[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 <linux/slab.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.fl4_dst = peer->srx.transport.sin.sin_addr.s_addr;
		fl.fl4_src = 0;
		fl.fl4_tos = 0;
		/* assume AFS.CM talking to AFS.FS */
		fl.fl_ip_sport = htons(7001);
		fl.fl_ip_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->dst);
	dst_release(&rt->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>
5a0e3ad6	19	#include <linux/slab.h>
17926a79 DH	20	#include <net/sock.h>
	21	#include <net/af_rxrpc.h>
	22	#include <net/ip.h>
224711df	23	#include <net/route.h>
17926a79 DH	24	#include "ar-internal.h"
	25
	26	static LIST_HEAD(rxrpc_peers);
	27	static DEFINE_RWLOCK(rxrpc_peer_lock);
	28	static DECLARE_WAIT_QUEUE_HEAD(rxrpc_peer_wq);
	29
	30	static void rxrpc_destroy_peer(struct work_struct *work);
	31
224711df DH	32	/*
	33	* assess the MTU size for the network interface through which this peer is
	34	* reached
	35	*/
	36	static void rxrpc_assess_MTU_size(struct rxrpc_peer *peer)
	37	{
	38	struct rtable *rt;
	39	struct flowi fl;
	40	int ret;
	41
	42	peer->if_mtu = 1500;
	43
	44	memset(&fl, 0, sizeof(fl));
	45
	46	switch (peer->srx.transport.family) {
	47	case AF_INET:
	48	fl.oif = 0;
	49	fl.proto = IPPROTO_UDP,
5811662b CG	50	fl.fl4_dst = peer->srx.transport.sin.sin_addr.s_addr;
	51	fl.fl4_src = 0;
	52	fl.fl4_tos = 0;
224711df	53	/* assume AFS.CM talking to AFS.FS */
5811662b CG	54	fl.fl_ip_sport = htons(7001);
5811662b CG	55	fl.fl_ip_dport = htons(7000);
224711df DH	56	break;
	57	default:
	58	BUG();
	59	}
	60
f206351a	61	ret = ip_route_output_key(&init_net, &rt, &fl);
224711df	62	if (ret < 0) {
a6a62b69	63	_leave(" [route err %d]", ret);
224711df DH	64	return;
	65	}
	66
d8d1f30b CG	67	peer->if_mtu = dst_mtu(&rt->dst);
d8d1f30b CG	68	dst_release(&rt->dst);
224711df	69
a6a62b69	70	_leave(" [if_mtu %u]", peer->if_mtu);
224711df DH	71	}
224711df DH	72
17926a79 DH	73	/*
	74	* allocate a new peer
	75	*/
	76	static struct rxrpc_peer rxrpc_alloc_peer(struct sockaddr_rxrpc srx,
	77	gfp_t gfp)
	78	{
	79	struct rxrpc_peer *peer;
	80
	81	_enter("");
	82
	83	peer = kzalloc(sizeof(struct rxrpc_peer), gfp);
	84	if (peer) {
	85	INIT_WORK(&peer->destroyer, &rxrpc_destroy_peer);
	86	INIT_LIST_HEAD(&peer->link);
	87	INIT_LIST_HEAD(&peer->error_targets);
	88	spin_lock_init(&peer->lock);
	89	atomic_set(&peer->usage, 1);
	90	peer->debug_id = atomic_inc_return(&rxrpc_debug_id);
	91	memcpy(&peer->srx, srx, sizeof(*srx));
	92
224711df DH	93	rxrpc_assess_MTU_size(peer);
224711df DH	94	peer->mtu = peer->if_mtu;
17926a79 DH	95
	96	if (srx->transport.family == AF_INET) {
	97	peer->hdrsize = sizeof(struct iphdr);
	98	switch (srx->transport_type) {
	99	case SOCK_DGRAM:
	100	peer->hdrsize += sizeof(struct udphdr);
	101	break;
	102	default:
	103	BUG();
	104	break;
	105	}
	106	} else {
	107	BUG();
	108	}
	109
	110	peer->hdrsize += sizeof(struct rxrpc_header);
	111	peer->maxdata = peer->mtu - peer->hdrsize;
	112	}
	113
	114	_leave(" = %p", peer);
	115	return peer;
	116	}
	117
	118	/*
	119	* obtain a remote transport endpoint for the specified address
	120	*/
	121	struct rxrpc_peer rxrpc_get_peer(struct sockaddr_rxrpc srx, gfp_t gfp)
	122	{
	123	struct rxrpc_peer peer, candidate;
	124	const char *new = "old";
	125	int usage;
	126
21454aaa	127	_enter("{%d,%d,%pI4+%hu}",
17926a79 DH	128	srx->transport_type,
17926a79 DH	129	srx->transport_len,
21454aaa	130	&srx->transport.sin.sin_addr,
17926a79 DH	131	ntohs(srx->transport.sin.sin_port));
	132
	133	/* search the peer list first */
	134	read_lock_bh(&rxrpc_peer_lock);
	135	list_for_each_entry(peer, &rxrpc_peers, link) {
	136	_debug("check PEER %d { u=%d t=%d l=%d }",
	137	peer->debug_id,
	138	atomic_read(&peer->usage),
	139	peer->srx.transport_type,
	140	peer->srx.transport_len);
	141
	142	if (atomic_read(&peer->usage) > 0 &&
	143	peer->srx.transport_type == srx->transport_type &&
	144	peer->srx.transport_len == srx->transport_len &&
	145	memcmp(&peer->srx.transport,
	146	&srx->transport,
	147	srx->transport_len) == 0)
	148	goto found_extant_peer;
	149	}
	150	read_unlock_bh(&rxrpc_peer_lock);
	151
	152	/* not yet present - create a candidate for a new record and then
	153	* redo the search */
	154	candidate = rxrpc_alloc_peer(srx, gfp);
	155	if (!candidate) {
	156	_leave(" = -ENOMEM");
	157	return ERR_PTR(-ENOMEM);
	158	}
	159
	160	write_lock_bh(&rxrpc_peer_lock);
	161
	162	list_for_each_entry(peer, &rxrpc_peers, link) {
	163	if (atomic_read(&peer->usage) > 0 &&
	164	peer->srx.transport_type == srx->transport_type &&
	165	peer->srx.transport_len == srx->transport_len &&
	166	memcmp(&peer->srx.transport,
	167	&srx->transport,
	168	srx->transport_len) == 0)
	169	goto found_extant_second;
	170	}
	171
	172	/* we can now add the new candidate to the list */
	173	peer = candidate;
	174	candidate = NULL;
	175
	176	list_add_tail(&peer->link, &rxrpc_peers);
	177	write_unlock_bh(&rxrpc_peer_lock);
	178	new = "new";
	179
	180	success:
21454aaa	181	_net("PEER %s %d {%d,%u,%pI4+%hu}",
17926a79 DH	182	new,
	183	peer->debug_id,
	184	peer->srx.transport_type,
	185	peer->srx.transport.family,
21454aaa	186	&peer->srx.transport.sin.sin_addr,
17926a79 DH	187	ntohs(peer->srx.transport.sin.sin_port));
	188
	189	_leave(" = %p {u=%d}", peer, atomic_read(&peer->usage));
	190	return peer;
	191
	192	/* we found the peer in the list immediately */
	193	found_extant_peer:
	194	usage = atomic_inc_return(&peer->usage);
	195	read_unlock_bh(&rxrpc_peer_lock);
	196	goto success;
	197
	198	/* we found the peer on the second time through the list */
	199	found_extant_second:
	200	usage = atomic_inc_return(&peer->usage);
	201	write_unlock_bh(&rxrpc_peer_lock);
	202	kfree(candidate);
	203	goto success;
	204	}
	205
	206	/*
	207	* find the peer associated with a packet
	208	*/
	209	struct rxrpc_peer rxrpc_find_peer(struct rxrpc_local local,
	210	__be32 addr, __be16 port)
	211	{
	212	struct rxrpc_peer *peer;
	213
	214	_enter("");
	215
	216	/* search the peer list */
	217	read_lock_bh(&rxrpc_peer_lock);
	218
	219	if (local->srx.transport.family == AF_INET &&
	220	local->srx.transport_type == SOCK_DGRAM
	221	) {
	222	list_for_each_entry(peer, &rxrpc_peers, link) {
	223	if (atomic_read(&peer->usage) > 0 &&
	224	peer->srx.transport_type == SOCK_DGRAM &&
	225	peer->srx.transport.family == AF_INET &&
	226	peer->srx.transport.sin.sin_port == port &&
	227	peer->srx.transport.sin.sin_addr.s_addr == addr)
	228	goto found_UDP_peer;
	229	}
	230
	231	goto new_UDP_peer;
	232	}
	233
	234	read_unlock_bh(&rxrpc_peer_lock);
	235	_leave(" = -EAFNOSUPPORT");
	236	return ERR_PTR(-EAFNOSUPPORT);
	237
	238	found_UDP_peer:
	239	_net("Rx UDP DGRAM from peer %d", peer->debug_id);
	240	atomic_inc(&peer->usage);
	241	read_unlock_bh(&rxrpc_peer_lock);
	242	_leave(" = %p", peer);
	243	return peer;
	244
	245	new_UDP_peer:
	246	_net("Rx UDP DGRAM from NEW peer %d", peer->debug_id);
	247	read_unlock_bh(&rxrpc_peer_lock);
	248	_leave(" = -EBUSY [new]");
	249	return ERR_PTR(-EBUSY);
	250	}
251
252	/*
253	* release a remote transport endpoint
254	*/
255	void rxrpc_put_peer(struct rxrpc_peer *peer)
256	{
257	_enter("%p{u=%d}", peer, atomic_read(&peer->usage));
258
259	ASSERTCMP(atomic_read(&peer->usage), >, 0);
260
261	if (likely(!atomic_dec_and_test(&peer->usage))) {
262	_leave(" [in use]");
263	return;
264	}
265
651350d1	266	rxrpc_queue_work(&peer->destroyer);
17926a79 DH	267	_leave("");
	268	}
	269
	270	/*
	271	* destroy a remote transport endpoint
	272	*/
	273	static void rxrpc_destroy_peer(struct work_struct *work)
	274	{
	275	struct rxrpc_peer *peer =
	276	container_of(work, struct rxrpc_peer, destroyer);
	277
	278	_enter("%p{%d}", peer, atomic_read(&peer->usage));
	279
	280	write_lock_bh(&rxrpc_peer_lock);
	281	list_del(&peer->link);
	282	write_unlock_bh(&rxrpc_peer_lock);
	283
	284	_net("DESTROY PEER %d", peer->debug_id);
	285	kfree(peer);
	286
	287	if (list_empty(&rxrpc_peers))
	288	wake_up_all(&rxrpc_peer_wq);
	289	_leave("");
	290	}
	291
	292	/*
	293	* preemptively destroy all the peer records from a transport endpoint rather
	294	* than waiting for them to time out
	295	*/
	296	void __exit rxrpc_destroy_all_peers(void)
	297	{
	298	DECLARE_WAITQUEUE(myself,current);
	299
	300	_enter("");
	301
	302	/* we simply have to wait for them to go away */
	303	if (!list_empty(&rxrpc_peers)) {
	304	set_current_state(TASK_UNINTERRUPTIBLE);
	305	add_wait_queue(&rxrpc_peer_wq, &myself);
	306
	307	while (!list_empty(&rxrpc_peers)) {
	308	schedule();
	309	set_current_state(TASK_UNINTERRUPTIBLE);
	310	}
	311
	312	remove_wait_queue(&rxrpc_peer_wq, &myself);
	313	set_current_state(TASK_RUNNING);
	314	}
	315
	316	_leave("");
	317	}