?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ata_timing: ensure t->cycle is always correct
[linux-2.6/verdex.git] / net / ipv4 / udp.c
blobcec0f2cc49b7357975e8b254df25709cfc261d90
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 *
10 * Authors: Ross Biro
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>
89 #include <linux/igmp.h>
90 #include <linux/in.h>
91 #include <linux/errno.h>
92 #include <linux/timer.h>
93 #include <linux/mm.h>
94 #include <linux/inet.h>
95 #include <linux/netdevice.h>
96 #include <net/tcp_states.h>
97 #include <linux/skbuff.h>
98 #include <linux/proc_fs.h>
99 #include <linux/seq_file.h>
100 #include <net/icmp.h>
101 #include <net/route.h>
102 #include <net/checksum.h>
103 #include <net/xfrm.h>
104 #include "udp_impl.h"
105
106 /*
107 * Snmp MIB for the UDP layer
108 */
109
110 DEFINE_SNMP_STAT(struct udp_mib, udp_statistics) __read_mostly;
111
112 struct hlist_head udp_hash[UDP_HTABLE_SIZE];
113 DEFINE_RWLOCK(udp_hash_lock);
114
115 static int udp_port_rover;
116
117 static inline int __udp_lib_lport_inuse(__u16 num, struct hlist_head udptable[])
118 {
119 struct sock *sk;
120 struct hlist_node *node;
121
122 sk_for_each(sk, node, &udptable[num & (UDP_HTABLE_SIZE - 1)])
123 if (sk->sk_hash == num)
124 return 1;
125 return 0;
126 }
127
128 /**
129 * __udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
130 *
131 * @sk: socket struct in question
132 * @snum: port number to look up
133 * @udptable: hash list table, must be of UDP_HTABLE_SIZE
134 * @port_rover: pointer to record of last unallocated port
135 * @saddr_comp: AF-dependent comparison of bound local IP addresses
136 */
137 int __udp_lib_get_port(struct sock *sk, unsigned short snum,
138 struct hlist_head udptable[], int *port_rover,
139 int (*saddr_comp)(const struct sock *sk1,
140 const struct sock *sk2 ) )
141 {
142 struct hlist_node *node;
143 struct hlist_head *head;
144 struct sock *sk2;
145 int error = 1;
146
147 write_lock_bh(&udp_hash_lock);
148 if (snum == 0) {
149 int best_size_so_far, best, result, i;
150
151 if (*port_rover > sysctl_local_port_range[1] ||
152 *port_rover < sysctl_local_port_range[0])
153 *port_rover = sysctl_local_port_range[0];
154 best_size_so_far = 32767;
155 best = result = *port_rover;
156 for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) {
157 int size;
158
159 head = &udptable[result & (UDP_HTABLE_SIZE - 1)];
160 if (hlist_empty(head)) {
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 goto gotit;
166 }
167 size = 0;
168 sk_for_each(sk2, node, head) {
169 if (++size >= best_size_so_far)
170 goto next;
171 }
172 best_size_so_far = size;
173 best = result;
174 next:
175 ;
176 }
177 result = best;
178 for (i = 0; i < (1 << 16) / UDP_HTABLE_SIZE;
179 i++, result += UDP_HTABLE_SIZE) {
180 if (result > sysctl_local_port_range[1])
181 result = sysctl_local_port_range[0]
182 + ((result - sysctl_local_port_range[0]) &
183 (UDP_HTABLE_SIZE - 1));
184 if (! __udp_lib_lport_inuse(result, udptable))
185 break;
186 }
187 if (i >= (1 << 16) / UDP_HTABLE_SIZE)
188 goto fail;
189 gotit:
190 *port_rover = snum = result;
191 } else {
192 head = &udptable[snum & (UDP_HTABLE_SIZE - 1)];
193
194 sk_for_each(sk2, node, head)
195 if (sk2->sk_hash == snum &&
196 sk2 != sk &&
197 (!sk2->sk_reuse || !sk->sk_reuse) &&
198 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if
199 || sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
200 (*saddr_comp)(sk, sk2) )
201 goto fail;
202 }
203 inet_sk(sk)->num = snum;
204 sk->sk_hash = snum;
205 if (sk_unhashed(sk)) {
206 head = &udptable[snum & (UDP_HTABLE_SIZE - 1)];
207 sk_add_node(sk, head);
208 sock_prot_inc_use(sk->sk_prot);
209 }
210 error = 0;
211 fail:
212 write_unlock_bh(&udp_hash_lock);
213 return error;
214 }
215
216 int udp_get_port(struct sock *sk, unsigned short snum,
217 int (*scmp)(const struct sock *, const struct sock *))
218 {
219 return __udp_lib_get_port(sk, snum, udp_hash, &udp_port_rover, scmp);
220 }
221
222 int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
223 {
224 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
225
226 return ( !ipv6_only_sock(sk2) &&
227 (!inet1->rcv_saddr || !inet2->rcv_saddr ||
228 inet1->rcv_saddr == inet2->rcv_saddr ));
229 }
230
231 static inline int udp_v4_get_port(struct sock *sk, unsigned short snum)
232 {
233 return udp_get_port(sk, snum, ipv4_rcv_saddr_equal);
234 }
235
236 /* UDP is nearly always wildcards out the wazoo, it makes no sense to try
237 * harder than this. -DaveM
238 */
239 static struct sock *__udp4_lib_lookup(__be32 saddr, __be16 sport,
240 __be32 daddr, __be16 dport,
241 int dif, struct hlist_head udptable[])
242 {
243 struct sock *sk, *result = NULL;
244 struct hlist_node *node;
245 unsigned short hnum = ntohs(dport);
246 int badness = -1;
247
248 read_lock(&udp_hash_lock);
249 sk_for_each(sk, node, &udptable[hnum & (UDP_HTABLE_SIZE - 1)]) {
250 struct inet_sock *inet = inet_sk(sk);
251
252 if (sk->sk_hash == hnum && !ipv6_only_sock(sk)) {
253 int score = (sk->sk_family == PF_INET ? 1 : 0);
254 if (inet->rcv_saddr) {
255 if (inet->rcv_saddr != daddr)
256 continue;
257 score+=2;
258 }
259 if (inet->daddr) {
260 if (inet->daddr != saddr)
261 continue;
262 score+=2;
263 }
264 if (inet->dport) {
265 if (inet->dport != sport)
266 continue;
267 score+=2;
268 }
269 if (sk->sk_bound_dev_if) {
270 if (sk->sk_bound_dev_if != dif)
271 continue;
272 score+=2;
273 }
274 if (score == 9) {
275 result = sk;
276 break;
277 } else if (score > badness) {
278 result = sk;
279 badness = score;
280 }
281 }
282 }
283 if (result)
284 sock_hold(result);
285 read_unlock(&udp_hash_lock);
286 return result;
287 }
288
289 static inline struct sock *udp_v4_mcast_next(struct sock *sk,
290 __be16 loc_port, __be32 loc_addr,
291 __be16 rmt_port, __be32 rmt_addr,
292 int dif)
293 {
294 struct hlist_node *node;
295 struct sock *s = sk;
296 unsigned short hnum = ntohs(loc_port);
297
298 sk_for_each_from(s, node) {
299 struct inet_sock *inet = inet_sk(s);
300
301 if (s->sk_hash != hnum ||
302 (inet->daddr && inet->daddr != rmt_addr) ||
303 (inet->dport != rmt_port && inet->dport) ||
304 (inet->rcv_saddr && inet->rcv_saddr != loc_addr) ||
305 ipv6_only_sock(s) ||
306 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
307 continue;
308 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
309 continue;
310 goto found;
311 }
312 s = NULL;
313 found:
314 return s;
315 }
316
317 /*
318 * This routine is called by the ICMP module when it gets some
319 * sort of error condition. If err < 0 then the socket should
320 * be closed and the error returned to the user. If err > 0
321 * it's just the icmp type << 8 | icmp code.
322 * Header points to the ip header of the error packet. We move
323 * on past this. Then (as it used to claim before adjustment)
324 * header points to the first 8 bytes of the udp header. We need
325 * to find the appropriate port.
326 */
327
328 void __udp4_lib_err(struct sk_buff *skb, u32 info, struct hlist_head udptable[])
329 {
330 struct inet_sock *inet;
331 struct iphdr *iph = (struct iphdr*)skb->data;
332 struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2));
333 const int type = icmp_hdr(skb)->type;
334 const int code = icmp_hdr(skb)->code;
335 struct sock *sk;
336 int harderr;
337 int err;
338
339 sk = __udp4_lib_lookup(iph->daddr, uh->dest, iph->saddr, uh->source,
340 skb->dev->ifindex, udptable );
341 if (sk == NULL) {
342 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
343 return; /* No socket for error */
344 }
345
346 err = 0;
347 harderr = 0;
348 inet = inet_sk(sk);
349
350 switch (type) {
351 default:
352 case ICMP_TIME_EXCEEDED:
353 err = EHOSTUNREACH;
354 break;
355 case ICMP_SOURCE_QUENCH:
356 goto out;
357 case ICMP_PARAMETERPROB:
358 err = EPROTO;
359 harderr = 1;
360 break;
361 case ICMP_DEST_UNREACH:
362 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
363 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
364 err = EMSGSIZE;
365 harderr = 1;
366 break;
367 }
368 goto out;
369 }
370 err = EHOSTUNREACH;
371 if (code <= NR_ICMP_UNREACH) {
372 harderr = icmp_err_convert[code].fatal;
373 err = icmp_err_convert[code].errno;
374 }
375 break;
376 }
377
378 /*
379 * RFC1122: OK. Passes ICMP errors back to application, as per
380 * 4.1.3.3.
381 */
382 if (!inet->recverr) {
383 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
384 goto out;
385 } else {
386 ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1));
387 }
388 sk->sk_err = err;
389 sk->sk_error_report(sk);
390 out:
391 sock_put(sk);
392 }
393
394 void udp_err(struct sk_buff *skb, u32 info)
395 {
396 return __udp4_lib_err(skb, info, udp_hash);
397 }
398
399 /*
400 * Throw away all pending data and cancel the corking. Socket is locked.
401 */
402 static void udp_flush_pending_frames(struct sock *sk)
403 {
404 struct udp_sock *up = udp_sk(sk);
405
406 if (up->pending) {
407 up->len = 0;
408 up->pending = 0;
409 ip_flush_pending_frames(sk);
410 }
411 }
412
413 /**
414 * udp4_hwcsum_outgoing - handle outgoing HW checksumming
415 * @sk: socket we are sending on
416 * @skb: sk_buff containing the filled-in UDP header
417 * (checksum field must be zeroed out)
418 */
419 static void udp4_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
420 __be32 src, __be32 dst, int len )
421 {
422 unsigned int offset;
423 struct udphdr *uh = udp_hdr(skb);
424 __wsum csum = 0;
425
426 if (skb_queue_len(&sk->sk_write_queue) == 1) {
427 /*
428 * Only one fragment on the socket.
429 */
430 skb->csum_start = skb_transport_header(skb) - skb->head;
431 skb->csum_offset = offsetof(struct udphdr, check);
432 uh->check = ~csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, 0);
433 } else {
434 /*
435 * HW-checksum won't work as there are two or more
436 * fragments on the socket so that all csums of sk_buffs
437 * should be together
438 */
439 offset = skb_transport_offset(skb);
440 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
441
442 skb->ip_summed = CHECKSUM_NONE;
443
444 skb_queue_walk(&sk->sk_write_queue, skb) {
445 csum = csum_add(csum, skb->csum);
446 }
447
448 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
449 if (uh->check == 0)
450 uh->check = CSUM_MANGLED_0;
451 }
452 }
453
454 /*
455 * Push out all pending data as one UDP datagram. Socket is locked.
456 */
457 static int udp_push_pending_frames(struct sock *sk)
458 {
459 struct udp_sock *up = udp_sk(sk);
460 struct inet_sock *inet = inet_sk(sk);
461 struct flowi *fl = &inet->cork.fl;
462 struct sk_buff *skb;
463 struct udphdr *uh;
464 int err = 0;
465 __wsum csum = 0;
466
467 /* Grab the skbuff where UDP header space exists. */
468 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
469 goto out;
470
471 /*
472 * Create a UDP header
473 */
474 uh = udp_hdr(skb);
475 uh->source = fl->fl_ip_sport;
476 uh->dest = fl->fl_ip_dport;
477 uh->len = htons(up->len);
478 uh->check = 0;
479
480 if (up->pcflag) /* UDP-Lite */
481 csum = udplite_csum_outgoing(sk, skb);
482
483 else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */
484
485 skb->ip_summed = CHECKSUM_NONE;
486 goto send;
487
488 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
489
490 udp4_hwcsum_outgoing(sk, skb, fl->fl4_src,fl->fl4_dst, up->len);
491 goto send;
492
493 } else /* `normal' UDP */
494 csum = udp_csum_outgoing(sk, skb);
495
496 /* add protocol-dependent pseudo-header */
497 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, up->len,
498 sk->sk_protocol, csum );
499 if (uh->check == 0)
500 uh->check = CSUM_MANGLED_0;
501
502 send:
503 err = ip_push_pending_frames(sk);
504 out:
505 up->len = 0;
506 up->pending = 0;
507 return err;
508 }
509
510 int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
511 size_t len)
512 {
513 struct inet_sock *inet = inet_sk(sk);
514 struct udp_sock *up = udp_sk(sk);
515 int ulen = len;
516 struct ipcm_cookie ipc;
517 struct rtable *rt = NULL;
518 int free = 0;
519 int connected = 0;
520 __be32 daddr, faddr, saddr;
521 __be16 dport;
522 u8 tos;
523 int err, is_udplite = up->pcflag;
524 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
525 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
526
527 if (len > 0xFFFF)
528 return -EMSGSIZE;
529
530 /*
531 * Check the flags.
532 */
533
534 if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */
535 return -EOPNOTSUPP;
536
537 ipc.opt = NULL;
538
539 if (up->pending) {
540 /*
541 * There are pending frames.
542 * The socket lock must be held while it's corked.
543 */
544 lock_sock(sk);
545 if (likely(up->pending)) {
546 if (unlikely(up->pending != AF_INET)) {
547 release_sock(sk);
548 return -EINVAL;
549 }
550 goto do_append_data;
551 }
552 release_sock(sk);
553 }
554 ulen += sizeof(struct udphdr);
555
556 /*
557 * Get and verify the address.
558 */
559 if (msg->msg_name) {
560 struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name;
561 if (msg->msg_namelen < sizeof(*usin))
562 return -EINVAL;
563 if (usin->sin_family != AF_INET) {
564 if (usin->sin_family != AF_UNSPEC)
565 return -EAFNOSUPPORT;
566 }
567
568 daddr = usin->sin_addr.s_addr;
569 dport = usin->sin_port;
570 if (dport == 0)
571 return -EINVAL;
572 } else {
573 if (sk->sk_state != TCP_ESTABLISHED)
574 return -EDESTADDRREQ;
575 daddr = inet->daddr;
576 dport = inet->dport;
577 /* Open fast path for connected socket.
578 Route will not be used, if at least one option is set.
579 */
580 connected = 1;
581 }
582 ipc.addr = inet->saddr;
583
584 ipc.oif = sk->sk_bound_dev_if;
585 if (msg->msg_controllen) {
586 err = ip_cmsg_send(msg, &ipc);
587 if (err)
588 return err;
589 if (ipc.opt)
590 free = 1;
591 connected = 0;
592 }
593 if (!ipc.opt)
594 ipc.opt = inet->opt;
595
596 saddr = ipc.addr;
597 ipc.addr = faddr = daddr;
598
599 if (ipc.opt && ipc.opt->srr) {
600 if (!daddr)
601 return -EINVAL;
602 faddr = ipc.opt->faddr;
603 connected = 0;
604 }
605 tos = RT_TOS(inet->tos);
606 if (sock_flag(sk, SOCK_LOCALROUTE) ||
607 (msg->msg_flags & MSG_DONTROUTE) ||
608 (ipc.opt && ipc.opt->is_strictroute)) {
609 tos |= RTO_ONLINK;
610 connected = 0;
611 }
612
613 if (MULTICAST(daddr)) {
614 if (!ipc.oif)
615 ipc.oif = inet->mc_index;
616 if (!saddr)
617 saddr = inet->mc_addr;
618 connected = 0;
619 }
620
621 if (connected)
622 rt = (struct rtable*)sk_dst_check(sk, 0);
623
624 if (rt == NULL) {
625 struct flowi fl = { .oif = ipc.oif,
626 .nl_u = { .ip4_u =
627 { .daddr = faddr,
628 .saddr = saddr,
629 .tos = tos } },
630 .proto = sk->sk_protocol,
631 .uli_u = { .ports =
632 { .sport = inet->sport,
633 .dport = dport } } };
634 security_sk_classify_flow(sk, &fl);
635 err = ip_route_output_flow(&rt, &fl, sk, 1);
636 if (err)
637 goto out;
638
639 err = -EACCES;
640 if ((rt->rt_flags & RTCF_BROADCAST) &&
641 !sock_flag(sk, SOCK_BROADCAST))
642 goto out;
643 if (connected)
644 sk_dst_set(sk, dst_clone(&rt->u.dst));
645 }
646
647 if (msg->msg_flags&MSG_CONFIRM)
648 goto do_confirm;
649 back_from_confirm:
650
651 saddr = rt->rt_src;
652 if (!ipc.addr)
653 daddr = ipc.addr = rt->rt_dst;
654
655 lock_sock(sk);
656 if (unlikely(up->pending)) {
657 /* The socket is already corked while preparing it. */
658 /* ... which is an evident application bug. --ANK */
659 release_sock(sk);
660
661 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
662 err = -EINVAL;
663 goto out;
664 }
665 /*
666 * Now cork the socket to pend data.
667 */
668 inet->cork.fl.fl4_dst = daddr;
669 inet->cork.fl.fl_ip_dport = dport;
670 inet->cork.fl.fl4_src = saddr;
671 inet->cork.fl.fl_ip_sport = inet->sport;
672 up->pending = AF_INET;
673
674 do_append_data:
675 up->len += ulen;
676 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
677 err = ip_append_data(sk, getfrag, msg->msg_iov, ulen,
678 sizeof(struct udphdr), &ipc, rt,
679 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
680 if (err)
681 udp_flush_pending_frames(sk);
682 else if (!corkreq)
683 err = udp_push_pending_frames(sk);
684 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
685 up->pending = 0;
686 release_sock(sk);
687
688 out:
689 ip_rt_put(rt);
690 if (free)
691 kfree(ipc.opt);
692 if (!err) {
693 UDP_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS, is_udplite);
694 return len;
695 }
696 /*
697 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
698 * ENOBUFS might not be good (it's not tunable per se), but otherwise
699 * we don't have a good statistic (IpOutDiscards but it can be too many
700 * things). We could add another new stat but at least for now that
701 * seems like overkill.
702 */
703 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
704 UDP_INC_STATS_USER(UDP_MIB_SNDBUFERRORS, is_udplite);
705 }
706 return err;
707
708 do_confirm:
709 dst_confirm(&rt->u.dst);
710 if (!(msg->msg_flags&MSG_PROBE) || len)
711 goto back_from_confirm;
712 err = 0;
713 goto out;
714 }
715
716 int udp_sendpage(struct sock *sk, struct page *page, int offset,
717 size_t size, int flags)
718 {
719 struct udp_sock *up = udp_sk(sk);
720 int ret;
721
722 if (!up->pending) {
723 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
724
725 /* Call udp_sendmsg to specify destination address which
726 * sendpage interface can't pass.
727 * This will succeed only when the socket is connected.
728 */
729 ret = udp_sendmsg(NULL, sk, &msg, 0);
730 if (ret < 0)
731 return ret;
732 }
733
734 lock_sock(sk);
735
736 if (unlikely(!up->pending)) {
737 release_sock(sk);
738
739 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n");
740 return -EINVAL;
741 }
742
743 ret = ip_append_page(sk, page, offset, size, flags);
744 if (ret == -EOPNOTSUPP) {
745 release_sock(sk);
746 return sock_no_sendpage(sk->sk_socket, page, offset,
747 size, flags);
748 }
749 if (ret < 0) {
750 udp_flush_pending_frames(sk);
751 goto out;
752 }
753
754 up->len += size;
755 if (!(up->corkflag || (flags&MSG_MORE)))
756 ret = udp_push_pending_frames(sk);
757 if (!ret)
758 ret = size;
759 out:
760 release_sock(sk);
761 return ret;
762 }
763
764 /*
765 * IOCTL requests applicable to the UDP protocol
766 */
767
768 int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
769 {
770 switch (cmd) {
771 case SIOCOUTQ:
772 {
773 int amount = atomic_read(&sk->sk_wmem_alloc);
774 return put_user(amount, (int __user *)arg);
775 }
776
777 case SIOCINQ:
778 {
779 struct sk_buff *skb;
780 unsigned long amount;
781
782 amount = 0;
783 spin_lock_bh(&sk->sk_receive_queue.lock);
784 skb = skb_peek(&sk->sk_receive_queue);
785 if (skb != NULL) {
786 /*
787 * We will only return the amount
788 * of this packet since that is all
789 * that will be read.
790 */
791 amount = skb->len - sizeof(struct udphdr);
792 }
793 spin_unlock_bh(&sk->sk_receive_queue.lock);
794 return put_user(amount, (int __user *)arg);
795 }
796
797 default:
798 return -ENOIOCTLCMD;
799 }
800
801 return 0;
802 }
803
804 /*
805 * This should be easy, if there is something there we
806 * return it, otherwise we block.
807 */
808
809 int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
810 size_t len, int noblock, int flags, int *addr_len)
811 {
812 struct inet_sock *inet = inet_sk(sk);
813 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
814 struct sk_buff *skb;
815 unsigned int ulen, copied;
816 int err;
817 int is_udplite = IS_UDPLITE(sk);
818
819 /*
820 * Check any passed addresses
821 */
822 if (addr_len)
823 *addr_len=sizeof(*sin);
824
825 if (flags & MSG_ERRQUEUE)
826 return ip_recv_error(sk, msg, len);
827
828 try_again:
829 skb = skb_recv_datagram(sk, flags, noblock, &err);
830 if (!skb)
831 goto out;
832
833 ulen = skb->len - sizeof(struct udphdr);
834 copied = len;
835 if (copied > ulen)
836 copied = ulen;
837 else if (copied < ulen)
838 msg->msg_flags |= MSG_TRUNC;
839
840 /*
841 * If checksum is needed at all, try to do it while copying the
842 * data. If the data is truncated, or if we only want a partial
843 * coverage checksum (UDP-Lite), do it before the copy.
844 */
845
846 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
847 if (udp_lib_checksum_complete(skb))
848 goto csum_copy_err;
849 }
850
851 if (skb_csum_unnecessary(skb))
852 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
853 msg->msg_iov, copied );
854 else {
855 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
856
857 if (err == -EINVAL)
858 goto csum_copy_err;
859 }
860
861 if (err)
862 goto out_free;
863
864 sock_recv_timestamp(msg, sk, skb);
865
866 /* Copy the address. */
867 if (sin)
868 {
869 sin->sin_family = AF_INET;
870 sin->sin_port = udp_hdr(skb)->source;
871 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
872 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
873 }
874 if (inet->cmsg_flags)
875 ip_cmsg_recv(msg, skb);
876
877 err = copied;
878 if (flags & MSG_TRUNC)
879 err = ulen;
880
881 out_free:
882 skb_free_datagram(sk, skb);
883 out:
884 return err;
885
886 csum_copy_err:
887 UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_udplite);
888
889 skb_kill_datagram(sk, skb, flags);
890
891 if (noblock)
892 return -EAGAIN;
893 goto try_again;
894 }
895
896
897 int udp_disconnect(struct sock *sk, int flags)
898 {
899 struct inet_sock *inet = inet_sk(sk);
900 /*
901 * 1003.1g - break association.
902 */
903
904 sk->sk_state = TCP_CLOSE;
905 inet->daddr = 0;
906 inet->dport = 0;
907 sk->sk_bound_dev_if = 0;
908 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
909 inet_reset_saddr(sk);
910
911 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
912 sk->sk_prot->unhash(sk);
913 inet->sport = 0;
914 }
915 sk_dst_reset(sk);
916 return 0;
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 */
924 static 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;
931 struct iphdr *iph;
932 int iphlen, len;
933
934 __u8 *udpdata;
935 __be32 *udpdata32;
936 __u16 encap_type = up->encap_type;
937
938 /* if we're overly short, let UDP handle it */
939 len = skb->len - sizeof(struct udphdr);
940 if (len <= 0)
941 return 1;
942
943 /* if this is not encapsulated socket, then just return now */
944 if (!encap_type)
945 return 1;
946
947 /* If this is a paged skb, make sure we pull up
948 * whatever data we need to look at. */
949 if (!pskb_may_pull(skb, sizeof(struct udphdr) + min(len, 8)))
950 return 1;
951
952 /* Now we can get the pointers */
953 uh = udp_hdr(skb);
954 udpdata = (__u8 *)uh + sizeof(struct udphdr);
955 udpdata32 = (__be32 *)udpdata;
956
957 switch (encap_type) {
958 default:
959 case UDP_ENCAP_ESPINUDP:
960 /* Check if this is a keepalive packet. If so, eat it. */
961 if (len == 1 && udpdata[0] == 0xff) {
962 return 0;
963 } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0) {
964 /* ESP Packet without Non-ESP header */
965 len = sizeof(struct udphdr);
966 } else
967 /* Must be an IKE packet.. pass it through */
968 return 1;
969 break;
970 case UDP_ENCAP_ESPINUDP_NON_IKE:
971 /* Check if this is a keepalive packet. If so, eat it. */
972 if (len == 1 && udpdata[0] == 0xff) {
973 return 0;
974 } else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) &&
975 udpdata32[0] == 0 && udpdata32[1] == 0) {
976
977 /* ESP Packet with Non-IKE marker */
978 len = sizeof(struct udphdr) + 2 * sizeof(u32);
979 } else
980 /* Must be an IKE packet.. pass it through */
981 return 1;
982 break;
983 }
984
985 /* At this point we are sure that this is an ESPinUDP packet,
986 * so we need to remove 'len' bytes from the packet (the UDP
987 * header and optional ESP marker bytes) and then modify the
988 * protocol to ESP, and then call into the transform receiver.
989 */
990 if (skb_cloned(skb) && pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
991 return 0;
992
993 /* Now we can update and verify the packet length... */
994 iph = ip_hdr(skb);
995 iphlen = iph->ihl << 2;
996 iph->tot_len = htons(ntohs(iph->tot_len) - len);
997 if (skb->len < iphlen + len) {
998 /* packet is too small!?! */
999 return 0;
1000 }
1001
1002 /* pull the data buffer up to the ESP header and set the
1003 * transport header to point to ESP. Keep UDP on the stack
1004 * for later.
1005 */
1006 __skb_pull(skb, len);
1007 skb_reset_transport_header(skb);
1008
1009 /* modify the protocol (it's ESP!) */
1010 iph->protocol = IPPROTO_ESP;
1011
1012 /* and let the caller know to send this into the ESP processor... */
1013 return -1;
1014 #endif
1015 }
1016
1017 /* returns:
1018 * -1: error
1019 * 0: success
1020 * >0: "udp encap" protocol resubmission
1021 *
1022 * Note that in the success and error cases, the skb is assumed to
1023 * have either been requeued or freed.
1024 */
1025 int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
1026 {
1027 struct udp_sock *up = udp_sk(sk);
1028 int rc;
1029
1030 /*
1031 * Charge it to the socket, dropping if the queue is full.
1032 */
1033 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1034 goto drop;
1035 nf_reset(skb);
1036
1037 if (up->encap_type) {
1038 /*
1039 * This is an encapsulation socket, so let's see if this is
1040 * an encapsulated packet.
1041 * If it's a keepalive packet, then just eat it.
1042 * If it's an encapsulateed packet, then pass it to the
1043 * IPsec xfrm input and return the response
1044 * appropriately. Otherwise, just fall through and
1045 * pass this up the UDP socket.
1046 */
1047 int ret;
1048
1049 ret = udp_encap_rcv(sk, skb);
1050 if (ret == 0) {
1051 /* Eat the packet .. */
1052 kfree_skb(skb);
1053 return 0;
1054 }
1055 if (ret < 0) {
1056 /* process the ESP packet */
1057 ret = xfrm4_rcv_encap(skb, up->encap_type);
1058 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS, up->pcflag);
1059 return -ret;
1060 }
1061 /* FALLTHROUGH -- it's a UDP Packet */
1062 }
1063
1064 /*
1065 * UDP-Lite specific tests, ignored on UDP sockets
1066 */
1067 if ((up->pcflag & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1068
1069 /*
1070 * MIB statistics other than incrementing the error count are
1071 * disabled for the following two types of errors: these depend
1072 * on the application settings, not on the functioning of the
1073 * protocol stack as such.
1074 *
1075 * RFC 3828 here recommends (sec 3.3): "There should also be a
1076 * way ... to ... at least let the receiving application block
1077 * delivery of packets with coverage values less than a value
1078 * provided by the application."
1079 */
1080 if (up->pcrlen == 0) { /* full coverage was set */
1081 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE: partial coverage "
1082 "%d while full coverage %d requested\n",
1083 UDP_SKB_CB(skb)->cscov, skb->len);
1084 goto drop;
1085 }
1086 /* The next case involves violating the min. coverage requested
1087 * by the receiver. This is subtle: if receiver wants x and x is
1088 * greater than the buffersize/MTU then receiver will complain
1089 * that it wants x while sender emits packets of smaller size y.
1090 * Therefore the above ...()->partial_cov statement is essential.
1091 */
1092 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
1093 LIMIT_NETDEBUG(KERN_WARNING
1094 "UDPLITE: coverage %d too small, need min %d\n",
1095 UDP_SKB_CB(skb)->cscov, up->pcrlen);
1096 goto drop;
1097 }
1098 }
1099
1100 if (sk->sk_filter) {
1101 if (udp_lib_checksum_complete(skb))
1102 goto drop;
1103 }
1104
1105 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
1106 /* Note that an ENOMEM error is charged twice */
1107 if (rc == -ENOMEM)
1108 UDP_INC_STATS_BH(UDP_MIB_RCVBUFERRORS, up->pcflag);
1109 goto drop;
1110 }
1111
1112 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS, up->pcflag);
1113 return 0;
1114
1115 drop:
1116 UDP_INC_STATS_BH(UDP_MIB_INERRORS, up->pcflag);
1117 kfree_skb(skb);
1118 return -1;
1119 }
1120
1121 /*
1122 * Multicasts and broadcasts go to each listener.
1123 *
1124 * Note: called only from the BH handler context,
1125 * so we don't need to lock the hashes.
1126 */
1127 static int __udp4_lib_mcast_deliver(struct sk_buff *skb,
1128 struct udphdr *uh,
1129 __be32 saddr, __be32 daddr,
1130 struct hlist_head udptable[])
1131 {
1132 struct sock *sk;
1133 int dif;
1134
1135 read_lock(&udp_hash_lock);
1136 sk = sk_head(&udptable[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
1137 dif = skb->dev->ifindex;
1138 sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
1139 if (sk) {
1140 struct sock *sknext = NULL;
1141
1142 do {
1143 struct sk_buff *skb1 = skb;
1144
1145 sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr,
1146 uh->source, saddr, dif);
1147 if (sknext)
1148 skb1 = skb_clone(skb, GFP_ATOMIC);
1149
1150 if (skb1) {
1151 int ret = udp_queue_rcv_skb(sk, skb1);
1152 if (ret > 0)
1153 /* we should probably re-process instead
1154 * of dropping packets here. */
1155 kfree_skb(skb1);
1156 }
1157 sk = sknext;
1158 } while (sknext);
1159 } else
1160 kfree_skb(skb);
1161 read_unlock(&udp_hash_lock);
1162 return 0;
1163 }
1164
1165 /* Initialize UDP checksum. If exited with zero value (success),
1166 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1167 * Otherwise, csum completion requires chacksumming packet body,
1168 * including udp header and folding it to skb->csum.
1169 */
1170 static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1171 int proto)
1172 {
1173 const struct iphdr *iph;
1174 int err;
1175
1176 UDP_SKB_CB(skb)->partial_cov = 0;
1177 UDP_SKB_CB(skb)->cscov = skb->len;
1178
1179 if (proto == IPPROTO_UDPLITE) {
1180 err = udplite_checksum_init(skb, uh);
1181 if (err)
1182 return err;
1183 }
1184
1185 iph = ip_hdr(skb);
1186 if (uh->check == 0) {
1187 skb->ip_summed = CHECKSUM_UNNECESSARY;
1188 } else if (skb->ip_summed == CHECKSUM_COMPLETE) {
1189 if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len,
1190 proto, skb->csum))
1191 skb->ip_summed = CHECKSUM_UNNECESSARY;
1192 }
1193 if (!skb_csum_unnecessary(skb))
1194 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
1195 skb->len, proto, 0);
1196 /* Probably, we should checksum udp header (it should be in cache
1197 * in any case) and data in tiny packets (< rx copybreak).
1198 */
1199
1200 return 0;
1201 }
1202
1203 /*
1204 * All we need to do is get the socket, and then do a checksum.
1205 */
1206
1207 int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[],
1208 int proto)
1209 {
1210 struct sock *sk;
1211 struct udphdr *uh = udp_hdr(skb);
1212 unsigned short ulen;
1213 struct rtable *rt = (struct rtable*)skb->dst;
1214 __be32 saddr = ip_hdr(skb)->saddr;
1215 __be32 daddr = ip_hdr(skb)->daddr;
1216
1217 /*
1218 * Validate the packet.
1219 */
1220 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1221 goto drop; /* No space for header. */
1222
1223 ulen = ntohs(uh->len);
1224 if (ulen > skb->len)
1225 goto short_packet;
1226
1227 if (proto == IPPROTO_UDP) {
1228 /* UDP validates ulen. */
1229 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1230 goto short_packet;
1231 uh = udp_hdr(skb);
1232 }
1233
1234 if (udp4_csum_init(skb, uh, proto))
1235 goto csum_error;
1236
1237 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1238 return __udp4_lib_mcast_deliver(skb, uh, saddr, daddr, udptable);
1239
1240 sk = __udp4_lib_lookup(saddr, uh->source, daddr, uh->dest,
1241 skb->dev->ifindex, udptable );
1242
1243 if (sk != NULL) {
1244 int ret = udp_queue_rcv_skb(sk, skb);
1245 sock_put(sk);
1246
1247 /* a return value > 0 means to resubmit the input, but
1248 * it wants the return to be -protocol, or 0
1249 */
1250 if (ret > 0)
1251 return -ret;
1252 return 0;
1253 }
1254
1255 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1256 goto drop;
1257 nf_reset(skb);
1258
1259 /* No socket. Drop packet silently, if checksum is wrong */
1260 if (udp_lib_checksum_complete(skb))
1261 goto csum_error;
1262
1263 UDP_INC_STATS_BH(UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
1264 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1265
1266 /*
1267 * Hmm. We got an UDP packet to a port to which we
1268 * don't wanna listen. Ignore it.
1269 */
1270 kfree_skb(skb);
1271 return 0;
1272
1273 short_packet:
1274 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n",
1275 proto == IPPROTO_UDPLITE ? "-Lite" : "",
1276 NIPQUAD(saddr),
1277 ntohs(uh->source),
1278 ulen,
1279 skb->len,
1280 NIPQUAD(daddr),
1281 ntohs(uh->dest));
1282 goto drop;
1283
1284 csum_error:
1285 /*
1286 * RFC1122: OK. Discards the bad packet silently (as far as
1287 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1288 */
1289 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n",
1290 proto == IPPROTO_UDPLITE ? "-Lite" : "",
1291 NIPQUAD(saddr),
1292 ntohs(uh->source),
1293 NIPQUAD(daddr),
1294 ntohs(uh->dest),
1295 ulen);
1296 drop:
1297 UDP_INC_STATS_BH(UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1298 kfree_skb(skb);
1299 return 0;
1300 }
1301
1302 int udp_rcv(struct sk_buff *skb)
1303 {
1304 return __udp4_lib_rcv(skb, udp_hash, IPPROTO_UDP);
1305 }
1306
1307 int udp_destroy_sock(struct sock *sk)
1308 {
1309 lock_sock(sk);
1310 udp_flush_pending_frames(sk);
1311 release_sock(sk);
1312 return 0;
1313 }
1314
1315 /*
1316 * Socket option code for UDP
1317 */
1318 int udp_lib_setsockopt(struct sock *sk, int level, int optname,
1319 char __user *optval, int optlen,
1320 int (*push_pending_frames)(struct sock *))
1321 {
1322 struct udp_sock *up = udp_sk(sk);
1323 int val;
1324 int err = 0;
1325
1326 if (optlen<sizeof(int))
1327 return -EINVAL;
1328
1329 if (get_user(val, (int __user *)optval))
1330 return -EFAULT;
1331
1332 switch (optname) {
1333 case UDP_CORK:
1334 if (val != 0) {
1335 up->corkflag = 1;
1336 } else {
1337 up->corkflag = 0;
1338 lock_sock(sk);
1339 (*push_pending_frames)(sk);
1340 release_sock(sk);
1341 }
1342 break;
1343
1344 case UDP_ENCAP:
1345 switch (val) {
1346 case 0:
1347 case UDP_ENCAP_ESPINUDP:
1348 case UDP_ENCAP_ESPINUDP_NON_IKE:
1349 up->encap_type = val;
1350 break;
1351 default:
1352 err = -ENOPROTOOPT;
1353 break;
1354 }
1355 break;
1356
1357 /*
1358 * UDP-Lite's partial checksum coverage (RFC 3828).
1359 */
1360 /* The sender sets actual checksum coverage length via this option.
1361 * The case coverage > packet length is handled by send module. */
1362 case UDPLITE_SEND_CSCOV:
1363 if (!up->pcflag) /* Disable the option on UDP sockets */
1364 return -ENOPROTOOPT;
1365 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
1366 val = 8;
1367 up->pcslen = val;
1368 up->pcflag |= UDPLITE_SEND_CC;
1369 break;
1370
1371 /* The receiver specifies a minimum checksum coverage value. To make
1372 * sense, this should be set to at least 8 (as done below). If zero is
1373 * used, this again means full checksum coverage. */
1374 case UDPLITE_RECV_CSCOV:
1375 if (!up->pcflag) /* Disable the option on UDP sockets */
1376 return -ENOPROTOOPT;
1377 if (val != 0 && val < 8) /* Avoid silly minimal values. */
1378 val = 8;
1379 up->pcrlen = val;
1380 up->pcflag |= UDPLITE_RECV_CC;
1381 break;
1382
1383 default:
1384 err = -ENOPROTOOPT;
1385 break;
1386 }
1387
1388 return err;
1389 }
1390
1391 int udp_setsockopt(struct sock *sk, int level, int optname,
1392 char __user *optval, int optlen)
1393 {
1394 if (level == SOL_UDP || level == SOL_UDPLITE)
1395 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1396 udp_push_pending_frames);
1397 return ip_setsockopt(sk, level, optname, optval, optlen);
1398 }
1399
1400 #ifdef CONFIG_COMPAT
1401 int compat_udp_setsockopt(struct sock *sk, int level, int optname,
1402 char __user *optval, int optlen)
1403 {
1404 if (level == SOL_UDP || level == SOL_UDPLITE)
1405 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1406 udp_push_pending_frames);
1407 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
1408 }
1409 #endif
1410
1411 int udp_lib_getsockopt(struct sock *sk, int level, int optname,
1412 char __user *optval, int __user *optlen)
1413 {
1414 struct udp_sock *up = udp_sk(sk);
1415 int val, len;
1416
1417 if (get_user(len,optlen))
1418 return -EFAULT;
1419
1420 len = min_t(unsigned int, len, sizeof(int));
1421
1422 if (len < 0)
1423 return -EINVAL;
1424
1425 switch (optname) {
1426 case UDP_CORK:
1427 val = up->corkflag;
1428 break;
1429
1430 case UDP_ENCAP:
1431 val = up->encap_type;
1432 break;
1433
1434 /* The following two cannot be changed on UDP sockets, the return is
1435 * always 0 (which corresponds to the full checksum coverage of UDP). */
1436 case UDPLITE_SEND_CSCOV:
1437 val = up->pcslen;
1438 break;
1439
1440 case UDPLITE_RECV_CSCOV:
1441 val = up->pcrlen;
1442 break;
1443
1444 default:
1445 return -ENOPROTOOPT;
1446 }
1447
1448 if (put_user(len, optlen))
1449 return -EFAULT;
1450 if (copy_to_user(optval, &val,len))
1451 return -EFAULT;
1452 return 0;
1453 }
1454
1455 int udp_getsockopt(struct sock *sk, int level, int optname,
1456 char __user *optval, int __user *optlen)
1457 {
1458 if (level == SOL_UDP || level == SOL_UDPLITE)
1459 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1460 return ip_getsockopt(sk, level, optname, optval, optlen);
1461 }
1462
1463 #ifdef CONFIG_COMPAT
1464 int compat_udp_getsockopt(struct sock *sk, int level, int optname,
1465 char __user *optval, int __user *optlen)
1466 {
1467 if (level == SOL_UDP || level == SOL_UDPLITE)
1468 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1469 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
1470 }
1471 #endif
1472 /**
1473 * udp_poll - wait for a UDP event.
1474 * @file - file struct
1475 * @sock - socket
1476 * @wait - poll table
1477 *
1478 * This is same as datagram poll, except for the special case of
1479 * blocking sockets. If application is using a blocking fd
1480 * and a packet with checksum error is in the queue;
1481 * then it could get return from select indicating data available
1482 * but then block when reading it. Add special case code
1483 * to work around these arguably broken applications.
1484 */
1485 unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1486 {
1487 unsigned int mask = datagram_poll(file, sock, wait);
1488 struct sock *sk = sock->sk;
1489 int is_lite = IS_UDPLITE(sk);
1490
1491 /* Check for false positives due to checksum errors */
1492 if ( (mask & POLLRDNORM) &&
1493 !(file->f_flags & O_NONBLOCK) &&
1494 !(sk->sk_shutdown & RCV_SHUTDOWN)){
1495 struct sk_buff_head *rcvq = &sk->sk_receive_queue;
1496 struct sk_buff *skb;
1497
1498 spin_lock_bh(&rcvq->lock);
1499 while ((skb = skb_peek(rcvq)) != NULL &&
1500 udp_lib_checksum_complete(skb)) {
1501 UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_lite);
1502 __skb_unlink(skb, rcvq);
1503 kfree_skb(skb);
1504 }
1505 spin_unlock_bh(&rcvq->lock);
1506
1507 /* nothing to see, move along */
1508 if (skb == NULL)
1509 mask &= ~(POLLIN | POLLRDNORM);
1510 }
1511
1512 return mask;
1513
1514 }
1515
1516 struct proto udp_prot = {
1517 .name = "UDP",
1518 .owner = THIS_MODULE,
1519 .close = udp_lib_close,
1520 .connect = ip4_datagram_connect,
1521 .disconnect = udp_disconnect,
1522 .ioctl = udp_ioctl,
1523 .destroy = udp_destroy_sock,
1524 .setsockopt = udp_setsockopt,
1525 .getsockopt = udp_getsockopt,
1526 .sendmsg = udp_sendmsg,
1527 .recvmsg = udp_recvmsg,
1528 .sendpage = udp_sendpage,
1529 .backlog_rcv = udp_queue_rcv_skb,
1530 .hash = udp_lib_hash,
1531 .unhash = udp_lib_unhash,
1532 .get_port = udp_v4_get_port,
1533 .obj_size = sizeof(struct udp_sock),
1534 #ifdef CONFIG_COMPAT
1535 .compat_setsockopt = compat_udp_setsockopt,
1536 .compat_getsockopt = compat_udp_getsockopt,
1537 #endif
1538 };
1539
1540 /* ------------------------------------------------------------------------ */
1541 #ifdef CONFIG_PROC_FS
1542
1543 static struct sock *udp_get_first(struct seq_file *seq)
1544 {
1545 struct sock *sk;
1546 struct udp_iter_state *state = seq->private;
1547
1548 for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) {
1549 struct hlist_node *node;
1550 sk_for_each(sk, node, state->hashtable + state->bucket) {
1551 if (sk->sk_family == state->family)
1552 goto found;
1553 }
1554 }
1555 sk = NULL;
1556 found:
1557 return sk;
1558 }
1559
1560 static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
1561 {
1562 struct udp_iter_state *state = seq->private;
1563
1564 do {
1565 sk = sk_next(sk);
1566 try_again:
1567 ;
1568 } while (sk && sk->sk_family != state->family);
1569
1570 if (!sk && ++state->bucket < UDP_HTABLE_SIZE) {
1571 sk = sk_head(state->hashtable + state->bucket);
1572 goto try_again;
1573 }
1574 return sk;
1575 }
1576
1577 static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
1578 {
1579 struct sock *sk = udp_get_first(seq);
1580
1581 if (sk)
1582 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
1583 --pos;
1584 return pos ? NULL : sk;
1585 }
1586
1587 static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
1588 {
1589 read_lock(&udp_hash_lock);
1590 return *pos ? udp_get_idx(seq, *pos-1) : (void *)1;
1591 }
1592
1593 static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1594 {
1595 struct sock *sk;
1596
1597 if (v == (void *)1)
1598 sk = udp_get_idx(seq, 0);
1599 else
1600 sk = udp_get_next(seq, v);
1601
1602 ++*pos;
1603 return sk;
1604 }
1605
1606 static void udp_seq_stop(struct seq_file *seq, void *v)
1607 {
1608 read_unlock(&udp_hash_lock);
1609 }
1610
1611 static int udp_seq_open(struct inode *inode, struct file *file)
1612 {
1613 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
1614 struct seq_file *seq;
1615 int rc = -ENOMEM;
1616 struct udp_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
1617
1618 if (!s)
1619 goto out;
1620 s->family = afinfo->family;
1621 s->hashtable = afinfo->hashtable;
1622 s->seq_ops.start = udp_seq_start;
1623 s->seq_ops.next = udp_seq_next;
1624 s->seq_ops.show = afinfo->seq_show;
1625 s->seq_ops.stop = udp_seq_stop;
1626
1627 rc = seq_open(file, &s->seq_ops);
1628 if (rc)
1629 goto out_kfree;
1630
1631 seq = file->private_data;
1632 seq->private = s;
1633 out:
1634 return rc;
1635 out_kfree:
1636 kfree(s);
1637 goto out;
1638 }
1639
1640 /* ------------------------------------------------------------------------ */
1641 int udp_proc_register(struct udp_seq_afinfo *afinfo)
1642 {
1643 struct proc_dir_entry *p;
1644 int rc = 0;
1645
1646 if (!afinfo)
1647 return -EINVAL;
1648 afinfo->seq_fops->owner = afinfo->owner;
1649 afinfo->seq_fops->open = udp_seq_open;
1650 afinfo->seq_fops->read = seq_read;
1651 afinfo->seq_fops->llseek = seq_lseek;
1652 afinfo->seq_fops->release = seq_release_private;
1653
1654 p = proc_net_fops_create(afinfo->name, S_IRUGO, afinfo->seq_fops);
1655 if (p)
1656 p->data = afinfo;
1657 else
1658 rc = -ENOMEM;
1659 return rc;
1660 }
1661
1662 void udp_proc_unregister(struct udp_seq_afinfo *afinfo)
1663 {
1664 if (!afinfo)
1665 return;
1666 proc_net_remove(afinfo->name);
1667 memset(afinfo->seq_fops, 0, sizeof(*afinfo->seq_fops));
1668 }
1669
1670 /* ------------------------------------------------------------------------ */
1671 static void udp4_format_sock(struct sock *sp, char *tmpbuf, int bucket)
1672 {
1673 struct inet_sock *inet = inet_sk(sp);
1674 __be32 dest = inet->daddr;
1675 __be32 src = inet->rcv_saddr;
1676 __u16 destp = ntohs(inet->dport);
1677 __u16 srcp = ntohs(inet->sport);
1678
1679 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
1680 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
1681 bucket, src, srcp, dest, destp, sp->sk_state,
1682 atomic_read(&sp->sk_wmem_alloc),
1683 atomic_read(&sp->sk_rmem_alloc),
1684 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
1685 atomic_read(&sp->sk_refcnt), sp);
1686 }
1687
1688 int udp4_seq_show(struct seq_file *seq, void *v)
1689 {
1690 if (v == SEQ_START_TOKEN)
1691 seq_printf(seq, "%-127s\n",
1692 " sl local_address rem_address st tx_queue "
1693 "rx_queue tr tm->when retrnsmt uid timeout "
1694 "inode");
1695 else {
1696 char tmpbuf[129];
1697 struct udp_iter_state *state = seq->private;
1698
1699 udp4_format_sock(v, tmpbuf, state->bucket);
1700 seq_printf(seq, "%-127s\n", tmpbuf);
1701 }
1702 return 0;
1703 }
1704
1705 /* ------------------------------------------------------------------------ */
1706 static struct file_operations udp4_seq_fops;
1707 static struct udp_seq_afinfo udp4_seq_afinfo = {
1708 .owner = THIS_MODULE,
1709 .name = "udp",
1710 .family = AF_INET,
1711 .hashtable = udp_hash,
1712 .seq_show = udp4_seq_show,
1713 .seq_fops = &udp4_seq_fops,
1714 };
1715
1716 int __init udp4_proc_init(void)
1717 {
1718 return udp_proc_register(&udp4_seq_afinfo);
1719 }
1720
1721 void udp4_proc_exit(void)
1722 {
1723 udp_proc_unregister(&udp4_seq_afinfo);
1724 }
1725 #endif /* CONFIG_PROC_FS */
1726
1727 EXPORT_SYMBOL(udp_disconnect);
1728 EXPORT_SYMBOL(udp_hash);
1729 EXPORT_SYMBOL(udp_hash_lock);
1730 EXPORT_SYMBOL(udp_ioctl);
1731 EXPORT_SYMBOL(udp_get_port);
1732 EXPORT_SYMBOL(udp_prot);
1733 EXPORT_SYMBOL(udp_sendmsg);
1734 EXPORT_SYMBOL(udp_lib_getsockopt);
1735 EXPORT_SYMBOL(udp_lib_setsockopt);
1736 EXPORT_SYMBOL(udp_poll);
1737
1738 #ifdef CONFIG_PROC_FS
1739 EXPORT_SYMBOL(udp_proc_register);
1740 EXPORT_SYMBOL(udp_proc_unregister);
1741 #endif
Verdex Pro support