]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * The User Datagram Protocol (UDP). | |
7 | * | |
02c30a84 | 8 | * Authors: Ross Biro |
1da177e4 LT |
9 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
10 | * Arnt Gulbrandsen, <agulbra@nvg.unit.no> | |
113aa838 | 11 | * Alan Cox, <alan@lxorguk.ukuu.org.uk> |
1da177e4 LT |
12 | * Hirokazu Takahashi, <taka@valinux.co.jp> |
13 | * | |
14 | * Fixes: | |
15 | * Alan Cox : verify_area() calls | |
16 | * Alan Cox : stopped close while in use off icmp | |
17 | * messages. Not a fix but a botch that | |
18 | * for udp at least is 'valid'. | |
19 | * Alan Cox : Fixed icmp handling properly | |
20 | * Alan Cox : Correct error for oversized datagrams | |
e905a9ed YH |
21 | * Alan Cox : Tidied select() semantics. |
22 | * Alan Cox : udp_err() fixed properly, also now | |
1da177e4 LT |
23 | * select and read wake correctly on errors |
24 | * Alan Cox : udp_send verify_area moved to avoid mem leak | |
25 | * Alan Cox : UDP can count its memory | |
26 | * Alan Cox : send to an unknown connection causes | |
27 | * an ECONNREFUSED off the icmp, but | |
28 | * does NOT close. | |
29 | * Alan Cox : Switched to new sk_buff handlers. No more backlog! | |
30 | * Alan Cox : Using generic datagram code. Even smaller and the PEEK | |
31 | * bug no longer crashes it. | |
32 | * Fred Van Kempen : Net2e support for sk->broadcast. | |
33 | * Alan Cox : Uses skb_free_datagram | |
34 | * Alan Cox : Added get/set sockopt support. | |
35 | * Alan Cox : Broadcasting without option set returns EACCES. | |
36 | * Alan Cox : No wakeup calls. Instead we now use the callbacks. | |
37 | * Alan Cox : Use ip_tos and ip_ttl | |
38 | * Alan Cox : SNMP Mibs | |
39 | * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support. | |
40 | * Matt Dillon : UDP length checks. | |
41 | * Alan Cox : Smarter af_inet used properly. | |
42 | * Alan Cox : Use new kernel side addressing. | |
43 | * Alan Cox : Incorrect return on truncated datagram receive. | |
44 | * Arnt Gulbrandsen : New udp_send and stuff | |
45 | * Alan Cox : Cache last socket | |
46 | * Alan Cox : Route cache | |
47 | * Jon Peatfield : Minor efficiency fix to sendto(). | |
48 | * Mike Shaver : RFC1122 checks. | |
49 | * Alan Cox : Nonblocking error fix. | |
50 | * Willy Konynenberg : Transparent proxying support. | |
51 | * Mike McLagan : Routing by source | |
52 | * David S. Miller : New socket lookup architecture. | |
53 | * Last socket cache retained as it | |
54 | * does have a high hit rate. | |
55 | * Olaf Kirch : Don't linearise iovec on sendmsg. | |
56 | * Andi Kleen : Some cleanups, cache destination entry | |
e905a9ed | 57 | * for connect. |
1da177e4 LT |
58 | * Vitaly E. Lavrov : Transparent proxy revived after year coma. |
59 | * Melvin Smith : Check msg_name not msg_namelen in sendto(), | |
60 | * return ENOTCONN for unconnected sockets (POSIX) | |
61 | * Janos Farkas : don't deliver multi/broadcasts to a different | |
62 | * bound-to-device socket | |
63 | * Hirokazu Takahashi : HW checksumming for outgoing UDP | |
64 | * datagrams. | |
65 | * Hirokazu Takahashi : sendfile() on UDP works now. | |
66 | * Arnaldo C. Melo : convert /proc/net/udp to seq_file | |
67 | * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which | |
68 | * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind | |
69 | * a single port at the same time. | |
70 | * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support | |
342f0234 | 71 | * James Chapman : Add L2TP encapsulation type. |
1da177e4 LT |
72 | * |
73 | * | |
74 | * This program is free software; you can redistribute it and/or | |
75 | * modify it under the terms of the GNU General Public License | |
76 | * as published by the Free Software Foundation; either version | |
77 | * 2 of the License, or (at your option) any later version. | |
78 | */ | |
e905a9ed | 79 | |
1da177e4 LT |
80 | #include <asm/system.h> |
81 | #include <asm/uaccess.h> | |
82 | #include <asm/ioctls.h> | |
95766fff | 83 | #include <linux/bootmem.h> |
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 | 106 | |
645ca708 ED |
107 | struct udp_table udp_table; |
108 | EXPORT_SYMBOL(udp_table); | |
1da177e4 | 109 | |
95766fff HA |
110 | int sysctl_udp_mem[3] __read_mostly; |
111 | int sysctl_udp_rmem_min __read_mostly; | |
112 | int sysctl_udp_wmem_min __read_mostly; | |
113 | ||
114 | EXPORT_SYMBOL(sysctl_udp_mem); | |
115 | EXPORT_SYMBOL(sysctl_udp_rmem_min); | |
116 | EXPORT_SYMBOL(sysctl_udp_wmem_min); | |
117 | ||
118 | atomic_t udp_memory_allocated; | |
119 | EXPORT_SYMBOL(udp_memory_allocated); | |
120 | ||
f24d43c0 | 121 | static int udp_lib_lport_inuse(struct net *net, __u16 num, |
645ca708 | 122 | const struct udp_hslot *hslot, |
f24d43c0 ED |
123 | struct sock *sk, |
124 | int (*saddr_comp)(const struct sock *sk1, | |
125 | const struct sock *sk2)) | |
1da177e4 | 126 | { |
f24d43c0 | 127 | struct sock *sk2; |
1da177e4 | 128 | struct hlist_node *node; |
25030a7f | 129 | |
645ca708 | 130 | sk_for_each(sk2, node, &hslot->head) |
f24d43c0 ED |
131 | if (net_eq(sock_net(sk2), net) && |
132 | sk2 != sk && | |
133 | sk2->sk_hash == num && | |
134 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
135 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if | |
136 | || sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
137 | (*saddr_comp)(sk, sk2)) | |
25030a7f GR |
138 | return 1; |
139 | return 0; | |
140 | } | |
141 | ||
142 | /** | |
6ba5a3c5 | 143 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
144 | * |
145 | * @sk: socket struct in question | |
146 | * @snum: port number to look up | |
df2bc459 | 147 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25030a7f | 148 | */ |
6ba5a3c5 | 149 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
df2bc459 DM |
150 | int (*saddr_comp)(const struct sock *sk1, |
151 | const struct sock *sk2 ) ) | |
25030a7f | 152 | { |
645ca708 ED |
153 | struct udp_hslot *hslot; |
154 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
25030a7f | 155 | int error = 1; |
3b1e0a65 | 156 | struct net *net = sock_net(sk); |
1da177e4 | 157 | |
32c1da70 | 158 | if (!snum) { |
9088c560 ED |
159 | int low, high, remaining; |
160 | unsigned rand; | |
161 | unsigned short first; | |
32c1da70 | 162 | |
227b60f5 | 163 | inet_get_local_port_range(&low, &high); |
a25de534 | 164 | remaining = (high - low) + 1; |
227b60f5 | 165 | |
9088c560 ED |
166 | rand = net_random(); |
167 | snum = first = rand % remaining + low; | |
168 | rand |= 1; | |
645ca708 ED |
169 | for (;;) { |
170 | hslot = &udptable->hash[udp_hashfn(net, snum)]; | |
171 | spin_lock_bh(&hslot->lock); | |
172 | if (!udp_lib_lport_inuse(net, snum, hslot, sk, saddr_comp)) | |
173 | break; | |
174 | spin_unlock_bh(&hslot->lock); | |
9088c560 ED |
175 | do { |
176 | snum = snum + rand; | |
177 | } while (snum < low || snum > high); | |
178 | if (snum == first) | |
179 | goto fail; | |
1da177e4 | 180 | } |
645ca708 ED |
181 | } else { |
182 | hslot = &udptable->hash[udp_hashfn(net, snum)]; | |
183 | spin_lock_bh(&hslot->lock); | |
184 | if (udp_lib_lport_inuse(net, snum, hslot, sk, saddr_comp)) | |
185 | goto fail_unlock; | |
186 | } | |
25030a7f | 187 | inet_sk(sk)->num = snum; |
df2bc459 | 188 | sk->sk_hash = snum; |
1da177e4 | 189 | if (sk_unhashed(sk)) { |
271b72c7 | 190 | sk_add_node_rcu(sk, &hslot->head); |
c29a0bc4 | 191 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
1da177e4 | 192 | } |
25030a7f | 193 | error = 0; |
645ca708 ED |
194 | fail_unlock: |
195 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 196 | fail: |
25030a7f GR |
197 | return error; |
198 | } | |
199 | ||
6ba5a3c5 | 200 | static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2) |
db8dac20 DM |
201 | { |
202 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
203 | ||
204 | return ( !ipv6_only_sock(sk2) && | |
205 | (!inet1->rcv_saddr || !inet2->rcv_saddr || | |
206 | inet1->rcv_saddr == inet2->rcv_saddr )); | |
207 | } | |
208 | ||
6ba5a3c5 | 209 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 210 | { |
6ba5a3c5 | 211 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal); |
db8dac20 DM |
212 | } |
213 | ||
645ca708 ED |
214 | static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr, |
215 | unsigned short hnum, | |
216 | __be16 sport, __be32 daddr, __be16 dport, int dif) | |
217 | { | |
218 | int score = -1; | |
219 | ||
220 | if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum && | |
221 | !ipv6_only_sock(sk)) { | |
222 | struct inet_sock *inet = inet_sk(sk); | |
223 | ||
224 | score = (sk->sk_family == PF_INET ? 1 : 0); | |
225 | if (inet->rcv_saddr) { | |
226 | if (inet->rcv_saddr != daddr) | |
227 | return -1; | |
228 | score += 2; | |
229 | } | |
230 | if (inet->daddr) { | |
231 | if (inet->daddr != saddr) | |
232 | return -1; | |
233 | score += 2; | |
234 | } | |
235 | if (inet->dport) { | |
236 | if (inet->dport != sport) | |
237 | return -1; | |
238 | score += 2; | |
239 | } | |
240 | if (sk->sk_bound_dev_if) { | |
241 | if (sk->sk_bound_dev_if != dif) | |
242 | return -1; | |
243 | score += 2; | |
244 | } | |
245 | } | |
246 | return score; | |
247 | } | |
248 | ||
db8dac20 DM |
249 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
250 | * harder than this. -DaveM | |
251 | */ | |
252 | static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, | |
253 | __be16 sport, __be32 daddr, __be16 dport, | |
645ca708 | 254 | int dif, struct udp_table *udptable) |
db8dac20 | 255 | { |
271b72c7 | 256 | struct sock *sk, *result; |
db8dac20 DM |
257 | struct hlist_node *node; |
258 | unsigned short hnum = ntohs(dport); | |
645ca708 ED |
259 | unsigned int hash = udp_hashfn(net, hnum); |
260 | struct udp_hslot *hslot = &udptable->hash[hash]; | |
271b72c7 | 261 | int score, badness; |
645ca708 | 262 | |
271b72c7 ED |
263 | rcu_read_lock(); |
264 | begin: | |
265 | result = NULL; | |
266 | badness = -1; | |
267 | sk_for_each_rcu(sk, node, &hslot->head) { | |
268 | /* | |
269 | * lockless reader, and SLAB_DESTROY_BY_RCU items: | |
270 | * We must check this item was not moved to another chain | |
271 | */ | |
272 | if (udp_hashfn(net, sk->sk_hash) != hash) | |
273 | goto begin; | |
645ca708 ED |
274 | score = compute_score(sk, net, saddr, hnum, sport, |
275 | daddr, dport, dif); | |
276 | if (score > badness) { | |
277 | result = sk; | |
278 | badness = score; | |
db8dac20 DM |
279 | } |
280 | } | |
271b72c7 ED |
281 | if (result) { |
282 | if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt))) | |
283 | result = NULL; | |
284 | else if (unlikely(compute_score(result, net, saddr, hnum, sport, | |
285 | daddr, dport, dif) < badness)) { | |
286 | sock_put(result); | |
287 | goto begin; | |
288 | } | |
289 | } | |
290 | rcu_read_unlock(); | |
db8dac20 DM |
291 | return result; |
292 | } | |
293 | ||
607c4aaf KK |
294 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
295 | __be16 sport, __be16 dport, | |
645ca708 | 296 | struct udp_table *udptable) |
607c4aaf | 297 | { |
23542618 | 298 | struct sock *sk; |
607c4aaf KK |
299 | const struct iphdr *iph = ip_hdr(skb); |
300 | ||
23542618 KK |
301 | if (unlikely(sk = skb_steal_sock(skb))) |
302 | return sk; | |
303 | else | |
304 | return __udp4_lib_lookup(dev_net(skb->dst->dev), iph->saddr, sport, | |
305 | iph->daddr, dport, inet_iif(skb), | |
306 | udptable); | |
607c4aaf KK |
307 | } |
308 | ||
bcd41303 KK |
309 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
310 | __be32 daddr, __be16 dport, int dif) | |
311 | { | |
645ca708 | 312 | return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table); |
bcd41303 KK |
313 | } |
314 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
315 | ||
db8dac20 DM |
316 | static inline struct sock *udp_v4_mcast_next(struct sock *sk, |
317 | __be16 loc_port, __be32 loc_addr, | |
318 | __be16 rmt_port, __be32 rmt_addr, | |
319 | int dif) | |
320 | { | |
321 | struct hlist_node *node; | |
322 | struct sock *s = sk; | |
323 | unsigned short hnum = ntohs(loc_port); | |
324 | ||
325 | sk_for_each_from(s, node) { | |
326 | struct inet_sock *inet = inet_sk(s); | |
327 | ||
328 | if (s->sk_hash != hnum || | |
329 | (inet->daddr && inet->daddr != rmt_addr) || | |
330 | (inet->dport != rmt_port && inet->dport) || | |
331 | (inet->rcv_saddr && inet->rcv_saddr != loc_addr) || | |
332 | ipv6_only_sock(s) || | |
333 | (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)) | |
334 | continue; | |
335 | if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif)) | |
336 | continue; | |
337 | goto found; | |
338 | } | |
339 | s = NULL; | |
340 | found: | |
341 | return s; | |
342 | } | |
343 | ||
344 | /* | |
345 | * This routine is called by the ICMP module when it gets some | |
346 | * sort of error condition. If err < 0 then the socket should | |
347 | * be closed and the error returned to the user. If err > 0 | |
348 | * it's just the icmp type << 8 | icmp code. | |
349 | * Header points to the ip header of the error packet. We move | |
350 | * on past this. Then (as it used to claim before adjustment) | |
351 | * header points to the first 8 bytes of the udp header. We need | |
352 | * to find the appropriate port. | |
353 | */ | |
354 | ||
645ca708 | 355 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
356 | { |
357 | struct inet_sock *inet; | |
358 | struct iphdr *iph = (struct iphdr*)skb->data; | |
359 | struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2)); | |
360 | const int type = icmp_hdr(skb)->type; | |
361 | const int code = icmp_hdr(skb)->code; | |
362 | struct sock *sk; | |
363 | int harderr; | |
364 | int err; | |
fd54d716 | 365 | struct net *net = dev_net(skb->dev); |
db8dac20 | 366 | |
fd54d716 | 367 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
db8dac20 DM |
368 | iph->saddr, uh->source, skb->dev->ifindex, udptable); |
369 | if (sk == NULL) { | |
dcfc23ca | 370 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
371 | return; /* No socket for error */ |
372 | } | |
373 | ||
374 | err = 0; | |
375 | harderr = 0; | |
376 | inet = inet_sk(sk); | |
377 | ||
378 | switch (type) { | |
379 | default: | |
380 | case ICMP_TIME_EXCEEDED: | |
381 | err = EHOSTUNREACH; | |
382 | break; | |
383 | case ICMP_SOURCE_QUENCH: | |
384 | goto out; | |
385 | case ICMP_PARAMETERPROB: | |
386 | err = EPROTO; | |
387 | harderr = 1; | |
388 | break; | |
389 | case ICMP_DEST_UNREACH: | |
390 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
391 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { | |
392 | err = EMSGSIZE; | |
393 | harderr = 1; | |
394 | break; | |
395 | } | |
396 | goto out; | |
397 | } | |
398 | err = EHOSTUNREACH; | |
399 | if (code <= NR_ICMP_UNREACH) { | |
400 | harderr = icmp_err_convert[code].fatal; | |
401 | err = icmp_err_convert[code].errno; | |
402 | } | |
403 | break; | |
404 | } | |
405 | ||
406 | /* | |
407 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
408 | * 4.1.3.3. | |
409 | */ | |
410 | if (!inet->recverr) { | |
411 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
412 | goto out; | |
413 | } else { | |
414 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1)); | |
415 | } | |
416 | sk->sk_err = err; | |
417 | sk->sk_error_report(sk); | |
418 | out: | |
419 | sock_put(sk); | |
420 | } | |
421 | ||
422 | void udp_err(struct sk_buff *skb, u32 info) | |
423 | { | |
645ca708 | 424 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
425 | } |
426 | ||
427 | /* | |
428 | * Throw away all pending data and cancel the corking. Socket is locked. | |
429 | */ | |
36d926b9 | 430 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
431 | { |
432 | struct udp_sock *up = udp_sk(sk); | |
433 | ||
434 | if (up->pending) { | |
435 | up->len = 0; | |
436 | up->pending = 0; | |
437 | ip_flush_pending_frames(sk); | |
438 | } | |
439 | } | |
36d926b9 | 440 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
441 | |
442 | /** | |
443 | * udp4_hwcsum_outgoing - handle outgoing HW checksumming | |
444 | * @sk: socket we are sending on | |
445 | * @skb: sk_buff containing the filled-in UDP header | |
446 | * (checksum field must be zeroed out) | |
447 | */ | |
448 | static void udp4_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb, | |
449 | __be32 src, __be32 dst, int len ) | |
450 | { | |
451 | unsigned int offset; | |
452 | struct udphdr *uh = udp_hdr(skb); | |
453 | __wsum csum = 0; | |
454 | ||
455 | if (skb_queue_len(&sk->sk_write_queue) == 1) { | |
456 | /* | |
457 | * Only one fragment on the socket. | |
458 | */ | |
459 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
460 | skb->csum_offset = offsetof(struct udphdr, check); | |
461 | uh->check = ~csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, 0); | |
462 | } else { | |
463 | /* | |
464 | * HW-checksum won't work as there are two or more | |
465 | * fragments on the socket so that all csums of sk_buffs | |
466 | * should be together | |
467 | */ | |
468 | offset = skb_transport_offset(skb); | |
469 | skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); | |
470 | ||
471 | skb->ip_summed = CHECKSUM_NONE; | |
472 | ||
473 | skb_queue_walk(&sk->sk_write_queue, skb) { | |
474 | csum = csum_add(csum, skb->csum); | |
475 | } | |
476 | ||
477 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); | |
478 | if (uh->check == 0) | |
479 | uh->check = CSUM_MANGLED_0; | |
480 | } | |
481 | } | |
482 | ||
483 | /* | |
484 | * Push out all pending data as one UDP datagram. Socket is locked. | |
485 | */ | |
486 | static int udp_push_pending_frames(struct sock *sk) | |
487 | { | |
488 | struct udp_sock *up = udp_sk(sk); | |
489 | struct inet_sock *inet = inet_sk(sk); | |
490 | struct flowi *fl = &inet->cork.fl; | |
491 | struct sk_buff *skb; | |
492 | struct udphdr *uh; | |
493 | int err = 0; | |
494 | int is_udplite = IS_UDPLITE(sk); | |
495 | __wsum csum = 0; | |
496 | ||
497 | /* Grab the skbuff where UDP header space exists. */ | |
498 | if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) | |
499 | goto out; | |
500 | ||
501 | /* | |
502 | * Create a UDP header | |
503 | */ | |
504 | uh = udp_hdr(skb); | |
505 | uh->source = fl->fl_ip_sport; | |
506 | uh->dest = fl->fl_ip_dport; | |
507 | uh->len = htons(up->len); | |
508 | uh->check = 0; | |
509 | ||
510 | if (is_udplite) /* UDP-Lite */ | |
511 | csum = udplite_csum_outgoing(sk, skb); | |
512 | ||
513 | else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */ | |
514 | ||
515 | skb->ip_summed = CHECKSUM_NONE; | |
516 | goto send; | |
517 | ||
518 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
519 | ||
520 | udp4_hwcsum_outgoing(sk, skb, fl->fl4_src,fl->fl4_dst, up->len); | |
521 | goto send; | |
522 | ||
523 | } else /* `normal' UDP */ | |
524 | csum = udp_csum_outgoing(sk, skb); | |
525 | ||
526 | /* add protocol-dependent pseudo-header */ | |
527 | uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, up->len, | |
528 | sk->sk_protocol, csum ); | |
529 | if (uh->check == 0) | |
530 | uh->check = CSUM_MANGLED_0; | |
531 | ||
532 | send: | |
533 | err = ip_push_pending_frames(sk); | |
534 | out: | |
535 | up->len = 0; | |
536 | up->pending = 0; | |
537 | if (!err) | |
629ca23c PE |
538 | UDP_INC_STATS_USER(sock_net(sk), |
539 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
db8dac20 DM |
540 | return err; |
541 | } | |
542 | ||
543 | int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
544 | size_t len) | |
545 | { | |
546 | struct inet_sock *inet = inet_sk(sk); | |
547 | struct udp_sock *up = udp_sk(sk); | |
548 | int ulen = len; | |
549 | struct ipcm_cookie ipc; | |
550 | struct rtable *rt = NULL; | |
551 | int free = 0; | |
552 | int connected = 0; | |
553 | __be32 daddr, faddr, saddr; | |
554 | __be16 dport; | |
555 | u8 tos; | |
556 | int err, is_udplite = IS_UDPLITE(sk); | |
557 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
558 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
559 | ||
560 | if (len > 0xFFFF) | |
561 | return -EMSGSIZE; | |
562 | ||
563 | /* | |
564 | * Check the flags. | |
565 | */ | |
566 | ||
567 | if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */ | |
568 | return -EOPNOTSUPP; | |
569 | ||
570 | ipc.opt = NULL; | |
571 | ||
572 | if (up->pending) { | |
573 | /* | |
574 | * There are pending frames. | |
575 | * The socket lock must be held while it's corked. | |
576 | */ | |
577 | lock_sock(sk); | |
578 | if (likely(up->pending)) { | |
579 | if (unlikely(up->pending != AF_INET)) { | |
580 | release_sock(sk); | |
581 | return -EINVAL; | |
582 | } | |
583 | goto do_append_data; | |
584 | } | |
585 | release_sock(sk); | |
586 | } | |
587 | ulen += sizeof(struct udphdr); | |
588 | ||
589 | /* | |
590 | * Get and verify the address. | |
591 | */ | |
592 | if (msg->msg_name) { | |
593 | struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name; | |
594 | if (msg->msg_namelen < sizeof(*usin)) | |
595 | return -EINVAL; | |
596 | if (usin->sin_family != AF_INET) { | |
597 | if (usin->sin_family != AF_UNSPEC) | |
598 | return -EAFNOSUPPORT; | |
599 | } | |
600 | ||
601 | daddr = usin->sin_addr.s_addr; | |
602 | dport = usin->sin_port; | |
603 | if (dport == 0) | |
604 | return -EINVAL; | |
605 | } else { | |
606 | if (sk->sk_state != TCP_ESTABLISHED) | |
607 | return -EDESTADDRREQ; | |
608 | daddr = inet->daddr; | |
609 | dport = inet->dport; | |
610 | /* Open fast path for connected socket. | |
611 | Route will not be used, if at least one option is set. | |
612 | */ | |
613 | connected = 1; | |
614 | } | |
615 | ipc.addr = inet->saddr; | |
616 | ||
617 | ipc.oif = sk->sk_bound_dev_if; | |
618 | if (msg->msg_controllen) { | |
3b1e0a65 | 619 | err = ip_cmsg_send(sock_net(sk), msg, &ipc); |
db8dac20 DM |
620 | if (err) |
621 | return err; | |
622 | if (ipc.opt) | |
623 | free = 1; | |
624 | connected = 0; | |
625 | } | |
626 | if (!ipc.opt) | |
627 | ipc.opt = inet->opt; | |
628 | ||
629 | saddr = ipc.addr; | |
630 | ipc.addr = faddr = daddr; | |
631 | ||
632 | if (ipc.opt && ipc.opt->srr) { | |
633 | if (!daddr) | |
634 | return -EINVAL; | |
635 | faddr = ipc.opt->faddr; | |
636 | connected = 0; | |
637 | } | |
638 | tos = RT_TOS(inet->tos); | |
639 | if (sock_flag(sk, SOCK_LOCALROUTE) || | |
640 | (msg->msg_flags & MSG_DONTROUTE) || | |
641 | (ipc.opt && ipc.opt->is_strictroute)) { | |
642 | tos |= RTO_ONLINK; | |
643 | connected = 0; | |
644 | } | |
645 | ||
646 | if (ipv4_is_multicast(daddr)) { | |
647 | if (!ipc.oif) | |
648 | ipc.oif = inet->mc_index; | |
649 | if (!saddr) | |
650 | saddr = inet->mc_addr; | |
651 | connected = 0; | |
652 | } | |
653 | ||
654 | if (connected) | |
655 | rt = (struct rtable*)sk_dst_check(sk, 0); | |
656 | ||
657 | if (rt == NULL) { | |
658 | struct flowi fl = { .oif = ipc.oif, | |
659 | .nl_u = { .ip4_u = | |
660 | { .daddr = faddr, | |
661 | .saddr = saddr, | |
662 | .tos = tos } }, | |
663 | .proto = sk->sk_protocol, | |
664 | .uli_u = { .ports = | |
665 | { .sport = inet->sport, | |
666 | .dport = dport } } }; | |
84a3aa00 PE |
667 | struct net *net = sock_net(sk); |
668 | ||
db8dac20 | 669 | security_sk_classify_flow(sk, &fl); |
84a3aa00 | 670 | err = ip_route_output_flow(net, &rt, &fl, sk, 1); |
db8dac20 DM |
671 | if (err) { |
672 | if (err == -ENETUNREACH) | |
7c73a6fa | 673 | IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
674 | goto out; |
675 | } | |
676 | ||
677 | err = -EACCES; | |
678 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
679 | !sock_flag(sk, SOCK_BROADCAST)) | |
680 | goto out; | |
681 | if (connected) | |
682 | sk_dst_set(sk, dst_clone(&rt->u.dst)); | |
683 | } | |
684 | ||
685 | if (msg->msg_flags&MSG_CONFIRM) | |
686 | goto do_confirm; | |
687 | back_from_confirm: | |
688 | ||
689 | saddr = rt->rt_src; | |
690 | if (!ipc.addr) | |
691 | daddr = ipc.addr = rt->rt_dst; | |
692 | ||
693 | lock_sock(sk); | |
694 | if (unlikely(up->pending)) { | |
695 | /* The socket is already corked while preparing it. */ | |
696 | /* ... which is an evident application bug. --ANK */ | |
697 | release_sock(sk); | |
698 | ||
699 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n"); | |
700 | err = -EINVAL; | |
701 | goto out; | |
702 | } | |
703 | /* | |
704 | * Now cork the socket to pend data. | |
705 | */ | |
706 | inet->cork.fl.fl4_dst = daddr; | |
707 | inet->cork.fl.fl_ip_dport = dport; | |
708 | inet->cork.fl.fl4_src = saddr; | |
709 | inet->cork.fl.fl_ip_sport = inet->sport; | |
710 | up->pending = AF_INET; | |
711 | ||
712 | do_append_data: | |
713 | up->len += ulen; | |
714 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; | |
715 | err = ip_append_data(sk, getfrag, msg->msg_iov, ulen, | |
716 | sizeof(struct udphdr), &ipc, rt, | |
717 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); | |
718 | if (err) | |
719 | udp_flush_pending_frames(sk); | |
720 | else if (!corkreq) | |
721 | err = udp_push_pending_frames(sk); | |
722 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
723 | up->pending = 0; | |
724 | release_sock(sk); | |
725 | ||
726 | out: | |
727 | ip_rt_put(rt); | |
728 | if (free) | |
729 | kfree(ipc.opt); | |
730 | if (!err) | |
731 | return len; | |
732 | /* | |
733 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
734 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
735 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
736 | * things). We could add another new stat but at least for now that | |
737 | * seems like overkill. | |
738 | */ | |
739 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
740 | UDP_INC_STATS_USER(sock_net(sk), |
741 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
742 | } |
743 | return err; | |
744 | ||
745 | do_confirm: | |
746 | dst_confirm(&rt->u.dst); | |
747 | if (!(msg->msg_flags&MSG_PROBE) || len) | |
748 | goto back_from_confirm; | |
749 | err = 0; | |
750 | goto out; | |
751 | } | |
752 | ||
753 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
754 | size_t size, int flags) | |
755 | { | |
756 | struct udp_sock *up = udp_sk(sk); | |
757 | int ret; | |
758 | ||
759 | if (!up->pending) { | |
760 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
761 | ||
762 | /* Call udp_sendmsg to specify destination address which | |
763 | * sendpage interface can't pass. | |
764 | * This will succeed only when the socket is connected. | |
765 | */ | |
766 | ret = udp_sendmsg(NULL, sk, &msg, 0); | |
767 | if (ret < 0) | |
768 | return ret; | |
769 | } | |
770 | ||
771 | lock_sock(sk); | |
772 | ||
773 | if (unlikely(!up->pending)) { | |
774 | release_sock(sk); | |
775 | ||
776 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n"); | |
777 | return -EINVAL; | |
778 | } | |
779 | ||
780 | ret = ip_append_page(sk, page, offset, size, flags); | |
781 | if (ret == -EOPNOTSUPP) { | |
782 | release_sock(sk); | |
783 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
784 | size, flags); | |
785 | } | |
786 | if (ret < 0) { | |
787 | udp_flush_pending_frames(sk); | |
788 | goto out; | |
789 | } | |
790 | ||
791 | up->len += size; | |
792 | if (!(up->corkflag || (flags&MSG_MORE))) | |
793 | ret = udp_push_pending_frames(sk); | |
794 | if (!ret) | |
795 | ret = size; | |
796 | out: | |
797 | release_sock(sk); | |
798 | return ret; | |
799 | } | |
800 | ||
1da177e4 LT |
801 | /* |
802 | * IOCTL requests applicable to the UDP protocol | |
803 | */ | |
e905a9ed | 804 | |
1da177e4 LT |
805 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
806 | { | |
6516c655 SH |
807 | switch (cmd) { |
808 | case SIOCOUTQ: | |
1da177e4 | 809 | { |
6516c655 SH |
810 | int amount = atomic_read(&sk->sk_wmem_alloc); |
811 | return put_user(amount, (int __user *)arg); | |
812 | } | |
1da177e4 | 813 | |
6516c655 SH |
814 | case SIOCINQ: |
815 | { | |
816 | struct sk_buff *skb; | |
817 | unsigned long amount; | |
818 | ||
819 | amount = 0; | |
820 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
821 | skb = skb_peek(&sk->sk_receive_queue); | |
822 | if (skb != NULL) { | |
823 | /* | |
824 | * We will only return the amount | |
825 | * of this packet since that is all | |
826 | * that will be read. | |
827 | */ | |
828 | amount = skb->len - sizeof(struct udphdr); | |
1da177e4 | 829 | } |
6516c655 SH |
830 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
831 | return put_user(amount, (int __user *)arg); | |
832 | } | |
1da177e4 | 833 | |
6516c655 SH |
834 | default: |
835 | return -ENOIOCTLCMD; | |
1da177e4 | 836 | } |
6516c655 SH |
837 | |
838 | return 0; | |
1da177e4 LT |
839 | } |
840 | ||
db8dac20 DM |
841 | /* |
842 | * This should be easy, if there is something there we | |
843 | * return it, otherwise we block. | |
844 | */ | |
845 | ||
846 | int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
847 | size_t len, int noblock, int flags, int *addr_len) | |
848 | { | |
849 | struct inet_sock *inet = inet_sk(sk); | |
850 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; | |
851 | struct sk_buff *skb; | |
852 | unsigned int ulen, copied; | |
853 | int peeked; | |
854 | int err; | |
855 | int is_udplite = IS_UDPLITE(sk); | |
856 | ||
857 | /* | |
858 | * Check any passed addresses | |
859 | */ | |
860 | if (addr_len) | |
861 | *addr_len=sizeof(*sin); | |
862 | ||
863 | if (flags & MSG_ERRQUEUE) | |
864 | return ip_recv_error(sk, msg, len); | |
865 | ||
866 | try_again: | |
867 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), | |
868 | &peeked, &err); | |
869 | if (!skb) | |
870 | goto out; | |
871 | ||
872 | ulen = skb->len - sizeof(struct udphdr); | |
873 | copied = len; | |
874 | if (copied > ulen) | |
875 | copied = ulen; | |
876 | else if (copied < ulen) | |
877 | msg->msg_flags |= MSG_TRUNC; | |
878 | ||
879 | /* | |
880 | * If checksum is needed at all, try to do it while copying the | |
881 | * data. If the data is truncated, or if we only want a partial | |
882 | * coverage checksum (UDP-Lite), do it before the copy. | |
883 | */ | |
884 | ||
885 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { | |
886 | if (udp_lib_checksum_complete(skb)) | |
887 | goto csum_copy_err; | |
888 | } | |
889 | ||
890 | if (skb_csum_unnecessary(skb)) | |
891 | err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), | |
892 | msg->msg_iov, copied ); | |
893 | else { | |
894 | err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov); | |
895 | ||
896 | if (err == -EINVAL) | |
897 | goto csum_copy_err; | |
898 | } | |
899 | ||
900 | if (err) | |
901 | goto out_free; | |
902 | ||
903 | if (!peeked) | |
629ca23c PE |
904 | UDP_INC_STATS_USER(sock_net(sk), |
905 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 DM |
906 | |
907 | sock_recv_timestamp(msg, sk, skb); | |
908 | ||
909 | /* Copy the address. */ | |
910 | if (sin) | |
911 | { | |
912 | sin->sin_family = AF_INET; | |
913 | sin->sin_port = udp_hdr(skb)->source; | |
914 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
915 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
916 | } | |
917 | if (inet->cmsg_flags) | |
918 | ip_cmsg_recv(msg, skb); | |
919 | ||
920 | err = copied; | |
921 | if (flags & MSG_TRUNC) | |
922 | err = ulen; | |
923 | ||
924 | out_free: | |
925 | lock_sock(sk); | |
926 | skb_free_datagram(sk, skb); | |
927 | release_sock(sk); | |
928 | out: | |
929 | return err; | |
930 | ||
931 | csum_copy_err: | |
932 | lock_sock(sk); | |
933 | if (!skb_kill_datagram(sk, skb, flags)) | |
629ca23c | 934 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
db8dac20 DM |
935 | release_sock(sk); |
936 | ||
937 | if (noblock) | |
938 | return -EAGAIN; | |
939 | goto try_again; | |
940 | } | |
941 | ||
942 | ||
1da177e4 LT |
943 | int udp_disconnect(struct sock *sk, int flags) |
944 | { | |
945 | struct inet_sock *inet = inet_sk(sk); | |
946 | /* | |
947 | * 1003.1g - break association. | |
948 | */ | |
e905a9ed | 949 | |
1da177e4 LT |
950 | sk->sk_state = TCP_CLOSE; |
951 | inet->daddr = 0; | |
952 | inet->dport = 0; | |
953 | sk->sk_bound_dev_if = 0; | |
954 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
955 | inet_reset_saddr(sk); | |
956 | ||
957 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
958 | sk->sk_prot->unhash(sk); | |
959 | inet->sport = 0; | |
960 | } | |
961 | sk_dst_reset(sk); | |
962 | return 0; | |
963 | } | |
964 | ||
645ca708 ED |
965 | void udp_lib_unhash(struct sock *sk) |
966 | { | |
967 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
968 | unsigned int hash = udp_hashfn(sock_net(sk), sk->sk_hash); | |
969 | struct udp_hslot *hslot = &udptable->hash[hash]; | |
970 | ||
971 | spin_lock(&hslot->lock); | |
271b72c7 | 972 | if (sk_del_node_init_rcu(sk)) { |
645ca708 ED |
973 | inet_sk(sk)->num = 0; |
974 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); | |
975 | } | |
976 | spin_unlock(&hslot->lock); | |
977 | } | |
978 | EXPORT_SYMBOL(udp_lib_unhash); | |
979 | ||
93821778 HX |
980 | static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
981 | { | |
982 | int is_udplite = IS_UDPLITE(sk); | |
983 | int rc; | |
984 | ||
985 | if ((rc = sock_queue_rcv_skb(sk, skb)) < 0) { | |
986 | /* Note that an ENOMEM error is charged twice */ | |
987 | if (rc == -ENOMEM) | |
988 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, | |
989 | is_udplite); | |
990 | goto drop; | |
991 | } | |
992 | ||
993 | return 0; | |
994 | ||
995 | drop: | |
996 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); | |
997 | kfree_skb(skb); | |
998 | return -1; | |
999 | } | |
1000 | ||
db8dac20 DM |
1001 | /* returns: |
1002 | * -1: error | |
1003 | * 0: success | |
1004 | * >0: "udp encap" protocol resubmission | |
1005 | * | |
1006 | * Note that in the success and error cases, the skb is assumed to | |
1007 | * have either been requeued or freed. | |
1008 | */ | |
1009 | int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb) | |
1010 | { | |
1011 | struct udp_sock *up = udp_sk(sk); | |
1012 | int rc; | |
1013 | int is_udplite = IS_UDPLITE(sk); | |
1014 | ||
1015 | /* | |
1016 | * Charge it to the socket, dropping if the queue is full. | |
1017 | */ | |
1018 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1019 | goto drop; | |
1020 | nf_reset(skb); | |
1021 | ||
1022 | if (up->encap_type) { | |
1023 | /* | |
1024 | * This is an encapsulation socket so pass the skb to | |
1025 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1026 | * fall through and pass this up the UDP socket. | |
1027 | * up->encap_rcv() returns the following value: | |
1028 | * =0 if skb was successfully passed to the encap | |
1029 | * handler or was discarded by it. | |
1030 | * >0 if skb should be passed on to UDP. | |
1031 | * <0 if skb should be resubmitted as proto -N | |
1032 | */ | |
1033 | ||
1034 | /* if we're overly short, let UDP handle it */ | |
1035 | if (skb->len > sizeof(struct udphdr) && | |
1036 | up->encap_rcv != NULL) { | |
1037 | int ret; | |
1038 | ||
1039 | ret = (*up->encap_rcv)(sk, skb); | |
1040 | if (ret <= 0) { | |
0283328e PE |
1041 | UDP_INC_STATS_BH(sock_net(sk), |
1042 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
1043 | is_udplite); |
1044 | return -ret; | |
1045 | } | |
1046 | } | |
1047 | ||
1048 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1049 | } | |
1050 | ||
1051 | /* | |
1052 | * UDP-Lite specific tests, ignored on UDP sockets | |
1053 | */ | |
1054 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1055 | ||
1056 | /* | |
1057 | * MIB statistics other than incrementing the error count are | |
1058 | * disabled for the following two types of errors: these depend | |
1059 | * on the application settings, not on the functioning of the | |
1060 | * protocol stack as such. | |
1061 | * | |
1062 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1063 | * way ... to ... at least let the receiving application block | |
1064 | * delivery of packets with coverage values less than a value | |
1065 | * provided by the application." | |
1066 | */ | |
1067 | if (up->pcrlen == 0) { /* full coverage was set */ | |
1068 | LIMIT_NETDEBUG(KERN_WARNING "UDPLITE: partial coverage " | |
1069 | "%d while full coverage %d requested\n", | |
1070 | UDP_SKB_CB(skb)->cscov, skb->len); | |
1071 | goto drop; | |
1072 | } | |
1073 | /* The next case involves violating the min. coverage requested | |
1074 | * by the receiver. This is subtle: if receiver wants x and x is | |
1075 | * greater than the buffersize/MTU then receiver will complain | |
1076 | * that it wants x while sender emits packets of smaller size y. | |
1077 | * Therefore the above ...()->partial_cov statement is essential. | |
1078 | */ | |
1079 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
1080 | LIMIT_NETDEBUG(KERN_WARNING | |
1081 | "UDPLITE: coverage %d too small, need min %d\n", | |
1082 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
1083 | goto drop; | |
1084 | } | |
1085 | } | |
1086 | ||
1087 | if (sk->sk_filter) { | |
1088 | if (udp_lib_checksum_complete(skb)) | |
1089 | goto drop; | |
1090 | } | |
1091 | ||
93821778 | 1092 | rc = 0; |
db8dac20 | 1093 | |
93821778 HX |
1094 | bh_lock_sock(sk); |
1095 | if (!sock_owned_by_user(sk)) | |
1096 | rc = __udp_queue_rcv_skb(sk, skb); | |
1097 | else | |
1098 | sk_add_backlog(sk, skb); | |
1099 | bh_unlock_sock(sk); | |
1100 | ||
1101 | return rc; | |
db8dac20 DM |
1102 | |
1103 | drop: | |
0283328e | 1104 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
db8dac20 DM |
1105 | kfree_skb(skb); |
1106 | return -1; | |
1107 | } | |
1108 | ||
1109 | /* | |
1110 | * Multicasts and broadcasts go to each listener. | |
1111 | * | |
1112 | * Note: called only from the BH handler context, | |
1113 | * so we don't need to lock the hashes. | |
1114 | */ | |
e3163493 | 1115 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1116 | struct udphdr *uh, |
1117 | __be32 saddr, __be32 daddr, | |
645ca708 | 1118 | struct udp_table *udptable) |
db8dac20 DM |
1119 | { |
1120 | struct sock *sk; | |
645ca708 | 1121 | struct udp_hslot *hslot = &udptable->hash[udp_hashfn(net, ntohs(uh->dest))]; |
db8dac20 DM |
1122 | int dif; |
1123 | ||
645ca708 ED |
1124 | spin_lock(&hslot->lock); |
1125 | sk = sk_head(&hslot->head); | |
db8dac20 DM |
1126 | dif = skb->dev->ifindex; |
1127 | sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif); | |
1128 | if (sk) { | |
1129 | struct sock *sknext = NULL; | |
1130 | ||
1131 | do { | |
1132 | struct sk_buff *skb1 = skb; | |
1133 | ||
1134 | sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr, | |
1135 | uh->source, saddr, dif); | |
1136 | if (sknext) | |
1137 | skb1 = skb_clone(skb, GFP_ATOMIC); | |
1138 | ||
1139 | if (skb1) { | |
93821778 | 1140 | int ret = udp_queue_rcv_skb(sk, skb1); |
db8dac20 DM |
1141 | if (ret > 0) |
1142 | /* we should probably re-process instead | |
1143 | * of dropping packets here. */ | |
1144 | kfree_skb(skb1); | |
1145 | } | |
1146 | sk = sknext; | |
1147 | } while (sknext); | |
1148 | } else | |
1149 | kfree_skb(skb); | |
645ca708 | 1150 | spin_unlock(&hslot->lock); |
db8dac20 DM |
1151 | return 0; |
1152 | } | |
1153 | ||
1154 | /* Initialize UDP checksum. If exited with zero value (success), | |
1155 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1156 | * Otherwise, csum completion requires chacksumming packet body, | |
1157 | * including udp header and folding it to skb->csum. | |
1158 | */ | |
1159 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1160 | int proto) | |
1161 | { | |
1162 | const struct iphdr *iph; | |
1163 | int err; | |
1164 | ||
1165 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1166 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1167 | ||
1168 | if (proto == IPPROTO_UDPLITE) { | |
1169 | err = udplite_checksum_init(skb, uh); | |
1170 | if (err) | |
1171 | return err; | |
1172 | } | |
1173 | ||
1174 | iph = ip_hdr(skb); | |
1175 | if (uh->check == 0) { | |
1176 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1177 | } else if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
1178 | if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, | |
1179 | proto, skb->csum)) | |
1180 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1181 | } | |
1182 | if (!skb_csum_unnecessary(skb)) | |
1183 | skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, | |
1184 | skb->len, proto, 0); | |
1185 | /* Probably, we should checksum udp header (it should be in cache | |
1186 | * in any case) and data in tiny packets (< rx copybreak). | |
1187 | */ | |
1188 | ||
1189 | return 0; | |
1190 | } | |
1191 | ||
1192 | /* | |
1193 | * All we need to do is get the socket, and then do a checksum. | |
1194 | */ | |
1195 | ||
645ca708 | 1196 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1197 | int proto) |
1198 | { | |
1199 | struct sock *sk; | |
1200 | struct udphdr *uh = udp_hdr(skb); | |
1201 | unsigned short ulen; | |
1202 | struct rtable *rt = (struct rtable*)skb->dst; | |
1203 | __be32 saddr = ip_hdr(skb)->saddr; | |
1204 | __be32 daddr = ip_hdr(skb)->daddr; | |
0283328e | 1205 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1206 | |
1207 | /* | |
1208 | * Validate the packet. | |
1209 | */ | |
1210 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1211 | goto drop; /* No space for header. */ | |
1212 | ||
1213 | ulen = ntohs(uh->len); | |
1214 | if (ulen > skb->len) | |
1215 | goto short_packet; | |
1216 | ||
1217 | if (proto == IPPROTO_UDP) { | |
1218 | /* UDP validates ulen. */ | |
1219 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1220 | goto short_packet; | |
1221 | uh = udp_hdr(skb); | |
1222 | } | |
1223 | ||
1224 | if (udp4_csum_init(skb, uh, proto)) | |
1225 | goto csum_error; | |
1226 | ||
1227 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) | |
e3163493 PE |
1228 | return __udp4_lib_mcast_deliver(net, skb, uh, |
1229 | saddr, daddr, udptable); | |
db8dac20 | 1230 | |
607c4aaf | 1231 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); |
db8dac20 DM |
1232 | |
1233 | if (sk != NULL) { | |
93821778 | 1234 | int ret = udp_queue_rcv_skb(sk, skb); |
db8dac20 DM |
1235 | sock_put(sk); |
1236 | ||
1237 | /* a return value > 0 means to resubmit the input, but | |
1238 | * it wants the return to be -protocol, or 0 | |
1239 | */ | |
1240 | if (ret > 0) | |
1241 | return -ret; | |
1242 | return 0; | |
1243 | } | |
1244 | ||
1245 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1246 | goto drop; | |
1247 | nf_reset(skb); | |
1248 | ||
1249 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1250 | if (udp_lib_checksum_complete(skb)) | |
1251 | goto csum_error; | |
1252 | ||
0283328e | 1253 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1254 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1255 | ||
1256 | /* | |
1257 | * Hmm. We got an UDP packet to a port to which we | |
1258 | * don't wanna listen. Ignore it. | |
1259 | */ | |
1260 | kfree_skb(skb); | |
1261 | return 0; | |
1262 | ||
1263 | short_packet: | |
a7d632b6 | 1264 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From " NIPQUAD_FMT ":%u %d/%d to " NIPQUAD_FMT ":%u\n", |
db8dac20 DM |
1265 | proto == IPPROTO_UDPLITE ? "-Lite" : "", |
1266 | NIPQUAD(saddr), | |
1267 | ntohs(uh->source), | |
1268 | ulen, | |
1269 | skb->len, | |
1270 | NIPQUAD(daddr), | |
1271 | ntohs(uh->dest)); | |
1272 | goto drop; | |
1273 | ||
1274 | csum_error: | |
1275 | /* | |
1276 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1277 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1278 | */ | |
a7d632b6 | 1279 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From " NIPQUAD_FMT ":%u to " NIPQUAD_FMT ":%u ulen %d\n", |
db8dac20 DM |
1280 | proto == IPPROTO_UDPLITE ? "-Lite" : "", |
1281 | NIPQUAD(saddr), | |
1282 | ntohs(uh->source), | |
1283 | NIPQUAD(daddr), | |
1284 | ntohs(uh->dest), | |
1285 | ulen); | |
1286 | drop: | |
0283328e | 1287 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1288 | kfree_skb(skb); |
1289 | return 0; | |
1290 | } | |
1291 | ||
1292 | int udp_rcv(struct sk_buff *skb) | |
1293 | { | |
645ca708 | 1294 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
1295 | } |
1296 | ||
7d06b2e0 | 1297 | void udp_destroy_sock(struct sock *sk) |
db8dac20 DM |
1298 | { |
1299 | lock_sock(sk); | |
1300 | udp_flush_pending_frames(sk); | |
1301 | release_sock(sk); | |
db8dac20 DM |
1302 | } |
1303 | ||
1da177e4 LT |
1304 | /* |
1305 | * Socket option code for UDP | |
1306 | */ | |
4c0a6cb0 GR |
1307 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
1308 | char __user *optval, int optlen, | |
1309 | int (*push_pending_frames)(struct sock *)) | |
1da177e4 LT |
1310 | { |
1311 | struct udp_sock *up = udp_sk(sk); | |
1312 | int val; | |
1313 | int err = 0; | |
b2bf1e26 | 1314 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 1315 | |
6516c655 | 1316 | if (optlen<sizeof(int)) |
1da177e4 LT |
1317 | return -EINVAL; |
1318 | ||
1319 | if (get_user(val, (int __user *)optval)) | |
1320 | return -EFAULT; | |
1321 | ||
6516c655 | 1322 | switch (optname) { |
1da177e4 LT |
1323 | case UDP_CORK: |
1324 | if (val != 0) { | |
1325 | up->corkflag = 1; | |
1326 | } else { | |
1327 | up->corkflag = 0; | |
1328 | lock_sock(sk); | |
4c0a6cb0 | 1329 | (*push_pending_frames)(sk); |
1da177e4 LT |
1330 | release_sock(sk); |
1331 | } | |
1332 | break; | |
e905a9ed | 1333 | |
1da177e4 LT |
1334 | case UDP_ENCAP: |
1335 | switch (val) { | |
1336 | case 0: | |
1337 | case UDP_ENCAP_ESPINUDP: | |
1338 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
1339 | up->encap_rcv = xfrm4_udp_encap_rcv; |
1340 | /* FALLTHROUGH */ | |
342f0234 | 1341 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 LT |
1342 | up->encap_type = val; |
1343 | break; | |
1344 | default: | |
1345 | err = -ENOPROTOOPT; | |
1346 | break; | |
1347 | } | |
1348 | break; | |
1349 | ||
ba4e58ec GR |
1350 | /* |
1351 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
1352 | */ | |
1353 | /* The sender sets actual checksum coverage length via this option. | |
1354 | * The case coverage > packet length is handled by send module. */ | |
1355 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 1356 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1357 | return -ENOPROTOOPT; |
1358 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
1359 | val = 8; | |
47112e25 GR |
1360 | else if (val > USHORT_MAX) |
1361 | val = USHORT_MAX; | |
ba4e58ec GR |
1362 | up->pcslen = val; |
1363 | up->pcflag |= UDPLITE_SEND_CC; | |
1364 | break; | |
1365 | ||
e905a9ed YH |
1366 | /* The receiver specifies a minimum checksum coverage value. To make |
1367 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
1368 | * used, this again means full checksum coverage. */ |
1369 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 1370 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1371 | return -ENOPROTOOPT; |
1372 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
1373 | val = 8; | |
47112e25 GR |
1374 | else if (val > USHORT_MAX) |
1375 | val = USHORT_MAX; | |
ba4e58ec GR |
1376 | up->pcrlen = val; |
1377 | up->pcflag |= UDPLITE_RECV_CC; | |
1378 | break; | |
1379 | ||
1da177e4 LT |
1380 | default: |
1381 | err = -ENOPROTOOPT; | |
1382 | break; | |
6516c655 | 1383 | } |
1da177e4 LT |
1384 | |
1385 | return err; | |
1386 | } | |
1387 | ||
db8dac20 DM |
1388 | int udp_setsockopt(struct sock *sk, int level, int optname, |
1389 | char __user *optval, int optlen) | |
1390 | { | |
1391 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1392 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1393 | udp_push_pending_frames); | |
1394 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
1395 | } | |
1396 | ||
1397 | #ifdef CONFIG_COMPAT | |
1398 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
1399 | char __user *optval, int optlen) | |
1400 | { | |
1401 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1402 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1403 | udp_push_pending_frames); | |
1404 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
1405 | } | |
1406 | #endif | |
1407 | ||
4c0a6cb0 GR |
1408 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
1409 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
1410 | { |
1411 | struct udp_sock *up = udp_sk(sk); | |
1412 | int val, len; | |
1413 | ||
6516c655 | 1414 | if (get_user(len,optlen)) |
1da177e4 LT |
1415 | return -EFAULT; |
1416 | ||
1417 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 1418 | |
6516c655 | 1419 | if (len < 0) |
1da177e4 LT |
1420 | return -EINVAL; |
1421 | ||
6516c655 | 1422 | switch (optname) { |
1da177e4 LT |
1423 | case UDP_CORK: |
1424 | val = up->corkflag; | |
1425 | break; | |
1426 | ||
1427 | case UDP_ENCAP: | |
1428 | val = up->encap_type; | |
1429 | break; | |
1430 | ||
ba4e58ec GR |
1431 | /* The following two cannot be changed on UDP sockets, the return is |
1432 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
1433 | case UDPLITE_SEND_CSCOV: | |
1434 | val = up->pcslen; | |
1435 | break; | |
1436 | ||
1437 | case UDPLITE_RECV_CSCOV: | |
1438 | val = up->pcrlen; | |
1439 | break; | |
1440 | ||
1da177e4 LT |
1441 | default: |
1442 | return -ENOPROTOOPT; | |
6516c655 | 1443 | } |
1da177e4 | 1444 | |
6516c655 | 1445 | if (put_user(len, optlen)) |
e905a9ed | 1446 | return -EFAULT; |
6516c655 | 1447 | if (copy_to_user(optval, &val,len)) |
1da177e4 | 1448 | return -EFAULT; |
e905a9ed | 1449 | return 0; |
1da177e4 LT |
1450 | } |
1451 | ||
db8dac20 DM |
1452 | int udp_getsockopt(struct sock *sk, int level, int optname, |
1453 | char __user *optval, int __user *optlen) | |
1454 | { | |
1455 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1456 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1457 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
1458 | } | |
1459 | ||
1460 | #ifdef CONFIG_COMPAT | |
1461 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
1462 | char __user *optval, int __user *optlen) | |
1463 | { | |
1464 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1465 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1466 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
1467 | } | |
1468 | #endif | |
1da177e4 LT |
1469 | /** |
1470 | * udp_poll - wait for a UDP event. | |
1471 | * @file - file struct | |
1472 | * @sock - socket | |
1473 | * @wait - poll table | |
1474 | * | |
e905a9ed | 1475 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
1476 | * blocking sockets. If application is using a blocking fd |
1477 | * and a packet with checksum error is in the queue; | |
1478 | * then it could get return from select indicating data available | |
1479 | * but then block when reading it. Add special case code | |
1480 | * to work around these arguably broken applications. | |
1481 | */ | |
1482 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
1483 | { | |
1484 | unsigned int mask = datagram_poll(file, sock, wait); | |
1485 | struct sock *sk = sock->sk; | |
ba4e58ec GR |
1486 | int is_lite = IS_UDPLITE(sk); |
1487 | ||
1da177e4 LT |
1488 | /* Check for false positives due to checksum errors */ |
1489 | if ( (mask & POLLRDNORM) && | |
1490 | !(file->f_flags & O_NONBLOCK) && | |
1491 | !(sk->sk_shutdown & RCV_SHUTDOWN)){ | |
1492 | struct sk_buff_head *rcvq = &sk->sk_receive_queue; | |
1493 | struct sk_buff *skb; | |
1494 | ||
208d8984 | 1495 | spin_lock_bh(&rcvq->lock); |
759e5d00 HX |
1496 | while ((skb = skb_peek(rcvq)) != NULL && |
1497 | udp_lib_checksum_complete(skb)) { | |
0283328e PE |
1498 | UDP_INC_STATS_BH(sock_net(sk), |
1499 | UDP_MIB_INERRORS, is_lite); | |
759e5d00 HX |
1500 | __skb_unlink(skb, rcvq); |
1501 | kfree_skb(skb); | |
1da177e4 | 1502 | } |
208d8984 | 1503 | spin_unlock_bh(&rcvq->lock); |
1da177e4 LT |
1504 | |
1505 | /* nothing to see, move along */ | |
1506 | if (skb == NULL) | |
1507 | mask &= ~(POLLIN | POLLRDNORM); | |
1508 | } | |
1509 | ||
1510 | return mask; | |
e905a9ed | 1511 | |
1da177e4 LT |
1512 | } |
1513 | ||
db8dac20 DM |
1514 | struct proto udp_prot = { |
1515 | .name = "UDP", | |
1516 | .owner = THIS_MODULE, | |
1517 | .close = udp_lib_close, | |
1518 | .connect = ip4_datagram_connect, | |
1519 | .disconnect = udp_disconnect, | |
1520 | .ioctl = udp_ioctl, | |
1521 | .destroy = udp_destroy_sock, | |
1522 | .setsockopt = udp_setsockopt, | |
1523 | .getsockopt = udp_getsockopt, | |
1524 | .sendmsg = udp_sendmsg, | |
1525 | .recvmsg = udp_recvmsg, | |
1526 | .sendpage = udp_sendpage, | |
93821778 | 1527 | .backlog_rcv = __udp_queue_rcv_skb, |
db8dac20 DM |
1528 | .hash = udp_lib_hash, |
1529 | .unhash = udp_lib_unhash, | |
1530 | .get_port = udp_v4_get_port, | |
1531 | .memory_allocated = &udp_memory_allocated, | |
1532 | .sysctl_mem = sysctl_udp_mem, | |
1533 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
1534 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
1535 | .obj_size = sizeof(struct udp_sock), | |
271b72c7 | 1536 | .slab_flags = SLAB_DESTROY_BY_RCU, |
645ca708 | 1537 | .h.udp_table = &udp_table, |
db8dac20 DM |
1538 | #ifdef CONFIG_COMPAT |
1539 | .compat_setsockopt = compat_udp_setsockopt, | |
1540 | .compat_getsockopt = compat_udp_getsockopt, | |
1541 | #endif | |
db8dac20 | 1542 | }; |
1da177e4 LT |
1543 | |
1544 | /* ------------------------------------------------------------------------ */ | |
1545 | #ifdef CONFIG_PROC_FS | |
1546 | ||
645ca708 | 1547 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
1548 | { |
1549 | struct sock *sk; | |
1550 | struct udp_iter_state *state = seq->private; | |
6f191efe | 1551 | struct net *net = seq_file_net(seq); |
1da177e4 | 1552 | |
645ca708 | 1553 | for (state->bucket = start; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) { |
1da177e4 | 1554 | struct hlist_node *node; |
645ca708 ED |
1555 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
1556 | spin_lock_bh(&hslot->lock); | |
1557 | sk_for_each(sk, node, &hslot->head) { | |
878628fb | 1558 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 1559 | continue; |
1da177e4 LT |
1560 | if (sk->sk_family == state->family) |
1561 | goto found; | |
1562 | } | |
645ca708 | 1563 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
1564 | } |
1565 | sk = NULL; | |
1566 | found: | |
1567 | return sk; | |
1568 | } | |
1569 | ||
1570 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
1571 | { | |
1572 | struct udp_iter_state *state = seq->private; | |
6f191efe | 1573 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1574 | |
1575 | do { | |
1576 | sk = sk_next(sk); | |
878628fb | 1577 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 1578 | |
645ca708 ED |
1579 | if (!sk) { |
1580 | spin_unlock(&state->udp_table->hash[state->bucket].lock); | |
1581 | return udp_get_first(seq, state->bucket + 1); | |
1da177e4 LT |
1582 | } |
1583 | return sk; | |
1584 | } | |
1585 | ||
1586 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
1587 | { | |
645ca708 | 1588 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
1589 | |
1590 | if (sk) | |
6516c655 | 1591 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
1592 | --pos; |
1593 | return pos ? NULL : sk; | |
1594 | } | |
1595 | ||
1596 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
1597 | { | |
b50660f1 | 1598 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
1599 | } |
1600 | ||
1601 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
1602 | { | |
1603 | struct sock *sk; | |
1604 | ||
b50660f1 | 1605 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
1606 | sk = udp_get_idx(seq, 0); |
1607 | else | |
1608 | sk = udp_get_next(seq, v); | |
1609 | ||
1610 | ++*pos; | |
1611 | return sk; | |
1612 | } | |
1613 | ||
1614 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
1615 | { | |
645ca708 ED |
1616 | struct udp_iter_state *state = seq->private; |
1617 | ||
1618 | if (state->bucket < UDP_HTABLE_SIZE) | |
1619 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); | |
1da177e4 LT |
1620 | } |
1621 | ||
1622 | static int udp_seq_open(struct inode *inode, struct file *file) | |
1623 | { | |
1624 | struct udp_seq_afinfo *afinfo = PDE(inode)->data; | |
a2be75c1 DL |
1625 | struct udp_iter_state *s; |
1626 | int err; | |
a91275ef | 1627 | |
a2be75c1 DL |
1628 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
1629 | sizeof(struct udp_iter_state)); | |
1630 | if (err < 0) | |
1631 | return err; | |
a91275ef | 1632 | |
a2be75c1 | 1633 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 1634 | s->family = afinfo->family; |
645ca708 | 1635 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 1636 | return err; |
a91275ef DL |
1637 | } |
1638 | ||
1da177e4 | 1639 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 1640 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
1641 | { |
1642 | struct proc_dir_entry *p; | |
1643 | int rc = 0; | |
1644 | ||
3ba9441b DL |
1645 | afinfo->seq_fops.open = udp_seq_open; |
1646 | afinfo->seq_fops.read = seq_read; | |
1647 | afinfo->seq_fops.llseek = seq_lseek; | |
1648 | afinfo->seq_fops.release = seq_release_net; | |
1da177e4 | 1649 | |
dda61925 DL |
1650 | afinfo->seq_ops.start = udp_seq_start; |
1651 | afinfo->seq_ops.next = udp_seq_next; | |
1652 | afinfo->seq_ops.stop = udp_seq_stop; | |
1653 | ||
84841c3c DL |
1654 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
1655 | &afinfo->seq_fops, afinfo); | |
1656 | if (!p) | |
1da177e4 LT |
1657 | rc = -ENOMEM; |
1658 | return rc; | |
1659 | } | |
1660 | ||
0c96d8c5 | 1661 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 1662 | { |
0c96d8c5 | 1663 | proc_net_remove(net, afinfo->name); |
1da177e4 | 1664 | } |
db8dac20 DM |
1665 | |
1666 | /* ------------------------------------------------------------------------ */ | |
5e659e4c PE |
1667 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
1668 | int bucket, int *len) | |
db8dac20 DM |
1669 | { |
1670 | struct inet_sock *inet = inet_sk(sp); | |
1671 | __be32 dest = inet->daddr; | |
1672 | __be32 src = inet->rcv_saddr; | |
1673 | __u16 destp = ntohs(inet->dport); | |
1674 | __u16 srcp = ntohs(inet->sport); | |
1675 | ||
5e659e4c | 1676 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
cb61cb9b | 1677 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d%n", |
db8dac20 DM |
1678 | bucket, src, srcp, dest, destp, sp->sk_state, |
1679 | atomic_read(&sp->sk_wmem_alloc), | |
1680 | atomic_read(&sp->sk_rmem_alloc), | |
1681 | 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp), | |
cb61cb9b ED |
1682 | atomic_read(&sp->sk_refcnt), sp, |
1683 | atomic_read(&sp->sk_drops), len); | |
db8dac20 DM |
1684 | } |
1685 | ||
1686 | int udp4_seq_show(struct seq_file *seq, void *v) | |
1687 | { | |
1688 | if (v == SEQ_START_TOKEN) | |
1689 | seq_printf(seq, "%-127s\n", | |
1690 | " sl local_address rem_address st tx_queue " | |
1691 | "rx_queue tr tm->when retrnsmt uid timeout " | |
cb61cb9b | 1692 | "inode ref pointer drops"); |
db8dac20 | 1693 | else { |
db8dac20 | 1694 | struct udp_iter_state *state = seq->private; |
5e659e4c | 1695 | int len; |
db8dac20 | 1696 | |
5e659e4c PE |
1697 | udp4_format_sock(v, seq, state->bucket, &len); |
1698 | seq_printf(seq, "%*s\n", 127 - len ,""); | |
db8dac20 DM |
1699 | } |
1700 | return 0; | |
1701 | } | |
1702 | ||
1703 | /* ------------------------------------------------------------------------ */ | |
db8dac20 | 1704 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
1705 | .name = "udp", |
1706 | .family = AF_INET, | |
645ca708 | 1707 | .udp_table = &udp_table, |
4ad96d39 DL |
1708 | .seq_fops = { |
1709 | .owner = THIS_MODULE, | |
1710 | }, | |
dda61925 DL |
1711 | .seq_ops = { |
1712 | .show = udp4_seq_show, | |
1713 | }, | |
db8dac20 DM |
1714 | }; |
1715 | ||
15439feb PE |
1716 | static int udp4_proc_init_net(struct net *net) |
1717 | { | |
1718 | return udp_proc_register(net, &udp4_seq_afinfo); | |
1719 | } | |
1720 | ||
1721 | static void udp4_proc_exit_net(struct net *net) | |
1722 | { | |
1723 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
1724 | } | |
1725 | ||
1726 | static struct pernet_operations udp4_net_ops = { | |
1727 | .init = udp4_proc_init_net, | |
1728 | .exit = udp4_proc_exit_net, | |
1729 | }; | |
1730 | ||
db8dac20 DM |
1731 | int __init udp4_proc_init(void) |
1732 | { | |
15439feb | 1733 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
1734 | } |
1735 | ||
1736 | void udp4_proc_exit(void) | |
1737 | { | |
15439feb | 1738 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 1739 | } |
1da177e4 LT |
1740 | #endif /* CONFIG_PROC_FS */ |
1741 | ||
645ca708 ED |
1742 | void __init udp_table_init(struct udp_table *table) |
1743 | { | |
1744 | int i; | |
1745 | ||
1746 | for (i = 0; i < UDP_HTABLE_SIZE; i++) { | |
1747 | INIT_HLIST_HEAD(&table->hash[i].head); | |
1748 | spin_lock_init(&table->hash[i].lock); | |
1749 | } | |
1750 | } | |
1751 | ||
95766fff HA |
1752 | void __init udp_init(void) |
1753 | { | |
1754 | unsigned long limit; | |
1755 | ||
645ca708 | 1756 | udp_table_init(&udp_table); |
95766fff HA |
1757 | /* Set the pressure threshold up by the same strategy of TCP. It is a |
1758 | * fraction of global memory that is up to 1/2 at 256 MB, decreasing | |
1759 | * toward zero with the amount of memory, with a floor of 128 pages. | |
1760 | */ | |
1761 | limit = min(nr_all_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT); | |
1762 | limit = (limit * (nr_all_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11); | |
1763 | limit = max(limit, 128UL); | |
1764 | sysctl_udp_mem[0] = limit / 4 * 3; | |
1765 | sysctl_udp_mem[1] = limit; | |
1766 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
1767 | ||
1768 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
1769 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
1770 | } | |
1771 | ||
1da177e4 | 1772 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1773 | EXPORT_SYMBOL(udp_ioctl); |
db8dac20 DM |
1774 | EXPORT_SYMBOL(udp_prot); |
1775 | EXPORT_SYMBOL(udp_sendmsg); | |
4c0a6cb0 GR |
1776 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1777 | EXPORT_SYMBOL(udp_lib_setsockopt); | |
1da177e4 | 1778 | EXPORT_SYMBOL(udp_poll); |
6ba5a3c5 | 1779 | EXPORT_SYMBOL(udp_lib_get_port); |
1da177e4 LT |
1780 | |
1781 | #ifdef CONFIG_PROC_FS | |
1782 | EXPORT_SYMBOL(udp_proc_register); | |
1783 | EXPORT_SYMBOL(udp_proc_unregister); | |
1784 | #endif |