1 // SPDX-License-Identifier: GPL-2.0-or-later
4 * Linux INET6 implementation
7 * Pedro Roque <roque@di.fc.ul.pt>
9 * Based on linux/ipv4/udp.c
12 * Hideaki YOSHIFUJI : sin6_scope_id support
13 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
14 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
15 * a single port at the same time.
16 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
17 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/udp6 to seq_file.
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/in6.h>
26 #include <linux/netdevice.h>
27 #include <linux/if_arp.h>
28 #include <linux/ipv6.h>
29 #include <linux/icmpv6.h>
30 #include <linux/init.h>
31 #include <linux/module.h>
32 #include <linux/skbuff.h>
33 #include <linux/slab.h>
34 #include <linux/uaccess.h>
35 #include <linux/indirect_call_wrapper.h>
37 #include <net/addrconf.h>
38 #include <net/ndisc.h>
39 #include <net/protocol.h>
40 #include <net/transp_v6.h>
41 #include <net/ip6_route.h>
44 #include <net/tcp_states.h>
45 #include <net/ip6_checksum.h>
46 #include <net/ip6_tunnel.h>
48 #include <net/inet_hashtables.h>
49 #include <net/inet6_hashtables.h>
50 #include <net/busy_poll.h>
51 #include <net/sock_reuseport.h>
53 #include <linux/proc_fs.h>
54 #include <linux/seq_file.h>
55 #include <trace/events/skb.h>
58 static u32 udp6_ehashfn(const struct net *net,
59 const struct in6_addr *laddr,
61 const struct in6_addr *faddr,
64 static u32 udp6_ehash_secret __read_mostly;
65 static u32 udp_ipv6_hash_secret __read_mostly;
69 net_get_random_once(&udp6_ehash_secret,
70 sizeof(udp6_ehash_secret));
71 net_get_random_once(&udp_ipv6_hash_secret,
72 sizeof(udp_ipv6_hash_secret));
74 lhash = (__force u32)laddr->s6_addr32[3];
75 fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret);
77 return __inet6_ehashfn(lhash, lport, fhash, fport,
78 udp_ipv6_hash_secret + net_hash_mix(net));
81 int udp_v6_get_port(struct sock *sk, unsigned short snum)
83 unsigned int hash2_nulladdr =
84 ipv6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
85 unsigned int hash2_partial =
86 ipv6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
88 /* precompute partial secondary hash */
89 udp_sk(sk)->udp_portaddr_hash = hash2_partial;
90 return udp_lib_get_port(sk, snum, hash2_nulladdr);
93 void udp_v6_rehash(struct sock *sk)
95 u16 new_hash = ipv6_portaddr_hash(sock_net(sk),
97 inet_sk(sk)->inet_num);
99 udp_lib_rehash(sk, new_hash);
102 static int compute_score(struct sock *sk, struct net *net,
103 const struct in6_addr *saddr, __be16 sport,
104 const struct in6_addr *daddr, unsigned short hnum,
108 struct inet_sock *inet;
111 if (!net_eq(sock_net(sk), net) ||
112 udp_sk(sk)->udp_port_hash != hnum ||
113 sk->sk_family != PF_INET6)
116 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
122 if (inet->inet_dport) {
123 if (inet->inet_dport != sport)
128 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
129 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
134 dev_match = udp_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif);
137 if (sk->sk_bound_dev_if)
140 if (READ_ONCE(sk->sk_incoming_cpu) == raw_smp_processor_id())
146 static struct sock *lookup_reuseport(struct net *net, struct sock *sk,
148 const struct in6_addr *saddr,
150 const struct in6_addr *daddr,
153 struct sock *reuse_sk = NULL;
156 if (sk->sk_reuseport && sk->sk_state != TCP_ESTABLISHED) {
157 hash = udp6_ehashfn(net, daddr, hnum, saddr, sport);
158 reuse_sk = reuseport_select_sock(sk, hash, skb,
159 sizeof(struct udphdr));
164 /* called with rcu_read_lock() */
165 static struct sock *udp6_lib_lookup2(struct net *net,
166 const struct in6_addr *saddr, __be16 sport,
167 const struct in6_addr *daddr, unsigned int hnum,
168 int dif, int sdif, struct udp_hslot *hslot2,
171 struct sock *sk, *result;
176 udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
177 score = compute_score(sk, net, saddr, sport,
178 daddr, hnum, dif, sdif);
179 if (score > badness) {
180 result = lookup_reuseport(net, sk, skb,
181 saddr, sport, daddr, hnum);
182 /* Fall back to scoring if group has connections */
183 if (result && !reuseport_has_conns(sk))
186 result = result ? : sk;
193 static inline struct sock *udp6_lookup_run_bpf(struct net *net,
194 struct udp_table *udptable,
196 const struct in6_addr *saddr,
198 const struct in6_addr *daddr,
201 struct sock *sk, *reuse_sk;
204 if (udptable != &udp_table)
205 return NULL; /* only UDP is supported */
207 no_reuseport = bpf_sk_lookup_run_v6(net, IPPROTO_UDP,
208 saddr, sport, daddr, hnum, &sk);
209 if (no_reuseport || IS_ERR_OR_NULL(sk))
212 reuse_sk = lookup_reuseport(net, sk, skb, saddr, sport, daddr, hnum);
218 /* rcu_read_lock() must be held */
219 struct sock *__udp6_lib_lookup(struct net *net,
220 const struct in6_addr *saddr, __be16 sport,
221 const struct in6_addr *daddr, __be16 dport,
222 int dif, int sdif, struct udp_table *udptable,
225 unsigned short hnum = ntohs(dport);
226 unsigned int hash2, slot2;
227 struct udp_hslot *hslot2;
228 struct sock *result, *sk;
230 hash2 = ipv6_portaddr_hash(net, daddr, hnum);
231 slot2 = hash2 & udptable->mask;
232 hslot2 = &udptable->hash2[slot2];
234 /* Lookup connected or non-wildcard sockets */
235 result = udp6_lib_lookup2(net, saddr, sport,
236 daddr, hnum, dif, sdif,
238 if (!IS_ERR_OR_NULL(result) && result->sk_state == TCP_ESTABLISHED)
241 /* Lookup redirect from BPF */
242 if (static_branch_unlikely(&bpf_sk_lookup_enabled)) {
243 sk = udp6_lookup_run_bpf(net, udptable, skb,
244 saddr, sport, daddr, hnum);
251 /* Got non-wildcard socket or error on first lookup */
255 /* Lookup wildcard sockets */
256 hash2 = ipv6_portaddr_hash(net, &in6addr_any, hnum);
257 slot2 = hash2 & udptable->mask;
258 hslot2 = &udptable->hash2[slot2];
260 result = udp6_lib_lookup2(net, saddr, sport,
261 &in6addr_any, hnum, dif, sdif,
268 EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
270 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
271 __be16 sport, __be16 dport,
272 struct udp_table *udptable)
274 const struct ipv6hdr *iph = ipv6_hdr(skb);
276 return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport,
277 &iph->daddr, dport, inet6_iif(skb),
278 inet6_sdif(skb), udptable, skb);
281 struct sock *udp6_lib_lookup_skb(const struct sk_buff *skb,
282 __be16 sport, __be16 dport)
284 const struct ipv6hdr *iph = ipv6_hdr(skb);
286 return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport,
287 &iph->daddr, dport, inet6_iif(skb),
288 inet6_sdif(skb), &udp_table, NULL);
291 /* Must be called under rcu_read_lock().
292 * Does increment socket refcount.
294 #if IS_ENABLED(CONFIG_NF_TPROXY_IPV6) || IS_ENABLED(CONFIG_NF_SOCKET_IPV6)
295 struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
296 const struct in6_addr *daddr, __be16 dport, int dif)
300 sk = __udp6_lib_lookup(net, saddr, sport, daddr, dport,
301 dif, 0, &udp_table, NULL);
302 if (sk && !refcount_inc_not_zero(&sk->sk_refcnt))
306 EXPORT_SYMBOL_GPL(udp6_lib_lookup);
309 /* do not use the scratch area len for jumbogram: their length execeeds the
310 * scratch area space; note that the IP6CB flags is still in the first
311 * cacheline, so checking for jumbograms is cheap
313 static int udp6_skb_len(struct sk_buff *skb)
315 return unlikely(inet6_is_jumbogram(skb)) ? skb->len : udp_skb_len(skb);
319 * This should be easy, if there is something there we
320 * return it, otherwise we block.
323 int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
324 int noblock, int flags, int *addr_len)
326 struct ipv6_pinfo *np = inet6_sk(sk);
327 struct inet_sock *inet = inet_sk(sk);
329 unsigned int ulen, copied;
330 int off, err, peeking = flags & MSG_PEEK;
331 int is_udplite = IS_UDPLITE(sk);
332 struct udp_mib __percpu *mib;
333 bool checksum_valid = false;
336 if (flags & MSG_ERRQUEUE)
337 return ipv6_recv_error(sk, msg, len, addr_len);
339 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
340 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
343 off = sk_peek_offset(sk, flags);
344 skb = __skb_recv_udp(sk, flags, noblock, &off, &err);
348 ulen = udp6_skb_len(skb);
350 if (copied > ulen - off)
352 else if (copied < ulen)
353 msg->msg_flags |= MSG_TRUNC;
355 is_udp4 = (skb->protocol == htons(ETH_P_IP));
356 mib = __UDPX_MIB(sk, is_udp4);
359 * If checksum is needed at all, try to do it while copying the
360 * data. If the data is truncated, or if we only want a partial
361 * coverage checksum (UDP-Lite), do it before the copy.
364 if (copied < ulen || peeking ||
365 (is_udplite && UDP_SKB_CB(skb)->partial_cov)) {
366 checksum_valid = udp_skb_csum_unnecessary(skb) ||
367 !__udp_lib_checksum_complete(skb);
372 if (checksum_valid || udp_skb_csum_unnecessary(skb)) {
373 if (udp_skb_is_linear(skb))
374 err = copy_linear_skb(skb, copied, off, &msg->msg_iter);
376 err = skb_copy_datagram_msg(skb, off, msg, copied);
378 err = skb_copy_and_csum_datagram_msg(skb, off, msg);
384 atomic_inc(&sk->sk_drops);
385 SNMP_INC_STATS(mib, UDP_MIB_INERRORS);
391 SNMP_INC_STATS(mib, UDP_MIB_INDATAGRAMS);
393 sock_recv_ts_and_drops(msg, sk, skb);
395 /* Copy the address. */
397 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
398 sin6->sin6_family = AF_INET6;
399 sin6->sin6_port = udp_hdr(skb)->source;
400 sin6->sin6_flowinfo = 0;
403 ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
405 sin6->sin6_scope_id = 0;
407 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
408 sin6->sin6_scope_id =
409 ipv6_iface_scope_id(&sin6->sin6_addr,
412 *addr_len = sizeof(*sin6);
414 BPF_CGROUP_RUN_PROG_UDP6_RECVMSG_LOCK(sk,
415 (struct sockaddr *)sin6);
418 if (udp_sk(sk)->gro_enabled)
419 udp_cmsg_recv(msg, sk, skb);
422 ip6_datagram_recv_common_ctl(sk, msg, skb);
425 if (inet->cmsg_flags)
426 ip_cmsg_recv_offset(msg, sk, skb,
427 sizeof(struct udphdr), off);
430 ip6_datagram_recv_specific_ctl(sk, msg, skb);
434 if (flags & MSG_TRUNC)
437 skb_consume_udp(sk, skb, peeking ? -err : err);
441 if (!__sk_queue_drop_skb(sk, &udp_sk(sk)->reader_queue, skb, flags,
442 udp_skb_destructor)) {
443 SNMP_INC_STATS(mib, UDP_MIB_CSUMERRORS);
444 SNMP_INC_STATS(mib, UDP_MIB_INERRORS);
448 /* starting over for a new packet, but check if we need to yield */
450 msg->msg_flags &= ~MSG_TRUNC;
454 DEFINE_STATIC_KEY_FALSE(udpv6_encap_needed_key);
455 void udpv6_encap_enable(void)
457 static_branch_inc(&udpv6_encap_needed_key);
459 EXPORT_SYMBOL(udpv6_encap_enable);
461 /* Handler for tunnels with arbitrary destination ports: no socket lookup, go
462 * through error handlers in encapsulations looking for a match.
464 static int __udp6_lib_err_encap_no_sk(struct sk_buff *skb,
465 struct inet6_skb_parm *opt,
466 u8 type, u8 code, int offset, __be32 info)
470 for (i = 0; i < MAX_IPTUN_ENCAP_OPS; i++) {
471 int (*handler)(struct sk_buff *skb, struct inet6_skb_parm *opt,
472 u8 type, u8 code, int offset, __be32 info);
473 const struct ip6_tnl_encap_ops *encap;
475 encap = rcu_dereference(ip6tun_encaps[i]);
478 handler = encap->err_handler;
479 if (handler && !handler(skb, opt, type, code, offset, info))
486 /* Try to match ICMP errors to UDP tunnels by looking up a socket without
487 * reversing source and destination port: this will match tunnels that force the
488 * same destination port on both endpoints (e.g. VXLAN, GENEVE). Note that
489 * lwtunnels might actually break this assumption by being configured with
490 * different destination ports on endpoints, in this case we won't be able to
491 * trace ICMP messages back to them.
493 * If this doesn't match any socket, probe tunnels with arbitrary destination
494 * ports (e.g. FoU, GUE): there, the receiving socket is useless, as the port
495 * we've sent packets to won't necessarily match the local destination port.
497 * Then ask the tunnel implementation to match the error against a valid
500 * Return an error if we can't find a match, the socket if we need further
501 * processing, zero otherwise.
503 static struct sock *__udp6_lib_err_encap(struct net *net,
504 const struct ipv6hdr *hdr, int offset,
506 struct udp_table *udptable,
509 struct inet6_skb_parm *opt,
510 u8 type, u8 code, __be32 info)
512 int (*lookup)(struct sock *sk, struct sk_buff *skb);
513 int network_offset, transport_offset;
516 network_offset = skb_network_offset(skb);
517 transport_offset = skb_transport_offset(skb);
519 /* Network header needs to point to the outer IPv6 header inside ICMP */
520 skb_reset_network_header(skb);
522 /* Transport header needs to point to the UDP header */
523 skb_set_transport_header(skb, offset);
528 lookup = READ_ONCE(up->encap_err_lookup);
529 if (lookup && lookup(sk, skb))
535 sk = __udp6_lib_lookup(net, &hdr->daddr, uh->source,
536 &hdr->saddr, uh->dest,
537 inet6_iif(skb), 0, udptable, skb);
541 lookup = READ_ONCE(up->encap_err_lookup);
542 if (!lookup || lookup(sk, skb))
548 sk = ERR_PTR(__udp6_lib_err_encap_no_sk(skb, opt, type, code,
552 skb_set_transport_header(skb, transport_offset);
553 skb_set_network_header(skb, network_offset);
558 int __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
559 u8 type, u8 code, int offset, __be32 info,
560 struct udp_table *udptable)
562 struct ipv6_pinfo *np;
563 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
564 const struct in6_addr *saddr = &hdr->saddr;
565 const struct in6_addr *daddr = seg6_get_daddr(skb, opt) ? : &hdr->daddr;
566 struct udphdr *uh = (struct udphdr *)(skb->data+offset);
571 struct net *net = dev_net(skb->dev);
573 sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source,
574 inet6_iif(skb), inet6_sdif(skb), udptable, NULL);
576 if (!sk || udp_sk(sk)->encap_type) {
577 /* No socket for error: try tunnels before discarding */
578 if (static_branch_unlikely(&udpv6_encap_needed_key)) {
579 sk = __udp6_lib_err_encap(net, hdr, offset, uh,
581 opt, type, code, info);
585 sk = ERR_PTR(-ENOENT);
588 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
596 harderr = icmpv6_err_convert(type, code, &err);
599 if (type == ICMPV6_PKT_TOOBIG) {
600 if (!ip6_sk_accept_pmtu(sk))
602 ip6_sk_update_pmtu(skb, sk, info);
603 if (np->pmtudisc != IPV6_PMTUDISC_DONT)
606 if (type == NDISC_REDIRECT) {
608 ip6_redirect(skb, sock_net(sk), inet6_iif(skb),
609 sk->sk_mark, sk->sk_uid);
611 ip6_sk_redirect(skb, sk);
616 /* Tunnels don't have an application socket: don't pass errors back */
618 if (udp_sk(sk)->encap_err_rcv)
619 udp_sk(sk)->encap_err_rcv(sk, skb, offset);
624 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
627 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
636 static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
640 if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
641 sock_rps_save_rxhash(sk, skb);
642 sk_mark_napi_id(sk, skb);
643 sk_incoming_cpu_update(sk);
645 sk_mark_napi_id_once(sk, skb);
648 rc = __udp_enqueue_schedule_skb(sk, skb);
650 int is_udplite = IS_UDPLITE(sk);
652 /* Note that an ENOMEM error is charged twice */
654 UDP6_INC_STATS(sock_net(sk),
655 UDP_MIB_RCVBUFERRORS, is_udplite);
657 UDP6_INC_STATS(sock_net(sk),
658 UDP_MIB_MEMERRORS, is_udplite);
659 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
667 static __inline__ int udpv6_err(struct sk_buff *skb,
668 struct inet6_skb_parm *opt, u8 type,
669 u8 code, int offset, __be32 info)
671 return __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
674 static int udpv6_queue_rcv_one_skb(struct sock *sk, struct sk_buff *skb)
676 struct udp_sock *up = udp_sk(sk);
677 int is_udplite = IS_UDPLITE(sk);
679 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
682 if (static_branch_unlikely(&udpv6_encap_needed_key) && up->encap_type) {
683 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
686 * This is an encapsulation socket so pass the skb to
687 * the socket's udp_encap_rcv() hook. Otherwise, just
688 * fall through and pass this up the UDP socket.
689 * up->encap_rcv() returns the following value:
690 * =0 if skb was successfully passed to the encap
691 * handler or was discarded by it.
692 * >0 if skb should be passed on to UDP.
693 * <0 if skb should be resubmitted as proto -N
696 /* if we're overly short, let UDP handle it */
697 encap_rcv = READ_ONCE(up->encap_rcv);
701 /* Verify checksum before giving to encap */
702 if (udp_lib_checksum_complete(skb))
705 ret = encap_rcv(sk, skb);
707 __UDP_INC_STATS(sock_net(sk),
714 /* FALLTHROUGH -- it's a UDP Packet */
718 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
720 if ((up->pcflag & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
722 if (up->pcrlen == 0) { /* full coverage was set */
723 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
724 UDP_SKB_CB(skb)->cscov, skb->len);
727 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
728 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
729 UDP_SKB_CB(skb)->cscov, up->pcrlen);
734 prefetch(&sk->sk_rmem_alloc);
735 if (rcu_access_pointer(sk->sk_filter) &&
736 udp_lib_checksum_complete(skb))
739 if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr)))
742 udp_csum_pull_header(skb);
746 return __udpv6_queue_rcv_skb(sk, skb);
749 __UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
751 __UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
752 atomic_inc(&sk->sk_drops);
757 static int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
759 struct sk_buff *next, *segs;
762 if (likely(!udp_unexpected_gso(sk, skb)))
763 return udpv6_queue_rcv_one_skb(sk, skb);
765 __skb_push(skb, -skb_mac_offset(skb));
766 segs = udp_rcv_segment(sk, skb, false);
767 skb_list_walk_safe(segs, skb, next) {
768 __skb_pull(skb, skb_transport_offset(skb));
770 udp_post_segment_fix_csum(skb);
771 ret = udpv6_queue_rcv_one_skb(sk, skb);
773 ip6_protocol_deliver_rcu(dev_net(skb->dev), skb, ret,
779 static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
780 __be16 loc_port, const struct in6_addr *loc_addr,
781 __be16 rmt_port, const struct in6_addr *rmt_addr,
782 int dif, int sdif, unsigned short hnum)
784 struct inet_sock *inet = inet_sk(sk);
786 if (!net_eq(sock_net(sk), net))
789 if (udp_sk(sk)->udp_port_hash != hnum ||
790 sk->sk_family != PF_INET6 ||
791 (inet->inet_dport && inet->inet_dport != rmt_port) ||
792 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
793 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
794 !udp_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif) ||
795 (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
796 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
798 if (!inet6_mc_check(sk, loc_addr, rmt_addr))
803 static void udp6_csum_zero_error(struct sk_buff *skb)
805 /* RFC 2460 section 8.1 says that we SHOULD log
806 * this error. Well, it is reasonable.
808 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
809 &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
810 &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
814 * Note: called only from the BH handler context,
815 * so we don't need to lock the hashes.
817 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
818 const struct in6_addr *saddr, const struct in6_addr *daddr,
819 struct udp_table *udptable, int proto)
821 struct sock *sk, *first = NULL;
822 const struct udphdr *uh = udp_hdr(skb);
823 unsigned short hnum = ntohs(uh->dest);
824 struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
825 unsigned int offset = offsetof(typeof(*sk), sk_node);
826 unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
827 int dif = inet6_iif(skb);
828 int sdif = inet6_sdif(skb);
829 struct hlist_node *node;
830 struct sk_buff *nskb;
833 hash2_any = ipv6_portaddr_hash(net, &in6addr_any, hnum) &
835 hash2 = ipv6_portaddr_hash(net, daddr, hnum) & udptable->mask;
837 hslot = &udptable->hash2[hash2];
838 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
841 sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
842 if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr,
843 uh->source, saddr, dif, sdif,
846 /* If zero checksum and no_check is not on for
847 * the socket then skip it.
849 if (!uh->check && !udp_sk(sk)->no_check6_rx)
855 nskb = skb_clone(skb, GFP_ATOMIC);
856 if (unlikely(!nskb)) {
857 atomic_inc(&sk->sk_drops);
858 __UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
860 __UDP6_INC_STATS(net, UDP_MIB_INERRORS,
865 if (udpv6_queue_rcv_skb(sk, nskb) > 0)
869 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
870 if (use_hash2 && hash2 != hash2_any) {
876 if (udpv6_queue_rcv_skb(first, skb) > 0)
880 __UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI,
881 proto == IPPROTO_UDPLITE);
886 static void udp6_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst)
888 if (udp_sk_rx_dst_set(sk, dst)) {
889 const struct rt6_info *rt = (const struct rt6_info *)dst;
891 sk->sk_rx_dst_cookie = rt6_get_cookie(rt);
895 /* wrapper for udp_queue_rcv_skb tacking care of csum conversion and
896 * return code conversion for ip layer consumption
898 static int udp6_unicast_rcv_skb(struct sock *sk, struct sk_buff *skb,
903 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
904 skb_checksum_try_convert(skb, IPPROTO_UDP, ip6_compute_pseudo);
906 ret = udpv6_queue_rcv_skb(sk, skb);
908 /* a return value > 0 means to resubmit the input */
914 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
917 const struct in6_addr *saddr, *daddr;
918 struct net *net = dev_net(skb->dev);
924 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
927 saddr = &ipv6_hdr(skb)->saddr;
928 daddr = &ipv6_hdr(skb)->daddr;
931 ulen = ntohs(uh->len);
935 if (proto == IPPROTO_UDP) {
936 /* UDP validates ulen. */
938 /* Check for jumbo payload */
942 if (ulen < sizeof(*uh))
945 if (ulen < skb->len) {
946 if (pskb_trim_rcsum(skb, ulen))
948 saddr = &ipv6_hdr(skb)->saddr;
949 daddr = &ipv6_hdr(skb)->daddr;
954 if (udp6_csum_init(skb, uh, proto))
957 /* Check if the socket is already available, e.g. due to early demux */
958 sk = skb_steal_sock(skb, &refcounted);
960 struct dst_entry *dst = skb_dst(skb);
963 if (unlikely(rcu_dereference(sk->sk_rx_dst) != dst))
964 udp6_sk_rx_dst_set(sk, dst);
966 if (!uh->check && !udp_sk(sk)->no_check6_rx) {
969 goto report_csum_error;
972 ret = udp6_unicast_rcv_skb(sk, skb, uh);
979 * Multicast receive code
981 if (ipv6_addr_is_multicast(daddr))
982 return __udp6_lib_mcast_deliver(net, skb,
983 saddr, daddr, udptable, proto);
986 sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
988 if (!uh->check && !udp_sk(sk)->no_check6_rx)
989 goto report_csum_error;
990 return udp6_unicast_rcv_skb(sk, skb, uh);
994 goto report_csum_error;
996 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
999 if (udp_lib_checksum_complete(skb))
1002 __UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
1003 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
1009 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
1010 proto == IPPROTO_UDPLITE ? "-Lite" : "",
1011 saddr, ntohs(uh->source),
1013 daddr, ntohs(uh->dest));
1017 udp6_csum_zero_error(skb);
1019 __UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
1021 __UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1027 static struct sock *__udp6_lib_demux_lookup(struct net *net,
1028 __be16 loc_port, const struct in6_addr *loc_addr,
1029 __be16 rmt_port, const struct in6_addr *rmt_addr,
1032 unsigned short hnum = ntohs(loc_port);
1033 unsigned int hash2 = ipv6_portaddr_hash(net, loc_addr, hnum);
1034 unsigned int slot2 = hash2 & udp_table.mask;
1035 struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
1036 const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
1039 udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
1040 if (sk->sk_state == TCP_ESTABLISHED &&
1041 inet6_match(net, sk, rmt_addr, loc_addr, ports, dif, sdif))
1043 /* Only check first socket in chain */
1049 void udp_v6_early_demux(struct sk_buff *skb)
1051 struct net *net = dev_net(skb->dev);
1052 const struct udphdr *uh;
1054 struct dst_entry *dst;
1055 int dif = skb->dev->ifindex;
1056 int sdif = inet6_sdif(skb);
1058 if (!pskb_may_pull(skb, skb_transport_offset(skb) +
1059 sizeof(struct udphdr)))
1064 if (skb->pkt_type == PACKET_HOST)
1065 sk = __udp6_lib_demux_lookup(net, uh->dest,
1066 &ipv6_hdr(skb)->daddr,
1067 uh->source, &ipv6_hdr(skb)->saddr,
1072 if (!sk || !refcount_inc_not_zero(&sk->sk_refcnt))
1076 skb->destructor = sock_efree;
1077 dst = rcu_dereference(sk->sk_rx_dst);
1080 dst = dst_check(dst, sk->sk_rx_dst_cookie);
1082 /* set noref for now.
1083 * any place which wants to hold dst has to call
1086 skb_dst_set_noref(skb, dst);
1090 INDIRECT_CALLABLE_SCOPE int udpv6_rcv(struct sk_buff *skb)
1092 return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
1096 * Throw away all pending data and cancel the corking. Socket is locked.
1098 static void udp_v6_flush_pending_frames(struct sock *sk)
1100 struct udp_sock *up = udp_sk(sk);
1102 if (up->pending == AF_INET)
1103 udp_flush_pending_frames(sk);
1104 else if (up->pending) {
1107 ip6_flush_pending_frames(sk);
1111 static int udpv6_pre_connect(struct sock *sk, struct sockaddr *uaddr,
1114 if (addr_len < offsetofend(struct sockaddr, sa_family))
1116 /* The following checks are replicated from __ip6_datagram_connect()
1117 * and intended to prevent BPF program called below from accessing
1118 * bytes that are out of the bound specified by user in addr_len.
1120 if (uaddr->sa_family == AF_INET) {
1121 if (__ipv6_only_sock(sk))
1122 return -EAFNOSUPPORT;
1123 return udp_pre_connect(sk, uaddr, addr_len);
1126 if (addr_len < SIN6_LEN_RFC2133)
1129 return BPF_CGROUP_RUN_PROG_INET6_CONNECT_LOCK(sk, uaddr);
1133 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
1134 * @sk: socket we are sending on
1135 * @skb: sk_buff containing the filled-in UDP header
1136 * (checksum field must be zeroed out)
1137 * @saddr: source address
1138 * @daddr: destination address
1139 * @len: length of packet
1141 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
1142 const struct in6_addr *saddr,
1143 const struct in6_addr *daddr, int len)
1145 unsigned int offset;
1146 struct udphdr *uh = udp_hdr(skb);
1147 struct sk_buff *frags = skb_shinfo(skb)->frag_list;
1151 /* Only one fragment on the socket. */
1152 skb->csum_start = skb_transport_header(skb) - skb->head;
1153 skb->csum_offset = offsetof(struct udphdr, check);
1154 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
1157 * HW-checksum won't work as there are two or more
1158 * fragments on the socket so that all csums of sk_buffs
1159 * should be together
1161 offset = skb_transport_offset(skb);
1162 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
1165 skb->ip_summed = CHECKSUM_NONE;
1168 csum = csum_add(csum, frags->csum);
1169 } while ((frags = frags->next));
1171 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
1174 uh->check = CSUM_MANGLED_0;
1182 static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6,
1183 struct inet_cork *cork)
1185 struct sock *sk = skb->sk;
1188 int is_udplite = IS_UDPLITE(sk);
1190 int offset = skb_transport_offset(skb);
1191 int len = skb->len - offset;
1192 int datalen = len - sizeof(*uh);
1195 * Create a UDP header
1198 uh->source = fl6->fl6_sport;
1199 uh->dest = fl6->fl6_dport;
1200 uh->len = htons(len);
1203 if (cork->gso_size) {
1204 const int hlen = skb_network_header_len(skb) +
1205 sizeof(struct udphdr);
1207 if (hlen + cork->gso_size > cork->fragsize) {
1211 if (datalen > cork->gso_size * UDP_MAX_SEGMENTS) {
1215 if (udp_sk(sk)->no_check6_tx) {
1219 if (skb->ip_summed != CHECKSUM_PARTIAL || is_udplite ||
1220 dst_xfrm(skb_dst(skb))) {
1225 if (datalen > cork->gso_size) {
1226 skb_shinfo(skb)->gso_size = cork->gso_size;
1227 skb_shinfo(skb)->gso_type = SKB_GSO_UDP_L4;
1228 skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(datalen,
1235 csum = udplite_csum(skb);
1236 else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */
1237 skb->ip_summed = CHECKSUM_NONE;
1239 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
1241 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
1244 csum = udp_csum(skb);
1246 /* add protocol-dependent pseudo-header */
1247 uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
1248 len, fl6->flowi6_proto, csum);
1250 uh->check = CSUM_MANGLED_0;
1253 err = ip6_send_skb(skb);
1255 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
1256 UDP6_INC_STATS(sock_net(sk),
1257 UDP_MIB_SNDBUFERRORS, is_udplite);
1261 UDP6_INC_STATS(sock_net(sk),
1262 UDP_MIB_OUTDATAGRAMS, is_udplite);
1267 static int udp_v6_push_pending_frames(struct sock *sk)
1269 struct sk_buff *skb;
1270 struct udp_sock *up = udp_sk(sk);
1274 if (up->pending == AF_INET)
1275 return udp_push_pending_frames(sk);
1277 /* ip6_finish_skb will release the cork, so make a copy of
1280 fl6 = inet_sk(sk)->cork.fl.u.ip6;
1282 skb = ip6_finish_skb(sk);
1286 err = udp_v6_send_skb(skb, &fl6, &inet_sk(sk)->cork.base);
1294 int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1296 struct ipv6_txoptions opt_space;
1297 struct udp_sock *up = udp_sk(sk);
1298 struct inet_sock *inet = inet_sk(sk);
1299 struct ipv6_pinfo *np = inet6_sk(sk);
1300 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1301 struct in6_addr *daddr, *final_p, final;
1302 struct ipv6_txoptions *opt = NULL;
1303 struct ipv6_txoptions *opt_to_free = NULL;
1304 struct ip6_flowlabel *flowlabel = NULL;
1306 struct dst_entry *dst;
1307 struct ipcm6_cookie ipc6;
1308 int addr_len = msg->msg_namelen;
1309 bool connected = false;
1311 int corkreq = READ_ONCE(up->corkflag) || msg->msg_flags&MSG_MORE;
1313 int is_udplite = IS_UDPLITE(sk);
1314 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1317 ipc6.gso_size = READ_ONCE(up->gso_size);
1318 ipc6.sockc.tsflags = sk->sk_tsflags;
1319 ipc6.sockc.mark = sk->sk_mark;
1321 /* destination address check */
1323 if (addr_len < offsetof(struct sockaddr, sa_data))
1326 switch (sin6->sin6_family) {
1328 if (addr_len < SIN6_LEN_RFC2133)
1330 daddr = &sin6->sin6_addr;
1331 if (ipv6_addr_any(daddr) &&
1332 ipv6_addr_v4mapped(&np->saddr))
1333 ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK),
1337 goto do_udp_sendmsg;
1339 msg->msg_name = sin6 = NULL;
1340 msg->msg_namelen = addr_len = 0;
1346 } else if (!up->pending) {
1347 if (sk->sk_state != TCP_ESTABLISHED)
1348 return -EDESTADDRREQ;
1349 daddr = &sk->sk_v6_daddr;
1354 if (ipv6_addr_v4mapped(daddr)) {
1355 struct sockaddr_in sin;
1356 sin.sin_family = AF_INET;
1357 sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1358 sin.sin_addr.s_addr = daddr->s6_addr32[3];
1359 msg->msg_name = &sin;
1360 msg->msg_namelen = sizeof(sin);
1362 if (__ipv6_only_sock(sk))
1363 return -ENETUNREACH;
1364 return udp_sendmsg(sk, msg, len);
1368 if (up->pending == AF_INET)
1369 return udp_sendmsg(sk, msg, len);
1371 /* Rough check on arithmetic overflow,
1372 better check is made in ip6_append_data().
1374 if (len > INT_MAX - sizeof(struct udphdr))
1377 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
1380 * There are pending frames.
1381 * The socket lock must be held while it's corked.
1384 if (likely(up->pending)) {
1385 if (unlikely(up->pending != AF_INET6)) {
1387 return -EAFNOSUPPORT;
1390 goto do_append_data;
1394 ulen += sizeof(struct udphdr);
1396 memset(&fl6, 0, sizeof(fl6));
1399 if (sin6->sin6_port == 0)
1402 fl6.fl6_dport = sin6->sin6_port;
1403 daddr = &sin6->sin6_addr;
1406 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1407 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1408 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1409 if (IS_ERR(flowlabel))
1415 * Otherwise it will be difficult to maintain
1418 if (sk->sk_state == TCP_ESTABLISHED &&
1419 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1420 daddr = &sk->sk_v6_daddr;
1422 if (addr_len >= sizeof(struct sockaddr_in6) &&
1423 sin6->sin6_scope_id &&
1424 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1425 fl6.flowi6_oif = sin6->sin6_scope_id;
1427 if (sk->sk_state != TCP_ESTABLISHED)
1428 return -EDESTADDRREQ;
1430 fl6.fl6_dport = inet->inet_dport;
1431 daddr = &sk->sk_v6_daddr;
1432 fl6.flowlabel = np->flow_label;
1436 if (!fl6.flowi6_oif)
1437 fl6.flowi6_oif = sk->sk_bound_dev_if;
1439 if (!fl6.flowi6_oif)
1440 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1442 fl6.flowi6_uid = sk->sk_uid;
1444 if (msg->msg_controllen) {
1446 memset(opt, 0, sizeof(struct ipv6_txoptions));
1447 opt->tot_len = sizeof(*opt);
1450 err = udp_cmsg_send(sk, msg, &ipc6.gso_size);
1452 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6,
1455 fl6_sock_release(flowlabel);
1458 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1459 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1460 if (IS_ERR(flowlabel))
1463 if (!(opt->opt_nflen|opt->opt_flen))
1468 opt = txopt_get(np);
1472 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1473 opt = ipv6_fixup_options(&opt_space, opt);
1476 fl6.flowi6_proto = sk->sk_protocol;
1477 fl6.flowi6_mark = ipc6.sockc.mark;
1479 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1480 fl6.saddr = np->saddr;
1481 fl6.fl6_sport = inet->inet_sport;
1483 if (cgroup_bpf_enabled(CGROUP_UDP6_SENDMSG) && !connected) {
1484 err = BPF_CGROUP_RUN_PROG_UDP6_SENDMSG_LOCK(sk,
1485 (struct sockaddr *)sin6, &fl6.saddr);
1489 if (ipv6_addr_v4mapped(&sin6->sin6_addr)) {
1490 /* BPF program rewrote IPv6-only by IPv4-mapped
1491 * IPv6. It's currently unsupported.
1496 if (sin6->sin6_port == 0) {
1497 /* BPF program set invalid port. Reject it. */
1501 fl6.fl6_dport = sin6->sin6_port;
1502 fl6.daddr = sin6->sin6_addr;
1506 if (ipv6_addr_any(&fl6.daddr))
1507 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1509 final_p = fl6_update_dst(&fl6, opt, &final);
1513 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1514 fl6.flowi6_oif = np->mcast_oif;
1516 } else if (!fl6.flowi6_oif)
1517 fl6.flowi6_oif = np->ucast_oif;
1519 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
1521 if (ipc6.tclass < 0)
1522 ipc6.tclass = np->tclass;
1524 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
1526 dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, connected);
1533 if (ipc6.hlimit < 0)
1534 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1536 if (msg->msg_flags&MSG_CONFIRM)
1540 /* Lockless fast path for the non-corking case */
1542 struct inet_cork_full cork;
1543 struct sk_buff *skb;
1545 skb = ip6_make_skb(sk, getfrag, msg, ulen,
1546 sizeof(struct udphdr), &ipc6,
1547 &fl6, (struct rt6_info *)dst,
1548 msg->msg_flags, &cork);
1550 if (!IS_ERR_OR_NULL(skb))
1551 err = udp_v6_send_skb(skb, &fl6, &cork.base);
1556 if (unlikely(up->pending)) {
1557 /* The socket is already corked while preparing it. */
1558 /* ... which is an evident application bug. --ANK */
1561 net_dbg_ratelimited("udp cork app bug 2\n");
1566 up->pending = AF_INET6;
1569 if (ipc6.dontfrag < 0)
1570 ipc6.dontfrag = np->dontfrag;
1572 err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr),
1573 &ipc6, &fl6, (struct rt6_info *)dst,
1574 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
1576 udp_v6_flush_pending_frames(sk);
1578 err = udp_v6_push_pending_frames(sk);
1579 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1583 err = np->recverr ? net_xmit_errno(err) : 0;
1589 fl6_sock_release(flowlabel);
1590 txopt_put(opt_to_free);
1594 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1595 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1596 * we don't have a good statistic (IpOutDiscards but it can be too many
1597 * things). We could add another new stat but at least for now that
1598 * seems like overkill.
1600 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1601 UDP6_INC_STATS(sock_net(sk),
1602 UDP_MIB_SNDBUFERRORS, is_udplite);
1607 if (msg->msg_flags & MSG_PROBE)
1608 dst_confirm_neigh(dst, &fl6.daddr);
1609 if (!(msg->msg_flags&MSG_PROBE) || len)
1610 goto back_from_confirm;
1615 void udpv6_destroy_sock(struct sock *sk)
1617 struct udp_sock *up = udp_sk(sk);
1620 /* protects from races with udp_abort() */
1621 sock_set_flag(sk, SOCK_DEAD);
1622 udp_v6_flush_pending_frames(sk);
1625 if (static_branch_unlikely(&udpv6_encap_needed_key)) {
1626 if (up->encap_type) {
1627 void (*encap_destroy)(struct sock *sk);
1628 encap_destroy = READ_ONCE(up->encap_destroy);
1632 if (up->encap_enabled) {
1633 static_branch_dec(&udpv6_encap_needed_key);
1634 udp_encap_disable();
1638 inet6_destroy_sock(sk);
1642 * Socket option code for UDP
1644 int udpv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1645 unsigned int optlen)
1647 if (level == SOL_UDP || level == SOL_UDPLITE)
1648 return udp_lib_setsockopt(sk, level, optname,
1650 udp_v6_push_pending_frames);
1651 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1654 int udpv6_getsockopt(struct sock *sk, int level, int optname,
1655 char __user *optval, int __user *optlen)
1657 if (level == SOL_UDP || level == SOL_UDPLITE)
1658 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1659 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1662 static const struct inet6_protocol udpv6_protocol = {
1663 .handler = udpv6_rcv,
1664 .err_handler = udpv6_err,
1665 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1668 /* ------------------------------------------------------------------------ */
1669 #ifdef CONFIG_PROC_FS
1670 int udp6_seq_show(struct seq_file *seq, void *v)
1672 if (v == SEQ_START_TOKEN) {
1673 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1675 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1676 struct inet_sock *inet = inet_sk(v);
1677 __u16 srcp = ntohs(inet->inet_sport);
1678 __u16 destp = ntohs(inet->inet_dport);
1679 __ip6_dgram_sock_seq_show(seq, v, srcp, destp,
1680 udp_rqueue_get(v), bucket);
1685 const struct seq_operations udp6_seq_ops = {
1686 .start = udp_seq_start,
1687 .next = udp_seq_next,
1688 .stop = udp_seq_stop,
1689 .show = udp6_seq_show,
1691 EXPORT_SYMBOL(udp6_seq_ops);
1693 static struct udp_seq_afinfo udp6_seq_afinfo = {
1695 .udp_table = &udp_table,
1698 int __net_init udp6_proc_init(struct net *net)
1700 if (!proc_create_net_data("udp6", 0444, net->proc_net, &udp6_seq_ops,
1701 sizeof(struct udp_iter_state), &udp6_seq_afinfo))
1706 void udp6_proc_exit(struct net *net)
1708 remove_proc_entry("udp6", net->proc_net);
1710 #endif /* CONFIG_PROC_FS */
1712 /* ------------------------------------------------------------------------ */
1714 struct proto udpv6_prot = {
1716 .owner = THIS_MODULE,
1717 .close = udp_lib_close,
1718 .pre_connect = udpv6_pre_connect,
1719 .connect = ip6_datagram_connect,
1720 .disconnect = udp_disconnect,
1722 .init = udp_init_sock,
1723 .destroy = udpv6_destroy_sock,
1724 .setsockopt = udpv6_setsockopt,
1725 .getsockopt = udpv6_getsockopt,
1726 .sendmsg = udpv6_sendmsg,
1727 .recvmsg = udpv6_recvmsg,
1728 .release_cb = ip6_datagram_release_cb,
1729 .hash = udp_lib_hash,
1730 .unhash = udp_lib_unhash,
1731 .rehash = udp_v6_rehash,
1732 .get_port = udp_v6_get_port,
1733 #ifdef CONFIG_BPF_SYSCALL
1734 .psock_update_sk_prot = udp_bpf_update_proto,
1736 .memory_allocated = &udp_memory_allocated,
1737 .sysctl_mem = sysctl_udp_mem,
1738 .sysctl_wmem_offset = offsetof(struct net, ipv4.sysctl_udp_wmem_min),
1739 .sysctl_rmem_offset = offsetof(struct net, ipv4.sysctl_udp_rmem_min),
1740 .obj_size = sizeof(struct udp6_sock),
1741 .h.udp_table = &udp_table,
1742 .diag_destroy = udp_abort,
1745 static struct inet_protosw udpv6_protosw = {
1747 .protocol = IPPROTO_UDP,
1748 .prot = &udpv6_prot,
1749 .ops = &inet6_dgram_ops,
1750 .flags = INET_PROTOSW_PERMANENT,
1753 int __init udpv6_init(void)
1757 ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1761 ret = inet6_register_protosw(&udpv6_protosw);
1763 goto out_udpv6_protocol;
1768 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1772 void udpv6_exit(void)
1774 inet6_unregister_protosw(&udpv6_protosw);
1775 inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);