]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * The User Datagram Protocol (UDP). | |
7 | * | |
02c30a84 | 8 | * Authors: Ross Biro |
1da177e4 LT |
9 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
10 | * Arnt Gulbrandsen, <agulbra@nvg.unit.no> | |
113aa838 | 11 | * Alan Cox, <alan@lxorguk.ukuu.org.uk> |
1da177e4 LT |
12 | * Hirokazu Takahashi, <taka@valinux.co.jp> |
13 | * | |
14 | * Fixes: | |
15 | * Alan Cox : verify_area() calls | |
16 | * Alan Cox : stopped close while in use off icmp | |
17 | * messages. Not a fix but a botch that | |
18 | * for udp at least is 'valid'. | |
19 | * Alan Cox : Fixed icmp handling properly | |
20 | * Alan Cox : Correct error for oversized datagrams | |
e905a9ed YH |
21 | * Alan Cox : Tidied select() semantics. |
22 | * Alan Cox : udp_err() fixed properly, also now | |
1da177e4 LT |
23 | * select and read wake correctly on errors |
24 | * Alan Cox : udp_send verify_area moved to avoid mem leak | |
25 | * Alan Cox : UDP can count its memory | |
26 | * Alan Cox : send to an unknown connection causes | |
27 | * an ECONNREFUSED off the icmp, but | |
28 | * does NOT close. | |
29 | * Alan Cox : Switched to new sk_buff handlers. No more backlog! | |
30 | * Alan Cox : Using generic datagram code. Even smaller and the PEEK | |
31 | * bug no longer crashes it. | |
32 | * Fred Van Kempen : Net2e support for sk->broadcast. | |
33 | * Alan Cox : Uses skb_free_datagram | |
34 | * Alan Cox : Added get/set sockopt support. | |
35 | * Alan Cox : Broadcasting without option set returns EACCES. | |
36 | * Alan Cox : No wakeup calls. Instead we now use the callbacks. | |
37 | * Alan Cox : Use ip_tos and ip_ttl | |
38 | * Alan Cox : SNMP Mibs | |
39 | * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support. | |
40 | * Matt Dillon : UDP length checks. | |
41 | * Alan Cox : Smarter af_inet used properly. | |
42 | * Alan Cox : Use new kernel side addressing. | |
43 | * Alan Cox : Incorrect return on truncated datagram receive. | |
44 | * Arnt Gulbrandsen : New udp_send and stuff | |
45 | * Alan Cox : Cache last socket | |
46 | * Alan Cox : Route cache | |
47 | * Jon Peatfield : Minor efficiency fix to sendto(). | |
48 | * Mike Shaver : RFC1122 checks. | |
49 | * Alan Cox : Nonblocking error fix. | |
50 | * Willy Konynenberg : Transparent proxying support. | |
51 | * Mike McLagan : Routing by source | |
52 | * David S. Miller : New socket lookup architecture. | |
53 | * Last socket cache retained as it | |
54 | * does have a high hit rate. | |
55 | * Olaf Kirch : Don't linearise iovec on sendmsg. | |
56 | * Andi Kleen : Some cleanups, cache destination entry | |
e905a9ed | 57 | * for connect. |
1da177e4 LT |
58 | * Vitaly E. Lavrov : Transparent proxy revived after year coma. |
59 | * Melvin Smith : Check msg_name not msg_namelen in sendto(), | |
60 | * return ENOTCONN for unconnected sockets (POSIX) | |
61 | * Janos Farkas : don't deliver multi/broadcasts to a different | |
62 | * bound-to-device socket | |
63 | * Hirokazu Takahashi : HW checksumming for outgoing UDP | |
64 | * datagrams. | |
65 | * Hirokazu Takahashi : sendfile() on UDP works now. | |
66 | * Arnaldo C. Melo : convert /proc/net/udp to seq_file | |
67 | * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which | |
68 | * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind | |
69 | * a single port at the same time. | |
70 | * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support | |
342f0234 | 71 | * James Chapman : Add L2TP encapsulation type. |
1da177e4 LT |
72 | * |
73 | * | |
74 | * This program is free software; you can redistribute it and/or | |
75 | * modify it under the terms of the GNU General Public License | |
76 | * as published by the Free Software Foundation; either version | |
77 | * 2 of the License, or (at your option) any later version. | |
78 | */ | |
e905a9ed | 79 | |
afd46503 JP |
80 | #define pr_fmt(fmt) "UDP: " fmt |
81 | ||
1da177e4 LT |
82 | #include <asm/uaccess.h> |
83 | #include <asm/ioctls.h> | |
95766fff | 84 | #include <linux/bootmem.h> |
8203efb3 ED |
85 | #include <linux/highmem.h> |
86 | #include <linux/swap.h> | |
1da177e4 LT |
87 | #include <linux/types.h> |
88 | #include <linux/fcntl.h> | |
89 | #include <linux/module.h> | |
90 | #include <linux/socket.h> | |
91 | #include <linux/sockios.h> | |
14c85021 | 92 | #include <linux/igmp.h> |
1da177e4 LT |
93 | #include <linux/in.h> |
94 | #include <linux/errno.h> | |
95 | #include <linux/timer.h> | |
96 | #include <linux/mm.h> | |
1da177e4 | 97 | #include <linux/inet.h> |
1da177e4 | 98 | #include <linux/netdevice.h> |
5a0e3ad6 | 99 | #include <linux/slab.h> |
c752f073 | 100 | #include <net/tcp_states.h> |
1da177e4 LT |
101 | #include <linux/skbuff.h> |
102 | #include <linux/proc_fs.h> | |
103 | #include <linux/seq_file.h> | |
457c4cbc | 104 | #include <net/net_namespace.h> |
1da177e4 | 105 | #include <net/icmp.h> |
421b3885 | 106 | #include <net/inet_hashtables.h> |
1da177e4 | 107 | #include <net/route.h> |
1da177e4 LT |
108 | #include <net/checksum.h> |
109 | #include <net/xfrm.h> | |
296f7ea7 | 110 | #include <trace/events/udp.h> |
447167bf | 111 | #include <linux/static_key.h> |
22911fc5 | 112 | #include <trace/events/skb.h> |
076bb0c8 | 113 | #include <net/busy_poll.h> |
ba4e58ec | 114 | #include "udp_impl.h" |
1da177e4 | 115 | |
f86dcc5a | 116 | struct udp_table udp_table __read_mostly; |
645ca708 | 117 | EXPORT_SYMBOL(udp_table); |
1da177e4 | 118 | |
8d987e5c | 119 | long sysctl_udp_mem[3] __read_mostly; |
95766fff | 120 | EXPORT_SYMBOL(sysctl_udp_mem); |
c482c568 ED |
121 | |
122 | int sysctl_udp_rmem_min __read_mostly; | |
95766fff | 123 | EXPORT_SYMBOL(sysctl_udp_rmem_min); |
c482c568 ED |
124 | |
125 | int sysctl_udp_wmem_min __read_mostly; | |
95766fff HA |
126 | EXPORT_SYMBOL(sysctl_udp_wmem_min); |
127 | ||
8d987e5c | 128 | atomic_long_t udp_memory_allocated; |
95766fff HA |
129 | EXPORT_SYMBOL(udp_memory_allocated); |
130 | ||
f86dcc5a ED |
131 | #define MAX_UDP_PORTS 65536 |
132 | #define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN) | |
98322f22 | 133 | |
f24d43c0 | 134 | static int udp_lib_lport_inuse(struct net *net, __u16 num, |
645ca708 | 135 | const struct udp_hslot *hslot, |
98322f22 | 136 | unsigned long *bitmap, |
f24d43c0 ED |
137 | struct sock *sk, |
138 | int (*saddr_comp)(const struct sock *sk1, | |
f86dcc5a ED |
139 | const struct sock *sk2), |
140 | unsigned int log) | |
1da177e4 | 141 | { |
f24d43c0 | 142 | struct sock *sk2; |
88ab1932 | 143 | struct hlist_nulls_node *node; |
ba418fa3 | 144 | kuid_t uid = sock_i_uid(sk); |
25030a7f | 145 | |
88ab1932 | 146 | sk_nulls_for_each(sk2, node, &hslot->head) |
9d4fb27d JP |
147 | if (net_eq(sock_net(sk2), net) && |
148 | sk2 != sk && | |
d4cada4a | 149 | (bitmap || udp_sk(sk2)->udp_port_hash == num) && |
9d4fb27d JP |
150 | (!sk2->sk_reuse || !sk->sk_reuse) && |
151 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
152 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 TH |
153 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
154 | !uid_eq(uid, sock_i_uid(sk2))) && | |
98322f22 ED |
155 | (*saddr_comp)(sk, sk2)) { |
156 | if (bitmap) | |
d4cada4a ED |
157 | __set_bit(udp_sk(sk2)->udp_port_hash >> log, |
158 | bitmap); | |
98322f22 ED |
159 | else |
160 | return 1; | |
161 | } | |
25030a7f GR |
162 | return 0; |
163 | } | |
164 | ||
30fff923 ED |
165 | /* |
166 | * Note: we still hold spinlock of primary hash chain, so no other writer | |
167 | * can insert/delete a socket with local_port == num | |
168 | */ | |
169 | static int udp_lib_lport_inuse2(struct net *net, __u16 num, | |
170 | struct udp_hslot *hslot2, | |
171 | struct sock *sk, | |
172 | int (*saddr_comp)(const struct sock *sk1, | |
173 | const struct sock *sk2)) | |
174 | { | |
175 | struct sock *sk2; | |
176 | struct hlist_nulls_node *node; | |
ba418fa3 | 177 | kuid_t uid = sock_i_uid(sk); |
30fff923 ED |
178 | int res = 0; |
179 | ||
180 | spin_lock(&hslot2->lock); | |
181 | udp_portaddr_for_each_entry(sk2, node, &hslot2->head) | |
9d4fb27d JP |
182 | if (net_eq(sock_net(sk2), net) && |
183 | sk2 != sk && | |
184 | (udp_sk(sk2)->udp_port_hash == num) && | |
185 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
186 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
187 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 TH |
188 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
189 | !uid_eq(uid, sock_i_uid(sk2))) && | |
30fff923 ED |
190 | (*saddr_comp)(sk, sk2)) { |
191 | res = 1; | |
192 | break; | |
193 | } | |
194 | spin_unlock(&hslot2->lock); | |
195 | return res; | |
196 | } | |
197 | ||
25030a7f | 198 | /** |
6ba5a3c5 | 199 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
200 | * |
201 | * @sk: socket struct in question | |
202 | * @snum: port number to look up | |
df2bc459 | 203 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25985edc | 204 | * @hash2_nulladdr: AF-dependent hash value in secondary hash chains, |
30fff923 | 205 | * with NULL address |
25030a7f | 206 | */ |
6ba5a3c5 | 207 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
df2bc459 | 208 | int (*saddr_comp)(const struct sock *sk1, |
30fff923 ED |
209 | const struct sock *sk2), |
210 | unsigned int hash2_nulladdr) | |
25030a7f | 211 | { |
512615b6 | 212 | struct udp_hslot *hslot, *hslot2; |
645ca708 | 213 | struct udp_table *udptable = sk->sk_prot->h.udp_table; |
25030a7f | 214 | int error = 1; |
3b1e0a65 | 215 | struct net *net = sock_net(sk); |
1da177e4 | 216 | |
32c1da70 | 217 | if (!snum) { |
9088c560 | 218 | int low, high, remaining; |
95c96174 | 219 | unsigned int rand; |
98322f22 ED |
220 | unsigned short first, last; |
221 | DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN); | |
32c1da70 | 222 | |
0bbf87d8 | 223 | inet_get_local_port_range(net, &low, &high); |
a25de534 | 224 | remaining = (high - low) + 1; |
227b60f5 | 225 | |
9088c560 | 226 | rand = net_random(); |
98322f22 ED |
227 | first = (((u64)rand * remaining) >> 32) + low; |
228 | /* | |
229 | * force rand to be an odd multiple of UDP_HTABLE_SIZE | |
230 | */ | |
f86dcc5a | 231 | rand = (rand | 1) * (udptable->mask + 1); |
5781b235 ED |
232 | last = first + udptable->mask + 1; |
233 | do { | |
f86dcc5a | 234 | hslot = udp_hashslot(udptable, net, first); |
98322f22 | 235 | bitmap_zero(bitmap, PORTS_PER_CHAIN); |
645ca708 | 236 | spin_lock_bh(&hslot->lock); |
98322f22 | 237 | udp_lib_lport_inuse(net, snum, hslot, bitmap, sk, |
f86dcc5a | 238 | saddr_comp, udptable->log); |
98322f22 ED |
239 | |
240 | snum = first; | |
241 | /* | |
242 | * Iterate on all possible values of snum for this hash. | |
243 | * Using steps of an odd multiple of UDP_HTABLE_SIZE | |
244 | * give us randomization and full range coverage. | |
245 | */ | |
9088c560 | 246 | do { |
98322f22 | 247 | if (low <= snum && snum <= high && |
e3826f1e AW |
248 | !test_bit(snum >> udptable->log, bitmap) && |
249 | !inet_is_reserved_local_port(snum)) | |
98322f22 ED |
250 | goto found; |
251 | snum += rand; | |
252 | } while (snum != first); | |
253 | spin_unlock_bh(&hslot->lock); | |
5781b235 | 254 | } while (++first != last); |
98322f22 | 255 | goto fail; |
645ca708 | 256 | } else { |
f86dcc5a | 257 | hslot = udp_hashslot(udptable, net, snum); |
645ca708 | 258 | spin_lock_bh(&hslot->lock); |
30fff923 ED |
259 | if (hslot->count > 10) { |
260 | int exist; | |
261 | unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum; | |
262 | ||
263 | slot2 &= udptable->mask; | |
264 | hash2_nulladdr &= udptable->mask; | |
265 | ||
266 | hslot2 = udp_hashslot2(udptable, slot2); | |
267 | if (hslot->count < hslot2->count) | |
268 | goto scan_primary_hash; | |
269 | ||
270 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
271 | sk, saddr_comp); | |
272 | if (!exist && (hash2_nulladdr != slot2)) { | |
273 | hslot2 = udp_hashslot2(udptable, hash2_nulladdr); | |
274 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
275 | sk, saddr_comp); | |
276 | } | |
277 | if (exist) | |
278 | goto fail_unlock; | |
279 | else | |
280 | goto found; | |
281 | } | |
282 | scan_primary_hash: | |
f86dcc5a ED |
283 | if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk, |
284 | saddr_comp, 0)) | |
645ca708 ED |
285 | goto fail_unlock; |
286 | } | |
98322f22 | 287 | found: |
c720c7e8 | 288 | inet_sk(sk)->inet_num = snum; |
d4cada4a ED |
289 | udp_sk(sk)->udp_port_hash = snum; |
290 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
1da177e4 | 291 | if (sk_unhashed(sk)) { |
88ab1932 | 292 | sk_nulls_add_node_rcu(sk, &hslot->head); |
fdcc8aa9 | 293 | hslot->count++; |
c29a0bc4 | 294 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
512615b6 ED |
295 | |
296 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
297 | spin_lock(&hslot2->lock); | |
298 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
299 | &hslot2->head); | |
300 | hslot2->count++; | |
301 | spin_unlock(&hslot2->lock); | |
1da177e4 | 302 | } |
25030a7f | 303 | error = 0; |
645ca708 ED |
304 | fail_unlock: |
305 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 306 | fail: |
25030a7f GR |
307 | return error; |
308 | } | |
c482c568 | 309 | EXPORT_SYMBOL(udp_lib_get_port); |
25030a7f | 310 | |
499923c7 | 311 | static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2) |
db8dac20 DM |
312 | { |
313 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
314 | ||
c482c568 | 315 | return (!ipv6_only_sock(sk2) && |
c720c7e8 ED |
316 | (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr || |
317 | inet1->inet_rcv_saddr == inet2->inet_rcv_saddr)); | |
db8dac20 DM |
318 | } |
319 | ||
d4cada4a ED |
320 | static unsigned int udp4_portaddr_hash(struct net *net, __be32 saddr, |
321 | unsigned int port) | |
322 | { | |
0eae88f3 | 323 | return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port; |
d4cada4a ED |
324 | } |
325 | ||
6ba5a3c5 | 326 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 327 | { |
30fff923 | 328 | unsigned int hash2_nulladdr = |
0eae88f3 | 329 | udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum); |
30fff923 ED |
330 | unsigned int hash2_partial = |
331 | udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0); | |
332 | ||
d4cada4a | 333 | /* precompute partial secondary hash */ |
30fff923 ED |
334 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
335 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr); | |
db8dac20 DM |
336 | } |
337 | ||
645ca708 ED |
338 | static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr, |
339 | unsigned short hnum, | |
340 | __be16 sport, __be32 daddr, __be16 dport, int dif) | |
341 | { | |
342 | int score = -1; | |
343 | ||
d4cada4a | 344 | if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum && |
645ca708 ED |
345 | !ipv6_only_sock(sk)) { |
346 | struct inet_sock *inet = inet_sk(sk); | |
347 | ||
ba418fa3 | 348 | score = (sk->sk_family == PF_INET ? 2 : 1); |
c720c7e8 ED |
349 | if (inet->inet_rcv_saddr) { |
350 | if (inet->inet_rcv_saddr != daddr) | |
645ca708 | 351 | return -1; |
ba418fa3 | 352 | score += 4; |
645ca708 | 353 | } |
c720c7e8 ED |
354 | if (inet->inet_daddr) { |
355 | if (inet->inet_daddr != saddr) | |
645ca708 | 356 | return -1; |
ba418fa3 | 357 | score += 4; |
645ca708 | 358 | } |
c720c7e8 ED |
359 | if (inet->inet_dport) { |
360 | if (inet->inet_dport != sport) | |
645ca708 | 361 | return -1; |
ba418fa3 | 362 | score += 4; |
645ca708 ED |
363 | } |
364 | if (sk->sk_bound_dev_if) { | |
365 | if (sk->sk_bound_dev_if != dif) | |
366 | return -1; | |
ba418fa3 | 367 | score += 4; |
645ca708 ED |
368 | } |
369 | } | |
370 | return score; | |
371 | } | |
372 | ||
5051ebd2 ED |
373 | /* |
374 | * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num) | |
375 | */ | |
5051ebd2 ED |
376 | static inline int compute_score2(struct sock *sk, struct net *net, |
377 | __be32 saddr, __be16 sport, | |
378 | __be32 daddr, unsigned int hnum, int dif) | |
379 | { | |
380 | int score = -1; | |
381 | ||
382 | if (net_eq(sock_net(sk), net) && !ipv6_only_sock(sk)) { | |
383 | struct inet_sock *inet = inet_sk(sk); | |
384 | ||
385 | if (inet->inet_rcv_saddr != daddr) | |
386 | return -1; | |
387 | if (inet->inet_num != hnum) | |
388 | return -1; | |
389 | ||
ba418fa3 | 390 | score = (sk->sk_family == PF_INET ? 2 : 1); |
5051ebd2 ED |
391 | if (inet->inet_daddr) { |
392 | if (inet->inet_daddr != saddr) | |
393 | return -1; | |
ba418fa3 | 394 | score += 4; |
5051ebd2 ED |
395 | } |
396 | if (inet->inet_dport) { | |
397 | if (inet->inet_dport != sport) | |
398 | return -1; | |
ba418fa3 | 399 | score += 4; |
5051ebd2 ED |
400 | } |
401 | if (sk->sk_bound_dev_if) { | |
402 | if (sk->sk_bound_dev_if != dif) | |
403 | return -1; | |
ba418fa3 | 404 | score += 4; |
5051ebd2 ED |
405 | } |
406 | } | |
407 | return score; | |
408 | } | |
409 | ||
5051ebd2 ED |
410 | |
411 | /* called with read_rcu_lock() */ | |
412 | static struct sock *udp4_lib_lookup2(struct net *net, | |
413 | __be32 saddr, __be16 sport, | |
414 | __be32 daddr, unsigned int hnum, int dif, | |
415 | struct udp_hslot *hslot2, unsigned int slot2) | |
416 | { | |
417 | struct sock *sk, *result; | |
418 | struct hlist_nulls_node *node; | |
ba418fa3 TH |
419 | int score, badness, matches = 0, reuseport = 0; |
420 | u32 hash = 0; | |
5051ebd2 ED |
421 | |
422 | begin: | |
423 | result = NULL; | |
ba418fa3 | 424 | badness = 0; |
5051ebd2 ED |
425 | udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { |
426 | score = compute_score2(sk, net, saddr, sport, | |
427 | daddr, hnum, dif); | |
428 | if (score > badness) { | |
429 | result = sk; | |
430 | badness = score; | |
ba418fa3 TH |
431 | reuseport = sk->sk_reuseport; |
432 | if (reuseport) { | |
433 | hash = inet_ehashfn(net, daddr, hnum, | |
7c0cadc6 | 434 | saddr, sport); |
ba418fa3 TH |
435 | matches = 1; |
436 | } | |
437 | } else if (score == badness && reuseport) { | |
438 | matches++; | |
439 | if (((u64)hash * matches) >> 32 == 0) | |
440 | result = sk; | |
441 | hash = next_pseudo_random32(hash); | |
5051ebd2 ED |
442 | } |
443 | } | |
444 | /* | |
445 | * if the nulls value we got at the end of this lookup is | |
446 | * not the expected one, we must restart lookup. | |
447 | * We probably met an item that was moved to another chain. | |
448 | */ | |
449 | if (get_nulls_value(node) != slot2) | |
450 | goto begin; | |
5051ebd2 | 451 | if (result) { |
c31504dc | 452 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
5051ebd2 ED |
453 | result = NULL; |
454 | else if (unlikely(compute_score2(result, net, saddr, sport, | |
455 | daddr, hnum, dif) < badness)) { | |
456 | sock_put(result); | |
457 | goto begin; | |
458 | } | |
459 | } | |
460 | return result; | |
461 | } | |
462 | ||
db8dac20 DM |
463 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
464 | * harder than this. -DaveM | |
465 | */ | |
fce82338 | 466 | struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, |
db8dac20 | 467 | __be16 sport, __be32 daddr, __be16 dport, |
645ca708 | 468 | int dif, struct udp_table *udptable) |
db8dac20 | 469 | { |
271b72c7 | 470 | struct sock *sk, *result; |
88ab1932 | 471 | struct hlist_nulls_node *node; |
db8dac20 | 472 | unsigned short hnum = ntohs(dport); |
5051ebd2 ED |
473 | unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); |
474 | struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; | |
ba418fa3 TH |
475 | int score, badness, matches = 0, reuseport = 0; |
476 | u32 hash = 0; | |
645ca708 | 477 | |
271b72c7 | 478 | rcu_read_lock(); |
5051ebd2 ED |
479 | if (hslot->count > 10) { |
480 | hash2 = udp4_portaddr_hash(net, daddr, hnum); | |
481 | slot2 = hash2 & udptable->mask; | |
482 | hslot2 = &udptable->hash2[slot2]; | |
483 | if (hslot->count < hslot2->count) | |
484 | goto begin; | |
485 | ||
486 | result = udp4_lib_lookup2(net, saddr, sport, | |
487 | daddr, hnum, dif, | |
488 | hslot2, slot2); | |
489 | if (!result) { | |
0eae88f3 | 490 | hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum); |
5051ebd2 ED |
491 | slot2 = hash2 & udptable->mask; |
492 | hslot2 = &udptable->hash2[slot2]; | |
493 | if (hslot->count < hslot2->count) | |
494 | goto begin; | |
495 | ||
1223c67c | 496 | result = udp4_lib_lookup2(net, saddr, sport, |
0eae88f3 | 497 | htonl(INADDR_ANY), hnum, dif, |
5051ebd2 ED |
498 | hslot2, slot2); |
499 | } | |
500 | rcu_read_unlock(); | |
501 | return result; | |
502 | } | |
271b72c7 ED |
503 | begin: |
504 | result = NULL; | |
ba418fa3 | 505 | badness = 0; |
88ab1932 | 506 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { |
645ca708 ED |
507 | score = compute_score(sk, net, saddr, hnum, sport, |
508 | daddr, dport, dif); | |
509 | if (score > badness) { | |
510 | result = sk; | |
511 | badness = score; | |
ba418fa3 TH |
512 | reuseport = sk->sk_reuseport; |
513 | if (reuseport) { | |
514 | hash = inet_ehashfn(net, daddr, hnum, | |
7c0cadc6 | 515 | saddr, sport); |
ba418fa3 TH |
516 | matches = 1; |
517 | } | |
518 | } else if (score == badness && reuseport) { | |
519 | matches++; | |
520 | if (((u64)hash * matches) >> 32 == 0) | |
521 | result = sk; | |
522 | hash = next_pseudo_random32(hash); | |
db8dac20 DM |
523 | } |
524 | } | |
88ab1932 ED |
525 | /* |
526 | * if the nulls value we got at the end of this lookup is | |
527 | * not the expected one, we must restart lookup. | |
528 | * We probably met an item that was moved to another chain. | |
529 | */ | |
5051ebd2 | 530 | if (get_nulls_value(node) != slot) |
88ab1932 ED |
531 | goto begin; |
532 | ||
271b72c7 | 533 | if (result) { |
c31504dc | 534 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
271b72c7 ED |
535 | result = NULL; |
536 | else if (unlikely(compute_score(result, net, saddr, hnum, sport, | |
537 | daddr, dport, dif) < badness)) { | |
538 | sock_put(result); | |
539 | goto begin; | |
540 | } | |
541 | } | |
542 | rcu_read_unlock(); | |
db8dac20 DM |
543 | return result; |
544 | } | |
fce82338 | 545 | EXPORT_SYMBOL_GPL(__udp4_lib_lookup); |
db8dac20 | 546 | |
607c4aaf KK |
547 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
548 | __be16 sport, __be16 dport, | |
645ca708 | 549 | struct udp_table *udptable) |
607c4aaf | 550 | { |
23542618 | 551 | struct sock *sk; |
607c4aaf KK |
552 | const struct iphdr *iph = ip_hdr(skb); |
553 | ||
23542618 KK |
554 | if (unlikely(sk = skb_steal_sock(skb))) |
555 | return sk; | |
556 | else | |
adf30907 | 557 | return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport, |
23542618 KK |
558 | iph->daddr, dport, inet_iif(skb), |
559 | udptable); | |
607c4aaf KK |
560 | } |
561 | ||
bcd41303 KK |
562 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
563 | __be32 daddr, __be16 dport, int dif) | |
564 | { | |
645ca708 | 565 | return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table); |
bcd41303 KK |
566 | } |
567 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
568 | ||
421b3885 SB |
569 | static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk, |
570 | __be16 loc_port, __be32 loc_addr, | |
571 | __be16 rmt_port, __be32 rmt_addr, | |
572 | int dif, unsigned short hnum) | |
573 | { | |
574 | struct inet_sock *inet = inet_sk(sk); | |
575 | ||
576 | if (!net_eq(sock_net(sk), net) || | |
577 | udp_sk(sk)->udp_port_hash != hnum || | |
578 | (inet->inet_daddr && inet->inet_daddr != rmt_addr) || | |
579 | (inet->inet_dport != rmt_port && inet->inet_dport) || | |
580 | (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) || | |
581 | ipv6_only_sock(sk) || | |
582 | (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) | |
583 | return false; | |
584 | if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif)) | |
585 | return false; | |
586 | return true; | |
587 | } | |
588 | ||
920a4611 | 589 | static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk, |
db8dac20 DM |
590 | __be16 loc_port, __be32 loc_addr, |
591 | __be16 rmt_port, __be32 rmt_addr, | |
592 | int dif) | |
593 | { | |
88ab1932 | 594 | struct hlist_nulls_node *node; |
db8dac20 DM |
595 | struct sock *s = sk; |
596 | unsigned short hnum = ntohs(loc_port); | |
597 | ||
88ab1932 | 598 | sk_nulls_for_each_from(s, node) { |
421b3885 SB |
599 | if (__udp_is_mcast_sock(net, s, |
600 | loc_port, loc_addr, | |
601 | rmt_port, rmt_addr, | |
602 | dif, hnum)) | |
603 | goto found; | |
db8dac20 DM |
604 | } |
605 | s = NULL; | |
606 | found: | |
607 | return s; | |
608 | } | |
609 | ||
610 | /* | |
611 | * This routine is called by the ICMP module when it gets some | |
612 | * sort of error condition. If err < 0 then the socket should | |
613 | * be closed and the error returned to the user. If err > 0 | |
614 | * it's just the icmp type << 8 | icmp code. | |
615 | * Header points to the ip header of the error packet. We move | |
616 | * on past this. Then (as it used to claim before adjustment) | |
617 | * header points to the first 8 bytes of the udp header. We need | |
618 | * to find the appropriate port. | |
619 | */ | |
620 | ||
645ca708 | 621 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
622 | { |
623 | struct inet_sock *inet; | |
b71d1d42 | 624 | const struct iphdr *iph = (const struct iphdr *)skb->data; |
c482c568 | 625 | struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2)); |
db8dac20 DM |
626 | const int type = icmp_hdr(skb)->type; |
627 | const int code = icmp_hdr(skb)->code; | |
628 | struct sock *sk; | |
629 | int harderr; | |
630 | int err; | |
fd54d716 | 631 | struct net *net = dev_net(skb->dev); |
db8dac20 | 632 | |
fd54d716 | 633 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
db8dac20 DM |
634 | iph->saddr, uh->source, skb->dev->ifindex, udptable); |
635 | if (sk == NULL) { | |
dcfc23ca | 636 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
637 | return; /* No socket for error */ |
638 | } | |
639 | ||
640 | err = 0; | |
641 | harderr = 0; | |
642 | inet = inet_sk(sk); | |
643 | ||
644 | switch (type) { | |
645 | default: | |
646 | case ICMP_TIME_EXCEEDED: | |
647 | err = EHOSTUNREACH; | |
648 | break; | |
649 | case ICMP_SOURCE_QUENCH: | |
650 | goto out; | |
651 | case ICMP_PARAMETERPROB: | |
652 | err = EPROTO; | |
653 | harderr = 1; | |
654 | break; | |
655 | case ICMP_DEST_UNREACH: | |
656 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
36393395 | 657 | ipv4_sk_update_pmtu(skb, sk, info); |
db8dac20 DM |
658 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { |
659 | err = EMSGSIZE; | |
660 | harderr = 1; | |
661 | break; | |
662 | } | |
663 | goto out; | |
664 | } | |
665 | err = EHOSTUNREACH; | |
666 | if (code <= NR_ICMP_UNREACH) { | |
667 | harderr = icmp_err_convert[code].fatal; | |
668 | err = icmp_err_convert[code].errno; | |
669 | } | |
670 | break; | |
55be7a9c DM |
671 | case ICMP_REDIRECT: |
672 | ipv4_sk_redirect(skb, sk); | |
1a462d18 | 673 | goto out; |
db8dac20 DM |
674 | } |
675 | ||
676 | /* | |
677 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
678 | * 4.1.3.3. | |
679 | */ | |
680 | if (!inet->recverr) { | |
681 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
682 | goto out; | |
b1faf566 | 683 | } else |
c482c568 | 684 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1)); |
b1faf566 | 685 | |
db8dac20 DM |
686 | sk->sk_err = err; |
687 | sk->sk_error_report(sk); | |
688 | out: | |
689 | sock_put(sk); | |
690 | } | |
691 | ||
692 | void udp_err(struct sk_buff *skb, u32 info) | |
693 | { | |
645ca708 | 694 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
695 | } |
696 | ||
697 | /* | |
698 | * Throw away all pending data and cancel the corking. Socket is locked. | |
699 | */ | |
36d926b9 | 700 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
701 | { |
702 | struct udp_sock *up = udp_sk(sk); | |
703 | ||
704 | if (up->pending) { | |
705 | up->len = 0; | |
706 | up->pending = 0; | |
707 | ip_flush_pending_frames(sk); | |
708 | } | |
709 | } | |
36d926b9 | 710 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
711 | |
712 | /** | |
f6b9664f | 713 | * udp4_hwcsum - handle outgoing HW checksumming |
db8dac20 DM |
714 | * @skb: sk_buff containing the filled-in UDP header |
715 | * (checksum field must be zeroed out) | |
f6b9664f HX |
716 | * @src: source IP address |
717 | * @dst: destination IP address | |
db8dac20 | 718 | */ |
c26bf4a5 | 719 | void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst) |
db8dac20 | 720 | { |
db8dac20 | 721 | struct udphdr *uh = udp_hdr(skb); |
f6b9664f HX |
722 | struct sk_buff *frags = skb_shinfo(skb)->frag_list; |
723 | int offset = skb_transport_offset(skb); | |
724 | int len = skb->len - offset; | |
725 | int hlen = len; | |
db8dac20 DM |
726 | __wsum csum = 0; |
727 | ||
f6b9664f | 728 | if (!frags) { |
db8dac20 DM |
729 | /* |
730 | * Only one fragment on the socket. | |
731 | */ | |
732 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
733 | skb->csum_offset = offsetof(struct udphdr, check); | |
f6b9664f HX |
734 | uh->check = ~csum_tcpudp_magic(src, dst, len, |
735 | IPPROTO_UDP, 0); | |
db8dac20 DM |
736 | } else { |
737 | /* | |
738 | * HW-checksum won't work as there are two or more | |
739 | * fragments on the socket so that all csums of sk_buffs | |
740 | * should be together | |
741 | */ | |
f6b9664f HX |
742 | do { |
743 | csum = csum_add(csum, frags->csum); | |
744 | hlen -= frags->len; | |
745 | } while ((frags = frags->next)); | |
db8dac20 | 746 | |
f6b9664f | 747 | csum = skb_checksum(skb, offset, hlen, csum); |
db8dac20 DM |
748 | skb->ip_summed = CHECKSUM_NONE; |
749 | ||
db8dac20 DM |
750 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); |
751 | if (uh->check == 0) | |
752 | uh->check = CSUM_MANGLED_0; | |
753 | } | |
754 | } | |
c26bf4a5 | 755 | EXPORT_SYMBOL_GPL(udp4_hwcsum); |
db8dac20 | 756 | |
79ab0531 | 757 | static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4) |
db8dac20 | 758 | { |
f6b9664f | 759 | struct sock *sk = skb->sk; |
db8dac20 | 760 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
761 | struct udphdr *uh; |
762 | int err = 0; | |
763 | int is_udplite = IS_UDPLITE(sk); | |
f6b9664f HX |
764 | int offset = skb_transport_offset(skb); |
765 | int len = skb->len - offset; | |
db8dac20 DM |
766 | __wsum csum = 0; |
767 | ||
db8dac20 DM |
768 | /* |
769 | * Create a UDP header | |
770 | */ | |
771 | uh = udp_hdr(skb); | |
f6b9664f | 772 | uh->source = inet->inet_sport; |
79ab0531 | 773 | uh->dest = fl4->fl4_dport; |
f6b9664f | 774 | uh->len = htons(len); |
db8dac20 DM |
775 | uh->check = 0; |
776 | ||
777 | if (is_udplite) /* UDP-Lite */ | |
f6b9664f | 778 | csum = udplite_csum(skb); |
db8dac20 DM |
779 | |
780 | else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */ | |
781 | ||
782 | skb->ip_summed = CHECKSUM_NONE; | |
783 | goto send; | |
784 | ||
785 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
786 | ||
79ab0531 | 787 | udp4_hwcsum(skb, fl4->saddr, fl4->daddr); |
db8dac20 DM |
788 | goto send; |
789 | ||
f6b9664f HX |
790 | } else |
791 | csum = udp_csum(skb); | |
db8dac20 DM |
792 | |
793 | /* add protocol-dependent pseudo-header */ | |
79ab0531 | 794 | uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len, |
c482c568 | 795 | sk->sk_protocol, csum); |
db8dac20 DM |
796 | if (uh->check == 0) |
797 | uh->check = CSUM_MANGLED_0; | |
798 | ||
799 | send: | |
b5ec8eea | 800 | err = ip_send_skb(sock_net(sk), skb); |
6ce9e7b5 ED |
801 | if (err) { |
802 | if (err == -ENOBUFS && !inet->recverr) { | |
803 | UDP_INC_STATS_USER(sock_net(sk), | |
804 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
805 | err = 0; | |
806 | } | |
807 | } else | |
808 | UDP_INC_STATS_USER(sock_net(sk), | |
809 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
f6b9664f HX |
810 | return err; |
811 | } | |
812 | ||
813 | /* | |
814 | * Push out all pending data as one UDP datagram. Socket is locked. | |
815 | */ | |
8822b64a | 816 | int udp_push_pending_frames(struct sock *sk) |
f6b9664f HX |
817 | { |
818 | struct udp_sock *up = udp_sk(sk); | |
819 | struct inet_sock *inet = inet_sk(sk); | |
b6f21b26 | 820 | struct flowi4 *fl4 = &inet->cork.fl.u.ip4; |
f6b9664f HX |
821 | struct sk_buff *skb; |
822 | int err = 0; | |
823 | ||
77968b78 | 824 | skb = ip_finish_skb(sk, fl4); |
f6b9664f HX |
825 | if (!skb) |
826 | goto out; | |
827 | ||
79ab0531 | 828 | err = udp_send_skb(skb, fl4); |
f6b9664f | 829 | |
db8dac20 DM |
830 | out: |
831 | up->len = 0; | |
832 | up->pending = 0; | |
db8dac20 DM |
833 | return err; |
834 | } | |
8822b64a | 835 | EXPORT_SYMBOL(udp_push_pending_frames); |
db8dac20 DM |
836 | |
837 | int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
838 | size_t len) | |
839 | { | |
840 | struct inet_sock *inet = inet_sk(sk); | |
841 | struct udp_sock *up = udp_sk(sk); | |
e474995f | 842 | struct flowi4 fl4_stack; |
b6f21b26 | 843 | struct flowi4 *fl4; |
db8dac20 DM |
844 | int ulen = len; |
845 | struct ipcm_cookie ipc; | |
846 | struct rtable *rt = NULL; | |
847 | int free = 0; | |
848 | int connected = 0; | |
849 | __be32 daddr, faddr, saddr; | |
850 | __be16 dport; | |
851 | u8 tos; | |
852 | int err, is_udplite = IS_UDPLITE(sk); | |
853 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
854 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
903ab86d | 855 | struct sk_buff *skb; |
f6d8bd05 | 856 | struct ip_options_data opt_copy; |
db8dac20 DM |
857 | |
858 | if (len > 0xFFFF) | |
859 | return -EMSGSIZE; | |
860 | ||
861 | /* | |
862 | * Check the flags. | |
863 | */ | |
864 | ||
c482c568 | 865 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */ |
db8dac20 DM |
866 | return -EOPNOTSUPP; |
867 | ||
868 | ipc.opt = NULL; | |
2244d07b | 869 | ipc.tx_flags = 0; |
aa661581 FF |
870 | ipc.ttl = 0; |
871 | ipc.tos = -1; | |
db8dac20 | 872 | |
903ab86d HX |
873 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
874 | ||
f5fca608 | 875 | fl4 = &inet->cork.fl.u.ip4; |
db8dac20 DM |
876 | if (up->pending) { |
877 | /* | |
878 | * There are pending frames. | |
879 | * The socket lock must be held while it's corked. | |
880 | */ | |
881 | lock_sock(sk); | |
882 | if (likely(up->pending)) { | |
883 | if (unlikely(up->pending != AF_INET)) { | |
884 | release_sock(sk); | |
885 | return -EINVAL; | |
886 | } | |
887 | goto do_append_data; | |
888 | } | |
889 | release_sock(sk); | |
890 | } | |
891 | ulen += sizeof(struct udphdr); | |
892 | ||
893 | /* | |
894 | * Get and verify the address. | |
895 | */ | |
896 | if (msg->msg_name) { | |
5e73ea1a | 897 | struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name; |
db8dac20 DM |
898 | if (msg->msg_namelen < sizeof(*usin)) |
899 | return -EINVAL; | |
900 | if (usin->sin_family != AF_INET) { | |
901 | if (usin->sin_family != AF_UNSPEC) | |
902 | return -EAFNOSUPPORT; | |
903 | } | |
904 | ||
905 | daddr = usin->sin_addr.s_addr; | |
906 | dport = usin->sin_port; | |
907 | if (dport == 0) | |
908 | return -EINVAL; | |
909 | } else { | |
910 | if (sk->sk_state != TCP_ESTABLISHED) | |
911 | return -EDESTADDRREQ; | |
c720c7e8 ED |
912 | daddr = inet->inet_daddr; |
913 | dport = inet->inet_dport; | |
db8dac20 DM |
914 | /* Open fast path for connected socket. |
915 | Route will not be used, if at least one option is set. | |
916 | */ | |
917 | connected = 1; | |
918 | } | |
c720c7e8 | 919 | ipc.addr = inet->inet_saddr; |
db8dac20 DM |
920 | |
921 | ipc.oif = sk->sk_bound_dev_if; | |
bf84a010 DB |
922 | |
923 | sock_tx_timestamp(sk, &ipc.tx_flags); | |
924 | ||
db8dac20 | 925 | if (msg->msg_controllen) { |
3b1e0a65 | 926 | err = ip_cmsg_send(sock_net(sk), msg, &ipc); |
db8dac20 DM |
927 | if (err) |
928 | return err; | |
929 | if (ipc.opt) | |
930 | free = 1; | |
931 | connected = 0; | |
932 | } | |
f6d8bd05 ED |
933 | if (!ipc.opt) { |
934 | struct ip_options_rcu *inet_opt; | |
935 | ||
936 | rcu_read_lock(); | |
937 | inet_opt = rcu_dereference(inet->inet_opt); | |
938 | if (inet_opt) { | |
939 | memcpy(&opt_copy, inet_opt, | |
940 | sizeof(*inet_opt) + inet_opt->opt.optlen); | |
941 | ipc.opt = &opt_copy.opt; | |
942 | } | |
943 | rcu_read_unlock(); | |
944 | } | |
db8dac20 DM |
945 | |
946 | saddr = ipc.addr; | |
947 | ipc.addr = faddr = daddr; | |
948 | ||
f6d8bd05 | 949 | if (ipc.opt && ipc.opt->opt.srr) { |
db8dac20 DM |
950 | if (!daddr) |
951 | return -EINVAL; | |
f6d8bd05 | 952 | faddr = ipc.opt->opt.faddr; |
db8dac20 DM |
953 | connected = 0; |
954 | } | |
aa661581 | 955 | tos = get_rttos(&ipc, inet); |
db8dac20 DM |
956 | if (sock_flag(sk, SOCK_LOCALROUTE) || |
957 | (msg->msg_flags & MSG_DONTROUTE) || | |
f6d8bd05 | 958 | (ipc.opt && ipc.opt->opt.is_strictroute)) { |
db8dac20 DM |
959 | tos |= RTO_ONLINK; |
960 | connected = 0; | |
961 | } | |
962 | ||
963 | if (ipv4_is_multicast(daddr)) { | |
964 | if (!ipc.oif) | |
965 | ipc.oif = inet->mc_index; | |
966 | if (!saddr) | |
967 | saddr = inet->mc_addr; | |
968 | connected = 0; | |
76e21053 EH |
969 | } else if (!ipc.oif) |
970 | ipc.oif = inet->uc_index; | |
db8dac20 DM |
971 | |
972 | if (connected) | |
c482c568 | 973 | rt = (struct rtable *)sk_dst_check(sk, 0); |
db8dac20 DM |
974 | |
975 | if (rt == NULL) { | |
84a3aa00 PE |
976 | struct net *net = sock_net(sk); |
977 | ||
e474995f DM |
978 | fl4 = &fl4_stack; |
979 | flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos, | |
c0951cbc DM |
980 | RT_SCOPE_UNIVERSE, sk->sk_protocol, |
981 | inet_sk_flowi_flags(sk)|FLOWI_FLAG_CAN_SLEEP, | |
982 | faddr, saddr, dport, inet->inet_sport); | |
983 | ||
e474995f DM |
984 | security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); |
985 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe DM |
986 | if (IS_ERR(rt)) { |
987 | err = PTR_ERR(rt); | |
06dc94b1 | 988 | rt = NULL; |
db8dac20 | 989 | if (err == -ENETUNREACH) |
7c73a6fa | 990 | IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
991 | goto out; |
992 | } | |
993 | ||
994 | err = -EACCES; | |
995 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
996 | !sock_flag(sk, SOCK_BROADCAST)) | |
997 | goto out; | |
998 | if (connected) | |
d8d1f30b | 999 | sk_dst_set(sk, dst_clone(&rt->dst)); |
db8dac20 DM |
1000 | } |
1001 | ||
1002 | if (msg->msg_flags&MSG_CONFIRM) | |
1003 | goto do_confirm; | |
1004 | back_from_confirm: | |
1005 | ||
e474995f | 1006 | saddr = fl4->saddr; |
db8dac20 | 1007 | if (!ipc.addr) |
e474995f | 1008 | daddr = ipc.addr = fl4->daddr; |
db8dac20 | 1009 | |
903ab86d HX |
1010 | /* Lockless fast path for the non-corking case. */ |
1011 | if (!corkreq) { | |
77968b78 | 1012 | skb = ip_make_skb(sk, fl4, getfrag, msg->msg_iov, ulen, |
903ab86d HX |
1013 | sizeof(struct udphdr), &ipc, &rt, |
1014 | msg->msg_flags); | |
1015 | err = PTR_ERR(skb); | |
50c3a487 | 1016 | if (!IS_ERR_OR_NULL(skb)) |
79ab0531 | 1017 | err = udp_send_skb(skb, fl4); |
903ab86d HX |
1018 | goto out; |
1019 | } | |
1020 | ||
db8dac20 DM |
1021 | lock_sock(sk); |
1022 | if (unlikely(up->pending)) { | |
1023 | /* The socket is already corked while preparing it. */ | |
1024 | /* ... which is an evident application bug. --ANK */ | |
1025 | release_sock(sk); | |
1026 | ||
afd46503 | 1027 | LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("cork app bug 2\n")); |
db8dac20 DM |
1028 | err = -EINVAL; |
1029 | goto out; | |
1030 | } | |
1031 | /* | |
1032 | * Now cork the socket to pend data. | |
1033 | */ | |
b6f21b26 DM |
1034 | fl4 = &inet->cork.fl.u.ip4; |
1035 | fl4->daddr = daddr; | |
1036 | fl4->saddr = saddr; | |
9cce96df DM |
1037 | fl4->fl4_dport = dport; |
1038 | fl4->fl4_sport = inet->inet_sport; | |
db8dac20 DM |
1039 | up->pending = AF_INET; |
1040 | ||
1041 | do_append_data: | |
1042 | up->len += ulen; | |
f5fca608 DM |
1043 | err = ip_append_data(sk, fl4, getfrag, msg->msg_iov, ulen, |
1044 | sizeof(struct udphdr), &ipc, &rt, | |
1045 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | |
db8dac20 DM |
1046 | if (err) |
1047 | udp_flush_pending_frames(sk); | |
1048 | else if (!corkreq) | |
1049 | err = udp_push_pending_frames(sk); | |
1050 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
1051 | up->pending = 0; | |
1052 | release_sock(sk); | |
1053 | ||
1054 | out: | |
1055 | ip_rt_put(rt); | |
1056 | if (free) | |
1057 | kfree(ipc.opt); | |
1058 | if (!err) | |
1059 | return len; | |
1060 | /* | |
1061 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1062 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1063 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1064 | * things). We could add another new stat but at least for now that | |
1065 | * seems like overkill. | |
1066 | */ | |
1067 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
1068 | UDP_INC_STATS_USER(sock_net(sk), |
1069 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
1070 | } |
1071 | return err; | |
1072 | ||
1073 | do_confirm: | |
d8d1f30b | 1074 | dst_confirm(&rt->dst); |
db8dac20 DM |
1075 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1076 | goto back_from_confirm; | |
1077 | err = 0; | |
1078 | goto out; | |
1079 | } | |
c482c568 | 1080 | EXPORT_SYMBOL(udp_sendmsg); |
db8dac20 DM |
1081 | |
1082 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
1083 | size_t size, int flags) | |
1084 | { | |
f5fca608 | 1085 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
1086 | struct udp_sock *up = udp_sk(sk); |
1087 | int ret; | |
1088 | ||
1089 | if (!up->pending) { | |
1090 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
1091 | ||
1092 | /* Call udp_sendmsg to specify destination address which | |
1093 | * sendpage interface can't pass. | |
1094 | * This will succeed only when the socket is connected. | |
1095 | */ | |
1096 | ret = udp_sendmsg(NULL, sk, &msg, 0); | |
1097 | if (ret < 0) | |
1098 | return ret; | |
1099 | } | |
1100 | ||
1101 | lock_sock(sk); | |
1102 | ||
1103 | if (unlikely(!up->pending)) { | |
1104 | release_sock(sk); | |
1105 | ||
afd46503 | 1106 | LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("udp cork app bug 3\n")); |
db8dac20 DM |
1107 | return -EINVAL; |
1108 | } | |
1109 | ||
f5fca608 DM |
1110 | ret = ip_append_page(sk, &inet->cork.fl.u.ip4, |
1111 | page, offset, size, flags); | |
db8dac20 DM |
1112 | if (ret == -EOPNOTSUPP) { |
1113 | release_sock(sk); | |
1114 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
1115 | size, flags); | |
1116 | } | |
1117 | if (ret < 0) { | |
1118 | udp_flush_pending_frames(sk); | |
1119 | goto out; | |
1120 | } | |
1121 | ||
1122 | up->len += size; | |
1123 | if (!(up->corkflag || (flags&MSG_MORE))) | |
1124 | ret = udp_push_pending_frames(sk); | |
1125 | if (!ret) | |
1126 | ret = size; | |
1127 | out: | |
1128 | release_sock(sk); | |
1129 | return ret; | |
1130 | } | |
1131 | ||
85584672 ED |
1132 | |
1133 | /** | |
1134 | * first_packet_length - return length of first packet in receive queue | |
1135 | * @sk: socket | |
1136 | * | |
1137 | * Drops all bad checksum frames, until a valid one is found. | |
1138 | * Returns the length of found skb, or 0 if none is found. | |
1139 | */ | |
1140 | static unsigned int first_packet_length(struct sock *sk) | |
1141 | { | |
1142 | struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue; | |
1143 | struct sk_buff *skb; | |
1144 | unsigned int res; | |
1145 | ||
1146 | __skb_queue_head_init(&list_kill); | |
1147 | ||
1148 | spin_lock_bh(&rcvq->lock); | |
1149 | while ((skb = skb_peek(rcvq)) != NULL && | |
1150 | udp_lib_checksum_complete(skb)) { | |
6a5dc9e5 ED |
1151 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, |
1152 | IS_UDPLITE(sk)); | |
85584672 ED |
1153 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, |
1154 | IS_UDPLITE(sk)); | |
8edf19c2 | 1155 | atomic_inc(&sk->sk_drops); |
85584672 ED |
1156 | __skb_unlink(skb, rcvq); |
1157 | __skb_queue_tail(&list_kill, skb); | |
1158 | } | |
1159 | res = skb ? skb->len : 0; | |
1160 | spin_unlock_bh(&rcvq->lock); | |
1161 | ||
1162 | if (!skb_queue_empty(&list_kill)) { | |
8a74ad60 ED |
1163 | bool slow = lock_sock_fast(sk); |
1164 | ||
85584672 ED |
1165 | __skb_queue_purge(&list_kill); |
1166 | sk_mem_reclaim_partial(sk); | |
8a74ad60 | 1167 | unlock_sock_fast(sk, slow); |
85584672 ED |
1168 | } |
1169 | return res; | |
1170 | } | |
1171 | ||
1da177e4 LT |
1172 | /* |
1173 | * IOCTL requests applicable to the UDP protocol | |
1174 | */ | |
e905a9ed | 1175 | |
1da177e4 LT |
1176 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
1177 | { | |
6516c655 SH |
1178 | switch (cmd) { |
1179 | case SIOCOUTQ: | |
1da177e4 | 1180 | { |
31e6d363 ED |
1181 | int amount = sk_wmem_alloc_get(sk); |
1182 | ||
6516c655 SH |
1183 | return put_user(amount, (int __user *)arg); |
1184 | } | |
1da177e4 | 1185 | |
6516c655 SH |
1186 | case SIOCINQ: |
1187 | { | |
85584672 | 1188 | unsigned int amount = first_packet_length(sk); |
6516c655 | 1189 | |
85584672 | 1190 | if (amount) |
6516c655 SH |
1191 | /* |
1192 | * We will only return the amount | |
1193 | * of this packet since that is all | |
1194 | * that will be read. | |
1195 | */ | |
85584672 ED |
1196 | amount -= sizeof(struct udphdr); |
1197 | ||
6516c655 SH |
1198 | return put_user(amount, (int __user *)arg); |
1199 | } | |
1da177e4 | 1200 | |
6516c655 SH |
1201 | default: |
1202 | return -ENOIOCTLCMD; | |
1da177e4 | 1203 | } |
6516c655 SH |
1204 | |
1205 | return 0; | |
1da177e4 | 1206 | } |
c482c568 | 1207 | EXPORT_SYMBOL(udp_ioctl); |
1da177e4 | 1208 | |
db8dac20 DM |
1209 | /* |
1210 | * This should be easy, if there is something there we | |
1211 | * return it, otherwise we block. | |
1212 | */ | |
1213 | ||
1214 | int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
1215 | size_t len, int noblock, int flags, int *addr_len) | |
1216 | { | |
1217 | struct inet_sock *inet = inet_sk(sk); | |
1218 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; | |
1219 | struct sk_buff *skb; | |
59c2cdae | 1220 | unsigned int ulen, copied; |
3f518bf7 | 1221 | int peeked, off = 0; |
db8dac20 DM |
1222 | int err; |
1223 | int is_udplite = IS_UDPLITE(sk); | |
8a74ad60 | 1224 | bool slow; |
db8dac20 DM |
1225 | |
1226 | /* | |
1227 | * Check any passed addresses | |
1228 | */ | |
1229 | if (addr_len) | |
c482c568 | 1230 | *addr_len = sizeof(*sin); |
db8dac20 DM |
1231 | |
1232 | if (flags & MSG_ERRQUEUE) | |
1233 | return ip_recv_error(sk, msg, len); | |
1234 | ||
1235 | try_again: | |
1236 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), | |
3f518bf7 | 1237 | &peeked, &off, &err); |
db8dac20 DM |
1238 | if (!skb) |
1239 | goto out; | |
1240 | ||
1241 | ulen = skb->len - sizeof(struct udphdr); | |
59c2cdae DM |
1242 | copied = len; |
1243 | if (copied > ulen) | |
1244 | copied = ulen; | |
1245 | else if (copied < ulen) | |
db8dac20 DM |
1246 | msg->msg_flags |= MSG_TRUNC; |
1247 | ||
1248 | /* | |
1249 | * If checksum is needed at all, try to do it while copying the | |
1250 | * data. If the data is truncated, or if we only want a partial | |
1251 | * coverage checksum (UDP-Lite), do it before the copy. | |
1252 | */ | |
1253 | ||
59c2cdae | 1254 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { |
db8dac20 DM |
1255 | if (udp_lib_checksum_complete(skb)) |
1256 | goto csum_copy_err; | |
1257 | } | |
1258 | ||
1259 | if (skb_csum_unnecessary(skb)) | |
1260 | err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), | |
59c2cdae | 1261 | msg->msg_iov, copied); |
db8dac20 | 1262 | else { |
c482c568 ED |
1263 | err = skb_copy_and_csum_datagram_iovec(skb, |
1264 | sizeof(struct udphdr), | |
1265 | msg->msg_iov); | |
db8dac20 DM |
1266 | |
1267 | if (err == -EINVAL) | |
1268 | goto csum_copy_err; | |
1269 | } | |
1270 | ||
22911fc5 ED |
1271 | if (unlikely(err)) { |
1272 | trace_kfree_skb(skb, udp_recvmsg); | |
979402b1 ED |
1273 | if (!peeked) { |
1274 | atomic_inc(&sk->sk_drops); | |
1275 | UDP_INC_STATS_USER(sock_net(sk), | |
1276 | UDP_MIB_INERRORS, is_udplite); | |
1277 | } | |
db8dac20 | 1278 | goto out_free; |
22911fc5 | 1279 | } |
db8dac20 DM |
1280 | |
1281 | if (!peeked) | |
629ca23c PE |
1282 | UDP_INC_STATS_USER(sock_net(sk), |
1283 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 | 1284 | |
3b885787 | 1285 | sock_recv_ts_and_drops(msg, sk, skb); |
db8dac20 DM |
1286 | |
1287 | /* Copy the address. */ | |
c482c568 | 1288 | if (sin) { |
db8dac20 DM |
1289 | sin->sin_family = AF_INET; |
1290 | sin->sin_port = udp_hdr(skb)->source; | |
1291 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
1292 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
1293 | } | |
1294 | if (inet->cmsg_flags) | |
1295 | ip_cmsg_recv(msg, skb); | |
1296 | ||
59c2cdae | 1297 | err = copied; |
db8dac20 DM |
1298 | if (flags & MSG_TRUNC) |
1299 | err = ulen; | |
1300 | ||
1301 | out_free: | |
9d410c79 | 1302 | skb_free_datagram_locked(sk, skb); |
db8dac20 DM |
1303 | out: |
1304 | return err; | |
1305 | ||
1306 | csum_copy_err: | |
8a74ad60 | 1307 | slow = lock_sock_fast(sk); |
6a5dc9e5 ED |
1308 | if (!skb_kill_datagram(sk, skb, flags)) { |
1309 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); | |
629ca23c | 1310 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
6a5dc9e5 | 1311 | } |
8a74ad60 | 1312 | unlock_sock_fast(sk, slow); |
db8dac20 DM |
1313 | |
1314 | if (noblock) | |
1315 | return -EAGAIN; | |
9cfaa8de XZ |
1316 | |
1317 | /* starting over for a new packet */ | |
1318 | msg->msg_flags &= ~MSG_TRUNC; | |
db8dac20 DM |
1319 | goto try_again; |
1320 | } | |
1321 | ||
1322 | ||
1da177e4 LT |
1323 | int udp_disconnect(struct sock *sk, int flags) |
1324 | { | |
1325 | struct inet_sock *inet = inet_sk(sk); | |
1326 | /* | |
1327 | * 1003.1g - break association. | |
1328 | */ | |
e905a9ed | 1329 | |
1da177e4 | 1330 | sk->sk_state = TCP_CLOSE; |
c720c7e8 ED |
1331 | inet->inet_daddr = 0; |
1332 | inet->inet_dport = 0; | |
bdeab991 | 1333 | sock_rps_reset_rxhash(sk); |
1da177e4 LT |
1334 | sk->sk_bound_dev_if = 0; |
1335 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
1336 | inet_reset_saddr(sk); | |
1337 | ||
1338 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
1339 | sk->sk_prot->unhash(sk); | |
c720c7e8 | 1340 | inet->inet_sport = 0; |
1da177e4 LT |
1341 | } |
1342 | sk_dst_reset(sk); | |
1343 | return 0; | |
1344 | } | |
c482c568 | 1345 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1346 | |
645ca708 ED |
1347 | void udp_lib_unhash(struct sock *sk) |
1348 | { | |
723b4610 ED |
1349 | if (sk_hashed(sk)) { |
1350 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
512615b6 ED |
1351 | struct udp_hslot *hslot, *hslot2; |
1352 | ||
1353 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1354 | udp_sk(sk)->udp_port_hash); | |
1355 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
645ca708 | 1356 | |
723b4610 ED |
1357 | spin_lock_bh(&hslot->lock); |
1358 | if (sk_nulls_del_node_init_rcu(sk)) { | |
fdcc8aa9 | 1359 | hslot->count--; |
c720c7e8 | 1360 | inet_sk(sk)->inet_num = 0; |
723b4610 | 1361 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
512615b6 ED |
1362 | |
1363 | spin_lock(&hslot2->lock); | |
1364 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1365 | hslot2->count--; | |
1366 | spin_unlock(&hslot2->lock); | |
723b4610 ED |
1367 | } |
1368 | spin_unlock_bh(&hslot->lock); | |
645ca708 | 1369 | } |
645ca708 ED |
1370 | } |
1371 | EXPORT_SYMBOL(udp_lib_unhash); | |
1372 | ||
719f8358 ED |
1373 | /* |
1374 | * inet_rcv_saddr was changed, we must rehash secondary hash | |
1375 | */ | |
1376 | void udp_lib_rehash(struct sock *sk, u16 newhash) | |
1377 | { | |
1378 | if (sk_hashed(sk)) { | |
1379 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
1380 | struct udp_hslot *hslot, *hslot2, *nhslot2; | |
1381 | ||
1382 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
1383 | nhslot2 = udp_hashslot2(udptable, newhash); | |
1384 | udp_sk(sk)->udp_portaddr_hash = newhash; | |
1385 | if (hslot2 != nhslot2) { | |
1386 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1387 | udp_sk(sk)->udp_port_hash); | |
1388 | /* we must lock primary chain too */ | |
1389 | spin_lock_bh(&hslot->lock); | |
1390 | ||
1391 | spin_lock(&hslot2->lock); | |
1392 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1393 | hslot2->count--; | |
1394 | spin_unlock(&hslot2->lock); | |
1395 | ||
1396 | spin_lock(&nhslot2->lock); | |
1397 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
1398 | &nhslot2->head); | |
1399 | nhslot2->count++; | |
1400 | spin_unlock(&nhslot2->lock); | |
1401 | ||
1402 | spin_unlock_bh(&hslot->lock); | |
1403 | } | |
1404 | } | |
1405 | } | |
1406 | EXPORT_SYMBOL(udp_lib_rehash); | |
1407 | ||
1408 | static void udp_v4_rehash(struct sock *sk) | |
1409 | { | |
1410 | u16 new_hash = udp4_portaddr_hash(sock_net(sk), | |
1411 | inet_sk(sk)->inet_rcv_saddr, | |
1412 | inet_sk(sk)->inet_num); | |
1413 | udp_lib_rehash(sk, new_hash); | |
1414 | } | |
1415 | ||
93821778 HX |
1416 | static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
1417 | { | |
fec5e652 | 1418 | int rc; |
766e9037 | 1419 | |
005ec974 | 1420 | if (inet_sk(sk)->inet_daddr) { |
bdeab991 | 1421 | sock_rps_save_rxhash(sk, skb); |
005ec974 SB |
1422 | sk_mark_napi_id(sk, skb); |
1423 | } | |
fec5e652 | 1424 | |
d826eb14 | 1425 | rc = sock_queue_rcv_skb(sk, skb); |
766e9037 ED |
1426 | if (rc < 0) { |
1427 | int is_udplite = IS_UDPLITE(sk); | |
93821778 | 1428 | |
93821778 | 1429 | /* Note that an ENOMEM error is charged twice */ |
766e9037 | 1430 | if (rc == -ENOMEM) |
93821778 HX |
1431 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1432 | is_udplite); | |
766e9037 ED |
1433 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
1434 | kfree_skb(skb); | |
296f7ea7 | 1435 | trace_udp_fail_queue_rcv_skb(rc, sk); |
766e9037 | 1436 | return -1; |
93821778 HX |
1437 | } |
1438 | ||
1439 | return 0; | |
1440 | ||
93821778 HX |
1441 | } |
1442 | ||
447167bf ED |
1443 | static struct static_key udp_encap_needed __read_mostly; |
1444 | void udp_encap_enable(void) | |
1445 | { | |
1446 | if (!static_key_enabled(&udp_encap_needed)) | |
1447 | static_key_slow_inc(&udp_encap_needed); | |
1448 | } | |
1449 | EXPORT_SYMBOL(udp_encap_enable); | |
1450 | ||
db8dac20 DM |
1451 | /* returns: |
1452 | * -1: error | |
1453 | * 0: success | |
1454 | * >0: "udp encap" protocol resubmission | |
1455 | * | |
1456 | * Note that in the success and error cases, the skb is assumed to | |
1457 | * have either been requeued or freed. | |
1458 | */ | |
c482c568 | 1459 | int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
db8dac20 DM |
1460 | { |
1461 | struct udp_sock *up = udp_sk(sk); | |
1462 | int rc; | |
1463 | int is_udplite = IS_UDPLITE(sk); | |
1464 | ||
1465 | /* | |
1466 | * Charge it to the socket, dropping if the queue is full. | |
1467 | */ | |
1468 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1469 | goto drop; | |
1470 | nf_reset(skb); | |
1471 | ||
447167bf | 1472 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
0ad92ad0 ED |
1473 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
1474 | ||
db8dac20 DM |
1475 | /* |
1476 | * This is an encapsulation socket so pass the skb to | |
1477 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1478 | * fall through and pass this up the UDP socket. | |
1479 | * up->encap_rcv() returns the following value: | |
1480 | * =0 if skb was successfully passed to the encap | |
1481 | * handler or was discarded by it. | |
1482 | * >0 if skb should be passed on to UDP. | |
1483 | * <0 if skb should be resubmitted as proto -N | |
1484 | */ | |
1485 | ||
1486 | /* if we're overly short, let UDP handle it */ | |
0ad92ad0 ED |
1487 | encap_rcv = ACCESS_ONCE(up->encap_rcv); |
1488 | if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) { | |
db8dac20 DM |
1489 | int ret; |
1490 | ||
0ad92ad0 | 1491 | ret = encap_rcv(sk, skb); |
db8dac20 | 1492 | if (ret <= 0) { |
0283328e PE |
1493 | UDP_INC_STATS_BH(sock_net(sk), |
1494 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
1495 | is_udplite); |
1496 | return -ret; | |
1497 | } | |
1498 | } | |
1499 | ||
1500 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1501 | } | |
1502 | ||
1503 | /* | |
1504 | * UDP-Lite specific tests, ignored on UDP sockets | |
1505 | */ | |
1506 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1507 | ||
1508 | /* | |
1509 | * MIB statistics other than incrementing the error count are | |
1510 | * disabled for the following two types of errors: these depend | |
1511 | * on the application settings, not on the functioning of the | |
1512 | * protocol stack as such. | |
1513 | * | |
1514 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1515 | * way ... to ... at least let the receiving application block | |
1516 | * delivery of packets with coverage values less than a value | |
1517 | * provided by the application." | |
1518 | */ | |
1519 | if (up->pcrlen == 0) { /* full coverage was set */ | |
afd46503 JP |
1520 | LIMIT_NETDEBUG(KERN_WARNING "UDPLite: partial coverage %d while full coverage %d requested\n", |
1521 | UDP_SKB_CB(skb)->cscov, skb->len); | |
db8dac20 DM |
1522 | goto drop; |
1523 | } | |
1524 | /* The next case involves violating the min. coverage requested | |
1525 | * by the receiver. This is subtle: if receiver wants x and x is | |
1526 | * greater than the buffersize/MTU then receiver will complain | |
1527 | * that it wants x while sender emits packets of smaller size y. | |
1528 | * Therefore the above ...()->partial_cov statement is essential. | |
1529 | */ | |
1530 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
afd46503 JP |
1531 | LIMIT_NETDEBUG(KERN_WARNING "UDPLite: coverage %d too small, need min %d\n", |
1532 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
db8dac20 DM |
1533 | goto drop; |
1534 | } | |
1535 | } | |
1536 | ||
33d480ce ED |
1537 | if (rcu_access_pointer(sk->sk_filter) && |
1538 | udp_lib_checksum_complete(skb)) | |
6a5dc9e5 | 1539 | goto csum_error; |
db8dac20 | 1540 | |
c377411f | 1541 | |
f545a38f | 1542 | if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf)) |
c377411f ED |
1543 | goto drop; |
1544 | ||
93821778 | 1545 | rc = 0; |
db8dac20 | 1546 | |
fbf8866d | 1547 | ipv4_pktinfo_prepare(sk, skb); |
93821778 HX |
1548 | bh_lock_sock(sk); |
1549 | if (!sock_owned_by_user(sk)) | |
1550 | rc = __udp_queue_rcv_skb(sk, skb); | |
f545a38f | 1551 | else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) { |
55349790 ZY |
1552 | bh_unlock_sock(sk); |
1553 | goto drop; | |
1554 | } | |
93821778 HX |
1555 | bh_unlock_sock(sk); |
1556 | ||
1557 | return rc; | |
db8dac20 | 1558 | |
6a5dc9e5 ED |
1559 | csum_error: |
1560 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); | |
db8dac20 | 1561 | drop: |
0283328e | 1562 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8edf19c2 | 1563 | atomic_inc(&sk->sk_drops); |
db8dac20 DM |
1564 | kfree_skb(skb); |
1565 | return -1; | |
1566 | } | |
1567 | ||
1240d137 ED |
1568 | |
1569 | static void flush_stack(struct sock **stack, unsigned int count, | |
1570 | struct sk_buff *skb, unsigned int final) | |
1571 | { | |
1572 | unsigned int i; | |
1573 | struct sk_buff *skb1 = NULL; | |
f6b8f32c | 1574 | struct sock *sk; |
1240d137 ED |
1575 | |
1576 | for (i = 0; i < count; i++) { | |
f6b8f32c | 1577 | sk = stack[i]; |
1240d137 ED |
1578 | if (likely(skb1 == NULL)) |
1579 | skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC); | |
1580 | ||
f6b8f32c ED |
1581 | if (!skb1) { |
1582 | atomic_inc(&sk->sk_drops); | |
1583 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, | |
1584 | IS_UDPLITE(sk)); | |
1585 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, | |
1586 | IS_UDPLITE(sk)); | |
1587 | } | |
1588 | ||
1589 | if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0) | |
1240d137 ED |
1590 | skb1 = NULL; |
1591 | } | |
1592 | if (unlikely(skb1)) | |
1593 | kfree_skb(skb1); | |
1594 | } | |
1595 | ||
421b3885 SB |
1596 | static void udp_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb) |
1597 | { | |
1598 | struct dst_entry *dst = skb_dst(skb); | |
1599 | ||
1600 | dst_hold(dst); | |
1601 | sk->sk_rx_dst = dst; | |
1602 | } | |
1603 | ||
db8dac20 DM |
1604 | /* |
1605 | * Multicasts and broadcasts go to each listener. | |
1606 | * | |
1240d137 | 1607 | * Note: called only from the BH handler context. |
db8dac20 | 1608 | */ |
e3163493 | 1609 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1610 | struct udphdr *uh, |
1611 | __be32 saddr, __be32 daddr, | |
645ca708 | 1612 | struct udp_table *udptable) |
db8dac20 | 1613 | { |
1240d137 | 1614 | struct sock *sk, *stack[256 / sizeof(struct sock *)]; |
f86dcc5a | 1615 | struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest)); |
db8dac20 | 1616 | int dif; |
1240d137 | 1617 | unsigned int i, count = 0; |
db8dac20 | 1618 | |
645ca708 | 1619 | spin_lock(&hslot->lock); |
88ab1932 | 1620 | sk = sk_nulls_head(&hslot->head); |
db8dac20 | 1621 | dif = skb->dev->ifindex; |
920a4611 | 1622 | sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif); |
1240d137 ED |
1623 | while (sk) { |
1624 | stack[count++] = sk; | |
1625 | sk = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest, | |
1626 | daddr, uh->source, saddr, dif); | |
1627 | if (unlikely(count == ARRAY_SIZE(stack))) { | |
1628 | if (!sk) | |
1629 | break; | |
1630 | flush_stack(stack, count, skb, ~0); | |
1631 | count = 0; | |
1632 | } | |
1633 | } | |
1634 | /* | |
1635 | * before releasing chain lock, we must take a reference on sockets | |
1636 | */ | |
1637 | for (i = 0; i < count; i++) | |
1638 | sock_hold(stack[i]); | |
1639 | ||
645ca708 | 1640 | spin_unlock(&hslot->lock); |
1240d137 ED |
1641 | |
1642 | /* | |
1643 | * do the slow work with no lock held | |
1644 | */ | |
1645 | if (count) { | |
1646 | flush_stack(stack, count, skb, count - 1); | |
1647 | ||
1648 | for (i = 0; i < count; i++) | |
1649 | sock_put(stack[i]); | |
1650 | } else { | |
1651 | kfree_skb(skb); | |
1652 | } | |
db8dac20 DM |
1653 | return 0; |
1654 | } | |
1655 | ||
1656 | /* Initialize UDP checksum. If exited with zero value (success), | |
1657 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1658 | * Otherwise, csum completion requires chacksumming packet body, | |
1659 | * including udp header and folding it to skb->csum. | |
1660 | */ | |
1661 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1662 | int proto) | |
1663 | { | |
1664 | const struct iphdr *iph; | |
1665 | int err; | |
1666 | ||
1667 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1668 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1669 | ||
1670 | if (proto == IPPROTO_UDPLITE) { | |
1671 | err = udplite_checksum_init(skb, uh); | |
1672 | if (err) | |
1673 | return err; | |
1674 | } | |
1675 | ||
1676 | iph = ip_hdr(skb); | |
1677 | if (uh->check == 0) { | |
1678 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1679 | } else if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
c482c568 | 1680 | if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, |
db8dac20 DM |
1681 | proto, skb->csum)) |
1682 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1683 | } | |
1684 | if (!skb_csum_unnecessary(skb)) | |
1685 | skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, | |
1686 | skb->len, proto, 0); | |
1687 | /* Probably, we should checksum udp header (it should be in cache | |
1688 | * in any case) and data in tiny packets (< rx copybreak). | |
1689 | */ | |
1690 | ||
1691 | return 0; | |
1692 | } | |
1693 | ||
1694 | /* | |
1695 | * All we need to do is get the socket, and then do a checksum. | |
1696 | */ | |
1697 | ||
645ca708 | 1698 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1699 | int proto) |
1700 | { | |
1701 | struct sock *sk; | |
7b5e56f9 | 1702 | struct udphdr *uh; |
db8dac20 | 1703 | unsigned short ulen; |
adf30907 | 1704 | struct rtable *rt = skb_rtable(skb); |
2783ef23 | 1705 | __be32 saddr, daddr; |
0283328e | 1706 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1707 | |
1708 | /* | |
1709 | * Validate the packet. | |
1710 | */ | |
1711 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1712 | goto drop; /* No space for header. */ | |
1713 | ||
7b5e56f9 | 1714 | uh = udp_hdr(skb); |
db8dac20 | 1715 | ulen = ntohs(uh->len); |
ccc2d97c BM |
1716 | saddr = ip_hdr(skb)->saddr; |
1717 | daddr = ip_hdr(skb)->daddr; | |
1718 | ||
db8dac20 DM |
1719 | if (ulen > skb->len) |
1720 | goto short_packet; | |
1721 | ||
1722 | if (proto == IPPROTO_UDP) { | |
1723 | /* UDP validates ulen. */ | |
1724 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1725 | goto short_packet; | |
1726 | uh = udp_hdr(skb); | |
1727 | } | |
1728 | ||
1729 | if (udp4_csum_init(skb, uh, proto)) | |
1730 | goto csum_error; | |
1731 | ||
421b3885 SB |
1732 | if (skb->sk) { |
1733 | int ret; | |
1734 | sk = skb->sk; | |
1735 | ||
1736 | if (unlikely(sk->sk_rx_dst == NULL)) | |
1737 | udp_sk_rx_dst_set(sk, skb); | |
db8dac20 | 1738 | |
421b3885 SB |
1739 | ret = udp_queue_rcv_skb(sk, skb); |
1740 | ||
1741 | /* a return value > 0 means to resubmit the input, but | |
1742 | * it wants the return to be -protocol, or 0 | |
1743 | */ | |
1744 | if (ret > 0) | |
1745 | return -ret; | |
1746 | return 0; | |
1747 | } else { | |
1748 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) | |
1749 | return __udp4_lib_mcast_deliver(net, skb, uh, | |
1750 | saddr, daddr, udptable); | |
1751 | ||
1752 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); | |
1753 | } | |
db8dac20 DM |
1754 | |
1755 | if (sk != NULL) { | |
a5b50476 ET |
1756 | int ret; |
1757 | ||
a5b50476 | 1758 | ret = udp_queue_rcv_skb(sk, skb); |
db8dac20 DM |
1759 | sock_put(sk); |
1760 | ||
1761 | /* a return value > 0 means to resubmit the input, but | |
1762 | * it wants the return to be -protocol, or 0 | |
1763 | */ | |
1764 | if (ret > 0) | |
1765 | return -ret; | |
1766 | return 0; | |
1767 | } | |
1768 | ||
1769 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1770 | goto drop; | |
1771 | nf_reset(skb); | |
1772 | ||
1773 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1774 | if (udp_lib_checksum_complete(skb)) | |
1775 | goto csum_error; | |
1776 | ||
0283328e | 1777 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1778 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1779 | ||
1780 | /* | |
1781 | * Hmm. We got an UDP packet to a port to which we | |
1782 | * don't wanna listen. Ignore it. | |
1783 | */ | |
1784 | kfree_skb(skb); | |
1785 | return 0; | |
1786 | ||
1787 | short_packet: | |
673d57e7 | 1788 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n", |
afd46503 JP |
1789 | proto == IPPROTO_UDPLITE ? "Lite" : "", |
1790 | &saddr, ntohs(uh->source), | |
1791 | ulen, skb->len, | |
1792 | &daddr, ntohs(uh->dest)); | |
db8dac20 DM |
1793 | goto drop; |
1794 | ||
1795 | csum_error: | |
1796 | /* | |
1797 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1798 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1799 | */ | |
673d57e7 | 1800 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n", |
afd46503 JP |
1801 | proto == IPPROTO_UDPLITE ? "Lite" : "", |
1802 | &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest), | |
db8dac20 | 1803 | ulen); |
6a5dc9e5 | 1804 | UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 | 1805 | drop: |
0283328e | 1806 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1807 | kfree_skb(skb); |
1808 | return 0; | |
1809 | } | |
1810 | ||
421b3885 SB |
1811 | /* We can only early demux multicast if there is a single matching socket. |
1812 | * If more than one socket found returns NULL | |
1813 | */ | |
1814 | static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net, | |
1815 | __be16 loc_port, __be32 loc_addr, | |
1816 | __be16 rmt_port, __be32 rmt_addr, | |
1817 | int dif) | |
1818 | { | |
1819 | struct sock *sk, *result; | |
1820 | struct hlist_nulls_node *node; | |
1821 | unsigned short hnum = ntohs(loc_port); | |
1822 | unsigned int count, slot = udp_hashfn(net, hnum, udp_table.mask); | |
1823 | struct udp_hslot *hslot = &udp_table.hash[slot]; | |
1824 | ||
1825 | rcu_read_lock(); | |
1826 | begin: | |
1827 | count = 0; | |
1828 | result = NULL; | |
1829 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { | |
1830 | if (__udp_is_mcast_sock(net, sk, | |
1831 | loc_port, loc_addr, | |
1832 | rmt_port, rmt_addr, | |
1833 | dif, hnum)) { | |
1834 | result = sk; | |
1835 | ++count; | |
1836 | } | |
1837 | } | |
1838 | /* | |
1839 | * if the nulls value we got at the end of this lookup is | |
1840 | * not the expected one, we must restart lookup. | |
1841 | * We probably met an item that was moved to another chain. | |
1842 | */ | |
1843 | if (get_nulls_value(node) != slot) | |
1844 | goto begin; | |
1845 | ||
1846 | if (result) { | |
1847 | if (count != 1 || | |
1848 | unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) | |
1849 | result = NULL; | |
f69b923a | 1850 | else if (unlikely(!__udp_is_mcast_sock(net, result, |
421b3885 SB |
1851 | loc_port, loc_addr, |
1852 | rmt_port, rmt_addr, | |
1853 | dif, hnum))) { | |
1854 | sock_put(result); | |
1855 | result = NULL; | |
1856 | } | |
1857 | } | |
1858 | rcu_read_unlock(); | |
1859 | return result; | |
1860 | } | |
1861 | ||
1862 | /* For unicast we should only early demux connected sockets or we can | |
1863 | * break forwarding setups. The chains here can be long so only check | |
1864 | * if the first socket is an exact match and if not move on. | |
1865 | */ | |
1866 | static struct sock *__udp4_lib_demux_lookup(struct net *net, | |
1867 | __be16 loc_port, __be32 loc_addr, | |
1868 | __be16 rmt_port, __be32 rmt_addr, | |
1869 | int dif) | |
1870 | { | |
1871 | struct sock *sk, *result; | |
1872 | struct hlist_nulls_node *node; | |
1873 | unsigned short hnum = ntohs(loc_port); | |
1874 | unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum); | |
1875 | unsigned int slot2 = hash2 & udp_table.mask; | |
1876 | struct udp_hslot *hslot2 = &udp_table.hash2[slot2]; | |
1877 | INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr) | |
1878 | const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum); | |
1879 | ||
1880 | rcu_read_lock(); | |
1881 | result = NULL; | |
1882 | udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { | |
1883 | if (INET_MATCH(sk, net, acookie, | |
1884 | rmt_addr, loc_addr, ports, dif)) | |
1885 | result = sk; | |
1886 | /* Only check first socket in chain */ | |
1887 | break; | |
1888 | } | |
1889 | ||
1890 | if (result) { | |
1891 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) | |
1892 | result = NULL; | |
1893 | else if (unlikely(!INET_MATCH(sk, net, acookie, | |
1894 | rmt_addr, loc_addr, | |
1895 | ports, dif))) { | |
1896 | sock_put(result); | |
1897 | result = NULL; | |
1898 | } | |
1899 | } | |
1900 | rcu_read_unlock(); | |
1901 | return result; | |
1902 | } | |
1903 | ||
1904 | void udp_v4_early_demux(struct sk_buff *skb) | |
1905 | { | |
1906 | const struct iphdr *iph = ip_hdr(skb); | |
1907 | const struct udphdr *uh = udp_hdr(skb); | |
1908 | struct sock *sk; | |
1909 | struct dst_entry *dst; | |
1910 | struct net *net = dev_net(skb->dev); | |
1911 | int dif = skb->dev->ifindex; | |
1912 | ||
1913 | /* validate the packet */ | |
1914 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr))) | |
1915 | return; | |
1916 | ||
1917 | if (skb->pkt_type == PACKET_BROADCAST || | |
1918 | skb->pkt_type == PACKET_MULTICAST) | |
1919 | sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr, | |
1920 | uh->source, iph->saddr, dif); | |
1921 | else if (skb->pkt_type == PACKET_HOST) | |
1922 | sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr, | |
1923 | uh->source, iph->saddr, dif); | |
1924 | else | |
1925 | return; | |
1926 | ||
1927 | if (!sk) | |
1928 | return; | |
1929 | ||
1930 | skb->sk = sk; | |
1931 | skb->destructor = sock_edemux; | |
1932 | dst = sk->sk_rx_dst; | |
1933 | ||
1934 | if (dst) | |
1935 | dst = dst_check(dst, 0); | |
1936 | if (dst) | |
1937 | skb_dst_set_noref(skb, dst); | |
1938 | } | |
1939 | ||
db8dac20 DM |
1940 | int udp_rcv(struct sk_buff *skb) |
1941 | { | |
645ca708 | 1942 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
1943 | } |
1944 | ||
7d06b2e0 | 1945 | void udp_destroy_sock(struct sock *sk) |
db8dac20 | 1946 | { |
44046a59 | 1947 | struct udp_sock *up = udp_sk(sk); |
8a74ad60 | 1948 | bool slow = lock_sock_fast(sk); |
db8dac20 | 1949 | udp_flush_pending_frames(sk); |
8a74ad60 | 1950 | unlock_sock_fast(sk, slow); |
44046a59 TP |
1951 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
1952 | void (*encap_destroy)(struct sock *sk); | |
1953 | encap_destroy = ACCESS_ONCE(up->encap_destroy); | |
1954 | if (encap_destroy) | |
1955 | encap_destroy(sk); | |
1956 | } | |
db8dac20 DM |
1957 | } |
1958 | ||
1da177e4 LT |
1959 | /* |
1960 | * Socket option code for UDP | |
1961 | */ | |
4c0a6cb0 | 1962 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 1963 | char __user *optval, unsigned int optlen, |
4c0a6cb0 | 1964 | int (*push_pending_frames)(struct sock *)) |
1da177e4 LT |
1965 | { |
1966 | struct udp_sock *up = udp_sk(sk); | |
1967 | int val; | |
1968 | int err = 0; | |
b2bf1e26 | 1969 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 1970 | |
c482c568 | 1971 | if (optlen < sizeof(int)) |
1da177e4 LT |
1972 | return -EINVAL; |
1973 | ||
1974 | if (get_user(val, (int __user *)optval)) | |
1975 | return -EFAULT; | |
1976 | ||
6516c655 | 1977 | switch (optname) { |
1da177e4 LT |
1978 | case UDP_CORK: |
1979 | if (val != 0) { | |
1980 | up->corkflag = 1; | |
1981 | } else { | |
1982 | up->corkflag = 0; | |
1983 | lock_sock(sk); | |
4c0a6cb0 | 1984 | (*push_pending_frames)(sk); |
1da177e4 LT |
1985 | release_sock(sk); |
1986 | } | |
1987 | break; | |
e905a9ed | 1988 | |
1da177e4 LT |
1989 | case UDP_ENCAP: |
1990 | switch (val) { | |
1991 | case 0: | |
1992 | case UDP_ENCAP_ESPINUDP: | |
1993 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
1994 | up->encap_rcv = xfrm4_udp_encap_rcv; |
1995 | /* FALLTHROUGH */ | |
342f0234 | 1996 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 | 1997 | up->encap_type = val; |
447167bf | 1998 | udp_encap_enable(); |
1da177e4 LT |
1999 | break; |
2000 | default: | |
2001 | err = -ENOPROTOOPT; | |
2002 | break; | |
2003 | } | |
2004 | break; | |
2005 | ||
ba4e58ec GR |
2006 | /* |
2007 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
2008 | */ | |
2009 | /* The sender sets actual checksum coverage length via this option. | |
2010 | * The case coverage > packet length is handled by send module. */ | |
2011 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 2012 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
2013 | return -ENOPROTOOPT; |
2014 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
2015 | val = 8; | |
4be929be AD |
2016 | else if (val > USHRT_MAX) |
2017 | val = USHRT_MAX; | |
ba4e58ec GR |
2018 | up->pcslen = val; |
2019 | up->pcflag |= UDPLITE_SEND_CC; | |
2020 | break; | |
2021 | ||
e905a9ed YH |
2022 | /* The receiver specifies a minimum checksum coverage value. To make |
2023 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
2024 | * used, this again means full checksum coverage. */ |
2025 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 2026 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
2027 | return -ENOPROTOOPT; |
2028 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
2029 | val = 8; | |
4be929be AD |
2030 | else if (val > USHRT_MAX) |
2031 | val = USHRT_MAX; | |
ba4e58ec GR |
2032 | up->pcrlen = val; |
2033 | up->pcflag |= UDPLITE_RECV_CC; | |
2034 | break; | |
2035 | ||
1da177e4 LT |
2036 | default: |
2037 | err = -ENOPROTOOPT; | |
2038 | break; | |
6516c655 | 2039 | } |
1da177e4 LT |
2040 | |
2041 | return err; | |
2042 | } | |
c482c568 | 2043 | EXPORT_SYMBOL(udp_lib_setsockopt); |
1da177e4 | 2044 | |
db8dac20 | 2045 | int udp_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 2046 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
2047 | { |
2048 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2049 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
2050 | udp_push_pending_frames); | |
2051 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
2052 | } | |
2053 | ||
2054 | #ifdef CONFIG_COMPAT | |
2055 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 2056 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
2057 | { |
2058 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2059 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
2060 | udp_push_pending_frames); | |
2061 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
2062 | } | |
2063 | #endif | |
2064 | ||
4c0a6cb0 GR |
2065 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
2066 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
2067 | { |
2068 | struct udp_sock *up = udp_sk(sk); | |
2069 | int val, len; | |
2070 | ||
c482c568 | 2071 | if (get_user(len, optlen)) |
1da177e4 LT |
2072 | return -EFAULT; |
2073 | ||
2074 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 2075 | |
6516c655 | 2076 | if (len < 0) |
1da177e4 LT |
2077 | return -EINVAL; |
2078 | ||
6516c655 | 2079 | switch (optname) { |
1da177e4 LT |
2080 | case UDP_CORK: |
2081 | val = up->corkflag; | |
2082 | break; | |
2083 | ||
2084 | case UDP_ENCAP: | |
2085 | val = up->encap_type; | |
2086 | break; | |
2087 | ||
ba4e58ec GR |
2088 | /* The following two cannot be changed on UDP sockets, the return is |
2089 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
2090 | case UDPLITE_SEND_CSCOV: | |
2091 | val = up->pcslen; | |
2092 | break; | |
2093 | ||
2094 | case UDPLITE_RECV_CSCOV: | |
2095 | val = up->pcrlen; | |
2096 | break; | |
2097 | ||
1da177e4 LT |
2098 | default: |
2099 | return -ENOPROTOOPT; | |
6516c655 | 2100 | } |
1da177e4 | 2101 | |
6516c655 | 2102 | if (put_user(len, optlen)) |
e905a9ed | 2103 | return -EFAULT; |
c482c568 | 2104 | if (copy_to_user(optval, &val, len)) |
1da177e4 | 2105 | return -EFAULT; |
e905a9ed | 2106 | return 0; |
1da177e4 | 2107 | } |
c482c568 | 2108 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1da177e4 | 2109 | |
db8dac20 DM |
2110 | int udp_getsockopt(struct sock *sk, int level, int optname, |
2111 | char __user *optval, int __user *optlen) | |
2112 | { | |
2113 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2114 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
2115 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
2116 | } | |
2117 | ||
2118 | #ifdef CONFIG_COMPAT | |
2119 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
2120 | char __user *optval, int __user *optlen) | |
2121 | { | |
2122 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
2123 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
2124 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
2125 | } | |
2126 | #endif | |
1da177e4 LT |
2127 | /** |
2128 | * udp_poll - wait for a UDP event. | |
2129 | * @file - file struct | |
2130 | * @sock - socket | |
2131 | * @wait - poll table | |
2132 | * | |
e905a9ed | 2133 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
2134 | * blocking sockets. If application is using a blocking fd |
2135 | * and a packet with checksum error is in the queue; | |
2136 | * then it could get return from select indicating data available | |
2137 | * but then block when reading it. Add special case code | |
2138 | * to work around these arguably broken applications. | |
2139 | */ | |
2140 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
2141 | { | |
2142 | unsigned int mask = datagram_poll(file, sock, wait); | |
2143 | struct sock *sk = sock->sk; | |
ba4e58ec | 2144 | |
c3f1dbaf DM |
2145 | sock_rps_record_flow(sk); |
2146 | ||
1da177e4 | 2147 | /* Check for false positives due to checksum errors */ |
85584672 ED |
2148 | if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) && |
2149 | !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk)) | |
2150 | mask &= ~(POLLIN | POLLRDNORM); | |
1da177e4 LT |
2151 | |
2152 | return mask; | |
e905a9ed | 2153 | |
1da177e4 | 2154 | } |
c482c568 | 2155 | EXPORT_SYMBOL(udp_poll); |
1da177e4 | 2156 | |
db8dac20 DM |
2157 | struct proto udp_prot = { |
2158 | .name = "UDP", | |
2159 | .owner = THIS_MODULE, | |
2160 | .close = udp_lib_close, | |
2161 | .connect = ip4_datagram_connect, | |
2162 | .disconnect = udp_disconnect, | |
2163 | .ioctl = udp_ioctl, | |
2164 | .destroy = udp_destroy_sock, | |
2165 | .setsockopt = udp_setsockopt, | |
2166 | .getsockopt = udp_getsockopt, | |
2167 | .sendmsg = udp_sendmsg, | |
2168 | .recvmsg = udp_recvmsg, | |
2169 | .sendpage = udp_sendpage, | |
93821778 | 2170 | .backlog_rcv = __udp_queue_rcv_skb, |
8141ed9f | 2171 | .release_cb = ip4_datagram_release_cb, |
db8dac20 DM |
2172 | .hash = udp_lib_hash, |
2173 | .unhash = udp_lib_unhash, | |
719f8358 | 2174 | .rehash = udp_v4_rehash, |
db8dac20 DM |
2175 | .get_port = udp_v4_get_port, |
2176 | .memory_allocated = &udp_memory_allocated, | |
2177 | .sysctl_mem = sysctl_udp_mem, | |
2178 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
2179 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
2180 | .obj_size = sizeof(struct udp_sock), | |
271b72c7 | 2181 | .slab_flags = SLAB_DESTROY_BY_RCU, |
645ca708 | 2182 | .h.udp_table = &udp_table, |
db8dac20 DM |
2183 | #ifdef CONFIG_COMPAT |
2184 | .compat_setsockopt = compat_udp_setsockopt, | |
2185 | .compat_getsockopt = compat_udp_getsockopt, | |
2186 | #endif | |
fcbdf09d | 2187 | .clear_sk = sk_prot_clear_portaddr_nulls, |
db8dac20 | 2188 | }; |
c482c568 | 2189 | EXPORT_SYMBOL(udp_prot); |
1da177e4 LT |
2190 | |
2191 | /* ------------------------------------------------------------------------ */ | |
2192 | #ifdef CONFIG_PROC_FS | |
2193 | ||
645ca708 | 2194 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
2195 | { |
2196 | struct sock *sk; | |
2197 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2198 | struct net *net = seq_file_net(seq); |
1da177e4 | 2199 | |
f86dcc5a ED |
2200 | for (state->bucket = start; state->bucket <= state->udp_table->mask; |
2201 | ++state->bucket) { | |
88ab1932 | 2202 | struct hlist_nulls_node *node; |
645ca708 | 2203 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
f86dcc5a ED |
2204 | |
2205 | if (hlist_nulls_empty(&hslot->head)) | |
2206 | continue; | |
2207 | ||
645ca708 | 2208 | spin_lock_bh(&hslot->lock); |
88ab1932 | 2209 | sk_nulls_for_each(sk, node, &hslot->head) { |
878628fb | 2210 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 2211 | continue; |
1da177e4 LT |
2212 | if (sk->sk_family == state->family) |
2213 | goto found; | |
2214 | } | |
645ca708 | 2215 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
2216 | } |
2217 | sk = NULL; | |
2218 | found: | |
2219 | return sk; | |
2220 | } | |
2221 | ||
2222 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
2223 | { | |
2224 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2225 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2226 | |
2227 | do { | |
88ab1932 | 2228 | sk = sk_nulls_next(sk); |
878628fb | 2229 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 2230 | |
645ca708 | 2231 | if (!sk) { |
f86dcc5a | 2232 | if (state->bucket <= state->udp_table->mask) |
30842f29 | 2233 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
645ca708 | 2234 | return udp_get_first(seq, state->bucket + 1); |
1da177e4 LT |
2235 | } |
2236 | return sk; | |
2237 | } | |
2238 | ||
2239 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
2240 | { | |
645ca708 | 2241 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
2242 | |
2243 | if (sk) | |
6516c655 | 2244 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
2245 | --pos; |
2246 | return pos ? NULL : sk; | |
2247 | } | |
2248 | ||
2249 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
2250 | { | |
30842f29 | 2251 | struct udp_iter_state *state = seq->private; |
f86dcc5a | 2252 | state->bucket = MAX_UDP_PORTS; |
30842f29 | 2253 | |
b50660f1 | 2254 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
2255 | } |
2256 | ||
2257 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2258 | { | |
2259 | struct sock *sk; | |
2260 | ||
b50660f1 | 2261 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2262 | sk = udp_get_idx(seq, 0); |
2263 | else | |
2264 | sk = udp_get_next(seq, v); | |
2265 | ||
2266 | ++*pos; | |
2267 | return sk; | |
2268 | } | |
2269 | ||
2270 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
2271 | { | |
645ca708 ED |
2272 | struct udp_iter_state *state = seq->private; |
2273 | ||
f86dcc5a | 2274 | if (state->bucket <= state->udp_table->mask) |
645ca708 | 2275 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
1da177e4 LT |
2276 | } |
2277 | ||
73cb88ec | 2278 | int udp_seq_open(struct inode *inode, struct file *file) |
1da177e4 | 2279 | { |
d9dda78b | 2280 | struct udp_seq_afinfo *afinfo = PDE_DATA(inode); |
a2be75c1 DL |
2281 | struct udp_iter_state *s; |
2282 | int err; | |
a91275ef | 2283 | |
a2be75c1 DL |
2284 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2285 | sizeof(struct udp_iter_state)); | |
2286 | if (err < 0) | |
2287 | return err; | |
a91275ef | 2288 | |
a2be75c1 | 2289 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2290 | s->family = afinfo->family; |
645ca708 | 2291 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 2292 | return err; |
a91275ef | 2293 | } |
73cb88ec | 2294 | EXPORT_SYMBOL(udp_seq_open); |
a91275ef | 2295 | |
1da177e4 | 2296 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 2297 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
2298 | { |
2299 | struct proc_dir_entry *p; | |
2300 | int rc = 0; | |
2301 | ||
dda61925 DL |
2302 | afinfo->seq_ops.start = udp_seq_start; |
2303 | afinfo->seq_ops.next = udp_seq_next; | |
2304 | afinfo->seq_ops.stop = udp_seq_stop; | |
2305 | ||
84841c3c | 2306 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2307 | afinfo->seq_fops, afinfo); |
84841c3c | 2308 | if (!p) |
1da177e4 LT |
2309 | rc = -ENOMEM; |
2310 | return rc; | |
2311 | } | |
c482c568 | 2312 | EXPORT_SYMBOL(udp_proc_register); |
1da177e4 | 2313 | |
0c96d8c5 | 2314 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 2315 | { |
ece31ffd | 2316 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2317 | } |
c482c568 | 2318 | EXPORT_SYMBOL(udp_proc_unregister); |
db8dac20 DM |
2319 | |
2320 | /* ------------------------------------------------------------------------ */ | |
5e659e4c PE |
2321 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
2322 | int bucket, int *len) | |
db8dac20 DM |
2323 | { |
2324 | struct inet_sock *inet = inet_sk(sp); | |
c720c7e8 ED |
2325 | __be32 dest = inet->inet_daddr; |
2326 | __be32 src = inet->inet_rcv_saddr; | |
2327 | __u16 destp = ntohs(inet->inet_dport); | |
2328 | __u16 srcp = ntohs(inet->inet_sport); | |
db8dac20 | 2329 | |
f86dcc5a | 2330 | seq_printf(f, "%5d: %08X:%04X %08X:%04X" |
d14c5ab6 | 2331 | " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d%n", |
db8dac20 | 2332 | bucket, src, srcp, dest, destp, sp->sk_state, |
31e6d363 ED |
2333 | sk_wmem_alloc_get(sp), |
2334 | sk_rmem_alloc_get(sp), | |
a7cb5a49 EB |
2335 | 0, 0L, 0, |
2336 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)), | |
2337 | 0, sock_i_ino(sp), | |
cb61cb9b ED |
2338 | atomic_read(&sp->sk_refcnt), sp, |
2339 | atomic_read(&sp->sk_drops), len); | |
db8dac20 DM |
2340 | } |
2341 | ||
2342 | int udp4_seq_show(struct seq_file *seq, void *v) | |
2343 | { | |
2344 | if (v == SEQ_START_TOKEN) | |
2345 | seq_printf(seq, "%-127s\n", | |
2346 | " sl local_address rem_address st tx_queue " | |
2347 | "rx_queue tr tm->when retrnsmt uid timeout " | |
cb61cb9b | 2348 | "inode ref pointer drops"); |
db8dac20 | 2349 | else { |
db8dac20 | 2350 | struct udp_iter_state *state = seq->private; |
5e659e4c | 2351 | int len; |
db8dac20 | 2352 | |
5e659e4c | 2353 | udp4_format_sock(v, seq, state->bucket, &len); |
c482c568 | 2354 | seq_printf(seq, "%*s\n", 127 - len, ""); |
db8dac20 DM |
2355 | } |
2356 | return 0; | |
2357 | } | |
2358 | ||
73cb88ec AV |
2359 | static const struct file_operations udp_afinfo_seq_fops = { |
2360 | .owner = THIS_MODULE, | |
2361 | .open = udp_seq_open, | |
2362 | .read = seq_read, | |
2363 | .llseek = seq_lseek, | |
2364 | .release = seq_release_net | |
2365 | }; | |
2366 | ||
db8dac20 | 2367 | /* ------------------------------------------------------------------------ */ |
db8dac20 | 2368 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
2369 | .name = "udp", |
2370 | .family = AF_INET, | |
645ca708 | 2371 | .udp_table = &udp_table, |
73cb88ec | 2372 | .seq_fops = &udp_afinfo_seq_fops, |
dda61925 DL |
2373 | .seq_ops = { |
2374 | .show = udp4_seq_show, | |
2375 | }, | |
db8dac20 DM |
2376 | }; |
2377 | ||
2c8c1e72 | 2378 | static int __net_init udp4_proc_init_net(struct net *net) |
15439feb PE |
2379 | { |
2380 | return udp_proc_register(net, &udp4_seq_afinfo); | |
2381 | } | |
2382 | ||
2c8c1e72 | 2383 | static void __net_exit udp4_proc_exit_net(struct net *net) |
15439feb PE |
2384 | { |
2385 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
2386 | } | |
2387 | ||
2388 | static struct pernet_operations udp4_net_ops = { | |
2389 | .init = udp4_proc_init_net, | |
2390 | .exit = udp4_proc_exit_net, | |
2391 | }; | |
2392 | ||
db8dac20 DM |
2393 | int __init udp4_proc_init(void) |
2394 | { | |
15439feb | 2395 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
2396 | } |
2397 | ||
2398 | void udp4_proc_exit(void) | |
2399 | { | |
15439feb | 2400 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 2401 | } |
1da177e4 LT |
2402 | #endif /* CONFIG_PROC_FS */ |
2403 | ||
f86dcc5a ED |
2404 | static __initdata unsigned long uhash_entries; |
2405 | static int __init set_uhash_entries(char *str) | |
645ca708 | 2406 | { |
413c27d8 EZ |
2407 | ssize_t ret; |
2408 | ||
f86dcc5a ED |
2409 | if (!str) |
2410 | return 0; | |
413c27d8 EZ |
2411 | |
2412 | ret = kstrtoul(str, 0, &uhash_entries); | |
2413 | if (ret) | |
2414 | return 0; | |
2415 | ||
f86dcc5a ED |
2416 | if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN) |
2417 | uhash_entries = UDP_HTABLE_SIZE_MIN; | |
2418 | return 1; | |
2419 | } | |
2420 | __setup("uhash_entries=", set_uhash_entries); | |
645ca708 | 2421 | |
f86dcc5a ED |
2422 | void __init udp_table_init(struct udp_table *table, const char *name) |
2423 | { | |
2424 | unsigned int i; | |
2425 | ||
31fe62b9 TB |
2426 | table->hash = alloc_large_system_hash(name, |
2427 | 2 * sizeof(struct udp_hslot), | |
2428 | uhash_entries, | |
2429 | 21, /* one slot per 2 MB */ | |
2430 | 0, | |
2431 | &table->log, | |
2432 | &table->mask, | |
2433 | UDP_HTABLE_SIZE_MIN, | |
2434 | 64 * 1024); | |
2435 | ||
512615b6 | 2436 | table->hash2 = table->hash + (table->mask + 1); |
f86dcc5a | 2437 | for (i = 0; i <= table->mask; i++) { |
88ab1932 | 2438 | INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i); |
fdcc8aa9 | 2439 | table->hash[i].count = 0; |
645ca708 ED |
2440 | spin_lock_init(&table->hash[i].lock); |
2441 | } | |
512615b6 ED |
2442 | for (i = 0; i <= table->mask; i++) { |
2443 | INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i); | |
2444 | table->hash2[i].count = 0; | |
2445 | spin_lock_init(&table->hash2[i].lock); | |
2446 | } | |
645ca708 ED |
2447 | } |
2448 | ||
95766fff HA |
2449 | void __init udp_init(void) |
2450 | { | |
f03d78db | 2451 | unsigned long limit; |
95766fff | 2452 | |
f86dcc5a | 2453 | udp_table_init(&udp_table, "UDP"); |
f03d78db | 2454 | limit = nr_free_buffer_pages() / 8; |
95766fff HA |
2455 | limit = max(limit, 128UL); |
2456 | sysctl_udp_mem[0] = limit / 4 * 3; | |
2457 | sysctl_udp_mem[1] = limit; | |
2458 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
2459 | ||
2460 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
2461 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
2462 | } | |
2463 | ||
da5bab07 DB |
2464 | struct sk_buff *skb_udp_tunnel_segment(struct sk_buff *skb, |
2465 | netdev_features_t features) | |
73136267 PS |
2466 | { |
2467 | struct sk_buff *segs = ERR_PTR(-EINVAL); | |
2468 | int mac_len = skb->mac_len; | |
2469 | int tnl_hlen = skb_inner_mac_header(skb) - skb_transport_header(skb); | |
0d05535d | 2470 | __be16 protocol = skb->protocol; |
73136267 | 2471 | netdev_features_t enc_features; |
0d05535d | 2472 | int outer_hlen; |
73136267 PS |
2473 | |
2474 | if (unlikely(!pskb_may_pull(skb, tnl_hlen))) | |
2475 | goto out; | |
2476 | ||
2477 | skb->encapsulation = 0; | |
2478 | __skb_pull(skb, tnl_hlen); | |
2479 | skb_reset_mac_header(skb); | |
2480 | skb_set_network_header(skb, skb_inner_network_offset(skb)); | |
2481 | skb->mac_len = skb_inner_network_offset(skb); | |
19acc327 | 2482 | skb->protocol = htons(ETH_P_TEB); |
73136267 PS |
2483 | |
2484 | /* segment inner packet. */ | |
2485 | enc_features = skb->dev->hw_enc_features & netif_skb_features(skb); | |
2486 | segs = skb_mac_gso_segment(skb, enc_features); | |
2487 | if (!segs || IS_ERR(segs)) | |
2488 | goto out; | |
2489 | ||
2490 | outer_hlen = skb_tnl_header_len(skb); | |
2491 | skb = segs; | |
2492 | do { | |
2493 | struct udphdr *uh; | |
2494 | int udp_offset = outer_hlen - tnl_hlen; | |
2495 | ||
cdbaa0bb AD |
2496 | skb_reset_inner_headers(skb); |
2497 | skb->encapsulation = 1; | |
2498 | ||
73136267 PS |
2499 | skb->mac_len = mac_len; |
2500 | ||
2501 | skb_push(skb, outer_hlen); | |
2502 | skb_reset_mac_header(skb); | |
2503 | skb_set_network_header(skb, mac_len); | |
2504 | skb_set_transport_header(skb, udp_offset); | |
2505 | uh = udp_hdr(skb); | |
2506 | uh->len = htons(skb->len - udp_offset); | |
2507 | ||
2508 | /* csum segment if tunnel sets skb with csum. */ | |
eb3c0d83 | 2509 | if (protocol == htons(ETH_P_IP) && unlikely(uh->check)) { |
73136267 PS |
2510 | struct iphdr *iph = ip_hdr(skb); |
2511 | ||
2512 | uh->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr, | |
2513 | skb->len - udp_offset, | |
2514 | IPPROTO_UDP, 0); | |
2515 | uh->check = csum_fold(skb_checksum(skb, udp_offset, | |
2516 | skb->len - udp_offset, 0)); | |
2517 | if (uh->check == 0) | |
2518 | uh->check = CSUM_MANGLED_0; | |
2519 | ||
eb3c0d83 CW |
2520 | } else if (protocol == htons(ETH_P_IPV6)) { |
2521 | struct ipv6hdr *ipv6h = ipv6_hdr(skb); | |
2522 | u32 len = skb->len - udp_offset; | |
2523 | ||
2524 | uh->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, | |
2525 | len, IPPROTO_UDP, 0); | |
2526 | uh->check = csum_fold(skb_checksum(skb, udp_offset, len, 0)); | |
2527 | if (uh->check == 0) | |
2528 | uh->check = CSUM_MANGLED_0; | |
2529 | skb->ip_summed = CHECKSUM_NONE; | |
73136267 | 2530 | } |
eb3c0d83 | 2531 | |
0d05535d | 2532 | skb->protocol = protocol; |
73136267 PS |
2533 | } while ((skb = skb->next)); |
2534 | out: | |
2535 | return segs; | |
2536 | } |