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

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

Version: ~ [ linux-5.10-rc1 ] ~ [ linux-5.9.1 ] ~ [ linux-5.8.16 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.72 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.152 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.202 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.240 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.240 ] ~ [ 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  *  net/dccp/ipv4.c
  3  *
  4  *  An implementation of the DCCP protocol
  5  *  Arnaldo Carvalho de Melo <acme@conectiva.com.br>
  6  *
  7  *      This program is free software; you can redistribute it and/or
  8  *      modify it under the terms of the GNU General Public License
  9  *      as published by the Free Software Foundation; either version
 10  *      2 of the License, or (at your option) any later version.
 11  */
 12 
 13 #include <linux/dccp.h>
 14 #include <linux/icmp.h>
 15 #include <linux/slab.h>
 16 #include <linux/module.h>
 17 #include <linux/skbuff.h>
 18 #include <linux/random.h>
 19 
 20 #include <net/icmp.h>
 21 #include <net/inet_common.h>
 22 #include <net/inet_hashtables.h>
 23 #include <net/inet_sock.h>
 24 #include <net/protocol.h>
 25 #include <net/sock.h>
 26 #include <net/timewait_sock.h>
 27 #include <net/tcp_states.h>
 28 #include <net/xfrm.h>
 29 #include <net/secure_seq.h>
 30 
 31 #include "ackvec.h"
 32 #include "ccid.h"
 33 #include "dccp.h"
 34 #include "feat.h"
 35 
 36 /*
 37  * The per-net dccp.v4_ctl_sk socket is used for responding to
 38  * the Out-of-the-blue (OOTB) packets. A control sock will be created
 39  * for this socket at the initialization time.
 40  */
 41 
 42 int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 43 {
 44         const struct sockaddr_in *usin = (struct sockaddr_in *)uaddr;
 45         struct inet_sock *inet = inet_sk(sk);
 46         struct dccp_sock *dp = dccp_sk(sk);
 47         __be16 orig_sport, orig_dport;
 48         __be32 daddr, nexthop;
 49         struct flowi4 *fl4;
 50         struct rtable *rt;
 51         int err;
 52         struct ip_options_rcu *inet_opt;
 53 
 54         dp->dccps_role = DCCP_ROLE_CLIENT;
 55 
 56         if (addr_len < sizeof(struct sockaddr_in))
 57                 return -EINVAL;
 58 
 59         if (usin->sin_family != AF_INET)
 60                 return -EAFNOSUPPORT;
 61 
 62         nexthop = daddr = usin->sin_addr.s_addr;
 63 
 64         inet_opt = rcu_dereference_protected(inet->inet_opt,
 65                                              sock_owned_by_user(sk));
 66         if (inet_opt != NULL && inet_opt->opt.srr) {
 67                 if (daddr == 0)
 68                         return -EINVAL;
 69                 nexthop = inet_opt->opt.faddr;
 70         }
 71 
 72         orig_sport = inet->inet_sport;
 73         orig_dport = usin->sin_port;
 74         fl4 = &inet->cork.fl.u.ip4;
 75         rt = ip_route_connect(fl4, nexthop, inet->inet_saddr,
 76                               RT_CONN_FLAGS(sk), sk->sk_bound_dev_if,
 77                               IPPROTO_DCCP,
 78                               orig_sport, orig_dport, sk, true);
 79         if (IS_ERR(rt))
 80                 return PTR_ERR(rt);
 81 
 82         if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) {
 83                 ip_rt_put(rt);
 84                 return -ENETUNREACH;
 85         }
 86 
 87         if (inet_opt == NULL || !inet_opt->opt.srr)
 88                 daddr = fl4->daddr;
 89 
 90         if (inet->inet_saddr == 0)
 91                 inet->inet_saddr = fl4->saddr;
 92         inet->inet_rcv_saddr = inet->inet_saddr;
 93 
 94         inet->inet_dport = usin->sin_port;
 95         inet->inet_daddr = daddr;
 96 
 97         inet_csk(sk)->icsk_ext_hdr_len = 0;
 98         if (inet_opt)
 99                 inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
100         /*
101          * Socket identity is still unknown (sport may be zero).
102          * However we set state to DCCP_REQUESTING and not releasing socket
103          * lock select source port, enter ourselves into the hash tables and
104          * complete initialization after this.
105          */
106         dccp_set_state(sk, DCCP_REQUESTING);
107         err = inet_hash_connect(&dccp_death_row, sk);
108         if (err != 0)
109                 goto failure;
110 
111         rt = ip_route_newports(fl4, rt, orig_sport, orig_dport,
112                                inet->inet_sport, inet->inet_dport, sk);
113         if (IS_ERR(rt)) {
114                 err = PTR_ERR(rt);
115                 rt = NULL;
116                 goto failure;
117         }
118         /* OK, now commit destination to socket.  */
119         sk_setup_caps(sk, &rt->dst);
120 
121         dp->dccps_iss = secure_dccp_sequence_number(inet->inet_saddr,
122                                                     inet->inet_daddr,
123                                                     inet->inet_sport,
124                                                     inet->inet_dport);
125         inet->inet_id = dp->dccps_iss ^ jiffies;
126 
127         err = dccp_connect(sk);
128         rt = NULL;
129         if (err != 0)
130                 goto failure;
131 out:
132         return err;
133 failure:
134         /*
135          * This unhashes the socket and releases the local port, if necessary.
136          */
137         dccp_set_state(sk, DCCP_CLOSED);
138         ip_rt_put(rt);
139         sk->sk_route_caps = 0;
140         inet->inet_dport = 0;
141         goto out;
142 }
143 
144 EXPORT_SYMBOL_GPL(dccp_v4_connect);
145 
146 /*
147  * This routine does path mtu discovery as defined in RFC1191.
148  */
149 static inline void dccp_do_pmtu_discovery(struct sock *sk,
150                                           const struct iphdr *iph,
151                                           u32 mtu)
152 {
153         struct dst_entry *dst;
154         const struct inet_sock *inet = inet_sk(sk);
155         const struct dccp_sock *dp = dccp_sk(sk);
156 
157         /* We are not interested in DCCP_LISTEN and request_socks (RESPONSEs
158          * send out by Linux are always < 576bytes so they should go through
159          * unfragmented).
160          */
161         if (sk->sk_state == DCCP_LISTEN)
162                 return;
163 
164         dst = inet_csk_update_pmtu(sk, mtu);
165         if (!dst)
166                 return;
167 
168         /* Something is about to be wrong... Remember soft error
169          * for the case, if this connection will not able to recover.
170          */
171         if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst))
172                 sk->sk_err_soft = EMSGSIZE;
173 
174         mtu = dst_mtu(dst);
175 
176         if (inet->pmtudisc != IP_PMTUDISC_DONT &&
177             inet_csk(sk)->icsk_pmtu_cookie > mtu) {
178                 dccp_sync_mss(sk, mtu);
179 
180                 /*
181                  * From RFC 4340, sec. 14.1:
182                  *
183                  *      DCCP-Sync packets are the best choice for upward
184                  *      probing, since DCCP-Sync probes do not risk application
185                  *      data loss.
186                  */
187                 dccp_send_sync(sk, dp->dccps_gsr, DCCP_PKT_SYNC);
188         } /* else let the usual retransmit timer handle it */
189 }
190 
191 static void dccp_do_redirect(struct sk_buff *skb, struct sock *sk)
192 {
193         struct dst_entry *dst = __sk_dst_check(sk, 0);
194 
195         if (dst)
196                 dst->ops->redirect(dst, sk, skb);
197 }
198 
199 /*
200  * This routine is called by the ICMP module when it gets some sort of error
201  * condition. If err < 0 then the socket should be closed and the error
202  * returned to the user. If err > 0 it's just the icmp type << 8 | icmp code.
203  * After adjustment header points to the first 8 bytes of the tcp header. We
204  * need to find the appropriate port.
205  *
206  * The locking strategy used here is very "optimistic". When someone else
207  * accesses the socket the ICMP is just dropped and for some paths there is no
208  * check at all. A more general error queue to queue errors for later handling
209  * is probably better.
210  */
211 static void dccp_v4_err(struct sk_buff *skb, u32 info)
212 {
213         const struct iphdr *iph = (struct iphdr *)skb->data;
214         const u8 offset = iph->ihl << 2;
215         const struct dccp_hdr *dh;
216         struct dccp_sock *dp;
217         struct inet_sock *inet;
218         const int type = icmp_hdr(skb)->type;
219         const int code = icmp_hdr(skb)->code;
220         struct sock *sk;
221         __u64 seq;
222         int err;
223         struct net *net = dev_net(skb->dev);
224 
225         /* Only need dccph_dport & dccph_sport which are the first
226          * 4 bytes in dccp header.
227          * Our caller (icmp_socket_deliver()) already pulled 8 bytes for us.
228          */
229         BUILD_BUG_ON(offsetofend(struct dccp_hdr, dccph_sport) > 8);
230         BUILD_BUG_ON(offsetofend(struct dccp_hdr, dccph_dport) > 8);
231         dh = (struct dccp_hdr *)(skb->data + offset);
232 
233         sk = inet_lookup(net, &dccp_hashinfo,
234                         iph->daddr, dh->dccph_dport,
235                         iph->saddr, dh->dccph_sport, inet_iif(skb));
236         if (sk == NULL) {
237                 ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
238                 return;
239         }
240 
241         if (sk->sk_state == DCCP_TIME_WAIT) {
242                 inet_twsk_put(inet_twsk(sk));
243                 return;
244         }
245 
246         bh_lock_sock(sk);
247         /* If too many ICMPs get dropped on busy
248          * servers this needs to be solved differently.
249          */
250         if (sock_owned_by_user(sk))
251                 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
252 
253         if (sk->sk_state == DCCP_CLOSED)
254                 goto out;
255 
256         dp = dccp_sk(sk);
257         seq = dccp_hdr_seq(dh);
258         if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
259             !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
260                 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
261                 goto out;
262         }
263 
264         switch (type) {
265         case ICMP_REDIRECT:
266                 if (!sock_owned_by_user(sk))
267                         dccp_do_redirect(skb, sk);
268                 goto out;
269         case ICMP_SOURCE_QUENCH:
270                 /* Just silently ignore these. */
271                 goto out;
272         case ICMP_PARAMETERPROB:
273                 err = EPROTO;
274                 break;
275         case ICMP_DEST_UNREACH:
276                 if (code > NR_ICMP_UNREACH)
277                         goto out;
278 
279                 if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */
280                         if (!sock_owned_by_user(sk))
281                                 dccp_do_pmtu_discovery(sk, iph, info);
282                         goto out;
283                 }
284 
285                 err = icmp_err_convert[code].errno;
286                 break;
287         case ICMP_TIME_EXCEEDED:
288                 err = EHOSTUNREACH;
289                 break;
290         default:
291                 goto out;
292         }
293 
294         switch (sk->sk_state) {
295                 struct request_sock *req , **prev;
296         case DCCP_LISTEN:
297                 if (sock_owned_by_user(sk))
298                         goto out;
299                 req = inet_csk_search_req(sk, &prev, dh->dccph_dport,
300                                           iph->daddr, iph->saddr);
301                 if (!req)
302                         goto out;
303 
304                 /*
305                  * ICMPs are not backlogged, hence we cannot get an established
306                  * socket here.
307                  */
308                 WARN_ON(req->sk);
309 
310                 if (!between48(seq, dccp_rsk(req)->dreq_iss,
311                                     dccp_rsk(req)->dreq_gss)) {
312                         NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
313                         goto out;
314                 }
315                 /*
316                  * Still in RESPOND, just remove it silently.
317                  * There is no good way to pass the error to the newly
318                  * created socket, and POSIX does not want network
319                  * errors returned from accept().
320                  */
321                 inet_csk_reqsk_queue_drop(sk, req, prev);
322                 goto out;
323 
324         case DCCP_REQUESTING:
325         case DCCP_RESPOND:
326                 if (!sock_owned_by_user(sk)) {
327                         DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
328                         sk->sk_err = err;
329 
330                         sk->sk_error_report(sk);
331 
332                         dccp_done(sk);
333                 } else
334                         sk->sk_err_soft = err;
335                 goto out;
336         }
337 
338         /* If we've already connected we will keep trying
339          * until we time out, or the user gives up.
340          *
341          * rfc1122 4.2.3.9 allows to consider as hard errors
342          * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too,
343          * but it is obsoleted by pmtu discovery).
344          *
345          * Note, that in modern internet, where routing is unreliable
346          * and in each dark corner broken firewalls sit, sending random
347          * errors ordered by their masters even this two messages finally lose
348          * their original sense (even Linux sends invalid PORT_UNREACHs)
349          *
350          * Now we are in compliance with RFCs.
351          *                                                      --ANK (980905)
352          */
353 
354         inet = inet_sk(sk);
355         if (!sock_owned_by_user(sk) && inet->recverr) {
356                 sk->sk_err = err;
357                 sk->sk_error_report(sk);
358         } else /* Only an error on timeout */
359                 sk->sk_err_soft = err;
360 out:
361         bh_unlock_sock(sk);
362         sock_put(sk);
363 }
364 
365 static inline __sum16 dccp_v4_csum_finish(struct sk_buff *skb,
366                                       __be32 src, __be32 dst)
367 {
368         return csum_tcpudp_magic(src, dst, skb->len, IPPROTO_DCCP, skb->csum);
369 }
370 
371 void dccp_v4_send_check(struct sock *sk, struct sk_buff *skb)
372 {
373         const struct inet_sock *inet = inet_sk(sk);
374         struct dccp_hdr *dh = dccp_hdr(skb);
375 
376         dccp_csum_outgoing(skb);
377         dh->dccph_checksum = dccp_v4_csum_finish(skb,
378                                                  inet->inet_saddr,
379                                                  inet->inet_daddr);
380 }
381 
382 EXPORT_SYMBOL_GPL(dccp_v4_send_check);
383 
384 static inline u64 dccp_v4_init_sequence(const struct sk_buff *skb)
385 {
386         return secure_dccp_sequence_number(ip_hdr(skb)->daddr,
387                                            ip_hdr(skb)->saddr,
388                                            dccp_hdr(skb)->dccph_dport,
389                                            dccp_hdr(skb)->dccph_sport);
390 }
391 
392 /*
393  * The three way handshake has completed - we got a valid ACK or DATAACK -
394  * now create the new socket.
395  *
396  * This is the equivalent of TCP's tcp_v4_syn_recv_sock
397  */
398 struct sock *dccp_v4_request_recv_sock(struct sock *sk, struct sk_buff *skb,
399                                        struct request_sock *req,
400                                        struct dst_entry *dst)
401 {
402         struct inet_request_sock *ireq;
403         struct inet_sock *newinet;
404         struct sock *newsk;
405 
406         if (sk_acceptq_is_full(sk))
407                 goto exit_overflow;
408 
409         newsk = dccp_create_openreq_child(sk, req, skb);
410         if (newsk == NULL)
411                 goto exit_nonewsk;
412 
413         newinet            = inet_sk(newsk);
414         ireq               = inet_rsk(req);
415         newinet->inet_daddr     = ireq->rmt_addr;
416         newinet->inet_rcv_saddr = ireq->loc_addr;
417         newinet->inet_saddr     = ireq->loc_addr;
418         newinet->inet_opt       = ireq->opt;
419         ireq->opt          = NULL;
420         newinet->mc_index  = inet_iif(skb);
421         newinet->mc_ttl    = ip_hdr(skb)->ttl;
422         newinet->inet_id   = jiffies;
423 
424         if (dst == NULL && (dst = inet_csk_route_child_sock(sk, newsk, req)) == NULL)
425                 goto put_and_exit;
426 
427         sk_setup_caps(newsk, dst);
428 
429         dccp_sync_mss(newsk, dst_mtu(dst));
430 
431         if (__inet_inherit_port(sk, newsk) < 0)
432                 goto put_and_exit;
433         __inet_hash_nolisten(newsk, NULL);
434 
435         return newsk;
436 
437 exit_overflow:
438         NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
439 exit_nonewsk:
440         dst_release(dst);
441 exit:
442         NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
443         return NULL;
444 put_and_exit:
445         inet_csk_prepare_forced_close(newsk);
446         dccp_done(newsk);
447         goto exit;
448 }
449 
450 EXPORT_SYMBOL_GPL(dccp_v4_request_recv_sock);
451 
452 static struct sock *dccp_v4_hnd_req(struct sock *sk, struct sk_buff *skb)
453 {
454         const struct dccp_hdr *dh = dccp_hdr(skb);
455         const struct iphdr *iph = ip_hdr(skb);
456         struct sock *nsk;
457         struct request_sock **prev;
458         /* Find possible connection requests. */
459         struct request_sock *req = inet_csk_search_req(sk, &prev,
460                                                        dh->dccph_sport,
461                                                        iph->saddr, iph->daddr);
462         if (req != NULL)
463                 return dccp_check_req(sk, skb, req, prev);
464 
465         nsk = inet_lookup_established(sock_net(sk), &dccp_hashinfo,
466                                       iph->saddr, dh->dccph_sport,
467                                       iph->daddr, dh->dccph_dport,
468                                       inet_iif(skb));
469         if (nsk != NULL) {
470                 if (nsk->sk_state != DCCP_TIME_WAIT) {
471                         bh_lock_sock(nsk);
472                         return nsk;
473                 }
474                 inet_twsk_put(inet_twsk(nsk));
475                 return NULL;
476         }
477 
478         return sk;
479 }
480 
481 static struct dst_entry* dccp_v4_route_skb(struct net *net, struct sock *sk,
482                                            struct sk_buff *skb)
483 {
484         struct rtable *rt;
485         const struct iphdr *iph = ip_hdr(skb);
486         struct flowi4 fl4 = {
487                 .flowi4_oif = inet_iif(skb),
488                 .daddr = iph->saddr,
489                 .saddr = iph->daddr,
490                 .flowi4_tos = RT_CONN_FLAGS(sk),
491                 .flowi4_proto = sk->sk_protocol,
492                 .fl4_sport = dccp_hdr(skb)->dccph_dport,
493                 .fl4_dport = dccp_hdr(skb)->dccph_sport,
494         };
495 
496         security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
497         rt = ip_route_output_flow(net, &fl4, sk);
498         if (IS_ERR(rt)) {
499                 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
500                 return NULL;
501         }
502 
503         return &rt->dst;
504 }
505 
506 static int dccp_v4_send_response(struct sock *sk, struct request_sock *req)
507 {
508         int err = -1;
509         struct sk_buff *skb;
510         struct dst_entry *dst;
511         struct flowi4 fl4;
512 
513         dst = inet_csk_route_req(sk, &fl4, req);
514         if (dst == NULL)
515                 goto out;
516 
517         skb = dccp_make_response(sk, dst, req);
518         if (skb != NULL) {
519                 const struct inet_request_sock *ireq = inet_rsk(req);
520                 struct dccp_hdr *dh = dccp_hdr(skb);
521 
522                 dh->dccph_checksum = dccp_v4_csum_finish(skb, ireq->loc_addr,
523                                                               ireq->rmt_addr);
524                 err = ip_build_and_send_pkt(skb, sk, ireq->loc_addr,
525                                             ireq->rmt_addr,
526                                             ireq->opt);
527                 err = net_xmit_eval(err);
528         }
529 
530 out:
531         dst_release(dst);
532         return err;
533 }
534 
535 static void dccp_v4_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb)
536 {
537         int err;
538         const struct iphdr *rxiph;
539         struct sk_buff *skb;
540         struct dst_entry *dst;
541         struct net *net = dev_net(skb_dst(rxskb)->dev);
542         struct sock *ctl_sk = net->dccp.v4_ctl_sk;
543 
544         /* Never send a reset in response to a reset. */
545         if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
546                 return;
547 
548         if (skb_rtable(rxskb)->rt_type != RTN_LOCAL)
549                 return;
550 
551         dst = dccp_v4_route_skb(net, ctl_sk, rxskb);
552         if (dst == NULL)
553                 return;
554 
555         skb = dccp_ctl_make_reset(ctl_sk, rxskb);
556         if (skb == NULL)
557                 goto out;
558 
559         rxiph = ip_hdr(rxskb);
560         dccp_hdr(skb)->dccph_checksum = dccp_v4_csum_finish(skb, rxiph->saddr,
561                                                                  rxiph->daddr);
562         skb_dst_set(skb, dst_clone(dst));
563 
564         bh_lock_sock(ctl_sk);
565         err = ip_build_and_send_pkt(skb, ctl_sk,
566                                     rxiph->daddr, rxiph->saddr, NULL);
567         bh_unlock_sock(ctl_sk);
568 
569         if (net_xmit_eval(err) == 0) {
570                 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
571                 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
572         }
573 out:
574          dst_release(dst);
575 }
576 
577 static void dccp_v4_reqsk_destructor(struct request_sock *req)
578 {
579         dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
580         kfree(inet_rsk(req)->opt);
581 }
582 
583 void dccp_syn_ack_timeout(struct sock *sk, struct request_sock *req)
584 {
585 }
586 EXPORT_SYMBOL(dccp_syn_ack_timeout);
587 
588 static struct request_sock_ops dccp_request_sock_ops __read_mostly = {
589         .family         = PF_INET,
590         .obj_size       = sizeof(struct dccp_request_sock),
591         .rtx_syn_ack    = dccp_v4_send_response,
592         .send_ack       = dccp_reqsk_send_ack,
593         .destructor     = dccp_v4_reqsk_destructor,
594         .send_reset     = dccp_v4_ctl_send_reset,
595         .syn_ack_timeout = dccp_syn_ack_timeout,
596 };
597 
598 int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
599 {
600         struct inet_request_sock *ireq;
601         struct request_sock *req;
602         struct dccp_request_sock *dreq;
603         const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
604         struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
605 
606         /* Never answer to DCCP_PKT_REQUESTs send to broadcast or multicast */
607         if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
608                 return 0;       /* discard, don't send a reset here */
609 
610         if (dccp_bad_service_code(sk, service)) {
611                 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
612                 goto drop;
613         }
614         /*
615          * TW buckets are converted to open requests without
616          * limitations, they conserve resources and peer is
617          * evidently real one.
618          */
619         dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
620         if (inet_csk_reqsk_queue_is_full(sk))
621                 goto drop;
622 
623         /*
624          * Accept backlog is full. If we have already queued enough
625          * of warm entries in syn queue, drop request. It is better than
626          * clogging syn queue with openreqs with exponentially increasing
627          * timeout.
628          */
629         if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
630                 goto drop;
631 
632         req = inet_reqsk_alloc(&dccp_request_sock_ops);
633         if (req == NULL)
634                 goto drop;
635 
636         if (dccp_reqsk_init(req, dccp_sk(sk), skb))
637                 goto drop_and_free;
638 
639         dreq = dccp_rsk(req);
640         if (dccp_parse_options(sk, dreq, skb))
641                 goto drop_and_free;
642 
643         if (security_inet_conn_request(sk, skb, req))
644                 goto drop_and_free;
645 
646         ireq = inet_rsk(req);
647         ireq->loc_addr = ip_hdr(skb)->daddr;
648         ireq->rmt_addr = ip_hdr(skb)->saddr;
649 
650         /*
651          * Step 3: Process LISTEN state
652          *
653          * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
654          *
655          * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
656          */
657         dreq->dreq_isr     = dcb->dccpd_seq;
658         dreq->dreq_gsr     = dreq->dreq_isr;
659         dreq->dreq_iss     = dccp_v4_init_sequence(skb);
660         dreq->dreq_gss     = dreq->dreq_iss;
661         dreq->dreq_service = service;
662 
663         if (dccp_v4_send_response(sk, req))
664                 goto drop_and_free;
665 
666         inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
667         return 0;
668 
669 drop_and_free:
670         reqsk_free(req);
671 drop:
672         DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
673         return -1;
674 }
675 
676 EXPORT_SYMBOL_GPL(dccp_v4_conn_request);
677 
678 int dccp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
679 {
680         struct dccp_hdr *dh = dccp_hdr(skb);
681 
682         if (sk->sk_state == DCCP_OPEN) { /* Fast path */
683                 if (dccp_rcv_established(sk, skb, dh, skb->len))
684                         goto reset;
685                 return 0;
686         }
687 
688         /*
689          *  Step 3: Process LISTEN state
690          *       If P.type == Request or P contains a valid Init Cookie option,
691          *            (* Must scan the packet's options to check for Init
692          *               Cookies.  Only Init Cookies are processed here,
693          *               however; other options are processed in Step 8.  This
694          *               scan need only be performed if the endpoint uses Init
695          *               Cookies *)
696          *            (* Generate a new socket and switch to that socket *)
697          *            Set S := new socket for this port pair
698          *            S.state = RESPOND
699          *            Choose S.ISS (initial seqno) or set from Init Cookies
700          *            Initialize S.GAR := S.ISS
701          *            Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
702          *            Continue with S.state == RESPOND
703          *            (* A Response packet will be generated in Step 11 *)
704          *       Otherwise,
705          *            Generate Reset(No Connection) unless P.type == Reset
706          *            Drop packet and return
707          *
708          * NOTE: the check for the packet types is done in
709          *       dccp_rcv_state_process
710          */
711         if (sk->sk_state == DCCP_LISTEN) {
712                 struct sock *nsk = dccp_v4_hnd_req(sk, skb);
713 
714                 if (nsk == NULL)
715                         goto discard;
716 
717                 if (nsk != sk) {
718                         if (dccp_child_process(sk, nsk, skb))
719                                 goto reset;
720                         return 0;
721                 }
722         }
723 
724         if (dccp_rcv_state_process(sk, skb, dh, skb->len))
725                 goto reset;
726         return 0;
727 
728 reset:
729         dccp_v4_ctl_send_reset(sk, skb);
730 discard:
731         kfree_skb(skb);
732         return 0;
733 }
734 
735 EXPORT_SYMBOL_GPL(dccp_v4_do_rcv);
736 
737 /**
738  *      dccp_invalid_packet  -  check for malformed packets
739  *      Implements RFC 4340, 8.5:  Step 1: Check header basics
740  *      Packets that fail these checks are ignored and do not receive Resets.
741  */
742 int dccp_invalid_packet(struct sk_buff *skb)
743 {
744         const struct dccp_hdr *dh;
745         unsigned int cscov;
746 
747         if (skb->pkt_type != PACKET_HOST)
748                 return 1;
749 
750         /* If the packet is shorter than 12 bytes, drop packet and return */
751         if (!pskb_may_pull(skb, sizeof(struct dccp_hdr))) {
752                 DCCP_WARN("pskb_may_pull failed\n");
753                 return 1;
754         }
755 
756         dh = dccp_hdr(skb);
757 
758         /* If P.type is not understood, drop packet and return */
759         if (dh->dccph_type >= DCCP_PKT_INVALID) {
760                 DCCP_WARN("invalid packet type\n");
761                 return 1;
762         }
763 
764         /*
765          * If P.Data Offset is too small for packet type, drop packet and return
766          */
767         if (dh->dccph_doff < dccp_hdr_len(skb) / sizeof(u32)) {
768                 DCCP_WARN("P.Data Offset(%u) too small\n", dh->dccph_doff);
769                 return 1;
770         }
771         /*
772          * If P.Data Offset is too too large for packet, drop packet and return
773          */
774         if (!pskb_may_pull(skb, dh->dccph_doff * sizeof(u32))) {
775                 DCCP_WARN("P.Data Offset(%u) too large\n", dh->dccph_doff);
776                 return 1;
777         }
778 
779         /*
780          * If P.type is not Data, Ack, or DataAck and P.X == 0 (the packet
781          * has short sequence numbers), drop packet and return
782          */
783         if ((dh->dccph_type < DCCP_PKT_DATA    ||
784             dh->dccph_type > DCCP_PKT_DATAACK) && dh->dccph_x == 0)  {
785                 DCCP_WARN("P.type (%s) not Data || [Data]Ack, while P.X == 0\n",
786                           dccp_packet_name(dh->dccph_type));
787                 return 1;
788         }
789 
790         /*
791          * If P.CsCov is too large for the packet size, drop packet and return.
792          * This must come _before_ checksumming (not as RFC 4340 suggests).
793          */
794         cscov = dccp_csum_coverage(skb);
795         if (cscov > skb->len) {
796                 DCCP_WARN("P.CsCov %u exceeds packet length %d\n",
797                           dh->dccph_cscov, skb->len);
798                 return 1;
799         }
800 
801         /* If header checksum is incorrect, drop packet and return.
802          * (This step is completed in the AF-dependent functions.) */
803         skb->csum = skb_checksum(skb, 0, cscov, 0);
804 
805         return 0;
806 }
807 
808 EXPORT_SYMBOL_GPL(dccp_invalid_packet);
809 
810 /* this is called when real data arrives */
811 static int dccp_v4_rcv(struct sk_buff *skb)
812 {
813         const struct dccp_hdr *dh;
814         const struct iphdr *iph;
815         struct sock *sk;
816         int min_cov;
817 
818         /* Step 1: Check header basics */
819 
820         if (dccp_invalid_packet(skb))
821                 goto discard_it;
822 
823         iph = ip_hdr(skb);
824         /* Step 1: If header checksum is incorrect, drop packet and return */
825         if (dccp_v4_csum_finish(skb, iph->saddr, iph->daddr)) {
826                 DCCP_WARN("dropped packet with invalid checksum\n");
827                 goto discard_it;
828         }
829 
830         dh = dccp_hdr(skb);
831 
832         DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(dh);
833         DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
834 
835         dccp_pr_debug("%8.8s src=%pI4@%-5d dst=%pI4@%-5d seq=%llu",
836                       dccp_packet_name(dh->dccph_type),
837                       &iph->saddr, ntohs(dh->dccph_sport),
838                       &iph->daddr, ntohs(dh->dccph_dport),
839                       (unsigned long long) DCCP_SKB_CB(skb)->dccpd_seq);
840 
841         if (dccp_packet_without_ack(skb)) {
842                 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
843                 dccp_pr_debug_cat("\n");
844         } else {
845                 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
846                 dccp_pr_debug_cat(", ack=%llu\n", (unsigned long long)
847                                   DCCP_SKB_CB(skb)->dccpd_ack_seq);
848         }
849 
850         /* Step 2:
851          *      Look up flow ID in table and get corresponding socket */
852         sk = __inet_lookup_skb(&dccp_hashinfo, skb,
853                                dh->dccph_sport, dh->dccph_dport);
854         /*
855          * Step 2:
856          *      If no socket ...
857          */
858         if (sk == NULL) {
859                 dccp_pr_debug("failed to look up flow ID in table and "
860                               "get corresponding socket\n");
861                 goto no_dccp_socket;
862         }
863 
864         /*
865          * Step 2:
866          *      ... or S.state == TIMEWAIT,
867          *              Generate Reset(No Connection) unless P.type == Reset
868          *              Drop packet and return
869          */
870         if (sk->sk_state == DCCP_TIME_WAIT) {
871                 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
872                 inet_twsk_put(inet_twsk(sk));
873                 goto no_dccp_socket;
874         }
875 
876         /*
877          * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
878          *      o if MinCsCov = 0, only packets with CsCov = 0 are accepted
879          *      o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
880          */
881         min_cov = dccp_sk(sk)->dccps_pcrlen;
882         if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov))  {
883                 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
884                               dh->dccph_cscov, min_cov);
885                 /* FIXME: "Such packets SHOULD be reported using Data Dropped
886                  *         options (Section 11.7) with Drop Code 0, Protocol
887                  *         Constraints."                                     */
888                 goto discard_and_relse;
889         }
890 
891         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
892                 goto discard_and_relse;
893         nf_reset(skb);
894 
895         return sk_receive_skb(sk, skb, 1);
896 
897 no_dccp_socket:
898         if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
899                 goto discard_it;
900         /*
901          * Step 2:
902          *      If no socket ...
903          *              Generate Reset(No Connection) unless P.type == Reset
904          *              Drop packet and return
905          */
906         if (dh->dccph_type != DCCP_PKT_RESET) {
907                 DCCP_SKB_CB(skb)->dccpd_reset_code =
908                                         DCCP_RESET_CODE_NO_CONNECTION;
909                 dccp_v4_ctl_send_reset(sk, skb);
910         }
911 
912 discard_it:
913         kfree_skb(skb);
914         return 0;
915 
916 discard_and_relse:
917         sock_put(sk);
918         goto discard_it;
919 }
920 
921 static const struct inet_connection_sock_af_ops dccp_ipv4_af_ops = {
922         .queue_xmit        = ip_queue_xmit,
923         .send_check        = dccp_v4_send_check,
924         .rebuild_header    = inet_sk_rebuild_header,
925         .conn_request      = dccp_v4_conn_request,
926         .syn_recv_sock     = dccp_v4_request_recv_sock,
927         .net_header_len    = sizeof(struct iphdr),
928         .setsockopt        = ip_setsockopt,
929         .getsockopt        = ip_getsockopt,
930         .addr2sockaddr     = inet_csk_addr2sockaddr,
931         .sockaddr_len      = sizeof(struct sockaddr_in),
932         .bind_conflict     = inet_csk_bind_conflict,
933 #ifdef CONFIG_COMPAT
934         .compat_setsockopt = compat_ip_setsockopt,
935         .compat_getsockopt = compat_ip_getsockopt,
936 #endif
937 };
938 
939 static int dccp_v4_init_sock(struct sock *sk)
940 {
941         static __u8 dccp_v4_ctl_sock_initialized;
942         int err = dccp_init_sock(sk, dccp_v4_ctl_sock_initialized);
943 
944         if (err == 0) {
945                 if (unlikely(!dccp_v4_ctl_sock_initialized))
946                         dccp_v4_ctl_sock_initialized = 1;
947                 inet_csk(sk)->icsk_af_ops = &dccp_ipv4_af_ops;
948         }
949 
950         return err;
951 }
952 
953 static struct timewait_sock_ops dccp_timewait_sock_ops = {
954         .twsk_obj_size  = sizeof(struct inet_timewait_sock),
955 };
956 
957 static struct proto dccp_v4_prot = {
958         .name                   = "DCCP",
959         .owner                  = THIS_MODULE,
960         .close                  = dccp_close,
961         .connect                = dccp_v4_connect,
962         .disconnect             = dccp_disconnect,
963         .ioctl                  = dccp_ioctl,
964         .init                   = dccp_v4_init_sock,
965         .setsockopt             = dccp_setsockopt,
966         .getsockopt             = dccp_getsockopt,
967         .sendmsg                = dccp_sendmsg,
968         .recvmsg                = dccp_recvmsg,
969         .backlog_rcv            = dccp_v4_do_rcv,
970         .hash                   = inet_hash,
971         .unhash                 = inet_unhash,
972         .accept                 = inet_csk_accept,
973         .get_port               = inet_csk_get_port,
974         .shutdown               = dccp_shutdown,
975         .destroy                = dccp_destroy_sock,
976         .orphan_count           = &dccp_orphan_count,
977         .max_header             = MAX_DCCP_HEADER,
978         .obj_size               = sizeof(struct dccp_sock),
979         .slab_flags             = SLAB_DESTROY_BY_RCU,
980         .rsk_prot               = &dccp_request_sock_ops,
981         .twsk_prot              = &dccp_timewait_sock_ops,
982         .h.hashinfo             = &dccp_hashinfo,
983 #ifdef CONFIG_COMPAT
984         .compat_setsockopt      = compat_dccp_setsockopt,
985         .compat_getsockopt      = compat_dccp_getsockopt,
986 #endif
987 };
988 
989 static const struct net_protocol dccp_v4_protocol = {
990         .handler        = dccp_v4_rcv,
991         .err_handler    = dccp_v4_err,
992         .no_policy      = 1,
993         .netns_ok       = 1,
994 };
995 
996 static const struct proto_ops inet_dccp_ops = {
997         .family            = PF_INET,
998         .owner             = THIS_MODULE,
999         .release           = inet_release,
1000         .bind              = inet_bind,
1001         .connect           = inet_stream_connect,
1002         .socketpair        = sock_no_socketpair,
1003         .accept            = inet_accept,
1004         .getname           = inet_getname,
1005         /* FIXME: work on tcp_poll to rename it to inet_csk_poll */
1006         .poll              = dccp_poll,
1007         .ioctl             = inet_ioctl,
1008         /* FIXME: work on inet_listen to rename it to sock_common_listen */
1009         .listen            = inet_dccp_listen,
1010         .shutdown          = inet_shutdown,
1011         .setsockopt        = sock_common_setsockopt,
1012         .getsockopt        = sock_common_getsockopt,
1013         .sendmsg           = inet_sendmsg,
1014         .recvmsg           = sock_common_recvmsg,
1015         .mmap              = sock_no_mmap,
1016         .sendpage          = sock_no_sendpage,
1017 #ifdef CONFIG_COMPAT
1018         .compat_setsockopt = compat_sock_common_setsockopt,
1019         .compat_getsockopt = compat_sock_common_getsockopt,
1020 #endif
1021 };
1022 
1023 static struct inet_protosw dccp_v4_protosw = {
1024         .type           = SOCK_DCCP,
1025         .protocol       = IPPROTO_DCCP,
1026         .prot           = &dccp_v4_prot,
1027         .ops            = &inet_dccp_ops,
1028         .no_check       = 0,
1029         .flags          = INET_PROTOSW_ICSK,
1030 };
1031 
1032 static int __net_init dccp_v4_init_net(struct net *net)
1033 {
1034         if (dccp_hashinfo.bhash == NULL)
1035                 return -ESOCKTNOSUPPORT;
1036 
1037         return inet_ctl_sock_create(&net->dccp.v4_ctl_sk, PF_INET,
1038                                     SOCK_DCCP, IPPROTO_DCCP, net);
1039 }
1040 
1041 static void __net_exit dccp_v4_exit_net(struct net *net)
1042 {
1043         inet_ctl_sock_destroy(net->dccp.v4_ctl_sk);
1044 }
1045 
1046 static struct pernet_operations dccp_v4_ops = {
1047         .init   = dccp_v4_init_net,
1048         .exit   = dccp_v4_exit_net,
1049 };
1050 
1051 static int __init dccp_v4_init(void)
1052 {
1053         int err = proto_register(&dccp_v4_prot, 1);
1054 
1055         if (err != 0)
1056                 goto out;
1057 
1058         err = inet_add_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
1059         if (err != 0)
1060                 goto out_proto_unregister;
1061 
1062         inet_register_protosw(&dccp_v4_protosw);
1063 
1064         err = register_pernet_subsys(&dccp_v4_ops);
1065         if (err)
1066                 goto out_destroy_ctl_sock;
1067 out:
1068         return err;
1069 out_destroy_ctl_sock:
1070         inet_unregister_protosw(&dccp_v4_protosw);
1071         inet_del_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
1072 out_proto_unregister:
1073         proto_unregister(&dccp_v4_prot);
1074         goto out;
1075 }
1076 
1077 static void __exit dccp_v4_exit(void)
1078 {
1079         unregister_pernet_subsys(&dccp_v4_ops);
1080         inet_unregister_protosw(&dccp_v4_protosw);
1081         inet_del_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
1082         proto_unregister(&dccp_v4_prot);
1083 }
1084 
1085 module_init(dccp_v4_init);
1086 module_exit(dccp_v4_exit);
1087 
1088 /*
1089  * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1090  * values directly, Also cover the case where the protocol is not specified,
1091  * i.e. net-pf-PF_INET-proto-0-type-SOCK_DCCP
1092  */
1093 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 33, 6);
1094 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 0, 6);
1095 MODULE_LICENSE("GPL");
1096 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1097 MODULE_DESCRIPTION("DCCP - Datagram Congestion Controlled Protocol");
1098 

~ [ 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