]>
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 | * | |
8 | * Version: $Id: udp.c,v 1.102 2002/02/01 22:01:04 davem Exp $ | |
9 | * | |
02c30a84 | 10 | * Authors: Ross Biro |
1da177e4 LT |
11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
12 | * Arnt Gulbrandsen, <agulbra@nvg.unit.no> | |
13 | * Alan Cox, <Alan.Cox@linux.org> | |
14 | * Hirokazu Takahashi, <taka@valinux.co.jp> | |
15 | * | |
16 | * Fixes: | |
17 | * Alan Cox : verify_area() calls | |
18 | * Alan Cox : stopped close while in use off icmp | |
19 | * messages. Not a fix but a botch that | |
20 | * for udp at least is 'valid'. | |
21 | * Alan Cox : Fixed icmp handling properly | |
22 | * Alan Cox : Correct error for oversized datagrams | |
e905a9ed YH |
23 | * Alan Cox : Tidied select() semantics. |
24 | * Alan Cox : udp_err() fixed properly, also now | |
1da177e4 LT |
25 | * select and read wake correctly on errors |
26 | * Alan Cox : udp_send verify_area moved to avoid mem leak | |
27 | * Alan Cox : UDP can count its memory | |
28 | * Alan Cox : send to an unknown connection causes | |
29 | * an ECONNREFUSED off the icmp, but | |
30 | * does NOT close. | |
31 | * Alan Cox : Switched to new sk_buff handlers. No more backlog! | |
32 | * Alan Cox : Using generic datagram code. Even smaller and the PEEK | |
33 | * bug no longer crashes it. | |
34 | * Fred Van Kempen : Net2e support for sk->broadcast. | |
35 | * Alan Cox : Uses skb_free_datagram | |
36 | * Alan Cox : Added get/set sockopt support. | |
37 | * Alan Cox : Broadcasting without option set returns EACCES. | |
38 | * Alan Cox : No wakeup calls. Instead we now use the callbacks. | |
39 | * Alan Cox : Use ip_tos and ip_ttl | |
40 | * Alan Cox : SNMP Mibs | |
41 | * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support. | |
42 | * Matt Dillon : UDP length checks. | |
43 | * Alan Cox : Smarter af_inet used properly. | |
44 | * Alan Cox : Use new kernel side addressing. | |
45 | * Alan Cox : Incorrect return on truncated datagram receive. | |
46 | * Arnt Gulbrandsen : New udp_send and stuff | |
47 | * Alan Cox : Cache last socket | |
48 | * Alan Cox : Route cache | |
49 | * Jon Peatfield : Minor efficiency fix to sendto(). | |
50 | * Mike Shaver : RFC1122 checks. | |
51 | * Alan Cox : Nonblocking error fix. | |
52 | * Willy Konynenberg : Transparent proxying support. | |
53 | * Mike McLagan : Routing by source | |
54 | * David S. Miller : New socket lookup architecture. | |
55 | * Last socket cache retained as it | |
56 | * does have a high hit rate. | |
57 | * Olaf Kirch : Don't linearise iovec on sendmsg. | |
58 | * Andi Kleen : Some cleanups, cache destination entry | |
e905a9ed | 59 | * for connect. |
1da177e4 LT |
60 | * Vitaly E. Lavrov : Transparent proxy revived after year coma. |
61 | * Melvin Smith : Check msg_name not msg_namelen in sendto(), | |
62 | * return ENOTCONN for unconnected sockets (POSIX) | |
63 | * Janos Farkas : don't deliver multi/broadcasts to a different | |
64 | * bound-to-device socket | |
65 | * Hirokazu Takahashi : HW checksumming for outgoing UDP | |
66 | * datagrams. | |
67 | * Hirokazu Takahashi : sendfile() on UDP works now. | |
68 | * Arnaldo C. Melo : convert /proc/net/udp to seq_file | |
69 | * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which | |
70 | * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind | |
71 | * a single port at the same time. | |
72 | * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support | |
342f0234 | 73 | * James Chapman : Add L2TP encapsulation type. |
1da177e4 LT |
74 | * |
75 | * | |
76 | * This program is free software; you can redistribute it and/or | |
77 | * modify it under the terms of the GNU General Public License | |
78 | * as published by the Free Software Foundation; either version | |
79 | * 2 of the License, or (at your option) any later version. | |
80 | */ | |
e905a9ed | 81 | |
1da177e4 LT |
82 | #include <asm/system.h> |
83 | #include <asm/uaccess.h> | |
84 | #include <asm/ioctls.h> | |
85 | #include <linux/types.h> | |
86 | #include <linux/fcntl.h> | |
87 | #include <linux/module.h> | |
88 | #include <linux/socket.h> | |
89 | #include <linux/sockios.h> | |
14c85021 | 90 | #include <linux/igmp.h> |
1da177e4 LT |
91 | #include <linux/in.h> |
92 | #include <linux/errno.h> | |
93 | #include <linux/timer.h> | |
94 | #include <linux/mm.h> | |
1da177e4 | 95 | #include <linux/inet.h> |
1da177e4 | 96 | #include <linux/netdevice.h> |
c752f073 | 97 | #include <net/tcp_states.h> |
1da177e4 LT |
98 | #include <linux/skbuff.h> |
99 | #include <linux/proc_fs.h> | |
100 | #include <linux/seq_file.h> | |
1da177e4 LT |
101 | #include <net/icmp.h> |
102 | #include <net/route.h> | |
1da177e4 LT |
103 | #include <net/checksum.h> |
104 | #include <net/xfrm.h> | |
ba4e58ec | 105 | #include "udp_impl.h" |
1da177e4 LT |
106 | |
107 | /* | |
108 | * Snmp MIB for the UDP layer | |
109 | */ | |
110 | ||
ba89966c | 111 | DEFINE_SNMP_STAT(struct udp_mib, udp_statistics) __read_mostly; |
1da177e4 LT |
112 | |
113 | struct hlist_head udp_hash[UDP_HTABLE_SIZE]; | |
114 | DEFINE_RWLOCK(udp_hash_lock); | |
115 | ||
bed53ea7 | 116 | static int udp_port_rover; |
1da177e4 | 117 | |
df2bc459 | 118 | static inline int __udp_lib_lport_inuse(__u16 num, struct hlist_head udptable[]) |
1da177e4 | 119 | { |
25030a7f | 120 | struct sock *sk; |
1da177e4 | 121 | struct hlist_node *node; |
25030a7f | 122 | |
df2bc459 DM |
123 | sk_for_each(sk, node, &udptable[num & (UDP_HTABLE_SIZE - 1)]) |
124 | if (sk->sk_hash == num) | |
25030a7f GR |
125 | return 1; |
126 | return 0; | |
127 | } | |
128 | ||
129 | /** | |
ba4e58ec | 130 | * __udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
131 | * |
132 | * @sk: socket struct in question | |
133 | * @snum: port number to look up | |
ba4e58ec GR |
134 | * @udptable: hash list table, must be of UDP_HTABLE_SIZE |
135 | * @port_rover: pointer to record of last unallocated port | |
df2bc459 | 136 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25030a7f | 137 | */ |
ba4e58ec GR |
138 | int __udp_lib_get_port(struct sock *sk, unsigned short snum, |
139 | struct hlist_head udptable[], int *port_rover, | |
df2bc459 DM |
140 | int (*saddr_comp)(const struct sock *sk1, |
141 | const struct sock *sk2 ) ) | |
25030a7f GR |
142 | { |
143 | struct hlist_node *node; | |
144 | struct hlist_head *head; | |
1da177e4 | 145 | struct sock *sk2; |
25030a7f | 146 | int error = 1; |
1da177e4 LT |
147 | |
148 | write_lock_bh(&udp_hash_lock); | |
149 | if (snum == 0) { | |
150 | int best_size_so_far, best, result, i; | |
151 | ||
ba4e58ec GR |
152 | if (*port_rover > sysctl_local_port_range[1] || |
153 | *port_rover < sysctl_local_port_range[0]) | |
154 | *port_rover = sysctl_local_port_range[0]; | |
1da177e4 | 155 | best_size_so_far = 32767; |
ba4e58ec | 156 | best = result = *port_rover; |
1da177e4 | 157 | for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) { |
1da177e4 LT |
158 | int size; |
159 | ||
df2bc459 | 160 | head = &udptable[result & (UDP_HTABLE_SIZE - 1)]; |
25030a7f | 161 | if (hlist_empty(head)) { |
1da177e4 LT |
162 | if (result > sysctl_local_port_range[1]) |
163 | result = sysctl_local_port_range[0] + | |
164 | ((result - sysctl_local_port_range[0]) & | |
165 | (UDP_HTABLE_SIZE - 1)); | |
166 | goto gotit; | |
167 | } | |
168 | size = 0; | |
5c668704 DM |
169 | sk_for_each(sk2, node, head) { |
170 | if (++size >= best_size_so_far) | |
171 | goto next; | |
172 | } | |
173 | best_size_so_far = size; | |
174 | best = result; | |
175 | next: | |
176 | ; | |
1da177e4 LT |
177 | } |
178 | result = best; | |
6516c655 SH |
179 | for (i = 0; i < (1 << 16) / UDP_HTABLE_SIZE; |
180 | i++, result += UDP_HTABLE_SIZE) { | |
1da177e4 LT |
181 | if (result > sysctl_local_port_range[1]) |
182 | result = sysctl_local_port_range[0] | |
183 | + ((result - sysctl_local_port_range[0]) & | |
184 | (UDP_HTABLE_SIZE - 1)); | |
df2bc459 | 185 | if (! __udp_lib_lport_inuse(result, udptable)) |
1da177e4 LT |
186 | break; |
187 | } | |
188 | if (i >= (1 << 16) / UDP_HTABLE_SIZE) | |
189 | goto fail; | |
190 | gotit: | |
ba4e58ec | 191 | *port_rover = snum = result; |
1da177e4 | 192 | } else { |
df2bc459 | 193 | head = &udptable[snum & (UDP_HTABLE_SIZE - 1)]; |
25030a7f GR |
194 | |
195 | sk_for_each(sk2, node, head) | |
df2bc459 DM |
196 | if (sk2->sk_hash == snum && |
197 | sk2 != sk && | |
198 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
199 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if | |
200 | || sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
201 | (*saddr_comp)(sk, sk2) ) | |
1da177e4 | 202 | goto fail; |
1da177e4 | 203 | } |
25030a7f | 204 | inet_sk(sk)->num = snum; |
df2bc459 | 205 | sk->sk_hash = snum; |
1da177e4 | 206 | if (sk_unhashed(sk)) { |
df2bc459 | 207 | head = &udptable[snum & (UDP_HTABLE_SIZE - 1)]; |
25030a7f | 208 | sk_add_node(sk, head); |
1da177e4 LT |
209 | sock_prot_inc_use(sk->sk_prot); |
210 | } | |
25030a7f | 211 | error = 0; |
1da177e4 LT |
212 | fail: |
213 | write_unlock_bh(&udp_hash_lock); | |
25030a7f GR |
214 | return error; |
215 | } | |
216 | ||
3fbe070a | 217 | int udp_get_port(struct sock *sk, unsigned short snum, |
df2bc459 | 218 | int (*scmp)(const struct sock *, const struct sock *)) |
ba4e58ec | 219 | { |
df2bc459 | 220 | return __udp_lib_get_port(sk, snum, udp_hash, &udp_port_rover, scmp); |
ba4e58ec GR |
221 | } |
222 | ||
df2bc459 | 223 | int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2) |
25030a7f GR |
224 | { |
225 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
226 | ||
227 | return ( !ipv6_only_sock(sk2) && | |
228 | (!inet1->rcv_saddr || !inet2->rcv_saddr || | |
229 | inet1->rcv_saddr == inet2->rcv_saddr )); | |
230 | } | |
231 | ||
232 | static inline int udp_v4_get_port(struct sock *sk, unsigned short snum) | |
233 | { | |
df2bc459 | 234 | return udp_get_port(sk, snum, ipv4_rcv_saddr_equal); |
1da177e4 LT |
235 | } |
236 | ||
1da177e4 LT |
237 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
238 | * harder than this. -DaveM | |
239 | */ | |
ba4e58ec GR |
240 | static struct sock *__udp4_lib_lookup(__be32 saddr, __be16 sport, |
241 | __be32 daddr, __be16 dport, | |
242 | int dif, struct hlist_head udptable[]) | |
1da177e4 LT |
243 | { |
244 | struct sock *sk, *result = NULL; | |
245 | struct hlist_node *node; | |
df2bc459 DM |
246 | unsigned short hnum = ntohs(dport); |
247 | int badness = -1; | |
1da177e4 | 248 | |
ba4e58ec | 249 | read_lock(&udp_hash_lock); |
df2bc459 | 250 | sk_for_each(sk, node, &udptable[hnum & (UDP_HTABLE_SIZE - 1)]) { |
1da177e4 LT |
251 | struct inet_sock *inet = inet_sk(sk); |
252 | ||
df2bc459 DM |
253 | if (sk->sk_hash == hnum && !ipv6_only_sock(sk)) { |
254 | int score = (sk->sk_family == PF_INET ? 1 : 0); | |
255 | if (inet->rcv_saddr) { | |
256 | if (inet->rcv_saddr != daddr) | |
257 | continue; | |
258 | score+=2; | |
259 | } | |
260 | if (inet->daddr) { | |
261 | if (inet->daddr != saddr) | |
262 | continue; | |
263 | score+=2; | |
264 | } | |
265 | if (inet->dport) { | |
266 | if (inet->dport != sport) | |
267 | continue; | |
268 | score+=2; | |
269 | } | |
270 | if (sk->sk_bound_dev_if) { | |
271 | if (sk->sk_bound_dev_if != dif) | |
272 | continue; | |
273 | score+=2; | |
274 | } | |
275 | if (score == 9) { | |
276 | result = sk; | |
277 | break; | |
278 | } else if (score > badness) { | |
279 | result = sk; | |
280 | badness = score; | |
281 | } | |
1da177e4 LT |
282 | } |
283 | } | |
ba4e58ec GR |
284 | if (result) |
285 | sock_hold(result); | |
1da177e4 | 286 | read_unlock(&udp_hash_lock); |
ba4e58ec | 287 | return result; |
1da177e4 LT |
288 | } |
289 | ||
df2bc459 DM |
290 | static inline struct sock *udp_v4_mcast_next(struct sock *sk, |
291 | __be16 loc_port, __be32 loc_addr, | |
b7b5f487 DM |
292 | __be16 rmt_port, __be32 rmt_addr, |
293 | int dif) | |
1da177e4 LT |
294 | { |
295 | struct hlist_node *node; | |
296 | struct sock *s = sk; | |
df2bc459 | 297 | unsigned short hnum = ntohs(loc_port); |
1da177e4 LT |
298 | |
299 | sk_for_each_from(s, node) { | |
300 | struct inet_sock *inet = inet_sk(s); | |
301 | ||
95f30b33 | 302 | if (s->sk_hash != hnum || |
1da177e4 LT |
303 | (inet->daddr && inet->daddr != rmt_addr) || |
304 | (inet->dport != rmt_port && inet->dport) || | |
305 | (inet->rcv_saddr && inet->rcv_saddr != loc_addr) || | |
306 | ipv6_only_sock(s) || | |
307 | (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)) | |
308 | continue; | |
309 | if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif)) | |
310 | continue; | |
311 | goto found; | |
e905a9ed | 312 | } |
1da177e4 LT |
313 | s = NULL; |
314 | found: | |
e905a9ed | 315 | return s; |
1da177e4 LT |
316 | } |
317 | ||
318 | /* | |
319 | * This routine is called by the ICMP module when it gets some | |
320 | * sort of error condition. If err < 0 then the socket should | |
321 | * be closed and the error returned to the user. If err > 0 | |
e905a9ed | 322 | * it's just the icmp type << 8 | icmp code. |
1da177e4 LT |
323 | * Header points to the ip header of the error packet. We move |
324 | * on past this. Then (as it used to claim before adjustment) | |
325 | * header points to the first 8 bytes of the udp header. We need | |
326 | * to find the appropriate port. | |
327 | */ | |
328 | ||
ba4e58ec | 329 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct hlist_head udptable[]) |
1da177e4 LT |
330 | { |
331 | struct inet_sock *inet; | |
332 | struct iphdr *iph = (struct iphdr*)skb->data; | |
333 | struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2)); | |
88c7664f ACM |
334 | const int type = icmp_hdr(skb)->type; |
335 | const int code = icmp_hdr(skb)->code; | |
1da177e4 LT |
336 | struct sock *sk; |
337 | int harderr; | |
338 | int err; | |
339 | ||
ba4e58ec GR |
340 | sk = __udp4_lib_lookup(iph->daddr, uh->dest, iph->saddr, uh->source, |
341 | skb->dev->ifindex, udptable ); | |
1da177e4 LT |
342 | if (sk == NULL) { |
343 | ICMP_INC_STATS_BH(ICMP_MIB_INERRORS); | |
e905a9ed | 344 | return; /* No socket for error */ |
1da177e4 LT |
345 | } |
346 | ||
347 | err = 0; | |
348 | harderr = 0; | |
349 | inet = inet_sk(sk); | |
350 | ||
351 | switch (type) { | |
352 | default: | |
353 | case ICMP_TIME_EXCEEDED: | |
354 | err = EHOSTUNREACH; | |
355 | break; | |
356 | case ICMP_SOURCE_QUENCH: | |
357 | goto out; | |
358 | case ICMP_PARAMETERPROB: | |
359 | err = EPROTO; | |
360 | harderr = 1; | |
361 | break; | |
362 | case ICMP_DEST_UNREACH: | |
363 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
364 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { | |
365 | err = EMSGSIZE; | |
366 | harderr = 1; | |
367 | break; | |
368 | } | |
369 | goto out; | |
370 | } | |
371 | err = EHOSTUNREACH; | |
372 | if (code <= NR_ICMP_UNREACH) { | |
373 | harderr = icmp_err_convert[code].fatal; | |
374 | err = icmp_err_convert[code].errno; | |
375 | } | |
376 | break; | |
377 | } | |
378 | ||
379 | /* | |
e905a9ed | 380 | * RFC1122: OK. Passes ICMP errors back to application, as per |
1da177e4 LT |
381 | * 4.1.3.3. |
382 | */ | |
383 | if (!inet->recverr) { | |
384 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
385 | goto out; | |
386 | } else { | |
387 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1)); | |
388 | } | |
389 | sk->sk_err = err; | |
390 | sk->sk_error_report(sk); | |
391 | out: | |
392 | sock_put(sk); | |
393 | } | |
394 | ||
3fbe070a | 395 | void udp_err(struct sk_buff *skb, u32 info) |
ba4e58ec GR |
396 | { |
397 | return __udp4_lib_err(skb, info, udp_hash); | |
398 | } | |
399 | ||
1da177e4 LT |
400 | /* |
401 | * Throw away all pending data and cancel the corking. Socket is locked. | |
402 | */ | |
403 | static void udp_flush_pending_frames(struct sock *sk) | |
404 | { | |
405 | struct udp_sock *up = udp_sk(sk); | |
406 | ||
407 | if (up->pending) { | |
408 | up->len = 0; | |
409 | up->pending = 0; | |
410 | ip_flush_pending_frames(sk); | |
411 | } | |
412 | } | |
413 | ||
ba4e58ec GR |
414 | /** |
415 | * udp4_hwcsum_outgoing - handle outgoing HW checksumming | |
416 | * @sk: socket we are sending on | |
417 | * @skb: sk_buff containing the filled-in UDP header | |
418 | * (checksum field must be zeroed out) | |
419 | */ | |
420 | static void udp4_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb, | |
421 | __be32 src, __be32 dst, int len ) | |
422 | { | |
6b11687e | 423 | unsigned int offset; |
4bedb452 | 424 | struct udphdr *uh = udp_hdr(skb); |
6b11687e | 425 | __wsum csum = 0; |
ba4e58ec GR |
426 | |
427 | if (skb_queue_len(&sk->sk_write_queue) == 1) { | |
428 | /* | |
429 | * Only one fragment on the socket. | |
430 | */ | |
663ead3b | 431 | skb->csum_start = skb_transport_header(skb) - skb->head; |
ff1dcadb | 432 | skb->csum_offset = offsetof(struct udphdr, check); |
ba4e58ec GR |
433 | uh->check = ~csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, 0); |
434 | } else { | |
435 | /* | |
436 | * HW-checksum won't work as there are two or more | |
437 | * fragments on the socket so that all csums of sk_buffs | |
438 | * should be together | |
439 | */ | |
ea2ae17d | 440 | offset = skb_transport_offset(skb); |
ba4e58ec GR |
441 | skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); |
442 | ||
443 | skb->ip_summed = CHECKSUM_NONE; | |
444 | ||
445 | skb_queue_walk(&sk->sk_write_queue, skb) { | |
446 | csum = csum_add(csum, skb->csum); | |
447 | } | |
448 | ||
449 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); | |
450 | if (uh->check == 0) | |
f6ab0288 | 451 | uh->check = CSUM_MANGLED_0; |
ba4e58ec GR |
452 | } |
453 | } | |
454 | ||
1da177e4 LT |
455 | /* |
456 | * Push out all pending data as one UDP datagram. Socket is locked. | |
457 | */ | |
4c0a6cb0 | 458 | static int udp_push_pending_frames(struct sock *sk) |
1da177e4 | 459 | { |
4c0a6cb0 | 460 | struct udp_sock *up = udp_sk(sk); |
1da177e4 LT |
461 | struct inet_sock *inet = inet_sk(sk); |
462 | struct flowi *fl = &inet->cork.fl; | |
463 | struct sk_buff *skb; | |
464 | struct udphdr *uh; | |
465 | int err = 0; | |
8e5200f5 | 466 | __wsum csum = 0; |
1da177e4 LT |
467 | |
468 | /* Grab the skbuff where UDP header space exists. */ | |
469 | if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) | |
470 | goto out; | |
471 | ||
472 | /* | |
473 | * Create a UDP header | |
474 | */ | |
4bedb452 | 475 | uh = udp_hdr(skb); |
1da177e4 LT |
476 | uh->source = fl->fl_ip_sport; |
477 | uh->dest = fl->fl_ip_dport; | |
478 | uh->len = htons(up->len); | |
479 | uh->check = 0; | |
480 | ||
ba4e58ec GR |
481 | if (up->pcflag) /* UDP-Lite */ |
482 | csum = udplite_csum_outgoing(sk, skb); | |
483 | ||
484 | else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */ | |
485 | ||
1da177e4 LT |
486 | skb->ip_summed = CHECKSUM_NONE; |
487 | goto send; | |
1da177e4 | 488 | |
ba4e58ec | 489 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ |
1da177e4 | 490 | |
ba4e58ec GR |
491 | udp4_hwcsum_outgoing(sk, skb, fl->fl4_src,fl->fl4_dst, up->len); |
492 | goto send; | |
493 | ||
494 | } else /* `normal' UDP */ | |
495 | csum = udp_csum_outgoing(sk, skb); | |
496 | ||
497 | /* add protocol-dependent pseudo-header */ | |
498 | uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, up->len, | |
499 | sk->sk_protocol, csum ); | |
500 | if (uh->check == 0) | |
f6ab0288 | 501 | uh->check = CSUM_MANGLED_0; |
1da177e4 | 502 | |
1da177e4 LT |
503 | send: |
504 | err = ip_push_pending_frames(sk); | |
505 | out: | |
506 | up->len = 0; | |
507 | up->pending = 0; | |
508 | return err; | |
509 | } | |
510 | ||
1da177e4 LT |
511 | int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, |
512 | size_t len) | |
513 | { | |
514 | struct inet_sock *inet = inet_sk(sk); | |
515 | struct udp_sock *up = udp_sk(sk); | |
516 | int ulen = len; | |
517 | struct ipcm_cookie ipc; | |
518 | struct rtable *rt = NULL; | |
519 | int free = 0; | |
520 | int connected = 0; | |
3ca3c68e | 521 | __be32 daddr, faddr, saddr; |
734ab87f | 522 | __be16 dport; |
1da177e4 | 523 | u8 tos; |
ba4e58ec | 524 | int err, is_udplite = up->pcflag; |
1da177e4 | 525 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; |
ba4e58ec | 526 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); |
1da177e4 LT |
527 | |
528 | if (len > 0xFFFF) | |
529 | return -EMSGSIZE; | |
530 | ||
e905a9ed | 531 | /* |
1da177e4 LT |
532 | * Check the flags. |
533 | */ | |
534 | ||
535 | if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */ | |
536 | return -EOPNOTSUPP; | |
537 | ||
538 | ipc.opt = NULL; | |
539 | ||
540 | if (up->pending) { | |
541 | /* | |
542 | * There are pending frames. | |
e905a9ed | 543 | * The socket lock must be held while it's corked. |
1da177e4 LT |
544 | */ |
545 | lock_sock(sk); | |
546 | if (likely(up->pending)) { | |
547 | if (unlikely(up->pending != AF_INET)) { | |
548 | release_sock(sk); | |
549 | return -EINVAL; | |
550 | } | |
e905a9ed | 551 | goto do_append_data; |
1da177e4 LT |
552 | } |
553 | release_sock(sk); | |
554 | } | |
555 | ulen += sizeof(struct udphdr); | |
556 | ||
557 | /* | |
e905a9ed | 558 | * Get and verify the address. |
1da177e4 LT |
559 | */ |
560 | if (msg->msg_name) { | |
561 | struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name; | |
562 | if (msg->msg_namelen < sizeof(*usin)) | |
563 | return -EINVAL; | |
564 | if (usin->sin_family != AF_INET) { | |
565 | if (usin->sin_family != AF_UNSPEC) | |
566 | return -EAFNOSUPPORT; | |
567 | } | |
568 | ||
569 | daddr = usin->sin_addr.s_addr; | |
570 | dport = usin->sin_port; | |
571 | if (dport == 0) | |
572 | return -EINVAL; | |
573 | } else { | |
574 | if (sk->sk_state != TCP_ESTABLISHED) | |
575 | return -EDESTADDRREQ; | |
576 | daddr = inet->daddr; | |
577 | dport = inet->dport; | |
578 | /* Open fast path for connected socket. | |
579 | Route will not be used, if at least one option is set. | |
580 | */ | |
581 | connected = 1; | |
e905a9ed | 582 | } |
1da177e4 LT |
583 | ipc.addr = inet->saddr; |
584 | ||
585 | ipc.oif = sk->sk_bound_dev_if; | |
586 | if (msg->msg_controllen) { | |
587 | err = ip_cmsg_send(msg, &ipc); | |
588 | if (err) | |
589 | return err; | |
590 | if (ipc.opt) | |
591 | free = 1; | |
592 | connected = 0; | |
593 | } | |
594 | if (!ipc.opt) | |
595 | ipc.opt = inet->opt; | |
596 | ||
597 | saddr = ipc.addr; | |
598 | ipc.addr = faddr = daddr; | |
599 | ||
600 | if (ipc.opt && ipc.opt->srr) { | |
601 | if (!daddr) | |
602 | return -EINVAL; | |
603 | faddr = ipc.opt->faddr; | |
604 | connected = 0; | |
605 | } | |
606 | tos = RT_TOS(inet->tos); | |
607 | if (sock_flag(sk, SOCK_LOCALROUTE) || | |
e905a9ed | 608 | (msg->msg_flags & MSG_DONTROUTE) || |
1da177e4 LT |
609 | (ipc.opt && ipc.opt->is_strictroute)) { |
610 | tos |= RTO_ONLINK; | |
611 | connected = 0; | |
612 | } | |
613 | ||
614 | if (MULTICAST(daddr)) { | |
615 | if (!ipc.oif) | |
616 | ipc.oif = inet->mc_index; | |
617 | if (!saddr) | |
618 | saddr = inet->mc_addr; | |
619 | connected = 0; | |
620 | } | |
621 | ||
622 | if (connected) | |
623 | rt = (struct rtable*)sk_dst_check(sk, 0); | |
624 | ||
625 | if (rt == NULL) { | |
626 | struct flowi fl = { .oif = ipc.oif, | |
627 | .nl_u = { .ip4_u = | |
628 | { .daddr = faddr, | |
629 | .saddr = saddr, | |
630 | .tos = tos } }, | |
ba4e58ec | 631 | .proto = sk->sk_protocol, |
1da177e4 LT |
632 | .uli_u = { .ports = |
633 | { .sport = inet->sport, | |
634 | .dport = dport } } }; | |
beb8d13b | 635 | security_sk_classify_flow(sk, &fl); |
8eb9086f | 636 | err = ip_route_output_flow(&rt, &fl, sk, 1); |
584bdf8c WD |
637 | if (err) { |
638 | if (err == -ENETUNREACH) | |
639 | IP_INC_STATS_BH(IPSTATS_MIB_OUTNOROUTES); | |
1da177e4 | 640 | goto out; |
584bdf8c | 641 | } |
1da177e4 LT |
642 | |
643 | err = -EACCES; | |
644 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
645 | !sock_flag(sk, SOCK_BROADCAST)) | |
646 | goto out; | |
647 | if (connected) | |
648 | sk_dst_set(sk, dst_clone(&rt->u.dst)); | |
649 | } | |
650 | ||
651 | if (msg->msg_flags&MSG_CONFIRM) | |
652 | goto do_confirm; | |
653 | back_from_confirm: | |
654 | ||
655 | saddr = rt->rt_src; | |
656 | if (!ipc.addr) | |
657 | daddr = ipc.addr = rt->rt_dst; | |
658 | ||
659 | lock_sock(sk); | |
660 | if (unlikely(up->pending)) { | |
661 | /* The socket is already corked while preparing it. */ | |
662 | /* ... which is an evident application bug. --ANK */ | |
663 | release_sock(sk); | |
664 | ||
64ce2073 | 665 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n"); |
1da177e4 LT |
666 | err = -EINVAL; |
667 | goto out; | |
668 | } | |
669 | /* | |
670 | * Now cork the socket to pend data. | |
671 | */ | |
672 | inet->cork.fl.fl4_dst = daddr; | |
673 | inet->cork.fl.fl_ip_dport = dport; | |
674 | inet->cork.fl.fl4_src = saddr; | |
675 | inet->cork.fl.fl_ip_sport = inet->sport; | |
676 | up->pending = AF_INET; | |
677 | ||
678 | do_append_data: | |
679 | up->len += ulen; | |
ba4e58ec GR |
680 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; |
681 | err = ip_append_data(sk, getfrag, msg->msg_iov, ulen, | |
682 | sizeof(struct udphdr), &ipc, rt, | |
1da177e4 LT |
683 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); |
684 | if (err) | |
685 | udp_flush_pending_frames(sk); | |
686 | else if (!corkreq) | |
4c0a6cb0 | 687 | err = udp_push_pending_frames(sk); |
1e0c14f4 HX |
688 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) |
689 | up->pending = 0; | |
1da177e4 LT |
690 | release_sock(sk); |
691 | ||
692 | out: | |
693 | ip_rt_put(rt); | |
694 | if (free) | |
695 | kfree(ipc.opt); | |
696 | if (!err) { | |
ba4e58ec | 697 | UDP_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS, is_udplite); |
1da177e4 LT |
698 | return len; |
699 | } | |
81aa646c MB |
700 | /* |
701 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
702 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
703 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
704 | * things). We could add another new stat but at least for now that | |
705 | * seems like overkill. | |
706 | */ | |
707 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
ba4e58ec | 708 | UDP_INC_STATS_USER(UDP_MIB_SNDBUFERRORS, is_udplite); |
81aa646c | 709 | } |
1da177e4 LT |
710 | return err; |
711 | ||
712 | do_confirm: | |
713 | dst_confirm(&rt->u.dst); | |
714 | if (!(msg->msg_flags&MSG_PROBE) || len) | |
715 | goto back_from_confirm; | |
716 | err = 0; | |
717 | goto out; | |
718 | } | |
719 | ||
ba4e58ec GR |
720 | int udp_sendpage(struct sock *sk, struct page *page, int offset, |
721 | size_t size, int flags) | |
1da177e4 LT |
722 | { |
723 | struct udp_sock *up = udp_sk(sk); | |
724 | int ret; | |
725 | ||
726 | if (!up->pending) { | |
727 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
728 | ||
729 | /* Call udp_sendmsg to specify destination address which | |
730 | * sendpage interface can't pass. | |
731 | * This will succeed only when the socket is connected. | |
732 | */ | |
733 | ret = udp_sendmsg(NULL, sk, &msg, 0); | |
734 | if (ret < 0) | |
735 | return ret; | |
736 | } | |
737 | ||
738 | lock_sock(sk); | |
739 | ||
740 | if (unlikely(!up->pending)) { | |
741 | release_sock(sk); | |
742 | ||
64ce2073 | 743 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n"); |
1da177e4 LT |
744 | return -EINVAL; |
745 | } | |
746 | ||
747 | ret = ip_append_page(sk, page, offset, size, flags); | |
748 | if (ret == -EOPNOTSUPP) { | |
749 | release_sock(sk); | |
750 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
751 | size, flags); | |
752 | } | |
753 | if (ret < 0) { | |
754 | udp_flush_pending_frames(sk); | |
755 | goto out; | |
756 | } | |
757 | ||
758 | up->len += size; | |
759 | if (!(up->corkflag || (flags&MSG_MORE))) | |
4c0a6cb0 | 760 | ret = udp_push_pending_frames(sk); |
1da177e4 LT |
761 | if (!ret) |
762 | ret = size; | |
763 | out: | |
764 | release_sock(sk); | |
765 | return ret; | |
766 | } | |
767 | ||
768 | /* | |
769 | * IOCTL requests applicable to the UDP protocol | |
770 | */ | |
e905a9ed | 771 | |
1da177e4 LT |
772 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
773 | { | |
6516c655 SH |
774 | switch (cmd) { |
775 | case SIOCOUTQ: | |
1da177e4 | 776 | { |
6516c655 SH |
777 | int amount = atomic_read(&sk->sk_wmem_alloc); |
778 | return put_user(amount, (int __user *)arg); | |
779 | } | |
1da177e4 | 780 | |
6516c655 SH |
781 | case SIOCINQ: |
782 | { | |
783 | struct sk_buff *skb; | |
784 | unsigned long amount; | |
785 | ||
786 | amount = 0; | |
787 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
788 | skb = skb_peek(&sk->sk_receive_queue); | |
789 | if (skb != NULL) { | |
790 | /* | |
791 | * We will only return the amount | |
792 | * of this packet since that is all | |
793 | * that will be read. | |
794 | */ | |
795 | amount = skb->len - sizeof(struct udphdr); | |
1da177e4 | 796 | } |
6516c655 SH |
797 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
798 | return put_user(amount, (int __user *)arg); | |
799 | } | |
1da177e4 | 800 | |
6516c655 SH |
801 | default: |
802 | return -ENOIOCTLCMD; | |
1da177e4 | 803 | } |
6516c655 SH |
804 | |
805 | return 0; | |
1da177e4 LT |
806 | } |
807 | ||
1da177e4 LT |
808 | /* |
809 | * This should be easy, if there is something there we | |
810 | * return it, otherwise we block. | |
811 | */ | |
812 | ||
ba4e58ec | 813 | int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, |
e905a9ed | 814 | size_t len, int noblock, int flags, int *addr_len) |
1da177e4 LT |
815 | { |
816 | struct inet_sock *inet = inet_sk(sk); | |
e905a9ed YH |
817 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; |
818 | struct sk_buff *skb; | |
759e5d00 HX |
819 | unsigned int ulen, copied; |
820 | int err; | |
821 | int is_udplite = IS_UDPLITE(sk); | |
1da177e4 LT |
822 | |
823 | /* | |
824 | * Check any passed addresses | |
825 | */ | |
826 | if (addr_len) | |
827 | *addr_len=sizeof(*sin); | |
828 | ||
829 | if (flags & MSG_ERRQUEUE) | |
830 | return ip_recv_error(sk, msg, len); | |
831 | ||
832 | try_again: | |
833 | skb = skb_recv_datagram(sk, flags, noblock, &err); | |
834 | if (!skb) | |
835 | goto out; | |
e905a9ed | 836 | |
759e5d00 HX |
837 | ulen = skb->len - sizeof(struct udphdr); |
838 | copied = len; | |
839 | if (copied > ulen) | |
840 | copied = ulen; | |
841 | else if (copied < ulen) | |
1da177e4 | 842 | msg->msg_flags |= MSG_TRUNC; |
1da177e4 | 843 | |
ba4e58ec | 844 | /* |
759e5d00 HX |
845 | * If checksum is needed at all, try to do it while copying the |
846 | * data. If the data is truncated, or if we only want a partial | |
847 | * coverage checksum (UDP-Lite), do it before the copy. | |
ba4e58ec | 848 | */ |
ba4e58ec | 849 | |
759e5d00 HX |
850 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { |
851 | if (udp_lib_checksum_complete(skb)) | |
1da177e4 | 852 | goto csum_copy_err; |
ba4e58ec GR |
853 | } |
854 | ||
60476372 | 855 | if (skb_csum_unnecessary(skb)) |
ba4e58ec GR |
856 | err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), |
857 | msg->msg_iov, copied ); | |
858 | else { | |
1da177e4 LT |
859 | err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov); |
860 | ||
861 | if (err == -EINVAL) | |
862 | goto csum_copy_err; | |
863 | } | |
864 | ||
865 | if (err) | |
866 | goto out_free; | |
867 | ||
868 | sock_recv_timestamp(msg, sk, skb); | |
869 | ||
870 | /* Copy the address. */ | |
871 | if (sin) | |
872 | { | |
873 | sin->sin_family = AF_INET; | |
4bedb452 | 874 | sin->sin_port = udp_hdr(skb)->source; |
eddc9ec5 | 875 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; |
1da177e4 | 876 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); |
e905a9ed | 877 | } |
1da177e4 LT |
878 | if (inet->cmsg_flags) |
879 | ip_cmsg_recv(msg, skb); | |
880 | ||
881 | err = copied; | |
882 | if (flags & MSG_TRUNC) | |
759e5d00 | 883 | err = ulen; |
e905a9ed | 884 | |
1da177e4 | 885 | out_free: |
e905a9ed | 886 | skb_free_datagram(sk, skb); |
1da177e4 | 887 | out: |
e905a9ed | 888 | return err; |
1da177e4 LT |
889 | |
890 | csum_copy_err: | |
ba4e58ec | 891 | UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_udplite); |
1da177e4 | 892 | |
3305b80c | 893 | skb_kill_datagram(sk, skb, flags); |
1da177e4 LT |
894 | |
895 | if (noblock) | |
e905a9ed | 896 | return -EAGAIN; |
1da177e4 LT |
897 | goto try_again; |
898 | } | |
899 | ||
900 | ||
901 | int udp_disconnect(struct sock *sk, int flags) | |
902 | { | |
903 | struct inet_sock *inet = inet_sk(sk); | |
904 | /* | |
905 | * 1003.1g - break association. | |
906 | */ | |
e905a9ed | 907 | |
1da177e4 LT |
908 | sk->sk_state = TCP_CLOSE; |
909 | inet->daddr = 0; | |
910 | inet->dport = 0; | |
911 | sk->sk_bound_dev_if = 0; | |
912 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
913 | inet_reset_saddr(sk); | |
914 | ||
915 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
916 | sk->sk_prot->unhash(sk); | |
917 | inet->sport = 0; | |
918 | } | |
919 | sk_dst_reset(sk); | |
920 | return 0; | |
921 | } | |
922 | ||
1da177e4 | 923 | /* return: |
59c51591 | 924 | * 1 if the UDP system should process it |
1da177e4 LT |
925 | * 0 if we should drop this packet |
926 | * -1 if it should get processed by xfrm4_rcv_encap | |
342f0234 | 927 | * -2 if it should get processed by l2tp |
1da177e4 LT |
928 | */ |
929 | static int udp_encap_rcv(struct sock * sk, struct sk_buff *skb) | |
930 | { | |
1da177e4 | 931 | struct udp_sock *up = udp_sk(sk); |
e905a9ed | 932 | struct udphdr *uh; |
1da177e4 LT |
933 | struct iphdr *iph; |
934 | int iphlen, len; | |
e905a9ed | 935 | |
753eab76 OK |
936 | __u8 *udpdata; |
937 | __be32 *udpdata32; | |
1da177e4 LT |
938 | __u16 encap_type = up->encap_type; |
939 | ||
940 | /* if we're overly short, let UDP handle it */ | |
753eab76 OK |
941 | len = skb->len - sizeof(struct udphdr); |
942 | if (len <= 0) | |
1da177e4 LT |
943 | return 1; |
944 | ||
945 | /* if this is not encapsulated socket, then just return now */ | |
946 | if (!encap_type) | |
947 | return 1; | |
948 | ||
753eab76 OK |
949 | /* If this is a paged skb, make sure we pull up |
950 | * whatever data we need to look at. */ | |
951 | if (!pskb_may_pull(skb, sizeof(struct udphdr) + min(len, 8))) | |
952 | return 1; | |
953 | ||
954 | /* Now we can get the pointers */ | |
4bedb452 | 955 | uh = udp_hdr(skb); |
753eab76 OK |
956 | udpdata = (__u8 *)uh + sizeof(struct udphdr); |
957 | udpdata32 = (__be32 *)udpdata; | |
1da177e4 LT |
958 | |
959 | switch (encap_type) { | |
960 | default: | |
961 | case UDP_ENCAP_ESPINUDP: | |
962 | /* Check if this is a keepalive packet. If so, eat it. */ | |
963 | if (len == 1 && udpdata[0] == 0xff) { | |
964 | return 0; | |
6516c655 | 965 | } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0) { |
1da177e4 LT |
966 | /* ESP Packet without Non-ESP header */ |
967 | len = sizeof(struct udphdr); | |
968 | } else | |
969 | /* Must be an IKE packet.. pass it through */ | |
970 | return 1; | |
971 | break; | |
972 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
973 | /* Check if this is a keepalive packet. If so, eat it. */ | |
974 | if (len == 1 && udpdata[0] == 0xff) { | |
975 | return 0; | |
976 | } else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) && | |
977 | udpdata32[0] == 0 && udpdata32[1] == 0) { | |
e905a9ed | 978 | |
1da177e4 LT |
979 | /* ESP Packet with Non-IKE marker */ |
980 | len = sizeof(struct udphdr) + 2 * sizeof(u32); | |
981 | } else | |
982 | /* Must be an IKE packet.. pass it through */ | |
983 | return 1; | |
984 | break; | |
342f0234 JC |
985 | case UDP_ENCAP_L2TPINUDP: |
986 | /* Let caller know to send this to l2tp */ | |
987 | return -2; | |
1da177e4 LT |
988 | } |
989 | ||
342f0234 JC |
990 | #ifndef CONFIG_XFRM |
991 | return 1; | |
992 | #else | |
1da177e4 LT |
993 | /* At this point we are sure that this is an ESPinUDP packet, |
994 | * so we need to remove 'len' bytes from the packet (the UDP | |
995 | * header and optional ESP marker bytes) and then modify the | |
996 | * protocol to ESP, and then call into the transform receiver. | |
997 | */ | |
4d78b6c7 HX |
998 | if (skb_cloned(skb) && pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) |
999 | return 0; | |
1da177e4 LT |
1000 | |
1001 | /* Now we can update and verify the packet length... */ | |
eddc9ec5 | 1002 | iph = ip_hdr(skb); |
1da177e4 LT |
1003 | iphlen = iph->ihl << 2; |
1004 | iph->tot_len = htons(ntohs(iph->tot_len) - len); | |
1005 | if (skb->len < iphlen + len) { | |
1006 | /* packet is too small!?! */ | |
1007 | return 0; | |
1008 | } | |
1009 | ||
1010 | /* pull the data buffer up to the ESP header and set the | |
1011 | * transport header to point to ESP. Keep UDP on the stack | |
1012 | * for later. | |
1013 | */ | |
badff6d0 ACM |
1014 | __skb_pull(skb, len); |
1015 | skb_reset_transport_header(skb); | |
1da177e4 LT |
1016 | |
1017 | /* modify the protocol (it's ESP!) */ | |
1018 | iph->protocol = IPPROTO_ESP; | |
1019 | ||
1020 | /* and let the caller know to send this into the ESP processor... */ | |
1021 | return -1; | |
1022 | #endif | |
1023 | } | |
1024 | ||
1025 | /* returns: | |
1026 | * -1: error | |
1027 | * 0: success | |
1028 | * >0: "udp encap" protocol resubmission | |
1029 | * | |
1030 | * Note that in the success and error cases, the skb is assumed to | |
1031 | * have either been requeued or freed. | |
1032 | */ | |
ba4e58ec | 1033 | int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb) |
1da177e4 LT |
1034 | { |
1035 | struct udp_sock *up = udp_sk(sk); | |
81aa646c | 1036 | int rc; |
1da177e4 LT |
1037 | |
1038 | /* | |
1039 | * Charge it to the socket, dropping if the queue is full. | |
1040 | */ | |
ba4e58ec GR |
1041 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) |
1042 | goto drop; | |
b59c2701 | 1043 | nf_reset(skb); |
1da177e4 LT |
1044 | |
1045 | if (up->encap_type) { | |
1046 | /* | |
1047 | * This is an encapsulation socket, so let's see if this is | |
1048 | * an encapsulated packet. | |
1049 | * If it's a keepalive packet, then just eat it. | |
1050 | * If it's an encapsulateed packet, then pass it to the | |
1051 | * IPsec xfrm input and return the response | |
1052 | * appropriately. Otherwise, just fall through and | |
1053 | * pass this up the UDP socket. | |
1054 | */ | |
1055 | int ret; | |
1056 | ||
1057 | ret = udp_encap_rcv(sk, skb); | |
1058 | if (ret == 0) { | |
1059 | /* Eat the packet .. */ | |
1060 | kfree_skb(skb); | |
1061 | return 0; | |
1062 | } | |
342f0234 | 1063 | if (ret == -1) { |
1da177e4 LT |
1064 | /* process the ESP packet */ |
1065 | ret = xfrm4_rcv_encap(skb, up->encap_type); | |
ba4e58ec | 1066 | UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS, up->pcflag); |
1da177e4 LT |
1067 | return -ret; |
1068 | } | |
342f0234 JC |
1069 | if (ret == -2) { |
1070 | /* process the L2TP packet */ | |
1071 | if (up->encap_rcv != NULL) { | |
1072 | ret = (*up->encap_rcv)(sk, skb); | |
1073 | if (ret <= 0) { | |
1074 | UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS, up->pcflag); | |
1075 | return ret; | |
1076 | } | |
1077 | ||
1078 | /* FALLTHROUGH -- pass up as UDP packet */ | |
1079 | } | |
1080 | } | |
1081 | ||
1da177e4 LT |
1082 | /* FALLTHROUGH -- it's a UDP Packet */ |
1083 | } | |
1084 | ||
ba4e58ec GR |
1085 | /* |
1086 | * UDP-Lite specific tests, ignored on UDP sockets | |
1087 | */ | |
1088 | if ((up->pcflag & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1089 | ||
1090 | /* | |
1091 | * MIB statistics other than incrementing the error count are | |
1092 | * disabled for the following two types of errors: these depend | |
1093 | * on the application settings, not on the functioning of the | |
1094 | * protocol stack as such. | |
1095 | * | |
1096 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1097 | * way ... to ... at least let the receiving application block | |
1098 | * delivery of packets with coverage values less than a value | |
1099 | * provided by the application." | |
1100 | */ | |
1101 | if (up->pcrlen == 0) { /* full coverage was set */ | |
1102 | LIMIT_NETDEBUG(KERN_WARNING "UDPLITE: partial coverage " | |
1103 | "%d while full coverage %d requested\n", | |
1104 | UDP_SKB_CB(skb)->cscov, skb->len); | |
1105 | goto drop; | |
1da177e4 | 1106 | } |
ba4e58ec GR |
1107 | /* The next case involves violating the min. coverage requested |
1108 | * by the receiver. This is subtle: if receiver wants x and x is | |
1109 | * greater than the buffersize/MTU then receiver will complain | |
1110 | * that it wants x while sender emits packets of smaller size y. | |
1111 | * Therefore the above ...()->partial_cov statement is essential. | |
1112 | */ | |
1113 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
1114 | LIMIT_NETDEBUG(KERN_WARNING | |
1115 | "UDPLITE: coverage %d too small, need min %d\n", | |
1116 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
1117 | goto drop; | |
1118 | } | |
1119 | } | |
1120 | ||
759e5d00 HX |
1121 | if (sk->sk_filter) { |
1122 | if (udp_lib_checksum_complete(skb)) | |
ba4e58ec | 1123 | goto drop; |
1da177e4 LT |
1124 | } |
1125 | ||
81aa646c MB |
1126 | if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) { |
1127 | /* Note that an ENOMEM error is charged twice */ | |
1128 | if (rc == -ENOMEM) | |
ba4e58ec GR |
1129 | UDP_INC_STATS_BH(UDP_MIB_RCVBUFERRORS, up->pcflag); |
1130 | goto drop; | |
1da177e4 | 1131 | } |
ba4e58ec GR |
1132 | |
1133 | UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS, up->pcflag); | |
1da177e4 | 1134 | return 0; |
ba4e58ec GR |
1135 | |
1136 | drop: | |
1137 | UDP_INC_STATS_BH(UDP_MIB_INERRORS, up->pcflag); | |
1138 | kfree_skb(skb); | |
1139 | return -1; | |
1da177e4 LT |
1140 | } |
1141 | ||
1142 | /* | |
1143 | * Multicasts and broadcasts go to each listener. | |
1144 | * | |
1145 | * Note: called only from the BH handler context, | |
1146 | * so we don't need to lock the hashes. | |
1147 | */ | |
ba4e58ec GR |
1148 | static int __udp4_lib_mcast_deliver(struct sk_buff *skb, |
1149 | struct udphdr *uh, | |
1150 | __be32 saddr, __be32 daddr, | |
1151 | struct hlist_head udptable[]) | |
1da177e4 | 1152 | { |
df2bc459 | 1153 | struct sock *sk; |
1da177e4 | 1154 | int dif; |
1da177e4 | 1155 | |
6aaf47fa | 1156 | read_lock(&udp_hash_lock); |
df2bc459 DM |
1157 | sk = sk_head(&udptable[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]); |
1158 | dif = skb->dev->ifindex; | |
1159 | sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif); | |
6aaf47fa | 1160 | if (sk) { |
df2bc459 DM |
1161 | struct sock *sknext = NULL; |
1162 | ||
1da177e4 LT |
1163 | do { |
1164 | struct sk_buff *skb1 = skb; | |
df2bc459 DM |
1165 | |
1166 | sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr, | |
1167 | uh->source, saddr, dif); | |
6516c655 | 1168 | if (sknext) |
1da177e4 LT |
1169 | skb1 = skb_clone(skb, GFP_ATOMIC); |
1170 | ||
6516c655 | 1171 | if (skb1) { |
1da177e4 LT |
1172 | int ret = udp_queue_rcv_skb(sk, skb1); |
1173 | if (ret > 0) | |
df2bc459 DM |
1174 | /* we should probably re-process instead |
1175 | * of dropping packets here. */ | |
1da177e4 LT |
1176 | kfree_skb(skb1); |
1177 | } | |
1178 | sk = sknext; | |
6516c655 | 1179 | } while (sknext); |
1da177e4 LT |
1180 | } else |
1181 | kfree_skb(skb); | |
1182 | read_unlock(&udp_hash_lock); | |
1183 | return 0; | |
1184 | } | |
1185 | ||
1186 | /* Initialize UDP checksum. If exited with zero value (success), | |
1187 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1188 | * Otherwise, csum completion requires chacksumming packet body, | |
1189 | * including udp header and folding it to skb->csum. | |
1190 | */ | |
759e5d00 HX |
1191 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, |
1192 | int proto) | |
1da177e4 | 1193 | { |
eddc9ec5 | 1194 | const struct iphdr *iph; |
759e5d00 HX |
1195 | int err; |
1196 | ||
1197 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1198 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1199 | ||
1200 | if (proto == IPPROTO_UDPLITE) { | |
1201 | err = udplite_checksum_init(skb, uh); | |
1202 | if (err) | |
1203 | return err; | |
1204 | } | |
1205 | ||
eddc9ec5 | 1206 | iph = ip_hdr(skb); |
1da177e4 LT |
1207 | if (uh->check == 0) { |
1208 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
84fa7933 | 1209 | } else if (skb->ip_summed == CHECKSUM_COMPLETE) { |
eddc9ec5 ACM |
1210 | if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, |
1211 | proto, skb->csum)) | |
fb286bb2 | 1212 | skb->ip_summed = CHECKSUM_UNNECESSARY; |
1da177e4 | 1213 | } |
60476372 | 1214 | if (!skb_csum_unnecessary(skb)) |
eddc9ec5 | 1215 | skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, |
759e5d00 | 1216 | skb->len, proto, 0); |
1da177e4 LT |
1217 | /* Probably, we should checksum udp header (it should be in cache |
1218 | * in any case) and data in tiny packets (< rx copybreak). | |
1219 | */ | |
ba4e58ec | 1220 | |
759e5d00 | 1221 | return 0; |
1da177e4 LT |
1222 | } |
1223 | ||
1224 | /* | |
e905a9ed | 1225 | * All we need to do is get the socket, and then do a checksum. |
1da177e4 | 1226 | */ |
e905a9ed | 1227 | |
ba4e58ec | 1228 | int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[], |
759e5d00 | 1229 | int proto) |
1da177e4 | 1230 | { |
e905a9ed | 1231 | struct sock *sk; |
4bedb452 | 1232 | struct udphdr *uh = udp_hdr(skb); |
1da177e4 LT |
1233 | unsigned short ulen; |
1234 | struct rtable *rt = (struct rtable*)skb->dst; | |
eddc9ec5 ACM |
1235 | __be32 saddr = ip_hdr(skb)->saddr; |
1236 | __be32 daddr = ip_hdr(skb)->daddr; | |
1da177e4 LT |
1237 | |
1238 | /* | |
ba4e58ec | 1239 | * Validate the packet. |
1da177e4 LT |
1240 | */ |
1241 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
ba4e58ec | 1242 | goto drop; /* No space for header. */ |
1da177e4 LT |
1243 | |
1244 | ulen = ntohs(uh->len); | |
ba4e58ec | 1245 | if (ulen > skb->len) |
1da177e4 LT |
1246 | goto short_packet; |
1247 | ||
759e5d00 HX |
1248 | if (proto == IPPROTO_UDP) { |
1249 | /* UDP validates ulen. */ | |
ba4e58ec GR |
1250 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) |
1251 | goto short_packet; | |
4bedb452 | 1252 | uh = udp_hdr(skb); |
ba4e58ec | 1253 | } |
1da177e4 | 1254 | |
759e5d00 HX |
1255 | if (udp4_csum_init(skb, uh, proto)) |
1256 | goto csum_error; | |
1257 | ||
6516c655 | 1258 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) |
ba4e58ec | 1259 | return __udp4_lib_mcast_deliver(skb, uh, saddr, daddr, udptable); |
1da177e4 | 1260 | |
ba4e58ec | 1261 | sk = __udp4_lib_lookup(saddr, uh->source, daddr, uh->dest, |
df2bc459 | 1262 | skb->dev->ifindex, udptable ); |
1da177e4 LT |
1263 | |
1264 | if (sk != NULL) { | |
1265 | int ret = udp_queue_rcv_skb(sk, skb); | |
1266 | sock_put(sk); | |
1267 | ||
1268 | /* a return value > 0 means to resubmit the input, but | |
ba4e58ec | 1269 | * it wants the return to be -protocol, or 0 |
1da177e4 LT |
1270 | */ |
1271 | if (ret > 0) | |
1272 | return -ret; | |
1273 | return 0; | |
1274 | } | |
1275 | ||
1276 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1277 | goto drop; | |
b59c2701 | 1278 | nf_reset(skb); |
1da177e4 LT |
1279 | |
1280 | /* No socket. Drop packet silently, if checksum is wrong */ | |
ba4e58ec | 1281 | if (udp_lib_checksum_complete(skb)) |
1da177e4 LT |
1282 | goto csum_error; |
1283 | ||
759e5d00 | 1284 | UDP_INC_STATS_BH(UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
1da177e4 LT |
1285 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1286 | ||
1287 | /* | |
1288 | * Hmm. We got an UDP packet to a port to which we | |
1289 | * don't wanna listen. Ignore it. | |
1290 | */ | |
1291 | kfree_skb(skb); | |
6516c655 | 1292 | return 0; |
1da177e4 LT |
1293 | |
1294 | short_packet: | |
ba4e58ec | 1295 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n", |
759e5d00 | 1296 | proto == IPPROTO_UDPLITE ? "-Lite" : "", |
64ce2073 PM |
1297 | NIPQUAD(saddr), |
1298 | ntohs(uh->source), | |
1299 | ulen, | |
ba4e58ec | 1300 | skb->len, |
64ce2073 PM |
1301 | NIPQUAD(daddr), |
1302 | ntohs(uh->dest)); | |
ba4e58ec | 1303 | goto drop; |
1da177e4 LT |
1304 | |
1305 | csum_error: | |
e905a9ed YH |
1306 | /* |
1307 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1308 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1da177e4 | 1309 | */ |
ba4e58ec | 1310 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n", |
759e5d00 | 1311 | proto == IPPROTO_UDPLITE ? "-Lite" : "", |
64ce2073 PM |
1312 | NIPQUAD(saddr), |
1313 | ntohs(uh->source), | |
1314 | NIPQUAD(daddr), | |
1315 | ntohs(uh->dest), | |
1316 | ulen); | |
1da177e4 | 1317 | drop: |
759e5d00 | 1318 | UDP_INC_STATS_BH(UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
1da177e4 | 1319 | kfree_skb(skb); |
6516c655 | 1320 | return 0; |
1da177e4 LT |
1321 | } |
1322 | ||
3fbe070a | 1323 | int udp_rcv(struct sk_buff *skb) |
ba4e58ec | 1324 | { |
759e5d00 | 1325 | return __udp4_lib_rcv(skb, udp_hash, IPPROTO_UDP); |
ba4e58ec GR |
1326 | } |
1327 | ||
1328 | int udp_destroy_sock(struct sock *sk) | |
1da177e4 LT |
1329 | { |
1330 | lock_sock(sk); | |
1331 | udp_flush_pending_frames(sk); | |
1332 | release_sock(sk); | |
1333 | return 0; | |
1334 | } | |
1335 | ||
1336 | /* | |
1337 | * Socket option code for UDP | |
1338 | */ | |
4c0a6cb0 GR |
1339 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
1340 | char __user *optval, int optlen, | |
1341 | int (*push_pending_frames)(struct sock *)) | |
1da177e4 LT |
1342 | { |
1343 | struct udp_sock *up = udp_sk(sk); | |
1344 | int val; | |
1345 | int err = 0; | |
1346 | ||
6516c655 | 1347 | if (optlen<sizeof(int)) |
1da177e4 LT |
1348 | return -EINVAL; |
1349 | ||
1350 | if (get_user(val, (int __user *)optval)) | |
1351 | return -EFAULT; | |
1352 | ||
6516c655 | 1353 | switch (optname) { |
1da177e4 LT |
1354 | case UDP_CORK: |
1355 | if (val != 0) { | |
1356 | up->corkflag = 1; | |
1357 | } else { | |
1358 | up->corkflag = 0; | |
1359 | lock_sock(sk); | |
4c0a6cb0 | 1360 | (*push_pending_frames)(sk); |
1da177e4 LT |
1361 | release_sock(sk); |
1362 | } | |
1363 | break; | |
e905a9ed | 1364 | |
1da177e4 LT |
1365 | case UDP_ENCAP: |
1366 | switch (val) { | |
1367 | case 0: | |
1368 | case UDP_ENCAP_ESPINUDP: | |
1369 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
342f0234 | 1370 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 LT |
1371 | up->encap_type = val; |
1372 | break; | |
1373 | default: | |
1374 | err = -ENOPROTOOPT; | |
1375 | break; | |
1376 | } | |
1377 | break; | |
1378 | ||
ba4e58ec GR |
1379 | /* |
1380 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
1381 | */ | |
1382 | /* The sender sets actual checksum coverage length via this option. | |
1383 | * The case coverage > packet length is handled by send module. */ | |
1384 | case UDPLITE_SEND_CSCOV: | |
1385 | if (!up->pcflag) /* Disable the option on UDP sockets */ | |
1386 | return -ENOPROTOOPT; | |
1387 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
1388 | val = 8; | |
1389 | up->pcslen = val; | |
1390 | up->pcflag |= UDPLITE_SEND_CC; | |
1391 | break; | |
1392 | ||
e905a9ed YH |
1393 | /* The receiver specifies a minimum checksum coverage value. To make |
1394 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
1395 | * used, this again means full checksum coverage. */ |
1396 | case UDPLITE_RECV_CSCOV: | |
1397 | if (!up->pcflag) /* Disable the option on UDP sockets */ | |
1398 | return -ENOPROTOOPT; | |
1399 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
1400 | val = 8; | |
1401 | up->pcrlen = val; | |
1402 | up->pcflag |= UDPLITE_RECV_CC; | |
1403 | break; | |
1404 | ||
1da177e4 LT |
1405 | default: |
1406 | err = -ENOPROTOOPT; | |
1407 | break; | |
6516c655 | 1408 | } |
1da177e4 LT |
1409 | |
1410 | return err; | |
1411 | } | |
1412 | ||
ba4e58ec GR |
1413 | int udp_setsockopt(struct sock *sk, int level, int optname, |
1414 | char __user *optval, int optlen) | |
3fdadf7d | 1415 | { |
ba4e58ec | 1416 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 GR |
1417 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, |
1418 | udp_push_pending_frames); | |
ba4e58ec | 1419 | return ip_setsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1420 | } |
1421 | ||
1422 | #ifdef CONFIG_COMPAT | |
ba4e58ec GR |
1423 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, |
1424 | char __user *optval, int optlen) | |
3fdadf7d | 1425 | { |
ba4e58ec | 1426 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 GR |
1427 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, |
1428 | udp_push_pending_frames); | |
ba4e58ec | 1429 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1430 | } |
1431 | #endif | |
1432 | ||
4c0a6cb0 GR |
1433 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
1434 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
1435 | { |
1436 | struct udp_sock *up = udp_sk(sk); | |
1437 | int val, len; | |
1438 | ||
6516c655 | 1439 | if (get_user(len,optlen)) |
1da177e4 LT |
1440 | return -EFAULT; |
1441 | ||
1442 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 1443 | |
6516c655 | 1444 | if (len < 0) |
1da177e4 LT |
1445 | return -EINVAL; |
1446 | ||
6516c655 | 1447 | switch (optname) { |
1da177e4 LT |
1448 | case UDP_CORK: |
1449 | val = up->corkflag; | |
1450 | break; | |
1451 | ||
1452 | case UDP_ENCAP: | |
1453 | val = up->encap_type; | |
1454 | break; | |
1455 | ||
ba4e58ec GR |
1456 | /* The following two cannot be changed on UDP sockets, the return is |
1457 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
1458 | case UDPLITE_SEND_CSCOV: | |
1459 | val = up->pcslen; | |
1460 | break; | |
1461 | ||
1462 | case UDPLITE_RECV_CSCOV: | |
1463 | val = up->pcrlen; | |
1464 | break; | |
1465 | ||
1da177e4 LT |
1466 | default: |
1467 | return -ENOPROTOOPT; | |
6516c655 | 1468 | } |
1da177e4 | 1469 | |
6516c655 | 1470 | if (put_user(len, optlen)) |
e905a9ed | 1471 | return -EFAULT; |
6516c655 | 1472 | if (copy_to_user(optval, &val,len)) |
1da177e4 | 1473 | return -EFAULT; |
e905a9ed | 1474 | return 0; |
1da177e4 LT |
1475 | } |
1476 | ||
ba4e58ec GR |
1477 | int udp_getsockopt(struct sock *sk, int level, int optname, |
1478 | char __user *optval, int __user *optlen) | |
3fdadf7d | 1479 | { |
ba4e58ec | 1480 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 | 1481 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); |
ba4e58ec | 1482 | return ip_getsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1483 | } |
1484 | ||
1485 | #ifdef CONFIG_COMPAT | |
ba4e58ec | 1486 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, |
543d9cfe | 1487 | char __user *optval, int __user *optlen) |
3fdadf7d | 1488 | { |
ba4e58ec | 1489 | if (level == SOL_UDP || level == SOL_UDPLITE) |
4c0a6cb0 | 1490 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); |
ba4e58ec | 1491 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
1492 | } |
1493 | #endif | |
1da177e4 LT |
1494 | /** |
1495 | * udp_poll - wait for a UDP event. | |
1496 | * @file - file struct | |
1497 | * @sock - socket | |
1498 | * @wait - poll table | |
1499 | * | |
e905a9ed | 1500 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
1501 | * blocking sockets. If application is using a blocking fd |
1502 | * and a packet with checksum error is in the queue; | |
1503 | * then it could get return from select indicating data available | |
1504 | * but then block when reading it. Add special case code | |
1505 | * to work around these arguably broken applications. | |
1506 | */ | |
1507 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
1508 | { | |
1509 | unsigned int mask = datagram_poll(file, sock, wait); | |
1510 | struct sock *sk = sock->sk; | |
ba4e58ec GR |
1511 | int is_lite = IS_UDPLITE(sk); |
1512 | ||
1da177e4 LT |
1513 | /* Check for false positives due to checksum errors */ |
1514 | if ( (mask & POLLRDNORM) && | |
1515 | !(file->f_flags & O_NONBLOCK) && | |
1516 | !(sk->sk_shutdown & RCV_SHUTDOWN)){ | |
1517 | struct sk_buff_head *rcvq = &sk->sk_receive_queue; | |
1518 | struct sk_buff *skb; | |
1519 | ||
208d8984 | 1520 | spin_lock_bh(&rcvq->lock); |
759e5d00 HX |
1521 | while ((skb = skb_peek(rcvq)) != NULL && |
1522 | udp_lib_checksum_complete(skb)) { | |
1523 | UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_lite); | |
1524 | __skb_unlink(skb, rcvq); | |
1525 | kfree_skb(skb); | |
1da177e4 | 1526 | } |
208d8984 | 1527 | spin_unlock_bh(&rcvq->lock); |
1da177e4 LT |
1528 | |
1529 | /* nothing to see, move along */ | |
1530 | if (skb == NULL) | |
1531 | mask &= ~(POLLIN | POLLRDNORM); | |
1532 | } | |
1533 | ||
1534 | return mask; | |
e905a9ed | 1535 | |
1da177e4 LT |
1536 | } |
1537 | ||
1538 | struct proto udp_prot = { | |
e905a9ed | 1539 | .name = "UDP", |
543d9cfe | 1540 | .owner = THIS_MODULE, |
ba4e58ec | 1541 | .close = udp_lib_close, |
543d9cfe ACM |
1542 | .connect = ip4_datagram_connect, |
1543 | .disconnect = udp_disconnect, | |
1544 | .ioctl = udp_ioctl, | |
1545 | .destroy = udp_destroy_sock, | |
1546 | .setsockopt = udp_setsockopt, | |
1547 | .getsockopt = udp_getsockopt, | |
1548 | .sendmsg = udp_sendmsg, | |
1549 | .recvmsg = udp_recvmsg, | |
1550 | .sendpage = udp_sendpage, | |
1551 | .backlog_rcv = udp_queue_rcv_skb, | |
ba4e58ec GR |
1552 | .hash = udp_lib_hash, |
1553 | .unhash = udp_lib_unhash, | |
543d9cfe ACM |
1554 | .get_port = udp_v4_get_port, |
1555 | .obj_size = sizeof(struct udp_sock), | |
3fdadf7d | 1556 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1557 | .compat_setsockopt = compat_udp_setsockopt, |
1558 | .compat_getsockopt = compat_udp_getsockopt, | |
3fdadf7d | 1559 | #endif |
1da177e4 LT |
1560 | }; |
1561 | ||
1562 | /* ------------------------------------------------------------------------ */ | |
1563 | #ifdef CONFIG_PROC_FS | |
1564 | ||
1565 | static struct sock *udp_get_first(struct seq_file *seq) | |
1566 | { | |
1567 | struct sock *sk; | |
1568 | struct udp_iter_state *state = seq->private; | |
1569 | ||
1570 | for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) { | |
1571 | struct hlist_node *node; | |
ba4e58ec | 1572 | sk_for_each(sk, node, state->hashtable + state->bucket) { |
1da177e4 LT |
1573 | if (sk->sk_family == state->family) |
1574 | goto found; | |
1575 | } | |
1576 | } | |
1577 | sk = NULL; | |
1578 | found: | |
1579 | return sk; | |
1580 | } | |
1581 | ||
1582 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
1583 | { | |
1584 | struct udp_iter_state *state = seq->private; | |
1585 | ||
1586 | do { | |
1587 | sk = sk_next(sk); | |
1588 | try_again: | |
1589 | ; | |
1590 | } while (sk && sk->sk_family != state->family); | |
1591 | ||
1592 | if (!sk && ++state->bucket < UDP_HTABLE_SIZE) { | |
ba4e58ec | 1593 | sk = sk_head(state->hashtable + state->bucket); |
1da177e4 LT |
1594 | goto try_again; |
1595 | } | |
1596 | return sk; | |
1597 | } | |
1598 | ||
1599 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
1600 | { | |
1601 | struct sock *sk = udp_get_first(seq); | |
1602 | ||
1603 | if (sk) | |
6516c655 | 1604 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
1605 | --pos; |
1606 | return pos ? NULL : sk; | |
1607 | } | |
1608 | ||
1609 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
1610 | { | |
1611 | read_lock(&udp_hash_lock); | |
1612 | return *pos ? udp_get_idx(seq, *pos-1) : (void *)1; | |
1613 | } | |
1614 | ||
1615 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
1616 | { | |
1617 | struct sock *sk; | |
1618 | ||
1619 | if (v == (void *)1) | |
1620 | sk = udp_get_idx(seq, 0); | |
1621 | else | |
1622 | sk = udp_get_next(seq, v); | |
1623 | ||
1624 | ++*pos; | |
1625 | return sk; | |
1626 | } | |
1627 | ||
1628 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
1629 | { | |
1630 | read_unlock(&udp_hash_lock); | |
1631 | } | |
1632 | ||
1633 | static int udp_seq_open(struct inode *inode, struct file *file) | |
1634 | { | |
1635 | struct udp_seq_afinfo *afinfo = PDE(inode)->data; | |
1636 | struct seq_file *seq; | |
1637 | int rc = -ENOMEM; | |
0da974f4 | 1638 | struct udp_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL); |
1da177e4 LT |
1639 | |
1640 | if (!s) | |
1641 | goto out; | |
1da177e4 | 1642 | s->family = afinfo->family; |
ba4e58ec | 1643 | s->hashtable = afinfo->hashtable; |
1da177e4 LT |
1644 | s->seq_ops.start = udp_seq_start; |
1645 | s->seq_ops.next = udp_seq_next; | |
1646 | s->seq_ops.show = afinfo->seq_show; | |
1647 | s->seq_ops.stop = udp_seq_stop; | |
1648 | ||
1649 | rc = seq_open(file, &s->seq_ops); | |
1650 | if (rc) | |
1651 | goto out_kfree; | |
1652 | ||
1653 | seq = file->private_data; | |
1654 | seq->private = s; | |
1655 | out: | |
1656 | return rc; | |
1657 | out_kfree: | |
1658 | kfree(s); | |
1659 | goto out; | |
1660 | } | |
1661 | ||
1662 | /* ------------------------------------------------------------------------ */ | |
1663 | int udp_proc_register(struct udp_seq_afinfo *afinfo) | |
1664 | { | |
1665 | struct proc_dir_entry *p; | |
1666 | int rc = 0; | |
1667 | ||
1668 | if (!afinfo) | |
1669 | return -EINVAL; | |
1670 | afinfo->seq_fops->owner = afinfo->owner; | |
1671 | afinfo->seq_fops->open = udp_seq_open; | |
1672 | afinfo->seq_fops->read = seq_read; | |
1673 | afinfo->seq_fops->llseek = seq_lseek; | |
1674 | afinfo->seq_fops->release = seq_release_private; | |
1675 | ||
1676 | p = proc_net_fops_create(afinfo->name, S_IRUGO, afinfo->seq_fops); | |
1677 | if (p) | |
1678 | p->data = afinfo; | |
1679 | else | |
1680 | rc = -ENOMEM; | |
1681 | return rc; | |
1682 | } | |
1683 | ||
1684 | void udp_proc_unregister(struct udp_seq_afinfo *afinfo) | |
1685 | { | |
1686 | if (!afinfo) | |
1687 | return; | |
1688 | proc_net_remove(afinfo->name); | |
1689 | memset(afinfo->seq_fops, 0, sizeof(*afinfo->seq_fops)); | |
1690 | } | |
1691 | ||
1692 | /* ------------------------------------------------------------------------ */ | |
1693 | static void udp4_format_sock(struct sock *sp, char *tmpbuf, int bucket) | |
1694 | { | |
1695 | struct inet_sock *inet = inet_sk(sp); | |
734ab87f AV |
1696 | __be32 dest = inet->daddr; |
1697 | __be32 src = inet->rcv_saddr; | |
1da177e4 LT |
1698 | __u16 destp = ntohs(inet->dport); |
1699 | __u16 srcp = ntohs(inet->sport); | |
1700 | ||
1701 | sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X" | |
1702 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p", | |
e905a9ed | 1703 | bucket, src, srcp, dest, destp, sp->sk_state, |
1da177e4 LT |
1704 | atomic_read(&sp->sk_wmem_alloc), |
1705 | atomic_read(&sp->sk_rmem_alloc), | |
1706 | 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp), | |
1707 | atomic_read(&sp->sk_refcnt), sp); | |
1708 | } | |
1709 | ||
ba4e58ec | 1710 | int udp4_seq_show(struct seq_file *seq, void *v) |
1da177e4 LT |
1711 | { |
1712 | if (v == SEQ_START_TOKEN) | |
1713 | seq_printf(seq, "%-127s\n", | |
1714 | " sl local_address rem_address st tx_queue " | |
1715 | "rx_queue tr tm->when retrnsmt uid timeout " | |
1716 | "inode"); | |
1717 | else { | |
1718 | char tmpbuf[129]; | |
1719 | struct udp_iter_state *state = seq->private; | |
1720 | ||
1721 | udp4_format_sock(v, tmpbuf, state->bucket); | |
1722 | seq_printf(seq, "%-127s\n", tmpbuf); | |
1723 | } | |
1724 | return 0; | |
1725 | } | |
1726 | ||
1727 | /* ------------------------------------------------------------------------ */ | |
1728 | static struct file_operations udp4_seq_fops; | |
1729 | static struct udp_seq_afinfo udp4_seq_afinfo = { | |
1730 | .owner = THIS_MODULE, | |
1731 | .name = "udp", | |
1732 | .family = AF_INET, | |
ba4e58ec | 1733 | .hashtable = udp_hash, |
1da177e4 LT |
1734 | .seq_show = udp4_seq_show, |
1735 | .seq_fops = &udp4_seq_fops, | |
1736 | }; | |
1737 | ||
1738 | int __init udp4_proc_init(void) | |
1739 | { | |
1740 | return udp_proc_register(&udp4_seq_afinfo); | |
1741 | } | |
1742 | ||
1743 | void udp4_proc_exit(void) | |
1744 | { | |
1745 | udp_proc_unregister(&udp4_seq_afinfo); | |
1746 | } | |
1747 | #endif /* CONFIG_PROC_FS */ | |
1748 | ||
1749 | EXPORT_SYMBOL(udp_disconnect); | |
1750 | EXPORT_SYMBOL(udp_hash); | |
1751 | EXPORT_SYMBOL(udp_hash_lock); | |
1752 | EXPORT_SYMBOL(udp_ioctl); | |
25030a7f | 1753 | EXPORT_SYMBOL(udp_get_port); |
1da177e4 LT |
1754 | EXPORT_SYMBOL(udp_prot); |
1755 | EXPORT_SYMBOL(udp_sendmsg); | |
4c0a6cb0 GR |
1756 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1757 | EXPORT_SYMBOL(udp_lib_setsockopt); | |
1da177e4 LT |
1758 | EXPORT_SYMBOL(udp_poll); |
1759 | ||
1760 | #ifdef CONFIG_PROC_FS | |
1761 | EXPORT_SYMBOL(udp_proc_register); | |
1762 | EXPORT_SYMBOL(udp_proc_unregister); | |
1763 | #endif |