1 /* transport.c: Rx Transport routines
3 * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
12 #include <linux/sched.h>
13 #include <linux/slab.h>
14 #include <linux/module.h>
15 #include <rxrpc/transport.h>
16 #include <rxrpc/peer.h>
17 #include <rxrpc/connection.h>
18 #include <rxrpc/call.h>
19 #include <rxrpc/message.h>
20 #include <rxrpc/krxiod.h>
21 #include <rxrpc/krxsecd.h>
22 #include <linux/udp.h>
24 #include <linux/in6.h>
25 #include <linux/icmp.h>
26 #include <linux/skbuff.h>
29 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
30 #include <linux/ipv6.h> /* this should _really_ be in errqueue.h.. */
32 #include <linux/errqueue.h>
33 #include <asm/uaccess.h>
34 #include <asm/checksum.h>
38 struct cmsghdr cmsg; /* control message header */
39 struct sock_extended_err ee; /* extended error information */
40 struct sockaddr_in icmp_src; /* ICMP packet source address */
43 static DEFINE_SPINLOCK(rxrpc_transports_lock);
44 static struct list_head rxrpc_transports = LIST_HEAD_INIT(rxrpc_transports);
46 __RXACCT_DECL(atomic_t rxrpc_transport_count);
47 LIST_HEAD(rxrpc_proc_transports);
48 DECLARE_RWSEM(rxrpc_proc_transports_sem);
50 static void rxrpc_data_ready(struct sock *sk, int count);
51 static void rxrpc_error_report(struct sock *sk);
52 static int rxrpc_trans_receive_new_call(struct rxrpc_transport *trans,
53 struct list_head *msgq);
54 static void rxrpc_trans_receive_error_report(struct rxrpc_transport *trans);
56 /*****************************************************************************/
58 * create a new transport endpoint using the specified UDP port
60 int rxrpc_create_transport(unsigned short port,
61 struct rxrpc_transport **_trans)
63 struct rxrpc_transport *trans;
64 struct sockaddr_in sin;
71 trans = kzalloc(sizeof(struct rxrpc_transport), GFP_KERNEL);
75 atomic_set(&trans->usage, 1);
76 INIT_LIST_HEAD(&trans->services);
77 INIT_LIST_HEAD(&trans->link);
78 INIT_LIST_HEAD(&trans->krxiodq_link);
79 spin_lock_init(&trans->lock);
80 INIT_LIST_HEAD(&trans->peer_active);
81 INIT_LIST_HEAD(&trans->peer_graveyard);
82 spin_lock_init(&trans->peer_gylock);
83 init_waitqueue_head(&trans->peer_gy_waitq);
84 rwlock_init(&trans->peer_lock);
85 atomic_set(&trans->peer_count, 0);
88 /* create a UDP socket to be my actual transport endpoint */
89 ret = sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &trans->socket);
93 /* use the specified port */
95 memset(&sin, 0, sizeof(sin));
96 sin.sin_family = AF_INET;
97 sin.sin_port = htons(port);
98 ret = trans->socket->ops->bind(trans->socket,
99 (struct sockaddr *) &sin,
108 ret = trans->socket->ops->setsockopt(trans->socket, SOL_IP, IP_RECVERR,
109 (char *) &opt, sizeof(opt));
112 spin_lock(&rxrpc_transports_lock);
113 list_add(&trans->link, &rxrpc_transports);
114 spin_unlock(&rxrpc_transports_lock);
116 /* set the socket up */
117 sock = trans->socket->sk;
118 sock->sk_user_data = trans;
119 sock->sk_data_ready = rxrpc_data_ready;
120 sock->sk_error_report = rxrpc_error_report;
122 down_write(&rxrpc_proc_transports_sem);
123 list_add_tail(&trans->proc_link, &rxrpc_proc_transports);
124 up_write(&rxrpc_proc_transports_sem);
126 __RXACCT(atomic_inc(&rxrpc_transport_count));
129 _leave(" = 0 (%p)", trans);
133 /* finish cleaning up the transport (not really needed here, but...) */
135 trans->socket->ops->shutdown(trans->socket, 2);
137 /* close the socket */
139 trans->socket->sk->sk_user_data = NULL;
140 sock_release(trans->socket);
141 trans->socket = NULL;
147 _leave(" = %d", ret);
149 } /* end rxrpc_create_transport() */
151 /*****************************************************************************/
153 * destroy a transport endpoint
155 void rxrpc_put_transport(struct rxrpc_transport *trans)
157 _enter("%p{u=%d p=%hu}",
158 trans, atomic_read(&trans->usage), trans->port);
160 BUG_ON(atomic_read(&trans->usage) <= 0);
162 /* to prevent a race, the decrement and the dequeue must be
163 * effectively atomic */
164 spin_lock(&rxrpc_transports_lock);
165 if (likely(!atomic_dec_and_test(&trans->usage))) {
166 spin_unlock(&rxrpc_transports_lock);
171 list_del(&trans->link);
172 spin_unlock(&rxrpc_transports_lock);
174 /* finish cleaning up the transport */
176 trans->socket->ops->shutdown(trans->socket, 2);
178 rxrpc_krxsecd_clear_transport(trans);
179 rxrpc_krxiod_dequeue_transport(trans);
181 /* discard all peer information */
182 rxrpc_peer_clearall(trans);
184 down_write(&rxrpc_proc_transports_sem);
185 list_del(&trans->proc_link);
186 up_write(&rxrpc_proc_transports_sem);
187 __RXACCT(atomic_dec(&rxrpc_transport_count));
189 /* close the socket */
191 trans->socket->sk->sk_user_data = NULL;
192 sock_release(trans->socket);
193 trans->socket = NULL;
199 } /* end rxrpc_put_transport() */
201 /*****************************************************************************/
203 * add a service to a transport to be listened upon
205 int rxrpc_add_service(struct rxrpc_transport *trans,
206 struct rxrpc_service *newsrv)
208 struct rxrpc_service *srv;
209 struct list_head *_p;
212 _enter("%p{%hu},%p{%hu}",
213 trans, trans->port, newsrv, newsrv->service_id);
215 /* verify that the service ID is not already present */
216 spin_lock(&trans->lock);
218 list_for_each(_p, &trans->services) {
219 srv = list_entry(_p, struct rxrpc_service, link);
220 if (srv->service_id == newsrv->service_id)
224 /* okay - add the transport to the list */
225 list_add_tail(&newsrv->link, &trans->services);
226 rxrpc_get_transport(trans);
230 spin_unlock(&trans->lock);
234 } /* end rxrpc_add_service() */
236 /*****************************************************************************/
238 * remove a service from a transport
240 void rxrpc_del_service(struct rxrpc_transport *trans, struct rxrpc_service *srv)
242 _enter("%p{%hu},%p{%hu}", trans, trans->port, srv, srv->service_id);
244 spin_lock(&trans->lock);
245 list_del(&srv->link);
246 spin_unlock(&trans->lock);
248 rxrpc_put_transport(trans);
251 } /* end rxrpc_del_service() */
253 /*****************************************************************************/
255 * INET callback when data has been received on the socket.
257 static void rxrpc_data_ready(struct sock *sk, int count)
259 struct rxrpc_transport *trans;
261 _enter("%p{t=%p},%d", sk, sk->sk_user_data, count);
263 /* queue the transport for attention by krxiod */
264 trans = (struct rxrpc_transport *) sk->sk_user_data;
266 rxrpc_krxiod_queue_transport(trans);
268 /* wake up anyone waiting on the socket */
269 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
270 wake_up_interruptible(sk->sk_sleep);
273 } /* end rxrpc_data_ready() */
275 /*****************************************************************************/
277 * INET callback when an ICMP error packet is received
278 * - sk->err is error (EHOSTUNREACH, EPROTO or EMSGSIZE)
280 static void rxrpc_error_report(struct sock *sk)
282 struct rxrpc_transport *trans;
284 _enter("%p{t=%p}", sk, sk->sk_user_data);
286 /* queue the transport for attention by krxiod */
287 trans = (struct rxrpc_transport *) sk->sk_user_data;
289 trans->error_rcvd = 1;
290 rxrpc_krxiod_queue_transport(trans);
293 /* wake up anyone waiting on the socket */
294 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
295 wake_up_interruptible(sk->sk_sleep);
298 } /* end rxrpc_error_report() */
300 /*****************************************************************************/
302 * split a message up, allocating message records and filling them in
303 * from the contents of a socket buffer
305 static int rxrpc_incoming_msg(struct rxrpc_transport *trans,
307 struct list_head *msgq)
309 struct rxrpc_message *msg;
314 msg = kzalloc(sizeof(struct rxrpc_message), GFP_KERNEL);
316 _leave(" = -ENOMEM");
320 atomic_set(&msg->usage, 1);
321 list_add_tail(&msg->link,msgq);
323 /* dig out the Rx routing parameters */
324 if (skb_copy_bits(pkt, sizeof(struct udphdr),
325 &msg->hdr, sizeof(msg->hdr)) < 0) {
331 msg->state = RXRPC_MSG_RECEIVED;
332 skb_get_timestamp(pkt, &msg->stamp);
333 if (msg->stamp.tv_sec == 0) {
334 do_gettimeofday(&msg->stamp);
336 sock_enable_timestamp(pkt->sk);
338 msg->seq = ntohl(msg->hdr.seq);
340 /* attach the packet */
344 msg->offset = sizeof(struct udphdr) + sizeof(struct rxrpc_header);
345 msg->dsize = msg->pkt->len - msg->offset;
347 _net("Rx Received packet from %s (%08x;%08x,%1x,%d,%s,%02x,%d,%d)",
348 msg->hdr.flags & RXRPC_CLIENT_INITIATED ? "client" : "server",
349 ntohl(msg->hdr.epoch),
350 (ntohl(msg->hdr.cid) & RXRPC_CIDMASK) >> RXRPC_CIDSHIFT,
351 ntohl(msg->hdr.cid) & RXRPC_CHANNELMASK,
352 ntohl(msg->hdr.callNumber),
353 rxrpc_pkts[msg->hdr.type],
355 ntohs(msg->hdr.serviceId),
356 msg->hdr.securityIndex);
358 __RXACCT(atomic_inc(&rxrpc_message_count));
360 /* split off jumbo packets */
361 while (msg->hdr.type == RXRPC_PACKET_TYPE_DATA &&
362 msg->hdr.flags & RXRPC_JUMBO_PACKET
364 struct rxrpc_jumbo_header jumbo;
365 struct rxrpc_message *jumbomsg = msg;
367 _debug("split jumbo packet");
369 /* quick sanity check */
372 RXRPC_JUMBO_DATALEN + sizeof(struct rxrpc_jumbo_header))
374 if (msg->hdr.flags & RXRPC_LAST_PACKET)
377 /* dig out the secondary header */
378 if (skb_copy_bits(pkt, msg->offset + RXRPC_JUMBO_DATALEN,
379 &jumbo, sizeof(jumbo)) < 0)
382 /* allocate a new message record */
384 msg = kmalloc(sizeof(struct rxrpc_message), GFP_KERNEL);
388 memcpy(msg, jumbomsg, sizeof(*msg));
389 list_add_tail(&msg->link, msgq);
391 /* adjust the jumbo packet */
392 jumbomsg->dsize = RXRPC_JUMBO_DATALEN;
394 /* attach the packet here too */
397 /* adjust the parameters */
399 msg->hdr.seq = htonl(msg->seq);
400 msg->hdr.serial = htonl(ntohl(msg->hdr.serial) + 1);
401 msg->offset += RXRPC_JUMBO_DATALEN +
402 sizeof(struct rxrpc_jumbo_header);
403 msg->dsize -= RXRPC_JUMBO_DATALEN +
404 sizeof(struct rxrpc_jumbo_header);
405 msg->hdr.flags = jumbo.flags;
406 msg->hdr._rsvd = jumbo._rsvd;
408 _net("Rx Split jumbo packet from %s"
409 " (%08x;%08x,%1x,%d,%s,%02x,%d,%d)",
410 msg->hdr.flags & RXRPC_CLIENT_INITIATED ? "client" : "server",
411 ntohl(msg->hdr.epoch),
412 (ntohl(msg->hdr.cid) & RXRPC_CIDMASK) >> RXRPC_CIDSHIFT,
413 ntohl(msg->hdr.cid) & RXRPC_CHANNELMASK,
414 ntohl(msg->hdr.callNumber),
415 rxrpc_pkts[msg->hdr.type],
417 ntohs(msg->hdr.serviceId),
418 msg->hdr.securityIndex);
420 __RXACCT(atomic_inc(&rxrpc_message_count));
423 _leave(" = 0 #%d", atomic_read(&rxrpc_message_count));
427 while (!list_empty(msgq)) {
428 msg = list_entry(msgq->next, struct rxrpc_message, link);
429 list_del_init(&msg->link);
431 rxrpc_put_message(msg);
434 _leave(" = %d", ret);
436 } /* end rxrpc_incoming_msg() */
438 /*****************************************************************************/
441 * - called from krxiod in process context
443 void rxrpc_trans_receive_packet(struct rxrpc_transport *trans)
445 struct rxrpc_message *msg;
446 struct rxrpc_peer *peer;
454 _enter("%p{%d}", trans, trans->port);
457 /* deal with outstanting errors first */
458 if (trans->error_rcvd)
459 rxrpc_trans_receive_error_report(trans);
461 /* attempt to receive a packet */
462 pkt = skb_recv_datagram(trans->socket->sk, 0, 1, &ret);
464 if (ret == -EAGAIN) {
469 /* an icmp error may have occurred */
470 rxrpc_krxiod_queue_transport(trans);
471 _leave(" error %d\n", ret);
475 /* we'll probably need to checksum it (didn't call
477 if (skb_checksum_complete(pkt)) {
479 rxrpc_krxiod_queue_transport(trans);
480 _leave(" CSUM failed");
484 addr = pkt->nh.iph->saddr;
485 port = pkt->h.uh->source;
487 _net("Rx Received UDP packet from %08x:%04hu",
488 ntohl(addr), ntohs(port));
490 /* unmarshall the Rx parameters and split jumbo packets */
491 ret = rxrpc_incoming_msg(trans, pkt, &msgq);
494 rxrpc_krxiod_queue_transport(trans);
495 _leave(" bad packet");
499 BUG_ON(list_empty(&msgq));
501 msg = list_entry(msgq.next, struct rxrpc_message, link);
503 /* locate the record for the peer from which it
505 ret = rxrpc_peer_lookup(trans, addr, &peer);
507 kdebug("Rx No connections from that peer");
508 rxrpc_trans_immediate_abort(trans, msg, -EINVAL);
512 /* try and find a matching connection */
513 ret = rxrpc_connection_lookup(peer, msg, &msg->conn);
515 kdebug("Rx Unknown Connection");
516 rxrpc_trans_immediate_abort(trans, msg, -EINVAL);
517 rxrpc_put_peer(peer);
520 rxrpc_put_peer(peer);
522 /* deal with the first packet of a new call */
523 if (msg->hdr.flags & RXRPC_CLIENT_INITIATED &&
524 msg->hdr.type == RXRPC_PACKET_TYPE_DATA &&
525 ntohl(msg->hdr.seq) == 1
527 _debug("Rx New server call");
528 rxrpc_trans_receive_new_call(trans, &msgq);
532 /* deal with subsequent packet(s) of call */
533 _debug("Rx Call packet");
534 while (!list_empty(&msgq)) {
535 msg = list_entry(msgq.next, struct rxrpc_message, link);
536 list_del_init(&msg->link);
538 ret = rxrpc_conn_receive_call_packet(msg->conn, NULL, msg);
540 rxrpc_trans_immediate_abort(trans, msg, ret);
541 rxrpc_put_message(msg);
545 rxrpc_put_message(msg);
550 /* dispose of the packets */
552 while (!list_empty(&msgq)) {
553 msg = list_entry(msgq.next, struct rxrpc_message, link);
554 list_del_init(&msg->link);
556 rxrpc_put_message(msg);
563 } /* end rxrpc_trans_receive_packet() */
565 /*****************************************************************************/
567 * accept a new call from a client trying to connect to one of my services
568 * - called in process context
570 static int rxrpc_trans_receive_new_call(struct rxrpc_transport *trans,
571 struct list_head *msgq)
573 struct rxrpc_message *msg;
577 /* only bother with the first packet */
578 msg = list_entry(msgq->next, struct rxrpc_message, link);
579 list_del_init(&msg->link);
580 rxrpc_krxsecd_queue_incoming_call(msg);
581 rxrpc_put_message(msg);
586 } /* end rxrpc_trans_receive_new_call() */
588 /*****************************************************************************/
590 * perform an immediate abort without connection or call structures
592 int rxrpc_trans_immediate_abort(struct rxrpc_transport *trans,
593 struct rxrpc_message *msg,
596 struct rxrpc_header ahdr;
597 struct sockaddr_in sin;
598 struct msghdr msghdr;
603 _enter("%p,%p,%d", trans, msg, error);
605 /* don't abort an abort packet */
606 if (msg->hdr.type == RXRPC_PACKET_TYPE_ABORT) {
611 _error = htonl(-error);
613 /* set up the message to be transmitted */
614 memcpy(&ahdr, &msg->hdr, sizeof(ahdr));
615 ahdr.epoch = msg->hdr.epoch;
616 ahdr.serial = htonl(1);
618 ahdr.type = RXRPC_PACKET_TYPE_ABORT;
619 ahdr.flags = RXRPC_LAST_PACKET;
620 ahdr.flags |= ~msg->hdr.flags & RXRPC_CLIENT_INITIATED;
622 iov[0].iov_len = sizeof(ahdr);
623 iov[0].iov_base = &ahdr;
624 iov[1].iov_len = sizeof(_error);
625 iov[1].iov_base = &_error;
627 len = sizeof(ahdr) + sizeof(_error);
629 memset(&sin,0,sizeof(sin));
630 sin.sin_family = AF_INET;
631 sin.sin_port = msg->pkt->h.uh->source;
632 sin.sin_addr.s_addr = msg->pkt->nh.iph->saddr;
634 msghdr.msg_name = &sin;
635 msghdr.msg_namelen = sizeof(sin);
636 msghdr.msg_control = NULL;
637 msghdr.msg_controllen = 0;
638 msghdr.msg_flags = MSG_DONTWAIT;
640 _net("Sending message type %d of %d bytes to %08x:%d",
643 ntohl(sin.sin_addr.s_addr),
644 ntohs(sin.sin_port));
646 /* send the message */
647 ret = kernel_sendmsg(trans->socket, &msghdr, iov, 2, len);
649 _leave(" = %d", ret);
651 } /* end rxrpc_trans_immediate_abort() */
653 /*****************************************************************************/
655 * receive an ICMP error report and percolate it to all connections
656 * heading to the affected host or port
658 static void rxrpc_trans_receive_error_report(struct rxrpc_transport *trans)
660 struct rxrpc_connection *conn;
661 struct sockaddr_in sin;
662 struct rxrpc_peer *peer;
663 struct list_head connq, *_p;
664 struct errormsg emsg;
672 trans->error_rcvd = 0;
674 /* try and receive an error message */
676 msg.msg_namelen = sizeof(sin);
677 msg.msg_control = &emsg;
678 msg.msg_controllen = sizeof(emsg);
681 err = kernel_recvmsg(trans->socket, &msg, NULL, 0, 0,
682 MSG_ERRQUEUE | MSG_DONTWAIT | MSG_TRUNC);
684 if (err == -EAGAIN) {
690 printk("%s: unable to recv an error report: %d\n",
696 msg.msg_controllen = (char *) msg.msg_control - (char *) &emsg;
698 if (msg.msg_controllen < sizeof(emsg.cmsg) ||
699 msg.msg_namelen < sizeof(sin)) {
700 printk("%s: short control message"
701 " (nlen=%u clen=%Zu fl=%x)\n",
709 _net("Rx Received control message"
710 " { len=%Zu level=%u type=%u }",
712 emsg.cmsg.cmsg_level,
713 emsg.cmsg.cmsg_type);
715 if (sin.sin_family != AF_INET) {
716 printk("Rx Ignoring error report with non-INET address"
722 _net("Rx Received message pertaining to host addr=%x port=%hu",
723 ntohl(sin.sin_addr.s_addr), ntohs(sin.sin_port));
725 if (emsg.cmsg.cmsg_level != SOL_IP ||
726 emsg.cmsg.cmsg_type != IP_RECVERR) {
727 printk("Rx Ignoring unknown error report"
728 " { level=%u type=%u }",
729 emsg.cmsg.cmsg_level,
730 emsg.cmsg.cmsg_type);
734 if (msg.msg_controllen < sizeof(emsg.cmsg) + sizeof(emsg.ee)) {
735 printk("%s: short error message (%Zu)\n",
736 __FUNCTION__, msg.msg_controllen);
743 switch (emsg.ee.ee_origin) {
744 case SO_EE_ORIGIN_ICMP:
746 switch (emsg.ee.ee_type) {
747 case ICMP_DEST_UNREACH:
748 switch (emsg.ee.ee_code) {
749 case ICMP_NET_UNREACH:
750 _net("Rx Received ICMP Network Unreachable");
754 case ICMP_HOST_UNREACH:
755 _net("Rx Received ICMP Host Unreachable");
759 case ICMP_PORT_UNREACH:
760 _net("Rx Received ICMP Port Unreachable");
763 case ICMP_NET_UNKNOWN:
764 _net("Rx Received ICMP Unknown Network");
768 case ICMP_HOST_UNKNOWN:
769 _net("Rx Received ICMP Unknown Host");
774 _net("Rx Received ICMP DestUnreach { code=%u }",
776 err = emsg.ee.ee_errno;
781 case ICMP_TIME_EXCEEDED:
782 _net("Rx Received ICMP TTL Exceeded");
783 err = emsg.ee.ee_errno;
787 _proto("Rx Received ICMP error { type=%u code=%u }",
788 emsg.ee.ee_type, emsg.ee.ee_code);
789 err = emsg.ee.ee_errno;
794 case SO_EE_ORIGIN_LOCAL:
795 _proto("Rx Received local error { error=%d }",
798 err = emsg.ee.ee_errno;
801 case SO_EE_ORIGIN_NONE:
802 case SO_EE_ORIGIN_ICMP6:
804 _proto("Rx Received error report { orig=%u }",
807 err = emsg.ee.ee_errno;
811 /* find all the connections between this transport and the
812 * affected destination */
813 INIT_LIST_HEAD(&connq);
815 if (rxrpc_peer_lookup(trans, sin.sin_addr.s_addr,
817 read_lock(&peer->conn_lock);
818 list_for_each(_p, &peer->conn_active) {
819 conn = list_entry(_p, struct rxrpc_connection,
821 if (port && conn->addr.sin_port != port)
823 if (!list_empty(&conn->err_link))
826 rxrpc_get_connection(conn);
827 list_add_tail(&conn->err_link, &connq);
829 read_unlock(&peer->conn_lock);
831 /* service all those connections */
832 while (!list_empty(&connq)) {
833 conn = list_entry(connq.next,
834 struct rxrpc_connection,
836 list_del(&conn->err_link);
838 rxrpc_conn_handle_error(conn, local, err);
840 rxrpc_put_connection(conn);
843 rxrpc_put_peer(peer);
849 } /* end rxrpc_trans_receive_error_report() */