blob: e0bd1013cb0d0053c3d4b0f5fe0bbd2befa9ad48 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * The User Datagram Protocol (UDP).
7 *
8 * Version: $Id: udp.c,v 1.102 2002/02/01 22:01:04 davem Exp $
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
13 * Alan Cox, <Alan.Cox@linux.org>
14 * Hirokazu Takahashi, <taka@valinux.co.jp>
15 *
16 * Fixes:
17 * Alan Cox : verify_area() calls
18 * Alan Cox : stopped close while in use off icmp
19 * messages. Not a fix but a botch that
20 * for udp at least is 'valid'.
21 * Alan Cox : Fixed icmp handling properly
22 * Alan Cox : Correct error for oversized datagrams
23 * Alan Cox : Tidied select() semantics.
24 * Alan Cox : udp_err() fixed properly, also now
25 * select and read wake correctly on errors
26 * Alan Cox : udp_send verify_area moved to avoid mem leak
27 * Alan Cox : UDP can count its memory
28 * Alan Cox : send to an unknown connection causes
29 * an ECONNREFUSED off the icmp, but
30 * does NOT close.
31 * Alan Cox : Switched to new sk_buff handlers. No more backlog!
32 * Alan Cox : Using generic datagram code. Even smaller and the PEEK
33 * bug no longer crashes it.
34 * Fred Van Kempen : Net2e support for sk->broadcast.
35 * Alan Cox : Uses skb_free_datagram
36 * Alan Cox : Added get/set sockopt support.
37 * Alan Cox : Broadcasting without option set returns EACCES.
38 * Alan Cox : No wakeup calls. Instead we now use the callbacks.
39 * Alan Cox : Use ip_tos and ip_ttl
40 * Alan Cox : SNMP Mibs
41 * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support.
42 * Matt Dillon : UDP length checks.
43 * Alan Cox : Smarter af_inet used properly.
44 * Alan Cox : Use new kernel side addressing.
45 * Alan Cox : Incorrect return on truncated datagram receive.
46 * Arnt Gulbrandsen : New udp_send and stuff
47 * Alan Cox : Cache last socket
48 * Alan Cox : Route cache
49 * Jon Peatfield : Minor efficiency fix to sendto().
50 * Mike Shaver : RFC1122 checks.
51 * Alan Cox : Nonblocking error fix.
52 * Willy Konynenberg : Transparent proxying support.
53 * Mike McLagan : Routing by source
54 * David S. Miller : New socket lookup architecture.
55 * Last socket cache retained as it
56 * does have a high hit rate.
57 * Olaf Kirch : Don't linearise iovec on sendmsg.
58 * Andi Kleen : Some cleanups, cache destination entry
59 * for connect.
60 * Vitaly E. Lavrov : Transparent proxy revived after year coma.
61 * Melvin Smith : Check msg_name not msg_namelen in sendto(),
62 * return ENOTCONN for unconnected sockets (POSIX)
63 * Janos Farkas : don't deliver multi/broadcasts to a different
64 * bound-to-device socket
65 * Hirokazu Takahashi : HW checksumming for outgoing UDP
66 * datagrams.
67 * Hirokazu Takahashi : sendfile() on UDP works now.
68 * Arnaldo C. Melo : convert /proc/net/udp to seq_file
69 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
70 * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind
71 * a single port at the same time.
72 * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support
73 *
74 *
75 * This program is free software; you can redistribute it and/or
76 * modify it under the terms of the GNU General Public License
77 * as published by the Free Software Foundation; either version
78 * 2 of the License, or (at your option) any later version.
79 */
80
81#include <asm/system.h>
82#include <asm/uaccess.h>
83#include <asm/ioctls.h>
84#include <linux/types.h>
85#include <linux/fcntl.h>
86#include <linux/module.h>
87#include <linux/socket.h>
88#include <linux/sockios.h>
89#include <linux/in.h>
90#include <linux/errno.h>
91#include <linux/timer.h>
92#include <linux/mm.h>
93#include <linux/config.h>
94#include <linux/inet.h>
95#include <linux/ipv6.h>
96#include <linux/netdevice.h>
97#include <net/snmp.h>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070098#include <net/ip.h>
99#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100#include <net/protocol.h>
101#include <linux/skbuff.h>
102#include <linux/proc_fs.h>
103#include <linux/seq_file.h>
104#include <net/sock.h>
105#include <net/udp.h>
106#include <net/icmp.h>
107#include <net/route.h>
108#include <net/inet_common.h>
109#include <net/checksum.h>
110#include <net/xfrm.h>
111
112/*
113 * Snmp MIB for the UDP layer
114 */
115
Eric Dumazetba899662005-08-26 12:05:31 -0700116DEFINE_SNMP_STAT(struct udp_mib, udp_statistics) __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
118struct hlist_head udp_hash[UDP_HTABLE_SIZE];
119DEFINE_RWLOCK(udp_hash_lock);
120
121/* Shared by v4/v6 udp. */
122int udp_port_rover;
123
124static int udp_v4_get_port(struct sock *sk, unsigned short snum)
125{
126 struct hlist_node *node;
127 struct sock *sk2;
128 struct inet_sock *inet = inet_sk(sk);
129
130 write_lock_bh(&udp_hash_lock);
131 if (snum == 0) {
132 int best_size_so_far, best, result, i;
133
134 if (udp_port_rover > sysctl_local_port_range[1] ||
135 udp_port_rover < sysctl_local_port_range[0])
136 udp_port_rover = sysctl_local_port_range[0];
137 best_size_so_far = 32767;
138 best = result = udp_port_rover;
139 for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) {
140 struct hlist_head *list;
141 int size;
142
143 list = &udp_hash[result & (UDP_HTABLE_SIZE - 1)];
144 if (hlist_empty(list)) {
145 if (result > sysctl_local_port_range[1])
146 result = sysctl_local_port_range[0] +
147 ((result - sysctl_local_port_range[0]) &
148 (UDP_HTABLE_SIZE - 1));
149 goto gotit;
150 }
151 size = 0;
152 sk_for_each(sk2, node, list)
153 if (++size >= best_size_so_far)
154 goto next;
155 best_size_so_far = size;
156 best = result;
157 next:;
158 }
159 result = best;
160 for(i = 0; i < (1 << 16) / UDP_HTABLE_SIZE; i++, result += UDP_HTABLE_SIZE) {
161 if (result > sysctl_local_port_range[1])
162 result = sysctl_local_port_range[0]
163 + ((result - sysctl_local_port_range[0]) &
164 (UDP_HTABLE_SIZE - 1));
165 if (!udp_lport_inuse(result))
166 break;
167 }
168 if (i >= (1 << 16) / UDP_HTABLE_SIZE)
169 goto fail;
170gotit:
171 udp_port_rover = snum = result;
172 } else {
173 sk_for_each(sk2, node,
174 &udp_hash[snum & (UDP_HTABLE_SIZE - 1)]) {
175 struct inet_sock *inet2 = inet_sk(sk2);
176
177 if (inet2->num == snum &&
178 sk2 != sk &&
179 !ipv6_only_sock(sk2) &&
180 (!sk2->sk_bound_dev_if ||
181 !sk->sk_bound_dev_if ||
182 sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
183 (!inet2->rcv_saddr ||
184 !inet->rcv_saddr ||
185 inet2->rcv_saddr == inet->rcv_saddr) &&
186 (!sk2->sk_reuse || !sk->sk_reuse))
187 goto fail;
188 }
189 }
190 inet->num = snum;
191 if (sk_unhashed(sk)) {
192 struct hlist_head *h = &udp_hash[snum & (UDP_HTABLE_SIZE - 1)];
193
194 sk_add_node(sk, h);
195 sock_prot_inc_use(sk->sk_prot);
196 }
197 write_unlock_bh(&udp_hash_lock);
198 return 0;
199
200fail:
201 write_unlock_bh(&udp_hash_lock);
202 return 1;
203}
204
205static void udp_v4_hash(struct sock *sk)
206{
207 BUG();
208}
209
210static void udp_v4_unhash(struct sock *sk)
211{
212 write_lock_bh(&udp_hash_lock);
213 if (sk_del_node_init(sk)) {
214 inet_sk(sk)->num = 0;
215 sock_prot_dec_use(sk->sk_prot);
216 }
217 write_unlock_bh(&udp_hash_lock);
218}
219
220/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
221 * harder than this. -DaveM
222 */
223static struct sock *udp_v4_lookup_longway(u32 saddr, u16 sport,
224 u32 daddr, u16 dport, int dif)
225{
226 struct sock *sk, *result = NULL;
227 struct hlist_node *node;
228 unsigned short hnum = ntohs(dport);
229 int badness = -1;
230
231 sk_for_each(sk, node, &udp_hash[hnum & (UDP_HTABLE_SIZE - 1)]) {
232 struct inet_sock *inet = inet_sk(sk);
233
234 if (inet->num == hnum && !ipv6_only_sock(sk)) {
235 int score = (sk->sk_family == PF_INET ? 1 : 0);
236 if (inet->rcv_saddr) {
237 if (inet->rcv_saddr != daddr)
238 continue;
239 score+=2;
240 }
241 if (inet->daddr) {
242 if (inet->daddr != saddr)
243 continue;
244 score+=2;
245 }
246 if (inet->dport) {
247 if (inet->dport != sport)
248 continue;
249 score+=2;
250 }
251 if (sk->sk_bound_dev_if) {
252 if (sk->sk_bound_dev_if != dif)
253 continue;
254 score+=2;
255 }
256 if(score == 9) {
257 result = sk;
258 break;
259 } else if(score > badness) {
260 result = sk;
261 badness = score;
262 }
263 }
264 }
265 return result;
266}
267
268static __inline__ struct sock *udp_v4_lookup(u32 saddr, u16 sport,
269 u32 daddr, u16 dport, int dif)
270{
271 struct sock *sk;
272
273 read_lock(&udp_hash_lock);
274 sk = udp_v4_lookup_longway(saddr, sport, daddr, dport, dif);
275 if (sk)
276 sock_hold(sk);
277 read_unlock(&udp_hash_lock);
278 return sk;
279}
280
281static inline struct sock *udp_v4_mcast_next(struct sock *sk,
282 u16 loc_port, u32 loc_addr,
283 u16 rmt_port, u32 rmt_addr,
284 int dif)
285{
286 struct hlist_node *node;
287 struct sock *s = sk;
288 unsigned short hnum = ntohs(loc_port);
289
290 sk_for_each_from(s, node) {
291 struct inet_sock *inet = inet_sk(s);
292
293 if (inet->num != hnum ||
294 (inet->daddr && inet->daddr != rmt_addr) ||
295 (inet->dport != rmt_port && inet->dport) ||
296 (inet->rcv_saddr && inet->rcv_saddr != loc_addr) ||
297 ipv6_only_sock(s) ||
298 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
299 continue;
300 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
301 continue;
302 goto found;
303 }
304 s = NULL;
305found:
306 return s;
307}
308
309/*
310 * This routine is called by the ICMP module when it gets some
311 * sort of error condition. If err < 0 then the socket should
312 * be closed and the error returned to the user. If err > 0
313 * it's just the icmp type << 8 | icmp code.
314 * Header points to the ip header of the error packet. We move
315 * on past this. Then (as it used to claim before adjustment)
316 * header points to the first 8 bytes of the udp header. We need
317 * to find the appropriate port.
318 */
319
320void udp_err(struct sk_buff *skb, u32 info)
321{
322 struct inet_sock *inet;
323 struct iphdr *iph = (struct iphdr*)skb->data;
324 struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2));
325 int type = skb->h.icmph->type;
326 int code = skb->h.icmph->code;
327 struct sock *sk;
328 int harderr;
329 int err;
330
331 sk = udp_v4_lookup(iph->daddr, uh->dest, iph->saddr, uh->source, skb->dev->ifindex);
332 if (sk == NULL) {
333 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
334 return; /* No socket for error */
335 }
336
337 err = 0;
338 harderr = 0;
339 inet = inet_sk(sk);
340
341 switch (type) {
342 default:
343 case ICMP_TIME_EXCEEDED:
344 err = EHOSTUNREACH;
345 break;
346 case ICMP_SOURCE_QUENCH:
347 goto out;
348 case ICMP_PARAMETERPROB:
349 err = EPROTO;
350 harderr = 1;
351 break;
352 case ICMP_DEST_UNREACH:
353 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
354 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
355 err = EMSGSIZE;
356 harderr = 1;
357 break;
358 }
359 goto out;
360 }
361 err = EHOSTUNREACH;
362 if (code <= NR_ICMP_UNREACH) {
363 harderr = icmp_err_convert[code].fatal;
364 err = icmp_err_convert[code].errno;
365 }
366 break;
367 }
368
369 /*
370 * RFC1122: OK. Passes ICMP errors back to application, as per
371 * 4.1.3.3.
372 */
373 if (!inet->recverr) {
374 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
375 goto out;
376 } else {
377 ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1));
378 }
379 sk->sk_err = err;
380 sk->sk_error_report(sk);
381out:
382 sock_put(sk);
383}
384
385/*
386 * Throw away all pending data and cancel the corking. Socket is locked.
387 */
388static void udp_flush_pending_frames(struct sock *sk)
389{
390 struct udp_sock *up = udp_sk(sk);
391
392 if (up->pending) {
393 up->len = 0;
394 up->pending = 0;
395 ip_flush_pending_frames(sk);
396 }
397}
398
399/*
400 * Push out all pending data as one UDP datagram. Socket is locked.
401 */
402static int udp_push_pending_frames(struct sock *sk, struct udp_sock *up)
403{
404 struct inet_sock *inet = inet_sk(sk);
405 struct flowi *fl = &inet->cork.fl;
406 struct sk_buff *skb;
407 struct udphdr *uh;
408 int err = 0;
409
410 /* Grab the skbuff where UDP header space exists. */
411 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
412 goto out;
413
414 /*
415 * Create a UDP header
416 */
417 uh = skb->h.uh;
418 uh->source = fl->fl_ip_sport;
419 uh->dest = fl->fl_ip_dport;
420 uh->len = htons(up->len);
421 uh->check = 0;
422
423 if (sk->sk_no_check == UDP_CSUM_NOXMIT) {
424 skb->ip_summed = CHECKSUM_NONE;
425 goto send;
426 }
427
428 if (skb_queue_len(&sk->sk_write_queue) == 1) {
429 /*
430 * Only one fragment on the socket.
431 */
432 if (skb->ip_summed == CHECKSUM_HW) {
433 skb->csum = offsetof(struct udphdr, check);
434 uh->check = ~csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
435 up->len, IPPROTO_UDP, 0);
436 } else {
437 skb->csum = csum_partial((char *)uh,
438 sizeof(struct udphdr), skb->csum);
439 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
440 up->len, IPPROTO_UDP, skb->csum);
441 if (uh->check == 0)
442 uh->check = -1;
443 }
444 } else {
445 unsigned int csum = 0;
446 /*
447 * HW-checksum won't work as there are two or more
448 * fragments on the socket so that all csums of sk_buffs
449 * should be together.
450 */
451 if (skb->ip_summed == CHECKSUM_HW) {
452 int offset = (unsigned char *)uh - skb->data;
453 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
454
455 skb->ip_summed = CHECKSUM_NONE;
456 } else {
457 skb->csum = csum_partial((char *)uh,
458 sizeof(struct udphdr), skb->csum);
459 }
460
461 skb_queue_walk(&sk->sk_write_queue, skb) {
462 csum = csum_add(csum, skb->csum);
463 }
464 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
465 up->len, IPPROTO_UDP, csum);
466 if (uh->check == 0)
467 uh->check = -1;
468 }
469send:
470 err = ip_push_pending_frames(sk);
471out:
472 up->len = 0;
473 up->pending = 0;
474 return err;
475}
476
477
478static unsigned short udp_check(struct udphdr *uh, int len, unsigned long saddr, unsigned long daddr, unsigned long base)
479{
480 return(csum_tcpudp_magic(saddr, daddr, len, IPPROTO_UDP, base));
481}
482
483int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
484 size_t len)
485{
486 struct inet_sock *inet = inet_sk(sk);
487 struct udp_sock *up = udp_sk(sk);
488 int ulen = len;
489 struct ipcm_cookie ipc;
490 struct rtable *rt = NULL;
491 int free = 0;
492 int connected = 0;
493 u32 daddr, faddr, saddr;
494 u16 dport;
495 u8 tos;
496 int err;
497 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
498
499 if (len > 0xFFFF)
500 return -EMSGSIZE;
501
502 /*
503 * Check the flags.
504 */
505
506 if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */
507 return -EOPNOTSUPP;
508
509 ipc.opt = NULL;
510
511 if (up->pending) {
512 /*
513 * There are pending frames.
514 * The socket lock must be held while it's corked.
515 */
516 lock_sock(sk);
517 if (likely(up->pending)) {
518 if (unlikely(up->pending != AF_INET)) {
519 release_sock(sk);
520 return -EINVAL;
521 }
522 goto do_append_data;
523 }
524 release_sock(sk);
525 }
526 ulen += sizeof(struct udphdr);
527
528 /*
529 * Get and verify the address.
530 */
531 if (msg->msg_name) {
532 struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name;
533 if (msg->msg_namelen < sizeof(*usin))
534 return -EINVAL;
535 if (usin->sin_family != AF_INET) {
536 if (usin->sin_family != AF_UNSPEC)
537 return -EAFNOSUPPORT;
538 }
539
540 daddr = usin->sin_addr.s_addr;
541 dport = usin->sin_port;
542 if (dport == 0)
543 return -EINVAL;
544 } else {
545 if (sk->sk_state != TCP_ESTABLISHED)
546 return -EDESTADDRREQ;
547 daddr = inet->daddr;
548 dport = inet->dport;
549 /* Open fast path for connected socket.
550 Route will not be used, if at least one option is set.
551 */
552 connected = 1;
553 }
554 ipc.addr = inet->saddr;
555
556 ipc.oif = sk->sk_bound_dev_if;
557 if (msg->msg_controllen) {
558 err = ip_cmsg_send(msg, &ipc);
559 if (err)
560 return err;
561 if (ipc.opt)
562 free = 1;
563 connected = 0;
564 }
565 if (!ipc.opt)
566 ipc.opt = inet->opt;
567
568 saddr = ipc.addr;
569 ipc.addr = faddr = daddr;
570
571 if (ipc.opt && ipc.opt->srr) {
572 if (!daddr)
573 return -EINVAL;
574 faddr = ipc.opt->faddr;
575 connected = 0;
576 }
577 tos = RT_TOS(inet->tos);
578 if (sock_flag(sk, SOCK_LOCALROUTE) ||
579 (msg->msg_flags & MSG_DONTROUTE) ||
580 (ipc.opt && ipc.opt->is_strictroute)) {
581 tos |= RTO_ONLINK;
582 connected = 0;
583 }
584
585 if (MULTICAST(daddr)) {
586 if (!ipc.oif)
587 ipc.oif = inet->mc_index;
588 if (!saddr)
589 saddr = inet->mc_addr;
590 connected = 0;
591 }
592
593 if (connected)
594 rt = (struct rtable*)sk_dst_check(sk, 0);
595
596 if (rt == NULL) {
597 struct flowi fl = { .oif = ipc.oif,
598 .nl_u = { .ip4_u =
599 { .daddr = faddr,
600 .saddr = saddr,
601 .tos = tos } },
602 .proto = IPPROTO_UDP,
603 .uli_u = { .ports =
604 { .sport = inet->sport,
605 .dport = dport } } };
606 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
607 if (err)
608 goto out;
609
610 err = -EACCES;
611 if ((rt->rt_flags & RTCF_BROADCAST) &&
612 !sock_flag(sk, SOCK_BROADCAST))
613 goto out;
614 if (connected)
615 sk_dst_set(sk, dst_clone(&rt->u.dst));
616 }
617
618 if (msg->msg_flags&MSG_CONFIRM)
619 goto do_confirm;
620back_from_confirm:
621
622 saddr = rt->rt_src;
623 if (!ipc.addr)
624 daddr = ipc.addr = rt->rt_dst;
625
626 lock_sock(sk);
627 if (unlikely(up->pending)) {
628 /* The socket is already corked while preparing it. */
629 /* ... which is an evident application bug. --ANK */
630 release_sock(sk);
631
Patrick McHardy64ce2072005-08-09 20:50:53 -0700632 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 err = -EINVAL;
634 goto out;
635 }
636 /*
637 * Now cork the socket to pend data.
638 */
639 inet->cork.fl.fl4_dst = daddr;
640 inet->cork.fl.fl_ip_dport = dport;
641 inet->cork.fl.fl4_src = saddr;
642 inet->cork.fl.fl_ip_sport = inet->sport;
643 up->pending = AF_INET;
644
645do_append_data:
646 up->len += ulen;
647 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, ulen,
648 sizeof(struct udphdr), &ipc, rt,
649 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
650 if (err)
651 udp_flush_pending_frames(sk);
652 else if (!corkreq)
653 err = udp_push_pending_frames(sk, up);
654 release_sock(sk);
655
656out:
657 ip_rt_put(rt);
658 if (free)
659 kfree(ipc.opt);
660 if (!err) {
661 UDP_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS);
662 return len;
663 }
664 return err;
665
666do_confirm:
667 dst_confirm(&rt->u.dst);
668 if (!(msg->msg_flags&MSG_PROBE) || len)
669 goto back_from_confirm;
670 err = 0;
671 goto out;
672}
673
674static int udp_sendpage(struct sock *sk, struct page *page, int offset,
675 size_t size, int flags)
676{
677 struct udp_sock *up = udp_sk(sk);
678 int ret;
679
680 if (!up->pending) {
681 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
682
683 /* Call udp_sendmsg to specify destination address which
684 * sendpage interface can't pass.
685 * This will succeed only when the socket is connected.
686 */
687 ret = udp_sendmsg(NULL, sk, &msg, 0);
688 if (ret < 0)
689 return ret;
690 }
691
692 lock_sock(sk);
693
694 if (unlikely(!up->pending)) {
695 release_sock(sk);
696
Patrick McHardy64ce2072005-08-09 20:50:53 -0700697 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 return -EINVAL;
699 }
700
701 ret = ip_append_page(sk, page, offset, size, flags);
702 if (ret == -EOPNOTSUPP) {
703 release_sock(sk);
704 return sock_no_sendpage(sk->sk_socket, page, offset,
705 size, flags);
706 }
707 if (ret < 0) {
708 udp_flush_pending_frames(sk);
709 goto out;
710 }
711
712 up->len += size;
713 if (!(up->corkflag || (flags&MSG_MORE)))
714 ret = udp_push_pending_frames(sk, up);
715 if (!ret)
716 ret = size;
717out:
718 release_sock(sk);
719 return ret;
720}
721
722/*
723 * IOCTL requests applicable to the UDP protocol
724 */
725
726int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
727{
728 switch(cmd)
729 {
730 case SIOCOUTQ:
731 {
732 int amount = atomic_read(&sk->sk_wmem_alloc);
733 return put_user(amount, (int __user *)arg);
734 }
735
736 case SIOCINQ:
737 {
738 struct sk_buff *skb;
739 unsigned long amount;
740
741 amount = 0;
Herbert Xu208d8982005-05-30 15:50:15 -0700742 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 skb = skb_peek(&sk->sk_receive_queue);
744 if (skb != NULL) {
745 /*
746 * We will only return the amount
747 * of this packet since that is all
748 * that will be read.
749 */
750 amount = skb->len - sizeof(struct udphdr);
751 }
Herbert Xu208d8982005-05-30 15:50:15 -0700752 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 return put_user(amount, (int __user *)arg);
754 }
755
756 default:
757 return -ENOIOCTLCMD;
758 }
759 return(0);
760}
761
762static __inline__ int __udp_checksum_complete(struct sk_buff *skb)
763{
764 return (unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum));
765}
766
767static __inline__ int udp_checksum_complete(struct sk_buff *skb)
768{
769 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
770 __udp_checksum_complete(skb);
771}
772
773/*
774 * This should be easy, if there is something there we
775 * return it, otherwise we block.
776 */
777
778static int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
779 size_t len, int noblock, int flags, int *addr_len)
780{
781 struct inet_sock *inet = inet_sk(sk);
782 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
783 struct sk_buff *skb;
784 int copied, err;
785
786 /*
787 * Check any passed addresses
788 */
789 if (addr_len)
790 *addr_len=sizeof(*sin);
791
792 if (flags & MSG_ERRQUEUE)
793 return ip_recv_error(sk, msg, len);
794
795try_again:
796 skb = skb_recv_datagram(sk, flags, noblock, &err);
797 if (!skb)
798 goto out;
799
800 copied = skb->len - sizeof(struct udphdr);
801 if (copied > len) {
802 copied = len;
803 msg->msg_flags |= MSG_TRUNC;
804 }
805
806 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
807 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
808 copied);
809 } else if (msg->msg_flags&MSG_TRUNC) {
810 if (__udp_checksum_complete(skb))
811 goto csum_copy_err;
812 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
813 copied);
814 } else {
815 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
816
817 if (err == -EINVAL)
818 goto csum_copy_err;
819 }
820
821 if (err)
822 goto out_free;
823
824 sock_recv_timestamp(msg, sk, skb);
825
826 /* Copy the address. */
827 if (sin)
828 {
829 sin->sin_family = AF_INET;
830 sin->sin_port = skb->h.uh->source;
831 sin->sin_addr.s_addr = skb->nh.iph->saddr;
832 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
833 }
834 if (inet->cmsg_flags)
835 ip_cmsg_recv(msg, skb);
836
837 err = copied;
838 if (flags & MSG_TRUNC)
839 err = skb->len - sizeof(struct udphdr);
840
841out_free:
842 skb_free_datagram(sk, skb);
843out:
844 return err;
845
846csum_copy_err:
847 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
848
849 /* Clear queue. */
850 if (flags&MSG_PEEK) {
851 int clear = 0;
Herbert Xu208d8982005-05-30 15:50:15 -0700852 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 if (skb == skb_peek(&sk->sk_receive_queue)) {
854 __skb_unlink(skb, &sk->sk_receive_queue);
855 clear = 1;
856 }
Herbert Xu208d8982005-05-30 15:50:15 -0700857 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 if (clear)
859 kfree_skb(skb);
860 }
861
862 skb_free_datagram(sk, skb);
863
864 if (noblock)
865 return -EAGAIN;
866 goto try_again;
867}
868
869
870int udp_disconnect(struct sock *sk, int flags)
871{
872 struct inet_sock *inet = inet_sk(sk);
873 /*
874 * 1003.1g - break association.
875 */
876
877 sk->sk_state = TCP_CLOSE;
878 inet->daddr = 0;
879 inet->dport = 0;
880 sk->sk_bound_dev_if = 0;
881 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
882 inet_reset_saddr(sk);
883
884 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
885 sk->sk_prot->unhash(sk);
886 inet->sport = 0;
887 }
888 sk_dst_reset(sk);
889 return 0;
890}
891
892static void udp_close(struct sock *sk, long timeout)
893{
894 sk_common_release(sk);
895}
896
897/* return:
898 * 1 if the the UDP system should process it
899 * 0 if we should drop this packet
900 * -1 if it should get processed by xfrm4_rcv_encap
901 */
902static int udp_encap_rcv(struct sock * sk, struct sk_buff *skb)
903{
904#ifndef CONFIG_XFRM
905 return 1;
906#else
907 struct udp_sock *up = udp_sk(sk);
908 struct udphdr *uh = skb->h.uh;
909 struct iphdr *iph;
910 int iphlen, len;
911
912 __u8 *udpdata = (__u8 *)uh + sizeof(struct udphdr);
913 __u32 *udpdata32 = (__u32 *)udpdata;
914 __u16 encap_type = up->encap_type;
915
916 /* if we're overly short, let UDP handle it */
917 if (udpdata > skb->tail)
918 return 1;
919
920 /* if this is not encapsulated socket, then just return now */
921 if (!encap_type)
922 return 1;
923
924 len = skb->tail - udpdata;
925
926 switch (encap_type) {
927 default:
928 case UDP_ENCAP_ESPINUDP:
929 /* Check if this is a keepalive packet. If so, eat it. */
930 if (len == 1 && udpdata[0] == 0xff) {
931 return 0;
932 } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0 ) {
933 /* ESP Packet without Non-ESP header */
934 len = sizeof(struct udphdr);
935 } else
936 /* Must be an IKE packet.. pass it through */
937 return 1;
938 break;
939 case UDP_ENCAP_ESPINUDP_NON_IKE:
940 /* Check if this is a keepalive packet. If so, eat it. */
941 if (len == 1 && udpdata[0] == 0xff) {
942 return 0;
943 } else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) &&
944 udpdata32[0] == 0 && udpdata32[1] == 0) {
945
946 /* ESP Packet with Non-IKE marker */
947 len = sizeof(struct udphdr) + 2 * sizeof(u32);
948 } else
949 /* Must be an IKE packet.. pass it through */
950 return 1;
951 break;
952 }
953
954 /* At this point we are sure that this is an ESPinUDP packet,
955 * so we need to remove 'len' bytes from the packet (the UDP
956 * header and optional ESP marker bytes) and then modify the
957 * protocol to ESP, and then call into the transform receiver.
958 */
Herbert Xu4d78b6c2005-04-19 22:48:59 -0700959 if (skb_cloned(skb) && pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
960 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961
962 /* Now we can update and verify the packet length... */
963 iph = skb->nh.iph;
964 iphlen = iph->ihl << 2;
965 iph->tot_len = htons(ntohs(iph->tot_len) - len);
966 if (skb->len < iphlen + len) {
967 /* packet is too small!?! */
968 return 0;
969 }
970
971 /* pull the data buffer up to the ESP header and set the
972 * transport header to point to ESP. Keep UDP on the stack
973 * for later.
974 */
975 skb->h.raw = skb_pull(skb, len);
976
977 /* modify the protocol (it's ESP!) */
978 iph->protocol = IPPROTO_ESP;
979
980 /* and let the caller know to send this into the ESP processor... */
981 return -1;
982#endif
983}
984
985/* returns:
986 * -1: error
987 * 0: success
988 * >0: "udp encap" protocol resubmission
989 *
990 * Note that in the success and error cases, the skb is assumed to
991 * have either been requeued or freed.
992 */
993static int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
994{
995 struct udp_sock *up = udp_sk(sk);
996
997 /*
998 * Charge it to the socket, dropping if the queue is full.
999 */
1000 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
1001 kfree_skb(skb);
1002 return -1;
1003 }
1004
1005 if (up->encap_type) {
1006 /*
1007 * This is an encapsulation socket, so let's see if this is
1008 * an encapsulated packet.
1009 * If it's a keepalive packet, then just eat it.
1010 * If it's an encapsulateed packet, then pass it to the
1011 * IPsec xfrm input and return the response
1012 * appropriately. Otherwise, just fall through and
1013 * pass this up the UDP socket.
1014 */
1015 int ret;
1016
1017 ret = udp_encap_rcv(sk, skb);
1018 if (ret == 0) {
1019 /* Eat the packet .. */
1020 kfree_skb(skb);
1021 return 0;
1022 }
1023 if (ret < 0) {
1024 /* process the ESP packet */
1025 ret = xfrm4_rcv_encap(skb, up->encap_type);
1026 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS);
1027 return -ret;
1028 }
1029 /* FALLTHROUGH -- it's a UDP Packet */
1030 }
1031
1032 if (sk->sk_filter && skb->ip_summed != CHECKSUM_UNNECESSARY) {
1033 if (__udp_checksum_complete(skb)) {
1034 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1035 kfree_skb(skb);
1036 return -1;
1037 }
1038 skb->ip_summed = CHECKSUM_UNNECESSARY;
1039 }
1040
1041 if (sock_queue_rcv_skb(sk,skb)<0) {
1042 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1043 kfree_skb(skb);
1044 return -1;
1045 }
1046 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS);
1047 return 0;
1048}
1049
1050/*
1051 * Multicasts and broadcasts go to each listener.
1052 *
1053 * Note: called only from the BH handler context,
1054 * so we don't need to lock the hashes.
1055 */
1056static int udp_v4_mcast_deliver(struct sk_buff *skb, struct udphdr *uh,
1057 u32 saddr, u32 daddr)
1058{
1059 struct sock *sk;
1060 int dif;
1061
1062 read_lock(&udp_hash_lock);
1063 sk = sk_head(&udp_hash[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
1064 dif = skb->dev->ifindex;
1065 sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
1066 if (sk) {
1067 struct sock *sknext = NULL;
1068
1069 do {
1070 struct sk_buff *skb1 = skb;
1071
1072 sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr,
1073 uh->source, saddr, dif);
1074 if(sknext)
1075 skb1 = skb_clone(skb, GFP_ATOMIC);
1076
1077 if(skb1) {
1078 int ret = udp_queue_rcv_skb(sk, skb1);
1079 if (ret > 0)
1080 /* we should probably re-process instead
1081 * of dropping packets here. */
1082 kfree_skb(skb1);
1083 }
1084 sk = sknext;
1085 } while(sknext);
1086 } else
1087 kfree_skb(skb);
1088 read_unlock(&udp_hash_lock);
1089 return 0;
1090}
1091
1092/* Initialize UDP checksum. If exited with zero value (success),
1093 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1094 * Otherwise, csum completion requires chacksumming packet body,
1095 * including udp header and folding it to skb->csum.
1096 */
1097static int udp_checksum_init(struct sk_buff *skb, struct udphdr *uh,
1098 unsigned short ulen, u32 saddr, u32 daddr)
1099{
1100 if (uh->check == 0) {
1101 skb->ip_summed = CHECKSUM_UNNECESSARY;
1102 } else if (skb->ip_summed == CHECKSUM_HW) {
1103 skb->ip_summed = CHECKSUM_UNNECESSARY;
1104 if (!udp_check(uh, ulen, saddr, daddr, skb->csum))
1105 return 0;
Patrick McHardy64ce2072005-08-09 20:50:53 -07001106 LIMIT_NETDEBUG(KERN_DEBUG "udp v4 hw csum failure.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 skb->ip_summed = CHECKSUM_NONE;
1108 }
1109 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
1110 skb->csum = csum_tcpudp_nofold(saddr, daddr, ulen, IPPROTO_UDP, 0);
1111 /* Probably, we should checksum udp header (it should be in cache
1112 * in any case) and data in tiny packets (< rx copybreak).
1113 */
1114 return 0;
1115}
1116
1117/*
1118 * All we need to do is get the socket, and then do a checksum.
1119 */
1120
1121int udp_rcv(struct sk_buff *skb)
1122{
1123 struct sock *sk;
1124 struct udphdr *uh;
1125 unsigned short ulen;
1126 struct rtable *rt = (struct rtable*)skb->dst;
1127 u32 saddr = skb->nh.iph->saddr;
1128 u32 daddr = skb->nh.iph->daddr;
1129 int len = skb->len;
1130
1131 /*
1132 * Validate the packet and the UDP length.
1133 */
1134 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1135 goto no_header;
1136
1137 uh = skb->h.uh;
1138
1139 ulen = ntohs(uh->len);
1140
1141 if (ulen > len || ulen < sizeof(*uh))
1142 goto short_packet;
1143
Stephen Hemmingere308e252005-09-08 12:32:21 -07001144 if (pskb_trim_rcsum(skb, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145 goto short_packet;
1146
1147 if (udp_checksum_init(skb, uh, ulen, saddr, daddr) < 0)
1148 goto csum_error;
1149
1150 if(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1151 return udp_v4_mcast_deliver(skb, uh, saddr, daddr);
1152
1153 sk = udp_v4_lookup(saddr, uh->source, daddr, uh->dest, skb->dev->ifindex);
1154
1155 if (sk != NULL) {
1156 int ret = udp_queue_rcv_skb(sk, skb);
1157 sock_put(sk);
1158
1159 /* a return value > 0 means to resubmit the input, but
1160 * it it wants the return to be -protocol, or 0
1161 */
1162 if (ret > 0)
1163 return -ret;
1164 return 0;
1165 }
1166
1167 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1168 goto drop;
1169
1170 /* No socket. Drop packet silently, if checksum is wrong */
1171 if (udp_checksum_complete(skb))
1172 goto csum_error;
1173
1174 UDP_INC_STATS_BH(UDP_MIB_NOPORTS);
1175 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1176
1177 /*
1178 * Hmm. We got an UDP packet to a port to which we
1179 * don't wanna listen. Ignore it.
1180 */
1181 kfree_skb(skb);
1182 return(0);
1183
1184short_packet:
Patrick McHardy64ce2072005-08-09 20:50:53 -07001185 LIMIT_NETDEBUG(KERN_DEBUG "UDP: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n",
1186 NIPQUAD(saddr),
1187 ntohs(uh->source),
1188 ulen,
1189 len,
1190 NIPQUAD(daddr),
1191 ntohs(uh->dest));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192no_header:
1193 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1194 kfree_skb(skb);
1195 return(0);
1196
1197csum_error:
1198 /*
1199 * RFC1122: OK. Discards the bad packet silently (as far as
1200 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1201 */
Patrick McHardy64ce2072005-08-09 20:50:53 -07001202 LIMIT_NETDEBUG(KERN_DEBUG "UDP: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n",
1203 NIPQUAD(saddr),
1204 ntohs(uh->source),
1205 NIPQUAD(daddr),
1206 ntohs(uh->dest),
1207 ulen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208drop:
1209 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1210 kfree_skb(skb);
1211 return(0);
1212}
1213
1214static int udp_destroy_sock(struct sock *sk)
1215{
1216 lock_sock(sk);
1217 udp_flush_pending_frames(sk);
1218 release_sock(sk);
1219 return 0;
1220}
1221
1222/*
1223 * Socket option code for UDP
1224 */
1225static int udp_setsockopt(struct sock *sk, int level, int optname,
1226 char __user *optval, int optlen)
1227{
1228 struct udp_sock *up = udp_sk(sk);
1229 int val;
1230 int err = 0;
1231
1232 if (level != SOL_UDP)
1233 return ip_setsockopt(sk, level, optname, optval, optlen);
1234
1235 if(optlen<sizeof(int))
1236 return -EINVAL;
1237
1238 if (get_user(val, (int __user *)optval))
1239 return -EFAULT;
1240
1241 switch(optname) {
1242 case UDP_CORK:
1243 if (val != 0) {
1244 up->corkflag = 1;
1245 } else {
1246 up->corkflag = 0;
1247 lock_sock(sk);
1248 udp_push_pending_frames(sk, up);
1249 release_sock(sk);
1250 }
1251 break;
1252
1253 case UDP_ENCAP:
1254 switch (val) {
1255 case 0:
1256 case UDP_ENCAP_ESPINUDP:
1257 case UDP_ENCAP_ESPINUDP_NON_IKE:
1258 up->encap_type = val;
1259 break;
1260 default:
1261 err = -ENOPROTOOPT;
1262 break;
1263 }
1264 break;
1265
1266 default:
1267 err = -ENOPROTOOPT;
1268 break;
1269 };
1270
1271 return err;
1272}
1273
1274static int udp_getsockopt(struct sock *sk, int level, int optname,
1275 char __user *optval, int __user *optlen)
1276{
1277 struct udp_sock *up = udp_sk(sk);
1278 int val, len;
1279
1280 if (level != SOL_UDP)
1281 return ip_getsockopt(sk, level, optname, optval, optlen);
1282
1283 if(get_user(len,optlen))
1284 return -EFAULT;
1285
1286 len = min_t(unsigned int, len, sizeof(int));
1287
1288 if(len < 0)
1289 return -EINVAL;
1290
1291 switch(optname) {
1292 case UDP_CORK:
1293 val = up->corkflag;
1294 break;
1295
1296 case UDP_ENCAP:
1297 val = up->encap_type;
1298 break;
1299
1300 default:
1301 return -ENOPROTOOPT;
1302 };
1303
1304 if(put_user(len, optlen))
1305 return -EFAULT;
1306 if(copy_to_user(optval, &val,len))
1307 return -EFAULT;
1308 return 0;
1309}
1310
1311/**
1312 * udp_poll - wait for a UDP event.
1313 * @file - file struct
1314 * @sock - socket
1315 * @wait - poll table
1316 *
1317 * This is same as datagram poll, except for the special case of
1318 * blocking sockets. If application is using a blocking fd
1319 * and a packet with checksum error is in the queue;
1320 * then it could get return from select indicating data available
1321 * but then block when reading it. Add special case code
1322 * to work around these arguably broken applications.
1323 */
1324unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1325{
1326 unsigned int mask = datagram_poll(file, sock, wait);
1327 struct sock *sk = sock->sk;
1328
1329 /* Check for false positives due to checksum errors */
1330 if ( (mask & POLLRDNORM) &&
1331 !(file->f_flags & O_NONBLOCK) &&
1332 !(sk->sk_shutdown & RCV_SHUTDOWN)){
1333 struct sk_buff_head *rcvq = &sk->sk_receive_queue;
1334 struct sk_buff *skb;
1335
Herbert Xu208d8982005-05-30 15:50:15 -07001336 spin_lock_bh(&rcvq->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 while ((skb = skb_peek(rcvq)) != NULL) {
1338 if (udp_checksum_complete(skb)) {
1339 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1340 __skb_unlink(skb, rcvq);
1341 kfree_skb(skb);
1342 } else {
1343 skb->ip_summed = CHECKSUM_UNNECESSARY;
1344 break;
1345 }
1346 }
Herbert Xu208d8982005-05-30 15:50:15 -07001347 spin_unlock_bh(&rcvq->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348
1349 /* nothing to see, move along */
1350 if (skb == NULL)
1351 mask &= ~(POLLIN | POLLRDNORM);
1352 }
1353
1354 return mask;
1355
1356}
1357
1358struct proto udp_prot = {
1359 .name = "UDP",
1360 .owner = THIS_MODULE,
1361 .close = udp_close,
1362 .connect = ip4_datagram_connect,
1363 .disconnect = udp_disconnect,
1364 .ioctl = udp_ioctl,
1365 .destroy = udp_destroy_sock,
1366 .setsockopt = udp_setsockopt,
1367 .getsockopt = udp_getsockopt,
1368 .sendmsg = udp_sendmsg,
1369 .recvmsg = udp_recvmsg,
1370 .sendpage = udp_sendpage,
1371 .backlog_rcv = udp_queue_rcv_skb,
1372 .hash = udp_v4_hash,
1373 .unhash = udp_v4_unhash,
1374 .get_port = udp_v4_get_port,
1375 .obj_size = sizeof(struct udp_sock),
1376};
1377
1378/* ------------------------------------------------------------------------ */
1379#ifdef CONFIG_PROC_FS
1380
1381static struct sock *udp_get_first(struct seq_file *seq)
1382{
1383 struct sock *sk;
1384 struct udp_iter_state *state = seq->private;
1385
1386 for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) {
1387 struct hlist_node *node;
1388 sk_for_each(sk, node, &udp_hash[state->bucket]) {
1389 if (sk->sk_family == state->family)
1390 goto found;
1391 }
1392 }
1393 sk = NULL;
1394found:
1395 return sk;
1396}
1397
1398static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
1399{
1400 struct udp_iter_state *state = seq->private;
1401
1402 do {
1403 sk = sk_next(sk);
1404try_again:
1405 ;
1406 } while (sk && sk->sk_family != state->family);
1407
1408 if (!sk && ++state->bucket < UDP_HTABLE_SIZE) {
1409 sk = sk_head(&udp_hash[state->bucket]);
1410 goto try_again;
1411 }
1412 return sk;
1413}
1414
1415static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
1416{
1417 struct sock *sk = udp_get_first(seq);
1418
1419 if (sk)
1420 while(pos && (sk = udp_get_next(seq, sk)) != NULL)
1421 --pos;
1422 return pos ? NULL : sk;
1423}
1424
1425static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
1426{
1427 read_lock(&udp_hash_lock);
1428 return *pos ? udp_get_idx(seq, *pos-1) : (void *)1;
1429}
1430
1431static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1432{
1433 struct sock *sk;
1434
1435 if (v == (void *)1)
1436 sk = udp_get_idx(seq, 0);
1437 else
1438 sk = udp_get_next(seq, v);
1439
1440 ++*pos;
1441 return sk;
1442}
1443
1444static void udp_seq_stop(struct seq_file *seq, void *v)
1445{
1446 read_unlock(&udp_hash_lock);
1447}
1448
1449static int udp_seq_open(struct inode *inode, struct file *file)
1450{
1451 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
1452 struct seq_file *seq;
1453 int rc = -ENOMEM;
1454 struct udp_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
1455
1456 if (!s)
1457 goto out;
1458 memset(s, 0, sizeof(*s));
1459 s->family = afinfo->family;
1460 s->seq_ops.start = udp_seq_start;
1461 s->seq_ops.next = udp_seq_next;
1462 s->seq_ops.show = afinfo->seq_show;
1463 s->seq_ops.stop = udp_seq_stop;
1464
1465 rc = seq_open(file, &s->seq_ops);
1466 if (rc)
1467 goto out_kfree;
1468
1469 seq = file->private_data;
1470 seq->private = s;
1471out:
1472 return rc;
1473out_kfree:
1474 kfree(s);
1475 goto out;
1476}
1477
1478/* ------------------------------------------------------------------------ */
1479int udp_proc_register(struct udp_seq_afinfo *afinfo)
1480{
1481 struct proc_dir_entry *p;
1482 int rc = 0;
1483
1484 if (!afinfo)
1485 return -EINVAL;
1486 afinfo->seq_fops->owner = afinfo->owner;
1487 afinfo->seq_fops->open = udp_seq_open;
1488 afinfo->seq_fops->read = seq_read;
1489 afinfo->seq_fops->llseek = seq_lseek;
1490 afinfo->seq_fops->release = seq_release_private;
1491
1492 p = proc_net_fops_create(afinfo->name, S_IRUGO, afinfo->seq_fops);
1493 if (p)
1494 p->data = afinfo;
1495 else
1496 rc = -ENOMEM;
1497 return rc;
1498}
1499
1500void udp_proc_unregister(struct udp_seq_afinfo *afinfo)
1501{
1502 if (!afinfo)
1503 return;
1504 proc_net_remove(afinfo->name);
1505 memset(afinfo->seq_fops, 0, sizeof(*afinfo->seq_fops));
1506}
1507
1508/* ------------------------------------------------------------------------ */
1509static void udp4_format_sock(struct sock *sp, char *tmpbuf, int bucket)
1510{
1511 struct inet_sock *inet = inet_sk(sp);
1512 unsigned int dest = inet->daddr;
1513 unsigned int src = inet->rcv_saddr;
1514 __u16 destp = ntohs(inet->dport);
1515 __u16 srcp = ntohs(inet->sport);
1516
1517 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
1518 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
1519 bucket, src, srcp, dest, destp, sp->sk_state,
1520 atomic_read(&sp->sk_wmem_alloc),
1521 atomic_read(&sp->sk_rmem_alloc),
1522 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
1523 atomic_read(&sp->sk_refcnt), sp);
1524}
1525
1526static int udp4_seq_show(struct seq_file *seq, void *v)
1527{
1528 if (v == SEQ_START_TOKEN)
1529 seq_printf(seq, "%-127s\n",
1530 " sl local_address rem_address st tx_queue "
1531 "rx_queue tr tm->when retrnsmt uid timeout "
1532 "inode");
1533 else {
1534 char tmpbuf[129];
1535 struct udp_iter_state *state = seq->private;
1536
1537 udp4_format_sock(v, tmpbuf, state->bucket);
1538 seq_printf(seq, "%-127s\n", tmpbuf);
1539 }
1540 return 0;
1541}
1542
1543/* ------------------------------------------------------------------------ */
1544static struct file_operations udp4_seq_fops;
1545static struct udp_seq_afinfo udp4_seq_afinfo = {
1546 .owner = THIS_MODULE,
1547 .name = "udp",
1548 .family = AF_INET,
1549 .seq_show = udp4_seq_show,
1550 .seq_fops = &udp4_seq_fops,
1551};
1552
1553int __init udp4_proc_init(void)
1554{
1555 return udp_proc_register(&udp4_seq_afinfo);
1556}
1557
1558void udp4_proc_exit(void)
1559{
1560 udp_proc_unregister(&udp4_seq_afinfo);
1561}
1562#endif /* CONFIG_PROC_FS */
1563
1564EXPORT_SYMBOL(udp_disconnect);
1565EXPORT_SYMBOL(udp_hash);
1566EXPORT_SYMBOL(udp_hash_lock);
1567EXPORT_SYMBOL(udp_ioctl);
1568EXPORT_SYMBOL(udp_port_rover);
1569EXPORT_SYMBOL(udp_prot);
1570EXPORT_SYMBOL(udp_sendmsg);
1571EXPORT_SYMBOL(udp_poll);
1572
1573#ifdef CONFIG_PROC_FS
1574EXPORT_SYMBOL(udp_proc_register);
1575EXPORT_SYMBOL(udp_proc_unregister);
1576#endif