?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[NET]: Supporting UDP-Lite (RFC 3828) in Linux
[linux-2.6/kvm.git] / net / ipv4 / udp.c
blob98ba75096175ff1a95c9bc3a8918812cd1450a20
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(__be16 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 (inet_sk(sk)->num == 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 best_size_so_far = size;
171 best = result;
172 }
173 }
174 result = best;
175 for(i = 0; i < (1 << 16) / UDP_HTABLE_SIZE; i++, result += UDP_HTABLE_SIZE) {
176 if (result > sysctl_local_port_range[1])
177 result = sysctl_local_port_range[0]
178 + ((result - sysctl_local_port_range[0]) &
179 (UDP_HTABLE_SIZE - 1));
180 if (! __udp_lib_lport_inuse(result, udptable))
181 break;
182 }
183 if (i >= (1 << 16) / UDP_HTABLE_SIZE)
184 goto fail;
185 gotit:
186 *port_rover = snum = result;
187 } else {
188 head = &udptable[snum & (UDP_HTABLE_SIZE - 1)];
189
190 sk_for_each(sk2, node, head)
191 if (inet_sk(sk2)->num == snum &&
192 sk2 != sk &&
193 (!sk2->sk_reuse || !sk->sk_reuse) &&
194 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if
195 || sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
196 (*saddr_comp)(sk, sk2) )
197 goto fail;
198 }
199 inet_sk(sk)->num = snum;
200 if (sk_unhashed(sk)) {
201 head = &udptable[snum & (UDP_HTABLE_SIZE - 1)];
202 sk_add_node(sk, head);
203 sock_prot_inc_use(sk->sk_prot);
204 }
205 error = 0;
206 fail:
207 write_unlock_bh(&udp_hash_lock);
208 return error;
209 }
210
211 __inline__ int udp_get_port(struct sock *sk, unsigned short snum,
212 int (*scmp)(const struct sock *, const struct sock *))
213 {
214 return __udp_lib_get_port(sk, snum, udp_hash, &udp_port_rover, scmp);
215 }
216
217 inline int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
218 {
219 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
220
221 return ( !ipv6_only_sock(sk2) &&
222 (!inet1->rcv_saddr || !inet2->rcv_saddr ||
223 inet1->rcv_saddr == inet2->rcv_saddr ));
224 }
225
226 static inline int udp_v4_get_port(struct sock *sk, unsigned short snum)
227 {
228 return udp_get_port(sk, snum, ipv4_rcv_saddr_equal);
229 }
230
231 /* UDP is nearly always wildcards out the wazoo, it makes no sense to try
232 * harder than this. -DaveM
233 */
234 static struct sock *__udp4_lib_lookup(__be32 saddr, __be16 sport,
235 __be32 daddr, __be16 dport,
236 int dif, struct hlist_head udptable[])
237 {
238 struct sock *sk, *result = NULL;
239 struct hlist_node *node;
240 unsigned short hnum = ntohs(dport);
241 int badness = -1;
242
243 read_lock(&udp_hash_lock);
244 sk_for_each(sk, node, &udptable[hnum & (UDP_HTABLE_SIZE - 1)]) {
245 struct inet_sock *inet = inet_sk(sk);
246
247 if (inet->num == hnum && !ipv6_only_sock(sk)) {
248 int score = (sk->sk_family == PF_INET ? 1 : 0);
249 if (inet->rcv_saddr) {
250 if (inet->rcv_saddr != daddr)
251 continue;
252 score+=2;
253 }
254 if (inet->daddr) {
255 if (inet->daddr != saddr)
256 continue;
257 score+=2;
258 }
259 if (inet->dport) {
260 if (inet->dport != sport)
261 continue;
262 score+=2;
263 }
264 if (sk->sk_bound_dev_if) {
265 if (sk->sk_bound_dev_if != dif)
266 continue;
267 score+=2;
268 }
269 if(score == 9) {
270 result = sk;
271 break;
272 } else if(score > badness) {
273 result = sk;
274 badness = score;
275 }
276 }
277 }
278 if (result)
279 sock_hold(result);
280 read_unlock(&udp_hash_lock);
281 return result;
282 }
283
284 static inline struct sock *udp_v4_mcast_next(struct sock *sk,
285 __be16 loc_port, __be32 loc_addr,
286 __be16 rmt_port, __be32 rmt_addr,
287 int dif)
288 {
289 struct hlist_node *node;
290 struct sock *s = sk;
291 unsigned short hnum = ntohs(loc_port);
292
293 sk_for_each_from(s, node) {
294 struct inet_sock *inet = inet_sk(s);
295
296 if (inet->num != hnum ||
297 (inet->daddr && inet->daddr != rmt_addr) ||
298 (inet->dport != rmt_port && inet->dport) ||
299 (inet->rcv_saddr && inet->rcv_saddr != loc_addr) ||
300 ipv6_only_sock(s) ||
301 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
302 continue;
303 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
304 continue;
305 goto found;
306 }
307 s = NULL;
308 found:
309 return s;
310 }
311
312 /*
313 * This routine is called by the ICMP module when it gets some
314 * sort of error condition. If err < 0 then the socket should
315 * be closed and the error returned to the user. If err > 0
316 * it's just the icmp type << 8 | icmp code.
317 * Header points to the ip header of the error packet. We move
318 * on past this. Then (as it used to claim before adjustment)
319 * header points to the first 8 bytes of the udp header. We need
320 * to find the appropriate port.
321 */
322
323 void __udp4_lib_err(struct sk_buff *skb, u32 info, struct hlist_head udptable[])
324 {
325 struct inet_sock *inet;
326 struct iphdr *iph = (struct iphdr*)skb->data;
327 struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2));
328 int type = skb->h.icmph->type;
329 int code = skb->h.icmph->code;
330 struct sock *sk;
331 int harderr;
332 int err;
333
334 sk = __udp4_lib_lookup(iph->daddr, uh->dest, iph->saddr, uh->source,
335 skb->dev->ifindex, udptable );
336 if (sk == NULL) {
337 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
338 return; /* No socket for error */
339 }
340
341 err = 0;
342 harderr = 0;
343 inet = inet_sk(sk);
344
345 switch (type) {
346 default:
347 case ICMP_TIME_EXCEEDED:
348 err = EHOSTUNREACH;
349 break;
350 case ICMP_SOURCE_QUENCH:
351 goto out;
352 case ICMP_PARAMETERPROB:
353 err = EPROTO;
354 harderr = 1;
355 break;
356 case ICMP_DEST_UNREACH:
357 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
358 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
359 err = EMSGSIZE;
360 harderr = 1;
361 break;
362 }
363 goto out;
364 }
365 err = EHOSTUNREACH;
366 if (code <= NR_ICMP_UNREACH) {
367 harderr = icmp_err_convert[code].fatal;
368 err = icmp_err_convert[code].errno;
369 }
370 break;
371 }
372
373 /*
374 * RFC1122: OK. Passes ICMP errors back to application, as per
375 * 4.1.3.3.
376 */
377 if (!inet->recverr) {
378 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
379 goto out;
380 } else {
381 ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1));
382 }
383 sk->sk_err = err;
384 sk->sk_error_report(sk);
385 out:
386 sock_put(sk);
387 }
388
389 __inline__ void udp_err(struct sk_buff *skb, u32 info)
390 {
391 return __udp4_lib_err(skb, info, udp_hash);
392 }
393
394 /*
395 * Throw away all pending data and cancel the corking. Socket is locked.
396 */
397 static void udp_flush_pending_frames(struct sock *sk)
398 {
399 struct udp_sock *up = udp_sk(sk);
400
401 if (up->pending) {
402 up->len = 0;
403 up->pending = 0;
404 ip_flush_pending_frames(sk);
405 }
406 }
407
408 /**
409 * udp4_hwcsum_outgoing - handle outgoing HW checksumming
410 * @sk: socket we are sending on
411 * @skb: sk_buff containing the filled-in UDP header
412 * (checksum field must be zeroed out)
413 */
414 static void udp4_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
415 __be32 src, __be32 dst, int len )
416 {
417 unsigned int csum = 0, offset;
418 struct udphdr *uh = skb->h.uh;
419
420 if (skb_queue_len(&sk->sk_write_queue) == 1) {
421 /*
422 * Only one fragment on the socket.
423 */
424 skb->csum = offsetof(struct udphdr, check);
425 uh->check = ~csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, 0);
426 } else {
427 /*
428 * HW-checksum won't work as there are two or more
429 * fragments on the socket so that all csums of sk_buffs
430 * should be together
431 */
432 offset = skb->h.raw - skb->data;
433 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
434
435 skb->ip_summed = CHECKSUM_NONE;
436
437 skb_queue_walk(&sk->sk_write_queue, skb) {
438 csum = csum_add(csum, skb->csum);
439 }
440
441 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
442 if (uh->check == 0)
443 uh->check = -1;
444 }
445 }
446
447 /*
448 * Push out all pending data as one UDP datagram. Socket is locked.
449 */
450 int udp_push_pending_frames(struct sock *sk, struct udp_sock *up)
451 {
452 struct inet_sock *inet = inet_sk(sk);
453 struct flowi *fl = &inet->cork.fl;
454 struct sk_buff *skb;
455 struct udphdr *uh;
456 int err = 0;
457 u32 csum = 0;
458
459 /* Grab the skbuff where UDP header space exists. */
460 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
461 goto out;
462
463 /*
464 * Create a UDP header
465 */
466 uh = skb->h.uh;
467 uh->source = fl->fl_ip_sport;
468 uh->dest = fl->fl_ip_dport;
469 uh->len = htons(up->len);
470 uh->check = 0;
471
472 if (up->pcflag) /* UDP-Lite */
473 csum = udplite_csum_outgoing(sk, skb);
474
475 else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */
476
477 skb->ip_summed = CHECKSUM_NONE;
478 goto send;
479
480 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
481
482 udp4_hwcsum_outgoing(sk, skb, fl->fl4_src,fl->fl4_dst, up->len);
483 goto send;
484
485 } else /* `normal' UDP */
486 csum = udp_csum_outgoing(sk, skb);
487
488 /* add protocol-dependent pseudo-header */
489 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, up->len,
490 sk->sk_protocol, csum );
491 if (uh->check == 0)
492 uh->check = -1;
493
494 send:
495 err = ip_push_pending_frames(sk);
496 out:
497 up->len = 0;
498 up->pending = 0;
499 return err;
500 }
501
502 int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
503 size_t len)
504 {
505 struct inet_sock *inet = inet_sk(sk);
506 struct udp_sock *up = udp_sk(sk);
507 int ulen = len;
508 struct ipcm_cookie ipc;
509 struct rtable *rt = NULL;
510 int free = 0;
511 int connected = 0;
512 __be32 daddr, faddr, saddr;
513 __be16 dport;
514 u8 tos;
515 int err, is_udplite = up->pcflag;
516 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
517 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
518
519 if (len > 0xFFFF)
520 return -EMSGSIZE;
521
522 /*
523 * Check the flags.
524 */
525
526 if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */
527 return -EOPNOTSUPP;
528
529 ipc.opt = NULL;
530
531 if (up->pending) {
532 /*
533 * There are pending frames.
534 * The socket lock must be held while it's corked.
535 */
536 lock_sock(sk);
537 if (likely(up->pending)) {
538 if (unlikely(up->pending != AF_INET)) {
539 release_sock(sk);
540 return -EINVAL;
541 }
542 goto do_append_data;
543 }
544 release_sock(sk);
545 }
546 ulen += sizeof(struct udphdr);
547
548 /*
549 * Get and verify the address.
550 */
551 if (msg->msg_name) {
552 struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name;
553 if (msg->msg_namelen < sizeof(*usin))
554 return -EINVAL;
555 if (usin->sin_family != AF_INET) {
556 if (usin->sin_family != AF_UNSPEC)
557 return -EAFNOSUPPORT;
558 }
559
560 daddr = usin->sin_addr.s_addr;
561 dport = usin->sin_port;
562 if (dport == 0)
563 return -EINVAL;
564 } else {
565 if (sk->sk_state != TCP_ESTABLISHED)
566 return -EDESTADDRREQ;
567 daddr = inet->daddr;
568 dport = inet->dport;
569 /* Open fast path for connected socket.
570 Route will not be used, if at least one option is set.
571 */
572 connected = 1;
573 }
574 ipc.addr = inet->saddr;
575
576 ipc.oif = sk->sk_bound_dev_if;
577 if (msg->msg_controllen) {
578 err = ip_cmsg_send(msg, &ipc);
579 if (err)
580 return err;
581 if (ipc.opt)
582 free = 1;
583 connected = 0;
584 }
585 if (!ipc.opt)
586 ipc.opt = inet->opt;
587
588 saddr = ipc.addr;
589 ipc.addr = faddr = daddr;
590
591 if (ipc.opt && ipc.opt->srr) {
592 if (!daddr)
593 return -EINVAL;
594 faddr = ipc.opt->faddr;
595 connected = 0;
596 }
597 tos = RT_TOS(inet->tos);
598 if (sock_flag(sk, SOCK_LOCALROUTE) ||
599 (msg->msg_flags & MSG_DONTROUTE) ||
600 (ipc.opt && ipc.opt->is_strictroute)) {
601 tos |= RTO_ONLINK;
602 connected = 0;
603 }
604
605 if (MULTICAST(daddr)) {
606 if (!ipc.oif)
607 ipc.oif = inet->mc_index;
608 if (!saddr)
609 saddr = inet->mc_addr;
610 connected = 0;
611 }
612
613 if (connected)
614 rt = (struct rtable*)sk_dst_check(sk, 0);
615
616 if (rt == NULL) {
617 struct flowi fl = { .oif = ipc.oif,
618 .nl_u = { .ip4_u =
619 { .daddr = faddr,
620 .saddr = saddr,
621 .tos = tos } },
622 .proto = sk->sk_protocol,
623 .uli_u = { .ports =
624 { .sport = inet->sport,
625 .dport = dport } } };
626 security_sk_classify_flow(sk, &fl);
627 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
628 if (err)
629 goto out;
630
631 err = -EACCES;
632 if ((rt->rt_flags & RTCF_BROADCAST) &&
633 !sock_flag(sk, SOCK_BROADCAST))
634 goto out;
635 if (connected)
636 sk_dst_set(sk, dst_clone(&rt->u.dst));
637 }
638
639 if (msg->msg_flags&MSG_CONFIRM)
640 goto do_confirm;
641 back_from_confirm:
642
643 saddr = rt->rt_src;
644 if (!ipc.addr)
645 daddr = ipc.addr = rt->rt_dst;
646
647 lock_sock(sk);
648 if (unlikely(up->pending)) {
649 /* The socket is already corked while preparing it. */
650 /* ... which is an evident application bug. --ANK */
651 release_sock(sk);
652
653 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
654 err = -EINVAL;
655 goto out;
656 }
657 /*
658 * Now cork the socket to pend data.
659 */
660 inet->cork.fl.fl4_dst = daddr;
661 inet->cork.fl.fl_ip_dport = dport;
662 inet->cork.fl.fl4_src = saddr;
663 inet->cork.fl.fl_ip_sport = inet->sport;
664 up->pending = AF_INET;
665
666 do_append_data:
667 up->len += ulen;
668 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
669 err = ip_append_data(sk, getfrag, msg->msg_iov, ulen,
670 sizeof(struct udphdr), &ipc, rt,
671 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
672 if (err)
673 udp_flush_pending_frames(sk);
674 else if (!corkreq)
675 err = udp_push_pending_frames(sk, up);
676 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
677 up->pending = 0;
678 release_sock(sk);
679
680 out:
681 ip_rt_put(rt);
682 if (free)
683 kfree(ipc.opt);
684 if (!err) {
685 UDP_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS, is_udplite);
686 return len;
687 }
688 /*
689 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
690 * ENOBUFS might not be good (it's not tunable per se), but otherwise
691 * we don't have a good statistic (IpOutDiscards but it can be too many
692 * things). We could add another new stat but at least for now that
693 * seems like overkill.
694 */
695 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
696 UDP_INC_STATS_USER(UDP_MIB_SNDBUFERRORS, is_udplite);
697 }
698 return err;
699
700 do_confirm:
701 dst_confirm(&rt->u.dst);
702 if (!(msg->msg_flags&MSG_PROBE) || len)
703 goto back_from_confirm;
704 err = 0;
705 goto out;
706 }
707
708 int udp_sendpage(struct sock *sk, struct page *page, int offset,
709 size_t size, int flags)
710 {
711 struct udp_sock *up = udp_sk(sk);
712 int ret;
713
714 if (!up->pending) {
715 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
716
717 /* Call udp_sendmsg to specify destination address which
718 * sendpage interface can't pass.
719 * This will succeed only when the socket is connected.
720 */
721 ret = udp_sendmsg(NULL, sk, &msg, 0);
722 if (ret < 0)
723 return ret;
724 }
725
726 lock_sock(sk);
727
728 if (unlikely(!up->pending)) {
729 release_sock(sk);
730
731 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n");
732 return -EINVAL;
733 }
734
735 ret = ip_append_page(sk, page, offset, size, flags);
736 if (ret == -EOPNOTSUPP) {
737 release_sock(sk);
738 return sock_no_sendpage(sk->sk_socket, page, offset,
739 size, flags);
740 }
741 if (ret < 0) {
742 udp_flush_pending_frames(sk);
743 goto out;
744 }
745
746 up->len += size;
747 if (!(up->corkflag || (flags&MSG_MORE)))
748 ret = udp_push_pending_frames(sk, up);
749 if (!ret)
750 ret = size;
751 out:
752 release_sock(sk);
753 return ret;
754 }
755
756 /*
757 * IOCTL requests applicable to the UDP protocol
758 */
759
760 int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
761 {
762 switch(cmd)
763 {
764 case SIOCOUTQ:
765 {
766 int amount = atomic_read(&sk->sk_wmem_alloc);
767 return put_user(amount, (int __user *)arg);
768 }
769
770 case SIOCINQ:
771 {
772 struct sk_buff *skb;
773 unsigned long amount;
774
775 amount = 0;
776 spin_lock_bh(&sk->sk_receive_queue.lock);
777 skb = skb_peek(&sk->sk_receive_queue);
778 if (skb != NULL) {
779 /*
780 * We will only return the amount
781 * of this packet since that is all
782 * that will be read.
783 */
784 amount = skb->len - sizeof(struct udphdr);
785 }
786 spin_unlock_bh(&sk->sk_receive_queue.lock);
787 return put_user(amount, (int __user *)arg);
788 }
789
790 default:
791 return -ENOIOCTLCMD;
792 }
793 return(0);
794 }
795
796 /*
797 * This should be easy, if there is something there we
798 * return it, otherwise we block.
799 */
800
801 int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
802 size_t len, int noblock, int flags, int *addr_len)
803 {
804 struct inet_sock *inet = inet_sk(sk);
805 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
806 struct sk_buff *skb;
807 int copied, err, copy_only, is_udplite = IS_UDPLITE(sk);
808
809 /*
810 * Check any passed addresses
811 */
812 if (addr_len)
813 *addr_len=sizeof(*sin);
814
815 if (flags & MSG_ERRQUEUE)
816 return ip_recv_error(sk, msg, len);
817
818 try_again:
819 skb = skb_recv_datagram(sk, flags, noblock, &err);
820 if (!skb)
821 goto out;
822
823 copied = skb->len - sizeof(struct udphdr);
824 if (copied > len) {
825 copied = len;
826 msg->msg_flags |= MSG_TRUNC;
827 }
828
829 /*
830 * Decide whether to checksum and/or copy data.
831 *
832 * UDP: checksum may have been computed in HW,
833 * (re-)compute it if message is truncated.
834 * UDP-Lite: always needs to checksum, no HW support.
835 */
836 copy_only = (skb->ip_summed==CHECKSUM_UNNECESSARY);
837
838 if (is_udplite || (!copy_only && msg->msg_flags&MSG_TRUNC)) {
839 if (__udp_lib_checksum_complete(skb))
840 goto csum_copy_err;
841 copy_only = 1;
842 }
843
844 if (copy_only)
845 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
846 msg->msg_iov, copied );
847 else {
848 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
849
850 if (err == -EINVAL)
851 goto csum_copy_err;
852 }
853
854 if (err)
855 goto out_free;
856
857 sock_recv_timestamp(msg, sk, skb);
858
859 /* Copy the address. */
860 if (sin)
861 {
862 sin->sin_family = AF_INET;
863 sin->sin_port = skb->h.uh->source;
864 sin->sin_addr.s_addr = skb->nh.iph->saddr;
865 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
866 }
867 if (inet->cmsg_flags)
868 ip_cmsg_recv(msg, skb);
869
870 err = copied;
871 if (flags & MSG_TRUNC)
872 err = skb->len - sizeof(struct udphdr);
873
874 out_free:
875 skb_free_datagram(sk, skb);
876 out:
877 return err;
878
879 csum_copy_err:
880 UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_udplite);
881
882 skb_kill_datagram(sk, skb, flags);
883
884 if (noblock)
885 return -EAGAIN;
886 goto try_again;
887 }
888
889
890 int udp_disconnect(struct sock *sk, int flags)
891 {
892 struct inet_sock *inet = inet_sk(sk);
893 /*
894 * 1003.1g - break association.
895 */
896
897 sk->sk_state = TCP_CLOSE;
898 inet->daddr = 0;
899 inet->dport = 0;
900 sk->sk_bound_dev_if = 0;
901 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
902 inet_reset_saddr(sk);
903
904 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
905 sk->sk_prot->unhash(sk);
906 inet->sport = 0;
907 }
908 sk_dst_reset(sk);
909 return 0;
910 }
911
912 /* return:
913 * 1 if the the UDP system should process it
914 * 0 if we should drop this packet
915 * -1 if it should get processed by xfrm4_rcv_encap
916 */
917 static int udp_encap_rcv(struct sock * sk, struct sk_buff *skb)
918 {
919 #ifndef CONFIG_XFRM
920 return 1;
921 #else
922 struct udp_sock *up = udp_sk(sk);
923 struct udphdr *uh;
924 struct iphdr *iph;
925 int iphlen, len;
926
927 __u8 *udpdata;
928 __be32 *udpdata32;
929 __u16 encap_type = up->encap_type;
930
931 /* if we're overly short, let UDP handle it */
932 len = skb->len - sizeof(struct udphdr);
933 if (len <= 0)
934 return 1;
935
936 /* if this is not encapsulated socket, then just return now */
937 if (!encap_type)
938 return 1;
939
940 /* If this is a paged skb, make sure we pull up
941 * whatever data we need to look at. */
942 if (!pskb_may_pull(skb, sizeof(struct udphdr) + min(len, 8)))
943 return 1;
944
945 /* Now we can get the pointers */
946 uh = skb->h.uh;
947 udpdata = (__u8 *)uh + sizeof(struct udphdr);
948 udpdata32 = (__be32 *)udpdata;
949
950 switch (encap_type) {
951 default:
952 case UDP_ENCAP_ESPINUDP:
953 /* Check if this is a keepalive packet. If so, eat it. */
954 if (len == 1 && udpdata[0] == 0xff) {
955 return 0;
956 } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0 ) {
957 /* ESP Packet without Non-ESP header */
958 len = sizeof(struct udphdr);
959 } else
960 /* Must be an IKE packet.. pass it through */
961 return 1;
962 break;
963 case UDP_ENCAP_ESPINUDP_NON_IKE:
964 /* Check if this is a keepalive packet. If so, eat it. */
965 if (len == 1 && udpdata[0] == 0xff) {
966 return 0;
967 } else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) &&
968 udpdata32[0] == 0 && udpdata32[1] == 0) {
969
970 /* ESP Packet with Non-IKE marker */
971 len = sizeof(struct udphdr) + 2 * sizeof(u32);
972 } else
973 /* Must be an IKE packet.. pass it through */
974 return 1;
975 break;
976 }
977
978 /* At this point we are sure that this is an ESPinUDP packet,
979 * so we need to remove 'len' bytes from the packet (the UDP
980 * header and optional ESP marker bytes) and then modify the
981 * protocol to ESP, and then call into the transform receiver.
982 */
983 if (skb_cloned(skb) && pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
984 return 0;
985
986 /* Now we can update and verify the packet length... */
987 iph = skb->nh.iph;
988 iphlen = iph->ihl << 2;
989 iph->tot_len = htons(ntohs(iph->tot_len) - len);
990 if (skb->len < iphlen + len) {
991 /* packet is too small!?! */
992 return 0;
993 }
994
995 /* pull the data buffer up to the ESP header and set the
996 * transport header to point to ESP. Keep UDP on the stack
997 * for later.
998 */
999 skb->h.raw = skb_pull(skb, len);
1000
1001 /* modify the protocol (it's ESP!) */
1002 iph->protocol = IPPROTO_ESP;
1003
1004 /* and let the caller know to send this into the ESP processor... */
1005 return -1;
1006 #endif
1007 }
1008
1009 /* returns:
1010 * -1: error
1011 * 0: success
1012 * >0: "udp encap" protocol resubmission
1013 *
1014 * Note that in the success and error cases, the skb is assumed to
1015 * have either been requeued or freed.
1016 */
1017 int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
1018 {
1019 struct udp_sock *up = udp_sk(sk);
1020 int rc;
1021
1022 /*
1023 * Charge it to the socket, dropping if the queue is full.
1024 */
1025 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1026 goto drop;
1027 nf_reset(skb);
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, up->pcflag);
1051 return -ret;
1052 }
1053 /* FALLTHROUGH -- it's a UDP Packet */
1054 }
1055
1056 /*
1057 * UDP-Lite specific tests, ignored on UDP sockets
1058 */
1059 if ((up->pcflag & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1060
1061 /*
1062 * MIB statistics other than incrementing the error count are
1063 * disabled for the following two types of errors: these depend
1064 * on the application settings, not on the functioning of the
1065 * protocol stack as such.
1066 *
1067 * RFC 3828 here recommends (sec 3.3): "There should also be a
1068 * way ... to ... at least let the receiving application block
1069 * delivery of packets with coverage values less than a value
1070 * provided by the application."
1071 */
1072 if (up->pcrlen == 0) { /* full coverage was set */
1073 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE: partial coverage "
1074 "%d while full coverage %d requested\n",
1075 UDP_SKB_CB(skb)->cscov, skb->len);
1076 goto drop;
1077 }
1078 /* The next case involves violating the min. coverage requested
1079 * by the receiver. This is subtle: if receiver wants x and x is
1080 * greater than the buffersize/MTU then receiver will complain
1081 * that it wants x while sender emits packets of smaller size y.
1082 * Therefore the above ...()->partial_cov statement is essential.
1083 */
1084 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
1085 LIMIT_NETDEBUG(KERN_WARNING
1086 "UDPLITE: coverage %d too small, need min %d\n",
1087 UDP_SKB_CB(skb)->cscov, up->pcrlen);
1088 goto drop;
1089 }
1090 }
1091
1092 if (sk->sk_filter && skb->ip_summed != CHECKSUM_UNNECESSARY) {
1093 if (__udp_lib_checksum_complete(skb))
1094 goto drop;
1095 skb->ip_summed = CHECKSUM_UNNECESSARY;
1096 }
1097
1098 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
1099 /* Note that an ENOMEM error is charged twice */
1100 if (rc == -ENOMEM)
1101 UDP_INC_STATS_BH(UDP_MIB_RCVBUFERRORS, up->pcflag);
1102 goto drop;
1103 }
1104
1105 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS, up->pcflag);
1106 return 0;
1107
1108 drop:
1109 UDP_INC_STATS_BH(UDP_MIB_INERRORS, up->pcflag);
1110 kfree_skb(skb);
1111 return -1;
1112 }
1113
1114 /*
1115 * Multicasts and broadcasts go to each listener.
1116 *
1117 * Note: called only from the BH handler context,
1118 * so we don't need to lock the hashes.
1119 */
1120 static int __udp4_lib_mcast_deliver(struct sk_buff *skb,
1121 struct udphdr *uh,
1122 __be32 saddr, __be32 daddr,
1123 struct hlist_head udptable[])
1124 {
1125 struct sock *sk;
1126 int dif;
1127
1128 read_lock(&udp_hash_lock);
1129 sk = sk_head(&udptable[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
1130 dif = skb->dev->ifindex;
1131 sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
1132 if (sk) {
1133 struct sock *sknext = NULL;
1134
1135 do {
1136 struct sk_buff *skb1 = skb;
1137
1138 sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr,
1139 uh->source, saddr, dif);
1140 if(sknext)
1141 skb1 = skb_clone(skb, GFP_ATOMIC);
1142
1143 if(skb1) {
1144 int ret = udp_queue_rcv_skb(sk, skb1);
1145 if (ret > 0)
1146 /* we should probably re-process instead
1147 * of dropping packets here. */
1148 kfree_skb(skb1);
1149 }
1150 sk = sknext;
1151 } while(sknext);
1152 } else
1153 kfree_skb(skb);
1154 read_unlock(&udp_hash_lock);
1155 return 0;
1156 }
1157
1158 /* Initialize UDP checksum. If exited with zero value (success),
1159 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1160 * Otherwise, csum completion requires chacksumming packet body,
1161 * including udp header and folding it to skb->csum.
1162 */
1163 static inline void udp4_csum_init(struct sk_buff *skb, struct udphdr *uh)
1164 {
1165 if (uh->check == 0) {
1166 skb->ip_summed = CHECKSUM_UNNECESSARY;
1167 } else if (skb->ip_summed == CHECKSUM_COMPLETE) {
1168 if (!csum_tcpudp_magic(skb->nh.iph->saddr, skb->nh.iph->daddr,
1169 skb->len, IPPROTO_UDP, skb->csum ))
1170 skb->ip_summed = CHECKSUM_UNNECESSARY;
1171 }
1172 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
1173 skb->csum = csum_tcpudp_nofold(skb->nh.iph->saddr,
1174 skb->nh.iph->daddr,
1175 skb->len, IPPROTO_UDP, 0);
1176 /* Probably, we should checksum udp header (it should be in cache
1177 * in any case) and data in tiny packets (< rx copybreak).
1178 */
1179
1180 /* UDP = UDP-Lite with a non-partial checksum coverage */
1181 UDP_SKB_CB(skb)->partial_cov = 0;
1182 }
1183
1184 /*
1185 * All we need to do is get the socket, and then do a checksum.
1186 */
1187
1188 int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[],
1189 int is_udplite)
1190 {
1191 struct sock *sk;
1192 struct udphdr *uh = skb->h.uh;
1193 unsigned short ulen;
1194 struct rtable *rt = (struct rtable*)skb->dst;
1195 __be32 saddr = skb->nh.iph->saddr;
1196 __be32 daddr = skb->nh.iph->daddr;
1197
1198 /*
1199 * Validate the packet.
1200 */
1201 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1202 goto drop; /* No space for header. */
1203
1204 ulen = ntohs(uh->len);
1205 if (ulen > skb->len)
1206 goto short_packet;
1207
1208 if(! is_udplite ) { /* UDP validates ulen. */
1209
1210 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1211 goto short_packet;
1212
1213 udp4_csum_init(skb, uh);
1214
1215 } else { /* UDP-Lite validates cscov. */
1216 if (udplite4_csum_init(skb, uh))
1217 goto csum_error;
1218 }
1219
1220 if(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1221 return __udp4_lib_mcast_deliver(skb, uh, saddr, daddr, udptable);
1222
1223 sk = __udp4_lib_lookup(saddr, uh->source, daddr, uh->dest,
1224 skb->dev->ifindex, udptable );
1225
1226 if (sk != NULL) {
1227 int ret = udp_queue_rcv_skb(sk, skb);
1228 sock_put(sk);
1229
1230 /* a return value > 0 means to resubmit the input, but
1231 * it wants the return to be -protocol, or 0
1232 */
1233 if (ret > 0)
1234 return -ret;
1235 return 0;
1236 }
1237
1238 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1239 goto drop;
1240 nf_reset(skb);
1241
1242 /* No socket. Drop packet silently, if checksum is wrong */
1243 if (udp_lib_checksum_complete(skb))
1244 goto csum_error;
1245
1246 UDP_INC_STATS_BH(UDP_MIB_NOPORTS, is_udplite);
1247 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1248
1249 /*
1250 * Hmm. We got an UDP packet to a port to which we
1251 * don't wanna listen. Ignore it.
1252 */
1253 kfree_skb(skb);
1254 return(0);
1255
1256 short_packet:
1257 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n",
1258 is_udplite? "-Lite" : "",
1259 NIPQUAD(saddr),
1260 ntohs(uh->source),
1261 ulen,
1262 skb->len,
1263 NIPQUAD(daddr),
1264 ntohs(uh->dest));
1265 goto drop;
1266
1267 csum_error:
1268 /*
1269 * RFC1122: OK. Discards the bad packet silently (as far as
1270 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1271 */
1272 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n",
1273 is_udplite? "-Lite" : "",
1274 NIPQUAD(saddr),
1275 ntohs(uh->source),
1276 NIPQUAD(daddr),
1277 ntohs(uh->dest),
1278 ulen);
1279 drop:
1280 UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_udplite);
1281 kfree_skb(skb);
1282 return(0);
1283 }
1284
1285 __inline__ int udp_rcv(struct sk_buff *skb)
1286 {
1287 return __udp4_lib_rcv(skb, udp_hash, 0);
1288 }
1289
1290 int udp_destroy_sock(struct sock *sk)
1291 {
1292 lock_sock(sk);
1293 udp_flush_pending_frames(sk);
1294 release_sock(sk);
1295 return 0;
1296 }
1297
1298 /*
1299 * Socket option code for UDP
1300 */
1301 static int do_udp_setsockopt(struct sock *sk, int level, int optname,
1302 char __user *optval, int optlen)
1303 {
1304 struct udp_sock *up = udp_sk(sk);
1305 int val;
1306 int err = 0;
1307
1308 if(optlen<sizeof(int))
1309 return -EINVAL;
1310
1311 if (get_user(val, (int __user *)optval))
1312 return -EFAULT;
1313
1314 switch(optname) {
1315 case UDP_CORK:
1316 if (val != 0) {
1317 up->corkflag = 1;
1318 } else {
1319 up->corkflag = 0;
1320 lock_sock(sk);
1321 udp_push_pending_frames(sk, up);
1322 release_sock(sk);
1323 }
1324 break;
1325
1326 case UDP_ENCAP:
1327 switch (val) {
1328 case 0:
1329 case UDP_ENCAP_ESPINUDP:
1330 case UDP_ENCAP_ESPINUDP_NON_IKE:
1331 up->encap_type = val;
1332 break;
1333 default:
1334 err = -ENOPROTOOPT;
1335 break;
1336 }
1337 break;
1338
1339 /*
1340 * UDP-Lite's partial checksum coverage (RFC 3828).
1341 */
1342 /* The sender sets actual checksum coverage length via this option.
1343 * The case coverage > packet length is handled by send module. */
1344 case UDPLITE_SEND_CSCOV:
1345 if (!up->pcflag) /* Disable the option on UDP sockets */
1346 return -ENOPROTOOPT;
1347 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
1348 val = 8;
1349 up->pcslen = val;
1350 up->pcflag |= UDPLITE_SEND_CC;
1351 break;
1352
1353 /* The receiver specifies a minimum checksum coverage value. To make
1354 * sense, this should be set to at least 8 (as done below). If zero is
1355 * used, this again means full checksum coverage. */
1356 case UDPLITE_RECV_CSCOV:
1357 if (!up->pcflag) /* Disable the option on UDP sockets */
1358 return -ENOPROTOOPT;
1359 if (val != 0 && val < 8) /* Avoid silly minimal values. */
1360 val = 8;
1361 up->pcrlen = val;
1362 up->pcflag |= UDPLITE_RECV_CC;
1363 break;
1364
1365 default:
1366 err = -ENOPROTOOPT;
1367 break;
1368 };
1369
1370 return err;
1371 }
1372
1373 int udp_setsockopt(struct sock *sk, int level, int optname,
1374 char __user *optval, int optlen)
1375 {
1376 if (level == SOL_UDP || level == SOL_UDPLITE)
1377 return do_udp_setsockopt(sk, level, optname, optval, optlen);
1378 return ip_setsockopt(sk, level, optname, optval, optlen);
1379 }
1380
1381 #ifdef CONFIG_COMPAT
1382 int compat_udp_setsockopt(struct sock *sk, int level, int optname,
1383 char __user *optval, int optlen)
1384 {
1385 if (level == SOL_UDP || level == SOL_UDPLITE)
1386 return do_udp_setsockopt(sk, level, optname, optval, optlen);
1387 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
1388 }
1389 #endif
1390
1391 static int do_udp_getsockopt(struct sock *sk, int level, int optname,
1392 char __user *optval, int __user *optlen)
1393 {
1394 struct udp_sock *up = udp_sk(sk);
1395 int val, len;
1396
1397 if(get_user(len,optlen))
1398 return -EFAULT;
1399
1400 len = min_t(unsigned int, len, sizeof(int));
1401
1402 if(len < 0)
1403 return -EINVAL;
1404
1405 switch(optname) {
1406 case UDP_CORK:
1407 val = up->corkflag;
1408 break;
1409
1410 case UDP_ENCAP:
1411 val = up->encap_type;
1412 break;
1413
1414 /* The following two cannot be changed on UDP sockets, the return is
1415 * always 0 (which corresponds to the full checksum coverage of UDP). */
1416 case UDPLITE_SEND_CSCOV:
1417 val = up->pcslen;
1418 break;
1419
1420 case UDPLITE_RECV_CSCOV:
1421 val = up->pcrlen;
1422 break;
1423
1424 default:
1425 return -ENOPROTOOPT;
1426 };
1427
1428 if(put_user(len, optlen))
1429 return -EFAULT;
1430 if(copy_to_user(optval, &val,len))
1431 return -EFAULT;
1432 return 0;
1433 }
1434
1435 int udp_getsockopt(struct sock *sk, int level, int optname,
1436 char __user *optval, int __user *optlen)
1437 {
1438 if (level == SOL_UDP || level == SOL_UDPLITE)
1439 return do_udp_getsockopt(sk, level, optname, optval, optlen);
1440 return ip_getsockopt(sk, level, optname, optval, optlen);
1441 }
1442
1443 #ifdef CONFIG_COMPAT
1444 int compat_udp_getsockopt(struct sock *sk, int level, int optname,
1445 char __user *optval, int __user *optlen)
1446 {
1447 if (level == SOL_UDP || level == SOL_UDPLITE)
1448 return do_udp_getsockopt(sk, level, optname, optval, optlen);
1449 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
1450 }
1451 #endif
1452 /**
1453 * udp_poll - wait for a UDP event.
1454 * @file - file struct
1455 * @sock - socket
1456 * @wait - poll table
1457 *
1458 * This is same as datagram poll, except for the special case of
1459 * blocking sockets. If application is using a blocking fd
1460 * and a packet with checksum error is in the queue;
1461 * then it could get return from select indicating data available
1462 * but then block when reading it. Add special case code
1463 * to work around these arguably broken applications.
1464 */
1465 unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1466 {
1467 unsigned int mask = datagram_poll(file, sock, wait);
1468 struct sock *sk = sock->sk;
1469 int is_lite = IS_UDPLITE(sk);
1470
1471 /* Check for false positives due to checksum errors */
1472 if ( (mask & POLLRDNORM) &&
1473 !(file->f_flags & O_NONBLOCK) &&
1474 !(sk->sk_shutdown & RCV_SHUTDOWN)){
1475 struct sk_buff_head *rcvq = &sk->sk_receive_queue;
1476 struct sk_buff *skb;
1477
1478 spin_lock_bh(&rcvq->lock);
1479 while ((skb = skb_peek(rcvq)) != NULL) {
1480 if (udp_lib_checksum_complete(skb)) {
1481 UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_lite);
1482 __skb_unlink(skb, rcvq);
1483 kfree_skb(skb);
1484 } else {
1485 skb->ip_summed = CHECKSUM_UNNECESSARY;
1486 break;
1487 }
1488 }
1489 spin_unlock_bh(&rcvq->lock);
1490
1491 /* nothing to see, move along */
1492 if (skb == NULL)
1493 mask &= ~(POLLIN | POLLRDNORM);
1494 }
1495
1496 return mask;
1497
1498 }
1499
1500 struct proto udp_prot = {
1501 .name = "UDP",
1502 .owner = THIS_MODULE,
1503 .close = udp_lib_close,
1504 .connect = ip4_datagram_connect,
1505 .disconnect = udp_disconnect,
1506 .ioctl = udp_ioctl,
1507 .destroy = udp_destroy_sock,
1508 .setsockopt = udp_setsockopt,
1509 .getsockopt = udp_getsockopt,
1510 .sendmsg = udp_sendmsg,
1511 .recvmsg = udp_recvmsg,
1512 .sendpage = udp_sendpage,
1513 .backlog_rcv = udp_queue_rcv_skb,
1514 .hash = udp_lib_hash,
1515 .unhash = udp_lib_unhash,
1516 .get_port = udp_v4_get_port,
1517 .obj_size = sizeof(struct udp_sock),
1518 #ifdef CONFIG_COMPAT
1519 .compat_setsockopt = compat_udp_setsockopt,
1520 .compat_getsockopt = compat_udp_getsockopt,
1521 #endif
1522 };
1523
1524 /* ------------------------------------------------------------------------ */
1525 #ifdef CONFIG_PROC_FS
1526
1527 static struct sock *udp_get_first(struct seq_file *seq)
1528 {
1529 struct sock *sk;
1530 struct udp_iter_state *state = seq->private;
1531
1532 for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) {
1533 struct hlist_node *node;
1534 sk_for_each(sk, node, state->hashtable + state->bucket) {
1535 if (sk->sk_family == state->family)
1536 goto found;
1537 }
1538 }
1539 sk = NULL;
1540 found:
1541 return sk;
1542 }
1543
1544 static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
1545 {
1546 struct udp_iter_state *state = seq->private;
1547
1548 do {
1549 sk = sk_next(sk);
1550 try_again:
1551 ;
1552 } while (sk && sk->sk_family != state->family);
1553
1554 if (!sk && ++state->bucket < UDP_HTABLE_SIZE) {
1555 sk = sk_head(state->hashtable + state->bucket);
1556 goto try_again;
1557 }
1558 return sk;
1559 }
1560
1561 static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
1562 {
1563 struct sock *sk = udp_get_first(seq);
1564
1565 if (sk)
1566 while(pos && (sk = udp_get_next(seq, sk)) != NULL)
1567 --pos;
1568 return pos ? NULL : sk;
1569 }
1570
1571 static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
1572 {
1573 read_lock(&udp_hash_lock);
1574 return *pos ? udp_get_idx(seq, *pos-1) : (void *)1;
1575 }
1576
1577 static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1578 {
1579 struct sock *sk;
1580
1581 if (v == (void *)1)
1582 sk = udp_get_idx(seq, 0);
1583 else
1584 sk = udp_get_next(seq, v);
1585
1586 ++*pos;
1587 return sk;
1588 }
1589
1590 static void udp_seq_stop(struct seq_file *seq, void *v)
1591 {
1592 read_unlock(&udp_hash_lock);
1593 }
1594
1595 static int udp_seq_open(struct inode *inode, struct file *file)
1596 {
1597 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
1598 struct seq_file *seq;
1599 int rc = -ENOMEM;
1600 struct udp_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
1601
1602 if (!s)
1603 goto out;
1604 s->family = afinfo->family;
1605 s->hashtable = afinfo->hashtable;
1606 s->seq_ops.start = udp_seq_start;
1607 s->seq_ops.next = udp_seq_next;
1608 s->seq_ops.show = afinfo->seq_show;
1609 s->seq_ops.stop = udp_seq_stop;
1610
1611 rc = seq_open(file, &s->seq_ops);
1612 if (rc)
1613 goto out_kfree;
1614
1615 seq = file->private_data;
1616 seq->private = s;
1617 out:
1618 return rc;
1619 out_kfree:
1620 kfree(s);
1621 goto out;
1622 }
1623
1624 /* ------------------------------------------------------------------------ */
1625 int udp_proc_register(struct udp_seq_afinfo *afinfo)
1626 {
1627 struct proc_dir_entry *p;
1628 int rc = 0;
1629
1630 if (!afinfo)
1631 return -EINVAL;
1632 afinfo->seq_fops->owner = afinfo->owner;
1633 afinfo->seq_fops->open = udp_seq_open;
1634 afinfo->seq_fops->read = seq_read;
1635 afinfo->seq_fops->llseek = seq_lseek;
1636 afinfo->seq_fops->release = seq_release_private;
1637
1638 p = proc_net_fops_create(afinfo->name, S_IRUGO, afinfo->seq_fops);
1639 if (p)
1640 p->data = afinfo;
1641 else
1642 rc = -ENOMEM;
1643 return rc;
1644 }
1645
1646 void udp_proc_unregister(struct udp_seq_afinfo *afinfo)
1647 {
1648 if (!afinfo)
1649 return;
1650 proc_net_remove(afinfo->name);
1651 memset(afinfo->seq_fops, 0, sizeof(*afinfo->seq_fops));
1652 }
1653
1654 /* ------------------------------------------------------------------------ */
1655 static void udp4_format_sock(struct sock *sp, char *tmpbuf, int bucket)
1656 {
1657 struct inet_sock *inet = inet_sk(sp);
1658 __be32 dest = inet->daddr;
1659 __be32 src = inet->rcv_saddr;
1660 __u16 destp = ntohs(inet->dport);
1661 __u16 srcp = ntohs(inet->sport);
1662
1663 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
1664 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
1665 bucket, src, srcp, dest, destp, sp->sk_state,
1666 atomic_read(&sp->sk_wmem_alloc),
1667 atomic_read(&sp->sk_rmem_alloc),
1668 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
1669 atomic_read(&sp->sk_refcnt), sp);
1670 }
1671
1672 int udp4_seq_show(struct seq_file *seq, void *v)
1673 {
1674 if (v == SEQ_START_TOKEN)
1675 seq_printf(seq, "%-127s\n",
1676 " sl local_address rem_address st tx_queue "
1677 "rx_queue tr tm->when retrnsmt uid timeout "
1678 "inode");
1679 else {
1680 char tmpbuf[129];
1681 struct udp_iter_state *state = seq->private;
1682
1683 udp4_format_sock(v, tmpbuf, state->bucket);
1684 seq_printf(seq, "%-127s\n", tmpbuf);
1685 }
1686 return 0;
1687 }
1688
1689 /* ------------------------------------------------------------------------ */
1690 static struct file_operations udp4_seq_fops;
1691 static struct udp_seq_afinfo udp4_seq_afinfo = {
1692 .owner = THIS_MODULE,
1693 .name = "udp",
1694 .family = AF_INET,
1695 .hashtable = udp_hash,
1696 .seq_show = udp4_seq_show,
1697 .seq_fops = &udp4_seq_fops,
1698 };
1699
1700 int __init udp4_proc_init(void)
1701 {
1702 return udp_proc_register(&udp4_seq_afinfo);
1703 }
1704
1705 void udp4_proc_exit(void)
1706 {
1707 udp_proc_unregister(&udp4_seq_afinfo);
1708 }
1709 #endif /* CONFIG_PROC_FS */
1710
1711 EXPORT_SYMBOL(udp_disconnect);
1712 EXPORT_SYMBOL(udp_hash);
1713 EXPORT_SYMBOL(udp_hash_lock);
1714 EXPORT_SYMBOL(udp_ioctl);
1715 EXPORT_SYMBOL(udp_get_port);
1716 EXPORT_SYMBOL(udp_prot);
1717 EXPORT_SYMBOL(udp_sendmsg);
1718 EXPORT_SYMBOL(udp_poll);
1719
1720 #ifdef CONFIG_PROC_FS
1721 EXPORT_SYMBOL(udp_proc_register);
1722 EXPORT_SYMBOL(udp_proc_unregister);
1723 #endif
kernel-based virtual machine