Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * The User Datagram Protocol (UDP). | |
7 | * | |
02c30a84 | 8 | * Authors: Ross Biro |
1da177e4 LT |
9 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
10 | * Arnt Gulbrandsen, <agulbra@nvg.unit.no> | |
113aa838 | 11 | * Alan Cox, <alan@lxorguk.ukuu.org.uk> |
1da177e4 LT |
12 | * Hirokazu Takahashi, <taka@valinux.co.jp> |
13 | * | |
14 | * Fixes: | |
15 | * Alan Cox : verify_area() calls | |
16 | * Alan Cox : stopped close while in use off icmp | |
17 | * messages. Not a fix but a botch that | |
18 | * for udp at least is 'valid'. | |
19 | * Alan Cox : Fixed icmp handling properly | |
20 | * Alan Cox : Correct error for oversized datagrams | |
e905a9ed YH |
21 | * Alan Cox : Tidied select() semantics. |
22 | * Alan Cox : udp_err() fixed properly, also now | |
1da177e4 LT |
23 | * select and read wake correctly on errors |
24 | * Alan Cox : udp_send verify_area moved to avoid mem leak | |
25 | * Alan Cox : UDP can count its memory | |
26 | * Alan Cox : send to an unknown connection causes | |
27 | * an ECONNREFUSED off the icmp, but | |
28 | * does NOT close. | |
29 | * Alan Cox : Switched to new sk_buff handlers. No more backlog! | |
30 | * Alan Cox : Using generic datagram code. Even smaller and the PEEK | |
31 | * bug no longer crashes it. | |
32 | * Fred Van Kempen : Net2e support for sk->broadcast. | |
33 | * Alan Cox : Uses skb_free_datagram | |
34 | * Alan Cox : Added get/set sockopt support. | |
35 | * Alan Cox : Broadcasting without option set returns EACCES. | |
36 | * Alan Cox : No wakeup calls. Instead we now use the callbacks. | |
37 | * Alan Cox : Use ip_tos and ip_ttl | |
38 | * Alan Cox : SNMP Mibs | |
39 | * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support. | |
40 | * Matt Dillon : UDP length checks. | |
41 | * Alan Cox : Smarter af_inet used properly. | |
42 | * Alan Cox : Use new kernel side addressing. | |
43 | * Alan Cox : Incorrect return on truncated datagram receive. | |
44 | * Arnt Gulbrandsen : New udp_send and stuff | |
45 | * Alan Cox : Cache last socket | |
46 | * Alan Cox : Route cache | |
47 | * Jon Peatfield : Minor efficiency fix to sendto(). | |
48 | * Mike Shaver : RFC1122 checks. | |
49 | * Alan Cox : Nonblocking error fix. | |
50 | * Willy Konynenberg : Transparent proxying support. | |
51 | * Mike McLagan : Routing by source | |
52 | * David S. Miller : New socket lookup architecture. | |
53 | * Last socket cache retained as it | |
54 | * does have a high hit rate. | |
55 | * Olaf Kirch : Don't linearise iovec on sendmsg. | |
56 | * Andi Kleen : Some cleanups, cache destination entry | |
e905a9ed | 57 | * for connect. |
1da177e4 LT |
58 | * Vitaly E. Lavrov : Transparent proxy revived after year coma. |
59 | * Melvin Smith : Check msg_name not msg_namelen in sendto(), | |
60 | * return ENOTCONN for unconnected sockets (POSIX) | |
61 | * Janos Farkas : don't deliver multi/broadcasts to a different | |
62 | * bound-to-device socket | |
63 | * Hirokazu Takahashi : HW checksumming for outgoing UDP | |
64 | * datagrams. | |
65 | * Hirokazu Takahashi : sendfile() on UDP works now. | |
66 | * Arnaldo C. Melo : convert /proc/net/udp to seq_file | |
67 | * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which | |
68 | * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind | |
69 | * a single port at the same time. | |
70 | * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support | |
342f0234 | 71 | * James Chapman : Add L2TP encapsulation type. |
1da177e4 LT |
72 | * |
73 | * | |
74 | * This program is free software; you can redistribute it and/or | |
75 | * modify it under the terms of the GNU General Public License | |
76 | * as published by the Free Software Foundation; either version | |
77 | * 2 of the License, or (at your option) any later version. | |
78 | */ | |
e905a9ed | 79 | |
1da177e4 LT |
80 | #include <asm/system.h> |
81 | #include <asm/uaccess.h> | |
82 | #include <asm/ioctls.h> | |
95766fff | 83 | #include <linux/bootmem.h> |
8203efb3 ED |
84 | #include <linux/highmem.h> |
85 | #include <linux/swap.h> | |
1da177e4 LT |
86 | #include <linux/types.h> |
87 | #include <linux/fcntl.h> | |
88 | #include <linux/module.h> | |
89 | #include <linux/socket.h> | |
90 | #include <linux/sockios.h> | |
14c85021 | 91 | #include <linux/igmp.h> |
1da177e4 LT |
92 | #include <linux/in.h> |
93 | #include <linux/errno.h> | |
94 | #include <linux/timer.h> | |
95 | #include <linux/mm.h> | |
1da177e4 | 96 | #include <linux/inet.h> |
1da177e4 | 97 | #include <linux/netdevice.h> |
c752f073 | 98 | #include <net/tcp_states.h> |
1da177e4 LT |
99 | #include <linux/skbuff.h> |
100 | #include <linux/proc_fs.h> | |
101 | #include <linux/seq_file.h> | |
457c4cbc | 102 | #include <net/net_namespace.h> |
1da177e4 LT |
103 | #include <net/icmp.h> |
104 | #include <net/route.h> | |
1da177e4 LT |
105 | #include <net/checksum.h> |
106 | #include <net/xfrm.h> | |
ba4e58ec | 107 | #include "udp_impl.h" |
1da177e4 | 108 | |
645ca708 ED |
109 | struct udp_table udp_table; |
110 | EXPORT_SYMBOL(udp_table); | |
1da177e4 | 111 | |
95766fff HA |
112 | int sysctl_udp_mem[3] __read_mostly; |
113 | int sysctl_udp_rmem_min __read_mostly; | |
114 | int sysctl_udp_wmem_min __read_mostly; | |
115 | ||
116 | EXPORT_SYMBOL(sysctl_udp_mem); | |
117 | EXPORT_SYMBOL(sysctl_udp_rmem_min); | |
118 | EXPORT_SYMBOL(sysctl_udp_wmem_min); | |
119 | ||
120 | atomic_t udp_memory_allocated; | |
121 | EXPORT_SYMBOL(udp_memory_allocated); | |
122 | ||
f24d43c0 | 123 | static int udp_lib_lport_inuse(struct net *net, __u16 num, |
645ca708 | 124 | const struct udp_hslot *hslot, |
f24d43c0 ED |
125 | struct sock *sk, |
126 | int (*saddr_comp)(const struct sock *sk1, | |
127 | const struct sock *sk2)) | |
1da177e4 | 128 | { |
f24d43c0 | 129 | struct sock *sk2; |
88ab1932 | 130 | struct hlist_nulls_node *node; |
25030a7f | 131 | |
88ab1932 | 132 | sk_nulls_for_each(sk2, node, &hslot->head) |
f24d43c0 ED |
133 | if (net_eq(sock_net(sk2), net) && |
134 | sk2 != sk && | |
135 | sk2->sk_hash == num && | |
136 | (!sk2->sk_reuse || !sk->sk_reuse) && | |
137 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if | |
138 | || sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
139 | (*saddr_comp)(sk, sk2)) | |
25030a7f GR |
140 | return 1; |
141 | return 0; | |
142 | } | |
143 | ||
144 | /** | |
6ba5a3c5 | 145 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
146 | * |
147 | * @sk: socket struct in question | |
148 | * @snum: port number to look up | |
df2bc459 | 149 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25030a7f | 150 | */ |
6ba5a3c5 | 151 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
df2bc459 DM |
152 | int (*saddr_comp)(const struct sock *sk1, |
153 | const struct sock *sk2 ) ) | |
25030a7f | 154 | { |
645ca708 ED |
155 | struct udp_hslot *hslot; |
156 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
25030a7f | 157 | int error = 1; |
3b1e0a65 | 158 | struct net *net = sock_net(sk); |
1da177e4 | 159 | |
32c1da70 | 160 | if (!snum) { |
9088c560 ED |
161 | int low, high, remaining; |
162 | unsigned rand; | |
163 | unsigned short first; | |
32c1da70 | 164 | |
227b60f5 | 165 | inet_get_local_port_range(&low, &high); |
a25de534 | 166 | remaining = (high - low) + 1; |
227b60f5 | 167 | |
9088c560 ED |
168 | rand = net_random(); |
169 | snum = first = rand % remaining + low; | |
170 | rand |= 1; | |
645ca708 ED |
171 | for (;;) { |
172 | hslot = &udptable->hash[udp_hashfn(net, snum)]; | |
173 | spin_lock_bh(&hslot->lock); | |
174 | if (!udp_lib_lport_inuse(net, snum, hslot, sk, saddr_comp)) | |
175 | break; | |
176 | spin_unlock_bh(&hslot->lock); | |
9088c560 ED |
177 | do { |
178 | snum = snum + rand; | |
179 | } while (snum < low || snum > high); | |
180 | if (snum == first) | |
181 | goto fail; | |
1da177e4 | 182 | } |
645ca708 ED |
183 | } else { |
184 | hslot = &udptable->hash[udp_hashfn(net, snum)]; | |
185 | spin_lock_bh(&hslot->lock); | |
186 | if (udp_lib_lport_inuse(net, snum, hslot, sk, saddr_comp)) | |
187 | goto fail_unlock; | |
188 | } | |
25030a7f | 189 | inet_sk(sk)->num = snum; |
df2bc459 | 190 | sk->sk_hash = snum; |
1da177e4 | 191 | if (sk_unhashed(sk)) { |
88ab1932 | 192 | sk_nulls_add_node_rcu(sk, &hslot->head); |
c29a0bc4 | 193 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
1da177e4 | 194 | } |
25030a7f | 195 | error = 0; |
645ca708 ED |
196 | fail_unlock: |
197 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 198 | fail: |
25030a7f GR |
199 | return error; |
200 | } | |
201 | ||
6ba5a3c5 | 202 | static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2) |
db8dac20 DM |
203 | { |
204 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
205 | ||
206 | return ( !ipv6_only_sock(sk2) && | |
207 | (!inet1->rcv_saddr || !inet2->rcv_saddr || | |
208 | inet1->rcv_saddr == inet2->rcv_saddr )); | |
209 | } | |
210 | ||
6ba5a3c5 | 211 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 212 | { |
6ba5a3c5 | 213 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal); |
db8dac20 DM |
214 | } |
215 | ||
645ca708 ED |
216 | static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr, |
217 | unsigned short hnum, | |
218 | __be16 sport, __be32 daddr, __be16 dport, int dif) | |
219 | { | |
220 | int score = -1; | |
221 | ||
222 | if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum && | |
223 | !ipv6_only_sock(sk)) { | |
224 | struct inet_sock *inet = inet_sk(sk); | |
225 | ||
226 | score = (sk->sk_family == PF_INET ? 1 : 0); | |
227 | if (inet->rcv_saddr) { | |
228 | if (inet->rcv_saddr != daddr) | |
229 | return -1; | |
230 | score += 2; | |
231 | } | |
232 | if (inet->daddr) { | |
233 | if (inet->daddr != saddr) | |
234 | return -1; | |
235 | score += 2; | |
236 | } | |
237 | if (inet->dport) { | |
238 | if (inet->dport != sport) | |
239 | return -1; | |
240 | score += 2; | |
241 | } | |
242 | if (sk->sk_bound_dev_if) { | |
243 | if (sk->sk_bound_dev_if != dif) | |
244 | return -1; | |
245 | score += 2; | |
246 | } | |
247 | } | |
248 | return score; | |
249 | } | |
250 | ||
db8dac20 DM |
251 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
252 | * harder than this. -DaveM | |
253 | */ | |
254 | static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, | |
255 | __be16 sport, __be32 daddr, __be16 dport, | |
645ca708 | 256 | int dif, struct udp_table *udptable) |
db8dac20 | 257 | { |
271b72c7 | 258 | struct sock *sk, *result; |
88ab1932 | 259 | struct hlist_nulls_node *node; |
db8dac20 | 260 | unsigned short hnum = ntohs(dport); |
645ca708 ED |
261 | unsigned int hash = udp_hashfn(net, hnum); |
262 | struct udp_hslot *hslot = &udptable->hash[hash]; | |
271b72c7 | 263 | int score, badness; |
645ca708 | 264 | |
271b72c7 ED |
265 | rcu_read_lock(); |
266 | begin: | |
267 | result = NULL; | |
268 | badness = -1; | |
88ab1932 | 269 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { |
645ca708 ED |
270 | score = compute_score(sk, net, saddr, hnum, sport, |
271 | daddr, dport, dif); | |
272 | if (score > badness) { | |
273 | result = sk; | |
274 | badness = score; | |
db8dac20 DM |
275 | } |
276 | } | |
88ab1932 ED |
277 | /* |
278 | * if the nulls value we got at the end of this lookup is | |
279 | * not the expected one, we must restart lookup. | |
280 | * We probably met an item that was moved to another chain. | |
281 | */ | |
282 | if (get_nulls_value(node) != hash) | |
283 | goto begin; | |
284 | ||
271b72c7 ED |
285 | if (result) { |
286 | if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt))) | |
287 | result = NULL; | |
288 | else if (unlikely(compute_score(result, net, saddr, hnum, sport, | |
289 | daddr, dport, dif) < badness)) { | |
290 | sock_put(result); | |
291 | goto begin; | |
292 | } | |
293 | } | |
294 | rcu_read_unlock(); | |
db8dac20 DM |
295 | return result; |
296 | } | |
297 | ||
607c4aaf KK |
298 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
299 | __be16 sport, __be16 dport, | |
645ca708 | 300 | struct udp_table *udptable) |
607c4aaf | 301 | { |
23542618 | 302 | struct sock *sk; |
607c4aaf KK |
303 | const struct iphdr *iph = ip_hdr(skb); |
304 | ||
23542618 KK |
305 | if (unlikely(sk = skb_steal_sock(skb))) |
306 | return sk; | |
307 | else | |
308 | return __udp4_lib_lookup(dev_net(skb->dst->dev), iph->saddr, sport, | |
309 | iph->daddr, dport, inet_iif(skb), | |
310 | udptable); | |
607c4aaf KK |
311 | } |
312 | ||
bcd41303 KK |
313 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
314 | __be32 daddr, __be16 dport, int dif) | |
315 | { | |
645ca708 | 316 | return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table); |
bcd41303 KK |
317 | } |
318 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
319 | ||
920a4611 | 320 | static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk, |
db8dac20 DM |
321 | __be16 loc_port, __be32 loc_addr, |
322 | __be16 rmt_port, __be32 rmt_addr, | |
323 | int dif) | |
324 | { | |
88ab1932 | 325 | struct hlist_nulls_node *node; |
db8dac20 DM |
326 | struct sock *s = sk; |
327 | unsigned short hnum = ntohs(loc_port); | |
328 | ||
88ab1932 | 329 | sk_nulls_for_each_from(s, node) { |
db8dac20 DM |
330 | struct inet_sock *inet = inet_sk(s); |
331 | ||
920a4611 ED |
332 | if (!net_eq(sock_net(s), net) || |
333 | s->sk_hash != hnum || | |
db8dac20 DM |
334 | (inet->daddr && inet->daddr != rmt_addr) || |
335 | (inet->dport != rmt_port && inet->dport) || | |
336 | (inet->rcv_saddr && inet->rcv_saddr != loc_addr) || | |
337 | ipv6_only_sock(s) || | |
338 | (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)) | |
339 | continue; | |
340 | if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif)) | |
341 | continue; | |
342 | goto found; | |
343 | } | |
344 | s = NULL; | |
345 | found: | |
346 | return s; | |
347 | } | |
348 | ||
349 | /* | |
350 | * This routine is called by the ICMP module when it gets some | |
351 | * sort of error condition. If err < 0 then the socket should | |
352 | * be closed and the error returned to the user. If err > 0 | |
353 | * it's just the icmp type << 8 | icmp code. | |
354 | * Header points to the ip header of the error packet. We move | |
355 | * on past this. Then (as it used to claim before adjustment) | |
356 | * header points to the first 8 bytes of the udp header. We need | |
357 | * to find the appropriate port. | |
358 | */ | |
359 | ||
645ca708 | 360 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
361 | { |
362 | struct inet_sock *inet; | |
363 | struct iphdr *iph = (struct iphdr*)skb->data; | |
364 | struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2)); | |
365 | const int type = icmp_hdr(skb)->type; | |
366 | const int code = icmp_hdr(skb)->code; | |
367 | struct sock *sk; | |
368 | int harderr; | |
369 | int err; | |
fd54d716 | 370 | struct net *net = dev_net(skb->dev); |
db8dac20 | 371 | |
fd54d716 | 372 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
db8dac20 DM |
373 | iph->saddr, uh->source, skb->dev->ifindex, udptable); |
374 | if (sk == NULL) { | |
dcfc23ca | 375 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
376 | return; /* No socket for error */ |
377 | } | |
378 | ||
379 | err = 0; | |
380 | harderr = 0; | |
381 | inet = inet_sk(sk); | |
382 | ||
383 | switch (type) { | |
384 | default: | |
385 | case ICMP_TIME_EXCEEDED: | |
386 | err = EHOSTUNREACH; | |
387 | break; | |
388 | case ICMP_SOURCE_QUENCH: | |
389 | goto out; | |
390 | case ICMP_PARAMETERPROB: | |
391 | err = EPROTO; | |
392 | harderr = 1; | |
393 | break; | |
394 | case ICMP_DEST_UNREACH: | |
395 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
396 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { | |
397 | err = EMSGSIZE; | |
398 | harderr = 1; | |
399 | break; | |
400 | } | |
401 | goto out; | |
402 | } | |
403 | err = EHOSTUNREACH; | |
404 | if (code <= NR_ICMP_UNREACH) { | |
405 | harderr = icmp_err_convert[code].fatal; | |
406 | err = icmp_err_convert[code].errno; | |
407 | } | |
408 | break; | |
409 | } | |
410 | ||
411 | /* | |
412 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
413 | * 4.1.3.3. | |
414 | */ | |
415 | if (!inet->recverr) { | |
416 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
417 | goto out; | |
418 | } else { | |
419 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1)); | |
420 | } | |
421 | sk->sk_err = err; | |
422 | sk->sk_error_report(sk); | |
423 | out: | |
424 | sock_put(sk); | |
425 | } | |
426 | ||
427 | void udp_err(struct sk_buff *skb, u32 info) | |
428 | { | |
645ca708 | 429 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
430 | } |
431 | ||
432 | /* | |
433 | * Throw away all pending data and cancel the corking. Socket is locked. | |
434 | */ | |
36d926b9 | 435 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
436 | { |
437 | struct udp_sock *up = udp_sk(sk); | |
438 | ||
439 | if (up->pending) { | |
440 | up->len = 0; | |
441 | up->pending = 0; | |
442 | ip_flush_pending_frames(sk); | |
443 | } | |
444 | } | |
36d926b9 | 445 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
446 | |
447 | /** | |
448 | * udp4_hwcsum_outgoing - handle outgoing HW checksumming | |
449 | * @sk: socket we are sending on | |
450 | * @skb: sk_buff containing the filled-in UDP header | |
451 | * (checksum field must be zeroed out) | |
452 | */ | |
453 | static void udp4_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb, | |
454 | __be32 src, __be32 dst, int len ) | |
455 | { | |
456 | unsigned int offset; | |
457 | struct udphdr *uh = udp_hdr(skb); | |
458 | __wsum csum = 0; | |
459 | ||
460 | if (skb_queue_len(&sk->sk_write_queue) == 1) { | |
461 | /* | |
462 | * Only one fragment on the socket. | |
463 | */ | |
464 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
465 | skb->csum_offset = offsetof(struct udphdr, check); | |
466 | uh->check = ~csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, 0); | |
467 | } else { | |
468 | /* | |
469 | * HW-checksum won't work as there are two or more | |
470 | * fragments on the socket so that all csums of sk_buffs | |
471 | * should be together | |
472 | */ | |
473 | offset = skb_transport_offset(skb); | |
474 | skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); | |
475 | ||
476 | skb->ip_summed = CHECKSUM_NONE; | |
477 | ||
478 | skb_queue_walk(&sk->sk_write_queue, skb) { | |
479 | csum = csum_add(csum, skb->csum); | |
480 | } | |
481 | ||
482 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); | |
483 | if (uh->check == 0) | |
484 | uh->check = CSUM_MANGLED_0; | |
485 | } | |
486 | } | |
487 | ||
488 | /* | |
489 | * Push out all pending data as one UDP datagram. Socket is locked. | |
490 | */ | |
491 | static int udp_push_pending_frames(struct sock *sk) | |
492 | { | |
493 | struct udp_sock *up = udp_sk(sk); | |
494 | struct inet_sock *inet = inet_sk(sk); | |
495 | struct flowi *fl = &inet->cork.fl; | |
496 | struct sk_buff *skb; | |
497 | struct udphdr *uh; | |
498 | int err = 0; | |
499 | int is_udplite = IS_UDPLITE(sk); | |
500 | __wsum csum = 0; | |
501 | ||
502 | /* Grab the skbuff where UDP header space exists. */ | |
503 | if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) | |
504 | goto out; | |
505 | ||
506 | /* | |
507 | * Create a UDP header | |
508 | */ | |
509 | uh = udp_hdr(skb); | |
510 | uh->source = fl->fl_ip_sport; | |
511 | uh->dest = fl->fl_ip_dport; | |
512 | uh->len = htons(up->len); | |
513 | uh->check = 0; | |
514 | ||
515 | if (is_udplite) /* UDP-Lite */ | |
516 | csum = udplite_csum_outgoing(sk, skb); | |
517 | ||
518 | else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */ | |
519 | ||
520 | skb->ip_summed = CHECKSUM_NONE; | |
521 | goto send; | |
522 | ||
523 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
524 | ||
525 | udp4_hwcsum_outgoing(sk, skb, fl->fl4_src,fl->fl4_dst, up->len); | |
526 | goto send; | |
527 | ||
528 | } else /* `normal' UDP */ | |
529 | csum = udp_csum_outgoing(sk, skb); | |
530 | ||
531 | /* add protocol-dependent pseudo-header */ | |
532 | uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, up->len, | |
533 | sk->sk_protocol, csum ); | |
534 | if (uh->check == 0) | |
535 | uh->check = CSUM_MANGLED_0; | |
536 | ||
537 | send: | |
538 | err = ip_push_pending_frames(sk); | |
539 | out: | |
540 | up->len = 0; | |
541 | up->pending = 0; | |
542 | if (!err) | |
629ca23c PE |
543 | UDP_INC_STATS_USER(sock_net(sk), |
544 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
db8dac20 DM |
545 | return err; |
546 | } | |
547 | ||
548 | int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
549 | size_t len) | |
550 | { | |
551 | struct inet_sock *inet = inet_sk(sk); | |
552 | struct udp_sock *up = udp_sk(sk); | |
553 | int ulen = len; | |
554 | struct ipcm_cookie ipc; | |
555 | struct rtable *rt = NULL; | |
556 | int free = 0; | |
557 | int connected = 0; | |
558 | __be32 daddr, faddr, saddr; | |
559 | __be16 dport; | |
560 | u8 tos; | |
561 | int err, is_udplite = IS_UDPLITE(sk); | |
562 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
563 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
564 | ||
565 | if (len > 0xFFFF) | |
566 | return -EMSGSIZE; | |
567 | ||
568 | /* | |
569 | * Check the flags. | |
570 | */ | |
571 | ||
572 | if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */ | |
573 | return -EOPNOTSUPP; | |
574 | ||
575 | ipc.opt = NULL; | |
576 | ||
577 | if (up->pending) { | |
578 | /* | |
579 | * There are pending frames. | |
580 | * The socket lock must be held while it's corked. | |
581 | */ | |
582 | lock_sock(sk); | |
583 | if (likely(up->pending)) { | |
584 | if (unlikely(up->pending != AF_INET)) { | |
585 | release_sock(sk); | |
586 | return -EINVAL; | |
587 | } | |
588 | goto do_append_data; | |
589 | } | |
590 | release_sock(sk); | |
591 | } | |
592 | ulen += sizeof(struct udphdr); | |
593 | ||
594 | /* | |
595 | * Get and verify the address. | |
596 | */ | |
597 | if (msg->msg_name) { | |
598 | struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name; | |
599 | if (msg->msg_namelen < sizeof(*usin)) | |
600 | return -EINVAL; | |
601 | if (usin->sin_family != AF_INET) { | |
602 | if (usin->sin_family != AF_UNSPEC) | |
603 | return -EAFNOSUPPORT; | |
604 | } | |
605 | ||
606 | daddr = usin->sin_addr.s_addr; | |
607 | dport = usin->sin_port; | |
608 | if (dport == 0) | |
609 | return -EINVAL; | |
610 | } else { | |
611 | if (sk->sk_state != TCP_ESTABLISHED) | |
612 | return -EDESTADDRREQ; | |
613 | daddr = inet->daddr; | |
614 | dport = inet->dport; | |
615 | /* Open fast path for connected socket. | |
616 | Route will not be used, if at least one option is set. | |
617 | */ | |
618 | connected = 1; | |
619 | } | |
620 | ipc.addr = inet->saddr; | |
621 | ||
622 | ipc.oif = sk->sk_bound_dev_if; | |
623 | if (msg->msg_controllen) { | |
3b1e0a65 | 624 | err = ip_cmsg_send(sock_net(sk), msg, &ipc); |
db8dac20 DM |
625 | if (err) |
626 | return err; | |
627 | if (ipc.opt) | |
628 | free = 1; | |
629 | connected = 0; | |
630 | } | |
631 | if (!ipc.opt) | |
632 | ipc.opt = inet->opt; | |
633 | ||
634 | saddr = ipc.addr; | |
635 | ipc.addr = faddr = daddr; | |
636 | ||
637 | if (ipc.opt && ipc.opt->srr) { | |
638 | if (!daddr) | |
639 | return -EINVAL; | |
640 | faddr = ipc.opt->faddr; | |
641 | connected = 0; | |
642 | } | |
643 | tos = RT_TOS(inet->tos); | |
644 | if (sock_flag(sk, SOCK_LOCALROUTE) || | |
645 | (msg->msg_flags & MSG_DONTROUTE) || | |
646 | (ipc.opt && ipc.opt->is_strictroute)) { | |
647 | tos |= RTO_ONLINK; | |
648 | connected = 0; | |
649 | } | |
650 | ||
651 | if (ipv4_is_multicast(daddr)) { | |
652 | if (!ipc.oif) | |
653 | ipc.oif = inet->mc_index; | |
654 | if (!saddr) | |
655 | saddr = inet->mc_addr; | |
656 | connected = 0; | |
657 | } | |
658 | ||
659 | if (connected) | |
660 | rt = (struct rtable*)sk_dst_check(sk, 0); | |
661 | ||
662 | if (rt == NULL) { | |
663 | struct flowi fl = { .oif = ipc.oif, | |
664 | .nl_u = { .ip4_u = | |
665 | { .daddr = faddr, | |
666 | .saddr = saddr, | |
667 | .tos = tos } }, | |
668 | .proto = sk->sk_protocol, | |
a134f85c | 669 | .flags = inet_sk_flowi_flags(sk), |
db8dac20 DM |
670 | .uli_u = { .ports = |
671 | { .sport = inet->sport, | |
672 | .dport = dport } } }; | |
84a3aa00 PE |
673 | struct net *net = sock_net(sk); |
674 | ||
db8dac20 | 675 | security_sk_classify_flow(sk, &fl); |
84a3aa00 | 676 | err = ip_route_output_flow(net, &rt, &fl, sk, 1); |
db8dac20 DM |
677 | if (err) { |
678 | if (err == -ENETUNREACH) | |
7c73a6fa | 679 | IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
680 | goto out; |
681 | } | |
682 | ||
683 | err = -EACCES; | |
684 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
685 | !sock_flag(sk, SOCK_BROADCAST)) | |
686 | goto out; | |
687 | if (connected) | |
688 | sk_dst_set(sk, dst_clone(&rt->u.dst)); | |
689 | } | |
690 | ||
691 | if (msg->msg_flags&MSG_CONFIRM) | |
692 | goto do_confirm; | |
693 | back_from_confirm: | |
694 | ||
695 | saddr = rt->rt_src; | |
696 | if (!ipc.addr) | |
697 | daddr = ipc.addr = rt->rt_dst; | |
698 | ||
699 | lock_sock(sk); | |
700 | if (unlikely(up->pending)) { | |
701 | /* The socket is already corked while preparing it. */ | |
702 | /* ... which is an evident application bug. --ANK */ | |
703 | release_sock(sk); | |
704 | ||
705 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n"); | |
706 | err = -EINVAL; | |
707 | goto out; | |
708 | } | |
709 | /* | |
710 | * Now cork the socket to pend data. | |
711 | */ | |
712 | inet->cork.fl.fl4_dst = daddr; | |
713 | inet->cork.fl.fl_ip_dport = dport; | |
714 | inet->cork.fl.fl4_src = saddr; | |
715 | inet->cork.fl.fl_ip_sport = inet->sport; | |
716 | up->pending = AF_INET; | |
717 | ||
718 | do_append_data: | |
719 | up->len += ulen; | |
720 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; | |
721 | err = ip_append_data(sk, getfrag, msg->msg_iov, ulen, | |
2e77d89b | 722 | sizeof(struct udphdr), &ipc, &rt, |
db8dac20 DM |
723 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); |
724 | if (err) | |
725 | udp_flush_pending_frames(sk); | |
726 | else if (!corkreq) | |
727 | err = udp_push_pending_frames(sk); | |
728 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
729 | up->pending = 0; | |
730 | release_sock(sk); | |
731 | ||
732 | out: | |
733 | ip_rt_put(rt); | |
734 | if (free) | |
735 | kfree(ipc.opt); | |
736 | if (!err) | |
737 | return len; | |
738 | /* | |
739 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
740 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
741 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
742 | * things). We could add another new stat but at least for now that | |
743 | * seems like overkill. | |
744 | */ | |
745 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
746 | UDP_INC_STATS_USER(sock_net(sk), |
747 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
748 | } |
749 | return err; | |
750 | ||
751 | do_confirm: | |
752 | dst_confirm(&rt->u.dst); | |
753 | if (!(msg->msg_flags&MSG_PROBE) || len) | |
754 | goto back_from_confirm; | |
755 | err = 0; | |
756 | goto out; | |
757 | } | |
758 | ||
759 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
760 | size_t size, int flags) | |
761 | { | |
762 | struct udp_sock *up = udp_sk(sk); | |
763 | int ret; | |
764 | ||
765 | if (!up->pending) { | |
766 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
767 | ||
768 | /* Call udp_sendmsg to specify destination address which | |
769 | * sendpage interface can't pass. | |
770 | * This will succeed only when the socket is connected. | |
771 | */ | |
772 | ret = udp_sendmsg(NULL, sk, &msg, 0); | |
773 | if (ret < 0) | |
774 | return ret; | |
775 | } | |
776 | ||
777 | lock_sock(sk); | |
778 | ||
779 | if (unlikely(!up->pending)) { | |
780 | release_sock(sk); | |
781 | ||
782 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n"); | |
783 | return -EINVAL; | |
784 | } | |
785 | ||
786 | ret = ip_append_page(sk, page, offset, size, flags); | |
787 | if (ret == -EOPNOTSUPP) { | |
788 | release_sock(sk); | |
789 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
790 | size, flags); | |
791 | } | |
792 | if (ret < 0) { | |
793 | udp_flush_pending_frames(sk); | |
794 | goto out; | |
795 | } | |
796 | ||
797 | up->len += size; | |
798 | if (!(up->corkflag || (flags&MSG_MORE))) | |
799 | ret = udp_push_pending_frames(sk); | |
800 | if (!ret) | |
801 | ret = size; | |
802 | out: | |
803 | release_sock(sk); | |
804 | return ret; | |
805 | } | |
806 | ||
1da177e4 LT |
807 | /* |
808 | * IOCTL requests applicable to the UDP protocol | |
809 | */ | |
e905a9ed | 810 | |
1da177e4 LT |
811 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
812 | { | |
6516c655 SH |
813 | switch (cmd) { |
814 | case SIOCOUTQ: | |
1da177e4 | 815 | { |
6516c655 SH |
816 | int amount = atomic_read(&sk->sk_wmem_alloc); |
817 | return put_user(amount, (int __user *)arg); | |
818 | } | |
1da177e4 | 819 | |
6516c655 SH |
820 | case SIOCINQ: |
821 | { | |
822 | struct sk_buff *skb; | |
823 | unsigned long amount; | |
824 | ||
825 | amount = 0; | |
826 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
827 | skb = skb_peek(&sk->sk_receive_queue); | |
828 | if (skb != NULL) { | |
829 | /* | |
830 | * We will only return the amount | |
831 | * of this packet since that is all | |
832 | * that will be read. | |
833 | */ | |
834 | amount = skb->len - sizeof(struct udphdr); | |
1da177e4 | 835 | } |
6516c655 SH |
836 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
837 | return put_user(amount, (int __user *)arg); | |
838 | } | |
1da177e4 | 839 | |
6516c655 SH |
840 | default: |
841 | return -ENOIOCTLCMD; | |
1da177e4 | 842 | } |
6516c655 SH |
843 | |
844 | return 0; | |
1da177e4 LT |
845 | } |
846 | ||
db8dac20 DM |
847 | /* |
848 | * This should be easy, if there is something there we | |
849 | * return it, otherwise we block. | |
850 | */ | |
851 | ||
852 | int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
853 | size_t len, int noblock, int flags, int *addr_len) | |
854 | { | |
855 | struct inet_sock *inet = inet_sk(sk); | |
856 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; | |
857 | struct sk_buff *skb; | |
858 | unsigned int ulen, copied; | |
859 | int peeked; | |
860 | int err; | |
861 | int is_udplite = IS_UDPLITE(sk); | |
862 | ||
863 | /* | |
864 | * Check any passed addresses | |
865 | */ | |
866 | if (addr_len) | |
867 | *addr_len=sizeof(*sin); | |
868 | ||
869 | if (flags & MSG_ERRQUEUE) | |
870 | return ip_recv_error(sk, msg, len); | |
871 | ||
872 | try_again: | |
873 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), | |
874 | &peeked, &err); | |
875 | if (!skb) | |
876 | goto out; | |
877 | ||
878 | ulen = skb->len - sizeof(struct udphdr); | |
879 | copied = len; | |
880 | if (copied > ulen) | |
881 | copied = ulen; | |
882 | else if (copied < ulen) | |
883 | msg->msg_flags |= MSG_TRUNC; | |
884 | ||
885 | /* | |
886 | * If checksum is needed at all, try to do it while copying the | |
887 | * data. If the data is truncated, or if we only want a partial | |
888 | * coverage checksum (UDP-Lite), do it before the copy. | |
889 | */ | |
890 | ||
891 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { | |
892 | if (udp_lib_checksum_complete(skb)) | |
893 | goto csum_copy_err; | |
894 | } | |
895 | ||
896 | if (skb_csum_unnecessary(skb)) | |
897 | err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), | |
898 | msg->msg_iov, copied ); | |
899 | else { | |
900 | err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov); | |
901 | ||
902 | if (err == -EINVAL) | |
903 | goto csum_copy_err; | |
904 | } | |
905 | ||
906 | if (err) | |
907 | goto out_free; | |
908 | ||
909 | if (!peeked) | |
629ca23c PE |
910 | UDP_INC_STATS_USER(sock_net(sk), |
911 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 DM |
912 | |
913 | sock_recv_timestamp(msg, sk, skb); | |
914 | ||
915 | /* Copy the address. */ | |
916 | if (sin) | |
917 | { | |
918 | sin->sin_family = AF_INET; | |
919 | sin->sin_port = udp_hdr(skb)->source; | |
920 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
921 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
922 | } | |
923 | if (inet->cmsg_flags) | |
924 | ip_cmsg_recv(msg, skb); | |
925 | ||
926 | err = copied; | |
927 | if (flags & MSG_TRUNC) | |
928 | err = ulen; | |
929 | ||
930 | out_free: | |
931 | lock_sock(sk); | |
932 | skb_free_datagram(sk, skb); | |
933 | release_sock(sk); | |
934 | out: | |
935 | return err; | |
936 | ||
937 | csum_copy_err: | |
938 | lock_sock(sk); | |
939 | if (!skb_kill_datagram(sk, skb, flags)) | |
629ca23c | 940 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
db8dac20 DM |
941 | release_sock(sk); |
942 | ||
943 | if (noblock) | |
944 | return -EAGAIN; | |
945 | goto try_again; | |
946 | } | |
947 | ||
948 | ||
1da177e4 LT |
949 | int udp_disconnect(struct sock *sk, int flags) |
950 | { | |
951 | struct inet_sock *inet = inet_sk(sk); | |
952 | /* | |
953 | * 1003.1g - break association. | |
954 | */ | |
e905a9ed | 955 | |
1da177e4 LT |
956 | sk->sk_state = TCP_CLOSE; |
957 | inet->daddr = 0; | |
958 | inet->dport = 0; | |
959 | sk->sk_bound_dev_if = 0; | |
960 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
961 | inet_reset_saddr(sk); | |
962 | ||
963 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
964 | sk->sk_prot->unhash(sk); | |
965 | inet->sport = 0; | |
966 | } | |
967 | sk_dst_reset(sk); | |
968 | return 0; | |
969 | } | |
970 | ||
645ca708 ED |
971 | void udp_lib_unhash(struct sock *sk) |
972 | { | |
723b4610 ED |
973 | if (sk_hashed(sk)) { |
974 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
975 | unsigned int hash = udp_hashfn(sock_net(sk), sk->sk_hash); | |
976 | struct udp_hslot *hslot = &udptable->hash[hash]; | |
645ca708 | 977 | |
723b4610 ED |
978 | spin_lock_bh(&hslot->lock); |
979 | if (sk_nulls_del_node_init_rcu(sk)) { | |
980 | inet_sk(sk)->num = 0; | |
981 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); | |
982 | } | |
983 | spin_unlock_bh(&hslot->lock); | |
645ca708 | 984 | } |
645ca708 ED |
985 | } |
986 | EXPORT_SYMBOL(udp_lib_unhash); | |
987 | ||
93821778 HX |
988 | static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
989 | { | |
990 | int is_udplite = IS_UDPLITE(sk); | |
991 | int rc; | |
992 | ||
993 | if ((rc = sock_queue_rcv_skb(sk, skb)) < 0) { | |
994 | /* Note that an ENOMEM error is charged twice */ | |
995 | if (rc == -ENOMEM) | |
996 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, | |
997 | is_udplite); | |
998 | goto drop; | |
999 | } | |
1000 | ||
1001 | return 0; | |
1002 | ||
1003 | drop: | |
1004 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); | |
1005 | kfree_skb(skb); | |
1006 | return -1; | |
1007 | } | |
1008 | ||
db8dac20 DM |
1009 | /* returns: |
1010 | * -1: error | |
1011 | * 0: success | |
1012 | * >0: "udp encap" protocol resubmission | |
1013 | * | |
1014 | * Note that in the success and error cases, the skb is assumed to | |
1015 | * have either been requeued or freed. | |
1016 | */ | |
1017 | int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb) | |
1018 | { | |
1019 | struct udp_sock *up = udp_sk(sk); | |
1020 | int rc; | |
1021 | int is_udplite = IS_UDPLITE(sk); | |
1022 | ||
1023 | /* | |
1024 | * Charge it to the socket, dropping if the queue is full. | |
1025 | */ | |
1026 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1027 | goto drop; | |
1028 | nf_reset(skb); | |
1029 | ||
1030 | if (up->encap_type) { | |
1031 | /* | |
1032 | * This is an encapsulation socket so pass the skb to | |
1033 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1034 | * fall through and pass this up the UDP socket. | |
1035 | * up->encap_rcv() returns the following value: | |
1036 | * =0 if skb was successfully passed to the encap | |
1037 | * handler or was discarded by it. | |
1038 | * >0 if skb should be passed on to UDP. | |
1039 | * <0 if skb should be resubmitted as proto -N | |
1040 | */ | |
1041 | ||
1042 | /* if we're overly short, let UDP handle it */ | |
1043 | if (skb->len > sizeof(struct udphdr) && | |
1044 | up->encap_rcv != NULL) { | |
1045 | int ret; | |
1046 | ||
1047 | ret = (*up->encap_rcv)(sk, skb); | |
1048 | if (ret <= 0) { | |
0283328e PE |
1049 | UDP_INC_STATS_BH(sock_net(sk), |
1050 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
1051 | is_udplite); |
1052 | return -ret; | |
1053 | } | |
1054 | } | |
1055 | ||
1056 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1057 | } | |
1058 | ||
1059 | /* | |
1060 | * UDP-Lite specific tests, ignored on UDP sockets | |
1061 | */ | |
1062 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1063 | ||
1064 | /* | |
1065 | * MIB statistics other than incrementing the error count are | |
1066 | * disabled for the following two types of errors: these depend | |
1067 | * on the application settings, not on the functioning of the | |
1068 | * protocol stack as such. | |
1069 | * | |
1070 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1071 | * way ... to ... at least let the receiving application block | |
1072 | * delivery of packets with coverage values less than a value | |
1073 | * provided by the application." | |
1074 | */ | |
1075 | if (up->pcrlen == 0) { /* full coverage was set */ | |
1076 | LIMIT_NETDEBUG(KERN_WARNING "UDPLITE: partial coverage " | |
1077 | "%d while full coverage %d requested\n", | |
1078 | UDP_SKB_CB(skb)->cscov, skb->len); | |
1079 | goto drop; | |
1080 | } | |
1081 | /* The next case involves violating the min. coverage requested | |
1082 | * by the receiver. This is subtle: if receiver wants x and x is | |
1083 | * greater than the buffersize/MTU then receiver will complain | |
1084 | * that it wants x while sender emits packets of smaller size y. | |
1085 | * Therefore the above ...()->partial_cov statement is essential. | |
1086 | */ | |
1087 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
1088 | LIMIT_NETDEBUG(KERN_WARNING | |
1089 | "UDPLITE: coverage %d too small, need min %d\n", | |
1090 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
1091 | goto drop; | |
1092 | } | |
1093 | } | |
1094 | ||
1095 | if (sk->sk_filter) { | |
1096 | if (udp_lib_checksum_complete(skb)) | |
1097 | goto drop; | |
1098 | } | |
1099 | ||
93821778 | 1100 | rc = 0; |
db8dac20 | 1101 | |
93821778 HX |
1102 | bh_lock_sock(sk); |
1103 | if (!sock_owned_by_user(sk)) | |
1104 | rc = __udp_queue_rcv_skb(sk, skb); | |
1105 | else | |
1106 | sk_add_backlog(sk, skb); | |
1107 | bh_unlock_sock(sk); | |
1108 | ||
1109 | return rc; | |
db8dac20 DM |
1110 | |
1111 | drop: | |
0283328e | 1112 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
db8dac20 DM |
1113 | kfree_skb(skb); |
1114 | return -1; | |
1115 | } | |
1116 | ||
1117 | /* | |
1118 | * Multicasts and broadcasts go to each listener. | |
1119 | * | |
1120 | * Note: called only from the BH handler context, | |
1121 | * so we don't need to lock the hashes. | |
1122 | */ | |
e3163493 | 1123 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1124 | struct udphdr *uh, |
1125 | __be32 saddr, __be32 daddr, | |
645ca708 | 1126 | struct udp_table *udptable) |
db8dac20 DM |
1127 | { |
1128 | struct sock *sk; | |
645ca708 | 1129 | struct udp_hslot *hslot = &udptable->hash[udp_hashfn(net, ntohs(uh->dest))]; |
db8dac20 DM |
1130 | int dif; |
1131 | ||
645ca708 | 1132 | spin_lock(&hslot->lock); |
88ab1932 | 1133 | sk = sk_nulls_head(&hslot->head); |
db8dac20 | 1134 | dif = skb->dev->ifindex; |
920a4611 | 1135 | sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif); |
db8dac20 DM |
1136 | if (sk) { |
1137 | struct sock *sknext = NULL; | |
1138 | ||
1139 | do { | |
1140 | struct sk_buff *skb1 = skb; | |
1141 | ||
88ab1932 | 1142 | sknext = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest, |
920a4611 ED |
1143 | daddr, uh->source, saddr, |
1144 | dif); | |
db8dac20 DM |
1145 | if (sknext) |
1146 | skb1 = skb_clone(skb, GFP_ATOMIC); | |
1147 | ||
1148 | if (skb1) { | |
93821778 | 1149 | int ret = udp_queue_rcv_skb(sk, skb1); |
db8dac20 DM |
1150 | if (ret > 0) |
1151 | /* we should probably re-process instead | |
1152 | * of dropping packets here. */ | |
1153 | kfree_skb(skb1); | |
1154 | } | |
1155 | sk = sknext; | |
1156 | } while (sknext); | |
1157 | } else | |
1158 | kfree_skb(skb); | |
645ca708 | 1159 | spin_unlock(&hslot->lock); |
db8dac20 DM |
1160 | return 0; |
1161 | } | |
1162 | ||
1163 | /* Initialize UDP checksum. If exited with zero value (success), | |
1164 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1165 | * Otherwise, csum completion requires chacksumming packet body, | |
1166 | * including udp header and folding it to skb->csum. | |
1167 | */ | |
1168 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1169 | int proto) | |
1170 | { | |
1171 | const struct iphdr *iph; | |
1172 | int err; | |
1173 | ||
1174 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1175 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1176 | ||
1177 | if (proto == IPPROTO_UDPLITE) { | |
1178 | err = udplite_checksum_init(skb, uh); | |
1179 | if (err) | |
1180 | return err; | |
1181 | } | |
1182 | ||
1183 | iph = ip_hdr(skb); | |
1184 | if (uh->check == 0) { | |
1185 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1186 | } else if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
1187 | if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, | |
1188 | proto, skb->csum)) | |
1189 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1190 | } | |
1191 | if (!skb_csum_unnecessary(skb)) | |
1192 | skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, | |
1193 | skb->len, proto, 0); | |
1194 | /* Probably, we should checksum udp header (it should be in cache | |
1195 | * in any case) and data in tiny packets (< rx copybreak). | |
1196 | */ | |
1197 | ||
1198 | return 0; | |
1199 | } | |
1200 | ||
1201 | /* | |
1202 | * All we need to do is get the socket, and then do a checksum. | |
1203 | */ | |
1204 | ||
645ca708 | 1205 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1206 | int proto) |
1207 | { | |
1208 | struct sock *sk; | |
1209 | struct udphdr *uh = udp_hdr(skb); | |
1210 | unsigned short ulen; | |
1211 | struct rtable *rt = (struct rtable*)skb->dst; | |
1212 | __be32 saddr = ip_hdr(skb)->saddr; | |
1213 | __be32 daddr = ip_hdr(skb)->daddr; | |
0283328e | 1214 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1215 | |
1216 | /* | |
1217 | * Validate the packet. | |
1218 | */ | |
1219 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1220 | goto drop; /* No space for header. */ | |
1221 | ||
1222 | ulen = ntohs(uh->len); | |
1223 | if (ulen > skb->len) | |
1224 | goto short_packet; | |
1225 | ||
1226 | if (proto == IPPROTO_UDP) { | |
1227 | /* UDP validates ulen. */ | |
1228 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1229 | goto short_packet; | |
1230 | uh = udp_hdr(skb); | |
1231 | } | |
1232 | ||
1233 | if (udp4_csum_init(skb, uh, proto)) | |
1234 | goto csum_error; | |
1235 | ||
1236 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) | |
e3163493 PE |
1237 | return __udp4_lib_mcast_deliver(net, skb, uh, |
1238 | saddr, daddr, udptable); | |
db8dac20 | 1239 | |
607c4aaf | 1240 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); |
db8dac20 DM |
1241 | |
1242 | if (sk != NULL) { | |
93821778 | 1243 | int ret = udp_queue_rcv_skb(sk, skb); |
db8dac20 DM |
1244 | sock_put(sk); |
1245 | ||
1246 | /* a return value > 0 means to resubmit the input, but | |
1247 | * it wants the return to be -protocol, or 0 | |
1248 | */ | |
1249 | if (ret > 0) | |
1250 | return -ret; | |
1251 | return 0; | |
1252 | } | |
1253 | ||
1254 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1255 | goto drop; | |
1256 | nf_reset(skb); | |
1257 | ||
1258 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1259 | if (udp_lib_checksum_complete(skb)) | |
1260 | goto csum_error; | |
1261 | ||
0283328e | 1262 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1263 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1264 | ||
1265 | /* | |
1266 | * Hmm. We got an UDP packet to a port to which we | |
1267 | * don't wanna listen. Ignore it. | |
1268 | */ | |
1269 | kfree_skb(skb); | |
1270 | return 0; | |
1271 | ||
1272 | short_packet: | |
673d57e7 | 1273 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n", |
db8dac20 | 1274 | proto == IPPROTO_UDPLITE ? "-Lite" : "", |
673d57e7 | 1275 | &saddr, |
db8dac20 DM |
1276 | ntohs(uh->source), |
1277 | ulen, | |
1278 | skb->len, | |
673d57e7 | 1279 | &daddr, |
db8dac20 DM |
1280 | ntohs(uh->dest)); |
1281 | goto drop; | |
1282 | ||
1283 | csum_error: | |
1284 | /* | |
1285 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1286 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1287 | */ | |
673d57e7 | 1288 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n", |
db8dac20 | 1289 | proto == IPPROTO_UDPLITE ? "-Lite" : "", |
673d57e7 | 1290 | &saddr, |
db8dac20 | 1291 | ntohs(uh->source), |
673d57e7 | 1292 | &daddr, |
db8dac20 DM |
1293 | ntohs(uh->dest), |
1294 | ulen); | |
1295 | drop: | |
0283328e | 1296 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1297 | kfree_skb(skb); |
1298 | return 0; | |
1299 | } | |
1300 | ||
1301 | int udp_rcv(struct sk_buff *skb) | |
1302 | { | |
645ca708 | 1303 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
1304 | } |
1305 | ||
7d06b2e0 | 1306 | void udp_destroy_sock(struct sock *sk) |
db8dac20 DM |
1307 | { |
1308 | lock_sock(sk); | |
1309 | udp_flush_pending_frames(sk); | |
1310 | release_sock(sk); | |
db8dac20 DM |
1311 | } |
1312 | ||
1da177e4 LT |
1313 | /* |
1314 | * Socket option code for UDP | |
1315 | */ | |
4c0a6cb0 GR |
1316 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
1317 | char __user *optval, int optlen, | |
1318 | int (*push_pending_frames)(struct sock *)) | |
1da177e4 LT |
1319 | { |
1320 | struct udp_sock *up = udp_sk(sk); | |
1321 | int val; | |
1322 | int err = 0; | |
b2bf1e26 | 1323 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 1324 | |
6516c655 | 1325 | if (optlen<sizeof(int)) |
1da177e4 LT |
1326 | return -EINVAL; |
1327 | ||
1328 | if (get_user(val, (int __user *)optval)) | |
1329 | return -EFAULT; | |
1330 | ||
6516c655 | 1331 | switch (optname) { |
1da177e4 LT |
1332 | case UDP_CORK: |
1333 | if (val != 0) { | |
1334 | up->corkflag = 1; | |
1335 | } else { | |
1336 | up->corkflag = 0; | |
1337 | lock_sock(sk); | |
4c0a6cb0 | 1338 | (*push_pending_frames)(sk); |
1da177e4 LT |
1339 | release_sock(sk); |
1340 | } | |
1341 | break; | |
e905a9ed | 1342 | |
1da177e4 LT |
1343 | case UDP_ENCAP: |
1344 | switch (val) { | |
1345 | case 0: | |
1346 | case UDP_ENCAP_ESPINUDP: | |
1347 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
1348 | up->encap_rcv = xfrm4_udp_encap_rcv; |
1349 | /* FALLTHROUGH */ | |
342f0234 | 1350 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 LT |
1351 | up->encap_type = val; |
1352 | break; | |
1353 | default: | |
1354 | err = -ENOPROTOOPT; | |
1355 | break; | |
1356 | } | |
1357 | break; | |
1358 | ||
ba4e58ec GR |
1359 | /* |
1360 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
1361 | */ | |
1362 | /* The sender sets actual checksum coverage length via this option. | |
1363 | * The case coverage > packet length is handled by send module. */ | |
1364 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 1365 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1366 | return -ENOPROTOOPT; |
1367 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
1368 | val = 8; | |
47112e25 GR |
1369 | else if (val > USHORT_MAX) |
1370 | val = USHORT_MAX; | |
ba4e58ec GR |
1371 | up->pcslen = val; |
1372 | up->pcflag |= UDPLITE_SEND_CC; | |
1373 | break; | |
1374 | ||
e905a9ed YH |
1375 | /* The receiver specifies a minimum checksum coverage value. To make |
1376 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
1377 | * used, this again means full checksum coverage. */ |
1378 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 1379 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1380 | return -ENOPROTOOPT; |
1381 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
1382 | val = 8; | |
47112e25 GR |
1383 | else if (val > USHORT_MAX) |
1384 | val = USHORT_MAX; | |
ba4e58ec GR |
1385 | up->pcrlen = val; |
1386 | up->pcflag |= UDPLITE_RECV_CC; | |
1387 | break; | |
1388 | ||
1da177e4 LT |
1389 | default: |
1390 | err = -ENOPROTOOPT; | |
1391 | break; | |
6516c655 | 1392 | } |
1da177e4 LT |
1393 | |
1394 | return err; | |
1395 | } | |
1396 | ||
db8dac20 DM |
1397 | int udp_setsockopt(struct sock *sk, int level, int optname, |
1398 | char __user *optval, int optlen) | |
1399 | { | |
1400 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1401 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1402 | udp_push_pending_frames); | |
1403 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
1404 | } | |
1405 | ||
1406 | #ifdef CONFIG_COMPAT | |
1407 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
1408 | char __user *optval, int optlen) | |
1409 | { | |
1410 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1411 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1412 | udp_push_pending_frames); | |
1413 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
1414 | } | |
1415 | #endif | |
1416 | ||
4c0a6cb0 GR |
1417 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
1418 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
1419 | { |
1420 | struct udp_sock *up = udp_sk(sk); | |
1421 | int val, len; | |
1422 | ||
6516c655 | 1423 | if (get_user(len,optlen)) |
1da177e4 LT |
1424 | return -EFAULT; |
1425 | ||
1426 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 1427 | |
6516c655 | 1428 | if (len < 0) |
1da177e4 LT |
1429 | return -EINVAL; |
1430 | ||
6516c655 | 1431 | switch (optname) { |
1da177e4 LT |
1432 | case UDP_CORK: |
1433 | val = up->corkflag; | |
1434 | break; | |
1435 | ||
1436 | case UDP_ENCAP: | |
1437 | val = up->encap_type; | |
1438 | break; | |
1439 | ||
ba4e58ec GR |
1440 | /* The following two cannot be changed on UDP sockets, the return is |
1441 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
1442 | case UDPLITE_SEND_CSCOV: | |
1443 | val = up->pcslen; | |
1444 | break; | |
1445 | ||
1446 | case UDPLITE_RECV_CSCOV: | |
1447 | val = up->pcrlen; | |
1448 | break; | |
1449 | ||
1da177e4 LT |
1450 | default: |
1451 | return -ENOPROTOOPT; | |
6516c655 | 1452 | } |
1da177e4 | 1453 | |
6516c655 | 1454 | if (put_user(len, optlen)) |
e905a9ed | 1455 | return -EFAULT; |
6516c655 | 1456 | if (copy_to_user(optval, &val,len)) |
1da177e4 | 1457 | return -EFAULT; |
e905a9ed | 1458 | return 0; |
1da177e4 LT |
1459 | } |
1460 | ||
db8dac20 DM |
1461 | int 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 ip_getsockopt(sk, level, optname, optval, optlen); | |
1467 | } | |
1468 | ||
1469 | #ifdef CONFIG_COMPAT | |
1470 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
1471 | char __user *optval, int __user *optlen) | |
1472 | { | |
1473 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1474 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1475 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
1476 | } | |
1477 | #endif | |
1da177e4 LT |
1478 | /** |
1479 | * udp_poll - wait for a UDP event. | |
1480 | * @file - file struct | |
1481 | * @sock - socket | |
1482 | * @wait - poll table | |
1483 | * | |
e905a9ed | 1484 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
1485 | * blocking sockets. If application is using a blocking fd |
1486 | * and a packet with checksum error is in the queue; | |
1487 | * then it could get return from select indicating data available | |
1488 | * but then block when reading it. Add special case code | |
1489 | * to work around these arguably broken applications. | |
1490 | */ | |
1491 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
1492 | { | |
1493 | unsigned int mask = datagram_poll(file, sock, wait); | |
1494 | struct sock *sk = sock->sk; | |
ba4e58ec GR |
1495 | int is_lite = IS_UDPLITE(sk); |
1496 | ||
1da177e4 LT |
1497 | /* Check for false positives due to checksum errors */ |
1498 | if ( (mask & POLLRDNORM) && | |
1499 | !(file->f_flags & O_NONBLOCK) && | |
1500 | !(sk->sk_shutdown & RCV_SHUTDOWN)){ | |
1501 | struct sk_buff_head *rcvq = &sk->sk_receive_queue; | |
1502 | struct sk_buff *skb; | |
1503 | ||
208d8984 | 1504 | spin_lock_bh(&rcvq->lock); |
759e5d00 HX |
1505 | while ((skb = skb_peek(rcvq)) != NULL && |
1506 | udp_lib_checksum_complete(skb)) { | |
0283328e PE |
1507 | UDP_INC_STATS_BH(sock_net(sk), |
1508 | UDP_MIB_INERRORS, is_lite); | |
759e5d00 HX |
1509 | __skb_unlink(skb, rcvq); |
1510 | kfree_skb(skb); | |
1da177e4 | 1511 | } |
208d8984 | 1512 | spin_unlock_bh(&rcvq->lock); |
1da177e4 LT |
1513 | |
1514 | /* nothing to see, move along */ | |
1515 | if (skb == NULL) | |
1516 | mask &= ~(POLLIN | POLLRDNORM); | |
1517 | } | |
1518 | ||
1519 | return mask; | |
e905a9ed | 1520 | |
1da177e4 LT |
1521 | } |
1522 | ||
db8dac20 DM |
1523 | struct proto udp_prot = { |
1524 | .name = "UDP", | |
1525 | .owner = THIS_MODULE, | |
1526 | .close = udp_lib_close, | |
1527 | .connect = ip4_datagram_connect, | |
1528 | .disconnect = udp_disconnect, | |
1529 | .ioctl = udp_ioctl, | |
1530 | .destroy = udp_destroy_sock, | |
1531 | .setsockopt = udp_setsockopt, | |
1532 | .getsockopt = udp_getsockopt, | |
1533 | .sendmsg = udp_sendmsg, | |
1534 | .recvmsg = udp_recvmsg, | |
1535 | .sendpage = udp_sendpage, | |
93821778 | 1536 | .backlog_rcv = __udp_queue_rcv_skb, |
db8dac20 DM |
1537 | .hash = udp_lib_hash, |
1538 | .unhash = udp_lib_unhash, | |
1539 | .get_port = udp_v4_get_port, | |
1540 | .memory_allocated = &udp_memory_allocated, | |
1541 | .sysctl_mem = sysctl_udp_mem, | |
1542 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
1543 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
1544 | .obj_size = sizeof(struct udp_sock), | |
271b72c7 | 1545 | .slab_flags = SLAB_DESTROY_BY_RCU, |
645ca708 | 1546 | .h.udp_table = &udp_table, |
db8dac20 DM |
1547 | #ifdef CONFIG_COMPAT |
1548 | .compat_setsockopt = compat_udp_setsockopt, | |
1549 | .compat_getsockopt = compat_udp_getsockopt, | |
1550 | #endif | |
db8dac20 | 1551 | }; |
1da177e4 LT |
1552 | |
1553 | /* ------------------------------------------------------------------------ */ | |
1554 | #ifdef CONFIG_PROC_FS | |
1555 | ||
645ca708 | 1556 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
1557 | { |
1558 | struct sock *sk; | |
1559 | struct udp_iter_state *state = seq->private; | |
6f191efe | 1560 | struct net *net = seq_file_net(seq); |
1da177e4 | 1561 | |
645ca708 | 1562 | for (state->bucket = start; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) { |
88ab1932 | 1563 | struct hlist_nulls_node *node; |
645ca708 ED |
1564 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
1565 | spin_lock_bh(&hslot->lock); | |
88ab1932 | 1566 | sk_nulls_for_each(sk, node, &hslot->head) { |
878628fb | 1567 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 1568 | continue; |
1da177e4 LT |
1569 | if (sk->sk_family == state->family) |
1570 | goto found; | |
1571 | } | |
645ca708 | 1572 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
1573 | } |
1574 | sk = NULL; | |
1575 | found: | |
1576 | return sk; | |
1577 | } | |
1578 | ||
1579 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
1580 | { | |
1581 | struct udp_iter_state *state = seq->private; | |
6f191efe | 1582 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1583 | |
1584 | do { | |
88ab1932 | 1585 | sk = sk_nulls_next(sk); |
878628fb | 1586 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 1587 | |
645ca708 | 1588 | if (!sk) { |
f52b5054 | 1589 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); |
645ca708 | 1590 | return udp_get_first(seq, state->bucket + 1); |
1da177e4 LT |
1591 | } |
1592 | return sk; | |
1593 | } | |
1594 | ||
1595 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
1596 | { | |
645ca708 | 1597 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
1598 | |
1599 | if (sk) | |
6516c655 | 1600 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
1601 | --pos; |
1602 | return pos ? NULL : sk; | |
1603 | } | |
1604 | ||
1605 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
1606 | { | |
b50660f1 | 1607 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
1608 | } |
1609 | ||
1610 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
1611 | { | |
1612 | struct sock *sk; | |
1613 | ||
b50660f1 | 1614 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
1615 | sk = udp_get_idx(seq, 0); |
1616 | else | |
1617 | sk = udp_get_next(seq, v); | |
1618 | ||
1619 | ++*pos; | |
1620 | return sk; | |
1621 | } | |
1622 | ||
1623 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
1624 | { | |
645ca708 ED |
1625 | struct udp_iter_state *state = seq->private; |
1626 | ||
1627 | if (state->bucket < UDP_HTABLE_SIZE) | |
1628 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); | |
1da177e4 LT |
1629 | } |
1630 | ||
1631 | static int udp_seq_open(struct inode *inode, struct file *file) | |
1632 | { | |
1633 | struct udp_seq_afinfo *afinfo = PDE(inode)->data; | |
a2be75c1 DL |
1634 | struct udp_iter_state *s; |
1635 | int err; | |
a91275ef | 1636 | |
a2be75c1 DL |
1637 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
1638 | sizeof(struct udp_iter_state)); | |
1639 | if (err < 0) | |
1640 | return err; | |
a91275ef | 1641 | |
a2be75c1 | 1642 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 1643 | s->family = afinfo->family; |
645ca708 | 1644 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 1645 | return err; |
a91275ef DL |
1646 | } |
1647 | ||
1da177e4 | 1648 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 1649 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
1650 | { |
1651 | struct proc_dir_entry *p; | |
1652 | int rc = 0; | |
1653 | ||
3ba9441b DL |
1654 | afinfo->seq_fops.open = udp_seq_open; |
1655 | afinfo->seq_fops.read = seq_read; | |
1656 | afinfo->seq_fops.llseek = seq_lseek; | |
1657 | afinfo->seq_fops.release = seq_release_net; | |
1da177e4 | 1658 | |
dda61925 DL |
1659 | afinfo->seq_ops.start = udp_seq_start; |
1660 | afinfo->seq_ops.next = udp_seq_next; | |
1661 | afinfo->seq_ops.stop = udp_seq_stop; | |
1662 | ||
84841c3c DL |
1663 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
1664 | &afinfo->seq_fops, afinfo); | |
1665 | if (!p) | |
1da177e4 LT |
1666 | rc = -ENOMEM; |
1667 | return rc; | |
1668 | } | |
1669 | ||
0c96d8c5 | 1670 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 1671 | { |
0c96d8c5 | 1672 | proc_net_remove(net, afinfo->name); |
1da177e4 | 1673 | } |
db8dac20 DM |
1674 | |
1675 | /* ------------------------------------------------------------------------ */ | |
5e659e4c PE |
1676 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
1677 | int bucket, int *len) | |
db8dac20 DM |
1678 | { |
1679 | struct inet_sock *inet = inet_sk(sp); | |
1680 | __be32 dest = inet->daddr; | |
1681 | __be32 src = inet->rcv_saddr; | |
1682 | __u16 destp = ntohs(inet->dport); | |
1683 | __u16 srcp = ntohs(inet->sport); | |
1684 | ||
5e659e4c | 1685 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
cb61cb9b | 1686 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d%n", |
db8dac20 DM |
1687 | bucket, src, srcp, dest, destp, sp->sk_state, |
1688 | atomic_read(&sp->sk_wmem_alloc), | |
1689 | atomic_read(&sp->sk_rmem_alloc), | |
1690 | 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp), | |
cb61cb9b ED |
1691 | atomic_read(&sp->sk_refcnt), sp, |
1692 | atomic_read(&sp->sk_drops), len); | |
db8dac20 DM |
1693 | } |
1694 | ||
1695 | int udp4_seq_show(struct seq_file *seq, void *v) | |
1696 | { | |
1697 | if (v == SEQ_START_TOKEN) | |
1698 | seq_printf(seq, "%-127s\n", | |
1699 | " sl local_address rem_address st tx_queue " | |
1700 | "rx_queue tr tm->when retrnsmt uid timeout " | |
cb61cb9b | 1701 | "inode ref pointer drops"); |
db8dac20 | 1702 | else { |
db8dac20 | 1703 | struct udp_iter_state *state = seq->private; |
5e659e4c | 1704 | int len; |
db8dac20 | 1705 | |
5e659e4c PE |
1706 | udp4_format_sock(v, seq, state->bucket, &len); |
1707 | seq_printf(seq, "%*s\n", 127 - len ,""); | |
db8dac20 DM |
1708 | } |
1709 | return 0; | |
1710 | } | |
1711 | ||
1712 | /* ------------------------------------------------------------------------ */ | |
db8dac20 | 1713 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
1714 | .name = "udp", |
1715 | .family = AF_INET, | |
645ca708 | 1716 | .udp_table = &udp_table, |
4ad96d39 DL |
1717 | .seq_fops = { |
1718 | .owner = THIS_MODULE, | |
1719 | }, | |
dda61925 DL |
1720 | .seq_ops = { |
1721 | .show = udp4_seq_show, | |
1722 | }, | |
db8dac20 DM |
1723 | }; |
1724 | ||
15439feb PE |
1725 | static int udp4_proc_init_net(struct net *net) |
1726 | { | |
1727 | return udp_proc_register(net, &udp4_seq_afinfo); | |
1728 | } | |
1729 | ||
1730 | static void udp4_proc_exit_net(struct net *net) | |
1731 | { | |
1732 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
1733 | } | |
1734 | ||
1735 | static struct pernet_operations udp4_net_ops = { | |
1736 | .init = udp4_proc_init_net, | |
1737 | .exit = udp4_proc_exit_net, | |
1738 | }; | |
1739 | ||
db8dac20 DM |
1740 | int __init udp4_proc_init(void) |
1741 | { | |
15439feb | 1742 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
1743 | } |
1744 | ||
1745 | void udp4_proc_exit(void) | |
1746 | { | |
15439feb | 1747 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 1748 | } |
1da177e4 LT |
1749 | #endif /* CONFIG_PROC_FS */ |
1750 | ||
645ca708 ED |
1751 | void __init udp_table_init(struct udp_table *table) |
1752 | { | |
1753 | int i; | |
1754 | ||
1755 | for (i = 0; i < UDP_HTABLE_SIZE; i++) { | |
88ab1932 | 1756 | INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i); |
645ca708 ED |
1757 | spin_lock_init(&table->hash[i].lock); |
1758 | } | |
1759 | } | |
1760 | ||
95766fff HA |
1761 | void __init udp_init(void) |
1762 | { | |
8203efb3 | 1763 | unsigned long nr_pages, limit; |
95766fff | 1764 | |
645ca708 | 1765 | udp_table_init(&udp_table); |
95766fff HA |
1766 | /* Set the pressure threshold up by the same strategy of TCP. It is a |
1767 | * fraction of global memory that is up to 1/2 at 256 MB, decreasing | |
1768 | * toward zero with the amount of memory, with a floor of 128 pages. | |
1769 | */ | |
8203efb3 ED |
1770 | nr_pages = totalram_pages - totalhigh_pages; |
1771 | limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT); | |
1772 | limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11); | |
95766fff HA |
1773 | limit = max(limit, 128UL); |
1774 | sysctl_udp_mem[0] = limit / 4 * 3; | |
1775 | sysctl_udp_mem[1] = limit; | |
1776 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
1777 | ||
1778 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
1779 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
1780 | } | |
1781 | ||
1da177e4 | 1782 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1783 | EXPORT_SYMBOL(udp_ioctl); |
db8dac20 DM |
1784 | EXPORT_SYMBOL(udp_prot); |
1785 | EXPORT_SYMBOL(udp_sendmsg); | |
4c0a6cb0 GR |
1786 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1787 | EXPORT_SYMBOL(udp_lib_setsockopt); | |
1da177e4 | 1788 | EXPORT_SYMBOL(udp_poll); |
6ba5a3c5 | 1789 | EXPORT_SYMBOL(udp_lib_get_port); |
1da177e4 LT |
1790 | |
1791 | #ifdef CONFIG_PROC_FS | |
1792 | EXPORT_SYMBOL(udp_proc_register); | |
1793 | EXPORT_SYMBOL(udp_proc_unregister); | |
1794 | #endif |