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, 1995
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 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95
67 * $FreeBSD: src/sys/netinet/udp_usrreq.c,v 1.64.2.18 2003/01/24 05:11:34 sam Exp $
68 * $DragonFly: src/sys/netinet/udp_usrreq.c,v 1.47 2008/11/11 10:46:58 sephe Exp $
71 #include "opt_ipsec.h"
72 #include "opt_inet6.h"
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/kernel.h>
77 #include <sys/malloc.h>
79 #include <sys/domain.h>
82 #include <sys/protosw.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
85 #include <sys/sysctl.h>
86 #include <sys/syslog.h>
87 #include <sys/in_cksum.h>
89 #include <sys/thread2.h>
90 #include <sys/socketvar2.h>
91 #include <sys/serialize.h>
93 #include <machine/stdarg.h>
96 #include <net/route.h>
97 #include <net/netmsg2.h>
99 #include <netinet/in.h>
100 #include <netinet/in_systm.h>
101 #include <netinet/ip.h>
103 #include <netinet/ip6.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/in_var.h>
107 #include <netinet/ip_var.h>
109 #include <netinet6/ip6_var.h>
111 #include <netinet/ip_icmp.h>
112 #include <netinet/icmp_var.h>
113 #include <netinet/udp.h>
114 #include <netinet/udp_var.h>
117 #include <netproto/ipsec/ipsec.h>
121 #include <netinet6/ipsec.h>
125 * UDP protocol implementation.
126 * Per RFC 768, August, 1980.
129 static int udpcksum
= 1;
131 static int udpcksum
= 0; /* XXX */
133 SYSCTL_INT(_net_inet_udp
, UDPCTL_CHECKSUM
, checksum
, CTLFLAG_RW
,
134 &udpcksum
, 0, "Enable checksumming of UDP packets");
137 SYSCTL_INT(_net_inet_udp
, OID_AUTO
, log_in_vain
, CTLFLAG_RW
,
138 &log_in_vain
, 0, "Log all incoming UDP packets");
140 static int blackhole
= 0;
141 SYSCTL_INT(_net_inet_udp
, OID_AUTO
, blackhole
, CTLFLAG_RW
,
142 &blackhole
, 0, "Do not send port unreachables for refused connects");
144 static int strict_mcast_mship
= 1;
145 SYSCTL_INT(_net_inet_udp
, OID_AUTO
, strict_mcast_mship
, CTLFLAG_RW
,
146 &strict_mcast_mship
, 0, "Only send multicast to member sockets");
148 int udp_sosnd_async
= 1;
149 SYSCTL_INT(_net_inet_udp
, OID_AUTO
, sosnd_async
, CTLFLAG_RW
,
150 &udp_sosnd_async
, 0, "UDP asynchronized pru_send");
152 struct inpcbinfo udbinfo
;
154 static struct netisr_barrier
*udbinfo_br
;
155 static struct lwkt_serialize udbinfo_slize
= LWKT_SERIALIZE_INITIALIZER
;
158 #define UDBHASHSIZE 16
161 struct udpstat udpstat
; /* from udp_var.h */
162 SYSCTL_STRUCT(_net_inet_udp
, UDPCTL_STATS
, stats
, CTLFLAG_RW
,
163 &udpstat
, udpstat
, "UDP statistics (struct udpstat, netinet/udp_var.h)");
165 static struct sockaddr_in udp_in
= { sizeof udp_in
, AF_INET
};
168 struct sockaddr_in6 uin6_sin
;
169 u_char uin6_init_done
: 1;
171 { sizeof udp_in6
.uin6_sin
, AF_INET6
},
175 struct ip6_hdr uip6_ip6
;
176 u_char uip6_init_done
: 1;
180 static void udp_append (struct inpcb
*last
, struct ip
*ip
,
181 struct mbuf
*n
, int off
);
183 static void ip_2_ip6_hdr (struct ip6_hdr
*ip6
, struct ip
*ip
);
186 static int udp_connect_oncpu(struct socket
*so
, struct thread
*td
,
187 struct sockaddr_in
*sin
, struct sockaddr_in
*if_sin
);
188 static int udp_output (struct inpcb
*, struct mbuf
*, struct sockaddr
*,
189 struct thread
*, int);
194 in_pcbinfo_init(&udbinfo
);
195 udbinfo
.hashbase
= hashinit(UDBHASHSIZE
, M_PCB
, &udbinfo
.hashmask
);
196 udbinfo
.porthashbase
= hashinit(UDBHASHSIZE
, M_PCB
,
197 &udbinfo
.porthashmask
);
198 udbinfo
.wildcardhashbase
= hashinit(UDBHASHSIZE
, M_PCB
,
199 &udbinfo
.wildcardhashmask
);
200 udbinfo
.ipi_size
= sizeof(struct inpcb
);
202 udbinfo_br
= netisr_barrier_create();
206 * Check multicast packets to make sure they are only sent to sockets with
207 * multicast memberships for the packet's destination address and arrival
208 * interface. Multicast packets to multicast-unaware sockets are also
211 * Returns 0 if the packet is acceptable, -1 if it is not.
214 check_multicast_membership(struct ip
*ip
, struct inpcb
*inp
, struct mbuf
*m
)
217 struct ip_moptions
*mopt
;
219 if (strict_mcast_mship
== 0 ||
220 !IN_MULTICAST(ntohl(ip
->ip_dst
.s_addr
))) {
223 mopt
= inp
->inp_moptions
;
226 for (mshipno
= 0; mshipno
< mopt
->imo_num_memberships
; ++mshipno
) {
227 struct in_multi
*maddr
= mopt
->imo_membership
[mshipno
];
229 if (ip
->ip_dst
.s_addr
== maddr
->inm_addr
.s_addr
&&
230 m
->m_pkthdr
.rcvif
== maddr
->inm_ifp
) {
238 udp_input(struct mbuf
**mp
, int *offp
, int proto
)
245 struct mbuf
*opts
= NULL
;
248 struct sockaddr
*append_sa
;
255 udpstat
.udps_ipackets
++;
258 * Strip IP options, if any; should skip this,
259 * make available to user, and use on returned packets,
260 * but we don't yet have a way to check the checksum
261 * with options still present.
263 if (iphlen
> sizeof(struct ip
)) {
265 iphlen
= sizeof(struct ip
);
269 * IP and UDP headers are together in first mbuf.
270 * Already checked and pulled up in ip_demux().
272 KASSERT(m
->m_len
>= iphlen
+ sizeof(struct udphdr
),
273 ("UDP header not in one mbuf"));
275 ip
= mtod(m
, struct ip
*);
276 uh
= (struct udphdr
*)((caddr_t
)ip
+ iphlen
);
278 /* destination port of 0 is illegal, based on RFC768. */
279 if (uh
->uh_dport
== 0)
283 * Make mbuf data length reflect UDP length.
284 * If not enough data to reflect UDP length, drop.
286 len
= ntohs((u_short
)uh
->uh_ulen
);
287 if (ip
->ip_len
!= len
) {
288 if (len
> ip
->ip_len
|| len
< sizeof(struct udphdr
)) {
289 udpstat
.udps_badlen
++;
292 m_adj(m
, len
- ip
->ip_len
);
293 /* ip->ip_len = len; */
296 * Save a copy of the IP header in case we want restore it
297 * for sending an ICMP error message in response.
302 * Checksum extended UDP header and data.
305 if (m
->m_pkthdr
.csum_flags
& CSUM_DATA_VALID
) {
306 if (m
->m_pkthdr
.csum_flags
& CSUM_PSEUDO_HDR
)
307 uh
->uh_sum
= m
->m_pkthdr
.csum_data
;
309 uh
->uh_sum
= in_pseudo(ip
->ip_src
.s_addr
,
310 ip
->ip_dst
.s_addr
, htonl((u_short
)len
+
311 m
->m_pkthdr
.csum_data
+ IPPROTO_UDP
));
312 uh
->uh_sum
^= 0xffff;
316 bcopy(((struct ipovly
*)ip
)->ih_x1
, b
, 9);
317 bzero(((struct ipovly
*)ip
)->ih_x1
, 9);
318 ((struct ipovly
*)ip
)->ih_len
= uh
->uh_ulen
;
319 uh
->uh_sum
= in_cksum(m
, len
+ sizeof(struct ip
));
320 bcopy(b
, ((struct ipovly
*)ip
)->ih_x1
, 9);
323 udpstat
.udps_badsum
++;
325 return(IPPROTO_DONE
);
328 udpstat
.udps_nosum
++;
330 if (IN_MULTICAST(ntohl(ip
->ip_dst
.s_addr
)) ||
331 in_broadcast(ip
->ip_dst
, m
->m_pkthdr
.rcvif
)) {
335 * Deliver a multicast or broadcast datagram to *all* sockets
336 * for which the local and remote addresses and ports match
337 * those of the incoming datagram. This allows more than
338 * one process to receive multi/broadcasts on the same port.
339 * (This really ought to be done for unicast datagrams as
340 * well, but that would cause problems with existing
341 * applications that open both address-specific sockets and
342 * a wildcard socket listening to the same port -- they would
343 * end up receiving duplicates of every unicast datagram.
344 * Those applications open the multiple sockets to overcome an
345 * inadequacy of the UDP socket interface, but for backwards
346 * compatibility we avoid the problem here rather than
347 * fixing the interface. Maybe 4.5BSD will remedy this?)
351 * Construct sockaddr format source address.
353 udp_in
.sin_port
= uh
->uh_sport
;
354 udp_in
.sin_addr
= ip
->ip_src
;
356 * Locate pcb(s) for datagram.
357 * (Algorithm copied from raw_intr().)
361 udp_in6
.uin6_init_done
= udp_ip6
.uip6_init_done
= 0;
363 LIST_FOREACH(inp
, &udbinfo
.pcblisthead
, inp_list
) {
364 KKASSERT((inp
->inp_flags
& INP_PLACEMARKER
) == 0);
366 if (!(inp
->inp_vflag
& INP_IPV4
))
369 if (inp
->inp_lport
!= uh
->uh_dport
)
371 if (inp
->inp_laddr
.s_addr
!= INADDR_ANY
) {
372 if (inp
->inp_laddr
.s_addr
!=
376 if (inp
->inp_faddr
.s_addr
!= INADDR_ANY
) {
377 if (inp
->inp_faddr
.s_addr
!=
379 inp
->inp_fport
!= uh
->uh_sport
)
383 if (check_multicast_membership(ip
, inp
, m
) < 0)
390 /* check AH/ESP integrity. */
391 if (ipsec4_in_reject_so(m
, last
->inp_socket
))
392 ipsecstat
.in_polvio
++;
393 /* do not inject data to pcb */
397 /* check AH/ESP integrity. */
398 if (ipsec4_in_reject(m
, last
))
401 #endif /*FAST_IPSEC*/
402 if ((n
= m_copypacket(m
, MB_DONTWAIT
)) != NULL
)
403 udp_append(last
, ip
, n
,
405 sizeof(struct udphdr
));
409 * Don't look for additional matches if this one does
410 * not have either the SO_REUSEPORT or SO_REUSEADDR
411 * socket options set. This heuristic avoids searching
412 * through all pcbs in the common case of a non-shared
413 * port. It * assumes that an application will never
414 * clear these options after setting them.
416 if (!(last
->inp_socket
->so_options
&
417 (SO_REUSEPORT
| SO_REUSEADDR
)))
423 * No matching pcb found; discard datagram.
424 * (No need to send an ICMP Port Unreachable
425 * for a broadcast or multicast datgram.)
427 udpstat
.udps_noportbcast
++;
431 /* check AH/ESP integrity. */
432 if (ipsec4_in_reject_so(m
, last
->inp_socket
)) {
433 ipsecstat
.in_polvio
++;
438 /* check AH/ESP integrity. */
439 if (ipsec4_in_reject(m
, last
))
441 #endif /*FAST_IPSEC*/
442 udp_append(last
, ip
, m
, iphlen
+ sizeof(struct udphdr
));
443 return(IPPROTO_DONE
);
446 * Locate pcb for datagram.
448 inp
= in_pcblookup_hash(&udbinfo
, ip
->ip_src
, uh
->uh_sport
,
449 ip
->ip_dst
, uh
->uh_dport
, 1, m
->m_pkthdr
.rcvif
);
452 char buf
[sizeof "aaa.bbb.ccc.ddd"];
454 strcpy(buf
, inet_ntoa(ip
->ip_dst
));
456 "Connection attempt to UDP %s:%d from %s:%d\n",
457 buf
, ntohs(uh
->uh_dport
), inet_ntoa(ip
->ip_src
),
458 ntohs(uh
->uh_sport
));
460 udpstat
.udps_noport
++;
461 if (m
->m_flags
& (M_BCAST
| M_MCAST
)) {
462 udpstat
.udps_noportbcast
++;
468 if (badport_bandlim(BANDLIM_ICMP_UNREACH
) < 0)
472 ip
->ip_len
+= iphlen
;
473 icmp_error(m
, ICMP_UNREACH
, ICMP_UNREACH_PORT
, 0, 0);
474 return(IPPROTO_DONE
);
477 if (ipsec4_in_reject_so(m
, inp
->inp_socket
)) {
478 ipsecstat
.in_polvio
++;
483 if (ipsec4_in_reject(m
, inp
))
485 #endif /*FAST_IPSEC*/
487 * Check the minimum TTL for socket.
489 if (ip
->ip_ttl
< inp
->inp_ip_minttl
)
493 * Construct sockaddr format source address.
494 * Stuff source address and datagram in user buffer.
496 udp_in
.sin_port
= uh
->uh_sport
;
497 udp_in
.sin_addr
= ip
->ip_src
;
498 if ((inp
->inp_flags
& INP_CONTROLOPTS
) ||
499 (inp
->inp_socket
->so_options
& SO_TIMESTAMP
)) {
501 if (inp
->inp_vflag
& INP_IPV6
) {
504 ip_2_ip6_hdr(&udp_ip6
.uip6_ip6
, ip
);
505 savedflags
= inp
->inp_flags
;
506 inp
->inp_flags
&= ~INP_UNMAPPABLEOPTS
;
507 ip6_savecontrol(inp
, &opts
, &udp_ip6
.uip6_ip6
, m
);
508 inp
->inp_flags
= savedflags
;
511 ip_savecontrol(inp
, &opts
, ip
, m
);
513 m_adj(m
, iphlen
+ sizeof(struct udphdr
));
515 if (inp
->inp_vflag
& INP_IPV6
) {
516 in6_sin_2_v4mapsin6(&udp_in
, &udp_in6
.uin6_sin
);
517 append_sa
= (struct sockaddr
*)&udp_in6
;
520 append_sa
= (struct sockaddr
*)&udp_in
;
522 lwkt_gettoken(&inp
->inp_socket
->so_rcv
.ssb_token
);
523 if (ssb_appendaddr(&inp
->inp_socket
->so_rcv
, append_sa
, m
, opts
) == 0) {
524 udpstat
.udps_fullsock
++;
525 lwkt_reltoken(&inp
->inp_socket
->so_rcv
.ssb_token
);
528 lwkt_reltoken(&inp
->inp_socket
->so_rcv
.ssb_token
);
529 sorwakeup(inp
->inp_socket
);
530 return(IPPROTO_DONE
);
535 return(IPPROTO_DONE
);
540 ip_2_ip6_hdr(struct ip6_hdr
*ip6
, struct ip
*ip
)
542 bzero(ip6
, sizeof *ip6
);
544 ip6
->ip6_vfc
= IPV6_VERSION
;
545 ip6
->ip6_plen
= ip
->ip_len
;
546 ip6
->ip6_nxt
= ip
->ip_p
;
547 ip6
->ip6_hlim
= ip
->ip_ttl
;
548 ip6
->ip6_src
.s6_addr32
[2] = ip6
->ip6_dst
.s6_addr32
[2] =
550 ip6
->ip6_src
.s6_addr32
[3] = ip
->ip_src
.s_addr
;
551 ip6
->ip6_dst
.s6_addr32
[3] = ip
->ip_dst
.s_addr
;
556 * subroutine of udp_input(), mainly for source code readability.
557 * caller must properly init udp_ip6 and udp_in6 beforehand.
560 udp_append(struct inpcb
*last
, struct ip
*ip
, struct mbuf
*n
, int off
)
562 struct sockaddr
*append_sa
;
563 struct mbuf
*opts
= NULL
;
565 if (last
->inp_flags
& INP_CONTROLOPTS
||
566 last
->inp_socket
->so_options
& SO_TIMESTAMP
) {
568 if (last
->inp_vflag
& INP_IPV6
) {
571 if (udp_ip6
.uip6_init_done
== 0) {
572 ip_2_ip6_hdr(&udp_ip6
.uip6_ip6
, ip
);
573 udp_ip6
.uip6_init_done
= 1;
575 savedflags
= last
->inp_flags
;
576 last
->inp_flags
&= ~INP_UNMAPPABLEOPTS
;
577 ip6_savecontrol(last
, &opts
, &udp_ip6
.uip6_ip6
, n
);
578 last
->inp_flags
= savedflags
;
581 ip_savecontrol(last
, &opts
, ip
, n
);
584 if (last
->inp_vflag
& INP_IPV6
) {
585 if (udp_in6
.uin6_init_done
== 0) {
586 in6_sin_2_v4mapsin6(&udp_in
, &udp_in6
.uin6_sin
);
587 udp_in6
.uin6_init_done
= 1;
589 append_sa
= (struct sockaddr
*)&udp_in6
.uin6_sin
;
592 append_sa
= (struct sockaddr
*)&udp_in
;
594 lwkt_gettoken(&last
->inp_socket
->so_rcv
.ssb_token
);
595 if (ssb_appendaddr(&last
->inp_socket
->so_rcv
, append_sa
, n
, opts
) == 0) {
599 udpstat
.udps_fullsock
++;
601 sorwakeup(last
->inp_socket
);
603 lwkt_reltoken(&last
->inp_socket
->so_rcv
.ssb_token
);
607 * Notify a udp user of an asynchronous error;
608 * just wake up so that he can collect error status.
611 udp_notify(struct inpcb
*inp
, int error
)
613 inp
->inp_socket
->so_error
= error
;
614 sorwakeup(inp
->inp_socket
);
615 sowwakeup(inp
->inp_socket
);
618 struct netmsg_udp_notify
{
619 struct netmsg_base base
;
620 void (*nm_notify
)(struct inpcb
*, int);
621 struct in_addr nm_faddr
;
626 udp_notifyall_oncpu(netmsg_t msg
)
628 struct netmsg_udp_notify
*nm
= (struct netmsg_udp_notify
*)msg
;
633 in_pcbnotifyall(&udbinfo
.pcblisthead
, nm
->nm_faddr
,
634 nm
->nm_arg
, nm
->nm_notify
);
635 lwkt_replymsg(&nm
->base
.lmsg
, 0);
638 /* XXX currently udp only runs on cpu 0 */
639 nextcpu
= mycpuid
+ 1;
640 if (nextcpu
< ncpus2
)
641 lwkt_forwardmsg(cpu_portfn(nextcpu
), &nm
->base
.lmsg
);
643 lwkt_replymsg(&nmsg
->base
.lmsg
, 0);
648 udp_rtchange(struct inpcb
*inp
, int err
)
651 /* XXX Nuke this, once UDP inpcbs are CPU localized */
652 if (inp
->inp_route
.ro_rt
&& inp
->inp_route
.ro_rt
->rt_cpuid
== mycpuid
) {
653 rtfree(inp
->inp_route
.ro_rt
);
654 inp
->inp_route
.ro_rt
= NULL
;
656 * A new route can be allocated the next time
657 * output is attempted.
661 in_rtchange(inp
, err
);
666 udp_ctlinput(netmsg_t msg
)
668 struct sockaddr
*sa
= msg
->ctlinput
.nm_arg
;
669 struct ip
*ip
= msg
->ctlinput
.nm_extra
;
670 int cmd
= msg
->ctlinput
.nm_cmd
;
672 void (*notify
) (struct inpcb
*, int) = udp_notify
;
673 struct in_addr faddr
;
676 KKASSERT(&curthread
->td_msgport
== cpu_portfn(0));
678 faddr
= ((struct sockaddr_in
*)sa
)->sin_addr
;
679 if (sa
->sa_family
!= AF_INET
|| faddr
.s_addr
== INADDR_ANY
)
682 if (PRC_IS_REDIRECT(cmd
)) {
684 notify
= udp_rtchange
;
685 } else if (cmd
== PRC_HOSTDEAD
) {
687 } else if ((unsigned)cmd
>= PRC_NCMDS
|| inetctlerrmap
[cmd
] == 0) {
692 uh
= (struct udphdr
*)((caddr_t
)ip
+ (ip
->ip_hl
<< 2));
693 inp
= in_pcblookup_hash(&udbinfo
, faddr
, uh
->uh_dport
,
694 ip
->ip_src
, uh
->uh_sport
, 0, NULL
);
695 if (inp
!= NULL
&& inp
->inp_socket
!= NULL
)
696 (*notify
)(inp
, inetctlerrmap
[cmd
]);
697 } else if (PRC_IS_REDIRECT(cmd
)) {
698 struct netmsg_udp_notify
*nm
;
700 KKASSERT(&curthread
->td_msgport
== cpu_portfn(0));
701 nm
= kmalloc(sizeof(*nm
), M_LWKTMSG
, M_INTWAIT
);
702 netmsg_init(&nm
->base
, NULL
, &netisr_afree_rport
,
703 0, udp_notifyall_oncpu
);
704 nm
->nm_faddr
= faddr
;
705 nm
->nm_arg
= inetctlerrmap
[cmd
];
706 nm
->nm_notify
= notify
;
707 lwkt_sendmsg(cpu_portfn(0), &nm
->base
.lmsg
);
710 * XXX We should forward msg upon PRC_HOSTHEAD and ip == NULL,
711 * once UDP inpcbs are CPU localized
713 KKASSERT(&curthread
->td_msgport
== cpu_portfn(0));
714 in_pcbnotifyall(&udbinfo
.pcblisthead
, faddr
, inetctlerrmap
[cmd
],
718 lwkt_replymsg(&msg
->lmsg
, 0);
722 udp_pcblist(SYSCTL_HANDLER_ARGS
)
728 error
= in_pcblist_global_nomarker(oidp
, arg1
, arg2
, req
, &xi
, &nxi
);
732 KKASSERT(xi
== NULL
);
736 KKASSERT(xi
== NULL
);
740 for (i
= 0; i
< nxi
; ++i
) {
741 error
= SYSCTL_OUT(req
, &xi
[i
], sizeof(xi
[i
]));
749 SYSCTL_PROC(_net_inet_udp
, UDPCTL_PCBLIST
, pcblist
, CTLFLAG_RD
, &udbinfo
, 0,
750 udp_pcblist
, "S,xinpcb", "List of active UDP sockets");
753 udp_getcred(SYSCTL_HANDLER_ARGS
)
755 struct sockaddr_in addrs
[2];
756 struct ucred cred0
, *cred
= NULL
;
760 error
= priv_check(req
->td
, PRIV_ROOT
);
763 error
= SYSCTL_IN(req
, addrs
, sizeof addrs
);
768 inp
= in_pcblookup_hash(&udbinfo
, addrs
[1].sin_addr
, addrs
[1].sin_port
,
769 addrs
[0].sin_addr
, addrs
[0].sin_port
, 1, NULL
);
770 if (inp
== NULL
|| inp
->inp_socket
== NULL
) {
773 if (inp
->inp_socket
->so_cred
!= NULL
) {
774 cred0
= *(inp
->inp_socket
->so_cred
);
783 return SYSCTL_OUT(req
, cred
, sizeof(struct ucred
));
786 SYSCTL_PROC(_net_inet_udp
, OID_AUTO
, getcred
, CTLTYPE_OPAQUE
|CTLFLAG_RW
,
787 0, 0, udp_getcred
, "S,ucred", "Get the ucred of a UDP connection");
790 udp_output(struct inpcb
*inp
, struct mbuf
*m
, struct sockaddr
*dstaddr
,
791 struct thread
*td
, int flags
)
794 int len
= m
->m_pkthdr
.len
;
795 struct sockaddr_in
*sin
; /* really is initialized before use */
796 int error
= 0, lport_any
= 0;
798 if (len
+ sizeof(struct udpiphdr
) > IP_MAXPACKET
) {
803 if (inp
->inp_lport
== 0) { /* unbound socket */
804 error
= in_pcbbind(inp
, NULL
, td
);
808 udbinfo_barrier_set();
809 in_pcbinswildcardhash(inp
);
810 udbinfo_barrier_rem();
814 if (dstaddr
!= NULL
) { /* destination address specified */
815 if (inp
->inp_faddr
.s_addr
!= INADDR_ANY
) {
816 /* already connected */
820 sin
= (struct sockaddr_in
*)dstaddr
;
821 if (!prison_remote_ip(td
, (struct sockaddr
*)&sin
)) {
822 error
= EAFNOSUPPORT
; /* IPv6 only jail */
826 if (inp
->inp_faddr
.s_addr
== INADDR_ANY
) {
827 /* no destination specified and not already connected */
835 * Calculate data length and get a mbuf
836 * for UDP and IP headers.
838 M_PREPEND(m
, sizeof(struct udpiphdr
), MB_DONTWAIT
);
845 * Fill in mbuf with extended UDP header
846 * and addresses and length put into network format.
848 ui
= mtod(m
, struct udpiphdr
*);
849 bzero(ui
->ui_x1
, sizeof ui
->ui_x1
); /* XXX still needed? */
850 ui
->ui_pr
= IPPROTO_UDP
;
853 * Set destination address.
855 if (dstaddr
!= NULL
) { /* use specified destination */
856 ui
->ui_dst
= sin
->sin_addr
;
857 ui
->ui_dport
= sin
->sin_port
;
858 } else { /* use connected destination */
859 ui
->ui_dst
= inp
->inp_faddr
;
860 ui
->ui_dport
= inp
->inp_fport
;
864 * Set source address.
866 if (inp
->inp_laddr
.s_addr
== INADDR_ANY
) {
867 struct sockaddr_in
*if_sin
;
869 if (dstaddr
== NULL
) {
871 * connect() had (or should have) failed because
872 * the interface had no IP address, but the
873 * application proceeded to call send() anyways.
879 /* Look up outgoing interface. */
880 if ((error
= in_pcbladdr(inp
, dstaddr
, &if_sin
, td
)))
882 ui
->ui_src
= if_sin
->sin_addr
; /* use address of interface */
884 ui
->ui_src
= inp
->inp_laddr
; /* use non-null bound address */
886 ui
->ui_sport
= inp
->inp_lport
;
887 KASSERT(inp
->inp_lport
!= 0, ("inp lport should have been bound"));
889 ui
->ui_ulen
= htons((u_short
)len
+ sizeof(struct udphdr
));
892 * Set up checksum and output datagram.
895 ui
->ui_sum
= in_pseudo(ui
->ui_src
.s_addr
, ui
->ui_dst
.s_addr
,
896 htons((u_short
)len
+ sizeof(struct udphdr
) + IPPROTO_UDP
));
897 m
->m_pkthdr
.csum_flags
= CSUM_UDP
;
898 m
->m_pkthdr
.csum_data
= offsetof(struct udphdr
, uh_sum
);
902 ((struct ip
*)ui
)->ip_len
= sizeof(struct udpiphdr
) + len
;
903 ((struct ip
*)ui
)->ip_ttl
= inp
->inp_ip_ttl
; /* XXX */
904 ((struct ip
*)ui
)->ip_tos
= inp
->inp_ip_tos
; /* XXX */
905 udpstat
.udps_opackets
++;
907 error
= ip_output(m
, inp
->inp_options
, &inp
->inp_route
,
908 (inp
->inp_socket
->so_options
& (SO_DONTROUTE
| SO_BROADCAST
)) |
909 flags
| IP_DEBUGROUTE
,
910 inp
->inp_moptions
, inp
);
913 * If this is the first data gram sent on an unbound and unconnected
914 * UDP socket, lport will be changed in this function. If target
915 * CPU after this lport changing is no longer the current CPU, then
916 * free the route entry allocated on the current CPU.
919 if (udp_addrcpu(inp
->inp_faddr
.s_addr
, inp
->inp_fport
,
920 inp
->inp_laddr
.s_addr
, inp
->inp_lport
) != mycpuid
) {
922 struct route
*ro
= &inp
->inp_route
;
924 if (ro
->ro_rt
!= NULL
)
926 bzero(ro
, sizeof(*ro
));
928 panic("UDP activity should only be in netisr0");
939 u_long udp_sendspace
= 9216; /* really max datagram size */
940 /* 40 1K datagrams */
941 SYSCTL_INT(_net_inet_udp
, UDPCTL_MAXDGRAM
, maxdgram
, CTLFLAG_RW
,
942 &udp_sendspace
, 0, "Maximum outgoing UDP datagram size");
944 u_long udp_recvspace
= 40 * (1024 +
946 sizeof(struct sockaddr_in6
)
948 sizeof(struct sockaddr_in
)
951 SYSCTL_INT(_net_inet_udp
, UDPCTL_RECVSPACE
, recvspace
, CTLFLAG_RW
,
952 &udp_recvspace
, 0, "Maximum incoming UDP datagram size");
955 * NOTE: (so) is referenced from soabort*() and netmsg_pru_abort()
956 * will sofree() it when we return.
959 udp_abort(netmsg_t msg
)
961 struct socket
*so
= msg
->abort
.base
.nm_so
;
965 KKASSERT(&curthread
->td_msgport
== cpu_portfn(0));
969 soisdisconnected(so
);
971 udbinfo_barrier_set();
973 udbinfo_barrier_rem();
978 lwkt_replymsg(&msg
->abort
.base
.lmsg
, error
);
982 udp_attach(netmsg_t msg
)
984 struct socket
*so
= msg
->attach
.base
.nm_so
;
985 struct pru_attach_info
*ai
= msg
->attach
.nm_ai
;
989 KKASSERT(&curthread
->td_msgport
== cpu_portfn(0));
996 error
= soreserve(so
, udp_sendspace
, udp_recvspace
, ai
->sb_rlimit
);
1000 udbinfo_barrier_set();
1001 error
= in_pcballoc(so
, &udbinfo
);
1002 udbinfo_barrier_rem();
1008 * Set default port for protocol processing prior to bind/connect.
1010 sosetport(so
, cpu_portfn(0));
1012 inp
= (struct inpcb
*)so
->so_pcb
;
1013 inp
->inp_vflag
|= INP_IPV4
;
1014 inp
->inp_ip_ttl
= ip_defttl
;
1017 lwkt_replymsg(&msg
->attach
.base
.lmsg
, error
);
1021 udp_bind(netmsg_t msg
)
1023 struct socket
*so
= msg
->bind
.base
.nm_so
;
1024 struct sockaddr
*nam
= msg
->bind
.nm_nam
;
1025 struct thread
*td
= msg
->bind
.nm_td
;
1026 struct sockaddr_in
*sin
= (struct sockaddr_in
*)nam
;
1032 error
= in_pcbbind(inp
, nam
, td
);
1034 if (sin
->sin_addr
.s_addr
!= INADDR_ANY
)
1035 inp
->inp_flags
|= INP_WASBOUND_NOTANY
;
1037 udbinfo_barrier_set();
1038 in_pcbinswildcardhash(inp
);
1039 udbinfo_barrier_rem();
1044 lwkt_replymsg(&msg
->bind
.base
.lmsg
, error
);
1048 udp_connect(netmsg_t msg
)
1050 struct socket
*so
= msg
->connect
.base
.nm_so
;
1051 struct sockaddr
*nam
= msg
->connect
.nm_nam
;
1052 struct thread
*td
= msg
->connect
.nm_td
;
1054 struct sockaddr_in
*sin
= (struct sockaddr_in
*)nam
;
1055 struct sockaddr_in
*if_sin
;
1059 KKASSERT(&curthread
->td_msgport
== cpu_portfn(0));
1067 if (msg
->connect
.nm_reconnect
& NMSG_RECONNECT_RECONNECT
) {
1068 panic("UDP does not support RECONNECT\n");
1070 msg
->connect
.nm_reconnect
&= ~NMSG_RECONNECT_RECONNECT
;
1071 in_pcblink(inp
, &udbinfo
);
1075 if (inp
->inp_faddr
.s_addr
!= INADDR_ANY
) {
1082 * Bind if we have to
1084 if (td
->td_proc
&& td
->td_proc
->p_ucred
->cr_prison
!= NULL
&&
1085 inp
->inp_laddr
.s_addr
== INADDR_ANY
) {
1086 error
= in_pcbbind(inp
, NULL
, td
);
1092 * Calculate the correct protocol processing thread. The connect
1093 * operation must run there.
1095 error
= in_pcbladdr(inp
, nam
, &if_sin
, td
);
1098 if (!prison_remote_ip(td
, nam
)) {
1099 error
= EAFNOSUPPORT
; /* IPv6 only jail */
1103 port
= udp_addrport(sin
->sin_addr
.s_addr
, sin
->sin_port
,
1104 inp
->inp_laddr
.s_addr
, inp
->inp_lport
);
1106 if (port
!= &curthread
->td_msgport
) {
1108 struct route
*ro
= &inp
->inp_route
;
1111 * in_pcbladdr() may have allocated a route entry for us
1112 * on the current CPU, but we need a route entry on the
1113 * inpcb's owner CPU, so free it here.
1115 if (ro
->ro_rt
!= NULL
)
1117 bzero(ro
, sizeof(*ro
));
1120 * We are moving the protocol processing port the socket
1121 * is on, we have to unlink here and re-link on the
1124 in_pcbunlink(so
->so_pcb
, &udbinfo
);
1125 /* in_pcbunlink(so->so_pcb, &udbinfo[mycpu->gd_cpuid]); */
1126 sosetport(so
, port
);
1127 msg
->connect
.nm_reconnect
|= NMSG_RECONNECT_RECONNECT
;
1128 msg
->connect
.base
.nm_dispatch
= udp_connect
;
1130 lwkt_forwardmsg(port
, &msg
->connect
.base
.lmsg
);
1131 /* msg invalid now */
1134 panic("UDP activity should only be in netisr0");
1138 KKASSERT(port
== &curthread
->td_msgport
);
1139 error
= udp_connect_oncpu(so
, td
, sin
, if_sin
);
1141 KKASSERT(msg
->connect
.nm_m
== NULL
);
1142 lwkt_replymsg(&msg
->connect
.base
.lmsg
, error
);
1146 udp_connect_oncpu(struct socket
*so
, struct thread
*td
,
1147 struct sockaddr_in
*sin
, struct sockaddr_in
*if_sin
)
1152 udbinfo_barrier_set();
1155 if (inp
->inp_flags
& INP_WILDCARD
)
1156 in_pcbremwildcardhash(inp
);
1157 error
= in_pcbconnect(inp
, (struct sockaddr
*)sin
, td
);
1161 * No more errors can occur, finish adjusting the socket
1162 * and change the processing port to reflect the connected
1163 * socket. Once set we can no longer safely mess with the
1167 } else if (error
== EAFNOSUPPORT
) { /* connection dissolved */
1169 * Follow traditional BSD behavior and retain
1170 * the local port binding. But, fix the old misbehavior
1171 * of overwriting any previously bound local address.
1173 if (!(inp
->inp_flags
& INP_WASBOUND_NOTANY
))
1174 inp
->inp_laddr
.s_addr
= INADDR_ANY
;
1175 in_pcbinswildcardhash(inp
);
1178 udbinfo_barrier_rem();
1183 udp_detach(netmsg_t msg
)
1185 struct socket
*so
= msg
->detach
.base
.nm_so
;
1189 KKASSERT(&curthread
->td_msgport
== cpu_portfn(0));
1193 udbinfo_barrier_set();
1195 udbinfo_barrier_rem();
1200 lwkt_replymsg(&msg
->detach
.base
.lmsg
, error
);
1204 udp_disconnect(netmsg_t msg
)
1206 struct socket
*so
= msg
->disconnect
.base
.nm_so
;
1211 KKASSERT(&curthread
->td_msgport
== cpu_portfn(0));
1218 if (inp
->inp_faddr
.s_addr
== INADDR_ANY
) {
1225 udbinfo_barrier_set();
1226 in_pcbdisconnect(inp
);
1227 udbinfo_barrier_rem();
1229 soclrstate(so
, SS_ISCONNECTED
); /* XXX */
1232 ro
= &inp
->inp_route
;
1233 if (ro
->ro_rt
!= NULL
)
1235 bzero(ro
, sizeof(*ro
));
1238 lwkt_replymsg(&msg
->disconnect
.base
.lmsg
, error
);
1242 udp_send(netmsg_t msg
)
1244 struct socket
*so
= msg
->send
.base
.nm_so
;
1245 struct mbuf
*m
= msg
->send
.nm_m
;
1246 struct sockaddr
*addr
= msg
->send
.nm_addr
;
1247 int pru_flags
= msg
->send
.nm_flags
;
1251 KKASSERT(&curthread
->td_msgport
== cpu_portfn(0));
1252 KKASSERT(msg
->send
.nm_control
== NULL
);
1256 struct thread
*td
= msg
->send
.nm_td
;
1259 if (pru_flags
& PRUS_DONTROUTE
)
1260 flags
|= SO_DONTROUTE
;
1261 error
= udp_output(inp
, m
, addr
, td
, flags
);
1267 if (pru_flags
& PRUS_FREEADDR
)
1268 kfree(addr
, M_SONAME
);
1270 if ((pru_flags
& PRUS_NOREPLY
) == 0)
1271 lwkt_replymsg(&msg
->send
.base
.lmsg
, error
);
1275 udp_shutdown(netmsg_t msg
)
1277 struct socket
*so
= msg
->shutdown
.base
.nm_so
;
1281 KKASSERT(&curthread
->td_msgport
== cpu_portfn(0));
1290 lwkt_replymsg(&msg
->shutdown
.base
.lmsg
, error
);
1296 lwkt_serialize_enter(&udbinfo_slize
);
1300 udbinfo_unlock(void)
1302 lwkt_serialize_exit(&udbinfo_slize
);
1306 udbinfo_barrier_set(void)
1308 netisr_barrier_set(udbinfo_br
);
1313 udbinfo_barrier_rem(void)
1316 netisr_barrier_rem(udbinfo_br
);
1319 struct pr_usrreqs udp_usrreqs
= {
1320 .pru_abort
= udp_abort
,
1321 .pru_accept
= pr_generic_notsupp
,
1322 .pru_attach
= udp_attach
,
1323 .pru_bind
= udp_bind
,
1324 .pru_connect
= udp_connect
,
1325 .pru_connect2
= pr_generic_notsupp
,
1326 .pru_control
= in_control_dispatch
,
1327 .pru_detach
= udp_detach
,
1328 .pru_disconnect
= udp_disconnect
,
1329 .pru_listen
= pr_generic_notsupp
,
1330 .pru_peeraddr
= in_setpeeraddr_dispatch
,
1331 .pru_rcvd
= pr_generic_notsupp
,
1332 .pru_rcvoob
= pr_generic_notsupp
,
1333 .pru_send
= udp_send
,
1334 .pru_sense
= pru_sense_null
,
1335 .pru_shutdown
= udp_shutdown
,
1336 .pru_sockaddr
= in_setsockaddr_dispatch
,
1337 .pru_sosend
= sosendudp
,
1338 .pru_soreceive
= soreceive