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 | ||
aef2feda | 20 | #include <linux/bpf-cgroup.h> |
1da177e4 LT |
21 | #include <linux/errno.h> |
22 | #include <linux/types.h> | |
23 | #include <linux/socket.h> | |
24 | #include <linux/sockios.h> | |
1da177e4 LT |
25 | #include <linux/net.h> |
26 | #include <linux/in6.h> | |
27 | #include <linux/netdevice.h> | |
28 | #include <linux/if_arp.h> | |
29 | #include <linux/ipv6.h> | |
30 | #include <linux/icmpv6.h> | |
31 | #include <linux/init.h> | |
1781f7f5 | 32 | #include <linux/module.h> |
3305b80c | 33 | #include <linux/skbuff.h> |
5a0e3ad6 | 34 | #include <linux/slab.h> |
7c0f6ba6 | 35 | #include <linux/uaccess.h> |
0e219ae4 | 36 | #include <linux/indirect_call_wrapper.h> |
1da177e4 | 37 | |
496611d7 | 38 | #include <net/addrconf.h> |
1da177e4 LT |
39 | #include <net/ndisc.h> |
40 | #include <net/protocol.h> | |
41 | #include <net/transp_v6.h> | |
42 | #include <net/ip6_route.h> | |
1da177e4 | 43 | #include <net/raw.h> |
222a011e | 44 | #include <net/seg6.h> |
c752f073 | 45 | #include <net/tcp_states.h> |
1da177e4 | 46 | #include <net/ip6_checksum.h> |
e7cc0824 | 47 | #include <net/ip6_tunnel.h> |
1da177e4 | 48 | #include <net/xfrm.h> |
0bd84065 | 49 | #include <net/inet_hashtables.h> |
72289b96 | 50 | #include <net/inet6_hashtables.h> |
076bb0c8 | 51 | #include <net/busy_poll.h> |
e32ea7e7 | 52 | #include <net/sock_reuseport.h> |
1da177e4 LT |
53 | |
54 | #include <linux/proc_fs.h> | |
55 | #include <linux/seq_file.h> | |
22911fc5 | 56 | #include <trace/events/skb.h> |
ba4e58ec | 57 | #include "udp_impl.h" |
1da177e4 | 58 | |
6eada011 ED |
59 | static u32 udp6_ehashfn(const struct net *net, |
60 | const struct in6_addr *laddr, | |
61 | const u16 lport, | |
62 | const struct in6_addr *faddr, | |
63 | const __be16 fport) | |
b50026b5 | 64 | { |
1bbdceef HFS |
65 | static u32 udp6_ehash_secret __read_mostly; |
66 | static u32 udp_ipv6_hash_secret __read_mostly; | |
67 | ||
68 | u32 lhash, fhash; | |
69 | ||
70 | net_get_random_once(&udp6_ehash_secret, | |
71 | sizeof(udp6_ehash_secret)); | |
72 | net_get_random_once(&udp_ipv6_hash_secret, | |
73 | sizeof(udp_ipv6_hash_secret)); | |
74 | ||
75 | lhash = (__force u32)laddr->s6_addr32[3]; | |
76 | fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret); | |
77 | ||
b50026b5 | 78 | return __inet6_ehashfn(lhash, lport, fhash, fport, |
1bbdceef | 79 | udp_ipv6_hash_secret + net_hash_mix(net)); |
b50026b5 HFS |
80 | } |
81 | ||
6ba5a3c5 | 82 | int udp_v6_get_port(struct sock *sk, unsigned short snum) |
1da177e4 | 83 | { |
30fff923 | 84 | unsigned int hash2_nulladdr = |
f0b1e64c | 85 | ipv6_portaddr_hash(sock_net(sk), &in6addr_any, snum); |
9a52e97e | 86 | unsigned int hash2_partial = |
f0b1e64c | 87 | ipv6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0); |
30fff923 | 88 | |
d4cada4a | 89 | /* precompute partial secondary hash */ |
30fff923 | 90 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
fe38d2a1 | 91 | return udp_lib_get_port(sk, snum, hash2_nulladdr); |
1da177e4 LT |
92 | } |
93 | ||
f7c46156 | 94 | void udp_v6_rehash(struct sock *sk) |
719f8358 | 95 | { |
f0b1e64c | 96 | u16 new_hash = ipv6_portaddr_hash(sock_net(sk), |
efe4208f | 97 | &sk->sk_v6_rcv_saddr, |
719f8358 ED |
98 | inet_sk(sk)->inet_num); |
99 | ||
100 | udp_lib_rehash(sk, new_hash); | |
101 | } | |
102 | ||
d1e37288 SX |
103 | static int compute_score(struct sock *sk, struct net *net, |
104 | const struct in6_addr *saddr, __be16 sport, | |
105 | const struct in6_addr *daddr, unsigned short hnum, | |
73545373 | 106 | int dif, int sdif) |
645ca708 | 107 | { |
60c04aec JP |
108 | int score; |
109 | struct inet_sock *inet; | |
6da5b0f0 | 110 | bool dev_match; |
60c04aec JP |
111 | |
112 | if (!net_eq(sock_net(sk), net) || | |
113 | udp_sk(sk)->udp_port_hash != hnum || | |
114 | sk->sk_family != PF_INET6) | |
115 | return -1; | |
116 | ||
23b0269e PO |
117 | if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr)) |
118 | return -1; | |
119 | ||
60c04aec JP |
120 | score = 0; |
121 | inet = inet_sk(sk); | |
122 | ||
123 | if (inet->inet_dport) { | |
124 | if (inet->inet_dport != sport) | |
125 | return -1; | |
126 | score++; | |
127 | } | |
128 | ||
60c04aec JP |
129 | if (!ipv6_addr_any(&sk->sk_v6_daddr)) { |
130 | if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr)) | |
131 | return -1; | |
132 | score++; | |
133 | } | |
134 | ||
6da5b0f0 MM |
135 | dev_match = udp_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif); |
136 | if (!dev_match) | |
137 | return -1; | |
8d6c414c MM |
138 | if (sk->sk_bound_dev_if) |
139 | score++; | |
60c04aec | 140 | |
7170a977 | 141 | if (READ_ONCE(sk->sk_incoming_cpu) == raw_smp_processor_id()) |
70da268b ED |
142 | score++; |
143 | ||
645ca708 ED |
144 | return score; |
145 | } | |
146 | ||
a57066b1 DM |
147 | static struct sock *lookup_reuseport(struct net *net, struct sock *sk, |
148 | struct sk_buff *skb, | |
149 | const struct in6_addr *saddr, | |
150 | __be16 sport, | |
151 | const struct in6_addr *daddr, | |
152 | unsigned int hnum) | |
2a08748c JS |
153 | { |
154 | struct sock *reuse_sk = NULL; | |
155 | u32 hash; | |
156 | ||
157 | if (sk->sk_reuseport && sk->sk_state != TCP_ESTABLISHED) { | |
158 | hash = udp6_ehashfn(net, daddr, hnum, saddr, sport); | |
159 | reuse_sk = reuseport_select_sock(sk, hash, skb, | |
160 | sizeof(struct udphdr)); | |
2a08748c JS |
161 | } |
162 | return reuse_sk; | |
163 | } | |
164 | ||
d1e37288 | 165 | /* called with rcu_read_lock() */ |
fddc17de ED |
166 | static struct sock *udp6_lib_lookup2(struct net *net, |
167 | const struct in6_addr *saddr, __be16 sport, | |
1801b570 | 168 | const struct in6_addr *daddr, unsigned int hnum, |
73545373 TB |
169 | int dif, int sdif, struct udp_hslot *hslot2, |
170 | struct sk_buff *skb) | |
fddc17de ED |
171 | { |
172 | struct sock *sk, *result; | |
e94a62f5 | 173 | int score, badness; |
fddc17de | 174 | |
fddc17de ED |
175 | result = NULL; |
176 | badness = -1; | |
ca065d0c | 177 | udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
d1e37288 | 178 | score = compute_score(sk, net, saddr, sport, |
73545373 | 179 | daddr, hnum, dif, sdif); |
fddc17de | 180 | if (score > badness) { |
2a08748c JS |
181 | result = lookup_reuseport(net, sk, skb, |
182 | saddr, sport, daddr, hnum); | |
a57066b1 DM |
183 | /* Fall back to scoring if group has connections */ |
184 | if (result && !reuseport_has_conns(sk, false)) | |
2a08748c JS |
185 | return result; |
186 | ||
a57066b1 | 187 | result = result ? : sk; |
ca065d0c | 188 | badness = score; |
fddc17de ED |
189 | } |
190 | } | |
fddc17de ED |
191 | return result; |
192 | } | |
193 | ||
6d4201b1 JS |
194 | static inline struct sock *udp6_lookup_run_bpf(struct net *net, |
195 | struct udp_table *udptable, | |
196 | struct sk_buff *skb, | |
197 | const struct in6_addr *saddr, | |
198 | __be16 sport, | |
199 | const struct in6_addr *daddr, | |
f8931565 | 200 | u16 hnum, const int dif) |
6d4201b1 JS |
201 | { |
202 | struct sock *sk, *reuse_sk; | |
203 | bool no_reuseport; | |
204 | ||
205 | if (udptable != &udp_table) | |
206 | return NULL; /* only UDP is supported */ | |
207 | ||
f8931565 MP |
208 | no_reuseport = bpf_sk_lookup_run_v6(net, IPPROTO_UDP, saddr, sport, |
209 | daddr, hnum, dif, &sk); | |
6d4201b1 JS |
210 | if (no_reuseport || IS_ERR_OR_NULL(sk)) |
211 | return sk; | |
212 | ||
213 | reuse_sk = lookup_reuseport(net, sk, skb, saddr, sport, daddr, hnum); | |
c64c9c28 | 214 | if (reuse_sk) |
6d4201b1 JS |
215 | sk = reuse_sk; |
216 | return sk; | |
217 | } | |
218 | ||
ca065d0c | 219 | /* rcu_read_lock() must be held */ |
fce82338 | 220 | struct sock *__udp6_lib_lookup(struct net *net, |
1801b570 DA |
221 | const struct in6_addr *saddr, __be16 sport, |
222 | const struct in6_addr *daddr, __be16 dport, | |
223 | int dif, int sdif, struct udp_table *udptable, | |
224 | struct sk_buff *skb) | |
1da177e4 | 225 | { |
1da177e4 | 226 | unsigned short hnum = ntohs(dport); |
23b0269e PO |
227 | unsigned int hash2, slot2; |
228 | struct udp_hslot *hslot2; | |
6d4201b1 | 229 | struct sock *result, *sk; |
645ca708 | 230 | |
23b0269e PO |
231 | hash2 = ipv6_portaddr_hash(net, daddr, hnum); |
232 | slot2 = hash2 & udptable->mask; | |
233 | hslot2 = &udptable->hash2[slot2]; | |
234 | ||
6d4201b1 | 235 | /* Lookup connected or non-wildcard sockets */ |
23b0269e | 236 | result = udp6_lib_lookup2(net, saddr, sport, |
73545373 | 237 | daddr, hnum, dif, sdif, |
23b0269e | 238 | hslot2, skb); |
6d4201b1 JS |
239 | if (!IS_ERR_OR_NULL(result) && result->sk_state == TCP_ESTABLISHED) |
240 | goto done; | |
241 | ||
242 | /* Lookup redirect from BPF */ | |
243 | if (static_branch_unlikely(&bpf_sk_lookup_enabled)) { | |
244 | sk = udp6_lookup_run_bpf(net, udptable, skb, | |
f8931565 | 245 | saddr, sport, daddr, hnum, dif); |
6d4201b1 JS |
246 | if (sk) { |
247 | result = sk; | |
248 | goto done; | |
249 | } | |
250 | } | |
23b0269e | 251 | |
6d4201b1 JS |
252 | /* Got non-wildcard socket or error on first lookup */ |
253 | if (result) | |
254 | goto done; | |
fddc17de | 255 | |
6d4201b1 JS |
256 | /* Lookup wildcard sockets */ |
257 | hash2 = ipv6_portaddr_hash(net, &in6addr_any, hnum); | |
258 | slot2 = hash2 & udptable->mask; | |
259 | hslot2 = &udptable->hash2[slot2]; | |
260 | ||
261 | result = udp6_lib_lookup2(net, saddr, sport, | |
262 | &in6addr_any, hnum, dif, sdif, | |
263 | hslot2, skb); | |
264 | done: | |
26f8113c | 265 | if (IS_ERR(result)) |
23b0269e | 266 | return NULL; |
1da177e4 LT |
267 | return result; |
268 | } | |
fce82338 | 269 | EXPORT_SYMBOL_GPL(__udp6_lib_lookup); |
1da177e4 | 270 | |
607c4aaf KK |
271 | static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb, |
272 | __be16 sport, __be16 dport, | |
645ca708 | 273 | struct udp_table *udptable) |
607c4aaf | 274 | { |
b71d1d42 | 275 | const struct ipv6hdr *iph = ipv6_hdr(skb); |
607c4aaf | 276 | |
ed7cbbce | 277 | return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport, |
adf30907 | 278 | &iph->daddr, dport, inet6_iif(skb), |
1801b570 | 279 | inet6_sdif(skb), udptable, skb); |
607c4aaf KK |
280 | } |
281 | ||
7b58e63e | 282 | struct sock *udp6_lib_lookup_skb(const struct sk_buff *skb, |
63058308 TH |
283 | __be16 sport, __be16 dport) |
284 | { | |
285 | const struct ipv6hdr *iph = ipv6_hdr(skb); | |
63058308 | 286 | |
ed7cbbce | 287 | return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport, |
63058308 | 288 | &iph->daddr, dport, inet6_iif(skb), |
257a525f | 289 | inet6_sdif(skb), &udp_table, NULL); |
63058308 | 290 | } |
63058308 | 291 | |
ca065d0c ED |
292 | /* Must be called under rcu_read_lock(). |
293 | * Does increment socket refcount. | |
294 | */ | |
6e86000c | 295 | #if IS_ENABLED(CONFIG_NF_TPROXY_IPV6) || IS_ENABLED(CONFIG_NF_SOCKET_IPV6) |
aa976fc0 BS |
296 | struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport, |
297 | const struct in6_addr *daddr, __be16 dport, int dif) | |
298 | { | |
ca065d0c ED |
299 | struct sock *sk; |
300 | ||
301 | sk = __udp6_lib_lookup(net, saddr, sport, daddr, dport, | |
1801b570 | 302 | dif, 0, &udp_table, NULL); |
41c6d650 | 303 | if (sk && !refcount_inc_not_zero(&sk->sk_refcnt)) |
ca065d0c ED |
304 | sk = NULL; |
305 | return sk; | |
aa976fc0 BS |
306 | } |
307 | EXPORT_SYMBOL_GPL(udp6_lib_lookup); | |
ca065d0c | 308 | #endif |
aa976fc0 | 309 | |
cb891fa6 PA |
310 | /* do not use the scratch area len for jumbogram: their length execeeds the |
311 | * scratch area space; note that the IP6CB flags is still in the first | |
312 | * cacheline, so checking for jumbograms is cheap | |
313 | */ | |
314 | static int udp6_skb_len(struct sk_buff *skb) | |
315 | { | |
316 | return unlikely(inet6_is_jumbogram(skb)) ? skb->len : udp_skb_len(skb); | |
317 | } | |
318 | ||
1da177e4 | 319 | /* |
67ba4152 IM |
320 | * This should be easy, if there is something there we |
321 | * return it, otherwise we block. | |
1da177e4 LT |
322 | */ |
323 | ||
1b784140 | 324 | int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, |
1da177e4 LT |
325 | int noblock, int flags, int *addr_len) |
326 | { | |
327 | struct ipv6_pinfo *np = inet6_sk(sk); | |
328 | struct inet_sock *inet = inet_sk(sk); | |
1ab1457c | 329 | struct sk_buff *skb; |
59c2cdae | 330 | unsigned int ulen, copied; |
fd69c399 | 331 | int off, err, peeking = flags & MSG_PEEK; |
759e5d00 | 332 | int is_udplite = IS_UDPLITE(sk); |
543fc3fb | 333 | struct udp_mib __percpu *mib; |
197c949e | 334 | bool checksum_valid = false; |
f26ba175 | 335 | int is_udp4; |
1da177e4 | 336 | |
1da177e4 | 337 | if (flags & MSG_ERRQUEUE) |
85fbaa75 | 338 | return ipv6_recv_error(sk, msg, len, addr_len); |
1da177e4 | 339 | |
4b340ae2 | 340 | if (np->rxpmtu && np->rxopt.bits.rxpmtu) |
85fbaa75 | 341 | return ipv6_recv_rxpmtu(sk, msg, len, addr_len); |
4b340ae2 | 342 | |
1da177e4 | 343 | try_again: |
a0917e0b | 344 | off = sk_peek_offset(sk, flags); |
fd69c399 | 345 | skb = __skb_recv_udp(sk, flags, noblock, &off, &err); |
1da177e4 | 346 | if (!skb) |
627d2d6b | 347 | return err; |
1da177e4 | 348 | |
cb891fa6 | 349 | ulen = udp6_skb_len(skb); |
59c2cdae | 350 | copied = len; |
627d2d6b | 351 | if (copied > ulen - off) |
352 | copied = ulen - off; | |
59c2cdae | 353 | else if (copied < ulen) |
1ab1457c | 354 | msg->msg_flags |= MSG_TRUNC; |
1da177e4 | 355 | |
f26ba175 | 356 | is_udp4 = (skb->protocol == htons(ETH_P_IP)); |
029a3743 | 357 | mib = __UDPX_MIB(sk, is_udp4); |
f26ba175 | 358 | |
ba4e58ec | 359 | /* |
759e5d00 HX |
360 | * If checksum is needed at all, try to do it while copying the |
361 | * data. If the data is truncated, or if we only want a partial | |
362 | * coverage checksum (UDP-Lite), do it before the copy. | |
ba4e58ec | 363 | */ |
ba4e58ec | 364 | |
d21dbdfe ED |
365 | if (copied < ulen || peeking || |
366 | (is_udplite && UDP_SKB_CB(skb)->partial_cov)) { | |
67a51780 PA |
367 | checksum_valid = udp_skb_csum_unnecessary(skb) || |
368 | !__udp_lib_checksum_complete(skb); | |
197c949e | 369 | if (!checksum_valid) |
1da177e4 | 370 | goto csum_copy_err; |
ba4e58ec GR |
371 | } |
372 | ||
67a51780 PA |
373 | if (checksum_valid || udp_skb_csum_unnecessary(skb)) { |
374 | if (udp_skb_is_linear(skb)) | |
375 | err = copy_linear_skb(skb, copied, off, &msg->msg_iter); | |
376 | else | |
377 | err = skb_copy_datagram_msg(skb, off, msg, copied); | |
378 | } else { | |
627d2d6b | 379 | err = skb_copy_and_csum_datagram_msg(skb, off, msg); |
1da177e4 LT |
380 | if (err == -EINVAL) |
381 | goto csum_copy_err; | |
382 | } | |
22911fc5 | 383 | if (unlikely(err)) { |
fd69c399 | 384 | if (!peeking) { |
979402b1 | 385 | atomic_inc(&sk->sk_drops); |
029a3743 | 386 | SNMP_INC_STATS(mib, UDP_MIB_INERRORS); |
979402b1 | 387 | } |
850cbadd | 388 | kfree_skb(skb); |
627d2d6b | 389 | return err; |
22911fc5 | 390 | } |
fd69c399 | 391 | if (!peeking) |
029a3743 | 392 | SNMP_INC_STATS(mib, UDP_MIB_INDATAGRAMS); |
cb75994e | 393 | |
3b885787 | 394 | sock_recv_ts_and_drops(msg, sk, skb); |
1da177e4 LT |
395 | |
396 | /* Copy the address. */ | |
397 | if (msg->msg_name) { | |
342dfc30 | 398 | DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
1da177e4 | 399 | sin6->sin6_family = AF_INET6; |
4bedb452 | 400 | sin6->sin6_port = udp_hdr(skb)->source; |
1da177e4 | 401 | sin6->sin6_flowinfo = 0; |
1da177e4 | 402 | |
842df073 | 403 | if (is_udp4) { |
b301e82c BH |
404 | ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, |
405 | &sin6->sin6_addr); | |
842df073 HFS |
406 | sin6->sin6_scope_id = 0; |
407 | } else { | |
4e3fd7a0 | 408 | sin6->sin6_addr = ipv6_hdr(skb)->saddr; |
842df073 HFS |
409 | sin6->sin6_scope_id = |
410 | ipv6_iface_scope_id(&sin6->sin6_addr, | |
4330487a | 411 | inet6_iif(skb)); |
1da177e4 | 412 | } |
bceaa902 | 413 | *addr_len = sizeof(*sin6); |
983695fa | 414 | |
a9ed15da SF |
415 | BPF_CGROUP_RUN_PROG_UDP6_RECVMSG_LOCK(sk, |
416 | (struct sockaddr *)sin6); | |
1da177e4 | 417 | } |
4b261c75 | 418 | |
bcd1665e PA |
419 | if (udp_sk(sk)->gro_enabled) |
420 | udp_cmsg_recv(msg, sk, skb); | |
421 | ||
4b261c75 HFS |
422 | if (np->rxopt.all) |
423 | ip6_datagram_recv_common_ctl(sk, msg, skb); | |
424 | ||
f26ba175 | 425 | if (is_udp4) { |
1da177e4 | 426 | if (inet->cmsg_flags) |
ad959036 | 427 | ip_cmsg_recv_offset(msg, sk, skb, |
10df8e61 | 428 | sizeof(struct udphdr), off); |
1da177e4 LT |
429 | } else { |
430 | if (np->rxopt.all) | |
4b261c75 | 431 | ip6_datagram_recv_specific_ctl(sk, msg, skb); |
1ab1457c | 432 | } |
1da177e4 | 433 | |
59c2cdae | 434 | err = copied; |
1da177e4 | 435 | if (flags & MSG_TRUNC) |
759e5d00 | 436 | err = ulen; |
1da177e4 | 437 | |
850cbadd | 438 | skb_consume_udp(sk, skb, peeking ? -err : err); |
1da177e4 LT |
439 | return err; |
440 | ||
441 | csum_copy_err: | |
2276f58a PA |
442 | if (!__sk_queue_drop_skb(sk, &udp_sk(sk)->reader_queue, skb, flags, |
443 | udp_skb_destructor)) { | |
029a3743 PA |
444 | SNMP_INC_STATS(mib, UDP_MIB_CSUMERRORS); |
445 | SNMP_INC_STATS(mib, UDP_MIB_INERRORS); | |
0856f939 | 446 | } |
850cbadd | 447 | kfree_skb(skb); |
1da177e4 | 448 | |
beb39db5 ED |
449 | /* starting over for a new packet, but check if we need to yield */ |
450 | cond_resched(); | |
9cfaa8de | 451 | msg->msg_flags &= ~MSG_TRUNC; |
1da177e4 LT |
452 | goto try_again; |
453 | } | |
454 | ||
a36e185e SB |
455 | DEFINE_STATIC_KEY_FALSE(udpv6_encap_needed_key); |
456 | void udpv6_encap_enable(void) | |
457 | { | |
9c480601 | 458 | static_branch_inc(&udpv6_encap_needed_key); |
a36e185e SB |
459 | } |
460 | EXPORT_SYMBOL(udpv6_encap_enable); | |
461 | ||
e7cc0824 SB |
462 | /* Handler for tunnels with arbitrary destination ports: no socket lookup, go |
463 | * through error handlers in encapsulations looking for a match. | |
464 | */ | |
465 | static int __udp6_lib_err_encap_no_sk(struct sk_buff *skb, | |
466 | struct inet6_skb_parm *opt, | |
424a7cd0 | 467 | u8 type, u8 code, int offset, __be32 info) |
e7cc0824 SB |
468 | { |
469 | int i; | |
470 | ||
471 | for (i = 0; i < MAX_IPTUN_ENCAP_OPS; i++) { | |
472 | int (*handler)(struct sk_buff *skb, struct inet6_skb_parm *opt, | |
424a7cd0 PA |
473 | u8 type, u8 code, int offset, __be32 info); |
474 | const struct ip6_tnl_encap_ops *encap; | |
e7cc0824 | 475 | |
424a7cd0 PA |
476 | encap = rcu_dereference(ip6tun_encaps[i]); |
477 | if (!encap) | |
e7cc0824 | 478 | continue; |
424a7cd0 | 479 | handler = encap->err_handler; |
e7cc0824 SB |
480 | if (handler && !handler(skb, opt, type, code, offset, info)) |
481 | return 0; | |
482 | } | |
483 | ||
484 | return -ENOENT; | |
485 | } | |
486 | ||
a36e185e SB |
487 | /* Try to match ICMP errors to UDP tunnels by looking up a socket without |
488 | * reversing source and destination port: this will match tunnels that force the | |
489 | * same destination port on both endpoints (e.g. VXLAN, GENEVE). Note that | |
490 | * lwtunnels might actually break this assumption by being configured with | |
491 | * different destination ports on endpoints, in this case we won't be able to | |
492 | * trace ICMP messages back to them. | |
493 | * | |
e7cc0824 SB |
494 | * If this doesn't match any socket, probe tunnels with arbitrary destination |
495 | * ports (e.g. FoU, GUE): there, the receiving socket is useless, as the port | |
496 | * we've sent packets to won't necessarily match the local destination port. | |
497 | * | |
a36e185e SB |
498 | * Then ask the tunnel implementation to match the error against a valid |
499 | * association. | |
500 | * | |
e7cc0824 SB |
501 | * Return an error if we can't find a match, the socket if we need further |
502 | * processing, zero otherwise. | |
a36e185e SB |
503 | */ |
504 | static struct sock *__udp6_lib_err_encap(struct net *net, | |
505 | const struct ipv6hdr *hdr, int offset, | |
506 | struct udphdr *uh, | |
507 | struct udp_table *udptable, | |
9bfce73c | 508 | struct sock *sk, |
e7cc0824 SB |
509 | struct sk_buff *skb, |
510 | struct inet6_skb_parm *opt, | |
511 | u8 type, u8 code, __be32 info) | |
a36e185e | 512 | { |
9bfce73c | 513 | int (*lookup)(struct sock *sk, struct sk_buff *skb); |
a36e185e | 514 | int network_offset, transport_offset; |
9bfce73c | 515 | struct udp_sock *up; |
a36e185e | 516 | |
a36e185e SB |
517 | network_offset = skb_network_offset(skb); |
518 | transport_offset = skb_transport_offset(skb); | |
519 | ||
520 | /* Network header needs to point to the outer IPv6 header inside ICMP */ | |
521 | skb_reset_network_header(skb); | |
522 | ||
523 | /* Transport header needs to point to the UDP header */ | |
524 | skb_set_transport_header(skb, offset); | |
525 | ||
9bfce73c VF |
526 | if (sk) { |
527 | up = udp_sk(sk); | |
528 | ||
529 | lookup = READ_ONCE(up->encap_err_lookup); | |
530 | if (lookup && lookup(sk, skb)) | |
531 | sk = NULL; | |
532 | ||
533 | goto out; | |
534 | } | |
535 | ||
e7cc0824 SB |
536 | sk = __udp6_lib_lookup(net, &hdr->daddr, uh->source, |
537 | &hdr->saddr, uh->dest, | |
538 | inet6_iif(skb), 0, udptable, skb); | |
539 | if (sk) { | |
9bfce73c | 540 | up = udp_sk(sk); |
e7cc0824 SB |
541 | |
542 | lookup = READ_ONCE(up->encap_err_lookup); | |
543 | if (!lookup || lookup(sk, skb)) | |
544 | sk = NULL; | |
545 | } | |
546 | ||
9bfce73c | 547 | out: |
e7cc0824 SB |
548 | if (!sk) { |
549 | sk = ERR_PTR(__udp6_lib_err_encap_no_sk(skb, opt, type, code, | |
550 | offset, info)); | |
551 | } | |
a36e185e SB |
552 | |
553 | skb_set_transport_header(skb, transport_offset); | |
554 | skb_set_network_header(skb, network_offset); | |
e7cc0824 | 555 | |
a36e185e SB |
556 | return sk; |
557 | } | |
558 | ||
32bbd879 SB |
559 | int __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt, |
560 | u8 type, u8 code, int offset, __be32 info, | |
561 | struct udp_table *udptable) | |
1da177e4 LT |
562 | { |
563 | struct ipv6_pinfo *np; | |
b71d1d42 ED |
564 | const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data; |
565 | const struct in6_addr *saddr = &hdr->saddr; | |
222a011e | 566 | const struct in6_addr *daddr = seg6_get_daddr(skb, opt) ? : &hdr->daddr; |
67ba4152 | 567 | struct udphdr *uh = (struct udphdr *)(skb->data+offset); |
a36e185e | 568 | bool tunnel = false; |
1da177e4 | 569 | struct sock *sk; |
e0d8c1b7 | 570 | int harderr; |
1da177e4 | 571 | int err; |
7304fe46 | 572 | struct net *net = dev_net(skb->dev); |
1da177e4 | 573 | |
e32ea7e7 | 574 | sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source, |
4ac30c4b | 575 | inet6_iif(skb), inet6_sdif(skb), udptable, NULL); |
9bfce73c | 576 | |
d26796ae | 577 | if (!sk || udp_sk(sk)->encap_type) { |
a36e185e SB |
578 | /* No socket for error: try tunnels before discarding */ |
579 | if (static_branch_unlikely(&udpv6_encap_needed_key)) { | |
580 | sk = __udp6_lib_err_encap(net, hdr, offset, uh, | |
9bfce73c | 581 | udptable, sk, skb, |
e7cc0824 SB |
582 | opt, type, code, info); |
583 | if (!sk) | |
584 | return 0; | |
9bfce73c VF |
585 | } else |
586 | sk = ERR_PTR(-ENOENT); | |
a36e185e | 587 | |
e7cc0824 | 588 | if (IS_ERR(sk)) { |
a36e185e SB |
589 | __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), |
590 | ICMP6_MIB_INERRORS); | |
e7cc0824 | 591 | return PTR_ERR(sk); |
a36e185e | 592 | } |
e7cc0824 | 593 | |
a36e185e | 594 | tunnel = true; |
7304fe46 | 595 | } |
1da177e4 | 596 | |
e0d8c1b7 WW |
597 | harderr = icmpv6_err_convert(type, code, &err); |
598 | np = inet6_sk(sk); | |
599 | ||
93b36cf3 HFS |
600 | if (type == ICMPV6_PKT_TOOBIG) { |
601 | if (!ip6_sk_accept_pmtu(sk)) | |
602 | goto out; | |
81aded24 | 603 | ip6_sk_update_pmtu(skb, sk, info); |
e0d8c1b7 WW |
604 | if (np->pmtudisc != IPV6_PMTUDISC_DONT) |
605 | harderr = 1; | |
93b36cf3 | 606 | } |
1a462d18 | 607 | if (type == NDISC_REDIRECT) { |
a36e185e SB |
608 | if (tunnel) { |
609 | ip6_redirect(skb, sock_net(sk), inet6_iif(skb), | |
610 | sk->sk_mark, sk->sk_uid); | |
611 | } else { | |
612 | ip6_sk_redirect(skb, sk); | |
613 | } | |
1a462d18 DJ |
614 | goto out; |
615 | } | |
81aded24 | 616 | |
a36e185e SB |
617 | /* Tunnels don't have an application socket: don't pass errors back */ |
618 | if (tunnel) | |
619 | goto out; | |
620 | ||
e0d8c1b7 WW |
621 | if (!np->recverr) { |
622 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
623 | goto out; | |
624 | } else { | |
1da177e4 | 625 | ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1)); |
e0d8c1b7 | 626 | } |
b1faf566 | 627 | |
1da177e4 | 628 | sk->sk_err = err; |
e3ae2365 | 629 | sk_error_report(sk); |
1da177e4 | 630 | out: |
32bbd879 | 631 | return 0; |
1da177e4 LT |
632 | } |
633 | ||
a3f96c47 | 634 | static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
f7ad74fe BL |
635 | { |
636 | int rc; | |
637 | ||
efe4208f | 638 | if (!ipv6_addr_any(&sk->sk_v6_daddr)) { |
f7ad74fe | 639 | sock_rps_save_rxhash(sk, skb); |
005ec974 | 640 | sk_mark_napi_id(sk, skb); |
2c8c56e1 | 641 | sk_incoming_cpu_update(sk); |
e68b6e50 ED |
642 | } else { |
643 | sk_mark_napi_id_once(sk, skb); | |
005ec974 | 644 | } |
f7ad74fe | 645 | |
850cbadd | 646 | rc = __udp_enqueue_schedule_skb(sk, skb); |
f7ad74fe BL |
647 | if (rc < 0) { |
648 | int is_udplite = IS_UDPLITE(sk); | |
649 | ||
650 | /* Note that an ENOMEM error is charged twice */ | |
651 | if (rc == -ENOMEM) | |
e61da9e2 | 652 | UDP6_INC_STATS(sock_net(sk), |
02c22347 | 653 | UDP_MIB_RCVBUFERRORS, is_udplite); |
a3ce2b10 MD |
654 | else |
655 | UDP6_INC_STATS(sock_net(sk), | |
656 | UDP_MIB_MEMERRORS, is_udplite); | |
e61da9e2 | 657 | UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
f7ad74fe BL |
658 | kfree_skb(skb); |
659 | return -1; | |
660 | } | |
850cbadd | 661 | |
f7ad74fe BL |
662 | return 0; |
663 | } | |
664 | ||
32bbd879 SB |
665 | static __inline__ int udpv6_err(struct sk_buff *skb, |
666 | struct inet6_skb_parm *opt, u8 type, | |
667 | u8 code, int offset, __be32 info) | |
ba4e58ec | 668 | { |
32bbd879 | 669 | return __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table); |
ba4e58ec GR |
670 | } |
671 | ||
cf329aa4 | 672 | static int udpv6_queue_rcv_one_skb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 673 | { |
ba4e58ec | 674 | struct udp_sock *up = udp_sk(sk); |
b2bf1e26 | 675 | int is_udplite = IS_UDPLITE(sk); |
a18135eb | 676 | |
ba4e58ec GR |
677 | if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) |
678 | goto drop; | |
1da177e4 | 679 | |
88ab3108 | 680 | if (static_branch_unlikely(&udpv6_encap_needed_key) && up->encap_type) { |
d7f3f621 BL |
681 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
682 | ||
683 | /* | |
684 | * This is an encapsulation socket so pass the skb to | |
685 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
686 | * fall through and pass this up the UDP socket. | |
687 | * up->encap_rcv() returns the following value: | |
688 | * =0 if skb was successfully passed to the encap | |
689 | * handler or was discarded by it. | |
690 | * >0 if skb should be passed on to UDP. | |
691 | * <0 if skb should be resubmitted as proto -N | |
692 | */ | |
693 | ||
694 | /* if we're overly short, let UDP handle it */ | |
6aa7de05 | 695 | encap_rcv = READ_ONCE(up->encap_rcv); |
e5aed006 | 696 | if (encap_rcv) { |
d7f3f621 BL |
697 | int ret; |
698 | ||
0a80966b TH |
699 | /* Verify checksum before giving to encap */ |
700 | if (udp_lib_checksum_complete(skb)) | |
701 | goto csum_error; | |
702 | ||
d7f3f621 BL |
703 | ret = encap_rcv(sk, skb); |
704 | if (ret <= 0) { | |
250962e4 MD |
705 | __UDP6_INC_STATS(sock_net(sk), |
706 | UDP_MIB_INDATAGRAMS, | |
707 | is_udplite); | |
d7f3f621 BL |
708 | return -ret; |
709 | } | |
710 | } | |
711 | ||
712 | /* FALLTHROUGH -- it's a UDP Packet */ | |
713 | } | |
714 | ||
ba4e58ec GR |
715 | /* |
716 | * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c). | |
717 | */ | |
b0a42277 | 718 | if ((up->pcflag & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { |
ba4e58ec GR |
719 | |
720 | if (up->pcrlen == 0) { /* full coverage was set */ | |
ba7a46f1 JP |
721 | net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n", |
722 | UDP_SKB_CB(skb)->cscov, skb->len); | |
ba4e58ec GR |
723 | goto drop; |
724 | } | |
725 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
ba7a46f1 JP |
726 | net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n", |
727 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
ba4e58ec GR |
728 | goto drop; |
729 | } | |
1da177e4 LT |
730 | } |
731 | ||
4b943fae | 732 | prefetch(&sk->sk_rmem_alloc); |
ce25d66a ED |
733 | if (rcu_access_pointer(sk->sk_filter) && |
734 | udp_lib_checksum_complete(skb)) | |
735 | goto csum_error; | |
736 | ||
ba66bbe5 | 737 | if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr))) |
a6127697 | 738 | goto drop; |
ba4e58ec | 739 | |
e6afc8ac | 740 | udp_csum_pull_header(skb); |
cb80ef46 | 741 | |
d826eb14 | 742 | skb_dst_drop(skb); |
cb75994e | 743 | |
850cbadd | 744 | return __udpv6_queue_rcv_skb(sk, skb); |
3e215c8d | 745 | |
6a5dc9e5 | 746 | csum_error: |
02c22347 | 747 | __UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); |
ba4e58ec | 748 | drop: |
02c22347 | 749 | __UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
cb80ef46 | 750 | atomic_inc(&sk->sk_drops); |
ba4e58ec GR |
751 | kfree_skb(skb); |
752 | return -1; | |
1da177e4 LT |
753 | } |
754 | ||
cf329aa4 PA |
755 | static int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
756 | { | |
757 | struct sk_buff *next, *segs; | |
758 | int ret; | |
759 | ||
760 | if (likely(!udp_unexpected_gso(sk, skb))) | |
761 | return udpv6_queue_rcv_one_skb(sk, skb); | |
762 | ||
763 | __skb_push(skb, -skb_mac_offset(skb)); | |
764 | segs = udp_rcv_segment(sk, skb, false); | |
1a186c14 | 765 | skb_list_walk_safe(segs, skb, next) { |
cf329aa4 PA |
766 | __skb_pull(skb, skb_transport_offset(skb)); |
767 | ||
000ac44d | 768 | udp_post_segment_fix_csum(skb); |
cf329aa4 PA |
769 | ret = udpv6_queue_rcv_one_skb(sk, skb); |
770 | if (ret > 0) | |
771 | ip6_protocol_deliver_rcu(dev_net(skb->dev), skb, ret, | |
772 | true); | |
773 | } | |
774 | return 0; | |
775 | } | |
776 | ||
5cf3d461 DH |
777 | static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk, |
778 | __be16 loc_port, const struct in6_addr *loc_addr, | |
779 | __be16 rmt_port, const struct in6_addr *rmt_addr, | |
7bd2db40 | 780 | int dif, int sdif, unsigned short hnum) |
1da177e4 | 781 | { |
5cf3d461 | 782 | struct inet_sock *inet = inet_sk(sk); |
1da177e4 | 783 | |
5cf3d461 DH |
784 | if (!net_eq(sock_net(sk), net)) |
785 | return false; | |
786 | ||
787 | if (udp_sk(sk)->udp_port_hash != hnum || | |
788 | sk->sk_family != PF_INET6 || | |
789 | (inet->inet_dport && inet->inet_dport != rmt_port) || | |
790 | (!ipv6_addr_any(&sk->sk_v6_daddr) && | |
791 | !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) || | |
7bd2db40 | 792 | !udp_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif) || |
33b4b015 HR |
793 | (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) && |
794 | !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))) | |
5cf3d461 DH |
795 | return false; |
796 | if (!inet6_mc_check(sk, loc_addr, rmt_addr)) | |
797 | return false; | |
798 | return true; | |
1da177e4 LT |
799 | } |
800 | ||
4068579e TH |
801 | static void udp6_csum_zero_error(struct sk_buff *skb) |
802 | { | |
803 | /* RFC 2460 section 8.1 says that we SHOULD log | |
804 | * this error. Well, it is reasonable. | |
805 | */ | |
ba7a46f1 JP |
806 | net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n", |
807 | &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source), | |
808 | &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest)); | |
4068579e TH |
809 | } |
810 | ||
1da177e4 LT |
811 | /* |
812 | * Note: called only from the BH handler context, | |
813 | * so we don't need to lock the hashes. | |
814 | */ | |
e3163493 | 815 | static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
b71d1d42 | 816 | const struct in6_addr *saddr, const struct in6_addr *daddr, |
36cbb245 | 817 | struct udp_table *udptable, int proto) |
1da177e4 | 818 | { |
ca065d0c | 819 | struct sock *sk, *first = NULL; |
4bedb452 | 820 | const struct udphdr *uh = udp_hdr(skb); |
5cf3d461 DH |
821 | unsigned short hnum = ntohs(uh->dest); |
822 | struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum); | |
ca065d0c | 823 | unsigned int offset = offsetof(typeof(*sk), sk_node); |
2dc41cff | 824 | unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10); |
ca065d0c | 825 | int dif = inet6_iif(skb); |
7bd2db40 | 826 | int sdif = inet6_sdif(skb); |
ca065d0c ED |
827 | struct hlist_node *node; |
828 | struct sk_buff *nskb; | |
2dc41cff DH |
829 | |
830 | if (use_hash2) { | |
f0b1e64c | 831 | hash2_any = ipv6_portaddr_hash(net, &in6addr_any, hnum) & |
73e2d5e3 | 832 | udptable->mask; |
f0b1e64c | 833 | hash2 = ipv6_portaddr_hash(net, daddr, hnum) & udptable->mask; |
2dc41cff | 834 | start_lookup: |
73e2d5e3 | 835 | hslot = &udptable->hash2[hash2]; |
2dc41cff DH |
836 | offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node); |
837 | } | |
1da177e4 | 838 | |
ca065d0c ED |
839 | sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) { |
840 | if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr, | |
7bd2db40 DM |
841 | uh->source, saddr, dif, sdif, |
842 | hnum)) | |
ca065d0c ED |
843 | continue; |
844 | /* If zero checksum and no_check is not on for | |
845 | * the socket then skip it. | |
846 | */ | |
847 | if (!uh->check && !udp_sk(sk)->no_check6_rx) | |
848 | continue; | |
849 | if (!first) { | |
850 | first = sk; | |
851 | continue; | |
852 | } | |
853 | nskb = skb_clone(skb, GFP_ATOMIC); | |
854 | if (unlikely(!nskb)) { | |
855 | atomic_inc(&sk->sk_drops); | |
02c22347 ED |
856 | __UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS, |
857 | IS_UDPLITE(sk)); | |
858 | __UDP6_INC_STATS(net, UDP_MIB_INERRORS, | |
859 | IS_UDPLITE(sk)); | |
ca065d0c | 860 | continue; |
95766fff | 861 | } |
a1ab77f9 | 862 | |
ca065d0c ED |
863 | if (udpv6_queue_rcv_skb(sk, nskb) > 0) |
864 | consume_skb(nskb); | |
865 | } | |
a1ab77f9 | 866 | |
2dc41cff DH |
867 | /* Also lookup *:port if we are using hash2 and haven't done so yet. */ |
868 | if (use_hash2 && hash2 != hash2_any) { | |
869 | hash2 = hash2_any; | |
870 | goto start_lookup; | |
871 | } | |
872 | ||
ca065d0c ED |
873 | if (first) { |
874 | if (udpv6_queue_rcv_skb(first, skb) > 0) | |
875 | consume_skb(skb); | |
a1ab77f9 | 876 | } else { |
ca065d0c | 877 | kfree_skb(skb); |
02c22347 ED |
878 | __UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI, |
879 | proto == IPPROTO_UDPLITE); | |
a1ab77f9 | 880 | } |
ba4e58ec | 881 | return 0; |
1da177e4 LT |
882 | } |
883 | ||
64f0f5d1 PA |
884 | static void udp6_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst) |
885 | { | |
886 | if (udp_sk_rx_dst_set(sk, dst)) { | |
887 | const struct rt6_info *rt = (const struct rt6_info *)dst; | |
888 | ||
ef57c161 | 889 | sk->sk_rx_dst_cookie = rt6_get_cookie(rt); |
64f0f5d1 PA |
890 | } |
891 | } | |
892 | ||
eb63f296 PA |
893 | /* wrapper for udp_queue_rcv_skb tacking care of csum conversion and |
894 | * return code conversion for ip layer consumption | |
895 | */ | |
896 | static int udp6_unicast_rcv_skb(struct sock *sk, struct sk_buff *skb, | |
897 | struct udphdr *uh) | |
898 | { | |
899 | int ret; | |
900 | ||
901 | if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk)) | |
e4aa33ad | 902 | skb_checksum_try_convert(skb, IPPROTO_UDP, ip6_compute_pseudo); |
eb63f296 PA |
903 | |
904 | ret = udpv6_queue_rcv_skb(sk, skb); | |
905 | ||
84dad559 | 906 | /* a return value > 0 means to resubmit the input */ |
eb63f296 | 907 | if (ret > 0) |
84dad559 | 908 | return ret; |
eb63f296 PA |
909 | return 0; |
910 | } | |
911 | ||
645ca708 | 912 | int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
759e5d00 | 913 | int proto) |
1da177e4 | 914 | { |
ca065d0c | 915 | const struct in6_addr *saddr, *daddr; |
3ffe533c | 916 | struct net *net = dev_net(skb->dev); |
1ab1457c | 917 | struct udphdr *uh; |
ca065d0c | 918 | struct sock *sk; |
71489e21 | 919 | bool refcounted; |
1da177e4 LT |
920 | u32 ulen = 0; |
921 | ||
922 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
d6bc0149 | 923 | goto discard; |
1da177e4 | 924 | |
0660e03f ACM |
925 | saddr = &ipv6_hdr(skb)->saddr; |
926 | daddr = &ipv6_hdr(skb)->daddr; | |
4bedb452 | 927 | uh = udp_hdr(skb); |
1da177e4 LT |
928 | |
929 | ulen = ntohs(uh->len); | |
ba4e58ec GR |
930 | if (ulen > skb->len) |
931 | goto short_packet; | |
1da177e4 | 932 | |
759e5d00 HX |
933 | if (proto == IPPROTO_UDP) { |
934 | /* UDP validates ulen. */ | |
1da177e4 | 935 | |
ba4e58ec GR |
936 | /* Check for jumbo payload */ |
937 | if (ulen == 0) | |
938 | ulen = skb->len; | |
1da177e4 | 939 | |
ba4e58ec GR |
940 | if (ulen < sizeof(*uh)) |
941 | goto short_packet; | |
1da177e4 | 942 | |
ba4e58ec GR |
943 | if (ulen < skb->len) { |
944 | if (pskb_trim_rcsum(skb, ulen)) | |
945 | goto short_packet; | |
0660e03f ACM |
946 | saddr = &ipv6_hdr(skb)->saddr; |
947 | daddr = &ipv6_hdr(skb)->daddr; | |
4bedb452 | 948 | uh = udp_hdr(skb); |
ba4e58ec | 949 | } |
ba4e58ec | 950 | } |
1da177e4 | 951 | |
759e5d00 | 952 | if (udp6_csum_init(skb, uh, proto)) |
6a5dc9e5 | 953 | goto csum_error; |
759e5d00 | 954 | |
c9f2c1ae | 955 | /* Check if the socket is already available, e.g. due to early demux */ |
71489e21 | 956 | sk = skb_steal_sock(skb, &refcounted); |
c9f2c1ae PA |
957 | if (sk) { |
958 | struct dst_entry *dst = skb_dst(skb); | |
959 | int ret; | |
960 | ||
8f905c0e | 961 | if (unlikely(rcu_dereference(sk->sk_rx_dst) != dst)) |
64f0f5d1 | 962 | udp6_sk_rx_dst_set(sk, dst); |
c9f2c1ae | 963 | |
eb63f296 | 964 | if (!uh->check && !udp_sk(sk)->no_check6_rx) { |
71489e21 JS |
965 | if (refcounted) |
966 | sock_put(sk); | |
eb63f296 PA |
967 | goto report_csum_error; |
968 | } | |
c9f2c1ae | 969 | |
eb63f296 | 970 | ret = udp6_unicast_rcv_skb(sk, skb, uh); |
71489e21 JS |
971 | if (refcounted) |
972 | sock_put(sk); | |
eb63f296 | 973 | return ret; |
c9f2c1ae PA |
974 | } |
975 | ||
1ab1457c YH |
976 | /* |
977 | * Multicast receive code | |
1da177e4 | 978 | */ |
ba4e58ec | 979 | if (ipv6_addr_is_multicast(daddr)) |
e3163493 | 980 | return __udp6_lib_mcast_deliver(net, skb, |
36cbb245 | 981 | saddr, daddr, udptable, proto); |
1da177e4 LT |
982 | |
983 | /* Unicast */ | |
607c4aaf | 984 | sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable); |
53b24b8f | 985 | if (sk) { |
eb63f296 PA |
986 | if (!uh->check && !udp_sk(sk)->no_check6_rx) |
987 | goto report_csum_error; | |
988 | return udp6_unicast_rcv_skb(sk, skb, uh); | |
1da177e4 | 989 | } |
1ab1457c | 990 | |
eb63f296 PA |
991 | if (!uh->check) |
992 | goto report_csum_error; | |
4068579e | 993 | |
cb80ef46 | 994 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) |
c377411f | 995 | goto discard; |
cb80ef46 BL |
996 | |
997 | if (udp_lib_checksum_complete(skb)) | |
6a5dc9e5 | 998 | goto csum_error; |
cb80ef46 | 999 | |
02c22347 | 1000 | __UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
cb80ef46 BL |
1001 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
1002 | ||
1003 | kfree_skb(skb); | |
add459aa | 1004 | return 0; |
1da177e4 | 1005 | |
1ab1457c | 1006 | short_packet: |
ba7a46f1 JP |
1007 | net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n", |
1008 | proto == IPPROTO_UDPLITE ? "-Lite" : "", | |
1009 | saddr, ntohs(uh->source), | |
1010 | ulen, skb->len, | |
1011 | daddr, ntohs(uh->dest)); | |
6a5dc9e5 | 1012 | goto discard; |
eb63f296 PA |
1013 | |
1014 | report_csum_error: | |
1015 | udp6_csum_zero_error(skb); | |
6a5dc9e5 | 1016 | csum_error: |
02c22347 | 1017 | __UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); |
1da177e4 | 1018 | discard: |
02c22347 | 1019 | __UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
1da177e4 | 1020 | kfree_skb(skb); |
add459aa | 1021 | return 0; |
1da177e4 | 1022 | } |
ba4e58ec | 1023 | |
0bd84065 | 1024 | |
5425077d | 1025 | static struct sock *__udp6_lib_demux_lookup(struct net *net, |
1026 | __be16 loc_port, const struct in6_addr *loc_addr, | |
1027 | __be16 rmt_port, const struct in6_addr *rmt_addr, | |
4297a0ef | 1028 | int dif, int sdif) |
5425077d | 1029 | { |
0bd84065 | 1030 | unsigned short hnum = ntohs(loc_port); |
f0b1e64c | 1031 | unsigned int hash2 = ipv6_portaddr_hash(net, loc_addr, hnum); |
0bd84065 | 1032 | unsigned int slot2 = hash2 & udp_table.mask; |
1033 | struct udp_hslot *hslot2 = &udp_table.hash2[slot2]; | |
1034 | const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum); | |
5425077d | 1035 | struct sock *sk; |
1036 | ||
0bd84065 | 1037 | udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
85cb73ff | 1038 | if (sk->sk_state == TCP_ESTABLISHED && |
4297a0ef | 1039 | INET6_MATCH(sk, net, rmt_addr, loc_addr, ports, dif, sdif)) |
0bd84065 | 1040 | return sk; |
1041 | /* Only check first socket in chain */ | |
1042 | break; | |
1043 | } | |
1044 | return NULL; | |
5425077d | 1045 | } |
1046 | ||
97ff7ffb | 1047 | INDIRECT_CALLABLE_SCOPE void udp_v6_early_demux(struct sk_buff *skb) |
5425077d | 1048 | { |
1049 | struct net *net = dev_net(skb->dev); | |
1050 | const struct udphdr *uh; | |
1051 | struct sock *sk; | |
1052 | struct dst_entry *dst; | |
1053 | int dif = skb->dev->ifindex; | |
4297a0ef | 1054 | int sdif = inet6_sdif(skb); |
5425077d | 1055 | |
1056 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + | |
1057 | sizeof(struct udphdr))) | |
1058 | return; | |
1059 | ||
1060 | uh = udp_hdr(skb); | |
1061 | ||
1062 | if (skb->pkt_type == PACKET_HOST) | |
1063 | sk = __udp6_lib_demux_lookup(net, uh->dest, | |
1064 | &ipv6_hdr(skb)->daddr, | |
1065 | uh->source, &ipv6_hdr(skb)->saddr, | |
4297a0ef | 1066 | dif, sdif); |
5425077d | 1067 | else |
1068 | return; | |
1069 | ||
41c6d650 | 1070 | if (!sk || !refcount_inc_not_zero(&sk->sk_refcnt)) |
5425077d | 1071 | return; |
1072 | ||
1073 | skb->sk = sk; | |
1074 | skb->destructor = sock_efree; | |
8f905c0e | 1075 | dst = rcu_dereference(sk->sk_rx_dst); |
5425077d | 1076 | |
1077 | if (dst) | |
ef57c161 | 1078 | dst = dst_check(dst, sk->sk_rx_dst_cookie); |
5425077d | 1079 | if (dst) { |
d24406c8 WW |
1080 | /* set noref for now. |
1081 | * any place which wants to hold dst has to call | |
1082 | * dst_hold_safe() | |
1083 | */ | |
1084 | skb_dst_set_noref(skb, dst); | |
5425077d | 1085 | } |
1086 | } | |
1087 | ||
0e219ae4 | 1088 | INDIRECT_CALLABLE_SCOPE int udpv6_rcv(struct sk_buff *skb) |
ba4e58ec | 1089 | { |
645ca708 | 1090 | return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
ba4e58ec GR |
1091 | } |
1092 | ||
1da177e4 LT |
1093 | /* |
1094 | * Throw away all pending data and cancel the corking. Socket is locked. | |
1095 | */ | |
1096 | static void udp_v6_flush_pending_frames(struct sock *sk) | |
1097 | { | |
1098 | struct udp_sock *up = udp_sk(sk); | |
1099 | ||
36d926b9 DL |
1100 | if (up->pending == AF_INET) |
1101 | udp_flush_pending_frames(sk); | |
1102 | else if (up->pending) { | |
1da177e4 LT |
1103 | up->len = 0; |
1104 | up->pending = 0; | |
1105 | ip6_flush_pending_frames(sk); | |
1ab1457c | 1106 | } |
1da177e4 LT |
1107 | } |
1108 | ||
d74bad4e AI |
1109 | static int udpv6_pre_connect(struct sock *sk, struct sockaddr *uaddr, |
1110 | int addr_len) | |
1111 | { | |
bddc028a TH |
1112 | if (addr_len < offsetofend(struct sockaddr, sa_family)) |
1113 | return -EINVAL; | |
d74bad4e AI |
1114 | /* The following checks are replicated from __ip6_datagram_connect() |
1115 | * and intended to prevent BPF program called below from accessing | |
1116 | * bytes that are out of the bound specified by user in addr_len. | |
1117 | */ | |
1118 | if (uaddr->sa_family == AF_INET) { | |
1119 | if (__ipv6_only_sock(sk)) | |
1120 | return -EAFNOSUPPORT; | |
1121 | return udp_pre_connect(sk, uaddr, addr_len); | |
1122 | } | |
1123 | ||
1124 | if (addr_len < SIN6_LEN_RFC2133) | |
1125 | return -EINVAL; | |
1126 | ||
1127 | return BPF_CGROUP_RUN_PROG_INET6_CONNECT_LOCK(sk, uaddr); | |
1128 | } | |
1129 | ||
493c6be3 | 1130 | /** |
67ba4152 IM |
1131 | * udp6_hwcsum_outgoing - handle outgoing HW checksumming |
1132 | * @sk: socket we are sending on | |
1133 | * @skb: sk_buff containing the filled-in UDP header | |
1134 | * (checksum field must be zeroed out) | |
b51cd7c8 AL |
1135 | * @saddr: source address |
1136 | * @daddr: destination address | |
1137 | * @len: length of packet | |
493c6be3 SS |
1138 | */ |
1139 | static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb, | |
1140 | const struct in6_addr *saddr, | |
1141 | const struct in6_addr *daddr, int len) | |
1142 | { | |
1143 | unsigned int offset; | |
1144 | struct udphdr *uh = udp_hdr(skb); | |
d39d938c | 1145 | struct sk_buff *frags = skb_shinfo(skb)->frag_list; |
493c6be3 SS |
1146 | __wsum csum = 0; |
1147 | ||
d39d938c | 1148 | if (!frags) { |
493c6be3 SS |
1149 | /* Only one fragment on the socket. */ |
1150 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
1151 | skb->csum_offset = offsetof(struct udphdr, check); | |
1152 | uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0); | |
1153 | } else { | |
1154 | /* | |
1155 | * HW-checksum won't work as there are two or more | |
1156 | * fragments on the socket so that all csums of sk_buffs | |
1157 | * should be together | |
1158 | */ | |
1159 | offset = skb_transport_offset(skb); | |
1160 | skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); | |
63ecc3d9 | 1161 | csum = skb->csum; |
493c6be3 SS |
1162 | |
1163 | skb->ip_summed = CHECKSUM_NONE; | |
1164 | ||
d39d938c VY |
1165 | do { |
1166 | csum = csum_add(csum, frags->csum); | |
1167 | } while ((frags = frags->next)); | |
493c6be3 SS |
1168 | |
1169 | uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, | |
1170 | csum); | |
1171 | if (uh->check == 0) | |
1172 | uh->check = CSUM_MANGLED_0; | |
1173 | } | |
1174 | } | |
1175 | ||
1da177e4 LT |
1176 | /* |
1177 | * Sending | |
1178 | */ | |
1179 | ||
bec1f6f6 WB |
1180 | static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6, |
1181 | struct inet_cork *cork) | |
1da177e4 | 1182 | { |
d39d938c | 1183 | struct sock *sk = skb->sk; |
1da177e4 | 1184 | struct udphdr *uh; |
1da177e4 | 1185 | int err = 0; |
b2bf1e26 | 1186 | int is_udplite = IS_UDPLITE(sk); |
868c86bc | 1187 | __wsum csum = 0; |
d39d938c VY |
1188 | int offset = skb_transport_offset(skb); |
1189 | int len = skb->len - offset; | |
4094871d | 1190 | int datalen = len - sizeof(*uh); |
1da177e4 LT |
1191 | |
1192 | /* | |
1193 | * Create a UDP header | |
1194 | */ | |
4bedb452 | 1195 | uh = udp_hdr(skb); |
1958b856 DM |
1196 | uh->source = fl6->fl6_sport; |
1197 | uh->dest = fl6->fl6_dport; | |
d39d938c | 1198 | uh->len = htons(len); |
1da177e4 LT |
1199 | uh->check = 0; |
1200 | ||
bec1f6f6 WB |
1201 | if (cork->gso_size) { |
1202 | const int hlen = skb_network_header_len(skb) + | |
1203 | sizeof(struct udphdr); | |
1204 | ||
0f149c9f WB |
1205 | if (hlen + cork->gso_size > cork->fragsize) { |
1206 | kfree_skb(skb); | |
bec1f6f6 | 1207 | return -EINVAL; |
0f149c9f | 1208 | } |
736ef37f | 1209 | if (datalen > cork->gso_size * UDP_MAX_SEGMENTS) { |
0f149c9f | 1210 | kfree_skb(skb); |
bec1f6f6 | 1211 | return -EINVAL; |
0f149c9f WB |
1212 | } |
1213 | if (udp_sk(sk)->no_check6_tx) { | |
1214 | kfree_skb(skb); | |
a8c744a8 | 1215 | return -EINVAL; |
0f149c9f | 1216 | } |
ff06342c | 1217 | if (skb->ip_summed != CHECKSUM_PARTIAL || is_udplite || |
0f149c9f WB |
1218 | dst_xfrm(skb_dst(skb))) { |
1219 | kfree_skb(skb); | |
bec1f6f6 | 1220 | return -EIO; |
0f149c9f | 1221 | } |
bec1f6f6 | 1222 | |
4094871d JH |
1223 | if (datalen > cork->gso_size) { |
1224 | skb_shinfo(skb)->gso_size = cork->gso_size; | |
1225 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP_L4; | |
1226 | skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(datalen, | |
1227 | cork->gso_size); | |
1228 | } | |
a8c744a8 | 1229 | goto csum_partial; |
bec1f6f6 WB |
1230 | } |
1231 | ||
b2bf1e26 | 1232 | if (is_udplite) |
d39d938c VY |
1233 | csum = udplite_csum(skb); |
1234 | else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */ | |
4068579e TH |
1235 | skb->ip_summed = CHECKSUM_NONE; |
1236 | goto send; | |
1237 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
a8c744a8 | 1238 | csum_partial: |
d39d938c | 1239 | udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len); |
493c6be3 SS |
1240 | goto send; |
1241 | } else | |
d39d938c | 1242 | csum = udp_csum(skb); |
1da177e4 | 1243 | |
ba4e58ec | 1244 | /* add protocol-dependent pseudo-header */ |
4c9483b2 | 1245 | uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, |
d39d938c | 1246 | len, fl6->flowi6_proto, csum); |
1da177e4 | 1247 | if (uh->check == 0) |
f6ab0288 | 1248 | uh->check = CSUM_MANGLED_0; |
1da177e4 | 1249 | |
493c6be3 | 1250 | send: |
d39d938c | 1251 | err = ip6_send_skb(skb); |
6ce9e7b5 ED |
1252 | if (err) { |
1253 | if (err == -ENOBUFS && !inet6_sk(sk)->recverr) { | |
6aef70a8 ED |
1254 | UDP6_INC_STATS(sock_net(sk), |
1255 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
6ce9e7b5 ED |
1256 | err = 0; |
1257 | } | |
6aef70a8 ED |
1258 | } else { |
1259 | UDP6_INC_STATS(sock_net(sk), | |
1260 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
1261 | } | |
d39d938c VY |
1262 | return err; |
1263 | } | |
1264 | ||
1265 | static int udp_v6_push_pending_frames(struct sock *sk) | |
1266 | { | |
1267 | struct sk_buff *skb; | |
1268 | struct udp_sock *up = udp_sk(sk); | |
d39d938c VY |
1269 | int err = 0; |
1270 | ||
1271 | if (up->pending == AF_INET) | |
1272 | return udp_push_pending_frames(sk); | |
1273 | ||
d39d938c VY |
1274 | skb = ip6_finish_skb(sk); |
1275 | if (!skb) | |
1276 | goto out; | |
1277 | ||
940ea00b PB |
1278 | err = udp_v6_send_skb(skb, &inet_sk(sk)->cork.fl.u.ip6, |
1279 | &inet_sk(sk)->cork.base); | |
1da177e4 LT |
1280 | out: |
1281 | up->len = 0; | |
1282 | up->pending = 0; | |
1283 | return err; | |
1284 | } | |
1285 | ||
1b784140 | 1286 | int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
1da177e4 LT |
1287 | { |
1288 | struct ipv6_txoptions opt_space; | |
1289 | struct udp_sock *up = udp_sk(sk); | |
1290 | struct inet_sock *inet = inet_sk(sk); | |
1291 | struct ipv6_pinfo *np = inet6_sk(sk); | |
342dfc30 | 1292 | DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
20c59de2 | 1293 | struct in6_addr *daddr, *final_p, final; |
1da177e4 | 1294 | struct ipv6_txoptions *opt = NULL; |
45f6fad8 | 1295 | struct ipv6_txoptions *opt_to_free = NULL; |
1da177e4 | 1296 | struct ip6_flowlabel *flowlabel = NULL; |
4c9483b2 | 1297 | struct flowi6 fl6; |
1da177e4 | 1298 | struct dst_entry *dst; |
26879da5 | 1299 | struct ipcm6_cookie ipc6; |
1da177e4 | 1300 | int addr_len = msg->msg_namelen; |
9f542f61 | 1301 | bool connected = false; |
1da177e4 | 1302 | int ulen = len; |
a9f59707 | 1303 | int corkreq = READ_ONCE(up->corkflag) || msg->msg_flags&MSG_MORE; |
1da177e4 | 1304 | int err; |
b2bf1e26 | 1305 | int is_udplite = IS_UDPLITE(sk); |
ba4e58ec | 1306 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); |
1da177e4 | 1307 | |
b515430a | 1308 | ipcm6_init(&ipc6); |
18a419ba | 1309 | ipc6.gso_size = READ_ONCE(up->gso_size); |
5fdaa88d | 1310 | ipc6.sockc.tsflags = sk->sk_tsflags; |
c6af0c22 | 1311 | ipc6.sockc.mark = sk->sk_mark; |
26879da5 | 1312 | |
1da177e4 LT |
1313 | /* destination address check */ |
1314 | if (sin6) { | |
1315 | if (addr_len < offsetof(struct sockaddr, sa_data)) | |
1316 | return -EINVAL; | |
1317 | ||
1318 | switch (sin6->sin6_family) { | |
1319 | case AF_INET6: | |
1320 | if (addr_len < SIN6_LEN_RFC2133) | |
1321 | return -EINVAL; | |
1322 | daddr = &sin6->sin6_addr; | |
052d2369 JL |
1323 | if (ipv6_addr_any(daddr) && |
1324 | ipv6_addr_v4mapped(&np->saddr)) | |
1325 | ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK), | |
1326 | daddr); | |
1da177e4 LT |
1327 | break; |
1328 | case AF_INET: | |
1329 | goto do_udp_sendmsg; | |
1330 | case AF_UNSPEC: | |
1331 | msg->msg_name = sin6 = NULL; | |
1332 | msg->msg_namelen = addr_len = 0; | |
1333 | daddr = NULL; | |
1334 | break; | |
1335 | default: | |
1336 | return -EINVAL; | |
1337 | } | |
1338 | } else if (!up->pending) { | |
1339 | if (sk->sk_state != TCP_ESTABLISHED) | |
1340 | return -EDESTADDRREQ; | |
efe4208f | 1341 | daddr = &sk->sk_v6_daddr; |
1ab1457c | 1342 | } else |
1da177e4 LT |
1343 | daddr = NULL; |
1344 | ||
1345 | if (daddr) { | |
e773e4fa | 1346 | if (ipv6_addr_v4mapped(daddr)) { |
1da177e4 LT |
1347 | struct sockaddr_in sin; |
1348 | sin.sin_family = AF_INET; | |
c720c7e8 | 1349 | sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport; |
1da177e4 LT |
1350 | sin.sin_addr.s_addr = daddr->s6_addr32[3]; |
1351 | msg->msg_name = &sin; | |
1352 | msg->msg_namelen = sizeof(sin); | |
1353 | do_udp_sendmsg: | |
1354 | if (__ipv6_only_sock(sk)) | |
1355 | return -ENETUNREACH; | |
1b784140 | 1356 | return udp_sendmsg(sk, msg, len); |
1da177e4 LT |
1357 | } |
1358 | } | |
1359 | ||
1da177e4 | 1360 | /* Rough check on arithmetic overflow, |
b59e139b | 1361 | better check is made in ip6_append_data(). |
1da177e4 LT |
1362 | */ |
1363 | if (len > INT_MAX - sizeof(struct udphdr)) | |
1364 | return -EMSGSIZE; | |
1ab1457c | 1365 | |
03485f2a | 1366 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
1da177e4 | 1367 | if (up->pending) { |
406c4a0a PB |
1368 | if (up->pending == AF_INET) |
1369 | return udp_sendmsg(sk, msg, len); | |
1da177e4 LT |
1370 | /* |
1371 | * There are pending frames. | |
1372 | * The socket lock must be held while it's corked. | |
1373 | */ | |
1374 | lock_sock(sk); | |
1375 | if (likely(up->pending)) { | |
1376 | if (unlikely(up->pending != AF_INET6)) { | |
1377 | release_sock(sk); | |
1378 | return -EAFNOSUPPORT; | |
1379 | } | |
1380 | dst = NULL; | |
1381 | goto do_append_data; | |
1382 | } | |
1383 | release_sock(sk); | |
1384 | } | |
1385 | ulen += sizeof(struct udphdr); | |
1386 | ||
4c9483b2 | 1387 | memset(&fl6, 0, sizeof(fl6)); |
1da177e4 LT |
1388 | |
1389 | if (sin6) { | |
1390 | if (sin6->sin6_port == 0) | |
1391 | return -EINVAL; | |
1392 | ||
1958b856 | 1393 | fl6.fl6_dport = sin6->sin6_port; |
1da177e4 LT |
1394 | daddr = &sin6->sin6_addr; |
1395 | ||
1396 | if (np->sndflow) { | |
4c9483b2 DM |
1397 | fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; |
1398 | if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { | |
1399 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); | |
59c820b2 | 1400 | if (IS_ERR(flowlabel)) |
1da177e4 | 1401 | return -EINVAL; |
1da177e4 LT |
1402 | } |
1403 | } | |
1404 | ||
1405 | /* | |
1406 | * Otherwise it will be difficult to maintain | |
1407 | * sk->sk_dst_cache. | |
1408 | */ | |
1409 | if (sk->sk_state == TCP_ESTABLISHED && | |
efe4208f ED |
1410 | ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) |
1411 | daddr = &sk->sk_v6_daddr; | |
1da177e4 LT |
1412 | |
1413 | if (addr_len >= sizeof(struct sockaddr_in6) && | |
1414 | sin6->sin6_scope_id && | |
842df073 | 1415 | __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) |
4c9483b2 | 1416 | fl6.flowi6_oif = sin6->sin6_scope_id; |
1da177e4 LT |
1417 | } else { |
1418 | if (sk->sk_state != TCP_ESTABLISHED) | |
1419 | return -EDESTADDRREQ; | |
1420 | ||
1958b856 | 1421 | fl6.fl6_dport = inet->inet_dport; |
efe4208f | 1422 | daddr = &sk->sk_v6_daddr; |
4c9483b2 | 1423 | fl6.flowlabel = np->flow_label; |
9f542f61 | 1424 | connected = true; |
1da177e4 LT |
1425 | } |
1426 | ||
4c9483b2 DM |
1427 | if (!fl6.flowi6_oif) |
1428 | fl6.flowi6_oif = sk->sk_bound_dev_if; | |
1da177e4 | 1429 | |
4c9483b2 DM |
1430 | if (!fl6.flowi6_oif) |
1431 | fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex; | |
9f690db7 | 1432 | |
e2d118a1 | 1433 | fl6.flowi6_uid = sk->sk_uid; |
51953d5b | 1434 | |
1da177e4 LT |
1435 | if (msg->msg_controllen) { |
1436 | opt = &opt_space; | |
1437 | memset(opt, 0, sizeof(struct ipv6_txoptions)); | |
1438 | opt->tot_len = sizeof(*opt); | |
26879da5 | 1439 | ipc6.opt = opt; |
1da177e4 | 1440 | |
2e8de857 WB |
1441 | err = udp_cmsg_send(sk, msg, &ipc6.gso_size); |
1442 | if (err > 0) | |
1443 | err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, | |
5fdaa88d | 1444 | &ipc6); |
1da177e4 LT |
1445 | if (err < 0) { |
1446 | fl6_sock_release(flowlabel); | |
1447 | return err; | |
1448 | } | |
4c9483b2 DM |
1449 | if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { |
1450 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); | |
59c820b2 | 1451 | if (IS_ERR(flowlabel)) |
1da177e4 LT |
1452 | return -EINVAL; |
1453 | } | |
1454 | if (!(opt->opt_nflen|opt->opt_flen)) | |
1455 | opt = NULL; | |
9f542f61 | 1456 | connected = false; |
1da177e4 | 1457 | } |
45f6fad8 ED |
1458 | if (!opt) { |
1459 | opt = txopt_get(np); | |
1460 | opt_to_free = opt; | |
1461 | } | |
df9890c3 YH |
1462 | if (flowlabel) |
1463 | opt = fl6_merge_options(&opt_space, flowlabel, opt); | |
1464 | opt = ipv6_fixup_options(&opt_space, opt); | |
26879da5 | 1465 | ipc6.opt = opt; |
1da177e4 | 1466 | |
4c9483b2 | 1467 | fl6.flowi6_proto = sk->sk_protocol; |
42dcfd85 | 1468 | fl6.flowi6_mark = ipc6.sockc.mark; |
e8e36984 | 1469 | fl6.daddr = *daddr; |
4c9483b2 | 1470 | if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) |
4e3fd7a0 | 1471 | fl6.saddr = np->saddr; |
1958b856 | 1472 | fl6.fl6_sport = inet->inet_sport; |
1ab1457c | 1473 | |
6fc88c35 | 1474 | if (cgroup_bpf_enabled(CGROUP_UDP6_SENDMSG) && !connected) { |
1cedee13 AI |
1475 | err = BPF_CGROUP_RUN_PROG_UDP6_SENDMSG_LOCK(sk, |
1476 | (struct sockaddr *)sin6, &fl6.saddr); | |
1477 | if (err) | |
1478 | goto out_no_dst; | |
1479 | if (sin6) { | |
1480 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) { | |
1481 | /* BPF program rewrote IPv6-only by IPv4-mapped | |
1482 | * IPv6. It's currently unsupported. | |
1483 | */ | |
1484 | err = -ENOTSUPP; | |
1485 | goto out_no_dst; | |
1486 | } | |
1487 | if (sin6->sin6_port == 0) { | |
1488 | /* BPF program set invalid port. Reject it. */ | |
1489 | err = -EINVAL; | |
1490 | goto out_no_dst; | |
1491 | } | |
1492 | fl6.fl6_dport = sin6->sin6_port; | |
1493 | fl6.daddr = sin6->sin6_addr; | |
1494 | } | |
1495 | } | |
1496 | ||
e8e36984 AI |
1497 | if (ipv6_addr_any(&fl6.daddr)) |
1498 | fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ | |
1499 | ||
4c9483b2 | 1500 | final_p = fl6_update_dst(&fl6, opt, &final); |
20c59de2 | 1501 | if (final_p) |
9f542f61 | 1502 | connected = false; |
1da177e4 | 1503 | |
4c9483b2 DM |
1504 | if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) { |
1505 | fl6.flowi6_oif = np->mcast_oif; | |
9f542f61 | 1506 | connected = false; |
c4062dfc EH |
1507 | } else if (!fl6.flowi6_oif) |
1508 | fl6.flowi6_oif = np->ucast_oif; | |
1da177e4 | 1509 | |
3df98d79 | 1510 | security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6)); |
beb8d13b | 1511 | |
38b7097b HFS |
1512 | if (ipc6.tclass < 0) |
1513 | ipc6.tclass = np->tclass; | |
1514 | ||
1515 | fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel); | |
1516 | ||
4f858c56 | 1517 | dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, connected); |
68d0c6d3 DM |
1518 | if (IS_ERR(dst)) { |
1519 | err = PTR_ERR(dst); | |
1520 | dst = NULL; | |
1da177e4 | 1521 | goto out; |
14e50e57 | 1522 | } |
1da177e4 | 1523 | |
26879da5 WW |
1524 | if (ipc6.hlimit < 0) |
1525 | ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); | |
1da177e4 LT |
1526 | |
1527 | if (msg->msg_flags&MSG_CONFIRM) | |
1528 | goto do_confirm; | |
1529 | back_from_confirm: | |
1530 | ||
03485f2a VY |
1531 | /* Lockless fast path for the non-corking case */ |
1532 | if (!corkreq) { | |
1cd7884d | 1533 | struct inet_cork_full cork; |
03485f2a VY |
1534 | struct sk_buff *skb; |
1535 | ||
f37a4cc6 PB |
1536 | cork.fl.u.ip6 = fl6; |
1537 | ||
03485f2a | 1538 | skb = ip6_make_skb(sk, getfrag, msg, ulen, |
26879da5 | 1539 | sizeof(struct udphdr), &ipc6, |
f37a4cc6 | 1540 | (struct rt6_info *)dst, |
5fdaa88d | 1541 | msg->msg_flags, &cork); |
03485f2a VY |
1542 | err = PTR_ERR(skb); |
1543 | if (!IS_ERR_OR_NULL(skb)) | |
bec1f6f6 | 1544 | err = udp_v6_send_skb(skb, &fl6, &cork.base); |
4f858c56 | 1545 | goto out; |
03485f2a VY |
1546 | } |
1547 | ||
1da177e4 LT |
1548 | lock_sock(sk); |
1549 | if (unlikely(up->pending)) { | |
1550 | /* The socket is already corked while preparing it. */ | |
1551 | /* ... which is an evident application bug. --ANK */ | |
1552 | release_sock(sk); | |
1553 | ||
ba7a46f1 | 1554 | net_dbg_ratelimited("udp cork app bug 2\n"); |
1da177e4 LT |
1555 | err = -EINVAL; |
1556 | goto out; | |
1557 | } | |
1558 | ||
1559 | up->pending = AF_INET6; | |
1560 | ||
1561 | do_append_data: | |
26879da5 WW |
1562 | if (ipc6.dontfrag < 0) |
1563 | ipc6.dontfrag = np->dontfrag; | |
1da177e4 | 1564 | up->len += ulen; |
26879da5 WW |
1565 | err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr), |
1566 | &ipc6, &fl6, (struct rt6_info *)dst, | |
5fdaa88d | 1567 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); |
1da177e4 LT |
1568 | if (err) |
1569 | udp_v6_flush_pending_frames(sk); | |
1570 | else if (!corkreq) | |
4c0a6cb0 | 1571 | err = udp_v6_push_pending_frames(sk); |
1e0c14f4 HX |
1572 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) |
1573 | up->pending = 0; | |
1da177e4 | 1574 | |
03485f2a VY |
1575 | if (err > 0) |
1576 | err = np->recverr ? net_xmit_errno(err) : 0; | |
1577 | release_sock(sk); | |
1578 | ||
1da177e4 | 1579 | out: |
a3c96089 | 1580 | dst_release(dst); |
1cedee13 | 1581 | out_no_dst: |
1da177e4 | 1582 | fl6_sock_release(flowlabel); |
45f6fad8 | 1583 | txopt_put(opt_to_free); |
cd562c98 | 1584 | if (!err) |
1da177e4 | 1585 | return len; |
a18135eb DM |
1586 | /* |
1587 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1588 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1589 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1590 | * things). We could add another new stat but at least for now that | |
1591 | * seems like overkill. | |
1592 | */ | |
1593 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
6aef70a8 ED |
1594 | UDP6_INC_STATS(sock_net(sk), |
1595 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
a18135eb | 1596 | } |
1da177e4 LT |
1597 | return err; |
1598 | ||
1599 | do_confirm: | |
0dec879f JA |
1600 | if (msg->msg_flags & MSG_PROBE) |
1601 | dst_confirm_neigh(dst, &fl6.daddr); | |
1da177e4 LT |
1602 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1603 | goto back_from_confirm; | |
1604 | err = 0; | |
1605 | goto out; | |
1606 | } | |
1607 | ||
7d06b2e0 | 1608 | void udpv6_destroy_sock(struct sock *sk) |
1da177e4 | 1609 | { |
44046a59 | 1610 | struct udp_sock *up = udp_sk(sk); |
1da177e4 | 1611 | lock_sock(sk); |
a8b897c7 PA |
1612 | |
1613 | /* protects from races with udp_abort() */ | |
1614 | sock_set_flag(sk, SOCK_DEAD); | |
1da177e4 LT |
1615 | udp_v6_flush_pending_frames(sk); |
1616 | release_sock(sk); | |
1617 | ||
60fb9567 PA |
1618 | if (static_branch_unlikely(&udpv6_encap_needed_key)) { |
1619 | if (up->encap_type) { | |
1620 | void (*encap_destroy)(struct sock *sk); | |
1621 | encap_destroy = READ_ONCE(up->encap_destroy); | |
1622 | if (encap_destroy) | |
1623 | encap_destroy(sk); | |
1624 | } | |
a4a600dd | 1625 | if (up->encap_enabled) { |
9c480601 | 1626 | static_branch_dec(&udpv6_encap_needed_key); |
a4a600dd XL |
1627 | udp_encap_disable(); |
1628 | } | |
44046a59 TP |
1629 | } |
1630 | ||
1da177e4 | 1631 | inet6_destroy_sock(sk); |
1da177e4 LT |
1632 | } |
1633 | ||
1634 | /* | |
1635 | * Socket option code for UDP | |
1636 | */ | |
a7b75c5a CH |
1637 | int udpv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval, |
1638 | unsigned int optlen) | |
3fdadf7d | 1639 | { |
db8dac20 | 1640 | if (level == SOL_UDP || level == SOL_UDPLITE) |
91ac1cca | 1641 | return udp_lib_setsockopt(sk, level, optname, |
a7b75c5a | 1642 | optval, optlen, |
4c0a6cb0 | 1643 | udp_v6_push_pending_frames); |
ba4e58ec | 1644 | return ipv6_setsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1645 | } |
1646 | ||
ba4e58ec GR |
1647 | int udpv6_getsockopt(struct sock *sk, int level, int optname, |
1648 | char __user *optval, int __user *optlen) | |
3fdadf7d | 1649 | { |
db8dac20 | 1650 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 | 1651 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); |
ba4e58ec | 1652 | return ipv6_getsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1653 | } |
1654 | ||
a8e3bb34 DA |
1655 | /* thinking of making this const? Don't. |
1656 | * early_demux can change based on sysctl. | |
1657 | */ | |
dddb64bc | 1658 | static struct inet6_protocol udpv6_protocol = { |
5425077d | 1659 | .early_demux = udp_v6_early_demux, |
dddb64bc | 1660 | .early_demux_handler = udp_v6_early_demux, |
1da177e4 LT |
1661 | .handler = udpv6_rcv, |
1662 | .err_handler = udpv6_err, | |
1663 | .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, | |
1664 | }; | |
1665 | ||
1666 | /* ------------------------------------------------------------------------ */ | |
1667 | #ifdef CONFIG_PROC_FS | |
ba4e58ec | 1668 | int udp6_seq_show(struct seq_file *seq, void *v) |
1da177e4 | 1669 | { |
17ef66af LC |
1670 | if (v == SEQ_START_TOKEN) { |
1671 | seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); | |
1672 | } else { | |
1673 | int bucket = ((struct udp_iter_state *)seq->private)->bucket; | |
1674 | struct inet_sock *inet = inet_sk(v); | |
1675 | __u16 srcp = ntohs(inet->inet_sport); | |
1676 | __u16 destp = ntohs(inet->inet_dport); | |
6c206b20 PA |
1677 | __ip6_dgram_sock_seq_show(seq, v, srcp, destp, |
1678 | udp_rqueue_get(v), bucket); | |
17ef66af | 1679 | } |
1da177e4 LT |
1680 | return 0; |
1681 | } | |
1682 | ||
c3506372 | 1683 | const struct seq_operations udp6_seq_ops = { |
a3d2599b CH |
1684 | .start = udp_seq_start, |
1685 | .next = udp_seq_next, | |
1686 | .stop = udp_seq_stop, | |
1687 | .show = udp6_seq_show, | |
1688 | }; | |
c3506372 | 1689 | EXPORT_SYMBOL(udp6_seq_ops); |
73cb88ec | 1690 | |
1da177e4 | 1691 | static struct udp_seq_afinfo udp6_seq_afinfo = { |
1da177e4 | 1692 | .family = AF_INET6, |
645ca708 | 1693 | .udp_table = &udp_table, |
1da177e4 LT |
1694 | }; |
1695 | ||
2c8c1e72 | 1696 | int __net_init udp6_proc_init(struct net *net) |
1da177e4 | 1697 | { |
c3506372 CH |
1698 | if (!proc_create_net_data("udp6", 0444, net->proc_net, &udp6_seq_ops, |
1699 | sizeof(struct udp_iter_state), &udp6_seq_afinfo)) | |
a3d2599b CH |
1700 | return -ENOMEM; |
1701 | return 0; | |
1da177e4 LT |
1702 | } |
1703 | ||
ec120da6 IM |
1704 | void udp6_proc_exit(struct net *net) |
1705 | { | |
a3d2599b | 1706 | remove_proc_entry("udp6", net->proc_net); |
1da177e4 LT |
1707 | } |
1708 | #endif /* CONFIG_PROC_FS */ | |
1709 | ||
1710 | /* ------------------------------------------------------------------------ */ | |
1711 | ||
1712 | struct proto udpv6_prot = { | |
1e802951 TZ |
1713 | .name = "UDPv6", |
1714 | .owner = THIS_MODULE, | |
1715 | .close = udp_lib_close, | |
d74bad4e | 1716 | .pre_connect = udpv6_pre_connect, |
1e802951 TZ |
1717 | .connect = ip6_datagram_connect, |
1718 | .disconnect = udp_disconnect, | |
1719 | .ioctl = udp_ioctl, | |
1720 | .init = udp_init_sock, | |
1721 | .destroy = udpv6_destroy_sock, | |
1722 | .setsockopt = udpv6_setsockopt, | |
1723 | .getsockopt = udpv6_getsockopt, | |
1724 | .sendmsg = udpv6_sendmsg, | |
1725 | .recvmsg = udpv6_recvmsg, | |
1726 | .release_cb = ip6_datagram_release_cb, | |
1727 | .hash = udp_lib_hash, | |
1728 | .unhash = udp_lib_unhash, | |
1729 | .rehash = udp_v6_rehash, | |
1730 | .get_port = udp_v6_get_port, | |
91a760b2 | 1731 | .put_port = udp_lib_unhash, |
8a59f9d1 CW |
1732 | #ifdef CONFIG_BPF_SYSCALL |
1733 | .psock_update_sk_prot = udp_bpf_update_proto, | |
1734 | #endif | |
1e802951 TZ |
1735 | .memory_allocated = &udp_memory_allocated, |
1736 | .sysctl_mem = sysctl_udp_mem, | |
1737 | .sysctl_wmem_offset = offsetof(struct net, ipv4.sysctl_udp_wmem_min), | |
1738 | .sysctl_rmem_offset = offsetof(struct net, ipv4.sysctl_udp_rmem_min), | |
1739 | .obj_size = sizeof(struct udp6_sock), | |
1740 | .h.udp_table = &udp_table, | |
1e802951 | 1741 | .diag_destroy = udp_abort, |
1da177e4 LT |
1742 | }; |
1743 | ||
1da177e4 LT |
1744 | static struct inet_protosw udpv6_protosw = { |
1745 | .type = SOCK_DGRAM, | |
1746 | .protocol = IPPROTO_UDP, | |
1747 | .prot = &udpv6_prot, | |
1748 | .ops = &inet6_dgram_ops, | |
1da177e4 LT |
1749 | .flags = INET_PROTOSW_PERMANENT, |
1750 | }; | |
1751 | ||
7f4e4868 | 1752 | int __init udpv6_init(void) |
1da177e4 | 1753 | { |
7f4e4868 DL |
1754 | int ret; |
1755 | ||
3336288a VY |
1756 | ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP); |
1757 | if (ret) | |
c6b641a4 | 1758 | goto out; |
3336288a | 1759 | |
7f4e4868 DL |
1760 | ret = inet6_register_protosw(&udpv6_protosw); |
1761 | if (ret) | |
1762 | goto out_udpv6_protocol; | |
1763 | out: | |
1764 | return ret; | |
1765 | ||
1766 | out_udpv6_protocol: | |
1767 | inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); | |
1768 | goto out; | |
1769 | } | |
1770 | ||
09f7709f | 1771 | void udpv6_exit(void) |
7f4e4868 DL |
1772 | { |
1773 | inet6_unregister_protosw(&udpv6_protosw); | |
1774 | inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); | |
1da177e4 | 1775 | } |