3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Based on linux/ipv4/udp.c
11 * Hideaki YOSHIFUJI : sin6_scope_id support
12 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
13 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
14 * a single port at the same time.
15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
16 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/udp6 to seq_file.
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/ipv6.h>
33 #include <linux/icmpv6.h>
34 #include <linux/init.h>
35 #include <linux/module.h>
36 #include <linux/skbuff.h>
37 #include <linux/slab.h>
38 #include <linux/uaccess.h>
40 #include <net/addrconf.h>
41 #include <net/ndisc.h>
42 #include <net/protocol.h>
43 #include <net/transp_v6.h>
44 #include <net/ip6_route.h>
46 #include <net/tcp_states.h>
47 #include <net/ip6_checksum.h>
48 #include <net/ip6_tunnel.h>
50 #include <net/inet_hashtables.h>
51 #include <net/inet6_hashtables.h>
52 #include <net/busy_poll.h>
53 #include <net/sock_reuseport.h>
55 #include <linux/proc_fs.h>
56 #include <linux/seq_file.h>
57 #include <trace/events/skb.h>
60 static bool udp6_lib_exact_dif_match(struct net
*net
, struct sk_buff
*skb
)
62 #if defined(CONFIG_NET_L3_MASTER_DEV)
63 if (!net
->ipv4
.sysctl_udp_l3mdev_accept
&&
64 skb
&& ipv6_l3mdev_skb(IP6CB(skb
)->flags
))
70 static u32
udp6_ehashfn(const struct net
*net
,
71 const struct in6_addr
*laddr
,
73 const struct in6_addr
*faddr
,
76 static u32 udp6_ehash_secret __read_mostly
;
77 static u32 udp_ipv6_hash_secret __read_mostly
;
81 net_get_random_once(&udp6_ehash_secret
,
82 sizeof(udp6_ehash_secret
));
83 net_get_random_once(&udp_ipv6_hash_secret
,
84 sizeof(udp_ipv6_hash_secret
));
86 lhash
= (__force u32
)laddr
->s6_addr32
[3];
87 fhash
= __ipv6_addr_jhash(faddr
, udp_ipv6_hash_secret
);
89 return __inet6_ehashfn(lhash
, lport
, fhash
, fport
,
90 udp_ipv6_hash_secret
+ net_hash_mix(net
));
93 int udp_v6_get_port(struct sock
*sk
, unsigned short snum
)
95 unsigned int hash2_nulladdr
=
96 ipv6_portaddr_hash(sock_net(sk
), &in6addr_any
, snum
);
97 unsigned int hash2_partial
=
98 ipv6_portaddr_hash(sock_net(sk
), &sk
->sk_v6_rcv_saddr
, 0);
100 /* precompute partial secondary hash */
101 udp_sk(sk
)->udp_portaddr_hash
= hash2_partial
;
102 return udp_lib_get_port(sk
, snum
, hash2_nulladdr
);
105 static void udp_v6_rehash(struct sock
*sk
)
107 u16 new_hash
= ipv6_portaddr_hash(sock_net(sk
),
108 &sk
->sk_v6_rcv_saddr
,
109 inet_sk(sk
)->inet_num
);
111 udp_lib_rehash(sk
, new_hash
);
114 static int compute_score(struct sock
*sk
, struct net
*net
,
115 const struct in6_addr
*saddr
, __be16 sport
,
116 const struct in6_addr
*daddr
, unsigned short hnum
,
117 int dif
, int sdif
, bool exact_dif
)
120 struct inet_sock
*inet
;
123 if (!net_eq(sock_net(sk
), net
) ||
124 udp_sk(sk
)->udp_port_hash
!= hnum
||
125 sk
->sk_family
!= PF_INET6
)
128 if (!ipv6_addr_equal(&sk
->sk_v6_rcv_saddr
, daddr
))
134 if (inet
->inet_dport
) {
135 if (inet
->inet_dport
!= sport
)
140 if (!ipv6_addr_any(&sk
->sk_v6_daddr
)) {
141 if (!ipv6_addr_equal(&sk
->sk_v6_daddr
, saddr
))
146 dev_match
= udp_sk_bound_dev_eq(net
, sk
->sk_bound_dev_if
, dif
, sdif
);
151 if (sk
->sk_incoming_cpu
== raw_smp_processor_id())
157 /* called with rcu_read_lock() */
158 static struct sock
*udp6_lib_lookup2(struct net
*net
,
159 const struct in6_addr
*saddr
, __be16 sport
,
160 const struct in6_addr
*daddr
, unsigned int hnum
,
161 int dif
, int sdif
, bool exact_dif
,
162 struct udp_hslot
*hslot2
, struct sk_buff
*skb
)
164 struct sock
*sk
, *result
;
170 udp_portaddr_for_each_entry_rcu(sk
, &hslot2
->head
) {
171 score
= compute_score(sk
, net
, saddr
, sport
,
172 daddr
, hnum
, dif
, sdif
, exact_dif
);
173 if (score
> badness
) {
174 if (sk
->sk_reuseport
) {
175 hash
= udp6_ehashfn(net
, daddr
, hnum
,
178 result
= reuseport_select_sock(sk
, hash
, skb
,
179 sizeof(struct udphdr
));
190 /* rcu_read_lock() must be held */
191 struct sock
*__udp6_lib_lookup(struct net
*net
,
192 const struct in6_addr
*saddr
, __be16 sport
,
193 const struct in6_addr
*daddr
, __be16 dport
,
194 int dif
, int sdif
, struct udp_table
*udptable
,
197 unsigned short hnum
= ntohs(dport
);
198 unsigned int hash2
, slot2
;
199 struct udp_hslot
*hslot2
;
201 bool exact_dif
= udp6_lib_exact_dif_match(net
, skb
);
203 hash2
= ipv6_portaddr_hash(net
, daddr
, hnum
);
204 slot2
= hash2
& udptable
->mask
;
205 hslot2
= &udptable
->hash2
[slot2
];
207 result
= udp6_lib_lookup2(net
, saddr
, sport
,
208 daddr
, hnum
, dif
, sdif
, exact_dif
,
211 hash2
= ipv6_portaddr_hash(net
, &in6addr_any
, hnum
);
212 slot2
= hash2
& udptable
->mask
;
214 hslot2
= &udptable
->hash2
[slot2
];
216 result
= udp6_lib_lookup2(net
, saddr
, sport
,
217 &in6addr_any
, hnum
, dif
, sdif
,
221 if (unlikely(IS_ERR(result
)))
225 EXPORT_SYMBOL_GPL(__udp6_lib_lookup
);
227 static struct sock
*__udp6_lib_lookup_skb(struct sk_buff
*skb
,
228 __be16 sport
, __be16 dport
,
229 struct udp_table
*udptable
)
231 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
233 return __udp6_lib_lookup(dev_net(skb
->dev
), &iph
->saddr
, sport
,
234 &iph
->daddr
, dport
, inet6_iif(skb
),
235 inet6_sdif(skb
), udptable
, skb
);
238 struct sock
*udp6_lib_lookup_skb(struct sk_buff
*skb
,
239 __be16 sport
, __be16 dport
)
241 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
243 return __udp6_lib_lookup(dev_net(skb
->dev
), &iph
->saddr
, sport
,
244 &iph
->daddr
, dport
, inet6_iif(skb
),
245 inet6_sdif(skb
), &udp_table
, skb
);
247 EXPORT_SYMBOL_GPL(udp6_lib_lookup_skb
);
249 /* Must be called under rcu_read_lock().
250 * Does increment socket refcount.
252 #if IS_ENABLED(CONFIG_NF_TPROXY_IPV6) || IS_ENABLED(CONFIG_NF_SOCKET_IPV6)
253 struct sock
*udp6_lib_lookup(struct net
*net
, const struct in6_addr
*saddr
, __be16 sport
,
254 const struct in6_addr
*daddr
, __be16 dport
, int dif
)
258 sk
= __udp6_lib_lookup(net
, saddr
, sport
, daddr
, dport
,
259 dif
, 0, &udp_table
, NULL
);
260 if (sk
&& !refcount_inc_not_zero(&sk
->sk_refcnt
))
264 EXPORT_SYMBOL_GPL(udp6_lib_lookup
);
267 /* do not use the scratch area len for jumbogram: their length execeeds the
268 * scratch area space; note that the IP6CB flags is still in the first
269 * cacheline, so checking for jumbograms is cheap
271 static int udp6_skb_len(struct sk_buff
*skb
)
273 return unlikely(inet6_is_jumbogram(skb
)) ? skb
->len
: udp_skb_len(skb
);
277 * This should be easy, if there is something there we
278 * return it, otherwise we block.
281 int udpv6_recvmsg(struct sock
*sk
, struct msghdr
*msg
, size_t len
,
282 int noblock
, int flags
, int *addr_len
)
284 struct ipv6_pinfo
*np
= inet6_sk(sk
);
285 struct inet_sock
*inet
= inet_sk(sk
);
287 unsigned int ulen
, copied
;
288 int peeked
, peeking
, off
;
290 int is_udplite
= IS_UDPLITE(sk
);
291 bool checksum_valid
= false;
295 if (flags
& MSG_ERRQUEUE
)
296 return ipv6_recv_error(sk
, msg
, len
, addr_len
);
298 if (np
->rxpmtu
&& np
->rxopt
.bits
.rxpmtu
)
299 return ipv6_recv_rxpmtu(sk
, msg
, len
, addr_len
);
302 peeking
= flags
& MSG_PEEK
;
303 off
= sk_peek_offset(sk
, flags
);
304 skb
= __skb_recv_udp(sk
, flags
, noblock
, &peeked
, &off
, &err
);
308 ulen
= udp6_skb_len(skb
);
310 if (copied
> ulen
- off
)
312 else if (copied
< ulen
)
313 msg
->msg_flags
|= MSG_TRUNC
;
315 is_udp4
= (skb
->protocol
== htons(ETH_P_IP
));
316 mib
= __UDPX_MIB(sk
, is_udp4
);
319 * If checksum is needed at all, try to do it while copying the
320 * data. If the data is truncated, or if we only want a partial
321 * coverage checksum (UDP-Lite), do it before the copy.
324 if (copied
< ulen
|| peeking
||
325 (is_udplite
&& UDP_SKB_CB(skb
)->partial_cov
)) {
326 checksum_valid
= udp_skb_csum_unnecessary(skb
) ||
327 !__udp_lib_checksum_complete(skb
);
332 if (checksum_valid
|| udp_skb_csum_unnecessary(skb
)) {
333 if (udp_skb_is_linear(skb
))
334 err
= copy_linear_skb(skb
, copied
, off
, &msg
->msg_iter
);
336 err
= skb_copy_datagram_msg(skb
, off
, msg
, copied
);
338 err
= skb_copy_and_csum_datagram_msg(skb
, off
, msg
);
344 atomic_inc(&sk
->sk_drops
);
345 SNMP_INC_STATS(mib
, UDP_MIB_INERRORS
);
351 SNMP_INC_STATS(mib
, UDP_MIB_INDATAGRAMS
);
353 sock_recv_ts_and_drops(msg
, sk
, skb
);
355 /* Copy the address. */
357 DECLARE_SOCKADDR(struct sockaddr_in6
*, sin6
, msg
->msg_name
);
358 sin6
->sin6_family
= AF_INET6
;
359 sin6
->sin6_port
= udp_hdr(skb
)->source
;
360 sin6
->sin6_flowinfo
= 0;
363 ipv6_addr_set_v4mapped(ip_hdr(skb
)->saddr
,
365 sin6
->sin6_scope_id
= 0;
367 sin6
->sin6_addr
= ipv6_hdr(skb
)->saddr
;
368 sin6
->sin6_scope_id
=
369 ipv6_iface_scope_id(&sin6
->sin6_addr
,
372 *addr_len
= sizeof(*sin6
);
375 if (udp_sk(sk
)->gro_enabled
)
376 udp_cmsg_recv(msg
, sk
, skb
);
379 ip6_datagram_recv_common_ctl(sk
, msg
, skb
);
382 if (inet
->cmsg_flags
)
383 ip_cmsg_recv_offset(msg
, sk
, skb
,
384 sizeof(struct udphdr
), off
);
387 ip6_datagram_recv_specific_ctl(sk
, msg
, skb
);
391 if (flags
& MSG_TRUNC
)
394 skb_consume_udp(sk
, skb
, peeking
? -err
: err
);
398 if (!__sk_queue_drop_skb(sk
, &udp_sk(sk
)->reader_queue
, skb
, flags
,
399 udp_skb_destructor
)) {
400 SNMP_INC_STATS(mib
, UDP_MIB_CSUMERRORS
);
401 SNMP_INC_STATS(mib
, UDP_MIB_INERRORS
);
405 /* starting over for a new packet, but check if we need to yield */
407 msg
->msg_flags
&= ~MSG_TRUNC
;
411 DEFINE_STATIC_KEY_FALSE(udpv6_encap_needed_key
);
412 void udpv6_encap_enable(void)
414 static_branch_inc(&udpv6_encap_needed_key
);
416 EXPORT_SYMBOL(udpv6_encap_enable
);
418 /* Handler for tunnels with arbitrary destination ports: no socket lookup, go
419 * through error handlers in encapsulations looking for a match.
421 static int __udp6_lib_err_encap_no_sk(struct sk_buff
*skb
,
422 struct inet6_skb_parm
*opt
,
423 u8 type
, u8 code
, int offset
, u32 info
)
427 for (i
= 0; i
< MAX_IPTUN_ENCAP_OPS
; i
++) {
428 int (*handler
)(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
429 u8 type
, u8 code
, int offset
, u32 info
);
431 if (!ip6tun_encaps
[i
])
433 handler
= rcu_dereference(ip6tun_encaps
[i
]->err_handler
);
434 if (handler
&& !handler(skb
, opt
, type
, code
, offset
, info
))
441 /* Try to match ICMP errors to UDP tunnels by looking up a socket without
442 * reversing source and destination port: this will match tunnels that force the
443 * same destination port on both endpoints (e.g. VXLAN, GENEVE). Note that
444 * lwtunnels might actually break this assumption by being configured with
445 * different destination ports on endpoints, in this case we won't be able to
446 * trace ICMP messages back to them.
448 * If this doesn't match any socket, probe tunnels with arbitrary destination
449 * ports (e.g. FoU, GUE): there, the receiving socket is useless, as the port
450 * we've sent packets to won't necessarily match the local destination port.
452 * Then ask the tunnel implementation to match the error against a valid
455 * Return an error if we can't find a match, the socket if we need further
456 * processing, zero otherwise.
458 static struct sock
*__udp6_lib_err_encap(struct net
*net
,
459 const struct ipv6hdr
*hdr
, int offset
,
461 struct udp_table
*udptable
,
463 struct inet6_skb_parm
*opt
,
464 u8 type
, u8 code
, __be32 info
)
466 int network_offset
, transport_offset
;
469 network_offset
= skb_network_offset(skb
);
470 transport_offset
= skb_transport_offset(skb
);
472 /* Network header needs to point to the outer IPv6 header inside ICMP */
473 skb_reset_network_header(skb
);
475 /* Transport header needs to point to the UDP header */
476 skb_set_transport_header(skb
, offset
);
478 sk
= __udp6_lib_lookup(net
, &hdr
->daddr
, uh
->source
,
479 &hdr
->saddr
, uh
->dest
,
480 inet6_iif(skb
), 0, udptable
, skb
);
482 int (*lookup
)(struct sock
*sk
, struct sk_buff
*skb
);
483 struct udp_sock
*up
= udp_sk(sk
);
485 lookup
= READ_ONCE(up
->encap_err_lookup
);
486 if (!lookup
|| lookup(sk
, skb
))
491 sk
= ERR_PTR(__udp6_lib_err_encap_no_sk(skb
, opt
, type
, code
,
495 skb_set_transport_header(skb
, transport_offset
);
496 skb_set_network_header(skb
, network_offset
);
501 int __udp6_lib_err(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
502 u8 type
, u8 code
, int offset
, __be32 info
,
503 struct udp_table
*udptable
)
505 struct ipv6_pinfo
*np
;
506 const struct ipv6hdr
*hdr
= (const struct ipv6hdr
*)skb
->data
;
507 const struct in6_addr
*saddr
= &hdr
->saddr
;
508 const struct in6_addr
*daddr
= &hdr
->daddr
;
509 struct udphdr
*uh
= (struct udphdr
*)(skb
->data
+offset
);
514 struct net
*net
= dev_net(skb
->dev
);
516 sk
= __udp6_lib_lookup(net
, daddr
, uh
->dest
, saddr
, uh
->source
,
517 inet6_iif(skb
), inet6_sdif(skb
), udptable
, skb
);
519 /* No socket for error: try tunnels before discarding */
520 sk
= ERR_PTR(-ENOENT
);
521 if (static_branch_unlikely(&udpv6_encap_needed_key
)) {
522 sk
= __udp6_lib_err_encap(net
, hdr
, offset
, uh
,
524 opt
, type
, code
, info
);
530 __ICMP6_INC_STATS(net
, __in6_dev_get(skb
->dev
),
538 harderr
= icmpv6_err_convert(type
, code
, &err
);
541 if (type
== ICMPV6_PKT_TOOBIG
) {
542 if (!ip6_sk_accept_pmtu(sk
))
544 ip6_sk_update_pmtu(skb
, sk
, info
);
545 if (np
->pmtudisc
!= IPV6_PMTUDISC_DONT
)
548 if (type
== NDISC_REDIRECT
) {
550 ip6_redirect(skb
, sock_net(sk
), inet6_iif(skb
),
551 sk
->sk_mark
, sk
->sk_uid
);
553 ip6_sk_redirect(skb
, sk
);
558 /* Tunnels don't have an application socket: don't pass errors back */
563 if (!harderr
|| sk
->sk_state
!= TCP_ESTABLISHED
)
566 ipv6_icmp_error(sk
, skb
, err
, uh
->dest
, ntohl(info
), (u8
*)(uh
+1));
570 sk
->sk_error_report(sk
);
575 static int __udpv6_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
579 if (!ipv6_addr_any(&sk
->sk_v6_daddr
)) {
580 sock_rps_save_rxhash(sk
, skb
);
581 sk_mark_napi_id(sk
, skb
);
582 sk_incoming_cpu_update(sk
);
584 sk_mark_napi_id_once(sk
, skb
);
587 rc
= __udp_enqueue_schedule_skb(sk
, skb
);
589 int is_udplite
= IS_UDPLITE(sk
);
591 /* Note that an ENOMEM error is charged twice */
593 UDP6_INC_STATS(sock_net(sk
),
594 UDP_MIB_RCVBUFERRORS
, is_udplite
);
595 UDP6_INC_STATS(sock_net(sk
), UDP_MIB_INERRORS
, is_udplite
);
603 static __inline__
int udpv6_err(struct sk_buff
*skb
,
604 struct inet6_skb_parm
*opt
, u8 type
,
605 u8 code
, int offset
, __be32 info
)
607 return __udp6_lib_err(skb
, opt
, type
, code
, offset
, info
, &udp_table
);
610 static int udpv6_queue_rcv_one_skb(struct sock
*sk
, struct sk_buff
*skb
)
612 struct udp_sock
*up
= udp_sk(sk
);
613 int is_udplite
= IS_UDPLITE(sk
);
615 if (!xfrm6_policy_check(sk
, XFRM_POLICY_IN
, skb
))
618 if (static_branch_unlikely(&udpv6_encap_needed_key
) && up
->encap_type
) {
619 int (*encap_rcv
)(struct sock
*sk
, struct sk_buff
*skb
);
622 * This is an encapsulation socket so pass the skb to
623 * the socket's udp_encap_rcv() hook. Otherwise, just
624 * fall through and pass this up the UDP socket.
625 * up->encap_rcv() returns the following value:
626 * =0 if skb was successfully passed to the encap
627 * handler or was discarded by it.
628 * >0 if skb should be passed on to UDP.
629 * <0 if skb should be resubmitted as proto -N
632 /* if we're overly short, let UDP handle it */
633 encap_rcv
= READ_ONCE(up
->encap_rcv
);
637 /* Verify checksum before giving to encap */
638 if (udp_lib_checksum_complete(skb
))
641 ret
= encap_rcv(sk
, skb
);
643 __UDP_INC_STATS(sock_net(sk
),
650 /* FALLTHROUGH -- it's a UDP Packet */
654 * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
656 if ((is_udplite
& UDPLITE_RECV_CC
) && UDP_SKB_CB(skb
)->partial_cov
) {
658 if (up
->pcrlen
== 0) { /* full coverage was set */
659 net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
660 UDP_SKB_CB(skb
)->cscov
, skb
->len
);
663 if (UDP_SKB_CB(skb
)->cscov
< up
->pcrlen
) {
664 net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
665 UDP_SKB_CB(skb
)->cscov
, up
->pcrlen
);
670 prefetch(&sk
->sk_rmem_alloc
);
671 if (rcu_access_pointer(sk
->sk_filter
) &&
672 udp_lib_checksum_complete(skb
))
675 if (sk_filter_trim_cap(sk
, skb
, sizeof(struct udphdr
)))
678 udp_csum_pull_header(skb
);
682 return __udpv6_queue_rcv_skb(sk
, skb
);
685 __UDP6_INC_STATS(sock_net(sk
), UDP_MIB_CSUMERRORS
, is_udplite
);
687 __UDP6_INC_STATS(sock_net(sk
), UDP_MIB_INERRORS
, is_udplite
);
688 atomic_inc(&sk
->sk_drops
);
693 static int udpv6_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
695 struct sk_buff
*next
, *segs
;
698 if (likely(!udp_unexpected_gso(sk
, skb
)))
699 return udpv6_queue_rcv_one_skb(sk
, skb
);
701 __skb_push(skb
, -skb_mac_offset(skb
));
702 segs
= udp_rcv_segment(sk
, skb
, false);
703 for (skb
= segs
; skb
; skb
= next
) {
705 __skb_pull(skb
, skb_transport_offset(skb
));
707 ret
= udpv6_queue_rcv_one_skb(sk
, skb
);
709 ip6_protocol_deliver_rcu(dev_net(skb
->dev
), skb
, ret
,
715 static bool __udp_v6_is_mcast_sock(struct net
*net
, struct sock
*sk
,
716 __be16 loc_port
, const struct in6_addr
*loc_addr
,
717 __be16 rmt_port
, const struct in6_addr
*rmt_addr
,
718 int dif
, int sdif
, unsigned short hnum
)
720 struct inet_sock
*inet
= inet_sk(sk
);
722 if (!net_eq(sock_net(sk
), net
))
725 if (udp_sk(sk
)->udp_port_hash
!= hnum
||
726 sk
->sk_family
!= PF_INET6
||
727 (inet
->inet_dport
&& inet
->inet_dport
!= rmt_port
) ||
728 (!ipv6_addr_any(&sk
->sk_v6_daddr
) &&
729 !ipv6_addr_equal(&sk
->sk_v6_daddr
, rmt_addr
)) ||
730 !udp_sk_bound_dev_eq(net
, sk
->sk_bound_dev_if
, dif
, sdif
) ||
731 (!ipv6_addr_any(&sk
->sk_v6_rcv_saddr
) &&
732 !ipv6_addr_equal(&sk
->sk_v6_rcv_saddr
, loc_addr
)))
734 if (!inet6_mc_check(sk
, loc_addr
, rmt_addr
))
739 static void udp6_csum_zero_error(struct sk_buff
*skb
)
741 /* RFC 2460 section 8.1 says that we SHOULD log
742 * this error. Well, it is reasonable.
744 net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
745 &ipv6_hdr(skb
)->saddr
, ntohs(udp_hdr(skb
)->source
),
746 &ipv6_hdr(skb
)->daddr
, ntohs(udp_hdr(skb
)->dest
));
750 * Note: called only from the BH handler context,
751 * so we don't need to lock the hashes.
753 static int __udp6_lib_mcast_deliver(struct net
*net
, struct sk_buff
*skb
,
754 const struct in6_addr
*saddr
, const struct in6_addr
*daddr
,
755 struct udp_table
*udptable
, int proto
)
757 struct sock
*sk
, *first
= NULL
;
758 const struct udphdr
*uh
= udp_hdr(skb
);
759 unsigned short hnum
= ntohs(uh
->dest
);
760 struct udp_hslot
*hslot
= udp_hashslot(udptable
, net
, hnum
);
761 unsigned int offset
= offsetof(typeof(*sk
), sk_node
);
762 unsigned int hash2
= 0, hash2_any
= 0, use_hash2
= (hslot
->count
> 10);
763 int dif
= inet6_iif(skb
);
764 int sdif
= inet6_sdif(skb
);
765 struct hlist_node
*node
;
766 struct sk_buff
*nskb
;
769 hash2_any
= ipv6_portaddr_hash(net
, &in6addr_any
, hnum
) &
771 hash2
= ipv6_portaddr_hash(net
, daddr
, hnum
) & udptable
->mask
;
773 hslot
= &udptable
->hash2
[hash2
];
774 offset
= offsetof(typeof(*sk
), __sk_common
.skc_portaddr_node
);
777 sk_for_each_entry_offset_rcu(sk
, node
, &hslot
->head
, offset
) {
778 if (!__udp_v6_is_mcast_sock(net
, sk
, uh
->dest
, daddr
,
779 uh
->source
, saddr
, dif
, sdif
,
782 /* If zero checksum and no_check is not on for
783 * the socket then skip it.
785 if (!uh
->check
&& !udp_sk(sk
)->no_check6_rx
)
791 nskb
= skb_clone(skb
, GFP_ATOMIC
);
792 if (unlikely(!nskb
)) {
793 atomic_inc(&sk
->sk_drops
);
794 __UDP6_INC_STATS(net
, UDP_MIB_RCVBUFERRORS
,
796 __UDP6_INC_STATS(net
, UDP_MIB_INERRORS
,
801 if (udpv6_queue_rcv_skb(sk
, nskb
) > 0)
805 /* Also lookup *:port if we are using hash2 and haven't done so yet. */
806 if (use_hash2
&& hash2
!= hash2_any
) {
812 if (udpv6_queue_rcv_skb(first
, skb
) > 0)
816 __UDP6_INC_STATS(net
, UDP_MIB_IGNOREDMULTI
,
817 proto
== IPPROTO_UDPLITE
);
822 static void udp6_sk_rx_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
824 if (udp_sk_rx_dst_set(sk
, dst
)) {
825 const struct rt6_info
*rt
= (const struct rt6_info
*)dst
;
827 inet6_sk(sk
)->rx_dst_cookie
= rt6_get_cookie(rt
);
831 /* wrapper for udp_queue_rcv_skb tacking care of csum conversion and
832 * return code conversion for ip layer consumption
834 static int udp6_unicast_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
,
839 if (inet_get_convert_csum(sk
) && uh
->check
&& !IS_UDPLITE(sk
))
840 skb_checksum_try_convert(skb
, IPPROTO_UDP
, uh
->check
,
843 ret
= udpv6_queue_rcv_skb(sk
, skb
);
845 /* a return value > 0 means to resubmit the input */
851 int __udp6_lib_rcv(struct sk_buff
*skb
, struct udp_table
*udptable
,
854 const struct in6_addr
*saddr
, *daddr
;
855 struct net
*net
= dev_net(skb
->dev
);
860 if (!pskb_may_pull(skb
, sizeof(struct udphdr
)))
863 saddr
= &ipv6_hdr(skb
)->saddr
;
864 daddr
= &ipv6_hdr(skb
)->daddr
;
867 ulen
= ntohs(uh
->len
);
871 if (proto
== IPPROTO_UDP
) {
872 /* UDP validates ulen. */
874 /* Check for jumbo payload */
878 if (ulen
< sizeof(*uh
))
881 if (ulen
< skb
->len
) {
882 if (pskb_trim_rcsum(skb
, ulen
))
884 saddr
= &ipv6_hdr(skb
)->saddr
;
885 daddr
= &ipv6_hdr(skb
)->daddr
;
890 if (udp6_csum_init(skb
, uh
, proto
))
893 /* Check if the socket is already available, e.g. due to early demux */
894 sk
= skb_steal_sock(skb
);
896 struct dst_entry
*dst
= skb_dst(skb
);
899 if (unlikely(sk
->sk_rx_dst
!= dst
))
900 udp6_sk_rx_dst_set(sk
, dst
);
902 if (!uh
->check
&& !udp_sk(sk
)->no_check6_rx
) {
904 goto report_csum_error
;
907 ret
= udp6_unicast_rcv_skb(sk
, skb
, uh
);
913 * Multicast receive code
915 if (ipv6_addr_is_multicast(daddr
))
916 return __udp6_lib_mcast_deliver(net
, skb
,
917 saddr
, daddr
, udptable
, proto
);
920 sk
= __udp6_lib_lookup_skb(skb
, uh
->source
, uh
->dest
, udptable
);
922 if (!uh
->check
&& !udp_sk(sk
)->no_check6_rx
)
923 goto report_csum_error
;
924 return udp6_unicast_rcv_skb(sk
, skb
, uh
);
928 goto report_csum_error
;
930 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
933 if (udp_lib_checksum_complete(skb
))
936 __UDP6_INC_STATS(net
, UDP_MIB_NOPORTS
, proto
== IPPROTO_UDPLITE
);
937 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_PORT_UNREACH
, 0);
943 net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
944 proto
== IPPROTO_UDPLITE
? "-Lite" : "",
945 saddr
, ntohs(uh
->source
),
947 daddr
, ntohs(uh
->dest
));
951 udp6_csum_zero_error(skb
);
953 __UDP6_INC_STATS(net
, UDP_MIB_CSUMERRORS
, proto
== IPPROTO_UDPLITE
);
955 __UDP6_INC_STATS(net
, UDP_MIB_INERRORS
, proto
== IPPROTO_UDPLITE
);
961 static struct sock
*__udp6_lib_demux_lookup(struct net
*net
,
962 __be16 loc_port
, const struct in6_addr
*loc_addr
,
963 __be16 rmt_port
, const struct in6_addr
*rmt_addr
,
966 unsigned short hnum
= ntohs(loc_port
);
967 unsigned int hash2
= ipv6_portaddr_hash(net
, loc_addr
, hnum
);
968 unsigned int slot2
= hash2
& udp_table
.mask
;
969 struct udp_hslot
*hslot2
= &udp_table
.hash2
[slot2
];
970 const __portpair ports
= INET_COMBINED_PORTS(rmt_port
, hnum
);
973 udp_portaddr_for_each_entry_rcu(sk
, &hslot2
->head
) {
974 if (sk
->sk_state
== TCP_ESTABLISHED
&&
975 INET6_MATCH(sk
, net
, rmt_addr
, loc_addr
, ports
, dif
, sdif
))
977 /* Only check first socket in chain */
983 static void udp_v6_early_demux(struct sk_buff
*skb
)
985 struct net
*net
= dev_net(skb
->dev
);
986 const struct udphdr
*uh
;
988 struct dst_entry
*dst
;
989 int dif
= skb
->dev
->ifindex
;
990 int sdif
= inet6_sdif(skb
);
992 if (!pskb_may_pull(skb
, skb_transport_offset(skb
) +
993 sizeof(struct udphdr
)))
998 if (skb
->pkt_type
== PACKET_HOST
)
999 sk
= __udp6_lib_demux_lookup(net
, uh
->dest
,
1000 &ipv6_hdr(skb
)->daddr
,
1001 uh
->source
, &ipv6_hdr(skb
)->saddr
,
1006 if (!sk
|| !refcount_inc_not_zero(&sk
->sk_refcnt
))
1010 skb
->destructor
= sock_efree
;
1011 dst
= READ_ONCE(sk
->sk_rx_dst
);
1014 dst
= dst_check(dst
, inet6_sk(sk
)->rx_dst_cookie
);
1016 /* set noref for now.
1017 * any place which wants to hold dst has to call
1020 skb_dst_set_noref(skb
, dst
);
1024 static __inline__
int udpv6_rcv(struct sk_buff
*skb
)
1026 return __udp6_lib_rcv(skb
, &udp_table
, IPPROTO_UDP
);
1030 * Throw away all pending data and cancel the corking. Socket is locked.
1032 static void udp_v6_flush_pending_frames(struct sock
*sk
)
1034 struct udp_sock
*up
= udp_sk(sk
);
1036 if (up
->pending
== AF_INET
)
1037 udp_flush_pending_frames(sk
);
1038 else if (up
->pending
) {
1041 ip6_flush_pending_frames(sk
);
1045 static int udpv6_pre_connect(struct sock
*sk
, struct sockaddr
*uaddr
,
1048 /* The following checks are replicated from __ip6_datagram_connect()
1049 * and intended to prevent BPF program called below from accessing
1050 * bytes that are out of the bound specified by user in addr_len.
1052 if (uaddr
->sa_family
== AF_INET
) {
1053 if (__ipv6_only_sock(sk
))
1054 return -EAFNOSUPPORT
;
1055 return udp_pre_connect(sk
, uaddr
, addr_len
);
1058 if (addr_len
< SIN6_LEN_RFC2133
)
1061 return BPF_CGROUP_RUN_PROG_INET6_CONNECT_LOCK(sk
, uaddr
);
1065 * udp6_hwcsum_outgoing - handle outgoing HW checksumming
1066 * @sk: socket we are sending on
1067 * @skb: sk_buff containing the filled-in UDP header
1068 * (checksum field must be zeroed out)
1070 static void udp6_hwcsum_outgoing(struct sock
*sk
, struct sk_buff
*skb
,
1071 const struct in6_addr
*saddr
,
1072 const struct in6_addr
*daddr
, int len
)
1074 unsigned int offset
;
1075 struct udphdr
*uh
= udp_hdr(skb
);
1076 struct sk_buff
*frags
= skb_shinfo(skb
)->frag_list
;
1080 /* Only one fragment on the socket. */
1081 skb
->csum_start
= skb_transport_header(skb
) - skb
->head
;
1082 skb
->csum_offset
= offsetof(struct udphdr
, check
);
1083 uh
->check
= ~csum_ipv6_magic(saddr
, daddr
, len
, IPPROTO_UDP
, 0);
1086 * HW-checksum won't work as there are two or more
1087 * fragments on the socket so that all csums of sk_buffs
1088 * should be together
1090 offset
= skb_transport_offset(skb
);
1091 skb
->csum
= skb_checksum(skb
, offset
, skb
->len
- offset
, 0);
1094 skb
->ip_summed
= CHECKSUM_NONE
;
1097 csum
= csum_add(csum
, frags
->csum
);
1098 } while ((frags
= frags
->next
));
1100 uh
->check
= csum_ipv6_magic(saddr
, daddr
, len
, IPPROTO_UDP
,
1103 uh
->check
= CSUM_MANGLED_0
;
1111 static int udp_v6_send_skb(struct sk_buff
*skb
, struct flowi6
*fl6
,
1112 struct inet_cork
*cork
)
1114 struct sock
*sk
= skb
->sk
;
1117 int is_udplite
= IS_UDPLITE(sk
);
1119 int offset
= skb_transport_offset(skb
);
1120 int len
= skb
->len
- offset
;
1123 * Create a UDP header
1126 uh
->source
= fl6
->fl6_sport
;
1127 uh
->dest
= fl6
->fl6_dport
;
1128 uh
->len
= htons(len
);
1131 if (cork
->gso_size
) {
1132 const int hlen
= skb_network_header_len(skb
) +
1133 sizeof(struct udphdr
);
1135 if (hlen
+ cork
->gso_size
> cork
->fragsize
)
1137 if (skb
->len
> cork
->gso_size
* UDP_MAX_SEGMENTS
)
1139 if (udp_sk(sk
)->no_check6_tx
)
1141 if (skb
->ip_summed
!= CHECKSUM_PARTIAL
|| is_udplite
||
1142 dst_xfrm(skb_dst(skb
)))
1145 skb_shinfo(skb
)->gso_size
= cork
->gso_size
;
1146 skb_shinfo(skb
)->gso_type
= SKB_GSO_UDP_L4
;
1151 csum
= udplite_csum(skb
);
1152 else if (udp_sk(sk
)->no_check6_tx
) { /* UDP csum disabled */
1153 skb
->ip_summed
= CHECKSUM_NONE
;
1155 } else if (skb
->ip_summed
== CHECKSUM_PARTIAL
) { /* UDP hardware csum */
1157 udp6_hwcsum_outgoing(sk
, skb
, &fl6
->saddr
, &fl6
->daddr
, len
);
1160 csum
= udp_csum(skb
);
1162 /* add protocol-dependent pseudo-header */
1163 uh
->check
= csum_ipv6_magic(&fl6
->saddr
, &fl6
->daddr
,
1164 len
, fl6
->flowi6_proto
, csum
);
1166 uh
->check
= CSUM_MANGLED_0
;
1169 err
= ip6_send_skb(skb
);
1171 if (err
== -ENOBUFS
&& !inet6_sk(sk
)->recverr
) {
1172 UDP6_INC_STATS(sock_net(sk
),
1173 UDP_MIB_SNDBUFERRORS
, is_udplite
);
1177 UDP6_INC_STATS(sock_net(sk
),
1178 UDP_MIB_OUTDATAGRAMS
, is_udplite
);
1183 static int udp_v6_push_pending_frames(struct sock
*sk
)
1185 struct sk_buff
*skb
;
1186 struct udp_sock
*up
= udp_sk(sk
);
1190 if (up
->pending
== AF_INET
)
1191 return udp_push_pending_frames(sk
);
1193 /* ip6_finish_skb will release the cork, so make a copy of
1196 fl6
= inet_sk(sk
)->cork
.fl
.u
.ip6
;
1198 skb
= ip6_finish_skb(sk
);
1202 err
= udp_v6_send_skb(skb
, &fl6
, &inet_sk(sk
)->cork
.base
);
1210 int udpv6_sendmsg(struct sock
*sk
, struct msghdr
*msg
, size_t len
)
1212 struct ipv6_txoptions opt_space
;
1213 struct udp_sock
*up
= udp_sk(sk
);
1214 struct inet_sock
*inet
= inet_sk(sk
);
1215 struct ipv6_pinfo
*np
= inet6_sk(sk
);
1216 DECLARE_SOCKADDR(struct sockaddr_in6
*, sin6
, msg
->msg_name
);
1217 struct in6_addr
*daddr
, *final_p
, final
;
1218 struct ipv6_txoptions
*opt
= NULL
;
1219 struct ipv6_txoptions
*opt_to_free
= NULL
;
1220 struct ip6_flowlabel
*flowlabel
= NULL
;
1222 struct dst_entry
*dst
;
1223 struct ipcm6_cookie ipc6
;
1224 int addr_len
= msg
->msg_namelen
;
1225 bool connected
= false;
1227 int corkreq
= up
->corkflag
|| msg
->msg_flags
&MSG_MORE
;
1229 int is_udplite
= IS_UDPLITE(sk
);
1230 int (*getfrag
)(void *, char *, int, int, int, struct sk_buff
*);
1233 ipc6
.gso_size
= up
->gso_size
;
1234 ipc6
.sockc
.tsflags
= sk
->sk_tsflags
;
1236 /* destination address check */
1238 if (addr_len
< offsetof(struct sockaddr
, sa_data
))
1241 switch (sin6
->sin6_family
) {
1243 if (addr_len
< SIN6_LEN_RFC2133
)
1245 daddr
= &sin6
->sin6_addr
;
1246 if (ipv6_addr_any(daddr
) &&
1247 ipv6_addr_v4mapped(&np
->saddr
))
1248 ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK
),
1252 goto do_udp_sendmsg
;
1254 msg
->msg_name
= sin6
= NULL
;
1255 msg
->msg_namelen
= addr_len
= 0;
1261 } else if (!up
->pending
) {
1262 if (sk
->sk_state
!= TCP_ESTABLISHED
)
1263 return -EDESTADDRREQ
;
1264 daddr
= &sk
->sk_v6_daddr
;
1269 if (ipv6_addr_v4mapped(daddr
)) {
1270 struct sockaddr_in sin
;
1271 sin
.sin_family
= AF_INET
;
1272 sin
.sin_port
= sin6
? sin6
->sin6_port
: inet
->inet_dport
;
1273 sin
.sin_addr
.s_addr
= daddr
->s6_addr32
[3];
1274 msg
->msg_name
= &sin
;
1275 msg
->msg_namelen
= sizeof(sin
);
1277 if (__ipv6_only_sock(sk
))
1278 return -ENETUNREACH
;
1279 return udp_sendmsg(sk
, msg
, len
);
1283 if (up
->pending
== AF_INET
)
1284 return udp_sendmsg(sk
, msg
, len
);
1286 /* Rough check on arithmetic overflow,
1287 better check is made in ip6_append_data().
1289 if (len
> INT_MAX
- sizeof(struct udphdr
))
1292 getfrag
= is_udplite
? udplite_getfrag
: ip_generic_getfrag
;
1295 * There are pending frames.
1296 * The socket lock must be held while it's corked.
1299 if (likely(up
->pending
)) {
1300 if (unlikely(up
->pending
!= AF_INET6
)) {
1302 return -EAFNOSUPPORT
;
1305 goto do_append_data
;
1309 ulen
+= sizeof(struct udphdr
);
1311 memset(&fl6
, 0, sizeof(fl6
));
1314 if (sin6
->sin6_port
== 0)
1317 fl6
.fl6_dport
= sin6
->sin6_port
;
1318 daddr
= &sin6
->sin6_addr
;
1321 fl6
.flowlabel
= sin6
->sin6_flowinfo
&IPV6_FLOWINFO_MASK
;
1322 if (fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) {
1323 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
1330 * Otherwise it will be difficult to maintain
1333 if (sk
->sk_state
== TCP_ESTABLISHED
&&
1334 ipv6_addr_equal(daddr
, &sk
->sk_v6_daddr
))
1335 daddr
= &sk
->sk_v6_daddr
;
1337 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
1338 sin6
->sin6_scope_id
&&
1339 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr
)))
1340 fl6
.flowi6_oif
= sin6
->sin6_scope_id
;
1342 if (sk
->sk_state
!= TCP_ESTABLISHED
)
1343 return -EDESTADDRREQ
;
1345 fl6
.fl6_dport
= inet
->inet_dport
;
1346 daddr
= &sk
->sk_v6_daddr
;
1347 fl6
.flowlabel
= np
->flow_label
;
1351 if (!fl6
.flowi6_oif
)
1352 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
1354 if (!fl6
.flowi6_oif
)
1355 fl6
.flowi6_oif
= np
->sticky_pktinfo
.ipi6_ifindex
;
1357 fl6
.flowi6_mark
= sk
->sk_mark
;
1358 fl6
.flowi6_uid
= sk
->sk_uid
;
1360 if (msg
->msg_controllen
) {
1362 memset(opt
, 0, sizeof(struct ipv6_txoptions
));
1363 opt
->tot_len
= sizeof(*opt
);
1366 err
= udp_cmsg_send(sk
, msg
, &ipc6
.gso_size
);
1368 err
= ip6_datagram_send_ctl(sock_net(sk
), sk
, msg
, &fl6
,
1371 fl6_sock_release(flowlabel
);
1374 if ((fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) && !flowlabel
) {
1375 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
1379 if (!(opt
->opt_nflen
|opt
->opt_flen
))
1384 opt
= txopt_get(np
);
1388 opt
= fl6_merge_options(&opt_space
, flowlabel
, opt
);
1389 opt
= ipv6_fixup_options(&opt_space
, opt
);
1392 fl6
.flowi6_proto
= sk
->sk_protocol
;
1393 if (!ipv6_addr_any(daddr
))
1396 fl6
.daddr
.s6_addr
[15] = 0x1; /* :: means loopback (BSD'ism) */
1397 if (ipv6_addr_any(&fl6
.saddr
) && !ipv6_addr_any(&np
->saddr
))
1398 fl6
.saddr
= np
->saddr
;
1399 fl6
.fl6_sport
= inet
->inet_sport
;
1401 if (cgroup_bpf_enabled
&& !connected
) {
1402 err
= BPF_CGROUP_RUN_PROG_UDP6_SENDMSG_LOCK(sk
,
1403 (struct sockaddr
*)sin6
, &fl6
.saddr
);
1407 if (ipv6_addr_v4mapped(&sin6
->sin6_addr
)) {
1408 /* BPF program rewrote IPv6-only by IPv4-mapped
1409 * IPv6. It's currently unsupported.
1414 if (sin6
->sin6_port
== 0) {
1415 /* BPF program set invalid port. Reject it. */
1419 fl6
.fl6_dport
= sin6
->sin6_port
;
1420 fl6
.daddr
= sin6
->sin6_addr
;
1424 final_p
= fl6_update_dst(&fl6
, opt
, &final
);
1428 if (!fl6
.flowi6_oif
&& ipv6_addr_is_multicast(&fl6
.daddr
)) {
1429 fl6
.flowi6_oif
= np
->mcast_oif
;
1431 } else if (!fl6
.flowi6_oif
)
1432 fl6
.flowi6_oif
= np
->ucast_oif
;
1434 security_sk_classify_flow(sk
, flowi6_to_flowi(&fl6
));
1436 if (ipc6
.tclass
< 0)
1437 ipc6
.tclass
= np
->tclass
;
1439 fl6
.flowlabel
= ip6_make_flowinfo(ipc6
.tclass
, fl6
.flowlabel
);
1441 dst
= ip6_sk_dst_lookup_flow(sk
, &fl6
, final_p
, connected
);
1448 if (ipc6
.hlimit
< 0)
1449 ipc6
.hlimit
= ip6_sk_dst_hoplimit(np
, &fl6
, dst
);
1451 if (msg
->msg_flags
&MSG_CONFIRM
)
1455 /* Lockless fast path for the non-corking case */
1457 struct inet_cork_full cork
;
1458 struct sk_buff
*skb
;
1460 skb
= ip6_make_skb(sk
, getfrag
, msg
, ulen
,
1461 sizeof(struct udphdr
), &ipc6
,
1462 &fl6
, (struct rt6_info
*)dst
,
1463 msg
->msg_flags
, &cork
);
1465 if (!IS_ERR_OR_NULL(skb
))
1466 err
= udp_v6_send_skb(skb
, &fl6
, &cork
.base
);
1471 if (unlikely(up
->pending
)) {
1472 /* The socket is already corked while preparing it. */
1473 /* ... which is an evident application bug. --ANK */
1476 net_dbg_ratelimited("udp cork app bug 2\n");
1481 up
->pending
= AF_INET6
;
1484 if (ipc6
.dontfrag
< 0)
1485 ipc6
.dontfrag
= np
->dontfrag
;
1487 err
= ip6_append_data(sk
, getfrag
, msg
, ulen
, sizeof(struct udphdr
),
1488 &ipc6
, &fl6
, (struct rt6_info
*)dst
,
1489 corkreq
? msg
->msg_flags
|MSG_MORE
: msg
->msg_flags
);
1491 udp_v6_flush_pending_frames(sk
);
1493 err
= udp_v6_push_pending_frames(sk
);
1494 else if (unlikely(skb_queue_empty(&sk
->sk_write_queue
)))
1498 err
= np
->recverr
? net_xmit_errno(err
) : 0;
1504 fl6_sock_release(flowlabel
);
1505 txopt_put(opt_to_free
);
1509 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
1510 * ENOBUFS might not be good (it's not tunable per se), but otherwise
1511 * we don't have a good statistic (IpOutDiscards but it can be too many
1512 * things). We could add another new stat but at least for now that
1513 * seems like overkill.
1515 if (err
== -ENOBUFS
|| test_bit(SOCK_NOSPACE
, &sk
->sk_socket
->flags
)) {
1516 UDP6_INC_STATS(sock_net(sk
),
1517 UDP_MIB_SNDBUFERRORS
, is_udplite
);
1522 if (msg
->msg_flags
& MSG_PROBE
)
1523 dst_confirm_neigh(dst
, &fl6
.daddr
);
1524 if (!(msg
->msg_flags
&MSG_PROBE
) || len
)
1525 goto back_from_confirm
;
1530 void udpv6_destroy_sock(struct sock
*sk
)
1532 struct udp_sock
*up
= udp_sk(sk
);
1534 udp_v6_flush_pending_frames(sk
);
1537 if (static_branch_unlikely(&udpv6_encap_needed_key
)) {
1538 if (up
->encap_type
) {
1539 void (*encap_destroy
)(struct sock
*sk
);
1540 encap_destroy
= READ_ONCE(up
->encap_destroy
);
1544 if (up
->encap_enabled
)
1545 static_branch_dec(&udpv6_encap_needed_key
);
1548 inet6_destroy_sock(sk
);
1552 * Socket option code for UDP
1554 int udpv6_setsockopt(struct sock
*sk
, int level
, int optname
,
1555 char __user
*optval
, unsigned int optlen
)
1557 if (level
== SOL_UDP
|| level
== SOL_UDPLITE
)
1558 return udp_lib_setsockopt(sk
, level
, optname
, optval
, optlen
,
1559 udp_v6_push_pending_frames
);
1560 return ipv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1563 #ifdef CONFIG_COMPAT
1564 int compat_udpv6_setsockopt(struct sock
*sk
, int level
, int optname
,
1565 char __user
*optval
, unsigned int optlen
)
1567 if (level
== SOL_UDP
|| level
== SOL_UDPLITE
)
1568 return udp_lib_setsockopt(sk
, level
, optname
, optval
, optlen
,
1569 udp_v6_push_pending_frames
);
1570 return compat_ipv6_setsockopt(sk
, level
, optname
, optval
, optlen
);
1574 int udpv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1575 char __user
*optval
, int __user
*optlen
)
1577 if (level
== SOL_UDP
|| level
== SOL_UDPLITE
)
1578 return udp_lib_getsockopt(sk
, level
, optname
, optval
, optlen
);
1579 return ipv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1582 #ifdef CONFIG_COMPAT
1583 int compat_udpv6_getsockopt(struct sock
*sk
, int level
, int optname
,
1584 char __user
*optval
, int __user
*optlen
)
1586 if (level
== SOL_UDP
|| level
== SOL_UDPLITE
)
1587 return udp_lib_getsockopt(sk
, level
, optname
, optval
, optlen
);
1588 return compat_ipv6_getsockopt(sk
, level
, optname
, optval
, optlen
);
1592 /* thinking of making this const? Don't.
1593 * early_demux can change based on sysctl.
1595 static struct inet6_protocol udpv6_protocol
= {
1596 .early_demux
= udp_v6_early_demux
,
1597 .early_demux_handler
= udp_v6_early_demux
,
1598 .handler
= udpv6_rcv
,
1599 .err_handler
= udpv6_err
,
1600 .flags
= INET6_PROTO_NOPOLICY
|INET6_PROTO_FINAL
,
1603 /* ------------------------------------------------------------------------ */
1604 #ifdef CONFIG_PROC_FS
1605 int udp6_seq_show(struct seq_file
*seq
, void *v
)
1607 if (v
== SEQ_START_TOKEN
) {
1608 seq_puts(seq
, IPV6_SEQ_DGRAM_HEADER
);
1610 int bucket
= ((struct udp_iter_state
*)seq
->private)->bucket
;
1611 struct inet_sock
*inet
= inet_sk(v
);
1612 __u16 srcp
= ntohs(inet
->inet_sport
);
1613 __u16 destp
= ntohs(inet
->inet_dport
);
1614 __ip6_dgram_sock_seq_show(seq
, v
, srcp
, destp
,
1615 udp_rqueue_get(v
), bucket
);
1620 const struct seq_operations udp6_seq_ops
= {
1621 .start
= udp_seq_start
,
1622 .next
= udp_seq_next
,
1623 .stop
= udp_seq_stop
,
1624 .show
= udp6_seq_show
,
1626 EXPORT_SYMBOL(udp6_seq_ops
);
1628 static struct udp_seq_afinfo udp6_seq_afinfo
= {
1630 .udp_table
= &udp_table
,
1633 int __net_init
udp6_proc_init(struct net
*net
)
1635 if (!proc_create_net_data("udp6", 0444, net
->proc_net
, &udp6_seq_ops
,
1636 sizeof(struct udp_iter_state
), &udp6_seq_afinfo
))
1641 void udp6_proc_exit(struct net
*net
)
1643 remove_proc_entry("udp6", net
->proc_net
);
1645 #endif /* CONFIG_PROC_FS */
1647 /* ------------------------------------------------------------------------ */
1649 struct proto udpv6_prot
= {
1651 .owner
= THIS_MODULE
,
1652 .close
= udp_lib_close
,
1653 .pre_connect
= udpv6_pre_connect
,
1654 .connect
= ip6_datagram_connect
,
1655 .disconnect
= udp_disconnect
,
1657 .init
= udp_init_sock
,
1658 .destroy
= udpv6_destroy_sock
,
1659 .setsockopt
= udpv6_setsockopt
,
1660 .getsockopt
= udpv6_getsockopt
,
1661 .sendmsg
= udpv6_sendmsg
,
1662 .recvmsg
= udpv6_recvmsg
,
1663 .release_cb
= ip6_datagram_release_cb
,
1664 .hash
= udp_lib_hash
,
1665 .unhash
= udp_lib_unhash
,
1666 .rehash
= udp_v6_rehash
,
1667 .get_port
= udp_v6_get_port
,
1668 .memory_allocated
= &udp_memory_allocated
,
1669 .sysctl_mem
= sysctl_udp_mem
,
1670 .sysctl_wmem_offset
= offsetof(struct net
, ipv4
.sysctl_udp_wmem_min
),
1671 .sysctl_rmem_offset
= offsetof(struct net
, ipv4
.sysctl_udp_rmem_min
),
1672 .obj_size
= sizeof(struct udp6_sock
),
1673 .h
.udp_table
= &udp_table
,
1674 #ifdef CONFIG_COMPAT
1675 .compat_setsockopt
= compat_udpv6_setsockopt
,
1676 .compat_getsockopt
= compat_udpv6_getsockopt
,
1678 .diag_destroy
= udp_abort
,
1681 static struct inet_protosw udpv6_protosw
= {
1683 .protocol
= IPPROTO_UDP
,
1684 .prot
= &udpv6_prot
,
1685 .ops
= &inet6_dgram_ops
,
1686 .flags
= INET_PROTOSW_PERMANENT
,
1689 int __init
udpv6_init(void)
1693 ret
= inet6_add_protocol(&udpv6_protocol
, IPPROTO_UDP
);
1697 ret
= inet6_register_protosw(&udpv6_protosw
);
1699 goto out_udpv6_protocol
;
1704 inet6_del_protocol(&udpv6_protocol
, IPPROTO_UDP
);
1708 void udpv6_exit(void)
1710 inet6_unregister_protosw(&udpv6_protosw
);
1711 inet6_del_protocol(&udpv6_protocol
, IPPROTO_UDP
);