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

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

Version: ~ [ linux-5.11-rc3 ] ~ [ linux-5.10.7 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.89 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.167 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.215 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.251 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.251 ] ~ [ 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-3.16.85 ] ~ [ 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-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/inetdevice.h>
 94 #include <linux/in.h>
 95 #include <linux/errno.h>
 96 #include <linux/timer.h>
 97 #include <linux/mm.h>
 98 #include <linux/inet.h>
 99 #include <linux/netdevice.h>
100 #include <linux/slab.h>
101 #include <net/tcp_states.h>
102 #include <linux/skbuff.h>
103 #include <linux/proc_fs.h>
104 #include <linux/seq_file.h>
105 #include <net/net_namespace.h>
106 #include <net/icmp.h>
107 #include <net/inet_hashtables.h>
108 #include <net/route.h>
109 #include <net/checksum.h>
110 #include <net/xfrm.h>
111 #include <trace/events/udp.h>
112 #include <linux/static_key.h>
113 #include <trace/events/skb.h>
114 #include <net/busy_poll.h>
115 #include "udp_impl.h"
116 #include <net/sock_reuseport.h>
117 
118 struct udp_table udp_table __read_mostly;
119 EXPORT_SYMBOL(udp_table);
120 
121 long sysctl_udp_mem[3] __read_mostly;
122 EXPORT_SYMBOL(sysctl_udp_mem);
123 
124 int sysctl_udp_rmem_min __read_mostly;
125 EXPORT_SYMBOL(sysctl_udp_rmem_min);
126 
127 int sysctl_udp_wmem_min __read_mostly;
128 EXPORT_SYMBOL(sysctl_udp_wmem_min);
129 
130 atomic_long_t udp_memory_allocated;
131 EXPORT_SYMBOL(udp_memory_allocated);
132 
133 #define MAX_UDP_PORTS 65536
134 #define PORTS_PER_CHAIN (MAX_UDP_PORTS / UDP_HTABLE_SIZE_MIN)
135 
136 static int udp_lib_lport_inuse(struct net *net, __u16 num,
137                                const struct udp_hslot *hslot,
138                                unsigned long *bitmap,
139                                struct sock *sk,
140                                int (*saddr_comp)(const struct sock *sk1,
141                                                  const struct sock *sk2,
142                                                  bool match_wildcard),
143                                unsigned int log)
144 {
145         struct sock *sk2;
146         kuid_t uid = sock_i_uid(sk);
147 
148         sk_for_each(sk2, &hslot->head) {
149                 if (net_eq(sock_net(sk2), net) &&
150                     sk2 != sk &&
151                     (bitmap || udp_sk(sk2)->udp_port_hash == num) &&
152                     (!sk2->sk_reuse || !sk->sk_reuse) &&
153                     (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
154                      sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
155                     (!sk2->sk_reuseport || !sk->sk_reuseport ||
156                      rcu_access_pointer(sk->sk_reuseport_cb) ||
157                      !uid_eq(uid, sock_i_uid(sk2))) &&
158                     saddr_comp(sk, sk2, true)) {
159                         if (!bitmap)
160                                 return 1;
161                         __set_bit(udp_sk(sk2)->udp_port_hash >> log, bitmap);
162                 }
163         }
164         return 0;
165 }
166 
167 /*
168  * Note: we still hold spinlock of primary hash chain, so no other writer
169  * can insert/delete a socket with local_port == num
170  */
171 static int udp_lib_lport_inuse2(struct net *net, __u16 num,
172                                 struct udp_hslot *hslot2,
173                                 struct sock *sk,
174                                 int (*saddr_comp)(const struct sock *sk1,
175                                                   const struct sock *sk2,
176                                                   bool match_wildcard))
177 {
178         struct sock *sk2;
179         kuid_t uid = sock_i_uid(sk);
180         int res = 0;
181 
182         spin_lock(&hslot2->lock);
183         udp_portaddr_for_each_entry(sk2, &hslot2->head) {
184                 if (net_eq(sock_net(sk2), net) &&
185                     sk2 != sk &&
186                     (udp_sk(sk2)->udp_port_hash == num) &&
187                     (!sk2->sk_reuse || !sk->sk_reuse) &&
188                     (!sk2->sk_bound_dev_if || !sk->sk_bound_dev_if ||
189                      sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
190                     (!sk2->sk_reuseport || !sk->sk_reuseport ||
191                      rcu_access_pointer(sk->sk_reuseport_cb) ||
192                      !uid_eq(uid, sock_i_uid(sk2))) &&
193                     saddr_comp(sk, sk2, true)) {
194                         res = 1;
195                         break;
196                 }
197         }
198         spin_unlock(&hslot2->lock);
199         return res;
200 }
201 
202 static int udp_reuseport_add_sock(struct sock *sk, struct udp_hslot *hslot,
203                                   int (*saddr_same)(const struct sock *sk1,
204                                                     const struct sock *sk2,
205                                                     bool match_wildcard))
206 {
207         struct net *net = sock_net(sk);
208         kuid_t uid = sock_i_uid(sk);
209         struct sock *sk2;
210 
211         sk_for_each(sk2, &hslot->head) {
212                 if (net_eq(sock_net(sk2), net) &&
213                     sk2 != sk &&
214                     sk2->sk_family == sk->sk_family &&
215                     ipv6_only_sock(sk2) == ipv6_only_sock(sk) &&
216                     (udp_sk(sk2)->udp_port_hash == udp_sk(sk)->udp_port_hash) &&
217                     (sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
218                     sk2->sk_reuseport && uid_eq(uid, sock_i_uid(sk2)) &&
219                     (*saddr_same)(sk, sk2, false)) {
220                         return reuseport_add_sock(sk, sk2);
221                 }
222         }
223 
224         /* Initial allocation may have already happened via setsockopt */
225         if (!rcu_access_pointer(sk->sk_reuseport_cb))
226                 return reuseport_alloc(sk);
227         return 0;
228 }
229 
230 /**
231  *  udp_lib_get_port  -  UDP/-Lite port lookup for IPv4 and IPv6
232  *
233  *  @sk:          socket struct in question
234  *  @snum:        port number to look up
235  *  @saddr_comp:  AF-dependent comparison of bound local IP addresses
236  *  @hash2_nulladdr: AF-dependent hash value in secondary hash chains,
237  *                   with NULL address
238  */
239 int udp_lib_get_port(struct sock *sk, unsigned short snum,
240                      int (*saddr_comp)(const struct sock *sk1,
241                                        const struct sock *sk2,
242                                        bool match_wildcard),
243                      unsigned int hash2_nulladdr)
244 {
245         struct udp_hslot *hslot, *hslot2;
246         struct udp_table *udptable = sk->sk_prot->h.udp_table;
247         int    error = 1;
248         struct net *net = sock_net(sk);
249 
250         if (!snum) {
251                 int low, high, remaining;
252                 unsigned int rand;
253                 unsigned short first, last;
254                 DECLARE_BITMAP(bitmap, PORTS_PER_CHAIN);
255 
256                 inet_get_local_port_range(net, &low, &high);
257                 remaining = (high - low) + 1;
258 
259                 rand = prandom_u32();
260                 first = reciprocal_scale(rand, remaining) + low;
261                 /*
262                  * force rand to be an odd multiple of UDP_HTABLE_SIZE
263                  */
264                 rand = (rand | 1) * (udptable->mask + 1);
265                 last = first + udptable->mask + 1;
266                 do {
267                         hslot = udp_hashslot(udptable, net, first);
268                         bitmap_zero(bitmap, PORTS_PER_CHAIN);
269                         spin_lock_bh(&hslot->lock);
270                         udp_lib_lport_inuse(net, snum, hslot, bitmap, sk,
271                                             saddr_comp, udptable->log);
272 
273                         snum = first;
274                         /*
275                          * Iterate on all possible values of snum for this hash.
276                          * Using steps of an odd multiple of UDP_HTABLE_SIZE
277                          * give us randomization and full range coverage.
278                          */
279                         do {
280                                 if (low <= snum && snum <= high &&
281                                     !test_bit(snum >> udptable->log, bitmap) &&
282                                     !inet_is_local_reserved_port(net, snum))
283                                         goto found;
284                                 snum += rand;
285                         } while (snum != first);
286                         spin_unlock_bh(&hslot->lock);
287                 } while (++first != last);
288                 goto fail;
289         } else {
290                 hslot = udp_hashslot(udptable, net, snum);
291                 spin_lock_bh(&hslot->lock);
292                 if (hslot->count > 10) {
293                         int exist;
294                         unsigned int slot2 = udp_sk(sk)->udp_portaddr_hash ^ snum;
295 
296                         slot2          &= udptable->mask;
297                         hash2_nulladdr &= udptable->mask;
298 
299                         hslot2 = udp_hashslot2(udptable, slot2);
300                         if (hslot->count < hslot2->count)
301                                 goto scan_primary_hash;
302 
303                         exist = udp_lib_lport_inuse2(net, snum, hslot2,
304                                                      sk, saddr_comp);
305                         if (!exist && (hash2_nulladdr != slot2)) {
306                                 hslot2 = udp_hashslot2(udptable, hash2_nulladdr);
307                                 exist = udp_lib_lport_inuse2(net, snum, hslot2,
308                                                              sk, saddr_comp);
309                         }
310                         if (exist)
311                                 goto fail_unlock;
312                         else
313                                 goto found;
314                 }
315 scan_primary_hash:
316                 if (udp_lib_lport_inuse(net, snum, hslot, NULL, sk,
317                                         saddr_comp, 0))
318                         goto fail_unlock;
319         }
320 found:
321         inet_sk(sk)->inet_num = snum;
322         udp_sk(sk)->udp_port_hash = snum;
323         udp_sk(sk)->udp_portaddr_hash ^= snum;
324         if (sk_unhashed(sk)) {
325                 if (sk->sk_reuseport &&
326                     udp_reuseport_add_sock(sk, hslot, saddr_comp)) {
327                         inet_sk(sk)->inet_num = 0;
328                         udp_sk(sk)->udp_port_hash = 0;
329                         udp_sk(sk)->udp_portaddr_hash ^= snum;
330                         goto fail_unlock;
331                 }
332 
333                 sk_add_node_rcu(sk, &hslot->head);
334                 hslot->count++;
335                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
336 
337                 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
338                 spin_lock(&hslot2->lock);
339                 if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
340                     sk->sk_family == AF_INET6)
341                         hlist_add_tail_rcu(&udp_sk(sk)->udp_portaddr_node,
342                                            &hslot2->head);
343                 else
344                         hlist_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
345                                            &hslot2->head);
346                 hslot2->count++;
347                 spin_unlock(&hslot2->lock);
348         }
349         sock_set_flag(sk, SOCK_RCU_FREE);
350         error = 0;
351 fail_unlock:
352         spin_unlock_bh(&hslot->lock);
353 fail:
354         return error;
355 }
356 EXPORT_SYMBOL(udp_lib_get_port);
357 
358 /* match_wildcard == true:  0.0.0.0 equals to any IPv4 addresses
359  * match_wildcard == false: addresses must be exactly the same, i.e.
360  *                          0.0.0.0 only equals to 0.0.0.0
361  */
362 int ipv4_rcv_saddr_equal(const struct sock *sk1, const struct sock *sk2,
363                          bool match_wildcard)
364 {
365         struct inet_sock *inet1 = inet_sk(sk1), *inet2 = inet_sk(sk2);
366 
367         if (!ipv6_only_sock(sk2)) {
368                 if (inet1->inet_rcv_saddr == inet2->inet_rcv_saddr)
369                         return 1;
370                 if (!inet1->inet_rcv_saddr || !inet2->inet_rcv_saddr)
371                         return match_wildcard;
372         }
373         return 0;
374 }
375 
376 static u32 udp4_portaddr_hash(const struct net *net, __be32 saddr,
377                               unsigned int port)
378 {
379         return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
380 }
381 
382 int udp_v4_get_port(struct sock *sk, unsigned short snum)
383 {
384         unsigned int hash2_nulladdr =
385                 udp4_portaddr_hash(sock_net(sk), htonl(INADDR_ANY), snum);
386         unsigned int hash2_partial =
387                 udp4_portaddr_hash(sock_net(sk), inet_sk(sk)->inet_rcv_saddr, 0);
388 
389         /* precompute partial secondary hash */
390         udp_sk(sk)->udp_portaddr_hash = hash2_partial;
391         return udp_lib_get_port(sk, snum, ipv4_rcv_saddr_equal, hash2_nulladdr);
392 }
393 
394 static int compute_score(struct sock *sk, struct net *net,
395                          __be32 saddr, __be16 sport,
396                          __be32 daddr, unsigned short hnum, int dif)
397 {
398         int score;
399         struct inet_sock *inet;
400 
401         if (!net_eq(sock_net(sk), net) ||
402             udp_sk(sk)->udp_port_hash != hnum ||
403             ipv6_only_sock(sk))
404                 return -1;
405 
406         score = (sk->sk_family == PF_INET) ? 2 : 1;
407         inet = inet_sk(sk);
408 
409         if (inet->inet_rcv_saddr) {
410                 if (inet->inet_rcv_saddr != daddr)
411                         return -1;
412                 score += 4;
413         }
414 
415         if (inet->inet_daddr) {
416                 if (inet->inet_daddr != saddr)
417                         return -1;
418                 score += 4;
419         }
420 
421         if (inet->inet_dport) {
422                 if (inet->inet_dport != sport)
423                         return -1;
424                 score += 4;
425         }
426 
427         if (sk->sk_bound_dev_if) {
428                 if (sk->sk_bound_dev_if != dif)
429                         return -1;
430                 score += 4;
431         }
432         if (sk->sk_incoming_cpu == raw_smp_processor_id())
433                 score++;
434         return score;
435 }
436 
437 static u32 udp_ehashfn(const struct net *net, const __be32 laddr,
438                        const __u16 lport, const __be32 faddr,
439                        const __be16 fport)
440 {
441         static u32 udp_ehash_secret __read_mostly;
442 
443         net_get_random_once(&udp_ehash_secret, sizeof(udp_ehash_secret));
444 
445         return __inet_ehashfn(laddr, lport, faddr, fport,
446                               udp_ehash_secret + net_hash_mix(net));
447 }
448 
449 /* called with rcu_read_lock() */
450 static struct sock *udp4_lib_lookup2(struct net *net,
451                 __be32 saddr, __be16 sport,
452                 __be32 daddr, unsigned int hnum, int dif,
453                 struct udp_hslot *hslot2,
454                 struct sk_buff *skb)
455 {
456         struct sock *sk, *result;
457         int score, badness, matches = 0, reuseport = 0;
458         u32 hash = 0;
459 
460         result = NULL;
461         badness = 0;
462         udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
463                 score = compute_score(sk, net, saddr, sport,
464                                       daddr, hnum, dif);
465                 if (score > badness) {
466                         reuseport = sk->sk_reuseport;
467                         if (reuseport) {
468                                 hash = udp_ehashfn(net, daddr, hnum,
469                                                    saddr, sport);
470                                 result = reuseport_select_sock(sk, hash, skb,
471                                                         sizeof(struct udphdr));
472                                 if (result)
473                                         return result;
474                                 matches = 1;
475                         }
476                         badness = score;
477                         result = sk;
478                 } else if (score == badness && reuseport) {
479                         matches++;
480                         if (reciprocal_scale(hash, matches) == 0)
481                                 result = sk;
482                         hash = next_pseudo_random32(hash);
483                 }
484         }
485         return result;
486 }
487 
488 /* UDP is nearly always wildcards out the wazoo, it makes no sense to try
489  * harder than this. -DaveM
490  */
491 struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr,
492                 __be16 sport, __be32 daddr, __be16 dport,
493                 int dif, struct udp_table *udptable, struct sk_buff *skb)
494 {
495         struct sock *sk, *result;
496         unsigned short hnum = ntohs(dport);
497         unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
498         struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
499         int score, badness, matches = 0, reuseport = 0;
500         u32 hash = 0;
501 
502         if (hslot->count > 10) {
503                 hash2 = udp4_portaddr_hash(net, daddr, hnum);
504                 slot2 = hash2 & udptable->mask;
505                 hslot2 = &udptable->hash2[slot2];
506                 if (hslot->count < hslot2->count)
507                         goto begin;
508 
509                 result = udp4_lib_lookup2(net, saddr, sport,
510                                           daddr, hnum, dif,
511                                           hslot2, skb);
512                 if (!result) {
513                         unsigned int old_slot2 = slot2;
514                         hash2 = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum);
515                         slot2 = hash2 & udptable->mask;
516                         /* avoid searching the same slot again. */
517                         if (unlikely(slot2 == old_slot2))
518                                 return result;
519 
520                         hslot2 = &udptable->hash2[slot2];
521                         if (hslot->count < hslot2->count)
522                                 goto begin;
523 
524                         result = udp4_lib_lookup2(net, saddr, sport,
525                                                   daddr, hnum, dif,
526                                                   hslot2, skb);
527                 }
528                 return result;
529         }
530 begin:
531         result = NULL;
532         badness = 0;
533         sk_for_each_rcu(sk, &hslot->head) {
534                 score = compute_score(sk, net, saddr, sport,
535                                       daddr, hnum, dif);
536                 if (score > badness) {
537                         reuseport = sk->sk_reuseport;
538                         if (reuseport) {
539                                 hash = udp_ehashfn(net, daddr, hnum,
540                                                    saddr, sport);
541                                 result = reuseport_select_sock(sk, hash, skb,
542                                                         sizeof(struct udphdr));
543                                 if (result)
544                                         return result;
545                                 matches = 1;
546                         }
547                         result = sk;
548                         badness = score;
549                 } else if (score == badness && reuseport) {
550                         matches++;
551                         if (reciprocal_scale(hash, matches) == 0)
552                                 result = sk;
553                         hash = next_pseudo_random32(hash);
554                 }
555         }
556         return result;
557 }
558 EXPORT_SYMBOL_GPL(__udp4_lib_lookup);
559 
560 static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
561                                                  __be16 sport, __be16 dport,
562                                                  struct udp_table *udptable)
563 {
564         const struct iphdr *iph = ip_hdr(skb);
565 
566         return __udp4_lib_lookup(dev_net(skb->dev), iph->saddr, sport,
567                                  iph->daddr, dport, inet_iif(skb),
568                                  udptable, skb);
569 }
570 
571 struct sock *udp4_lib_lookup_skb(struct sk_buff *skb,
572                                  __be16 sport, __be16 dport)
573 {
574         return __udp4_lib_lookup_skb(skb, sport, dport, &udp_table);
575 }
576 EXPORT_SYMBOL_GPL(udp4_lib_lookup_skb);
577 
578 /* Must be called under rcu_read_lock().
579  * Does increment socket refcount.
580  */
581 #if IS_ENABLED(CONFIG_NETFILTER_XT_MATCH_SOCKET) || \
582     IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY)
583 struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
584                              __be32 daddr, __be16 dport, int dif)
585 {
586         struct sock *sk;
587 
588         sk = __udp4_lib_lookup(net, saddr, sport, daddr, dport,
589                                dif, &udp_table, NULL);
590         if (sk && !atomic_inc_not_zero(&sk->sk_refcnt))
591                 sk = NULL;
592         return sk;
593 }
594 EXPORT_SYMBOL_GPL(udp4_lib_lookup);
595 #endif
596 
597 static inline bool __udp_is_mcast_sock(struct net *net, struct sock *sk,
598                                        __be16 loc_port, __be32 loc_addr,
599                                        __be16 rmt_port, __be32 rmt_addr,
600                                        int dif, unsigned short hnum)
601 {
602         struct inet_sock *inet = inet_sk(sk);
603 
604         if (!net_eq(sock_net(sk), net) ||
605             udp_sk(sk)->udp_port_hash != hnum ||
606             (inet->inet_daddr && inet->inet_daddr != rmt_addr) ||
607             (inet->inet_dport != rmt_port && inet->inet_dport) ||
608             (inet->inet_rcv_saddr && inet->inet_rcv_saddr != loc_addr) ||
609             ipv6_only_sock(sk) ||
610             (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
611                 return false;
612         if (!ip_mc_sf_allow(sk, loc_addr, rmt_addr, dif))
613                 return false;
614         return true;
615 }
616 
617 /*
618  * This routine is called by the ICMP module when it gets some
619  * sort of error condition.  If err < 0 then the socket should
620  * be closed and the error returned to the user.  If err > 0
621  * it's just the icmp type << 8 | icmp code.
622  * Header points to the ip header of the error packet. We move
623  * on past this. Then (as it used to claim before adjustment)
624  * header points to the first 8 bytes of the udp header.  We need
625  * to find the appropriate port.
626  */
627 
628 void __udp4_lib_err(struct sk_buff *skb, u32 info, struct udp_table *udptable)
629 {
630         struct inet_sock *inet;
631         const struct iphdr *iph = (const struct iphdr *)skb->data;
632         struct udphdr *uh = (struct udphdr *)(skb->data+(iph->ihl<<2));
633         const int type = icmp_hdr(skb)->type;
634         const int code = icmp_hdr(skb)->code;
635         struct sock *sk;
636         int harderr;
637         int err;
638         struct net *net = dev_net(skb->dev);
639 
640         sk = __udp4_lib_lookup(net, iph->daddr, uh->dest,
641                         iph->saddr, uh->source, skb->dev->ifindex, udptable,
642                         NULL);
643         if (!sk) {
644                 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
645                 return; /* No socket for error */
646         }
647 
648         err = 0;
649         harderr = 0;
650         inet = inet_sk(sk);
651 
652         switch (type) {
653         default:
654         case ICMP_TIME_EXCEEDED:
655                 err = EHOSTUNREACH;
656                 break;
657         case ICMP_SOURCE_QUENCH:
658                 goto out;
659         case ICMP_PARAMETERPROB:
660                 err = EPROTO;
661                 harderr = 1;
662                 break;
663         case ICMP_DEST_UNREACH:
664                 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
665                         ipv4_sk_update_pmtu(skb, sk, info);
666                         if (inet->pmtudisc != IP_PMTUDISC_DONT) {
667                                 err = EMSGSIZE;
668                                 harderr = 1;
669                                 break;
670                         }
671                         goto out;
672                 }
673                 err = EHOSTUNREACH;
674                 if (code <= NR_ICMP_UNREACH) {
675                         harderr = icmp_err_convert[code].fatal;
676                         err = icmp_err_convert[code].errno;
677                 }
678                 break;
679         case ICMP_REDIRECT:
680                 ipv4_sk_redirect(skb, sk);
681                 goto out;
682         }
683 
684         /*
685          *      RFC1122: OK.  Passes ICMP errors back to application, as per
686          *      4.1.3.3.
687          */
688         if (!inet->recverr) {
689                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
690                         goto out;
691         } else
692                 ip_icmp_error(sk, skb, err, uh->dest, info, (u8 *)(uh+1));
693 
694         sk->sk_err = err;
695         sk->sk_error_report(sk);
696 out:
697         return;
698 }
699 
700 void udp_err(struct sk_buff *skb, u32 info)
701 {
702         __udp4_lib_err(skb, info, &udp_table);
703 }
704 
705 /*
706  * Throw away all pending data and cancel the corking. Socket is locked.
707  */
708 void udp_flush_pending_frames(struct sock *sk)
709 {
710         struct udp_sock *up = udp_sk(sk);
711 
712         if (up->pending) {
713                 up->len = 0;
714                 up->pending = 0;
715                 ip_flush_pending_frames(sk);
716         }
717 }
718 EXPORT_SYMBOL(udp_flush_pending_frames);
719 
720 /**
721  *      udp4_hwcsum  -  handle outgoing HW checksumming
722  *      @skb:   sk_buff containing the filled-in UDP header
723  *              (checksum field must be zeroed out)
724  *      @src:   source IP address
725  *      @dst:   destination IP address
726  */
727 void udp4_hwcsum(struct sk_buff *skb, __be32 src, __be32 dst)
728 {
729         struct udphdr *uh = udp_hdr(skb);
730         int offset = skb_transport_offset(skb);
731         int len = skb->len - offset;
732         int hlen = len;
733         __wsum csum = 0;
734 
735         if (!skb_has_frag_list(skb)) {
736                 /*
737                  * Only one fragment on the socket.
738                  */
739                 skb->csum_start = skb_transport_header(skb) - skb->head;
740                 skb->csum_offset = offsetof(struct udphdr, check);
741                 uh->check = ~csum_tcpudp_magic(src, dst, len,
742                                                IPPROTO_UDP, 0);
743         } else {
744                 struct sk_buff *frags;
745 
746                 /*
747                  * HW-checksum won't work as there are two or more
748                  * fragments on the socket so that all csums of sk_buffs
749                  * should be together
750                  */
751                 skb_walk_frags(skb, frags) {
752                         csum = csum_add(csum, frags->csum);
753                         hlen -= frags->len;
754                 }
755 
756                 csum = skb_checksum(skb, offset, hlen, csum);
757                 skb->ip_summed = CHECKSUM_NONE;
758 
759                 uh->check = csum_tcpudp_magic(src, dst, len, IPPROTO_UDP, csum);
760                 if (uh->check == 0)
761                         uh->check = CSUM_MANGLED_0;
762         }
763 }
764 EXPORT_SYMBOL_GPL(udp4_hwcsum);
765 
766 /* Function to set UDP checksum for an IPv4 UDP packet. This is intended
767  * for the simple case like when setting the checksum for a UDP tunnel.
768  */
769 void udp_set_csum(bool nocheck, struct sk_buff *skb,
770                   __be32 saddr, __be32 daddr, int len)
771 {
772         struct udphdr *uh = udp_hdr(skb);
773 
774         if (nocheck) {
775                 uh->check = 0;
776         } else if (skb_is_gso(skb)) {
777                 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
778         } else if (skb->ip_summed == CHECKSUM_PARTIAL) {
779                 uh->check = 0;
780                 uh->check = udp_v4_check(len, saddr, daddr, lco_csum(skb));
781                 if (uh->check == 0)
782                         uh->check = CSUM_MANGLED_0;
783         } else {
784                 skb->ip_summed = CHECKSUM_PARTIAL;
785                 skb->csum_start = skb_transport_header(skb) - skb->head;
786                 skb->csum_offset = offsetof(struct udphdr, check);
787                 uh->check = ~udp_v4_check(len, saddr, daddr, 0);
788         }
789 }
790 EXPORT_SYMBOL(udp_set_csum);
791 
792 static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
793 {
794         struct sock *sk = skb->sk;
795         struct inet_sock *inet = inet_sk(sk);
796         struct udphdr *uh;
797         int err = 0;
798         int is_udplite = IS_UDPLITE(sk);
799         int offset = skb_transport_offset(skb);
800         int len = skb->len - offset;
801         __wsum csum = 0;
802 
803         /*
804          * Create a UDP header
805          */
806         uh = udp_hdr(skb);
807         uh->source = inet->inet_sport;
808         uh->dest = fl4->fl4_dport;
809         uh->len = htons(len);
810         uh->check = 0;
811 
812         if (is_udplite)                                  /*     UDP-Lite      */
813                 csum = udplite_csum(skb);
814 
815         else if (sk->sk_no_check_tx) {   /* UDP csum disabled */
816 
817                 skb->ip_summed = CHECKSUM_NONE;
818                 goto send;
819 
820         } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
821 
822                 udp4_hwcsum(skb, fl4->saddr, fl4->daddr);
823                 goto send;
824 
825         } else
826                 csum = udp_csum(skb);
827 
828         /* add protocol-dependent pseudo-header */
829         uh->check = csum_tcpudp_magic(fl4->saddr, fl4->daddr, len,
830                                       sk->sk_protocol, csum);
831         if (uh->check == 0)
832                 uh->check = CSUM_MANGLED_0;
833 
834 send:
835         err = ip_send_skb(sock_net(sk), skb);
836         if (err) {
837                 if (err == -ENOBUFS && !inet->recverr) {
838                         UDP_INC_STATS(sock_net(sk),
839                                       UDP_MIB_SNDBUFERRORS, is_udplite);
840                         err = 0;
841                 }
842         } else
843                 UDP_INC_STATS(sock_net(sk),
844                               UDP_MIB_OUTDATAGRAMS, is_udplite);
845         return err;
846 }
847 
848 /*
849  * Push out all pending data as one UDP datagram. Socket is locked.
850  */
851 int udp_push_pending_frames(struct sock *sk)
852 {
853         struct udp_sock  *up = udp_sk(sk);
854         struct inet_sock *inet = inet_sk(sk);
855         struct flowi4 *fl4 = &inet->cork.fl.u.ip4;
856         struct sk_buff *skb;
857         int err = 0;
858 
859         skb = ip_finish_skb(sk, fl4);
860         if (!skb)
861                 goto out;
862 
863         err = udp_send_skb(skb, fl4);
864 
865 out:
866         up->len = 0;
867         up->pending = 0;
868         return err;
869 }
870 EXPORT_SYMBOL(udp_push_pending_frames);
871 
872 int udp_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
873 {
874         struct inet_sock *inet = inet_sk(sk);
875         struct udp_sock *up = udp_sk(sk);
876         struct flowi4 fl4_stack;
877         struct flowi4 *fl4;
878         int ulen = len;
879         struct ipcm_cookie ipc;
880         struct rtable *rt = NULL;
881         int free = 0;
882         int connected = 0;
883         __be32 daddr, faddr, saddr;
884         __be16 dport;
885         u8  tos;
886         int err, is_udplite = IS_UDPLITE(sk);
887         int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
888         int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
889         struct sk_buff *skb;
890         struct ip_options_data opt_copy;
891 
892         if (len > 0xFFFF)
893                 return -EMSGSIZE;
894 
895         /*
896          *      Check the flags.
897          */
898 
899         if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message compatibility */
900                 return -EOPNOTSUPP;
901 
902         ipc.opt = NULL;
903         ipc.tx_flags = 0;
904         ipc.ttl = 0;
905         ipc.tos = -1;
906 
907         getfrag = is_udplite ? udplite_getfrag : ip_generic_getfrag;
908 
909         fl4 = &inet->cork.fl.u.ip4;
910         if (up->pending) {
911                 /*
912                  * There are pending frames.
913                  * The socket lock must be held while it's corked.
914                  */
915                 lock_sock(sk);
916                 if (likely(up->pending)) {
917                         if (unlikely(up->pending != AF_INET)) {
918                                 release_sock(sk);
919                                 return -EINVAL;
920                         }
921                         goto do_append_data;
922                 }
923                 release_sock(sk);
924         }
925         ulen += sizeof(struct udphdr);
926 
927         /*
928          *      Get and verify the address.
929          */
930         if (msg->msg_name) {
931                 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
932                 if (msg->msg_namelen < sizeof(*usin))
933                         return -EINVAL;
934                 if (usin->sin_family != AF_INET) {
935                         if (usin->sin_family != AF_UNSPEC)
936                                 return -EAFNOSUPPORT;
937                 }
938 
939                 daddr = usin->sin_addr.s_addr;
940                 dport = usin->sin_port;
941                 if (dport == 0)
942                         return -EINVAL;
943         } else {
944                 if (sk->sk_state != TCP_ESTABLISHED)
945                         return -EDESTADDRREQ;
946                 daddr = inet->inet_daddr;
947                 dport = inet->inet_dport;
948                 /* Open fast path for connected socket.
949                    Route will not be used, if at least one option is set.
950                  */
951                 connected = 1;
952         }
953 
954         ipc.sockc.tsflags = sk->sk_tsflags;
955         ipc.addr = inet->inet_saddr;
956         ipc.oif = sk->sk_bound_dev_if;
957 
958         if (msg->msg_controllen) {
959                 err = ip_cmsg_send(sk, msg, &ipc, sk->sk_family == AF_INET6);
960                 if (unlikely(err)) {
961                         kfree(ipc.opt);
962                         return err;
963                 }
964                 if (ipc.opt)
965                         free = 1;
966                 connected = 0;
967         }
968         if (!ipc.opt) {
969                 struct ip_options_rcu *inet_opt;
970 
971                 rcu_read_lock();
972                 inet_opt = rcu_dereference(inet->inet_opt);
973                 if (inet_opt) {
974                         memcpy(&opt_copy, inet_opt,
975                                sizeof(*inet_opt) + inet_opt->opt.optlen);
976                         ipc.opt = &opt_copy.opt;
977                 }
978                 rcu_read_unlock();
979         }
980 
981         saddr = ipc.addr;
982         ipc.addr = faddr = daddr;
983 
984         sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
985 
986         if (ipc.opt && ipc.opt->opt.srr) {
987                 if (!daddr)
988                         return -EINVAL;
989                 faddr = ipc.opt->opt.faddr;
990                 connected = 0;
991         }
992         tos = get_rttos(&ipc, inet);
993         if (sock_flag(sk, SOCK_LOCALROUTE) ||
994             (msg->msg_flags & MSG_DONTROUTE) ||
995             (ipc.opt && ipc.opt->opt.is_strictroute)) {
996                 tos |= RTO_ONLINK;
997                 connected = 0;
998         }
999 
1000         if (ipv4_is_multicast(daddr)) {
1001                 if (!ipc.oif)
1002                         ipc.oif = inet->mc_index;
1003                 if (!saddr)
1004                         saddr = inet->mc_addr;
1005                 connected = 0;
1006         } else if (!ipc.oif)
1007                 ipc.oif = inet->uc_index;
1008 
1009         if (connected)
1010                 rt = (struct rtable *)sk_dst_check(sk, 0);
1011 
1012         if (!rt) {
1013                 struct net *net = sock_net(sk);
1014                 __u8 flow_flags = inet_sk_flowi_flags(sk);
1015 
1016                 fl4 = &fl4_stack;
1017 
1018                 flowi4_init_output(fl4, ipc.oif, sk->sk_mark, tos,
1019                                    RT_SCOPE_UNIVERSE, sk->sk_protocol,
1020                                    flow_flags,
1021                                    faddr, saddr, dport, inet->inet_sport);
1022 
1023                 if (!saddr && ipc.oif) {
1024                         err = l3mdev_get_saddr(net, ipc.oif, fl4);
1025                         if (err < 0)
1026                                 goto out;
1027                 }
1028 
1029                 security_sk_classify_flow(sk, flowi4_to_flowi(fl4));
1030                 rt = ip_route_output_flow(net, fl4, sk);
1031                 if (IS_ERR(rt)) {
1032                         err = PTR_ERR(rt);
1033                         rt = NULL;
1034                         if (err == -ENETUNREACH)
1035                                 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
1036                         goto out;
1037                 }
1038 
1039                 err = -EACCES;
1040                 if ((rt->rt_flags & RTCF_BROADCAST) &&
1041                     !sock_flag(sk, SOCK_BROADCAST))
1042                         goto out;
1043                 if (connected)
1044                         sk_dst_set(sk, dst_clone(&rt->dst));
1045         }
1046 
1047         if (msg->msg_flags&MSG_CONFIRM)
1048                 goto do_confirm;
1049 back_from_confirm:
1050 
1051         saddr = fl4->saddr;
1052         if (!ipc.addr)
1053                 daddr = ipc.addr = fl4->daddr;
1054 
1055         /* Lockless fast path for the non-corking case. */
1056         if (!corkreq) {
1057                 skb = ip_make_skb(sk, fl4, getfrag, msg, ulen,
1058                                   sizeof(struct udphdr), &ipc, &rt,
1059                                   msg->msg_flags);
1060                 err = PTR_ERR(skb);
1061                 if (!IS_ERR_OR_NULL(skb))
1062                         err = udp_send_skb(skb, fl4);
1063                 goto out;
1064         }
1065 
1066         lock_sock(sk);
1067         if (unlikely(up->pending)) {
1068                 /* The socket is already corked while preparing it. */
1069                 /* ... which is an evident application bug. --ANK */
1070                 release_sock(sk);
1071 
1072                 net_dbg_ratelimited("cork app bug 2\n");
1073                 err = -EINVAL;
1074                 goto out;
1075         }
1076         /*
1077          *      Now cork the socket to pend data.
1078          */
1079         fl4 = &inet->cork.fl.u.ip4;
1080         fl4->daddr = daddr;
1081         fl4->saddr = saddr;
1082         fl4->fl4_dport = dport;
1083         fl4->fl4_sport = inet->inet_sport;
1084         up->pending = AF_INET;
1085 
1086 do_append_data:
1087         up->len += ulen;
1088         err = ip_append_data(sk, fl4, getfrag, msg, ulen,
1089                              sizeof(struct udphdr), &ipc, &rt,
1090                              corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
1091         if (err)
1092                 udp_flush_pending_frames(sk);
1093         else if (!corkreq)
1094                 err = udp_push_pending_frames(sk);
1095         else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1096                 up->pending = 0;
1097         release_sock(sk);
1098 
1099 out:
1100         ip_rt_put(rt);
1101         if (free)
1102                 kfree(ipc.opt);
1103         if (!err)
1104                 return len;
1105         /*
1106          * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space.  Reporting
1107          * ENOBUFS might not be good (it's not tunable per se), but otherwise
1108          * we don't have a good statistic (IpOutDiscards but it can be too many
1109          * things).  We could add another new stat but at least for now that
1110          * seems like overkill.
1111          */
1112         if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1113                 UDP_INC_STATS(sock_net(sk),
1114                               UDP_MIB_SNDBUFERRORS, is_udplite);
1115         }
1116         return err;
1117 
1118 do_confirm:
1119         dst_confirm(&rt->dst);
1120         if (!(msg->msg_flags&MSG_PROBE) || len)
1121                 goto back_from_confirm;
1122         err = 0;
1123         goto out;
1124 }
1125 EXPORT_SYMBOL(udp_sendmsg);
1126 
1127 int udp_sendpage(struct sock *sk, struct page *page, int offset,
1128                  size_t size, int flags)
1129 {
1130         struct inet_sock *inet = inet_sk(sk);
1131         struct udp_sock *up = udp_sk(sk);
1132         int ret;
1133 
1134         if (flags & MSG_SENDPAGE_NOTLAST)
1135                 flags |= MSG_MORE;
1136 
1137         if (!up->pending) {
1138                 struct msghdr msg = {   .msg_flags = flags|MSG_MORE };
1139 
1140                 /* Call udp_sendmsg to specify destination address which
1141                  * sendpage interface can't pass.
1142                  * This will succeed only when the socket is connected.
1143                  */
1144                 ret = udp_sendmsg(sk, &msg, 0);
1145                 if (ret < 0)
1146                         return ret;
1147         }
1148 
1149         lock_sock(sk);
1150 
1151         if (unlikely(!up->pending)) {
1152                 release_sock(sk);
1153 
1154                 net_dbg_ratelimited("udp cork app bug 3\n");
1155                 return -EINVAL;
1156         }
1157 
1158         ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
1159                              page, offset, size, flags);
1160         if (ret == -EOPNOTSUPP) {
1161                 release_sock(sk);
1162                 return sock_no_sendpage(sk->sk_socket, page, offset,
1163                                         size, flags);
1164         }
1165         if (ret < 0) {
1166                 udp_flush_pending_frames(sk);
1167                 goto out;
1168         }
1169 
1170         up->len += size;
1171         if (!(up->corkflag || (flags&MSG_MORE)))
1172                 ret = udp_push_pending_frames(sk);
1173         if (!ret)
1174                 ret = size;
1175 out:
1176         release_sock(sk);
1177         return ret;
1178 }
1179 
1180 /**
1181  *      first_packet_length     - return length of first packet in receive queue
1182  *      @sk: socket
1183  *
1184  *      Drops all bad checksum frames, until a valid one is found.
1185  *      Returns the length of found skb, or -1 if none is found.
1186  */
1187 static int first_packet_length(struct sock *sk)
1188 {
1189         struct sk_buff_head list_kill, *rcvq = &sk->sk_receive_queue;
1190         struct sk_buff *skb;
1191         int res;
1192 
1193         __skb_queue_head_init(&list_kill);
1194 
1195         spin_lock_bh(&rcvq->lock);
1196         while ((skb = skb_peek(rcvq)) != NULL &&
1197                 udp_lib_checksum_complete(skb)) {
1198                 __UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS,
1199                                 IS_UDPLITE(sk));
1200                 __UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
1201                                 IS_UDPLITE(sk));
1202                 atomic_inc(&sk->sk_drops);
1203                 __skb_unlink(skb, rcvq);
1204                 __skb_queue_tail(&list_kill, skb);
1205         }
1206         res = skb ? skb->len : -1;
1207         spin_unlock_bh(&rcvq->lock);
1208 
1209         if (!skb_queue_empty(&list_kill)) {
1210                 bool slow = lock_sock_fast(sk);
1211 
1212                 __skb_queue_purge(&list_kill);
1213                 sk_mem_reclaim_partial(sk);
1214                 unlock_sock_fast(sk, slow);
1215         }
1216         return res;
1217 }
1218 
1219 /*
1220  *      IOCTL requests applicable to the UDP protocol
1221  */
1222 
1223 int udp_ioctl(struct sock *sk, int cmd, unsigned long arg)
1224 {
1225         switch (cmd) {
1226         case SIOCOUTQ:
1227         {
1228                 int amount = sk_wmem_alloc_get(sk);
1229 
1230                 return put_user(amount, (int __user *)arg);
1231         }
1232 
1233         case SIOCINQ:
1234         {
1235                 int amount = max_t(int, 0, first_packet_length(sk));
1236 
1237                 return put_user(amount, (int __user *)arg);
1238         }
1239 
1240         default:
1241                 return -ENOIOCTLCMD;
1242         }
1243 
1244         return 0;
1245 }
1246 EXPORT_SYMBOL(udp_ioctl);
1247 
1248 /*
1249  *      This should be easy, if there is something there we
1250  *      return it, otherwise we block.
1251  */
1252 
1253 int udp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
1254                 int flags, int *addr_len)
1255 {
1256         struct inet_sock *inet = inet_sk(sk);
1257         DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
1258         struct sk_buff *skb;
1259         unsigned int ulen, copied;
1260         int peeked, peeking, off;
1261         int err;
1262         int is_udplite = IS_UDPLITE(sk);
1263         bool checksum_valid = false;
1264         bool slow;
1265 
1266         if (flags & MSG_ERRQUEUE)
1267                 return ip_recv_error(sk, msg, len, addr_len);
1268 
1269 try_again:
1270         peeking = off = sk_peek_offset(sk, flags);
1271         skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
1272                                   &peeked, &off, &err);
1273         if (!skb)
1274                 return err;
1275         if (ccs_socket_post_recvmsg_permission(sk, skb, flags))
1276                 return -EAGAIN; /* Hope less harmful than -EPERM. */
1277 
1278         ulen = skb->len;
1279         copied = len;
1280         if (copied > ulen - off)
1281                 copied = ulen - off;
1282         else if (copied < ulen)
1283                 msg->msg_flags |= MSG_TRUNC;
1284 
1285         /*
1286          * If checksum is needed at all, try to do it while copying the
1287          * data.  If the data is truncated, or if we only want a partial
1288          * coverage checksum (UDP-Lite), do it before the copy.
1289          */
1290 
1291         if (copied < ulen || UDP_SKB_CB(skb)->partial_cov || peeking) {
1292                 checksum_valid = !udp_lib_checksum_complete(skb);
1293                 if (!checksum_valid)
1294                         goto csum_copy_err;
1295         }
1296 
1297         if (checksum_valid || skb_csum_unnecessary(skb))
1298                 err = skb_copy_datagram_msg(skb, off, msg, copied);
1299         else {
1300                 err = skb_copy_and_csum_datagram_msg(skb, off, msg);
1301 
1302                 if (err == -EINVAL)
1303                         goto csum_copy_err;
1304         }
1305 
1306         if (unlikely(err)) {
1307                 trace_kfree_skb(skb, udp_recvmsg);
1308                 if (!peeked) {
1309                         atomic_inc(&sk->sk_drops);
1310                         UDP_INC_STATS(sock_net(sk),
1311                                       UDP_MIB_INERRORS, is_udplite);
1312                 }
1313                 skb_free_datagram_locked(sk, skb);
1314                 return err;
1315         }
1316 
1317         if (!peeked)
1318                 UDP_INC_STATS(sock_net(sk),
1319                               UDP_MIB_INDATAGRAMS, is_udplite);
1320 
1321         sock_recv_ts_and_drops(msg, sk, skb);
1322 
1323         /* Copy the address. */
1324         if (sin) {
1325                 sin->sin_family = AF_INET;
1326                 sin->sin_port = udp_hdr(skb)->source;
1327                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
1328                 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
1329                 *addr_len = sizeof(*sin);
1330         }
1331         if (inet->cmsg_flags)
1332                 ip_cmsg_recv_offset(msg, skb, sizeof(struct udphdr), off);
1333 
1334         err = copied;
1335         if (flags & MSG_TRUNC)
1336                 err = ulen;
1337 
1338         __skb_free_datagram_locked(sk, skb, peeking ? -err : err);
1339         return err;
1340 
1341 csum_copy_err:
1342         slow = lock_sock_fast(sk);
1343         if (!skb_kill_datagram(sk, skb, flags)) {
1344                 UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
1345                 UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1346         }
1347         unlock_sock_fast(sk, slow);
1348 
1349         /* starting over for a new packet, but check if we need to yield */
1350         cond_resched();
1351         msg->msg_flags &= ~MSG_TRUNC;
1352         goto try_again;
1353 }
1354 
1355 int udp_disconnect(struct sock *sk, int flags)
1356 {
1357         struct inet_sock *inet = inet_sk(sk);
1358         /*
1359          *      1003.1g - break association.
1360          */
1361 
1362         sk->sk_state = TCP_CLOSE;
1363         inet->inet_daddr = 0;
1364         inet->inet_dport = 0;
1365         sock_rps_reset_rxhash(sk);
1366         sk->sk_bound_dev_if = 0;
1367         if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
1368                 inet_reset_saddr(sk);
1369 
1370         if (!(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) {
1371                 sk->sk_prot->unhash(sk);
1372                 inet->inet_sport = 0;
1373         }
1374         sk_dst_reset(sk);
1375         return 0;
1376 }
1377 EXPORT_SYMBOL(udp_disconnect);
1378 
1379 void udp_lib_unhash(struct sock *sk)
1380 {
1381         if (sk_hashed(sk)) {
1382                 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1383                 struct udp_hslot *hslot, *hslot2;
1384 
1385                 hslot  = udp_hashslot(udptable, sock_net(sk),
1386                                       udp_sk(sk)->udp_port_hash);
1387                 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1388 
1389                 spin_lock_bh(&hslot->lock);
1390                 if (rcu_access_pointer(sk->sk_reuseport_cb))
1391                         reuseport_detach_sock(sk);
1392                 if (sk_del_node_init_rcu(sk)) {
1393                         hslot->count--;
1394                         inet_sk(sk)->inet_num = 0;
1395                         sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
1396 
1397                         spin_lock(&hslot2->lock);
1398                         hlist_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1399                         hslot2->count--;
1400                         spin_unlock(&hslot2->lock);
1401                 }
1402                 spin_unlock_bh(&hslot->lock);
1403         }
1404 }
1405 EXPORT_SYMBOL(udp_lib_unhash);
1406 
1407 /*
1408  * inet_rcv_saddr was changed, we must rehash secondary hash
1409  */
1410 void udp_lib_rehash(struct sock *sk, u16 newhash)
1411 {
1412         if (sk_hashed(sk)) {
1413                 struct udp_table *udptable = sk->sk_prot->h.udp_table;
1414                 struct udp_hslot *hslot, *hslot2, *nhslot2;
1415 
1416                 hslot2 = udp_hashslot2(udptable, udp_sk(sk)->udp_portaddr_hash);
1417                 nhslot2 = udp_hashslot2(udptable, newhash);
1418                 udp_sk(sk)->udp_portaddr_hash = newhash;
1419 
1420                 if (hslot2 != nhslot2 ||
1421                     rcu_access_pointer(sk->sk_reuseport_cb)) {
1422                         hslot = udp_hashslot(udptable, sock_net(sk),
1423                                              udp_sk(sk)->udp_port_hash);
1424                         /* we must lock primary chain too */
1425                         spin_lock_bh(&hslot->lock);
1426                         if (rcu_access_pointer(sk->sk_reuseport_cb))
1427                                 reuseport_detach_sock(sk);
1428 
1429                         if (hslot2 != nhslot2) {
1430                                 spin_lock(&hslot2->lock);
1431                                 hlist_del_init_rcu(&udp_sk(sk)->udp_portaddr_node);
1432                                 hslot2->count--;
1433                                 spin_unlock(&hslot2->lock);
1434 
1435                                 spin_lock(&nhslot2->lock);
1436                                 hlist_add_head_rcu(&udp_sk(sk)->udp_portaddr_node,
1437                                                          &nhslot2->head);
1438                                 nhslot2->count++;
1439                                 spin_unlock(&nhslot2->lock);
1440                         }
1441 
1442                         spin_unlock_bh(&hslot->lock);
1443                 }
1444         }
1445 }
1446 EXPORT_SYMBOL(udp_lib_rehash);
1447 
1448 static void udp_v4_rehash(struct sock *sk)
1449 {
1450         u16 new_hash = udp4_portaddr_hash(sock_net(sk),
1451                                           inet_sk(sk)->inet_rcv_saddr,
1452                                           inet_sk(sk)->inet_num);
1453         udp_lib_rehash(sk, new_hash);
1454 }
1455 
1456 int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1457 {
1458         int rc;
1459 
1460         if (inet_sk(sk)->inet_daddr) {
1461                 sock_rps_save_rxhash(sk, skb);
1462                 sk_mark_napi_id(sk, skb);
1463                 sk_incoming_cpu_update(sk);
1464         }
1465 
1466         rc = __sock_queue_rcv_skb(sk, skb);
1467         if (rc < 0) {
1468                 int is_udplite = IS_UDPLITE(sk);
1469 
1470                 /* Note that an ENOMEM error is charged twice */
1471                 if (rc == -ENOMEM)
1472                         UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1473                                         is_udplite);
1474                 UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1475                 kfree_skb(skb);
1476                 trace_udp_fail_queue_rcv_skb(rc, sk);
1477                 return -1;
1478         }
1479 
1480         return 0;
1481 
1482 }
1483 
1484 static struct static_key udp_encap_needed __read_mostly;
1485 void udp_encap_enable(void)
1486 {
1487         if (!static_key_enabled(&udp_encap_needed))
1488                 static_key_slow_inc(&udp_encap_needed);
1489 }
1490 EXPORT_SYMBOL(udp_encap_enable);
1491 
1492 /* returns:
1493  *  -1: error
1494  *   0: success
1495  *  >0: "udp encap" protocol resubmission
1496  *
1497  * Note that in the success and error cases, the skb is assumed to
1498  * have either been requeued or freed.
1499  */
1500 int udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1501 {
1502         struct udp_sock *up = udp_sk(sk);
1503         int rc;
1504         int is_udplite = IS_UDPLITE(sk);
1505 
1506         /*
1507          *      Charge it to the socket, dropping if the queue is full.
1508          */
1509         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
1510                 goto drop;
1511         nf_reset(skb);
1512 
1513         if (static_key_false(&udp_encap_needed) && up->encap_type) {
1514                 int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
1515 
1516                 /*
1517                  * This is an encapsulation socket so pass the skb to
1518                  * the socket's udp_encap_rcv() hook. Otherwise, just
1519                  * fall through and pass this up the UDP socket.
1520                  * up->encap_rcv() returns the following value:
1521                  * =0 if skb was successfully passed to the encap
1522                  *    handler or was discarded by it.
1523                  * >0 if skb should be passed on to UDP.
1524                  * <0 if skb should be resubmitted as proto -N
1525                  */
1526 
1527                 /* if we're overly short, let UDP handle it */
1528                 encap_rcv = ACCESS_ONCE(up->encap_rcv);
1529                 if (encap_rcv) {
1530                         int ret;
1531 
1532                         /* Verify checksum before giving to encap */
1533                         if (udp_lib_checksum_complete(skb))
1534                                 goto csum_error;
1535 
1536                         ret = encap_rcv(sk, skb);
1537                         if (ret <= 0) {
1538                                 __UDP_INC_STATS(sock_net(sk),
1539                                                 UDP_MIB_INDATAGRAMS,
1540                                                 is_udplite);
1541                                 return -ret;
1542                         }
1543                 }
1544 
1545                 /* FALLTHROUGH -- it's a UDP Packet */
1546         }
1547 
1548         /*
1549          *      UDP-Lite specific tests, ignored on UDP sockets
1550          */
1551         if ((is_udplite & UDPLITE_RECV_CC)  &&  UDP_SKB_CB(skb)->partial_cov) {
1552 
1553                 /*
1554                  * MIB statistics other than incrementing the error count are
1555                  * disabled for the following two types of errors: these depend
1556                  * on the application settings, not on the functioning of the
1557                  * protocol stack as such.
1558                  *
1559                  * RFC 3828 here recommends (sec 3.3): "There should also be a
1560                  * way ... to ... at least let the receiving application block
1561                  * delivery of packets with coverage values less than a value
1562                  * provided by the application."
1563                  */
1564                 if (up->pcrlen == 0) {          /* full coverage was set  */
1565                         net_dbg_ratelimited("UDPLite: partial coverage %d while full coverage %d requested\n",
1566                                             UDP_SKB_CB(skb)->cscov, skb->len);
1567                         goto drop;
1568                 }
1569                 /* The next case involves violating the min. coverage requested
1570                  * by the receiver. This is subtle: if receiver wants x and x is
1571                  * greater than the buffersize/MTU then receiver will complain
1572                  * that it wants x while sender emits packets of smaller size y.
1573                  * Therefore the above ...()->partial_cov statement is essential.
1574                  */
1575                 if (UDP_SKB_CB(skb)->cscov  <  up->pcrlen) {
1576                         net_dbg_ratelimited("UDPLite: coverage %d too small, need min %d\n",
1577                                             UDP_SKB_CB(skb)->cscov, up->pcrlen);
1578                         goto drop;
1579                 }
1580         }
1581 
1582         if (rcu_access_pointer(sk->sk_filter) &&
1583             udp_lib_checksum_complete(skb))
1584                         goto csum_error;
1585 
1586         if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr)))
1587                 goto drop;
1588 
1589         udp_csum_pull_header(skb);
1590         if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
1591                 __UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS,
1592                                 is_udplite);
1593                 goto drop;
1594         }
1595 
1596         rc = 0;
1597 
1598         ipv4_pktinfo_prepare(sk, skb);
1599         bh_lock_sock(sk);
1600         if (!sock_owned_by_user(sk))
1601                 rc = __udp_queue_rcv_skb(sk, skb);
1602         else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
1603                 bh_unlock_sock(sk);
1604                 goto drop;
1605         }
1606         bh_unlock_sock(sk);
1607 
1608         return rc;
1609 
1610 csum_error:
1611         __UDP_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
1612 drop:
1613         __UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
1614         atomic_inc(&sk->sk_drops);
1615         kfree_skb(skb);
1616         return -1;
1617 }
1618 
1619 /* For TCP sockets, sk_rx_dst is protected by socket lock
1620  * For UDP, we use xchg() to guard against concurrent changes.
1621  */
1622 static void udp_sk_rx_dst_set(struct sock *sk, struct dst_entry *dst)
1623 {
1624         struct dst_entry *old;
1625 
1626         dst_hold(dst);
1627         old = xchg(&sk->sk_rx_dst, dst);
1628         dst_release(old);
1629 }
1630 
1631 /*
1632  *      Multicasts and broadcasts go to each listener.
1633  *
1634  *      Note: called only from the BH handler context.
1635  */
1636 static int __udp4_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
1637                                     struct udphdr  *uh,
1638                                     __be32 saddr, __be32 daddr,
1639                                     struct udp_table *udptable,
1640                                     int proto)
1641 {
1642         struct sock *sk, *first = NULL;
1643         unsigned short hnum = ntohs(uh->dest);
1644         struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
1645         unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
1646         unsigned int offset = offsetof(typeof(*sk), sk_node);
1647         int dif = skb->dev->ifindex;
1648         struct hlist_node *node;
1649         struct sk_buff *nskb;
1650 
1651         if (use_hash2) {
1652                 hash2_any = udp4_portaddr_hash(net, htonl(INADDR_ANY), hnum) &
1653                             udp_table.mask;
1654                 hash2 = udp4_portaddr_hash(net, daddr, hnum) & udp_table.mask;
1655 start_lookup:
1656                 hslot = &udp_table.hash2[hash2];
1657                 offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
1658         }
1659 
1660         sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
1661                 if (!__udp_is_mcast_sock(net, sk, uh->dest, daddr,
1662                                          uh->source, saddr, dif, hnum))
1663                         continue;
1664 
1665                 if (!first) {
1666                         first = sk;
1667                         continue;
1668                 }
1669                 nskb = skb_clone(skb, GFP_ATOMIC);
1670 
1671                 if (unlikely(!nskb)) {
1672                         atomic_inc(&sk->sk_drops);
1673                         __UDP_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
1674                                         IS_UDPLITE(sk));
1675                         __UDP_INC_STATS(net, UDP_MIB_INERRORS,
1676                                         IS_UDPLITE(sk));
1677                         continue;
1678                 }
1679                 if (udp_queue_rcv_skb(sk, nskb) > 0)
1680                         consume_skb(nskb);
1681         }
1682 
1683         /* Also lookup *:port if we are using hash2 and haven't done so yet. */
1684         if (use_hash2 && hash2 != hash2_any) {
1685                 hash2 = hash2_any;
1686                 goto start_lookup;
1687         }
1688 
1689         if (first) {
1690                 if (udp_queue_rcv_skb(first, skb) > 0)
1691                         consume_skb(skb);
1692         } else {
1693                 kfree_skb(skb);
1694                 __UDP_INC_STATS(net, UDP_MIB_IGNOREDMULTI,
1695                                 proto == IPPROTO_UDPLITE);
1696         }
1697         return 0;
1698 }
1699 
1700 /* Initialize UDP checksum. If exited with zero value (success),
1701  * CHECKSUM_UNNECESSARY means, that no more checks are required.
1702  * Otherwise, csum completion requires chacksumming packet body,
1703  * including udp header and folding it to skb->csum.
1704  */
1705 static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
1706                                  int proto)
1707 {
1708         int err;
1709 
1710         UDP_SKB_CB(skb)->partial_cov = 0;
1711         UDP_SKB_CB(skb)->cscov = skb->len;
1712 
1713         if (proto == IPPROTO_UDPLITE) {
1714                 err = udplite_checksum_init(skb, uh);
1715                 if (err)
1716                         return err;
1717         }
1718 
1719         /* Note, we are only interested in != 0 or == 0, thus the
1720          * force to int.
1721          */
1722         return (__force int)skb_checksum_init_zero_check(skb, proto, uh->check,
1723                                                          inet_compute_pseudo);
1724 }
1725 
1726 /*
1727  *      All we need to do is get the socket, and then do a checksum.
1728  */
1729 
1730 int __udp4_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
1731                    int proto)
1732 {
1733         struct sock *sk;
1734         struct udphdr *uh;
1735         unsigned short ulen;
1736         struct rtable *rt = skb_rtable(skb);
1737         __be32 saddr, daddr;
1738         struct net *net = dev_net(skb->dev);
1739 
1740         /*
1741          *  Validate the packet.
1742          */
1743         if (!pskb_may_pull(skb, sizeof(struct udphdr)))
1744                 goto drop;              /* No space for header. */
1745 
1746         uh   = udp_hdr(skb);
1747         ulen = ntohs(uh->len);
1748         saddr = ip_hdr(skb)->saddr;
1749         daddr = ip_hdr(skb)->daddr;
1750 
1751         if (ulen > skb->len)
1752                 goto short_packet;
1753 
1754         if (proto == IPPROTO_UDP) {
1755                 /* UDP validates ulen. */
1756                 if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
1757                         goto short_packet;
1758                 uh = udp_hdr(skb);
1759         }
1760 
1761         if (udp4_csum_init(skb, uh, proto))
1762                 goto csum_error;
1763 
1764         sk = skb_steal_sock(skb);
1765         if (sk) {
1766                 struct dst_entry *dst = skb_dst(skb);
1767                 int ret;
1768 
1769                 if (unlikely(sk->sk_rx_dst != dst))
1770                         udp_sk_rx_dst_set(sk, dst);
1771 
1772                 ret = udp_queue_rcv_skb(sk, skb);
1773                 sock_put(sk);
1774                 /* a return value > 0 means to resubmit the input, but
1775                  * it wants the return to be -protocol, or 0
1776                  */
1777                 if (ret > 0)
1778                         return -ret;
1779                 return 0;
1780         }
1781 
1782         if (rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
1783                 return __udp4_lib_mcast_deliver(net, skb, uh,
1784                                                 saddr, daddr, udptable, proto);
1785 
1786         sk = __udp4_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
1787         if (sk) {
1788                 int ret;
1789 
1790                 if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
1791                         skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
1792                                                  inet_compute_pseudo);
1793 
1794                 ret = udp_queue_rcv_skb(sk, skb);
1795 
1796                 /* a return value > 0 means to resubmit the input, but
1797                  * it wants the return to be -protocol, or 0
1798                  */
1799                 if (ret > 0)
1800                         return -ret;
1801                 return 0;
1802         }
1803 
1804         if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
1805                 goto drop;
1806         nf_reset(skb);
1807 
1808         /* No socket. Drop packet silently, if checksum is wrong */
1809         if (udp_lib_checksum_complete(skb))
1810                 goto csum_error;
1811 
1812         __UDP_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
1813         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
1814 
1815         /*
1816          * Hmm.  We got an UDP packet to a port to which we
1817          * don't wanna listen.  Ignore it.
1818          */
1819         kfree_skb(skb);
1820         return 0;
1821 
1822 short_packet:
1823         net_dbg_ratelimited("UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
1824                             proto == IPPROTO_UDPLITE ? "Lite" : "",
1825                             &saddr, ntohs(uh->source),
1826                             ulen, skb->len,
1827                             &daddr, ntohs(uh->dest));
1828         goto drop;
1829 
1830 csum_error:
1831         /*
1832          * RFC1122: OK.  Discards the bad packet silently (as far as
1833          * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1834          */
1835         net_dbg_ratelimited("UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
1836                             proto == IPPROTO_UDPLITE ? "Lite" : "",
1837                             &saddr, ntohs(uh->source), &daddr, ntohs(uh->dest),
1838                             ulen);
1839         __UDP_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
1840 drop:
1841         __UDP_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
1842         kfree_skb(skb);
1843         return 0;
1844 }
1845 
1846 /* We can only early demux multicast if there is a single matching socket.
1847  * If more than one socket found returns NULL
1848  */
1849 static struct sock *__udp4_lib_mcast_demux_lookup(struct net *net,
1850                                                   __be16 loc_port, __be32 loc_addr,
1851                                                   __be16 rmt_port, __be32 rmt_addr,
1852                                                   int dif)
1853 {
1854         struct sock *sk, *result;
1855         unsigned short hnum = ntohs(loc_port);
1856         unsigned int slot = udp_hashfn(net, hnum, udp_table.mask);
1857         struct udp_hslot *hslot = &udp_table.hash[slot];
1858 
1859         /* Do not bother scanning a too big list */
1860         if (hslot->count > 10)
1861                 return NULL;
1862 
1863         result = NULL;
1864         sk_for_each_rcu(sk, &hslot->head) {
1865                 if (__udp_is_mcast_sock(net, sk, loc_port, loc_addr,
1866                                         rmt_port, rmt_addr, dif, hnum)) {
1867                         if (result)
1868                                 return NULL;
1869                         result = sk;
1870                 }
1871         }
1872 
1873         return result;
1874 }
1875 
1876 /* For unicast we should only early demux connected sockets or we can
1877  * break forwarding setups.  The chains here can be long so only check
1878  * if the first socket is an exact match and if not move on.
1879  */
1880 static struct sock *__udp4_lib_demux_lookup(struct net *net,
1881                                             __be16 loc_port, __be32 loc_addr,
1882                                             __be16 rmt_port, __be32 rmt_addr,
1883                                             int dif)
1884 {
1885         unsigned short hnum = ntohs(loc_port);
1886         unsigned int hash2 = udp4_portaddr_hash(net, loc_addr, hnum);
1887         unsigned int slot2 = hash2 & udp_table.mask;
1888         struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
1889         INET_ADDR_COOKIE(acookie, rmt_addr, loc_addr);
1890         const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
1891         struct sock *sk;
1892 
1893         udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
1894                 if (INET_MATCH(sk, net, acookie, rmt_addr,
1895                                loc_addr, ports, dif))
1896                         return sk;
1897                 /* Only check first socket in chain */
1898                 break;
1899         }
1900         return NULL;
1901 }
1902 
1903 void udp_v4_early_demux(struct sk_buff *skb)
1904 {
1905         struct net *net = dev_net(skb->dev);
1906         const struct iphdr *iph;
1907         const struct udphdr *uh;
1908         struct sock *sk = NULL;
1909         struct dst_entry *dst;
1910         int dif = skb->dev->ifindex;
1911         int ours;
1912 
1913         /* validate the packet */
1914         if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct udphdr)))
1915                 return;
1916 
1917         iph = ip_hdr(skb);
1918         uh = udp_hdr(skb);
1919 
1920         if (skb->pkt_type == PACKET_BROADCAST ||
1921             skb->pkt_type == PACKET_MULTICAST) {
1922                 struct in_device *in_dev = __in_dev_get_rcu(skb->dev);
1923 
1924                 if (!in_dev)
1925                         return;
1926 
1927                 /* we are supposed to accept bcast packets */
1928                 if (skb->pkt_type == PACKET_MULTICAST) {
1929                         ours = ip_check_mc_rcu(in_dev, iph->daddr, iph->saddr,
1930                                                iph->protocol);
1931                         if (!ours)
1932                                 return;
1933                 }
1934 
1935                 sk = __udp4_lib_mcast_demux_lookup(net, uh->dest, iph->daddr,
1936                                                    uh->source, iph->saddr, dif);
1937         } else if (skb->pkt_type == PACKET_HOST) {
1938                 sk = __udp4_lib_demux_lookup(net, uh->dest, iph->daddr,
1939                                              uh->source, iph->saddr, dif);
1940         }
1941 
1942         if (!sk || !atomic_inc_not_zero_hint(&sk->sk_refcnt, 2))
1943                 return;
1944 
1945         skb->sk = sk;
1946         skb->destructor = sock_efree;
1947         dst = READ_ONCE(sk->sk_rx_dst);
1948 
1949         if (dst)
1950                 dst = dst_check(dst, 0);
1951         if (dst) {
1952                 /* DST_NOCACHE can not be used without taking a reference */
1953                 if (dst->flags & DST_NOCACHE) {
1954                         if (likely(atomic_inc_not_zero(&dst->__refcnt)))
1955                                 skb_dst_set(skb, dst);
1956                 } else {
1957                         skb_dst_set_noref(skb, dst);
1958                 }
1959         }
1960 }
1961 
1962 int udp_rcv(struct sk_buff *skb)
1963 {
1964         return __udp4_lib_rcv(skb, &udp_table, IPPROTO_UDP);
1965 }
1966 
1967 void udp_destroy_sock(struct sock *sk)
1968 {
1969         struct udp_sock *up = udp_sk(sk);
1970         bool slow = lock_sock_fast(sk);
1971         udp_flush_pending_frames(sk);
1972         unlock_sock_fast(sk, slow);
1973         if (static_key_false(&udp_encap_needed) && up->encap_type) {
1974                 void (*encap_destroy)(struct sock *sk);
1975                 encap_destroy = ACCESS_ONCE(up->encap_destroy);
1976                 if (encap_destroy)
1977                         encap_destroy(sk);
1978         }
1979 }
1980 
1981 /*
1982  *      Socket option code for UDP
1983  */
1984 int udp_lib_setsockopt(struct sock *sk, int level, int optname,
1985                        char __user *optval, unsigned int optlen,
1986                        int (*push_pending_frames)(struct sock *))
1987 {
1988         struct udp_sock *up = udp_sk(sk);
1989         int val, valbool;
1990         int err = 0;
1991         int is_udplite = IS_UDPLITE(sk);
1992 
1993         if (optlen < sizeof(int))
1994                 return -EINVAL;
1995 
1996         if (get_user(val, (int __user *)optval))
1997                 return -EFAULT;
1998 
1999         valbool = val ? 1 : 0;
2000 
2001         switch (optname) {
2002         case UDP_CORK:
2003                 if (val != 0) {
2004                         up->corkflag = 1;
2005                 } else {
2006                         up->corkflag = 0;
2007                         lock_sock(sk);
2008                         push_pending_frames(sk);
2009                         release_sock(sk);
2010                 }
2011                 break;
2012 
2013         case UDP_ENCAP:
2014                 switch (val) {
2015                 case 0:
2016                 case UDP_ENCAP_ESPINUDP:
2017                 case UDP_ENCAP_ESPINUDP_NON_IKE:
2018                         up->encap_rcv = xfrm4_udp_encap_rcv;
2019                         /* FALLTHROUGH */
2020                 case UDP_ENCAP_L2TPINUDP:
2021                         up->encap_type = val;
2022                         udp_encap_enable();
2023                         break;
2024                 default:
2025                         err = -ENOPROTOOPT;
2026                         break;
2027                 }
2028                 break;
2029 
2030         case UDP_NO_CHECK6_TX:
2031                 up->no_check6_tx = valbool;
2032                 break;
2033 
2034         case UDP_NO_CHECK6_RX:
2035                 up->no_check6_rx = valbool;
2036                 break;
2037 
2038         /*
2039          *      UDP-Lite's partial checksum coverage (RFC 3828).
2040          */
2041         /* The sender sets actual checksum coverage length via this option.
2042          * The case coverage > packet length is handled by send module. */
2043         case UDPLITE_SEND_CSCOV:
2044                 if (!is_udplite)         /* Disable the option on UDP sockets */
2045                         return -ENOPROTOOPT;
2046                 if (val != 0 && val < 8) /* Illegal coverage: use default (8) */
2047                         val = 8;
2048                 else if (val > USHRT_MAX)
2049                         val = USHRT_MAX;
2050                 up->pcslen = val;
2051                 up->pcflag |= UDPLITE_SEND_CC;
2052                 break;
2053 
2054         /* The receiver specifies a minimum checksum coverage value. To make
2055          * sense, this should be set to at least 8 (as done below). If zero is
2056          * used, this again means full checksum coverage.                     */
2057         case UDPLITE_RECV_CSCOV:
2058                 if (!is_udplite)         /* Disable the option on UDP sockets */
2059                         return -ENOPROTOOPT;
2060                 if (val != 0 && val < 8) /* Avoid silly minimal values.       */
2061                         val = 8;
2062                 else if (val > USHRT_MAX)
2063                         val = USHRT_MAX;
2064                 up->pcrlen = val;
2065                 up->pcflag |= UDPLITE_RECV_CC;
2066                 break;
2067 
2068         default:
2069                 err = -ENOPROTOOPT;
2070                 break;
2071         }
2072 
2073         return err;
2074 }
2075 EXPORT_SYMBOL(udp_lib_setsockopt);
2076 
2077 int udp_setsockopt(struct sock *sk, int level, int optname,
2078                    char __user *optval, unsigned int optlen)
2079 {
2080         if (level == SOL_UDP  ||  level == SOL_UDPLITE)
2081                 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2082                                           udp_push_pending_frames);
2083         return ip_setsockopt(sk, level, optname, optval, optlen);
2084 }
2085 
2086 #ifdef CONFIG_COMPAT
2087 int compat_udp_setsockopt(struct sock *sk, int level, int optname,
2088                           char __user *optval, unsigned int optlen)
2089 {
2090         if (level == SOL_UDP  ||  level == SOL_UDPLITE)
2091                 return udp_lib_setsockopt(sk, level, optname, optval, optlen,
2092                                           udp_push_pending_frames);
2093         return compat_ip_setsockopt(sk, level, optname, optval, optlen);
2094 }
2095 #endif
2096 
2097 int udp_lib_getsockopt(struct sock *sk, int level, int optname,
2098                        char __user *optval, int __user *optlen)
2099 {
2100         struct udp_sock *up = udp_sk(sk);
2101         int val, len;
2102 
2103         if (get_user(len, optlen))
2104                 return -EFAULT;
2105 
2106         len = min_t(unsigned int, len, sizeof(int));
2107 
2108         if (len < 0)
2109                 return -EINVAL;
2110 
2111         switch (optname) {
2112         case UDP_CORK:
2113                 val = up->corkflag;
2114                 break;
2115 
2116         case UDP_ENCAP:
2117                 val = up->encap_type;
2118                 break;
2119 
2120         case UDP_NO_CHECK6_TX:
2121                 val = up->no_check6_tx;
2122                 break;
2123 
2124         case UDP_NO_CHECK6_RX:
2125                 val = up->no_check6_rx;
2126                 break;
2127 
2128         /* The following two cannot be changed on UDP sockets, the return is
2129          * always 0 (which corresponds to the full checksum coverage of UDP). */
2130         case UDPLITE_SEND_CSCOV:
2131                 val = up->pcslen;
2132                 break;
2133 
2134         case UDPLITE_RECV_CSCOV:
2135                 val = up->pcrlen;
2136                 break;
2137 
2138         default:
2139                 return -ENOPROTOOPT;
2140         }
2141 
2142         if (put_user(len, optlen))
2143                 return -EFAULT;
2144         if (copy_to_user(optval, &val, len))
2145                 return -EFAULT;
2146         return 0;
2147 }
2148 EXPORT_SYMBOL(udp_lib_getsockopt);
2149 
2150 int udp_getsockopt(struct sock *sk, int level, int optname,
2151                    char __user *optval, int __user *optlen)
2152 {
2153         if (level == SOL_UDP  ||  level == SOL_UDPLITE)
2154                 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2155         return ip_getsockopt(sk, level, optname, optval, optlen);
2156 }
2157 
2158 #ifdef CONFIG_COMPAT
2159 int compat_udp_getsockopt(struct sock *sk, int level, int optname,
2160                                  char __user *optval, int __user *optlen)
2161 {
2162         if (level == SOL_UDP  ||  level == SOL_UDPLITE)
2163                 return udp_lib_getsockopt(sk, level, optname, optval, optlen);
2164         return compat_ip_getsockopt(sk, level, optname, optval, optlen);
2165 }
2166 #endif
2167 /**
2168  *      udp_poll - wait for a UDP event.
2169  *      @file - file struct
2170  *      @sock - socket
2171  *      @wait - poll table
2172  *
2173  *      This is same as datagram poll, except for the special case of
2174  *      blocking sockets. If application is using a blocking fd
2175  *      and a packet with checksum error is in the queue;
2176  *      then it could get return from select indicating data available
2177  *      but then block when reading it. Add special case code
2178  *      to work around these arguably broken applications.
2179  */
2180 unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
2181 {
2182         unsigned int mask = datagram_poll(file, sock, wait);
2183         struct sock *sk = sock->sk;
2184 
2185         sock_rps_record_flow(sk);
2186 
2187         /* Check for false positives due to checksum errors */
2188         if ((mask & POLLRDNORM) && !(file->f_flags & O_NONBLOCK) &&
2189             !(sk->sk_shutdown & RCV_SHUTDOWN) && first_packet_length(sk) == -1)
2190                 mask &= ~(POLLIN | POLLRDNORM);
2191 
2192         return mask;
2193 
2194 }
2195 EXPORT_SYMBOL(udp_poll);
2196 
2197 struct proto udp_prot = {
2198         .name              = "UDP",
2199         .owner             = THIS_MODULE,
2200         .close             = udp_lib_close,
2201         .connect           = ip4_datagram_connect,
2202         .disconnect        = udp_disconnect,
2203         .ioctl             = udp_ioctl,
2204         .destroy           = udp_destroy_sock,
2205         .setsockopt        = udp_setsockopt,
2206         .getsockopt        = udp_getsockopt,
2207         .sendmsg           = udp_sendmsg,
2208         .recvmsg           = udp_recvmsg,
2209         .sendpage          = udp_sendpage,
2210         .backlog_rcv       = __udp_queue_rcv_skb,
2211         .release_cb        = ip4_datagram_release_cb,
2212         .hash              = udp_lib_hash,
2213         .unhash            = udp_lib_unhash,
2214         .rehash            = udp_v4_rehash,
2215         .get_port          = udp_v4_get_port,
2216         .memory_allocated  = &udp_memory_allocated,
2217         .sysctl_mem        = sysctl_udp_mem,
2218         .sysctl_wmem       = &sysctl_udp_wmem_min,
2219         .sysctl_rmem       = &sysctl_udp_rmem_min,
2220         .obj_size          = sizeof(struct udp_sock),
2221         .h.udp_table       = &udp_table,
2222 #ifdef CONFIG_COMPAT
2223         .compat_setsockopt = compat_udp_setsockopt,
2224         .compat_getsockopt = compat_udp_getsockopt,
2225 #endif
2226         .clear_sk          = sk_prot_clear_portaddr_nulls,
2227 };
2228 EXPORT_SYMBOL(udp_prot);
2229 
2230 /* ------------------------------------------------------------------------ */
2231 #ifdef CONFIG_PROC_FS
2232 
2233 static struct sock *udp_get_first(struct seq_file *seq, int start)
2234 {
2235         struct sock *sk;
2236         struct udp_iter_state *state = seq->private;
2237         struct net *net = seq_file_net(seq);
2238 
2239         for (state->bucket = start; state->bucket <= state->udp_table->mask;
2240              ++state->bucket) {
2241                 struct udp_hslot *hslot = &state->udp_table->hash[state->bucket];
2242 
2243                 if (hlist_empty(&hslot->head))
2244                         continue;
2245 
2246                 spin_lock_bh(&hslot->lock);
2247                 sk_for_each(sk, &hslot->head) {
2248                         if (!net_eq(sock_net(sk), net))
2249                                 continue;
2250                         if (sk->sk_family == state->family)
2251                                 goto found;
2252                 }
2253                 spin_unlock_bh(&hslot->lock);
2254         }
2255         sk = NULL;
2256 found:
2257         return sk;
2258 }
2259 
2260 static struct sock *udp_get_next(struct seq_file *seq, struct sock *sk)
2261 {
2262         struct udp_iter_state *state = seq->private;
2263         struct net *net = seq_file_net(seq);
2264 
2265         do {
2266                 sk = sk_next(sk);
2267         } while (sk && (!net_eq(sock_net(sk), net) || sk->sk_family != state->family));
2268 
2269         if (!sk) {
2270                 if (state->bucket <= state->udp_table->mask)
2271                         spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
2272                 return udp_get_first(seq, state->bucket + 1);
2273         }
2274         return sk;
2275 }
2276 
2277 static struct sock *udp_get_idx(struct seq_file *seq, loff_t pos)
2278 {
2279         struct sock *sk = udp_get_first(seq, 0);
2280 
2281         if (sk)
2282                 while (pos && (sk = udp_get_next(seq, sk)) != NULL)
2283                         --pos;
2284         return pos ? NULL : sk;
2285 }
2286 
2287 static void *udp_seq_start(struct seq_file *seq, loff_t *pos)
2288 {
2289         struct udp_iter_state *state = seq->private;
2290         state->bucket = MAX_UDP_PORTS;
2291 
2292         return *pos ? udp_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
2293 }
2294 
2295 static void *udp_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2296 {
2297         struct sock *sk;
2298 
2299         if (v == SEQ_START_TOKEN)
2300                 sk = udp_get_idx(seq, 0);
2301         else
2302                 sk = udp_get_next(seq, v);
2303 
2304         ++*pos;
2305         return sk;
2306 }
2307 
2308 static void udp_seq_stop(struct seq_file *seq, void *v)
2309 {
2310         struct udp_iter_state *state = seq->private;
2311 
2312         if (state->bucket <= state->udp_table->mask)
2313                 spin_unlock_bh(&state->udp_table->hash[state->bucket].lock);
2314 }
2315 
2316 int udp_seq_open(struct inode *inode, struct file *file)
2317 {
2318         struct udp_seq_afinfo *afinfo = PDE_DATA(inode);
2319         struct udp_iter_state *s;
2320         int err;
2321 
2322         err = seq_open_net(inode, file, &afinfo->seq_ops,
2323                            sizeof(struct udp_iter_state));
2324         if (err < 0)
2325                 return err;
2326 
2327         s = ((struct seq_file *)file->private_data)->private;
2328         s->family               = afinfo->family;
2329         s->udp_table            = afinfo->udp_table;
2330         return err;
2331 }
2332 EXPORT_SYMBOL(udp_seq_open);
2333 
2334 /* ------------------------------------------------------------------------ */
2335 int udp_proc_register(struct net *net, struct udp_seq_afinfo *afinfo)
2336 {
2337         struct proc_dir_entry *p;
2338         int rc = 0;
2339 
2340         afinfo->seq_ops.start           = udp_seq_start;
2341         afinfo->seq_ops.next            = udp_seq_next;
2342         afinfo->seq_ops.stop            = udp_seq_stop;
2343 
2344         p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
2345                              afinfo->seq_fops, afinfo);
2346         if (!p)
2347                 rc = -ENOMEM;
2348         return rc;
2349 }
2350 EXPORT_SYMBOL(udp_proc_register);
2351 
2352 void udp_proc_unregister(struct net *net, struct udp_seq_afinfo *afinfo)
2353 {
2354         remove_proc_entry(afinfo->name, net->proc_net);
2355 }
2356 EXPORT_SYMBOL(udp_proc_unregister);
2357 
2358 /* ------------------------------------------------------------------------ */
2359 static void udp4_format_sock(struct sock *sp, struct seq_file *f,
2360                 int bucket)
2361 {
2362         struct inet_sock *inet = inet_sk(sp);
2363         __be32 dest = inet->inet_daddr;
2364         __be32 src  = inet->inet_rcv_saddr;
2365         __u16 destp       = ntohs(inet->inet_dport);
2366         __u16 srcp        = ntohs(inet->inet_sport);
2367 
2368         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
2369                 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
2370                 bucket, src, srcp, dest, destp, sp->sk_state,
2371                 sk_wmem_alloc_get(sp),
2372                 sk_rmem_alloc_get(sp),
2373                 0, 0L, 0,
2374                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
2375                 0, sock_i_ino(sp),
2376                 atomic_read(&sp->sk_refcnt), sp,
2377                 atomic_read(&sp->sk_drops));
2378 }
2379 
2380 int udp4_seq_show(struct seq_file *seq, void *v)
2381 {
2382         seq_setwidth(seq, 127);
2383         if (v == SEQ_START_TOKEN)
2384                 seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
2385                            "rx_queue tr tm->when retrnsmt   uid  timeout "
2386                            "inode ref pointer drops");
2387         else {
2388                 struct udp_iter_state *state = seq->private;
2389 
2390                 udp4_format_sock(v, seq, state->bucket);
2391         }
2392         seq_pad(seq, '\n');
2393         return 0;
2394 }
2395 
2396 static const struct file_operations udp_afinfo_seq_fops = {
2397         .owner    = THIS_MODULE,
2398         .open     = udp_seq_open,
2399         .read     = seq_read,
2400         .llseek   = seq_lseek,
2401         .release  = seq_release_net
2402 };
2403 
2404 /* ------------------------------------------------------------------------ */
2405 static struct udp_seq_afinfo udp4_seq_afinfo = {
2406         .name           = "udp",
2407         .family         = AF_INET,
2408         .udp_table      = &udp_table,
2409         .seq_fops       = &udp_afinfo_seq_fops,
2410         .seq_ops        = {
2411                 .show           = udp4_seq_show,
2412         },
2413 };
2414 
2415 static int __net_init udp4_proc_init_net(struct net *net)
2416 {
2417         return udp_proc_register(net, &udp4_seq_afinfo);
2418 }
2419 
2420 static void __net_exit udp4_proc_exit_net(struct net *net)
2421 {
2422         udp_proc_unregister(net, &udp4_seq_afinfo);
2423 }
2424 
2425 static struct pernet_operations udp4_net_ops = {
2426         .init = udp4_proc_init_net,
2427         .exit = udp4_proc_exit_net,
2428 };
2429 
2430 int __init udp4_proc_init(void)
2431 {
2432         return register_pernet_subsys(&udp4_net_ops);
2433 }
2434 
2435 void udp4_proc_exit(void)
2436 {
2437         unregister_pernet_subsys(&udp4_net_ops);
2438 }
2439 #endif /* CONFIG_PROC_FS */
2440 
2441 static __initdata unsigned long uhash_entries;
2442 static int __init set_uhash_entries(char *str)
2443 {
2444         ssize_t ret;
2445 
2446         if (!str)
2447                 return 0;
2448 
2449         ret = kstrtoul(str, 0, &uhash_entries);
2450         if (ret)
2451                 return 0;
2452 
2453         if (uhash_entries && uhash_entries < UDP_HTABLE_SIZE_MIN)
2454                 uhash_entries = UDP_HTABLE_SIZE_MIN;
2455         return 1;
2456 }
2457 __setup("uhash_entries=", set_uhash_entries);
2458 
2459 void __init udp_table_init(struct udp_table *table, const char *name)
2460 {
2461         unsigned int i;
2462 
2463         table->hash = alloc_large_system_hash(name,
2464                                               2 * sizeof(struct udp_hslot),
2465                                               uhash_entries,
2466                                               21, /* one slot per 2 MB */
2467                                               0,
2468                                               &table->log,
2469                                               &table->mask,
2470                                               UDP_HTABLE_SIZE_MIN,
2471                                               64 * 1024);
2472 
2473         table->hash2 = table->hash + (table->mask + 1);
2474         for (i = 0; i <= table->mask; i++) {
2475                 INIT_HLIST_HEAD(&table->hash[i].head);
2476                 table->hash[i].count = 0;
2477                 spin_lock_init(&table->hash[i].lock);
2478         }
2479         for (i = 0; i <= table->mask; i++) {
2480                 INIT_HLIST_HEAD(&table->hash2[i].head);
2481                 table->hash2[i].count = 0;
2482                 spin_lock_init(&table->hash2[i].lock);
2483         }
2484 }
2485 
2486 u32 udp_flow_hashrnd(void)
2487 {
2488         static u32 hashrnd __read_mostly;
2489 
2490         net_get_random_once(&hashrnd, sizeof(hashrnd));
2491 
2492         return hashrnd;
2493 }
2494 EXPORT_SYMBOL(udp_flow_hashrnd);
2495 
2496 void __init udp_init(void)
2497 {
2498         unsigned long limit;
2499 
2500         udp_table_init(&udp_table, "UDP");
2501         limit = nr_free_buffer_pages() / 8;
2502         limit = max(limit, 128UL);
2503         sysctl_udp_mem[0] = limit / 4 * 3;
2504         sysctl_udp_mem[1] = limit;
2505         sysctl_udp_mem[2] = sysctl_udp_mem[0] * 2;
2506 
2507         sysctl_udp_rmem_min = SK_MEM_QUANTUM;
2508         sysctl_udp_wmem_min = SK_MEM_QUANTUM;
2509 }
2510 

~ [ 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