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> |
1da177e4 LT |
93 | #include <linux/in.h> |
94 | #include <linux/errno.h> | |
95 | #include <linux/timer.h> | |
96 | #include <linux/mm.h> | |
1da177e4 | 97 | #include <linux/inet.h> |
1da177e4 | 98 | #include <linux/netdevice.h> |
5a0e3ad6 | 99 | #include <linux/slab.h> |
c752f073 | 100 | #include <net/tcp_states.h> |
1da177e4 | 101 | #include <linux/skbuff.h> |
57c67ff4 | 102 | #include <linux/netdevice.h> |
1da177e4 LT |
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" |
1da177e4 | 116 | |
f86dcc5a | 117 | struct udp_table udp_table __read_mostly; |
645ca708 | 118 | EXPORT_SYMBOL(udp_table); |
1da177e4 | 119 | |
8d987e5c | 120 | long sysctl_udp_mem[3] __read_mostly; |
95766fff | 121 | EXPORT_SYMBOL(sysctl_udp_mem); |
c482c568 ED |
122 | |
123 | int sysctl_udp_rmem_min __read_mostly; | |
95766fff | 124 | EXPORT_SYMBOL(sysctl_udp_rmem_min); |
c482c568 ED |
125 | |
126 | int sysctl_udp_wmem_min __read_mostly; | |
95766fff HA |
127 | EXPORT_SYMBOL(sysctl_udp_wmem_min); |
128 | ||
8d987e5c | 129 | atomic_long_t udp_memory_allocated; |
95766fff HA |
130 | EXPORT_SYMBOL(udp_memory_allocated); |
131 | ||
f86dcc5a ED |
132 | #define MAX_UDP_PORTS 65536 |
133 | #define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN) | |
98322f22 | 134 | |
f24d43c0 | 135 | static int udp_lib_lport_inuse(struct net *net, __u16 num, |
645ca708 | 136 | const struct udp_hslot *hslot, |
98322f22 | 137 | unsigned long *bitmap, |
f24d43c0 ED |
138 | struct sock *sk, |
139 | int (*saddr_comp)(const struct sock *sk1, | |
f86dcc5a ED |
140 | const struct sock *sk2), |
141 | unsigned int log) | |
1da177e4 | 142 | { |
f24d43c0 | 143 | struct sock *sk2; |
88ab1932 | 144 | struct hlist_nulls_node *node; |
ba418fa3 | 145 | kuid_t uid = sock_i_uid(sk); |
25030a7f | 146 | |
4243cdc2 | 147 | sk_nulls_for_each(sk2, node, &hslot->head) { |
9d4fb27d JP |
148 | if (net_eq(sock_net(sk2), net) && |
149 | sk2 != sk && | |
d4cada4a | 150 | (bitmap || udp_sk(sk2)->udp_port_hash == num) && |
9d4fb27d JP |
151 | (!sk2->sk_reuse || !sk->sk_reuse) && |
152 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
153 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 | 154 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
4243cdc2 JP |
155 | !uid_eq(uid, sock_i_uid(sk2))) && |
156 | saddr_comp(sk, sk2)) { | |
157 | if (!bitmap) | |
98322f22 | 158 | return 1; |
4243cdc2 | 159 | __set_bit(udp_sk(sk2)->udp_port_hash >> log, bitmap); |
98322f22 | 160 | } |
4243cdc2 | 161 | } |
25030a7f GR |
162 | return 0; |
163 | } | |
164 | ||
30fff923 ED |
165 | /* |
166 | * Note: we still hold spinlock of primary hash chain, so no other writer | |
167 | * can insert/delete a socket with local_port == num | |
168 | */ | |
169 | static int udp_lib_lport_inuse2(struct net *net, __u16 num, | |
4243cdc2 JP |
170 | struct udp_hslot *hslot2, |
171 | struct sock *sk, | |
172 | int (*saddr_comp)(const struct sock *sk1, | |
173 | const struct sock *sk2)) | |
30fff923 ED |
174 | { |
175 | struct sock *sk2; | |
176 | struct hlist_nulls_node *node; | |
ba418fa3 | 177 | kuid_t uid = sock_i_uid(sk); |
30fff923 ED |
178 | int res = 0; |
179 | ||
180 | spin_lock(&hslot2->lock); | |
4243cdc2 | 181 | udp_portaddr_for_each_entry(sk2, node, &hslot2->head) { |
9d4fb27d JP |
182 | if (net_eq(sock_net(sk2), net) && |
183 | sk2 != sk && | |
184 | (udp_sk(sk2)->udp_port_hash == num) && | |
185 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
186 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
187 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 | 188 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
4243cdc2 JP |
189 | !uid_eq(uid, sock_i_uid(sk2))) && |
190 | saddr_comp(sk, sk2)) { | |
30fff923 ED |
191 | res = 1; |
192 | break; | |
193 | } | |
4243cdc2 | 194 | } |
30fff923 ED |
195 | spin_unlock(&hslot2->lock); |
196 | return res; | |
197 | } | |
198 | ||
25030a7f | 199 | /** |
6ba5a3c5 | 200 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
201 | * |
202 | * @sk: socket struct in question | |
203 | * @snum: port number to look up | |
df2bc459 | 204 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25985edc | 205 | * @hash2_nulladdr: AF-dependent hash value in secondary hash chains, |
30fff923 | 206 | * with NULL address |
25030a7f | 207 | */ |
6ba5a3c5 | 208 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
4243cdc2 JP |
209 | int (*saddr_comp)(const struct sock *sk1, |
210 | const struct sock *sk2), | |
30fff923 | 211 | unsigned int hash2_nulladdr) |
25030a7f | 212 | { |
512615b6 | 213 | struct udp_hslot *hslot, *hslot2; |
645ca708 | 214 | struct udp_table *udptable = sk->sk_prot->h.udp_table; |
25030a7f | 215 | int error = 1; |
3b1e0a65 | 216 | struct net *net = sock_net(sk); |
1da177e4 | 217 | |
32c1da70 | 218 | if (!snum) { |
9088c560 | 219 | int low, high, remaining; |
95c96174 | 220 | unsigned int rand; |
98322f22 ED |
221 | unsigned short first, last; |
222 | DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN); | |
32c1da70 | 223 | |
0bbf87d8 | 224 | inet_get_local_port_range(net, &low, &high); |
a25de534 | 225 | remaining = (high - low) + 1; |
227b60f5 | 226 | |
63862b5b | 227 | rand = prandom_u32(); |
8fc54f68 | 228 | first = reciprocal_scale(rand, remaining) + low; |
98322f22 ED |
229 | /* |
230 | * force rand to be an odd multiple of UDP_HTABLE_SIZE | |
231 | */ | |
f86dcc5a | 232 | rand = (rand | 1) * (udptable->mask + 1); |
5781b235 ED |
233 | last = first + udptable->mask + 1; |
234 | do { | |
f86dcc5a | 235 | hslot = udp_hashslot(udptable, net, first); |
98322f22 | 236 | bitmap_zero(bitmap, PORTS_PER_CHAIN); |
645ca708 | 237 | spin_lock_bh(&hslot->lock); |
98322f22 | 238 | udp_lib_lport_inuse(net, snum, hslot, bitmap, sk, |
f86dcc5a | 239 | saddr_comp, udptable->log); |
98322f22 ED |
240 | |
241 | snum = first; | |
242 | /* | |
243 | * Iterate on all possible values of snum for this hash. | |
244 | * Using steps of an odd multiple of UDP_HTABLE_SIZE | |
245 | * give us randomization and full range coverage. | |
246 | */ | |
9088c560 | 247 | do { |
98322f22 | 248 | if (low <= snum && snum <= high && |
e3826f1e | 249 | !test_bit(snum >> udptable->log, bitmap) && |
122ff243 | 250 | !inet_is_local_reserved_port(net, snum)) |
98322f22 ED |
251 | goto found; |
252 | snum += rand; | |
253 | } while (snum != first); | |
254 | spin_unlock_bh(&hslot->lock); | |
5781b235 | 255 | } while (++first != last); |
98322f22 | 256 | goto fail; |
645ca708 | 257 | } else { |
f86dcc5a | 258 | hslot = udp_hashslot(udptable, net, snum); |
645ca708 | 259 | spin_lock_bh(&hslot->lock); |
30fff923 ED |
260 | if (hslot->count > 10) { |
261 | int exist; | |
262 | unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum; | |
263 | ||
264 | slot2 &= udptable->mask; | |
265 | hash2_nulladdr &= udptable->mask; | |
266 | ||
267 | hslot2 = udp_hashslot2(udptable, slot2); | |
268 | if (hslot->count < hslot2->count) | |
269 | goto scan_primary_hash; | |
270 | ||
271 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
272 | sk, saddr_comp); | |
273 | if (!exist && (hash2_nulladdr != slot2)) { | |
274 | hslot2 = udp_hashslot2(udptable, hash2_nulladdr); | |
275 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
276 | sk, saddr_comp); | |
277 | } | |
278 | if (exist) | |
279 | goto fail_unlock; | |
280 | else | |
281 | goto found; | |
282 | } | |
283 | scan_primary_hash: | |
f86dcc5a ED |
284 | if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk, |
285 | saddr_comp, 0)) | |
645ca708 ED |
286 | goto fail_unlock; |
287 | } | |
98322f22 | 288 | found: |
c720c7e8 | 289 | inet_sk(sk)->inet_num = snum; |
d4cada4a ED |
290 | udp_sk(sk)->udp_port_hash = snum; |
291 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
1da177e4 | 292 | if (sk_unhashed(sk)) { |
88ab1932 | 293 | sk_nulls_add_node_rcu(sk, &hslot->head); |
fdcc8aa9 | 294 | hslot->count++; |
c29a0bc4 | 295 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
512615b6 ED |
296 | |
297 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
298 | spin_lock(&hslot2->lock); | |
299 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
300 | &hslot2->head); | |
301 | hslot2->count++; | |
302 | spin_unlock(&hslot2->lock); | |
1da177e4 | 303 | } |
25030a7f | 304 | error = 0; |
645ca708 ED |
305 | fail_unlock: |
306 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 307 | fail: |
25030a7f GR |
308 | return error; |
309 | } | |
c482c568 | 310 | EXPORT_SYMBOL(udp_lib_get_port); |
25030a7f | 311 | |
499923c7 | 312 | static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2) |
db8dac20 DM |
313 | { |
314 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
315 | ||
c482c568 | 316 | return (!ipv6_only_sock(sk2) && |
c720c7e8 ED |
317 | (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr || |
318 | inet1->inet_rcv_saddr == inet2->inet_rcv_saddr)); | |
db8dac20 DM |
319 | } |
320 | ||
6eada011 ED |
321 | static u32 udp4_portaddr_hash(const struct net *net, __be32 saddr, |
322 | unsigned int port) | |
d4cada4a | 323 | { |
0eae88f3 | 324 | return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port; |
d4cada4a ED |
325 | } |
326 | ||
6ba5a3c5 | 327 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 328 | { |
30fff923 | 329 | unsigned int hash2_nulladdr = |
0eae88f3 | 330 | udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum); |
30fff923 ED |
331 | unsigned int hash2_partial = |
332 | udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0); | |
333 | ||
d4cada4a | 334 | /* precompute partial secondary hash */ |
30fff923 ED |
335 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
336 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr); | |
db8dac20 DM |
337 | } |
338 | ||
60c04aec JP |
339 | static inline int compute_score(struct sock *sk, struct net *net, |
340 | __be32 saddr, unsigned short hnum, __be16 sport, | |
341 | __be32 daddr, __be16 dport, int dif) | |
645ca708 | 342 | { |
60c04aec JP |
343 | int score; |
344 | struct inet_sock *inet; | |
645ca708 | 345 | |
60c04aec JP |
346 | if (!net_eq(sock_net(sk), net) || |
347 | udp_sk(sk)->udp_port_hash != hnum || | |
348 | ipv6_only_sock(sk)) | |
349 | return -1; | |
645ca708 | 350 | |
60c04aec JP |
351 | score = (sk->sk_family == PF_INET) ? 2 : 1; |
352 | inet = inet_sk(sk); | |
353 | ||
354 | if (inet->inet_rcv_saddr) { | |
355 | if (inet->inet_rcv_saddr != daddr) | |
356 | return -1; | |
357 | score += 4; | |
645ca708 | 358 | } |
60c04aec JP |
359 | |
360 | if (inet->inet_daddr) { | |
361 | if (inet->inet_daddr != saddr) | |
362 | return -1; | |
363 | score += 4; | |
364 | } | |
365 | ||
366 | if (inet->inet_dport) { | |
367 | if (inet->inet_dport != sport) | |
368 | return -1; | |
369 | score += 4; | |
370 | } | |
371 | ||
372 | if (sk->sk_bound_dev_if) { | |
373 | if (sk->sk_bound_dev_if != dif) | |
374 | return -1; | |
375 | score += 4; | |
376 | } | |
377 | ||
645ca708 ED |
378 | return score; |
379 | } | |
380 | ||
5051ebd2 ED |
381 | /* |
382 | * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num) | |
383 | */ | |
5051ebd2 ED |
384 | static inline int compute_score2(struct sock *sk, struct net *net, |
385 | __be32 saddr, __be16 sport, | |
386 | __be32 daddr, unsigned int hnum, int dif) | |
387 | { | |
60c04aec JP |
388 | int score; |
389 | struct inet_sock *inet; | |
5051ebd2 | 390 | |
60c04aec JP |
391 | if (!net_eq(sock_net(sk), net) || |
392 | ipv6_only_sock(sk)) | |
393 | return -1; | |
5051ebd2 | 394 | |
60c04aec JP |
395 | inet = inet_sk(sk); |
396 | ||
397 | if (inet->inet_rcv_saddr != daddr || | |
398 | inet->inet_num != hnum) | |
399 | return -1; | |
400 | ||
401 | score = (sk->sk_family == PF_INET) ? 2 : 1; | |
402 | ||
403 | if (inet->inet_daddr) { | |
404 | if (inet->inet_daddr != saddr) | |
5051ebd2 | 405 | return -1; |
60c04aec JP |
406 | score += 4; |
407 | } | |
408 | ||
409 | if (inet->inet_dport) { | |
410 | if (inet->inet_dport != sport) | |
5051ebd2 | 411 | return -1; |
60c04aec JP |
412 | score += 4; |
413 | } | |
5051ebd2 | 414 | |
60c04aec JP |
415 | if (sk->sk_bound_dev_if) { |
416 | if (sk->sk_bound_dev_if != dif) | |
417 | return -1; | |
418 | score += 4; | |
5051ebd2 | 419 | } |
60c04aec | 420 | |
5051ebd2 ED |
421 | return score; |
422 | } | |
423 | ||
6eada011 ED |
424 | static u32 udp_ehashfn(const struct net *net, const __be32 laddr, |
425 | const __u16 lport, const __be32 faddr, | |
426 | const __be16 fport) | |
65cd8033 | 427 | { |
1bbdceef HFS |
428 | static u32 udp_ehash_secret __read_mostly; |
429 | ||
430 | net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret)); | |
431 | ||
65cd8033 | 432 | return __inet_ehashfn(laddr, lport, faddr, fport, |
1bbdceef | 433 | udp_ehash_secret + net_hash_mix(net)); |
65cd8033 HFS |
434 | } |
435 | ||
5051ebd2 ED |
436 | /* called with read_rcu_lock() */ |
437 | static struct sock *udp4_lib_lookup2(struct net *net, | |
438 | __be32 saddr, __be16 sport, | |
439 | __be32 daddr, unsigned int hnum, int dif, | |
440 | struct udp_hslot *hslot2, unsigned int slot2) | |
441 | { | |
442 | struct sock *sk, *result; | |
443 | struct hlist_nulls_node *node; | |
ba418fa3 TH |
444 | int score, badness, matches = 0, reuseport = 0; |
445 | u32 hash = 0; | |
5051ebd2 ED |
446 | |
447 | begin: | |
448 | result = NULL; | |
ba418fa3 | 449 | badness = 0; |
5051ebd2 ED |
450 | udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { |
451 | score = compute_score2(sk, net, saddr, sport, | |
452 | daddr, hnum, dif); | |
453 | if (score > badness) { | |
454 | result = sk; | |
455 | badness = score; | |
ba418fa3 TH |
456 | reuseport = sk->sk_reuseport; |
457 | if (reuseport) { | |
65cd8033 HFS |
458 | hash = udp_ehashfn(net, daddr, hnum, |
459 | saddr, sport); | |
ba418fa3 TH |
460 | matches = 1; |
461 | } | |
462 | } else if (score == badness && reuseport) { | |
463 | matches++; | |
8fc54f68 | 464 | if (reciprocal_scale(hash, matches) == 0) |
ba418fa3 TH |
465 | result = sk; |
466 | hash = next_pseudo_random32(hash); | |
5051ebd2 ED |
467 | } |
468 | } | |
469 | /* | |
470 | * if the nulls value we got at the end of this lookup is | |
471 | * not the expected one, we must restart lookup. | |
472 | * We probably met an item that was moved to another chain. | |
473 | */ | |
474 | if (get_nulls_value(node) != slot2) | |
475 | goto begin; | |
5051ebd2 | 476 | if (result) { |
c31504dc | 477 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
5051ebd2 ED |
478 | result = NULL; |
479 | else if (unlikely(compute_score2(result, net, saddr, sport, | |
480 | daddr, hnum, dif) < badness)) { | |
481 | sock_put(result); | |
482 | goto begin; | |
483 | } | |
484 | } | |
485 | return result; | |
486 | } | |
487 | ||
db8dac20 DM |
488 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
489 | * harder than this. -DaveM | |
490 | */ | |
fce82338 | 491 | struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, |
db8dac20 | 492 | __be16 sport, __be32 daddr, __be16 dport, |
645ca708 | 493 | int dif, struct udp_table *udptable) |
db8dac20 | 494 | { |
271b72c7 | 495 | struct sock *sk, *result; |
88ab1932 | 496 | struct hlist_nulls_node *node; |
db8dac20 | 497 | unsigned short hnum = ntohs(dport); |
5051ebd2 ED |
498 | unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); |
499 | struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; | |
ba418fa3 TH |
500 | int score, badness, matches = 0, reuseport = 0; |
501 | u32 hash = 0; | |
645ca708 | 502 | |
271b72c7 | 503 | rcu_read_lock(); |
5051ebd2 ED |
504 | if (hslot->count > 10) { |
505 | hash2 = udp4_portaddr_hash(net, daddr, hnum); | |
506 | slot2 = hash2 & udptable->mask; | |
507 | hslot2 = &udptable->hash2[slot2]; | |
508 | if (hslot->count < hslot2->count) | |
509 | goto begin; | |
510 | ||
511 | result = udp4_lib_lookup2(net, saddr, sport, | |
512 | daddr, hnum, dif, | |
513 | hslot2, slot2); | |
514 | if (!result) { | |
0eae88f3 | 515 | hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum); |
5051ebd2 ED |
516 | slot2 = hash2 & udptable->mask; |
517 | hslot2 = &udptable->hash2[slot2]; | |
518 | if (hslot->count < hslot2->count) | |
519 | goto begin; | |
520 | ||
1223c67c | 521 | result = udp4_lib_lookup2(net, saddr, sport, |
0eae88f3 | 522 | htonl(INADDR_ANY), hnum, dif, |
5051ebd2 ED |
523 | hslot2, slot2); |
524 | } | |
525 | rcu_read_unlock(); | |
526 | return result; | |
527 | } | |
271b72c7 ED |
528 | begin: |
529 | result = NULL; | |
ba418fa3 | 530 | badness = 0; |
88ab1932 | 531 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { |
645ca708 ED |
532 | score = compute_score(sk, net, saddr, hnum, sport, |
533 | daddr, dport, dif); | |
534 | if (score > badness) { | |
535 | result = sk; | |
536 | badness = score; | |
ba418fa3 TH |
537 | reuseport = sk->sk_reuseport; |
538 | if (reuseport) { | |
65cd8033 HFS |
539 | hash = udp_ehashfn(net, daddr, hnum, |
540 | saddr, sport); | |
ba418fa3 TH |
541 | matches = 1; |
542 | } | |
543 | } else if (score == badness && reuseport) { | |
544 | matches++; | |
8fc54f68 | 545 | if (reciprocal_scale(hash, matches) == 0) |
ba418fa3 TH |
546 | result = sk; |
547 | hash = next_pseudo_random32(hash); | |
db8dac20 DM |
548 | } |
549 | } | |
88ab1932 ED |
550 | /* |
551 | * if the nulls value we got at the end of this lookup is | |
552 | * not the expected one, we must restart lookup. | |
553 | * We probably met an item that was moved to another chain. | |
554 | */ | |
5051ebd2 | 555 | if (get_nulls_value(node) != slot) |
88ab1932 ED |
556 | goto begin; |
557 | ||
271b72c7 | 558 | if (result) { |
c31504dc | 559 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
271b72c7 ED |
560 | result = NULL; |
561 | else if (unlikely(compute_score(result, net, saddr, hnum, sport, | |
562 | daddr, dport, dif) < badness)) { | |
563 | sock_put(result); | |
564 | goto begin; | |
565 | } | |
566 | } | |
567 | rcu_read_unlock(); | |
db8dac20 DM |
568 | return result; |
569 | } | |
fce82338 | 570 | EXPORT_SYMBOL_GPL(__udp4_lib_lookup); |
db8dac20 | 571 | |
607c4aaf KK |
572 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
573 | __be16 sport, __be16 dport, | |
645ca708 | 574 | struct udp_table *udptable) |
607c4aaf KK |
575 | { |
576 | const struct iphdr *iph = ip_hdr(skb); | |
577 | ||
8afdd99a ED |
578 | return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport, |
579 | iph->daddr, dport, inet_iif(skb), | |
580 | udptable); | |
607c4aaf KK |
581 | } |
582 | ||
bcd41303 KK |
583 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
584 | __be32 daddr, __be16 dport, int dif) | |
585 | { | |
645ca708 | 586 | return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table); |
bcd41303 KK |
587 | } |
588 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
589 | ||
421b3885 SB |
590 | static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk, |
591 | __be16 loc_port, __be32 loc_addr, | |
592 | __be16 rmt_port, __be32 rmt_addr, | |
593 | int dif, unsigned short hnum) | |
594 | { | |
595 | struct inet_sock *inet = inet_sk(sk); | |
596 | ||
597 | if (!net_eq(sock_net(sk), net) || | |
598 | udp_sk(sk)->udp_port_hash != hnum || | |
599 | (inet->inet_daddr && inet->inet_daddr != rmt_addr) || | |
600 | (inet->inet_dport != rmt_port && inet->inet_dport) || | |
601 | (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) || | |
602 | ipv6_only_sock(sk) || | |
603 | (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) | |
604 | return false; | |
605 | if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif)) | |
606 | return false; | |
607 | return true; | |
608 | } | |
609 | ||
db8dac20 DM |
610 | /* |
611 | * This routine is called by the ICMP module when it gets some | |
612 | * sort of error condition. If err < 0 then the socket should | |
613 | * be closed and the error returned to the user. If err > 0 | |
614 | * it's just the icmp type << 8 | icmp code. | |
615 | * Header points to the ip header of the error packet. We move | |
616 | * on past this. Then (as it used to claim before adjustment) | |
617 | * header points to the first 8 bytes of the udp header. We need | |
618 | * to find the appropriate port. | |
619 | */ | |
620 | ||
645ca708 | 621 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
622 | { |
623 | struct inet_sock *inet; | |
b71d1d42 | 624 | const struct iphdr *iph = (const struct iphdr *)skb->data; |
c482c568 | 625 | struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2)); |
db8dac20 DM |
626 | const int type = icmp_hdr(skb)->type; |
627 | const int code = icmp_hdr(skb)->code; | |
628 | struct sock *sk; | |
629 | int harderr; | |
630 | int err; | |
fd54d716 | 631 | struct net *net = dev_net(skb->dev); |
db8dac20 | 632 | |
fd54d716 | 633 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
db8dac20 | 634 | iph->saddr, uh->source, skb->dev->ifindex, udptable); |
51456b29 | 635 | if (!sk) { |
dcfc23ca | 636 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
637 | return; /* No socket for error */ |
638 | } | |
639 | ||
640 | err = 0; | |
641 | harderr = 0; | |
642 | inet = inet_sk(sk); | |
643 | ||
644 | switch (type) { | |
645 | default: | |
646 | case ICMP_TIME_EXCEEDED: | |
647 | err = EHOSTUNREACH; | |
648 | break; | |
649 | case ICMP_SOURCE_QUENCH: | |
650 | goto out; | |
651 | case ICMP_PARAMETERPROB: | |
652 | err = EPROTO; | |
653 | harderr = 1; | |
654 | break; | |
655 | case ICMP_DEST_UNREACH: | |
656 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
36393395 | 657 | ipv4_sk_update_pmtu(skb, sk, info); |
db8dac20 DM |
658 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { |
659 | err = EMSGSIZE; | |
660 | harderr = 1; | |
661 | break; | |
662 | } | |
663 | goto out; | |
664 | } | |
665 | err = EHOSTUNREACH; | |
666 | if (code <= NR_ICMP_UNREACH) { | |
667 | harderr = icmp_err_convert[code].fatal; | |
668 | err = icmp_err_convert[code].errno; | |
669 | } | |
670 | break; | |
55be7a9c DM |
671 | case ICMP_REDIRECT: |
672 | ipv4_sk_redirect(skb, sk); | |
1a462d18 | 673 | goto out; |
db8dac20 DM |
674 | } |
675 | ||
676 | /* | |
677 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
678 | * 4.1.3.3. | |
679 | */ | |
680 | if (!inet->recverr) { | |
681 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
682 | goto out; | |
b1faf566 | 683 | } else |
c482c568 | 684 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1)); |
b1faf566 | 685 | |
db8dac20 DM |
686 | sk->sk_err = err; |
687 | sk->sk_error_report(sk); | |
688 | out: | |
689 | sock_put(sk); | |
690 | } | |
691 | ||
692 | void udp_err(struct sk_buff *skb, u32 info) | |
693 | { | |
645ca708 | 694 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
695 | } |
696 | ||
697 | /* | |
698 | * Throw away all pending data and cancel the corking. Socket is locked. | |
699 | */ | |
36d926b9 | 700 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
701 | { |
702 | struct udp_sock *up = udp_sk(sk); | |
703 | ||
704 | if (up->pending) { | |
705 | up->len = 0; | |
706 | up->pending = 0; | |
707 | ip_flush_pending_frames(sk); | |
708 | } | |
709 | } | |
36d926b9 | 710 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
711 | |
712 | /** | |
f6b9664f | 713 | * udp4_hwcsum - handle outgoing HW checksumming |
db8dac20 DM |
714 | * @skb: sk_buff containing the filled-in UDP header |
715 | * (checksum field must be zeroed out) | |
f6b9664f HX |
716 | * @src: source IP address |
717 | * @dst: destination IP address | |
db8dac20 | 718 | */ |
c26bf4a5 | 719 | void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst) |
db8dac20 | 720 | { |
db8dac20 | 721 | struct udphdr *uh = udp_hdr(skb); |
f6b9664f HX |
722 | int offset = skb_transport_offset(skb); |
723 | int len = skb->len - offset; | |
724 | int hlen = len; | |
db8dac20 DM |
725 | __wsum csum = 0; |
726 | ||
ebbe495f | 727 | if (!skb_has_frag_list(skb)) { |
db8dac20 DM |
728 | /* |
729 | * Only one fragment on the socket. | |
730 | */ | |
731 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
732 | skb->csum_offset = offsetof(struct udphdr, check); | |
f6b9664f HX |
733 | uh->check = ~csum_tcpudp_magic(src, dst, len, |
734 | IPPROTO_UDP, 0); | |
db8dac20 | 735 | } else { |
ebbe495f WC |
736 | struct sk_buff *frags; |
737 | ||
db8dac20 DM |
738 | /* |
739 | * HW-checksum won't work as there are two or more | |
740 | * fragments on the socket so that all csums of sk_buffs | |
741 | * should be together | |
742 | */ | |
ebbe495f | 743 | skb_walk_frags(skb, frags) { |
f6b9664f HX |
744 | csum = csum_add(csum, frags->csum); |
745 | hlen -= frags->len; | |
ebbe495f | 746 | } |
db8dac20 | 747 | |
f6b9664f | 748 | csum = skb_checksum(skb, offset, hlen, csum); |
db8dac20 DM |
749 | skb->ip_summed = CHECKSUM_NONE; |
750 | ||
db8dac20 DM |
751 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); |
752 | if (uh->check == 0) | |
753 | uh->check = CSUM_MANGLED_0; | |
754 | } | |
755 | } | |
c26bf4a5 | 756 | EXPORT_SYMBOL_GPL(udp4_hwcsum); |
db8dac20 | 757 | |
af5fcba7 TH |
758 | /* Function to set UDP checksum for an IPv4 UDP packet. This is intended |
759 | * for the simple case like when setting the checksum for a UDP tunnel. | |
760 | */ | |
761 | void udp_set_csum(bool nocheck, struct sk_buff *skb, | |
762 | __be32 saddr, __be32 daddr, int len) | |
763 | { | |
764 | struct udphdr *uh = udp_hdr(skb); | |
765 | ||
766 | if (nocheck) | |
767 | uh->check = 0; | |
768 | else if (skb_is_gso(skb)) | |
769 | uh->check = ~udp_v4_check(len, saddr, daddr, 0); | |
770 | else if (skb_dst(skb) && skb_dst(skb)->dev && | |
771 | (skb_dst(skb)->dev->features & NETIF_F_V4_CSUM)) { | |
772 | ||
773 | BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL); | |
774 | ||
775 | skb->ip_summed = CHECKSUM_PARTIAL; | |
776 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
777 | skb->csum_offset = offsetof(struct udphdr, check); | |
778 | uh->check = ~udp_v4_check(len, saddr, daddr, 0); | |
779 | } else { | |
780 | __wsum csum; | |
781 | ||
782 | BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL); | |
783 | ||
784 | uh->check = 0; | |
785 | csum = skb_checksum(skb, 0, len, 0); | |
786 | uh->check = udp_v4_check(len, saddr, daddr, csum); | |
787 | if (uh->check == 0) | |
788 | uh->check = CSUM_MANGLED_0; | |
789 | ||
790 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
791 | } | |
792 | } | |
793 | EXPORT_SYMBOL(udp_set_csum); | |
794 | ||
79ab0531 | 795 | static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4) |
db8dac20 | 796 | { |
f6b9664f | 797 | struct sock *sk = skb->sk; |
db8dac20 | 798 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
799 | struct udphdr *uh; |
800 | int err = 0; | |
801 | int is_udplite = IS_UDPLITE(sk); | |
f6b9664f HX |
802 | int offset = skb_transport_offset(skb); |
803 | int len = skb->len - offset; | |
db8dac20 DM |
804 | __wsum csum = 0; |
805 | ||
db8dac20 DM |
806 | /* |
807 | * Create a UDP header | |
808 | */ | |
809 | uh = udp_hdr(skb); | |
f6b9664f | 810 | uh->source = inet->inet_sport; |
79ab0531 | 811 | uh->dest = fl4->fl4_dport; |
f6b9664f | 812 | uh->len = htons(len); |
db8dac20 DM |
813 | uh->check = 0; |
814 | ||
815 | if (is_udplite) /* UDP-Lite */ | |
f6b9664f | 816 | csum = udplite_csum(skb); |
db8dac20 | 817 | |
28448b80 | 818 | else if (sk->sk_no_check_tx) { /* UDP csum disabled */ |
db8dac20 DM |
819 | |
820 | skb->ip_summed = CHECKSUM_NONE; | |
821 | goto send; | |
822 | ||
823 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
824 | ||
79ab0531 | 825 | udp4_hwcsum(skb, fl4->saddr, fl4->daddr); |
db8dac20 DM |
826 | goto send; |
827 | ||
f6b9664f HX |
828 | } else |
829 | csum = udp_csum(skb); | |
db8dac20 DM |
830 | |
831 | /* add protocol-dependent pseudo-header */ | |
79ab0531 | 832 | uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len, |
c482c568 | 833 | sk->sk_protocol, csum); |
db8dac20 DM |
834 | if (uh->check == 0) |
835 | uh->check = CSUM_MANGLED_0; | |
836 | ||
837 | send: | |
b5ec8eea | 838 | err = ip_send_skb(sock_net(sk), skb); |
6ce9e7b5 ED |
839 | if (err) { |
840 | if (err == -ENOBUFS && !inet->recverr) { | |
841 | UDP_INC_STATS_USER(sock_net(sk), | |
842 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
843 | err = 0; | |
844 | } | |
845 | } else | |
846 | UDP_INC_STATS_USER(sock_net(sk), | |
847 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
f6b9664f HX |
848 | return err; |
849 | } | |
850 | ||
851 | /* | |
852 | * Push out all pending data as one UDP datagram. Socket is locked. | |
853 | */ | |
8822b64a | 854 | int udp_push_pending_frames(struct sock *sk) |
f6b9664f HX |
855 | { |
856 | struct udp_sock *up = udp_sk(sk); | |
857 | struct inet_sock *inet = inet_sk(sk); | |
b6f21b26 | 858 | struct flowi4 *fl4 = &inet->cork.fl.u.ip4; |
f6b9664f HX |
859 | struct sk_buff *skb; |
860 | int err = 0; | |
861 | ||
77968b78 | 862 | skb = ip_finish_skb(sk, fl4); |
f6b9664f HX |
863 | if (!skb) |
864 | goto out; | |
865 | ||
79ab0531 | 866 | err = udp_send_skb(skb, fl4); |
f6b9664f | 867 | |
db8dac20 DM |
868 | out: |
869 | up->len = 0; | |
870 | up->pending = 0; | |
db8dac20 DM |
871 | return err; |
872 | } | |
8822b64a | 873 | EXPORT_SYMBOL(udp_push_pending_frames); |
db8dac20 | 874 | |
1b784140 | 875 | int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
db8dac20 DM |
876 | { |
877 | struct inet_sock *inet = inet_sk(sk); | |
878 | struct udp_sock *up = udp_sk(sk); | |
e474995f | 879 | struct flowi4 fl4_stack; |
b6f21b26 | 880 | struct flowi4 *fl4; |
db8dac20 DM |
881 | int ulen = len; |
882 | struct ipcm_cookie ipc; | |
883 | struct rtable *rt = NULL; | |
884 | int free = 0; | |
885 | int connected = 0; | |
886 | __be32 daddr, faddr, saddr; | |
887 | __be16 dport; | |
888 | u8 tos; | |
889 | int err, is_udplite = IS_UDPLITE(sk); | |
890 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
891 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
903ab86d | 892 | struct sk_buff *skb; |
f6d8bd05 | 893 | struct ip_options_data opt_copy; |
db8dac20 DM |
894 | |
895 | if (len > 0xFFFF) | |
896 | return -EMSGSIZE; | |
897 | ||
898 | /* | |
899 | * Check the flags. | |
900 | */ | |
901 | ||
c482c568 | 902 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */ |
db8dac20 DM |
903 | return -EOPNOTSUPP; |
904 | ||
905 | ipc.opt = NULL; | |
2244d07b | 906 | ipc.tx_flags = 0; |
aa661581 FF |
907 | ipc.ttl = 0; |
908 | ipc.tos = -1; | |
db8dac20 | 909 | |
903ab86d HX |
910 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
911 | ||
f5fca608 | 912 | fl4 = &inet->cork.fl.u.ip4; |
db8dac20 DM |
913 | if (up->pending) { |
914 | /* | |
915 | * There are pending frames. | |
916 | * The socket lock must be held while it's corked. | |
917 | */ | |
918 | lock_sock(sk); | |
919 | if (likely(up->pending)) { | |
920 | if (unlikely(up->pending != AF_INET)) { | |
921 | release_sock(sk); | |
922 | return -EINVAL; | |
923 | } | |
924 | goto do_append_data; | |
925 | } | |
926 | release_sock(sk); | |
927 | } | |
928 | ulen += sizeof(struct udphdr); | |
929 | ||
930 | /* | |
931 | * Get and verify the address. | |
932 | */ | |
933 | if (msg->msg_name) { | |
342dfc30 | 934 | DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); |
db8dac20 DM |
935 | if (msg->msg_namelen < sizeof(*usin)) |
936 | return -EINVAL; | |
937 | if (usin->sin_family != AF_INET) { | |
938 | if (usin->sin_family != AF_UNSPEC) | |
939 | return -EAFNOSUPPORT; | |
940 | } | |
941 | ||
942 | daddr = usin->sin_addr.s_addr; | |
943 | dport = usin->sin_port; | |
944 | if (dport == 0) | |
945 | return -EINVAL; | |
946 | } else { | |
947 | if (sk->sk_state != TCP_ESTABLISHED) | |
948 | return -EDESTADDRREQ; | |
c720c7e8 ED |
949 | daddr = inet->inet_daddr; |
950 | dport = inet->inet_dport; | |
db8dac20 DM |
951 | /* Open fast path for connected socket. |
952 | Route will not be used, if at least one option is set. | |
953 | */ | |
954 | connected = 1; | |
955 | } | |
c720c7e8 | 956 | ipc.addr = inet->inet_saddr; |
db8dac20 DM |
957 | |
958 | ipc.oif = sk->sk_bound_dev_if; | |
bf84a010 DB |
959 | |
960 | sock_tx_timestamp(sk, &ipc.tx_flags); | |
961 | ||
db8dac20 | 962 | if (msg->msg_controllen) { |
c8e6ad08 HFS |
963 | err = ip_cmsg_send(sock_net(sk), msg, &ipc, |
964 | sk->sk_family == AF_INET6); | |
db8dac20 DM |
965 | if (err) |
966 | return err; | |
967 | if (ipc.opt) | |
968 | free = 1; | |
969 | connected = 0; | |
970 | } | |
f6d8bd05 ED |
971 | if (!ipc.opt) { |
972 | struct ip_options_rcu *inet_opt; | |
973 | ||
974 | rcu_read_lock(); | |
975 | inet_opt = rcu_dereference(inet->inet_opt); | |
976 | if (inet_opt) { | |
977 | memcpy(&opt_copy, inet_opt, | |
978 | sizeof(*inet_opt) + inet_opt->opt.optlen); | |
979 | ipc.opt = &opt_copy.opt; | |
980 | } | |
981 | rcu_read_unlock(); | |
982 | } | |
db8dac20 DM |
983 | |
984 | saddr = ipc.addr; | |
985 | ipc.addr = faddr = daddr; | |
986 | ||
f6d8bd05 | 987 | if (ipc.opt && ipc.opt->opt.srr) { |
db8dac20 DM |
988 | if (!daddr) |
989 | return -EINVAL; | |
f6d8bd05 | 990 | faddr = ipc.opt->opt.faddr; |
db8dac20 DM |
991 | connected = 0; |
992 | } | |
aa661581 | 993 | tos = get_rttos(&ipc, inet); |
db8dac20 DM |
994 | if (sock_flag(sk, SOCK_LOCALROUTE) || |
995 | (msg->msg_flags & MSG_DONTROUTE) || | |
f6d8bd05 | 996 | (ipc.opt && ipc.opt->opt.is_strictroute)) { |
db8dac20 DM |
997 | tos |= RTO_ONLINK; |
998 | connected = 0; | |
999 | } | |
1000 | ||
1001 | if (ipv4_is_multicast(daddr)) { | |
1002 | if (!ipc.oif) | |
1003 | ipc.oif = inet->mc_index; | |
1004 | if (!saddr) | |
1005 | saddr = inet->mc_addr; | |
1006 | connected = 0; | |
76e21053 EH |
1007 | } else if (!ipc.oif) |
1008 | ipc.oif = inet->uc_index; | |
db8dac20 DM |
1009 | |
1010 | if (connected) | |
c482c568 | 1011 | rt = (struct rtable *)sk_dst_check(sk, 0); |
db8dac20 | 1012 | |
51456b29 | 1013 | if (!rt) { |
84a3aa00 PE |
1014 | struct net *net = sock_net(sk); |
1015 | ||
e474995f DM |
1016 | fl4 = &fl4_stack; |
1017 | flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos, | |
c0951cbc | 1018 | RT_SCOPE_UNIVERSE, sk->sk_protocol, |
0e0d44ab | 1019 | inet_sk_flowi_flags(sk), |
c0951cbc DM |
1020 | faddr, saddr, dport, inet->inet_sport); |
1021 | ||
e474995f DM |
1022 | security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); |
1023 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe DM |
1024 | if (IS_ERR(rt)) { |
1025 | err = PTR_ERR(rt); | |
06dc94b1 | 1026 | rt = NULL; |
db8dac20 | 1027 | if (err == -ENETUNREACH) |
f1d8cba6 | 1028 | IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
1029 | goto out; |
1030 | } | |
1031 | ||
1032 | err = -EACCES; | |
1033 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
1034 | !sock_flag(sk, SOCK_BROADCAST)) | |
1035 | goto out; | |
1036 | if (connected) | |
d8d1f30b | 1037 | sk_dst_set(sk, dst_clone(&rt->dst)); |
db8dac20 DM |
1038 | } |
1039 | ||
1040 | if (msg->msg_flags&MSG_CONFIRM) | |
1041 | goto do_confirm; | |
1042 | back_from_confirm: | |
1043 | ||
e474995f | 1044 | saddr = fl4->saddr; |
db8dac20 | 1045 | if (!ipc.addr) |
e474995f | 1046 | daddr = ipc.addr = fl4->daddr; |
db8dac20 | 1047 | |
903ab86d HX |
1048 | /* Lockless fast path for the non-corking case. */ |
1049 | if (!corkreq) { | |
f69e6d13 | 1050 | skb = ip_make_skb(sk, fl4, getfrag, msg, ulen, |
903ab86d HX |
1051 | sizeof(struct udphdr), &ipc, &rt, |
1052 | msg->msg_flags); | |
1053 | err = PTR_ERR(skb); | |
50c3a487 | 1054 | if (!IS_ERR_OR_NULL(skb)) |
79ab0531 | 1055 | err = udp_send_skb(skb, fl4); |
903ab86d HX |
1056 | goto out; |
1057 | } | |
1058 | ||
db8dac20 DM |
1059 | lock_sock(sk); |
1060 | if (unlikely(up->pending)) { | |
1061 | /* The socket is already corked while preparing it. */ | |
1062 | /* ... which is an evident application bug. --ANK */ | |
1063 | release_sock(sk); | |
1064 | ||
ba7a46f1 | 1065 | net_dbg_ratelimited("cork app bug 2\n"); |
db8dac20 DM |
1066 | err = -EINVAL; |
1067 | goto out; | |
1068 | } | |
1069 | /* | |
1070 | * Now cork the socket to pend data. | |
1071 | */ | |
b6f21b26 DM |
1072 | fl4 = &inet->cork.fl.u.ip4; |
1073 | fl4->daddr = daddr; | |
1074 | fl4->saddr = saddr; | |
9cce96df DM |
1075 | fl4->fl4_dport = dport; |
1076 | fl4->fl4_sport = inet->inet_sport; | |
db8dac20 DM |
1077 | up->pending = AF_INET; |
1078 | ||
1079 | do_append_data: | |
1080 | up->len += ulen; | |
f69e6d13 | 1081 | err = ip_append_data(sk, fl4, getfrag, msg, ulen, |
f5fca608 DM |
1082 | sizeof(struct udphdr), &ipc, &rt, |
1083 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | |
db8dac20 DM |
1084 | if (err) |
1085 | udp_flush_pending_frames(sk); | |
1086 | else if (!corkreq) | |
1087 | err = udp_push_pending_frames(sk); | |
1088 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
1089 | up->pending = 0; | |
1090 | release_sock(sk); | |
1091 | ||
1092 | out: | |
1093 | ip_rt_put(rt); | |
1094 | if (free) | |
1095 | kfree(ipc.opt); | |
1096 | if (!err) | |
1097 | return len; | |
1098 | /* | |
1099 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1100 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1101 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1102 | * things). We could add another new stat but at least for now that | |
1103 | * seems like overkill. | |
1104 | */ | |
1105 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
1106 | UDP_INC_STATS_USER(sock_net(sk), |
1107 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
1108 | } |
1109 | return err; | |
1110 | ||
1111 | do_confirm: | |
d8d1f30b | 1112 | dst_confirm(&rt->dst); |
db8dac20 DM |
1113 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1114 | goto back_from_confirm; | |
1115 | err = 0; | |
1116 | goto out; | |
1117 | } | |
c482c568 | 1118 | EXPORT_SYMBOL(udp_sendmsg); |
db8dac20 DM |
1119 | |
1120 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
1121 | size_t size, int flags) | |
1122 | { | |
f5fca608 | 1123 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
1124 | struct udp_sock *up = udp_sk(sk); |
1125 | int ret; | |
1126 | ||
d3f7d56a SL |
1127 | if (flags & MSG_SENDPAGE_NOTLAST) |
1128 | flags |= MSG_MORE; | |
1129 | ||
db8dac20 DM |
1130 | if (!up->pending) { |
1131 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
1132 | ||
1133 | /* Call udp_sendmsg to specify destination address which | |
1134 | * sendpage interface can't pass. | |
1135 | * This will succeed only when the socket is connected. | |
1136 | */ | |
1b784140 | 1137 | ret = udp_sendmsg(sk, &msg, 0); |
db8dac20 DM |
1138 | if (ret < 0) |
1139 | return ret; | |
1140 | } | |
1141 | ||
1142 | lock_sock(sk); | |
1143 | ||
1144 | if (unlikely(!up->pending)) { | |
1145 | release_sock(sk); | |
1146 | ||
ba7a46f1 | 1147 | net_dbg_ratelimited("udp cork app bug 3\n"); |
db8dac20 DM |
1148 | return -EINVAL; |
1149 | } | |
1150 | ||
f5fca608 DM |
1151 | ret = ip_append_page(sk, &inet->cork.fl.u.ip4, |
1152 | page, offset, size, flags); | |
db8dac20 DM |
1153 | if (ret == -EOPNOTSUPP) { |
1154 | release_sock(sk); | |
1155 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
1156 | size, flags); | |
1157 | } | |
1158 | if (ret < 0) { | |
1159 | udp_flush_pending_frames(sk); | |
1160 | goto out; | |
1161 | } | |
1162 | ||
1163 | up->len += size; | |
1164 | if (!(up->corkflag || (flags&MSG_MORE))) | |
1165 | ret = udp_push_pending_frames(sk); | |
1166 | if (!ret) | |
1167 | ret = size; | |
1168 | out: | |
1169 | release_sock(sk); | |
1170 | return ret; | |
1171 | } | |
1172 | ||
85584672 ED |
1173 | /** |
1174 | * first_packet_length - return length of first packet in receive queue | |
1175 | * @sk: socket | |
1176 | * | |
1177 | * Drops all bad checksum frames, until a valid one is found. | |
1178 | * Returns the length of found skb, or 0 if none is found. | |
1179 | */ | |
1180 | static unsigned int first_packet_length(struct sock *sk) | |
1181 | { | |
1182 | struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue; | |
1183 | struct sk_buff *skb; | |
1184 | unsigned int res; | |
1185 | ||
1186 | __skb_queue_head_init(&list_kill); | |
1187 | ||
1188 | spin_lock_bh(&rcvq->lock); | |
1189 | while ((skb = skb_peek(rcvq)) != NULL && | |
1190 | udp_lib_checksum_complete(skb)) { | |
6a5dc9e5 ED |
1191 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, |
1192 | IS_UDPLITE(sk)); | |
85584672 ED |
1193 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, |
1194 | IS_UDPLITE(sk)); | |
8edf19c2 | 1195 | atomic_inc(&sk->sk_drops); |
85584672 ED |
1196 | __skb_unlink(skb, rcvq); |
1197 | __skb_queue_tail(&list_kill, skb); | |
1198 | } | |
1199 | res = skb ? skb->len : 0; | |
1200 | spin_unlock_bh(&rcvq->lock); | |
1201 | ||
1202 | if (!skb_queue_empty(&list_kill)) { | |
8a74ad60 ED |
1203 | bool slow = lock_sock_fast(sk); |
1204 | ||
85584672 ED |
1205 | __skb_queue_purge(&list_kill); |
1206 | sk_mem_reclaim_partial(sk); | |
8a74ad60 | 1207 | unlock_sock_fast(sk, slow); |
85584672 ED |
1208 | } |
1209 | return res; | |
1210 | } | |
1211 | ||
1da177e4 LT |
1212 | /* |
1213 | * IOCTL requests applicable to the UDP protocol | |
1214 | */ | |
e905a9ed | 1215 | |
1da177e4 LT |
1216 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
1217 | { | |
6516c655 SH |
1218 | switch (cmd) { |
1219 | case SIOCOUTQ: | |
1da177e4 | 1220 | { |
31e6d363 ED |
1221 | int amount = sk_wmem_alloc_get(sk); |
1222 | ||
6516c655 SH |
1223 | return put_user(amount, (int __user *)arg); |
1224 | } | |
1da177e4 | 1225 | |
6516c655 SH |
1226 | case SIOCINQ: |
1227 | { | |
85584672 | 1228 | unsigned int amount = first_packet_length(sk); |
6516c655 | 1229 | |
85584672 | 1230 | if (amount) |
6516c655 SH |
1231 | /* |
1232 | * We will only return the amount | |
1233 | * of this packet since that is all | |
1234 | * that will be read. | |
1235 | */ | |
85584672 ED |
1236 | amount -= sizeof(struct udphdr); |
1237 | ||
6516c655 SH |
1238 | return put_user(amount, (int __user *)arg); |
1239 | } | |
1da177e4 | 1240 | |
6516c655 SH |
1241 | default: |
1242 | return -ENOIOCTLCMD; | |
1da177e4 | 1243 | } |
6516c655 SH |
1244 | |
1245 | return 0; | |
1da177e4 | 1246 | } |
c482c568 | 1247 | EXPORT_SYMBOL(udp_ioctl); |
1da177e4 | 1248 | |
db8dac20 DM |
1249 | /* |
1250 | * This should be easy, if there is something there we | |
1251 | * return it, otherwise we block. | |
1252 | */ | |
1253 | ||
1b784140 YX |
1254 | int udp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock, |
1255 | int flags, int *addr_len) | |
db8dac20 DM |
1256 | { |
1257 | struct inet_sock *inet = inet_sk(sk); | |
342dfc30 | 1258 | DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); |
db8dac20 | 1259 | struct sk_buff *skb; |
59c2cdae | 1260 | unsigned int ulen, copied; |
3f518bf7 | 1261 | int peeked, off = 0; |
db8dac20 DM |
1262 | int err; |
1263 | int is_udplite = IS_UDPLITE(sk); | |
8a74ad60 | 1264 | bool slow; |
db8dac20 | 1265 | |
db8dac20 | 1266 | if (flags & MSG_ERRQUEUE) |
85fbaa75 | 1267 | return ip_recv_error(sk, msg, len, addr_len); |
db8dac20 DM |
1268 | |
1269 | try_again: | |
1270 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), | |
3f518bf7 | 1271 | &peeked, &off, &err); |
db8dac20 DM |
1272 | if (!skb) |
1273 | goto out; | |
1274 | ||
1275 | ulen = skb->len - sizeof(struct udphdr); | |
59c2cdae DM |
1276 | copied = len; |
1277 | if (copied > ulen) | |
1278 | copied = ulen; | |
1279 | else if (copied < ulen) | |
db8dac20 DM |
1280 | msg->msg_flags |= MSG_TRUNC; |
1281 | ||
1282 | /* | |
1283 | * If checksum is needed at all, try to do it while copying the | |
1284 | * data. If the data is truncated, or if we only want a partial | |
1285 | * coverage checksum (UDP-Lite), do it before the copy. | |
1286 | */ | |
1287 | ||
59c2cdae | 1288 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { |
db8dac20 DM |
1289 | if (udp_lib_checksum_complete(skb)) |
1290 | goto csum_copy_err; | |
1291 | } | |
1292 | ||
1293 | if (skb_csum_unnecessary(skb)) | |
51f3d02b DM |
1294 | err = skb_copy_datagram_msg(skb, sizeof(struct udphdr), |
1295 | msg, copied); | |
db8dac20 | 1296 | else { |
227158db AV |
1297 | err = skb_copy_and_csum_datagram_msg(skb, sizeof(struct udphdr), |
1298 | msg); | |
db8dac20 DM |
1299 | |
1300 | if (err == -EINVAL) | |
1301 | goto csum_copy_err; | |
1302 | } | |
1303 | ||
22911fc5 ED |
1304 | if (unlikely(err)) { |
1305 | trace_kfree_skb(skb, udp_recvmsg); | |
979402b1 ED |
1306 | if (!peeked) { |
1307 | atomic_inc(&sk->sk_drops); | |
1308 | UDP_INC_STATS_USER(sock_net(sk), | |
1309 | UDP_MIB_INERRORS, is_udplite); | |
1310 | } | |
db8dac20 | 1311 | goto out_free; |
22911fc5 | 1312 | } |
db8dac20 DM |
1313 | |
1314 | if (!peeked) | |
629ca23c PE |
1315 | UDP_INC_STATS_USER(sock_net(sk), |
1316 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 | 1317 | |
3b885787 | 1318 | sock_recv_ts_and_drops(msg, sk, skb); |
db8dac20 DM |
1319 | |
1320 | /* Copy the address. */ | |
c482c568 | 1321 | if (sin) { |
db8dac20 DM |
1322 | sin->sin_family = AF_INET; |
1323 | sin->sin_port = udp_hdr(skb)->source; | |
1324 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
1325 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
bceaa902 | 1326 | *addr_len = sizeof(*sin); |
db8dac20 DM |
1327 | } |
1328 | if (inet->cmsg_flags) | |
ad6f939a | 1329 | ip_cmsg_recv_offset(msg, skb, sizeof(struct udphdr)); |
db8dac20 | 1330 | |
59c2cdae | 1331 | err = copied; |
db8dac20 DM |
1332 | if (flags & MSG_TRUNC) |
1333 | err = ulen; | |
1334 | ||
1335 | out_free: | |
9d410c79 | 1336 | skb_free_datagram_locked(sk, skb); |
db8dac20 DM |
1337 | out: |
1338 | return err; | |
1339 | ||
1340 | csum_copy_err: | |
8a74ad60 | 1341 | slow = lock_sock_fast(sk); |
6a5dc9e5 ED |
1342 | if (!skb_kill_datagram(sk, skb, flags)) { |
1343 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); | |
629ca23c | 1344 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
6a5dc9e5 | 1345 | } |
8a74ad60 | 1346 | unlock_sock_fast(sk, slow); |
db8dac20 DM |
1347 | |
1348 | if (noblock) | |
1349 | return -EAGAIN; | |
9cfaa8de XZ |
1350 | |
1351 | /* starting over for a new packet */ | |
1352 | msg->msg_flags &= ~MSG_TRUNC; | |
db8dac20 DM |
1353 | goto try_again; |
1354 | } | |
1355 | ||
1da177e4 LT |
1356 | int udp_disconnect(struct sock *sk, int flags) |
1357 | { | |
1358 | struct inet_sock *inet = inet_sk(sk); | |
1359 | /* | |
1360 | * 1003.1g - break association. | |
1361 | */ | |
e905a9ed | 1362 | |
1da177e4 | 1363 | sk->sk_state = TCP_CLOSE; |
c720c7e8 ED |
1364 | inet->inet_daddr = 0; |
1365 | inet->inet_dport = 0; | |
bdeab991 | 1366 | sock_rps_reset_rxhash(sk); |
1da177e4 LT |
1367 | sk->sk_bound_dev_if = 0; |
1368 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
1369 | inet_reset_saddr(sk); | |
1370 | ||
1371 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
1372 | sk->sk_prot->unhash(sk); | |
c720c7e8 | 1373 | inet->inet_sport = 0; |
1da177e4 LT |
1374 | } |
1375 | sk_dst_reset(sk); | |
1376 | return 0; | |
1377 | } | |
c482c568 | 1378 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1379 | |
645ca708 ED |
1380 | void udp_lib_unhash(struct sock *sk) |
1381 | { | |
723b4610 ED |
1382 | if (sk_hashed(sk)) { |
1383 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
512615b6 ED |
1384 | struct udp_hslot *hslot, *hslot2; |
1385 | ||
1386 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1387 | udp_sk(sk)->udp_port_hash); | |
1388 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
645ca708 | 1389 | |
723b4610 ED |
1390 | spin_lock_bh(&hslot->lock); |
1391 | if (sk_nulls_del_node_init_rcu(sk)) { | |
fdcc8aa9 | 1392 | hslot->count--; |
c720c7e8 | 1393 | inet_sk(sk)->inet_num = 0; |
723b4610 | 1394 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
512615b6 ED |
1395 | |
1396 | spin_lock(&hslot2->lock); | |
1397 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1398 | hslot2->count--; | |
1399 | spin_unlock(&hslot2->lock); | |
723b4610 ED |
1400 | } |
1401 | spin_unlock_bh(&hslot->lock); | |
645ca708 | 1402 | } |
645ca708 ED |
1403 | } |
1404 | EXPORT_SYMBOL(udp_lib_unhash); | |
1405 | ||
719f8358 ED |
1406 | /* |
1407 | * inet_rcv_saddr was changed, we must rehash secondary hash | |
1408 | */ | |
1409 | void udp_lib_rehash(struct sock *sk, u16 newhash) | |
1410 | { | |
1411 | if (sk_hashed(sk)) { | |
1412 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
1413 | struct udp_hslot *hslot, *hslot2, *nhslot2; | |
1414 | ||
1415 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
1416 | nhslot2 = udp_hashslot2(udptable, newhash); | |
1417 | udp_sk(sk)->udp_portaddr_hash = newhash; | |
1418 | if (hslot2 != nhslot2) { | |
1419 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1420 | udp_sk(sk)->udp_port_hash); | |
1421 | /* we must lock primary chain too */ | |
1422 | spin_lock_bh(&hslot->lock); | |
1423 | ||
1424 | spin_lock(&hslot2->lock); | |
1425 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1426 | hslot2->count--; | |
1427 | spin_unlock(&hslot2->lock); | |
1428 | ||
1429 | spin_lock(&nhslot2->lock); | |
1430 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
1431 | &nhslot2->head); | |
1432 | nhslot2->count++; | |
1433 | spin_unlock(&nhslot2->lock); | |
1434 | ||
1435 | spin_unlock_bh(&hslot->lock); | |
1436 | } | |
1437 | } | |
1438 | } | |
1439 | EXPORT_SYMBOL(udp_lib_rehash); | |
1440 | ||
1441 | static void udp_v4_rehash(struct sock *sk) | |
1442 | { | |
1443 | u16 new_hash = udp4_portaddr_hash(sock_net(sk), | |
1444 | inet_sk(sk)->inet_rcv_saddr, | |
1445 | inet_sk(sk)->inet_num); | |
1446 | udp_lib_rehash(sk, new_hash); | |
1447 | } | |
1448 | ||
93821778 HX |
1449 | static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
1450 | { | |
fec5e652 | 1451 | int rc; |
766e9037 | 1452 | |
005ec974 | 1453 | if (inet_sk(sk)->inet_daddr) { |
bdeab991 | 1454 | sock_rps_save_rxhash(sk, skb); |
005ec974 | 1455 | sk_mark_napi_id(sk, skb); |
2c8c56e1 | 1456 | sk_incoming_cpu_update(sk); |
005ec974 | 1457 | } |
fec5e652 | 1458 | |
d826eb14 | 1459 | rc = sock_queue_rcv_skb(sk, skb); |
766e9037 ED |
1460 | if (rc < 0) { |
1461 | int is_udplite = IS_UDPLITE(sk); | |
93821778 | 1462 | |
93821778 | 1463 | /* Note that an ENOMEM error is charged twice */ |
766e9037 | 1464 | if (rc == -ENOMEM) |
93821778 HX |
1465 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1466 | is_udplite); | |
766e9037 ED |
1467 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
1468 | kfree_skb(skb); | |
296f7ea7 | 1469 | trace_udp_fail_queue_rcv_skb(rc, sk); |
766e9037 | 1470 | return -1; |
93821778 HX |
1471 | } |
1472 | ||
1473 | return 0; | |
1474 | ||
93821778 HX |
1475 | } |
1476 | ||
447167bf ED |
1477 | static struct static_key udp_encap_needed __read_mostly; |
1478 | void udp_encap_enable(void) | |
1479 | { | |
1480 | if (!static_key_enabled(&udp_encap_needed)) | |
1481 | static_key_slow_inc(&udp_encap_needed); | |
1482 | } | |
1483 | EXPORT_SYMBOL(udp_encap_enable); | |
1484 | ||
db8dac20 DM |
1485 | /* returns: |
1486 | * -1: error | |
1487 | * 0: success | |
1488 | * >0: "udp encap" protocol resubmission | |
1489 | * | |
1490 | * Note that in the success and error cases, the skb is assumed to | |
1491 | * have either been requeued or freed. | |
1492 | */ | |
c482c568 | 1493 | int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
db8dac20 DM |
1494 | { |
1495 | struct udp_sock *up = udp_sk(sk); | |
1496 | int rc; | |
1497 | int is_udplite = IS_UDPLITE(sk); | |
1498 | ||
1499 | /* | |
1500 | * Charge it to the socket, dropping if the queue is full. | |
1501 | */ | |
1502 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1503 | goto drop; | |
1504 | nf_reset(skb); | |
1505 | ||
447167bf | 1506 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
0ad92ad0 ED |
1507 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
1508 | ||
db8dac20 DM |
1509 | /* |
1510 | * This is an encapsulation socket so pass the skb to | |
1511 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1512 | * fall through and pass this up the UDP socket. | |
1513 | * up->encap_rcv() returns the following value: | |
1514 | * =0 if skb was successfully passed to the encap | |
1515 | * handler or was discarded by it. | |
1516 | * >0 if skb should be passed on to UDP. | |
1517 | * <0 if skb should be resubmitted as proto -N | |
1518 | */ | |
1519 | ||
1520 | /* if we're overly short, let UDP handle it */ | |
0ad92ad0 | 1521 | encap_rcv = ACCESS_ONCE(up->encap_rcv); |
00db4124 | 1522 | if (skb->len > sizeof(struct udphdr) && encap_rcv) { |
db8dac20 DM |
1523 | int ret; |
1524 | ||
0a80966b TH |
1525 | /* Verify checksum before giving to encap */ |
1526 | if (udp_lib_checksum_complete(skb)) | |
1527 | goto csum_error; | |
1528 | ||
0ad92ad0 | 1529 | ret = encap_rcv(sk, skb); |
db8dac20 | 1530 | if (ret <= 0) { |
0283328e PE |
1531 | UDP_INC_STATS_BH(sock_net(sk), |
1532 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
1533 | is_udplite); |
1534 | return -ret; | |
1535 | } | |
1536 | } | |
1537 | ||
1538 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1539 | } | |
1540 | ||
1541 | /* | |
1542 | * UDP-Lite specific tests, ignored on UDP sockets | |
1543 | */ | |
1544 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1545 | ||
1546 | /* | |
1547 | * MIB statistics other than incrementing the error count are | |
1548 | * disabled for the following two types of errors: these depend | |
1549 | * on the application settings, not on the functioning of the | |
1550 | * protocol stack as such. | |
1551 | * | |
1552 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1553 | * way ... to ... at least let the receiving application block | |
1554 | * delivery of packets with coverage values less than a value | |
1555 | * provided by the application." | |
1556 | */ | |
1557 | if (up->pcrlen == 0) { /* full coverage was set */ | |
ba7a46f1 JP |
1558 | net_dbg_ratelimited("UDPLite: partial coverage %d while full coverage %d requested\n", |
1559 | UDP_SKB_CB(skb)->cscov, skb->len); | |
db8dac20 DM |
1560 | goto drop; |
1561 | } | |
1562 | /* The next case involves violating the min. coverage requested | |
1563 | * by the receiver. This is subtle: if receiver wants x and x is | |
1564 | * greater than the buffersize/MTU then receiver will complain | |
1565 | * that it wants x while sender emits packets of smaller size y. | |
1566 | * Therefore the above ...()->partial_cov statement is essential. | |
1567 | */ | |
1568 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
ba7a46f1 JP |
1569 | net_dbg_ratelimited("UDPLite: coverage %d too small, need min %d\n", |
1570 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
db8dac20 DM |
1571 | goto drop; |
1572 | } | |
1573 | } | |
1574 | ||
33d480ce ED |
1575 | if (rcu_access_pointer(sk->sk_filter) && |
1576 | udp_lib_checksum_complete(skb)) | |
6a5dc9e5 | 1577 | goto csum_error; |
db8dac20 | 1578 | |
274f482d | 1579 | if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) { |
3e215c8d JL |
1580 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1581 | is_udplite); | |
c377411f | 1582 | goto drop; |
3e215c8d | 1583 | } |
c377411f | 1584 | |
93821778 | 1585 | rc = 0; |
db8dac20 | 1586 | |
fbf8866d | 1587 | ipv4_pktinfo_prepare(sk, skb); |
93821778 HX |
1588 | bh_lock_sock(sk); |
1589 | if (!sock_owned_by_user(sk)) | |
1590 | rc = __udp_queue_rcv_skb(sk, skb); | |
f545a38f | 1591 | else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) { |
55349790 ZY |
1592 | bh_unlock_sock(sk); |
1593 | goto drop; | |
1594 | } | |
93821778 HX |
1595 | bh_unlock_sock(sk); |
1596 | ||
1597 | return rc; | |
db8dac20 | 1598 | |
6a5dc9e5 ED |
1599 | csum_error: |
1600 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); | |
db8dac20 | 1601 | drop: |
0283328e | 1602 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8edf19c2 | 1603 | atomic_inc(&sk->sk_drops); |
db8dac20 DM |
1604 | kfree_skb(skb); |
1605 | return -1; | |
1606 | } | |
1607 | ||
1240d137 ED |
1608 | static void flush_stack(struct sock **stack, unsigned int count, |
1609 | struct sk_buff *skb, unsigned int final) | |
1610 | { | |
1611 | unsigned int i; | |
1612 | struct sk_buff *skb1 = NULL; | |
f6b8f32c | 1613 | struct sock *sk; |
1240d137 ED |
1614 | |
1615 | for (i = 0; i < count; i++) { | |
f6b8f32c | 1616 | sk = stack[i]; |
51456b29 | 1617 | if (likely(!skb1)) |
1240d137 ED |
1618 | skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC); |
1619 | ||
f6b8f32c ED |
1620 | if (!skb1) { |
1621 | atomic_inc(&sk->sk_drops); | |
1622 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, | |
1623 | IS_UDPLITE(sk)); | |
1624 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, | |
1625 | IS_UDPLITE(sk)); | |
1626 | } | |
1627 | ||
1628 | if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0) | |
1240d137 | 1629 | skb1 = NULL; |
2dc41cff DH |
1630 | |
1631 | sock_put(sk); | |
1240d137 ED |
1632 | } |
1633 | if (unlikely(skb1)) | |
1634 | kfree_skb(skb1); | |
1635 | } | |
1636 | ||
97502231 | 1637 | /* For TCP sockets, sk_rx_dst is protected by socket lock |
e47eb5df | 1638 | * For UDP, we use xchg() to guard against concurrent changes. |
97502231 ED |
1639 | */ |
1640 | static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst) | |
421b3885 | 1641 | { |
97502231 ED |
1642 | struct dst_entry *old; |
1643 | ||
e47eb5df ED |
1644 | dst_hold(dst); |
1645 | old = xchg(&sk->sk_rx_dst, dst); | |
1646 | dst_release(old); | |
421b3885 SB |
1647 | } |
1648 | ||
db8dac20 DM |
1649 | /* |
1650 | * Multicasts and broadcasts go to each listener. | |
1651 | * | |
1240d137 | 1652 | * Note: called only from the BH handler context. |
db8dac20 | 1653 | */ |
e3163493 | 1654 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1655 | struct udphdr *uh, |
1656 | __be32 saddr, __be32 daddr, | |
36cbb245 RJ |
1657 | struct udp_table *udptable, |
1658 | int proto) | |
db8dac20 | 1659 | { |
1240d137 | 1660 | struct sock *sk, *stack[256 / sizeof(struct sock *)]; |
5cf3d461 DH |
1661 | struct hlist_nulls_node *node; |
1662 | unsigned short hnum = ntohs(uh->dest); | |
1663 | struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum); | |
1664 | int dif = skb->dev->ifindex; | |
2dc41cff DH |
1665 | unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node); |
1666 | unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10); | |
36cbb245 | 1667 | bool inner_flushed = false; |
2dc41cff DH |
1668 | |
1669 | if (use_hash2) { | |
1670 | hash2_any = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum) & | |
1671 | udp_table.mask; | |
1672 | hash2 = udp4_portaddr_hash(net, daddr, hnum) & udp_table.mask; | |
1673 | start_lookup: | |
1674 | hslot = &udp_table.hash2[hash2]; | |
1675 | offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node); | |
1676 | } | |
db8dac20 | 1677 | |
645ca708 | 1678 | spin_lock(&hslot->lock); |
2dc41cff | 1679 | sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) { |
5cf3d461 DH |
1680 | if (__udp_is_mcast_sock(net, sk, |
1681 | uh->dest, daddr, | |
1682 | uh->source, saddr, | |
1683 | dif, hnum)) { | |
1684 | if (unlikely(count == ARRAY_SIZE(stack))) { | |
1685 | flush_stack(stack, count, skb, ~0); | |
36cbb245 | 1686 | inner_flushed = true; |
5cf3d461 DH |
1687 | count = 0; |
1688 | } | |
1689 | stack[count++] = sk; | |
2dc41cff | 1690 | sock_hold(sk); |
1240d137 ED |
1691 | } |
1692 | } | |
1240d137 | 1693 | |
645ca708 | 1694 | spin_unlock(&hslot->lock); |
1240d137 | 1695 | |
2dc41cff DH |
1696 | /* Also lookup *:port if we are using hash2 and haven't done so yet. */ |
1697 | if (use_hash2 && hash2 != hash2_any) { | |
1698 | hash2 = hash2_any; | |
1699 | goto start_lookup; | |
1700 | } | |
1701 | ||
1240d137 ED |
1702 | /* |
1703 | * do the slow work with no lock held | |
1704 | */ | |
1705 | if (count) { | |
1706 | flush_stack(stack, count, skb, count - 1); | |
1240d137 | 1707 | } else { |
36cbb245 RJ |
1708 | if (!inner_flushed) |
1709 | UDP_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI, | |
1710 | proto == IPPROTO_UDPLITE); | |
1711 | consume_skb(skb); | |
1240d137 | 1712 | } |
db8dac20 DM |
1713 | return 0; |
1714 | } | |
1715 | ||
1716 | /* Initialize UDP checksum. If exited with zero value (success), | |
1717 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1718 | * Otherwise, csum completion requires chacksumming packet body, | |
1719 | * including udp header and folding it to skb->csum. | |
1720 | */ | |
1721 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1722 | int proto) | |
1723 | { | |
db8dac20 DM |
1724 | int err; |
1725 | ||
1726 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1727 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1728 | ||
1729 | if (proto == IPPROTO_UDPLITE) { | |
1730 | err = udplite_checksum_init(skb, uh); | |
1731 | if (err) | |
1732 | return err; | |
1733 | } | |
1734 | ||
ed70fcfc TH |
1735 | return skb_checksum_init_zero_check(skb, proto, uh->check, |
1736 | inet_compute_pseudo); | |
db8dac20 DM |
1737 | } |
1738 | ||
1739 | /* | |
1740 | * All we need to do is get the socket, and then do a checksum. | |
1741 | */ | |
1742 | ||
645ca708 | 1743 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1744 | int proto) |
1745 | { | |
1746 | struct sock *sk; | |
7b5e56f9 | 1747 | struct udphdr *uh; |
db8dac20 | 1748 | unsigned short ulen; |
adf30907 | 1749 | struct rtable *rt = skb_rtable(skb); |
2783ef23 | 1750 | __be32 saddr, daddr; |
0283328e | 1751 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1752 | |
1753 | /* | |
1754 | * Validate the packet. | |
1755 | */ | |
1756 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1757 | goto drop; /* No space for header. */ | |
1758 | ||
7b5e56f9 | 1759 | uh = udp_hdr(skb); |
db8dac20 | 1760 | ulen = ntohs(uh->len); |
ccc2d97c BM |
1761 | saddr = ip_hdr(skb)->saddr; |
1762 | daddr = ip_hdr(skb)->daddr; | |
1763 | ||
db8dac20 DM |
1764 | if (ulen > skb->len) |
1765 | goto short_packet; | |
1766 | ||
1767 | if (proto == IPPROTO_UDP) { | |
1768 | /* UDP validates ulen. */ | |
1769 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1770 | goto short_packet; | |
1771 | uh = udp_hdr(skb); | |
1772 | } | |
1773 | ||
1774 | if (udp4_csum_init(skb, uh, proto)) | |
1775 | goto csum_error; | |
1776 | ||
8afdd99a ED |
1777 | sk = skb_steal_sock(skb); |
1778 | if (sk) { | |
97502231 | 1779 | struct dst_entry *dst = skb_dst(skb); |
421b3885 | 1780 | int ret; |
421b3885 | 1781 | |
97502231 ED |
1782 | if (unlikely(sk->sk_rx_dst != dst)) |
1783 | udp_sk_rx_dst_set(sk, dst); | |
db8dac20 | 1784 | |
421b3885 | 1785 | ret = udp_queue_rcv_skb(sk, skb); |
8afdd99a | 1786 | sock_put(sk); |
421b3885 SB |
1787 | /* a return value > 0 means to resubmit the input, but |
1788 | * it wants the return to be -protocol, or 0 | |
1789 | */ | |
1790 | if (ret > 0) | |
1791 | return -ret; | |
1792 | return 0; | |
421b3885 | 1793 | } |
db8dac20 | 1794 | |
c18450a5 FF |
1795 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) |
1796 | return __udp4_lib_mcast_deliver(net, skb, uh, | |
36cbb245 | 1797 | saddr, daddr, udptable, proto); |
c18450a5 FF |
1798 | |
1799 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); | |
00db4124 | 1800 | if (sk) { |
a5b50476 ET |
1801 | int ret; |
1802 | ||
224d019c | 1803 | if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk)) |
2abb7cdc TH |
1804 | skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check, |
1805 | inet_compute_pseudo); | |
1806 | ||
a5b50476 | 1807 | ret = udp_queue_rcv_skb(sk, skb); |
db8dac20 DM |
1808 | sock_put(sk); |
1809 | ||
1810 | /* a return value > 0 means to resubmit the input, but | |
1811 | * it wants the return to be -protocol, or 0 | |
1812 | */ | |
1813 | if (ret > 0) | |
1814 | return -ret; | |
1815 | return 0; | |
1816 | } | |
1817 | ||
1818 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1819 | goto drop; | |
1820 | nf_reset(skb); | |
1821 | ||
1822 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1823 | if (udp_lib_checksum_complete(skb)) | |
1824 | goto csum_error; | |
1825 | ||
0283328e | 1826 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1827 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1828 | ||
1829 | /* | |
1830 | * Hmm. We got an UDP packet to a port to which we | |
1831 | * don't wanna listen. Ignore it. | |
1832 | */ | |
1833 | kfree_skb(skb); | |
1834 | return 0; | |
1835 | ||
1836 | short_packet: | |
ba7a46f1 JP |
1837 | net_dbg_ratelimited("UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n", |
1838 | proto == IPPROTO_UDPLITE ? "Lite" : "", | |
1839 | &saddr, ntohs(uh->source), | |
1840 | ulen, skb->len, | |
1841 | &daddr, ntohs(uh->dest)); | |
db8dac20 DM |
1842 | goto drop; |
1843 | ||
1844 | csum_error: | |
1845 | /* | |
1846 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1847 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1848 | */ | |
ba7a46f1 JP |
1849 | net_dbg_ratelimited("UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n", |
1850 | proto == IPPROTO_UDPLITE ? "Lite" : "", | |
1851 | &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest), | |
1852 | ulen); | |
6a5dc9e5 | 1853 | UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 | 1854 | drop: |
0283328e | 1855 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1856 | kfree_skb(skb); |
1857 | return 0; | |
1858 | } | |
1859 | ||
421b3885 SB |
1860 | /* We can only early demux multicast if there is a single matching socket. |
1861 | * If more than one socket found returns NULL | |
1862 | */ | |
1863 | static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net, | |
1864 | __be16 loc_port, __be32 loc_addr, | |
1865 | __be16 rmt_port, __be32 rmt_addr, | |
1866 | int dif) | |
1867 | { | |
1868 | struct sock *sk, *result; | |
1869 | struct hlist_nulls_node *node; | |
1870 | unsigned short hnum = ntohs(loc_port); | |
1871 | unsigned int count, slot = udp_hashfn(net, hnum, udp_table.mask); | |
1872 | struct udp_hslot *hslot = &udp_table.hash[slot]; | |
1873 | ||
63c6f81c ED |
1874 | /* Do not bother scanning a too big list */ |
1875 | if (hslot->count > 10) | |
1876 | return NULL; | |
1877 | ||
421b3885 SB |
1878 | rcu_read_lock(); |
1879 | begin: | |
1880 | count = 0; | |
1881 | result = NULL; | |
1882 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { | |
1883 | if (__udp_is_mcast_sock(net, sk, | |
1884 | loc_port, loc_addr, | |
1885 | rmt_port, rmt_addr, | |
1886 | dif, hnum)) { | |
1887 | result = sk; | |
1888 | ++count; | |
1889 | } | |
1890 | } | |
1891 | /* | |
1892 | * if the nulls value we got at the end of this lookup is | |
1893 | * not the expected one, we must restart lookup. | |
1894 | * We probably met an item that was moved to another chain. | |
1895 | */ | |
1896 | if (get_nulls_value(node) != slot) | |
1897 | goto begin; | |
1898 | ||
1899 | if (result) { | |
1900 | if (count != 1 || | |
1901 | unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) | |
1902 | result = NULL; | |
f69b923a | 1903 | else if (unlikely(!__udp_is_mcast_sock(net, result, |
421b3885 SB |
1904 | loc_port, loc_addr, |
1905 | rmt_port, rmt_addr, | |
1906 | dif, hnum))) { | |
1907 | sock_put(result); | |
1908 | result = NULL; | |
1909 | } | |
1910 | } | |
1911 | rcu_read_unlock(); | |
1912 | return result; | |
1913 | } | |
1914 | ||
1915 | /* For unicast we should only early demux connected sockets or we can | |
1916 | * break forwarding setups. The chains here can be long so only check | |
1917 | * if the first socket is an exact match and if not move on. | |
1918 | */ | |
1919 | static struct sock *__udp4_lib_demux_lookup(struct net *net, | |
1920 | __be16 loc_port, __be32 loc_addr, | |
1921 | __be16 rmt_port, __be32 rmt_addr, | |
1922 | int dif) | |
1923 | { | |
1924 | struct sock *sk, *result; | |
1925 | struct hlist_nulls_node *node; | |
1926 | unsigned short hnum = ntohs(loc_port); | |
1927 | unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum); | |
1928 | unsigned int slot2 = hash2 & udp_table.mask; | |
1929 | struct udp_hslot *hslot2 = &udp_table.hash2[slot2]; | |
c7228317 | 1930 | INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr); |
421b3885 SB |
1931 | const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum); |
1932 | ||
1933 | rcu_read_lock(); | |
1934 | result = NULL; | |
1935 | udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { | |
1936 | if (INET_MATCH(sk, net, acookie, | |
1937 | rmt_addr, loc_addr, ports, dif)) | |
1938 | result = sk; | |
1939 | /* Only check first socket in chain */ | |
1940 | break; | |
1941 | } | |
1942 | ||
1943 | if (result) { | |
1944 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) | |
1945 | result = NULL; | |
1946 | else if (unlikely(!INET_MATCH(sk, net, acookie, | |
1947 | rmt_addr, loc_addr, | |
1948 | ports, dif))) { | |
1949 | sock_put(result); | |
1950 | result = NULL; | |
1951 | } | |
1952 | } | |
1953 | rcu_read_unlock(); | |
1954 | return result; | |
1955 | } | |
1956 | ||
1957 | void udp_v4_early_demux(struct sk_buff *skb) | |
1958 | { | |
610438b7 ED |
1959 | struct net *net = dev_net(skb->dev); |
1960 | const struct iphdr *iph; | |
1961 | const struct udphdr *uh; | |
421b3885 SB |
1962 | struct sock *sk; |
1963 | struct dst_entry *dst; | |
421b3885 SB |
1964 | int dif = skb->dev->ifindex; |
1965 | ||
1966 | /* validate the packet */ | |
1967 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr))) | |
1968 | return; | |
1969 | ||
610438b7 ED |
1970 | iph = ip_hdr(skb); |
1971 | uh = udp_hdr(skb); | |
1972 | ||
421b3885 SB |
1973 | if (skb->pkt_type == PACKET_BROADCAST || |
1974 | skb->pkt_type == PACKET_MULTICAST) | |
1975 | sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr, | |
1976 | uh->source, iph->saddr, dif); | |
1977 | else if (skb->pkt_type == PACKET_HOST) | |
1978 | sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr, | |
1979 | uh->source, iph->saddr, dif); | |
1980 | else | |
1981 | return; | |
1982 | ||
1983 | if (!sk) | |
1984 | return; | |
1985 | ||
1986 | skb->sk = sk; | |
82eabd9e | 1987 | skb->destructor = sock_efree; |
421b3885 SB |
1988 | dst = sk->sk_rx_dst; |
1989 | ||
1990 | if (dst) | |
1991 | dst = dst_check(dst, 0); | |
1992 | if (dst) | |
1993 | skb_dst_set_noref(skb, dst); | |
1994 | } | |
1995 | ||
db8dac20 DM |
1996 | int udp_rcv(struct sk_buff *skb) |
1997 | { | |
645ca708 | 1998 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
1999 | } |
2000 | ||
7d06b2e0 | 2001 | void udp_destroy_sock(struct sock *sk) |
db8dac20 | 2002 | { |
44046a59 | 2003 | struct udp_sock *up = udp_sk(sk); |
8a74ad60 | 2004 | bool slow = lock_sock_fast(sk); |
db8dac20 | 2005 | udp_flush_pending_frames(sk); |
8a74ad60 | 2006 | unlock_sock_fast(sk, slow); |
44046a59 TP |
2007 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
2008 | void (*encap_destroy)(struct sock *sk); | |
2009 | encap_destroy = ACCESS_ONCE(up->encap_destroy); | |
2010 | if (encap_destroy) | |
2011 | encap_destroy(sk); | |
2012 | } | |
db8dac20 DM |
2013 | } |
2014 | ||
1da177e4 LT |
2015 | /* |
2016 | * Socket option code for UDP | |
2017 | */ | |
4c0a6cb0 | 2018 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 2019 | char __user *optval, unsigned int optlen, |
4c0a6cb0 | 2020 | int (*push_pending_frames)(struct sock *)) |
1da177e4 LT |
2021 | { |
2022 | struct udp_sock *up = udp_sk(sk); | |
1c19448c | 2023 | int val, valbool; |
1da177e4 | 2024 | int err = 0; |
b2bf1e26 | 2025 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 2026 | |
c482c568 | 2027 | if (optlen < sizeof(int)) |
1da177e4 LT |
2028 | return -EINVAL; |
2029 | ||
2030 | if (get_user(val, (int __user *)optval)) | |
2031 | return -EFAULT; | |
2032 | ||
1c19448c TH |
2033 | valbool = val ? 1 : 0; |
2034 | ||
6516c655 | 2035 | switch (optname) { |
1da177e4 LT |
2036 | case UDP_CORK: |
2037 | if (val != 0) { | |
2038 | up->corkflag = 1; | |
2039 | } else { | |
2040 | up->corkflag = 0; | |
2041 | lock_sock(sk); | |
4243cdc2 | 2042 | push_pending_frames(sk); |
1da177e4 LT |
2043 | release_sock(sk); |
2044 | } | |
2045 | break; | |
e905a9ed | 2046 | |
1da177e4 LT |
2047 | case UDP_ENCAP: |
2048 | switch (val) { | |
2049 | case 0: | |
2050 | case UDP_ENCAP_ESPINUDP: | |
2051 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
2052 | up->encap_rcv = xfrm4_udp_encap_rcv; |
2053 | /* FALLTHROUGH */ | |
342f0234 | 2054 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 | 2055 | up->encap_type = val; |
447167bf | 2056 | udp_encap_enable(); |
1da177e4 LT |
2057 | break; |
2058 | default: | |
2059 | err = -ENOPROTOOPT; | |
2060 | break; | |
2061 | } | |
2062 | break; | |
2063 | ||
1c19448c TH |
2064 | case UDP_NO_CHECK6_TX: |
2065 | up->no_check6_tx = valbool; | |
2066 | break; | |
2067 | ||
2068 | case UDP_NO_CHECK6_RX: | |
2069 | up->no_check6_rx = valbool; | |
2070 | break; | |
2071 | ||
ba4e58ec GR |
2072 | /* |
2073 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
2074 | */ | |
2075 | /* The sender sets actual checksum coverage length via this option. | |
2076 | * The case coverage > packet length is handled by send module. */ | |
2077 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 2078 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
2079 | return -ENOPROTOOPT; |
2080 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
2081 | val = 8; | |
4be929be AD |
2082 | else if (val > USHRT_MAX) |
2083 | val = USHRT_MAX; | |
ba4e58ec GR |
2084 | up->pcslen = val; |
2085 | up->pcflag |= UDPLITE_SEND_CC; | |
2086 | break; | |
2087 | ||
e905a9ed YH |
2088 | /* The receiver specifies a minimum checksum coverage value. To make |
2089 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
2090 | * used, this again means full checksum coverage. */ |
2091 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 2092 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
2093 | return -ENOPROTOOPT; |
2094 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
2095 | val = 8; | |
4be929be AD |
2096 | else if (val > USHRT_MAX) |
2097 | val = USHRT_MAX; | |
ba4e58ec GR |
2098 | up->pcrlen = val; |
2099 | up->pcflag |= UDPLITE_RECV_CC; | |
2100 | break; | |
2101 | ||
1da177e4 LT |
2102 | default: |
2103 | err = -ENOPROTOOPT; | |
2104 | break; | |
6516c655 | 2105 | } |
1da177e4 LT |
2106 | |
2107 | return err; | |
2108 | } | |
c482c568 | 2109 | EXPORT_SYMBOL(udp_lib_setsockopt); |
1da177e4 | 2110 | |
db8dac20 | 2111 | int udp_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 2112 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
2113 | { |
2114 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2115 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
2116 | udp_push_pending_frames); | |
2117 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
2118 | } | |
2119 | ||
2120 | #ifdef CONFIG_COMPAT | |
2121 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 2122 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
2123 | { |
2124 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2125 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
2126 | udp_push_pending_frames); | |
2127 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
2128 | } | |
2129 | #endif | |
2130 | ||
4c0a6cb0 GR |
2131 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
2132 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
2133 | { |
2134 | struct udp_sock *up = udp_sk(sk); | |
2135 | int val, len; | |
2136 | ||
c482c568 | 2137 | if (get_user(len, optlen)) |
1da177e4 LT |
2138 | return -EFAULT; |
2139 | ||
2140 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 2141 | |
6516c655 | 2142 | if (len < 0) |
1da177e4 LT |
2143 | return -EINVAL; |
2144 | ||
6516c655 | 2145 | switch (optname) { |
1da177e4 LT |
2146 | case UDP_CORK: |
2147 | val = up->corkflag; | |
2148 | break; | |
2149 | ||
2150 | case UDP_ENCAP: | |
2151 | val = up->encap_type; | |
2152 | break; | |
2153 | ||
1c19448c TH |
2154 | case UDP_NO_CHECK6_TX: |
2155 | val = up->no_check6_tx; | |
2156 | break; | |
2157 | ||
2158 | case UDP_NO_CHECK6_RX: | |
2159 | val = up->no_check6_rx; | |
2160 | break; | |
2161 | ||
ba4e58ec GR |
2162 | /* The following two cannot be changed on UDP sockets, the return is |
2163 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
2164 | case UDPLITE_SEND_CSCOV: | |
2165 | val = up->pcslen; | |
2166 | break; | |
2167 | ||
2168 | case UDPLITE_RECV_CSCOV: | |
2169 | val = up->pcrlen; | |
2170 | break; | |
2171 | ||
1da177e4 LT |
2172 | default: |
2173 | return -ENOPROTOOPT; | |
6516c655 | 2174 | } |
1da177e4 | 2175 | |
6516c655 | 2176 | if (put_user(len, optlen)) |
e905a9ed | 2177 | return -EFAULT; |
c482c568 | 2178 | if (copy_to_user(optval, &val, len)) |
1da177e4 | 2179 | return -EFAULT; |
e905a9ed | 2180 | return 0; |
1da177e4 | 2181 | } |
c482c568 | 2182 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1da177e4 | 2183 | |
db8dac20 DM |
2184 | int udp_getsockopt(struct sock *sk, int level, int optname, |
2185 | char __user *optval, int __user *optlen) | |
2186 | { | |
2187 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2188 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
2189 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
2190 | } | |
2191 | ||
2192 | #ifdef CONFIG_COMPAT | |
2193 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
2194 | char __user *optval, int __user *optlen) | |
2195 | { | |
2196 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2197 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
2198 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
2199 | } | |
2200 | #endif | |
1da177e4 LT |
2201 | /** |
2202 | * udp_poll - wait for a UDP event. | |
2203 | * @file - file struct | |
2204 | * @sock - socket | |
2205 | * @wait - poll table | |
2206 | * | |
e905a9ed | 2207 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
2208 | * blocking sockets. If application is using a blocking fd |
2209 | * and a packet with checksum error is in the queue; | |
2210 | * then it could get return from select indicating data available | |
2211 | * but then block when reading it. Add special case code | |
2212 | * to work around these arguably broken applications. | |
2213 | */ | |
2214 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
2215 | { | |
2216 | unsigned int mask = datagram_poll(file, sock, wait); | |
2217 | struct sock *sk = sock->sk; | |
ba4e58ec | 2218 | |
c3f1dbaf DM |
2219 | sock_rps_record_flow(sk); |
2220 | ||
1da177e4 | 2221 | /* Check for false positives due to checksum errors */ |
85584672 ED |
2222 | if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) && |
2223 | !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk)) | |
2224 | mask &= ~(POLLIN | POLLRDNORM); | |
1da177e4 LT |
2225 | |
2226 | return mask; | |
e905a9ed | 2227 | |
1da177e4 | 2228 | } |
c482c568 | 2229 | EXPORT_SYMBOL(udp_poll); |
1da177e4 | 2230 | |
db8dac20 DM |
2231 | struct proto udp_prot = { |
2232 | .name = "UDP", | |
2233 | .owner = THIS_MODULE, | |
2234 | .close = udp_lib_close, | |
2235 | .connect = ip4_datagram_connect, | |
2236 | .disconnect = udp_disconnect, | |
2237 | .ioctl = udp_ioctl, | |
2238 | .destroy = udp_destroy_sock, | |
2239 | .setsockopt = udp_setsockopt, | |
2240 | .getsockopt = udp_getsockopt, | |
2241 | .sendmsg = udp_sendmsg, | |
2242 | .recvmsg = udp_recvmsg, | |
2243 | .sendpage = udp_sendpage, | |
93821778 | 2244 | .backlog_rcv = __udp_queue_rcv_skb, |
8141ed9f | 2245 | .release_cb = ip4_datagram_release_cb, |
db8dac20 DM |
2246 | .hash = udp_lib_hash, |
2247 | .unhash = udp_lib_unhash, | |
719f8358 | 2248 | .rehash = udp_v4_rehash, |
db8dac20 DM |
2249 | .get_port = udp_v4_get_port, |
2250 | .memory_allocated = &udp_memory_allocated, | |
2251 | .sysctl_mem = sysctl_udp_mem, | |
2252 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
2253 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
2254 | .obj_size = sizeof(struct udp_sock), | |
271b72c7 | 2255 | .slab_flags = SLAB_DESTROY_BY_RCU, |
645ca708 | 2256 | .h.udp_table = &udp_table, |
db8dac20 DM |
2257 | #ifdef CONFIG_COMPAT |
2258 | .compat_setsockopt = compat_udp_setsockopt, | |
2259 | .compat_getsockopt = compat_udp_getsockopt, | |
2260 | #endif | |
fcbdf09d | 2261 | .clear_sk = sk_prot_clear_portaddr_nulls, |
db8dac20 | 2262 | }; |
c482c568 | 2263 | EXPORT_SYMBOL(udp_prot); |
1da177e4 LT |
2264 | |
2265 | /* ------------------------------------------------------------------------ */ | |
2266 | #ifdef CONFIG_PROC_FS | |
2267 | ||
645ca708 | 2268 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
2269 | { |
2270 | struct sock *sk; | |
2271 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2272 | struct net *net = seq_file_net(seq); |
1da177e4 | 2273 | |
f86dcc5a ED |
2274 | for (state->bucket = start; state->bucket <= state->udp_table->mask; |
2275 | ++state->bucket) { | |
88ab1932 | 2276 | struct hlist_nulls_node *node; |
645ca708 | 2277 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
f86dcc5a ED |
2278 | |
2279 | if (hlist_nulls_empty(&hslot->head)) | |
2280 | continue; | |
2281 | ||
645ca708 | 2282 | spin_lock_bh(&hslot->lock); |
88ab1932 | 2283 | sk_nulls_for_each(sk, node, &hslot->head) { |
878628fb | 2284 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 2285 | continue; |
1da177e4 LT |
2286 | if (sk->sk_family == state->family) |
2287 | goto found; | |
2288 | } | |
645ca708 | 2289 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
2290 | } |
2291 | sk = NULL; | |
2292 | found: | |
2293 | return sk; | |
2294 | } | |
2295 | ||
2296 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
2297 | { | |
2298 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2299 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2300 | |
2301 | do { | |
88ab1932 | 2302 | sk = sk_nulls_next(sk); |
878628fb | 2303 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 2304 | |
645ca708 | 2305 | if (!sk) { |
f86dcc5a | 2306 | if (state->bucket <= state->udp_table->mask) |
30842f29 | 2307 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
645ca708 | 2308 | return udp_get_first(seq, state->bucket + 1); |
1da177e4 LT |
2309 | } |
2310 | return sk; | |
2311 | } | |
2312 | ||
2313 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
2314 | { | |
645ca708 | 2315 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
2316 | |
2317 | if (sk) | |
6516c655 | 2318 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
2319 | --pos; |
2320 | return pos ? NULL : sk; | |
2321 | } | |
2322 | ||
2323 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
2324 | { | |
30842f29 | 2325 | struct udp_iter_state *state = seq->private; |
f86dcc5a | 2326 | state->bucket = MAX_UDP_PORTS; |
30842f29 | 2327 | |
b50660f1 | 2328 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
2329 | } |
2330 | ||
2331 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2332 | { | |
2333 | struct sock *sk; | |
2334 | ||
b50660f1 | 2335 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2336 | sk = udp_get_idx(seq, 0); |
2337 | else | |
2338 | sk = udp_get_next(seq, v); | |
2339 | ||
2340 | ++*pos; | |
2341 | return sk; | |
2342 | } | |
2343 | ||
2344 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
2345 | { | |
645ca708 ED |
2346 | struct udp_iter_state *state = seq->private; |
2347 | ||
f86dcc5a | 2348 | if (state->bucket <= state->udp_table->mask) |
645ca708 | 2349 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
1da177e4 LT |
2350 | } |
2351 | ||
73cb88ec | 2352 | int udp_seq_open(struct inode *inode, struct file *file) |
1da177e4 | 2353 | { |
d9dda78b | 2354 | struct udp_seq_afinfo *afinfo = PDE_DATA(inode); |
a2be75c1 DL |
2355 | struct udp_iter_state *s; |
2356 | int err; | |
a91275ef | 2357 | |
a2be75c1 DL |
2358 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2359 | sizeof(struct udp_iter_state)); | |
2360 | if (err < 0) | |
2361 | return err; | |
a91275ef | 2362 | |
a2be75c1 | 2363 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2364 | s->family = afinfo->family; |
645ca708 | 2365 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 2366 | return err; |
a91275ef | 2367 | } |
73cb88ec | 2368 | EXPORT_SYMBOL(udp_seq_open); |
a91275ef | 2369 | |
1da177e4 | 2370 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 2371 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
2372 | { |
2373 | struct proc_dir_entry *p; | |
2374 | int rc = 0; | |
2375 | ||
dda61925 DL |
2376 | afinfo->seq_ops.start = udp_seq_start; |
2377 | afinfo->seq_ops.next = udp_seq_next; | |
2378 | afinfo->seq_ops.stop = udp_seq_stop; | |
2379 | ||
84841c3c | 2380 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2381 | afinfo->seq_fops, afinfo); |
84841c3c | 2382 | if (!p) |
1da177e4 LT |
2383 | rc = -ENOMEM; |
2384 | return rc; | |
2385 | } | |
c482c568 | 2386 | EXPORT_SYMBOL(udp_proc_register); |
1da177e4 | 2387 | |
0c96d8c5 | 2388 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 2389 | { |
ece31ffd | 2390 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2391 | } |
c482c568 | 2392 | EXPORT_SYMBOL(udp_proc_unregister); |
db8dac20 DM |
2393 | |
2394 | /* ------------------------------------------------------------------------ */ | |
5e659e4c | 2395 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
652586df | 2396 | int bucket) |
db8dac20 DM |
2397 | { |
2398 | struct inet_sock *inet = inet_sk(sp); | |
c720c7e8 ED |
2399 | __be32 dest = inet->inet_daddr; |
2400 | __be32 src = inet->inet_rcv_saddr; | |
2401 | __u16 destp = ntohs(inet->inet_dport); | |
2402 | __u16 srcp = ntohs(inet->inet_sport); | |
db8dac20 | 2403 | |
f86dcc5a | 2404 | seq_printf(f, "%5d: %08X:%04X %08X:%04X" |
652586df | 2405 | " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d", |
db8dac20 | 2406 | bucket, src, srcp, dest, destp, sp->sk_state, |
31e6d363 ED |
2407 | sk_wmem_alloc_get(sp), |
2408 | sk_rmem_alloc_get(sp), | |
a7cb5a49 EB |
2409 | 0, 0L, 0, |
2410 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)), | |
2411 | 0, sock_i_ino(sp), | |
cb61cb9b | 2412 | atomic_read(&sp->sk_refcnt), sp, |
652586df | 2413 | atomic_read(&sp->sk_drops)); |
db8dac20 DM |
2414 | } |
2415 | ||
2416 | int udp4_seq_show(struct seq_file *seq, void *v) | |
2417 | { | |
652586df | 2418 | seq_setwidth(seq, 127); |
db8dac20 | 2419 | if (v == SEQ_START_TOKEN) |
652586df | 2420 | seq_puts(seq, " sl local_address rem_address st tx_queue " |
db8dac20 | 2421 | "rx_queue tr tm->when retrnsmt uid timeout " |
cb61cb9b | 2422 | "inode ref pointer drops"); |
db8dac20 | 2423 | else { |
db8dac20 DM |
2424 | struct udp_iter_state *state = seq->private; |
2425 | ||
652586df | 2426 | udp4_format_sock(v, seq, state->bucket); |
db8dac20 | 2427 | } |
652586df | 2428 | seq_pad(seq, '\n'); |
db8dac20 DM |
2429 | return 0; |
2430 | } | |
2431 | ||
73cb88ec AV |
2432 | static const struct file_operations udp_afinfo_seq_fops = { |
2433 | .owner = THIS_MODULE, | |
2434 | .open = udp_seq_open, | |
2435 | .read = seq_read, | |
2436 | .llseek = seq_lseek, | |
2437 | .release = seq_release_net | |
2438 | }; | |
2439 | ||
db8dac20 | 2440 | /* ------------------------------------------------------------------------ */ |
db8dac20 | 2441 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
2442 | .name = "udp", |
2443 | .family = AF_INET, | |
645ca708 | 2444 | .udp_table = &udp_table, |
73cb88ec | 2445 | .seq_fops = &udp_afinfo_seq_fops, |
dda61925 DL |
2446 | .seq_ops = { |
2447 | .show = udp4_seq_show, | |
2448 | }, | |
db8dac20 DM |
2449 | }; |
2450 | ||
2c8c1e72 | 2451 | static int __net_init udp4_proc_init_net(struct net *net) |
15439feb PE |
2452 | { |
2453 | return udp_proc_register(net, &udp4_seq_afinfo); | |
2454 | } | |
2455 | ||
2c8c1e72 | 2456 | static void __net_exit udp4_proc_exit_net(struct net *net) |
15439feb PE |
2457 | { |
2458 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
2459 | } | |
2460 | ||
2461 | static struct pernet_operations udp4_net_ops = { | |
2462 | .init = udp4_proc_init_net, | |
2463 | .exit = udp4_proc_exit_net, | |
2464 | }; | |
2465 | ||
db8dac20 DM |
2466 | int __init udp4_proc_init(void) |
2467 | { | |
15439feb | 2468 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
2469 | } |
2470 | ||
2471 | void udp4_proc_exit(void) | |
2472 | { | |
15439feb | 2473 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 2474 | } |
1da177e4 LT |
2475 | #endif /* CONFIG_PROC_FS */ |
2476 | ||
f86dcc5a ED |
2477 | static __initdata unsigned long uhash_entries; |
2478 | static int __init set_uhash_entries(char *str) | |
645ca708 | 2479 | { |
413c27d8 EZ |
2480 | ssize_t ret; |
2481 | ||
f86dcc5a ED |
2482 | if (!str) |
2483 | return 0; | |
413c27d8 EZ |
2484 | |
2485 | ret = kstrtoul(str, 0, &uhash_entries); | |
2486 | if (ret) | |
2487 | return 0; | |
2488 | ||
f86dcc5a ED |
2489 | if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN) |
2490 | uhash_entries = UDP_HTABLE_SIZE_MIN; | |
2491 | return 1; | |
2492 | } | |
2493 | __setup("uhash_entries=", set_uhash_entries); | |
645ca708 | 2494 | |
f86dcc5a ED |
2495 | void __init udp_table_init(struct udp_table *table, const char *name) |
2496 | { | |
2497 | unsigned int i; | |
2498 | ||
31fe62b9 TB |
2499 | table->hash = alloc_large_system_hash(name, |
2500 | 2 * sizeof(struct udp_hslot), | |
2501 | uhash_entries, | |
2502 | 21, /* one slot per 2 MB */ | |
2503 | 0, | |
2504 | &table->log, | |
2505 | &table->mask, | |
2506 | UDP_HTABLE_SIZE_MIN, | |
2507 | 64 * 1024); | |
2508 | ||
512615b6 | 2509 | table->hash2 = table->hash + (table->mask + 1); |
f86dcc5a | 2510 | for (i = 0; i <= table->mask; i++) { |
88ab1932 | 2511 | INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i); |
fdcc8aa9 | 2512 | table->hash[i].count = 0; |
645ca708 ED |
2513 | spin_lock_init(&table->hash[i].lock); |
2514 | } | |
512615b6 ED |
2515 | for (i = 0; i <= table->mask; i++) { |
2516 | INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i); | |
2517 | table->hash2[i].count = 0; | |
2518 | spin_lock_init(&table->hash2[i].lock); | |
2519 | } | |
645ca708 ED |
2520 | } |
2521 | ||
723b8e46 TH |
2522 | u32 udp_flow_hashrnd(void) |
2523 | { | |
2524 | static u32 hashrnd __read_mostly; | |
2525 | ||
2526 | net_get_random_once(&hashrnd, sizeof(hashrnd)); | |
2527 | ||
2528 | return hashrnd; | |
2529 | } | |
2530 | EXPORT_SYMBOL(udp_flow_hashrnd); | |
2531 | ||
95766fff HA |
2532 | void __init udp_init(void) |
2533 | { | |
f03d78db | 2534 | unsigned long limit; |
95766fff | 2535 | |
f86dcc5a | 2536 | udp_table_init(&udp_table, "UDP"); |
f03d78db | 2537 | limit = nr_free_buffer_pages() / 8; |
95766fff HA |
2538 | limit = max(limit, 128UL); |
2539 | sysctl_udp_mem[0] = limit / 4 * 3; | |
2540 | sysctl_udp_mem[1] = limit; | |
2541 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
2542 | ||
2543 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
2544 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
2545 | } |