[net-next-2.6.git] / net / rxrpc / ar-peer.c

/* RxRPC remote transport endpoint management
 *
 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
 * Written by David Howells (dhowells@redhat.com)
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version
 * 2 of the License, or (at your option) any later version.
 */

#include <linux/module.h>
#include <linux/net.h>
#include <linux/skbuff.h>
#include <linux/udp.h>
#include <linux/in.h>
#include <linux/in6.h>
#include <linux/icmp.h>
#include <net/sock.h>
#include <net/af_rxrpc.h>
#include <net/ip.h>
#include <net/route.h>
#include "ar-internal.h"

static LIST_HEAD(rxrpc_peers);
static DEFINE_RWLOCK(rxrpc_peer_lock);
static DECLARE_WAIT_QUEUE_HEAD(rxrpc_peer_wq);

static void rxrpc_destroy_peer(struct work_struct *work);

/*
 * assess the MTU size for the network interface through which this peer is
 * reached
 */
static void rxrpc_assess_MTU_size(struct rxrpc_peer *peer)
{
	struct rtable *rt;
	struct flowi fl;
	int ret;

	peer->if_mtu = 1500;

	memset(&fl, 0, sizeof(fl));

	switch (peer->srx.transport.family) {
	case AF_INET:
		fl.oif = 0;
		fl.proto = IPPROTO_UDP,
		fl.nl_u.ip4_u.saddr = 0;
		fl.nl_u.ip4_u.daddr = peer->srx.transport.sin.sin_addr.s_addr;
		fl.nl_u.ip4_u.tos = 0;
		/* assume AFS.CM talking to AFS.FS */
		fl.uli_u.ports.sport = htons(7001);
		fl.uli_u.ports.dport = htons(7000);
		break;
	default:
		BUG();
	}

	ret = ip_route_output_key(&init_net, &rt, &fl);
	if (ret < 0) {
		_leave(" [route err %d]", ret);
		return;
	}

	peer->if_mtu = dst_mtu(&rt->u.dst);
	dst_release(&rt->u.dst);

	_leave(" [if_mtu %u]", peer->if_mtu);
}

/*
 * allocate a new peer
 */
static struct rxrpc_peer *rxrpc_alloc_peer(struct sockaddr_rxrpc *srx,
					   gfp_t gfp)
{
	struct rxrpc_peer *peer;

	_enter("");

	peer = kzalloc(sizeof(struct rxrpc_peer), gfp);
	if (peer) {
		INIT_WORK(&peer->destroyer, &rxrpc_destroy_peer);
		INIT_LIST_HEAD(&peer->link);
		INIT_LIST_HEAD(&peer->error_targets);
		spin_lock_init(&peer->lock);
		atomic_set(&peer->usage, 1);
		peer->debug_id = atomic_inc_return(&rxrpc_debug_id);
		memcpy(&peer->srx, srx, sizeof(*srx));

		rxrpc_assess_MTU_size(peer);
		peer->mtu = peer->if_mtu;

		if (srx->transport.family == AF_INET) {
			peer->hdrsize = sizeof(struct iphdr);
			switch (srx->transport_type) {
			case SOCK_DGRAM:
				peer->hdrsize += sizeof(struct udphdr);
				break;
			default:
				BUG();
				break;
			}
		} else {
			BUG();
		}

		peer->hdrsize += sizeof(struct rxrpc_header);
		peer->maxdata = peer->mtu - peer->hdrsize;
	}

	_leave(" = %p", peer);
	return peer;
}

/*
 * obtain a remote transport endpoint for the specified address
 */
struct rxrpc_peer *rxrpc_get_peer(struct sockaddr_rxrpc *srx, gfp_t gfp)
{
	struct rxrpc_peer *peer, *candidate;
	const char *new = "old";
	int usage;

	_enter("{%d,%d,%pI4+%hu}",
	       srx->transport_type,
	       srx->transport_len,
	       &srx->transport.sin.sin_addr,
	       ntohs(srx->transport.sin.sin_port));

	/* search the peer list first */
	read_lock_bh(&rxrpc_peer_lock);
	list_for_each_entry(peer, &rxrpc_peers, link) {
		_debug("check PEER %d { u=%d t=%d l=%d }",
		       peer->debug_id,
		       atomic_read(&peer->usage),
		       peer->srx.transport_type,
		       peer->srx.transport_len);

		if (atomic_read(&peer->usage) > 0 &&
		    peer->srx.transport_type == srx->transport_type &&
		    peer->srx.transport_len == srx->transport_len &&
		    memcmp(&peer->srx.transport,
			   &srx->transport,
			   srx->transport_len) == 0)
			goto found_extant_peer;
	}
	read_unlock_bh(&rxrpc_peer_lock);

	/* not yet present - create a candidate for a new record and then
	 * redo the search */
	candidate = rxrpc_alloc_peer(srx, gfp);
	if (!candidate) {
		_leave(" = -ENOMEM");
		return ERR_PTR(-ENOMEM);
	}

	write_lock_bh(&rxrpc_peer_lock);

	list_for_each_entry(peer, &rxrpc_peers, link) {
		if (atomic_read(&peer->usage) > 0 &&
		    peer->srx.transport_type == srx->transport_type &&
		    peer->srx.transport_len == srx->transport_len &&
		    memcmp(&peer->srx.transport,
			   &srx->transport,
			   srx->transport_len) == 0)
			goto found_extant_second;
	}

	/* we can now add the new candidate to the list */
	peer = candidate;
	candidate = NULL;

	list_add_tail(&peer->link, &rxrpc_peers);
	write_unlock_bh(&rxrpc_peer_lock);
	new = "new";

success:
	_net("PEER %s %d {%d,%u,%pI4+%hu}",
	     new,
	     peer->debug_id,
	     peer->srx.transport_type,
	     peer->srx.transport.family,
	     &peer->srx.transport.sin.sin_addr,
	     ntohs(peer->srx.transport.sin.sin_port));

	_leave(" = %p {u=%d}", peer, atomic_read(&peer->usage));
	return peer;

	/* we found the peer in the list immediately */
found_extant_peer:
	usage = atomic_inc_return(&peer->usage);
	read_unlock_bh(&rxrpc_peer_lock);
	goto success;

	/* we found the peer on the second time through the list */
found_extant_second:
	usage = atomic_inc_return(&peer->usage);
	write_unlock_bh(&rxrpc_peer_lock);
	kfree(candidate);
	goto success;
}

/*
 * find the peer associated with a packet
 */
struct rxrpc_peer *rxrpc_find_peer(struct rxrpc_local *local,
				   __be32 addr, __be16 port)
{
	struct rxrpc_peer *peer;

	_enter("");

	/* search the peer list */
	read_lock_bh(&rxrpc_peer_lock);

	if (local->srx.transport.family == AF_INET &&
	    local->srx.transport_type == SOCK_DGRAM
	    ) {
		list_for_each_entry(peer, &rxrpc_peers, link) {
			if (atomic_read(&peer->usage) > 0 &&
			    peer->srx.transport_type == SOCK_DGRAM &&
			    peer->srx.transport.family == AF_INET &&
			    peer->srx.transport.sin.sin_port == port &&
			    peer->srx.transport.sin.sin_addr.s_addr == addr)
				goto found_UDP_peer;
		}

		goto new_UDP_peer;
	}

	read_unlock_bh(&rxrpc_peer_lock);
	_leave(" = -EAFNOSUPPORT");
	return ERR_PTR(-EAFNOSUPPORT);

found_UDP_peer:
	_net("Rx UDP DGRAM from peer %d", peer->debug_id);
	atomic_inc(&peer->usage);
	read_unlock_bh(&rxrpc_peer_lock);
	_leave(" = %p", peer);
	return peer;

new_UDP_peer:
	_net("Rx UDP DGRAM from NEW peer %d", peer->debug_id);
	read_unlock_bh(&rxrpc_peer_lock);
	_leave(" = -EBUSY [new]");
	return ERR_PTR(-EBUSY);
}

/*
 * release a remote transport endpoint
 */
void rxrpc_put_peer(struct rxrpc_peer *peer)
{
	_enter("%p{u=%d}", peer, atomic_read(&peer->usage));

	ASSERTCMP(atomic_read(&peer->usage), >, 0);

	if (likely(!atomic_dec_and_test(&peer->usage))) {
		_leave(" [in use]");
		return;
	}

	rxrpc_queue_work(&peer->destroyer);
	_leave("");
}

/*
 * destroy a remote transport endpoint
 */
static void rxrpc_destroy_peer(struct work_struct *work)
{
	struct rxrpc_peer *peer =
		container_of(work, struct rxrpc_peer, destroyer);

	_enter("%p{%d}", peer, atomic_read(&peer->usage));

	write_lock_bh(&rxrpc_peer_lock);
	list_del(&peer->link);
	write_unlock_bh(&rxrpc_peer_lock);

	_net("DESTROY PEER %d", peer->debug_id);
	kfree(peer);

	if (list_empty(&rxrpc_peers))
		wake_up_all(&rxrpc_peer_wq);
	_leave("");
}

/*
 * preemptively destroy all the peer records from a transport endpoint rather
 * than waiting for them to time out
 */
void __exit rxrpc_destroy_all_peers(void)
{
	DECLARE_WAITQUEUE(myself,current);

	_enter("");

	/* we simply have to wait for them to go away */
	if (!list_empty(&rxrpc_peers)) {
		set_current_state(TASK_UNINTERRUPTIBLE);
		add_wait_queue(&rxrpc_peer_wq, &myself);

		while (!list_empty(&rxrpc_peers)) {
			schedule();
			set_current_state(TASK_UNINTERRUPTIBLE);
		}

		remove_wait_queue(&rxrpc_peer_wq, &myself);
		set_current_state(TASK_RUNNING);
	}

	_leave("");
}
Commit	Line	Data
17926a79 DH	1	/* RxRPC remote transport endpoint management
	2	*
	3	* Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
	4	* Written by David Howells (dhowells@redhat.com)
	5	*
	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.
	10	*/
	11
	12	#include <linux/module.h>
	13	#include <linux/net.h>
	14	#include <linux/skbuff.h>
	15	#include <linux/udp.h>
	16	#include <linux/in.h>
	17	#include <linux/in6.h>
	18	#include <linux/icmp.h>
	19	#include <net/sock.h>
	20	#include <net/af_rxrpc.h>
	21	#include <net/ip.h>
224711df	22	#include <net/route.h>
17926a79 DH	23	#include "ar-internal.h"
	24
	25	static LIST_HEAD(rxrpc_peers);
	26	static DEFINE_RWLOCK(rxrpc_peer_lock);
	27	static DECLARE_WAIT_QUEUE_HEAD(rxrpc_peer_wq);
	28
	29	static void rxrpc_destroy_peer(struct work_struct *work);
	30
224711df DH	31	/*
	32	* assess the MTU size for the network interface through which this peer is
	33	* reached
	34	*/
	35	static void rxrpc_assess_MTU_size(struct rxrpc_peer *peer)
	36	{
	37	struct rtable *rt;
	38	struct flowi fl;
	39	int ret;
	40
	41	peer->if_mtu = 1500;
	42
	43	memset(&fl, 0, sizeof(fl));
	44
	45	switch (peer->srx.transport.family) {
	46	case AF_INET:
	47	fl.oif = 0;
	48	fl.proto = IPPROTO_UDP,
	49	fl.nl_u.ip4_u.saddr = 0;
	50	fl.nl_u.ip4_u.daddr = peer->srx.transport.sin.sin_addr.s_addr;
	51	fl.nl_u.ip4_u.tos = 0;
	52	/* assume AFS.CM talking to AFS.FS */
	53	fl.uli_u.ports.sport = htons(7001);
	54	fl.uli_u.ports.dport = htons(7000);
	55	break;
	56	default:
	57	BUG();
	58	}
	59
f206351a	60	ret = ip_route_output_key(&init_net, &rt, &fl);
224711df	61	if (ret < 0) {
a6a62b69	62	_leave(" [route err %d]", ret);
224711df DH	63	return;
	64	}
	65
	66	peer->if_mtu = dst_mtu(&rt->u.dst);
	67	dst_release(&rt->u.dst);
	68
a6a62b69	69	_leave(" [if_mtu %u]", peer->if_mtu);
224711df DH	70	}
224711df DH	71
17926a79 DH	72	/*
	73	* allocate a new peer
	74	*/
	75	static struct rxrpc_peer rxrpc_alloc_peer(struct sockaddr_rxrpc srx,
	76	gfp_t gfp)
	77	{
	78	struct rxrpc_peer *peer;
	79
	80	_enter("");
	81
	82	peer = kzalloc(sizeof(struct rxrpc_peer), gfp);
	83	if (peer) {
	84	INIT_WORK(&peer->destroyer, &rxrpc_destroy_peer);
	85	INIT_LIST_HEAD(&peer->link);
	86	INIT_LIST_HEAD(&peer->error_targets);
	87	spin_lock_init(&peer->lock);
	88	atomic_set(&peer->usage, 1);
	89	peer->debug_id = atomic_inc_return(&rxrpc_debug_id);
	90	memcpy(&peer->srx, srx, sizeof(*srx));
	91
224711df DH	92	rxrpc_assess_MTU_size(peer);
224711df DH	93	peer->mtu = peer->if_mtu;
17926a79 DH	94
	95	if (srx->transport.family == AF_INET) {
	96	peer->hdrsize = sizeof(struct iphdr);
	97	switch (srx->transport_type) {
	98	case SOCK_DGRAM:
	99	peer->hdrsize += sizeof(struct udphdr);
	100	break;
	101	default:
	102	BUG();
	103	break;
	104	}
	105	} else {
	106	BUG();
	107	}
	108
	109	peer->hdrsize += sizeof(struct rxrpc_header);
	110	peer->maxdata = peer->mtu - peer->hdrsize;
	111	}
	112
	113	_leave(" = %p", peer);
	114	return peer;
	115	}
	116
	117	/*
	118	* obtain a remote transport endpoint for the specified address
	119	*/
	120	struct rxrpc_peer rxrpc_get_peer(struct sockaddr_rxrpc srx, gfp_t gfp)
	121	{
	122	struct rxrpc_peer peer, candidate;
	123	const char *new = "old";
	124	int usage;
	125
21454aaa	126	_enter("{%d,%d,%pI4+%hu}",
17926a79 DH	127	srx->transport_type,
17926a79 DH	128	srx->transport_len,
21454aaa	129	&srx->transport.sin.sin_addr,
17926a79 DH	130	ntohs(srx->transport.sin.sin_port));
	131
	132	/* search the peer list first */
	133	read_lock_bh(&rxrpc_peer_lock);
	134	list_for_each_entry(peer, &rxrpc_peers, link) {
	135	_debug("check PEER %d { u=%d t=%d l=%d }",
	136	peer->debug_id,
	137	atomic_read(&peer->usage),
	138	peer->srx.transport_type,
	139	peer->srx.transport_len);
	140
	141	if (atomic_read(&peer->usage) > 0 &&
	142	peer->srx.transport_type == srx->transport_type &&
	143	peer->srx.transport_len == srx->transport_len &&
	144	memcmp(&peer->srx.transport,
	145	&srx->transport,
	146	srx->transport_len) == 0)
	147	goto found_extant_peer;
	148	}
	149	read_unlock_bh(&rxrpc_peer_lock);
	150
	151	/* not yet present - create a candidate for a new record and then
	152	* redo the search */
	153	candidate = rxrpc_alloc_peer(srx, gfp);
	154	if (!candidate) {
	155	_leave(" = -ENOMEM");
	156	return ERR_PTR(-ENOMEM);
	157	}
	158
	159	write_lock_bh(&rxrpc_peer_lock);
	160
	161	list_for_each_entry(peer, &rxrpc_peers, link) {
	162	if (atomic_read(&peer->usage) > 0 &&
	163	peer->srx.transport_type == srx->transport_type &&
	164	peer->srx.transport_len == srx->transport_len &&
	165	memcmp(&peer->srx.transport,
	166	&srx->transport,
	167	srx->transport_len) == 0)
	168	goto found_extant_second;
	169	}
	170
	171	/* we can now add the new candidate to the list */
	172	peer = candidate;
	173	candidate = NULL;
	174
	175	list_add_tail(&peer->link, &rxrpc_peers);
	176	write_unlock_bh(&rxrpc_peer_lock);
	177	new = "new";
	178
	179	success:
21454aaa	180	_net("PEER %s %d {%d,%u,%pI4+%hu}",
17926a79 DH	181	new,
	182	peer->debug_id,
	183	peer->srx.transport_type,
	184	peer->srx.transport.family,
21454aaa	185	&peer->srx.transport.sin.sin_addr,
17926a79 DH	186	ntohs(peer->srx.transport.sin.sin_port));
	187
	188	_leave(" = %p {u=%d}", peer, atomic_read(&peer->usage));
	189	return peer;
	190
	191	/* we found the peer in the list immediately */
	192	found_extant_peer:
	193	usage = atomic_inc_return(&peer->usage);
	194	read_unlock_bh(&rxrpc_peer_lock);
	195	goto success;
	196
	197	/* we found the peer on the second time through the list */
	198	found_extant_second:
	199	usage = atomic_inc_return(&peer->usage);
	200	write_unlock_bh(&rxrpc_peer_lock);
	201	kfree(candidate);
	202	goto success;
	203	}
	204
	205	/*
	206	* find the peer associated with a packet
	207	*/
	208	struct rxrpc_peer rxrpc_find_peer(struct rxrpc_local local,
	209	__be32 addr, __be16 port)
	210	{
	211	struct rxrpc_peer *peer;
	212
	213	_enter("");
	214
	215	/* search the peer list */
	216	read_lock_bh(&rxrpc_peer_lock);
	217
	218	if (local->srx.transport.family == AF_INET &&
	219	local->srx.transport_type == SOCK_DGRAM
	220	) {
	221	list_for_each_entry(peer, &rxrpc_peers, link) {
	222	if (atomic_read(&peer->usage) > 0 &&
	223	peer->srx.transport_type == SOCK_DGRAM &&
	224	peer->srx.transport.family == AF_INET &&
	225	peer->srx.transport.sin.sin_port == port &&
	226	peer->srx.transport.sin.sin_addr.s_addr == addr)
	227	goto found_UDP_peer;
	228	}
	229
	230	goto new_UDP_peer;
	231	}
	232
	233	read_unlock_bh(&rxrpc_peer_lock);
	234	_leave(" = -EAFNOSUPPORT");
	235	return ERR_PTR(-EAFNOSUPPORT);
	236
	237	found_UDP_peer:
	238	_net("Rx UDP DGRAM from peer %d", peer->debug_id);
	239	atomic_inc(&peer->usage);
	240	read_unlock_bh(&rxrpc_peer_lock);
	241	_leave(" = %p", peer);
	242	return peer;
	243
	244	new_UDP_peer:
	245	_net("Rx UDP DGRAM from NEW peer %d", peer->debug_id);
	246	read_unlock_bh(&rxrpc_peer_lock);
	247	_leave(" = -EBUSY [new]");
	248	return ERR_PTR(-EBUSY);
	249	}
250
251	/*
252	* release a remote transport endpoint
253	*/
254	void rxrpc_put_peer(struct rxrpc_peer *peer)
255	{
256	_enter("%p{u=%d}", peer, atomic_read(&peer->usage));
257
258	ASSERTCMP(atomic_read(&peer->usage), >, 0);
259
260	if (likely(!atomic_dec_and_test(&peer->usage))) {
261	_leave(" [in use]");
262	return;
263	}
264
651350d1	265	rxrpc_queue_work(&peer->destroyer);
17926a79 DH	266	_leave("");
	267	}
	268
	269	/*
	270	* destroy a remote transport endpoint
	271	*/
	272	static void rxrpc_destroy_peer(struct work_struct *work)
	273	{
	274	struct rxrpc_peer *peer =
	275	container_of(work, struct rxrpc_peer, destroyer);
	276
	277	_enter("%p{%d}", peer, atomic_read(&peer->usage));
	278
	279	write_lock_bh(&rxrpc_peer_lock);
	280	list_del(&peer->link);
	281	write_unlock_bh(&rxrpc_peer_lock);
	282
	283	_net("DESTROY PEER %d", peer->debug_id);
	284	kfree(peer);
	285
	286	if (list_empty(&rxrpc_peers))
	287	wake_up_all(&rxrpc_peer_wq);
	288	_leave("");
	289	}
	290
	291	/*
	292	* preemptively destroy all the peer records from a transport endpoint rather
	293	* than waiting for them to time out
	294	*/
	295	void __exit rxrpc_destroy_all_peers(void)
	296	{
	297	DECLARE_WAITQUEUE(myself,current);
	298
	299	_enter("");
	300
	301	/* we simply have to wait for them to go away */
	302	if (!list_empty(&rxrpc_peers)) {
	303	set_current_state(TASK_UNINTERRUPTIBLE);
	304	add_wait_queue(&rxrpc_peer_wq, &myself);
	305
	306	while (!list_empty(&rxrpc_peers)) {
	307	schedule();
	308	set_current_state(TASK_UNINTERRUPTIBLE);
	309	}
	310
	311	remove_wait_queue(&rxrpc_peer_wq, &myself);
	312	set_current_state(TASK_RUNNING);
	313	}
	314
	315	_leave("");
	316	}