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 | ||
98322f22 ED |
123 | #define PORTS_PER_CHAIN (65536 / UDP_HTABLE_SIZE) |
124 | ||
f24d43c0 | 125 | static int udp_lib_lport_inuse(struct net *net, __u16 num, |
645ca708 | 126 | const struct udp_hslot *hslot, |
98322f22 | 127 | unsigned long *bitmap, |
f24d43c0 ED |
128 | struct sock *sk, |
129 | int (*saddr_comp)(const struct sock *sk1, | |
130 | const struct sock *sk2)) | |
1da177e4 | 131 | { |
f24d43c0 | 132 | struct sock *sk2; |
88ab1932 | 133 | struct hlist_nulls_node *node; |
25030a7f | 134 | |
88ab1932 | 135 | sk_nulls_for_each(sk2, node, &hslot->head) |
f24d43c0 ED |
136 | if (net_eq(sock_net(sk2), net) && |
137 | sk2 != sk && | |
98322f22 | 138 | (bitmap || sk2->sk_hash == num) && |
f24d43c0 ED |
139 | (!sk2->sk_reuse || !sk->sk_reuse) && |
140 | (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if | |
141 | || sk2->sk_bound_dev_if == sk->sk_bound_dev_if) && | |
98322f22 ED |
142 | (*saddr_comp)(sk, sk2)) { |
143 | if (bitmap) | |
144 | __set_bit(sk2->sk_hash / UDP_HTABLE_SIZE, | |
145 | bitmap); | |
146 | else | |
147 | return 1; | |
148 | } | |
25030a7f GR |
149 | return 0; |
150 | } | |
151 | ||
152 | /** | |
6ba5a3c5 | 153 | * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6 |
25030a7f GR |
154 | * |
155 | * @sk: socket struct in question | |
156 | * @snum: port number to look up | |
df2bc459 | 157 | * @saddr_comp: AF-dependent comparison of bound local IP addresses |
25030a7f | 158 | */ |
6ba5a3c5 | 159 | int udp_lib_get_port(struct sock *sk, unsigned short snum, |
df2bc459 DM |
160 | int (*saddr_comp)(const struct sock *sk1, |
161 | const struct sock *sk2 ) ) | |
25030a7f | 162 | { |
645ca708 ED |
163 | struct udp_hslot *hslot; |
164 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
25030a7f | 165 | int error = 1; |
3b1e0a65 | 166 | struct net *net = sock_net(sk); |
1da177e4 | 167 | |
32c1da70 | 168 | if (!snum) { |
9088c560 ED |
169 | int low, high, remaining; |
170 | unsigned rand; | |
98322f22 ED |
171 | unsigned short first, last; |
172 | DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN); | |
32c1da70 | 173 | |
227b60f5 | 174 | inet_get_local_port_range(&low, &high); |
a25de534 | 175 | remaining = (high - low) + 1; |
227b60f5 | 176 | |
9088c560 | 177 | rand = net_random(); |
98322f22 ED |
178 | first = (((u64)rand * remaining) >> 32) + low; |
179 | /* | |
180 | * force rand to be an odd multiple of UDP_HTABLE_SIZE | |
181 | */ | |
182 | rand = (rand | 1) * UDP_HTABLE_SIZE; | |
183 | for (last = first + UDP_HTABLE_SIZE; first != last; first++) { | |
184 | hslot = &udptable->hash[udp_hashfn(net, first)]; | |
185 | bitmap_zero(bitmap, PORTS_PER_CHAIN); | |
645ca708 | 186 | spin_lock_bh(&hslot->lock); |
98322f22 ED |
187 | udp_lib_lport_inuse(net, snum, hslot, bitmap, sk, |
188 | saddr_comp); | |
189 | ||
190 | snum = first; | |
191 | /* | |
192 | * Iterate on all possible values of snum for this hash. | |
193 | * Using steps of an odd multiple of UDP_HTABLE_SIZE | |
194 | * give us randomization and full range coverage. | |
195 | */ | |
9088c560 | 196 | do { |
98322f22 ED |
197 | if (low <= snum && snum <= high && |
198 | !test_bit(snum / UDP_HTABLE_SIZE, bitmap)) | |
199 | goto found; | |
200 | snum += rand; | |
201 | } while (snum != first); | |
202 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 203 | } |
98322f22 | 204 | goto fail; |
645ca708 ED |
205 | } else { |
206 | hslot = &udptable->hash[udp_hashfn(net, snum)]; | |
207 | spin_lock_bh(&hslot->lock); | |
98322f22 | 208 | if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk, saddr_comp)) |
645ca708 ED |
209 | goto fail_unlock; |
210 | } | |
98322f22 | 211 | found: |
25030a7f | 212 | inet_sk(sk)->num = snum; |
df2bc459 | 213 | sk->sk_hash = snum; |
1da177e4 | 214 | if (sk_unhashed(sk)) { |
88ab1932 | 215 | sk_nulls_add_node_rcu(sk, &hslot->head); |
c29a0bc4 | 216 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1); |
1da177e4 | 217 | } |
25030a7f | 218 | error = 0; |
645ca708 ED |
219 | fail_unlock: |
220 | spin_unlock_bh(&hslot->lock); | |
1da177e4 | 221 | fail: |
25030a7f GR |
222 | return error; |
223 | } | |
224 | ||
499923c7 | 225 | static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2) |
db8dac20 DM |
226 | { |
227 | struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2); | |
228 | ||
229 | return ( !ipv6_only_sock(sk2) && | |
230 | (!inet1->rcv_saddr || !inet2->rcv_saddr || | |
231 | inet1->rcv_saddr == inet2->rcv_saddr )); | |
232 | } | |
233 | ||
6ba5a3c5 | 234 | int udp_v4_get_port(struct sock *sk, unsigned short snum) |
db8dac20 | 235 | { |
6ba5a3c5 | 236 | return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal); |
db8dac20 DM |
237 | } |
238 | ||
645ca708 ED |
239 | static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr, |
240 | unsigned short hnum, | |
241 | __be16 sport, __be32 daddr, __be16 dport, int dif) | |
242 | { | |
243 | int score = -1; | |
244 | ||
245 | if (net_eq(sock_net(sk), net) && sk->sk_hash == hnum && | |
246 | !ipv6_only_sock(sk)) { | |
247 | struct inet_sock *inet = inet_sk(sk); | |
248 | ||
249 | score = (sk->sk_family == PF_INET ? 1 : 0); | |
250 | if (inet->rcv_saddr) { | |
251 | if (inet->rcv_saddr != daddr) | |
252 | return -1; | |
253 | score += 2; | |
254 | } | |
255 | if (inet->daddr) { | |
256 | if (inet->daddr != saddr) | |
257 | return -1; | |
258 | score += 2; | |
259 | } | |
260 | if (inet->dport) { | |
261 | if (inet->dport != sport) | |
262 | return -1; | |
263 | score += 2; | |
264 | } | |
265 | if (sk->sk_bound_dev_if) { | |
266 | if (sk->sk_bound_dev_if != dif) | |
267 | return -1; | |
268 | score += 2; | |
269 | } | |
270 | } | |
271 | return score; | |
272 | } | |
273 | ||
db8dac20 DM |
274 | /* UDP is nearly always wildcards out the wazoo, it makes no sense to try |
275 | * harder than this. -DaveM | |
276 | */ | |
277 | static struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, | |
278 | __be16 sport, __be32 daddr, __be16 dport, | |
645ca708 | 279 | int dif, struct udp_table *udptable) |
db8dac20 | 280 | { |
271b72c7 | 281 | struct sock *sk, *result; |
88ab1932 | 282 | struct hlist_nulls_node *node; |
db8dac20 | 283 | unsigned short hnum = ntohs(dport); |
645ca708 ED |
284 | unsigned int hash = udp_hashfn(net, hnum); |
285 | struct udp_hslot *hslot = &udptable->hash[hash]; | |
271b72c7 | 286 | int score, badness; |
645ca708 | 287 | |
271b72c7 ED |
288 | rcu_read_lock(); |
289 | begin: | |
290 | result = NULL; | |
291 | badness = -1; | |
88ab1932 | 292 | sk_nulls_for_each_rcu(sk, node, &hslot->head) { |
645ca708 ED |
293 | score = compute_score(sk, net, saddr, hnum, sport, |
294 | daddr, dport, dif); | |
295 | if (score > badness) { | |
296 | result = sk; | |
297 | badness = score; | |
db8dac20 DM |
298 | } |
299 | } | |
88ab1932 ED |
300 | /* |
301 | * if the nulls value we got at the end of this lookup is | |
302 | * not the expected one, we must restart lookup. | |
303 | * We probably met an item that was moved to another chain. | |
304 | */ | |
305 | if (get_nulls_value(node) != hash) | |
306 | goto begin; | |
307 | ||
271b72c7 ED |
308 | if (result) { |
309 | if (unlikely(!atomic_inc_not_zero(&result->sk_refcnt))) | |
310 | result = NULL; | |
311 | else if (unlikely(compute_score(result, net, saddr, hnum, sport, | |
312 | daddr, dport, dif) < badness)) { | |
313 | sock_put(result); | |
314 | goto begin; | |
315 | } | |
316 | } | |
317 | rcu_read_unlock(); | |
db8dac20 DM |
318 | return result; |
319 | } | |
320 | ||
607c4aaf KK |
321 | static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb, |
322 | __be16 sport, __be16 dport, | |
645ca708 | 323 | struct udp_table *udptable) |
607c4aaf | 324 | { |
23542618 | 325 | struct sock *sk; |
607c4aaf KK |
326 | const struct iphdr *iph = ip_hdr(skb); |
327 | ||
23542618 KK |
328 | if (unlikely(sk = skb_steal_sock(skb))) |
329 | return sk; | |
330 | else | |
331 | return __udp4_lib_lookup(dev_net(skb->dst->dev), iph->saddr, sport, | |
332 | iph->daddr, dport, inet_iif(skb), | |
333 | udptable); | |
607c4aaf KK |
334 | } |
335 | ||
bcd41303 KK |
336 | struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport, |
337 | __be32 daddr, __be16 dport, int dif) | |
338 | { | |
645ca708 | 339 | return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table); |
bcd41303 KK |
340 | } |
341 | EXPORT_SYMBOL_GPL(udp4_lib_lookup); | |
342 | ||
920a4611 | 343 | static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk, |
db8dac20 DM |
344 | __be16 loc_port, __be32 loc_addr, |
345 | __be16 rmt_port, __be32 rmt_addr, | |
346 | int dif) | |
347 | { | |
88ab1932 | 348 | struct hlist_nulls_node *node; |
db8dac20 DM |
349 | struct sock *s = sk; |
350 | unsigned short hnum = ntohs(loc_port); | |
351 | ||
88ab1932 | 352 | sk_nulls_for_each_from(s, node) { |
db8dac20 DM |
353 | struct inet_sock *inet = inet_sk(s); |
354 | ||
920a4611 ED |
355 | if (!net_eq(sock_net(s), net) || |
356 | s->sk_hash != hnum || | |
db8dac20 DM |
357 | (inet->daddr && inet->daddr != rmt_addr) || |
358 | (inet->dport != rmt_port && inet->dport) || | |
359 | (inet->rcv_saddr && inet->rcv_saddr != loc_addr) || | |
360 | ipv6_only_sock(s) || | |
361 | (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)) | |
362 | continue; | |
363 | if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif)) | |
364 | continue; | |
365 | goto found; | |
366 | } | |
367 | s = NULL; | |
368 | found: | |
369 | return s; | |
370 | } | |
371 | ||
372 | /* | |
373 | * This routine is called by the ICMP module when it gets some | |
374 | * sort of error condition. If err < 0 then the socket should | |
375 | * be closed and the error returned to the user. If err > 0 | |
376 | * it's just the icmp type << 8 | icmp code. | |
377 | * Header points to the ip header of the error packet. We move | |
378 | * on past this. Then (as it used to claim before adjustment) | |
379 | * header points to the first 8 bytes of the udp header. We need | |
380 | * to find the appropriate port. | |
381 | */ | |
382 | ||
645ca708 | 383 | void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable) |
db8dac20 DM |
384 | { |
385 | struct inet_sock *inet; | |
386 | struct iphdr *iph = (struct iphdr*)skb->data; | |
387 | struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2)); | |
388 | const int type = icmp_hdr(skb)->type; | |
389 | const int code = icmp_hdr(skb)->code; | |
390 | struct sock *sk; | |
391 | int harderr; | |
392 | int err; | |
fd54d716 | 393 | struct net *net = dev_net(skb->dev); |
db8dac20 | 394 | |
fd54d716 | 395 | sk = __udp4_lib_lookup(net, iph->daddr, uh->dest, |
db8dac20 DM |
396 | iph->saddr, uh->source, skb->dev->ifindex, udptable); |
397 | if (sk == NULL) { | |
dcfc23ca | 398 | ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS); |
db8dac20 DM |
399 | return; /* No socket for error */ |
400 | } | |
401 | ||
402 | err = 0; | |
403 | harderr = 0; | |
404 | inet = inet_sk(sk); | |
405 | ||
406 | switch (type) { | |
407 | default: | |
408 | case ICMP_TIME_EXCEEDED: | |
409 | err = EHOSTUNREACH; | |
410 | break; | |
411 | case ICMP_SOURCE_QUENCH: | |
412 | goto out; | |
413 | case ICMP_PARAMETERPROB: | |
414 | err = EPROTO; | |
415 | harderr = 1; | |
416 | break; | |
417 | case ICMP_DEST_UNREACH: | |
418 | if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ | |
419 | if (inet->pmtudisc != IP_PMTUDISC_DONT) { | |
420 | err = EMSGSIZE; | |
421 | harderr = 1; | |
422 | break; | |
423 | } | |
424 | goto out; | |
425 | } | |
426 | err = EHOSTUNREACH; | |
427 | if (code <= NR_ICMP_UNREACH) { | |
428 | harderr = icmp_err_convert[code].fatal; | |
429 | err = icmp_err_convert[code].errno; | |
430 | } | |
431 | break; | |
432 | } | |
433 | ||
434 | /* | |
435 | * RFC1122: OK. Passes ICMP errors back to application, as per | |
436 | * 4.1.3.3. | |
437 | */ | |
438 | if (!inet->recverr) { | |
439 | if (!harderr || sk->sk_state != TCP_ESTABLISHED) | |
440 | goto out; | |
441 | } else { | |
442 | ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1)); | |
443 | } | |
444 | sk->sk_err = err; | |
445 | sk->sk_error_report(sk); | |
446 | out: | |
447 | sock_put(sk); | |
448 | } | |
449 | ||
450 | void udp_err(struct sk_buff *skb, u32 info) | |
451 | { | |
645ca708 | 452 | __udp4_lib_err(skb, info, &udp_table); |
db8dac20 DM |
453 | } |
454 | ||
455 | /* | |
456 | * Throw away all pending data and cancel the corking. Socket is locked. | |
457 | */ | |
36d926b9 | 458 | void udp_flush_pending_frames(struct sock *sk) |
db8dac20 DM |
459 | { |
460 | struct udp_sock *up = udp_sk(sk); | |
461 | ||
462 | if (up->pending) { | |
463 | up->len = 0; | |
464 | up->pending = 0; | |
465 | ip_flush_pending_frames(sk); | |
466 | } | |
467 | } | |
36d926b9 | 468 | EXPORT_SYMBOL(udp_flush_pending_frames); |
db8dac20 DM |
469 | |
470 | /** | |
471 | * udp4_hwcsum_outgoing - handle outgoing HW checksumming | |
472 | * @sk: socket we are sending on | |
473 | * @skb: sk_buff containing the filled-in UDP header | |
474 | * (checksum field must be zeroed out) | |
475 | */ | |
476 | static void udp4_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb, | |
477 | __be32 src, __be32 dst, int len ) | |
478 | { | |
479 | unsigned int offset; | |
480 | struct udphdr *uh = udp_hdr(skb); | |
481 | __wsum csum = 0; | |
482 | ||
483 | if (skb_queue_len(&sk->sk_write_queue) == 1) { | |
484 | /* | |
485 | * Only one fragment on the socket. | |
486 | */ | |
487 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
488 | skb->csum_offset = offsetof(struct udphdr, check); | |
489 | uh->check = ~csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, 0); | |
490 | } else { | |
491 | /* | |
492 | * HW-checksum won't work as there are two or more | |
493 | * fragments on the socket so that all csums of sk_buffs | |
494 | * should be together | |
495 | */ | |
496 | offset = skb_transport_offset(skb); | |
497 | skb->csum = skb_checksum(skb, offset, skb->len - offset, 0); | |
498 | ||
499 | skb->ip_summed = CHECKSUM_NONE; | |
500 | ||
501 | skb_queue_walk(&sk->sk_write_queue, skb) { | |
502 | csum = csum_add(csum, skb->csum); | |
503 | } | |
504 | ||
505 | uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum); | |
506 | if (uh->check == 0) | |
507 | uh->check = CSUM_MANGLED_0; | |
508 | } | |
509 | } | |
510 | ||
511 | /* | |
512 | * Push out all pending data as one UDP datagram. Socket is locked. | |
513 | */ | |
514 | static int udp_push_pending_frames(struct sock *sk) | |
515 | { | |
516 | struct udp_sock *up = udp_sk(sk); | |
517 | struct inet_sock *inet = inet_sk(sk); | |
518 | struct flowi *fl = &inet->cork.fl; | |
519 | struct sk_buff *skb; | |
520 | struct udphdr *uh; | |
521 | int err = 0; | |
522 | int is_udplite = IS_UDPLITE(sk); | |
523 | __wsum csum = 0; | |
524 | ||
525 | /* Grab the skbuff where UDP header space exists. */ | |
526 | if ((skb = skb_peek(&sk->sk_write_queue)) == NULL) | |
527 | goto out; | |
528 | ||
529 | /* | |
530 | * Create a UDP header | |
531 | */ | |
532 | uh = udp_hdr(skb); | |
533 | uh->source = fl->fl_ip_sport; | |
534 | uh->dest = fl->fl_ip_dport; | |
535 | uh->len = htons(up->len); | |
536 | uh->check = 0; | |
537 | ||
538 | if (is_udplite) /* UDP-Lite */ | |
539 | csum = udplite_csum_outgoing(sk, skb); | |
540 | ||
541 | else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */ | |
542 | ||
543 | skb->ip_summed = CHECKSUM_NONE; | |
544 | goto send; | |
545 | ||
546 | } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */ | |
547 | ||
548 | udp4_hwcsum_outgoing(sk, skb, fl->fl4_src,fl->fl4_dst, up->len); | |
549 | goto send; | |
550 | ||
551 | } else /* `normal' UDP */ | |
552 | csum = udp_csum_outgoing(sk, skb); | |
553 | ||
554 | /* add protocol-dependent pseudo-header */ | |
555 | uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, up->len, | |
556 | sk->sk_protocol, csum ); | |
557 | if (uh->check == 0) | |
558 | uh->check = CSUM_MANGLED_0; | |
559 | ||
560 | send: | |
561 | err = ip_push_pending_frames(sk); | |
562 | out: | |
563 | up->len = 0; | |
564 | up->pending = 0; | |
565 | if (!err) | |
629ca23c PE |
566 | UDP_INC_STATS_USER(sock_net(sk), |
567 | UDP_MIB_OUTDATAGRAMS, is_udplite); | |
db8dac20 DM |
568 | return err; |
569 | } | |
570 | ||
571 | int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
572 | size_t len) | |
573 | { | |
574 | struct inet_sock *inet = inet_sk(sk); | |
575 | struct udp_sock *up = udp_sk(sk); | |
576 | int ulen = len; | |
577 | struct ipcm_cookie ipc; | |
578 | struct rtable *rt = NULL; | |
579 | int free = 0; | |
580 | int connected = 0; | |
581 | __be32 daddr, faddr, saddr; | |
582 | __be16 dport; | |
583 | u8 tos; | |
584 | int err, is_udplite = IS_UDPLITE(sk); | |
585 | int corkreq = up->corkflag || msg->msg_flags&MSG_MORE; | |
586 | int (*getfrag)(void *, char *, int, int, int, struct sk_buff *); | |
587 | ||
588 | if (len > 0xFFFF) | |
589 | return -EMSGSIZE; | |
590 | ||
591 | /* | |
592 | * Check the flags. | |
593 | */ | |
594 | ||
595 | if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */ | |
596 | return -EOPNOTSUPP; | |
597 | ||
598 | ipc.opt = NULL; | |
51f31cab | 599 | ipc.shtx.flags = 0; |
db8dac20 DM |
600 | |
601 | if (up->pending) { | |
602 | /* | |
603 | * There are pending frames. | |
604 | * The socket lock must be held while it's corked. | |
605 | */ | |
606 | lock_sock(sk); | |
607 | if (likely(up->pending)) { | |
608 | if (unlikely(up->pending != AF_INET)) { | |
609 | release_sock(sk); | |
610 | return -EINVAL; | |
611 | } | |
612 | goto do_append_data; | |
613 | } | |
614 | release_sock(sk); | |
615 | } | |
616 | ulen += sizeof(struct udphdr); | |
617 | ||
618 | /* | |
619 | * Get and verify the address. | |
620 | */ | |
621 | if (msg->msg_name) { | |
622 | struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name; | |
623 | if (msg->msg_namelen < sizeof(*usin)) | |
624 | return -EINVAL; | |
625 | if (usin->sin_family != AF_INET) { | |
626 | if (usin->sin_family != AF_UNSPEC) | |
627 | return -EAFNOSUPPORT; | |
628 | } | |
629 | ||
630 | daddr = usin->sin_addr.s_addr; | |
631 | dport = usin->sin_port; | |
632 | if (dport == 0) | |
633 | return -EINVAL; | |
634 | } else { | |
635 | if (sk->sk_state != TCP_ESTABLISHED) | |
636 | return -EDESTADDRREQ; | |
637 | daddr = inet->daddr; | |
638 | dport = inet->dport; | |
639 | /* Open fast path for connected socket. | |
640 | Route will not be used, if at least one option is set. | |
641 | */ | |
642 | connected = 1; | |
643 | } | |
644 | ipc.addr = inet->saddr; | |
645 | ||
646 | ipc.oif = sk->sk_bound_dev_if; | |
51f31cab PO |
647 | err = sock_tx_timestamp(msg, sk, &ipc.shtx); |
648 | if (err) | |
649 | return err; | |
db8dac20 | 650 | if (msg->msg_controllen) { |
3b1e0a65 | 651 | err = ip_cmsg_send(sock_net(sk), msg, &ipc); |
db8dac20 DM |
652 | if (err) |
653 | return err; | |
654 | if (ipc.opt) | |
655 | free = 1; | |
656 | connected = 0; | |
657 | } | |
658 | if (!ipc.opt) | |
659 | ipc.opt = inet->opt; | |
660 | ||
661 | saddr = ipc.addr; | |
662 | ipc.addr = faddr = daddr; | |
663 | ||
664 | if (ipc.opt && ipc.opt->srr) { | |
665 | if (!daddr) | |
666 | return -EINVAL; | |
667 | faddr = ipc.opt->faddr; | |
668 | connected = 0; | |
669 | } | |
670 | tos = RT_TOS(inet->tos); | |
671 | if (sock_flag(sk, SOCK_LOCALROUTE) || | |
672 | (msg->msg_flags & MSG_DONTROUTE) || | |
673 | (ipc.opt && ipc.opt->is_strictroute)) { | |
674 | tos |= RTO_ONLINK; | |
675 | connected = 0; | |
676 | } | |
677 | ||
678 | if (ipv4_is_multicast(daddr)) { | |
679 | if (!ipc.oif) | |
680 | ipc.oif = inet->mc_index; | |
681 | if (!saddr) | |
682 | saddr = inet->mc_addr; | |
683 | connected = 0; | |
684 | } | |
685 | ||
686 | if (connected) | |
687 | rt = (struct rtable*)sk_dst_check(sk, 0); | |
688 | ||
689 | if (rt == NULL) { | |
690 | struct flowi fl = { .oif = ipc.oif, | |
691 | .nl_u = { .ip4_u = | |
692 | { .daddr = faddr, | |
693 | .saddr = saddr, | |
694 | .tos = tos } }, | |
695 | .proto = sk->sk_protocol, | |
a134f85c | 696 | .flags = inet_sk_flowi_flags(sk), |
db8dac20 DM |
697 | .uli_u = { .ports = |
698 | { .sport = inet->sport, | |
699 | .dport = dport } } }; | |
84a3aa00 PE |
700 | struct net *net = sock_net(sk); |
701 | ||
db8dac20 | 702 | security_sk_classify_flow(sk, &fl); |
84a3aa00 | 703 | err = ip_route_output_flow(net, &rt, &fl, sk, 1); |
db8dac20 DM |
704 | if (err) { |
705 | if (err == -ENETUNREACH) | |
7c73a6fa | 706 | IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES); |
db8dac20 DM |
707 | goto out; |
708 | } | |
709 | ||
710 | err = -EACCES; | |
711 | if ((rt->rt_flags & RTCF_BROADCAST) && | |
712 | !sock_flag(sk, SOCK_BROADCAST)) | |
713 | goto out; | |
714 | if (connected) | |
715 | sk_dst_set(sk, dst_clone(&rt->u.dst)); | |
716 | } | |
717 | ||
718 | if (msg->msg_flags&MSG_CONFIRM) | |
719 | goto do_confirm; | |
720 | back_from_confirm: | |
721 | ||
722 | saddr = rt->rt_src; | |
723 | if (!ipc.addr) | |
724 | daddr = ipc.addr = rt->rt_dst; | |
725 | ||
726 | lock_sock(sk); | |
727 | if (unlikely(up->pending)) { | |
728 | /* The socket is already corked while preparing it. */ | |
729 | /* ... which is an evident application bug. --ANK */ | |
730 | release_sock(sk); | |
731 | ||
732 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n"); | |
733 | err = -EINVAL; | |
734 | goto out; | |
735 | } | |
736 | /* | |
737 | * Now cork the socket to pend data. | |
738 | */ | |
739 | inet->cork.fl.fl4_dst = daddr; | |
740 | inet->cork.fl.fl_ip_dport = dport; | |
741 | inet->cork.fl.fl4_src = saddr; | |
742 | inet->cork.fl.fl_ip_sport = inet->sport; | |
743 | up->pending = AF_INET; | |
744 | ||
745 | do_append_data: | |
746 | up->len += ulen; | |
747 | getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag; | |
748 | err = ip_append_data(sk, getfrag, msg->msg_iov, ulen, | |
2e77d89b | 749 | sizeof(struct udphdr), &ipc, &rt, |
db8dac20 DM |
750 | corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags); |
751 | if (err) | |
752 | udp_flush_pending_frames(sk); | |
753 | else if (!corkreq) | |
754 | err = udp_push_pending_frames(sk); | |
755 | else if (unlikely(skb_queue_empty(&sk->sk_write_queue))) | |
756 | up->pending = 0; | |
757 | release_sock(sk); | |
758 | ||
759 | out: | |
760 | ip_rt_put(rt); | |
761 | if (free) | |
762 | kfree(ipc.opt); | |
763 | if (!err) | |
764 | return len; | |
765 | /* | |
766 | * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting | |
767 | * ENOBUFS might not be good (it's not tunable per se), but otherwise | |
768 | * we don't have a good statistic (IpOutDiscards but it can be too many | |
769 | * things). We could add another new stat but at least for now that | |
770 | * seems like overkill. | |
771 | */ | |
772 | if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) { | |
629ca23c PE |
773 | UDP_INC_STATS_USER(sock_net(sk), |
774 | UDP_MIB_SNDBUFERRORS, is_udplite); | |
db8dac20 DM |
775 | } |
776 | return err; | |
777 | ||
778 | do_confirm: | |
779 | dst_confirm(&rt->u.dst); | |
780 | if (!(msg->msg_flags&MSG_PROBE) || len) | |
781 | goto back_from_confirm; | |
782 | err = 0; | |
783 | goto out; | |
784 | } | |
785 | ||
786 | int udp_sendpage(struct sock *sk, struct page *page, int offset, | |
787 | size_t size, int flags) | |
788 | { | |
789 | struct udp_sock *up = udp_sk(sk); | |
790 | int ret; | |
791 | ||
792 | if (!up->pending) { | |
793 | struct msghdr msg = { .msg_flags = flags|MSG_MORE }; | |
794 | ||
795 | /* Call udp_sendmsg to specify destination address which | |
796 | * sendpage interface can't pass. | |
797 | * This will succeed only when the socket is connected. | |
798 | */ | |
799 | ret = udp_sendmsg(NULL, sk, &msg, 0); | |
800 | if (ret < 0) | |
801 | return ret; | |
802 | } | |
803 | ||
804 | lock_sock(sk); | |
805 | ||
806 | if (unlikely(!up->pending)) { | |
807 | release_sock(sk); | |
808 | ||
809 | LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n"); | |
810 | return -EINVAL; | |
811 | } | |
812 | ||
813 | ret = ip_append_page(sk, page, offset, size, flags); | |
814 | if (ret == -EOPNOTSUPP) { | |
815 | release_sock(sk); | |
816 | return sock_no_sendpage(sk->sk_socket, page, offset, | |
817 | size, flags); | |
818 | } | |
819 | if (ret < 0) { | |
820 | udp_flush_pending_frames(sk); | |
821 | goto out; | |
822 | } | |
823 | ||
824 | up->len += size; | |
825 | if (!(up->corkflag || (flags&MSG_MORE))) | |
826 | ret = udp_push_pending_frames(sk); | |
827 | if (!ret) | |
828 | ret = size; | |
829 | out: | |
830 | release_sock(sk); | |
831 | return ret; | |
832 | } | |
833 | ||
1da177e4 LT |
834 | /* |
835 | * IOCTL requests applicable to the UDP protocol | |
836 | */ | |
e905a9ed | 837 | |
1da177e4 LT |
838 | int udp_ioctl(struct sock *sk, int cmd, unsigned long arg) |
839 | { | |
6516c655 SH |
840 | switch (cmd) { |
841 | case SIOCOUTQ: | |
1da177e4 | 842 | { |
6516c655 SH |
843 | int amount = atomic_read(&sk->sk_wmem_alloc); |
844 | return put_user(amount, (int __user *)arg); | |
845 | } | |
1da177e4 | 846 | |
6516c655 SH |
847 | case SIOCINQ: |
848 | { | |
849 | struct sk_buff *skb; | |
850 | unsigned long amount; | |
851 | ||
852 | amount = 0; | |
853 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
854 | skb = skb_peek(&sk->sk_receive_queue); | |
855 | if (skb != NULL) { | |
856 | /* | |
857 | * We will only return the amount | |
858 | * of this packet since that is all | |
859 | * that will be read. | |
860 | */ | |
861 | amount = skb->len - sizeof(struct udphdr); | |
1da177e4 | 862 | } |
6516c655 SH |
863 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
864 | return put_user(amount, (int __user *)arg); | |
865 | } | |
1da177e4 | 866 | |
6516c655 SH |
867 | default: |
868 | return -ENOIOCTLCMD; | |
1da177e4 | 869 | } |
6516c655 SH |
870 | |
871 | return 0; | |
1da177e4 LT |
872 | } |
873 | ||
db8dac20 DM |
874 | /* |
875 | * This should be easy, if there is something there we | |
876 | * return it, otherwise we block. | |
877 | */ | |
878 | ||
879 | int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
880 | size_t len, int noblock, int flags, int *addr_len) | |
881 | { | |
882 | struct inet_sock *inet = inet_sk(sk); | |
883 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; | |
884 | struct sk_buff *skb; | |
885 | unsigned int ulen, copied; | |
886 | int peeked; | |
887 | int err; | |
888 | int is_udplite = IS_UDPLITE(sk); | |
889 | ||
890 | /* | |
891 | * Check any passed addresses | |
892 | */ | |
893 | if (addr_len) | |
894 | *addr_len=sizeof(*sin); | |
895 | ||
896 | if (flags & MSG_ERRQUEUE) | |
897 | return ip_recv_error(sk, msg, len); | |
898 | ||
899 | try_again: | |
900 | skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0), | |
901 | &peeked, &err); | |
902 | if (!skb) | |
903 | goto out; | |
904 | ||
905 | ulen = skb->len - sizeof(struct udphdr); | |
906 | copied = len; | |
907 | if (copied > ulen) | |
908 | copied = ulen; | |
909 | else if (copied < ulen) | |
910 | msg->msg_flags |= MSG_TRUNC; | |
911 | ||
912 | /* | |
913 | * If checksum is needed at all, try to do it while copying the | |
914 | * data. If the data is truncated, or if we only want a partial | |
915 | * coverage checksum (UDP-Lite), do it before the copy. | |
916 | */ | |
917 | ||
918 | if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) { | |
919 | if (udp_lib_checksum_complete(skb)) | |
920 | goto csum_copy_err; | |
921 | } | |
922 | ||
923 | if (skb_csum_unnecessary(skb)) | |
924 | err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), | |
925 | msg->msg_iov, copied ); | |
926 | else { | |
927 | err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov); | |
928 | ||
929 | if (err == -EINVAL) | |
930 | goto csum_copy_err; | |
931 | } | |
932 | ||
933 | if (err) | |
934 | goto out_free; | |
935 | ||
936 | if (!peeked) | |
629ca23c PE |
937 | UDP_INC_STATS_USER(sock_net(sk), |
938 | UDP_MIB_INDATAGRAMS, is_udplite); | |
db8dac20 DM |
939 | |
940 | sock_recv_timestamp(msg, sk, skb); | |
941 | ||
942 | /* Copy the address. */ | |
943 | if (sin) | |
944 | { | |
945 | sin->sin_family = AF_INET; | |
946 | sin->sin_port = udp_hdr(skb)->source; | |
947 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; | |
948 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
949 | } | |
950 | if (inet->cmsg_flags) | |
951 | ip_cmsg_recv(msg, skb); | |
952 | ||
953 | err = copied; | |
954 | if (flags & MSG_TRUNC) | |
955 | err = ulen; | |
956 | ||
957 | out_free: | |
958 | lock_sock(sk); | |
959 | skb_free_datagram(sk, skb); | |
960 | release_sock(sk); | |
961 | out: | |
962 | return err; | |
963 | ||
964 | csum_copy_err: | |
965 | lock_sock(sk); | |
966 | if (!skb_kill_datagram(sk, skb, flags)) | |
629ca23c | 967 | UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
db8dac20 DM |
968 | release_sock(sk); |
969 | ||
970 | if (noblock) | |
971 | return -EAGAIN; | |
972 | goto try_again; | |
973 | } | |
974 | ||
975 | ||
1da177e4 LT |
976 | int udp_disconnect(struct sock *sk, int flags) |
977 | { | |
978 | struct inet_sock *inet = inet_sk(sk); | |
979 | /* | |
980 | * 1003.1g - break association. | |
981 | */ | |
e905a9ed | 982 | |
1da177e4 LT |
983 | sk->sk_state = TCP_CLOSE; |
984 | inet->daddr = 0; | |
985 | inet->dport = 0; | |
986 | sk->sk_bound_dev_if = 0; | |
987 | if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) | |
988 | inet_reset_saddr(sk); | |
989 | ||
990 | if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) { | |
991 | sk->sk_prot->unhash(sk); | |
992 | inet->sport = 0; | |
993 | } | |
994 | sk_dst_reset(sk); | |
995 | return 0; | |
996 | } | |
997 | ||
645ca708 ED |
998 | void udp_lib_unhash(struct sock *sk) |
999 | { | |
723b4610 ED |
1000 | if (sk_hashed(sk)) { |
1001 | struct udp_table *udptable = sk->sk_prot->h.udp_table; | |
1002 | unsigned int hash = udp_hashfn(sock_net(sk), sk->sk_hash); | |
1003 | struct udp_hslot *hslot = &udptable->hash[hash]; | |
645ca708 | 1004 | |
723b4610 ED |
1005 | spin_lock_bh(&hslot->lock); |
1006 | if (sk_nulls_del_node_init_rcu(sk)) { | |
1007 | inet_sk(sk)->num = 0; | |
1008 | sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); | |
1009 | } | |
1010 | spin_unlock_bh(&hslot->lock); | |
645ca708 | 1011 | } |
645ca708 ED |
1012 | } |
1013 | EXPORT_SYMBOL(udp_lib_unhash); | |
1014 | ||
93821778 HX |
1015 | static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) |
1016 | { | |
1017 | int is_udplite = IS_UDPLITE(sk); | |
1018 | int rc; | |
1019 | ||
1020 | if ((rc = sock_queue_rcv_skb(sk, skb)) < 0) { | |
1021 | /* Note that an ENOMEM error is charged twice */ | |
e408b8dc | 1022 | if (rc == -ENOMEM) { |
93821778 HX |
1023 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS, |
1024 | is_udplite); | |
e408b8dc ED |
1025 | atomic_inc(&sk->sk_drops); |
1026 | } | |
93821778 HX |
1027 | goto drop; |
1028 | } | |
1029 | ||
1030 | return 0; | |
1031 | ||
1032 | drop: | |
1033 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); | |
1034 | kfree_skb(skb); | |
1035 | return -1; | |
1036 | } | |
1037 | ||
db8dac20 DM |
1038 | /* returns: |
1039 | * -1: error | |
1040 | * 0: success | |
1041 | * >0: "udp encap" protocol resubmission | |
1042 | * | |
1043 | * Note that in the success and error cases, the skb is assumed to | |
1044 | * have either been requeued or freed. | |
1045 | */ | |
1046 | int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb) | |
1047 | { | |
1048 | struct udp_sock *up = udp_sk(sk); | |
1049 | int rc; | |
1050 | int is_udplite = IS_UDPLITE(sk); | |
1051 | ||
1052 | /* | |
1053 | * Charge it to the socket, dropping if the queue is full. | |
1054 | */ | |
1055 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1056 | goto drop; | |
1057 | nf_reset(skb); | |
1058 | ||
1059 | if (up->encap_type) { | |
1060 | /* | |
1061 | * This is an encapsulation socket so pass the skb to | |
1062 | * the socket's udp_encap_rcv() hook. Otherwise, just | |
1063 | * fall through and pass this up the UDP socket. | |
1064 | * up->encap_rcv() returns the following value: | |
1065 | * =0 if skb was successfully passed to the encap | |
1066 | * handler or was discarded by it. | |
1067 | * >0 if skb should be passed on to UDP. | |
1068 | * <0 if skb should be resubmitted as proto -N | |
1069 | */ | |
1070 | ||
1071 | /* if we're overly short, let UDP handle it */ | |
1072 | if (skb->len > sizeof(struct udphdr) && | |
1073 | up->encap_rcv != NULL) { | |
1074 | int ret; | |
1075 | ||
1076 | ret = (*up->encap_rcv)(sk, skb); | |
1077 | if (ret <= 0) { | |
0283328e PE |
1078 | UDP_INC_STATS_BH(sock_net(sk), |
1079 | UDP_MIB_INDATAGRAMS, | |
db8dac20 DM |
1080 | is_udplite); |
1081 | return -ret; | |
1082 | } | |
1083 | } | |
1084 | ||
1085 | /* FALLTHROUGH -- it's a UDP Packet */ | |
1086 | } | |
1087 | ||
1088 | /* | |
1089 | * UDP-Lite specific tests, ignored on UDP sockets | |
1090 | */ | |
1091 | if ((is_udplite & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) { | |
1092 | ||
1093 | /* | |
1094 | * MIB statistics other than incrementing the error count are | |
1095 | * disabled for the following two types of errors: these depend | |
1096 | * on the application settings, not on the functioning of the | |
1097 | * protocol stack as such. | |
1098 | * | |
1099 | * RFC 3828 here recommends (sec 3.3): "There should also be a | |
1100 | * way ... to ... at least let the receiving application block | |
1101 | * delivery of packets with coverage values less than a value | |
1102 | * provided by the application." | |
1103 | */ | |
1104 | if (up->pcrlen == 0) { /* full coverage was set */ | |
1105 | LIMIT_NETDEBUG(KERN_WARNING "UDPLITE: partial coverage " | |
1106 | "%d while full coverage %d requested\n", | |
1107 | UDP_SKB_CB(skb)->cscov, skb->len); | |
1108 | goto drop; | |
1109 | } | |
1110 | /* The next case involves violating the min. coverage requested | |
1111 | * by the receiver. This is subtle: if receiver wants x and x is | |
1112 | * greater than the buffersize/MTU then receiver will complain | |
1113 | * that it wants x while sender emits packets of smaller size y. | |
1114 | * Therefore the above ...()->partial_cov statement is essential. | |
1115 | */ | |
1116 | if (UDP_SKB_CB(skb)->cscov < up->pcrlen) { | |
1117 | LIMIT_NETDEBUG(KERN_WARNING | |
1118 | "UDPLITE: coverage %d too small, need min %d\n", | |
1119 | UDP_SKB_CB(skb)->cscov, up->pcrlen); | |
1120 | goto drop; | |
1121 | } | |
1122 | } | |
1123 | ||
1124 | if (sk->sk_filter) { | |
1125 | if (udp_lib_checksum_complete(skb)) | |
1126 | goto drop; | |
1127 | } | |
1128 | ||
93821778 | 1129 | rc = 0; |
db8dac20 | 1130 | |
93821778 HX |
1131 | bh_lock_sock(sk); |
1132 | if (!sock_owned_by_user(sk)) | |
1133 | rc = __udp_queue_rcv_skb(sk, skb); | |
1134 | else | |
1135 | sk_add_backlog(sk, skb); | |
1136 | bh_unlock_sock(sk); | |
1137 | ||
1138 | return rc; | |
db8dac20 DM |
1139 | |
1140 | drop: | |
0283328e | 1141 | UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite); |
db8dac20 DM |
1142 | kfree_skb(skb); |
1143 | return -1; | |
1144 | } | |
1145 | ||
1146 | /* | |
1147 | * Multicasts and broadcasts go to each listener. | |
1148 | * | |
1149 | * Note: called only from the BH handler context, | |
1150 | * so we don't need to lock the hashes. | |
1151 | */ | |
e3163493 | 1152 | static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb, |
db8dac20 DM |
1153 | struct udphdr *uh, |
1154 | __be32 saddr, __be32 daddr, | |
645ca708 | 1155 | struct udp_table *udptable) |
db8dac20 DM |
1156 | { |
1157 | struct sock *sk; | |
645ca708 | 1158 | struct udp_hslot *hslot = &udptable->hash[udp_hashfn(net, ntohs(uh->dest))]; |
db8dac20 DM |
1159 | int dif; |
1160 | ||
645ca708 | 1161 | spin_lock(&hslot->lock); |
88ab1932 | 1162 | sk = sk_nulls_head(&hslot->head); |
db8dac20 | 1163 | dif = skb->dev->ifindex; |
920a4611 | 1164 | sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif); |
db8dac20 DM |
1165 | if (sk) { |
1166 | struct sock *sknext = NULL; | |
1167 | ||
1168 | do { | |
1169 | struct sk_buff *skb1 = skb; | |
1170 | ||
88ab1932 | 1171 | sknext = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest, |
920a4611 ED |
1172 | daddr, uh->source, saddr, |
1173 | dif); | |
db8dac20 DM |
1174 | if (sknext) |
1175 | skb1 = skb_clone(skb, GFP_ATOMIC); | |
1176 | ||
1177 | if (skb1) { | |
93821778 | 1178 | int ret = udp_queue_rcv_skb(sk, skb1); |
db8dac20 DM |
1179 | if (ret > 0) |
1180 | /* we should probably re-process instead | |
1181 | * of dropping packets here. */ | |
1182 | kfree_skb(skb1); | |
1183 | } | |
1184 | sk = sknext; | |
1185 | } while (sknext); | |
1186 | } else | |
ead2ceb0 | 1187 | consume_skb(skb); |
645ca708 | 1188 | spin_unlock(&hslot->lock); |
db8dac20 DM |
1189 | return 0; |
1190 | } | |
1191 | ||
1192 | /* Initialize UDP checksum. If exited with zero value (success), | |
1193 | * CHECKSUM_UNNECESSARY means, that no more checks are required. | |
1194 | * Otherwise, csum completion requires chacksumming packet body, | |
1195 | * including udp header and folding it to skb->csum. | |
1196 | */ | |
1197 | static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh, | |
1198 | int proto) | |
1199 | { | |
1200 | const struct iphdr *iph; | |
1201 | int err; | |
1202 | ||
1203 | UDP_SKB_CB(skb)->partial_cov = 0; | |
1204 | UDP_SKB_CB(skb)->cscov = skb->len; | |
1205 | ||
1206 | if (proto == IPPROTO_UDPLITE) { | |
1207 | err = udplite_checksum_init(skb, uh); | |
1208 | if (err) | |
1209 | return err; | |
1210 | } | |
1211 | ||
1212 | iph = ip_hdr(skb); | |
1213 | if (uh->check == 0) { | |
1214 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1215 | } else if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
1216 | if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len, | |
1217 | proto, skb->csum)) | |
1218 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1219 | } | |
1220 | if (!skb_csum_unnecessary(skb)) | |
1221 | skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, | |
1222 | skb->len, proto, 0); | |
1223 | /* Probably, we should checksum udp header (it should be in cache | |
1224 | * in any case) and data in tiny packets (< rx copybreak). | |
1225 | */ | |
1226 | ||
1227 | return 0; | |
1228 | } | |
1229 | ||
1230 | /* | |
1231 | * All we need to do is get the socket, and then do a checksum. | |
1232 | */ | |
1233 | ||
645ca708 | 1234 | int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable, |
db8dac20 DM |
1235 | int proto) |
1236 | { | |
1237 | struct sock *sk; | |
7b5e56f9 | 1238 | struct udphdr *uh; |
db8dac20 DM |
1239 | unsigned short ulen; |
1240 | struct rtable *rt = (struct rtable*)skb->dst; | |
2783ef23 | 1241 | __be32 saddr, daddr; |
0283328e | 1242 | struct net *net = dev_net(skb->dev); |
db8dac20 DM |
1243 | |
1244 | /* | |
1245 | * Validate the packet. | |
1246 | */ | |
1247 | if (!pskb_may_pull(skb, sizeof(struct udphdr))) | |
1248 | goto drop; /* No space for header. */ | |
1249 | ||
7b5e56f9 | 1250 | uh = udp_hdr(skb); |
db8dac20 DM |
1251 | ulen = ntohs(uh->len); |
1252 | if (ulen > skb->len) | |
1253 | goto short_packet; | |
1254 | ||
1255 | if (proto == IPPROTO_UDP) { | |
1256 | /* UDP validates ulen. */ | |
1257 | if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen)) | |
1258 | goto short_packet; | |
1259 | uh = udp_hdr(skb); | |
1260 | } | |
1261 | ||
1262 | if (udp4_csum_init(skb, uh, proto)) | |
1263 | goto csum_error; | |
1264 | ||
2783ef23 JDB |
1265 | saddr = ip_hdr(skb)->saddr; |
1266 | daddr = ip_hdr(skb)->daddr; | |
1267 | ||
db8dac20 | 1268 | if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST)) |
e3163493 PE |
1269 | return __udp4_lib_mcast_deliver(net, skb, uh, |
1270 | saddr, daddr, udptable); | |
db8dac20 | 1271 | |
607c4aaf | 1272 | sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable); |
db8dac20 DM |
1273 | |
1274 | if (sk != NULL) { | |
93821778 | 1275 | int ret = udp_queue_rcv_skb(sk, skb); |
db8dac20 DM |
1276 | sock_put(sk); |
1277 | ||
1278 | /* a return value > 0 means to resubmit the input, but | |
1279 | * it wants the return to be -protocol, or 0 | |
1280 | */ | |
1281 | if (ret > 0) | |
1282 | return -ret; | |
1283 | return 0; | |
1284 | } | |
1285 | ||
1286 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1287 | goto drop; | |
1288 | nf_reset(skb); | |
1289 | ||
1290 | /* No socket. Drop packet silently, if checksum is wrong */ | |
1291 | if (udp_lib_checksum_complete(skb)) | |
1292 | goto csum_error; | |
1293 | ||
0283328e | 1294 | UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1295 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
1296 | ||
1297 | /* | |
1298 | * Hmm. We got an UDP packet to a port to which we | |
1299 | * don't wanna listen. Ignore it. | |
1300 | */ | |
1301 | kfree_skb(skb); | |
1302 | return 0; | |
1303 | ||
1304 | short_packet: | |
673d57e7 | 1305 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n", |
db8dac20 | 1306 | proto == IPPROTO_UDPLITE ? "-Lite" : "", |
673d57e7 | 1307 | &saddr, |
db8dac20 DM |
1308 | ntohs(uh->source), |
1309 | ulen, | |
1310 | skb->len, | |
673d57e7 | 1311 | &daddr, |
db8dac20 DM |
1312 | ntohs(uh->dest)); |
1313 | goto drop; | |
1314 | ||
1315 | csum_error: | |
1316 | /* | |
1317 | * RFC1122: OK. Discards the bad packet silently (as far as | |
1318 | * the network is concerned, anyway) as per 4.1.3.4 (MUST). | |
1319 | */ | |
673d57e7 | 1320 | LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n", |
db8dac20 | 1321 | proto == IPPROTO_UDPLITE ? "-Lite" : "", |
673d57e7 | 1322 | &saddr, |
db8dac20 | 1323 | ntohs(uh->source), |
673d57e7 | 1324 | &daddr, |
db8dac20 DM |
1325 | ntohs(uh->dest), |
1326 | ulen); | |
1327 | drop: | |
0283328e | 1328 | UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE); |
db8dac20 DM |
1329 | kfree_skb(skb); |
1330 | return 0; | |
1331 | } | |
1332 | ||
1333 | int udp_rcv(struct sk_buff *skb) | |
1334 | { | |
645ca708 | 1335 | return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP); |
db8dac20 DM |
1336 | } |
1337 | ||
7d06b2e0 | 1338 | void udp_destroy_sock(struct sock *sk) |
db8dac20 DM |
1339 | { |
1340 | lock_sock(sk); | |
1341 | udp_flush_pending_frames(sk); | |
1342 | release_sock(sk); | |
db8dac20 DM |
1343 | } |
1344 | ||
1da177e4 LT |
1345 | /* |
1346 | * Socket option code for UDP | |
1347 | */ | |
4c0a6cb0 GR |
1348 | int udp_lib_setsockopt(struct sock *sk, int level, int optname, |
1349 | char __user *optval, int optlen, | |
1350 | int (*push_pending_frames)(struct sock *)) | |
1da177e4 LT |
1351 | { |
1352 | struct udp_sock *up = udp_sk(sk); | |
1353 | int val; | |
1354 | int err = 0; | |
b2bf1e26 | 1355 | int is_udplite = IS_UDPLITE(sk); |
1da177e4 | 1356 | |
6516c655 | 1357 | if (optlen<sizeof(int)) |
1da177e4 LT |
1358 | return -EINVAL; |
1359 | ||
1360 | if (get_user(val, (int __user *)optval)) | |
1361 | return -EFAULT; | |
1362 | ||
6516c655 | 1363 | switch (optname) { |
1da177e4 LT |
1364 | case UDP_CORK: |
1365 | if (val != 0) { | |
1366 | up->corkflag = 1; | |
1367 | } else { | |
1368 | up->corkflag = 0; | |
1369 | lock_sock(sk); | |
4c0a6cb0 | 1370 | (*push_pending_frames)(sk); |
1da177e4 LT |
1371 | release_sock(sk); |
1372 | } | |
1373 | break; | |
e905a9ed | 1374 | |
1da177e4 LT |
1375 | case UDP_ENCAP: |
1376 | switch (val) { | |
1377 | case 0: | |
1378 | case UDP_ENCAP_ESPINUDP: | |
1379 | case UDP_ENCAP_ESPINUDP_NON_IKE: | |
067b207b JC |
1380 | up->encap_rcv = xfrm4_udp_encap_rcv; |
1381 | /* FALLTHROUGH */ | |
342f0234 | 1382 | case UDP_ENCAP_L2TPINUDP: |
1da177e4 LT |
1383 | up->encap_type = val; |
1384 | break; | |
1385 | default: | |
1386 | err = -ENOPROTOOPT; | |
1387 | break; | |
1388 | } | |
1389 | break; | |
1390 | ||
ba4e58ec GR |
1391 | /* |
1392 | * UDP-Lite's partial checksum coverage (RFC 3828). | |
1393 | */ | |
1394 | /* The sender sets actual checksum coverage length via this option. | |
1395 | * The case coverage > packet length is handled by send module. */ | |
1396 | case UDPLITE_SEND_CSCOV: | |
b2bf1e26 | 1397 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1398 | return -ENOPROTOOPT; |
1399 | if (val != 0 && val < 8) /* Illegal coverage: use default (8) */ | |
1400 | val = 8; | |
47112e25 GR |
1401 | else if (val > USHORT_MAX) |
1402 | val = USHORT_MAX; | |
ba4e58ec GR |
1403 | up->pcslen = val; |
1404 | up->pcflag |= UDPLITE_SEND_CC; | |
1405 | break; | |
1406 | ||
e905a9ed YH |
1407 | /* The receiver specifies a minimum checksum coverage value. To make |
1408 | * sense, this should be set to at least 8 (as done below). If zero is | |
ba4e58ec GR |
1409 | * used, this again means full checksum coverage. */ |
1410 | case UDPLITE_RECV_CSCOV: | |
b2bf1e26 | 1411 | if (!is_udplite) /* Disable the option on UDP sockets */ |
ba4e58ec GR |
1412 | return -ENOPROTOOPT; |
1413 | if (val != 0 && val < 8) /* Avoid silly minimal values. */ | |
1414 | val = 8; | |
47112e25 GR |
1415 | else if (val > USHORT_MAX) |
1416 | val = USHORT_MAX; | |
ba4e58ec GR |
1417 | up->pcrlen = val; |
1418 | up->pcflag |= UDPLITE_RECV_CC; | |
1419 | break; | |
1420 | ||
1da177e4 LT |
1421 | default: |
1422 | err = -ENOPROTOOPT; | |
1423 | break; | |
6516c655 | 1424 | } |
1da177e4 LT |
1425 | |
1426 | return err; | |
1427 | } | |
1428 | ||
db8dac20 DM |
1429 | int udp_setsockopt(struct sock *sk, int level, int optname, |
1430 | char __user *optval, int optlen) | |
1431 | { | |
1432 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1433 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1434 | udp_push_pending_frames); | |
1435 | return ip_setsockopt(sk, level, optname, optval, optlen); | |
1436 | } | |
1437 | ||
1438 | #ifdef CONFIG_COMPAT | |
1439 | int compat_udp_setsockopt(struct sock *sk, int level, int optname, | |
1440 | char __user *optval, int optlen) | |
1441 | { | |
1442 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1443 | return udp_lib_setsockopt(sk, level, optname, optval, optlen, | |
1444 | udp_push_pending_frames); | |
1445 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); | |
1446 | } | |
1447 | #endif | |
1448 | ||
4c0a6cb0 GR |
1449 | int udp_lib_getsockopt(struct sock *sk, int level, int optname, |
1450 | char __user *optval, int __user *optlen) | |
1da177e4 LT |
1451 | { |
1452 | struct udp_sock *up = udp_sk(sk); | |
1453 | int val, len; | |
1454 | ||
6516c655 | 1455 | if (get_user(len,optlen)) |
1da177e4 LT |
1456 | return -EFAULT; |
1457 | ||
1458 | len = min_t(unsigned int, len, sizeof(int)); | |
e905a9ed | 1459 | |
6516c655 | 1460 | if (len < 0) |
1da177e4 LT |
1461 | return -EINVAL; |
1462 | ||
6516c655 | 1463 | switch (optname) { |
1da177e4 LT |
1464 | case UDP_CORK: |
1465 | val = up->corkflag; | |
1466 | break; | |
1467 | ||
1468 | case UDP_ENCAP: | |
1469 | val = up->encap_type; | |
1470 | break; | |
1471 | ||
ba4e58ec GR |
1472 | /* The following two cannot be changed on UDP sockets, the return is |
1473 | * always 0 (which corresponds to the full checksum coverage of UDP). */ | |
1474 | case UDPLITE_SEND_CSCOV: | |
1475 | val = up->pcslen; | |
1476 | break; | |
1477 | ||
1478 | case UDPLITE_RECV_CSCOV: | |
1479 | val = up->pcrlen; | |
1480 | break; | |
1481 | ||
1da177e4 LT |
1482 | default: |
1483 | return -ENOPROTOOPT; | |
6516c655 | 1484 | } |
1da177e4 | 1485 | |
6516c655 | 1486 | if (put_user(len, optlen)) |
e905a9ed | 1487 | return -EFAULT; |
6516c655 | 1488 | if (copy_to_user(optval, &val,len)) |
1da177e4 | 1489 | return -EFAULT; |
e905a9ed | 1490 | return 0; |
1da177e4 LT |
1491 | } |
1492 | ||
db8dac20 DM |
1493 | int udp_getsockopt(struct sock *sk, int level, int optname, |
1494 | char __user *optval, int __user *optlen) | |
1495 | { | |
1496 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1497 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1498 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
1499 | } | |
1500 | ||
1501 | #ifdef CONFIG_COMPAT | |
1502 | int compat_udp_getsockopt(struct sock *sk, int level, int optname, | |
1503 | char __user *optval, int __user *optlen) | |
1504 | { | |
1505 | if (level == SOL_UDP || level == SOL_UDPLITE) | |
1506 | return udp_lib_getsockopt(sk, level, optname, optval, optlen); | |
1507 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); | |
1508 | } | |
1509 | #endif | |
1da177e4 LT |
1510 | /** |
1511 | * udp_poll - wait for a UDP event. | |
1512 | * @file - file struct | |
1513 | * @sock - socket | |
1514 | * @wait - poll table | |
1515 | * | |
e905a9ed | 1516 | * This is same as datagram poll, except for the special case of |
1da177e4 LT |
1517 | * blocking sockets. If application is using a blocking fd |
1518 | * and a packet with checksum error is in the queue; | |
1519 | * then it could get return from select indicating data available | |
1520 | * but then block when reading it. Add special case code | |
1521 | * to work around these arguably broken applications. | |
1522 | */ | |
1523 | unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait) | |
1524 | { | |
1525 | unsigned int mask = datagram_poll(file, sock, wait); | |
1526 | struct sock *sk = sock->sk; | |
ba4e58ec GR |
1527 | int is_lite = IS_UDPLITE(sk); |
1528 | ||
1da177e4 LT |
1529 | /* Check for false positives due to checksum errors */ |
1530 | if ( (mask & POLLRDNORM) && | |
1531 | !(file->f_flags & O_NONBLOCK) && | |
1532 | !(sk->sk_shutdown & RCV_SHUTDOWN)){ | |
1533 | struct sk_buff_head *rcvq = &sk->sk_receive_queue; | |
1534 | struct sk_buff *skb; | |
1535 | ||
208d8984 | 1536 | spin_lock_bh(&rcvq->lock); |
759e5d00 HX |
1537 | while ((skb = skb_peek(rcvq)) != NULL && |
1538 | udp_lib_checksum_complete(skb)) { | |
0283328e PE |
1539 | UDP_INC_STATS_BH(sock_net(sk), |
1540 | UDP_MIB_INERRORS, is_lite); | |
759e5d00 HX |
1541 | __skb_unlink(skb, rcvq); |
1542 | kfree_skb(skb); | |
1da177e4 | 1543 | } |
208d8984 | 1544 | spin_unlock_bh(&rcvq->lock); |
1da177e4 LT |
1545 | |
1546 | /* nothing to see, move along */ | |
1547 | if (skb == NULL) | |
1548 | mask &= ~(POLLIN | POLLRDNORM); | |
1549 | } | |
1550 | ||
1551 | return mask; | |
e905a9ed | 1552 | |
1da177e4 LT |
1553 | } |
1554 | ||
db8dac20 DM |
1555 | struct proto udp_prot = { |
1556 | .name = "UDP", | |
1557 | .owner = THIS_MODULE, | |
1558 | .close = udp_lib_close, | |
1559 | .connect = ip4_datagram_connect, | |
1560 | .disconnect = udp_disconnect, | |
1561 | .ioctl = udp_ioctl, | |
1562 | .destroy = udp_destroy_sock, | |
1563 | .setsockopt = udp_setsockopt, | |
1564 | .getsockopt = udp_getsockopt, | |
1565 | .sendmsg = udp_sendmsg, | |
1566 | .recvmsg = udp_recvmsg, | |
1567 | .sendpage = udp_sendpage, | |
93821778 | 1568 | .backlog_rcv = __udp_queue_rcv_skb, |
db8dac20 DM |
1569 | .hash = udp_lib_hash, |
1570 | .unhash = udp_lib_unhash, | |
1571 | .get_port = udp_v4_get_port, | |
1572 | .memory_allocated = &udp_memory_allocated, | |
1573 | .sysctl_mem = sysctl_udp_mem, | |
1574 | .sysctl_wmem = &sysctl_udp_wmem_min, | |
1575 | .sysctl_rmem = &sysctl_udp_rmem_min, | |
1576 | .obj_size = sizeof(struct udp_sock), | |
271b72c7 | 1577 | .slab_flags = SLAB_DESTROY_BY_RCU, |
645ca708 | 1578 | .h.udp_table = &udp_table, |
db8dac20 DM |
1579 | #ifdef CONFIG_COMPAT |
1580 | .compat_setsockopt = compat_udp_setsockopt, | |
1581 | .compat_getsockopt = compat_udp_getsockopt, | |
1582 | #endif | |
db8dac20 | 1583 | }; |
1da177e4 LT |
1584 | |
1585 | /* ------------------------------------------------------------------------ */ | |
1586 | #ifdef CONFIG_PROC_FS | |
1587 | ||
645ca708 | 1588 | static struct sock *udp_get_first(struct seq_file *seq, int start) |
1da177e4 LT |
1589 | { |
1590 | struct sock *sk; | |
1591 | struct udp_iter_state *state = seq->private; | |
6f191efe | 1592 | struct net *net = seq_file_net(seq); |
1da177e4 | 1593 | |
645ca708 | 1594 | for (state->bucket = start; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) { |
88ab1932 | 1595 | struct hlist_nulls_node *node; |
645ca708 ED |
1596 | struct udp_hslot *hslot = &state->udp_table->hash[state->bucket]; |
1597 | spin_lock_bh(&hslot->lock); | |
88ab1932 | 1598 | sk_nulls_for_each(sk, node, &hslot->head) { |
878628fb | 1599 | if (!net_eq(sock_net(sk), net)) |
a91275ef | 1600 | continue; |
1da177e4 LT |
1601 | if (sk->sk_family == state->family) |
1602 | goto found; | |
1603 | } | |
645ca708 | 1604 | spin_unlock_bh(&hslot->lock); |
1da177e4 LT |
1605 | } |
1606 | sk = NULL; | |
1607 | found: | |
1608 | return sk; | |
1609 | } | |
1610 | ||
1611 | static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk) | |
1612 | { | |
1613 | struct udp_iter_state *state = seq->private; | |
6f191efe | 1614 | struct net *net = seq_file_net(seq); |
1da177e4 LT |
1615 | |
1616 | do { | |
88ab1932 | 1617 | sk = sk_nulls_next(sk); |
878628fb | 1618 | } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family)); |
1da177e4 | 1619 | |
645ca708 | 1620 | if (!sk) { |
30842f29 VM |
1621 | if (state->bucket < UDP_HTABLE_SIZE) |
1622 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); | |
645ca708 | 1623 | return udp_get_first(seq, state->bucket + 1); |
1da177e4 LT |
1624 | } |
1625 | return sk; | |
1626 | } | |
1627 | ||
1628 | static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos) | |
1629 | { | |
645ca708 | 1630 | struct sock *sk = udp_get_first(seq, 0); |
1da177e4 LT |
1631 | |
1632 | if (sk) | |
6516c655 | 1633 | while (pos && (sk = udp_get_next(seq, sk)) != NULL) |
1da177e4 LT |
1634 | --pos; |
1635 | return pos ? NULL : sk; | |
1636 | } | |
1637 | ||
1638 | static void *udp_seq_start(struct seq_file *seq, loff_t *pos) | |
1639 | { | |
30842f29 VM |
1640 | struct udp_iter_state *state = seq->private; |
1641 | state->bucket = UDP_HTABLE_SIZE; | |
1642 | ||
b50660f1 | 1643 | return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN; |
1da177e4 LT |
1644 | } |
1645 | ||
1646 | static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
1647 | { | |
1648 | struct sock *sk; | |
1649 | ||
b50660f1 | 1650 | if (v == SEQ_START_TOKEN) |
1da177e4 LT |
1651 | sk = udp_get_idx(seq, 0); |
1652 | else | |
1653 | sk = udp_get_next(seq, v); | |
1654 | ||
1655 | ++*pos; | |
1656 | return sk; | |
1657 | } | |
1658 | ||
1659 | static void udp_seq_stop(struct seq_file *seq, void *v) | |
1660 | { | |
645ca708 ED |
1661 | struct udp_iter_state *state = seq->private; |
1662 | ||
1663 | if (state->bucket < UDP_HTABLE_SIZE) | |
1664 | spin_unlock_bh(&state->udp_table->hash[state->bucket].lock); | |
1da177e4 LT |
1665 | } |
1666 | ||
1667 | static int udp_seq_open(struct inode *inode, struct file *file) | |
1668 | { | |
1669 | struct udp_seq_afinfo *afinfo = PDE(inode)->data; | |
a2be75c1 DL |
1670 | struct udp_iter_state *s; |
1671 | int err; | |
a91275ef | 1672 | |
a2be75c1 DL |
1673 | err = seq_open_net(inode, file, &afinfo->seq_ops, |
1674 | sizeof(struct udp_iter_state)); | |
1675 | if (err < 0) | |
1676 | return err; | |
a91275ef | 1677 | |
a2be75c1 | 1678 | s = ((struct seq_file *)file->private_data)->private; |
1da177e4 | 1679 | s->family = afinfo->family; |
645ca708 | 1680 | s->udp_table = afinfo->udp_table; |
a2be75c1 | 1681 | return err; |
a91275ef DL |
1682 | } |
1683 | ||
1da177e4 | 1684 | /* ------------------------------------------------------------------------ */ |
0c96d8c5 | 1685 | int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 LT |
1686 | { |
1687 | struct proc_dir_entry *p; | |
1688 | int rc = 0; | |
1689 | ||
3ba9441b DL |
1690 | afinfo->seq_fops.open = udp_seq_open; |
1691 | afinfo->seq_fops.read = seq_read; | |
1692 | afinfo->seq_fops.llseek = seq_lseek; | |
1693 | afinfo->seq_fops.release = seq_release_net; | |
1da177e4 | 1694 | |
dda61925 DL |
1695 | afinfo->seq_ops.start = udp_seq_start; |
1696 | afinfo->seq_ops.next = udp_seq_next; | |
1697 | afinfo->seq_ops.stop = udp_seq_stop; | |
1698 | ||
84841c3c DL |
1699 | p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, |
1700 | &afinfo->seq_fops, afinfo); | |
1701 | if (!p) | |
1da177e4 LT |
1702 | rc = -ENOMEM; |
1703 | return rc; | |
1704 | } | |
1705 | ||
0c96d8c5 | 1706 | void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo) |
1da177e4 | 1707 | { |
0c96d8c5 | 1708 | proc_net_remove(net, afinfo->name); |
1da177e4 | 1709 | } |
db8dac20 DM |
1710 | |
1711 | /* ------------------------------------------------------------------------ */ | |
5e659e4c PE |
1712 | static void udp4_format_sock(struct sock *sp, struct seq_file *f, |
1713 | int bucket, int *len) | |
db8dac20 DM |
1714 | { |
1715 | struct inet_sock *inet = inet_sk(sp); | |
1716 | __be32 dest = inet->daddr; | |
1717 | __be32 src = inet->rcv_saddr; | |
1718 | __u16 destp = ntohs(inet->dport); | |
1719 | __u16 srcp = ntohs(inet->sport); | |
1720 | ||
5e659e4c | 1721 | seq_printf(f, "%4d: %08X:%04X %08X:%04X" |
cb61cb9b | 1722 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d%n", |
db8dac20 DM |
1723 | bucket, src, srcp, dest, destp, sp->sk_state, |
1724 | atomic_read(&sp->sk_wmem_alloc), | |
1725 | atomic_read(&sp->sk_rmem_alloc), | |
1726 | 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp), | |
cb61cb9b ED |
1727 | atomic_read(&sp->sk_refcnt), sp, |
1728 | atomic_read(&sp->sk_drops), len); | |
db8dac20 DM |
1729 | } |
1730 | ||
1731 | int udp4_seq_show(struct seq_file *seq, void *v) | |
1732 | { | |
1733 | if (v == SEQ_START_TOKEN) | |
1734 | seq_printf(seq, "%-127s\n", | |
1735 | " sl local_address rem_address st tx_queue " | |
1736 | "rx_queue tr tm->when retrnsmt uid timeout " | |
cb61cb9b | 1737 | "inode ref pointer drops"); |
db8dac20 | 1738 | else { |
db8dac20 | 1739 | struct udp_iter_state *state = seq->private; |
5e659e4c | 1740 | int len; |
db8dac20 | 1741 | |
5e659e4c PE |
1742 | udp4_format_sock(v, seq, state->bucket, &len); |
1743 | seq_printf(seq, "%*s\n", 127 - len ,""); | |
db8dac20 DM |
1744 | } |
1745 | return 0; | |
1746 | } | |
1747 | ||
1748 | /* ------------------------------------------------------------------------ */ | |
db8dac20 | 1749 | static struct udp_seq_afinfo udp4_seq_afinfo = { |
db8dac20 DM |
1750 | .name = "udp", |
1751 | .family = AF_INET, | |
645ca708 | 1752 | .udp_table = &udp_table, |
4ad96d39 DL |
1753 | .seq_fops = { |
1754 | .owner = THIS_MODULE, | |
1755 | }, | |
dda61925 DL |
1756 | .seq_ops = { |
1757 | .show = udp4_seq_show, | |
1758 | }, | |
db8dac20 DM |
1759 | }; |
1760 | ||
15439feb PE |
1761 | static int udp4_proc_init_net(struct net *net) |
1762 | { | |
1763 | return udp_proc_register(net, &udp4_seq_afinfo); | |
1764 | } | |
1765 | ||
1766 | static void udp4_proc_exit_net(struct net *net) | |
1767 | { | |
1768 | udp_proc_unregister(net, &udp4_seq_afinfo); | |
1769 | } | |
1770 | ||
1771 | static struct pernet_operations udp4_net_ops = { | |
1772 | .init = udp4_proc_init_net, | |
1773 | .exit = udp4_proc_exit_net, | |
1774 | }; | |
1775 | ||
db8dac20 DM |
1776 | int __init udp4_proc_init(void) |
1777 | { | |
15439feb | 1778 | return register_pernet_subsys(&udp4_net_ops); |
db8dac20 DM |
1779 | } |
1780 | ||
1781 | void udp4_proc_exit(void) | |
1782 | { | |
15439feb | 1783 | unregister_pernet_subsys(&udp4_net_ops); |
db8dac20 | 1784 | } |
1da177e4 LT |
1785 | #endif /* CONFIG_PROC_FS */ |
1786 | ||
645ca708 ED |
1787 | void __init udp_table_init(struct udp_table *table) |
1788 | { | |
1789 | int i; | |
1790 | ||
1791 | for (i = 0; i < UDP_HTABLE_SIZE; i++) { | |
88ab1932 | 1792 | INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i); |
645ca708 ED |
1793 | spin_lock_init(&table->hash[i].lock); |
1794 | } | |
1795 | } | |
1796 | ||
95766fff HA |
1797 | void __init udp_init(void) |
1798 | { | |
8203efb3 | 1799 | unsigned long nr_pages, limit; |
95766fff | 1800 | |
645ca708 | 1801 | udp_table_init(&udp_table); |
95766fff HA |
1802 | /* Set the pressure threshold up by the same strategy of TCP. It is a |
1803 | * fraction of global memory that is up to 1/2 at 256 MB, decreasing | |
1804 | * toward zero with the amount of memory, with a floor of 128 pages. | |
1805 | */ | |
8203efb3 ED |
1806 | nr_pages = totalram_pages - totalhigh_pages; |
1807 | limit = min(nr_pages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT); | |
1808 | limit = (limit * (nr_pages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11); | |
95766fff HA |
1809 | limit = max(limit, 128UL); |
1810 | sysctl_udp_mem[0] = limit / 4 * 3; | |
1811 | sysctl_udp_mem[1] = limit; | |
1812 | sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2; | |
1813 | ||
1814 | sysctl_udp_rmem_min = SK_MEM_QUANTUM; | |
1815 | sysctl_udp_wmem_min = SK_MEM_QUANTUM; | |
1816 | } | |
1817 | ||
1da177e4 | 1818 | EXPORT_SYMBOL(udp_disconnect); |
1da177e4 | 1819 | EXPORT_SYMBOL(udp_ioctl); |
db8dac20 DM |
1820 | EXPORT_SYMBOL(udp_prot); |
1821 | EXPORT_SYMBOL(udp_sendmsg); | |
4c0a6cb0 GR |
1822 | EXPORT_SYMBOL(udp_lib_getsockopt); |
1823 | EXPORT_SYMBOL(udp_lib_setsockopt); | |
1da177e4 | 1824 | EXPORT_SYMBOL(udp_poll); |
6ba5a3c5 | 1825 | EXPORT_SYMBOL(udp_lib_get_port); |
1da177e4 LT |
1826 | |
1827 | #ifdef CONFIG_PROC_FS | |
1828 | EXPORT_SYMBOL(udp_proc_register); | |
1829 | EXPORT_SYMBOL(udp_proc_unregister); | |
1830 | #endif |