~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/ipv6/udp.c

Version: ~ [ linux-5.13-rc1 ] ~ [ linux-5.12.2 ] ~ [ linux-5.11.19 ] ~ [ linux-5.10.35 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.117 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.190 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.232 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.268 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.268 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  *      UDP over IPv6
  3  *      Linux INET6 implementation
  4  *
  5  *      Authors:
  6  *      Pedro Roque             <roque@di.fc.ul.pt>
  7  *
  8  *      Based on linux/ipv4/udp.c
  9  *
 10  *      Fixes:
 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.
 17  *
 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.
 22  */
 23 
 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 <asm/uaccess.h>
 39 
 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>
 45 #include <net/raw.h>
 46 #include <net/tcp_states.h>
 47 #include <net/ip6_checksum.h>
 48 #include <net/xfrm.h>
 49 #include <net/inet6_hashtables.h>
 50 #include <net/busy_poll.h>
 51 #include <net/sock_reuseport.h>
 52 
 53 #include <linux/proc_fs.h>
 54 #include <linux/seq_file.h>
 55 #include <trace/events/skb.h>
 56 #include "udp_impl.h"
 57 
 58 static u32 udp6_ehashfn(const struct net *net,
 59                         const struct in6_addr *laddr,
 60                         const u16 lport,
 61                         const struct in6_addr *faddr,
 62                         const __be16 fport)
 63 {
 64         static u32 udp6_ehash_secret __read_mostly;
 65         static u32 udp_ipv6_hash_secret __read_mostly;
 66 
 67         u32 lhash, fhash;
 68 
 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));
 73 
 74         lhash = (__force u32)laddr->s6_addr32[3];
 75         fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret);
 76 
 77         return __inet6_ehashfn(lhash, lport, fhash, fport,
 78                                udp_ipv6_hash_secret + net_hash_mix(net));
 79 }
 80 
 81 static u32 udp6_portaddr_hash(const struct net *net,
 82                               const struct in6_addr *addr6,
 83                               unsigned int port)
 84 {
 85         unsigned int hash, mix = net_hash_mix(net);
 86 
 87         if (ipv6_addr_any(addr6))
 88                 hash = jhash_1word(0, mix);
 89         else if (ipv6_addr_v4mapped(addr6))
 90                 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
 91         else
 92                 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
 93 
 94         return hash ^ port;
 95 }
 96 
 97 int udp_v6_get_port(struct sock *sk, unsigned short snum)
 98 {
 99         unsigned int hash2_nulladdr =
100                 udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
101         unsigned int hash2_partial =
102                 udp6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
103 
104         /* precompute partial secondary hash */
105         udp_sk(sk)->udp_portaddr_hash = hash2_partial;
106         return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
107 }
108 
109 static void udp_v6_rehash(struct sock *sk)
110 {
111         u16 new_hash = udp6_portaddr_hash(sock_net(sk),
112                                           &sk->sk_v6_rcv_saddr,
113                                           inet_sk(sk)->inet_num);
114 
115         udp_lib_rehash(sk, new_hash);
116 }
117 
118 static int compute_score(struct sock *sk, struct net *net,
119                          const struct in6_addr *saddr, __be16 sport,
120                          const struct in6_addr *daddr, unsigned short hnum,
121                          int dif)
122 {
123         int score;
124         struct inet_sock *inet;
125 
126         if (!net_eq(sock_net(sk), net) ||
127             udp_sk(sk)->udp_port_hash != hnum ||
128             sk->sk_family != PF_INET6)
129                 return -1;
130 
131         score = 0;
132         inet = inet_sk(sk);
133 
134         if (inet->inet_dport) {
135                 if (inet->inet_dport != sport)
136                         return -1;
137                 score++;
138         }
139 
140         if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
141                 if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
142                         return -1;
143                 score++;
144         }
145 
146         if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
147                 if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
148                         return -1;
149                 score++;
150         }
151 
152         if (sk->sk_bound_dev_if) {
153                 if (sk->sk_bound_dev_if != dif)
154                         return -1;
155                 score++;
156         }
157 
158         if (sk->sk_incoming_cpu == raw_smp_processor_id())
159                 score++;
160 
161         return score;
162 }
163 
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, int dif,
168                 struct udp_hslot *hslot2,
169                 struct sk_buff *skb)
170 {
171         struct sock *sk, *result;
172         int score, badness, matches = 0, reuseport = 0;
173         u32 hash = 0;
174 
175         result = NULL;
176         badness = -1;
177         udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
178                 score = compute_score(sk, net, saddr, sport,
179                                       daddr, hnum, dif);
180                 if (score > badness) {
181                         reuseport = sk->sk_reuseport;
182                         if (reuseport) {
183                                 hash = udp6_ehashfn(net, daddr, hnum,
184                                                     saddr, sport);
185 
186                                 result = reuseport_select_sock(sk, hash, skb,
187                                                         sizeof(struct udphdr));
188                                 if (result)
189                                         return result;
190                                 matches = 1;
191                         }
192                         result = sk;
193                         badness = score;
194                 } else if (score == badness && reuseport) {
195                         matches++;
196                         if (reciprocal_scale(hash, matches) == 0)
197                                 result = sk;
198                         hash = next_pseudo_random32(hash);
199                 }
200         }
201         return result;
202 }
203 
204 /* rcu_read_lock() must be held */
205 struct sock *__udp6_lib_lookup(struct net *net,
206                                       const struct in6_addr *saddr, __be16 sport,
207                                       const struct in6_addr *daddr, __be16 dport,
208                                       int dif, struct udp_table *udptable,
209                                       struct sk_buff *skb)
210 {
211         struct sock *sk, *result;
212         unsigned short hnum = ntohs(dport);
213         unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
214         struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
215         int score, badness, matches = 0, reuseport = 0;
216         u32 hash = 0;
217 
218         if (hslot->count > 10) {
219                 hash2 = udp6_portaddr_hash(net, daddr, hnum);
220                 slot2 = hash2 & udptable->mask;
221                 hslot2 = &udptable->hash2[slot2];
222                 if (hslot->count < hslot2->count)
223                         goto begin;
224 
225                 result = udp6_lib_lookup2(net, saddr, sport,
226                                           daddr, hnum, dif,
227                                           hslot2, skb);
228                 if (!result) {
229                         unsigned int old_slot2 = slot2;
230                         hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
231                         slot2 = hash2 & udptable->mask;
232                         /* avoid searching the same slot again. */
233                         if (unlikely(slot2 == old_slot2))
234                                 return result;
235 
236                         hslot2 = &udptable->hash2[slot2];
237                         if (hslot->count < hslot2->count)
238                                 goto begin;
239 
240                         result = udp6_lib_lookup2(net, saddr, sport,
241                                                   daddr, hnum, dif,
242                                                   hslot2, skb);
243                 }
244                 return result;
245         }
246 begin:
247         result = NULL;
248         badness = -1;
249         sk_for_each_rcu(sk, &hslot->head) {
250                 score = compute_score(sk, net, saddr, sport, daddr, hnum, dif);
251                 if (score > badness) {
252                         reuseport = sk->sk_reuseport;
253                         if (reuseport) {
254                                 hash = udp6_ehashfn(net, daddr, hnum,
255                                                     saddr, sport);
256                                 result = reuseport_select_sock(sk, hash, skb,
257                                                         sizeof(struct udphdr));
258                                 if (result)
259                                         return result;
260                                 matches = 1;
261                         }
262                         result = sk;
263                         badness = score;
264                 } else if (score == badness && reuseport) {
265                         matches++;
266                         if (reciprocal_scale(hash, matches) == 0)
267                                 result = sk;
268                         hash = next_pseudo_random32(hash);
269                 }
270         }
271         return result;
272 }
273 EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
274 
275 static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
276                                           __be16 sport, __be16 dport,
277                                           struct udp_table *udptable)
278 {
279         const struct ipv6hdr *iph = ipv6_hdr(skb);
280         struct sock *sk;
281 
282         sk = skb_steal_sock(skb);
283         if (unlikely(sk))
284                 return sk;
285         return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport,
286                                  &iph->daddr, dport, inet6_iif(skb),
287                                  udptable, skb);
288 }
289 
290 struct sock *udp6_lib_lookup_skb(struct sk_buff *skb,
291                                  __be16 sport, __be16 dport)
292 {
293         const struct ipv6hdr *iph = ipv6_hdr(skb);
294 
295         return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport,
296                                  &iph->daddr, dport, inet6_iif(skb),
297                                  &udp_table, skb);
298 }
299 EXPORT_SYMBOL_GPL(udp6_lib_lookup_skb);
300 
301 /* Must be called under rcu_read_lock().
302  * Does increment socket refcount.
303  */
304 #if IS_ENABLED(CONFIG_NETFILTER_XT_MATCH_SOCKET) || \
305     IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY)
306 struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
307                              const struct in6_addr *daddr, __be16 dport, int dif)
308 {
309         struct sock *sk;
310 
311         sk =  __udp6_lib_lookup(net, saddr, sport, daddr, dport,
312                                 dif, &udp_table, NULL);
313         if (sk && !atomic_inc_not_zero(&sk->sk_refcnt))
314                 sk = NULL;
315         return sk;
316 }
317 EXPORT_SYMBOL_GPL(udp6_lib_lookup);
318 #endif
319 
320 /*
321  *      This should be easy, if there is something there we
322  *      return it, otherwise we block.
323  */
324 
325 int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
326                   int noblock, int flags, int *addr_len)
327 {
328         struct ipv6_pinfo *np = inet6_sk(sk);
329         struct inet_sock *inet = inet_sk(sk);
330         struct sk_buff *skb;
331         unsigned int ulen, copied;
332         int peeked, peeking, off;
333         int err;
334         int is_udplite = IS_UDPLITE(sk);
335         bool checksum_valid = false;
336         int is_udp4;
337         bool slow;
338 
339         if (flags & MSG_ERRQUEUE)
340                 return ipv6_recv_error(sk, msg, len, addr_len);
341 
342         if (np->rxpmtu && np->rxopt.bits.rxpmtu)
343                 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
344 
345 try_again:
346         peeking = off = sk_peek_offset(sk, flags);
347         skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
348                                   &peeked, &off, &err);
349         if (!skb)
350                 return err;
351         if (ccs_socket_post_recvmsg_permission(sk, skb, flags))
352                 return -EAGAIN; /* Hope less harmful than -EPERM. */
353 
354         ulen = skb->len;
355         copied = len;
356         if (copied > ulen - off)
357                 copied = ulen - off;
358         else if (copied < ulen)
359                 msg->msg_flags |= MSG_TRUNC;
360 
361         is_udp4 = (skb->protocol == htons(ETH_P_IP));
362 
363         /*
364          * If checksum is needed at all, try to do it while copying the
365          * data.  If the data is truncated, or if we only want a partial
366          * coverage checksum (UDP-Lite), do it before the copy.
367          */
368 
369         if (copied < ulen || UDP_SKB_CB(skb)->partial_cov || peeking) {
370                 checksum_valid = !udp_lib_checksum_complete(skb);
371                 if (!checksum_valid)
372                         goto csum_copy_err;
373         }
374 
375         if (checksum_valid || skb_csum_unnecessary(skb))
376                 err = skb_copy_datagram_msg(skb, off, msg, copied);
377         else {
378                 err = skb_copy_and_csum_datagram_msg(skb, off, msg);
379                 if (err == -EINVAL)
380                         goto csum_copy_err;
381         }
382         if (unlikely(err)) {
383                 trace_kfree_skb(skb, udpv6_recvmsg);
384                 if (!peeked) {
385                         atomic_inc(&sk->sk_drops);
386                         if (is_udp4)
387                                 UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
388                                               is_udplite);
389                         else
390                                 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
391                                                is_udplite);
392                 }
393                 skb_free_datagram_locked(sk, skb);
394                 return err;
395         }
396         if (!peeked) {
397                 if (is_udp4)
398                         UDP_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS,
399                                       is_udplite);
400                 else
401                         UDP6_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS,
402                                        is_udplite);
403         }
404 
405         sock_recv_ts_and_drops(msg, sk, skb);
406 
407         /* Copy the address. */
408         if (msg->msg_name) {
409                 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
410                 sin6->sin6_family = AF_INET6;
411                 sin6->sin6_port = udp_hdr(skb)->source;
412                 sin6->sin6_flowinfo = 0;
413 
414                 if (is_udp4) {
415                         ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
416                                                &sin6->sin6_addr);
417                         sin6->sin6_scope_id = 0;
418                 } else {
419                         sin6->sin6_addr = ipv6_hdr(skb)->saddr;
420                         sin6->sin6_scope_id =
421                                 ipv6_iface_scope_id(&sin6->sin6_addr,
422                                                     inet6_iif(skb));
423                 }
424                 *addr_len = sizeof(*sin6);
425         }
426 
427         if (np->rxopt.all)
428                 ip6_datagram_recv_common_ctl(sk, msg, skb);
429 
430         if (is_udp4) {
431                 if (inet->cmsg_flags)
432                         ip_cmsg_recv(msg, skb);
433         } else {
434                 if (np->rxopt.all)
435                         ip6_datagram_recv_specific_ctl(sk, msg, skb);
436         }
437 
438         err = copied;
439         if (flags & MSG_TRUNC)
440                 err = ulen;
441 
442         __skb_free_datagram_locked(sk, skb, peeking ? -err : err);
443         return err;
444 
445 csum_copy_err:
446         slow = lock_sock_fast(sk);
447         if (!skb_kill_datagram(sk, skb, flags)) {
448                 if (is_udp4) {
449                         UDP_INC_STATS(sock_net(sk),
450                                       UDP_MIB_CSUMERRORS, is_udplite);
451                         UDP_INC_STATS(sock_net(sk),
452                                       UDP_MIB_INERRORS, is_udplite);
453                 } else {
454                         UDP6_INC_STATS(sock_net(sk),
455                                        UDP_MIB_CSUMERRORS, is_udplite);
456                         UDP6_INC_STATS(sock_net(sk),
457                                        UDP_MIB_INERRORS, is_udplite);
458                 }
459         }
460         unlock_sock_fast(sk, slow);
461 
462         /* starting over for a new packet, but check if we need to yield */
463         cond_resched();
464         msg->msg_flags &= ~MSG_TRUNC;
465         goto try_again;
466 }
467 
468 void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
469                     u8 type, u8 code, int offset, __be32 info,
470                     struct udp_table *udptable)
471 {
472         struct ipv6_pinfo *np;
473         const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
474         const struct in6_addr *saddr = &hdr->saddr;
475         const struct in6_addr *daddr = &hdr->daddr;
476         struct udphdr *uh = (struct udphdr *)(skb->data+offset);
477         struct sock *sk;
478         int harderr;
479         int err;
480         struct net *net = dev_net(skb->dev);
481 
482         sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source,
483                                inet6_iif(skb), udptable, skb);
484         if (!sk) {
485                 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
486                                   ICMP6_MIB_INERRORS);
487                 return;
488         }
489 
490         harderr = icmpv6_err_convert(type, code, &err);
491         np = inet6_sk(sk);
492 
493         if (type == ICMPV6_PKT_TOOBIG) {
494                 if (!ip6_sk_accept_pmtu(sk))
495                         goto out;
496                 ip6_sk_update_pmtu(skb, sk, info);
497                 if (np->pmtudisc != IPV6_PMTUDISC_DONT)
498                         harderr = 1;
499         }
500         if (type == NDISC_REDIRECT) {
501                 ip6_sk_redirect(skb, sk);
502                 goto out;
503         }
504 
505         if (!np->recverr) {
506                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
507                         goto out;
508         } else {
509                 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
510         }
511 
512         sk->sk_err = err;
513         sk->sk_error_report(sk);
514 out:
515         return;
516 }
517 
518 static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
519 {
520         int rc;
521 
522         if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
523                 sock_rps_save_rxhash(sk, skb);
524                 sk_mark_napi_id(sk, skb);
525                 sk_incoming_cpu_update(sk);
526         }
527 
528         rc = __sock_queue_rcv_skb(sk, skb);
529         if (rc < 0) {
530                 int is_udplite = IS_UDPLITE(sk);
531 
532                 /* Note that an ENOMEM error is charged twice */
533                 if (rc == -ENOMEM)
534                         UDP6_INC_STATS(sock_net(sk),
535                                          UDP_MIB_RCVBUFERRORS, is_udplite);
536                 UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
537                 kfree_skb(skb);
538                 return -1;
539         }
540         return 0;
541 }
542 
543 static __inline__ void udpv6_err(struct sk_buff *skb,
544                                  struct inet6_skb_parm *opt, u8 type,
545                                  u8 code, int offset, __be32 info)
546 {
547         __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
548 }
549 
550 static struct static_key udpv6_encap_needed __read_mostly;
551 void udpv6_encap_enable(void)
552 {
553         if (!static_key_enabled(&udpv6_encap_needed))
554                 static_key_slow_inc(&udpv6_encap_needed);
555 }
556 EXPORT_SYMBOL(udpv6_encap_enable);
557 
558 int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
559 {
560         struct udp_sock *up = udp_sk(sk);
561         int rc;
562         int is_udplite = IS_UDPLITE(sk);
563 
564         if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
565                 goto drop;
566 
567         if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
568                 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
569 
570                 /*
571                  * This is an encapsulation socket so pass the skb to
572                  * the socket's udp_encap_rcv() hook. Otherwise, just
573                  * fall through and pass this up the UDP socket.
574                  * up->encap_rcv() returns the following value:
575                  * =0 if skb was successfully passed to the encap
576                  *    handler or was discarded by it.
577                  * >0 if skb should be passed on to UDP.
578                  * <0 if skb should be resubmitted as proto -N
579                  */
580 
581                 /* if we're overly short, let UDP handle it */
582                 encap_rcv = ACCESS_ONCE(up->encap_rcv);
583                 if (encap_rcv) {
584                         int ret;
585 
586                         /* Verify checksum before giving to encap */
587                         if (udp_lib_checksum_complete(skb))
588                                 goto csum_error;
589 
590                         ret = encap_rcv(sk, skb);
591                         if (ret <= 0) {
592                                 __UDP_INC_STATS(sock_net(sk),
593                                                 UDP_MIB_INDATAGRAMS,
594                                                 is_udplite);
595                                 return -ret;
596                         }
597                 }
598 
599                 /* FALLTHROUGH -- it's a UDP Packet */
600         }
601 
602         /*
603          * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
604          */
605         if ((is_udplite & UDPLITE_RECV_CC)  &&  UDP_SKB_CB(skb)->partial_cov) {
606 
607                 if (up->pcrlen == 0) {          /* full coverage was set  */
608                         net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
609                                             UDP_SKB_CB(skb)->cscov, skb->len);
610                         goto drop;
611                 }
612                 if (UDP_SKB_CB(skb)->cscov  <  up->pcrlen) {
613                         net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
614                                             UDP_SKB_CB(skb)->cscov, up->pcrlen);
615                         goto drop;
616                 }
617         }
618 
619         if (rcu_access_pointer(sk->sk_filter) &&
620             udp_lib_checksum_complete(skb))
621                 goto csum_error;
622 
623         if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr)))
624                 goto drop;
625 
626         udp_csum_pull_header(skb);
627         if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
628                 __UDP6_INC_STATS(sock_net(sk),
629                                  UDP_MIB_RCVBUFERRORS, is_udplite);
630                 goto drop;
631         }
632 
633         skb_dst_drop(skb);
634 
635         bh_lock_sock(sk);
636         rc = 0;
637         if (!sock_owned_by_user(sk))
638                 rc = __udpv6_queue_rcv_skb(sk, skb);
639         else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
640                 bh_unlock_sock(sk);
641                 goto drop;
642         }
643         bh_unlock_sock(sk);
644 
645         return rc;
646 
647 csum_error:
648         __UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
649 drop:
650         __UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
651         atomic_inc(&sk->sk_drops);
652         kfree_skb(skb);
653         return -1;
654 }
655 
656 static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
657                                    __be16 loc_port, const struct in6_addr *loc_addr,
658                                    __be16 rmt_port, const struct in6_addr *rmt_addr,
659                                    int dif, unsigned short hnum)
660 {
661         struct inet_sock *inet = inet_sk(sk);
662 
663         if (!net_eq(sock_net(sk), net))
664                 return false;
665 
666         if (udp_sk(sk)->udp_port_hash != hnum ||
667             sk->sk_family != PF_INET6 ||
668             (inet->inet_dport && inet->inet_dport != rmt_port) ||
669             (!ipv6_addr_any(&sk->sk_v6_daddr) &&
670                     !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
671             (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) ||
672             (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
673                     !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
674                 return false;
675         if (!inet6_mc_check(sk, loc_addr, rmt_addr))
676                 return false;
677         return true;
678 }
679 
680 static void udp6_csum_zero_error(struct sk_buff *skb)
681 {
682         /* RFC 2460 section 8.1 says that we SHOULD log
683          * this error. Well, it is reasonable.
684          */
685         net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
686                             &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
687                             &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
688 }
689 
690 /*
691  * Note: called only from the BH handler context,
692  * so we don't need to lock the hashes.
693  */
694 static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
695                 const struct in6_addr *saddr, const struct in6_addr *daddr,
696                 struct udp_table *udptable, int proto)
697 {
698         struct sock *sk, *first = NULL;
699         const struct udphdr *uh = udp_hdr(skb);
700         unsigned short hnum = ntohs(uh->dest);
701         struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
702         unsigned int offset = offsetof(typeof(*sk), sk_node);
703         unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
704         int dif = inet6_iif(skb);
705         struct hlist_node *node;
706         struct sk_buff *nskb;
707 
708         if (use_hash2) {
709                 hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
710                             udp_table.mask;
711                 hash2 = udp6_portaddr_hash(net, daddr, hnum) & udp_table.mask;
712 start_lookup:
713                 hslot = &udp_table.hash2[hash2];
714                 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
715         }
716 
717         sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
718                 if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr,
719                                             uh->source, saddr, dif, hnum))
720                         continue;
721                 /* If zero checksum and no_check is not on for
722                  * the socket then skip it.
723                  */
724                 if (!uh->check && !udp_sk(sk)->no_check6_rx)
725                         continue;
726                 if (!first) {
727                         first = sk;
728                         continue;
729                 }
730                 nskb = skb_clone(skb, GFP_ATOMIC);
731                 if (unlikely(!nskb)) {
732                         atomic_inc(&sk->sk_drops);
733                         __UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
734                                          IS_UDPLITE(sk));
735                         __UDP6_INC_STATS(net, UDP_MIB_INERRORS,
736                                          IS_UDPLITE(sk));
737                         continue;
738                 }
739 
740                 if (udpv6_queue_rcv_skb(sk, nskb) > 0)
741                         consume_skb(nskb);
742         }
743 
744         /* Also lookup *:port if we are using hash2 and haven't done so yet. */
745         if (use_hash2 && hash2 != hash2_any) {
746                 hash2 = hash2_any;
747                 goto start_lookup;
748         }
749 
750         if (first) {
751                 if (udpv6_queue_rcv_skb(first, skb) > 0)
752                         consume_skb(skb);
753         } else {
754                 kfree_skb(skb);
755                 __UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI,
756                                  proto == IPPROTO_UDPLITE);
757         }
758         return 0;
759 }
760 
761 int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
762                    int proto)
763 {
764         const struct in6_addr *saddr, *daddr;
765         struct net *net = dev_net(skb->dev);
766         struct udphdr *uh;
767         struct sock *sk;
768         u32 ulen = 0;
769 
770         if (!pskb_may_pull(skb, sizeof(struct udphdr)))
771                 goto discard;
772 
773         saddr = &ipv6_hdr(skb)->saddr;
774         daddr = &ipv6_hdr(skb)->daddr;
775         uh = udp_hdr(skb);
776 
777         ulen = ntohs(uh->len);
778         if (ulen > skb->len)
779                 goto short_packet;
780 
781         if (proto == IPPROTO_UDP) {
782                 /* UDP validates ulen. */
783 
784                 /* Check for jumbo payload */
785                 if (ulen == 0)
786                         ulen = skb->len;
787 
788                 if (ulen < sizeof(*uh))
789                         goto short_packet;
790 
791                 if (ulen < skb->len) {
792                         if (pskb_trim_rcsum(skb, ulen))
793                                 goto short_packet;
794                         saddr = &ipv6_hdr(skb)->saddr;
795                         daddr = &ipv6_hdr(skb)->daddr;
796                         uh = udp_hdr(skb);
797                 }
798         }
799 
800         if (udp6_csum_init(skb, uh, proto))
801                 goto csum_error;
802 
803         /*
804          *      Multicast receive code
805          */
806         if (ipv6_addr_is_multicast(daddr))
807                 return __udp6_lib_mcast_deliver(net, skb,
808                                 saddr, daddr, udptable, proto);
809 
810         /* Unicast */
811 
812         /*
813          * check socket cache ... must talk to Alan about his plans
814          * for sock caches... i'll skip this for now.
815          */
816         sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
817         if (sk) {
818                 int ret;
819 
820                 if (!uh->check && !udp_sk(sk)->no_check6_rx) {
821                         udp6_csum_zero_error(skb);
822                         goto csum_error;
823                 }
824 
825                 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
826                         skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
827                                                  ip6_compute_pseudo);
828 
829                 ret = udpv6_queue_rcv_skb(sk, skb);
830 
831                 /* a return value > 0 means to resubmit the input */
832                 if (ret > 0)
833                         return ret;
834 
835                 return 0;
836         }
837 
838         if (!uh->check) {
839                 udp6_csum_zero_error(skb);
840                 goto csum_error;
841         }
842 
843         if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
844                 goto discard;
845 
846         if (udp_lib_checksum_complete(skb))
847                 goto csum_error;
848 
849         __UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
850         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
851 
852         kfree_skb(skb);
853         return 0;
854 
855 short_packet:
856         net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
857                             proto == IPPROTO_UDPLITE ? "-Lite" : "",
858                             saddr, ntohs(uh->source),
859                             ulen, skb->len,
860                             daddr, ntohs(uh->dest));
861         goto discard;
862 csum_error:
863         __UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
864 discard:
865         __UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
866         kfree_skb(skb);
867         return 0;
868 }
869 
870 static __inline__ int udpv6_rcv(struct sk_buff *skb)
871 {
872         return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
873 }
874 
875 /*
876  * Throw away all pending data and cancel the corking. Socket is locked.
877  */
878 static void udp_v6_flush_pending_frames(struct sock *sk)
879 {
880         struct udp_sock *up = udp_sk(sk);
881 
882         if (up->pending == AF_INET)
883                 udp_flush_pending_frames(sk);
884         else if (up->pending) {
885                 up->len = 0;
886                 up->pending = 0;
887                 ip6_flush_pending_frames(sk);
888         }
889 }
890 
891 /**
892  *      udp6_hwcsum_outgoing  -  handle outgoing HW checksumming
893  *      @sk:    socket we are sending on
894  *      @skb:   sk_buff containing the filled-in UDP header
895  *              (checksum field must be zeroed out)
896  */
897 static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
898                                  const struct in6_addr *saddr,
899                                  const struct in6_addr *daddr, int len)
900 {
901         unsigned int offset;
902         struct udphdr *uh = udp_hdr(skb);
903         struct sk_buff *frags = skb_shinfo(skb)->frag_list;
904         __wsum csum = 0;
905 
906         if (!frags) {
907                 /* Only one fragment on the socket.  */
908                 skb->csum_start = skb_transport_header(skb) - skb->head;
909                 skb->csum_offset = offsetof(struct udphdr, check);
910                 uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
911         } else {
912                 /*
913                  * HW-checksum won't work as there are two or more
914                  * fragments on the socket so that all csums of sk_buffs
915                  * should be together
916                  */
917                 offset = skb_transport_offset(skb);
918                 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
919 
920                 skb->ip_summed = CHECKSUM_NONE;
921 
922                 do {
923                         csum = csum_add(csum, frags->csum);
924                 } while ((frags = frags->next));
925 
926                 uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
927                                             csum);
928                 if (uh->check == 0)
929                         uh->check = CSUM_MANGLED_0;
930         }
931 }
932 
933 /*
934  *      Sending
935  */
936 
937 static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6)
938 {
939         struct sock *sk = skb->sk;
940         struct udphdr *uh;
941         int err = 0;
942         int is_udplite = IS_UDPLITE(sk);
943         __wsum csum = 0;
944         int offset = skb_transport_offset(skb);
945         int len = skb->len - offset;
946 
947         /*
948          * Create a UDP header
949          */
950         uh = udp_hdr(skb);
951         uh->source = fl6->fl6_sport;
952         uh->dest = fl6->fl6_dport;
953         uh->len = htons(len);
954         uh->check = 0;
955 
956         if (is_udplite)
957                 csum = udplite_csum(skb);
958         else if (udp_sk(sk)->no_check6_tx) {   /* UDP csum disabled */
959                 skb->ip_summed = CHECKSUM_NONE;
960                 goto send;
961         } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
962                 udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
963                 goto send;
964         } else
965                 csum = udp_csum(skb);
966 
967         /* add protocol-dependent pseudo-header */
968         uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
969                                     len, fl6->flowi6_proto, csum);
970         if (uh->check == 0)
971                 uh->check = CSUM_MANGLED_0;
972 
973 send:
974         err = ip6_send_skb(skb);
975         if (err) {
976                 if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
977                         UDP6_INC_STATS(sock_net(sk),
978                                        UDP_MIB_SNDBUFERRORS, is_udplite);
979                         err = 0;
980                 }
981         } else {
982                 UDP6_INC_STATS(sock_net(sk),
983                                UDP_MIB_OUTDATAGRAMS, is_udplite);
984         }
985         return err;
986 }
987 
988 static int udp_v6_push_pending_frames(struct sock *sk)
989 {
990         struct sk_buff *skb;
991         struct udp_sock  *up = udp_sk(sk);
992         struct flowi6 fl6;
993         int err = 0;
994 
995         if (up->pending == AF_INET)
996                 return udp_push_pending_frames(sk);
997 
998         /* ip6_finish_skb will release the cork, so make a copy of
999          * fl6 here.
1000          */
1001         fl6 = inet_sk(sk)->cork.fl.u.ip6;
1002 
1003         skb = ip6_finish_skb(sk);
1004         if (!skb)
1005                 goto out;
1006 
1007         err = udp_v6_send_skb(skb, &fl6);
1008 
1009 out:
1010         up->len = 0;
1011         up->pending = 0;
1012         return err;
1013 }
1014 
1015 int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1016 {
1017         struct ipv6_txoptions opt_space;
1018         struct udp_sock *up = udp_sk(sk);
1019         struct inet_sock *inet = inet_sk(sk);
1020         struct ipv6_pinfo *np = inet6_sk(sk);
1021         DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1022         struct in6_addr *daddr, *final_p, final;
1023         struct ipv6_txoptions *opt = NULL;
1024         struct ipv6_txoptions *opt_to_free = NULL;
1025         struct ip6_flowlabel *flowlabel = NULL;
1026         struct flowi6 fl6;
1027         struct dst_entry *dst;
1028         struct ipcm6_cookie ipc6;
1029         int addr_len = msg->msg_namelen;
1030         int ulen = len;
1031         int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1032         int err;
1033         int connected = 0;
1034         int is_udplite = IS_UDPLITE(sk);
1035         int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1036         struct sockcm_cookie sockc;
1037 
1038         ipc6.hlimit = -1;
1039         ipc6.tclass = -1;
1040         ipc6.dontfrag = -1;
1041 
1042         /* destination address check */
1043         if (sin6) {
1044                 if (addr_len < offsetof(struct sockaddr, sa_data))
1045                         return -EINVAL;
1046 
1047                 switch (sin6->sin6_family) {
1048                 case AF_INET6:
1049                         if (addr_len < SIN6_LEN_RFC2133)
1050                                 return -EINVAL;
1051                         daddr = &sin6->sin6_addr;
1052                         break;
1053                 case AF_INET:
1054                         goto do_udp_sendmsg;
1055                 case AF_UNSPEC:
1056                         msg->msg_name = sin6 = NULL;
1057                         msg->msg_namelen = addr_len = 0;
1058                         daddr = NULL;
1059                         break;
1060                 default:
1061                         return -EINVAL;
1062                 }
1063         } else if (!up->pending) {
1064                 if (sk->sk_state != TCP_ESTABLISHED)
1065                         return -EDESTADDRREQ;
1066                 daddr = &sk->sk_v6_daddr;
1067         } else
1068                 daddr = NULL;
1069 
1070         if (daddr) {
1071                 if (ipv6_addr_v4mapped(daddr)) {
1072                         struct sockaddr_in sin;
1073                         sin.sin_family = AF_INET;
1074                         sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1075                         sin.sin_addr.s_addr = daddr->s6_addr32[3];
1076                         msg->msg_name = &sin;
1077                         msg->msg_namelen = sizeof(sin);
1078 do_udp_sendmsg:
1079                         if (__ipv6_only_sock(sk))
1080                                 return -ENETUNREACH;
1081                         return udp_sendmsg(sk, msg, len);
1082                 }
1083         }
1084 
1085         if (up->pending == AF_INET)
1086                 return udp_sendmsg(sk, msg, len);
1087 
1088         /* Rough check on arithmetic overflow,
1089            better check is made in ip6_append_data().
1090            */
1091         if (len > INT_MAX - sizeof(struct udphdr))
1092                 return -EMSGSIZE;
1093 
1094         getfrag  =  is_udplite ?  udplite_getfrag : ip_generic_getfrag;
1095         if (up->pending) {
1096                 /*
1097                  * There are pending frames.
1098                  * The socket lock must be held while it's corked.
1099                  */
1100                 lock_sock(sk);
1101                 if (likely(up->pending)) {
1102                         if (unlikely(up->pending != AF_INET6)) {
1103                                 release_sock(sk);
1104                                 return -EAFNOSUPPORT;
1105                         }
1106                         dst = NULL;
1107                         goto do_append_data;
1108                 }
1109                 release_sock(sk);
1110         }
1111         ulen += sizeof(struct udphdr);
1112 
1113         memset(&fl6, 0, sizeof(fl6));
1114 
1115         if (sin6) {
1116                 if (sin6->sin6_port == 0)
1117                         return -EINVAL;
1118 
1119                 fl6.fl6_dport = sin6->sin6_port;
1120                 daddr = &sin6->sin6_addr;
1121 
1122                 if (np->sndflow) {
1123                         fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1124                         if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1125                                 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1126                                 if (!flowlabel)
1127                                         return -EINVAL;
1128                         }
1129                 }
1130 
1131                 /*
1132                  * Otherwise it will be difficult to maintain
1133                  * sk->sk_dst_cache.
1134                  */
1135                 if (sk->sk_state == TCP_ESTABLISHED &&
1136                     ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1137                         daddr = &sk->sk_v6_daddr;
1138 
1139                 if (addr_len >= sizeof(struct sockaddr_in6) &&
1140                     sin6->sin6_scope_id &&
1141                     __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1142                         fl6.flowi6_oif = sin6->sin6_scope_id;
1143         } else {
1144                 if (sk->sk_state != TCP_ESTABLISHED)
1145                         return -EDESTADDRREQ;
1146 
1147                 fl6.fl6_dport = inet->inet_dport;
1148                 daddr = &sk->sk_v6_daddr;
1149                 fl6.flowlabel = np->flow_label;
1150                 connected = 1;
1151         }
1152 
1153         if (!fl6.flowi6_oif)
1154                 fl6.flowi6_oif = sk->sk_bound_dev_if;
1155 
1156         if (!fl6.flowi6_oif)
1157                 fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1158 
1159         fl6.flowi6_mark = sk->sk_mark;
1160         sockc.tsflags = sk->sk_tsflags;
1161 
1162         if (msg->msg_controllen) {
1163                 opt = &opt_space;
1164                 memset(opt, 0, sizeof(struct ipv6_txoptions));
1165                 opt->tot_len = sizeof(*opt);
1166                 ipc6.opt = opt;
1167 
1168                 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6, &sockc);
1169                 if (err < 0) {
1170                         fl6_sock_release(flowlabel);
1171                         return err;
1172                 }
1173                 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1174                         flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1175                         if (!flowlabel)
1176                                 return -EINVAL;
1177                 }
1178                 if (!(opt->opt_nflen|opt->opt_flen))
1179                         opt = NULL;
1180                 connected = 0;
1181         }
1182         if (!opt) {
1183                 opt = txopt_get(np);
1184                 opt_to_free = opt;
1185         }
1186         if (flowlabel)
1187                 opt = fl6_merge_options(&opt_space, flowlabel, opt);
1188         opt = ipv6_fixup_options(&opt_space, opt);
1189         ipc6.opt = opt;
1190 
1191         fl6.flowi6_proto = sk->sk_protocol;
1192         if (!ipv6_addr_any(daddr))
1193                 fl6.daddr = *daddr;
1194         else
1195                 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1196         if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1197                 fl6.saddr = np->saddr;
1198         fl6.fl6_sport = inet->inet_sport;
1199 
1200         final_p = fl6_update_dst(&fl6, opt, &final);
1201         if (final_p)
1202                 connected = 0;
1203 
1204         if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1205                 fl6.flowi6_oif = np->mcast_oif;
1206                 connected = 0;
1207         } else if (!fl6.flowi6_oif)
1208                 fl6.flowi6_oif = np->ucast_oif;
1209 
1210         security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
1211 
1212         dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
1213         if (IS_ERR(dst)) {
1214                 err = PTR_ERR(dst);
1215                 dst = NULL;
1216                 goto out;
1217         }
1218 
1219         if (ipc6.hlimit < 0)
1220                 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1221 
1222         if (ipc6.tclass < 0)
1223                 ipc6.tclass = np->tclass;
1224 
1225         if (msg->msg_flags&MSG_CONFIRM)
1226                 goto do_confirm;
1227 back_from_confirm:
1228 
1229         /* Lockless fast path for the non-corking case */
1230         if (!corkreq) {
1231                 struct sk_buff *skb;
1232 
1233                 skb = ip6_make_skb(sk, getfrag, msg, ulen,
1234                                    sizeof(struct udphdr), &ipc6,
1235                                    &fl6, (struct rt6_info *)dst,
1236                                    msg->msg_flags, &sockc);
1237                 err = PTR_ERR(skb);
1238                 if (!IS_ERR_OR_NULL(skb))
1239                         err = udp_v6_send_skb(skb, &fl6);
1240                 goto release_dst;
1241         }
1242 
1243         lock_sock(sk);
1244         if (unlikely(up->pending)) {
1245                 /* The socket is already corked while preparing it. */
1246                 /* ... which is an evident application bug. --ANK */
1247                 release_sock(sk);
1248 
1249                 net_dbg_ratelimited("udp cork app bug 2\n");
1250                 err = -EINVAL;
1251                 goto out;
1252         }
1253 
1254         up->pending = AF_INET6;
1255 
1256 do_append_data:
1257         if (ipc6.dontfrag < 0)
1258                 ipc6.dontfrag = np->dontfrag;
1259         up->len += ulen;
1260         err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr),
1261                               &ipc6, &fl6, (struct rt6_info *)dst,
1262                               corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, &sockc);
1263         if (err)
1264                 udp_v6_flush_pending_frames(sk);
1265         else if (!corkreq)
1266                 err = udp_v6_push_pending_frames(sk);
1267         else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1268                 up->pending = 0;
1269 
1270         if (err > 0)
1271                 err = np->recverr ? net_xmit_errno(err) : 0;
1272         release_sock(sk);
1273 
1274 release_dst:
1275         if (dst) {
1276                 if (connected) {
1277                         ip6_dst_store(sk, dst,
1278                                       ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1279                                       &sk->sk_v6_daddr : NULL,
1280 #ifdef CONFIG_IPV6_SUBTREES
1281                                       ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
1282                                       &np->saddr :
1283 #endif
1284                                       NULL);
1285                 } else {
1286                         dst_release(dst);
1287                 }
1288                 dst = NULL;
1289         }
1290 
1291 out:
1292         dst_release(dst);
1293         fl6_sock_release(flowlabel);
1294         txopt_put(opt_to_free);
1295         if (!err)
1296                 return len;
1297         /*
1298          * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space.  Reporting
1299          * ENOBUFS might not be good (it's not tunable per se), but otherwise
1300          * we don't have a good statistic (IpOutDiscards but it can be too many
1301          * things).  We could add another new stat but at least for now that
1302          * seems like overkill.
1303          */
1304         if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1305                 UDP6_INC_STATS(sock_net(sk),
1306                                UDP_MIB_SNDBUFERRORS, is_udplite);
1307         }
1308         return err;
1309 
1310 do_confirm:
1311         dst_confirm(dst);
1312         if (!(msg->msg_flags&MSG_PROBE) || len)
1313                 goto back_from_confirm;
1314         err = 0;
1315         goto out;
1316 }
1317 
1318 void udpv6_destroy_sock(struct sock *sk)
1319 {
1320         struct udp_sock *up = udp_sk(sk);
1321         lock_sock(sk);
1322         udp_v6_flush_pending_frames(sk);
1323         release_sock(sk);
1324 
1325         if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1326                 void (*encap_destroy)(struct sock *sk);
1327                 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1328                 if (encap_destroy)
1329                         encap_destroy(sk);
1330         }
1331 
1332         inet6_destroy_sock(sk);
1333 }
1334 
1335 /*
1336  *      Socket option code for UDP
1337  */
1338 int udpv6_setsockopt(struct sock *sk, int level, int optname,
1339                      char __user *optval, unsigned int optlen)
1340 {
1341         if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1342                 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1343                                           udp_v6_push_pending_frames);
1344         return ipv6_setsockopt(sk, level, optname, optval, optlen);
1345 }
1346 
1347 #ifdef CONFIG_COMPAT
1348 int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1349                             char __user *optval, unsigned int optlen)
1350 {
1351         if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1352                 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1353                                           udp_v6_push_pending_frames);
1354         return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1355 }
1356 #endif
1357 
1358 int udpv6_getsockopt(struct sock *sk, int level, int optname,
1359                      char __user *optval, int __user *optlen)
1360 {
1361         if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1362                 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1363         return ipv6_getsockopt(sk, level, optname, optval, optlen);
1364 }
1365 
1366 #ifdef CONFIG_COMPAT
1367 int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1368                             char __user *optval, int __user *optlen)
1369 {
1370         if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1371                 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1372         return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1373 }
1374 #endif
1375 
1376 static const struct inet6_protocol udpv6_protocol = {
1377         .handler        =       udpv6_rcv,
1378         .err_handler    =       udpv6_err,
1379         .flags          =       INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1380 };
1381 
1382 /* ------------------------------------------------------------------------ */
1383 #ifdef CONFIG_PROC_FS
1384 int udp6_seq_show(struct seq_file *seq, void *v)
1385 {
1386         if (v == SEQ_START_TOKEN) {
1387                 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1388         } else {
1389                 int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1390                 struct inet_sock *inet = inet_sk(v);
1391                 __u16 srcp = ntohs(inet->inet_sport);
1392                 __u16 destp = ntohs(inet->inet_dport);
1393                 ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1394         }
1395         return 0;
1396 }
1397 
1398 static const struct file_operations udp6_afinfo_seq_fops = {
1399         .owner    = THIS_MODULE,
1400         .open     = udp_seq_open,
1401         .read     = seq_read,
1402         .llseek   = seq_lseek,
1403         .release  = seq_release_net
1404 };
1405 
1406 static struct udp_seq_afinfo udp6_seq_afinfo = {
1407         .name           = "udp6",
1408         .family         = AF_INET6,
1409         .udp_table      = &udp_table,
1410         .seq_fops       = &udp6_afinfo_seq_fops,
1411         .seq_ops        = {
1412                 .show           = udp6_seq_show,
1413         },
1414 };
1415 
1416 int __net_init udp6_proc_init(struct net *net)
1417 {
1418         return udp_proc_register(net, &udp6_seq_afinfo);
1419 }
1420 
1421 void udp6_proc_exit(struct net *net)
1422 {
1423         udp_proc_unregister(net, &udp6_seq_afinfo);
1424 }
1425 #endif /* CONFIG_PROC_FS */
1426 
1427 void udp_v6_clear_sk(struct sock *sk, int size)
1428 {
1429         struct inet_sock *inet = inet_sk(sk);
1430 
1431         /* we do not want to clear pinet6 field, because of RCU lookups */
1432         sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1433 
1434         size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1435         memset(&inet->pinet6 + 1, 0, size);
1436 }
1437 
1438 /* ------------------------------------------------------------------------ */
1439 
1440 struct proto udpv6_prot = {
1441         .name              = "UDPv6",
1442         .owner             = THIS_MODULE,
1443         .close             = udp_lib_close,
1444         .connect           = ip6_datagram_connect,
1445         .disconnect        = udp_disconnect,
1446         .ioctl             = udp_ioctl,
1447         .destroy           = udpv6_destroy_sock,
1448         .setsockopt        = udpv6_setsockopt,
1449         .getsockopt        = udpv6_getsockopt,
1450         .sendmsg           = udpv6_sendmsg,
1451         .recvmsg           = udpv6_recvmsg,
1452         .backlog_rcv       = __udpv6_queue_rcv_skb,
1453         .release_cb        = ip6_datagram_release_cb,
1454         .hash              = udp_lib_hash,
1455         .unhash            = udp_lib_unhash,
1456         .rehash            = udp_v6_rehash,
1457         .get_port          = udp_v6_get_port,
1458         .memory_allocated  = &udp_memory_allocated,
1459         .sysctl_mem        = sysctl_udp_mem,
1460         .sysctl_wmem       = &sysctl_udp_wmem_min,
1461         .sysctl_rmem       = &sysctl_udp_rmem_min,
1462         .obj_size          = sizeof(struct udp6_sock),
1463         .slab_flags        = SLAB_DESTROY_BY_RCU,
1464         .h.udp_table       = &udp_table,
1465 #ifdef CONFIG_COMPAT
1466         .compat_setsockopt = compat_udpv6_setsockopt,
1467         .compat_getsockopt = compat_udpv6_getsockopt,
1468 #endif
1469         .clear_sk          = udp_v6_clear_sk,
1470 };
1471 
1472 static struct inet_protosw udpv6_protosw = {
1473         .type =      SOCK_DGRAM,
1474         .protocol =  IPPROTO_UDP,
1475         .prot =      &udpv6_prot,
1476         .ops =       &inet6_dgram_ops,
1477         .flags =     INET_PROTOSW_PERMANENT,
1478 };
1479 
1480 int __init udpv6_init(void)
1481 {
1482         int ret;
1483 
1484         ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1485         if (ret)
1486                 goto out;
1487 
1488         ret = inet6_register_protosw(&udpv6_protosw);
1489         if (ret)
1490                 goto out_udpv6_protocol;
1491 out:
1492         return ret;
1493 
1494 out_udpv6_protocol:
1495         inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1496         goto out;
1497 }
1498 
1499 void udpv6_exit(void)
1500 {
1501         inet6_unregister_protosw(&udpv6_protosw);
1502         inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1503 }
1504 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp