]>
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 | ||
1da177e4 LT |
82 | #include <asm/uaccess.h> |
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> |
1da177e4 | 117 | |
f86dcc5a | 118 | struct udp_table udp_table __read_mostly; |
645ca708 | 119 | EXPORT_SYMBOL(udp_table); |
1da177e4 | 120 | |
8d987e5c | 121 | long sysctl_udp_mem[3] __read_mostly; |
95766fff | 122 | EXPORT_SYMBOL(sysctl_udp_mem); |
c482c568 ED |
123 | |
124 | int sysctl_udp_rmem_min __read_mostly; | |
95766fff | 125 | EXPORT_SYMBOL(sysctl_udp_rmem_min); |
c482c568 ED |
126 | |
127 | int sysctl_udp_wmem_min __read_mostly; | |
95766fff HA |
128 | EXPORT_SYMBOL(sysctl_udp_wmem_min); |
129 | ||
8d987e5c | 130 | atomic_long_t udp_memory_allocated; |
95766fff HA |
131 | EXPORT_SYMBOL(udp_memory_allocated); |
132 | ||
f86dcc5a ED |
133 | #define MAX_UDP_PORTS 65536 |
134 | #define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN) | |
98322f22 | 135 | |
f24d43c0 | 136 | static int udp_lib_lport_inuse(struct net *net, __u16 num, |
645ca708 | 137 | const struct udp_hslot *hslot, |
98322f22 | 138 | unsigned long *bitmap, |
f24d43c0 ED |
139 | struct sock *sk, |
140 | int (*saddr_comp)(const struct sock *sk1, | |
e32ea7e7 CG |
141 | const struct sock *sk2, |
142 | bool match_wildcard), | |
f86dcc5a | 143 | unsigned int log) |
1da177e4 | 144 | { |
f24d43c0 | 145 | struct sock *sk2; |
88ab1932 | 146 | struct hlist_nulls_node *node; |
ba418fa3 | 147 | kuid_t uid = sock_i_uid(sk); |
25030a7f | 148 | |
4243cdc2 | 149 | sk_nulls_for_each(sk2, node, &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) && | |
ba418fa3 | 156 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
e32ea7e7 | 157 | rcu_access_pointer(sk->sk_reuseport_cb) || |
4243cdc2 | 158 | !uid_eq(uid, sock_i_uid(sk2))) && |
e32ea7e7 | 159 | saddr_comp(sk, sk2, true)) { |
4243cdc2 | 160 | if (!bitmap) |
98322f22 | 161 | return 1; |
4243cdc2 | 162 | __set_bit(udp_sk(sk2)->udp_port_hash >> log, bitmap); |
98322f22 | 163 | } |
4243cdc2 | 164 | } |
25030a7f GR |
165 | return 0; |
166 | } | |
167 | ||
30fff923 ED |
168 | /* |
169 | * Note: we still hold spinlock of primary hash chain, so no other writer | |
170 | * can insert/delete a socket with local_port == num | |
171 | */ | |
172 | static int udp_lib_lport_inuse2(struct net *net, __u16 num, | |
4243cdc2 JP |
173 | struct udp_hslot *hslot2, |
174 | struct sock *sk, | |
175 | int (*saddr_comp)(const struct sock *sk1, | |
e32ea7e7 CG |
176 | const struct sock *sk2, |
177 | bool match_wildcard)) | |
30fff923 ED |
178 | { |
179 | struct sock *sk2; | |
180 | struct hlist_nulls_node *node; | |
ba418fa3 | 181 | kuid_t uid = sock_i_uid(sk); |
30fff923 ED |
182 | int res = 0; |
183 | ||
184 | spin_lock(&hslot2->lock); | |
4243cdc2 | 185 | udp_portaddr_for_each_entry(sk2, node, &hslot2->head) { |
9d4fb27d JP |
186 | if (net_eq(sock_net(sk2), net) && |
187 | sk2 != sk && | |
188 | (udp_sk(sk2)->udp_port_hash == num) && | |
189 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
190 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
191 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 | 192 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
e32ea7e7 | 193 | rcu_access_pointer(sk->sk_reuseport_cb) || |
4243cdc2 | 194 | !uid_eq(uid, sock_i_uid(sk2))) && |
e32ea7e7 | 195 | saddr_comp(sk, sk2, true)) { |
30fff923 ED |
196 | res = 1; |
197 | break; | |
198 | } | |
4243cdc2 | 199 | } |
30fff923 ED |
200 | spin_unlock(&hslot2->lock); |
201 | return res; | |
202 | } | |
203 | ||
e32ea7e7 CG |
204 | static int udp_reuseport_add_sock(struct sock *sk, struct udp_hslot *hslot, |
205 | int (*saddr_same)(const struct sock *sk1, | |
206 | const struct sock *sk2, | |
207 | bool match_wildcard)) | |
208 | { | |
209 | struct net *net = sock_net(sk); | |
210 | struct hlist_nulls_node *node; | |
211 | kuid_t uid = sock_i_uid(sk); | |
212 | struct sock *sk2; | |
213 | ||
214 | sk_nulls_for_each(sk2, node, &hslot->head) { | |
215 | if (net_eq(sock_net(sk2), net) && | |
216 | sk2 != sk && | |
217 | sk2->sk_family == sk->sk_family && | |
218 | ipv6_only_sock(sk2) == ipv6_only_sock(sk) && | |
219 | (udp_sk(sk2)->udp_port_hash == udp_sk(sk)->udp_port_hash) && | |
220 | (sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
221 | sk2->sk_reuseport && uid_eq(uid, sock_i_uid(sk2)) && | |
222 | (*saddr_same)(sk, sk2, false)) { | |
223 | return reuseport_add_sock(sk, sk2); | |
224 | } | |
225 | } | |
226 | ||
227 | /* Initial allocation may have already happened via setsockopt */ | |
228 | if (!rcu_access_pointer(sk->sk_reuseport_cb)) | |
229 | return reuseport_alloc(sk); | |
230 | return 0; | |
231 | } | |
232 | ||
25030a7f | 233 | /** |
6ba5a3c5 | 234 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
235 | * |
236 | * @sk: socket struct in question | |
237 | * @snum: port number to look up | |
df2bc459 | 238 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25985edc | 239 | * @hash2_nulladdr: AF-dependent hash value in secondary hash chains, |
30fff923 | 240 | * with NULL address |
25030a7f | 241 | */ |
6ba5a3c5 | 242 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
4243cdc2 | 243 | int (*saddr_comp)(const struct sock *sk1, |
e32ea7e7 CG |
244 | const struct sock *sk2, |
245 | bool match_wildcard), | |
30fff923 | 246 | unsigned int hash2_nulladdr) |
25030a7f | 247 | { |
512615b6 | 248 | struct udp_hslot *hslot, *hslot2; |
645ca708 | 249 | struct udp_table *udptable = sk->sk_prot->h.udp_table; |
25030a7f | 250 | int error = 1; |
3b1e0a65 | 251 | struct net *net = sock_net(sk); |
1da177e4 | 252 | |
32c1da70 | 253 | if (!snum) { |
9088c560 | 254 | int low, high, remaining; |
95c96174 | 255 | unsigned int rand; |
98322f22 ED |
256 | unsigned short first, last; |
257 | DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN); | |
32c1da70 | 258 | |
0bbf87d8 | 259 | inet_get_local_port_range(net, &low, &high); |
a25de534 | 260 | remaining = (high - low) + 1; |
227b60f5 | 261 | |
63862b5b | 262 | rand = prandom_u32(); |
8fc54f68 | 263 | first = reciprocal_scale(rand, remaining) + low; |
98322f22 ED |
264 | /* |
265 | * force rand to be an odd multiple of UDP_HTABLE_SIZE | |
266 | */ | |
f86dcc5a | 267 | rand = (rand | 1) * (udptable->mask + 1); |
5781b235 ED |
268 | last = first + udptable->mask + 1; |
269 | do { | |
f86dcc5a | 270 | hslot = udp_hashslot(udptable, net, first); |
98322f22 | 271 | bitmap_zero(bitmap, PORTS_PER_CHAIN); |
645ca708 | 272 | spin_lock_bh(&hslot->lock); |
98322f22 | 273 | udp_lib_lport_inuse(net, snum, hslot, bitmap, sk, |
f86dcc5a | 274 | saddr_comp, udptable->log); |
98322f22 ED |
275 | |
276 | snum = first; | |
277 | /* | |
278 | * Iterate on all possible values of snum for this hash. | |
279 | * Using steps of an odd multiple of UDP_HTABLE_SIZE | |
280 | * give us randomization and full range coverage. | |
281 | */ | |
9088c560 | 282 | do { |
98322f22 | 283 | if (low <= snum && snum <= high && |
e3826f1e | 284 | !test_bit(snum >> udptable->log, bitmap) && |
122ff243 | 285 | !inet_is_local_reserved_port(net, snum)) |
98322f22 ED |
286 | goto found; |
287 | snum += rand; | |
288 | } while (snum != first); | |
289 | spin_unlock_bh(&hslot->lock); | |
5781b235 | 290 | } while (++first != last); |
98322f22 | 291 | goto fail; |
645ca708 | 292 | } else { |
f86dcc5a | 293 | hslot = udp_hashslot(udptable, net, snum); |
645ca708 | 294 | spin_lock_bh(&hslot->lock); |
30fff923 ED |
295 | if (hslot->count > 10) { |
296 | int exist; | |
297 | unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum; | |
298 | ||
299 | slot2 &= udptable->mask; | |
300 | hash2_nulladdr &= udptable->mask; | |
301 | ||
302 | hslot2 = udp_hashslot2(udptable, slot2); | |
303 | if (hslot->count < hslot2->count) | |
304 | goto scan_primary_hash; | |
305 | ||
306 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
307 | sk, saddr_comp); | |
308 | if (!exist && (hash2_nulladdr != slot2)) { | |
309 | hslot2 = udp_hashslot2(udptable, hash2_nulladdr); | |
310 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
311 | sk, saddr_comp); | |
312 | } | |
313 | if (exist) | |
314 | goto fail_unlock; | |
315 | else | |
316 | goto found; | |
317 | } | |
318 | scan_primary_hash: | |
f86dcc5a ED |
319 | if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk, |
320 | saddr_comp, 0)) | |
645ca708 ED |
321 | goto fail_unlock; |
322 | } | |
98322f22 | 323 | found: |
c720c7e8 | 324 | inet_sk(sk)->inet_num = snum; |
d4cada4a ED |
325 | udp_sk(sk)->udp_port_hash = snum; |
326 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
1da177e4 | 327 | if (sk_unhashed(sk)) { |
e32ea7e7 CG |
328 | if (sk->sk_reuseport && |
329 | udp_reuseport_add_sock(sk, hslot, saddr_comp)) { | |
330 | inet_sk(sk)->inet_num = 0; | |
331 | udp_sk(sk)->udp_port_hash = 0; | |
332 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
333 | goto fail_unlock; | |
334 | } | |
335 | ||
88ab1932 | 336 | sk_nulls_add_node_rcu(sk, &hslot->head); |
fdcc8aa9 | 337 | hslot->count++; |
c29a0bc4 | 338 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
512615b6 ED |
339 | |
340 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
341 | spin_lock(&hslot2->lock); | |
342 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
343 | &hslot2->head); | |
344 | hslot2->count++; | |
345 | spin_unlock(&hslot2->lock); | |
1da177e4 | 346 | } |
25030a7f | 347 | error = 0; |
645ca708 ED |
348 | fail_unlock: |
349 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 350 | fail: |
25030a7f GR |
351 | return error; |
352 | } | |
c482c568 | 353 | EXPORT_SYMBOL(udp_lib_get_port); |
25030a7f | 354 | |
e32ea7e7 CG |
355 | /* match_wildcard == true: 0.0.0.0 equals to any IPv4 addresses |
356 | * match_wildcard == false: addresses must be exactly the same, i.e. | |
357 | * 0.0.0.0 only equals to 0.0.0.0 | |
358 | */ | |
359 | static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2, | |
360 | bool match_wildcard) | |
db8dac20 DM |
361 | { |
362 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
363 | ||
e32ea7e7 CG |
364 | if (!ipv6_only_sock(sk2)) { |
365 | if (inet1->inet_rcv_saddr == inet2->inet_rcv_saddr) | |
366 | return 1; | |
367 | if (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr) | |
368 | return match_wildcard; | |
369 | } | |
370 | return 0; | |
db8dac20 DM |
371 | } |
372 | ||
6eada011 ED |
373 | static u32 udp4_portaddr_hash(const struct net *net, __be32 saddr, |
374 | unsigned int port) | |
d4cada4a | 375 | { |
0eae88f3 | 376 | return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port; |
d4cada4a ED |
377 | } |
378 | ||
6ba5a3c5 | 379 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 380 | { |
30fff923 | 381 | unsigned int hash2_nulladdr = |
0eae88f3 | 382 | udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum); |
30fff923 ED |
383 | unsigned int hash2_partial = |
384 | udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0); | |
385 | ||
d4cada4a | 386 | /* precompute partial secondary hash */ |
30fff923 ED |
387 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
388 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr); | |
db8dac20 DM |
389 | } |
390 | ||
60c04aec JP |
391 | static inline int compute_score(struct sock *sk, struct net *net, |
392 | __be32 saddr, unsigned short hnum, __be16 sport, | |
393 | __be32 daddr, __be16 dport, int dif) | |
645ca708 | 394 | { |
60c04aec JP |
395 | int score; |
396 | struct inet_sock *inet; | |
645ca708 | 397 | |
60c04aec JP |
398 | if (!net_eq(sock_net(sk), net) || |
399 | udp_sk(sk)->udp_port_hash != hnum || | |
400 | ipv6_only_sock(sk)) | |
401 | return -1; | |
645ca708 | 402 | |
60c04aec JP |
403 | score = (sk->sk_family == PF_INET) ? 2 : 1; |
404 | inet = inet_sk(sk); | |
405 | ||
406 | if (inet->inet_rcv_saddr) { | |
407 | if (inet->inet_rcv_saddr != daddr) | |
408 | return -1; | |
409 | score += 4; | |
645ca708 | 410 | } |
60c04aec JP |
411 | |
412 | if (inet->inet_daddr) { | |
413 | if (inet->inet_daddr != saddr) | |
414 | return -1; | |
415 | score += 4; | |
416 | } | |
417 | ||
418 | if (inet->inet_dport) { | |
419 | if (inet->inet_dport != sport) | |
420 | return -1; | |
421 | score += 4; | |
422 | } | |
423 | ||
424 | if (sk->sk_bound_dev_if) { | |
425 | if (sk->sk_bound_dev_if != dif) | |
426 | return -1; | |
427 | score += 4; | |
428 | } | |
70da268b ED |
429 | if (sk->sk_incoming_cpu == raw_smp_processor_id()) |
430 | score++; | |
645ca708 ED |
431 | return score; |
432 | } | |
433 | ||
5051ebd2 ED |
434 | /* |
435 | * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num) | |
436 | */ | |
5051ebd2 ED |
437 | static inline int compute_score2(struct sock *sk, struct net *net, |
438 | __be32 saddr, __be16 sport, | |
439 | __be32 daddr, unsigned int hnum, int dif) | |
440 | { | |
60c04aec JP |
441 | int score; |
442 | struct inet_sock *inet; | |
5051ebd2 | 443 | |
60c04aec JP |
444 | if (!net_eq(sock_net(sk), net) || |
445 | ipv6_only_sock(sk)) | |
446 | return -1; | |
5051ebd2 | 447 | |
60c04aec JP |
448 | inet = inet_sk(sk); |
449 | ||
450 | if (inet->inet_rcv_saddr != daddr || | |
451 | inet->inet_num != hnum) | |
452 | return -1; | |
453 | ||
454 | score = (sk->sk_family == PF_INET) ? 2 : 1; | |
455 | ||
456 | if (inet->inet_daddr) { | |
457 | if (inet->inet_daddr != saddr) | |
5051ebd2 | 458 | return -1; |
60c04aec JP |
459 | score += 4; |
460 | } | |
461 | ||
462 | if (inet->inet_dport) { | |
463 | if (inet->inet_dport != sport) | |
5051ebd2 | 464 | return -1; |
60c04aec JP |
465 | score += 4; |
466 | } | |
5051ebd2 | 467 | |
60c04aec JP |
468 | if (sk->sk_bound_dev_if) { |
469 | if (sk->sk_bound_dev_if != dif) | |
470 | return -1; | |
471 | score += 4; | |
5051ebd2 | 472 | } |
60c04aec | 473 | |
70da268b ED |
474 | if (sk->sk_incoming_cpu == raw_smp_processor_id()) |
475 | score++; | |
476 | ||
5051ebd2 ED |
477 | return score; |
478 | } | |
479 | ||
6eada011 ED |
480 | static u32 udp_ehashfn(const struct net *net, const __be32 laddr, |
481 | const __u16 lport, const __be32 faddr, | |
482 | const __be16 fport) | |
65cd8033 | 483 | { |
1bbdceef HFS |
484 | static u32 udp_ehash_secret __read_mostly; |
485 | ||
486 | net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret)); | |
487 | ||
65cd8033 | 488 | return __inet_ehashfn(laddr, lport, faddr, fport, |
1bbdceef | 489 | udp_ehash_secret + net_hash_mix(net)); |
65cd8033 HFS |
490 | } |
491 | ||
5051ebd2 ED |
492 | /* called with read_rcu_lock() */ |
493 | static struct sock *udp4_lib_lookup2(struct net *net, | |
494 | __be32 saddr, __be16 sport, | |
495 | __be32 daddr, unsigned int hnum, int dif, | |
496 | struct udp_hslot *hslot2, unsigned int slot2) | |
497 | { | |
498 | struct sock *sk, *result; | |
499 | struct hlist_nulls_node *node; | |
ba418fa3 TH |
500 | int score, badness, matches = 0, reuseport = 0; |
501 | u32 hash = 0; | |
5051ebd2 ED |
502 | |
503 | begin: | |
504 | result = NULL; | |
ba418fa3 | 505 | badness = 0; |
5051ebd2 ED |
506 | udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { |
507 | score = compute_score2(sk, net, saddr, sport, | |
508 | daddr, hnum, dif); | |
509 | if (score > badness) { | |
510 | result = sk; | |
511 | badness = score; | |
ba418fa3 TH |
512 | reuseport = sk->sk_reuseport; |
513 | if (reuseport) { | |
e32ea7e7 | 514 | struct sock *sk2; |
65cd8033 HFS |
515 | hash = udp_ehashfn(net, daddr, hnum, |
516 | saddr, sport); | |
e32ea7e7 CG |
517 | sk2 = reuseport_select_sock(sk, hash); |
518 | if (sk2) { | |
519 | result = sk2; | |
520 | goto found; | |
521 | } | |
ba418fa3 TH |
522 | matches = 1; |
523 | } | |
524 | } else if (score == badness && reuseport) { | |
525 | matches++; | |
8fc54f68 | 526 | if (reciprocal_scale(hash, matches) == 0) |
ba418fa3 TH |
527 | result = sk; |
528 | hash = next_pseudo_random32(hash); | |
5051ebd2 ED |
529 | } |
530 | } | |
531 | /* | |
532 | * if the nulls value we got at the end of this lookup is | |
533 | * not the expected one, we must restart lookup. | |
534 | * We probably met an item that was moved to another chain. | |
535 | */ | |
536 | if (get_nulls_value(node) != slot2) | |
537 | goto begin; | |
5051ebd2 | 538 | if (result) { |
e32ea7e7 | 539 | found: |
c31504dc | 540 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
5051ebd2 ED |
541 | result = NULL; |
542 | else if (unlikely(compute_score2(result, net, saddr, sport, | |
543 | daddr, hnum, dif) < badness)) { | |
544 | sock_put(result); | |
545 | goto begin; | |
546 | } | |
547 | } | |
548 | return result; | |
549 | } | |
550 | ||
db8dac20 DM |
551 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
552 | * harder than this. -DaveM | |
553 | */ | |
fce82338 | 554 | struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, |
db8dac20 | 555 | __be16 sport, __be32 daddr, __be16 dport, |
645ca708 | 556 | int dif, struct udp_table *udptable) |
db8dac20 | 557 | { |
271b72c7 | 558 | struct sock *sk, *result; |
88ab1932 | 559 | struct hlist_nulls_node *node; |
db8dac20 | 560 | unsigned short hnum = ntohs(dport); |
5051ebd2 ED |
561 | unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); |
562 | struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; | |
ba418fa3 TH |
563 | int score, badness, matches = 0, reuseport = 0; |
564 | u32 hash = 0; | |
645ca708 | 565 | |
271b72c7 | 566 | rcu_read_lock(); |
5051ebd2 ED |
567 | if (hslot->count > 10) { |
568 | hash2 = udp4_portaddr_hash(net, daddr, hnum); | |
569 | slot2 = hash2 & udptable->mask; | |
570 | hslot2 = &udptable->hash2[slot2]; | |
571 | if (hslot->count < hslot2->count) | |
572 | goto begin; | |
573 | ||
574 | result = udp4_lib_lookup2(net, saddr, sport, | |
575 | daddr, hnum, dif, | |
576 | hslot2, slot2); | |
577 | if (!result) { | |
0eae88f3 | 578 | hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum); |
5051ebd2 ED |
579 | slot2 = hash2 & udptable->mask; |
580 | hslot2 = &udptable->hash2[slot2]; | |
581 | if (hslot->count < hslot2->count) | |
582 | goto begin; | |
583 | ||
1223c67c | 584 | result = udp4_lib_lookup2(net, saddr, sport, |
0eae88f3 | 585 | htonl(INADDR_ANY), hnum, dif, |
5051ebd2 ED |
586 | hslot2, slot2); |
587 | } | |
588 | rcu_read_unlock(); | |
589 | return result; | |
590 | } | |
271b72c7 ED |
591 | begin: |
592 | result = NULL; | |
ba418fa3 | 593 | badness = 0; |
88ab1932 | 594 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { |
645ca708 ED |
595 | score = compute_score(sk, net, saddr, hnum, sport, |
596 | daddr, dport, dif); | |
597 | if (score > badness) { | |
598 | result = sk; | |
599 | badness = score; | |
ba418fa3 TH |
600 | reuseport = sk->sk_reuseport; |
601 | if (reuseport) { | |
e32ea7e7 | 602 | struct sock *sk2; |
65cd8033 HFS |
603 | hash = udp_ehashfn(net, daddr, hnum, |
604 | saddr, sport); | |
e32ea7e7 CG |
605 | sk2 = reuseport_select_sock(sk, hash); |
606 | if (sk2) { | |
607 | result = sk2; | |
608 | goto found; | |
609 | } | |
ba418fa3 TH |
610 | matches = 1; |
611 | } | |
612 | } else if (score == badness && reuseport) { | |
613 | matches++; | |
8fc54f68 | 614 | if (reciprocal_scale(hash, matches) == 0) |
ba418fa3 TH |
615 | result = sk; |
616 | hash = next_pseudo_random32(hash); | |
db8dac20 DM |
617 | } |
618 | } | |
88ab1932 ED |
619 | /* |
620 | * if the nulls value we got at the end of this lookup is | |
621 | * not the expected one, we must restart lookup. | |
622 | * We probably met an item that was moved to another chain. | |
623 | */ | |
5051ebd2 | 624 | if (get_nulls_value(node) != slot) |
88ab1932 ED |
625 | goto begin; |
626 | ||
271b72c7 | 627 | if (result) { |
e32ea7e7 | 628 | found: |
c31504dc | 629 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
271b72c7 ED |
630 | result = NULL; |
631 | else if (unlikely(compute_score(result, net, saddr, hnum, sport, | |
632 | daddr, dport, dif) < badness)) { | |
633 | sock_put(result); | |
634 | goto begin; | |
635 | } | |
636 | } | |
637 | rcu_read_unlock(); | |
db8dac20 DM |
638 | return result; |
639 | } | |
fce82338 | 640 | EXPORT_SYMBOL_GPL(__udp4_lib_lookup); |
db8dac20 | 641 | |
607c4aaf KK |
642 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
643 | __be16 sport, __be16 dport, | |
645ca708 | 644 | struct udp_table *udptable) |
607c4aaf KK |
645 | { |
646 | const struct iphdr *iph = ip_hdr(skb); | |
647 | ||
8afdd99a ED |
648 | return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport, |
649 | iph->daddr, dport, inet_iif(skb), | |
650 | udptable); | |
607c4aaf KK |
651 | } |
652 | ||
bcd41303 KK |
653 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
654 | __be32 daddr, __be16 dport, int dif) | |
655 | { | |
e32ea7e7 CG |
656 | return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, |
657 | &udp_table); | |
bcd41303 KK |
658 | } |
659 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
660 | ||
421b3885 SB |
661 | static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk, |
662 | __be16 loc_port, __be32 loc_addr, | |
663 | __be16 rmt_port, __be32 rmt_addr, | |
664 | int dif, unsigned short hnum) | |
665 | { | |
666 | struct inet_sock *inet = inet_sk(sk); | |
667 | ||
668 | if (!net_eq(sock_net(sk), net) || | |
669 | udp_sk(sk)->udp_port_hash != hnum || | |
670 | (inet->inet_daddr && inet->inet_daddr != rmt_addr) || | |
671 | (inet->inet_dport != rmt_port && inet->inet_dport) || | |
672 | (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) || | |
673 | ipv6_only_sock(sk) || | |
674 | (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) | |
675 | return false; | |
676 | if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif)) | |
677 | return false; | |
678 | return true; | |
679 | } | |
680 | ||
db8dac20 DM |
681 | /* |
682 | * This routine is called by the ICMP module when it gets some | |
683 | * sort of error condition. If err < 0 then the socket should | |
684 | * be closed and the error returned to the user. If err > 0 | |
685 | * it's just the icmp type << 8 | icmp code. | |
686 | * Header points to the ip header of the error packet. We move | |
687 | * on past this. Then (as it used to claim before adjustment) | |
688 | * header points to the first 8 bytes of the udp header. We need | |
689 | * to find the appropriate port. | |
690 | */ | |
691 | ||
645ca708 | 692 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
693 | { |
694 | struct inet_sock *inet; | |
b71d1d42 | 695 | const struct iphdr *iph = (const struct iphdr *)skb->data; |
c482c568 | 696 | struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2)); |
db8dac20 DM |
697 | const int type = icmp_hdr(skb)->type; |
698 | const int code = icmp_hdr(skb)->code; | |
699 | struct sock *sk; | |
700 | int harderr; | |
701 | int err; | |
fd54d716 | 702 | struct net *net = dev_net(skb->dev); |
db8dac20 | 703 | |
fd54d716 | 704 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
db8dac20 | 705 | iph->saddr, uh->source, skb->dev->ifindex, udptable); |
51456b29 | 706 | if (!sk) { |
dcfc23ca | 707 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
708 | return; /* No socket for error */ |
709 | } | |
710 | ||
711 | err = 0; | |
712 | harderr = 0; | |
713 | inet = inet_sk(sk); | |
714 | ||
715 | switch (type) { | |
716 | default: | |
717 | case ICMP_TIME_EXCEEDED: | |
718 | err = EHOSTUNREACH; | |
719 | break; | |
720 | case ICMP_SOURCE_QUENCH: | |
721 | goto out; | |
722 | case ICMP_PARAMETERPROB: | |
723 | err = EPROTO; | |
724 | harderr = 1; | |
725 | break; | |
726 | case ICMP_DEST_UNREACH: | |
727 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
36393395 | 728 | ipv4_sk_update_pmtu(skb, sk, info); |
db8dac20 DM |
729 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { |
730 | err = EMSGSIZE; | |
731 | harderr = 1; | |
732 | break; | |
733 | } | |
734 | goto out; | |
735 | } | |
736 | err = EHOSTUNREACH; | |
737 | if (code <= NR_ICMP_UNREACH) { | |
738 | harderr = icmp_err_convert[code].fatal; | |
739 | err = icmp_err_convert[code].errno; | |
740 | } | |
741 | break; | |
55be7a9c DM |
742 | case ICMP_REDIRECT: |
743 | ipv4_sk_redirect(skb, sk); | |
1a462d18 | 744 | goto out; |
db8dac20 DM |
745 | } |
746 | ||
747 | /* | |
748 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
749 | * 4.1.3.3. | |
750 | */ | |
751 | if (!inet->recverr) { | |
752 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
753 | goto out; | |
b1faf566 | 754 | } else |
c482c568 | 755 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1)); |
b1faf566 | 756 | |
db8dac20 DM |
757 | sk->sk_err = err; |
758 | sk->sk_error_report(sk); | |
759 | out: | |
760 | sock_put(sk); | |
761 | } | |
762 | ||
763 | void udp_err(struct sk_buff *skb, u32 info) | |
764 | { | |
645ca708 | 765 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
766 | } |
767 | ||
768 | /* | |
769 | * Throw away all pending data and cancel the corking. Socket is locked. | |
770 | */ | |
36d926b9 | 771 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
772 | { |
773 | struct udp_sock *up = udp_sk(sk); | |
774 | ||
775 | if (up->pending) { | |
776 | up->len = 0; | |
777 | up->pending = 0; | |
778 | ip_flush_pending_frames(sk); | |
779 | } | |
780 | } | |
36d926b9 | 781 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
782 | |
783 | /** | |
f6b9664f | 784 | * udp4_hwcsum - handle outgoing HW checksumming |
db8dac20 DM |
785 | * @skb: sk_buff containing the filled-in UDP header |
786 | * (checksum field must be zeroed out) | |
f6b9664f HX |
787 | * @src: source IP address |
788 | * @dst: destination IP address | |
db8dac20 | 789 | */ |
c26bf4a5 | 790 | void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst) |
db8dac20 | 791 | { |
db8dac20 | 792 | struct udphdr *uh = udp_hdr(skb); |
f6b9664f HX |
793 | int offset = skb_transport_offset(skb); |
794 | int len = skb->len - offset; | |
795 | int hlen = len; | |
db8dac20 DM |
796 | __wsum csum = 0; |
797 | ||
ebbe495f | 798 | if (!skb_has_frag_list(skb)) { |
db8dac20 DM |
799 | /* |
800 | * Only one fragment on the socket. | |
801 | */ | |
802 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
803 | skb->csum_offset = offsetof(struct udphdr, check); | |
f6b9664f HX |
804 | uh->check = ~csum_tcpudp_magic(src, dst, len, |
805 | IPPROTO_UDP, 0); | |
db8dac20 | 806 | } else { |
ebbe495f WC |
807 | struct sk_buff *frags; |
808 | ||
db8dac20 DM |
809 | /* |
810 | * HW-checksum won't work as there are two or more | |
811 | * fragments on the socket so that all csums of sk_buffs | |
812 | * should be together | |
813 | */ | |
ebbe495f | 814 | skb_walk_frags(skb, frags) { |
f6b9664f HX |
815 | csum = csum_add(csum, frags->csum); |
816 | hlen -= frags->len; | |
ebbe495f | 817 | } |
db8dac20 | 818 | |
f6b9664f | 819 | csum = skb_checksum(skb, offset, hlen, csum); |
db8dac20 DM |
820 | skb->ip_summed = CHECKSUM_NONE; |
821 | ||
db8dac20 DM |
822 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); |
823 | if (uh->check == 0) | |
824 | uh->check = CSUM_MANGLED_0; | |
825 | } | |
826 | } | |
c26bf4a5 | 827 | EXPORT_SYMBOL_GPL(udp4_hwcsum); |
db8dac20 | 828 | |
af5fcba7 TH |
829 | /* Function to set UDP checksum for an IPv4 UDP packet. This is intended |
830 | * for the simple case like when setting the checksum for a UDP tunnel. | |
831 | */ | |
832 | void udp_set_csum(bool nocheck, struct sk_buff *skb, | |
833 | __be32 saddr, __be32 daddr, int len) | |
834 | { | |
835 | struct udphdr *uh = udp_hdr(skb); | |
836 | ||
837 | if (nocheck) | |
838 | uh->check = 0; | |
839 | else if (skb_is_gso(skb)) | |
840 | uh->check = ~udp_v4_check(len, saddr, daddr, 0); | |
841 | else if (skb_dst(skb) && skb_dst(skb)->dev && | |
c8cd0989 TH |
842 | (skb_dst(skb)->dev->features & |
843 | (NETIF_F_IP_CSUM | NETIF_F_HW_CSUM))) { | |
af5fcba7 TH |
844 | |
845 | BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL); | |
846 | ||
847 | skb->ip_summed = CHECKSUM_PARTIAL; | |
848 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
849 | skb->csum_offset = offsetof(struct udphdr, check); | |
850 | uh->check = ~udp_v4_check(len, saddr, daddr, 0); | |
851 | } else { | |
852 | __wsum csum; | |
853 | ||
854 | BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL); | |
855 | ||
856 | uh->check = 0; | |
857 | csum = skb_checksum(skb, 0, len, 0); | |
858 | uh->check = udp_v4_check(len, saddr, daddr, csum); | |
859 | if (uh->check == 0) | |
860 | uh->check = CSUM_MANGLED_0; | |
861 | ||
862 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
863 | } | |
864 | } | |
865 | EXPORT_SYMBOL(udp_set_csum); | |
866 | ||
79ab0531 | 867 | static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4) |
db8dac20 | 868 | { |
f6b9664f | 869 | struct sock *sk = skb->sk; |
db8dac20 | 870 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
871 | struct udphdr *uh; |
872 | int err = 0; | |
873 | int is_udplite = IS_UDPLITE(sk); | |
f6b9664f HX |
874 | int offset = skb_transport_offset(skb); |
875 | int len = skb->len - offset; | |
db8dac20 DM |
876 | __wsum csum = 0; |
877 | ||
db8dac20 DM |
878 | /* |
879 | * Create a UDP header | |
880 | */ | |
881 | uh = udp_hdr(skb); | |
f6b9664f | 882 | uh->source = inet->inet_sport; |
79ab0531 | 883 | uh->dest = fl4->fl4_dport; |
f6b9664f | 884 | uh->len = htons(len); |
db8dac20 DM |
885 | uh->check = 0; |
886 | ||
887 | if (is_udplite) /* UDP-Lite */ | |
f6b9664f | 888 | csum = udplite_csum(skb); |
db8dac20 | 889 | |
28448b80 | 890 | else if (sk->sk_no_check_tx) { /* UDP csum disabled */ |
db8dac20 DM |
891 | |
892 | skb->ip_summed = CHECKSUM_NONE; | |
893 | goto send; | |
894 | ||
895 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
896 | ||
79ab0531 | 897 | udp4_hwcsum(skb, fl4->saddr, fl4->daddr); |
db8dac20 DM |
898 | goto send; |
899 | ||
f6b9664f HX |
900 | } else |
901 | csum = udp_csum(skb); | |
db8dac20 DM |
902 | |
903 | /* add protocol-dependent pseudo-header */ | |
79ab0531 | 904 | uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len, |
c482c568 | 905 | sk->sk_protocol, csum); |
db8dac20 DM |
906 | if (uh->check == 0) |
907 | uh->check = CSUM_MANGLED_0; | |
908 | ||
909 | send: | |
b5ec8eea | 910 | err = ip_send_skb(sock_net(sk), skb); |
6ce9e7b5 ED |
911 | if (err) { |
912 | if (err == -ENOBUFS && !inet->recverr) { | |
913 | UDP_INC_STATS_USER(sock_net(sk), | |
914 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
915 | err = 0; | |
916 | } | |
917 | } else | |
918 | UDP_INC_STATS_USER(sock_net(sk), | |
919 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
f6b9664f HX |
920 | return err; |
921 | } | |
922 | ||
923 | /* | |
924 | * Push out all pending data as one UDP datagram. Socket is locked. | |
925 | */ | |
8822b64a | 926 | int udp_push_pending_frames(struct sock *sk) |
f6b9664f HX |
927 | { |
928 | struct udp_sock *up = udp_sk(sk); | |
929 | struct inet_sock *inet = inet_sk(sk); | |
b6f21b26 | 930 | struct flowi4 *fl4 = &inet->cork.fl.u.ip4; |
f6b9664f HX |
931 | struct sk_buff *skb; |
932 | int err = 0; | |
933 | ||
77968b78 | 934 | skb = ip_finish_skb(sk, fl4); |
f6b9664f HX |
935 | if (!skb) |
936 | goto out; | |
937 | ||
79ab0531 | 938 | err = udp_send_skb(skb, fl4); |
f6b9664f | 939 | |
db8dac20 DM |
940 | out: |
941 | up->len = 0; | |
942 | up->pending = 0; | |
db8dac20 DM |
943 | return err; |
944 | } | |
8822b64a | 945 | EXPORT_SYMBOL(udp_push_pending_frames); |
db8dac20 | 946 | |
1b784140 | 947 | int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
db8dac20 DM |
948 | { |
949 | struct inet_sock *inet = inet_sk(sk); | |
950 | struct udp_sock *up = udp_sk(sk); | |
e474995f | 951 | struct flowi4 fl4_stack; |
b6f21b26 | 952 | struct flowi4 *fl4; |
db8dac20 DM |
953 | int ulen = len; |
954 | struct ipcm_cookie ipc; | |
955 | struct rtable *rt = NULL; | |
956 | int free = 0; | |
957 | int connected = 0; | |
958 | __be32 daddr, faddr, saddr; | |
959 | __be16 dport; | |
960 | u8 tos; | |
961 | int err, is_udplite = IS_UDPLITE(sk); | |
962 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
963 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
903ab86d | 964 | struct sk_buff *skb; |
f6d8bd05 | 965 | struct ip_options_data opt_copy; |
db8dac20 DM |
966 | |
967 | if (len > 0xFFFF) | |
968 | return -EMSGSIZE; | |
969 | ||
970 | /* | |
971 | * Check the flags. | |
972 | */ | |
973 | ||
c482c568 | 974 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */ |
db8dac20 DM |
975 | return -EOPNOTSUPP; |
976 | ||
977 | ipc.opt = NULL; | |
2244d07b | 978 | ipc.tx_flags = 0; |
aa661581 FF |
979 | ipc.ttl = 0; |
980 | ipc.tos = -1; | |
db8dac20 | 981 | |
903ab86d HX |
982 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
983 | ||
f5fca608 | 984 | fl4 = &inet->cork.fl.u.ip4; |
db8dac20 DM |
985 | if (up->pending) { |
986 | /* | |
987 | * There are pending frames. | |
988 | * The socket lock must be held while it's corked. | |
989 | */ | |
990 | lock_sock(sk); | |
991 | if (likely(up->pending)) { | |
992 | if (unlikely(up->pending != AF_INET)) { | |
993 | release_sock(sk); | |
994 | return -EINVAL; | |
995 | } | |
996 | goto do_append_data; | |
997 | } | |
998 | release_sock(sk); | |
999 | } | |
1000 | ulen += sizeof(struct udphdr); | |
1001 | ||
1002 | /* | |
1003 | * Get and verify the address. | |
1004 | */ | |
1005 | if (msg->msg_name) { | |
342dfc30 | 1006 | DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); |
db8dac20 DM |
1007 | if (msg->msg_namelen < sizeof(*usin)) |
1008 | return -EINVAL; | |
1009 | if (usin->sin_family != AF_INET) { | |
1010 | if (usin->sin_family != AF_UNSPEC) | |
1011 | return -EAFNOSUPPORT; | |
1012 | } | |
1013 | ||
1014 | daddr = usin->sin_addr.s_addr; | |
1015 | dport = usin->sin_port; | |
1016 | if (dport == 0) | |
1017 | return -EINVAL; | |
1018 | } else { | |
1019 | if (sk->sk_state != TCP_ESTABLISHED) | |
1020 | return -EDESTADDRREQ; | |
c720c7e8 ED |
1021 | daddr = inet->inet_daddr; |
1022 | dport = inet->inet_dport; | |
db8dac20 DM |
1023 | /* Open fast path for connected socket. |
1024 | Route will not be used, if at least one option is set. | |
1025 | */ | |
1026 | connected = 1; | |
1027 | } | |
c720c7e8 | 1028 | ipc.addr = inet->inet_saddr; |
db8dac20 DM |
1029 | |
1030 | ipc.oif = sk->sk_bound_dev_if; | |
bf84a010 DB |
1031 | |
1032 | sock_tx_timestamp(sk, &ipc.tx_flags); | |
1033 | ||
db8dac20 | 1034 | if (msg->msg_controllen) { |
c8e6ad08 HFS |
1035 | err = ip_cmsg_send(sock_net(sk), msg, &ipc, |
1036 | sk->sk_family == AF_INET6); | |
db8dac20 DM |
1037 | if (err) |
1038 | return err; | |
1039 | if (ipc.opt) | |
1040 | free = 1; | |
1041 | connected = 0; | |
1042 | } | |
f6d8bd05 ED |
1043 | if (!ipc.opt) { |
1044 | struct ip_options_rcu *inet_opt; | |
1045 | ||
1046 | rcu_read_lock(); | |
1047 | inet_opt = rcu_dereference(inet->inet_opt); | |
1048 | if (inet_opt) { | |
1049 | memcpy(&opt_copy, inet_opt, | |
1050 | sizeof(*inet_opt) + inet_opt->opt.optlen); | |
1051 | ipc.opt = &opt_copy.opt; | |
1052 | } | |
1053 | rcu_read_unlock(); | |
1054 | } | |
db8dac20 DM |
1055 | |
1056 | saddr = ipc.addr; | |
1057 | ipc.addr = faddr = daddr; | |
1058 | ||
f6d8bd05 | 1059 | if (ipc.opt && ipc.opt->opt.srr) { |
db8dac20 DM |
1060 | if (!daddr) |
1061 | return -EINVAL; | |
f6d8bd05 | 1062 | faddr = ipc.opt->opt.faddr; |
db8dac20 DM |
1063 | connected = 0; |
1064 | } | |
aa661581 | 1065 | tos = get_rttos(&ipc, inet); |
db8dac20 DM |
1066 | if (sock_flag(sk, SOCK_LOCALROUTE) || |
1067 | (msg->msg_flags & MSG_DONTROUTE) || | |
f6d8bd05 | 1068 | (ipc.opt && ipc.opt->opt.is_strictroute)) { |
db8dac20 DM |
1069 | tos |= RTO_ONLINK; |
1070 | connected = 0; | |
1071 | } | |
1072 | ||
1073 | if (ipv4_is_multicast(daddr)) { | |
1074 | if (!ipc.oif) | |
1075 | ipc.oif = inet->mc_index; | |
1076 | if (!saddr) | |
1077 | saddr = inet->mc_addr; | |
1078 | connected = 0; | |
76e21053 EH |
1079 | } else if (!ipc.oif) |
1080 | ipc.oif = inet->uc_index; | |
db8dac20 DM |
1081 | |
1082 | if (connected) | |
c482c568 | 1083 | rt = (struct rtable *)sk_dst_check(sk, 0); |
db8dac20 | 1084 | |
51456b29 | 1085 | if (!rt) { |
84a3aa00 | 1086 | struct net *net = sock_net(sk); |
9a24abfa | 1087 | __u8 flow_flags = inet_sk_flowi_flags(sk); |
84a3aa00 | 1088 | |
e474995f | 1089 | fl4 = &fl4_stack; |
9a24abfa | 1090 | |
e474995f | 1091 | flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos, |
c0951cbc | 1092 | RT_SCOPE_UNIVERSE, sk->sk_protocol, |
9a24abfa | 1093 | flow_flags, |
c0951cbc DM |
1094 | faddr, saddr, dport, inet->inet_sport); |
1095 | ||
8cbb512c DA |
1096 | if (!saddr && ipc.oif) |
1097 | l3mdev_get_saddr(net, ipc.oif, fl4); | |
1098 | ||
e474995f DM |
1099 | security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); |
1100 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe DM |
1101 | if (IS_ERR(rt)) { |
1102 | err = PTR_ERR(rt); | |
06dc94b1 | 1103 | rt = NULL; |
db8dac20 | 1104 | if (err == -ENETUNREACH) |
f1d8cba6 | 1105 | IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
1106 | goto out; |
1107 | } | |
1108 | ||
1109 | err = -EACCES; | |
1110 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
1111 | !sock_flag(sk, SOCK_BROADCAST)) | |
1112 | goto out; | |
1113 | if (connected) | |
d8d1f30b | 1114 | sk_dst_set(sk, dst_clone(&rt->dst)); |
db8dac20 DM |
1115 | } |
1116 | ||
1117 | if (msg->msg_flags&MSG_CONFIRM) | |
1118 | goto do_confirm; | |
1119 | back_from_confirm: | |
1120 | ||
e474995f | 1121 | saddr = fl4->saddr; |
db8dac20 | 1122 | if (!ipc.addr) |
e474995f | 1123 | daddr = ipc.addr = fl4->daddr; |
db8dac20 | 1124 | |
903ab86d HX |
1125 | /* Lockless fast path for the non-corking case. */ |
1126 | if (!corkreq) { | |
f69e6d13 | 1127 | skb = ip_make_skb(sk, fl4, getfrag, msg, ulen, |
903ab86d HX |
1128 | sizeof(struct udphdr), &ipc, &rt, |
1129 | msg->msg_flags); | |
1130 | err = PTR_ERR(skb); | |
50c3a487 | 1131 | if (!IS_ERR_OR_NULL(skb)) |
79ab0531 | 1132 | err = udp_send_skb(skb, fl4); |
903ab86d HX |
1133 | goto out; |
1134 | } | |
1135 | ||
db8dac20 DM |
1136 | lock_sock(sk); |
1137 | if (unlikely(up->pending)) { | |
1138 | /* The socket is already corked while preparing it. */ | |
1139 | /* ... which is an evident application bug. --ANK */ | |
1140 | release_sock(sk); | |
1141 | ||
ba7a46f1 | 1142 | net_dbg_ratelimited("cork app bug 2\n"); |
db8dac20 DM |
1143 | err = -EINVAL; |
1144 | goto out; | |
1145 | } | |
1146 | /* | |
1147 | * Now cork the socket to pend data. | |
1148 | */ | |
b6f21b26 DM |
1149 | fl4 = &inet->cork.fl.u.ip4; |
1150 | fl4->daddr = daddr; | |
1151 | fl4->saddr = saddr; | |
9cce96df DM |
1152 | fl4->fl4_dport = dport; |
1153 | fl4->fl4_sport = inet->inet_sport; | |
db8dac20 DM |
1154 | up->pending = AF_INET; |
1155 | ||
1156 | do_append_data: | |
1157 | up->len += ulen; | |
f69e6d13 | 1158 | err = ip_append_data(sk, fl4, getfrag, msg, ulen, |
f5fca608 DM |
1159 | sizeof(struct udphdr), &ipc, &rt, |
1160 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | |
db8dac20 DM |
1161 | if (err) |
1162 | udp_flush_pending_frames(sk); | |
1163 | else if (!corkreq) | |
1164 | err = udp_push_pending_frames(sk); | |
1165 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
1166 | up->pending = 0; | |
1167 | release_sock(sk); | |
1168 | ||
1169 | out: | |
1170 | ip_rt_put(rt); | |
1171 | if (free) | |
1172 | kfree(ipc.opt); | |
1173 | if (!err) | |
1174 | return len; | |
1175 | /* | |
1176 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1177 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1178 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1179 | * things). We could add another new stat but at least for now that | |
1180 | * seems like overkill. | |
1181 | */ | |
1182 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
1183 | UDP_INC_STATS_USER(sock_net(sk), |
1184 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
1185 | } |
1186 | return err; | |
1187 | ||
1188 | do_confirm: | |
d8d1f30b | 1189 | dst_confirm(&rt->dst); |
db8dac20 DM |
1190 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1191 | goto back_from_confirm; | |
1192 | err = 0; | |
1193 | goto out; | |
1194 | } | |
c482c568 | 1195 | EXPORT_SYMBOL(udp_sendmsg); |
db8dac20 DM |
1196 | |
1197 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
1198 | size_t size, int flags) | |
1199 | { | |
f5fca608 | 1200 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
1201 | struct udp_sock *up = udp_sk(sk); |
1202 | int ret; | |
1203 | ||
d3f7d56a SL |
1204 | if (flags & MSG_SENDPAGE_NOTLAST) |
1205 | flags |= MSG_MORE; | |
1206 | ||
db8dac20 DM |
1207 | if (!up->pending) { |
1208 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
1209 | ||
1210 | /* Call udp_sendmsg to specify destination address which | |
1211 | * sendpage interface can't pass. | |
1212 | * This will succeed only when the socket is connected. | |
1213 | */ | |
1b784140 | 1214 | ret = udp_sendmsg(sk, &msg, 0); |
db8dac20 DM |
1215 | if (ret < 0) |
1216 | return ret; | |
1217 | } | |
1218 | ||
1219 | lock_sock(sk); | |
1220 | ||
1221 | if (unlikely(!up->pending)) { | |
1222 | release_sock(sk); | |
1223 | ||
ba7a46f1 | 1224 | net_dbg_ratelimited("udp cork app bug 3\n"); |
db8dac20 DM |
1225 | return -EINVAL; |
1226 | } | |
1227 | ||
f5fca608 DM |
1228 | ret = ip_append_page(sk, &inet->cork.fl.u.ip4, |
1229 | page, offset, size, flags); | |
db8dac20 DM |
1230 | if (ret == -EOPNOTSUPP) { |
1231 | release_sock(sk); | |
1232 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
1233 | size, flags); | |
1234 | } | |
1235 | if (ret < 0) { | |
1236 | udp_flush_pending_frames(sk); | |
1237 | goto out; | |
1238 | } | |
1239 | ||
1240 | up->len += size; | |
1241 | if (!(up->corkflag || (flags&MSG_MORE))) | |
1242 | ret = udp_push_pending_frames(sk); | |
1243 | if (!ret) | |
1244 | ret = size; | |
1245 | out: | |
1246 | release_sock(sk); | |
1247 | return ret; | |
1248 | } | |
1249 | ||
85584672 ED |
1250 | /** |
1251 | * first_packet_length - return length of first packet in receive queue | |
1252 | * @sk: socket | |
1253 | * | |
1254 | * Drops all bad checksum frames, until a valid one is found. | |
1255 | * Returns the length of found skb, or 0 if none is found. | |
1256 | */ | |
1257 | static unsigned int first_packet_length(struct sock *sk) | |
1258 | { | |
1259 | struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue; | |
1260 | struct sk_buff *skb; | |
1261 | unsigned int res; | |
1262 | ||
1263 | __skb_queue_head_init(&list_kill); | |
1264 | ||
1265 | spin_lock_bh(&rcvq->lock); | |
1266 | while ((skb = skb_peek(rcvq)) != NULL && | |
1267 | udp_lib_checksum_complete(skb)) { | |
6a5dc9e5 ED |
1268 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, |
1269 | IS_UDPLITE(sk)); | |
85584672 ED |
1270 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, |
1271 | IS_UDPLITE(sk)); | |
8edf19c2 | 1272 | atomic_inc(&sk->sk_drops); |
85584672 ED |
1273 | __skb_unlink(skb, rcvq); |
1274 | __skb_queue_tail(&list_kill, skb); | |
1275 | } | |
1276 | res = skb ? skb->len : 0; | |
1277 | spin_unlock_bh(&rcvq->lock); | |
1278 | ||
1279 | if (!skb_queue_empty(&list_kill)) { | |
8a74ad60 ED |
1280 | bool slow = lock_sock_fast(sk); |
1281 | ||
85584672 ED |
1282 | __skb_queue_purge(&list_kill); |
1283 | sk_mem_reclaim_partial(sk); | |
8a74ad60 | 1284 | unlock_sock_fast(sk, slow); |
85584672 ED |
1285 | } |
1286 | return res; | |
1287 | } | |
1288 | ||
1da177e4 LT |
1289 | /* |
1290 | * IOCTL requests applicable to the UDP protocol | |
1291 | */ | |
e905a9ed | 1292 | |
1da177e4 LT |
1293 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
1294 | { | |
6516c655 SH |
1295 | switch (cmd) { |
1296 | case SIOCOUTQ: | |
1da177e4 | 1297 | { |
31e6d363 ED |
1298 | int amount = sk_wmem_alloc_get(sk); |
1299 | ||
6516c655 SH |
1300 | return put_user(amount, (int __user *)arg); |
1301 | } | |
1da177e4 | 1302 | |
6516c655 SH |
1303 | case SIOCINQ: |
1304 | { | |
85584672 | 1305 | unsigned int amount = first_packet_length(sk); |
6516c655 | 1306 | |
85584672 | 1307 | if (amount) |
6516c655 SH |
1308 | /* |
1309 | * We will only return the amount | |
1310 | * of this packet since that is all | |
1311 | * that will be read. | |
1312 | */ | |
85584672 ED |
1313 | amount -= sizeof(struct udphdr); |
1314 | ||
6516c655 SH |
1315 | return put_user(amount, (int __user *)arg); |
1316 | } | |
1da177e4 | 1317 | |
6516c655 SH |
1318 | default: |
1319 | return -ENOIOCTLCMD; | |
1da177e4 | 1320 | } |
6516c655 SH |
1321 | |
1322 | return 0; | |
1da177e4 | 1323 | } |
c482c568 | 1324 | EXPORT_SYMBOL(udp_ioctl); |
1da177e4 | 1325 | |
db8dac20 DM |
1326 | /* |
1327 | * This should be easy, if there is something there we | |
1328 | * return it, otherwise we block. | |
1329 | */ | |
1330 | ||
1b784140 YX |
1331 | int udp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock, |
1332 | int flags, int *addr_len) | |
db8dac20 DM |
1333 | { |
1334 | struct inet_sock *inet = inet_sk(sk); | |
342dfc30 | 1335 | DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); |
db8dac20 | 1336 | struct sk_buff *skb; |
59c2cdae | 1337 | unsigned int ulen, copied; |
3f518bf7 | 1338 | int peeked, off = 0; |
db8dac20 DM |
1339 | int err; |
1340 | int is_udplite = IS_UDPLITE(sk); | |
197c949e | 1341 | bool checksum_valid = false; |
8a74ad60 | 1342 | bool slow; |
db8dac20 | 1343 | |
db8dac20 | 1344 | if (flags & MSG_ERRQUEUE) |
85fbaa75 | 1345 | return ip_recv_error(sk, msg, len, addr_len); |
db8dac20 DM |
1346 | |
1347 | try_again: | |
1348 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), | |
3f518bf7 | 1349 | &peeked, &off, &err); |
db8dac20 DM |
1350 | if (!skb) |
1351 | goto out; | |
1352 | ||
1353 | ulen = skb->len - sizeof(struct udphdr); | |
59c2cdae DM |
1354 | copied = len; |
1355 | if (copied > ulen) | |
1356 | copied = ulen; | |
1357 | else if (copied < ulen) | |
db8dac20 DM |
1358 | msg->msg_flags |= MSG_TRUNC; |
1359 | ||
1360 | /* | |
1361 | * If checksum is needed at all, try to do it while copying the | |
1362 | * data. If the data is truncated, or if we only want a partial | |
1363 | * coverage checksum (UDP-Lite), do it before the copy. | |
1364 | */ | |
1365 | ||
59c2cdae | 1366 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { |
197c949e ED |
1367 | checksum_valid = !udp_lib_checksum_complete(skb); |
1368 | if (!checksum_valid) | |
db8dac20 DM |
1369 | goto csum_copy_err; |
1370 | } | |
1371 | ||
197c949e | 1372 | if (checksum_valid || skb_csum_unnecessary(skb)) |
51f3d02b DM |
1373 | err = skb_copy_datagram_msg(skb, sizeof(struct udphdr), |
1374 | msg, copied); | |
db8dac20 | 1375 | else { |
227158db AV |
1376 | err = skb_copy_and_csum_datagram_msg(skb, sizeof(struct udphdr), |
1377 | msg); | |
db8dac20 DM |
1378 | |
1379 | if (err == -EINVAL) | |
1380 | goto csum_copy_err; | |
1381 | } | |
1382 | ||
22911fc5 ED |
1383 | if (unlikely(err)) { |
1384 | trace_kfree_skb(skb, udp_recvmsg); | |
979402b1 ED |
1385 | if (!peeked) { |
1386 | atomic_inc(&sk->sk_drops); | |
1387 | UDP_INC_STATS_USER(sock_net(sk), | |
1388 | UDP_MIB_INERRORS, is_udplite); | |
1389 | } | |
db8dac20 | 1390 | goto out_free; |
22911fc5 | 1391 | } |
db8dac20 DM |
1392 | |
1393 | if (!peeked) | |
629ca23c PE |
1394 | UDP_INC_STATS_USER(sock_net(sk), |
1395 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 | 1396 | |
3b885787 | 1397 | sock_recv_ts_and_drops(msg, sk, skb); |
db8dac20 DM |
1398 | |
1399 | /* Copy the address. */ | |
c482c568 | 1400 | if (sin) { |
db8dac20 DM |
1401 | sin->sin_family = AF_INET; |
1402 | sin->sin_port = udp_hdr(skb)->source; | |
1403 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
1404 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
bceaa902 | 1405 | *addr_len = sizeof(*sin); |
db8dac20 DM |
1406 | } |
1407 | if (inet->cmsg_flags) | |
ad6f939a | 1408 | ip_cmsg_recv_offset(msg, skb, sizeof(struct udphdr)); |
db8dac20 | 1409 | |
59c2cdae | 1410 | err = copied; |
db8dac20 DM |
1411 | if (flags & MSG_TRUNC) |
1412 | err = ulen; | |
1413 | ||
1414 | out_free: | |
9d410c79 | 1415 | skb_free_datagram_locked(sk, skb); |
db8dac20 DM |
1416 | out: |
1417 | return err; | |
1418 | ||
1419 | csum_copy_err: | |
8a74ad60 | 1420 | slow = lock_sock_fast(sk); |
6a5dc9e5 ED |
1421 | if (!skb_kill_datagram(sk, skb, flags)) { |
1422 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); | |
629ca23c | 1423 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
6a5dc9e5 | 1424 | } |
8a74ad60 | 1425 | unlock_sock_fast(sk, slow); |
db8dac20 | 1426 | |
beb39db5 ED |
1427 | /* starting over for a new packet, but check if we need to yield */ |
1428 | cond_resched(); | |
9cfaa8de | 1429 | msg->msg_flags &= ~MSG_TRUNC; |
db8dac20 DM |
1430 | goto try_again; |
1431 | } | |
1432 | ||
1da177e4 LT |
1433 | int udp_disconnect(struct sock *sk, int flags) |
1434 | { | |
1435 | struct inet_sock *inet = inet_sk(sk); | |
1436 | /* | |
1437 | * 1003.1g - break association. | |
1438 | */ | |
e905a9ed | 1439 | |
1da177e4 | 1440 | sk->sk_state = TCP_CLOSE; |
c720c7e8 ED |
1441 | inet->inet_daddr = 0; |
1442 | inet->inet_dport = 0; | |
bdeab991 | 1443 | sock_rps_reset_rxhash(sk); |
1da177e4 LT |
1444 | sk->sk_bound_dev_if = 0; |
1445 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
1446 | inet_reset_saddr(sk); | |
1447 | ||
1448 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
1449 | sk->sk_prot->unhash(sk); | |
c720c7e8 | 1450 | inet->inet_sport = 0; |
1da177e4 LT |
1451 | } |
1452 | sk_dst_reset(sk); | |
1453 | return 0; | |
1454 | } | |
c482c568 | 1455 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1456 | |
645ca708 ED |
1457 | void udp_lib_unhash(struct sock *sk) |
1458 | { | |
723b4610 ED |
1459 | if (sk_hashed(sk)) { |
1460 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
512615b6 ED |
1461 | struct udp_hslot *hslot, *hslot2; |
1462 | ||
1463 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1464 | udp_sk(sk)->udp_port_hash); | |
1465 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
645ca708 | 1466 | |
723b4610 | 1467 | spin_lock_bh(&hslot->lock); |
e32ea7e7 CG |
1468 | if (rcu_access_pointer(sk->sk_reuseport_cb)) |
1469 | reuseport_detach_sock(sk); | |
723b4610 | 1470 | if (sk_nulls_del_node_init_rcu(sk)) { |
fdcc8aa9 | 1471 | hslot->count--; |
c720c7e8 | 1472 | inet_sk(sk)->inet_num = 0; |
723b4610 | 1473 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
512615b6 ED |
1474 | |
1475 | spin_lock(&hslot2->lock); | |
1476 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1477 | hslot2->count--; | |
1478 | spin_unlock(&hslot2->lock); | |
723b4610 ED |
1479 | } |
1480 | spin_unlock_bh(&hslot->lock); | |
645ca708 | 1481 | } |
645ca708 ED |
1482 | } |
1483 | EXPORT_SYMBOL(udp_lib_unhash); | |
1484 | ||
719f8358 ED |
1485 | /* |
1486 | * inet_rcv_saddr was changed, we must rehash secondary hash | |
1487 | */ | |
1488 | void udp_lib_rehash(struct sock *sk, u16 newhash) | |
1489 | { | |
1490 | if (sk_hashed(sk)) { | |
1491 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
1492 | struct udp_hslot *hslot, *hslot2, *nhslot2; | |
1493 | ||
1494 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
1495 | nhslot2 = udp_hashslot2(udptable, newhash); | |
1496 | udp_sk(sk)->udp_portaddr_hash = newhash; | |
e32ea7e7 CG |
1497 | |
1498 | if (hslot2 != nhslot2 || | |
1499 | rcu_access_pointer(sk->sk_reuseport_cb)) { | |
719f8358 ED |
1500 | hslot = udp_hashslot(udptable, sock_net(sk), |
1501 | udp_sk(sk)->udp_port_hash); | |
1502 | /* we must lock primary chain too */ | |
1503 | spin_lock_bh(&hslot->lock); | |
e32ea7e7 CG |
1504 | if (rcu_access_pointer(sk->sk_reuseport_cb)) |
1505 | reuseport_detach_sock(sk); | |
1506 | ||
1507 | if (hslot2 != nhslot2) { | |
1508 | spin_lock(&hslot2->lock); | |
1509 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1510 | hslot2->count--; | |
1511 | spin_unlock(&hslot2->lock); | |
1512 | ||
1513 | spin_lock(&nhslot2->lock); | |
1514 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
1515 | &nhslot2->head); | |
1516 | nhslot2->count++; | |
1517 | spin_unlock(&nhslot2->lock); | |
1518 | } | |
719f8358 ED |
1519 | |
1520 | spin_unlock_bh(&hslot->lock); | |
1521 | } | |
1522 | } | |
1523 | } | |
1524 | EXPORT_SYMBOL(udp_lib_rehash); | |
1525 | ||
1526 | static void udp_v4_rehash(struct sock *sk) | |
1527 | { | |
1528 | u16 new_hash = udp4_portaddr_hash(sock_net(sk), | |
1529 | inet_sk(sk)->inet_rcv_saddr, | |
1530 | inet_sk(sk)->inet_num); | |
1531 | udp_lib_rehash(sk, new_hash); | |
1532 | } | |
1533 | ||
93821778 HX |
1534 | static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
1535 | { | |
fec5e652 | 1536 | int rc; |
766e9037 | 1537 | |
005ec974 | 1538 | if (inet_sk(sk)->inet_daddr) { |
bdeab991 | 1539 | sock_rps_save_rxhash(sk, skb); |
005ec974 | 1540 | sk_mark_napi_id(sk, skb); |
2c8c56e1 | 1541 | sk_incoming_cpu_update(sk); |
005ec974 | 1542 | } |
fec5e652 | 1543 | |
d826eb14 | 1544 | rc = sock_queue_rcv_skb(sk, skb); |
766e9037 ED |
1545 | if (rc < 0) { |
1546 | int is_udplite = IS_UDPLITE(sk); | |
93821778 | 1547 | |
93821778 | 1548 | /* Note that an ENOMEM error is charged twice */ |
766e9037 | 1549 | if (rc == -ENOMEM) |
93821778 HX |
1550 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1551 | is_udplite); | |
766e9037 ED |
1552 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
1553 | kfree_skb(skb); | |
296f7ea7 | 1554 | trace_udp_fail_queue_rcv_skb(rc, sk); |
766e9037 | 1555 | return -1; |
93821778 HX |
1556 | } |
1557 | ||
1558 | return 0; | |
1559 | ||
93821778 HX |
1560 | } |
1561 | ||
447167bf ED |
1562 | static struct static_key udp_encap_needed __read_mostly; |
1563 | void udp_encap_enable(void) | |
1564 | { | |
1565 | if (!static_key_enabled(&udp_encap_needed)) | |
1566 | static_key_slow_inc(&udp_encap_needed); | |
1567 | } | |
1568 | EXPORT_SYMBOL(udp_encap_enable); | |
1569 | ||
db8dac20 DM |
1570 | /* returns: |
1571 | * -1: error | |
1572 | * 0: success | |
1573 | * >0: "udp encap" protocol resubmission | |
1574 | * | |
1575 | * Note that in the success and error cases, the skb is assumed to | |
1576 | * have either been requeued or freed. | |
1577 | */ | |
c482c568 | 1578 | int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
db8dac20 DM |
1579 | { |
1580 | struct udp_sock *up = udp_sk(sk); | |
1581 | int rc; | |
1582 | int is_udplite = IS_UDPLITE(sk); | |
1583 | ||
1584 | /* | |
1585 | * Charge it to the socket, dropping if the queue is full. | |
1586 | */ | |
1587 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1588 | goto drop; | |
1589 | nf_reset(skb); | |
1590 | ||
447167bf | 1591 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
0ad92ad0 ED |
1592 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
1593 | ||
db8dac20 DM |
1594 | /* |
1595 | * This is an encapsulation socket so pass the skb to | |
1596 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1597 | * fall through and pass this up the UDP socket. | |
1598 | * up->encap_rcv() returns the following value: | |
1599 | * =0 if skb was successfully passed to the encap | |
1600 | * handler or was discarded by it. | |
1601 | * >0 if skb should be passed on to UDP. | |
1602 | * <0 if skb should be resubmitted as proto -N | |
1603 | */ | |
1604 | ||
1605 | /* if we're overly short, let UDP handle it */ | |
0ad92ad0 | 1606 | encap_rcv = ACCESS_ONCE(up->encap_rcv); |
00db4124 | 1607 | if (skb->len > sizeof(struct udphdr) && encap_rcv) { |
db8dac20 DM |
1608 | int ret; |
1609 | ||
0a80966b TH |
1610 | /* Verify checksum before giving to encap */ |
1611 | if (udp_lib_checksum_complete(skb)) | |
1612 | goto csum_error; | |
1613 | ||
0ad92ad0 | 1614 | ret = encap_rcv(sk, skb); |
db8dac20 | 1615 | if (ret <= 0) { |
0283328e PE |
1616 | UDP_INC_STATS_BH(sock_net(sk), |
1617 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
1618 | is_udplite); |
1619 | return -ret; | |
1620 | } | |
1621 | } | |
1622 | ||
1623 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1624 | } | |
1625 | ||
1626 | /* | |
1627 | * UDP-Lite specific tests, ignored on UDP sockets | |
1628 | */ | |
1629 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1630 | ||
1631 | /* | |
1632 | * MIB statistics other than incrementing the error count are | |
1633 | * disabled for the following two types of errors: these depend | |
1634 | * on the application settings, not on the functioning of the | |
1635 | * protocol stack as such. | |
1636 | * | |
1637 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1638 | * way ... to ... at least let the receiving application block | |
1639 | * delivery of packets with coverage values less than a value | |
1640 | * provided by the application." | |
1641 | */ | |
1642 | if (up->pcrlen == 0) { /* full coverage was set */ | |
ba7a46f1 JP |
1643 | net_dbg_ratelimited("UDPLite: partial coverage %d while full coverage %d requested\n", |
1644 | UDP_SKB_CB(skb)->cscov, skb->len); | |
db8dac20 DM |
1645 | goto drop; |
1646 | } | |
1647 | /* The next case involves violating the min. coverage requested | |
1648 | * by the receiver. This is subtle: if receiver wants x and x is | |
1649 | * greater than the buffersize/MTU then receiver will complain | |
1650 | * that it wants x while sender emits packets of smaller size y. | |
1651 | * Therefore the above ...()->partial_cov statement is essential. | |
1652 | */ | |
1653 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
ba7a46f1 JP |
1654 | net_dbg_ratelimited("UDPLite: coverage %d too small, need min %d\n", |
1655 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
db8dac20 DM |
1656 | goto drop; |
1657 | } | |
1658 | } | |
1659 | ||
33d480ce ED |
1660 | if (rcu_access_pointer(sk->sk_filter) && |
1661 | udp_lib_checksum_complete(skb)) | |
6a5dc9e5 | 1662 | goto csum_error; |
db8dac20 | 1663 | |
274f482d | 1664 | if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) { |
3e215c8d JL |
1665 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1666 | is_udplite); | |
c377411f | 1667 | goto drop; |
3e215c8d | 1668 | } |
c377411f | 1669 | |
93821778 | 1670 | rc = 0; |
db8dac20 | 1671 | |
fbf8866d | 1672 | ipv4_pktinfo_prepare(sk, skb); |
93821778 HX |
1673 | bh_lock_sock(sk); |
1674 | if (!sock_owned_by_user(sk)) | |
1675 | rc = __udp_queue_rcv_skb(sk, skb); | |
f545a38f | 1676 | else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) { |
55349790 ZY |
1677 | bh_unlock_sock(sk); |
1678 | goto drop; | |
1679 | } | |
93821778 HX |
1680 | bh_unlock_sock(sk); |
1681 | ||
1682 | return rc; | |
db8dac20 | 1683 | |
6a5dc9e5 ED |
1684 | csum_error: |
1685 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); | |
db8dac20 | 1686 | drop: |
0283328e | 1687 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8edf19c2 | 1688 | atomic_inc(&sk->sk_drops); |
db8dac20 DM |
1689 | kfree_skb(skb); |
1690 | return -1; | |
1691 | } | |
1692 | ||
1240d137 ED |
1693 | static void flush_stack(struct sock **stack, unsigned int count, |
1694 | struct sk_buff *skb, unsigned int final) | |
1695 | { | |
1696 | unsigned int i; | |
1697 | struct sk_buff *skb1 = NULL; | |
f6b8f32c | 1698 | struct sock *sk; |
1240d137 ED |
1699 | |
1700 | for (i = 0; i < count; i++) { | |
f6b8f32c | 1701 | sk = stack[i]; |
51456b29 | 1702 | if (likely(!skb1)) |
1240d137 ED |
1703 | skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC); |
1704 | ||
f6b8f32c ED |
1705 | if (!skb1) { |
1706 | atomic_inc(&sk->sk_drops); | |
1707 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, | |
1708 | IS_UDPLITE(sk)); | |
1709 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, | |
1710 | IS_UDPLITE(sk)); | |
1711 | } | |
1712 | ||
1713 | if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0) | |
1240d137 | 1714 | skb1 = NULL; |
2dc41cff DH |
1715 | |
1716 | sock_put(sk); | |
1240d137 ED |
1717 | } |
1718 | if (unlikely(skb1)) | |
1719 | kfree_skb(skb1); | |
1720 | } | |
1721 | ||
97502231 | 1722 | /* For TCP sockets, sk_rx_dst is protected by socket lock |
e47eb5df | 1723 | * For UDP, we use xchg() to guard against concurrent changes. |
97502231 ED |
1724 | */ |
1725 | static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst) | |
421b3885 | 1726 | { |
97502231 ED |
1727 | struct dst_entry *old; |
1728 | ||
e47eb5df ED |
1729 | dst_hold(dst); |
1730 | old = xchg(&sk->sk_rx_dst, dst); | |
1731 | dst_release(old); | |
421b3885 SB |
1732 | } |
1733 | ||
db8dac20 DM |
1734 | /* |
1735 | * Multicasts and broadcasts go to each listener. | |
1736 | * | |
1240d137 | 1737 | * Note: called only from the BH handler context. |
db8dac20 | 1738 | */ |
e3163493 | 1739 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1740 | struct udphdr *uh, |
1741 | __be32 saddr, __be32 daddr, | |
36cbb245 RJ |
1742 | struct udp_table *udptable, |
1743 | int proto) | |
db8dac20 | 1744 | { |
1240d137 | 1745 | struct sock *sk, *stack[256 / sizeof(struct sock *)]; |
5cf3d461 DH |
1746 | struct hlist_nulls_node *node; |
1747 | unsigned short hnum = ntohs(uh->dest); | |
1748 | struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum); | |
1749 | int dif = skb->dev->ifindex; | |
2dc41cff DH |
1750 | unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node); |
1751 | unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10); | |
36cbb245 | 1752 | bool inner_flushed = false; |
2dc41cff DH |
1753 | |
1754 | if (use_hash2) { | |
1755 | hash2_any = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum) & | |
1756 | udp_table.mask; | |
1757 | hash2 = udp4_portaddr_hash(net, daddr, hnum) & udp_table.mask; | |
1758 | start_lookup: | |
1759 | hslot = &udp_table.hash2[hash2]; | |
1760 | offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node); | |
1761 | } | |
db8dac20 | 1762 | |
645ca708 | 1763 | spin_lock(&hslot->lock); |
2dc41cff | 1764 | sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) { |
5cf3d461 DH |
1765 | if (__udp_is_mcast_sock(net, sk, |
1766 | uh->dest, daddr, | |
1767 | uh->source, saddr, | |
1768 | dif, hnum)) { | |
1769 | if (unlikely(count == ARRAY_SIZE(stack))) { | |
1770 | flush_stack(stack, count, skb, ~0); | |
36cbb245 | 1771 | inner_flushed = true; |
5cf3d461 DH |
1772 | count = 0; |
1773 | } | |
1774 | stack[count++] = sk; | |
2dc41cff | 1775 | sock_hold(sk); |
1240d137 ED |
1776 | } |
1777 | } | |
1240d137 | 1778 | |
645ca708 | 1779 | spin_unlock(&hslot->lock); |
1240d137 | 1780 | |
2dc41cff DH |
1781 | /* Also lookup *:port if we are using hash2 and haven't done so yet. */ |
1782 | if (use_hash2 && hash2 != hash2_any) { | |
1783 | hash2 = hash2_any; | |
1784 | goto start_lookup; | |
1785 | } | |
1786 | ||
1240d137 ED |
1787 | /* |
1788 | * do the slow work with no lock held | |
1789 | */ | |
1790 | if (count) { | |
1791 | flush_stack(stack, count, skb, count - 1); | |
1240d137 | 1792 | } else { |
36cbb245 RJ |
1793 | if (!inner_flushed) |
1794 | UDP_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI, | |
1795 | proto == IPPROTO_UDPLITE); | |
1796 | consume_skb(skb); | |
1240d137 | 1797 | } |
db8dac20 DM |
1798 | return 0; |
1799 | } | |
1800 | ||
1801 | /* Initialize UDP checksum. If exited with zero value (success), | |
1802 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1803 | * Otherwise, csum completion requires chacksumming packet body, | |
1804 | * including udp header and folding it to skb->csum. | |
1805 | */ | |
1806 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1807 | int proto) | |
1808 | { | |
db8dac20 DM |
1809 | int err; |
1810 | ||
1811 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1812 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1813 | ||
1814 | if (proto == IPPROTO_UDPLITE) { | |
1815 | err = udplite_checksum_init(skb, uh); | |
1816 | if (err) | |
1817 | return err; | |
1818 | } | |
1819 | ||
ed70fcfc TH |
1820 | return skb_checksum_init_zero_check(skb, proto, uh->check, |
1821 | inet_compute_pseudo); | |
db8dac20 DM |
1822 | } |
1823 | ||
1824 | /* | |
1825 | * All we need to do is get the socket, and then do a checksum. | |
1826 | */ | |
1827 | ||
645ca708 | 1828 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1829 | int proto) |
1830 | { | |
1831 | struct sock *sk; | |
7b5e56f9 | 1832 | struct udphdr *uh; |
db8dac20 | 1833 | unsigned short ulen; |
adf30907 | 1834 | struct rtable *rt = skb_rtable(skb); |
2783ef23 | 1835 | __be32 saddr, daddr; |
0283328e | 1836 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1837 | |
1838 | /* | |
1839 | * Validate the packet. | |
1840 | */ | |
1841 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1842 | goto drop; /* No space for header. */ | |
1843 | ||
7b5e56f9 | 1844 | uh = udp_hdr(skb); |
db8dac20 | 1845 | ulen = ntohs(uh->len); |
ccc2d97c BM |
1846 | saddr = ip_hdr(skb)->saddr; |
1847 | daddr = ip_hdr(skb)->daddr; | |
1848 | ||
db8dac20 DM |
1849 | if (ulen > skb->len) |
1850 | goto short_packet; | |
1851 | ||
1852 | if (proto == IPPROTO_UDP) { | |
1853 | /* UDP validates ulen. */ | |
1854 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1855 | goto short_packet; | |
1856 | uh = udp_hdr(skb); | |
1857 | } | |
1858 | ||
1859 | if (udp4_csum_init(skb, uh, proto)) | |
1860 | goto csum_error; | |
1861 | ||
8afdd99a ED |
1862 | sk = skb_steal_sock(skb); |
1863 | if (sk) { | |
97502231 | 1864 | struct dst_entry *dst = skb_dst(skb); |
421b3885 | 1865 | int ret; |
421b3885 | 1866 | |
97502231 ED |
1867 | if (unlikely(sk->sk_rx_dst != dst)) |
1868 | udp_sk_rx_dst_set(sk, dst); | |
db8dac20 | 1869 | |
421b3885 | 1870 | ret = udp_queue_rcv_skb(sk, skb); |
8afdd99a | 1871 | sock_put(sk); |
421b3885 SB |
1872 | /* a return value > 0 means to resubmit the input, but |
1873 | * it wants the return to be -protocol, or 0 | |
1874 | */ | |
1875 | if (ret > 0) | |
1876 | return -ret; | |
1877 | return 0; | |
421b3885 | 1878 | } |
db8dac20 | 1879 | |
c18450a5 FF |
1880 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) |
1881 | return __udp4_lib_mcast_deliver(net, skb, uh, | |
36cbb245 | 1882 | saddr, daddr, udptable, proto); |
c18450a5 FF |
1883 | |
1884 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); | |
00db4124 | 1885 | if (sk) { |
a5b50476 ET |
1886 | int ret; |
1887 | ||
224d019c | 1888 | if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk)) |
2abb7cdc TH |
1889 | skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check, |
1890 | inet_compute_pseudo); | |
1891 | ||
a5b50476 | 1892 | ret = udp_queue_rcv_skb(sk, skb); |
db8dac20 DM |
1893 | sock_put(sk); |
1894 | ||
1895 | /* a return value > 0 means to resubmit the input, but | |
1896 | * it wants the return to be -protocol, or 0 | |
1897 | */ | |
1898 | if (ret > 0) | |
1899 | return -ret; | |
1900 | return 0; | |
1901 | } | |
1902 | ||
1903 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1904 | goto drop; | |
1905 | nf_reset(skb); | |
1906 | ||
1907 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1908 | if (udp_lib_checksum_complete(skb)) | |
1909 | goto csum_error; | |
1910 | ||
0283328e | 1911 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1912 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1913 | ||
1914 | /* | |
1915 | * Hmm. We got an UDP packet to a port to which we | |
1916 | * don't wanna listen. Ignore it. | |
1917 | */ | |
1918 | kfree_skb(skb); | |
1919 | return 0; | |
1920 | ||
1921 | short_packet: | |
ba7a46f1 JP |
1922 | net_dbg_ratelimited("UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n", |
1923 | proto == IPPROTO_UDPLITE ? "Lite" : "", | |
1924 | &saddr, ntohs(uh->source), | |
1925 | ulen, skb->len, | |
1926 | &daddr, ntohs(uh->dest)); | |
db8dac20 DM |
1927 | goto drop; |
1928 | ||
1929 | csum_error: | |
1930 | /* | |
1931 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1932 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1933 | */ | |
ba7a46f1 JP |
1934 | net_dbg_ratelimited("UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n", |
1935 | proto == IPPROTO_UDPLITE ? "Lite" : "", | |
1936 | &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest), | |
1937 | ulen); | |
6a5dc9e5 | 1938 | UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 | 1939 | drop: |
0283328e | 1940 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1941 | kfree_skb(skb); |
1942 | return 0; | |
1943 | } | |
1944 | ||
421b3885 SB |
1945 | /* We can only early demux multicast if there is a single matching socket. |
1946 | * If more than one socket found returns NULL | |
1947 | */ | |
1948 | static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net, | |
1949 | __be16 loc_port, __be32 loc_addr, | |
1950 | __be16 rmt_port, __be32 rmt_addr, | |
1951 | int dif) | |
1952 | { | |
1953 | struct sock *sk, *result; | |
1954 | struct hlist_nulls_node *node; | |
1955 | unsigned short hnum = ntohs(loc_port); | |
1956 | unsigned int count, slot = udp_hashfn(net, hnum, udp_table.mask); | |
1957 | struct udp_hslot *hslot = &udp_table.hash[slot]; | |
1958 | ||
63c6f81c ED |
1959 | /* Do not bother scanning a too big list */ |
1960 | if (hslot->count > 10) | |
1961 | return NULL; | |
1962 | ||
421b3885 SB |
1963 | rcu_read_lock(); |
1964 | begin: | |
1965 | count = 0; | |
1966 | result = NULL; | |
1967 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { | |
1968 | if (__udp_is_mcast_sock(net, sk, | |
1969 | loc_port, loc_addr, | |
1970 | rmt_port, rmt_addr, | |
1971 | dif, hnum)) { | |
1972 | result = sk; | |
1973 | ++count; | |
1974 | } | |
1975 | } | |
1976 | /* | |
1977 | * if the nulls value we got at the end of this lookup is | |
1978 | * not the expected one, we must restart lookup. | |
1979 | * We probably met an item that was moved to another chain. | |
1980 | */ | |
1981 | if (get_nulls_value(node) != slot) | |
1982 | goto begin; | |
1983 | ||
1984 | if (result) { | |
1985 | if (count != 1 || | |
1986 | unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) | |
1987 | result = NULL; | |
f69b923a | 1988 | else if (unlikely(!__udp_is_mcast_sock(net, result, |
421b3885 SB |
1989 | loc_port, loc_addr, |
1990 | rmt_port, rmt_addr, | |
1991 | dif, hnum))) { | |
1992 | sock_put(result); | |
1993 | result = NULL; | |
1994 | } | |
1995 | } | |
1996 | rcu_read_unlock(); | |
1997 | return result; | |
1998 | } | |
1999 | ||
2000 | /* For unicast we should only early demux connected sockets or we can | |
2001 | * break forwarding setups. The chains here can be long so only check | |
2002 | * if the first socket is an exact match and if not move on. | |
2003 | */ | |
2004 | static struct sock *__udp4_lib_demux_lookup(struct net *net, | |
2005 | __be16 loc_port, __be32 loc_addr, | |
2006 | __be16 rmt_port, __be32 rmt_addr, | |
2007 | int dif) | |
2008 | { | |
2009 | struct sock *sk, *result; | |
2010 | struct hlist_nulls_node *node; | |
2011 | unsigned short hnum = ntohs(loc_port); | |
2012 | unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum); | |
2013 | unsigned int slot2 = hash2 & udp_table.mask; | |
2014 | struct udp_hslot *hslot2 = &udp_table.hash2[slot2]; | |
c7228317 | 2015 | INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr); |
421b3885 SB |
2016 | const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum); |
2017 | ||
2018 | rcu_read_lock(); | |
2019 | result = NULL; | |
2020 | udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { | |
2021 | if (INET_MATCH(sk, net, acookie, | |
2022 | rmt_addr, loc_addr, ports, dif)) | |
2023 | result = sk; | |
2024 | /* Only check first socket in chain */ | |
2025 | break; | |
2026 | } | |
2027 | ||
2028 | if (result) { | |
2029 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) | |
2030 | result = NULL; | |
2031 | else if (unlikely(!INET_MATCH(sk, net, acookie, | |
2032 | rmt_addr, loc_addr, | |
2033 | ports, dif))) { | |
2034 | sock_put(result); | |
2035 | result = NULL; | |
2036 | } | |
2037 | } | |
2038 | rcu_read_unlock(); | |
2039 | return result; | |
2040 | } | |
2041 | ||
2042 | void udp_v4_early_demux(struct sk_buff *skb) | |
2043 | { | |
610438b7 ED |
2044 | struct net *net = dev_net(skb->dev); |
2045 | const struct iphdr *iph; | |
2046 | const struct udphdr *uh; | |
421b3885 SB |
2047 | struct sock *sk; |
2048 | struct dst_entry *dst; | |
421b3885 | 2049 | int dif = skb->dev->ifindex; |
6e540309 | 2050 | int ours; |
421b3885 SB |
2051 | |
2052 | /* validate the packet */ | |
2053 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr))) | |
2054 | return; | |
2055 | ||
610438b7 ED |
2056 | iph = ip_hdr(skb); |
2057 | uh = udp_hdr(skb); | |
2058 | ||
421b3885 | 2059 | if (skb->pkt_type == PACKET_BROADCAST || |
6e540309 SB |
2060 | skb->pkt_type == PACKET_MULTICAST) { |
2061 | struct in_device *in_dev = __in_dev_get_rcu(skb->dev); | |
2062 | ||
2063 | if (!in_dev) | |
2064 | return; | |
2065 | ||
2066 | ours = ip_check_mc_rcu(in_dev, iph->daddr, iph->saddr, | |
2067 | iph->protocol); | |
2068 | if (!ours) | |
2069 | return; | |
421b3885 SB |
2070 | sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr, |
2071 | uh->source, iph->saddr, dif); | |
6e540309 | 2072 | } else if (skb->pkt_type == PACKET_HOST) { |
421b3885 SB |
2073 | sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr, |
2074 | uh->source, iph->saddr, dif); | |
6e540309 | 2075 | } else { |
421b3885 | 2076 | return; |
6e540309 | 2077 | } |
421b3885 SB |
2078 | |
2079 | if (!sk) | |
2080 | return; | |
2081 | ||
2082 | skb->sk = sk; | |
82eabd9e | 2083 | skb->destructor = sock_efree; |
10e2eb87 | 2084 | dst = READ_ONCE(sk->sk_rx_dst); |
421b3885 SB |
2085 | |
2086 | if (dst) | |
2087 | dst = dst_check(dst, 0); | |
10e2eb87 ED |
2088 | if (dst) { |
2089 | /* DST_NOCACHE can not be used without taking a reference */ | |
2090 | if (dst->flags & DST_NOCACHE) { | |
2091 | if (likely(atomic_inc_not_zero(&dst->__refcnt))) | |
2092 | skb_dst_set(skb, dst); | |
2093 | } else { | |
2094 | skb_dst_set_noref(skb, dst); | |
2095 | } | |
2096 | } | |
421b3885 SB |
2097 | } |
2098 | ||
db8dac20 DM |
2099 | int udp_rcv(struct sk_buff *skb) |
2100 | { | |
645ca708 | 2101 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
2102 | } |
2103 | ||
7d06b2e0 | 2104 | void udp_destroy_sock(struct sock *sk) |
db8dac20 | 2105 | { |
44046a59 | 2106 | struct udp_sock *up = udp_sk(sk); |
8a74ad60 | 2107 | bool slow = lock_sock_fast(sk); |
db8dac20 | 2108 | udp_flush_pending_frames(sk); |
8a74ad60 | 2109 | unlock_sock_fast(sk, slow); |
44046a59 TP |
2110 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
2111 | void (*encap_destroy)(struct sock *sk); | |
2112 | encap_destroy = ACCESS_ONCE(up->encap_destroy); | |
2113 | if (encap_destroy) | |
2114 | encap_destroy(sk); | |
2115 | } | |
db8dac20 DM |
2116 | } |
2117 | ||
1da177e4 LT |
2118 | /* |
2119 | * Socket option code for UDP | |
2120 | */ | |
4c0a6cb0 | 2121 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 2122 | char __user *optval, unsigned int optlen, |
4c0a6cb0 | 2123 | int (*push_pending_frames)(struct sock *)) |
1da177e4 LT |
2124 | { |
2125 | struct udp_sock *up = udp_sk(sk); | |
1c19448c | 2126 | int val, valbool; |
1da177e4 | 2127 | int err = 0; |
b2bf1e26 | 2128 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 2129 | |
c482c568 | 2130 | if (optlen < sizeof(int)) |
1da177e4 LT |
2131 | return -EINVAL; |
2132 | ||
2133 | if (get_user(val, (int __user *)optval)) | |
2134 | return -EFAULT; | |
2135 | ||
1c19448c TH |
2136 | valbool = val ? 1 : 0; |
2137 | ||
6516c655 | 2138 | switch (optname) { |
1da177e4 LT |
2139 | case UDP_CORK: |
2140 | if (val != 0) { | |
2141 | up->corkflag = 1; | |
2142 | } else { | |
2143 | up->corkflag = 0; | |
2144 | lock_sock(sk); | |
4243cdc2 | 2145 | push_pending_frames(sk); |
1da177e4 LT |
2146 | release_sock(sk); |
2147 | } | |
2148 | break; | |
e905a9ed | 2149 | |
1da177e4 LT |
2150 | case UDP_ENCAP: |
2151 | switch (val) { | |
2152 | case 0: | |
2153 | case UDP_ENCAP_ESPINUDP: | |
2154 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
2155 | up->encap_rcv = xfrm4_udp_encap_rcv; |
2156 | /* FALLTHROUGH */ | |
342f0234 | 2157 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 | 2158 | up->encap_type = val; |
447167bf | 2159 | udp_encap_enable(); |
1da177e4 LT |
2160 | break; |
2161 | default: | |
2162 | err = -ENOPROTOOPT; | |
2163 | break; | |
2164 | } | |
2165 | break; | |
2166 | ||
1c19448c TH |
2167 | case UDP_NO_CHECK6_TX: |
2168 | up->no_check6_tx = valbool; | |
2169 | break; | |
2170 | ||
2171 | case UDP_NO_CHECK6_RX: | |
2172 | up->no_check6_rx = valbool; | |
2173 | break; | |
2174 | ||
ba4e58ec GR |
2175 | /* |
2176 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
2177 | */ | |
2178 | /* The sender sets actual checksum coverage length via this option. | |
2179 | * The case coverage > packet length is handled by send module. */ | |
2180 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 2181 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
2182 | return -ENOPROTOOPT; |
2183 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
2184 | val = 8; | |
4be929be AD |
2185 | else if (val > USHRT_MAX) |
2186 | val = USHRT_MAX; | |
ba4e58ec GR |
2187 | up->pcslen = val; |
2188 | up->pcflag |= UDPLITE_SEND_CC; | |
2189 | break; | |
2190 | ||
e905a9ed YH |
2191 | /* The receiver specifies a minimum checksum coverage value. To make |
2192 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
2193 | * used, this again means full checksum coverage. */ |
2194 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 2195 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
2196 | return -ENOPROTOOPT; |
2197 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
2198 | val = 8; | |
4be929be AD |
2199 | else if (val > USHRT_MAX) |
2200 | val = USHRT_MAX; | |
ba4e58ec GR |
2201 | up->pcrlen = val; |
2202 | up->pcflag |= UDPLITE_RECV_CC; | |
2203 | break; | |
2204 | ||
1da177e4 LT |
2205 | default: |
2206 | err = -ENOPROTOOPT; | |
2207 | break; | |
6516c655 | 2208 | } |
1da177e4 LT |
2209 | |
2210 | return err; | |
2211 | } | |
c482c568 | 2212 | EXPORT_SYMBOL(udp_lib_setsockopt); |
1da177e4 | 2213 | |
db8dac20 | 2214 | int udp_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 2215 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
2216 | { |
2217 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2218 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
2219 | udp_push_pending_frames); | |
2220 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
2221 | } | |
2222 | ||
2223 | #ifdef CONFIG_COMPAT | |
2224 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 2225 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
2226 | { |
2227 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2228 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
2229 | udp_push_pending_frames); | |
2230 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
2231 | } | |
2232 | #endif | |
2233 | ||
4c0a6cb0 GR |
2234 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
2235 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
2236 | { |
2237 | struct udp_sock *up = udp_sk(sk); | |
2238 | int val, len; | |
2239 | ||
c482c568 | 2240 | if (get_user(len, optlen)) |
1da177e4 LT |
2241 | return -EFAULT; |
2242 | ||
2243 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 2244 | |
6516c655 | 2245 | if (len < 0) |
1da177e4 LT |
2246 | return -EINVAL; |
2247 | ||
6516c655 | 2248 | switch (optname) { |
1da177e4 LT |
2249 | case UDP_CORK: |
2250 | val = up->corkflag; | |
2251 | break; | |
2252 | ||
2253 | case UDP_ENCAP: | |
2254 | val = up->encap_type; | |
2255 | break; | |
2256 | ||
1c19448c TH |
2257 | case UDP_NO_CHECK6_TX: |
2258 | val = up->no_check6_tx; | |
2259 | break; | |
2260 | ||
2261 | case UDP_NO_CHECK6_RX: | |
2262 | val = up->no_check6_rx; | |
2263 | break; | |
2264 | ||
ba4e58ec GR |
2265 | /* The following two cannot be changed on UDP sockets, the return is |
2266 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
2267 | case UDPLITE_SEND_CSCOV: | |
2268 | val = up->pcslen; | |
2269 | break; | |
2270 | ||
2271 | case UDPLITE_RECV_CSCOV: | |
2272 | val = up->pcrlen; | |
2273 | break; | |
2274 | ||
1da177e4 LT |
2275 | default: |
2276 | return -ENOPROTOOPT; | |
6516c655 | 2277 | } |
1da177e4 | 2278 | |
6516c655 | 2279 | if (put_user(len, optlen)) |
e905a9ed | 2280 | return -EFAULT; |
c482c568 | 2281 | if (copy_to_user(optval, &val, len)) |
1da177e4 | 2282 | return -EFAULT; |
e905a9ed | 2283 | return 0; |
1da177e4 | 2284 | } |
c482c568 | 2285 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1da177e4 | 2286 | |
db8dac20 DM |
2287 | int udp_getsockopt(struct sock *sk, int level, int optname, |
2288 | char __user *optval, int __user *optlen) | |
2289 | { | |
2290 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2291 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
2292 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
2293 | } | |
2294 | ||
2295 | #ifdef CONFIG_COMPAT | |
2296 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
2297 | char __user *optval, int __user *optlen) | |
2298 | { | |
2299 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2300 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
2301 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
2302 | } | |
2303 | #endif | |
1da177e4 LT |
2304 | /** |
2305 | * udp_poll - wait for a UDP event. | |
2306 | * @file - file struct | |
2307 | * @sock - socket | |
2308 | * @wait - poll table | |
2309 | * | |
e905a9ed | 2310 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
2311 | * blocking sockets. If application is using a blocking fd |
2312 | * and a packet with checksum error is in the queue; | |
2313 | * then it could get return from select indicating data available | |
2314 | * but then block when reading it. Add special case code | |
2315 | * to work around these arguably broken applications. | |
2316 | */ | |
2317 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
2318 | { | |
2319 | unsigned int mask = datagram_poll(file, sock, wait); | |
2320 | struct sock *sk = sock->sk; | |
ba4e58ec | 2321 | |
c3f1dbaf DM |
2322 | sock_rps_record_flow(sk); |
2323 | ||
1da177e4 | 2324 | /* Check for false positives due to checksum errors */ |
85584672 ED |
2325 | if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) && |
2326 | !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk)) | |
2327 | mask &= ~(POLLIN | POLLRDNORM); | |
1da177e4 LT |
2328 | |
2329 | return mask; | |
e905a9ed | 2330 | |
1da177e4 | 2331 | } |
c482c568 | 2332 | EXPORT_SYMBOL(udp_poll); |
1da177e4 | 2333 | |
db8dac20 DM |
2334 | struct proto udp_prot = { |
2335 | .name = "UDP", | |
2336 | .owner = THIS_MODULE, | |
2337 | .close = udp_lib_close, | |
2338 | .connect = ip4_datagram_connect, | |
2339 | .disconnect = udp_disconnect, | |
2340 | .ioctl = udp_ioctl, | |
2341 | .destroy = udp_destroy_sock, | |
2342 | .setsockopt = udp_setsockopt, | |
2343 | .getsockopt = udp_getsockopt, | |
2344 | .sendmsg = udp_sendmsg, | |
2345 | .recvmsg = udp_recvmsg, | |
2346 | .sendpage = udp_sendpage, | |
93821778 | 2347 | .backlog_rcv = __udp_queue_rcv_skb, |
8141ed9f | 2348 | .release_cb = ip4_datagram_release_cb, |
db8dac20 DM |
2349 | .hash = udp_lib_hash, |
2350 | .unhash = udp_lib_unhash, | |
719f8358 | 2351 | .rehash = udp_v4_rehash, |
db8dac20 DM |
2352 | .get_port = udp_v4_get_port, |
2353 | .memory_allocated = &udp_memory_allocated, | |
2354 | .sysctl_mem = sysctl_udp_mem, | |
2355 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
2356 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
2357 | .obj_size = sizeof(struct udp_sock), | |
271b72c7 | 2358 | .slab_flags = SLAB_DESTROY_BY_RCU, |
645ca708 | 2359 | .h.udp_table = &udp_table, |
db8dac20 DM |
2360 | #ifdef CONFIG_COMPAT |
2361 | .compat_setsockopt = compat_udp_setsockopt, | |
2362 | .compat_getsockopt = compat_udp_getsockopt, | |
2363 | #endif | |
fcbdf09d | 2364 | .clear_sk = sk_prot_clear_portaddr_nulls, |
db8dac20 | 2365 | }; |
c482c568 | 2366 | EXPORT_SYMBOL(udp_prot); |
1da177e4 LT |
2367 | |
2368 | /* ------------------------------------------------------------------------ */ | |
2369 | #ifdef CONFIG_PROC_FS | |
2370 | ||
645ca708 | 2371 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
2372 | { |
2373 | struct sock *sk; | |
2374 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2375 | struct net *net = seq_file_net(seq); |
1da177e4 | 2376 | |
f86dcc5a ED |
2377 | for (state->bucket = start; state->bucket <= state->udp_table->mask; |
2378 | ++state->bucket) { | |
88ab1932 | 2379 | struct hlist_nulls_node *node; |
645ca708 | 2380 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
f86dcc5a ED |
2381 | |
2382 | if (hlist_nulls_empty(&hslot->head)) | |
2383 | continue; | |
2384 | ||
645ca708 | 2385 | spin_lock_bh(&hslot->lock); |
88ab1932 | 2386 | sk_nulls_for_each(sk, node, &hslot->head) { |
878628fb | 2387 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 2388 | continue; |
1da177e4 LT |
2389 | if (sk->sk_family == state->family) |
2390 | goto found; | |
2391 | } | |
645ca708 | 2392 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
2393 | } |
2394 | sk = NULL; | |
2395 | found: | |
2396 | return sk; | |
2397 | } | |
2398 | ||
2399 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
2400 | { | |
2401 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2402 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2403 | |
2404 | do { | |
88ab1932 | 2405 | sk = sk_nulls_next(sk); |
878628fb | 2406 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 2407 | |
645ca708 | 2408 | if (!sk) { |
f86dcc5a | 2409 | if (state->bucket <= state->udp_table->mask) |
30842f29 | 2410 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
645ca708 | 2411 | return udp_get_first(seq, state->bucket + 1); |
1da177e4 LT |
2412 | } |
2413 | return sk; | |
2414 | } | |
2415 | ||
2416 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
2417 | { | |
645ca708 | 2418 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
2419 | |
2420 | if (sk) | |
6516c655 | 2421 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
2422 | --pos; |
2423 | return pos ? NULL : sk; | |
2424 | } | |
2425 | ||
2426 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
2427 | { | |
30842f29 | 2428 | struct udp_iter_state *state = seq->private; |
f86dcc5a | 2429 | state->bucket = MAX_UDP_PORTS; |
30842f29 | 2430 | |
b50660f1 | 2431 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
2432 | } |
2433 | ||
2434 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2435 | { | |
2436 | struct sock *sk; | |
2437 | ||
b50660f1 | 2438 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2439 | sk = udp_get_idx(seq, 0); |
2440 | else | |
2441 | sk = udp_get_next(seq, v); | |
2442 | ||
2443 | ++*pos; | |
2444 | return sk; | |
2445 | } | |
2446 | ||
2447 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
2448 | { | |
645ca708 ED |
2449 | struct udp_iter_state *state = seq->private; |
2450 | ||
f86dcc5a | 2451 | if (state->bucket <= state->udp_table->mask) |
645ca708 | 2452 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
1da177e4 LT |
2453 | } |
2454 | ||
73cb88ec | 2455 | int udp_seq_open(struct inode *inode, struct file *file) |
1da177e4 | 2456 | { |
d9dda78b | 2457 | struct udp_seq_afinfo *afinfo = PDE_DATA(inode); |
a2be75c1 DL |
2458 | struct udp_iter_state *s; |
2459 | int err; | |
a91275ef | 2460 | |
a2be75c1 DL |
2461 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2462 | sizeof(struct udp_iter_state)); | |
2463 | if (err < 0) | |
2464 | return err; | |
a91275ef | 2465 | |
a2be75c1 | 2466 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2467 | s->family = afinfo->family; |
645ca708 | 2468 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 2469 | return err; |
a91275ef | 2470 | } |
73cb88ec | 2471 | EXPORT_SYMBOL(udp_seq_open); |
a91275ef | 2472 | |
1da177e4 | 2473 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 2474 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
2475 | { |
2476 | struct proc_dir_entry *p; | |
2477 | int rc = 0; | |
2478 | ||
dda61925 DL |
2479 | afinfo->seq_ops.start = udp_seq_start; |
2480 | afinfo->seq_ops.next = udp_seq_next; | |
2481 | afinfo->seq_ops.stop = udp_seq_stop; | |
2482 | ||
84841c3c | 2483 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2484 | afinfo->seq_fops, afinfo); |
84841c3c | 2485 | if (!p) |
1da177e4 LT |
2486 | rc = -ENOMEM; |
2487 | return rc; | |
2488 | } | |
c482c568 | 2489 | EXPORT_SYMBOL(udp_proc_register); |
1da177e4 | 2490 | |
0c96d8c5 | 2491 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 2492 | { |
ece31ffd | 2493 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2494 | } |
c482c568 | 2495 | EXPORT_SYMBOL(udp_proc_unregister); |
db8dac20 DM |
2496 | |
2497 | /* ------------------------------------------------------------------------ */ | |
5e659e4c | 2498 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
652586df | 2499 | int bucket) |
db8dac20 DM |
2500 | { |
2501 | struct inet_sock *inet = inet_sk(sp); | |
c720c7e8 ED |
2502 | __be32 dest = inet->inet_daddr; |
2503 | __be32 src = inet->inet_rcv_saddr; | |
2504 | __u16 destp = ntohs(inet->inet_dport); | |
2505 | __u16 srcp = ntohs(inet->inet_sport); | |
db8dac20 | 2506 | |
f86dcc5a | 2507 | seq_printf(f, "%5d: %08X:%04X %08X:%04X" |
652586df | 2508 | " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d", |
db8dac20 | 2509 | bucket, src, srcp, dest, destp, sp->sk_state, |
31e6d363 ED |
2510 | sk_wmem_alloc_get(sp), |
2511 | sk_rmem_alloc_get(sp), | |
a7cb5a49 EB |
2512 | 0, 0L, 0, |
2513 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)), | |
2514 | 0, sock_i_ino(sp), | |
cb61cb9b | 2515 | atomic_read(&sp->sk_refcnt), sp, |
652586df | 2516 | atomic_read(&sp->sk_drops)); |
db8dac20 DM |
2517 | } |
2518 | ||
2519 | int udp4_seq_show(struct seq_file *seq, void *v) | |
2520 | { | |
652586df | 2521 | seq_setwidth(seq, 127); |
db8dac20 | 2522 | if (v == SEQ_START_TOKEN) |
652586df | 2523 | seq_puts(seq, " sl local_address rem_address st tx_queue " |
db8dac20 | 2524 | "rx_queue tr tm->when retrnsmt uid timeout " |
cb61cb9b | 2525 | "inode ref pointer drops"); |
db8dac20 | 2526 | else { |
db8dac20 DM |
2527 | struct udp_iter_state *state = seq->private; |
2528 | ||
652586df | 2529 | udp4_format_sock(v, seq, state->bucket); |
db8dac20 | 2530 | } |
652586df | 2531 | seq_pad(seq, '\n'); |
db8dac20 DM |
2532 | return 0; |
2533 | } | |
2534 | ||
73cb88ec AV |
2535 | static const struct file_operations udp_afinfo_seq_fops = { |
2536 | .owner = THIS_MODULE, | |
2537 | .open = udp_seq_open, | |
2538 | .read = seq_read, | |
2539 | .llseek = seq_lseek, | |
2540 | .release = seq_release_net | |
2541 | }; | |
2542 | ||
db8dac20 | 2543 | /* ------------------------------------------------------------------------ */ |
db8dac20 | 2544 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
2545 | .name = "udp", |
2546 | .family = AF_INET, | |
645ca708 | 2547 | .udp_table = &udp_table, |
73cb88ec | 2548 | .seq_fops = &udp_afinfo_seq_fops, |
dda61925 DL |
2549 | .seq_ops = { |
2550 | .show = udp4_seq_show, | |
2551 | }, | |
db8dac20 DM |
2552 | }; |
2553 | ||
2c8c1e72 | 2554 | static int __net_init udp4_proc_init_net(struct net *net) |
15439feb PE |
2555 | { |
2556 | return udp_proc_register(net, &udp4_seq_afinfo); | |
2557 | } | |
2558 | ||
2c8c1e72 | 2559 | static void __net_exit udp4_proc_exit_net(struct net *net) |
15439feb PE |
2560 | { |
2561 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
2562 | } | |
2563 | ||
2564 | static struct pernet_operations udp4_net_ops = { | |
2565 | .init = udp4_proc_init_net, | |
2566 | .exit = udp4_proc_exit_net, | |
2567 | }; | |
2568 | ||
db8dac20 DM |
2569 | int __init udp4_proc_init(void) |
2570 | { | |
15439feb | 2571 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
2572 | } |
2573 | ||
2574 | void udp4_proc_exit(void) | |
2575 | { | |
15439feb | 2576 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 2577 | } |
1da177e4 LT |
2578 | #endif /* CONFIG_PROC_FS */ |
2579 | ||
f86dcc5a ED |
2580 | static __initdata unsigned long uhash_entries; |
2581 | static int __init set_uhash_entries(char *str) | |
645ca708 | 2582 | { |
413c27d8 EZ |
2583 | ssize_t ret; |
2584 | ||
f86dcc5a ED |
2585 | if (!str) |
2586 | return 0; | |
413c27d8 EZ |
2587 | |
2588 | ret = kstrtoul(str, 0, &uhash_entries); | |
2589 | if (ret) | |
2590 | return 0; | |
2591 | ||
f86dcc5a ED |
2592 | if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN) |
2593 | uhash_entries = UDP_HTABLE_SIZE_MIN; | |
2594 | return 1; | |
2595 | } | |
2596 | __setup("uhash_entries=", set_uhash_entries); | |
645ca708 | 2597 | |
f86dcc5a ED |
2598 | void __init udp_table_init(struct udp_table *table, const char *name) |
2599 | { | |
2600 | unsigned int i; | |
2601 | ||
31fe62b9 TB |
2602 | table->hash = alloc_large_system_hash(name, |
2603 | 2 * sizeof(struct udp_hslot), | |
2604 | uhash_entries, | |
2605 | 21, /* one slot per 2 MB */ | |
2606 | 0, | |
2607 | &table->log, | |
2608 | &table->mask, | |
2609 | UDP_HTABLE_SIZE_MIN, | |
2610 | 64 * 1024); | |
2611 | ||
512615b6 | 2612 | table->hash2 = table->hash + (table->mask + 1); |
f86dcc5a | 2613 | for (i = 0; i <= table->mask; i++) { |
88ab1932 | 2614 | INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i); |
fdcc8aa9 | 2615 | table->hash[i].count = 0; |
645ca708 ED |
2616 | spin_lock_init(&table->hash[i].lock); |
2617 | } | |
512615b6 ED |
2618 | for (i = 0; i <= table->mask; i++) { |
2619 | INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i); | |
2620 | table->hash2[i].count = 0; | |
2621 | spin_lock_init(&table->hash2[i].lock); | |
2622 | } | |
645ca708 ED |
2623 | } |
2624 | ||
723b8e46 TH |
2625 | u32 udp_flow_hashrnd(void) |
2626 | { | |
2627 | static u32 hashrnd __read_mostly; | |
2628 | ||
2629 | net_get_random_once(&hashrnd, sizeof(hashrnd)); | |
2630 | ||
2631 | return hashrnd; | |
2632 | } | |
2633 | EXPORT_SYMBOL(udp_flow_hashrnd); | |
2634 | ||
95766fff HA |
2635 | void __init udp_init(void) |
2636 | { | |
f03d78db | 2637 | unsigned long limit; |
95766fff | 2638 | |
f86dcc5a | 2639 | udp_table_init(&udp_table, "UDP"); |
f03d78db | 2640 | limit = nr_free_buffer_pages() / 8; |
95766fff HA |
2641 | limit = max(limit, 128UL); |
2642 | sysctl_udp_mem[0] = limit / 4 * 3; | |
2643 | sysctl_udp_mem[1] = limit; | |
2644 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
2645 | ||
2646 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
2647 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
2648 | } |