Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * UDP over IPv6 | |
1ab1457c | 3 | * Linux INET6 implementation |
1da177e4 LT |
4 | * |
5 | * Authors: | |
1ab1457c | 6 | * Pedro Roque <roque@di.fc.ul.pt> |
1da177e4 LT |
7 | * |
8 | * Based on linux/ipv4/udp.c | |
9 | * | |
1da177e4 LT |
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 | ||
1da177e4 LT |
24 | #include <linux/errno.h> |
25 | #include <linux/types.h> | |
26 | #include <linux/socket.h> | |
27 | #include <linux/sockios.h> | |
1da177e4 LT |
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> | |
1781f7f5 | 35 | #include <linux/module.h> |
3305b80c | 36 | #include <linux/skbuff.h> |
5a0e3ad6 | 37 | #include <linux/slab.h> |
7c0f6ba6 | 38 | #include <linux/uaccess.h> |
1da177e4 | 39 | |
496611d7 | 40 | #include <net/addrconf.h> |
1da177e4 LT |
41 | #include <net/ndisc.h> |
42 | #include <net/protocol.h> | |
43 | #include <net/transp_v6.h> | |
44 | #include <net/ip6_route.h> | |
1da177e4 | 45 | #include <net/raw.h> |
c752f073 | 46 | #include <net/tcp_states.h> |
1da177e4 | 47 | #include <net/ip6_checksum.h> |
e7cc0824 | 48 | #include <net/ip6_tunnel.h> |
1da177e4 | 49 | #include <net/xfrm.h> |
0bd84065 | 50 | #include <net/inet_hashtables.h> |
72289b96 | 51 | #include <net/inet6_hashtables.h> |
076bb0c8 | 52 | #include <net/busy_poll.h> |
e32ea7e7 | 53 | #include <net/sock_reuseport.h> |
1da177e4 LT |
54 | |
55 | #include <linux/proc_fs.h> | |
56 | #include <linux/seq_file.h> | |
22911fc5 | 57 | #include <trace/events/skb.h> |
ba4e58ec | 58 | #include "udp_impl.h" |
1da177e4 | 59 | |
63a6fff3 RS |
60 | static bool udp6_lib_exact_dif_match(struct net *net, struct sk_buff *skb) |
61 | { | |
62 | #if defined(CONFIG_NET_L3_MASTER_DEV) | |
63 | if (!net->ipv4.sysctl_udp_l3mdev_accept && | |
64 | skb && ipv6_l3mdev_skb(IP6CB(skb)->flags)) | |
65 | return true; | |
66 | #endif | |
67 | return false; | |
68 | } | |
69 | ||
6eada011 ED |
70 | static u32 udp6_ehashfn(const struct net *net, |
71 | const struct in6_addr *laddr, | |
72 | const u16 lport, | |
73 | const struct in6_addr *faddr, | |
74 | const __be16 fport) | |
b50026b5 | 75 | { |
1bbdceef HFS |
76 | static u32 udp6_ehash_secret __read_mostly; |
77 | static u32 udp_ipv6_hash_secret __read_mostly; | |
78 | ||
79 | u32 lhash, fhash; | |
80 | ||
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)); | |
85 | ||
86 | lhash = (__force u32)laddr->s6_addr32[3]; | |
87 | fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret); | |
88 | ||
b50026b5 | 89 | return __inet6_ehashfn(lhash, lport, fhash, fport, |
1bbdceef | 90 | udp_ipv6_hash_secret + net_hash_mix(net)); |
b50026b5 HFS |
91 | } |
92 | ||
6ba5a3c5 | 93 | int udp_v6_get_port(struct sock *sk, unsigned short snum) |
1da177e4 | 94 | { |
30fff923 | 95 | unsigned int hash2_nulladdr = |
f0b1e64c | 96 | ipv6_portaddr_hash(sock_net(sk), &in6addr_any, snum); |
9a52e97e | 97 | unsigned int hash2_partial = |
f0b1e64c | 98 | ipv6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0); |
30fff923 | 99 | |
d4cada4a | 100 | /* precompute partial secondary hash */ |
30fff923 | 101 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
fe38d2a1 | 102 | return udp_lib_get_port(sk, snum, hash2_nulladdr); |
1da177e4 LT |
103 | } |
104 | ||
719f8358 ED |
105 | static void udp_v6_rehash(struct sock *sk) |
106 | { | |
f0b1e64c | 107 | u16 new_hash = ipv6_portaddr_hash(sock_net(sk), |
efe4208f | 108 | &sk->sk_v6_rcv_saddr, |
719f8358 ED |
109 | inet_sk(sk)->inet_num); |
110 | ||
111 | udp_lib_rehash(sk, new_hash); | |
112 | } | |
113 | ||
d1e37288 SX |
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, | |
1801b570 | 117 | int dif, int sdif, bool exact_dif) |
645ca708 | 118 | { |
60c04aec JP |
119 | int score; |
120 | struct inet_sock *inet; | |
6da5b0f0 | 121 | bool dev_match; |
60c04aec JP |
122 | |
123 | if (!net_eq(sock_net(sk), net) || | |
124 | udp_sk(sk)->udp_port_hash != hnum || | |
125 | sk->sk_family != PF_INET6) | |
126 | return -1; | |
127 | ||
23b0269e PO |
128 | if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr)) |
129 | return -1; | |
130 | ||
60c04aec JP |
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 | ||
60c04aec JP |
140 | if (!ipv6_addr_any(&sk->sk_v6_daddr)) { |
141 | if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr)) | |
142 | return -1; | |
143 | score++; | |
144 | } | |
145 | ||
6da5b0f0 MM |
146 | dev_match = udp_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif); |
147 | if (!dev_match) | |
148 | return -1; | |
149 | score++; | |
60c04aec | 150 | |
70da268b ED |
151 | if (sk->sk_incoming_cpu == raw_smp_processor_id()) |
152 | score++; | |
153 | ||
645ca708 ED |
154 | return score; |
155 | } | |
156 | ||
d1e37288 | 157 | /* called with rcu_read_lock() */ |
fddc17de ED |
158 | static struct sock *udp6_lib_lookup2(struct net *net, |
159 | const struct in6_addr *saddr, __be16 sport, | |
1801b570 DA |
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) | |
fddc17de ED |
163 | { |
164 | struct sock *sk, *result; | |
e94a62f5 | 165 | int score, badness; |
72289b96 | 166 | u32 hash = 0; |
fddc17de | 167 | |
fddc17de ED |
168 | result = NULL; |
169 | badness = -1; | |
ca065d0c | 170 | udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
d1e37288 | 171 | score = compute_score(sk, net, saddr, sport, |
1801b570 | 172 | daddr, hnum, dif, sdif, exact_dif); |
fddc17de | 173 | if (score > badness) { |
e94a62f5 | 174 | if (sk->sk_reuseport) { |
b50026b5 HFS |
175 | hash = udp6_ehashfn(net, daddr, hnum, |
176 | saddr, sport); | |
ed0dfffd | 177 | |
ca065d0c | 178 | result = reuseport_select_sock(sk, hash, skb, |
ed0dfffd | 179 | sizeof(struct udphdr)); |
ca065d0c ED |
180 | if (result) |
181 | return result; | |
70da268b | 182 | } |
ca065d0c ED |
183 | result = sk; |
184 | badness = score; | |
fddc17de ED |
185 | } |
186 | } | |
fddc17de ED |
187 | return result; |
188 | } | |
189 | ||
ca065d0c | 190 | /* rcu_read_lock() must be held */ |
fce82338 | 191 | struct sock *__udp6_lib_lookup(struct net *net, |
1801b570 DA |
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, | |
195 | struct sk_buff *skb) | |
1da177e4 | 196 | { |
1da177e4 | 197 | unsigned short hnum = ntohs(dport); |
23b0269e PO |
198 | unsigned int hash2, slot2; |
199 | struct udp_hslot *hslot2; | |
200 | struct sock *result; | |
63a6fff3 | 201 | bool exact_dif = udp6_lib_exact_dif_match(net, skb); |
645ca708 | 202 | |
23b0269e PO |
203 | hash2 = ipv6_portaddr_hash(net, daddr, hnum); |
204 | slot2 = hash2 & udptable->mask; | |
205 | hslot2 = &udptable->hash2[slot2]; | |
206 | ||
207 | result = udp6_lib_lookup2(net, saddr, sport, | |
208 | daddr, hnum, dif, sdif, exact_dif, | |
209 | hslot2, skb); | |
210 | if (!result) { | |
211 | hash2 = ipv6_portaddr_hash(net, &in6addr_any, hnum); | |
fddc17de | 212 | slot2 = hash2 & udptable->mask; |
23b0269e | 213 | |
fddc17de | 214 | hslot2 = &udptable->hash2[slot2]; |
fddc17de ED |
215 | |
216 | result = udp6_lib_lookup2(net, saddr, sport, | |
23b0269e PO |
217 | &in6addr_any, hnum, dif, sdif, |
218 | exact_dif, hslot2, | |
219 | skb); | |
1da177e4 | 220 | } |
23b0269e PO |
221 | if (unlikely(IS_ERR(result))) |
222 | return NULL; | |
1da177e4 LT |
223 | return result; |
224 | } | |
fce82338 | 225 | EXPORT_SYMBOL_GPL(__udp6_lib_lookup); |
1da177e4 | 226 | |
607c4aaf KK |
227 | static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb, |
228 | __be16 sport, __be16 dport, | |
645ca708 | 229 | struct udp_table *udptable) |
607c4aaf | 230 | { |
b71d1d42 | 231 | const struct ipv6hdr *iph = ipv6_hdr(skb); |
607c4aaf | 232 | |
ed7cbbce | 233 | return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport, |
adf30907 | 234 | &iph->daddr, dport, inet6_iif(skb), |
1801b570 | 235 | inet6_sdif(skb), udptable, skb); |
607c4aaf KK |
236 | } |
237 | ||
63058308 TH |
238 | struct sock *udp6_lib_lookup_skb(struct sk_buff *skb, |
239 | __be16 sport, __be16 dport) | |
240 | { | |
241 | const struct ipv6hdr *iph = ipv6_hdr(skb); | |
63058308 | 242 | |
ed7cbbce | 243 | return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport, |
63058308 | 244 | &iph->daddr, dport, inet6_iif(skb), |
1801b570 | 245 | inet6_sdif(skb), &udp_table, skb); |
63058308 TH |
246 | } |
247 | EXPORT_SYMBOL_GPL(udp6_lib_lookup_skb); | |
248 | ||
ca065d0c ED |
249 | /* Must be called under rcu_read_lock(). |
250 | * Does increment socket refcount. | |
251 | */ | |
6e86000c | 252 | #if IS_ENABLED(CONFIG_NF_TPROXY_IPV6) || IS_ENABLED(CONFIG_NF_SOCKET_IPV6) |
aa976fc0 BS |
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) | |
255 | { | |
ca065d0c ED |
256 | struct sock *sk; |
257 | ||
258 | sk = __udp6_lib_lookup(net, saddr, sport, daddr, dport, | |
1801b570 | 259 | dif, 0, &udp_table, NULL); |
41c6d650 | 260 | if (sk && !refcount_inc_not_zero(&sk->sk_refcnt)) |
ca065d0c ED |
261 | sk = NULL; |
262 | return sk; | |
aa976fc0 BS |
263 | } |
264 | EXPORT_SYMBOL_GPL(udp6_lib_lookup); | |
ca065d0c | 265 | #endif |
aa976fc0 | 266 | |
cb891fa6 PA |
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 | |
270 | */ | |
271 | static int udp6_skb_len(struct sk_buff *skb) | |
272 | { | |
273 | return unlikely(inet6_is_jumbogram(skb)) ? skb->len : udp_skb_len(skb); | |
274 | } | |
275 | ||
1da177e4 | 276 | /* |
67ba4152 IM |
277 | * This should be easy, if there is something there we |
278 | * return it, otherwise we block. | |
1da177e4 LT |
279 | */ |
280 | ||
1b784140 | 281 | int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, |
1da177e4 LT |
282 | int noblock, int flags, int *addr_len) |
283 | { | |
284 | struct ipv6_pinfo *np = inet6_sk(sk); | |
285 | struct inet_sock *inet = inet_sk(sk); | |
1ab1457c | 286 | struct sk_buff *skb; |
59c2cdae | 287 | unsigned int ulen, copied; |
627d2d6b | 288 | int peeked, peeking, off; |
759e5d00 HX |
289 | int err; |
290 | int is_udplite = IS_UDPLITE(sk); | |
197c949e | 291 | bool checksum_valid = false; |
029a3743 | 292 | struct udp_mib *mib; |
f26ba175 | 293 | int is_udp4; |
1da177e4 | 294 | |
1da177e4 | 295 | if (flags & MSG_ERRQUEUE) |
85fbaa75 | 296 | return ipv6_recv_error(sk, msg, len, addr_len); |
1da177e4 | 297 | |
4b340ae2 | 298 | if (np->rxpmtu && np->rxopt.bits.rxpmtu) |
85fbaa75 | 299 | return ipv6_recv_rxpmtu(sk, msg, len, addr_len); |
4b340ae2 | 300 | |
1da177e4 | 301 | try_again: |
a0917e0b MD |
302 | peeking = flags & MSG_PEEK; |
303 | off = sk_peek_offset(sk, flags); | |
7c13f97f | 304 | skb = __skb_recv_udp(sk, flags, noblock, &peeked, &off, &err); |
1da177e4 | 305 | if (!skb) |
627d2d6b | 306 | return err; |
1da177e4 | 307 | |
cb891fa6 | 308 | ulen = udp6_skb_len(skb); |
59c2cdae | 309 | copied = len; |
627d2d6b | 310 | if (copied > ulen - off) |
311 | copied = ulen - off; | |
59c2cdae | 312 | else if (copied < ulen) |
1ab1457c | 313 | msg->msg_flags |= MSG_TRUNC; |
1da177e4 | 314 | |
f26ba175 | 315 | is_udp4 = (skb->protocol == htons(ETH_P_IP)); |
029a3743 | 316 | mib = __UDPX_MIB(sk, is_udp4); |
f26ba175 | 317 | |
ba4e58ec | 318 | /* |
759e5d00 HX |
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. | |
ba4e58ec | 322 | */ |
ba4e58ec | 323 | |
d21dbdfe ED |
324 | if (copied < ulen || peeking || |
325 | (is_udplite && UDP_SKB_CB(skb)->partial_cov)) { | |
67a51780 PA |
326 | checksum_valid = udp_skb_csum_unnecessary(skb) || |
327 | !__udp_lib_checksum_complete(skb); | |
197c949e | 328 | if (!checksum_valid) |
1da177e4 | 329 | goto csum_copy_err; |
ba4e58ec GR |
330 | } |
331 | ||
67a51780 PA |
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); | |
335 | else | |
336 | err = skb_copy_datagram_msg(skb, off, msg, copied); | |
337 | } else { | |
627d2d6b | 338 | err = skb_copy_and_csum_datagram_msg(skb, off, msg); |
1da177e4 LT |
339 | if (err == -EINVAL) |
340 | goto csum_copy_err; | |
341 | } | |
22911fc5 | 342 | if (unlikely(err)) { |
979402b1 ED |
343 | if (!peeked) { |
344 | atomic_inc(&sk->sk_drops); | |
029a3743 | 345 | SNMP_INC_STATS(mib, UDP_MIB_INERRORS); |
979402b1 | 346 | } |
850cbadd | 347 | kfree_skb(skb); |
627d2d6b | 348 | return err; |
22911fc5 | 349 | } |
029a3743 PA |
350 | if (!peeked) |
351 | SNMP_INC_STATS(mib, UDP_MIB_INDATAGRAMS); | |
cb75994e | 352 | |
3b885787 | 353 | sock_recv_ts_and_drops(msg, sk, skb); |
1da177e4 LT |
354 | |
355 | /* Copy the address. */ | |
356 | if (msg->msg_name) { | |
342dfc30 | 357 | DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
1da177e4 | 358 | sin6->sin6_family = AF_INET6; |
4bedb452 | 359 | sin6->sin6_port = udp_hdr(skb)->source; |
1da177e4 | 360 | sin6->sin6_flowinfo = 0; |
1da177e4 | 361 | |
842df073 | 362 | if (is_udp4) { |
b301e82c BH |
363 | ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, |
364 | &sin6->sin6_addr); | |
842df073 HFS |
365 | sin6->sin6_scope_id = 0; |
366 | } else { | |
4e3fd7a0 | 367 | sin6->sin6_addr = ipv6_hdr(skb)->saddr; |
842df073 HFS |
368 | sin6->sin6_scope_id = |
369 | ipv6_iface_scope_id(&sin6->sin6_addr, | |
4330487a | 370 | inet6_iif(skb)); |
1da177e4 | 371 | } |
bceaa902 | 372 | *addr_len = sizeof(*sin6); |
1da177e4 | 373 | } |
4b261c75 | 374 | |
bcd1665e PA |
375 | if (udp_sk(sk)->gro_enabled) |
376 | udp_cmsg_recv(msg, sk, skb); | |
377 | ||
4b261c75 HFS |
378 | if (np->rxopt.all) |
379 | ip6_datagram_recv_common_ctl(sk, msg, skb); | |
380 | ||
f26ba175 | 381 | if (is_udp4) { |
1da177e4 | 382 | if (inet->cmsg_flags) |
ad959036 | 383 | ip_cmsg_recv_offset(msg, sk, skb, |
10df8e61 | 384 | sizeof(struct udphdr), off); |
1da177e4 LT |
385 | } else { |
386 | if (np->rxopt.all) | |
4b261c75 | 387 | ip6_datagram_recv_specific_ctl(sk, msg, skb); |
1ab1457c | 388 | } |
1da177e4 | 389 | |
59c2cdae | 390 | err = copied; |
1da177e4 | 391 | if (flags & MSG_TRUNC) |
759e5d00 | 392 | err = ulen; |
1da177e4 | 393 | |
850cbadd | 394 | skb_consume_udp(sk, skb, peeking ? -err : err); |
1da177e4 LT |
395 | return err; |
396 | ||
397 | csum_copy_err: | |
2276f58a PA |
398 | if (!__sk_queue_drop_skb(sk, &udp_sk(sk)->reader_queue, skb, flags, |
399 | udp_skb_destructor)) { | |
029a3743 PA |
400 | SNMP_INC_STATS(mib, UDP_MIB_CSUMERRORS); |
401 | SNMP_INC_STATS(mib, UDP_MIB_INERRORS); | |
0856f939 | 402 | } |
850cbadd | 403 | kfree_skb(skb); |
1da177e4 | 404 | |
beb39db5 ED |
405 | /* starting over for a new packet, but check if we need to yield */ |
406 | cond_resched(); | |
9cfaa8de | 407 | msg->msg_flags &= ~MSG_TRUNC; |
1da177e4 LT |
408 | goto try_again; |
409 | } | |
410 | ||
a36e185e SB |
411 | DEFINE_STATIC_KEY_FALSE(udpv6_encap_needed_key); |
412 | void udpv6_encap_enable(void) | |
413 | { | |
9c480601 | 414 | static_branch_inc(&udpv6_encap_needed_key); |
a36e185e SB |
415 | } |
416 | EXPORT_SYMBOL(udpv6_encap_enable); | |
417 | ||
e7cc0824 SB |
418 | /* Handler for tunnels with arbitrary destination ports: no socket lookup, go |
419 | * through error handlers in encapsulations looking for a match. | |
420 | */ | |
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) | |
424 | { | |
425 | int i; | |
426 | ||
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); | |
430 | ||
431 | if (!ip6tun_encaps[i]) | |
432 | continue; | |
433 | handler = rcu_dereference(ip6tun_encaps[i]->err_handler); | |
434 | if (handler && !handler(skb, opt, type, code, offset, info)) | |
435 | return 0; | |
436 | } | |
437 | ||
438 | return -ENOENT; | |
439 | } | |
440 | ||
a36e185e SB |
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. | |
447 | * | |
e7cc0824 SB |
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. | |
451 | * | |
a36e185e SB |
452 | * Then ask the tunnel implementation to match the error against a valid |
453 | * association. | |
454 | * | |
e7cc0824 SB |
455 | * Return an error if we can't find a match, the socket if we need further |
456 | * processing, zero otherwise. | |
a36e185e SB |
457 | */ |
458 | static struct sock *__udp6_lib_err_encap(struct net *net, | |
459 | const struct ipv6hdr *hdr, int offset, | |
460 | struct udphdr *uh, | |
461 | struct udp_table *udptable, | |
e7cc0824 SB |
462 | struct sk_buff *skb, |
463 | struct inet6_skb_parm *opt, | |
464 | u8 type, u8 code, __be32 info) | |
a36e185e | 465 | { |
a36e185e | 466 | int network_offset, transport_offset; |
a36e185e SB |
467 | struct sock *sk; |
468 | ||
a36e185e SB |
469 | network_offset = skb_network_offset(skb); |
470 | transport_offset = skb_transport_offset(skb); | |
471 | ||
472 | /* Network header needs to point to the outer IPv6 header inside ICMP */ | |
473 | skb_reset_network_header(skb); | |
474 | ||
475 | /* Transport header needs to point to the UDP header */ | |
476 | skb_set_transport_header(skb, offset); | |
477 | ||
e7cc0824 SB |
478 | sk = __udp6_lib_lookup(net, &hdr->daddr, uh->source, |
479 | &hdr->saddr, uh->dest, | |
480 | inet6_iif(skb), 0, udptable, skb); | |
481 | if (sk) { | |
482 | int (*lookup)(struct sock *sk, struct sk_buff *skb); | |
483 | struct udp_sock *up = udp_sk(sk); | |
484 | ||
485 | lookup = READ_ONCE(up->encap_err_lookup); | |
486 | if (!lookup || lookup(sk, skb)) | |
487 | sk = NULL; | |
488 | } | |
489 | ||
490 | if (!sk) { | |
491 | sk = ERR_PTR(__udp6_lib_err_encap_no_sk(skb, opt, type, code, | |
492 | offset, info)); | |
493 | } | |
a36e185e SB |
494 | |
495 | skb_set_transport_header(skb, transport_offset); | |
496 | skb_set_network_header(skb, network_offset); | |
e7cc0824 | 497 | |
a36e185e SB |
498 | return sk; |
499 | } | |
500 | ||
32bbd879 SB |
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) | |
1da177e4 LT |
504 | { |
505 | struct ipv6_pinfo *np; | |
b71d1d42 ED |
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; | |
67ba4152 | 509 | struct udphdr *uh = (struct udphdr *)(skb->data+offset); |
a36e185e | 510 | bool tunnel = false; |
1da177e4 | 511 | struct sock *sk; |
e0d8c1b7 | 512 | int harderr; |
1da177e4 | 513 | int err; |
7304fe46 | 514 | struct net *net = dev_net(skb->dev); |
1da177e4 | 515 | |
e32ea7e7 | 516 | sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source, |
f64bf6b8 | 517 | inet6_iif(skb), inet6_sdif(skb), udptable, skb); |
63159f29 | 518 | if (!sk) { |
a36e185e | 519 | /* No socket for error: try tunnels before discarding */ |
e7cc0824 | 520 | sk = ERR_PTR(-ENOENT); |
a36e185e SB |
521 | if (static_branch_unlikely(&udpv6_encap_needed_key)) { |
522 | sk = __udp6_lib_err_encap(net, hdr, offset, uh, | |
e7cc0824 SB |
523 | udptable, skb, |
524 | opt, type, code, info); | |
525 | if (!sk) | |
526 | return 0; | |
a36e185e SB |
527 | } |
528 | ||
e7cc0824 | 529 | if (IS_ERR(sk)) { |
a36e185e SB |
530 | __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), |
531 | ICMP6_MIB_INERRORS); | |
e7cc0824 | 532 | return PTR_ERR(sk); |
a36e185e | 533 | } |
e7cc0824 | 534 | |
a36e185e | 535 | tunnel = true; |
7304fe46 | 536 | } |
1da177e4 | 537 | |
e0d8c1b7 WW |
538 | harderr = icmpv6_err_convert(type, code, &err); |
539 | np = inet6_sk(sk); | |
540 | ||
93b36cf3 HFS |
541 | if (type == ICMPV6_PKT_TOOBIG) { |
542 | if (!ip6_sk_accept_pmtu(sk)) | |
543 | goto out; | |
81aded24 | 544 | ip6_sk_update_pmtu(skb, sk, info); |
e0d8c1b7 WW |
545 | if (np->pmtudisc != IPV6_PMTUDISC_DONT) |
546 | harderr = 1; | |
93b36cf3 | 547 | } |
1a462d18 | 548 | if (type == NDISC_REDIRECT) { |
a36e185e SB |
549 | if (tunnel) { |
550 | ip6_redirect(skb, sock_net(sk), inet6_iif(skb), | |
551 | sk->sk_mark, sk->sk_uid); | |
552 | } else { | |
553 | ip6_sk_redirect(skb, sk); | |
554 | } | |
1a462d18 DJ |
555 | goto out; |
556 | } | |
81aded24 | 557 | |
a36e185e SB |
558 | /* Tunnels don't have an application socket: don't pass errors back */ |
559 | if (tunnel) | |
560 | goto out; | |
561 | ||
e0d8c1b7 WW |
562 | if (!np->recverr) { |
563 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
564 | goto out; | |
565 | } else { | |
1da177e4 | 566 | ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1)); |
e0d8c1b7 | 567 | } |
b1faf566 | 568 | |
1da177e4 LT |
569 | sk->sk_err = err; |
570 | sk->sk_error_report(sk); | |
571 | out: | |
32bbd879 | 572 | return 0; |
1da177e4 LT |
573 | } |
574 | ||
a3f96c47 | 575 | static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
f7ad74fe BL |
576 | { |
577 | int rc; | |
578 | ||
efe4208f | 579 | if (!ipv6_addr_any(&sk->sk_v6_daddr)) { |
f7ad74fe | 580 | sock_rps_save_rxhash(sk, skb); |
005ec974 | 581 | sk_mark_napi_id(sk, skb); |
2c8c56e1 | 582 | sk_incoming_cpu_update(sk); |
e68b6e50 ED |
583 | } else { |
584 | sk_mark_napi_id_once(sk, skb); | |
005ec974 | 585 | } |
f7ad74fe | 586 | |
850cbadd | 587 | rc = __udp_enqueue_schedule_skb(sk, skb); |
f7ad74fe BL |
588 | if (rc < 0) { |
589 | int is_udplite = IS_UDPLITE(sk); | |
590 | ||
591 | /* Note that an ENOMEM error is charged twice */ | |
592 | if (rc == -ENOMEM) | |
e61da9e2 | 593 | UDP6_INC_STATS(sock_net(sk), |
02c22347 | 594 | UDP_MIB_RCVBUFERRORS, is_udplite); |
e61da9e2 | 595 | UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
f7ad74fe BL |
596 | kfree_skb(skb); |
597 | return -1; | |
598 | } | |
850cbadd | 599 | |
f7ad74fe BL |
600 | return 0; |
601 | } | |
602 | ||
32bbd879 SB |
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) | |
ba4e58ec | 606 | { |
32bbd879 | 607 | return __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table); |
ba4e58ec GR |
608 | } |
609 | ||
cf329aa4 | 610 | static int udpv6_queue_rcv_one_skb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 611 | { |
ba4e58ec | 612 | struct udp_sock *up = udp_sk(sk); |
b2bf1e26 | 613 | int is_udplite = IS_UDPLITE(sk); |
a18135eb | 614 | |
ba4e58ec GR |
615 | if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) |
616 | goto drop; | |
1da177e4 | 617 | |
88ab3108 | 618 | if (static_branch_unlikely(&udpv6_encap_needed_key) && up->encap_type) { |
d7f3f621 BL |
619 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
620 | ||
621 | /* | |
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 | |
630 | */ | |
631 | ||
632 | /* if we're overly short, let UDP handle it */ | |
6aa7de05 | 633 | encap_rcv = READ_ONCE(up->encap_rcv); |
e5aed006 | 634 | if (encap_rcv) { |
d7f3f621 BL |
635 | int ret; |
636 | ||
0a80966b TH |
637 | /* Verify checksum before giving to encap */ |
638 | if (udp_lib_checksum_complete(skb)) | |
639 | goto csum_error; | |
640 | ||
d7f3f621 BL |
641 | ret = encap_rcv(sk, skb); |
642 | if (ret <= 0) { | |
02c22347 ED |
643 | __UDP_INC_STATS(sock_net(sk), |
644 | UDP_MIB_INDATAGRAMS, | |
645 | is_udplite); | |
d7f3f621 BL |
646 | return -ret; |
647 | } | |
648 | } | |
649 | ||
650 | /* FALLTHROUGH -- it's a UDP Packet */ | |
651 | } | |
652 | ||
ba4e58ec GR |
653 | /* |
654 | * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c). | |
655 | */ | |
b2bf1e26 | 656 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { |
ba4e58ec GR |
657 | |
658 | if (up->pcrlen == 0) { /* full coverage was set */ | |
ba7a46f1 JP |
659 | net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n", |
660 | UDP_SKB_CB(skb)->cscov, skb->len); | |
ba4e58ec GR |
661 | goto drop; |
662 | } | |
663 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
ba7a46f1 JP |
664 | net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n", |
665 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
ba4e58ec GR |
666 | goto drop; |
667 | } | |
1da177e4 LT |
668 | } |
669 | ||
4b943fae | 670 | prefetch(&sk->sk_rmem_alloc); |
ce25d66a ED |
671 | if (rcu_access_pointer(sk->sk_filter) && |
672 | udp_lib_checksum_complete(skb)) | |
673 | goto csum_error; | |
674 | ||
ba66bbe5 | 675 | if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr))) |
a6127697 | 676 | goto drop; |
ba4e58ec | 677 | |
e6afc8ac | 678 | udp_csum_pull_header(skb); |
cb80ef46 | 679 | |
d826eb14 | 680 | skb_dst_drop(skb); |
cb75994e | 681 | |
850cbadd | 682 | return __udpv6_queue_rcv_skb(sk, skb); |
3e215c8d | 683 | |
6a5dc9e5 | 684 | csum_error: |
02c22347 | 685 | __UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); |
ba4e58ec | 686 | drop: |
02c22347 | 687 | __UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
cb80ef46 | 688 | atomic_inc(&sk->sk_drops); |
ba4e58ec GR |
689 | kfree_skb(skb); |
690 | return -1; | |
1da177e4 LT |
691 | } |
692 | ||
cf329aa4 PA |
693 | static int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
694 | { | |
695 | struct sk_buff *next, *segs; | |
696 | int ret; | |
697 | ||
698 | if (likely(!udp_unexpected_gso(sk, skb))) | |
699 | return udpv6_queue_rcv_one_skb(sk, skb); | |
700 | ||
701 | __skb_push(skb, -skb_mac_offset(skb)); | |
702 | segs = udp_rcv_segment(sk, skb, false); | |
703 | for (skb = segs; skb; skb = next) { | |
704 | next = skb->next; | |
705 | __skb_pull(skb, skb_transport_offset(skb)); | |
706 | ||
707 | ret = udpv6_queue_rcv_one_skb(sk, skb); | |
708 | if (ret > 0) | |
709 | ip6_protocol_deliver_rcu(dev_net(skb->dev), skb, ret, | |
710 | true); | |
711 | } | |
712 | return 0; | |
713 | } | |
714 | ||
5cf3d461 DH |
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, | |
7bd2db40 | 718 | int dif, int sdif, unsigned short hnum) |
1da177e4 | 719 | { |
5cf3d461 | 720 | struct inet_sock *inet = inet_sk(sk); |
1da177e4 | 721 | |
5cf3d461 DH |
722 | if (!net_eq(sock_net(sk), net)) |
723 | return false; | |
724 | ||
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)) || | |
7bd2db40 | 730 | !udp_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif) || |
33b4b015 HR |
731 | (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) && |
732 | !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))) | |
5cf3d461 DH |
733 | return false; |
734 | if (!inet6_mc_check(sk, loc_addr, rmt_addr)) | |
735 | return false; | |
736 | return true; | |
1da177e4 LT |
737 | } |
738 | ||
4068579e TH |
739 | static void udp6_csum_zero_error(struct sk_buff *skb) |
740 | { | |
741 | /* RFC 2460 section 8.1 says that we SHOULD log | |
742 | * this error. Well, it is reasonable. | |
743 | */ | |
ba7a46f1 JP |
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)); | |
4068579e TH |
747 | } |
748 | ||
1da177e4 LT |
749 | /* |
750 | * Note: called only from the BH handler context, | |
751 | * so we don't need to lock the hashes. | |
752 | */ | |
e3163493 | 753 | static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
b71d1d42 | 754 | const struct in6_addr *saddr, const struct in6_addr *daddr, |
36cbb245 | 755 | struct udp_table *udptable, int proto) |
1da177e4 | 756 | { |
ca065d0c | 757 | struct sock *sk, *first = NULL; |
4bedb452 | 758 | const struct udphdr *uh = udp_hdr(skb); |
5cf3d461 DH |
759 | unsigned short hnum = ntohs(uh->dest); |
760 | struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum); | |
ca065d0c | 761 | unsigned int offset = offsetof(typeof(*sk), sk_node); |
2dc41cff | 762 | unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10); |
ca065d0c | 763 | int dif = inet6_iif(skb); |
7bd2db40 | 764 | int sdif = inet6_sdif(skb); |
ca065d0c ED |
765 | struct hlist_node *node; |
766 | struct sk_buff *nskb; | |
2dc41cff DH |
767 | |
768 | if (use_hash2) { | |
f0b1e64c | 769 | hash2_any = ipv6_portaddr_hash(net, &in6addr_any, hnum) & |
73e2d5e3 | 770 | udptable->mask; |
f0b1e64c | 771 | hash2 = ipv6_portaddr_hash(net, daddr, hnum) & udptable->mask; |
2dc41cff | 772 | start_lookup: |
73e2d5e3 | 773 | hslot = &udptable->hash2[hash2]; |
2dc41cff DH |
774 | offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node); |
775 | } | |
1da177e4 | 776 | |
ca065d0c ED |
777 | sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) { |
778 | if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr, | |
7bd2db40 DM |
779 | uh->source, saddr, dif, sdif, |
780 | hnum)) | |
ca065d0c ED |
781 | continue; |
782 | /* If zero checksum and no_check is not on for | |
783 | * the socket then skip it. | |
784 | */ | |
785 | if (!uh->check && !udp_sk(sk)->no_check6_rx) | |
786 | continue; | |
787 | if (!first) { | |
788 | first = sk; | |
789 | continue; | |
790 | } | |
791 | nskb = skb_clone(skb, GFP_ATOMIC); | |
792 | if (unlikely(!nskb)) { | |
793 | atomic_inc(&sk->sk_drops); | |
02c22347 ED |
794 | __UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS, |
795 | IS_UDPLITE(sk)); | |
796 | __UDP6_INC_STATS(net, UDP_MIB_INERRORS, | |
797 | IS_UDPLITE(sk)); | |
ca065d0c | 798 | continue; |
95766fff | 799 | } |
a1ab77f9 | 800 | |
ca065d0c ED |
801 | if (udpv6_queue_rcv_skb(sk, nskb) > 0) |
802 | consume_skb(nskb); | |
803 | } | |
a1ab77f9 | 804 | |
2dc41cff DH |
805 | /* Also lookup *:port if we are using hash2 and haven't done so yet. */ |
806 | if (use_hash2 && hash2 != hash2_any) { | |
807 | hash2 = hash2_any; | |
808 | goto start_lookup; | |
809 | } | |
810 | ||
ca065d0c ED |
811 | if (first) { |
812 | if (udpv6_queue_rcv_skb(first, skb) > 0) | |
813 | consume_skb(skb); | |
a1ab77f9 | 814 | } else { |
ca065d0c | 815 | kfree_skb(skb); |
02c22347 ED |
816 | __UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI, |
817 | proto == IPPROTO_UDPLITE); | |
a1ab77f9 | 818 | } |
ba4e58ec | 819 | return 0; |
1da177e4 LT |
820 | } |
821 | ||
64f0f5d1 PA |
822 | static void udp6_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst) |
823 | { | |
824 | if (udp_sk_rx_dst_set(sk, dst)) { | |
825 | const struct rt6_info *rt = (const struct rt6_info *)dst; | |
826 | ||
827 | inet6_sk(sk)->rx_dst_cookie = rt6_get_cookie(rt); | |
828 | } | |
829 | } | |
830 | ||
eb63f296 PA |
831 | /* wrapper for udp_queue_rcv_skb tacking care of csum conversion and |
832 | * return code conversion for ip layer consumption | |
833 | */ | |
834 | static int udp6_unicast_rcv_skb(struct sock *sk, struct sk_buff *skb, | |
835 | struct udphdr *uh) | |
836 | { | |
837 | int ret; | |
838 | ||
839 | if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk)) | |
840 | skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check, | |
841 | ip6_compute_pseudo); | |
842 | ||
843 | ret = udpv6_queue_rcv_skb(sk, skb); | |
844 | ||
84dad559 | 845 | /* a return value > 0 means to resubmit the input */ |
eb63f296 | 846 | if (ret > 0) |
84dad559 | 847 | return ret; |
eb63f296 PA |
848 | return 0; |
849 | } | |
850 | ||
645ca708 | 851 | int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
759e5d00 | 852 | int proto) |
1da177e4 | 853 | { |
ca065d0c | 854 | const struct in6_addr *saddr, *daddr; |
3ffe533c | 855 | struct net *net = dev_net(skb->dev); |
1ab1457c | 856 | struct udphdr *uh; |
ca065d0c | 857 | struct sock *sk; |
1da177e4 LT |
858 | u32 ulen = 0; |
859 | ||
860 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
d6bc0149 | 861 | goto discard; |
1da177e4 | 862 | |
0660e03f ACM |
863 | saddr = &ipv6_hdr(skb)->saddr; |
864 | daddr = &ipv6_hdr(skb)->daddr; | |
4bedb452 | 865 | uh = udp_hdr(skb); |
1da177e4 LT |
866 | |
867 | ulen = ntohs(uh->len); | |
ba4e58ec GR |
868 | if (ulen > skb->len) |
869 | goto short_packet; | |
1da177e4 | 870 | |
759e5d00 HX |
871 | if (proto == IPPROTO_UDP) { |
872 | /* UDP validates ulen. */ | |
1da177e4 | 873 | |
ba4e58ec GR |
874 | /* Check for jumbo payload */ |
875 | if (ulen == 0) | |
876 | ulen = skb->len; | |
1da177e4 | 877 | |
ba4e58ec GR |
878 | if (ulen < sizeof(*uh)) |
879 | goto short_packet; | |
1da177e4 | 880 | |
ba4e58ec GR |
881 | if (ulen < skb->len) { |
882 | if (pskb_trim_rcsum(skb, ulen)) | |
883 | goto short_packet; | |
0660e03f ACM |
884 | saddr = &ipv6_hdr(skb)->saddr; |
885 | daddr = &ipv6_hdr(skb)->daddr; | |
4bedb452 | 886 | uh = udp_hdr(skb); |
ba4e58ec | 887 | } |
ba4e58ec | 888 | } |
1da177e4 | 889 | |
759e5d00 | 890 | if (udp6_csum_init(skb, uh, proto)) |
6a5dc9e5 | 891 | goto csum_error; |
759e5d00 | 892 | |
c9f2c1ae PA |
893 | /* Check if the socket is already available, e.g. due to early demux */ |
894 | sk = skb_steal_sock(skb); | |
895 | if (sk) { | |
896 | struct dst_entry *dst = skb_dst(skb); | |
897 | int ret; | |
898 | ||
899 | if (unlikely(sk->sk_rx_dst != dst)) | |
64f0f5d1 | 900 | udp6_sk_rx_dst_set(sk, dst); |
c9f2c1ae | 901 | |
eb63f296 PA |
902 | if (!uh->check && !udp_sk(sk)->no_check6_rx) { |
903 | sock_put(sk); | |
904 | goto report_csum_error; | |
905 | } | |
c9f2c1ae | 906 | |
eb63f296 PA |
907 | ret = udp6_unicast_rcv_skb(sk, skb, uh); |
908 | sock_put(sk); | |
909 | return ret; | |
c9f2c1ae PA |
910 | } |
911 | ||
1ab1457c YH |
912 | /* |
913 | * Multicast receive code | |
1da177e4 | 914 | */ |
ba4e58ec | 915 | if (ipv6_addr_is_multicast(daddr)) |
e3163493 | 916 | return __udp6_lib_mcast_deliver(net, skb, |
36cbb245 | 917 | saddr, daddr, udptable, proto); |
1da177e4 LT |
918 | |
919 | /* Unicast */ | |
607c4aaf | 920 | sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable); |
53b24b8f | 921 | if (sk) { |
eb63f296 PA |
922 | if (!uh->check && !udp_sk(sk)->no_check6_rx) |
923 | goto report_csum_error; | |
924 | return udp6_unicast_rcv_skb(sk, skb, uh); | |
1da177e4 | 925 | } |
1ab1457c | 926 | |
eb63f296 PA |
927 | if (!uh->check) |
928 | goto report_csum_error; | |
4068579e | 929 | |
cb80ef46 | 930 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) |
c377411f | 931 | goto discard; |
cb80ef46 BL |
932 | |
933 | if (udp_lib_checksum_complete(skb)) | |
6a5dc9e5 | 934 | goto csum_error; |
cb80ef46 | 935 | |
02c22347 | 936 | __UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
cb80ef46 BL |
937 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
938 | ||
939 | kfree_skb(skb); | |
add459aa | 940 | return 0; |
1da177e4 | 941 | |
1ab1457c | 942 | short_packet: |
ba7a46f1 JP |
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), | |
946 | ulen, skb->len, | |
947 | daddr, ntohs(uh->dest)); | |
6a5dc9e5 | 948 | goto discard; |
eb63f296 PA |
949 | |
950 | report_csum_error: | |
951 | udp6_csum_zero_error(skb); | |
6a5dc9e5 | 952 | csum_error: |
02c22347 | 953 | __UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); |
1da177e4 | 954 | discard: |
02c22347 | 955 | __UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
1da177e4 | 956 | kfree_skb(skb); |
add459aa | 957 | return 0; |
1da177e4 | 958 | } |
ba4e58ec | 959 | |
0bd84065 | 960 | |
5425077d | 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, | |
4297a0ef | 964 | int dif, int sdif) |
5425077d | 965 | { |
0bd84065 | 966 | unsigned short hnum = ntohs(loc_port); |
f0b1e64c | 967 | unsigned int hash2 = ipv6_portaddr_hash(net, loc_addr, hnum); |
0bd84065 | 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); | |
5425077d | 971 | struct sock *sk; |
972 | ||
0bd84065 | 973 | udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
85cb73ff | 974 | if (sk->sk_state == TCP_ESTABLISHED && |
4297a0ef | 975 | INET6_MATCH(sk, net, rmt_addr, loc_addr, ports, dif, sdif)) |
0bd84065 | 976 | return sk; |
977 | /* Only check first socket in chain */ | |
978 | break; | |
979 | } | |
980 | return NULL; | |
5425077d | 981 | } |
982 | ||
983 | static void udp_v6_early_demux(struct sk_buff *skb) | |
984 | { | |
985 | struct net *net = dev_net(skb->dev); | |
986 | const struct udphdr *uh; | |
987 | struct sock *sk; | |
988 | struct dst_entry *dst; | |
989 | int dif = skb->dev->ifindex; | |
4297a0ef | 990 | int sdif = inet6_sdif(skb); |
5425077d | 991 | |
992 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + | |
993 | sizeof(struct udphdr))) | |
994 | return; | |
995 | ||
996 | uh = udp_hdr(skb); | |
997 | ||
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, | |
4297a0ef | 1002 | dif, sdif); |
5425077d | 1003 | else |
1004 | return; | |
1005 | ||
41c6d650 | 1006 | if (!sk || !refcount_inc_not_zero(&sk->sk_refcnt)) |
5425077d | 1007 | return; |
1008 | ||
1009 | skb->sk = sk; | |
1010 | skb->destructor = sock_efree; | |
1011 | dst = READ_ONCE(sk->sk_rx_dst); | |
1012 | ||
1013 | if (dst) | |
1014 | dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie); | |
1015 | if (dst) { | |
d24406c8 WW |
1016 | /* set noref for now. |
1017 | * any place which wants to hold dst has to call | |
1018 | * dst_hold_safe() | |
1019 | */ | |
1020 | skb_dst_set_noref(skb, dst); | |
5425077d | 1021 | } |
1022 | } | |
1023 | ||
e5bbef20 | 1024 | static __inline__ int udpv6_rcv(struct sk_buff *skb) |
ba4e58ec | 1025 | { |
645ca708 | 1026 | return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
ba4e58ec GR |
1027 | } |
1028 | ||
1da177e4 LT |
1029 | /* |
1030 | * Throw away all pending data and cancel the corking. Socket is locked. | |
1031 | */ | |
1032 | static void udp_v6_flush_pending_frames(struct sock *sk) | |
1033 | { | |
1034 | struct udp_sock *up = udp_sk(sk); | |
1035 | ||
36d926b9 DL |
1036 | if (up->pending == AF_INET) |
1037 | udp_flush_pending_frames(sk); | |
1038 | else if (up->pending) { | |
1da177e4 LT |
1039 | up->len = 0; |
1040 | up->pending = 0; | |
1041 | ip6_flush_pending_frames(sk); | |
1ab1457c | 1042 | } |
1da177e4 LT |
1043 | } |
1044 | ||
d74bad4e AI |
1045 | static int udpv6_pre_connect(struct sock *sk, struct sockaddr *uaddr, |
1046 | int addr_len) | |
1047 | { | |
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. | |
1051 | */ | |
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); | |
1056 | } | |
1057 | ||
1058 | if (addr_len < SIN6_LEN_RFC2133) | |
1059 | return -EINVAL; | |
1060 | ||
1061 | return BPF_CGROUP_RUN_PROG_INET6_CONNECT_LOCK(sk, uaddr); | |
1062 | } | |
1063 | ||
493c6be3 | 1064 | /** |
67ba4152 IM |
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) | |
493c6be3 SS |
1069 | */ |
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) | |
1073 | { | |
1074 | unsigned int offset; | |
1075 | struct udphdr *uh = udp_hdr(skb); | |
d39d938c | 1076 | struct sk_buff *frags = skb_shinfo(skb)->frag_list; |
493c6be3 SS |
1077 | __wsum csum = 0; |
1078 | ||
d39d938c | 1079 | if (!frags) { |
493c6be3 SS |
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); | |
1084 | } else { | |
1085 | /* | |
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 | |
1089 | */ | |
1090 | offset = skb_transport_offset(skb); | |
1091 | skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); | |
63ecc3d9 | 1092 | csum = skb->csum; |
493c6be3 SS |
1093 | |
1094 | skb->ip_summed = CHECKSUM_NONE; | |
1095 | ||
d39d938c VY |
1096 | do { |
1097 | csum = csum_add(csum, frags->csum); | |
1098 | } while ((frags = frags->next)); | |
493c6be3 SS |
1099 | |
1100 | uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, | |
1101 | csum); | |
1102 | if (uh->check == 0) | |
1103 | uh->check = CSUM_MANGLED_0; | |
1104 | } | |
1105 | } | |
1106 | ||
1da177e4 LT |
1107 | /* |
1108 | * Sending | |
1109 | */ | |
1110 | ||
bec1f6f6 WB |
1111 | static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6, |
1112 | struct inet_cork *cork) | |
1da177e4 | 1113 | { |
d39d938c | 1114 | struct sock *sk = skb->sk; |
1da177e4 | 1115 | struct udphdr *uh; |
1da177e4 | 1116 | int err = 0; |
b2bf1e26 | 1117 | int is_udplite = IS_UDPLITE(sk); |
868c86bc | 1118 | __wsum csum = 0; |
d39d938c VY |
1119 | int offset = skb_transport_offset(skb); |
1120 | int len = skb->len - offset; | |
1da177e4 LT |
1121 | |
1122 | /* | |
1123 | * Create a UDP header | |
1124 | */ | |
4bedb452 | 1125 | uh = udp_hdr(skb); |
1958b856 DM |
1126 | uh->source = fl6->fl6_sport; |
1127 | uh->dest = fl6->fl6_dport; | |
d39d938c | 1128 | uh->len = htons(len); |
1da177e4 LT |
1129 | uh->check = 0; |
1130 | ||
bec1f6f6 WB |
1131 | if (cork->gso_size) { |
1132 | const int hlen = skb_network_header_len(skb) + | |
1133 | sizeof(struct udphdr); | |
1134 | ||
0f149c9f WB |
1135 | if (hlen + cork->gso_size > cork->fragsize) { |
1136 | kfree_skb(skb); | |
bec1f6f6 | 1137 | return -EINVAL; |
0f149c9f WB |
1138 | } |
1139 | if (skb->len > cork->gso_size * UDP_MAX_SEGMENTS) { | |
1140 | kfree_skb(skb); | |
bec1f6f6 | 1141 | return -EINVAL; |
0f149c9f WB |
1142 | } |
1143 | if (udp_sk(sk)->no_check6_tx) { | |
1144 | kfree_skb(skb); | |
a8c744a8 | 1145 | return -EINVAL; |
0f149c9f | 1146 | } |
ff06342c | 1147 | if (skb->ip_summed != CHECKSUM_PARTIAL || is_udplite || |
0f149c9f WB |
1148 | dst_xfrm(skb_dst(skb))) { |
1149 | kfree_skb(skb); | |
bec1f6f6 | 1150 | return -EIO; |
0f149c9f | 1151 | } |
bec1f6f6 WB |
1152 | |
1153 | skb_shinfo(skb)->gso_size = cork->gso_size; | |
1154 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP_L4; | |
a8c744a8 | 1155 | goto csum_partial; |
bec1f6f6 WB |
1156 | } |
1157 | ||
b2bf1e26 | 1158 | if (is_udplite) |
d39d938c VY |
1159 | csum = udplite_csum(skb); |
1160 | else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */ | |
4068579e TH |
1161 | skb->ip_summed = CHECKSUM_NONE; |
1162 | goto send; | |
1163 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
a8c744a8 | 1164 | csum_partial: |
d39d938c | 1165 | udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len); |
493c6be3 SS |
1166 | goto send; |
1167 | } else | |
d39d938c | 1168 | csum = udp_csum(skb); |
1da177e4 | 1169 | |
ba4e58ec | 1170 | /* add protocol-dependent pseudo-header */ |
4c9483b2 | 1171 | uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, |
d39d938c | 1172 | len, fl6->flowi6_proto, csum); |
1da177e4 | 1173 | if (uh->check == 0) |
f6ab0288 | 1174 | uh->check = CSUM_MANGLED_0; |
1da177e4 | 1175 | |
493c6be3 | 1176 | send: |
d39d938c | 1177 | err = ip6_send_skb(skb); |
6ce9e7b5 ED |
1178 | if (err) { |
1179 | if (err == -ENOBUFS && !inet6_sk(sk)->recverr) { | |
6aef70a8 ED |
1180 | UDP6_INC_STATS(sock_net(sk), |
1181 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
6ce9e7b5 ED |
1182 | err = 0; |
1183 | } | |
6aef70a8 ED |
1184 | } else { |
1185 | UDP6_INC_STATS(sock_net(sk), | |
1186 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
1187 | } | |
d39d938c VY |
1188 | return err; |
1189 | } | |
1190 | ||
1191 | static int udp_v6_push_pending_frames(struct sock *sk) | |
1192 | { | |
1193 | struct sk_buff *skb; | |
1194 | struct udp_sock *up = udp_sk(sk); | |
1195 | struct flowi6 fl6; | |
1196 | int err = 0; | |
1197 | ||
1198 | if (up->pending == AF_INET) | |
1199 | return udp_push_pending_frames(sk); | |
1200 | ||
1201 | /* ip6_finish_skb will release the cork, so make a copy of | |
1202 | * fl6 here. | |
1203 | */ | |
1204 | fl6 = inet_sk(sk)->cork.fl.u.ip6; | |
1205 | ||
1206 | skb = ip6_finish_skb(sk); | |
1207 | if (!skb) | |
1208 | goto out; | |
1209 | ||
bec1f6f6 | 1210 | err = udp_v6_send_skb(skb, &fl6, &inet_sk(sk)->cork.base); |
d39d938c | 1211 | |
1da177e4 LT |
1212 | out: |
1213 | up->len = 0; | |
1214 | up->pending = 0; | |
1215 | return err; | |
1216 | } | |
1217 | ||
1b784140 | 1218 | int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
1da177e4 LT |
1219 | { |
1220 | struct ipv6_txoptions opt_space; | |
1221 | struct udp_sock *up = udp_sk(sk); | |
1222 | struct inet_sock *inet = inet_sk(sk); | |
1223 | struct ipv6_pinfo *np = inet6_sk(sk); | |
342dfc30 | 1224 | DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
20c59de2 | 1225 | struct in6_addr *daddr, *final_p, final; |
1da177e4 | 1226 | struct ipv6_txoptions *opt = NULL; |
45f6fad8 | 1227 | struct ipv6_txoptions *opt_to_free = NULL; |
1da177e4 | 1228 | struct ip6_flowlabel *flowlabel = NULL; |
4c9483b2 | 1229 | struct flowi6 fl6; |
1da177e4 | 1230 | struct dst_entry *dst; |
26879da5 | 1231 | struct ipcm6_cookie ipc6; |
1da177e4 | 1232 | int addr_len = msg->msg_namelen; |
9f542f61 | 1233 | bool connected = false; |
1da177e4 | 1234 | int ulen = len; |
1da177e4 LT |
1235 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; |
1236 | int err; | |
b2bf1e26 | 1237 | int is_udplite = IS_UDPLITE(sk); |
ba4e58ec | 1238 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); |
1da177e4 | 1239 | |
b515430a | 1240 | ipcm6_init(&ipc6); |
bec1f6f6 | 1241 | ipc6.gso_size = up->gso_size; |
5fdaa88d | 1242 | ipc6.sockc.tsflags = sk->sk_tsflags; |
26879da5 | 1243 | |
1da177e4 LT |
1244 | /* destination address check */ |
1245 | if (sin6) { | |
1246 | if (addr_len < offsetof(struct sockaddr, sa_data)) | |
1247 | return -EINVAL; | |
1248 | ||
1249 | switch (sin6->sin6_family) { | |
1250 | case AF_INET6: | |
1251 | if (addr_len < SIN6_LEN_RFC2133) | |
1252 | return -EINVAL; | |
1253 | daddr = &sin6->sin6_addr; | |
052d2369 JL |
1254 | if (ipv6_addr_any(daddr) && |
1255 | ipv6_addr_v4mapped(&np->saddr)) | |
1256 | ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK), | |
1257 | daddr); | |
1da177e4 LT |
1258 | break; |
1259 | case AF_INET: | |
1260 | goto do_udp_sendmsg; | |
1261 | case AF_UNSPEC: | |
1262 | msg->msg_name = sin6 = NULL; | |
1263 | msg->msg_namelen = addr_len = 0; | |
1264 | daddr = NULL; | |
1265 | break; | |
1266 | default: | |
1267 | return -EINVAL; | |
1268 | } | |
1269 | } else if (!up->pending) { | |
1270 | if (sk->sk_state != TCP_ESTABLISHED) | |
1271 | return -EDESTADDRREQ; | |
efe4208f | 1272 | daddr = &sk->sk_v6_daddr; |
1ab1457c | 1273 | } else |
1da177e4 LT |
1274 | daddr = NULL; |
1275 | ||
1276 | if (daddr) { | |
e773e4fa | 1277 | if (ipv6_addr_v4mapped(daddr)) { |
1da177e4 LT |
1278 | struct sockaddr_in sin; |
1279 | sin.sin_family = AF_INET; | |
c720c7e8 | 1280 | sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport; |
1da177e4 LT |
1281 | sin.sin_addr.s_addr = daddr->s6_addr32[3]; |
1282 | msg->msg_name = &sin; | |
1283 | msg->msg_namelen = sizeof(sin); | |
1284 | do_udp_sendmsg: | |
1285 | if (__ipv6_only_sock(sk)) | |
1286 | return -ENETUNREACH; | |
1b784140 | 1287 | return udp_sendmsg(sk, msg, len); |
1da177e4 LT |
1288 | } |
1289 | } | |
1290 | ||
1291 | if (up->pending == AF_INET) | |
1b784140 | 1292 | return udp_sendmsg(sk, msg, len); |
1da177e4 LT |
1293 | |
1294 | /* Rough check on arithmetic overflow, | |
b59e139b | 1295 | better check is made in ip6_append_data(). |
1da177e4 LT |
1296 | */ |
1297 | if (len > INT_MAX - sizeof(struct udphdr)) | |
1298 | return -EMSGSIZE; | |
1ab1457c | 1299 | |
03485f2a | 1300 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
1da177e4 LT |
1301 | if (up->pending) { |
1302 | /* | |
1303 | * There are pending frames. | |
1304 | * The socket lock must be held while it's corked. | |
1305 | */ | |
1306 | lock_sock(sk); | |
1307 | if (likely(up->pending)) { | |
1308 | if (unlikely(up->pending != AF_INET6)) { | |
1309 | release_sock(sk); | |
1310 | return -EAFNOSUPPORT; | |
1311 | } | |
1312 | dst = NULL; | |
1313 | goto do_append_data; | |
1314 | } | |
1315 | release_sock(sk); | |
1316 | } | |
1317 | ulen += sizeof(struct udphdr); | |
1318 | ||
4c9483b2 | 1319 | memset(&fl6, 0, sizeof(fl6)); |
1da177e4 LT |
1320 | |
1321 | if (sin6) { | |
1322 | if (sin6->sin6_port == 0) | |
1323 | return -EINVAL; | |
1324 | ||
1958b856 | 1325 | fl6.fl6_dport = sin6->sin6_port; |
1da177e4 LT |
1326 | daddr = &sin6->sin6_addr; |
1327 | ||
1328 | if (np->sndflow) { | |
4c9483b2 DM |
1329 | fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; |
1330 | if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { | |
1331 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); | |
63159f29 | 1332 | if (!flowlabel) |
1da177e4 | 1333 | return -EINVAL; |
1da177e4 LT |
1334 | } |
1335 | } | |
1336 | ||
1337 | /* | |
1338 | * Otherwise it will be difficult to maintain | |
1339 | * sk->sk_dst_cache. | |
1340 | */ | |
1341 | if (sk->sk_state == TCP_ESTABLISHED && | |
efe4208f ED |
1342 | ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) |
1343 | daddr = &sk->sk_v6_daddr; | |
1da177e4 LT |
1344 | |
1345 | if (addr_len >= sizeof(struct sockaddr_in6) && | |
1346 | sin6->sin6_scope_id && | |
842df073 | 1347 | __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) |
4c9483b2 | 1348 | fl6.flowi6_oif = sin6->sin6_scope_id; |
1da177e4 LT |
1349 | } else { |
1350 | if (sk->sk_state != TCP_ESTABLISHED) | |
1351 | return -EDESTADDRREQ; | |
1352 | ||
1958b856 | 1353 | fl6.fl6_dport = inet->inet_dport; |
efe4208f | 1354 | daddr = &sk->sk_v6_daddr; |
4c9483b2 | 1355 | fl6.flowlabel = np->flow_label; |
9f542f61 | 1356 | connected = true; |
1da177e4 LT |
1357 | } |
1358 | ||
4c9483b2 DM |
1359 | if (!fl6.flowi6_oif) |
1360 | fl6.flowi6_oif = sk->sk_bound_dev_if; | |
1da177e4 | 1361 | |
4c9483b2 DM |
1362 | if (!fl6.flowi6_oif) |
1363 | fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex; | |
9f690db7 | 1364 | |
4c9483b2 | 1365 | fl6.flowi6_mark = sk->sk_mark; |
e2d118a1 | 1366 | fl6.flowi6_uid = sk->sk_uid; |
51953d5b | 1367 | |
1da177e4 LT |
1368 | if (msg->msg_controllen) { |
1369 | opt = &opt_space; | |
1370 | memset(opt, 0, sizeof(struct ipv6_txoptions)); | |
1371 | opt->tot_len = sizeof(*opt); | |
26879da5 | 1372 | ipc6.opt = opt; |
1da177e4 | 1373 | |
2e8de857 WB |
1374 | err = udp_cmsg_send(sk, msg, &ipc6.gso_size); |
1375 | if (err > 0) | |
1376 | err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, | |
5fdaa88d | 1377 | &ipc6); |
1da177e4 LT |
1378 | if (err < 0) { |
1379 | fl6_sock_release(flowlabel); | |
1380 | return err; | |
1381 | } | |
4c9483b2 DM |
1382 | if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { |
1383 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); | |
63159f29 | 1384 | if (!flowlabel) |
1da177e4 LT |
1385 | return -EINVAL; |
1386 | } | |
1387 | if (!(opt->opt_nflen|opt->opt_flen)) | |
1388 | opt = NULL; | |
9f542f61 | 1389 | connected = false; |
1da177e4 | 1390 | } |
45f6fad8 ED |
1391 | if (!opt) { |
1392 | opt = txopt_get(np); | |
1393 | opt_to_free = opt; | |
1394 | } | |
df9890c3 YH |
1395 | if (flowlabel) |
1396 | opt = fl6_merge_options(&opt_space, flowlabel, opt); | |
1397 | opt = ipv6_fixup_options(&opt_space, opt); | |
26879da5 | 1398 | ipc6.opt = opt; |
1da177e4 | 1399 | |
4c9483b2 | 1400 | fl6.flowi6_proto = sk->sk_protocol; |
e8e36984 | 1401 | fl6.daddr = *daddr; |
4c9483b2 | 1402 | if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) |
4e3fd7a0 | 1403 | fl6.saddr = np->saddr; |
1958b856 | 1404 | fl6.fl6_sport = inet->inet_sport; |
1ab1457c | 1405 | |
1cedee13 AI |
1406 | if (cgroup_bpf_enabled && !connected) { |
1407 | err = BPF_CGROUP_RUN_PROG_UDP6_SENDMSG_LOCK(sk, | |
1408 | (struct sockaddr *)sin6, &fl6.saddr); | |
1409 | if (err) | |
1410 | goto out_no_dst; | |
1411 | if (sin6) { | |
1412 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) { | |
1413 | /* BPF program rewrote IPv6-only by IPv4-mapped | |
1414 | * IPv6. It's currently unsupported. | |
1415 | */ | |
1416 | err = -ENOTSUPP; | |
1417 | goto out_no_dst; | |
1418 | } | |
1419 | if (sin6->sin6_port == 0) { | |
1420 | /* BPF program set invalid port. Reject it. */ | |
1421 | err = -EINVAL; | |
1422 | goto out_no_dst; | |
1423 | } | |
1424 | fl6.fl6_dport = sin6->sin6_port; | |
1425 | fl6.daddr = sin6->sin6_addr; | |
1426 | } | |
1427 | } | |
1428 | ||
e8e36984 AI |
1429 | if (ipv6_addr_any(&fl6.daddr)) |
1430 | fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ | |
1431 | ||
4c9483b2 | 1432 | final_p = fl6_update_dst(&fl6, opt, &final); |
20c59de2 | 1433 | if (final_p) |
9f542f61 | 1434 | connected = false; |
1da177e4 | 1435 | |
4c9483b2 DM |
1436 | if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) { |
1437 | fl6.flowi6_oif = np->mcast_oif; | |
9f542f61 | 1438 | connected = false; |
c4062dfc EH |
1439 | } else if (!fl6.flowi6_oif) |
1440 | fl6.flowi6_oif = np->ucast_oif; | |
1da177e4 | 1441 | |
4c9483b2 | 1442 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); |
beb8d13b | 1443 | |
38b7097b HFS |
1444 | if (ipc6.tclass < 0) |
1445 | ipc6.tclass = np->tclass; | |
1446 | ||
1447 | fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel); | |
1448 | ||
4f858c56 | 1449 | dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, connected); |
68d0c6d3 DM |
1450 | if (IS_ERR(dst)) { |
1451 | err = PTR_ERR(dst); | |
1452 | dst = NULL; | |
1da177e4 | 1453 | goto out; |
14e50e57 | 1454 | } |
1da177e4 | 1455 | |
26879da5 WW |
1456 | if (ipc6.hlimit < 0) |
1457 | ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); | |
1da177e4 LT |
1458 | |
1459 | if (msg->msg_flags&MSG_CONFIRM) | |
1460 | goto do_confirm; | |
1461 | back_from_confirm: | |
1462 | ||
03485f2a VY |
1463 | /* Lockless fast path for the non-corking case */ |
1464 | if (!corkreq) { | |
1cd7884d | 1465 | struct inet_cork_full cork; |
03485f2a VY |
1466 | struct sk_buff *skb; |
1467 | ||
1468 | skb = ip6_make_skb(sk, getfrag, msg, ulen, | |
26879da5 | 1469 | sizeof(struct udphdr), &ipc6, |
03485f2a | 1470 | &fl6, (struct rt6_info *)dst, |
5fdaa88d | 1471 | msg->msg_flags, &cork); |
03485f2a VY |
1472 | err = PTR_ERR(skb); |
1473 | if (!IS_ERR_OR_NULL(skb)) | |
bec1f6f6 | 1474 | err = udp_v6_send_skb(skb, &fl6, &cork.base); |
4f858c56 | 1475 | goto out; |
03485f2a VY |
1476 | } |
1477 | ||
1da177e4 LT |
1478 | lock_sock(sk); |
1479 | if (unlikely(up->pending)) { | |
1480 | /* The socket is already corked while preparing it. */ | |
1481 | /* ... which is an evident application bug. --ANK */ | |
1482 | release_sock(sk); | |
1483 | ||
ba7a46f1 | 1484 | net_dbg_ratelimited("udp cork app bug 2\n"); |
1da177e4 LT |
1485 | err = -EINVAL; |
1486 | goto out; | |
1487 | } | |
1488 | ||
1489 | up->pending = AF_INET6; | |
1490 | ||
1491 | do_append_data: | |
26879da5 WW |
1492 | if (ipc6.dontfrag < 0) |
1493 | ipc6.dontfrag = np->dontfrag; | |
1da177e4 | 1494 | up->len += ulen; |
26879da5 WW |
1495 | err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr), |
1496 | &ipc6, &fl6, (struct rt6_info *)dst, | |
5fdaa88d | 1497 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); |
1da177e4 LT |
1498 | if (err) |
1499 | udp_v6_flush_pending_frames(sk); | |
1500 | else if (!corkreq) | |
4c0a6cb0 | 1501 | err = udp_v6_push_pending_frames(sk); |
1e0c14f4 HX |
1502 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) |
1503 | up->pending = 0; | |
1da177e4 | 1504 | |
03485f2a VY |
1505 | if (err > 0) |
1506 | err = np->recverr ? net_xmit_errno(err) : 0; | |
1507 | release_sock(sk); | |
1508 | ||
1da177e4 | 1509 | out: |
a3c96089 | 1510 | dst_release(dst); |
1cedee13 | 1511 | out_no_dst: |
1da177e4 | 1512 | fl6_sock_release(flowlabel); |
45f6fad8 | 1513 | txopt_put(opt_to_free); |
cd562c98 | 1514 | if (!err) |
1da177e4 | 1515 | return len; |
a18135eb DM |
1516 | /* |
1517 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1518 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1519 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1520 | * things). We could add another new stat but at least for now that | |
1521 | * seems like overkill. | |
1522 | */ | |
1523 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
6aef70a8 ED |
1524 | UDP6_INC_STATS(sock_net(sk), |
1525 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
a18135eb | 1526 | } |
1da177e4 LT |
1527 | return err; |
1528 | ||
1529 | do_confirm: | |
0dec879f JA |
1530 | if (msg->msg_flags & MSG_PROBE) |
1531 | dst_confirm_neigh(dst, &fl6.daddr); | |
1da177e4 LT |
1532 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1533 | goto back_from_confirm; | |
1534 | err = 0; | |
1535 | goto out; | |
1536 | } | |
1537 | ||
7d06b2e0 | 1538 | void udpv6_destroy_sock(struct sock *sk) |
1da177e4 | 1539 | { |
44046a59 | 1540 | struct udp_sock *up = udp_sk(sk); |
1da177e4 LT |
1541 | lock_sock(sk); |
1542 | udp_v6_flush_pending_frames(sk); | |
1543 | release_sock(sk); | |
1544 | ||
60fb9567 PA |
1545 | if (static_branch_unlikely(&udpv6_encap_needed_key)) { |
1546 | if (up->encap_type) { | |
1547 | void (*encap_destroy)(struct sock *sk); | |
1548 | encap_destroy = READ_ONCE(up->encap_destroy); | |
1549 | if (encap_destroy) | |
1550 | encap_destroy(sk); | |
1551 | } | |
1552 | if (up->encap_enabled) | |
9c480601 | 1553 | static_branch_dec(&udpv6_encap_needed_key); |
44046a59 TP |
1554 | } |
1555 | ||
1da177e4 | 1556 | inet6_destroy_sock(sk); |
1da177e4 LT |
1557 | } |
1558 | ||
1559 | /* | |
1560 | * Socket option code for UDP | |
1561 | */ | |
ba4e58ec | 1562 | int udpv6_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 1563 | char __user *optval, unsigned int optlen) |
3fdadf7d | 1564 | { |
db8dac20 | 1565 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 GR |
1566 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, |
1567 | udp_v6_push_pending_frames); | |
ba4e58ec | 1568 | return ipv6_setsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1569 | } |
1570 | ||
1571 | #ifdef CONFIG_COMPAT | |
ba4e58ec | 1572 | int compat_udpv6_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 1573 | char __user *optval, unsigned int optlen) |
3fdadf7d | 1574 | { |
db8dac20 | 1575 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 GR |
1576 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, |
1577 | udp_v6_push_pending_frames); | |
ba4e58ec | 1578 | return compat_ipv6_setsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1579 | } |
1580 | #endif | |
1581 | ||
ba4e58ec GR |
1582 | int udpv6_getsockopt(struct sock *sk, int level, int optname, |
1583 | char __user *optval, int __user *optlen) | |
3fdadf7d | 1584 | { |
db8dac20 | 1585 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 | 1586 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); |
ba4e58ec | 1587 | return ipv6_getsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1588 | } |
1589 | ||
1590 | #ifdef CONFIG_COMPAT | |
ba4e58ec GR |
1591 | int compat_udpv6_getsockopt(struct sock *sk, int level, int optname, |
1592 | char __user *optval, int __user *optlen) | |
3fdadf7d | 1593 | { |
db8dac20 | 1594 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 | 1595 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); |
ba4e58ec | 1596 | return compat_ipv6_getsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1597 | } |
1598 | #endif | |
1599 | ||
a8e3bb34 DA |
1600 | /* thinking of making this const? Don't. |
1601 | * early_demux can change based on sysctl. | |
1602 | */ | |
dddb64bc | 1603 | static struct inet6_protocol udpv6_protocol = { |
5425077d | 1604 | .early_demux = udp_v6_early_demux, |
dddb64bc | 1605 | .early_demux_handler = udp_v6_early_demux, |
1da177e4 LT |
1606 | .handler = udpv6_rcv, |
1607 | .err_handler = udpv6_err, | |
1608 | .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, | |
1609 | }; | |
1610 | ||
1611 | /* ------------------------------------------------------------------------ */ | |
1612 | #ifdef CONFIG_PROC_FS | |
ba4e58ec | 1613 | int udp6_seq_show(struct seq_file *seq, void *v) |
1da177e4 | 1614 | { |
17ef66af LC |
1615 | if (v == SEQ_START_TOKEN) { |
1616 | seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); | |
1617 | } else { | |
1618 | int bucket = ((struct udp_iter_state *)seq->private)->bucket; | |
1619 | struct inet_sock *inet = inet_sk(v); | |
1620 | __u16 srcp = ntohs(inet->inet_sport); | |
1621 | __u16 destp = ntohs(inet->inet_dport); | |
6c206b20 PA |
1622 | __ip6_dgram_sock_seq_show(seq, v, srcp, destp, |
1623 | udp_rqueue_get(v), bucket); | |
17ef66af | 1624 | } |
1da177e4 LT |
1625 | return 0; |
1626 | } | |
1627 | ||
c3506372 | 1628 | const struct seq_operations udp6_seq_ops = { |
a3d2599b CH |
1629 | .start = udp_seq_start, |
1630 | .next = udp_seq_next, | |
1631 | .stop = udp_seq_stop, | |
1632 | .show = udp6_seq_show, | |
1633 | }; | |
c3506372 | 1634 | EXPORT_SYMBOL(udp6_seq_ops); |
73cb88ec | 1635 | |
1da177e4 | 1636 | static struct udp_seq_afinfo udp6_seq_afinfo = { |
1da177e4 | 1637 | .family = AF_INET6, |
645ca708 | 1638 | .udp_table = &udp_table, |
1da177e4 LT |
1639 | }; |
1640 | ||
2c8c1e72 | 1641 | int __net_init udp6_proc_init(struct net *net) |
1da177e4 | 1642 | { |
c3506372 CH |
1643 | if (!proc_create_net_data("udp6", 0444, net->proc_net, &udp6_seq_ops, |
1644 | sizeof(struct udp_iter_state), &udp6_seq_afinfo)) | |
a3d2599b CH |
1645 | return -ENOMEM; |
1646 | return 0; | |
1da177e4 LT |
1647 | } |
1648 | ||
ec120da6 IM |
1649 | void udp6_proc_exit(struct net *net) |
1650 | { | |
a3d2599b | 1651 | remove_proc_entry("udp6", net->proc_net); |
1da177e4 LT |
1652 | } |
1653 | #endif /* CONFIG_PROC_FS */ | |
1654 | ||
1655 | /* ------------------------------------------------------------------------ */ | |
1656 | ||
1657 | struct proto udpv6_prot = { | |
1e802951 TZ |
1658 | .name = "UDPv6", |
1659 | .owner = THIS_MODULE, | |
1660 | .close = udp_lib_close, | |
d74bad4e | 1661 | .pre_connect = udpv6_pre_connect, |
1e802951 TZ |
1662 | .connect = ip6_datagram_connect, |
1663 | .disconnect = udp_disconnect, | |
1664 | .ioctl = udp_ioctl, | |
1665 | .init = udp_init_sock, | |
1666 | .destroy = udpv6_destroy_sock, | |
1667 | .setsockopt = udpv6_setsockopt, | |
1668 | .getsockopt = udpv6_getsockopt, | |
1669 | .sendmsg = udpv6_sendmsg, | |
1670 | .recvmsg = udpv6_recvmsg, | |
1671 | .release_cb = ip6_datagram_release_cb, | |
1672 | .hash = udp_lib_hash, | |
1673 | .unhash = udp_lib_unhash, | |
1674 | .rehash = udp_v6_rehash, | |
1675 | .get_port = udp_v6_get_port, | |
1676 | .memory_allocated = &udp_memory_allocated, | |
1677 | .sysctl_mem = sysctl_udp_mem, | |
1678 | .sysctl_wmem_offset = offsetof(struct net, ipv4.sysctl_udp_wmem_min), | |
1679 | .sysctl_rmem_offset = offsetof(struct net, ipv4.sysctl_udp_rmem_min), | |
1680 | .obj_size = sizeof(struct udp6_sock), | |
1681 | .h.udp_table = &udp_table, | |
3fdadf7d | 1682 | #ifdef CONFIG_COMPAT |
1e802951 TZ |
1683 | .compat_setsockopt = compat_udpv6_setsockopt, |
1684 | .compat_getsockopt = compat_udpv6_getsockopt, | |
3fdadf7d | 1685 | #endif |
1e802951 | 1686 | .diag_destroy = udp_abort, |
1da177e4 LT |
1687 | }; |
1688 | ||
1da177e4 LT |
1689 | static struct inet_protosw udpv6_protosw = { |
1690 | .type = SOCK_DGRAM, | |
1691 | .protocol = IPPROTO_UDP, | |
1692 | .prot = &udpv6_prot, | |
1693 | .ops = &inet6_dgram_ops, | |
1da177e4 LT |
1694 | .flags = INET_PROTOSW_PERMANENT, |
1695 | }; | |
1696 | ||
7f4e4868 | 1697 | int __init udpv6_init(void) |
1da177e4 | 1698 | { |
7f4e4868 DL |
1699 | int ret; |
1700 | ||
3336288a VY |
1701 | ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP); |
1702 | if (ret) | |
c6b641a4 | 1703 | goto out; |
3336288a | 1704 | |
7f4e4868 DL |
1705 | ret = inet6_register_protosw(&udpv6_protosw); |
1706 | if (ret) | |
1707 | goto out_udpv6_protocol; | |
1708 | out: | |
1709 | return ret; | |
1710 | ||
1711 | out_udpv6_protocol: | |
1712 | inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); | |
1713 | goto out; | |
1714 | } | |
1715 | ||
09f7709f | 1716 | void udpv6_exit(void) |
7f4e4868 DL |
1717 | { |
1718 | inet6_unregister_protosw(&udpv6_protosw); | |
1719 | inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); | |
1da177e4 | 1720 | } |