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 | |
1da177e4 LT |
80 | #include <asm/system.h> |
81 | #include <asm/uaccess.h> | |
82 | #include <asm/ioctls.h> | |
95766fff | 83 | #include <linux/bootmem.h> |
8203efb3 ED |
84 | #include <linux/highmem.h> |
85 | #include <linux/swap.h> | |
1da177e4 LT |
86 | #include <linux/types.h> |
87 | #include <linux/fcntl.h> | |
88 | #include <linux/module.h> | |
89 | #include <linux/socket.h> | |
90 | #include <linux/sockios.h> | |
14c85021 | 91 | #include <linux/igmp.h> |
1da177e4 LT |
92 | #include <linux/in.h> |
93 | #include <linux/errno.h> | |
94 | #include <linux/timer.h> | |
95 | #include <linux/mm.h> | |
1da177e4 | 96 | #include <linux/inet.h> |
1da177e4 | 97 | #include <linux/netdevice.h> |
5a0e3ad6 | 98 | #include <linux/slab.h> |
c752f073 | 99 | #include <net/tcp_states.h> |
1da177e4 LT |
100 | #include <linux/skbuff.h> |
101 | #include <linux/proc_fs.h> | |
102 | #include <linux/seq_file.h> | |
457c4cbc | 103 | #include <net/net_namespace.h> |
1da177e4 LT |
104 | #include <net/icmp.h> |
105 | #include <net/route.h> | |
1da177e4 LT |
106 | #include <net/checksum.h> |
107 | #include <net/xfrm.h> | |
ba4e58ec | 108 | #include "udp_impl.h" |
1da177e4 | 109 | |
f86dcc5a | 110 | struct udp_table udp_table __read_mostly; |
645ca708 | 111 | EXPORT_SYMBOL(udp_table); |
1da177e4 | 112 | |
8d987e5c | 113 | long sysctl_udp_mem[3] __read_mostly; |
95766fff | 114 | EXPORT_SYMBOL(sysctl_udp_mem); |
c482c568 ED |
115 | |
116 | int sysctl_udp_rmem_min __read_mostly; | |
95766fff | 117 | EXPORT_SYMBOL(sysctl_udp_rmem_min); |
c482c568 ED |
118 | |
119 | int sysctl_udp_wmem_min __read_mostly; | |
95766fff HA |
120 | EXPORT_SYMBOL(sysctl_udp_wmem_min); |
121 | ||
8d987e5c | 122 | atomic_long_t udp_memory_allocated; |
95766fff HA |
123 | EXPORT_SYMBOL(udp_memory_allocated); |
124 | ||
f86dcc5a ED |
125 | #define MAX_UDP_PORTS 65536 |
126 | #define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN) | |
98322f22 | 127 | |
f24d43c0 | 128 | static int udp_lib_lport_inuse(struct net *net, __u16 num, |
645ca708 | 129 | const struct udp_hslot *hslot, |
98322f22 | 130 | unsigned long *bitmap, |
f24d43c0 ED |
131 | struct sock *sk, |
132 | int (*saddr_comp)(const struct sock *sk1, | |
f86dcc5a ED |
133 | const struct sock *sk2), |
134 | unsigned int log) | |
1da177e4 | 135 | { |
f24d43c0 | 136 | struct sock *sk2; |
88ab1932 | 137 | struct hlist_nulls_node *node; |
25030a7f | 138 | |
88ab1932 | 139 | sk_nulls_for_each(sk2, node, &hslot->head) |
9d4fb27d JP |
140 | if (net_eq(sock_net(sk2), net) && |
141 | sk2 != sk && | |
d4cada4a | 142 | (bitmap || udp_sk(sk2)->udp_port_hash == num) && |
9d4fb27d JP |
143 | (!sk2->sk_reuse || !sk->sk_reuse) && |
144 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
145 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
98322f22 ED |
146 | (*saddr_comp)(sk, sk2)) { |
147 | if (bitmap) | |
d4cada4a ED |
148 | __set_bit(udp_sk(sk2)->udp_port_hash >> log, |
149 | bitmap); | |
98322f22 ED |
150 | else |
151 | return 1; | |
152 | } | |
25030a7f GR |
153 | return 0; |
154 | } | |
155 | ||
30fff923 ED |
156 | /* |
157 | * Note: we still hold spinlock of primary hash chain, so no other writer | |
158 | * can insert/delete a socket with local_port == num | |
159 | */ | |
160 | static int udp_lib_lport_inuse2(struct net *net, __u16 num, | |
161 | struct udp_hslot *hslot2, | |
162 | struct sock *sk, | |
163 | int (*saddr_comp)(const struct sock *sk1, | |
164 | const struct sock *sk2)) | |
165 | { | |
166 | struct sock *sk2; | |
167 | struct hlist_nulls_node *node; | |
168 | int res = 0; | |
169 | ||
170 | spin_lock(&hslot2->lock); | |
171 | udp_portaddr_for_each_entry(sk2, node, &hslot2->head) | |
9d4fb27d JP |
172 | if (net_eq(sock_net(sk2), net) && |
173 | sk2 != sk && | |
174 | (udp_sk(sk2)->udp_port_hash == num) && | |
175 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
176 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
177 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
30fff923 ED |
178 | (*saddr_comp)(sk, sk2)) { |
179 | res = 1; | |
180 | break; | |
181 | } | |
182 | spin_unlock(&hslot2->lock); | |
183 | return res; | |
184 | } | |
185 | ||
25030a7f | 186 | /** |
6ba5a3c5 | 187 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
188 | * |
189 | * @sk: socket struct in question | |
190 | * @snum: port number to look up | |
df2bc459 | 191 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25985edc | 192 | * @hash2_nulladdr: AF-dependent hash value in secondary hash chains, |
30fff923 | 193 | * with NULL address |
25030a7f | 194 | */ |
6ba5a3c5 | 195 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
df2bc459 | 196 | int (*saddr_comp)(const struct sock *sk1, |
30fff923 ED |
197 | const struct sock *sk2), |
198 | unsigned int hash2_nulladdr) | |
25030a7f | 199 | { |
512615b6 | 200 | struct udp_hslot *hslot, *hslot2; |
645ca708 | 201 | struct udp_table *udptable = sk->sk_prot->h.udp_table; |
25030a7f | 202 | int error = 1; |
3b1e0a65 | 203 | struct net *net = sock_net(sk); |
1da177e4 | 204 | |
32c1da70 | 205 | if (!snum) { |
9088c560 ED |
206 | int low, high, remaining; |
207 | unsigned rand; | |
98322f22 ED |
208 | unsigned short first, last; |
209 | DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN); | |
32c1da70 | 210 | |
227b60f5 | 211 | inet_get_local_port_range(&low, &high); |
a25de534 | 212 | remaining = (high - low) + 1; |
227b60f5 | 213 | |
9088c560 | 214 | rand = net_random(); |
98322f22 ED |
215 | first = (((u64)rand * remaining) >> 32) + low; |
216 | /* | |
217 | * force rand to be an odd multiple of UDP_HTABLE_SIZE | |
218 | */ | |
f86dcc5a | 219 | rand = (rand | 1) * (udptable->mask + 1); |
5781b235 ED |
220 | last = first + udptable->mask + 1; |
221 | do { | |
f86dcc5a | 222 | hslot = udp_hashslot(udptable, net, first); |
98322f22 | 223 | bitmap_zero(bitmap, PORTS_PER_CHAIN); |
645ca708 | 224 | spin_lock_bh(&hslot->lock); |
98322f22 | 225 | udp_lib_lport_inuse(net, snum, hslot, bitmap, sk, |
f86dcc5a | 226 | saddr_comp, udptable->log); |
98322f22 ED |
227 | |
228 | snum = first; | |
229 | /* | |
230 | * Iterate on all possible values of snum for this hash. | |
231 | * Using steps of an odd multiple of UDP_HTABLE_SIZE | |
232 | * give us randomization and full range coverage. | |
233 | */ | |
9088c560 | 234 | do { |
98322f22 | 235 | if (low <= snum && snum <= high && |
e3826f1e AW |
236 | !test_bit(snum >> udptable->log, bitmap) && |
237 | !inet_is_reserved_local_port(snum)) | |
98322f22 ED |
238 | goto found; |
239 | snum += rand; | |
240 | } while (snum != first); | |
241 | spin_unlock_bh(&hslot->lock); | |
5781b235 | 242 | } while (++first != last); |
98322f22 | 243 | goto fail; |
645ca708 | 244 | } else { |
f86dcc5a | 245 | hslot = udp_hashslot(udptable, net, snum); |
645ca708 | 246 | spin_lock_bh(&hslot->lock); |
30fff923 ED |
247 | if (hslot->count > 10) { |
248 | int exist; | |
249 | unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum; | |
250 | ||
251 | slot2 &= udptable->mask; | |
252 | hash2_nulladdr &= udptable->mask; | |
253 | ||
254 | hslot2 = udp_hashslot2(udptable, slot2); | |
255 | if (hslot->count < hslot2->count) | |
256 | goto scan_primary_hash; | |
257 | ||
258 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
259 | sk, saddr_comp); | |
260 | if (!exist && (hash2_nulladdr != slot2)) { | |
261 | hslot2 = udp_hashslot2(udptable, hash2_nulladdr); | |
262 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
263 | sk, saddr_comp); | |
264 | } | |
265 | if (exist) | |
266 | goto fail_unlock; | |
267 | else | |
268 | goto found; | |
269 | } | |
270 | scan_primary_hash: | |
f86dcc5a ED |
271 | if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk, |
272 | saddr_comp, 0)) | |
645ca708 ED |
273 | goto fail_unlock; |
274 | } | |
98322f22 | 275 | found: |
c720c7e8 | 276 | inet_sk(sk)->inet_num = snum; |
d4cada4a ED |
277 | udp_sk(sk)->udp_port_hash = snum; |
278 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
1da177e4 | 279 | if (sk_unhashed(sk)) { |
88ab1932 | 280 | sk_nulls_add_node_rcu(sk, &hslot->head); |
fdcc8aa9 | 281 | hslot->count++; |
c29a0bc4 | 282 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
512615b6 ED |
283 | |
284 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
285 | spin_lock(&hslot2->lock); | |
286 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
287 | &hslot2->head); | |
288 | hslot2->count++; | |
289 | spin_unlock(&hslot2->lock); | |
1da177e4 | 290 | } |
25030a7f | 291 | error = 0; |
645ca708 ED |
292 | fail_unlock: |
293 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 294 | fail: |
25030a7f GR |
295 | return error; |
296 | } | |
c482c568 | 297 | EXPORT_SYMBOL(udp_lib_get_port); |
25030a7f | 298 | |
499923c7 | 299 | static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2) |
db8dac20 DM |
300 | { |
301 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
302 | ||
c482c568 | 303 | return (!ipv6_only_sock(sk2) && |
c720c7e8 ED |
304 | (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr || |
305 | inet1->inet_rcv_saddr == inet2->inet_rcv_saddr)); | |
db8dac20 DM |
306 | } |
307 | ||
d4cada4a ED |
308 | static unsigned int udp4_portaddr_hash(struct net *net, __be32 saddr, |
309 | unsigned int port) | |
310 | { | |
0eae88f3 | 311 | return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port; |
d4cada4a ED |
312 | } |
313 | ||
6ba5a3c5 | 314 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 315 | { |
30fff923 | 316 | unsigned int hash2_nulladdr = |
0eae88f3 | 317 | udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum); |
30fff923 ED |
318 | unsigned int hash2_partial = |
319 | udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0); | |
320 | ||
d4cada4a | 321 | /* precompute partial secondary hash */ |
30fff923 ED |
322 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
323 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr); | |
db8dac20 DM |
324 | } |
325 | ||
645ca708 ED |
326 | static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr, |
327 | unsigned short hnum, | |
328 | __be16 sport, __be32 daddr, __be16 dport, int dif) | |
329 | { | |
330 | int score = -1; | |
331 | ||
d4cada4a | 332 | if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum && |
645ca708 ED |
333 | !ipv6_only_sock(sk)) { |
334 | struct inet_sock *inet = inet_sk(sk); | |
335 | ||
336 | score = (sk->sk_family == PF_INET ? 1 : 0); | |
c720c7e8 ED |
337 | if (inet->inet_rcv_saddr) { |
338 | if (inet->inet_rcv_saddr != daddr) | |
645ca708 ED |
339 | return -1; |
340 | score += 2; | |
341 | } | |
c720c7e8 ED |
342 | if (inet->inet_daddr) { |
343 | if (inet->inet_daddr != saddr) | |
645ca708 ED |
344 | return -1; |
345 | score += 2; | |
346 | } | |
c720c7e8 ED |
347 | if (inet->inet_dport) { |
348 | if (inet->inet_dport != sport) | |
645ca708 ED |
349 | return -1; |
350 | score += 2; | |
351 | } | |
352 | if (sk->sk_bound_dev_if) { | |
353 | if (sk->sk_bound_dev_if != dif) | |
354 | return -1; | |
355 | score += 2; | |
356 | } | |
357 | } | |
358 | return score; | |
359 | } | |
360 | ||
5051ebd2 ED |
361 | /* |
362 | * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num) | |
363 | */ | |
364 | #define SCORE2_MAX (1 + 2 + 2 + 2) | |
365 | static inline int compute_score2(struct sock *sk, struct net *net, | |
366 | __be32 saddr, __be16 sport, | |
367 | __be32 daddr, unsigned int hnum, int dif) | |
368 | { | |
369 | int score = -1; | |
370 | ||
371 | if (net_eq(sock_net(sk), net) && !ipv6_only_sock(sk)) { | |
372 | struct inet_sock *inet = inet_sk(sk); | |
373 | ||
374 | if (inet->inet_rcv_saddr != daddr) | |
375 | return -1; | |
376 | if (inet->inet_num != hnum) | |
377 | return -1; | |
378 | ||
379 | score = (sk->sk_family == PF_INET ? 1 : 0); | |
380 | if (inet->inet_daddr) { | |
381 | if (inet->inet_daddr != saddr) | |
382 | return -1; | |
383 | score += 2; | |
384 | } | |
385 | if (inet->inet_dport) { | |
386 | if (inet->inet_dport != sport) | |
387 | return -1; | |
388 | score += 2; | |
389 | } | |
390 | if (sk->sk_bound_dev_if) { | |
391 | if (sk->sk_bound_dev_if != dif) | |
392 | return -1; | |
393 | score += 2; | |
394 | } | |
395 | } | |
396 | return score; | |
397 | } | |
398 | ||
5051ebd2 ED |
399 | |
400 | /* called with read_rcu_lock() */ | |
401 | static struct sock *udp4_lib_lookup2(struct net *net, | |
402 | __be32 saddr, __be16 sport, | |
403 | __be32 daddr, unsigned int hnum, int dif, | |
404 | struct udp_hslot *hslot2, unsigned int slot2) | |
405 | { | |
406 | struct sock *sk, *result; | |
407 | struct hlist_nulls_node *node; | |
408 | int score, badness; | |
409 | ||
410 | begin: | |
411 | result = NULL; | |
412 | badness = -1; | |
413 | udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { | |
414 | score = compute_score2(sk, net, saddr, sport, | |
415 | daddr, hnum, dif); | |
416 | if (score > badness) { | |
417 | result = sk; | |
418 | badness = score; | |
419 | if (score == SCORE2_MAX) | |
420 | goto exact_match; | |
421 | } | |
422 | } | |
423 | /* | |
424 | * if the nulls value we got at the end of this lookup is | |
425 | * not the expected one, we must restart lookup. | |
426 | * We probably met an item that was moved to another chain. | |
427 | */ | |
428 | if (get_nulls_value(node) != slot2) | |
429 | goto begin; | |
430 | ||
431 | if (result) { | |
432 | exact_match: | |
c31504dc | 433 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
5051ebd2 ED |
434 | result = NULL; |
435 | else if (unlikely(compute_score2(result, net, saddr, sport, | |
436 | daddr, hnum, dif) < badness)) { | |
437 | sock_put(result); | |
438 | goto begin; | |
439 | } | |
440 | } | |
441 | return result; | |
442 | } | |
443 | ||
db8dac20 DM |
444 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
445 | * harder than this. -DaveM | |
446 | */ | |
447 | static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, | |
448 | __be16 sport, __be32 daddr, __be16 dport, | |
645ca708 | 449 | int dif, struct udp_table *udptable) |
db8dac20 | 450 | { |
271b72c7 | 451 | struct sock *sk, *result; |
88ab1932 | 452 | struct hlist_nulls_node *node; |
db8dac20 | 453 | unsigned short hnum = ntohs(dport); |
5051ebd2 ED |
454 | unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); |
455 | struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; | |
271b72c7 | 456 | int score, badness; |
645ca708 | 457 | |
271b72c7 | 458 | rcu_read_lock(); |
5051ebd2 ED |
459 | if (hslot->count > 10) { |
460 | hash2 = udp4_portaddr_hash(net, daddr, hnum); | |
461 | slot2 = hash2 & udptable->mask; | |
462 | hslot2 = &udptable->hash2[slot2]; | |
463 | if (hslot->count < hslot2->count) | |
464 | goto begin; | |
465 | ||
466 | result = udp4_lib_lookup2(net, saddr, sport, | |
467 | daddr, hnum, dif, | |
468 | hslot2, slot2); | |
469 | if (!result) { | |
0eae88f3 | 470 | hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum); |
5051ebd2 ED |
471 | slot2 = hash2 & udptable->mask; |
472 | hslot2 = &udptable->hash2[slot2]; | |
473 | if (hslot->count < hslot2->count) | |
474 | goto begin; | |
475 | ||
1223c67c | 476 | result = udp4_lib_lookup2(net, saddr, sport, |
0eae88f3 | 477 | htonl(INADDR_ANY), hnum, dif, |
5051ebd2 ED |
478 | hslot2, slot2); |
479 | } | |
480 | rcu_read_unlock(); | |
481 | return result; | |
482 | } | |
271b72c7 ED |
483 | begin: |
484 | result = NULL; | |
485 | badness = -1; | |
88ab1932 | 486 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { |
645ca708 ED |
487 | score = compute_score(sk, net, saddr, hnum, sport, |
488 | daddr, dport, dif); | |
489 | if (score > badness) { | |
490 | result = sk; | |
491 | badness = score; | |
db8dac20 DM |
492 | } |
493 | } | |
88ab1932 ED |
494 | /* |
495 | * if the nulls value we got at the end of this lookup is | |
496 | * not the expected one, we must restart lookup. | |
497 | * We probably met an item that was moved to another chain. | |
498 | */ | |
5051ebd2 | 499 | if (get_nulls_value(node) != slot) |
88ab1932 ED |
500 | goto begin; |
501 | ||
271b72c7 | 502 | if (result) { |
c31504dc | 503 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
271b72c7 ED |
504 | result = NULL; |
505 | else if (unlikely(compute_score(result, net, saddr, hnum, sport, | |
506 | daddr, dport, dif) < badness)) { | |
507 | sock_put(result); | |
508 | goto begin; | |
509 | } | |
510 | } | |
511 | rcu_read_unlock(); | |
db8dac20 DM |
512 | return result; |
513 | } | |
514 | ||
607c4aaf KK |
515 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
516 | __be16 sport, __be16 dport, | |
645ca708 | 517 | struct udp_table *udptable) |
607c4aaf | 518 | { |
23542618 | 519 | struct sock *sk; |
607c4aaf KK |
520 | const struct iphdr *iph = ip_hdr(skb); |
521 | ||
23542618 KK |
522 | if (unlikely(sk = skb_steal_sock(skb))) |
523 | return sk; | |
524 | else | |
adf30907 | 525 | return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport, |
23542618 KK |
526 | iph->daddr, dport, inet_iif(skb), |
527 | udptable); | |
607c4aaf KK |
528 | } |
529 | ||
bcd41303 KK |
530 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
531 | __be32 daddr, __be16 dport, int dif) | |
532 | { | |
645ca708 | 533 | return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table); |
bcd41303 KK |
534 | } |
535 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
536 | ||
920a4611 | 537 | static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk, |
db8dac20 DM |
538 | __be16 loc_port, __be32 loc_addr, |
539 | __be16 rmt_port, __be32 rmt_addr, | |
540 | int dif) | |
541 | { | |
88ab1932 | 542 | struct hlist_nulls_node *node; |
db8dac20 DM |
543 | struct sock *s = sk; |
544 | unsigned short hnum = ntohs(loc_port); | |
545 | ||
88ab1932 | 546 | sk_nulls_for_each_from(s, node) { |
db8dac20 DM |
547 | struct inet_sock *inet = inet_sk(s); |
548 | ||
9d4fb27d JP |
549 | if (!net_eq(sock_net(s), net) || |
550 | udp_sk(s)->udp_port_hash != hnum || | |
551 | (inet->inet_daddr && inet->inet_daddr != rmt_addr) || | |
552 | (inet->inet_dport != rmt_port && inet->inet_dport) || | |
553 | (inet->inet_rcv_saddr && | |
554 | inet->inet_rcv_saddr != loc_addr) || | |
555 | ipv6_only_sock(s) || | |
db8dac20 DM |
556 | (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)) |
557 | continue; | |
558 | if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif)) | |
559 | continue; | |
560 | goto found; | |
561 | } | |
562 | s = NULL; | |
563 | found: | |
564 | return s; | |
565 | } | |
566 | ||
567 | /* | |
568 | * This routine is called by the ICMP module when it gets some | |
569 | * sort of error condition. If err < 0 then the socket should | |
570 | * be closed and the error returned to the user. If err > 0 | |
571 | * it's just the icmp type << 8 | icmp code. | |
572 | * Header points to the ip header of the error packet. We move | |
573 | * on past this. Then (as it used to claim before adjustment) | |
574 | * header points to the first 8 bytes of the udp header. We need | |
575 | * to find the appropriate port. | |
576 | */ | |
577 | ||
645ca708 | 578 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
579 | { |
580 | struct inet_sock *inet; | |
b71d1d42 | 581 | const struct iphdr *iph = (const struct iphdr *)skb->data; |
c482c568 | 582 | struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2)); |
db8dac20 DM |
583 | const int type = icmp_hdr(skb)->type; |
584 | const int code = icmp_hdr(skb)->code; | |
585 | struct sock *sk; | |
586 | int harderr; | |
587 | int err; | |
fd54d716 | 588 | struct net *net = dev_net(skb->dev); |
db8dac20 | 589 | |
fd54d716 | 590 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
db8dac20 DM |
591 | iph->saddr, uh->source, skb->dev->ifindex, udptable); |
592 | if (sk == NULL) { | |
dcfc23ca | 593 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
594 | return; /* No socket for error */ |
595 | } | |
596 | ||
597 | err = 0; | |
598 | harderr = 0; | |
599 | inet = inet_sk(sk); | |
600 | ||
601 | switch (type) { | |
602 | default: | |
603 | case ICMP_TIME_EXCEEDED: | |
604 | err = EHOSTUNREACH; | |
605 | break; | |
606 | case ICMP_SOURCE_QUENCH: | |
607 | goto out; | |
608 | case ICMP_PARAMETERPROB: | |
609 | err = EPROTO; | |
610 | harderr = 1; | |
611 | break; | |
612 | case ICMP_DEST_UNREACH: | |
613 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
614 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { | |
615 | err = EMSGSIZE; | |
616 | harderr = 1; | |
617 | break; | |
618 | } | |
619 | goto out; | |
620 | } | |
621 | err = EHOSTUNREACH; | |
622 | if (code <= NR_ICMP_UNREACH) { | |
623 | harderr = icmp_err_convert[code].fatal; | |
624 | err = icmp_err_convert[code].errno; | |
625 | } | |
626 | break; | |
627 | } | |
628 | ||
629 | /* | |
630 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
631 | * 4.1.3.3. | |
632 | */ | |
633 | if (!inet->recverr) { | |
634 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
635 | goto out; | |
b1faf566 | 636 | } else |
c482c568 | 637 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1)); |
b1faf566 | 638 | |
db8dac20 DM |
639 | sk->sk_err = err; |
640 | sk->sk_error_report(sk); | |
641 | out: | |
642 | sock_put(sk); | |
643 | } | |
644 | ||
645 | void udp_err(struct sk_buff *skb, u32 info) | |
646 | { | |
645ca708 | 647 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
648 | } |
649 | ||
650 | /* | |
651 | * Throw away all pending data and cancel the corking. Socket is locked. | |
652 | */ | |
36d926b9 | 653 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
654 | { |
655 | struct udp_sock *up = udp_sk(sk); | |
656 | ||
657 | if (up->pending) { | |
658 | up->len = 0; | |
659 | up->pending = 0; | |
660 | ip_flush_pending_frames(sk); | |
661 | } | |
662 | } | |
36d926b9 | 663 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
664 | |
665 | /** | |
f6b9664f | 666 | * udp4_hwcsum - handle outgoing HW checksumming |
db8dac20 DM |
667 | * @skb: sk_buff containing the filled-in UDP header |
668 | * (checksum field must be zeroed out) | |
f6b9664f HX |
669 | * @src: source IP address |
670 | * @dst: destination IP address | |
db8dac20 | 671 | */ |
f6b9664f | 672 | static void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst) |
db8dac20 | 673 | { |
db8dac20 | 674 | struct udphdr *uh = udp_hdr(skb); |
f6b9664f HX |
675 | struct sk_buff *frags = skb_shinfo(skb)->frag_list; |
676 | int offset = skb_transport_offset(skb); | |
677 | int len = skb->len - offset; | |
678 | int hlen = len; | |
db8dac20 DM |
679 | __wsum csum = 0; |
680 | ||
f6b9664f | 681 | if (!frags) { |
db8dac20 DM |
682 | /* |
683 | * Only one fragment on the socket. | |
684 | */ | |
685 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
686 | skb->csum_offset = offsetof(struct udphdr, check); | |
f6b9664f HX |
687 | uh->check = ~csum_tcpudp_magic(src, dst, len, |
688 | IPPROTO_UDP, 0); | |
db8dac20 DM |
689 | } else { |
690 | /* | |
691 | * HW-checksum won't work as there are two or more | |
692 | * fragments on the socket so that all csums of sk_buffs | |
693 | * should be together | |
694 | */ | |
f6b9664f HX |
695 | do { |
696 | csum = csum_add(csum, frags->csum); | |
697 | hlen -= frags->len; | |
698 | } while ((frags = frags->next)); | |
db8dac20 | 699 | |
f6b9664f | 700 | csum = skb_checksum(skb, offset, hlen, csum); |
db8dac20 DM |
701 | skb->ip_summed = CHECKSUM_NONE; |
702 | ||
db8dac20 DM |
703 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); |
704 | if (uh->check == 0) | |
705 | uh->check = CSUM_MANGLED_0; | |
706 | } | |
707 | } | |
708 | ||
f6b9664f | 709 | static int udp_send_skb(struct sk_buff *skb, __be32 daddr, __be32 dport) |
db8dac20 | 710 | { |
f6b9664f | 711 | struct sock *sk = skb->sk; |
db8dac20 | 712 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 | 713 | struct udphdr *uh; |
f6b9664f | 714 | struct rtable *rt = (struct rtable *)skb_dst(skb); |
db8dac20 DM |
715 | int err = 0; |
716 | int is_udplite = IS_UDPLITE(sk); | |
f6b9664f HX |
717 | int offset = skb_transport_offset(skb); |
718 | int len = skb->len - offset; | |
db8dac20 DM |
719 | __wsum csum = 0; |
720 | ||
db8dac20 DM |
721 | /* |
722 | * Create a UDP header | |
723 | */ | |
724 | uh = udp_hdr(skb); | |
f6b9664f HX |
725 | uh->source = inet->inet_sport; |
726 | uh->dest = dport; | |
727 | uh->len = htons(len); | |
db8dac20 DM |
728 | uh->check = 0; |
729 | ||
730 | if (is_udplite) /* UDP-Lite */ | |
f6b9664f | 731 | csum = udplite_csum(skb); |
db8dac20 DM |
732 | |
733 | else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */ | |
734 | ||
735 | skb->ip_summed = CHECKSUM_NONE; | |
736 | goto send; | |
737 | ||
738 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
739 | ||
f6b9664f | 740 | udp4_hwcsum(skb, rt->rt_src, daddr); |
db8dac20 DM |
741 | goto send; |
742 | ||
f6b9664f HX |
743 | } else |
744 | csum = udp_csum(skb); | |
db8dac20 DM |
745 | |
746 | /* add protocol-dependent pseudo-header */ | |
f6b9664f | 747 | uh->check = csum_tcpudp_magic(rt->rt_src, daddr, len, |
c482c568 | 748 | sk->sk_protocol, csum); |
db8dac20 DM |
749 | if (uh->check == 0) |
750 | uh->check = CSUM_MANGLED_0; | |
751 | ||
752 | send: | |
f6b9664f | 753 | err = ip_send_skb(skb); |
6ce9e7b5 ED |
754 | if (err) { |
755 | if (err == -ENOBUFS && !inet->recverr) { | |
756 | UDP_INC_STATS_USER(sock_net(sk), | |
757 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
758 | err = 0; | |
759 | } | |
760 | } else | |
761 | UDP_INC_STATS_USER(sock_net(sk), | |
762 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
f6b9664f HX |
763 | return err; |
764 | } | |
765 | ||
766 | /* | |
767 | * Push out all pending data as one UDP datagram. Socket is locked. | |
768 | */ | |
769 | static int udp_push_pending_frames(struct sock *sk) | |
770 | { | |
771 | struct udp_sock *up = udp_sk(sk); | |
772 | struct inet_sock *inet = inet_sk(sk); | |
b6f21b26 | 773 | struct flowi4 *fl4 = &inet->cork.fl.u.ip4; |
f6b9664f HX |
774 | struct sk_buff *skb; |
775 | int err = 0; | |
776 | ||
77968b78 | 777 | skb = ip_finish_skb(sk, fl4); |
f6b9664f HX |
778 | if (!skb) |
779 | goto out; | |
780 | ||
9cce96df | 781 | err = udp_send_skb(skb, fl4->daddr, fl4->fl4_dport); |
f6b9664f | 782 | |
db8dac20 DM |
783 | out: |
784 | up->len = 0; | |
785 | up->pending = 0; | |
db8dac20 DM |
786 | return err; |
787 | } | |
788 | ||
789 | int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
790 | size_t len) | |
791 | { | |
792 | struct inet_sock *inet = inet_sk(sk); | |
793 | struct udp_sock *up = udp_sk(sk); | |
e474995f | 794 | struct flowi4 fl4_stack; |
b6f21b26 | 795 | struct flowi4 *fl4; |
db8dac20 DM |
796 | int ulen = len; |
797 | struct ipcm_cookie ipc; | |
798 | struct rtable *rt = NULL; | |
799 | int free = 0; | |
800 | int connected = 0; | |
801 | __be32 daddr, faddr, saddr; | |
802 | __be16 dport; | |
803 | u8 tos; | |
804 | int err, is_udplite = IS_UDPLITE(sk); | |
805 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
806 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
903ab86d | 807 | struct sk_buff *skb; |
f6d8bd05 | 808 | struct ip_options_data opt_copy; |
db8dac20 DM |
809 | |
810 | if (len > 0xFFFF) | |
811 | return -EMSGSIZE; | |
812 | ||
813 | /* | |
814 | * Check the flags. | |
815 | */ | |
816 | ||
c482c568 | 817 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */ |
db8dac20 DM |
818 | return -EOPNOTSUPP; |
819 | ||
820 | ipc.opt = NULL; | |
2244d07b | 821 | ipc.tx_flags = 0; |
db8dac20 | 822 | |
903ab86d HX |
823 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
824 | ||
f5fca608 | 825 | fl4 = &inet->cork.fl.u.ip4; |
db8dac20 DM |
826 | if (up->pending) { |
827 | /* | |
828 | * There are pending frames. | |
829 | * The socket lock must be held while it's corked. | |
830 | */ | |
831 | lock_sock(sk); | |
832 | if (likely(up->pending)) { | |
833 | if (unlikely(up->pending != AF_INET)) { | |
834 | release_sock(sk); | |
835 | return -EINVAL; | |
836 | } | |
837 | goto do_append_data; | |
838 | } | |
839 | release_sock(sk); | |
840 | } | |
841 | ulen += sizeof(struct udphdr); | |
842 | ||
843 | /* | |
844 | * Get and verify the address. | |
845 | */ | |
846 | if (msg->msg_name) { | |
c482c568 | 847 | struct sockaddr_in * usin = (struct sockaddr_in *)msg->msg_name; |
db8dac20 DM |
848 | if (msg->msg_namelen < sizeof(*usin)) |
849 | return -EINVAL; | |
850 | if (usin->sin_family != AF_INET) { | |
851 | if (usin->sin_family != AF_UNSPEC) | |
852 | return -EAFNOSUPPORT; | |
853 | } | |
854 | ||
855 | daddr = usin->sin_addr.s_addr; | |
856 | dport = usin->sin_port; | |
857 | if (dport == 0) | |
858 | return -EINVAL; | |
859 | } else { | |
860 | if (sk->sk_state != TCP_ESTABLISHED) | |
861 | return -EDESTADDRREQ; | |
c720c7e8 ED |
862 | daddr = inet->inet_daddr; |
863 | dport = inet->inet_dport; | |
db8dac20 DM |
864 | /* Open fast path for connected socket. |
865 | Route will not be used, if at least one option is set. | |
866 | */ | |
867 | connected = 1; | |
868 | } | |
c720c7e8 | 869 | ipc.addr = inet->inet_saddr; |
db8dac20 DM |
870 | |
871 | ipc.oif = sk->sk_bound_dev_if; | |
2244d07b | 872 | err = sock_tx_timestamp(sk, &ipc.tx_flags); |
51f31cab PO |
873 | if (err) |
874 | return err; | |
db8dac20 | 875 | if (msg->msg_controllen) { |
3b1e0a65 | 876 | err = ip_cmsg_send(sock_net(sk), msg, &ipc); |
db8dac20 DM |
877 | if (err) |
878 | return err; | |
879 | if (ipc.opt) | |
880 | free = 1; | |
881 | connected = 0; | |
882 | } | |
f6d8bd05 ED |
883 | if (!ipc.opt) { |
884 | struct ip_options_rcu *inet_opt; | |
885 | ||
886 | rcu_read_lock(); | |
887 | inet_opt = rcu_dereference(inet->inet_opt); | |
888 | if (inet_opt) { | |
889 | memcpy(&opt_copy, inet_opt, | |
890 | sizeof(*inet_opt) + inet_opt->opt.optlen); | |
891 | ipc.opt = &opt_copy.opt; | |
892 | } | |
893 | rcu_read_unlock(); | |
894 | } | |
db8dac20 DM |
895 | |
896 | saddr = ipc.addr; | |
897 | ipc.addr = faddr = daddr; | |
898 | ||
f6d8bd05 | 899 | if (ipc.opt && ipc.opt->opt.srr) { |
db8dac20 DM |
900 | if (!daddr) |
901 | return -EINVAL; | |
f6d8bd05 | 902 | faddr = ipc.opt->opt.faddr; |
db8dac20 DM |
903 | connected = 0; |
904 | } | |
905 | tos = RT_TOS(inet->tos); | |
906 | if (sock_flag(sk, SOCK_LOCALROUTE) || | |
907 | (msg->msg_flags & MSG_DONTROUTE) || | |
f6d8bd05 | 908 | (ipc.opt && ipc.opt->opt.is_strictroute)) { |
db8dac20 DM |
909 | tos |= RTO_ONLINK; |
910 | connected = 0; | |
911 | } | |
912 | ||
913 | if (ipv4_is_multicast(daddr)) { | |
914 | if (!ipc.oif) | |
915 | ipc.oif = inet->mc_index; | |
916 | if (!saddr) | |
917 | saddr = inet->mc_addr; | |
918 | connected = 0; | |
919 | } | |
920 | ||
921 | if (connected) | |
c482c568 | 922 | rt = (struct rtable *)sk_dst_check(sk, 0); |
db8dac20 DM |
923 | |
924 | if (rt == NULL) { | |
84a3aa00 PE |
925 | struct net *net = sock_net(sk); |
926 | ||
e474995f DM |
927 | fl4 = &fl4_stack; |
928 | flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos, | |
c0951cbc DM |
929 | RT_SCOPE_UNIVERSE, sk->sk_protocol, |
930 | inet_sk_flowi_flags(sk)|FLOWI_FLAG_CAN_SLEEP, | |
931 | faddr, saddr, dport, inet->inet_sport); | |
932 | ||
e474995f DM |
933 | security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); |
934 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe DM |
935 | if (IS_ERR(rt)) { |
936 | err = PTR_ERR(rt); | |
06dc94b1 | 937 | rt = NULL; |
db8dac20 | 938 | if (err == -ENETUNREACH) |
7c73a6fa | 939 | IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
940 | goto out; |
941 | } | |
942 | ||
943 | err = -EACCES; | |
944 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
945 | !sock_flag(sk, SOCK_BROADCAST)) | |
946 | goto out; | |
947 | if (connected) | |
d8d1f30b | 948 | sk_dst_set(sk, dst_clone(&rt->dst)); |
db8dac20 DM |
949 | } |
950 | ||
951 | if (msg->msg_flags&MSG_CONFIRM) | |
952 | goto do_confirm; | |
953 | back_from_confirm: | |
954 | ||
e474995f | 955 | saddr = fl4->saddr; |
db8dac20 | 956 | if (!ipc.addr) |
e474995f | 957 | daddr = ipc.addr = fl4->daddr; |
db8dac20 | 958 | |
903ab86d HX |
959 | /* Lockless fast path for the non-corking case. */ |
960 | if (!corkreq) { | |
77968b78 | 961 | skb = ip_make_skb(sk, fl4, getfrag, msg->msg_iov, ulen, |
903ab86d HX |
962 | sizeof(struct udphdr), &ipc, &rt, |
963 | msg->msg_flags); | |
964 | err = PTR_ERR(skb); | |
965 | if (skb && !IS_ERR(skb)) | |
966 | err = udp_send_skb(skb, daddr, dport); | |
967 | goto out; | |
968 | } | |
969 | ||
db8dac20 DM |
970 | lock_sock(sk); |
971 | if (unlikely(up->pending)) { | |
972 | /* The socket is already corked while preparing it. */ | |
973 | /* ... which is an evident application bug. --ANK */ | |
974 | release_sock(sk); | |
975 | ||
976 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n"); | |
977 | err = -EINVAL; | |
978 | goto out; | |
979 | } | |
980 | /* | |
981 | * Now cork the socket to pend data. | |
982 | */ | |
b6f21b26 DM |
983 | fl4 = &inet->cork.fl.u.ip4; |
984 | fl4->daddr = daddr; | |
985 | fl4->saddr = saddr; | |
9cce96df DM |
986 | fl4->fl4_dport = dport; |
987 | fl4->fl4_sport = inet->inet_sport; | |
db8dac20 DM |
988 | up->pending = AF_INET; |
989 | ||
990 | do_append_data: | |
991 | up->len += ulen; | |
f5fca608 DM |
992 | err = ip_append_data(sk, fl4, getfrag, msg->msg_iov, ulen, |
993 | sizeof(struct udphdr), &ipc, &rt, | |
994 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | |
db8dac20 DM |
995 | if (err) |
996 | udp_flush_pending_frames(sk); | |
997 | else if (!corkreq) | |
998 | err = udp_push_pending_frames(sk); | |
999 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
1000 | up->pending = 0; | |
1001 | release_sock(sk); | |
1002 | ||
1003 | out: | |
1004 | ip_rt_put(rt); | |
1005 | if (free) | |
1006 | kfree(ipc.opt); | |
1007 | if (!err) | |
1008 | return len; | |
1009 | /* | |
1010 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1011 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1012 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1013 | * things). We could add another new stat but at least for now that | |
1014 | * seems like overkill. | |
1015 | */ | |
1016 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
1017 | UDP_INC_STATS_USER(sock_net(sk), |
1018 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
1019 | } |
1020 | return err; | |
1021 | ||
1022 | do_confirm: | |
d8d1f30b | 1023 | dst_confirm(&rt->dst); |
db8dac20 DM |
1024 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1025 | goto back_from_confirm; | |
1026 | err = 0; | |
1027 | goto out; | |
1028 | } | |
c482c568 | 1029 | EXPORT_SYMBOL(udp_sendmsg); |
db8dac20 DM |
1030 | |
1031 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
1032 | size_t size, int flags) | |
1033 | { | |
f5fca608 | 1034 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
1035 | struct udp_sock *up = udp_sk(sk); |
1036 | int ret; | |
1037 | ||
1038 | if (!up->pending) { | |
1039 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
1040 | ||
1041 | /* Call udp_sendmsg to specify destination address which | |
1042 | * sendpage interface can't pass. | |
1043 | * This will succeed only when the socket is connected. | |
1044 | */ | |
1045 | ret = udp_sendmsg(NULL, sk, &msg, 0); | |
1046 | if (ret < 0) | |
1047 | return ret; | |
1048 | } | |
1049 | ||
1050 | lock_sock(sk); | |
1051 | ||
1052 | if (unlikely(!up->pending)) { | |
1053 | release_sock(sk); | |
1054 | ||
1055 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n"); | |
1056 | return -EINVAL; | |
1057 | } | |
1058 | ||
f5fca608 DM |
1059 | ret = ip_append_page(sk, &inet->cork.fl.u.ip4, |
1060 | page, offset, size, flags); | |
db8dac20 DM |
1061 | if (ret == -EOPNOTSUPP) { |
1062 | release_sock(sk); | |
1063 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
1064 | size, flags); | |
1065 | } | |
1066 | if (ret < 0) { | |
1067 | udp_flush_pending_frames(sk); | |
1068 | goto out; | |
1069 | } | |
1070 | ||
1071 | up->len += size; | |
1072 | if (!(up->corkflag || (flags&MSG_MORE))) | |
1073 | ret = udp_push_pending_frames(sk); | |
1074 | if (!ret) | |
1075 | ret = size; | |
1076 | out: | |
1077 | release_sock(sk); | |
1078 | return ret; | |
1079 | } | |
1080 | ||
85584672 ED |
1081 | |
1082 | /** | |
1083 | * first_packet_length - return length of first packet in receive queue | |
1084 | * @sk: socket | |
1085 | * | |
1086 | * Drops all bad checksum frames, until a valid one is found. | |
1087 | * Returns the length of found skb, or 0 if none is found. | |
1088 | */ | |
1089 | static unsigned int first_packet_length(struct sock *sk) | |
1090 | { | |
1091 | struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue; | |
1092 | struct sk_buff *skb; | |
1093 | unsigned int res; | |
1094 | ||
1095 | __skb_queue_head_init(&list_kill); | |
1096 | ||
1097 | spin_lock_bh(&rcvq->lock); | |
1098 | while ((skb = skb_peek(rcvq)) != NULL && | |
1099 | udp_lib_checksum_complete(skb)) { | |
1100 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, | |
1101 | IS_UDPLITE(sk)); | |
8edf19c2 | 1102 | atomic_inc(&sk->sk_drops); |
85584672 ED |
1103 | __skb_unlink(skb, rcvq); |
1104 | __skb_queue_tail(&list_kill, skb); | |
1105 | } | |
1106 | res = skb ? skb->len : 0; | |
1107 | spin_unlock_bh(&rcvq->lock); | |
1108 | ||
1109 | if (!skb_queue_empty(&list_kill)) { | |
8a74ad60 ED |
1110 | bool slow = lock_sock_fast(sk); |
1111 | ||
85584672 ED |
1112 | __skb_queue_purge(&list_kill); |
1113 | sk_mem_reclaim_partial(sk); | |
8a74ad60 | 1114 | unlock_sock_fast(sk, slow); |
85584672 ED |
1115 | } |
1116 | return res; | |
1117 | } | |
1118 | ||
1da177e4 LT |
1119 | /* |
1120 | * IOCTL requests applicable to the UDP protocol | |
1121 | */ | |
e905a9ed | 1122 | |
1da177e4 LT |
1123 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
1124 | { | |
6516c655 SH |
1125 | switch (cmd) { |
1126 | case SIOCOUTQ: | |
1da177e4 | 1127 | { |
31e6d363 ED |
1128 | int amount = sk_wmem_alloc_get(sk); |
1129 | ||
6516c655 SH |
1130 | return put_user(amount, (int __user *)arg); |
1131 | } | |
1da177e4 | 1132 | |
6516c655 SH |
1133 | case SIOCINQ: |
1134 | { | |
85584672 | 1135 | unsigned int amount = first_packet_length(sk); |
6516c655 | 1136 | |
85584672 | 1137 | if (amount) |
6516c655 SH |
1138 | /* |
1139 | * We will only return the amount | |
1140 | * of this packet since that is all | |
1141 | * that will be read. | |
1142 | */ | |
85584672 ED |
1143 | amount -= sizeof(struct udphdr); |
1144 | ||
6516c655 SH |
1145 | return put_user(amount, (int __user *)arg); |
1146 | } | |
1da177e4 | 1147 | |
6516c655 SH |
1148 | default: |
1149 | return -ENOIOCTLCMD; | |
1da177e4 | 1150 | } |
6516c655 SH |
1151 | |
1152 | return 0; | |
1da177e4 | 1153 | } |
c482c568 | 1154 | EXPORT_SYMBOL(udp_ioctl); |
1da177e4 | 1155 | |
db8dac20 DM |
1156 | /* |
1157 | * This should be easy, if there is something there we | |
1158 | * return it, otherwise we block. | |
1159 | */ | |
1160 | ||
1161 | int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
1162 | size_t len, int noblock, int flags, int *addr_len) | |
1163 | { | |
1164 | struct inet_sock *inet = inet_sk(sk); | |
1165 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; | |
1166 | struct sk_buff *skb; | |
81d54ec8 | 1167 | unsigned int ulen; |
db8dac20 DM |
1168 | int peeked; |
1169 | int err; | |
1170 | int is_udplite = IS_UDPLITE(sk); | |
8a74ad60 | 1171 | bool slow; |
db8dac20 DM |
1172 | |
1173 | /* | |
1174 | * Check any passed addresses | |
1175 | */ | |
1176 | if (addr_len) | |
c482c568 | 1177 | *addr_len = sizeof(*sin); |
db8dac20 DM |
1178 | |
1179 | if (flags & MSG_ERRQUEUE) | |
1180 | return ip_recv_error(sk, msg, len); | |
1181 | ||
1182 | try_again: | |
1183 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), | |
1184 | &peeked, &err); | |
1185 | if (!skb) | |
1186 | goto out; | |
1187 | ||
1188 | ulen = skb->len - sizeof(struct udphdr); | |
81d54ec8 GR |
1189 | if (len > ulen) |
1190 | len = ulen; | |
1191 | else if (len < ulen) | |
db8dac20 DM |
1192 | msg->msg_flags |= MSG_TRUNC; |
1193 | ||
1194 | /* | |
1195 | * If checksum is needed at all, try to do it while copying the | |
1196 | * data. If the data is truncated, or if we only want a partial | |
1197 | * coverage checksum (UDP-Lite), do it before the copy. | |
1198 | */ | |
1199 | ||
81d54ec8 | 1200 | if (len < ulen || UDP_SKB_CB(skb)->partial_cov) { |
db8dac20 DM |
1201 | if (udp_lib_checksum_complete(skb)) |
1202 | goto csum_copy_err; | |
1203 | } | |
1204 | ||
1205 | if (skb_csum_unnecessary(skb)) | |
1206 | err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), | |
81d54ec8 | 1207 | msg->msg_iov, len); |
db8dac20 | 1208 | else { |
c482c568 ED |
1209 | err = skb_copy_and_csum_datagram_iovec(skb, |
1210 | sizeof(struct udphdr), | |
1211 | msg->msg_iov); | |
db8dac20 DM |
1212 | |
1213 | if (err == -EINVAL) | |
1214 | goto csum_copy_err; | |
1215 | } | |
1216 | ||
1217 | if (err) | |
1218 | goto out_free; | |
1219 | ||
1220 | if (!peeked) | |
629ca23c PE |
1221 | UDP_INC_STATS_USER(sock_net(sk), |
1222 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 | 1223 | |
3b885787 | 1224 | sock_recv_ts_and_drops(msg, sk, skb); |
db8dac20 DM |
1225 | |
1226 | /* Copy the address. */ | |
c482c568 | 1227 | if (sin) { |
db8dac20 DM |
1228 | sin->sin_family = AF_INET; |
1229 | sin->sin_port = udp_hdr(skb)->source; | |
1230 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
1231 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
1232 | } | |
1233 | if (inet->cmsg_flags) | |
1234 | ip_cmsg_recv(msg, skb); | |
1235 | ||
81d54ec8 | 1236 | err = len; |
db8dac20 DM |
1237 | if (flags & MSG_TRUNC) |
1238 | err = ulen; | |
1239 | ||
1240 | out_free: | |
9d410c79 | 1241 | skb_free_datagram_locked(sk, skb); |
db8dac20 DM |
1242 | out: |
1243 | return err; | |
1244 | ||
1245 | csum_copy_err: | |
8a74ad60 | 1246 | slow = lock_sock_fast(sk); |
db8dac20 | 1247 | if (!skb_kill_datagram(sk, skb, flags)) |
629ca23c | 1248 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8a74ad60 | 1249 | unlock_sock_fast(sk, slow); |
db8dac20 DM |
1250 | |
1251 | if (noblock) | |
1252 | return -EAGAIN; | |
1253 | goto try_again; | |
1254 | } | |
1255 | ||
1256 | ||
1da177e4 LT |
1257 | int udp_disconnect(struct sock *sk, int flags) |
1258 | { | |
1259 | struct inet_sock *inet = inet_sk(sk); | |
1260 | /* | |
1261 | * 1003.1g - break association. | |
1262 | */ | |
e905a9ed | 1263 | |
1da177e4 | 1264 | sk->sk_state = TCP_CLOSE; |
c720c7e8 ED |
1265 | inet->inet_daddr = 0; |
1266 | inet->inet_dport = 0; | |
c58dc01b | 1267 | sock_rps_save_rxhash(sk, 0); |
1da177e4 LT |
1268 | sk->sk_bound_dev_if = 0; |
1269 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
1270 | inet_reset_saddr(sk); | |
1271 | ||
1272 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
1273 | sk->sk_prot->unhash(sk); | |
c720c7e8 | 1274 | inet->inet_sport = 0; |
1da177e4 LT |
1275 | } |
1276 | sk_dst_reset(sk); | |
1277 | return 0; | |
1278 | } | |
c482c568 | 1279 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1280 | |
645ca708 ED |
1281 | void udp_lib_unhash(struct sock *sk) |
1282 | { | |
723b4610 ED |
1283 | if (sk_hashed(sk)) { |
1284 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
512615b6 ED |
1285 | struct udp_hslot *hslot, *hslot2; |
1286 | ||
1287 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1288 | udp_sk(sk)->udp_port_hash); | |
1289 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
645ca708 | 1290 | |
723b4610 ED |
1291 | spin_lock_bh(&hslot->lock); |
1292 | if (sk_nulls_del_node_init_rcu(sk)) { | |
fdcc8aa9 | 1293 | hslot->count--; |
c720c7e8 | 1294 | inet_sk(sk)->inet_num = 0; |
723b4610 | 1295 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
512615b6 ED |
1296 | |
1297 | spin_lock(&hslot2->lock); | |
1298 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1299 | hslot2->count--; | |
1300 | spin_unlock(&hslot2->lock); | |
723b4610 ED |
1301 | } |
1302 | spin_unlock_bh(&hslot->lock); | |
645ca708 | 1303 | } |
645ca708 ED |
1304 | } |
1305 | EXPORT_SYMBOL(udp_lib_unhash); | |
1306 | ||
719f8358 ED |
1307 | /* |
1308 | * inet_rcv_saddr was changed, we must rehash secondary hash | |
1309 | */ | |
1310 | void udp_lib_rehash(struct sock *sk, u16 newhash) | |
1311 | { | |
1312 | if (sk_hashed(sk)) { | |
1313 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
1314 | struct udp_hslot *hslot, *hslot2, *nhslot2; | |
1315 | ||
1316 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
1317 | nhslot2 = udp_hashslot2(udptable, newhash); | |
1318 | udp_sk(sk)->udp_portaddr_hash = newhash; | |
1319 | if (hslot2 != nhslot2) { | |
1320 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1321 | udp_sk(sk)->udp_port_hash); | |
1322 | /* we must lock primary chain too */ | |
1323 | spin_lock_bh(&hslot->lock); | |
1324 | ||
1325 | spin_lock(&hslot2->lock); | |
1326 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1327 | hslot2->count--; | |
1328 | spin_unlock(&hslot2->lock); | |
1329 | ||
1330 | spin_lock(&nhslot2->lock); | |
1331 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
1332 | &nhslot2->head); | |
1333 | nhslot2->count++; | |
1334 | spin_unlock(&nhslot2->lock); | |
1335 | ||
1336 | spin_unlock_bh(&hslot->lock); | |
1337 | } | |
1338 | } | |
1339 | } | |
1340 | EXPORT_SYMBOL(udp_lib_rehash); | |
1341 | ||
1342 | static void udp_v4_rehash(struct sock *sk) | |
1343 | { | |
1344 | u16 new_hash = udp4_portaddr_hash(sock_net(sk), | |
1345 | inet_sk(sk)->inet_rcv_saddr, | |
1346 | inet_sk(sk)->inet_num); | |
1347 | udp_lib_rehash(sk, new_hash); | |
1348 | } | |
1349 | ||
93821778 HX |
1350 | static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
1351 | { | |
fec5e652 | 1352 | int rc; |
766e9037 | 1353 | |
fec5e652 | 1354 | if (inet_sk(sk)->inet_daddr) |
c58dc01b | 1355 | sock_rps_save_rxhash(sk, skb->rxhash); |
fec5e652 | 1356 | |
f84af32c | 1357 | rc = ip_queue_rcv_skb(sk, skb); |
766e9037 ED |
1358 | if (rc < 0) { |
1359 | int is_udplite = IS_UDPLITE(sk); | |
93821778 | 1360 | |
93821778 | 1361 | /* Note that an ENOMEM error is charged twice */ |
766e9037 | 1362 | if (rc == -ENOMEM) |
93821778 HX |
1363 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1364 | is_udplite); | |
766e9037 ED |
1365 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
1366 | kfree_skb(skb); | |
1367 | return -1; | |
93821778 HX |
1368 | } |
1369 | ||
1370 | return 0; | |
1371 | ||
93821778 HX |
1372 | } |
1373 | ||
db8dac20 DM |
1374 | /* returns: |
1375 | * -1: error | |
1376 | * 0: success | |
1377 | * >0: "udp encap" protocol resubmission | |
1378 | * | |
1379 | * Note that in the success and error cases, the skb is assumed to | |
1380 | * have either been requeued or freed. | |
1381 | */ | |
c482c568 | 1382 | int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
db8dac20 DM |
1383 | { |
1384 | struct udp_sock *up = udp_sk(sk); | |
1385 | int rc; | |
1386 | int is_udplite = IS_UDPLITE(sk); | |
1387 | ||
1388 | /* | |
1389 | * Charge it to the socket, dropping if the queue is full. | |
1390 | */ | |
1391 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1392 | goto drop; | |
1393 | nf_reset(skb); | |
1394 | ||
1395 | if (up->encap_type) { | |
1396 | /* | |
1397 | * This is an encapsulation socket so pass the skb to | |
1398 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1399 | * fall through and pass this up the UDP socket. | |
1400 | * up->encap_rcv() returns the following value: | |
1401 | * =0 if skb was successfully passed to the encap | |
1402 | * handler or was discarded by it. | |
1403 | * >0 if skb should be passed on to UDP. | |
1404 | * <0 if skb should be resubmitted as proto -N | |
1405 | */ | |
1406 | ||
1407 | /* if we're overly short, let UDP handle it */ | |
1408 | if (skb->len > sizeof(struct udphdr) && | |
1409 | up->encap_rcv != NULL) { | |
1410 | int ret; | |
1411 | ||
1412 | ret = (*up->encap_rcv)(sk, skb); | |
1413 | if (ret <= 0) { | |
0283328e PE |
1414 | UDP_INC_STATS_BH(sock_net(sk), |
1415 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
1416 | is_udplite); |
1417 | return -ret; | |
1418 | } | |
1419 | } | |
1420 | ||
1421 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1422 | } | |
1423 | ||
1424 | /* | |
1425 | * UDP-Lite specific tests, ignored on UDP sockets | |
1426 | */ | |
1427 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1428 | ||
1429 | /* | |
1430 | * MIB statistics other than incrementing the error count are | |
1431 | * disabled for the following two types of errors: these depend | |
1432 | * on the application settings, not on the functioning of the | |
1433 | * protocol stack as such. | |
1434 | * | |
1435 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1436 | * way ... to ... at least let the receiving application block | |
1437 | * delivery of packets with coverage values less than a value | |
1438 | * provided by the application." | |
1439 | */ | |
1440 | if (up->pcrlen == 0) { /* full coverage was set */ | |
1441 | LIMIT_NETDEBUG(KERN_WARNING "UDPLITE: partial coverage " | |
1442 | "%d while full coverage %d requested\n", | |
1443 | UDP_SKB_CB(skb)->cscov, skb->len); | |
1444 | goto drop; | |
1445 | } | |
1446 | /* The next case involves violating the min. coverage requested | |
1447 | * by the receiver. This is subtle: if receiver wants x and x is | |
1448 | * greater than the buffersize/MTU then receiver will complain | |
1449 | * that it wants x while sender emits packets of smaller size y. | |
1450 | * Therefore the above ...()->partial_cov statement is essential. | |
1451 | */ | |
1452 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
1453 | LIMIT_NETDEBUG(KERN_WARNING | |
1454 | "UDPLITE: coverage %d too small, need min %d\n", | |
1455 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
1456 | goto drop; | |
1457 | } | |
1458 | } | |
1459 | ||
0d7da9dd | 1460 | if (rcu_dereference_raw(sk->sk_filter)) { |
db8dac20 DM |
1461 | if (udp_lib_checksum_complete(skb)) |
1462 | goto drop; | |
1463 | } | |
1464 | ||
c377411f ED |
1465 | |
1466 | if (sk_rcvqueues_full(sk, skb)) | |
1467 | goto drop; | |
1468 | ||
93821778 | 1469 | rc = 0; |
db8dac20 | 1470 | |
93821778 HX |
1471 | bh_lock_sock(sk); |
1472 | if (!sock_owned_by_user(sk)) | |
1473 | rc = __udp_queue_rcv_skb(sk, skb); | |
a3a858ff | 1474 | else if (sk_add_backlog(sk, skb)) { |
55349790 ZY |
1475 | bh_unlock_sock(sk); |
1476 | goto drop; | |
1477 | } | |
93821778 HX |
1478 | bh_unlock_sock(sk); |
1479 | ||
1480 | return rc; | |
db8dac20 DM |
1481 | |
1482 | drop: | |
0283328e | 1483 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8edf19c2 | 1484 | atomic_inc(&sk->sk_drops); |
db8dac20 DM |
1485 | kfree_skb(skb); |
1486 | return -1; | |
1487 | } | |
1488 | ||
1240d137 ED |
1489 | |
1490 | static void flush_stack(struct sock **stack, unsigned int count, | |
1491 | struct sk_buff *skb, unsigned int final) | |
1492 | { | |
1493 | unsigned int i; | |
1494 | struct sk_buff *skb1 = NULL; | |
f6b8f32c | 1495 | struct sock *sk; |
1240d137 ED |
1496 | |
1497 | for (i = 0; i < count; i++) { | |
f6b8f32c | 1498 | sk = stack[i]; |
1240d137 ED |
1499 | if (likely(skb1 == NULL)) |
1500 | skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC); | |
1501 | ||
f6b8f32c ED |
1502 | if (!skb1) { |
1503 | atomic_inc(&sk->sk_drops); | |
1504 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, | |
1505 | IS_UDPLITE(sk)); | |
1506 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, | |
1507 | IS_UDPLITE(sk)); | |
1508 | } | |
1509 | ||
1510 | if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0) | |
1240d137 ED |
1511 | skb1 = NULL; |
1512 | } | |
1513 | if (unlikely(skb1)) | |
1514 | kfree_skb(skb1); | |
1515 | } | |
1516 | ||
db8dac20 DM |
1517 | /* |
1518 | * Multicasts and broadcasts go to each listener. | |
1519 | * | |
1240d137 | 1520 | * Note: called only from the BH handler context. |
db8dac20 | 1521 | */ |
e3163493 | 1522 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1523 | struct udphdr *uh, |
1524 | __be32 saddr, __be32 daddr, | |
645ca708 | 1525 | struct udp_table *udptable) |
db8dac20 | 1526 | { |
1240d137 | 1527 | struct sock *sk, *stack[256 / sizeof(struct sock *)]; |
f86dcc5a | 1528 | struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest)); |
db8dac20 | 1529 | int dif; |
1240d137 | 1530 | unsigned int i, count = 0; |
db8dac20 | 1531 | |
645ca708 | 1532 | spin_lock(&hslot->lock); |
88ab1932 | 1533 | sk = sk_nulls_head(&hslot->head); |
db8dac20 | 1534 | dif = skb->dev->ifindex; |
920a4611 | 1535 | sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif); |
1240d137 ED |
1536 | while (sk) { |
1537 | stack[count++] = sk; | |
1538 | sk = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest, | |
1539 | daddr, uh->source, saddr, dif); | |
1540 | if (unlikely(count == ARRAY_SIZE(stack))) { | |
1541 | if (!sk) | |
1542 | break; | |
1543 | flush_stack(stack, count, skb, ~0); | |
1544 | count = 0; | |
1545 | } | |
1546 | } | |
1547 | /* | |
1548 | * before releasing chain lock, we must take a reference on sockets | |
1549 | */ | |
1550 | for (i = 0; i < count; i++) | |
1551 | sock_hold(stack[i]); | |
1552 | ||
645ca708 | 1553 | spin_unlock(&hslot->lock); |
1240d137 ED |
1554 | |
1555 | /* | |
1556 | * do the slow work with no lock held | |
1557 | */ | |
1558 | if (count) { | |
1559 | flush_stack(stack, count, skb, count - 1); | |
1560 | ||
1561 | for (i = 0; i < count; i++) | |
1562 | sock_put(stack[i]); | |
1563 | } else { | |
1564 | kfree_skb(skb); | |
1565 | } | |
db8dac20 DM |
1566 | return 0; |
1567 | } | |
1568 | ||
1569 | /* Initialize UDP checksum. If exited with zero value (success), | |
1570 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1571 | * Otherwise, csum completion requires chacksumming packet body, | |
1572 | * including udp header and folding it to skb->csum. | |
1573 | */ | |
1574 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1575 | int proto) | |
1576 | { | |
1577 | const struct iphdr *iph; | |
1578 | int err; | |
1579 | ||
1580 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1581 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1582 | ||
1583 | if (proto == IPPROTO_UDPLITE) { | |
1584 | err = udplite_checksum_init(skb, uh); | |
1585 | if (err) | |
1586 | return err; | |
1587 | } | |
1588 | ||
1589 | iph = ip_hdr(skb); | |
1590 | if (uh->check == 0) { | |
1591 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1592 | } else if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
c482c568 | 1593 | if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, |
db8dac20 DM |
1594 | proto, skb->csum)) |
1595 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1596 | } | |
1597 | if (!skb_csum_unnecessary(skb)) | |
1598 | skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, | |
1599 | skb->len, proto, 0); | |
1600 | /* Probably, we should checksum udp header (it should be in cache | |
1601 | * in any case) and data in tiny packets (< rx copybreak). | |
1602 | */ | |
1603 | ||
1604 | return 0; | |
1605 | } | |
1606 | ||
1607 | /* | |
1608 | * All we need to do is get the socket, and then do a checksum. | |
1609 | */ | |
1610 | ||
645ca708 | 1611 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1612 | int proto) |
1613 | { | |
1614 | struct sock *sk; | |
7b5e56f9 | 1615 | struct udphdr *uh; |
db8dac20 | 1616 | unsigned short ulen; |
adf30907 | 1617 | struct rtable *rt = skb_rtable(skb); |
2783ef23 | 1618 | __be32 saddr, daddr; |
0283328e | 1619 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1620 | |
1621 | /* | |
1622 | * Validate the packet. | |
1623 | */ | |
1624 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1625 | goto drop; /* No space for header. */ | |
1626 | ||
7b5e56f9 | 1627 | uh = udp_hdr(skb); |
db8dac20 | 1628 | ulen = ntohs(uh->len); |
ccc2d97c BM |
1629 | saddr = ip_hdr(skb)->saddr; |
1630 | daddr = ip_hdr(skb)->daddr; | |
1631 | ||
db8dac20 DM |
1632 | if (ulen > skb->len) |
1633 | goto short_packet; | |
1634 | ||
1635 | if (proto == IPPROTO_UDP) { | |
1636 | /* UDP validates ulen. */ | |
1637 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1638 | goto short_packet; | |
1639 | uh = udp_hdr(skb); | |
1640 | } | |
1641 | ||
1642 | if (udp4_csum_init(skb, uh, proto)) | |
1643 | goto csum_error; | |
1644 | ||
1645 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) | |
e3163493 PE |
1646 | return __udp4_lib_mcast_deliver(net, skb, uh, |
1647 | saddr, daddr, udptable); | |
db8dac20 | 1648 | |
607c4aaf | 1649 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); |
db8dac20 DM |
1650 | |
1651 | if (sk != NULL) { | |
93821778 | 1652 | int ret = udp_queue_rcv_skb(sk, skb); |
db8dac20 DM |
1653 | sock_put(sk); |
1654 | ||
1655 | /* a return value > 0 means to resubmit the input, but | |
1656 | * it wants the return to be -protocol, or 0 | |
1657 | */ | |
1658 | if (ret > 0) | |
1659 | return -ret; | |
1660 | return 0; | |
1661 | } | |
1662 | ||
1663 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1664 | goto drop; | |
1665 | nf_reset(skb); | |
1666 | ||
1667 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1668 | if (udp_lib_checksum_complete(skb)) | |
1669 | goto csum_error; | |
1670 | ||
0283328e | 1671 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1672 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1673 | ||
1674 | /* | |
1675 | * Hmm. We got an UDP packet to a port to which we | |
1676 | * don't wanna listen. Ignore it. | |
1677 | */ | |
1678 | kfree_skb(skb); | |
1679 | return 0; | |
1680 | ||
1681 | short_packet: | |
673d57e7 | 1682 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n", |
db8dac20 | 1683 | proto == IPPROTO_UDPLITE ? "-Lite" : "", |
673d57e7 | 1684 | &saddr, |
db8dac20 DM |
1685 | ntohs(uh->source), |
1686 | ulen, | |
1687 | skb->len, | |
673d57e7 | 1688 | &daddr, |
db8dac20 DM |
1689 | ntohs(uh->dest)); |
1690 | goto drop; | |
1691 | ||
1692 | csum_error: | |
1693 | /* | |
1694 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1695 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1696 | */ | |
673d57e7 | 1697 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n", |
db8dac20 | 1698 | proto == IPPROTO_UDPLITE ? "-Lite" : "", |
673d57e7 | 1699 | &saddr, |
db8dac20 | 1700 | ntohs(uh->source), |
673d57e7 | 1701 | &daddr, |
db8dac20 DM |
1702 | ntohs(uh->dest), |
1703 | ulen); | |
1704 | drop: | |
0283328e | 1705 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1706 | kfree_skb(skb); |
1707 | return 0; | |
1708 | } | |
1709 | ||
1710 | int udp_rcv(struct sk_buff *skb) | |
1711 | { | |
645ca708 | 1712 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
1713 | } |
1714 | ||
7d06b2e0 | 1715 | void udp_destroy_sock(struct sock *sk) |
db8dac20 | 1716 | { |
8a74ad60 | 1717 | bool slow = lock_sock_fast(sk); |
db8dac20 | 1718 | udp_flush_pending_frames(sk); |
8a74ad60 | 1719 | unlock_sock_fast(sk, slow); |
db8dac20 DM |
1720 | } |
1721 | ||
1da177e4 LT |
1722 | /* |
1723 | * Socket option code for UDP | |
1724 | */ | |
4c0a6cb0 | 1725 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 1726 | char __user *optval, unsigned int optlen, |
4c0a6cb0 | 1727 | int (*push_pending_frames)(struct sock *)) |
1da177e4 LT |
1728 | { |
1729 | struct udp_sock *up = udp_sk(sk); | |
1730 | int val; | |
1731 | int err = 0; | |
b2bf1e26 | 1732 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 1733 | |
c482c568 | 1734 | if (optlen < sizeof(int)) |
1da177e4 LT |
1735 | return -EINVAL; |
1736 | ||
1737 | if (get_user(val, (int __user *)optval)) | |
1738 | return -EFAULT; | |
1739 | ||
6516c655 | 1740 | switch (optname) { |
1da177e4 LT |
1741 | case UDP_CORK: |
1742 | if (val != 0) { | |
1743 | up->corkflag = 1; | |
1744 | } else { | |
1745 | up->corkflag = 0; | |
1746 | lock_sock(sk); | |
4c0a6cb0 | 1747 | (*push_pending_frames)(sk); |
1da177e4 LT |
1748 | release_sock(sk); |
1749 | } | |
1750 | break; | |
e905a9ed | 1751 | |
1da177e4 LT |
1752 | case UDP_ENCAP: |
1753 | switch (val) { | |
1754 | case 0: | |
1755 | case UDP_ENCAP_ESPINUDP: | |
1756 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
1757 | up->encap_rcv = xfrm4_udp_encap_rcv; |
1758 | /* FALLTHROUGH */ | |
342f0234 | 1759 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 LT |
1760 | up->encap_type = val; |
1761 | break; | |
1762 | default: | |
1763 | err = -ENOPROTOOPT; | |
1764 | break; | |
1765 | } | |
1766 | break; | |
1767 | ||
ba4e58ec GR |
1768 | /* |
1769 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
1770 | */ | |
1771 | /* The sender sets actual checksum coverage length via this option. | |
1772 | * The case coverage > packet length is handled by send module. */ | |
1773 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 1774 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1775 | return -ENOPROTOOPT; |
1776 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
1777 | val = 8; | |
4be929be AD |
1778 | else if (val > USHRT_MAX) |
1779 | val = USHRT_MAX; | |
ba4e58ec GR |
1780 | up->pcslen = val; |
1781 | up->pcflag |= UDPLITE_SEND_CC; | |
1782 | break; | |
1783 | ||
e905a9ed YH |
1784 | /* The receiver specifies a minimum checksum coverage value. To make |
1785 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
1786 | * used, this again means full checksum coverage. */ |
1787 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 1788 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1789 | return -ENOPROTOOPT; |
1790 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
1791 | val = 8; | |
4be929be AD |
1792 | else if (val > USHRT_MAX) |
1793 | val = USHRT_MAX; | |
ba4e58ec GR |
1794 | up->pcrlen = val; |
1795 | up->pcflag |= UDPLITE_RECV_CC; | |
1796 | break; | |
1797 | ||
1da177e4 LT |
1798 | default: |
1799 | err = -ENOPROTOOPT; | |
1800 | break; | |
6516c655 | 1801 | } |
1da177e4 LT |
1802 | |
1803 | return err; | |
1804 | } | |
c482c568 | 1805 | EXPORT_SYMBOL(udp_lib_setsockopt); |
1da177e4 | 1806 | |
db8dac20 | 1807 | int udp_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 1808 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
1809 | { |
1810 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1811 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1812 | udp_push_pending_frames); | |
1813 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
1814 | } | |
1815 | ||
1816 | #ifdef CONFIG_COMPAT | |
1817 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 1818 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
1819 | { |
1820 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1821 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1822 | udp_push_pending_frames); | |
1823 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
1824 | } | |
1825 | #endif | |
1826 | ||
4c0a6cb0 GR |
1827 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
1828 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
1829 | { |
1830 | struct udp_sock *up = udp_sk(sk); | |
1831 | int val, len; | |
1832 | ||
c482c568 | 1833 | if (get_user(len, optlen)) |
1da177e4 LT |
1834 | return -EFAULT; |
1835 | ||
1836 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 1837 | |
6516c655 | 1838 | if (len < 0) |
1da177e4 LT |
1839 | return -EINVAL; |
1840 | ||
6516c655 | 1841 | switch (optname) { |
1da177e4 LT |
1842 | case UDP_CORK: |
1843 | val = up->corkflag; | |
1844 | break; | |
1845 | ||
1846 | case UDP_ENCAP: | |
1847 | val = up->encap_type; | |
1848 | break; | |
1849 | ||
ba4e58ec GR |
1850 | /* The following two cannot be changed on UDP sockets, the return is |
1851 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
1852 | case UDPLITE_SEND_CSCOV: | |
1853 | val = up->pcslen; | |
1854 | break; | |
1855 | ||
1856 | case UDPLITE_RECV_CSCOV: | |
1857 | val = up->pcrlen; | |
1858 | break; | |
1859 | ||
1da177e4 LT |
1860 | default: |
1861 | return -ENOPROTOOPT; | |
6516c655 | 1862 | } |
1da177e4 | 1863 | |
6516c655 | 1864 | if (put_user(len, optlen)) |
e905a9ed | 1865 | return -EFAULT; |
c482c568 | 1866 | if (copy_to_user(optval, &val, len)) |
1da177e4 | 1867 | return -EFAULT; |
e905a9ed | 1868 | return 0; |
1da177e4 | 1869 | } |
c482c568 | 1870 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1da177e4 | 1871 | |
db8dac20 DM |
1872 | int udp_getsockopt(struct sock *sk, int level, int optname, |
1873 | char __user *optval, int __user *optlen) | |
1874 | { | |
1875 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1876 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1877 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
1878 | } | |
1879 | ||
1880 | #ifdef CONFIG_COMPAT | |
1881 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
1882 | char __user *optval, int __user *optlen) | |
1883 | { | |
1884 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1885 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1886 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
1887 | } | |
1888 | #endif | |
1da177e4 LT |
1889 | /** |
1890 | * udp_poll - wait for a UDP event. | |
1891 | * @file - file struct | |
1892 | * @sock - socket | |
1893 | * @wait - poll table | |
1894 | * | |
e905a9ed | 1895 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
1896 | * blocking sockets. If application is using a blocking fd |
1897 | * and a packet with checksum error is in the queue; | |
1898 | * then it could get return from select indicating data available | |
1899 | * but then block when reading it. Add special case code | |
1900 | * to work around these arguably broken applications. | |
1901 | */ | |
1902 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
1903 | { | |
1904 | unsigned int mask = datagram_poll(file, sock, wait); | |
1905 | struct sock *sk = sock->sk; | |
ba4e58ec | 1906 | |
1da177e4 | 1907 | /* Check for false positives due to checksum errors */ |
85584672 ED |
1908 | if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) && |
1909 | !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk)) | |
1910 | mask &= ~(POLLIN | POLLRDNORM); | |
1da177e4 LT |
1911 | |
1912 | return mask; | |
e905a9ed | 1913 | |
1da177e4 | 1914 | } |
c482c568 | 1915 | EXPORT_SYMBOL(udp_poll); |
1da177e4 | 1916 | |
db8dac20 DM |
1917 | struct proto udp_prot = { |
1918 | .name = "UDP", | |
1919 | .owner = THIS_MODULE, | |
1920 | .close = udp_lib_close, | |
1921 | .connect = ip4_datagram_connect, | |
1922 | .disconnect = udp_disconnect, | |
1923 | .ioctl = udp_ioctl, | |
1924 | .destroy = udp_destroy_sock, | |
1925 | .setsockopt = udp_setsockopt, | |
1926 | .getsockopt = udp_getsockopt, | |
1927 | .sendmsg = udp_sendmsg, | |
1928 | .recvmsg = udp_recvmsg, | |
1929 | .sendpage = udp_sendpage, | |
93821778 | 1930 | .backlog_rcv = __udp_queue_rcv_skb, |
db8dac20 DM |
1931 | .hash = udp_lib_hash, |
1932 | .unhash = udp_lib_unhash, | |
719f8358 | 1933 | .rehash = udp_v4_rehash, |
db8dac20 DM |
1934 | .get_port = udp_v4_get_port, |
1935 | .memory_allocated = &udp_memory_allocated, | |
1936 | .sysctl_mem = sysctl_udp_mem, | |
1937 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
1938 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
1939 | .obj_size = sizeof(struct udp_sock), | |
271b72c7 | 1940 | .slab_flags = SLAB_DESTROY_BY_RCU, |
645ca708 | 1941 | .h.udp_table = &udp_table, |
db8dac20 DM |
1942 | #ifdef CONFIG_COMPAT |
1943 | .compat_setsockopt = compat_udp_setsockopt, | |
1944 | .compat_getsockopt = compat_udp_getsockopt, | |
1945 | #endif | |
fcbdf09d | 1946 | .clear_sk = sk_prot_clear_portaddr_nulls, |
db8dac20 | 1947 | }; |
c482c568 | 1948 | EXPORT_SYMBOL(udp_prot); |
1da177e4 LT |
1949 | |
1950 | /* ------------------------------------------------------------------------ */ | |
1951 | #ifdef CONFIG_PROC_FS | |
1952 | ||
645ca708 | 1953 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
1954 | { |
1955 | struct sock *sk; | |
1956 | struct udp_iter_state *state = seq->private; | |
6f191efe | 1957 | struct net *net = seq_file_net(seq); |
1da177e4 | 1958 | |
f86dcc5a ED |
1959 | for (state->bucket = start; state->bucket <= state->udp_table->mask; |
1960 | ++state->bucket) { | |
88ab1932 | 1961 | struct hlist_nulls_node *node; |
645ca708 | 1962 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
f86dcc5a ED |
1963 | |
1964 | if (hlist_nulls_empty(&hslot->head)) | |
1965 | continue; | |
1966 | ||
645ca708 | 1967 | spin_lock_bh(&hslot->lock); |
88ab1932 | 1968 | sk_nulls_for_each(sk, node, &hslot->head) { |
878628fb | 1969 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 1970 | continue; |
1da177e4 LT |
1971 | if (sk->sk_family == state->family) |
1972 | goto found; | |
1973 | } | |
645ca708 | 1974 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
1975 | } |
1976 | sk = NULL; | |
1977 | found: | |
1978 | return sk; | |
1979 | } | |
1980 | ||
1981 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
1982 | { | |
1983 | struct udp_iter_state *state = seq->private; | |
6f191efe | 1984 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1985 | |
1986 | do { | |
88ab1932 | 1987 | sk = sk_nulls_next(sk); |
878628fb | 1988 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 1989 | |
645ca708 | 1990 | if (!sk) { |
f86dcc5a | 1991 | if (state->bucket <= state->udp_table->mask) |
30842f29 | 1992 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
645ca708 | 1993 | return udp_get_first(seq, state->bucket + 1); |
1da177e4 LT |
1994 | } |
1995 | return sk; | |
1996 | } | |
1997 | ||
1998 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
1999 | { | |
645ca708 | 2000 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
2001 | |
2002 | if (sk) | |
6516c655 | 2003 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
2004 | --pos; |
2005 | return pos ? NULL : sk; | |
2006 | } | |
2007 | ||
2008 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
2009 | { | |
30842f29 | 2010 | struct udp_iter_state *state = seq->private; |
f86dcc5a | 2011 | state->bucket = MAX_UDP_PORTS; |
30842f29 | 2012 | |
b50660f1 | 2013 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
2014 | } |
2015 | ||
2016 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2017 | { | |
2018 | struct sock *sk; | |
2019 | ||
b50660f1 | 2020 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2021 | sk = udp_get_idx(seq, 0); |
2022 | else | |
2023 | sk = udp_get_next(seq, v); | |
2024 | ||
2025 | ++*pos; | |
2026 | return sk; | |
2027 | } | |
2028 | ||
2029 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
2030 | { | |
645ca708 ED |
2031 | struct udp_iter_state *state = seq->private; |
2032 | ||
f86dcc5a | 2033 | if (state->bucket <= state->udp_table->mask) |
645ca708 | 2034 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
1da177e4 LT |
2035 | } |
2036 | ||
2037 | static int udp_seq_open(struct inode *inode, struct file *file) | |
2038 | { | |
2039 | struct udp_seq_afinfo *afinfo = PDE(inode)->data; | |
a2be75c1 DL |
2040 | struct udp_iter_state *s; |
2041 | int err; | |
a91275ef | 2042 | |
a2be75c1 DL |
2043 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2044 | sizeof(struct udp_iter_state)); | |
2045 | if (err < 0) | |
2046 | return err; | |
a91275ef | 2047 | |
a2be75c1 | 2048 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2049 | s->family = afinfo->family; |
645ca708 | 2050 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 2051 | return err; |
a91275ef DL |
2052 | } |
2053 | ||
1da177e4 | 2054 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 2055 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
2056 | { |
2057 | struct proc_dir_entry *p; | |
2058 | int rc = 0; | |
2059 | ||
3ba9441b DL |
2060 | afinfo->seq_fops.open = udp_seq_open; |
2061 | afinfo->seq_fops.read = seq_read; | |
2062 | afinfo->seq_fops.llseek = seq_lseek; | |
2063 | afinfo->seq_fops.release = seq_release_net; | |
1da177e4 | 2064 | |
dda61925 DL |
2065 | afinfo->seq_ops.start = udp_seq_start; |
2066 | afinfo->seq_ops.next = udp_seq_next; | |
2067 | afinfo->seq_ops.stop = udp_seq_stop; | |
2068 | ||
84841c3c DL |
2069 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
2070 | &afinfo->seq_fops, afinfo); | |
2071 | if (!p) | |
1da177e4 LT |
2072 | rc = -ENOMEM; |
2073 | return rc; | |
2074 | } | |
c482c568 | 2075 | EXPORT_SYMBOL(udp_proc_register); |
1da177e4 | 2076 | |
0c96d8c5 | 2077 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 2078 | { |
0c96d8c5 | 2079 | proc_net_remove(net, afinfo->name); |
1da177e4 | 2080 | } |
c482c568 | 2081 | EXPORT_SYMBOL(udp_proc_unregister); |
db8dac20 DM |
2082 | |
2083 | /* ------------------------------------------------------------------------ */ | |
5e659e4c PE |
2084 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
2085 | int bucket, int *len) | |
db8dac20 DM |
2086 | { |
2087 | struct inet_sock *inet = inet_sk(sp); | |
c720c7e8 ED |
2088 | __be32 dest = inet->inet_daddr; |
2089 | __be32 src = inet->inet_rcv_saddr; | |
2090 | __u16 destp = ntohs(inet->inet_dport); | |
2091 | __u16 srcp = ntohs(inet->inet_sport); | |
db8dac20 | 2092 | |
f86dcc5a | 2093 | seq_printf(f, "%5d: %08X:%04X %08X:%04X" |
cb61cb9b | 2094 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d%n", |
db8dac20 | 2095 | bucket, src, srcp, dest, destp, sp->sk_state, |
31e6d363 ED |
2096 | sk_wmem_alloc_get(sp), |
2097 | sk_rmem_alloc_get(sp), | |
db8dac20 | 2098 | 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp), |
cb61cb9b ED |
2099 | atomic_read(&sp->sk_refcnt), sp, |
2100 | atomic_read(&sp->sk_drops), len); | |
db8dac20 DM |
2101 | } |
2102 | ||
2103 | int udp4_seq_show(struct seq_file *seq, void *v) | |
2104 | { | |
2105 | if (v == SEQ_START_TOKEN) | |
2106 | seq_printf(seq, "%-127s\n", | |
2107 | " sl local_address rem_address st tx_queue " | |
2108 | "rx_queue tr tm->when retrnsmt uid timeout " | |
cb61cb9b | 2109 | "inode ref pointer drops"); |
db8dac20 | 2110 | else { |
db8dac20 | 2111 | struct udp_iter_state *state = seq->private; |
5e659e4c | 2112 | int len; |
db8dac20 | 2113 | |
5e659e4c | 2114 | udp4_format_sock(v, seq, state->bucket, &len); |
c482c568 | 2115 | seq_printf(seq, "%*s\n", 127 - len, ""); |
db8dac20 DM |
2116 | } |
2117 | return 0; | |
2118 | } | |
2119 | ||
2120 | /* ------------------------------------------------------------------------ */ | |
db8dac20 | 2121 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
2122 | .name = "udp", |
2123 | .family = AF_INET, | |
645ca708 | 2124 | .udp_table = &udp_table, |
4ad96d39 DL |
2125 | .seq_fops = { |
2126 | .owner = THIS_MODULE, | |
2127 | }, | |
dda61925 DL |
2128 | .seq_ops = { |
2129 | .show = udp4_seq_show, | |
2130 | }, | |
db8dac20 DM |
2131 | }; |
2132 | ||
2c8c1e72 | 2133 | static int __net_init udp4_proc_init_net(struct net *net) |
15439feb PE |
2134 | { |
2135 | return udp_proc_register(net, &udp4_seq_afinfo); | |
2136 | } | |
2137 | ||
2c8c1e72 | 2138 | static void __net_exit udp4_proc_exit_net(struct net *net) |
15439feb PE |
2139 | { |
2140 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
2141 | } | |
2142 | ||
2143 | static struct pernet_operations udp4_net_ops = { | |
2144 | .init = udp4_proc_init_net, | |
2145 | .exit = udp4_proc_exit_net, | |
2146 | }; | |
2147 | ||
db8dac20 DM |
2148 | int __init udp4_proc_init(void) |
2149 | { | |
15439feb | 2150 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
2151 | } |
2152 | ||
2153 | void udp4_proc_exit(void) | |
2154 | { | |
15439feb | 2155 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 2156 | } |
1da177e4 LT |
2157 | #endif /* CONFIG_PROC_FS */ |
2158 | ||
f86dcc5a ED |
2159 | static __initdata unsigned long uhash_entries; |
2160 | static int __init set_uhash_entries(char *str) | |
645ca708 | 2161 | { |
f86dcc5a ED |
2162 | if (!str) |
2163 | return 0; | |
2164 | uhash_entries = simple_strtoul(str, &str, 0); | |
2165 | if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN) | |
2166 | uhash_entries = UDP_HTABLE_SIZE_MIN; | |
2167 | return 1; | |
2168 | } | |
2169 | __setup("uhash_entries=", set_uhash_entries); | |
645ca708 | 2170 | |
f86dcc5a ED |
2171 | void __init udp_table_init(struct udp_table *table, const char *name) |
2172 | { | |
2173 | unsigned int i; | |
2174 | ||
2175 | if (!CONFIG_BASE_SMALL) | |
2176 | table->hash = alloc_large_system_hash(name, | |
512615b6 | 2177 | 2 * sizeof(struct udp_hslot), |
f86dcc5a ED |
2178 | uhash_entries, |
2179 | 21, /* one slot per 2 MB */ | |
2180 | 0, | |
2181 | &table->log, | |
2182 | &table->mask, | |
2183 | 64 * 1024); | |
2184 | /* | |
2185 | * Make sure hash table has the minimum size | |
2186 | */ | |
2187 | if (CONFIG_BASE_SMALL || table->mask < UDP_HTABLE_SIZE_MIN - 1) { | |
2188 | table->hash = kmalloc(UDP_HTABLE_SIZE_MIN * | |
512615b6 | 2189 | 2 * sizeof(struct udp_hslot), GFP_KERNEL); |
f86dcc5a ED |
2190 | if (!table->hash) |
2191 | panic(name); | |
2192 | table->log = ilog2(UDP_HTABLE_SIZE_MIN); | |
2193 | table->mask = UDP_HTABLE_SIZE_MIN - 1; | |
2194 | } | |
512615b6 | 2195 | table->hash2 = table->hash + (table->mask + 1); |
f86dcc5a | 2196 | for (i = 0; i <= table->mask; i++) { |
88ab1932 | 2197 | INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i); |
fdcc8aa9 | 2198 | table->hash[i].count = 0; |
645ca708 ED |
2199 | spin_lock_init(&table->hash[i].lock); |
2200 | } | |
512615b6 ED |
2201 | for (i = 0; i <= table->mask; i++) { |
2202 | INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i); | |
2203 | table->hash2[i].count = 0; | |
2204 | spin_lock_init(&table->hash2[i].lock); | |
2205 | } | |
645ca708 ED |
2206 | } |
2207 | ||
95766fff HA |
2208 | void __init udp_init(void) |
2209 | { | |
8203efb3 | 2210 | unsigned long nr_pages, limit; |
95766fff | 2211 | |
f86dcc5a | 2212 | udp_table_init(&udp_table, "UDP"); |
95766fff HA |
2213 | /* Set the pressure threshold up by the same strategy of TCP. It is a |
2214 | * fraction of global memory that is up to 1/2 at 256 MB, decreasing | |
2215 | * toward zero with the amount of memory, with a floor of 128 pages. | |
2216 | */ | |
8203efb3 ED |
2217 | nr_pages = totalram_pages - totalhigh_pages; |
2218 | limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT); | |
2219 | limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11); | |
95766fff HA |
2220 | limit = max(limit, 128UL); |
2221 | sysctl_udp_mem[0] = limit / 4 * 3; | |
2222 | sysctl_udp_mem[1] = limit; | |
2223 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
2224 | ||
2225 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
2226 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
2227 | } | |
2228 | ||
d7ca4cc0 SS |
2229 | int udp4_ufo_send_check(struct sk_buff *skb) |
2230 | { | |
2231 | const struct iphdr *iph; | |
2232 | struct udphdr *uh; | |
2233 | ||
2234 | if (!pskb_may_pull(skb, sizeof(*uh))) | |
2235 | return -EINVAL; | |
2236 | ||
2237 | iph = ip_hdr(skb); | |
2238 | uh = udp_hdr(skb); | |
2239 | ||
2240 | uh->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, | |
2241 | IPPROTO_UDP, 0); | |
2242 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
2243 | skb->csum_offset = offsetof(struct udphdr, check); | |
2244 | skb->ip_summed = CHECKSUM_PARTIAL; | |
2245 | return 0; | |
2246 | } | |
2247 | ||
04ed3e74 | 2248 | struct sk_buff *udp4_ufo_fragment(struct sk_buff *skb, u32 features) |
d7ca4cc0 SS |
2249 | { |
2250 | struct sk_buff *segs = ERR_PTR(-EINVAL); | |
2251 | unsigned int mss; | |
2252 | int offset; | |
2253 | __wsum csum; | |
2254 | ||
2255 | mss = skb_shinfo(skb)->gso_size; | |
2256 | if (unlikely(skb->len <= mss)) | |
2257 | goto out; | |
2258 | ||
2259 | if (skb_gso_ok(skb, features | NETIF_F_GSO_ROBUST)) { | |
2260 | /* Packet is from an untrusted source, reset gso_segs. */ | |
2261 | int type = skb_shinfo(skb)->gso_type; | |
2262 | ||
2263 | if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY) || | |
2264 | !(type & (SKB_GSO_UDP)))) | |
2265 | goto out; | |
2266 | ||
2267 | skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss); | |
2268 | ||
2269 | segs = NULL; | |
2270 | goto out; | |
2271 | } | |
2272 | ||
2273 | /* Do software UFO. Complete and fill in the UDP checksum as HW cannot | |
2274 | * do checksum of UDP packets sent as multiple IP fragments. | |
2275 | */ | |
55508d60 | 2276 | offset = skb_checksum_start_offset(skb); |
c482c568 | 2277 | csum = skb_checksum(skb, offset, skb->len - offset, 0); |
d7ca4cc0 SS |
2278 | offset += skb->csum_offset; |
2279 | *(__sum16 *)(skb->data + offset) = csum_fold(csum); | |
2280 | skb->ip_summed = CHECKSUM_NONE; | |
2281 | ||
2282 | /* Fragment the skb. IP headers of the fragments are updated in | |
2283 | * inet_gso_segment() | |
2284 | */ | |
2285 | segs = skb_segment(skb, features); | |
2286 | out: | |
2287 | return segs; | |
2288 | } | |
2289 |