?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[UDP]: Use __skb_pull since we have checked it won't fail with pskb_may_pull
[pohmelfs.git] / net / ipv4 / udp.c
blob1bbf5510cf3a9185cbb2d64a05b8e9cf4e49d1ba
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 int type = skb->h.icmph->type;
334 int code = skb->h.icmph->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 = skb->h.uh;
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_offset = offsetof(struct udphdr, check);
431 uh->check = ~csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, 0);
432 } else {
433 /*
434 * HW-checksum won't work as there are two or more
435 * fragments on the socket so that all csums of sk_buffs
436 * should be together
437 */
438 offset = skb->h.raw - skb->data;
439 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
440
441 skb->ip_summed = CHECKSUM_NONE;
442
443 skb_queue_walk(&sk->sk_write_queue, skb) {
444 csum = csum_add(csum, skb->csum);
445 }
446
447 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
448 if (uh->check == 0)
449 uh->check = CSUM_MANGLED_0;
450 }
451 }
452
453 /*
454 * Push out all pending data as one UDP datagram. Socket is locked.
455 */
456 static int udp_push_pending_frames(struct sock *sk)
457 {
458 struct udp_sock *up = udp_sk(sk);
459 struct inet_sock *inet = inet_sk(sk);
460 struct flowi *fl = &inet->cork.fl;
461 struct sk_buff *skb;
462 struct udphdr *uh;
463 int err = 0;
464 __wsum csum = 0;
465
466 /* Grab the skbuff where UDP header space exists. */
467 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
468 goto out;
469
470 /*
471 * Create a UDP header
472 */
473 uh = skb->h.uh;
474 uh->source = fl->fl_ip_sport;
475 uh->dest = fl->fl_ip_dport;
476 uh->len = htons(up->len);
477 uh->check = 0;
478
479 if (up->pcflag) /* UDP-Lite */
480 csum = udplite_csum_outgoing(sk, skb);
481
482 else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */
483
484 skb->ip_summed = CHECKSUM_NONE;
485 goto send;
486
487 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
488
489 udp4_hwcsum_outgoing(sk, skb, fl->fl4_src,fl->fl4_dst, up->len);
490 goto send;
491
492 } else /* `normal' UDP */
493 csum = udp_csum_outgoing(sk, skb);
494
495 /* add protocol-dependent pseudo-header */
496 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, up->len,
497 sk->sk_protocol, csum );
498 if (uh->check == 0)
499 uh->check = CSUM_MANGLED_0;
500
501 send:
502 err = ip_push_pending_frames(sk);
503 out:
504 up->len = 0;
505 up->pending = 0;
506 return err;
507 }
508
509 int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
510 size_t len)
511 {
512 struct inet_sock *inet = inet_sk(sk);
513 struct udp_sock *up = udp_sk(sk);
514 int ulen = len;
515 struct ipcm_cookie ipc;
516 struct rtable *rt = NULL;
517 int free = 0;
518 int connected = 0;
519 __be32 daddr, faddr, saddr;
520 __be16 dport;
521 u8 tos;
522 int err, is_udplite = up->pcflag;
523 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
524 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
525
526 if (len > 0xFFFF)
527 return -EMSGSIZE;
528
529 /*
530 * Check the flags.
531 */
532
533 if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */
534 return -EOPNOTSUPP;
535
536 ipc.opt = NULL;
537
538 if (up->pending) {
539 /*
540 * There are pending frames.
541 * The socket lock must be held while it's corked.
542 */
543 lock_sock(sk);
544 if (likely(up->pending)) {
545 if (unlikely(up->pending != AF_INET)) {
546 release_sock(sk);
547 return -EINVAL;
548 }
549 goto do_append_data;
550 }
551 release_sock(sk);
552 }
553 ulen += sizeof(struct udphdr);
554
555 /*
556 * Get and verify the address.
557 */
558 if (msg->msg_name) {
559 struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name;
560 if (msg->msg_namelen < sizeof(*usin))
561 return -EINVAL;
562 if (usin->sin_family != AF_INET) {
563 if (usin->sin_family != AF_UNSPEC)
564 return -EAFNOSUPPORT;
565 }
566
567 daddr = usin->sin_addr.s_addr;
568 dport = usin->sin_port;
569 if (dport == 0)
570 return -EINVAL;
571 } else {
572 if (sk->sk_state != TCP_ESTABLISHED)
573 return -EDESTADDRREQ;
574 daddr = inet->daddr;
575 dport = inet->dport;
576 /* Open fast path for connected socket.
577 Route will not be used, if at least one option is set.
578 */
579 connected = 1;
580 }
581 ipc.addr = inet->saddr;
582
583 ipc.oif = sk->sk_bound_dev_if;
584 if (msg->msg_controllen) {
585 err = ip_cmsg_send(msg, &ipc);
586 if (err)
587 return err;
588 if (ipc.opt)
589 free = 1;
590 connected = 0;
591 }
592 if (!ipc.opt)
593 ipc.opt = inet->opt;
594
595 saddr = ipc.addr;
596 ipc.addr = faddr = daddr;
597
598 if (ipc.opt && ipc.opt->srr) {
599 if (!daddr)
600 return -EINVAL;
601 faddr = ipc.opt->faddr;
602 connected = 0;
603 }
604 tos = RT_TOS(inet->tos);
605 if (sock_flag(sk, SOCK_LOCALROUTE) ||
606 (msg->msg_flags & MSG_DONTROUTE) ||
607 (ipc.opt && ipc.opt->is_strictroute)) {
608 tos |= RTO_ONLINK;
609 connected = 0;
610 }
611
612 if (MULTICAST(daddr)) {
613 if (!ipc.oif)
614 ipc.oif = inet->mc_index;
615 if (!saddr)
616 saddr = inet->mc_addr;
617 connected = 0;
618 }
619
620 if (connected)
621 rt = (struct rtable*)sk_dst_check(sk, 0);
622
623 if (rt == NULL) {
624 struct flowi fl = { .oif = ipc.oif,
625 .nl_u = { .ip4_u =
626 { .daddr = faddr,
627 .saddr = saddr,
628 .tos = tos } },
629 .proto = sk->sk_protocol,
630 .uli_u = { .ports =
631 { .sport = inet->sport,
632 .dport = dport } } };
633 security_sk_classify_flow(sk, &fl);
634 err = ip_route_output_flow(&rt, &fl, sk, 1);
635 if (err)
636 goto out;
637
638 err = -EACCES;
639 if ((rt->rt_flags & RTCF_BROADCAST) &&
640 !sock_flag(sk, SOCK_BROADCAST))
641 goto out;
642 if (connected)
643 sk_dst_set(sk, dst_clone(&rt->u.dst));
644 }
645
646 if (msg->msg_flags&MSG_CONFIRM)
647 goto do_confirm;
648 back_from_confirm:
649
650 saddr = rt->rt_src;
651 if (!ipc.addr)
652 daddr = ipc.addr = rt->rt_dst;
653
654 lock_sock(sk);
655 if (unlikely(up->pending)) {
656 /* The socket is already corked while preparing it. */
657 /* ... which is an evident application bug. --ANK */
658 release_sock(sk);
659
660 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
661 err = -EINVAL;
662 goto out;
663 }
664 /*
665 * Now cork the socket to pend data.
666 */
667 inet->cork.fl.fl4_dst = daddr;
668 inet->cork.fl.fl_ip_dport = dport;
669 inet->cork.fl.fl4_src = saddr;
670 inet->cork.fl.fl_ip_sport = inet->sport;
671 up->pending = AF_INET;
672
673 do_append_data:
674 up->len += ulen;
675 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
676 err = ip_append_data(sk, getfrag, msg->msg_iov, ulen,
677 sizeof(struct udphdr), &ipc, rt,
678 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
679 if (err)
680 udp_flush_pending_frames(sk);
681 else if (!corkreq)
682 err = udp_push_pending_frames(sk);
683 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
684 up->pending = 0;
685 release_sock(sk);
686
687 out:
688 ip_rt_put(rt);
689 if (free)
690 kfree(ipc.opt);
691 if (!err) {
692 UDP_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS, is_udplite);
693 return len;
694 }
695 /*
696 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
697 * ENOBUFS might not be good (it's not tunable per se), but otherwise
698 * we don't have a good statistic (IpOutDiscards but it can be too many
699 * things). We could add another new stat but at least for now that
700 * seems like overkill.
701 */
702 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
703 UDP_INC_STATS_USER(UDP_MIB_SNDBUFERRORS, is_udplite);
704 }
705 return err;
706
707 do_confirm:
708 dst_confirm(&rt->u.dst);
709 if (!(msg->msg_flags&MSG_PROBE) || len)
710 goto back_from_confirm;
711 err = 0;
712 goto out;
713 }
714
715 int udp_sendpage(struct sock *sk, struct page *page, int offset,
716 size_t size, int flags)
717 {
718 struct udp_sock *up = udp_sk(sk);
719 int ret;
720
721 if (!up->pending) {
722 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
723
724 /* Call udp_sendmsg to specify destination address which
725 * sendpage interface can't pass.
726 * This will succeed only when the socket is connected.
727 */
728 ret = udp_sendmsg(NULL, sk, &msg, 0);
729 if (ret < 0)
730 return ret;
731 }
732
733 lock_sock(sk);
734
735 if (unlikely(!up->pending)) {
736 release_sock(sk);
737
738 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n");
739 return -EINVAL;
740 }
741
742 ret = ip_append_page(sk, page, offset, size, flags);
743 if (ret == -EOPNOTSUPP) {
744 release_sock(sk);
745 return sock_no_sendpage(sk->sk_socket, page, offset,
746 size, flags);
747 }
748 if (ret < 0) {
749 udp_flush_pending_frames(sk);
750 goto out;
751 }
752
753 up->len += size;
754 if (!(up->corkflag || (flags&MSG_MORE)))
755 ret = udp_push_pending_frames(sk);
756 if (!ret)
757 ret = size;
758 out:
759 release_sock(sk);
760 return ret;
761 }
762
763 /*
764 * IOCTL requests applicable to the UDP protocol
765 */
766
767 int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
768 {
769 switch (cmd) {
770 case SIOCOUTQ:
771 {
772 int amount = atomic_read(&sk->sk_wmem_alloc);
773 return put_user(amount, (int __user *)arg);
774 }
775
776 case SIOCINQ:
777 {
778 struct sk_buff *skb;
779 unsigned long amount;
780
781 amount = 0;
782 spin_lock_bh(&sk->sk_receive_queue.lock);
783 skb = skb_peek(&sk->sk_receive_queue);
784 if (skb != NULL) {
785 /*
786 * We will only return the amount
787 * of this packet since that is all
788 * that will be read.
789 */
790 amount = skb->len - sizeof(struct udphdr);
791 }
792 spin_unlock_bh(&sk->sk_receive_queue.lock);
793 return put_user(amount, (int __user *)arg);
794 }
795
796 default:
797 return -ENOIOCTLCMD;
798 }
799
800 return 0;
801 }
802
803 /*
804 * This should be easy, if there is something there we
805 * return it, otherwise we block.
806 */
807
808 int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
809 size_t len, int noblock, int flags, int *addr_len)
810 {
811 struct inet_sock *inet = inet_sk(sk);
812 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
813 struct sk_buff *skb;
814 unsigned int ulen, copied;
815 int err;
816 int is_udplite = IS_UDPLITE(sk);
817
818 /*
819 * Check any passed addresses
820 */
821 if (addr_len)
822 *addr_len=sizeof(*sin);
823
824 if (flags & MSG_ERRQUEUE)
825 return ip_recv_error(sk, msg, len);
826
827 try_again:
828 skb = skb_recv_datagram(sk, flags, noblock, &err);
829 if (!skb)
830 goto out;
831
832 ulen = skb->len - sizeof(struct udphdr);
833 copied = len;
834 if (copied > ulen)
835 copied = ulen;
836 else if (copied < ulen)
837 msg->msg_flags |= MSG_TRUNC;
838
839 /*
840 * If checksum is needed at all, try to do it while copying the
841 * data. If the data is truncated, or if we only want a partial
842 * coverage checksum (UDP-Lite), do it before the copy.
843 */
844
845 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
846 if (udp_lib_checksum_complete(skb))
847 goto csum_copy_err;
848 }
849
850 if (skb->ip_summed == CHECKSUM_UNNECESSARY)
851 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
852 msg->msg_iov, copied );
853 else {
854 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
855
856 if (err == -EINVAL)
857 goto csum_copy_err;
858 }
859
860 if (err)
861 goto out_free;
862
863 sock_recv_timestamp(msg, sk, skb);
864
865 /* Copy the address. */
866 if (sin)
867 {
868 sin->sin_family = AF_INET;
869 sin->sin_port = skb->h.uh->source;
870 sin->sin_addr.s_addr = skb->nh.iph->saddr;
871 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
872 }
873 if (inet->cmsg_flags)
874 ip_cmsg_recv(msg, skb);
875
876 err = copied;
877 if (flags & MSG_TRUNC)
878 err = ulen;
879
880 out_free:
881 skb_free_datagram(sk, skb);
882 out:
883 return err;
884
885 csum_copy_err:
886 UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_udplite);
887
888 skb_kill_datagram(sk, skb, flags);
889
890 if (noblock)
891 return -EAGAIN;
892 goto try_again;
893 }
894
895
896 int udp_disconnect(struct sock *sk, int flags)
897 {
898 struct inet_sock *inet = inet_sk(sk);
899 /*
900 * 1003.1g - break association.
901 */
902
903 sk->sk_state = TCP_CLOSE;
904 inet->daddr = 0;
905 inet->dport = 0;
906 sk->sk_bound_dev_if = 0;
907 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
908 inet_reset_saddr(sk);
909
910 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
911 sk->sk_prot->unhash(sk);
912 inet->sport = 0;
913 }
914 sk_dst_reset(sk);
915 return 0;
916 }
917
918 /* return:
919 * 1 if the the UDP system should process it
920 * 0 if we should drop this packet
921 * -1 if it should get processed by xfrm4_rcv_encap
922 */
923 static int udp_encap_rcv(struct sock * sk, struct sk_buff *skb)
924 {
925 #ifndef CONFIG_XFRM
926 return 1;
927 #else
928 struct udp_sock *up = udp_sk(sk);
929 struct udphdr *uh;
930 struct iphdr *iph;
931 int iphlen, len;
932
933 __u8 *udpdata;
934 __be32 *udpdata32;
935 __u16 encap_type = up->encap_type;
936
937 /* if we're overly short, let UDP handle it */
938 len = skb->len - sizeof(struct udphdr);
939 if (len <= 0)
940 return 1;
941
942 /* if this is not encapsulated socket, then just return now */
943 if (!encap_type)
944 return 1;
945
946 /* If this is a paged skb, make sure we pull up
947 * whatever data we need to look at. */
948 if (!pskb_may_pull(skb, sizeof(struct udphdr) + min(len, 8)))
949 return 1;
950
951 /* Now we can get the pointers */
952 uh = skb->h.uh;
953 udpdata = (__u8 *)uh + sizeof(struct udphdr);
954 udpdata32 = (__be32 *)udpdata;
955
956 switch (encap_type) {
957 default:
958 case UDP_ENCAP_ESPINUDP:
959 /* Check if this is a keepalive packet. If so, eat it. */
960 if (len == 1 && udpdata[0] == 0xff) {
961 return 0;
962 } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0) {
963 /* ESP Packet without Non-ESP header */
964 len = sizeof(struct udphdr);
965 } else
966 /* Must be an IKE packet.. pass it through */
967 return 1;
968 break;
969 case UDP_ENCAP_ESPINUDP_NON_IKE:
970 /* Check if this is a keepalive packet. If so, eat it. */
971 if (len == 1 && udpdata[0] == 0xff) {
972 return 0;
973 } else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) &&
974 udpdata32[0] == 0 && udpdata32[1] == 0) {
975
976 /* ESP Packet with Non-IKE marker */
977 len = sizeof(struct udphdr) + 2 * sizeof(u32);
978 } else
979 /* Must be an IKE packet.. pass it through */
980 return 1;
981 break;
982 }
983
984 /* At this point we are sure that this is an ESPinUDP packet,
985 * so we need to remove 'len' bytes from the packet (the UDP
986 * header and optional ESP marker bytes) and then modify the
987 * protocol to ESP, and then call into the transform receiver.
988 */
989 if (skb_cloned(skb) && pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
990 return 0;
991
992 /* Now we can update and verify the packet length... */
993 iph = skb->nh.iph;
994 iphlen = iph->ihl << 2;
995 iph->tot_len = htons(ntohs(iph->tot_len) - len);
996 if (skb->len < iphlen + len) {
997 /* packet is too small!?! */
998 return 0;
999 }
1000
1001 /* pull the data buffer up to the ESP header and set the
1002 * transport header to point to ESP. Keep UDP on the stack
1003 * for later.
1004 */
1005 skb->h.raw = __skb_pull(skb, len);
1006
1007 /* modify the protocol (it's ESP!) */
1008 iph->protocol = IPPROTO_ESP;
1009
1010 /* and let the caller know to send this into the ESP processor... */
1011 return -1;
1012 #endif
1013 }
1014
1015 /* returns:
1016 * -1: error
1017 * 0: success
1018 * >0: "udp encap" protocol resubmission
1019 *
1020 * Note that in the success and error cases, the skb is assumed to
1021 * have either been requeued or freed.
1022 */
1023 int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
1024 {
1025 struct udp_sock *up = udp_sk(sk);
1026 int rc;
1027
1028 /*
1029 * Charge it to the socket, dropping if the queue is full.
1030 */
1031 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1032 goto drop;
1033 nf_reset(skb);
1034
1035 if (up->encap_type) {
1036 /*
1037 * This is an encapsulation socket, so let's see if this is
1038 * an encapsulated packet.
1039 * If it's a keepalive packet, then just eat it.
1040 * If it's an encapsulateed packet, then pass it to the
1041 * IPsec xfrm input and return the response
1042 * appropriately. Otherwise, just fall through and
1043 * pass this up the UDP socket.
1044 */
1045 int ret;
1046
1047 ret = udp_encap_rcv(sk, skb);
1048 if (ret == 0) {
1049 /* Eat the packet .. */
1050 kfree_skb(skb);
1051 return 0;
1052 }
1053 if (ret < 0) {
1054 /* process the ESP packet */
1055 ret = xfrm4_rcv_encap(skb, up->encap_type);
1056 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS, up->pcflag);
1057 return -ret;
1058 }
1059 /* FALLTHROUGH -- it's a UDP Packet */
1060 }
1061
1062 /*
1063 * UDP-Lite specific tests, ignored on UDP sockets
1064 */
1065 if ((up->pcflag & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1066
1067 /*
1068 * MIB statistics other than incrementing the error count are
1069 * disabled for the following two types of errors: these depend
1070 * on the application settings, not on the functioning of the
1071 * protocol stack as such.
1072 *
1073 * RFC 3828 here recommends (sec 3.3): "There should also be a
1074 * way ... to ... at least let the receiving application block
1075 * delivery of packets with coverage values less than a value
1076 * provided by the application."
1077 */
1078 if (up->pcrlen == 0) { /* full coverage was set */
1079 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE: partial coverage "
1080 "%d while full coverage %d requested\n",
1081 UDP_SKB_CB(skb)->cscov, skb->len);
1082 goto drop;
1083 }
1084 /* The next case involves violating the min. coverage requested
1085 * by the receiver. This is subtle: if receiver wants x and x is
1086 * greater than the buffersize/MTU then receiver will complain
1087 * that it wants x while sender emits packets of smaller size y.
1088 * Therefore the above ...()->partial_cov statement is essential.
1089 */
1090 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
1091 LIMIT_NETDEBUG(KERN_WARNING
1092 "UDPLITE: coverage %d too small, need min %d\n",
1093 UDP_SKB_CB(skb)->cscov, up->pcrlen);
1094 goto drop;
1095 }
1096 }
1097
1098 if (sk->sk_filter) {
1099 if (udp_lib_checksum_complete(skb))
1100 goto drop;
1101 }
1102
1103 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
1104 /* Note that an ENOMEM error is charged twice */
1105 if (rc == -ENOMEM)
1106 UDP_INC_STATS_BH(UDP_MIB_RCVBUFERRORS, up->pcflag);
1107 goto drop;
1108 }
1109
1110 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS, up->pcflag);
1111 return 0;
1112
1113 drop:
1114 UDP_INC_STATS_BH(UDP_MIB_INERRORS, up->pcflag);
1115 kfree_skb(skb);
1116 return -1;
1117 }
1118
1119 /*
1120 * Multicasts and broadcasts go to each listener.
1121 *
1122 * Note: called only from the BH handler context,
1123 * so we don't need to lock the hashes.
1124 */
1125 static int __udp4_lib_mcast_deliver(struct sk_buff *skb,
1126 struct udphdr *uh,
1127 __be32 saddr, __be32 daddr,
1128 struct hlist_head udptable[])
1129 {
1130 struct sock *sk;
1131 int dif;
1132
1133 read_lock(&udp_hash_lock);
1134 sk = sk_head(&udptable[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
1135 dif = skb->dev->ifindex;
1136 sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
1137 if (sk) {
1138 struct sock *sknext = NULL;
1139
1140 do {
1141 struct sk_buff *skb1 = skb;
1142
1143 sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr,
1144 uh->source, saddr, dif);
1145 if (sknext)
1146 skb1 = skb_clone(skb, GFP_ATOMIC);
1147
1148 if (skb1) {
1149 int ret = udp_queue_rcv_skb(sk, skb1);
1150 if (ret > 0)
1151 /* we should probably re-process instead
1152 * of dropping packets here. */
1153 kfree_skb(skb1);
1154 }
1155 sk = sknext;
1156 } while (sknext);
1157 } else
1158 kfree_skb(skb);
1159 read_unlock(&udp_hash_lock);
1160 return 0;
1161 }
1162
1163 /* Initialize UDP checksum. If exited with zero value (success),
1164 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1165 * Otherwise, csum completion requires chacksumming packet body,
1166 * including udp header and folding it to skb->csum.
1167 */
1168 static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1169 int proto)
1170 {
1171 int err;
1172
1173 UDP_SKB_CB(skb)->partial_cov = 0;
1174 UDP_SKB_CB(skb)->cscov = skb->len;
1175
1176 if (proto == IPPROTO_UDPLITE) {
1177 err = udplite_checksum_init(skb, uh);
1178 if (err)
1179 return err;
1180 }
1181
1182 if (uh->check == 0) {
1183 skb->ip_summed = CHECKSUM_UNNECESSARY;
1184 } else if (skb->ip_summed == CHECKSUM_COMPLETE) {
1185 if (!csum_tcpudp_magic(skb->nh.iph->saddr, skb->nh.iph->daddr,
1186 skb->len, proto, skb->csum))
1187 skb->ip_summed = CHECKSUM_UNNECESSARY;
1188 }
1189 if (skb->ip_summed != CHECKSUM_UNNECESSARY)
1190 skb->csum = csum_tcpudp_nofold(skb->nh.iph->saddr,
1191 skb->nh.iph->daddr,
1192 skb->len, proto, 0);
1193 /* Probably, we should checksum udp header (it should be in cache
1194 * in any case) and data in tiny packets (< rx copybreak).
1195 */
1196
1197 return 0;
1198 }
1199
1200 /*
1201 * All we need to do is get the socket, and then do a checksum.
1202 */
1203
1204 int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[],
1205 int proto)
1206 {
1207 struct sock *sk;
1208 struct udphdr *uh = skb->h.uh;
1209 unsigned short ulen;
1210 struct rtable *rt = (struct rtable*)skb->dst;
1211 __be32 saddr = skb->nh.iph->saddr;
1212 __be32 daddr = skb->nh.iph->daddr;
1213
1214 /*
1215 * Validate the packet.
1216 */
1217 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1218 goto drop; /* No space for header. */
1219
1220 ulen = ntohs(uh->len);
1221 if (ulen > skb->len)
1222 goto short_packet;
1223
1224 if (proto == IPPROTO_UDP) {
1225 /* UDP validates ulen. */
1226 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1227 goto short_packet;
1228 uh = skb->h.uh;
1229 }
1230
1231 if (udp4_csum_init(skb, uh, proto))
1232 goto csum_error;
1233
1234 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1235 return __udp4_lib_mcast_deliver(skb, uh, saddr, daddr, udptable);
1236
1237 sk = __udp4_lib_lookup(saddr, uh->source, daddr, uh->dest,
1238 skb->dev->ifindex, udptable );
1239
1240 if (sk != NULL) {
1241 int ret = udp_queue_rcv_skb(sk, skb);
1242 sock_put(sk);
1243
1244 /* a return value > 0 means to resubmit the input, but
1245 * it wants the return to be -protocol, or 0
1246 */
1247 if (ret > 0)
1248 return -ret;
1249 return 0;
1250 }
1251
1252 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1253 goto drop;
1254 nf_reset(skb);
1255
1256 /* No socket. Drop packet silently, if checksum is wrong */
1257 if (udp_lib_checksum_complete(skb))
1258 goto csum_error;
1259
1260 UDP_INC_STATS_BH(UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
1261 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1262
1263 /*
1264 * Hmm. We got an UDP packet to a port to which we
1265 * don't wanna listen. Ignore it.
1266 */
1267 kfree_skb(skb);
1268 return 0;
1269
1270 short_packet:
1271 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n",
1272 proto == IPPROTO_UDPLITE ? "-Lite" : "",
1273 NIPQUAD(saddr),
1274 ntohs(uh->source),
1275 ulen,
1276 skb->len,
1277 NIPQUAD(daddr),
1278 ntohs(uh->dest));
1279 goto drop;
1280
1281 csum_error:
1282 /*
1283 * RFC1122: OK. Discards the bad packet silently (as far as
1284 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1285 */
1286 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n",
1287 proto == IPPROTO_UDPLITE ? "-Lite" : "",
1288 NIPQUAD(saddr),
1289 ntohs(uh->source),
1290 NIPQUAD(daddr),
1291 ntohs(uh->dest),
1292 ulen);
1293 drop:
1294 UDP_INC_STATS_BH(UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1295 kfree_skb(skb);
1296 return 0;
1297 }
1298
1299 int udp_rcv(struct sk_buff *skb)
1300 {
1301 return __udp4_lib_rcv(skb, udp_hash, IPPROTO_UDP);
1302 }
1303
1304 int udp_destroy_sock(struct sock *sk)
1305 {
1306 lock_sock(sk);
1307 udp_flush_pending_frames(sk);
1308 release_sock(sk);
1309 return 0;
1310 }
1311
1312 /*
1313 * Socket option code for UDP
1314 */
1315 int udp_lib_setsockopt(struct sock *sk, int level, int optname,
1316 char __user *optval, int optlen,
1317 int (*push_pending_frames)(struct sock *))
1318 {
1319 struct udp_sock *up = udp_sk(sk);
1320 int val;
1321 int err = 0;
1322
1323 if (optlen<sizeof(int))
1324 return -EINVAL;
1325
1326 if (get_user(val, (int __user *)optval))
1327 return -EFAULT;
1328
1329 switch (optname) {
1330 case UDP_CORK:
1331 if (val != 0) {
1332 up->corkflag = 1;
1333 } else {
1334 up->corkflag = 0;
1335 lock_sock(sk);
1336 (*push_pending_frames)(sk);
1337 release_sock(sk);
1338 }
1339 break;
1340
1341 case UDP_ENCAP:
1342 switch (val) {
1343 case 0:
1344 case UDP_ENCAP_ESPINUDP:
1345 case UDP_ENCAP_ESPINUDP_NON_IKE:
1346 up->encap_type = val;
1347 break;
1348 default:
1349 err = -ENOPROTOOPT;
1350 break;
1351 }
1352 break;
1353
1354 /*
1355 * UDP-Lite's partial checksum coverage (RFC 3828).
1356 */
1357 /* The sender sets actual checksum coverage length via this option.
1358 * The case coverage > packet length is handled by send module. */
1359 case UDPLITE_SEND_CSCOV:
1360 if (!up->pcflag) /* Disable the option on UDP sockets */
1361 return -ENOPROTOOPT;
1362 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
1363 val = 8;
1364 up->pcslen = val;
1365 up->pcflag |= UDPLITE_SEND_CC;
1366 break;
1367
1368 /* The receiver specifies a minimum checksum coverage value. To make
1369 * sense, this should be set to at least 8 (as done below). If zero is
1370 * used, this again means full checksum coverage. */
1371 case UDPLITE_RECV_CSCOV:
1372 if (!up->pcflag) /* Disable the option on UDP sockets */
1373 return -ENOPROTOOPT;
1374 if (val != 0 && val < 8) /* Avoid silly minimal values. */
1375 val = 8;
1376 up->pcrlen = val;
1377 up->pcflag |= UDPLITE_RECV_CC;
1378 break;
1379
1380 default:
1381 err = -ENOPROTOOPT;
1382 break;
1383 }
1384
1385 return err;
1386 }
1387
1388 int udp_setsockopt(struct sock *sk, int level, int optname,
1389 char __user *optval, int optlen)
1390 {
1391 if (level == SOL_UDP || level == SOL_UDPLITE)
1392 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1393 udp_push_pending_frames);
1394 return ip_setsockopt(sk, level, optname, optval, optlen);
1395 }
1396
1397 #ifdef CONFIG_COMPAT
1398 int compat_udp_setsockopt(struct sock *sk, int level, int optname,
1399 char __user *optval, int optlen)
1400 {
1401 if (level == SOL_UDP || level == SOL_UDPLITE)
1402 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1403 udp_push_pending_frames);
1404 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
1405 }
1406 #endif
1407
1408 int udp_lib_getsockopt(struct sock *sk, int level, int optname,
1409 char __user *optval, int __user *optlen)
1410 {
1411 struct udp_sock *up = udp_sk(sk);
1412 int val, len;
1413
1414 if (get_user(len,optlen))
1415 return -EFAULT;
1416
1417 len = min_t(unsigned int, len, sizeof(int));
1418
1419 if (len < 0)
1420 return -EINVAL;
1421
1422 switch (optname) {
1423 case UDP_CORK:
1424 val = up->corkflag;
1425 break;
1426
1427 case UDP_ENCAP:
1428 val = up->encap_type;
1429 break;
1430
1431 /* The following two cannot be changed on UDP sockets, the return is
1432 * always 0 (which corresponds to the full checksum coverage of UDP). */
1433 case UDPLITE_SEND_CSCOV:
1434 val = up->pcslen;
1435 break;
1436
1437 case UDPLITE_RECV_CSCOV:
1438 val = up->pcrlen;
1439 break;
1440
1441 default:
1442 return -ENOPROTOOPT;
1443 }
1444
1445 if (put_user(len, optlen))
1446 return -EFAULT;
1447 if (copy_to_user(optval, &val,len))
1448 return -EFAULT;
1449 return 0;
1450 }
1451
1452 int udp_getsockopt(struct sock *sk, int level, int optname,
1453 char __user *optval, int __user *optlen)
1454 {
1455 if (level == SOL_UDP || level == SOL_UDPLITE)
1456 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1457 return ip_getsockopt(sk, level, optname, optval, optlen);
1458 }
1459
1460 #ifdef CONFIG_COMPAT
1461 int compat_udp_getsockopt(struct sock *sk, int level, int optname,
1462 char __user *optval, int __user *optlen)
1463 {
1464 if (level == SOL_UDP || level == SOL_UDPLITE)
1465 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1466 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
1467 }
1468 #endif
1469 /**
1470 * udp_poll - wait for a UDP event.
1471 * @file - file struct
1472 * @sock - socket
1473 * @wait - poll table
1474 *
1475 * This is same as datagram poll, except for the special case of
1476 * blocking sockets. If application is using a blocking fd
1477 * and a packet with checksum error is in the queue;
1478 * then it could get return from select indicating data available
1479 * but then block when reading it. Add special case code
1480 * to work around these arguably broken applications.
1481 */
1482 unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1483 {
1484 unsigned int mask = datagram_poll(file, sock, wait);
1485 struct sock *sk = sock->sk;
1486 int is_lite = IS_UDPLITE(sk);
1487
1488 /* Check for false positives due to checksum errors */
1489 if ( (mask & POLLRDNORM) &&
1490 !(file->f_flags & O_NONBLOCK) &&
1491 !(sk->sk_shutdown & RCV_SHUTDOWN)){
1492 struct sk_buff_head *rcvq = &sk->sk_receive_queue;
1493 struct sk_buff *skb;
1494
1495 spin_lock_bh(&rcvq->lock);
1496 while ((skb = skb_peek(rcvq)) != NULL &&
1497 udp_lib_checksum_complete(skb)) {
1498 UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_lite);
1499 __skb_unlink(skb, rcvq);
1500 kfree_skb(skb);
1501 }
1502 spin_unlock_bh(&rcvq->lock);
1503
1504 /* nothing to see, move along */
1505 if (skb == NULL)
1506 mask &= ~(POLLIN | POLLRDNORM);
1507 }
1508
1509 return mask;
1510
1511 }
1512
1513 struct proto udp_prot = {
1514 .name = "UDP",
1515 .owner = THIS_MODULE,
1516 .close = udp_lib_close,
1517 .connect = ip4_datagram_connect,
1518 .disconnect = udp_disconnect,
1519 .ioctl = udp_ioctl,
1520 .destroy = udp_destroy_sock,
1521 .setsockopt = udp_setsockopt,
1522 .getsockopt = udp_getsockopt,
1523 .sendmsg = udp_sendmsg,
1524 .recvmsg = udp_recvmsg,
1525 .sendpage = udp_sendpage,
1526 .backlog_rcv = udp_queue_rcv_skb,
1527 .hash = udp_lib_hash,
1528 .unhash = udp_lib_unhash,
1529 .get_port = udp_v4_get_port,
1530 .obj_size = sizeof(struct udp_sock),
1531 #ifdef CONFIG_COMPAT
1532 .compat_setsockopt = compat_udp_setsockopt,
1533 .compat_getsockopt = compat_udp_getsockopt,
1534 #endif
1535 };
1536
1537 /* ------------------------------------------------------------------------ */
1538 #ifdef CONFIG_PROC_FS
1539
1540 static struct sock *udp_get_first(struct seq_file *seq)
1541 {
1542 struct sock *sk;
1543 struct udp_iter_state *state = seq->private;
1544
1545 for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) {
1546 struct hlist_node *node;
1547 sk_for_each(sk, node, state->hashtable + state->bucket) {
1548 if (sk->sk_family == state->family)
1549 goto found;
1550 }
1551 }
1552 sk = NULL;
1553 found:
1554 return sk;
1555 }
1556
1557 static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
1558 {
1559 struct udp_iter_state *state = seq->private;
1560
1561 do {
1562 sk = sk_next(sk);
1563 try_again:
1564 ;
1565 } while (sk && sk->sk_family != state->family);
1566
1567 if (!sk && ++state->bucket < UDP_HTABLE_SIZE) {
1568 sk = sk_head(state->hashtable + state->bucket);
1569 goto try_again;
1570 }
1571 return sk;
1572 }
1573
1574 static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
1575 {
1576 struct sock *sk = udp_get_first(seq);
1577
1578 if (sk)
1579 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
1580 --pos;
1581 return pos ? NULL : sk;
1582 }
1583
1584 static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
1585 {
1586 read_lock(&udp_hash_lock);
1587 return *pos ? udp_get_idx(seq, *pos-1) : (void *)1;
1588 }
1589
1590 static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1591 {
1592 struct sock *sk;
1593
1594 if (v == (void *)1)
1595 sk = udp_get_idx(seq, 0);
1596 else
1597 sk = udp_get_next(seq, v);
1598
1599 ++*pos;
1600 return sk;
1601 }
1602
1603 static void udp_seq_stop(struct seq_file *seq, void *v)
1604 {
1605 read_unlock(&udp_hash_lock);
1606 }
1607
1608 static int udp_seq_open(struct inode *inode, struct file *file)
1609 {
1610 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
1611 struct seq_file *seq;
1612 int rc = -ENOMEM;
1613 struct udp_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
1614
1615 if (!s)
1616 goto out;
1617 s->family = afinfo->family;
1618 s->hashtable = afinfo->hashtable;
1619 s->seq_ops.start = udp_seq_start;
1620 s->seq_ops.next = udp_seq_next;
1621 s->seq_ops.show = afinfo->seq_show;
1622 s->seq_ops.stop = udp_seq_stop;
1623
1624 rc = seq_open(file, &s->seq_ops);
1625 if (rc)
1626 goto out_kfree;
1627
1628 seq = file->private_data;
1629 seq->private = s;
1630 out:
1631 return rc;
1632 out_kfree:
1633 kfree(s);
1634 goto out;
1635 }
1636
1637 /* ------------------------------------------------------------------------ */
1638 int udp_proc_register(struct udp_seq_afinfo *afinfo)
1639 {
1640 struct proc_dir_entry *p;
1641 int rc = 0;
1642
1643 if (!afinfo)
1644 return -EINVAL;
1645 afinfo->seq_fops->owner = afinfo->owner;
1646 afinfo->seq_fops->open = udp_seq_open;
1647 afinfo->seq_fops->read = seq_read;
1648 afinfo->seq_fops->llseek = seq_lseek;
1649 afinfo->seq_fops->release = seq_release_private;
1650
1651 p = proc_net_fops_create(afinfo->name, S_IRUGO, afinfo->seq_fops);
1652 if (p)
1653 p->data = afinfo;
1654 else
1655 rc = -ENOMEM;
1656 return rc;
1657 }
1658
1659 void udp_proc_unregister(struct udp_seq_afinfo *afinfo)
1660 {
1661 if (!afinfo)
1662 return;
1663 proc_net_remove(afinfo->name);
1664 memset(afinfo->seq_fops, 0, sizeof(*afinfo->seq_fops));
1665 }
1666
1667 /* ------------------------------------------------------------------------ */
1668 static void udp4_format_sock(struct sock *sp, char *tmpbuf, int bucket)
1669 {
1670 struct inet_sock *inet = inet_sk(sp);
1671 __be32 dest = inet->daddr;
1672 __be32 src = inet->rcv_saddr;
1673 __u16 destp = ntohs(inet->dport);
1674 __u16 srcp = ntohs(inet->sport);
1675
1676 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
1677 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
1678 bucket, src, srcp, dest, destp, sp->sk_state,
1679 atomic_read(&sp->sk_wmem_alloc),
1680 atomic_read(&sp->sk_rmem_alloc),
1681 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
1682 atomic_read(&sp->sk_refcnt), sp);
1683 }
1684
1685 int udp4_seq_show(struct seq_file *seq, void *v)
1686 {
1687 if (v == SEQ_START_TOKEN)
1688 seq_printf(seq, "%-127s\n",
1689 " sl local_address rem_address st tx_queue "
1690 "rx_queue tr tm->when retrnsmt uid timeout "
1691 "inode");
1692 else {
1693 char tmpbuf[129];
1694 struct udp_iter_state *state = seq->private;
1695
1696 udp4_format_sock(v, tmpbuf, state->bucket);
1697 seq_printf(seq, "%-127s\n", tmpbuf);
1698 }
1699 return 0;
1700 }
1701
1702 /* ------------------------------------------------------------------------ */
1703 static struct file_operations udp4_seq_fops;
1704 static struct udp_seq_afinfo udp4_seq_afinfo = {
1705 .owner = THIS_MODULE,
1706 .name = "udp",
1707 .family = AF_INET,
1708 .hashtable = udp_hash,
1709 .seq_show = udp4_seq_show,
1710 .seq_fops = &udp4_seq_fops,
1711 };
1712
1713 int __init udp4_proc_init(void)
1714 {
1715 return udp_proc_register(&udp4_seq_afinfo);
1716 }
1717
1718 void udp4_proc_exit(void)
1719 {
1720 udp_proc_unregister(&udp4_seq_afinfo);
1721 }
1722 #endif /* CONFIG_PROC_FS */
1723
1724 EXPORT_SYMBOL(udp_disconnect);
1725 EXPORT_SYMBOL(udp_hash);
1726 EXPORT_SYMBOL(udp_hash_lock);
1727 EXPORT_SYMBOL(udp_ioctl);
1728 EXPORT_SYMBOL(udp_get_port);
1729 EXPORT_SYMBOL(udp_prot);
1730 EXPORT_SYMBOL(udp_sendmsg);
1731 EXPORT_SYMBOL(udp_lib_getsockopt);
1732 EXPORT_SYMBOL(udp_lib_setsockopt);
1733 EXPORT_SYMBOL(udp_poll);
1734
1735 #ifdef CONFIG_PROC_FS
1736 EXPORT_SYMBOL(udp_proc_register);
1737 EXPORT_SYMBOL(udp_proc_unregister);
1738 #endif