[linux-2.6-block.git] / net / ipv4 / inetpeer.c

/*
 *		INETPEER - A storage for permanent information about peers
 *
 *  This source is covered by the GNU GPL, the same as all kernel sources.
 *
 *  Authors:	Andrey V. Savochkin <saw@msu.ru>
 */

#include <linux/cache.h>
#include <linux/module.h>
#include <linux/types.h>
#include <linux/slab.h>
#include <linux/interrupt.h>
#include <linux/spinlock.h>
#include <linux/random.h>
#include <linux/timer.h>
#include <linux/time.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/net.h>
#include <linux/workqueue.h>
#include <net/ip.h>
#include <net/inetpeer.h>
#include <net/secure_seq.h>

/*
 *  Theory of operations.
 *  We keep one entry for each peer IP address.  The nodes contains long-living
 *  information about the peer which doesn't depend on routes.
 *
 *  Nodes are removed only when reference counter goes to 0.
 *  When it's happened the node may be removed when a sufficient amount of
 *  time has been passed since its last use.  The less-recently-used entry can
 *  also be removed if the pool is overloaded i.e. if the total amount of
 *  entries is greater-or-equal than the threshold.
 *
 *  Node pool is organised as an RB tree.
 *  Such an implementation has been chosen not just for fun.  It's a way to
 *  prevent easy and efficient DoS attacks by creating hash collisions.  A huge
 *  amount of long living nodes in a single hash slot would significantly delay
 *  lookups performed with disabled BHs.
 *
 *  Serialisation issues.
 *  1.  Nodes may appear in the tree only with the pool lock held.
 *  2.  Nodes may disappear from the tree only with the pool lock held
 *      AND reference count being 0.
 *  3.  Global variable peer_total is modified under the pool lock.
 *  4.  struct inet_peer fields modification:
 *		rb_node: pool lock
 *		refcnt: atomically against modifications on other CPU;
 *		   usually under some other lock to prevent node disappearing
 *		daddr: unchangeable
 */

static struct kmem_cache *peer_cachep __ro_after_init;

void inet_peer_base_init(struct inet_peer_base *bp)
{
	bp->rb_root = RB_ROOT;
	seqlock_init(&bp->lock);
	bp->total = 0;
}
EXPORT_SYMBOL_GPL(inet_peer_base_init);

#define PEER_MAX_GC 32

/* Exported for sysctl_net_ipv4.  */
int inet_peer_threshold __read_mostly;	/* start to throw entries more
					 * aggressively at this stage */
int inet_peer_minttl __read_mostly = 120 * HZ;	/* TTL under high load: 120 sec */
int inet_peer_maxttl __read_mostly = 10 * 60 * HZ;	/* usual time to live: 10 min */

/* Called from ip_output.c:ip_init  */
void __init inet_initpeers(void)
{
	u64 nr_entries;

	 /* 1% of physical memory */
	nr_entries = div64_ul((u64)totalram_pages() << PAGE_SHIFT,
			      100 * L1_CACHE_ALIGN(sizeof(struct inet_peer)));

	inet_peer_threshold = clamp_val(nr_entries, 4096, 65536 + 128);

	peer_cachep = kmem_cache_create("inet_peer_cache",
			sizeof(struct inet_peer),
			0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
			NULL);
}

/* Called with rcu_read_lock() or base->lock held */
static struct inet_peer *lookup(const struct inetpeer_addr *daddr,
				struct inet_peer_base *base,
				unsigned int seq,
				struct inet_peer *gc_stack[],
				unsigned int *gc_cnt,
				struct rb_node **parent_p,
				struct rb_node ***pp_p)
{
	struct rb_node **pp, *parent, *next;
	struct inet_peer *p;

	pp = &base->rb_root.rb_node;
	parent = NULL;
	while (1) {
		int cmp;

		next = rcu_dereference_raw(*pp);
		if (!next)
			break;
		parent = next;
		p = rb_entry(parent, struct inet_peer, rb_node);
		cmp = inetpeer_addr_cmp(daddr, &p->daddr);
		if (cmp == 0) {
			if (!refcount_inc_not_zero(&p->refcnt))
				break;
			return p;
		}
		if (gc_stack) {
			if (*gc_cnt < PEER_MAX_GC)
				gc_stack[(*gc_cnt)++] = p;
		} else if (unlikely(read_seqretry(&base->lock, seq))) {
			break;
		}
		if (cmp == -1)
			pp = &next->rb_left;
		else
			pp = &next->rb_right;
	}
	*parent_p = parent;
	*pp_p = pp;
	return NULL;
}

static void inetpeer_free_rcu(struct rcu_head *head)
{
	kmem_cache_free(peer_cachep, container_of(head, struct inet_peer, rcu));
}

/* perform garbage collect on all items stacked during a lookup */
static void inet_peer_gc(struct inet_peer_base *base,
			 struct inet_peer *gc_stack[],
			 unsigned int gc_cnt)
{
	int peer_threshold, peer_maxttl, peer_minttl;
	struct inet_peer *p;
	__u32 delta, ttl;
	int i;

	peer_threshold = READ_ONCE(inet_peer_threshold);
	peer_maxttl = READ_ONCE(inet_peer_maxttl);
	peer_minttl = READ_ONCE(inet_peer_minttl);

	if (base->total >= peer_threshold)
		ttl = 0; /* be aggressive */
	else
		ttl = peer_maxttl - (peer_maxttl - peer_minttl) / HZ *
			base->total / peer_threshold * HZ;
	for (i = 0; i < gc_cnt; i++) {
		p = gc_stack[i];

		/* The READ_ONCE() pairs with the WRITE_ONCE()
		 * in inet_putpeer()
		 */
		delta = (__u32)jiffies - READ_ONCE(p->dtime);

		if (delta < ttl || !refcount_dec_if_one(&p->refcnt))
			gc_stack[i] = NULL;
	}
	for (i = 0; i < gc_cnt; i++) {
		p = gc_stack[i];
		if (p) {
			rb_erase(&p->rb_node, &base->rb_root);
			base->total--;
			call_rcu(&p->rcu, inetpeer_free_rcu);
		}
	}
}

struct inet_peer *inet_getpeer(struct inet_peer_base *base,
			       const struct inetpeer_addr *daddr,
			       int create)
{
	struct inet_peer *p, *gc_stack[PEER_MAX_GC];
	struct rb_node **pp, *parent;
	unsigned int gc_cnt, seq;
	int invalidated;

	/* Attempt a lockless lookup first.
	 * Because of a concurrent writer, we might not find an existing entry.
	 */
	rcu_read_lock();
	seq = read_seqbegin(&base->lock);
	p = lookup(daddr, base, seq, NULL, &gc_cnt, &parent, &pp);
	invalidated = read_seqretry(&base->lock, seq);
	rcu_read_unlock();

	if (p)
		return p;

	/* If no writer did a change during our lookup, we can return early. */
	if (!create && !invalidated)
		return NULL;

	/* retry an exact lookup, taking the lock before.
	 * At least, nodes should be hot in our cache.
	 */
	parent = NULL;
	write_seqlock_bh(&base->lock);

	gc_cnt = 0;
	p = lookup(daddr, base, seq, gc_stack, &gc_cnt, &parent, &pp);
	if (!p && create) {
		p = kmem_cache_alloc(peer_cachep, GFP_ATOMIC);
		if (p) {
			p->daddr = *daddr;
			p->dtime = (__u32)jiffies;
			refcount_set(&p->refcnt, 2);
			atomic_set(&p->rid, 0);
			p->metrics[RTAX_LOCK-1] = INETPEER_METRICS_NEW;
			p->rate_tokens = 0;
			p->n_redirects = 0;
			/* 60*HZ is arbitrary, but chosen enough high so that the first
			 * calculation of tokens is at its maximum.
			 */
			p->rate_last = jiffies - 60*HZ;

			rb_link_node(&p->rb_node, parent, pp);
			rb_insert_color(&p->rb_node, &base->rb_root);
			base->total++;
		}
	}
	if (gc_cnt)
		inet_peer_gc(base, gc_stack, gc_cnt);
	write_sequnlock_bh(&base->lock);

	return p;
}
EXPORT_SYMBOL_GPL(inet_getpeer);

void inet_putpeer(struct inet_peer *p)
{
	/* The WRITE_ONCE() pairs with itself (we run lockless)
	 * and the READ_ONCE() in inet_peer_gc()
	 */
	WRITE_ONCE(p->dtime, (__u32)jiffies);

	if (refcount_dec_and_test(&p->refcnt))
		call_rcu(&p->rcu, inetpeer_free_rcu);
}
EXPORT_SYMBOL_GPL(inet_putpeer);

/*
 *	Check transmit rate limitation for given message.
 *	The rate information is held in the inet_peer entries now.
 *	This function is generic and could be used for other purposes
 *	too. It uses a Token bucket filter as suggested by Alexey Kuznetsov.
 *
 *	Note that the same inet_peer fields are modified by functions in
 *	route.c too, but these work for packet destinations while xrlim_allow
 *	works for icmp destinations. This means the rate limiting information
 *	for one "ip object" is shared - and these ICMPs are twice limited:
 *	by source and by destination.
 *
 *	RFC 1812: 4.3.2.8 SHOULD be able to limit error message rate
 *			  SHOULD allow setting of rate limits
 *
 * 	Shared between ICMPv4 and ICMPv6.
 */
#define XRLIM_BURST_FACTOR 6
bool inet_peer_xrlim_allow(struct inet_peer *peer, int timeout)
{
	unsigned long now, token;
	bool rc = false;

	if (!peer)
		return true;

	token = peer->rate_tokens;
	now = jiffies;
	token += now - peer->rate_last;
	peer->rate_last = now;
	if (token > XRLIM_BURST_FACTOR * timeout)
		token = XRLIM_BURST_FACTOR * timeout;
	if (token >= timeout) {
		token -= timeout;
		rc = true;
	}
	peer->rate_tokens = token;
	return rc;
}
EXPORT_SYMBOL(inet_peer_xrlim_allow);

void inetpeer_invalidate_tree(struct inet_peer_base *base)
{
	struct rb_node *p = rb_first(&base->rb_root);

	while (p) {
		struct inet_peer *peer = rb_entry(p, struct inet_peer, rb_node);

		p = rb_next(p);
		rb_erase(&peer->rb_node, &base->rb_root);
		inet_putpeer(peer);
		cond_resched();
	}

	base->total = 0;
}
EXPORT_SYMBOL(inetpeer_invalidate_tree);
Commit	Line	Data
1da177e4 LT	1	/*
	2	* INETPEER - A storage for permanent information about peers
	3	*
	4	* This source is covered by the GNU GPL, the same as all kernel sources.
	5	*
1da177e4 LT	6	* Authors: Andrey V. Savochkin <saw@msu.ru>
	7	*/
	8
08009a76	9	#include <linux/cache.h>
1da177e4 LT	10	#include <linux/module.h>
	11	#include <linux/types.h>
	12	#include <linux/slab.h>
	13	#include <linux/interrupt.h>
	14	#include <linux/spinlock.h>
	15	#include <linux/random.h>
1da177e4 LT	16	#include <linux/timer.h>
	17	#include <linux/time.h>
	18	#include <linux/kernel.h>
	19	#include <linux/mm.h>
	20	#include <linux/net.h>
5faa5df1	21	#include <linux/workqueue.h>
20380731	22	#include <net/ip.h>
1da177e4	23	#include <net/inetpeer.h>
6e5714ea	24	#include <net/secure_seq.h>
1da177e4 LT	25
	26	/*
	27	* Theory of operations.
	28	* We keep one entry for each peer IP address. The nodes contains long-living
	29	* information about the peer which doesn't depend on routes.
1da177e4	30	*
1da177e4 LT	31	* Nodes are removed only when reference counter goes to 0.
	32	* When it's happened the node may be removed when a sufficient amount of
	33	* time has been passed since its last use. The less-recently-used entry can
	34	* also be removed if the pool is overloaded i.e. if the total amount of
	35	* entries is greater-or-equal than the threshold.
	36	*
b145425f	37	* Node pool is organised as an RB tree.
1da177e4 LT	38	* Such an implementation has been chosen not just for fun. It's a way to
	39	* prevent easy and efficient DoS attacks by creating hash collisions. A huge
	40	* amount of long living nodes in a single hash slot would significantly delay
	41	* lookups performed with disabled BHs.
	42	*
	43	* Serialisation issues.
aa1039e7 ED	44	* 1. Nodes may appear in the tree only with the pool lock held.
aa1039e7 ED	45	* 2. Nodes may disappear from the tree only with the pool lock held
1da177e4	46	* AND reference count being 0.
4b9d9be8 ED	47	* 3. Global variable peer_total is modified under the pool lock.
4b9d9be8 ED	48	* 4. struct inet_peer fields modification:
b145425f	49	* rb_node: pool lock
1da177e4 LT	50	* refcnt: atomically against modifications on other CPU;
1da177e4 LT	51	* usually under some other lock to prevent node disappearing
582a72da	52	* daddr: unchangeable
1da177e4 LT	53	*/
1da177e4 LT	54
08009a76	55	static struct kmem_cache *peer_cachep __ro_after_init;
1da177e4	56
c3426b47 DM	57	void inet_peer_base_init(struct inet_peer_base *bp)
c3426b47 DM	58	{
b145425f	59	bp->rb_root = RB_ROOT;
c3426b47 DM	60	seqlock_init(&bp->lock);
	61	bp->total = 0;
	62	}
	63	EXPORT_SYMBOL_GPL(inet_peer_base_init);
021e9299	64
b145425f	65	#define PEER_MAX_GC 32
1da177e4	66
1da177e4	67	/* Exported for sysctl_net_ipv4. */
8bd2a055	68	int inet_peer_threshold __read_mostly; /* start to throw entries more
1da177e4	69	* aggressively at this stage */
243bbcaa ED	70	int inet_peer_minttl __read_mostly = 120 * HZ; /* TTL under high load: 120 sec */
243bbcaa ED	71	int inet_peer_maxttl __read_mostly = 10 * 60 * HZ; /* usual time to live: 10 min */
1da177e4	72
1da177e4 LT	73	/* Called from ip_output.c:ip_init */
	74	void __init inet_initpeers(void)
	75	{
8bd2a055	76	u64 nr_entries;
1da177e4	77
8bd2a055 YD	78	/* 1% of physical memory */
	79	nr_entries = div64_ul((u64)totalram_pages() << PAGE_SHIFT,
	80	100 * L1_CACHE_ALIGN(sizeof(struct inet_peer)));
	81
	82	inet_peer_threshold = clamp_val(nr_entries, 4096, 65536 + 128);
1da177e4 LT	83
	84	peer_cachep = kmem_cache_create("inet_peer_cache",
	85	sizeof(struct inet_peer),
317fe0e6	86	0, SLAB_HWCACHE_ALIGN \| SLAB_PANIC,
20c2df83	87	NULL);
1da177e4 LT	88	}
1da177e4 LT	89
b145425f ED	90	/* Called with rcu_read_lock() or base->lock held */
	91	static struct inet_peer lookup(const struct inetpeer_addr daddr,
	92	struct inet_peer_base *base,
	93	unsigned int seq,
	94	struct inet_peer *gc_stack[],
	95	unsigned int *gc_cnt,
	96	struct rb_node **parent_p,
	97	struct rb_node ***pp_p)
aa1039e7	98	{
4cc5b44b	99	struct rb_node *pp, parent, *next;
b145425f ED	100	struct inet_peer *p;
	101
	102	pp = &base->rb_root.rb_node;
	103	parent = NULL;
4cc5b44b	104	while (1) {
b145425f	105	int cmp;
aa1039e7	106
4cc5b44b ED	107	next = rcu_dereference_raw(*pp);
	108	if (!next)
	109	break;
	110	parent = next;
b145425f ED	111	p = rb_entry(parent, struct inet_peer, rb_node);
b145425f ED	112	cmp = inetpeer_addr_cmp(daddr, &p->daddr);
02663045	113	if (cmp == 0) {
b145425f ED	114	if (!refcount_inc_not_zero(&p->refcnt))
	115	break;
	116	return p;
	117	}
	118	if (gc_stack) {
	119	if (*gc_cnt < PEER_MAX_GC)
	120	gc_stack[(*gc_cnt)++] = p;
	121	} else if (unlikely(read_seqretry(&base->lock, seq))) {
	122	break;
aa1039e7	123	}
02663045	124	if (cmp == -1)
35f493b8	125	pp = &next->rb_left;
aa1039e7	126	else
35f493b8	127	pp = &next->rb_right;
aa1039e7	128	}
b145425f ED	129	*parent_p = parent;
b145425f ED	130	*pp_p = pp;
aa1039e7 ED	131	return NULL;
	132	}
	133
aa1039e7 ED	134	static void inetpeer_free_rcu(struct rcu_head *head)
	135	{
	136	kmem_cache_free(peer_cachep, container_of(head, struct inet_peer, rcu));
	137	}
	138
4b9d9be8	139	/* perform garbage collect on all items stacked during a lookup */
b145425f ED	140	static void inet_peer_gc(struct inet_peer_base *base,
	141	struct inet_peer *gc_stack[],
	142	unsigned int gc_cnt)
98158f5a	143	{
3d32edf1	144	int peer_threshold, peer_maxttl, peer_minttl;
b145425f	145	struct inet_peer *p;
4b9d9be8	146	__u32 delta, ttl;
b145425f	147	int i;
d71209de	148
3d32edf1 KI	149	peer_threshold = READ_ONCE(inet_peer_threshold);
	150	peer_maxttl = READ_ONCE(inet_peer_maxttl);
	151	peer_minttl = READ_ONCE(inet_peer_minttl);
	152
	153	if (base->total >= peer_threshold)
4b9d9be8 ED	154	ttl = 0; /* be aggressive */
4b9d9be8 ED	155	else
3d32edf1 KI	156	ttl = peer_maxttl - (peer_maxttl - peer_minttl) / HZ *
3d32edf1 KI	157	base->total / peer_threshold * HZ;
b145425f ED	158	for (i = 0; i < gc_cnt; i++) {
b145425f ED	159	p = gc_stack[i];
71685eb4 ED	160
	161	/* The READ_ONCE() pairs with the WRITE_ONCE()
	162	* in inet_putpeer()
	163	*/
	164	delta = (__u32)jiffies - READ_ONCE(p->dtime);
	165
b145425f ED	166	if (delta < ttl \|\| !refcount_dec_if_one(&p->refcnt))
b145425f ED	167	gc_stack[i] = NULL;
1da177e4	168	}
b145425f ED	169	for (i = 0; i < gc_cnt; i++) {
	170	p = gc_stack[i];
	171	if (p) {
	172	rb_erase(&p->rb_node, &base->rb_root);
	173	base->total--;
	174	call_rcu(&p->rcu, inetpeer_free_rcu);
	175	}
4b9d9be8	176	}
1da177e4 LT	177	}
1da177e4 LT	178
c0efc887	179	struct inet_peer inet_getpeer(struct inet_peer_base base,
c8a627ed G	180	const struct inetpeer_addr *daddr,
c8a627ed G	181	int create)
1da177e4	182	{
b145425f ED	183	struct inet_peer p, gc_stack[PEER_MAX_GC];
	184	struct rb_node *pp, parent;
	185	unsigned int gc_cnt, seq;
	186	int invalidated;
1da177e4	187
4b9d9be8	188	/* Attempt a lockless lookup first.
aa1039e7 ED	189	* Because of a concurrent writer, we might not find an existing entry.
aa1039e7 ED	190	*/
7b46ac4e	191	rcu_read_lock();
b145425f ED	192	seq = read_seqbegin(&base->lock);
	193	p = lookup(daddr, base, seq, NULL, &gc_cnt, &parent, &pp);
	194	invalidated = read_seqretry(&base->lock, seq);
7b46ac4e	195	rcu_read_unlock();
aa1039e7	196
4b9d9be8	197	if (p)
aa1039e7	198	return p;
1da177e4	199
65e8354e ED	200	/* If no writer did a change during our lookup, we can return early. */
	201	if (!create && !invalidated)
	202	return NULL;
	203
aa1039e7 ED	204	/* retry an exact lookup, taking the lock before.
	205	* At least, nodes should be hot in our cache.
	206	*/
b145425f	207	parent = NULL;
65e8354e	208	write_seqlock_bh(&base->lock);
aa1039e7	209
b145425f ED	210	gc_cnt = 0;
	211	p = lookup(daddr, base, seq, gc_stack, &gc_cnt, &parent, &pp);
	212	if (!p && create) {
	213	p = kmem_cache_alloc(peer_cachep, GFP_ATOMIC);
	214	if (p) {
	215	p->daddr = *daddr;
b6a37e5e	216	p->dtime = (__u32)jiffies;
b145425f ED	217	refcount_set(&p->refcnt, 2);
	218	atomic_set(&p->rid, 0);
	219	p->metrics[RTAX_LOCK-1] = INETPEER_METRICS_NEW;
	220	p->rate_tokens = 0;
c09551c6	221	p->n_redirects = 0;
b145425f ED	222	/* 60*HZ is arbitrary, but chosen enough high so that the first
	223	* calculation of tokens is at its maximum.
	224	*/
	225	p->rate_last = jiffies - 60*HZ;
	226
	227	rb_link_node(&p->rb_node, parent, pp);
	228	rb_insert_color(&p->rb_node, &base->rb_root);
	229	base->total++;
	230	}
aa1039e7	231	}
b145425f ED	232	if (gc_cnt)
b145425f ED	233	inet_peer_gc(base, gc_stack, gc_cnt);
65e8354e	234	write_sequnlock_bh(&base->lock);
1da177e4	235
1da177e4 LT	236	return p;
1da177e4 LT	237	}
b3419363	238	EXPORT_SYMBOL_GPL(inet_getpeer);
98158f5a	239
4663afe2 ED	240	void inet_putpeer(struct inet_peer *p)
4663afe2 ED	241	{
71685eb4 ED	242	/* The WRITE_ONCE() pairs with itself (we run lockless)
	243	* and the READ_ONCE() in inet_peer_gc()
	244	*/
	245	WRITE_ONCE(p->dtime, (__u32)jiffies);
b145425f ED	246
	247	if (refcount_dec_and_test(&p->refcnt))
	248	call_rcu(&p->rcu, inetpeer_free_rcu);
4663afe2	249	}
b3419363	250	EXPORT_SYMBOL_GPL(inet_putpeer);
92d86829 DM	251
	252	/*
	253	* Check transmit rate limitation for given message.
	254	* The rate information is held in the inet_peer entries now.
	255	* This function is generic and could be used for other purposes
	256	* too. It uses a Token bucket filter as suggested by Alexey Kuznetsov.
	257	*
	258	* Note that the same inet_peer fields are modified by functions in
	259	* route.c too, but these work for packet destinations while xrlim_allow
	260	* works for icmp destinations. This means the rate limiting information
	261	* for one "ip object" is shared - and these ICMPs are twice limited:
	262	* by source and by destination.
	263	*
	264	* RFC 1812: 4.3.2.8 SHOULD be able to limit error message rate
	265	* SHOULD allow setting of rate limits
	266	*
	267	* Shared between ICMPv4 and ICMPv6.
	268	*/
	269	#define XRLIM_BURST_FACTOR 6
	270	bool inet_peer_xrlim_allow(struct inet_peer *peer, int timeout)
	271	{
	272	unsigned long now, token;
	273	bool rc = false;
	274
	275	if (!peer)
	276	return true;
	277
	278	token = peer->rate_tokens;
	279	now = jiffies;
	280	token += now - peer->rate_last;
	281	peer->rate_last = now;
	282	if (token > XRLIM_BURST_FACTOR * timeout)
	283	token = XRLIM_BURST_FACTOR * timeout;
	284	if (token >= timeout) {
	285	token -= timeout;
	286	rc = true;
	287	}
	288	peer->rate_tokens = token;
	289	return rc;
	290	}
	291	EXPORT_SYMBOL(inet_peer_xrlim_allow);
5faa5df1	292
56a6b248	293	void inetpeer_invalidate_tree(struct inet_peer_base *base)
5faa5df1	294	{
8f1975e3	295	struct rb_node *p = rb_first(&base->rb_root);
5faa5df1	296
8f1975e3 ED	297	while (p) {
	298	struct inet_peer *peer = rb_entry(p, struct inet_peer, rb_node);
	299
	300	p = rb_next(p);
	301	rb_erase(&peer->rb_node, &base->rb_root);
	302	inet_putpeer(peer);
b145425f	303	cond_resched();
5faa5df1 SK	304	}
5faa5df1 SK	305
b145425f	306	base->total = 0;
5faa5df1 SK	307	}
5faa5df1 SK	308	EXPORT_SYMBOL(inetpeer_invalidate_tree);