| 1 |
// SPDX-License-Identifier: GPL-2.0-or-later |
| 2 |
/* |
| 3 |
* UDP over IPv6 |
| 4 |
* Linux INET6 implementation |
| 5 |
* |
| 6 |
* Authors: |
| 7 |
* Pedro Roque <[email protected]> |
| 8 |
* |
| 9 |
* Based on linux/ipv4/udp.c |
| 10 |
* |
| 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. |
| 18 |
*/ |
| 19 |
|
| 20 |
#include <linux/bpf-cgroup.h> |
| 21 |
#include <linux/errno.h> |
| 22 |
#include <linux/types.h> |
| 23 |
#include <linux/socket.h> |
| 24 |
#include <linux/sockios.h> |
| 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> |
| 32 |
#include <linux/module.h> |
| 33 |
#include <linux/skbuff.h> |
| 34 |
#include <linux/slab.h> |
| 35 |
#include <linux/uaccess.h> |
| 36 |
#include <linux/indirect_call_wrapper.h> |
| 37 |
#include <trace/events/udp.h> |
| 38 |
|
| 39 |
#include <net/addrconf.h> |
| 40 |
#include <net/ndisc.h> |
| 41 |
#include <net/protocol.h> |
| 42 |
#include <net/transp_v6.h> |
| 43 |
#include <net/ip6_route.h> |
| 44 |
#include <net/raw.h> |
| 45 |
#include <net/seg6.h> |
| 46 |
#include <net/tcp_states.h> |
| 47 |
#include <net/ip6_checksum.h> |
| 48 |
#include <net/ip6_tunnel.h> |
| 49 |
#include <net/udp_tunnel.h> |
| 50 |
#include <net/xfrm.h> |
| 51 |
#include <net/inet_hashtables.h> |
| 52 |
#include <net/inet6_hashtables.h> |
| 53 |
#include <net/busy_poll.h> |
| 54 |
#include <net/sock_reuseport.h> |
| 55 |
#include <net/gro.h> |
| 56 |
|
| 57 |
#include <linux/proc_fs.h> |
| 58 |
#include <linux/seq_file.h> |
| 59 |
#include <trace/events/skb.h> |
| 60 |
#include "udp_impl.h" |
| 61 |
|
| 62 |
static void udpv6_destruct_sock(struct sock *sk) |
| 63 |
{ |
| 64 |
udp_destruct_common(sk); |
| 65 |
inet6_sock_destruct(sk); |
| 66 |
} |
| 67 |
|
| 68 |
int udpv6_init_sock(struct sock *sk) |
| 69 |
{ |
| 70 |
int res = udp_lib_init_sock(sk); |
| 71 |
|
| 72 |
sk->sk_destruct = udpv6_destruct_sock; |
| 73 |
set_bit(SOCK_SUPPORT_ZC, &sk->sk_socket->flags); |
| 74 |
return res; |
| 75 |
} |
| 76 |
|
| 77 |
INDIRECT_CALLABLE_SCOPE |
| 78 |
u32 udp6_ehashfn(const struct net *net, |
| 79 |
const struct in6_addr *laddr, |
| 80 |
const u16 lport, |
| 81 |
const struct in6_addr *faddr, |
| 82 |
const __be16 fport) |
| 83 |
{ |
| 84 |
u32 lhash, fhash; |
| 85 |
|
| 86 |
net_get_random_once(&udp6_ehash_secret, |
| 87 |
sizeof(udp6_ehash_secret)); |
| 88 |
net_get_random_once(&udp_ipv6_hash_secret, |
| 89 |
sizeof(udp_ipv6_hash_secret)); |
| 90 |
|
| 91 |
lhash = (__force u32)laddr->s6_addr32[3]; |
| 92 |
fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret); |
| 93 |
|
| 94 |
return __inet6_ehashfn(lhash, lport, fhash, fport, |
| 95 |
udp6_ehash_secret + net_hash_mix(net)); |
| 96 |
} |
| 97 |
|
| 98 |
int udp_v6_get_port(struct sock *sk, unsigned short snum) |
| 99 |
{ |
| 100 |
unsigned int hash2_nulladdr = |
| 101 |
ipv6_portaddr_hash(sock_net(sk), &in6addr_any, snum); |
| 102 |
unsigned int hash2_partial = |
| 103 |
ipv6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0); |
| 104 |
|
| 105 |
/* precompute partial secondary hash */ |
| 106 |
udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
| 107 |
return udp_lib_get_port(sk, snum, hash2_nulladdr); |
| 108 |
} |
| 109 |
|
| 110 |
void udp_v6_rehash(struct sock *sk) |
| 111 |
{ |
| 112 |
u16 new_hash = ipv6_portaddr_hash(sock_net(sk), |
| 113 |
&sk->sk_v6_rcv_saddr, |
| 114 |
inet_sk(sk)->inet_num); |
| 115 |
u16 new_hash4; |
| 116 |
|
| 117 |
if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr)) { |
| 118 |
new_hash4 = udp_ehashfn(sock_net(sk), |
| 119 |
sk->sk_rcv_saddr, sk->sk_num, |
| 120 |
sk->sk_daddr, sk->sk_dport); |
| 121 |
} else { |
| 122 |
new_hash4 = udp6_ehashfn(sock_net(sk), |
| 123 |
&sk->sk_v6_rcv_saddr, sk->sk_num, |
| 124 |
&sk->sk_v6_daddr, sk->sk_dport); |
| 125 |
} |
| 126 |
|
| 127 |
udp_lib_rehash(sk, new_hash, new_hash4); |
| 128 |
} |
| 129 |
|
| 130 |
static int compute_score(struct sock *sk, const struct net *net, |
| 131 |
const struct in6_addr *saddr, __be16 sport, |
| 132 |
const struct in6_addr *daddr, unsigned short hnum, |
| 133 |
int dif, int sdif) |
| 134 |
{ |
| 135 |
int bound_dev_if, score; |
| 136 |
struct inet_sock *inet; |
| 137 |
bool dev_match; |
| 138 |
|
| 139 |
if (!net_eq(sock_net(sk), net) || |
| 140 |
udp_sk(sk)->udp_port_hash != hnum || |
| 141 |
sk->sk_family != PF_INET6) |
| 142 |
return -1; |
| 143 |
|
| 144 |
if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr)) |
| 145 |
return -1; |
| 146 |
|
| 147 |
score = 0; |
| 148 |
inet = inet_sk(sk); |
| 149 |
|
| 150 |
if (inet->inet_dport) { |
| 151 |
if (inet->inet_dport != sport) |
| 152 |
return -1; |
| 153 |
score++; |
| 154 |
} |
| 155 |
|
| 156 |
if (!ipv6_addr_any(&sk->sk_v6_daddr)) { |
| 157 |
if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr)) |
| 158 |
return -1; |
| 159 |
score++; |
| 160 |
} |
| 161 |
|
| 162 |
bound_dev_if = READ_ONCE(sk->sk_bound_dev_if); |
| 163 |
dev_match = udp_sk_bound_dev_eq(net, bound_dev_if, dif, sdif); |
| 164 |
if (!dev_match) |
| 165 |
return -1; |
| 166 |
if (bound_dev_if) |
| 167 |
score++; |
| 168 |
|
| 169 |
if (READ_ONCE(sk->sk_incoming_cpu) == raw_smp_processor_id()) |
| 170 |
score++; |
| 171 |
|
| 172 |
return score; |
| 173 |
} |
| 174 |
|
| 175 |
/** |
| 176 |
* udp6_lib_lookup1() - Simplified lookup using primary hash (destination port) |
| 177 |
* @net: Network namespace |
| 178 |
* @saddr: Source address, network order |
| 179 |
* @sport: Source port, network order |
| 180 |
* @daddr: Destination address, network order |
| 181 |
* @hnum: Destination port, host order |
| 182 |
* @dif: Destination interface index |
| 183 |
* @sdif: Destination bridge port index, if relevant |
| 184 |
* @udptable: Set of UDP hash tables |
| 185 |
* |
| 186 |
* Simplified lookup to be used as fallback if no sockets are found due to a |
| 187 |
* potential race between (receive) address change, and lookup happening before |
| 188 |
* the rehash operation. This function ignores SO_REUSEPORT groups while scoring |
| 189 |
* result sockets, because if we have one, we don't need the fallback at all. |
| 190 |
* |
| 191 |
* Called under rcu_read_lock(). |
| 192 |
* |
| 193 |
* Return: socket with highest matching score if any, NULL if none |
| 194 |
*/ |
| 195 |
static struct sock *udp6_lib_lookup1(const struct net *net, |
| 196 |
const struct in6_addr *saddr, __be16 sport, |
| 197 |
const struct in6_addr *daddr, |
| 198 |
unsigned int hnum, int dif, int sdif, |
| 199 |
const struct udp_table *udptable) |
| 200 |
{ |
| 201 |
unsigned int slot = udp_hashfn(net, hnum, udptable->mask); |
| 202 |
struct udp_hslot *hslot = &udptable->hash[slot]; |
| 203 |
struct sock *sk, *result = NULL; |
| 204 |
int score, badness = 0; |
| 205 |
|
| 206 |
sk_for_each_rcu(sk, &hslot->head) { |
| 207 |
score = compute_score(sk, net, |
| 208 |
saddr, sport, daddr, hnum, dif, sdif); |
| 209 |
if (score > badness) { |
| 210 |
result = sk; |
| 211 |
badness = score; |
| 212 |
} |
| 213 |
} |
| 214 |
|
| 215 |
return result; |
| 216 |
} |
| 217 |
|
| 218 |
/* called with rcu_read_lock() */ |
| 219 |
static struct sock *udp6_lib_lookup2(const struct net *net, |
| 220 |
const struct in6_addr *saddr, __be16 sport, |
| 221 |
const struct in6_addr *daddr, unsigned int hnum, |
| 222 |
int dif, int sdif, struct udp_hslot *hslot2, |
| 223 |
struct sk_buff *skb) |
| 224 |
{ |
| 225 |
struct sock *sk, *result; |
| 226 |
int score, badness; |
| 227 |
bool need_rescore; |
| 228 |
|
| 229 |
result = NULL; |
| 230 |
badness = -1; |
| 231 |
udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
| 232 |
need_rescore = false; |
| 233 |
rescore: |
| 234 |
score = compute_score(need_rescore ? result : sk, net, saddr, |
| 235 |
sport, daddr, hnum, dif, sdif); |
| 236 |
if (score > badness) { |
| 237 |
badness = score; |
| 238 |
|
| 239 |
if (need_rescore) |
| 240 |
continue; |
| 241 |
|
| 242 |
if (sk->sk_state == TCP_ESTABLISHED) { |
| 243 |
result = sk; |
| 244 |
continue; |
| 245 |
} |
| 246 |
|
| 247 |
result = inet6_lookup_reuseport(net, sk, skb, sizeof(struct udphdr), |
| 248 |
saddr, sport, daddr, hnum, udp6_ehashfn); |
| 249 |
if (!result) { |
| 250 |
result = sk; |
| 251 |
continue; |
| 252 |
} |
| 253 |
|
| 254 |
/* Fall back to scoring if group has connections */ |
| 255 |
if (!reuseport_has_conns(sk)) |
| 256 |
return result; |
| 257 |
|
| 258 |
/* Reuseport logic returned an error, keep original score. */ |
| 259 |
if (IS_ERR(result)) |
| 260 |
continue; |
| 261 |
|
| 262 |
/* compute_score is too long of a function to be |
| 263 |
* inlined, and calling it again here yields |
| 264 |
* measurable overhead for some |
| 265 |
* workloads. Work around it by jumping |
| 266 |
* backwards to rescore 'result'. |
| 267 |
*/ |
| 268 |
need_rescore = true; |
| 269 |
goto rescore; |
| 270 |
} |
| 271 |
} |
| 272 |
return result; |
| 273 |
} |
| 274 |
|
| 275 |
#if IS_ENABLED(CONFIG_BASE_SMALL) |
| 276 |
static struct sock *udp6_lib_lookup4(const struct net *net, |
| 277 |
const struct in6_addr *saddr, __be16 sport, |
| 278 |
const struct in6_addr *daddr, |
| 279 |
unsigned int hnum, int dif, int sdif, |
| 280 |
struct udp_table *udptable) |
| 281 |
{ |
| 282 |
return NULL; |
| 283 |
} |
| 284 |
|
| 285 |
static void udp6_hash4(struct sock *sk) |
| 286 |
{ |
| 287 |
} |
| 288 |
#else /* !CONFIG_BASE_SMALL */ |
| 289 |
static struct sock *udp6_lib_lookup4(const struct net *net, |
| 290 |
const struct in6_addr *saddr, __be16 sport, |
| 291 |
const struct in6_addr *daddr, |
| 292 |
unsigned int hnum, int dif, int sdif, |
| 293 |
struct udp_table *udptable) |
| 294 |
{ |
| 295 |
const __portpair ports = INET_COMBINED_PORTS(sport, hnum); |
| 296 |
const struct hlist_nulls_node *node; |
| 297 |
struct udp_hslot *hslot4; |
| 298 |
unsigned int hash4, slot; |
| 299 |
struct udp_sock *up; |
| 300 |
struct sock *sk; |
| 301 |
|
| 302 |
hash4 = udp6_ehashfn(net, daddr, hnum, saddr, sport); |
| 303 |
slot = hash4 & udptable->mask; |
| 304 |
hslot4 = &udptable->hash4[slot]; |
| 305 |
|
| 306 |
begin: |
| 307 |
udp_lrpa_for_each_entry_rcu(up, node, &hslot4->nulls_head) { |
| 308 |
sk = (struct sock *)up; |
| 309 |
if (inet6_match(net, sk, saddr, daddr, ports, dif, sdif)) |
| 310 |
return sk; |
| 311 |
} |
| 312 |
|
| 313 |
/* if the nulls value we got at the end of this lookup is not the |
| 314 |
* expected one, we must restart lookup. We probably met an item that |
| 315 |
* was moved to another chain due to rehash. |
| 316 |
*/ |
| 317 |
if (get_nulls_value(node) != slot) |
| 318 |
goto begin; |
| 319 |
|
| 320 |
return NULL; |
| 321 |
} |
| 322 |
|
| 323 |
static void udp6_hash4(struct sock *sk) |
| 324 |
{ |
| 325 |
struct net *net = sock_net(sk); |
| 326 |
unsigned int hash; |
| 327 |
|
| 328 |
if (ipv6_addr_v4mapped(&sk->sk_v6_rcv_saddr)) { |
| 329 |
udp4_hash4(sk); |
| 330 |
return; |
| 331 |
} |
| 332 |
|
| 333 |
if (sk_unhashed(sk) || ipv6_addr_any(&sk->sk_v6_rcv_saddr)) |
| 334 |
return; |
| 335 |
|
| 336 |
hash = udp6_ehashfn(net, &sk->sk_v6_rcv_saddr, sk->sk_num, |
| 337 |
&sk->sk_v6_daddr, sk->sk_dport); |
| 338 |
|
| 339 |
udp_lib_hash4(sk, hash); |
| 340 |
} |
| 341 |
#endif /* CONFIG_BASE_SMALL */ |
| 342 |
|
| 343 |
/* rcu_read_lock() must be held */ |
| 344 |
struct sock *__udp6_lib_lookup(const struct net *net, |
| 345 |
const struct in6_addr *saddr, __be16 sport, |
| 346 |
const struct in6_addr *daddr, __be16 dport, |
| 347 |
int dif, int sdif, struct udp_table *udptable, |
| 348 |
struct sk_buff *skb) |
| 349 |
{ |
| 350 |
unsigned short hnum = ntohs(dport); |
| 351 |
struct udp_hslot *hslot2; |
| 352 |
struct sock *result, *sk; |
| 353 |
unsigned int hash2; |
| 354 |
|
| 355 |
hash2 = ipv6_portaddr_hash(net, daddr, hnum); |
| 356 |
hslot2 = udp_hashslot2(udptable, hash2); |
| 357 |
|
| 358 |
if (udp_has_hash4(hslot2)) { |
| 359 |
result = udp6_lib_lookup4(net, saddr, sport, daddr, hnum, |
| 360 |
dif, sdif, udptable); |
| 361 |
if (result) /* udp6_lib_lookup4 return sk or NULL */ |
| 362 |
return result; |
| 363 |
} |
| 364 |
|
| 365 |
/* Lookup connected or non-wildcard sockets */ |
| 366 |
result = udp6_lib_lookup2(net, saddr, sport, |
| 367 |
daddr, hnum, dif, sdif, |
| 368 |
hslot2, skb); |
| 369 |
if (!IS_ERR_OR_NULL(result) && result->sk_state == TCP_ESTABLISHED) |
| 370 |
goto done; |
| 371 |
|
| 372 |
/* Lookup redirect from BPF */ |
| 373 |
if (static_branch_unlikely(&bpf_sk_lookup_enabled) && |
| 374 |
udptable == net->ipv4.udp_table) { |
| 375 |
sk = inet6_lookup_run_sk_lookup(net, IPPROTO_UDP, skb, sizeof(struct udphdr), |
| 376 |
saddr, sport, daddr, hnum, dif, |
| 377 |
udp6_ehashfn); |
| 378 |
if (sk) { |
| 379 |
result = sk; |
| 380 |
goto done; |
| 381 |
} |
| 382 |
} |
| 383 |
|
| 384 |
/* Got non-wildcard socket or error on first lookup */ |
| 385 |
if (result) |
| 386 |
goto done; |
| 387 |
|
| 388 |
/* Lookup wildcard sockets */ |
| 389 |
hash2 = ipv6_portaddr_hash(net, &in6addr_any, hnum); |
| 390 |
hslot2 = udp_hashslot2(udptable, hash2); |
| 391 |
|
| 392 |
result = udp6_lib_lookup2(net, saddr, sport, |
| 393 |
&in6addr_any, hnum, dif, sdif, |
| 394 |
hslot2, skb); |
| 395 |
if (!IS_ERR_OR_NULL(result)) |
| 396 |
goto done; |
| 397 |
|
| 398 |
/* Cover address change/lookup/rehash race: see __udp4_lib_lookup() */ |
| 399 |
result = udp6_lib_lookup1(net, saddr, sport, daddr, hnum, dif, sdif, |
| 400 |
udptable); |
| 401 |
|
| 402 |
done: |
| 403 |
if (IS_ERR(result)) |
| 404 |
return NULL; |
| 405 |
return result; |
| 406 |
} |
| 407 |
EXPORT_SYMBOL_GPL(__udp6_lib_lookup); |
| 408 |
|
| 409 |
static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb, |
| 410 |
__be16 sport, __be16 dport, |
| 411 |
struct udp_table *udptable) |
| 412 |
{ |
| 413 |
const struct ipv6hdr *iph = ipv6_hdr(skb); |
| 414 |
|
| 415 |
return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport, |
| 416 |
&iph->daddr, dport, inet6_iif(skb), |
| 417 |
inet6_sdif(skb), udptable, skb); |
| 418 |
} |
| 419 |
|
| 420 |
struct sock *udp6_lib_lookup_skb(const struct sk_buff *skb, |
| 421 |
__be16 sport, __be16 dport) |
| 422 |
{ |
| 423 |
const u16 offset = NAPI_GRO_CB(skb)->network_offsets[skb->encapsulation]; |
| 424 |
const struct ipv6hdr *iph = (struct ipv6hdr *)(skb->data + offset); |
| 425 |
struct net *net = dev_net(skb->dev); |
| 426 |
int iif, sdif; |
| 427 |
|
| 428 |
inet6_get_iif_sdif(skb, &iif, &sdif); |
| 429 |
|
| 430 |
return __udp6_lib_lookup(net, &iph->saddr, sport, |
| 431 |
&iph->daddr, dport, iif, |
| 432 |
sdif, net->ipv4.udp_table, NULL); |
| 433 |
} |
| 434 |
|
| 435 |
/* Must be called under rcu_read_lock(). |
| 436 |
* Does increment socket refcount. |
| 437 |
*/ |
| 438 |
#if IS_ENABLED(CONFIG_NF_TPROXY_IPV6) || IS_ENABLED(CONFIG_NF_SOCKET_IPV6) |
| 439 |
struct sock *udp6_lib_lookup(const struct net *net, const struct in6_addr *saddr, __be16 sport, |
| 440 |
const struct in6_addr *daddr, __be16 dport, int dif) |
| 441 |
{ |
| 442 |
struct sock *sk; |
| 443 |
|
| 444 |
sk = __udp6_lib_lookup(net, saddr, sport, daddr, dport, |
| 445 |
dif, 0, net->ipv4.udp_table, NULL); |
| 446 |
if (sk && !refcount_inc_not_zero(&sk->sk_refcnt)) |
| 447 |
sk = NULL; |
| 448 |
return sk; |
| 449 |
} |
| 450 |
EXPORT_SYMBOL_GPL(udp6_lib_lookup); |
| 451 |
#endif |
| 452 |
|
| 453 |
/* do not use the scratch area len for jumbogram: their length exceeds the |
| 454 |
* scratch area space; note that the IP6CB flags is still in the first |
| 455 |
* cacheline, so checking for jumbograms is cheap |
| 456 |
*/ |
| 457 |
static int udp6_skb_len(struct sk_buff *skb) |
| 458 |
{ |
| 459 |
return unlikely(inet6_is_jumbogram(skb)) ? skb->len : udp_skb_len(skb); |
| 460 |
} |
| 461 |
|
| 462 |
/* |
| 463 |
* This should be easy, if there is something there we |
| 464 |
* return it, otherwise we block. |
| 465 |
*/ |
| 466 |
|
| 467 |
int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, |
| 468 |
int flags, int *addr_len) |
| 469 |
{ |
| 470 |
struct ipv6_pinfo *np = inet6_sk(sk); |
| 471 |
struct inet_sock *inet = inet_sk(sk); |
| 472 |
struct sk_buff *skb; |
| 473 |
unsigned int ulen, copied; |
| 474 |
int off, err, peeking = flags & MSG_PEEK; |
| 475 |
int is_udplite = IS_UDPLITE(sk); |
| 476 |
struct udp_mib __percpu *mib; |
| 477 |
bool checksum_valid = false; |
| 478 |
int is_udp4; |
| 479 |
|
| 480 |
if (flags & MSG_ERRQUEUE) |
| 481 |
return ipv6_recv_error(sk, msg, len, addr_len); |
| 482 |
|
| 483 |
if (np->rxopt.bits.rxpmtu && READ_ONCE(np->rxpmtu)) |
| 484 |
return ipv6_recv_rxpmtu(sk, msg, len, addr_len); |
| 485 |
|
| 486 |
try_again: |
| 487 |
off = sk_peek_offset(sk, flags); |
| 488 |
skb = __skb_recv_udp(sk, flags, &off, &err); |
| 489 |
if (!skb) |
| 490 |
return err; |
| 491 |
|
| 492 |
ulen = udp6_skb_len(skb); |
| 493 |
copied = len; |
| 494 |
if (copied > ulen - off) |
| 495 |
copied = ulen - off; |
| 496 |
else if (copied < ulen) |
| 497 |
msg->msg_flags |= MSG_TRUNC; |
| 498 |
|
| 499 |
is_udp4 = (skb->protocol == htons(ETH_P_IP)); |
| 500 |
mib = __UDPX_MIB(sk, is_udp4); |
| 501 |
|
| 502 |
/* |
| 503 |
* If checksum is needed at all, try to do it while copying the |
| 504 |
* data. If the data is truncated, or if we only want a partial |
| 505 |
* coverage checksum (UDP-Lite), do it before the copy. |
| 506 |
*/ |
| 507 |
|
| 508 |
if (copied < ulen || peeking || |
| 509 |
(is_udplite && UDP_SKB_CB(skb)->partial_cov)) { |
| 510 |
checksum_valid = udp_skb_csum_unnecessary(skb) || |
| 511 |
!__udp_lib_checksum_complete(skb); |
| 512 |
if (!checksum_valid) |
| 513 |
goto csum_copy_err; |
| 514 |
} |
| 515 |
|
| 516 |
if (checksum_valid || udp_skb_csum_unnecessary(skb)) { |
| 517 |
if (udp_skb_is_linear(skb)) |
| 518 |
err = copy_linear_skb(skb, copied, off, &msg->msg_iter); |
| 519 |
else |
| 520 |
err = skb_copy_datagram_msg(skb, off, msg, copied); |
| 521 |
} else { |
| 522 |
err = skb_copy_and_csum_datagram_msg(skb, off, msg); |
| 523 |
if (err == -EINVAL) |
| 524 |
goto csum_copy_err; |
| 525 |
} |
| 526 |
if (unlikely(err)) { |
| 527 |
if (!peeking) { |
| 528 |
udp_drops_inc(sk); |
| 529 |
SNMP_INC_STATS(mib, UDP_MIB_INERRORS); |
| 530 |
} |
| 531 |
kfree_skb(skb); |
| 532 |
return err; |
| 533 |
} |
| 534 |
if (!peeking) |
| 535 |
SNMP_INC_STATS(mib, UDP_MIB_INDATAGRAMS); |
| 536 |
|
| 537 |
sock_recv_cmsgs(msg, sk, skb); |
| 538 |
|
| 539 |
/* Copy the address. */ |
| 540 |
if (msg->msg_name) { |
| 541 |
DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
| 542 |
sin6->sin6_family = AF_INET6; |
| 543 |
sin6->sin6_port = udp_hdr(skb)->source; |
| 544 |
sin6->sin6_flowinfo = 0; |
| 545 |
|
| 546 |
if (is_udp4) { |
| 547 |
ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr, |
| 548 |
&sin6->sin6_addr); |
| 549 |
sin6->sin6_scope_id = 0; |
| 550 |
} else { |
| 551 |
sin6->sin6_addr = ipv6_hdr(skb)->saddr; |
| 552 |
sin6->sin6_scope_id = |
| 553 |
ipv6_iface_scope_id(&sin6->sin6_addr, |
| 554 |
inet6_iif(skb)); |
| 555 |
} |
| 556 |
*addr_len = sizeof(*sin6); |
| 557 |
|
| 558 |
BPF_CGROUP_RUN_PROG_UDP6_RECVMSG_LOCK(sk, |
| 559 |
(struct sockaddr *)sin6, |
| 560 |
addr_len); |
| 561 |
} |
| 562 |
|
| 563 |
if (udp_test_bit(GRO_ENABLED, sk)) |
| 564 |
udp_cmsg_recv(msg, sk, skb); |
| 565 |
|
| 566 |
if (np->rxopt.all) |
| 567 |
ip6_datagram_recv_common_ctl(sk, msg, skb); |
| 568 |
|
| 569 |
if (is_udp4) { |
| 570 |
if (inet_cmsg_flags(inet)) |
| 571 |
ip_cmsg_recv_offset(msg, sk, skb, |
| 572 |
sizeof(struct udphdr), off); |
| 573 |
} else { |
| 574 |
if (np->rxopt.all) |
| 575 |
ip6_datagram_recv_specific_ctl(sk, msg, skb); |
| 576 |
} |
| 577 |
|
| 578 |
err = copied; |
| 579 |
if (flags & MSG_TRUNC) |
| 580 |
err = ulen; |
| 581 |
|
| 582 |
skb_consume_udp(sk, skb, peeking ? -err : err); |
| 583 |
return err; |
| 584 |
|
| 585 |
csum_copy_err: |
| 586 |
if (!__sk_queue_drop_skb(sk, &udp_sk(sk)->reader_queue, skb, flags, |
| 587 |
udp_skb_destructor)) { |
| 588 |
SNMP_INC_STATS(mib, UDP_MIB_CSUMERRORS); |
| 589 |
SNMP_INC_STATS(mib, UDP_MIB_INERRORS); |
| 590 |
} |
| 591 |
kfree_skb_reason(skb, SKB_DROP_REASON_UDP_CSUM); |
| 592 |
|
| 593 |
/* starting over for a new packet, but check if we need to yield */ |
| 594 |
cond_resched(); |
| 595 |
msg->msg_flags &= ~MSG_TRUNC; |
| 596 |
goto try_again; |
| 597 |
} |
| 598 |
|
| 599 |
DECLARE_STATIC_KEY_FALSE(udpv6_encap_needed_key); |
| 600 |
void udpv6_encap_enable(void) |
| 601 |
{ |
| 602 |
static_branch_inc(&udpv6_encap_needed_key); |
| 603 |
} |
| 604 |
EXPORT_SYMBOL(udpv6_encap_enable); |
| 605 |
|
| 606 |
/* Handler for tunnels with arbitrary destination ports: no socket lookup, go |
| 607 |
* through error handlers in encapsulations looking for a match. |
| 608 |
*/ |
| 609 |
static int __udp6_lib_err_encap_no_sk(struct sk_buff *skb, |
| 610 |
struct inet6_skb_parm *opt, |
| 611 |
u8 type, u8 code, int offset, __be32 info) |
| 612 |
{ |
| 613 |
int i; |
| 614 |
|
| 615 |
for (i = 0; i < MAX_IPTUN_ENCAP_OPS; i++) { |
| 616 |
int (*handler)(struct sk_buff *skb, struct inet6_skb_parm *opt, |
| 617 |
u8 type, u8 code, int offset, __be32 info); |
| 618 |
const struct ip6_tnl_encap_ops *encap; |
| 619 |
|
| 620 |
encap = rcu_dereference(ip6tun_encaps[i]); |
| 621 |
if (!encap) |
| 622 |
continue; |
| 623 |
handler = encap->err_handler; |
| 624 |
if (handler && !handler(skb, opt, type, code, offset, info)) |
| 625 |
return 0; |
| 626 |
} |
| 627 |
|
| 628 |
return -ENOENT; |
| 629 |
} |
| 630 |
|
| 631 |
/* Try to match ICMP errors to UDP tunnels by looking up a socket without |
| 632 |
* reversing source and destination port: this will match tunnels that force the |
| 633 |
* same destination port on both endpoints (e.g. VXLAN, GENEVE). Note that |
| 634 |
* lwtunnels might actually break this assumption by being configured with |
| 635 |
* different destination ports on endpoints, in this case we won't be able to |
| 636 |
* trace ICMP messages back to them. |
| 637 |
* |
| 638 |
* If this doesn't match any socket, probe tunnels with arbitrary destination |
| 639 |
* ports (e.g. FoU, GUE): there, the receiving socket is useless, as the port |
| 640 |
* we've sent packets to won't necessarily match the local destination port. |
| 641 |
* |
| 642 |
* Then ask the tunnel implementation to match the error against a valid |
| 643 |
* association. |
| 644 |
* |
| 645 |
* Return an error if we can't find a match, the socket if we need further |
| 646 |
* processing, zero otherwise. |
| 647 |
*/ |
| 648 |
static struct sock *__udp6_lib_err_encap(struct net *net, |
| 649 |
const struct ipv6hdr *hdr, int offset, |
| 650 |
struct udphdr *uh, |
| 651 |
struct udp_table *udptable, |
| 652 |
struct sock *sk, |
| 653 |
struct sk_buff *skb, |
| 654 |
struct inet6_skb_parm *opt, |
| 655 |
u8 type, u8 code, __be32 info) |
| 656 |
{ |
| 657 |
int (*lookup)(struct sock *sk, struct sk_buff *skb); |
| 658 |
int network_offset, transport_offset; |
| 659 |
struct udp_sock *up; |
| 660 |
|
| 661 |
network_offset = skb_network_offset(skb); |
| 662 |
transport_offset = skb_transport_offset(skb); |
| 663 |
|
| 664 |
/* Network header needs to point to the outer IPv6 header inside ICMP */ |
| 665 |
skb_reset_network_header(skb); |
| 666 |
|
| 667 |
/* Transport header needs to point to the UDP header */ |
| 668 |
skb_set_transport_header(skb, offset); |
| 669 |
|
| 670 |
if (sk) { |
| 671 |
up = udp_sk(sk); |
| 672 |
|
| 673 |
lookup = READ_ONCE(up->encap_err_lookup); |
| 674 |
if (lookup && lookup(sk, skb)) |
| 675 |
sk = NULL; |
| 676 |
|
| 677 |
goto out; |
| 678 |
} |
| 679 |
|
| 680 |
sk = __udp6_lib_lookup(net, &hdr->daddr, uh->source, |
| 681 |
&hdr->saddr, uh->dest, |
| 682 |
inet6_iif(skb), 0, udptable, skb); |
| 683 |
if (sk) { |
| 684 |
up = udp_sk(sk); |
| 685 |
|
| 686 |
lookup = READ_ONCE(up->encap_err_lookup); |
| 687 |
if (!lookup || lookup(sk, skb)) |
| 688 |
sk = NULL; |
| 689 |
} |
| 690 |
|
| 691 |
out: |
| 692 |
if (!sk) { |
| 693 |
sk = ERR_PTR(__udp6_lib_err_encap_no_sk(skb, opt, type, code, |
| 694 |
offset, info)); |
| 695 |
} |
| 696 |
|
| 697 |
skb_set_transport_header(skb, transport_offset); |
| 698 |
skb_set_network_header(skb, network_offset); |
| 699 |
|
| 700 |
return sk; |
| 701 |
} |
| 702 |
|
| 703 |
int __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt, |
| 704 |
u8 type, u8 code, int offset, __be32 info, |
| 705 |
struct udp_table *udptable) |
| 706 |
{ |
| 707 |
struct ipv6_pinfo *np; |
| 708 |
const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data; |
| 709 |
const struct in6_addr *saddr = &hdr->saddr; |
| 710 |
const struct in6_addr *daddr = seg6_get_daddr(skb, opt) ? : &hdr->daddr; |
| 711 |
struct udphdr *uh = (struct udphdr *)(skb->data+offset); |
| 712 |
bool tunnel = false; |
| 713 |
struct sock *sk; |
| 714 |
int harderr; |
| 715 |
int err; |
| 716 |
struct net *net = dev_net(skb->dev); |
| 717 |
|
| 718 |
sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source, |
| 719 |
inet6_iif(skb), inet6_sdif(skb), udptable, NULL); |
| 720 |
|
| 721 |
if (!sk || READ_ONCE(udp_sk(sk)->encap_type)) { |
| 722 |
/* No socket for error: try tunnels before discarding */ |
| 723 |
if (static_branch_unlikely(&udpv6_encap_needed_key)) { |
| 724 |
sk = __udp6_lib_err_encap(net, hdr, offset, uh, |
| 725 |
udptable, sk, skb, |
| 726 |
opt, type, code, info); |
| 727 |
if (!sk) |
| 728 |
return 0; |
| 729 |
} else |
| 730 |
sk = ERR_PTR(-ENOENT); |
| 731 |
|
| 732 |
if (IS_ERR(sk)) { |
| 733 |
__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), |
| 734 |
ICMP6_MIB_INERRORS); |
| 735 |
return PTR_ERR(sk); |
| 736 |
} |
| 737 |
|
| 738 |
tunnel = true; |
| 739 |
} |
| 740 |
|
| 741 |
harderr = icmpv6_err_convert(type, code, &err); |
| 742 |
np = inet6_sk(sk); |
| 743 |
|
| 744 |
if (type == ICMPV6_PKT_TOOBIG) { |
| 745 |
if (!ip6_sk_accept_pmtu(sk)) |
| 746 |
goto out; |
| 747 |
ip6_sk_update_pmtu(skb, sk, info); |
| 748 |
if (READ_ONCE(np->pmtudisc) != IPV6_PMTUDISC_DONT) |
| 749 |
harderr = 1; |
| 750 |
} |
| 751 |
if (type == NDISC_REDIRECT) { |
| 752 |
if (tunnel) { |
| 753 |
ip6_redirect(skb, sock_net(sk), inet6_iif(skb), |
| 754 |
READ_ONCE(sk->sk_mark), |
| 755 |
sk_uid(sk)); |
| 756 |
} else { |
| 757 |
ip6_sk_redirect(skb, sk); |
| 758 |
} |
| 759 |
goto out; |
| 760 |
} |
| 761 |
|
| 762 |
/* Tunnels don't have an application socket: don't pass errors back */ |
| 763 |
if (tunnel) { |
| 764 |
if (udp_sk(sk)->encap_err_rcv) |
| 765 |
udp_sk(sk)->encap_err_rcv(sk, skb, err, uh->dest, |
| 766 |
ntohl(info), (u8 *)(uh+1)); |
| 767 |
goto out; |
| 768 |
} |
| 769 |
|
| 770 |
if (!inet6_test_bit(RECVERR6, sk)) { |
| 771 |
if (!harderr || sk->sk_state != TCP_ESTABLISHED) |
| 772 |
goto out; |
| 773 |
} else { |
| 774 |
ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1)); |
| 775 |
} |
| 776 |
|
| 777 |
sk->sk_err = err; |
| 778 |
sk_error_report(sk); |
| 779 |
out: |
| 780 |
return 0; |
| 781 |
} |
| 782 |
|
| 783 |
static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
| 784 |
{ |
| 785 |
int rc; |
| 786 |
|
| 787 |
if (!ipv6_addr_any(&sk->sk_v6_daddr)) { |
| 788 |
sock_rps_save_rxhash(sk, skb); |
| 789 |
sk_mark_napi_id(sk, skb); |
| 790 |
sk_incoming_cpu_update(sk); |
| 791 |
} else { |
| 792 |
sk_mark_napi_id_once(sk, skb); |
| 793 |
} |
| 794 |
|
| 795 |
rc = __udp_enqueue_schedule_skb(sk, skb); |
| 796 |
if (rc < 0) { |
| 797 |
int is_udplite = IS_UDPLITE(sk); |
| 798 |
enum skb_drop_reason drop_reason; |
| 799 |
|
| 800 |
/* Note that an ENOMEM error is charged twice */ |
| 801 |
if (rc == -ENOMEM) { |
| 802 |
UDP6_INC_STATS(sock_net(sk), |
| 803 |
UDP_MIB_RCVBUFERRORS, is_udplite); |
| 804 |
drop_reason = SKB_DROP_REASON_SOCKET_RCVBUFF; |
| 805 |
} else { |
| 806 |
UDP6_INC_STATS(sock_net(sk), |
| 807 |
UDP_MIB_MEMERRORS, is_udplite); |
| 808 |
drop_reason = SKB_DROP_REASON_PROTO_MEM; |
| 809 |
} |
| 810 |
UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
| 811 |
trace_udp_fail_queue_rcv_skb(rc, sk, skb); |
| 812 |
sk_skb_reason_drop(sk, skb, drop_reason); |
| 813 |
return -1; |
| 814 |
} |
| 815 |
|
| 816 |
return 0; |
| 817 |
} |
| 818 |
|
| 819 |
static __inline__ int udpv6_err(struct sk_buff *skb, |
| 820 |
struct inet6_skb_parm *opt, u8 type, |
| 821 |
u8 code, int offset, __be32 info) |
| 822 |
{ |
| 823 |
return __udp6_lib_err(skb, opt, type, code, offset, info, |
| 824 |
dev_net(skb->dev)->ipv4.udp_table); |
| 825 |
} |
| 826 |
|
| 827 |
static int udpv6_queue_rcv_one_skb(struct sock *sk, struct sk_buff *skb) |
| 828 |
{ |
| 829 |
enum skb_drop_reason drop_reason = SKB_DROP_REASON_NOT_SPECIFIED; |
| 830 |
struct udp_sock *up = udp_sk(sk); |
| 831 |
int is_udplite = IS_UDPLITE(sk); |
| 832 |
|
| 833 |
if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { |
| 834 |
drop_reason = SKB_DROP_REASON_XFRM_POLICY; |
| 835 |
goto drop; |
| 836 |
} |
| 837 |
nf_reset_ct(skb); |
| 838 |
|
| 839 |
if (static_branch_unlikely(&udpv6_encap_needed_key) && |
| 840 |
READ_ONCE(up->encap_type)) { |
| 841 |
int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
| 842 |
|
| 843 |
/* |
| 844 |
* This is an encapsulation socket so pass the skb to |
| 845 |
* the socket's udp_encap_rcv() hook. Otherwise, just |
| 846 |
* fall through and pass this up the UDP socket. |
| 847 |
* up->encap_rcv() returns the following value: |
| 848 |
* =0 if skb was successfully passed to the encap |
| 849 |
* handler or was discarded by it. |
| 850 |
* >0 if skb should be passed on to UDP. |
| 851 |
* <0 if skb should be resubmitted as proto -N |
| 852 |
*/ |
| 853 |
|
| 854 |
/* if we're overly short, let UDP handle it */ |
| 855 |
encap_rcv = READ_ONCE(up->encap_rcv); |
| 856 |
if (encap_rcv) { |
| 857 |
int ret; |
| 858 |
|
| 859 |
/* Verify checksum before giving to encap */ |
| 860 |
if (udp_lib_checksum_complete(skb)) |
| 861 |
goto csum_error; |
| 862 |
|
| 863 |
ret = encap_rcv(sk, skb); |
| 864 |
if (ret <= 0) { |
| 865 |
__UDP6_INC_STATS(sock_net(sk), |
| 866 |
UDP_MIB_INDATAGRAMS, |
| 867 |
is_udplite); |
| 868 |
return -ret; |
| 869 |
} |
| 870 |
} |
| 871 |
|
| 872 |
/* FALLTHROUGH -- it's a UDP Packet */ |
| 873 |
} |
| 874 |
|
| 875 |
/* |
| 876 |
* UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c). |
| 877 |
*/ |
| 878 |
if (unlikely(udp_test_bit(UDPLITE_RECV_CC, sk) && |
| 879 |
UDP_SKB_CB(skb)->partial_cov)) { |
| 880 |
u16 pcrlen = READ_ONCE(up->pcrlen); |
| 881 |
|
| 882 |
if (pcrlen == 0) { /* full coverage was set */ |
| 883 |
net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n", |
| 884 |
UDP_SKB_CB(skb)->cscov, skb->len); |
| 885 |
goto drop; |
| 886 |
} |
| 887 |
if (UDP_SKB_CB(skb)->cscov < pcrlen) { |
| 888 |
net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n", |
| 889 |
UDP_SKB_CB(skb)->cscov, pcrlen); |
| 890 |
goto drop; |
| 891 |
} |
| 892 |
} |
| 893 |
|
| 894 |
prefetch(&sk->sk_rmem_alloc); |
| 895 |
if (rcu_access_pointer(sk->sk_filter) && |
| 896 |
udp_lib_checksum_complete(skb)) |
| 897 |
goto csum_error; |
| 898 |
|
| 899 |
if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr), &drop_reason)) |
| 900 |
goto drop; |
| 901 |
|
| 902 |
udp_csum_pull_header(skb); |
| 903 |
|
| 904 |
skb_dst_drop(skb); |
| 905 |
|
| 906 |
return __udpv6_queue_rcv_skb(sk, skb); |
| 907 |
|
| 908 |
csum_error: |
| 909 |
drop_reason = SKB_DROP_REASON_UDP_CSUM; |
| 910 |
__UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); |
| 911 |
drop: |
| 912 |
__UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
| 913 |
udp_drops_inc(sk); |
| 914 |
sk_skb_reason_drop(sk, skb, drop_reason); |
| 915 |
return -1; |
| 916 |
} |
| 917 |
|
| 918 |
static int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
| 919 |
{ |
| 920 |
struct sk_buff *next, *segs; |
| 921 |
int ret; |
| 922 |
|
| 923 |
if (likely(!udp_unexpected_gso(sk, skb))) |
| 924 |
return udpv6_queue_rcv_one_skb(sk, skb); |
| 925 |
|
| 926 |
__skb_push(skb, -skb_mac_offset(skb)); |
| 927 |
segs = udp_rcv_segment(sk, skb, false); |
| 928 |
skb_list_walk_safe(segs, skb, next) { |
| 929 |
__skb_pull(skb, skb_transport_offset(skb)); |
| 930 |
|
| 931 |
udp_post_segment_fix_csum(skb); |
| 932 |
ret = udpv6_queue_rcv_one_skb(sk, skb); |
| 933 |
if (ret > 0) |
| 934 |
ip6_protocol_deliver_rcu(dev_net(skb->dev), skb, ret, |
| 935 |
true); |
| 936 |
} |
| 937 |
return 0; |
| 938 |
} |
| 939 |
|
| 940 |
static bool __udp_v6_is_mcast_sock(struct net *net, const struct sock *sk, |
| 941 |
__be16 loc_port, const struct in6_addr *loc_addr, |
| 942 |
__be16 rmt_port, const struct in6_addr *rmt_addr, |
| 943 |
int dif, int sdif, unsigned short hnum) |
| 944 |
{ |
| 945 |
const struct inet_sock *inet = inet_sk(sk); |
| 946 |
|
| 947 |
if (!net_eq(sock_net(sk), net)) |
| 948 |
return false; |
| 949 |
|
| 950 |
if (udp_sk(sk)->udp_port_hash != hnum || |
| 951 |
sk->sk_family != PF_INET6 || |
| 952 |
(inet->inet_dport && inet->inet_dport != rmt_port) || |
| 953 |
(!ipv6_addr_any(&sk->sk_v6_daddr) && |
| 954 |
!ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) || |
| 955 |
!udp_sk_bound_dev_eq(net, READ_ONCE(sk->sk_bound_dev_if), dif, sdif) || |
| 956 |
(!ipv6_addr_any(&sk->sk_v6_rcv_saddr) && |
| 957 |
!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))) |
| 958 |
return false; |
| 959 |
if (!inet6_mc_check(sk, loc_addr, rmt_addr)) |
| 960 |
return false; |
| 961 |
return true; |
| 962 |
} |
| 963 |
|
| 964 |
static void udp6_csum_zero_error(struct sk_buff *skb) |
| 965 |
{ |
| 966 |
/* RFC 2460 section 8.1 says that we SHOULD log |
| 967 |
* this error. Well, it is reasonable. |
| 968 |
*/ |
| 969 |
net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n", |
| 970 |
&ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source), |
| 971 |
&ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest)); |
| 972 |
} |
| 973 |
|
| 974 |
/* |
| 975 |
* Note: called only from the BH handler context, |
| 976 |
* so we don't need to lock the hashes. |
| 977 |
*/ |
| 978 |
static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
| 979 |
const struct in6_addr *saddr, const struct in6_addr *daddr, |
| 980 |
struct udp_table *udptable, int proto) |
| 981 |
{ |
| 982 |
struct sock *sk, *first = NULL; |
| 983 |
const struct udphdr *uh = udp_hdr(skb); |
| 984 |
unsigned short hnum = ntohs(uh->dest); |
| 985 |
struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum); |
| 986 |
unsigned int offset = offsetof(typeof(*sk), sk_node); |
| 987 |
unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10); |
| 988 |
int dif = inet6_iif(skb); |
| 989 |
int sdif = inet6_sdif(skb); |
| 990 |
struct hlist_node *node; |
| 991 |
struct sk_buff *nskb; |
| 992 |
|
| 993 |
if (use_hash2) { |
| 994 |
hash2_any = ipv6_portaddr_hash(net, &in6addr_any, hnum) & |
| 995 |
udptable->mask; |
| 996 |
hash2 = ipv6_portaddr_hash(net, daddr, hnum) & udptable->mask; |
| 997 |
start_lookup: |
| 998 |
hslot = &udptable->hash2[hash2].hslot; |
| 999 |
offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node); |
| 1000 |
} |
| 1001 |
|
| 1002 |
sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) { |
| 1003 |
if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr, |
| 1004 |
uh->source, saddr, dif, sdif, |
| 1005 |
hnum)) |
| 1006 |
continue; |
| 1007 |
/* If zero checksum and no_check is not on for |
| 1008 |
* the socket then skip it. |
| 1009 |
*/ |
| 1010 |
if (!uh->check && !udp_get_no_check6_rx(sk)) |
| 1011 |
continue; |
| 1012 |
if (!first) { |
| 1013 |
first = sk; |
| 1014 |
continue; |
| 1015 |
} |
| 1016 |
nskb = skb_clone(skb, GFP_ATOMIC); |
| 1017 |
if (unlikely(!nskb)) { |
| 1018 |
udp_drops_inc(sk); |
| 1019 |
__UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS, |
| 1020 |
IS_UDPLITE(sk)); |
| 1021 |
__UDP6_INC_STATS(net, UDP_MIB_INERRORS, |
| 1022 |
IS_UDPLITE(sk)); |
| 1023 |
continue; |
| 1024 |
} |
| 1025 |
|
| 1026 |
if (udpv6_queue_rcv_skb(sk, nskb) > 0) |
| 1027 |
consume_skb(nskb); |
| 1028 |
} |
| 1029 |
|
| 1030 |
/* Also lookup *:port if we are using hash2 and haven't done so yet. */ |
| 1031 |
if (use_hash2 && hash2 != hash2_any) { |
| 1032 |
hash2 = hash2_any; |
| 1033 |
goto start_lookup; |
| 1034 |
} |
| 1035 |
|
| 1036 |
if (first) { |
| 1037 |
if (udpv6_queue_rcv_skb(first, skb) > 0) |
| 1038 |
consume_skb(skb); |
| 1039 |
} else { |
| 1040 |
kfree_skb(skb); |
| 1041 |
__UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI, |
| 1042 |
proto == IPPROTO_UDPLITE); |
| 1043 |
} |
| 1044 |
return 0; |
| 1045 |
} |
| 1046 |
|
| 1047 |
static void udp6_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst) |
| 1048 |
{ |
| 1049 |
if (udp_sk_rx_dst_set(sk, dst)) |
| 1050 |
sk->sk_rx_dst_cookie = rt6_get_cookie(dst_rt6_info(dst)); |
| 1051 |
} |
| 1052 |
|
| 1053 |
/* wrapper for udp_queue_rcv_skb taking care of csum conversion and |
| 1054 |
* return code conversion for ip layer consumption |
| 1055 |
*/ |
| 1056 |
static int udp6_unicast_rcv_skb(struct sock *sk, struct sk_buff *skb, |
| 1057 |
struct udphdr *uh) |
| 1058 |
{ |
| 1059 |
int ret; |
| 1060 |
|
| 1061 |
if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk)) |
| 1062 |
skb_checksum_try_convert(skb, IPPROTO_UDP, ip6_compute_pseudo); |
| 1063 |
|
| 1064 |
ret = udpv6_queue_rcv_skb(sk, skb); |
| 1065 |
|
| 1066 |
/* a return value > 0 means to resubmit the input */ |
| 1067 |
if (ret > 0) |
| 1068 |
return ret; |
| 1069 |
return 0; |
| 1070 |
} |
| 1071 |
|
| 1072 |
int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
| 1073 |
int proto) |
| 1074 |
{ |
| 1075 |
enum skb_drop_reason reason = SKB_DROP_REASON_NOT_SPECIFIED; |
| 1076 |
const struct in6_addr *saddr, *daddr; |
| 1077 |
struct net *net = dev_net(skb->dev); |
| 1078 |
struct sock *sk = NULL; |
| 1079 |
struct udphdr *uh; |
| 1080 |
bool refcounted; |
| 1081 |
u32 ulen = 0; |
| 1082 |
|
| 1083 |
if (!pskb_may_pull(skb, sizeof(struct udphdr))) |
| 1084 |
goto discard; |
| 1085 |
|
| 1086 |
saddr = &ipv6_hdr(skb)->saddr; |
| 1087 |
daddr = &ipv6_hdr(skb)->daddr; |
| 1088 |
uh = udp_hdr(skb); |
| 1089 |
|
| 1090 |
ulen = ntohs(uh->len); |
| 1091 |
if (ulen > skb->len) |
| 1092 |
goto short_packet; |
| 1093 |
|
| 1094 |
if (proto == IPPROTO_UDP) { |
| 1095 |
/* UDP validates ulen. */ |
| 1096 |
|
| 1097 |
/* Check for jumbo payload */ |
| 1098 |
if (ulen == 0) |
| 1099 |
ulen = skb->len; |
| 1100 |
|
| 1101 |
if (ulen < sizeof(*uh)) |
| 1102 |
goto short_packet; |
| 1103 |
|
| 1104 |
if (ulen < skb->len) { |
| 1105 |
if (pskb_trim_rcsum(skb, ulen)) |
| 1106 |
goto short_packet; |
| 1107 |
saddr = &ipv6_hdr(skb)->saddr; |
| 1108 |
daddr = &ipv6_hdr(skb)->daddr; |
| 1109 |
uh = udp_hdr(skb); |
| 1110 |
} |
| 1111 |
} |
| 1112 |
|
| 1113 |
if (udp6_csum_init(skb, uh, proto)) |
| 1114 |
goto csum_error; |
| 1115 |
|
| 1116 |
/* Check if the socket is already available, e.g. due to early demux */ |
| 1117 |
sk = inet6_steal_sock(net, skb, sizeof(struct udphdr), saddr, uh->source, daddr, uh->dest, |
| 1118 |
&refcounted, udp6_ehashfn); |
| 1119 |
if (IS_ERR(sk)) |
| 1120 |
goto no_sk; |
| 1121 |
|
| 1122 |
if (sk) { |
| 1123 |
struct dst_entry *dst = skb_dst(skb); |
| 1124 |
int ret; |
| 1125 |
|
| 1126 |
if (unlikely(rcu_dereference(sk->sk_rx_dst) != dst)) |
| 1127 |
udp6_sk_rx_dst_set(sk, dst); |
| 1128 |
|
| 1129 |
if (!uh->check && !udp_get_no_check6_rx(sk)) { |
| 1130 |
if (refcounted) |
| 1131 |
sock_put(sk); |
| 1132 |
goto report_csum_error; |
| 1133 |
} |
| 1134 |
|
| 1135 |
ret = udp6_unicast_rcv_skb(sk, skb, uh); |
| 1136 |
if (refcounted) |
| 1137 |
sock_put(sk); |
| 1138 |
return ret; |
| 1139 |
} |
| 1140 |
|
| 1141 |
/* |
| 1142 |
* Multicast receive code |
| 1143 |
*/ |
| 1144 |
if (ipv6_addr_is_multicast(daddr)) |
| 1145 |
return __udp6_lib_mcast_deliver(net, skb, |
| 1146 |
saddr, daddr, udptable, proto); |
| 1147 |
|
| 1148 |
/* Unicast */ |
| 1149 |
sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable); |
| 1150 |
if (sk) { |
| 1151 |
if (!uh->check && !udp_get_no_check6_rx(sk)) |
| 1152 |
goto report_csum_error; |
| 1153 |
return udp6_unicast_rcv_skb(sk, skb, uh); |
| 1154 |
} |
| 1155 |
no_sk: |
| 1156 |
reason = SKB_DROP_REASON_NO_SOCKET; |
| 1157 |
|
| 1158 |
if (!uh->check) |
| 1159 |
goto report_csum_error; |
| 1160 |
|
| 1161 |
if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) |
| 1162 |
goto discard; |
| 1163 |
nf_reset_ct(skb); |
| 1164 |
|
| 1165 |
if (udp_lib_checksum_complete(skb)) |
| 1166 |
goto csum_error; |
| 1167 |
|
| 1168 |
__UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
| 1169 |
icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
| 1170 |
|
| 1171 |
sk_skb_reason_drop(sk, skb, reason); |
| 1172 |
return 0; |
| 1173 |
|
| 1174 |
short_packet: |
| 1175 |
if (reason == SKB_DROP_REASON_NOT_SPECIFIED) |
| 1176 |
reason = SKB_DROP_REASON_PKT_TOO_SMALL; |
| 1177 |
net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n", |
| 1178 |
proto == IPPROTO_UDPLITE ? "-Lite" : "", |
| 1179 |
saddr, ntohs(uh->source), |
| 1180 |
ulen, skb->len, |
| 1181 |
daddr, ntohs(uh->dest)); |
| 1182 |
goto discard; |
| 1183 |
|
| 1184 |
report_csum_error: |
| 1185 |
udp6_csum_zero_error(skb); |
| 1186 |
csum_error: |
| 1187 |
if (reason == SKB_DROP_REASON_NOT_SPECIFIED) |
| 1188 |
reason = SKB_DROP_REASON_UDP_CSUM; |
| 1189 |
__UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); |
| 1190 |
discard: |
| 1191 |
__UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
| 1192 |
sk_skb_reason_drop(sk, skb, reason); |
| 1193 |
return 0; |
| 1194 |
} |
| 1195 |
|
| 1196 |
|
| 1197 |
static struct sock *__udp6_lib_demux_lookup(struct net *net, |
| 1198 |
__be16 loc_port, const struct in6_addr *loc_addr, |
| 1199 |
__be16 rmt_port, const struct in6_addr *rmt_addr, |
| 1200 |
int dif, int sdif) |
| 1201 |
{ |
| 1202 |
struct udp_table *udptable = net->ipv4.udp_table; |
| 1203 |
unsigned short hnum = ntohs(loc_port); |
| 1204 |
struct udp_hslot *hslot2; |
| 1205 |
unsigned int hash2; |
| 1206 |
__portpair ports; |
| 1207 |
struct sock *sk; |
| 1208 |
|
| 1209 |
hash2 = ipv6_portaddr_hash(net, loc_addr, hnum); |
| 1210 |
hslot2 = udp_hashslot2(udptable, hash2); |
| 1211 |
ports = INET_COMBINED_PORTS(rmt_port, hnum); |
| 1212 |
|
| 1213 |
udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
| 1214 |
if (sk->sk_state == TCP_ESTABLISHED && |
| 1215 |
inet6_match(net, sk, rmt_addr, loc_addr, ports, dif, sdif)) |
| 1216 |
return sk; |
| 1217 |
/* Only check first socket in chain */ |
| 1218 |
break; |
| 1219 |
} |
| 1220 |
return NULL; |
| 1221 |
} |
| 1222 |
|
| 1223 |
void udp_v6_early_demux(struct sk_buff *skb) |
| 1224 |
{ |
| 1225 |
struct net *net = dev_net(skb->dev); |
| 1226 |
const struct udphdr *uh; |
| 1227 |
struct sock *sk; |
| 1228 |
struct dst_entry *dst; |
| 1229 |
int dif = skb->dev->ifindex; |
| 1230 |
int sdif = inet6_sdif(skb); |
| 1231 |
|
| 1232 |
if (!pskb_may_pull(skb, skb_transport_offset(skb) + |
| 1233 |
sizeof(struct udphdr))) |
| 1234 |
return; |
| 1235 |
|
| 1236 |
uh = udp_hdr(skb); |
| 1237 |
|
| 1238 |
if (skb->pkt_type == PACKET_HOST) |
| 1239 |
sk = __udp6_lib_demux_lookup(net, uh->dest, |
| 1240 |
&ipv6_hdr(skb)->daddr, |
| 1241 |
uh->source, &ipv6_hdr(skb)->saddr, |
| 1242 |
dif, sdif); |
| 1243 |
else |
| 1244 |
return; |
| 1245 |
|
| 1246 |
if (!sk) |
| 1247 |
return; |
| 1248 |
|
| 1249 |
skb->sk = sk; |
| 1250 |
DEBUG_NET_WARN_ON_ONCE(sk_is_refcounted(sk)); |
| 1251 |
skb->destructor = sock_pfree; |
| 1252 |
dst = rcu_dereference(sk->sk_rx_dst); |
| 1253 |
|
| 1254 |
if (dst) |
| 1255 |
dst = dst_check(dst, sk->sk_rx_dst_cookie); |
| 1256 |
if (dst) { |
| 1257 |
/* set noref for now. |
| 1258 |
* any place which wants to hold dst has to call |
| 1259 |
* dst_hold_safe() |
| 1260 |
*/ |
| 1261 |
skb_dst_set_noref(skb, dst); |
| 1262 |
} |
| 1263 |
} |
| 1264 |
|
| 1265 |
INDIRECT_CALLABLE_SCOPE int udpv6_rcv(struct sk_buff *skb) |
| 1266 |
{ |
| 1267 |
return __udp6_lib_rcv(skb, dev_net(skb->dev)->ipv4.udp_table, IPPROTO_UDP); |
| 1268 |
} |
| 1269 |
|
| 1270 |
/* |
| 1271 |
* Throw away all pending data and cancel the corking. Socket is locked. |
| 1272 |
*/ |
| 1273 |
static void udp_v6_flush_pending_frames(struct sock *sk) |
| 1274 |
{ |
| 1275 |
struct udp_sock *up = udp_sk(sk); |
| 1276 |
|
| 1277 |
if (up->pending == AF_INET) |
| 1278 |
udp_flush_pending_frames(sk); |
| 1279 |
else if (up->pending) { |
| 1280 |
up->len = 0; |
| 1281 |
WRITE_ONCE(up->pending, 0); |
| 1282 |
ip6_flush_pending_frames(sk); |
| 1283 |
} |
| 1284 |
} |
| 1285 |
|
| 1286 |
static int udpv6_pre_connect(struct sock *sk, struct sockaddr_unsized *uaddr, |
| 1287 |
int addr_len) |
| 1288 |
{ |
| 1289 |
if (addr_len < offsetofend(struct sockaddr, sa_family)) |
| 1290 |
return -EINVAL; |
| 1291 |
/* The following checks are replicated from __ip6_datagram_connect() |
| 1292 |
* and intended to prevent BPF program called below from accessing |
| 1293 |
* bytes that are out of the bound specified by user in addr_len. |
| 1294 |
*/ |
| 1295 |
if (uaddr->sa_family == AF_INET) { |
| 1296 |
if (ipv6_only_sock(sk)) |
| 1297 |
return -EAFNOSUPPORT; |
| 1298 |
return udp_pre_connect(sk, uaddr, addr_len); |
| 1299 |
} |
| 1300 |
|
| 1301 |
if (addr_len < SIN6_LEN_RFC2133) |
| 1302 |
return -EINVAL; |
| 1303 |
|
| 1304 |
return BPF_CGROUP_RUN_PROG_INET6_CONNECT_LOCK(sk, uaddr, &addr_len); |
| 1305 |
} |
| 1306 |
|
| 1307 |
static int udpv6_connect(struct sock *sk, struct sockaddr_unsized *uaddr, |
| 1308 |
int addr_len) |
| 1309 |
{ |
| 1310 |
int res; |
| 1311 |
|
| 1312 |
lock_sock(sk); |
| 1313 |
res = __ip6_datagram_connect(sk, uaddr, addr_len); |
| 1314 |
if (!res) |
| 1315 |
udp6_hash4(sk); |
| 1316 |
release_sock(sk); |
| 1317 |
return res; |
| 1318 |
} |
| 1319 |
|
| 1320 |
/** |
| 1321 |
* udp6_hwcsum_outgoing - handle outgoing HW checksumming |
| 1322 |
* @sk: socket we are sending on |
| 1323 |
* @skb: sk_buff containing the filled-in UDP header |
| 1324 |
* (checksum field must be zeroed out) |
| 1325 |
* @saddr: source address |
| 1326 |
* @daddr: destination address |
| 1327 |
* @len: length of packet |
| 1328 |
*/ |
| 1329 |
static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb, |
| 1330 |
const struct in6_addr *saddr, |
| 1331 |
const struct in6_addr *daddr, int len) |
| 1332 |
{ |
| 1333 |
unsigned int offset; |
| 1334 |
struct udphdr *uh = udp_hdr(skb); |
| 1335 |
struct sk_buff *frags = skb_shinfo(skb)->frag_list; |
| 1336 |
__wsum csum = 0; |
| 1337 |
|
| 1338 |
if (!frags) { |
| 1339 |
/* Only one fragment on the socket. */ |
| 1340 |
skb->csum_start = skb_transport_header(skb) - skb->head; |
| 1341 |
skb->csum_offset = offsetof(struct udphdr, check); |
| 1342 |
uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0); |
| 1343 |
} else { |
| 1344 |
/* |
| 1345 |
* HW-checksum won't work as there are two or more |
| 1346 |
* fragments on the socket so that all csums of sk_buffs |
| 1347 |
* should be together |
| 1348 |
*/ |
| 1349 |
offset = skb_transport_offset(skb); |
| 1350 |
skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); |
| 1351 |
csum = skb->csum; |
| 1352 |
|
| 1353 |
skb->ip_summed = CHECKSUM_NONE; |
| 1354 |
|
| 1355 |
do { |
| 1356 |
csum = csum_add(csum, frags->csum); |
| 1357 |
} while ((frags = frags->next)); |
| 1358 |
|
| 1359 |
uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, |
| 1360 |
csum); |
| 1361 |
if (uh->check == 0) |
| 1362 |
uh->check = CSUM_MANGLED_0; |
| 1363 |
} |
| 1364 |
} |
| 1365 |
|
| 1366 |
/* |
| 1367 |
* Sending |
| 1368 |
*/ |
| 1369 |
|
| 1370 |
static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6, |
| 1371 |
struct inet_cork *cork) |
| 1372 |
{ |
| 1373 |
struct sock *sk = skb->sk; |
| 1374 |
struct udphdr *uh; |
| 1375 |
int err = 0; |
| 1376 |
int is_udplite = IS_UDPLITE(sk); |
| 1377 |
__wsum csum = 0; |
| 1378 |
int offset = skb_transport_offset(skb); |
| 1379 |
int len = skb->len - offset; |
| 1380 |
int datalen = len - sizeof(*uh); |
| 1381 |
|
| 1382 |
/* |
| 1383 |
* Create a UDP header |
| 1384 |
*/ |
| 1385 |
uh = udp_hdr(skb); |
| 1386 |
uh->source = fl6->fl6_sport; |
| 1387 |
uh->dest = fl6->fl6_dport; |
| 1388 |
uh->len = htons(len); |
| 1389 |
uh->check = 0; |
| 1390 |
|
| 1391 |
if (cork->gso_size) { |
| 1392 |
const int hlen = skb_network_header_len(skb) + |
| 1393 |
sizeof(struct udphdr); |
| 1394 |
|
| 1395 |
if (hlen + min(datalen, cork->gso_size) > cork->fragsize) { |
| 1396 |
kfree_skb(skb); |
| 1397 |
return -EMSGSIZE; |
| 1398 |
} |
| 1399 |
if (datalen > cork->gso_size * UDP_MAX_SEGMENTS) { |
| 1400 |
kfree_skb(skb); |
| 1401 |
return -EINVAL; |
| 1402 |
} |
| 1403 |
if (udp_get_no_check6_tx(sk)) { |
| 1404 |
kfree_skb(skb); |
| 1405 |
return -EINVAL; |
| 1406 |
} |
| 1407 |
if (is_udplite || dst_xfrm(skb_dst(skb))) { |
| 1408 |
kfree_skb(skb); |
| 1409 |
return -EIO; |
| 1410 |
} |
| 1411 |
|
| 1412 |
if (datalen > cork->gso_size) { |
| 1413 |
skb_shinfo(skb)->gso_size = cork->gso_size; |
| 1414 |
skb_shinfo(skb)->gso_type = SKB_GSO_UDP_L4; |
| 1415 |
skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(datalen, |
| 1416 |
cork->gso_size); |
| 1417 |
|
| 1418 |
/* Don't checksum the payload, skb will get segmented */ |
| 1419 |
goto csum_partial; |
| 1420 |
} |
| 1421 |
} |
| 1422 |
|
| 1423 |
if (is_udplite) |
| 1424 |
csum = udplite_csum(skb); |
| 1425 |
else if (udp_get_no_check6_tx(sk)) { /* UDP csum disabled */ |
| 1426 |
skb->ip_summed = CHECKSUM_NONE; |
| 1427 |
goto send; |
| 1428 |
} else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ |
| 1429 |
csum_partial: |
| 1430 |
udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len); |
| 1431 |
goto send; |
| 1432 |
} else |
| 1433 |
csum = udp_csum(skb); |
| 1434 |
|
| 1435 |
/* add protocol-dependent pseudo-header */ |
| 1436 |
uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, |
| 1437 |
len, fl6->flowi6_proto, csum); |
| 1438 |
if (uh->check == 0) |
| 1439 |
uh->check = CSUM_MANGLED_0; |
| 1440 |
|
| 1441 |
send: |
| 1442 |
err = ip6_send_skb(skb); |
| 1443 |
if (unlikely(err)) { |
| 1444 |
if (err == -ENOBUFS && !inet6_test_bit(RECVERR6, sk)) { |
| 1445 |
UDP6_INC_STATS(sock_net(sk), |
| 1446 |
UDP_MIB_SNDBUFERRORS, is_udplite); |
| 1447 |
err = 0; |
| 1448 |
} |
| 1449 |
} else { |
| 1450 |
UDP6_INC_STATS(sock_net(sk), |
| 1451 |
UDP_MIB_OUTDATAGRAMS, is_udplite); |
| 1452 |
} |
| 1453 |
return err; |
| 1454 |
} |
| 1455 |
|
| 1456 |
static int udp_v6_push_pending_frames(struct sock *sk) |
| 1457 |
{ |
| 1458 |
struct sk_buff *skb; |
| 1459 |
struct udp_sock *up = udp_sk(sk); |
| 1460 |
int err = 0; |
| 1461 |
|
| 1462 |
if (up->pending == AF_INET) |
| 1463 |
return udp_push_pending_frames(sk); |
| 1464 |
|
| 1465 |
skb = ip6_finish_skb(sk); |
| 1466 |
if (!skb) |
| 1467 |
goto out; |
| 1468 |
|
| 1469 |
err = udp_v6_send_skb(skb, &inet_sk(sk)->cork.fl.u.ip6, |
| 1470 |
&inet_sk(sk)->cork.base); |
| 1471 |
out: |
| 1472 |
up->len = 0; |
| 1473 |
WRITE_ONCE(up->pending, 0); |
| 1474 |
return err; |
| 1475 |
} |
| 1476 |
|
| 1477 |
int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
| 1478 |
{ |
| 1479 |
struct ipv6_txoptions opt_space; |
| 1480 |
struct udp_sock *up = udp_sk(sk); |
| 1481 |
struct inet_sock *inet = inet_sk(sk); |
| 1482 |
struct ipv6_pinfo *np = inet6_sk(sk); |
| 1483 |
DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
| 1484 |
struct in6_addr *daddr, *final_p, final; |
| 1485 |
struct ipv6_txoptions *opt = NULL; |
| 1486 |
struct ipv6_txoptions *opt_to_free = NULL; |
| 1487 |
struct ip6_flowlabel *flowlabel = NULL; |
| 1488 |
struct inet_cork_full cork; |
| 1489 |
struct flowi6 *fl6 = &cork.fl.u.ip6; |
| 1490 |
struct dst_entry *dst; |
| 1491 |
struct ipcm6_cookie ipc6; |
| 1492 |
int addr_len = msg->msg_namelen; |
| 1493 |
bool connected = false; |
| 1494 |
int ulen = len; |
| 1495 |
int corkreq = udp_test_bit(CORK, sk) || msg->msg_flags & MSG_MORE; |
| 1496 |
int err; |
| 1497 |
int is_udplite = IS_UDPLITE(sk); |
| 1498 |
int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); |
| 1499 |
|
| 1500 |
ipcm6_init_sk(&ipc6, sk); |
| 1501 |
ipc6.gso_size = READ_ONCE(up->gso_size); |
| 1502 |
|
| 1503 |
/* destination address check */ |
| 1504 |
if (sin6) { |
| 1505 |
if (addr_len < offsetof(struct sockaddr, sa_data)) |
| 1506 |
return -EINVAL; |
| 1507 |
|
| 1508 |
switch (sin6->sin6_family) { |
| 1509 |
case AF_INET6: |
| 1510 |
if (addr_len < SIN6_LEN_RFC2133) |
| 1511 |
return -EINVAL; |
| 1512 |
daddr = &sin6->sin6_addr; |
| 1513 |
if (ipv6_addr_any(daddr) && |
| 1514 |
ipv6_addr_v4mapped(&np->saddr)) |
| 1515 |
ipv6_addr_set_v4mapped(htonl(INADDR_LOOPBACK), |
| 1516 |
daddr); |
| 1517 |
break; |
| 1518 |
case AF_INET: |
| 1519 |
goto do_udp_sendmsg; |
| 1520 |
case AF_UNSPEC: |
| 1521 |
msg->msg_name = sin6 = NULL; |
| 1522 |
msg->msg_namelen = addr_len = 0; |
| 1523 |
daddr = NULL; |
| 1524 |
break; |
| 1525 |
default: |
| 1526 |
return -EINVAL; |
| 1527 |
} |
| 1528 |
} else if (!READ_ONCE(up->pending)) { |
| 1529 |
if (sk->sk_state != TCP_ESTABLISHED) |
| 1530 |
return -EDESTADDRREQ; |
| 1531 |
daddr = &sk->sk_v6_daddr; |
| 1532 |
} else |
| 1533 |
daddr = NULL; |
| 1534 |
|
| 1535 |
if (daddr) { |
| 1536 |
if (ipv6_addr_v4mapped(daddr)) { |
| 1537 |
struct sockaddr_in sin; |
| 1538 |
sin.sin_family = AF_INET; |
| 1539 |
sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport; |
| 1540 |
sin.sin_addr.s_addr = daddr->s6_addr32[3]; |
| 1541 |
msg->msg_name = &sin; |
| 1542 |
msg->msg_namelen = sizeof(sin); |
| 1543 |
do_udp_sendmsg: |
| 1544 |
err = ipv6_only_sock(sk) ? |
| 1545 |
-ENETUNREACH : udp_sendmsg(sk, msg, len); |
| 1546 |
msg->msg_name = sin6; |
| 1547 |
msg->msg_namelen = addr_len; |
| 1548 |
return err; |
| 1549 |
} |
| 1550 |
} |
| 1551 |
|
| 1552 |
/* Rough check on arithmetic overflow, |
| 1553 |
better check is made in ip6_append_data(). |
| 1554 |
*/ |
| 1555 |
if (len > INT_MAX - sizeof(struct udphdr)) |
| 1556 |
return -EMSGSIZE; |
| 1557 |
|
| 1558 |
getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
| 1559 |
if (READ_ONCE(up->pending)) { |
| 1560 |
if (READ_ONCE(up->pending) == AF_INET) |
| 1561 |
return udp_sendmsg(sk, msg, len); |
| 1562 |
/* |
| 1563 |
* There are pending frames. |
| 1564 |
* The socket lock must be held while it's corked. |
| 1565 |
*/ |
| 1566 |
lock_sock(sk); |
| 1567 |
if (likely(up->pending)) { |
| 1568 |
if (unlikely(up->pending != AF_INET6)) { |
| 1569 |
release_sock(sk); |
| 1570 |
return -EAFNOSUPPORT; |
| 1571 |
} |
| 1572 |
dst = NULL; |
| 1573 |
goto do_append_data; |
| 1574 |
} |
| 1575 |
release_sock(sk); |
| 1576 |
} |
| 1577 |
ulen += sizeof(struct udphdr); |
| 1578 |
|
| 1579 |
memset(fl6, 0, sizeof(*fl6)); |
| 1580 |
|
| 1581 |
if (sin6) { |
| 1582 |
if (sin6->sin6_port == 0) |
| 1583 |
return -EINVAL; |
| 1584 |
|
| 1585 |
fl6->fl6_dport = sin6->sin6_port; |
| 1586 |
daddr = &sin6->sin6_addr; |
| 1587 |
|
| 1588 |
if (inet6_test_bit(SNDFLOW, sk)) { |
| 1589 |
fl6->flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; |
| 1590 |
if (fl6->flowlabel & IPV6_FLOWLABEL_MASK) { |
| 1591 |
flowlabel = fl6_sock_lookup(sk, fl6->flowlabel); |
| 1592 |
if (IS_ERR(flowlabel)) |
| 1593 |
return -EINVAL; |
| 1594 |
} |
| 1595 |
} |
| 1596 |
|
| 1597 |
/* |
| 1598 |
* Otherwise it will be difficult to maintain |
| 1599 |
* sk->sk_dst_cache. |
| 1600 |
*/ |
| 1601 |
if (sk->sk_state == TCP_ESTABLISHED && |
| 1602 |
ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) |
| 1603 |
daddr = &sk->sk_v6_daddr; |
| 1604 |
|
| 1605 |
if (addr_len >= sizeof(struct sockaddr_in6) && |
| 1606 |
sin6->sin6_scope_id && |
| 1607 |
__ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) |
| 1608 |
fl6->flowi6_oif = sin6->sin6_scope_id; |
| 1609 |
} else { |
| 1610 |
if (sk->sk_state != TCP_ESTABLISHED) |
| 1611 |
return -EDESTADDRREQ; |
| 1612 |
|
| 1613 |
fl6->fl6_dport = inet->inet_dport; |
| 1614 |
daddr = &sk->sk_v6_daddr; |
| 1615 |
fl6->flowlabel = np->flow_label; |
| 1616 |
connected = true; |
| 1617 |
} |
| 1618 |
|
| 1619 |
if (!fl6->flowi6_oif) |
| 1620 |
fl6->flowi6_oif = READ_ONCE(sk->sk_bound_dev_if); |
| 1621 |
|
| 1622 |
if (!fl6->flowi6_oif) |
| 1623 |
fl6->flowi6_oif = np->sticky_pktinfo.ipi6_ifindex; |
| 1624 |
|
| 1625 |
fl6->flowi6_uid = sk_uid(sk); |
| 1626 |
|
| 1627 |
if (msg->msg_controllen) { |
| 1628 |
opt = &opt_space; |
| 1629 |
memset(opt, 0, sizeof(struct ipv6_txoptions)); |
| 1630 |
opt->tot_len = sizeof(*opt); |
| 1631 |
ipc6.opt = opt; |
| 1632 |
|
| 1633 |
err = udp_cmsg_send(sk, msg, &ipc6.gso_size); |
| 1634 |
if (err > 0) { |
| 1635 |
err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, fl6, |
| 1636 |
&ipc6); |
| 1637 |
connected = false; |
| 1638 |
} |
| 1639 |
if (err < 0) { |
| 1640 |
fl6_sock_release(flowlabel); |
| 1641 |
return err; |
| 1642 |
} |
| 1643 |
if ((fl6->flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { |
| 1644 |
flowlabel = fl6_sock_lookup(sk, fl6->flowlabel); |
| 1645 |
if (IS_ERR(flowlabel)) |
| 1646 |
return -EINVAL; |
| 1647 |
} |
| 1648 |
if (!(opt->opt_nflen|opt->opt_flen)) |
| 1649 |
opt = NULL; |
| 1650 |
} |
| 1651 |
if (!opt) { |
| 1652 |
opt = txopt_get(np); |
| 1653 |
opt_to_free = opt; |
| 1654 |
} |
| 1655 |
if (flowlabel) |
| 1656 |
opt = fl6_merge_options(&opt_space, flowlabel, opt); |
| 1657 |
opt = ipv6_fixup_options(&opt_space, opt); |
| 1658 |
ipc6.opt = opt; |
| 1659 |
|
| 1660 |
fl6->flowi6_proto = sk->sk_protocol; |
| 1661 |
fl6->flowi6_mark = ipc6.sockc.mark; |
| 1662 |
fl6->daddr = *daddr; |
| 1663 |
if (ipv6_addr_any(&fl6->saddr) && !ipv6_addr_any(&np->saddr)) |
| 1664 |
fl6->saddr = np->saddr; |
| 1665 |
fl6->fl6_sport = inet->inet_sport; |
| 1666 |
|
| 1667 |
if (cgroup_bpf_enabled(CGROUP_UDP6_SENDMSG) && !connected) { |
| 1668 |
err = BPF_CGROUP_RUN_PROG_UDP6_SENDMSG_LOCK(sk, |
| 1669 |
(struct sockaddr *)sin6, |
| 1670 |
&addr_len, |
| 1671 |
&fl6->saddr); |
| 1672 |
if (err) |
| 1673 |
goto out_no_dst; |
| 1674 |
if (sin6) { |
| 1675 |
if (ipv6_addr_v4mapped(&sin6->sin6_addr)) { |
| 1676 |
/* BPF program rewrote IPv6-only by IPv4-mapped |
| 1677 |
* IPv6. It's currently unsupported. |
| 1678 |
*/ |
| 1679 |
err = -ENOTSUPP; |
| 1680 |
goto out_no_dst; |
| 1681 |
} |
| 1682 |
if (sin6->sin6_port == 0) { |
| 1683 |
/* BPF program set invalid port. Reject it. */ |
| 1684 |
err = -EINVAL; |
| 1685 |
goto out_no_dst; |
| 1686 |
} |
| 1687 |
fl6->fl6_dport = sin6->sin6_port; |
| 1688 |
fl6->daddr = sin6->sin6_addr; |
| 1689 |
} |
| 1690 |
} |
| 1691 |
|
| 1692 |
if (ipv6_addr_any(&fl6->daddr)) |
| 1693 |
fl6->daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ |
| 1694 |
|
| 1695 |
final_p = fl6_update_dst(fl6, opt, &final); |
| 1696 |
if (final_p) |
| 1697 |
connected = false; |
| 1698 |
|
| 1699 |
if (!fl6->flowi6_oif && ipv6_addr_is_multicast(&fl6->daddr)) { |
| 1700 |
fl6->flowi6_oif = READ_ONCE(np->mcast_oif); |
| 1701 |
connected = false; |
| 1702 |
} else if (!fl6->flowi6_oif) |
| 1703 |
fl6->flowi6_oif = READ_ONCE(np->ucast_oif); |
| 1704 |
|
| 1705 |
security_sk_classify_flow(sk, flowi6_to_flowi_common(fl6)); |
| 1706 |
|
| 1707 |
fl6->flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6->flowlabel); |
| 1708 |
|
| 1709 |
dst = ip6_sk_dst_lookup_flow(sk, fl6, final_p, connected); |
| 1710 |
if (IS_ERR(dst)) { |
| 1711 |
err = PTR_ERR(dst); |
| 1712 |
dst = NULL; |
| 1713 |
goto out; |
| 1714 |
} |
| 1715 |
|
| 1716 |
if (ipc6.hlimit < 0) |
| 1717 |
ipc6.hlimit = ip6_sk_dst_hoplimit(np, fl6, dst); |
| 1718 |
|
| 1719 |
if (msg->msg_flags&MSG_CONFIRM) |
| 1720 |
goto do_confirm; |
| 1721 |
back_from_confirm: |
| 1722 |
|
| 1723 |
/* Lockless fast path for the non-corking case */ |
| 1724 |
if (!corkreq) { |
| 1725 |
struct sk_buff *skb; |
| 1726 |
|
| 1727 |
skb = ip6_make_skb(sk, getfrag, msg, ulen, |
| 1728 |
sizeof(struct udphdr), &ipc6, |
| 1729 |
dst_rt6_info(dst), |
| 1730 |
msg->msg_flags, &cork); |
| 1731 |
err = PTR_ERR(skb); |
| 1732 |
if (!IS_ERR_OR_NULL(skb)) |
| 1733 |
err = udp_v6_send_skb(skb, fl6, &cork.base); |
| 1734 |
/* ip6_make_skb steals dst reference */ |
| 1735 |
goto out_no_dst; |
| 1736 |
} |
| 1737 |
|
| 1738 |
lock_sock(sk); |
| 1739 |
if (unlikely(up->pending)) { |
| 1740 |
/* The socket is already corked while preparing it. */ |
| 1741 |
/* ... which is an evident application bug. --ANK */ |
| 1742 |
release_sock(sk); |
| 1743 |
|
| 1744 |
net_dbg_ratelimited("udp cork app bug 2\n"); |
| 1745 |
err = -EINVAL; |
| 1746 |
goto out; |
| 1747 |
} |
| 1748 |
|
| 1749 |
WRITE_ONCE(up->pending, AF_INET6); |
| 1750 |
|
| 1751 |
do_append_data: |
| 1752 |
up->len += ulen; |
| 1753 |
err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr), |
| 1754 |
&ipc6, fl6, dst_rt6_info(dst), |
| 1755 |
corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); |
| 1756 |
if (err) |
| 1757 |
udp_v6_flush_pending_frames(sk); |
| 1758 |
else if (!corkreq) |
| 1759 |
err = udp_v6_push_pending_frames(sk); |
| 1760 |
else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) |
| 1761 |
WRITE_ONCE(up->pending, 0); |
| 1762 |
|
| 1763 |
if (err > 0) |
| 1764 |
err = inet6_test_bit(RECVERR6, sk) ? net_xmit_errno(err) : 0; |
| 1765 |
release_sock(sk); |
| 1766 |
|
| 1767 |
out: |
| 1768 |
dst_release(dst); |
| 1769 |
out_no_dst: |
| 1770 |
fl6_sock_release(flowlabel); |
| 1771 |
txopt_put(opt_to_free); |
| 1772 |
if (!err) |
| 1773 |
return len; |
| 1774 |
/* |
| 1775 |
* ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting |
| 1776 |
* ENOBUFS might not be good (it's not tunable per se), but otherwise |
| 1777 |
* we don't have a good statistic (IpOutDiscards but it can be too many |
| 1778 |
* things). We could add another new stat but at least for now that |
| 1779 |
* seems like overkill. |
| 1780 |
*/ |
| 1781 |
if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { |
| 1782 |
UDP6_INC_STATS(sock_net(sk), |
| 1783 |
UDP_MIB_SNDBUFERRORS, is_udplite); |
| 1784 |
} |
| 1785 |
return err; |
| 1786 |
|
| 1787 |
do_confirm: |
| 1788 |
if (msg->msg_flags & MSG_PROBE) |
| 1789 |
dst_confirm_neigh(dst, &fl6->daddr); |
| 1790 |
if (!(msg->msg_flags&MSG_PROBE) || len) |
| 1791 |
goto back_from_confirm; |
| 1792 |
err = 0; |
| 1793 |
goto out; |
| 1794 |
} |
| 1795 |
EXPORT_SYMBOL(udpv6_sendmsg); |
| 1796 |
|
| 1797 |
static void udpv6_splice_eof(struct socket *sock) |
| 1798 |
{ |
| 1799 |
struct sock *sk = sock->sk; |
| 1800 |
struct udp_sock *up = udp_sk(sk); |
| 1801 |
|
| 1802 |
if (!READ_ONCE(up->pending) || udp_test_bit(CORK, sk)) |
| 1803 |
return; |
| 1804 |
|
| 1805 |
lock_sock(sk); |
| 1806 |
if (up->pending && !udp_test_bit(CORK, sk)) |
| 1807 |
udp_v6_push_pending_frames(sk); |
| 1808 |
release_sock(sk); |
| 1809 |
} |
| 1810 |
|
| 1811 |
void udpv6_destroy_sock(struct sock *sk) |
| 1812 |
{ |
| 1813 |
struct udp_sock *up = udp_sk(sk); |
| 1814 |
lock_sock(sk); |
| 1815 |
|
| 1816 |
/* protects from races with udp_abort() */ |
| 1817 |
sock_set_flag(sk, SOCK_DEAD); |
| 1818 |
udp_v6_flush_pending_frames(sk); |
| 1819 |
release_sock(sk); |
| 1820 |
|
| 1821 |
if (static_branch_unlikely(&udpv6_encap_needed_key)) { |
| 1822 |
if (up->encap_type) { |
| 1823 |
void (*encap_destroy)(struct sock *sk); |
| 1824 |
encap_destroy = READ_ONCE(up->encap_destroy); |
| 1825 |
if (encap_destroy) |
| 1826 |
encap_destroy(sk); |
| 1827 |
} |
| 1828 |
if (udp_test_bit(ENCAP_ENABLED, sk)) { |
| 1829 |
static_branch_dec(&udpv6_encap_needed_key); |
| 1830 |
udp_encap_disable(); |
| 1831 |
udp_tunnel_cleanup_gro(sk); |
| 1832 |
} |
| 1833 |
} |
| 1834 |
} |
| 1835 |
|
| 1836 |
/* |
| 1837 |
* Socket option code for UDP |
| 1838 |
*/ |
| 1839 |
int udpv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval, |
| 1840 |
unsigned int optlen) |
| 1841 |
{ |
| 1842 |
if (level == SOL_UDP || level == SOL_UDPLITE || level == SOL_SOCKET) |
| 1843 |
return udp_lib_setsockopt(sk, level, optname, |
| 1844 |
optval, optlen, |
| 1845 |
udp_v6_push_pending_frames); |
| 1846 |
return ipv6_setsockopt(sk, level, optname, optval, optlen); |
| 1847 |
} |
| 1848 |
|
| 1849 |
int udpv6_getsockopt(struct sock *sk, int level, int optname, |
| 1850 |
char __user *optval, int __user *optlen) |
| 1851 |
{ |
| 1852 |
if (level == SOL_UDP || level == SOL_UDPLITE) |
| 1853 |
return udp_lib_getsockopt(sk, level, optname, optval, optlen); |
| 1854 |
return ipv6_getsockopt(sk, level, optname, optval, optlen); |
| 1855 |
} |
| 1856 |
|
| 1857 |
|
| 1858 |
/* ------------------------------------------------------------------------ */ |
| 1859 |
#ifdef CONFIG_PROC_FS |
| 1860 |
int udp6_seq_show(struct seq_file *seq, void *v) |
| 1861 |
{ |
| 1862 |
if (v == SEQ_START_TOKEN) { |
| 1863 |
seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); |
| 1864 |
} else { |
| 1865 |
int bucket = ((struct udp_iter_state *)seq->private)->bucket; |
| 1866 |
const struct inet_sock *inet = inet_sk((const struct sock *)v); |
| 1867 |
__u16 srcp = ntohs(inet->inet_sport); |
| 1868 |
__u16 destp = ntohs(inet->inet_dport); |
| 1869 |
__ip6_dgram_sock_seq_show(seq, v, srcp, destp, |
| 1870 |
udp_rqueue_get(v), bucket); |
| 1871 |
} |
| 1872 |
return 0; |
| 1873 |
} |
| 1874 |
|
| 1875 |
const struct seq_operations udp6_seq_ops = { |
| 1876 |
.start = udp_seq_start, |
| 1877 |
.next = udp_seq_next, |
| 1878 |
.stop = udp_seq_stop, |
| 1879 |
.show = udp6_seq_show, |
| 1880 |
}; |
| 1881 |
EXPORT_SYMBOL(udp6_seq_ops); |
| 1882 |
|
| 1883 |
static struct udp_seq_afinfo udp6_seq_afinfo = { |
| 1884 |
.family = AF_INET6, |
| 1885 |
.udp_table = NULL, |
| 1886 |
}; |
| 1887 |
|
| 1888 |
int __net_init udp6_proc_init(struct net *net) |
| 1889 |
{ |
| 1890 |
if (!proc_create_net_data("udp6", 0444, net->proc_net, &udp6_seq_ops, |
| 1891 |
sizeof(struct udp_iter_state), &udp6_seq_afinfo)) |
| 1892 |
return -ENOMEM; |
| 1893 |
return 0; |
| 1894 |
} |
| 1895 |
|
| 1896 |
void udp6_proc_exit(struct net *net) |
| 1897 |
{ |
| 1898 |
remove_proc_entry("udp6", net->proc_net); |
| 1899 |
} |
| 1900 |
#endif /* CONFIG_PROC_FS */ |
| 1901 |
|
| 1902 |
/* ------------------------------------------------------------------------ */ |
| 1903 |
|
| 1904 |
struct proto udpv6_prot = { |
| 1905 |
.name = "UDPv6", |
| 1906 |
.owner = THIS_MODULE, |
| 1907 |
.close = udp_lib_close, |
| 1908 |
.pre_connect = udpv6_pre_connect, |
| 1909 |
.connect = udpv6_connect, |
| 1910 |
.disconnect = udp_disconnect, |
| 1911 |
.ioctl = udp_ioctl, |
| 1912 |
.init = udpv6_init_sock, |
| 1913 |
.destroy = udpv6_destroy_sock, |
| 1914 |
.setsockopt = udpv6_setsockopt, |
| 1915 |
.getsockopt = udpv6_getsockopt, |
| 1916 |
.sendmsg = udpv6_sendmsg, |
| 1917 |
.recvmsg = udpv6_recvmsg, |
| 1918 |
.splice_eof = udpv6_splice_eof, |
| 1919 |
.release_cb = ip6_datagram_release_cb, |
| 1920 |
.hash = udp_lib_hash, |
| 1921 |
.unhash = udp_lib_unhash, |
| 1922 |
.rehash = udp_v6_rehash, |
| 1923 |
.get_port = udp_v6_get_port, |
| 1924 |
.put_port = udp_lib_unhash, |
| 1925 |
#ifdef CONFIG_BPF_SYSCALL |
| 1926 |
.psock_update_sk_prot = udp_bpf_update_proto, |
| 1927 |
#endif |
| 1928 |
|
| 1929 |
.memory_allocated = &net_aligned_data.udp_memory_allocated, |
| 1930 |
.per_cpu_fw_alloc = &udp_memory_per_cpu_fw_alloc, |
| 1931 |
|
| 1932 |
.sysctl_mem = sysctl_udp_mem, |
| 1933 |
.sysctl_wmem_offset = offsetof(struct net, ipv4.sysctl_udp_wmem_min), |
| 1934 |
.sysctl_rmem_offset = offsetof(struct net, ipv4.sysctl_udp_rmem_min), |
| 1935 |
.obj_size = sizeof(struct udp6_sock), |
| 1936 |
.ipv6_pinfo_offset = offsetof(struct udp6_sock, inet6), |
| 1937 |
.h.udp_table = NULL, |
| 1938 |
.diag_destroy = udp_abort, |
| 1939 |
}; |
| 1940 |
|
| 1941 |
static struct inet_protosw udpv6_protosw = { |
| 1942 |
.type = SOCK_DGRAM, |
| 1943 |
.protocol = IPPROTO_UDP, |
| 1944 |
.prot = &udpv6_prot, |
| 1945 |
.ops = &inet6_dgram_ops, |
| 1946 |
.flags = INET_PROTOSW_PERMANENT, |
| 1947 |
}; |
| 1948 |
|
| 1949 |
int __init udpv6_init(void) |
| 1950 |
{ |
| 1951 |
int ret; |
| 1952 |
|
| 1953 |
net_hotdata.udpv6_protocol = (struct inet6_protocol) { |
| 1954 |
.handler = udpv6_rcv, |
| 1955 |
.err_handler = udpv6_err, |
| 1956 |
.flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL, |
| 1957 |
}; |
| 1958 |
ret = inet6_add_protocol(&net_hotdata.udpv6_protocol, IPPROTO_UDP); |
| 1959 |
if (ret) |
| 1960 |
goto out; |
| 1961 |
|
| 1962 |
ret = inet6_register_protosw(&udpv6_protosw); |
| 1963 |
if (ret) |
| 1964 |
goto out_udpv6_protocol; |
| 1965 |
out: |
| 1966 |
return ret; |
| 1967 |
|
| 1968 |
out_udpv6_protocol: |
| 1969 |
inet6_del_protocol(&net_hotdata.udpv6_protocol, IPPROTO_UDP); |
| 1970 |
goto out; |
| 1971 |
} |
| 1972 |
|
| 1973 |
void udpv6_exit(void) |
| 1974 |
{ |
| 1975 |
inet6_unregister_protosw(&udpv6_protosw); |
| 1976 |
inet6_del_protocol(&net_hotdata.udpv6_protocol, IPPROTO_UDP); |
| 1977 |
} |