2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * The User Datagram Protocol (UDP).
8 * Version: $Id: udp.c,v 1.102 2002/02/01 22:01:04 davem Exp $
10 * Authors: Ross Biro, <bir7@leland.Stanford.Edu>
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
13 * Alan Cox, <Alan.Cox@linux.org>
14 * Hirokazu Takahashi, <taka@valinux.co.jp>
17 * Alan Cox : verify_area() calls
18 * Alan Cox : stopped close while in use off icmp
19 * messages. Not a fix but a botch that
20 * for udp at least is 'valid'.
21 * Alan Cox : Fixed icmp handling properly
22 * Alan Cox : Correct error for oversized datagrams
23 * Alan Cox : Tidied select() semantics.
24 * Alan Cox : udp_err() fixed properly, also now
25 * select and read wake correctly on errors
26 * Alan Cox : udp_send verify_area moved to avoid mem leak
27 * Alan Cox : UDP can count its memory
28 * Alan Cox : send to an unknown connection causes
29 * an ECONNREFUSED off the icmp, but
31 * Alan Cox : Switched to new sk_buff handlers. No more backlog!
32 * Alan Cox : Using generic datagram code. Even smaller and the PEEK
33 * bug no longer crashes it.
34 * Fred Van Kempen : Net2e support for sk->broadcast.
35 * Alan Cox : Uses skb_free_datagram
36 * Alan Cox : Added get/set sockopt support.
37 * Alan Cox : Broadcasting without option set returns EACCES.
38 * Alan Cox : No wakeup calls. Instead we now use the callbacks.
39 * Alan Cox : Use ip_tos and ip_ttl
40 * Alan Cox : SNMP Mibs
41 * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support.
42 * Matt Dillon : UDP length checks.
43 * Alan Cox : Smarter af_inet used properly.
44 * Alan Cox : Use new kernel side addressing.
45 * Alan Cox : Incorrect return on truncated datagram receive.
46 * Arnt Gulbrandsen : New udp_send and stuff
47 * Alan Cox : Cache last socket
48 * Alan Cox : Route cache
49 * Jon Peatfield : Minor efficiency fix to sendto().
50 * Mike Shaver : RFC1122 checks.
51 * Alan Cox : Nonblocking error fix.
52 * Willy Konynenberg : Transparent proxying support.
53 * Mike McLagan : Routing by source
54 * David S. Miller : New socket lookup architecture.
55 * Last socket cache retained as it
56 * does have a high hit rate.
57 * Olaf Kirch : Don't linearise iovec on sendmsg.
58 * Andi Kleen : Some cleanups, cache destination entry
60 * Vitaly E. Lavrov : Transparent proxy revived after year coma.
61 * Melvin Smith : Check msg_name not msg_namelen in sendto(),
62 * return ENOTCONN for unconnected sockets (POSIX)
63 * Janos Farkas : don't deliver multi/broadcasts to a different
64 * bound-to-device socket
65 * Hirokazu Takahashi : HW checksumming for outgoing UDP
67 * Hirokazu Takahashi : sendfile() on UDP works now.
68 * Arnaldo C. Melo : convert /proc/net/udp to seq_file
69 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
70 * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind
71 * a single port at the same time.
72 * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support
75 * This program is free software; you can redistribute it and/or
76 * modify it under the terms of the GNU General Public License
77 * as published by the Free Software Foundation; either version
78 * 2 of the License, or (at your option) any later version.
81 #include <asm/system.h>
82 #include <asm/uaccess.h>
83 #include <asm/ioctls.h>
84 #include <linux/types.h>
85 #include <linux/fcntl.h>
86 #include <linux/socket.h>
87 #include <linux/sockios.h>
89 #include <linux/errno.h>
90 #include <linux/timer.h>
92 #include <linux/config.h>
93 #include <linux/inet.h>
94 #include <linux/ipv6.h>
95 #include <linux/netdevice.h>
98 #include <net/protocol.h>
99 #include <linux/skbuff.h>
100 #include <linux/proc_fs.h>
101 #include <linux/seq_file.h>
102 #include <net/sock.h>
104 #include <net/icmp.h>
105 #include <net/route.h>
106 #include <net/inet_common.h>
107 #include <net/checksum.h>
108 #include <net/xfrm.h>
111 * Snmp MIB for the UDP layer
114 DEFINE_SNMP_STAT(struct udp_mib
, udp_statistics
);
116 struct hlist_head udp_hash
[UDP_HTABLE_SIZE
];
117 rwlock_t udp_hash_lock
= RW_LOCK_UNLOCKED
;
119 /* Shared by v4/v6 udp. */
122 static int udp_v4_get_port(struct sock
*sk
, unsigned short snum
)
124 struct hlist_node
*node
;
126 struct inet_opt
*inet
= inet_sk(sk
);
128 write_lock_bh(&udp_hash_lock
);
130 int best_size_so_far
, best
, result
, i
;
132 if (udp_port_rover
> sysctl_local_port_range
[1] ||
133 udp_port_rover
< sysctl_local_port_range
[0])
134 udp_port_rover
= sysctl_local_port_range
[0];
135 best_size_so_far
= 32767;
136 best
= result
= udp_port_rover
;
137 for (i
= 0; i
< UDP_HTABLE_SIZE
; i
++, result
++) {
138 struct hlist_head
*list
;
141 list
= &udp_hash
[result
& (UDP_HTABLE_SIZE
- 1)];
142 if (hlist_empty(list
)) {
143 if (result
> sysctl_local_port_range
[1])
144 result
= sysctl_local_port_range
[0] +
145 ((result
- sysctl_local_port_range
[0]) &
146 (UDP_HTABLE_SIZE
- 1));
150 sk_for_each(sk2
, node
, list
)
151 if (++size
>= best_size_so_far
)
153 best_size_so_far
= size
;
158 for(i
= 0; i
< (1 << 16) / UDP_HTABLE_SIZE
; i
++, result
+= UDP_HTABLE_SIZE
) {
159 if (result
> sysctl_local_port_range
[1])
160 result
= sysctl_local_port_range
[0]
161 + ((result
- sysctl_local_port_range
[0]) &
162 (UDP_HTABLE_SIZE
- 1));
163 if (!udp_lport_inuse(result
))
166 if (i
>= (1 << 16) / UDP_HTABLE_SIZE
)
169 udp_port_rover
= snum
= result
;
171 sk_for_each(sk2
, node
,
172 &udp_hash
[snum
& (UDP_HTABLE_SIZE
- 1)]) {
173 struct inet_opt
*inet2
= inet_sk(sk2
);
175 if (inet2
->num
== snum
&&
177 !ipv6_only_sock(sk2
) &&
178 sk2
->sk_bound_dev_if
== sk
->sk_bound_dev_if
&&
179 (!inet2
->rcv_saddr
||
181 inet2
->rcv_saddr
== inet
->rcv_saddr
) &&
182 (!sk2
->sk_reuse
|| !sk
->sk_reuse
))
187 if (sk_unhashed(sk
)) {
188 struct hlist_head
*h
= &udp_hash
[snum
& (UDP_HTABLE_SIZE
- 1)];
191 sock_prot_inc_use(sk
->sk_prot
);
193 write_unlock_bh(&udp_hash_lock
);
197 write_unlock_bh(&udp_hash_lock
);
201 static void udp_v4_hash(struct sock
*sk
)
206 static void udp_v4_unhash(struct sock
*sk
)
208 write_lock_bh(&udp_hash_lock
);
209 if (sk_del_node_init(sk
)) {
210 inet_sk(sk
)->num
= 0;
211 sock_prot_dec_use(sk
->sk_prot
);
213 write_unlock_bh(&udp_hash_lock
);
216 /* UDP is nearly always wildcards out the wazoo, it makes no sense to try
217 * harder than this. -DaveM
219 struct sock
*udp_v4_lookup_longway(u32 saddr
, u16 sport
, u32 daddr
, u16 dport
, int dif
)
221 struct sock
*sk
, *result
= NULL
;
222 struct hlist_node
*node
;
223 unsigned short hnum
= ntohs(dport
);
226 sk_for_each(sk
, node
, &udp_hash
[hnum
& (UDP_HTABLE_SIZE
- 1)]) {
227 struct inet_opt
*inet
= inet_sk(sk
);
229 if (inet
->num
== hnum
&& !ipv6_only_sock(sk
)) {
230 int score
= (sk
->sk_family
== PF_INET
? 1 : 0);
231 if (inet
->rcv_saddr
) {
232 if (inet
->rcv_saddr
!= daddr
)
237 if (inet
->daddr
!= saddr
)
242 if (inet
->dport
!= sport
)
246 if (sk
->sk_bound_dev_if
) {
247 if (sk
->sk_bound_dev_if
!= dif
)
254 } else if(score
> badness
) {
263 __inline__
struct sock
*udp_v4_lookup(u32 saddr
, u16 sport
, u32 daddr
, u16 dport
, int dif
)
267 read_lock(&udp_hash_lock
);
268 sk
= udp_v4_lookup_longway(saddr
, sport
, daddr
, dport
, dif
);
271 read_unlock(&udp_hash_lock
);
275 static inline struct sock
*udp_v4_mcast_next(struct sock
*sk
,
276 u16 loc_port
, u32 loc_addr
,
277 u16 rmt_port
, u32 rmt_addr
,
280 struct hlist_node
*node
;
282 unsigned short hnum
= ntohs(loc_port
);
284 sk_for_each_from(s
, node
) {
285 struct inet_opt
*inet
= inet_sk(s
);
287 if (inet
->num
!= hnum
||
288 (inet
->daddr
&& inet
->daddr
!= rmt_addr
) ||
289 (inet
->dport
!= rmt_port
&& inet
->dport
) ||
290 (inet
->rcv_saddr
&& inet
->rcv_saddr
!= loc_addr
) ||
292 (s
->sk_bound_dev_if
&& s
->sk_bound_dev_if
!= dif
))
294 if (!ip_mc_sf_allow(sk
, loc_addr
, rmt_addr
, dif
))
304 * This routine is called by the ICMP module when it gets some
305 * sort of error condition. If err < 0 then the socket should
306 * be closed and the error returned to the user. If err > 0
307 * it's just the icmp type << 8 | icmp code.
308 * Header points to the ip header of the error packet. We move
309 * on past this. Then (as it used to claim before adjustment)
310 * header points to the first 8 bytes of the udp header. We need
311 * to find the appropriate port.
314 void udp_err(struct sk_buff
*skb
, u32 info
)
316 struct inet_opt
*inet
;
317 struct iphdr
*iph
= (struct iphdr
*)skb
->data
;
318 struct udphdr
*uh
= (struct udphdr
*)(skb
->data
+(iph
->ihl
<<2));
319 int type
= skb
->h
.icmph
->type
;
320 int code
= skb
->h
.icmph
->code
;
325 sk
= udp_v4_lookup(iph
->daddr
, uh
->dest
, iph
->saddr
, uh
->source
, skb
->dev
->ifindex
);
327 ICMP_INC_STATS_BH(IcmpInErrors
);
328 return; /* No socket for error */
337 case ICMP_TIME_EXCEEDED
:
340 case ICMP_SOURCE_QUENCH
:
342 case ICMP_PARAMETERPROB
:
346 case ICMP_DEST_UNREACH
:
347 if (code
== ICMP_FRAG_NEEDED
) { /* Path MTU discovery */
348 if (inet
->pmtudisc
!= IP_PMTUDISC_DONT
) {
356 if (code
<= NR_ICMP_UNREACH
) {
357 harderr
= icmp_err_convert
[code
].fatal
;
358 err
= icmp_err_convert
[code
].errno
;
364 * RFC1122: OK. Passes ICMP errors back to application, as per
367 if (!inet
->recverr
) {
368 if (!harderr
|| sk
->sk_state
!= TCP_ESTABLISHED
)
371 ip_icmp_error(sk
, skb
, err
, uh
->dest
, info
, (u8
*)(uh
+1));
374 sk
->sk_error_report(sk
);
380 * Throw away all pending data and cancel the corking. Socket is locked.
382 static void udp_flush_pending_frames(struct sock
*sk
)
384 struct udp_opt
*up
= udp_sk(sk
);
388 ip_flush_pending_frames(sk
);
393 * Push out all pending data as one UDP datagram. Socket is locked.
395 static int udp_push_pending_frames(struct sock
*sk
, struct udp_opt
*up
)
401 /* Grab the skbuff where UDP header space exists. */
402 if ((skb
= skb_peek(&sk
->sk_write_queue
)) == NULL
)
406 * Create a UDP header
409 uh
->source
= up
->sport
;
410 uh
->dest
= up
->dport
;
411 uh
->len
= htons(up
->len
);
414 if (sk
->sk_no_check
== UDP_CSUM_NOXMIT
) {
415 skb
->ip_summed
= CHECKSUM_NONE
;
419 if (skb_queue_len(&sk
->sk_write_queue
) == 1) {
421 * Only one fragment on the socket.
423 if (skb
->ip_summed
== CHECKSUM_HW
) {
424 skb
->csum
= offsetof(struct udphdr
, check
);
425 uh
->check
= ~csum_tcpudp_magic(up
->saddr
, up
->daddr
,
426 up
->len
, IPPROTO_UDP
, 0);
428 skb
->csum
= csum_partial((char *)uh
,
429 sizeof(struct udphdr
), skb
->csum
);
430 uh
->check
= csum_tcpudp_magic(up
->saddr
, up
->daddr
,
431 up
->len
, IPPROTO_UDP
, skb
->csum
);
436 unsigned int csum
= 0;
438 * HW-checksum won't work as there are two or more
439 * fragments on the socket so that all csums of sk_buffs
440 * should be together.
442 if (skb
->ip_summed
== CHECKSUM_HW
) {
443 int offset
= (unsigned char *)uh
- skb
->data
;
444 skb
->csum
= skb_checksum(skb
, offset
, skb
->len
- offset
, 0);
446 skb
->ip_summed
= CHECKSUM_NONE
;
448 skb
->csum
= csum_partial((char *)uh
,
449 sizeof(struct udphdr
), skb
->csum
);
452 skb_queue_walk(&sk
->sk_write_queue
, skb
) {
453 csum
= csum_add(csum
, skb
->csum
);
455 uh
->check
= csum_tcpudp_magic(up
->saddr
, up
->daddr
,
456 up
->len
, IPPROTO_UDP
, csum
);
461 err
= ip_push_pending_frames(sk
);
469 static unsigned short udp_check(struct udphdr
*uh
, int len
, unsigned long saddr
, unsigned long daddr
, unsigned long base
)
471 return(csum_tcpudp_magic(saddr
, daddr
, len
, IPPROTO_UDP
, base
));
474 int udp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
477 struct inet_opt
*inet
= inet_sk(sk
);
478 struct udp_opt
*up
= udp_sk(sk
);
480 struct ipcm_cookie ipc
;
481 struct rtable
*rt
= NULL
;
484 u32 daddr
, faddr
, saddr
;
488 int corkreq
= up
->corkflag
|| msg
->msg_flags
&MSG_MORE
;
490 /* This check is ONLY to check for arithmetic overflow
491 on integer(!) len. Not more! Real check will be made
494 BTW socket.c -> af_*.c -> ... make multiple
495 invalid conversions size_t -> int. We MUST repair it f.e.
496 by replacing all of them with size_t and revise all
497 the places sort of len += sizeof(struct iphdr)
498 If len was ULONG_MAX-10 it would be cathastrophe --ANK
501 if (len
< 0 || len
> 0xFFFF)
508 if (msg
->msg_flags
&MSG_OOB
) /* Mirror BSD error message compatibility */
515 * There are pending frames.
516 * The socket lock must be held while it's corked.
519 if (likely(up
->pending
))
523 ulen
+= sizeof(struct udphdr
);
526 * Get and verify the address.
529 struct sockaddr_in
* usin
= (struct sockaddr_in
*)msg
->msg_name
;
530 if (msg
->msg_namelen
< sizeof(*usin
))
532 if (usin
->sin_family
!= AF_INET
) {
533 if (usin
->sin_family
!= AF_UNSPEC
)
537 daddr
= usin
->sin_addr
.s_addr
;
538 dport
= usin
->sin_port
;
542 if (sk
->sk_state
!= TCP_ESTABLISHED
)
543 return -EDESTADDRREQ
;
546 /* Open fast path for connected socket.
547 Route will not be used, if at least one option is set.
551 ipc
.addr
= inet
->saddr
;
553 ipc
.oif
= sk
->sk_bound_dev_if
;
554 if (msg
->msg_controllen
) {
555 err
= ip_cmsg_send(msg
, &ipc
);
566 ipc
.addr
= faddr
= daddr
;
568 if (ipc
.opt
&& ipc
.opt
->srr
) {
571 faddr
= ipc
.opt
->faddr
;
574 tos
= RT_TOS(inet
->tos
);
575 if (sk
->sk_localroute
|| (msg
->msg_flags
& MSG_DONTROUTE
) ||
576 (ipc
.opt
&& ipc
.opt
->is_strictroute
)) {
581 if (MULTICAST(daddr
)) {
583 ipc
.oif
= inet
->mc_index
;
585 saddr
= inet
->mc_addr
;
590 rt
= (struct rtable
*)sk_dst_check(sk
, 0);
593 struct flowi fl
= { .oif
= ipc
.oif
,
598 .proto
= IPPROTO_UDP
,
600 { .sport
= inet
->sport
,
601 .dport
= dport
} } };
602 err
= ip_route_output_flow(&rt
, &fl
, sk
, !(msg
->msg_flags
&MSG_DONTWAIT
));
607 if ((rt
->rt_flags
& RTCF_BROADCAST
) &&
608 !sock_flag(sk
, SOCK_BROADCAST
))
611 sk_dst_set(sk
, dst_clone(&rt
->u
.dst
));
614 if (msg
->msg_flags
&MSG_CONFIRM
)
620 daddr
= ipc
.addr
= rt
->rt_dst
;
623 if (unlikely(up
->pending
)) {
624 /* The socket is already corked while preparing it. */
625 /* ... which is an evident application bug. --ANK */
628 NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG
"udp cork app bug 2\n"));
633 * Now cork the socket to pend data.
638 up
->sport
= inet
->sport
;
643 err
= ip_append_data(sk
, ip_generic_getfrag
, msg
->msg_iov
, ulen
,
644 sizeof(struct udphdr
), &ipc
, rt
,
645 corkreq
? msg
->msg_flags
|MSG_MORE
: msg
->msg_flags
);
647 udp_flush_pending_frames(sk
);
649 err
= udp_push_pending_frames(sk
, up
);
657 UDP_INC_STATS_USER(UdpOutDatagrams
);
663 dst_confirm(&rt
->u
.dst
);
664 if (!(msg
->msg_flags
&MSG_PROBE
) || len
)
665 goto back_from_confirm
;
670 int udp_sendpage(struct sock
*sk
, struct page
*page
, int offset
, size_t size
, int flags
)
672 struct udp_opt
*up
= udp_sk(sk
);
676 struct msghdr msg
= { .msg_flags
= flags
|MSG_MORE
};
678 /* Call udp_sendmsg to specify destination address which
679 * sendpage interface can't pass.
680 * This will succeed only when the socket is connected.
682 ret
= udp_sendmsg(NULL
, sk
, &msg
, 0);
689 if (unlikely(!up
->pending
)) {
692 NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG
"udp cork app bug 3\n"));
696 ret
= ip_append_page(sk
, page
, offset
, size
, flags
);
697 if (ret
== -EOPNOTSUPP
) {
699 return sock_no_sendpage(sk
->sk_socket
, page
, offset
,
703 udp_flush_pending_frames(sk
);
708 if (!(up
->corkflag
|| (flags
&MSG_MORE
)))
709 ret
= udp_push_pending_frames(sk
, up
);
718 * IOCTL requests applicable to the UDP protocol
721 int udp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
727 int amount
= atomic_read(&sk
->sk_wmem_alloc
);
728 return put_user(amount
, (int *)arg
);
734 unsigned long amount
;
737 spin_lock_irq(&sk
->sk_receive_queue
.lock
);
738 skb
= skb_peek(&sk
->sk_receive_queue
);
741 * We will only return the amount
742 * of this packet since that is all
745 amount
= skb
->len
- sizeof(struct udphdr
);
747 spin_unlock_irq(&sk
->sk_receive_queue
.lock
);
748 return put_user(amount
, (int *)arg
);
757 static __inline__
int __udp_checksum_complete(struct sk_buff
*skb
)
759 return (unsigned short)csum_fold(skb_checksum(skb
, 0, skb
->len
, skb
->csum
));
762 static __inline__
int udp_checksum_complete(struct sk_buff
*skb
)
764 return skb
->ip_summed
!= CHECKSUM_UNNECESSARY
&&
765 __udp_checksum_complete(skb
);
769 * This should be easy, if there is something there we
770 * return it, otherwise we block.
773 int udp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
774 int len
, int noblock
, int flags
, int *addr_len
)
776 struct inet_opt
*inet
= inet_sk(sk
);
777 struct sockaddr_in
*sin
= (struct sockaddr_in
*)msg
->msg_name
;
782 * Check any passed addresses
785 *addr_len
=sizeof(*sin
);
787 if (flags
& MSG_ERRQUEUE
)
788 return ip_recv_error(sk
, msg
, len
);
790 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
794 copied
= skb
->len
- sizeof(struct udphdr
);
797 msg
->msg_flags
|= MSG_TRUNC
;
800 if (skb
->ip_summed
==CHECKSUM_UNNECESSARY
) {
801 err
= skb_copy_datagram_iovec(skb
, sizeof(struct udphdr
), msg
->msg_iov
,
803 } else if (msg
->msg_flags
&MSG_TRUNC
) {
804 if (__udp_checksum_complete(skb
))
806 err
= skb_copy_datagram_iovec(skb
, sizeof(struct udphdr
), msg
->msg_iov
,
809 err
= skb_copy_and_csum_datagram_iovec(skb
, sizeof(struct udphdr
), msg
->msg_iov
);
818 sock_recv_timestamp(msg
, sk
, skb
);
820 /* Copy the address. */
823 sin
->sin_family
= AF_INET
;
824 sin
->sin_port
= skb
->h
.uh
->source
;
825 sin
->sin_addr
.s_addr
= skb
->nh
.iph
->saddr
;
826 memset(sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
828 if (inet
->cmsg_flags
)
829 ip_cmsg_recv(msg
, skb
);
833 skb_free_datagram(sk
, skb
);
838 UDP_INC_STATS_BH(UdpInErrors
);
841 if (flags
&MSG_PEEK
) {
843 spin_lock_irq(&sk
->sk_receive_queue
.lock
);
844 if (skb
== skb_peek(&sk
->sk_receive_queue
)) {
845 __skb_unlink(skb
, &sk
->sk_receive_queue
);
848 spin_unlock_irq(&sk
->sk_receive_queue
.lock
);
853 skb_free_datagram(sk
, skb
);
858 int udp_connect(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
860 struct inet_opt
*inet
= inet_sk(sk
);
861 struct sockaddr_in
*usin
= (struct sockaddr_in
*) uaddr
;
868 if (addr_len
< sizeof(*usin
))
871 if (usin
->sin_family
!= AF_INET
)
872 return -EAFNOSUPPORT
;
876 oif
= sk
->sk_bound_dev_if
;
878 if (MULTICAST(usin
->sin_addr
.s_addr
)) {
880 oif
= inet
->mc_index
;
882 saddr
= inet
->mc_addr
;
884 err
= ip_route_connect(&rt
, usin
->sin_addr
.s_addr
, saddr
,
885 RT_CONN_FLAGS(sk
), oif
,
887 inet
->sport
, usin
->sin_port
, sk
);
890 if ((rt
->rt_flags
& RTCF_BROADCAST
) && !sock_flag(sk
, SOCK_BROADCAST
)) {
895 inet
->saddr
= rt
->rt_src
; /* Update source address */
896 if (!inet
->rcv_saddr
)
897 inet
->rcv_saddr
= rt
->rt_src
;
898 inet
->daddr
= rt
->rt_dst
;
899 inet
->dport
= usin
->sin_port
;
900 sk
->sk_state
= TCP_ESTABLISHED
;
903 sk_dst_set(sk
, &rt
->u
.dst
);
907 int udp_disconnect(struct sock
*sk
, int flags
)
909 struct inet_opt
*inet
= inet_sk(sk
);
911 * 1003.1g - break association.
914 sk
->sk_state
= TCP_CLOSE
;
917 sk
->sk_bound_dev_if
= 0;
918 if (!(sk
->sk_userlocks
& SOCK_BINDADDR_LOCK
))
919 inet_reset_saddr(sk
);
921 if (!(sk
->sk_userlocks
& SOCK_BINDPORT_LOCK
)) {
922 sk
->sk_prot
->unhash(sk
);
929 static void udp_close(struct sock
*sk
, long timeout
)
931 inet_sock_release(sk
);
935 * 1 if the the UDP system should process it
936 * 0 if we should drop this packet
937 * -1 if it should get processed by xfrm4_rcv_encap
939 static int udp_encap_rcv(struct sock
* sk
, struct sk_buff
*skb
)
941 struct udp_opt
*up
= udp_sk(sk
);
942 struct udphdr
*uh
= skb
->h
.uh
;
946 __u8
*udpdata
= (__u8
*)uh
+ sizeof(struct udphdr
);
947 __u32
*udpdata32
= (__u32
*)udpdata
;
948 __u16 encap_type
= up
->encap_type
;
950 /* if we're overly short, let UDP handle it */
951 if (udpdata
> skb
->tail
)
954 /* if this is not encapsulated socket, then just return now */
958 len
= skb
->tail
- udpdata
;
960 switch (encap_type
) {
961 case UDP_ENCAP_ESPINUDP
:
962 /* Check if this is a keepalive packet. If so, eat it. */
963 if (len
== 1 && udpdata
[0] == 0xff) {
965 } else if (len
> sizeof(struct ip_esp_hdr
) && udpdata32
[0] != 0 ) {
966 /* ESP Packet without Non-ESP header */
967 len
= sizeof(struct udphdr
);
969 /* Must be an IKE packet.. pass it through */
972 /* At this point we are sure that this is an ESPinUDP packet,
973 * so we need to remove 'len' bytes from the packet (the UDP
974 * header and optional ESP marker bytes) and then modify the
975 * protocol to ESP, and then call into the transform receiver.
978 /* Now we can update and verify the packet length... */
980 iphlen
= iph
->ihl
<< 2;
981 iph
->tot_len
= htons(ntohs(iph
->tot_len
) - len
);
982 if (skb
->len
< iphlen
+ len
) {
983 /* packet is too small!?! */
987 /* pull the data buffer up to the ESP header and set the
988 * transport header to point to ESP. Keep UDP on the stack
991 skb
->h
.raw
= skb_pull(skb
, len
);
993 /* modify the protocol (it's ESP!) */
994 iph
->protocol
= IPPROTO_ESP
;
996 /* and let the caller know to send this into the ESP processor... */
1000 printk(KERN_INFO
"udp_encap_rcv(): Unhandled UDP encap type: %u\n",
1009 * >0: "udp encap" protocol resubmission
1011 * Note that in the success and error cases, the skb is assumed to
1012 * have either been requeued or freed.
1014 static int udp_queue_rcv_skb(struct sock
* sk
, struct sk_buff
*skb
)
1016 struct udp_opt
*up
= udp_sk(sk
);
1019 * Charge it to the socket, dropping if the queue is full.
1021 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
)) {
1026 if (up
->encap_type
) {
1028 * This is an encapsulation socket, so let's see if this is
1029 * an encapsulated packet.
1030 * If it's a keepalive packet, then just eat it.
1031 * If it's an encapsulateed packet, then pass it to the
1032 * IPsec xfrm input and return the response
1033 * appropriately. Otherwise, just fall through and
1034 * pass this up the UDP socket.
1038 ret
= udp_encap_rcv(sk
, skb
);
1040 /* Eat the packet .. */
1045 /* process the ESP packet */
1046 ret
= xfrm4_rcv_encap(skb
, up
->encap_type
);
1047 UDP_INC_STATS_BH(UdpInDatagrams
);
1050 /* FALLTHROUGH -- it's a UDP Packet */
1053 if (sk
->sk_filter
&& skb
->ip_summed
!= CHECKSUM_UNNECESSARY
) {
1054 if (__udp_checksum_complete(skb
)) {
1055 UDP_INC_STATS_BH(UdpInErrors
);
1059 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1062 if (sock_queue_rcv_skb(sk
,skb
)<0) {
1063 UDP_INC_STATS_BH(UdpInErrors
);
1067 UDP_INC_STATS_BH(UdpInDatagrams
);
1072 * Multicasts and broadcasts go to each listener.
1074 * Note: called only from the BH handler context,
1075 * so we don't need to lock the hashes.
1077 static int udp_v4_mcast_deliver(struct sk_buff
*skb
, struct udphdr
*uh
,
1078 u32 saddr
, u32 daddr
)
1083 read_lock(&udp_hash_lock
);
1084 sk
= sk_head(&udp_hash
[ntohs(uh
->dest
) & (UDP_HTABLE_SIZE
- 1)]);
1085 dif
= skb
->dev
->ifindex
;
1086 sk
= udp_v4_mcast_next(sk
, uh
->dest
, daddr
, uh
->source
, saddr
, dif
);
1088 struct sock
*sknext
= NULL
;
1091 struct sk_buff
*skb1
= skb
;
1093 sknext
= udp_v4_mcast_next(sk_next(sk
), uh
->dest
, daddr
,
1094 uh
->source
, saddr
, dif
);
1096 skb1
= skb_clone(skb
, GFP_ATOMIC
);
1099 int ret
= udp_queue_rcv_skb(sk
, skb1
);
1101 /* we should probably re-process instead
1102 * of dropping packets here. */
1109 read_unlock(&udp_hash_lock
);
1113 /* Initialize UDP checksum. If exited with zero value (success),
1114 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1115 * Otherwise, csum completion requires chacksumming packet body,
1116 * including udp header and folding it to skb->csum.
1118 static int udp_checksum_init(struct sk_buff
*skb
, struct udphdr
*uh
,
1119 unsigned short ulen
, u32 saddr
, u32 daddr
)
1121 if (uh
->check
== 0) {
1122 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1123 } else if (skb
->ip_summed
== CHECKSUM_HW
) {
1124 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1125 if (!udp_check(uh
, ulen
, saddr
, daddr
, skb
->csum
))
1127 NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG
"udp v4 hw csum failure.\n"));
1128 skb
->ip_summed
= CHECKSUM_NONE
;
1130 if (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
)
1131 skb
->csum
= csum_tcpudp_nofold(saddr
, daddr
, ulen
, IPPROTO_UDP
, 0);
1132 /* Probably, we should checksum udp header (it should be in cache
1133 * in any case) and data in tiny packets (< rx copybreak).
1139 * All we need to do is get the socket, and then do a checksum.
1142 int udp_rcv(struct sk_buff
*skb
)
1146 unsigned short ulen
;
1147 struct rtable
*rt
= (struct rtable
*)skb
->dst
;
1148 u32 saddr
= skb
->nh
.iph
->saddr
;
1149 u32 daddr
= skb
->nh
.iph
->daddr
;
1153 * Validate the packet and the UDP length.
1155 if (!pskb_may_pull(skb
, sizeof(struct udphdr
)))
1160 ulen
= ntohs(uh
->len
);
1162 if (ulen
> len
|| ulen
< sizeof(*uh
))
1165 if (pskb_trim(skb
, ulen
))
1168 if (udp_checksum_init(skb
, uh
, ulen
, saddr
, daddr
) < 0)
1171 if(rt
->rt_flags
& (RTCF_BROADCAST
|RTCF_MULTICAST
))
1172 return udp_v4_mcast_deliver(skb
, uh
, saddr
, daddr
);
1174 sk
= udp_v4_lookup(saddr
, uh
->source
, daddr
, uh
->dest
, skb
->dev
->ifindex
);
1177 int ret
= udp_queue_rcv_skb(sk
, skb
);
1180 /* a return value > 0 means to resubmit the input, but
1181 * it it wants the return to be -protocol, or 0
1188 if (!xfrm4_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
1191 /* No socket. Drop packet silently, if checksum is wrong */
1192 if (udp_checksum_complete(skb
))
1195 UDP_INC_STATS_BH(UdpNoPorts
);
1196 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_PORT_UNREACH
, 0);
1199 * Hmm. We got an UDP packet to a port to which we
1200 * don't wanna listen. Ignore it.
1206 NETDEBUG(if (net_ratelimit())
1207 printk(KERN_DEBUG
"UDP: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n",
1215 UDP_INC_STATS_BH(UdpInErrors
);
1221 * RFC1122: OK. Discards the bad packet silently (as far as
1222 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1224 NETDEBUG(if (net_ratelimit())
1225 printk(KERN_DEBUG
"UDP: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n",
1232 UDP_INC_STATS_BH(UdpInErrors
);
1237 static int udp_destroy_sock(struct sock
*sk
)
1240 udp_flush_pending_frames(sk
);
1246 * Socket option code for UDP
1248 static int udp_setsockopt(struct sock
*sk
, int level
, int optname
,
1249 char *optval
, int optlen
)
1251 struct udp_opt
*up
= udp_sk(sk
);
1255 if (level
!= SOL_UDP
)
1256 return ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
1258 if(optlen
<sizeof(int))
1261 if (get_user(val
, (int *)optval
))
1271 udp_push_pending_frames(sk
, up
);
1277 up
->encap_type
= val
;
1288 static int udp_getsockopt(struct sock
*sk
, int level
, int optname
,
1289 char *optval
, int *optlen
)
1291 struct udp_opt
*up
= udp_sk(sk
);
1294 if (level
!= SOL_UDP
)
1295 return ip_getsockopt(sk
, level
, optname
, optval
, optlen
);
1297 if(get_user(len
,optlen
))
1300 len
= min_t(unsigned int, len
, sizeof(int));
1311 val
= up
->encap_type
;
1315 return -ENOPROTOOPT
;
1318 if(put_user(len
, optlen
))
1320 if(copy_to_user(optval
, &val
,len
))
1326 struct proto udp_prot
= {
1329 .connect
= udp_connect
,
1330 .disconnect
= udp_disconnect
,
1332 .destroy
= udp_destroy_sock
,
1333 .setsockopt
= udp_setsockopt
,
1334 .getsockopt
= udp_getsockopt
,
1335 .sendmsg
= udp_sendmsg
,
1336 .recvmsg
= udp_recvmsg
,
1337 .sendpage
= udp_sendpage
,
1338 .backlog_rcv
= udp_queue_rcv_skb
,
1339 .hash
= udp_v4_hash
,
1340 .unhash
= udp_v4_unhash
,
1341 .get_port
= udp_v4_get_port
,
1344 /* ------------------------------------------------------------------------ */
1345 #ifdef CONFIG_PROC_FS
1347 static __inline__
struct sock
*udp_get_bucket(struct seq_file
*seq
, loff_t
*pos
)
1351 struct hlist_node
*node
;
1353 struct udp_iter_state
*state
= seq
->private;
1355 for (; state
->bucket
< UDP_HTABLE_SIZE
; ++state
->bucket
) {
1357 sk_for_each(sk
, node
, &udp_hash
[state
->bucket
]) {
1358 if (sk
->sk_family
!= state
->family
) {
1375 static void *udp_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1377 read_lock(&udp_hash_lock
);
1378 return *pos
? udp_get_bucket(seq
, pos
) : (void *)1;
1381 static void *udp_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1384 struct hlist_node
*node
;
1385 struct udp_iter_state
*state
;
1387 if (v
== (void *)1) {
1388 sk
= udp_get_bucket(seq
, pos
);
1392 state
= seq
->private;
1395 sk_for_each_continue(sk
, node
)
1396 if (sk
->sk_family
== state
->family
)
1399 if (++state
->bucket
>= UDP_HTABLE_SIZE
)
1403 sk
= udp_get_bucket(seq
, pos
);
1409 static void udp_seq_stop(struct seq_file
*seq
, void *v
)
1411 read_unlock(&udp_hash_lock
);
1414 static int udp_seq_open(struct inode
*inode
, struct file
*file
)
1416 struct udp_seq_afinfo
*afinfo
= PDE(inode
)->data
;
1417 struct seq_file
*seq
;
1419 struct udp_iter_state
*s
= kmalloc(sizeof(*s
), GFP_KERNEL
);
1423 memset(s
, 0, sizeof(*s
));
1424 s
->family
= afinfo
->family
;
1425 s
->seq_ops
.start
= udp_seq_start
;
1426 s
->seq_ops
.next
= udp_seq_next
;
1427 s
->seq_ops
.show
= afinfo
->seq_show
;
1428 s
->seq_ops
.stop
= udp_seq_stop
;
1430 rc
= seq_open(file
, &s
->seq_ops
);
1434 seq
= file
->private_data
;
1443 /* ------------------------------------------------------------------------ */
1444 int udp_proc_register(struct udp_seq_afinfo
*afinfo
)
1446 struct proc_dir_entry
*p
;
1451 afinfo
->seq_fops
->owner
= afinfo
->owner
;
1452 afinfo
->seq_fops
->open
= udp_seq_open
;
1453 afinfo
->seq_fops
->read
= seq_read
;
1454 afinfo
->seq_fops
->llseek
= seq_lseek
;
1455 afinfo
->seq_fops
->release
= seq_release_private
;
1457 p
= create_proc_entry(afinfo
->name
, S_IRUGO
, proc_net
);
1460 p
->proc_fops
= afinfo
->seq_fops
;
1466 void udp_proc_unregister(struct udp_seq_afinfo
*afinfo
)
1470 remove_proc_entry(afinfo
->name
, proc_net
);
1471 memset(afinfo
->seq_fops
, 0, sizeof(*afinfo
->seq_fops
));
1474 /* ------------------------------------------------------------------------ */
1475 static void udp4_format_sock(struct sock
*sp
, char *tmpbuf
, int bucket
)
1477 struct inet_opt
*inet
= inet_sk(sp
);
1478 unsigned int dest
= inet
->daddr
;
1479 unsigned int src
= inet
->rcv_saddr
;
1480 __u16 destp
= ntohs(inet
->dport
);
1481 __u16 srcp
= ntohs(inet
->sport
);
1483 sprintf(tmpbuf
, "%4d: %08X:%04X %08X:%04X"
1484 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
1485 bucket
, src
, srcp
, dest
, destp
, sp
->sk_state
,
1486 atomic_read(&sp
->sk_wmem_alloc
),
1487 atomic_read(&sp
->sk_rmem_alloc
),
1488 0, 0L, 0, sock_i_uid(sp
), 0, sock_i_ino(sp
),
1489 atomic_read(&sp
->sk_refcnt
), sp
);
1492 static int udp4_seq_show(struct seq_file
*seq
, void *v
)
1495 seq_printf(seq
, "%-127s\n",
1496 " sl local_address rem_address st tx_queue "
1497 "rx_queue tr tm->when retrnsmt uid timeout "
1501 struct udp_iter_state
*state
= seq
->private;
1503 udp4_format_sock(v
, tmpbuf
, state
->bucket
);
1504 seq_printf(seq
, "%-127s\n", tmpbuf
);
1509 /* ------------------------------------------------------------------------ */
1510 static struct file_operations udp4_seq_fops
;
1511 static struct udp_seq_afinfo udp4_seq_afinfo
= {
1512 .owner
= THIS_MODULE
,
1515 .seq_show
= udp4_seq_show
,
1516 .seq_fops
= &udp4_seq_fops
,
1519 int __init
udp4_proc_init(void)
1521 return udp_proc_register(&udp4_seq_afinfo
);
1524 void udp4_proc_exit(void)
1526 udp_proc_unregister(&udp4_seq_afinfo
);
1528 #endif /* CONFIG_PROC_FS */