]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * The User Datagram Protocol (UDP). | |
7 | * | |
02c30a84 | 8 | * Authors: Ross Biro |
1da177e4 LT |
9 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
10 | * Arnt Gulbrandsen, <agulbra@nvg.unit.no> | |
113aa838 | 11 | * Alan Cox, <alan@lxorguk.ukuu.org.uk> |
1da177e4 LT |
12 | * Hirokazu Takahashi, <taka@valinux.co.jp> |
13 | * | |
14 | * Fixes: | |
15 | * Alan Cox : verify_area() calls | |
16 | * Alan Cox : stopped close while in use off icmp | |
17 | * messages. Not a fix but a botch that | |
18 | * for udp at least is 'valid'. | |
19 | * Alan Cox : Fixed icmp handling properly | |
20 | * Alan Cox : Correct error for oversized datagrams | |
e905a9ed YH |
21 | * Alan Cox : Tidied select() semantics. |
22 | * Alan Cox : udp_err() fixed properly, also now | |
1da177e4 LT |
23 | * select and read wake correctly on errors |
24 | * Alan Cox : udp_send verify_area moved to avoid mem leak | |
25 | * Alan Cox : UDP can count its memory | |
26 | * Alan Cox : send to an unknown connection causes | |
27 | * an ECONNREFUSED off the icmp, but | |
28 | * does NOT close. | |
29 | * Alan Cox : Switched to new sk_buff handlers. No more backlog! | |
30 | * Alan Cox : Using generic datagram code. Even smaller and the PEEK | |
31 | * bug no longer crashes it. | |
32 | * Fred Van Kempen : Net2e support for sk->broadcast. | |
33 | * Alan Cox : Uses skb_free_datagram | |
34 | * Alan Cox : Added get/set sockopt support. | |
35 | * Alan Cox : Broadcasting without option set returns EACCES. | |
36 | * Alan Cox : No wakeup calls. Instead we now use the callbacks. | |
37 | * Alan Cox : Use ip_tos and ip_ttl | |
38 | * Alan Cox : SNMP Mibs | |
39 | * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support. | |
40 | * Matt Dillon : UDP length checks. | |
41 | * Alan Cox : Smarter af_inet used properly. | |
42 | * Alan Cox : Use new kernel side addressing. | |
43 | * Alan Cox : Incorrect return on truncated datagram receive. | |
44 | * Arnt Gulbrandsen : New udp_send and stuff | |
45 | * Alan Cox : Cache last socket | |
46 | * Alan Cox : Route cache | |
47 | * Jon Peatfield : Minor efficiency fix to sendto(). | |
48 | * Mike Shaver : RFC1122 checks. | |
49 | * Alan Cox : Nonblocking error fix. | |
50 | * Willy Konynenberg : Transparent proxying support. | |
51 | * Mike McLagan : Routing by source | |
52 | * David S. Miller : New socket lookup architecture. | |
53 | * Last socket cache retained as it | |
54 | * does have a high hit rate. | |
55 | * Olaf Kirch : Don't linearise iovec on sendmsg. | |
56 | * Andi Kleen : Some cleanups, cache destination entry | |
e905a9ed | 57 | * for connect. |
1da177e4 LT |
58 | * Vitaly E. Lavrov : Transparent proxy revived after year coma. |
59 | * Melvin Smith : Check msg_name not msg_namelen in sendto(), | |
60 | * return ENOTCONN for unconnected sockets (POSIX) | |
61 | * Janos Farkas : don't deliver multi/broadcasts to a different | |
62 | * bound-to-device socket | |
63 | * Hirokazu Takahashi : HW checksumming for outgoing UDP | |
64 | * datagrams. | |
65 | * Hirokazu Takahashi : sendfile() on UDP works now. | |
66 | * Arnaldo C. Melo : convert /proc/net/udp to seq_file | |
67 | * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which | |
68 | * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind | |
69 | * a single port at the same time. | |
70 | * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support | |
342f0234 | 71 | * James Chapman : Add L2TP encapsulation type. |
1da177e4 LT |
72 | * |
73 | * | |
74 | * This program is free software; you can redistribute it and/or | |
75 | * modify it under the terms of the GNU General Public License | |
76 | * as published by the Free Software Foundation; either version | |
77 | * 2 of the License, or (at your option) any later version. | |
78 | */ | |
e905a9ed | 79 | |
afd46503 JP |
80 | #define pr_fmt(fmt) "UDP: " fmt |
81 | ||
1da177e4 LT |
82 | #include <asm/uaccess.h> |
83 | #include <asm/ioctls.h> | |
95766fff | 84 | #include <linux/bootmem.h> |
8203efb3 ED |
85 | #include <linux/highmem.h> |
86 | #include <linux/swap.h> | |
1da177e4 LT |
87 | #include <linux/types.h> |
88 | #include <linux/fcntl.h> | |
89 | #include <linux/module.h> | |
90 | #include <linux/socket.h> | |
91 | #include <linux/sockios.h> | |
14c85021 | 92 | #include <linux/igmp.h> |
1da177e4 LT |
93 | #include <linux/in.h> |
94 | #include <linux/errno.h> | |
95 | #include <linux/timer.h> | |
96 | #include <linux/mm.h> | |
1da177e4 | 97 | #include <linux/inet.h> |
1da177e4 | 98 | #include <linux/netdevice.h> |
5a0e3ad6 | 99 | #include <linux/slab.h> |
c752f073 | 100 | #include <net/tcp_states.h> |
1da177e4 LT |
101 | #include <linux/skbuff.h> |
102 | #include <linux/proc_fs.h> | |
103 | #include <linux/seq_file.h> | |
457c4cbc | 104 | #include <net/net_namespace.h> |
1da177e4 LT |
105 | #include <net/icmp.h> |
106 | #include <net/route.h> | |
1da177e4 LT |
107 | #include <net/checksum.h> |
108 | #include <net/xfrm.h> | |
296f7ea7 | 109 | #include <trace/events/udp.h> |
447167bf | 110 | #include <linux/static_key.h> |
22911fc5 | 111 | #include <trace/events/skb.h> |
a5b50476 | 112 | #include <net/ll_poll.h> |
ba4e58ec | 113 | #include "udp_impl.h" |
1da177e4 | 114 | |
f86dcc5a | 115 | struct udp_table udp_table __read_mostly; |
645ca708 | 116 | EXPORT_SYMBOL(udp_table); |
1da177e4 | 117 | |
8d987e5c | 118 | long sysctl_udp_mem[3] __read_mostly; |
95766fff | 119 | EXPORT_SYMBOL(sysctl_udp_mem); |
c482c568 ED |
120 | |
121 | int sysctl_udp_rmem_min __read_mostly; | |
95766fff | 122 | EXPORT_SYMBOL(sysctl_udp_rmem_min); |
c482c568 ED |
123 | |
124 | int sysctl_udp_wmem_min __read_mostly; | |
95766fff HA |
125 | EXPORT_SYMBOL(sysctl_udp_wmem_min); |
126 | ||
8d987e5c | 127 | atomic_long_t udp_memory_allocated; |
95766fff HA |
128 | EXPORT_SYMBOL(udp_memory_allocated); |
129 | ||
f86dcc5a ED |
130 | #define MAX_UDP_PORTS 65536 |
131 | #define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN) | |
98322f22 | 132 | |
f24d43c0 | 133 | static int udp_lib_lport_inuse(struct net *net, __u16 num, |
645ca708 | 134 | const struct udp_hslot *hslot, |
98322f22 | 135 | unsigned long *bitmap, |
f24d43c0 ED |
136 | struct sock *sk, |
137 | int (*saddr_comp)(const struct sock *sk1, | |
f86dcc5a ED |
138 | const struct sock *sk2), |
139 | unsigned int log) | |
1da177e4 | 140 | { |
f24d43c0 | 141 | struct sock *sk2; |
88ab1932 | 142 | struct hlist_nulls_node *node; |
ba418fa3 | 143 | kuid_t uid = sock_i_uid(sk); |
25030a7f | 144 | |
88ab1932 | 145 | sk_nulls_for_each(sk2, node, &hslot->head) |
9d4fb27d JP |
146 | if (net_eq(sock_net(sk2), net) && |
147 | sk2 != sk && | |
d4cada4a | 148 | (bitmap || udp_sk(sk2)->udp_port_hash == num) && |
9d4fb27d JP |
149 | (!sk2->sk_reuse || !sk->sk_reuse) && |
150 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
151 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 TH |
152 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
153 | !uid_eq(uid, sock_i_uid(sk2))) && | |
98322f22 ED |
154 | (*saddr_comp)(sk, sk2)) { |
155 | if (bitmap) | |
d4cada4a ED |
156 | __set_bit(udp_sk(sk2)->udp_port_hash >> log, |
157 | bitmap); | |
98322f22 ED |
158 | else |
159 | return 1; | |
160 | } | |
25030a7f GR |
161 | return 0; |
162 | } | |
163 | ||
30fff923 ED |
164 | /* |
165 | * Note: we still hold spinlock of primary hash chain, so no other writer | |
166 | * can insert/delete a socket with local_port == num | |
167 | */ | |
168 | static int udp_lib_lport_inuse2(struct net *net, __u16 num, | |
169 | struct udp_hslot *hslot2, | |
170 | struct sock *sk, | |
171 | int (*saddr_comp)(const struct sock *sk1, | |
172 | const struct sock *sk2)) | |
173 | { | |
174 | struct sock *sk2; | |
175 | struct hlist_nulls_node *node; | |
ba418fa3 | 176 | kuid_t uid = sock_i_uid(sk); |
30fff923 ED |
177 | int res = 0; |
178 | ||
179 | spin_lock(&hslot2->lock); | |
180 | udp_portaddr_for_each_entry(sk2, node, &hslot2->head) | |
9d4fb27d JP |
181 | if (net_eq(sock_net(sk2), net) && |
182 | sk2 != sk && | |
183 | (udp_sk(sk2)->udp_port_hash == num) && | |
184 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
185 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if || | |
186 | sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
ba418fa3 TH |
187 | (!sk2->sk_reuseport || !sk->sk_reuseport || |
188 | !uid_eq(uid, sock_i_uid(sk2))) && | |
30fff923 ED |
189 | (*saddr_comp)(sk, sk2)) { |
190 | res = 1; | |
191 | break; | |
192 | } | |
193 | spin_unlock(&hslot2->lock); | |
194 | return res; | |
195 | } | |
196 | ||
25030a7f | 197 | /** |
6ba5a3c5 | 198 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
199 | * |
200 | * @sk: socket struct in question | |
201 | * @snum: port number to look up | |
df2bc459 | 202 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25985edc | 203 | * @hash2_nulladdr: AF-dependent hash value in secondary hash chains, |
30fff923 | 204 | * with NULL address |
25030a7f | 205 | */ |
6ba5a3c5 | 206 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
df2bc459 | 207 | int (*saddr_comp)(const struct sock *sk1, |
30fff923 ED |
208 | const struct sock *sk2), |
209 | unsigned int hash2_nulladdr) | |
25030a7f | 210 | { |
512615b6 | 211 | struct udp_hslot *hslot, *hslot2; |
645ca708 | 212 | struct udp_table *udptable = sk->sk_prot->h.udp_table; |
25030a7f | 213 | int error = 1; |
3b1e0a65 | 214 | struct net *net = sock_net(sk); |
1da177e4 | 215 | |
32c1da70 | 216 | if (!snum) { |
9088c560 | 217 | int low, high, remaining; |
95c96174 | 218 | unsigned int rand; |
98322f22 ED |
219 | unsigned short first, last; |
220 | DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN); | |
32c1da70 | 221 | |
227b60f5 | 222 | inet_get_local_port_range(&low, &high); |
a25de534 | 223 | remaining = (high - low) + 1; |
227b60f5 | 224 | |
9088c560 | 225 | rand = net_random(); |
98322f22 ED |
226 | first = (((u64)rand * remaining) >> 32) + low; |
227 | /* | |
228 | * force rand to be an odd multiple of UDP_HTABLE_SIZE | |
229 | */ | |
f86dcc5a | 230 | rand = (rand | 1) * (udptable->mask + 1); |
5781b235 ED |
231 | last = first + udptable->mask + 1; |
232 | do { | |
f86dcc5a | 233 | hslot = udp_hashslot(udptable, net, first); |
98322f22 | 234 | bitmap_zero(bitmap, PORTS_PER_CHAIN); |
645ca708 | 235 | spin_lock_bh(&hslot->lock); |
98322f22 | 236 | udp_lib_lport_inuse(net, snum, hslot, bitmap, sk, |
f86dcc5a | 237 | saddr_comp, udptable->log); |
98322f22 ED |
238 | |
239 | snum = first; | |
240 | /* | |
241 | * Iterate on all possible values of snum for this hash. | |
242 | * Using steps of an odd multiple of UDP_HTABLE_SIZE | |
243 | * give us randomization and full range coverage. | |
244 | */ | |
9088c560 | 245 | do { |
98322f22 | 246 | if (low <= snum && snum <= high && |
e3826f1e AW |
247 | !test_bit(snum >> udptable->log, bitmap) && |
248 | !inet_is_reserved_local_port(snum)) | |
98322f22 ED |
249 | goto found; |
250 | snum += rand; | |
251 | } while (snum != first); | |
252 | spin_unlock_bh(&hslot->lock); | |
5781b235 | 253 | } while (++first != last); |
98322f22 | 254 | goto fail; |
645ca708 | 255 | } else { |
f86dcc5a | 256 | hslot = udp_hashslot(udptable, net, snum); |
645ca708 | 257 | spin_lock_bh(&hslot->lock); |
30fff923 ED |
258 | if (hslot->count > 10) { |
259 | int exist; | |
260 | unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum; | |
261 | ||
262 | slot2 &= udptable->mask; | |
263 | hash2_nulladdr &= udptable->mask; | |
264 | ||
265 | hslot2 = udp_hashslot2(udptable, slot2); | |
266 | if (hslot->count < hslot2->count) | |
267 | goto scan_primary_hash; | |
268 | ||
269 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
270 | sk, saddr_comp); | |
271 | if (!exist && (hash2_nulladdr != slot2)) { | |
272 | hslot2 = udp_hashslot2(udptable, hash2_nulladdr); | |
273 | exist = udp_lib_lport_inuse2(net, snum, hslot2, | |
274 | sk, saddr_comp); | |
275 | } | |
276 | if (exist) | |
277 | goto fail_unlock; | |
278 | else | |
279 | goto found; | |
280 | } | |
281 | scan_primary_hash: | |
f86dcc5a ED |
282 | if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk, |
283 | saddr_comp, 0)) | |
645ca708 ED |
284 | goto fail_unlock; |
285 | } | |
98322f22 | 286 | found: |
c720c7e8 | 287 | inet_sk(sk)->inet_num = snum; |
d4cada4a ED |
288 | udp_sk(sk)->udp_port_hash = snum; |
289 | udp_sk(sk)->udp_portaddr_hash ^= snum; | |
1da177e4 | 290 | if (sk_unhashed(sk)) { |
88ab1932 | 291 | sk_nulls_add_node_rcu(sk, &hslot->head); |
fdcc8aa9 | 292 | hslot->count++; |
c29a0bc4 | 293 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
512615b6 ED |
294 | |
295 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
296 | spin_lock(&hslot2->lock); | |
297 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
298 | &hslot2->head); | |
299 | hslot2->count++; | |
300 | spin_unlock(&hslot2->lock); | |
1da177e4 | 301 | } |
25030a7f | 302 | error = 0; |
645ca708 ED |
303 | fail_unlock: |
304 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 305 | fail: |
25030a7f GR |
306 | return error; |
307 | } | |
c482c568 | 308 | EXPORT_SYMBOL(udp_lib_get_port); |
25030a7f | 309 | |
499923c7 | 310 | static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2) |
db8dac20 DM |
311 | { |
312 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
313 | ||
c482c568 | 314 | return (!ipv6_only_sock(sk2) && |
c720c7e8 ED |
315 | (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr || |
316 | inet1->inet_rcv_saddr == inet2->inet_rcv_saddr)); | |
db8dac20 DM |
317 | } |
318 | ||
d4cada4a ED |
319 | static unsigned int udp4_portaddr_hash(struct net *net, __be32 saddr, |
320 | unsigned int port) | |
321 | { | |
0eae88f3 | 322 | return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port; |
d4cada4a ED |
323 | } |
324 | ||
6ba5a3c5 | 325 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 326 | { |
30fff923 | 327 | unsigned int hash2_nulladdr = |
0eae88f3 | 328 | udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum); |
30fff923 ED |
329 | unsigned int hash2_partial = |
330 | udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0); | |
331 | ||
d4cada4a | 332 | /* precompute partial secondary hash */ |
30fff923 ED |
333 | udp_sk(sk)->udp_portaddr_hash = hash2_partial; |
334 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr); | |
db8dac20 DM |
335 | } |
336 | ||
645ca708 ED |
337 | static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr, |
338 | unsigned short hnum, | |
339 | __be16 sport, __be32 daddr, __be16 dport, int dif) | |
340 | { | |
341 | int score = -1; | |
342 | ||
d4cada4a | 343 | if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum && |
645ca708 ED |
344 | !ipv6_only_sock(sk)) { |
345 | struct inet_sock *inet = inet_sk(sk); | |
346 | ||
ba418fa3 | 347 | score = (sk->sk_family == PF_INET ? 2 : 1); |
c720c7e8 ED |
348 | if (inet->inet_rcv_saddr) { |
349 | if (inet->inet_rcv_saddr != daddr) | |
645ca708 | 350 | return -1; |
ba418fa3 | 351 | score += 4; |
645ca708 | 352 | } |
c720c7e8 ED |
353 | if (inet->inet_daddr) { |
354 | if (inet->inet_daddr != saddr) | |
645ca708 | 355 | return -1; |
ba418fa3 | 356 | score += 4; |
645ca708 | 357 | } |
c720c7e8 ED |
358 | if (inet->inet_dport) { |
359 | if (inet->inet_dport != sport) | |
645ca708 | 360 | return -1; |
ba418fa3 | 361 | score += 4; |
645ca708 ED |
362 | } |
363 | if (sk->sk_bound_dev_if) { | |
364 | if (sk->sk_bound_dev_if != dif) | |
365 | return -1; | |
ba418fa3 | 366 | score += 4; |
645ca708 ED |
367 | } |
368 | } | |
369 | return score; | |
370 | } | |
371 | ||
5051ebd2 ED |
372 | /* |
373 | * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num) | |
374 | */ | |
5051ebd2 ED |
375 | static inline int compute_score2(struct sock *sk, struct net *net, |
376 | __be32 saddr, __be16 sport, | |
377 | __be32 daddr, unsigned int hnum, int dif) | |
378 | { | |
379 | int score = -1; | |
380 | ||
381 | if (net_eq(sock_net(sk), net) && !ipv6_only_sock(sk)) { | |
382 | struct inet_sock *inet = inet_sk(sk); | |
383 | ||
384 | if (inet->inet_rcv_saddr != daddr) | |
385 | return -1; | |
386 | if (inet->inet_num != hnum) | |
387 | return -1; | |
388 | ||
ba418fa3 | 389 | score = (sk->sk_family == PF_INET ? 2 : 1); |
5051ebd2 ED |
390 | if (inet->inet_daddr) { |
391 | if (inet->inet_daddr != saddr) | |
392 | return -1; | |
ba418fa3 | 393 | score += 4; |
5051ebd2 ED |
394 | } |
395 | if (inet->inet_dport) { | |
396 | if (inet->inet_dport != sport) | |
397 | return -1; | |
ba418fa3 | 398 | score += 4; |
5051ebd2 ED |
399 | } |
400 | if (sk->sk_bound_dev_if) { | |
401 | if (sk->sk_bound_dev_if != dif) | |
402 | return -1; | |
ba418fa3 | 403 | score += 4; |
5051ebd2 ED |
404 | } |
405 | } | |
406 | return score; | |
407 | } | |
408 | ||
5051ebd2 ED |
409 | |
410 | /* called with read_rcu_lock() */ | |
411 | static struct sock *udp4_lib_lookup2(struct net *net, | |
412 | __be32 saddr, __be16 sport, | |
413 | __be32 daddr, unsigned int hnum, int dif, | |
414 | struct udp_hslot *hslot2, unsigned int slot2) | |
415 | { | |
416 | struct sock *sk, *result; | |
417 | struct hlist_nulls_node *node; | |
ba418fa3 TH |
418 | int score, badness, matches = 0, reuseport = 0; |
419 | u32 hash = 0; | |
5051ebd2 ED |
420 | |
421 | begin: | |
422 | result = NULL; | |
ba418fa3 | 423 | badness = 0; |
5051ebd2 ED |
424 | udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) { |
425 | score = compute_score2(sk, net, saddr, sport, | |
426 | daddr, hnum, dif); | |
427 | if (score > badness) { | |
428 | result = sk; | |
429 | badness = score; | |
ba418fa3 TH |
430 | reuseport = sk->sk_reuseport; |
431 | if (reuseport) { | |
432 | hash = inet_ehashfn(net, daddr, hnum, | |
7c0cadc6 | 433 | saddr, sport); |
ba418fa3 TH |
434 | matches = 1; |
435 | } | |
436 | } else if (score == badness && reuseport) { | |
437 | matches++; | |
438 | if (((u64)hash * matches) >> 32 == 0) | |
439 | result = sk; | |
440 | hash = next_pseudo_random32(hash); | |
5051ebd2 ED |
441 | } |
442 | } | |
443 | /* | |
444 | * if the nulls value we got at the end of this lookup is | |
445 | * not the expected one, we must restart lookup. | |
446 | * We probably met an item that was moved to another chain. | |
447 | */ | |
448 | if (get_nulls_value(node) != slot2) | |
449 | goto begin; | |
5051ebd2 | 450 | if (result) { |
c31504dc | 451 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
5051ebd2 ED |
452 | result = NULL; |
453 | else if (unlikely(compute_score2(result, net, saddr, sport, | |
454 | daddr, hnum, dif) < badness)) { | |
455 | sock_put(result); | |
456 | goto begin; | |
457 | } | |
458 | } | |
459 | return result; | |
460 | } | |
461 | ||
db8dac20 DM |
462 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
463 | * harder than this. -DaveM | |
464 | */ | |
fce82338 | 465 | struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, |
db8dac20 | 466 | __be16 sport, __be32 daddr, __be16 dport, |
645ca708 | 467 | int dif, struct udp_table *udptable) |
db8dac20 | 468 | { |
271b72c7 | 469 | struct sock *sk, *result; |
88ab1932 | 470 | struct hlist_nulls_node *node; |
db8dac20 | 471 | unsigned short hnum = ntohs(dport); |
5051ebd2 ED |
472 | unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask); |
473 | struct udp_hslot *hslot2, *hslot = &udptable->hash[slot]; | |
ba418fa3 TH |
474 | int score, badness, matches = 0, reuseport = 0; |
475 | u32 hash = 0; | |
645ca708 | 476 | |
271b72c7 | 477 | rcu_read_lock(); |
5051ebd2 ED |
478 | if (hslot->count > 10) { |
479 | hash2 = udp4_portaddr_hash(net, daddr, hnum); | |
480 | slot2 = hash2 & udptable->mask; | |
481 | hslot2 = &udptable->hash2[slot2]; | |
482 | if (hslot->count < hslot2->count) | |
483 | goto begin; | |
484 | ||
485 | result = udp4_lib_lookup2(net, saddr, sport, | |
486 | daddr, hnum, dif, | |
487 | hslot2, slot2); | |
488 | if (!result) { | |
0eae88f3 | 489 | hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum); |
5051ebd2 ED |
490 | slot2 = hash2 & udptable->mask; |
491 | hslot2 = &udptable->hash2[slot2]; | |
492 | if (hslot->count < hslot2->count) | |
493 | goto begin; | |
494 | ||
1223c67c | 495 | result = udp4_lib_lookup2(net, saddr, sport, |
0eae88f3 | 496 | htonl(INADDR_ANY), hnum, dif, |
5051ebd2 ED |
497 | hslot2, slot2); |
498 | } | |
499 | rcu_read_unlock(); | |
500 | return result; | |
501 | } | |
271b72c7 ED |
502 | begin: |
503 | result = NULL; | |
ba418fa3 | 504 | badness = 0; |
88ab1932 | 505 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { |
645ca708 ED |
506 | score = compute_score(sk, net, saddr, hnum, sport, |
507 | daddr, dport, dif); | |
508 | if (score > badness) { | |
509 | result = sk; | |
510 | badness = score; | |
ba418fa3 TH |
511 | reuseport = sk->sk_reuseport; |
512 | if (reuseport) { | |
513 | hash = inet_ehashfn(net, daddr, hnum, | |
7c0cadc6 | 514 | saddr, sport); |
ba418fa3 TH |
515 | matches = 1; |
516 | } | |
517 | } else if (score == badness && reuseport) { | |
518 | matches++; | |
519 | if (((u64)hash * matches) >> 32 == 0) | |
520 | result = sk; | |
521 | hash = next_pseudo_random32(hash); | |
db8dac20 DM |
522 | } |
523 | } | |
88ab1932 ED |
524 | /* |
525 | * if the nulls value we got at the end of this lookup is | |
526 | * not the expected one, we must restart lookup. | |
527 | * We probably met an item that was moved to another chain. | |
528 | */ | |
5051ebd2 | 529 | if (get_nulls_value(node) != slot) |
88ab1932 ED |
530 | goto begin; |
531 | ||
271b72c7 | 532 | if (result) { |
c31504dc | 533 | if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2))) |
271b72c7 ED |
534 | result = NULL; |
535 | else if (unlikely(compute_score(result, net, saddr, hnum, sport, | |
536 | daddr, dport, dif) < badness)) { | |
537 | sock_put(result); | |
538 | goto begin; | |
539 | } | |
540 | } | |
541 | rcu_read_unlock(); | |
db8dac20 DM |
542 | return result; |
543 | } | |
fce82338 | 544 | EXPORT_SYMBOL_GPL(__udp4_lib_lookup); |
db8dac20 | 545 | |
607c4aaf KK |
546 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
547 | __be16 sport, __be16 dport, | |
645ca708 | 548 | struct udp_table *udptable) |
607c4aaf | 549 | { |
23542618 | 550 | struct sock *sk; |
607c4aaf KK |
551 | const struct iphdr *iph = ip_hdr(skb); |
552 | ||
23542618 KK |
553 | if (unlikely(sk = skb_steal_sock(skb))) |
554 | return sk; | |
555 | else | |
adf30907 | 556 | return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport, |
23542618 KK |
557 | iph->daddr, dport, inet_iif(skb), |
558 | udptable); | |
607c4aaf KK |
559 | } |
560 | ||
bcd41303 KK |
561 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
562 | __be32 daddr, __be16 dport, int dif) | |
563 | { | |
645ca708 | 564 | return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table); |
bcd41303 KK |
565 | } |
566 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
567 | ||
920a4611 | 568 | static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk, |
db8dac20 DM |
569 | __be16 loc_port, __be32 loc_addr, |
570 | __be16 rmt_port, __be32 rmt_addr, | |
571 | int dif) | |
572 | { | |
88ab1932 | 573 | struct hlist_nulls_node *node; |
db8dac20 DM |
574 | struct sock *s = sk; |
575 | unsigned short hnum = ntohs(loc_port); | |
576 | ||
88ab1932 | 577 | sk_nulls_for_each_from(s, node) { |
db8dac20 DM |
578 | struct inet_sock *inet = inet_sk(s); |
579 | ||
9d4fb27d JP |
580 | if (!net_eq(sock_net(s), net) || |
581 | udp_sk(s)->udp_port_hash != hnum || | |
582 | (inet->inet_daddr && inet->inet_daddr != rmt_addr) || | |
583 | (inet->inet_dport != rmt_port && inet->inet_dport) || | |
584 | (inet->inet_rcv_saddr && | |
585 | inet->inet_rcv_saddr != loc_addr) || | |
586 | ipv6_only_sock(s) || | |
db8dac20 DM |
587 | (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)) |
588 | continue; | |
589 | if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif)) | |
590 | continue; | |
591 | goto found; | |
592 | } | |
593 | s = NULL; | |
594 | found: | |
595 | return s; | |
596 | } | |
597 | ||
598 | /* | |
599 | * This routine is called by the ICMP module when it gets some | |
600 | * sort of error condition. If err < 0 then the socket should | |
601 | * be closed and the error returned to the user. If err > 0 | |
602 | * it's just the icmp type << 8 | icmp code. | |
603 | * Header points to the ip header of the error packet. We move | |
604 | * on past this. Then (as it used to claim before adjustment) | |
605 | * header points to the first 8 bytes of the udp header. We need | |
606 | * to find the appropriate port. | |
607 | */ | |
608 | ||
645ca708 | 609 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
610 | { |
611 | struct inet_sock *inet; | |
b71d1d42 | 612 | const struct iphdr *iph = (const struct iphdr *)skb->data; |
c482c568 | 613 | struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2)); |
db8dac20 DM |
614 | const int type = icmp_hdr(skb)->type; |
615 | const int code = icmp_hdr(skb)->code; | |
616 | struct sock *sk; | |
617 | int harderr; | |
618 | int err; | |
fd54d716 | 619 | struct net *net = dev_net(skb->dev); |
db8dac20 | 620 | |
fd54d716 | 621 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
db8dac20 DM |
622 | iph->saddr, uh->source, skb->dev->ifindex, udptable); |
623 | if (sk == NULL) { | |
dcfc23ca | 624 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
625 | return; /* No socket for error */ |
626 | } | |
627 | ||
628 | err = 0; | |
629 | harderr = 0; | |
630 | inet = inet_sk(sk); | |
631 | ||
632 | switch (type) { | |
633 | default: | |
634 | case ICMP_TIME_EXCEEDED: | |
635 | err = EHOSTUNREACH; | |
636 | break; | |
637 | case ICMP_SOURCE_QUENCH: | |
638 | goto out; | |
639 | case ICMP_PARAMETERPROB: | |
640 | err = EPROTO; | |
641 | harderr = 1; | |
642 | break; | |
643 | case ICMP_DEST_UNREACH: | |
644 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
36393395 | 645 | ipv4_sk_update_pmtu(skb, sk, info); |
db8dac20 DM |
646 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { |
647 | err = EMSGSIZE; | |
648 | harderr = 1; | |
649 | break; | |
650 | } | |
651 | goto out; | |
652 | } | |
653 | err = EHOSTUNREACH; | |
654 | if (code <= NR_ICMP_UNREACH) { | |
655 | harderr = icmp_err_convert[code].fatal; | |
656 | err = icmp_err_convert[code].errno; | |
657 | } | |
658 | break; | |
55be7a9c DM |
659 | case ICMP_REDIRECT: |
660 | ipv4_sk_redirect(skb, sk); | |
661 | break; | |
db8dac20 DM |
662 | } |
663 | ||
664 | /* | |
665 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
666 | * 4.1.3.3. | |
667 | */ | |
668 | if (!inet->recverr) { | |
669 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
670 | goto out; | |
b1faf566 | 671 | } else |
c482c568 | 672 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1)); |
b1faf566 | 673 | |
db8dac20 DM |
674 | sk->sk_err = err; |
675 | sk->sk_error_report(sk); | |
676 | out: | |
677 | sock_put(sk); | |
678 | } | |
679 | ||
680 | void udp_err(struct sk_buff *skb, u32 info) | |
681 | { | |
645ca708 | 682 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
683 | } |
684 | ||
685 | /* | |
686 | * Throw away all pending data and cancel the corking. Socket is locked. | |
687 | */ | |
36d926b9 | 688 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
689 | { |
690 | struct udp_sock *up = udp_sk(sk); | |
691 | ||
692 | if (up->pending) { | |
693 | up->len = 0; | |
694 | up->pending = 0; | |
695 | ip_flush_pending_frames(sk); | |
696 | } | |
697 | } | |
36d926b9 | 698 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
699 | |
700 | /** | |
f6b9664f | 701 | * udp4_hwcsum - handle outgoing HW checksumming |
db8dac20 DM |
702 | * @skb: sk_buff containing the filled-in UDP header |
703 | * (checksum field must be zeroed out) | |
f6b9664f HX |
704 | * @src: source IP address |
705 | * @dst: destination IP address | |
db8dac20 | 706 | */ |
f6b9664f | 707 | static void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst) |
db8dac20 | 708 | { |
db8dac20 | 709 | struct udphdr *uh = udp_hdr(skb); |
f6b9664f HX |
710 | struct sk_buff *frags = skb_shinfo(skb)->frag_list; |
711 | int offset = skb_transport_offset(skb); | |
712 | int len = skb->len - offset; | |
713 | int hlen = len; | |
db8dac20 DM |
714 | __wsum csum = 0; |
715 | ||
f6b9664f | 716 | if (!frags) { |
db8dac20 DM |
717 | /* |
718 | * Only one fragment on the socket. | |
719 | */ | |
720 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
721 | skb->csum_offset = offsetof(struct udphdr, check); | |
f6b9664f HX |
722 | uh->check = ~csum_tcpudp_magic(src, dst, len, |
723 | IPPROTO_UDP, 0); | |
db8dac20 DM |
724 | } else { |
725 | /* | |
726 | * HW-checksum won't work as there are two or more | |
727 | * fragments on the socket so that all csums of sk_buffs | |
728 | * should be together | |
729 | */ | |
f6b9664f HX |
730 | do { |
731 | csum = csum_add(csum, frags->csum); | |
732 | hlen -= frags->len; | |
733 | } while ((frags = frags->next)); | |
db8dac20 | 734 | |
f6b9664f | 735 | csum = skb_checksum(skb, offset, hlen, csum); |
db8dac20 DM |
736 | skb->ip_summed = CHECKSUM_NONE; |
737 | ||
db8dac20 DM |
738 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); |
739 | if (uh->check == 0) | |
740 | uh->check = CSUM_MANGLED_0; | |
741 | } | |
742 | } | |
743 | ||
79ab0531 | 744 | static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4) |
db8dac20 | 745 | { |
f6b9664f | 746 | struct sock *sk = skb->sk; |
db8dac20 | 747 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
748 | struct udphdr *uh; |
749 | int err = 0; | |
750 | int is_udplite = IS_UDPLITE(sk); | |
f6b9664f HX |
751 | int offset = skb_transport_offset(skb); |
752 | int len = skb->len - offset; | |
db8dac20 DM |
753 | __wsum csum = 0; |
754 | ||
db8dac20 DM |
755 | /* |
756 | * Create a UDP header | |
757 | */ | |
758 | uh = udp_hdr(skb); | |
f6b9664f | 759 | uh->source = inet->inet_sport; |
79ab0531 | 760 | uh->dest = fl4->fl4_dport; |
f6b9664f | 761 | uh->len = htons(len); |
db8dac20 DM |
762 | uh->check = 0; |
763 | ||
764 | if (is_udplite) /* UDP-Lite */ | |
f6b9664f | 765 | csum = udplite_csum(skb); |
db8dac20 DM |
766 | |
767 | else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */ | |
768 | ||
769 | skb->ip_summed = CHECKSUM_NONE; | |
770 | goto send; | |
771 | ||
772 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
773 | ||
79ab0531 | 774 | udp4_hwcsum(skb, fl4->saddr, fl4->daddr); |
db8dac20 DM |
775 | goto send; |
776 | ||
f6b9664f HX |
777 | } else |
778 | csum = udp_csum(skb); | |
db8dac20 DM |
779 | |
780 | /* add protocol-dependent pseudo-header */ | |
79ab0531 | 781 | uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len, |
c482c568 | 782 | sk->sk_protocol, csum); |
db8dac20 DM |
783 | if (uh->check == 0) |
784 | uh->check = CSUM_MANGLED_0; | |
785 | ||
786 | send: | |
b5ec8eea | 787 | err = ip_send_skb(sock_net(sk), skb); |
6ce9e7b5 ED |
788 | if (err) { |
789 | if (err == -ENOBUFS && !inet->recverr) { | |
790 | UDP_INC_STATS_USER(sock_net(sk), | |
791 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
792 | err = 0; | |
793 | } | |
794 | } else | |
795 | UDP_INC_STATS_USER(sock_net(sk), | |
796 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
f6b9664f HX |
797 | return err; |
798 | } | |
799 | ||
800 | /* | |
801 | * Push out all pending data as one UDP datagram. Socket is locked. | |
802 | */ | |
803 | static int udp_push_pending_frames(struct sock *sk) | |
804 | { | |
805 | struct udp_sock *up = udp_sk(sk); | |
806 | struct inet_sock *inet = inet_sk(sk); | |
b6f21b26 | 807 | struct flowi4 *fl4 = &inet->cork.fl.u.ip4; |
f6b9664f HX |
808 | struct sk_buff *skb; |
809 | int err = 0; | |
810 | ||
77968b78 | 811 | skb = ip_finish_skb(sk, fl4); |
f6b9664f HX |
812 | if (!skb) |
813 | goto out; | |
814 | ||
79ab0531 | 815 | err = udp_send_skb(skb, fl4); |
f6b9664f | 816 | |
db8dac20 DM |
817 | out: |
818 | up->len = 0; | |
819 | up->pending = 0; | |
db8dac20 DM |
820 | return err; |
821 | } | |
822 | ||
823 | int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
824 | size_t len) | |
825 | { | |
826 | struct inet_sock *inet = inet_sk(sk); | |
827 | struct udp_sock *up = udp_sk(sk); | |
e474995f | 828 | struct flowi4 fl4_stack; |
b6f21b26 | 829 | struct flowi4 *fl4; |
db8dac20 DM |
830 | int ulen = len; |
831 | struct ipcm_cookie ipc; | |
832 | struct rtable *rt = NULL; | |
833 | int free = 0; | |
834 | int connected = 0; | |
835 | __be32 daddr, faddr, saddr; | |
836 | __be16 dport; | |
837 | u8 tos; | |
838 | int err, is_udplite = IS_UDPLITE(sk); | |
839 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
840 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
903ab86d | 841 | struct sk_buff *skb; |
f6d8bd05 | 842 | struct ip_options_data opt_copy; |
db8dac20 DM |
843 | |
844 | if (len > 0xFFFF) | |
845 | return -EMSGSIZE; | |
846 | ||
847 | /* | |
848 | * Check the flags. | |
849 | */ | |
850 | ||
c482c568 | 851 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */ |
db8dac20 DM |
852 | return -EOPNOTSUPP; |
853 | ||
854 | ipc.opt = NULL; | |
2244d07b | 855 | ipc.tx_flags = 0; |
db8dac20 | 856 | |
903ab86d HX |
857 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
858 | ||
f5fca608 | 859 | fl4 = &inet->cork.fl.u.ip4; |
db8dac20 DM |
860 | if (up->pending) { |
861 | /* | |
862 | * There are pending frames. | |
863 | * The socket lock must be held while it's corked. | |
864 | */ | |
865 | lock_sock(sk); | |
866 | if (likely(up->pending)) { | |
867 | if (unlikely(up->pending != AF_INET)) { | |
868 | release_sock(sk); | |
869 | return -EINVAL; | |
870 | } | |
871 | goto do_append_data; | |
872 | } | |
873 | release_sock(sk); | |
874 | } | |
875 | ulen += sizeof(struct udphdr); | |
876 | ||
877 | /* | |
878 | * Get and verify the address. | |
879 | */ | |
880 | if (msg->msg_name) { | |
5e73ea1a | 881 | struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name; |
db8dac20 DM |
882 | if (msg->msg_namelen < sizeof(*usin)) |
883 | return -EINVAL; | |
884 | if (usin->sin_family != AF_INET) { | |
885 | if (usin->sin_family != AF_UNSPEC) | |
886 | return -EAFNOSUPPORT; | |
887 | } | |
888 | ||
889 | daddr = usin->sin_addr.s_addr; | |
890 | dport = usin->sin_port; | |
891 | if (dport == 0) | |
892 | return -EINVAL; | |
893 | } else { | |
894 | if (sk->sk_state != TCP_ESTABLISHED) | |
895 | return -EDESTADDRREQ; | |
c720c7e8 ED |
896 | daddr = inet->inet_daddr; |
897 | dport = inet->inet_dport; | |
db8dac20 DM |
898 | /* Open fast path for connected socket. |
899 | Route will not be used, if at least one option is set. | |
900 | */ | |
901 | connected = 1; | |
902 | } | |
c720c7e8 | 903 | ipc.addr = inet->inet_saddr; |
db8dac20 DM |
904 | |
905 | ipc.oif = sk->sk_bound_dev_if; | |
bf84a010 DB |
906 | |
907 | sock_tx_timestamp(sk, &ipc.tx_flags); | |
908 | ||
db8dac20 | 909 | if (msg->msg_controllen) { |
3b1e0a65 | 910 | err = ip_cmsg_send(sock_net(sk), msg, &ipc); |
db8dac20 DM |
911 | if (err) |
912 | return err; | |
913 | if (ipc.opt) | |
914 | free = 1; | |
915 | connected = 0; | |
916 | } | |
f6d8bd05 ED |
917 | if (!ipc.opt) { |
918 | struct ip_options_rcu *inet_opt; | |
919 | ||
920 | rcu_read_lock(); | |
921 | inet_opt = rcu_dereference(inet->inet_opt); | |
922 | if (inet_opt) { | |
923 | memcpy(&opt_copy, inet_opt, | |
924 | sizeof(*inet_opt) + inet_opt->opt.optlen); | |
925 | ipc.opt = &opt_copy.opt; | |
926 | } | |
927 | rcu_read_unlock(); | |
928 | } | |
db8dac20 DM |
929 | |
930 | saddr = ipc.addr; | |
931 | ipc.addr = faddr = daddr; | |
932 | ||
f6d8bd05 | 933 | if (ipc.opt && ipc.opt->opt.srr) { |
db8dac20 DM |
934 | if (!daddr) |
935 | return -EINVAL; | |
f6d8bd05 | 936 | faddr = ipc.opt->opt.faddr; |
db8dac20 DM |
937 | connected = 0; |
938 | } | |
939 | tos = RT_TOS(inet->tos); | |
940 | if (sock_flag(sk, SOCK_LOCALROUTE) || | |
941 | (msg->msg_flags & MSG_DONTROUTE) || | |
f6d8bd05 | 942 | (ipc.opt && ipc.opt->opt.is_strictroute)) { |
db8dac20 DM |
943 | tos |= RTO_ONLINK; |
944 | connected = 0; | |
945 | } | |
946 | ||
947 | if (ipv4_is_multicast(daddr)) { | |
948 | if (!ipc.oif) | |
949 | ipc.oif = inet->mc_index; | |
950 | if (!saddr) | |
951 | saddr = inet->mc_addr; | |
952 | connected = 0; | |
76e21053 EH |
953 | } else if (!ipc.oif) |
954 | ipc.oif = inet->uc_index; | |
db8dac20 DM |
955 | |
956 | if (connected) | |
c482c568 | 957 | rt = (struct rtable *)sk_dst_check(sk, 0); |
db8dac20 DM |
958 | |
959 | if (rt == NULL) { | |
84a3aa00 PE |
960 | struct net *net = sock_net(sk); |
961 | ||
e474995f DM |
962 | fl4 = &fl4_stack; |
963 | flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos, | |
c0951cbc DM |
964 | RT_SCOPE_UNIVERSE, sk->sk_protocol, |
965 | inet_sk_flowi_flags(sk)|FLOWI_FLAG_CAN_SLEEP, | |
966 | faddr, saddr, dport, inet->inet_sport); | |
967 | ||
e474995f DM |
968 | security_sk_classify_flow(sk, flowi4_to_flowi(fl4)); |
969 | rt = ip_route_output_flow(net, fl4, sk); | |
b23dd4fe DM |
970 | if (IS_ERR(rt)) { |
971 | err = PTR_ERR(rt); | |
06dc94b1 | 972 | rt = NULL; |
db8dac20 | 973 | if (err == -ENETUNREACH) |
7c73a6fa | 974 | IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
975 | goto out; |
976 | } | |
977 | ||
978 | err = -EACCES; | |
979 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
980 | !sock_flag(sk, SOCK_BROADCAST)) | |
981 | goto out; | |
982 | if (connected) | |
d8d1f30b | 983 | sk_dst_set(sk, dst_clone(&rt->dst)); |
db8dac20 DM |
984 | } |
985 | ||
986 | if (msg->msg_flags&MSG_CONFIRM) | |
987 | goto do_confirm; | |
988 | back_from_confirm: | |
989 | ||
e474995f | 990 | saddr = fl4->saddr; |
db8dac20 | 991 | if (!ipc.addr) |
e474995f | 992 | daddr = ipc.addr = fl4->daddr; |
db8dac20 | 993 | |
903ab86d HX |
994 | /* Lockless fast path for the non-corking case. */ |
995 | if (!corkreq) { | |
77968b78 | 996 | skb = ip_make_skb(sk, fl4, getfrag, msg->msg_iov, ulen, |
903ab86d HX |
997 | sizeof(struct udphdr), &ipc, &rt, |
998 | msg->msg_flags); | |
999 | err = PTR_ERR(skb); | |
50c3a487 | 1000 | if (!IS_ERR_OR_NULL(skb)) |
79ab0531 | 1001 | err = udp_send_skb(skb, fl4); |
903ab86d HX |
1002 | goto out; |
1003 | } | |
1004 | ||
db8dac20 DM |
1005 | lock_sock(sk); |
1006 | if (unlikely(up->pending)) { | |
1007 | /* The socket is already corked while preparing it. */ | |
1008 | /* ... which is an evident application bug. --ANK */ | |
1009 | release_sock(sk); | |
1010 | ||
afd46503 | 1011 | LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("cork app bug 2\n")); |
db8dac20 DM |
1012 | err = -EINVAL; |
1013 | goto out; | |
1014 | } | |
1015 | /* | |
1016 | * Now cork the socket to pend data. | |
1017 | */ | |
b6f21b26 DM |
1018 | fl4 = &inet->cork.fl.u.ip4; |
1019 | fl4->daddr = daddr; | |
1020 | fl4->saddr = saddr; | |
9cce96df DM |
1021 | fl4->fl4_dport = dport; |
1022 | fl4->fl4_sport = inet->inet_sport; | |
db8dac20 DM |
1023 | up->pending = AF_INET; |
1024 | ||
1025 | do_append_data: | |
1026 | up->len += ulen; | |
f5fca608 DM |
1027 | err = ip_append_data(sk, fl4, getfrag, msg->msg_iov, ulen, |
1028 | sizeof(struct udphdr), &ipc, &rt, | |
1029 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | |
db8dac20 DM |
1030 | if (err) |
1031 | udp_flush_pending_frames(sk); | |
1032 | else if (!corkreq) | |
1033 | err = udp_push_pending_frames(sk); | |
1034 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
1035 | up->pending = 0; | |
1036 | release_sock(sk); | |
1037 | ||
1038 | out: | |
1039 | ip_rt_put(rt); | |
1040 | if (free) | |
1041 | kfree(ipc.opt); | |
1042 | if (!err) | |
1043 | return len; | |
1044 | /* | |
1045 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
1046 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
1047 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
1048 | * things). We could add another new stat but at least for now that | |
1049 | * seems like overkill. | |
1050 | */ | |
1051 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
1052 | UDP_INC_STATS_USER(sock_net(sk), |
1053 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
1054 | } |
1055 | return err; | |
1056 | ||
1057 | do_confirm: | |
d8d1f30b | 1058 | dst_confirm(&rt->dst); |
db8dac20 DM |
1059 | if (!(msg->msg_flags&MSG_PROBE) || len) |
1060 | goto back_from_confirm; | |
1061 | err = 0; | |
1062 | goto out; | |
1063 | } | |
c482c568 | 1064 | EXPORT_SYMBOL(udp_sendmsg); |
db8dac20 DM |
1065 | |
1066 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
1067 | size_t size, int flags) | |
1068 | { | |
f5fca608 | 1069 | struct inet_sock *inet = inet_sk(sk); |
db8dac20 DM |
1070 | struct udp_sock *up = udp_sk(sk); |
1071 | int ret; | |
1072 | ||
1073 | if (!up->pending) { | |
1074 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
1075 | ||
1076 | /* Call udp_sendmsg to specify destination address which | |
1077 | * sendpage interface can't pass. | |
1078 | * This will succeed only when the socket is connected. | |
1079 | */ | |
1080 | ret = udp_sendmsg(NULL, sk, &msg, 0); | |
1081 | if (ret < 0) | |
1082 | return ret; | |
1083 | } | |
1084 | ||
1085 | lock_sock(sk); | |
1086 | ||
1087 | if (unlikely(!up->pending)) { | |
1088 | release_sock(sk); | |
1089 | ||
afd46503 | 1090 | LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("udp cork app bug 3\n")); |
db8dac20 DM |
1091 | return -EINVAL; |
1092 | } | |
1093 | ||
f5fca608 DM |
1094 | ret = ip_append_page(sk, &inet->cork.fl.u.ip4, |
1095 | page, offset, size, flags); | |
db8dac20 DM |
1096 | if (ret == -EOPNOTSUPP) { |
1097 | release_sock(sk); | |
1098 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
1099 | size, flags); | |
1100 | } | |
1101 | if (ret < 0) { | |
1102 | udp_flush_pending_frames(sk); | |
1103 | goto out; | |
1104 | } | |
1105 | ||
1106 | up->len += size; | |
1107 | if (!(up->corkflag || (flags&MSG_MORE))) | |
1108 | ret = udp_push_pending_frames(sk); | |
1109 | if (!ret) | |
1110 | ret = size; | |
1111 | out: | |
1112 | release_sock(sk); | |
1113 | return ret; | |
1114 | } | |
1115 | ||
85584672 ED |
1116 | |
1117 | /** | |
1118 | * first_packet_length - return length of first packet in receive queue | |
1119 | * @sk: socket | |
1120 | * | |
1121 | * Drops all bad checksum frames, until a valid one is found. | |
1122 | * Returns the length of found skb, or 0 if none is found. | |
1123 | */ | |
1124 | static unsigned int first_packet_length(struct sock *sk) | |
1125 | { | |
1126 | struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue; | |
1127 | struct sk_buff *skb; | |
1128 | unsigned int res; | |
1129 | ||
1130 | __skb_queue_head_init(&list_kill); | |
1131 | ||
1132 | spin_lock_bh(&rcvq->lock); | |
1133 | while ((skb = skb_peek(rcvq)) != NULL && | |
1134 | udp_lib_checksum_complete(skb)) { | |
6a5dc9e5 ED |
1135 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, |
1136 | IS_UDPLITE(sk)); | |
85584672 ED |
1137 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, |
1138 | IS_UDPLITE(sk)); | |
8edf19c2 | 1139 | atomic_inc(&sk->sk_drops); |
85584672 ED |
1140 | __skb_unlink(skb, rcvq); |
1141 | __skb_queue_tail(&list_kill, skb); | |
1142 | } | |
1143 | res = skb ? skb->len : 0; | |
1144 | spin_unlock_bh(&rcvq->lock); | |
1145 | ||
1146 | if (!skb_queue_empty(&list_kill)) { | |
8a74ad60 ED |
1147 | bool slow = lock_sock_fast(sk); |
1148 | ||
85584672 ED |
1149 | __skb_queue_purge(&list_kill); |
1150 | sk_mem_reclaim_partial(sk); | |
8a74ad60 | 1151 | unlock_sock_fast(sk, slow); |
85584672 ED |
1152 | } |
1153 | return res; | |
1154 | } | |
1155 | ||
1da177e4 LT |
1156 | /* |
1157 | * IOCTL requests applicable to the UDP protocol | |
1158 | */ | |
e905a9ed | 1159 | |
1da177e4 LT |
1160 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
1161 | { | |
6516c655 SH |
1162 | switch (cmd) { |
1163 | case SIOCOUTQ: | |
1da177e4 | 1164 | { |
31e6d363 ED |
1165 | int amount = sk_wmem_alloc_get(sk); |
1166 | ||
6516c655 SH |
1167 | return put_user(amount, (int __user *)arg); |
1168 | } | |
1da177e4 | 1169 | |
6516c655 SH |
1170 | case SIOCINQ: |
1171 | { | |
85584672 | 1172 | unsigned int amount = first_packet_length(sk); |
6516c655 | 1173 | |
85584672 | 1174 | if (amount) |
6516c655 SH |
1175 | /* |
1176 | * We will only return the amount | |
1177 | * of this packet since that is all | |
1178 | * that will be read. | |
1179 | */ | |
85584672 ED |
1180 | amount -= sizeof(struct udphdr); |
1181 | ||
6516c655 SH |
1182 | return put_user(amount, (int __user *)arg); |
1183 | } | |
1da177e4 | 1184 | |
6516c655 SH |
1185 | default: |
1186 | return -ENOIOCTLCMD; | |
1da177e4 | 1187 | } |
6516c655 SH |
1188 | |
1189 | return 0; | |
1da177e4 | 1190 | } |
c482c568 | 1191 | EXPORT_SYMBOL(udp_ioctl); |
1da177e4 | 1192 | |
db8dac20 DM |
1193 | /* |
1194 | * This should be easy, if there is something there we | |
1195 | * return it, otherwise we block. | |
1196 | */ | |
1197 | ||
1198 | int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
1199 | size_t len, int noblock, int flags, int *addr_len) | |
1200 | { | |
1201 | struct inet_sock *inet = inet_sk(sk); | |
1202 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; | |
1203 | struct sk_buff *skb; | |
59c2cdae | 1204 | unsigned int ulen, copied; |
3f518bf7 | 1205 | int peeked, off = 0; |
db8dac20 DM |
1206 | int err; |
1207 | int is_udplite = IS_UDPLITE(sk); | |
8a74ad60 | 1208 | bool slow; |
db8dac20 DM |
1209 | |
1210 | /* | |
1211 | * Check any passed addresses | |
1212 | */ | |
1213 | if (addr_len) | |
c482c568 | 1214 | *addr_len = sizeof(*sin); |
db8dac20 DM |
1215 | |
1216 | if (flags & MSG_ERRQUEUE) | |
1217 | return ip_recv_error(sk, msg, len); | |
1218 | ||
1219 | try_again: | |
1220 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), | |
3f518bf7 | 1221 | &peeked, &off, &err); |
db8dac20 DM |
1222 | if (!skb) |
1223 | goto out; | |
1224 | ||
1225 | ulen = skb->len - sizeof(struct udphdr); | |
59c2cdae DM |
1226 | copied = len; |
1227 | if (copied > ulen) | |
1228 | copied = ulen; | |
1229 | else if (copied < ulen) | |
db8dac20 DM |
1230 | msg->msg_flags |= MSG_TRUNC; |
1231 | ||
1232 | /* | |
1233 | * If checksum is needed at all, try to do it while copying the | |
1234 | * data. If the data is truncated, or if we only want a partial | |
1235 | * coverage checksum (UDP-Lite), do it before the copy. | |
1236 | */ | |
1237 | ||
59c2cdae | 1238 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { |
db8dac20 DM |
1239 | if (udp_lib_checksum_complete(skb)) |
1240 | goto csum_copy_err; | |
1241 | } | |
1242 | ||
1243 | if (skb_csum_unnecessary(skb)) | |
1244 | err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), | |
59c2cdae | 1245 | msg->msg_iov, copied); |
db8dac20 | 1246 | else { |
c482c568 ED |
1247 | err = skb_copy_and_csum_datagram_iovec(skb, |
1248 | sizeof(struct udphdr), | |
1249 | msg->msg_iov); | |
db8dac20 DM |
1250 | |
1251 | if (err == -EINVAL) | |
1252 | goto csum_copy_err; | |
1253 | } | |
1254 | ||
22911fc5 ED |
1255 | if (unlikely(err)) { |
1256 | trace_kfree_skb(skb, udp_recvmsg); | |
979402b1 ED |
1257 | if (!peeked) { |
1258 | atomic_inc(&sk->sk_drops); | |
1259 | UDP_INC_STATS_USER(sock_net(sk), | |
1260 | UDP_MIB_INERRORS, is_udplite); | |
1261 | } | |
db8dac20 | 1262 | goto out_free; |
22911fc5 | 1263 | } |
db8dac20 DM |
1264 | |
1265 | if (!peeked) | |
629ca23c PE |
1266 | UDP_INC_STATS_USER(sock_net(sk), |
1267 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 | 1268 | |
3b885787 | 1269 | sock_recv_ts_and_drops(msg, sk, skb); |
db8dac20 DM |
1270 | |
1271 | /* Copy the address. */ | |
c482c568 | 1272 | if (sin) { |
db8dac20 DM |
1273 | sin->sin_family = AF_INET; |
1274 | sin->sin_port = udp_hdr(skb)->source; | |
1275 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
1276 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
1277 | } | |
1278 | if (inet->cmsg_flags) | |
1279 | ip_cmsg_recv(msg, skb); | |
1280 | ||
59c2cdae | 1281 | err = copied; |
db8dac20 DM |
1282 | if (flags & MSG_TRUNC) |
1283 | err = ulen; | |
1284 | ||
1285 | out_free: | |
9d410c79 | 1286 | skb_free_datagram_locked(sk, skb); |
db8dac20 DM |
1287 | out: |
1288 | return err; | |
1289 | ||
1290 | csum_copy_err: | |
8a74ad60 | 1291 | slow = lock_sock_fast(sk); |
6a5dc9e5 ED |
1292 | if (!skb_kill_datagram(sk, skb, flags)) { |
1293 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); | |
629ca23c | 1294 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
6a5dc9e5 | 1295 | } |
8a74ad60 | 1296 | unlock_sock_fast(sk, slow); |
db8dac20 DM |
1297 | |
1298 | if (noblock) | |
1299 | return -EAGAIN; | |
9cfaa8de XZ |
1300 | |
1301 | /* starting over for a new packet */ | |
1302 | msg->msg_flags &= ~MSG_TRUNC; | |
db8dac20 DM |
1303 | goto try_again; |
1304 | } | |
1305 | ||
1306 | ||
1da177e4 LT |
1307 | int udp_disconnect(struct sock *sk, int flags) |
1308 | { | |
1309 | struct inet_sock *inet = inet_sk(sk); | |
1310 | /* | |
1311 | * 1003.1g - break association. | |
1312 | */ | |
e905a9ed | 1313 | |
1da177e4 | 1314 | sk->sk_state = TCP_CLOSE; |
c720c7e8 ED |
1315 | inet->inet_daddr = 0; |
1316 | inet->inet_dport = 0; | |
bdeab991 | 1317 | sock_rps_reset_rxhash(sk); |
1da177e4 LT |
1318 | sk->sk_bound_dev_if = 0; |
1319 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
1320 | inet_reset_saddr(sk); | |
1321 | ||
1322 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
1323 | sk->sk_prot->unhash(sk); | |
c720c7e8 | 1324 | inet->inet_sport = 0; |
1da177e4 LT |
1325 | } |
1326 | sk_dst_reset(sk); | |
1327 | return 0; | |
1328 | } | |
c482c568 | 1329 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1330 | |
645ca708 ED |
1331 | void udp_lib_unhash(struct sock *sk) |
1332 | { | |
723b4610 ED |
1333 | if (sk_hashed(sk)) { |
1334 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
512615b6 ED |
1335 | struct udp_hslot *hslot, *hslot2; |
1336 | ||
1337 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1338 | udp_sk(sk)->udp_port_hash); | |
1339 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
645ca708 | 1340 | |
723b4610 ED |
1341 | spin_lock_bh(&hslot->lock); |
1342 | if (sk_nulls_del_node_init_rcu(sk)) { | |
fdcc8aa9 | 1343 | hslot->count--; |
c720c7e8 | 1344 | inet_sk(sk)->inet_num = 0; |
723b4610 | 1345 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); |
512615b6 ED |
1346 | |
1347 | spin_lock(&hslot2->lock); | |
1348 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1349 | hslot2->count--; | |
1350 | spin_unlock(&hslot2->lock); | |
723b4610 ED |
1351 | } |
1352 | spin_unlock_bh(&hslot->lock); | |
645ca708 | 1353 | } |
645ca708 ED |
1354 | } |
1355 | EXPORT_SYMBOL(udp_lib_unhash); | |
1356 | ||
719f8358 ED |
1357 | /* |
1358 | * inet_rcv_saddr was changed, we must rehash secondary hash | |
1359 | */ | |
1360 | void udp_lib_rehash(struct sock *sk, u16 newhash) | |
1361 | { | |
1362 | if (sk_hashed(sk)) { | |
1363 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
1364 | struct udp_hslot *hslot, *hslot2, *nhslot2; | |
1365 | ||
1366 | hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash); | |
1367 | nhslot2 = udp_hashslot2(udptable, newhash); | |
1368 | udp_sk(sk)->udp_portaddr_hash = newhash; | |
1369 | if (hslot2 != nhslot2) { | |
1370 | hslot = udp_hashslot(udptable, sock_net(sk), | |
1371 | udp_sk(sk)->udp_port_hash); | |
1372 | /* we must lock primary chain too */ | |
1373 | spin_lock_bh(&hslot->lock); | |
1374 | ||
1375 | spin_lock(&hslot2->lock); | |
1376 | hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node); | |
1377 | hslot2->count--; | |
1378 | spin_unlock(&hslot2->lock); | |
1379 | ||
1380 | spin_lock(&nhslot2->lock); | |
1381 | hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node, | |
1382 | &nhslot2->head); | |
1383 | nhslot2->count++; | |
1384 | spin_unlock(&nhslot2->lock); | |
1385 | ||
1386 | spin_unlock_bh(&hslot->lock); | |
1387 | } | |
1388 | } | |
1389 | } | |
1390 | EXPORT_SYMBOL(udp_lib_rehash); | |
1391 | ||
1392 | static void udp_v4_rehash(struct sock *sk) | |
1393 | { | |
1394 | u16 new_hash = udp4_portaddr_hash(sock_net(sk), | |
1395 | inet_sk(sk)->inet_rcv_saddr, | |
1396 | inet_sk(sk)->inet_num); | |
1397 | udp_lib_rehash(sk, new_hash); | |
1398 | } | |
1399 | ||
93821778 HX |
1400 | static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
1401 | { | |
fec5e652 | 1402 | int rc; |
766e9037 | 1403 | |
fec5e652 | 1404 | if (inet_sk(sk)->inet_daddr) |
bdeab991 | 1405 | sock_rps_save_rxhash(sk, skb); |
fec5e652 | 1406 | |
d826eb14 | 1407 | rc = sock_queue_rcv_skb(sk, skb); |
766e9037 ED |
1408 | if (rc < 0) { |
1409 | int is_udplite = IS_UDPLITE(sk); | |
93821778 | 1410 | |
93821778 | 1411 | /* Note that an ENOMEM error is charged twice */ |
766e9037 | 1412 | if (rc == -ENOMEM) |
93821778 HX |
1413 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1414 | is_udplite); | |
766e9037 ED |
1415 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
1416 | kfree_skb(skb); | |
296f7ea7 | 1417 | trace_udp_fail_queue_rcv_skb(rc, sk); |
766e9037 | 1418 | return -1; |
93821778 HX |
1419 | } |
1420 | ||
1421 | return 0; | |
1422 | ||
93821778 HX |
1423 | } |
1424 | ||
447167bf ED |
1425 | static struct static_key udp_encap_needed __read_mostly; |
1426 | void udp_encap_enable(void) | |
1427 | { | |
1428 | if (!static_key_enabled(&udp_encap_needed)) | |
1429 | static_key_slow_inc(&udp_encap_needed); | |
1430 | } | |
1431 | EXPORT_SYMBOL(udp_encap_enable); | |
1432 | ||
db8dac20 DM |
1433 | /* returns: |
1434 | * -1: error | |
1435 | * 0: success | |
1436 | * >0: "udp encap" protocol resubmission | |
1437 | * | |
1438 | * Note that in the success and error cases, the skb is assumed to | |
1439 | * have either been requeued or freed. | |
1440 | */ | |
c482c568 | 1441 | int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
db8dac20 DM |
1442 | { |
1443 | struct udp_sock *up = udp_sk(sk); | |
1444 | int rc; | |
1445 | int is_udplite = IS_UDPLITE(sk); | |
1446 | ||
1447 | /* | |
1448 | * Charge it to the socket, dropping if the queue is full. | |
1449 | */ | |
1450 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1451 | goto drop; | |
1452 | nf_reset(skb); | |
1453 | ||
447167bf | 1454 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
0ad92ad0 ED |
1455 | int (*encap_rcv)(struct sock *sk, struct sk_buff *skb); |
1456 | ||
db8dac20 DM |
1457 | /* |
1458 | * This is an encapsulation socket so pass the skb to | |
1459 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1460 | * fall through and pass this up the UDP socket. | |
1461 | * up->encap_rcv() returns the following value: | |
1462 | * =0 if skb was successfully passed to the encap | |
1463 | * handler or was discarded by it. | |
1464 | * >0 if skb should be passed on to UDP. | |
1465 | * <0 if skb should be resubmitted as proto -N | |
1466 | */ | |
1467 | ||
1468 | /* if we're overly short, let UDP handle it */ | |
0ad92ad0 ED |
1469 | encap_rcv = ACCESS_ONCE(up->encap_rcv); |
1470 | if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) { | |
db8dac20 DM |
1471 | int ret; |
1472 | ||
0ad92ad0 | 1473 | ret = encap_rcv(sk, skb); |
db8dac20 | 1474 | if (ret <= 0) { |
0283328e PE |
1475 | UDP_INC_STATS_BH(sock_net(sk), |
1476 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
1477 | is_udplite); |
1478 | return -ret; | |
1479 | } | |
1480 | } | |
1481 | ||
1482 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1483 | } | |
1484 | ||
1485 | /* | |
1486 | * UDP-Lite specific tests, ignored on UDP sockets | |
1487 | */ | |
1488 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1489 | ||
1490 | /* | |
1491 | * MIB statistics other than incrementing the error count are | |
1492 | * disabled for the following two types of errors: these depend | |
1493 | * on the application settings, not on the functioning of the | |
1494 | * protocol stack as such. | |
1495 | * | |
1496 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1497 | * way ... to ... at least let the receiving application block | |
1498 | * delivery of packets with coverage values less than a value | |
1499 | * provided by the application." | |
1500 | */ | |
1501 | if (up->pcrlen == 0) { /* full coverage was set */ | |
afd46503 JP |
1502 | LIMIT_NETDEBUG(KERN_WARNING "UDPLite: partial coverage %d while full coverage %d requested\n", |
1503 | UDP_SKB_CB(skb)->cscov, skb->len); | |
db8dac20 DM |
1504 | goto drop; |
1505 | } | |
1506 | /* The next case involves violating the min. coverage requested | |
1507 | * by the receiver. This is subtle: if receiver wants x and x is | |
1508 | * greater than the buffersize/MTU then receiver will complain | |
1509 | * that it wants x while sender emits packets of smaller size y. | |
1510 | * Therefore the above ...()->partial_cov statement is essential. | |
1511 | */ | |
1512 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
afd46503 JP |
1513 | LIMIT_NETDEBUG(KERN_WARNING "UDPLite: coverage %d too small, need min %d\n", |
1514 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
db8dac20 DM |
1515 | goto drop; |
1516 | } | |
1517 | } | |
1518 | ||
33d480ce ED |
1519 | if (rcu_access_pointer(sk->sk_filter) && |
1520 | udp_lib_checksum_complete(skb)) | |
6a5dc9e5 | 1521 | goto csum_error; |
db8dac20 | 1522 | |
c377411f | 1523 | |
f545a38f | 1524 | if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf)) |
c377411f ED |
1525 | goto drop; |
1526 | ||
93821778 | 1527 | rc = 0; |
db8dac20 | 1528 | |
d826eb14 | 1529 | ipv4_pktinfo_prepare(skb); |
93821778 HX |
1530 | bh_lock_sock(sk); |
1531 | if (!sock_owned_by_user(sk)) | |
1532 | rc = __udp_queue_rcv_skb(sk, skb); | |
f545a38f | 1533 | else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) { |
55349790 ZY |
1534 | bh_unlock_sock(sk); |
1535 | goto drop; | |
1536 | } | |
93821778 HX |
1537 | bh_unlock_sock(sk); |
1538 | ||
1539 | return rc; | |
db8dac20 | 1540 | |
6a5dc9e5 ED |
1541 | csum_error: |
1542 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite); | |
db8dac20 | 1543 | drop: |
0283328e | 1544 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
8edf19c2 | 1545 | atomic_inc(&sk->sk_drops); |
db8dac20 DM |
1546 | kfree_skb(skb); |
1547 | return -1; | |
1548 | } | |
1549 | ||
1240d137 ED |
1550 | |
1551 | static void flush_stack(struct sock **stack, unsigned int count, | |
1552 | struct sk_buff *skb, unsigned int final) | |
1553 | { | |
1554 | unsigned int i; | |
1555 | struct sk_buff *skb1 = NULL; | |
f6b8f32c | 1556 | struct sock *sk; |
1240d137 ED |
1557 | |
1558 | for (i = 0; i < count; i++) { | |
f6b8f32c | 1559 | sk = stack[i]; |
1240d137 ED |
1560 | if (likely(skb1 == NULL)) |
1561 | skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC); | |
1562 | ||
f6b8f32c ED |
1563 | if (!skb1) { |
1564 | atomic_inc(&sk->sk_drops); | |
1565 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, | |
1566 | IS_UDPLITE(sk)); | |
1567 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, | |
1568 | IS_UDPLITE(sk)); | |
1569 | } | |
1570 | ||
1571 | if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0) | |
1240d137 ED |
1572 | skb1 = NULL; |
1573 | } | |
1574 | if (unlikely(skb1)) | |
1575 | kfree_skb(skb1); | |
1576 | } | |
1577 | ||
db8dac20 DM |
1578 | /* |
1579 | * Multicasts and broadcasts go to each listener. | |
1580 | * | |
1240d137 | 1581 | * Note: called only from the BH handler context. |
db8dac20 | 1582 | */ |
e3163493 | 1583 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1584 | struct udphdr *uh, |
1585 | __be32 saddr, __be32 daddr, | |
645ca708 | 1586 | struct udp_table *udptable) |
db8dac20 | 1587 | { |
1240d137 | 1588 | struct sock *sk, *stack[256 / sizeof(struct sock *)]; |
f86dcc5a | 1589 | struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest)); |
db8dac20 | 1590 | int dif; |
1240d137 | 1591 | unsigned int i, count = 0; |
db8dac20 | 1592 | |
645ca708 | 1593 | spin_lock(&hslot->lock); |
88ab1932 | 1594 | sk = sk_nulls_head(&hslot->head); |
db8dac20 | 1595 | dif = skb->dev->ifindex; |
920a4611 | 1596 | sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif); |
1240d137 ED |
1597 | while (sk) { |
1598 | stack[count++] = sk; | |
1599 | sk = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest, | |
1600 | daddr, uh->source, saddr, dif); | |
1601 | if (unlikely(count == ARRAY_SIZE(stack))) { | |
1602 | if (!sk) | |
1603 | break; | |
1604 | flush_stack(stack, count, skb, ~0); | |
1605 | count = 0; | |
1606 | } | |
1607 | } | |
1608 | /* | |
1609 | * before releasing chain lock, we must take a reference on sockets | |
1610 | */ | |
1611 | for (i = 0; i < count; i++) | |
1612 | sock_hold(stack[i]); | |
1613 | ||
645ca708 | 1614 | spin_unlock(&hslot->lock); |
1240d137 ED |
1615 | |
1616 | /* | |
1617 | * do the slow work with no lock held | |
1618 | */ | |
1619 | if (count) { | |
1620 | flush_stack(stack, count, skb, count - 1); | |
1621 | ||
1622 | for (i = 0; i < count; i++) | |
1623 | sock_put(stack[i]); | |
1624 | } else { | |
1625 | kfree_skb(skb); | |
1626 | } | |
db8dac20 DM |
1627 | return 0; |
1628 | } | |
1629 | ||
1630 | /* Initialize UDP checksum. If exited with zero value (success), | |
1631 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1632 | * Otherwise, csum completion requires chacksumming packet body, | |
1633 | * including udp header and folding it to skb->csum. | |
1634 | */ | |
1635 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1636 | int proto) | |
1637 | { | |
1638 | const struct iphdr *iph; | |
1639 | int err; | |
1640 | ||
1641 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1642 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1643 | ||
1644 | if (proto == IPPROTO_UDPLITE) { | |
1645 | err = udplite_checksum_init(skb, uh); | |
1646 | if (err) | |
1647 | return err; | |
1648 | } | |
1649 | ||
1650 | iph = ip_hdr(skb); | |
1651 | if (uh->check == 0) { | |
1652 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1653 | } else if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
c482c568 | 1654 | if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, |
db8dac20 DM |
1655 | proto, skb->csum)) |
1656 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1657 | } | |
1658 | if (!skb_csum_unnecessary(skb)) | |
1659 | skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, | |
1660 | skb->len, proto, 0); | |
1661 | /* Probably, we should checksum udp header (it should be in cache | |
1662 | * in any case) and data in tiny packets (< rx copybreak). | |
1663 | */ | |
1664 | ||
1665 | return 0; | |
1666 | } | |
1667 | ||
1668 | /* | |
1669 | * All we need to do is get the socket, and then do a checksum. | |
1670 | */ | |
1671 | ||
645ca708 | 1672 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1673 | int proto) |
1674 | { | |
1675 | struct sock *sk; | |
7b5e56f9 | 1676 | struct udphdr *uh; |
db8dac20 | 1677 | unsigned short ulen; |
adf30907 | 1678 | struct rtable *rt = skb_rtable(skb); |
2783ef23 | 1679 | __be32 saddr, daddr; |
0283328e | 1680 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1681 | |
1682 | /* | |
1683 | * Validate the packet. | |
1684 | */ | |
1685 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1686 | goto drop; /* No space for header. */ | |
1687 | ||
7b5e56f9 | 1688 | uh = udp_hdr(skb); |
db8dac20 | 1689 | ulen = ntohs(uh->len); |
ccc2d97c BM |
1690 | saddr = ip_hdr(skb)->saddr; |
1691 | daddr = ip_hdr(skb)->daddr; | |
1692 | ||
db8dac20 DM |
1693 | if (ulen > skb->len) |
1694 | goto short_packet; | |
1695 | ||
1696 | if (proto == IPPROTO_UDP) { | |
1697 | /* UDP validates ulen. */ | |
1698 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1699 | goto short_packet; | |
1700 | uh = udp_hdr(skb); | |
1701 | } | |
1702 | ||
1703 | if (udp4_csum_init(skb, uh, proto)) | |
1704 | goto csum_error; | |
1705 | ||
1706 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) | |
e3163493 PE |
1707 | return __udp4_lib_mcast_deliver(net, skb, uh, |
1708 | saddr, daddr, udptable); | |
db8dac20 | 1709 | |
607c4aaf | 1710 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); |
db8dac20 DM |
1711 | |
1712 | if (sk != NULL) { | |
a5b50476 ET |
1713 | int ret; |
1714 | ||
1715 | sk_mark_ll(sk, skb); | |
1716 | ret = udp_queue_rcv_skb(sk, skb); | |
db8dac20 DM |
1717 | sock_put(sk); |
1718 | ||
1719 | /* a return value > 0 means to resubmit the input, but | |
1720 | * it wants the return to be -protocol, or 0 | |
1721 | */ | |
1722 | if (ret > 0) | |
1723 | return -ret; | |
1724 | return 0; | |
1725 | } | |
1726 | ||
1727 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1728 | goto drop; | |
1729 | nf_reset(skb); | |
1730 | ||
1731 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1732 | if (udp_lib_checksum_complete(skb)) | |
1733 | goto csum_error; | |
1734 | ||
0283328e | 1735 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1736 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1737 | ||
1738 | /* | |
1739 | * Hmm. We got an UDP packet to a port to which we | |
1740 | * don't wanna listen. Ignore it. | |
1741 | */ | |
1742 | kfree_skb(skb); | |
1743 | return 0; | |
1744 | ||
1745 | short_packet: | |
673d57e7 | 1746 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n", |
afd46503 JP |
1747 | proto == IPPROTO_UDPLITE ? "Lite" : "", |
1748 | &saddr, ntohs(uh->source), | |
1749 | ulen, skb->len, | |
1750 | &daddr, ntohs(uh->dest)); | |
db8dac20 DM |
1751 | goto drop; |
1752 | ||
1753 | csum_error: | |
1754 | /* | |
1755 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1756 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1757 | */ | |
673d57e7 | 1758 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n", |
afd46503 JP |
1759 | proto == IPPROTO_UDPLITE ? "Lite" : "", |
1760 | &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest), | |
db8dac20 | 1761 | ulen); |
6a5dc9e5 | 1762 | UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 | 1763 | drop: |
0283328e | 1764 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1765 | kfree_skb(skb); |
1766 | return 0; | |
1767 | } | |
1768 | ||
1769 | int udp_rcv(struct sk_buff *skb) | |
1770 | { | |
645ca708 | 1771 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
1772 | } |
1773 | ||
7d06b2e0 | 1774 | void udp_destroy_sock(struct sock *sk) |
db8dac20 | 1775 | { |
44046a59 | 1776 | struct udp_sock *up = udp_sk(sk); |
8a74ad60 | 1777 | bool slow = lock_sock_fast(sk); |
db8dac20 | 1778 | udp_flush_pending_frames(sk); |
8a74ad60 | 1779 | unlock_sock_fast(sk, slow); |
44046a59 TP |
1780 | if (static_key_false(&udp_encap_needed) && up->encap_type) { |
1781 | void (*encap_destroy)(struct sock *sk); | |
1782 | encap_destroy = ACCESS_ONCE(up->encap_destroy); | |
1783 | if (encap_destroy) | |
1784 | encap_destroy(sk); | |
1785 | } | |
db8dac20 DM |
1786 | } |
1787 | ||
1da177e4 LT |
1788 | /* |
1789 | * Socket option code for UDP | |
1790 | */ | |
4c0a6cb0 | 1791 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 1792 | char __user *optval, unsigned int optlen, |
4c0a6cb0 | 1793 | int (*push_pending_frames)(struct sock *)) |
1da177e4 LT |
1794 | { |
1795 | struct udp_sock *up = udp_sk(sk); | |
1796 | int val; | |
1797 | int err = 0; | |
b2bf1e26 | 1798 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 1799 | |
c482c568 | 1800 | if (optlen < sizeof(int)) |
1da177e4 LT |
1801 | return -EINVAL; |
1802 | ||
1803 | if (get_user(val, (int __user *)optval)) | |
1804 | return -EFAULT; | |
1805 | ||
6516c655 | 1806 | switch (optname) { |
1da177e4 LT |
1807 | case UDP_CORK: |
1808 | if (val != 0) { | |
1809 | up->corkflag = 1; | |
1810 | } else { | |
1811 | up->corkflag = 0; | |
1812 | lock_sock(sk); | |
4c0a6cb0 | 1813 | (*push_pending_frames)(sk); |
1da177e4 LT |
1814 | release_sock(sk); |
1815 | } | |
1816 | break; | |
e905a9ed | 1817 | |
1da177e4 LT |
1818 | case UDP_ENCAP: |
1819 | switch (val) { | |
1820 | case 0: | |
1821 | case UDP_ENCAP_ESPINUDP: | |
1822 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
1823 | up->encap_rcv = xfrm4_udp_encap_rcv; |
1824 | /* FALLTHROUGH */ | |
342f0234 | 1825 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 | 1826 | up->encap_type = val; |
447167bf | 1827 | udp_encap_enable(); |
1da177e4 LT |
1828 | break; |
1829 | default: | |
1830 | err = -ENOPROTOOPT; | |
1831 | break; | |
1832 | } | |
1833 | break; | |
1834 | ||
ba4e58ec GR |
1835 | /* |
1836 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
1837 | */ | |
1838 | /* The sender sets actual checksum coverage length via this option. | |
1839 | * The case coverage > packet length is handled by send module. */ | |
1840 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 1841 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1842 | return -ENOPROTOOPT; |
1843 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
1844 | val = 8; | |
4be929be AD |
1845 | else if (val > USHRT_MAX) |
1846 | val = USHRT_MAX; | |
ba4e58ec GR |
1847 | up->pcslen = val; |
1848 | up->pcflag |= UDPLITE_SEND_CC; | |
1849 | break; | |
1850 | ||
e905a9ed YH |
1851 | /* The receiver specifies a minimum checksum coverage value. To make |
1852 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
1853 | * used, this again means full checksum coverage. */ |
1854 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 1855 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1856 | return -ENOPROTOOPT; |
1857 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
1858 | val = 8; | |
4be929be AD |
1859 | else if (val > USHRT_MAX) |
1860 | val = USHRT_MAX; | |
ba4e58ec GR |
1861 | up->pcrlen = val; |
1862 | up->pcflag |= UDPLITE_RECV_CC; | |
1863 | break; | |
1864 | ||
1da177e4 LT |
1865 | default: |
1866 | err = -ENOPROTOOPT; | |
1867 | break; | |
6516c655 | 1868 | } |
1da177e4 LT |
1869 | |
1870 | return err; | |
1871 | } | |
c482c568 | 1872 | EXPORT_SYMBOL(udp_lib_setsockopt); |
1da177e4 | 1873 | |
db8dac20 | 1874 | int udp_setsockopt(struct sock *sk, int level, int optname, |
b7058842 | 1875 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
1876 | { |
1877 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1878 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1879 | udp_push_pending_frames); | |
1880 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
1881 | } | |
1882 | ||
1883 | #ifdef CONFIG_COMPAT | |
1884 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
b7058842 | 1885 | char __user *optval, unsigned int optlen) |
db8dac20 DM |
1886 | { |
1887 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1888 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1889 | udp_push_pending_frames); | |
1890 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
1891 | } | |
1892 | #endif | |
1893 | ||
4c0a6cb0 GR |
1894 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
1895 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
1896 | { |
1897 | struct udp_sock *up = udp_sk(sk); | |
1898 | int val, len; | |
1899 | ||
c482c568 | 1900 | if (get_user(len, optlen)) |
1da177e4 LT |
1901 | return -EFAULT; |
1902 | ||
1903 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 1904 | |
6516c655 | 1905 | if (len < 0) |
1da177e4 LT |
1906 | return -EINVAL; |
1907 | ||
6516c655 | 1908 | switch (optname) { |
1da177e4 LT |
1909 | case UDP_CORK: |
1910 | val = up->corkflag; | |
1911 | break; | |
1912 | ||
1913 | case UDP_ENCAP: | |
1914 | val = up->encap_type; | |
1915 | break; | |
1916 | ||
ba4e58ec GR |
1917 | /* The following two cannot be changed on UDP sockets, the return is |
1918 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
1919 | case UDPLITE_SEND_CSCOV: | |
1920 | val = up->pcslen; | |
1921 | break; | |
1922 | ||
1923 | case UDPLITE_RECV_CSCOV: | |
1924 | val = up->pcrlen; | |
1925 | break; | |
1926 | ||
1da177e4 LT |
1927 | default: |
1928 | return -ENOPROTOOPT; | |
6516c655 | 1929 | } |
1da177e4 | 1930 | |
6516c655 | 1931 | if (put_user(len, optlen)) |
e905a9ed | 1932 | return -EFAULT; |
c482c568 | 1933 | if (copy_to_user(optval, &val, len)) |
1da177e4 | 1934 | return -EFAULT; |
e905a9ed | 1935 | return 0; |
1da177e4 | 1936 | } |
c482c568 | 1937 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1da177e4 | 1938 | |
db8dac20 DM |
1939 | int udp_getsockopt(struct sock *sk, int level, int optname, |
1940 | char __user *optval, int __user *optlen) | |
1941 | { | |
1942 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1943 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1944 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
1945 | } | |
1946 | ||
1947 | #ifdef CONFIG_COMPAT | |
1948 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
1949 | char __user *optval, int __user *optlen) | |
1950 | { | |
1951 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1952 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1953 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
1954 | } | |
1955 | #endif | |
1da177e4 LT |
1956 | /** |
1957 | * udp_poll - wait for a UDP event. | |
1958 | * @file - file struct | |
1959 | * @sock - socket | |
1960 | * @wait - poll table | |
1961 | * | |
e905a9ed | 1962 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
1963 | * blocking sockets. If application is using a blocking fd |
1964 | * and a packet with checksum error is in the queue; | |
1965 | * then it could get return from select indicating data available | |
1966 | * but then block when reading it. Add special case code | |
1967 | * to work around these arguably broken applications. | |
1968 | */ | |
1969 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
1970 | { | |
1971 | unsigned int mask = datagram_poll(file, sock, wait); | |
1972 | struct sock *sk = sock->sk; | |
ba4e58ec | 1973 | |
c3f1dbaf DM |
1974 | sock_rps_record_flow(sk); |
1975 | ||
1da177e4 | 1976 | /* Check for false positives due to checksum errors */ |
85584672 ED |
1977 | if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) && |
1978 | !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk)) | |
1979 | mask &= ~(POLLIN | POLLRDNORM); | |
1da177e4 LT |
1980 | |
1981 | return mask; | |
e905a9ed | 1982 | |
1da177e4 | 1983 | } |
c482c568 | 1984 | EXPORT_SYMBOL(udp_poll); |
1da177e4 | 1985 | |
db8dac20 DM |
1986 | struct proto udp_prot = { |
1987 | .name = "UDP", | |
1988 | .owner = THIS_MODULE, | |
1989 | .close = udp_lib_close, | |
1990 | .connect = ip4_datagram_connect, | |
1991 | .disconnect = udp_disconnect, | |
1992 | .ioctl = udp_ioctl, | |
1993 | .destroy = udp_destroy_sock, | |
1994 | .setsockopt = udp_setsockopt, | |
1995 | .getsockopt = udp_getsockopt, | |
1996 | .sendmsg = udp_sendmsg, | |
1997 | .recvmsg = udp_recvmsg, | |
1998 | .sendpage = udp_sendpage, | |
93821778 | 1999 | .backlog_rcv = __udp_queue_rcv_skb, |
8141ed9f | 2000 | .release_cb = ip4_datagram_release_cb, |
db8dac20 DM |
2001 | .hash = udp_lib_hash, |
2002 | .unhash = udp_lib_unhash, | |
719f8358 | 2003 | .rehash = udp_v4_rehash, |
db8dac20 DM |
2004 | .get_port = udp_v4_get_port, |
2005 | .memory_allocated = &udp_memory_allocated, | |
2006 | .sysctl_mem = sysctl_udp_mem, | |
2007 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
2008 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
2009 | .obj_size = sizeof(struct udp_sock), | |
271b72c7 | 2010 | .slab_flags = SLAB_DESTROY_BY_RCU, |
645ca708 | 2011 | .h.udp_table = &udp_table, |
db8dac20 DM |
2012 | #ifdef CONFIG_COMPAT |
2013 | .compat_setsockopt = compat_udp_setsockopt, | |
2014 | .compat_getsockopt = compat_udp_getsockopt, | |
2015 | #endif | |
fcbdf09d | 2016 | .clear_sk = sk_prot_clear_portaddr_nulls, |
db8dac20 | 2017 | }; |
c482c568 | 2018 | EXPORT_SYMBOL(udp_prot); |
1da177e4 LT |
2019 | |
2020 | /* ------------------------------------------------------------------------ */ | |
2021 | #ifdef CONFIG_PROC_FS | |
2022 | ||
645ca708 | 2023 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
2024 | { |
2025 | struct sock *sk; | |
2026 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2027 | struct net *net = seq_file_net(seq); |
1da177e4 | 2028 | |
f86dcc5a ED |
2029 | for (state->bucket = start; state->bucket <= state->udp_table->mask; |
2030 | ++state->bucket) { | |
88ab1932 | 2031 | struct hlist_nulls_node *node; |
645ca708 | 2032 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
f86dcc5a ED |
2033 | |
2034 | if (hlist_nulls_empty(&hslot->head)) | |
2035 | continue; | |
2036 | ||
645ca708 | 2037 | spin_lock_bh(&hslot->lock); |
88ab1932 | 2038 | sk_nulls_for_each(sk, node, &hslot->head) { |
878628fb | 2039 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 2040 | continue; |
1da177e4 LT |
2041 | if (sk->sk_family == state->family) |
2042 | goto found; | |
2043 | } | |
645ca708 | 2044 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
2045 | } |
2046 | sk = NULL; | |
2047 | found: | |
2048 | return sk; | |
2049 | } | |
2050 | ||
2051 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
2052 | { | |
2053 | struct udp_iter_state *state = seq->private; | |
6f191efe | 2054 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
2055 | |
2056 | do { | |
88ab1932 | 2057 | sk = sk_nulls_next(sk); |
878628fb | 2058 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 2059 | |
645ca708 | 2060 | if (!sk) { |
f86dcc5a | 2061 | if (state->bucket <= state->udp_table->mask) |
30842f29 | 2062 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
645ca708 | 2063 | return udp_get_first(seq, state->bucket + 1); |
1da177e4 LT |
2064 | } |
2065 | return sk; | |
2066 | } | |
2067 | ||
2068 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
2069 | { | |
645ca708 | 2070 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
2071 | |
2072 | if (sk) | |
6516c655 | 2073 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
2074 | --pos; |
2075 | return pos ? NULL : sk; | |
2076 | } | |
2077 | ||
2078 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
2079 | { | |
30842f29 | 2080 | struct udp_iter_state *state = seq->private; |
f86dcc5a | 2081 | state->bucket = MAX_UDP_PORTS; |
30842f29 | 2082 | |
b50660f1 | 2083 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
2084 | } |
2085 | ||
2086 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2087 | { | |
2088 | struct sock *sk; | |
2089 | ||
b50660f1 | 2090 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
2091 | sk = udp_get_idx(seq, 0); |
2092 | else | |
2093 | sk = udp_get_next(seq, v); | |
2094 | ||
2095 | ++*pos; | |
2096 | return sk; | |
2097 | } | |
2098 | ||
2099 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
2100 | { | |
645ca708 ED |
2101 | struct udp_iter_state *state = seq->private; |
2102 | ||
f86dcc5a | 2103 | if (state->bucket <= state->udp_table->mask) |
645ca708 | 2104 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
1da177e4 LT |
2105 | } |
2106 | ||
73cb88ec | 2107 | int udp_seq_open(struct inode *inode, struct file *file) |
1da177e4 | 2108 | { |
d9dda78b | 2109 | struct udp_seq_afinfo *afinfo = PDE_DATA(inode); |
a2be75c1 DL |
2110 | struct udp_iter_state *s; |
2111 | int err; | |
a91275ef | 2112 | |
a2be75c1 DL |
2113 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
2114 | sizeof(struct udp_iter_state)); | |
2115 | if (err < 0) | |
2116 | return err; | |
a91275ef | 2117 | |
a2be75c1 | 2118 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 2119 | s->family = afinfo->family; |
645ca708 | 2120 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 2121 | return err; |
a91275ef | 2122 | } |
73cb88ec | 2123 | EXPORT_SYMBOL(udp_seq_open); |
a91275ef | 2124 | |
1da177e4 | 2125 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 2126 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
2127 | { |
2128 | struct proc_dir_entry *p; | |
2129 | int rc = 0; | |
2130 | ||
dda61925 DL |
2131 | afinfo->seq_ops.start = udp_seq_start; |
2132 | afinfo->seq_ops.next = udp_seq_next; | |
2133 | afinfo->seq_ops.stop = udp_seq_stop; | |
2134 | ||
84841c3c | 2135 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
73cb88ec | 2136 | afinfo->seq_fops, afinfo); |
84841c3c | 2137 | if (!p) |
1da177e4 LT |
2138 | rc = -ENOMEM; |
2139 | return rc; | |
2140 | } | |
c482c568 | 2141 | EXPORT_SYMBOL(udp_proc_register); |
1da177e4 | 2142 | |
0c96d8c5 | 2143 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 2144 | { |
ece31ffd | 2145 | remove_proc_entry(afinfo->name, net->proc_net); |
1da177e4 | 2146 | } |
c482c568 | 2147 | EXPORT_SYMBOL(udp_proc_unregister); |
db8dac20 DM |
2148 | |
2149 | /* ------------------------------------------------------------------------ */ | |
5e659e4c PE |
2150 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
2151 | int bucket, int *len) | |
db8dac20 DM |
2152 | { |
2153 | struct inet_sock *inet = inet_sk(sp); | |
c720c7e8 ED |
2154 | __be32 dest = inet->inet_daddr; |
2155 | __be32 src = inet->inet_rcv_saddr; | |
2156 | __u16 destp = ntohs(inet->inet_dport); | |
2157 | __u16 srcp = ntohs(inet->inet_sport); | |
db8dac20 | 2158 | |
f86dcc5a | 2159 | seq_printf(f, "%5d: %08X:%04X %08X:%04X" |
71338aa7 | 2160 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n", |
db8dac20 | 2161 | bucket, src, srcp, dest, destp, sp->sk_state, |
31e6d363 ED |
2162 | sk_wmem_alloc_get(sp), |
2163 | sk_rmem_alloc_get(sp), | |
a7cb5a49 EB |
2164 | 0, 0L, 0, |
2165 | from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)), | |
2166 | 0, sock_i_ino(sp), | |
cb61cb9b ED |
2167 | atomic_read(&sp->sk_refcnt), sp, |
2168 | atomic_read(&sp->sk_drops), len); | |
db8dac20 DM |
2169 | } |
2170 | ||
2171 | int udp4_seq_show(struct seq_file *seq, void *v) | |
2172 | { | |
2173 | if (v == SEQ_START_TOKEN) | |
2174 | seq_printf(seq, "%-127s\n", | |
2175 | " sl local_address rem_address st tx_queue " | |
2176 | "rx_queue tr tm->when retrnsmt uid timeout " | |
cb61cb9b | 2177 | "inode ref pointer drops"); |
db8dac20 | 2178 | else { |
db8dac20 | 2179 | struct udp_iter_state *state = seq->private; |
5e659e4c | 2180 | int len; |
db8dac20 | 2181 | |
5e659e4c | 2182 | udp4_format_sock(v, seq, state->bucket, &len); |
c482c568 | 2183 | seq_printf(seq, "%*s\n", 127 - len, ""); |
db8dac20 DM |
2184 | } |
2185 | return 0; | |
2186 | } | |
2187 | ||
73cb88ec AV |
2188 | static const struct file_operations udp_afinfo_seq_fops = { |
2189 | .owner = THIS_MODULE, | |
2190 | .open = udp_seq_open, | |
2191 | .read = seq_read, | |
2192 | .llseek = seq_lseek, | |
2193 | .release = seq_release_net | |
2194 | }; | |
2195 | ||
db8dac20 | 2196 | /* ------------------------------------------------------------------------ */ |
db8dac20 | 2197 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
2198 | .name = "udp", |
2199 | .family = AF_INET, | |
645ca708 | 2200 | .udp_table = &udp_table, |
73cb88ec | 2201 | .seq_fops = &udp_afinfo_seq_fops, |
dda61925 DL |
2202 | .seq_ops = { |
2203 | .show = udp4_seq_show, | |
2204 | }, | |
db8dac20 DM |
2205 | }; |
2206 | ||
2c8c1e72 | 2207 | static int __net_init udp4_proc_init_net(struct net *net) |
15439feb PE |
2208 | { |
2209 | return udp_proc_register(net, &udp4_seq_afinfo); | |
2210 | } | |
2211 | ||
2c8c1e72 | 2212 | static void __net_exit udp4_proc_exit_net(struct net *net) |
15439feb PE |
2213 | { |
2214 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
2215 | } | |
2216 | ||
2217 | static struct pernet_operations udp4_net_ops = { | |
2218 | .init = udp4_proc_init_net, | |
2219 | .exit = udp4_proc_exit_net, | |
2220 | }; | |
2221 | ||
db8dac20 DM |
2222 | int __init udp4_proc_init(void) |
2223 | { | |
15439feb | 2224 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
2225 | } |
2226 | ||
2227 | void udp4_proc_exit(void) | |
2228 | { | |
15439feb | 2229 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 2230 | } |
1da177e4 LT |
2231 | #endif /* CONFIG_PROC_FS */ |
2232 | ||
f86dcc5a ED |
2233 | static __initdata unsigned long uhash_entries; |
2234 | static int __init set_uhash_entries(char *str) | |
645ca708 | 2235 | { |
413c27d8 EZ |
2236 | ssize_t ret; |
2237 | ||
f86dcc5a ED |
2238 | if (!str) |
2239 | return 0; | |
413c27d8 EZ |
2240 | |
2241 | ret = kstrtoul(str, 0, &uhash_entries); | |
2242 | if (ret) | |
2243 | return 0; | |
2244 | ||
f86dcc5a ED |
2245 | if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN) |
2246 | uhash_entries = UDP_HTABLE_SIZE_MIN; | |
2247 | return 1; | |
2248 | } | |
2249 | __setup("uhash_entries=", set_uhash_entries); | |
645ca708 | 2250 | |
f86dcc5a ED |
2251 | void __init udp_table_init(struct udp_table *table, const char *name) |
2252 | { | |
2253 | unsigned int i; | |
2254 | ||
31fe62b9 TB |
2255 | table->hash = alloc_large_system_hash(name, |
2256 | 2 * sizeof(struct udp_hslot), | |
2257 | uhash_entries, | |
2258 | 21, /* one slot per 2 MB */ | |
2259 | 0, | |
2260 | &table->log, | |
2261 | &table->mask, | |
2262 | UDP_HTABLE_SIZE_MIN, | |
2263 | 64 * 1024); | |
2264 | ||
512615b6 | 2265 | table->hash2 = table->hash + (table->mask + 1); |
f86dcc5a | 2266 | for (i = 0; i <= table->mask; i++) { |
88ab1932 | 2267 | INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i); |
fdcc8aa9 | 2268 | table->hash[i].count = 0; |
645ca708 ED |
2269 | spin_lock_init(&table->hash[i].lock); |
2270 | } | |
512615b6 ED |
2271 | for (i = 0; i <= table->mask; i++) { |
2272 | INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i); | |
2273 | table->hash2[i].count = 0; | |
2274 | spin_lock_init(&table->hash2[i].lock); | |
2275 | } | |
645ca708 ED |
2276 | } |
2277 | ||
95766fff HA |
2278 | void __init udp_init(void) |
2279 | { | |
f03d78db | 2280 | unsigned long limit; |
95766fff | 2281 | |
f86dcc5a | 2282 | udp_table_init(&udp_table, "UDP"); |
f03d78db | 2283 | limit = nr_free_buffer_pages() / 8; |
95766fff HA |
2284 | limit = max(limit, 128UL); |
2285 | sysctl_udp_mem[0] = limit / 4 * 3; | |
2286 | sysctl_udp_mem[1] = limit; | |
2287 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
2288 | ||
2289 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
2290 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
2291 | } | |
2292 | ||
da5bab07 DB |
2293 | struct sk_buff *skb_udp_tunnel_segment(struct sk_buff *skb, |
2294 | netdev_features_t features) | |
73136267 PS |
2295 | { |
2296 | struct sk_buff *segs = ERR_PTR(-EINVAL); | |
2297 | int mac_len = skb->mac_len; | |
2298 | int tnl_hlen = skb_inner_mac_header(skb) - skb_transport_header(skb); | |
0d05535d | 2299 | __be16 protocol = skb->protocol; |
73136267 | 2300 | netdev_features_t enc_features; |
0d05535d | 2301 | int outer_hlen; |
73136267 PS |
2302 | |
2303 | if (unlikely(!pskb_may_pull(skb, tnl_hlen))) | |
2304 | goto out; | |
2305 | ||
2306 | skb->encapsulation = 0; | |
2307 | __skb_pull(skb, tnl_hlen); | |
2308 | skb_reset_mac_header(skb); | |
2309 | skb_set_network_header(skb, skb_inner_network_offset(skb)); | |
2310 | skb->mac_len = skb_inner_network_offset(skb); | |
19acc327 | 2311 | skb->protocol = htons(ETH_P_TEB); |
73136267 PS |
2312 | |
2313 | /* segment inner packet. */ | |
2314 | enc_features = skb->dev->hw_enc_features & netif_skb_features(skb); | |
2315 | segs = skb_mac_gso_segment(skb, enc_features); | |
2316 | if (!segs || IS_ERR(segs)) | |
2317 | goto out; | |
2318 | ||
2319 | outer_hlen = skb_tnl_header_len(skb); | |
2320 | skb = segs; | |
2321 | do { | |
2322 | struct udphdr *uh; | |
2323 | int udp_offset = outer_hlen - tnl_hlen; | |
2324 | ||
2325 | skb->mac_len = mac_len; | |
2326 | ||
2327 | skb_push(skb, outer_hlen); | |
2328 | skb_reset_mac_header(skb); | |
2329 | skb_set_network_header(skb, mac_len); | |
2330 | skb_set_transport_header(skb, udp_offset); | |
2331 | uh = udp_hdr(skb); | |
2332 | uh->len = htons(skb->len - udp_offset); | |
2333 | ||
2334 | /* csum segment if tunnel sets skb with csum. */ | |
2335 | if (unlikely(uh->check)) { | |
2336 | struct iphdr *iph = ip_hdr(skb); | |
2337 | ||
2338 | uh->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr, | |
2339 | skb->len - udp_offset, | |
2340 | IPPROTO_UDP, 0); | |
2341 | uh->check = csum_fold(skb_checksum(skb, udp_offset, | |
2342 | skb->len - udp_offset, 0)); | |
2343 | if (uh->check == 0) | |
2344 | uh->check = CSUM_MANGLED_0; | |
2345 | ||
2346 | } | |
2347 | skb->ip_summed = CHECKSUM_NONE; | |
0d05535d | 2348 | skb->protocol = protocol; |
73136267 PS |
2349 | } while ((skb = skb->next)); |
2350 | out: | |
2351 | return segs; | |
2352 | } |