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>
110 #include <linux/vs_base.h>
113 * Snmp MIB for the UDP layer
116 DEFINE_SNMP_STAT(struct udp_mib, udp_statistics);
118 struct hlist_head udp_hash[UDP_HTABLE_SIZE];
119 rwlock_t udp_hash_lock = RW_LOCK_UNLOCKED;
121 /* Shared by v4/v6 udp. */
124 static int udp_v4_get_port(struct sock *sk, unsigned short snum)
126 struct hlist_node *node;
128 struct inet_opt *inet = inet_sk(sk);
130 write_lock_bh(&udp_hash_lock);
132 int best_size_so_far, best, result, i;
134 if (udp_port_rover > sysctl_local_port_range[1] ||
135 udp_port_rover < sysctl_local_port_range[0])
136 udp_port_rover = sysctl_local_port_range[0];
137 best_size_so_far = 32767;
138 best = result = udp_port_rover;
139 for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) {
140 struct hlist_head *list;
143 list = &udp_hash[result & (UDP_HTABLE_SIZE - 1)];
144 if (hlist_empty(list)) {
145 if (result > sysctl_local_port_range[1])
146 result = sysctl_local_port_range[0] +
147 ((result - sysctl_local_port_range[0]) &
148 (UDP_HTABLE_SIZE - 1));
152 sk_for_each(sk2, node, list)
153 if (++size >= best_size_so_far)
155 best_size_so_far = size;
160 for(i = 0; i < (1 << 16) / UDP_HTABLE_SIZE; i++, result += UDP_HTABLE_SIZE) {
161 if (result > sysctl_local_port_range[1])
162 result = sysctl_local_port_range[0]
163 + ((result - sysctl_local_port_range[0]) &
164 (UDP_HTABLE_SIZE - 1));
165 if (!udp_lport_inuse(result))
168 if (i >= (1 << 16) / UDP_HTABLE_SIZE)
171 udp_port_rover = snum = result;
173 sk_for_each(sk2, node,
174 &udp_hash[snum & (UDP_HTABLE_SIZE - 1)]) {
175 struct inet_opt *inet2 = inet_sk(sk2);
177 if (inet2->num == snum &&
179 !ipv6_only_sock(sk2) &&
180 (!sk2->sk_bound_dev_if ||
181 !sk->sk_bound_dev_if ||
182 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
183 (!inet2->rcv_saddr ||
185 inet2->rcv_saddr == inet->rcv_saddr) &&
186 (!sk2->sk_reuse || !sk->sk_reuse))
191 if (sk_unhashed(sk)) {
192 struct hlist_head *h = &udp_hash[snum & (UDP_HTABLE_SIZE - 1)];
195 sock_prot_inc_use(sk->sk_prot);
197 write_unlock_bh(&udp_hash_lock);
201 write_unlock_bh(&udp_hash_lock);
205 static void udp_v4_hash(struct sock *sk)
210 static void udp_v4_unhash(struct sock *sk)
212 write_lock_bh(&udp_hash_lock);
213 if (sk_del_node_init(sk)) {
214 inet_sk(sk)->num = 0;
215 sock_prot_dec_use(sk->sk_prot);
217 write_unlock_bh(&udp_hash_lock);
220 /* UDP is nearly always wildcards out the wazoo, it makes no sense to try
221 * harder than this. -DaveM
223 struct sock *udp_v4_lookup_longway(u32 saddr, u16 sport, 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_opt *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 __inline__ struct sock *udp_v4_lookup(u32 saddr, u16 sport, u32 daddr, u16 dport, int dif)
271 read_lock(&udp_hash_lock);
272 sk = udp_v4_lookup_longway(saddr, sport, daddr, dport, dif);
275 read_unlock(&udp_hash_lock);
279 static inline struct sock *udp_v4_mcast_next(struct sock *sk,
280 u16 loc_port, u32 loc_addr,
281 u16 rmt_port, u32 rmt_addr,
284 struct hlist_node *node;
286 unsigned short hnum = ntohs(loc_port);
288 sk_for_each_from(s, node) {
289 struct inet_opt *inet = inet_sk(s);
291 if (inet->num != hnum ||
292 (inet->daddr && inet->daddr != rmt_addr) ||
293 (inet->dport != rmt_port && inet->dport) ||
294 (inet->rcv_saddr && inet->rcv_saddr != loc_addr) ||
296 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
298 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
308 * This routine is called by the ICMP module when it gets some
309 * sort of error condition. If err < 0 then the socket should
310 * be closed and the error returned to the user. If err > 0
311 * it's just the icmp type << 8 | icmp code.
312 * Header points to the ip header of the error packet. We move
313 * on past this. Then (as it used to claim before adjustment)
314 * header points to the first 8 bytes of the udp header. We need
315 * to find the appropriate port.
318 void udp_err(struct sk_buff *skb, u32 info)
320 struct inet_opt *inet;
321 struct iphdr *iph = (struct iphdr*)skb->data;
322 struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2));
323 int type = skb->h.icmph->type;
324 int code = skb->h.icmph->code;
329 sk = udp_v4_lookup(iph->daddr, uh->dest, iph->saddr, uh->source, skb->dev->ifindex);
331 ICMP_INC_STATS_BH(IcmpInErrors);
332 return; /* No socket for error */
341 case ICMP_TIME_EXCEEDED:
344 case ICMP_SOURCE_QUENCH:
346 case ICMP_PARAMETERPROB:
350 case ICMP_DEST_UNREACH:
351 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
352 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
360 if (code <= NR_ICMP_UNREACH) {
361 harderr = icmp_err_convert[code].fatal;
362 err = icmp_err_convert[code].errno;
368 * RFC1122: OK. Passes ICMP errors back to application, as per
371 if (!inet->recverr) {
372 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
375 ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1));
378 sk->sk_error_report(sk);
384 * Throw away all pending data and cancel the corking. Socket is locked.
386 static void udp_flush_pending_frames(struct sock *sk)
388 struct udp_opt *up = udp_sk(sk);
393 ip_flush_pending_frames(sk);
398 * Push out all pending data as one UDP datagram. Socket is locked.
400 static int udp_push_pending_frames(struct sock *sk, struct udp_opt *up)
402 struct inet_opt *inet = inet_sk(sk);
403 struct flowi *fl = &inet->cork.fl;
408 /* Grab the skbuff where UDP header space exists. */
409 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
413 * Create a UDP header
416 uh->source = fl->fl_ip_sport;
417 uh->dest = fl->fl_ip_dport;
418 uh->len = htons(up->len);
421 if (sk->sk_no_check == UDP_CSUM_NOXMIT) {
422 skb->ip_summed = CHECKSUM_NONE;
426 if (skb_queue_len(&sk->sk_write_queue) == 1) {
428 * Only one fragment on the socket.
430 if (skb->ip_summed == CHECKSUM_HW) {
431 skb->csum = offsetof(struct udphdr, check);
432 uh->check = ~csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
433 up->len, IPPROTO_UDP, 0);
435 skb->csum = csum_partial((char *)uh,
436 sizeof(struct udphdr), skb->csum);
437 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
438 up->len, IPPROTO_UDP, skb->csum);
443 unsigned int csum = 0;
445 * HW-checksum won't work as there are two or more
446 * fragments on the socket so that all csums of sk_buffs
447 * should be together.
449 if (skb->ip_summed == CHECKSUM_HW) {
450 int offset = (unsigned char *)uh - skb->data;
451 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
453 skb->ip_summed = CHECKSUM_NONE;
455 skb->csum = csum_partial((char *)uh,
456 sizeof(struct udphdr), skb->csum);
459 skb_queue_walk(&sk->sk_write_queue, skb) {
460 csum = csum_add(csum, skb->csum);
462 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
463 up->len, IPPROTO_UDP, csum);
468 err = ip_push_pending_frames(sk);
476 static unsigned short udp_check(struct udphdr *uh, int len, unsigned long saddr, unsigned long daddr, unsigned long base)
478 return(csum_tcpudp_magic(saddr, daddr, len, IPPROTO_UDP, base));
481 int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
484 struct inet_opt *inet = inet_sk(sk);
485 struct udp_opt *up = udp_sk(sk);
487 struct ipcm_cookie ipc;
488 struct rtable *rt = NULL;
491 u32 daddr, faddr, saddr;
495 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
504 if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */
511 * There are pending frames.
512 * The socket lock must be held while it's corked.
515 if (likely(up->pending)) {
516 if (unlikely(up->pending != AF_INET)) {
524 ulen += sizeof(struct udphdr);
527 * Get and verify the address.
530 struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name;
531 if (msg->msg_namelen < sizeof(*usin))
533 if (usin->sin_family != AF_INET) {
534 if (usin->sin_family != AF_UNSPEC)
538 daddr = usin->sin_addr.s_addr;
539 dport = usin->sin_port;
543 if (sk->sk_state != TCP_ESTABLISHED)
544 return -EDESTADDRREQ;
547 /* Open fast path for connected socket.
548 Route will not be used, if at least one option is set.
552 ipc.addr = inet->saddr;
554 ipc.oif = sk->sk_bound_dev_if;
555 if (msg->msg_controllen) {
556 err = ip_cmsg_send(msg, &ipc);
567 ipc.addr = faddr = daddr;
569 if (ipc.opt && ipc.opt->srr) {
572 faddr = ipc.opt->faddr;
575 tos = RT_TOS(inet->tos);
576 if (sk->sk_localroute || (msg->msg_flags & MSG_DONTROUTE) ||
577 (ipc.opt && ipc.opt->is_strictroute)) {
582 if (MULTICAST(daddr)) {
584 ipc.oif = inet->mc_index;
586 saddr = inet->mc_addr;
591 rt = (struct rtable*)sk_dst_check(sk, 0);
594 struct flowi fl = { .oif = ipc.oif,
599 .proto = IPPROTO_UDP,
601 { .sport = inet->sport,
602 .dport = dport } } };
603 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
608 if ((rt->rt_flags & RTCF_BROADCAST) &&
609 !sock_flag(sk, SOCK_BROADCAST))
612 sk_dst_set(sk, dst_clone(&rt->u.dst));
615 if (msg->msg_flags&MSG_CONFIRM)
621 daddr = ipc.addr = rt->rt_dst;
624 if (unlikely(up->pending)) {
625 /* The socket is already corked while preparing it. */
626 /* ... which is an evident application bug. --ANK */
629 NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG "udp cork app bug 2\n"));
634 * Now cork the socket to pend data.
636 inet->cork.fl.fl4_dst = daddr;
637 inet->cork.fl.fl_ip_dport = dport;
638 inet->cork.fl.fl4_src = saddr;
639 inet->cork.fl.fl_ip_sport = inet->sport;
640 up->pending = AF_INET;
644 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, ulen,
645 sizeof(struct udphdr), &ipc, rt,
646 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
648 udp_flush_pending_frames(sk);
650 err = udp_push_pending_frames(sk, up);
658 UDP_INC_STATS_USER(UdpOutDatagrams);
664 dst_confirm(&rt->u.dst);
665 if (!(msg->msg_flags&MSG_PROBE) || len)
666 goto back_from_confirm;
671 int udp_sendpage(struct sock *sk, struct page *page, int offset, size_t size, int flags)
673 struct udp_opt *up = udp_sk(sk);
677 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
679 /* Call udp_sendmsg to specify destination address which
680 * sendpage interface can't pass.
681 * This will succeed only when the socket is connected.
683 ret = udp_sendmsg(NULL, sk, &msg, 0);
690 if (unlikely(!up->pending)) {
693 NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG "udp cork app bug 3\n"));
697 ret = ip_append_page(sk, page, offset, size, flags);
698 if (ret == -EOPNOTSUPP) {
700 return sock_no_sendpage(sk->sk_socket, page, offset,
704 udp_flush_pending_frames(sk);
709 if (!(up->corkflag || (flags&MSG_MORE)))
710 ret = udp_push_pending_frames(sk, up);
719 * IOCTL requests applicable to the UDP protocol
722 int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
728 int amount = atomic_read(&sk->sk_wmem_alloc);
729 return put_user(amount, (int __user *)arg);
735 unsigned long amount;
738 spin_lock_irq(&sk->sk_receive_queue.lock);
739 skb = skb_peek(&sk->sk_receive_queue);
742 * We will only return the amount
743 * of this packet since that is all
746 amount = skb->len - sizeof(struct udphdr);
748 spin_unlock_irq(&sk->sk_receive_queue.lock);
749 return put_user(amount, (int __user *)arg);
758 static __inline__ int __udp_checksum_complete(struct sk_buff *skb)
760 return (unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum));
763 static __inline__ int udp_checksum_complete(struct sk_buff *skb)
765 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
766 __udp_checksum_complete(skb);
770 * This should be easy, if there is something there we
771 * return it, otherwise we block.
774 int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
775 size_t len, int noblock, int flags, int *addr_len)
777 struct inet_opt *inet = inet_sk(sk);
778 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
783 * Check any passed addresses
786 *addr_len=sizeof(*sin);
788 if (flags & MSG_ERRQUEUE)
789 return ip_recv_error(sk, msg, len);
792 skb = skb_recv_datagram(sk, flags, noblock, &err);
796 copied = skb->len - sizeof(struct udphdr);
799 msg->msg_flags |= MSG_TRUNC;
802 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
803 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
805 } else if (msg->msg_flags&MSG_TRUNC) {
806 if (__udp_checksum_complete(skb))
808 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
811 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
820 sock_recv_timestamp(msg, sk, skb);
822 /* Copy the address. */
825 sin->sin_family = AF_INET;
826 sin->sin_port = skb->h.uh->source;
827 sin->sin_addr.s_addr = skb->nh.iph->saddr;
828 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
830 if (inet->cmsg_flags)
831 ip_cmsg_recv(msg, skb);
835 skb_free_datagram(sk, skb);
840 UDP_INC_STATS_BH(UdpInErrors);
843 if (flags&MSG_PEEK) {
845 spin_lock_irq(&sk->sk_receive_queue.lock);
846 if (skb == skb_peek(&sk->sk_receive_queue)) {
847 __skb_unlink(skb, &sk->sk_receive_queue);
850 spin_unlock_irq(&sk->sk_receive_queue.lock);
855 skb_free_datagram(sk, skb);
862 int udp_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
864 struct inet_opt *inet = inet_sk(sk);
865 struct sockaddr_in *usin = (struct sockaddr_in *) uaddr;
872 if (addr_len < sizeof(*usin))
875 if (usin->sin_family != AF_INET)
876 return -EAFNOSUPPORT;
880 oif = sk->sk_bound_dev_if;
882 if (MULTICAST(usin->sin_addr.s_addr)) {
884 oif = inet->mc_index;
886 saddr = inet->mc_addr;
888 err = ip_route_connect(&rt, usin->sin_addr.s_addr, saddr,
889 RT_CONN_FLAGS(sk), oif,
891 inet->sport, usin->sin_port, sk);
894 if ((rt->rt_flags & RTCF_BROADCAST) && !sock_flag(sk, SOCK_BROADCAST)) {
899 inet->saddr = rt->rt_src; /* Update source address */
900 if (!inet->rcv_saddr)
901 inet->rcv_saddr = rt->rt_src;
902 inet->daddr = rt->rt_dst;
903 inet->dport = usin->sin_port;
904 sk->sk_state = TCP_ESTABLISHED;
907 sk_dst_set(sk, &rt->u.dst);
911 int udp_disconnect(struct sock *sk, int flags)
913 struct inet_opt *inet = inet_sk(sk);
915 * 1003.1g - break association.
918 sk->sk_state = TCP_CLOSE;
921 sk->sk_bound_dev_if = 0;
922 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
923 inet_reset_saddr(sk);
925 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
926 sk->sk_prot->unhash(sk);
933 static void udp_close(struct sock *sk, long timeout)
935 sk_common_release(sk);
939 * 1 if the the UDP system should process it
940 * 0 if we should drop this packet
941 * -1 if it should get processed by xfrm4_rcv_encap
943 static int udp_encap_rcv(struct sock * sk, struct sk_buff *skb)
948 struct udp_opt *up = udp_sk(sk);
949 struct udphdr *uh = skb->h.uh;
953 __u8 *udpdata = (__u8 *)uh + sizeof(struct udphdr);
954 __u32 *udpdata32 = (__u32 *)udpdata;
955 __u16 encap_type = up->encap_type;
957 /* if we're overly short, let UDP handle it */
958 if (udpdata > skb->tail)
961 /* if this is not encapsulated socket, then just return now */
965 len = skb->tail - udpdata;
967 switch (encap_type) {
969 case UDP_ENCAP_ESPINUDP:
970 /* Check if this is a keepalive packet. If so, eat it. */
971 if (len == 1 && udpdata[0] == 0xff) {
973 } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0 ) {
974 /* ESP Packet without Non-ESP header */
975 len = sizeof(struct udphdr);
977 /* Must be an IKE packet.. pass it through */
980 case UDP_ENCAP_ESPINUDP_NON_IKE:
981 /* Check if this is a keepalive packet. If so, eat it. */
982 if (len == 1 && udpdata[0] == 0xff) {
984 } else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) &&
985 udpdata32[0] == 0 && udpdata32[1] == 0) {
987 /* ESP Packet with Non-IKE marker */
988 len = sizeof(struct udphdr) + 2 * sizeof(u32);
990 /* Must be an IKE packet.. pass it through */
994 /* At this point we are sure that this is an ESPinUDP packet,
995 * so we need to remove 'len' bytes from the packet (the UDP
996 * header and optional ESP marker bytes) and then modify the
997 * protocol to ESP, and then call into the transform receiver.
1000 /* Now we can update and verify the packet length... */
1002 iphlen = iph->ihl << 2;
1003 iph->tot_len = htons(ntohs(iph->tot_len) - len);
1004 if (skb->len < iphlen + len) {
1005 /* packet is too small!?! */
1009 /* pull the data buffer up to the ESP header and set the
1010 * transport header to point to ESP. Keep UDP on the stack
1013 skb->h.raw = skb_pull(skb, len);
1015 /* modify the protocol (it's ESP!) */
1016 iph->protocol = IPPROTO_ESP;
1018 /* and let the caller know to send this into the ESP processor... */
1026 * >0: "udp encap" protocol resubmission
1028 * Note that in the success and error cases, the skb is assumed to
1029 * have either been requeued or freed.
1031 static int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
1033 struct udp_opt *up = udp_sk(sk);
1036 * Charge it to the socket, dropping if the queue is full.
1038 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
1043 if (up->encap_type) {
1045 * This is an encapsulation socket, so let's see if this is
1046 * an encapsulated packet.
1047 * If it's a keepalive packet, then just eat it.
1048 * If it's an encapsulateed packet, then pass it to the
1049 * IPsec xfrm input and return the response
1050 * appropriately. Otherwise, just fall through and
1051 * pass this up the UDP socket.
1055 ret = udp_encap_rcv(sk, skb);
1057 /* Eat the packet .. */
1062 /* process the ESP packet */
1063 ret = xfrm4_rcv_encap(skb, up->encap_type);
1064 UDP_INC_STATS_BH(UdpInDatagrams);
1067 /* FALLTHROUGH -- it's a UDP Packet */
1070 if (sk->sk_filter && skb->ip_summed != CHECKSUM_UNNECESSARY) {
1071 if (__udp_checksum_complete(skb)) {
1072 UDP_INC_STATS_BH(UdpInErrors);
1076 skb->ip_summed = CHECKSUM_UNNECESSARY;
1079 if (sock_queue_rcv_skb(sk,skb)<0) {
1080 UDP_INC_STATS_BH(UdpInErrors);
1084 UDP_INC_STATS_BH(UdpInDatagrams);
1089 * Multicasts and broadcasts go to each listener.
1091 * Note: called only from the BH handler context,
1092 * so we don't need to lock the hashes.
1094 static int udp_v4_mcast_deliver(struct sk_buff *skb, struct udphdr *uh,
1095 u32 saddr, u32 daddr)
1100 read_lock(&udp_hash_lock);
1101 sk = sk_head(&udp_hash[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
1102 dif = skb->dev->ifindex;
1103 sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
1105 struct sock *sknext = NULL;
1108 struct sk_buff *skb1 = skb;
1110 sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr,
1111 uh->source, saddr, dif);
1113 skb1 = skb_clone(skb, GFP_ATOMIC);
1116 int ret = udp_queue_rcv_skb(sk, skb1);
1118 /* we should probably re-process instead
1119 * of dropping packets here. */
1126 read_unlock(&udp_hash_lock);
1130 /* Initialize UDP checksum. If exited with zero value (success),
1131 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1132 * Otherwise, csum completion requires chacksumming packet body,
1133 * including udp header and folding it to skb->csum.
1135 static int udp_checksum_init(struct sk_buff *skb, struct udphdr *uh,
1136 unsigned short ulen, u32 saddr, u32 daddr)
1138 if (uh->check == 0) {
1139 skb->ip_summed = CHECKSUM_UNNECESSARY;
1140 } else if (skb->ip_summed == CHECKSUM_HW) {
1141 skb->ip_summed = CHECKSUM_UNNECESSARY;
1142 if (!udp_check(uh, ulen, saddr, daddr, skb->csum))
1144 NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG "udp v4 hw csum failure.\n"));
1145 skb->ip_summed = CHECKSUM_NONE;
1147 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
1148 skb->csum = csum_tcpudp_nofold(saddr, daddr, ulen, IPPROTO_UDP, 0);
1149 /* Probably, we should checksum udp header (it should be in cache
1150 * in any case) and data in tiny packets (< rx copybreak).
1156 * All we need to do is get the socket, and then do a checksum.
1159 int udp_rcv(struct sk_buff *skb)
1163 unsigned short ulen;
1164 struct rtable *rt = (struct rtable*)skb->dst;
1165 u32 saddr = skb->nh.iph->saddr;
1166 u32 daddr = skb->nh.iph->daddr;
1170 * Validate the packet and the UDP length.
1172 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1177 ulen = ntohs(uh->len);
1179 if (ulen > len || ulen < sizeof(*uh))
1182 if (pskb_trim(skb, ulen))
1185 if (udp_checksum_init(skb, uh, ulen, saddr, daddr) < 0)
1188 if(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1189 return udp_v4_mcast_deliver(skb, uh, saddr, daddr);
1191 sk = udp_v4_lookup(saddr, uh->source, daddr, uh->dest, skb->dev->ifindex);
1194 int ret = udp_queue_rcv_skb(sk, skb);
1197 /* a return value > 0 means to resubmit the input, but
1198 * it it wants the return to be -protocol, or 0
1205 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1208 /* No socket. Drop packet silently, if checksum is wrong */
1209 if (udp_checksum_complete(skb))
1212 UDP_INC_STATS_BH(UdpNoPorts);
1213 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1216 * Hmm. We got an UDP packet to a port to which we
1217 * don't wanna listen. Ignore it.
1223 NETDEBUG(if (net_ratelimit())
1224 printk(KERN_DEBUG "UDP: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n",
1232 UDP_INC_STATS_BH(UdpInErrors);
1238 * RFC1122: OK. Discards the bad packet silently (as far as
1239 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1241 NETDEBUG(if (net_ratelimit())
1242 printk(KERN_DEBUG "UDP: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n",
1249 UDP_INC_STATS_BH(UdpInErrors);
1254 static int udp_destroy_sock(struct sock *sk)
1257 udp_flush_pending_frames(sk);
1263 * Socket option code for UDP
1265 static int udp_setsockopt(struct sock *sk, int level, int optname,
1266 char __user *optval, int optlen)
1268 struct udp_opt *up = udp_sk(sk);
1272 if (level != SOL_UDP)
1273 return ip_setsockopt(sk, level, optname, optval, optlen);
1275 if(optlen<sizeof(int))
1278 if (get_user(val, (int __user *)optval))
1288 udp_push_pending_frames(sk, up);
1296 case UDP_ENCAP_ESPINUDP:
1297 case UDP_ENCAP_ESPINUDP_NON_IKE:
1298 up->encap_type = val;
1314 static int udp_getsockopt(struct sock *sk, int level, int optname,
1315 char __user *optval, int __user *optlen)
1317 struct udp_opt *up = udp_sk(sk);
1320 if (level != SOL_UDP)
1321 return ip_getsockopt(sk, level, optname, optval, optlen);
1323 if(get_user(len,optlen))
1326 len = min_t(unsigned int, len, sizeof(int));
1337 val = up->encap_type;
1341 return -ENOPROTOOPT;
1344 if(put_user(len, optlen))
1346 if(copy_to_user(optval, &val,len))
1352 struct proto udp_prot = {
1355 .connect = udp_connect,
1356 .disconnect = udp_disconnect,
1358 .destroy = udp_destroy_sock,
1359 .setsockopt = udp_setsockopt,
1360 .getsockopt = udp_getsockopt,
1361 .sendmsg = udp_sendmsg,
1362 .recvmsg = udp_recvmsg,
1363 .sendpage = udp_sendpage,
1364 .backlog_rcv = udp_queue_rcv_skb,
1365 .hash = udp_v4_hash,
1366 .unhash = udp_v4_unhash,
1367 .get_port = udp_v4_get_port,
1370 /* ------------------------------------------------------------------------ */
1371 #ifdef CONFIG_PROC_FS
1373 static struct sock *udp_get_first(struct seq_file *seq)
1376 struct udp_iter_state *state = seq->private;
1378 for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) {
1379 struct hlist_node *node;
1381 sk_for_each(sk, node, &udp_hash[state->bucket]) {
1382 if (sk->sk_family == state->family &&
1383 vx_check(sk->sk_xid, VX_WATCH|VX_IDENT))
1392 static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
1394 struct udp_iter_state *state = seq->private;
1400 } while (sk && (sk->sk_family != state->family ||
1401 !vx_check(sk->sk_xid, VX_WATCH|VX_IDENT)));
1403 if (!sk && ++state->bucket < UDP_HTABLE_SIZE) {
1404 sk = sk_head(&udp_hash[state->bucket]);
1410 static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
1412 struct sock *sk = udp_get_first(seq);
1415 while(pos && (sk = udp_get_next(seq, sk)) != NULL)
1417 return pos ? NULL : sk;
1420 static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
1422 read_lock(&udp_hash_lock);
1423 return *pos ? udp_get_idx(seq, *pos-1) : (void *)1;
1426 static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1431 sk = udp_get_idx(seq, 0);
1433 sk = udp_get_next(seq, v);
1439 static void udp_seq_stop(struct seq_file *seq, void *v)
1441 read_unlock(&udp_hash_lock);
1444 static int udp_seq_open(struct inode *inode, struct file *file)
1446 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
1447 struct seq_file *seq;
1449 struct udp_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
1453 memset(s, 0, sizeof(*s));
1454 s->family = afinfo->family;
1455 s->seq_ops.start = udp_seq_start;
1456 s->seq_ops.next = udp_seq_next;
1457 s->seq_ops.show = afinfo->seq_show;
1458 s->seq_ops.stop = udp_seq_stop;
1460 rc = seq_open(file, &s->seq_ops);
1464 seq = file->private_data;
1473 /* ------------------------------------------------------------------------ */
1474 int udp_proc_register(struct udp_seq_afinfo *afinfo)
1476 struct proc_dir_entry *p;
1481 afinfo->seq_fops->owner = afinfo->owner;
1482 afinfo->seq_fops->open = udp_seq_open;
1483 afinfo->seq_fops->read = seq_read;
1484 afinfo->seq_fops->llseek = seq_lseek;
1485 afinfo->seq_fops->release = seq_release_private;
1487 p = proc_net_fops_create(afinfo->name, S_IRUGO, afinfo->seq_fops);
1495 void udp_proc_unregister(struct udp_seq_afinfo *afinfo)
1499 proc_net_remove(afinfo->name);
1500 memset(afinfo->seq_fops, 0, sizeof(*afinfo->seq_fops));
1503 /* ------------------------------------------------------------------------ */
1504 static void udp4_format_sock(struct sock *sp, char *tmpbuf, int bucket)
1506 struct inet_opt *inet = inet_sk(sp);
1507 unsigned int dest = inet->daddr;
1508 unsigned int src = inet->rcv_saddr;
1509 __u16 destp = ntohs(inet->dport);
1510 __u16 srcp = ntohs(inet->sport);
1512 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
1513 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
1514 bucket, src, srcp, dest, destp, sp->sk_state,
1515 atomic_read(&sp->sk_wmem_alloc),
1516 atomic_read(&sp->sk_rmem_alloc),
1517 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
1518 atomic_read(&sp->sk_refcnt), sp);
1521 static int udp4_seq_show(struct seq_file *seq, void *v)
1523 if (v == SEQ_START_TOKEN)
1524 seq_printf(seq, "%-127s\n",
1525 " sl local_address rem_address st tx_queue "
1526 "rx_queue tr tm->when retrnsmt uid timeout "
1530 struct udp_iter_state *state = seq->private;
1532 udp4_format_sock(v, tmpbuf, state->bucket);
1533 seq_printf(seq, "%-127s\n", tmpbuf);
1538 /* ------------------------------------------------------------------------ */
1539 static struct file_operations udp4_seq_fops;
1540 static struct udp_seq_afinfo udp4_seq_afinfo = {
1541 .owner = THIS_MODULE,
1544 .seq_show = udp4_seq_show,
1545 .seq_fops = &udp4_seq_fops,
1548 int __init udp4_proc_init(void)
1550 return udp_proc_register(&udp4_seq_afinfo);
1553 void udp4_proc_exit(void)
1555 udp_proc_unregister(&udp4_seq_afinfo);
1557 #endif /* CONFIG_PROC_FS */
1559 EXPORT_SYMBOL(udp_connect);
1560 EXPORT_SYMBOL(udp_disconnect);
1561 EXPORT_SYMBOL(udp_hash);
1562 EXPORT_SYMBOL(udp_hash_lock);
1563 EXPORT_SYMBOL(udp_ioctl);
1564 EXPORT_SYMBOL(udp_port_rover);
1565 EXPORT_SYMBOL(udp_prot);
1566 EXPORT_SYMBOL(udp_sendmsg);
1568 #ifdef CONFIG_PROC_FS
1569 EXPORT_SYMBOL(udp_proc_register);
1570 EXPORT_SYMBOL(udp_proc_unregister);