]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * The User Datagram Protocol (UDP). | |
7 | * | |
02c30a84 | 8 | * Authors: Ross Biro |
1da177e4 LT |
9 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
10 | * Arnt Gulbrandsen, <agulbra@nvg.unit.no> | |
113aa838 | 11 | * Alan Cox, <alan@lxorguk.ukuu.org.uk> |
1da177e4 LT |
12 | * Hirokazu Takahashi, <taka@valinux.co.jp> |
13 | * | |
14 | * Fixes: | |
15 | * Alan Cox : verify_area() calls | |
16 | * Alan Cox : stopped close while in use off icmp | |
17 | * messages. Not a fix but a botch that | |
18 | * for udp at least is 'valid'. | |
19 | * Alan Cox : Fixed icmp handling properly | |
20 | * Alan Cox : Correct error for oversized datagrams | |
e905a9ed YH |
21 | * Alan Cox : Tidied select() semantics. |
22 | * Alan Cox : udp_err() fixed properly, also now | |
1da177e4 LT |
23 | * select and read wake correctly on errors |
24 | * Alan Cox : udp_send verify_area moved to avoid mem leak | |
25 | * Alan Cox : UDP can count its memory | |
26 | * Alan Cox : send to an unknown connection causes | |
27 | * an ECONNREFUSED off the icmp, but | |
28 | * does NOT close. | |
29 | * Alan Cox : Switched to new sk_buff handlers. No more backlog! | |
30 | * Alan Cox : Using generic datagram code. Even smaller and the PEEK | |
31 | * bug no longer crashes it. | |
32 | * Fred Van Kempen : Net2e support for sk->broadcast. | |
33 | * Alan Cox : Uses skb_free_datagram | |
34 | * Alan Cox : Added get/set sockopt support. | |
35 | * Alan Cox : Broadcasting without option set returns EACCES. | |
36 | * Alan Cox : No wakeup calls. Instead we now use the callbacks. | |
37 | * Alan Cox : Use ip_tos and ip_ttl | |
38 | * Alan Cox : SNMP Mibs | |
39 | * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support. | |
40 | * Matt Dillon : UDP length checks. | |
41 | * Alan Cox : Smarter af_inet used properly. | |
42 | * Alan Cox : Use new kernel side addressing. | |
43 | * Alan Cox : Incorrect return on truncated datagram receive. | |
44 | * Arnt Gulbrandsen : New udp_send and stuff | |
45 | * Alan Cox : Cache last socket | |
46 | * Alan Cox : Route cache | |
47 | * Jon Peatfield : Minor efficiency fix to sendto(). | |
48 | * Mike Shaver : RFC1122 checks. | |
49 | * Alan Cox : Nonblocking error fix. | |
50 | * Willy Konynenberg : Transparent proxying support. | |
51 | * Mike McLagan : Routing by source | |
52 | * David S. Miller : New socket lookup architecture. | |
53 | * Last socket cache retained as it | |
54 | * does have a high hit rate. | |
55 | * Olaf Kirch : Don't linearise iovec on sendmsg. | |
56 | * Andi Kleen : Some cleanups, cache destination entry | |
e905a9ed | 57 | * for connect. |
1da177e4 LT |
58 | * Vitaly E. Lavrov : Transparent proxy revived after year coma. |
59 | * Melvin Smith : Check msg_name not msg_namelen in sendto(), | |
60 | * return ENOTCONN for unconnected sockets (POSIX) | |
61 | * Janos Farkas : don't deliver multi/broadcasts to a different | |
62 | * bound-to-device socket | |
63 | * Hirokazu Takahashi : HW checksumming for outgoing UDP | |
64 | * datagrams. | |
65 | * Hirokazu Takahashi : sendfile() on UDP works now. | |
66 | * Arnaldo C. Melo : convert /proc/net/udp to seq_file | |
67 | * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which | |
68 | * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind | |
69 | * a single port at the same time. | |
70 | * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support | |
342f0234 | 71 | * James Chapman : Add L2TP encapsulation type. |
1da177e4 LT |
72 | * |
73 | * | |
74 | * This program is free software; you can redistribute it and/or | |
75 | * modify it under the terms of the GNU General Public License | |
76 | * as published by the Free Software Foundation; either version | |
77 | * 2 of the License, or (at your option) any later version. | |
78 | */ | |
e905a9ed | 79 | |
afd46503 JP |
80 | #define pr_fmt(fmt) "UDP: " fmt |
81 | ||
7c0f6ba6 | 82 | #include <linux/uaccess.h> |
1da177e4 | 83 | #include <asm/ioctls.h> |
95766fff | 84 | #include <linux/bootmem.h> |
8203efb3 ED |
85 | #include <linux/highmem.h> |
86 | #include <linux/swap.h> | |
1da177e4 LT |
87 | #include <linux/types.h> |
88 | #include <linux/fcntl.h> | |
89 | #include <linux/module.h> | |
90 | #include <linux/socket.h> | |
91 | #include <linux/sockios.h> | |
14c85021 | 92 | #include <linux/igmp.h> |
6e540309 | 93 | #include <linux/inetdevice.h> |
1da177e4 LT |
94 | #include <linux/in.h> |
95 | #include <linux/errno.h> | |
96 | #include <linux/timer.h> | |
97 | #include <linux/mm.h> | |
1da177e4 | 98 | #include <linux/inet.h> |
1da177e4 | 99 | #include <linux/netdevice.h> |
5a0e3ad6 | 100 | #include <linux/slab.h> |
c752f073 | 101 | #include <net/tcp_states.h> |
1da177e4 LT |
102 | #include <linux/skbuff.h> |
103 | #include <linux/proc_fs.h> | |
104 | #include <linux/seq_file.h> | |
457c4cbc | 105 | #include <net/net_namespace.h> |
1da177e4 | 106 | #include <net/icmp.h> |
421b3885 | 107 | #include <net/inet_hashtables.h> |
1da177e4 | 108 | #include <net/route.h> |
1da177e4 LT |
109 | #include <net/checksum.h> |
110 | #include <net/xfrm.h> | |
296f7ea7 | 111 | #include <trace/events/udp.h> |
447167bf | 112 | #include <linux/static_key.h> |
22911fc5 | 113 | #include <trace/events/skb.h> |
076bb0c8 | 114 | #include <net/busy_poll.h> |
ba4e58ec | 115 | #include "udp_impl.h" |
e32ea7e7 | 116 | #include <net/sock_reuseport.h> |
217375a0 | 117 | #include <net/addrconf.h> |
1da177e4 | 118 | |
f86dcc5a | 119 | struct udp_table udp_table __read_mostly; |
645ca708 | 120 | EXPORT_SYMBOL(udp_table); |
1da177e4 | 121 | |
8d987e5c | 122 | long sysctl_udp_mem[3] __read_mostly; |
95766fff | 123 | EXPORT_SYMBOL(sysctl_udp_mem); |
c482c568 ED |
124 | |
125 | int sysctl_udp_rmem_min __read_mostly; | |
95766fff | 126 | EXPORT_SYMBOL(sysctl_udp_rmem_min); |
c482c568 ED |
127 | |
128 | int sysctl_udp_wmem_min __read_mostly; | |
95766fff HA |
129 | EXPORT_SYMBOL(sysctl_udp_wmem_min); |
130 | ||
8d987e5c | 131 | atomic_long_t udp_memory_allocated; |
95766fff HA |
132 | EXPORT_SYMBOL(udp_memory_allocated); |
133 | ||
f86dcc5a ED |
134 | #define MAX_UDP_PORTS 65536 |
135 | #define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN) | |
98322f22 | 136 | |
f24d43c0 | 137 | static int udp_lib_lport_inuse(struct net *net, __u16 num, |
645ca708 | 138 | const struct udp_hslot *hslot, |
98322f22 | 139 | unsigned long *bitmap, |
f24d43c0 ED |
140 | struct sock *sk, |
141 | int (*saddr_comp)(const struct sock *sk1, | |
e32ea7e7 CG |
142 | const struct sock *sk2, |
143 | bool match_wildcard), | |
f86dcc5a | 144 | unsigned int log) |
1da177e4 | 145 | { |
f24d43c0 | 146 | struct sock *sk2; |
ba418fa3 | 147 | kuid_t uid = sock_i_uid(sk); |
25030a7f | 148 | |
ca065d0c | 149 | sk_for_each(sk2, &hslot->head) { |
9d4fb27d JP |
150 | if (net_eq(sock_net(sk2), net) && |
151 | sk2 != sk && | |
d4cada4a | 152 | (bitmap || udp_sk(sk2)->udp_port_hash == num) && |
9d4fb27d JP |
153 | (!sk2->sk_reuse || !sk->sk_reuse) && |
154 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
155 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
e32ea7e7 | 156 | saddr_comp(sk, sk2, true)) { |
df560056 EG |
157 | if (sk2->sk_reuseport && sk->sk_reuseport && |
158 | !rcu_access_pointer(sk->sk_reuseport_cb) && | |
159 | uid_eq(uid, sock_i_uid(sk2))) { | |
160 | if (!bitmap) | |
161 | return 0; | |
162 | } else { | |
163 | if (!bitmap) | |
164 | return 1; | |
165 | __set_bit(udp_sk(sk2)->udp_port_hash >> log, | |
166 | bitmap); | |
167 | } | |
98322f22 | 168 | } |
4243cdc2 | 169 | } |
25030a7f GR |
170 | return 0; |
171 | } | |
172 | ||
30fff923 ED |
173 | /* |
174 | * Note: we still hold spinlock of primary hash chain, so no other writer | |
175 | * can insert/delete a socket with local_port == num | |
176 | */ | |
177 | static int udp_lib_lport_inuse2(struct net *net, __u16 num, | |
4243cdc2 JP |
178 | struct udp_hslot *hslot2, |
179 | struct sock *sk, | |
180 | int (*saddr_comp)(const struct sock *sk1, | |
e32ea7e7 CG |
181 | const struct sock *sk2, |
182 | bool match_wildcard)) | |
30fff923 ED |
183 | { |
184 | struct sock *sk2; | |
ba418fa3 | 185 | kuid_t uid = sock_i_uid(sk); |
30fff923 ED |
186 | int res = 0; |
187 | ||
188 | spin_lock(&hslot2->lock); | |
ca065d0c | 189 | udp_portaddr_for_each_entry(sk2, &hslot2->head) { |
9d4fb27d JP |
190 | if (net_eq(sock_net(sk2), net) && |
191 | sk2 != sk && | |
192 | (udp_sk(sk2)->udp_port_hash == num) && | |
193 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
194 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
195 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
e32ea7e7 | 196 | saddr_comp(sk, sk2, true)) { |
df560056 EG |
197 | if (sk2->sk_reuseport && sk->sk_reuseport && |
198 | !rcu_access_pointer(sk->sk_reuseport_cb) && | |
199 | uid_eq(uid, sock_i_uid(sk2))) { | |
200 | res = 0; | |
201 | } else { | |
202 | res = 1; | |
203 | } | |
30fff923 ED |
204 | break; |
205 | } | |
4243cdc2 | 206 | } |
30fff923 ED |
207 | spin_unlock(&hslot2->lock); |
208 | return res; | |
209 | } | |
210 | ||
e32ea7e7 CG |
211 | static int udp_reuseport_add_sock(struct sock *sk, struct udp_hslot *hslot, |
212 | int (*saddr_same)(const struct sock *sk1, | |
213 | const struct sock *sk2, | |
214 | bool match_wildcard)) | |
215 | { | |
216 | struct net *net = sock_net(sk); | |
e32ea7e7 CG |
217 | kuid_t uid = sock_i_uid(sk); |
218 | struct sock *sk2; | |
219 | ||
ca065d0c | 220 | sk_for_each(sk2, &hslot->head) { |
e32ea7e7 CG |
221 | if (net_eq(sock_net(sk2), net) && |
222 | sk2 != sk && | |
223 | sk2->sk_family == sk->sk_family && | |
224 | ipv6_only_sock(sk2) == ipv6_only_sock(sk) && | |
225 | (udp_sk(sk2)->udp_port_hash == udp_sk(sk)->udp_port_hash) && | |
226 | (sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
227 | sk2->sk_reuseport && uid_eq(uid, sock_i_uid(sk2)) && | |
228 | (*saddr_same)(sk, sk2, false)) { | |
229 | return reuseport_add_sock(sk, sk2); | |
230 | } | |
231 | } | |
232 | ||
233 | /* Initial allocation may have already happened via setsockopt */ | |
234 | if (!rcu_access_pointer(sk->sk_reuseport_cb)) | |
235 | return reuseport_alloc(sk); | |
236 | return 0; | |
237 | } | |
238 | ||
25030a7f | 239 | /** |
6ba5a3c5 | 240 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
241 | * |
242 | * @sk: socket struct in question | |
243 | * @snum: port number to look up | |
df2bc459 | 244 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25985edc | 245 | * @hash2_nulladdr: AF-dependent hash value in secondary hash chains, |
30fff923 | 246 | * with NULL address |
25030a7f | 247 | */ |
6ba5a3c5 | 248 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
4243cdc2 | 249 | int (*saddr_comp)(const struct sock *sk1, |
e32ea7e7 CG |
250 | const struct sock *sk2, |
251 | bool match_wildcard), | |
30fff923 | 252 | unsigned int hash2_nulladdr) |
25030a7f | 253 | { |
512615b6 | 254 | struct udp_hslot *hslot, *hslot2; |
645ca708 | 255 | struct udp_table *udptable = sk->sk_prot->h.udp_table; |
25030a7f | 256 | int error = 1; |
3b1e0a65 | 257 | struct net *net = sock_net(sk); |
1da177e4 | 258 | |
32c1da70 | 259 | if (!snum) { |
9088c560 | 260 | int low, high, remaining; |
95c96174 | 261 | unsigned int rand; |
98322f22 ED |
262 | unsigned short first, last; |
263 | DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN); | |
32c1da70 | 264 | |
0bbf87d8 | 265 | inet_get_local_port_range(net, &low, &high); |
a25de534 | 266 | remaining = (high - low) + 1; |
227b60f5 | 267 | |
63862b5b | 268 | rand = prandom_u32(); |
8fc54f68 | 269 | first = reciprocal_scale(rand, remaining) + low; |
98322f22 ED |
270 | /* |
271 | * force rand to be an odd multiple of UDP_HTABLE_SIZE | |
272 | */ | |
f86dcc5a | 273 | rand = (rand | 1) * (udptable->mask + 1); |
5781b235 ED |
274 | last = first + udptable->mask + 1; |
275 | do { | |
f86dcc5a | 276 | hslot = udp_hashslot(udptable, net, first); |
98322f22 | 277 | bitmap_zero(bitmap, PORTS_PER_CHAIN); |
645ca708 | 278 | spin_lock_bh(&hslot->lock); |
98322f22 | 279 | udp_lib_lport_inuse(net, snum, hslot, bitmap, sk, |
f86dcc5a | 280 | saddr_comp, udptable->log); |
98322f22 ED |
281 | |
282 | snum = first; | |
283 | /* | |
284 | * Iterate on all possible values of snum for this hash. | |
285 | * Using steps of an odd multiple of UDP_HTABLE_SIZE | |
286 | * give us randomization and full range coverage. | |
287 | */ | |
9088c560 | 288 | do { |
98322f22 | 289 | if (low <= snum && snum <= high && |
e3826f1e | 290 | !test_bit(snum >> udptable->log, bitmap) && |
122ff243 | 291 | !inet_is_local_reserved_port(net, snum)) |
98322f22 ED |
292 | goto found; |
293 | snum += rand; | |
294 | } while (snum != first); | |
295 | spin_unlock_bh(&hslot->lock); | |
df560056 | 296 | cond_resched(); |
5781b235 | 297 | } while (++first != last); |
98322f22 | 298 | goto fail; |
645ca708 | 299 | } else { |
f86dcc5a | 300 | hslot = udp_hashslot(udptable, net, snum); |
645ca708 | 301 | spin_lock_bh(&hslot->lock); |
30fff923 ED |
302 | if (hslot->count > 10) { |
303 | int exist; | |
304 | unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum; | |
305 | ||
306 | slot2 &= udptable->mask; | |
307 | hash2_nulladdr &= udptable->mask; | |
308 | ||
309 | hslot2 = udp_hashslot2(udptable, slot2); | |
310 | if (hslot->count < hslot2->count) | |
311 | goto scan_primary_hash; | |
312 | ||
313 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
314 | sk, saddr_comp); | |
315 | if (!exist && (hash2_nulladdr != slot2)) { | |
316 | hslot2 = udp_hashslot2(udptable, hash2_nulladdr); | |
317 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
318 | sk, saddr_comp); | |
319 | } | |
320 | if (exist) | |
321 | goto fail_unlock; | |
322 | else | |
323 | goto found; | |
324 | } | |
325 | scan_primary_hash: | |
f86dcc5a ED |
326 | if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk, |
327 | saddr_comp, 0)) | |
645ca708 ED |
328 | goto fail_unlock; |
329 | } | |
98322f22 | 330 | found: |
c720c7e8 | 331 | inet_sk(sk)->inet_num = snum; |
d4cada4a ED |
332 | udp_sk(sk)->udp_port_hash = snum; |
333 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
1da177e4 | 334 | if (sk_unhashed(sk)) { |
e32ea7e7 CG |
335 | if (sk->sk_reuseport && |
336 | udp_reuseport_add_sock(sk, hslot, saddr_comp)) { | |
337 | inet_sk(sk)->inet_num = 0; | |
338 | udp_sk(sk)->udp_port_hash = 0; | |
339 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
340 | goto fail_unlock; | |
341 | } | |
342 | ||
ca065d0c | 343 | sk_add_node_rcu(sk, &hslot->head); |
fdcc8aa9 | 344 | hslot->count++; |
c29a0bc4 | 345 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
512615b6 ED |
346 | |
347 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
348 | spin_lock(&hslot2->lock); | |
d894ba18 | 349 | if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport && |
1602f49b DM |
350 | sk->sk_family == AF_INET6) |
351 | hlist_add_tail_rcu(&udp_sk(sk)->udp_portaddr_node, | |
352 | &hslot2->head); | |
d894ba18 | 353 | else |
1602f49b DM |
354 | hlist_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, |
355 | &hslot2->head); | |
512615b6 ED |
356 | hslot2->count++; |
357 | spin_unlock(&hslot2->lock); | |
1da177e4 | 358 | } |
ca065d0c | 359 | sock_set_flag(sk, SOCK_RCU_FREE); |
25030a7f | 360 | error = 0; |
645ca708 ED |
361 | fail_unlock: |
362 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 363 | fail: |
25030a7f GR |
364 | return error; |
365 | } | |
c482c568 | 366 | EXPORT_SYMBOL(udp_lib_get_port); |
25030a7f | 367 | |
e32ea7e7 CG |
368 | /* match_wildcard == true: 0.0.0.0 equals to any IPv4 addresses |
369 | * match_wildcard == false: addresses must be exactly the same, i.e. | |
370 | * 0.0.0.0 only equals to 0.0.0.0 | |
371 | */ | |
c125e80b CG |
372 | int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2, |
373 | bool match_wildcard) | |
db8dac20 DM |
374 | { |
375 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
376 | ||
e32ea7e7 CG |
377 | if (!ipv6_only_sock(sk2)) { |
378 | if (inet1->inet_rcv_saddr == inet2->inet_rcv_saddr) | |
379 | return 1; | |
380 | if (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr) | |
381 | return match_wildcard; | |
382 | } | |
383 | return 0; | |
db8dac20 DM |
384 | } |
385 | ||
6eada011 ED |
386 | static u32 udp4_portaddr_hash(const struct net *net, __be32 saddr, |
387 | unsigned int port) | |
d4cada4a | 388 | { |
0eae88f3 | 389 | return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port; |
d4cada4a ED |
390 | } |
391 | ||
6ba5a3c5 | 392 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 393 | { |
30fff923 | 394 | unsigned int hash2_nulladdr = |
0eae88f3 | 395 | udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum); |
30fff923 ED |
396 | unsigned int hash2_partial = |
397 | udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0); | |
398 | ||
d4cada4a | 399 | /* precompute partial secondary hash */ |
30fff923 ED |
400 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
401 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr); | |
db8dac20 DM |
402 | } |
403 | ||
d1e37288 SX |
404 | static int compute_score(struct sock *sk, struct net *net, |
405 | __be32 saddr, __be16 sport, | |
406 | __be32 daddr, unsigned short hnum, int dif) | |
645ca708 | 407 | { |
60c04aec JP |
408 | int score; |
409 | struct inet_sock *inet; | |
645ca708 | 410 | |
60c04aec JP |
411 | if (!net_eq(sock_net(sk), net) || |
412 | udp_sk(sk)->udp_port_hash != hnum || | |
413 | ipv6_only_sock(sk)) | |
414 | return -1; | |
645ca708 | 415 | |
60c04aec JP |
416 | score = (sk->sk_family == PF_INET) ? 2 : 1; |
417 | inet = inet_sk(sk); | |
418 | ||
419 | if (inet->inet_rcv_saddr) { | |
420 | if (inet->inet_rcv_saddr != daddr) | |
421 | return -1; | |
422 | score += 4; | |
645ca708 | 423 | } |
60c04aec JP |
424 | |
425 | if (inet->inet_daddr) { | |
426 | if (inet->inet_daddr != saddr) | |
427 | return -1; | |
428 | score += 4; | |
429 | } | |
430 | ||
431 | if (inet->inet_dport) { | |
432 | if (inet->inet_dport != sport) | |
433 | return -1; | |
434 | score += 4; | |
435 | } | |
436 | ||
437 | if (sk->sk_bound_dev_if) { | |
438 | if (sk->sk_bound_dev_if != dif) | |
439 | return -1; | |
440 | score += 4; | |
441 | } | |
70da268b ED |
442 | if (sk->sk_incoming_cpu == raw_smp_processor_id()) |
443 | score++; | |
645ca708 ED |
444 | return score; |
445 | } | |
446 | ||
6eada011 ED |
447 | static u32 udp_ehashfn(const struct net *net, const __be32 laddr, |
448 | const __u16 lport, const __be32 faddr, | |
449 | const __be16 fport) | |
65cd8033 | 450 | { |
1bbdceef HFS |
451 | static u32 udp_ehash_secret __read_mostly; |
452 | ||
453 | net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret)); | |
454 | ||
65cd8033 | 455 | return __inet_ehashfn(laddr, lport, faddr, fport, |
1bbdceef | 456 | udp_ehash_secret + net_hash_mix(net)); |
65cd8033 HFS |
457 | } |
458 | ||
d1e37288 | 459 | /* called with rcu_read_lock() */ |
5051ebd2 ED |
460 | static struct sock *udp4_lib_lookup2(struct net *net, |
461 | __be32 saddr, __be16 sport, | |
462 | __be32 daddr, unsigned int hnum, int dif, | |
d1e37288 | 463 | struct udp_hslot *hslot2, |
1134158b | 464 | struct sk_buff *skb) |
5051ebd2 ED |
465 | { |
466 | struct sock *sk, *result; | |
ba418fa3 TH |
467 | int score, badness, matches = 0, reuseport = 0; |
468 | u32 hash = 0; | |
5051ebd2 | 469 | |
5051ebd2 | 470 | result = NULL; |
ba418fa3 | 471 | badness = 0; |
ca065d0c | 472 | udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
d1e37288 | 473 | score = compute_score(sk, net, saddr, sport, |
5051ebd2 ED |
474 | daddr, hnum, dif); |
475 | if (score > badness) { | |
ba418fa3 TH |
476 | reuseport = sk->sk_reuseport; |
477 | if (reuseport) { | |
65cd8033 HFS |
478 | hash = udp_ehashfn(net, daddr, hnum, |
479 | saddr, sport); | |
ca065d0c | 480 | result = reuseport_select_sock(sk, hash, skb, |
ed0dfffd | 481 | sizeof(struct udphdr)); |
ca065d0c ED |
482 | if (result) |
483 | return result; | |
ba418fa3 TH |
484 | matches = 1; |
485 | } | |
ca065d0c ED |
486 | badness = score; |
487 | result = sk; | |
ba418fa3 TH |
488 | } else if (score == badness && reuseport) { |
489 | matches++; | |
8fc54f68 | 490 | if (reciprocal_scale(hash, matches) == 0) |
ba418fa3 TH |
491 | result = sk; |
492 | hash = next_pseudo_random32(hash); | |
5051ebd2 ED |
493 | } |
494 | } | |
5051ebd2 ED |
495 | return result; |
496 | } | |
497 | ||
db8dac20 DM |
498 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
499 | * harder than this. -DaveM | |
500 | */ | |
fce82338 | 501 | struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, |
db8dac20 | 502 | __be16 sport, __be32 daddr, __be16 dport, |
538950a1 | 503 | int dif, struct udp_table *udptable, struct sk_buff *skb) |
db8dac20 | 504 | { |
271b72c7 | 505 | struct sock *sk, *result; |
db8dac20 | 506 | unsigned short hnum = ntohs(dport); |
5051ebd2 ED |
507 | unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); |
508 | struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; | |
ba418fa3 TH |
509 | int score, badness, matches = 0, reuseport = 0; |
510 | u32 hash = 0; | |
645ca708 | 511 | |
5051ebd2 ED |
512 | if (hslot->count > 10) { |
513 | hash2 = udp4_portaddr_hash(net, daddr, hnum); | |
514 | slot2 = hash2 & udptable->mask; | |
515 | hslot2 = &udptable->hash2[slot2]; | |
516 | if (hslot->count < hslot2->count) | |
517 | goto begin; | |
518 | ||
519 | result = udp4_lib_lookup2(net, saddr, sport, | |
520 | daddr, hnum, dif, | |
d1e37288 | 521 | hslot2, skb); |
5051ebd2 | 522 | if (!result) { |
d1e37288 | 523 | unsigned int old_slot2 = slot2; |
0eae88f3 | 524 | hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum); |
5051ebd2 | 525 | slot2 = hash2 & udptable->mask; |
d1e37288 SX |
526 | /* avoid searching the same slot again. */ |
527 | if (unlikely(slot2 == old_slot2)) | |
528 | return result; | |
529 | ||
5051ebd2 ED |
530 | hslot2 = &udptable->hash2[slot2]; |
531 | if (hslot->count < hslot2->count) | |
532 | goto begin; | |
533 | ||
1223c67c | 534 | result = udp4_lib_lookup2(net, saddr, sport, |
d1e37288 SX |
535 | daddr, hnum, dif, |
536 | hslot2, skb); | |
5051ebd2 | 537 | } |
5051ebd2 ED |
538 | return result; |
539 | } | |
271b72c7 ED |
540 | begin: |
541 | result = NULL; | |
ba418fa3 | 542 | badness = 0; |
ca065d0c | 543 | sk_for_each_rcu(sk, &hslot->head) { |
d1e37288 SX |
544 | score = compute_score(sk, net, saddr, sport, |
545 | daddr, hnum, dif); | |
645ca708 | 546 | if (score > badness) { |
ba418fa3 TH |
547 | reuseport = sk->sk_reuseport; |
548 | if (reuseport) { | |
65cd8033 HFS |
549 | hash = udp_ehashfn(net, daddr, hnum, |
550 | saddr, sport); | |
ca065d0c | 551 | result = reuseport_select_sock(sk, hash, skb, |
538950a1 | 552 | sizeof(struct udphdr)); |
ca065d0c ED |
553 | if (result) |
554 | return result; | |
ba418fa3 TH |
555 | matches = 1; |
556 | } | |
ca065d0c ED |
557 | result = sk; |
558 | badness = score; | |
ba418fa3 TH |
559 | } else if (score == badness && reuseport) { |
560 | matches++; | |
8fc54f68 | 561 | if (reciprocal_scale(hash, matches) == 0) |
ba418fa3 TH |
562 | result = sk; |
563 | hash = next_pseudo_random32(hash); | |
db8dac20 DM |
564 | } |
565 | } | |
db8dac20 DM |
566 | return result; |
567 | } | |
fce82338 | 568 | EXPORT_SYMBOL_GPL(__udp4_lib_lookup); |
db8dac20 | 569 | |
607c4aaf KK |
570 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
571 | __be16 sport, __be16 dport, | |
645ca708 | 572 | struct udp_table *udptable) |
607c4aaf KK |
573 | { |
574 | const struct iphdr *iph = ip_hdr(skb); | |
575 | ||
ed7cbbce | 576 | return __udp4_lib_lookup(dev_net(skb->dev), iph->saddr, sport, |
8afdd99a | 577 | iph->daddr, dport, inet_iif(skb), |
538950a1 | 578 | udptable, skb); |
607c4aaf KK |
579 | } |
580 | ||
63058308 TH |
581 | struct sock *udp4_lib_lookup_skb(struct sk_buff *skb, |
582 | __be16 sport, __be16 dport) | |
583 | { | |
ed7cbbce | 584 | return __udp4_lib_lookup_skb(skb, sport, dport, &udp_table); |
63058308 TH |
585 | } |
586 | EXPORT_SYMBOL_GPL(udp4_lib_lookup_skb); | |
587 | ||
ca065d0c ED |
588 | /* Must be called under rcu_read_lock(). |
589 | * Does increment socket refcount. | |
590 | */ | |
591 | #if IS_ENABLED(CONFIG_NETFILTER_XT_MATCH_SOCKET) || \ | |
30f58158 AB |
592 | IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY) || \ |
593 | IS_ENABLED(CONFIG_NF_SOCKET_IPV4) | |
bcd41303 KK |
594 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
595 | __be32 daddr, __be16 dport, int dif) | |
596 | { | |
ca065d0c ED |
597 | struct sock *sk; |
598 | ||
599 | sk = __udp4_lib_lookup(net, saddr, sport, daddr, dport, | |
600 | dif, &udp_table, NULL); | |
601 | if (sk && !atomic_inc_not_zero(&sk->sk_refcnt)) | |
602 | sk = NULL; | |
603 | return sk; | |
bcd41303 KK |
604 | } |
605 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
ca065d0c | 606 | #endif |
bcd41303 | 607 | |
421b3885 SB |
608 | static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk, |
609 | __be16 loc_port, __be32 loc_addr, | |
610 | __be16 rmt_port, __be32 rmt_addr, | |
611 | int dif, unsigned short hnum) | |
612 | { | |
613 | struct inet_sock *inet = inet_sk(sk); | |
614 | ||
615 | if (!net_eq(sock_net(sk), net) || | |
616 | udp_sk(sk)->udp_port_hash != hnum || | |
617 | (inet->inet_daddr && inet->inet_daddr != rmt_addr) || | |
618 | (inet->inet_dport != rmt_port && inet->inet_dport) || | |
619 | (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) || | |
620 | ipv6_only_sock(sk) || | |
621 | (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) | |
622 | return false; | |
623 | if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif)) | |
624 | return false; | |
625 | return true; | |
626 | } | |
627 | ||
db8dac20 DM |
628 | /* |
629 | * This routine is called by the ICMP module when it gets some | |
630 | * sort of error condition. If err < 0 then the socket should | |
631 | * be closed and the error returned to the user. If err > 0 | |
632 | * it's just the icmp type << 8 | icmp code. | |
633 | * Header points to the ip header of the error packet. We move | |
634 | * on past this. Then (as it used to claim before adjustment) | |
635 | * header points to the first 8 bytes of the udp header. We need | |
636 | * to find the appropriate port. | |
637 | */ | |
638 | ||
645ca708 | 639 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
640 | { |
641 | struct inet_sock *inet; | |
b71d1d42 | 642 | const struct iphdr *iph = (const struct iphdr *)skb->data; |
c482c568 | 643 | struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2)); |
db8dac20 DM |
644 | const int type = icmp_hdr(skb)->type; |
645 | const int code = icmp_hdr(skb)->code; | |
646 | struct sock *sk; | |
647 | int harderr; | |
648 | int err; | |
fd54d716 | 649 | struct net *net = dev_net(skb->dev); |
db8dac20 | 650 | |
fd54d716 | 651 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
538950a1 CG |
652 | iph->saddr, uh->source, skb->dev->ifindex, udptable, |
653 | NULL); | |
51456b29 | 654 | if (!sk) { |
5d3848bc | 655 | __ICMP_INC_STATS(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
656 | return; /* No socket for error */ |
657 | } | |
658 | ||
659 | err = 0; | |
660 | harderr = 0; | |
661 | inet = inet_sk(sk); | |
662 | ||
663 | switch (type) { | |
664 | default: | |
665 | case ICMP_TIME_EXCEEDED: | |
666 | err = EHOSTUNREACH; | |
667 | break; | |
668 | case ICMP_SOURCE_QUENCH: | |
669 | goto out; | |
670 | case ICMP_PARAMETERPROB: | |
671 | err = EPROTO; | |
672 | harderr = 1; | |
673 | break; | |
674 | case ICMP_DEST_UNREACH: | |
675 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
36393395 | 676 | ipv4_sk_update_pmtu(skb, sk, info); |
db8dac20 DM |
677 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { |
678 | err = EMSGSIZE; | |
679 | harderr = 1; | |
680 | break; | |
681 | } | |
682 | goto out; | |
683 | } | |
684 | err = EHOSTUNREACH; | |
685 | if (code <= NR_ICMP_UNREACH) { | |
686 | harderr = icmp_err_convert[code].fatal; | |
687 | err = icmp_err_convert[code].errno; | |
688 | } | |
689 | break; | |
55be7a9c DM |
690 | case ICMP_REDIRECT: |
691 | ipv4_sk_redirect(skb, sk); | |
1a462d18 | 692 | goto out; |
db8dac20 DM |
693 | } |
694 | ||
695 | /* | |
696 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
697 | * 4.1.3.3. | |
698 | */ | |
699 | if (!inet->recverr) { | |
700 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
701 | goto out; | |
b1faf566 | 702 | } else |
c482c568 | 703 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1)); |
b1faf566 | 704 | |
db8dac20 DM |
705 | sk->sk_err = err; |
706 | sk->sk_error_report(sk); | |
707 | out: | |
ca065d0c | 708 | return; |
db8dac20 DM |
709 | } |
710 | ||
711 | void udp_err(struct sk_buff *skb, u32 info) | |
712 | { | |
645ca708 | 713 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
714 | } |
715 | ||
716 | /* | |
717 | * Throw away all pending data and cancel the corking. Socket is locked. | |
718 | */ | |
36d926b9 | 719 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
720 | { |
721 | struct udp_sock *up = udp_sk(sk); | |
722 | ||
723 | if (up->pending) { | |
724 | up->len = 0; | |
725 | up->pending = 0; | |
726 | ip_flush_pending_frames(sk); | |
727 | } | |
728 | } | |
36d926b9 | 729 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
730 | |
731 | /** | |
f6b9664f | 732 | * udp4_hwcsum - handle outgoing HW checksumming |
db8dac20 DM |
733 | * @skb: sk_buff containing the filled-in UDP header |
734 | * (checksum field must be zeroed out) | |
f6b9664f HX |
735 | * @src: source IP address |
736 | * @dst: destination IP address | |
db8dac20 | 737 | */ |
c26bf4a5 | 738 | void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst) |
db8dac20 | 739 | { |
db8dac20 | 740 | struct udphdr *uh = udp_hdr(skb); |
f6b9664f HX |
741 | int offset = skb_transport_offset(skb); |
742 | int len = skb->len - offset; | |
743 | int hlen = len; | |
db8dac20 DM |
744 | __wsum csum = 0; |
745 | ||
ebbe495f | 746 | if (!skb_has_frag_list(skb)) { |
db8dac20 DM |
747 | /* |
748 | * Only one fragment on the socket. | |
749 | */ | |
750 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
751 | skb->csum_offset = offsetof(struct udphdr, check); | |
f6b9664f HX |
752 | uh->check = ~csum_tcpudp_magic(src, dst, len, |
753 | IPPROTO_UDP, 0); | |
db8dac20 | 754 | } else { |
ebbe495f WC |
755 | struct sk_buff *frags; |
756 | ||
db8dac20 DM |
757 | /* |
758 | * HW-checksum won't work as there are two or more | |
759 | * fragments on the socket so that all csums of sk_buffs | |
760 | * should be together | |
761 | */ | |
ebbe495f | 762 | skb_walk_frags(skb, frags) { |
f6b9664f HX |
763 | csum = csum_add(csum, frags->csum); |
764 | hlen -= frags->len; | |
ebbe495f | 765 | } |
db8dac20 | 766 | |
f6b9664f | 767 | csum = skb_checksum(skb, offset, hlen, csum); |
db8dac20 DM |
768 | skb->ip_summed = CHECKSUM_NONE; |
769 | ||
db8dac20 DM |
770 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); |
771 | if (uh->check == 0) | |
772 | uh->check = CSUM_MANGLED_0; | |
773 | } | |
774 | } | |
c26bf4a5 | 775 | EXPORT_SYMBOL_GPL(udp4_hwcsum); |
db8dac20 | 776 | |
af5fcba7 TH |
777 | /* Function to set UDP checksum for an IPv4 UDP packet. This is intended |
778 | * for the simple case like when setting the checksum for a UDP tunnel. | |
779 | */ | |
780 | void udp_set_csum(bool nocheck, struct sk_buff *skb, | |
781 | __be32 saddr, __be32 daddr, int len) | |
782 | { | |
783 | struct udphdr *uh = udp_hdr(skb); | |
784 | ||
179bc67f | 785 | if (nocheck) { |
af5fcba7 | 786 | uh->check = 0; |
179bc67f | 787 | } else if (skb_is_gso(skb)) { |
af5fcba7 | 788 | uh->check = ~udp_v4_check(len, saddr, daddr, 0); |
179bc67f EC |
789 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { |
790 | uh->check = 0; | |
791 | uh->check = udp_v4_check(len, saddr, daddr, lco_csum(skb)); | |
792 | if (uh->check == 0) | |
793 | uh->check = CSUM_MANGLED_0; | |
d75f1306 | 794 | } else { |
af5fcba7 TH |
795 | skb->ip_summed = CHECKSUM_PARTIAL; |
796 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
797 | skb->csum_offset = offsetof(struct udphdr, check); | |
798 | uh->check = ~udp_v4_check(len, saddr, daddr, 0); | |
af5fcba7 TH |
799 | } |
800 | } | |
801 | EXPORT_SYMBOL(udp_set_csum); | |
802 | ||
79ab0531 | 803 | static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4) |
db8dac20 | 804 | { |
f6b9664f | 805 | struct sock *sk = skb->sk; |
db8dac20 | 806 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
807 | struct udphdr *uh; |
808 | int err = 0; | |
809 | int is_udplite = IS_UDPLITE(sk); | |
f6b9664f HX |
810 | int offset = skb_transport_offset(skb); |
811 | int len = skb->len - offset; | |
db8dac20 DM |
812 | __wsum csum = 0; |
813 | ||
db8dac20 DM |
814 | /* |
815 | * Create a UDP header | |
816 | */ | |
817 | uh = udp_hdr(skb); | |
f6b9664f | 818 | uh->source = inet->inet_sport; |
79ab0531 | 819 | uh->dest = fl4->fl4_dport; |
f6b9664f | 820 | uh->len = htons(len); |
db8dac20 DM |
821 | uh->check = 0; |
822 | ||
823 | if (is_udplite) /* UDP-Lite */ | |
f6b9664f | 824 | csum = udplite_csum(skb); |
db8dac20 | 825 | |
28448b80 | 826 | else if (sk->sk_no_check_tx) { /* UDP csum disabled */ |
db8dac20 DM |
827 | |
828 | skb->ip_summed = CHECKSUM_NONE; | |
829 | goto send; | |
830 | ||
831 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
832 | ||
79ab0531 | 833 | udp4_hwcsum(skb, fl4->saddr, fl4->daddr); |
db8dac20 DM |
834 | goto send; |
835 | ||
f6b9664f HX |
836 | } else |
837 | csum = udp_csum(skb); | |
db8dac20 DM |
838 | |
839 | /* add protocol-dependent pseudo-header */ | |
79ab0531 | 840 | uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len, |
c482c568 | 841 | sk->sk_protocol, csum); |
db8dac20 DM |
842 | if (uh->check == 0) |
843 | uh->check = CSUM_MANGLED_0; | |
844 | ||
845 | send: | |
b5ec8eea | 846 | err = ip_send_skb(sock_net(sk), skb); |
6ce9e7b5 ED |
847 | if (err) { |
848 | if (err == -ENOBUFS && !inet->recverr) { | |
6aef70a8 ED |
849 | UDP_INC_STATS(sock_net(sk), |
850 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
6ce9e7b5 ED |
851 | err = 0; |
852 | } | |
853 | } else | |
6aef70a8 ED |
854 | UDP_INC_STATS(sock_net(sk), |
855 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
f6b9664f HX |
856 | return err; |
857 | } | |
858 | ||
859 | /* | |
860 | * Push out all pending data as one UDP datagram. Socket is locked. | |
861 | */ | |
8822b64a | 862 | int udp_push_pending_frames(struct sock *sk) |
f6b9664f HX |
863 | { |
864 | struct udp_sock *up = udp_sk(sk); | |
865 | struct inet_sock *inet = inet_sk(sk); | |
b6f21b26 | 866 | struct flowi4 *fl4 = &inet->cork.fl.u.ip4; |
f6b9664f HX |
867 | struct sk_buff *skb; |
868 | int err = 0; | |
869 | ||
77968b78 | 870 | skb = ip_finish_skb(sk, fl4); |
f6b9664f HX |
871 | if (!skb) |
872 | goto out; | |
873 | ||
79ab0531 | 874 | err = udp_send_skb(skb, fl4); |
f6b9664f | 875 | |
db8dac20 DM |
876 | out: |
877 | up->len = 0; | |
878 | up->pending = 0; | |
db8dac20 DM |
879 | return err; |
880 | } | |
8822b64a | 881 | EXPORT_SYMBOL(udp_push_pending_frames); |
db8dac20 | 882 | |
1b784140 | 883 | int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
db8dac20 DM |
884 | { |
885 | struct inet_sock *inet = inet_sk(sk); | |
886 | struct udp_sock *up = udp_sk(sk); | |
e474995f | 887 | struct flowi4 fl4_stack; |
b6f21b26 | 888 | struct flowi4 *fl4; |
db8dac20 DM |
889 | int ulen = len; |
890 | struct ipcm_cookie ipc; | |
891 | struct rtable *rt = NULL; | |
892 | int free = 0; | |
893 | int connected = 0; | |
894 | __be32 daddr, faddr, saddr; | |
895 | __be16 dport; | |
896 | u8 tos; | |
897 | int err, is_udplite = IS_UDPLITE(sk); | |
898 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
899 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
903ab86d | 900 | struct sk_buff *skb; |
f6d8bd05 | 901 | struct ip_options_data opt_copy; |
db8dac20 DM |
902 | |
903 | if (len > 0xFFFF) | |
904 | return -EMSGSIZE; | |
905 | ||
906 | /* | |
907 | * Check the flags. | |
908 | */ | |
909 | ||
c482c568 | 910 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */ |
db8dac20 DM |
911 | return -EOPNOTSUPP; |
912 | ||
913 | ipc.opt = NULL; | |
2244d07b | 914 | ipc.tx_flags = 0; |
aa661581 FF |
915 | ipc.ttl = 0; |
916 | ipc.tos = -1; | |
db8dac20 | 917 | |
903ab86d HX |
918 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
919 | ||
f5fca608 | 920 | fl4 = &inet->cork.fl.u.ip4; |
db8dac20 DM |
921 | if (up->pending) { |
922 | /* | |
923 | * There are pending frames. | |
924 | * The socket lock must be held while it's corked. | |
925 | */ | |
926 | lock_sock(sk); | |
927 | if (likely(up->pending)) { | |
928 | if (unlikely(up->pending != AF_INET)) { | |
929 | release_sock(sk); | |
930 | return -EINVAL; | |
931 | } | |
932 | goto do_append_data; | |
933 | } | |
934 | release_sock(sk); | |
935 | } | |
936 | ulen += sizeof(struct udphdr); | |
937 | ||
938 | /* | |
939 | * Get and verify the address. | |
940 | */ | |
941 | if (msg->msg_name) { | |
342dfc30 | 942 | DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); |
db8dac20 DM |
943 | if (msg->msg_namelen < sizeof(*usin)) |
944 | return -EINVAL; | |
945 | if (usin->sin_family != AF_INET) { | |
946 | if (usin->sin_family != AF_UNSPEC) | |
947 | return -EAFNOSUPPORT; | |
948 | } | |
949 | ||
950 | daddr = usin->sin_addr.s_addr; | |
951 | dport = usin->sin_port; | |
952 | if (dport == 0) | |
953 | return -EINVAL; | |
954 | } else { | |
955 | if (sk->sk_state != TCP_ESTABLISHED) | |
956 | return -EDESTADDRREQ; | |
c720c7e8 ED |
957 | daddr = inet->inet_daddr; |
958 | dport = inet->inet_dport; | |
db8dac20 DM |
959 | /* Open fast path for connected socket. |
960 | Route will not be used, if at least one option is set. | |
961 | */ | |
962 | connected = 1; | |
963 | } | |
db8dac20 | 964 | |
c14ac945 SHY |
965 | ipc.sockc.tsflags = sk->sk_tsflags; |
966 | ipc.addr = inet->inet_saddr; | |
db8dac20 | 967 | ipc.oif = sk->sk_bound_dev_if; |
bf84a010 | 968 | |
db8dac20 | 969 | if (msg->msg_controllen) { |
24025c46 | 970 | err = ip_cmsg_send(sk, msg, &ipc, sk->sk_family == AF_INET6); |
91948309 ED |
971 | if (unlikely(err)) { |
972 | kfree(ipc.opt); | |
db8dac20 | 973 | return err; |
91948309 | 974 | } |
db8dac20 DM |
975 | if (ipc.opt) |
976 | free = 1; | |
977 | connected = 0; | |
978 | } | |
f6d8bd05 ED |
979 | if (!ipc.opt) { |
980 | struct ip_options_rcu *inet_opt; | |
981 | ||
982 | rcu_read_lock(); | |
983 | inet_opt = rcu_dereference(inet->inet_opt); | |
984 | if (inet_opt) { | |
985 | memcpy(&opt_copy, inet_opt, | |
986 | sizeof(*inet_opt) + inet_opt->opt.optlen); | |
987 | ipc.opt = &opt_copy.opt; | |
988 | } | |
989 | rcu_read_unlock(); | |
990 | } | |
db8dac20 DM |
991 | |
992 | saddr = ipc.addr; | |
993 | ipc.addr = faddr = daddr; | |
994 | ||
c14ac945 SHY |
995 | sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags); |
996 | ||
f6d8bd05 | 997 | if (ipc.opt && ipc.opt->opt.srr) { |
db8dac20 DM |
998 | if (!daddr) |
999 | return -EINVAL; | |
f6d8bd05 | 1000 | faddr = ipc.opt->opt.faddr; |
db8dac20 DM |
1001 | connected = 0; |
1002 | } | |
aa661581 | 1003 | tos = get_rttos(&ipc, inet); |
db8dac20 DM |
1004 | if (sock_flag(sk, SOCK_LOCALROUTE) || |
1005 | (msg->msg_flags & MSG_DONTROUTE) || | |
f6d8bd05 | 1006 | (ipc.opt && ipc.opt->opt.is_strictroute)) { |
db8dac20 DM |
1007 | tos |= RTO_ONLINK; |
1008 | connected = 0; | |
1009 | } | |
1010 | ||
1011 | if (ipv4_is_multicast(daddr)) { | |
1012 | if (!ipc.oif) | |
1013 | ipc.oif = inet->mc_index; | |
1014 | if (!saddr) | |
1015 | saddr = inet->mc_addr; | |
1016 | connected = 0; | |
76e21053 EH |
1017 | } else if (!ipc.oif) |
1018 | ipc.oif = inet->uc_index; | |
db8dac20 DM |
1019 | |
1020 | if (connected) | |
c482c568 | 1021 | rt = (struct rtable *)sk_dst_check(sk, 0); |
db8dac20 | 1022 | |
51456b29 | 1023 | if (!rt) { |
84a3aa00 | 1024 | struct net *net = sock_net(sk); |
9a24abfa | 1025 | __u8 flow_flags = inet_sk_flowi_flags(sk); |
84a3aa00 | 1026 | |
e474995f | 1027 | fl4 = &fl4_stack; |
9a24abfa | 1028 | |
e474995f | 1029 | flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos, |
c0951cbc | 1030 | RT_SCOPE_UNIVERSE, sk->sk_protocol, |
9a24abfa | 1031 | flow_flags, |
e2d118a1 LC |
1032 | faddr, saddr, dport, inet->inet_sport, |
1033 | sk->sk_uid); | |
c0951cbc | 1034 | |
e474995f DM |
1035 | security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); |
1036 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe DM |
1037 | if (IS_ERR(rt)) { |
1038 | err = PTR_ERR(rt); | |
06dc94b1 | 1039 | rt = NULL; |
db8dac20 | 1040 | if (err == -ENETUNREACH) |
f1d8cba6 | 1041 | IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
1042 | goto out; |
1043 | } | |
1044 | ||
1045 | err = -EACCES; | |
1046 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
1047 | !sock_flag(sk, SOCK_BROADCAST)) | |
1048 | goto out; | |
1049 | if (connected) | |
d8d1f30b | 1050 | sk_dst_set(sk, dst_clone(&rt->dst)); |
db8dac20 DM |
1051 | } |
1052 | ||
1053 | if (msg->msg_flags&MSG_CONFIRM) | |
1054 | goto do_confirm; | |
1055 | back_from_confirm: | |
1056 | ||
e474995f | 1057 | saddr = fl4->saddr; |
db8dac20 | 1058 | if (!ipc.addr) |
e474995f | 1059 | daddr = ipc.addr = fl4->daddr; |
db8dac20 | 1060 | |
903ab86d HX |
1061 | /* Lockless fast path for the non-corking case. */ |
1062 | if (!corkreq) { | |
f69e6d13 | 1063 | skb = ip_make_skb(sk, fl4, getfrag, msg, ulen, |
903ab86d HX |
1064 | sizeof(struct udphdr), &ipc, &rt, |
1065 | msg->msg_flags); | |
1066 | err = PTR_ERR(skb); | |
50c3a487 | 1067 | if (!IS_ERR_OR_NULL(skb)) |
79ab0531 | 1068 | err = udp_send_skb(skb, fl4); |
903ab86d HX |
1069 | goto out; |
1070 | } | |
1071 | ||
db8dac20 DM |
1072 | lock_sock(sk); |
1073 | if (unlikely(up->pending)) { | |
1074 | /* The socket is already corked while preparing it. */ | |
1075 | /* ... which is an evident application bug. --ANK */ | |
1076 | release_sock(sk); | |
1077 | ||
ba7a46f1 | 1078 | net_dbg_ratelimited("cork app bug 2\n"); |
db8dac20 DM |
1079 | err = -EINVAL; |
1080 | goto out; | |
1081 | } | |
1082 | /* | |
1083 | * Now cork the socket to pend data. | |
1084 | */ | |
b6f21b26 DM |
1085 | fl4 = &inet->cork.fl.u.ip4; |
1086 | fl4->daddr = daddr; | |
1087 | fl4->saddr = saddr; | |
9cce96df DM |
1088 | fl4->fl4_dport = dport; |
1089 | fl4->fl4_sport = inet->inet_sport; | |
db8dac20 DM |
1090 | up->pending = AF_INET; |
1091 | ||
1092 | do_append_data: | |
1093 | up->len += ulen; | |
f69e6d13 | 1094 | err = ip_append_data(sk, fl4, getfrag, msg, ulen, |
f5fca608 DM |
1095 | sizeof(struct udphdr), &ipc, &rt, |
1096 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | |
db8dac20 DM |
1097 | if (err) |
1098 | udp_flush_pending_frames(sk); | |
1099 | else if (!corkreq) | |
1100 | err = udp_push_pending_frames(sk); | |
1101 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
1102 | up->pending = 0; | |
1103 | release_sock(sk); | |
1104 | ||
1105 | out: | |
1106 | ip_rt_put(rt); | |
1107 | if (free) | |
1108 | kfree(ipc.opt); | |
1109 | if (!err) | |
1110 | return len; | |
1111 | /* | |
1112 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1113 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1114 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1115 | * things). We could add another new stat but at least for now that | |
1116 | * seems like overkill. | |
1117 | */ | |
1118 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
6aef70a8 ED |
1119 | UDP_INC_STATS(sock_net(sk), |
1120 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
1121 | } |
1122 | return err; | |
1123 | ||
1124 | do_confirm: | |
d8d1f30b | 1125 | dst_confirm(&rt->dst); |
db8dac20 DM |
1126 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1127 | goto back_from_confirm; | |
1128 | err = 0; | |
1129 | goto out; | |
1130 | } | |
c482c568 | 1131 | EXPORT_SYMBOL(udp_sendmsg); |
db8dac20 DM |
1132 | |
1133 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
1134 | size_t size, int flags) | |
1135 | { | |
f5fca608 | 1136 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
1137 | struct udp_sock *up = udp_sk(sk); |
1138 | int ret; | |
1139 | ||
d3f7d56a SL |
1140 | if (flags & MSG_SENDPAGE_NOTLAST) |
1141 | flags |= MSG_MORE; | |
1142 | ||
db8dac20 DM |
1143 | if (!up->pending) { |
1144 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
1145 | ||
1146 | /* Call udp_sendmsg to specify destination address which | |
1147 | * sendpage interface can't pass. | |
1148 | * This will succeed only when the socket is connected. | |
1149 | */ | |
1b784140 | 1150 | ret = udp_sendmsg(sk, &msg, 0); |
db8dac20 DM |
1151 | if (ret < 0) |
1152 | return ret; | |
1153 | } | |
1154 | ||
1155 | lock_sock(sk); | |
1156 | ||
1157 | if (unlikely(!up->pending)) { | |
1158 | release_sock(sk); | |
1159 | ||
ba7a46f1 | 1160 | net_dbg_ratelimited("udp cork app bug 3\n"); |
db8dac20 DM |
1161 | return -EINVAL; |
1162 | } | |
1163 | ||
f5fca608 DM |
1164 | ret = ip_append_page(sk, &inet->cork.fl.u.ip4, |
1165 | page, offset, size, flags); | |
db8dac20 DM |
1166 | if (ret == -EOPNOTSUPP) { |
1167 | release_sock(sk); | |
1168 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
1169 | size, flags); | |
1170 | } | |
1171 | if (ret < 0) { | |
1172 | udp_flush_pending_frames(sk); | |
1173 | goto out; | |
1174 | } | |
1175 | ||
1176 | up->len += size; | |
1177 | if (!(up->corkflag || (flags&MSG_MORE))) | |
1178 | ret = udp_push_pending_frames(sk); | |
1179 | if (!ret) | |
1180 | ret = size; | |
1181 | out: | |
1182 | release_sock(sk); | |
1183 | return ret; | |
1184 | } | |
1185 | ||
7c13f97f | 1186 | /* fully reclaim rmem/fwd memory allocated for skb */ |
f970bd9e PA |
1187 | static void udp_rmem_release(struct sock *sk, int size, int partial) |
1188 | { | |
6b229cf7 | 1189 | struct udp_sock *up = udp_sk(sk); |
f970bd9e PA |
1190 | int amt; |
1191 | ||
6b229cf7 ED |
1192 | if (likely(partial)) { |
1193 | up->forward_deficit += size; | |
1194 | size = up->forward_deficit; | |
1195 | if (size < (sk->sk_rcvbuf >> 2) && | |
1196 | !skb_queue_empty(&sk->sk_receive_queue)) | |
1197 | return; | |
1198 | } else { | |
1199 | size += up->forward_deficit; | |
1200 | } | |
1201 | up->forward_deficit = 0; | |
1202 | ||
f970bd9e PA |
1203 | sk->sk_forward_alloc += size; |
1204 | amt = (sk->sk_forward_alloc - partial) & ~(SK_MEM_QUANTUM - 1); | |
1205 | sk->sk_forward_alloc -= amt; | |
f970bd9e PA |
1206 | |
1207 | if (amt) | |
1208 | __sk_mem_reduce_allocated(sk, amt >> SK_MEM_QUANTUM_SHIFT); | |
02ab0d13 ED |
1209 | |
1210 | atomic_sub(size, &sk->sk_rmem_alloc); | |
f970bd9e PA |
1211 | } |
1212 | ||
c84d9490 ED |
1213 | /* Note: called with sk_receive_queue.lock held. |
1214 | * Instead of using skb->truesize here, find a copy of it in skb->dev_scratch | |
1215 | * This avoids a cache line miss while receive_queue lock is held. | |
1216 | * Look at __udp_enqueue_schedule_skb() to find where this copy is done. | |
1217 | */ | |
7c13f97f | 1218 | void udp_skb_destructor(struct sock *sk, struct sk_buff *skb) |
f970bd9e | 1219 | { |
c84d9490 | 1220 | udp_rmem_release(sk, skb->dev_scratch, 1); |
f970bd9e | 1221 | } |
7c13f97f | 1222 | EXPORT_SYMBOL(udp_skb_destructor); |
f970bd9e | 1223 | |
4b272750 ED |
1224 | /* Idea of busylocks is to let producers grab an extra spinlock |
1225 | * to relieve pressure on the receive_queue spinlock shared by consumer. | |
1226 | * Under flood, this means that only one producer can be in line | |
1227 | * trying to acquire the receive_queue spinlock. | |
1228 | * These busylock can be allocated on a per cpu manner, instead of a | |
1229 | * per socket one (that would consume a cache line per socket) | |
1230 | */ | |
1231 | static int udp_busylocks_log __read_mostly; | |
1232 | static spinlock_t *udp_busylocks __read_mostly; | |
1233 | ||
1234 | static spinlock_t *busylock_acquire(void *ptr) | |
1235 | { | |
1236 | spinlock_t *busy; | |
1237 | ||
1238 | busy = udp_busylocks + hash_ptr(ptr, udp_busylocks_log); | |
1239 | spin_lock(busy); | |
1240 | return busy; | |
1241 | } | |
1242 | ||
1243 | static void busylock_release(spinlock_t *busy) | |
1244 | { | |
1245 | if (busy) | |
1246 | spin_unlock(busy); | |
1247 | } | |
1248 | ||
f970bd9e PA |
1249 | int __udp_enqueue_schedule_skb(struct sock *sk, struct sk_buff *skb) |
1250 | { | |
1251 | struct sk_buff_head *list = &sk->sk_receive_queue; | |
1252 | int rmem, delta, amt, err = -ENOMEM; | |
4b272750 | 1253 | spinlock_t *busy = NULL; |
c8c8b127 | 1254 | int size; |
f970bd9e PA |
1255 | |
1256 | /* try to avoid the costly atomic add/sub pair when the receive | |
1257 | * queue is full; always allow at least a packet | |
1258 | */ | |
1259 | rmem = atomic_read(&sk->sk_rmem_alloc); | |
363dc73a | 1260 | if (rmem > sk->sk_rcvbuf) |
f970bd9e PA |
1261 | goto drop; |
1262 | ||
c8c8b127 ED |
1263 | /* Under mem pressure, it might be helpful to help udp_recvmsg() |
1264 | * having linear skbs : | |
1265 | * - Reduce memory overhead and thus increase receive queue capacity | |
1266 | * - Less cache line misses at copyout() time | |
1267 | * - Less work at consume_skb() (less alien page frag freeing) | |
1268 | */ | |
4b272750 | 1269 | if (rmem > (sk->sk_rcvbuf >> 1)) { |
c8c8b127 | 1270 | skb_condense(skb); |
4b272750 ED |
1271 | |
1272 | busy = busylock_acquire(sk); | |
1273 | } | |
c8c8b127 | 1274 | size = skb->truesize; |
c84d9490 ED |
1275 | /* Copy skb->truesize into skb->dev_scratch to avoid a cache line miss |
1276 | * in udp_skb_destructor() | |
1277 | */ | |
1278 | skb->dev_scratch = size; | |
c8c8b127 | 1279 | |
f970bd9e PA |
1280 | /* we drop only if the receive buf is full and the receive |
1281 | * queue contains some other skb | |
1282 | */ | |
1283 | rmem = atomic_add_return(size, &sk->sk_rmem_alloc); | |
363dc73a | 1284 | if (rmem > (size + sk->sk_rcvbuf)) |
f970bd9e PA |
1285 | goto uncharge_drop; |
1286 | ||
1287 | spin_lock(&list->lock); | |
1288 | if (size >= sk->sk_forward_alloc) { | |
1289 | amt = sk_mem_pages(size); | |
1290 | delta = amt << SK_MEM_QUANTUM_SHIFT; | |
1291 | if (!__sk_mem_raise_allocated(sk, delta, amt, SK_MEM_RECV)) { | |
1292 | err = -ENOBUFS; | |
1293 | spin_unlock(&list->lock); | |
1294 | goto uncharge_drop; | |
1295 | } | |
1296 | ||
1297 | sk->sk_forward_alloc += delta; | |
1298 | } | |
1299 | ||
1300 | sk->sk_forward_alloc -= size; | |
1301 | ||
7c13f97f PA |
1302 | /* no need to setup a destructor, we will explicitly release the |
1303 | * forward allocated memory on dequeue | |
1304 | */ | |
f970bd9e PA |
1305 | sock_skb_set_dropcount(sk, skb); |
1306 | ||
1307 | __skb_queue_tail(list, skb); | |
1308 | spin_unlock(&list->lock); | |
1309 | ||
1310 | if (!sock_flag(sk, SOCK_DEAD)) | |
1311 | sk->sk_data_ready(sk); | |
1312 | ||
4b272750 | 1313 | busylock_release(busy); |
f970bd9e PA |
1314 | return 0; |
1315 | ||
1316 | uncharge_drop: | |
1317 | atomic_sub(skb->truesize, &sk->sk_rmem_alloc); | |
1318 | ||
1319 | drop: | |
1320 | atomic_inc(&sk->sk_drops); | |
4b272750 | 1321 | busylock_release(busy); |
f970bd9e PA |
1322 | return err; |
1323 | } | |
1324 | EXPORT_SYMBOL_GPL(__udp_enqueue_schedule_skb); | |
1325 | ||
c915fe13 | 1326 | void udp_destruct_sock(struct sock *sk) |
f970bd9e PA |
1327 | { |
1328 | /* reclaim completely the forward allocated memory */ | |
7c13f97f PA |
1329 | unsigned int total = 0; |
1330 | struct sk_buff *skb; | |
1331 | ||
1332 | while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) { | |
1333 | total += skb->truesize; | |
1334 | kfree_skb(skb); | |
1335 | } | |
1336 | udp_rmem_release(sk, total, 0); | |
1337 | ||
f970bd9e PA |
1338 | inet_sock_destruct(sk); |
1339 | } | |
c915fe13 | 1340 | EXPORT_SYMBOL_GPL(udp_destruct_sock); |
f970bd9e PA |
1341 | |
1342 | int udp_init_sock(struct sock *sk) | |
1343 | { | |
1344 | sk->sk_destruct = udp_destruct_sock; | |
1345 | return 0; | |
1346 | } | |
1347 | EXPORT_SYMBOL_GPL(udp_init_sock); | |
1348 | ||
1349 | void skb_consume_udp(struct sock *sk, struct sk_buff *skb, int len) | |
1350 | { | |
1351 | if (unlikely(READ_ONCE(sk->sk_peek_off) >= 0)) { | |
1352 | bool slow = lock_sock_fast(sk); | |
1353 | ||
1354 | sk_peek_offset_bwd(sk, len); | |
1355 | unlock_sock_fast(sk, slow); | |
1356 | } | |
1357 | consume_skb(skb); | |
1358 | } | |
1359 | EXPORT_SYMBOL_GPL(skb_consume_udp); | |
1360 | ||
85584672 ED |
1361 | /** |
1362 | * first_packet_length - return length of first packet in receive queue | |
1363 | * @sk: socket | |
1364 | * | |
1365 | * Drops all bad checksum frames, until a valid one is found. | |
e83c6744 | 1366 | * Returns the length of found skb, or -1 if none is found. |
85584672 | 1367 | */ |
e83c6744 | 1368 | static int first_packet_length(struct sock *sk) |
85584672 | 1369 | { |
7c13f97f | 1370 | struct sk_buff_head *rcvq = &sk->sk_receive_queue; |
85584672 | 1371 | struct sk_buff *skb; |
7c13f97f | 1372 | int total = 0; |
e83c6744 | 1373 | int res; |
85584672 | 1374 | |
85584672 ED |
1375 | spin_lock_bh(&rcvq->lock); |
1376 | while ((skb = skb_peek(rcvq)) != NULL && | |
1377 | udp_lib_checksum_complete(skb)) { | |
02c22347 ED |
1378 | __UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, |
1379 | IS_UDPLITE(sk)); | |
1380 | __UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, | |
1381 | IS_UDPLITE(sk)); | |
8edf19c2 | 1382 | atomic_inc(&sk->sk_drops); |
85584672 | 1383 | __skb_unlink(skb, rcvq); |
7c13f97f PA |
1384 | total += skb->truesize; |
1385 | kfree_skb(skb); | |
85584672 | 1386 | } |
e83c6744 | 1387 | res = skb ? skb->len : -1; |
7c13f97f PA |
1388 | if (total) |
1389 | udp_rmem_release(sk, total, 1); | |
85584672 | 1390 | spin_unlock_bh(&rcvq->lock); |
85584672 ED |
1391 | return res; |
1392 | } | |
1393 | ||
1da177e4 LT |
1394 | /* |
1395 | * IOCTL requests applicable to the UDP protocol | |
1396 | */ | |
e905a9ed | 1397 | |
1da177e4 LT |
1398 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
1399 | { | |
6516c655 SH |
1400 | switch (cmd) { |
1401 | case SIOCOUTQ: | |
1da177e4 | 1402 | { |
31e6d363 ED |
1403 | int amount = sk_wmem_alloc_get(sk); |
1404 | ||
6516c655 SH |
1405 | return put_user(amount, (int __user *)arg); |
1406 | } | |
1da177e4 | 1407 | |
6516c655 SH |
1408 | case SIOCINQ: |
1409 | { | |
e83c6744 | 1410 | int amount = max_t(int, 0, first_packet_length(sk)); |
6516c655 | 1411 | |
6516c655 SH |
1412 | return put_user(amount, (int __user *)arg); |
1413 | } | |
1da177e4 | 1414 | |
6516c655 SH |
1415 | default: |
1416 | return -ENOIOCTLCMD; | |
1da177e4 | 1417 | } |
6516c655 SH |
1418 | |
1419 | return 0; | |
1da177e4 | 1420 | } |
c482c568 | 1421 | EXPORT_SYMBOL(udp_ioctl); |
1da177e4 | 1422 | |
db8dac20 DM |
1423 | /* |
1424 | * This should be easy, if there is something there we | |
1425 | * return it, otherwise we block. | |
1426 | */ | |
1427 | ||
1b784140 YX |
1428 | int udp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock, |
1429 | int flags, int *addr_len) | |
db8dac20 DM |
1430 | { |
1431 | struct inet_sock *inet = inet_sk(sk); | |
342dfc30 | 1432 | DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); |
db8dac20 | 1433 | struct sk_buff *skb; |
59c2cdae | 1434 | unsigned int ulen, copied; |
627d2d6b | 1435 | int peeked, peeking, off; |
db8dac20 DM |
1436 | int err; |
1437 | int is_udplite = IS_UDPLITE(sk); | |
197c949e | 1438 | bool checksum_valid = false; |
db8dac20 | 1439 | |
db8dac20 | 1440 | if (flags & MSG_ERRQUEUE) |
85fbaa75 | 1441 | return ip_recv_error(sk, msg, len, addr_len); |
db8dac20 DM |
1442 | |
1443 | try_again: | |
627d2d6b | 1444 | peeking = off = sk_peek_offset(sk, flags); |
7c13f97f | 1445 | skb = __skb_recv_udp(sk, flags, noblock, &peeked, &off, &err); |
db8dac20 | 1446 | if (!skb) |
627d2d6b | 1447 | return err; |
db8dac20 | 1448 | |
e6afc8ac | 1449 | ulen = skb->len; |
59c2cdae | 1450 | copied = len; |
627d2d6b | 1451 | if (copied > ulen - off) |
1452 | copied = ulen - off; | |
59c2cdae | 1453 | else if (copied < ulen) |
db8dac20 DM |
1454 | msg->msg_flags |= MSG_TRUNC; |
1455 | ||
1456 | /* | |
1457 | * If checksum is needed at all, try to do it while copying the | |
1458 | * data. If the data is truncated, or if we only want a partial | |
1459 | * coverage checksum (UDP-Lite), do it before the copy. | |
1460 | */ | |
1461 | ||
d21dbdfe ED |
1462 | if (copied < ulen || peeking || |
1463 | (is_udplite && UDP_SKB_CB(skb)->partial_cov)) { | |
197c949e ED |
1464 | checksum_valid = !udp_lib_checksum_complete(skb); |
1465 | if (!checksum_valid) | |
db8dac20 DM |
1466 | goto csum_copy_err; |
1467 | } | |
1468 | ||
197c949e | 1469 | if (checksum_valid || skb_csum_unnecessary(skb)) |
627d2d6b | 1470 | err = skb_copy_datagram_msg(skb, off, msg, copied); |
db8dac20 | 1471 | else { |
627d2d6b | 1472 | err = skb_copy_and_csum_datagram_msg(skb, off, msg); |
db8dac20 DM |
1473 | |
1474 | if (err == -EINVAL) | |
1475 | goto csum_copy_err; | |
1476 | } | |
1477 | ||
22911fc5 | 1478 | if (unlikely(err)) { |
979402b1 ED |
1479 | if (!peeked) { |
1480 | atomic_inc(&sk->sk_drops); | |
6aef70a8 ED |
1481 | UDP_INC_STATS(sock_net(sk), |
1482 | UDP_MIB_INERRORS, is_udplite); | |
979402b1 | 1483 | } |
850cbadd | 1484 | kfree_skb(skb); |
627d2d6b | 1485 | return err; |
22911fc5 | 1486 | } |
db8dac20 DM |
1487 | |
1488 | if (!peeked) | |
6aef70a8 ED |
1489 | UDP_INC_STATS(sock_net(sk), |
1490 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 | 1491 | |
3b885787 | 1492 | sock_recv_ts_and_drops(msg, sk, skb); |
db8dac20 DM |
1493 | |
1494 | /* Copy the address. */ | |
c482c568 | 1495 | if (sin) { |
db8dac20 DM |
1496 | sin->sin_family = AF_INET; |
1497 | sin->sin_port = udp_hdr(skb)->source; | |
1498 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
1499 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
bceaa902 | 1500 | *addr_len = sizeof(*sin); |
db8dac20 DM |
1501 | } |
1502 | if (inet->cmsg_flags) | |
ad959036 | 1503 | ip_cmsg_recv_offset(msg, sk, skb, sizeof(struct udphdr), off); |
db8dac20 | 1504 | |
59c2cdae | 1505 | err = copied; |
db8dac20 DM |
1506 | if (flags & MSG_TRUNC) |
1507 | err = ulen; | |
1508 | ||
850cbadd | 1509 | skb_consume_udp(sk, skb, peeking ? -err : err); |
db8dac20 DM |
1510 | return err; |
1511 | ||
1512 | csum_copy_err: | |
850cbadd | 1513 | if (!__sk_queue_drop_skb(sk, skb, flags)) { |
6aef70a8 ED |
1514 | UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); |
1515 | UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); | |
6a5dc9e5 | 1516 | } |
850cbadd | 1517 | kfree_skb(skb); |
db8dac20 | 1518 | |
beb39db5 ED |
1519 | /* starting over for a new packet, but check if we need to yield */ |
1520 | cond_resched(); | |
9cfaa8de | 1521 | msg->msg_flags &= ~MSG_TRUNC; |
db8dac20 DM |
1522 | goto try_again; |
1523 | } | |
1524 | ||
286c72de | 1525 | int __udp_disconnect(struct sock *sk, int flags) |
1da177e4 LT |
1526 | { |
1527 | struct inet_sock *inet = inet_sk(sk); | |
1528 | /* | |
1529 | * 1003.1g - break association. | |
1530 | */ | |
e905a9ed | 1531 | |
1da177e4 | 1532 | sk->sk_state = TCP_CLOSE; |
c720c7e8 ED |
1533 | inet->inet_daddr = 0; |
1534 | inet->inet_dport = 0; | |
bdeab991 | 1535 | sock_rps_reset_rxhash(sk); |
1da177e4 LT |
1536 | sk->sk_bound_dev_if = 0; |
1537 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
1538 | inet_reset_saddr(sk); | |
1539 | ||
1540 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
1541 | sk->sk_prot->unhash(sk); | |
c720c7e8 | 1542 | inet->inet_sport = 0; |
1da177e4 LT |
1543 | } |
1544 | sk_dst_reset(sk); | |
1545 | return 0; | |
1546 | } | |
286c72de ED |
1547 | EXPORT_SYMBOL(__udp_disconnect); |
1548 | ||
1549 | int udp_disconnect(struct sock *sk, int flags) | |
1550 | { | |
1551 | lock_sock(sk); | |
1552 | __udp_disconnect(sk, flags); | |
1553 | release_sock(sk); | |
1554 | return 0; | |
1555 | } | |
c482c568 | 1556 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1557 | |
645ca708 ED |
1558 | void udp_lib_unhash(struct sock *sk) |
1559 | { | |
723b4610 ED |
1560 | if (sk_hashed(sk)) { |
1561 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
512615b6 ED |
1562 | struct udp_hslot *hslot, *hslot2; |
1563 | ||
1564 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1565 | udp_sk(sk)->udp_port_hash); | |
1566 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
645ca708 | 1567 | |
723b4610 | 1568 | spin_lock_bh(&hslot->lock); |
e32ea7e7 CG |
1569 | if (rcu_access_pointer(sk->sk_reuseport_cb)) |
1570 | reuseport_detach_sock(sk); | |
ca065d0c | 1571 | if (sk_del_node_init_rcu(sk)) { |
fdcc8aa9 | 1572 | hslot->count--; |
c720c7e8 | 1573 | inet_sk(sk)->inet_num = 0; |
723b4610 | 1574 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
512615b6 ED |
1575 | |
1576 | spin_lock(&hslot2->lock); | |
ca065d0c | 1577 | hlist_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); |
512615b6 ED |
1578 | hslot2->count--; |
1579 | spin_unlock(&hslot2->lock); | |
723b4610 ED |
1580 | } |
1581 | spin_unlock_bh(&hslot->lock); | |
645ca708 | 1582 | } |
645ca708 ED |
1583 | } |
1584 | EXPORT_SYMBOL(udp_lib_unhash); | |
1585 | ||
719f8358 ED |
1586 | /* |
1587 | * inet_rcv_saddr was changed, we must rehash secondary hash | |
1588 | */ | |
1589 | void udp_lib_rehash(struct sock *sk, u16 newhash) | |
1590 | { | |
1591 | if (sk_hashed(sk)) { | |
1592 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
1593 | struct udp_hslot *hslot, *hslot2, *nhslot2; | |
1594 | ||
1595 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
1596 | nhslot2 = udp_hashslot2(udptable, newhash); | |
1597 | udp_sk(sk)->udp_portaddr_hash = newhash; | |
e32ea7e7 CG |
1598 | |
1599 | if (hslot2 != nhslot2 || | |
1600 | rcu_access_pointer(sk->sk_reuseport_cb)) { | |
719f8358 ED |
1601 | hslot = udp_hashslot(udptable, sock_net(sk), |
1602 | udp_sk(sk)->udp_port_hash); | |
1603 | /* we must lock primary chain too */ | |
1604 | spin_lock_bh(&hslot->lock); | |
e32ea7e7 CG |
1605 | if (rcu_access_pointer(sk->sk_reuseport_cb)) |
1606 | reuseport_detach_sock(sk); | |
1607 | ||
1608 | if (hslot2 != nhslot2) { | |
1609 | spin_lock(&hslot2->lock); | |
ca065d0c | 1610 | hlist_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); |
e32ea7e7 CG |
1611 | hslot2->count--; |
1612 | spin_unlock(&hslot2->lock); | |
1613 | ||
1614 | spin_lock(&nhslot2->lock); | |
ca065d0c | 1615 | hlist_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, |
e32ea7e7 CG |
1616 | &nhslot2->head); |
1617 | nhslot2->count++; | |
1618 | spin_unlock(&nhslot2->lock); | |
1619 | } | |
719f8358 ED |
1620 | |
1621 | spin_unlock_bh(&hslot->lock); | |
1622 | } | |
1623 | } | |
1624 | } | |
1625 | EXPORT_SYMBOL(udp_lib_rehash); | |
1626 | ||
1627 | static void udp_v4_rehash(struct sock *sk) | |
1628 | { | |
1629 | u16 new_hash = udp4_portaddr_hash(sock_net(sk), | |
1630 | inet_sk(sk)->inet_rcv_saddr, | |
1631 | inet_sk(sk)->inet_num); | |
1632 | udp_lib_rehash(sk, new_hash); | |
1633 | } | |
1634 | ||
30c7be26 | 1635 | int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
93821778 | 1636 | { |
fec5e652 | 1637 | int rc; |
766e9037 | 1638 | |
005ec974 | 1639 | if (inet_sk(sk)->inet_daddr) { |
bdeab991 | 1640 | sock_rps_save_rxhash(sk, skb); |
005ec974 | 1641 | sk_mark_napi_id(sk, skb); |
2c8c56e1 | 1642 | sk_incoming_cpu_update(sk); |
e68b6e50 ED |
1643 | } else { |
1644 | sk_mark_napi_id_once(sk, skb); | |
005ec974 | 1645 | } |
fec5e652 | 1646 | |
850cbadd | 1647 | rc = __udp_enqueue_schedule_skb(sk, skb); |
766e9037 ED |
1648 | if (rc < 0) { |
1649 | int is_udplite = IS_UDPLITE(sk); | |
93821778 | 1650 | |
93821778 | 1651 | /* Note that an ENOMEM error is charged twice */ |
766e9037 | 1652 | if (rc == -ENOMEM) |
e61da9e2 | 1653 | UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
02c22347 | 1654 | is_udplite); |
e61da9e2 | 1655 | UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
766e9037 | 1656 | kfree_skb(skb); |
296f7ea7 | 1657 | trace_udp_fail_queue_rcv_skb(rc, sk); |
766e9037 | 1658 | return -1; |
93821778 HX |
1659 | } |
1660 | ||
1661 | return 0; | |
93821778 HX |
1662 | } |
1663 | ||
447167bf ED |
1664 | static struct static_key udp_encap_needed __read_mostly; |
1665 | void udp_encap_enable(void) | |
1666 | { | |
1667 | if (!static_key_enabled(&udp_encap_needed)) | |
1668 | static_key_slow_inc(&udp_encap_needed); | |
1669 | } | |
1670 | EXPORT_SYMBOL(udp_encap_enable); | |
1671 | ||
db8dac20 DM |
1672 | /* returns: |
1673 | * -1: error | |
1674 | * 0: success | |
1675 | * >0: "udp encap" protocol resubmission | |
1676 | * | |
1677 | * Note that in the success and error cases, the skb is assumed to | |
1678 | * have either been requeued or freed. | |
1679 | */ | |
c482c568 | 1680 | int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
db8dac20 DM |
1681 | { |
1682 | struct udp_sock *up = udp_sk(sk); | |
db8dac20 DM |
1683 | int is_udplite = IS_UDPLITE(sk); |
1684 | ||
1685 | /* | |
1686 | * Charge it to the socket, dropping if the queue is full. | |
1687 | */ | |
1688 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1689 | goto drop; | |
1690 | nf_reset(skb); | |
1691 | ||
447167bf | 1692 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
0ad92ad0 ED |
1693 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
1694 | ||
db8dac20 DM |
1695 | /* |
1696 | * This is an encapsulation socket so pass the skb to | |
1697 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1698 | * fall through and pass this up the UDP socket. | |
1699 | * up->encap_rcv() returns the following value: | |
1700 | * =0 if skb was successfully passed to the encap | |
1701 | * handler or was discarded by it. | |
1702 | * >0 if skb should be passed on to UDP. | |
1703 | * <0 if skb should be resubmitted as proto -N | |
1704 | */ | |
1705 | ||
1706 | /* if we're overly short, let UDP handle it */ | |
0ad92ad0 | 1707 | encap_rcv = ACCESS_ONCE(up->encap_rcv); |
e5aed006 | 1708 | if (encap_rcv) { |
db8dac20 DM |
1709 | int ret; |
1710 | ||
0a80966b TH |
1711 | /* Verify checksum before giving to encap */ |
1712 | if (udp_lib_checksum_complete(skb)) | |
1713 | goto csum_error; | |
1714 | ||
0ad92ad0 | 1715 | ret = encap_rcv(sk, skb); |
db8dac20 | 1716 | if (ret <= 0) { |
02c22347 ED |
1717 | __UDP_INC_STATS(sock_net(sk), |
1718 | UDP_MIB_INDATAGRAMS, | |
1719 | is_udplite); | |
db8dac20 DM |
1720 | return -ret; |
1721 | } | |
1722 | } | |
1723 | ||
1724 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1725 | } | |
1726 | ||
1727 | /* | |
1728 | * UDP-Lite specific tests, ignored on UDP sockets | |
1729 | */ | |
1730 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1731 | ||
1732 | /* | |
1733 | * MIB statistics other than incrementing the error count are | |
1734 | * disabled for the following two types of errors: these depend | |
1735 | * on the application settings, not on the functioning of the | |
1736 | * protocol stack as such. | |
1737 | * | |
1738 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1739 | * way ... to ... at least let the receiving application block | |
1740 | * delivery of packets with coverage values less than a value | |
1741 | * provided by the application." | |
1742 | */ | |
1743 | if (up->pcrlen == 0) { /* full coverage was set */ | |
ba7a46f1 JP |
1744 | net_dbg_ratelimited("UDPLite: partial coverage %d while full coverage %d requested\n", |
1745 | UDP_SKB_CB(skb)->cscov, skb->len); | |
db8dac20 DM |
1746 | goto drop; |
1747 | } | |
1748 | /* The next case involves violating the min. coverage requested | |
1749 | * by the receiver. This is subtle: if receiver wants x and x is | |
1750 | * greater than the buffersize/MTU then receiver will complain | |
1751 | * that it wants x while sender emits packets of smaller size y. | |
1752 | * Therefore the above ...()->partial_cov statement is essential. | |
1753 | */ | |
1754 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
ba7a46f1 JP |
1755 | net_dbg_ratelimited("UDPLite: coverage %d too small, need min %d\n", |
1756 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
db8dac20 DM |
1757 | goto drop; |
1758 | } | |
1759 | } | |
1760 | ||
ce25d66a ED |
1761 | if (rcu_access_pointer(sk->sk_filter) && |
1762 | udp_lib_checksum_complete(skb)) | |
e6afc8ac | 1763 | goto csum_error; |
ce25d66a | 1764 | |
ba66bbe5 | 1765 | if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr))) |
a6127697 | 1766 | goto drop; |
db8dac20 | 1767 | |
e6afc8ac | 1768 | udp_csum_pull_header(skb); |
db8dac20 | 1769 | |
fbf8866d | 1770 | ipv4_pktinfo_prepare(sk, skb); |
850cbadd | 1771 | return __udp_queue_rcv_skb(sk, skb); |
db8dac20 | 1772 | |
6a5dc9e5 | 1773 | csum_error: |
02c22347 | 1774 | __UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); |
db8dac20 | 1775 | drop: |
02c22347 | 1776 | __UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8edf19c2 | 1777 | atomic_inc(&sk->sk_drops); |
db8dac20 DM |
1778 | kfree_skb(skb); |
1779 | return -1; | |
1780 | } | |
1781 | ||
97502231 | 1782 | /* For TCP sockets, sk_rx_dst is protected by socket lock |
e47eb5df | 1783 | * For UDP, we use xchg() to guard against concurrent changes. |
97502231 ED |
1784 | */ |
1785 | static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst) | |
421b3885 | 1786 | { |
97502231 ED |
1787 | struct dst_entry *old; |
1788 | ||
e47eb5df ED |
1789 | dst_hold(dst); |
1790 | old = xchg(&sk->sk_rx_dst, dst); | |
1791 | dst_release(old); | |
421b3885 SB |
1792 | } |
1793 | ||
db8dac20 DM |
1794 | /* |
1795 | * Multicasts and broadcasts go to each listener. | |
1796 | * | |
1240d137 | 1797 | * Note: called only from the BH handler context. |
db8dac20 | 1798 | */ |
e3163493 | 1799 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1800 | struct udphdr *uh, |
1801 | __be32 saddr, __be32 daddr, | |
36cbb245 RJ |
1802 | struct udp_table *udptable, |
1803 | int proto) | |
db8dac20 | 1804 | { |
ca065d0c | 1805 | struct sock *sk, *first = NULL; |
5cf3d461 DH |
1806 | unsigned short hnum = ntohs(uh->dest); |
1807 | struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum); | |
2dc41cff | 1808 | unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10); |
ca065d0c ED |
1809 | unsigned int offset = offsetof(typeof(*sk), sk_node); |
1810 | int dif = skb->dev->ifindex; | |
1811 | struct hlist_node *node; | |
1812 | struct sk_buff *nskb; | |
2dc41cff DH |
1813 | |
1814 | if (use_hash2) { | |
1815 | hash2_any = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum) & | |
73e2d5e3 PN |
1816 | udptable->mask; |
1817 | hash2 = udp4_portaddr_hash(net, daddr, hnum) & udptable->mask; | |
2dc41cff | 1818 | start_lookup: |
73e2d5e3 | 1819 | hslot = &udptable->hash2[hash2]; |
2dc41cff DH |
1820 | offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node); |
1821 | } | |
db8dac20 | 1822 | |
ca065d0c ED |
1823 | sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) { |
1824 | if (!__udp_is_mcast_sock(net, sk, uh->dest, daddr, | |
1825 | uh->source, saddr, dif, hnum)) | |
1826 | continue; | |
1827 | ||
1828 | if (!first) { | |
1829 | first = sk; | |
1830 | continue; | |
1240d137 | 1831 | } |
ca065d0c | 1832 | nskb = skb_clone(skb, GFP_ATOMIC); |
1240d137 | 1833 | |
ca065d0c ED |
1834 | if (unlikely(!nskb)) { |
1835 | atomic_inc(&sk->sk_drops); | |
02c22347 ED |
1836 | __UDP_INC_STATS(net, UDP_MIB_RCVBUFERRORS, |
1837 | IS_UDPLITE(sk)); | |
1838 | __UDP_INC_STATS(net, UDP_MIB_INERRORS, | |
1839 | IS_UDPLITE(sk)); | |
ca065d0c ED |
1840 | continue; |
1841 | } | |
1842 | if (udp_queue_rcv_skb(sk, nskb) > 0) | |
1843 | consume_skb(nskb); | |
1844 | } | |
1240d137 | 1845 | |
2dc41cff DH |
1846 | /* Also lookup *:port if we are using hash2 and haven't done so yet. */ |
1847 | if (use_hash2 && hash2 != hash2_any) { | |
1848 | hash2 = hash2_any; | |
1849 | goto start_lookup; | |
1850 | } | |
1851 | ||
ca065d0c ED |
1852 | if (first) { |
1853 | if (udp_queue_rcv_skb(first, skb) > 0) | |
1854 | consume_skb(skb); | |
1240d137 | 1855 | } else { |
ca065d0c | 1856 | kfree_skb(skb); |
02c22347 ED |
1857 | __UDP_INC_STATS(net, UDP_MIB_IGNOREDMULTI, |
1858 | proto == IPPROTO_UDPLITE); | |
1240d137 | 1859 | } |
db8dac20 DM |
1860 | return 0; |
1861 | } | |
1862 | ||
1863 | /* Initialize UDP checksum. If exited with zero value (success), | |
1864 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1865 | * Otherwise, csum completion requires chacksumming packet body, | |
1866 | * including udp header and folding it to skb->csum. | |
1867 | */ | |
1868 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1869 | int proto) | |
1870 | { | |
db8dac20 DM |
1871 | int err; |
1872 | ||
1873 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1874 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1875 | ||
1876 | if (proto == IPPROTO_UDPLITE) { | |
1877 | err = udplite_checksum_init(skb, uh); | |
1878 | if (err) | |
1879 | return err; | |
1880 | } | |
1881 | ||
b46d9f62 HFS |
1882 | /* Note, we are only interested in != 0 or == 0, thus the |
1883 | * force to int. | |
1884 | */ | |
1885 | return (__force int)skb_checksum_init_zero_check(skb, proto, uh->check, | |
1886 | inet_compute_pseudo); | |
db8dac20 DM |
1887 | } |
1888 | ||
1889 | /* | |
1890 | * All we need to do is get the socket, and then do a checksum. | |
1891 | */ | |
1892 | ||
645ca708 | 1893 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1894 | int proto) |
1895 | { | |
1896 | struct sock *sk; | |
7b5e56f9 | 1897 | struct udphdr *uh; |
db8dac20 | 1898 | unsigned short ulen; |
adf30907 | 1899 | struct rtable *rt = skb_rtable(skb); |
2783ef23 | 1900 | __be32 saddr, daddr; |
0283328e | 1901 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1902 | |
1903 | /* | |
1904 | * Validate the packet. | |
1905 | */ | |
1906 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1907 | goto drop; /* No space for header. */ | |
1908 | ||
7b5e56f9 | 1909 | uh = udp_hdr(skb); |
db8dac20 | 1910 | ulen = ntohs(uh->len); |
ccc2d97c BM |
1911 | saddr = ip_hdr(skb)->saddr; |
1912 | daddr = ip_hdr(skb)->daddr; | |
1913 | ||
db8dac20 DM |
1914 | if (ulen > skb->len) |
1915 | goto short_packet; | |
1916 | ||
1917 | if (proto == IPPROTO_UDP) { | |
1918 | /* UDP validates ulen. */ | |
1919 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1920 | goto short_packet; | |
1921 | uh = udp_hdr(skb); | |
1922 | } | |
1923 | ||
1924 | if (udp4_csum_init(skb, uh, proto)) | |
1925 | goto csum_error; | |
1926 | ||
8afdd99a ED |
1927 | sk = skb_steal_sock(skb); |
1928 | if (sk) { | |
97502231 | 1929 | struct dst_entry *dst = skb_dst(skb); |
421b3885 | 1930 | int ret; |
421b3885 | 1931 | |
97502231 ED |
1932 | if (unlikely(sk->sk_rx_dst != dst)) |
1933 | udp_sk_rx_dst_set(sk, dst); | |
db8dac20 | 1934 | |
421b3885 | 1935 | ret = udp_queue_rcv_skb(sk, skb); |
8afdd99a | 1936 | sock_put(sk); |
421b3885 SB |
1937 | /* a return value > 0 means to resubmit the input, but |
1938 | * it wants the return to be -protocol, or 0 | |
1939 | */ | |
1940 | if (ret > 0) | |
1941 | return -ret; | |
1942 | return 0; | |
421b3885 | 1943 | } |
db8dac20 | 1944 | |
c18450a5 FF |
1945 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) |
1946 | return __udp4_lib_mcast_deliver(net, skb, uh, | |
36cbb245 | 1947 | saddr, daddr, udptable, proto); |
c18450a5 FF |
1948 | |
1949 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); | |
00db4124 | 1950 | if (sk) { |
a5b50476 ET |
1951 | int ret; |
1952 | ||
224d019c | 1953 | if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk)) |
2abb7cdc TH |
1954 | skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check, |
1955 | inet_compute_pseudo); | |
1956 | ||
a5b50476 | 1957 | ret = udp_queue_rcv_skb(sk, skb); |
db8dac20 DM |
1958 | |
1959 | /* a return value > 0 means to resubmit the input, but | |
1960 | * it wants the return to be -protocol, or 0 | |
1961 | */ | |
1962 | if (ret > 0) | |
1963 | return -ret; | |
1964 | return 0; | |
1965 | } | |
1966 | ||
1967 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1968 | goto drop; | |
1969 | nf_reset(skb); | |
1970 | ||
1971 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1972 | if (udp_lib_checksum_complete(skb)) | |
1973 | goto csum_error; | |
1974 | ||
02c22347 | 1975 | __UDP_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1976 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1977 | ||
1978 | /* | |
1979 | * Hmm. We got an UDP packet to a port to which we | |
1980 | * don't wanna listen. Ignore it. | |
1981 | */ | |
1982 | kfree_skb(skb); | |
1983 | return 0; | |
1984 | ||
1985 | short_packet: | |
ba7a46f1 JP |
1986 | net_dbg_ratelimited("UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n", |
1987 | proto == IPPROTO_UDPLITE ? "Lite" : "", | |
1988 | &saddr, ntohs(uh->source), | |
1989 | ulen, skb->len, | |
1990 | &daddr, ntohs(uh->dest)); | |
db8dac20 DM |
1991 | goto drop; |
1992 | ||
1993 | csum_error: | |
1994 | /* | |
1995 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1996 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1997 | */ | |
ba7a46f1 JP |
1998 | net_dbg_ratelimited("UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n", |
1999 | proto == IPPROTO_UDPLITE ? "Lite" : "", | |
2000 | &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest), | |
2001 | ulen); | |
02c22347 | 2002 | __UDP_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 | 2003 | drop: |
02c22347 | 2004 | __UDP_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
2005 | kfree_skb(skb); |
2006 | return 0; | |
2007 | } | |
2008 | ||
421b3885 SB |
2009 | /* We can only early demux multicast if there is a single matching socket. |
2010 | * If more than one socket found returns NULL | |
2011 | */ | |
2012 | static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net, | |
2013 | __be16 loc_port, __be32 loc_addr, | |
2014 | __be16 rmt_port, __be32 rmt_addr, | |
2015 | int dif) | |
2016 | { | |
2017 | struct sock *sk, *result; | |
421b3885 | 2018 | unsigned short hnum = ntohs(loc_port); |
ca065d0c | 2019 | unsigned int slot = udp_hashfn(net, hnum, udp_table.mask); |
421b3885 SB |
2020 | struct udp_hslot *hslot = &udp_table.hash[slot]; |
2021 | ||
63c6f81c ED |
2022 | /* Do not bother scanning a too big list */ |
2023 | if (hslot->count > 10) | |
2024 | return NULL; | |
2025 | ||
421b3885 | 2026 | result = NULL; |
ca065d0c ED |
2027 | sk_for_each_rcu(sk, &hslot->head) { |
2028 | if (__udp_is_mcast_sock(net, sk, loc_port, loc_addr, | |
2029 | rmt_port, rmt_addr, dif, hnum)) { | |
2030 | if (result) | |
2031 | return NULL; | |
421b3885 | 2032 | result = sk; |
421b3885 SB |
2033 | } |
2034 | } | |
ca065d0c | 2035 | |
421b3885 SB |
2036 | return result; |
2037 | } | |
2038 | ||
2039 | /* For unicast we should only early demux connected sockets or we can | |
2040 | * break forwarding setups. The chains here can be long so only check | |
2041 | * if the first socket is an exact match and if not move on. | |
2042 | */ | |
2043 | static struct sock *__udp4_lib_demux_lookup(struct net *net, | |
2044 | __be16 loc_port, __be32 loc_addr, | |
2045 | __be16 rmt_port, __be32 rmt_addr, | |
2046 | int dif) | |
2047 | { | |
421b3885 SB |
2048 | unsigned short hnum = ntohs(loc_port); |
2049 | unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum); | |
2050 | unsigned int slot2 = hash2 & udp_table.mask; | |
2051 | struct udp_hslot *hslot2 = &udp_table.hash2[slot2]; | |
c7228317 | 2052 | INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr); |
421b3885 | 2053 | const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum); |
ca065d0c | 2054 | struct sock *sk; |
421b3885 | 2055 | |
ca065d0c ED |
2056 | udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
2057 | if (INET_MATCH(sk, net, acookie, rmt_addr, | |
2058 | loc_addr, ports, dif)) | |
2059 | return sk; | |
421b3885 SB |
2060 | /* Only check first socket in chain */ |
2061 | break; | |
2062 | } | |
ca065d0c | 2063 | return NULL; |
421b3885 SB |
2064 | } |
2065 | ||
2066 | void udp_v4_early_demux(struct sk_buff *skb) | |
2067 | { | |
610438b7 ED |
2068 | struct net *net = dev_net(skb->dev); |
2069 | const struct iphdr *iph; | |
2070 | const struct udphdr *uh; | |
ca065d0c | 2071 | struct sock *sk = NULL; |
421b3885 | 2072 | struct dst_entry *dst; |
421b3885 | 2073 | int dif = skb->dev->ifindex; |
6e540309 | 2074 | int ours; |
421b3885 SB |
2075 | |
2076 | /* validate the packet */ | |
2077 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr))) | |
2078 | return; | |
2079 | ||
610438b7 ED |
2080 | iph = ip_hdr(skb); |
2081 | uh = udp_hdr(skb); | |
2082 | ||
421b3885 | 2083 | if (skb->pkt_type == PACKET_BROADCAST || |
6e540309 SB |
2084 | skb->pkt_type == PACKET_MULTICAST) { |
2085 | struct in_device *in_dev = __in_dev_get_rcu(skb->dev); | |
2086 | ||
2087 | if (!in_dev) | |
2088 | return; | |
2089 | ||
ad0ea198 PA |
2090 | /* we are supposed to accept bcast packets */ |
2091 | if (skb->pkt_type == PACKET_MULTICAST) { | |
2092 | ours = ip_check_mc_rcu(in_dev, iph->daddr, iph->saddr, | |
2093 | iph->protocol); | |
2094 | if (!ours) | |
2095 | return; | |
2096 | } | |
2097 | ||
421b3885 SB |
2098 | sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr, |
2099 | uh->source, iph->saddr, dif); | |
6e540309 | 2100 | } else if (skb->pkt_type == PACKET_HOST) { |
421b3885 SB |
2101 | sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr, |
2102 | uh->source, iph->saddr, dif); | |
6e540309 | 2103 | } |
421b3885 | 2104 | |
ca065d0c | 2105 | if (!sk || !atomic_inc_not_zero_hint(&sk->sk_refcnt, 2)) |
421b3885 SB |
2106 | return; |
2107 | ||
2108 | skb->sk = sk; | |
82eabd9e | 2109 | skb->destructor = sock_efree; |
10e2eb87 | 2110 | dst = READ_ONCE(sk->sk_rx_dst); |
421b3885 SB |
2111 | |
2112 | if (dst) | |
2113 | dst = dst_check(dst, 0); | |
10e2eb87 ED |
2114 | if (dst) { |
2115 | /* DST_NOCACHE can not be used without taking a reference */ | |
2116 | if (dst->flags & DST_NOCACHE) { | |
2117 | if (likely(atomic_inc_not_zero(&dst->__refcnt))) | |
2118 | skb_dst_set(skb, dst); | |
2119 | } else { | |
2120 | skb_dst_set_noref(skb, dst); | |
2121 | } | |
2122 | } | |
421b3885 SB |
2123 | } |
2124 | ||
db8dac20 DM |
2125 | int udp_rcv(struct sk_buff *skb) |
2126 | { | |
645ca708 | 2127 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
2128 | } |
2129 | ||
7d06b2e0 | 2130 | void udp_destroy_sock(struct sock *sk) |
db8dac20 | 2131 | { |
44046a59 | 2132 | struct udp_sock *up = udp_sk(sk); |
8a74ad60 | 2133 | bool slow = lock_sock_fast(sk); |
db8dac20 | 2134 | udp_flush_pending_frames(sk); |
8a74ad60 | 2135 | unlock_sock_fast(sk, slow); |
44046a59 TP |
2136 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
2137 | void (*encap_destroy)(struct sock *sk); | |
2138 | encap_destroy = ACCESS_ONCE(up->encap_destroy); | |
2139 | if (encap_destroy) | |
2140 | encap_destroy(sk); | |
2141 | } | |
db8dac20 DM |
2142 | } |
2143 | ||
1da177e4 LT |
2144 | /* |
2145 | * Socket option code for UDP | |
2146 | */ | |
4c0a6cb0 | 2147 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 2148 | char __user *optval, unsigned int optlen, |
4c0a6cb0 | 2149 | int (*push_pending_frames)(struct sock *)) |
1da177e4 LT |
2150 | { |
2151 | struct udp_sock *up = udp_sk(sk); | |
1c19448c | 2152 | int val, valbool; |
1da177e4 | 2153 | int err = 0; |
b2bf1e26 | 2154 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 2155 | |
c482c568 | 2156 | if (optlen < sizeof(int)) |
1da177e4 LT |
2157 | return -EINVAL; |
2158 | ||
2159 | if (get_user(val, (int __user *)optval)) | |
2160 | return -EFAULT; | |
2161 | ||
1c19448c TH |
2162 | valbool = val ? 1 : 0; |
2163 | ||
6516c655 | 2164 | switch (optname) { |
1da177e4 LT |
2165 | case UDP_CORK: |
2166 | if (val != 0) { | |
2167 | up->corkflag = 1; | |
2168 | } else { | |
2169 | up->corkflag = 0; | |
2170 | lock_sock(sk); | |
4243cdc2 | 2171 | push_pending_frames(sk); |
1da177e4 LT |
2172 | release_sock(sk); |
2173 | } | |
2174 | break; | |
e905a9ed | 2175 | |
1da177e4 LT |
2176 | case UDP_ENCAP: |
2177 | switch (val) { | |
2178 | case 0: | |
2179 | case UDP_ENCAP_ESPINUDP: | |
2180 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
2181 | up->encap_rcv = xfrm4_udp_encap_rcv; |
2182 | /* FALLTHROUGH */ | |
342f0234 | 2183 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 | 2184 | up->encap_type = val; |
447167bf | 2185 | udp_encap_enable(); |
1da177e4 LT |
2186 | break; |
2187 | default: | |
2188 | err = -ENOPROTOOPT; | |
2189 | break; | |
2190 | } | |
2191 | break; | |
2192 | ||
1c19448c TH |
2193 | case UDP_NO_CHECK6_TX: |
2194 | up->no_check6_tx = valbool; | |
2195 | break; | |
2196 | ||
2197 | case UDP_NO_CHECK6_RX: | |
2198 | up->no_check6_rx = valbool; | |
2199 | break; | |
2200 | ||
ba4e58ec GR |
2201 | /* |
2202 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
2203 | */ | |
2204 | /* The sender sets actual checksum coverage length via this option. | |
2205 | * The case coverage > packet length is handled by send module. */ | |
2206 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 2207 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
2208 | return -ENOPROTOOPT; |
2209 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
2210 | val = 8; | |
4be929be AD |
2211 | else if (val > USHRT_MAX) |
2212 | val = USHRT_MAX; | |
ba4e58ec GR |
2213 | up->pcslen = val; |
2214 | up->pcflag |= UDPLITE_SEND_CC; | |
2215 | break; | |
2216 | ||
e905a9ed YH |
2217 | /* The receiver specifies a minimum checksum coverage value. To make |
2218 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
2219 | * used, this again means full checksum coverage. */ |
2220 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 2221 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
2222 | return -ENOPROTOOPT; |
2223 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
2224 | val = 8; | |
4be929be AD |
2225 | else if (val > USHRT_MAX) |
2226 | val = USHRT_MAX; | |
ba4e58ec GR |
2227 | up->pcrlen = val; |
2228 | up->pcflag |= UDPLITE_RECV_CC; | |
2229 | break; | |
2230 | ||
1da177e4 LT |
2231 | default: |
2232 | err = -ENOPROTOOPT; | |
2233 | break; | |
6516c655 | 2234 | } |
1da177e4 LT |
2235 | |
2236 | return err; | |
2237 | } | |
c482c568 | 2238 | EXPORT_SYMBOL(udp_lib_setsockopt); |
1da177e4 | 2239 | |
db8dac20 | 2240 | int udp_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 2241 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
2242 | { |
2243 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2244 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
2245 | udp_push_pending_frames); | |
2246 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
2247 | } | |
2248 | ||
2249 | #ifdef CONFIG_COMPAT | |
2250 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 2251 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
2252 | { |
2253 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2254 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
2255 | udp_push_pending_frames); | |
2256 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
2257 | } | |
2258 | #endif | |
2259 | ||
4c0a6cb0 GR |
2260 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
2261 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
2262 | { |
2263 | struct udp_sock *up = udp_sk(sk); | |
2264 | int val, len; | |
2265 | ||
c482c568 | 2266 | if (get_user(len, optlen)) |
1da177e4 LT |
2267 | return -EFAULT; |
2268 | ||
2269 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 2270 | |
6516c655 | 2271 | if (len < 0) |
1da177e4 LT |
2272 | return -EINVAL; |
2273 | ||
6516c655 | 2274 | switch (optname) { |
1da177e4 LT |
2275 | case UDP_CORK: |
2276 | val = up->corkflag; | |
2277 | break; | |
2278 | ||
2279 | case UDP_ENCAP: | |
2280 | val = up->encap_type; | |
2281 | break; | |
2282 | ||
1c19448c TH |
2283 | case UDP_NO_CHECK6_TX: |
2284 | val = up->no_check6_tx; | |
2285 | break; | |
2286 | ||
2287 | case UDP_NO_CHECK6_RX: | |
2288 | val = up->no_check6_rx; | |
2289 | break; | |
2290 | ||
ba4e58ec GR |
2291 | /* The following two cannot be changed on UDP sockets, the return is |
2292 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
2293 | case UDPLITE_SEND_CSCOV: | |
2294 | val = up->pcslen; | |
2295 | break; | |
2296 | ||
2297 | case UDPLITE_RECV_CSCOV: | |
2298 | val = up->pcrlen; | |
2299 | break; | |
2300 | ||
1da177e4 LT |
2301 | default: |
2302 | return -ENOPROTOOPT; | |
6516c655 | 2303 | } |
1da177e4 | 2304 | |
6516c655 | 2305 | if (put_user(len, optlen)) |
e905a9ed | 2306 | return -EFAULT; |
c482c568 | 2307 | if (copy_to_user(optval, &val, len)) |
1da177e4 | 2308 | return -EFAULT; |
e905a9ed | 2309 | return 0; |
1da177e4 | 2310 | } |
c482c568 | 2311 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1da177e4 | 2312 | |
db8dac20 DM |
2313 | int udp_getsockopt(struct sock *sk, int level, int optname, |
2314 | char __user *optval, int __user *optlen) | |
2315 | { | |
2316 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2317 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
2318 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
2319 | } | |
2320 | ||
2321 | #ifdef CONFIG_COMPAT | |
2322 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
2323 | char __user *optval, int __user *optlen) | |
2324 | { | |
2325 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2326 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
2327 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
2328 | } | |
2329 | #endif | |
1da177e4 LT |
2330 | /** |
2331 | * udp_poll - wait for a UDP event. | |
2332 | * @file - file struct | |
2333 | * @sock - socket | |
2334 | * @wait - poll table | |
2335 | * | |
e905a9ed | 2336 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
2337 | * blocking sockets. If application is using a blocking fd |
2338 | * and a packet with checksum error is in the queue; | |
2339 | * then it could get return from select indicating data available | |
2340 | * but then block when reading it. Add special case code | |
2341 | * to work around these arguably broken applications. | |
2342 | */ | |
2343 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
2344 | { | |
2345 | unsigned int mask = datagram_poll(file, sock, wait); | |
2346 | struct sock *sk = sock->sk; | |
ba4e58ec | 2347 | |
c3f1dbaf DM |
2348 | sock_rps_record_flow(sk); |
2349 | ||
1da177e4 | 2350 | /* Check for false positives due to checksum errors */ |
85584672 | 2351 | if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) && |
e83c6744 | 2352 | !(sk->sk_shutdown & RCV_SHUTDOWN) && first_packet_length(sk) == -1) |
85584672 | 2353 | mask &= ~(POLLIN | POLLRDNORM); |
1da177e4 LT |
2354 | |
2355 | return mask; | |
e905a9ed | 2356 | |
1da177e4 | 2357 | } |
c482c568 | 2358 | EXPORT_SYMBOL(udp_poll); |
1da177e4 | 2359 | |
5d77dca8 DA |
2360 | int udp_abort(struct sock *sk, int err) |
2361 | { | |
2362 | lock_sock(sk); | |
2363 | ||
2364 | sk->sk_err = err; | |
2365 | sk->sk_error_report(sk); | |
286c72de | 2366 | __udp_disconnect(sk, 0); |
5d77dca8 DA |
2367 | |
2368 | release_sock(sk); | |
2369 | ||
2370 | return 0; | |
2371 | } | |
2372 | EXPORT_SYMBOL_GPL(udp_abort); | |
2373 | ||
db8dac20 DM |
2374 | struct proto udp_prot = { |
2375 | .name = "UDP", | |
2376 | .owner = THIS_MODULE, | |
2377 | .close = udp_lib_close, | |
2378 | .connect = ip4_datagram_connect, | |
2379 | .disconnect = udp_disconnect, | |
2380 | .ioctl = udp_ioctl, | |
850cbadd | 2381 | .init = udp_init_sock, |
db8dac20 DM |
2382 | .destroy = udp_destroy_sock, |
2383 | .setsockopt = udp_setsockopt, | |
2384 | .getsockopt = udp_getsockopt, | |
2385 | .sendmsg = udp_sendmsg, | |
2386 | .recvmsg = udp_recvmsg, | |
2387 | .sendpage = udp_sendpage, | |
8141ed9f | 2388 | .release_cb = ip4_datagram_release_cb, |
db8dac20 DM |
2389 | .hash = udp_lib_hash, |
2390 | .unhash = udp_lib_unhash, | |
719f8358 | 2391 | .rehash = udp_v4_rehash, |
db8dac20 DM |
2392 | .get_port = udp_v4_get_port, |
2393 | .memory_allocated = &udp_memory_allocated, | |
2394 | .sysctl_mem = sysctl_udp_mem, | |
2395 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
2396 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
2397 | .obj_size = sizeof(struct udp_sock), | |
645ca708 | 2398 | .h.udp_table = &udp_table, |
db8dac20 DM |
2399 | #ifdef CONFIG_COMPAT |
2400 | .compat_setsockopt = compat_udp_setsockopt, | |
2401 | .compat_getsockopt = compat_udp_getsockopt, | |
2402 | #endif | |
5d77dca8 | 2403 | .diag_destroy = udp_abort, |
db8dac20 | 2404 | }; |
c482c568 | 2405 | EXPORT_SYMBOL(udp_prot); |
1da177e4 LT |
2406 | |
2407 | /* ------------------------------------------------------------------------ */ | |
2408 | #ifdef CONFIG_PROC_FS | |
2409 | ||
645ca708 | 2410 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
2411 | { |
2412 | struct sock *sk; | |
2413 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2414 | struct net *net = seq_file_net(seq); |
1da177e4 | 2415 | |
f86dcc5a ED |
2416 | for (state->bucket = start; state->bucket <= state->udp_table->mask; |
2417 | ++state->bucket) { | |
645ca708 | 2418 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
f86dcc5a | 2419 | |
ca065d0c | 2420 | if (hlist_empty(&hslot->head)) |
f86dcc5a ED |
2421 | continue; |
2422 | ||
645ca708 | 2423 | spin_lock_bh(&hslot->lock); |
ca065d0c | 2424 | sk_for_each(sk, &hslot->head) { |
878628fb | 2425 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 2426 | continue; |
1da177e4 LT |
2427 | if (sk->sk_family == state->family) |
2428 | goto found; | |
2429 | } | |
645ca708 | 2430 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
2431 | } |
2432 | sk = NULL; | |
2433 | found: | |
2434 | return sk; | |
2435 | } | |
2436 | ||
2437 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
2438 | { | |
2439 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2440 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2441 | |
2442 | do { | |
ca065d0c | 2443 | sk = sk_next(sk); |
878628fb | 2444 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 2445 | |
645ca708 | 2446 | if (!sk) { |
f86dcc5a | 2447 | if (state->bucket <= state->udp_table->mask) |
30842f29 | 2448 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
645ca708 | 2449 | return udp_get_first(seq, state->bucket + 1); |
1da177e4 LT |
2450 | } |
2451 | return sk; | |
2452 | } | |
2453 | ||
2454 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
2455 | { | |
645ca708 | 2456 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
2457 | |
2458 | if (sk) | |
6516c655 | 2459 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
2460 | --pos; |
2461 | return pos ? NULL : sk; | |
2462 | } | |
2463 | ||
2464 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
2465 | { | |
30842f29 | 2466 | struct udp_iter_state *state = seq->private; |
f86dcc5a | 2467 | state->bucket = MAX_UDP_PORTS; |
30842f29 | 2468 | |
b50660f1 | 2469 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
2470 | } |
2471 | ||
2472 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2473 | { | |
2474 | struct sock *sk; | |
2475 | ||
b50660f1 | 2476 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2477 | sk = udp_get_idx(seq, 0); |
2478 | else | |
2479 | sk = udp_get_next(seq, v); | |
2480 | ||
2481 | ++*pos; | |
2482 | return sk; | |
2483 | } | |
2484 | ||
2485 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
2486 | { | |
645ca708 ED |
2487 | struct udp_iter_state *state = seq->private; |
2488 | ||
f86dcc5a | 2489 | if (state->bucket <= state->udp_table->mask) |
645ca708 | 2490 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
1da177e4 LT |
2491 | } |
2492 | ||
73cb88ec | 2493 | int udp_seq_open(struct inode *inode, struct file *file) |
1da177e4 | 2494 | { |
d9dda78b | 2495 | struct udp_seq_afinfo *afinfo = PDE_DATA(inode); |
a2be75c1 DL |
2496 | struct udp_iter_state *s; |
2497 | int err; | |
a91275ef | 2498 | |
a2be75c1 DL |
2499 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2500 | sizeof(struct udp_iter_state)); | |
2501 | if (err < 0) | |
2502 | return err; | |
a91275ef | 2503 | |
a2be75c1 | 2504 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2505 | s->family = afinfo->family; |
645ca708 | 2506 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 2507 | return err; |
a91275ef | 2508 | } |
73cb88ec | 2509 | EXPORT_SYMBOL(udp_seq_open); |
a91275ef | 2510 | |
1da177e4 | 2511 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 2512 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
2513 | { |
2514 | struct proc_dir_entry *p; | |
2515 | int rc = 0; | |
2516 | ||
dda61925 DL |
2517 | afinfo->seq_ops.start = udp_seq_start; |
2518 | afinfo->seq_ops.next = udp_seq_next; | |
2519 | afinfo->seq_ops.stop = udp_seq_stop; | |
2520 | ||
84841c3c | 2521 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2522 | afinfo->seq_fops, afinfo); |
84841c3c | 2523 | if (!p) |
1da177e4 LT |
2524 | rc = -ENOMEM; |
2525 | return rc; | |
2526 | } | |
c482c568 | 2527 | EXPORT_SYMBOL(udp_proc_register); |
1da177e4 | 2528 | |
0c96d8c5 | 2529 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 2530 | { |
ece31ffd | 2531 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2532 | } |
c482c568 | 2533 | EXPORT_SYMBOL(udp_proc_unregister); |
db8dac20 DM |
2534 | |
2535 | /* ------------------------------------------------------------------------ */ | |
5e659e4c | 2536 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
652586df | 2537 | int bucket) |
db8dac20 DM |
2538 | { |
2539 | struct inet_sock *inet = inet_sk(sp); | |
c720c7e8 ED |
2540 | __be32 dest = inet->inet_daddr; |
2541 | __be32 src = inet->inet_rcv_saddr; | |
2542 | __u16 destp = ntohs(inet->inet_dport); | |
2543 | __u16 srcp = ntohs(inet->inet_sport); | |
db8dac20 | 2544 | |
f86dcc5a | 2545 | seq_printf(f, "%5d: %08X:%04X %08X:%04X" |
652586df | 2546 | " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d", |
db8dac20 | 2547 | bucket, src, srcp, dest, destp, sp->sk_state, |
31e6d363 ED |
2548 | sk_wmem_alloc_get(sp), |
2549 | sk_rmem_alloc_get(sp), | |
a7cb5a49 EB |
2550 | 0, 0L, 0, |
2551 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)), | |
2552 | 0, sock_i_ino(sp), | |
cb61cb9b | 2553 | atomic_read(&sp->sk_refcnt), sp, |
652586df | 2554 | atomic_read(&sp->sk_drops)); |
db8dac20 DM |
2555 | } |
2556 | ||
2557 | int udp4_seq_show(struct seq_file *seq, void *v) | |
2558 | { | |
652586df | 2559 | seq_setwidth(seq, 127); |
db8dac20 | 2560 | if (v == SEQ_START_TOKEN) |
652586df | 2561 | seq_puts(seq, " sl local_address rem_address st tx_queue " |
db8dac20 | 2562 | "rx_queue tr tm->when retrnsmt uid timeout " |
cb61cb9b | 2563 | "inode ref pointer drops"); |
db8dac20 | 2564 | else { |
db8dac20 DM |
2565 | struct udp_iter_state *state = seq->private; |
2566 | ||
652586df | 2567 | udp4_format_sock(v, seq, state->bucket); |
db8dac20 | 2568 | } |
652586df | 2569 | seq_pad(seq, '\n'); |
db8dac20 DM |
2570 | return 0; |
2571 | } | |
2572 | ||
73cb88ec AV |
2573 | static const struct file_operations udp_afinfo_seq_fops = { |
2574 | .owner = THIS_MODULE, | |
2575 | .open = udp_seq_open, | |
2576 | .read = seq_read, | |
2577 | .llseek = seq_lseek, | |
2578 | .release = seq_release_net | |
2579 | }; | |
2580 | ||
db8dac20 | 2581 | /* ------------------------------------------------------------------------ */ |
db8dac20 | 2582 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
2583 | .name = "udp", |
2584 | .family = AF_INET, | |
645ca708 | 2585 | .udp_table = &udp_table, |
73cb88ec | 2586 | .seq_fops = &udp_afinfo_seq_fops, |
dda61925 DL |
2587 | .seq_ops = { |
2588 | .show = udp4_seq_show, | |
2589 | }, | |
db8dac20 DM |
2590 | }; |
2591 | ||
2c8c1e72 | 2592 | static int __net_init udp4_proc_init_net(struct net *net) |
15439feb PE |
2593 | { |
2594 | return udp_proc_register(net, &udp4_seq_afinfo); | |
2595 | } | |
2596 | ||
2c8c1e72 | 2597 | static void __net_exit udp4_proc_exit_net(struct net *net) |
15439feb PE |
2598 | { |
2599 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
2600 | } | |
2601 | ||
2602 | static struct pernet_operations udp4_net_ops = { | |
2603 | .init = udp4_proc_init_net, | |
2604 | .exit = udp4_proc_exit_net, | |
2605 | }; | |
2606 | ||
db8dac20 DM |
2607 | int __init udp4_proc_init(void) |
2608 | { | |
15439feb | 2609 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
2610 | } |
2611 | ||
2612 | void udp4_proc_exit(void) | |
2613 | { | |
15439feb | 2614 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 2615 | } |
1da177e4 LT |
2616 | #endif /* CONFIG_PROC_FS */ |
2617 | ||
f86dcc5a ED |
2618 | static __initdata unsigned long uhash_entries; |
2619 | static int __init set_uhash_entries(char *str) | |
645ca708 | 2620 | { |
413c27d8 EZ |
2621 | ssize_t ret; |
2622 | ||
f86dcc5a ED |
2623 | if (!str) |
2624 | return 0; | |
413c27d8 EZ |
2625 | |
2626 | ret = kstrtoul(str, 0, &uhash_entries); | |
2627 | if (ret) | |
2628 | return 0; | |
2629 | ||
f86dcc5a ED |
2630 | if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN) |
2631 | uhash_entries = UDP_HTABLE_SIZE_MIN; | |
2632 | return 1; | |
2633 | } | |
2634 | __setup("uhash_entries=", set_uhash_entries); | |
645ca708 | 2635 | |
f86dcc5a ED |
2636 | void __init udp_table_init(struct udp_table *table, const char *name) |
2637 | { | |
2638 | unsigned int i; | |
2639 | ||
31fe62b9 TB |
2640 | table->hash = alloc_large_system_hash(name, |
2641 | 2 * sizeof(struct udp_hslot), | |
2642 | uhash_entries, | |
2643 | 21, /* one slot per 2 MB */ | |
2644 | 0, | |
2645 | &table->log, | |
2646 | &table->mask, | |
2647 | UDP_HTABLE_SIZE_MIN, | |
2648 | 64 * 1024); | |
2649 | ||
512615b6 | 2650 | table->hash2 = table->hash + (table->mask + 1); |
f86dcc5a | 2651 | for (i = 0; i <= table->mask; i++) { |
ca065d0c | 2652 | INIT_HLIST_HEAD(&table->hash[i].head); |
fdcc8aa9 | 2653 | table->hash[i].count = 0; |
645ca708 ED |
2654 | spin_lock_init(&table->hash[i].lock); |
2655 | } | |
512615b6 | 2656 | for (i = 0; i <= table->mask; i++) { |
ca065d0c | 2657 | INIT_HLIST_HEAD(&table->hash2[i].head); |
512615b6 ED |
2658 | table->hash2[i].count = 0; |
2659 | spin_lock_init(&table->hash2[i].lock); | |
2660 | } | |
645ca708 ED |
2661 | } |
2662 | ||
723b8e46 TH |
2663 | u32 udp_flow_hashrnd(void) |
2664 | { | |
2665 | static u32 hashrnd __read_mostly; | |
2666 | ||
2667 | net_get_random_once(&hashrnd, sizeof(hashrnd)); | |
2668 | ||
2669 | return hashrnd; | |
2670 | } | |
2671 | EXPORT_SYMBOL(udp_flow_hashrnd); | |
2672 | ||
95766fff HA |
2673 | void __init udp_init(void) |
2674 | { | |
f03d78db | 2675 | unsigned long limit; |
4b272750 | 2676 | unsigned int i; |
95766fff | 2677 | |
f86dcc5a | 2678 | udp_table_init(&udp_table, "UDP"); |
f03d78db | 2679 | limit = nr_free_buffer_pages() / 8; |
95766fff HA |
2680 | limit = max(limit, 128UL); |
2681 | sysctl_udp_mem[0] = limit / 4 * 3; | |
2682 | sysctl_udp_mem[1] = limit; | |
2683 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
2684 | ||
2685 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
2686 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
4b272750 ED |
2687 | |
2688 | /* 16 spinlocks per cpu */ | |
2689 | udp_busylocks_log = ilog2(nr_cpu_ids) + 4; | |
2690 | udp_busylocks = kmalloc(sizeof(spinlock_t) << udp_busylocks_log, | |
2691 | GFP_KERNEL); | |
2692 | if (!udp_busylocks) | |
2693 | panic("UDP: failed to alloc udp_busylocks\n"); | |
2694 | for (i = 0; i < (1U << udp_busylocks_log); i++) | |
2695 | spin_lock_init(udp_busylocks + i); | |
95766fff | 2696 | } |