[net-next-2.6.git] / net / ipv4 / inet_fragment.c

/*
 * inet fragments management
 *
 *		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.
 *
 * 		Authors:	Pavel Emelyanov <xemul@openvz.org>
 *				Started as consolidation of ipv4/ip_fragment.c,
 *				ipv6/reassembly. and ipv6 nf conntrack reassembly
 */

#include <linux/list.h>
#include <linux/spinlock.h>
#include <linux/module.h>
#include <linux/timer.h>
#include <linux/mm.h>
#include <linux/random.h>
#include <linux/skbuff.h>
#include <linux/rtnetlink.h>

#include <net/inet_frag.h>

static void inet_frag_secret_rebuild(unsigned long dummy)
{
	struct inet_frags *f = (struct inet_frags *)dummy;
	unsigned long now = jiffies;
	int i;

	write_lock(&f->lock);
	get_random_bytes(&f->rnd, sizeof(u32));
	for (i = 0; i < INETFRAGS_HASHSZ; i++) {
		struct inet_frag_queue *q;
		struct hlist_node *p, *n;

		hlist_for_each_entry_safe(q, p, n, &f->hash[i], list) {
			unsigned int hval = f->hashfn(q);

			if (hval != i) {
				hlist_del(&q->list);

				/* Relink to new hash chain. */
				hlist_add_head(&q->list, &f->hash[hval]);
			}
		}
	}
	write_unlock(&f->lock);

	mod_timer(&f->secret_timer, now + f->secret_interval);
}

void inet_frags_init(struct inet_frags *f)
{
	int i;

	for (i = 0; i < INETFRAGS_HASHSZ; i++)
		INIT_HLIST_HEAD(&f->hash[i]);

	rwlock_init(&f->lock);

	f->rnd = (u32) ((num_physpages ^ (num_physpages>>7)) ^
				   (jiffies ^ (jiffies >> 6)));

	setup_timer(&f->secret_timer, inet_frag_secret_rebuild,
			(unsigned long)f);
	f->secret_timer.expires = jiffies + f->secret_interval;
	add_timer(&f->secret_timer);
}
EXPORT_SYMBOL(inet_frags_init);

void inet_frags_init_net(struct netns_frags *nf)
{
	nf->nqueues = 0;
	atomic_set(&nf->mem, 0);
	INIT_LIST_HEAD(&nf->lru_list);
}
EXPORT_SYMBOL(inet_frags_init_net);

void inet_frags_fini(struct inet_frags *f)
{
	del_timer(&f->secret_timer);
}
EXPORT_SYMBOL(inet_frags_fini);

void inet_frags_exit_net(struct netns_frags *nf, struct inet_frags *f)
{
	nf->low_thresh = 0;

	local_bh_disable();
	inet_frag_evictor(nf, f);
	local_bh_enable();
}
EXPORT_SYMBOL(inet_frags_exit_net);

static inline void fq_unlink(struct inet_frag_queue *fq, struct inet_frags *f)
{
	write_lock(&f->lock);
	hlist_del(&fq->list);
	list_del(&fq->lru_list);
	fq->net->nqueues--;
	write_unlock(&f->lock);
}

void inet_frag_kill(struct inet_frag_queue *fq, struct inet_frags *f)
{
	if (del_timer(&fq->timer))
		atomic_dec(&fq->refcnt);

	if (!(fq->last_in & INET_FRAG_COMPLETE)) {
		fq_unlink(fq, f);
		atomic_dec(&fq->refcnt);
		fq->last_in |= INET_FRAG_COMPLETE;
	}
}

EXPORT_SYMBOL(inet_frag_kill);

static inline void frag_kfree_skb(struct netns_frags *nf, struct inet_frags *f,
		struct sk_buff *skb, int *work)
{
	if (work)
		*work -= skb->truesize;

	atomic_sub(skb->truesize, &nf->mem);
	if (f->skb_free)
		f->skb_free(skb);
	kfree_skb(skb);
}

void inet_frag_destroy(struct inet_frag_queue *q, struct inet_frags *f,
					int *work)
{
	struct sk_buff *fp;
	struct netns_frags *nf;

	WARN_ON(!(q->last_in & INET_FRAG_COMPLETE));
	WARN_ON(del_timer(&q->timer) != 0);

	/* Release all fragment data. */
	fp = q->fragments;
	nf = q->net;
	while (fp) {
		struct sk_buff *xp = fp->next;

		frag_kfree_skb(nf, f, fp, work);
		fp = xp;
	}

	if (work)
		*work -= f->qsize;
	atomic_sub(f->qsize, &nf->mem);

	if (f->destructor)
		f->destructor(q);
	kfree(q);

}
EXPORT_SYMBOL(inet_frag_destroy);

int inet_frag_evictor(struct netns_frags *nf, struct inet_frags *f)
{
	struct inet_frag_queue *q;
	int work, evicted = 0;

	work = atomic_read(&nf->mem) - nf->low_thresh;
	while (work > 0) {
		read_lock(&f->lock);
		if (list_empty(&nf->lru_list)) {
			read_unlock(&f->lock);
			break;
		}

		q = list_first_entry(&nf->lru_list,
				struct inet_frag_queue, lru_list);
		atomic_inc(&q->refcnt);
		read_unlock(&f->lock);

		spin_lock(&q->lock);
		if (!(q->last_in & INET_FRAG_COMPLETE))
			inet_frag_kill(q, f);
		spin_unlock(&q->lock);

		if (atomic_dec_and_test(&q->refcnt))
			inet_frag_destroy(q, f, &work);
		evicted++;
	}

	return evicted;
}
EXPORT_SYMBOL(inet_frag_evictor);

static struct inet_frag_queue *inet_frag_intern(struct netns_frags *nf,
		struct inet_frag_queue *qp_in, struct inet_frags *f,
		void *arg)
{
	struct inet_frag_queue *qp;
#ifdef CONFIG_SMP
	struct hlist_node *n;
#endif
	unsigned int hash;

	write_lock(&f->lock);
	/*
	 * While we stayed w/o the lock other CPU could update
	 * the rnd seed, so we need to re-calculate the hash
	 * chain. Fortunatelly the qp_in can be used to get one.
	 */
	hash = f->hashfn(qp_in);
#ifdef CONFIG_SMP
	/* With SMP race we have to recheck hash table, because
	 * such entry could be created on other cpu, while we
	 * promoted read lock to write lock.
	 */
	hlist_for_each_entry(qp, n, &f->hash[hash], list) {
		if (qp->net == nf && f->match(qp, arg)) {
			atomic_inc(&qp->refcnt);
			write_unlock(&f->lock);
			qp_in->last_in |= INET_FRAG_COMPLETE;
			inet_frag_put(qp_in, f);
			return qp;
		}
	}
#endif
	qp = qp_in;
	if (!mod_timer(&qp->timer, jiffies + nf->timeout))
		atomic_inc(&qp->refcnt);

	atomic_inc(&qp->refcnt);
	hlist_add_head(&qp->list, &f->hash[hash]);
	list_add_tail(&qp->lru_list, &nf->lru_list);
	nf->nqueues++;
	write_unlock(&f->lock);
	return qp;
}

static struct inet_frag_queue *inet_frag_alloc(struct netns_frags *nf,
		struct inet_frags *f, void *arg)
{
	struct inet_frag_queue *q;

	q = kzalloc(f->qsize, GFP_ATOMIC);
	if (q == NULL)
		return NULL;

	f->constructor(q, arg);
	atomic_add(f->qsize, &nf->mem);
	setup_timer(&q->timer, f->frag_expire, (unsigned long)q);
	spin_lock_init(&q->lock);
	atomic_set(&q->refcnt, 1);
	q->net = nf;

	return q;
}

static struct inet_frag_queue *inet_frag_create(struct netns_frags *nf,
		struct inet_frags *f, void *arg)
{
	struct inet_frag_queue *q;

	q = inet_frag_alloc(nf, f, arg);
	if (q == NULL)
		return NULL;

	return inet_frag_intern(nf, q, f, arg);
}

struct inet_frag_queue *inet_frag_find(struct netns_frags *nf,
		struct inet_frags *f, void *key, unsigned int hash)
	__releases(&f->lock)
{
	struct inet_frag_queue *q;
	struct hlist_node *n;

	hlist_for_each_entry(q, n, &f->hash[hash], list) {
		if (q->net == nf && f->match(q, key)) {
			atomic_inc(&q->refcnt);
			read_unlock(&f->lock);
			return q;
		}
	}
	read_unlock(&f->lock);

	return inet_frag_create(nf, f, key);
}
EXPORT_SYMBOL(inet_frag_find);
Commit	Line	Data
7eb95156 PE	1	/*
	2	* inet fragments management
	3	*
	4	* This program is free software; you can redistribute it and/or
	5	* modify it under the terms of the GNU General Public License
	6	* as published by the Free Software Foundation; either version
	7	* 2 of the License, or (at your option) any later version.
	8	*
	9	* Authors: Pavel Emelyanov <xemul@openvz.org>
	10	* Started as consolidation of ipv4/ip_fragment.c,
	11	* ipv6/reassembly. and ipv6 nf conntrack reassembly
	12	*/
	13
	14	#include <linux/list.h>
	15	#include <linux/spinlock.h>
	16	#include <linux/module.h>
	17	#include <linux/timer.h>
	18	#include <linux/mm.h>
321a3a99	19	#include <linux/random.h>
1e4b8287 PE	20	#include <linux/skbuff.h>
1e4b8287 PE	21	#include <linux/rtnetlink.h>
7eb95156 PE	22
	23	#include <net/inet_frag.h>
	24
321a3a99 PE	25	static void inet_frag_secret_rebuild(unsigned long dummy)
	26	{
	27	struct inet_frags f = (struct inet_frags )dummy;
	28	unsigned long now = jiffies;
	29	int i;
	30
	31	write_lock(&f->lock);
	32	get_random_bytes(&f->rnd, sizeof(u32));
	33	for (i = 0; i < INETFRAGS_HASHSZ; i++) {
	34	struct inet_frag_queue *q;
	35	struct hlist_node p, n;
	36
	37	hlist_for_each_entry_safe(q, p, n, &f->hash[i], list) {
	38	unsigned int hval = f->hashfn(q);
	39
	40	if (hval != i) {
	41	hlist_del(&q->list);
	42
	43	/* Relink to new hash chain. */
	44	hlist_add_head(&q->list, &f->hash[hval]);
	45	}
	46	}
	47	}
	48	write_unlock(&f->lock);
	49
3b4bc4a2	50	mod_timer(&f->secret_timer, now + f->secret_interval);
321a3a99 PE	51	}
321a3a99 PE	52
7eb95156 PE	53	void inet_frags_init(struct inet_frags *f)
	54	{
	55	int i;
	56
	57	for (i = 0; i < INETFRAGS_HASHSZ; i++)
	58	INIT_HLIST_HEAD(&f->hash[i]);
	59
7eb95156 PE	60	rwlock_init(&f->lock);
	61
	62	f->rnd = (u32) ((num_physpages ^ (num_physpages>>7)) ^
	63	(jiffies ^ (jiffies >> 6)));
	64
b24b8a24 PE	65	setup_timer(&f->secret_timer, inet_frag_secret_rebuild,
b24b8a24 PE	66	(unsigned long)f);
3b4bc4a2	67	f->secret_timer.expires = jiffies + f->secret_interval;
321a3a99	68	add_timer(&f->secret_timer);
7eb95156 PE	69	}
	70	EXPORT_SYMBOL(inet_frags_init);
	71
e5a2bb84 PE	72	void inet_frags_init_net(struct netns_frags *nf)
	73	{
	74	nf->nqueues = 0;
6ddc0822	75	atomic_set(&nf->mem, 0);
3140c25c	76	INIT_LIST_HEAD(&nf->lru_list);
e5a2bb84 PE	77	}
	78	EXPORT_SYMBOL(inet_frags_init_net);
	79
7eb95156 PE	80	void inet_frags_fini(struct inet_frags *f)
7eb95156 PE	81	{
321a3a99	82	del_timer(&f->secret_timer);
7eb95156 PE	83	}
7eb95156 PE	84	EXPORT_SYMBOL(inet_frags_fini);
277e650d	85
81566e83 PE	86	void inet_frags_exit_net(struct netns_frags nf, struct inet_frags f)
	87	{
	88	nf->low_thresh = 0;
e8e16b70 DM	89
e8e16b70 DM	90	local_bh_disable();
81566e83	91	inet_frag_evictor(nf, f);
e8e16b70	92	local_bh_enable();
81566e83 PE	93	}
	94	EXPORT_SYMBOL(inet_frags_exit_net);
	95
277e650d PE	96	static inline void fq_unlink(struct inet_frag_queue fq, struct inet_frags f)
	97	{
	98	write_lock(&f->lock);
	99	hlist_del(&fq->list);
	100	list_del(&fq->lru_list);
e5a2bb84	101	fq->net->nqueues--;
277e650d PE	102	write_unlock(&f->lock);
	103	}
	104
	105	void inet_frag_kill(struct inet_frag_queue fq, struct inet_frags f)
	106	{
	107	if (del_timer(&fq->timer))
	108	atomic_dec(&fq->refcnt);
	109
bc578a54	110	if (!(fq->last_in & INET_FRAG_COMPLETE)) {
277e650d PE	111	fq_unlink(fq, f);
277e650d PE	112	atomic_dec(&fq->refcnt);
bc578a54	113	fq->last_in \|= INET_FRAG_COMPLETE;
277e650d PE	114	}
	115	}
	116
	117	EXPORT_SYMBOL(inet_frag_kill);
1e4b8287	118
6ddc0822 PE	119	static inline void frag_kfree_skb(struct netns_frags nf, struct inet_frags f,
6ddc0822 PE	120	struct sk_buff skb, int work)
1e4b8287 PE	121	{
	122	if (work)
	123	*work -= skb->truesize;
	124
6ddc0822	125	atomic_sub(skb->truesize, &nf->mem);
1e4b8287 PE	126	if (f->skb_free)
	127	f->skb_free(skb);
	128	kfree_skb(skb);
	129	}
	130
	131	void inet_frag_destroy(struct inet_frag_queue q, struct inet_frags f,
	132	int *work)
	133	{
	134	struct sk_buff *fp;
6ddc0822	135	struct netns_frags *nf;
1e4b8287	136
547b792c IJ	137	WARN_ON(!(q->last_in & INET_FRAG_COMPLETE));
547b792c IJ	138	WARN_ON(del_timer(&q->timer) != 0);
1e4b8287 PE	139
	140	/* Release all fragment data. */
	141	fp = q->fragments;
6ddc0822	142	nf = q->net;
1e4b8287 PE	143	while (fp) {
	144	struct sk_buff *xp = fp->next;
	145
6ddc0822	146	frag_kfree_skb(nf, f, fp, work);
1e4b8287 PE	147	fp = xp;
	148	}
	149
	150	if (work)
	151	*work -= f->qsize;
6ddc0822	152	atomic_sub(f->qsize, &nf->mem);
1e4b8287	153
c9547709 PE	154	if (f->destructor)
	155	f->destructor(q);
	156	kfree(q);
1e4b8287 PE	157
	158	}
	159	EXPORT_SYMBOL(inet_frag_destroy);
8e7999c4	160
6ddc0822	161	int inet_frag_evictor(struct netns_frags nf, struct inet_frags f)
8e7999c4 PE	162	{
	163	struct inet_frag_queue *q;
	164	int work, evicted = 0;
	165
e31e0bdc	166	work = atomic_read(&nf->mem) - nf->low_thresh;
8e7999c4 PE	167	while (work > 0) {
8e7999c4 PE	168	read_lock(&f->lock);
3140c25c	169	if (list_empty(&nf->lru_list)) {
8e7999c4 PE	170	read_unlock(&f->lock);
	171	break;
	172	}
	173
3140c25c	174	q = list_first_entry(&nf->lru_list,
8e7999c4 PE	175	struct inet_frag_queue, lru_list);
	176	atomic_inc(&q->refcnt);
	177	read_unlock(&f->lock);
	178
	179	spin_lock(&q->lock);
bc578a54	180	if (!(q->last_in & INET_FRAG_COMPLETE))
8e7999c4 PE	181	inet_frag_kill(q, f);
	182	spin_unlock(&q->lock);
	183
	184	if (atomic_dec_and_test(&q->refcnt))
	185	inet_frag_destroy(q, f, &work);
	186	evicted++;
	187	}
	188
	189	return evicted;
	190	}
	191	EXPORT_SYMBOL(inet_frag_evictor);
2588fe1d	192
ac18e750 PE	193	static struct inet_frag_queue inet_frag_intern(struct netns_frags nf,
ac18e750 PE	194	struct inet_frag_queue qp_in, struct inet_frags f,
9a375803	195	void *arg)
2588fe1d PE	196	{
	197	struct inet_frag_queue *qp;
	198	#ifdef CONFIG_SMP
	199	struct hlist_node *n;
	200	#endif
9a375803	201	unsigned int hash;
2588fe1d PE	202
2588fe1d PE	203	write_lock(&f->lock);
9a375803 PE	204	/*
	205	* While we stayed w/o the lock other CPU could update
	206	* the rnd seed, so we need to re-calculate the hash
	207	* chain. Fortunatelly the qp_in can be used to get one.
	208	*/
	209	hash = f->hashfn(qp_in);
2588fe1d PE	210	#ifdef CONFIG_SMP
	211	/* With SMP race we have to recheck hash table, because
	212	* such entry could be created on other cpu, while we
	213	* promoted read lock to write lock.
	214	*/
	215	hlist_for_each_entry(qp, n, &f->hash[hash], list) {
ac18e750	216	if (qp->net == nf && f->match(qp, arg)) {
2588fe1d PE	217	atomic_inc(&qp->refcnt);
2588fe1d PE	218	write_unlock(&f->lock);
bc578a54	219	qp_in->last_in \|= INET_FRAG_COMPLETE;
2588fe1d PE	220	inet_frag_put(qp_in, f);
	221	return qp;
	222	}
	223	}
	224	#endif
	225	qp = qp_in;
b2fd5321	226	if (!mod_timer(&qp->timer, jiffies + nf->timeout))
2588fe1d PE	227	atomic_inc(&qp->refcnt);
	228
	229	atomic_inc(&qp->refcnt);
	230	hlist_add_head(&qp->list, &f->hash[hash]);
3140c25c	231	list_add_tail(&qp->lru_list, &nf->lru_list);
e5a2bb84	232	nf->nqueues++;
2588fe1d PE	233	write_unlock(&f->lock);
	234	return qp;
	235	}
e521db9d	236
ac18e750 PE	237	static struct inet_frag_queue inet_frag_alloc(struct netns_frags nf,
ac18e750 PE	238	struct inet_frags f, void arg)
e521db9d PE	239	{
	240	struct inet_frag_queue *q;
	241
	242	q = kzalloc(f->qsize, GFP_ATOMIC);
	243	if (q == NULL)
	244	return NULL;
	245
c6fda282	246	f->constructor(q, arg);
6ddc0822	247	atomic_add(f->qsize, &nf->mem);
e521db9d PE	248	setup_timer(&q->timer, f->frag_expire, (unsigned long)q);
	249	spin_lock_init(&q->lock);
	250	atomic_set(&q->refcnt, 1);
ac18e750	251	q->net = nf;
e521db9d PE	252
	253	return q;
	254	}
c6fda282	255
ac18e750	256	static struct inet_frag_queue inet_frag_create(struct netns_frags nf,
9a375803	257	struct inet_frags f, void arg)
c6fda282 PE	258	{
	259	struct inet_frag_queue *q;
	260
ac18e750	261	q = inet_frag_alloc(nf, f, arg);
c6fda282 PE	262	if (q == NULL)
	263	return NULL;
	264
9a375803	265	return inet_frag_intern(nf, q, f, arg);
c6fda282	266	}
abd6523d	267
ac18e750 PE	268	struct inet_frag_queue inet_frag_find(struct netns_frags nf,
ac18e750 PE	269	struct inet_frags f, void key, unsigned int hash)
56bca31f	270	__releases(&f->lock)
abd6523d PE	271	{
	272	struct inet_frag_queue *q;
	273	struct hlist_node *n;
	274
abd6523d	275	hlist_for_each_entry(q, n, &f->hash[hash], list) {
ac18e750	276	if (q->net == nf && f->match(q, key)) {
abd6523d PE	277	atomic_inc(&q->refcnt);
	278	read_unlock(&f->lock);
	279	return q;
	280	}
	281	}
	282	read_unlock(&f->lock);
	283
9a375803	284	return inet_frag_create(nf, f, key);
abd6523d PE	285	}
abd6523d PE	286	EXPORT_SYMBOL(inet_frag_find);