]>
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> |
ba4e58ec | 110 | #include "udp_impl.h" |
1da177e4 | 111 | |
f86dcc5a | 112 | struct udp_table udp_table __read_mostly; |
645ca708 | 113 | EXPORT_SYMBOL(udp_table); |
1da177e4 | 114 | |
8d987e5c | 115 | long sysctl_udp_mem[3] __read_mostly; |
95766fff | 116 | EXPORT_SYMBOL(sysctl_udp_mem); |
c482c568 ED |
117 | |
118 | int sysctl_udp_rmem_min __read_mostly; | |
95766fff | 119 | EXPORT_SYMBOL(sysctl_udp_rmem_min); |
c482c568 ED |
120 | |
121 | int sysctl_udp_wmem_min __read_mostly; | |
95766fff HA |
122 | EXPORT_SYMBOL(sysctl_udp_wmem_min); |
123 | ||
8d987e5c | 124 | atomic_long_t udp_memory_allocated; |
95766fff HA |
125 | EXPORT_SYMBOL(udp_memory_allocated); |
126 | ||
f86dcc5a ED |
127 | #define MAX_UDP_PORTS 65536 |
128 | #define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN) | |
98322f22 | 129 | |
f24d43c0 | 130 | static int udp_lib_lport_inuse(struct net *net, __u16 num, |
645ca708 | 131 | const struct udp_hslot *hslot, |
98322f22 | 132 | unsigned long *bitmap, |
f24d43c0 ED |
133 | struct sock *sk, |
134 | int (*saddr_comp)(const struct sock *sk1, | |
f86dcc5a ED |
135 | const struct sock *sk2), |
136 | unsigned int log) | |
1da177e4 | 137 | { |
f24d43c0 | 138 | struct sock *sk2; |
88ab1932 | 139 | struct hlist_nulls_node *node; |
25030a7f | 140 | |
88ab1932 | 141 | sk_nulls_for_each(sk2, node, &hslot->head) |
9d4fb27d JP |
142 | if (net_eq(sock_net(sk2), net) && |
143 | sk2 != sk && | |
d4cada4a | 144 | (bitmap || udp_sk(sk2)->udp_port_hash == num) && |
9d4fb27d JP |
145 | (!sk2->sk_reuse || !sk->sk_reuse) && |
146 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
147 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
98322f22 ED |
148 | (*saddr_comp)(sk, sk2)) { |
149 | if (bitmap) | |
d4cada4a ED |
150 | __set_bit(udp_sk(sk2)->udp_port_hash >> log, |
151 | bitmap); | |
98322f22 ED |
152 | else |
153 | return 1; | |
154 | } | |
25030a7f GR |
155 | return 0; |
156 | } | |
157 | ||
30fff923 ED |
158 | /* |
159 | * Note: we still hold spinlock of primary hash chain, so no other writer | |
160 | * can insert/delete a socket with local_port == num | |
161 | */ | |
162 | static int udp_lib_lport_inuse2(struct net *net, __u16 num, | |
163 | struct udp_hslot *hslot2, | |
164 | struct sock *sk, | |
165 | int (*saddr_comp)(const struct sock *sk1, | |
166 | const struct sock *sk2)) | |
167 | { | |
168 | struct sock *sk2; | |
169 | struct hlist_nulls_node *node; | |
170 | int res = 0; | |
171 | ||
172 | spin_lock(&hslot2->lock); | |
173 | udp_portaddr_for_each_entry(sk2, node, &hslot2->head) | |
9d4fb27d JP |
174 | if (net_eq(sock_net(sk2), net) && |
175 | sk2 != sk && | |
176 | (udp_sk(sk2)->udp_port_hash == num) && | |
177 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
178 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
179 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
30fff923 ED |
180 | (*saddr_comp)(sk, sk2)) { |
181 | res = 1; | |
182 | break; | |
183 | } | |
184 | spin_unlock(&hslot2->lock); | |
185 | return res; | |
186 | } | |
187 | ||
25030a7f | 188 | /** |
6ba5a3c5 | 189 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
190 | * |
191 | * @sk: socket struct in question | |
192 | * @snum: port number to look up | |
df2bc459 | 193 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25985edc | 194 | * @hash2_nulladdr: AF-dependent hash value in secondary hash chains, |
30fff923 | 195 | * with NULL address |
25030a7f | 196 | */ |
6ba5a3c5 | 197 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
df2bc459 | 198 | int (*saddr_comp)(const struct sock *sk1, |
30fff923 ED |
199 | const struct sock *sk2), |
200 | unsigned int hash2_nulladdr) | |
25030a7f | 201 | { |
512615b6 | 202 | struct udp_hslot *hslot, *hslot2; |
645ca708 | 203 | struct udp_table *udptable = sk->sk_prot->h.udp_table; |
25030a7f | 204 | int error = 1; |
3b1e0a65 | 205 | struct net *net = sock_net(sk); |
1da177e4 | 206 | |
32c1da70 | 207 | if (!snum) { |
9088c560 ED |
208 | int low, high, remaining; |
209 | unsigned rand; | |
98322f22 ED |
210 | unsigned short first, last; |
211 | DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN); | |
32c1da70 | 212 | |
227b60f5 | 213 | inet_get_local_port_range(&low, &high); |
a25de534 | 214 | remaining = (high - low) + 1; |
227b60f5 | 215 | |
9088c560 | 216 | rand = net_random(); |
98322f22 ED |
217 | first = (((u64)rand * remaining) >> 32) + low; |
218 | /* | |
219 | * force rand to be an odd multiple of UDP_HTABLE_SIZE | |
220 | */ | |
f86dcc5a | 221 | rand = (rand | 1) * (udptable->mask + 1); |
5781b235 ED |
222 | last = first + udptable->mask + 1; |
223 | do { | |
f86dcc5a | 224 | hslot = udp_hashslot(udptable, net, first); |
98322f22 | 225 | bitmap_zero(bitmap, PORTS_PER_CHAIN); |
645ca708 | 226 | spin_lock_bh(&hslot->lock); |
98322f22 | 227 | udp_lib_lport_inuse(net, snum, hslot, bitmap, sk, |
f86dcc5a | 228 | saddr_comp, udptable->log); |
98322f22 ED |
229 | |
230 | snum = first; | |
231 | /* | |
232 | * Iterate on all possible values of snum for this hash. | |
233 | * Using steps of an odd multiple of UDP_HTABLE_SIZE | |
234 | * give us randomization and full range coverage. | |
235 | */ | |
9088c560 | 236 | do { |
98322f22 | 237 | if (low <= snum && snum <= high && |
e3826f1e AW |
238 | !test_bit(snum >> udptable->log, bitmap) && |
239 | !inet_is_reserved_local_port(snum)) | |
98322f22 ED |
240 | goto found; |
241 | snum += rand; | |
242 | } while (snum != first); | |
243 | spin_unlock_bh(&hslot->lock); | |
5781b235 | 244 | } while (++first != last); |
98322f22 | 245 | goto fail; |
645ca708 | 246 | } else { |
f86dcc5a | 247 | hslot = udp_hashslot(udptable, net, snum); |
645ca708 | 248 | spin_lock_bh(&hslot->lock); |
30fff923 ED |
249 | if (hslot->count > 10) { |
250 | int exist; | |
251 | unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum; | |
252 | ||
253 | slot2 &= udptable->mask; | |
254 | hash2_nulladdr &= udptable->mask; | |
255 | ||
256 | hslot2 = udp_hashslot2(udptable, slot2); | |
257 | if (hslot->count < hslot2->count) | |
258 | goto scan_primary_hash; | |
259 | ||
260 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
261 | sk, saddr_comp); | |
262 | if (!exist && (hash2_nulladdr != slot2)) { | |
263 | hslot2 = udp_hashslot2(udptable, hash2_nulladdr); | |
264 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
265 | sk, saddr_comp); | |
266 | } | |
267 | if (exist) | |
268 | goto fail_unlock; | |
269 | else | |
270 | goto found; | |
271 | } | |
272 | scan_primary_hash: | |
f86dcc5a ED |
273 | if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk, |
274 | saddr_comp, 0)) | |
645ca708 ED |
275 | goto fail_unlock; |
276 | } | |
98322f22 | 277 | found: |
c720c7e8 | 278 | inet_sk(sk)->inet_num = snum; |
d4cada4a ED |
279 | udp_sk(sk)->udp_port_hash = snum; |
280 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
1da177e4 | 281 | if (sk_unhashed(sk)) { |
88ab1932 | 282 | sk_nulls_add_node_rcu(sk, &hslot->head); |
fdcc8aa9 | 283 | hslot->count++; |
c29a0bc4 | 284 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
512615b6 ED |
285 | |
286 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
287 | spin_lock(&hslot2->lock); | |
288 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
289 | &hslot2->head); | |
290 | hslot2->count++; | |
291 | spin_unlock(&hslot2->lock); | |
1da177e4 | 292 | } |
25030a7f | 293 | error = 0; |
645ca708 ED |
294 | fail_unlock: |
295 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 296 | fail: |
25030a7f GR |
297 | return error; |
298 | } | |
c482c568 | 299 | EXPORT_SYMBOL(udp_lib_get_port); |
25030a7f | 300 | |
499923c7 | 301 | static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2) |
db8dac20 DM |
302 | { |
303 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
304 | ||
c482c568 | 305 | return (!ipv6_only_sock(sk2) && |
c720c7e8 ED |
306 | (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr || |
307 | inet1->inet_rcv_saddr == inet2->inet_rcv_saddr)); | |
db8dac20 DM |
308 | } |
309 | ||
d4cada4a ED |
310 | static unsigned int udp4_portaddr_hash(struct net *net, __be32 saddr, |
311 | unsigned int port) | |
312 | { | |
0eae88f3 | 313 | return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port; |
d4cada4a ED |
314 | } |
315 | ||
6ba5a3c5 | 316 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 317 | { |
30fff923 | 318 | unsigned int hash2_nulladdr = |
0eae88f3 | 319 | udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum); |
30fff923 ED |
320 | unsigned int hash2_partial = |
321 | udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0); | |
322 | ||
d4cada4a | 323 | /* precompute partial secondary hash */ |
30fff923 ED |
324 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
325 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr); | |
db8dac20 DM |
326 | } |
327 | ||
645ca708 ED |
328 | static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr, |
329 | unsigned short hnum, | |
330 | __be16 sport, __be32 daddr, __be16 dport, int dif) | |
331 | { | |
332 | int score = -1; | |
333 | ||
d4cada4a | 334 | if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum && |
645ca708 ED |
335 | !ipv6_only_sock(sk)) { |
336 | struct inet_sock *inet = inet_sk(sk); | |
337 | ||
338 | score = (sk->sk_family == PF_INET ? 1 : 0); | |
c720c7e8 ED |
339 | if (inet->inet_rcv_saddr) { |
340 | if (inet->inet_rcv_saddr != daddr) | |
645ca708 ED |
341 | return -1; |
342 | score += 2; | |
343 | } | |
c720c7e8 ED |
344 | if (inet->inet_daddr) { |
345 | if (inet->inet_daddr != saddr) | |
645ca708 ED |
346 | return -1; |
347 | score += 2; | |
348 | } | |
c720c7e8 ED |
349 | if (inet->inet_dport) { |
350 | if (inet->inet_dport != sport) | |
645ca708 ED |
351 | return -1; |
352 | score += 2; | |
353 | } | |
354 | if (sk->sk_bound_dev_if) { | |
355 | if (sk->sk_bound_dev_if != dif) | |
356 | return -1; | |
357 | score += 2; | |
358 | } | |
359 | } | |
360 | return score; | |
361 | } | |
362 | ||
5051ebd2 ED |
363 | /* |
364 | * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num) | |
365 | */ | |
366 | #define SCORE2_MAX (1 + 2 + 2 + 2) | |
367 | static inline int compute_score2(struct sock *sk, struct net *net, | |
368 | __be32 saddr, __be16 sport, | |
369 | __be32 daddr, unsigned int hnum, int dif) | |
370 | { | |
371 | int score = -1; | |
372 | ||
373 | if (net_eq(sock_net(sk), net) && !ipv6_only_sock(sk)) { | |
374 | struct inet_sock *inet = inet_sk(sk); | |
375 | ||
376 | if (inet->inet_rcv_saddr != daddr) | |
377 | return -1; | |
378 | if (inet->inet_num != hnum) | |
379 | return -1; | |
380 | ||
381 | score = (sk->sk_family == PF_INET ? 1 : 0); | |
382 | if (inet->inet_daddr) { | |
383 | if (inet->inet_daddr != saddr) | |
384 | return -1; | |
385 | score += 2; | |
386 | } | |
387 | if (inet->inet_dport) { | |
388 | if (inet->inet_dport != sport) | |
389 | return -1; | |
390 | score += 2; | |
391 | } | |
392 | if (sk->sk_bound_dev_if) { | |
393 | if (sk->sk_bound_dev_if != dif) | |
394 | return -1; | |
395 | score += 2; | |
396 | } | |
397 | } | |
398 | return score; | |
399 | } | |
400 | ||
5051ebd2 ED |
401 | |
402 | /* called with read_rcu_lock() */ | |
403 | static struct sock *udp4_lib_lookup2(struct net *net, | |
404 | __be32 saddr, __be16 sport, | |
405 | __be32 daddr, unsigned int hnum, int dif, | |
406 | struct udp_hslot *hslot2, unsigned int slot2) | |
407 | { | |
408 | struct sock *sk, *result; | |
409 | struct hlist_nulls_node *node; | |
410 | int score, badness; | |
411 | ||
412 | begin: | |
413 | result = NULL; | |
414 | badness = -1; | |
415 | udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { | |
416 | score = compute_score2(sk, net, saddr, sport, | |
417 | daddr, hnum, dif); | |
418 | if (score > badness) { | |
419 | result = sk; | |
420 | badness = score; | |
421 | if (score == SCORE2_MAX) | |
422 | goto exact_match; | |
423 | } | |
424 | } | |
425 | /* | |
426 | * if the nulls value we got at the end of this lookup is | |
427 | * not the expected one, we must restart lookup. | |
428 | * We probably met an item that was moved to another chain. | |
429 | */ | |
430 | if (get_nulls_value(node) != slot2) | |
431 | goto begin; | |
432 | ||
433 | if (result) { | |
434 | exact_match: | |
c31504dc | 435 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
5051ebd2 ED |
436 | result = NULL; |
437 | else if (unlikely(compute_score2(result, net, saddr, sport, | |
438 | daddr, hnum, dif) < badness)) { | |
439 | sock_put(result); | |
440 | goto begin; | |
441 | } | |
442 | } | |
443 | return result; | |
444 | } | |
445 | ||
db8dac20 DM |
446 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
447 | * harder than this. -DaveM | |
448 | */ | |
fce82338 | 449 | struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, |
db8dac20 | 450 | __be16 sport, __be32 daddr, __be16 dport, |
645ca708 | 451 | int dif, struct udp_table *udptable) |
db8dac20 | 452 | { |
271b72c7 | 453 | struct sock *sk, *result; |
88ab1932 | 454 | struct hlist_nulls_node *node; |
db8dac20 | 455 | unsigned short hnum = ntohs(dport); |
5051ebd2 ED |
456 | unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); |
457 | struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; | |
271b72c7 | 458 | int score, badness; |
645ca708 | 459 | |
271b72c7 | 460 | rcu_read_lock(); |
5051ebd2 ED |
461 | if (hslot->count > 10) { |
462 | hash2 = udp4_portaddr_hash(net, daddr, hnum); | |
463 | slot2 = hash2 & udptable->mask; | |
464 | hslot2 = &udptable->hash2[slot2]; | |
465 | if (hslot->count < hslot2->count) | |
466 | goto begin; | |
467 | ||
468 | result = udp4_lib_lookup2(net, saddr, sport, | |
469 | daddr, hnum, dif, | |
470 | hslot2, slot2); | |
471 | if (!result) { | |
0eae88f3 | 472 | hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum); |
5051ebd2 ED |
473 | slot2 = hash2 & udptable->mask; |
474 | hslot2 = &udptable->hash2[slot2]; | |
475 | if (hslot->count < hslot2->count) | |
476 | goto begin; | |
477 | ||
1223c67c | 478 | result = udp4_lib_lookup2(net, saddr, sport, |
0eae88f3 | 479 | htonl(INADDR_ANY), hnum, dif, |
5051ebd2 ED |
480 | hslot2, slot2); |
481 | } | |
482 | rcu_read_unlock(); | |
483 | return result; | |
484 | } | |
271b72c7 ED |
485 | begin: |
486 | result = NULL; | |
487 | badness = -1; | |
88ab1932 | 488 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { |
645ca708 ED |
489 | score = compute_score(sk, net, saddr, hnum, sport, |
490 | daddr, dport, dif); | |
491 | if (score > badness) { | |
492 | result = sk; | |
493 | badness = score; | |
db8dac20 DM |
494 | } |
495 | } | |
88ab1932 ED |
496 | /* |
497 | * if the nulls value we got at the end of this lookup is | |
498 | * not the expected one, we must restart lookup. | |
499 | * We probably met an item that was moved to another chain. | |
500 | */ | |
5051ebd2 | 501 | if (get_nulls_value(node) != slot) |
88ab1932 ED |
502 | goto begin; |
503 | ||
271b72c7 | 504 | if (result) { |
c31504dc | 505 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
271b72c7 ED |
506 | result = NULL; |
507 | else if (unlikely(compute_score(result, net, saddr, hnum, sport, | |
508 | daddr, dport, dif) < badness)) { | |
509 | sock_put(result); | |
510 | goto begin; | |
511 | } | |
512 | } | |
513 | rcu_read_unlock(); | |
db8dac20 DM |
514 | return result; |
515 | } | |
fce82338 | 516 | EXPORT_SYMBOL_GPL(__udp4_lib_lookup); |
db8dac20 | 517 | |
607c4aaf KK |
518 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
519 | __be16 sport, __be16 dport, | |
645ca708 | 520 | struct udp_table *udptable) |
607c4aaf | 521 | { |
23542618 | 522 | struct sock *sk; |
607c4aaf KK |
523 | const struct iphdr *iph = ip_hdr(skb); |
524 | ||
23542618 KK |
525 | if (unlikely(sk = skb_steal_sock(skb))) |
526 | return sk; | |
527 | else | |
adf30907 | 528 | return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport, |
23542618 KK |
529 | iph->daddr, dport, inet_iif(skb), |
530 | udptable); | |
607c4aaf KK |
531 | } |
532 | ||
bcd41303 KK |
533 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
534 | __be32 daddr, __be16 dport, int dif) | |
535 | { | |
645ca708 | 536 | return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table); |
bcd41303 KK |
537 | } |
538 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
539 | ||
920a4611 | 540 | static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk, |
db8dac20 DM |
541 | __be16 loc_port, __be32 loc_addr, |
542 | __be16 rmt_port, __be32 rmt_addr, | |
543 | int dif) | |
544 | { | |
88ab1932 | 545 | struct hlist_nulls_node *node; |
db8dac20 DM |
546 | struct sock *s = sk; |
547 | unsigned short hnum = ntohs(loc_port); | |
548 | ||
88ab1932 | 549 | sk_nulls_for_each_from(s, node) { |
db8dac20 DM |
550 | struct inet_sock *inet = inet_sk(s); |
551 | ||
9d4fb27d JP |
552 | if (!net_eq(sock_net(s), net) || |
553 | udp_sk(s)->udp_port_hash != hnum || | |
554 | (inet->inet_daddr && inet->inet_daddr != rmt_addr) || | |
555 | (inet->inet_dport != rmt_port && inet->inet_dport) || | |
556 | (inet->inet_rcv_saddr && | |
557 | inet->inet_rcv_saddr != loc_addr) || | |
558 | ipv6_only_sock(s) || | |
db8dac20 DM |
559 | (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)) |
560 | continue; | |
561 | if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif)) | |
562 | continue; | |
563 | goto found; | |
564 | } | |
565 | s = NULL; | |
566 | found: | |
567 | return s; | |
568 | } | |
569 | ||
570 | /* | |
571 | * This routine is called by the ICMP module when it gets some | |
572 | * sort of error condition. If err < 0 then the socket should | |
573 | * be closed and the error returned to the user. If err > 0 | |
574 | * it's just the icmp type << 8 | icmp code. | |
575 | * Header points to the ip header of the error packet. We move | |
576 | * on past this. Then (as it used to claim before adjustment) | |
577 | * header points to the first 8 bytes of the udp header. We need | |
578 | * to find the appropriate port. | |
579 | */ | |
580 | ||
645ca708 | 581 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
582 | { |
583 | struct inet_sock *inet; | |
b71d1d42 | 584 | const struct iphdr *iph = (const struct iphdr *)skb->data; |
c482c568 | 585 | struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2)); |
db8dac20 DM |
586 | const int type = icmp_hdr(skb)->type; |
587 | const int code = icmp_hdr(skb)->code; | |
588 | struct sock *sk; | |
589 | int harderr; | |
590 | int err; | |
fd54d716 | 591 | struct net *net = dev_net(skb->dev); |
db8dac20 | 592 | |
fd54d716 | 593 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
db8dac20 DM |
594 | iph->saddr, uh->source, skb->dev->ifindex, udptable); |
595 | if (sk == NULL) { | |
dcfc23ca | 596 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
597 | return; /* No socket for error */ |
598 | } | |
599 | ||
600 | err = 0; | |
601 | harderr = 0; | |
602 | inet = inet_sk(sk); | |
603 | ||
604 | switch (type) { | |
605 | default: | |
606 | case ICMP_TIME_EXCEEDED: | |
607 | err = EHOSTUNREACH; | |
608 | break; | |
609 | case ICMP_SOURCE_QUENCH: | |
610 | goto out; | |
611 | case ICMP_PARAMETERPROB: | |
612 | err = EPROTO; | |
613 | harderr = 1; | |
614 | break; | |
615 | case ICMP_DEST_UNREACH: | |
616 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
617 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { | |
618 | err = EMSGSIZE; | |
619 | harderr = 1; | |
620 | break; | |
621 | } | |
622 | goto out; | |
623 | } | |
624 | err = EHOSTUNREACH; | |
625 | if (code <= NR_ICMP_UNREACH) { | |
626 | harderr = icmp_err_convert[code].fatal; | |
627 | err = icmp_err_convert[code].errno; | |
628 | } | |
629 | break; | |
630 | } | |
631 | ||
632 | /* | |
633 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
634 | * 4.1.3.3. | |
635 | */ | |
636 | if (!inet->recverr) { | |
637 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
638 | goto out; | |
b1faf566 | 639 | } else |
c482c568 | 640 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1)); |
b1faf566 | 641 | |
db8dac20 DM |
642 | sk->sk_err = err; |
643 | sk->sk_error_report(sk); | |
644 | out: | |
645 | sock_put(sk); | |
646 | } | |
647 | ||
648 | void udp_err(struct sk_buff *skb, u32 info) | |
649 | { | |
645ca708 | 650 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
651 | } |
652 | ||
653 | /* | |
654 | * Throw away all pending data and cancel the corking. Socket is locked. | |
655 | */ | |
36d926b9 | 656 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
657 | { |
658 | struct udp_sock *up = udp_sk(sk); | |
659 | ||
660 | if (up->pending) { | |
661 | up->len = 0; | |
662 | up->pending = 0; | |
663 | ip_flush_pending_frames(sk); | |
664 | } | |
665 | } | |
36d926b9 | 666 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
667 | |
668 | /** | |
f6b9664f | 669 | * udp4_hwcsum - handle outgoing HW checksumming |
db8dac20 DM |
670 | * @skb: sk_buff containing the filled-in UDP header |
671 | * (checksum field must be zeroed out) | |
f6b9664f HX |
672 | * @src: source IP address |
673 | * @dst: destination IP address | |
db8dac20 | 674 | */ |
f6b9664f | 675 | static void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst) |
db8dac20 | 676 | { |
db8dac20 | 677 | struct udphdr *uh = udp_hdr(skb); |
f6b9664f HX |
678 | struct sk_buff *frags = skb_shinfo(skb)->frag_list; |
679 | int offset = skb_transport_offset(skb); | |
680 | int len = skb->len - offset; | |
681 | int hlen = len; | |
db8dac20 DM |
682 | __wsum csum = 0; |
683 | ||
f6b9664f | 684 | if (!frags) { |
db8dac20 DM |
685 | /* |
686 | * Only one fragment on the socket. | |
687 | */ | |
688 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
689 | skb->csum_offset = offsetof(struct udphdr, check); | |
f6b9664f HX |
690 | uh->check = ~csum_tcpudp_magic(src, dst, len, |
691 | IPPROTO_UDP, 0); | |
db8dac20 DM |
692 | } else { |
693 | /* | |
694 | * HW-checksum won't work as there are two or more | |
695 | * fragments on the socket so that all csums of sk_buffs | |
696 | * should be together | |
697 | */ | |
f6b9664f HX |
698 | do { |
699 | csum = csum_add(csum, frags->csum); | |
700 | hlen -= frags->len; | |
701 | } while ((frags = frags->next)); | |
db8dac20 | 702 | |
f6b9664f | 703 | csum = skb_checksum(skb, offset, hlen, csum); |
db8dac20 DM |
704 | skb->ip_summed = CHECKSUM_NONE; |
705 | ||
db8dac20 DM |
706 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); |
707 | if (uh->check == 0) | |
708 | uh->check = CSUM_MANGLED_0; | |
709 | } | |
710 | } | |
711 | ||
79ab0531 | 712 | static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4) |
db8dac20 | 713 | { |
f6b9664f | 714 | struct sock *sk = skb->sk; |
db8dac20 | 715 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
716 | struct udphdr *uh; |
717 | int err = 0; | |
718 | int is_udplite = IS_UDPLITE(sk); | |
f6b9664f HX |
719 | int offset = skb_transport_offset(skb); |
720 | int len = skb->len - offset; | |
db8dac20 DM |
721 | __wsum csum = 0; |
722 | ||
db8dac20 DM |
723 | /* |
724 | * Create a UDP header | |
725 | */ | |
726 | uh = udp_hdr(skb); | |
f6b9664f | 727 | uh->source = inet->inet_sport; |
79ab0531 | 728 | uh->dest = fl4->fl4_dport; |
f6b9664f | 729 | uh->len = htons(len); |
db8dac20 DM |
730 | uh->check = 0; |
731 | ||
732 | if (is_udplite) /* UDP-Lite */ | |
f6b9664f | 733 | csum = udplite_csum(skb); |
db8dac20 DM |
734 | |
735 | else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */ | |
736 | ||
737 | skb->ip_summed = CHECKSUM_NONE; | |
738 | goto send; | |
739 | ||
740 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
741 | ||
79ab0531 | 742 | udp4_hwcsum(skb, fl4->saddr, fl4->daddr); |
db8dac20 DM |
743 | goto send; |
744 | ||
f6b9664f HX |
745 | } else |
746 | csum = udp_csum(skb); | |
db8dac20 DM |
747 | |
748 | /* add protocol-dependent pseudo-header */ | |
79ab0531 | 749 | uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len, |
c482c568 | 750 | sk->sk_protocol, csum); |
db8dac20 DM |
751 | if (uh->check == 0) |
752 | uh->check = CSUM_MANGLED_0; | |
753 | ||
754 | send: | |
f6b9664f | 755 | err = ip_send_skb(skb); |
6ce9e7b5 ED |
756 | if (err) { |
757 | if (err == -ENOBUFS && !inet->recverr) { | |
758 | UDP_INC_STATS_USER(sock_net(sk), | |
759 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
760 | err = 0; | |
761 | } | |
762 | } else | |
763 | UDP_INC_STATS_USER(sock_net(sk), | |
764 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
f6b9664f HX |
765 | return err; |
766 | } | |
767 | ||
768 | /* | |
769 | * Push out all pending data as one UDP datagram. Socket is locked. | |
770 | */ | |
771 | static int udp_push_pending_frames(struct sock *sk) | |
772 | { | |
773 | struct udp_sock *up = udp_sk(sk); | |
774 | struct inet_sock *inet = inet_sk(sk); | |
b6f21b26 | 775 | struct flowi4 *fl4 = &inet->cork.fl.u.ip4; |
f6b9664f HX |
776 | struct sk_buff *skb; |
777 | int err = 0; | |
778 | ||
77968b78 | 779 | skb = ip_finish_skb(sk, fl4); |
f6b9664f HX |
780 | if (!skb) |
781 | goto out; | |
782 | ||
79ab0531 | 783 | err = udp_send_skb(skb, fl4); |
f6b9664f | 784 | |
db8dac20 DM |
785 | out: |
786 | up->len = 0; | |
787 | up->pending = 0; | |
db8dac20 DM |
788 | return err; |
789 | } | |
790 | ||
791 | int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
792 | size_t len) | |
793 | { | |
794 | struct inet_sock *inet = inet_sk(sk); | |
795 | struct udp_sock *up = udp_sk(sk); | |
e474995f | 796 | struct flowi4 fl4_stack; |
b6f21b26 | 797 | struct flowi4 *fl4; |
db8dac20 DM |
798 | int ulen = len; |
799 | struct ipcm_cookie ipc; | |
800 | struct rtable *rt = NULL; | |
801 | int free = 0; | |
802 | int connected = 0; | |
803 | __be32 daddr, faddr, saddr; | |
804 | __be16 dport; | |
805 | u8 tos; | |
806 | int err, is_udplite = IS_UDPLITE(sk); | |
807 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
808 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
903ab86d | 809 | struct sk_buff *skb; |
f6d8bd05 | 810 | struct ip_options_data opt_copy; |
db8dac20 DM |
811 | |
812 | if (len > 0xFFFF) | |
813 | return -EMSGSIZE; | |
814 | ||
815 | /* | |
816 | * Check the flags. | |
817 | */ | |
818 | ||
c482c568 | 819 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */ |
db8dac20 DM |
820 | return -EOPNOTSUPP; |
821 | ||
822 | ipc.opt = NULL; | |
2244d07b | 823 | ipc.tx_flags = 0; |
db8dac20 | 824 | |
903ab86d HX |
825 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
826 | ||
f5fca608 | 827 | fl4 = &inet->cork.fl.u.ip4; |
db8dac20 DM |
828 | if (up->pending) { |
829 | /* | |
830 | * There are pending frames. | |
831 | * The socket lock must be held while it's corked. | |
832 | */ | |
833 | lock_sock(sk); | |
834 | if (likely(up->pending)) { | |
835 | if (unlikely(up->pending != AF_INET)) { | |
836 | release_sock(sk); | |
837 | return -EINVAL; | |
838 | } | |
839 | goto do_append_data; | |
840 | } | |
841 | release_sock(sk); | |
842 | } | |
843 | ulen += sizeof(struct udphdr); | |
844 | ||
845 | /* | |
846 | * Get and verify the address. | |
847 | */ | |
848 | if (msg->msg_name) { | |
c482c568 | 849 | struct sockaddr_in * usin = (struct sockaddr_in *)msg->msg_name; |
db8dac20 DM |
850 | if (msg->msg_namelen < sizeof(*usin)) |
851 | return -EINVAL; | |
852 | if (usin->sin_family != AF_INET) { | |
853 | if (usin->sin_family != AF_UNSPEC) | |
854 | return -EAFNOSUPPORT; | |
855 | } | |
856 | ||
857 | daddr = usin->sin_addr.s_addr; | |
858 | dport = usin->sin_port; | |
859 | if (dport == 0) | |
860 | return -EINVAL; | |
861 | } else { | |
862 | if (sk->sk_state != TCP_ESTABLISHED) | |
863 | return -EDESTADDRREQ; | |
c720c7e8 ED |
864 | daddr = inet->inet_daddr; |
865 | dport = inet->inet_dport; | |
db8dac20 DM |
866 | /* Open fast path for connected socket. |
867 | Route will not be used, if at least one option is set. | |
868 | */ | |
869 | connected = 1; | |
870 | } | |
c720c7e8 | 871 | ipc.addr = inet->inet_saddr; |
db8dac20 DM |
872 | |
873 | ipc.oif = sk->sk_bound_dev_if; | |
2244d07b | 874 | err = sock_tx_timestamp(sk, &ipc.tx_flags); |
51f31cab PO |
875 | if (err) |
876 | return err; | |
db8dac20 | 877 | if (msg->msg_controllen) { |
3b1e0a65 | 878 | err = ip_cmsg_send(sock_net(sk), msg, &ipc); |
db8dac20 DM |
879 | if (err) |
880 | return err; | |
881 | if (ipc.opt) | |
882 | free = 1; | |
883 | connected = 0; | |
884 | } | |
f6d8bd05 ED |
885 | if (!ipc.opt) { |
886 | struct ip_options_rcu *inet_opt; | |
887 | ||
888 | rcu_read_lock(); | |
889 | inet_opt = rcu_dereference(inet->inet_opt); | |
890 | if (inet_opt) { | |
891 | memcpy(&opt_copy, inet_opt, | |
892 | sizeof(*inet_opt) + inet_opt->opt.optlen); | |
893 | ipc.opt = &opt_copy.opt; | |
894 | } | |
895 | rcu_read_unlock(); | |
896 | } | |
db8dac20 DM |
897 | |
898 | saddr = ipc.addr; | |
899 | ipc.addr = faddr = daddr; | |
900 | ||
f6d8bd05 | 901 | if (ipc.opt && ipc.opt->opt.srr) { |
db8dac20 DM |
902 | if (!daddr) |
903 | return -EINVAL; | |
f6d8bd05 | 904 | faddr = ipc.opt->opt.faddr; |
db8dac20 DM |
905 | connected = 0; |
906 | } | |
907 | tos = RT_TOS(inet->tos); | |
908 | if (sock_flag(sk, SOCK_LOCALROUTE) || | |
909 | (msg->msg_flags & MSG_DONTROUTE) || | |
f6d8bd05 | 910 | (ipc.opt && ipc.opt->opt.is_strictroute)) { |
db8dac20 DM |
911 | tos |= RTO_ONLINK; |
912 | connected = 0; | |
913 | } | |
914 | ||
915 | if (ipv4_is_multicast(daddr)) { | |
916 | if (!ipc.oif) | |
917 | ipc.oif = inet->mc_index; | |
918 | if (!saddr) | |
919 | saddr = inet->mc_addr; | |
920 | connected = 0; | |
76e21053 EH |
921 | } else if (!ipc.oif) |
922 | ipc.oif = inet->uc_index; | |
db8dac20 DM |
923 | |
924 | if (connected) | |
c482c568 | 925 | rt = (struct rtable *)sk_dst_check(sk, 0); |
db8dac20 DM |
926 | |
927 | if (rt == NULL) { | |
84a3aa00 PE |
928 | struct net *net = sock_net(sk); |
929 | ||
e474995f DM |
930 | fl4 = &fl4_stack; |
931 | flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos, | |
c0951cbc DM |
932 | RT_SCOPE_UNIVERSE, sk->sk_protocol, |
933 | inet_sk_flowi_flags(sk)|FLOWI_FLAG_CAN_SLEEP, | |
934 | faddr, saddr, dport, inet->inet_sport); | |
935 | ||
e474995f DM |
936 | security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); |
937 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe DM |
938 | if (IS_ERR(rt)) { |
939 | err = PTR_ERR(rt); | |
06dc94b1 | 940 | rt = NULL; |
db8dac20 | 941 | if (err == -ENETUNREACH) |
7c73a6fa | 942 | IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
943 | goto out; |
944 | } | |
945 | ||
946 | err = -EACCES; | |
947 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
948 | !sock_flag(sk, SOCK_BROADCAST)) | |
949 | goto out; | |
950 | if (connected) | |
d8d1f30b | 951 | sk_dst_set(sk, dst_clone(&rt->dst)); |
db8dac20 DM |
952 | } |
953 | ||
954 | if (msg->msg_flags&MSG_CONFIRM) | |
955 | goto do_confirm; | |
956 | back_from_confirm: | |
957 | ||
e474995f | 958 | saddr = fl4->saddr; |
db8dac20 | 959 | if (!ipc.addr) |
e474995f | 960 | daddr = ipc.addr = fl4->daddr; |
db8dac20 | 961 | |
903ab86d HX |
962 | /* Lockless fast path for the non-corking case. */ |
963 | if (!corkreq) { | |
77968b78 | 964 | skb = ip_make_skb(sk, fl4, getfrag, msg->msg_iov, ulen, |
903ab86d HX |
965 | sizeof(struct udphdr), &ipc, &rt, |
966 | msg->msg_flags); | |
967 | err = PTR_ERR(skb); | |
968 | if (skb && !IS_ERR(skb)) | |
79ab0531 | 969 | err = udp_send_skb(skb, fl4); |
903ab86d HX |
970 | goto out; |
971 | } | |
972 | ||
db8dac20 DM |
973 | lock_sock(sk); |
974 | if (unlikely(up->pending)) { | |
975 | /* The socket is already corked while preparing it. */ | |
976 | /* ... which is an evident application bug. --ANK */ | |
977 | release_sock(sk); | |
978 | ||
afd46503 | 979 | LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("cork app bug 2\n")); |
db8dac20 DM |
980 | err = -EINVAL; |
981 | goto out; | |
982 | } | |
983 | /* | |
984 | * Now cork the socket to pend data. | |
985 | */ | |
b6f21b26 DM |
986 | fl4 = &inet->cork.fl.u.ip4; |
987 | fl4->daddr = daddr; | |
988 | fl4->saddr = saddr; | |
9cce96df DM |
989 | fl4->fl4_dport = dport; |
990 | fl4->fl4_sport = inet->inet_sport; | |
db8dac20 DM |
991 | up->pending = AF_INET; |
992 | ||
993 | do_append_data: | |
994 | up->len += ulen; | |
f5fca608 DM |
995 | err = ip_append_data(sk, fl4, getfrag, msg->msg_iov, ulen, |
996 | sizeof(struct udphdr), &ipc, &rt, | |
997 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | |
db8dac20 DM |
998 | if (err) |
999 | udp_flush_pending_frames(sk); | |
1000 | else if (!corkreq) | |
1001 | err = udp_push_pending_frames(sk); | |
1002 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
1003 | up->pending = 0; | |
1004 | release_sock(sk); | |
1005 | ||
1006 | out: | |
1007 | ip_rt_put(rt); | |
1008 | if (free) | |
1009 | kfree(ipc.opt); | |
1010 | if (!err) | |
1011 | return len; | |
1012 | /* | |
1013 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1014 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1015 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1016 | * things). We could add another new stat but at least for now that | |
1017 | * seems like overkill. | |
1018 | */ | |
1019 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
1020 | UDP_INC_STATS_USER(sock_net(sk), |
1021 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
1022 | } |
1023 | return err; | |
1024 | ||
1025 | do_confirm: | |
d8d1f30b | 1026 | dst_confirm(&rt->dst); |
db8dac20 DM |
1027 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1028 | goto back_from_confirm; | |
1029 | err = 0; | |
1030 | goto out; | |
1031 | } | |
c482c568 | 1032 | EXPORT_SYMBOL(udp_sendmsg); |
db8dac20 DM |
1033 | |
1034 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
1035 | size_t size, int flags) | |
1036 | { | |
f5fca608 | 1037 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
1038 | struct udp_sock *up = udp_sk(sk); |
1039 | int ret; | |
1040 | ||
1041 | if (!up->pending) { | |
1042 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
1043 | ||
1044 | /* Call udp_sendmsg to specify destination address which | |
1045 | * sendpage interface can't pass. | |
1046 | * This will succeed only when the socket is connected. | |
1047 | */ | |
1048 | ret = udp_sendmsg(NULL, sk, &msg, 0); | |
1049 | if (ret < 0) | |
1050 | return ret; | |
1051 | } | |
1052 | ||
1053 | lock_sock(sk); | |
1054 | ||
1055 | if (unlikely(!up->pending)) { | |
1056 | release_sock(sk); | |
1057 | ||
afd46503 | 1058 | LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("udp cork app bug 3\n")); |
db8dac20 DM |
1059 | return -EINVAL; |
1060 | } | |
1061 | ||
f5fca608 DM |
1062 | ret = ip_append_page(sk, &inet->cork.fl.u.ip4, |
1063 | page, offset, size, flags); | |
db8dac20 DM |
1064 | if (ret == -EOPNOTSUPP) { |
1065 | release_sock(sk); | |
1066 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
1067 | size, flags); | |
1068 | } | |
1069 | if (ret < 0) { | |
1070 | udp_flush_pending_frames(sk); | |
1071 | goto out; | |
1072 | } | |
1073 | ||
1074 | up->len += size; | |
1075 | if (!(up->corkflag || (flags&MSG_MORE))) | |
1076 | ret = udp_push_pending_frames(sk); | |
1077 | if (!ret) | |
1078 | ret = size; | |
1079 | out: | |
1080 | release_sock(sk); | |
1081 | return ret; | |
1082 | } | |
1083 | ||
85584672 ED |
1084 | |
1085 | /** | |
1086 | * first_packet_length - return length of first packet in receive queue | |
1087 | * @sk: socket | |
1088 | * | |
1089 | * Drops all bad checksum frames, until a valid one is found. | |
1090 | * Returns the length of found skb, or 0 if none is found. | |
1091 | */ | |
1092 | static unsigned int first_packet_length(struct sock *sk) | |
1093 | { | |
1094 | struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue; | |
1095 | struct sk_buff *skb; | |
1096 | unsigned int res; | |
1097 | ||
1098 | __skb_queue_head_init(&list_kill); | |
1099 | ||
1100 | spin_lock_bh(&rcvq->lock); | |
1101 | while ((skb = skb_peek(rcvq)) != NULL && | |
1102 | udp_lib_checksum_complete(skb)) { | |
1103 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, | |
1104 | IS_UDPLITE(sk)); | |
8edf19c2 | 1105 | atomic_inc(&sk->sk_drops); |
85584672 ED |
1106 | __skb_unlink(skb, rcvq); |
1107 | __skb_queue_tail(&list_kill, skb); | |
1108 | } | |
1109 | res = skb ? skb->len : 0; | |
1110 | spin_unlock_bh(&rcvq->lock); | |
1111 | ||
1112 | if (!skb_queue_empty(&list_kill)) { | |
8a74ad60 ED |
1113 | bool slow = lock_sock_fast(sk); |
1114 | ||
85584672 ED |
1115 | __skb_queue_purge(&list_kill); |
1116 | sk_mem_reclaim_partial(sk); | |
8a74ad60 | 1117 | unlock_sock_fast(sk, slow); |
85584672 ED |
1118 | } |
1119 | return res; | |
1120 | } | |
1121 | ||
1da177e4 LT |
1122 | /* |
1123 | * IOCTL requests applicable to the UDP protocol | |
1124 | */ | |
e905a9ed | 1125 | |
1da177e4 LT |
1126 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
1127 | { | |
6516c655 SH |
1128 | switch (cmd) { |
1129 | case SIOCOUTQ: | |
1da177e4 | 1130 | { |
31e6d363 ED |
1131 | int amount = sk_wmem_alloc_get(sk); |
1132 | ||
6516c655 SH |
1133 | return put_user(amount, (int __user *)arg); |
1134 | } | |
1da177e4 | 1135 | |
6516c655 SH |
1136 | case SIOCINQ: |
1137 | { | |
85584672 | 1138 | unsigned int amount = first_packet_length(sk); |
6516c655 | 1139 | |
85584672 | 1140 | if (amount) |
6516c655 SH |
1141 | /* |
1142 | * We will only return the amount | |
1143 | * of this packet since that is all | |
1144 | * that will be read. | |
1145 | */ | |
85584672 ED |
1146 | amount -= sizeof(struct udphdr); |
1147 | ||
6516c655 SH |
1148 | return put_user(amount, (int __user *)arg); |
1149 | } | |
1da177e4 | 1150 | |
6516c655 SH |
1151 | default: |
1152 | return -ENOIOCTLCMD; | |
1da177e4 | 1153 | } |
6516c655 SH |
1154 | |
1155 | return 0; | |
1da177e4 | 1156 | } |
c482c568 | 1157 | EXPORT_SYMBOL(udp_ioctl); |
1da177e4 | 1158 | |
db8dac20 DM |
1159 | /* |
1160 | * This should be easy, if there is something there we | |
1161 | * return it, otherwise we block. | |
1162 | */ | |
1163 | ||
1164 | int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
1165 | size_t len, int noblock, int flags, int *addr_len) | |
1166 | { | |
1167 | struct inet_sock *inet = inet_sk(sk); | |
1168 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; | |
1169 | struct sk_buff *skb; | |
59c2cdae | 1170 | unsigned int ulen, copied; |
3f518bf7 | 1171 | int peeked, off = 0; |
db8dac20 DM |
1172 | int err; |
1173 | int is_udplite = IS_UDPLITE(sk); | |
8a74ad60 | 1174 | bool slow; |
db8dac20 DM |
1175 | |
1176 | /* | |
1177 | * Check any passed addresses | |
1178 | */ | |
1179 | if (addr_len) | |
c482c568 | 1180 | *addr_len = sizeof(*sin); |
db8dac20 DM |
1181 | |
1182 | if (flags & MSG_ERRQUEUE) | |
1183 | return ip_recv_error(sk, msg, len); | |
1184 | ||
1185 | try_again: | |
1186 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), | |
3f518bf7 | 1187 | &peeked, &off, &err); |
db8dac20 DM |
1188 | if (!skb) |
1189 | goto out; | |
1190 | ||
1191 | ulen = skb->len - sizeof(struct udphdr); | |
59c2cdae DM |
1192 | copied = len; |
1193 | if (copied > ulen) | |
1194 | copied = ulen; | |
1195 | else if (copied < ulen) | |
db8dac20 DM |
1196 | msg->msg_flags |= MSG_TRUNC; |
1197 | ||
1198 | /* | |
1199 | * If checksum is needed at all, try to do it while copying the | |
1200 | * data. If the data is truncated, or if we only want a partial | |
1201 | * coverage checksum (UDP-Lite), do it before the copy. | |
1202 | */ | |
1203 | ||
59c2cdae | 1204 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { |
db8dac20 DM |
1205 | if (udp_lib_checksum_complete(skb)) |
1206 | goto csum_copy_err; | |
1207 | } | |
1208 | ||
1209 | if (skb_csum_unnecessary(skb)) | |
1210 | err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), | |
59c2cdae | 1211 | msg->msg_iov, copied); |
db8dac20 | 1212 | else { |
c482c568 ED |
1213 | err = skb_copy_and_csum_datagram_iovec(skb, |
1214 | sizeof(struct udphdr), | |
1215 | msg->msg_iov); | |
db8dac20 DM |
1216 | |
1217 | if (err == -EINVAL) | |
1218 | goto csum_copy_err; | |
1219 | } | |
1220 | ||
1221 | if (err) | |
1222 | goto out_free; | |
1223 | ||
1224 | if (!peeked) | |
629ca23c PE |
1225 | UDP_INC_STATS_USER(sock_net(sk), |
1226 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 | 1227 | |
3b885787 | 1228 | sock_recv_ts_and_drops(msg, sk, skb); |
db8dac20 DM |
1229 | |
1230 | /* Copy the address. */ | |
c482c568 | 1231 | if (sin) { |
db8dac20 DM |
1232 | sin->sin_family = AF_INET; |
1233 | sin->sin_port = udp_hdr(skb)->source; | |
1234 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
1235 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
1236 | } | |
1237 | if (inet->cmsg_flags) | |
1238 | ip_cmsg_recv(msg, skb); | |
1239 | ||
59c2cdae | 1240 | err = copied; |
db8dac20 DM |
1241 | if (flags & MSG_TRUNC) |
1242 | err = ulen; | |
1243 | ||
1244 | out_free: | |
9d410c79 | 1245 | skb_free_datagram_locked(sk, skb); |
db8dac20 DM |
1246 | out: |
1247 | return err; | |
1248 | ||
1249 | csum_copy_err: | |
8a74ad60 | 1250 | slow = lock_sock_fast(sk); |
db8dac20 | 1251 | if (!skb_kill_datagram(sk, skb, flags)) |
629ca23c | 1252 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8a74ad60 | 1253 | unlock_sock_fast(sk, slow); |
db8dac20 DM |
1254 | |
1255 | if (noblock) | |
1256 | return -EAGAIN; | |
9cfaa8de XZ |
1257 | |
1258 | /* starting over for a new packet */ | |
1259 | msg->msg_flags &= ~MSG_TRUNC; | |
db8dac20 DM |
1260 | goto try_again; |
1261 | } | |
1262 | ||
1263 | ||
1da177e4 LT |
1264 | int udp_disconnect(struct sock *sk, int flags) |
1265 | { | |
1266 | struct inet_sock *inet = inet_sk(sk); | |
1267 | /* | |
1268 | * 1003.1g - break association. | |
1269 | */ | |
e905a9ed | 1270 | |
1da177e4 | 1271 | sk->sk_state = TCP_CLOSE; |
c720c7e8 ED |
1272 | inet->inet_daddr = 0; |
1273 | inet->inet_dport = 0; | |
bdeab991 | 1274 | sock_rps_reset_rxhash(sk); |
1da177e4 LT |
1275 | sk->sk_bound_dev_if = 0; |
1276 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
1277 | inet_reset_saddr(sk); | |
1278 | ||
1279 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
1280 | sk->sk_prot->unhash(sk); | |
c720c7e8 | 1281 | inet->inet_sport = 0; |
1da177e4 LT |
1282 | } |
1283 | sk_dst_reset(sk); | |
1284 | return 0; | |
1285 | } | |
c482c568 | 1286 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1287 | |
645ca708 ED |
1288 | void udp_lib_unhash(struct sock *sk) |
1289 | { | |
723b4610 ED |
1290 | if (sk_hashed(sk)) { |
1291 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
512615b6 ED |
1292 | struct udp_hslot *hslot, *hslot2; |
1293 | ||
1294 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1295 | udp_sk(sk)->udp_port_hash); | |
1296 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
645ca708 | 1297 | |
723b4610 ED |
1298 | spin_lock_bh(&hslot->lock); |
1299 | if (sk_nulls_del_node_init_rcu(sk)) { | |
fdcc8aa9 | 1300 | hslot->count--; |
c720c7e8 | 1301 | inet_sk(sk)->inet_num = 0; |
723b4610 | 1302 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
512615b6 ED |
1303 | |
1304 | spin_lock(&hslot2->lock); | |
1305 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1306 | hslot2->count--; | |
1307 | spin_unlock(&hslot2->lock); | |
723b4610 ED |
1308 | } |
1309 | spin_unlock_bh(&hslot->lock); | |
645ca708 | 1310 | } |
645ca708 ED |
1311 | } |
1312 | EXPORT_SYMBOL(udp_lib_unhash); | |
1313 | ||
719f8358 ED |
1314 | /* |
1315 | * inet_rcv_saddr was changed, we must rehash secondary hash | |
1316 | */ | |
1317 | void udp_lib_rehash(struct sock *sk, u16 newhash) | |
1318 | { | |
1319 | if (sk_hashed(sk)) { | |
1320 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
1321 | struct udp_hslot *hslot, *hslot2, *nhslot2; | |
1322 | ||
1323 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
1324 | nhslot2 = udp_hashslot2(udptable, newhash); | |
1325 | udp_sk(sk)->udp_portaddr_hash = newhash; | |
1326 | if (hslot2 != nhslot2) { | |
1327 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1328 | udp_sk(sk)->udp_port_hash); | |
1329 | /* we must lock primary chain too */ | |
1330 | spin_lock_bh(&hslot->lock); | |
1331 | ||
1332 | spin_lock(&hslot2->lock); | |
1333 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1334 | hslot2->count--; | |
1335 | spin_unlock(&hslot2->lock); | |
1336 | ||
1337 | spin_lock(&nhslot2->lock); | |
1338 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
1339 | &nhslot2->head); | |
1340 | nhslot2->count++; | |
1341 | spin_unlock(&nhslot2->lock); | |
1342 | ||
1343 | spin_unlock_bh(&hslot->lock); | |
1344 | } | |
1345 | } | |
1346 | } | |
1347 | EXPORT_SYMBOL(udp_lib_rehash); | |
1348 | ||
1349 | static void udp_v4_rehash(struct sock *sk) | |
1350 | { | |
1351 | u16 new_hash = udp4_portaddr_hash(sock_net(sk), | |
1352 | inet_sk(sk)->inet_rcv_saddr, | |
1353 | inet_sk(sk)->inet_num); | |
1354 | udp_lib_rehash(sk, new_hash); | |
1355 | } | |
1356 | ||
93821778 HX |
1357 | static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
1358 | { | |
fec5e652 | 1359 | int rc; |
766e9037 | 1360 | |
fec5e652 | 1361 | if (inet_sk(sk)->inet_daddr) |
bdeab991 | 1362 | sock_rps_save_rxhash(sk, skb); |
fec5e652 | 1363 | |
d826eb14 | 1364 | rc = sock_queue_rcv_skb(sk, skb); |
766e9037 ED |
1365 | if (rc < 0) { |
1366 | int is_udplite = IS_UDPLITE(sk); | |
93821778 | 1367 | |
93821778 | 1368 | /* Note that an ENOMEM error is charged twice */ |
766e9037 | 1369 | if (rc == -ENOMEM) |
93821778 HX |
1370 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1371 | is_udplite); | |
766e9037 ED |
1372 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
1373 | kfree_skb(skb); | |
296f7ea7 | 1374 | trace_udp_fail_queue_rcv_skb(rc, sk); |
766e9037 | 1375 | return -1; |
93821778 HX |
1376 | } |
1377 | ||
1378 | return 0; | |
1379 | ||
93821778 HX |
1380 | } |
1381 | ||
db8dac20 DM |
1382 | /* returns: |
1383 | * -1: error | |
1384 | * 0: success | |
1385 | * >0: "udp encap" protocol resubmission | |
1386 | * | |
1387 | * Note that in the success and error cases, the skb is assumed to | |
1388 | * have either been requeued or freed. | |
1389 | */ | |
c482c568 | 1390 | int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
db8dac20 DM |
1391 | { |
1392 | struct udp_sock *up = udp_sk(sk); | |
1393 | int rc; | |
1394 | int is_udplite = IS_UDPLITE(sk); | |
1395 | ||
1396 | /* | |
1397 | * Charge it to the socket, dropping if the queue is full. | |
1398 | */ | |
1399 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1400 | goto drop; | |
1401 | nf_reset(skb); | |
1402 | ||
1403 | if (up->encap_type) { | |
0ad92ad0 ED |
1404 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
1405 | ||
db8dac20 DM |
1406 | /* |
1407 | * This is an encapsulation socket so pass the skb to | |
1408 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1409 | * fall through and pass this up the UDP socket. | |
1410 | * up->encap_rcv() returns the following value: | |
1411 | * =0 if skb was successfully passed to the encap | |
1412 | * handler or was discarded by it. | |
1413 | * >0 if skb should be passed on to UDP. | |
1414 | * <0 if skb should be resubmitted as proto -N | |
1415 | */ | |
1416 | ||
1417 | /* if we're overly short, let UDP handle it */ | |
0ad92ad0 ED |
1418 | encap_rcv = ACCESS_ONCE(up->encap_rcv); |
1419 | if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) { | |
db8dac20 DM |
1420 | int ret; |
1421 | ||
0ad92ad0 | 1422 | ret = encap_rcv(sk, skb); |
db8dac20 | 1423 | if (ret <= 0) { |
0283328e PE |
1424 | UDP_INC_STATS_BH(sock_net(sk), |
1425 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
1426 | is_udplite); |
1427 | return -ret; | |
1428 | } | |
1429 | } | |
1430 | ||
1431 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1432 | } | |
1433 | ||
1434 | /* | |
1435 | * UDP-Lite specific tests, ignored on UDP sockets | |
1436 | */ | |
1437 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1438 | ||
1439 | /* | |
1440 | * MIB statistics other than incrementing the error count are | |
1441 | * disabled for the following two types of errors: these depend | |
1442 | * on the application settings, not on the functioning of the | |
1443 | * protocol stack as such. | |
1444 | * | |
1445 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1446 | * way ... to ... at least let the receiving application block | |
1447 | * delivery of packets with coverage values less than a value | |
1448 | * provided by the application." | |
1449 | */ | |
1450 | if (up->pcrlen == 0) { /* full coverage was set */ | |
afd46503 JP |
1451 | LIMIT_NETDEBUG(KERN_WARNING "UDPLite: partial coverage %d while full coverage %d requested\n", |
1452 | UDP_SKB_CB(skb)->cscov, skb->len); | |
db8dac20 DM |
1453 | goto drop; |
1454 | } | |
1455 | /* The next case involves violating the min. coverage requested | |
1456 | * by the receiver. This is subtle: if receiver wants x and x is | |
1457 | * greater than the buffersize/MTU then receiver will complain | |
1458 | * that it wants x while sender emits packets of smaller size y. | |
1459 | * Therefore the above ...()->partial_cov statement is essential. | |
1460 | */ | |
1461 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
afd46503 JP |
1462 | LIMIT_NETDEBUG(KERN_WARNING "UDPLite: coverage %d too small, need min %d\n", |
1463 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
db8dac20 DM |
1464 | goto drop; |
1465 | } | |
1466 | } | |
1467 | ||
33d480ce ED |
1468 | if (rcu_access_pointer(sk->sk_filter) && |
1469 | udp_lib_checksum_complete(skb)) | |
1470 | goto drop; | |
db8dac20 | 1471 | |
c377411f ED |
1472 | |
1473 | if (sk_rcvqueues_full(sk, skb)) | |
1474 | goto drop; | |
1475 | ||
93821778 | 1476 | rc = 0; |
db8dac20 | 1477 | |
d826eb14 | 1478 | ipv4_pktinfo_prepare(skb); |
93821778 HX |
1479 | bh_lock_sock(sk); |
1480 | if (!sock_owned_by_user(sk)) | |
1481 | rc = __udp_queue_rcv_skb(sk, skb); | |
a3a858ff | 1482 | else if (sk_add_backlog(sk, skb)) { |
55349790 ZY |
1483 | bh_unlock_sock(sk); |
1484 | goto drop; | |
1485 | } | |
93821778 HX |
1486 | bh_unlock_sock(sk); |
1487 | ||
1488 | return rc; | |
db8dac20 DM |
1489 | |
1490 | drop: | |
0283328e | 1491 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8edf19c2 | 1492 | atomic_inc(&sk->sk_drops); |
db8dac20 DM |
1493 | kfree_skb(skb); |
1494 | return -1; | |
1495 | } | |
1496 | ||
1240d137 ED |
1497 | |
1498 | static void flush_stack(struct sock **stack, unsigned int count, | |
1499 | struct sk_buff *skb, unsigned int final) | |
1500 | { | |
1501 | unsigned int i; | |
1502 | struct sk_buff *skb1 = NULL; | |
f6b8f32c | 1503 | struct sock *sk; |
1240d137 ED |
1504 | |
1505 | for (i = 0; i < count; i++) { | |
f6b8f32c | 1506 | sk = stack[i]; |
1240d137 ED |
1507 | if (likely(skb1 == NULL)) |
1508 | skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC); | |
1509 | ||
f6b8f32c ED |
1510 | if (!skb1) { |
1511 | atomic_inc(&sk->sk_drops); | |
1512 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, | |
1513 | IS_UDPLITE(sk)); | |
1514 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, | |
1515 | IS_UDPLITE(sk)); | |
1516 | } | |
1517 | ||
1518 | if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0) | |
1240d137 ED |
1519 | skb1 = NULL; |
1520 | } | |
1521 | if (unlikely(skb1)) | |
1522 | kfree_skb(skb1); | |
1523 | } | |
1524 | ||
db8dac20 DM |
1525 | /* |
1526 | * Multicasts and broadcasts go to each listener. | |
1527 | * | |
1240d137 | 1528 | * Note: called only from the BH handler context. |
db8dac20 | 1529 | */ |
e3163493 | 1530 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1531 | struct udphdr *uh, |
1532 | __be32 saddr, __be32 daddr, | |
645ca708 | 1533 | struct udp_table *udptable) |
db8dac20 | 1534 | { |
1240d137 | 1535 | struct sock *sk, *stack[256 / sizeof(struct sock *)]; |
f86dcc5a | 1536 | struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest)); |
db8dac20 | 1537 | int dif; |
1240d137 | 1538 | unsigned int i, count = 0; |
db8dac20 | 1539 | |
645ca708 | 1540 | spin_lock(&hslot->lock); |
88ab1932 | 1541 | sk = sk_nulls_head(&hslot->head); |
db8dac20 | 1542 | dif = skb->dev->ifindex; |
920a4611 | 1543 | sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif); |
1240d137 ED |
1544 | while (sk) { |
1545 | stack[count++] = sk; | |
1546 | sk = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest, | |
1547 | daddr, uh->source, saddr, dif); | |
1548 | if (unlikely(count == ARRAY_SIZE(stack))) { | |
1549 | if (!sk) | |
1550 | break; | |
1551 | flush_stack(stack, count, skb, ~0); | |
1552 | count = 0; | |
1553 | } | |
1554 | } | |
1555 | /* | |
1556 | * before releasing chain lock, we must take a reference on sockets | |
1557 | */ | |
1558 | for (i = 0; i < count; i++) | |
1559 | sock_hold(stack[i]); | |
1560 | ||
645ca708 | 1561 | spin_unlock(&hslot->lock); |
1240d137 ED |
1562 | |
1563 | /* | |
1564 | * do the slow work with no lock held | |
1565 | */ | |
1566 | if (count) { | |
1567 | flush_stack(stack, count, skb, count - 1); | |
1568 | ||
1569 | for (i = 0; i < count; i++) | |
1570 | sock_put(stack[i]); | |
1571 | } else { | |
1572 | kfree_skb(skb); | |
1573 | } | |
db8dac20 DM |
1574 | return 0; |
1575 | } | |
1576 | ||
1577 | /* Initialize UDP checksum. If exited with zero value (success), | |
1578 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1579 | * Otherwise, csum completion requires chacksumming packet body, | |
1580 | * including udp header and folding it to skb->csum. | |
1581 | */ | |
1582 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1583 | int proto) | |
1584 | { | |
1585 | const struct iphdr *iph; | |
1586 | int err; | |
1587 | ||
1588 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1589 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1590 | ||
1591 | if (proto == IPPROTO_UDPLITE) { | |
1592 | err = udplite_checksum_init(skb, uh); | |
1593 | if (err) | |
1594 | return err; | |
1595 | } | |
1596 | ||
1597 | iph = ip_hdr(skb); | |
1598 | if (uh->check == 0) { | |
1599 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1600 | } else if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
c482c568 | 1601 | if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, |
db8dac20 DM |
1602 | proto, skb->csum)) |
1603 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1604 | } | |
1605 | if (!skb_csum_unnecessary(skb)) | |
1606 | skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, | |
1607 | skb->len, proto, 0); | |
1608 | /* Probably, we should checksum udp header (it should be in cache | |
1609 | * in any case) and data in tiny packets (< rx copybreak). | |
1610 | */ | |
1611 | ||
1612 | return 0; | |
1613 | } | |
1614 | ||
1615 | /* | |
1616 | * All we need to do is get the socket, and then do a checksum. | |
1617 | */ | |
1618 | ||
645ca708 | 1619 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1620 | int proto) |
1621 | { | |
1622 | struct sock *sk; | |
7b5e56f9 | 1623 | struct udphdr *uh; |
db8dac20 | 1624 | unsigned short ulen; |
adf30907 | 1625 | struct rtable *rt = skb_rtable(skb); |
2783ef23 | 1626 | __be32 saddr, daddr; |
0283328e | 1627 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1628 | |
1629 | /* | |
1630 | * Validate the packet. | |
1631 | */ | |
1632 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1633 | goto drop; /* No space for header. */ | |
1634 | ||
7b5e56f9 | 1635 | uh = udp_hdr(skb); |
db8dac20 | 1636 | ulen = ntohs(uh->len); |
ccc2d97c BM |
1637 | saddr = ip_hdr(skb)->saddr; |
1638 | daddr = ip_hdr(skb)->daddr; | |
1639 | ||
db8dac20 DM |
1640 | if (ulen > skb->len) |
1641 | goto short_packet; | |
1642 | ||
1643 | if (proto == IPPROTO_UDP) { | |
1644 | /* UDP validates ulen. */ | |
1645 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1646 | goto short_packet; | |
1647 | uh = udp_hdr(skb); | |
1648 | } | |
1649 | ||
1650 | if (udp4_csum_init(skb, uh, proto)) | |
1651 | goto csum_error; | |
1652 | ||
1653 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) | |
e3163493 PE |
1654 | return __udp4_lib_mcast_deliver(net, skb, uh, |
1655 | saddr, daddr, udptable); | |
db8dac20 | 1656 | |
607c4aaf | 1657 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); |
db8dac20 DM |
1658 | |
1659 | if (sk != NULL) { | |
93821778 | 1660 | int ret = udp_queue_rcv_skb(sk, skb); |
db8dac20 DM |
1661 | sock_put(sk); |
1662 | ||
1663 | /* a return value > 0 means to resubmit the input, but | |
1664 | * it wants the return to be -protocol, or 0 | |
1665 | */ | |
1666 | if (ret > 0) | |
1667 | return -ret; | |
1668 | return 0; | |
1669 | } | |
1670 | ||
1671 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1672 | goto drop; | |
1673 | nf_reset(skb); | |
1674 | ||
1675 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1676 | if (udp_lib_checksum_complete(skb)) | |
1677 | goto csum_error; | |
1678 | ||
0283328e | 1679 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1680 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1681 | ||
1682 | /* | |
1683 | * Hmm. We got an UDP packet to a port to which we | |
1684 | * don't wanna listen. Ignore it. | |
1685 | */ | |
1686 | kfree_skb(skb); | |
1687 | return 0; | |
1688 | ||
1689 | short_packet: | |
673d57e7 | 1690 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n", |
afd46503 JP |
1691 | proto == IPPROTO_UDPLITE ? "Lite" : "", |
1692 | &saddr, ntohs(uh->source), | |
1693 | ulen, skb->len, | |
1694 | &daddr, ntohs(uh->dest)); | |
db8dac20 DM |
1695 | goto drop; |
1696 | ||
1697 | csum_error: | |
1698 | /* | |
1699 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1700 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1701 | */ | |
673d57e7 | 1702 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n", |
afd46503 JP |
1703 | proto == IPPROTO_UDPLITE ? "Lite" : "", |
1704 | &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest), | |
db8dac20 DM |
1705 | ulen); |
1706 | drop: | |
0283328e | 1707 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1708 | kfree_skb(skb); |
1709 | return 0; | |
1710 | } | |
1711 | ||
1712 | int udp_rcv(struct sk_buff *skb) | |
1713 | { | |
645ca708 | 1714 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
1715 | } |
1716 | ||
7d06b2e0 | 1717 | void udp_destroy_sock(struct sock *sk) |
db8dac20 | 1718 | { |
8a74ad60 | 1719 | bool slow = lock_sock_fast(sk); |
db8dac20 | 1720 | udp_flush_pending_frames(sk); |
8a74ad60 | 1721 | unlock_sock_fast(sk, slow); |
db8dac20 DM |
1722 | } |
1723 | ||
1da177e4 LT |
1724 | /* |
1725 | * Socket option code for UDP | |
1726 | */ | |
4c0a6cb0 | 1727 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 1728 | char __user *optval, unsigned int optlen, |
4c0a6cb0 | 1729 | int (*push_pending_frames)(struct sock *)) |
1da177e4 LT |
1730 | { |
1731 | struct udp_sock *up = udp_sk(sk); | |
1732 | int val; | |
1733 | int err = 0; | |
b2bf1e26 | 1734 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 1735 | |
c482c568 | 1736 | if (optlen < sizeof(int)) |
1da177e4 LT |
1737 | return -EINVAL; |
1738 | ||
1739 | if (get_user(val, (int __user *)optval)) | |
1740 | return -EFAULT; | |
1741 | ||
6516c655 | 1742 | switch (optname) { |
1da177e4 LT |
1743 | case UDP_CORK: |
1744 | if (val != 0) { | |
1745 | up->corkflag = 1; | |
1746 | } else { | |
1747 | up->corkflag = 0; | |
1748 | lock_sock(sk); | |
4c0a6cb0 | 1749 | (*push_pending_frames)(sk); |
1da177e4 LT |
1750 | release_sock(sk); |
1751 | } | |
1752 | break; | |
e905a9ed | 1753 | |
1da177e4 LT |
1754 | case UDP_ENCAP: |
1755 | switch (val) { | |
1756 | case 0: | |
1757 | case UDP_ENCAP_ESPINUDP: | |
1758 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
1759 | up->encap_rcv = xfrm4_udp_encap_rcv; |
1760 | /* FALLTHROUGH */ | |
342f0234 | 1761 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 LT |
1762 | up->encap_type = val; |
1763 | break; | |
1764 | default: | |
1765 | err = -ENOPROTOOPT; | |
1766 | break; | |
1767 | } | |
1768 | break; | |
1769 | ||
ba4e58ec GR |
1770 | /* |
1771 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
1772 | */ | |
1773 | /* The sender sets actual checksum coverage length via this option. | |
1774 | * The case coverage > packet length is handled by send module. */ | |
1775 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 1776 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1777 | return -ENOPROTOOPT; |
1778 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
1779 | val = 8; | |
4be929be AD |
1780 | else if (val > USHRT_MAX) |
1781 | val = USHRT_MAX; | |
ba4e58ec GR |
1782 | up->pcslen = val; |
1783 | up->pcflag |= UDPLITE_SEND_CC; | |
1784 | break; | |
1785 | ||
e905a9ed YH |
1786 | /* The receiver specifies a minimum checksum coverage value. To make |
1787 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
1788 | * used, this again means full checksum coverage. */ |
1789 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 1790 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1791 | return -ENOPROTOOPT; |
1792 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
1793 | val = 8; | |
4be929be AD |
1794 | else if (val > USHRT_MAX) |
1795 | val = USHRT_MAX; | |
ba4e58ec GR |
1796 | up->pcrlen = val; |
1797 | up->pcflag |= UDPLITE_RECV_CC; | |
1798 | break; | |
1799 | ||
1da177e4 LT |
1800 | default: |
1801 | err = -ENOPROTOOPT; | |
1802 | break; | |
6516c655 | 1803 | } |
1da177e4 LT |
1804 | |
1805 | return err; | |
1806 | } | |
c482c568 | 1807 | EXPORT_SYMBOL(udp_lib_setsockopt); |
1da177e4 | 1808 | |
db8dac20 | 1809 | int udp_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 1810 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
1811 | { |
1812 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1813 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1814 | udp_push_pending_frames); | |
1815 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
1816 | } | |
1817 | ||
1818 | #ifdef CONFIG_COMPAT | |
1819 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 1820 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
1821 | { |
1822 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1823 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1824 | udp_push_pending_frames); | |
1825 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
1826 | } | |
1827 | #endif | |
1828 | ||
4c0a6cb0 GR |
1829 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
1830 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
1831 | { |
1832 | struct udp_sock *up = udp_sk(sk); | |
1833 | int val, len; | |
1834 | ||
c482c568 | 1835 | if (get_user(len, optlen)) |
1da177e4 LT |
1836 | return -EFAULT; |
1837 | ||
1838 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 1839 | |
6516c655 | 1840 | if (len < 0) |
1da177e4 LT |
1841 | return -EINVAL; |
1842 | ||
6516c655 | 1843 | switch (optname) { |
1da177e4 LT |
1844 | case UDP_CORK: |
1845 | val = up->corkflag; | |
1846 | break; | |
1847 | ||
1848 | case UDP_ENCAP: | |
1849 | val = up->encap_type; | |
1850 | break; | |
1851 | ||
ba4e58ec GR |
1852 | /* The following two cannot be changed on UDP sockets, the return is |
1853 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
1854 | case UDPLITE_SEND_CSCOV: | |
1855 | val = up->pcslen; | |
1856 | break; | |
1857 | ||
1858 | case UDPLITE_RECV_CSCOV: | |
1859 | val = up->pcrlen; | |
1860 | break; | |
1861 | ||
1da177e4 LT |
1862 | default: |
1863 | return -ENOPROTOOPT; | |
6516c655 | 1864 | } |
1da177e4 | 1865 | |
6516c655 | 1866 | if (put_user(len, optlen)) |
e905a9ed | 1867 | return -EFAULT; |
c482c568 | 1868 | if (copy_to_user(optval, &val, len)) |
1da177e4 | 1869 | return -EFAULT; |
e905a9ed | 1870 | return 0; |
1da177e4 | 1871 | } |
c482c568 | 1872 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1da177e4 | 1873 | |
db8dac20 DM |
1874 | int udp_getsockopt(struct sock *sk, int level, int optname, |
1875 | char __user *optval, int __user *optlen) | |
1876 | { | |
1877 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1878 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1879 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
1880 | } | |
1881 | ||
1882 | #ifdef CONFIG_COMPAT | |
1883 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
1884 | char __user *optval, int __user *optlen) | |
1885 | { | |
1886 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1887 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1888 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
1889 | } | |
1890 | #endif | |
1da177e4 LT |
1891 | /** |
1892 | * udp_poll - wait for a UDP event. | |
1893 | * @file - file struct | |
1894 | * @sock - socket | |
1895 | * @wait - poll table | |
1896 | * | |
e905a9ed | 1897 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
1898 | * blocking sockets. If application is using a blocking fd |
1899 | * and a packet with checksum error is in the queue; | |
1900 | * then it could get return from select indicating data available | |
1901 | * but then block when reading it. Add special case code | |
1902 | * to work around these arguably broken applications. | |
1903 | */ | |
1904 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
1905 | { | |
1906 | unsigned int mask = datagram_poll(file, sock, wait); | |
1907 | struct sock *sk = sock->sk; | |
ba4e58ec | 1908 | |
1da177e4 | 1909 | /* Check for false positives due to checksum errors */ |
85584672 ED |
1910 | if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) && |
1911 | !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk)) | |
1912 | mask &= ~(POLLIN | POLLRDNORM); | |
1da177e4 LT |
1913 | |
1914 | return mask; | |
e905a9ed | 1915 | |
1da177e4 | 1916 | } |
c482c568 | 1917 | EXPORT_SYMBOL(udp_poll); |
1da177e4 | 1918 | |
db8dac20 DM |
1919 | struct proto udp_prot = { |
1920 | .name = "UDP", | |
1921 | .owner = THIS_MODULE, | |
1922 | .close = udp_lib_close, | |
1923 | .connect = ip4_datagram_connect, | |
1924 | .disconnect = udp_disconnect, | |
1925 | .ioctl = udp_ioctl, | |
1926 | .destroy = udp_destroy_sock, | |
1927 | .setsockopt = udp_setsockopt, | |
1928 | .getsockopt = udp_getsockopt, | |
1929 | .sendmsg = udp_sendmsg, | |
1930 | .recvmsg = udp_recvmsg, | |
1931 | .sendpage = udp_sendpage, | |
93821778 | 1932 | .backlog_rcv = __udp_queue_rcv_skb, |
db8dac20 DM |
1933 | .hash = udp_lib_hash, |
1934 | .unhash = udp_lib_unhash, | |
719f8358 | 1935 | .rehash = udp_v4_rehash, |
db8dac20 DM |
1936 | .get_port = udp_v4_get_port, |
1937 | .memory_allocated = &udp_memory_allocated, | |
1938 | .sysctl_mem = sysctl_udp_mem, | |
1939 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
1940 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
1941 | .obj_size = sizeof(struct udp_sock), | |
271b72c7 | 1942 | .slab_flags = SLAB_DESTROY_BY_RCU, |
645ca708 | 1943 | .h.udp_table = &udp_table, |
db8dac20 DM |
1944 | #ifdef CONFIG_COMPAT |
1945 | .compat_setsockopt = compat_udp_setsockopt, | |
1946 | .compat_getsockopt = compat_udp_getsockopt, | |
1947 | #endif | |
fcbdf09d | 1948 | .clear_sk = sk_prot_clear_portaddr_nulls, |
db8dac20 | 1949 | }; |
c482c568 | 1950 | EXPORT_SYMBOL(udp_prot); |
1da177e4 LT |
1951 | |
1952 | /* ------------------------------------------------------------------------ */ | |
1953 | #ifdef CONFIG_PROC_FS | |
1954 | ||
645ca708 | 1955 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
1956 | { |
1957 | struct sock *sk; | |
1958 | struct udp_iter_state *state = seq->private; | |
6f191efe | 1959 | struct net *net = seq_file_net(seq); |
1da177e4 | 1960 | |
f86dcc5a ED |
1961 | for (state->bucket = start; state->bucket <= state->udp_table->mask; |
1962 | ++state->bucket) { | |
88ab1932 | 1963 | struct hlist_nulls_node *node; |
645ca708 | 1964 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
f86dcc5a ED |
1965 | |
1966 | if (hlist_nulls_empty(&hslot->head)) | |
1967 | continue; | |
1968 | ||
645ca708 | 1969 | spin_lock_bh(&hslot->lock); |
88ab1932 | 1970 | sk_nulls_for_each(sk, node, &hslot->head) { |
878628fb | 1971 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 1972 | continue; |
1da177e4 LT |
1973 | if (sk->sk_family == state->family) |
1974 | goto found; | |
1975 | } | |
645ca708 | 1976 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
1977 | } |
1978 | sk = NULL; | |
1979 | found: | |
1980 | return sk; | |
1981 | } | |
1982 | ||
1983 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
1984 | { | |
1985 | struct udp_iter_state *state = seq->private; | |
6f191efe | 1986 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1987 | |
1988 | do { | |
88ab1932 | 1989 | sk = sk_nulls_next(sk); |
878628fb | 1990 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 1991 | |
645ca708 | 1992 | if (!sk) { |
f86dcc5a | 1993 | if (state->bucket <= state->udp_table->mask) |
30842f29 | 1994 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
645ca708 | 1995 | return udp_get_first(seq, state->bucket + 1); |
1da177e4 LT |
1996 | } |
1997 | return sk; | |
1998 | } | |
1999 | ||
2000 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
2001 | { | |
645ca708 | 2002 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
2003 | |
2004 | if (sk) | |
6516c655 | 2005 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
2006 | --pos; |
2007 | return pos ? NULL : sk; | |
2008 | } | |
2009 | ||
2010 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
2011 | { | |
30842f29 | 2012 | struct udp_iter_state *state = seq->private; |
f86dcc5a | 2013 | state->bucket = MAX_UDP_PORTS; |
30842f29 | 2014 | |
b50660f1 | 2015 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
2016 | } |
2017 | ||
2018 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2019 | { | |
2020 | struct sock *sk; | |
2021 | ||
b50660f1 | 2022 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2023 | sk = udp_get_idx(seq, 0); |
2024 | else | |
2025 | sk = udp_get_next(seq, v); | |
2026 | ||
2027 | ++*pos; | |
2028 | return sk; | |
2029 | } | |
2030 | ||
2031 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
2032 | { | |
645ca708 ED |
2033 | struct udp_iter_state *state = seq->private; |
2034 | ||
f86dcc5a | 2035 | if (state->bucket <= state->udp_table->mask) |
645ca708 | 2036 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
1da177e4 LT |
2037 | } |
2038 | ||
73cb88ec | 2039 | int udp_seq_open(struct inode *inode, struct file *file) |
1da177e4 LT |
2040 | { |
2041 | struct udp_seq_afinfo *afinfo = PDE(inode)->data; | |
a2be75c1 DL |
2042 | struct udp_iter_state *s; |
2043 | int err; | |
a91275ef | 2044 | |
a2be75c1 DL |
2045 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2046 | sizeof(struct udp_iter_state)); | |
2047 | if (err < 0) | |
2048 | return err; | |
a91275ef | 2049 | |
a2be75c1 | 2050 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2051 | s->family = afinfo->family; |
645ca708 | 2052 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 2053 | return err; |
a91275ef | 2054 | } |
73cb88ec | 2055 | EXPORT_SYMBOL(udp_seq_open); |
a91275ef | 2056 | |
1da177e4 | 2057 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 2058 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
2059 | { |
2060 | struct proc_dir_entry *p; | |
2061 | int rc = 0; | |
2062 | ||
dda61925 DL |
2063 | afinfo->seq_ops.start = udp_seq_start; |
2064 | afinfo->seq_ops.next = udp_seq_next; | |
2065 | afinfo->seq_ops.stop = udp_seq_stop; | |
2066 | ||
84841c3c | 2067 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2068 | afinfo->seq_fops, afinfo); |
84841c3c | 2069 | if (!p) |
1da177e4 LT |
2070 | rc = -ENOMEM; |
2071 | return rc; | |
2072 | } | |
c482c568 | 2073 | EXPORT_SYMBOL(udp_proc_register); |
1da177e4 | 2074 | |
0c96d8c5 | 2075 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 2076 | { |
0c96d8c5 | 2077 | proc_net_remove(net, afinfo->name); |
1da177e4 | 2078 | } |
c482c568 | 2079 | EXPORT_SYMBOL(udp_proc_unregister); |
db8dac20 DM |
2080 | |
2081 | /* ------------------------------------------------------------------------ */ | |
5e659e4c PE |
2082 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
2083 | int bucket, int *len) | |
db8dac20 DM |
2084 | { |
2085 | struct inet_sock *inet = inet_sk(sp); | |
c720c7e8 ED |
2086 | __be32 dest = inet->inet_daddr; |
2087 | __be32 src = inet->inet_rcv_saddr; | |
2088 | __u16 destp = ntohs(inet->inet_dport); | |
2089 | __u16 srcp = ntohs(inet->inet_sport); | |
db8dac20 | 2090 | |
f86dcc5a | 2091 | seq_printf(f, "%5d: %08X:%04X %08X:%04X" |
71338aa7 | 2092 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n", |
db8dac20 | 2093 | bucket, src, srcp, dest, destp, sp->sk_state, |
31e6d363 ED |
2094 | sk_wmem_alloc_get(sp), |
2095 | sk_rmem_alloc_get(sp), | |
db8dac20 | 2096 | 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp), |
cb61cb9b ED |
2097 | atomic_read(&sp->sk_refcnt), sp, |
2098 | atomic_read(&sp->sk_drops), len); | |
db8dac20 DM |
2099 | } |
2100 | ||
2101 | int udp4_seq_show(struct seq_file *seq, void *v) | |
2102 | { | |
2103 | if (v == SEQ_START_TOKEN) | |
2104 | seq_printf(seq, "%-127s\n", | |
2105 | " sl local_address rem_address st tx_queue " | |
2106 | "rx_queue tr tm->when retrnsmt uid timeout " | |
cb61cb9b | 2107 | "inode ref pointer drops"); |
db8dac20 | 2108 | else { |
db8dac20 | 2109 | struct udp_iter_state *state = seq->private; |
5e659e4c | 2110 | int len; |
db8dac20 | 2111 | |
5e659e4c | 2112 | udp4_format_sock(v, seq, state->bucket, &len); |
c482c568 | 2113 | seq_printf(seq, "%*s\n", 127 - len, ""); |
db8dac20 DM |
2114 | } |
2115 | return 0; | |
2116 | } | |
2117 | ||
73cb88ec AV |
2118 | static const struct file_operations udp_afinfo_seq_fops = { |
2119 | .owner = THIS_MODULE, | |
2120 | .open = udp_seq_open, | |
2121 | .read = seq_read, | |
2122 | .llseek = seq_lseek, | |
2123 | .release = seq_release_net | |
2124 | }; | |
2125 | ||
db8dac20 | 2126 | /* ------------------------------------------------------------------------ */ |
db8dac20 | 2127 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
2128 | .name = "udp", |
2129 | .family = AF_INET, | |
645ca708 | 2130 | .udp_table = &udp_table, |
73cb88ec | 2131 | .seq_fops = &udp_afinfo_seq_fops, |
dda61925 DL |
2132 | .seq_ops = { |
2133 | .show = udp4_seq_show, | |
2134 | }, | |
db8dac20 DM |
2135 | }; |
2136 | ||
2c8c1e72 | 2137 | static int __net_init udp4_proc_init_net(struct net *net) |
15439feb PE |
2138 | { |
2139 | return udp_proc_register(net, &udp4_seq_afinfo); | |
2140 | } | |
2141 | ||
2c8c1e72 | 2142 | static void __net_exit udp4_proc_exit_net(struct net *net) |
15439feb PE |
2143 | { |
2144 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
2145 | } | |
2146 | ||
2147 | static struct pernet_operations udp4_net_ops = { | |
2148 | .init = udp4_proc_init_net, | |
2149 | .exit = udp4_proc_exit_net, | |
2150 | }; | |
2151 | ||
db8dac20 DM |
2152 | int __init udp4_proc_init(void) |
2153 | { | |
15439feb | 2154 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
2155 | } |
2156 | ||
2157 | void udp4_proc_exit(void) | |
2158 | { | |
15439feb | 2159 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 2160 | } |
1da177e4 LT |
2161 | #endif /* CONFIG_PROC_FS */ |
2162 | ||
f86dcc5a ED |
2163 | static __initdata unsigned long uhash_entries; |
2164 | static int __init set_uhash_entries(char *str) | |
645ca708 | 2165 | { |
f86dcc5a ED |
2166 | if (!str) |
2167 | return 0; | |
2168 | uhash_entries = simple_strtoul(str, &str, 0); | |
2169 | if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN) | |
2170 | uhash_entries = UDP_HTABLE_SIZE_MIN; | |
2171 | return 1; | |
2172 | } | |
2173 | __setup("uhash_entries=", set_uhash_entries); | |
645ca708 | 2174 | |
f86dcc5a ED |
2175 | void __init udp_table_init(struct udp_table *table, const char *name) |
2176 | { | |
2177 | unsigned int i; | |
2178 | ||
2179 | if (!CONFIG_BASE_SMALL) | |
2180 | table->hash = alloc_large_system_hash(name, | |
512615b6 | 2181 | 2 * sizeof(struct udp_hslot), |
f86dcc5a ED |
2182 | uhash_entries, |
2183 | 21, /* one slot per 2 MB */ | |
2184 | 0, | |
2185 | &table->log, | |
2186 | &table->mask, | |
2187 | 64 * 1024); | |
2188 | /* | |
2189 | * Make sure hash table has the minimum size | |
2190 | */ | |
2191 | if (CONFIG_BASE_SMALL || table->mask < UDP_HTABLE_SIZE_MIN - 1) { | |
2192 | table->hash = kmalloc(UDP_HTABLE_SIZE_MIN * | |
512615b6 | 2193 | 2 * sizeof(struct udp_hslot), GFP_KERNEL); |
f86dcc5a ED |
2194 | if (!table->hash) |
2195 | panic(name); | |
2196 | table->log = ilog2(UDP_HTABLE_SIZE_MIN); | |
2197 | table->mask = UDP_HTABLE_SIZE_MIN - 1; | |
2198 | } | |
512615b6 | 2199 | table->hash2 = table->hash + (table->mask + 1); |
f86dcc5a | 2200 | for (i = 0; i <= table->mask; i++) { |
88ab1932 | 2201 | INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i); |
fdcc8aa9 | 2202 | table->hash[i].count = 0; |
645ca708 ED |
2203 | spin_lock_init(&table->hash[i].lock); |
2204 | } | |
512615b6 ED |
2205 | for (i = 0; i <= table->mask; i++) { |
2206 | INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i); | |
2207 | table->hash2[i].count = 0; | |
2208 | spin_lock_init(&table->hash2[i].lock); | |
2209 | } | |
645ca708 ED |
2210 | } |
2211 | ||
95766fff HA |
2212 | void __init udp_init(void) |
2213 | { | |
f03d78db | 2214 | unsigned long limit; |
95766fff | 2215 | |
f86dcc5a | 2216 | udp_table_init(&udp_table, "UDP"); |
f03d78db | 2217 | limit = nr_free_buffer_pages() / 8; |
95766fff HA |
2218 | limit = max(limit, 128UL); |
2219 | sysctl_udp_mem[0] = limit / 4 * 3; | |
2220 | sysctl_udp_mem[1] = limit; | |
2221 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
2222 | ||
2223 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
2224 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
2225 | } | |
2226 | ||
d7ca4cc0 SS |
2227 | int udp4_ufo_send_check(struct sk_buff *skb) |
2228 | { | |
2229 | const struct iphdr *iph; | |
2230 | struct udphdr *uh; | |
2231 | ||
2232 | if (!pskb_may_pull(skb, sizeof(*uh))) | |
2233 | return -EINVAL; | |
2234 | ||
2235 | iph = ip_hdr(skb); | |
2236 | uh = udp_hdr(skb); | |
2237 | ||
2238 | uh->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, | |
2239 | IPPROTO_UDP, 0); | |
2240 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
2241 | skb->csum_offset = offsetof(struct udphdr, check); | |
2242 | skb->ip_summed = CHECKSUM_PARTIAL; | |
2243 | return 0; | |
2244 | } | |
2245 | ||
c8f44aff MM |
2246 | struct sk_buff *udp4_ufo_fragment(struct sk_buff *skb, |
2247 | netdev_features_t features) | |
d7ca4cc0 SS |
2248 | { |
2249 | struct sk_buff *segs = ERR_PTR(-EINVAL); | |
2250 | unsigned int mss; | |
2251 | int offset; | |
2252 | __wsum csum; | |
2253 | ||
2254 | mss = skb_shinfo(skb)->gso_size; | |
2255 | if (unlikely(skb->len <= mss)) | |
2256 | goto out; | |
2257 | ||
2258 | if (skb_gso_ok(skb, features | NETIF_F_GSO_ROBUST)) { | |
2259 | /* Packet is from an untrusted source, reset gso_segs. */ | |
2260 | int type = skb_shinfo(skb)->gso_type; | |
2261 | ||
2262 | if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY) || | |
2263 | !(type & (SKB_GSO_UDP)))) | |
2264 | goto out; | |
2265 | ||
2266 | skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss); | |
2267 | ||
2268 | segs = NULL; | |
2269 | goto out; | |
2270 | } | |
2271 | ||
2272 | /* Do software UFO. Complete and fill in the UDP checksum as HW cannot | |
2273 | * do checksum of UDP packets sent as multiple IP fragments. | |
2274 | */ | |
55508d60 | 2275 | offset = skb_checksum_start_offset(skb); |
c482c568 | 2276 | csum = skb_checksum(skb, offset, skb->len - offset, 0); |
d7ca4cc0 SS |
2277 | offset += skb->csum_offset; |
2278 | *(__sum16 *)(skb->data + offset) = csum_fold(csum); | |
2279 | skb->ip_summed = CHECKSUM_NONE; | |
2280 | ||
2281 | /* Fragment the skb. IP headers of the fragments are updated in | |
2282 | * inet_gso_segment() | |
2283 | */ | |
2284 | segs = skb_segment(skb, features); | |
2285 | out: | |
2286 | return segs; | |
2287 | } | |
2288 |