]>
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> | |
11 | * Alan Cox, <Alan.Cox@linux.org> | |
12 | * Hirokazu Takahashi, <taka@valinux.co.jp> | |
13 | * | |
14 | * Fixes: | |
15 | * Alan Cox : verify_area() calls | |
16 | * Alan Cox : stopped close while in use off icmp | |
17 | * messages. Not a fix but a botch that | |
18 | * for udp at least is 'valid'. | |
19 | * Alan Cox : Fixed icmp handling properly | |
20 | * Alan Cox : Correct error for oversized datagrams | |
e905a9ed YH |
21 | * Alan Cox : Tidied select() semantics. |
22 | * Alan Cox : udp_err() fixed properly, also now | |
1da177e4 LT |
23 | * select and read wake correctly on errors |
24 | * Alan Cox : udp_send verify_area moved to avoid mem leak | |
25 | * Alan Cox : UDP can count its memory | |
26 | * Alan Cox : send to an unknown connection causes | |
27 | * an ECONNREFUSED off the icmp, but | |
28 | * does NOT close. | |
29 | * Alan Cox : Switched to new sk_buff handlers. No more backlog! | |
30 | * Alan Cox : Using generic datagram code. Even smaller and the PEEK | |
31 | * bug no longer crashes it. | |
32 | * Fred Van Kempen : Net2e support for sk->broadcast. | |
33 | * Alan Cox : Uses skb_free_datagram | |
34 | * Alan Cox : Added get/set sockopt support. | |
35 | * Alan Cox : Broadcasting without option set returns EACCES. | |
36 | * Alan Cox : No wakeup calls. Instead we now use the callbacks. | |
37 | * Alan Cox : Use ip_tos and ip_ttl | |
38 | * Alan Cox : SNMP Mibs | |
39 | * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support. | |
40 | * Matt Dillon : UDP length checks. | |
41 | * Alan Cox : Smarter af_inet used properly. | |
42 | * Alan Cox : Use new kernel side addressing. | |
43 | * Alan Cox : Incorrect return on truncated datagram receive. | |
44 | * Arnt Gulbrandsen : New udp_send and stuff | |
45 | * Alan Cox : Cache last socket | |
46 | * Alan Cox : Route cache | |
47 | * Jon Peatfield : Minor efficiency fix to sendto(). | |
48 | * Mike Shaver : RFC1122 checks. | |
49 | * Alan Cox : Nonblocking error fix. | |
50 | * Willy Konynenberg : Transparent proxying support. | |
51 | * Mike McLagan : Routing by source | |
52 | * David S. Miller : New socket lookup architecture. | |
53 | * Last socket cache retained as it | |
54 | * does have a high hit rate. | |
55 | * Olaf Kirch : Don't linearise iovec on sendmsg. | |
56 | * Andi Kleen : Some cleanups, cache destination entry | |
e905a9ed | 57 | * for connect. |
1da177e4 LT |
58 | * Vitaly E. Lavrov : Transparent proxy revived after year coma. |
59 | * Melvin Smith : Check msg_name not msg_namelen in sendto(), | |
60 | * return ENOTCONN for unconnected sockets (POSIX) | |
61 | * Janos Farkas : don't deliver multi/broadcasts to a different | |
62 | * bound-to-device socket | |
63 | * Hirokazu Takahashi : HW checksumming for outgoing UDP | |
64 | * datagrams. | |
65 | * Hirokazu Takahashi : sendfile() on UDP works now. | |
66 | * Arnaldo C. Melo : convert /proc/net/udp to seq_file | |
67 | * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which | |
68 | * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind | |
69 | * a single port at the same time. | |
70 | * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support | |
342f0234 | 71 | * James Chapman : Add L2TP encapsulation type. |
1da177e4 LT |
72 | * |
73 | * | |
74 | * This program is free software; you can redistribute it and/or | |
75 | * modify it under the terms of the GNU General Public License | |
76 | * as published by the Free Software Foundation; either version | |
77 | * 2 of the License, or (at your option) any later version. | |
78 | */ | |
e905a9ed | 79 | |
1da177e4 LT |
80 | #include <asm/system.h> |
81 | #include <asm/uaccess.h> | |
82 | #include <asm/ioctls.h> | |
95766fff | 83 | #include <linux/bootmem.h> |
1da177e4 LT |
84 | #include <linux/types.h> |
85 | #include <linux/fcntl.h> | |
86 | #include <linux/module.h> | |
87 | #include <linux/socket.h> | |
88 | #include <linux/sockios.h> | |
14c85021 | 89 | #include <linux/igmp.h> |
1da177e4 LT |
90 | #include <linux/in.h> |
91 | #include <linux/errno.h> | |
92 | #include <linux/timer.h> | |
93 | #include <linux/mm.h> | |
1da177e4 | 94 | #include <linux/inet.h> |
1da177e4 | 95 | #include <linux/netdevice.h> |
c752f073 | 96 | #include <net/tcp_states.h> |
1da177e4 LT |
97 | #include <linux/skbuff.h> |
98 | #include <linux/proc_fs.h> | |
99 | #include <linux/seq_file.h> | |
457c4cbc | 100 | #include <net/net_namespace.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; |
1781f7f5 | 112 | EXPORT_SYMBOL(udp_statistics); |
1da177e4 | 113 | |
9055e051 HX |
114 | DEFINE_SNMP_STAT(struct udp_mib, udp_stats_in6) __read_mostly; |
115 | EXPORT_SYMBOL(udp_stats_in6); | |
116 | ||
1da177e4 LT |
117 | struct hlist_head udp_hash[UDP_HTABLE_SIZE]; |
118 | DEFINE_RWLOCK(udp_hash_lock); | |
119 | ||
95766fff HA |
120 | int sysctl_udp_mem[3] __read_mostly; |
121 | int sysctl_udp_rmem_min __read_mostly; | |
122 | int sysctl_udp_wmem_min __read_mostly; | |
123 | ||
124 | EXPORT_SYMBOL(sysctl_udp_mem); | |
125 | EXPORT_SYMBOL(sysctl_udp_rmem_min); | |
126 | EXPORT_SYMBOL(sysctl_udp_wmem_min); | |
127 | ||
128 | atomic_t udp_memory_allocated; | |
129 | EXPORT_SYMBOL(udp_memory_allocated); | |
130 | ||
fa4d3c62 | 131 | static inline int __udp_lib_lport_inuse(struct net *net, __u16 num, |
32c1da70 | 132 | const struct hlist_head udptable[]) |
1da177e4 | 133 | { |
25030a7f | 134 | struct sock *sk; |
1da177e4 | 135 | struct hlist_node *node; |
25030a7f | 136 | |
19c7578f | 137 | sk_for_each(sk, node, &udptable[udp_hashfn(net, num)]) |
878628fb | 138 | if (net_eq(sock_net(sk), net) && sk->sk_hash == num) |
25030a7f GR |
139 | return 1; |
140 | return 0; | |
141 | } | |
142 | ||
143 | /** | |
6ba5a3c5 | 144 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
145 | * |
146 | * @sk: socket struct in question | |
147 | * @snum: port number to look up | |
df2bc459 | 148 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25030a7f | 149 | */ |
6ba5a3c5 | 150 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
df2bc459 DM |
151 | int (*saddr_comp)(const struct sock *sk1, |
152 | const struct sock *sk2 ) ) | |
25030a7f | 153 | { |
6ba5a3c5 | 154 | struct hlist_head *udptable = sk->sk_prot->h.udp_hash; |
25030a7f GR |
155 | struct hlist_node *node; |
156 | struct hlist_head *head; | |
1da177e4 | 157 | struct sock *sk2; |
25030a7f | 158 | int error = 1; |
3b1e0a65 | 159 | struct net *net = sock_net(sk); |
1da177e4 LT |
160 | |
161 | write_lock_bh(&udp_hash_lock); | |
32c1da70 SH |
162 | |
163 | if (!snum) { | |
a25de534 | 164 | int i, low, high, remaining; |
32c1da70 SH |
165 | unsigned rover, best, best_size_so_far; |
166 | ||
227b60f5 | 167 | inet_get_local_port_range(&low, &high); |
a25de534 | 168 | remaining = (high - low) + 1; |
227b60f5 | 169 | |
32c1da70 | 170 | best_size_so_far = UINT_MAX; |
a25de534 | 171 | best = rover = net_random() % remaining + low; |
32c1da70 SH |
172 | |
173 | /* 1st pass: look for empty (or shortest) hash chain */ | |
174 | for (i = 0; i < UDP_HTABLE_SIZE; i++) { | |
175 | int size = 0; | |
176 | ||
19c7578f | 177 | head = &udptable[udp_hashfn(net, rover)]; |
32c1da70 | 178 | if (hlist_empty(head)) |
1da177e4 | 179 | goto gotit; |
32c1da70 | 180 | |
5c668704 DM |
181 | sk_for_each(sk2, node, head) { |
182 | if (++size >= best_size_so_far) | |
183 | goto next; | |
184 | } | |
185 | best_size_so_far = size; | |
32c1da70 | 186 | best = rover; |
5c668704 | 187 | next: |
32c1da70 SH |
188 | /* fold back if end of range */ |
189 | if (++rover > high) | |
190 | rover = low + ((rover - low) | |
191 | & (UDP_HTABLE_SIZE - 1)); | |
192 | ||
193 | ||
1da177e4 | 194 | } |
32c1da70 SH |
195 | |
196 | /* 2nd pass: find hole in shortest hash chain */ | |
197 | rover = best; | |
198 | for (i = 0; i < (1 << 16) / UDP_HTABLE_SIZE; i++) { | |
fa4d3c62 | 199 | if (! __udp_lib_lport_inuse(net, rover, udptable)) |
32c1da70 SH |
200 | goto gotit; |
201 | rover += UDP_HTABLE_SIZE; | |
202 | if (rover > high) | |
203 | rover = low + ((rover - low) | |
204 | & (UDP_HTABLE_SIZE - 1)); | |
1da177e4 | 205 | } |
32c1da70 SH |
206 | |
207 | ||
208 | /* All ports in use! */ | |
209 | goto fail; | |
210 | ||
1da177e4 | 211 | gotit: |
32c1da70 | 212 | snum = rover; |
1da177e4 | 213 | } else { |
19c7578f | 214 | head = &udptable[udp_hashfn(net, snum)]; |
25030a7f GR |
215 | |
216 | sk_for_each(sk2, node, head) | |
df2bc459 DM |
217 | if (sk2->sk_hash == snum && |
218 | sk2 != sk && | |
878628fb | 219 | net_eq(sock_net(sk2), net) && |
df2bc459 DM |
220 | (!sk2->sk_reuse || !sk->sk_reuse) && |
221 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if | |
222 | || sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
223 | (*saddr_comp)(sk, sk2) ) | |
1da177e4 | 224 | goto fail; |
1da177e4 | 225 | } |
32c1da70 | 226 | |
25030a7f | 227 | inet_sk(sk)->num = snum; |
df2bc459 | 228 | sk->sk_hash = snum; |
1da177e4 | 229 | if (sk_unhashed(sk)) { |
19c7578f | 230 | head = &udptable[udp_hashfn(net, snum)]; |
25030a7f | 231 | sk_add_node(sk, head); |
c29a0bc4 | 232 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
1da177e4 | 233 | } |
25030a7f | 234 | error = 0; |
1da177e4 LT |
235 | fail: |
236 | write_unlock_bh(&udp_hash_lock); | |
25030a7f GR |
237 | return error; |
238 | } | |
239 | ||
6ba5a3c5 | 240 | static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2) |
db8dac20 DM |
241 | { |
242 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
243 | ||
244 | return ( !ipv6_only_sock(sk2) && | |
245 | (!inet1->rcv_saddr || !inet2->rcv_saddr || | |
246 | inet1->rcv_saddr == inet2->rcv_saddr )); | |
247 | } | |
248 | ||
6ba5a3c5 | 249 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 250 | { |
6ba5a3c5 | 251 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal); |
db8dac20 DM |
252 | } |
253 | ||
254 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try | |
255 | * harder than this. -DaveM | |
256 | */ | |
257 | static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, | |
258 | __be16 sport, __be32 daddr, __be16 dport, | |
259 | int dif, struct hlist_head udptable[]) | |
260 | { | |
261 | struct sock *sk, *result = NULL; | |
262 | struct hlist_node *node; | |
263 | unsigned short hnum = ntohs(dport); | |
264 | int badness = -1; | |
265 | ||
266 | read_lock(&udp_hash_lock); | |
19c7578f | 267 | sk_for_each(sk, node, &udptable[udp_hashfn(net, hnum)]) { |
db8dac20 DM |
268 | struct inet_sock *inet = inet_sk(sk); |
269 | ||
878628fb | 270 | if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum && |
db8dac20 DM |
271 | !ipv6_only_sock(sk)) { |
272 | int score = (sk->sk_family == PF_INET ? 1 : 0); | |
273 | if (inet->rcv_saddr) { | |
274 | if (inet->rcv_saddr != daddr) | |
275 | continue; | |
276 | score+=2; | |
277 | } | |
278 | if (inet->daddr) { | |
279 | if (inet->daddr != saddr) | |
280 | continue; | |
281 | score+=2; | |
282 | } | |
283 | if (inet->dport) { | |
284 | if (inet->dport != sport) | |
285 | continue; | |
286 | score+=2; | |
287 | } | |
288 | if (sk->sk_bound_dev_if) { | |
289 | if (sk->sk_bound_dev_if != dif) | |
290 | continue; | |
291 | score+=2; | |
292 | } | |
293 | if (score == 9) { | |
294 | result = sk; | |
295 | break; | |
296 | } else if (score > badness) { | |
297 | result = sk; | |
298 | badness = score; | |
299 | } | |
300 | } | |
301 | } | |
302 | if (result) | |
303 | sock_hold(result); | |
304 | read_unlock(&udp_hash_lock); | |
305 | return result; | |
306 | } | |
307 | ||
308 | static inline struct sock *udp_v4_mcast_next(struct sock *sk, | |
309 | __be16 loc_port, __be32 loc_addr, | |
310 | __be16 rmt_port, __be32 rmt_addr, | |
311 | int dif) | |
312 | { | |
313 | struct hlist_node *node; | |
314 | struct sock *s = sk; | |
315 | unsigned short hnum = ntohs(loc_port); | |
316 | ||
317 | sk_for_each_from(s, node) { | |
318 | struct inet_sock *inet = inet_sk(s); | |
319 | ||
320 | if (s->sk_hash != hnum || | |
321 | (inet->daddr && inet->daddr != rmt_addr) || | |
322 | (inet->dport != rmt_port && inet->dport) || | |
323 | (inet->rcv_saddr && inet->rcv_saddr != loc_addr) || | |
324 | ipv6_only_sock(s) || | |
325 | (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)) | |
326 | continue; | |
327 | if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif)) | |
328 | continue; | |
329 | goto found; | |
330 | } | |
331 | s = NULL; | |
332 | found: | |
333 | return s; | |
334 | } | |
335 | ||
336 | /* | |
337 | * This routine is called by the ICMP module when it gets some | |
338 | * sort of error condition. If err < 0 then the socket should | |
339 | * be closed and the error returned to the user. If err > 0 | |
340 | * it's just the icmp type << 8 | icmp code. | |
341 | * Header points to the ip header of the error packet. We move | |
342 | * on past this. Then (as it used to claim before adjustment) | |
343 | * header points to the first 8 bytes of the udp header. We need | |
344 | * to find the appropriate port. | |
345 | */ | |
346 | ||
347 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct hlist_head udptable[]) | |
348 | { | |
349 | struct inet_sock *inet; | |
350 | struct iphdr *iph = (struct iphdr*)skb->data; | |
351 | struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2)); | |
352 | const int type = icmp_hdr(skb)->type; | |
353 | const int code = icmp_hdr(skb)->code; | |
354 | struct sock *sk; | |
355 | int harderr; | |
356 | int err; | |
fd54d716 | 357 | struct net *net = dev_net(skb->dev); |
db8dac20 | 358 | |
fd54d716 | 359 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
db8dac20 DM |
360 | iph->saddr, uh->source, skb->dev->ifindex, udptable); |
361 | if (sk == NULL) { | |
362 | ICMP_INC_STATS_BH(ICMP_MIB_INERRORS); | |
363 | return; /* No socket for error */ | |
364 | } | |
365 | ||
366 | err = 0; | |
367 | harderr = 0; | |
368 | inet = inet_sk(sk); | |
369 | ||
370 | switch (type) { | |
371 | default: | |
372 | case ICMP_TIME_EXCEEDED: | |
373 | err = EHOSTUNREACH; | |
374 | break; | |
375 | case ICMP_SOURCE_QUENCH: | |
376 | goto out; | |
377 | case ICMP_PARAMETERPROB: | |
378 | err = EPROTO; | |
379 | harderr = 1; | |
380 | break; | |
381 | case ICMP_DEST_UNREACH: | |
382 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
383 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { | |
384 | err = EMSGSIZE; | |
385 | harderr = 1; | |
386 | break; | |
387 | } | |
388 | goto out; | |
389 | } | |
390 | err = EHOSTUNREACH; | |
391 | if (code <= NR_ICMP_UNREACH) { | |
392 | harderr = icmp_err_convert[code].fatal; | |
393 | err = icmp_err_convert[code].errno; | |
394 | } | |
395 | break; | |
396 | } | |
397 | ||
398 | /* | |
399 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
400 | * 4.1.3.3. | |
401 | */ | |
402 | if (!inet->recverr) { | |
403 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
404 | goto out; | |
405 | } else { | |
406 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1)); | |
407 | } | |
408 | sk->sk_err = err; | |
409 | sk->sk_error_report(sk); | |
410 | out: | |
411 | sock_put(sk); | |
412 | } | |
413 | ||
414 | void udp_err(struct sk_buff *skb, u32 info) | |
415 | { | |
416 | __udp4_lib_err(skb, info, udp_hash); | |
417 | } | |
418 | ||
419 | /* | |
420 | * Throw away all pending data and cancel the corking. Socket is locked. | |
421 | */ | |
36d926b9 | 422 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
423 | { |
424 | struct udp_sock *up = udp_sk(sk); | |
425 | ||
426 | if (up->pending) { | |
427 | up->len = 0; | |
428 | up->pending = 0; | |
429 | ip_flush_pending_frames(sk); | |
430 | } | |
431 | } | |
36d926b9 | 432 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
433 | |
434 | /** | |
435 | * udp4_hwcsum_outgoing - handle outgoing HW checksumming | |
436 | * @sk: socket we are sending on | |
437 | * @skb: sk_buff containing the filled-in UDP header | |
438 | * (checksum field must be zeroed out) | |
439 | */ | |
440 | static void udp4_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb, | |
441 | __be32 src, __be32 dst, int len ) | |
442 | { | |
443 | unsigned int offset; | |
444 | struct udphdr *uh = udp_hdr(skb); | |
445 | __wsum csum = 0; | |
446 | ||
447 | if (skb_queue_len(&sk->sk_write_queue) == 1) { | |
448 | /* | |
449 | * Only one fragment on the socket. | |
450 | */ | |
451 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
452 | skb->csum_offset = offsetof(struct udphdr, check); | |
453 | uh->check = ~csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, 0); | |
454 | } else { | |
455 | /* | |
456 | * HW-checksum won't work as there are two or more | |
457 | * fragments on the socket so that all csums of sk_buffs | |
458 | * should be together | |
459 | */ | |
460 | offset = skb_transport_offset(skb); | |
461 | skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); | |
462 | ||
463 | skb->ip_summed = CHECKSUM_NONE; | |
464 | ||
465 | skb_queue_walk(&sk->sk_write_queue, skb) { | |
466 | csum = csum_add(csum, skb->csum); | |
467 | } | |
468 | ||
469 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); | |
470 | if (uh->check == 0) | |
471 | uh->check = CSUM_MANGLED_0; | |
472 | } | |
473 | } | |
474 | ||
475 | /* | |
476 | * Push out all pending data as one UDP datagram. Socket is locked. | |
477 | */ | |
478 | static int udp_push_pending_frames(struct sock *sk) | |
479 | { | |
480 | struct udp_sock *up = udp_sk(sk); | |
481 | struct inet_sock *inet = inet_sk(sk); | |
482 | struct flowi *fl = &inet->cork.fl; | |
483 | struct sk_buff *skb; | |
484 | struct udphdr *uh; | |
485 | int err = 0; | |
486 | int is_udplite = IS_UDPLITE(sk); | |
487 | __wsum csum = 0; | |
488 | ||
489 | /* Grab the skbuff where UDP header space exists. */ | |
490 | if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) | |
491 | goto out; | |
492 | ||
493 | /* | |
494 | * Create a UDP header | |
495 | */ | |
496 | uh = udp_hdr(skb); | |
497 | uh->source = fl->fl_ip_sport; | |
498 | uh->dest = fl->fl_ip_dport; | |
499 | uh->len = htons(up->len); | |
500 | uh->check = 0; | |
501 | ||
502 | if (is_udplite) /* UDP-Lite */ | |
503 | csum = udplite_csum_outgoing(sk, skb); | |
504 | ||
505 | else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */ | |
506 | ||
507 | skb->ip_summed = CHECKSUM_NONE; | |
508 | goto send; | |
509 | ||
510 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
511 | ||
512 | udp4_hwcsum_outgoing(sk, skb, fl->fl4_src,fl->fl4_dst, up->len); | |
513 | goto send; | |
514 | ||
515 | } else /* `normal' UDP */ | |
516 | csum = udp_csum_outgoing(sk, skb); | |
517 | ||
518 | /* add protocol-dependent pseudo-header */ | |
519 | uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, up->len, | |
520 | sk->sk_protocol, csum ); | |
521 | if (uh->check == 0) | |
522 | uh->check = CSUM_MANGLED_0; | |
523 | ||
524 | send: | |
525 | err = ip_push_pending_frames(sk); | |
526 | out: | |
527 | up->len = 0; | |
528 | up->pending = 0; | |
529 | if (!err) | |
629ca23c PE |
530 | UDP_INC_STATS_USER(sock_net(sk), |
531 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
db8dac20 DM |
532 | return err; |
533 | } | |
534 | ||
535 | int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
536 | size_t len) | |
537 | { | |
538 | struct inet_sock *inet = inet_sk(sk); | |
539 | struct udp_sock *up = udp_sk(sk); | |
540 | int ulen = len; | |
541 | struct ipcm_cookie ipc; | |
542 | struct rtable *rt = NULL; | |
543 | int free = 0; | |
544 | int connected = 0; | |
545 | __be32 daddr, faddr, saddr; | |
546 | __be16 dport; | |
547 | u8 tos; | |
548 | int err, is_udplite = IS_UDPLITE(sk); | |
549 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
550 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
551 | ||
552 | if (len > 0xFFFF) | |
553 | return -EMSGSIZE; | |
554 | ||
555 | /* | |
556 | * Check the flags. | |
557 | */ | |
558 | ||
559 | if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */ | |
560 | return -EOPNOTSUPP; | |
561 | ||
562 | ipc.opt = NULL; | |
563 | ||
564 | if (up->pending) { | |
565 | /* | |
566 | * There are pending frames. | |
567 | * The socket lock must be held while it's corked. | |
568 | */ | |
569 | lock_sock(sk); | |
570 | if (likely(up->pending)) { | |
571 | if (unlikely(up->pending != AF_INET)) { | |
572 | release_sock(sk); | |
573 | return -EINVAL; | |
574 | } | |
575 | goto do_append_data; | |
576 | } | |
577 | release_sock(sk); | |
578 | } | |
579 | ulen += sizeof(struct udphdr); | |
580 | ||
581 | /* | |
582 | * Get and verify the address. | |
583 | */ | |
584 | if (msg->msg_name) { | |
585 | struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name; | |
586 | if (msg->msg_namelen < sizeof(*usin)) | |
587 | return -EINVAL; | |
588 | if (usin->sin_family != AF_INET) { | |
589 | if (usin->sin_family != AF_UNSPEC) | |
590 | return -EAFNOSUPPORT; | |
591 | } | |
592 | ||
593 | daddr = usin->sin_addr.s_addr; | |
594 | dport = usin->sin_port; | |
595 | if (dport == 0) | |
596 | return -EINVAL; | |
597 | } else { | |
598 | if (sk->sk_state != TCP_ESTABLISHED) | |
599 | return -EDESTADDRREQ; | |
600 | daddr = inet->daddr; | |
601 | dport = inet->dport; | |
602 | /* Open fast path for connected socket. | |
603 | Route will not be used, if at least one option is set. | |
604 | */ | |
605 | connected = 1; | |
606 | } | |
607 | ipc.addr = inet->saddr; | |
608 | ||
609 | ipc.oif = sk->sk_bound_dev_if; | |
610 | if (msg->msg_controllen) { | |
3b1e0a65 | 611 | err = ip_cmsg_send(sock_net(sk), msg, &ipc); |
db8dac20 DM |
612 | if (err) |
613 | return err; | |
614 | if (ipc.opt) | |
615 | free = 1; | |
616 | connected = 0; | |
617 | } | |
618 | if (!ipc.opt) | |
619 | ipc.opt = inet->opt; | |
620 | ||
621 | saddr = ipc.addr; | |
622 | ipc.addr = faddr = daddr; | |
623 | ||
624 | if (ipc.opt && ipc.opt->srr) { | |
625 | if (!daddr) | |
626 | return -EINVAL; | |
627 | faddr = ipc.opt->faddr; | |
628 | connected = 0; | |
629 | } | |
630 | tos = RT_TOS(inet->tos); | |
631 | if (sock_flag(sk, SOCK_LOCALROUTE) || | |
632 | (msg->msg_flags & MSG_DONTROUTE) || | |
633 | (ipc.opt && ipc.opt->is_strictroute)) { | |
634 | tos |= RTO_ONLINK; | |
635 | connected = 0; | |
636 | } | |
637 | ||
638 | if (ipv4_is_multicast(daddr)) { | |
639 | if (!ipc.oif) | |
640 | ipc.oif = inet->mc_index; | |
641 | if (!saddr) | |
642 | saddr = inet->mc_addr; | |
643 | connected = 0; | |
644 | } | |
645 | ||
646 | if (connected) | |
647 | rt = (struct rtable*)sk_dst_check(sk, 0); | |
648 | ||
649 | if (rt == NULL) { | |
650 | struct flowi fl = { .oif = ipc.oif, | |
651 | .nl_u = { .ip4_u = | |
652 | { .daddr = faddr, | |
653 | .saddr = saddr, | |
654 | .tos = tos } }, | |
655 | .proto = sk->sk_protocol, | |
656 | .uli_u = { .ports = | |
657 | { .sport = inet->sport, | |
658 | .dport = dport } } }; | |
659 | security_sk_classify_flow(sk, &fl); | |
3b1e0a65 | 660 | err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, 1); |
db8dac20 DM |
661 | if (err) { |
662 | if (err == -ENETUNREACH) | |
663 | IP_INC_STATS_BH(IPSTATS_MIB_OUTNOROUTES); | |
664 | goto out; | |
665 | } | |
666 | ||
667 | err = -EACCES; | |
668 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
669 | !sock_flag(sk, SOCK_BROADCAST)) | |
670 | goto out; | |
671 | if (connected) | |
672 | sk_dst_set(sk, dst_clone(&rt->u.dst)); | |
673 | } | |
674 | ||
675 | if (msg->msg_flags&MSG_CONFIRM) | |
676 | goto do_confirm; | |
677 | back_from_confirm: | |
678 | ||
679 | saddr = rt->rt_src; | |
680 | if (!ipc.addr) | |
681 | daddr = ipc.addr = rt->rt_dst; | |
682 | ||
683 | lock_sock(sk); | |
684 | if (unlikely(up->pending)) { | |
685 | /* The socket is already corked while preparing it. */ | |
686 | /* ... which is an evident application bug. --ANK */ | |
687 | release_sock(sk); | |
688 | ||
689 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n"); | |
690 | err = -EINVAL; | |
691 | goto out; | |
692 | } | |
693 | /* | |
694 | * Now cork the socket to pend data. | |
695 | */ | |
696 | inet->cork.fl.fl4_dst = daddr; | |
697 | inet->cork.fl.fl_ip_dport = dport; | |
698 | inet->cork.fl.fl4_src = saddr; | |
699 | inet->cork.fl.fl_ip_sport = inet->sport; | |
700 | up->pending = AF_INET; | |
701 | ||
702 | do_append_data: | |
703 | up->len += ulen; | |
704 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; | |
705 | err = ip_append_data(sk, getfrag, msg->msg_iov, ulen, | |
706 | sizeof(struct udphdr), &ipc, rt, | |
707 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | |
708 | if (err) | |
709 | udp_flush_pending_frames(sk); | |
710 | else if (!corkreq) | |
711 | err = udp_push_pending_frames(sk); | |
712 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
713 | up->pending = 0; | |
714 | release_sock(sk); | |
715 | ||
716 | out: | |
717 | ip_rt_put(rt); | |
718 | if (free) | |
719 | kfree(ipc.opt); | |
720 | if (!err) | |
721 | return len; | |
722 | /* | |
723 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
724 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
725 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
726 | * things). We could add another new stat but at least for now that | |
727 | * seems like overkill. | |
728 | */ | |
729 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
730 | UDP_INC_STATS_USER(sock_net(sk), |
731 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
732 | } |
733 | return err; | |
734 | ||
735 | do_confirm: | |
736 | dst_confirm(&rt->u.dst); | |
737 | if (!(msg->msg_flags&MSG_PROBE) || len) | |
738 | goto back_from_confirm; | |
739 | err = 0; | |
740 | goto out; | |
741 | } | |
742 | ||
743 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
744 | size_t size, int flags) | |
745 | { | |
746 | struct udp_sock *up = udp_sk(sk); | |
747 | int ret; | |
748 | ||
749 | if (!up->pending) { | |
750 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
751 | ||
752 | /* Call udp_sendmsg to specify destination address which | |
753 | * sendpage interface can't pass. | |
754 | * This will succeed only when the socket is connected. | |
755 | */ | |
756 | ret = udp_sendmsg(NULL, sk, &msg, 0); | |
757 | if (ret < 0) | |
758 | return ret; | |
759 | } | |
760 | ||
761 | lock_sock(sk); | |
762 | ||
763 | if (unlikely(!up->pending)) { | |
764 | release_sock(sk); | |
765 | ||
766 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n"); | |
767 | return -EINVAL; | |
768 | } | |
769 | ||
770 | ret = ip_append_page(sk, page, offset, size, flags); | |
771 | if (ret == -EOPNOTSUPP) { | |
772 | release_sock(sk); | |
773 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
774 | size, flags); | |
775 | } | |
776 | if (ret < 0) { | |
777 | udp_flush_pending_frames(sk); | |
778 | goto out; | |
779 | } | |
780 | ||
781 | up->len += size; | |
782 | if (!(up->corkflag || (flags&MSG_MORE))) | |
783 | ret = udp_push_pending_frames(sk); | |
784 | if (!ret) | |
785 | ret = size; | |
786 | out: | |
787 | release_sock(sk); | |
788 | return ret; | |
789 | } | |
790 | ||
1da177e4 LT |
791 | /* |
792 | * IOCTL requests applicable to the UDP protocol | |
793 | */ | |
e905a9ed | 794 | |
1da177e4 LT |
795 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
796 | { | |
6516c655 SH |
797 | switch (cmd) { |
798 | case SIOCOUTQ: | |
1da177e4 | 799 | { |
6516c655 SH |
800 | int amount = atomic_read(&sk->sk_wmem_alloc); |
801 | return put_user(amount, (int __user *)arg); | |
802 | } | |
1da177e4 | 803 | |
6516c655 SH |
804 | case SIOCINQ: |
805 | { | |
806 | struct sk_buff *skb; | |
807 | unsigned long amount; | |
808 | ||
809 | amount = 0; | |
810 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
811 | skb = skb_peek(&sk->sk_receive_queue); | |
812 | if (skb != NULL) { | |
813 | /* | |
814 | * We will only return the amount | |
815 | * of this packet since that is all | |
816 | * that will be read. | |
817 | */ | |
818 | amount = skb->len - sizeof(struct udphdr); | |
1da177e4 | 819 | } |
6516c655 SH |
820 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
821 | return put_user(amount, (int __user *)arg); | |
822 | } | |
1da177e4 | 823 | |
6516c655 SH |
824 | default: |
825 | return -ENOIOCTLCMD; | |
1da177e4 | 826 | } |
6516c655 SH |
827 | |
828 | return 0; | |
1da177e4 LT |
829 | } |
830 | ||
db8dac20 DM |
831 | /* |
832 | * This should be easy, if there is something there we | |
833 | * return it, otherwise we block. | |
834 | */ | |
835 | ||
836 | int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
837 | size_t len, int noblock, int flags, int *addr_len) | |
838 | { | |
839 | struct inet_sock *inet = inet_sk(sk); | |
840 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; | |
841 | struct sk_buff *skb; | |
842 | unsigned int ulen, copied; | |
843 | int peeked; | |
844 | int err; | |
845 | int is_udplite = IS_UDPLITE(sk); | |
846 | ||
847 | /* | |
848 | * Check any passed addresses | |
849 | */ | |
850 | if (addr_len) | |
851 | *addr_len=sizeof(*sin); | |
852 | ||
853 | if (flags & MSG_ERRQUEUE) | |
854 | return ip_recv_error(sk, msg, len); | |
855 | ||
856 | try_again: | |
857 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), | |
858 | &peeked, &err); | |
859 | if (!skb) | |
860 | goto out; | |
861 | ||
862 | ulen = skb->len - sizeof(struct udphdr); | |
863 | copied = len; | |
864 | if (copied > ulen) | |
865 | copied = ulen; | |
866 | else if (copied < ulen) | |
867 | msg->msg_flags |= MSG_TRUNC; | |
868 | ||
869 | /* | |
870 | * If checksum is needed at all, try to do it while copying the | |
871 | * data. If the data is truncated, or if we only want a partial | |
872 | * coverage checksum (UDP-Lite), do it before the copy. | |
873 | */ | |
874 | ||
875 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { | |
876 | if (udp_lib_checksum_complete(skb)) | |
877 | goto csum_copy_err; | |
878 | } | |
879 | ||
880 | if (skb_csum_unnecessary(skb)) | |
881 | err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), | |
882 | msg->msg_iov, copied ); | |
883 | else { | |
884 | err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov); | |
885 | ||
886 | if (err == -EINVAL) | |
887 | goto csum_copy_err; | |
888 | } | |
889 | ||
890 | if (err) | |
891 | goto out_free; | |
892 | ||
893 | if (!peeked) | |
629ca23c PE |
894 | UDP_INC_STATS_USER(sock_net(sk), |
895 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 DM |
896 | |
897 | sock_recv_timestamp(msg, sk, skb); | |
898 | ||
899 | /* Copy the address. */ | |
900 | if (sin) | |
901 | { | |
902 | sin->sin_family = AF_INET; | |
903 | sin->sin_port = udp_hdr(skb)->source; | |
904 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
905 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
906 | } | |
907 | if (inet->cmsg_flags) | |
908 | ip_cmsg_recv(msg, skb); | |
909 | ||
910 | err = copied; | |
911 | if (flags & MSG_TRUNC) | |
912 | err = ulen; | |
913 | ||
914 | out_free: | |
915 | lock_sock(sk); | |
916 | skb_free_datagram(sk, skb); | |
917 | release_sock(sk); | |
918 | out: | |
919 | return err; | |
920 | ||
921 | csum_copy_err: | |
922 | lock_sock(sk); | |
923 | if (!skb_kill_datagram(sk, skb, flags)) | |
629ca23c | 924 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
db8dac20 DM |
925 | release_sock(sk); |
926 | ||
927 | if (noblock) | |
928 | return -EAGAIN; | |
929 | goto try_again; | |
930 | } | |
931 | ||
932 | ||
1da177e4 LT |
933 | int udp_disconnect(struct sock *sk, int flags) |
934 | { | |
935 | struct inet_sock *inet = inet_sk(sk); | |
936 | /* | |
937 | * 1003.1g - break association. | |
938 | */ | |
e905a9ed | 939 | |
1da177e4 LT |
940 | sk->sk_state = TCP_CLOSE; |
941 | inet->daddr = 0; | |
942 | inet->dport = 0; | |
943 | sk->sk_bound_dev_if = 0; | |
944 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
945 | inet_reset_saddr(sk); | |
946 | ||
947 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
948 | sk->sk_prot->unhash(sk); | |
949 | inet->sport = 0; | |
950 | } | |
951 | sk_dst_reset(sk); | |
952 | return 0; | |
953 | } | |
954 | ||
db8dac20 DM |
955 | /* returns: |
956 | * -1: error | |
957 | * 0: success | |
958 | * >0: "udp encap" protocol resubmission | |
959 | * | |
960 | * Note that in the success and error cases, the skb is assumed to | |
961 | * have either been requeued or freed. | |
962 | */ | |
963 | int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb) | |
964 | { | |
965 | struct udp_sock *up = udp_sk(sk); | |
966 | int rc; | |
967 | int is_udplite = IS_UDPLITE(sk); | |
968 | ||
969 | /* | |
970 | * Charge it to the socket, dropping if the queue is full. | |
971 | */ | |
972 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
973 | goto drop; | |
974 | nf_reset(skb); | |
975 | ||
976 | if (up->encap_type) { | |
977 | /* | |
978 | * This is an encapsulation socket so pass the skb to | |
979 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
980 | * fall through and pass this up the UDP socket. | |
981 | * up->encap_rcv() returns the following value: | |
982 | * =0 if skb was successfully passed to the encap | |
983 | * handler or was discarded by it. | |
984 | * >0 if skb should be passed on to UDP. | |
985 | * <0 if skb should be resubmitted as proto -N | |
986 | */ | |
987 | ||
988 | /* if we're overly short, let UDP handle it */ | |
989 | if (skb->len > sizeof(struct udphdr) && | |
990 | up->encap_rcv != NULL) { | |
991 | int ret; | |
992 | ||
993 | ret = (*up->encap_rcv)(sk, skb); | |
994 | if (ret <= 0) { | |
0283328e PE |
995 | UDP_INC_STATS_BH(sock_net(sk), |
996 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
997 | is_udplite); |
998 | return -ret; | |
999 | } | |
1000 | } | |
1001 | ||
1002 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1003 | } | |
1004 | ||
1005 | /* | |
1006 | * UDP-Lite specific tests, ignored on UDP sockets | |
1007 | */ | |
1008 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1009 | ||
1010 | /* | |
1011 | * MIB statistics other than incrementing the error count are | |
1012 | * disabled for the following two types of errors: these depend | |
1013 | * on the application settings, not on the functioning of the | |
1014 | * protocol stack as such. | |
1015 | * | |
1016 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1017 | * way ... to ... at least let the receiving application block | |
1018 | * delivery of packets with coverage values less than a value | |
1019 | * provided by the application." | |
1020 | */ | |
1021 | if (up->pcrlen == 0) { /* full coverage was set */ | |
1022 | LIMIT_NETDEBUG(KERN_WARNING "UDPLITE: partial coverage " | |
1023 | "%d while full coverage %d requested\n", | |
1024 | UDP_SKB_CB(skb)->cscov, skb->len); | |
1025 | goto drop; | |
1026 | } | |
1027 | /* The next case involves violating the min. coverage requested | |
1028 | * by the receiver. This is subtle: if receiver wants x and x is | |
1029 | * greater than the buffersize/MTU then receiver will complain | |
1030 | * that it wants x while sender emits packets of smaller size y. | |
1031 | * Therefore the above ...()->partial_cov statement is essential. | |
1032 | */ | |
1033 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
1034 | LIMIT_NETDEBUG(KERN_WARNING | |
1035 | "UDPLITE: coverage %d too small, need min %d\n", | |
1036 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
1037 | goto drop; | |
1038 | } | |
1039 | } | |
1040 | ||
1041 | if (sk->sk_filter) { | |
1042 | if (udp_lib_checksum_complete(skb)) | |
1043 | goto drop; | |
1044 | } | |
1045 | ||
1046 | if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) { | |
1047 | /* Note that an ENOMEM error is charged twice */ | |
cb61cb9b | 1048 | if (rc == -ENOMEM) { |
0283328e PE |
1049 | UDP_INC_STATS_BH(sock_net(sk), |
1050 | UDP_MIB_RCVBUFERRORS, is_udplite); | |
cb61cb9b ED |
1051 | atomic_inc(&sk->sk_drops); |
1052 | } | |
db8dac20 DM |
1053 | goto drop; |
1054 | } | |
1055 | ||
1056 | return 0; | |
1057 | ||
1058 | drop: | |
0283328e | 1059 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
db8dac20 DM |
1060 | kfree_skb(skb); |
1061 | return -1; | |
1062 | } | |
1063 | ||
1064 | /* | |
1065 | * Multicasts and broadcasts go to each listener. | |
1066 | * | |
1067 | * Note: called only from the BH handler context, | |
1068 | * so we don't need to lock the hashes. | |
1069 | */ | |
e3163493 | 1070 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1071 | struct udphdr *uh, |
1072 | __be32 saddr, __be32 daddr, | |
1073 | struct hlist_head udptable[]) | |
1074 | { | |
1075 | struct sock *sk; | |
1076 | int dif; | |
1077 | ||
1078 | read_lock(&udp_hash_lock); | |
19c7578f | 1079 | sk = sk_head(&udptable[udp_hashfn(net, ntohs(uh->dest))]); |
db8dac20 DM |
1080 | dif = skb->dev->ifindex; |
1081 | sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif); | |
1082 | if (sk) { | |
1083 | struct sock *sknext = NULL; | |
1084 | ||
1085 | do { | |
1086 | struct sk_buff *skb1 = skb; | |
1087 | ||
1088 | sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr, | |
1089 | uh->source, saddr, dif); | |
1090 | if (sknext) | |
1091 | skb1 = skb_clone(skb, GFP_ATOMIC); | |
1092 | ||
1093 | if (skb1) { | |
1094 | int ret = 0; | |
1095 | ||
1096 | bh_lock_sock_nested(sk); | |
1097 | if (!sock_owned_by_user(sk)) | |
1098 | ret = udp_queue_rcv_skb(sk, skb1); | |
1099 | else | |
1100 | sk_add_backlog(sk, skb1); | |
1101 | bh_unlock_sock(sk); | |
1102 | ||
1103 | if (ret > 0) | |
1104 | /* we should probably re-process instead | |
1105 | * of dropping packets here. */ | |
1106 | kfree_skb(skb1); | |
1107 | } | |
1108 | sk = sknext; | |
1109 | } while (sknext); | |
1110 | } else | |
1111 | kfree_skb(skb); | |
1112 | read_unlock(&udp_hash_lock); | |
1113 | return 0; | |
1114 | } | |
1115 | ||
1116 | /* Initialize UDP checksum. If exited with zero value (success), | |
1117 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1118 | * Otherwise, csum completion requires chacksumming packet body, | |
1119 | * including udp header and folding it to skb->csum. | |
1120 | */ | |
1121 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1122 | int proto) | |
1123 | { | |
1124 | const struct iphdr *iph; | |
1125 | int err; | |
1126 | ||
1127 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1128 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1129 | ||
1130 | if (proto == IPPROTO_UDPLITE) { | |
1131 | err = udplite_checksum_init(skb, uh); | |
1132 | if (err) | |
1133 | return err; | |
1134 | } | |
1135 | ||
1136 | iph = ip_hdr(skb); | |
1137 | if (uh->check == 0) { | |
1138 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1139 | } else if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
1140 | if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, | |
1141 | proto, skb->csum)) | |
1142 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1143 | } | |
1144 | if (!skb_csum_unnecessary(skb)) | |
1145 | skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, | |
1146 | skb->len, proto, 0); | |
1147 | /* Probably, we should checksum udp header (it should be in cache | |
1148 | * in any case) and data in tiny packets (< rx copybreak). | |
1149 | */ | |
1150 | ||
1151 | return 0; | |
1152 | } | |
1153 | ||
1154 | /* | |
1155 | * All we need to do is get the socket, and then do a checksum. | |
1156 | */ | |
1157 | ||
1158 | int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[], | |
1159 | int proto) | |
1160 | { | |
1161 | struct sock *sk; | |
1162 | struct udphdr *uh = udp_hdr(skb); | |
1163 | unsigned short ulen; | |
1164 | struct rtable *rt = (struct rtable*)skb->dst; | |
1165 | __be32 saddr = ip_hdr(skb)->saddr; | |
1166 | __be32 daddr = ip_hdr(skb)->daddr; | |
0283328e | 1167 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1168 | |
1169 | /* | |
1170 | * Validate the packet. | |
1171 | */ | |
1172 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1173 | goto drop; /* No space for header. */ | |
1174 | ||
1175 | ulen = ntohs(uh->len); | |
1176 | if (ulen > skb->len) | |
1177 | goto short_packet; | |
1178 | ||
1179 | if (proto == IPPROTO_UDP) { | |
1180 | /* UDP validates ulen. */ | |
1181 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1182 | goto short_packet; | |
1183 | uh = udp_hdr(skb); | |
1184 | } | |
1185 | ||
1186 | if (udp4_csum_init(skb, uh, proto)) | |
1187 | goto csum_error; | |
1188 | ||
1189 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) | |
e3163493 PE |
1190 | return __udp4_lib_mcast_deliver(net, skb, uh, |
1191 | saddr, daddr, udptable); | |
db8dac20 | 1192 | |
e3163493 | 1193 | sk = __udp4_lib_lookup(net, saddr, uh->source, daddr, |
db8dac20 DM |
1194 | uh->dest, inet_iif(skb), udptable); |
1195 | ||
1196 | if (sk != NULL) { | |
1197 | int ret = 0; | |
1198 | bh_lock_sock_nested(sk); | |
1199 | if (!sock_owned_by_user(sk)) | |
1200 | ret = udp_queue_rcv_skb(sk, skb); | |
1201 | else | |
1202 | sk_add_backlog(sk, skb); | |
1203 | bh_unlock_sock(sk); | |
1204 | sock_put(sk); | |
1205 | ||
1206 | /* a return value > 0 means to resubmit the input, but | |
1207 | * it wants the return to be -protocol, or 0 | |
1208 | */ | |
1209 | if (ret > 0) | |
1210 | return -ret; | |
1211 | return 0; | |
1212 | } | |
1213 | ||
1214 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1215 | goto drop; | |
1216 | nf_reset(skb); | |
1217 | ||
1218 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1219 | if (udp_lib_checksum_complete(skb)) | |
1220 | goto csum_error; | |
1221 | ||
0283328e | 1222 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1223 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1224 | ||
1225 | /* | |
1226 | * Hmm. We got an UDP packet to a port to which we | |
1227 | * don't wanna listen. Ignore it. | |
1228 | */ | |
1229 | kfree_skb(skb); | |
1230 | return 0; | |
1231 | ||
1232 | short_packet: | |
a7d632b6 | 1233 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From " NIPQUAD_FMT ":%u %d/%d to " NIPQUAD_FMT ":%u\n", |
db8dac20 DM |
1234 | proto == IPPROTO_UDPLITE ? "-Lite" : "", |
1235 | NIPQUAD(saddr), | |
1236 | ntohs(uh->source), | |
1237 | ulen, | |
1238 | skb->len, | |
1239 | NIPQUAD(daddr), | |
1240 | ntohs(uh->dest)); | |
1241 | goto drop; | |
1242 | ||
1243 | csum_error: | |
1244 | /* | |
1245 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1246 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1247 | */ | |
a7d632b6 | 1248 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From " NIPQUAD_FMT ":%u to " NIPQUAD_FMT ":%u ulen %d\n", |
db8dac20 DM |
1249 | proto == IPPROTO_UDPLITE ? "-Lite" : "", |
1250 | NIPQUAD(saddr), | |
1251 | ntohs(uh->source), | |
1252 | NIPQUAD(daddr), | |
1253 | ntohs(uh->dest), | |
1254 | ulen); | |
1255 | drop: | |
0283328e | 1256 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1257 | kfree_skb(skb); |
1258 | return 0; | |
1259 | } | |
1260 | ||
1261 | int udp_rcv(struct sk_buff *skb) | |
1262 | { | |
1263 | return __udp4_lib_rcv(skb, udp_hash, IPPROTO_UDP); | |
1264 | } | |
1265 | ||
7d06b2e0 | 1266 | void udp_destroy_sock(struct sock *sk) |
db8dac20 DM |
1267 | { |
1268 | lock_sock(sk); | |
1269 | udp_flush_pending_frames(sk); | |
1270 | release_sock(sk); | |
db8dac20 DM |
1271 | } |
1272 | ||
1da177e4 LT |
1273 | /* |
1274 | * Socket option code for UDP | |
1275 | */ | |
4c0a6cb0 GR |
1276 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
1277 | char __user *optval, int optlen, | |
1278 | int (*push_pending_frames)(struct sock *)) | |
1da177e4 LT |
1279 | { |
1280 | struct udp_sock *up = udp_sk(sk); | |
1281 | int val; | |
1282 | int err = 0; | |
b2bf1e26 | 1283 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 1284 | |
6516c655 | 1285 | if (optlen<sizeof(int)) |
1da177e4 LT |
1286 | return -EINVAL; |
1287 | ||
1288 | if (get_user(val, (int __user *)optval)) | |
1289 | return -EFAULT; | |
1290 | ||
6516c655 | 1291 | switch (optname) { |
1da177e4 LT |
1292 | case UDP_CORK: |
1293 | if (val != 0) { | |
1294 | up->corkflag = 1; | |
1295 | } else { | |
1296 | up->corkflag = 0; | |
1297 | lock_sock(sk); | |
4c0a6cb0 | 1298 | (*push_pending_frames)(sk); |
1da177e4 LT |
1299 | release_sock(sk); |
1300 | } | |
1301 | break; | |
e905a9ed | 1302 | |
1da177e4 LT |
1303 | case UDP_ENCAP: |
1304 | switch (val) { | |
1305 | case 0: | |
1306 | case UDP_ENCAP_ESPINUDP: | |
1307 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
1308 | up->encap_rcv = xfrm4_udp_encap_rcv; |
1309 | /* FALLTHROUGH */ | |
342f0234 | 1310 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 LT |
1311 | up->encap_type = val; |
1312 | break; | |
1313 | default: | |
1314 | err = -ENOPROTOOPT; | |
1315 | break; | |
1316 | } | |
1317 | break; | |
1318 | ||
ba4e58ec GR |
1319 | /* |
1320 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
1321 | */ | |
1322 | /* The sender sets actual checksum coverage length via this option. | |
1323 | * The case coverage > packet length is handled by send module. */ | |
1324 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 1325 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1326 | return -ENOPROTOOPT; |
1327 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
1328 | val = 8; | |
1329 | up->pcslen = val; | |
1330 | up->pcflag |= UDPLITE_SEND_CC; | |
1331 | break; | |
1332 | ||
e905a9ed YH |
1333 | /* The receiver specifies a minimum checksum coverage value. To make |
1334 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
1335 | * used, this again means full checksum coverage. */ |
1336 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 1337 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1338 | return -ENOPROTOOPT; |
1339 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
1340 | val = 8; | |
1341 | up->pcrlen = val; | |
1342 | up->pcflag |= UDPLITE_RECV_CC; | |
1343 | break; | |
1344 | ||
1da177e4 LT |
1345 | default: |
1346 | err = -ENOPROTOOPT; | |
1347 | break; | |
6516c655 | 1348 | } |
1da177e4 LT |
1349 | |
1350 | return err; | |
1351 | } | |
1352 | ||
db8dac20 DM |
1353 | int udp_setsockopt(struct sock *sk, int level, int optname, |
1354 | char __user *optval, int optlen) | |
1355 | { | |
1356 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1357 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1358 | udp_push_pending_frames); | |
1359 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
1360 | } | |
1361 | ||
1362 | #ifdef CONFIG_COMPAT | |
1363 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
1364 | char __user *optval, int optlen) | |
1365 | { | |
1366 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1367 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1368 | udp_push_pending_frames); | |
1369 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
1370 | } | |
1371 | #endif | |
1372 | ||
4c0a6cb0 GR |
1373 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
1374 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
1375 | { |
1376 | struct udp_sock *up = udp_sk(sk); | |
1377 | int val, len; | |
1378 | ||
6516c655 | 1379 | if (get_user(len,optlen)) |
1da177e4 LT |
1380 | return -EFAULT; |
1381 | ||
1382 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 1383 | |
6516c655 | 1384 | if (len < 0) |
1da177e4 LT |
1385 | return -EINVAL; |
1386 | ||
6516c655 | 1387 | switch (optname) { |
1da177e4 LT |
1388 | case UDP_CORK: |
1389 | val = up->corkflag; | |
1390 | break; | |
1391 | ||
1392 | case UDP_ENCAP: | |
1393 | val = up->encap_type; | |
1394 | break; | |
1395 | ||
ba4e58ec GR |
1396 | /* The following two cannot be changed on UDP sockets, the return is |
1397 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
1398 | case UDPLITE_SEND_CSCOV: | |
1399 | val = up->pcslen; | |
1400 | break; | |
1401 | ||
1402 | case UDPLITE_RECV_CSCOV: | |
1403 | val = up->pcrlen; | |
1404 | break; | |
1405 | ||
1da177e4 LT |
1406 | default: |
1407 | return -ENOPROTOOPT; | |
6516c655 | 1408 | } |
1da177e4 | 1409 | |
6516c655 | 1410 | if (put_user(len, optlen)) |
e905a9ed | 1411 | return -EFAULT; |
6516c655 | 1412 | if (copy_to_user(optval, &val,len)) |
1da177e4 | 1413 | return -EFAULT; |
e905a9ed | 1414 | return 0; |
1da177e4 LT |
1415 | } |
1416 | ||
db8dac20 DM |
1417 | int udp_getsockopt(struct sock *sk, int level, int optname, |
1418 | char __user *optval, int __user *optlen) | |
1419 | { | |
1420 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1421 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1422 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
1423 | } | |
1424 | ||
1425 | #ifdef CONFIG_COMPAT | |
1426 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
1427 | char __user *optval, int __user *optlen) | |
1428 | { | |
1429 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1430 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1431 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
1432 | } | |
1433 | #endif | |
1da177e4 LT |
1434 | /** |
1435 | * udp_poll - wait for a UDP event. | |
1436 | * @file - file struct | |
1437 | * @sock - socket | |
1438 | * @wait - poll table | |
1439 | * | |
e905a9ed | 1440 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
1441 | * blocking sockets. If application is using a blocking fd |
1442 | * and a packet with checksum error is in the queue; | |
1443 | * then it could get return from select indicating data available | |
1444 | * but then block when reading it. Add special case code | |
1445 | * to work around these arguably broken applications. | |
1446 | */ | |
1447 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
1448 | { | |
1449 | unsigned int mask = datagram_poll(file, sock, wait); | |
1450 | struct sock *sk = sock->sk; | |
ba4e58ec GR |
1451 | int is_lite = IS_UDPLITE(sk); |
1452 | ||
1da177e4 LT |
1453 | /* Check for false positives due to checksum errors */ |
1454 | if ( (mask & POLLRDNORM) && | |
1455 | !(file->f_flags & O_NONBLOCK) && | |
1456 | !(sk->sk_shutdown & RCV_SHUTDOWN)){ | |
1457 | struct sk_buff_head *rcvq = &sk->sk_receive_queue; | |
1458 | struct sk_buff *skb; | |
1459 | ||
208d8984 | 1460 | spin_lock_bh(&rcvq->lock); |
759e5d00 HX |
1461 | while ((skb = skb_peek(rcvq)) != NULL && |
1462 | udp_lib_checksum_complete(skb)) { | |
0283328e PE |
1463 | UDP_INC_STATS_BH(sock_net(sk), |
1464 | UDP_MIB_INERRORS, is_lite); | |
759e5d00 HX |
1465 | __skb_unlink(skb, rcvq); |
1466 | kfree_skb(skb); | |
1da177e4 | 1467 | } |
208d8984 | 1468 | spin_unlock_bh(&rcvq->lock); |
1da177e4 LT |
1469 | |
1470 | /* nothing to see, move along */ | |
1471 | if (skb == NULL) | |
1472 | mask &= ~(POLLIN | POLLRDNORM); | |
1473 | } | |
1474 | ||
1475 | return mask; | |
e905a9ed | 1476 | |
1da177e4 LT |
1477 | } |
1478 | ||
db8dac20 DM |
1479 | struct proto udp_prot = { |
1480 | .name = "UDP", | |
1481 | .owner = THIS_MODULE, | |
1482 | .close = udp_lib_close, | |
1483 | .connect = ip4_datagram_connect, | |
1484 | .disconnect = udp_disconnect, | |
1485 | .ioctl = udp_ioctl, | |
1486 | .destroy = udp_destroy_sock, | |
1487 | .setsockopt = udp_setsockopt, | |
1488 | .getsockopt = udp_getsockopt, | |
1489 | .sendmsg = udp_sendmsg, | |
1490 | .recvmsg = udp_recvmsg, | |
1491 | .sendpage = udp_sendpage, | |
1492 | .backlog_rcv = udp_queue_rcv_skb, | |
1493 | .hash = udp_lib_hash, | |
1494 | .unhash = udp_lib_unhash, | |
1495 | .get_port = udp_v4_get_port, | |
1496 | .memory_allocated = &udp_memory_allocated, | |
1497 | .sysctl_mem = sysctl_udp_mem, | |
1498 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
1499 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
1500 | .obj_size = sizeof(struct udp_sock), | |
6ba5a3c5 | 1501 | .h.udp_hash = udp_hash, |
db8dac20 DM |
1502 | #ifdef CONFIG_COMPAT |
1503 | .compat_setsockopt = compat_udp_setsockopt, | |
1504 | .compat_getsockopt = compat_udp_getsockopt, | |
1505 | #endif | |
db8dac20 | 1506 | }; |
1da177e4 LT |
1507 | |
1508 | /* ------------------------------------------------------------------------ */ | |
1509 | #ifdef CONFIG_PROC_FS | |
1510 | ||
1511 | static struct sock *udp_get_first(struct seq_file *seq) | |
1512 | { | |
1513 | struct sock *sk; | |
1514 | struct udp_iter_state *state = seq->private; | |
6f191efe | 1515 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1516 | |
1517 | for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) { | |
1518 | struct hlist_node *node; | |
ba4e58ec | 1519 | sk_for_each(sk, node, state->hashtable + state->bucket) { |
878628fb | 1520 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 1521 | continue; |
1da177e4 LT |
1522 | if (sk->sk_family == state->family) |
1523 | goto found; | |
1524 | } | |
1525 | } | |
1526 | sk = NULL; | |
1527 | found: | |
1528 | return sk; | |
1529 | } | |
1530 | ||
1531 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
1532 | { | |
1533 | struct udp_iter_state *state = seq->private; | |
6f191efe | 1534 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1535 | |
1536 | do { | |
1537 | sk = sk_next(sk); | |
1538 | try_again: | |
1539 | ; | |
878628fb | 1540 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 LT |
1541 | |
1542 | if (!sk && ++state->bucket < UDP_HTABLE_SIZE) { | |
ba4e58ec | 1543 | sk = sk_head(state->hashtable + state->bucket); |
1da177e4 LT |
1544 | goto try_again; |
1545 | } | |
1546 | return sk; | |
1547 | } | |
1548 | ||
1549 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
1550 | { | |
1551 | struct sock *sk = udp_get_first(seq); | |
1552 | ||
1553 | if (sk) | |
6516c655 | 1554 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
1555 | --pos; |
1556 | return pos ? NULL : sk; | |
1557 | } | |
1558 | ||
1559 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
9a429c49 | 1560 | __acquires(udp_hash_lock) |
1da177e4 LT |
1561 | { |
1562 | read_lock(&udp_hash_lock); | |
b50660f1 | 1563 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
1564 | } |
1565 | ||
1566 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
1567 | { | |
1568 | struct sock *sk; | |
1569 | ||
b50660f1 | 1570 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
1571 | sk = udp_get_idx(seq, 0); |
1572 | else | |
1573 | sk = udp_get_next(seq, v); | |
1574 | ||
1575 | ++*pos; | |
1576 | return sk; | |
1577 | } | |
1578 | ||
1579 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
9a429c49 | 1580 | __releases(udp_hash_lock) |
1da177e4 LT |
1581 | { |
1582 | read_unlock(&udp_hash_lock); | |
1583 | } | |
1584 | ||
1585 | static int udp_seq_open(struct inode *inode, struct file *file) | |
1586 | { | |
1587 | struct udp_seq_afinfo *afinfo = PDE(inode)->data; | |
a2be75c1 DL |
1588 | struct udp_iter_state *s; |
1589 | int err; | |
a91275ef | 1590 | |
a2be75c1 DL |
1591 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
1592 | sizeof(struct udp_iter_state)); | |
1593 | if (err < 0) | |
1594 | return err; | |
a91275ef | 1595 | |
a2be75c1 | 1596 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 1597 | s->family = afinfo->family; |
ba4e58ec | 1598 | s->hashtable = afinfo->hashtable; |
a2be75c1 | 1599 | return err; |
a91275ef DL |
1600 | } |
1601 | ||
1da177e4 | 1602 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 1603 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
1604 | { |
1605 | struct proc_dir_entry *p; | |
1606 | int rc = 0; | |
1607 | ||
3ba9441b DL |
1608 | afinfo->seq_fops.open = udp_seq_open; |
1609 | afinfo->seq_fops.read = seq_read; | |
1610 | afinfo->seq_fops.llseek = seq_lseek; | |
1611 | afinfo->seq_fops.release = seq_release_net; | |
1da177e4 | 1612 | |
dda61925 DL |
1613 | afinfo->seq_ops.start = udp_seq_start; |
1614 | afinfo->seq_ops.next = udp_seq_next; | |
1615 | afinfo->seq_ops.stop = udp_seq_stop; | |
1616 | ||
84841c3c DL |
1617 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
1618 | &afinfo->seq_fops, afinfo); | |
1619 | if (!p) | |
1da177e4 LT |
1620 | rc = -ENOMEM; |
1621 | return rc; | |
1622 | } | |
1623 | ||
0c96d8c5 | 1624 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 1625 | { |
0c96d8c5 | 1626 | proc_net_remove(net, afinfo->name); |
1da177e4 | 1627 | } |
db8dac20 DM |
1628 | |
1629 | /* ------------------------------------------------------------------------ */ | |
5e659e4c PE |
1630 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
1631 | int bucket, int *len) | |
db8dac20 DM |
1632 | { |
1633 | struct inet_sock *inet = inet_sk(sp); | |
1634 | __be32 dest = inet->daddr; | |
1635 | __be32 src = inet->rcv_saddr; | |
1636 | __u16 destp = ntohs(inet->dport); | |
1637 | __u16 srcp = ntohs(inet->sport); | |
1638 | ||
5e659e4c | 1639 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
cb61cb9b | 1640 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d%n", |
db8dac20 DM |
1641 | bucket, src, srcp, dest, destp, sp->sk_state, |
1642 | atomic_read(&sp->sk_wmem_alloc), | |
1643 | atomic_read(&sp->sk_rmem_alloc), | |
1644 | 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp), | |
cb61cb9b ED |
1645 | atomic_read(&sp->sk_refcnt), sp, |
1646 | atomic_read(&sp->sk_drops), len); | |
db8dac20 DM |
1647 | } |
1648 | ||
1649 | int udp4_seq_show(struct seq_file *seq, void *v) | |
1650 | { | |
1651 | if (v == SEQ_START_TOKEN) | |
1652 | seq_printf(seq, "%-127s\n", | |
1653 | " sl local_address rem_address st tx_queue " | |
1654 | "rx_queue tr tm->when retrnsmt uid timeout " | |
cb61cb9b | 1655 | "inode ref pointer drops"); |
db8dac20 | 1656 | else { |
db8dac20 | 1657 | struct udp_iter_state *state = seq->private; |
5e659e4c | 1658 | int len; |
db8dac20 | 1659 | |
5e659e4c PE |
1660 | udp4_format_sock(v, seq, state->bucket, &len); |
1661 | seq_printf(seq, "%*s\n", 127 - len ,""); | |
db8dac20 DM |
1662 | } |
1663 | return 0; | |
1664 | } | |
1665 | ||
1666 | /* ------------------------------------------------------------------------ */ | |
db8dac20 | 1667 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
1668 | .name = "udp", |
1669 | .family = AF_INET, | |
1670 | .hashtable = udp_hash, | |
4ad96d39 DL |
1671 | .seq_fops = { |
1672 | .owner = THIS_MODULE, | |
1673 | }, | |
dda61925 DL |
1674 | .seq_ops = { |
1675 | .show = udp4_seq_show, | |
1676 | }, | |
db8dac20 DM |
1677 | }; |
1678 | ||
15439feb PE |
1679 | static int udp4_proc_init_net(struct net *net) |
1680 | { | |
1681 | return udp_proc_register(net, &udp4_seq_afinfo); | |
1682 | } | |
1683 | ||
1684 | static void udp4_proc_exit_net(struct net *net) | |
1685 | { | |
1686 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
1687 | } | |
1688 | ||
1689 | static struct pernet_operations udp4_net_ops = { | |
1690 | .init = udp4_proc_init_net, | |
1691 | .exit = udp4_proc_exit_net, | |
1692 | }; | |
1693 | ||
db8dac20 DM |
1694 | int __init udp4_proc_init(void) |
1695 | { | |
15439feb | 1696 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
1697 | } |
1698 | ||
1699 | void udp4_proc_exit(void) | |
1700 | { | |
15439feb | 1701 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 1702 | } |
1da177e4 LT |
1703 | #endif /* CONFIG_PROC_FS */ |
1704 | ||
95766fff HA |
1705 | void __init udp_init(void) |
1706 | { | |
1707 | unsigned long limit; | |
1708 | ||
1709 | /* Set the pressure threshold up by the same strategy of TCP. It is a | |
1710 | * fraction of global memory that is up to 1/2 at 256 MB, decreasing | |
1711 | * toward zero with the amount of memory, with a floor of 128 pages. | |
1712 | */ | |
1713 | limit = min(nr_all_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT); | |
1714 | limit = (limit * (nr_all_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11); | |
1715 | limit = max(limit, 128UL); | |
1716 | sysctl_udp_mem[0] = limit / 4 * 3; | |
1717 | sysctl_udp_mem[1] = limit; | |
1718 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
1719 | ||
1720 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
1721 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
1722 | } | |
1723 | ||
1da177e4 LT |
1724 | EXPORT_SYMBOL(udp_disconnect); |
1725 | EXPORT_SYMBOL(udp_hash); | |
1726 | EXPORT_SYMBOL(udp_hash_lock); | |
1727 | EXPORT_SYMBOL(udp_ioctl); | |
db8dac20 DM |
1728 | EXPORT_SYMBOL(udp_prot); |
1729 | EXPORT_SYMBOL(udp_sendmsg); | |
4c0a6cb0 GR |
1730 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1731 | EXPORT_SYMBOL(udp_lib_setsockopt); | |
1da177e4 | 1732 | EXPORT_SYMBOL(udp_poll); |
6ba5a3c5 | 1733 | EXPORT_SYMBOL(udp_lib_get_port); |
1da177e4 LT |
1734 | |
1735 | #ifdef CONFIG_PROC_FS | |
1736 | EXPORT_SYMBOL(udp_proc_register); | |
1737 | EXPORT_SYMBOL(udp_proc_unregister); | |
1738 | #endif |