~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/ipv4/udp.c

Version: ~ [ linux-5.3-rc5 ] ~ [ linux-5.2.9 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.67 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.139 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.189 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.189 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.72 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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  * Authors:     Ross Biro
  9  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 10  *              Arnt Gulbrandsen, <agulbra@nvg.unit.no>
 11  *              Alan Cox, <alan@lxorguk.ukuu.org.uk>
 12  *              Hirokazu Takahashi, <taka@valinux.co.jp>
 13  *
 14  * Fixes:
 15  *              Alan Cox        :       verify_area() calls
 16  *              Alan Cox        :       stopped close while in use off icmp
 17  *                                      messages. Not a fix but a botch that
 18  *                                      for udp at least is 'valid'.
 19  *              Alan Cox        :       Fixed icmp handling properly
 20  *              Alan Cox        :       Correct error for oversized datagrams
 21  *              Alan Cox        :       Tidied select() semantics.
 22  *              Alan Cox        :       udp_err() fixed properly, also now
 23  *                                      select and read wake correctly on errors
 24  *              Alan Cox        :       udp_send verify_area moved to avoid mem leak
 25  *              Alan Cox        :       UDP can count its memory
 26  *              Alan Cox        :       send to an unknown connection causes
 27  *                                      an ECONNREFUSED off the icmp, but
 28  *                                      does NOT close.
 29  *              Alan Cox        :       Switched to new sk_buff handlers. No more backlog!
 30  *              Alan Cox        :       Using generic datagram code. Even smaller and the PEEK
 31  *                                      bug no longer crashes it.
 32  *              Fred Van Kempen :       Net2e support for sk->broadcast.
 33  *              Alan Cox        :       Uses skb_free_datagram
 34  *              Alan Cox        :       Added get/set sockopt support.
 35  *              Alan Cox        :       Broadcasting without option set returns EACCES.
 36  *              Alan Cox        :       No wakeup calls. Instead we now use the callbacks.
 37  *              Alan Cox        :       Use ip_tos and ip_ttl
 38  *              Alan Cox        :       SNMP Mibs
 39  *              Alan Cox        :       MSG_DONTROUTE, and 0.0.0.0 support.
 40  *              Matt Dillon     :       UDP length checks.
 41  *              Alan Cox        :       Smarter af_inet used properly.
 42  *              Alan Cox        :       Use new kernel side addressing.
 43  *              Alan Cox        :       Incorrect return on truncated datagram receive.
 44  *      Arnt Gulbrandsen        :       New udp_send and stuff
 45  *              Alan Cox        :       Cache last socket
 46  *              Alan Cox        :       Route cache
 47  *              Jon Peatfield   :       Minor efficiency fix to sendto().
 48  *              Mike Shaver     :       RFC1122 checks.
 49  *              Alan Cox        :       Nonblocking error fix.
 50  *      Willy Konynenberg       :       Transparent proxying support.
 51  *              Mike McLagan    :       Routing by source
 52  *              David S. Miller :       New socket lookup architecture.
 53  *                                      Last socket cache retained as it
 54  *                                      does have a high hit rate.
 55  *              Olaf Kirch      :       Don't linearise iovec on sendmsg.
 56  *              Andi Kleen      :       Some cleanups, cache destination entry
 57  *                                      for connect.
 58  *      Vitaly E. Lavrov        :       Transparent proxy revived after year coma.
 59  *              Melvin Smith    :       Check msg_name not msg_namelen in sendto(),
 60  *                                      return ENOTCONN for unconnected sockets (POSIX)
 61  *              Janos Farkas    :       don't deliver multi/broadcasts to a different
 62  *                                      bound-to-device socket
 63  *      Hirokazu Takahashi      :       HW checksumming for outgoing UDP
 64  *                                      datagrams.
 65  *      Hirokazu Takahashi      :       sendfile() on UDP works now.
 66  *              Arnaldo C. Melo :       convert /proc/net/udp to seq_file
 67  *      YOSHIFUJI Hideaki @USAGI and:   Support IPV6_V6ONLY socket option, which
 68  *      Alexey Kuznetsov:               allow both IPv4 and IPv6 sockets to bind
 69  *                                      a single port at the same time.
 70  *      Derek Atkins <derek@ihtfp.com>: Add Encapulation Support
 71  *      James Chapman           :       Add L2TP encapsulation type.
 72  *
 73  *
 74  *              This program is free software; you can redistribute it and/or
 75  *              modify it under the terms of the GNU General Public License
 76  *              as published by the Free Software Foundation; either version
 77  *              2 of the License, or (at your option) any later version.
 78  */
 79 
 80 #define pr_fmt(fmt) "UDP: " fmt
 81 
 82 #include <asm/uaccess.h>
 83 #include <asm/ioctls.h>
 84 #include <linux/bootmem.h>
 85 #include <linux/highmem.h>
 86 #include <linux/swap.h>
 87 #include <linux/types.h>
 88 #include <linux/fcntl.h>
 89 #include <linux/module.h>
 90 #include <linux/socket.h>
 91 #include <linux/sockios.h>
 92 #include <linux/igmp.h>
 93 #include <linux/in.h>
 94 #include <linux/errno.h>
 95 #include <linux/timer.h>
 96 #include <linux/mm.h>
 97 #include <linux/inet.h>
 98 #include <linux/netdevice.h>
 99 #include <linux/slab.h>
100 #include <net/tcp_states.h>
101 #include <linux/skbuff.h>
102 #include <linux/proc_fs.h>
103 #include <linux/seq_file.h>
104 #include <net/net_namespace.h>
105 #include <net/icmp.h>
106 #include <net/route.h>
107 #include <net/checksum.h>
108 #include <net/xfrm.h>
109 #include <trace/events/udp.h>
110 #include <linux/static_key.h>
111 #include <trace/events/skb.h>
112 #include <net/busy_poll.h>
113 #include "udp_impl.h"
114 
115 struct udp_table udp_table __read_mostly;
116 EXPORT_SYMBOL(udp_table);
117 
118 long sysctl_udp_mem[3] __read_mostly;
119 EXPORT_SYMBOL(sysctl_udp_mem);
120 
121 int sysctl_udp_rmem_min __read_mostly;
122 EXPORT_SYMBOL(sysctl_udp_rmem_min);
123 
124 int sysctl_udp_wmem_min __read_mostly;
125 EXPORT_SYMBOL(sysctl_udp_wmem_min);
126 
127 atomic_long_t udp_memory_allocated;
128 EXPORT_SYMBOL(udp_memory_allocated);
129 
130 #define MAX_UDP_PORTS 65536
131 #define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN)
132 
133 static int udp_lib_lport_inuse(struct net *net, __u16 num,
134                                const struct udp_hslot *hslot,
135                                unsigned long *bitmap,
136                                struct sock *sk,
137                                int (*saddr_comp)(const struct sock *sk1,
138                                                  const struct sock *sk2),
139                                unsigned int log)
140 {
141         struct sock *sk2;
142         struct hlist_nulls_node *node;
143         kuid_t uid = sock_i_uid(sk);
144 
145         sk_nulls_for_each(sk2, node, &hslot->head)
146                 if (net_eq(sock_net(sk2), net) &&
147                     sk2 != sk &&
148                     (bitmap || udp_sk(sk2)->udp_port_hash == num) &&
149                     (!sk2->sk_reuse || !sk->sk_reuse) &&
150                     (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
151                      sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
152                     (!sk2->sk_reuseport || !sk->sk_reuseport ||
153                       !uid_eq(uid, sock_i_uid(sk2))) &&
154                     (*saddr_comp)(sk, sk2)) {
155                         if (bitmap)
156                                 __set_bit(udp_sk(sk2)->udp_port_hash >> log,
157                                           bitmap);
158                         else
159                                 return 1;
160                 }
161         return 0;
162 }
163 
164 /*
165  * Note: we still hold spinlock of primary hash chain, so no other writer
166  * can insert/delete a socket with local_port == num
167  */
168 static int udp_lib_lport_inuse2(struct net *net, __u16 num,
169                                struct udp_hslot *hslot2,
170                                struct sock *sk,
171                                int (*saddr_comp)(const struct sock *sk1,
172                                                  const struct sock *sk2))
173 {
174         struct sock *sk2;
175         struct hlist_nulls_node *node;
176         kuid_t uid = sock_i_uid(sk);
177         int res = 0;
178 
179         spin_lock(&hslot2->lock);
180         udp_portaddr_for_each_entry(sk2, node, &hslot2->head)
181                 if (net_eq(sock_net(sk2), net) &&
182                     sk2 != sk &&
183                     (udp_sk(sk2)->udp_port_hash == num) &&
184                     (!sk2->sk_reuse || !sk->sk_reuse) &&
185                     (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
186                      sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
187                     (!sk2->sk_reuseport || !sk->sk_reuseport ||
188                       !uid_eq(uid, sock_i_uid(sk2))) &&
189                     (*saddr_comp)(sk, sk2)) {
190                         res = 1;
191                         break;
192                 }
193         spin_unlock(&hslot2->lock);
194         return res;
195 }
196 
197 /**
198  *  udp_lib_get_port  -  UDP/-Lite port lookup for IPv4 and IPv6
199  *
200  *  @sk:          socket struct in question
201  *  @snum:        port number to look up
202  *  @saddr_comp:  AF-dependent comparison of bound local IP addresses
203  *  @hash2_nulladdr: AF-dependent hash value in secondary hash chains,
204  *                   with NULL address
205  */
206 int udp_lib_get_port(struct sock *sk, unsigned short snum,
207                        int (*saddr_comp)(const struct sock *sk1,
208                                          const struct sock *sk2),
209                      unsigned int hash2_nulladdr)
210 {
211         struct udp_hslot *hslot, *hslot2;
212         struct udp_table *udptable = sk->sk_prot->h.udp_table;
213         int    error = 1;
214         struct net *net = sock_net(sk);
215 
216         if (!snum) {
217                 int low, high, remaining;
218                 unsigned int rand;
219                 unsigned short first, last;
220                 DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN);
221 
222                 inet_get_local_port_range(&low, &high);
223                 remaining = (high - low) + 1;
224 
225                 rand = net_random();
226                 first = (((u64)rand * remaining) >> 32) + low;
227                 /*
228                  * force rand to be an odd multiple of UDP_HTABLE_SIZE
229                  */
230                 rand = (rand | 1) * (udptable->mask + 1);
231                 last = first + udptable->mask + 1;
232                 do {
233                         hslot = udp_hashslot(udptable, net, first);
234                         bitmap_zero(bitmap, PORTS_PER_CHAIN);
235                         spin_lock_bh(&hslot->lock);
236                         udp_lib_lport_inuse(net, snum, hslot, bitmap, sk,
237                                             saddr_comp, udptable->log);
238 
239                         snum = first;
240                         /*
241                          * Iterate on all possible values of snum for this hash.
242                          * Using steps of an odd multiple of UDP_HTABLE_SIZE
243                          * give us randomization and full range coverage.
244                          */
245                         do {
246                                 if (low <= snum && snum <= high &&
247                                     !test_bit(snum >> udptable->log, bitmap) &&
248                                     !inet_is_reserved_local_port(snum))
249                                         goto found;
250                                 snum += rand;
251                         } while (snum != first);
252                         spin_unlock_bh(&hslot->lock);
253                 } while (++first != last);
254                 goto fail;
255         } else {
256                 hslot = udp_hashslot(udptable, net, snum);
257                 spin_lock_bh(&hslot->lock);
258                 if (hslot->count > 10) {
259                         int exist;
260                         unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum;
261 
262                         slot2          &= udptable->mask;
263                         hash2_nulladdr &= udptable->mask;
264 
265                         hslot2 = udp_hashslot2(udptable, slot2);
266                         if (hslot->count < hslot2->count)
267                                 goto scan_primary_hash;
268 
269                         exist = udp_lib_lport_inuse2(net, snum, hslot2,
270                                                      sk, saddr_comp);
271                         if (!exist && (hash2_nulladdr != slot2)) {
272                                 hslot2 = udp_hashslot2(udptable, hash2_nulladdr);
273                                 exist = udp_lib_lport_inuse2(net, snum, hslot2,
274                                                              sk, saddr_comp);
275                         }
276                         if (exist)
277                                 goto fail_unlock;
278                         else
279                                 goto found;
280                 }
281 scan_primary_hash:
282                 if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk,
283                                         saddr_comp, 0))
284                         goto fail_unlock;
285         }
286 found:
287         inet_sk(sk)->inet_num = snum;
288         udp_sk(sk)->udp_port_hash = snum;
289         udp_sk(sk)->udp_portaddr_hash ^= snum;
290         if (sk_unhashed(sk)) {
291                 sk_nulls_add_node_rcu(sk, &hslot->head);
292                 hslot->count++;
293                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
294 
295                 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
296                 spin_lock(&hslot2->lock);
297                 hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
298                                          &hslot2->head);
299                 hslot2->count++;
300                 spin_unlock(&hslot2->lock);
301         }
302         error = 0;
303 fail_unlock:
304         spin_unlock_bh(&hslot->lock);
305 fail:
306         return error;
307 }
308 EXPORT_SYMBOL(udp_lib_get_port);
309 
310 static int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2)
311 {
312         struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
313 
314         return  (!ipv6_only_sock(sk2)  &&
315                  (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr ||
316                    inet1->inet_rcv_saddr == inet2->inet_rcv_saddr));
317 }
318 
319 static unsigned int udp4_portaddr_hash(struct net *net, __be32 saddr,
320                                        unsigned int port)
321 {
322         return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
323 }
324 
325 int udp_v4_get_port(struct sock *sk, unsigned short snum)
326 {
327         unsigned int hash2_nulladdr =
328                 udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum);
329         unsigned int hash2_partial =
330                 udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0);
331 
332         /* precompute partial secondary hash */
333         udp_sk(sk)->udp_portaddr_hash = hash2_partial;
334         return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr);
335 }
336 
337 static inline int compute_score(struct sock *sk, struct net *net, __be32 saddr,
338                          unsigned short hnum,
339                          __be16 sport, __be32 daddr, __be16 dport, int dif)
340 {
341         int score = -1;
342 
343         if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
344                         !ipv6_only_sock(sk)) {
345                 struct inet_sock *inet = inet_sk(sk);
346 
347                 score = (sk->sk_family == PF_INET ? 2 : 1);
348                 if (inet->inet_rcv_saddr) {
349                         if (inet->inet_rcv_saddr != daddr)
350                                 return -1;
351                         score += 4;
352                 }
353                 if (inet->inet_daddr) {
354                         if (inet->inet_daddr != saddr)
355                                 return -1;
356                         score += 4;
357                 }
358                 if (inet->inet_dport) {
359                         if (inet->inet_dport != sport)
360                                 return -1;
361                         score += 4;
362                 }
363                 if (sk->sk_bound_dev_if) {
364                         if (sk->sk_bound_dev_if != dif)
365                                 return -1;
366                         score += 4;
367                 }
368         }
369         return score;
370 }
371 
372 /*
373  * In this second variant, we check (daddr, dport) matches (inet_rcv_sadd, inet_num)
374  */
375 static inline int compute_score2(struct sock *sk, struct net *net,
376                                  __be32 saddr, __be16 sport,
377                                  __be32 daddr, unsigned int hnum, int dif)
378 {
379         int score = -1;
380 
381         if (net_eq(sock_net(sk), net) && !ipv6_only_sock(sk)) {
382                 struct inet_sock *inet = inet_sk(sk);
383 
384                 if (inet->inet_rcv_saddr != daddr)
385                         return -1;
386                 if (inet->inet_num != hnum)
387                         return -1;
388 
389                 score = (sk->sk_family == PF_INET ? 2 : 1);
390                 if (inet->inet_daddr) {
391                         if (inet->inet_daddr != saddr)
392                                 return -1;
393                         score += 4;
394                 }
395                 if (inet->inet_dport) {
396                         if (inet->inet_dport != sport)
397                                 return -1;
398                         score += 4;
399                 }
400                 if (sk->sk_bound_dev_if) {
401                         if (sk->sk_bound_dev_if != dif)
402                                 return -1;
403                         score += 4;
404                 }
405         }
406         return score;
407 }
408 
409 
410 /* called with read_rcu_lock() */
411 static struct sock *udp4_lib_lookup2(struct net *net,
412                 __be32 saddr, __be16 sport,
413                 __be32 daddr, unsigned int hnum, int dif,
414                 struct udp_hslot *hslot2, unsigned int slot2)
415 {
416         struct sock *sk, *result;
417         struct hlist_nulls_node *node;
418         int score, badness, matches = 0, reuseport = 0;
419         u32 hash = 0;
420 
421 begin:
422         result = NULL;
423         badness = 0;
424         udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
425                 score = compute_score2(sk, net, saddr, sport,
426                                       daddr, hnum, dif);
427                 if (score > badness) {
428                         result = sk;
429                         badness = score;
430                         reuseport = sk->sk_reuseport;
431                         if (reuseport) {
432                                 hash = inet_ehashfn(net, daddr, hnum,
433                                                     saddr, sport);
434                                 matches = 1;
435                         }
436                 } else if (score == badness && reuseport) {
437                         matches++;
438                         if (((u64)hash * matches) >> 32 == 0)
439                                 result = sk;
440                         hash = next_pseudo_random32(hash);
441                 }
442         }
443         /*
444          * if the nulls value we got at the end of this lookup is
445          * not the expected one, we must restart lookup.
446          * We probably met an item that was moved to another chain.
447          */
448         if (get_nulls_value(node) != slot2)
449                 goto begin;
450         if (result) {
451                 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
452                         result = NULL;
453                 else if (unlikely(compute_score2(result, net, saddr, sport,
454                                   daddr, hnum, dif) < badness)) {
455                         sock_put(result);
456                         goto begin;
457                 }
458         }
459         return result;
460 }
461 
462 /* UDP is nearly always wildcards out the wazoo, it makes no sense to try
463  * harder than this. -DaveM
464  */
465 struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
466                 __be16 sport, __be32 daddr, __be16 dport,
467                 int dif, struct udp_table *udptable)
468 {
469         struct sock *sk, *result;
470         struct hlist_nulls_node *node;
471         unsigned short hnum = ntohs(dport);
472         unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
473         struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
474         int score, badness, matches = 0, reuseport = 0;
475         u32 hash = 0;
476 
477         rcu_read_lock();
478         if (hslot->count > 10) {
479                 hash2 = udp4_portaddr_hash(net, daddr, hnum);
480                 slot2 = hash2 & udptable->mask;
481                 hslot2 = &udptable->hash2[slot2];
482                 if (hslot->count < hslot2->count)
483                         goto begin;
484 
485                 result = udp4_lib_lookup2(net, saddr, sport,
486                                           daddr, hnum, dif,
487                                           hslot2, slot2);
488                 if (!result) {
489                         hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum);
490                         slot2 = hash2 & udptable->mask;
491                         hslot2 = &udptable->hash2[slot2];
492                         if (hslot->count < hslot2->count)
493                                 goto begin;
494 
495                         result = udp4_lib_lookup2(net, saddr, sport,
496                                                   htonl(INADDR_ANY), hnum, dif,
497                                                   hslot2, slot2);
498                 }
499                 rcu_read_unlock();
500                 return result;
501         }
502 begin:
503         result = NULL;
504         badness = 0;
505         sk_nulls_for_each_rcu(sk, node, &hslot->head) {
506                 score = compute_score(sk, net, saddr, hnum, sport,
507                                       daddr, dport, dif);
508                 if (score > badness) {
509                         result = sk;
510                         badness = score;
511                         reuseport = sk->sk_reuseport;
512                         if (reuseport) {
513                                 hash = inet_ehashfn(net, daddr, hnum,
514                                                     saddr, sport);
515                                 matches = 1;
516                         }
517                 } else if (score == badness && reuseport) {
518                         matches++;
519                         if (((u64)hash * matches) >> 32 == 0)
520                                 result = sk;
521                         hash = next_pseudo_random32(hash);
522                 }
523         }
524         /*
525          * if the nulls value we got at the end of this lookup is
526          * not the expected one, we must restart lookup.
527          * We probably met an item that was moved to another chain.
528          */
529         if (get_nulls_value(node) != slot)
530                 goto begin;
531 
532         if (result) {
533                 if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
534                         result = NULL;
535                 else if (unlikely(compute_score(result, net, saddr, hnum, sport,
536                                   daddr, dport, dif) < badness)) {
537                         sock_put(result);
538                         goto begin;
539                 }
540         }
541         rcu_read_unlock();
542         return result;
543 }
544 EXPORT_SYMBOL_GPL(__udp4_lib_lookup);
545 
546 static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
547                                                  __be16 sport, __be16 dport,
548                                                  struct udp_table *udptable)
549 {
550         struct sock *sk;
551         const struct iphdr *iph = ip_hdr(skb);
552 
553         if (unlikely(sk = skb_steal_sock(skb)))
554                 return sk;
555         else
556                 return __udp4_lib_lookup(dev_net(skb_dst(skb)->dev), iph->saddr, sport,
557                                          iph->daddr, dport, inet_iif(skb),
558                                          udptable);
559 }
560 
561 struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
562                              __be32 daddr, __be16 dport, int dif)
563 {
564         return __udp4_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
565 }
566 EXPORT_SYMBOL_GPL(udp4_lib_lookup);
567 
568 static inline struct sock *udp_v4_mcast_next(struct net *net, struct sock *sk,
569                                              __be16 loc_port, __be32 loc_addr,
570                                              __be16 rmt_port, __be32 rmt_addr,
571                                              int dif)
572 {
573         struct hlist_nulls_node *node;
574         struct sock *s = sk;
575         unsigned short hnum = ntohs(loc_port);
576 
577         sk_nulls_for_each_from(s, node) {
578                 struct inet_sock *inet = inet_sk(s);
579 
580                 if (!net_eq(sock_net(s), net) ||
581                     udp_sk(s)->udp_port_hash != hnum ||
582                     (inet->inet_daddr && inet->inet_daddr != rmt_addr) ||
583                     (inet->inet_dport != rmt_port && inet->inet_dport) ||
584                     (inet->inet_rcv_saddr &&
585                      inet->inet_rcv_saddr != loc_addr) ||
586                     ipv6_only_sock(s) ||
587                     (s->sk_bound_dev_if && s->sk_bound_dev_if != dif))
588                         continue;
589                 if (!ip_mc_sf_allow(s, loc_addr, rmt_addr, dif))
590                         continue;
591                 goto found;
592         }
593         s = NULL;
594 found:
595         return s;
596 }
597 
598 /*
599  * This routine is called by the ICMP module when it gets some
600  * sort of error condition.  If err < 0 then the socket should
601  * be closed and the error returned to the user.  If err > 0
602  * it's just the icmp type << 8 | icmp code.
603  * Header points to the ip header of the error packet. We move
604  * on past this. Then (as it used to claim before adjustment)
605  * header points to the first 8 bytes of the udp header.  We need
606  * to find the appropriate port.
607  */
608 
609 void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
610 {
611         struct inet_sock *inet;
612         const struct iphdr *iph = (const struct iphdr *)skb->data;
613         struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2));
614         const int type = icmp_hdr(skb)->type;
615         const int code = icmp_hdr(skb)->code;
616         struct sock *sk;
617         int harderr;
618         int err;
619         struct net *net = dev_net(skb->dev);
620 
621         sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
622                         iph->saddr, uh->source, skb->dev->ifindex, udptable);
623         if (sk == NULL) {
624                 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
625                 return; /* No socket for error */
626         }
627 
628         err = 0;
629         harderr = 0;
630         inet = inet_sk(sk);
631 
632         switch (type) {
633         default:
634         case ICMP_TIME_EXCEEDED:
635                 err = EHOSTUNREACH;
636                 break;
637         case ICMP_SOURCE_QUENCH:
638                 goto out;
639         case ICMP_PARAMETERPROB:
640                 err = EPROTO;
641                 harderr = 1;
642                 break;
643         case ICMP_DEST_UNREACH:
644                 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
645                         ipv4_sk_update_pmtu(skb, sk, info);
646                         if (inet->pmtudisc != IP_PMTUDISC_DONT) {
647                                 err = EMSGSIZE;
648                                 harderr = 1;
649                                 break;
650                         }
651                         goto out;
652                 }
653                 err = EHOSTUNREACH;
654                 if (code <= NR_ICMP_UNREACH) {
655                         harderr = icmp_err_convert[code].fatal;
656                         err = icmp_err_convert[code].errno;
657                 }
658                 break;
659         case ICMP_REDIRECT:
660                 ipv4_sk_redirect(skb, sk);
661                 break;
662         }
663 
664         /*
665          *      RFC1122: OK.  Passes ICMP errors back to application, as per
666          *      4.1.3.3.
667          */
668         if (!inet->recverr) {
669                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
670                         goto out;
671         } else
672                 ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
673 
674         sk->sk_err = err;
675         sk->sk_error_report(sk);
676 out:
677         sock_put(sk);
678 }
679 
680 void udp_err(struct sk_buff *skb, u32 info)
681 {
682         __udp4_lib_err(skb, info, &udp_table);
683 }
684 
685 /*
686  * Throw away all pending data and cancel the corking. Socket is locked.
687  */
688 void udp_flush_pending_frames(struct sock *sk)
689 {
690         struct udp_sock *up = udp_sk(sk);
691 
692         if (up->pending) {
693                 up->len = 0;
694                 up->pending = 0;
695                 ip_flush_pending_frames(sk);
696         }
697 }
698 EXPORT_SYMBOL(udp_flush_pending_frames);
699 
700 /**
701  *      udp4_hwcsum  -  handle outgoing HW checksumming
702  *      @skb:   sk_buff containing the filled-in UDP header
703  *              (checksum field must be zeroed out)
704  *      @src:   source IP address
705  *      @dst:   destination IP address
706  */
707 static void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst)
708 {
709         struct udphdr *uh = udp_hdr(skb);
710         struct sk_buff *frags = skb_shinfo(skb)->frag_list;
711         int offset = skb_transport_offset(skb);
712         int len = skb->len - offset;
713         int hlen = len;
714         __wsum csum = 0;
715 
716         if (!frags) {
717                 /*
718                  * Only one fragment on the socket.
719                  */
720                 skb->csum_start = skb_transport_header(skb) - skb->head;
721                 skb->csum_offset = offsetof(struct udphdr, check);
722                 uh->check = ~csum_tcpudp_magic(src, dst, len,
723                                                IPPROTO_UDP, 0);
724         } else {
725                 /*
726                  * HW-checksum won't work as there are two or more
727                  * fragments on the socket so that all csums of sk_buffs
728                  * should be together
729                  */
730                 do {
731                         csum = csum_add(csum, frags->csum);
732                         hlen -= frags->len;
733                 } while ((frags = frags->next));
734 
735                 csum = skb_checksum(skb, offset, hlen, csum);
736                 skb->ip_summed = CHECKSUM_NONE;
737 
738                 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
739                 if (uh->check == 0)
740                         uh->check = CSUM_MANGLED_0;
741         }
742 }
743 
744 static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
745 {
746         struct sock *sk = skb->sk;
747         struct inet_sock *inet = inet_sk(sk);
748         struct udphdr *uh;
749         int err = 0;
750         int is_udplite = IS_UDPLITE(sk);
751         int offset = skb_transport_offset(skb);
752         int len = skb->len - offset;
753         __wsum csum = 0;
754 
755         /*
756          * Create a UDP header
757          */
758         uh = udp_hdr(skb);
759         uh->source = inet->inet_sport;
760         uh->dest = fl4->fl4_dport;
761         uh->len = htons(len);
762         uh->check = 0;
763 
764         if (is_udplite)                                  /*     UDP-Lite      */
765                 csum = udplite_csum(skb);
766 
767         else if (sk->sk_no_check == UDP_CSUM_NOXMIT) {   /* UDP csum disabled */
768 
769                 skb->ip_summed = CHECKSUM_NONE;
770                 goto send;
771 
772         } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
773 
774                 udp4_hwcsum(skb, fl4->saddr, fl4->daddr);
775                 goto send;
776 
777         } else
778                 csum = udp_csum(skb);
779 
780         /* add protocol-dependent pseudo-header */
781         uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len,
782                                       sk->sk_protocol, csum);
783         if (uh->check == 0)
784                 uh->check = CSUM_MANGLED_0;
785 
786 send:
787         err = ip_send_skb(sock_net(sk), skb);
788         if (err) {
789                 if (err == -ENOBUFS && !inet->recverr) {
790                         UDP_INC_STATS_USER(sock_net(sk),
791                                            UDP_MIB_SNDBUFERRORS, is_udplite);
792                         err = 0;
793                 }
794         } else
795                 UDP_INC_STATS_USER(sock_net(sk),
796                                    UDP_MIB_OUTDATAGRAMS, is_udplite);
797         return err;
798 }
799 
800 /*
801  * Push out all pending data as one UDP datagram. Socket is locked.
802  */
803 int udp_push_pending_frames(struct sock *sk)
804 {
805         struct udp_sock  *up = udp_sk(sk);
806         struct inet_sock *inet = inet_sk(sk);
807         struct flowi4 *fl4 = &inet->cork.fl.u.ip4;
808         struct sk_buff *skb;
809         int err = 0;
810 
811         skb = ip_finish_skb(sk, fl4);
812         if (!skb)
813                 goto out;
814 
815         err = udp_send_skb(skb, fl4);
816 
817 out:
818         up->len = 0;
819         up->pending = 0;
820         return err;
821 }
822 EXPORT_SYMBOL(udp_push_pending_frames);
823 
824 int udp_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
825                 size_t len)
826 {
827         struct inet_sock *inet = inet_sk(sk);
828         struct udp_sock *up = udp_sk(sk);
829         struct flowi4 fl4_stack;
830         struct flowi4 *fl4;
831         int ulen = len;
832         struct ipcm_cookie ipc;
833         struct rtable *rt = NULL;
834         int free = 0;
835         int connected = 0;
836         __be32 daddr, faddr, saddr;
837         __be16 dport;
838         u8  tos;
839         int err, is_udplite = IS_UDPLITE(sk);
840         int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
841         int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
842         struct sk_buff *skb;
843         struct ip_options_data opt_copy;
844 
845         if (len > 0xFFFF)
846                 return -EMSGSIZE;
847 
848         /*
849          *      Check the flags.
850          */
851 
852         if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */
853                 return -EOPNOTSUPP;
854 
855         ipc.opt = NULL;
856         ipc.tx_flags = 0;
857 
858         getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
859 
860         fl4 = &inet->cork.fl.u.ip4;
861         if (up->pending) {
862                 /*
863                  * There are pending frames.
864                  * The socket lock must be held while it's corked.
865                  */
866                 lock_sock(sk);
867                 if (likely(up->pending)) {
868                         if (unlikely(up->pending != AF_INET)) {
869                                 release_sock(sk);
870                                 return -EINVAL;
871                         }
872                         goto do_append_data;
873                 }
874                 release_sock(sk);
875         }
876         ulen += sizeof(struct udphdr);
877 
878         /*
879          *      Get and verify the address.
880          */
881         if (msg->msg_name) {
882                 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
883                 if (msg->msg_namelen < sizeof(*usin))
884                         return -EINVAL;
885                 if (usin->sin_family != AF_INET) {
886                         if (usin->sin_family != AF_UNSPEC)
887                                 return -EAFNOSUPPORT;
888                 }
889 
890                 daddr = usin->sin_addr.s_addr;
891                 dport = usin->sin_port;
892                 if (dport == 0)
893                         return -EINVAL;
894         } else {
895                 if (sk->sk_state != TCP_ESTABLISHED)
896                         return -EDESTADDRREQ;
897                 daddr = inet->inet_daddr;
898                 dport = inet->inet_dport;
899                 /* Open fast path for connected socket.
900                    Route will not be used, if at least one option is set.
901                  */
902                 connected = 1;
903         }
904         ipc.addr = inet->inet_saddr;
905 
906         ipc.oif = sk->sk_bound_dev_if;
907 
908         sock_tx_timestamp(sk, &ipc.tx_flags);
909 
910         if (msg->msg_controllen) {
911                 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
912                 if (err)
913                         return err;
914                 if (ipc.opt)
915                         free = 1;
916                 connected = 0;
917         }
918         if (!ipc.opt) {
919                 struct ip_options_rcu *inet_opt;
920 
921                 rcu_read_lock();
922                 inet_opt = rcu_dereference(inet->inet_opt);
923                 if (inet_opt) {
924                         memcpy(&opt_copy, inet_opt,
925                                sizeof(*inet_opt) + inet_opt->opt.optlen);
926                         ipc.opt = &opt_copy.opt;
927                 }
928                 rcu_read_unlock();
929         }
930 
931         saddr = ipc.addr;
932         ipc.addr = faddr = daddr;
933 
934         if (ipc.opt && ipc.opt->opt.srr) {
935                 if (!daddr)
936                         return -EINVAL;
937                 faddr = ipc.opt->opt.faddr;
938                 connected = 0;
939         }
940         tos = RT_TOS(inet->tos);
941         if (sock_flag(sk, SOCK_LOCALROUTE) ||
942             (msg->msg_flags & MSG_DONTROUTE) ||
943             (ipc.opt && ipc.opt->opt.is_strictroute)) {
944                 tos |= RTO_ONLINK;
945                 connected = 0;
946         }
947 
948         if (ipv4_is_multicast(daddr)) {
949                 if (!ipc.oif)
950                         ipc.oif = inet->mc_index;
951                 if (!saddr)
952                         saddr = inet->mc_addr;
953                 connected = 0;
954         } else if (!ipc.oif)
955                 ipc.oif = inet->uc_index;
956 
957         if (connected)
958                 rt = (struct rtable *)sk_dst_check(sk, 0);
959 
960         if (rt == NULL) {
961                 struct net *net = sock_net(sk);
962 
963                 fl4 = &fl4_stack;
964                 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
965                                    RT_SCOPE_UNIVERSE, sk->sk_protocol,
966                                    inet_sk_flowi_flags(sk)|FLOWI_FLAG_CAN_SLEEP,
967                                    faddr, saddr, dport, inet->inet_sport);
968 
969                 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
970                 rt = ip_route_output_flow(net, fl4, sk);
971                 if (IS_ERR(rt)) {
972                         err = PTR_ERR(rt);
973                         rt = NULL;
974                         if (err == -ENETUNREACH)
975                                 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
976                         goto out;
977                 }
978 
979                 err = -EACCES;
980                 if ((rt->rt_flags & RTCF_BROADCAST) &&
981                     !sock_flag(sk, SOCK_BROADCAST))
982                         goto out;
983                 if (connected)
984                         sk_dst_set(sk, dst_clone(&rt->dst));
985         }
986 
987         if (msg->msg_flags&MSG_CONFIRM)
988                 goto do_confirm;
989 back_from_confirm:
990 
991         saddr = fl4->saddr;
992         if (!ipc.addr)
993                 daddr = ipc.addr = fl4->daddr;
994 
995         /* Lockless fast path for the non-corking case. */
996         if (!corkreq) {
997                 skb = ip_make_skb(sk, fl4, getfrag, msg->msg_iov, ulen,
998                                   sizeof(struct udphdr), &ipc, &rt,
999                                   msg->msg_flags);
1000                 err = PTR_ERR(skb);
1001                 if (!IS_ERR_OR_NULL(skb))
1002                         err = udp_send_skb(skb, fl4);
1003                 goto out;
1004         }
1005 
1006         lock_sock(sk);
1007         if (unlikely(up->pending)) {
1008                 /* The socket is already corked while preparing it. */
1009                 /* ... which is an evident application bug. --ANK */
1010                 release_sock(sk);
1011 
1012                 LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("cork app bug 2\n"));
1013                 err = -EINVAL;
1014                 goto out;
1015         }
1016         /*
1017          *      Now cork the socket to pend data.
1018          */
1019         fl4 = &inet->cork.fl.u.ip4;
1020         fl4->daddr = daddr;
1021         fl4->saddr = saddr;
1022         fl4->fl4_dport = dport;
1023         fl4->fl4_sport = inet->inet_sport;
1024         up->pending = AF_INET;
1025 
1026 do_append_data:
1027         up->len += ulen;
1028         err = ip_append_data(sk, fl4, getfrag, msg->msg_iov, ulen,
1029                              sizeof(struct udphdr), &ipc, &rt,
1030                              corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
1031         if (err)
1032                 udp_flush_pending_frames(sk);
1033         else if (!corkreq)
1034                 err = udp_push_pending_frames(sk);
1035         else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1036                 up->pending = 0;
1037         release_sock(sk);
1038 
1039 out:
1040         ip_rt_put(rt);
1041         if (free)
1042                 kfree(ipc.opt);
1043         if (!err)
1044                 return len;
1045         /*
1046          * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space.  Reporting
1047          * ENOBUFS might not be good (it's not tunable per se), but otherwise
1048          * we don't have a good statistic (IpOutDiscards but it can be too many
1049          * things).  We could add another new stat but at least for now that
1050          * seems like overkill.
1051          */
1052         if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1053                 UDP_INC_STATS_USER(sock_net(sk),
1054                                 UDP_MIB_SNDBUFERRORS, is_udplite);
1055         }
1056         return err;
1057 
1058 do_confirm:
1059         dst_confirm(&rt->dst);
1060         if (!(msg->msg_flags&MSG_PROBE) || len)
1061                 goto back_from_confirm;
1062         err = 0;
1063         goto out;
1064 }
1065 EXPORT_SYMBOL(udp_sendmsg);
1066 
1067 int udp_sendpage(struct sock *sk, struct page *page, int offset,
1068                  size_t size, int flags)
1069 {
1070         struct inet_sock *inet = inet_sk(sk);
1071         struct udp_sock *up = udp_sk(sk);
1072         int ret;
1073 
1074         if (!up->pending) {
1075                 struct msghdr msg = {   .msg_flags = flags|MSG_MORE };
1076 
1077                 /* Call udp_sendmsg to specify destination address which
1078                  * sendpage interface can't pass.
1079                  * This will succeed only when the socket is connected.
1080                  */
1081                 ret = udp_sendmsg(NULL, sk, &msg, 0);
1082                 if (ret < 0)
1083                         return ret;
1084         }
1085 
1086         lock_sock(sk);
1087 
1088         if (unlikely(!up->pending)) {
1089                 release_sock(sk);
1090 
1091                 LIMIT_NETDEBUG(KERN_DEBUG pr_fmt("udp cork app bug 3\n"));
1092                 return -EINVAL;
1093         }
1094 
1095         ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
1096                              page, offset, size, flags);
1097         if (ret == -EOPNOTSUPP) {
1098                 release_sock(sk);
1099                 return sock_no_sendpage(sk->sk_socket, page, offset,
1100                                         size, flags);
1101         }
1102         if (ret < 0) {
1103                 udp_flush_pending_frames(sk);
1104                 goto out;
1105         }
1106 
1107         up->len += size;
1108         if (!(up->corkflag || (flags&MSG_MORE)))
1109                 ret = udp_push_pending_frames(sk);
1110         if (!ret)
1111                 ret = size;
1112 out:
1113         release_sock(sk);
1114         return ret;
1115 }
1116 
1117 
1118 /**
1119  *      first_packet_length     - return length of first packet in receive queue
1120  *      @sk: socket
1121  *
1122  *      Drops all bad checksum frames, until a valid one is found.
1123  *      Returns the length of found skb, or 0 if none is found.
1124  */
1125 static unsigned int first_packet_length(struct sock *sk)
1126 {
1127         struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue;
1128         struct sk_buff *skb;
1129         unsigned int res;
1130 
1131         __skb_queue_head_init(&list_kill);
1132 
1133         spin_lock_bh(&rcvq->lock);
1134         while ((skb = skb_peek(rcvq)) != NULL &&
1135                 udp_lib_checksum_complete(skb)) {
1136                 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS,
1137                                  IS_UDPLITE(sk));
1138                 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1139                                  IS_UDPLITE(sk));
1140                 atomic_inc(&sk->sk_drops);
1141                 __skb_unlink(skb, rcvq);
1142                 __skb_queue_tail(&list_kill, skb);
1143         }
1144         res = skb ? skb->len : 0;
1145         spin_unlock_bh(&rcvq->lock);
1146 
1147         if (!skb_queue_empty(&list_kill)) {
1148                 bool slow = lock_sock_fast(sk);
1149 
1150                 __skb_queue_purge(&list_kill);
1151                 sk_mem_reclaim_partial(sk);
1152                 unlock_sock_fast(sk, slow);
1153         }
1154         return res;
1155 }
1156 
1157 /*
1158  *      IOCTL requests applicable to the UDP protocol
1159  */
1160 
1161 int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1162 {
1163         switch (cmd) {
1164         case SIOCOUTQ:
1165         {
1166                 int amount = sk_wmem_alloc_get(sk);
1167 
1168                 return put_user(amount, (int __user *)arg);
1169         }
1170 
1171         case SIOCINQ:
1172         {
1173                 unsigned int amount = first_packet_length(sk);
1174 
1175                 if (amount)
1176                         /*
1177                          * We will only return the amount
1178                          * of this packet since that is all
1179                          * that will be read.
1180                          */
1181                         amount -= sizeof(struct udphdr);
1182 
1183                 return put_user(amount, (int __user *)arg);
1184         }
1185 
1186         default:
1187                 return -ENOIOCTLCMD;
1188         }
1189 
1190         return 0;
1191 }
1192 EXPORT_SYMBOL(udp_ioctl);
1193 
1194 /*
1195  *      This should be easy, if there is something there we
1196  *      return it, otherwise we block.
1197  */
1198 
1199 int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
1200                 size_t len, int noblock, int flags, int *addr_len)
1201 {
1202         struct inet_sock *inet = inet_sk(sk);
1203         struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
1204         struct sk_buff *skb;
1205         unsigned int ulen, copied;
1206         int peeked, off = 0;
1207         int err;
1208         int is_udplite = IS_UDPLITE(sk);
1209         bool slow;
1210 
1211         /*
1212          *      Check any passed addresses
1213          */
1214         if (addr_len)
1215                 *addr_len = sizeof(*sin);
1216 
1217         if (flags & MSG_ERRQUEUE)
1218                 return ip_recv_error(sk, msg, len);
1219 
1220 try_again:
1221         skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
1222                                   &peeked, &off, &err);
1223         if (!skb)
1224                 goto out;
1225         if (ccs_socket_post_recvmsg_permission(sk, skb, flags)) {
1226                 err = -EAGAIN; /* Hope less harmful than -EPERM. */
1227                 goto out;
1228         }
1229 
1230         ulen = skb->len - sizeof(struct udphdr);
1231         copied = len;
1232         if (copied > ulen)
1233                 copied = ulen;
1234         else if (copied < ulen)
1235                 msg->msg_flags |= MSG_TRUNC;
1236 
1237         /*
1238          * If checksum is needed at all, try to do it while copying the
1239          * data.  If the data is truncated, or if we only want a partial
1240          * coverage checksum (UDP-Lite), do it before the copy.
1241          */
1242 
1243         if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
1244                 if (udp_lib_checksum_complete(skb))
1245                         goto csum_copy_err;
1246         }
1247 
1248         if (skb_csum_unnecessary(skb))
1249                 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
1250                                               msg->msg_iov, copied);
1251         else {
1252                 err = skb_copy_and_csum_datagram_iovec(skb,
1253                                                        sizeof(struct udphdr),
1254                                                        msg->msg_iov);
1255 
1256                 if (err == -EINVAL)
1257                         goto csum_copy_err;
1258         }
1259 
1260         if (unlikely(err)) {
1261                 trace_kfree_skb(skb, udp_recvmsg);
1262                 if (!peeked) {
1263                         atomic_inc(&sk->sk_drops);
1264                         UDP_INC_STATS_USER(sock_net(sk),
1265                                            UDP_MIB_INERRORS, is_udplite);
1266                 }
1267                 goto out_free;
1268         }
1269 
1270         if (!peeked)
1271                 UDP_INC_STATS_USER(sock_net(sk),
1272                                 UDP_MIB_INDATAGRAMS, is_udplite);
1273 
1274         sock_recv_ts_and_drops(msg, sk, skb);
1275 
1276         /* Copy the address. */
1277         if (sin) {
1278                 sin->sin_family = AF_INET;
1279                 sin->sin_port = udp_hdr(skb)->source;
1280                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1281                 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
1282         }
1283         if (inet->cmsg_flags)
1284                 ip_cmsg_recv(msg, skb);
1285 
1286         err = copied;
1287         if (flags & MSG_TRUNC)
1288                 err = ulen;
1289 
1290 out_free:
1291         skb_free_datagram_locked(sk, skb);
1292 out:
1293         return err;
1294 
1295 csum_copy_err:
1296         slow = lock_sock_fast(sk);
1297         if (!skb_kill_datagram(sk, skb, flags)) {
1298                 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
1299                 UDP_INC_STATS_USER(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1300         }
1301         unlock_sock_fast(sk, slow);
1302 
1303         if (noblock)
1304                 return -EAGAIN;
1305 
1306         /* starting over for a new packet */
1307         msg->msg_flags &= ~MSG_TRUNC;
1308         goto try_again;
1309 }
1310 
1311 
1312 int udp_disconnect(struct sock *sk, int flags)
1313 {
1314         struct inet_sock *inet = inet_sk(sk);
1315         /*
1316          *      1003.1g - break association.
1317          */
1318 
1319         sk->sk_state = TCP_CLOSE;
1320         inet->inet_daddr = 0;
1321         inet->inet_dport = 0;
1322         sock_rps_reset_rxhash(sk);
1323         sk->sk_bound_dev_if = 0;
1324         if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
1325                 inet_reset_saddr(sk);
1326 
1327         if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
1328                 sk->sk_prot->unhash(sk);
1329                 inet->inet_sport = 0;
1330         }
1331         sk_dst_reset(sk);
1332         return 0;
1333 }
1334 EXPORT_SYMBOL(udp_disconnect);
1335 
1336 void udp_lib_unhash(struct sock *sk)
1337 {
1338         if (sk_hashed(sk)) {
1339                 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1340                 struct udp_hslot *hslot, *hslot2;
1341 
1342                 hslot  = udp_hashslot(udptable, sock_net(sk),
1343                                       udp_sk(sk)->udp_port_hash);
1344                 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1345 
1346                 spin_lock_bh(&hslot->lock);
1347                 if (sk_nulls_del_node_init_rcu(sk)) {
1348                         hslot->count--;
1349                         inet_sk(sk)->inet_num = 0;
1350                         sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
1351 
1352                         spin_lock(&hslot2->lock);
1353                         hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1354                         hslot2->count--;
1355                         spin_unlock(&hslot2->lock);
1356                 }
1357                 spin_unlock_bh(&hslot->lock);
1358         }
1359 }
1360 EXPORT_SYMBOL(udp_lib_unhash);
1361 
1362 /*
1363  * inet_rcv_saddr was changed, we must rehash secondary hash
1364  */
1365 void udp_lib_rehash(struct sock *sk, u16 newhash)
1366 {
1367         if (sk_hashed(sk)) {
1368                 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1369                 struct udp_hslot *hslot, *hslot2, *nhslot2;
1370 
1371                 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1372                 nhslot2 = udp_hashslot2(udptable, newhash);
1373                 udp_sk(sk)->udp_portaddr_hash = newhash;
1374                 if (hslot2 != nhslot2) {
1375                         hslot = udp_hashslot(udptable, sock_net(sk),
1376                                              udp_sk(sk)->udp_port_hash);
1377                         /* we must lock primary chain too */
1378                         spin_lock_bh(&hslot->lock);
1379 
1380                         spin_lock(&hslot2->lock);
1381                         hlist_nulls_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1382                         hslot2->count--;
1383                         spin_unlock(&hslot2->lock);
1384 
1385                         spin_lock(&nhslot2->lock);
1386                         hlist_nulls_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
1387                                                  &nhslot2->head);
1388                         nhslot2->count++;
1389                         spin_unlock(&nhslot2->lock);
1390 
1391                         spin_unlock_bh(&hslot->lock);
1392                 }
1393         }
1394 }
1395 EXPORT_SYMBOL(udp_lib_rehash);
1396 
1397 static void udp_v4_rehash(struct sock *sk)
1398 {
1399         u16 new_hash = udp4_portaddr_hash(sock_net(sk),
1400                                           inet_sk(sk)->inet_rcv_saddr,
1401                                           inet_sk(sk)->inet_num);
1402         udp_lib_rehash(sk, new_hash);
1403 }
1404 
1405 static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1406 {
1407         int rc;
1408 
1409         if (inet_sk(sk)->inet_daddr)
1410                 sock_rps_save_rxhash(sk, skb);
1411 
1412         rc = sock_queue_rcv_skb(sk, skb);
1413         if (rc < 0) {
1414                 int is_udplite = IS_UDPLITE(sk);
1415 
1416                 /* Note that an ENOMEM error is charged twice */
1417                 if (rc == -ENOMEM)
1418                         UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1419                                          is_udplite);
1420                 UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1421                 kfree_skb(skb);
1422                 trace_udp_fail_queue_rcv_skb(rc, sk);
1423                 return -1;
1424         }
1425 
1426         return 0;
1427 
1428 }
1429 
1430 static struct static_key udp_encap_needed __read_mostly;
1431 void udp_encap_enable(void)
1432 {
1433         if (!static_key_enabled(&udp_encap_needed))
1434                 static_key_slow_inc(&udp_encap_needed);
1435 }
1436 EXPORT_SYMBOL(udp_encap_enable);
1437 
1438 /* returns:
1439  *  -1: error
1440  *   0: success
1441  *  >0: "udp encap" protocol resubmission
1442  *
1443  * Note that in the success and error cases, the skb is assumed to
1444  * have either been requeued or freed.
1445  */
1446 int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1447 {
1448         struct udp_sock *up = udp_sk(sk);
1449         int rc;
1450         int is_udplite = IS_UDPLITE(sk);
1451 
1452         /*
1453          *      Charge it to the socket, dropping if the queue is full.
1454          */
1455         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1456                 goto drop;
1457         nf_reset(skb);
1458 
1459         if (static_key_false(&udp_encap_needed) && up->encap_type) {
1460                 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
1461 
1462                 /*
1463                  * This is an encapsulation socket so pass the skb to
1464                  * the socket's udp_encap_rcv() hook. Otherwise, just
1465                  * fall through and pass this up the UDP socket.
1466                  * up->encap_rcv() returns the following value:
1467                  * =0 if skb was successfully passed to the encap
1468                  *    handler or was discarded by it.
1469                  * >0 if skb should be passed on to UDP.
1470                  * <0 if skb should be resubmitted as proto -N
1471                  */
1472 
1473                 /* if we're overly short, let UDP handle it */
1474                 encap_rcv = ACCESS_ONCE(up->encap_rcv);
1475                 if (skb->len > sizeof(struct udphdr) && encap_rcv != NULL) {
1476                         int ret;
1477 
1478                         ret = encap_rcv(sk, skb);
1479                         if (ret <= 0) {
1480                                 UDP_INC_STATS_BH(sock_net(sk),
1481                                                  UDP_MIB_INDATAGRAMS,
1482                                                  is_udplite);
1483                                 return -ret;
1484                         }
1485                 }
1486 
1487                 /* FALLTHROUGH -- it's a UDP Packet */
1488         }
1489 
1490         /*
1491          *      UDP-Lite specific tests, ignored on UDP sockets
1492          */
1493         if ((is_udplite & UDPLITE_RECV_CC)  &&  UDP_SKB_CB(skb)->partial_cov) {
1494 
1495                 /*
1496                  * MIB statistics other than incrementing the error count are
1497                  * disabled for the following two types of errors: these depend
1498                  * on the application settings, not on the functioning of the
1499                  * protocol stack as such.
1500                  *
1501                  * RFC 3828 here recommends (sec 3.3): "There should also be a
1502                  * way ... to ... at least let the receiving application block
1503                  * delivery of packets with coverage values less than a value
1504                  * provided by the application."
1505                  */
1506                 if (up->pcrlen == 0) {          /* full coverage was set  */
1507                         LIMIT_NETDEBUG(KERN_WARNING "UDPLite: partial coverage %d while full coverage %d requested\n",
1508                                        UDP_SKB_CB(skb)->cscov, skb->len);
1509                         goto drop;
1510                 }
1511                 /* The next case involves violating the min. coverage requested
1512                  * by the receiver. This is subtle: if receiver wants x and x is
1513                  * greater than the buffersize/MTU then receiver will complain
1514                  * that it wants x while sender emits packets of smaller size y.
1515                  * Therefore the above ...()->partial_cov statement is essential.
1516                  */
1517                 if (UDP_SKB_CB(skb)->cscov  <  up->pcrlen) {
1518                         LIMIT_NETDEBUG(KERN_WARNING "UDPLite: coverage %d too small, need min %d\n",
1519                                        UDP_SKB_CB(skb)->cscov, up->pcrlen);
1520                         goto drop;
1521                 }
1522         }
1523 
1524         if (rcu_access_pointer(sk->sk_filter) &&
1525             udp_lib_checksum_complete(skb))
1526                 goto csum_error;
1527 
1528 
1529         if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf))
1530                 goto drop;
1531 
1532         rc = 0;
1533 
1534         ipv4_pktinfo_prepare(skb);
1535         bh_lock_sock(sk);
1536         if (!sock_owned_by_user(sk))
1537                 rc = __udp_queue_rcv_skb(sk, skb);
1538         else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
1539                 bh_unlock_sock(sk);
1540                 goto drop;
1541         }
1542         bh_unlock_sock(sk);
1543 
1544         return rc;
1545 
1546 csum_error:
1547         UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
1548 drop:
1549         UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1550         atomic_inc(&sk->sk_drops);
1551         kfree_skb(skb);
1552         return -1;
1553 }
1554 
1555 
1556 static void flush_stack(struct sock **stack, unsigned int count,
1557                         struct sk_buff *skb, unsigned int final)
1558 {
1559         unsigned int i;
1560         struct sk_buff *skb1 = NULL;
1561         struct sock *sk;
1562 
1563         for (i = 0; i < count; i++) {
1564                 sk = stack[i];
1565                 if (likely(skb1 == NULL))
1566                         skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
1567 
1568                 if (!skb1) {
1569                         atomic_inc(&sk->sk_drops);
1570                         UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1571                                          IS_UDPLITE(sk));
1572                         UDP_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS,
1573                                          IS_UDPLITE(sk));
1574                 }
1575 
1576                 if (skb1 && udp_queue_rcv_skb(sk, skb1) <= 0)
1577                         skb1 = NULL;
1578         }
1579         if (unlikely(skb1))
1580                 kfree_skb(skb1);
1581 }
1582 
1583 /*
1584  *      Multicasts and broadcasts go to each listener.
1585  *
1586  *      Note: called only from the BH handler context.
1587  */
1588 static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
1589                                     struct udphdr  *uh,
1590                                     __be32 saddr, __be32 daddr,
1591                                     struct udp_table *udptable)
1592 {
1593         struct sock *sk, *stack[256 / sizeof(struct sock *)];
1594         struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
1595         int dif;
1596         unsigned int i, count = 0;
1597 
1598         spin_lock(&hslot->lock);
1599         sk = sk_nulls_head(&hslot->head);
1600         dif = skb->dev->ifindex;
1601         sk = udp_v4_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
1602         while (sk) {
1603                 stack[count++] = sk;
1604                 sk = udp_v4_mcast_next(net, sk_nulls_next(sk), uh->dest,
1605                                        daddr, uh->source, saddr, dif);
1606                 if (unlikely(count == ARRAY_SIZE(stack))) {
1607                         if (!sk)
1608                                 break;
1609                         flush_stack(stack, count, skb, ~0);
1610                         count = 0;
1611                 }
1612         }
1613         /*
1614          * before releasing chain lock, we must take a reference on sockets
1615          */
1616         for (i = 0; i < count; i++)
1617                 sock_hold(stack[i]);
1618 
1619         spin_unlock(&hslot->lock);
1620 
1621         /*
1622          * do the slow work with no lock held
1623          */
1624         if (count) {
1625                 flush_stack(stack, count, skb, count - 1);
1626 
1627                 for (i = 0; i < count; i++)
1628                         sock_put(stack[i]);
1629         } else {
1630                 kfree_skb(skb);
1631         }
1632         return 0;
1633 }
1634 
1635 /* Initialize UDP checksum. If exited with zero value (success),
1636  * CHECKSUM_UNNECESSARY means, that no more checks are required.
1637  * Otherwise, csum completion requires chacksumming packet body,
1638  * including udp header and folding it to skb->csum.
1639  */
1640 static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1641                                  int proto)
1642 {
1643         const struct iphdr *iph;
1644         int err;
1645 
1646         UDP_SKB_CB(skb)->partial_cov = 0;
1647         UDP_SKB_CB(skb)->cscov = skb->len;
1648 
1649         if (proto == IPPROTO_UDPLITE) {
1650                 err = udplite_checksum_init(skb, uh);
1651                 if (err)
1652                         return err;
1653         }
1654 
1655         iph = ip_hdr(skb);
1656         if (uh->check == 0) {
1657                 skb->ip_summed = CHECKSUM_UNNECESSARY;
1658         } else if (skb->ip_summed == CHECKSUM_COMPLETE) {
1659                 if (!csum_tcpudp_magic(iph->saddr, iph->daddr, skb->len,
1660                                       proto, skb->csum))
1661                         skb->ip_summed = CHECKSUM_UNNECESSARY;
1662         }
1663         if (!skb_csum_unnecessary(skb))
1664                 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
1665                                                skb->len, proto, 0);
1666         /* Probably, we should checksum udp header (it should be in cache
1667          * in any case) and data in tiny packets (< rx copybreak).
1668          */
1669 
1670         return 0;
1671 }
1672 
1673 /*
1674  *      All we need to do is get the socket, and then do a checksum.
1675  */
1676 
1677 int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
1678                    int proto)
1679 {
1680         struct sock *sk;
1681         struct udphdr *uh;
1682         unsigned short ulen;
1683         struct rtable *rt = skb_rtable(skb);
1684         __be32 saddr, daddr;
1685         struct net *net = dev_net(skb->dev);
1686 
1687         /*
1688          *  Validate the packet.
1689          */
1690         if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1691                 goto drop;              /* No space for header. */
1692 
1693         uh   = udp_hdr(skb);
1694         ulen = ntohs(uh->len);
1695         saddr = ip_hdr(skb)->saddr;
1696         daddr = ip_hdr(skb)->daddr;
1697 
1698         if (ulen > skb->len)
1699                 goto short_packet;
1700 
1701         if (proto == IPPROTO_UDP) {
1702                 /* UDP validates ulen. */
1703                 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1704                         goto short_packet;
1705                 uh = udp_hdr(skb);
1706         }
1707 
1708         if (udp4_csum_init(skb, uh, proto))
1709                 goto csum_error;
1710 
1711         if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1712                 return __udp4_lib_mcast_deliver(net, skb, uh,
1713                                 saddr, daddr, udptable);
1714 
1715         sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
1716 
1717         if (sk != NULL) {
1718                 int ret;
1719 
1720                 sk_mark_napi_id(sk, skb);
1721                 ret = udp_queue_rcv_skb(sk, skb);
1722                 sock_put(sk);
1723 
1724                 /* a return value > 0 means to resubmit the input, but
1725                  * it wants the return to be -protocol, or 0
1726                  */
1727                 if (ret > 0)
1728                         return -ret;
1729                 return 0;
1730         }
1731 
1732         if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1733                 goto drop;
1734         nf_reset(skb);
1735 
1736         /* No socket. Drop packet silently, if checksum is wrong */
1737         if (udp_lib_checksum_complete(skb))
1738                 goto csum_error;
1739 
1740         UDP_INC_STATS_BH(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
1741         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1742 
1743         /*
1744          * Hmm.  We got an UDP packet to a port to which we
1745          * don't wanna listen.  Ignore it.
1746          */
1747         kfree_skb(skb);
1748         return 0;
1749 
1750 short_packet:
1751         LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
1752                        proto == IPPROTO_UDPLITE ? "Lite" : "",
1753                        &saddr, ntohs(uh->source),
1754                        ulen, skb->len,
1755                        &daddr, ntohs(uh->dest));
1756         goto drop;
1757 
1758 csum_error:
1759         /*
1760          * RFC1122: OK.  Discards the bad packet silently (as far as
1761          * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1762          */
1763         LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
1764                        proto == IPPROTO_UDPLITE ? "Lite" : "",
1765                        &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest),
1766                        ulen);
1767         UDP_INC_STATS_BH(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
1768 drop:
1769         UDP_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1770         kfree_skb(skb);
1771         return 0;
1772 }
1773 
1774 int udp_rcv(struct sk_buff *skb)
1775 {
1776         return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
1777 }
1778 
1779 void udp_destroy_sock(struct sock *sk)
1780 {
1781         struct udp_sock *up = udp_sk(sk);
1782         bool slow = lock_sock_fast(sk);
1783         udp_flush_pending_frames(sk);
1784         unlock_sock_fast(sk, slow);
1785         if (static_key_false(&udp_encap_needed) && up->encap_type) {
1786                 void (*encap_destroy)(struct sock *sk);
1787                 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1788                 if (encap_destroy)
1789                         encap_destroy(sk);
1790         }
1791 }
1792 
1793 /*
1794  *      Socket option code for UDP
1795  */
1796 int udp_lib_setsockopt(struct sock *sk, int level, int optname,
1797                        char __user *optval, unsigned int optlen,
1798                        int (*push_pending_frames)(struct sock *))
1799 {
1800         struct udp_sock *up = udp_sk(sk);
1801         int val;
1802         int err = 0;
1803         int is_udplite = IS_UDPLITE(sk);
1804 
1805         if (optlen < sizeof(int))
1806                 return -EINVAL;
1807 
1808         if (get_user(val, (int __user *)optval))
1809                 return -EFAULT;
1810 
1811         switch (optname) {
1812         case UDP_CORK:
1813                 if (val != 0) {
1814                         up->corkflag = 1;
1815                 } else {
1816                         up->corkflag = 0;
1817                         lock_sock(sk);
1818                         (*push_pending_frames)(sk);
1819                         release_sock(sk);
1820                 }
1821                 break;
1822 
1823         case UDP_ENCAP:
1824                 switch (val) {
1825                 case 0:
1826                 case UDP_ENCAP_ESPINUDP:
1827                 case UDP_ENCAP_ESPINUDP_NON_IKE:
1828                         up->encap_rcv = xfrm4_udp_encap_rcv;
1829                         /* FALLTHROUGH */
1830                 case UDP_ENCAP_L2TPINUDP:
1831                         up->encap_type = val;
1832                         udp_encap_enable();
1833                         break;
1834                 default:
1835                         err = -ENOPROTOOPT;
1836                         break;
1837                 }
1838                 break;
1839 
1840         /*
1841          *      UDP-Lite's partial checksum coverage (RFC 3828).
1842          */
1843         /* The sender sets actual checksum coverage length via this option.
1844          * The case coverage > packet length is handled by send module. */
1845         case UDPLITE_SEND_CSCOV:
1846                 if (!is_udplite)         /* Disable the option on UDP sockets */
1847                         return -ENOPROTOOPT;
1848                 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
1849                         val = 8;
1850                 else if (val > USHRT_MAX)
1851                         val = USHRT_MAX;
1852                 up->pcslen = val;
1853                 up->pcflag |= UDPLITE_SEND_CC;
1854                 break;
1855 
1856         /* The receiver specifies a minimum checksum coverage value. To make
1857          * sense, this should be set to at least 8 (as done below). If zero is
1858          * used, this again means full checksum coverage.                     */
1859         case UDPLITE_RECV_CSCOV:
1860                 if (!is_udplite)         /* Disable the option on UDP sockets */
1861                         return -ENOPROTOOPT;
1862                 if (val != 0 && val < 8) /* Avoid silly minimal values.       */
1863                         val = 8;
1864                 else if (val > USHRT_MAX)
1865                         val = USHRT_MAX;
1866                 up->pcrlen = val;
1867                 up->pcflag |= UDPLITE_RECV_CC;
1868                 break;
1869 
1870         default:
1871                 err = -ENOPROTOOPT;
1872                 break;
1873         }
1874 
1875         return err;
1876 }
1877 EXPORT_SYMBOL(udp_lib_setsockopt);
1878 
1879 int udp_setsockopt(struct sock *sk, int level, int optname,
1880                    char __user *optval, unsigned int optlen)
1881 {
1882         if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1883                 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1884                                           udp_push_pending_frames);
1885         return ip_setsockopt(sk, level, optname, optval, optlen);
1886 }
1887 
1888 #ifdef CONFIG_COMPAT
1889 int compat_udp_setsockopt(struct sock *sk, int level, int optname,
1890                           char __user *optval, unsigned int optlen)
1891 {
1892         if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1893                 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1894                                           udp_push_pending_frames);
1895         return compat_ip_setsockopt(sk, level, optname, optval, optlen);
1896 }
1897 #endif
1898 
1899 int udp_lib_getsockopt(struct sock *sk, int level, int optname,
1900                        char __user *optval, int __user *optlen)
1901 {
1902         struct udp_sock *up = udp_sk(sk);
1903         int val, len;
1904 
1905         if (get_user(len, optlen))
1906                 return -EFAULT;
1907 
1908         len = min_t(unsigned int, len, sizeof(int));
1909 
1910         if (len < 0)
1911                 return -EINVAL;
1912 
1913         switch (optname) {
1914         case UDP_CORK:
1915                 val = up->corkflag;
1916                 break;
1917 
1918         case UDP_ENCAP:
1919                 val = up->encap_type;
1920                 break;
1921 
1922         /* The following two cannot be changed on UDP sockets, the return is
1923          * always 0 (which corresponds to the full checksum coverage of UDP). */
1924         case UDPLITE_SEND_CSCOV:
1925                 val = up->pcslen;
1926                 break;
1927 
1928         case UDPLITE_RECV_CSCOV:
1929                 val = up->pcrlen;
1930                 break;
1931 
1932         default:
1933                 return -ENOPROTOOPT;
1934         }
1935 
1936         if (put_user(len, optlen))
1937                 return -EFAULT;
1938         if (copy_to_user(optval, &val, len))
1939                 return -EFAULT;
1940         return 0;
1941 }
1942 EXPORT_SYMBOL(udp_lib_getsockopt);
1943 
1944 int udp_getsockopt(struct sock *sk, int level, int optname,
1945                    char __user *optval, int __user *optlen)
1946 {
1947         if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1948                 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1949         return ip_getsockopt(sk, level, optname, optval, optlen);
1950 }
1951 
1952 #ifdef CONFIG_COMPAT
1953 int compat_udp_getsockopt(struct sock *sk, int level, int optname,
1954                                  char __user *optval, int __user *optlen)
1955 {
1956         if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1957                 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1958         return compat_ip_getsockopt(sk, level, optname, optval, optlen);
1959 }
1960 #endif
1961 /**
1962  *      udp_poll - wait for a UDP event.
1963  *      @file - file struct
1964  *      @sock - socket
1965  *      @wait - poll table
1966  *
1967  *      This is same as datagram poll, except for the special case of
1968  *      blocking sockets. If application is using a blocking fd
1969  *      and a packet with checksum error is in the queue;
1970  *      then it could get return from select indicating data available
1971  *      but then block when reading it. Add special case code
1972  *      to work around these arguably broken applications.
1973  */
1974 unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
1975 {
1976         unsigned int mask = datagram_poll(file, sock, wait);
1977         struct sock *sk = sock->sk;
1978 
1979         sock_rps_record_flow(sk);
1980 
1981         /* Check for false positives due to checksum errors */
1982         if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) &&
1983             !(sk->sk_shutdown & RCV_SHUTDOWN) && !first_packet_length(sk))
1984                 mask &= ~(POLLIN | POLLRDNORM);
1985 
1986         return mask;
1987 
1988 }
1989 EXPORT_SYMBOL(udp_poll);
1990 
1991 struct proto udp_prot = {
1992         .name              = "UDP",
1993         .owner             = THIS_MODULE,
1994         .close             = udp_lib_close,
1995         .connect           = ip4_datagram_connect,
1996         .disconnect        = udp_disconnect,
1997         .ioctl             = udp_ioctl,
1998         .destroy           = udp_destroy_sock,
1999         .setsockopt        = udp_setsockopt,
2000         .getsockopt        = udp_getsockopt,
2001         .sendmsg           = udp_sendmsg,
2002         .recvmsg           = udp_recvmsg,
2003         .sendpage          = udp_sendpage,
2004         .backlog_rcv       = __udp_queue_rcv_skb,
2005         .release_cb        = ip4_datagram_release_cb,
2006         .hash              = udp_lib_hash,
2007         .unhash            = udp_lib_unhash,
2008         .rehash            = udp_v4_rehash,
2009         .get_port          = udp_v4_get_port,
2010         .memory_allocated  = &udp_memory_allocated,
2011         .sysctl_mem        = sysctl_udp_mem,
2012         .sysctl_wmem       = &sysctl_udp_wmem_min,
2013         .sysctl_rmem       = &sysctl_udp_rmem_min,
2014         .obj_size          = sizeof(struct udp_sock),
2015         .slab_flags        = SLAB_DESTROY_BY_RCU,
2016         .h.udp_table       = &udp_table,
2017 #ifdef CONFIG_COMPAT
2018         .compat_setsockopt = compat_udp_setsockopt,
2019         .compat_getsockopt = compat_udp_getsockopt,
2020 #endif
2021         .clear_sk          = sk_prot_clear_portaddr_nulls,
2022 };
2023 EXPORT_SYMBOL(udp_prot);
2024 
2025 /* ------------------------------------------------------------------------ */
2026 #ifdef CONFIG_PROC_FS
2027 
2028 static struct sock *udp_get_first(struct seq_file *seq, int start)
2029 {
2030         struct sock *sk;
2031         struct udp_iter_state *state = seq->private;
2032         struct net *net = seq_file_net(seq);
2033 
2034         for (state->bucket = start; state->bucket <= state->udp_table->mask;
2035              ++state->bucket) {
2036                 struct hlist_nulls_node *node;
2037                 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
2038 
2039                 if (hlist_nulls_empty(&hslot->head))
2040                         continue;
2041 
2042                 spin_lock_bh(&hslot->lock);
2043                 sk_nulls_for_each(sk, node, &hslot->head) {
2044                         if (!net_eq(sock_net(sk), net))
2045                                 continue;
2046                         if (sk->sk_family == state->family)
2047                                 goto found;
2048                 }
2049                 spin_unlock_bh(&hslot->lock);
2050         }
2051         sk = NULL;
2052 found:
2053         return sk;
2054 }
2055 
2056 static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
2057 {
2058         struct udp_iter_state *state = seq->private;
2059         struct net *net = seq_file_net(seq);
2060 
2061         do {
2062                 sk = sk_nulls_next(sk);
2063         } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
2064 
2065         if (!sk) {
2066                 if (state->bucket <= state->udp_table->mask)
2067                         spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
2068                 return udp_get_first(seq, state->bucket + 1);
2069         }
2070         return sk;
2071 }
2072 
2073 static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
2074 {
2075         struct sock *sk = udp_get_first(seq, 0);
2076 
2077         if (sk)
2078                 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
2079                         --pos;
2080         return pos ? NULL : sk;
2081 }
2082 
2083 static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
2084 {
2085         struct udp_iter_state *state = seq->private;
2086         state->bucket = MAX_UDP_PORTS;
2087 
2088         return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
2089 }
2090 
2091 static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2092 {
2093         struct sock *sk;
2094 
2095         if (v == SEQ_START_TOKEN)
2096                 sk = udp_get_idx(seq, 0);
2097         else
2098                 sk = udp_get_next(seq, v);
2099 
2100         ++*pos;
2101         return sk;
2102 }
2103 
2104 static void udp_seq_stop(struct seq_file *seq, void *v)
2105 {
2106         struct udp_iter_state *state = seq->private;
2107 
2108         if (state->bucket <= state->udp_table->mask)
2109                 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
2110 }
2111 
2112 int udp_seq_open(struct inode *inode, struct file *file)
2113 {
2114         struct udp_seq_afinfo *afinfo = PDE_DATA(inode);
2115         struct udp_iter_state *s;
2116         int err;
2117 
2118         err = seq_open_net(inode, file, &afinfo->seq_ops,
2119                            sizeof(struct udp_iter_state));
2120         if (err < 0)
2121                 return err;
2122 
2123         s = ((struct seq_file *)file->private_data)->private;
2124         s->family               = afinfo->family;
2125         s->udp_table            = afinfo->udp_table;
2126         return err;
2127 }
2128 EXPORT_SYMBOL(udp_seq_open);
2129 
2130 /* ------------------------------------------------------------------------ */
2131 int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
2132 {
2133         struct proc_dir_entry *p;
2134         int rc = 0;
2135 
2136         afinfo->seq_ops.start           = udp_seq_start;
2137         afinfo->seq_ops.next            = udp_seq_next;
2138         afinfo->seq_ops.stop            = udp_seq_stop;
2139 
2140         p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
2141                              afinfo->seq_fops, afinfo);
2142         if (!p)
2143                 rc = -ENOMEM;
2144         return rc;
2145 }
2146 EXPORT_SYMBOL(udp_proc_register);
2147 
2148 void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
2149 {
2150         remove_proc_entry(afinfo->name, net->proc_net);
2151 }
2152 EXPORT_SYMBOL(udp_proc_unregister);
2153 
2154 /* ------------------------------------------------------------------------ */
2155 static void udp4_format_sock(struct sock *sp, struct seq_file *f,
2156                 int bucket, int *len)
2157 {
2158         struct inet_sock *inet = inet_sk(sp);
2159         __be32 dest = inet->inet_daddr;
2160         __be32 src  = inet->inet_rcv_saddr;
2161         __u16 destp       = ntohs(inet->inet_dport);
2162         __u16 srcp        = ntohs(inet->inet_sport);
2163 
2164         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
2165                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n",
2166                 bucket, src, srcp, dest, destp, sp->sk_state,
2167                 sk_wmem_alloc_get(sp),
2168                 sk_rmem_alloc_get(sp),
2169                 0, 0L, 0,
2170                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
2171                 0, sock_i_ino(sp),
2172                 atomic_read(&sp->sk_refcnt), sp,
2173                 atomic_read(&sp->sk_drops), len);
2174 }
2175 
2176 int udp4_seq_show(struct seq_file *seq, void *v)
2177 {
2178         if (v == SEQ_START_TOKEN)
2179                 seq_printf(seq, "%-127s\n",
2180                            "  sl  local_address rem_address   st tx_queue "
2181                            "rx_queue tr tm->when retrnsmt   uid  timeout "
2182                            "inode ref pointer drops");
2183         else {
2184                 struct udp_iter_state *state = seq->private;
2185                 int len;
2186 
2187                 udp4_format_sock(v, seq, state->bucket, &len);
2188                 seq_printf(seq, "%*s\n", 127 - len, "");
2189         }
2190         return 0;
2191 }
2192 
2193 static const struct file_operations udp_afinfo_seq_fops = {
2194         .owner    = THIS_MODULE,
2195         .open     = udp_seq_open,
2196         .read     = seq_read,
2197         .llseek   = seq_lseek,
2198         .release  = seq_release_net
2199 };
2200 
2201 /* ------------------------------------------------------------------------ */
2202 static struct udp_seq_afinfo udp4_seq_afinfo = {
2203         .name           = "udp",
2204         .family         = AF_INET,
2205         .udp_table      = &udp_table,
2206         .seq_fops       = &udp_afinfo_seq_fops,
2207         .seq_ops        = {
2208                 .show           = udp4_seq_show,
2209         },
2210 };
2211 
2212 static int __net_init udp4_proc_init_net(struct net *net)
2213 {
2214         return udp_proc_register(net, &udp4_seq_afinfo);
2215 }
2216 
2217 static void __net_exit udp4_proc_exit_net(struct net *net)
2218 {
2219         udp_proc_unregister(net, &udp4_seq_afinfo);
2220 }
2221 
2222 static struct pernet_operations udp4_net_ops = {
2223         .init = udp4_proc_init_net,
2224         .exit = udp4_proc_exit_net,
2225 };
2226 
2227 int __init udp4_proc_init(void)
2228 {
2229         return register_pernet_subsys(&udp4_net_ops);
2230 }
2231 
2232 void udp4_proc_exit(void)
2233 {
2234         unregister_pernet_subsys(&udp4_net_ops);
2235 }
2236 #endif /* CONFIG_PROC_FS */
2237 
2238 static __initdata unsigned long uhash_entries;
2239 static int __init set_uhash_entries(char *str)
2240 {
2241         ssize_t ret;
2242 
2243         if (!str)
2244                 return 0;
2245 
2246         ret = kstrtoul(str, 0, &uhash_entries);
2247         if (ret)
2248                 return 0;
2249 
2250         if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN)
2251                 uhash_entries = UDP_HTABLE_SIZE_MIN;
2252         return 1;
2253 }
2254 __setup("uhash_entries=", set_uhash_entries);
2255 
2256 void __init udp_table_init(struct udp_table *table, const char *name)
2257 {
2258         unsigned int i;
2259 
2260         table->hash = alloc_large_system_hash(name,
2261                                               2 * sizeof(struct udp_hslot),
2262                                               uhash_entries,
2263                                               21, /* one slot per 2 MB */
2264                                               0,
2265                                               &table->log,
2266                                               &table->mask,
2267                                               UDP_HTABLE_SIZE_MIN,
2268                                               64 * 1024);
2269 
2270         table->hash2 = table->hash + (table->mask + 1);
2271         for (i = 0; i <= table->mask; i++) {
2272                 INIT_HLIST_NULLS_HEAD(&table->hash[i].head, i);
2273                 table->hash[i].count = 0;
2274                 spin_lock_init(&table->hash[i].lock);
2275         }
2276         for (i = 0; i <= table->mask; i++) {
2277                 INIT_HLIST_NULLS_HEAD(&table->hash2[i].head, i);
2278                 table->hash2[i].count = 0;
2279                 spin_lock_init(&table->hash2[i].lock);
2280         }
2281 }
2282 
2283 void __init udp_init(void)
2284 {
2285         unsigned long limit;
2286 
2287         udp_table_init(&udp_table, "UDP");
2288         limit = nr_free_buffer_pages() / 8;
2289         limit = max(limit, 128UL);
2290         sysctl_udp_mem[0] = limit / 4 * 3;
2291         sysctl_udp_mem[1] = limit;
2292         sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
2293 
2294         sysctl_udp_rmem_min = SK_MEM_QUANTUM;
2295         sysctl_udp_wmem_min = SK_MEM_QUANTUM;
2296 }
2297 
2298 struct sk_buff *skb_udp_tunnel_segment(struct sk_buff *skb,
2299                                        netdev_features_t features)
2300 {
2301         struct sk_buff *segs = ERR_PTR(-EINVAL);
2302         int mac_len = skb->mac_len;
2303         int tnl_hlen = skb_inner_mac_header(skb) - skb_transport_header(skb);
2304         __be16 protocol = skb->protocol;
2305         netdev_features_t enc_features;
2306         int outer_hlen;
2307 
2308         if (unlikely(!pskb_may_pull(skb, tnl_hlen)))
2309                 goto out;
2310 
2311         skb->encapsulation = 0;
2312         __skb_pull(skb, tnl_hlen);
2313         skb_reset_mac_header(skb);
2314         skb_set_network_header(skb, skb_inner_network_offset(skb));
2315         skb->mac_len = skb_inner_network_offset(skb);
2316         skb->protocol = htons(ETH_P_TEB);
2317 
2318         /* segment inner packet. */
2319         enc_features = skb->dev->hw_enc_features & netif_skb_features(skb);
2320         segs = skb_mac_gso_segment(skb, enc_features);
2321         if (!segs || IS_ERR(segs))
2322                 goto out;
2323 
2324         outer_hlen = skb_tnl_header_len(skb);
2325         skb = segs;
2326         do {
2327                 struct udphdr *uh;
2328                 int udp_offset = outer_hlen - tnl_hlen;
2329 
2330                 skb_reset_inner_headers(skb);
2331                 skb->encapsulation = 1;
2332 
2333                 skb->mac_len = mac_len;
2334 
2335                 skb_push(skb, outer_hlen);
2336                 skb_reset_mac_header(skb);
2337                 skb_set_network_header(skb, mac_len);
2338                 skb_set_transport_header(skb, udp_offset);
2339                 uh = udp_hdr(skb);
2340                 uh->len = htons(skb->len - udp_offset);
2341 
2342                 /* csum segment if tunnel sets skb with csum. */
2343                 if (unlikely(uh->check)) {
2344                         struct iphdr *iph = ip_hdr(skb);
2345 
2346                         uh->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
2347                                                        skb->len - udp_offset,
2348                                                        IPPROTO_UDP, 0);
2349                         uh->check = csum_fold(skb_checksum(skb, udp_offset,
2350                                                            skb->len - udp_offset, 0));
2351                         if (uh->check == 0)
2352                                 uh->check = CSUM_MANGLED_0;
2353 
2354                 }
2355                 skb->protocol = protocol;
2356         } while ((skb = skb->next));
2357 out:
2358         return segs;
2359 }
2360 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp