]>
git.saurik.com Git - apple/xnu.git/blob - bsd/kern/uipc_socket.c
2 * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
4 * @APPLE_LICENSE_HEADER_START@
6 * The contents of this file constitute Original Code as defined in and
7 * are subject to the Apple Public Source License Version 1.1 (the
8 * "License"). You may not use this file except in compliance with the
9 * License. Please obtain a copy of the License at
10 * http://www.apple.com/publicsource and read it before using this file.
12 * This Original Code and all software distributed under the License are
13 * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
14 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
15 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the
17 * License for the specific language governing rights and limitations
20 * @APPLE_LICENSE_HEADER_END@
22 /* Copyright (c) 1998, 1999 Apple Computer, Inc. All Rights Reserved */
23 /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
25 * Copyright (c) 1982, 1986, 1988, 1990, 1993
26 * The Regents of the University of California. All rights reserved.
28 * Redistribution and use in source and binary forms, with or without
29 * modification, are permitted provided that the following conditions
31 * 1. Redistributions of source code must retain the above copyright
32 * notice, this list of conditions and the following disclaimer.
33 * 2. Redistributions in binary form must reproduce the above copyright
34 * notice, this list of conditions and the following disclaimer in the
35 * documentation and/or other materials provided with the distribution.
36 * 3. All advertising materials mentioning features or use of this software
37 * must display the following acknowledgement:
38 * This product includes software developed by the University of
39 * California, Berkeley and its contributors.
40 * 4. Neither the name of the University nor the names of its contributors
41 * may be used to endorse or promote products derived from this software
42 * without specific prior written permission.
44 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
45 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
46 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
47 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
48 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
49 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
50 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
51 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
52 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
53 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
56 * @(#)uipc_socket.c 8.6 (Berkeley) 5/2/95
59 #include <sys/param.h>
60 #include <sys/systm.h>
62 #include <sys/fcntl.h>
63 #include <sys/malloc.h>
65 #include <sys/domain.h>
66 #include <sys/kernel.h>
68 #include <sys/protosw.h>
69 #include <sys/socket.h>
70 #include <sys/socketvar.h>
71 #include <sys/resourcevar.h>
72 #include <sys/signalvar.h>
73 #include <sys/sysctl.h>
76 #include <sys/kdebug.h>
77 #include <net/route.h>
78 #include <netinet/in.h>
79 #include <netinet/in_pcb.h>
80 #include <kern/zalloc.h>
81 #include <machine/limits.h>
84 int so_cache_timeouts
= 0;
85 int so_cache_max_freed
= 0;
86 int cached_sock_count
= 0;
87 struct socket
*socket_cache_head
= 0;
88 struct socket
*socket_cache_tail
= 0;
89 u_long so_cache_time
= 0;
90 int so_cache_init_done
= 0;
91 struct zone
*so_cache_zone
;
92 extern int get_inpcb_str_size();
93 extern int get_tcp_str_size();
95 #include <machine/limits.h>
98 int socket_zone
= M_SOCKET
;
99 so_gen_t so_gencnt
; /* generation count for sockets */
101 MALLOC_DEFINE(M_SONAME
, "soname", "socket name");
102 MALLOC_DEFINE(M_PCB
, "pcb", "protocol control block");
104 #define DBG_LAYER_IN_BEG NETDBG_CODE(DBG_NETSOCK, 0)
105 #define DBG_LAYER_IN_END NETDBG_CODE(DBG_NETSOCK, 2)
106 #define DBG_LAYER_OUT_BEG NETDBG_CODE(DBG_NETSOCK, 1)
107 #define DBG_LAYER_OUT_END NETDBG_CODE(DBG_NETSOCK, 3)
108 #define DBG_FNC_SOSEND NETDBG_CODE(DBG_NETSOCK, (4 << 8) | 1)
109 #define DBG_FNC_SORECEIVE NETDBG_CODE(DBG_NETSOCK, (8 << 8))
110 #define DBG_FNC_SOSHUTDOWN NETDBG_CODE(DBG_NETSOCK, (9 << 8))
113 SYSCTL_DECL(_kern_ipc
);
115 static int somaxconn
= SOMAXCONN
;
116 SYSCTL_INT(_kern_ipc
, KIPC_SOMAXCONN
, somaxconn
, CTLFLAG_RW
, &somaxconn
,
119 /* Should we get a maximum also ??? */
120 static int sosendmaxchain
= 65536;
121 static int sosendminchain
= 16384;
122 SYSCTL_INT(_kern_ipc
, OID_AUTO
, sosendminchain
, CTLFLAG_RW
, &sosendminchain
,
125 void so_cache_timer();
128 * Socket operation routines.
129 * These routines are called by the routines in
130 * sys_socket.c or from a system process, and
131 * implement the semantics of socket operations by
132 * switching out to the protocol specific routines.
139 so_cache_init_done
= 1;
141 timeout(so_cache_timer
, NULL
, (SO_CACHE_FLUSH_INTERVAL
* hz
));
142 str_size
= (vm_size_t
)( sizeof(struct socket
) + 4 +
143 get_inpcb_str_size() + 4 +
145 so_cache_zone
= zinit (str_size
, 120000*str_size
, 8192, "socache zone");
147 kprintf("cached_sock_alloc -- so_cache_zone size is %x\n", str_size
);
152 void cached_sock_alloc(so
, waitok
)
159 register u_long offset
;
163 if (cached_sock_count
) {
165 *so
= socket_cache_head
;
167 panic("cached_sock_alloc: cached sock is null");
169 socket_cache_head
= socket_cache_head
->cache_next
;
170 if (socket_cache_head
)
171 socket_cache_head
->cache_prev
= 0;
173 socket_cache_tail
= 0;
176 temp
= (*so
)->so_saved_pcb
;
177 bzero((caddr_t
)*so
, sizeof(struct socket
));
179 kprintf("cached_sock_alloc - retreiving cached sock %x - count == %d\n", *so
,
182 (*so
)->so_saved_pcb
= temp
;
186 kprintf("Allocating cached sock %x from memory\n", *so
);
191 *so
= (struct socket
*) zalloc(so_cache_zone
);
193 *so
= (struct socket
*) zalloc_noblock(so_cache_zone
);
198 bzero((caddr_t
)*so
, sizeof(struct socket
));
201 * Define offsets for extra structures into our single block of
202 * memory. Align extra structures on longword boundaries.
206 offset
= (u_long
) *so
;
207 offset
+= sizeof(struct socket
);
210 offset
&= 0xfffffffc;
212 (*so
)->so_saved_pcb
= (caddr_t
) offset
;
213 offset
+= get_inpcb_str_size();
216 offset
&= 0xfffffffc;
219 ((struct inpcb
*) (*so
)->so_saved_pcb
)->inp_saved_ppcb
= (caddr_t
) offset
;
221 kprintf("Allocating cached socket - %x, pcb=%x tcpcb=%x\n", *so
,
223 ((struct inpcb
*)(*so
)->so_saved_pcb
)->inp_saved_ppcb
);
227 (*so
)->cached_in_sock_layer
= 1;
231 void cached_sock_free(so
)
238 if (++cached_sock_count
> MAX_CACHED_SOCKETS
) {
242 kprintf("Freeing overflowed cached socket %x\n", so
);
244 zfree(so_cache_zone
, (vm_offset_t
) so
);
248 kprintf("Freeing socket %x into cache\n", so
);
250 if (so_cache_hw
< cached_sock_count
)
251 so_cache_hw
= cached_sock_count
;
253 so
->cache_next
= socket_cache_head
;
255 if (socket_cache_head
)
256 socket_cache_head
->cache_prev
= so
;
258 socket_cache_tail
= so
;
260 so
->cache_timestamp
= so_cache_time
;
261 socket_cache_head
= so
;
266 kprintf("Freed cached sock %x into cache - count is %d\n", so
, cached_sock_count
);
273 void so_cache_timer()
275 register struct socket
*p
;
277 register int n_freed
= 0;
278 boolean_t funnel_state
;
280 funnel_state
= thread_funnel_set(network_flock
, TRUE
);
286 while (p
= socket_cache_tail
)
288 if ((so_cache_time
- p
->cache_timestamp
) < SO_CACHE_TIME_LIMIT
)
293 if (socket_cache_tail
= p
->cache_prev
)
294 p
->cache_prev
->cache_next
= 0;
295 if (--cached_sock_count
== 0)
296 socket_cache_head
= 0;
300 zfree(so_cache_zone
, (vm_offset_t
) p
);
303 if (++n_freed
>= SO_CACHE_MAX_FREE_BATCH
)
305 so_cache_max_freed
++;
311 timeout(so_cache_timer
, NULL
, (SO_CACHE_FLUSH_INTERVAL
* hz
));
313 (void) thread_funnel_set(network_flock
, FALSE
);
319 * Get a socket structure from our zone, and initialize it.
320 * We don't implement `waitok' yet (see comments in uipc_domain.c).
321 * Note that it would probably be better to allocate socket
322 * and PCB at the same time, but I'm not convinced that all
323 * the protocols can be easily modified to do this.
326 soalloc(waitok
, dom
, type
)
333 if ((dom
== PF_INET
) && (type
== SOCK_STREAM
))
334 cached_sock_alloc(&so
, waitok
);
337 so
= _MALLOC_ZONE(sizeof(*so
), socket_zone
, M_WAITOK
);
339 bzero(so
, sizeof *so
);
341 /* XXX race condition for reentrant kernel */
344 so
->so_gencnt
= ++so_gencnt
;
345 so
->so_zone
= socket_zone
;
352 socreate(dom
, aso
, type
, proto
)
359 struct proc
*p
= current_proc();
360 register struct protosw
*prp
;
362 register int error
= 0;
365 prp
= pffindproto(dom
, proto
, type
);
367 prp
= pffindtype(dom
, type
);
368 if (prp
== 0 || prp
->pr_usrreqs
->pru_attach
== 0)
369 return (EPROTONOSUPPORT
);
370 if (prp
->pr_type
!= type
)
372 so
= soalloc(p
!= 0, dom
, type
);
376 TAILQ_INIT(&so
->so_incomp
);
377 TAILQ_INIT(&so
->so_comp
);
381 if (p
->p_ucred
->cr_uid
== 0)
382 so
->so_state
= SS_PRIV
;
384 so
->so_uid
= p
->p_ucred
->cr_uid
;
388 so
->so_rcv
.sb_flags
|= SB_RECV
; /* XXX */
389 if (prp
->pr_sfilter
.tqh_first
)
390 error
= sfilter_init(so
);
392 error
= (*prp
->pr_usrreqs
->pru_attach
)(so
, proto
, p
);
395 so
->so_state
|= SS_NOFDREF
;
399 prp
->pr_domain
->dom_refs
++;
400 so
->so_rcv
.sb_so
= so
->so_snd
.sb_so
= so
;
401 TAILQ_INIT(&so
->so_evlist
);
409 struct sockaddr
*nam
;
412 struct proc
*p
= current_proc();
417 error
= (*so
->so_proto
->pr_usrreqs
->pru_bind
)(so
, nam
, p
);
418 if (error
== 0) /* ??? */
419 { kp
= sotokextcb(so
);
421 { if (kp
->e_soif
&& kp
->e_soif
->sf_sobind
)
422 { error
= (*kp
->e_soif
->sf_sobind
)(so
, nam
, kp
);
424 { if (error
== EJUSTRETURN
)
441 so
->so_gencnt
= ++so_gencnt
;
443 if (so
->cached_in_sock_layer
== 1)
444 cached_sock_free(so
);
446 _FREE_ZONE(so
, sizeof(*so
), so
->so_zone
);
450 solisten(so
, backlog
)
451 register struct socket
*so
;
456 struct proc
*p
= current_proc();
460 error
= (*so
->so_proto
->pr_usrreqs
->pru_listen
)(so
, p
);
465 if (TAILQ_EMPTY(&so
->so_comp
))
466 so
->so_options
|= SO_ACCEPTCONN
;
467 if (backlog
< 0 || backlog
> somaxconn
)
469 so
->so_qlimit
= backlog
;
473 if (kp
->e_soif
&& kp
->e_soif
->sf_solisten
)
474 { error
= (*kp
->e_soif
->sf_solisten
)(so
, kp
);
476 { if (error
== EJUSTRETURN
)
492 register struct socket
*so
;
495 struct socket
*head
= so
->so_head
;
499 { if (kp
->e_soif
&& kp
->e_soif
->sf_sofree
)
500 { error
= (*kp
->e_soif
->sf_sofree
)(so
, kp
);
502 selthreadclear(&so
->so_snd
.sb_sel
);
503 selthreadclear(&so
->so_rcv
.sb_sel
);
504 return; /* void fn */
510 if (so
->so_pcb
|| (so
->so_state
& SS_NOFDREF
) == 0) {
511 selthreadclear(&so
->so_snd
.sb_sel
);
512 selthreadclear(&so
->so_rcv
.sb_sel
);
516 if (so
->so_state
& SS_INCOMP
) {
517 TAILQ_REMOVE(&head
->so_incomp
, so
, so_list
);
519 } else if (so
->so_state
& SS_COMP
) {
521 * We must not decommission a socket that's
522 * on the accept(2) queue. If we do, then
523 * accept(2) may hang after select(2) indicated
524 * that the listening socket was ready.
526 selthreadclear(&so
->so_snd
.sb_sel
);
527 selthreadclear(&so
->so_rcv
.sb_sel
);
530 panic("sofree: not queued");
533 so
->so_state
&= ~(SS_INCOMP
|SS_COMP
);
537 selthreadclear(&so
->so_snd
.sb_sel
);
538 sbrelease(&so
->so_snd
);
545 * Close a socket on last file table reference removal.
546 * Initiate disconnect if connected.
547 * Free socket when disconnect complete.
551 register struct socket
*so
;
553 int s
= splnet(); /* conservative */
558 funsetown(so
->so_pgid
);
562 { if (kp
->e_soif
&& kp
->e_soif
->sf_soclose
)
563 { error
= (*kp
->e_soif
->sf_soclose
)(so
, kp
);
566 return((error
== EJUSTRETURN
) ? 0 : error
);
572 if (so
->so_options
& SO_ACCEPTCONN
) {
573 struct socket
*sp
, *sonext
;
575 sp
= TAILQ_FIRST(&so
->so_incomp
);
576 for (; sp
!= NULL
; sp
= sonext
) {
577 sonext
= TAILQ_NEXT(sp
, so_list
);
580 for (sp
= TAILQ_FIRST(&so
->so_comp
); sp
!= NULL
; sp
= sonext
) {
581 sonext
= TAILQ_NEXT(sp
, so_list
);
582 /* Dequeue from so_comp since sofree() won't do it */
583 TAILQ_REMOVE(&so
->so_comp
, sp
, so_list
);
585 sp
->so_state
&= ~SS_COMP
;
593 if (so
->so_state
& SS_ISCONNECTED
) {
594 if ((so
->so_state
& SS_ISDISCONNECTING
) == 0) {
595 error
= sodisconnect(so
);
599 if (so
->so_options
& SO_LINGER
) {
600 if ((so
->so_state
& SS_ISDISCONNECTING
) &&
601 (so
->so_state
& SS_NBIO
))
603 while (so
->so_state
& SS_ISCONNECTED
) {
604 error
= tsleep((caddr_t
)&so
->so_timeo
,
605 PSOCK
| PCATCH
, "soclos", so
->so_linger
);
613 int error2
= (*so
->so_proto
->pr_usrreqs
->pru_detach
)(so
);
618 if (so
->so_pcb
&& so
->so_state
& SS_NOFDREF
)
619 panic("soclose: NOFDREF");
620 so
->so_state
|= SS_NOFDREF
;
621 so
->so_proto
->pr_domain
->dom_refs
--;
629 * Must be called at splnet...
636 return (*so
->so_proto
->pr_usrreqs
->pru_abort
)(so
);
641 register struct socket
*so
;
642 struct sockaddr
**nam
;
647 if ((so
->so_state
& SS_NOFDREF
) == 0)
648 panic("soaccept: !NOFDREF");
649 so
->so_state
&= ~SS_NOFDREF
;
650 error
= (*so
->so_proto
->pr_usrreqs
->pru_accept
)(so
, nam
);
652 { kp
= sotokextcb(so
);
654 if (kp
->e_soif
&& kp
->e_soif
->sf_soaccept
)
655 { error
= (*kp
->e_soif
->sf_soaccept
)(so
, nam
, kp
);
657 { if (error
== EJUSTRETURN
)
674 register struct socket
*so
;
675 struct sockaddr
*nam
;
680 struct proc
*p
= current_proc();
683 if (so
->so_options
& SO_ACCEPTCONN
)
687 * If protocol is connection-based, can only connect once.
688 * Otherwise, if connected, try to disconnect first.
689 * This allows user to disconnect by connecting to, e.g.,
692 if (so
->so_state
& (SS_ISCONNECTED
|SS_ISCONNECTING
) &&
693 ((so
->so_proto
->pr_flags
& PR_CONNREQUIRED
) ||
694 (error
= sodisconnect(so
))))
697 error
= (*so
->so_proto
->pr_usrreqs
->pru_connect
)(so
, nam
, p
);
703 if (kp
->e_soif
&& kp
->e_soif
->sf_soconnect
)
704 { error
= (*kp
->e_soif
->sf_soconnect
)(so
, nam
, kp
);
706 { if (error
== EJUSTRETURN
)
723 register struct socket
*so1
;
730 error
= (*so1
->so_proto
->pr_usrreqs
->pru_connect2
)(so1
, so2
);
732 { kp
= sotokextcb(so1
);
734 { if (kp
->e_soif
&& kp
->e_soif
->sf_soconnect2
)
735 { error
= (*kp
->e_soif
->sf_soconnect2
)(so1
, so2
, kp
);
737 { if (error
== EJUSTRETURN
)
752 register struct socket
*so
;
758 if ((so
->so_state
& SS_ISCONNECTED
) == 0) {
762 if (so
->so_state
& SS_ISDISCONNECTING
) {
766 error
= (*so
->so_proto
->pr_usrreqs
->pru_disconnect
)(so
);
769 { kp
= sotokextcb(so
);
771 { if (kp
->e_soif
&& kp
->e_soif
->sf_sodisconnect
)
772 { error
= (*kp
->e_soif
->sf_sodisconnect
)(so
, kp
);
774 { if (error
== EJUSTRETURN
)
789 #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_DONTWAIT : M_WAIT)
792 * If send must go all at once and message is larger than
793 * send buffering, then hard error.
794 * Lock against other senders.
795 * If must go all at once and not enough room now, then
796 * inform user that this would block and do nothing.
797 * Otherwise, if nonblocking, send as much as possible.
798 * The data to be sent is described by "uio" if nonzero,
799 * otherwise by the mbuf chain "top" (which must be null
800 * if uio is not). Data provided in mbuf chain must be small
801 * enough to send all at once.
803 * Returns nonzero on error, timeout or signal; callers
804 * must check for short counts if EINTR/ERESTART are returned.
805 * Data and control buffers are freed on return.
807 * MSG_HOLD: go thru most of sosend(), but just enqueue the mbuf
808 * MSG_SEND: go thru as for MSG_HOLD on current fragment, then
809 * point at the mbuf chain being constructed and go from there.
812 sosend(so
, addr
, uio
, top
, control
, flags
)
813 register struct socket
*so
;
814 struct sockaddr
*addr
;
817 struct mbuf
*control
;
822 register struct mbuf
*m
, *freelist
= NULL
;
823 register long space
, len
, resid
;
824 int clen
= 0, error
, s
, dontroute
, mlen
, sendflags
;
825 int atomic
= sosendallatonce(so
) || top
;
826 struct proc
*p
= current_proc();
830 resid
= uio
->uio_resid
;
832 resid
= top
->m_pkthdr
.len
;
834 KERNEL_DEBUG((DBG_FNC_SOSEND
| DBG_FUNC_START
),
839 so
->so_snd
.sb_hiwat
);
842 * In theory resid should be unsigned.
843 * However, space must be signed, as it might be less than 0
844 * if we over-committed, and we must use a signed comparison
845 * of space and resid. On the other hand, a negative resid
846 * causes us to loop sending 0-length segments to the protocol.
848 * Also check to make sure that MSG_EOR isn't used on SOCK_STREAM
849 * type sockets since that's an error.
851 if (resid
< 0 || so
->so_type
== SOCK_STREAM
&& (flags
& MSG_EOR
)) {
857 (flags
& MSG_DONTROUTE
) && (so
->so_options
& SO_DONTROUTE
) == 0 &&
858 (so
->so_proto
->pr_flags
& PR_ATOMIC
);
860 p
->p_stats
->p_ru
.ru_msgsnd
++;
862 clen
= control
->m_len
;
863 #define snderr(errno) { error = errno; splx(s); goto release; }
866 error
= sblock(&so
->so_snd
, SBLOCKWAIT(flags
));
871 if (so
->so_state
& SS_CANTSENDMORE
)
874 error
= so
->so_error
;
879 if ((so
->so_state
& SS_ISCONNECTED
) == 0) {
881 * `sendto' and `sendmsg' is allowed on a connection-
882 * based socket if it supports implied connect.
883 * Return ENOTCONN if not connected and no address is
886 if ((so
->so_proto
->pr_flags
& PR_CONNREQUIRED
) &&
887 (so
->so_proto
->pr_flags
& PR_IMPLOPCL
) == 0) {
888 if ((so
->so_state
& SS_ISCONFIRMING
) == 0 &&
889 !(resid
== 0 && clen
!= 0))
891 } else if (addr
== 0 && !(flags
&MSG_HOLD
))
892 snderr(so
->so_proto
->pr_flags
& PR_CONNREQUIRED
?
893 ENOTCONN
: EDESTADDRREQ
);
895 space
= sbspace(&so
->so_snd
);
898 if ((atomic
&& resid
> so
->so_snd
.sb_hiwat
) ||
899 clen
> so
->so_snd
.sb_hiwat
)
901 if (space
< resid
+ clen
&& uio
&&
902 (atomic
|| space
< so
->so_snd
.sb_lowat
|| space
< clen
)) {
903 if (so
->so_state
& SS_NBIO
)
905 sbunlock(&so
->so_snd
);
906 error
= sbwait(&so
->so_snd
);
919 * Data is prepackaged in "top".
923 top
->m_flags
|= M_EOR
;
925 boolean_t dropped_funnel
= FALSE
;
929 bytes_to_copy
= min(resid
, space
);
931 if (sosendminchain
> 0) {
932 if (bytes_to_copy
>= sosendminchain
) {
933 dropped_funnel
= TRUE
;
934 (void)thread_funnel_set(network_flock
, FALSE
);
938 chainlength
= sosendmaxchain
;
942 if (bytes_to_copy
>= MINCLSIZE
) {
943 if ((m
= freelist
) == NULL
) {
949 num_needed
= bytes_to_copy
/ MCLBYTES
;
951 if ((bytes_to_copy
- (num_needed
* MCLBYTES
)) >= MINCLSIZE
)
954 if ((freelist
= m_getpackets(num_needed
, hdrs_needed
, M_WAIT
)) == NULL
)
955 goto getpackets_failed
;
958 freelist
= m
->m_next
;
962 len
= min(mlen
, bytes_to_copy
);
966 MGETHDR(m
, M_WAIT
, MT_DATA
);
969 m
->m_pkthdr
.rcvif
= (struct ifnet
*)0;
971 MGET(m
, M_WAIT
, MT_DATA
);
974 len
= min(mlen
, bytes_to_copy
);
976 * For datagram protocols, leave room
977 * for protocol headers in first mbuf.
979 if (atomic
&& top
== 0 && len
< mlen
)
986 error
= uiomove(mtod(m
, caddr_t
), (int)len
, uio
);
988 resid
= uio
->uio_resid
;
992 top
->m_pkthdr
.len
+= len
;
998 top
->m_flags
|= M_EOR
;
1001 bytes_to_copy
= min(resid
, space
);
1003 } while (space
> 0 && (chainlength
< sosendmaxchain
|| atomic
|| resid
< MINCLSIZE
));
1005 if (dropped_funnel
== TRUE
)
1006 (void)thread_funnel_set(network_flock
, TRUE
);
1011 if (flags
& (MSG_HOLD
|MSG_SEND
))
1012 { /* Enqueue for later, go away if HOLD */
1013 register struct mbuf
*mb1
;
1014 if (so
->so_temp
&& (flags
& MSG_FLUSH
))
1015 { m_freem(so
->so_temp
);
1019 so
->so_tail
->m_next
= top
;
1033 so
->so_options
|= SO_DONTROUTE
;
1034 s
= splnet(); /* XXX */
1035 kp
= sotokextcb(so
);
1036 /* Compute flags here, for pru_send and NKEs */
1037 sendflags
= (flags
& MSG_OOB
) ? PRUS_OOB
:
1039 * If the user set MSG_EOF, the protocol
1040 * understands this flag and nothing left to
1041 * send then use PRU_SEND_EOF instead of PRU_SEND.
1043 ((flags
& MSG_EOF
) &&
1044 (so
->so_proto
->pr_flags
& PR_IMPLOPCL
) &&
1047 /* If there is more to send set PRUS_MORETOCOME */
1048 (resid
> 0 && space
> 0) ? PRUS_MORETOCOME
: 0;
1050 { if (kp
->e_soif
&& kp
->e_soif
->sf_sosend
)
1051 { error
= (*kp
->e_soif
->sf_sosend
)(so
, &addr
,
1058 if (error
== EJUSTRETURN
)
1059 { sbunlock(&so
->so_snd
);
1062 m_freem_list(freelist
);
1071 error
= (*so
->so_proto
->pr_usrreqs
->pru_send
)(so
,
1072 sendflags
, top
, addr
, control
, p
);
1074 if (flags
& MSG_SEND
)
1078 so
->so_options
&= ~SO_DONTROUTE
;
1085 } while (resid
&& space
> 0);
1089 sbunlock(&so
->so_snd
);
1096 m_freem_list(freelist
);
1098 KERNEL_DEBUG(DBG_FNC_SOSEND
| DBG_FUNC_END
,
1109 * Implement receive operations on a socket.
1110 * We depend on the way that records are added to the sockbuf
1111 * by sbappend*. In particular, each record (mbufs linked through m_next)
1112 * must begin with an address if the protocol so specifies,
1113 * followed by an optional mbuf or mbufs containing ancillary data,
1114 * and then zero or more mbufs of data.
1115 * In order to avoid blocking network interrupts for the entire time here,
1116 * we splx() while doing the actual copy to user space.
1117 * Although the sockbuf is locked, new data may still be appended,
1118 * and thus we must maintain consistency of the sockbuf during that time.
1120 * The caller may receive the data as a single mbuf chain by supplying
1121 * an mbuf **mp0 for use in returning the chain. The uio is then used
1122 * only for the count in uio_resid.
1125 soreceive(so
, psa
, uio
, mp0
, controlp
, flagsp
)
1126 register struct socket
*so
;
1127 struct sockaddr
**psa
;
1130 struct mbuf
**controlp
;
1133 register struct mbuf
*m
, **mp
;
1134 register struct mbuf
*free_list
, *ml
;
1135 register int flags
, len
, error
, s
, offset
;
1136 struct protosw
*pr
= so
->so_proto
;
1137 struct mbuf
*nextrecord
;
1139 int orig_resid
= uio
->uio_resid
;
1142 KERNEL_DEBUG(DBG_FNC_SORECEIVE
| DBG_FUNC_START
,
1146 so
->so_rcv
.sb_lowat
,
1147 so
->so_rcv
.sb_hiwat
);
1149 kp
= sotokextcb(so
);
1151 { if (kp
->e_soif
&& kp
->e_soif
->sf_soreceive
)
1152 { error
= (*kp
->e_soif
->sf_soreceive
)(so
, psa
, &uio
,
1156 return((error
== EJUSTRETURN
) ? 0 : error
);
1167 flags
= *flagsp
&~ MSG_EOR
;
1171 * When SO_WANTOOBFLAG is set we try to get out-of-band data
1172 * regardless of the flags argument. Here is the case were
1173 * out-of-band data is not inline.
1175 if ((flags
& MSG_OOB
) ||
1176 ((so
->so_options
& SO_WANTOOBFLAG
) != 0 &&
1177 (so
->so_options
& SO_OOBINLINE
) == 0 &&
1178 (so
->so_oobmark
|| (so
->so_state
& SS_RCVATMARK
)))) {
1179 m
= m_get(M_WAIT
, MT_DATA
);
1180 error
= (*pr
->pr_usrreqs
->pru_rcvoob
)(so
, m
, flags
& MSG_PEEK
);
1184 error
= uiomove(mtod(m
, caddr_t
),
1185 (int) min(uio
->uio_resid
, m
->m_len
), uio
);
1187 } while (uio
->uio_resid
&& error
== 0 && m
);
1191 if ((so
->so_options
& SO_WANTOOBFLAG
) != 0) {
1192 if (error
== EWOULDBLOCK
|| error
== EINVAL
) {
1194 * Let's try to get normal data:
1195 * EWOULDBLOCK: out-of-band data not receive yet;
1196 * EINVAL: out-of-band data already read.
1200 } else if (error
== 0 && flagsp
)
1203 KERNEL_DEBUG(DBG_FNC_SORECEIVE
| DBG_FUNC_END
, error
,0,0,0,0);
1208 *mp
= (struct mbuf
*)0;
1209 if (so
->so_state
& SS_ISCONFIRMING
&& uio
->uio_resid
)
1210 (*pr
->pr_usrreqs
->pru_rcvd
)(so
, 0);
1213 if (error
= sblock(&so
->so_rcv
, SBLOCKWAIT(flags
)))
1215 KERNEL_DEBUG(DBG_FNC_SORECEIVE
| DBG_FUNC_END
, error
,0,0,0,0);
1220 m
= so
->so_rcv
.sb_mb
;
1222 * If we have less data than requested, block awaiting more
1223 * (subject to any timeout) if:
1224 * 1. the current count is less than the low water mark, or
1225 * 2. MSG_WAITALL is set, and it is possible to do the entire
1226 * receive operation at once if we block (resid <= hiwat).
1227 * 3. MSG_DONTWAIT is not set
1228 * If MSG_WAITALL is set but resid is larger than the receive buffer,
1229 * we have to do the receive in sections, and thus risk returning
1230 * a short count if a timeout or signal occurs after we start.
1232 if (m
== 0 || (((flags
& MSG_DONTWAIT
) == 0 &&
1233 so
->so_rcv
.sb_cc
< uio
->uio_resid
) &&
1234 (so
->so_rcv
.sb_cc
< so
->so_rcv
.sb_lowat
||
1235 ((flags
& MSG_WAITALL
) && uio
->uio_resid
<= so
->so_rcv
.sb_hiwat
)) &&
1236 m
->m_nextpkt
== 0 && (pr
->pr_flags
& PR_ATOMIC
) == 0)) {
1237 KASSERT(m
!= 0 || !so
->so_rcv
.sb_cc
, ("receive 1"));
1241 error
= so
->so_error
;
1242 if ((flags
& MSG_PEEK
) == 0)
1246 if (so
->so_state
& SS_CANTRCVMORE
) {
1252 for (; m
; m
= m
->m_next
)
1253 if (m
->m_type
== MT_OOBDATA
|| (m
->m_flags
& M_EOR
)) {
1254 m
= so
->so_rcv
.sb_mb
;
1257 if ((so
->so_state
& (SS_ISCONNECTED
|SS_ISCONNECTING
)) == 0 &&
1258 (so
->so_proto
->pr_flags
& PR_CONNREQUIRED
)) {
1262 if (uio
->uio_resid
== 0)
1264 if ((so
->so_state
& SS_NBIO
) || (flags
& MSG_DONTWAIT
)) {
1265 error
= EWOULDBLOCK
;
1268 sbunlock(&so
->so_rcv
);
1270 printf("Waiting for socket data\n");
1271 error
= sbwait(&so
->so_rcv
);
1273 printf("SORECEIVE - sbwait returned %d\n", error
);
1277 KERNEL_DEBUG(DBG_FNC_SORECEIVE
| DBG_FUNC_END
, error
,0,0,0,0);
1283 #ifdef notyet /* XXXX */
1285 uio
->uio_procp
->p_stats
->p_ru
.ru_msgrcv
++;
1287 nextrecord
= m
->m_nextpkt
;
1288 if ((pr
->pr_flags
& PR_ADDR
) && m
->m_type
== MT_SONAME
) {
1289 KASSERT(m
->m_type
== MT_SONAME
, ("receive 1a"));
1292 *psa
= dup_sockaddr(mtod(m
, struct sockaddr
*),
1294 if (flags
& MSG_PEEK
) {
1297 sbfree(&so
->so_rcv
, m
);
1298 MFREE(m
, so
->so_rcv
.sb_mb
);
1299 m
= so
->so_rcv
.sb_mb
;
1302 while (m
&& m
->m_type
== MT_CONTROL
&& error
== 0) {
1303 if (flags
& MSG_PEEK
) {
1305 *controlp
= m_copy(m
, 0, m
->m_len
);
1308 sbfree(&so
->so_rcv
, m
);
1310 if (pr
->pr_domain
->dom_externalize
&&
1311 mtod(m
, struct cmsghdr
*)->cmsg_type
==
1313 error
= (*pr
->pr_domain
->dom_externalize
)(m
);
1315 so
->so_rcv
.sb_mb
= m
->m_next
;
1317 m
= so
->so_rcv
.sb_mb
;
1319 MFREE(m
, so
->so_rcv
.sb_mb
);
1320 m
= so
->so_rcv
.sb_mb
;
1325 controlp
= &(*controlp
)->m_next
;
1329 if ((flags
& MSG_PEEK
) == 0)
1330 m
->m_nextpkt
= nextrecord
;
1332 if (type
== MT_OOBDATA
)
1339 ml
= (struct mbuf
*)0;
1341 while (m
&& uio
->uio_resid
> 0 && error
== 0) {
1342 if (m
->m_type
== MT_OOBDATA
) {
1343 if (type
!= MT_OOBDATA
)
1345 } else if (type
== MT_OOBDATA
)
1349 * This assertion needs rework. The trouble is Appletalk is uses many
1350 * mbuf types (NOT listed in mbuf.h!) which will trigger this panic.
1351 * For now just remove the assertion... CSM 9/98
1354 KASSERT(m
->m_type
== MT_DATA
|| m
->m_type
== MT_HEADER
,
1358 * Make sure to allways set MSG_OOB event when getting
1359 * out of band data inline.
1361 if ((so
->so_options
& SO_WANTOOBFLAG
) != 0 &&
1362 (so
->so_options
& SO_OOBINLINE
) != 0 &&
1363 (so
->so_state
& SS_RCVATMARK
) != 0) {
1366 so
->so_state
&= ~SS_RCVATMARK
;
1367 len
= uio
->uio_resid
;
1368 if (so
->so_oobmark
&& len
> so
->so_oobmark
- offset
)
1369 len
= so
->so_oobmark
- offset
;
1370 if (len
> m
->m_len
- moff
)
1371 len
= m
->m_len
- moff
;
1373 * If mp is set, just pass back the mbufs.
1374 * Otherwise copy them out via the uio, then free.
1375 * Sockbuf must be consistent here (points to current mbuf,
1376 * it points to next record) when we drop priority;
1377 * we must note any additions to the sockbuf when we
1378 * block interrupts again.
1382 error
= uiomove(mtod(m
, caddr_t
) + moff
, (int)len
, uio
);
1387 uio
->uio_resid
-= len
;
1388 if (len
== m
->m_len
- moff
) {
1389 if (m
->m_flags
& M_EOR
)
1391 if (flags
& MSG_PEEK
) {
1395 nextrecord
= m
->m_nextpkt
;
1396 sbfree(&so
->so_rcv
, m
);
1400 so
->so_rcv
.sb_mb
= m
= m
->m_next
;
1401 *mp
= (struct mbuf
*)0;
1408 m
->m_nextpkt
= nextrecord
;
1411 if (flags
& MSG_PEEK
)
1415 *mp
= m_copym(m
, 0, len
, M_WAIT
);
1418 so
->so_rcv
.sb_cc
-= len
;
1421 if (so
->so_oobmark
) {
1422 if ((flags
& MSG_PEEK
) == 0) {
1423 so
->so_oobmark
-= len
;
1424 if (so
->so_oobmark
== 0) {
1425 so
->so_state
|= SS_RCVATMARK
;
1426 postevent(so
, 0, EV_OOB
);
1431 if (offset
== so
->so_oobmark
)
1435 if (flags
& MSG_EOR
)
1438 * If the MSG_WAITALL flag is set (for non-atomic socket),
1439 * we must not quit until "uio->uio_resid == 0" or an error
1440 * termination. If a signal/timeout occurs, return
1441 * with a short count but without error.
1442 * Keep sockbuf locked against other readers.
1444 while (flags
& MSG_WAITALL
&& m
== 0 && uio
->uio_resid
> 0 &&
1445 !sosendallatonce(so
) && !nextrecord
) {
1446 if (so
->so_error
|| so
->so_state
& SS_CANTRCVMORE
)
1450 so
->so_rcv
.sb_mb
= ml
->m_next
;
1451 ml
->m_next
= (struct mbuf
*)0;
1452 m_freem_list(free_list
);
1454 error
= sbwait(&so
->so_rcv
);
1456 sbunlock(&so
->so_rcv
);
1458 KERNEL_DEBUG(DBG_FNC_SORECEIVE
| DBG_FUNC_END
, 0,0,0,0,0);
1461 m
= so
->so_rcv
.sb_mb
;
1463 nextrecord
= m
->m_nextpkt
;
1466 ml
= (struct mbuf
*)0;
1470 so
->so_rcv
.sb_mb
= ml
->m_next
;
1471 ml
->m_next
= (struct mbuf
*)0;
1472 m_freem_list(free_list
);
1475 if (m
&& pr
->pr_flags
& PR_ATOMIC
) {
1476 if (so
->so_options
& SO_DONTTRUNC
)
1477 flags
|= MSG_RCVMORE
;
1479 { flags
|= MSG_TRUNC
;
1480 if ((flags
& MSG_PEEK
) == 0)
1481 (void) sbdroprecord(&so
->so_rcv
);
1484 if ((flags
& MSG_PEEK
) == 0) {
1486 so
->so_rcv
.sb_mb
= nextrecord
;
1487 if (pr
->pr_flags
& PR_WANTRCVD
&& so
->so_pcb
)
1488 (*pr
->pr_usrreqs
->pru_rcvd
)(so
, flags
);
1490 if ((so
->so_options
& SO_WANTMORE
) && so
->so_rcv
.sb_cc
> 0)
1491 flags
|= MSG_HAVEMORE
;
1492 if (orig_resid
== uio
->uio_resid
&& orig_resid
&&
1493 (flags
& MSG_EOR
) == 0 && (so
->so_state
& SS_CANTRCVMORE
) == 0) {
1494 sbunlock(&so
->so_rcv
);
1502 sbunlock(&so
->so_rcv
);
1505 KERNEL_DEBUG(DBG_FNC_SORECEIVE
| DBG_FUNC_END
,
1517 register struct socket
*so
;
1520 register struct protosw
*pr
= so
->so_proto
;
1525 KERNEL_DEBUG(DBG_FNC_SOSHUTDOWN
| DBG_FUNC_START
, 0,0,0,0,0);
1526 kp
= sotokextcb(so
);
1528 { if (kp
->e_soif
&& kp
->e_soif
->sf_soshutdown
)
1529 { ret
= (*kp
->e_soif
->sf_soshutdown
)(so
, how
, kp
);
1531 return((ret
== EJUSTRETURN
) ? 0 : ret
);
1539 postevent(so
, 0, EV_RCLOSED
);
1542 ret
= ((*pr
->pr_usrreqs
->pru_shutdown
)(so
));
1543 postevent(so
, 0, EV_WCLOSED
);
1544 KERNEL_DEBUG(DBG_FNC_SOSHUTDOWN
| DBG_FUNC_END
, 0,0,0,0,0);
1548 KERNEL_DEBUG(DBG_FNC_SOSHUTDOWN
| DBG_FUNC_END
, 0,0,0,0,0);
1554 register struct socket
*so
;
1556 register struct sockbuf
*sb
= &so
->so_rcv
;
1557 register struct protosw
*pr
= so
->so_proto
;
1558 register int s
, error
;
1562 kp
= sotokextcb(so
);
1564 { if (kp
->e_soif
&& kp
->e_soif
->sf_sorflush
)
1565 { if ((*kp
->e_soif
->sf_sorflush
)(so
, kp
))
1571 sb
->sb_flags
|= SB_NOINTR
;
1572 (void) sblock(sb
, M_WAIT
);
1576 selthreadclear(&sb
->sb_sel
);
1578 bzero((caddr_t
)sb
, sizeof (*sb
));
1580 if (pr
->pr_flags
& PR_RIGHTS
&& pr
->pr_domain
->dom_dispose
)
1581 (*pr
->pr_domain
->dom_dispose
)(asb
.sb_mb
);
1586 * Perhaps this routine, and sooptcopyout(), below, ought to come in
1587 * an additional variant to handle the case where the option value needs
1588 * to be some kind of integer, but not a specific size.
1589 * In addition to their use here, these functions are also called by the
1590 * protocol-level pr_ctloutput() routines.
1593 sooptcopyin(sopt
, buf
, len
, minlen
)
1594 struct sockopt
*sopt
;
1602 * If the user gives us more than we wanted, we ignore it,
1603 * but if we don't get the minimum length the caller
1604 * wants, we return EINVAL. On success, sopt->sopt_valsize
1605 * is set to however much we actually retrieved.
1607 if ((valsize
= sopt
->sopt_valsize
) < minlen
)
1610 sopt
->sopt_valsize
= valsize
= len
;
1612 if (sopt
->sopt_p
!= 0)
1613 return (copyin(sopt
->sopt_val
, buf
, valsize
));
1615 bcopy(sopt
->sopt_val
, buf
, valsize
);
1622 struct sockopt
*sopt
;
1630 kp
= sotokextcb(so
);
1632 { if (kp
->e_soif
&& kp
->e_soif
->sf_socontrol
)
1633 { error
= (*kp
->e_soif
->sf_socontrol
)(so
, sopt
, kp
);
1635 return((error
== EJUSTRETURN
) ? 0 : error
);
1641 if (sopt
->sopt_level
!= SOL_SOCKET
) {
1642 if (so
->so_proto
&& so
->so_proto
->pr_ctloutput
)
1643 return ((*so
->so_proto
->pr_ctloutput
)
1645 error
= ENOPROTOOPT
;
1647 switch (sopt
->sopt_name
) {
1649 error
= sooptcopyin(sopt
, &l
, sizeof l
, sizeof l
);
1653 so
->so_linger
= l
.l_linger
;
1655 so
->so_options
|= SO_LINGER
;
1657 so
->so_options
&= ~SO_LINGER
;
1663 case SO_USELOOPBACK
:
1671 case SO_WANTOOBFLAG
:
1672 error
= sooptcopyin(sopt
, &optval
, sizeof optval
,
1677 so
->so_options
|= sopt
->sopt_name
;
1679 so
->so_options
&= ~sopt
->sopt_name
;
1686 error
= sooptcopyin(sopt
, &optval
, sizeof optval
,
1692 * Values < 1 make no sense for any of these
1693 * options, so disallow them.
1700 switch (sopt
->sopt_name
) {
1703 if (sbreserve(sopt
->sopt_name
== SO_SNDBUF
?
1704 &so
->so_snd
: &so
->so_rcv
,
1705 (u_long
) optval
) == 0) {
1712 * Make sure the low-water is never greater than
1716 so
->so_snd
.sb_lowat
=
1717 (optval
> so
->so_snd
.sb_hiwat
) ?
1718 so
->so_snd
.sb_hiwat
: optval
;
1721 so
->so_rcv
.sb_lowat
=
1722 (optval
> so
->so_rcv
.sb_hiwat
) ?
1723 so
->so_rcv
.sb_hiwat
: optval
;
1730 error
= sooptcopyin(sopt
, &tv
, sizeof tv
,
1735 if (tv
.tv_sec
> SHRT_MAX
/ hz
- hz
) {
1739 val
= tv
.tv_sec
* hz
+ tv
.tv_usec
/ tick
;
1741 switch (sopt
->sopt_name
) {
1743 so
->so_snd
.sb_timeo
= val
;
1746 so
->so_rcv
.sb_timeo
= val
;
1752 { struct so_nke nke
;
1753 struct NFDescriptor
*nf1
, *nf2
= NULL
;
1755 error
= sooptcopyin(sopt
, &nke
,
1756 sizeof nke
, sizeof nke
);
1760 error
= nke_insert(so
, &nke
);
1765 error
= ENOPROTOOPT
;
1768 if (error
== 0 && so
->so_proto
&& so
->so_proto
->pr_ctloutput
) {
1769 (void) ((*so
->so_proto
->pr_ctloutput
)
1777 /* Helper routine for getsockopt */
1779 sooptcopyout(sopt
, buf
, len
)
1780 struct sockopt
*sopt
;
1790 * Documented get behavior is that we always return a value,
1791 * possibly truncated to fit in the user's buffer.
1792 * Traditional behavior is that we always tell the user
1793 * precisely how much we copied, rather than something useful
1794 * like the total amount we had available for her.
1795 * Note that this interface is not idempotent; the entire answer must
1796 * generated ahead of time.
1798 valsize
= min(len
, sopt
->sopt_valsize
);
1799 sopt
->sopt_valsize
= valsize
;
1800 if (sopt
->sopt_val
!= 0) {
1801 if (sopt
->sopt_p
!= 0)
1802 error
= copyout(buf
, sopt
->sopt_val
, valsize
);
1804 bcopy(buf
, sopt
->sopt_val
, valsize
);
1812 struct sockopt
*sopt
;
1820 kp
= sotokextcb(so
);
1822 { if (kp
->e_soif
&& kp
->e_soif
->sf_socontrol
)
1823 { error
= (*kp
->e_soif
->sf_socontrol
)(so
, sopt
, kp
);
1825 return((error
== EJUSTRETURN
) ? 0 : error
);
1831 if (sopt
->sopt_level
!= SOL_SOCKET
) {
1832 if (so
->so_proto
&& so
->so_proto
->pr_ctloutput
) {
1833 return ((*so
->so_proto
->pr_ctloutput
)
1836 return (ENOPROTOOPT
);
1838 switch (sopt
->sopt_name
) {
1840 l
.l_onoff
= so
->so_options
& SO_LINGER
;
1841 l
.l_linger
= so
->so_linger
;
1842 error
= sooptcopyout(sopt
, &l
, sizeof l
);
1845 case SO_USELOOPBACK
:
1856 case SO_WANTOOBFLAG
:
1857 optval
= so
->so_options
& sopt
->sopt_name
;
1859 error
= sooptcopyout(sopt
, &optval
, sizeof optval
);
1863 optval
= so
->so_type
;
1871 m1
= so
->so_rcv
.sb_mb
;
1872 if (so
->so_proto
->pr_flags
& PR_ATOMIC
)
1875 kprintf("SKT CC: %d\n", so
->so_rcv
.sb_cc
);
1878 { if (m1
->m_type
== MT_DATA
)
1879 pkt_total
+= m1
->m_len
;
1881 kprintf("CNT: %d/%d\n", m1
->m_len
, pkt_total
);
1887 optval
= so
->so_rcv
.sb_cc
;
1889 kprintf("RTN: %d\n", optval
);
1894 optval
= so
->so_error
;
1899 optval
= so
->so_snd
.sb_hiwat
;
1903 optval
= so
->so_rcv
.sb_hiwat
;
1907 optval
= so
->so_snd
.sb_lowat
;
1911 optval
= so
->so_rcv
.sb_lowat
;
1916 optval
= (sopt
->sopt_name
== SO_SNDTIMEO
?
1917 so
->so_snd
.sb_timeo
: so
->so_rcv
.sb_timeo
);
1919 tv
.tv_sec
= optval
/ hz
;
1920 tv
.tv_usec
= (optval
% hz
) * tick
;
1921 error
= sooptcopyout(sopt
, &tv
, sizeof tv
);
1925 error
= ENOPROTOOPT
;
1934 register struct socket
*so
;
1940 kp
= sotokextcb(so
);
1942 { if (kp
->e_soif
&& kp
->e_soif
->sf_sohasoutofband
)
1943 { if ((*kp
->e_soif
->sf_sohasoutofband
)(so
, kp
))
1948 if (so
->so_pgid
< 0)
1949 gsignal(-so
->so_pgid
, SIGURG
);
1950 else if (so
->so_pgid
> 0 && (p
= pfind(so
->so_pgid
)) != 0)
1952 selwakeup(&so
->so_rcv
.sb_sel
);
1956 * Network filter support
1958 /* Run the list of filters, creating extension control blocks */
1959 sfilter_init(register struct socket
*so
)
1960 { struct kextcb
*kp
, **kpp
;
1961 struct protosw
*prp
;
1962 struct NFDescriptor
*nfp
;
1965 nfp
= prp
->pr_sfilter
.tqh_first
; /* non-null */
1969 { MALLOC(kp
, struct kextcb
*, sizeof(*kp
),
1972 return(ENOBUFS
); /* so_free will clean up */
1978 kp
->e_soif
= nfp
->nf_soif
;
1979 kp
->e_sout
= nfp
->nf_soutil
;
1981 * Ignore return value for create
1982 * Everyone gets a chance at startup
1984 if (kp
->e_soif
&& kp
->e_soif
->sf_socreate
)
1985 (*kp
->e_soif
->sf_socreate
)(so
, prp
, kp
);
1986 nfp
= nfp
->nf_next
.tqe_next
;
1993 * Run the list of filters, freeing extension control blocks
1994 * Assumes the soif/soutil blocks have been handled.
1996 sfilter_term(struct socket
*so
)
1997 { struct kextcb
*kp
, *kp1
;
2003 * Ignore return code on termination; everyone must
2006 if (kp
->e_soif
&& kp
->e_soif
->sf_sofree
)
2007 kp
->e_soif
->sf_sofree(so
, kp
);
2016 sopoll(struct socket
*so
, int events
, struct ucred
*cred
, void * wql
)
2018 struct proc
*p
= current_proc();
2022 if (events
& (POLLIN
| POLLRDNORM
))
2024 revents
|= events
& (POLLIN
| POLLRDNORM
);
2026 if (events
& (POLLOUT
| POLLWRNORM
))
2027 if (sowriteable(so
))
2028 revents
|= events
& (POLLOUT
| POLLWRNORM
);
2030 if (events
& (POLLPRI
| POLLRDBAND
))
2031 if (so
->so_oobmark
|| (so
->so_state
& SS_RCVATMARK
))
2032 revents
|= events
& (POLLPRI
| POLLRDBAND
);
2035 if (events
& (POLLIN
| POLLPRI
| POLLRDNORM
| POLLRDBAND
)) {
2036 so
->so_rcv
.sb_flags
|= SB_SEL
;
2037 selrecord(p
, &so
->so_rcv
.sb_sel
, wql
);
2040 if (events
& (POLLOUT
| POLLWRNORM
)) {
2041 so
->so_snd
.sb_flags
|= SB_SEL
;
2042 selrecord(p
, &so
->so_snd
.sb_sel
, wql
);
2050 /*#### IPv6 Integration. Added new routines */
2052 sooptgetm(struct sockopt
*sopt
, struct mbuf
**mp
)
2054 struct mbuf
*m
, *m_prev
;
2055 int sopt_size
= sopt
->sopt_valsize
;
2057 MGET(m
, sopt
->sopt_p
? M_WAIT
: M_DONTWAIT
, MT_DATA
);
2060 if (sopt_size
> MLEN
) {
2061 MCLGET(m
, sopt
->sopt_p
? M_WAIT
: M_DONTWAIT
);
2062 if ((m
->m_flags
& M_EXT
) == 0) {
2066 m
->m_len
= min(MCLBYTES
, sopt_size
);
2068 m
->m_len
= min(MLEN
, sopt_size
);
2070 sopt_size
-= m
->m_len
;
2075 MGET(m
, sopt
->sopt_p
? M_WAIT
: M_DONTWAIT
, MT_DATA
);
2080 if (sopt_size
> MLEN
) {
2081 MCLGET(m
, sopt
->sopt_p
? M_WAIT
: M_DONTWAIT
);
2082 if ((m
->m_flags
& M_EXT
) == 0) {
2086 m
->m_len
= min(MCLBYTES
, sopt_size
);
2088 m
->m_len
= min(MLEN
, sopt_size
);
2090 sopt_size
-= m
->m_len
;
2097 /* XXX; copyin sopt data into mbuf chain for (__FreeBSD__ < 3) routines. */
2099 sooptmcopyin(struct sockopt
*sopt
, struct mbuf
*m
)
2101 struct mbuf
*m0
= m
;
2103 if (sopt
->sopt_val
== NULL
)
2105 while (m
!= NULL
&& sopt
->sopt_valsize
>= m
->m_len
) {
2106 if (sopt
->sopt_p
!= NULL
) {
2109 error
= copyin(sopt
->sopt_val
, mtod(m
, char *),
2116 bcopy(sopt
->sopt_val
, mtod(m
, char *), m
->m_len
);
2117 sopt
->sopt_valsize
-= m
->m_len
;
2118 (caddr_t
)sopt
->sopt_val
+= m
->m_len
;
2121 if (m
!= NULL
) /* should be allocated enoughly at ip6_sooptmcopyin() */
2122 panic("sooptmcopyin");
2126 /* XXX; copyout mbuf chain data into soopt for (__FreeBSD__ < 3) routines. */
2128 sooptmcopyout(struct sockopt
*sopt
, struct mbuf
*m
)
2130 struct mbuf
*m0
= m
;
2133 if (sopt
->sopt_val
== NULL
)
2135 while (m
!= NULL
&& sopt
->sopt_valsize
>= m
->m_len
) {
2136 if (sopt
->sopt_p
!= NULL
) {
2139 error
= copyout(mtod(m
, char *), sopt
->sopt_val
,
2146 bcopy(mtod(m
, char *), sopt
->sopt_val
, m
->m_len
);
2147 sopt
->sopt_valsize
-= m
->m_len
;
2148 (caddr_t
)sopt
->sopt_val
+= m
->m_len
;
2149 valsize
+= m
->m_len
;
2153 /* enough soopt buffer should be given from user-land */
2157 sopt
->sopt_valsize
= valsize
;