2 * Copyright (c) 2004 Jeffrey M. Hsu. All rights reserved.
3 * Copyright (c) 2004 The DragonFly Project. All rights reserved.
5 * This code is derived from software contributed to The DragonFly Project
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. Neither the name of The DragonFly Project nor the names of its
17 * contributors may be used to endorse or promote products derived
18 * from this software without specific, prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * Copyright (c) 1982, 1986, 1988, 1990, 1993
36 * The Regents of the University of California. All rights reserved.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66 * @(#)uipc_socket.c 8.3 (Berkeley) 4/15/94
67 * $FreeBSD: src/sys/kern/uipc_socket.c,v 1.68.2.24 2003/11/11 17:18:18 silby Exp $
68 * $DragonFly: src/sys/kern/uipc_socket.c,v 1.55 2008/09/02 16:17:52 dillon Exp $
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/fcntl.h>
77 #include <sys/malloc.h>
79 #include <sys/domain.h>
80 #include <sys/file.h> /* for struct knote */
81 #include <sys/kernel.h>
82 #include <sys/malloc.h>
83 #include <sys/event.h>
86 #include <sys/protosw.h>
87 #include <sys/socket.h>
88 #include <sys/socketvar.h>
89 #include <sys/socketops.h>
90 #include <sys/resourcevar.h>
91 #include <sys/signalvar.h>
92 #include <sys/sysctl.h>
95 #include <vm/vm_zone.h>
98 #include <sys/thread2.h>
99 #include <sys/socketvar2.h>
101 #include <machine/limits.h>
104 static int do_setopt_accept_filter(struct socket
*so
, struct sockopt
*sopt
);
107 static void filt_sordetach(struct knote
*kn
);
108 static int filt_soread(struct knote
*kn
, long hint
);
109 static void filt_sowdetach(struct knote
*kn
);
110 static int filt_sowrite(struct knote
*kn
, long hint
);
111 static int filt_solisten(struct knote
*kn
, long hint
);
113 static struct filterops solisten_filtops
=
114 { 1, NULL
, filt_sordetach
, filt_solisten
};
115 static struct filterops soread_filtops
=
116 { 1, NULL
, filt_sordetach
, filt_soread
};
117 static struct filterops sowrite_filtops
=
118 { 1, NULL
, filt_sowdetach
, filt_sowrite
};
120 MALLOC_DEFINE(M_SOCKET
, "socket", "socket struct");
121 MALLOC_DEFINE(M_SONAME
, "soname", "socket name");
122 MALLOC_DEFINE(M_PCB
, "pcb", "protocol control block");
125 static int somaxconn
= SOMAXCONN
;
126 SYSCTL_INT(_kern_ipc
, KIPC_SOMAXCONN
, somaxconn
, CTLFLAG_RW
,
127 &somaxconn
, 0, "Maximum pending socket connection queue size");
130 * Socket operation routines.
131 * These routines are called by the routines in
132 * sys_socket.c or from a system process, and
133 * implement the semantics of socket operations by
134 * switching out to the protocol specific routines.
138 * Get a socket structure, and initialize it.
139 * Note that it would probably be better to allocate socket
140 * and PCB at the same time, but I'm not convinced that all
141 * the protocols can be easily modified to do this.
149 waitmask
= waitok
? M_WAITOK
: M_NOWAIT
;
150 so
= kmalloc(sizeof(struct socket
), M_SOCKET
, M_ZERO
|waitmask
);
152 /* XXX race condition for reentrant kernel */
153 TAILQ_INIT(&so
->so_aiojobq
);
154 TAILQ_INIT(&so
->so_rcv
.ssb_sel
.si_mlist
);
155 TAILQ_INIT(&so
->so_snd
.ssb_sel
.si_mlist
);
161 socreate(int dom
, struct socket
**aso
, int type
,
162 int proto
, struct thread
*td
)
164 struct proc
*p
= td
->td_proc
;
167 struct pru_attach_info ai
;
171 prp
= pffindproto(dom
, proto
, type
);
173 prp
= pffindtype(dom
, type
);
175 if (prp
== 0 || prp
->pr_usrreqs
->pru_attach
== 0)
176 return (EPROTONOSUPPORT
);
178 if (p
->p_ucred
->cr_prison
&& jail_socket_unixiproute_only
&&
179 prp
->pr_domain
->dom_family
!= PF_LOCAL
&&
180 prp
->pr_domain
->dom_family
!= PF_INET
&&
181 prp
->pr_domain
->dom_family
!= PF_INET6
&&
182 prp
->pr_domain
->dom_family
!= PF_ROUTE
) {
183 return (EPROTONOSUPPORT
);
186 if (prp
->pr_type
!= type
)
188 so
= soalloc(p
!= 0);
192 TAILQ_INIT(&so
->so_incomp
);
193 TAILQ_INIT(&so
->so_comp
);
195 so
->so_cred
= crhold(p
->p_ucred
);
197 ai
.sb_rlimit
= &p
->p_rlimit
[RLIMIT_SBSIZE
];
198 ai
.p_ucred
= p
->p_ucred
;
199 ai
.fd_rdir
= p
->p_fd
->fd_rdir
;
200 error
= so_pru_attach(so
, proto
, &ai
);
202 so
->so_state
|= SS_NOFDREF
;
211 sobind(struct socket
*so
, struct sockaddr
*nam
, struct thread
*td
)
216 error
= so_pru_bind(so
, nam
, td
);
222 sodealloc(struct socket
*so
)
224 if (so
->so_rcv
.ssb_hiwat
)
225 (void)chgsbsize(so
->so_cred
->cr_uidinfo
,
226 &so
->so_rcv
.ssb_hiwat
, 0, RLIM_INFINITY
);
227 if (so
->so_snd
.ssb_hiwat
)
228 (void)chgsbsize(so
->so_cred
->cr_uidinfo
,
229 &so
->so_snd
.ssb_hiwat
, 0, RLIM_INFINITY
);
231 /* remove accept filter if present */
232 if (so
->so_accf
!= NULL
)
233 do_setopt_accept_filter(so
, NULL
);
240 solisten(struct socket
*so
, int backlog
, struct thread
*td
)
244 short oldopt
, oldqlimit
;
248 if (so
->so_state
& (SS_ISCONNECTED
| SS_ISCONNECTING
)) {
254 oldopt
= so
->so_options
;
255 oldqlimit
= so
->so_qlimit
;
258 if (TAILQ_EMPTY(&so
->so_comp
))
259 so
->so_options
|= SO_ACCEPTCONN
;
260 if (backlog
< 0 || backlog
> somaxconn
)
262 so
->so_qlimit
= backlog
;
263 /* SCTP needs to look at tweak both the inbound backlog parameter AND
264 * the so_options (UDP model both connect's and gets inbound
265 * connections .. implicitly).
267 error
= so_pru_listen(so
, td
);
270 /* Restore the params */
271 so
->so_options
= oldopt
;
272 so
->so_qlimit
= oldqlimit
;
282 * Destroy a disconnected socket. This routine is a NOP if entities
283 * still have a reference on the socket:
285 * so_pcb - The protocol stack still has a reference
286 * SS_NOFDREF - There is no longer a file pointer reference
287 * SS_ABORTING - An abort netmsg is in-flight
290 sofree(struct socket
*so
)
292 struct socket
*head
= so
->so_head
;
294 if (so
->so_pcb
|| (so
->so_state
& SS_NOFDREF
) == 0)
296 if (so
->so_state
& SS_ABORTING
)
299 if (so
->so_state
& SS_INCOMP
) {
300 TAILQ_REMOVE(&head
->so_incomp
, so
, so_list
);
302 } else if (so
->so_state
& SS_COMP
) {
304 * We must not decommission a socket that's
305 * on the accept(2) queue. If we do, then
306 * accept(2) may hang after select(2) indicated
307 * that the listening socket was ready.
311 panic("sofree: not queued");
313 so
->so_state
&= ~SS_INCOMP
;
316 ssb_release(&so
->so_snd
, so
);
322 * Close a socket on last file table reference removal.
323 * Initiate disconnect if connected.
324 * Free socket when disconnect complete.
327 soclose(struct socket
*so
, int fflag
)
332 funsetown(so
->so_sigio
);
333 if (so
->so_pcb
== NULL
)
335 if (so
->so_state
& SS_ISCONNECTED
) {
336 if ((so
->so_state
& SS_ISDISCONNECTING
) == 0) {
337 error
= sodisconnect(so
);
341 if (so
->so_options
& SO_LINGER
) {
342 if ((so
->so_state
& SS_ISDISCONNECTING
) &&
345 while (so
->so_state
& SS_ISCONNECTED
) {
346 error
= tsleep((caddr_t
)&so
->so_timeo
,
347 PCATCH
, "soclos", so
->so_linger
* hz
);
357 error2
= so_pru_detach(so
);
362 if (so
->so_options
& SO_ACCEPTCONN
) {
365 while ((sp
= TAILQ_FIRST(&so
->so_incomp
)) != NULL
) {
366 TAILQ_REMOVE(&so
->so_incomp
, sp
, so_list
);
367 sp
->so_state
&= ~SS_INCOMP
;
372 while ((sp
= TAILQ_FIRST(&so
->so_comp
)) != NULL
) {
373 TAILQ_REMOVE(&so
->so_comp
, sp
, so_list
);
374 sp
->so_state
&= ~SS_COMP
;
380 if (so
->so_state
& SS_NOFDREF
)
381 panic("soclose: NOFDREF");
382 so
->so_state
|= SS_NOFDREF
;
389 * Abort and destroy a socket. Only one abort can be in progress
390 * at any given moment.
393 soabort(struct socket
*so
)
395 if ((so
->so_state
& SS_ABORTING
) == 0) {
396 so
->so_state
|= SS_ABORTING
;
402 soaborta(struct socket
*so
)
404 if ((so
->so_state
& SS_ABORTING
) == 0) {
405 so
->so_state
|= SS_ABORTING
;
411 soabort_oncpu(struct socket
*so
)
413 if ((so
->so_state
& SS_ABORTING
) == 0) {
414 so
->so_state
|= SS_ABORTING
;
415 so_pru_abort_oncpu(so
);
420 soaccept(struct socket
*so
, struct sockaddr
**nam
)
425 if ((so
->so_state
& SS_NOFDREF
) == 0)
426 panic("soaccept: !NOFDREF");
427 so
->so_state
&= ~SS_NOFDREF
;
428 error
= so_pru_accept(so
, nam
);
434 soconnect(struct socket
*so
, struct sockaddr
*nam
, struct thread
*td
)
438 if (so
->so_options
& SO_ACCEPTCONN
)
442 * If protocol is connection-based, can only connect once.
443 * Otherwise, if connected, try to disconnect first.
444 * This allows user to disconnect by connecting to, e.g.,
447 if (so
->so_state
& (SS_ISCONNECTED
|SS_ISCONNECTING
) &&
448 ((so
->so_proto
->pr_flags
& PR_CONNREQUIRED
) ||
449 (error
= sodisconnect(so
)))) {
453 * Prevent accumulated error from previous connection
457 error
= so_pru_connect(so
, nam
, td
);
464 soconnect2(struct socket
*so1
, struct socket
*so2
)
469 error
= so_pru_connect2(so1
, so2
);
475 sodisconnect(struct socket
*so
)
480 if ((so
->so_state
& SS_ISCONNECTED
) == 0) {
484 if (so
->so_state
& SS_ISDISCONNECTING
) {
488 error
= so_pru_disconnect(so
);
494 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
497 * If send must go all at once and message is larger than
498 * send buffering, then hard error.
499 * Lock against other senders.
500 * If must go all at once and not enough room now, then
501 * inform user that this would block and do nothing.
502 * Otherwise, if nonblocking, send as much as possible.
503 * The data to be sent is described by "uio" if nonzero,
504 * otherwise by the mbuf chain "top" (which must be null
505 * if uio is not). Data provided in mbuf chain must be small
506 * enough to send all at once.
508 * Returns nonzero on error, timeout or signal; callers
509 * must check for short counts if EINTR/ERESTART are returned.
510 * Data and control buffers are freed on return.
513 sosend(struct socket
*so
, struct sockaddr
*addr
, struct uio
*uio
,
514 struct mbuf
*top
, struct mbuf
*control
, int flags
,
519 long space
, len
, resid
;
520 int clen
= 0, error
, dontroute
, mlen
;
521 int atomic
= sosendallatonce(so
) || top
;
525 resid
= uio
->uio_resid
;
527 resid
= top
->m_pkthdr
.len
;
529 * In theory resid should be unsigned.
530 * However, space must be signed, as it might be less than 0
531 * if we over-committed, and we must use a signed comparison
532 * of space and resid. On the other hand, a negative resid
533 * causes us to loop sending 0-length segments to the protocol.
535 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
536 * type sockets since that's an error.
538 if (resid
< 0 || (so
->so_type
== SOCK_STREAM
&& (flags
& MSG_EOR
))) {
544 (flags
& MSG_DONTROUTE
) && (so
->so_options
& SO_DONTROUTE
) == 0 &&
545 (so
->so_proto
->pr_flags
& PR_ATOMIC
);
546 if (td
->td_lwp
!= NULL
)
547 td
->td_lwp
->lwp_ru
.ru_msgsnd
++;
549 clen
= control
->m_len
;
550 #define gotoerr(errcode) { error = errcode; crit_exit(); goto release; }
553 error
= ssb_lock(&so
->so_snd
, SBLOCKWAIT(flags
));
558 if (so
->so_state
& SS_CANTSENDMORE
)
561 error
= so
->so_error
;
566 if ((so
->so_state
& SS_ISCONNECTED
) == 0) {
568 * `sendto' and `sendmsg' is allowed on a connection-
569 * based socket if it supports implied connect.
570 * Return ENOTCONN if not connected and no address is
573 if ((so
->so_proto
->pr_flags
& PR_CONNREQUIRED
) &&
574 (so
->so_proto
->pr_flags
& PR_IMPLOPCL
) == 0) {
575 if ((so
->so_state
& SS_ISCONFIRMING
) == 0 &&
576 !(resid
== 0 && clen
!= 0))
578 } else if (addr
== 0)
579 gotoerr(so
->so_proto
->pr_flags
& PR_CONNREQUIRED
?
580 ENOTCONN
: EDESTADDRREQ
);
582 if ((atomic
&& resid
> so
->so_snd
.ssb_hiwat
) ||
583 clen
> so
->so_snd
.ssb_hiwat
) {
586 space
= ssb_space(&so
->so_snd
);
589 if (space
< resid
+ clen
&& uio
&&
590 (atomic
|| space
< so
->so_snd
.ssb_lowat
|| space
< clen
)) {
591 if (flags
& (MSG_FNONBLOCKING
|MSG_DONTWAIT
))
592 gotoerr(EWOULDBLOCK
);
593 ssb_unlock(&so
->so_snd
);
594 error
= ssb_wait(&so
->so_snd
);
606 * Data is prepackaged in "top".
610 top
->m_flags
|= M_EOR
;
612 m
= m_getl(resid
, MB_WAIT
, MT_DATA
,
613 top
== NULL
? M_PKTHDR
: 0, &mlen
);
616 m
->m_pkthdr
.rcvif
= (struct ifnet
*)0;
618 len
= min(min(mlen
, resid
), space
);
619 if (resid
< MINCLSIZE
) {
621 * For datagram protocols, leave room
622 * for protocol headers in first mbuf.
624 if (atomic
&& top
== 0 && len
< mlen
)
628 error
= uiomove(mtod(m
, caddr_t
), (int)len
, uio
);
629 resid
= uio
->uio_resid
;
632 top
->m_pkthdr
.len
+= len
;
638 top
->m_flags
|= M_EOR
;
641 } while (space
> 0 && atomic
);
643 so
->so_options
|= SO_DONTROUTE
;
644 if (flags
& MSG_OOB
) {
645 pru_flags
= PRUS_OOB
;
646 } else if ((flags
& MSG_EOF
) &&
647 (so
->so_proto
->pr_flags
& PR_IMPLOPCL
) &&
650 * If the user set MSG_EOF, the protocol
651 * understands this flag and nothing left to
652 * send then use PRU_SEND_EOF instead of PRU_SEND.
654 pru_flags
= PRUS_EOF
;
655 } else if (resid
> 0 && space
> 0) {
656 /* If there is more to send, set PRUS_MORETOCOME */
657 pru_flags
= PRUS_MORETOCOME
;
663 * XXX all the SS_CANTSENDMORE checks previously
664 * done could be out of date. We could have recieved
665 * a reset packet in an interrupt or maybe we slept
666 * while doing page faults in uiomove() etc. We could
667 * probably recheck again inside the splnet() protection
668 * here, but there are probably other places that this
669 * also happens. We must rethink this.
671 error
= so_pru_send(so
, pru_flags
, top
, addr
, control
, td
);
674 so
->so_options
&= ~SO_DONTROUTE
;
681 } while (resid
&& space
> 0);
685 ssb_unlock(&so
->so_snd
);
695 * A specialization of sosend() for UDP based on protocol-specific knowledge:
696 * so->so_proto->pr_flags has the PR_ATOMIC field set. This means that
697 * sosendallatonce() returns true,
698 * the "atomic" variable is true,
699 * and sosendudp() blocks until space is available for the entire send.
700 * so->so_proto->pr_flags does not have the PR_CONNREQUIRED or
701 * PR_IMPLOPCL flags set.
702 * UDP has no out-of-band data.
703 * UDP has no control data.
704 * UDP does not support MSG_EOR.
707 sosendudp(struct socket
*so
, struct sockaddr
*addr
, struct uio
*uio
,
708 struct mbuf
*top
, struct mbuf
*control
, int flags
, struct thread
*td
)
711 boolean_t dontroute
; /* temporary SO_DONTROUTE setting */
713 if (td
->td_lwp
!= NULL
)
714 td
->td_lwp
->lwp_ru
.ru_msgsnd
++;
718 KASSERT((uio
&& !top
) || (top
&& !uio
), ("bad arguments to sosendudp"));
719 resid
= uio
? uio
->uio_resid
: top
->m_pkthdr
.len
;
722 error
= ssb_lock(&so
->so_snd
, SBLOCKWAIT(flags
));
727 if (so
->so_state
& SS_CANTSENDMORE
)
730 error
= so
->so_error
;
735 if (!(so
->so_state
& SS_ISCONNECTED
) && addr
== NULL
)
736 gotoerr(EDESTADDRREQ
);
737 if (resid
> so
->so_snd
.ssb_hiwat
)
739 if (uio
&& ssb_space(&so
->so_snd
) < resid
) {
740 if (flags
& (MSG_FNONBLOCKING
|MSG_DONTWAIT
))
741 gotoerr(EWOULDBLOCK
);
742 ssb_unlock(&so
->so_snd
);
743 error
= ssb_wait(&so
->so_snd
);
752 top
= m_uiomove(uio
);
757 dontroute
= (flags
& MSG_DONTROUTE
) && !(so
->so_options
& SO_DONTROUTE
);
759 so
->so_options
|= SO_DONTROUTE
;
761 error
= so_pru_send(so
, 0, top
, addr
, NULL
, td
);
762 top
= NULL
; /* sent or freed in lower layer */
765 so
->so_options
&= ~SO_DONTROUTE
;
768 ssb_unlock(&so
->so_snd
);
776 * Implement receive operations on a socket.
777 * We depend on the way that records are added to the signalsockbuf
778 * by sbappend*. In particular, each record (mbufs linked through m_next)
779 * must begin with an address if the protocol so specifies,
780 * followed by an optional mbuf or mbufs containing ancillary data,
781 * and then zero or more mbufs of data.
782 * In order to avoid blocking network interrupts for the entire time here,
783 * we exit the critical section while doing the actual copy to user space.
784 * Although the signalsockbuf is locked, new data may still be appended,
785 * and thus we must maintain consistency of the signalsockbuf during that time.
787 * The caller may receive the data as a single mbuf chain by supplying
788 * an mbuf **mp0 for use in returning the chain. The uio is then used
789 * only for the count in uio_resid.
792 soreceive(struct socket
*so
, struct sockaddr
**psa
, struct uio
*uio
,
793 struct sockbuf
*sio
, struct mbuf
**controlp
, int *flagsp
)
796 struct mbuf
*free_chain
= NULL
;
797 int flags
, len
, error
, offset
;
798 struct protosw
*pr
= so
->so_proto
;
800 int resid
, orig_resid
;
803 resid
= uio
->uio_resid
;
805 resid
= (int)(sio
->sb_climit
- sio
->sb_cc
);
813 flags
= *flagsp
&~ MSG_EOR
;
816 if (flags
& MSG_OOB
) {
817 m
= m_get(MB_WAIT
, MT_DATA
);
820 error
= so_pru_rcvoob(so
, m
, flags
& MSG_PEEK
);
827 } while (resid
> 0 && m
);
830 uio
->uio_resid
= resid
;
831 error
= uiomove(mtod(m
, caddr_t
),
832 (int)min(resid
, m
->m_len
), uio
);
833 resid
= uio
->uio_resid
;
835 } while (uio
->uio_resid
&& error
== 0 && m
);
842 if (so
->so_state
& SS_ISCONFIRMING
&& resid
)
847 error
= ssb_lock(&so
->so_rcv
, SBLOCKWAIT(flags
));
851 m
= so
->so_rcv
.ssb_mb
;
853 * If we have less data than requested, block awaiting more
854 * (subject to any timeout) if:
855 * 1. the current count is less than the low water mark, or
856 * 2. MSG_WAITALL is set, and it is possible to do the entire
857 * receive operation at once if we block (resid <= hiwat).
858 * 3. MSG_DONTWAIT is not set
859 * If MSG_WAITALL is set but resid is larger than the receive buffer,
860 * we have to do the receive in sections, and thus risk returning
861 * a short count if a timeout or signal occurs after we start.
863 if (m
== NULL
|| (((flags
& MSG_DONTWAIT
) == 0 &&
864 so
->so_rcv
.ssb_cc
< resid
) &&
865 (so
->so_rcv
.ssb_cc
< so
->so_rcv
.ssb_lowat
||
866 ((flags
& MSG_WAITALL
) && resid
<= so
->so_rcv
.ssb_hiwat
)) &&
867 m
->m_nextpkt
== 0 && (pr
->pr_flags
& PR_ATOMIC
) == 0)) {
868 KASSERT(m
!= NULL
|| !so
->so_rcv
.ssb_cc
, ("receive 1"));
872 error
= so
->so_error
;
873 if ((flags
& MSG_PEEK
) == 0)
877 if (so
->so_state
& SS_CANTRCVMORE
) {
883 for (; m
; m
= m
->m_next
) {
884 if (m
->m_type
== MT_OOBDATA
|| (m
->m_flags
& M_EOR
)) {
885 m
= so
->so_rcv
.ssb_mb
;
889 if ((so
->so_state
& (SS_ISCONNECTED
|SS_ISCONNECTING
)) == 0 &&
890 (pr
->pr_flags
& PR_CONNREQUIRED
)) {
896 if (flags
& (MSG_FNONBLOCKING
|MSG_DONTWAIT
)) {
900 ssb_unlock(&so
->so_rcv
);
901 error
= ssb_wait(&so
->so_rcv
);
908 if (uio
&& uio
->uio_td
&& uio
->uio_td
->td_proc
)
909 uio
->uio_td
->td_lwp
->lwp_ru
.ru_msgrcv
++;
912 * note: m should be == sb_mb here. Cache the next record while
913 * cleaning up. Note that calling m_free*() will break out critical
916 KKASSERT(m
== so
->so_rcv
.ssb_mb
);
919 * Skip any address mbufs prepending the record.
921 if (pr
->pr_flags
& PR_ADDR
) {
922 KASSERT(m
->m_type
== MT_SONAME
, ("receive 1a"));
925 *psa
= dup_sockaddr(mtod(m
, struct sockaddr
*));
926 if (flags
& MSG_PEEK
)
929 m
= sbunlinkmbuf(&so
->so_rcv
.sb
, m
, &free_chain
);
933 * Skip any control mbufs prepending the record.
936 if (pr
->pr_flags
& PR_ADDR_OPT
) {
938 * For SCTP we may be getting a
939 * whole message OR a partial delivery.
941 if (m
&& m
->m_type
== MT_SONAME
) {
944 *psa
= dup_sockaddr(mtod(m
, struct sockaddr
*));
945 if (flags
& MSG_PEEK
)
948 m
= sbunlinkmbuf(&so
->so_rcv
.sb
, m
, &free_chain
);
952 while (m
&& m
->m_type
== MT_CONTROL
&& error
== 0) {
953 if (flags
& MSG_PEEK
) {
955 *controlp
= m_copy(m
, 0, m
->m_len
);
956 m
= m
->m_next
; /* XXX race */
959 n
= sbunlinkmbuf(&so
->so_rcv
.sb
, m
, NULL
);
960 if (pr
->pr_domain
->dom_externalize
&&
961 mtod(m
, struct cmsghdr
*)->cmsg_type
==
963 error
= (*pr
->pr_domain
->dom_externalize
)(m
);
967 m
= sbunlinkmbuf(&so
->so_rcv
.sb
, m
, &free_chain
);
970 if (controlp
&& *controlp
) {
972 controlp
= &(*controlp
)->m_next
;
981 if (type
== MT_OOBDATA
)
986 * Copy to the UIO or mbuf return chain (*mp).
990 while (m
&& resid
> 0 && error
== 0) {
991 if (m
->m_type
== MT_OOBDATA
) {
992 if (type
!= MT_OOBDATA
)
994 } else if (type
== MT_OOBDATA
)
997 KASSERT(m
->m_type
== MT_DATA
|| m
->m_type
== MT_HEADER
,
999 so
->so_state
&= ~SS_RCVATMARK
;
1001 if (so
->so_oobmark
&& len
> so
->so_oobmark
- offset
)
1002 len
= so
->so_oobmark
- offset
;
1003 if (len
> m
->m_len
- moff
)
1004 len
= m
->m_len
- moff
;
1007 * Copy out to the UIO or pass the mbufs back to the SIO.
1008 * The SIO is dealt with when we eat the mbuf, but deal
1009 * with the resid here either way.
1013 uio
->uio_resid
= resid
;
1014 error
= uiomove(mtod(m
, caddr_t
) + moff
, len
, uio
);
1015 resid
= uio
->uio_resid
;
1024 * Eat the entire mbuf or just a piece of it
1026 if (len
== m
->m_len
- moff
) {
1027 if (m
->m_flags
& M_EOR
)
1030 if (m
->m_flags
& M_NOTIFICATION
)
1031 flags
|= MSG_NOTIFICATION
;
1033 if (flags
& MSG_PEEK
) {
1038 n
= sbunlinkmbuf(&so
->so_rcv
.sb
, m
, NULL
);
1042 m
= sbunlinkmbuf(&so
->so_rcv
.sb
, m
, &free_chain
);
1046 if (flags
& MSG_PEEK
) {
1050 n
= m_copym(m
, 0, len
, MB_WAIT
);
1056 so
->so_rcv
.ssb_cc
-= len
;
1059 if (so
->so_oobmark
) {
1060 if ((flags
& MSG_PEEK
) == 0) {
1061 so
->so_oobmark
-= len
;
1062 if (so
->so_oobmark
== 0) {
1063 so
->so_state
|= SS_RCVATMARK
;
1068 if (offset
== so
->so_oobmark
)
1072 if (flags
& MSG_EOR
)
1075 * If the MSG_WAITALL flag is set (for non-atomic socket),
1076 * we must not quit until resid == 0 or an error
1077 * termination. If a signal/timeout occurs, return
1078 * with a short count but without error.
1079 * Keep signalsockbuf locked against other readers.
1081 while ((flags
& MSG_WAITALL
) && m
== NULL
&&
1082 resid
> 0 && !sosendallatonce(so
) &&
1083 so
->so_rcv
.ssb_mb
== NULL
) {
1084 if (so
->so_error
|| so
->so_state
& SS_CANTRCVMORE
)
1087 * The window might have closed to zero, make
1088 * sure we send an ack now that we've drained
1089 * the buffer or we might end up blocking until
1090 * the idle takes over (5 seconds).
1092 if (pr
->pr_flags
& PR_WANTRCVD
&& so
->so_pcb
)
1093 so_pru_rcvd(so
, flags
);
1094 error
= ssb_wait(&so
->so_rcv
);
1096 ssb_unlock(&so
->so_rcv
);
1100 m
= so
->so_rcv
.ssb_mb
;
1105 * If an atomic read was requested but unread data still remains
1106 * in the record, set MSG_TRUNC.
1108 if (m
&& pr
->pr_flags
& PR_ATOMIC
)
1112 * Cleanup. If an atomic read was requested drop any unread data.
1114 if ((flags
& MSG_PEEK
) == 0) {
1115 if (m
&& (pr
->pr_flags
& PR_ATOMIC
))
1116 sbdroprecord(&so
->so_rcv
.sb
);
1117 if ((pr
->pr_flags
& PR_WANTRCVD
) && so
->so_pcb
)
1118 so_pru_rcvd(so
, flags
);
1121 if (orig_resid
== resid
&& orig_resid
&&
1122 (flags
& MSG_EOR
) == 0 && (so
->so_state
& SS_CANTRCVMORE
) == 0) {
1123 ssb_unlock(&so
->so_rcv
);
1131 ssb_unlock(&so
->so_rcv
);
1135 m_freem(free_chain
);
1140 soshutdown(struct socket
*so
, int how
)
1142 if (!(how
== SHUT_RD
|| how
== SHUT_WR
|| how
== SHUT_RDWR
))
1148 return (so_pru_shutdown(so
));
1153 sorflush(struct socket
*so
)
1155 struct signalsockbuf
*ssb
= &so
->so_rcv
;
1156 struct protosw
*pr
= so
->so_proto
;
1157 struct signalsockbuf asb
;
1159 ssb
->ssb_flags
|= SSB_NOINTR
;
1160 (void) ssb_lock(ssb
, M_WAITOK
);
1166 bzero((caddr_t
)ssb
, sizeof (*ssb
));
1167 if (asb
.ssb_flags
& SSB_KNOTE
) {
1168 ssb
->ssb_sel
.si_note
= asb
.ssb_sel
.si_note
;
1169 ssb
->ssb_flags
= SSB_KNOTE
;
1173 if (pr
->pr_flags
& PR_RIGHTS
&& pr
->pr_domain
->dom_dispose
)
1174 (*pr
->pr_domain
->dom_dispose
)(asb
.ssb_mb
);
1175 ssb_release(&asb
, so
);
1180 do_setopt_accept_filter(struct socket
*so
, struct sockopt
*sopt
)
1182 struct accept_filter_arg
*afap
= NULL
;
1183 struct accept_filter
*afp
;
1184 struct so_accf
*af
= so
->so_accf
;
1187 /* do not set/remove accept filters on non listen sockets */
1188 if ((so
->so_options
& SO_ACCEPTCONN
) == 0) {
1193 /* removing the filter */
1196 if (af
->so_accept_filter
!= NULL
&&
1197 af
->so_accept_filter
->accf_destroy
!= NULL
) {
1198 af
->so_accept_filter
->accf_destroy(so
);
1200 if (af
->so_accept_filter_str
!= NULL
) {
1201 FREE(af
->so_accept_filter_str
, M_ACCF
);
1206 so
->so_options
&= ~SO_ACCEPTFILTER
;
1209 /* adding a filter */
1210 /* must remove previous filter first */
1215 /* don't put large objects on the kernel stack */
1216 MALLOC(afap
, struct accept_filter_arg
*, sizeof(*afap
), M_TEMP
, M_WAITOK
);
1217 error
= sooptcopyin(sopt
, afap
, sizeof *afap
, sizeof *afap
);
1218 afap
->af_name
[sizeof(afap
->af_name
)-1] = '\0';
1219 afap
->af_arg
[sizeof(afap
->af_arg
)-1] = '\0';
1222 afp
= accept_filt_get(afap
->af_name
);
1227 MALLOC(af
, struct so_accf
*, sizeof(*af
), M_ACCF
, M_WAITOK
| M_ZERO
);
1228 if (afp
->accf_create
!= NULL
) {
1229 if (afap
->af_name
[0] != '\0') {
1230 int len
= strlen(afap
->af_name
) + 1;
1232 MALLOC(af
->so_accept_filter_str
, char *, len
, M_ACCF
, M_WAITOK
);
1233 strcpy(af
->so_accept_filter_str
, afap
->af_name
);
1235 af
->so_accept_filter_arg
= afp
->accf_create(so
, afap
->af_arg
);
1236 if (af
->so_accept_filter_arg
== NULL
) {
1237 FREE(af
->so_accept_filter_str
, M_ACCF
);
1244 af
->so_accept_filter
= afp
;
1246 so
->so_options
|= SO_ACCEPTFILTER
;
1255 * Perhaps this routine, and sooptcopyout(), below, ought to come in
1256 * an additional variant to handle the case where the option value needs
1257 * to be some kind of integer, but not a specific size.
1258 * In addition to their use here, these functions are also called by the
1259 * protocol-level pr_ctloutput() routines.
1262 sooptcopyin(struct sockopt
*sopt
, void *buf
, size_t len
, size_t minlen
)
1264 return soopt_to_kbuf(sopt
, buf
, len
, minlen
);
1268 soopt_to_kbuf(struct sockopt
*sopt
, void *buf
, size_t len
, size_t minlen
)
1272 KKASSERT(!sopt
->sopt_val
|| kva_p(sopt
->sopt_val
));
1273 KKASSERT(kva_p(buf
));
1276 * If the user gives us more than we wanted, we ignore it,
1277 * but if we don't get the minimum length the caller
1278 * wants, we return EINVAL. On success, sopt->sopt_valsize
1279 * is set to however much we actually retrieved.
1281 if ((valsize
= sopt
->sopt_valsize
) < minlen
)
1284 sopt
->sopt_valsize
= valsize
= len
;
1286 bcopy(sopt
->sopt_val
, buf
, valsize
);
1292 sosetopt(struct socket
*so
, struct sockopt
*sopt
)
1300 sopt
->sopt_dir
= SOPT_SET
;
1301 if (sopt
->sopt_level
!= SOL_SOCKET
) {
1302 if (so
->so_proto
&& so
->so_proto
->pr_ctloutput
) {
1303 return (so_pru_ctloutput(so
, sopt
));
1305 error
= ENOPROTOOPT
;
1307 switch (sopt
->sopt_name
) {
1309 case SO_ACCEPTFILTER
:
1310 error
= do_setopt_accept_filter(so
, sopt
);
1316 error
= sooptcopyin(sopt
, &l
, sizeof l
, sizeof l
);
1320 so
->so_linger
= l
.l_linger
;
1322 so
->so_options
|= SO_LINGER
;
1324 so
->so_options
&= ~SO_LINGER
;
1330 case SO_USELOOPBACK
:
1336 error
= sooptcopyin(sopt
, &optval
, sizeof optval
,
1341 so
->so_options
|= sopt
->sopt_name
;
1343 so
->so_options
&= ~sopt
->sopt_name
;
1350 error
= sooptcopyin(sopt
, &optval
, sizeof optval
,
1356 * Values < 1 make no sense for any of these
1357 * options, so disallow them.
1364 switch (sopt
->sopt_name
) {
1367 if (ssb_reserve(sopt
->sopt_name
== SO_SNDBUF
?
1368 &so
->so_snd
: &so
->so_rcv
, (u_long
)optval
,
1370 &curproc
->p_rlimit
[RLIMIT_SBSIZE
]) == 0) {
1377 * Make sure the low-water is never greater than
1381 so
->so_snd
.ssb_lowat
=
1382 (optval
> so
->so_snd
.ssb_hiwat
) ?
1383 so
->so_snd
.ssb_hiwat
: optval
;
1386 so
->so_rcv
.ssb_lowat
=
1387 (optval
> so
->so_rcv
.ssb_hiwat
) ?
1388 so
->so_rcv
.ssb_hiwat
: optval
;
1395 error
= sooptcopyin(sopt
, &tv
, sizeof tv
,
1400 /* assert(hz > 0); */
1401 if (tv
.tv_sec
< 0 || tv
.tv_sec
> SHRT_MAX
/ hz
||
1402 tv
.tv_usec
< 0 || tv
.tv_usec
>= 1000000) {
1406 /* assert(tick > 0); */
1407 /* assert(ULONG_MAX - SHRT_MAX >= 1000000); */
1408 val
= (u_long
)(tv
.tv_sec
* hz
) + tv
.tv_usec
/ tick
;
1409 if (val
> SHRT_MAX
) {
1413 if (val
== 0 && tv
.tv_usec
!= 0)
1416 switch (sopt
->sopt_name
) {
1418 so
->so_snd
.ssb_timeo
= val
;
1421 so
->so_rcv
.ssb_timeo
= val
;
1426 error
= ENOPROTOOPT
;
1429 if (error
== 0 && so
->so_proto
&& so
->so_proto
->pr_ctloutput
) {
1430 (void) so_pru_ctloutput(so
, sopt
);
1437 /* Helper routine for getsockopt */
1439 sooptcopyout(struct sockopt
*sopt
, const void *buf
, size_t len
)
1441 soopt_from_kbuf(sopt
, buf
, len
);
1446 soopt_from_kbuf(struct sockopt
*sopt
, const void *buf
, size_t len
)
1450 KKASSERT(!sopt
->sopt_val
|| kva_p(sopt
->sopt_val
));
1451 KKASSERT(kva_p(buf
));
1454 * Documented get behavior is that we always return a value,
1455 * possibly truncated to fit in the user's buffer.
1456 * Traditional behavior is that we always tell the user
1457 * precisely how much we copied, rather than something useful
1458 * like the total amount we had available for her.
1459 * Note that this interface is not idempotent; the entire answer must
1460 * generated ahead of time.
1462 valsize
= min(len
, sopt
->sopt_valsize
);
1463 sopt
->sopt_valsize
= valsize
;
1464 if (sopt
->sopt_val
!= 0) {
1465 bcopy(buf
, sopt
->sopt_val
, valsize
);
1470 sogetopt(struct socket
*so
, struct sockopt
*sopt
)
1476 struct accept_filter_arg
*afap
;
1480 sopt
->sopt_dir
= SOPT_GET
;
1481 if (sopt
->sopt_level
!= SOL_SOCKET
) {
1482 if (so
->so_proto
&& so
->so_proto
->pr_ctloutput
) {
1483 return (so_pru_ctloutput(so
, sopt
));
1485 return (ENOPROTOOPT
);
1487 switch (sopt
->sopt_name
) {
1489 case SO_ACCEPTFILTER
:
1490 if ((so
->so_options
& SO_ACCEPTCONN
) == 0)
1492 MALLOC(afap
, struct accept_filter_arg
*, sizeof(*afap
),
1493 M_TEMP
, M_WAITOK
| M_ZERO
);
1494 if ((so
->so_options
& SO_ACCEPTFILTER
) != 0) {
1495 strcpy(afap
->af_name
, so
->so_accf
->so_accept_filter
->accf_name
);
1496 if (so
->so_accf
->so_accept_filter_str
!= NULL
)
1497 strcpy(afap
->af_arg
, so
->so_accf
->so_accept_filter_str
);
1499 error
= sooptcopyout(sopt
, afap
, sizeof(*afap
));
1505 l
.l_onoff
= so
->so_options
& SO_LINGER
;
1506 l
.l_linger
= so
->so_linger
;
1507 error
= sooptcopyout(sopt
, &l
, sizeof l
);
1510 case SO_USELOOPBACK
:
1519 optval
= so
->so_options
& sopt
->sopt_name
;
1521 error
= sooptcopyout(sopt
, &optval
, sizeof optval
);
1525 optval
= so
->so_type
;
1529 optval
= so
->so_error
;
1534 optval
= so
->so_snd
.ssb_hiwat
;
1538 optval
= so
->so_rcv
.ssb_hiwat
;
1542 optval
= so
->so_snd
.ssb_lowat
;
1546 optval
= so
->so_rcv
.ssb_lowat
;
1551 optval
= (sopt
->sopt_name
== SO_SNDTIMEO
?
1552 so
->so_snd
.ssb_timeo
: so
->so_rcv
.ssb_timeo
);
1554 tv
.tv_sec
= optval
/ hz
;
1555 tv
.tv_usec
= (optval
% hz
) * tick
;
1556 error
= sooptcopyout(sopt
, &tv
, sizeof tv
);
1560 error
= ENOPROTOOPT
;
1567 /* XXX; prepare mbuf for (__FreeBSD__ < 3) routines. */
1569 soopt_getm(struct sockopt
*sopt
, struct mbuf
**mp
)
1571 struct mbuf
*m
, *m_prev
;
1572 int sopt_size
= sopt
->sopt_valsize
, msize
;
1574 m
= m_getl(sopt_size
, sopt
->sopt_td
? MB_WAIT
: MB_DONTWAIT
, MT_DATA
,
1578 m
->m_len
= min(msize
, sopt_size
);
1579 sopt_size
-= m
->m_len
;
1583 while (sopt_size
> 0) {
1584 m
= m_getl(sopt_size
, sopt
->sopt_td
? MB_WAIT
: MB_DONTWAIT
,
1585 MT_DATA
, 0, &msize
);
1590 m
->m_len
= min(msize
, sopt_size
);
1591 sopt_size
-= m
->m_len
;
1598 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
1600 soopt_mcopyin(struct sockopt
*sopt
, struct mbuf
*m
)
1602 soopt_to_mbuf(sopt
, m
);
1607 soopt_to_mbuf(struct sockopt
*sopt
, struct mbuf
*m
)
1612 KKASSERT(!sopt
->sopt_val
|| kva_p(sopt
->sopt_val
));
1614 if (sopt
->sopt_val
== NULL
)
1616 val
= sopt
->sopt_val
;
1617 valsize
= sopt
->sopt_valsize
;
1618 while (m
!= NULL
&& valsize
>= m
->m_len
) {
1619 bcopy(val
, mtod(m
, char *), m
->m_len
);
1620 valsize
-= m
->m_len
;
1621 val
= (caddr_t
)val
+ m
->m_len
;
1624 if (m
!= NULL
) /* should be allocated enoughly at ip6_sooptmcopyin() */
1625 panic("ip6_sooptmcopyin");
1628 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
1630 soopt_mcopyout(struct sockopt
*sopt
, struct mbuf
*m
)
1632 return soopt_from_mbuf(sopt
, m
);
1636 soopt_from_mbuf(struct sockopt
*sopt
, struct mbuf
*m
)
1638 struct mbuf
*m0
= m
;
1643 KKASSERT(!sopt
->sopt_val
|| kva_p(sopt
->sopt_val
));
1645 if (sopt
->sopt_val
== NULL
)
1647 val
= sopt
->sopt_val
;
1648 maxsize
= sopt
->sopt_valsize
;
1649 while (m
!= NULL
&& maxsize
>= m
->m_len
) {
1650 bcopy(mtod(m
, char *), val
, m
->m_len
);
1651 maxsize
-= m
->m_len
;
1652 val
= (caddr_t
)val
+ m
->m_len
;
1653 valsize
+= m
->m_len
;
1657 /* enough soopt buffer should be given from user-land */
1661 sopt
->sopt_valsize
= valsize
;
1666 sohasoutofband(struct socket
*so
)
1668 if (so
->so_sigio
!= NULL
)
1669 pgsigio(so
->so_sigio
, SIGURG
, 0);
1670 selwakeup(&so
->so_rcv
.ssb_sel
);
1674 sopoll(struct socket
*so
, int events
, struct ucred
*cred
, struct thread
*td
)
1680 if (events
& (POLLIN
| POLLRDNORM
))
1682 revents
|= events
& (POLLIN
| POLLRDNORM
);
1684 if (events
& POLLINIGNEOF
)
1685 if (so
->so_rcv
.ssb_cc
>= so
->so_rcv
.ssb_lowat
||
1686 !TAILQ_EMPTY(&so
->so_comp
) || so
->so_error
)
1687 revents
|= POLLINIGNEOF
;
1689 if (events
& (POLLOUT
| POLLWRNORM
))
1690 if (sowriteable(so
))
1691 revents
|= events
& (POLLOUT
| POLLWRNORM
);
1693 if (events
& (POLLPRI
| POLLRDBAND
))
1694 if (so
->so_oobmark
|| (so
->so_state
& SS_RCVATMARK
))
1695 revents
|= events
& (POLLPRI
| POLLRDBAND
);
1699 (POLLIN
| POLLINIGNEOF
| POLLPRI
| POLLRDNORM
|
1701 selrecord(td
, &so
->so_rcv
.ssb_sel
);
1702 so
->so_rcv
.ssb_flags
|= SSB_SEL
;
1705 if (events
& (POLLOUT
| POLLWRNORM
)) {
1706 selrecord(td
, &so
->so_snd
.ssb_sel
);
1707 so
->so_snd
.ssb_flags
|= SSB_SEL
;
1716 sokqfilter(struct file
*fp
, struct knote
*kn
)
1718 struct socket
*so
= (struct socket
*)kn
->kn_fp
->f_data
;
1719 struct signalsockbuf
*ssb
;
1721 switch (kn
->kn_filter
) {
1723 if (so
->so_options
& SO_ACCEPTCONN
)
1724 kn
->kn_fop
= &solisten_filtops
;
1726 kn
->kn_fop
= &soread_filtops
;
1730 kn
->kn_fop
= &sowrite_filtops
;
1738 SLIST_INSERT_HEAD(&ssb
->ssb_sel
.si_note
, kn
, kn_selnext
);
1739 ssb
->ssb_flags
|= SSB_KNOTE
;
1745 filt_sordetach(struct knote
*kn
)
1747 struct socket
*so
= (struct socket
*)kn
->kn_fp
->f_data
;
1750 SLIST_REMOVE(&so
->so_rcv
.ssb_sel
.si_note
, kn
, knote
, kn_selnext
);
1751 if (SLIST_EMPTY(&so
->so_rcv
.ssb_sel
.si_note
))
1752 so
->so_rcv
.ssb_flags
&= ~SSB_KNOTE
;
1758 filt_soread(struct knote
*kn
, long hint
)
1760 struct socket
*so
= (struct socket
*)kn
->kn_fp
->f_data
;
1762 kn
->kn_data
= so
->so_rcv
.ssb_cc
;
1763 if (so
->so_state
& SS_CANTRCVMORE
) {
1764 kn
->kn_flags
|= EV_EOF
;
1765 kn
->kn_fflags
= so
->so_error
;
1768 if (so
->so_error
) /* temporary udp error */
1770 if (kn
->kn_sfflags
& NOTE_LOWAT
)
1771 return (kn
->kn_data
>= kn
->kn_sdata
);
1772 return (kn
->kn_data
>= so
->so_rcv
.ssb_lowat
);
1776 filt_sowdetach(struct knote
*kn
)
1778 struct socket
*so
= (struct socket
*)kn
->kn_fp
->f_data
;
1781 SLIST_REMOVE(&so
->so_snd
.ssb_sel
.si_note
, kn
, knote
, kn_selnext
);
1782 if (SLIST_EMPTY(&so
->so_snd
.ssb_sel
.si_note
))
1783 so
->so_snd
.ssb_flags
&= ~SSB_KNOTE
;
1789 filt_sowrite(struct knote
*kn
, long hint
)
1791 struct socket
*so
= (struct socket
*)kn
->kn_fp
->f_data
;
1793 kn
->kn_data
= ssb_space(&so
->so_snd
);
1794 if (so
->so_state
& SS_CANTSENDMORE
) {
1795 kn
->kn_flags
|= EV_EOF
;
1796 kn
->kn_fflags
= so
->so_error
;
1799 if (so
->so_error
) /* temporary udp error */
1801 if (((so
->so_state
& SS_ISCONNECTED
) == 0) &&
1802 (so
->so_proto
->pr_flags
& PR_CONNREQUIRED
))
1804 if (kn
->kn_sfflags
& NOTE_LOWAT
)
1805 return (kn
->kn_data
>= kn
->kn_sdata
);
1806 return (kn
->kn_data
>= so
->so_snd
.ssb_lowat
);
1811 filt_solisten(struct knote
*kn
, long hint
)
1813 struct socket
*so
= (struct socket
*)kn
->kn_fp
->f_data
;
1815 kn
->kn_data
= so
->so_qlen
;
1816 return (! TAILQ_EMPTY(&so
->so_comp
));