?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[XFS] Kill off xfs_count_bits
[linux-2.6.git] / net / ipv4 / udp.c
blobfacb7e29304e7df9a7a172d2e4ad2f63acfa1074
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * The User Datagram Protocol (UDP).
7 *
8 * Version: $Id: udp.c,v 1.102 2002/02/01 22:01:04 davem Exp $
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
13 * Alan Cox, <Alan.Cox@linux.org>
14 * Hirokazu Takahashi, <taka@valinux.co.jp>
15 *
16 * Fixes:
17 * Alan Cox : verify_area() calls
18 * Alan Cox : stopped close while in use off icmp
19 * messages. Not a fix but a botch that
20 * for udp at least is 'valid'.
21 * Alan Cox : Fixed icmp handling properly
22 * Alan Cox : Correct error for oversized datagrams
23 * Alan Cox : Tidied select() semantics.
24 * Alan Cox : udp_err() fixed properly, also now
25 * select and read wake correctly on errors
26 * Alan Cox : udp_send verify_area moved to avoid mem leak
27 * Alan Cox : UDP can count its memory
28 * Alan Cox : send to an unknown connection causes
29 * an ECONNREFUSED off the icmp, but
30 * does NOT close.
31 * Alan Cox : Switched to new sk_buff handlers. No more backlog!
32 * Alan Cox : Using generic datagram code. Even smaller and the PEEK
33 * bug no longer crashes it.
34 * Fred Van Kempen : Net2e support for sk->broadcast.
35 * Alan Cox : Uses skb_free_datagram
36 * Alan Cox : Added get/set sockopt support.
37 * Alan Cox : Broadcasting without option set returns EACCES.
38 * Alan Cox : No wakeup calls. Instead we now use the callbacks.
39 * Alan Cox : Use ip_tos and ip_ttl
40 * Alan Cox : SNMP Mibs
41 * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support.
42 * Matt Dillon : UDP length checks.
43 * Alan Cox : Smarter af_inet used properly.
44 * Alan Cox : Use new kernel side addressing.
45 * Alan Cox : Incorrect return on truncated datagram receive.
46 * Arnt Gulbrandsen : New udp_send and stuff
47 * Alan Cox : Cache last socket
48 * Alan Cox : Route cache
49 * Jon Peatfield : Minor efficiency fix to sendto().
50 * Mike Shaver : RFC1122 checks.
51 * Alan Cox : Nonblocking error fix.
52 * Willy Konynenberg : Transparent proxying support.
53 * Mike McLagan : Routing by source
54 * David S. Miller : New socket lookup architecture.
55 * Last socket cache retained as it
56 * does have a high hit rate.
57 * Olaf Kirch : Don't linearise iovec on sendmsg.
58 * Andi Kleen : Some cleanups, cache destination entry
59 * for connect.
60 * Vitaly E. Lavrov : Transparent proxy revived after year coma.
61 * Melvin Smith : Check msg_name not msg_namelen in sendto(),
62 * return ENOTCONN for unconnected sockets (POSIX)
63 * Janos Farkas : don't deliver multi/broadcasts to a different
64 * bound-to-device socket
65 * Hirokazu Takahashi : HW checksumming for outgoing UDP
66 * datagrams.
67 * Hirokazu Takahashi : sendfile() on UDP works now.
68 * Arnaldo C. Melo : convert /proc/net/udp to seq_file
69 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
70 * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind
71 * a single port at the same time.
72 * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support
73 *
74 *
75 * This program is free software; you can redistribute it and/or
76 * modify it under the terms of the GNU General Public License
77 * as published by the Free Software Foundation; either version
78 * 2 of the License, or (at your option) any later version.
79 */
80
81 #include <asm/system.h>
82 #include <asm/uaccess.h>
83 #include <asm/ioctls.h>
84 #include <linux/types.h>
85 #include <linux/fcntl.h>
86 #include <linux/module.h>
87 #include <linux/socket.h>
88 #include <linux/sockios.h>
89 #include <linux/igmp.h>
90 #include <linux/in.h>
91 #include <linux/errno.h>
92 #include <linux/timer.h>
93 #include <linux/mm.h>
94 #include <linux/inet.h>
95 #include <linux/netdevice.h>
96 #include <net/tcp_states.h>
97 #include <linux/skbuff.h>
98 #include <linux/proc_fs.h>
99 #include <linux/seq_file.h>
100 #include <net/icmp.h>
101 #include <net/route.h>
102 #include <net/checksum.h>
103 #include <net/xfrm.h>
104 #include "udp_impl.h"
105
106 /*
107 * Snmp MIB for the UDP layer
108 */
109
110 DEFINE_SNMP_STAT(struct udp_mib, udp_statistics) __read_mostly;
111
112 struct hlist_head udp_hash[UDP_HTABLE_SIZE];
113 DEFINE_RWLOCK(udp_hash_lock);
114
115 static int udp_port_rover;
116
117 static inline int __udp_lib_lport_inuse(__u16 num, struct hlist_head udptable[])
118 {
119 struct sock *sk;
120 struct hlist_node *node;
121
122 sk_for_each(sk, node, &udptable[num & (UDP_HTABLE_SIZE - 1)])
123 if (sk->sk_hash == num)
124 return 1;
125 return 0;
126 }
127
128 /**
129 * __udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
130 *
131 * @sk: socket struct in question
132 * @snum: port number to look up
133 * @udptable: hash list table, must be of UDP_HTABLE_SIZE
134 * @port_rover: pointer to record of last unallocated port
135 * @saddr_comp: AF-dependent comparison of bound local IP addresses
136 */
137 int __udp_lib_get_port(struct sock *sk, unsigned short snum,
138 struct hlist_head udptable[], int *port_rover,
139 int (*saddr_comp)(const struct sock *sk1,
140 const struct sock *sk2 ) )
141 {
142 struct hlist_node *node;
143 struct hlist_head *head;
144 struct sock *sk2;
145 int error = 1;
146
147 write_lock_bh(&udp_hash_lock);
148 if (snum == 0) {
149 int best_size_so_far, best, result, i;
150
151 if (*port_rover > sysctl_local_port_range[1] ||
152 *port_rover < sysctl_local_port_range[0])
153 *port_rover = sysctl_local_port_range[0];
154 best_size_so_far = 32767;
155 best = result = *port_rover;
156 for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) {
157 int size;
158
159 head = &udptable[result & (UDP_HTABLE_SIZE - 1)];
160 if (hlist_empty(head)) {
161 if (result > sysctl_local_port_range[1])
162 result = sysctl_local_port_range[0] +
163 ((result - sysctl_local_port_range[0]) &
164 (UDP_HTABLE_SIZE - 1));
165 goto gotit;
166 }
167 size = 0;
168 sk_for_each(sk2, node, head) {
169 if (++size >= best_size_so_far)
170 goto next;
171 }
172 best_size_so_far = size;
173 best = result;
174 next:
175 ;
176 }
177 result = best;
178 for (i = 0; i < (1 << 16) / UDP_HTABLE_SIZE;
179 i++, result += UDP_HTABLE_SIZE) {
180 if (result > sysctl_local_port_range[1])
181 result = sysctl_local_port_range[0]
182 + ((result - sysctl_local_port_range[0]) &
183 (UDP_HTABLE_SIZE - 1));
184 if (! __udp_lib_lport_inuse(result, udptable))
185 break;
186 }
187 if (i >= (1 << 16) / UDP_HTABLE_SIZE)
188 goto fail;
189 gotit:
190 *port_rover = snum = result;
191 } else {
192 head = &udptable[snum & (UDP_HTABLE_SIZE - 1)];
193
194 sk_for_each(sk2, node, head)
195 if (sk2->sk_hash == snum &&
196 sk2 != sk &&
197 (!sk2->sk_reuse || !sk->sk_reuse) &&
198 (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if
199 || sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
200 (*saddr_comp)(sk, sk2) )
201 goto fail;
202 }
203 inet_sk(sk)->num = snum;
204 sk->sk_hash = snum;
205 if (sk_unhashed(sk)) {
206 head = &udptable[snum & (UDP_HTABLE_SIZE - 1)];
207 sk_add_node(sk, head);
208 sock_prot_inc_use(sk->sk_prot);
209 }
210 error = 0;
211 fail:
212 write_unlock_bh(&udp_hash_lock);
213 return error;
214 }
215
216 int udp_get_port(struct sock *sk, unsigned short snum,
217 int (*scmp)(const struct sock *, const struct sock *))
218 {
219 return __udp_lib_get_port(sk, snum, udp_hash, &udp_port_rover, scmp);
220 }
221
222 int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
223 {
224 struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
225
226 return ( !ipv6_only_sock(sk2) &&
227 (!inet1->rcv_saddr || !inet2->rcv_saddr ||
228 inet1->rcv_saddr == inet2->rcv_saddr ));
229 }
230
231 static inline int udp_v4_get_port(struct sock *sk, unsigned short snum)
232 {
233 return udp_get_port(sk, snum, ipv4_rcv_saddr_equal);
234 }
235
236 /* UDP is nearly always wildcards out the wazoo, it makes no sense to try
237 * harder than this. -DaveM
238 */
239 static struct sock *__udp4_lib_lookup(__be32 saddr, __be16 sport,
240 __be32 daddr, __be16 dport,
241 int dif, struct hlist_head udptable[])
242 {
243 struct sock *sk, *result = NULL;
244 struct hlist_node *node;
245 unsigned short hnum = ntohs(dport);
246 int badness = -1;
247
248 read_lock(&udp_hash_lock);
249 sk_for_each(sk, node, &udptable[hnum & (UDP_HTABLE_SIZE - 1)]) {
250 struct inet_sock *inet = inet_sk(sk);
251
252 if (sk->sk_hash == hnum && !ipv6_only_sock(sk)) {
253 int score = (sk->sk_family == PF_INET ? 1 : 0);
254 if (inet->rcv_saddr) {
255 if (inet->rcv_saddr != daddr)
256 continue;
257 score+=2;
258 }
259 if (inet->daddr) {
260 if (inet->daddr != saddr)
261 continue;
262 score+=2;
263 }
264 if (inet->dport) {
265 if (inet->dport != sport)
266 continue;
267 score+=2;
268 }
269 if (sk->sk_bound_dev_if) {
270 if (sk->sk_bound_dev_if != dif)
271 continue;
272 score+=2;
273 }
274 if (score == 9) {
275 result = sk;
276 break;
277 } else if (score > badness) {
278 result = sk;
279 badness = score;
280 }
281 }
282 }
283 if (result)
284 sock_hold(result);
285 read_unlock(&udp_hash_lock);
286 return result;
287 }
288
289 static inline struct sock *udp_v4_mcast_next(struct sock *sk,
290 __be16 loc_port, __be32 loc_addr,
291 __be16 rmt_port, __be32 rmt_addr,
292 int dif)
293 {
294 struct hlist_node *node;
295 struct sock *s = sk;
296 unsigned short hnum = ntohs(loc_port);
297
298 sk_for_each_from(s, node) {
299 struct inet_sock *inet = inet_sk(s);
300
301 if (s->sk_hash != hnum ||
302 (inet->daddr && inet->daddr != rmt_addr) ||
303 (inet->dport != rmt_port && inet->dport) ||
304 (inet->rcv_saddr && inet->rcv_saddr != loc_addr) ||
305 ipv6_only_sock(s) ||
306 (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
307 continue;
308 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
309 continue;
310 goto found;
311 }
312 s = NULL;
313 found:
314 return s;
315 }
316
317 /*
318 * This routine is called by the ICMP module when it gets some
319 * sort of error condition. If err < 0 then the socket should
320 * be closed and the error returned to the user. If err > 0
321 * it's just the icmp type << 8 | icmp code.
322 * Header points to the ip header of the error packet. We move
323 * on past this. Then (as it used to claim before adjustment)
324 * header points to the first 8 bytes of the udp header. We need
325 * to find the appropriate port.
326 */
327
328 void __udp4_lib_err(struct sk_buff *skb, u32 info, struct hlist_head udptable[])
329 {
330 struct inet_sock *inet;
331 struct iphdr *iph = (struct iphdr*)skb->data;
332 struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2));
333 const int type = icmp_hdr(skb)->type;
334 const int code = icmp_hdr(skb)->code;
335 struct sock *sk;
336 int harderr;
337 int err;
338
339 sk = __udp4_lib_lookup(iph->daddr, uh->dest, iph->saddr, uh->source,
340 skb->dev->ifindex, udptable );
341 if (sk == NULL) {
342 ICMP_INC_STATS_BH(ICMP_MIB_INERRORS);
343 return; /* No socket for error */
344 }
345
346 err = 0;
347 harderr = 0;
348 inet = inet_sk(sk);
349
350 switch (type) {
351 default:
352 case ICMP_TIME_EXCEEDED:
353 err = EHOSTUNREACH;
354 break;
355 case ICMP_SOURCE_QUENCH:
356 goto out;
357 case ICMP_PARAMETERPROB:
358 err = EPROTO;
359 harderr = 1;
360 break;
361 case ICMP_DEST_UNREACH:
362 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
363 if (inet->pmtudisc != IP_PMTUDISC_DONT) {
364 err = EMSGSIZE;
365 harderr = 1;
366 break;
367 }
368 goto out;
369 }
370 err = EHOSTUNREACH;
371 if (code <= NR_ICMP_UNREACH) {
372 harderr = icmp_err_convert[code].fatal;
373 err = icmp_err_convert[code].errno;
374 }
375 break;
376 }
377
378 /*
379 * RFC1122: OK. Passes ICMP errors back to application, as per
380 * 4.1.3.3.
381 */
382 if (!inet->recverr) {
383 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
384 goto out;
385 } else {
386 ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1));
387 }
388 sk->sk_err = err;
389 sk->sk_error_report(sk);
390 out:
391 sock_put(sk);
392 }
393
394 void udp_err(struct sk_buff *skb, u32 info)
395 {
396 return __udp4_lib_err(skb, info, udp_hash);
397 }
398
399 /*
400 * Throw away all pending data and cancel the corking. Socket is locked.
401 */
402 static void udp_flush_pending_frames(struct sock *sk)
403 {
404 struct udp_sock *up = udp_sk(sk);
405
406 if (up->pending) {
407 up->len = 0;
408 up->pending = 0;
409 ip_flush_pending_frames(sk);
410 }
411 }
412
413 /**
414 * udp4_hwcsum_outgoing - handle outgoing HW checksumming
415 * @sk: socket we are sending on
416 * @skb: sk_buff containing the filled-in UDP header
417 * (checksum field must be zeroed out)
418 */
419 static void udp4_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
420 __be32 src, __be32 dst, int len )
421 {
422 unsigned int offset;
423 struct udphdr *uh = udp_hdr(skb);
424 __wsum csum = 0;
425
426 if (skb_queue_len(&sk->sk_write_queue) == 1) {
427 /*
428 * Only one fragment on the socket.
429 */
430 skb->csum_start = skb_transport_header(skb) - skb->head;
431 skb->csum_offset = offsetof(struct udphdr, check);
432 uh->check = ~csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, 0);
433 } else {
434 /*
435 * HW-checksum won't work as there are two or more
436 * fragments on the socket so that all csums of sk_buffs
437 * should be together
438 */
439 offset = skb_transport_offset(skb);
440 skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
441
442 skb->ip_summed = CHECKSUM_NONE;
443
444 skb_queue_walk(&sk->sk_write_queue, skb) {
445 csum = csum_add(csum, skb->csum);
446 }
447
448 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
449 if (uh->check == 0)
450 uh->check = CSUM_MANGLED_0;
451 }
452 }
453
454 /*
455 * Push out all pending data as one UDP datagram. Socket is locked.
456 */
457 static int udp_push_pending_frames(struct sock *sk)
458 {
459 struct udp_sock *up = udp_sk(sk);
460 struct inet_sock *inet = inet_sk(sk);
461 struct flowi *fl = &inet->cork.fl;
462 struct sk_buff *skb;
463 struct udphdr *uh;
464 int err = 0;
465 __wsum csum = 0;
466
467 /* Grab the skbuff where UDP header space exists. */
468 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
469 goto out;
470
471 /*
472 * Create a UDP header
473 */
474 uh = udp_hdr(skb);
475 uh->source = fl->fl_ip_sport;
476 uh->dest = fl->fl_ip_dport;
477 uh->len = htons(up->len);
478 uh->check = 0;
479
480 if (up->pcflag) /* UDP-Lite */
481 csum = udplite_csum_outgoing(sk, skb);
482
483 else if (sk->sk_no_check == UDP_CSUM_NOXMIT) { /* UDP csum disabled */
484
485 skb->ip_summed = CHECKSUM_NONE;
486 goto send;
487
488 } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
489
490 udp4_hwcsum_outgoing(sk, skb, fl->fl4_src,fl->fl4_dst, up->len);
491 goto send;
492
493 } else /* `normal' UDP */
494 csum = udp_csum_outgoing(sk, skb);
495
496 /* add protocol-dependent pseudo-header */
497 uh->check = csum_tcpudp_magic(fl->fl4_src, fl->fl4_dst, up->len,
498 sk->sk_protocol, csum );
499 if (uh->check == 0)
500 uh->check = CSUM_MANGLED_0;
501
502 send:
503 err = ip_push_pending_frames(sk);
504 out:
505 up->len = 0;
506 up->pending = 0;
507 return err;
508 }
509
510 int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
511 size_t len)
512 {
513 struct inet_sock *inet = inet_sk(sk);
514 struct udp_sock *up = udp_sk(sk);
515 int ulen = len;
516 struct ipcm_cookie ipc;
517 struct rtable *rt = NULL;
518 int free = 0;
519 int connected = 0;
520 __be32 daddr, faddr, saddr;
521 __be16 dport;
522 u8 tos;
523 int err, is_udplite = up->pcflag;
524 int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
525 int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
526
527 if (len > 0xFFFF)
528 return -EMSGSIZE;
529
530 /*
531 * Check the flags.
532 */
533
534 if (msg->msg_flags&MSG_OOB) /* Mirror BSD error message compatibility */
535 return -EOPNOTSUPP;
536
537 ipc.opt = NULL;
538
539 if (up->pending) {
540 /*
541 * There are pending frames.
542 * The socket lock must be held while it's corked.
543 */
544 lock_sock(sk);
545 if (likely(up->pending)) {
546 if (unlikely(up->pending != AF_INET)) {
547 release_sock(sk);
548 return -EINVAL;
549 }
550 goto do_append_data;
551 }
552 release_sock(sk);
553 }
554 ulen += sizeof(struct udphdr);
555
556 /*
557 * Get and verify the address.
558 */
559 if (msg->msg_name) {
560 struct sockaddr_in * usin = (struct sockaddr_in*)msg->msg_name;
561 if (msg->msg_namelen < sizeof(*usin))
562 return -EINVAL;
563 if (usin->sin_family != AF_INET) {
564 if (usin->sin_family != AF_UNSPEC)
565 return -EAFNOSUPPORT;
566 }
567
568 daddr = usin->sin_addr.s_addr;
569 dport = usin->sin_port;
570 if (dport == 0)
571 return -EINVAL;
572 } else {
573 if (sk->sk_state != TCP_ESTABLISHED)
574 return -EDESTADDRREQ;
575 daddr = inet->daddr;
576 dport = inet->dport;
577 /* Open fast path for connected socket.
578 Route will not be used, if at least one option is set.
579 */
580 connected = 1;
581 }
582 ipc.addr = inet->saddr;
583
584 ipc.oif = sk->sk_bound_dev_if;
585 if (msg->msg_controllen) {
586 err = ip_cmsg_send(msg, &ipc);
587 if (err)
588 return err;
589 if (ipc.opt)
590 free = 1;
591 connected = 0;
592 }
593 if (!ipc.opt)
594 ipc.opt = inet->opt;
595
596 saddr = ipc.addr;
597 ipc.addr = faddr = daddr;
598
599 if (ipc.opt && ipc.opt->srr) {
600 if (!daddr)
601 return -EINVAL;
602 faddr = ipc.opt->faddr;
603 connected = 0;
604 }
605 tos = RT_TOS(inet->tos);
606 if (sock_flag(sk, SOCK_LOCALROUTE) ||
607 (msg->msg_flags & MSG_DONTROUTE) ||
608 (ipc.opt && ipc.opt->is_strictroute)) {
609 tos |= RTO_ONLINK;
610 connected = 0;
611 }
612
613 if (MULTICAST(daddr)) {
614 if (!ipc.oif)
615 ipc.oif = inet->mc_index;
616 if (!saddr)
617 saddr = inet->mc_addr;
618 connected = 0;
619 }
620
621 if (connected)
622 rt = (struct rtable*)sk_dst_check(sk, 0);
623
624 if (rt == NULL) {
625 struct flowi fl = { .oif = ipc.oif,
626 .nl_u = { .ip4_u =
627 { .daddr = faddr,
628 .saddr = saddr,
629 .tos = tos } },
630 .proto = sk->sk_protocol,
631 .uli_u = { .ports =
632 { .sport = inet->sport,
633 .dport = dport } } };
634 security_sk_classify_flow(sk, &fl);
635 err = ip_route_output_flow(&rt, &fl, sk, 1);
636 if (err) {
637 if (err == -ENETUNREACH)
638 IP_INC_STATS_BH(IPSTATS_MIB_OUTNOROUTES);
639 goto out;
640 }
641
642 err = -EACCES;
643 if ((rt->rt_flags & RTCF_BROADCAST) &&
644 !sock_flag(sk, SOCK_BROADCAST))
645 goto out;
646 if (connected)
647 sk_dst_set(sk, dst_clone(&rt->u.dst));
648 }
649
650 if (msg->msg_flags&MSG_CONFIRM)
651 goto do_confirm;
652 back_from_confirm:
653
654 saddr = rt->rt_src;
655 if (!ipc.addr)
656 daddr = ipc.addr = rt->rt_dst;
657
658 lock_sock(sk);
659 if (unlikely(up->pending)) {
660 /* The socket is already corked while preparing it. */
661 /* ... which is an evident application bug. --ANK */
662 release_sock(sk);
663
664 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
665 err = -EINVAL;
666 goto out;
667 }
668 /*
669 * Now cork the socket to pend data.
670 */
671 inet->cork.fl.fl4_dst = daddr;
672 inet->cork.fl.fl_ip_dport = dport;
673 inet->cork.fl.fl4_src = saddr;
674 inet->cork.fl.fl_ip_sport = inet->sport;
675 up->pending = AF_INET;
676
677 do_append_data:
678 up->len += ulen;
679 getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
680 err = ip_append_data(sk, getfrag, msg->msg_iov, ulen,
681 sizeof(struct udphdr), &ipc, rt,
682 corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
683 if (err)
684 udp_flush_pending_frames(sk);
685 else if (!corkreq)
686 err = udp_push_pending_frames(sk);
687 else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
688 up->pending = 0;
689 release_sock(sk);
690
691 out:
692 ip_rt_put(rt);
693 if (free)
694 kfree(ipc.opt);
695 if (!err) {
696 UDP_INC_STATS_USER(UDP_MIB_OUTDATAGRAMS, is_udplite);
697 return len;
698 }
699 /*
700 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
701 * ENOBUFS might not be good (it's not tunable per se), but otherwise
702 * we don't have a good statistic (IpOutDiscards but it can be too many
703 * things). We could add another new stat but at least for now that
704 * seems like overkill.
705 */
706 if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
707 UDP_INC_STATS_USER(UDP_MIB_SNDBUFERRORS, is_udplite);
708 }
709 return err;
710
711 do_confirm:
712 dst_confirm(&rt->u.dst);
713 if (!(msg->msg_flags&MSG_PROBE) || len)
714 goto back_from_confirm;
715 err = 0;
716 goto out;
717 }
718
719 int udp_sendpage(struct sock *sk, struct page *page, int offset,
720 size_t size, int flags)
721 {
722 struct udp_sock *up = udp_sk(sk);
723 int ret;
724
725 if (!up->pending) {
726 struct msghdr msg = { .msg_flags = flags|MSG_MORE };
727
728 /* Call udp_sendmsg to specify destination address which
729 * sendpage interface can't pass.
730 * This will succeed only when the socket is connected.
731 */
732 ret = udp_sendmsg(NULL, sk, &msg, 0);
733 if (ret < 0)
734 return ret;
735 }
736
737 lock_sock(sk);
738
739 if (unlikely(!up->pending)) {
740 release_sock(sk);
741
742 LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 3\n");
743 return -EINVAL;
744 }
745
746 ret = ip_append_page(sk, page, offset, size, flags);
747 if (ret == -EOPNOTSUPP) {
748 release_sock(sk);
749 return sock_no_sendpage(sk->sk_socket, page, offset,
750 size, flags);
751 }
752 if (ret < 0) {
753 udp_flush_pending_frames(sk);
754 goto out;
755 }
756
757 up->len += size;
758 if (!(up->corkflag || (flags&MSG_MORE)))
759 ret = udp_push_pending_frames(sk);
760 if (!ret)
761 ret = size;
762 out:
763 release_sock(sk);
764 return ret;
765 }
766
767 /*
768 * IOCTL requests applicable to the UDP protocol
769 */
770
771 int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
772 {
773 switch (cmd) {
774 case SIOCOUTQ:
775 {
776 int amount = atomic_read(&sk->sk_wmem_alloc);
777 return put_user(amount, (int __user *)arg);
778 }
779
780 case SIOCINQ:
781 {
782 struct sk_buff *skb;
783 unsigned long amount;
784
785 amount = 0;
786 spin_lock_bh(&sk->sk_receive_queue.lock);
787 skb = skb_peek(&sk->sk_receive_queue);
788 if (skb != NULL) {
789 /*
790 * We will only return the amount
791 * of this packet since that is all
792 * that will be read.
793 */
794 amount = skb->len - sizeof(struct udphdr);
795 }
796 spin_unlock_bh(&sk->sk_receive_queue.lock);
797 return put_user(amount, (int __user *)arg);
798 }
799
800 default:
801 return -ENOIOCTLCMD;
802 }
803
804 return 0;
805 }
806
807 /*
808 * This should be easy, if there is something there we
809 * return it, otherwise we block.
810 */
811
812 int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
813 size_t len, int noblock, int flags, int *addr_len)
814 {
815 struct inet_sock *inet = inet_sk(sk);
816 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
817 struct sk_buff *skb;
818 unsigned int ulen, copied;
819 int err;
820 int is_udplite = IS_UDPLITE(sk);
821
822 /*
823 * Check any passed addresses
824 */
825 if (addr_len)
826 *addr_len=sizeof(*sin);
827
828 if (flags & MSG_ERRQUEUE)
829 return ip_recv_error(sk, msg, len);
830
831 try_again:
832 skb = skb_recv_datagram(sk, flags, noblock, &err);
833 if (!skb)
834 goto out;
835
836 ulen = skb->len - sizeof(struct udphdr);
837 copied = len;
838 if (copied > ulen)
839 copied = ulen;
840 else if (copied < ulen)
841 msg->msg_flags |= MSG_TRUNC;
842
843 /*
844 * If checksum is needed at all, try to do it while copying the
845 * data. If the data is truncated, or if we only want a partial
846 * coverage checksum (UDP-Lite), do it before the copy.
847 */
848
849 if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
850 if (udp_lib_checksum_complete(skb))
851 goto csum_copy_err;
852 }
853
854 if (skb_csum_unnecessary(skb))
855 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
856 msg->msg_iov, copied );
857 else {
858 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
859
860 if (err == -EINVAL)
861 goto csum_copy_err;
862 }
863
864 if (err)
865 goto out_free;
866
867 sock_recv_timestamp(msg, sk, skb);
868
869 /* Copy the address. */
870 if (sin)
871 {
872 sin->sin_family = AF_INET;
873 sin->sin_port = udp_hdr(skb)->source;
874 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
875 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
876 }
877 if (inet->cmsg_flags)
878 ip_cmsg_recv(msg, skb);
879
880 err = copied;
881 if (flags & MSG_TRUNC)
882 err = ulen;
883
884 out_free:
885 skb_free_datagram(sk, skb);
886 out:
887 return err;
888
889 csum_copy_err:
890 UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_udplite);
891
892 skb_kill_datagram(sk, skb, flags);
893
894 if (noblock)
895 return -EAGAIN;
896 goto try_again;
897 }
898
899
900 int udp_disconnect(struct sock *sk, int flags)
901 {
902 struct inet_sock *inet = inet_sk(sk);
903 /*
904 * 1003.1g - break association.
905 */
906
907 sk->sk_state = TCP_CLOSE;
908 inet->daddr = 0;
909 inet->dport = 0;
910 sk->sk_bound_dev_if = 0;
911 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
912 inet_reset_saddr(sk);
913
914 if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
915 sk->sk_prot->unhash(sk);
916 inet->sport = 0;
917 }
918 sk_dst_reset(sk);
919 return 0;
920 }
921
922 /* return:
923 * 1 if the UDP system should process it
924 * 0 if we should drop this packet
925 * -1 if it should get processed by xfrm4_rcv_encap
926 */
927 static int udp_encap_rcv(struct sock * sk, struct sk_buff *skb)
928 {
929 #ifndef CONFIG_XFRM
930 return 1;
931 #else
932 struct udp_sock *up = udp_sk(sk);
933 struct udphdr *uh;
934 struct iphdr *iph;
935 int iphlen, len;
936
937 __u8 *udpdata;
938 __be32 *udpdata32;
939 __u16 encap_type = up->encap_type;
940
941 /* if we're overly short, let UDP handle it */
942 len = skb->len - sizeof(struct udphdr);
943 if (len <= 0)
944 return 1;
945
946 /* if this is not encapsulated socket, then just return now */
947 if (!encap_type)
948 return 1;
949
950 /* If this is a paged skb, make sure we pull up
951 * whatever data we need to look at. */
952 if (!pskb_may_pull(skb, sizeof(struct udphdr) + min(len, 8)))
953 return 1;
954
955 /* Now we can get the pointers */
956 uh = udp_hdr(skb);
957 udpdata = (__u8 *)uh + sizeof(struct udphdr);
958 udpdata32 = (__be32 *)udpdata;
959
960 switch (encap_type) {
961 default:
962 case UDP_ENCAP_ESPINUDP:
963 /* Check if this is a keepalive packet. If so, eat it. */
964 if (len == 1 && udpdata[0] == 0xff) {
965 return 0;
966 } else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0) {
967 /* ESP Packet without Non-ESP header */
968 len = sizeof(struct udphdr);
969 } else
970 /* Must be an IKE packet.. pass it through */
971 return 1;
972 break;
973 case UDP_ENCAP_ESPINUDP_NON_IKE:
974 /* Check if this is a keepalive packet. If so, eat it. */
975 if (len == 1 && udpdata[0] == 0xff) {
976 return 0;
977 } else if (len > 2 * sizeof(u32) + sizeof(struct ip_esp_hdr) &&
978 udpdata32[0] == 0 && udpdata32[1] == 0) {
979
980 /* ESP Packet with Non-IKE marker */
981 len = sizeof(struct udphdr) + 2 * sizeof(u32);
982 } else
983 /* Must be an IKE packet.. pass it through */
984 return 1;
985 break;
986 }
987
988 /* At this point we are sure that this is an ESPinUDP packet,
989 * so we need to remove 'len' bytes from the packet (the UDP
990 * header and optional ESP marker bytes) and then modify the
991 * protocol to ESP, and then call into the transform receiver.
992 */
993 if (skb_cloned(skb) && pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
994 return 0;
995
996 /* Now we can update and verify the packet length... */
997 iph = ip_hdr(skb);
998 iphlen = iph->ihl << 2;
999 iph->tot_len = htons(ntohs(iph->tot_len) - len);
1000 if (skb->len < iphlen + len) {
1001 /* packet is too small!?! */
1002 return 0;
1003 }
1004
1005 /* pull the data buffer up to the ESP header and set the
1006 * transport header to point to ESP. Keep UDP on the stack
1007 * for later.
1008 */
1009 __skb_pull(skb, len);
1010 skb_reset_transport_header(skb);
1011
1012 /* modify the protocol (it's ESP!) */
1013 iph->protocol = IPPROTO_ESP;
1014
1015 /* and let the caller know to send this into the ESP processor... */
1016 return -1;
1017 #endif
1018 }
1019
1020 /* returns:
1021 * -1: error
1022 * 0: success
1023 * >0: "udp encap" protocol resubmission
1024 *
1025 * Note that in the success and error cases, the skb is assumed to
1026 * have either been requeued or freed.
1027 */
1028 int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
1029 {
1030 struct udp_sock *up = udp_sk(sk);
1031 int rc;
1032
1033 /*
1034 * Charge it to the socket, dropping if the queue is full.
1035 */
1036 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1037 goto drop;
1038 nf_reset(skb);
1039
1040 if (up->encap_type) {
1041 /*
1042 * This is an encapsulation socket, so let's see if this is
1043 * an encapsulated packet.
1044 * If it's a keepalive packet, then just eat it.
1045 * If it's an encapsulateed packet, then pass it to the
1046 * IPsec xfrm input and return the response
1047 * appropriately. Otherwise, just fall through and
1048 * pass this up the UDP socket.
1049 */
1050 int ret;
1051
1052 ret = udp_encap_rcv(sk, skb);
1053 if (ret == 0) {
1054 /* Eat the packet .. */
1055 kfree_skb(skb);
1056 return 0;
1057 }
1058 if (ret < 0) {
1059 /* process the ESP packet */
1060 ret = xfrm4_rcv_encap(skb, up->encap_type);
1061 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS, up->pcflag);
1062 return -ret;
1063 }
1064 /* FALLTHROUGH -- it's a UDP Packet */
1065 }
1066
1067 /*
1068 * UDP-Lite specific tests, ignored on UDP sockets
1069 */
1070 if ((up->pcflag & UDPLITE_RECV_CC) && UDP_SKB_CB(skb)->partial_cov) {
1071
1072 /*
1073 * MIB statistics other than incrementing the error count are
1074 * disabled for the following two types of errors: these depend
1075 * on the application settings, not on the functioning of the
1076 * protocol stack as such.
1077 *
1078 * RFC 3828 here recommends (sec 3.3): "There should also be a
1079 * way ... to ... at least let the receiving application block
1080 * delivery of packets with coverage values less than a value
1081 * provided by the application."
1082 */
1083 if (up->pcrlen == 0) { /* full coverage was set */
1084 LIMIT_NETDEBUG(KERN_WARNING "UDPLITE: partial coverage "
1085 "%d while full coverage %d requested\n",
1086 UDP_SKB_CB(skb)->cscov, skb->len);
1087 goto drop;
1088 }
1089 /* The next case involves violating the min. coverage requested
1090 * by the receiver. This is subtle: if receiver wants x and x is
1091 * greater than the buffersize/MTU then receiver will complain
1092 * that it wants x while sender emits packets of smaller size y.
1093 * Therefore the above ...()->partial_cov statement is essential.
1094 */
1095 if (UDP_SKB_CB(skb)->cscov < up->pcrlen) {
1096 LIMIT_NETDEBUG(KERN_WARNING
1097 "UDPLITE: coverage %d too small, need min %d\n",
1098 UDP_SKB_CB(skb)->cscov, up->pcrlen);
1099 goto drop;
1100 }
1101 }
1102
1103 if (sk->sk_filter) {
1104 if (udp_lib_checksum_complete(skb))
1105 goto drop;
1106 }
1107
1108 if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
1109 /* Note that an ENOMEM error is charged twice */
1110 if (rc == -ENOMEM)
1111 UDP_INC_STATS_BH(UDP_MIB_RCVBUFERRORS, up->pcflag);
1112 goto drop;
1113 }
1114
1115 UDP_INC_STATS_BH(UDP_MIB_INDATAGRAMS, up->pcflag);
1116 return 0;
1117
1118 drop:
1119 UDP_INC_STATS_BH(UDP_MIB_INERRORS, up->pcflag);
1120 kfree_skb(skb);
1121 return -1;
1122 }
1123
1124 /*
1125 * Multicasts and broadcasts go to each listener.
1126 *
1127 * Note: called only from the BH handler context,
1128 * so we don't need to lock the hashes.
1129 */
1130 static int __udp4_lib_mcast_deliver(struct sk_buff *skb,
1131 struct udphdr *uh,
1132 __be32 saddr, __be32 daddr,
1133 struct hlist_head udptable[])
1134 {
1135 struct sock *sk;
1136 int dif;
1137
1138 read_lock(&udp_hash_lock);
1139 sk = sk_head(&udptable[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
1140 dif = skb->dev->ifindex;
1141 sk = udp_v4_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
1142 if (sk) {
1143 struct sock *sknext = NULL;
1144
1145 do {
1146 struct sk_buff *skb1 = skb;
1147
1148 sknext = udp_v4_mcast_next(sk_next(sk), uh->dest, daddr,
1149 uh->source, saddr, dif);
1150 if (sknext)
1151 skb1 = skb_clone(skb, GFP_ATOMIC);
1152
1153 if (skb1) {
1154 int ret = udp_queue_rcv_skb(sk, skb1);
1155 if (ret > 0)
1156 /* we should probably re-process instead
1157 * of dropping packets here. */
1158 kfree_skb(skb1);
1159 }
1160 sk = sknext;
1161 } while (sknext);
1162 } else
1163 kfree_skb(skb);
1164 read_unlock(&udp_hash_lock);
1165 return 0;
1166 }
1167
1168 /* Initialize UDP checksum. If exited with zero value (success),
1169 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1170 * Otherwise, csum completion requires chacksumming packet body,
1171 * including udp header and folding it to skb->csum.
1172 */
1173 static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1174 int proto)
1175 {
1176 const struct iphdr *iph;
1177 int err;
1178
1179 UDP_SKB_CB(skb)->partial_cov = 0;
1180 UDP_SKB_CB(skb)->cscov = skb->len;
1181
1182 if (proto == IPPROTO_UDPLITE) {
1183 err = udplite_checksum_init(skb, uh);
1184 if (err)
1185 return err;
1186 }
1187
1188 iph = ip_hdr(skb);
1189 if (uh->check == 0) {
1190 skb->ip_summed = CHECKSUM_UNNECESSARY;
1191 } else if (skb->ip_summed == CHECKSUM_COMPLETE) {
1192 if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len,
1193 proto, skb->csum))
1194 skb->ip_summed = CHECKSUM_UNNECESSARY;
1195 }
1196 if (!skb_csum_unnecessary(skb))
1197 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
1198 skb->len, proto, 0);
1199 /* Probably, we should checksum udp header (it should be in cache
1200 * in any case) and data in tiny packets (< rx copybreak).
1201 */
1202
1203 return 0;
1204 }
1205
1206 /*
1207 * All we need to do is get the socket, and then do a checksum.
1208 */
1209
1210 int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[],
1211 int proto)
1212 {
1213 struct sock *sk;
1214 struct udphdr *uh = udp_hdr(skb);
1215 unsigned short ulen;
1216 struct rtable *rt = (struct rtable*)skb->dst;
1217 __be32 saddr = ip_hdr(skb)->saddr;
1218 __be32 daddr = ip_hdr(skb)->daddr;
1219
1220 /*
1221 * Validate the packet.
1222 */
1223 if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1224 goto drop; /* No space for header. */
1225
1226 ulen = ntohs(uh->len);
1227 if (ulen > skb->len)
1228 goto short_packet;
1229
1230 if (proto == IPPROTO_UDP) {
1231 /* UDP validates ulen. */
1232 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1233 goto short_packet;
1234 uh = udp_hdr(skb);
1235 }
1236
1237 if (udp4_csum_init(skb, uh, proto))
1238 goto csum_error;
1239
1240 if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1241 return __udp4_lib_mcast_deliver(skb, uh, saddr, daddr, udptable);
1242
1243 sk = __udp4_lib_lookup(saddr, uh->source, daddr, uh->dest,
1244 skb->dev->ifindex, udptable );
1245
1246 if (sk != NULL) {
1247 int ret = udp_queue_rcv_skb(sk, skb);
1248 sock_put(sk);
1249
1250 /* a return value > 0 means to resubmit the input, but
1251 * it wants the return to be -protocol, or 0
1252 */
1253 if (ret > 0)
1254 return -ret;
1255 return 0;
1256 }
1257
1258 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1259 goto drop;
1260 nf_reset(skb);
1261
1262 /* No socket. Drop packet silently, if checksum is wrong */
1263 if (udp_lib_checksum_complete(skb))
1264 goto csum_error;
1265
1266 UDP_INC_STATS_BH(UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
1267 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1268
1269 /*
1270 * Hmm. We got an UDP packet to a port to which we
1271 * don't wanna listen. Ignore it.
1272 */
1273 kfree_skb(skb);
1274 return 0;
1275
1276 short_packet:
1277 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n",
1278 proto == IPPROTO_UDPLITE ? "-Lite" : "",
1279 NIPQUAD(saddr),
1280 ntohs(uh->source),
1281 ulen,
1282 skb->len,
1283 NIPQUAD(daddr),
1284 ntohs(uh->dest));
1285 goto drop;
1286
1287 csum_error:
1288 /*
1289 * RFC1122: OK. Discards the bad packet silently (as far as
1290 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1291 */
1292 LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n",
1293 proto == IPPROTO_UDPLITE ? "-Lite" : "",
1294 NIPQUAD(saddr),
1295 ntohs(uh->source),
1296 NIPQUAD(daddr),
1297 ntohs(uh->dest),
1298 ulen);
1299 drop:
1300 UDP_INC_STATS_BH(UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1301 kfree_skb(skb);
1302 return 0;
1303 }
1304
1305 int udp_rcv(struct sk_buff *skb)
1306 {
1307 return __udp4_lib_rcv(skb, udp_hash, IPPROTO_UDP);
1308 }
1309
1310 int udp_destroy_sock(struct sock *sk)
1311 {
1312 lock_sock(sk);
1313 udp_flush_pending_frames(sk);
1314 release_sock(sk);
1315 return 0;
1316 }
1317
1318 /*
1319 * Socket option code for UDP
1320 */
1321 int udp_lib_setsockopt(struct sock *sk, int level, int optname,
1322 char __user *optval, int optlen,
1323 int (*push_pending_frames)(struct sock *))
1324 {
1325 struct udp_sock *up = udp_sk(sk);
1326 int val;
1327 int err = 0;
1328
1329 if (optlen<sizeof(int))
1330 return -EINVAL;
1331
1332 if (get_user(val, (int __user *)optval))
1333 return -EFAULT;
1334
1335 switch (optname) {
1336 case UDP_CORK:
1337 if (val != 0) {
1338 up->corkflag = 1;
1339 } else {
1340 up->corkflag = 0;
1341 lock_sock(sk);
1342 (*push_pending_frames)(sk);
1343 release_sock(sk);
1344 }
1345 break;
1346
1347 case UDP_ENCAP:
1348 switch (val) {
1349 case 0:
1350 case UDP_ENCAP_ESPINUDP:
1351 case UDP_ENCAP_ESPINUDP_NON_IKE:
1352 up->encap_type = val;
1353 break;
1354 default:
1355 err = -ENOPROTOOPT;
1356 break;
1357 }
1358 break;
1359
1360 /*
1361 * UDP-Lite's partial checksum coverage (RFC 3828).
1362 */
1363 /* The sender sets actual checksum coverage length via this option.
1364 * The case coverage > packet length is handled by send module. */
1365 case UDPLITE_SEND_CSCOV:
1366 if (!up->pcflag) /* Disable the option on UDP sockets */
1367 return -ENOPROTOOPT;
1368 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
1369 val = 8;
1370 up->pcslen = val;
1371 up->pcflag |= UDPLITE_SEND_CC;
1372 break;
1373
1374 /* The receiver specifies a minimum checksum coverage value. To make
1375 * sense, this should be set to at least 8 (as done below). If zero is
1376 * used, this again means full checksum coverage. */
1377 case UDPLITE_RECV_CSCOV:
1378 if (!up->pcflag) /* Disable the option on UDP sockets */
1379 return -ENOPROTOOPT;
1380 if (val != 0 && val < 8) /* Avoid silly minimal values. */
1381 val = 8;
1382 up->pcrlen = val;
1383 up->pcflag |= UDPLITE_RECV_CC;
1384 break;
1385
1386 default:
1387 err = -ENOPROTOOPT;
1388 break;
1389 }
1390
1391 return err;
1392 }
1393
1394 int udp_setsockopt(struct sock *sk, int level, int optname,
1395 char __user *optval, int optlen)
1396 {
1397 if (level == SOL_UDP || level == SOL_UDPLITE)
1398 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1399 udp_push_pending_frames);
1400 return ip_setsockopt(sk, level, optname, optval, optlen);
1401 }
1402
1403 #ifdef CONFIG_COMPAT
1404 int compat_udp_setsockopt(struct sock *sk, int level, int optname,
1405 char __user *optval, int optlen)
1406 {
1407 if (level == SOL_UDP || level == SOL_UDPLITE)
1408 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1409 udp_push_pending_frames);
1410 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
1411 }
1412 #endif
1413
1414 int udp_lib_getsockopt(struct sock *sk, int level, int optname,
1415 char __user *optval, int __user *optlen)
1416 {
1417 struct udp_sock *up = udp_sk(sk);
1418 int val, len;
1419
1420 if (get_user(len,optlen))
1421 return -EFAULT;
1422
1423 len = min_t(unsigned int, len, sizeof(int));
1424
1425 if (len < 0)
1426 return -EINVAL;
1427
1428 switch (optname) {
1429 case UDP_CORK:
1430 val = up->corkflag;
1431 break;
1432
1433 case UDP_ENCAP:
1434 val = up->encap_type;
1435 break;
1436
1437 /* The following two cannot be changed on UDP sockets, the return is
1438 * always 0 (which corresponds to the full checksum coverage of UDP). */
1439 case UDPLITE_SEND_CSCOV:
1440 val = up->pcslen;
1441 break;
1442
1443 case UDPLITE_RECV_CSCOV:
1444 val = up->pcrlen;
1445 break;
1446
1447 default:
1448 return -ENOPROTOOPT;
1449 }
1450
1451 if (put_user(len, optlen))
1452 return -EFAULT;
1453 if (copy_to_user(optval, &val,len))
1454 return -EFAULT;
1455 return 0;
1456 }
1457
1458 int udp_getsockopt(struct sock *sk, int level, int optname,
1459 char __user *optval, int __user *optlen)
1460 {
1461 if (level == SOL_UDP || level == SOL_UDPLITE)
1462 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1463 return ip_getsockopt(sk, level, optname, optval, optlen);
1464 }
1465
1466 #ifdef CONFIG_COMPAT
1467 int compat_udp_getsockopt(struct sock *sk, int level, int optname,
1468 char __user *optval, int __user *optlen)
1469 {
1470 if (level == SOL_UDP || level == SOL_UDPLITE)
1471 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1472 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
1473 }
1474 #endif
1475 /**
1476 * udp_poll - wait for a UDP event.
1477 * @file - file struct
1478 * @sock - socket
1479 * @wait - poll table
1480 *
1481 * This is same as datagram poll, except for the special case of
1482 * blocking sockets. If application is using a blocking fd
1483 * and a packet with checksum error is in the queue;
1484 * then it could get return from select indicating data available
1485 * but then block when reading it. Add special case code
1486 * to work around these arguably broken applications.
1487 */
1488 unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1489 {
1490 unsigned int mask = datagram_poll(file, sock, wait);
1491 struct sock *sk = sock->sk;
1492 int is_lite = IS_UDPLITE(sk);
1493
1494 /* Check for false positives due to checksum errors */
1495 if ( (mask & POLLRDNORM) &&
1496 !(file->f_flags & O_NONBLOCK) &&
1497 !(sk->sk_shutdown & RCV_SHUTDOWN)){
1498 struct sk_buff_head *rcvq = &sk->sk_receive_queue;
1499 struct sk_buff *skb;
1500
1501 spin_lock_bh(&rcvq->lock);
1502 while ((skb = skb_peek(rcvq)) != NULL &&
1503 udp_lib_checksum_complete(skb)) {
1504 UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_lite);
1505 __skb_unlink(skb, rcvq);
1506 kfree_skb(skb);
1507 }
1508 spin_unlock_bh(&rcvq->lock);
1509
1510 /* nothing to see, move along */
1511 if (skb == NULL)
1512 mask &= ~(POLLIN | POLLRDNORM);
1513 }
1514
1515 return mask;
1516
1517 }
1518
1519 struct proto udp_prot = {
1520 .name = "UDP",
1521 .owner = THIS_MODULE,
1522 .close = udp_lib_close,
1523 .connect = ip4_datagram_connect,
1524 .disconnect = udp_disconnect,
1525 .ioctl = udp_ioctl,
1526 .destroy = udp_destroy_sock,
1527 .setsockopt = udp_setsockopt,
1528 .getsockopt = udp_getsockopt,
1529 .sendmsg = udp_sendmsg,
1530 .recvmsg = udp_recvmsg,
1531 .sendpage = udp_sendpage,
1532 .backlog_rcv = udp_queue_rcv_skb,
1533 .hash = udp_lib_hash,
1534 .unhash = udp_lib_unhash,
1535 .get_port = udp_v4_get_port,
1536 .obj_size = sizeof(struct udp_sock),
1537 #ifdef CONFIG_COMPAT
1538 .compat_setsockopt = compat_udp_setsockopt,
1539 .compat_getsockopt = compat_udp_getsockopt,
1540 #endif
1541 };
1542
1543 /* ------------------------------------------------------------------------ */
1544 #ifdef CONFIG_PROC_FS
1545
1546 static struct sock *udp_get_first(struct seq_file *seq)
1547 {
1548 struct sock *sk;
1549 struct udp_iter_state *state = seq->private;
1550
1551 for (state->bucket = 0; state->bucket < UDP_HTABLE_SIZE; ++state->bucket) {
1552 struct hlist_node *node;
1553 sk_for_each(sk, node, state->hashtable + state->bucket) {
1554 if (sk->sk_family == state->family)
1555 goto found;
1556 }
1557 }
1558 sk = NULL;
1559 found:
1560 return sk;
1561 }
1562
1563 static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
1564 {
1565 struct udp_iter_state *state = seq->private;
1566
1567 do {
1568 sk = sk_next(sk);
1569 try_again:
1570 ;
1571 } while (sk && sk->sk_family != state->family);
1572
1573 if (!sk && ++state->bucket < UDP_HTABLE_SIZE) {
1574 sk = sk_head(state->hashtable + state->bucket);
1575 goto try_again;
1576 }
1577 return sk;
1578 }
1579
1580 static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
1581 {
1582 struct sock *sk = udp_get_first(seq);
1583
1584 if (sk)
1585 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
1586 --pos;
1587 return pos ? NULL : sk;
1588 }
1589
1590 static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
1591 {
1592 read_lock(&udp_hash_lock);
1593 return *pos ? udp_get_idx(seq, *pos-1) : (void *)1;
1594 }
1595
1596 static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1597 {
1598 struct sock *sk;
1599
1600 if (v == (void *)1)
1601 sk = udp_get_idx(seq, 0);
1602 else
1603 sk = udp_get_next(seq, v);
1604
1605 ++*pos;
1606 return sk;
1607 }
1608
1609 static void udp_seq_stop(struct seq_file *seq, void *v)
1610 {
1611 read_unlock(&udp_hash_lock);
1612 }
1613
1614 static int udp_seq_open(struct inode *inode, struct file *file)
1615 {
1616 struct udp_seq_afinfo *afinfo = PDE(inode)->data;
1617 struct seq_file *seq;
1618 int rc = -ENOMEM;
1619 struct udp_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
1620
1621 if (!s)
1622 goto out;
1623 s->family = afinfo->family;
1624 s->hashtable = afinfo->hashtable;
1625 s->seq_ops.start = udp_seq_start;
1626 s->seq_ops.next = udp_seq_next;
1627 s->seq_ops.show = afinfo->seq_show;
1628 s->seq_ops.stop = udp_seq_stop;
1629
1630 rc = seq_open(file, &s->seq_ops);
1631 if (rc)
1632 goto out_kfree;
1633
1634 seq = file->private_data;
1635 seq->private = s;
1636 out:
1637 return rc;
1638 out_kfree:
1639 kfree(s);
1640 goto out;
1641 }
1642
1643 /* ------------------------------------------------------------------------ */
1644 int udp_proc_register(struct udp_seq_afinfo *afinfo)
1645 {
1646 struct proc_dir_entry *p;
1647 int rc = 0;
1648
1649 if (!afinfo)
1650 return -EINVAL;
1651 afinfo->seq_fops->owner = afinfo->owner;
1652 afinfo->seq_fops->open = udp_seq_open;
1653 afinfo->seq_fops->read = seq_read;
1654 afinfo->seq_fops->llseek = seq_lseek;
1655 afinfo->seq_fops->release = seq_release_private;
1656
1657 p = proc_net_fops_create(afinfo->name, S_IRUGO, afinfo->seq_fops);
1658 if (p)
1659 p->data = afinfo;
1660 else
1661 rc = -ENOMEM;
1662 return rc;
1663 }
1664
1665 void udp_proc_unregister(struct udp_seq_afinfo *afinfo)
1666 {
1667 if (!afinfo)
1668 return;
1669 proc_net_remove(afinfo->name);
1670 memset(afinfo->seq_fops, 0, sizeof(*afinfo->seq_fops));
1671 }
1672
1673 /* ------------------------------------------------------------------------ */
1674 static void udp4_format_sock(struct sock *sp, char *tmpbuf, int bucket)
1675 {
1676 struct inet_sock *inet = inet_sk(sp);
1677 __be32 dest = inet->daddr;
1678 __be32 src = inet->rcv_saddr;
1679 __u16 destp = ntohs(inet->dport);
1680 __u16 srcp = ntohs(inet->sport);
1681
1682 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
1683 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
1684 bucket, src, srcp, dest, destp, sp->sk_state,
1685 atomic_read(&sp->sk_wmem_alloc),
1686 atomic_read(&sp->sk_rmem_alloc),
1687 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
1688 atomic_read(&sp->sk_refcnt), sp);
1689 }
1690
1691 int udp4_seq_show(struct seq_file *seq, void *v)
1692 {
1693 if (v == SEQ_START_TOKEN)
1694 seq_printf(seq, "%-127s\n",
1695 " sl local_address rem_address st tx_queue "
1696 "rx_queue tr tm->when retrnsmt uid timeout "
1697 "inode");
1698 else {
1699 char tmpbuf[129];
1700 struct udp_iter_state *state = seq->private;
1701
1702 udp4_format_sock(v, tmpbuf, state->bucket);
1703 seq_printf(seq, "%-127s\n", tmpbuf);
1704 }
1705 return 0;
1706 }
1707
1708 /* ------------------------------------------------------------------------ */
1709 static struct file_operations udp4_seq_fops;
1710 static struct udp_seq_afinfo udp4_seq_afinfo = {
1711 .owner = THIS_MODULE,
1712 .name = "udp",
1713 .family = AF_INET,
1714 .hashtable = udp_hash,
1715 .seq_show = udp4_seq_show,
1716 .seq_fops = &udp4_seq_fops,
1717 };
1718
1719 int __init udp4_proc_init(void)
1720 {
1721 return udp_proc_register(&udp4_seq_afinfo);
1722 }
1723
1724 void udp4_proc_exit(void)
1725 {
1726 udp_proc_unregister(&udp4_seq_afinfo);
1727 }
1728 #endif /* CONFIG_PROC_FS */
1729
1730 EXPORT_SYMBOL(udp_disconnect);
1731 EXPORT_SYMBOL(udp_hash);
1732 EXPORT_SYMBOL(udp_hash_lock);
1733 EXPORT_SYMBOL(udp_ioctl);
1734 EXPORT_SYMBOL(udp_get_port);
1735 EXPORT_SYMBOL(udp_prot);
1736 EXPORT_SYMBOL(udp_sendmsg);
1737 EXPORT_SYMBOL(udp_lib_getsockopt);
1738 EXPORT_SYMBOL(udp_lib_setsockopt);
1739 EXPORT_SYMBOL(udp_poll);
1740
1741 #ifdef CONFIG_PROC_FS
1742 EXPORT_SYMBOL(udp_proc_register);
1743 EXPORT_SYMBOL(udp_proc_unregister);
1744 #endif
Linus' kernel tree