[mirror_ubuntu-zesty-kernel.git] / net / core / dst.c

/*
 * net/core/dst.c	Protocol independent destination cache.
 *
 * Authors:		Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
 *
 */

#include <linux/bitops.h>
#include <linux/errno.h>
#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/module.h>
#include <linux/netdevice.h>
#include <linux/skbuff.h>
#include <linux/string.h>
#include <linux/types.h>

#include <net/dst.h>

/* Locking strategy:
 * 1) Garbage collection state of dead destination cache
 *    entries is protected by dst_lock.
 * 2) GC is run only from BH context, and is the only remover
 *    of entries.
 * 3) Entries are added to the garbage list from both BH
 *    and non-BH context, so local BH disabling is needed.
 * 4) All operations modify state, so a spinlock is used.
 */
static struct dst_entry 	*dst_garbage_list;
#if RT_CACHE_DEBUG >= 2
static atomic_t			 dst_total = ATOMIC_INIT(0);
#endif
static DEFINE_SPINLOCK(dst_lock);

static unsigned long dst_gc_timer_expires;
static unsigned long dst_gc_timer_inc = DST_GC_MAX;
static void dst_run_gc(unsigned long);
static void ___dst_free(struct dst_entry * dst);

static DEFINE_TIMER(dst_gc_timer, dst_run_gc, DST_GC_MIN, 0);

static void dst_run_gc(unsigned long dummy)
{
	int    delayed = 0;
	int    work_performed;
	struct dst_entry * dst, **dstp;

	if (!spin_trylock(&dst_lock)) {
		mod_timer(&dst_gc_timer, jiffies + HZ/10);
		return;
	}

	del_timer(&dst_gc_timer);
	dstp = &dst_garbage_list;
	work_performed = 0;
	while ((dst = *dstp) != NULL) {
		if (atomic_read(&dst->__refcnt)) {
			dstp = &dst->next;
			delayed++;
			continue;
		}
		*dstp = dst->next;
		work_performed = 1;

		dst = dst_destroy(dst);
		if (dst) {
			/* NOHASH and still referenced. Unless it is already
			 * on gc list, invalidate it and add to gc list.
			 *
			 * Note: this is temporary. Actually, NOHASH dst's
			 * must be obsoleted when parent is obsoleted.
			 * But we do not have state "obsoleted, but
			 * referenced by parent", so it is right.
			 */
			if (dst->obsolete > 1)
				continue;

			___dst_free(dst);
			dst->next = *dstp;
			*dstp = dst;
			dstp = &dst->next;
		}
	}
	if (!dst_garbage_list) {
		dst_gc_timer_inc = DST_GC_MAX;
		goto out;
	}
	if (!work_performed) {
		if ((dst_gc_timer_expires += dst_gc_timer_inc) > DST_GC_MAX)
			dst_gc_timer_expires = DST_GC_MAX;
		dst_gc_timer_inc += DST_GC_INC;
	} else {
		dst_gc_timer_inc = DST_GC_INC;
		dst_gc_timer_expires = DST_GC_MIN;
	}
#if RT_CACHE_DEBUG >= 2
	printk("dst_total: %d/%d %ld\n",
	       atomic_read(&dst_total), delayed,  dst_gc_timer_expires);
#endif
	/* if the next desired timer is more than 4 seconds in the future
	 * then round the timer to whole seconds
	 */
	if (dst_gc_timer_expires > 4*HZ)
		mod_timer(&dst_gc_timer,
			round_jiffies(jiffies + dst_gc_timer_expires));
	else
		mod_timer(&dst_gc_timer, jiffies + dst_gc_timer_expires);

out:
	spin_unlock(&dst_lock);
}

static int dst_discard_in(struct sk_buff *skb)
{
	kfree_skb(skb);
	return 0;
}

static int dst_discard_out(struct sk_buff *skb)
{
	kfree_skb(skb);
	return 0;
}

void * dst_alloc(struct dst_ops * ops)
{
	struct dst_entry * dst;

	if (ops->gc && atomic_read(&ops->entries) > ops->gc_thresh) {
		if (ops->gc())
			return NULL;
	}
	dst = kmem_cache_zalloc(ops->kmem_cachep, GFP_ATOMIC);
	if (!dst)
		return NULL;
	atomic_set(&dst->__refcnt, 0);
	dst->ops = ops;
	dst->lastuse = jiffies;
	dst->path = dst;
	dst->input = dst_discard_in;
	dst->output = dst_discard_out;
#if RT_CACHE_DEBUG >= 2
	atomic_inc(&dst_total);
#endif
	atomic_inc(&ops->entries);
	return dst;
}

static void ___dst_free(struct dst_entry * dst)
{
	/* The first case (dev==NULL) is required, when
	   protocol module is unloaded.
	 */
	if (dst->dev == NULL || !(dst->dev->flags&IFF_UP)) {
		dst->input = dst_discard_in;
		dst->output = dst_discard_out;
	}
	dst->obsolete = 2;
}

void __dst_free(struct dst_entry * dst)
{
	spin_lock_bh(&dst_lock);
	___dst_free(dst);
	dst->next = dst_garbage_list;
	dst_garbage_list = dst;
	if (dst_gc_timer_inc > DST_GC_INC) {
		dst_gc_timer_inc = DST_GC_INC;
		dst_gc_timer_expires = DST_GC_MIN;
		mod_timer(&dst_gc_timer, jiffies + dst_gc_timer_expires);
	}
	spin_unlock_bh(&dst_lock);
}

struct dst_entry *dst_destroy(struct dst_entry * dst)
{
	struct dst_entry *child;
	struct neighbour *neigh;
	struct hh_cache *hh;

	smp_rmb();

again:
	neigh = dst->neighbour;
	hh = dst->hh;
	child = dst->child;

	dst->hh = NULL;
	if (hh && atomic_dec_and_test(&hh->hh_refcnt))
		kfree(hh);

	if (neigh) {
		dst->neighbour = NULL;
		neigh_release(neigh);
	}

	atomic_dec(&dst->ops->entries);

	if (dst->ops->destroy)
		dst->ops->destroy(dst);
	if (dst->dev)
		dev_put(dst->dev);
#if RT_CACHE_DEBUG >= 2
	atomic_dec(&dst_total);
#endif
	kmem_cache_free(dst->ops->kmem_cachep, dst);

	dst = child;
	if (dst) {
		int nohash = dst->flags & DST_NOHASH;

		if (atomic_dec_and_test(&dst->__refcnt)) {
			/* We were real parent of this dst, so kill child. */
			if (nohash)
				goto again;
		} else {
			/* Child is still referenced, return it for freeing. */
			if (nohash)
				return dst;
			/* Child is still in his hash table */
		}
	}
	return NULL;
}

/* Dirty hack. We did it in 2.2 (in __dst_free),
 * we have _very_ good reasons not to repeat
 * this mistake in 2.3, but we have no choice
 * now. _It_ _is_ _explicit_ _deliberate_
 * _race_ _condition_.
 *
 * Commented and originally written by Alexey.
 */
static inline void dst_ifdown(struct dst_entry *dst, struct net_device *dev,
			      int unregister)
{
	if (dst->ops->ifdown)
		dst->ops->ifdown(dst, dev, unregister);

	if (dev != dst->dev)
		return;

	if (!unregister) {
		dst->input = dst_discard_in;
		dst->output = dst_discard_out;
	} else {
		dst->dev = &loopback_dev;
		dev_hold(&loopback_dev);
		dev_put(dev);
		if (dst->neighbour && dst->neighbour->dev == dev) {
			dst->neighbour->dev = &loopback_dev;
			dev_put(dev);
			dev_hold(&loopback_dev);
		}
	}
}

static int dst_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
{
	struct net_device *dev = ptr;
	struct dst_entry *dst;

	switch (event) {
	case NETDEV_UNREGISTER:
	case NETDEV_DOWN:
		spin_lock_bh(&dst_lock);
		for (dst = dst_garbage_list; dst; dst = dst->next) {
			dst_ifdown(dst, dev, event != NETDEV_DOWN);
		}
		spin_unlock_bh(&dst_lock);
		break;
	}
	return NOTIFY_DONE;
}

static struct notifier_block dst_dev_notifier = {
	.notifier_call	= dst_dev_event,
};

void __init dst_init(void)
{
	register_netdevice_notifier(&dst_dev_notifier);
}

EXPORT_SYMBOL(__dst_free);
EXPORT_SYMBOL(dst_alloc);
EXPORT_SYMBOL(dst_destroy);
Commit	Line	Data
1da177e4 LT	1	/*
	2	* net/core/dst.c Protocol independent destination cache.
	3	*
	4	* Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
	5	*
	6	*/
	7
	8	#include <linux/bitops.h>
	9	#include <linux/errno.h>
	10	#include <linux/init.h>
	11	#include <linux/kernel.h>
	12	#include <linux/mm.h>
	13	#include <linux/module.h>
	14	#include <linux/netdevice.h>
1da177e4 LT	15	#include <linux/skbuff.h>
	16	#include <linux/string.h>
	17	#include <linux/types.h>
	18
	19	#include <net/dst.h>
	20
	21	/* Locking strategy:
	22	* 1) Garbage collection state of dead destination cache
	23	* entries is protected by dst_lock.
	24	* 2) GC is run only from BH context, and is the only remover
	25	* of entries.
	26	* 3) Entries are added to the garbage list from both BH
	27	* and non-BH context, so local BH disabling is needed.
	28	* 4) All operations modify state, so a spinlock is used.
	29	*/
	30	static struct dst_entry *dst_garbage_list;
4ec93edb	31	#if RT_CACHE_DEBUG >= 2
1da177e4 LT	32	static atomic_t dst_total = ATOMIC_INIT(0);
	33	#endif
	34	static DEFINE_SPINLOCK(dst_lock);
	35
	36	static unsigned long dst_gc_timer_expires;
	37	static unsigned long dst_gc_timer_inc = DST_GC_MAX;
	38	static void dst_run_gc(unsigned long);
	39	static void ___dst_free(struct dst_entry * dst);
	40
8d06afab	41	static DEFINE_TIMER(dst_gc_timer, dst_run_gc, DST_GC_MIN, 0);
1da177e4 LT	42
	43	static void dst_run_gc(unsigned long dummy)
	44	{
	45	int delayed = 0;
f0098f78	46	int work_performed;
1da177e4 LT	47	struct dst_entry * dst, **dstp;
	48
	49	if (!spin_trylock(&dst_lock)) {
	50	mod_timer(&dst_gc_timer, jiffies + HZ/10);
	51	return;
	52	}
	53
1da177e4 LT	54	del_timer(&dst_gc_timer);
1da177e4 LT	55	dstp = &dst_garbage_list;
f0098f78	56	work_performed = 0;
1da177e4 LT	57	while ((dst = *dstp) != NULL) {
	58	if (atomic_read(&dst->__refcnt)) {
	59	dstp = &dst->next;
	60	delayed++;
	61	continue;
	62	}
	63	*dstp = dst->next;
f0098f78	64	work_performed = 1;
1da177e4 LT	65
	66	dst = dst_destroy(dst);
	67	if (dst) {
	68	/* NOHASH and still referenced. Unless it is already
	69	* on gc list, invalidate it and add to gc list.
	70	*
	71	* Note: this is temporary. Actually, NOHASH dst's
	72	* must be obsoleted when parent is obsoleted.
	73	* But we do not have state "obsoleted, but
	74	* referenced by parent", so it is right.
	75	*/
	76	if (dst->obsolete > 1)
	77	continue;
	78
	79	___dst_free(dst);
	80	dst->next = *dstp;
	81	*dstp = dst;
	82	dstp = &dst->next;
	83	}
	84	}
	85	if (!dst_garbage_list) {
	86	dst_gc_timer_inc = DST_GC_MAX;
	87	goto out;
	88	}
f0098f78 DL	89	if (!work_performed) {
	90	if ((dst_gc_timer_expires += dst_gc_timer_inc) > DST_GC_MAX)
	91	dst_gc_timer_expires = DST_GC_MAX;
	92	dst_gc_timer_inc += DST_GC_INC;
	93	} else {
	94	dst_gc_timer_inc = DST_GC_INC;
	95	dst_gc_timer_expires = DST_GC_MIN;
	96	}
1da177e4 LT	97	#if RT_CACHE_DEBUG >= 2
	98	printk("dst_total: %d/%d %ld\n",
	99	atomic_read(&dst_total), delayed, dst_gc_timer_expires);
	100	#endif
f5a6e01c AV	101	/* if the next desired timer is more than 4 seconds in the future
	102	* then round the timer to whole seconds
	103	*/
	104	if (dst_gc_timer_expires > 4*HZ)
	105	mod_timer(&dst_gc_timer,
	106	round_jiffies(jiffies + dst_gc_timer_expires));
	107	else
	108	mod_timer(&dst_gc_timer, jiffies + dst_gc_timer_expires);
1da177e4 LT	109
	110	out:
	111	spin_unlock(&dst_lock);
	112	}
	113
	114	static int dst_discard_in(struct sk_buff *skb)
	115	{
	116	kfree_skb(skb);
	117	return 0;
	118	}
	119
	120	static int dst_discard_out(struct sk_buff *skb)
	121	{
	122	kfree_skb(skb);
	123	return 0;
	124	}
	125
	126	void * dst_alloc(struct dst_ops * ops)
	127	{
	128	struct dst_entry * dst;
	129
	130	if (ops->gc && atomic_read(&ops->entries) > ops->gc_thresh) {
	131	if (ops->gc())
	132	return NULL;
	133	}
c3762229	134	dst = kmem_cache_zalloc(ops->kmem_cachep, GFP_ATOMIC);
1da177e4 LT	135	if (!dst)
1da177e4 LT	136	return NULL;
1da177e4 LT	137	atomic_set(&dst->__refcnt, 0);
	138	dst->ops = ops;
	139	dst->lastuse = jiffies;
	140	dst->path = dst;
	141	dst->input = dst_discard_in;
	142	dst->output = dst_discard_out;
4ec93edb	143	#if RT_CACHE_DEBUG >= 2
1da177e4 LT	144	atomic_inc(&dst_total);
	145	#endif
	146	atomic_inc(&ops->entries);
	147	return dst;
	148	}
	149
	150	static void ___dst_free(struct dst_entry * dst)
	151	{
	152	/* The first case (dev==NULL) is required, when
	153	protocol module is unloaded.
	154	*/
	155	if (dst->dev == NULL \|\| !(dst->dev->flags&IFF_UP)) {
	156	dst->input = dst_discard_in;
	157	dst->output = dst_discard_out;
	158	}
	159	dst->obsolete = 2;
	160	}
	161
	162	void __dst_free(struct dst_entry * dst)
	163	{
	164	spin_lock_bh(&dst_lock);
	165	___dst_free(dst);
	166	dst->next = dst_garbage_list;
	167	dst_garbage_list = dst;
	168	if (dst_gc_timer_inc > DST_GC_INC) {
	169	dst_gc_timer_inc = DST_GC_INC;
	170	dst_gc_timer_expires = DST_GC_MIN;
	171	mod_timer(&dst_gc_timer, jiffies + dst_gc_timer_expires);
	172	}
	173	spin_unlock_bh(&dst_lock);
	174	}
	175
	176	struct dst_entry dst_destroy(struct dst_entry dst)
	177	{
	178	struct dst_entry *child;
	179	struct neighbour *neigh;
	180	struct hh_cache *hh;
	181
	182	smp_rmb();
	183
	184	again:
	185	neigh = dst->neighbour;
	186	hh = dst->hh;
	187	child = dst->child;
	188
	189	dst->hh = NULL;
	190	if (hh && atomic_dec_and_test(&hh->hh_refcnt))
	191	kfree(hh);
	192
	193	if (neigh) {
	194	dst->neighbour = NULL;
	195	neigh_release(neigh);
	196	}
	197
	198	atomic_dec(&dst->ops->entries);
	199
	200	if (dst->ops->destroy)
	201	dst->ops->destroy(dst);
	202	if (dst->dev)
	203	dev_put(dst->dev);
4ec93edb	204	#if RT_CACHE_DEBUG >= 2
1da177e4 LT	205	atomic_dec(&dst_total);
	206	#endif
	207	kmem_cache_free(dst->ops->kmem_cachep, dst);
	208
	209	dst = child;
	210	if (dst) {
6775cab9 HX	211	int nohash = dst->flags & DST_NOHASH;
6775cab9 HX	212
1da177e4 LT	213	if (atomic_dec_and_test(&dst->__refcnt)) {
1da177e4 LT	214	/* We were real parent of this dst, so kill child. */
6775cab9	215	if (nohash)
1da177e4 LT	216	goto again;
	217	} else {
	218	/* Child is still referenced, return it for freeing. */
6775cab9	219	if (nohash)
1da177e4 LT	220	return dst;
	221	/* Child is still in his hash table */
	222	}
	223	}
	224	return NULL;
	225	}
	226
	227	/* Dirty hack. We did it in 2.2 (in __dst_free),
	228	* we have _very_ good reasons not to repeat
	229	* this mistake in 2.3, but we have no choice
	230	* now. _It_ _is_ _explicit_ _deliberate_
	231	* _race_ _condition_.
	232	*
	233	* Commented and originally written by Alexey.
	234	*/
	235	static inline void dst_ifdown(struct dst_entry dst, struct net_device dev,
	236	int unregister)
	237	{
	238	if (dst->ops->ifdown)
	239	dst->ops->ifdown(dst, dev, unregister);
	240
	241	if (dev != dst->dev)
	242	return;
	243
	244	if (!unregister) {
	245	dst->input = dst_discard_in;
	246	dst->output = dst_discard_out;
	247	} else {
	248	dst->dev = &loopback_dev;
	249	dev_hold(&loopback_dev);
	250	dev_put(dev);
	251	if (dst->neighbour && dst->neighbour->dev == dev) {
	252	dst->neighbour->dev = &loopback_dev;
	253	dev_put(dev);
	254	dev_hold(&loopback_dev);
	255	}
	256	}
	257	}
	258
	259	static int dst_dev_event(struct notifier_block this, unsigned long event, void ptr)
	260	{
	261	struct net_device *dev = ptr;
	262	struct dst_entry *dst;
	263
	264	switch (event) {
	265	case NETDEV_UNREGISTER:
	266	case NETDEV_DOWN:
	267	spin_lock_bh(&dst_lock);
	268	for (dst = dst_garbage_list; dst; dst = dst->next) {
	269	dst_ifdown(dst, dev, event != NETDEV_DOWN);
	270	}
	271	spin_unlock_bh(&dst_lock);
	272	break;
	273	}
	274	return NOTIFY_DONE;
	275	}
	276
	277	static struct notifier_block dst_dev_notifier = {
	278	.notifier_call = dst_dev_event,
	279	};
	280
	281	void __init dst_init(void)
	282	{
	283	register_netdevice_notifier(&dst_dev_notifier);
284	}
285
286	EXPORT_SYMBOL(__dst_free);
287	EXPORT_SYMBOL(dst_alloc);
288	EXPORT_SYMBOL(dst_destroy);