Commit | Line | Data |
---|---|---|
2874c5fd | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
1da177e4 LT |
2 | /* |
3 | * UDP over IPv6 | |
1ab1457c | 4 | * Linux INET6 implementation |
1da177e4 LT |
5 | * |
6 | * Authors: | |
1ab1457c | 7 | * Pedro Roque <roque@di.fc.ul.pt> |
1da177e4 LT |
8 | * |
9 | * Based on linux/ipv4/udp.c | |
10 | * | |
1da177e4 LT |
11 | * Fixes: |
12 | * Hideaki YOSHIFUJI : sin6_scope_id support | |
13 | * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which | |
14 | * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind | |
15 | * a single port at the same time. | |
16 | * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data | |
17 | * YOSHIFUJI Hideaki @USAGI: convert /proc/net/udp6 to seq_file. | |
1da177e4 LT |
18 | */ |
19 | ||
1da177e4 LT |
20 | #include <linux/errno.h> |
21 | #include <linux/types.h> | |
22 | #include <linux/socket.h> | |
23 | #include <linux/sockios.h> | |
1da177e4 LT |
24 | #include <linux/net.h> |
25 | #include <linux/in6.h> | |
26 | #include <linux/netdevice.h> | |
27 | #include <linux/if_arp.h> | |
28 | #include <linux/ipv6.h> | |
29 | #include <linux/icmpv6.h> | |
30 | #include <linux/init.h> | |
1781f7f5 | 31 | #include <linux/module.h> |
3305b80c | 32 | #include <linux/skbuff.h> |
5a0e3ad6 | 33 | #include <linux/slab.h> |
7c0f6ba6 | 34 | #include <linux/uaccess.h> |
0e219ae4 | 35 | #include <linux/indirect_call_wrapper.h> |
1da177e4 | 36 | |
496611d7 | 37 | #include <net/addrconf.h> |
1da177e4 LT |
38 | #include <net/ndisc.h> |
39 | #include <net/protocol.h> | |
40 | #include <net/transp_v6.h> | |
41 | #include <net/ip6_route.h> | |
1da177e4 | 42 | #include <net/raw.h> |
c752f073 | 43 | #include <net/tcp_states.h> |
1da177e4 | 44 | #include <net/ip6_checksum.h> |
e7cc0824 | 45 | #include <net/ip6_tunnel.h> |
1da177e4 | 46 | #include <net/xfrm.h> |
0bd84065 | 47 | #include <net/inet_hashtables.h> |
72289b96 | 48 | #include <net/inet6_hashtables.h> |
076bb0c8 | 49 | #include <net/busy_poll.h> |
e32ea7e7 | 50 | #include <net/sock_reuseport.h> |
1da177e4 LT |
51 | |
52 | #include <linux/proc_fs.h> | |
53 | #include <linux/seq_file.h> | |
22911fc5 | 54 | #include <trace/events/skb.h> |
ba4e58ec | 55 | #include "udp_impl.h" |
1da177e4 | 56 | |
6eada011 ED |
57 | static u32 udp6_ehashfn(const struct net *net, |
58 | const struct in6_addr *laddr, | |
59 | const u16 lport, | |
60 | const struct in6_addr *faddr, | |
61 | const __be16 fport) | |
b50026b5 | 62 | { |
1bbdceef HFS |
63 | static u32 udp6_ehash_secret __read_mostly; |
64 | static u32 udp_ipv6_hash_secret __read_mostly; | |
65 | ||
66 | u32 lhash, fhash; | |
67 | ||
68 | net_get_random_once(&udp6_ehash_secret, | |
69 | sizeof(udp6_ehash_secret)); | |
70 | net_get_random_once(&udp_ipv6_hash_secret, | |
71 | sizeof(udp_ipv6_hash_secret)); | |
72 | ||
73 | lhash = (__force u32)laddr->s6_addr32[3]; | |
74 | fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret); | |
75 | ||
b50026b5 | 76 | return __inet6_ehashfn(lhash, lport, fhash, fport, |
1bbdceef | 77 | udp_ipv6_hash_secret + net_hash_mix(net)); |
b50026b5 HFS |
78 | } |
79 | ||
6ba5a3c5 | 80 | int udp_v6_get_port(struct sock *sk, unsigned short snum) |
1da177e4 | 81 | { |
30fff923 | 82 | unsigned int hash2_nulladdr = |
f0b1e64c | 83 | ipv6_portaddr_hash(sock_net(sk), &in6addr_any, snum); |
9a52e97e | 84 | unsigned int hash2_partial = |
f0b1e64c | 85 | ipv6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0); |
30fff923 | 86 | |
d4cada4a | 87 | /* precompute partial secondary hash */ |
30fff923 | 88 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
fe38d2a1 | 89 | return udp_lib_get_port(sk, snum, hash2_nulladdr); |
1da177e4 LT |
90 | } |
91 | ||
f7c46156 | 92 | void udp_v6_rehash(struct sock *sk) |
719f8358 | 93 | { |
f0b1e64c | 94 | u16 new_hash = ipv6_portaddr_hash(sock_net(sk), |
efe4208f | 95 | &sk->sk_v6_rcv_saddr, |
719f8358 ED |
96 | inet_sk(sk)->inet_num); |
97 | ||
98 | udp_lib_rehash(sk, new_hash); | |
99 | } | |
100 | ||
d1e37288 SX |
101 | static int compute_score(struct sock *sk, struct net *net, |
102 | const struct in6_addr *saddr, __be16 sport, | |
103 | const struct in6_addr *daddr, unsigned short hnum, | |
73545373 | 104 | int dif, int sdif) |
645ca708 | 105 | { |
60c04aec JP |
106 | int score; |
107 | struct inet_sock *inet; | |
6da5b0f0 | 108 | bool dev_match; |
60c04aec JP |
109 | |
110 | if (!net_eq(sock_net(sk), net) || | |
111 | udp_sk(sk)->udp_port_hash != hnum || | |
112 | sk->sk_family != PF_INET6) | |
113 | return -1; | |
114 | ||
23b0269e PO |
115 | if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr)) |
116 | return -1; | |
117 | ||
60c04aec JP |
118 | score = 0; |
119 | inet = inet_sk(sk); | |
120 | ||
121 | if (inet->inet_dport) { | |
122 | if (inet->inet_dport != sport) | |
123 | return -1; | |
124 | score++; | |
125 | } | |
126 | ||
60c04aec JP |
127 | if (!ipv6_addr_any(&sk->sk_v6_daddr)) { |
128 | if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr)) | |
129 | return -1; | |
130 | score++; | |
131 | } | |
132 | ||
6da5b0f0 MM |
133 | dev_match = udp_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif); |
134 | if (!dev_match) | |
135 | return -1; | |
136 | score++; | |
60c04aec | 137 | |
70da268b ED |
138 | if (sk->sk_incoming_cpu == raw_smp_processor_id()) |
139 | score++; | |
140 | ||
645ca708 ED |
141 | return score; |
142 | } | |
143 | ||
d1e37288 | 144 | /* called with rcu_read_lock() */ |
fddc17de ED |
145 | static struct sock *udp6_lib_lookup2(struct net *net, |
146 | const struct in6_addr *saddr, __be16 sport, | |
1801b570 | 147 | const struct in6_addr *daddr, unsigned int hnum, |
73545373 TB |
148 | int dif, int sdif, struct udp_hslot *hslot2, |
149 | struct sk_buff *skb) | |
fddc17de ED |
150 | { |
151 | struct sock *sk, *result; | |
e94a62f5 | 152 | int score, badness; |
72289b96 | 153 | u32 hash = 0; |
fddc17de | 154 | |
fddc17de ED |
155 | result = NULL; |
156 | badness = -1; | |
ca065d0c | 157 | udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
d1e37288 | 158 | score = compute_score(sk, net, saddr, sport, |
73545373 | 159 | daddr, hnum, dif, sdif); |
fddc17de | 160 | if (score > badness) { |
e94a62f5 | 161 | if (sk->sk_reuseport) { |
b50026b5 HFS |
162 | hash = udp6_ehashfn(net, daddr, hnum, |
163 | saddr, sport); | |
ed0dfffd | 164 | |
ca065d0c | 165 | result = reuseport_select_sock(sk, hash, skb, |
ed0dfffd | 166 | sizeof(struct udphdr)); |
ca065d0c ED |
167 | if (result) |
168 | return result; | |
70da268b | 169 | } |
ca065d0c ED |
170 | result = sk; |
171 | badness = score; | |
fddc17de ED |
172 | } |
173 | } | |
fddc17de ED |
174 | return result; |
175 | } | |
176 | ||
ca065d0c | 177 | /* rcu_read_lock() must be held */ |
fce82338 | 178 | struct sock *__udp6_lib_lookup(struct net *net, |
1801b570 DA |
179 | const struct in6_addr *saddr, __be16 sport, |
180 | const struct in6_addr *daddr, __be16 dport, | |
181 | int dif, int sdif, struct udp_table *udptable, | |
182 | struct sk_buff *skb) | |
1da177e4 | 183 | { |
1da177e4 | 184 | unsigned short hnum = ntohs(dport); |
23b0269e PO |
185 | unsigned int hash2, slot2; |
186 | struct udp_hslot *hslot2; | |
187 | struct sock *result; | |
645ca708 | 188 | |
23b0269e PO |
189 | hash2 = ipv6_portaddr_hash(net, daddr, hnum); |
190 | slot2 = hash2 & udptable->mask; | |
191 | hslot2 = &udptable->hash2[slot2]; | |
192 | ||
193 | result = udp6_lib_lookup2(net, saddr, sport, | |
73545373 | 194 | daddr, hnum, dif, sdif, |
23b0269e PO |
195 | hslot2, skb); |
196 | if (!result) { | |
197 | hash2 = ipv6_portaddr_hash(net, &in6addr_any, hnum); | |
fddc17de | 198 | slot2 = hash2 & udptable->mask; |
23b0269e | 199 | |
fddc17de | 200 | hslot2 = &udptable->hash2[slot2]; |
fddc17de ED |
201 | |
202 | result = udp6_lib_lookup2(net, saddr, sport, | |
23b0269e | 203 | &in6addr_any, hnum, dif, sdif, |
73545373 | 204 | hslot2, skb); |
1da177e4 | 205 | } |
26f8113c | 206 | if (IS_ERR(result)) |
23b0269e | 207 | return NULL; |
1da177e4 LT |
208 | return result; |
209 | } | |
fce82338 | 210 | EXPORT_SYMBOL_GPL(__udp6_lib_lookup); |
1da177e4 | 211 | |
607c4aaf KK |
212 | static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb, |
213 | __be16 sport, __be16 dport, | |
645ca708 | 214 | struct udp_table *udptable) |
607c4aaf | 215 | { |
b71d1d42 | 216 | const struct ipv6hdr *iph = ipv6_hdr(skb); |
607c4aaf | 217 | |
ed7cbbce | 218 | return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport, |
adf30907 | 219 | &iph->daddr, dport, inet6_iif(skb), |
1801b570 | 220 | inet6_sdif(skb), udptable, skb); |
607c4aaf KK |
221 | } |
222 | ||
63058308 TH |
223 | struct sock *udp6_lib_lookup_skb(struct sk_buff *skb, |
224 | __be16 sport, __be16 dport) | |
225 | { | |
226 | const struct ipv6hdr *iph = ipv6_hdr(skb); | |
63058308 | 227 | |
ed7cbbce | 228 | return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport, |
63058308 | 229 | &iph->daddr, dport, inet6_iif(skb), |
257a525f | 230 | inet6_sdif(skb), &udp_table, NULL); |
63058308 TH |
231 | } |
232 | EXPORT_SYMBOL_GPL(udp6_lib_lookup_skb); | |
233 | ||
ca065d0c ED |
234 | /* Must be called under rcu_read_lock(). |
235 | * Does increment socket refcount. | |
236 | */ | |
6e86000c | 237 | #if IS_ENABLED(CONFIG_NF_TPROXY_IPV6) || IS_ENABLED(CONFIG_NF_SOCKET_IPV6) |
aa976fc0 BS |
238 | struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport, |
239 | const struct in6_addr *daddr, __be16 dport, int dif) | |
240 | { | |
ca065d0c ED |
241 | struct sock *sk; |
242 | ||
243 | sk = __udp6_lib_lookup(net, saddr, sport, daddr, dport, | |
1801b570 | 244 | dif, 0, &udp_table, NULL); |
41c6d650 | 245 | if (sk && !refcount_inc_not_zero(&sk->sk_refcnt)) |
ca065d0c ED |
246 | sk = NULL; |
247 | return sk; | |
aa976fc0 BS |
248 | } |
249 | EXPORT_SYMBOL_GPL(udp6_lib_lookup); | |
ca065d0c | 250 | #endif |
aa976fc0 | 251 | |
cb891fa6 PA |
252 | /* do not use the scratch area len for jumbogram: their length execeeds the |
253 | * scratch area space; note that the IP6CB flags is still in the first | |
254 | * cacheline, so checking for jumbograms is cheap | |
255 | */ | |
256 | static int udp6_skb_len(struct sk_buff *skb) | |
257 | { | |
258 | return unlikely(inet6_is_jumbogram(skb)) ? skb->len : udp_skb_len(skb); | |
259 | } | |
260 | ||
1da177e4 | 261 | /* |
67ba4152 IM |
262 | * This should be easy, if there is something there we |
263 | * return it, otherwise we block. | |
1da177e4 LT |
264 | */ |
265 | ||
1b784140 | 266 | int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, |
1da177e4 LT |
267 | int noblock, int flags, int *addr_len) |
268 | { | |
269 | struct ipv6_pinfo *np = inet6_sk(sk); | |
270 | struct inet_sock *inet = inet_sk(sk); | |
1ab1457c | 271 | struct sk_buff *skb; |
59c2cdae | 272 | unsigned int ulen, copied; |
fd69c399 | 273 | int off, err, peeking = flags & MSG_PEEK; |
759e5d00 | 274 | int is_udplite = IS_UDPLITE(sk); |
543fc3fb | 275 | struct udp_mib __percpu *mib; |
197c949e | 276 | bool checksum_valid = false; |
f26ba175 | 277 | int is_udp4; |
1da177e4 | 278 | |
1da177e4 | 279 | if (flags & MSG_ERRQUEUE) |
85fbaa75 | 280 | return ipv6_recv_error(sk, msg, len, addr_len); |
1da177e4 | 281 | |
4b340ae2 | 282 | if (np->rxpmtu && np->rxopt.bits.rxpmtu) |
85fbaa75 | 283 | return ipv6_recv_rxpmtu(sk, msg, len, addr_len); |
4b340ae2 | 284 | |
1da177e4 | 285 | try_again: |
a0917e0b | 286 | off = sk_peek_offset(sk, flags); |
fd69c399 | 287 | skb = __skb_recv_udp(sk, flags, noblock, &off, &err); |
1da177e4 | 288 | if (!skb) |
627d2d6b | 289 | return err; |
1da177e4 | 290 | |
cb891fa6 | 291 | ulen = udp6_skb_len(skb); |
59c2cdae | 292 | copied = len; |
627d2d6b | 293 | if (copied > ulen - off) |
294 | copied = ulen - off; | |
59c2cdae | 295 | else if (copied < ulen) |
1ab1457c | 296 | msg->msg_flags |= MSG_TRUNC; |
1da177e4 | 297 | |
f26ba175 | 298 | is_udp4 = (skb->protocol == htons(ETH_P_IP)); |
029a3743 | 299 | mib = __UDPX_MIB(sk, is_udp4); |
f26ba175 | 300 | |
ba4e58ec | 301 | /* |
759e5d00 HX |
302 | * If checksum is needed at all, try to do it while copying the |
303 | * data. If the data is truncated, or if we only want a partial | |
304 | * coverage checksum (UDP-Lite), do it before the copy. | |
ba4e58ec | 305 | */ |
ba4e58ec | 306 | |
d21dbdfe ED |
307 | if (copied < ulen || peeking || |
308 | (is_udplite && UDP_SKB_CB(skb)->partial_cov)) { | |
67a51780 PA |
309 | checksum_valid = udp_skb_csum_unnecessary(skb) || |
310 | !__udp_lib_checksum_complete(skb); | |
197c949e | 311 | if (!checksum_valid) |
1da177e4 | 312 | goto csum_copy_err; |
ba4e58ec GR |
313 | } |
314 | ||
67a51780 PA |
315 | if (checksum_valid || udp_skb_csum_unnecessary(skb)) { |
316 | if (udp_skb_is_linear(skb)) | |
317 | err = copy_linear_skb(skb, copied, off, &msg->msg_iter); | |
318 | else | |
319 | err = skb_copy_datagram_msg(skb, off, msg, copied); | |
320 | } else { | |
627d2d6b | 321 | err = skb_copy_and_csum_datagram_msg(skb, off, msg); |
1da177e4 LT |
322 | if (err == -EINVAL) |
323 | goto csum_copy_err; | |
324 | } | |
22911fc5 | 325 | if (unlikely(err)) { |
fd69c399 | 326 | if (!peeking) { |
979402b1 | 327 | atomic_inc(&sk->sk_drops); |
029a3743 | 328 | SNMP_INC_STATS(mib, UDP_MIB_INERRORS); |
979402b1 | 329 | } |
850cbadd | 330 | kfree_skb(skb); |
627d2d6b | 331 | return err; |
22911fc5 | 332 | } |
fd69c399 | 333 | if (!peeking) |
029a3743 | 334 | SNMP_INC_STATS(mib, UDP_MIB_INDATAGRAMS); |
cb75994e | 335 | |
3b885787 | 336 | sock_recv_ts_and_drops(msg, sk, skb); |
1da177e4 LT |
337 | |
338 | /* Copy the address. */ | |
339 | if (msg->msg_name) { | |
342dfc30 | 340 | DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
1da177e4 | 341 | sin6->sin6_family = AF_INET6; |
4bedb452 | 342 | sin6->sin6_port = udp_hdr(skb)->source; |
1da177e4 | 343 | sin6->sin6_flowinfo = 0; |
1da177e4 | 344 | |
842df073 | 345 | if (is_udp4) { |
b301e82c BH |
346 | ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, |
347 | &sin6->sin6_addr); | |
842df073 HFS |
348 | sin6->sin6_scope_id = 0; |
349 | } else { | |
4e3fd7a0 | 350 | sin6->sin6_addr = ipv6_hdr(skb)->saddr; |
842df073 HFS |
351 | sin6->sin6_scope_id = |
352 | ipv6_iface_scope_id(&sin6->sin6_addr, | |
4330487a | 353 | inet6_iif(skb)); |
1da177e4 | 354 | } |
bceaa902 | 355 | *addr_len = sizeof(*sin6); |
983695fa DB |
356 | |
357 | if (cgroup_bpf_enabled) | |
358 | BPF_CGROUP_RUN_PROG_UDP6_RECVMSG_LOCK(sk, | |
359 | (struct sockaddr *)sin6); | |
1da177e4 | 360 | } |
4b261c75 | 361 | |
bcd1665e PA |
362 | if (udp_sk(sk)->gro_enabled) |
363 | udp_cmsg_recv(msg, sk, skb); | |
364 | ||
4b261c75 HFS |
365 | if (np->rxopt.all) |
366 | ip6_datagram_recv_common_ctl(sk, msg, skb); | |
367 | ||
f26ba175 | 368 | if (is_udp4) { |
1da177e4 | 369 | if (inet->cmsg_flags) |
ad959036 | 370 | ip_cmsg_recv_offset(msg, sk, skb, |
10df8e61 | 371 | sizeof(struct udphdr), off); |
1da177e4 LT |
372 | } else { |
373 | if (np->rxopt.all) | |
4b261c75 | 374 | ip6_datagram_recv_specific_ctl(sk, msg, skb); |
1ab1457c | 375 | } |
1da177e4 | 376 | |
59c2cdae | 377 | err = copied; |
1da177e4 | 378 | if (flags & MSG_TRUNC) |
759e5d00 | 379 | err = ulen; |
1da177e4 | 380 | |
850cbadd | 381 | skb_consume_udp(sk, skb, peeking ? -err : err); |
1da177e4 LT |
382 | return err; |
383 | ||
384 | csum_copy_err: | |
2276f58a PA |
385 | if (!__sk_queue_drop_skb(sk, &udp_sk(sk)->reader_queue, skb, flags, |
386 | udp_skb_destructor)) { | |
029a3743 PA |
387 | SNMP_INC_STATS(mib, UDP_MIB_CSUMERRORS); |
388 | SNMP_INC_STATS(mib, UDP_MIB_INERRORS); | |
0856f939 | 389 | } |
850cbadd | 390 | kfree_skb(skb); |
1da177e4 | 391 | |
beb39db5 ED |
392 | /* starting over for a new packet, but check if we need to yield */ |
393 | cond_resched(); | |
9cfaa8de | 394 | msg->msg_flags &= ~MSG_TRUNC; |
1da177e4 LT |
395 | goto try_again; |
396 | } | |
397 | ||
a36e185e SB |
398 | DEFINE_STATIC_KEY_FALSE(udpv6_encap_needed_key); |
399 | void udpv6_encap_enable(void) | |
400 | { | |
9c480601 | 401 | static_branch_inc(&udpv6_encap_needed_key); |
a36e185e SB |
402 | } |
403 | EXPORT_SYMBOL(udpv6_encap_enable); | |
404 | ||
e7cc0824 SB |
405 | /* Handler for tunnels with arbitrary destination ports: no socket lookup, go |
406 | * through error handlers in encapsulations looking for a match. | |
407 | */ | |
408 | static int __udp6_lib_err_encap_no_sk(struct sk_buff *skb, | |
409 | struct inet6_skb_parm *opt, | |
424a7cd0 | 410 | u8 type, u8 code, int offset, __be32 info) |
e7cc0824 SB |
411 | { |
412 | int i; | |
413 | ||
414 | for (i = 0; i < MAX_IPTUN_ENCAP_OPS; i++) { | |
415 | int (*handler)(struct sk_buff *skb, struct inet6_skb_parm *opt, | |
424a7cd0 PA |
416 | u8 type, u8 code, int offset, __be32 info); |
417 | const struct ip6_tnl_encap_ops *encap; | |
e7cc0824 | 418 | |
424a7cd0 PA |
419 | encap = rcu_dereference(ip6tun_encaps[i]); |
420 | if (!encap) | |
e7cc0824 | 421 | continue; |
424a7cd0 | 422 | handler = encap->err_handler; |
e7cc0824 SB |
423 | if (handler && !handler(skb, opt, type, code, offset, info)) |
424 | return 0; | |
425 | } | |
426 | ||
427 | return -ENOENT; | |
428 | } | |
429 | ||
a36e185e SB |
430 | /* Try to match ICMP errors to UDP tunnels by looking up a socket without |
431 | * reversing source and destination port: this will match tunnels that force the | |
432 | * same destination port on both endpoints (e.g. VXLAN, GENEVE). Note that | |
433 | * lwtunnels might actually break this assumption by being configured with | |
434 | * different destination ports on endpoints, in this case we won't be able to | |
435 | * trace ICMP messages back to them. | |
436 | * | |
e7cc0824 SB |
437 | * If this doesn't match any socket, probe tunnels with arbitrary destination |
438 | * ports (e.g. FoU, GUE): there, the receiving socket is useless, as the port | |
439 | * we've sent packets to won't necessarily match the local destination port. | |
440 | * | |
a36e185e SB |
441 | * Then ask the tunnel implementation to match the error against a valid |
442 | * association. | |
443 | * | |
e7cc0824 SB |
444 | * Return an error if we can't find a match, the socket if we need further |
445 | * processing, zero otherwise. | |
a36e185e SB |
446 | */ |
447 | static struct sock *__udp6_lib_err_encap(struct net *net, | |
448 | const struct ipv6hdr *hdr, int offset, | |
449 | struct udphdr *uh, | |
450 | struct udp_table *udptable, | |
e7cc0824 SB |
451 | struct sk_buff *skb, |
452 | struct inet6_skb_parm *opt, | |
453 | u8 type, u8 code, __be32 info) | |
a36e185e | 454 | { |
a36e185e | 455 | int network_offset, transport_offset; |
a36e185e SB |
456 | struct sock *sk; |
457 | ||
a36e185e SB |
458 | network_offset = skb_network_offset(skb); |
459 | transport_offset = skb_transport_offset(skb); | |
460 | ||
461 | /* Network header needs to point to the outer IPv6 header inside ICMP */ | |
462 | skb_reset_network_header(skb); | |
463 | ||
464 | /* Transport header needs to point to the UDP header */ | |
465 | skb_set_transport_header(skb, offset); | |
466 | ||
e7cc0824 SB |
467 | sk = __udp6_lib_lookup(net, &hdr->daddr, uh->source, |
468 | &hdr->saddr, uh->dest, | |
469 | inet6_iif(skb), 0, udptable, skb); | |
470 | if (sk) { | |
471 | int (*lookup)(struct sock *sk, struct sk_buff *skb); | |
472 | struct udp_sock *up = udp_sk(sk); | |
473 | ||
474 | lookup = READ_ONCE(up->encap_err_lookup); | |
475 | if (!lookup || lookup(sk, skb)) | |
476 | sk = NULL; | |
477 | } | |
478 | ||
479 | if (!sk) { | |
480 | sk = ERR_PTR(__udp6_lib_err_encap_no_sk(skb, opt, type, code, | |
481 | offset, info)); | |
482 | } | |
a36e185e SB |
483 | |
484 | skb_set_transport_header(skb, transport_offset); | |
485 | skb_set_network_header(skb, network_offset); | |
e7cc0824 | 486 | |
a36e185e SB |
487 | return sk; |
488 | } | |
489 | ||
32bbd879 SB |
490 | int __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt, |
491 | u8 type, u8 code, int offset, __be32 info, | |
492 | struct udp_table *udptable) | |
1da177e4 LT |
493 | { |
494 | struct ipv6_pinfo *np; | |
b71d1d42 ED |
495 | const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data; |
496 | const struct in6_addr *saddr = &hdr->saddr; | |
497 | const struct in6_addr *daddr = &hdr->daddr; | |
67ba4152 | 498 | struct udphdr *uh = (struct udphdr *)(skb->data+offset); |
a36e185e | 499 | bool tunnel = false; |
1da177e4 | 500 | struct sock *sk; |
e0d8c1b7 | 501 | int harderr; |
1da177e4 | 502 | int err; |
7304fe46 | 503 | struct net *net = dev_net(skb->dev); |
1da177e4 | 504 | |
e32ea7e7 | 505 | sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source, |
4ac30c4b | 506 | inet6_iif(skb), inet6_sdif(skb), udptable, NULL); |
63159f29 | 507 | if (!sk) { |
a36e185e | 508 | /* No socket for error: try tunnels before discarding */ |
e7cc0824 | 509 | sk = ERR_PTR(-ENOENT); |
a36e185e SB |
510 | if (static_branch_unlikely(&udpv6_encap_needed_key)) { |
511 | sk = __udp6_lib_err_encap(net, hdr, offset, uh, | |
e7cc0824 SB |
512 | udptable, skb, |
513 | opt, type, code, info); | |
514 | if (!sk) | |
515 | return 0; | |
a36e185e SB |
516 | } |
517 | ||
e7cc0824 | 518 | if (IS_ERR(sk)) { |
a36e185e SB |
519 | __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), |
520 | ICMP6_MIB_INERRORS); | |
e7cc0824 | 521 | return PTR_ERR(sk); |
a36e185e | 522 | } |
e7cc0824 | 523 | |
a36e185e | 524 | tunnel = true; |
7304fe46 | 525 | } |
1da177e4 | 526 | |
e0d8c1b7 WW |
527 | harderr = icmpv6_err_convert(type, code, &err); |
528 | np = inet6_sk(sk); | |
529 | ||
93b36cf3 HFS |
530 | if (type == ICMPV6_PKT_TOOBIG) { |
531 | if (!ip6_sk_accept_pmtu(sk)) | |
532 | goto out; | |
81aded24 | 533 | ip6_sk_update_pmtu(skb, sk, info); |
e0d8c1b7 WW |
534 | if (np->pmtudisc != IPV6_PMTUDISC_DONT) |
535 | harderr = 1; | |
93b36cf3 | 536 | } |
1a462d18 | 537 | if (type == NDISC_REDIRECT) { |
a36e185e SB |
538 | if (tunnel) { |
539 | ip6_redirect(skb, sock_net(sk), inet6_iif(skb), | |
540 | sk->sk_mark, sk->sk_uid); | |
541 | } else { | |
542 | ip6_sk_redirect(skb, sk); | |
543 | } | |
1a462d18 DJ |
544 | goto out; |
545 | } | |
81aded24 | 546 | |
a36e185e SB |
547 | /* Tunnels don't have an application socket: don't pass errors back */ |
548 | if (tunnel) | |
549 | goto out; | |
550 | ||
e0d8c1b7 WW |
551 | if (!np->recverr) { |
552 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
553 | goto out; | |
554 | } else { | |
1da177e4 | 555 | ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1)); |
e0d8c1b7 | 556 | } |
b1faf566 | 557 | |
1da177e4 LT |
558 | sk->sk_err = err; |
559 | sk->sk_error_report(sk); | |
560 | out: | |
32bbd879 | 561 | return 0; |
1da177e4 LT |
562 | } |
563 | ||
a3f96c47 | 564 | static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
f7ad74fe BL |
565 | { |
566 | int rc; | |
567 | ||
efe4208f | 568 | if (!ipv6_addr_any(&sk->sk_v6_daddr)) { |
f7ad74fe | 569 | sock_rps_save_rxhash(sk, skb); |
005ec974 | 570 | sk_mark_napi_id(sk, skb); |
2c8c56e1 | 571 | sk_incoming_cpu_update(sk); |
e68b6e50 ED |
572 | } else { |
573 | sk_mark_napi_id_once(sk, skb); | |
005ec974 | 574 | } |
f7ad74fe | 575 | |
850cbadd | 576 | rc = __udp_enqueue_schedule_skb(sk, skb); |
f7ad74fe BL |
577 | if (rc < 0) { |
578 | int is_udplite = IS_UDPLITE(sk); | |
579 | ||
580 | /* Note that an ENOMEM error is charged twice */ | |
581 | if (rc == -ENOMEM) | |
e61da9e2 | 582 | UDP6_INC_STATS(sock_net(sk), |
02c22347 | 583 | UDP_MIB_RCVBUFERRORS, is_udplite); |
e61da9e2 | 584 | UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
f7ad74fe BL |
585 | kfree_skb(skb); |
586 | return -1; | |
587 | } | |
850cbadd | 588 | |
f7ad74fe BL |
589 | return 0; |
590 | } | |
591 | ||
32bbd879 SB |
592 | static __inline__ int udpv6_err(struct sk_buff *skb, |
593 | struct inet6_skb_parm *opt, u8 type, | |
594 | u8 code, int offset, __be32 info) | |
ba4e58ec | 595 | { |
32bbd879 | 596 | return __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table); |
ba4e58ec GR |
597 | } |
598 | ||
cf329aa4 | 599 | static int udpv6_queue_rcv_one_skb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 600 | { |
ba4e58ec | 601 | struct udp_sock *up = udp_sk(sk); |
b2bf1e26 | 602 | int is_udplite = IS_UDPLITE(sk); |
a18135eb | 603 | |
ba4e58ec GR |
604 | if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) |
605 | goto drop; | |
1da177e4 | 606 | |
88ab3108 | 607 | if (static_branch_unlikely(&udpv6_encap_needed_key) && up->encap_type) { |
d7f3f621 BL |
608 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
609 | ||
610 | /* | |
611 | * This is an encapsulation socket so pass the skb to | |
612 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
613 | * fall through and pass this up the UDP socket. | |
614 | * up->encap_rcv() returns the following value: | |
615 | * =0 if skb was successfully passed to the encap | |
616 | * handler or was discarded by it. | |
617 | * >0 if skb should be passed on to UDP. | |
618 | * <0 if skb should be resubmitted as proto -N | |
619 | */ | |
620 | ||
621 | /* if we're overly short, let UDP handle it */ | |
6aa7de05 | 622 | encap_rcv = READ_ONCE(up->encap_rcv); |
e5aed006 | 623 | if (encap_rcv) { |
d7f3f621 BL |
624 | int ret; |
625 | ||
0a80966b TH |
626 | /* Verify checksum before giving to encap */ |
627 | if (udp_lib_checksum_complete(skb)) | |
628 | goto csum_error; | |
629 | ||
d7f3f621 BL |
630 | ret = encap_rcv(sk, skb); |
631 | if (ret <= 0) { | |
02c22347 ED |
632 | __UDP_INC_STATS(sock_net(sk), |
633 | UDP_MIB_INDATAGRAMS, | |
634 | is_udplite); | |
d7f3f621 BL |
635 | return -ret; |
636 | } | |
637 | } | |
638 | ||
639 | /* FALLTHROUGH -- it's a UDP Packet */ | |
640 | } | |
641 | ||
ba4e58ec GR |
642 | /* |
643 | * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c). | |
644 | */ | |
b2bf1e26 | 645 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { |
ba4e58ec GR |
646 | |
647 | if (up->pcrlen == 0) { /* full coverage was set */ | |
ba7a46f1 JP |
648 | net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n", |
649 | UDP_SKB_CB(skb)->cscov, skb->len); | |
ba4e58ec GR |
650 | goto drop; |
651 | } | |
652 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
ba7a46f1 JP |
653 | net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n", |
654 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
ba4e58ec GR |
655 | goto drop; |
656 | } | |
1da177e4 LT |
657 | } |
658 | ||
4b943fae | 659 | prefetch(&sk->sk_rmem_alloc); |
ce25d66a ED |
660 | if (rcu_access_pointer(sk->sk_filter) && |
661 | udp_lib_checksum_complete(skb)) | |
662 | goto csum_error; | |
663 | ||
ba66bbe5 | 664 | if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr))) |
a6127697 | 665 | goto drop; |
ba4e58ec | 666 | |
e6afc8ac | 667 | udp_csum_pull_header(skb); |
cb80ef46 | 668 | |
d826eb14 | 669 | skb_dst_drop(skb); |
cb75994e | 670 | |
850cbadd | 671 | return __udpv6_queue_rcv_skb(sk, skb); |
3e215c8d | 672 | |
6a5dc9e5 | 673 | csum_error: |
02c22347 | 674 | __UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); |
ba4e58ec | 675 | drop: |
02c22347 | 676 | __UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
cb80ef46 | 677 | atomic_inc(&sk->sk_drops); |
ba4e58ec GR |
678 | kfree_skb(skb); |
679 | return -1; | |
1da177e4 LT |
680 | } |
681 | ||
cf329aa4 PA |
682 | static int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
683 | { | |
684 | struct sk_buff *next, *segs; | |
685 | int ret; | |
686 | ||
687 | if (likely(!udp_unexpected_gso(sk, skb))) | |
688 | return udpv6_queue_rcv_one_skb(sk, skb); | |
689 | ||
690 | __skb_push(skb, -skb_mac_offset(skb)); | |
691 | segs = udp_rcv_segment(sk, skb, false); | |
692 | for (skb = segs; skb; skb = next) { | |
693 | next = skb->next; | |
694 | __skb_pull(skb, skb_transport_offset(skb)); | |
695 | ||
696 | ret = udpv6_queue_rcv_one_skb(sk, skb); | |
697 | if (ret > 0) | |
698 | ip6_protocol_deliver_rcu(dev_net(skb->dev), skb, ret, | |
699 | true); | |
700 | } | |
701 | return 0; | |
702 | } | |
703 | ||
5cf3d461 DH |
704 | static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk, |
705 | __be16 loc_port, const struct in6_addr *loc_addr, | |
706 | __be16 rmt_port, const struct in6_addr *rmt_addr, | |
7bd2db40 | 707 | int dif, int sdif, unsigned short hnum) |
1da177e4 | 708 | { |
5cf3d461 | 709 | struct inet_sock *inet = inet_sk(sk); |
1da177e4 | 710 | |
5cf3d461 DH |
711 | if (!net_eq(sock_net(sk), net)) |
712 | return false; | |
713 | ||
714 | if (udp_sk(sk)->udp_port_hash != hnum || | |
715 | sk->sk_family != PF_INET6 || | |
716 | (inet->inet_dport && inet->inet_dport != rmt_port) || | |
717 | (!ipv6_addr_any(&sk->sk_v6_daddr) && | |
718 | !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) || | |
7bd2db40 | 719 | !udp_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif) || |
33b4b015 HR |
720 | (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) && |
721 | !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))) | |
5cf3d461 DH |
722 | return false; |
723 | if (!inet6_mc_check(sk, loc_addr, rmt_addr)) | |
724 | return false; | |
725 | return true; | |
1da177e4 LT |
726 | } |
727 | ||
4068579e TH |
728 | static void udp6_csum_zero_error(struct sk_buff *skb) |
729 | { | |
730 | /* RFC 2460 section 8.1 says that we SHOULD log | |
731 | * this error. Well, it is reasonable. | |
732 | */ | |
ba7a46f1 JP |
733 | net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n", |
734 | &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source), | |
735 | &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest)); | |
4068579e TH |
736 | } |
737 | ||
1da177e4 LT |
738 | /* |
739 | * Note: called only from the BH handler context, | |
740 | * so we don't need to lock the hashes. | |
741 | */ | |
e3163493 | 742 | static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
b71d1d42 | 743 | const struct in6_addr *saddr, const struct in6_addr *daddr, |
36cbb245 | 744 | struct udp_table *udptable, int proto) |
1da177e4 | 745 | { |
ca065d0c | 746 | struct sock *sk, *first = NULL; |
4bedb452 | 747 | const struct udphdr *uh = udp_hdr(skb); |
5cf3d461 DH |
748 | unsigned short hnum = ntohs(uh->dest); |
749 | struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum); | |
ca065d0c | 750 | unsigned int offset = offsetof(typeof(*sk), sk_node); |
2dc41cff | 751 | unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10); |
ca065d0c | 752 | int dif = inet6_iif(skb); |
7bd2db40 | 753 | int sdif = inet6_sdif(skb); |
ca065d0c ED |
754 | struct hlist_node *node; |
755 | struct sk_buff *nskb; | |
2dc41cff DH |
756 | |
757 | if (use_hash2) { | |
f0b1e64c | 758 | hash2_any = ipv6_portaddr_hash(net, &in6addr_any, hnum) & |
73e2d5e3 | 759 | udptable->mask; |
f0b1e64c | 760 | hash2 = ipv6_portaddr_hash(net, daddr, hnum) & udptable->mask; |
2dc41cff | 761 | start_lookup: |
73e2d5e3 | 762 | hslot = &udptable->hash2[hash2]; |
2dc41cff DH |
763 | offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node); |
764 | } | |
1da177e4 | 765 | |
ca065d0c ED |
766 | sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) { |
767 | if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr, | |
7bd2db40 DM |
768 | uh->source, saddr, dif, sdif, |
769 | hnum)) | |
ca065d0c ED |
770 | continue; |
771 | /* If zero checksum and no_check is not on for | |
772 | * the socket then skip it. | |
773 | */ | |
774 | if (!uh->check && !udp_sk(sk)->no_check6_rx) | |
775 | continue; | |
776 | if (!first) { | |
777 | first = sk; | |
778 | continue; | |
779 | } | |
780 | nskb = skb_clone(skb, GFP_ATOMIC); | |
781 | if (unlikely(!nskb)) { | |
782 | atomic_inc(&sk->sk_drops); | |
02c22347 ED |
783 | __UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS, |
784 | IS_UDPLITE(sk)); | |
785 | __UDP6_INC_STATS(net, UDP_MIB_INERRORS, | |
786 | IS_UDPLITE(sk)); | |
ca065d0c | 787 | continue; |
95766fff | 788 | } |
a1ab77f9 | 789 | |
ca065d0c ED |
790 | if (udpv6_queue_rcv_skb(sk, nskb) > 0) |
791 | consume_skb(nskb); | |
792 | } | |
a1ab77f9 | 793 | |
2dc41cff DH |
794 | /* Also lookup *:port if we are using hash2 and haven't done so yet. */ |
795 | if (use_hash2 && hash2 != hash2_any) { | |
796 | hash2 = hash2_any; | |
797 | goto start_lookup; | |
798 | } | |
799 | ||
ca065d0c ED |
800 | if (first) { |
801 | if (udpv6_queue_rcv_skb(first, skb) > 0) | |
802 | consume_skb(skb); | |
a1ab77f9 | 803 | } else { |
ca065d0c | 804 | kfree_skb(skb); |
02c22347 ED |
805 | __UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI, |
806 | proto == IPPROTO_UDPLITE); | |
a1ab77f9 | 807 | } |
ba4e58ec | 808 | return 0; |
1da177e4 LT |
809 | } |
810 | ||
64f0f5d1 PA |
811 | static void udp6_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst) |
812 | { | |
813 | if (udp_sk_rx_dst_set(sk, dst)) { | |
814 | const struct rt6_info *rt = (const struct rt6_info *)dst; | |
815 | ||
816 | inet6_sk(sk)->rx_dst_cookie = rt6_get_cookie(rt); | |
817 | } | |
818 | } | |
819 | ||
eb63f296 PA |
820 | /* wrapper for udp_queue_rcv_skb tacking care of csum conversion and |
821 | * return code conversion for ip layer consumption | |
822 | */ | |
823 | static int udp6_unicast_rcv_skb(struct sock *sk, struct sk_buff *skb, | |
824 | struct udphdr *uh) | |
825 | { | |
826 | int ret; | |
827 | ||
828 | if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk)) | |
e4aa33ad | 829 | skb_checksum_try_convert(skb, IPPROTO_UDP, ip6_compute_pseudo); |
eb63f296 PA |
830 | |
831 | ret = udpv6_queue_rcv_skb(sk, skb); | |
832 | ||
84dad559 | 833 | /* a return value > 0 means to resubmit the input */ |
eb63f296 | 834 | if (ret > 0) |
84dad559 | 835 | return ret; |
eb63f296 PA |
836 | return 0; |
837 | } | |
838 | ||
645ca708 | 839 | int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
759e5d00 | 840 | int proto) |
1da177e4 | 841 | { |
ca065d0c | 842 | const struct in6_addr *saddr, *daddr; |
3ffe533c | 843 | struct net *net = dev_net(skb->dev); |
1ab1457c | 844 | struct udphdr *uh; |
ca065d0c | 845 | struct sock *sk; |
1da177e4 LT |
846 | u32 ulen = 0; |
847 | ||
848 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
d6bc0149 | 849 | goto discard; |
1da177e4 | 850 | |
0660e03f ACM |
851 | saddr = &ipv6_hdr(skb)->saddr; |
852 | daddr = &ipv6_hdr(skb)->daddr; | |
4bedb452 | 853 | uh = udp_hdr(skb); |
1da177e4 LT |
854 | |
855 | ulen = ntohs(uh->len); | |
ba4e58ec GR |
856 | if (ulen > skb->len) |
857 | goto short_packet; | |
1da177e4 | 858 | |
759e5d00 HX |
859 | if (proto == IPPROTO_UDP) { |
860 | /* UDP validates ulen. */ | |
1da177e4 | 861 | |
ba4e58ec GR |
862 | /* Check for jumbo payload */ |
863 | if (ulen == 0) | |
864 | ulen = skb->len; | |
1da177e4 | 865 | |
ba4e58ec GR |
866 | if (ulen < sizeof(*uh)) |
867 | goto short_packet; | |
1da177e4 | 868 | |
ba4e58ec GR |
869 | if (ulen < skb->len) { |
870 | if (pskb_trim_rcsum(skb, ulen)) | |
871 | goto short_packet; | |
0660e03f ACM |
872 | saddr = &ipv6_hdr(skb)->saddr; |
873 | daddr = &ipv6_hdr(skb)->daddr; | |
4bedb452 | 874 | uh = udp_hdr(skb); |
ba4e58ec | 875 | } |
ba4e58ec | 876 | } |
1da177e4 | 877 | |
759e5d00 | 878 | if (udp6_csum_init(skb, uh, proto)) |
6a5dc9e5 | 879 | goto csum_error; |
759e5d00 | 880 | |
c9f2c1ae PA |
881 | /* Check if the socket is already available, e.g. due to early demux */ |
882 | sk = skb_steal_sock(skb); | |
883 | if (sk) { | |
884 | struct dst_entry *dst = skb_dst(skb); | |
885 | int ret; | |
886 | ||
887 | if (unlikely(sk->sk_rx_dst != dst)) | |
64f0f5d1 | 888 | udp6_sk_rx_dst_set(sk, dst); |
c9f2c1ae | 889 | |
eb63f296 PA |
890 | if (!uh->check && !udp_sk(sk)->no_check6_rx) { |
891 | sock_put(sk); | |
892 | goto report_csum_error; | |
893 | } | |
c9f2c1ae | 894 | |
eb63f296 PA |
895 | ret = udp6_unicast_rcv_skb(sk, skb, uh); |
896 | sock_put(sk); | |
897 | return ret; | |
c9f2c1ae PA |
898 | } |
899 | ||
1ab1457c YH |
900 | /* |
901 | * Multicast receive code | |
1da177e4 | 902 | */ |
ba4e58ec | 903 | if (ipv6_addr_is_multicast(daddr)) |
e3163493 | 904 | return __udp6_lib_mcast_deliver(net, skb, |
36cbb245 | 905 | saddr, daddr, udptable, proto); |
1da177e4 LT |
906 | |
907 | /* Unicast */ | |
607c4aaf | 908 | sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable); |
53b24b8f | 909 | if (sk) { |
eb63f296 PA |
910 | if (!uh->check && !udp_sk(sk)->no_check6_rx) |
911 | goto report_csum_error; | |
912 | return udp6_unicast_rcv_skb(sk, skb, uh); | |
1da177e4 | 913 | } |
1ab1457c | 914 | |
eb63f296 PA |
915 | if (!uh->check) |
916 | goto report_csum_error; | |
4068579e | 917 | |
cb80ef46 | 918 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) |
c377411f | 919 | goto discard; |
cb80ef46 BL |
920 | |
921 | if (udp_lib_checksum_complete(skb)) | |
6a5dc9e5 | 922 | goto csum_error; |
cb80ef46 | 923 | |
02c22347 | 924 | __UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
cb80ef46 BL |
925 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
926 | ||
927 | kfree_skb(skb); | |
add459aa | 928 | return 0; |
1da177e4 | 929 | |
1ab1457c | 930 | short_packet: |
ba7a46f1 JP |
931 | net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n", |
932 | proto == IPPROTO_UDPLITE ? "-Lite" : "", | |
933 | saddr, ntohs(uh->source), | |
934 | ulen, skb->len, | |
935 | daddr, ntohs(uh->dest)); | |
6a5dc9e5 | 936 | goto discard; |
eb63f296 PA |
937 | |
938 | report_csum_error: | |
939 | udp6_csum_zero_error(skb); | |
6a5dc9e5 | 940 | csum_error: |
02c22347 | 941 | __UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); |
1da177e4 | 942 | discard: |
02c22347 | 943 | __UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
1da177e4 | 944 | kfree_skb(skb); |
add459aa | 945 | return 0; |
1da177e4 | 946 | } |
ba4e58ec | 947 | |
0bd84065 | 948 | |
5425077d | 949 | static struct sock *__udp6_lib_demux_lookup(struct net *net, |
950 | __be16 loc_port, const struct in6_addr *loc_addr, | |
951 | __be16 rmt_port, const struct in6_addr *rmt_addr, | |
4297a0ef | 952 | int dif, int sdif) |
5425077d | 953 | { |
0bd84065 | 954 | unsigned short hnum = ntohs(loc_port); |
f0b1e64c | 955 | unsigned int hash2 = ipv6_portaddr_hash(net, loc_addr, hnum); |
0bd84065 | 956 | unsigned int slot2 = hash2 & udp_table.mask; |
957 | struct udp_hslot *hslot2 = &udp_table.hash2[slot2]; | |
958 | const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum); | |
5425077d | 959 | struct sock *sk; |
960 | ||
0bd84065 | 961 | udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
85cb73ff | 962 | if (sk->sk_state == TCP_ESTABLISHED && |
4297a0ef | 963 | INET6_MATCH(sk, net, rmt_addr, loc_addr, ports, dif, sdif)) |
0bd84065 | 964 | return sk; |
965 | /* Only check first socket in chain */ | |
966 | break; | |
967 | } | |
968 | return NULL; | |
5425077d | 969 | } |
970 | ||
97ff7ffb | 971 | INDIRECT_CALLABLE_SCOPE void udp_v6_early_demux(struct sk_buff *skb) |
5425077d | 972 | { |
973 | struct net *net = dev_net(skb->dev); | |
974 | const struct udphdr *uh; | |
975 | struct sock *sk; | |
976 | struct dst_entry *dst; | |
977 | int dif = skb->dev->ifindex; | |
4297a0ef | 978 | int sdif = inet6_sdif(skb); |
5425077d | 979 | |
980 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + | |
981 | sizeof(struct udphdr))) | |
982 | return; | |
983 | ||
984 | uh = udp_hdr(skb); | |
985 | ||
986 | if (skb->pkt_type == PACKET_HOST) | |
987 | sk = __udp6_lib_demux_lookup(net, uh->dest, | |
988 | &ipv6_hdr(skb)->daddr, | |
989 | uh->source, &ipv6_hdr(skb)->saddr, | |
4297a0ef | 990 | dif, sdif); |
5425077d | 991 | else |
992 | return; | |
993 | ||
41c6d650 | 994 | if (!sk || !refcount_inc_not_zero(&sk->sk_refcnt)) |
5425077d | 995 | return; |
996 | ||
997 | skb->sk = sk; | |
998 | skb->destructor = sock_efree; | |
999 | dst = READ_ONCE(sk->sk_rx_dst); | |
1000 | ||
1001 | if (dst) | |
1002 | dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie); | |
1003 | if (dst) { | |
d24406c8 WW |
1004 | /* set noref for now. |
1005 | * any place which wants to hold dst has to call | |
1006 | * dst_hold_safe() | |
1007 | */ | |
1008 | skb_dst_set_noref(skb, dst); | |
5425077d | 1009 | } |
1010 | } | |
1011 | ||
0e219ae4 | 1012 | INDIRECT_CALLABLE_SCOPE int udpv6_rcv(struct sk_buff *skb) |
ba4e58ec | 1013 | { |
645ca708 | 1014 | return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
ba4e58ec GR |
1015 | } |
1016 | ||
1da177e4 LT |
1017 | /* |
1018 | * Throw away all pending data and cancel the corking. Socket is locked. | |
1019 | */ | |
1020 | static void udp_v6_flush_pending_frames(struct sock *sk) | |
1021 | { | |
1022 | struct udp_sock *up = udp_sk(sk); | |
1023 | ||
36d926b9 DL |
1024 | if (up->pending == AF_INET) |
1025 | udp_flush_pending_frames(sk); | |
1026 | else if (up->pending) { | |
1da177e4 LT |
1027 | up->len = 0; |
1028 | up->pending = 0; | |
1029 | ip6_flush_pending_frames(sk); | |
1ab1457c | 1030 | } |
1da177e4 LT |
1031 | } |
1032 | ||
d74bad4e AI |
1033 | static int udpv6_pre_connect(struct sock *sk, struct sockaddr *uaddr, |
1034 | int addr_len) | |
1035 | { | |
bddc028a TH |
1036 | if (addr_len < offsetofend(struct sockaddr, sa_family)) |
1037 | return -EINVAL; | |
d74bad4e AI |
1038 | /* The following checks are replicated from __ip6_datagram_connect() |
1039 | * and intended to prevent BPF program called below from accessing | |
1040 | * bytes that are out of the bound specified by user in addr_len. | |
1041 | */ | |
1042 | if (uaddr->sa_family == AF_INET) { | |
1043 | if (__ipv6_only_sock(sk)) | |
1044 | return -EAFNOSUPPORT; | |
1045 | return udp_pre_connect(sk, uaddr, addr_len); | |
1046 | } | |
1047 | ||
1048 | if (addr_len < SIN6_LEN_RFC2133) | |
1049 | return -EINVAL; | |
1050 | ||
1051 | return BPF_CGROUP_RUN_PROG_INET6_CONNECT_LOCK(sk, uaddr); | |
1052 | } | |
1053 | ||
493c6be3 | 1054 | /** |
67ba4152 IM |
1055 | * udp6_hwcsum_outgoing - handle outgoing HW checksumming |
1056 | * @sk: socket we are sending on | |
1057 | * @skb: sk_buff containing the filled-in UDP header | |
1058 | * (checksum field must be zeroed out) | |
493c6be3 SS |
1059 | */ |
1060 | static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb, | |
1061 | const struct in6_addr *saddr, | |
1062 | const struct in6_addr *daddr, int len) | |
1063 | { | |
1064 | unsigned int offset; | |
1065 | struct udphdr *uh = udp_hdr(skb); | |
d39d938c | 1066 | struct sk_buff *frags = skb_shinfo(skb)->frag_list; |
493c6be3 SS |
1067 | __wsum csum = 0; |
1068 | ||
d39d938c | 1069 | if (!frags) { |
493c6be3 SS |
1070 | /* Only one fragment on the socket. */ |
1071 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
1072 | skb->csum_offset = offsetof(struct udphdr, check); | |
1073 | uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0); | |
1074 | } else { | |
1075 | /* | |
1076 | * HW-checksum won't work as there are two or more | |
1077 | * fragments on the socket so that all csums of sk_buffs | |
1078 | * should be together | |
1079 | */ | |
1080 | offset = skb_transport_offset(skb); | |
1081 | skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); | |
63ecc3d9 | 1082 | csum = skb->csum; |
493c6be3 SS |
1083 | |
1084 | skb->ip_summed = CHECKSUM_NONE; | |
1085 | ||
d39d938c VY |
1086 | do { |
1087 | csum = csum_add(csum, frags->csum); | |
1088 | } while ((frags = frags->next)); | |
493c6be3 SS |
1089 | |
1090 | uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, | |
1091 | csum); | |
1092 | if (uh->check == 0) | |
1093 | uh->check = CSUM_MANGLED_0; | |
1094 | } | |
1095 | } | |
1096 | ||
1da177e4 LT |
1097 | /* |
1098 | * Sending | |
1099 | */ | |
1100 | ||
bec1f6f6 WB |
1101 | static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6, |
1102 | struct inet_cork *cork) | |
1da177e4 | 1103 | { |
d39d938c | 1104 | struct sock *sk = skb->sk; |
1da177e4 | 1105 | struct udphdr *uh; |
1da177e4 | 1106 | int err = 0; |
b2bf1e26 | 1107 | int is_udplite = IS_UDPLITE(sk); |
868c86bc | 1108 | __wsum csum = 0; |
d39d938c VY |
1109 | int offset = skb_transport_offset(skb); |
1110 | int len = skb->len - offset; | |
1da177e4 LT |
1111 | |
1112 | /* | |
1113 | * Create a UDP header | |
1114 | */ | |
4bedb452 | 1115 | uh = udp_hdr(skb); |
1958b856 DM |
1116 | uh->source = fl6->fl6_sport; |
1117 | uh->dest = fl6->fl6_dport; | |
d39d938c | 1118 | uh->len = htons(len); |
1da177e4 LT |
1119 | uh->check = 0; |
1120 | ||
bec1f6f6 WB |
1121 | if (cork->gso_size) { |
1122 | const int hlen = skb_network_header_len(skb) + | |
1123 | sizeof(struct udphdr); | |
1124 | ||
0f149c9f WB |
1125 | if (hlen + cork->gso_size > cork->fragsize) { |
1126 | kfree_skb(skb); | |
bec1f6f6 | 1127 | return -EINVAL; |
0f149c9f WB |
1128 | } |
1129 | if (skb->len > cork->gso_size * UDP_MAX_SEGMENTS) { | |
1130 | kfree_skb(skb); | |
bec1f6f6 | 1131 | return -EINVAL; |
0f149c9f WB |
1132 | } |
1133 | if (udp_sk(sk)->no_check6_tx) { | |
1134 | kfree_skb(skb); | |
a8c744a8 | 1135 | return -EINVAL; |
0f149c9f | 1136 | } |
ff06342c | 1137 | if (skb->ip_summed != CHECKSUM_PARTIAL || is_udplite || |
0f149c9f WB |
1138 | dst_xfrm(skb_dst(skb))) { |
1139 | kfree_skb(skb); | |
bec1f6f6 | 1140 | return -EIO; |
0f149c9f | 1141 | } |
bec1f6f6 WB |
1142 | |
1143 | skb_shinfo(skb)->gso_size = cork->gso_size; | |
1144 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP_L4; | |
a8c744a8 | 1145 | goto csum_partial; |
bec1f6f6 WB |
1146 | } |
1147 | ||
b2bf1e26 | 1148 | if (is_udplite) |
d39d938c VY |
1149 | csum = udplite_csum(skb); |
1150 | else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */ | |
4068579e TH |
1151 | skb->ip_summed = CHECKSUM_NONE; |
1152 | goto send; | |
1153 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
a8c744a8 | 1154 | csum_partial: |
d39d938c | 1155 | udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len); |
493c6be3 SS |
1156 | goto send; |
1157 | } else | |
d39d938c | 1158 | csum = udp_csum(skb); |
1da177e4 | 1159 | |
ba4e58ec | 1160 | /* add protocol-dependent pseudo-header */ |
4c9483b2 | 1161 | uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, |
d39d938c | 1162 | len, fl6->flowi6_proto, csum); |
1da177e4 | 1163 | if (uh->check == 0) |
f6ab0288 | 1164 | uh->check = CSUM_MANGLED_0; |
1da177e4 | 1165 | |
493c6be3 | 1166 | send: |
d39d938c | 1167 | err = ip6_send_skb(skb); |
6ce9e7b5 ED |
1168 | if (err) { |
1169 | if (err == -ENOBUFS && !inet6_sk(sk)->recverr) { | |
6aef70a8 ED |
1170 | UDP6_INC_STATS(sock_net(sk), |
1171 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
6ce9e7b5 ED |
1172 | err = 0; |
1173 | } | |
6aef70a8 ED |
1174 | } else { |
1175 | UDP6_INC_STATS(sock_net(sk), | |
1176 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
1177 | } | |
d39d938c VY |
1178 | return err; |
1179 | } | |
1180 | ||
1181 | static int udp_v6_push_pending_frames(struct sock *sk) | |
1182 | { | |
1183 | struct sk_buff *skb; | |
1184 | struct udp_sock *up = udp_sk(sk); | |
1185 | struct flowi6 fl6; | |
1186 | int err = 0; | |
1187 | ||
1188 | if (up->pending == AF_INET) | |
1189 | return udp_push_pending_frames(sk); | |
1190 | ||
1191 | /* ip6_finish_skb will release the cork, so make a copy of | |
1192 | * fl6 here. | |
1193 | */ | |
1194 | fl6 = inet_sk(sk)->cork.fl.u.ip6; | |
1195 | ||
1196 | skb = ip6_finish_skb(sk); | |
1197 | if (!skb) | |
1198 | goto out; | |
1199 | ||
bec1f6f6 | 1200 | err = udp_v6_send_skb(skb, &fl6, &inet_sk(sk)->cork.base); |
d39d938c | 1201 | |
1da177e4 LT |
1202 | out: |
1203 | up->len = 0; | |
1204 | up->pending = 0; | |
1205 | return err; | |
1206 | } | |
1207 | ||
1b784140 | 1208 | int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
1da177e4 LT |
1209 | { |
1210 | struct ipv6_txoptions opt_space; | |
1211 | struct udp_sock *up = udp_sk(sk); | |
1212 | struct inet_sock *inet = inet_sk(sk); | |
1213 | struct ipv6_pinfo *np = inet6_sk(sk); | |
342dfc30 | 1214 | DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
20c59de2 | 1215 | struct in6_addr *daddr, *final_p, final; |
1da177e4 | 1216 | struct ipv6_txoptions *opt = NULL; |
45f6fad8 | 1217 | struct ipv6_txoptions *opt_to_free = NULL; |
1da177e4 | 1218 | struct ip6_flowlabel *flowlabel = NULL; |
4c9483b2 | 1219 | struct flowi6 fl6; |
1da177e4 | 1220 | struct dst_entry *dst; |
26879da5 | 1221 | struct ipcm6_cookie ipc6; |
1da177e4 | 1222 | int addr_len = msg->msg_namelen; |
9f542f61 | 1223 | bool connected = false; |
1da177e4 | 1224 | int ulen = len; |
1da177e4 LT |
1225 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; |
1226 | int err; | |
b2bf1e26 | 1227 | int is_udplite = IS_UDPLITE(sk); |
ba4e58ec | 1228 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); |
1da177e4 | 1229 | |
b515430a | 1230 | ipcm6_init(&ipc6); |
bec1f6f6 | 1231 | ipc6.gso_size = up->gso_size; |
5fdaa88d | 1232 | ipc6.sockc.tsflags = sk->sk_tsflags; |
26879da5 | 1233 | |
1da177e4 LT |
1234 | /* destination address check */ |
1235 | if (sin6) { | |
1236 | if (addr_len < offsetof(struct sockaddr, sa_data)) | |
1237 | return -EINVAL; | |
1238 | ||
1239 | switch (sin6->sin6_family) { | |
1240 | case AF_INET6: | |
1241 | if (addr_len < SIN6_LEN_RFC2133) | |
1242 | return -EINVAL; | |
1243 | daddr = &sin6->sin6_addr; | |
052d2369 JL |
1244 | if (ipv6_addr_any(daddr) && |
1245 | ipv6_addr_v4mapped(&np->saddr)) | |
1246 | ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK), | |
1247 | daddr); | |
1da177e4 LT |
1248 | break; |
1249 | case AF_INET: | |
1250 | goto do_udp_sendmsg; | |
1251 | case AF_UNSPEC: | |
1252 | msg->msg_name = sin6 = NULL; | |
1253 | msg->msg_namelen = addr_len = 0; | |
1254 | daddr = NULL; | |
1255 | break; | |
1256 | default: | |
1257 | return -EINVAL; | |
1258 | } | |
1259 | } else if (!up->pending) { | |
1260 | if (sk->sk_state != TCP_ESTABLISHED) | |
1261 | return -EDESTADDRREQ; | |
efe4208f | 1262 | daddr = &sk->sk_v6_daddr; |
1ab1457c | 1263 | } else |
1da177e4 LT |
1264 | daddr = NULL; |
1265 | ||
1266 | if (daddr) { | |
e773e4fa | 1267 | if (ipv6_addr_v4mapped(daddr)) { |
1da177e4 LT |
1268 | struct sockaddr_in sin; |
1269 | sin.sin_family = AF_INET; | |
c720c7e8 | 1270 | sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport; |
1da177e4 LT |
1271 | sin.sin_addr.s_addr = daddr->s6_addr32[3]; |
1272 | msg->msg_name = &sin; | |
1273 | msg->msg_namelen = sizeof(sin); | |
1274 | do_udp_sendmsg: | |
1275 | if (__ipv6_only_sock(sk)) | |
1276 | return -ENETUNREACH; | |
1b784140 | 1277 | return udp_sendmsg(sk, msg, len); |
1da177e4 LT |
1278 | } |
1279 | } | |
1280 | ||
1281 | if (up->pending == AF_INET) | |
1b784140 | 1282 | return udp_sendmsg(sk, msg, len); |
1da177e4 LT |
1283 | |
1284 | /* Rough check on arithmetic overflow, | |
b59e139b | 1285 | better check is made in ip6_append_data(). |
1da177e4 LT |
1286 | */ |
1287 | if (len > INT_MAX - sizeof(struct udphdr)) | |
1288 | return -EMSGSIZE; | |
1ab1457c | 1289 | |
03485f2a | 1290 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
1da177e4 LT |
1291 | if (up->pending) { |
1292 | /* | |
1293 | * There are pending frames. | |
1294 | * The socket lock must be held while it's corked. | |
1295 | */ | |
1296 | lock_sock(sk); | |
1297 | if (likely(up->pending)) { | |
1298 | if (unlikely(up->pending != AF_INET6)) { | |
1299 | release_sock(sk); | |
1300 | return -EAFNOSUPPORT; | |
1301 | } | |
1302 | dst = NULL; | |
1303 | goto do_append_data; | |
1304 | } | |
1305 | release_sock(sk); | |
1306 | } | |
1307 | ulen += sizeof(struct udphdr); | |
1308 | ||
4c9483b2 | 1309 | memset(&fl6, 0, sizeof(fl6)); |
1da177e4 LT |
1310 | |
1311 | if (sin6) { | |
1312 | if (sin6->sin6_port == 0) | |
1313 | return -EINVAL; | |
1314 | ||
1958b856 | 1315 | fl6.fl6_dport = sin6->sin6_port; |
1da177e4 LT |
1316 | daddr = &sin6->sin6_addr; |
1317 | ||
1318 | if (np->sndflow) { | |
4c9483b2 DM |
1319 | fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; |
1320 | if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { | |
1321 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); | |
59c820b2 | 1322 | if (IS_ERR(flowlabel)) |
1da177e4 | 1323 | return -EINVAL; |
1da177e4 LT |
1324 | } |
1325 | } | |
1326 | ||
1327 | /* | |
1328 | * Otherwise it will be difficult to maintain | |
1329 | * sk->sk_dst_cache. | |
1330 | */ | |
1331 | if (sk->sk_state == TCP_ESTABLISHED && | |
efe4208f ED |
1332 | ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) |
1333 | daddr = &sk->sk_v6_daddr; | |
1da177e4 LT |
1334 | |
1335 | if (addr_len >= sizeof(struct sockaddr_in6) && | |
1336 | sin6->sin6_scope_id && | |
842df073 | 1337 | __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) |
4c9483b2 | 1338 | fl6.flowi6_oif = sin6->sin6_scope_id; |
1da177e4 LT |
1339 | } else { |
1340 | if (sk->sk_state != TCP_ESTABLISHED) | |
1341 | return -EDESTADDRREQ; | |
1342 | ||
1958b856 | 1343 | fl6.fl6_dport = inet->inet_dport; |
efe4208f | 1344 | daddr = &sk->sk_v6_daddr; |
4c9483b2 | 1345 | fl6.flowlabel = np->flow_label; |
9f542f61 | 1346 | connected = true; |
1da177e4 LT |
1347 | } |
1348 | ||
4c9483b2 DM |
1349 | if (!fl6.flowi6_oif) |
1350 | fl6.flowi6_oif = sk->sk_bound_dev_if; | |
1da177e4 | 1351 | |
4c9483b2 DM |
1352 | if (!fl6.flowi6_oif) |
1353 | fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex; | |
9f690db7 | 1354 | |
4c9483b2 | 1355 | fl6.flowi6_mark = sk->sk_mark; |
e2d118a1 | 1356 | fl6.flowi6_uid = sk->sk_uid; |
51953d5b | 1357 | |
1da177e4 LT |
1358 | if (msg->msg_controllen) { |
1359 | opt = &opt_space; | |
1360 | memset(opt, 0, sizeof(struct ipv6_txoptions)); | |
1361 | opt->tot_len = sizeof(*opt); | |
26879da5 | 1362 | ipc6.opt = opt; |
1da177e4 | 1363 | |
2e8de857 WB |
1364 | err = udp_cmsg_send(sk, msg, &ipc6.gso_size); |
1365 | if (err > 0) | |
1366 | err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, | |
5fdaa88d | 1367 | &ipc6); |
1da177e4 LT |
1368 | if (err < 0) { |
1369 | fl6_sock_release(flowlabel); | |
1370 | return err; | |
1371 | } | |
4c9483b2 DM |
1372 | if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { |
1373 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); | |
59c820b2 | 1374 | if (IS_ERR(flowlabel)) |
1da177e4 LT |
1375 | return -EINVAL; |
1376 | } | |
1377 | if (!(opt->opt_nflen|opt->opt_flen)) | |
1378 | opt = NULL; | |
9f542f61 | 1379 | connected = false; |
1da177e4 | 1380 | } |
45f6fad8 ED |
1381 | if (!opt) { |
1382 | opt = txopt_get(np); | |
1383 | opt_to_free = opt; | |
1384 | } | |
df9890c3 YH |
1385 | if (flowlabel) |
1386 | opt = fl6_merge_options(&opt_space, flowlabel, opt); | |
1387 | opt = ipv6_fixup_options(&opt_space, opt); | |
26879da5 | 1388 | ipc6.opt = opt; |
1da177e4 | 1389 | |
4c9483b2 | 1390 | fl6.flowi6_proto = sk->sk_protocol; |
e8e36984 | 1391 | fl6.daddr = *daddr; |
4c9483b2 | 1392 | if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) |
4e3fd7a0 | 1393 | fl6.saddr = np->saddr; |
1958b856 | 1394 | fl6.fl6_sport = inet->inet_sport; |
1ab1457c | 1395 | |
1cedee13 AI |
1396 | if (cgroup_bpf_enabled && !connected) { |
1397 | err = BPF_CGROUP_RUN_PROG_UDP6_SENDMSG_LOCK(sk, | |
1398 | (struct sockaddr *)sin6, &fl6.saddr); | |
1399 | if (err) | |
1400 | goto out_no_dst; | |
1401 | if (sin6) { | |
1402 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) { | |
1403 | /* BPF program rewrote IPv6-only by IPv4-mapped | |
1404 | * IPv6. It's currently unsupported. | |
1405 | */ | |
1406 | err = -ENOTSUPP; | |
1407 | goto out_no_dst; | |
1408 | } | |
1409 | if (sin6->sin6_port == 0) { | |
1410 | /* BPF program set invalid port. Reject it. */ | |
1411 | err = -EINVAL; | |
1412 | goto out_no_dst; | |
1413 | } | |
1414 | fl6.fl6_dport = sin6->sin6_port; | |
1415 | fl6.daddr = sin6->sin6_addr; | |
1416 | } | |
1417 | } | |
1418 | ||
e8e36984 AI |
1419 | if (ipv6_addr_any(&fl6.daddr)) |
1420 | fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ | |
1421 | ||
4c9483b2 | 1422 | final_p = fl6_update_dst(&fl6, opt, &final); |
20c59de2 | 1423 | if (final_p) |
9f542f61 | 1424 | connected = false; |
1da177e4 | 1425 | |
4c9483b2 DM |
1426 | if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) { |
1427 | fl6.flowi6_oif = np->mcast_oif; | |
9f542f61 | 1428 | connected = false; |
c4062dfc EH |
1429 | } else if (!fl6.flowi6_oif) |
1430 | fl6.flowi6_oif = np->ucast_oif; | |
1da177e4 | 1431 | |
4c9483b2 | 1432 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); |
beb8d13b | 1433 | |
38b7097b HFS |
1434 | if (ipc6.tclass < 0) |
1435 | ipc6.tclass = np->tclass; | |
1436 | ||
1437 | fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel); | |
1438 | ||
4f858c56 | 1439 | dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, connected); |
68d0c6d3 DM |
1440 | if (IS_ERR(dst)) { |
1441 | err = PTR_ERR(dst); | |
1442 | dst = NULL; | |
1da177e4 | 1443 | goto out; |
14e50e57 | 1444 | } |
1da177e4 | 1445 | |
26879da5 WW |
1446 | if (ipc6.hlimit < 0) |
1447 | ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); | |
1da177e4 LT |
1448 | |
1449 | if (msg->msg_flags&MSG_CONFIRM) | |
1450 | goto do_confirm; | |
1451 | back_from_confirm: | |
1452 | ||
03485f2a VY |
1453 | /* Lockless fast path for the non-corking case */ |
1454 | if (!corkreq) { | |
1cd7884d | 1455 | struct inet_cork_full cork; |
03485f2a VY |
1456 | struct sk_buff *skb; |
1457 | ||
1458 | skb = ip6_make_skb(sk, getfrag, msg, ulen, | |
26879da5 | 1459 | sizeof(struct udphdr), &ipc6, |
03485f2a | 1460 | &fl6, (struct rt6_info *)dst, |
5fdaa88d | 1461 | msg->msg_flags, &cork); |
03485f2a VY |
1462 | err = PTR_ERR(skb); |
1463 | if (!IS_ERR_OR_NULL(skb)) | |
bec1f6f6 | 1464 | err = udp_v6_send_skb(skb, &fl6, &cork.base); |
4f858c56 | 1465 | goto out; |
03485f2a VY |
1466 | } |
1467 | ||
1da177e4 LT |
1468 | lock_sock(sk); |
1469 | if (unlikely(up->pending)) { | |
1470 | /* The socket is already corked while preparing it. */ | |
1471 | /* ... which is an evident application bug. --ANK */ | |
1472 | release_sock(sk); | |
1473 | ||
ba7a46f1 | 1474 | net_dbg_ratelimited("udp cork app bug 2\n"); |
1da177e4 LT |
1475 | err = -EINVAL; |
1476 | goto out; | |
1477 | } | |
1478 | ||
1479 | up->pending = AF_INET6; | |
1480 | ||
1481 | do_append_data: | |
26879da5 WW |
1482 | if (ipc6.dontfrag < 0) |
1483 | ipc6.dontfrag = np->dontfrag; | |
1da177e4 | 1484 | up->len += ulen; |
26879da5 WW |
1485 | err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr), |
1486 | &ipc6, &fl6, (struct rt6_info *)dst, | |
5fdaa88d | 1487 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); |
1da177e4 LT |
1488 | if (err) |
1489 | udp_v6_flush_pending_frames(sk); | |
1490 | else if (!corkreq) | |
4c0a6cb0 | 1491 | err = udp_v6_push_pending_frames(sk); |
1e0c14f4 HX |
1492 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) |
1493 | up->pending = 0; | |
1da177e4 | 1494 | |
03485f2a VY |
1495 | if (err > 0) |
1496 | err = np->recverr ? net_xmit_errno(err) : 0; | |
1497 | release_sock(sk); | |
1498 | ||
1da177e4 | 1499 | out: |
a3c96089 | 1500 | dst_release(dst); |
1cedee13 | 1501 | out_no_dst: |
1da177e4 | 1502 | fl6_sock_release(flowlabel); |
45f6fad8 | 1503 | txopt_put(opt_to_free); |
cd562c98 | 1504 | if (!err) |
1da177e4 | 1505 | return len; |
a18135eb DM |
1506 | /* |
1507 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1508 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1509 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1510 | * things). We could add another new stat but at least for now that | |
1511 | * seems like overkill. | |
1512 | */ | |
1513 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
6aef70a8 ED |
1514 | UDP6_INC_STATS(sock_net(sk), |
1515 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
a18135eb | 1516 | } |
1da177e4 LT |
1517 | return err; |
1518 | ||
1519 | do_confirm: | |
0dec879f JA |
1520 | if (msg->msg_flags & MSG_PROBE) |
1521 | dst_confirm_neigh(dst, &fl6.daddr); | |
1da177e4 LT |
1522 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1523 | goto back_from_confirm; | |
1524 | err = 0; | |
1525 | goto out; | |
1526 | } | |
1527 | ||
7d06b2e0 | 1528 | void udpv6_destroy_sock(struct sock *sk) |
1da177e4 | 1529 | { |
44046a59 | 1530 | struct udp_sock *up = udp_sk(sk); |
1da177e4 LT |
1531 | lock_sock(sk); |
1532 | udp_v6_flush_pending_frames(sk); | |
1533 | release_sock(sk); | |
1534 | ||
60fb9567 PA |
1535 | if (static_branch_unlikely(&udpv6_encap_needed_key)) { |
1536 | if (up->encap_type) { | |
1537 | void (*encap_destroy)(struct sock *sk); | |
1538 | encap_destroy = READ_ONCE(up->encap_destroy); | |
1539 | if (encap_destroy) | |
1540 | encap_destroy(sk); | |
1541 | } | |
1542 | if (up->encap_enabled) | |
9c480601 | 1543 | static_branch_dec(&udpv6_encap_needed_key); |
44046a59 TP |
1544 | } |
1545 | ||
1da177e4 | 1546 | inet6_destroy_sock(sk); |
1da177e4 LT |
1547 | } |
1548 | ||
1549 | /* | |
1550 | * Socket option code for UDP | |
1551 | */ | |
ba4e58ec | 1552 | int udpv6_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 1553 | char __user *optval, unsigned int optlen) |
3fdadf7d | 1554 | { |
db8dac20 | 1555 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 GR |
1556 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, |
1557 | udp_v6_push_pending_frames); | |
ba4e58ec | 1558 | return ipv6_setsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1559 | } |
1560 | ||
1561 | #ifdef CONFIG_COMPAT | |
ba4e58ec | 1562 | int compat_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 compat_ipv6_setsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1569 | } |
1570 | #endif | |
1571 | ||
ba4e58ec GR |
1572 | int udpv6_getsockopt(struct sock *sk, int level, int optname, |
1573 | char __user *optval, int __user *optlen) | |
3fdadf7d | 1574 | { |
db8dac20 | 1575 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 | 1576 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); |
ba4e58ec | 1577 | return ipv6_getsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1578 | } |
1579 | ||
1580 | #ifdef CONFIG_COMPAT | |
ba4e58ec GR |
1581 | int compat_udpv6_getsockopt(struct sock *sk, int level, int optname, |
1582 | char __user *optval, int __user *optlen) | |
3fdadf7d | 1583 | { |
db8dac20 | 1584 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 | 1585 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); |
ba4e58ec | 1586 | return compat_ipv6_getsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1587 | } |
1588 | #endif | |
1589 | ||
a8e3bb34 DA |
1590 | /* thinking of making this const? Don't. |
1591 | * early_demux can change based on sysctl. | |
1592 | */ | |
dddb64bc | 1593 | static struct inet6_protocol udpv6_protocol = { |
5425077d | 1594 | .early_demux = udp_v6_early_demux, |
dddb64bc | 1595 | .early_demux_handler = udp_v6_early_demux, |
1da177e4 LT |
1596 | .handler = udpv6_rcv, |
1597 | .err_handler = udpv6_err, | |
1598 | .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, | |
1599 | }; | |
1600 | ||
1601 | /* ------------------------------------------------------------------------ */ | |
1602 | #ifdef CONFIG_PROC_FS | |
ba4e58ec | 1603 | int udp6_seq_show(struct seq_file *seq, void *v) |
1da177e4 | 1604 | { |
17ef66af LC |
1605 | if (v == SEQ_START_TOKEN) { |
1606 | seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); | |
1607 | } else { | |
1608 | int bucket = ((struct udp_iter_state *)seq->private)->bucket; | |
1609 | struct inet_sock *inet = inet_sk(v); | |
1610 | __u16 srcp = ntohs(inet->inet_sport); | |
1611 | __u16 destp = ntohs(inet->inet_dport); | |
6c206b20 PA |
1612 | __ip6_dgram_sock_seq_show(seq, v, srcp, destp, |
1613 | udp_rqueue_get(v), bucket); | |
17ef66af | 1614 | } |
1da177e4 LT |
1615 | return 0; |
1616 | } | |
1617 | ||
c3506372 | 1618 | const struct seq_operations udp6_seq_ops = { |
a3d2599b CH |
1619 | .start = udp_seq_start, |
1620 | .next = udp_seq_next, | |
1621 | .stop = udp_seq_stop, | |
1622 | .show = udp6_seq_show, | |
1623 | }; | |
c3506372 | 1624 | EXPORT_SYMBOL(udp6_seq_ops); |
73cb88ec | 1625 | |
1da177e4 | 1626 | static struct udp_seq_afinfo udp6_seq_afinfo = { |
1da177e4 | 1627 | .family = AF_INET6, |
645ca708 | 1628 | .udp_table = &udp_table, |
1da177e4 LT |
1629 | }; |
1630 | ||
2c8c1e72 | 1631 | int __net_init udp6_proc_init(struct net *net) |
1da177e4 | 1632 | { |
c3506372 CH |
1633 | if (!proc_create_net_data("udp6", 0444, net->proc_net, &udp6_seq_ops, |
1634 | sizeof(struct udp_iter_state), &udp6_seq_afinfo)) | |
a3d2599b CH |
1635 | return -ENOMEM; |
1636 | return 0; | |
1da177e4 LT |
1637 | } |
1638 | ||
ec120da6 IM |
1639 | void udp6_proc_exit(struct net *net) |
1640 | { | |
a3d2599b | 1641 | remove_proc_entry("udp6", net->proc_net); |
1da177e4 LT |
1642 | } |
1643 | #endif /* CONFIG_PROC_FS */ | |
1644 | ||
1645 | /* ------------------------------------------------------------------------ */ | |
1646 | ||
1647 | struct proto udpv6_prot = { | |
1e802951 TZ |
1648 | .name = "UDPv6", |
1649 | .owner = THIS_MODULE, | |
1650 | .close = udp_lib_close, | |
d74bad4e | 1651 | .pre_connect = udpv6_pre_connect, |
1e802951 TZ |
1652 | .connect = ip6_datagram_connect, |
1653 | .disconnect = udp_disconnect, | |
1654 | .ioctl = udp_ioctl, | |
1655 | .init = udp_init_sock, | |
1656 | .destroy = udpv6_destroy_sock, | |
1657 | .setsockopt = udpv6_setsockopt, | |
1658 | .getsockopt = udpv6_getsockopt, | |
1659 | .sendmsg = udpv6_sendmsg, | |
1660 | .recvmsg = udpv6_recvmsg, | |
1661 | .release_cb = ip6_datagram_release_cb, | |
1662 | .hash = udp_lib_hash, | |
1663 | .unhash = udp_lib_unhash, | |
1664 | .rehash = udp_v6_rehash, | |
1665 | .get_port = udp_v6_get_port, | |
1666 | .memory_allocated = &udp_memory_allocated, | |
1667 | .sysctl_mem = sysctl_udp_mem, | |
1668 | .sysctl_wmem_offset = offsetof(struct net, ipv4.sysctl_udp_wmem_min), | |
1669 | .sysctl_rmem_offset = offsetof(struct net, ipv4.sysctl_udp_rmem_min), | |
1670 | .obj_size = sizeof(struct udp6_sock), | |
1671 | .h.udp_table = &udp_table, | |
3fdadf7d | 1672 | #ifdef CONFIG_COMPAT |
1e802951 TZ |
1673 | .compat_setsockopt = compat_udpv6_setsockopt, |
1674 | .compat_getsockopt = compat_udpv6_getsockopt, | |
3fdadf7d | 1675 | #endif |
1e802951 | 1676 | .diag_destroy = udp_abort, |
1da177e4 LT |
1677 | }; |
1678 | ||
1da177e4 LT |
1679 | static struct inet_protosw udpv6_protosw = { |
1680 | .type = SOCK_DGRAM, | |
1681 | .protocol = IPPROTO_UDP, | |
1682 | .prot = &udpv6_prot, | |
1683 | .ops = &inet6_dgram_ops, | |
1da177e4 LT |
1684 | .flags = INET_PROTOSW_PERMANENT, |
1685 | }; | |
1686 | ||
7f4e4868 | 1687 | int __init udpv6_init(void) |
1da177e4 | 1688 | { |
7f4e4868 DL |
1689 | int ret; |
1690 | ||
3336288a VY |
1691 | ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP); |
1692 | if (ret) | |
c6b641a4 | 1693 | goto out; |
3336288a | 1694 | |
7f4e4868 DL |
1695 | ret = inet6_register_protosw(&udpv6_protosw); |
1696 | if (ret) | |
1697 | goto out_udpv6_protocol; | |
1698 | out: | |
1699 | return ret; | |
1700 | ||
1701 | out_udpv6_protocol: | |
1702 | inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); | |
1703 | goto out; | |
1704 | } | |
1705 | ||
09f7709f | 1706 | void udpv6_exit(void) |
7f4e4868 DL |
1707 | { |
1708 | inet6_unregister_protosw(&udpv6_protosw); | |
1709 | inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); | |
1da177e4 | 1710 | } |