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 | |
7170a977 | 138 | if (READ_ONCE(sk->sk_incoming_cpu) == raw_smp_processor_id()) |
70da268b ED |
139 | score++; |
140 | ||
645ca708 ED |
141 | return score; |
142 | } | |
143 | ||
a57066b1 DM |
144 | static struct sock *lookup_reuseport(struct net *net, struct sock *sk, |
145 | struct sk_buff *skb, | |
146 | const struct in6_addr *saddr, | |
147 | __be16 sport, | |
148 | const struct in6_addr *daddr, | |
149 | unsigned int hnum) | |
2a08748c JS |
150 | { |
151 | struct sock *reuse_sk = NULL; | |
152 | u32 hash; | |
153 | ||
154 | if (sk->sk_reuseport && sk->sk_state != TCP_ESTABLISHED) { | |
155 | hash = udp6_ehashfn(net, daddr, hnum, saddr, sport); | |
156 | reuse_sk = reuseport_select_sock(sk, hash, skb, | |
157 | sizeof(struct udphdr)); | |
2a08748c JS |
158 | } |
159 | return reuse_sk; | |
160 | } | |
161 | ||
d1e37288 | 162 | /* called with rcu_read_lock() */ |
fddc17de ED |
163 | static struct sock *udp6_lib_lookup2(struct net *net, |
164 | const struct in6_addr *saddr, __be16 sport, | |
1801b570 | 165 | const struct in6_addr *daddr, unsigned int hnum, |
73545373 TB |
166 | int dif, int sdif, struct udp_hslot *hslot2, |
167 | struct sk_buff *skb) | |
fddc17de ED |
168 | { |
169 | struct sock *sk, *result; | |
e94a62f5 | 170 | int score, badness; |
fddc17de | 171 | |
fddc17de ED |
172 | result = NULL; |
173 | badness = -1; | |
ca065d0c | 174 | udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
d1e37288 | 175 | score = compute_score(sk, net, saddr, sport, |
73545373 | 176 | daddr, hnum, dif, sdif); |
fddc17de | 177 | if (score > badness) { |
2a08748c JS |
178 | result = lookup_reuseport(net, sk, skb, |
179 | saddr, sport, daddr, hnum); | |
a57066b1 DM |
180 | /* Fall back to scoring if group has connections */ |
181 | if (result && !reuseport_has_conns(sk, false)) | |
2a08748c JS |
182 | return result; |
183 | ||
a57066b1 | 184 | result = result ? : sk; |
ca065d0c | 185 | badness = score; |
fddc17de ED |
186 | } |
187 | } | |
fddc17de ED |
188 | return result; |
189 | } | |
190 | ||
6d4201b1 JS |
191 | static inline struct sock *udp6_lookup_run_bpf(struct net *net, |
192 | struct udp_table *udptable, | |
193 | struct sk_buff *skb, | |
194 | const struct in6_addr *saddr, | |
195 | __be16 sport, | |
196 | const struct in6_addr *daddr, | |
197 | u16 hnum) | |
198 | { | |
199 | struct sock *sk, *reuse_sk; | |
200 | bool no_reuseport; | |
201 | ||
202 | if (udptable != &udp_table) | |
203 | return NULL; /* only UDP is supported */ | |
204 | ||
205 | no_reuseport = bpf_sk_lookup_run_v6(net, IPPROTO_UDP, | |
206 | saddr, sport, daddr, hnum, &sk); | |
207 | if (no_reuseport || IS_ERR_OR_NULL(sk)) | |
208 | return sk; | |
209 | ||
210 | reuse_sk = lookup_reuseport(net, sk, skb, saddr, sport, daddr, hnum); | |
c64c9c28 | 211 | if (reuse_sk) |
6d4201b1 JS |
212 | sk = reuse_sk; |
213 | return sk; | |
214 | } | |
215 | ||
ca065d0c | 216 | /* rcu_read_lock() must be held */ |
fce82338 | 217 | struct sock *__udp6_lib_lookup(struct net *net, |
1801b570 DA |
218 | const struct in6_addr *saddr, __be16 sport, |
219 | const struct in6_addr *daddr, __be16 dport, | |
220 | int dif, int sdif, struct udp_table *udptable, | |
221 | struct sk_buff *skb) | |
1da177e4 | 222 | { |
1da177e4 | 223 | unsigned short hnum = ntohs(dport); |
23b0269e PO |
224 | unsigned int hash2, slot2; |
225 | struct udp_hslot *hslot2; | |
6d4201b1 | 226 | struct sock *result, *sk; |
645ca708 | 227 | |
23b0269e PO |
228 | hash2 = ipv6_portaddr_hash(net, daddr, hnum); |
229 | slot2 = hash2 & udptable->mask; | |
230 | hslot2 = &udptable->hash2[slot2]; | |
231 | ||
6d4201b1 | 232 | /* Lookup connected or non-wildcard sockets */ |
23b0269e | 233 | result = udp6_lib_lookup2(net, saddr, sport, |
73545373 | 234 | daddr, hnum, dif, sdif, |
23b0269e | 235 | hslot2, skb); |
6d4201b1 JS |
236 | if (!IS_ERR_OR_NULL(result) && result->sk_state == TCP_ESTABLISHED) |
237 | goto done; | |
238 | ||
239 | /* Lookup redirect from BPF */ | |
240 | if (static_branch_unlikely(&bpf_sk_lookup_enabled)) { | |
241 | sk = udp6_lookup_run_bpf(net, udptable, skb, | |
242 | saddr, sport, daddr, hnum); | |
243 | if (sk) { | |
244 | result = sk; | |
245 | goto done; | |
246 | } | |
247 | } | |
23b0269e | 248 | |
6d4201b1 JS |
249 | /* Got non-wildcard socket or error on first lookup */ |
250 | if (result) | |
251 | goto done; | |
fddc17de | 252 | |
6d4201b1 JS |
253 | /* Lookup wildcard sockets */ |
254 | hash2 = ipv6_portaddr_hash(net, &in6addr_any, hnum); | |
255 | slot2 = hash2 & udptable->mask; | |
256 | hslot2 = &udptable->hash2[slot2]; | |
257 | ||
258 | result = udp6_lib_lookup2(net, saddr, sport, | |
259 | &in6addr_any, hnum, dif, sdif, | |
260 | hslot2, skb); | |
261 | done: | |
26f8113c | 262 | if (IS_ERR(result)) |
23b0269e | 263 | return NULL; |
1da177e4 LT |
264 | return result; |
265 | } | |
fce82338 | 266 | EXPORT_SYMBOL_GPL(__udp6_lib_lookup); |
1da177e4 | 267 | |
607c4aaf KK |
268 | static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb, |
269 | __be16 sport, __be16 dport, | |
645ca708 | 270 | struct udp_table *udptable) |
607c4aaf | 271 | { |
b71d1d42 | 272 | const struct ipv6hdr *iph = ipv6_hdr(skb); |
607c4aaf | 273 | |
ed7cbbce | 274 | return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport, |
adf30907 | 275 | &iph->daddr, dport, inet6_iif(skb), |
1801b570 | 276 | inet6_sdif(skb), udptable, skb); |
607c4aaf KK |
277 | } |
278 | ||
63058308 TH |
279 | struct sock *udp6_lib_lookup_skb(struct sk_buff *skb, |
280 | __be16 sport, __be16 dport) | |
281 | { | |
282 | const struct ipv6hdr *iph = ipv6_hdr(skb); | |
63058308 | 283 | |
ed7cbbce | 284 | return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport, |
63058308 | 285 | &iph->daddr, dport, inet6_iif(skb), |
257a525f | 286 | inet6_sdif(skb), &udp_table, NULL); |
63058308 TH |
287 | } |
288 | EXPORT_SYMBOL_GPL(udp6_lib_lookup_skb); | |
289 | ||
ca065d0c ED |
290 | /* Must be called under rcu_read_lock(). |
291 | * Does increment socket refcount. | |
292 | */ | |
6e86000c | 293 | #if IS_ENABLED(CONFIG_NF_TPROXY_IPV6) || IS_ENABLED(CONFIG_NF_SOCKET_IPV6) |
aa976fc0 BS |
294 | struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport, |
295 | const struct in6_addr *daddr, __be16 dport, int dif) | |
296 | { | |
ca065d0c ED |
297 | struct sock *sk; |
298 | ||
299 | sk = __udp6_lib_lookup(net, saddr, sport, daddr, dport, | |
1801b570 | 300 | dif, 0, &udp_table, NULL); |
41c6d650 | 301 | if (sk && !refcount_inc_not_zero(&sk->sk_refcnt)) |
ca065d0c ED |
302 | sk = NULL; |
303 | return sk; | |
aa976fc0 BS |
304 | } |
305 | EXPORT_SYMBOL_GPL(udp6_lib_lookup); | |
ca065d0c | 306 | #endif |
aa976fc0 | 307 | |
cb891fa6 PA |
308 | /* do not use the scratch area len for jumbogram: their length execeeds the |
309 | * scratch area space; note that the IP6CB flags is still in the first | |
310 | * cacheline, so checking for jumbograms is cheap | |
311 | */ | |
312 | static int udp6_skb_len(struct sk_buff *skb) | |
313 | { | |
314 | return unlikely(inet6_is_jumbogram(skb)) ? skb->len : udp_skb_len(skb); | |
315 | } | |
316 | ||
1da177e4 | 317 | /* |
67ba4152 IM |
318 | * This should be easy, if there is something there we |
319 | * return it, otherwise we block. | |
1da177e4 LT |
320 | */ |
321 | ||
1b784140 | 322 | int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, |
1da177e4 LT |
323 | int noblock, int flags, int *addr_len) |
324 | { | |
325 | struct ipv6_pinfo *np = inet6_sk(sk); | |
326 | struct inet_sock *inet = inet_sk(sk); | |
1ab1457c | 327 | struct sk_buff *skb; |
59c2cdae | 328 | unsigned int ulen, copied; |
fd69c399 | 329 | int off, err, peeking = flags & MSG_PEEK; |
759e5d00 | 330 | int is_udplite = IS_UDPLITE(sk); |
543fc3fb | 331 | struct udp_mib __percpu *mib; |
197c949e | 332 | bool checksum_valid = false; |
f26ba175 | 333 | int is_udp4; |
1da177e4 | 334 | |
1da177e4 | 335 | if (flags & MSG_ERRQUEUE) |
85fbaa75 | 336 | return ipv6_recv_error(sk, msg, len, addr_len); |
1da177e4 | 337 | |
4b340ae2 | 338 | if (np->rxpmtu && np->rxopt.bits.rxpmtu) |
85fbaa75 | 339 | return ipv6_recv_rxpmtu(sk, msg, len, addr_len); |
4b340ae2 | 340 | |
1da177e4 | 341 | try_again: |
a0917e0b | 342 | off = sk_peek_offset(sk, flags); |
fd69c399 | 343 | skb = __skb_recv_udp(sk, flags, noblock, &off, &err); |
1da177e4 | 344 | if (!skb) |
627d2d6b | 345 | return err; |
1da177e4 | 346 | |
cb891fa6 | 347 | ulen = udp6_skb_len(skb); |
59c2cdae | 348 | copied = len; |
627d2d6b | 349 | if (copied > ulen - off) |
350 | copied = ulen - off; | |
59c2cdae | 351 | else if (copied < ulen) |
1ab1457c | 352 | msg->msg_flags |= MSG_TRUNC; |
1da177e4 | 353 | |
f26ba175 | 354 | is_udp4 = (skb->protocol == htons(ETH_P_IP)); |
029a3743 | 355 | mib = __UDPX_MIB(sk, is_udp4); |
f26ba175 | 356 | |
ba4e58ec | 357 | /* |
759e5d00 HX |
358 | * If checksum is needed at all, try to do it while copying the |
359 | * data. If the data is truncated, or if we only want a partial | |
360 | * coverage checksum (UDP-Lite), do it before the copy. | |
ba4e58ec | 361 | */ |
ba4e58ec | 362 | |
d21dbdfe ED |
363 | if (copied < ulen || peeking || |
364 | (is_udplite && UDP_SKB_CB(skb)->partial_cov)) { | |
67a51780 PA |
365 | checksum_valid = udp_skb_csum_unnecessary(skb) || |
366 | !__udp_lib_checksum_complete(skb); | |
197c949e | 367 | if (!checksum_valid) |
1da177e4 | 368 | goto csum_copy_err; |
ba4e58ec GR |
369 | } |
370 | ||
67a51780 PA |
371 | if (checksum_valid || udp_skb_csum_unnecessary(skb)) { |
372 | if (udp_skb_is_linear(skb)) | |
373 | err = copy_linear_skb(skb, copied, off, &msg->msg_iter); | |
374 | else | |
375 | err = skb_copy_datagram_msg(skb, off, msg, copied); | |
376 | } else { | |
627d2d6b | 377 | err = skb_copy_and_csum_datagram_msg(skb, off, msg); |
1da177e4 LT |
378 | if (err == -EINVAL) |
379 | goto csum_copy_err; | |
380 | } | |
22911fc5 | 381 | if (unlikely(err)) { |
fd69c399 | 382 | if (!peeking) { |
979402b1 | 383 | atomic_inc(&sk->sk_drops); |
029a3743 | 384 | SNMP_INC_STATS(mib, UDP_MIB_INERRORS); |
979402b1 | 385 | } |
850cbadd | 386 | kfree_skb(skb); |
627d2d6b | 387 | return err; |
22911fc5 | 388 | } |
fd69c399 | 389 | if (!peeking) |
029a3743 | 390 | SNMP_INC_STATS(mib, UDP_MIB_INDATAGRAMS); |
cb75994e | 391 | |
3b885787 | 392 | sock_recv_ts_and_drops(msg, sk, skb); |
1da177e4 LT |
393 | |
394 | /* Copy the address. */ | |
395 | if (msg->msg_name) { | |
342dfc30 | 396 | DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
1da177e4 | 397 | sin6->sin6_family = AF_INET6; |
4bedb452 | 398 | sin6->sin6_port = udp_hdr(skb)->source; |
1da177e4 | 399 | sin6->sin6_flowinfo = 0; |
1da177e4 | 400 | |
842df073 | 401 | if (is_udp4) { |
b301e82c BH |
402 | ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, |
403 | &sin6->sin6_addr); | |
842df073 HFS |
404 | sin6->sin6_scope_id = 0; |
405 | } else { | |
4e3fd7a0 | 406 | sin6->sin6_addr = ipv6_hdr(skb)->saddr; |
842df073 HFS |
407 | sin6->sin6_scope_id = |
408 | ipv6_iface_scope_id(&sin6->sin6_addr, | |
4330487a | 409 | inet6_iif(skb)); |
1da177e4 | 410 | } |
bceaa902 | 411 | *addr_len = sizeof(*sin6); |
983695fa DB |
412 | |
413 | if (cgroup_bpf_enabled) | |
414 | BPF_CGROUP_RUN_PROG_UDP6_RECVMSG_LOCK(sk, | |
415 | (struct sockaddr *)sin6); | |
1da177e4 | 416 | } |
4b261c75 | 417 | |
bcd1665e PA |
418 | if (udp_sk(sk)->gro_enabled) |
419 | udp_cmsg_recv(msg, sk, skb); | |
420 | ||
4b261c75 HFS |
421 | if (np->rxopt.all) |
422 | ip6_datagram_recv_common_ctl(sk, msg, skb); | |
423 | ||
f26ba175 | 424 | if (is_udp4) { |
1da177e4 | 425 | if (inet->cmsg_flags) |
ad959036 | 426 | ip_cmsg_recv_offset(msg, sk, skb, |
10df8e61 | 427 | sizeof(struct udphdr), off); |
1da177e4 LT |
428 | } else { |
429 | if (np->rxopt.all) | |
4b261c75 | 430 | ip6_datagram_recv_specific_ctl(sk, msg, skb); |
1ab1457c | 431 | } |
1da177e4 | 432 | |
59c2cdae | 433 | err = copied; |
1da177e4 | 434 | if (flags & MSG_TRUNC) |
759e5d00 | 435 | err = ulen; |
1da177e4 | 436 | |
850cbadd | 437 | skb_consume_udp(sk, skb, peeking ? -err : err); |
1da177e4 LT |
438 | return err; |
439 | ||
440 | csum_copy_err: | |
2276f58a PA |
441 | if (!__sk_queue_drop_skb(sk, &udp_sk(sk)->reader_queue, skb, flags, |
442 | udp_skb_destructor)) { | |
029a3743 PA |
443 | SNMP_INC_STATS(mib, UDP_MIB_CSUMERRORS); |
444 | SNMP_INC_STATS(mib, UDP_MIB_INERRORS); | |
0856f939 | 445 | } |
850cbadd | 446 | kfree_skb(skb); |
1da177e4 | 447 | |
beb39db5 ED |
448 | /* starting over for a new packet, but check if we need to yield */ |
449 | cond_resched(); | |
9cfaa8de | 450 | msg->msg_flags &= ~MSG_TRUNC; |
1da177e4 LT |
451 | goto try_again; |
452 | } | |
453 | ||
a36e185e SB |
454 | DEFINE_STATIC_KEY_FALSE(udpv6_encap_needed_key); |
455 | void udpv6_encap_enable(void) | |
456 | { | |
9c480601 | 457 | static_branch_inc(&udpv6_encap_needed_key); |
a36e185e SB |
458 | } |
459 | EXPORT_SYMBOL(udpv6_encap_enable); | |
460 | ||
e7cc0824 SB |
461 | /* Handler for tunnels with arbitrary destination ports: no socket lookup, go |
462 | * through error handlers in encapsulations looking for a match. | |
463 | */ | |
464 | static int __udp6_lib_err_encap_no_sk(struct sk_buff *skb, | |
465 | struct inet6_skb_parm *opt, | |
424a7cd0 | 466 | u8 type, u8 code, int offset, __be32 info) |
e7cc0824 SB |
467 | { |
468 | int i; | |
469 | ||
470 | for (i = 0; i < MAX_IPTUN_ENCAP_OPS; i++) { | |
471 | int (*handler)(struct sk_buff *skb, struct inet6_skb_parm *opt, | |
424a7cd0 PA |
472 | u8 type, u8 code, int offset, __be32 info); |
473 | const struct ip6_tnl_encap_ops *encap; | |
e7cc0824 | 474 | |
424a7cd0 PA |
475 | encap = rcu_dereference(ip6tun_encaps[i]); |
476 | if (!encap) | |
e7cc0824 | 477 | continue; |
424a7cd0 | 478 | handler = encap->err_handler; |
e7cc0824 SB |
479 | if (handler && !handler(skb, opt, type, code, offset, info)) |
480 | return 0; | |
481 | } | |
482 | ||
483 | return -ENOENT; | |
484 | } | |
485 | ||
a36e185e SB |
486 | /* Try to match ICMP errors to UDP tunnels by looking up a socket without |
487 | * reversing source and destination port: this will match tunnels that force the | |
488 | * same destination port on both endpoints (e.g. VXLAN, GENEVE). Note that | |
489 | * lwtunnels might actually break this assumption by being configured with | |
490 | * different destination ports on endpoints, in this case we won't be able to | |
491 | * trace ICMP messages back to them. | |
492 | * | |
e7cc0824 SB |
493 | * If this doesn't match any socket, probe tunnels with arbitrary destination |
494 | * ports (e.g. FoU, GUE): there, the receiving socket is useless, as the port | |
495 | * we've sent packets to won't necessarily match the local destination port. | |
496 | * | |
a36e185e SB |
497 | * Then ask the tunnel implementation to match the error against a valid |
498 | * association. | |
499 | * | |
e7cc0824 SB |
500 | * Return an error if we can't find a match, the socket if we need further |
501 | * processing, zero otherwise. | |
a36e185e SB |
502 | */ |
503 | static struct sock *__udp6_lib_err_encap(struct net *net, | |
504 | const struct ipv6hdr *hdr, int offset, | |
505 | struct udphdr *uh, | |
506 | struct udp_table *udptable, | |
e7cc0824 SB |
507 | struct sk_buff *skb, |
508 | struct inet6_skb_parm *opt, | |
509 | u8 type, u8 code, __be32 info) | |
a36e185e | 510 | { |
a36e185e | 511 | int network_offset, transport_offset; |
a36e185e SB |
512 | struct sock *sk; |
513 | ||
a36e185e SB |
514 | network_offset = skb_network_offset(skb); |
515 | transport_offset = skb_transport_offset(skb); | |
516 | ||
517 | /* Network header needs to point to the outer IPv6 header inside ICMP */ | |
518 | skb_reset_network_header(skb); | |
519 | ||
520 | /* Transport header needs to point to the UDP header */ | |
521 | skb_set_transport_header(skb, offset); | |
522 | ||
e7cc0824 SB |
523 | sk = __udp6_lib_lookup(net, &hdr->daddr, uh->source, |
524 | &hdr->saddr, uh->dest, | |
525 | inet6_iif(skb), 0, udptable, skb); | |
526 | if (sk) { | |
527 | int (*lookup)(struct sock *sk, struct sk_buff *skb); | |
528 | struct udp_sock *up = udp_sk(sk); | |
529 | ||
530 | lookup = READ_ONCE(up->encap_err_lookup); | |
531 | if (!lookup || lookup(sk, skb)) | |
532 | sk = NULL; | |
533 | } | |
534 | ||
535 | if (!sk) { | |
536 | sk = ERR_PTR(__udp6_lib_err_encap_no_sk(skb, opt, type, code, | |
537 | offset, info)); | |
538 | } | |
a36e185e SB |
539 | |
540 | skb_set_transport_header(skb, transport_offset); | |
541 | skb_set_network_header(skb, network_offset); | |
e7cc0824 | 542 | |
a36e185e SB |
543 | return sk; |
544 | } | |
545 | ||
32bbd879 SB |
546 | int __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt, |
547 | u8 type, u8 code, int offset, __be32 info, | |
548 | struct udp_table *udptable) | |
1da177e4 LT |
549 | { |
550 | struct ipv6_pinfo *np; | |
b71d1d42 ED |
551 | const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data; |
552 | const struct in6_addr *saddr = &hdr->saddr; | |
553 | const struct in6_addr *daddr = &hdr->daddr; | |
67ba4152 | 554 | struct udphdr *uh = (struct udphdr *)(skb->data+offset); |
a36e185e | 555 | bool tunnel = false; |
1da177e4 | 556 | struct sock *sk; |
e0d8c1b7 | 557 | int harderr; |
1da177e4 | 558 | int err; |
7304fe46 | 559 | struct net *net = dev_net(skb->dev); |
1da177e4 | 560 | |
e32ea7e7 | 561 | sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source, |
4ac30c4b | 562 | inet6_iif(skb), inet6_sdif(skb), udptable, NULL); |
d26796ae | 563 | if (!sk || udp_sk(sk)->encap_type) { |
a36e185e | 564 | /* No socket for error: try tunnels before discarding */ |
e7cc0824 | 565 | sk = ERR_PTR(-ENOENT); |
a36e185e SB |
566 | if (static_branch_unlikely(&udpv6_encap_needed_key)) { |
567 | sk = __udp6_lib_err_encap(net, hdr, offset, uh, | |
e7cc0824 SB |
568 | udptable, skb, |
569 | opt, type, code, info); | |
570 | if (!sk) | |
571 | return 0; | |
a36e185e SB |
572 | } |
573 | ||
e7cc0824 | 574 | if (IS_ERR(sk)) { |
a36e185e SB |
575 | __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), |
576 | ICMP6_MIB_INERRORS); | |
e7cc0824 | 577 | return PTR_ERR(sk); |
a36e185e | 578 | } |
e7cc0824 | 579 | |
a36e185e | 580 | tunnel = true; |
7304fe46 | 581 | } |
1da177e4 | 582 | |
e0d8c1b7 WW |
583 | harderr = icmpv6_err_convert(type, code, &err); |
584 | np = inet6_sk(sk); | |
585 | ||
93b36cf3 HFS |
586 | if (type == ICMPV6_PKT_TOOBIG) { |
587 | if (!ip6_sk_accept_pmtu(sk)) | |
588 | goto out; | |
81aded24 | 589 | ip6_sk_update_pmtu(skb, sk, info); |
e0d8c1b7 WW |
590 | if (np->pmtudisc != IPV6_PMTUDISC_DONT) |
591 | harderr = 1; | |
93b36cf3 | 592 | } |
1a462d18 | 593 | if (type == NDISC_REDIRECT) { |
a36e185e SB |
594 | if (tunnel) { |
595 | ip6_redirect(skb, sock_net(sk), inet6_iif(skb), | |
596 | sk->sk_mark, sk->sk_uid); | |
597 | } else { | |
598 | ip6_sk_redirect(skb, sk); | |
599 | } | |
1a462d18 DJ |
600 | goto out; |
601 | } | |
81aded24 | 602 | |
a36e185e SB |
603 | /* Tunnels don't have an application socket: don't pass errors back */ |
604 | if (tunnel) | |
605 | goto out; | |
606 | ||
e0d8c1b7 WW |
607 | if (!np->recverr) { |
608 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
609 | goto out; | |
610 | } else { | |
1da177e4 | 611 | ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1)); |
e0d8c1b7 | 612 | } |
b1faf566 | 613 | |
1da177e4 LT |
614 | sk->sk_err = err; |
615 | sk->sk_error_report(sk); | |
616 | out: | |
32bbd879 | 617 | return 0; |
1da177e4 LT |
618 | } |
619 | ||
a3f96c47 | 620 | static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
f7ad74fe BL |
621 | { |
622 | int rc; | |
623 | ||
efe4208f | 624 | if (!ipv6_addr_any(&sk->sk_v6_daddr)) { |
f7ad74fe | 625 | sock_rps_save_rxhash(sk, skb); |
005ec974 | 626 | sk_mark_napi_id(sk, skb); |
2c8c56e1 | 627 | sk_incoming_cpu_update(sk); |
e68b6e50 ED |
628 | } else { |
629 | sk_mark_napi_id_once(sk, skb); | |
005ec974 | 630 | } |
f7ad74fe | 631 | |
850cbadd | 632 | rc = __udp_enqueue_schedule_skb(sk, skb); |
f7ad74fe BL |
633 | if (rc < 0) { |
634 | int is_udplite = IS_UDPLITE(sk); | |
635 | ||
636 | /* Note that an ENOMEM error is charged twice */ | |
637 | if (rc == -ENOMEM) | |
e61da9e2 | 638 | UDP6_INC_STATS(sock_net(sk), |
02c22347 | 639 | UDP_MIB_RCVBUFERRORS, is_udplite); |
a3ce2b10 MD |
640 | else |
641 | UDP6_INC_STATS(sock_net(sk), | |
642 | UDP_MIB_MEMERRORS, is_udplite); | |
e61da9e2 | 643 | UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
f7ad74fe BL |
644 | kfree_skb(skb); |
645 | return -1; | |
646 | } | |
850cbadd | 647 | |
f7ad74fe BL |
648 | return 0; |
649 | } | |
650 | ||
32bbd879 SB |
651 | static __inline__ int udpv6_err(struct sk_buff *skb, |
652 | struct inet6_skb_parm *opt, u8 type, | |
653 | u8 code, int offset, __be32 info) | |
ba4e58ec | 654 | { |
32bbd879 | 655 | return __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table); |
ba4e58ec GR |
656 | } |
657 | ||
cf329aa4 | 658 | static int udpv6_queue_rcv_one_skb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 659 | { |
ba4e58ec | 660 | struct udp_sock *up = udp_sk(sk); |
b2bf1e26 | 661 | int is_udplite = IS_UDPLITE(sk); |
a18135eb | 662 | |
ba4e58ec GR |
663 | if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) |
664 | goto drop; | |
1da177e4 | 665 | |
88ab3108 | 666 | if (static_branch_unlikely(&udpv6_encap_needed_key) && up->encap_type) { |
d7f3f621 BL |
667 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
668 | ||
669 | /* | |
670 | * This is an encapsulation socket so pass the skb to | |
671 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
672 | * fall through and pass this up the UDP socket. | |
673 | * up->encap_rcv() returns the following value: | |
674 | * =0 if skb was successfully passed to the encap | |
675 | * handler or was discarded by it. | |
676 | * >0 if skb should be passed on to UDP. | |
677 | * <0 if skb should be resubmitted as proto -N | |
678 | */ | |
679 | ||
680 | /* if we're overly short, let UDP handle it */ | |
6aa7de05 | 681 | encap_rcv = READ_ONCE(up->encap_rcv); |
e5aed006 | 682 | if (encap_rcv) { |
d7f3f621 BL |
683 | int ret; |
684 | ||
0a80966b TH |
685 | /* Verify checksum before giving to encap */ |
686 | if (udp_lib_checksum_complete(skb)) | |
687 | goto csum_error; | |
688 | ||
d7f3f621 BL |
689 | ret = encap_rcv(sk, skb); |
690 | if (ret <= 0) { | |
02c22347 ED |
691 | __UDP_INC_STATS(sock_net(sk), |
692 | UDP_MIB_INDATAGRAMS, | |
693 | is_udplite); | |
d7f3f621 BL |
694 | return -ret; |
695 | } | |
696 | } | |
697 | ||
698 | /* FALLTHROUGH -- it's a UDP Packet */ | |
699 | } | |
700 | ||
ba4e58ec GR |
701 | /* |
702 | * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c). | |
703 | */ | |
b0a42277 | 704 | if ((up->pcflag & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { |
ba4e58ec GR |
705 | |
706 | if (up->pcrlen == 0) { /* full coverage was set */ | |
ba7a46f1 JP |
707 | net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n", |
708 | UDP_SKB_CB(skb)->cscov, skb->len); | |
ba4e58ec GR |
709 | goto drop; |
710 | } | |
711 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
ba7a46f1 JP |
712 | net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n", |
713 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
ba4e58ec GR |
714 | goto drop; |
715 | } | |
1da177e4 LT |
716 | } |
717 | ||
4b943fae | 718 | prefetch(&sk->sk_rmem_alloc); |
ce25d66a ED |
719 | if (rcu_access_pointer(sk->sk_filter) && |
720 | udp_lib_checksum_complete(skb)) | |
721 | goto csum_error; | |
722 | ||
ba66bbe5 | 723 | if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr))) |
a6127697 | 724 | goto drop; |
ba4e58ec | 725 | |
e6afc8ac | 726 | udp_csum_pull_header(skb); |
cb80ef46 | 727 | |
d826eb14 | 728 | skb_dst_drop(skb); |
cb75994e | 729 | |
850cbadd | 730 | return __udpv6_queue_rcv_skb(sk, skb); |
3e215c8d | 731 | |
6a5dc9e5 | 732 | csum_error: |
02c22347 | 733 | __UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); |
ba4e58ec | 734 | drop: |
02c22347 | 735 | __UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
cb80ef46 | 736 | atomic_inc(&sk->sk_drops); |
ba4e58ec GR |
737 | kfree_skb(skb); |
738 | return -1; | |
1da177e4 LT |
739 | } |
740 | ||
cf329aa4 PA |
741 | static int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
742 | { | |
743 | struct sk_buff *next, *segs; | |
744 | int ret; | |
745 | ||
746 | if (likely(!udp_unexpected_gso(sk, skb))) | |
747 | return udpv6_queue_rcv_one_skb(sk, skb); | |
748 | ||
749 | __skb_push(skb, -skb_mac_offset(skb)); | |
750 | segs = udp_rcv_segment(sk, skb, false); | |
1a186c14 | 751 | skb_list_walk_safe(segs, skb, next) { |
cf329aa4 PA |
752 | __skb_pull(skb, skb_transport_offset(skb)); |
753 | ||
754 | ret = udpv6_queue_rcv_one_skb(sk, skb); | |
755 | if (ret > 0) | |
756 | ip6_protocol_deliver_rcu(dev_net(skb->dev), skb, ret, | |
757 | true); | |
758 | } | |
759 | return 0; | |
760 | } | |
761 | ||
5cf3d461 DH |
762 | static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk, |
763 | __be16 loc_port, const struct in6_addr *loc_addr, | |
764 | __be16 rmt_port, const struct in6_addr *rmt_addr, | |
7bd2db40 | 765 | int dif, int sdif, unsigned short hnum) |
1da177e4 | 766 | { |
5cf3d461 | 767 | struct inet_sock *inet = inet_sk(sk); |
1da177e4 | 768 | |
5cf3d461 DH |
769 | if (!net_eq(sock_net(sk), net)) |
770 | return false; | |
771 | ||
772 | if (udp_sk(sk)->udp_port_hash != hnum || | |
773 | sk->sk_family != PF_INET6 || | |
774 | (inet->inet_dport && inet->inet_dport != rmt_port) || | |
775 | (!ipv6_addr_any(&sk->sk_v6_daddr) && | |
776 | !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) || | |
7bd2db40 | 777 | !udp_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif) || |
33b4b015 HR |
778 | (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) && |
779 | !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))) | |
5cf3d461 DH |
780 | return false; |
781 | if (!inet6_mc_check(sk, loc_addr, rmt_addr)) | |
782 | return false; | |
783 | return true; | |
1da177e4 LT |
784 | } |
785 | ||
4068579e TH |
786 | static void udp6_csum_zero_error(struct sk_buff *skb) |
787 | { | |
788 | /* RFC 2460 section 8.1 says that we SHOULD log | |
789 | * this error. Well, it is reasonable. | |
790 | */ | |
ba7a46f1 JP |
791 | net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n", |
792 | &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source), | |
793 | &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest)); | |
4068579e TH |
794 | } |
795 | ||
1da177e4 LT |
796 | /* |
797 | * Note: called only from the BH handler context, | |
798 | * so we don't need to lock the hashes. | |
799 | */ | |
e3163493 | 800 | static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
b71d1d42 | 801 | const struct in6_addr *saddr, const struct in6_addr *daddr, |
36cbb245 | 802 | struct udp_table *udptable, int proto) |
1da177e4 | 803 | { |
ca065d0c | 804 | struct sock *sk, *first = NULL; |
4bedb452 | 805 | const struct udphdr *uh = udp_hdr(skb); |
5cf3d461 DH |
806 | unsigned short hnum = ntohs(uh->dest); |
807 | struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum); | |
ca065d0c | 808 | unsigned int offset = offsetof(typeof(*sk), sk_node); |
2dc41cff | 809 | unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10); |
ca065d0c | 810 | int dif = inet6_iif(skb); |
7bd2db40 | 811 | int sdif = inet6_sdif(skb); |
ca065d0c ED |
812 | struct hlist_node *node; |
813 | struct sk_buff *nskb; | |
2dc41cff DH |
814 | |
815 | if (use_hash2) { | |
f0b1e64c | 816 | hash2_any = ipv6_portaddr_hash(net, &in6addr_any, hnum) & |
73e2d5e3 | 817 | udptable->mask; |
f0b1e64c | 818 | hash2 = ipv6_portaddr_hash(net, daddr, hnum) & udptable->mask; |
2dc41cff | 819 | start_lookup: |
73e2d5e3 | 820 | hslot = &udptable->hash2[hash2]; |
2dc41cff DH |
821 | offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node); |
822 | } | |
1da177e4 | 823 | |
ca065d0c ED |
824 | sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) { |
825 | if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr, | |
7bd2db40 DM |
826 | uh->source, saddr, dif, sdif, |
827 | hnum)) | |
ca065d0c ED |
828 | continue; |
829 | /* If zero checksum and no_check is not on for | |
830 | * the socket then skip it. | |
831 | */ | |
832 | if (!uh->check && !udp_sk(sk)->no_check6_rx) | |
833 | continue; | |
834 | if (!first) { | |
835 | first = sk; | |
836 | continue; | |
837 | } | |
838 | nskb = skb_clone(skb, GFP_ATOMIC); | |
839 | if (unlikely(!nskb)) { | |
840 | atomic_inc(&sk->sk_drops); | |
02c22347 ED |
841 | __UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS, |
842 | IS_UDPLITE(sk)); | |
843 | __UDP6_INC_STATS(net, UDP_MIB_INERRORS, | |
844 | IS_UDPLITE(sk)); | |
ca065d0c | 845 | continue; |
95766fff | 846 | } |
a1ab77f9 | 847 | |
ca065d0c ED |
848 | if (udpv6_queue_rcv_skb(sk, nskb) > 0) |
849 | consume_skb(nskb); | |
850 | } | |
a1ab77f9 | 851 | |
2dc41cff DH |
852 | /* Also lookup *:port if we are using hash2 and haven't done so yet. */ |
853 | if (use_hash2 && hash2 != hash2_any) { | |
854 | hash2 = hash2_any; | |
855 | goto start_lookup; | |
856 | } | |
857 | ||
ca065d0c ED |
858 | if (first) { |
859 | if (udpv6_queue_rcv_skb(first, skb) > 0) | |
860 | consume_skb(skb); | |
a1ab77f9 | 861 | } else { |
ca065d0c | 862 | kfree_skb(skb); |
02c22347 ED |
863 | __UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI, |
864 | proto == IPPROTO_UDPLITE); | |
a1ab77f9 | 865 | } |
ba4e58ec | 866 | return 0; |
1da177e4 LT |
867 | } |
868 | ||
64f0f5d1 PA |
869 | static void udp6_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst) |
870 | { | |
871 | if (udp_sk_rx_dst_set(sk, dst)) { | |
872 | const struct rt6_info *rt = (const struct rt6_info *)dst; | |
873 | ||
874 | inet6_sk(sk)->rx_dst_cookie = rt6_get_cookie(rt); | |
875 | } | |
876 | } | |
877 | ||
eb63f296 PA |
878 | /* wrapper for udp_queue_rcv_skb tacking care of csum conversion and |
879 | * return code conversion for ip layer consumption | |
880 | */ | |
881 | static int udp6_unicast_rcv_skb(struct sock *sk, struct sk_buff *skb, | |
882 | struct udphdr *uh) | |
883 | { | |
884 | int ret; | |
885 | ||
886 | if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk)) | |
e4aa33ad | 887 | skb_checksum_try_convert(skb, IPPROTO_UDP, ip6_compute_pseudo); |
eb63f296 PA |
888 | |
889 | ret = udpv6_queue_rcv_skb(sk, skb); | |
890 | ||
84dad559 | 891 | /* a return value > 0 means to resubmit the input */ |
eb63f296 | 892 | if (ret > 0) |
84dad559 | 893 | return ret; |
eb63f296 PA |
894 | return 0; |
895 | } | |
896 | ||
645ca708 | 897 | int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
759e5d00 | 898 | int proto) |
1da177e4 | 899 | { |
ca065d0c | 900 | const struct in6_addr *saddr, *daddr; |
3ffe533c | 901 | struct net *net = dev_net(skb->dev); |
1ab1457c | 902 | struct udphdr *uh; |
ca065d0c | 903 | struct sock *sk; |
71489e21 | 904 | bool refcounted; |
1da177e4 LT |
905 | u32 ulen = 0; |
906 | ||
907 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
d6bc0149 | 908 | goto discard; |
1da177e4 | 909 | |
0660e03f ACM |
910 | saddr = &ipv6_hdr(skb)->saddr; |
911 | daddr = &ipv6_hdr(skb)->daddr; | |
4bedb452 | 912 | uh = udp_hdr(skb); |
1da177e4 LT |
913 | |
914 | ulen = ntohs(uh->len); | |
ba4e58ec GR |
915 | if (ulen > skb->len) |
916 | goto short_packet; | |
1da177e4 | 917 | |
759e5d00 HX |
918 | if (proto == IPPROTO_UDP) { |
919 | /* UDP validates ulen. */ | |
1da177e4 | 920 | |
ba4e58ec GR |
921 | /* Check for jumbo payload */ |
922 | if (ulen == 0) | |
923 | ulen = skb->len; | |
1da177e4 | 924 | |
ba4e58ec GR |
925 | if (ulen < sizeof(*uh)) |
926 | goto short_packet; | |
1da177e4 | 927 | |
ba4e58ec GR |
928 | if (ulen < skb->len) { |
929 | if (pskb_trim_rcsum(skb, ulen)) | |
930 | goto short_packet; | |
0660e03f ACM |
931 | saddr = &ipv6_hdr(skb)->saddr; |
932 | daddr = &ipv6_hdr(skb)->daddr; | |
4bedb452 | 933 | uh = udp_hdr(skb); |
ba4e58ec | 934 | } |
ba4e58ec | 935 | } |
1da177e4 | 936 | |
759e5d00 | 937 | if (udp6_csum_init(skb, uh, proto)) |
6a5dc9e5 | 938 | goto csum_error; |
759e5d00 | 939 | |
c9f2c1ae | 940 | /* Check if the socket is already available, e.g. due to early demux */ |
71489e21 | 941 | sk = skb_steal_sock(skb, &refcounted); |
c9f2c1ae PA |
942 | if (sk) { |
943 | struct dst_entry *dst = skb_dst(skb); | |
944 | int ret; | |
945 | ||
946 | if (unlikely(sk->sk_rx_dst != dst)) | |
64f0f5d1 | 947 | udp6_sk_rx_dst_set(sk, dst); |
c9f2c1ae | 948 | |
eb63f296 | 949 | if (!uh->check && !udp_sk(sk)->no_check6_rx) { |
71489e21 JS |
950 | if (refcounted) |
951 | sock_put(sk); | |
eb63f296 PA |
952 | goto report_csum_error; |
953 | } | |
c9f2c1ae | 954 | |
eb63f296 | 955 | ret = udp6_unicast_rcv_skb(sk, skb, uh); |
71489e21 JS |
956 | if (refcounted) |
957 | sock_put(sk); | |
eb63f296 | 958 | return ret; |
c9f2c1ae PA |
959 | } |
960 | ||
1ab1457c YH |
961 | /* |
962 | * Multicast receive code | |
1da177e4 | 963 | */ |
ba4e58ec | 964 | if (ipv6_addr_is_multicast(daddr)) |
e3163493 | 965 | return __udp6_lib_mcast_deliver(net, skb, |
36cbb245 | 966 | saddr, daddr, udptable, proto); |
1da177e4 LT |
967 | |
968 | /* Unicast */ | |
607c4aaf | 969 | sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable); |
53b24b8f | 970 | if (sk) { |
eb63f296 PA |
971 | if (!uh->check && !udp_sk(sk)->no_check6_rx) |
972 | goto report_csum_error; | |
973 | return udp6_unicast_rcv_skb(sk, skb, uh); | |
1da177e4 | 974 | } |
1ab1457c | 975 | |
eb63f296 PA |
976 | if (!uh->check) |
977 | goto report_csum_error; | |
4068579e | 978 | |
cb80ef46 | 979 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) |
c377411f | 980 | goto discard; |
cb80ef46 BL |
981 | |
982 | if (udp_lib_checksum_complete(skb)) | |
6a5dc9e5 | 983 | goto csum_error; |
cb80ef46 | 984 | |
02c22347 | 985 | __UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
cb80ef46 BL |
986 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
987 | ||
988 | kfree_skb(skb); | |
add459aa | 989 | return 0; |
1da177e4 | 990 | |
1ab1457c | 991 | short_packet: |
ba7a46f1 JP |
992 | net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n", |
993 | proto == IPPROTO_UDPLITE ? "-Lite" : "", | |
994 | saddr, ntohs(uh->source), | |
995 | ulen, skb->len, | |
996 | daddr, ntohs(uh->dest)); | |
6a5dc9e5 | 997 | goto discard; |
eb63f296 PA |
998 | |
999 | report_csum_error: | |
1000 | udp6_csum_zero_error(skb); | |
6a5dc9e5 | 1001 | csum_error: |
02c22347 | 1002 | __UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); |
1da177e4 | 1003 | discard: |
02c22347 | 1004 | __UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
1da177e4 | 1005 | kfree_skb(skb); |
add459aa | 1006 | return 0; |
1da177e4 | 1007 | } |
ba4e58ec | 1008 | |
0bd84065 | 1009 | |
5425077d | 1010 | static struct sock *__udp6_lib_demux_lookup(struct net *net, |
1011 | __be16 loc_port, const struct in6_addr *loc_addr, | |
1012 | __be16 rmt_port, const struct in6_addr *rmt_addr, | |
4297a0ef | 1013 | int dif, int sdif) |
5425077d | 1014 | { |
0bd84065 | 1015 | unsigned short hnum = ntohs(loc_port); |
f0b1e64c | 1016 | unsigned int hash2 = ipv6_portaddr_hash(net, loc_addr, hnum); |
0bd84065 | 1017 | unsigned int slot2 = hash2 & udp_table.mask; |
1018 | struct udp_hslot *hslot2 = &udp_table.hash2[slot2]; | |
1019 | const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum); | |
5425077d | 1020 | struct sock *sk; |
1021 | ||
0bd84065 | 1022 | udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
85cb73ff | 1023 | if (sk->sk_state == TCP_ESTABLISHED && |
4297a0ef | 1024 | INET6_MATCH(sk, net, rmt_addr, loc_addr, ports, dif, sdif)) |
0bd84065 | 1025 | return sk; |
1026 | /* Only check first socket in chain */ | |
1027 | break; | |
1028 | } | |
1029 | return NULL; | |
5425077d | 1030 | } |
1031 | ||
97ff7ffb | 1032 | INDIRECT_CALLABLE_SCOPE void udp_v6_early_demux(struct sk_buff *skb) |
5425077d | 1033 | { |
1034 | struct net *net = dev_net(skb->dev); | |
1035 | const struct udphdr *uh; | |
1036 | struct sock *sk; | |
1037 | struct dst_entry *dst; | |
1038 | int dif = skb->dev->ifindex; | |
4297a0ef | 1039 | int sdif = inet6_sdif(skb); |
5425077d | 1040 | |
1041 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + | |
1042 | sizeof(struct udphdr))) | |
1043 | return; | |
1044 | ||
1045 | uh = udp_hdr(skb); | |
1046 | ||
1047 | if (skb->pkt_type == PACKET_HOST) | |
1048 | sk = __udp6_lib_demux_lookup(net, uh->dest, | |
1049 | &ipv6_hdr(skb)->daddr, | |
1050 | uh->source, &ipv6_hdr(skb)->saddr, | |
4297a0ef | 1051 | dif, sdif); |
5425077d | 1052 | else |
1053 | return; | |
1054 | ||
41c6d650 | 1055 | if (!sk || !refcount_inc_not_zero(&sk->sk_refcnt)) |
5425077d | 1056 | return; |
1057 | ||
1058 | skb->sk = sk; | |
1059 | skb->destructor = sock_efree; | |
1060 | dst = READ_ONCE(sk->sk_rx_dst); | |
1061 | ||
1062 | if (dst) | |
1063 | dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie); | |
1064 | if (dst) { | |
d24406c8 WW |
1065 | /* set noref for now. |
1066 | * any place which wants to hold dst has to call | |
1067 | * dst_hold_safe() | |
1068 | */ | |
1069 | skb_dst_set_noref(skb, dst); | |
5425077d | 1070 | } |
1071 | } | |
1072 | ||
0e219ae4 | 1073 | INDIRECT_CALLABLE_SCOPE int udpv6_rcv(struct sk_buff *skb) |
ba4e58ec | 1074 | { |
645ca708 | 1075 | return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
ba4e58ec GR |
1076 | } |
1077 | ||
1da177e4 LT |
1078 | /* |
1079 | * Throw away all pending data and cancel the corking. Socket is locked. | |
1080 | */ | |
1081 | static void udp_v6_flush_pending_frames(struct sock *sk) | |
1082 | { | |
1083 | struct udp_sock *up = udp_sk(sk); | |
1084 | ||
36d926b9 DL |
1085 | if (up->pending == AF_INET) |
1086 | udp_flush_pending_frames(sk); | |
1087 | else if (up->pending) { | |
1da177e4 LT |
1088 | up->len = 0; |
1089 | up->pending = 0; | |
1090 | ip6_flush_pending_frames(sk); | |
1ab1457c | 1091 | } |
1da177e4 LT |
1092 | } |
1093 | ||
d74bad4e AI |
1094 | static int udpv6_pre_connect(struct sock *sk, struct sockaddr *uaddr, |
1095 | int addr_len) | |
1096 | { | |
bddc028a TH |
1097 | if (addr_len < offsetofend(struct sockaddr, sa_family)) |
1098 | return -EINVAL; | |
d74bad4e AI |
1099 | /* The following checks are replicated from __ip6_datagram_connect() |
1100 | * and intended to prevent BPF program called below from accessing | |
1101 | * bytes that are out of the bound specified by user in addr_len. | |
1102 | */ | |
1103 | if (uaddr->sa_family == AF_INET) { | |
1104 | if (__ipv6_only_sock(sk)) | |
1105 | return -EAFNOSUPPORT; | |
1106 | return udp_pre_connect(sk, uaddr, addr_len); | |
1107 | } | |
1108 | ||
1109 | if (addr_len < SIN6_LEN_RFC2133) | |
1110 | return -EINVAL; | |
1111 | ||
1112 | return BPF_CGROUP_RUN_PROG_INET6_CONNECT_LOCK(sk, uaddr); | |
1113 | } | |
1114 | ||
493c6be3 | 1115 | /** |
67ba4152 IM |
1116 | * udp6_hwcsum_outgoing - handle outgoing HW checksumming |
1117 | * @sk: socket we are sending on | |
1118 | * @skb: sk_buff containing the filled-in UDP header | |
1119 | * (checksum field must be zeroed out) | |
b51cd7c8 AL |
1120 | * @saddr: source address |
1121 | * @daddr: destination address | |
1122 | * @len: length of packet | |
493c6be3 SS |
1123 | */ |
1124 | static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb, | |
1125 | const struct in6_addr *saddr, | |
1126 | const struct in6_addr *daddr, int len) | |
1127 | { | |
1128 | unsigned int offset; | |
1129 | struct udphdr *uh = udp_hdr(skb); | |
d39d938c | 1130 | struct sk_buff *frags = skb_shinfo(skb)->frag_list; |
493c6be3 SS |
1131 | __wsum csum = 0; |
1132 | ||
d39d938c | 1133 | if (!frags) { |
493c6be3 SS |
1134 | /* Only one fragment on the socket. */ |
1135 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
1136 | skb->csum_offset = offsetof(struct udphdr, check); | |
1137 | uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0); | |
1138 | } else { | |
1139 | /* | |
1140 | * HW-checksum won't work as there are two or more | |
1141 | * fragments on the socket so that all csums of sk_buffs | |
1142 | * should be together | |
1143 | */ | |
1144 | offset = skb_transport_offset(skb); | |
1145 | skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); | |
63ecc3d9 | 1146 | csum = skb->csum; |
493c6be3 SS |
1147 | |
1148 | skb->ip_summed = CHECKSUM_NONE; | |
1149 | ||
d39d938c VY |
1150 | do { |
1151 | csum = csum_add(csum, frags->csum); | |
1152 | } while ((frags = frags->next)); | |
493c6be3 SS |
1153 | |
1154 | uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, | |
1155 | csum); | |
1156 | if (uh->check == 0) | |
1157 | uh->check = CSUM_MANGLED_0; | |
1158 | } | |
1159 | } | |
1160 | ||
1da177e4 LT |
1161 | /* |
1162 | * Sending | |
1163 | */ | |
1164 | ||
bec1f6f6 WB |
1165 | static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6, |
1166 | struct inet_cork *cork) | |
1da177e4 | 1167 | { |
d39d938c | 1168 | struct sock *sk = skb->sk; |
1da177e4 | 1169 | struct udphdr *uh; |
1da177e4 | 1170 | int err = 0; |
b2bf1e26 | 1171 | int is_udplite = IS_UDPLITE(sk); |
868c86bc | 1172 | __wsum csum = 0; |
d39d938c VY |
1173 | int offset = skb_transport_offset(skb); |
1174 | int len = skb->len - offset; | |
4094871d | 1175 | int datalen = len - sizeof(*uh); |
1da177e4 LT |
1176 | |
1177 | /* | |
1178 | * Create a UDP header | |
1179 | */ | |
4bedb452 | 1180 | uh = udp_hdr(skb); |
1958b856 DM |
1181 | uh->source = fl6->fl6_sport; |
1182 | uh->dest = fl6->fl6_dport; | |
d39d938c | 1183 | uh->len = htons(len); |
1da177e4 LT |
1184 | uh->check = 0; |
1185 | ||
bec1f6f6 WB |
1186 | if (cork->gso_size) { |
1187 | const int hlen = skb_network_header_len(skb) + | |
1188 | sizeof(struct udphdr); | |
1189 | ||
0f149c9f WB |
1190 | if (hlen + cork->gso_size > cork->fragsize) { |
1191 | kfree_skb(skb); | |
bec1f6f6 | 1192 | return -EINVAL; |
0f149c9f WB |
1193 | } |
1194 | if (skb->len > cork->gso_size * UDP_MAX_SEGMENTS) { | |
1195 | kfree_skb(skb); | |
bec1f6f6 | 1196 | return -EINVAL; |
0f149c9f WB |
1197 | } |
1198 | if (udp_sk(sk)->no_check6_tx) { | |
1199 | kfree_skb(skb); | |
a8c744a8 | 1200 | return -EINVAL; |
0f149c9f | 1201 | } |
ff06342c | 1202 | if (skb->ip_summed != CHECKSUM_PARTIAL || is_udplite || |
0f149c9f WB |
1203 | dst_xfrm(skb_dst(skb))) { |
1204 | kfree_skb(skb); | |
bec1f6f6 | 1205 | return -EIO; |
0f149c9f | 1206 | } |
bec1f6f6 | 1207 | |
4094871d JH |
1208 | if (datalen > cork->gso_size) { |
1209 | skb_shinfo(skb)->gso_size = cork->gso_size; | |
1210 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP_L4; | |
1211 | skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(datalen, | |
1212 | cork->gso_size); | |
1213 | } | |
a8c744a8 | 1214 | goto csum_partial; |
bec1f6f6 WB |
1215 | } |
1216 | ||
b2bf1e26 | 1217 | if (is_udplite) |
d39d938c VY |
1218 | csum = udplite_csum(skb); |
1219 | else if (udp_sk(sk)->no_check6_tx) { /* UDP csum disabled */ | |
4068579e TH |
1220 | skb->ip_summed = CHECKSUM_NONE; |
1221 | goto send; | |
1222 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
a8c744a8 | 1223 | csum_partial: |
d39d938c | 1224 | udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len); |
493c6be3 SS |
1225 | goto send; |
1226 | } else | |
d39d938c | 1227 | csum = udp_csum(skb); |
1da177e4 | 1228 | |
ba4e58ec | 1229 | /* add protocol-dependent pseudo-header */ |
4c9483b2 | 1230 | uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, |
d39d938c | 1231 | len, fl6->flowi6_proto, csum); |
1da177e4 | 1232 | if (uh->check == 0) |
f6ab0288 | 1233 | uh->check = CSUM_MANGLED_0; |
1da177e4 | 1234 | |
493c6be3 | 1235 | send: |
d39d938c | 1236 | err = ip6_send_skb(skb); |
6ce9e7b5 ED |
1237 | if (err) { |
1238 | if (err == -ENOBUFS && !inet6_sk(sk)->recverr) { | |
6aef70a8 ED |
1239 | UDP6_INC_STATS(sock_net(sk), |
1240 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
6ce9e7b5 ED |
1241 | err = 0; |
1242 | } | |
6aef70a8 ED |
1243 | } else { |
1244 | UDP6_INC_STATS(sock_net(sk), | |
1245 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
1246 | } | |
d39d938c VY |
1247 | return err; |
1248 | } | |
1249 | ||
1250 | static int udp_v6_push_pending_frames(struct sock *sk) | |
1251 | { | |
1252 | struct sk_buff *skb; | |
1253 | struct udp_sock *up = udp_sk(sk); | |
1254 | struct flowi6 fl6; | |
1255 | int err = 0; | |
1256 | ||
1257 | if (up->pending == AF_INET) | |
1258 | return udp_push_pending_frames(sk); | |
1259 | ||
1260 | /* ip6_finish_skb will release the cork, so make a copy of | |
1261 | * fl6 here. | |
1262 | */ | |
1263 | fl6 = inet_sk(sk)->cork.fl.u.ip6; | |
1264 | ||
1265 | skb = ip6_finish_skb(sk); | |
1266 | if (!skb) | |
1267 | goto out; | |
1268 | ||
bec1f6f6 | 1269 | err = udp_v6_send_skb(skb, &fl6, &inet_sk(sk)->cork.base); |
d39d938c | 1270 | |
1da177e4 LT |
1271 | out: |
1272 | up->len = 0; | |
1273 | up->pending = 0; | |
1274 | return err; | |
1275 | } | |
1276 | ||
1b784140 | 1277 | int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
1da177e4 LT |
1278 | { |
1279 | struct ipv6_txoptions opt_space; | |
1280 | struct udp_sock *up = udp_sk(sk); | |
1281 | struct inet_sock *inet = inet_sk(sk); | |
1282 | struct ipv6_pinfo *np = inet6_sk(sk); | |
342dfc30 | 1283 | DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
20c59de2 | 1284 | struct in6_addr *daddr, *final_p, final; |
1da177e4 | 1285 | struct ipv6_txoptions *opt = NULL; |
45f6fad8 | 1286 | struct ipv6_txoptions *opt_to_free = NULL; |
1da177e4 | 1287 | struct ip6_flowlabel *flowlabel = NULL; |
4c9483b2 | 1288 | struct flowi6 fl6; |
1da177e4 | 1289 | struct dst_entry *dst; |
26879da5 | 1290 | struct ipcm6_cookie ipc6; |
1da177e4 | 1291 | int addr_len = msg->msg_namelen; |
9f542f61 | 1292 | bool connected = false; |
1da177e4 | 1293 | int ulen = len; |
1da177e4 LT |
1294 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; |
1295 | int err; | |
b2bf1e26 | 1296 | int is_udplite = IS_UDPLITE(sk); |
ba4e58ec | 1297 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); |
1da177e4 | 1298 | |
b515430a | 1299 | ipcm6_init(&ipc6); |
bec1f6f6 | 1300 | ipc6.gso_size = up->gso_size; |
5fdaa88d | 1301 | ipc6.sockc.tsflags = sk->sk_tsflags; |
c6af0c22 | 1302 | ipc6.sockc.mark = sk->sk_mark; |
26879da5 | 1303 | |
1da177e4 LT |
1304 | /* destination address check */ |
1305 | if (sin6) { | |
1306 | if (addr_len < offsetof(struct sockaddr, sa_data)) | |
1307 | return -EINVAL; | |
1308 | ||
1309 | switch (sin6->sin6_family) { | |
1310 | case AF_INET6: | |
1311 | if (addr_len < SIN6_LEN_RFC2133) | |
1312 | return -EINVAL; | |
1313 | daddr = &sin6->sin6_addr; | |
052d2369 JL |
1314 | if (ipv6_addr_any(daddr) && |
1315 | ipv6_addr_v4mapped(&np->saddr)) | |
1316 | ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK), | |
1317 | daddr); | |
1da177e4 LT |
1318 | break; |
1319 | case AF_INET: | |
1320 | goto do_udp_sendmsg; | |
1321 | case AF_UNSPEC: | |
1322 | msg->msg_name = sin6 = NULL; | |
1323 | msg->msg_namelen = addr_len = 0; | |
1324 | daddr = NULL; | |
1325 | break; | |
1326 | default: | |
1327 | return -EINVAL; | |
1328 | } | |
1329 | } else if (!up->pending) { | |
1330 | if (sk->sk_state != TCP_ESTABLISHED) | |
1331 | return -EDESTADDRREQ; | |
efe4208f | 1332 | daddr = &sk->sk_v6_daddr; |
1ab1457c | 1333 | } else |
1da177e4 LT |
1334 | daddr = NULL; |
1335 | ||
1336 | if (daddr) { | |
e773e4fa | 1337 | if (ipv6_addr_v4mapped(daddr)) { |
1da177e4 LT |
1338 | struct sockaddr_in sin; |
1339 | sin.sin_family = AF_INET; | |
c720c7e8 | 1340 | sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport; |
1da177e4 LT |
1341 | sin.sin_addr.s_addr = daddr->s6_addr32[3]; |
1342 | msg->msg_name = &sin; | |
1343 | msg->msg_namelen = sizeof(sin); | |
1344 | do_udp_sendmsg: | |
1345 | if (__ipv6_only_sock(sk)) | |
1346 | return -ENETUNREACH; | |
1b784140 | 1347 | return udp_sendmsg(sk, msg, len); |
1da177e4 LT |
1348 | } |
1349 | } | |
1350 | ||
1351 | if (up->pending == AF_INET) | |
1b784140 | 1352 | return udp_sendmsg(sk, msg, len); |
1da177e4 LT |
1353 | |
1354 | /* Rough check on arithmetic overflow, | |
b59e139b | 1355 | better check is made in ip6_append_data(). |
1da177e4 LT |
1356 | */ |
1357 | if (len > INT_MAX - sizeof(struct udphdr)) | |
1358 | return -EMSGSIZE; | |
1ab1457c | 1359 | |
03485f2a | 1360 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
1da177e4 LT |
1361 | if (up->pending) { |
1362 | /* | |
1363 | * There are pending frames. | |
1364 | * The socket lock must be held while it's corked. | |
1365 | */ | |
1366 | lock_sock(sk); | |
1367 | if (likely(up->pending)) { | |
1368 | if (unlikely(up->pending != AF_INET6)) { | |
1369 | release_sock(sk); | |
1370 | return -EAFNOSUPPORT; | |
1371 | } | |
1372 | dst = NULL; | |
1373 | goto do_append_data; | |
1374 | } | |
1375 | release_sock(sk); | |
1376 | } | |
1377 | ulen += sizeof(struct udphdr); | |
1378 | ||
4c9483b2 | 1379 | memset(&fl6, 0, sizeof(fl6)); |
1da177e4 LT |
1380 | |
1381 | if (sin6) { | |
1382 | if (sin6->sin6_port == 0) | |
1383 | return -EINVAL; | |
1384 | ||
1958b856 | 1385 | fl6.fl6_dport = sin6->sin6_port; |
1da177e4 LT |
1386 | daddr = &sin6->sin6_addr; |
1387 | ||
1388 | if (np->sndflow) { | |
4c9483b2 DM |
1389 | fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; |
1390 | if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { | |
1391 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); | |
59c820b2 | 1392 | if (IS_ERR(flowlabel)) |
1da177e4 | 1393 | return -EINVAL; |
1da177e4 LT |
1394 | } |
1395 | } | |
1396 | ||
1397 | /* | |
1398 | * Otherwise it will be difficult to maintain | |
1399 | * sk->sk_dst_cache. | |
1400 | */ | |
1401 | if (sk->sk_state == TCP_ESTABLISHED && | |
efe4208f ED |
1402 | ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) |
1403 | daddr = &sk->sk_v6_daddr; | |
1da177e4 LT |
1404 | |
1405 | if (addr_len >= sizeof(struct sockaddr_in6) && | |
1406 | sin6->sin6_scope_id && | |
842df073 | 1407 | __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) |
4c9483b2 | 1408 | fl6.flowi6_oif = sin6->sin6_scope_id; |
1da177e4 LT |
1409 | } else { |
1410 | if (sk->sk_state != TCP_ESTABLISHED) | |
1411 | return -EDESTADDRREQ; | |
1412 | ||
1958b856 | 1413 | fl6.fl6_dport = inet->inet_dport; |
efe4208f | 1414 | daddr = &sk->sk_v6_daddr; |
4c9483b2 | 1415 | fl6.flowlabel = np->flow_label; |
9f542f61 | 1416 | connected = true; |
1da177e4 LT |
1417 | } |
1418 | ||
4c9483b2 DM |
1419 | if (!fl6.flowi6_oif) |
1420 | fl6.flowi6_oif = sk->sk_bound_dev_if; | |
1da177e4 | 1421 | |
4c9483b2 DM |
1422 | if (!fl6.flowi6_oif) |
1423 | fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex; | |
9f690db7 | 1424 | |
c6af0c22 | 1425 | fl6.flowi6_mark = ipc6.sockc.mark; |
e2d118a1 | 1426 | fl6.flowi6_uid = sk->sk_uid; |
51953d5b | 1427 | |
1da177e4 LT |
1428 | if (msg->msg_controllen) { |
1429 | opt = &opt_space; | |
1430 | memset(opt, 0, sizeof(struct ipv6_txoptions)); | |
1431 | opt->tot_len = sizeof(*opt); | |
26879da5 | 1432 | ipc6.opt = opt; |
1da177e4 | 1433 | |
2e8de857 WB |
1434 | err = udp_cmsg_send(sk, msg, &ipc6.gso_size); |
1435 | if (err > 0) | |
1436 | err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, | |
5fdaa88d | 1437 | &ipc6); |
1da177e4 LT |
1438 | if (err < 0) { |
1439 | fl6_sock_release(flowlabel); | |
1440 | return err; | |
1441 | } | |
4c9483b2 DM |
1442 | if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { |
1443 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); | |
59c820b2 | 1444 | if (IS_ERR(flowlabel)) |
1da177e4 LT |
1445 | return -EINVAL; |
1446 | } | |
1447 | if (!(opt->opt_nflen|opt->opt_flen)) | |
1448 | opt = NULL; | |
9f542f61 | 1449 | connected = false; |
1da177e4 | 1450 | } |
45f6fad8 ED |
1451 | if (!opt) { |
1452 | opt = txopt_get(np); | |
1453 | opt_to_free = opt; | |
1454 | } | |
df9890c3 YH |
1455 | if (flowlabel) |
1456 | opt = fl6_merge_options(&opt_space, flowlabel, opt); | |
1457 | opt = ipv6_fixup_options(&opt_space, opt); | |
26879da5 | 1458 | ipc6.opt = opt; |
1da177e4 | 1459 | |
4c9483b2 | 1460 | fl6.flowi6_proto = sk->sk_protocol; |
e8e36984 | 1461 | fl6.daddr = *daddr; |
4c9483b2 | 1462 | if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) |
4e3fd7a0 | 1463 | fl6.saddr = np->saddr; |
1958b856 | 1464 | fl6.fl6_sport = inet->inet_sport; |
1ab1457c | 1465 | |
1cedee13 AI |
1466 | if (cgroup_bpf_enabled && !connected) { |
1467 | err = BPF_CGROUP_RUN_PROG_UDP6_SENDMSG_LOCK(sk, | |
1468 | (struct sockaddr *)sin6, &fl6.saddr); | |
1469 | if (err) | |
1470 | goto out_no_dst; | |
1471 | if (sin6) { | |
1472 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) { | |
1473 | /* BPF program rewrote IPv6-only by IPv4-mapped | |
1474 | * IPv6. It's currently unsupported. | |
1475 | */ | |
1476 | err = -ENOTSUPP; | |
1477 | goto out_no_dst; | |
1478 | } | |
1479 | if (sin6->sin6_port == 0) { | |
1480 | /* BPF program set invalid port. Reject it. */ | |
1481 | err = -EINVAL; | |
1482 | goto out_no_dst; | |
1483 | } | |
1484 | fl6.fl6_dport = sin6->sin6_port; | |
1485 | fl6.daddr = sin6->sin6_addr; | |
1486 | } | |
1487 | } | |
1488 | ||
e8e36984 AI |
1489 | if (ipv6_addr_any(&fl6.daddr)) |
1490 | fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ | |
1491 | ||
4c9483b2 | 1492 | final_p = fl6_update_dst(&fl6, opt, &final); |
20c59de2 | 1493 | if (final_p) |
9f542f61 | 1494 | connected = false; |
1da177e4 | 1495 | |
4c9483b2 DM |
1496 | if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) { |
1497 | fl6.flowi6_oif = np->mcast_oif; | |
9f542f61 | 1498 | connected = false; |
c4062dfc EH |
1499 | } else if (!fl6.flowi6_oif) |
1500 | fl6.flowi6_oif = np->ucast_oif; | |
1da177e4 | 1501 | |
4c9483b2 | 1502 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); |
beb8d13b | 1503 | |
38b7097b HFS |
1504 | if (ipc6.tclass < 0) |
1505 | ipc6.tclass = np->tclass; | |
1506 | ||
1507 | fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel); | |
1508 | ||
4f858c56 | 1509 | dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, connected); |
68d0c6d3 DM |
1510 | if (IS_ERR(dst)) { |
1511 | err = PTR_ERR(dst); | |
1512 | dst = NULL; | |
1da177e4 | 1513 | goto out; |
14e50e57 | 1514 | } |
1da177e4 | 1515 | |
26879da5 WW |
1516 | if (ipc6.hlimit < 0) |
1517 | ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); | |
1da177e4 LT |
1518 | |
1519 | if (msg->msg_flags&MSG_CONFIRM) | |
1520 | goto do_confirm; | |
1521 | back_from_confirm: | |
1522 | ||
03485f2a VY |
1523 | /* Lockless fast path for the non-corking case */ |
1524 | if (!corkreq) { | |
1cd7884d | 1525 | struct inet_cork_full cork; |
03485f2a VY |
1526 | struct sk_buff *skb; |
1527 | ||
1528 | skb = ip6_make_skb(sk, getfrag, msg, ulen, | |
26879da5 | 1529 | sizeof(struct udphdr), &ipc6, |
03485f2a | 1530 | &fl6, (struct rt6_info *)dst, |
5fdaa88d | 1531 | msg->msg_flags, &cork); |
03485f2a VY |
1532 | err = PTR_ERR(skb); |
1533 | if (!IS_ERR_OR_NULL(skb)) | |
bec1f6f6 | 1534 | err = udp_v6_send_skb(skb, &fl6, &cork.base); |
4f858c56 | 1535 | goto out; |
03485f2a VY |
1536 | } |
1537 | ||
1da177e4 LT |
1538 | lock_sock(sk); |
1539 | if (unlikely(up->pending)) { | |
1540 | /* The socket is already corked while preparing it. */ | |
1541 | /* ... which is an evident application bug. --ANK */ | |
1542 | release_sock(sk); | |
1543 | ||
ba7a46f1 | 1544 | net_dbg_ratelimited("udp cork app bug 2\n"); |
1da177e4 LT |
1545 | err = -EINVAL; |
1546 | goto out; | |
1547 | } | |
1548 | ||
1549 | up->pending = AF_INET6; | |
1550 | ||
1551 | do_append_data: | |
26879da5 WW |
1552 | if (ipc6.dontfrag < 0) |
1553 | ipc6.dontfrag = np->dontfrag; | |
1da177e4 | 1554 | up->len += ulen; |
26879da5 WW |
1555 | err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr), |
1556 | &ipc6, &fl6, (struct rt6_info *)dst, | |
5fdaa88d | 1557 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); |
1da177e4 LT |
1558 | if (err) |
1559 | udp_v6_flush_pending_frames(sk); | |
1560 | else if (!corkreq) | |
4c0a6cb0 | 1561 | err = udp_v6_push_pending_frames(sk); |
1e0c14f4 HX |
1562 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) |
1563 | up->pending = 0; | |
1da177e4 | 1564 | |
03485f2a VY |
1565 | if (err > 0) |
1566 | err = np->recverr ? net_xmit_errno(err) : 0; | |
1567 | release_sock(sk); | |
1568 | ||
1da177e4 | 1569 | out: |
a3c96089 | 1570 | dst_release(dst); |
1cedee13 | 1571 | out_no_dst: |
1da177e4 | 1572 | fl6_sock_release(flowlabel); |
45f6fad8 | 1573 | txopt_put(opt_to_free); |
cd562c98 | 1574 | if (!err) |
1da177e4 | 1575 | return len; |
a18135eb DM |
1576 | /* |
1577 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1578 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1579 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1580 | * things). We could add another new stat but at least for now that | |
1581 | * seems like overkill. | |
1582 | */ | |
1583 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
6aef70a8 ED |
1584 | UDP6_INC_STATS(sock_net(sk), |
1585 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
a18135eb | 1586 | } |
1da177e4 LT |
1587 | return err; |
1588 | ||
1589 | do_confirm: | |
0dec879f JA |
1590 | if (msg->msg_flags & MSG_PROBE) |
1591 | dst_confirm_neigh(dst, &fl6.daddr); | |
1da177e4 LT |
1592 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1593 | goto back_from_confirm; | |
1594 | err = 0; | |
1595 | goto out; | |
1596 | } | |
1597 | ||
7d06b2e0 | 1598 | void udpv6_destroy_sock(struct sock *sk) |
1da177e4 | 1599 | { |
44046a59 | 1600 | struct udp_sock *up = udp_sk(sk); |
1da177e4 LT |
1601 | lock_sock(sk); |
1602 | udp_v6_flush_pending_frames(sk); | |
1603 | release_sock(sk); | |
1604 | ||
60fb9567 PA |
1605 | if (static_branch_unlikely(&udpv6_encap_needed_key)) { |
1606 | if (up->encap_type) { | |
1607 | void (*encap_destroy)(struct sock *sk); | |
1608 | encap_destroy = READ_ONCE(up->encap_destroy); | |
1609 | if (encap_destroy) | |
1610 | encap_destroy(sk); | |
1611 | } | |
1612 | if (up->encap_enabled) | |
9c480601 | 1613 | static_branch_dec(&udpv6_encap_needed_key); |
44046a59 TP |
1614 | } |
1615 | ||
1da177e4 | 1616 | inet6_destroy_sock(sk); |
1da177e4 LT |
1617 | } |
1618 | ||
1619 | /* | |
1620 | * Socket option code for UDP | |
1621 | */ | |
a7b75c5a CH |
1622 | int udpv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval, |
1623 | unsigned int optlen) | |
3fdadf7d | 1624 | { |
db8dac20 | 1625 | if (level == SOL_UDP || level == SOL_UDPLITE) |
91ac1cca | 1626 | return udp_lib_setsockopt(sk, level, optname, |
a7b75c5a | 1627 | optval, optlen, |
4c0a6cb0 | 1628 | udp_v6_push_pending_frames); |
ba4e58ec | 1629 | return ipv6_setsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1630 | } |
1631 | ||
ba4e58ec GR |
1632 | int udpv6_getsockopt(struct sock *sk, int level, int optname, |
1633 | char __user *optval, int __user *optlen) | |
3fdadf7d | 1634 | { |
db8dac20 | 1635 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 | 1636 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); |
ba4e58ec | 1637 | return ipv6_getsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1638 | } |
1639 | ||
a8e3bb34 DA |
1640 | /* thinking of making this const? Don't. |
1641 | * early_demux can change based on sysctl. | |
1642 | */ | |
dddb64bc | 1643 | static struct inet6_protocol udpv6_protocol = { |
5425077d | 1644 | .early_demux = udp_v6_early_demux, |
dddb64bc | 1645 | .early_demux_handler = udp_v6_early_demux, |
1da177e4 LT |
1646 | .handler = udpv6_rcv, |
1647 | .err_handler = udpv6_err, | |
1648 | .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, | |
1649 | }; | |
1650 | ||
1651 | /* ------------------------------------------------------------------------ */ | |
1652 | #ifdef CONFIG_PROC_FS | |
ba4e58ec | 1653 | int udp6_seq_show(struct seq_file *seq, void *v) |
1da177e4 | 1654 | { |
17ef66af LC |
1655 | if (v == SEQ_START_TOKEN) { |
1656 | seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); | |
1657 | } else { | |
1658 | int bucket = ((struct udp_iter_state *)seq->private)->bucket; | |
1659 | struct inet_sock *inet = inet_sk(v); | |
1660 | __u16 srcp = ntohs(inet->inet_sport); | |
1661 | __u16 destp = ntohs(inet->inet_dport); | |
6c206b20 PA |
1662 | __ip6_dgram_sock_seq_show(seq, v, srcp, destp, |
1663 | udp_rqueue_get(v), bucket); | |
17ef66af | 1664 | } |
1da177e4 LT |
1665 | return 0; |
1666 | } | |
1667 | ||
c3506372 | 1668 | const struct seq_operations udp6_seq_ops = { |
a3d2599b CH |
1669 | .start = udp_seq_start, |
1670 | .next = udp_seq_next, | |
1671 | .stop = udp_seq_stop, | |
1672 | .show = udp6_seq_show, | |
1673 | }; | |
c3506372 | 1674 | EXPORT_SYMBOL(udp6_seq_ops); |
73cb88ec | 1675 | |
1da177e4 | 1676 | static struct udp_seq_afinfo udp6_seq_afinfo = { |
1da177e4 | 1677 | .family = AF_INET6, |
645ca708 | 1678 | .udp_table = &udp_table, |
1da177e4 LT |
1679 | }; |
1680 | ||
2c8c1e72 | 1681 | int __net_init udp6_proc_init(struct net *net) |
1da177e4 | 1682 | { |
c3506372 CH |
1683 | if (!proc_create_net_data("udp6", 0444, net->proc_net, &udp6_seq_ops, |
1684 | sizeof(struct udp_iter_state), &udp6_seq_afinfo)) | |
a3d2599b CH |
1685 | return -ENOMEM; |
1686 | return 0; | |
1da177e4 LT |
1687 | } |
1688 | ||
ec120da6 IM |
1689 | void udp6_proc_exit(struct net *net) |
1690 | { | |
a3d2599b | 1691 | remove_proc_entry("udp6", net->proc_net); |
1da177e4 LT |
1692 | } |
1693 | #endif /* CONFIG_PROC_FS */ | |
1694 | ||
1695 | /* ------------------------------------------------------------------------ */ | |
1696 | ||
1697 | struct proto udpv6_prot = { | |
1e802951 TZ |
1698 | .name = "UDPv6", |
1699 | .owner = THIS_MODULE, | |
1700 | .close = udp_lib_close, | |
d74bad4e | 1701 | .pre_connect = udpv6_pre_connect, |
1e802951 TZ |
1702 | .connect = ip6_datagram_connect, |
1703 | .disconnect = udp_disconnect, | |
1704 | .ioctl = udp_ioctl, | |
1705 | .init = udp_init_sock, | |
1706 | .destroy = udpv6_destroy_sock, | |
1707 | .setsockopt = udpv6_setsockopt, | |
1708 | .getsockopt = udpv6_getsockopt, | |
1709 | .sendmsg = udpv6_sendmsg, | |
1710 | .recvmsg = udpv6_recvmsg, | |
1711 | .release_cb = ip6_datagram_release_cb, | |
1712 | .hash = udp_lib_hash, | |
1713 | .unhash = udp_lib_unhash, | |
1714 | .rehash = udp_v6_rehash, | |
1715 | .get_port = udp_v6_get_port, | |
1716 | .memory_allocated = &udp_memory_allocated, | |
1717 | .sysctl_mem = sysctl_udp_mem, | |
1718 | .sysctl_wmem_offset = offsetof(struct net, ipv4.sysctl_udp_wmem_min), | |
1719 | .sysctl_rmem_offset = offsetof(struct net, ipv4.sysctl_udp_rmem_min), | |
1720 | .obj_size = sizeof(struct udp6_sock), | |
1721 | .h.udp_table = &udp_table, | |
1e802951 | 1722 | .diag_destroy = udp_abort, |
1da177e4 LT |
1723 | }; |
1724 | ||
1da177e4 LT |
1725 | static struct inet_protosw udpv6_protosw = { |
1726 | .type = SOCK_DGRAM, | |
1727 | .protocol = IPPROTO_UDP, | |
1728 | .prot = &udpv6_prot, | |
1729 | .ops = &inet6_dgram_ops, | |
1da177e4 LT |
1730 | .flags = INET_PROTOSW_PERMANENT, |
1731 | }; | |
1732 | ||
7f4e4868 | 1733 | int __init udpv6_init(void) |
1da177e4 | 1734 | { |
7f4e4868 DL |
1735 | int ret; |
1736 | ||
3336288a VY |
1737 | ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP); |
1738 | if (ret) | |
c6b641a4 | 1739 | goto out; |
3336288a | 1740 | |
7f4e4868 DL |
1741 | ret = inet6_register_protosw(&udpv6_protosw); |
1742 | if (ret) | |
1743 | goto out_udpv6_protocol; | |
1744 | out: | |
1745 | return ret; | |
1746 | ||
1747 | out_udpv6_protocol: | |
1748 | inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); | |
1749 | goto out; | |
1750 | } | |
1751 | ||
09f7709f | 1752 | void udpv6_exit(void) |
7f4e4868 DL |
1753 | { |
1754 | inet6_unregister_protosw(&udpv6_protosw); | |
1755 | inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP); | |
1da177e4 | 1756 | } |