starting with the requested
1564 * offset. Get a file page (do I/O if necessary), map the file page
1565 * into an sf_buf, attach an mbuf header to the sf_buf, and queue
1568 for (off
= offset
; ; off
+= xfsize
, *sbytes
+= xfsize
+ hbytes
) {
1572 pindex
= OFF_TO_IDX(off
);
1575 * Calculate the amount to transfer. Not to exceed a page,
1576 * the EOF, or the passed in nbytes.
1578 xfsize
= vp
->v_filesize
- off
;
1579 if (xfsize
> PAGE_SIZE
)
1581 pgoff
= (vm_offset_t
)(off
& PAGE_MASK
);
1582 if (PAGE_SIZE
- pgoff
< xfsize
)
1583 xfsize
= PAGE_SIZE
- pgoff
;
1584 if (nbytes
&& xfsize
> (nbytes
- *sbytes
))
1585 xfsize
= nbytes
- *sbytes
;
1589 * Optimize the non-blocking case by looking at the socket space
1590 * before going to the extra work of constituting the sf_buf.
1592 if ((fp
->f_flag
& FNONBLOCK
) && ssb_space(&so
->so_snd
) <= 0) {
1593 if (so
->so_state
& SS_CANTSENDMORE
)
1597 ssb_unlock(&so
->so_snd
);
1601 * Attempt to look up the page.
1603 * Allocate if not found, wait and loop if busy, then
1604 * wire the page. critical section protection is
1605 * required to maintain the object association (an
1606 * interrupt can free the page) through to the
1607 * vm_page_wire() call.
1610 pg
= vm_page_lookup(obj
, pindex
);
1612 pg
= vm_page_alloc(obj
, pindex
, VM_ALLOC_NORMAL
);
1619 } else if (vm_page_sleep_busy(pg
, TRUE
, "sfpbsy")) {
1627 * If page is not valid for what we need, initiate I/O
1630 if (!pg
->valid
|| !vm_page_is_valid(pg
, pgoff
, xfsize
)) {
1636 * Ensure that our page is still around when the I/O
1639 vm_page_io_start(pg
);
1642 * Get the page from backing store.
1644 bsize
= vp
->v_mount
->mnt_stat
.f_iosize
;
1645 auio
.uio_iov
= &aiov
;
1646 auio
.uio_iovcnt
= 1;
1648 aiov
.iov_len
= MAXBSIZE
;
1649 auio
.uio_resid
= MAXBSIZE
;
1650 auio
.uio_offset
= trunc_page(off
);
1651 auio
.uio_segflg
= UIO_NOCOPY
;
1652 auio
.uio_rw
= UIO_READ
;
1654 vn_lock(vp
, LK_SHARED
| LK_RETRY
);
1655 error
= VOP_READ(vp
, &auio
,
1656 IO_VMIO
| ((MAXBSIZE
/ bsize
) << 16),
1659 vm_page_flag_clear(pg
, PG_ZERO
);
1660 vm_page_io_finish(pg
);
1663 vm_page_unwire(pg
, 0);
1664 vm_page_try_to_free(pg
);
1666 ssb_unlock(&so
->so_snd
);
1673 * Get a sendfile buf. We usually wait as long as necessary,
1674 * but this wait can be interrupted.
1676 if ((sf
= sf_buf_alloc(pg
, SFB_CATCH
)) == NULL
) {
1678 vm_page_unwire(pg
, 0);
1679 vm_page_try_to_free(pg
);
1681 ssb_unlock(&so
->so_snd
);
1687 * Get an mbuf header and set it up as having external storage.
1689 MGETHDR(m
, MB_WAIT
, MT_DATA
);
1693 ssb_unlock(&so
->so_snd
);
1698 * sfm is a temporary hack, use a per-cpu cache for this.
1700 sfm
= kmalloc(sizeof(struct sfbuf_mref
), M_SENDFILE
, M_WAITOK
);
1702 sfm
->mref_count
= 1;
1703 lwkt_serialize_init(&sfm
->serializer
);
1705 m
->m_ext
.ext_free
= sf_buf_mfree
;
1706 m
->m_ext
.ext_ref
= sf_buf_mref
;
1707 m
->m_ext
.ext_arg
= sfm
;
1708 m
->m_ext
.ext_buf
= (void *)sf
->kva
;
1709 m
->m_ext
.ext_size
= PAGE_SIZE
;
1710 m
->m_data
= (char *) sf
->kva
+ pgoff
;
1711 m
->m_flags
|= M_EXT
;
1712 m
->m_pkthdr
.len
= m
->m_len
= xfsize
;
1713 KKASSERT((m
->m_flags
& (M_EXT_CLUSTER
)) == 0);
1715 if (mheader
!= NULL
) {
1716 hbytes
= mheader
->m_pkthdr
.len
;
1717 mheader
->m_pkthdr
.len
+= m
->m_pkthdr
.len
;
1725 * Add the buffer to the socket buffer chain.
1730 * Make sure that the socket is still able to take more data.
1731 * CANTSENDMORE being true usually means that the connection
1732 * was closed. so_error is true when an error was sensed after
1734 * The state is checked after the page mapping and buffer
1735 * allocation above since those operations may block and make
1736 * any socket checks stale. From this point forward, nothing
1737 * blocks before the pru_send (or more accurately, any blocking
1738 * results in a loop back to here to re-check).
1740 if ((so
->so_state
& SS_CANTSENDMORE
) || so
->so_error
) {
1741 if (so
->so_state
& SS_CANTSENDMORE
) {
1744 error
= so
->so_error
;
1748 ssb_unlock(&so
->so_snd
);
1753 * Wait for socket space to become available. We do this just
1754 * after checking the connection state above in order to avoid
1755 * a race condition with ssb_wait().
1757 if (ssb_space(&so
->so_snd
) < so
->so_snd
.ssb_lowat
) {
1758 if (fp
->f_flag
& FNONBLOCK
) {
1760 ssb_unlock(&so
->so_snd
);
1765 error
= ssb_wait(&so
->so_snd
);
1767 * An error from ssb_wait usually indicates that we've
1768 * been interrupted by a signal. If we've sent anything
1769 * then return bytes sent, otherwise return the error.
1773 ssb_unlock(&so
->so_snd
);
1779 error
= so_pru_send(so
, 0, m
, NULL
, NULL
, td
);
1782 ssb_unlock(&so
->so_snd
);
1786 if (mheader
!= NULL
) {
1787 *sbytes
+= mheader
->m_pkthdr
.len
;
1788 error
= so_pru_send(so
, 0, mheader
, NULL
, NULL
, td
);
1791 ssb_unlock(&so
->so_snd
);
1796 if (mheader
!= NULL
)
1802 sys_sctp_peeloff(struct sctp_peeloff_args
*uap
)
1805 struct thread
*td
= curthread
;
1806 struct proc
*p
= td
->td_proc
;
1807 struct file
*lfp
= NULL
;
1808 struct file
*nfp
= NULL
;
1810 struct socket
*head
, *so
;
1813 short fflag
; /* type must match fp->f_flag */
1815 assoc_id
= uap
->name
;
1816 error
= holdsock(p
->p_fd
, uap
->sd
, &lfp
);
1821 head
= (struct socket
*)lfp
->f_data
;
1822 error
= sctp_can_peel_off(head
, assoc_id
);
1828 * At this point we know we do have a assoc to pull
1829 * we proceed to get the fd setup. This may block
1833 fflag
= lfp
->f_flag
;
1834 error
= falloc(p
, &nfp
, &fd
);
1837 * Probably ran out of file descriptors. Put the
1838 * unaccepted connection back onto the queue and
1839 * do another wakeup so some other process might
1840 * have a chance at it.
1845 uap
->sysmsg_result
= fd
;
1847 so
= sctp_get_peeloff(head
, assoc_id
, &error
);
1850 * Either someone else peeled it off OR
1851 * we can't get a socket.
1855 so
->so_state
&= ~SS_COMP
;
1856 so
->so_state
&= ~SS_NOFDREF
;
1858 if (head
->so_sigio
!= NULL
)
1859 fsetown(fgetown(head
->so_sigio
), &so
->so_sigio
);
1861 nfp
->f_type
= DTYPE_SOCKET
;
1862 nfp
->f_flag
= fflag
;
1863 nfp
->f_ops
= &socketops
;
1868 * Assign the file pointer to the reserved descriptor, or clear
1869 * the reserved descriptor if an error occured.
1872 fsetfd(p
, NULL
, fd
);
1877 * Release explicitly held references before returning.