Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * The User Datagram Protocol (UDP). | |
7 | * | |
02c30a84 | 8 | * Authors: Ross Biro |
1da177e4 LT |
9 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
10 | * Arnt Gulbrandsen, <agulbra@nvg.unit.no> | |
113aa838 | 11 | * Alan Cox, <alan@lxorguk.ukuu.org.uk> |
1da177e4 LT |
12 | * Hirokazu Takahashi, <taka@valinux.co.jp> |
13 | * | |
14 | * Fixes: | |
15 | * Alan Cox : verify_area() calls | |
16 | * Alan Cox : stopped close while in use off icmp | |
17 | * messages. Not a fix but a botch that | |
18 | * for udp at least is 'valid'. | |
19 | * Alan Cox : Fixed icmp handling properly | |
20 | * Alan Cox : Correct error for oversized datagrams | |
e905a9ed YH |
21 | * Alan Cox : Tidied select() semantics. |
22 | * Alan Cox : udp_err() fixed properly, also now | |
1da177e4 LT |
23 | * select and read wake correctly on errors |
24 | * Alan Cox : udp_send verify_area moved to avoid mem leak | |
25 | * Alan Cox : UDP can count its memory | |
26 | * Alan Cox : send to an unknown connection causes | |
27 | * an ECONNREFUSED off the icmp, but | |
28 | * does NOT close. | |
29 | * Alan Cox : Switched to new sk_buff handlers. No more backlog! | |
30 | * Alan Cox : Using generic datagram code. Even smaller and the PEEK | |
31 | * bug no longer crashes it. | |
32 | * Fred Van Kempen : Net2e support for sk->broadcast. | |
33 | * Alan Cox : Uses skb_free_datagram | |
34 | * Alan Cox : Added get/set sockopt support. | |
35 | * Alan Cox : Broadcasting without option set returns EACCES. | |
36 | * Alan Cox : No wakeup calls. Instead we now use the callbacks. | |
37 | * Alan Cox : Use ip_tos and ip_ttl | |
38 | * Alan Cox : SNMP Mibs | |
39 | * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support. | |
40 | * Matt Dillon : UDP length checks. | |
41 | * Alan Cox : Smarter af_inet used properly. | |
42 | * Alan Cox : Use new kernel side addressing. | |
43 | * Alan Cox : Incorrect return on truncated datagram receive. | |
44 | * Arnt Gulbrandsen : New udp_send and stuff | |
45 | * Alan Cox : Cache last socket | |
46 | * Alan Cox : Route cache | |
47 | * Jon Peatfield : Minor efficiency fix to sendto(). | |
48 | * Mike Shaver : RFC1122 checks. | |
49 | * Alan Cox : Nonblocking error fix. | |
50 | * Willy Konynenberg : Transparent proxying support. | |
51 | * Mike McLagan : Routing by source | |
52 | * David S. Miller : New socket lookup architecture. | |
53 | * Last socket cache retained as it | |
54 | * does have a high hit rate. | |
55 | * Olaf Kirch : Don't linearise iovec on sendmsg. | |
56 | * Andi Kleen : Some cleanups, cache destination entry | |
e905a9ed | 57 | * for connect. |
1da177e4 LT |
58 | * Vitaly E. Lavrov : Transparent proxy revived after year coma. |
59 | * Melvin Smith : Check msg_name not msg_namelen in sendto(), | |
60 | * return ENOTCONN for unconnected sockets (POSIX) | |
61 | * Janos Farkas : don't deliver multi/broadcasts to a different | |
62 | * bound-to-device socket | |
63 | * Hirokazu Takahashi : HW checksumming for outgoing UDP | |
64 | * datagrams. | |
65 | * Hirokazu Takahashi : sendfile() on UDP works now. | |
66 | * Arnaldo C. Melo : convert /proc/net/udp to seq_file | |
67 | * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which | |
68 | * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind | |
69 | * a single port at the same time. | |
70 | * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support | |
342f0234 | 71 | * James Chapman : Add L2TP encapsulation type. |
1da177e4 LT |
72 | * |
73 | * | |
74 | * This program is free software; you can redistribute it and/or | |
75 | * modify it under the terms of the GNU General Public License | |
76 | * as published by the Free Software Foundation; either version | |
77 | * 2 of the License, or (at your option) any later version. | |
78 | */ | |
e905a9ed | 79 | |
afd46503 JP |
80 | #define pr_fmt(fmt) "UDP: " fmt |
81 | ||
1da177e4 LT |
82 | #include <asm/uaccess.h> |
83 | #include <asm/ioctls.h> | |
95766fff | 84 | #include <linux/bootmem.h> |
8203efb3 ED |
85 | #include <linux/highmem.h> |
86 | #include <linux/swap.h> | |
1da177e4 LT |
87 | #include <linux/types.h> |
88 | #include <linux/fcntl.h> | |
89 | #include <linux/module.h> | |
90 | #include <linux/socket.h> | |
91 | #include <linux/sockios.h> | |
14c85021 | 92 | #include <linux/igmp.h> |
6e540309 | 93 | #include <linux/inetdevice.h> |
1da177e4 LT |
94 | #include <linux/in.h> |
95 | #include <linux/errno.h> | |
96 | #include <linux/timer.h> | |
97 | #include <linux/mm.h> | |
1da177e4 | 98 | #include <linux/inet.h> |
1da177e4 | 99 | #include <linux/netdevice.h> |
5a0e3ad6 | 100 | #include <linux/slab.h> |
c752f073 | 101 | #include <net/tcp_states.h> |
1da177e4 LT |
102 | #include <linux/skbuff.h> |
103 | #include <linux/proc_fs.h> | |
104 | #include <linux/seq_file.h> | |
457c4cbc | 105 | #include <net/net_namespace.h> |
1da177e4 | 106 | #include <net/icmp.h> |
421b3885 | 107 | #include <net/inet_hashtables.h> |
1da177e4 | 108 | #include <net/route.h> |
1da177e4 LT |
109 | #include <net/checksum.h> |
110 | #include <net/xfrm.h> | |
296f7ea7 | 111 | #include <trace/events/udp.h> |
447167bf | 112 | #include <linux/static_key.h> |
22911fc5 | 113 | #include <trace/events/skb.h> |
076bb0c8 | 114 | #include <net/busy_poll.h> |
ba4e58ec | 115 | #include "udp_impl.h" |
e32ea7e7 | 116 | #include <net/sock_reuseport.h> |
1da177e4 | 117 | |
f86dcc5a | 118 | struct udp_table udp_table __read_mostly; |
645ca708 | 119 | EXPORT_SYMBOL(udp_table); |
1da177e4 | 120 | |
8d987e5c | 121 | long sysctl_udp_mem[3] __read_mostly; |
95766fff | 122 | EXPORT_SYMBOL(sysctl_udp_mem); |
c482c568 ED |
123 | |
124 | int sysctl_udp_rmem_min __read_mostly; | |
95766fff | 125 | EXPORT_SYMBOL(sysctl_udp_rmem_min); |
c482c568 ED |
126 | |
127 | int sysctl_udp_wmem_min __read_mostly; | |
95766fff HA |
128 | EXPORT_SYMBOL(sysctl_udp_wmem_min); |
129 | ||
8d987e5c | 130 | atomic_long_t udp_memory_allocated; |
95766fff HA |
131 | EXPORT_SYMBOL(udp_memory_allocated); |
132 | ||
f86dcc5a ED |
133 | #define MAX_UDP_PORTS 65536 |
134 | #define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN) | |
98322f22 | 135 | |
f24d43c0 | 136 | static int udp_lib_lport_inuse(struct net *net, __u16 num, |
645ca708 | 137 | const struct udp_hslot *hslot, |
98322f22 | 138 | unsigned long *bitmap, |
f24d43c0 ED |
139 | struct sock *sk, |
140 | int (*saddr_comp)(const struct sock *sk1, | |
e32ea7e7 CG |
141 | const struct sock *sk2, |
142 | bool match_wildcard), | |
f86dcc5a | 143 | unsigned int log) |
1da177e4 | 144 | { |
f24d43c0 | 145 | struct sock *sk2; |
88ab1932 | 146 | struct hlist_nulls_node *node; |
ba418fa3 | 147 | kuid_t uid = sock_i_uid(sk); |
25030a7f | 148 | |
4243cdc2 | 149 | sk_nulls_for_each(sk2, node, &hslot->head) { |
9d4fb27d JP |
150 | if (net_eq(sock_net(sk2), net) && |
151 | sk2 != sk && | |
d4cada4a | 152 | (bitmap || udp_sk(sk2)->udp_port_hash == num) && |
9d4fb27d JP |
153 | (!sk2->sk_reuse || !sk->sk_reuse) && |
154 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
155 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 | 156 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
e32ea7e7 | 157 | rcu_access_pointer(sk->sk_reuseport_cb) || |
4243cdc2 | 158 | !uid_eq(uid, sock_i_uid(sk2))) && |
e32ea7e7 | 159 | saddr_comp(sk, sk2, true)) { |
4243cdc2 | 160 | if (!bitmap) |
98322f22 | 161 | return 1; |
4243cdc2 | 162 | __set_bit(udp_sk(sk2)->udp_port_hash >> log, bitmap); |
98322f22 | 163 | } |
4243cdc2 | 164 | } |
25030a7f GR |
165 | return 0; |
166 | } | |
167 | ||
30fff923 ED |
168 | /* |
169 | * Note: we still hold spinlock of primary hash chain, so no other writer | |
170 | * can insert/delete a socket with local_port == num | |
171 | */ | |
172 | static int udp_lib_lport_inuse2(struct net *net, __u16 num, | |
4243cdc2 JP |
173 | struct udp_hslot *hslot2, |
174 | struct sock *sk, | |
175 | int (*saddr_comp)(const struct sock *sk1, | |
e32ea7e7 CG |
176 | const struct sock *sk2, |
177 | bool match_wildcard)) | |
30fff923 ED |
178 | { |
179 | struct sock *sk2; | |
180 | struct hlist_nulls_node *node; | |
ba418fa3 | 181 | kuid_t uid = sock_i_uid(sk); |
30fff923 ED |
182 | int res = 0; |
183 | ||
184 | spin_lock(&hslot2->lock); | |
4243cdc2 | 185 | udp_portaddr_for_each_entry(sk2, node, &hslot2->head) { |
9d4fb27d JP |
186 | if (net_eq(sock_net(sk2), net) && |
187 | sk2 != sk && | |
188 | (udp_sk(sk2)->udp_port_hash == num) && | |
189 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
190 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
191 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 | 192 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
e32ea7e7 | 193 | rcu_access_pointer(sk->sk_reuseport_cb) || |
4243cdc2 | 194 | !uid_eq(uid, sock_i_uid(sk2))) && |
e32ea7e7 | 195 | saddr_comp(sk, sk2, true)) { |
30fff923 ED |
196 | res = 1; |
197 | break; | |
198 | } | |
4243cdc2 | 199 | } |
30fff923 ED |
200 | spin_unlock(&hslot2->lock); |
201 | return res; | |
202 | } | |
203 | ||
e32ea7e7 CG |
204 | static int udp_reuseport_add_sock(struct sock *sk, struct udp_hslot *hslot, |
205 | int (*saddr_same)(const struct sock *sk1, | |
206 | const struct sock *sk2, | |
207 | bool match_wildcard)) | |
208 | { | |
209 | struct net *net = sock_net(sk); | |
210 | struct hlist_nulls_node *node; | |
211 | kuid_t uid = sock_i_uid(sk); | |
212 | struct sock *sk2; | |
213 | ||
214 | sk_nulls_for_each(sk2, node, &hslot->head) { | |
215 | if (net_eq(sock_net(sk2), net) && | |
216 | sk2 != sk && | |
217 | sk2->sk_family == sk->sk_family && | |
218 | ipv6_only_sock(sk2) == ipv6_only_sock(sk) && | |
219 | (udp_sk(sk2)->udp_port_hash == udp_sk(sk)->udp_port_hash) && | |
220 | (sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
221 | sk2->sk_reuseport && uid_eq(uid, sock_i_uid(sk2)) && | |
222 | (*saddr_same)(sk, sk2, false)) { | |
223 | return reuseport_add_sock(sk, sk2); | |
224 | } | |
225 | } | |
226 | ||
227 | /* Initial allocation may have already happened via setsockopt */ | |
228 | if (!rcu_access_pointer(sk->sk_reuseport_cb)) | |
229 | return reuseport_alloc(sk); | |
230 | return 0; | |
231 | } | |
232 | ||
25030a7f | 233 | /** |
6ba5a3c5 | 234 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
235 | * |
236 | * @sk: socket struct in question | |
237 | * @snum: port number to look up | |
df2bc459 | 238 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25985edc | 239 | * @hash2_nulladdr: AF-dependent hash value in secondary hash chains, |
30fff923 | 240 | * with NULL address |
25030a7f | 241 | */ |
6ba5a3c5 | 242 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
4243cdc2 | 243 | int (*saddr_comp)(const struct sock *sk1, |
e32ea7e7 CG |
244 | const struct sock *sk2, |
245 | bool match_wildcard), | |
30fff923 | 246 | unsigned int hash2_nulladdr) |
25030a7f | 247 | { |
512615b6 | 248 | struct udp_hslot *hslot, *hslot2; |
645ca708 | 249 | struct udp_table *udptable = sk->sk_prot->h.udp_table; |
25030a7f | 250 | int error = 1; |
3b1e0a65 | 251 | struct net *net = sock_net(sk); |
1da177e4 | 252 | |
32c1da70 | 253 | if (!snum) { |
9088c560 | 254 | int low, high, remaining; |
95c96174 | 255 | unsigned int rand; |
98322f22 ED |
256 | unsigned short first, last; |
257 | DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN); | |
32c1da70 | 258 | |
0bbf87d8 | 259 | inet_get_local_port_range(net, &low, &high); |
a25de534 | 260 | remaining = (high - low) + 1; |
227b60f5 | 261 | |
63862b5b | 262 | rand = prandom_u32(); |
8fc54f68 | 263 | first = reciprocal_scale(rand, remaining) + low; |
98322f22 ED |
264 | /* |
265 | * force rand to be an odd multiple of UDP_HTABLE_SIZE | |
266 | */ | |
f86dcc5a | 267 | rand = (rand | 1) * (udptable->mask + 1); |
5781b235 ED |
268 | last = first + udptable->mask + 1; |
269 | do { | |
f86dcc5a | 270 | hslot = udp_hashslot(udptable, net, first); |
98322f22 | 271 | bitmap_zero(bitmap, PORTS_PER_CHAIN); |
645ca708 | 272 | spin_lock_bh(&hslot->lock); |
98322f22 | 273 | udp_lib_lport_inuse(net, snum, hslot, bitmap, sk, |
f86dcc5a | 274 | saddr_comp, udptable->log); |
98322f22 ED |
275 | |
276 | snum = first; | |
277 | /* | |
278 | * Iterate on all possible values of snum for this hash. | |
279 | * Using steps of an odd multiple of UDP_HTABLE_SIZE | |
280 | * give us randomization and full range coverage. | |
281 | */ | |
9088c560 | 282 | do { |
98322f22 | 283 | if (low <= snum && snum <= high && |
e3826f1e | 284 | !test_bit(snum >> udptable->log, bitmap) && |
122ff243 | 285 | !inet_is_local_reserved_port(net, snum)) |
98322f22 ED |
286 | goto found; |
287 | snum += rand; | |
288 | } while (snum != first); | |
289 | spin_unlock_bh(&hslot->lock); | |
5781b235 | 290 | } while (++first != last); |
98322f22 | 291 | goto fail; |
645ca708 | 292 | } else { |
f86dcc5a | 293 | hslot = udp_hashslot(udptable, net, snum); |
645ca708 | 294 | spin_lock_bh(&hslot->lock); |
30fff923 ED |
295 | if (hslot->count > 10) { |
296 | int exist; | |
297 | unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum; | |
298 | ||
299 | slot2 &= udptable->mask; | |
300 | hash2_nulladdr &= udptable->mask; | |
301 | ||
302 | hslot2 = udp_hashslot2(udptable, slot2); | |
303 | if (hslot->count < hslot2->count) | |
304 | goto scan_primary_hash; | |
305 | ||
306 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
307 | sk, saddr_comp); | |
308 | if (!exist && (hash2_nulladdr != slot2)) { | |
309 | hslot2 = udp_hashslot2(udptable, hash2_nulladdr); | |
310 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
311 | sk, saddr_comp); | |
312 | } | |
313 | if (exist) | |
314 | goto fail_unlock; | |
315 | else | |
316 | goto found; | |
317 | } | |
318 | scan_primary_hash: | |
f86dcc5a ED |
319 | if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk, |
320 | saddr_comp, 0)) | |
645ca708 ED |
321 | goto fail_unlock; |
322 | } | |
98322f22 | 323 | found: |
c720c7e8 | 324 | inet_sk(sk)->inet_num = snum; |
d4cada4a ED |
325 | udp_sk(sk)->udp_port_hash = snum; |
326 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
1da177e4 | 327 | if (sk_unhashed(sk)) { |
e32ea7e7 CG |
328 | if (sk->sk_reuseport && |
329 | udp_reuseport_add_sock(sk, hslot, saddr_comp)) { | |
330 | inet_sk(sk)->inet_num = 0; | |
331 | udp_sk(sk)->udp_port_hash = 0; | |
332 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
333 | goto fail_unlock; | |
334 | } | |
335 | ||
88ab1932 | 336 | sk_nulls_add_node_rcu(sk, &hslot->head); |
fdcc8aa9 | 337 | hslot->count++; |
c29a0bc4 | 338 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
512615b6 ED |
339 | |
340 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
341 | spin_lock(&hslot2->lock); | |
342 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
343 | &hslot2->head); | |
344 | hslot2->count++; | |
345 | spin_unlock(&hslot2->lock); | |
1da177e4 | 346 | } |
25030a7f | 347 | error = 0; |
645ca708 ED |
348 | fail_unlock: |
349 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 350 | fail: |
25030a7f GR |
351 | return error; |
352 | } | |
c482c568 | 353 | EXPORT_SYMBOL(udp_lib_get_port); |
25030a7f | 354 | |
e32ea7e7 CG |
355 | /* match_wildcard == true: 0.0.0.0 equals to any IPv4 addresses |
356 | * match_wildcard == false: addresses must be exactly the same, i.e. | |
357 | * 0.0.0.0 only equals to 0.0.0.0 | |
358 | */ | |
359 | static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2, | |
360 | bool match_wildcard) | |
db8dac20 DM |
361 | { |
362 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
363 | ||
e32ea7e7 CG |
364 | if (!ipv6_only_sock(sk2)) { |
365 | if (inet1->inet_rcv_saddr == inet2->inet_rcv_saddr) | |
366 | return 1; | |
367 | if (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr) | |
368 | return match_wildcard; | |
369 | } | |
370 | return 0; | |
db8dac20 DM |
371 | } |
372 | ||
6eada011 ED |
373 | static u32 udp4_portaddr_hash(const struct net *net, __be32 saddr, |
374 | unsigned int port) | |
d4cada4a | 375 | { |
0eae88f3 | 376 | return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port; |
d4cada4a ED |
377 | } |
378 | ||
6ba5a3c5 | 379 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 380 | { |
30fff923 | 381 | unsigned int hash2_nulladdr = |
0eae88f3 | 382 | udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum); |
30fff923 ED |
383 | unsigned int hash2_partial = |
384 | udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0); | |
385 | ||
d4cada4a | 386 | /* precompute partial secondary hash */ |
30fff923 ED |
387 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
388 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr); | |
db8dac20 DM |
389 | } |
390 | ||
60c04aec JP |
391 | static inline int compute_score(struct sock *sk, struct net *net, |
392 | __be32 saddr, unsigned short hnum, __be16 sport, | |
393 | __be32 daddr, __be16 dport, int dif) | |
645ca708 | 394 | { |
60c04aec JP |
395 | int score; |
396 | struct inet_sock *inet; | |
645ca708 | 397 | |
60c04aec JP |
398 | if (!net_eq(sock_net(sk), net) || |
399 | udp_sk(sk)->udp_port_hash != hnum || | |
400 | ipv6_only_sock(sk)) | |
401 | return -1; | |
645ca708 | 402 | |
60c04aec JP |
403 | score = (sk->sk_family == PF_INET) ? 2 : 1; |
404 | inet = inet_sk(sk); | |
405 | ||
406 | if (inet->inet_rcv_saddr) { | |
407 | if (inet->inet_rcv_saddr != daddr) | |
408 | return -1; | |
409 | score += 4; | |
645ca708 | 410 | } |
60c04aec JP |
411 | |
412 | if (inet->inet_daddr) { | |
413 | if (inet->inet_daddr != saddr) | |
414 | return -1; | |
415 | score += 4; | |
416 | } | |
417 | ||
418 | if (inet->inet_dport) { | |
419 | if (inet->inet_dport != sport) | |
420 | return -1; | |
421 | score += 4; | |
422 | } | |
423 | ||
424 | if (sk->sk_bound_dev_if) { | |
425 | if (sk->sk_bound_dev_if != dif) | |
426 | return -1; | |
427 | score += 4; | |
428 | } | |
70da268b ED |
429 | if (sk->sk_incoming_cpu == raw_smp_processor_id()) |
430 | score++; | |
645ca708 ED |
431 | return score; |
432 | } | |
433 | ||
5051ebd2 ED |
434 | /* |
435 | * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num) | |
436 | */ | |
5051ebd2 ED |
437 | static inline int compute_score2(struct sock *sk, struct net *net, |
438 | __be32 saddr, __be16 sport, | |
439 | __be32 daddr, unsigned int hnum, int dif) | |
440 | { | |
60c04aec JP |
441 | int score; |
442 | struct inet_sock *inet; | |
5051ebd2 | 443 | |
60c04aec JP |
444 | if (!net_eq(sock_net(sk), net) || |
445 | ipv6_only_sock(sk)) | |
446 | return -1; | |
5051ebd2 | 447 | |
60c04aec JP |
448 | inet = inet_sk(sk); |
449 | ||
450 | if (inet->inet_rcv_saddr != daddr || | |
451 | inet->inet_num != hnum) | |
452 | return -1; | |
453 | ||
454 | score = (sk->sk_family == PF_INET) ? 2 : 1; | |
455 | ||
456 | if (inet->inet_daddr) { | |
457 | if (inet->inet_daddr != saddr) | |
5051ebd2 | 458 | return -1; |
60c04aec JP |
459 | score += 4; |
460 | } | |
461 | ||
462 | if (inet->inet_dport) { | |
463 | if (inet->inet_dport != sport) | |
5051ebd2 | 464 | return -1; |
60c04aec JP |
465 | score += 4; |
466 | } | |
5051ebd2 | 467 | |
60c04aec JP |
468 | if (sk->sk_bound_dev_if) { |
469 | if (sk->sk_bound_dev_if != dif) | |
470 | return -1; | |
471 | score += 4; | |
5051ebd2 | 472 | } |
60c04aec | 473 | |
70da268b ED |
474 | if (sk->sk_incoming_cpu == raw_smp_processor_id()) |
475 | score++; | |
476 | ||
5051ebd2 ED |
477 | return score; |
478 | } | |
479 | ||
6eada011 ED |
480 | static u32 udp_ehashfn(const struct net *net, const __be32 laddr, |
481 | const __u16 lport, const __be32 faddr, | |
482 | const __be16 fport) | |
65cd8033 | 483 | { |
1bbdceef HFS |
484 | static u32 udp_ehash_secret __read_mostly; |
485 | ||
486 | net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret)); | |
487 | ||
65cd8033 | 488 | return __inet_ehashfn(laddr, lport, faddr, fport, |
1bbdceef | 489 | udp_ehash_secret + net_hash_mix(net)); |
65cd8033 HFS |
490 | } |
491 | ||
5051ebd2 ED |
492 | /* called with read_rcu_lock() */ |
493 | static struct sock *udp4_lib_lookup2(struct net *net, | |
494 | __be32 saddr, __be16 sport, | |
495 | __be32 daddr, unsigned int hnum, int dif, | |
1134158b CG |
496 | struct udp_hslot *hslot2, unsigned int slot2, |
497 | struct sk_buff *skb) | |
5051ebd2 ED |
498 | { |
499 | struct sock *sk, *result; | |
500 | struct hlist_nulls_node *node; | |
ba418fa3 TH |
501 | int score, badness, matches = 0, reuseport = 0; |
502 | u32 hash = 0; | |
5051ebd2 ED |
503 | |
504 | begin: | |
505 | result = NULL; | |
ba418fa3 | 506 | badness = 0; |
5051ebd2 ED |
507 | udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { |
508 | score = compute_score2(sk, net, saddr, sport, | |
509 | daddr, hnum, dif); | |
510 | if (score > badness) { | |
511 | result = sk; | |
512 | badness = score; | |
ba418fa3 TH |
513 | reuseport = sk->sk_reuseport; |
514 | if (reuseport) { | |
e32ea7e7 | 515 | struct sock *sk2; |
65cd8033 HFS |
516 | hash = udp_ehashfn(net, daddr, hnum, |
517 | saddr, sport); | |
1134158b CG |
518 | sk2 = reuseport_select_sock(sk, hash, skb, |
519 | sizeof(struct udphdr)); | |
e32ea7e7 CG |
520 | if (sk2) { |
521 | result = sk2; | |
522 | goto found; | |
523 | } | |
ba418fa3 TH |
524 | matches = 1; |
525 | } | |
526 | } else if (score == badness && reuseport) { | |
527 | matches++; | |
8fc54f68 | 528 | if (reciprocal_scale(hash, matches) == 0) |
ba418fa3 TH |
529 | result = sk; |
530 | hash = next_pseudo_random32(hash); | |
5051ebd2 ED |
531 | } |
532 | } | |
533 | /* | |
534 | * if the nulls value we got at the end of this lookup is | |
535 | * not the expected one, we must restart lookup. | |
536 | * We probably met an item that was moved to another chain. | |
537 | */ | |
538 | if (get_nulls_value(node) != slot2) | |
539 | goto begin; | |
5051ebd2 | 540 | if (result) { |
e32ea7e7 | 541 | found: |
c31504dc | 542 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
5051ebd2 ED |
543 | result = NULL; |
544 | else if (unlikely(compute_score2(result, net, saddr, sport, | |
545 | daddr, hnum, dif) < badness)) { | |
546 | sock_put(result); | |
547 | goto begin; | |
548 | } | |
549 | } | |
550 | return result; | |
551 | } | |
552 | ||
db8dac20 DM |
553 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
554 | * harder than this. -DaveM | |
555 | */ | |
fce82338 | 556 | struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, |
db8dac20 | 557 | __be16 sport, __be32 daddr, __be16 dport, |
538950a1 | 558 | int dif, struct udp_table *udptable, struct sk_buff *skb) |
db8dac20 | 559 | { |
271b72c7 | 560 | struct sock *sk, *result; |
88ab1932 | 561 | struct hlist_nulls_node *node; |
db8dac20 | 562 | unsigned short hnum = ntohs(dport); |
5051ebd2 ED |
563 | unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); |
564 | struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; | |
ba418fa3 TH |
565 | int score, badness, matches = 0, reuseport = 0; |
566 | u32 hash = 0; | |
645ca708 | 567 | |
271b72c7 | 568 | rcu_read_lock(); |
5051ebd2 ED |
569 | if (hslot->count > 10) { |
570 | hash2 = udp4_portaddr_hash(net, daddr, hnum); | |
571 | slot2 = hash2 & udptable->mask; | |
572 | hslot2 = &udptable->hash2[slot2]; | |
573 | if (hslot->count < hslot2->count) | |
574 | goto begin; | |
575 | ||
576 | result = udp4_lib_lookup2(net, saddr, sport, | |
577 | daddr, hnum, dif, | |
1134158b | 578 | hslot2, slot2, skb); |
5051ebd2 | 579 | if (!result) { |
0eae88f3 | 580 | hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum); |
5051ebd2 ED |
581 | slot2 = hash2 & udptable->mask; |
582 | hslot2 = &udptable->hash2[slot2]; | |
583 | if (hslot->count < hslot2->count) | |
584 | goto begin; | |
585 | ||
1223c67c | 586 | result = udp4_lib_lookup2(net, saddr, sport, |
0eae88f3 | 587 | htonl(INADDR_ANY), hnum, dif, |
1134158b | 588 | hslot2, slot2, skb); |
5051ebd2 ED |
589 | } |
590 | rcu_read_unlock(); | |
591 | return result; | |
592 | } | |
271b72c7 ED |
593 | begin: |
594 | result = NULL; | |
ba418fa3 | 595 | badness = 0; |
88ab1932 | 596 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { |
645ca708 ED |
597 | score = compute_score(sk, net, saddr, hnum, sport, |
598 | daddr, dport, dif); | |
599 | if (score > badness) { | |
600 | result = sk; | |
601 | badness = score; | |
ba418fa3 TH |
602 | reuseport = sk->sk_reuseport; |
603 | if (reuseport) { | |
e32ea7e7 | 604 | struct sock *sk2; |
65cd8033 HFS |
605 | hash = udp_ehashfn(net, daddr, hnum, |
606 | saddr, sport); | |
538950a1 CG |
607 | sk2 = reuseport_select_sock(sk, hash, skb, |
608 | sizeof(struct udphdr)); | |
e32ea7e7 CG |
609 | if (sk2) { |
610 | result = sk2; | |
611 | goto found; | |
612 | } | |
ba418fa3 TH |
613 | matches = 1; |
614 | } | |
615 | } else if (score == badness && reuseport) { | |
616 | matches++; | |
8fc54f68 | 617 | if (reciprocal_scale(hash, matches) == 0) |
ba418fa3 TH |
618 | result = sk; |
619 | hash = next_pseudo_random32(hash); | |
db8dac20 DM |
620 | } |
621 | } | |
88ab1932 ED |
622 | /* |
623 | * if the nulls value we got at the end of this lookup is | |
624 | * not the expected one, we must restart lookup. | |
625 | * We probably met an item that was moved to another chain. | |
626 | */ | |
5051ebd2 | 627 | if (get_nulls_value(node) != slot) |
88ab1932 ED |
628 | goto begin; |
629 | ||
271b72c7 | 630 | if (result) { |
e32ea7e7 | 631 | found: |
c31504dc | 632 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
271b72c7 ED |
633 | result = NULL; |
634 | else if (unlikely(compute_score(result, net, saddr, hnum, sport, | |
635 | daddr, dport, dif) < badness)) { | |
636 | sock_put(result); | |
637 | goto begin; | |
638 | } | |
639 | } | |
640 | rcu_read_unlock(); | |
db8dac20 DM |
641 | return result; |
642 | } | |
fce82338 | 643 | EXPORT_SYMBOL_GPL(__udp4_lib_lookup); |
db8dac20 | 644 | |
607c4aaf KK |
645 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
646 | __be16 sport, __be16 dport, | |
645ca708 | 647 | struct udp_table *udptable) |
607c4aaf KK |
648 | { |
649 | const struct iphdr *iph = ip_hdr(skb); | |
650 | ||
8afdd99a ED |
651 | return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport, |
652 | iph->daddr, dport, inet_iif(skb), | |
538950a1 | 653 | udptable, skb); |
607c4aaf KK |
654 | } |
655 | ||
bcd41303 KK |
656 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
657 | __be32 daddr, __be16 dport, int dif) | |
658 | { | |
e32ea7e7 | 659 | return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, |
538950a1 | 660 | &udp_table, NULL); |
bcd41303 KK |
661 | } |
662 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
663 | ||
421b3885 SB |
664 | static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk, |
665 | __be16 loc_port, __be32 loc_addr, | |
666 | __be16 rmt_port, __be32 rmt_addr, | |
667 | int dif, unsigned short hnum) | |
668 | { | |
669 | struct inet_sock *inet = inet_sk(sk); | |
670 | ||
671 | if (!net_eq(sock_net(sk), net) || | |
672 | udp_sk(sk)->udp_port_hash != hnum || | |
673 | (inet->inet_daddr && inet->inet_daddr != rmt_addr) || | |
674 | (inet->inet_dport != rmt_port && inet->inet_dport) || | |
675 | (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) || | |
676 | ipv6_only_sock(sk) || | |
677 | (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) | |
678 | return false; | |
679 | if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif)) | |
680 | return false; | |
681 | return true; | |
682 | } | |
683 | ||
db8dac20 DM |
684 | /* |
685 | * This routine is called by the ICMP module when it gets some | |
686 | * sort of error condition. If err < 0 then the socket should | |
687 | * be closed and the error returned to the user. If err > 0 | |
688 | * it's just the icmp type << 8 | icmp code. | |
689 | * Header points to the ip header of the error packet. We move | |
690 | * on past this. Then (as it used to claim before adjustment) | |
691 | * header points to the first 8 bytes of the udp header. We need | |
692 | * to find the appropriate port. | |
693 | */ | |
694 | ||
645ca708 | 695 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
696 | { |
697 | struct inet_sock *inet; | |
b71d1d42 | 698 | const struct iphdr *iph = (const struct iphdr *)skb->data; |
c482c568 | 699 | struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2)); |
db8dac20 DM |
700 | const int type = icmp_hdr(skb)->type; |
701 | const int code = icmp_hdr(skb)->code; | |
702 | struct sock *sk; | |
703 | int harderr; | |
704 | int err; | |
fd54d716 | 705 | struct net *net = dev_net(skb->dev); |
db8dac20 | 706 | |
fd54d716 | 707 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
538950a1 CG |
708 | iph->saddr, uh->source, skb->dev->ifindex, udptable, |
709 | NULL); | |
51456b29 | 710 | if (!sk) { |
dcfc23ca | 711 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
712 | return; /* No socket for error */ |
713 | } | |
714 | ||
715 | err = 0; | |
716 | harderr = 0; | |
717 | inet = inet_sk(sk); | |
718 | ||
719 | switch (type) { | |
720 | default: | |
721 | case ICMP_TIME_EXCEEDED: | |
722 | err = EHOSTUNREACH; | |
723 | break; | |
724 | case ICMP_SOURCE_QUENCH: | |
725 | goto out; | |
726 | case ICMP_PARAMETERPROB: | |
727 | err = EPROTO; | |
728 | harderr = 1; | |
729 | break; | |
730 | case ICMP_DEST_UNREACH: | |
731 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
36393395 | 732 | ipv4_sk_update_pmtu(skb, sk, info); |
db8dac20 DM |
733 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { |
734 | err = EMSGSIZE; | |
735 | harderr = 1; | |
736 | break; | |
737 | } | |
738 | goto out; | |
739 | } | |
740 | err = EHOSTUNREACH; | |
741 | if (code <= NR_ICMP_UNREACH) { | |
742 | harderr = icmp_err_convert[code].fatal; | |
743 | err = icmp_err_convert[code].errno; | |
744 | } | |
745 | break; | |
55be7a9c DM |
746 | case ICMP_REDIRECT: |
747 | ipv4_sk_redirect(skb, sk); | |
1a462d18 | 748 | goto out; |
db8dac20 DM |
749 | } |
750 | ||
751 | /* | |
752 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
753 | * 4.1.3.3. | |
754 | */ | |
755 | if (!inet->recverr) { | |
756 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
757 | goto out; | |
b1faf566 | 758 | } else |
c482c568 | 759 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1)); |
b1faf566 | 760 | |
db8dac20 DM |
761 | sk->sk_err = err; |
762 | sk->sk_error_report(sk); | |
763 | out: | |
764 | sock_put(sk); | |
765 | } | |
766 | ||
767 | void udp_err(struct sk_buff *skb, u32 info) | |
768 | { | |
645ca708 | 769 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
770 | } |
771 | ||
772 | /* | |
773 | * Throw away all pending data and cancel the corking. Socket is locked. | |
774 | */ | |
36d926b9 | 775 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
776 | { |
777 | struct udp_sock *up = udp_sk(sk); | |
778 | ||
779 | if (up->pending) { | |
780 | up->len = 0; | |
781 | up->pending = 0; | |
782 | ip_flush_pending_frames(sk); | |
783 | } | |
784 | } | |
36d926b9 | 785 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
786 | |
787 | /** | |
f6b9664f | 788 | * udp4_hwcsum - handle outgoing HW checksumming |
db8dac20 DM |
789 | * @skb: sk_buff containing the filled-in UDP header |
790 | * (checksum field must be zeroed out) | |
f6b9664f HX |
791 | * @src: source IP address |
792 | * @dst: destination IP address | |
db8dac20 | 793 | */ |
c26bf4a5 | 794 | void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst) |
db8dac20 | 795 | { |
db8dac20 | 796 | struct udphdr *uh = udp_hdr(skb); |
f6b9664f HX |
797 | int offset = skb_transport_offset(skb); |
798 | int len = skb->len - offset; | |
799 | int hlen = len; | |
db8dac20 DM |
800 | __wsum csum = 0; |
801 | ||
ebbe495f | 802 | if (!skb_has_frag_list(skb)) { |
db8dac20 DM |
803 | /* |
804 | * Only one fragment on the socket. | |
805 | */ | |
806 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
807 | skb->csum_offset = offsetof(struct udphdr, check); | |
f6b9664f HX |
808 | uh->check = ~csum_tcpudp_magic(src, dst, len, |
809 | IPPROTO_UDP, 0); | |
db8dac20 | 810 | } else { |
ebbe495f WC |
811 | struct sk_buff *frags; |
812 | ||
db8dac20 DM |
813 | /* |
814 | * HW-checksum won't work as there are two or more | |
815 | * fragments on the socket so that all csums of sk_buffs | |
816 | * should be together | |
817 | */ | |
ebbe495f | 818 | skb_walk_frags(skb, frags) { |
f6b9664f HX |
819 | csum = csum_add(csum, frags->csum); |
820 | hlen -= frags->len; | |
ebbe495f | 821 | } |
db8dac20 | 822 | |
f6b9664f | 823 | csum = skb_checksum(skb, offset, hlen, csum); |
db8dac20 DM |
824 | skb->ip_summed = CHECKSUM_NONE; |
825 | ||
db8dac20 DM |
826 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); |
827 | if (uh->check == 0) | |
828 | uh->check = CSUM_MANGLED_0; | |
829 | } | |
830 | } | |
c26bf4a5 | 831 | EXPORT_SYMBOL_GPL(udp4_hwcsum); |
db8dac20 | 832 | |
af5fcba7 TH |
833 | /* Function to set UDP checksum for an IPv4 UDP packet. This is intended |
834 | * for the simple case like when setting the checksum for a UDP tunnel. | |
835 | */ | |
836 | void udp_set_csum(bool nocheck, struct sk_buff *skb, | |
837 | __be32 saddr, __be32 daddr, int len) | |
838 | { | |
839 | struct udphdr *uh = udp_hdr(skb); | |
840 | ||
841 | if (nocheck) | |
842 | uh->check = 0; | |
843 | else if (skb_is_gso(skb)) | |
844 | uh->check = ~udp_v4_check(len, saddr, daddr, 0); | |
845 | else if (skb_dst(skb) && skb_dst(skb)->dev && | |
c8cd0989 TH |
846 | (skb_dst(skb)->dev->features & |
847 | (NETIF_F_IP_CSUM | NETIF_F_HW_CSUM))) { | |
af5fcba7 TH |
848 | |
849 | BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL); | |
850 | ||
851 | skb->ip_summed = CHECKSUM_PARTIAL; | |
852 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
853 | skb->csum_offset = offsetof(struct udphdr, check); | |
854 | uh->check = ~udp_v4_check(len, saddr, daddr, 0); | |
855 | } else { | |
856 | __wsum csum; | |
857 | ||
858 | BUG_ON(skb->ip_summed == CHECKSUM_PARTIAL); | |
859 | ||
860 | uh->check = 0; | |
861 | csum = skb_checksum(skb, 0, len, 0); | |
862 | uh->check = udp_v4_check(len, saddr, daddr, csum); | |
863 | if (uh->check == 0) | |
864 | uh->check = CSUM_MANGLED_0; | |
865 | ||
866 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
867 | } | |
868 | } | |
869 | EXPORT_SYMBOL(udp_set_csum); | |
870 | ||
79ab0531 | 871 | static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4) |
db8dac20 | 872 | { |
f6b9664f | 873 | struct sock *sk = skb->sk; |
db8dac20 | 874 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
875 | struct udphdr *uh; |
876 | int err = 0; | |
877 | int is_udplite = IS_UDPLITE(sk); | |
f6b9664f HX |
878 | int offset = skb_transport_offset(skb); |
879 | int len = skb->len - offset; | |
db8dac20 DM |
880 | __wsum csum = 0; |
881 | ||
db8dac20 DM |
882 | /* |
883 | * Create a UDP header | |
884 | */ | |
885 | uh = udp_hdr(skb); | |
f6b9664f | 886 | uh->source = inet->inet_sport; |
79ab0531 | 887 | uh->dest = fl4->fl4_dport; |
f6b9664f | 888 | uh->len = htons(len); |
db8dac20 DM |
889 | uh->check = 0; |
890 | ||
891 | if (is_udplite) /* UDP-Lite */ | |
f6b9664f | 892 | csum = udplite_csum(skb); |
db8dac20 | 893 | |
28448b80 | 894 | else if (sk->sk_no_check_tx) { /* UDP csum disabled */ |
db8dac20 DM |
895 | |
896 | skb->ip_summed = CHECKSUM_NONE; | |
897 | goto send; | |
898 | ||
899 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
900 | ||
79ab0531 | 901 | udp4_hwcsum(skb, fl4->saddr, fl4->daddr); |
db8dac20 DM |
902 | goto send; |
903 | ||
f6b9664f HX |
904 | } else |
905 | csum = udp_csum(skb); | |
db8dac20 DM |
906 | |
907 | /* add protocol-dependent pseudo-header */ | |
79ab0531 | 908 | uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len, |
c482c568 | 909 | sk->sk_protocol, csum); |
db8dac20 DM |
910 | if (uh->check == 0) |
911 | uh->check = CSUM_MANGLED_0; | |
912 | ||
913 | send: | |
b5ec8eea | 914 | err = ip_send_skb(sock_net(sk), skb); |
6ce9e7b5 ED |
915 | if (err) { |
916 | if (err == -ENOBUFS && !inet->recverr) { | |
917 | UDP_INC_STATS_USER(sock_net(sk), | |
918 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
919 | err = 0; | |
920 | } | |
921 | } else | |
922 | UDP_INC_STATS_USER(sock_net(sk), | |
923 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
f6b9664f HX |
924 | return err; |
925 | } | |
926 | ||
927 | /* | |
928 | * Push out all pending data as one UDP datagram. Socket is locked. | |
929 | */ | |
8822b64a | 930 | int udp_push_pending_frames(struct sock *sk) |
f6b9664f HX |
931 | { |
932 | struct udp_sock *up = udp_sk(sk); | |
933 | struct inet_sock *inet = inet_sk(sk); | |
b6f21b26 | 934 | struct flowi4 *fl4 = &inet->cork.fl.u.ip4; |
f6b9664f HX |
935 | struct sk_buff *skb; |
936 | int err = 0; | |
937 | ||
77968b78 | 938 | skb = ip_finish_skb(sk, fl4); |
f6b9664f HX |
939 | if (!skb) |
940 | goto out; | |
941 | ||
79ab0531 | 942 | err = udp_send_skb(skb, fl4); |
f6b9664f | 943 | |
db8dac20 DM |
944 | out: |
945 | up->len = 0; | |
946 | up->pending = 0; | |
db8dac20 DM |
947 | return err; |
948 | } | |
8822b64a | 949 | EXPORT_SYMBOL(udp_push_pending_frames); |
db8dac20 | 950 | |
1b784140 | 951 | int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
db8dac20 DM |
952 | { |
953 | struct inet_sock *inet = inet_sk(sk); | |
954 | struct udp_sock *up = udp_sk(sk); | |
e474995f | 955 | struct flowi4 fl4_stack; |
b6f21b26 | 956 | struct flowi4 *fl4; |
db8dac20 DM |
957 | int ulen = len; |
958 | struct ipcm_cookie ipc; | |
959 | struct rtable *rt = NULL; | |
960 | int free = 0; | |
961 | int connected = 0; | |
962 | __be32 daddr, faddr, saddr; | |
963 | __be16 dport; | |
964 | u8 tos; | |
965 | int err, is_udplite = IS_UDPLITE(sk); | |
966 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
967 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
903ab86d | 968 | struct sk_buff *skb; |
f6d8bd05 | 969 | struct ip_options_data opt_copy; |
db8dac20 DM |
970 | |
971 | if (len > 0xFFFF) | |
972 | return -EMSGSIZE; | |
973 | ||
974 | /* | |
975 | * Check the flags. | |
976 | */ | |
977 | ||
c482c568 | 978 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */ |
db8dac20 DM |
979 | return -EOPNOTSUPP; |
980 | ||
981 | ipc.opt = NULL; | |
2244d07b | 982 | ipc.tx_flags = 0; |
aa661581 FF |
983 | ipc.ttl = 0; |
984 | ipc.tos = -1; | |
db8dac20 | 985 | |
903ab86d HX |
986 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
987 | ||
f5fca608 | 988 | fl4 = &inet->cork.fl.u.ip4; |
db8dac20 DM |
989 | if (up->pending) { |
990 | /* | |
991 | * There are pending frames. | |
992 | * The socket lock must be held while it's corked. | |
993 | */ | |
994 | lock_sock(sk); | |
995 | if (likely(up->pending)) { | |
996 | if (unlikely(up->pending != AF_INET)) { | |
997 | release_sock(sk); | |
998 | return -EINVAL; | |
999 | } | |
1000 | goto do_append_data; | |
1001 | } | |
1002 | release_sock(sk); | |
1003 | } | |
1004 | ulen += sizeof(struct udphdr); | |
1005 | ||
1006 | /* | |
1007 | * Get and verify the address. | |
1008 | */ | |
1009 | if (msg->msg_name) { | |
342dfc30 | 1010 | DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); |
db8dac20 DM |
1011 | if (msg->msg_namelen < sizeof(*usin)) |
1012 | return -EINVAL; | |
1013 | if (usin->sin_family != AF_INET) { | |
1014 | if (usin->sin_family != AF_UNSPEC) | |
1015 | return -EAFNOSUPPORT; | |
1016 | } | |
1017 | ||
1018 | daddr = usin->sin_addr.s_addr; | |
1019 | dport = usin->sin_port; | |
1020 | if (dport == 0) | |
1021 | return -EINVAL; | |
1022 | } else { | |
1023 | if (sk->sk_state != TCP_ESTABLISHED) | |
1024 | return -EDESTADDRREQ; | |
c720c7e8 ED |
1025 | daddr = inet->inet_daddr; |
1026 | dport = inet->inet_dport; | |
db8dac20 DM |
1027 | /* Open fast path for connected socket. |
1028 | Route will not be used, if at least one option is set. | |
1029 | */ | |
1030 | connected = 1; | |
1031 | } | |
c720c7e8 | 1032 | ipc.addr = inet->inet_saddr; |
db8dac20 DM |
1033 | |
1034 | ipc.oif = sk->sk_bound_dev_if; | |
bf84a010 DB |
1035 | |
1036 | sock_tx_timestamp(sk, &ipc.tx_flags); | |
1037 | ||
db8dac20 | 1038 | if (msg->msg_controllen) { |
c8e6ad08 HFS |
1039 | err = ip_cmsg_send(sock_net(sk), msg, &ipc, |
1040 | sk->sk_family == AF_INET6); | |
db8dac20 DM |
1041 | if (err) |
1042 | return err; | |
1043 | if (ipc.opt) | |
1044 | free = 1; | |
1045 | connected = 0; | |
1046 | } | |
f6d8bd05 ED |
1047 | if (!ipc.opt) { |
1048 | struct ip_options_rcu *inet_opt; | |
1049 | ||
1050 | rcu_read_lock(); | |
1051 | inet_opt = rcu_dereference(inet->inet_opt); | |
1052 | if (inet_opt) { | |
1053 | memcpy(&opt_copy, inet_opt, | |
1054 | sizeof(*inet_opt) + inet_opt->opt.optlen); | |
1055 | ipc.opt = &opt_copy.opt; | |
1056 | } | |
1057 | rcu_read_unlock(); | |
1058 | } | |
db8dac20 DM |
1059 | |
1060 | saddr = ipc.addr; | |
1061 | ipc.addr = faddr = daddr; | |
1062 | ||
f6d8bd05 | 1063 | if (ipc.opt && ipc.opt->opt.srr) { |
db8dac20 DM |
1064 | if (!daddr) |
1065 | return -EINVAL; | |
f6d8bd05 | 1066 | faddr = ipc.opt->opt.faddr; |
db8dac20 DM |
1067 | connected = 0; |
1068 | } | |
aa661581 | 1069 | tos = get_rttos(&ipc, inet); |
db8dac20 DM |
1070 | if (sock_flag(sk, SOCK_LOCALROUTE) || |
1071 | (msg->msg_flags & MSG_DONTROUTE) || | |
f6d8bd05 | 1072 | (ipc.opt && ipc.opt->opt.is_strictroute)) { |
db8dac20 DM |
1073 | tos |= RTO_ONLINK; |
1074 | connected = 0; | |
1075 | } | |
1076 | ||
1077 | if (ipv4_is_multicast(daddr)) { | |
1078 | if (!ipc.oif) | |
1079 | ipc.oif = inet->mc_index; | |
1080 | if (!saddr) | |
1081 | saddr = inet->mc_addr; | |
1082 | connected = 0; | |
76e21053 EH |
1083 | } else if (!ipc.oif) |
1084 | ipc.oif = inet->uc_index; | |
db8dac20 DM |
1085 | |
1086 | if (connected) | |
c482c568 | 1087 | rt = (struct rtable *)sk_dst_check(sk, 0); |
db8dac20 | 1088 | |
51456b29 | 1089 | if (!rt) { |
84a3aa00 | 1090 | struct net *net = sock_net(sk); |
9a24abfa | 1091 | __u8 flow_flags = inet_sk_flowi_flags(sk); |
84a3aa00 | 1092 | |
e474995f | 1093 | fl4 = &fl4_stack; |
9a24abfa | 1094 | |
e474995f | 1095 | flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos, |
c0951cbc | 1096 | RT_SCOPE_UNIVERSE, sk->sk_protocol, |
9a24abfa | 1097 | flow_flags, |
c0951cbc DM |
1098 | faddr, saddr, dport, inet->inet_sport); |
1099 | ||
b5bdacf3 DA |
1100 | if (!saddr && ipc.oif) { |
1101 | err = l3mdev_get_saddr(net, ipc.oif, fl4); | |
1102 | if (err < 0) | |
1103 | goto out; | |
1104 | } | |
8cbb512c | 1105 | |
e474995f DM |
1106 | security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); |
1107 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe DM |
1108 | if (IS_ERR(rt)) { |
1109 | err = PTR_ERR(rt); | |
06dc94b1 | 1110 | rt = NULL; |
db8dac20 | 1111 | if (err == -ENETUNREACH) |
f1d8cba6 | 1112 | IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
1113 | goto out; |
1114 | } | |
1115 | ||
1116 | err = -EACCES; | |
1117 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
1118 | !sock_flag(sk, SOCK_BROADCAST)) | |
1119 | goto out; | |
1120 | if (connected) | |
d8d1f30b | 1121 | sk_dst_set(sk, dst_clone(&rt->dst)); |
db8dac20 DM |
1122 | } |
1123 | ||
1124 | if (msg->msg_flags&MSG_CONFIRM) | |
1125 | goto do_confirm; | |
1126 | back_from_confirm: | |
1127 | ||
e474995f | 1128 | saddr = fl4->saddr; |
db8dac20 | 1129 | if (!ipc.addr) |
e474995f | 1130 | daddr = ipc.addr = fl4->daddr; |
db8dac20 | 1131 | |
903ab86d HX |
1132 | /* Lockless fast path for the non-corking case. */ |
1133 | if (!corkreq) { | |
f69e6d13 | 1134 | skb = ip_make_skb(sk, fl4, getfrag, msg, ulen, |
903ab86d HX |
1135 | sizeof(struct udphdr), &ipc, &rt, |
1136 | msg->msg_flags); | |
1137 | err = PTR_ERR(skb); | |
50c3a487 | 1138 | if (!IS_ERR_OR_NULL(skb)) |
79ab0531 | 1139 | err = udp_send_skb(skb, fl4); |
903ab86d HX |
1140 | goto out; |
1141 | } | |
1142 | ||
db8dac20 DM |
1143 | lock_sock(sk); |
1144 | if (unlikely(up->pending)) { | |
1145 | /* The socket is already corked while preparing it. */ | |
1146 | /* ... which is an evident application bug. --ANK */ | |
1147 | release_sock(sk); | |
1148 | ||
ba7a46f1 | 1149 | net_dbg_ratelimited("cork app bug 2\n"); |
db8dac20 DM |
1150 | err = -EINVAL; |
1151 | goto out; | |
1152 | } | |
1153 | /* | |
1154 | * Now cork the socket to pend data. | |
1155 | */ | |
b6f21b26 DM |
1156 | fl4 = &inet->cork.fl.u.ip4; |
1157 | fl4->daddr = daddr; | |
1158 | fl4->saddr = saddr; | |
9cce96df DM |
1159 | fl4->fl4_dport = dport; |
1160 | fl4->fl4_sport = inet->inet_sport; | |
db8dac20 DM |
1161 | up->pending = AF_INET; |
1162 | ||
1163 | do_append_data: | |
1164 | up->len += ulen; | |
f69e6d13 | 1165 | err = ip_append_data(sk, fl4, getfrag, msg, ulen, |
f5fca608 DM |
1166 | sizeof(struct udphdr), &ipc, &rt, |
1167 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | |
db8dac20 DM |
1168 | if (err) |
1169 | udp_flush_pending_frames(sk); | |
1170 | else if (!corkreq) | |
1171 | err = udp_push_pending_frames(sk); | |
1172 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
1173 | up->pending = 0; | |
1174 | release_sock(sk); | |
1175 | ||
1176 | out: | |
1177 | ip_rt_put(rt); | |
1178 | if (free) | |
1179 | kfree(ipc.opt); | |
1180 | if (!err) | |
1181 | return len; | |
1182 | /* | |
1183 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1184 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1185 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1186 | * things). We could add another new stat but at least for now that | |
1187 | * seems like overkill. | |
1188 | */ | |
1189 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
1190 | UDP_INC_STATS_USER(sock_net(sk), |
1191 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
1192 | } |
1193 | return err; | |
1194 | ||
1195 | do_confirm: | |
d8d1f30b | 1196 | dst_confirm(&rt->dst); |
db8dac20 DM |
1197 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1198 | goto back_from_confirm; | |
1199 | err = 0; | |
1200 | goto out; | |
1201 | } | |
c482c568 | 1202 | EXPORT_SYMBOL(udp_sendmsg); |
db8dac20 DM |
1203 | |
1204 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
1205 | size_t size, int flags) | |
1206 | { | |
f5fca608 | 1207 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
1208 | struct udp_sock *up = udp_sk(sk); |
1209 | int ret; | |
1210 | ||
d3f7d56a SL |
1211 | if (flags & MSG_SENDPAGE_NOTLAST) |
1212 | flags |= MSG_MORE; | |
1213 | ||
db8dac20 DM |
1214 | if (!up->pending) { |
1215 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
1216 | ||
1217 | /* Call udp_sendmsg to specify destination address which | |
1218 | * sendpage interface can't pass. | |
1219 | * This will succeed only when the socket is connected. | |
1220 | */ | |
1b784140 | 1221 | ret = udp_sendmsg(sk, &msg, 0); |
db8dac20 DM |
1222 | if (ret < 0) |
1223 | return ret; | |
1224 | } | |
1225 | ||
1226 | lock_sock(sk); | |
1227 | ||
1228 | if (unlikely(!up->pending)) { | |
1229 | release_sock(sk); | |
1230 | ||
ba7a46f1 | 1231 | net_dbg_ratelimited("udp cork app bug 3\n"); |
db8dac20 DM |
1232 | return -EINVAL; |
1233 | } | |
1234 | ||
f5fca608 DM |
1235 | ret = ip_append_page(sk, &inet->cork.fl.u.ip4, |
1236 | page, offset, size, flags); | |
db8dac20 DM |
1237 | if (ret == -EOPNOTSUPP) { |
1238 | release_sock(sk); | |
1239 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
1240 | size, flags); | |
1241 | } | |
1242 | if (ret < 0) { | |
1243 | udp_flush_pending_frames(sk); | |
1244 | goto out; | |
1245 | } | |
1246 | ||
1247 | up->len += size; | |
1248 | if (!(up->corkflag || (flags&MSG_MORE))) | |
1249 | ret = udp_push_pending_frames(sk); | |
1250 | if (!ret) | |
1251 | ret = size; | |
1252 | out: | |
1253 | release_sock(sk); | |
1254 | return ret; | |
1255 | } | |
1256 | ||
85584672 ED |
1257 | /** |
1258 | * first_packet_length - return length of first packet in receive queue | |
1259 | * @sk: socket | |
1260 | * | |
1261 | * Drops all bad checksum frames, until a valid one is found. | |
1262 | * Returns the length of found skb, or 0 if none is found. | |
1263 | */ | |
1264 | static unsigned int first_packet_length(struct sock *sk) | |
1265 | { | |
1266 | struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue; | |
1267 | struct sk_buff *skb; | |
1268 | unsigned int res; | |
1269 | ||
1270 | __skb_queue_head_init(&list_kill); | |
1271 | ||
1272 | spin_lock_bh(&rcvq->lock); | |
1273 | while ((skb = skb_peek(rcvq)) != NULL && | |
1274 | udp_lib_checksum_complete(skb)) { | |
6a5dc9e5 ED |
1275 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, |
1276 | IS_UDPLITE(sk)); | |
85584672 ED |
1277 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, |
1278 | IS_UDPLITE(sk)); | |
8edf19c2 | 1279 | atomic_inc(&sk->sk_drops); |
85584672 ED |
1280 | __skb_unlink(skb, rcvq); |
1281 | __skb_queue_tail(&list_kill, skb); | |
1282 | } | |
1283 | res = skb ? skb->len : 0; | |
1284 | spin_unlock_bh(&rcvq->lock); | |
1285 | ||
1286 | if (!skb_queue_empty(&list_kill)) { | |
8a74ad60 ED |
1287 | bool slow = lock_sock_fast(sk); |
1288 | ||
85584672 ED |
1289 | __skb_queue_purge(&list_kill); |
1290 | sk_mem_reclaim_partial(sk); | |
8a74ad60 | 1291 | unlock_sock_fast(sk, slow); |
85584672 ED |
1292 | } |
1293 | return res; | |
1294 | } | |
1295 | ||
1da177e4 LT |
1296 | /* |
1297 | * IOCTL requests applicable to the UDP protocol | |
1298 | */ | |
e905a9ed | 1299 | |
1da177e4 LT |
1300 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
1301 | { | |
6516c655 SH |
1302 | switch (cmd) { |
1303 | case SIOCOUTQ: | |
1da177e4 | 1304 | { |
31e6d363 ED |
1305 | int amount = sk_wmem_alloc_get(sk); |
1306 | ||
6516c655 SH |
1307 | return put_user(amount, (int __user *)arg); |
1308 | } | |
1da177e4 | 1309 | |
6516c655 SH |
1310 | case SIOCINQ: |
1311 | { | |
85584672 | 1312 | unsigned int amount = first_packet_length(sk); |
6516c655 | 1313 | |
85584672 | 1314 | if (amount) |
6516c655 SH |
1315 | /* |
1316 | * We will only return the amount | |
1317 | * of this packet since that is all | |
1318 | * that will be read. | |
1319 | */ | |
85584672 ED |
1320 | amount -= sizeof(struct udphdr); |
1321 | ||
6516c655 SH |
1322 | return put_user(amount, (int __user *)arg); |
1323 | } | |
1da177e4 | 1324 | |
6516c655 SH |
1325 | default: |
1326 | return -ENOIOCTLCMD; | |
1da177e4 | 1327 | } |
6516c655 SH |
1328 | |
1329 | return 0; | |
1da177e4 | 1330 | } |
c482c568 | 1331 | EXPORT_SYMBOL(udp_ioctl); |
1da177e4 | 1332 | |
db8dac20 DM |
1333 | /* |
1334 | * This should be easy, if there is something there we | |
1335 | * return it, otherwise we block. | |
1336 | */ | |
1337 | ||
1b784140 YX |
1338 | int udp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock, |
1339 | int flags, int *addr_len) | |
db8dac20 DM |
1340 | { |
1341 | struct inet_sock *inet = inet_sk(sk); | |
342dfc30 | 1342 | DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); |
db8dac20 | 1343 | struct sk_buff *skb; |
59c2cdae | 1344 | unsigned int ulen, copied; |
3f518bf7 | 1345 | int peeked, off = 0; |
db8dac20 DM |
1346 | int err; |
1347 | int is_udplite = IS_UDPLITE(sk); | |
197c949e | 1348 | bool checksum_valid = false; |
8a74ad60 | 1349 | bool slow; |
db8dac20 | 1350 | |
db8dac20 | 1351 | if (flags & MSG_ERRQUEUE) |
85fbaa75 | 1352 | return ip_recv_error(sk, msg, len, addr_len); |
db8dac20 DM |
1353 | |
1354 | try_again: | |
1355 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), | |
3f518bf7 | 1356 | &peeked, &off, &err); |
db8dac20 DM |
1357 | if (!skb) |
1358 | goto out; | |
1359 | ||
1360 | ulen = skb->len - sizeof(struct udphdr); | |
59c2cdae DM |
1361 | copied = len; |
1362 | if (copied > ulen) | |
1363 | copied = ulen; | |
1364 | else if (copied < ulen) | |
db8dac20 DM |
1365 | msg->msg_flags |= MSG_TRUNC; |
1366 | ||
1367 | /* | |
1368 | * If checksum is needed at all, try to do it while copying the | |
1369 | * data. If the data is truncated, or if we only want a partial | |
1370 | * coverage checksum (UDP-Lite), do it before the copy. | |
1371 | */ | |
1372 | ||
59c2cdae | 1373 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { |
197c949e ED |
1374 | checksum_valid = !udp_lib_checksum_complete(skb); |
1375 | if (!checksum_valid) | |
db8dac20 DM |
1376 | goto csum_copy_err; |
1377 | } | |
1378 | ||
197c949e | 1379 | if (checksum_valid || skb_csum_unnecessary(skb)) |
51f3d02b DM |
1380 | err = skb_copy_datagram_msg(skb, sizeof(struct udphdr), |
1381 | msg, copied); | |
db8dac20 | 1382 | else { |
227158db AV |
1383 | err = skb_copy_and_csum_datagram_msg(skb, sizeof(struct udphdr), |
1384 | msg); | |
db8dac20 DM |
1385 | |
1386 | if (err == -EINVAL) | |
1387 | goto csum_copy_err; | |
1388 | } | |
1389 | ||
22911fc5 ED |
1390 | if (unlikely(err)) { |
1391 | trace_kfree_skb(skb, udp_recvmsg); | |
979402b1 ED |
1392 | if (!peeked) { |
1393 | atomic_inc(&sk->sk_drops); | |
1394 | UDP_INC_STATS_USER(sock_net(sk), | |
1395 | UDP_MIB_INERRORS, is_udplite); | |
1396 | } | |
db8dac20 | 1397 | goto out_free; |
22911fc5 | 1398 | } |
db8dac20 DM |
1399 | |
1400 | if (!peeked) | |
629ca23c PE |
1401 | UDP_INC_STATS_USER(sock_net(sk), |
1402 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 | 1403 | |
3b885787 | 1404 | sock_recv_ts_and_drops(msg, sk, skb); |
db8dac20 DM |
1405 | |
1406 | /* Copy the address. */ | |
c482c568 | 1407 | if (sin) { |
db8dac20 DM |
1408 | sin->sin_family = AF_INET; |
1409 | sin->sin_port = udp_hdr(skb)->source; | |
1410 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
1411 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
bceaa902 | 1412 | *addr_len = sizeof(*sin); |
db8dac20 DM |
1413 | } |
1414 | if (inet->cmsg_flags) | |
ad6f939a | 1415 | ip_cmsg_recv_offset(msg, skb, sizeof(struct udphdr)); |
db8dac20 | 1416 | |
59c2cdae | 1417 | err = copied; |
db8dac20 DM |
1418 | if (flags & MSG_TRUNC) |
1419 | err = ulen; | |
1420 | ||
1421 | out_free: | |
9d410c79 | 1422 | skb_free_datagram_locked(sk, skb); |
db8dac20 DM |
1423 | out: |
1424 | return err; | |
1425 | ||
1426 | csum_copy_err: | |
8a74ad60 | 1427 | slow = lock_sock_fast(sk); |
6a5dc9e5 ED |
1428 | if (!skb_kill_datagram(sk, skb, flags)) { |
1429 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); | |
629ca23c | 1430 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
6a5dc9e5 | 1431 | } |
8a74ad60 | 1432 | unlock_sock_fast(sk, slow); |
db8dac20 | 1433 | |
beb39db5 ED |
1434 | /* starting over for a new packet, but check if we need to yield */ |
1435 | cond_resched(); | |
9cfaa8de | 1436 | msg->msg_flags &= ~MSG_TRUNC; |
db8dac20 DM |
1437 | goto try_again; |
1438 | } | |
1439 | ||
1da177e4 LT |
1440 | int udp_disconnect(struct sock *sk, int flags) |
1441 | { | |
1442 | struct inet_sock *inet = inet_sk(sk); | |
1443 | /* | |
1444 | * 1003.1g - break association. | |
1445 | */ | |
e905a9ed | 1446 | |
1da177e4 | 1447 | sk->sk_state = TCP_CLOSE; |
c720c7e8 ED |
1448 | inet->inet_daddr = 0; |
1449 | inet->inet_dport = 0; | |
bdeab991 | 1450 | sock_rps_reset_rxhash(sk); |
1da177e4 LT |
1451 | sk->sk_bound_dev_if = 0; |
1452 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
1453 | inet_reset_saddr(sk); | |
1454 | ||
1455 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
1456 | sk->sk_prot->unhash(sk); | |
c720c7e8 | 1457 | inet->inet_sport = 0; |
1da177e4 LT |
1458 | } |
1459 | sk_dst_reset(sk); | |
1460 | return 0; | |
1461 | } | |
c482c568 | 1462 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1463 | |
645ca708 ED |
1464 | void udp_lib_unhash(struct sock *sk) |
1465 | { | |
723b4610 ED |
1466 | if (sk_hashed(sk)) { |
1467 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
512615b6 ED |
1468 | struct udp_hslot *hslot, *hslot2; |
1469 | ||
1470 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1471 | udp_sk(sk)->udp_port_hash); | |
1472 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
645ca708 | 1473 | |
723b4610 | 1474 | spin_lock_bh(&hslot->lock); |
e32ea7e7 CG |
1475 | if (rcu_access_pointer(sk->sk_reuseport_cb)) |
1476 | reuseport_detach_sock(sk); | |
723b4610 | 1477 | if (sk_nulls_del_node_init_rcu(sk)) { |
fdcc8aa9 | 1478 | hslot->count--; |
c720c7e8 | 1479 | inet_sk(sk)->inet_num = 0; |
723b4610 | 1480 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
512615b6 ED |
1481 | |
1482 | spin_lock(&hslot2->lock); | |
1483 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1484 | hslot2->count--; | |
1485 | spin_unlock(&hslot2->lock); | |
723b4610 ED |
1486 | } |
1487 | spin_unlock_bh(&hslot->lock); | |
645ca708 | 1488 | } |
645ca708 ED |
1489 | } |
1490 | EXPORT_SYMBOL(udp_lib_unhash); | |
1491 | ||
719f8358 ED |
1492 | /* |
1493 | * inet_rcv_saddr was changed, we must rehash secondary hash | |
1494 | */ | |
1495 | void udp_lib_rehash(struct sock *sk, u16 newhash) | |
1496 | { | |
1497 | if (sk_hashed(sk)) { | |
1498 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
1499 | struct udp_hslot *hslot, *hslot2, *nhslot2; | |
1500 | ||
1501 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
1502 | nhslot2 = udp_hashslot2(udptable, newhash); | |
1503 | udp_sk(sk)->udp_portaddr_hash = newhash; | |
e32ea7e7 CG |
1504 | |
1505 | if (hslot2 != nhslot2 || | |
1506 | rcu_access_pointer(sk->sk_reuseport_cb)) { | |
719f8358 ED |
1507 | hslot = udp_hashslot(udptable, sock_net(sk), |
1508 | udp_sk(sk)->udp_port_hash); | |
1509 | /* we must lock primary chain too */ | |
1510 | spin_lock_bh(&hslot->lock); | |
e32ea7e7 CG |
1511 | if (rcu_access_pointer(sk->sk_reuseport_cb)) |
1512 | reuseport_detach_sock(sk); | |
1513 | ||
1514 | if (hslot2 != nhslot2) { | |
1515 | spin_lock(&hslot2->lock); | |
1516 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1517 | hslot2->count--; | |
1518 | spin_unlock(&hslot2->lock); | |
1519 | ||
1520 | spin_lock(&nhslot2->lock); | |
1521 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
1522 | &nhslot2->head); | |
1523 | nhslot2->count++; | |
1524 | spin_unlock(&nhslot2->lock); | |
1525 | } | |
719f8358 ED |
1526 | |
1527 | spin_unlock_bh(&hslot->lock); | |
1528 | } | |
1529 | } | |
1530 | } | |
1531 | EXPORT_SYMBOL(udp_lib_rehash); | |
1532 | ||
1533 | static void udp_v4_rehash(struct sock *sk) | |
1534 | { | |
1535 | u16 new_hash = udp4_portaddr_hash(sock_net(sk), | |
1536 | inet_sk(sk)->inet_rcv_saddr, | |
1537 | inet_sk(sk)->inet_num); | |
1538 | udp_lib_rehash(sk, new_hash); | |
1539 | } | |
1540 | ||
93821778 HX |
1541 | static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
1542 | { | |
fec5e652 | 1543 | int rc; |
766e9037 | 1544 | |
005ec974 | 1545 | if (inet_sk(sk)->inet_daddr) { |
bdeab991 | 1546 | sock_rps_save_rxhash(sk, skb); |
005ec974 | 1547 | sk_mark_napi_id(sk, skb); |
2c8c56e1 | 1548 | sk_incoming_cpu_update(sk); |
005ec974 | 1549 | } |
fec5e652 | 1550 | |
d826eb14 | 1551 | rc = sock_queue_rcv_skb(sk, skb); |
766e9037 ED |
1552 | if (rc < 0) { |
1553 | int is_udplite = IS_UDPLITE(sk); | |
93821778 | 1554 | |
93821778 | 1555 | /* Note that an ENOMEM error is charged twice */ |
766e9037 | 1556 | if (rc == -ENOMEM) |
93821778 HX |
1557 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1558 | is_udplite); | |
766e9037 ED |
1559 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
1560 | kfree_skb(skb); | |
296f7ea7 | 1561 | trace_udp_fail_queue_rcv_skb(rc, sk); |
766e9037 | 1562 | return -1; |
93821778 HX |
1563 | } |
1564 | ||
1565 | return 0; | |
1566 | ||
93821778 HX |
1567 | } |
1568 | ||
447167bf ED |
1569 | static struct static_key udp_encap_needed __read_mostly; |
1570 | void udp_encap_enable(void) | |
1571 | { | |
1572 | if (!static_key_enabled(&udp_encap_needed)) | |
1573 | static_key_slow_inc(&udp_encap_needed); | |
1574 | } | |
1575 | EXPORT_SYMBOL(udp_encap_enable); | |
1576 | ||
db8dac20 DM |
1577 | /* returns: |
1578 | * -1: error | |
1579 | * 0: success | |
1580 | * >0: "udp encap" protocol resubmission | |
1581 | * | |
1582 | * Note that in the success and error cases, the skb is assumed to | |
1583 | * have either been requeued or freed. | |
1584 | */ | |
c482c568 | 1585 | int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
db8dac20 DM |
1586 | { |
1587 | struct udp_sock *up = udp_sk(sk); | |
1588 | int rc; | |
1589 | int is_udplite = IS_UDPLITE(sk); | |
1590 | ||
1591 | /* | |
1592 | * Charge it to the socket, dropping if the queue is full. | |
1593 | */ | |
1594 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1595 | goto drop; | |
1596 | nf_reset(skb); | |
1597 | ||
447167bf | 1598 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
0ad92ad0 ED |
1599 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
1600 | ||
db8dac20 DM |
1601 | /* |
1602 | * This is an encapsulation socket so pass the skb to | |
1603 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1604 | * fall through and pass this up the UDP socket. | |
1605 | * up->encap_rcv() returns the following value: | |
1606 | * =0 if skb was successfully passed to the encap | |
1607 | * handler or was discarded by it. | |
1608 | * >0 if skb should be passed on to UDP. | |
1609 | * <0 if skb should be resubmitted as proto -N | |
1610 | */ | |
1611 | ||
1612 | /* if we're overly short, let UDP handle it */ | |
0ad92ad0 | 1613 | encap_rcv = ACCESS_ONCE(up->encap_rcv); |
00db4124 | 1614 | if (skb->len > sizeof(struct udphdr) && encap_rcv) { |
db8dac20 DM |
1615 | int ret; |
1616 | ||
0a80966b TH |
1617 | /* Verify checksum before giving to encap */ |
1618 | if (udp_lib_checksum_complete(skb)) | |
1619 | goto csum_error; | |
1620 | ||
0ad92ad0 | 1621 | ret = encap_rcv(sk, skb); |
db8dac20 | 1622 | if (ret <= 0) { |
0283328e PE |
1623 | UDP_INC_STATS_BH(sock_net(sk), |
1624 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
1625 | is_udplite); |
1626 | return -ret; | |
1627 | } | |
1628 | } | |
1629 | ||
1630 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1631 | } | |
1632 | ||
1633 | /* | |
1634 | * UDP-Lite specific tests, ignored on UDP sockets | |
1635 | */ | |
1636 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1637 | ||
1638 | /* | |
1639 | * MIB statistics other than incrementing the error count are | |
1640 | * disabled for the following two types of errors: these depend | |
1641 | * on the application settings, not on the functioning of the | |
1642 | * protocol stack as such. | |
1643 | * | |
1644 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1645 | * way ... to ... at least let the receiving application block | |
1646 | * delivery of packets with coverage values less than a value | |
1647 | * provided by the application." | |
1648 | */ | |
1649 | if (up->pcrlen == 0) { /* full coverage was set */ | |
ba7a46f1 JP |
1650 | net_dbg_ratelimited("UDPLite: partial coverage %d while full coverage %d requested\n", |
1651 | UDP_SKB_CB(skb)->cscov, skb->len); | |
db8dac20 DM |
1652 | goto drop; |
1653 | } | |
1654 | /* The next case involves violating the min. coverage requested | |
1655 | * by the receiver. This is subtle: if receiver wants x and x is | |
1656 | * greater than the buffersize/MTU then receiver will complain | |
1657 | * that it wants x while sender emits packets of smaller size y. | |
1658 | * Therefore the above ...()->partial_cov statement is essential. | |
1659 | */ | |
1660 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
ba7a46f1 JP |
1661 | net_dbg_ratelimited("UDPLite: coverage %d too small, need min %d\n", |
1662 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
db8dac20 DM |
1663 | goto drop; |
1664 | } | |
1665 | } | |
1666 | ||
33d480ce ED |
1667 | if (rcu_access_pointer(sk->sk_filter) && |
1668 | udp_lib_checksum_complete(skb)) | |
6a5dc9e5 | 1669 | goto csum_error; |
db8dac20 | 1670 | |
274f482d | 1671 | if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) { |
3e215c8d JL |
1672 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1673 | is_udplite); | |
c377411f | 1674 | goto drop; |
3e215c8d | 1675 | } |
c377411f | 1676 | |
93821778 | 1677 | rc = 0; |
db8dac20 | 1678 | |
fbf8866d | 1679 | ipv4_pktinfo_prepare(sk, skb); |
93821778 HX |
1680 | bh_lock_sock(sk); |
1681 | if (!sock_owned_by_user(sk)) | |
1682 | rc = __udp_queue_rcv_skb(sk, skb); | |
f545a38f | 1683 | else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) { |
55349790 ZY |
1684 | bh_unlock_sock(sk); |
1685 | goto drop; | |
1686 | } | |
93821778 HX |
1687 | bh_unlock_sock(sk); |
1688 | ||
1689 | return rc; | |
db8dac20 | 1690 | |
6a5dc9e5 ED |
1691 | csum_error: |
1692 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); | |
db8dac20 | 1693 | drop: |
0283328e | 1694 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8edf19c2 | 1695 | atomic_inc(&sk->sk_drops); |
db8dac20 DM |
1696 | kfree_skb(skb); |
1697 | return -1; | |
1698 | } | |
1699 | ||
1240d137 ED |
1700 | static void flush_stack(struct sock **stack, unsigned int count, |
1701 | struct sk_buff *skb, unsigned int final) | |
1702 | { | |
1703 | unsigned int i; | |
1704 | struct sk_buff *skb1 = NULL; | |
f6b8f32c | 1705 | struct sock *sk; |
1240d137 ED |
1706 | |
1707 | for (i = 0; i < count; i++) { | |
f6b8f32c | 1708 | sk = stack[i]; |
51456b29 | 1709 | if (likely(!skb1)) |
1240d137 ED |
1710 | skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC); |
1711 | ||
f6b8f32c ED |
1712 | if (!skb1) { |
1713 | atomic_inc(&sk->sk_drops); | |
1714 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, | |
1715 | IS_UDPLITE(sk)); | |
1716 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, | |
1717 | IS_UDPLITE(sk)); | |
1718 | } | |
1719 | ||
1720 | if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0) | |
1240d137 | 1721 | skb1 = NULL; |
2dc41cff DH |
1722 | |
1723 | sock_put(sk); | |
1240d137 ED |
1724 | } |
1725 | if (unlikely(skb1)) | |
1726 | kfree_skb(skb1); | |
1727 | } | |
1728 | ||
97502231 | 1729 | /* For TCP sockets, sk_rx_dst is protected by socket lock |
e47eb5df | 1730 | * For UDP, we use xchg() to guard against concurrent changes. |
97502231 ED |
1731 | */ |
1732 | static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst) | |
421b3885 | 1733 | { |
97502231 ED |
1734 | struct dst_entry *old; |
1735 | ||
e47eb5df ED |
1736 | dst_hold(dst); |
1737 | old = xchg(&sk->sk_rx_dst, dst); | |
1738 | dst_release(old); | |
421b3885 SB |
1739 | } |
1740 | ||
db8dac20 DM |
1741 | /* |
1742 | * Multicasts and broadcasts go to each listener. | |
1743 | * | |
1240d137 | 1744 | * Note: called only from the BH handler context. |
db8dac20 | 1745 | */ |
e3163493 | 1746 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1747 | struct udphdr *uh, |
1748 | __be32 saddr, __be32 daddr, | |
36cbb245 RJ |
1749 | struct udp_table *udptable, |
1750 | int proto) | |
db8dac20 | 1751 | { |
1240d137 | 1752 | struct sock *sk, *stack[256 / sizeof(struct sock *)]; |
5cf3d461 DH |
1753 | struct hlist_nulls_node *node; |
1754 | unsigned short hnum = ntohs(uh->dest); | |
1755 | struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum); | |
1756 | int dif = skb->dev->ifindex; | |
2dc41cff DH |
1757 | unsigned int count = 0, offset = offsetof(typeof(*sk), sk_nulls_node); |
1758 | unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10); | |
36cbb245 | 1759 | bool inner_flushed = false; |
2dc41cff DH |
1760 | |
1761 | if (use_hash2) { | |
1762 | hash2_any = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum) & | |
1763 | udp_table.mask; | |
1764 | hash2 = udp4_portaddr_hash(net, daddr, hnum) & udp_table.mask; | |
1765 | start_lookup: | |
1766 | hslot = &udp_table.hash2[hash2]; | |
1767 | offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node); | |
1768 | } | |
db8dac20 | 1769 | |
645ca708 | 1770 | spin_lock(&hslot->lock); |
2dc41cff | 1771 | sk_nulls_for_each_entry_offset(sk, node, &hslot->head, offset) { |
5cf3d461 DH |
1772 | if (__udp_is_mcast_sock(net, sk, |
1773 | uh->dest, daddr, | |
1774 | uh->source, saddr, | |
1775 | dif, hnum)) { | |
1776 | if (unlikely(count == ARRAY_SIZE(stack))) { | |
1777 | flush_stack(stack, count, skb, ~0); | |
36cbb245 | 1778 | inner_flushed = true; |
5cf3d461 DH |
1779 | count = 0; |
1780 | } | |
1781 | stack[count++] = sk; | |
2dc41cff | 1782 | sock_hold(sk); |
1240d137 ED |
1783 | } |
1784 | } | |
1240d137 | 1785 | |
645ca708 | 1786 | spin_unlock(&hslot->lock); |
1240d137 | 1787 | |
2dc41cff DH |
1788 | /* Also lookup *:port if we are using hash2 and haven't done so yet. */ |
1789 | if (use_hash2 && hash2 != hash2_any) { | |
1790 | hash2 = hash2_any; | |
1791 | goto start_lookup; | |
1792 | } | |
1793 | ||
1240d137 ED |
1794 | /* |
1795 | * do the slow work with no lock held | |
1796 | */ | |
1797 | if (count) { | |
1798 | flush_stack(stack, count, skb, count - 1); | |
1240d137 | 1799 | } else { |
36cbb245 RJ |
1800 | if (!inner_flushed) |
1801 | UDP_INC_STATS_BH(net, UDP_MIB_IGNOREDMULTI, | |
1802 | proto == IPPROTO_UDPLITE); | |
1803 | consume_skb(skb); | |
1240d137 | 1804 | } |
db8dac20 DM |
1805 | return 0; |
1806 | } | |
1807 | ||
1808 | /* Initialize UDP checksum. If exited with zero value (success), | |
1809 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1810 | * Otherwise, csum completion requires chacksumming packet body, | |
1811 | * including udp header and folding it to skb->csum. | |
1812 | */ | |
1813 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1814 | int proto) | |
1815 | { | |
db8dac20 DM |
1816 | int err; |
1817 | ||
1818 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1819 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1820 | ||
1821 | if (proto == IPPROTO_UDPLITE) { | |
1822 | err = udplite_checksum_init(skb, uh); | |
1823 | if (err) | |
1824 | return err; | |
1825 | } | |
1826 | ||
ed70fcfc TH |
1827 | return skb_checksum_init_zero_check(skb, proto, uh->check, |
1828 | inet_compute_pseudo); | |
db8dac20 DM |
1829 | } |
1830 | ||
1831 | /* | |
1832 | * All we need to do is get the socket, and then do a checksum. | |
1833 | */ | |
1834 | ||
645ca708 | 1835 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1836 | int proto) |
1837 | { | |
1838 | struct sock *sk; | |
7b5e56f9 | 1839 | struct udphdr *uh; |
db8dac20 | 1840 | unsigned short ulen; |
adf30907 | 1841 | struct rtable *rt = skb_rtable(skb); |
2783ef23 | 1842 | __be32 saddr, daddr; |
0283328e | 1843 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1844 | |
1845 | /* | |
1846 | * Validate the packet. | |
1847 | */ | |
1848 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1849 | goto drop; /* No space for header. */ | |
1850 | ||
7b5e56f9 | 1851 | uh = udp_hdr(skb); |
db8dac20 | 1852 | ulen = ntohs(uh->len); |
ccc2d97c BM |
1853 | saddr = ip_hdr(skb)->saddr; |
1854 | daddr = ip_hdr(skb)->daddr; | |
1855 | ||
db8dac20 DM |
1856 | if (ulen > skb->len) |
1857 | goto short_packet; | |
1858 | ||
1859 | if (proto == IPPROTO_UDP) { | |
1860 | /* UDP validates ulen. */ | |
1861 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1862 | goto short_packet; | |
1863 | uh = udp_hdr(skb); | |
1864 | } | |
1865 | ||
1866 | if (udp4_csum_init(skb, uh, proto)) | |
1867 | goto csum_error; | |
1868 | ||
8afdd99a ED |
1869 | sk = skb_steal_sock(skb); |
1870 | if (sk) { | |
97502231 | 1871 | struct dst_entry *dst = skb_dst(skb); |
421b3885 | 1872 | int ret; |
421b3885 | 1873 | |
97502231 ED |
1874 | if (unlikely(sk->sk_rx_dst != dst)) |
1875 | udp_sk_rx_dst_set(sk, dst); | |
db8dac20 | 1876 | |
421b3885 | 1877 | ret = udp_queue_rcv_skb(sk, skb); |
8afdd99a | 1878 | sock_put(sk); |
421b3885 SB |
1879 | /* a return value > 0 means to resubmit the input, but |
1880 | * it wants the return to be -protocol, or 0 | |
1881 | */ | |
1882 | if (ret > 0) | |
1883 | return -ret; | |
1884 | return 0; | |
421b3885 | 1885 | } |
db8dac20 | 1886 | |
c18450a5 FF |
1887 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) |
1888 | return __udp4_lib_mcast_deliver(net, skb, uh, | |
36cbb245 | 1889 | saddr, daddr, udptable, proto); |
c18450a5 FF |
1890 | |
1891 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); | |
00db4124 | 1892 | if (sk) { |
a5b50476 ET |
1893 | int ret; |
1894 | ||
224d019c | 1895 | if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk)) |
2abb7cdc TH |
1896 | skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check, |
1897 | inet_compute_pseudo); | |
1898 | ||
a5b50476 | 1899 | ret = udp_queue_rcv_skb(sk, skb); |
db8dac20 DM |
1900 | sock_put(sk); |
1901 | ||
1902 | /* a return value > 0 means to resubmit the input, but | |
1903 | * it wants the return to be -protocol, or 0 | |
1904 | */ | |
1905 | if (ret > 0) | |
1906 | return -ret; | |
1907 | return 0; | |
1908 | } | |
1909 | ||
1910 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1911 | goto drop; | |
1912 | nf_reset(skb); | |
1913 | ||
1914 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1915 | if (udp_lib_checksum_complete(skb)) | |
1916 | goto csum_error; | |
1917 | ||
0283328e | 1918 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1919 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1920 | ||
1921 | /* | |
1922 | * Hmm. We got an UDP packet to a port to which we | |
1923 | * don't wanna listen. Ignore it. | |
1924 | */ | |
1925 | kfree_skb(skb); | |
1926 | return 0; | |
1927 | ||
1928 | short_packet: | |
ba7a46f1 JP |
1929 | net_dbg_ratelimited("UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n", |
1930 | proto == IPPROTO_UDPLITE ? "Lite" : "", | |
1931 | &saddr, ntohs(uh->source), | |
1932 | ulen, skb->len, | |
1933 | &daddr, ntohs(uh->dest)); | |
db8dac20 DM |
1934 | goto drop; |
1935 | ||
1936 | csum_error: | |
1937 | /* | |
1938 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1939 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1940 | */ | |
ba7a46f1 JP |
1941 | net_dbg_ratelimited("UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n", |
1942 | proto == IPPROTO_UDPLITE ? "Lite" : "", | |
1943 | &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest), | |
1944 | ulen); | |
6a5dc9e5 | 1945 | UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 | 1946 | drop: |
0283328e | 1947 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1948 | kfree_skb(skb); |
1949 | return 0; | |
1950 | } | |
1951 | ||
421b3885 SB |
1952 | /* We can only early demux multicast if there is a single matching socket. |
1953 | * If more than one socket found returns NULL | |
1954 | */ | |
1955 | static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net, | |
1956 | __be16 loc_port, __be32 loc_addr, | |
1957 | __be16 rmt_port, __be32 rmt_addr, | |
1958 | int dif) | |
1959 | { | |
1960 | struct sock *sk, *result; | |
1961 | struct hlist_nulls_node *node; | |
1962 | unsigned short hnum = ntohs(loc_port); | |
1963 | unsigned int count, slot = udp_hashfn(net, hnum, udp_table.mask); | |
1964 | struct udp_hslot *hslot = &udp_table.hash[slot]; | |
1965 | ||
63c6f81c ED |
1966 | /* Do not bother scanning a too big list */ |
1967 | if (hslot->count > 10) | |
1968 | return NULL; | |
1969 | ||
421b3885 SB |
1970 | rcu_read_lock(); |
1971 | begin: | |
1972 | count = 0; | |
1973 | result = NULL; | |
1974 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { | |
1975 | if (__udp_is_mcast_sock(net, sk, | |
1976 | loc_port, loc_addr, | |
1977 | rmt_port, rmt_addr, | |
1978 | dif, hnum)) { | |
1979 | result = sk; | |
1980 | ++count; | |
1981 | } | |
1982 | } | |
1983 | /* | |
1984 | * if the nulls value we got at the end of this lookup is | |
1985 | * not the expected one, we must restart lookup. | |
1986 | * We probably met an item that was moved to another chain. | |
1987 | */ | |
1988 | if (get_nulls_value(node) != slot) | |
1989 | goto begin; | |
1990 | ||
1991 | if (result) { | |
1992 | if (count != 1 || | |
1993 | unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) | |
1994 | result = NULL; | |
f69b923a | 1995 | else if (unlikely(!__udp_is_mcast_sock(net, result, |
421b3885 SB |
1996 | loc_port, loc_addr, |
1997 | rmt_port, rmt_addr, | |
1998 | dif, hnum))) { | |
1999 | sock_put(result); | |
2000 | result = NULL; | |
2001 | } | |
2002 | } | |
2003 | rcu_read_unlock(); | |
2004 | return result; | |
2005 | } | |
2006 | ||
2007 | /* For unicast we should only early demux connected sockets or we can | |
2008 | * break forwarding setups. The chains here can be long so only check | |
2009 | * if the first socket is an exact match and if not move on. | |
2010 | */ | |
2011 | static struct sock *__udp4_lib_demux_lookup(struct net *net, | |
2012 | __be16 loc_port, __be32 loc_addr, | |
2013 | __be16 rmt_port, __be32 rmt_addr, | |
2014 | int dif) | |
2015 | { | |
2016 | struct sock *sk, *result; | |
2017 | struct hlist_nulls_node *node; | |
2018 | unsigned short hnum = ntohs(loc_port); | |
2019 | unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum); | |
2020 | unsigned int slot2 = hash2 & udp_table.mask; | |
2021 | struct udp_hslot *hslot2 = &udp_table.hash2[slot2]; | |
c7228317 | 2022 | INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr); |
421b3885 SB |
2023 | const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum); |
2024 | ||
2025 | rcu_read_lock(); | |
2026 | result = NULL; | |
2027 | udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { | |
2028 | if (INET_MATCH(sk, net, acookie, | |
2029 | rmt_addr, loc_addr, ports, dif)) | |
2030 | result = sk; | |
2031 | /* Only check first socket in chain */ | |
2032 | break; | |
2033 | } | |
2034 | ||
2035 | if (result) { | |
2036 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) | |
2037 | result = NULL; | |
2038 | else if (unlikely(!INET_MATCH(sk, net, acookie, | |
2039 | rmt_addr, loc_addr, | |
2040 | ports, dif))) { | |
2041 | sock_put(result); | |
2042 | result = NULL; | |
2043 | } | |
2044 | } | |
2045 | rcu_read_unlock(); | |
2046 | return result; | |
2047 | } | |
2048 | ||
2049 | void udp_v4_early_demux(struct sk_buff *skb) | |
2050 | { | |
610438b7 ED |
2051 | struct net *net = dev_net(skb->dev); |
2052 | const struct iphdr *iph; | |
2053 | const struct udphdr *uh; | |
421b3885 SB |
2054 | struct sock *sk; |
2055 | struct dst_entry *dst; | |
421b3885 | 2056 | int dif = skb->dev->ifindex; |
6e540309 | 2057 | int ours; |
421b3885 SB |
2058 | |
2059 | /* validate the packet */ | |
2060 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr))) | |
2061 | return; | |
2062 | ||
610438b7 ED |
2063 | iph = ip_hdr(skb); |
2064 | uh = udp_hdr(skb); | |
2065 | ||
421b3885 | 2066 | if (skb->pkt_type == PACKET_BROADCAST || |
6e540309 SB |
2067 | skb->pkt_type == PACKET_MULTICAST) { |
2068 | struct in_device *in_dev = __in_dev_get_rcu(skb->dev); | |
2069 | ||
2070 | if (!in_dev) | |
2071 | return; | |
2072 | ||
2073 | ours = ip_check_mc_rcu(in_dev, iph->daddr, iph->saddr, | |
2074 | iph->protocol); | |
2075 | if (!ours) | |
2076 | return; | |
421b3885 SB |
2077 | sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr, |
2078 | uh->source, iph->saddr, dif); | |
6e540309 | 2079 | } else if (skb->pkt_type == PACKET_HOST) { |
421b3885 SB |
2080 | sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr, |
2081 | uh->source, iph->saddr, dif); | |
6e540309 | 2082 | } else { |
421b3885 | 2083 | return; |
6e540309 | 2084 | } |
421b3885 SB |
2085 | |
2086 | if (!sk) | |
2087 | return; | |
2088 | ||
2089 | skb->sk = sk; | |
82eabd9e | 2090 | skb->destructor = sock_efree; |
10e2eb87 | 2091 | dst = READ_ONCE(sk->sk_rx_dst); |
421b3885 SB |
2092 | |
2093 | if (dst) | |
2094 | dst = dst_check(dst, 0); | |
10e2eb87 ED |
2095 | if (dst) { |
2096 | /* DST_NOCACHE can not be used without taking a reference */ | |
2097 | if (dst->flags & DST_NOCACHE) { | |
2098 | if (likely(atomic_inc_not_zero(&dst->__refcnt))) | |
2099 | skb_dst_set(skb, dst); | |
2100 | } else { | |
2101 | skb_dst_set_noref(skb, dst); | |
2102 | } | |
2103 | } | |
421b3885 SB |
2104 | } |
2105 | ||
db8dac20 DM |
2106 | int udp_rcv(struct sk_buff *skb) |
2107 | { | |
645ca708 | 2108 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
2109 | } |
2110 | ||
7d06b2e0 | 2111 | void udp_destroy_sock(struct sock *sk) |
db8dac20 | 2112 | { |
44046a59 | 2113 | struct udp_sock *up = udp_sk(sk); |
8a74ad60 | 2114 | bool slow = lock_sock_fast(sk); |
db8dac20 | 2115 | udp_flush_pending_frames(sk); |
8a74ad60 | 2116 | unlock_sock_fast(sk, slow); |
44046a59 TP |
2117 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
2118 | void (*encap_destroy)(struct sock *sk); | |
2119 | encap_destroy = ACCESS_ONCE(up->encap_destroy); | |
2120 | if (encap_destroy) | |
2121 | encap_destroy(sk); | |
2122 | } | |
db8dac20 DM |
2123 | } |
2124 | ||
1da177e4 LT |
2125 | /* |
2126 | * Socket option code for UDP | |
2127 | */ | |
4c0a6cb0 | 2128 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 2129 | char __user *optval, unsigned int optlen, |
4c0a6cb0 | 2130 | int (*push_pending_frames)(struct sock *)) |
1da177e4 LT |
2131 | { |
2132 | struct udp_sock *up = udp_sk(sk); | |
1c19448c | 2133 | int val, valbool; |
1da177e4 | 2134 | int err = 0; |
b2bf1e26 | 2135 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 2136 | |
c482c568 | 2137 | if (optlen < sizeof(int)) |
1da177e4 LT |
2138 | return -EINVAL; |
2139 | ||
2140 | if (get_user(val, (int __user *)optval)) | |
2141 | return -EFAULT; | |
2142 | ||
1c19448c TH |
2143 | valbool = val ? 1 : 0; |
2144 | ||
6516c655 | 2145 | switch (optname) { |
1da177e4 LT |
2146 | case UDP_CORK: |
2147 | if (val != 0) { | |
2148 | up->corkflag = 1; | |
2149 | } else { | |
2150 | up->corkflag = 0; | |
2151 | lock_sock(sk); | |
4243cdc2 | 2152 | push_pending_frames(sk); |
1da177e4 LT |
2153 | release_sock(sk); |
2154 | } | |
2155 | break; | |
e905a9ed | 2156 | |
1da177e4 LT |
2157 | case UDP_ENCAP: |
2158 | switch (val) { | |
2159 | case 0: | |
2160 | case UDP_ENCAP_ESPINUDP: | |
2161 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
2162 | up->encap_rcv = xfrm4_udp_encap_rcv; |
2163 | /* FALLTHROUGH */ | |
342f0234 | 2164 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 | 2165 | up->encap_type = val; |
447167bf | 2166 | udp_encap_enable(); |
1da177e4 LT |
2167 | break; |
2168 | default: | |
2169 | err = -ENOPROTOOPT; | |
2170 | break; | |
2171 | } | |
2172 | break; | |
2173 | ||
1c19448c TH |
2174 | case UDP_NO_CHECK6_TX: |
2175 | up->no_check6_tx = valbool; | |
2176 | break; | |
2177 | ||
2178 | case UDP_NO_CHECK6_RX: | |
2179 | up->no_check6_rx = valbool; | |
2180 | break; | |
2181 | ||
ba4e58ec GR |
2182 | /* |
2183 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
2184 | */ | |
2185 | /* The sender sets actual checksum coverage length via this option. | |
2186 | * The case coverage > packet length is handled by send module. */ | |
2187 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 2188 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
2189 | return -ENOPROTOOPT; |
2190 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
2191 | val = 8; | |
4be929be AD |
2192 | else if (val > USHRT_MAX) |
2193 | val = USHRT_MAX; | |
ba4e58ec GR |
2194 | up->pcslen = val; |
2195 | up->pcflag |= UDPLITE_SEND_CC; | |
2196 | break; | |
2197 | ||
e905a9ed YH |
2198 | /* The receiver specifies a minimum checksum coverage value. To make |
2199 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
2200 | * used, this again means full checksum coverage. */ |
2201 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 2202 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
2203 | return -ENOPROTOOPT; |
2204 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
2205 | val = 8; | |
4be929be AD |
2206 | else if (val > USHRT_MAX) |
2207 | val = USHRT_MAX; | |
ba4e58ec GR |
2208 | up->pcrlen = val; |
2209 | up->pcflag |= UDPLITE_RECV_CC; | |
2210 | break; | |
2211 | ||
1da177e4 LT |
2212 | default: |
2213 | err = -ENOPROTOOPT; | |
2214 | break; | |
6516c655 | 2215 | } |
1da177e4 LT |
2216 | |
2217 | return err; | |
2218 | } | |
c482c568 | 2219 | EXPORT_SYMBOL(udp_lib_setsockopt); |
1da177e4 | 2220 | |
db8dac20 | 2221 | int udp_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 2222 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
2223 | { |
2224 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2225 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
2226 | udp_push_pending_frames); | |
2227 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
2228 | } | |
2229 | ||
2230 | #ifdef CONFIG_COMPAT | |
2231 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 2232 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
2233 | { |
2234 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2235 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
2236 | udp_push_pending_frames); | |
2237 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
2238 | } | |
2239 | #endif | |
2240 | ||
4c0a6cb0 GR |
2241 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
2242 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
2243 | { |
2244 | struct udp_sock *up = udp_sk(sk); | |
2245 | int val, len; | |
2246 | ||
c482c568 | 2247 | if (get_user(len, optlen)) |
1da177e4 LT |
2248 | return -EFAULT; |
2249 | ||
2250 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 2251 | |
6516c655 | 2252 | if (len < 0) |
1da177e4 LT |
2253 | return -EINVAL; |
2254 | ||
6516c655 | 2255 | switch (optname) { |
1da177e4 LT |
2256 | case UDP_CORK: |
2257 | val = up->corkflag; | |
2258 | break; | |
2259 | ||
2260 | case UDP_ENCAP: | |
2261 | val = up->encap_type; | |
2262 | break; | |
2263 | ||
1c19448c TH |
2264 | case UDP_NO_CHECK6_TX: |
2265 | val = up->no_check6_tx; | |
2266 | break; | |
2267 | ||
2268 | case UDP_NO_CHECK6_RX: | |
2269 | val = up->no_check6_rx; | |
2270 | break; | |
2271 | ||
ba4e58ec GR |
2272 | /* The following two cannot be changed on UDP sockets, the return is |
2273 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
2274 | case UDPLITE_SEND_CSCOV: | |
2275 | val = up->pcslen; | |
2276 | break; | |
2277 | ||
2278 | case UDPLITE_RECV_CSCOV: | |
2279 | val = up->pcrlen; | |
2280 | break; | |
2281 | ||
1da177e4 LT |
2282 | default: |
2283 | return -ENOPROTOOPT; | |
6516c655 | 2284 | } |
1da177e4 | 2285 | |
6516c655 | 2286 | if (put_user(len, optlen)) |
e905a9ed | 2287 | return -EFAULT; |
c482c568 | 2288 | if (copy_to_user(optval, &val, len)) |
1da177e4 | 2289 | return -EFAULT; |
e905a9ed | 2290 | return 0; |
1da177e4 | 2291 | } |
c482c568 | 2292 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1da177e4 | 2293 | |
db8dac20 DM |
2294 | int udp_getsockopt(struct sock *sk, int level, int optname, |
2295 | char __user *optval, int __user *optlen) | |
2296 | { | |
2297 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2298 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
2299 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
2300 | } | |
2301 | ||
2302 | #ifdef CONFIG_COMPAT | |
2303 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
2304 | char __user *optval, int __user *optlen) | |
2305 | { | |
2306 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2307 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
2308 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
2309 | } | |
2310 | #endif | |
1da177e4 LT |
2311 | /** |
2312 | * udp_poll - wait for a UDP event. | |
2313 | * @file - file struct | |
2314 | * @sock - socket | |
2315 | * @wait - poll table | |
2316 | * | |
e905a9ed | 2317 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
2318 | * blocking sockets. If application is using a blocking fd |
2319 | * and a packet with checksum error is in the queue; | |
2320 | * then it could get return from select indicating data available | |
2321 | * but then block when reading it. Add special case code | |
2322 | * to work around these arguably broken applications. | |
2323 | */ | |
2324 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
2325 | { | |
2326 | unsigned int mask = datagram_poll(file, sock, wait); | |
2327 | struct sock *sk = sock->sk; | |
ba4e58ec | 2328 | |
c3f1dbaf DM |
2329 | sock_rps_record_flow(sk); |
2330 | ||
1da177e4 | 2331 | /* Check for false positives due to checksum errors */ |
85584672 ED |
2332 | if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) && |
2333 | !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk)) | |
2334 | mask &= ~(POLLIN | POLLRDNORM); | |
1da177e4 LT |
2335 | |
2336 | return mask; | |
e905a9ed | 2337 | |
1da177e4 | 2338 | } |
c482c568 | 2339 | EXPORT_SYMBOL(udp_poll); |
1da177e4 | 2340 | |
db8dac20 DM |
2341 | struct proto udp_prot = { |
2342 | .name = "UDP", | |
2343 | .owner = THIS_MODULE, | |
2344 | .close = udp_lib_close, | |
2345 | .connect = ip4_datagram_connect, | |
2346 | .disconnect = udp_disconnect, | |
2347 | .ioctl = udp_ioctl, | |
2348 | .destroy = udp_destroy_sock, | |
2349 | .setsockopt = udp_setsockopt, | |
2350 | .getsockopt = udp_getsockopt, | |
2351 | .sendmsg = udp_sendmsg, | |
2352 | .recvmsg = udp_recvmsg, | |
2353 | .sendpage = udp_sendpage, | |
93821778 | 2354 | .backlog_rcv = __udp_queue_rcv_skb, |
8141ed9f | 2355 | .release_cb = ip4_datagram_release_cb, |
db8dac20 DM |
2356 | .hash = udp_lib_hash, |
2357 | .unhash = udp_lib_unhash, | |
719f8358 | 2358 | .rehash = udp_v4_rehash, |
db8dac20 DM |
2359 | .get_port = udp_v4_get_port, |
2360 | .memory_allocated = &udp_memory_allocated, | |
2361 | .sysctl_mem = sysctl_udp_mem, | |
2362 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
2363 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
2364 | .obj_size = sizeof(struct udp_sock), | |
271b72c7 | 2365 | .slab_flags = SLAB_DESTROY_BY_RCU, |
645ca708 | 2366 | .h.udp_table = &udp_table, |
db8dac20 DM |
2367 | #ifdef CONFIG_COMPAT |
2368 | .compat_setsockopt = compat_udp_setsockopt, | |
2369 | .compat_getsockopt = compat_udp_getsockopt, | |
2370 | #endif | |
fcbdf09d | 2371 | .clear_sk = sk_prot_clear_portaddr_nulls, |
db8dac20 | 2372 | }; |
c482c568 | 2373 | EXPORT_SYMBOL(udp_prot); |
1da177e4 LT |
2374 | |
2375 | /* ------------------------------------------------------------------------ */ | |
2376 | #ifdef CONFIG_PROC_FS | |
2377 | ||
645ca708 | 2378 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
2379 | { |
2380 | struct sock *sk; | |
2381 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2382 | struct net *net = seq_file_net(seq); |
1da177e4 | 2383 | |
f86dcc5a ED |
2384 | for (state->bucket = start; state->bucket <= state->udp_table->mask; |
2385 | ++state->bucket) { | |
88ab1932 | 2386 | struct hlist_nulls_node *node; |
645ca708 | 2387 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
f86dcc5a ED |
2388 | |
2389 | if (hlist_nulls_empty(&hslot->head)) | |
2390 | continue; | |
2391 | ||
645ca708 | 2392 | spin_lock_bh(&hslot->lock); |
88ab1932 | 2393 | sk_nulls_for_each(sk, node, &hslot->head) { |
878628fb | 2394 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 2395 | continue; |
1da177e4 LT |
2396 | if (sk->sk_family == state->family) |
2397 | goto found; | |
2398 | } | |
645ca708 | 2399 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
2400 | } |
2401 | sk = NULL; | |
2402 | found: | |
2403 | return sk; | |
2404 | } | |
2405 | ||
2406 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
2407 | { | |
2408 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2409 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2410 | |
2411 | do { | |
88ab1932 | 2412 | sk = sk_nulls_next(sk); |
878628fb | 2413 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 2414 | |
645ca708 | 2415 | if (!sk) { |
f86dcc5a | 2416 | if (state->bucket <= state->udp_table->mask) |
30842f29 | 2417 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
645ca708 | 2418 | return udp_get_first(seq, state->bucket + 1); |
1da177e4 LT |
2419 | } |
2420 | return sk; | |
2421 | } | |
2422 | ||
2423 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
2424 | { | |
645ca708 | 2425 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
2426 | |
2427 | if (sk) | |
6516c655 | 2428 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
2429 | --pos; |
2430 | return pos ? NULL : sk; | |
2431 | } | |
2432 | ||
2433 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
2434 | { | |
30842f29 | 2435 | struct udp_iter_state *state = seq->private; |
f86dcc5a | 2436 | state->bucket = MAX_UDP_PORTS; |
30842f29 | 2437 | |
b50660f1 | 2438 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
2439 | } |
2440 | ||
2441 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2442 | { | |
2443 | struct sock *sk; | |
2444 | ||
b50660f1 | 2445 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2446 | sk = udp_get_idx(seq, 0); |
2447 | else | |
2448 | sk = udp_get_next(seq, v); | |
2449 | ||
2450 | ++*pos; | |
2451 | return sk; | |
2452 | } | |
2453 | ||
2454 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
2455 | { | |
645ca708 ED |
2456 | struct udp_iter_state *state = seq->private; |
2457 | ||
f86dcc5a | 2458 | if (state->bucket <= state->udp_table->mask) |
645ca708 | 2459 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
1da177e4 LT |
2460 | } |
2461 | ||
73cb88ec | 2462 | int udp_seq_open(struct inode *inode, struct file *file) |
1da177e4 | 2463 | { |
d9dda78b | 2464 | struct udp_seq_afinfo *afinfo = PDE_DATA(inode); |
a2be75c1 DL |
2465 | struct udp_iter_state *s; |
2466 | int err; | |
a91275ef | 2467 | |
a2be75c1 DL |
2468 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2469 | sizeof(struct udp_iter_state)); | |
2470 | if (err < 0) | |
2471 | return err; | |
a91275ef | 2472 | |
a2be75c1 | 2473 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2474 | s->family = afinfo->family; |
645ca708 | 2475 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 2476 | return err; |
a91275ef | 2477 | } |
73cb88ec | 2478 | EXPORT_SYMBOL(udp_seq_open); |
a91275ef | 2479 | |
1da177e4 | 2480 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 2481 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
2482 | { |
2483 | struct proc_dir_entry *p; | |
2484 | int rc = 0; | |
2485 | ||
dda61925 DL |
2486 | afinfo->seq_ops.start = udp_seq_start; |
2487 | afinfo->seq_ops.next = udp_seq_next; | |
2488 | afinfo->seq_ops.stop = udp_seq_stop; | |
2489 | ||
84841c3c | 2490 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2491 | afinfo->seq_fops, afinfo); |
84841c3c | 2492 | if (!p) |
1da177e4 LT |
2493 | rc = -ENOMEM; |
2494 | return rc; | |
2495 | } | |
c482c568 | 2496 | EXPORT_SYMBOL(udp_proc_register); |
1da177e4 | 2497 | |
0c96d8c5 | 2498 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 2499 | { |
ece31ffd | 2500 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2501 | } |
c482c568 | 2502 | EXPORT_SYMBOL(udp_proc_unregister); |
db8dac20 DM |
2503 | |
2504 | /* ------------------------------------------------------------------------ */ | |
5e659e4c | 2505 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
652586df | 2506 | int bucket) |
db8dac20 DM |
2507 | { |
2508 | struct inet_sock *inet = inet_sk(sp); | |
c720c7e8 ED |
2509 | __be32 dest = inet->inet_daddr; |
2510 | __be32 src = inet->inet_rcv_saddr; | |
2511 | __u16 destp = ntohs(inet->inet_dport); | |
2512 | __u16 srcp = ntohs(inet->inet_sport); | |
db8dac20 | 2513 | |
f86dcc5a | 2514 | seq_printf(f, "%5d: %08X:%04X %08X:%04X" |
652586df | 2515 | " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d", |
db8dac20 | 2516 | bucket, src, srcp, dest, destp, sp->sk_state, |
31e6d363 ED |
2517 | sk_wmem_alloc_get(sp), |
2518 | sk_rmem_alloc_get(sp), | |
a7cb5a49 EB |
2519 | 0, 0L, 0, |
2520 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)), | |
2521 | 0, sock_i_ino(sp), | |
cb61cb9b | 2522 | atomic_read(&sp->sk_refcnt), sp, |
652586df | 2523 | atomic_read(&sp->sk_drops)); |
db8dac20 DM |
2524 | } |
2525 | ||
2526 | int udp4_seq_show(struct seq_file *seq, void *v) | |
2527 | { | |
652586df | 2528 | seq_setwidth(seq, 127); |
db8dac20 | 2529 | if (v == SEQ_START_TOKEN) |
652586df | 2530 | seq_puts(seq, " sl local_address rem_address st tx_queue " |
db8dac20 | 2531 | "rx_queue tr tm->when retrnsmt uid timeout " |
cb61cb9b | 2532 | "inode ref pointer drops"); |
db8dac20 | 2533 | else { |
db8dac20 DM |
2534 | struct udp_iter_state *state = seq->private; |
2535 | ||
652586df | 2536 | udp4_format_sock(v, seq, state->bucket); |
db8dac20 | 2537 | } |
652586df | 2538 | seq_pad(seq, '\n'); |
db8dac20 DM |
2539 | return 0; |
2540 | } | |
2541 | ||
73cb88ec AV |
2542 | static const struct file_operations udp_afinfo_seq_fops = { |
2543 | .owner = THIS_MODULE, | |
2544 | .open = udp_seq_open, | |
2545 | .read = seq_read, | |
2546 | .llseek = seq_lseek, | |
2547 | .release = seq_release_net | |
2548 | }; | |
2549 | ||
db8dac20 | 2550 | /* ------------------------------------------------------------------------ */ |
db8dac20 | 2551 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
2552 | .name = "udp", |
2553 | .family = AF_INET, | |
645ca708 | 2554 | .udp_table = &udp_table, |
73cb88ec | 2555 | .seq_fops = &udp_afinfo_seq_fops, |
dda61925 DL |
2556 | .seq_ops = { |
2557 | .show = udp4_seq_show, | |
2558 | }, | |
db8dac20 DM |
2559 | }; |
2560 | ||
2c8c1e72 | 2561 | static int __net_init udp4_proc_init_net(struct net *net) |
15439feb PE |
2562 | { |
2563 | return udp_proc_register(net, &udp4_seq_afinfo); | |
2564 | } | |
2565 | ||
2c8c1e72 | 2566 | static void __net_exit udp4_proc_exit_net(struct net *net) |
15439feb PE |
2567 | { |
2568 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
2569 | } | |
2570 | ||
2571 | static struct pernet_operations udp4_net_ops = { | |
2572 | .init = udp4_proc_init_net, | |
2573 | .exit = udp4_proc_exit_net, | |
2574 | }; | |
2575 | ||
db8dac20 DM |
2576 | int __init udp4_proc_init(void) |
2577 | { | |
15439feb | 2578 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
2579 | } |
2580 | ||
2581 | void udp4_proc_exit(void) | |
2582 | { | |
15439feb | 2583 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 2584 | } |
1da177e4 LT |
2585 | #endif /* CONFIG_PROC_FS */ |
2586 | ||
f86dcc5a ED |
2587 | static __initdata unsigned long uhash_entries; |
2588 | static int __init set_uhash_entries(char *str) | |
645ca708 | 2589 | { |
413c27d8 EZ |
2590 | ssize_t ret; |
2591 | ||
f86dcc5a ED |
2592 | if (!str) |
2593 | return 0; | |
413c27d8 EZ |
2594 | |
2595 | ret = kstrtoul(str, 0, &uhash_entries); | |
2596 | if (ret) | |
2597 | return 0; | |
2598 | ||
f86dcc5a ED |
2599 | if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN) |
2600 | uhash_entries = UDP_HTABLE_SIZE_MIN; | |
2601 | return 1; | |
2602 | } | |
2603 | __setup("uhash_entries=", set_uhash_entries); | |
645ca708 | 2604 | |
f86dcc5a ED |
2605 | void __init udp_table_init(struct udp_table *table, const char *name) |
2606 | { | |
2607 | unsigned int i; | |
2608 | ||
31fe62b9 TB |
2609 | table->hash = alloc_large_system_hash(name, |
2610 | 2 * sizeof(struct udp_hslot), | |
2611 | uhash_entries, | |
2612 | 21, /* one slot per 2 MB */ | |
2613 | 0, | |
2614 | &table->log, | |
2615 | &table->mask, | |
2616 | UDP_HTABLE_SIZE_MIN, | |
2617 | 64 * 1024); | |
2618 | ||
512615b6 | 2619 | table->hash2 = table->hash + (table->mask + 1); |
f86dcc5a | 2620 | for (i = 0; i <= table->mask; i++) { |
88ab1932 | 2621 | INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i); |
fdcc8aa9 | 2622 | table->hash[i].count = 0; |
645ca708 ED |
2623 | spin_lock_init(&table->hash[i].lock); |
2624 | } | |
512615b6 ED |
2625 | for (i = 0; i <= table->mask; i++) { |
2626 | INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i); | |
2627 | table->hash2[i].count = 0; | |
2628 | spin_lock_init(&table->hash2[i].lock); | |
2629 | } | |
645ca708 ED |
2630 | } |
2631 | ||
723b8e46 TH |
2632 | u32 udp_flow_hashrnd(void) |
2633 | { | |
2634 | static u32 hashrnd __read_mostly; | |
2635 | ||
2636 | net_get_random_once(&hashrnd, sizeof(hashrnd)); | |
2637 | ||
2638 | return hashrnd; | |
2639 | } | |
2640 | EXPORT_SYMBOL(udp_flow_hashrnd); | |
2641 | ||
95766fff HA |
2642 | void __init udp_init(void) |
2643 | { | |
f03d78db | 2644 | unsigned long limit; |
95766fff | 2645 | |
f86dcc5a | 2646 | udp_table_init(&udp_table, "UDP"); |
f03d78db | 2647 | limit = nr_free_buffer_pages() / 8; |
95766fff HA |
2648 | limit = max(limit, 128UL); |
2649 | sysctl_udp_mem[0] = limit / 4 * 3; | |
2650 | sysctl_udp_mem[1] = limit; | |
2651 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
2652 | ||
2653 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
2654 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
2655 | } |