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 $
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/module.h>
87 #include <linux/socket.h>
88 #include <linux/sockios.h>
90 #include <linux/errno.h>
91 #include <linux/timer.h>
93 #include <linux/config.h>
94 #include <linux/inet.h>
95 #include <linux/ipv6.h>
96 #include <linux/netdevice.h>
99 #include <net/protocol.h>
100 #include <linux/skbuff.h>
101 #include <linux/proc_fs.h>
102 #include <linux/seq_file.h>
103 #include <net/sock.h>
105 #include <net/icmp.h>
106 #include <net/route.h>
107 #include <net/inet_common.h>
108 #include <net/checksum.h>
109 #include <net/xfrm.h>
112 * Snmp MIB for the UDP layer
115 DEFINE_SNMP_STAT(struct udp_mib
, udp_statistics
);
117 struct hlist_head udp_hash
[UDP_HTABLE_SIZE
];
118 DEFINE_RWLOCK(udp_hash_lock
);
120 /* Shared by v4/v6 udp. */
123 static int udp_v4_get_port(struct sock
*sk
, unsigned short snum
)
125 struct hlist_node
*node
;
127 struct inet_sock
*inet
= inet_sk(sk
);
129 write_lock_bh(&udp_hash_lock
);
131 int best_size_so_far
, best
, result
, i
;
133 if (udp_port_rover
> sysctl_local_port_range
[1] ||
134 udp_port_rover
< sysctl_local_port_range
[0])
135 udp_port_rover
= sysctl_local_port_range
[0];
136 best_size_so_far
= 32767;
137 best
= result
= udp_port_rover
;
138 for (i
= 0; i
< UDP_HTABLE_SIZE
; i
++, result
++) {
139 struct hlist_head
*list
;
142 list
= &udp_hash
[result
& (UDP_HTABLE_SIZE
- 1)];
143 if (hlist_empty(list
)) {
144 if (result
> sysctl_local_port_range
[1])
145 result
= sysctl_local_port_range
[0] +
146 ((result
- sysctl_local_port_range
[0]) &
147 (UDP_HTABLE_SIZE
- 1));
151 sk_for_each(sk2
, node
, list
)
152 if (++size
>= best_size_so_far
)
154 best_size_so_far
= size
;
159 for(i
= 0; i
< (1 << 16) / UDP_HTABLE_SIZE
; i
++, result
+= UDP_HTABLE_SIZE
) {
160 if (result
> sysctl_local_port_range
[1])
161 result
= sysctl_local_port_range
[0]
162 + ((result
- sysctl_local_port_range
[0]) &
163 (UDP_HTABLE_SIZE
- 1));
164 if (!udp_lport_inuse(result
))
167 if (i
>= (1 << 16) / UDP_HTABLE_SIZE
)
170 udp_port_rover
= snum
= result
;
172 sk_for_each(sk2
, node
,
173 &udp_hash
[snum
& (UDP_HTABLE_SIZE
- 1)]) {
174 struct inet_sock
*inet2
= inet_sk(sk2
);
176 if (inet2
->num
== snum
&&
178 !ipv6_only_sock(sk2
) &&
179 (!sk2
->sk_bound_dev_if
||
180 !sk
->sk_bound_dev_if
||
181 sk2
->sk_bound_dev_if
== sk
->sk_bound_dev_if
) &&
182 (!inet2
->rcv_saddr
||
184 inet2
->rcv_saddr
== inet
->rcv_saddr
) &&
185 (!sk2
->sk_reuse
|| !sk
->sk_reuse
))
190 if (sk_unhashed(sk
)) {
191 struct hlist_head
*h
= &udp_hash
[snum
& (UDP_HTABLE_SIZE
- 1)];
194 sock_prot_inc_use(sk
->sk_prot
);
196 write_unlock_bh(&udp_hash_lock
);
200 write_unlock_bh(&udp_hash_lock
);
204 static void udp_v4_hash(struct sock
*sk
)
209 static void udp_v4_unhash(struct sock
*sk
)
211 write_lock_bh(&udp_hash_lock
);
212 if (sk_del_node_init(sk
)) {
213 inet_sk(sk
)->num
= 0;
214 sock_prot_dec_use(sk
->sk_prot
);
216 write_unlock_bh(&udp_hash_lock
);
219 /* UDP is nearly always wildcards out the wazoo, it makes no sense to try
220 * harder than this. -DaveM
222 static struct sock
*udp_v4_lookup_longway(u32 saddr
, u16 sport
,
223 u32 daddr
, u16 dport
, int dif
)
225 struct sock
*sk
, *result
= NULL
;
226 struct hlist_node
*node
;
227 unsigned short hnum
= ntohs(dport
);
230 sk_for_each(sk
, node
, &udp_hash
[hnum
& (UDP_HTABLE_SIZE
- 1)]) {
231 struct inet_sock
*inet
= inet_sk(sk
);
233 if (inet
->num
== hnum
&& !ipv6_only_sock(sk
)) {
234 int score
= (sk
->sk_family
== PF_INET
? 1 : 0);
235 if (inet
->rcv_saddr
) {
236 if (inet
->rcv_saddr
!= daddr
)
241 if (inet
->daddr
!= saddr
)
246 if (inet
->dport
!= sport
)
250 if (sk
->sk_bound_dev_if
) {
251 if (sk
->sk_bound_dev_if
!= dif
)
258 } else if(score
> badness
) {
267 static __inline__
struct sock
*udp_v4_lookup(u32 saddr
, u16 sport
,
268 u32 daddr
, u16 dport
, int dif
)
272 read_lock(&udp_hash_lock
);
273 sk
= udp_v4_lookup_longway(saddr
, sport
, daddr
, dport
, dif
);
276 read_unlock(&udp_hash_lock
);
280 static inline struct sock
*udp_v4_mcast_next(struct sock
*sk
,
281 u16 loc_port
, u32 loc_addr
,
282 u16 rmt_port
, u32 rmt_addr
,
285 struct hlist_node
*node
;
287 unsigned short hnum
= ntohs(loc_port
);
289 sk_for_each_from(s
, node
) {
290 struct inet_sock
*inet
= inet_sk(s
);
292 if (inet
->num
!= hnum
||
293 (inet
->daddr
&& inet
->daddr
!= rmt_addr
) ||
294 (inet
->dport
!= rmt_port
&& inet
->dport
) ||
295 (inet
->rcv_saddr
&& inet
->rcv_saddr
!= loc_addr
) ||
297 (s
->sk_bound_dev_if
&& s
->sk_bound_dev_if
!= dif
))
299 if (!ip_mc_sf_allow(s
, loc_addr
, rmt_addr
, dif
))
309 * This routine is called by the ICMP module when it gets some
310 * sort of error condition. If err < 0 then the socket should
311 * be closed and the error returned to the user. If err > 0
312 * it's just the icmp type << 8 | icmp code.
313 * Header points to the ip header of the error packet. We move
314 * on past this. Then (as it used to claim before adjustment)
315 * header points to the first 8 bytes of the udp header. We need
316 * to find the appropriate port.
319 void udp_err(struct sk_buff
*skb
, u32 info
)
321 struct inet_sock
*inet
;
322 struct iphdr
*iph
= (struct iphdr
*)skb
->data
;
323 struct udphdr
*uh
= (struct udphdr
*)(skb
->data
+(iph
->ihl
<<2));
324 int type
= skb
->h
.icmph
->type
;
325 int code
= skb
->h
.icmph
->code
;
330 sk
= udp_v4_lookup(iph
->daddr
, uh
->dest
, iph
->saddr
, uh
->source
, skb
->dev
->ifindex
);
332 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS
);
333 return; /* No socket for error */
342 case ICMP_TIME_EXCEEDED
:
345 case ICMP_SOURCE_QUENCH
:
347 case ICMP_PARAMETERPROB
:
351 case ICMP_DEST_UNREACH
:
352 if (code
== ICMP_FRAG_NEEDED
) { /* Path MTU discovery */
353 if (inet
->pmtudisc
!= IP_PMTUDISC_DONT
) {
361 if (code
<= NR_ICMP_UNREACH
) {
362 harderr
= icmp_err_convert
[code
].fatal
;
363 err
= icmp_err_convert
[code
].errno
;
369 * RFC1122: OK. Passes ICMP errors back to application, as per
372 if (!inet
->recverr
) {
373 if (!harderr
|| sk
->sk_state
!= TCP_ESTABLISHED
)
376 ip_icmp_error(sk
, skb
, err
, uh
->dest
, info
, (u8
*)(uh
+1));
379 sk
->sk_error_report(sk
);
385 * Throw away all pending data and cancel the corking. Socket is locked.
387 static void udp_flush_pending_frames(struct sock
*sk
)
389 struct udp_sock
*up
= udp_sk(sk
);
394 ip_flush_pending_frames(sk
);
399 * Push out all pending data as one UDP datagram. Socket is locked.
401 static int udp_push_pending_frames(struct sock
*sk
, struct udp_sock
*up
)
403 struct inet_sock
*inet
= inet_sk(sk
);
404 struct flowi
*fl
= &inet
->cork
.fl
;
409 /* Grab the skbuff where UDP header space exists. */
410 if ((skb
= skb_peek(&sk
->sk_write_queue
)) == NULL
)
414 * Create a UDP header
417 uh
->source
= fl
->fl_ip_sport
;
418 uh
->dest
= fl
->fl_ip_dport
;
419 uh
->len
= htons(up
->len
);
422 if (sk
->sk_no_check
== UDP_CSUM_NOXMIT
) {
423 skb
->ip_summed
= CHECKSUM_NONE
;
427 if (skb_queue_len(&sk
->sk_write_queue
) == 1) {
429 * Only one fragment on the socket.
431 if (skb
->ip_summed
== CHECKSUM_HW
) {
432 skb
->csum
= offsetof(struct udphdr
, check
);
433 uh
->check
= ~csum_tcpudp_magic(fl
->fl4_src
, fl
->fl4_dst
,
434 up
->len
, IPPROTO_UDP
, 0);
436 skb
->csum
= csum_partial((char *)uh
,
437 sizeof(struct udphdr
), skb
->csum
);
438 uh
->check
= csum_tcpudp_magic(fl
->fl4_src
, fl
->fl4_dst
,
439 up
->len
, IPPROTO_UDP
, skb
->csum
);
444 unsigned int csum
= 0;
446 * HW-checksum won't work as there are two or more
447 * fragments on the socket so that all csums of sk_buffs
448 * should be together.
450 if (skb
->ip_summed
== CHECKSUM_HW
) {
451 int offset
= (unsigned char *)uh
- skb
->data
;
452 skb
->csum
= skb_checksum(skb
, offset
, skb
->len
- offset
, 0);
454 skb
->ip_summed
= CHECKSUM_NONE
;
456 skb
->csum
= csum_partial((char *)uh
,
457 sizeof(struct udphdr
), skb
->csum
);
460 skb_queue_walk(&sk
->sk_write_queue
, skb
) {
461 csum
= csum_add(csum
, skb
->csum
);
463 uh
->check
= csum_tcpudp_magic(fl
->fl4_src
, fl
->fl4_dst
,
464 up
->len
, IPPROTO_UDP
, csum
);
469 err
= ip_push_pending_frames(sk
);
477 static unsigned short udp_check(struct udphdr
*uh
, int len
, unsigned long saddr
, unsigned long daddr
, unsigned long base
)
479 return(csum_tcpudp_magic(saddr
, daddr
, len
, IPPROTO_UDP
, base
));
482 int udp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
485 struct inet_sock
*inet
= inet_sk(sk
);
486 struct udp_sock
*up
= udp_sk(sk
);
488 struct ipcm_cookie ipc
;
489 struct rtable
*rt
= NULL
;
492 u32 daddr
, faddr
, saddr
;
496 int corkreq
= up
->corkflag
|| msg
->msg_flags
&MSG_MORE
;
505 if (msg
->msg_flags
&MSG_OOB
) /* Mirror BSD error message compatibility */
512 * There are pending frames.
513 * The socket lock must be held while it's corked.
516 if (likely(up
->pending
)) {
517 if (unlikely(up
->pending
!= AF_INET
)) {
525 ulen
+= sizeof(struct udphdr
);
528 * Get and verify the address.
531 struct sockaddr_in
* usin
= (struct sockaddr_in
*)msg
->msg_name
;
532 if (msg
->msg_namelen
< sizeof(*usin
))
534 if (usin
->sin_family
!= AF_INET
) {
535 if (usin
->sin_family
!= AF_UNSPEC
)
536 return -EAFNOSUPPORT
;
539 daddr
= usin
->sin_addr
.s_addr
;
540 dport
= usin
->sin_port
;
544 if (sk
->sk_state
!= TCP_ESTABLISHED
)
545 return -EDESTADDRREQ
;
548 /* Open fast path for connected socket.
549 Route will not be used, if at least one option is set.
553 ipc
.addr
= inet
->saddr
;
555 ipc
.oif
= sk
->sk_bound_dev_if
;
556 if (msg
->msg_controllen
) {
557 err
= ip_cmsg_send(msg
, &ipc
);
568 ipc
.addr
= faddr
= daddr
;
570 if (ipc
.opt
&& ipc
.opt
->srr
) {
573 faddr
= ipc
.opt
->faddr
;
576 tos
= RT_TOS(inet
->tos
);
577 if (sock_flag(sk
, SOCK_LOCALROUTE
) ||
578 (msg
->msg_flags
& MSG_DONTROUTE
) ||
579 (ipc
.opt
&& ipc
.opt
->is_strictroute
)) {
584 if (MULTICAST(daddr
)) {
586 ipc
.oif
= inet
->mc_index
;
588 saddr
= inet
->mc_addr
;
593 rt
= (struct rtable
*)sk_dst_check(sk
, 0);
596 struct flowi fl
= { .oif
= ipc
.oif
,
601 .proto
= IPPROTO_UDP
,
603 { .sport
= inet
->sport
,
604 .dport
= dport
} } };
605 err
= ip_route_output_flow(&rt
, &fl
, sk
, !(msg
->msg_flags
&MSG_DONTWAIT
));
610 if ((rt
->rt_flags
& RTCF_BROADCAST
) &&
611 !sock_flag(sk
, SOCK_BROADCAST
))
614 sk_dst_set(sk
, dst_clone(&rt
->u
.dst
));
617 if (msg
->msg_flags
&MSG_CONFIRM
)
623 daddr
= ipc
.addr
= rt
->rt_dst
;
626 if (unlikely(up
->pending
)) {
627 /* The socket is already corked while preparing it. */
628 /* ... which is an evident application bug. --ANK */
631 NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG
"udp cork app bug 2\n"));
636 * Now cork the socket to pend data.
638 inet
->cork
.fl
.fl4_dst
= daddr
;
639 inet
->cork
.fl
.fl_ip_dport
= dport
;
640 inet
->cork
.fl
.fl4_src
= saddr
;
641 inet
->cork
.fl
.fl_ip_sport
= inet
->sport
;
642 up
->pending
= AF_INET
;
646 err
= ip_append_data(sk
, ip_generic_getfrag
, msg
->msg_iov
, ulen
,
647 sizeof(struct udphdr
), &ipc
, rt
,
648 corkreq
? msg
->msg_flags
|MSG_MORE
: msg
->msg_flags
);
650 udp_flush_pending_frames(sk
);
652 err
= udp_push_pending_frames(sk
, up
);
660 UDP_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS
);
666 dst_confirm(&rt
->u
.dst
);
667 if (!(msg
->msg_flags
&MSG_PROBE
) || len
)
668 goto back_from_confirm
;
673 static int udp_sendpage(struct sock
*sk
, struct page
*page
, int offset
,
674 size_t size
, int flags
)
676 struct udp_sock
*up
= udp_sk(sk
);
680 struct msghdr msg
= { .msg_flags
= flags
|MSG_MORE
};
682 /* Call udp_sendmsg to specify destination address which
683 * sendpage interface can't pass.
684 * This will succeed only when the socket is connected.
686 ret
= udp_sendmsg(NULL
, sk
, &msg
, 0);
693 if (unlikely(!up
->pending
)) {
696 NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG
"udp cork app bug 3\n"));
700 ret
= ip_append_page(sk
, page
, offset
, size
, flags
);
701 if (ret
== -EOPNOTSUPP
) {
703 return sock_no_sendpage(sk
->sk_socket
, page
, offset
,
707 udp_flush_pending_frames(sk
);
712 if (!(up
->corkflag
|| (flags
&MSG_MORE
)))
713 ret
= udp_push_pending_frames(sk
, up
);
722 * IOCTL requests applicable to the UDP protocol
725 int udp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
731 int amount
= atomic_read(&sk
->sk_wmem_alloc
);
732 return put_user(amount
, (int __user
*)arg
);
738 unsigned long amount
;
741 spin_lock_irq(&sk
->sk_receive_queue
.lock
);
742 skb
= skb_peek(&sk
->sk_receive_queue
);
745 * We will only return the amount
746 * of this packet since that is all
749 amount
= skb
->len
- sizeof(struct udphdr
);
751 spin_unlock_irq(&sk
->sk_receive_queue
.lock
);
752 return put_user(amount
, (int __user
*)arg
);
761 static __inline__
int __udp_checksum_complete(struct sk_buff
*skb
)
763 return (unsigned short)csum_fold(skb_checksum(skb
, 0, skb
->len
, skb
->csum
));
766 static __inline__
int udp_checksum_complete(struct sk_buff
*skb
)
768 return skb
->ip_summed
!= CHECKSUM_UNNECESSARY
&&
769 __udp_checksum_complete(skb
);
773 * This should be easy, if there is something there we
774 * return it, otherwise we block.
777 static int udp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
778 size_t len
, int noblock
, int flags
, int *addr_len
)
780 struct inet_sock
*inet
= inet_sk(sk
);
781 struct sockaddr_in
*sin
= (struct sockaddr_in
*)msg
->msg_name
;
786 * Check any passed addresses
789 *addr_len
=sizeof(*sin
);
791 if (flags
& MSG_ERRQUEUE
)
792 return ip_recv_error(sk
, msg
, len
);
795 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
799 copied
= skb
->len
- sizeof(struct udphdr
);
802 msg
->msg_flags
|= MSG_TRUNC
;
805 if (skb
->ip_summed
==CHECKSUM_UNNECESSARY
) {
806 err
= skb_copy_datagram_iovec(skb
, sizeof(struct udphdr
), msg
->msg_iov
,
808 } else if (msg
->msg_flags
&MSG_TRUNC
) {
809 if (__udp_checksum_complete(skb
))
811 err
= skb_copy_datagram_iovec(skb
, sizeof(struct udphdr
), msg
->msg_iov
,
814 err
= skb_copy_and_csum_datagram_iovec(skb
, sizeof(struct udphdr
), msg
->msg_iov
);
823 sock_recv_timestamp(msg
, sk
, skb
);
825 /* Copy the address. */
828 sin
->sin_family
= AF_INET
;
829 sin
->sin_port
= skb
->h
.uh
->source
;
830 sin
->sin_addr
.s_addr
= skb
->nh
.iph
->saddr
;
831 memset(sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
833 if (inet
->cmsg_flags
)
834 ip_cmsg_recv(msg
, skb
);
837 if (flags
& MSG_TRUNC
)
838 err
= skb
->len
- sizeof(struct udphdr
);
841 skb_free_datagram(sk
, skb
);
846 UDP_INC_STATS_BH(UDP_MIB_INERRORS
);
849 if (flags
&MSG_PEEK
) {
851 spin_lock_irq(&sk
->sk_receive_queue
.lock
);
852 if (skb
== skb_peek(&sk
->sk_receive_queue
)) {
853 __skb_unlink(skb
, &sk
->sk_receive_queue
);
856 spin_unlock_irq(&sk
->sk_receive_queue
.lock
);
861 skb_free_datagram(sk
, skb
);
869 int udp_disconnect(struct sock
*sk
, int flags
)
871 struct inet_sock
*inet
= inet_sk(sk
);
873 * 1003.1g - break association.
876 sk
->sk_state
= TCP_CLOSE
;
879 sk
->sk_bound_dev_if
= 0;
880 if (!(sk
->sk_userlocks
& SOCK_BINDADDR_LOCK
))
881 inet_reset_saddr(sk
);
883 if (!(sk
->sk_userlocks
& SOCK_BINDPORT_LOCK
)) {
884 sk
->sk_prot
->unhash(sk
);
891 static void udp_close(struct sock
*sk
, long timeout
)
893 sk_common_release(sk
);
897 * 1 if the the UDP system should process it
898 * 0 if we should drop this packet
899 * -1 if it should get processed by xfrm4_rcv_encap
901 static int udp_encap_rcv(struct sock
* sk
, struct sk_buff
*skb
)
906 struct udp_sock
*up
= udp_sk(sk
);
907 struct udphdr
*uh
= skb
->h
.uh
;
911 __u8
*udpdata
= (__u8
*)uh
+ sizeof(struct udphdr
);
912 __u32
*udpdata32
= (__u32
*)udpdata
;
913 __u16 encap_type
= up
->encap_type
;
915 /* if we're overly short, let UDP handle it */
916 if (udpdata
> skb
->tail
)
919 /* if this is not encapsulated socket, then just return now */
923 len
= skb
->tail
- udpdata
;
925 switch (encap_type
) {
927 case UDP_ENCAP_ESPINUDP
:
928 /* Check if this is a keepalive packet. If so, eat it. */
929 if (len
== 1 && udpdata
[0] == 0xff) {
931 } else if (len
> sizeof(struct ip_esp_hdr
) && udpdata32
[0] != 0 ) {
932 /* ESP Packet without Non-ESP header */
933 len
= sizeof(struct udphdr
);
935 /* Must be an IKE packet.. pass it through */
938 case UDP_ENCAP_ESPINUDP_NON_IKE
:
939 /* Check if this is a keepalive packet. If so, eat it. */
940 if (len
== 1 && udpdata
[0] == 0xff) {
942 } else if (len
> 2 * sizeof(u32
) + sizeof(struct ip_esp_hdr
) &&
943 udpdata32
[0] == 0 && udpdata32
[1] == 0) {
945 /* ESP Packet with Non-IKE marker */
946 len
= sizeof(struct udphdr
) + 2 * sizeof(u32
);
948 /* Must be an IKE packet.. pass it through */
953 /* At this point we are sure that this is an ESPinUDP packet,
954 * so we need to remove 'len' bytes from the packet (the UDP
955 * header and optional ESP marker bytes) and then modify the
956 * protocol to ESP, and then call into the transform receiver.
958 if (skb_cloned(skb
) && pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
))
961 /* Now we can update and verify the packet length... */
963 iphlen
= iph
->ihl
<< 2;
964 iph
->tot_len
= htons(ntohs(iph
->tot_len
) - len
);
965 if (skb
->len
< iphlen
+ len
) {
966 /* packet is too small!?! */
970 /* pull the data buffer up to the ESP header and set the
971 * transport header to point to ESP. Keep UDP on the stack
974 skb
->h
.raw
= skb_pull(skb
, len
);
976 /* modify the protocol (it's ESP!) */
977 iph
->protocol
= IPPROTO_ESP
;
979 /* and let the caller know to send this into the ESP processor... */
987 * >0: "udp encap" protocol resubmission
989 * Note that in the success and error cases, the skb is assumed to
990 * have either been requeued or freed.
992 static int udp_queue_rcv_skb(struct sock
* sk
, struct sk_buff
*skb
)
994 struct udp_sock
*up
= udp_sk(sk
);
997 * Charge it to the socket, dropping if the queue is full.
999 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
)) {
1004 if (up
->encap_type
) {
1006 * This is an encapsulation socket, so let's see if this is
1007 * an encapsulated packet.
1008 * If it's a keepalive packet, then just eat it.
1009 * If it's an encapsulateed packet, then pass it to the
1010 * IPsec xfrm input and return the response
1011 * appropriately. Otherwise, just fall through and
1012 * pass this up the UDP socket.
1016 ret
= udp_encap_rcv(sk
, skb
);
1018 /* Eat the packet .. */
1023 /* process the ESP packet */
1024 ret
= xfrm4_rcv_encap(skb
, up
->encap_type
);
1025 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS
);
1028 /* FALLTHROUGH -- it's a UDP Packet */
1031 if (sk
->sk_filter
&& skb
->ip_summed
!= CHECKSUM_UNNECESSARY
) {
1032 if (__udp_checksum_complete(skb
)) {
1033 UDP_INC_STATS_BH(UDP_MIB_INERRORS
);
1037 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1040 if (sock_queue_rcv_skb(sk
,skb
)<0) {
1041 UDP_INC_STATS_BH(UDP_MIB_INERRORS
);
1045 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS
);
1050 * Multicasts and broadcasts go to each listener.
1052 * Note: called only from the BH handler context,
1053 * so we don't need to lock the hashes.
1055 static int udp_v4_mcast_deliver(struct sk_buff
*skb
, struct udphdr
*uh
,
1056 u32 saddr
, u32 daddr
)
1061 read_lock(&udp_hash_lock
);
1062 sk
= sk_head(&udp_hash
[ntohs(uh
->dest
) & (UDP_HTABLE_SIZE
- 1)]);
1063 dif
= skb
->dev
->ifindex
;
1064 sk
= udp_v4_mcast_next(sk
, uh
->dest
, daddr
, uh
->source
, saddr
, dif
);
1066 struct sock
*sknext
= NULL
;
1069 struct sk_buff
*skb1
= skb
;
1071 sknext
= udp_v4_mcast_next(sk_next(sk
), uh
->dest
, daddr
,
1072 uh
->source
, saddr
, dif
);
1074 skb1
= skb_clone(skb
, GFP_ATOMIC
);
1077 int ret
= udp_queue_rcv_skb(sk
, skb1
);
1079 /* we should probably re-process instead
1080 * of dropping packets here. */
1087 read_unlock(&udp_hash_lock
);
1091 /* Initialize UDP checksum. If exited with zero value (success),
1092 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1093 * Otherwise, csum completion requires chacksumming packet body,
1094 * including udp header and folding it to skb->csum.
1096 static int udp_checksum_init(struct sk_buff
*skb
, struct udphdr
*uh
,
1097 unsigned short ulen
, u32 saddr
, u32 daddr
)
1099 if (uh
->check
== 0) {
1100 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1101 } else if (skb
->ip_summed
== CHECKSUM_HW
) {
1102 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1103 if (!udp_check(uh
, ulen
, saddr
, daddr
, skb
->csum
))
1105 NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG
"udp v4 hw csum failure.\n"));
1106 skb
->ip_summed
= CHECKSUM_NONE
;
1108 if (skb
->ip_summed
!= CHECKSUM_UNNECESSARY
)
1109 skb
->csum
= csum_tcpudp_nofold(saddr
, daddr
, ulen
, IPPROTO_UDP
, 0);
1110 /* Probably, we should checksum udp header (it should be in cache
1111 * in any case) and data in tiny packets (< rx copybreak).
1117 * All we need to do is get the socket, and then do a checksum.
1120 int udp_rcv(struct sk_buff
*skb
)
1124 unsigned short ulen
;
1125 struct rtable
*rt
= (struct rtable
*)skb
->dst
;
1126 u32 saddr
= skb
->nh
.iph
->saddr
;
1127 u32 daddr
= skb
->nh
.iph
->daddr
;
1131 * Validate the packet and the UDP length.
1133 if (!pskb_may_pull(skb
, sizeof(struct udphdr
)))
1138 ulen
= ntohs(uh
->len
);
1140 if (ulen
> len
|| ulen
< sizeof(*uh
))
1143 if (pskb_trim(skb
, ulen
))
1146 if (udp_checksum_init(skb
, uh
, ulen
, saddr
, daddr
) < 0)
1149 if(rt
->rt_flags
& (RTCF_BROADCAST
|RTCF_MULTICAST
))
1150 return udp_v4_mcast_deliver(skb
, uh
, saddr
, daddr
);
1152 sk
= udp_v4_lookup(saddr
, uh
->source
, daddr
, uh
->dest
, skb
->dev
->ifindex
);
1155 int ret
= udp_queue_rcv_skb(sk
, skb
);
1158 /* a return value > 0 means to resubmit the input, but
1159 * it it wants the return to be -protocol, or 0
1166 if (!xfrm4_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
1169 /* No socket. Drop packet silently, if checksum is wrong */
1170 if (udp_checksum_complete(skb
))
1173 UDP_INC_STATS_BH(UDP_MIB_NOPORTS
);
1174 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_PORT_UNREACH
, 0);
1177 * Hmm. We got an UDP packet to a port to which we
1178 * don't wanna listen. Ignore it.
1184 NETDEBUG(if (net_ratelimit())
1185 printk(KERN_DEBUG
"UDP: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n",
1193 UDP_INC_STATS_BH(UDP_MIB_INERRORS
);
1199 * RFC1122: OK. Discards the bad packet silently (as far as
1200 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1202 NETDEBUG(if (net_ratelimit())
1203 printk(KERN_DEBUG
"UDP: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n",
1210 UDP_INC_STATS_BH(UDP_MIB_INERRORS
);
1215 static int udp_destroy_sock(struct sock
*sk
)
1218 udp_flush_pending_frames(sk
);
1224 * Socket option code for UDP
1226 static int udp_setsockopt(struct sock
*sk
, int level
, int optname
,
1227 char __user
*optval
, int optlen
)
1229 struct udp_sock
*up
= udp_sk(sk
);
1233 if (level
!= SOL_UDP
)
1234 return ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
1236 if(optlen
<sizeof(int))
1239 if (get_user(val
, (int __user
*)optval
))
1249 udp_push_pending_frames(sk
, up
);
1257 case UDP_ENCAP_ESPINUDP
:
1258 case UDP_ENCAP_ESPINUDP_NON_IKE
:
1259 up
->encap_type
= val
;
1275 static int udp_getsockopt(struct sock
*sk
, int level
, int optname
,
1276 char __user
*optval
, int __user
*optlen
)
1278 struct udp_sock
*up
= udp_sk(sk
);
1281 if (level
!= SOL_UDP
)
1282 return ip_getsockopt(sk
, level
, optname
, optval
, optlen
);
1284 if(get_user(len
,optlen
))
1287 len
= min_t(unsigned int, len
, sizeof(int));
1298 val
= up
->encap_type
;
1302 return -ENOPROTOOPT
;
1305 if(put_user(len
, optlen
))
1307 if(copy_to_user(optval
, &val
,len
))
1313 * udp_poll - wait for a UDP event.
1314 * @file - file struct
1316 * @wait - poll table
1318 * This is same as datagram poll, except for the special case of
1319 * blocking sockets. If application is using a blocking fd
1320 * and a packet with checksum error is in the queue;
1321 * then it could get return from select indicating data available
1322 * but then block when reading it. Add special case code
1323 * to work around these arguably broken applications.
1325 unsigned int udp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
1327 unsigned int mask
= datagram_poll(file
, sock
, wait
);
1328 struct sock
*sk
= sock
->sk
;
1330 /* Check for false positives due to checksum errors */
1331 if ( (mask
& POLLRDNORM
) &&
1332 !(file
->f_flags
& O_NONBLOCK
) &&
1333 !(sk
->sk_shutdown
& RCV_SHUTDOWN
)){
1334 struct sk_buff_head
*rcvq
= &sk
->sk_receive_queue
;
1335 struct sk_buff
*skb
;
1337 spin_lock_irq(&rcvq
->lock
);
1338 while ((skb
= skb_peek(rcvq
)) != NULL
) {
1339 if (udp_checksum_complete(skb
)) {
1340 UDP_INC_STATS_BH(UDP_MIB_INERRORS
);
1341 __skb_unlink(skb
, rcvq
);
1344 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1348 spin_unlock_irq(&rcvq
->lock
);
1350 /* nothing to see, move along */
1352 mask
&= ~(POLLIN
| POLLRDNORM
);
1359 struct proto udp_prot
= {
1361 .owner
= THIS_MODULE
,
1363 .connect
= ip4_datagram_connect
,
1364 .disconnect
= udp_disconnect
,
1366 .destroy
= udp_destroy_sock
,
1367 .setsockopt
= udp_setsockopt
,
1368 .getsockopt
= udp_getsockopt
,
1369 .sendmsg
= udp_sendmsg
,
1370 .recvmsg
= udp_recvmsg
,
1371 .sendpage
= udp_sendpage
,
1372 .backlog_rcv
= udp_queue_rcv_skb
,
1373 .hash
= udp_v4_hash
,
1374 .unhash
= udp_v4_unhash
,
1375 .get_port
= udp_v4_get_port
,
1376 .obj_size
= sizeof(struct udp_sock
),
1379 /* ------------------------------------------------------------------------ */
1380 #ifdef CONFIG_PROC_FS
1382 static struct sock
*udp_get_first(struct seq_file
*seq
)
1385 struct udp_iter_state
*state
= seq
->private;
1387 for (state
->bucket
= 0; state
->bucket
< UDP_HTABLE_SIZE
; ++state
->bucket
) {
1388 struct hlist_node
*node
;
1389 sk_for_each(sk
, node
, &udp_hash
[state
->bucket
]) {
1390 if (sk
->sk_family
== state
->family
)
1399 static struct sock
*udp_get_next(struct seq_file
*seq
, struct sock
*sk
)
1401 struct udp_iter_state
*state
= seq
->private;
1407 } while (sk
&& sk
->sk_family
!= state
->family
);
1409 if (!sk
&& ++state
->bucket
< UDP_HTABLE_SIZE
) {
1410 sk
= sk_head(&udp_hash
[state
->bucket
]);
1416 static struct sock
*udp_get_idx(struct seq_file
*seq
, loff_t pos
)
1418 struct sock
*sk
= udp_get_first(seq
);
1421 while(pos
&& (sk
= udp_get_next(seq
, sk
)) != NULL
)
1423 return pos
? NULL
: sk
;
1426 static void *udp_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1428 read_lock(&udp_hash_lock
);
1429 return *pos
? udp_get_idx(seq
, *pos
-1) : (void *)1;
1432 static void *udp_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1437 sk
= udp_get_idx(seq
, 0);
1439 sk
= udp_get_next(seq
, v
);
1445 static void udp_seq_stop(struct seq_file
*seq
, void *v
)
1447 read_unlock(&udp_hash_lock
);
1450 static int udp_seq_open(struct inode
*inode
, struct file
*file
)
1452 struct udp_seq_afinfo
*afinfo
= PDE(inode
)->data
;
1453 struct seq_file
*seq
;
1455 struct udp_iter_state
*s
= kmalloc(sizeof(*s
), GFP_KERNEL
);
1459 memset(s
, 0, sizeof(*s
));
1460 s
->family
= afinfo
->family
;
1461 s
->seq_ops
.start
= udp_seq_start
;
1462 s
->seq_ops
.next
= udp_seq_next
;
1463 s
->seq_ops
.show
= afinfo
->seq_show
;
1464 s
->seq_ops
.stop
= udp_seq_stop
;
1466 rc
= seq_open(file
, &s
->seq_ops
);
1470 seq
= file
->private_data
;
1479 /* ------------------------------------------------------------------------ */
1480 int udp_proc_register(struct udp_seq_afinfo
*afinfo
)
1482 struct proc_dir_entry
*p
;
1487 afinfo
->seq_fops
->owner
= afinfo
->owner
;
1488 afinfo
->seq_fops
->open
= udp_seq_open
;
1489 afinfo
->seq_fops
->read
= seq_read
;
1490 afinfo
->seq_fops
->llseek
= seq_lseek
;
1491 afinfo
->seq_fops
->release
= seq_release_private
;
1493 p
= proc_net_fops_create(afinfo
->name
, S_IRUGO
, afinfo
->seq_fops
);
1501 void udp_proc_unregister(struct udp_seq_afinfo
*afinfo
)
1505 proc_net_remove(afinfo
->name
);
1506 memset(afinfo
->seq_fops
, 0, sizeof(*afinfo
->seq_fops
));
1509 /* ------------------------------------------------------------------------ */
1510 static void udp4_format_sock(struct sock
*sp
, char *tmpbuf
, int bucket
)
1512 struct inet_sock
*inet
= inet_sk(sp
);
1513 unsigned int dest
= inet
->daddr
;
1514 unsigned int src
= inet
->rcv_saddr
;
1515 __u16 destp
= ntohs(inet
->dport
);
1516 __u16 srcp
= ntohs(inet
->sport
);
1518 sprintf(tmpbuf
, "%4d: %08X:%04X %08X:%04X"
1519 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
1520 bucket
, src
, srcp
, dest
, destp
, sp
->sk_state
,
1521 atomic_read(&sp
->sk_wmem_alloc
),
1522 atomic_read(&sp
->sk_rmem_alloc
),
1523 0, 0L, 0, sock_i_uid(sp
), 0, sock_i_ino(sp
),
1524 atomic_read(&sp
->sk_refcnt
), sp
);
1527 static int udp4_seq_show(struct seq_file
*seq
, void *v
)
1529 if (v
== SEQ_START_TOKEN
)
1530 seq_printf(seq
, "%-127s\n",
1531 " sl local_address rem_address st tx_queue "
1532 "rx_queue tr tm->when retrnsmt uid timeout "
1536 struct udp_iter_state
*state
= seq
->private;
1538 udp4_format_sock(v
, tmpbuf
, state
->bucket
);
1539 seq_printf(seq
, "%-127s\n", tmpbuf
);
1544 /* ------------------------------------------------------------------------ */
1545 static struct file_operations udp4_seq_fops
;
1546 static struct udp_seq_afinfo udp4_seq_afinfo
= {
1547 .owner
= THIS_MODULE
,
1550 .seq_show
= udp4_seq_show
,
1551 .seq_fops
= &udp4_seq_fops
,
1554 int __init
udp4_proc_init(void)
1556 return udp_proc_register(&udp4_seq_afinfo
);
1559 void udp4_proc_exit(void)
1561 udp_proc_unregister(&udp4_seq_afinfo
);
1563 #endif /* CONFIG_PROC_FS */
1565 EXPORT_SYMBOL(udp_disconnect
);
1566 EXPORT_SYMBOL(udp_hash
);
1567 EXPORT_SYMBOL(udp_hash_lock
);
1568 EXPORT_SYMBOL(udp_ioctl
);
1569 EXPORT_SYMBOL(udp_port_rover
);
1570 EXPORT_SYMBOL(udp_prot
);
1571 EXPORT_SYMBOL(udp_sendmsg
);
1572 EXPORT_SYMBOL(udp_poll
);
1574 #ifdef CONFIG_PROC_FS
1575 EXPORT_SYMBOL(udp_proc_register
);
1576 EXPORT_SYMBOL(udp_proc_unregister
);