]>
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; |
ba418fa3 | 146 | kuid_t uid = sock_i_uid(sk); |
25030a7f | 147 | |
ca065d0c | 148 | sk_for_each(sk2, &hslot->head) { |
9d4fb27d JP |
149 | if (net_eq(sock_net(sk2), net) && |
150 | sk2 != sk && | |
d4cada4a | 151 | (bitmap || udp_sk(sk2)->udp_port_hash == num) && |
9d4fb27d JP |
152 | (!sk2->sk_reuse || !sk->sk_reuse) && |
153 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
154 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 | 155 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
e32ea7e7 | 156 | rcu_access_pointer(sk->sk_reuseport_cb) || |
4243cdc2 | 157 | !uid_eq(uid, sock_i_uid(sk2))) && |
e32ea7e7 | 158 | saddr_comp(sk, sk2, true)) { |
4243cdc2 | 159 | if (!bitmap) |
98322f22 | 160 | return 1; |
4243cdc2 | 161 | __set_bit(udp_sk(sk2)->udp_port_hash >> log, bitmap); |
98322f22 | 162 | } |
4243cdc2 | 163 | } |
25030a7f GR |
164 | return 0; |
165 | } | |
166 | ||
30fff923 ED |
167 | /* |
168 | * Note: we still hold spinlock of primary hash chain, so no other writer | |
169 | * can insert/delete a socket with local_port == num | |
170 | */ | |
171 | static int udp_lib_lport_inuse2(struct net *net, __u16 num, | |
4243cdc2 JP |
172 | struct udp_hslot *hslot2, |
173 | struct sock *sk, | |
174 | int (*saddr_comp)(const struct sock *sk1, | |
e32ea7e7 CG |
175 | const struct sock *sk2, |
176 | bool match_wildcard)) | |
30fff923 ED |
177 | { |
178 | struct sock *sk2; | |
ba418fa3 | 179 | kuid_t uid = sock_i_uid(sk); |
30fff923 ED |
180 | int res = 0; |
181 | ||
182 | spin_lock(&hslot2->lock); | |
ca065d0c | 183 | udp_portaddr_for_each_entry(sk2, &hslot2->head) { |
9d4fb27d JP |
184 | if (net_eq(sock_net(sk2), net) && |
185 | sk2 != sk && | |
186 | (udp_sk(sk2)->udp_port_hash == num) && | |
187 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
188 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
189 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 | 190 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
e32ea7e7 | 191 | rcu_access_pointer(sk->sk_reuseport_cb) || |
4243cdc2 | 192 | !uid_eq(uid, sock_i_uid(sk2))) && |
e32ea7e7 | 193 | saddr_comp(sk, sk2, true)) { |
30fff923 ED |
194 | res = 1; |
195 | break; | |
196 | } | |
4243cdc2 | 197 | } |
30fff923 ED |
198 | spin_unlock(&hslot2->lock); |
199 | return res; | |
200 | } | |
201 | ||
e32ea7e7 CG |
202 | static int udp_reuseport_add_sock(struct sock *sk, struct udp_hslot *hslot, |
203 | int (*saddr_same)(const struct sock *sk1, | |
204 | const struct sock *sk2, | |
205 | bool match_wildcard)) | |
206 | { | |
207 | struct net *net = sock_net(sk); | |
e32ea7e7 CG |
208 | kuid_t uid = sock_i_uid(sk); |
209 | struct sock *sk2; | |
210 | ||
ca065d0c | 211 | sk_for_each(sk2, &hslot->head) { |
e32ea7e7 CG |
212 | if (net_eq(sock_net(sk2), net) && |
213 | sk2 != sk && | |
214 | sk2->sk_family == sk->sk_family && | |
215 | ipv6_only_sock(sk2) == ipv6_only_sock(sk) && | |
216 | (udp_sk(sk2)->udp_port_hash == udp_sk(sk)->udp_port_hash) && | |
217 | (sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
218 | sk2->sk_reuseport && uid_eq(uid, sock_i_uid(sk2)) && | |
219 | (*saddr_same)(sk, sk2, false)) { | |
220 | return reuseport_add_sock(sk, sk2); | |
221 | } | |
222 | } | |
223 | ||
224 | /* Initial allocation may have already happened via setsockopt */ | |
225 | if (!rcu_access_pointer(sk->sk_reuseport_cb)) | |
226 | return reuseport_alloc(sk); | |
227 | return 0; | |
228 | } | |
229 | ||
25030a7f | 230 | /** |
6ba5a3c5 | 231 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
232 | * |
233 | * @sk: socket struct in question | |
234 | * @snum: port number to look up | |
df2bc459 | 235 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25985edc | 236 | * @hash2_nulladdr: AF-dependent hash value in secondary hash chains, |
30fff923 | 237 | * with NULL address |
25030a7f | 238 | */ |
6ba5a3c5 | 239 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
4243cdc2 | 240 | int (*saddr_comp)(const struct sock *sk1, |
e32ea7e7 CG |
241 | const struct sock *sk2, |
242 | bool match_wildcard), | |
30fff923 | 243 | unsigned int hash2_nulladdr) |
25030a7f | 244 | { |
512615b6 | 245 | struct udp_hslot *hslot, *hslot2; |
645ca708 | 246 | struct udp_table *udptable = sk->sk_prot->h.udp_table; |
25030a7f | 247 | int error = 1; |
3b1e0a65 | 248 | struct net *net = sock_net(sk); |
1da177e4 | 249 | |
32c1da70 | 250 | if (!snum) { |
9088c560 | 251 | int low, high, remaining; |
95c96174 | 252 | unsigned int rand; |
98322f22 ED |
253 | unsigned short first, last; |
254 | DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN); | |
32c1da70 | 255 | |
0bbf87d8 | 256 | inet_get_local_port_range(net, &low, &high); |
a25de534 | 257 | remaining = (high - low) + 1; |
227b60f5 | 258 | |
63862b5b | 259 | rand = prandom_u32(); |
8fc54f68 | 260 | first = reciprocal_scale(rand, remaining) + low; |
98322f22 ED |
261 | /* |
262 | * force rand to be an odd multiple of UDP_HTABLE_SIZE | |
263 | */ | |
f86dcc5a | 264 | rand = (rand | 1) * (udptable->mask + 1); |
5781b235 ED |
265 | last = first + udptable->mask + 1; |
266 | do { | |
f86dcc5a | 267 | hslot = udp_hashslot(udptable, net, first); |
98322f22 | 268 | bitmap_zero(bitmap, PORTS_PER_CHAIN); |
645ca708 | 269 | spin_lock_bh(&hslot->lock); |
98322f22 | 270 | udp_lib_lport_inuse(net, snum, hslot, bitmap, sk, |
f86dcc5a | 271 | saddr_comp, udptable->log); |
98322f22 ED |
272 | |
273 | snum = first; | |
274 | /* | |
275 | * Iterate on all possible values of snum for this hash. | |
276 | * Using steps of an odd multiple of UDP_HTABLE_SIZE | |
277 | * give us randomization and full range coverage. | |
278 | */ | |
9088c560 | 279 | do { |
98322f22 | 280 | if (low <= snum && snum <= high && |
e3826f1e | 281 | !test_bit(snum >> udptable->log, bitmap) && |
122ff243 | 282 | !inet_is_local_reserved_port(net, snum)) |
98322f22 ED |
283 | goto found; |
284 | snum += rand; | |
285 | } while (snum != first); | |
286 | spin_unlock_bh(&hslot->lock); | |
5781b235 | 287 | } while (++first != last); |
98322f22 | 288 | goto fail; |
645ca708 | 289 | } else { |
f86dcc5a | 290 | hslot = udp_hashslot(udptable, net, snum); |
645ca708 | 291 | spin_lock_bh(&hslot->lock); |
30fff923 ED |
292 | if (hslot->count > 10) { |
293 | int exist; | |
294 | unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum; | |
295 | ||
296 | slot2 &= udptable->mask; | |
297 | hash2_nulladdr &= udptable->mask; | |
298 | ||
299 | hslot2 = udp_hashslot2(udptable, slot2); | |
300 | if (hslot->count < hslot2->count) | |
301 | goto scan_primary_hash; | |
302 | ||
303 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
304 | sk, saddr_comp); | |
305 | if (!exist && (hash2_nulladdr != slot2)) { | |
306 | hslot2 = udp_hashslot2(udptable, hash2_nulladdr); | |
307 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
308 | sk, saddr_comp); | |
309 | } | |
310 | if (exist) | |
311 | goto fail_unlock; | |
312 | else | |
313 | goto found; | |
314 | } | |
315 | scan_primary_hash: | |
f86dcc5a ED |
316 | if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk, |
317 | saddr_comp, 0)) | |
645ca708 ED |
318 | goto fail_unlock; |
319 | } | |
98322f22 | 320 | found: |
c720c7e8 | 321 | inet_sk(sk)->inet_num = snum; |
d4cada4a ED |
322 | udp_sk(sk)->udp_port_hash = snum; |
323 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
1da177e4 | 324 | if (sk_unhashed(sk)) { |
e32ea7e7 CG |
325 | if (sk->sk_reuseport && |
326 | udp_reuseport_add_sock(sk, hslot, saddr_comp)) { | |
327 | inet_sk(sk)->inet_num = 0; | |
328 | udp_sk(sk)->udp_port_hash = 0; | |
329 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
330 | goto fail_unlock; | |
331 | } | |
332 | ||
ca065d0c | 333 | sk_add_node_rcu(sk, &hslot->head); |
fdcc8aa9 | 334 | hslot->count++; |
c29a0bc4 | 335 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
512615b6 ED |
336 | |
337 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
338 | spin_lock(&hslot2->lock); | |
ca065d0c | 339 | hlist_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, |
512615b6 ED |
340 | &hslot2->head); |
341 | hslot2->count++; | |
342 | spin_unlock(&hslot2->lock); | |
1da177e4 | 343 | } |
ca065d0c | 344 | sock_set_flag(sk, SOCK_RCU_FREE); |
25030a7f | 345 | error = 0; |
645ca708 ED |
346 | fail_unlock: |
347 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 348 | fail: |
25030a7f GR |
349 | return error; |
350 | } | |
c482c568 | 351 | EXPORT_SYMBOL(udp_lib_get_port); |
25030a7f | 352 | |
e32ea7e7 CG |
353 | /* match_wildcard == true: 0.0.0.0 equals to any IPv4 addresses |
354 | * match_wildcard == false: addresses must be exactly the same, i.e. | |
355 | * 0.0.0.0 only equals to 0.0.0.0 | |
356 | */ | |
c125e80b CG |
357 | int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2, |
358 | bool match_wildcard) | |
db8dac20 DM |
359 | { |
360 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
361 | ||
e32ea7e7 CG |
362 | if (!ipv6_only_sock(sk2)) { |
363 | if (inet1->inet_rcv_saddr == inet2->inet_rcv_saddr) | |
364 | return 1; | |
365 | if (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr) | |
366 | return match_wildcard; | |
367 | } | |
368 | return 0; | |
db8dac20 DM |
369 | } |
370 | ||
6eada011 ED |
371 | static u32 udp4_portaddr_hash(const struct net *net, __be32 saddr, |
372 | unsigned int port) | |
d4cada4a | 373 | { |
0eae88f3 | 374 | return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port; |
d4cada4a ED |
375 | } |
376 | ||
6ba5a3c5 | 377 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 378 | { |
30fff923 | 379 | unsigned int hash2_nulladdr = |
0eae88f3 | 380 | udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum); |
30fff923 ED |
381 | unsigned int hash2_partial = |
382 | udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0); | |
383 | ||
d4cada4a | 384 | /* precompute partial secondary hash */ |
30fff923 ED |
385 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
386 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr); | |
db8dac20 DM |
387 | } |
388 | ||
60c04aec JP |
389 | static inline int compute_score(struct sock *sk, struct net *net, |
390 | __be32 saddr, unsigned short hnum, __be16 sport, | |
391 | __be32 daddr, __be16 dport, int dif) | |
645ca708 | 392 | { |
60c04aec JP |
393 | int score; |
394 | struct inet_sock *inet; | |
645ca708 | 395 | |
60c04aec JP |
396 | if (!net_eq(sock_net(sk), net) || |
397 | udp_sk(sk)->udp_port_hash != hnum || | |
398 | ipv6_only_sock(sk)) | |
399 | return -1; | |
645ca708 | 400 | |
60c04aec JP |
401 | score = (sk->sk_family == PF_INET) ? 2 : 1; |
402 | inet = inet_sk(sk); | |
403 | ||
404 | if (inet->inet_rcv_saddr) { | |
405 | if (inet->inet_rcv_saddr != daddr) | |
406 | return -1; | |
407 | score += 4; | |
645ca708 | 408 | } |
60c04aec JP |
409 | |
410 | if (inet->inet_daddr) { | |
411 | if (inet->inet_daddr != saddr) | |
412 | return -1; | |
413 | score += 4; | |
414 | } | |
415 | ||
416 | if (inet->inet_dport) { | |
417 | if (inet->inet_dport != sport) | |
418 | return -1; | |
419 | score += 4; | |
420 | } | |
421 | ||
422 | if (sk->sk_bound_dev_if) { | |
423 | if (sk->sk_bound_dev_if != dif) | |
424 | return -1; | |
425 | score += 4; | |
426 | } | |
70da268b ED |
427 | if (sk->sk_incoming_cpu == raw_smp_processor_id()) |
428 | score++; | |
645ca708 ED |
429 | return score; |
430 | } | |
431 | ||
5051ebd2 ED |
432 | /* |
433 | * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num) | |
434 | */ | |
5051ebd2 ED |
435 | static inline int compute_score2(struct sock *sk, struct net *net, |
436 | __be32 saddr, __be16 sport, | |
437 | __be32 daddr, unsigned int hnum, int dif) | |
438 | { | |
60c04aec JP |
439 | int score; |
440 | struct inet_sock *inet; | |
5051ebd2 | 441 | |
60c04aec JP |
442 | if (!net_eq(sock_net(sk), net) || |
443 | ipv6_only_sock(sk)) | |
444 | return -1; | |
5051ebd2 | 445 | |
60c04aec JP |
446 | inet = inet_sk(sk); |
447 | ||
448 | if (inet->inet_rcv_saddr != daddr || | |
449 | inet->inet_num != hnum) | |
450 | return -1; | |
451 | ||
452 | score = (sk->sk_family == PF_INET) ? 2 : 1; | |
453 | ||
454 | if (inet->inet_daddr) { | |
455 | if (inet->inet_daddr != saddr) | |
5051ebd2 | 456 | return -1; |
60c04aec JP |
457 | score += 4; |
458 | } | |
459 | ||
460 | if (inet->inet_dport) { | |
461 | if (inet->inet_dport != sport) | |
5051ebd2 | 462 | return -1; |
60c04aec JP |
463 | score += 4; |
464 | } | |
5051ebd2 | 465 | |
60c04aec JP |
466 | if (sk->sk_bound_dev_if) { |
467 | if (sk->sk_bound_dev_if != dif) | |
468 | return -1; | |
469 | score += 4; | |
5051ebd2 | 470 | } |
60c04aec | 471 | |
70da268b ED |
472 | if (sk->sk_incoming_cpu == raw_smp_processor_id()) |
473 | score++; | |
474 | ||
5051ebd2 ED |
475 | return score; |
476 | } | |
477 | ||
6eada011 ED |
478 | static u32 udp_ehashfn(const struct net *net, const __be32 laddr, |
479 | const __u16 lport, const __be32 faddr, | |
480 | const __be16 fport) | |
65cd8033 | 481 | { |
1bbdceef HFS |
482 | static u32 udp_ehash_secret __read_mostly; |
483 | ||
484 | net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret)); | |
485 | ||
65cd8033 | 486 | return __inet_ehashfn(laddr, lport, faddr, fport, |
1bbdceef | 487 | udp_ehash_secret + net_hash_mix(net)); |
65cd8033 HFS |
488 | } |
489 | ||
5051ebd2 ED |
490 | /* called with read_rcu_lock() */ |
491 | static struct sock *udp4_lib_lookup2(struct net *net, | |
492 | __be32 saddr, __be16 sport, | |
493 | __be32 daddr, unsigned int hnum, int dif, | |
1134158b CG |
494 | struct udp_hslot *hslot2, unsigned int slot2, |
495 | struct sk_buff *skb) | |
5051ebd2 ED |
496 | { |
497 | struct sock *sk, *result; | |
ba418fa3 TH |
498 | int score, badness, matches = 0, reuseport = 0; |
499 | u32 hash = 0; | |
5051ebd2 | 500 | |
5051ebd2 | 501 | result = NULL; |
ba418fa3 | 502 | badness = 0; |
ca065d0c | 503 | udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
5051ebd2 ED |
504 | score = compute_score2(sk, net, saddr, sport, |
505 | daddr, hnum, dif); | |
506 | if (score > badness) { | |
ba418fa3 TH |
507 | reuseport = sk->sk_reuseport; |
508 | if (reuseport) { | |
65cd8033 HFS |
509 | hash = udp_ehashfn(net, daddr, hnum, |
510 | saddr, sport); | |
ca065d0c | 511 | result = reuseport_select_sock(sk, hash, skb, |
ed0dfffd | 512 | sizeof(struct udphdr)); |
ca065d0c ED |
513 | if (result) |
514 | return result; | |
ba418fa3 TH |
515 | matches = 1; |
516 | } | |
ca065d0c ED |
517 | badness = score; |
518 | result = sk; | |
ba418fa3 TH |
519 | } else if (score == badness && reuseport) { |
520 | matches++; | |
8fc54f68 | 521 | if (reciprocal_scale(hash, matches) == 0) |
ba418fa3 TH |
522 | result = sk; |
523 | hash = next_pseudo_random32(hash); | |
5051ebd2 ED |
524 | } |
525 | } | |
5051ebd2 ED |
526 | return result; |
527 | } | |
528 | ||
db8dac20 DM |
529 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
530 | * harder than this. -DaveM | |
531 | */ | |
fce82338 | 532 | struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, |
db8dac20 | 533 | __be16 sport, __be32 daddr, __be16 dport, |
538950a1 | 534 | int dif, struct udp_table *udptable, struct sk_buff *skb) |
db8dac20 | 535 | { |
271b72c7 | 536 | struct sock *sk, *result; |
db8dac20 | 537 | unsigned short hnum = ntohs(dport); |
5051ebd2 ED |
538 | unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); |
539 | struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; | |
ba418fa3 TH |
540 | int score, badness, matches = 0, reuseport = 0; |
541 | u32 hash = 0; | |
645ca708 | 542 | |
5051ebd2 ED |
543 | if (hslot->count > 10) { |
544 | hash2 = udp4_portaddr_hash(net, daddr, hnum); | |
545 | slot2 = hash2 & udptable->mask; | |
546 | hslot2 = &udptable->hash2[slot2]; | |
547 | if (hslot->count < hslot2->count) | |
548 | goto begin; | |
549 | ||
550 | result = udp4_lib_lookup2(net, saddr, sport, | |
551 | daddr, hnum, dif, | |
1134158b | 552 | hslot2, slot2, skb); |
5051ebd2 | 553 | if (!result) { |
0eae88f3 | 554 | hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum); |
5051ebd2 ED |
555 | slot2 = hash2 & udptable->mask; |
556 | hslot2 = &udptable->hash2[slot2]; | |
557 | if (hslot->count < hslot2->count) | |
558 | goto begin; | |
559 | ||
1223c67c | 560 | result = udp4_lib_lookup2(net, saddr, sport, |
0eae88f3 | 561 | htonl(INADDR_ANY), hnum, dif, |
1134158b | 562 | hslot2, slot2, skb); |
5051ebd2 | 563 | } |
5051ebd2 ED |
564 | return result; |
565 | } | |
271b72c7 ED |
566 | begin: |
567 | result = NULL; | |
ba418fa3 | 568 | badness = 0; |
ca065d0c | 569 | sk_for_each_rcu(sk, &hslot->head) { |
645ca708 ED |
570 | score = compute_score(sk, net, saddr, hnum, sport, |
571 | daddr, dport, dif); | |
572 | if (score > badness) { | |
ba418fa3 TH |
573 | reuseport = sk->sk_reuseport; |
574 | if (reuseport) { | |
65cd8033 HFS |
575 | hash = udp_ehashfn(net, daddr, hnum, |
576 | saddr, sport); | |
ca065d0c | 577 | result = reuseport_select_sock(sk, hash, skb, |
538950a1 | 578 | sizeof(struct udphdr)); |
ca065d0c ED |
579 | if (result) |
580 | return result; | |
ba418fa3 TH |
581 | matches = 1; |
582 | } | |
ca065d0c ED |
583 | result = sk; |
584 | badness = score; | |
ba418fa3 TH |
585 | } else if (score == badness && reuseport) { |
586 | matches++; | |
8fc54f68 | 587 | if (reciprocal_scale(hash, matches) == 0) |
ba418fa3 TH |
588 | result = sk; |
589 | hash = next_pseudo_random32(hash); | |
db8dac20 DM |
590 | } |
591 | } | |
db8dac20 DM |
592 | return result; |
593 | } | |
fce82338 | 594 | EXPORT_SYMBOL_GPL(__udp4_lib_lookup); |
db8dac20 | 595 | |
607c4aaf KK |
596 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
597 | __be16 sport, __be16 dport, | |
645ca708 | 598 | struct udp_table *udptable) |
607c4aaf KK |
599 | { |
600 | const struct iphdr *iph = ip_hdr(skb); | |
601 | ||
8afdd99a ED |
602 | return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport, |
603 | iph->daddr, dport, inet_iif(skb), | |
538950a1 | 604 | udptable, skb); |
607c4aaf KK |
605 | } |
606 | ||
63058308 TH |
607 | struct sock *udp4_lib_lookup_skb(struct sk_buff *skb, |
608 | __be16 sport, __be16 dport) | |
609 | { | |
610 | const struct iphdr *iph = ip_hdr(skb); | |
611 | const struct net_device *dev = | |
612 | skb_dst(skb) ? skb_dst(skb)->dev : skb->dev; | |
613 | ||
614 | return __udp4_lib_lookup(dev_net(dev), iph->saddr, sport, | |
615 | iph->daddr, dport, inet_iif(skb), | |
616 | &udp_table, skb); | |
617 | } | |
618 | EXPORT_SYMBOL_GPL(udp4_lib_lookup_skb); | |
619 | ||
ca065d0c ED |
620 | /* Must be called under rcu_read_lock(). |
621 | * Does increment socket refcount. | |
622 | */ | |
623 | #if IS_ENABLED(CONFIG_NETFILTER_XT_MATCH_SOCKET) || \ | |
624 | IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY) | |
bcd41303 KK |
625 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
626 | __be32 daddr, __be16 dport, int dif) | |
627 | { | |
ca065d0c ED |
628 | struct sock *sk; |
629 | ||
630 | sk = __udp4_lib_lookup(net, saddr, sport, daddr, dport, | |
631 | dif, &udp_table, NULL); | |
632 | if (sk && !atomic_inc_not_zero(&sk->sk_refcnt)) | |
633 | sk = NULL; | |
634 | return sk; | |
bcd41303 KK |
635 | } |
636 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
ca065d0c | 637 | #endif |
bcd41303 | 638 | |
421b3885 SB |
639 | static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk, |
640 | __be16 loc_port, __be32 loc_addr, | |
641 | __be16 rmt_port, __be32 rmt_addr, | |
642 | int dif, unsigned short hnum) | |
643 | { | |
644 | struct inet_sock *inet = inet_sk(sk); | |
645 | ||
646 | if (!net_eq(sock_net(sk), net) || | |
647 | udp_sk(sk)->udp_port_hash != hnum || | |
648 | (inet->inet_daddr && inet->inet_daddr != rmt_addr) || | |
649 | (inet->inet_dport != rmt_port && inet->inet_dport) || | |
650 | (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) || | |
651 | ipv6_only_sock(sk) || | |
652 | (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) | |
653 | return false; | |
654 | if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif)) | |
655 | return false; | |
656 | return true; | |
657 | } | |
658 | ||
db8dac20 DM |
659 | /* |
660 | * This routine is called by the ICMP module when it gets some | |
661 | * sort of error condition. If err < 0 then the socket should | |
662 | * be closed and the error returned to the user. If err > 0 | |
663 | * it's just the icmp type << 8 | icmp code. | |
664 | * Header points to the ip header of the error packet. We move | |
665 | * on past this. Then (as it used to claim before adjustment) | |
666 | * header points to the first 8 bytes of the udp header. We need | |
667 | * to find the appropriate port. | |
668 | */ | |
669 | ||
645ca708 | 670 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
671 | { |
672 | struct inet_sock *inet; | |
b71d1d42 | 673 | const struct iphdr *iph = (const struct iphdr *)skb->data; |
c482c568 | 674 | struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2)); |
db8dac20 DM |
675 | const int type = icmp_hdr(skb)->type; |
676 | const int code = icmp_hdr(skb)->code; | |
677 | struct sock *sk; | |
678 | int harderr; | |
679 | int err; | |
fd54d716 | 680 | struct net *net = dev_net(skb->dev); |
db8dac20 | 681 | |
fd54d716 | 682 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
538950a1 CG |
683 | iph->saddr, uh->source, skb->dev->ifindex, udptable, |
684 | NULL); | |
51456b29 | 685 | if (!sk) { |
dcfc23ca | 686 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
687 | return; /* No socket for error */ |
688 | } | |
689 | ||
690 | err = 0; | |
691 | harderr = 0; | |
692 | inet = inet_sk(sk); | |
693 | ||
694 | switch (type) { | |
695 | default: | |
696 | case ICMP_TIME_EXCEEDED: | |
697 | err = EHOSTUNREACH; | |
698 | break; | |
699 | case ICMP_SOURCE_QUENCH: | |
700 | goto out; | |
701 | case ICMP_PARAMETERPROB: | |
702 | err = EPROTO; | |
703 | harderr = 1; | |
704 | break; | |
705 | case ICMP_DEST_UNREACH: | |
706 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
36393395 | 707 | ipv4_sk_update_pmtu(skb, sk, info); |
db8dac20 DM |
708 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { |
709 | err = EMSGSIZE; | |
710 | harderr = 1; | |
711 | break; | |
712 | } | |
713 | goto out; | |
714 | } | |
715 | err = EHOSTUNREACH; | |
716 | if (code <= NR_ICMP_UNREACH) { | |
717 | harderr = icmp_err_convert[code].fatal; | |
718 | err = icmp_err_convert[code].errno; | |
719 | } | |
720 | break; | |
55be7a9c DM |
721 | case ICMP_REDIRECT: |
722 | ipv4_sk_redirect(skb, sk); | |
1a462d18 | 723 | goto out; |
db8dac20 DM |
724 | } |
725 | ||
726 | /* | |
727 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
728 | * 4.1.3.3. | |
729 | */ | |
730 | if (!inet->recverr) { | |
731 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
732 | goto out; | |
b1faf566 | 733 | } else |
c482c568 | 734 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1)); |
b1faf566 | 735 | |
db8dac20 DM |
736 | sk->sk_err = err; |
737 | sk->sk_error_report(sk); | |
738 | out: | |
ca065d0c | 739 | return; |
db8dac20 DM |
740 | } |
741 | ||
742 | void udp_err(struct sk_buff *skb, u32 info) | |
743 | { | |
645ca708 | 744 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
745 | } |
746 | ||
747 | /* | |
748 | * Throw away all pending data and cancel the corking. Socket is locked. | |
749 | */ | |
36d926b9 | 750 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
751 | { |
752 | struct udp_sock *up = udp_sk(sk); | |
753 | ||
754 | if (up->pending) { | |
755 | up->len = 0; | |
756 | up->pending = 0; | |
757 | ip_flush_pending_frames(sk); | |
758 | } | |
759 | } | |
36d926b9 | 760 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
761 | |
762 | /** | |
f6b9664f | 763 | * udp4_hwcsum - handle outgoing HW checksumming |
db8dac20 DM |
764 | * @skb: sk_buff containing the filled-in UDP header |
765 | * (checksum field must be zeroed out) | |
f6b9664f HX |
766 | * @src: source IP address |
767 | * @dst: destination IP address | |
db8dac20 | 768 | */ |
c26bf4a5 | 769 | void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst) |
db8dac20 | 770 | { |
db8dac20 | 771 | struct udphdr *uh = udp_hdr(skb); |
f6b9664f HX |
772 | int offset = skb_transport_offset(skb); |
773 | int len = skb->len - offset; | |
774 | int hlen = len; | |
db8dac20 DM |
775 | __wsum csum = 0; |
776 | ||
ebbe495f | 777 | if (!skb_has_frag_list(skb)) { |
db8dac20 DM |
778 | /* |
779 | * Only one fragment on the socket. | |
780 | */ | |
781 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
782 | skb->csum_offset = offsetof(struct udphdr, check); | |
f6b9664f HX |
783 | uh->check = ~csum_tcpudp_magic(src, dst, len, |
784 | IPPROTO_UDP, 0); | |
db8dac20 | 785 | } else { |
ebbe495f WC |
786 | struct sk_buff *frags; |
787 | ||
db8dac20 DM |
788 | /* |
789 | * HW-checksum won't work as there are two or more | |
790 | * fragments on the socket so that all csums of sk_buffs | |
791 | * should be together | |
792 | */ | |
ebbe495f | 793 | skb_walk_frags(skb, frags) { |
f6b9664f HX |
794 | csum = csum_add(csum, frags->csum); |
795 | hlen -= frags->len; | |
ebbe495f | 796 | } |
db8dac20 | 797 | |
f6b9664f | 798 | csum = skb_checksum(skb, offset, hlen, csum); |
db8dac20 DM |
799 | skb->ip_summed = CHECKSUM_NONE; |
800 | ||
db8dac20 DM |
801 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); |
802 | if (uh->check == 0) | |
803 | uh->check = CSUM_MANGLED_0; | |
804 | } | |
805 | } | |
c26bf4a5 | 806 | EXPORT_SYMBOL_GPL(udp4_hwcsum); |
db8dac20 | 807 | |
af5fcba7 TH |
808 | /* Function to set UDP checksum for an IPv4 UDP packet. This is intended |
809 | * for the simple case like when setting the checksum for a UDP tunnel. | |
810 | */ | |
811 | void udp_set_csum(bool nocheck, struct sk_buff *skb, | |
812 | __be32 saddr, __be32 daddr, int len) | |
813 | { | |
814 | struct udphdr *uh = udp_hdr(skb); | |
815 | ||
179bc67f | 816 | if (nocheck) { |
af5fcba7 | 817 | uh->check = 0; |
179bc67f | 818 | } else if (skb_is_gso(skb)) { |
af5fcba7 | 819 | uh->check = ~udp_v4_check(len, saddr, daddr, 0); |
179bc67f EC |
820 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { |
821 | uh->check = 0; | |
822 | uh->check = udp_v4_check(len, saddr, daddr, lco_csum(skb)); | |
823 | if (uh->check == 0) | |
824 | uh->check = CSUM_MANGLED_0; | |
d75f1306 | 825 | } else { |
af5fcba7 TH |
826 | skb->ip_summed = CHECKSUM_PARTIAL; |
827 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
828 | skb->csum_offset = offsetof(struct udphdr, check); | |
829 | uh->check = ~udp_v4_check(len, saddr, daddr, 0); | |
af5fcba7 TH |
830 | } |
831 | } | |
832 | EXPORT_SYMBOL(udp_set_csum); | |
833 | ||
79ab0531 | 834 | static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4) |
db8dac20 | 835 | { |
f6b9664f | 836 | struct sock *sk = skb->sk; |
db8dac20 | 837 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
838 | struct udphdr *uh; |
839 | int err = 0; | |
840 | int is_udplite = IS_UDPLITE(sk); | |
f6b9664f HX |
841 | int offset = skb_transport_offset(skb); |
842 | int len = skb->len - offset; | |
db8dac20 DM |
843 | __wsum csum = 0; |
844 | ||
db8dac20 DM |
845 | /* |
846 | * Create a UDP header | |
847 | */ | |
848 | uh = udp_hdr(skb); | |
f6b9664f | 849 | uh->source = inet->inet_sport; |
79ab0531 | 850 | uh->dest = fl4->fl4_dport; |
f6b9664f | 851 | uh->len = htons(len); |
db8dac20 DM |
852 | uh->check = 0; |
853 | ||
854 | if (is_udplite) /* UDP-Lite */ | |
f6b9664f | 855 | csum = udplite_csum(skb); |
db8dac20 | 856 | |
28448b80 | 857 | else if (sk->sk_no_check_tx) { /* UDP csum disabled */ |
db8dac20 DM |
858 | |
859 | skb->ip_summed = CHECKSUM_NONE; | |
860 | goto send; | |
861 | ||
862 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
863 | ||
79ab0531 | 864 | udp4_hwcsum(skb, fl4->saddr, fl4->daddr); |
db8dac20 DM |
865 | goto send; |
866 | ||
f6b9664f HX |
867 | } else |
868 | csum = udp_csum(skb); | |
db8dac20 DM |
869 | |
870 | /* add protocol-dependent pseudo-header */ | |
79ab0531 | 871 | uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len, |
c482c568 | 872 | sk->sk_protocol, csum); |
db8dac20 DM |
873 | if (uh->check == 0) |
874 | uh->check = CSUM_MANGLED_0; | |
875 | ||
876 | send: | |
b5ec8eea | 877 | err = ip_send_skb(sock_net(sk), skb); |
6ce9e7b5 ED |
878 | if (err) { |
879 | if (err == -ENOBUFS && !inet->recverr) { | |
880 | UDP_INC_STATS_USER(sock_net(sk), | |
881 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
882 | err = 0; | |
883 | } | |
884 | } else | |
885 | UDP_INC_STATS_USER(sock_net(sk), | |
886 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
f6b9664f HX |
887 | return err; |
888 | } | |
889 | ||
890 | /* | |
891 | * Push out all pending data as one UDP datagram. Socket is locked. | |
892 | */ | |
8822b64a | 893 | int udp_push_pending_frames(struct sock *sk) |
f6b9664f HX |
894 | { |
895 | struct udp_sock *up = udp_sk(sk); | |
896 | struct inet_sock *inet = inet_sk(sk); | |
b6f21b26 | 897 | struct flowi4 *fl4 = &inet->cork.fl.u.ip4; |
f6b9664f HX |
898 | struct sk_buff *skb; |
899 | int err = 0; | |
900 | ||
77968b78 | 901 | skb = ip_finish_skb(sk, fl4); |
f6b9664f HX |
902 | if (!skb) |
903 | goto out; | |
904 | ||
79ab0531 | 905 | err = udp_send_skb(skb, fl4); |
f6b9664f | 906 | |
db8dac20 DM |
907 | out: |
908 | up->len = 0; | |
909 | up->pending = 0; | |
db8dac20 DM |
910 | return err; |
911 | } | |
8822b64a | 912 | EXPORT_SYMBOL(udp_push_pending_frames); |
db8dac20 | 913 | |
1b784140 | 914 | int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
db8dac20 DM |
915 | { |
916 | struct inet_sock *inet = inet_sk(sk); | |
917 | struct udp_sock *up = udp_sk(sk); | |
e474995f | 918 | struct flowi4 fl4_stack; |
b6f21b26 | 919 | struct flowi4 *fl4; |
db8dac20 DM |
920 | int ulen = len; |
921 | struct ipcm_cookie ipc; | |
922 | struct rtable *rt = NULL; | |
923 | int free = 0; | |
924 | int connected = 0; | |
925 | __be32 daddr, faddr, saddr; | |
926 | __be16 dport; | |
927 | u8 tos; | |
928 | int err, is_udplite = IS_UDPLITE(sk); | |
929 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
930 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
903ab86d | 931 | struct sk_buff *skb; |
f6d8bd05 | 932 | struct ip_options_data opt_copy; |
db8dac20 DM |
933 | |
934 | if (len > 0xFFFF) | |
935 | return -EMSGSIZE; | |
936 | ||
937 | /* | |
938 | * Check the flags. | |
939 | */ | |
940 | ||
c482c568 | 941 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */ |
db8dac20 DM |
942 | return -EOPNOTSUPP; |
943 | ||
944 | ipc.opt = NULL; | |
2244d07b | 945 | ipc.tx_flags = 0; |
aa661581 FF |
946 | ipc.ttl = 0; |
947 | ipc.tos = -1; | |
db8dac20 | 948 | |
903ab86d HX |
949 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
950 | ||
f5fca608 | 951 | fl4 = &inet->cork.fl.u.ip4; |
db8dac20 DM |
952 | if (up->pending) { |
953 | /* | |
954 | * There are pending frames. | |
955 | * The socket lock must be held while it's corked. | |
956 | */ | |
957 | lock_sock(sk); | |
958 | if (likely(up->pending)) { | |
959 | if (unlikely(up->pending != AF_INET)) { | |
960 | release_sock(sk); | |
961 | return -EINVAL; | |
962 | } | |
963 | goto do_append_data; | |
964 | } | |
965 | release_sock(sk); | |
966 | } | |
967 | ulen += sizeof(struct udphdr); | |
968 | ||
969 | /* | |
970 | * Get and verify the address. | |
971 | */ | |
972 | if (msg->msg_name) { | |
342dfc30 | 973 | DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); |
db8dac20 DM |
974 | if (msg->msg_namelen < sizeof(*usin)) |
975 | return -EINVAL; | |
976 | if (usin->sin_family != AF_INET) { | |
977 | if (usin->sin_family != AF_UNSPEC) | |
978 | return -EAFNOSUPPORT; | |
979 | } | |
980 | ||
981 | daddr = usin->sin_addr.s_addr; | |
982 | dport = usin->sin_port; | |
983 | if (dport == 0) | |
984 | return -EINVAL; | |
985 | } else { | |
986 | if (sk->sk_state != TCP_ESTABLISHED) | |
987 | return -EDESTADDRREQ; | |
c720c7e8 ED |
988 | daddr = inet->inet_daddr; |
989 | dport = inet->inet_dport; | |
db8dac20 DM |
990 | /* Open fast path for connected socket. |
991 | Route will not be used, if at least one option is set. | |
992 | */ | |
993 | connected = 1; | |
994 | } | |
db8dac20 | 995 | |
c14ac945 SHY |
996 | ipc.sockc.tsflags = sk->sk_tsflags; |
997 | ipc.addr = inet->inet_saddr; | |
db8dac20 | 998 | ipc.oif = sk->sk_bound_dev_if; |
bf84a010 | 999 | |
db8dac20 | 1000 | if (msg->msg_controllen) { |
24025c46 | 1001 | err = ip_cmsg_send(sk, msg, &ipc, sk->sk_family == AF_INET6); |
91948309 ED |
1002 | if (unlikely(err)) { |
1003 | kfree(ipc.opt); | |
db8dac20 | 1004 | return err; |
91948309 | 1005 | } |
db8dac20 DM |
1006 | if (ipc.opt) |
1007 | free = 1; | |
1008 | connected = 0; | |
1009 | } | |
f6d8bd05 ED |
1010 | if (!ipc.opt) { |
1011 | struct ip_options_rcu *inet_opt; | |
1012 | ||
1013 | rcu_read_lock(); | |
1014 | inet_opt = rcu_dereference(inet->inet_opt); | |
1015 | if (inet_opt) { | |
1016 | memcpy(&opt_copy, inet_opt, | |
1017 | sizeof(*inet_opt) + inet_opt->opt.optlen); | |
1018 | ipc.opt = &opt_copy.opt; | |
1019 | } | |
1020 | rcu_read_unlock(); | |
1021 | } | |
db8dac20 DM |
1022 | |
1023 | saddr = ipc.addr; | |
1024 | ipc.addr = faddr = daddr; | |
1025 | ||
c14ac945 SHY |
1026 | sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags); |
1027 | ||
f6d8bd05 | 1028 | if (ipc.opt && ipc.opt->opt.srr) { |
db8dac20 DM |
1029 | if (!daddr) |
1030 | return -EINVAL; | |
f6d8bd05 | 1031 | faddr = ipc.opt->opt.faddr; |
db8dac20 DM |
1032 | connected = 0; |
1033 | } | |
aa661581 | 1034 | tos = get_rttos(&ipc, inet); |
db8dac20 DM |
1035 | if (sock_flag(sk, SOCK_LOCALROUTE) || |
1036 | (msg->msg_flags & MSG_DONTROUTE) || | |
f6d8bd05 | 1037 | (ipc.opt && ipc.opt->opt.is_strictroute)) { |
db8dac20 DM |
1038 | tos |= RTO_ONLINK; |
1039 | connected = 0; | |
1040 | } | |
1041 | ||
1042 | if (ipv4_is_multicast(daddr)) { | |
1043 | if (!ipc.oif) | |
1044 | ipc.oif = inet->mc_index; | |
1045 | if (!saddr) | |
1046 | saddr = inet->mc_addr; | |
1047 | connected = 0; | |
76e21053 EH |
1048 | } else if (!ipc.oif) |
1049 | ipc.oif = inet->uc_index; | |
db8dac20 DM |
1050 | |
1051 | if (connected) | |
c482c568 | 1052 | rt = (struct rtable *)sk_dst_check(sk, 0); |
db8dac20 | 1053 | |
51456b29 | 1054 | if (!rt) { |
84a3aa00 | 1055 | struct net *net = sock_net(sk); |
9a24abfa | 1056 | __u8 flow_flags = inet_sk_flowi_flags(sk); |
84a3aa00 | 1057 | |
e474995f | 1058 | fl4 = &fl4_stack; |
9a24abfa | 1059 | |
e474995f | 1060 | flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos, |
c0951cbc | 1061 | RT_SCOPE_UNIVERSE, sk->sk_protocol, |
9a24abfa | 1062 | flow_flags, |
c0951cbc DM |
1063 | faddr, saddr, dport, inet->inet_sport); |
1064 | ||
b5bdacf3 DA |
1065 | if (!saddr && ipc.oif) { |
1066 | err = l3mdev_get_saddr(net, ipc.oif, fl4); | |
1067 | if (err < 0) | |
1068 | goto out; | |
1069 | } | |
8cbb512c | 1070 | |
e474995f DM |
1071 | security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); |
1072 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe DM |
1073 | if (IS_ERR(rt)) { |
1074 | err = PTR_ERR(rt); | |
06dc94b1 | 1075 | rt = NULL; |
db8dac20 | 1076 | if (err == -ENETUNREACH) |
f1d8cba6 | 1077 | IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
1078 | goto out; |
1079 | } | |
1080 | ||
1081 | err = -EACCES; | |
1082 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
1083 | !sock_flag(sk, SOCK_BROADCAST)) | |
1084 | goto out; | |
1085 | if (connected) | |
d8d1f30b | 1086 | sk_dst_set(sk, dst_clone(&rt->dst)); |
db8dac20 DM |
1087 | } |
1088 | ||
1089 | if (msg->msg_flags&MSG_CONFIRM) | |
1090 | goto do_confirm; | |
1091 | back_from_confirm: | |
1092 | ||
e474995f | 1093 | saddr = fl4->saddr; |
db8dac20 | 1094 | if (!ipc.addr) |
e474995f | 1095 | daddr = ipc.addr = fl4->daddr; |
db8dac20 | 1096 | |
903ab86d HX |
1097 | /* Lockless fast path for the non-corking case. */ |
1098 | if (!corkreq) { | |
f69e6d13 | 1099 | skb = ip_make_skb(sk, fl4, getfrag, msg, ulen, |
903ab86d HX |
1100 | sizeof(struct udphdr), &ipc, &rt, |
1101 | msg->msg_flags); | |
1102 | err = PTR_ERR(skb); | |
50c3a487 | 1103 | if (!IS_ERR_OR_NULL(skb)) |
79ab0531 | 1104 | err = udp_send_skb(skb, fl4); |
903ab86d HX |
1105 | goto out; |
1106 | } | |
1107 | ||
db8dac20 DM |
1108 | lock_sock(sk); |
1109 | if (unlikely(up->pending)) { | |
1110 | /* The socket is already corked while preparing it. */ | |
1111 | /* ... which is an evident application bug. --ANK */ | |
1112 | release_sock(sk); | |
1113 | ||
ba7a46f1 | 1114 | net_dbg_ratelimited("cork app bug 2\n"); |
db8dac20 DM |
1115 | err = -EINVAL; |
1116 | goto out; | |
1117 | } | |
1118 | /* | |
1119 | * Now cork the socket to pend data. | |
1120 | */ | |
b6f21b26 DM |
1121 | fl4 = &inet->cork.fl.u.ip4; |
1122 | fl4->daddr = daddr; | |
1123 | fl4->saddr = saddr; | |
9cce96df DM |
1124 | fl4->fl4_dport = dport; |
1125 | fl4->fl4_sport = inet->inet_sport; | |
db8dac20 DM |
1126 | up->pending = AF_INET; |
1127 | ||
1128 | do_append_data: | |
1129 | up->len += ulen; | |
f69e6d13 | 1130 | err = ip_append_data(sk, fl4, getfrag, msg, ulen, |
f5fca608 DM |
1131 | sizeof(struct udphdr), &ipc, &rt, |
1132 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | |
db8dac20 DM |
1133 | if (err) |
1134 | udp_flush_pending_frames(sk); | |
1135 | else if (!corkreq) | |
1136 | err = udp_push_pending_frames(sk); | |
1137 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
1138 | up->pending = 0; | |
1139 | release_sock(sk); | |
1140 | ||
1141 | out: | |
1142 | ip_rt_put(rt); | |
1143 | if (free) | |
1144 | kfree(ipc.opt); | |
1145 | if (!err) | |
1146 | return len; | |
1147 | /* | |
1148 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1149 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1150 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1151 | * things). We could add another new stat but at least for now that | |
1152 | * seems like overkill. | |
1153 | */ | |
1154 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
1155 | UDP_INC_STATS_USER(sock_net(sk), |
1156 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
1157 | } |
1158 | return err; | |
1159 | ||
1160 | do_confirm: | |
d8d1f30b | 1161 | dst_confirm(&rt->dst); |
db8dac20 DM |
1162 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1163 | goto back_from_confirm; | |
1164 | err = 0; | |
1165 | goto out; | |
1166 | } | |
c482c568 | 1167 | EXPORT_SYMBOL(udp_sendmsg); |
db8dac20 DM |
1168 | |
1169 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
1170 | size_t size, int flags) | |
1171 | { | |
f5fca608 | 1172 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
1173 | struct udp_sock *up = udp_sk(sk); |
1174 | int ret; | |
1175 | ||
d3f7d56a SL |
1176 | if (flags & MSG_SENDPAGE_NOTLAST) |
1177 | flags |= MSG_MORE; | |
1178 | ||
db8dac20 DM |
1179 | if (!up->pending) { |
1180 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
1181 | ||
1182 | /* Call udp_sendmsg to specify destination address which | |
1183 | * sendpage interface can't pass. | |
1184 | * This will succeed only when the socket is connected. | |
1185 | */ | |
1b784140 | 1186 | ret = udp_sendmsg(sk, &msg, 0); |
db8dac20 DM |
1187 | if (ret < 0) |
1188 | return ret; | |
1189 | } | |
1190 | ||
1191 | lock_sock(sk); | |
1192 | ||
1193 | if (unlikely(!up->pending)) { | |
1194 | release_sock(sk); | |
1195 | ||
ba7a46f1 | 1196 | net_dbg_ratelimited("udp cork app bug 3\n"); |
db8dac20 DM |
1197 | return -EINVAL; |
1198 | } | |
1199 | ||
f5fca608 DM |
1200 | ret = ip_append_page(sk, &inet->cork.fl.u.ip4, |
1201 | page, offset, size, flags); | |
db8dac20 DM |
1202 | if (ret == -EOPNOTSUPP) { |
1203 | release_sock(sk); | |
1204 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
1205 | size, flags); | |
1206 | } | |
1207 | if (ret < 0) { | |
1208 | udp_flush_pending_frames(sk); | |
1209 | goto out; | |
1210 | } | |
1211 | ||
1212 | up->len += size; | |
1213 | if (!(up->corkflag || (flags&MSG_MORE))) | |
1214 | ret = udp_push_pending_frames(sk); | |
1215 | if (!ret) | |
1216 | ret = size; | |
1217 | out: | |
1218 | release_sock(sk); | |
1219 | return ret; | |
1220 | } | |
1221 | ||
85584672 ED |
1222 | /** |
1223 | * first_packet_length - return length of first packet in receive queue | |
1224 | * @sk: socket | |
1225 | * | |
1226 | * Drops all bad checksum frames, until a valid one is found. | |
1227 | * Returns the length of found skb, or 0 if none is found. | |
1228 | */ | |
1229 | static unsigned int first_packet_length(struct sock *sk) | |
1230 | { | |
1231 | struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue; | |
1232 | struct sk_buff *skb; | |
1233 | unsigned int res; | |
1234 | ||
1235 | __skb_queue_head_init(&list_kill); | |
1236 | ||
1237 | spin_lock_bh(&rcvq->lock); | |
1238 | while ((skb = skb_peek(rcvq)) != NULL && | |
1239 | udp_lib_checksum_complete(skb)) { | |
6a5dc9e5 ED |
1240 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, |
1241 | IS_UDPLITE(sk)); | |
85584672 ED |
1242 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, |
1243 | IS_UDPLITE(sk)); | |
8edf19c2 | 1244 | atomic_inc(&sk->sk_drops); |
85584672 ED |
1245 | __skb_unlink(skb, rcvq); |
1246 | __skb_queue_tail(&list_kill, skb); | |
1247 | } | |
1248 | res = skb ? skb->len : 0; | |
1249 | spin_unlock_bh(&rcvq->lock); | |
1250 | ||
1251 | if (!skb_queue_empty(&list_kill)) { | |
8a74ad60 ED |
1252 | bool slow = lock_sock_fast(sk); |
1253 | ||
85584672 ED |
1254 | __skb_queue_purge(&list_kill); |
1255 | sk_mem_reclaim_partial(sk); | |
8a74ad60 | 1256 | unlock_sock_fast(sk, slow); |
85584672 ED |
1257 | } |
1258 | return res; | |
1259 | } | |
1260 | ||
1da177e4 LT |
1261 | /* |
1262 | * IOCTL requests applicable to the UDP protocol | |
1263 | */ | |
e905a9ed | 1264 | |
1da177e4 LT |
1265 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
1266 | { | |
6516c655 SH |
1267 | switch (cmd) { |
1268 | case SIOCOUTQ: | |
1da177e4 | 1269 | { |
31e6d363 ED |
1270 | int amount = sk_wmem_alloc_get(sk); |
1271 | ||
6516c655 SH |
1272 | return put_user(amount, (int __user *)arg); |
1273 | } | |
1da177e4 | 1274 | |
6516c655 SH |
1275 | case SIOCINQ: |
1276 | { | |
85584672 | 1277 | unsigned int amount = first_packet_length(sk); |
6516c655 | 1278 | |
6516c655 SH |
1279 | return put_user(amount, (int __user *)arg); |
1280 | } | |
1da177e4 | 1281 | |
6516c655 SH |
1282 | default: |
1283 | return -ENOIOCTLCMD; | |
1da177e4 | 1284 | } |
6516c655 SH |
1285 | |
1286 | return 0; | |
1da177e4 | 1287 | } |
c482c568 | 1288 | EXPORT_SYMBOL(udp_ioctl); |
1da177e4 | 1289 | |
db8dac20 DM |
1290 | /* |
1291 | * This should be easy, if there is something there we | |
1292 | * return it, otherwise we block. | |
1293 | */ | |
1294 | ||
1b784140 YX |
1295 | int udp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock, |
1296 | int flags, int *addr_len) | |
db8dac20 DM |
1297 | { |
1298 | struct inet_sock *inet = inet_sk(sk); | |
342dfc30 | 1299 | DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); |
db8dac20 | 1300 | struct sk_buff *skb; |
59c2cdae | 1301 | unsigned int ulen, copied; |
627d2d6b | 1302 | int peeked, peeking, off; |
db8dac20 DM |
1303 | int err; |
1304 | int is_udplite = IS_UDPLITE(sk); | |
197c949e | 1305 | bool checksum_valid = false; |
8a74ad60 | 1306 | bool slow; |
db8dac20 | 1307 | |
db8dac20 | 1308 | if (flags & MSG_ERRQUEUE) |
85fbaa75 | 1309 | return ip_recv_error(sk, msg, len, addr_len); |
db8dac20 DM |
1310 | |
1311 | try_again: | |
627d2d6b | 1312 | peeking = off = sk_peek_offset(sk, flags); |
db8dac20 | 1313 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), |
3f518bf7 | 1314 | &peeked, &off, &err); |
db8dac20 | 1315 | if (!skb) |
627d2d6b | 1316 | return err; |
db8dac20 | 1317 | |
e6afc8ac | 1318 | ulen = skb->len; |
59c2cdae | 1319 | copied = len; |
627d2d6b | 1320 | if (copied > ulen - off) |
1321 | copied = ulen - off; | |
59c2cdae | 1322 | else if (copied < ulen) |
db8dac20 DM |
1323 | msg->msg_flags |= MSG_TRUNC; |
1324 | ||
1325 | /* | |
1326 | * If checksum is needed at all, try to do it while copying the | |
1327 | * data. If the data is truncated, or if we only want a partial | |
1328 | * coverage checksum (UDP-Lite), do it before the copy. | |
1329 | */ | |
1330 | ||
627d2d6b | 1331 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov || peeking) { |
197c949e ED |
1332 | checksum_valid = !udp_lib_checksum_complete(skb); |
1333 | if (!checksum_valid) | |
db8dac20 DM |
1334 | goto csum_copy_err; |
1335 | } | |
1336 | ||
197c949e | 1337 | if (checksum_valid || skb_csum_unnecessary(skb)) |
627d2d6b | 1338 | err = skb_copy_datagram_msg(skb, off, msg, copied); |
db8dac20 | 1339 | else { |
627d2d6b | 1340 | err = skb_copy_and_csum_datagram_msg(skb, off, msg); |
db8dac20 DM |
1341 | |
1342 | if (err == -EINVAL) | |
1343 | goto csum_copy_err; | |
1344 | } | |
1345 | ||
22911fc5 ED |
1346 | if (unlikely(err)) { |
1347 | trace_kfree_skb(skb, udp_recvmsg); | |
979402b1 ED |
1348 | if (!peeked) { |
1349 | atomic_inc(&sk->sk_drops); | |
1350 | UDP_INC_STATS_USER(sock_net(sk), | |
1351 | UDP_MIB_INERRORS, is_udplite); | |
1352 | } | |
627d2d6b | 1353 | skb_free_datagram_locked(sk, skb); |
1354 | return err; | |
22911fc5 | 1355 | } |
db8dac20 DM |
1356 | |
1357 | if (!peeked) | |
629ca23c PE |
1358 | UDP_INC_STATS_USER(sock_net(sk), |
1359 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 | 1360 | |
3b885787 | 1361 | sock_recv_ts_and_drops(msg, sk, skb); |
db8dac20 DM |
1362 | |
1363 | /* Copy the address. */ | |
c482c568 | 1364 | if (sin) { |
db8dac20 DM |
1365 | sin->sin_family = AF_INET; |
1366 | sin->sin_port = udp_hdr(skb)->source; | |
1367 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
1368 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
bceaa902 | 1369 | *addr_len = sizeof(*sin); |
db8dac20 DM |
1370 | } |
1371 | if (inet->cmsg_flags) | |
31c2e492 | 1372 | ip_cmsg_recv_offset(msg, skb, sizeof(struct udphdr) + off); |
db8dac20 | 1373 | |
59c2cdae | 1374 | err = copied; |
db8dac20 DM |
1375 | if (flags & MSG_TRUNC) |
1376 | err = ulen; | |
1377 | ||
627d2d6b | 1378 | __skb_free_datagram_locked(sk, skb, peeking ? -err : err); |
db8dac20 DM |
1379 | return err; |
1380 | ||
1381 | csum_copy_err: | |
8a74ad60 | 1382 | slow = lock_sock_fast(sk); |
6a5dc9e5 ED |
1383 | if (!skb_kill_datagram(sk, skb, flags)) { |
1384 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); | |
629ca23c | 1385 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
6a5dc9e5 | 1386 | } |
8a74ad60 | 1387 | unlock_sock_fast(sk, slow); |
db8dac20 | 1388 | |
beb39db5 ED |
1389 | /* starting over for a new packet, but check if we need to yield */ |
1390 | cond_resched(); | |
9cfaa8de | 1391 | msg->msg_flags &= ~MSG_TRUNC; |
db8dac20 DM |
1392 | goto try_again; |
1393 | } | |
1394 | ||
1da177e4 LT |
1395 | int udp_disconnect(struct sock *sk, int flags) |
1396 | { | |
1397 | struct inet_sock *inet = inet_sk(sk); | |
1398 | /* | |
1399 | * 1003.1g - break association. | |
1400 | */ | |
e905a9ed | 1401 | |
1da177e4 | 1402 | sk->sk_state = TCP_CLOSE; |
c720c7e8 ED |
1403 | inet->inet_daddr = 0; |
1404 | inet->inet_dport = 0; | |
bdeab991 | 1405 | sock_rps_reset_rxhash(sk); |
1da177e4 LT |
1406 | sk->sk_bound_dev_if = 0; |
1407 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
1408 | inet_reset_saddr(sk); | |
1409 | ||
1410 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
1411 | sk->sk_prot->unhash(sk); | |
c720c7e8 | 1412 | inet->inet_sport = 0; |
1da177e4 LT |
1413 | } |
1414 | sk_dst_reset(sk); | |
1415 | return 0; | |
1416 | } | |
c482c568 | 1417 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1418 | |
645ca708 ED |
1419 | void udp_lib_unhash(struct sock *sk) |
1420 | { | |
723b4610 ED |
1421 | if (sk_hashed(sk)) { |
1422 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
512615b6 ED |
1423 | struct udp_hslot *hslot, *hslot2; |
1424 | ||
1425 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1426 | udp_sk(sk)->udp_port_hash); | |
1427 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
645ca708 | 1428 | |
723b4610 | 1429 | spin_lock_bh(&hslot->lock); |
e32ea7e7 CG |
1430 | if (rcu_access_pointer(sk->sk_reuseport_cb)) |
1431 | reuseport_detach_sock(sk); | |
ca065d0c | 1432 | if (sk_del_node_init_rcu(sk)) { |
fdcc8aa9 | 1433 | hslot->count--; |
c720c7e8 | 1434 | inet_sk(sk)->inet_num = 0; |
723b4610 | 1435 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
512615b6 ED |
1436 | |
1437 | spin_lock(&hslot2->lock); | |
ca065d0c | 1438 | hlist_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); |
512615b6 ED |
1439 | hslot2->count--; |
1440 | spin_unlock(&hslot2->lock); | |
723b4610 ED |
1441 | } |
1442 | spin_unlock_bh(&hslot->lock); | |
645ca708 | 1443 | } |
645ca708 ED |
1444 | } |
1445 | EXPORT_SYMBOL(udp_lib_unhash); | |
1446 | ||
719f8358 ED |
1447 | /* |
1448 | * inet_rcv_saddr was changed, we must rehash secondary hash | |
1449 | */ | |
1450 | void udp_lib_rehash(struct sock *sk, u16 newhash) | |
1451 | { | |
1452 | if (sk_hashed(sk)) { | |
1453 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
1454 | struct udp_hslot *hslot, *hslot2, *nhslot2; | |
1455 | ||
1456 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
1457 | nhslot2 = udp_hashslot2(udptable, newhash); | |
1458 | udp_sk(sk)->udp_portaddr_hash = newhash; | |
e32ea7e7 CG |
1459 | |
1460 | if (hslot2 != nhslot2 || | |
1461 | rcu_access_pointer(sk->sk_reuseport_cb)) { | |
719f8358 ED |
1462 | hslot = udp_hashslot(udptable, sock_net(sk), |
1463 | udp_sk(sk)->udp_port_hash); | |
1464 | /* we must lock primary chain too */ | |
1465 | spin_lock_bh(&hslot->lock); | |
e32ea7e7 CG |
1466 | if (rcu_access_pointer(sk->sk_reuseport_cb)) |
1467 | reuseport_detach_sock(sk); | |
1468 | ||
1469 | if (hslot2 != nhslot2) { | |
1470 | spin_lock(&hslot2->lock); | |
ca065d0c | 1471 | hlist_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); |
e32ea7e7 CG |
1472 | hslot2->count--; |
1473 | spin_unlock(&hslot2->lock); | |
1474 | ||
1475 | spin_lock(&nhslot2->lock); | |
ca065d0c | 1476 | hlist_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, |
e32ea7e7 CG |
1477 | &nhslot2->head); |
1478 | nhslot2->count++; | |
1479 | spin_unlock(&nhslot2->lock); | |
1480 | } | |
719f8358 ED |
1481 | |
1482 | spin_unlock_bh(&hslot->lock); | |
1483 | } | |
1484 | } | |
1485 | } | |
1486 | EXPORT_SYMBOL(udp_lib_rehash); | |
1487 | ||
1488 | static void udp_v4_rehash(struct sock *sk) | |
1489 | { | |
1490 | u16 new_hash = udp4_portaddr_hash(sock_net(sk), | |
1491 | inet_sk(sk)->inet_rcv_saddr, | |
1492 | inet_sk(sk)->inet_num); | |
1493 | udp_lib_rehash(sk, new_hash); | |
1494 | } | |
1495 | ||
93821778 HX |
1496 | static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
1497 | { | |
fec5e652 | 1498 | int rc; |
766e9037 | 1499 | |
005ec974 | 1500 | if (inet_sk(sk)->inet_daddr) { |
bdeab991 | 1501 | sock_rps_save_rxhash(sk, skb); |
005ec974 | 1502 | sk_mark_napi_id(sk, skb); |
2c8c56e1 | 1503 | sk_incoming_cpu_update(sk); |
005ec974 | 1504 | } |
fec5e652 | 1505 | |
e6afc8ac | 1506 | rc = __sock_queue_rcv_skb(sk, skb); |
766e9037 ED |
1507 | if (rc < 0) { |
1508 | int is_udplite = IS_UDPLITE(sk); | |
93821778 | 1509 | |
93821778 | 1510 | /* Note that an ENOMEM error is charged twice */ |
766e9037 | 1511 | if (rc == -ENOMEM) |
93821778 HX |
1512 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1513 | is_udplite); | |
766e9037 ED |
1514 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
1515 | kfree_skb(skb); | |
296f7ea7 | 1516 | trace_udp_fail_queue_rcv_skb(rc, sk); |
766e9037 | 1517 | return -1; |
93821778 HX |
1518 | } |
1519 | ||
1520 | return 0; | |
1521 | ||
93821778 HX |
1522 | } |
1523 | ||
447167bf ED |
1524 | static struct static_key udp_encap_needed __read_mostly; |
1525 | void udp_encap_enable(void) | |
1526 | { | |
1527 | if (!static_key_enabled(&udp_encap_needed)) | |
1528 | static_key_slow_inc(&udp_encap_needed); | |
1529 | } | |
1530 | EXPORT_SYMBOL(udp_encap_enable); | |
1531 | ||
db8dac20 DM |
1532 | /* returns: |
1533 | * -1: error | |
1534 | * 0: success | |
1535 | * >0: "udp encap" protocol resubmission | |
1536 | * | |
1537 | * Note that in the success and error cases, the skb is assumed to | |
1538 | * have either been requeued or freed. | |
1539 | */ | |
c482c568 | 1540 | int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
db8dac20 DM |
1541 | { |
1542 | struct udp_sock *up = udp_sk(sk); | |
1543 | int rc; | |
1544 | int is_udplite = IS_UDPLITE(sk); | |
1545 | ||
1546 | /* | |
1547 | * Charge it to the socket, dropping if the queue is full. | |
1548 | */ | |
1549 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1550 | goto drop; | |
1551 | nf_reset(skb); | |
1552 | ||
447167bf | 1553 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
0ad92ad0 ED |
1554 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
1555 | ||
db8dac20 DM |
1556 | /* |
1557 | * This is an encapsulation socket so pass the skb to | |
1558 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1559 | * fall through and pass this up the UDP socket. | |
1560 | * up->encap_rcv() returns the following value: | |
1561 | * =0 if skb was successfully passed to the encap | |
1562 | * handler or was discarded by it. | |
1563 | * >0 if skb should be passed on to UDP. | |
1564 | * <0 if skb should be resubmitted as proto -N | |
1565 | */ | |
1566 | ||
1567 | /* if we're overly short, let UDP handle it */ | |
0ad92ad0 | 1568 | encap_rcv = ACCESS_ONCE(up->encap_rcv); |
00db4124 | 1569 | if (skb->len > sizeof(struct udphdr) && encap_rcv) { |
db8dac20 DM |
1570 | int ret; |
1571 | ||
0a80966b TH |
1572 | /* Verify checksum before giving to encap */ |
1573 | if (udp_lib_checksum_complete(skb)) | |
1574 | goto csum_error; | |
1575 | ||
0ad92ad0 | 1576 | ret = encap_rcv(sk, skb); |
db8dac20 | 1577 | if (ret <= 0) { |
0283328e PE |
1578 | UDP_INC_STATS_BH(sock_net(sk), |
1579 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
1580 | is_udplite); |
1581 | return -ret; | |
1582 | } | |
1583 | } | |
1584 | ||
1585 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1586 | } | |
1587 | ||
1588 | /* | |
1589 | * UDP-Lite specific tests, ignored on UDP sockets | |
1590 | */ | |
1591 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1592 | ||
1593 | /* | |
1594 | * MIB statistics other than incrementing the error count are | |
1595 | * disabled for the following two types of errors: these depend | |
1596 | * on the application settings, not on the functioning of the | |
1597 | * protocol stack as such. | |
1598 | * | |
1599 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1600 | * way ... to ... at least let the receiving application block | |
1601 | * delivery of packets with coverage values less than a value | |
1602 | * provided by the application." | |
1603 | */ | |
1604 | if (up->pcrlen == 0) { /* full coverage was set */ | |
ba7a46f1 JP |
1605 | net_dbg_ratelimited("UDPLite: partial coverage %d while full coverage %d requested\n", |
1606 | UDP_SKB_CB(skb)->cscov, skb->len); | |
db8dac20 DM |
1607 | goto drop; |
1608 | } | |
1609 | /* The next case involves violating the min. coverage requested | |
1610 | * by the receiver. This is subtle: if receiver wants x and x is | |
1611 | * greater than the buffersize/MTU then receiver will complain | |
1612 | * that it wants x while sender emits packets of smaller size y. | |
1613 | * Therefore the above ...()->partial_cov statement is essential. | |
1614 | */ | |
1615 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
ba7a46f1 JP |
1616 | net_dbg_ratelimited("UDPLite: coverage %d too small, need min %d\n", |
1617 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
db8dac20 DM |
1618 | goto drop; |
1619 | } | |
1620 | } | |
1621 | ||
e6afc8ac | 1622 | if (rcu_access_pointer(sk->sk_filter)) { |
1623 | if (udp_lib_checksum_complete(skb)) | |
1624 | goto csum_error; | |
1625 | if (sk_filter(sk, skb)) | |
1626 | goto drop; | |
1627 | } | |
db8dac20 | 1628 | |
e6afc8ac | 1629 | udp_csum_pull_header(skb); |
274f482d | 1630 | if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) { |
3e215c8d JL |
1631 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1632 | is_udplite); | |
c377411f | 1633 | goto drop; |
3e215c8d | 1634 | } |
c377411f | 1635 | |
93821778 | 1636 | rc = 0; |
db8dac20 | 1637 | |
fbf8866d | 1638 | ipv4_pktinfo_prepare(sk, skb); |
93821778 HX |
1639 | bh_lock_sock(sk); |
1640 | if (!sock_owned_by_user(sk)) | |
1641 | rc = __udp_queue_rcv_skb(sk, skb); | |
f545a38f | 1642 | else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) { |
55349790 ZY |
1643 | bh_unlock_sock(sk); |
1644 | goto drop; | |
1645 | } | |
93821778 HX |
1646 | bh_unlock_sock(sk); |
1647 | ||
1648 | return rc; | |
db8dac20 | 1649 | |
6a5dc9e5 ED |
1650 | csum_error: |
1651 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); | |
db8dac20 | 1652 | drop: |
0283328e | 1653 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8edf19c2 | 1654 | atomic_inc(&sk->sk_drops); |
db8dac20 DM |
1655 | kfree_skb(skb); |
1656 | return -1; | |
1657 | } | |
1658 | ||
97502231 | 1659 | /* For TCP sockets, sk_rx_dst is protected by socket lock |
e47eb5df | 1660 | * For UDP, we use xchg() to guard against concurrent changes. |
97502231 ED |
1661 | */ |
1662 | static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst) | |
421b3885 | 1663 | { |
97502231 ED |
1664 | struct dst_entry *old; |
1665 | ||
e47eb5df ED |
1666 | dst_hold(dst); |
1667 | old = xchg(&sk->sk_rx_dst, dst); | |
1668 | dst_release(old); | |
421b3885 SB |
1669 | } |
1670 | ||
db8dac20 DM |
1671 | /* |
1672 | * Multicasts and broadcasts go to each listener. | |
1673 | * | |
1240d137 | 1674 | * Note: called only from the BH handler context. |
db8dac20 | 1675 | */ |
e3163493 | 1676 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1677 | struct udphdr *uh, |
1678 | __be32 saddr, __be32 daddr, | |
36cbb245 RJ |
1679 | struct udp_table *udptable, |
1680 | int proto) | |
db8dac20 | 1681 | { |
ca065d0c | 1682 | struct sock *sk, *first = NULL; |
5cf3d461 DH |
1683 | unsigned short hnum = ntohs(uh->dest); |
1684 | struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum); | |
2dc41cff | 1685 | unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10); |
ca065d0c ED |
1686 | unsigned int offset = offsetof(typeof(*sk), sk_node); |
1687 | int dif = skb->dev->ifindex; | |
1688 | struct hlist_node *node; | |
1689 | struct sk_buff *nskb; | |
2dc41cff DH |
1690 | |
1691 | if (use_hash2) { | |
1692 | hash2_any = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum) & | |
1693 | udp_table.mask; | |
1694 | hash2 = udp4_portaddr_hash(net, daddr, hnum) & udp_table.mask; | |
1695 | start_lookup: | |
1696 | hslot = &udp_table.hash2[hash2]; | |
1697 | offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node); | |
1698 | } | |
db8dac20 | 1699 | |
ca065d0c ED |
1700 | sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) { |
1701 | if (!__udp_is_mcast_sock(net, sk, uh->dest, daddr, | |
1702 | uh->source, saddr, dif, hnum)) | |
1703 | continue; | |
1704 | ||
1705 | if (!first) { | |
1706 | first = sk; | |
1707 | continue; | |
1240d137 | 1708 | } |
ca065d0c | 1709 | nskb = skb_clone(skb, GFP_ATOMIC); |
1240d137 | 1710 | |
ca065d0c ED |
1711 | if (unlikely(!nskb)) { |
1712 | atomic_inc(&sk->sk_drops); | |
1713 | UDP_INC_STATS_BH(net, UDP_MIB_RCVBUFERRORS, | |
1714 | IS_UDPLITE(sk)); | |
1715 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, | |
1716 | IS_UDPLITE(sk)); | |
1717 | continue; | |
1718 | } | |
1719 | if (udp_queue_rcv_skb(sk, nskb) > 0) | |
1720 | consume_skb(nskb); | |
1721 | } | |
1240d137 | 1722 | |
2dc41cff DH |
1723 | /* Also lookup *:port if we are using hash2 and haven't done so yet. */ |
1724 | if (use_hash2 && hash2 != hash2_any) { | |
1725 | hash2 = hash2_any; | |
1726 | goto start_lookup; | |
1727 | } | |
1728 | ||
ca065d0c ED |
1729 | if (first) { |
1730 | if (udp_queue_rcv_skb(first, skb) > 0) | |
1731 | consume_skb(skb); | |
1240d137 | 1732 | } else { |
ca065d0c ED |
1733 | kfree_skb(skb); |
1734 | UDP_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI, | |
1735 | proto == IPPROTO_UDPLITE); | |
1240d137 | 1736 | } |
db8dac20 DM |
1737 | return 0; |
1738 | } | |
1739 | ||
1740 | /* Initialize UDP checksum. If exited with zero value (success), | |
1741 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1742 | * Otherwise, csum completion requires chacksumming packet body, | |
1743 | * including udp header and folding it to skb->csum. | |
1744 | */ | |
1745 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1746 | int proto) | |
1747 | { | |
db8dac20 DM |
1748 | int err; |
1749 | ||
1750 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1751 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1752 | ||
1753 | if (proto == IPPROTO_UDPLITE) { | |
1754 | err = udplite_checksum_init(skb, uh); | |
1755 | if (err) | |
1756 | return err; | |
1757 | } | |
1758 | ||
ed70fcfc TH |
1759 | return skb_checksum_init_zero_check(skb, proto, uh->check, |
1760 | inet_compute_pseudo); | |
db8dac20 DM |
1761 | } |
1762 | ||
1763 | /* | |
1764 | * All we need to do is get the socket, and then do a checksum. | |
1765 | */ | |
1766 | ||
645ca708 | 1767 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1768 | int proto) |
1769 | { | |
1770 | struct sock *sk; | |
7b5e56f9 | 1771 | struct udphdr *uh; |
db8dac20 | 1772 | unsigned short ulen; |
adf30907 | 1773 | struct rtable *rt = skb_rtable(skb); |
2783ef23 | 1774 | __be32 saddr, daddr; |
0283328e | 1775 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1776 | |
1777 | /* | |
1778 | * Validate the packet. | |
1779 | */ | |
1780 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1781 | goto drop; /* No space for header. */ | |
1782 | ||
7b5e56f9 | 1783 | uh = udp_hdr(skb); |
db8dac20 | 1784 | ulen = ntohs(uh->len); |
ccc2d97c BM |
1785 | saddr = ip_hdr(skb)->saddr; |
1786 | daddr = ip_hdr(skb)->daddr; | |
1787 | ||
db8dac20 DM |
1788 | if (ulen > skb->len) |
1789 | goto short_packet; | |
1790 | ||
1791 | if (proto == IPPROTO_UDP) { | |
1792 | /* UDP validates ulen. */ | |
1793 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1794 | goto short_packet; | |
1795 | uh = udp_hdr(skb); | |
1796 | } | |
1797 | ||
1798 | if (udp4_csum_init(skb, uh, proto)) | |
1799 | goto csum_error; | |
1800 | ||
8afdd99a ED |
1801 | sk = skb_steal_sock(skb); |
1802 | if (sk) { | |
97502231 | 1803 | struct dst_entry *dst = skb_dst(skb); |
421b3885 | 1804 | int ret; |
421b3885 | 1805 | |
97502231 ED |
1806 | if (unlikely(sk->sk_rx_dst != dst)) |
1807 | udp_sk_rx_dst_set(sk, dst); | |
db8dac20 | 1808 | |
421b3885 | 1809 | ret = udp_queue_rcv_skb(sk, skb); |
8afdd99a | 1810 | sock_put(sk); |
421b3885 SB |
1811 | /* a return value > 0 means to resubmit the input, but |
1812 | * it wants the return to be -protocol, or 0 | |
1813 | */ | |
1814 | if (ret > 0) | |
1815 | return -ret; | |
1816 | return 0; | |
421b3885 | 1817 | } |
db8dac20 | 1818 | |
c18450a5 FF |
1819 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) |
1820 | return __udp4_lib_mcast_deliver(net, skb, uh, | |
36cbb245 | 1821 | saddr, daddr, udptable, proto); |
c18450a5 FF |
1822 | |
1823 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); | |
00db4124 | 1824 | if (sk) { |
a5b50476 ET |
1825 | int ret; |
1826 | ||
224d019c | 1827 | if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk)) |
2abb7cdc TH |
1828 | skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check, |
1829 | inet_compute_pseudo); | |
1830 | ||
a5b50476 | 1831 | ret = udp_queue_rcv_skb(sk, skb); |
db8dac20 DM |
1832 | |
1833 | /* a return value > 0 means to resubmit the input, but | |
1834 | * it wants the return to be -protocol, or 0 | |
1835 | */ | |
1836 | if (ret > 0) | |
1837 | return -ret; | |
1838 | return 0; | |
1839 | } | |
1840 | ||
1841 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1842 | goto drop; | |
1843 | nf_reset(skb); | |
1844 | ||
1845 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1846 | if (udp_lib_checksum_complete(skb)) | |
1847 | goto csum_error; | |
1848 | ||
0283328e | 1849 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1850 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1851 | ||
1852 | /* | |
1853 | * Hmm. We got an UDP packet to a port to which we | |
1854 | * don't wanna listen. Ignore it. | |
1855 | */ | |
1856 | kfree_skb(skb); | |
1857 | return 0; | |
1858 | ||
1859 | short_packet: | |
ba7a46f1 JP |
1860 | net_dbg_ratelimited("UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n", |
1861 | proto == IPPROTO_UDPLITE ? "Lite" : "", | |
1862 | &saddr, ntohs(uh->source), | |
1863 | ulen, skb->len, | |
1864 | &daddr, ntohs(uh->dest)); | |
db8dac20 DM |
1865 | goto drop; |
1866 | ||
1867 | csum_error: | |
1868 | /* | |
1869 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1870 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1871 | */ | |
ba7a46f1 JP |
1872 | net_dbg_ratelimited("UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n", |
1873 | proto == IPPROTO_UDPLITE ? "Lite" : "", | |
1874 | &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest), | |
1875 | ulen); | |
6a5dc9e5 | 1876 | UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 | 1877 | drop: |
0283328e | 1878 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1879 | kfree_skb(skb); |
1880 | return 0; | |
1881 | } | |
1882 | ||
421b3885 SB |
1883 | /* We can only early demux multicast if there is a single matching socket. |
1884 | * If more than one socket found returns NULL | |
1885 | */ | |
1886 | static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net, | |
1887 | __be16 loc_port, __be32 loc_addr, | |
1888 | __be16 rmt_port, __be32 rmt_addr, | |
1889 | int dif) | |
1890 | { | |
1891 | struct sock *sk, *result; | |
421b3885 | 1892 | unsigned short hnum = ntohs(loc_port); |
ca065d0c | 1893 | unsigned int slot = udp_hashfn(net, hnum, udp_table.mask); |
421b3885 SB |
1894 | struct udp_hslot *hslot = &udp_table.hash[slot]; |
1895 | ||
63c6f81c ED |
1896 | /* Do not bother scanning a too big list */ |
1897 | if (hslot->count > 10) | |
1898 | return NULL; | |
1899 | ||
421b3885 | 1900 | result = NULL; |
ca065d0c ED |
1901 | sk_for_each_rcu(sk, &hslot->head) { |
1902 | if (__udp_is_mcast_sock(net, sk, loc_port, loc_addr, | |
1903 | rmt_port, rmt_addr, dif, hnum)) { | |
1904 | if (result) | |
1905 | return NULL; | |
421b3885 | 1906 | result = sk; |
421b3885 SB |
1907 | } |
1908 | } | |
ca065d0c | 1909 | |
421b3885 SB |
1910 | return result; |
1911 | } | |
1912 | ||
1913 | /* For unicast we should only early demux connected sockets or we can | |
1914 | * break forwarding setups. The chains here can be long so only check | |
1915 | * if the first socket is an exact match and if not move on. | |
1916 | */ | |
1917 | static struct sock *__udp4_lib_demux_lookup(struct net *net, | |
1918 | __be16 loc_port, __be32 loc_addr, | |
1919 | __be16 rmt_port, __be32 rmt_addr, | |
1920 | int dif) | |
1921 | { | |
421b3885 SB |
1922 | unsigned short hnum = ntohs(loc_port); |
1923 | unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum); | |
1924 | unsigned int slot2 = hash2 & udp_table.mask; | |
1925 | struct udp_hslot *hslot2 = &udp_table.hash2[slot2]; | |
c7228317 | 1926 | INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr); |
421b3885 | 1927 | const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum); |
ca065d0c | 1928 | struct sock *sk; |
421b3885 | 1929 | |
ca065d0c ED |
1930 | udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) { |
1931 | if (INET_MATCH(sk, net, acookie, rmt_addr, | |
1932 | loc_addr, ports, dif)) | |
1933 | return sk; | |
421b3885 SB |
1934 | /* Only check first socket in chain */ |
1935 | break; | |
1936 | } | |
ca065d0c | 1937 | return NULL; |
421b3885 SB |
1938 | } |
1939 | ||
1940 | void udp_v4_early_demux(struct sk_buff *skb) | |
1941 | { | |
610438b7 ED |
1942 | struct net *net = dev_net(skb->dev); |
1943 | const struct iphdr *iph; | |
1944 | const struct udphdr *uh; | |
ca065d0c | 1945 | struct sock *sk = NULL; |
421b3885 | 1946 | struct dst_entry *dst; |
421b3885 | 1947 | int dif = skb->dev->ifindex; |
6e540309 | 1948 | int ours; |
421b3885 SB |
1949 | |
1950 | /* validate the packet */ | |
1951 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr))) | |
1952 | return; | |
1953 | ||
610438b7 ED |
1954 | iph = ip_hdr(skb); |
1955 | uh = udp_hdr(skb); | |
1956 | ||
421b3885 | 1957 | if (skb->pkt_type == PACKET_BROADCAST || |
6e540309 SB |
1958 | skb->pkt_type == PACKET_MULTICAST) { |
1959 | struct in_device *in_dev = __in_dev_get_rcu(skb->dev); | |
1960 | ||
1961 | if (!in_dev) | |
1962 | return; | |
1963 | ||
ad0ea198 PA |
1964 | /* we are supposed to accept bcast packets */ |
1965 | if (skb->pkt_type == PACKET_MULTICAST) { | |
1966 | ours = ip_check_mc_rcu(in_dev, iph->daddr, iph->saddr, | |
1967 | iph->protocol); | |
1968 | if (!ours) | |
1969 | return; | |
1970 | } | |
1971 | ||
421b3885 SB |
1972 | sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr, |
1973 | uh->source, iph->saddr, dif); | |
6e540309 | 1974 | } else if (skb->pkt_type == PACKET_HOST) { |
421b3885 SB |
1975 | sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr, |
1976 | uh->source, iph->saddr, dif); | |
6e540309 | 1977 | } |
421b3885 | 1978 | |
ca065d0c | 1979 | if (!sk || !atomic_inc_not_zero_hint(&sk->sk_refcnt, 2)) |
421b3885 SB |
1980 | return; |
1981 | ||
1982 | skb->sk = sk; | |
82eabd9e | 1983 | skb->destructor = sock_efree; |
10e2eb87 | 1984 | dst = READ_ONCE(sk->sk_rx_dst); |
421b3885 SB |
1985 | |
1986 | if (dst) | |
1987 | dst = dst_check(dst, 0); | |
10e2eb87 ED |
1988 | if (dst) { |
1989 | /* DST_NOCACHE can not be used without taking a reference */ | |
1990 | if (dst->flags & DST_NOCACHE) { | |
1991 | if (likely(atomic_inc_not_zero(&dst->__refcnt))) | |
1992 | skb_dst_set(skb, dst); | |
1993 | } else { | |
1994 | skb_dst_set_noref(skb, dst); | |
1995 | } | |
1996 | } | |
421b3885 SB |
1997 | } |
1998 | ||
db8dac20 DM |
1999 | int udp_rcv(struct sk_buff *skb) |
2000 | { | |
645ca708 | 2001 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
2002 | } |
2003 | ||
7d06b2e0 | 2004 | void udp_destroy_sock(struct sock *sk) |
db8dac20 | 2005 | { |
44046a59 | 2006 | struct udp_sock *up = udp_sk(sk); |
8a74ad60 | 2007 | bool slow = lock_sock_fast(sk); |
db8dac20 | 2008 | udp_flush_pending_frames(sk); |
8a74ad60 | 2009 | unlock_sock_fast(sk, slow); |
44046a59 TP |
2010 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
2011 | void (*encap_destroy)(struct sock *sk); | |
2012 | encap_destroy = ACCESS_ONCE(up->encap_destroy); | |
2013 | if (encap_destroy) | |
2014 | encap_destroy(sk); | |
2015 | } | |
db8dac20 DM |
2016 | } |
2017 | ||
1da177e4 LT |
2018 | /* |
2019 | * Socket option code for UDP | |
2020 | */ | |
4c0a6cb0 | 2021 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 2022 | char __user *optval, unsigned int optlen, |
4c0a6cb0 | 2023 | int (*push_pending_frames)(struct sock *)) |
1da177e4 LT |
2024 | { |
2025 | struct udp_sock *up = udp_sk(sk); | |
1c19448c | 2026 | int val, valbool; |
1da177e4 | 2027 | int err = 0; |
b2bf1e26 | 2028 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 2029 | |
c482c568 | 2030 | if (optlen < sizeof(int)) |
1da177e4 LT |
2031 | return -EINVAL; |
2032 | ||
2033 | if (get_user(val, (int __user *)optval)) | |
2034 | return -EFAULT; | |
2035 | ||
1c19448c TH |
2036 | valbool = val ? 1 : 0; |
2037 | ||
6516c655 | 2038 | switch (optname) { |
1da177e4 LT |
2039 | case UDP_CORK: |
2040 | if (val != 0) { | |
2041 | up->corkflag = 1; | |
2042 | } else { | |
2043 | up->corkflag = 0; | |
2044 | lock_sock(sk); | |
4243cdc2 | 2045 | push_pending_frames(sk); |
1da177e4 LT |
2046 | release_sock(sk); |
2047 | } | |
2048 | break; | |
e905a9ed | 2049 | |
1da177e4 LT |
2050 | case UDP_ENCAP: |
2051 | switch (val) { | |
2052 | case 0: | |
2053 | case UDP_ENCAP_ESPINUDP: | |
2054 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
2055 | up->encap_rcv = xfrm4_udp_encap_rcv; |
2056 | /* FALLTHROUGH */ | |
342f0234 | 2057 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 | 2058 | up->encap_type = val; |
447167bf | 2059 | udp_encap_enable(); |
1da177e4 LT |
2060 | break; |
2061 | default: | |
2062 | err = -ENOPROTOOPT; | |
2063 | break; | |
2064 | } | |
2065 | break; | |
2066 | ||
1c19448c TH |
2067 | case UDP_NO_CHECK6_TX: |
2068 | up->no_check6_tx = valbool; | |
2069 | break; | |
2070 | ||
2071 | case UDP_NO_CHECK6_RX: | |
2072 | up->no_check6_rx = valbool; | |
2073 | break; | |
2074 | ||
ba4e58ec GR |
2075 | /* |
2076 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
2077 | */ | |
2078 | /* The sender sets actual checksum coverage length via this option. | |
2079 | * The case coverage > packet length is handled by send module. */ | |
2080 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 2081 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
2082 | return -ENOPROTOOPT; |
2083 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
2084 | val = 8; | |
4be929be AD |
2085 | else if (val > USHRT_MAX) |
2086 | val = USHRT_MAX; | |
ba4e58ec GR |
2087 | up->pcslen = val; |
2088 | up->pcflag |= UDPLITE_SEND_CC; | |
2089 | break; | |
2090 | ||
e905a9ed YH |
2091 | /* The receiver specifies a minimum checksum coverage value. To make |
2092 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
2093 | * used, this again means full checksum coverage. */ |
2094 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 2095 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
2096 | return -ENOPROTOOPT; |
2097 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
2098 | val = 8; | |
4be929be AD |
2099 | else if (val > USHRT_MAX) |
2100 | val = USHRT_MAX; | |
ba4e58ec GR |
2101 | up->pcrlen = val; |
2102 | up->pcflag |= UDPLITE_RECV_CC; | |
2103 | break; | |
2104 | ||
1da177e4 LT |
2105 | default: |
2106 | err = -ENOPROTOOPT; | |
2107 | break; | |
6516c655 | 2108 | } |
1da177e4 LT |
2109 | |
2110 | return err; | |
2111 | } | |
c482c568 | 2112 | EXPORT_SYMBOL(udp_lib_setsockopt); |
1da177e4 | 2113 | |
db8dac20 | 2114 | int udp_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 2115 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
2116 | { |
2117 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2118 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
2119 | udp_push_pending_frames); | |
2120 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
2121 | } | |
2122 | ||
2123 | #ifdef CONFIG_COMPAT | |
2124 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 2125 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
2126 | { |
2127 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2128 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
2129 | udp_push_pending_frames); | |
2130 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
2131 | } | |
2132 | #endif | |
2133 | ||
4c0a6cb0 GR |
2134 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
2135 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
2136 | { |
2137 | struct udp_sock *up = udp_sk(sk); | |
2138 | int val, len; | |
2139 | ||
c482c568 | 2140 | if (get_user(len, optlen)) |
1da177e4 LT |
2141 | return -EFAULT; |
2142 | ||
2143 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 2144 | |
6516c655 | 2145 | if (len < 0) |
1da177e4 LT |
2146 | return -EINVAL; |
2147 | ||
6516c655 | 2148 | switch (optname) { |
1da177e4 LT |
2149 | case UDP_CORK: |
2150 | val = up->corkflag; | |
2151 | break; | |
2152 | ||
2153 | case UDP_ENCAP: | |
2154 | val = up->encap_type; | |
2155 | break; | |
2156 | ||
1c19448c TH |
2157 | case UDP_NO_CHECK6_TX: |
2158 | val = up->no_check6_tx; | |
2159 | break; | |
2160 | ||
2161 | case UDP_NO_CHECK6_RX: | |
2162 | val = up->no_check6_rx; | |
2163 | break; | |
2164 | ||
ba4e58ec GR |
2165 | /* The following two cannot be changed on UDP sockets, the return is |
2166 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
2167 | case UDPLITE_SEND_CSCOV: | |
2168 | val = up->pcslen; | |
2169 | break; | |
2170 | ||
2171 | case UDPLITE_RECV_CSCOV: | |
2172 | val = up->pcrlen; | |
2173 | break; | |
2174 | ||
1da177e4 LT |
2175 | default: |
2176 | return -ENOPROTOOPT; | |
6516c655 | 2177 | } |
1da177e4 | 2178 | |
6516c655 | 2179 | if (put_user(len, optlen)) |
e905a9ed | 2180 | return -EFAULT; |
c482c568 | 2181 | if (copy_to_user(optval, &val, len)) |
1da177e4 | 2182 | return -EFAULT; |
e905a9ed | 2183 | return 0; |
1da177e4 | 2184 | } |
c482c568 | 2185 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1da177e4 | 2186 | |
db8dac20 DM |
2187 | int udp_getsockopt(struct sock *sk, int level, int optname, |
2188 | char __user *optval, int __user *optlen) | |
2189 | { | |
2190 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2191 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
2192 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
2193 | } | |
2194 | ||
2195 | #ifdef CONFIG_COMPAT | |
2196 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
2197 | char __user *optval, int __user *optlen) | |
2198 | { | |
2199 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2200 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
2201 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
2202 | } | |
2203 | #endif | |
1da177e4 LT |
2204 | /** |
2205 | * udp_poll - wait for a UDP event. | |
2206 | * @file - file struct | |
2207 | * @sock - socket | |
2208 | * @wait - poll table | |
2209 | * | |
e905a9ed | 2210 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
2211 | * blocking sockets. If application is using a blocking fd |
2212 | * and a packet with checksum error is in the queue; | |
2213 | * then it could get return from select indicating data available | |
2214 | * but then block when reading it. Add special case code | |
2215 | * to work around these arguably broken applications. | |
2216 | */ | |
2217 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
2218 | { | |
2219 | unsigned int mask = datagram_poll(file, sock, wait); | |
2220 | struct sock *sk = sock->sk; | |
ba4e58ec | 2221 | |
c3f1dbaf DM |
2222 | sock_rps_record_flow(sk); |
2223 | ||
1da177e4 | 2224 | /* Check for false positives due to checksum errors */ |
85584672 ED |
2225 | if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) && |
2226 | !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk)) | |
2227 | mask &= ~(POLLIN | POLLRDNORM); | |
1da177e4 LT |
2228 | |
2229 | return mask; | |
e905a9ed | 2230 | |
1da177e4 | 2231 | } |
c482c568 | 2232 | EXPORT_SYMBOL(udp_poll); |
1da177e4 | 2233 | |
db8dac20 DM |
2234 | struct proto udp_prot = { |
2235 | .name = "UDP", | |
2236 | .owner = THIS_MODULE, | |
2237 | .close = udp_lib_close, | |
2238 | .connect = ip4_datagram_connect, | |
2239 | .disconnect = udp_disconnect, | |
2240 | .ioctl = udp_ioctl, | |
2241 | .destroy = udp_destroy_sock, | |
2242 | .setsockopt = udp_setsockopt, | |
2243 | .getsockopt = udp_getsockopt, | |
2244 | .sendmsg = udp_sendmsg, | |
2245 | .recvmsg = udp_recvmsg, | |
2246 | .sendpage = udp_sendpage, | |
93821778 | 2247 | .backlog_rcv = __udp_queue_rcv_skb, |
8141ed9f | 2248 | .release_cb = ip4_datagram_release_cb, |
db8dac20 DM |
2249 | .hash = udp_lib_hash, |
2250 | .unhash = udp_lib_unhash, | |
719f8358 | 2251 | .rehash = udp_v4_rehash, |
db8dac20 DM |
2252 | .get_port = udp_v4_get_port, |
2253 | .memory_allocated = &udp_memory_allocated, | |
2254 | .sysctl_mem = sysctl_udp_mem, | |
2255 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
2256 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
2257 | .obj_size = sizeof(struct udp_sock), | |
271b72c7 | 2258 | .slab_flags = SLAB_DESTROY_BY_RCU, |
645ca708 | 2259 | .h.udp_table = &udp_table, |
db8dac20 DM |
2260 | #ifdef CONFIG_COMPAT |
2261 | .compat_setsockopt = compat_udp_setsockopt, | |
2262 | .compat_getsockopt = compat_udp_getsockopt, | |
2263 | #endif | |
fcbdf09d | 2264 | .clear_sk = sk_prot_clear_portaddr_nulls, |
db8dac20 | 2265 | }; |
c482c568 | 2266 | EXPORT_SYMBOL(udp_prot); |
1da177e4 LT |
2267 | |
2268 | /* ------------------------------------------------------------------------ */ | |
2269 | #ifdef CONFIG_PROC_FS | |
2270 | ||
645ca708 | 2271 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
2272 | { |
2273 | struct sock *sk; | |
2274 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2275 | struct net *net = seq_file_net(seq); |
1da177e4 | 2276 | |
f86dcc5a ED |
2277 | for (state->bucket = start; state->bucket <= state->udp_table->mask; |
2278 | ++state->bucket) { | |
645ca708 | 2279 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
f86dcc5a | 2280 | |
ca065d0c | 2281 | if (hlist_empty(&hslot->head)) |
f86dcc5a ED |
2282 | continue; |
2283 | ||
645ca708 | 2284 | spin_lock_bh(&hslot->lock); |
ca065d0c | 2285 | sk_for_each(sk, &hslot->head) { |
878628fb | 2286 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 2287 | continue; |
1da177e4 LT |
2288 | if (sk->sk_family == state->family) |
2289 | goto found; | |
2290 | } | |
645ca708 | 2291 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
2292 | } |
2293 | sk = NULL; | |
2294 | found: | |
2295 | return sk; | |
2296 | } | |
2297 | ||
2298 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
2299 | { | |
2300 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2301 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2302 | |
2303 | do { | |
ca065d0c | 2304 | sk = sk_next(sk); |
878628fb | 2305 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 2306 | |
645ca708 | 2307 | if (!sk) { |
f86dcc5a | 2308 | if (state->bucket <= state->udp_table->mask) |
30842f29 | 2309 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
645ca708 | 2310 | return udp_get_first(seq, state->bucket + 1); |
1da177e4 LT |
2311 | } |
2312 | return sk; | |
2313 | } | |
2314 | ||
2315 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
2316 | { | |
645ca708 | 2317 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
2318 | |
2319 | if (sk) | |
6516c655 | 2320 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
2321 | --pos; |
2322 | return pos ? NULL : sk; | |
2323 | } | |
2324 | ||
2325 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
2326 | { | |
30842f29 | 2327 | struct udp_iter_state *state = seq->private; |
f86dcc5a | 2328 | state->bucket = MAX_UDP_PORTS; |
30842f29 | 2329 | |
b50660f1 | 2330 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
2331 | } |
2332 | ||
2333 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2334 | { | |
2335 | struct sock *sk; | |
2336 | ||
b50660f1 | 2337 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2338 | sk = udp_get_idx(seq, 0); |
2339 | else | |
2340 | sk = udp_get_next(seq, v); | |
2341 | ||
2342 | ++*pos; | |
2343 | return sk; | |
2344 | } | |
2345 | ||
2346 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
2347 | { | |
645ca708 ED |
2348 | struct udp_iter_state *state = seq->private; |
2349 | ||
f86dcc5a | 2350 | if (state->bucket <= state->udp_table->mask) |
645ca708 | 2351 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
1da177e4 LT |
2352 | } |
2353 | ||
73cb88ec | 2354 | int udp_seq_open(struct inode *inode, struct file *file) |
1da177e4 | 2355 | { |
d9dda78b | 2356 | struct udp_seq_afinfo *afinfo = PDE_DATA(inode); |
a2be75c1 DL |
2357 | struct udp_iter_state *s; |
2358 | int err; | |
a91275ef | 2359 | |
a2be75c1 DL |
2360 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2361 | sizeof(struct udp_iter_state)); | |
2362 | if (err < 0) | |
2363 | return err; | |
a91275ef | 2364 | |
a2be75c1 | 2365 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2366 | s->family = afinfo->family; |
645ca708 | 2367 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 2368 | return err; |
a91275ef | 2369 | } |
73cb88ec | 2370 | EXPORT_SYMBOL(udp_seq_open); |
a91275ef | 2371 | |
1da177e4 | 2372 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 2373 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
2374 | { |
2375 | struct proc_dir_entry *p; | |
2376 | int rc = 0; | |
2377 | ||
dda61925 DL |
2378 | afinfo->seq_ops.start = udp_seq_start; |
2379 | afinfo->seq_ops.next = udp_seq_next; | |
2380 | afinfo->seq_ops.stop = udp_seq_stop; | |
2381 | ||
84841c3c | 2382 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2383 | afinfo->seq_fops, afinfo); |
84841c3c | 2384 | if (!p) |
1da177e4 LT |
2385 | rc = -ENOMEM; |
2386 | return rc; | |
2387 | } | |
c482c568 | 2388 | EXPORT_SYMBOL(udp_proc_register); |
1da177e4 | 2389 | |
0c96d8c5 | 2390 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 2391 | { |
ece31ffd | 2392 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2393 | } |
c482c568 | 2394 | EXPORT_SYMBOL(udp_proc_unregister); |
db8dac20 DM |
2395 | |
2396 | /* ------------------------------------------------------------------------ */ | |
5e659e4c | 2397 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
652586df | 2398 | int bucket) |
db8dac20 DM |
2399 | { |
2400 | struct inet_sock *inet = inet_sk(sp); | |
c720c7e8 ED |
2401 | __be32 dest = inet->inet_daddr; |
2402 | __be32 src = inet->inet_rcv_saddr; | |
2403 | __u16 destp = ntohs(inet->inet_dport); | |
2404 | __u16 srcp = ntohs(inet->inet_sport); | |
db8dac20 | 2405 | |
f86dcc5a | 2406 | seq_printf(f, "%5d: %08X:%04X %08X:%04X" |
652586df | 2407 | " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d", |
db8dac20 | 2408 | bucket, src, srcp, dest, destp, sp->sk_state, |
31e6d363 ED |
2409 | sk_wmem_alloc_get(sp), |
2410 | sk_rmem_alloc_get(sp), | |
a7cb5a49 EB |
2411 | 0, 0L, 0, |
2412 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)), | |
2413 | 0, sock_i_ino(sp), | |
cb61cb9b | 2414 | atomic_read(&sp->sk_refcnt), sp, |
652586df | 2415 | atomic_read(&sp->sk_drops)); |
db8dac20 DM |
2416 | } |
2417 | ||
2418 | int udp4_seq_show(struct seq_file *seq, void *v) | |
2419 | { | |
652586df | 2420 | seq_setwidth(seq, 127); |
db8dac20 | 2421 | if (v == SEQ_START_TOKEN) |
652586df | 2422 | seq_puts(seq, " sl local_address rem_address st tx_queue " |
db8dac20 | 2423 | "rx_queue tr tm->when retrnsmt uid timeout " |
cb61cb9b | 2424 | "inode ref pointer drops"); |
db8dac20 | 2425 | else { |
db8dac20 DM |
2426 | struct udp_iter_state *state = seq->private; |
2427 | ||
652586df | 2428 | udp4_format_sock(v, seq, state->bucket); |
db8dac20 | 2429 | } |
652586df | 2430 | seq_pad(seq, '\n'); |
db8dac20 DM |
2431 | return 0; |
2432 | } | |
2433 | ||
73cb88ec AV |
2434 | static const struct file_operations udp_afinfo_seq_fops = { |
2435 | .owner = THIS_MODULE, | |
2436 | .open = udp_seq_open, | |
2437 | .read = seq_read, | |
2438 | .llseek = seq_lseek, | |
2439 | .release = seq_release_net | |
2440 | }; | |
2441 | ||
db8dac20 | 2442 | /* ------------------------------------------------------------------------ */ |
db8dac20 | 2443 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
2444 | .name = "udp", |
2445 | .family = AF_INET, | |
645ca708 | 2446 | .udp_table = &udp_table, |
73cb88ec | 2447 | .seq_fops = &udp_afinfo_seq_fops, |
dda61925 DL |
2448 | .seq_ops = { |
2449 | .show = udp4_seq_show, | |
2450 | }, | |
db8dac20 DM |
2451 | }; |
2452 | ||
2c8c1e72 | 2453 | static int __net_init udp4_proc_init_net(struct net *net) |
15439feb PE |
2454 | { |
2455 | return udp_proc_register(net, &udp4_seq_afinfo); | |
2456 | } | |
2457 | ||
2c8c1e72 | 2458 | static void __net_exit udp4_proc_exit_net(struct net *net) |
15439feb PE |
2459 | { |
2460 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
2461 | } | |
2462 | ||
2463 | static struct pernet_operations udp4_net_ops = { | |
2464 | .init = udp4_proc_init_net, | |
2465 | .exit = udp4_proc_exit_net, | |
2466 | }; | |
2467 | ||
db8dac20 DM |
2468 | int __init udp4_proc_init(void) |
2469 | { | |
15439feb | 2470 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
2471 | } |
2472 | ||
2473 | void udp4_proc_exit(void) | |
2474 | { | |
15439feb | 2475 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 2476 | } |
1da177e4 LT |
2477 | #endif /* CONFIG_PROC_FS */ |
2478 | ||
f86dcc5a ED |
2479 | static __initdata unsigned long uhash_entries; |
2480 | static int __init set_uhash_entries(char *str) | |
645ca708 | 2481 | { |
413c27d8 EZ |
2482 | ssize_t ret; |
2483 | ||
f86dcc5a ED |
2484 | if (!str) |
2485 | return 0; | |
413c27d8 EZ |
2486 | |
2487 | ret = kstrtoul(str, 0, &uhash_entries); | |
2488 | if (ret) | |
2489 | return 0; | |
2490 | ||
f86dcc5a ED |
2491 | if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN) |
2492 | uhash_entries = UDP_HTABLE_SIZE_MIN; | |
2493 | return 1; | |
2494 | } | |
2495 | __setup("uhash_entries=", set_uhash_entries); | |
645ca708 | 2496 | |
f86dcc5a ED |
2497 | void __init udp_table_init(struct udp_table *table, const char *name) |
2498 | { | |
2499 | unsigned int i; | |
2500 | ||
31fe62b9 TB |
2501 | table->hash = alloc_large_system_hash(name, |
2502 | 2 * sizeof(struct udp_hslot), | |
2503 | uhash_entries, | |
2504 | 21, /* one slot per 2 MB */ | |
2505 | 0, | |
2506 | &table->log, | |
2507 | &table->mask, | |
2508 | UDP_HTABLE_SIZE_MIN, | |
2509 | 64 * 1024); | |
2510 | ||
512615b6 | 2511 | table->hash2 = table->hash + (table->mask + 1); |
f86dcc5a | 2512 | for (i = 0; i <= table->mask; i++) { |
ca065d0c | 2513 | INIT_HLIST_HEAD(&table->hash[i].head); |
fdcc8aa9 | 2514 | table->hash[i].count = 0; |
645ca708 ED |
2515 | spin_lock_init(&table->hash[i].lock); |
2516 | } | |
512615b6 | 2517 | for (i = 0; i <= table->mask; i++) { |
ca065d0c | 2518 | INIT_HLIST_HEAD(&table->hash2[i].head); |
512615b6 ED |
2519 | table->hash2[i].count = 0; |
2520 | spin_lock_init(&table->hash2[i].lock); | |
2521 | } | |
645ca708 ED |
2522 | } |
2523 | ||
723b8e46 TH |
2524 | u32 udp_flow_hashrnd(void) |
2525 | { | |
2526 | static u32 hashrnd __read_mostly; | |
2527 | ||
2528 | net_get_random_once(&hashrnd, sizeof(hashrnd)); | |
2529 | ||
2530 | return hashrnd; | |
2531 | } | |
2532 | EXPORT_SYMBOL(udp_flow_hashrnd); | |
2533 | ||
95766fff HA |
2534 | void __init udp_init(void) |
2535 | { | |
f03d78db | 2536 | unsigned long limit; |
95766fff | 2537 | |
f86dcc5a | 2538 | udp_table_init(&udp_table, "UDP"); |
f03d78db | 2539 | limit = nr_free_buffer_pages() / 8; |
95766fff HA |
2540 | limit = max(limit, 128UL); |
2541 | sysctl_udp_mem[0] = limit / 4 * 3; | |
2542 | sysctl_udp_mem[1] = limit; | |
2543 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
2544 | ||
2545 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
2546 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
2547 | } |