1 /* SCTP kernel implementation
2 * Copyright (c) 1999-2000 Cisco, Inc.
3 * Copyright (c) 1999-2001 Motorola, Inc.
4 * Copyright (c) 2001-2003 International Business Machines, Corp.
5 * Copyright (c) 2001 Intel Corp.
6 * Copyright (c) 2001 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions handle all input from the IP layer into SCTP.
13 * This SCTP implementation is free software;
14 * you can redistribute it and/or modify it under the terms of
15 * the GNU General Public License as published by
16 * the Free Software Foundation; either version 2, or (at your option)
19 * This SCTP implementation is distributed in the hope that it
20 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
21 * ************************
22 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
23 * See the GNU General Public License for more details.
25 * You should have received a copy of the GNU General Public License
26 * along with GNU CC; see the file COPYING. If not, see
27 * <http://www.gnu.org/licenses/>.
29 * Please send any bug reports or fixes you make to the
31 * lksctp developers <linux-sctp@vger.kernel.org>
33 * Written or modified by:
34 * La Monte H.P. Yarroll <piggy@acm.org>
35 * Karl Knutson <karl@athena.chicago.il.us>
36 * Xingang Guo <xingang.guo@intel.com>
37 * Jon Grimm <jgrimm@us.ibm.com>
38 * Hui Huang <hui.huang@nokia.com>
39 * Daisy Chang <daisyc@us.ibm.com>
40 * Sridhar Samudrala <sri@us.ibm.com>
41 * Ardelle Fan <ardelle.fan@intel.com>
44 #include <linux/types.h>
45 #include <linux/list.h> /* For struct list_head */
46 #include <linux/socket.h>
48 #include <linux/time.h> /* For struct timeval */
49 #include <linux/slab.h>
55 #include <net/sctp/sctp.h>
56 #include <net/sctp/sm.h>
57 #include <net/sctp/checksum.h>
58 #include <net/net_namespace.h>
59 #include <linux/rhashtable.h>
60 #include <net/sock_reuseport.h>
62 /* Forward declarations for internal helpers. */
63 static int sctp_rcv_ootb(struct sk_buff
*);
64 static struct sctp_association
*__sctp_rcv_lookup(struct net
*net
,
66 const union sctp_addr
*paddr
,
67 const union sctp_addr
*laddr
,
68 struct sctp_transport
**transportp
);
69 static struct sctp_endpoint
*__sctp_rcv_lookup_endpoint(
70 struct net
*net
, struct sk_buff
*skb
,
71 const union sctp_addr
*laddr
,
72 const union sctp_addr
*daddr
);
73 static struct sctp_association
*__sctp_lookup_association(
75 const union sctp_addr
*local
,
76 const union sctp_addr
*peer
,
77 struct sctp_transport
**pt
);
79 static int sctp_add_backlog(struct sock
*sk
, struct sk_buff
*skb
);
82 /* Calculate the SCTP checksum of an SCTP packet. */
83 static inline int sctp_rcv_checksum(struct net
*net
, struct sk_buff
*skb
)
85 struct sctphdr
*sh
= sctp_hdr(skb
);
86 __le32 cmp
= sh
->checksum
;
87 __le32 val
= sctp_compute_cksum(skb
, 0);
90 /* CRC failure, dump it. */
91 __SCTP_INC_STATS(net
, SCTP_MIB_CHECKSUMERRORS
);
98 * This is the routine which IP calls when receiving an SCTP packet.
100 int sctp_rcv(struct sk_buff
*skb
)
103 struct sctp_association
*asoc
;
104 struct sctp_endpoint
*ep
= NULL
;
105 struct sctp_ep_common
*rcvr
;
106 struct sctp_transport
*transport
= NULL
;
107 struct sctp_chunk
*chunk
;
109 union sctp_addr dest
;
112 struct net
*net
= dev_net(skb
->dev
);
113 bool is_gso
= skb_is_gso(skb
) && skb_is_gso_sctp(skb
);
115 if (skb
->pkt_type
!= PACKET_HOST
)
118 __SCTP_INC_STATS(net
, SCTP_MIB_INSCTPPACKS
);
120 /* If packet is too small to contain a single chunk, let's not
121 * waste time on it anymore.
123 if (skb
->len
< sizeof(struct sctphdr
) + sizeof(struct sctp_chunkhdr
) +
124 skb_transport_offset(skb
))
127 /* If the packet is fragmented and we need to do crc checking,
128 * it's better to just linearize it otherwise crc computing
131 if ((!is_gso
&& skb_linearize(skb
)) ||
132 !pskb_may_pull(skb
, sizeof(struct sctphdr
)))
135 /* Pull up the IP header. */
136 __skb_pull(skb
, skb_transport_offset(skb
));
138 skb
->csum_valid
= 0; /* Previous value not applicable */
139 if (skb_csum_unnecessary(skb
))
140 __skb_decr_checksum_unnecessary(skb
);
141 else if (!sctp_checksum_disable
&&
143 sctp_rcv_checksum(net
, skb
) < 0)
147 __skb_pull(skb
, sizeof(struct sctphdr
));
149 family
= ipver2af(ip_hdr(skb
)->version
);
150 af
= sctp_get_af_specific(family
);
153 SCTP_INPUT_CB(skb
)->af
= af
;
155 /* Initialize local addresses for lookups. */
156 af
->from_skb(&src
, skb
, 1);
157 af
->from_skb(&dest
, skb
, 0);
159 /* If the packet is to or from a non-unicast address,
160 * silently discard the packet.
162 * This is not clearly defined in the RFC except in section
163 * 8.4 - OOTB handling. However, based on the book "Stream Control
164 * Transmission Protocol" 2.1, "It is important to note that the
165 * IP address of an SCTP transport address must be a routable
166 * unicast address. In other words, IP multicast addresses and
167 * IP broadcast addresses cannot be used in an SCTP transport
170 if (!af
->addr_valid(&src
, NULL
, skb
) ||
171 !af
->addr_valid(&dest
, NULL
, skb
))
174 asoc
= __sctp_rcv_lookup(net
, skb
, &src
, &dest
, &transport
);
177 ep
= __sctp_rcv_lookup_endpoint(net
, skb
, &dest
, &src
);
179 /* Retrieve the common input handling substructure. */
180 rcvr
= asoc
? &asoc
->base
: &ep
->base
;
184 * If a frame arrives on an interface and the receiving socket is
185 * bound to another interface, via SO_BINDTODEVICE, treat it as OOTB
187 if (sk
->sk_bound_dev_if
&& (sk
->sk_bound_dev_if
!= af
->skb_iif(skb
))) {
189 sctp_transport_put(transport
);
193 sctp_endpoint_put(ep
);
196 sk
= net
->sctp
.ctl_sock
;
197 ep
= sctp_sk(sk
)->ep
;
198 sctp_endpoint_hold(ep
);
203 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
204 * An SCTP packet is called an "out of the blue" (OOTB)
205 * packet if it is correctly formed, i.e., passed the
206 * receiver's checksum check, but the receiver is not
207 * able to identify the association to which this
211 if (sctp_rcv_ootb(skb
)) {
212 __SCTP_INC_STATS(net
, SCTP_MIB_OUTOFBLUES
);
213 goto discard_release
;
217 if (!xfrm_policy_check(sk
, XFRM_POLICY_IN
, skb
, family
))
218 goto discard_release
;
221 if (sk_filter(sk
, skb
))
222 goto discard_release
;
224 /* Create an SCTP packet structure. */
225 chunk
= sctp_chunkify(skb
, asoc
, sk
, GFP_ATOMIC
);
227 goto discard_release
;
228 SCTP_INPUT_CB(skb
)->chunk
= chunk
;
230 /* Remember what endpoint is to handle this packet. */
233 /* Remember the SCTP header. */
234 chunk
->sctp_hdr
= sctp_hdr(skb
);
236 /* Set the source and destination addresses of the incoming chunk. */
237 sctp_init_addrs(chunk
, &src
, &dest
);
239 /* Remember where we came from. */
240 chunk
->transport
= transport
;
242 /* Acquire access to the sock lock. Note: We are safe from other
243 * bottom halves on this lock, but a user may be in the lock too,
244 * so check if it is busy.
248 if (sk
!= rcvr
->sk
) {
249 /* Our cached sk is different from the rcvr->sk. This is
250 * because migrate()/accept() may have moved the association
251 * to a new socket and released all the sockets. So now we
252 * are holding a lock on the old socket while the user may
253 * be doing something with the new socket. Switch our veiw
261 if (sock_owned_by_user(sk
)) {
262 if (sctp_add_backlog(sk
, skb
)) {
264 sctp_chunk_free(chunk
);
265 skb
= NULL
; /* sctp_chunk_free already freed the skb */
266 goto discard_release
;
268 __SCTP_INC_STATS(net
, SCTP_MIB_IN_PKT_BACKLOG
);
270 __SCTP_INC_STATS(net
, SCTP_MIB_IN_PKT_SOFTIRQ
);
271 sctp_inq_push(&chunk
->rcvr
->inqueue
, chunk
);
276 /* Release the asoc/ep ref we took in the lookup calls. */
278 sctp_transport_put(transport
);
280 sctp_endpoint_put(ep
);
285 __SCTP_INC_STATS(net
, SCTP_MIB_IN_PKT_DISCARDS
);
290 /* Release the asoc/ep ref we took in the lookup calls. */
292 sctp_transport_put(transport
);
294 sctp_endpoint_put(ep
);
299 /* Process the backlog queue of the socket. Every skb on
300 * the backlog holds a ref on an association or endpoint.
301 * We hold this ref throughout the state machine to make
302 * sure that the structure we need is still around.
304 int sctp_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
306 struct sctp_chunk
*chunk
= SCTP_INPUT_CB(skb
)->chunk
;
307 struct sctp_inq
*inqueue
= &chunk
->rcvr
->inqueue
;
308 struct sctp_transport
*t
= chunk
->transport
;
309 struct sctp_ep_common
*rcvr
= NULL
;
314 /* If the rcvr is dead then the association or endpoint
315 * has been deleted and we can safely drop the chunk
316 * and refs that we are holding.
319 sctp_chunk_free(chunk
);
323 if (unlikely(rcvr
->sk
!= sk
)) {
324 /* In this case, the association moved from one socket to
325 * another. We are currently sitting on the backlog of the
326 * old socket, so we need to move.
327 * However, since we are here in the process context we
328 * need to take make sure that the user doesn't own
329 * the new socket when we process the packet.
330 * If the new socket is user-owned, queue the chunk to the
331 * backlog of the new socket without dropping any refs.
332 * Otherwise, we can safely push the chunk on the inqueue.
339 if (sock_owned_by_user(sk
)) {
340 if (sk_add_backlog(sk
, skb
, sk
->sk_rcvbuf
))
341 sctp_chunk_free(chunk
);
345 sctp_inq_push(inqueue
, chunk
);
350 /* If the chunk was backloged again, don't drop refs */
354 sctp_inq_push(inqueue
, chunk
);
358 /* Release the refs we took in sctp_add_backlog */
359 if (SCTP_EP_TYPE_ASSOCIATION
== rcvr
->type
)
360 sctp_transport_put(t
);
361 else if (SCTP_EP_TYPE_SOCKET
== rcvr
->type
)
362 sctp_endpoint_put(sctp_ep(rcvr
));
369 static int sctp_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
371 struct sctp_chunk
*chunk
= SCTP_INPUT_CB(skb
)->chunk
;
372 struct sctp_transport
*t
= chunk
->transport
;
373 struct sctp_ep_common
*rcvr
= chunk
->rcvr
;
376 ret
= sk_add_backlog(sk
, skb
, sk
->sk_rcvbuf
);
378 /* Hold the assoc/ep while hanging on the backlog queue.
379 * This way, we know structures we need will not disappear
382 if (SCTP_EP_TYPE_ASSOCIATION
== rcvr
->type
)
383 sctp_transport_hold(t
);
384 else if (SCTP_EP_TYPE_SOCKET
== rcvr
->type
)
385 sctp_endpoint_hold(sctp_ep(rcvr
));
393 /* Handle icmp frag needed error. */
394 void sctp_icmp_frag_needed(struct sock
*sk
, struct sctp_association
*asoc
,
395 struct sctp_transport
*t
, __u32 pmtu
)
397 if (!t
|| (t
->pathmtu
<= pmtu
))
400 if (sock_owned_by_user(sk
)) {
401 atomic_set(&t
->mtu_info
, pmtu
);
402 asoc
->pmtu_pending
= 1;
407 if (!(t
->param_flags
& SPP_PMTUD_ENABLE
))
408 /* We can't allow retransmitting in such case, as the
409 * retransmission would be sized just as before, and thus we
410 * would get another icmp, and retransmit again.
414 /* Update transports view of the MTU. Return if no update was needed.
415 * If an update wasn't needed/possible, it also doesn't make sense to
416 * try to retransmit now.
418 if (!sctp_transport_update_pmtu(t
, pmtu
))
421 /* Update association pmtu. */
422 sctp_assoc_sync_pmtu(asoc
);
424 /* Retransmit with the new pmtu setting. */
425 sctp_retransmit(&asoc
->outqueue
, t
, SCTP_RTXR_PMTUD
);
428 void sctp_icmp_redirect(struct sock
*sk
, struct sctp_transport
*t
,
431 struct dst_entry
*dst
;
433 if (sock_owned_by_user(sk
) || !t
)
435 dst
= sctp_transport_dst_check(t
);
437 dst
->ops
->redirect(dst
, sk
, skb
);
441 * SCTP Implementer's Guide, 2.37 ICMP handling procedures
443 * ICMP8) If the ICMP code is a "Unrecognized next header type encountered"
444 * or a "Protocol Unreachable" treat this message as an abort
445 * with the T bit set.
447 * This function sends an event to the state machine, which will abort the
451 void sctp_icmp_proto_unreachable(struct sock
*sk
,
452 struct sctp_association
*asoc
,
453 struct sctp_transport
*t
)
455 if (sock_owned_by_user(sk
)) {
456 if (timer_pending(&t
->proto_unreach_timer
))
459 if (!mod_timer(&t
->proto_unreach_timer
,
461 sctp_association_hold(asoc
);
464 struct net
*net
= sock_net(sk
);
466 pr_debug("%s: unrecognized next header type "
467 "encountered!\n", __func__
);
469 if (del_timer(&t
->proto_unreach_timer
))
470 sctp_association_put(asoc
);
472 sctp_do_sm(net
, SCTP_EVENT_T_OTHER
,
473 SCTP_ST_OTHER(SCTP_EVENT_ICMP_PROTO_UNREACH
),
474 asoc
->state
, asoc
->ep
, asoc
, t
,
479 /* Common lookup code for icmp/icmpv6 error handler. */
480 struct sock
*sctp_err_lookup(struct net
*net
, int family
, struct sk_buff
*skb
,
481 struct sctphdr
*sctphdr
,
482 struct sctp_association
**app
,
483 struct sctp_transport
**tpp
)
485 struct sctp_init_chunk
*chunkhdr
, _chunkhdr
;
486 union sctp_addr saddr
;
487 union sctp_addr daddr
;
489 struct sock
*sk
= NULL
;
490 struct sctp_association
*asoc
;
491 struct sctp_transport
*transport
= NULL
;
492 __u32 vtag
= ntohl(sctphdr
->vtag
);
494 *app
= NULL
; *tpp
= NULL
;
496 af
= sctp_get_af_specific(family
);
501 /* Initialize local addresses for lookups. */
502 af
->from_skb(&saddr
, skb
, 1);
503 af
->from_skb(&daddr
, skb
, 0);
505 /* Look for an association that matches the incoming ICMP error
508 asoc
= __sctp_lookup_association(net
, &saddr
, &daddr
, &transport
);
514 /* RFC 4960, Appendix C. ICMP Handling
516 * ICMP6) An implementation MUST validate that the Verification Tag
517 * contained in the ICMP message matches the Verification Tag of
518 * the peer. If the Verification Tag is not 0 and does NOT
519 * match, discard the ICMP message. If it is 0 and the ICMP
520 * message contains enough bytes to verify that the chunk type is
521 * an INIT chunk and that the Initiate Tag matches the tag of the
522 * peer, continue with ICMP7. If the ICMP message is too short
523 * or the chunk type or the Initiate Tag does not match, silently
524 * discard the packet.
527 /* chunk header + first 4 octects of init header */
528 chunkhdr
= skb_header_pointer(skb
, skb_transport_offset(skb
) +
529 sizeof(struct sctphdr
),
530 sizeof(struct sctp_chunkhdr
) +
531 sizeof(__be32
), &_chunkhdr
);
533 chunkhdr
->chunk_hdr
.type
!= SCTP_CID_INIT
||
534 ntohl(chunkhdr
->init_hdr
.init_tag
) != asoc
->c
.my_vtag
)
537 } else if (vtag
!= asoc
->c
.peer_vtag
) {
543 /* If too many ICMPs get dropped on busy
544 * servers this needs to be solved differently.
546 if (sock_owned_by_user(sk
))
547 __NET_INC_STATS(net
, LINUX_MIB_LOCKDROPPEDICMPS
);
554 sctp_transport_put(transport
);
558 /* Common cleanup code for icmp/icmpv6 error handler. */
559 void sctp_err_finish(struct sock
*sk
, struct sctp_transport
*t
)
562 sctp_transport_put(t
);
566 * This routine is called by the ICMP module when it gets some
567 * sort of error condition. If err < 0 then the socket should
568 * be closed and the error returned to the user. If err > 0
569 * it's just the icmp type << 8 | icmp code. After adjustment
570 * header points to the first 8 bytes of the sctp header. We need
571 * to find the appropriate port.
573 * The locking strategy used here is very "optimistic". When
574 * someone else accesses the socket the ICMP is just dropped
575 * and for some paths there is no check at all.
576 * A more general error queue to queue errors for later handling
577 * is probably better.
580 int sctp_v4_err(struct sk_buff
*skb
, __u32 info
)
582 const struct iphdr
*iph
= (const struct iphdr
*)skb
->data
;
583 const int ihlen
= iph
->ihl
* 4;
584 const int type
= icmp_hdr(skb
)->type
;
585 const int code
= icmp_hdr(skb
)->code
;
587 struct sctp_association
*asoc
= NULL
;
588 struct sctp_transport
*transport
;
589 struct inet_sock
*inet
;
590 __u16 saveip
, savesctp
;
592 struct net
*net
= dev_net(skb
->dev
);
594 /* Fix up skb to look at the embedded net header. */
595 saveip
= skb
->network_header
;
596 savesctp
= skb
->transport_header
;
597 skb_reset_network_header(skb
);
598 skb_set_transport_header(skb
, ihlen
);
599 sk
= sctp_err_lookup(net
, AF_INET
, skb
, sctp_hdr(skb
), &asoc
, &transport
);
600 /* Put back, the original values. */
601 skb
->network_header
= saveip
;
602 skb
->transport_header
= savesctp
;
604 __ICMP_INC_STATS(net
, ICMP_MIB_INERRORS
);
607 /* Warning: The sock lock is held. Remember to call
612 case ICMP_PARAMETERPROB
:
615 case ICMP_DEST_UNREACH
:
616 if (code
> NR_ICMP_UNREACH
)
619 /* PMTU discovery (RFC1191) */
620 if (ICMP_FRAG_NEEDED
== code
) {
621 sctp_icmp_frag_needed(sk
, asoc
, transport
,
625 if (ICMP_PROT_UNREACH
== code
) {
626 sctp_icmp_proto_unreachable(sk
, asoc
,
631 err
= icmp_err_convert
[code
].errno
;
633 case ICMP_TIME_EXCEEDED
:
634 /* Ignore any time exceeded errors due to fragment reassembly
637 if (ICMP_EXC_FRAGTIME
== code
)
643 sctp_icmp_redirect(sk
, transport
, skb
);
644 /* Fall through to out_unlock. */
650 if (!sock_owned_by_user(sk
) && inet
->recverr
) {
652 sk
->sk_error_report(sk
);
653 } else { /* Only an error on timeout */
654 sk
->sk_err_soft
= err
;
658 sctp_err_finish(sk
, transport
);
663 * RFC 2960, 8.4 - Handle "Out of the blue" Packets.
665 * This function scans all the chunks in the OOTB packet to determine if
666 * the packet should be discarded right away. If a response might be needed
667 * for this packet, or, if further processing is possible, the packet will
668 * be queued to a proper inqueue for the next phase of handling.
671 * Return 0 - If further processing is needed.
672 * Return 1 - If the packet can be discarded right away.
674 static int sctp_rcv_ootb(struct sk_buff
*skb
)
676 struct sctp_chunkhdr
*ch
, _ch
;
677 int ch_end
, offset
= 0;
679 /* Scan through all the chunks in the packet. */
681 /* Make sure we have at least the header there */
682 if (offset
+ sizeof(_ch
) > skb
->len
)
685 ch
= skb_header_pointer(skb
, offset
, sizeof(*ch
), &_ch
);
687 /* Break out if chunk length is less then minimal. */
688 if (ntohs(ch
->length
) < sizeof(_ch
))
691 ch_end
= offset
+ SCTP_PAD4(ntohs(ch
->length
));
692 if (ch_end
> skb
->len
)
695 /* RFC 8.4, 2) If the OOTB packet contains an ABORT chunk, the
696 * receiver MUST silently discard the OOTB packet and take no
699 if (SCTP_CID_ABORT
== ch
->type
)
702 /* RFC 8.4, 6) If the packet contains a SHUTDOWN COMPLETE
703 * chunk, the receiver should silently discard the packet
704 * and take no further action.
706 if (SCTP_CID_SHUTDOWN_COMPLETE
== ch
->type
)
710 * This will discard packets with INIT chunk bundled as
711 * subsequent chunks in the packet. When INIT is first,
712 * the normal INIT processing will discard the chunk.
714 if (SCTP_CID_INIT
== ch
->type
&& (void *)ch
!= skb
->data
)
718 } while (ch_end
< skb
->len
);
726 /* Insert endpoint into the hash table. */
727 static int __sctp_hash_endpoint(struct sctp_endpoint
*ep
)
729 struct sock
*sk
= ep
->base
.sk
;
730 struct net
*net
= sock_net(sk
);
731 struct sctp_hashbucket
*head
;
732 struct sctp_ep_common
*epb
;
735 epb
->hashent
= sctp_ep_hashfn(net
, epb
->bind_addr
.port
);
736 head
= &sctp_ep_hashtable
[epb
->hashent
];
738 if (sk
->sk_reuseport
) {
739 bool any
= sctp_is_ep_boundall(sk
);
740 struct sctp_ep_common
*epb2
;
741 struct list_head
*list
;
742 int cnt
= 0, err
= 1;
744 list_for_each(list
, &ep
->base
.bind_addr
.address_list
)
747 sctp_for_each_hentry(epb2
, &head
->chain
) {
748 struct sock
*sk2
= epb2
->sk
;
750 if (!net_eq(sock_net(sk2
), net
) || sk2
== sk
||
751 !uid_eq(sock_i_uid(sk2
), sock_i_uid(sk
)) ||
755 err
= sctp_bind_addrs_check(sctp_sk(sk2
),
758 err
= reuseport_add_sock(sk
, sk2
, any
);
762 } else if (err
< 0) {
768 err
= reuseport_alloc(sk
, any
);
774 write_lock(&head
->lock
);
775 hlist_add_head(&epb
->node
, &head
->chain
);
776 write_unlock(&head
->lock
);
780 /* Add an endpoint to the hash. Local BH-safe. */
781 int sctp_hash_endpoint(struct sctp_endpoint
*ep
)
786 err
= __sctp_hash_endpoint(ep
);
792 /* Remove endpoint from the hash table. */
793 static void __sctp_unhash_endpoint(struct sctp_endpoint
*ep
)
795 struct sock
*sk
= ep
->base
.sk
;
796 struct sctp_hashbucket
*head
;
797 struct sctp_ep_common
*epb
;
801 epb
->hashent
= sctp_ep_hashfn(sock_net(sk
), epb
->bind_addr
.port
);
803 head
= &sctp_ep_hashtable
[epb
->hashent
];
805 if (rcu_access_pointer(sk
->sk_reuseport_cb
))
806 reuseport_detach_sock(sk
);
808 write_lock(&head
->lock
);
809 hlist_del_init(&epb
->node
);
810 write_unlock(&head
->lock
);
813 /* Remove endpoint from the hash. Local BH-safe. */
814 void sctp_unhash_endpoint(struct sctp_endpoint
*ep
)
817 __sctp_unhash_endpoint(ep
);
821 static inline __u32
sctp_hashfn(const struct net
*net
, __be16 lport
,
822 const union sctp_addr
*paddr
, __u32 seed
)
826 if (paddr
->sa
.sa_family
== AF_INET6
)
827 addr
= jhash(&paddr
->v6
.sin6_addr
, 16, seed
);
829 addr
= (__force __u32
)paddr
->v4
.sin_addr
.s_addr
;
831 return jhash_3words(addr
, ((__force __u32
)paddr
->v4
.sin_port
) << 16 |
832 (__force __u32
)lport
, net_hash_mix(net
), seed
);
835 /* Look up an endpoint. */
836 static struct sctp_endpoint
*__sctp_rcv_lookup_endpoint(
837 struct net
*net
, struct sk_buff
*skb
,
838 const union sctp_addr
*laddr
,
839 const union sctp_addr
*paddr
)
841 struct sctp_hashbucket
*head
;
842 struct sctp_ep_common
*epb
;
843 struct sctp_endpoint
*ep
;
848 lport
= laddr
->v4
.sin_port
;
849 hash
= sctp_ep_hashfn(net
, ntohs(lport
));
850 head
= &sctp_ep_hashtable
[hash
];
851 read_lock(&head
->lock
);
852 sctp_for_each_hentry(epb
, &head
->chain
) {
854 if (sctp_endpoint_is_match(ep
, net
, laddr
))
858 ep
= sctp_sk(net
->sctp
.ctl_sock
)->ep
;
862 if (sk
->sk_reuseport
) {
863 __u32 phash
= sctp_hashfn(net
, lport
, paddr
, 0);
865 sk
= reuseport_select_sock(sk
, phash
, skb
,
866 sizeof(struct sctphdr
));
868 ep
= sctp_sk(sk
)->ep
;
870 sctp_endpoint_hold(ep
);
871 read_unlock(&head
->lock
);
875 /* rhashtable for transport */
876 struct sctp_hash_cmp_arg
{
877 const union sctp_addr
*paddr
;
878 const struct net
*net
;
882 static inline int sctp_hash_cmp(struct rhashtable_compare_arg
*arg
,
885 struct sctp_transport
*t
= (struct sctp_transport
*)ptr
;
886 const struct sctp_hash_cmp_arg
*x
= arg
->key
;
889 if (!sctp_cmp_addr_exact(&t
->ipaddr
, x
->paddr
))
891 if (!sctp_transport_hold(t
))
894 if (!net_eq(sock_net(t
->asoc
->base
.sk
), x
->net
))
896 if (x
->lport
!= htons(t
->asoc
->base
.bind_addr
.port
))
901 sctp_transport_put(t
);
905 static inline __u32
sctp_hash_obj(const void *data
, u32 len
, u32 seed
)
907 const struct sctp_transport
*t
= data
;
909 return sctp_hashfn(sock_net(t
->asoc
->base
.sk
),
910 htons(t
->asoc
->base
.bind_addr
.port
),
914 static inline __u32
sctp_hash_key(const void *data
, u32 len
, u32 seed
)
916 const struct sctp_hash_cmp_arg
*x
= data
;
918 return sctp_hashfn(x
->net
, x
->lport
, x
->paddr
, seed
);
921 static const struct rhashtable_params sctp_hash_params
= {
922 .head_offset
= offsetof(struct sctp_transport
, node
),
923 .hashfn
= sctp_hash_key
,
924 .obj_hashfn
= sctp_hash_obj
,
925 .obj_cmpfn
= sctp_hash_cmp
,
926 .automatic_shrinking
= true,
929 int sctp_transport_hashtable_init(void)
931 return rhltable_init(&sctp_transport_hashtable
, &sctp_hash_params
);
934 void sctp_transport_hashtable_destroy(void)
936 rhltable_destroy(&sctp_transport_hashtable
);
939 int sctp_hash_transport(struct sctp_transport
*t
)
941 struct sctp_transport
*transport
;
942 struct rhlist_head
*tmp
, *list
;
943 struct sctp_hash_cmp_arg arg
;
949 arg
.net
= sock_net(t
->asoc
->base
.sk
);
950 arg
.paddr
= &t
->ipaddr
;
951 arg
.lport
= htons(t
->asoc
->base
.bind_addr
.port
);
954 list
= rhltable_lookup(&sctp_transport_hashtable
, &arg
,
957 rhl_for_each_entry_rcu(transport
, tmp
, list
, node
)
958 if (transport
->asoc
->ep
== t
->asoc
->ep
) {
964 err
= rhltable_insert_key(&sctp_transport_hashtable
, &arg
,
965 &t
->node
, sctp_hash_params
);
967 pr_err_once("insert transport fail, errno %d\n", err
);
972 void sctp_unhash_transport(struct sctp_transport
*t
)
977 rhltable_remove(&sctp_transport_hashtable
, &t
->node
,
981 /* return a transport with holding it */
982 struct sctp_transport
*sctp_addrs_lookup_transport(
984 const union sctp_addr
*laddr
,
985 const union sctp_addr
*paddr
)
987 struct rhlist_head
*tmp
, *list
;
988 struct sctp_transport
*t
;
989 struct sctp_hash_cmp_arg arg
= {
992 .lport
= laddr
->v4
.sin_port
,
995 list
= rhltable_lookup(&sctp_transport_hashtable
, &arg
,
998 rhl_for_each_entry_rcu(t
, tmp
, list
, node
) {
999 if (!sctp_transport_hold(t
))
1002 if (sctp_bind_addr_match(&t
->asoc
->base
.bind_addr
,
1003 laddr
, sctp_sk(t
->asoc
->base
.sk
)))
1005 sctp_transport_put(t
);
1011 /* return a transport without holding it, as it's only used under sock lock */
1012 struct sctp_transport
*sctp_epaddr_lookup_transport(
1013 const struct sctp_endpoint
*ep
,
1014 const union sctp_addr
*paddr
)
1016 struct net
*net
= sock_net(ep
->base
.sk
);
1017 struct rhlist_head
*tmp
, *list
;
1018 struct sctp_transport
*t
;
1019 struct sctp_hash_cmp_arg arg
= {
1022 .lport
= htons(ep
->base
.bind_addr
.port
),
1025 list
= rhltable_lookup(&sctp_transport_hashtable
, &arg
,
1028 rhl_for_each_entry_rcu(t
, tmp
, list
, node
)
1029 if (ep
== t
->asoc
->ep
)
1035 /* Look up an association. */
1036 static struct sctp_association
*__sctp_lookup_association(
1038 const union sctp_addr
*local
,
1039 const union sctp_addr
*peer
,
1040 struct sctp_transport
**pt
)
1042 struct sctp_transport
*t
;
1043 struct sctp_association
*asoc
= NULL
;
1045 t
= sctp_addrs_lookup_transport(net
, local
, peer
);
1056 /* Look up an association. protected by RCU read lock */
1058 struct sctp_association
*sctp_lookup_association(struct net
*net
,
1059 const union sctp_addr
*laddr
,
1060 const union sctp_addr
*paddr
,
1061 struct sctp_transport
**transportp
)
1063 struct sctp_association
*asoc
;
1066 asoc
= __sctp_lookup_association(net
, laddr
, paddr
, transportp
);
1072 /* Is there an association matching the given local and peer addresses? */
1073 bool sctp_has_association(struct net
*net
,
1074 const union sctp_addr
*laddr
,
1075 const union sctp_addr
*paddr
)
1077 struct sctp_transport
*transport
;
1079 if (sctp_lookup_association(net
, laddr
, paddr
, &transport
)) {
1080 sctp_transport_put(transport
);
1088 * SCTP Implementors Guide, 2.18 Handling of address
1089 * parameters within the INIT or INIT-ACK.
1091 * D) When searching for a matching TCB upon reception of an INIT
1092 * or INIT-ACK chunk the receiver SHOULD use not only the
1093 * source address of the packet (containing the INIT or
1094 * INIT-ACK) but the receiver SHOULD also use all valid
1095 * address parameters contained within the chunk.
1097 * 2.18.3 Solution description
1099 * This new text clearly specifies to an implementor the need
1100 * to look within the INIT or INIT-ACK. Any implementation that
1101 * does not do this, may not be able to establish associations
1102 * in certain circumstances.
1105 static struct sctp_association
*__sctp_rcv_init_lookup(struct net
*net
,
1106 struct sk_buff
*skb
,
1107 const union sctp_addr
*laddr
, struct sctp_transport
**transportp
)
1109 struct sctp_association
*asoc
;
1110 union sctp_addr addr
;
1111 union sctp_addr
*paddr
= &addr
;
1112 struct sctphdr
*sh
= sctp_hdr(skb
);
1113 union sctp_params params
;
1114 struct sctp_init_chunk
*init
;
1118 * This code will NOT touch anything inside the chunk--it is
1119 * strictly READ-ONLY.
1121 * RFC 2960 3 SCTP packet Format
1123 * Multiple chunks can be bundled into one SCTP packet up to
1124 * the MTU size, except for the INIT, INIT ACK, and SHUTDOWN
1125 * COMPLETE chunks. These chunks MUST NOT be bundled with any
1126 * other chunk in a packet. See Section 6.10 for more details
1127 * on chunk bundling.
1130 /* Find the start of the TLVs and the end of the chunk. This is
1131 * the region we search for address parameters.
1133 init
= (struct sctp_init_chunk
*)skb
->data
;
1135 /* Walk the parameters looking for embedded addresses. */
1136 sctp_walk_params(params
, init
, init_hdr
.params
) {
1138 /* Note: Ignoring hostname addresses. */
1139 af
= sctp_get_af_specific(param_type2af(params
.p
->type
));
1143 af
->from_addr_param(paddr
, params
.addr
, sh
->source
, 0);
1145 asoc
= __sctp_lookup_association(net
, laddr
, paddr
, transportp
);
1153 /* ADD-IP, Section 5.2
1154 * When an endpoint receives an ASCONF Chunk from the remote peer
1155 * special procedures may be needed to identify the association the
1156 * ASCONF Chunk is associated with. To properly find the association
1157 * the following procedures SHOULD be followed:
1159 * D2) If the association is not found, use the address found in the
1160 * Address Parameter TLV combined with the port number found in the
1161 * SCTP common header. If found proceed to rule D4.
1163 * D2-ext) If more than one ASCONF Chunks are packed together, use the
1164 * address found in the ASCONF Address Parameter TLV of each of the
1165 * subsequent ASCONF Chunks. If found, proceed to rule D4.
1167 static struct sctp_association
*__sctp_rcv_asconf_lookup(
1169 struct sctp_chunkhdr
*ch
,
1170 const union sctp_addr
*laddr
,
1172 struct sctp_transport
**transportp
)
1174 struct sctp_addip_chunk
*asconf
= (struct sctp_addip_chunk
*)ch
;
1176 union sctp_addr_param
*param
;
1177 union sctp_addr paddr
;
1179 /* Skip over the ADDIP header and find the Address parameter */
1180 param
= (union sctp_addr_param
*)(asconf
+ 1);
1182 af
= sctp_get_af_specific(param_type2af(param
->p
.type
));
1186 af
->from_addr_param(&paddr
, param
, peer_port
, 0);
1188 return __sctp_lookup_association(net
, laddr
, &paddr
, transportp
);
1192 /* SCTP-AUTH, Section 6.3:
1193 * If the receiver does not find a STCB for a packet containing an AUTH
1194 * chunk as the first chunk and not a COOKIE-ECHO chunk as the second
1195 * chunk, it MUST use the chunks after the AUTH chunk to look up an existing
1198 * This means that any chunks that can help us identify the association need
1199 * to be looked at to find this association.
1201 static struct sctp_association
*__sctp_rcv_walk_lookup(struct net
*net
,
1202 struct sk_buff
*skb
,
1203 const union sctp_addr
*laddr
,
1204 struct sctp_transport
**transportp
)
1206 struct sctp_association
*asoc
= NULL
;
1207 struct sctp_chunkhdr
*ch
;
1209 unsigned int chunk_num
= 1;
1212 /* Walk through the chunks looking for AUTH or ASCONF chunks
1213 * to help us find the association.
1215 ch
= (struct sctp_chunkhdr
*)skb
->data
;
1217 /* Break out if chunk length is less then minimal. */
1218 if (ntohs(ch
->length
) < sizeof(*ch
))
1221 ch_end
= ((__u8
*)ch
) + SCTP_PAD4(ntohs(ch
->length
));
1222 if (ch_end
> skb_tail_pointer(skb
))
1227 have_auth
= chunk_num
;
1230 case SCTP_CID_COOKIE_ECHO
:
1231 /* If a packet arrives containing an AUTH chunk as
1232 * a first chunk, a COOKIE-ECHO chunk as the second
1233 * chunk, and possibly more chunks after them, and
1234 * the receiver does not have an STCB for that
1235 * packet, then authentication is based on
1236 * the contents of the COOKIE- ECHO chunk.
1238 if (have_auth
== 1 && chunk_num
== 2)
1242 case SCTP_CID_ASCONF
:
1243 if (have_auth
|| net
->sctp
.addip_noauth
)
1244 asoc
= __sctp_rcv_asconf_lookup(
1246 sctp_hdr(skb
)->source
,
1255 ch
= (struct sctp_chunkhdr
*)ch_end
;
1257 } while (ch_end
< skb_tail_pointer(skb
));
1263 * There are circumstances when we need to look inside the SCTP packet
1264 * for information to help us find the association. Examples
1265 * include looking inside of INIT/INIT-ACK chunks or after the AUTH
1268 static struct sctp_association
*__sctp_rcv_lookup_harder(struct net
*net
,
1269 struct sk_buff
*skb
,
1270 const union sctp_addr
*laddr
,
1271 struct sctp_transport
**transportp
)
1273 struct sctp_chunkhdr
*ch
;
1275 /* We do not allow GSO frames here as we need to linearize and
1276 * then cannot guarantee frame boundaries. This shouldn't be an
1277 * issue as packets hitting this are mostly INIT or INIT-ACK and
1278 * those cannot be on GSO-style anyway.
1280 if (skb_is_gso(skb
) && skb_is_gso_sctp(skb
))
1283 ch
= (struct sctp_chunkhdr
*)skb
->data
;
1285 /* The code below will attempt to walk the chunk and extract
1286 * parameter information. Before we do that, we need to verify
1287 * that the chunk length doesn't cause overflow. Otherwise, we'll
1290 if (SCTP_PAD4(ntohs(ch
->length
)) > skb
->len
)
1293 /* If this is INIT/INIT-ACK look inside the chunk too. */
1294 if (ch
->type
== SCTP_CID_INIT
|| ch
->type
== SCTP_CID_INIT_ACK
)
1295 return __sctp_rcv_init_lookup(net
, skb
, laddr
, transportp
);
1297 return __sctp_rcv_walk_lookup(net
, skb
, laddr
, transportp
);
1300 /* Lookup an association for an inbound skb. */
1301 static struct sctp_association
*__sctp_rcv_lookup(struct net
*net
,
1302 struct sk_buff
*skb
,
1303 const union sctp_addr
*paddr
,
1304 const union sctp_addr
*laddr
,
1305 struct sctp_transport
**transportp
)
1307 struct sctp_association
*asoc
;
1309 asoc
= __sctp_lookup_association(net
, laddr
, paddr
, transportp
);
1313 /* Further lookup for INIT/INIT-ACK packets.
1314 * SCTP Implementors Guide, 2.18 Handling of address
1315 * parameters within the INIT or INIT-ACK.
1317 asoc
= __sctp_rcv_lookup_harder(net
, skb
, laddr
, transportp
);
1321 if (paddr
->sa
.sa_family
== AF_INET
)
1322 pr_debug("sctp: asoc not found for src:%pI4:%d dst:%pI4:%d\n",
1323 &laddr
->v4
.sin_addr
, ntohs(laddr
->v4
.sin_port
),
1324 &paddr
->v4
.sin_addr
, ntohs(paddr
->v4
.sin_port
));
1326 pr_debug("sctp: asoc not found for src:%pI6:%d dst:%pI6:%d\n",
1327 &laddr
->v6
.sin6_addr
, ntohs(laddr
->v6
.sin6_port
),
1328 &paddr
->v6
.sin6_addr
, ntohs(paddr
->v6
.sin6_port
));