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 LT |
105 | #include <net/icmp.h> |
106 | #include <net/route.h> | |
1da177e4 LT |
107 | #include <net/checksum.h> |
108 | #include <net/xfrm.h> | |
296f7ea7 | 109 | #include <trace/events/udp.h> |
447167bf | 110 | #include <linux/static_key.h> |
22911fc5 | 111 | #include <trace/events/skb.h> |
ba4e58ec | 112 | #include "udp_impl.h" |
1da177e4 | 113 | |
f86dcc5a | 114 | struct udp_table udp_table __read_mostly; |
645ca708 | 115 | EXPORT_SYMBOL(udp_table); |
1da177e4 | 116 | |
8d987e5c | 117 | long sysctl_udp_mem[3] __read_mostly; |
95766fff | 118 | EXPORT_SYMBOL(sysctl_udp_mem); |
c482c568 ED |
119 | |
120 | int sysctl_udp_rmem_min __read_mostly; | |
95766fff | 121 | EXPORT_SYMBOL(sysctl_udp_rmem_min); |
c482c568 ED |
122 | |
123 | int sysctl_udp_wmem_min __read_mostly; | |
95766fff HA |
124 | EXPORT_SYMBOL(sysctl_udp_wmem_min); |
125 | ||
8d987e5c | 126 | atomic_long_t udp_memory_allocated; |
95766fff HA |
127 | EXPORT_SYMBOL(udp_memory_allocated); |
128 | ||
f86dcc5a ED |
129 | #define MAX_UDP_PORTS 65536 |
130 | #define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN) | |
98322f22 | 131 | |
f24d43c0 | 132 | static int udp_lib_lport_inuse(struct net *net, __u16 num, |
645ca708 | 133 | const struct udp_hslot *hslot, |
98322f22 | 134 | unsigned long *bitmap, |
f24d43c0 ED |
135 | struct sock *sk, |
136 | int (*saddr_comp)(const struct sock *sk1, | |
f86dcc5a ED |
137 | const struct sock *sk2), |
138 | unsigned int log) | |
1da177e4 | 139 | { |
f24d43c0 | 140 | struct sock *sk2; |
88ab1932 | 141 | struct hlist_nulls_node *node; |
ba418fa3 | 142 | kuid_t uid = sock_i_uid(sk); |
25030a7f | 143 | |
88ab1932 | 144 | sk_nulls_for_each(sk2, node, &hslot->head) |
9d4fb27d JP |
145 | if (net_eq(sock_net(sk2), net) && |
146 | sk2 != sk && | |
d4cada4a | 147 | (bitmap || udp_sk(sk2)->udp_port_hash == num) && |
9d4fb27d JP |
148 | (!sk2->sk_reuse || !sk->sk_reuse) && |
149 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
150 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 TH |
151 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
152 | !uid_eq(uid, sock_i_uid(sk2))) && | |
98322f22 ED |
153 | (*saddr_comp)(sk, sk2)) { |
154 | if (bitmap) | |
d4cada4a ED |
155 | __set_bit(udp_sk(sk2)->udp_port_hash >> log, |
156 | bitmap); | |
98322f22 ED |
157 | else |
158 | return 1; | |
159 | } | |
25030a7f GR |
160 | return 0; |
161 | } | |
162 | ||
30fff923 ED |
163 | /* |
164 | * Note: we still hold spinlock of primary hash chain, so no other writer | |
165 | * can insert/delete a socket with local_port == num | |
166 | */ | |
167 | static int udp_lib_lport_inuse2(struct net *net, __u16 num, | |
168 | struct udp_hslot *hslot2, | |
169 | struct sock *sk, | |
170 | int (*saddr_comp)(const struct sock *sk1, | |
171 | const struct sock *sk2)) | |
172 | { | |
173 | struct sock *sk2; | |
174 | struct hlist_nulls_node *node; | |
ba418fa3 | 175 | kuid_t uid = sock_i_uid(sk); |
30fff923 ED |
176 | int res = 0; |
177 | ||
178 | spin_lock(&hslot2->lock); | |
179 | udp_portaddr_for_each_entry(sk2, node, &hslot2->head) | |
9d4fb27d JP |
180 | if (net_eq(sock_net(sk2), net) && |
181 | sk2 != sk && | |
182 | (udp_sk(sk2)->udp_port_hash == num) && | |
183 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
184 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
185 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 TH |
186 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
187 | !uid_eq(uid, sock_i_uid(sk2))) && | |
30fff923 ED |
188 | (*saddr_comp)(sk, sk2)) { |
189 | res = 1; | |
190 | break; | |
191 | } | |
192 | spin_unlock(&hslot2->lock); | |
193 | return res; | |
194 | } | |
195 | ||
25030a7f | 196 | /** |
6ba5a3c5 | 197 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
198 | * |
199 | * @sk: socket struct in question | |
200 | * @snum: port number to look up | |
df2bc459 | 201 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25985edc | 202 | * @hash2_nulladdr: AF-dependent hash value in secondary hash chains, |
30fff923 | 203 | * with NULL address |
25030a7f | 204 | */ |
6ba5a3c5 | 205 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
df2bc459 | 206 | int (*saddr_comp)(const struct sock *sk1, |
30fff923 ED |
207 | const struct sock *sk2), |
208 | unsigned int hash2_nulladdr) | |
25030a7f | 209 | { |
512615b6 | 210 | struct udp_hslot *hslot, *hslot2; |
645ca708 | 211 | struct udp_table *udptable = sk->sk_prot->h.udp_table; |
25030a7f | 212 | int error = 1; |
3b1e0a65 | 213 | struct net *net = sock_net(sk); |
1da177e4 | 214 | |
32c1da70 | 215 | if (!snum) { |
9088c560 | 216 | int low, high, remaining; |
95c96174 | 217 | unsigned int rand; |
98322f22 ED |
218 | unsigned short first, last; |
219 | DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN); | |
32c1da70 | 220 | |
227b60f5 | 221 | inet_get_local_port_range(&low, &high); |
a25de534 | 222 | remaining = (high - low) + 1; |
227b60f5 | 223 | |
9088c560 | 224 | rand = net_random(); |
98322f22 ED |
225 | first = (((u64)rand * remaining) >> 32) + low; |
226 | /* | |
227 | * force rand to be an odd multiple of UDP_HTABLE_SIZE | |
228 | */ | |
f86dcc5a | 229 | rand = (rand | 1) * (udptable->mask + 1); |
5781b235 ED |
230 | last = first + udptable->mask + 1; |
231 | do { | |
f86dcc5a | 232 | hslot = udp_hashslot(udptable, net, first); |
98322f22 | 233 | bitmap_zero(bitmap, PORTS_PER_CHAIN); |
645ca708 | 234 | spin_lock_bh(&hslot->lock); |
98322f22 | 235 | udp_lib_lport_inuse(net, snum, hslot, bitmap, sk, |
f86dcc5a | 236 | saddr_comp, udptable->log); |
98322f22 ED |
237 | |
238 | snum = first; | |
239 | /* | |
240 | * Iterate on all possible values of snum for this hash. | |
241 | * Using steps of an odd multiple of UDP_HTABLE_SIZE | |
242 | * give us randomization and full range coverage. | |
243 | */ | |
9088c560 | 244 | do { |
98322f22 | 245 | if (low <= snum && snum <= high && |
e3826f1e AW |
246 | !test_bit(snum >> udptable->log, bitmap) && |
247 | !inet_is_reserved_local_port(snum)) | |
98322f22 ED |
248 | goto found; |
249 | snum += rand; | |
250 | } while (snum != first); | |
251 | spin_unlock_bh(&hslot->lock); | |
5781b235 | 252 | } while (++first != last); |
98322f22 | 253 | goto fail; |
645ca708 | 254 | } else { |
f86dcc5a | 255 | hslot = udp_hashslot(udptable, net, snum); |
645ca708 | 256 | spin_lock_bh(&hslot->lock); |
30fff923 ED |
257 | if (hslot->count > 10) { |
258 | int exist; | |
259 | unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum; | |
260 | ||
261 | slot2 &= udptable->mask; | |
262 | hash2_nulladdr &= udptable->mask; | |
263 | ||
264 | hslot2 = udp_hashslot2(udptable, slot2); | |
265 | if (hslot->count < hslot2->count) | |
266 | goto scan_primary_hash; | |
267 | ||
268 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
269 | sk, saddr_comp); | |
270 | if (!exist && (hash2_nulladdr != slot2)) { | |
271 | hslot2 = udp_hashslot2(udptable, hash2_nulladdr); | |
272 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
273 | sk, saddr_comp); | |
274 | } | |
275 | if (exist) | |
276 | goto fail_unlock; | |
277 | else | |
278 | goto found; | |
279 | } | |
280 | scan_primary_hash: | |
f86dcc5a ED |
281 | if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk, |
282 | saddr_comp, 0)) | |
645ca708 ED |
283 | goto fail_unlock; |
284 | } | |
98322f22 | 285 | found: |
c720c7e8 | 286 | inet_sk(sk)->inet_num = snum; |
d4cada4a ED |
287 | udp_sk(sk)->udp_port_hash = snum; |
288 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
1da177e4 | 289 | if (sk_unhashed(sk)) { |
88ab1932 | 290 | sk_nulls_add_node_rcu(sk, &hslot->head); |
fdcc8aa9 | 291 | hslot->count++; |
c29a0bc4 | 292 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
512615b6 ED |
293 | |
294 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
295 | spin_lock(&hslot2->lock); | |
296 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
297 | &hslot2->head); | |
298 | hslot2->count++; | |
299 | spin_unlock(&hslot2->lock); | |
1da177e4 | 300 | } |
25030a7f | 301 | error = 0; |
645ca708 ED |
302 | fail_unlock: |
303 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 304 | fail: |
25030a7f GR |
305 | return error; |
306 | } | |
c482c568 | 307 | EXPORT_SYMBOL(udp_lib_get_port); |
25030a7f | 308 | |
499923c7 | 309 | static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2) |
db8dac20 DM |
310 | { |
311 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
312 | ||
c482c568 | 313 | return (!ipv6_only_sock(sk2) && |
c720c7e8 ED |
314 | (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr || |
315 | inet1->inet_rcv_saddr == inet2->inet_rcv_saddr)); | |
db8dac20 DM |
316 | } |
317 | ||
d4cada4a ED |
318 | static unsigned int udp4_portaddr_hash(struct net *net, __be32 saddr, |
319 | unsigned int port) | |
320 | { | |
0eae88f3 | 321 | return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port; |
d4cada4a ED |
322 | } |
323 | ||
6ba5a3c5 | 324 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 325 | { |
30fff923 | 326 | unsigned int hash2_nulladdr = |
0eae88f3 | 327 | udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum); |
30fff923 ED |
328 | unsigned int hash2_partial = |
329 | udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0); | |
330 | ||
d4cada4a | 331 | /* precompute partial secondary hash */ |
30fff923 ED |
332 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
333 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr); | |
db8dac20 DM |
334 | } |
335 | ||
645ca708 ED |
336 | static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr, |
337 | unsigned short hnum, | |
338 | __be16 sport, __be32 daddr, __be16 dport, int dif) | |
339 | { | |
340 | int score = -1; | |
341 | ||
d4cada4a | 342 | if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum && |
645ca708 ED |
343 | !ipv6_only_sock(sk)) { |
344 | struct inet_sock *inet = inet_sk(sk); | |
345 | ||
ba418fa3 | 346 | score = (sk->sk_family == PF_INET ? 2 : 1); |
c720c7e8 ED |
347 | if (inet->inet_rcv_saddr) { |
348 | if (inet->inet_rcv_saddr != daddr) | |
645ca708 | 349 | return -1; |
ba418fa3 | 350 | score += 4; |
645ca708 | 351 | } |
c720c7e8 ED |
352 | if (inet->inet_daddr) { |
353 | if (inet->inet_daddr != saddr) | |
645ca708 | 354 | return -1; |
ba418fa3 | 355 | score += 4; |
645ca708 | 356 | } |
c720c7e8 ED |
357 | if (inet->inet_dport) { |
358 | if (inet->inet_dport != sport) | |
645ca708 | 359 | return -1; |
ba418fa3 | 360 | score += 4; |
645ca708 ED |
361 | } |
362 | if (sk->sk_bound_dev_if) { | |
363 | if (sk->sk_bound_dev_if != dif) | |
364 | return -1; | |
ba418fa3 | 365 | score += 4; |
645ca708 ED |
366 | } |
367 | } | |
368 | return score; | |
369 | } | |
370 | ||
5051ebd2 ED |
371 | /* |
372 | * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num) | |
373 | */ | |
5051ebd2 ED |
374 | static inline int compute_score2(struct sock *sk, struct net *net, |
375 | __be32 saddr, __be16 sport, | |
376 | __be32 daddr, unsigned int hnum, int dif) | |
377 | { | |
378 | int score = -1; | |
379 | ||
380 | if (net_eq(sock_net(sk), net) && !ipv6_only_sock(sk)) { | |
381 | struct inet_sock *inet = inet_sk(sk); | |
382 | ||
383 | if (inet->inet_rcv_saddr != daddr) | |
384 | return -1; | |
385 | if (inet->inet_num != hnum) | |
386 | return -1; | |
387 | ||
ba418fa3 | 388 | score = (sk->sk_family == PF_INET ? 2 : 1); |
5051ebd2 ED |
389 | if (inet->inet_daddr) { |
390 | if (inet->inet_daddr != saddr) | |
391 | return -1; | |
ba418fa3 | 392 | score += 4; |
5051ebd2 ED |
393 | } |
394 | if (inet->inet_dport) { | |
395 | if (inet->inet_dport != sport) | |
396 | return -1; | |
ba418fa3 | 397 | score += 4; |
5051ebd2 ED |
398 | } |
399 | if (sk->sk_bound_dev_if) { | |
400 | if (sk->sk_bound_dev_if != dif) | |
401 | return -1; | |
ba418fa3 | 402 | score += 4; |
5051ebd2 ED |
403 | } |
404 | } | |
405 | return score; | |
406 | } | |
407 | ||
5051ebd2 ED |
408 | |
409 | /* called with read_rcu_lock() */ | |
410 | static struct sock *udp4_lib_lookup2(struct net *net, | |
411 | __be32 saddr, __be16 sport, | |
412 | __be32 daddr, unsigned int hnum, int dif, | |
413 | struct udp_hslot *hslot2, unsigned int slot2) | |
414 | { | |
415 | struct sock *sk, *result; | |
416 | struct hlist_nulls_node *node; | |
ba418fa3 TH |
417 | int score, badness, matches = 0, reuseport = 0; |
418 | u32 hash = 0; | |
5051ebd2 ED |
419 | |
420 | begin: | |
421 | result = NULL; | |
ba418fa3 | 422 | badness = 0; |
5051ebd2 ED |
423 | udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { |
424 | score = compute_score2(sk, net, saddr, sport, | |
425 | daddr, hnum, dif); | |
426 | if (score > badness) { | |
427 | result = sk; | |
428 | badness = score; | |
ba418fa3 TH |
429 | reuseport = sk->sk_reuseport; |
430 | if (reuseport) { | |
431 | hash = inet_ehashfn(net, daddr, hnum, | |
432 | saddr, htons(sport)); | |
433 | matches = 1; | |
434 | } | |
435 | } else if (score == badness && reuseport) { | |
436 | matches++; | |
437 | if (((u64)hash * matches) >> 32 == 0) | |
438 | result = sk; | |
439 | hash = next_pseudo_random32(hash); | |
5051ebd2 ED |
440 | } |
441 | } | |
442 | /* | |
443 | * if the nulls value we got at the end of this lookup is | |
444 | * not the expected one, we must restart lookup. | |
445 | * We probably met an item that was moved to another chain. | |
446 | */ | |
447 | if (get_nulls_value(node) != slot2) | |
448 | goto begin; | |
5051ebd2 | 449 | if (result) { |
c31504dc | 450 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
5051ebd2 ED |
451 | result = NULL; |
452 | else if (unlikely(compute_score2(result, net, saddr, sport, | |
453 | daddr, hnum, dif) < badness)) { | |
454 | sock_put(result); | |
455 | goto begin; | |
456 | } | |
457 | } | |
458 | return result; | |
459 | } | |
460 | ||
db8dac20 DM |
461 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
462 | * harder than this. -DaveM | |
463 | */ | |
fce82338 | 464 | struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, |
db8dac20 | 465 | __be16 sport, __be32 daddr, __be16 dport, |
645ca708 | 466 | int dif, struct udp_table *udptable) |
db8dac20 | 467 | { |
271b72c7 | 468 | struct sock *sk, *result; |
88ab1932 | 469 | struct hlist_nulls_node *node; |
db8dac20 | 470 | unsigned short hnum = ntohs(dport); |
5051ebd2 ED |
471 | unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); |
472 | struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; | |
ba418fa3 TH |
473 | int score, badness, matches = 0, reuseport = 0; |
474 | u32 hash = 0; | |
645ca708 | 475 | |
271b72c7 | 476 | rcu_read_lock(); |
5051ebd2 ED |
477 | if (hslot->count > 10) { |
478 | hash2 = udp4_portaddr_hash(net, daddr, hnum); | |
479 | slot2 = hash2 & udptable->mask; | |
480 | hslot2 = &udptable->hash2[slot2]; | |
481 | if (hslot->count < hslot2->count) | |
482 | goto begin; | |
483 | ||
484 | result = udp4_lib_lookup2(net, saddr, sport, | |
485 | daddr, hnum, dif, | |
486 | hslot2, slot2); | |
487 | if (!result) { | |
0eae88f3 | 488 | hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum); |
5051ebd2 ED |
489 | slot2 = hash2 & udptable->mask; |
490 | hslot2 = &udptable->hash2[slot2]; | |
491 | if (hslot->count < hslot2->count) | |
492 | goto begin; | |
493 | ||
1223c67c | 494 | result = udp4_lib_lookup2(net, saddr, sport, |
0eae88f3 | 495 | htonl(INADDR_ANY), hnum, dif, |
5051ebd2 ED |
496 | hslot2, slot2); |
497 | } | |
498 | rcu_read_unlock(); | |
499 | return result; | |
500 | } | |
271b72c7 ED |
501 | begin: |
502 | result = NULL; | |
ba418fa3 | 503 | badness = 0; |
88ab1932 | 504 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { |
645ca708 ED |
505 | score = compute_score(sk, net, saddr, hnum, sport, |
506 | daddr, dport, dif); | |
507 | if (score > badness) { | |
508 | result = sk; | |
509 | badness = score; | |
ba418fa3 TH |
510 | reuseport = sk->sk_reuseport; |
511 | if (reuseport) { | |
512 | hash = inet_ehashfn(net, daddr, hnum, | |
513 | saddr, htons(sport)); | |
514 | matches = 1; | |
515 | } | |
516 | } else if (score == badness && reuseport) { | |
517 | matches++; | |
518 | if (((u64)hash * matches) >> 32 == 0) | |
519 | result = sk; | |
520 | hash = next_pseudo_random32(hash); | |
db8dac20 DM |
521 | } |
522 | } | |
88ab1932 ED |
523 | /* |
524 | * if the nulls value we got at the end of this lookup is | |
525 | * not the expected one, we must restart lookup. | |
526 | * We probably met an item that was moved to another chain. | |
527 | */ | |
5051ebd2 | 528 | if (get_nulls_value(node) != slot) |
88ab1932 ED |
529 | goto begin; |
530 | ||
271b72c7 | 531 | if (result) { |
c31504dc | 532 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
271b72c7 ED |
533 | result = NULL; |
534 | else if (unlikely(compute_score(result, net, saddr, hnum, sport, | |
535 | daddr, dport, dif) < badness)) { | |
536 | sock_put(result); | |
537 | goto begin; | |
538 | } | |
539 | } | |
540 | rcu_read_unlock(); | |
db8dac20 DM |
541 | return result; |
542 | } | |
fce82338 | 543 | EXPORT_SYMBOL_GPL(__udp4_lib_lookup); |
db8dac20 | 544 | |
607c4aaf KK |
545 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
546 | __be16 sport, __be16 dport, | |
645ca708 | 547 | struct udp_table *udptable) |
607c4aaf | 548 | { |
23542618 | 549 | struct sock *sk; |
607c4aaf KK |
550 | const struct iphdr *iph = ip_hdr(skb); |
551 | ||
23542618 KK |
552 | if (unlikely(sk = skb_steal_sock(skb))) |
553 | return sk; | |
554 | else | |
adf30907 | 555 | return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport, |
23542618 KK |
556 | iph->daddr, dport, inet_iif(skb), |
557 | udptable); | |
607c4aaf KK |
558 | } |
559 | ||
bcd41303 KK |
560 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
561 | __be32 daddr, __be16 dport, int dif) | |
562 | { | |
645ca708 | 563 | return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table); |
bcd41303 KK |
564 | } |
565 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
566 | ||
920a4611 | 567 | static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk, |
db8dac20 DM |
568 | __be16 loc_port, __be32 loc_addr, |
569 | __be16 rmt_port, __be32 rmt_addr, | |
570 | int dif) | |
571 | { | |
88ab1932 | 572 | struct hlist_nulls_node *node; |
db8dac20 DM |
573 | struct sock *s = sk; |
574 | unsigned short hnum = ntohs(loc_port); | |
575 | ||
88ab1932 | 576 | sk_nulls_for_each_from(s, node) { |
db8dac20 DM |
577 | struct inet_sock *inet = inet_sk(s); |
578 | ||
9d4fb27d JP |
579 | if (!net_eq(sock_net(s), net) || |
580 | udp_sk(s)->udp_port_hash != hnum || | |
581 | (inet->inet_daddr && inet->inet_daddr != rmt_addr) || | |
582 | (inet->inet_dport != rmt_port && inet->inet_dport) || | |
583 | (inet->inet_rcv_saddr && | |
584 | inet->inet_rcv_saddr != loc_addr) || | |
585 | ipv6_only_sock(s) || | |
db8dac20 DM |
586 | (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)) |
587 | continue; | |
588 | if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif)) | |
589 | continue; | |
590 | goto found; | |
591 | } | |
592 | s = NULL; | |
593 | found: | |
594 | return s; | |
595 | } | |
596 | ||
597 | /* | |
598 | * This routine is called by the ICMP module when it gets some | |
599 | * sort of error condition. If err < 0 then the socket should | |
600 | * be closed and the error returned to the user. If err > 0 | |
601 | * it's just the icmp type << 8 | icmp code. | |
602 | * Header points to the ip header of the error packet. We move | |
603 | * on past this. Then (as it used to claim before adjustment) | |
604 | * header points to the first 8 bytes of the udp header. We need | |
605 | * to find the appropriate port. | |
606 | */ | |
607 | ||
645ca708 | 608 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
609 | { |
610 | struct inet_sock *inet; | |
b71d1d42 | 611 | const struct iphdr *iph = (const struct iphdr *)skb->data; |
c482c568 | 612 | struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2)); |
db8dac20 DM |
613 | const int type = icmp_hdr(skb)->type; |
614 | const int code = icmp_hdr(skb)->code; | |
615 | struct sock *sk; | |
616 | int harderr; | |
617 | int err; | |
fd54d716 | 618 | struct net *net = dev_net(skb->dev); |
db8dac20 | 619 | |
fd54d716 | 620 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
db8dac20 DM |
621 | iph->saddr, uh->source, skb->dev->ifindex, udptable); |
622 | if (sk == NULL) { | |
dcfc23ca | 623 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
624 | return; /* No socket for error */ |
625 | } | |
626 | ||
627 | err = 0; | |
628 | harderr = 0; | |
629 | inet = inet_sk(sk); | |
630 | ||
631 | switch (type) { | |
632 | default: | |
633 | case ICMP_TIME_EXCEEDED: | |
634 | err = EHOSTUNREACH; | |
635 | break; | |
636 | case ICMP_SOURCE_QUENCH: | |
637 | goto out; | |
638 | case ICMP_PARAMETERPROB: | |
639 | err = EPROTO; | |
640 | harderr = 1; | |
641 | break; | |
642 | case ICMP_DEST_UNREACH: | |
643 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
36393395 | 644 | ipv4_sk_update_pmtu(skb, sk, info); |
db8dac20 DM |
645 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { |
646 | err = EMSGSIZE; | |
647 | harderr = 1; | |
648 | break; | |
649 | } | |
650 | goto out; | |
651 | } | |
652 | err = EHOSTUNREACH; | |
653 | if (code <= NR_ICMP_UNREACH) { | |
654 | harderr = icmp_err_convert[code].fatal; | |
655 | err = icmp_err_convert[code].errno; | |
656 | } | |
657 | break; | |
55be7a9c DM |
658 | case ICMP_REDIRECT: |
659 | ipv4_sk_redirect(skb, sk); | |
660 | break; | |
db8dac20 DM |
661 | } |
662 | ||
663 | /* | |
664 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
665 | * 4.1.3.3. | |
666 | */ | |
667 | if (!inet->recverr) { | |
668 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
669 | goto out; | |
b1faf566 | 670 | } else |
c482c568 | 671 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1)); |
b1faf566 | 672 | |
db8dac20 DM |
673 | sk->sk_err = err; |
674 | sk->sk_error_report(sk); | |
675 | out: | |
676 | sock_put(sk); | |
677 | } | |
678 | ||
679 | void udp_err(struct sk_buff *skb, u32 info) | |
680 | { | |
645ca708 | 681 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
682 | } |
683 | ||
684 | /* | |
685 | * Throw away all pending data and cancel the corking. Socket is locked. | |
686 | */ | |
36d926b9 | 687 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
688 | { |
689 | struct udp_sock *up = udp_sk(sk); | |
690 | ||
691 | if (up->pending) { | |
692 | up->len = 0; | |
693 | up->pending = 0; | |
694 | ip_flush_pending_frames(sk); | |
695 | } | |
696 | } | |
36d926b9 | 697 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
698 | |
699 | /** | |
f6b9664f | 700 | * udp4_hwcsum - handle outgoing HW checksumming |
db8dac20 DM |
701 | * @skb: sk_buff containing the filled-in UDP header |
702 | * (checksum field must be zeroed out) | |
f6b9664f HX |
703 | * @src: source IP address |
704 | * @dst: destination IP address | |
db8dac20 | 705 | */ |
f6b9664f | 706 | static void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst) |
db8dac20 | 707 | { |
db8dac20 | 708 | struct udphdr *uh = udp_hdr(skb); |
f6b9664f HX |
709 | struct sk_buff *frags = skb_shinfo(skb)->frag_list; |
710 | int offset = skb_transport_offset(skb); | |
711 | int len = skb->len - offset; | |
712 | int hlen = len; | |
db8dac20 DM |
713 | __wsum csum = 0; |
714 | ||
f6b9664f | 715 | if (!frags) { |
db8dac20 DM |
716 | /* |
717 | * Only one fragment on the socket. | |
718 | */ | |
719 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
720 | skb->csum_offset = offsetof(struct udphdr, check); | |
f6b9664f HX |
721 | uh->check = ~csum_tcpudp_magic(src, dst, len, |
722 | IPPROTO_UDP, 0); | |
db8dac20 DM |
723 | } else { |
724 | /* | |
725 | * HW-checksum won't work as there are two or more | |
726 | * fragments on the socket so that all csums of sk_buffs | |
727 | * should be together | |
728 | */ | |
f6b9664f HX |
729 | do { |
730 | csum = csum_add(csum, frags->csum); | |
731 | hlen -= frags->len; | |
732 | } while ((frags = frags->next)); | |
db8dac20 | 733 | |
f6b9664f | 734 | csum = skb_checksum(skb, offset, hlen, csum); |
db8dac20 DM |
735 | skb->ip_summed = CHECKSUM_NONE; |
736 | ||
db8dac20 DM |
737 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); |
738 | if (uh->check == 0) | |
739 | uh->check = CSUM_MANGLED_0; | |
740 | } | |
741 | } | |
742 | ||
79ab0531 | 743 | static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4) |
db8dac20 | 744 | { |
f6b9664f | 745 | struct sock *sk = skb->sk; |
db8dac20 | 746 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
747 | struct udphdr *uh; |
748 | int err = 0; | |
749 | int is_udplite = IS_UDPLITE(sk); | |
f6b9664f HX |
750 | int offset = skb_transport_offset(skb); |
751 | int len = skb->len - offset; | |
db8dac20 DM |
752 | __wsum csum = 0; |
753 | ||
db8dac20 DM |
754 | /* |
755 | * Create a UDP header | |
756 | */ | |
757 | uh = udp_hdr(skb); | |
f6b9664f | 758 | uh->source = inet->inet_sport; |
79ab0531 | 759 | uh->dest = fl4->fl4_dport; |
f6b9664f | 760 | uh->len = htons(len); |
db8dac20 DM |
761 | uh->check = 0; |
762 | ||
763 | if (is_udplite) /* UDP-Lite */ | |
f6b9664f | 764 | csum = udplite_csum(skb); |
db8dac20 DM |
765 | |
766 | else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */ | |
767 | ||
768 | skb->ip_summed = CHECKSUM_NONE; | |
769 | goto send; | |
770 | ||
771 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
772 | ||
79ab0531 | 773 | udp4_hwcsum(skb, fl4->saddr, fl4->daddr); |
db8dac20 DM |
774 | goto send; |
775 | ||
f6b9664f HX |
776 | } else |
777 | csum = udp_csum(skb); | |
db8dac20 DM |
778 | |
779 | /* add protocol-dependent pseudo-header */ | |
79ab0531 | 780 | uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len, |
c482c568 | 781 | sk->sk_protocol, csum); |
db8dac20 DM |
782 | if (uh->check == 0) |
783 | uh->check = CSUM_MANGLED_0; | |
784 | ||
785 | send: | |
b5ec8eea | 786 | err = ip_send_skb(sock_net(sk), skb); |
6ce9e7b5 ED |
787 | if (err) { |
788 | if (err == -ENOBUFS && !inet->recverr) { | |
789 | UDP_INC_STATS_USER(sock_net(sk), | |
790 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
791 | err = 0; | |
792 | } | |
793 | } else | |
794 | UDP_INC_STATS_USER(sock_net(sk), | |
795 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
f6b9664f HX |
796 | return err; |
797 | } | |
798 | ||
799 | /* | |
800 | * Push out all pending data as one UDP datagram. Socket is locked. | |
801 | */ | |
802 | static int udp_push_pending_frames(struct sock *sk) | |
803 | { | |
804 | struct udp_sock *up = udp_sk(sk); | |
805 | struct inet_sock *inet = inet_sk(sk); | |
b6f21b26 | 806 | struct flowi4 *fl4 = &inet->cork.fl.u.ip4; |
f6b9664f HX |
807 | struct sk_buff *skb; |
808 | int err = 0; | |
809 | ||
77968b78 | 810 | skb = ip_finish_skb(sk, fl4); |
f6b9664f HX |
811 | if (!skb) |
812 | goto out; | |
813 | ||
79ab0531 | 814 | err = udp_send_skb(skb, fl4); |
f6b9664f | 815 | |
db8dac20 DM |
816 | out: |
817 | up->len = 0; | |
818 | up->pending = 0; | |
db8dac20 DM |
819 | return err; |
820 | } | |
821 | ||
822 | int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
823 | size_t len) | |
824 | { | |
825 | struct inet_sock *inet = inet_sk(sk); | |
826 | struct udp_sock *up = udp_sk(sk); | |
e474995f | 827 | struct flowi4 fl4_stack; |
b6f21b26 | 828 | struct flowi4 *fl4; |
db8dac20 DM |
829 | int ulen = len; |
830 | struct ipcm_cookie ipc; | |
831 | struct rtable *rt = NULL; | |
832 | int free = 0; | |
833 | int connected = 0; | |
834 | __be32 daddr, faddr, saddr; | |
835 | __be16 dport; | |
836 | u8 tos; | |
837 | int err, is_udplite = IS_UDPLITE(sk); | |
838 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
839 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
903ab86d | 840 | struct sk_buff *skb; |
f6d8bd05 | 841 | struct ip_options_data opt_copy; |
db8dac20 DM |
842 | |
843 | if (len > 0xFFFF) | |
844 | return -EMSGSIZE; | |
845 | ||
846 | /* | |
847 | * Check the flags. | |
848 | */ | |
849 | ||
c482c568 | 850 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */ |
db8dac20 DM |
851 | return -EOPNOTSUPP; |
852 | ||
853 | ipc.opt = NULL; | |
2244d07b | 854 | ipc.tx_flags = 0; |
db8dac20 | 855 | |
903ab86d HX |
856 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
857 | ||
f5fca608 | 858 | fl4 = &inet->cork.fl.u.ip4; |
db8dac20 DM |
859 | if (up->pending) { |
860 | /* | |
861 | * There are pending frames. | |
862 | * The socket lock must be held while it's corked. | |
863 | */ | |
864 | lock_sock(sk); | |
865 | if (likely(up->pending)) { | |
866 | if (unlikely(up->pending != AF_INET)) { | |
867 | release_sock(sk); | |
868 | return -EINVAL; | |
869 | } | |
870 | goto do_append_data; | |
871 | } | |
872 | release_sock(sk); | |
873 | } | |
874 | ulen += sizeof(struct udphdr); | |
875 | ||
876 | /* | |
877 | * Get and verify the address. | |
878 | */ | |
879 | if (msg->msg_name) { | |
5e73ea1a | 880 | struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name; |
db8dac20 DM |
881 | if (msg->msg_namelen < sizeof(*usin)) |
882 | return -EINVAL; | |
883 | if (usin->sin_family != AF_INET) { | |
884 | if (usin->sin_family != AF_UNSPEC) | |
885 | return -EAFNOSUPPORT; | |
886 | } | |
887 | ||
888 | daddr = usin->sin_addr.s_addr; | |
889 | dport = usin->sin_port; | |
890 | if (dport == 0) | |
891 | return -EINVAL; | |
892 | } else { | |
893 | if (sk->sk_state != TCP_ESTABLISHED) | |
894 | return -EDESTADDRREQ; | |
c720c7e8 ED |
895 | daddr = inet->inet_daddr; |
896 | dport = inet->inet_dport; | |
db8dac20 DM |
897 | /* Open fast path for connected socket. |
898 | Route will not be used, if at least one option is set. | |
899 | */ | |
900 | connected = 1; | |
901 | } | |
c720c7e8 | 902 | ipc.addr = inet->inet_saddr; |
db8dac20 DM |
903 | |
904 | ipc.oif = sk->sk_bound_dev_if; | |
2244d07b | 905 | err = sock_tx_timestamp(sk, &ipc.tx_flags); |
51f31cab PO |
906 | if (err) |
907 | return err; | |
db8dac20 | 908 | if (msg->msg_controllen) { |
3b1e0a65 | 909 | err = ip_cmsg_send(sock_net(sk), msg, &ipc); |
db8dac20 DM |
910 | if (err) |
911 | return err; | |
912 | if (ipc.opt) | |
913 | free = 1; | |
914 | connected = 0; | |
915 | } | |
f6d8bd05 ED |
916 | if (!ipc.opt) { |
917 | struct ip_options_rcu *inet_opt; | |
918 | ||
919 | rcu_read_lock(); | |
920 | inet_opt = rcu_dereference(inet->inet_opt); | |
921 | if (inet_opt) { | |
922 | memcpy(&opt_copy, inet_opt, | |
923 | sizeof(*inet_opt) + inet_opt->opt.optlen); | |
924 | ipc.opt = &opt_copy.opt; | |
925 | } | |
926 | rcu_read_unlock(); | |
927 | } | |
db8dac20 DM |
928 | |
929 | saddr = ipc.addr; | |
930 | ipc.addr = faddr = daddr; | |
931 | ||
f6d8bd05 | 932 | if (ipc.opt && ipc.opt->opt.srr) { |
db8dac20 DM |
933 | if (!daddr) |
934 | return -EINVAL; | |
f6d8bd05 | 935 | faddr = ipc.opt->opt.faddr; |
db8dac20 DM |
936 | connected = 0; |
937 | } | |
938 | tos = RT_TOS(inet->tos); | |
939 | if (sock_flag(sk, SOCK_LOCALROUTE) || | |
940 | (msg->msg_flags & MSG_DONTROUTE) || | |
f6d8bd05 | 941 | (ipc.opt && ipc.opt->opt.is_strictroute)) { |
db8dac20 DM |
942 | tos |= RTO_ONLINK; |
943 | connected = 0; | |
944 | } | |
945 | ||
946 | if (ipv4_is_multicast(daddr)) { | |
947 | if (!ipc.oif) | |
948 | ipc.oif = inet->mc_index; | |
949 | if (!saddr) | |
950 | saddr = inet->mc_addr; | |
951 | connected = 0; | |
76e21053 EH |
952 | } else if (!ipc.oif) |
953 | ipc.oif = inet->uc_index; | |
db8dac20 DM |
954 | |
955 | if (connected) | |
c482c568 | 956 | rt = (struct rtable *)sk_dst_check(sk, 0); |
db8dac20 DM |
957 | |
958 | if (rt == NULL) { | |
84a3aa00 PE |
959 | struct net *net = sock_net(sk); |
960 | ||
e474995f DM |
961 | fl4 = &fl4_stack; |
962 | flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos, | |
c0951cbc DM |
963 | RT_SCOPE_UNIVERSE, sk->sk_protocol, |
964 | inet_sk_flowi_flags(sk)|FLOWI_FLAG_CAN_SLEEP, | |
965 | faddr, saddr, dport, inet->inet_sport); | |
966 | ||
e474995f DM |
967 | security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); |
968 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe DM |
969 | if (IS_ERR(rt)) { |
970 | err = PTR_ERR(rt); | |
06dc94b1 | 971 | rt = NULL; |
db8dac20 | 972 | if (err == -ENETUNREACH) |
7c73a6fa | 973 | IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
974 | goto out; |
975 | } | |
976 | ||
977 | err = -EACCES; | |
978 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
979 | !sock_flag(sk, SOCK_BROADCAST)) | |
980 | goto out; | |
981 | if (connected) | |
d8d1f30b | 982 | sk_dst_set(sk, dst_clone(&rt->dst)); |
db8dac20 DM |
983 | } |
984 | ||
985 | if (msg->msg_flags&MSG_CONFIRM) | |
986 | goto do_confirm; | |
987 | back_from_confirm: | |
988 | ||
e474995f | 989 | saddr = fl4->saddr; |
db8dac20 | 990 | if (!ipc.addr) |
e474995f | 991 | daddr = ipc.addr = fl4->daddr; |
db8dac20 | 992 | |
903ab86d HX |
993 | /* Lockless fast path for the non-corking case. */ |
994 | if (!corkreq) { | |
77968b78 | 995 | skb = ip_make_skb(sk, fl4, getfrag, msg->msg_iov, ulen, |
903ab86d HX |
996 | sizeof(struct udphdr), &ipc, &rt, |
997 | msg->msg_flags); | |
998 | err = PTR_ERR(skb); | |
50c3a487 | 999 | if (!IS_ERR_OR_NULL(skb)) |
79ab0531 | 1000 | err = udp_send_skb(skb, fl4); |
903ab86d HX |
1001 | goto out; |
1002 | } | |
1003 | ||
db8dac20 DM |
1004 | lock_sock(sk); |
1005 | if (unlikely(up->pending)) { | |
1006 | /* The socket is already corked while preparing it. */ | |
1007 | /* ... which is an evident application bug. --ANK */ | |
1008 | release_sock(sk); | |
1009 | ||
afd46503 | 1010 | LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("cork app bug 2\n")); |
db8dac20 DM |
1011 | err = -EINVAL; |
1012 | goto out; | |
1013 | } | |
1014 | /* | |
1015 | * Now cork the socket to pend data. | |
1016 | */ | |
b6f21b26 DM |
1017 | fl4 = &inet->cork.fl.u.ip4; |
1018 | fl4->daddr = daddr; | |
1019 | fl4->saddr = saddr; | |
9cce96df DM |
1020 | fl4->fl4_dport = dport; |
1021 | fl4->fl4_sport = inet->inet_sport; | |
db8dac20 DM |
1022 | up->pending = AF_INET; |
1023 | ||
1024 | do_append_data: | |
1025 | up->len += ulen; | |
f5fca608 DM |
1026 | err = ip_append_data(sk, fl4, getfrag, msg->msg_iov, ulen, |
1027 | sizeof(struct udphdr), &ipc, &rt, | |
1028 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | |
db8dac20 DM |
1029 | if (err) |
1030 | udp_flush_pending_frames(sk); | |
1031 | else if (!corkreq) | |
1032 | err = udp_push_pending_frames(sk); | |
1033 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
1034 | up->pending = 0; | |
1035 | release_sock(sk); | |
1036 | ||
1037 | out: | |
1038 | ip_rt_put(rt); | |
1039 | if (free) | |
1040 | kfree(ipc.opt); | |
1041 | if (!err) | |
1042 | return len; | |
1043 | /* | |
1044 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1045 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1046 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1047 | * things). We could add another new stat but at least for now that | |
1048 | * seems like overkill. | |
1049 | */ | |
1050 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
1051 | UDP_INC_STATS_USER(sock_net(sk), |
1052 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
1053 | } |
1054 | return err; | |
1055 | ||
1056 | do_confirm: | |
d8d1f30b | 1057 | dst_confirm(&rt->dst); |
db8dac20 DM |
1058 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1059 | goto back_from_confirm; | |
1060 | err = 0; | |
1061 | goto out; | |
1062 | } | |
c482c568 | 1063 | EXPORT_SYMBOL(udp_sendmsg); |
db8dac20 DM |
1064 | |
1065 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
1066 | size_t size, int flags) | |
1067 | { | |
f5fca608 | 1068 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
1069 | struct udp_sock *up = udp_sk(sk); |
1070 | int ret; | |
1071 | ||
1072 | if (!up->pending) { | |
1073 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
1074 | ||
1075 | /* Call udp_sendmsg to specify destination address which | |
1076 | * sendpage interface can't pass. | |
1077 | * This will succeed only when the socket is connected. | |
1078 | */ | |
1079 | ret = udp_sendmsg(NULL, sk, &msg, 0); | |
1080 | if (ret < 0) | |
1081 | return ret; | |
1082 | } | |
1083 | ||
1084 | lock_sock(sk); | |
1085 | ||
1086 | if (unlikely(!up->pending)) { | |
1087 | release_sock(sk); | |
1088 | ||
afd46503 | 1089 | LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("udp cork app bug 3\n")); |
db8dac20 DM |
1090 | return -EINVAL; |
1091 | } | |
1092 | ||
f5fca608 DM |
1093 | ret = ip_append_page(sk, &inet->cork.fl.u.ip4, |
1094 | page, offset, size, flags); | |
db8dac20 DM |
1095 | if (ret == -EOPNOTSUPP) { |
1096 | release_sock(sk); | |
1097 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
1098 | size, flags); | |
1099 | } | |
1100 | if (ret < 0) { | |
1101 | udp_flush_pending_frames(sk); | |
1102 | goto out; | |
1103 | } | |
1104 | ||
1105 | up->len += size; | |
1106 | if (!(up->corkflag || (flags&MSG_MORE))) | |
1107 | ret = udp_push_pending_frames(sk); | |
1108 | if (!ret) | |
1109 | ret = size; | |
1110 | out: | |
1111 | release_sock(sk); | |
1112 | return ret; | |
1113 | } | |
1114 | ||
85584672 ED |
1115 | |
1116 | /** | |
1117 | * first_packet_length - return length of first packet in receive queue | |
1118 | * @sk: socket | |
1119 | * | |
1120 | * Drops all bad checksum frames, until a valid one is found. | |
1121 | * Returns the length of found skb, or 0 if none is found. | |
1122 | */ | |
1123 | static unsigned int first_packet_length(struct sock *sk) | |
1124 | { | |
1125 | struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue; | |
1126 | struct sk_buff *skb; | |
1127 | unsigned int res; | |
1128 | ||
1129 | __skb_queue_head_init(&list_kill); | |
1130 | ||
1131 | spin_lock_bh(&rcvq->lock); | |
1132 | while ((skb = skb_peek(rcvq)) != NULL && | |
1133 | udp_lib_checksum_complete(skb)) { | |
1134 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, | |
1135 | IS_UDPLITE(sk)); | |
8edf19c2 | 1136 | atomic_inc(&sk->sk_drops); |
85584672 ED |
1137 | __skb_unlink(skb, rcvq); |
1138 | __skb_queue_tail(&list_kill, skb); | |
1139 | } | |
1140 | res = skb ? skb->len : 0; | |
1141 | spin_unlock_bh(&rcvq->lock); | |
1142 | ||
1143 | if (!skb_queue_empty(&list_kill)) { | |
8a74ad60 ED |
1144 | bool slow = lock_sock_fast(sk); |
1145 | ||
85584672 ED |
1146 | __skb_queue_purge(&list_kill); |
1147 | sk_mem_reclaim_partial(sk); | |
8a74ad60 | 1148 | unlock_sock_fast(sk, slow); |
85584672 ED |
1149 | } |
1150 | return res; | |
1151 | } | |
1152 | ||
1da177e4 LT |
1153 | /* |
1154 | * IOCTL requests applicable to the UDP protocol | |
1155 | */ | |
e905a9ed | 1156 | |
1da177e4 LT |
1157 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
1158 | { | |
6516c655 SH |
1159 | switch (cmd) { |
1160 | case SIOCOUTQ: | |
1da177e4 | 1161 | { |
31e6d363 ED |
1162 | int amount = sk_wmem_alloc_get(sk); |
1163 | ||
6516c655 SH |
1164 | return put_user(amount, (int __user *)arg); |
1165 | } | |
1da177e4 | 1166 | |
6516c655 SH |
1167 | case SIOCINQ: |
1168 | { | |
85584672 | 1169 | unsigned int amount = first_packet_length(sk); |
6516c655 | 1170 | |
85584672 | 1171 | if (amount) |
6516c655 SH |
1172 | /* |
1173 | * We will only return the amount | |
1174 | * of this packet since that is all | |
1175 | * that will be read. | |
1176 | */ | |
85584672 ED |
1177 | amount -= sizeof(struct udphdr); |
1178 | ||
6516c655 SH |
1179 | return put_user(amount, (int __user *)arg); |
1180 | } | |
1da177e4 | 1181 | |
6516c655 SH |
1182 | default: |
1183 | return -ENOIOCTLCMD; | |
1da177e4 | 1184 | } |
6516c655 SH |
1185 | |
1186 | return 0; | |
1da177e4 | 1187 | } |
c482c568 | 1188 | EXPORT_SYMBOL(udp_ioctl); |
1da177e4 | 1189 | |
db8dac20 DM |
1190 | /* |
1191 | * This should be easy, if there is something there we | |
1192 | * return it, otherwise we block. | |
1193 | */ | |
1194 | ||
1195 | int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
1196 | size_t len, int noblock, int flags, int *addr_len) | |
1197 | { | |
1198 | struct inet_sock *inet = inet_sk(sk); | |
1199 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; | |
1200 | struct sk_buff *skb; | |
59c2cdae | 1201 | unsigned int ulen, copied; |
3f518bf7 | 1202 | int peeked, off = 0; |
db8dac20 DM |
1203 | int err; |
1204 | int is_udplite = IS_UDPLITE(sk); | |
8a74ad60 | 1205 | bool slow; |
db8dac20 DM |
1206 | |
1207 | /* | |
1208 | * Check any passed addresses | |
1209 | */ | |
1210 | if (addr_len) | |
c482c568 | 1211 | *addr_len = sizeof(*sin); |
db8dac20 DM |
1212 | |
1213 | if (flags & MSG_ERRQUEUE) | |
1214 | return ip_recv_error(sk, msg, len); | |
1215 | ||
1216 | try_again: | |
1217 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), | |
3f518bf7 | 1218 | &peeked, &off, &err); |
db8dac20 DM |
1219 | if (!skb) |
1220 | goto out; | |
1221 | ||
1222 | ulen = skb->len - sizeof(struct udphdr); | |
59c2cdae DM |
1223 | copied = len; |
1224 | if (copied > ulen) | |
1225 | copied = ulen; | |
1226 | else if (copied < ulen) | |
db8dac20 DM |
1227 | msg->msg_flags |= MSG_TRUNC; |
1228 | ||
1229 | /* | |
1230 | * If checksum is needed at all, try to do it while copying the | |
1231 | * data. If the data is truncated, or if we only want a partial | |
1232 | * coverage checksum (UDP-Lite), do it before the copy. | |
1233 | */ | |
1234 | ||
59c2cdae | 1235 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { |
db8dac20 DM |
1236 | if (udp_lib_checksum_complete(skb)) |
1237 | goto csum_copy_err; | |
1238 | } | |
1239 | ||
1240 | if (skb_csum_unnecessary(skb)) | |
1241 | err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), | |
59c2cdae | 1242 | msg->msg_iov, copied); |
db8dac20 | 1243 | else { |
c482c568 ED |
1244 | err = skb_copy_and_csum_datagram_iovec(skb, |
1245 | sizeof(struct udphdr), | |
1246 | msg->msg_iov); | |
db8dac20 DM |
1247 | |
1248 | if (err == -EINVAL) | |
1249 | goto csum_copy_err; | |
1250 | } | |
1251 | ||
22911fc5 ED |
1252 | if (unlikely(err)) { |
1253 | trace_kfree_skb(skb, udp_recvmsg); | |
979402b1 ED |
1254 | if (!peeked) { |
1255 | atomic_inc(&sk->sk_drops); | |
1256 | UDP_INC_STATS_USER(sock_net(sk), | |
1257 | UDP_MIB_INERRORS, is_udplite); | |
1258 | } | |
db8dac20 | 1259 | goto out_free; |
22911fc5 | 1260 | } |
db8dac20 DM |
1261 | |
1262 | if (!peeked) | |
629ca23c PE |
1263 | UDP_INC_STATS_USER(sock_net(sk), |
1264 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 | 1265 | |
3b885787 | 1266 | sock_recv_ts_and_drops(msg, sk, skb); |
db8dac20 DM |
1267 | |
1268 | /* Copy the address. */ | |
c482c568 | 1269 | if (sin) { |
db8dac20 DM |
1270 | sin->sin_family = AF_INET; |
1271 | sin->sin_port = udp_hdr(skb)->source; | |
1272 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
1273 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
1274 | } | |
1275 | if (inet->cmsg_flags) | |
1276 | ip_cmsg_recv(msg, skb); | |
1277 | ||
59c2cdae | 1278 | err = copied; |
db8dac20 DM |
1279 | if (flags & MSG_TRUNC) |
1280 | err = ulen; | |
1281 | ||
1282 | out_free: | |
9d410c79 | 1283 | skb_free_datagram_locked(sk, skb); |
db8dac20 DM |
1284 | out: |
1285 | return err; | |
1286 | ||
1287 | csum_copy_err: | |
8a74ad60 | 1288 | slow = lock_sock_fast(sk); |
db8dac20 | 1289 | if (!skb_kill_datagram(sk, skb, flags)) |
629ca23c | 1290 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8a74ad60 | 1291 | unlock_sock_fast(sk, slow); |
db8dac20 DM |
1292 | |
1293 | if (noblock) | |
1294 | return -EAGAIN; | |
9cfaa8de XZ |
1295 | |
1296 | /* starting over for a new packet */ | |
1297 | msg->msg_flags &= ~MSG_TRUNC; | |
db8dac20 DM |
1298 | goto try_again; |
1299 | } | |
1300 | ||
1301 | ||
1da177e4 LT |
1302 | int udp_disconnect(struct sock *sk, int flags) |
1303 | { | |
1304 | struct inet_sock *inet = inet_sk(sk); | |
1305 | /* | |
1306 | * 1003.1g - break association. | |
1307 | */ | |
e905a9ed | 1308 | |
1da177e4 | 1309 | sk->sk_state = TCP_CLOSE; |
c720c7e8 ED |
1310 | inet->inet_daddr = 0; |
1311 | inet->inet_dport = 0; | |
bdeab991 | 1312 | sock_rps_reset_rxhash(sk); |
1da177e4 LT |
1313 | sk->sk_bound_dev_if = 0; |
1314 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
1315 | inet_reset_saddr(sk); | |
1316 | ||
1317 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
1318 | sk->sk_prot->unhash(sk); | |
c720c7e8 | 1319 | inet->inet_sport = 0; |
1da177e4 LT |
1320 | } |
1321 | sk_dst_reset(sk); | |
1322 | return 0; | |
1323 | } | |
c482c568 | 1324 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1325 | |
645ca708 ED |
1326 | void udp_lib_unhash(struct sock *sk) |
1327 | { | |
723b4610 ED |
1328 | if (sk_hashed(sk)) { |
1329 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
512615b6 ED |
1330 | struct udp_hslot *hslot, *hslot2; |
1331 | ||
1332 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1333 | udp_sk(sk)->udp_port_hash); | |
1334 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
645ca708 | 1335 | |
723b4610 ED |
1336 | spin_lock_bh(&hslot->lock); |
1337 | if (sk_nulls_del_node_init_rcu(sk)) { | |
fdcc8aa9 | 1338 | hslot->count--; |
c720c7e8 | 1339 | inet_sk(sk)->inet_num = 0; |
723b4610 | 1340 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
512615b6 ED |
1341 | |
1342 | spin_lock(&hslot2->lock); | |
1343 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1344 | hslot2->count--; | |
1345 | spin_unlock(&hslot2->lock); | |
723b4610 ED |
1346 | } |
1347 | spin_unlock_bh(&hslot->lock); | |
645ca708 | 1348 | } |
645ca708 ED |
1349 | } |
1350 | EXPORT_SYMBOL(udp_lib_unhash); | |
1351 | ||
719f8358 ED |
1352 | /* |
1353 | * inet_rcv_saddr was changed, we must rehash secondary hash | |
1354 | */ | |
1355 | void udp_lib_rehash(struct sock *sk, u16 newhash) | |
1356 | { | |
1357 | if (sk_hashed(sk)) { | |
1358 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
1359 | struct udp_hslot *hslot, *hslot2, *nhslot2; | |
1360 | ||
1361 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
1362 | nhslot2 = udp_hashslot2(udptable, newhash); | |
1363 | udp_sk(sk)->udp_portaddr_hash = newhash; | |
1364 | if (hslot2 != nhslot2) { | |
1365 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1366 | udp_sk(sk)->udp_port_hash); | |
1367 | /* we must lock primary chain too */ | |
1368 | spin_lock_bh(&hslot->lock); | |
1369 | ||
1370 | spin_lock(&hslot2->lock); | |
1371 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1372 | hslot2->count--; | |
1373 | spin_unlock(&hslot2->lock); | |
1374 | ||
1375 | spin_lock(&nhslot2->lock); | |
1376 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
1377 | &nhslot2->head); | |
1378 | nhslot2->count++; | |
1379 | spin_unlock(&nhslot2->lock); | |
1380 | ||
1381 | spin_unlock_bh(&hslot->lock); | |
1382 | } | |
1383 | } | |
1384 | } | |
1385 | EXPORT_SYMBOL(udp_lib_rehash); | |
1386 | ||
1387 | static void udp_v4_rehash(struct sock *sk) | |
1388 | { | |
1389 | u16 new_hash = udp4_portaddr_hash(sock_net(sk), | |
1390 | inet_sk(sk)->inet_rcv_saddr, | |
1391 | inet_sk(sk)->inet_num); | |
1392 | udp_lib_rehash(sk, new_hash); | |
1393 | } | |
1394 | ||
93821778 HX |
1395 | static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
1396 | { | |
fec5e652 | 1397 | int rc; |
766e9037 | 1398 | |
fec5e652 | 1399 | if (inet_sk(sk)->inet_daddr) |
bdeab991 | 1400 | sock_rps_save_rxhash(sk, skb); |
fec5e652 | 1401 | |
d826eb14 | 1402 | rc = sock_queue_rcv_skb(sk, skb); |
766e9037 ED |
1403 | if (rc < 0) { |
1404 | int is_udplite = IS_UDPLITE(sk); | |
93821778 | 1405 | |
93821778 | 1406 | /* Note that an ENOMEM error is charged twice */ |
766e9037 | 1407 | if (rc == -ENOMEM) |
93821778 HX |
1408 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1409 | is_udplite); | |
766e9037 ED |
1410 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
1411 | kfree_skb(skb); | |
296f7ea7 | 1412 | trace_udp_fail_queue_rcv_skb(rc, sk); |
766e9037 | 1413 | return -1; |
93821778 HX |
1414 | } |
1415 | ||
1416 | return 0; | |
1417 | ||
93821778 HX |
1418 | } |
1419 | ||
447167bf ED |
1420 | static struct static_key udp_encap_needed __read_mostly; |
1421 | void udp_encap_enable(void) | |
1422 | { | |
1423 | if (!static_key_enabled(&udp_encap_needed)) | |
1424 | static_key_slow_inc(&udp_encap_needed); | |
1425 | } | |
1426 | EXPORT_SYMBOL(udp_encap_enable); | |
1427 | ||
db8dac20 DM |
1428 | /* returns: |
1429 | * -1: error | |
1430 | * 0: success | |
1431 | * >0: "udp encap" protocol resubmission | |
1432 | * | |
1433 | * Note that in the success and error cases, the skb is assumed to | |
1434 | * have either been requeued or freed. | |
1435 | */ | |
c482c568 | 1436 | int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
db8dac20 DM |
1437 | { |
1438 | struct udp_sock *up = udp_sk(sk); | |
1439 | int rc; | |
1440 | int is_udplite = IS_UDPLITE(sk); | |
1441 | ||
1442 | /* | |
1443 | * Charge it to the socket, dropping if the queue is full. | |
1444 | */ | |
1445 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1446 | goto drop; | |
1447 | nf_reset(skb); | |
1448 | ||
447167bf | 1449 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
0ad92ad0 ED |
1450 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
1451 | ||
db8dac20 DM |
1452 | /* |
1453 | * This is an encapsulation socket so pass the skb to | |
1454 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1455 | * fall through and pass this up the UDP socket. | |
1456 | * up->encap_rcv() returns the following value: | |
1457 | * =0 if skb was successfully passed to the encap | |
1458 | * handler or was discarded by it. | |
1459 | * >0 if skb should be passed on to UDP. | |
1460 | * <0 if skb should be resubmitted as proto -N | |
1461 | */ | |
1462 | ||
1463 | /* if we're overly short, let UDP handle it */ | |
0ad92ad0 ED |
1464 | encap_rcv = ACCESS_ONCE(up->encap_rcv); |
1465 | if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) { | |
db8dac20 DM |
1466 | int ret; |
1467 | ||
0ad92ad0 | 1468 | ret = encap_rcv(sk, skb); |
db8dac20 | 1469 | if (ret <= 0) { |
0283328e PE |
1470 | UDP_INC_STATS_BH(sock_net(sk), |
1471 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
1472 | is_udplite); |
1473 | return -ret; | |
1474 | } | |
1475 | } | |
1476 | ||
1477 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1478 | } | |
1479 | ||
1480 | /* | |
1481 | * UDP-Lite specific tests, ignored on UDP sockets | |
1482 | */ | |
1483 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1484 | ||
1485 | /* | |
1486 | * MIB statistics other than incrementing the error count are | |
1487 | * disabled for the following two types of errors: these depend | |
1488 | * on the application settings, not on the functioning of the | |
1489 | * protocol stack as such. | |
1490 | * | |
1491 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1492 | * way ... to ... at least let the receiving application block | |
1493 | * delivery of packets with coverage values less than a value | |
1494 | * provided by the application." | |
1495 | */ | |
1496 | if (up->pcrlen == 0) { /* full coverage was set */ | |
afd46503 JP |
1497 | LIMIT_NETDEBUG(KERN_WARNING "UDPLite: partial coverage %d while full coverage %d requested\n", |
1498 | UDP_SKB_CB(skb)->cscov, skb->len); | |
db8dac20 DM |
1499 | goto drop; |
1500 | } | |
1501 | /* The next case involves violating the min. coverage requested | |
1502 | * by the receiver. This is subtle: if receiver wants x and x is | |
1503 | * greater than the buffersize/MTU then receiver will complain | |
1504 | * that it wants x while sender emits packets of smaller size y. | |
1505 | * Therefore the above ...()->partial_cov statement is essential. | |
1506 | */ | |
1507 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
afd46503 JP |
1508 | LIMIT_NETDEBUG(KERN_WARNING "UDPLite: coverage %d too small, need min %d\n", |
1509 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
db8dac20 DM |
1510 | goto drop; |
1511 | } | |
1512 | } | |
1513 | ||
33d480ce ED |
1514 | if (rcu_access_pointer(sk->sk_filter) && |
1515 | udp_lib_checksum_complete(skb)) | |
1516 | goto drop; | |
db8dac20 | 1517 | |
c377411f | 1518 | |
f545a38f | 1519 | if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf)) |
c377411f ED |
1520 | goto drop; |
1521 | ||
93821778 | 1522 | rc = 0; |
db8dac20 | 1523 | |
d826eb14 | 1524 | ipv4_pktinfo_prepare(skb); |
93821778 HX |
1525 | bh_lock_sock(sk); |
1526 | if (!sock_owned_by_user(sk)) | |
1527 | rc = __udp_queue_rcv_skb(sk, skb); | |
f545a38f | 1528 | else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) { |
55349790 ZY |
1529 | bh_unlock_sock(sk); |
1530 | goto drop; | |
1531 | } | |
93821778 HX |
1532 | bh_unlock_sock(sk); |
1533 | ||
1534 | return rc; | |
db8dac20 DM |
1535 | |
1536 | drop: | |
0283328e | 1537 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8edf19c2 | 1538 | atomic_inc(&sk->sk_drops); |
db8dac20 DM |
1539 | kfree_skb(skb); |
1540 | return -1; | |
1541 | } | |
1542 | ||
1240d137 ED |
1543 | |
1544 | static void flush_stack(struct sock **stack, unsigned int count, | |
1545 | struct sk_buff *skb, unsigned int final) | |
1546 | { | |
1547 | unsigned int i; | |
1548 | struct sk_buff *skb1 = NULL; | |
f6b8f32c | 1549 | struct sock *sk; |
1240d137 ED |
1550 | |
1551 | for (i = 0; i < count; i++) { | |
f6b8f32c | 1552 | sk = stack[i]; |
1240d137 ED |
1553 | if (likely(skb1 == NULL)) |
1554 | skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC); | |
1555 | ||
f6b8f32c ED |
1556 | if (!skb1) { |
1557 | atomic_inc(&sk->sk_drops); | |
1558 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, | |
1559 | IS_UDPLITE(sk)); | |
1560 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, | |
1561 | IS_UDPLITE(sk)); | |
1562 | } | |
1563 | ||
1564 | if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0) | |
1240d137 ED |
1565 | skb1 = NULL; |
1566 | } | |
1567 | if (unlikely(skb1)) | |
1568 | kfree_skb(skb1); | |
1569 | } | |
1570 | ||
db8dac20 DM |
1571 | /* |
1572 | * Multicasts and broadcasts go to each listener. | |
1573 | * | |
1240d137 | 1574 | * Note: called only from the BH handler context. |
db8dac20 | 1575 | */ |
e3163493 | 1576 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1577 | struct udphdr *uh, |
1578 | __be32 saddr, __be32 daddr, | |
645ca708 | 1579 | struct udp_table *udptable) |
db8dac20 | 1580 | { |
1240d137 | 1581 | struct sock *sk, *stack[256 / sizeof(struct sock *)]; |
f86dcc5a | 1582 | struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest)); |
db8dac20 | 1583 | int dif; |
1240d137 | 1584 | unsigned int i, count = 0; |
db8dac20 | 1585 | |
645ca708 | 1586 | spin_lock(&hslot->lock); |
88ab1932 | 1587 | sk = sk_nulls_head(&hslot->head); |
db8dac20 | 1588 | dif = skb->dev->ifindex; |
920a4611 | 1589 | sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif); |
1240d137 ED |
1590 | while (sk) { |
1591 | stack[count++] = sk; | |
1592 | sk = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest, | |
1593 | daddr, uh->source, saddr, dif); | |
1594 | if (unlikely(count == ARRAY_SIZE(stack))) { | |
1595 | if (!sk) | |
1596 | break; | |
1597 | flush_stack(stack, count, skb, ~0); | |
1598 | count = 0; | |
1599 | } | |
1600 | } | |
1601 | /* | |
1602 | * before releasing chain lock, we must take a reference on sockets | |
1603 | */ | |
1604 | for (i = 0; i < count; i++) | |
1605 | sock_hold(stack[i]); | |
1606 | ||
645ca708 | 1607 | spin_unlock(&hslot->lock); |
1240d137 ED |
1608 | |
1609 | /* | |
1610 | * do the slow work with no lock held | |
1611 | */ | |
1612 | if (count) { | |
1613 | flush_stack(stack, count, skb, count - 1); | |
1614 | ||
1615 | for (i = 0; i < count; i++) | |
1616 | sock_put(stack[i]); | |
1617 | } else { | |
1618 | kfree_skb(skb); | |
1619 | } | |
db8dac20 DM |
1620 | return 0; |
1621 | } | |
1622 | ||
1623 | /* Initialize UDP checksum. If exited with zero value (success), | |
1624 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1625 | * Otherwise, csum completion requires chacksumming packet body, | |
1626 | * including udp header and folding it to skb->csum. | |
1627 | */ | |
1628 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1629 | int proto) | |
1630 | { | |
1631 | const struct iphdr *iph; | |
1632 | int err; | |
1633 | ||
1634 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1635 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1636 | ||
1637 | if (proto == IPPROTO_UDPLITE) { | |
1638 | err = udplite_checksum_init(skb, uh); | |
1639 | if (err) | |
1640 | return err; | |
1641 | } | |
1642 | ||
1643 | iph = ip_hdr(skb); | |
1644 | if (uh->check == 0) { | |
1645 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1646 | } else if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
c482c568 | 1647 | if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, |
db8dac20 DM |
1648 | proto, skb->csum)) |
1649 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1650 | } | |
1651 | if (!skb_csum_unnecessary(skb)) | |
1652 | skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, | |
1653 | skb->len, proto, 0); | |
1654 | /* Probably, we should checksum udp header (it should be in cache | |
1655 | * in any case) and data in tiny packets (< rx copybreak). | |
1656 | */ | |
1657 | ||
1658 | return 0; | |
1659 | } | |
1660 | ||
1661 | /* | |
1662 | * All we need to do is get the socket, and then do a checksum. | |
1663 | */ | |
1664 | ||
645ca708 | 1665 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1666 | int proto) |
1667 | { | |
1668 | struct sock *sk; | |
7b5e56f9 | 1669 | struct udphdr *uh; |
db8dac20 | 1670 | unsigned short ulen; |
adf30907 | 1671 | struct rtable *rt = skb_rtable(skb); |
2783ef23 | 1672 | __be32 saddr, daddr; |
0283328e | 1673 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1674 | |
1675 | /* | |
1676 | * Validate the packet. | |
1677 | */ | |
1678 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1679 | goto drop; /* No space for header. */ | |
1680 | ||
7b5e56f9 | 1681 | uh = udp_hdr(skb); |
db8dac20 | 1682 | ulen = ntohs(uh->len); |
ccc2d97c BM |
1683 | saddr = ip_hdr(skb)->saddr; |
1684 | daddr = ip_hdr(skb)->daddr; | |
1685 | ||
db8dac20 DM |
1686 | if (ulen > skb->len) |
1687 | goto short_packet; | |
1688 | ||
1689 | if (proto == IPPROTO_UDP) { | |
1690 | /* UDP validates ulen. */ | |
1691 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1692 | goto short_packet; | |
1693 | uh = udp_hdr(skb); | |
1694 | } | |
1695 | ||
1696 | if (udp4_csum_init(skb, uh, proto)) | |
1697 | goto csum_error; | |
1698 | ||
1699 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) | |
e3163493 PE |
1700 | return __udp4_lib_mcast_deliver(net, skb, uh, |
1701 | saddr, daddr, udptable); | |
db8dac20 | 1702 | |
607c4aaf | 1703 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); |
db8dac20 DM |
1704 | |
1705 | if (sk != NULL) { | |
93821778 | 1706 | int ret = udp_queue_rcv_skb(sk, skb); |
db8dac20 DM |
1707 | sock_put(sk); |
1708 | ||
1709 | /* a return value > 0 means to resubmit the input, but | |
1710 | * it wants the return to be -protocol, or 0 | |
1711 | */ | |
1712 | if (ret > 0) | |
1713 | return -ret; | |
1714 | return 0; | |
1715 | } | |
1716 | ||
1717 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1718 | goto drop; | |
1719 | nf_reset(skb); | |
1720 | ||
1721 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1722 | if (udp_lib_checksum_complete(skb)) | |
1723 | goto csum_error; | |
1724 | ||
0283328e | 1725 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1726 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1727 | ||
1728 | /* | |
1729 | * Hmm. We got an UDP packet to a port to which we | |
1730 | * don't wanna listen. Ignore it. | |
1731 | */ | |
1732 | kfree_skb(skb); | |
1733 | return 0; | |
1734 | ||
1735 | short_packet: | |
673d57e7 | 1736 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n", |
afd46503 JP |
1737 | proto == IPPROTO_UDPLITE ? "Lite" : "", |
1738 | &saddr, ntohs(uh->source), | |
1739 | ulen, skb->len, | |
1740 | &daddr, ntohs(uh->dest)); | |
db8dac20 DM |
1741 | goto drop; |
1742 | ||
1743 | csum_error: | |
1744 | /* | |
1745 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1746 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1747 | */ | |
673d57e7 | 1748 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n", |
afd46503 JP |
1749 | proto == IPPROTO_UDPLITE ? "Lite" : "", |
1750 | &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest), | |
db8dac20 DM |
1751 | ulen); |
1752 | drop: | |
0283328e | 1753 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1754 | kfree_skb(skb); |
1755 | return 0; | |
1756 | } | |
1757 | ||
1758 | int udp_rcv(struct sk_buff *skb) | |
1759 | { | |
645ca708 | 1760 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
1761 | } |
1762 | ||
7d06b2e0 | 1763 | void udp_destroy_sock(struct sock *sk) |
db8dac20 | 1764 | { |
8a74ad60 | 1765 | bool slow = lock_sock_fast(sk); |
db8dac20 | 1766 | udp_flush_pending_frames(sk); |
8a74ad60 | 1767 | unlock_sock_fast(sk, slow); |
db8dac20 DM |
1768 | } |
1769 | ||
1da177e4 LT |
1770 | /* |
1771 | * Socket option code for UDP | |
1772 | */ | |
4c0a6cb0 | 1773 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 1774 | char __user *optval, unsigned int optlen, |
4c0a6cb0 | 1775 | int (*push_pending_frames)(struct sock *)) |
1da177e4 LT |
1776 | { |
1777 | struct udp_sock *up = udp_sk(sk); | |
1778 | int val; | |
1779 | int err = 0; | |
b2bf1e26 | 1780 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 1781 | |
c482c568 | 1782 | if (optlen < sizeof(int)) |
1da177e4 LT |
1783 | return -EINVAL; |
1784 | ||
1785 | if (get_user(val, (int __user *)optval)) | |
1786 | return -EFAULT; | |
1787 | ||
6516c655 | 1788 | switch (optname) { |
1da177e4 LT |
1789 | case UDP_CORK: |
1790 | if (val != 0) { | |
1791 | up->corkflag = 1; | |
1792 | } else { | |
1793 | up->corkflag = 0; | |
1794 | lock_sock(sk); | |
4c0a6cb0 | 1795 | (*push_pending_frames)(sk); |
1da177e4 LT |
1796 | release_sock(sk); |
1797 | } | |
1798 | break; | |
e905a9ed | 1799 | |
1da177e4 LT |
1800 | case UDP_ENCAP: |
1801 | switch (val) { | |
1802 | case 0: | |
1803 | case UDP_ENCAP_ESPINUDP: | |
1804 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
1805 | up->encap_rcv = xfrm4_udp_encap_rcv; |
1806 | /* FALLTHROUGH */ | |
342f0234 | 1807 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 | 1808 | up->encap_type = val; |
447167bf | 1809 | udp_encap_enable(); |
1da177e4 LT |
1810 | break; |
1811 | default: | |
1812 | err = -ENOPROTOOPT; | |
1813 | break; | |
1814 | } | |
1815 | break; | |
1816 | ||
ba4e58ec GR |
1817 | /* |
1818 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
1819 | */ | |
1820 | /* The sender sets actual checksum coverage length via this option. | |
1821 | * The case coverage > packet length is handled by send module. */ | |
1822 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 1823 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1824 | return -ENOPROTOOPT; |
1825 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
1826 | val = 8; | |
4be929be AD |
1827 | else if (val > USHRT_MAX) |
1828 | val = USHRT_MAX; | |
ba4e58ec GR |
1829 | up->pcslen = val; |
1830 | up->pcflag |= UDPLITE_SEND_CC; | |
1831 | break; | |
1832 | ||
e905a9ed YH |
1833 | /* The receiver specifies a minimum checksum coverage value. To make |
1834 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
1835 | * used, this again means full checksum coverage. */ |
1836 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 1837 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1838 | return -ENOPROTOOPT; |
1839 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
1840 | val = 8; | |
4be929be AD |
1841 | else if (val > USHRT_MAX) |
1842 | val = USHRT_MAX; | |
ba4e58ec GR |
1843 | up->pcrlen = val; |
1844 | up->pcflag |= UDPLITE_RECV_CC; | |
1845 | break; | |
1846 | ||
1da177e4 LT |
1847 | default: |
1848 | err = -ENOPROTOOPT; | |
1849 | break; | |
6516c655 | 1850 | } |
1da177e4 LT |
1851 | |
1852 | return err; | |
1853 | } | |
c482c568 | 1854 | EXPORT_SYMBOL(udp_lib_setsockopt); |
1da177e4 | 1855 | |
db8dac20 | 1856 | int udp_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 1857 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
1858 | { |
1859 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1860 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1861 | udp_push_pending_frames); | |
1862 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
1863 | } | |
1864 | ||
1865 | #ifdef CONFIG_COMPAT | |
1866 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 1867 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
1868 | { |
1869 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1870 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1871 | udp_push_pending_frames); | |
1872 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
1873 | } | |
1874 | #endif | |
1875 | ||
4c0a6cb0 GR |
1876 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
1877 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
1878 | { |
1879 | struct udp_sock *up = udp_sk(sk); | |
1880 | int val, len; | |
1881 | ||
c482c568 | 1882 | if (get_user(len, optlen)) |
1da177e4 LT |
1883 | return -EFAULT; |
1884 | ||
1885 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 1886 | |
6516c655 | 1887 | if (len < 0) |
1da177e4 LT |
1888 | return -EINVAL; |
1889 | ||
6516c655 | 1890 | switch (optname) { |
1da177e4 LT |
1891 | case UDP_CORK: |
1892 | val = up->corkflag; | |
1893 | break; | |
1894 | ||
1895 | case UDP_ENCAP: | |
1896 | val = up->encap_type; | |
1897 | break; | |
1898 | ||
ba4e58ec GR |
1899 | /* The following two cannot be changed on UDP sockets, the return is |
1900 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
1901 | case UDPLITE_SEND_CSCOV: | |
1902 | val = up->pcslen; | |
1903 | break; | |
1904 | ||
1905 | case UDPLITE_RECV_CSCOV: | |
1906 | val = up->pcrlen; | |
1907 | break; | |
1908 | ||
1da177e4 LT |
1909 | default: |
1910 | return -ENOPROTOOPT; | |
6516c655 | 1911 | } |
1da177e4 | 1912 | |
6516c655 | 1913 | if (put_user(len, optlen)) |
e905a9ed | 1914 | return -EFAULT; |
c482c568 | 1915 | if (copy_to_user(optval, &val, len)) |
1da177e4 | 1916 | return -EFAULT; |
e905a9ed | 1917 | return 0; |
1da177e4 | 1918 | } |
c482c568 | 1919 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1da177e4 | 1920 | |
db8dac20 DM |
1921 | int udp_getsockopt(struct sock *sk, int level, int optname, |
1922 | char __user *optval, int __user *optlen) | |
1923 | { | |
1924 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1925 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1926 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
1927 | } | |
1928 | ||
1929 | #ifdef CONFIG_COMPAT | |
1930 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
1931 | char __user *optval, int __user *optlen) | |
1932 | { | |
1933 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1934 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1935 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
1936 | } | |
1937 | #endif | |
1da177e4 LT |
1938 | /** |
1939 | * udp_poll - wait for a UDP event. | |
1940 | * @file - file struct | |
1941 | * @sock - socket | |
1942 | * @wait - poll table | |
1943 | * | |
e905a9ed | 1944 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
1945 | * blocking sockets. If application is using a blocking fd |
1946 | * and a packet with checksum error is in the queue; | |
1947 | * then it could get return from select indicating data available | |
1948 | * but then block when reading it. Add special case code | |
1949 | * to work around these arguably broken applications. | |
1950 | */ | |
1951 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
1952 | { | |
1953 | unsigned int mask = datagram_poll(file, sock, wait); | |
1954 | struct sock *sk = sock->sk; | |
ba4e58ec | 1955 | |
1da177e4 | 1956 | /* Check for false positives due to checksum errors */ |
85584672 ED |
1957 | if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) && |
1958 | !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk)) | |
1959 | mask &= ~(POLLIN | POLLRDNORM); | |
1da177e4 LT |
1960 | |
1961 | return mask; | |
e905a9ed | 1962 | |
1da177e4 | 1963 | } |
c482c568 | 1964 | EXPORT_SYMBOL(udp_poll); |
1da177e4 | 1965 | |
db8dac20 DM |
1966 | struct proto udp_prot = { |
1967 | .name = "UDP", | |
1968 | .owner = THIS_MODULE, | |
1969 | .close = udp_lib_close, | |
1970 | .connect = ip4_datagram_connect, | |
1971 | .disconnect = udp_disconnect, | |
1972 | .ioctl = udp_ioctl, | |
1973 | .destroy = udp_destroy_sock, | |
1974 | .setsockopt = udp_setsockopt, | |
1975 | .getsockopt = udp_getsockopt, | |
1976 | .sendmsg = udp_sendmsg, | |
1977 | .recvmsg = udp_recvmsg, | |
1978 | .sendpage = udp_sendpage, | |
93821778 | 1979 | .backlog_rcv = __udp_queue_rcv_skb, |
8141ed9f | 1980 | .release_cb = ip4_datagram_release_cb, |
db8dac20 DM |
1981 | .hash = udp_lib_hash, |
1982 | .unhash = udp_lib_unhash, | |
719f8358 | 1983 | .rehash = udp_v4_rehash, |
db8dac20 DM |
1984 | .get_port = udp_v4_get_port, |
1985 | .memory_allocated = &udp_memory_allocated, | |
1986 | .sysctl_mem = sysctl_udp_mem, | |
1987 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
1988 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
1989 | .obj_size = sizeof(struct udp_sock), | |
271b72c7 | 1990 | .slab_flags = SLAB_DESTROY_BY_RCU, |
645ca708 | 1991 | .h.udp_table = &udp_table, |
db8dac20 DM |
1992 | #ifdef CONFIG_COMPAT |
1993 | .compat_setsockopt = compat_udp_setsockopt, | |
1994 | .compat_getsockopt = compat_udp_getsockopt, | |
1995 | #endif | |
fcbdf09d | 1996 | .clear_sk = sk_prot_clear_portaddr_nulls, |
db8dac20 | 1997 | }; |
c482c568 | 1998 | EXPORT_SYMBOL(udp_prot); |
1da177e4 LT |
1999 | |
2000 | /* ------------------------------------------------------------------------ */ | |
2001 | #ifdef CONFIG_PROC_FS | |
2002 | ||
645ca708 | 2003 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
2004 | { |
2005 | struct sock *sk; | |
2006 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2007 | struct net *net = seq_file_net(seq); |
1da177e4 | 2008 | |
f86dcc5a ED |
2009 | for (state->bucket = start; state->bucket <= state->udp_table->mask; |
2010 | ++state->bucket) { | |
88ab1932 | 2011 | struct hlist_nulls_node *node; |
645ca708 | 2012 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
f86dcc5a ED |
2013 | |
2014 | if (hlist_nulls_empty(&hslot->head)) | |
2015 | continue; | |
2016 | ||
645ca708 | 2017 | spin_lock_bh(&hslot->lock); |
88ab1932 | 2018 | sk_nulls_for_each(sk, node, &hslot->head) { |
878628fb | 2019 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 2020 | continue; |
1da177e4 LT |
2021 | if (sk->sk_family == state->family) |
2022 | goto found; | |
2023 | } | |
645ca708 | 2024 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
2025 | } |
2026 | sk = NULL; | |
2027 | found: | |
2028 | return sk; | |
2029 | } | |
2030 | ||
2031 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
2032 | { | |
2033 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2034 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2035 | |
2036 | do { | |
88ab1932 | 2037 | sk = sk_nulls_next(sk); |
878628fb | 2038 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 2039 | |
645ca708 | 2040 | if (!sk) { |
f86dcc5a | 2041 | if (state->bucket <= state->udp_table->mask) |
30842f29 | 2042 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
645ca708 | 2043 | return udp_get_first(seq, state->bucket + 1); |
1da177e4 LT |
2044 | } |
2045 | return sk; | |
2046 | } | |
2047 | ||
2048 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
2049 | { | |
645ca708 | 2050 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
2051 | |
2052 | if (sk) | |
6516c655 | 2053 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
2054 | --pos; |
2055 | return pos ? NULL : sk; | |
2056 | } | |
2057 | ||
2058 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
2059 | { | |
30842f29 | 2060 | struct udp_iter_state *state = seq->private; |
f86dcc5a | 2061 | state->bucket = MAX_UDP_PORTS; |
30842f29 | 2062 | |
b50660f1 | 2063 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
2064 | } |
2065 | ||
2066 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2067 | { | |
2068 | struct sock *sk; | |
2069 | ||
b50660f1 | 2070 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2071 | sk = udp_get_idx(seq, 0); |
2072 | else | |
2073 | sk = udp_get_next(seq, v); | |
2074 | ||
2075 | ++*pos; | |
2076 | return sk; | |
2077 | } | |
2078 | ||
2079 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
2080 | { | |
645ca708 ED |
2081 | struct udp_iter_state *state = seq->private; |
2082 | ||
f86dcc5a | 2083 | if (state->bucket <= state->udp_table->mask) |
645ca708 | 2084 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
1da177e4 LT |
2085 | } |
2086 | ||
73cb88ec | 2087 | int udp_seq_open(struct inode *inode, struct file *file) |
1da177e4 LT |
2088 | { |
2089 | struct udp_seq_afinfo *afinfo = PDE(inode)->data; | |
a2be75c1 DL |
2090 | struct udp_iter_state *s; |
2091 | int err; | |
a91275ef | 2092 | |
a2be75c1 DL |
2093 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2094 | sizeof(struct udp_iter_state)); | |
2095 | if (err < 0) | |
2096 | return err; | |
a91275ef | 2097 | |
a2be75c1 | 2098 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2099 | s->family = afinfo->family; |
645ca708 | 2100 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 2101 | return err; |
a91275ef | 2102 | } |
73cb88ec | 2103 | EXPORT_SYMBOL(udp_seq_open); |
a91275ef | 2104 | |
1da177e4 | 2105 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 2106 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
2107 | { |
2108 | struct proc_dir_entry *p; | |
2109 | int rc = 0; | |
2110 | ||
dda61925 DL |
2111 | afinfo->seq_ops.start = udp_seq_start; |
2112 | afinfo->seq_ops.next = udp_seq_next; | |
2113 | afinfo->seq_ops.stop = udp_seq_stop; | |
2114 | ||
84841c3c | 2115 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2116 | afinfo->seq_fops, afinfo); |
84841c3c | 2117 | if (!p) |
1da177e4 LT |
2118 | rc = -ENOMEM; |
2119 | return rc; | |
2120 | } | |
c482c568 | 2121 | EXPORT_SYMBOL(udp_proc_register); |
1da177e4 | 2122 | |
0c96d8c5 | 2123 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 2124 | { |
ece31ffd | 2125 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2126 | } |
c482c568 | 2127 | EXPORT_SYMBOL(udp_proc_unregister); |
db8dac20 DM |
2128 | |
2129 | /* ------------------------------------------------------------------------ */ | |
5e659e4c PE |
2130 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
2131 | int bucket, int *len) | |
db8dac20 DM |
2132 | { |
2133 | struct inet_sock *inet = inet_sk(sp); | |
c720c7e8 ED |
2134 | __be32 dest = inet->inet_daddr; |
2135 | __be32 src = inet->inet_rcv_saddr; | |
2136 | __u16 destp = ntohs(inet->inet_dport); | |
2137 | __u16 srcp = ntohs(inet->inet_sport); | |
db8dac20 | 2138 | |
f86dcc5a | 2139 | seq_printf(f, "%5d: %08X:%04X %08X:%04X" |
71338aa7 | 2140 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n", |
db8dac20 | 2141 | bucket, src, srcp, dest, destp, sp->sk_state, |
31e6d363 ED |
2142 | sk_wmem_alloc_get(sp), |
2143 | sk_rmem_alloc_get(sp), | |
a7cb5a49 EB |
2144 | 0, 0L, 0, |
2145 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)), | |
2146 | 0, sock_i_ino(sp), | |
cb61cb9b ED |
2147 | atomic_read(&sp->sk_refcnt), sp, |
2148 | atomic_read(&sp->sk_drops), len); | |
db8dac20 DM |
2149 | } |
2150 | ||
2151 | int udp4_seq_show(struct seq_file *seq, void *v) | |
2152 | { | |
2153 | if (v == SEQ_START_TOKEN) | |
2154 | seq_printf(seq, "%-127s\n", | |
2155 | " sl local_address rem_address st tx_queue " | |
2156 | "rx_queue tr tm->when retrnsmt uid timeout " | |
cb61cb9b | 2157 | "inode ref pointer drops"); |
db8dac20 | 2158 | else { |
db8dac20 | 2159 | struct udp_iter_state *state = seq->private; |
5e659e4c | 2160 | int len; |
db8dac20 | 2161 | |
5e659e4c | 2162 | udp4_format_sock(v, seq, state->bucket, &len); |
c482c568 | 2163 | seq_printf(seq, "%*s\n", 127 - len, ""); |
db8dac20 DM |
2164 | } |
2165 | return 0; | |
2166 | } | |
2167 | ||
73cb88ec AV |
2168 | static const struct file_operations udp_afinfo_seq_fops = { |
2169 | .owner = THIS_MODULE, | |
2170 | .open = udp_seq_open, | |
2171 | .read = seq_read, | |
2172 | .llseek = seq_lseek, | |
2173 | .release = seq_release_net | |
2174 | }; | |
2175 | ||
db8dac20 | 2176 | /* ------------------------------------------------------------------------ */ |
db8dac20 | 2177 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
2178 | .name = "udp", |
2179 | .family = AF_INET, | |
645ca708 | 2180 | .udp_table = &udp_table, |
73cb88ec | 2181 | .seq_fops = &udp_afinfo_seq_fops, |
dda61925 DL |
2182 | .seq_ops = { |
2183 | .show = udp4_seq_show, | |
2184 | }, | |
db8dac20 DM |
2185 | }; |
2186 | ||
2c8c1e72 | 2187 | static int __net_init udp4_proc_init_net(struct net *net) |
15439feb PE |
2188 | { |
2189 | return udp_proc_register(net, &udp4_seq_afinfo); | |
2190 | } | |
2191 | ||
2c8c1e72 | 2192 | static void __net_exit udp4_proc_exit_net(struct net *net) |
15439feb PE |
2193 | { |
2194 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
2195 | } | |
2196 | ||
2197 | static struct pernet_operations udp4_net_ops = { | |
2198 | .init = udp4_proc_init_net, | |
2199 | .exit = udp4_proc_exit_net, | |
2200 | }; | |
2201 | ||
db8dac20 DM |
2202 | int __init udp4_proc_init(void) |
2203 | { | |
15439feb | 2204 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
2205 | } |
2206 | ||
2207 | void udp4_proc_exit(void) | |
2208 | { | |
15439feb | 2209 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 2210 | } |
1da177e4 LT |
2211 | #endif /* CONFIG_PROC_FS */ |
2212 | ||
f86dcc5a ED |
2213 | static __initdata unsigned long uhash_entries; |
2214 | static int __init set_uhash_entries(char *str) | |
645ca708 | 2215 | { |
413c27d8 EZ |
2216 | ssize_t ret; |
2217 | ||
f86dcc5a ED |
2218 | if (!str) |
2219 | return 0; | |
413c27d8 EZ |
2220 | |
2221 | ret = kstrtoul(str, 0, &uhash_entries); | |
2222 | if (ret) | |
2223 | return 0; | |
2224 | ||
f86dcc5a ED |
2225 | if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN) |
2226 | uhash_entries = UDP_HTABLE_SIZE_MIN; | |
2227 | return 1; | |
2228 | } | |
2229 | __setup("uhash_entries=", set_uhash_entries); | |
645ca708 | 2230 | |
f86dcc5a ED |
2231 | void __init udp_table_init(struct udp_table *table, const char *name) |
2232 | { | |
2233 | unsigned int i; | |
2234 | ||
31fe62b9 TB |
2235 | table->hash = alloc_large_system_hash(name, |
2236 | 2 * sizeof(struct udp_hslot), | |
2237 | uhash_entries, | |
2238 | 21, /* one slot per 2 MB */ | |
2239 | 0, | |
2240 | &table->log, | |
2241 | &table->mask, | |
2242 | UDP_HTABLE_SIZE_MIN, | |
2243 | 64 * 1024); | |
2244 | ||
512615b6 | 2245 | table->hash2 = table->hash + (table->mask + 1); |
f86dcc5a | 2246 | for (i = 0; i <= table->mask; i++) { |
88ab1932 | 2247 | INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i); |
fdcc8aa9 | 2248 | table->hash[i].count = 0; |
645ca708 ED |
2249 | spin_lock_init(&table->hash[i].lock); |
2250 | } | |
512615b6 ED |
2251 | for (i = 0; i <= table->mask; i++) { |
2252 | INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i); | |
2253 | table->hash2[i].count = 0; | |
2254 | spin_lock_init(&table->hash2[i].lock); | |
2255 | } | |
645ca708 ED |
2256 | } |
2257 | ||
95766fff HA |
2258 | void __init udp_init(void) |
2259 | { | |
f03d78db | 2260 | unsigned long limit; |
95766fff | 2261 | |
f86dcc5a | 2262 | udp_table_init(&udp_table, "UDP"); |
f03d78db | 2263 | limit = nr_free_buffer_pages() / 8; |
95766fff HA |
2264 | limit = max(limit, 128UL); |
2265 | sysctl_udp_mem[0] = limit / 4 * 3; | |
2266 | sysctl_udp_mem[1] = limit; | |
2267 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
2268 | ||
2269 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
2270 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
2271 | } | |
2272 | ||
d7ca4cc0 SS |
2273 | int udp4_ufo_send_check(struct sk_buff *skb) |
2274 | { | |
2275 | const struct iphdr *iph; | |
2276 | struct udphdr *uh; | |
2277 | ||
2278 | if (!pskb_may_pull(skb, sizeof(*uh))) | |
2279 | return -EINVAL; | |
2280 | ||
2281 | iph = ip_hdr(skb); | |
2282 | uh = udp_hdr(skb); | |
2283 | ||
2284 | uh->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, | |
2285 | IPPROTO_UDP, 0); | |
2286 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
2287 | skb->csum_offset = offsetof(struct udphdr, check); | |
2288 | skb->ip_summed = CHECKSUM_PARTIAL; | |
2289 | return 0; | |
2290 | } | |
2291 | ||
c8f44aff MM |
2292 | struct sk_buff *udp4_ufo_fragment(struct sk_buff *skb, |
2293 | netdev_features_t features) | |
d7ca4cc0 SS |
2294 | { |
2295 | struct sk_buff *segs = ERR_PTR(-EINVAL); | |
2296 | unsigned int mss; | |
2297 | int offset; | |
2298 | __wsum csum; | |
2299 | ||
2300 | mss = skb_shinfo(skb)->gso_size; | |
2301 | if (unlikely(skb->len <= mss)) | |
2302 | goto out; | |
2303 | ||
2304 | if (skb_gso_ok(skb, features | NETIF_F_GSO_ROBUST)) { | |
2305 | /* Packet is from an untrusted source, reset gso_segs. */ | |
2306 | int type = skb_shinfo(skb)->gso_type; | |
2307 | ||
68c33163 PS |
2308 | if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY | |
2309 | SKB_GSO_GRE) || | |
d7ca4cc0 SS |
2310 | !(type & (SKB_GSO_UDP)))) |
2311 | goto out; | |
2312 | ||
2313 | skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss); | |
2314 | ||
2315 | segs = NULL; | |
2316 | goto out; | |
2317 | } | |
2318 | ||
2319 | /* Do software UFO. Complete and fill in the UDP checksum as HW cannot | |
2320 | * do checksum of UDP packets sent as multiple IP fragments. | |
2321 | */ | |
55508d60 | 2322 | offset = skb_checksum_start_offset(skb); |
c482c568 | 2323 | csum = skb_checksum(skb, offset, skb->len - offset, 0); |
d7ca4cc0 SS |
2324 | offset += skb->csum_offset; |
2325 | *(__sum16 *)(skb->data + offset) = csum_fold(csum); | |
2326 | skb->ip_summed = CHECKSUM_NONE; | |
2327 | ||
2328 | /* Fragment the skb. IP headers of the fragments are updated in | |
2329 | * inet_gso_segment() | |
2330 | */ | |
2331 | segs = skb_segment(skb, features); | |
2332 | out: | |
2333 | return segs; | |
2334 | } | |
2335 |