]> bbs.cooldavid.org Git - net-next-2.6.git/blame - net/ipv4/udp.c
[UDP]: Unify UDPv4 and UDPv6 ->get_port()
[net-next-2.6.git] / net / ipv4 / udp.c
CommitLineData
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 *
8 * Version: $Id: udp.c,v 1.102 2002/02/01 22:01:04 davem Exp $
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * 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>
14c85021 89#include <linux/igmp.h>
1da177e4
LT
90#include <linux/in.h>
91#include <linux/errno.h>
92#include <linux/timer.h>
93#include <linux/mm.h>
1da177e4
LT
94#include <linux/inet.h>
95#include <linux/ipv6.h>
96#include <linux/netdevice.h>
97#include <net/snmp.h>
c752f073
ACM
98#include <net/ip.h>
99#include <net/tcp_states.h>
1da177e4
LT
100#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
ba89966c 116DEFINE_SNMP_STAT(struct udp_mib, udp_statistics) __read_mostly;
1da177e4
LT
117
118struct hlist_head udp_hash[UDP_HTABLE_SIZE];
119DEFINE_RWLOCK(udp_hash_lock);
120
25030a7f 121/* Shared by v4/v6 udp_get_port */
1da177e4
LT
122int udp_port_rover;
123
25030a7f 124static inline int udp_lport_inuse(u16 num)
1da177e4 125{
25030a7f 126 struct sock *sk;
1da177e4 127 struct hlist_node *node;
25030a7f
GR
128
129 sk_for_each(sk, node, &udp_hash[num & (UDP_HTABLE_SIZE - 1)])
130 if (inet_sk(sk)->num == num)
131 return 1;
132 return 0;
133}
134
135/**
136 * udp_get_port - common port lookup for IPv4 and IPv6
137 *
138 * @sk: socket struct in question
139 * @snum: port number to look up
140 * @saddr_comp: AF-dependent comparison of bound local IP addresses
141 */
142int udp_get_port(struct sock *sk, unsigned short snum,
143 int (*saddr_cmp)(struct sock *sk1, struct sock *sk2))
144{
145 struct hlist_node *node;
146 struct hlist_head *head;
1da177e4 147 struct sock *sk2;
25030a7f 148 int error = 1;
1da177e4
LT
149
150 write_lock_bh(&udp_hash_lock);
151 if (snum == 0) {
152 int best_size_so_far, best, result, i;
153
154 if (udp_port_rover > sysctl_local_port_range[1] ||
155 udp_port_rover < sysctl_local_port_range[0])
156 udp_port_rover = sysctl_local_port_range[0];
157 best_size_so_far = 32767;
158 best = result = udp_port_rover;
159 for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) {
1da177e4
LT
160 int size;
161
25030a7f
GR
162 head = &udp_hash[result & (UDP_HTABLE_SIZE - 1)];
163 if (hlist_empty(head)) {
1da177e4
LT
164 if (result > sysctl_local_port_range[1])
165 result = sysctl_local_port_range[0] +
166 ((result - sysctl_local_port_range[0]) &
167 (UDP_HTABLE_SIZE - 1));
168 goto gotit;
169 }
170 size = 0;
25030a7f
GR
171 sk_for_each(sk2, node, head)
172 if (++size < best_size_so_far) {
173 best_size_so_far = size;
174 best = result;
175 }
1da177e4
LT
176 }
177 result = best;
178 for(i = 0; i < (1 << 16) / UDP_HTABLE_SIZE; i++, result += UDP_HTABLE_SIZE) {
179 if (result > sysctl_local_port_range[1])
180 result = sysctl_local_port_range[0]
181 + ((result - sysctl_local_port_range[0]) &
182 (UDP_HTABLE_SIZE - 1));
183 if (!udp_lport_inuse(result))
184 break;
185 }
186 if (i >= (1 << 16) / UDP_HTABLE_SIZE)
187 goto fail;
188gotit:
189 udp_port_rover = snum = result;
190 } else {
25030a7f
GR
191 head = &udp_hash[snum & (UDP_HTABLE_SIZE - 1)];
192
193 sk_for_each(sk2, node, head)
194 if (inet_sk(sk2)->num == snum &&
195 sk2 != sk &&
196 (!sk2->sk_reuse || !sk->sk_reuse) &&
197 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if
198 || sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
199 (*saddr_cmp)(sk, sk2) )
1da177e4 200 goto fail;
1da177e4 201 }
25030a7f 202 inet_sk(sk)->num = snum;
1da177e4 203 if (sk_unhashed(sk)) {
25030a7f
GR
204 head = &udp_hash[snum & (UDP_HTABLE_SIZE - 1)];
205 sk_add_node(sk, head);
1da177e4
LT
206 sock_prot_inc_use(sk->sk_prot);
207 }
25030a7f 208 error = 0;
1da177e4
LT
209fail:
210 write_unlock_bh(&udp_hash_lock);
25030a7f
GR
211 return error;
212}
213
214static inline int ipv4_rcv_saddr_equal(struct sock *sk1, struct sock *sk2)
215{
216 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
217
218 return ( !ipv6_only_sock(sk2) &&
219 (!inet1->rcv_saddr || !inet2->rcv_saddr ||
220 inet1->rcv_saddr == inet2->rcv_saddr ));
221}
222
223static inline int udp_v4_get_port(struct sock *sk, unsigned short snum)
224{
225 return udp_get_port(sk, snum, ipv4_rcv_saddr_equal);
1da177e4
LT
226}
227
25030a7f 228
1da177e4
LT
229static void udp_v4_hash(struct sock *sk)
230{
231 BUG();
232}
233
234static void udp_v4_unhash(struct sock *sk)
235{
236 write_lock_bh(&udp_hash_lock);
237 if (sk_del_node_init(sk)) {
238 inet_sk(sk)->num = 0;
239 sock_prot_dec_use(sk->sk_prot);
240 }
241 write_unlock_bh(&udp_hash_lock);
242}
243
244/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
245 * harder than this. -DaveM
246 */
247static struct sock *udp_v4_lookup_longway(u32 saddr, u16 sport,
248 u32 daddr, u16 dport, int dif)
249{
250 struct sock *sk, *result = NULL;
251 struct hlist_node *node;
252 unsigned short hnum = ntohs(dport);
253 int badness = -1;
254
255 sk_for_each(sk, node, &udp_hash[hnum & (UDP_HTABLE_SIZE - 1)]) {
256 struct inet_sock *inet = inet_sk(sk);
257
258 if (inet->num == hnum && !ipv6_only_sock(sk)) {
259 int score = (sk->sk_family == PF_INET ? 1 : 0);
260 if (inet->rcv_saddr) {
261 if (inet->rcv_saddr != daddr)
262 continue;
263 score+=2;
264 }
265 if (inet->daddr) {
266 if (inet->daddr != saddr)
267 continue;
268 score+=2;
269 }
270 if (inet->dport) {
271 if (inet->dport != sport)
272 continue;
273 score+=2;
274 }
275 if (sk->sk_bound_dev_if) {
276 if (sk->sk_bound_dev_if != dif)
277 continue;
278 score+=2;
279 }
280 if(score == 9) {
281 result = sk;
282 break;
283 } else if(score > badness) {
284 result = sk;
285 badness = score;
286 }
287 }
288 }
289 return result;
290}
291
292static __inline__ struct sock *udp_v4_lookup(u32 saddr, u16 sport,
293 u32 daddr, u16 dport, int dif)
294{
295 struct sock *sk;
296
297 read_lock(&udp_hash_lock);
298 sk = udp_v4_lookup_longway(saddr, sport, daddr, dport, dif);
299 if (sk)
300 sock_hold(sk);
301 read_unlock(&udp_hash_lock);
302 return sk;
303}
304
305static inline struct sock *udp_v4_mcast_next(struct sock *sk,
306 u16 loc_port, u32 loc_addr,
307 u16 rmt_port, u32 rmt_addr,
308 int dif)
309{
310 struct hlist_node *node;
311 struct sock *s = sk;
312 unsigned short hnum = ntohs(loc_port);
313
314 sk_for_each_from(s, node) {
315 struct inet_sock *inet = inet_sk(s);
316
317 if (inet->num != hnum ||
318 (inet->daddr && inet->daddr != rmt_addr) ||
319 (inet->dport != rmt_port && inet->dport) ||
320 (inet->rcv_saddr && inet->rcv_saddr != loc_addr) ||
321 ipv6_only_sock(s) ||
322 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
323 continue;
324 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
325 continue;
326 goto found;
327 }
328 s = NULL;
329found:
330 return s;
331}
332
333/*
334 * This routine is called by the ICMP module when it gets some
335 * sort of error condition. If err < 0 then the socket should
336 * be closed and the error returned to the user. If err > 0
337 * it's just the icmp type << 8 | icmp code.
338 * Header points to the ip header of the error packet. We move
339 * on past this. Then (as it used to claim before adjustment)
340 * header points to the first 8 bytes of the udp header. We need
341 * to find the appropriate port.
342 */
343
344void udp_err(struct sk_buff *skb, u32 info)
345{
346 struct inet_sock *inet;
347 struct iphdr *iph = (struct iphdr*)skb->data;
348 struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2));
349 int type = skb->h.icmph->type;
350 int code = skb->h.icmph->code;
351 struct sock *sk;
352 int harderr;
353 int err;
354
355 sk = udp_v4_lookup(iph->daddr, uh->dest, iph->saddr, uh->source, skb->dev->ifindex);
356 if (sk == NULL) {
357 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
358 return; /* No socket for error */
359 }
360
361 err = 0;
362 harderr = 0;
363 inet = inet_sk(sk);
364
365 switch (type) {
366 default:
367 case ICMP_TIME_EXCEEDED:
368 err = EHOSTUNREACH;
369 break;
370 case ICMP_SOURCE_QUENCH:
371 goto out;
372 case ICMP_PARAMETERPROB:
373 err = EPROTO;
374 harderr = 1;
375 break;
376 case ICMP_DEST_UNREACH:
377 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
378 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
379 err = EMSGSIZE;
380 harderr = 1;
381 break;
382 }
383 goto out;
384 }
385 err = EHOSTUNREACH;
386 if (code <= NR_ICMP_UNREACH) {
387 harderr = icmp_err_convert[code].fatal;
388 err = icmp_err_convert[code].errno;
389 }
390 break;
391 }
392
393 /*
394 * RFC1122: OK. Passes ICMP errors back to application, as per
395 * 4.1.3.3.
396 */
397 if (!inet->recverr) {
398 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
399 goto out;
400 } else {
401 ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1));
402 }
403 sk->sk_err = err;
404 sk->sk_error_report(sk);
405out:
406 sock_put(sk);
407}
408
409/*
410 * Throw away all pending data and cancel the corking. Socket is locked.
411 */
412static void udp_flush_pending_frames(struct sock *sk)
413{
414 struct udp_sock *up = udp_sk(sk);
415
416 if (up->pending) {
417 up->len = 0;
418 up->pending = 0;
419 ip_flush_pending_frames(sk);
420 }
421}
422
423/*
424 * Push out all pending data as one UDP datagram. Socket is locked.
425 */
426static int udp_push_pending_frames(struct sock *sk, struct udp_sock *up)
427{
428 struct inet_sock *inet = inet_sk(sk);
429 struct flowi *fl = &inet->cork.fl;
430 struct sk_buff *skb;
431 struct udphdr *uh;
432 int err = 0;
433
434 /* Grab the skbuff where UDP header space exists. */
435 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
436 goto out;
437
438 /*
439 * Create a UDP header
440 */
441 uh = skb->h.uh;
442 uh->source = fl->fl_ip_sport;
443 uh->dest = fl->fl_ip_dport;
444 uh->len = htons(up->len);
445 uh->check = 0;
446
447 if (sk->sk_no_check == UDP_CSUM_NOXMIT) {
448 skb->ip_summed = CHECKSUM_NONE;
449 goto send;
450 }
451
452 if (skb_queue_len(&sk->sk_write_queue) == 1) {
453 /*
454 * Only one fragment on the socket.
455 */
84fa7933 456 if (skb->ip_summed == CHECKSUM_PARTIAL) {
1da177e4
LT
457 skb->csum = offsetof(struct udphdr, check);
458 uh->check = ~csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
459 up->len, IPPROTO_UDP, 0);
460 } else {
461 skb->csum = csum_partial((char *)uh,
462 sizeof(struct udphdr), skb->csum);
463 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
464 up->len, IPPROTO_UDP, skb->csum);
465 if (uh->check == 0)
466 uh->check = -1;
467 }
468 } else {
469 unsigned int csum = 0;
470 /*
471 * HW-checksum won't work as there are two or more
472 * fragments on the socket so that all csums of sk_buffs
473 * should be together.
474 */
84fa7933 475 if (skb->ip_summed == CHECKSUM_PARTIAL) {
1da177e4
LT
476 int offset = (unsigned char *)uh - skb->data;
477 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
478
479 skb->ip_summed = CHECKSUM_NONE;
480 } else {
481 skb->csum = csum_partial((char *)uh,
482 sizeof(struct udphdr), skb->csum);
483 }
484
485 skb_queue_walk(&sk->sk_write_queue, skb) {
486 csum = csum_add(csum, skb->csum);
487 }
488 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst,
489 up->len, IPPROTO_UDP, csum);
490 if (uh->check == 0)
491 uh->check = -1;
492 }
493send:
494 err = ip_push_pending_frames(sk);
495out:
496 up->len = 0;
497 up->pending = 0;
498 return err;
499}
500
501
502static unsigned short udp_check(struct udphdr *uh, int len, unsigned long saddr, unsigned long daddr, unsigned long base)
503{
504 return(csum_tcpudp_magic(saddr, daddr, len, IPPROTO_UDP, base));
505}
506
507int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
508 size_t len)
509{
510 struct inet_sock *inet = inet_sk(sk);
511 struct udp_sock *up = udp_sk(sk);
512 int ulen = len;
513 struct ipcm_cookie ipc;
514 struct rtable *rt = NULL;
515 int free = 0;
516 int connected = 0;
517 u32 daddr, faddr, saddr;
518 u16 dport;
519 u8 tos;
520 int err;
521 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
522
523 if (len > 0xFFFF)
524 return -EMSGSIZE;
525
526 /*
527 * Check the flags.
528 */
529
530 if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */
531 return -EOPNOTSUPP;
532
533 ipc.opt = NULL;
534
535 if (up->pending) {
536 /*
537 * There are pending frames.
538 * The socket lock must be held while it's corked.
539 */
540 lock_sock(sk);
541 if (likely(up->pending)) {
542 if (unlikely(up->pending != AF_INET)) {
543 release_sock(sk);
544 return -EINVAL;
545 }
546 goto do_append_data;
547 }
548 release_sock(sk);
549 }
550 ulen += sizeof(struct udphdr);
551
552 /*
553 * Get and verify the address.
554 */
555 if (msg->msg_name) {
556 struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name;
557 if (msg->msg_namelen < sizeof(*usin))
558 return -EINVAL;
559 if (usin->sin_family != AF_INET) {
560 if (usin->sin_family != AF_UNSPEC)
561 return -EAFNOSUPPORT;
562 }
563
564 daddr = usin->sin_addr.s_addr;
565 dport = usin->sin_port;
566 if (dport == 0)
567 return -EINVAL;
568 } else {
569 if (sk->sk_state != TCP_ESTABLISHED)
570 return -EDESTADDRREQ;
571 daddr = inet->daddr;
572 dport = inet->dport;
573 /* Open fast path for connected socket.
574 Route will not be used, if at least one option is set.
575 */
576 connected = 1;
577 }
578 ipc.addr = inet->saddr;
579
580 ipc.oif = sk->sk_bound_dev_if;
581 if (msg->msg_controllen) {
582 err = ip_cmsg_send(msg, &ipc);
583 if (err)
584 return err;
585 if (ipc.opt)
586 free = 1;
587 connected = 0;
588 }
589 if (!ipc.opt)
590 ipc.opt = inet->opt;
591
592 saddr = ipc.addr;
593 ipc.addr = faddr = daddr;
594
595 if (ipc.opt && ipc.opt->srr) {
596 if (!daddr)
597 return -EINVAL;
598 faddr = ipc.opt->faddr;
599 connected = 0;
600 }
601 tos = RT_TOS(inet->tos);
602 if (sock_flag(sk, SOCK_LOCALROUTE) ||
603 (msg->msg_flags & MSG_DONTROUTE) ||
604 (ipc.opt && ipc.opt->is_strictroute)) {
605 tos |= RTO_ONLINK;
606 connected = 0;
607 }
608
609 if (MULTICAST(daddr)) {
610 if (!ipc.oif)
611 ipc.oif = inet->mc_index;
612 if (!saddr)
613 saddr = inet->mc_addr;
614 connected = 0;
615 }
616
617 if (connected)
618 rt = (struct rtable*)sk_dst_check(sk, 0);
619
620 if (rt == NULL) {
621 struct flowi fl = { .oif = ipc.oif,
622 .nl_u = { .ip4_u =
623 { .daddr = faddr,
624 .saddr = saddr,
625 .tos = tos } },
626 .proto = IPPROTO_UDP,
627 .uli_u = { .ports =
628 { .sport = inet->sport,
629 .dport = dport } } };
beb8d13b 630 security_sk_classify_flow(sk, &fl);
1da177e4
LT
631 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
632 if (err)
633 goto out;
634
635 err = -EACCES;
636 if ((rt->rt_flags & RTCF_BROADCAST) &&
637 !sock_flag(sk, SOCK_BROADCAST))
638 goto out;
639 if (connected)
640 sk_dst_set(sk, dst_clone(&rt->u.dst));
641 }
642
643 if (msg->msg_flags&MSG_CONFIRM)
644 goto do_confirm;
645back_from_confirm:
646
647 saddr = rt->rt_src;
648 if (!ipc.addr)
649 daddr = ipc.addr = rt->rt_dst;
650
651 lock_sock(sk);
652 if (unlikely(up->pending)) {
653 /* The socket is already corked while preparing it. */
654 /* ... which is an evident application bug. --ANK */
655 release_sock(sk);
656
64ce2073 657 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
1da177e4
LT
658 err = -EINVAL;
659 goto out;
660 }
661 /*
662 * Now cork the socket to pend data.
663 */
664 inet->cork.fl.fl4_dst = daddr;
665 inet->cork.fl.fl_ip_dport = dport;
666 inet->cork.fl.fl4_src = saddr;
667 inet->cork.fl.fl_ip_sport = inet->sport;
668 up->pending = AF_INET;
669
670do_append_data:
671 up->len += ulen;
672 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, ulen,
673 sizeof(struct udphdr), &ipc, rt,
674 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
675 if (err)
676 udp_flush_pending_frames(sk);
677 else if (!corkreq)
678 err = udp_push_pending_frames(sk, up);
679 release_sock(sk);
680
681out:
682 ip_rt_put(rt);
683 if (free)
684 kfree(ipc.opt);
685 if (!err) {
686 UDP_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS);
687 return len;
688 }
81aa646c
MB
689 /*
690 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
691 * ENOBUFS might not be good (it's not tunable per se), but otherwise
692 * we don't have a good statistic (IpOutDiscards but it can be too many
693 * things). We could add another new stat but at least for now that
694 * seems like overkill.
695 */
696 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
697 UDP_INC_STATS_USER(UDP_MIB_SNDBUFERRORS);
698 }
1da177e4
LT
699 return err;
700
701do_confirm:
702 dst_confirm(&rt->u.dst);
703 if (!(msg->msg_flags&MSG_PROBE) || len)
704 goto back_from_confirm;
705 err = 0;
706 goto out;
707}
708
709static int udp_sendpage(struct sock *sk, struct page *page, int offset,
710 size_t size, int flags)
711{
712 struct udp_sock *up = udp_sk(sk);
713 int ret;
714
715 if (!up->pending) {
716 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
717
718 /* Call udp_sendmsg to specify destination address which
719 * sendpage interface can't pass.
720 * This will succeed only when the socket is connected.
721 */
722 ret = udp_sendmsg(NULL, sk, &msg, 0);
723 if (ret < 0)
724 return ret;
725 }
726
727 lock_sock(sk);
728
729 if (unlikely(!up->pending)) {
730 release_sock(sk);
731
64ce2073 732 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n");
1da177e4
LT
733 return -EINVAL;
734 }
735
736 ret = ip_append_page(sk, page, offset, size, flags);
737 if (ret == -EOPNOTSUPP) {
738 release_sock(sk);
739 return sock_no_sendpage(sk->sk_socket, page, offset,
740 size, flags);
741 }
742 if (ret < 0) {
743 udp_flush_pending_frames(sk);
744 goto out;
745 }
746
747 up->len += size;
748 if (!(up->corkflag || (flags&MSG_MORE)))
749 ret = udp_push_pending_frames(sk, up);
750 if (!ret)
751 ret = size;
752out:
753 release_sock(sk);
754 return ret;
755}
756
757/*
758 * IOCTL requests applicable to the UDP protocol
759 */
760
761int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
762{
763 switch(cmd)
764 {
765 case SIOCOUTQ:
766 {
767 int amount = atomic_read(&sk->sk_wmem_alloc);
768 return put_user(amount, (int __user *)arg);
769 }
770
771 case SIOCINQ:
772 {
773 struct sk_buff *skb;
774 unsigned long amount;
775
776 amount = 0;
208d8984 777 spin_lock_bh(&sk->sk_receive_queue.lock);
1da177e4
LT
778 skb = skb_peek(&sk->sk_receive_queue);
779 if (skb != NULL) {
780 /*
781 * We will only return the amount
782 * of this packet since that is all
783 * that will be read.
784 */
785 amount = skb->len - sizeof(struct udphdr);
786 }
208d8984 787 spin_unlock_bh(&sk->sk_receive_queue.lock);
1da177e4
LT
788 return put_user(amount, (int __user *)arg);
789 }
790
791 default:
792 return -ENOIOCTLCMD;
793 }
794 return(0);
795}
796
797static __inline__ int __udp_checksum_complete(struct sk_buff *skb)
798{
fb286bb2 799 return __skb_checksum_complete(skb);
1da177e4
LT
800}
801
802static __inline__ int udp_checksum_complete(struct sk_buff *skb)
803{
804 return skb->ip_summed != CHECKSUM_UNNECESSARY &&
805 __udp_checksum_complete(skb);
806}
807
808/*
809 * This should be easy, if there is something there we
810 * return it, otherwise we block.
811 */
812
813static int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
814 size_t len, int noblock, int flags, int *addr_len)
815{
816 struct inet_sock *inet = inet_sk(sk);
817 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
818 struct sk_buff *skb;
819 int copied, err;
820
821 /*
822 * Check any passed addresses
823 */
824 if (addr_len)
825 *addr_len=sizeof(*sin);
826
827 if (flags & MSG_ERRQUEUE)
828 return ip_recv_error(sk, msg, len);
829
830try_again:
831 skb = skb_recv_datagram(sk, flags, noblock, &err);
832 if (!skb)
833 goto out;
834
835 copied = skb->len - sizeof(struct udphdr);
836 if (copied > len) {
837 copied = len;
838 msg->msg_flags |= MSG_TRUNC;
839 }
840
841 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
842 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
843 copied);
844 } else if (msg->msg_flags&MSG_TRUNC) {
845 if (__udp_checksum_complete(skb))
846 goto csum_copy_err;
847 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
848 copied);
849 } else {
850 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
851
852 if (err == -EINVAL)
853 goto csum_copy_err;
854 }
855
856 if (err)
857 goto out_free;
858
859 sock_recv_timestamp(msg, sk, skb);
860
861 /* Copy the address. */
862 if (sin)
863 {
864 sin->sin_family = AF_INET;
865 sin->sin_port = skb->h.uh->source;
866 sin->sin_addr.s_addr = skb->nh.iph->saddr;
867 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
868 }
869 if (inet->cmsg_flags)
870 ip_cmsg_recv(msg, skb);
871
872 err = copied;
873 if (flags & MSG_TRUNC)
874 err = skb->len - sizeof(struct udphdr);
875
876out_free:
877 skb_free_datagram(sk, skb);
878out:
879 return err;
880
881csum_copy_err:
882 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
883
3305b80c 884 skb_kill_datagram(sk, skb, flags);
1da177e4
LT
885
886 if (noblock)
887 return -EAGAIN;
888 goto try_again;
889}
890
891
892int udp_disconnect(struct sock *sk, int flags)
893{
894 struct inet_sock *inet = inet_sk(sk);
895 /*
896 * 1003.1g - break association.
897 */
898
899 sk->sk_state = TCP_CLOSE;
900 inet->daddr = 0;
901 inet->dport = 0;
902 sk->sk_bound_dev_if = 0;
903 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
904 inet_reset_saddr(sk);
905
906 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
907 sk->sk_prot->unhash(sk);
908 inet->sport = 0;
909 }
910 sk_dst_reset(sk);
911 return 0;
912}
913
914static void udp_close(struct sock *sk, long timeout)
915{
916 sk_common_release(sk);
917}
918
919/* return:
920 * 1 if the the UDP system should process it
921 * 0 if we should drop this packet
922 * -1 if it should get processed by xfrm4_rcv_encap
923 */
924static int udp_encap_rcv(struct sock * sk, struct sk_buff *skb)
925{
926#ifndef CONFIG_XFRM
927 return 1;
928#else
929 struct udp_sock *up = udp_sk(sk);
930 struct udphdr *uh = skb->h.uh;
931 struct iphdr *iph;
932 int iphlen, len;
933
934 __u8 *udpdata = (__u8 *)uh + sizeof(struct udphdr);
935 __u32 *udpdata32 = (__u32 *)udpdata;
936 __u16 encap_type = up->encap_type;
937
938 /* if we're overly short, let UDP handle it */
939 if (udpdata > skb->tail)
940 return 1;
941
942 /* if this is not encapsulated socket, then just return now */
943 if (!encap_type)
944 return 1;
945
946 len = skb->tail - udpdata;
947
948 switch (encap_type) {
949 default:
950 case UDP_ENCAP_ESPINUDP:
951 /* Check if this is a keepalive packet. If so, eat it. */
952 if (len == 1 && udpdata[0] == 0xff) {
953 return 0;
954 } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0 ) {
955 /* ESP Packet without Non-ESP header */
956 len = sizeof(struct udphdr);
957 } else
958 /* Must be an IKE packet.. pass it through */
959 return 1;
960 break;
961 case UDP_ENCAP_ESPINUDP_NON_IKE:
962 /* Check if this is a keepalive packet. If so, eat it. */
963 if (len == 1 && udpdata[0] == 0xff) {
964 return 0;
965 } else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) &&
966 udpdata32[0] == 0 && udpdata32[1] == 0) {
967
968 /* ESP Packet with Non-IKE marker */
969 len = sizeof(struct udphdr) + 2 * sizeof(u32);
970 } else
971 /* Must be an IKE packet.. pass it through */
972 return 1;
973 break;
974 }
975
976 /* At this point we are sure that this is an ESPinUDP packet,
977 * so we need to remove 'len' bytes from the packet (the UDP
978 * header and optional ESP marker bytes) and then modify the
979 * protocol to ESP, and then call into the transform receiver.
980 */
4d78b6c7
HX
981 if (skb_cloned(skb) && pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
982 return 0;
1da177e4
LT
983
984 /* Now we can update and verify the packet length... */
985 iph = skb->nh.iph;
986 iphlen = iph->ihl << 2;
987 iph->tot_len = htons(ntohs(iph->tot_len) - len);
988 if (skb->len < iphlen + len) {
989 /* packet is too small!?! */
990 return 0;
991 }
992
993 /* pull the data buffer up to the ESP header and set the
994 * transport header to point to ESP. Keep UDP on the stack
995 * for later.
996 */
997 skb->h.raw = skb_pull(skb, len);
998
999 /* modify the protocol (it's ESP!) */
1000 iph->protocol = IPPROTO_ESP;
1001
1002 /* and let the caller know to send this into the ESP processor... */
1003 return -1;
1004#endif
1005}
1006
1007/* returns:
1008 * -1: error
1009 * 0: success
1010 * >0: "udp encap" protocol resubmission
1011 *
1012 * Note that in the success and error cases, the skb is assumed to
1013 * have either been requeued or freed.
1014 */
1015static int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
1016{
1017 struct udp_sock *up = udp_sk(sk);
81aa646c 1018 int rc;
1da177e4
LT
1019
1020 /*
1021 * Charge it to the socket, dropping if the queue is full.
1022 */
1023 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
1024 kfree_skb(skb);
1025 return -1;
1026 }
b59c2701 1027 nf_reset(skb);
1da177e4
LT
1028
1029 if (up->encap_type) {
1030 /*
1031 * This is an encapsulation socket, so let's see if this is
1032 * an encapsulated packet.
1033 * If it's a keepalive packet, then just eat it.
1034 * If it's an encapsulateed packet, then pass it to the
1035 * IPsec xfrm input and return the response
1036 * appropriately. Otherwise, just fall through and
1037 * pass this up the UDP socket.
1038 */
1039 int ret;
1040
1041 ret = udp_encap_rcv(sk, skb);
1042 if (ret == 0) {
1043 /* Eat the packet .. */
1044 kfree_skb(skb);
1045 return 0;
1046 }
1047 if (ret < 0) {
1048 /* process the ESP packet */
1049 ret = xfrm4_rcv_encap(skb, up->encap_type);
1050 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS);
1051 return -ret;
1052 }
1053 /* FALLTHROUGH -- it's a UDP Packet */
1054 }
1055
1056 if (sk->sk_filter && skb->ip_summed != CHECKSUM_UNNECESSARY) {
1057 if (__udp_checksum_complete(skb)) {
1058 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1059 kfree_skb(skb);
1060 return -1;
1061 }
1062 skb->ip_summed = CHECKSUM_UNNECESSARY;
1063 }
1064
81aa646c
MB
1065 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
1066 /* Note that an ENOMEM error is charged twice */
1067 if (rc == -ENOMEM)
1068 UDP_INC_STATS_BH(UDP_MIB_RCVBUFERRORS);
1da177e4
LT
1069 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1070 kfree_skb(skb);
1071 return -1;
1072 }
1073 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS);
1074 return 0;
1075}
1076
1077/*
1078 * Multicasts and broadcasts go to each listener.
1079 *
1080 * Note: called only from the BH handler context,
1081 * so we don't need to lock the hashes.
1082 */
1083static int udp_v4_mcast_deliver(struct sk_buff *skb, struct udphdr *uh,
1084 u32 saddr, u32 daddr)
1085{
1086 struct sock *sk;
1087 int dif;
1088
1089 read_lock(&udp_hash_lock);
1090 sk = sk_head(&udp_hash[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
1091 dif = skb->dev->ifindex;
1092 sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
1093 if (sk) {
1094 struct sock *sknext = NULL;
1095
1096 do {
1097 struct sk_buff *skb1 = skb;
1098
1099 sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr,
1100 uh->source, saddr, dif);
1101 if(sknext)
1102 skb1 = skb_clone(skb, GFP_ATOMIC);
1103
1104 if(skb1) {
1105 int ret = udp_queue_rcv_skb(sk, skb1);
1106 if (ret > 0)
1107 /* we should probably re-process instead
1108 * of dropping packets here. */
1109 kfree_skb(skb1);
1110 }
1111 sk = sknext;
1112 } while(sknext);
1113 } else
1114 kfree_skb(skb);
1115 read_unlock(&udp_hash_lock);
1116 return 0;
1117}
1118
1119/* Initialize UDP checksum. If exited with zero value (success),
1120 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1121 * Otherwise, csum completion requires chacksumming packet body,
1122 * including udp header and folding it to skb->csum.
1123 */
65a45441 1124static void udp_checksum_init(struct sk_buff *skb, struct udphdr *uh,
1da177e4
LT
1125 unsigned short ulen, u32 saddr, u32 daddr)
1126{
1127 if (uh->check == 0) {
1128 skb->ip_summed = CHECKSUM_UNNECESSARY;
84fa7933 1129 } else if (skb->ip_summed == CHECKSUM_COMPLETE) {
1da177e4 1130 if (!udp_check(uh, ulen, saddr, daddr, skb->csum))
fb286bb2 1131 skb->ip_summed = CHECKSUM_UNNECESSARY;
1da177e4
LT
1132 }
1133 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
1134 skb->csum = csum_tcpudp_nofold(saddr, daddr, ulen, IPPROTO_UDP, 0);
1135 /* Probably, we should checksum udp header (it should be in cache
1136 * in any case) and data in tiny packets (< rx copybreak).
1137 */
1da177e4
LT
1138}
1139
1140/*
1141 * All we need to do is get the socket, and then do a checksum.
1142 */
1143
1144int udp_rcv(struct sk_buff *skb)
1145{
1146 struct sock *sk;
1147 struct udphdr *uh;
1148 unsigned short ulen;
1149 struct rtable *rt = (struct rtable*)skb->dst;
1150 u32 saddr = skb->nh.iph->saddr;
1151 u32 daddr = skb->nh.iph->daddr;
1152 int len = skb->len;
1153
1154 /*
1155 * Validate the packet and the UDP length.
1156 */
1157 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1158 goto no_header;
1159
1160 uh = skb->h.uh;
1161
1162 ulen = ntohs(uh->len);
1163
1164 if (ulen > len || ulen < sizeof(*uh))
1165 goto short_packet;
1166
e308e25c 1167 if (pskb_trim_rcsum(skb, ulen))
1da177e4
LT
1168 goto short_packet;
1169
65a45441 1170 udp_checksum_init(skb, uh, ulen, saddr, daddr);
1da177e4
LT
1171
1172 if(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1173 return udp_v4_mcast_deliver(skb, uh, saddr, daddr);
1174
1175 sk = udp_v4_lookup(saddr, uh->source, daddr, uh->dest, skb->dev->ifindex);
1176
1177 if (sk != NULL) {
1178 int ret = udp_queue_rcv_skb(sk, skb);
1179 sock_put(sk);
1180
1181 /* a return value > 0 means to resubmit the input, but
1182 * it it wants the return to be -protocol, or 0
1183 */
1184 if (ret > 0)
1185 return -ret;
1186 return 0;
1187 }
1188
1189 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1190 goto drop;
b59c2701 1191 nf_reset(skb);
1da177e4
LT
1192
1193 /* No socket. Drop packet silently, if checksum is wrong */
1194 if (udp_checksum_complete(skb))
1195 goto csum_error;
1196
1197 UDP_INC_STATS_BH(UDP_MIB_NOPORTS);
1198 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1199
1200 /*
1201 * Hmm. We got an UDP packet to a port to which we
1202 * don't wanna listen. Ignore it.
1203 */
1204 kfree_skb(skb);
1205 return(0);
1206
1207short_packet:
64ce2073
PM
1208 LIMIT_NETDEBUG(KERN_DEBUG "UDP: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n",
1209 NIPQUAD(saddr),
1210 ntohs(uh->source),
1211 ulen,
1212 len,
1213 NIPQUAD(daddr),
1214 ntohs(uh->dest));
1da177e4
LT
1215no_header:
1216 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1217 kfree_skb(skb);
1218 return(0);
1219
1220csum_error:
1221 /*
1222 * RFC1122: OK. Discards the bad packet silently (as far as
1223 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1224 */
64ce2073
PM
1225 LIMIT_NETDEBUG(KERN_DEBUG "UDP: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n",
1226 NIPQUAD(saddr),
1227 ntohs(uh->source),
1228 NIPQUAD(daddr),
1229 ntohs(uh->dest),
1230 ulen);
1da177e4
LT
1231drop:
1232 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1233 kfree_skb(skb);
1234 return(0);
1235}
1236
1237static int udp_destroy_sock(struct sock *sk)
1238{
1239 lock_sock(sk);
1240 udp_flush_pending_frames(sk);
1241 release_sock(sk);
1242 return 0;
1243}
1244
1245/*
1246 * Socket option code for UDP
1247 */
3fdadf7d 1248static int do_udp_setsockopt(struct sock *sk, int level, int optname,
1da177e4
LT
1249 char __user *optval, int optlen)
1250{
1251 struct udp_sock *up = udp_sk(sk);
1252 int val;
1253 int err = 0;
1254
1da177e4
LT
1255 if(optlen<sizeof(int))
1256 return -EINVAL;
1257
1258 if (get_user(val, (int __user *)optval))
1259 return -EFAULT;
1260
1261 switch(optname) {
1262 case UDP_CORK:
1263 if (val != 0) {
1264 up->corkflag = 1;
1265 } else {
1266 up->corkflag = 0;
1267 lock_sock(sk);
1268 udp_push_pending_frames(sk, up);
1269 release_sock(sk);
1270 }
1271 break;
1272
1273 case UDP_ENCAP:
1274 switch (val) {
1275 case 0:
1276 case UDP_ENCAP_ESPINUDP:
1277 case UDP_ENCAP_ESPINUDP_NON_IKE:
1278 up->encap_type = val;
1279 break;
1280 default:
1281 err = -ENOPROTOOPT;
1282 break;
1283 }
1284 break;
1285
1286 default:
1287 err = -ENOPROTOOPT;
1288 break;
1289 };
1290
1291 return err;
1292}
1293
3fdadf7d
DM
1294static int udp_setsockopt(struct sock *sk, int level, int optname,
1295 char __user *optval, int optlen)
1296{
1297 if (level != SOL_UDP)
1298 return ip_setsockopt(sk, level, optname, optval, optlen);
1299 return do_udp_setsockopt(sk, level, optname, optval, optlen);
1300}
1301
1302#ifdef CONFIG_COMPAT
1303static int compat_udp_setsockopt(struct sock *sk, int level, int optname,
543d9cfe 1304 char __user *optval, int optlen)
3fdadf7d
DM
1305{
1306 if (level != SOL_UDP)
543d9cfe 1307 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1308 return do_udp_setsockopt(sk, level, optname, optval, optlen);
1309}
1310#endif
1311
1312static int do_udp_getsockopt(struct sock *sk, int level, int optname,
1da177e4
LT
1313 char __user *optval, int __user *optlen)
1314{
1315 struct udp_sock *up = udp_sk(sk);
1316 int val, len;
1317
1da177e4
LT
1318 if(get_user(len,optlen))
1319 return -EFAULT;
1320
1321 len = min_t(unsigned int, len, sizeof(int));
1322
1323 if(len < 0)
1324 return -EINVAL;
1325
1326 switch(optname) {
1327 case UDP_CORK:
1328 val = up->corkflag;
1329 break;
1330
1331 case UDP_ENCAP:
1332 val = up->encap_type;
1333 break;
1334
1335 default:
1336 return -ENOPROTOOPT;
1337 };
1338
1339 if(put_user(len, optlen))
1340 return -EFAULT;
1341 if(copy_to_user(optval, &val,len))
1342 return -EFAULT;
1343 return 0;
1344}
1345
3fdadf7d
DM
1346static int udp_getsockopt(struct sock *sk, int level, int optname,
1347 char __user *optval, int __user *optlen)
1348{
1349 if (level != SOL_UDP)
1350 return ip_getsockopt(sk, level, optname, optval, optlen);
1351 return do_udp_getsockopt(sk, level, optname, optval, optlen);
1352}
1353
1354#ifdef CONFIG_COMPAT
1355static int compat_udp_getsockopt(struct sock *sk, int level, int optname,
543d9cfe 1356 char __user *optval, int __user *optlen)
3fdadf7d
DM
1357{
1358 if (level != SOL_UDP)
543d9cfe 1359 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM
1360 return do_udp_getsockopt(sk, level, optname, optval, optlen);
1361}
1362#endif
1da177e4
LT
1363/**
1364 * udp_poll - wait for a UDP event.
1365 * @file - file struct
1366 * @sock - socket
1367 * @wait - poll table
1368 *
1369 * This is same as datagram poll, except for the special case of
1370 * blocking sockets. If application is using a blocking fd
1371 * and a packet with checksum error is in the queue;
1372 * then it could get return from select indicating data available
1373 * but then block when reading it. Add special case code
1374 * to work around these arguably broken applications.
1375 */
1376unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1377{
1378 unsigned int mask = datagram_poll(file, sock, wait);
1379 struct sock *sk = sock->sk;
1380
1381 /* Check for false positives due to checksum errors */
1382 if ( (mask & POLLRDNORM) &&
1383 !(file->f_flags & O_NONBLOCK) &&
1384 !(sk->sk_shutdown & RCV_SHUTDOWN)){
1385 struct sk_buff_head *rcvq = &sk->sk_receive_queue;
1386 struct sk_buff *skb;
1387
208d8984 1388 spin_lock_bh(&rcvq->lock);
1da177e4
LT
1389 while ((skb = skb_peek(rcvq)) != NULL) {
1390 if (udp_checksum_complete(skb)) {
1391 UDP_INC_STATS_BH(UDP_MIB_INERRORS);
1392 __skb_unlink(skb, rcvq);
1393 kfree_skb(skb);
1394 } else {
1395 skb->ip_summed = CHECKSUM_UNNECESSARY;
1396 break;
1397 }
1398 }
208d8984 1399 spin_unlock_bh(&rcvq->lock);
1da177e4
LT
1400
1401 /* nothing to see, move along */
1402 if (skb == NULL)
1403 mask &= ~(POLLIN | POLLRDNORM);
1404 }
1405
1406 return mask;
1407
1408}
1409
1410struct proto udp_prot = {
543d9cfe
ACM
1411 .name = "UDP",
1412 .owner = THIS_MODULE,
1413 .close = udp_close,
1414 .connect = ip4_datagram_connect,
1415 .disconnect = udp_disconnect,
1416 .ioctl = udp_ioctl,
1417 .destroy = udp_destroy_sock,
1418 .setsockopt = udp_setsockopt,
1419 .getsockopt = udp_getsockopt,
1420 .sendmsg = udp_sendmsg,
1421 .recvmsg = udp_recvmsg,
1422 .sendpage = udp_sendpage,
1423 .backlog_rcv = udp_queue_rcv_skb,
1424 .hash = udp_v4_hash,
1425 .unhash = udp_v4_unhash,
1426 .get_port = udp_v4_get_port,
1427 .obj_size = sizeof(struct udp_sock),
3fdadf7d 1428#ifdef CONFIG_COMPAT
543d9cfe
ACM
1429 .compat_setsockopt = compat_udp_setsockopt,
1430 .compat_getsockopt = compat_udp_getsockopt,
3fdadf7d 1431#endif
1da177e4
LT
1432};
1433
1434/* ------------------------------------------------------------------------ */
1435#ifdef CONFIG_PROC_FS
1436
1437static struct sock *udp_get_first(struct seq_file *seq)
1438{
1439 struct sock *sk;
1440 struct udp_iter_state *state = seq->private;
1441
1442 for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) {
1443 struct hlist_node *node;
1444 sk_for_each(sk, node, &udp_hash[state->bucket]) {
1445 if (sk->sk_family == state->family)
1446 goto found;
1447 }
1448 }
1449 sk = NULL;
1450found:
1451 return sk;
1452}
1453
1454static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
1455{
1456 struct udp_iter_state *state = seq->private;
1457
1458 do {
1459 sk = sk_next(sk);
1460try_again:
1461 ;
1462 } while (sk && sk->sk_family != state->family);
1463
1464 if (!sk && ++state->bucket < UDP_HTABLE_SIZE) {
1465 sk = sk_head(&udp_hash[state->bucket]);
1466 goto try_again;
1467 }
1468 return sk;
1469}
1470
1471static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
1472{
1473 struct sock *sk = udp_get_first(seq);
1474
1475 if (sk)
1476 while(pos && (sk = udp_get_next(seq, sk)) != NULL)
1477 --pos;
1478 return pos ? NULL : sk;
1479}
1480
1481static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
1482{
1483 read_lock(&udp_hash_lock);
1484 return *pos ? udp_get_idx(seq, *pos-1) : (void *)1;
1485}
1486
1487static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1488{
1489 struct sock *sk;
1490
1491 if (v == (void *)1)
1492 sk = udp_get_idx(seq, 0);
1493 else
1494 sk = udp_get_next(seq, v);
1495
1496 ++*pos;
1497 return sk;
1498}
1499
1500static void udp_seq_stop(struct seq_file *seq, void *v)
1501{
1502 read_unlock(&udp_hash_lock);
1503}
1504
1505static int udp_seq_open(struct inode *inode, struct file *file)
1506{
1507 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
1508 struct seq_file *seq;
1509 int rc = -ENOMEM;
0da974f4 1510 struct udp_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
1da177e4
LT
1511
1512 if (!s)
1513 goto out;
1da177e4
LT
1514 s->family = afinfo->family;
1515 s->seq_ops.start = udp_seq_start;
1516 s->seq_ops.next = udp_seq_next;
1517 s->seq_ops.show = afinfo->seq_show;
1518 s->seq_ops.stop = udp_seq_stop;
1519
1520 rc = seq_open(file, &s->seq_ops);
1521 if (rc)
1522 goto out_kfree;
1523
1524 seq = file->private_data;
1525 seq->private = s;
1526out:
1527 return rc;
1528out_kfree:
1529 kfree(s);
1530 goto out;
1531}
1532
1533/* ------------------------------------------------------------------------ */
1534int udp_proc_register(struct udp_seq_afinfo *afinfo)
1535{
1536 struct proc_dir_entry *p;
1537 int rc = 0;
1538
1539 if (!afinfo)
1540 return -EINVAL;
1541 afinfo->seq_fops->owner = afinfo->owner;
1542 afinfo->seq_fops->open = udp_seq_open;
1543 afinfo->seq_fops->read = seq_read;
1544 afinfo->seq_fops->llseek = seq_lseek;
1545 afinfo->seq_fops->release = seq_release_private;
1546
1547 p = proc_net_fops_create(afinfo->name, S_IRUGO, afinfo->seq_fops);
1548 if (p)
1549 p->data = afinfo;
1550 else
1551 rc = -ENOMEM;
1552 return rc;
1553}
1554
1555void udp_proc_unregister(struct udp_seq_afinfo *afinfo)
1556{
1557 if (!afinfo)
1558 return;
1559 proc_net_remove(afinfo->name);
1560 memset(afinfo->seq_fops, 0, sizeof(*afinfo->seq_fops));
1561}
1562
1563/* ------------------------------------------------------------------------ */
1564static void udp4_format_sock(struct sock *sp, char *tmpbuf, int bucket)
1565{
1566 struct inet_sock *inet = inet_sk(sp);
1567 unsigned int dest = inet->daddr;
1568 unsigned int src = inet->rcv_saddr;
1569 __u16 destp = ntohs(inet->dport);
1570 __u16 srcp = ntohs(inet->sport);
1571
1572 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
1573 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
1574 bucket, src, srcp, dest, destp, sp->sk_state,
1575 atomic_read(&sp->sk_wmem_alloc),
1576 atomic_read(&sp->sk_rmem_alloc),
1577 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
1578 atomic_read(&sp->sk_refcnt), sp);
1579}
1580
1581static int udp4_seq_show(struct seq_file *seq, void *v)
1582{
1583 if (v == SEQ_START_TOKEN)
1584 seq_printf(seq, "%-127s\n",
1585 " sl local_address rem_address st tx_queue "
1586 "rx_queue tr tm->when retrnsmt uid timeout "
1587 "inode");
1588 else {
1589 char tmpbuf[129];
1590 struct udp_iter_state *state = seq->private;
1591
1592 udp4_format_sock(v, tmpbuf, state->bucket);
1593 seq_printf(seq, "%-127s\n", tmpbuf);
1594 }
1595 return 0;
1596}
1597
1598/* ------------------------------------------------------------------------ */
1599static struct file_operations udp4_seq_fops;
1600static struct udp_seq_afinfo udp4_seq_afinfo = {
1601 .owner = THIS_MODULE,
1602 .name = "udp",
1603 .family = AF_INET,
1604 .seq_show = udp4_seq_show,
1605 .seq_fops = &udp4_seq_fops,
1606};
1607
1608int __init udp4_proc_init(void)
1609{
1610 return udp_proc_register(&udp4_seq_afinfo);
1611}
1612
1613void udp4_proc_exit(void)
1614{
1615 udp_proc_unregister(&udp4_seq_afinfo);
1616}
1617#endif /* CONFIG_PROC_FS */
1618
1619EXPORT_SYMBOL(udp_disconnect);
1620EXPORT_SYMBOL(udp_hash);
1621EXPORT_SYMBOL(udp_hash_lock);
1622EXPORT_SYMBOL(udp_ioctl);
25030a7f 1623EXPORT_SYMBOL(udp_get_port);
1da177e4
LT
1624EXPORT_SYMBOL(udp_prot);
1625EXPORT_SYMBOL(udp_sendmsg);
1626EXPORT_SYMBOL(udp_poll);
1627
1628#ifdef CONFIG_PROC_FS
1629EXPORT_SYMBOL(udp_proc_register);
1630EXPORT_SYMBOL(udp_proc_unregister);
1631#endif