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/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 rwlock_t udp_hash_lock = RW_LOCK_UNLOCKED;
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_opt *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_opt *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 struct sock *udp_v4_lookup_longway(u32 saddr, u16 sport, u32 daddr, u16 dport, int dif)
224 struct sock *sk, *result = NULL;
225 struct hlist_node *node;
226 unsigned short hnum = ntohs(dport);
229 sk_for_each(sk, node, &udp_hash[hnum & (UDP_HTABLE_SIZE - 1)]) {
230 struct inet_opt *inet = inet_sk(sk);
232 if (inet->num == hnum && !ipv6_only_sock(sk)) {
233 int score = (sk->sk_family == PF_INET ? 1 : 0);
234 if (inet->rcv_saddr) {
235 if (inet->rcv_saddr != daddr)
240 if (inet->daddr != saddr)
245 if (inet->dport != sport)
249 if (sk->sk_bound_dev_if) {
250 if (sk->sk_bound_dev_if != dif)
257 } else if(score > badness) {
266 __inline__ struct sock *udp_v4_lookup(u32 saddr, u16 sport, u32 daddr, u16 dport, int dif)
270 read_lock(&udp_hash_lock);
271 sk = udp_v4_lookup_longway(saddr, sport, daddr, dport, dif);
274 read_unlock(&udp_hash_lock);
278 static inline struct sock *udp_v4_mcast_next(struct sock *sk,
279 u16 loc_port, u32 loc_addr,
280 u16 rmt_port, u32 rmt_addr,
283 struct hlist_node *node;
285 unsigned short hnum = ntohs(loc_port);
287 sk_for_each_from(s, node) {
288 struct inet_opt *inet = inet_sk(s);
290 if (inet->num != hnum ||
291 (inet->daddr && inet->daddr != rmt_addr) ||
292 (inet->dport != rmt_port && inet->dport) ||
293 (inet->rcv_saddr && inet->rcv_saddr != loc_addr) ||
295 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
297 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
307 * This routine is called by the ICMP module when it gets some
308 * sort of error condition. If err < 0 then the socket should
309 * be closed and the error returned to the user. If err > 0
310 * it's just the icmp type << 8 | icmp code.
311 * Header points to the ip header of the error packet. We move
312 * on past this. Then (as it used to claim before adjustment)
313 * header points to the first 8 bytes of the udp header. We need
314 * to find the appropriate port.
317 void udp_err(struct sk_buff *skb, u32 info)
319 struct inet_opt *inet;
320 struct iphdr *iph = (struct iphdr*)skb->data;
321 struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2));
322 int type = skb->h.icmph->type;
323 int code = skb->h.icmph->code;
328 sk = udp_v4_lookup(iph->daddr, uh->dest, iph->saddr, uh->source, skb->dev->ifindex);
330 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
331 return; /* No socket for error */
340 case ICMP_TIME_EXCEEDED:
343 case ICMP_SOURCE_QUENCH:
345 case ICMP_PARAMETERPROB:
349 case ICMP_DEST_UNREACH:
350 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
351 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
359 if (code <= NR_ICMP_UNREACH) {
360 harderr = icmp_err_convert[code].fatal;
361 err = icmp_err_convert[code].errno;
367 * RFC1122: OK. Passes ICMP errors back to application, as per
370 if (!inet->recverr) {
371 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
374 ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1));
377 sk->sk_error_report(sk);
383 * Throw away all pending data and cancel the corking. Socket is locked.
385 static void udp_flush_pending_frames(struct sock *sk)
387 struct udp_opt *up = udp_sk(sk);
392 ip_flush_pending_frames(sk);
397 * Push out all pending data as one UDP datagram. Socket is locked.
399 static int udp_push_pending_frames(struct sock *sk, struct udp_opt *up)
401 struct inet_opt *inet = inet_sk(sk);
402 struct flowi *fl = &inet->cork.fl;
407 /* Grab the skbuff where UDP header space exists. */
408 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
412 * Create a UDP header
415 uh->source = fl->fl_ip_sport;
416 uh->dest = fl->fl_ip_dport;
417 uh->len = htons(up->len);
420 if (sk->sk_no_check == UDP_CSUM_NOXMIT) {
421 skb->ip_summed = CHECKSUM_NONE;
425 if (skb_queue_len(&sk->sk_write_queue) == 1) {
427 * Only one fragment on the socket.
429 if (skb->ip_summed == CHECKSUM_HW) {
430 skb->csum = offsetof(struct udphdr, check);
431 uh->check = ~csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
432 up->len, IPPROTO_UDP, 0);
434 skb->csum = csum_partial((char *)uh,
435 sizeof(struct udphdr), skb->csum);
436 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
437 up->len, IPPROTO_UDP, skb->csum);
442 unsigned int csum = 0;
444 * HW-checksum won't work as there are two or more
445 * fragments on the socket so that all csums of sk_buffs
446 * should be together.
448 if (skb->ip_summed == CHECKSUM_HW) {
449 int offset = (unsigned char *)uh - skb->data;
450 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
452 skb->ip_summed = CHECKSUM_NONE;
454 skb->csum = csum_partial((char *)uh,
455 sizeof(struct udphdr), skb->csum);
458 skb_queue_walk(&sk->sk_write_queue, skb) {
459 csum = csum_add(csum, skb->csum);
461 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
462 up->len, IPPROTO_UDP, csum);
467 err = ip_push_pending_frames(sk);
475 static unsigned short udp_check(struct udphdr *uh, int len, unsigned long saddr, unsigned long daddr, unsigned long base)
477 return(csum_tcpudp_magic(saddr, daddr, len, IPPROTO_UDP, base));
480 int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
483 struct inet_opt *inet = inet_sk(sk);
484 struct udp_opt *up = udp_sk(sk);
486 struct ipcm_cookie ipc;
487 struct rtable *rt = NULL;
490 u32 daddr, faddr, saddr;
494 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
503 if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */
510 * There are pending frames.
511 * The socket lock must be held while it's corked.
514 if (likely(up->pending)) {
515 if (unlikely(up->pending != AF_INET)) {
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.
635 inet->cork.fl.fl4_dst = daddr;
636 inet->cork.fl.fl_ip_dport = dport;
637 inet->cork.fl.fl4_src = saddr;
638 inet->cork.fl.fl_ip_sport = inet->sport;
639 up->pending = AF_INET;
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(UDP_MIB_OUTDATAGRAMS);
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 __user *)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 __user *)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 size_t 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);
791 skb = skb_recv_datagram(sk, flags, noblock, &err);
795 copied = skb->len - sizeof(struct udphdr);
798 msg->msg_flags |= MSG_TRUNC;
801 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
802 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
804 } else if (msg->msg_flags&MSG_TRUNC) {
805 if (__udp_checksum_complete(skb))
807 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
810 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
819 sock_recv_timestamp(msg, sk, skb);
821 /* Copy the address. */
824 sin->sin_family = AF_INET;
825 sin->sin_port = skb->h.uh->source;
826 sin->sin_addr.s_addr = skb->nh.iph->saddr;
827 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
829 if (inet->cmsg_flags)
830 ip_cmsg_recv(msg, skb);
833 if (flags & MSG_TRUNC)
834 err = skb->len - sizeof(struct udphdr);
837 skb_free_datagram(sk, skb);
842 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
845 if (flags&MSG_PEEK) {
847 spin_lock_irq(&sk->sk_receive_queue.lock);
848 if (skb == skb_peek(&sk->sk_receive_queue)) {
849 __skb_unlink(skb, &sk->sk_receive_queue);
852 spin_unlock_irq(&sk->sk_receive_queue.lock);
857 skb_free_datagram(sk, skb);
865 int udp_disconnect(struct sock *sk, int flags)
867 struct inet_opt *inet = inet_sk(sk);
869 * 1003.1g - break association.
872 sk->sk_state = TCP_CLOSE;
875 sk->sk_bound_dev_if = 0;
876 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
877 inet_reset_saddr(sk);
879 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
880 sk->sk_prot->unhash(sk);
887 static void udp_close(struct sock *sk, long timeout)
889 sk_common_release(sk);
893 * 1 if the the UDP system should process it
894 * 0 if we should drop this packet
895 * -1 if it should get processed by xfrm4_rcv_encap
897 static int udp_encap_rcv(struct sock * sk, struct sk_buff *skb)
902 struct udp_opt *up = udp_sk(sk);
903 struct udphdr *uh = skb->h.uh;
907 __u8 *udpdata = (__u8 *)uh + sizeof(struct udphdr);
908 __u32 *udpdata32 = (__u32 *)udpdata;
909 __u16 encap_type = up->encap_type;
911 /* if we're overly short, let UDP handle it */
912 if (udpdata > skb->tail)
915 /* if this is not encapsulated socket, then just return now */
919 len = skb->tail - udpdata;
921 switch (encap_type) {
923 case UDP_ENCAP_ESPINUDP:
924 /* Check if this is a keepalive packet. If so, eat it. */
925 if (len == 1 && udpdata[0] == 0xff) {
927 } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0 ) {
928 /* ESP Packet without Non-ESP header */
929 len = sizeof(struct udphdr);
931 /* Must be an IKE packet.. pass it through */
934 case UDP_ENCAP_ESPINUDP_NON_IKE:
935 /* Check if this is a keepalive packet. If so, eat it. */
936 if (len == 1 && udpdata[0] == 0xff) {
938 } else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) &&
939 udpdata32[0] == 0 && udpdata32[1] == 0) {
941 /* ESP Packet with Non-IKE marker */
942 len = sizeof(struct udphdr) + 2 * sizeof(u32);
944 /* Must be an IKE packet.. pass it through */
949 /* At this point we are sure that this is an ESPinUDP packet,
950 * so we need to remove 'len' bytes from the packet (the UDP
951 * header and optional ESP marker bytes) and then modify the
952 * protocol to ESP, and then call into the transform receiver.
955 /* Now we can update and verify the packet length... */
957 iphlen = iph->ihl << 2;
958 iph->tot_len = htons(ntohs(iph->tot_len) - len);
959 if (skb->len < iphlen + len) {
960 /* packet is too small!?! */
964 /* pull the data buffer up to the ESP header and set the
965 * transport header to point to ESP. Keep UDP on the stack
968 skb->h.raw = skb_pull(skb, len);
970 /* modify the protocol (it's ESP!) */
971 iph->protocol = IPPROTO_ESP;
973 /* and let the caller know to send this into the ESP processor... */
981 * >0: "udp encap" protocol resubmission
983 * Note that in the success and error cases, the skb is assumed to
984 * have either been requeued or freed.
986 static int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
988 struct udp_opt *up = udp_sk(sk);
991 * Charge it to the socket, dropping if the queue is full.
993 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
998 if (up->encap_type) {
1000 * This is an encapsulation socket, so let's see if this is
1001 * an encapsulated packet.
1002 * If it's a keepalive packet, then just eat it.
1003 * If it's an encapsulateed packet, then pass it to the
1004 * IPsec xfrm input and return the response
1005 * appropriately. Otherwise, just fall through and
1006 * pass this up the UDP socket.
1010 ret = udp_encap_rcv(sk, skb);
1012 /* Eat the packet .. */
1017 /* process the ESP packet */
1018 ret = xfrm4_rcv_encap(skb, up->encap_type);
1019 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS);
1022 /* FALLTHROUGH -- it's a UDP Packet */
1025 if (sk->sk_filter && skb->ip_summed != CHECKSUM_UNNECESSARY) {
1026 if (__udp_checksum_complete(skb)) {
1027 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1031 skb->ip_summed = CHECKSUM_UNNECESSARY;
1034 if (sock_queue_rcv_skb(sk,skb)<0) {
1035 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1039 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS);
1044 * Multicasts and broadcasts go to each listener.
1046 * Note: called only from the BH handler context,
1047 * so we don't need to lock the hashes.
1049 static int udp_v4_mcast_deliver(struct sk_buff *skb, struct udphdr *uh,
1050 u32 saddr, u32 daddr)
1055 read_lock(&udp_hash_lock);
1056 sk = sk_head(&udp_hash[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
1057 dif = skb->dev->ifindex;
1058 sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
1060 struct sock *sknext = NULL;
1063 struct sk_buff *skb1 = skb;
1065 sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr,
1066 uh->source, saddr, dif);
1068 skb1 = skb_clone(skb, GFP_ATOMIC);
1071 int ret = udp_queue_rcv_skb(sk, skb1);
1073 /* we should probably re-process instead
1074 * of dropping packets here. */
1081 read_unlock(&udp_hash_lock);
1085 /* Initialize UDP checksum. If exited with zero value (success),
1086 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1087 * Otherwise, csum completion requires chacksumming packet body,
1088 * including udp header and folding it to skb->csum.
1090 static int udp_checksum_init(struct sk_buff *skb, struct udphdr *uh,
1091 unsigned short ulen, u32 saddr, u32 daddr)
1093 if (uh->check == 0) {
1094 skb->ip_summed = CHECKSUM_UNNECESSARY;
1095 } else if (skb->ip_summed == CHECKSUM_HW) {
1096 skb->ip_summed = CHECKSUM_UNNECESSARY;
1097 if (!udp_check(uh, ulen, saddr, daddr, skb->csum))
1099 NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG "udp v4 hw csum failure.\n"));
1100 skb->ip_summed = CHECKSUM_NONE;
1102 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
1103 skb->csum = csum_tcpudp_nofold(saddr, daddr, ulen, IPPROTO_UDP, 0);
1104 /* Probably, we should checksum udp header (it should be in cache
1105 * in any case) and data in tiny packets (< rx copybreak).
1110 /* XXX (mef) need to generalize the IPOD stuff. Right now I am borrowing
1111 from the ICMP infrastructure. */
1112 #ifdef CONFIG_ICMP_IPOD
1113 #include <linux/reboot.h>
1115 extern int sysctl_icmp_ipod_version;
1116 extern int sysctl_icmp_ipod_enabled;
1117 extern u32 sysctl_icmp_ipod_host;
1118 extern u32 sysctl_icmp_ipod_mask;
1119 extern char sysctl_icmp_ipod_key[32+1];
1120 #define IPOD_CHECK_KEY \
1121 (sysctl_icmp_ipod_key[0] != 0)
1122 #define IPOD_VALID_KEY(d) \
1123 (strncmp(sysctl_icmp_ipod_key, (char *)(d), strlen(sysctl_icmp_ipod_key)) == 0)
1125 static void udp_ping_of_death(struct sk_buff *skb, struct udphdr *uh, u32 saddr)
1130 * If IPOD not enabled or wrong UDP IPOD port, ignore.
1132 if (!sysctl_icmp_ipod_enabled || (ntohs(uh->dest) != 664))
1136 printk(KERN_INFO "IPOD: got udp pod request, host=%u.%u.%u.%u\n", NIPQUAD(saddr));
1141 * First check the source address info.
1142 * If host not set, ignore.
1144 if (sysctl_icmp_ipod_host != 0xffffffff &&
1145 (ntohl(saddr) & sysctl_icmp_ipod_mask) == sysctl_icmp_ipod_host) {
1147 * Now check the key if enabled.
1148 * If packet doesn't contain enough data or key
1149 * is otherwise invalid, ignore.
1151 if (IPOD_CHECK_KEY) {
1152 if (pskb_may_pull(skb, sizeof(sysctl_icmp_ipod_key)+sizeof(struct udphdr)-1)){
1155 for (i=0;i<32+1;i++){
1156 printk("%c",((char*)skb->data)[i+sizeof(struct udphdr)]);
1160 if (IPOD_VALID_KEY(skb->data+sizeof(struct udphdr)))
1168 sysctl_icmp_ipod_enabled = 0;
1169 printk(KERN_CRIT "IPOD: reboot forced by %u.%u.%u.%u...\n",
1171 machine_restart(NULL);
1173 printk(KERN_WARNING "IPOD: from %u.%u.%u.%u rejected\n",
1180 * All we need to do is get the socket, and then do a checksum.
1183 int udp_rcv(struct sk_buff *skb)
1187 unsigned short ulen;
1188 struct rtable *rt = (struct rtable*)skb->dst;
1189 u32 saddr = skb->nh.iph->saddr;
1190 u32 daddr = skb->nh.iph->daddr;
1194 * Validate the packet and the UDP length.
1196 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1201 ulen = ntohs(uh->len);
1203 if (ulen > len || ulen < sizeof(*uh))
1206 if (pskb_trim(skb, ulen))
1209 if (udp_checksum_init(skb, uh, ulen, saddr, daddr) < 0)
1212 if(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1213 return udp_v4_mcast_deliver(skb, uh, saddr, daddr);
1215 #ifdef CONFIG_ICMP_IPOD
1216 udp_ping_of_death(skb, uh, saddr);
1219 sk = udp_v4_lookup(saddr, uh->source, daddr, uh->dest, skb->dev->ifindex);
1222 int ret = udp_queue_rcv_skb(sk, skb);
1225 /* a return value > 0 means to resubmit the input, but
1226 * it it wants the return to be -protocol, or 0
1233 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1236 /* No socket. Drop packet silently, if checksum is wrong */
1237 if (udp_checksum_complete(skb))
1240 UDP_INC_STATS_BH(UDP_MIB_NOPORTS);
1241 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1244 * Hmm. We got an UDP packet to a port to which we
1245 * don't wanna listen. Ignore it.
1251 NETDEBUG(if (net_ratelimit())
1252 printk(KERN_DEBUG "UDP: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n",
1260 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1266 * RFC1122: OK. Discards the bad packet silently (as far as
1267 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1269 NETDEBUG(if (net_ratelimit())
1270 printk(KERN_DEBUG "UDP: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n",
1277 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1282 static int udp_destroy_sock(struct sock *sk)
1285 udp_flush_pending_frames(sk);
1291 * Socket option code for UDP
1293 static int udp_setsockopt(struct sock *sk, int level, int optname,
1294 char __user *optval, int optlen)
1296 struct udp_opt *up = udp_sk(sk);
1300 if (level != SOL_UDP)
1301 return ip_setsockopt(sk, level, optname, optval, optlen);
1303 if(optlen<sizeof(int))
1306 if (get_user(val, (int __user *)optval))
1316 udp_push_pending_frames(sk, up);
1324 case UDP_ENCAP_ESPINUDP:
1325 case UDP_ENCAP_ESPINUDP_NON_IKE:
1326 up->encap_type = val;
1342 static int udp_getsockopt(struct sock *sk, int level, int optname,
1343 char __user *optval, int __user *optlen)
1345 struct udp_opt *up = udp_sk(sk);
1348 if (level != SOL_UDP)
1349 return ip_getsockopt(sk, level, optname, optval, optlen);
1351 if(get_user(len,optlen))
1354 len = min_t(unsigned int, len, sizeof(int));
1365 val = up->encap_type;
1369 return -ENOPROTOOPT;
1372 if(put_user(len, optlen))
1374 if(copy_to_user(optval, &val,len))
1380 struct proto udp_prot = {
1383 .connect = ip4_datagram_connect,
1384 .disconnect = udp_disconnect,
1386 .destroy = udp_destroy_sock,
1387 .setsockopt = udp_setsockopt,
1388 .getsockopt = udp_getsockopt,
1389 .sendmsg = udp_sendmsg,
1390 .recvmsg = udp_recvmsg,
1391 .sendpage = udp_sendpage,
1392 .backlog_rcv = udp_queue_rcv_skb,
1393 .hash = udp_v4_hash,
1394 .unhash = udp_v4_unhash,
1395 .get_port = udp_v4_get_port,
1398 /* ------------------------------------------------------------------------ */
1399 #ifdef CONFIG_PROC_FS
1401 static struct sock *udp_get_first(struct seq_file *seq)
1404 struct udp_iter_state *state = seq->private;
1406 for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) {
1407 struct hlist_node *node;
1409 sk_for_each(sk, node, &udp_hash[state->bucket]) {
1410 if (sk->sk_family == state->family &&
1411 vx_check(sk->sk_xid, VX_WATCH|VX_IDENT))
1420 static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
1422 struct udp_iter_state *state = seq->private;
1428 } while (sk && (sk->sk_family != state->family ||
1429 !vx_check(sk->sk_xid, VX_WATCH|VX_IDENT)));
1431 if (!sk && ++state->bucket < UDP_HTABLE_SIZE) {
1432 sk = sk_head(&udp_hash[state->bucket]);
1438 static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
1440 struct sock *sk = udp_get_first(seq);
1443 while(pos && (sk = udp_get_next(seq, sk)) != NULL)
1445 return pos ? NULL : sk;
1448 static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
1450 read_lock(&udp_hash_lock);
1451 return *pos ? udp_get_idx(seq, *pos-1) : (void *)1;
1454 static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1459 sk = udp_get_idx(seq, 0);
1461 sk = udp_get_next(seq, v);
1467 static void udp_seq_stop(struct seq_file *seq, void *v)
1469 read_unlock(&udp_hash_lock);
1472 static int udp_seq_open(struct inode *inode, struct file *file)
1474 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
1475 struct seq_file *seq;
1477 struct udp_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
1481 memset(s, 0, sizeof(*s));
1482 s->family = afinfo->family;
1483 s->seq_ops.start = udp_seq_start;
1484 s->seq_ops.next = udp_seq_next;
1485 s->seq_ops.show = afinfo->seq_show;
1486 s->seq_ops.stop = udp_seq_stop;
1488 rc = seq_open(file, &s->seq_ops);
1492 seq = file->private_data;
1501 /* ------------------------------------------------------------------------ */
1502 int udp_proc_register(struct udp_seq_afinfo *afinfo)
1504 struct proc_dir_entry *p;
1509 afinfo->seq_fops->owner = afinfo->owner;
1510 afinfo->seq_fops->open = udp_seq_open;
1511 afinfo->seq_fops->read = seq_read;
1512 afinfo->seq_fops->llseek = seq_lseek;
1513 afinfo->seq_fops->release = seq_release_private;
1515 p = proc_net_fops_create(afinfo->name, S_IRUGO, afinfo->seq_fops);
1523 void udp_proc_unregister(struct udp_seq_afinfo *afinfo)
1527 proc_net_remove(afinfo->name);
1528 memset(afinfo->seq_fops, 0, sizeof(*afinfo->seq_fops));
1531 /* ------------------------------------------------------------------------ */
1532 static void udp4_format_sock(struct sock *sp, char *tmpbuf, int bucket)
1534 struct inet_opt *inet = inet_sk(sp);
1535 unsigned int dest = inet->daddr;
1536 unsigned int src = inet->rcv_saddr;
1537 __u16 destp = ntohs(inet->dport);
1538 __u16 srcp = ntohs(inet->sport);
1540 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
1541 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
1542 bucket, src, srcp, dest, destp, sp->sk_state,
1543 atomic_read(&sp->sk_wmem_alloc),
1544 atomic_read(&sp->sk_rmem_alloc),
1545 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
1546 atomic_read(&sp->sk_refcnt), sp);
1549 static int udp4_seq_show(struct seq_file *seq, void *v)
1551 if (v == SEQ_START_TOKEN)
1552 seq_printf(seq, "%-127s\n",
1553 " sl local_address rem_address st tx_queue "
1554 "rx_queue tr tm->when retrnsmt uid timeout "
1558 struct udp_iter_state *state = seq->private;
1560 udp4_format_sock(v, tmpbuf, state->bucket);
1561 seq_printf(seq, "%-127s\n", tmpbuf);
1566 /* ------------------------------------------------------------------------ */
1567 static struct file_operations udp4_seq_fops;
1568 static struct udp_seq_afinfo udp4_seq_afinfo = {
1569 .owner = THIS_MODULE,
1572 .seq_show = udp4_seq_show,
1573 .seq_fops = &udp4_seq_fops,
1576 int __init udp4_proc_init(void)
1578 return udp_proc_register(&udp4_seq_afinfo);
1581 void udp4_proc_exit(void)
1583 udp_proc_unregister(&udp4_seq_afinfo);
1585 #endif /* CONFIG_PROC_FS */
1587 EXPORT_SYMBOL(udp_disconnect);
1588 EXPORT_SYMBOL(udp_hash);
1589 EXPORT_SYMBOL(udp_hash_lock);
1590 EXPORT_SYMBOL(udp_ioctl);
1591 EXPORT_SYMBOL(udp_port_rover);
1592 EXPORT_SYMBOL(udp_prot);
1593 EXPORT_SYMBOL(udp_sendmsg);
1595 #ifdef CONFIG_PROC_FS
1596 EXPORT_SYMBOL(udp_proc_register);
1597 EXPORT_SYMBOL(udp_proc_unregister);