[linux-2.6-block.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/workqueue.h>
#include <linux/mm.h>
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/netdevice.h>
#include <linux/skbuff.h>
#include <linux/string.h>
#include <linux/types.h>
#include <net/net_namespace.h>
#include <linux/sched.h>
#include <linux/prefetch.h>
#include <net/lwtunnel.h>

#include <net/dst.h>
#include <net/dst_metadata.h>

/*
 * Theory of operations:
 * 1) We use a list, protected by a spinlock, to add
 *    new entries from both BH and non-BH context.
 * 2) In order to keep spinlock held for a small delay,
 *    we use a second list where are stored long lived
 *    entries, that are handled by the garbage collect thread
 *    fired by a workqueue.
 * 3) This list is guarded by a mutex,
 *    so that the gc_task and dst_dev_event() can be synchronized.
 */

/*
 * We want to keep lock & list close together
 * to dirty as few cache lines as possible in __dst_free().
 * As this is not a very strong hint, we dont force an alignment on SMP.
 */
static struct {
	spinlock_t		lock;
	struct dst_entry	*list;
	unsigned long		timer_inc;
	unsigned long		timer_expires;
} dst_garbage = {
	.lock = __SPIN_LOCK_UNLOCKED(dst_garbage.lock),
	.timer_inc = DST_GC_MAX,
};
static void dst_gc_task(struct work_struct *work);
static void ___dst_free(struct dst_entry *dst);

static DECLARE_DELAYED_WORK(dst_gc_work, dst_gc_task);

static DEFINE_MUTEX(dst_gc_mutex);
/*
 * long lived entries are maintained in this list, guarded by dst_gc_mutex
 */
static struct dst_entry         *dst_busy_list;

static void dst_gc_task(struct work_struct *work)
{
	int    delayed = 0;
	int    work_performed = 0;
	unsigned long expires = ~0L;
	struct dst_entry *dst, *next, head;
	struct dst_entry *last = &head;

	mutex_lock(&dst_gc_mutex);
	next = dst_busy_list;

loop:
	while ((dst = next) != NULL) {
		next = dst->next;
		prefetch(&next->next);
		cond_resched();
		if (likely(atomic_read(&dst->__refcnt))) {
			last->next = dst;
			last = dst;
			delayed++;
			continue;
		}
		work_performed++;

		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 > 0)
				continue;

			___dst_free(dst);
			dst->next = next;
			next = dst;
		}
	}

	spin_lock_bh(&dst_garbage.lock);
	next = dst_garbage.list;
	if (next) {
		dst_garbage.list = NULL;
		spin_unlock_bh(&dst_garbage.lock);
		goto loop;
	}
	last->next = NULL;
	dst_busy_list = head.next;
	if (!dst_busy_list)
		dst_garbage.timer_inc = DST_GC_MAX;
	else {
		/*
		 * if we freed less than 1/10 of delayed entries,
		 * we can sleep longer.
		 */
		if (work_performed <= delayed/10) {
			dst_garbage.timer_expires += dst_garbage.timer_inc;
			if (dst_garbage.timer_expires > DST_GC_MAX)
				dst_garbage.timer_expires = DST_GC_MAX;
			dst_garbage.timer_inc += DST_GC_INC;
		} else {
			dst_garbage.timer_inc = DST_GC_INC;
			dst_garbage.timer_expires = DST_GC_MIN;
		}
		expires = dst_garbage.timer_expires;
		/*
		 * if the next desired timer is more than 4 seconds in the
		 * future then round the timer to whole seconds
		 */
		if (expires > 4*HZ)
			expires = round_jiffies_relative(expires);
		schedule_delayed_work(&dst_gc_work, expires);
	}

	spin_unlock_bh(&dst_garbage.lock);
	mutex_unlock(&dst_gc_mutex);
}

int dst_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
{
	kfree_skb(skb);
	return 0;
}
EXPORT_SYMBOL(dst_discard_out);

const struct dst_metrics dst_default_metrics = {
	/* This initializer is needed to force linker to place this variable
	 * into const section. Otherwise it might end into bss section.
	 * We really want to avoid false sharing on this variable, and catch
	 * any writes on it.
	 */
	.refcnt = ATOMIC_INIT(1),
};

void dst_init(struct dst_entry *dst, struct dst_ops *ops,
	      struct net_device *dev, int initial_ref, int initial_obsolete,
	      unsigned short flags)
{
	dst->child = NULL;
	dst->dev = dev;
	if (dev)
		dev_hold(dev);
	dst->ops = ops;
	dst_init_metrics(dst, dst_default_metrics.metrics, true);
	dst->expires = 0UL;
	dst->path = dst;
	dst->from = NULL;
#ifdef CONFIG_XFRM
	dst->xfrm = NULL;
#endif
	dst->input = dst_discard;
	dst->output = dst_discard_out;
	dst->error = 0;
	dst->obsolete = initial_obsolete;
	dst->header_len = 0;
	dst->trailer_len = 0;
#ifdef CONFIG_IP_ROUTE_CLASSID
	dst->tclassid = 0;
#endif
	dst->lwtstate = NULL;
	atomic_set(&dst->__refcnt, initial_ref);
	dst->__use = 0;
	dst->lastuse = jiffies;
	dst->flags = flags;
	dst->next = NULL;
	if (!(flags & DST_NOCOUNT))
		dst_entries_add(ops, 1);
}
EXPORT_SYMBOL(dst_init);

void *dst_alloc(struct dst_ops *ops, struct net_device *dev,
		int initial_ref, int initial_obsolete, unsigned short flags)
{
	struct dst_entry *dst;

	if (ops->gc && dst_entries_get_fast(ops) > ops->gc_thresh) {
		if (ops->gc(ops))
			return NULL;
	}

	dst = kmem_cache_alloc(ops->kmem_cachep, GFP_ATOMIC);
	if (!dst)
		return NULL;

	dst_init(dst, ops, dev, initial_ref, initial_obsolete, flags);

	return dst;
}
EXPORT_SYMBOL(dst_alloc);

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;
		dst->output = dst_discard_out;
	}
	dst->obsolete = DST_OBSOLETE_DEAD;
}

void __dst_free(struct dst_entry *dst)
{
	spin_lock_bh(&dst_garbage.lock);
	___dst_free(dst);
	dst->next = dst_garbage.list;
	dst_garbage.list = dst;
	if (dst_garbage.timer_inc > DST_GC_INC) {
		dst_garbage.timer_inc = DST_GC_INC;
		dst_garbage.timer_expires = DST_GC_MIN;
		mod_delayed_work(system_wq, &dst_gc_work,
				 dst_garbage.timer_expires);
	}
	spin_unlock_bh(&dst_garbage.lock);
}
EXPORT_SYMBOL(__dst_free);

struct dst_entry *dst_destroy(struct dst_entry * dst)
{
	struct dst_entry *child;

	smp_rmb();

again:
	child = dst->child;

	if (!(dst->flags & DST_NOCOUNT))
		dst_entries_add(dst->ops, -1);

	if (dst->ops->destroy)
		dst->ops->destroy(dst);
	if (dst->dev)
		dev_put(dst->dev);

	lwtstate_put(dst->lwtstate);

	if (dst->flags & DST_METADATA)
		metadata_dst_free((struct metadata_dst *)dst);
	else
		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;
}
EXPORT_SYMBOL(dst_destroy);

static void dst_destroy_rcu(struct rcu_head *head)
{
	struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);

	dst = dst_destroy(dst);
	if (dst)
		__dst_free(dst);
}

void dst_release(struct dst_entry *dst)
{
	if (dst) {
		int newrefcnt;
		unsigned short nocache = dst->flags & DST_NOCACHE;

		newrefcnt = atomic_dec_return(&dst->__refcnt);
		if (unlikely(newrefcnt < 0))
			net_warn_ratelimited("%s: dst:%p refcnt:%d\n",
					     __func__, dst, newrefcnt);
		if (!newrefcnt && unlikely(nocache))
			call_rcu(&dst->rcu_head, dst_destroy_rcu);
	}
}
EXPORT_SYMBOL(dst_release);

u32 *dst_cow_metrics_generic(struct dst_entry *dst, unsigned long old)
{
	struct dst_metrics *p = kmalloc(sizeof(*p), GFP_ATOMIC);

	if (p) {
		struct dst_metrics *old_p = (struct dst_metrics *)__DST_METRICS_PTR(old);
		unsigned long prev, new;

		atomic_set(&p->refcnt, 1);
		memcpy(p->metrics, old_p->metrics, sizeof(p->metrics));

		new = (unsigned long) p;
		prev = cmpxchg(&dst->_metrics, old, new);

		if (prev != old) {
			kfree(p);
			p = (struct dst_metrics *)__DST_METRICS_PTR(prev);
			if (prev & DST_METRICS_READ_ONLY)
				p = NULL;
		} else if (prev & DST_METRICS_REFCOUNTED) {
			if (atomic_dec_and_test(&old_p->refcnt))
				kfree(old_p);
		}
	}
	BUILD_BUG_ON(offsetof(struct dst_metrics, metrics) != 0);
	return (u32 *)p;
}
EXPORT_SYMBOL(dst_cow_metrics_generic);

/* Caller asserts that dst_metrics_read_only(dst) is false.  */
void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old)
{
	unsigned long prev, new;

	new = ((unsigned long) &dst_default_metrics) | DST_METRICS_READ_ONLY;
	prev = cmpxchg(&dst->_metrics, old, new);
	if (prev == old)
		kfree(__DST_METRICS_PTR(old));
}
EXPORT_SYMBOL(__dst_destroy_metrics_generic);

static struct dst_ops md_dst_ops = {
	.family =		AF_UNSPEC,
};

static int dst_md_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
{
	WARN_ONCE(1, "Attempting to call output on metadata dst\n");
	kfree_skb(skb);
	return 0;
}

static int dst_md_discard(struct sk_buff *skb)
{
	WARN_ONCE(1, "Attempting to call input on metadata dst\n");
	kfree_skb(skb);
	return 0;
}

static void __metadata_dst_init(struct metadata_dst *md_dst, u8 optslen)
{
	struct dst_entry *dst;

	dst = &md_dst->dst;
	dst_init(dst, &md_dst_ops, NULL, 1, DST_OBSOLETE_NONE,
		 DST_METADATA | DST_NOCACHE | DST_NOCOUNT);

	dst->input = dst_md_discard;
	dst->output = dst_md_discard_out;

	memset(dst + 1, 0, sizeof(*md_dst) + optslen - sizeof(*dst));
}

struct metadata_dst *metadata_dst_alloc(u8 optslen, gfp_t flags)
{
	struct metadata_dst *md_dst;

	md_dst = kmalloc(sizeof(*md_dst) + optslen, flags);
	if (!md_dst)
		return NULL;

	__metadata_dst_init(md_dst, optslen);

	return md_dst;
}
EXPORT_SYMBOL_GPL(metadata_dst_alloc);

void metadata_dst_free(struct metadata_dst *md_dst)
{
#ifdef CONFIG_DST_CACHE
	dst_cache_destroy(&md_dst->u.tun_info.dst_cache);
#endif
	kfree(md_dst);
}

struct metadata_dst __percpu *metadata_dst_alloc_percpu(u8 optslen, gfp_t flags)
{
	int cpu;
	struct metadata_dst __percpu *md_dst;

	md_dst = __alloc_percpu_gfp(sizeof(struct metadata_dst) + optslen,
				    __alignof__(struct metadata_dst), flags);
	if (!md_dst)
		return NULL;

	for_each_possible_cpu(cpu)
		__metadata_dst_init(per_cpu_ptr(md_dst, cpu), optslen);

	return md_dst;
}
EXPORT_SYMBOL_GPL(metadata_dst_alloc_percpu);

/* 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 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;
		dst->output = dst_discard_out;
	} else {
		dst->dev = dev_net(dst->dev)->loopback_dev;
		dev_hold(dst->dev);
		dev_put(dev);
	}
}

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

	switch (event) {
	case NETDEV_UNREGISTER_FINAL:
	case NETDEV_DOWN:
		mutex_lock(&dst_gc_mutex);
		for (dst = dst_busy_list; dst; dst = dst->next) {
			last = dst;
			dst_ifdown(dst, dev, event != NETDEV_DOWN);
		}

		spin_lock_bh(&dst_garbage.lock);
		dst = dst_garbage.list;
		dst_garbage.list = NULL;
		/* The code in dst_ifdown places a hold on the loopback device.
		 * If the gc entry processing is set to expire after a lengthy
		 * interval, this hold can cause netdev_wait_allrefs() to hang
		 * out and wait for a long time -- until the the loopback
		 * interface is released.  If we're really unlucky, it'll emit
		 * pr_emerg messages to console too.  Reset the interval here,
		 * so dst cleanups occur in a more timely fashion.
		 */
		if (dst_garbage.timer_inc > DST_GC_INC) {
			dst_garbage.timer_inc = DST_GC_INC;
			dst_garbage.timer_expires = DST_GC_MIN;
			mod_delayed_work(system_wq, &dst_gc_work,
					 dst_garbage.timer_expires);
		}
		spin_unlock_bh(&dst_garbage.lock);

		if (last)
			last->next = dst;
		else
			dst_busy_list = dst;
		for (; dst; dst = dst->next)
			dst_ifdown(dst, dev, event != NETDEV_DOWN);
		mutex_unlock(&dst_gc_mutex);
		break;
	}
	return NOTIFY_DONE;
}

static struct notifier_block dst_dev_notifier = {
	.notifier_call	= dst_dev_event,
	.priority = -10, /* must be called after other network notifiers */
};

void __init dst_subsys_init(void)
{
	register_netdevice_notifier(&dst_dev_notifier);
}
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>
86bba269	12	#include <linux/workqueue.h>
1da177e4 LT	13	#include <linux/mm.h>
1da177e4 LT	14	#include <linux/module.h>
5a0e3ad6	15	#include <linux/slab.h>
1da177e4	16	#include <linux/netdevice.h>
1da177e4 LT	17	#include <linux/skbuff.h>
	18	#include <linux/string.h>
	19	#include <linux/types.h>
e9dc8653	20	#include <net/net_namespace.h>
2fc1b5dd	21	#include <linux/sched.h>
268bb0ce	22	#include <linux/prefetch.h>
61adedf3	23	#include <net/lwtunnel.h>
1da177e4 LT	24
1da177e4 LT	25	#include <net/dst.h>
f38a9eb1	26	#include <net/dst_metadata.h>
1da177e4	27
86bba269 ED	28	/*
	29	* Theory of operations:
	30	* 1) We use a list, protected by a spinlock, to add
	31	* new entries from both BH and non-BH context.
	32	* 2) In order to keep spinlock held for a small delay,
	33	* we use a second list where are stored long lived
	34	* entries, that are handled by the garbage collect thread
	35	* fired by a workqueue.
	36	* 3) This list is guarded by a mutex,
	37	* so that the gc_task and dst_dev_event() can be synchronized.
1da177e4	38	*/
1da177e4	39
86bba269 ED	40	/*
	41	* We want to keep lock & list close together
	42	* to dirty as few cache lines as possible in __dst_free().
	43	* As this is not a very strong hint, we dont force an alignment on SMP.
	44	*/
	45	static struct {
	46	spinlock_t lock;
598ed936	47	struct dst_entry *list;
86bba269 ED	48	unsigned long timer_inc;
	49	unsigned long timer_expires;
	50	} dst_garbage = {
	51	.lock = __SPIN_LOCK_UNLOCKED(dst_garbage.lock),
	52	.timer_inc = DST_GC_MAX,
	53	};
	54	static void dst_gc_task(struct work_struct *work);
598ed936	55	static void ___dst_free(struct dst_entry *dst);
1da177e4	56
86bba269	57	static DECLARE_DELAYED_WORK(dst_gc_work, dst_gc_task);
1da177e4	58
86bba269 ED	59	static DEFINE_MUTEX(dst_gc_mutex);
	60	/*
	61	* long lived entries are maintained in this list, guarded by dst_gc_mutex
	62	*/
	63	static struct dst_entry *dst_busy_list;
	64
	65	static void dst_gc_task(struct work_struct *work)
1da177e4 LT	66	{
1da177e4 LT	67	int delayed = 0;
86bba269 ED	68	int work_performed = 0;
	69	unsigned long expires = ~0L;
	70	struct dst_entry dst, next, head;
	71	struct dst_entry *last = &head;
1da177e4	72
86bba269 ED	73	mutex_lock(&dst_gc_mutex);
86bba269 ED	74	next = dst_busy_list;
1da177e4	75
86bba269 ED	76	loop:
	77	while ((dst = next) != NULL) {
	78	next = dst->next;
	79	prefetch(&next->next);
2fc1b5dd	80	cond_resched();
86bba269 ED	81	if (likely(atomic_read(&dst->__refcnt))) {
	82	last->next = dst;
	83	last = dst;
1da177e4 LT	84	delayed++;
	85	continue;
	86	}
86bba269	87	work_performed++;
1da177e4 LT	88
	89	dst = dst_destroy(dst);
	90	if (dst) {
	91	/* NOHASH and still referenced. Unless it is already
	92	* on gc list, invalidate it and add to gc list.
	93	*
	94	* Note: this is temporary. Actually, NOHASH dst's
	95	* must be obsoleted when parent is obsoleted.
	96	* But we do not have state "obsoleted, but
	97	* referenced by parent", so it is right.
	98	*/
f5b0a874	99	if (dst->obsolete > 0)
1da177e4 LT	100	continue;
	101
	102	___dst_free(dst);
86bba269 ED	103	dst->next = next;
86bba269 ED	104	next = dst;
1da177e4 LT	105	}
1da177e4 LT	106	}
86bba269 ED	107
	108	spin_lock_bh(&dst_garbage.lock);
	109	next = dst_garbage.list;
	110	if (next) {
	111	dst_garbage.list = NULL;
	112	spin_unlock_bh(&dst_garbage.lock);
	113	goto loop;
1da177e4	114	}
86bba269 ED	115	last->next = NULL;
	116	dst_busy_list = head.next;
	117	if (!dst_busy_list)
	118	dst_garbage.timer_inc = DST_GC_MAX;
	119	else {
	120	/*
	121	* if we freed less than 1/10 of delayed entries,
	122	* we can sleep longer.
	123	*/
	124	if (work_performed <= delayed/10) {
	125	dst_garbage.timer_expires += dst_garbage.timer_inc;
	126	if (dst_garbage.timer_expires > DST_GC_MAX)
	127	dst_garbage.timer_expires = DST_GC_MAX;
	128	dst_garbage.timer_inc += DST_GC_INC;
	129	} else {
	130	dst_garbage.timer_inc = DST_GC_INC;
	131	dst_garbage.timer_expires = DST_GC_MIN;
	132	}
	133	expires = dst_garbage.timer_expires;
	134	/*
598ed936	135	* if the next desired timer is more than 4 seconds in the
598ed936	136	* future then round the timer to whole seconds
86bba269 ED	137	*/
	138	if (expires > 4*HZ)
	139	expires = round_jiffies_relative(expires);
	140	schedule_delayed_work(&dst_gc_work, expires);
f0098f78	141	}
86bba269 ED	142
	143	spin_unlock_bh(&dst_garbage.lock);
	144	mutex_unlock(&dst_gc_mutex);
1da177e4 LT	145	}
1da177e4 LT	146
ede2059d	147	int dst_discard_out(struct net net, struct sock sk, struct sk_buff *skb)
1da177e4 LT	148	{
	149	kfree_skb(skb);
	150	return 0;
	151	}
ede2059d	152	EXPORT_SYMBOL(dst_discard_out);
1da177e4	153
3fb07daf	154	const struct dst_metrics dst_default_metrics = {
a37e6e34 ED	155	/* This initializer is needed to force linker to place this variable
	156	* into const section. Otherwise it might end into bss section.
	157	* We really want to avoid false sharing on this variable, and catch
	158	* any writes on it.
	159	*/
3fb07daf	160	.refcnt = ATOMIC_INIT(1),
a37e6e34 ED	161	};
a37e6e34 ED	162
f38a9eb1 TG	163	void dst_init(struct dst_entry dst, struct dst_ops ops,
	164	struct net_device *dev, int initial_ref, int initial_obsolete,
	165	unsigned short flags)
1da177e4	166	{
cf911662	167	dst->child = NULL;
5c1e6aa3 DM	168	dst->dev = dev;
	169	if (dev)
	170	dev_hold(dev);
1da177e4	171	dst->ops = ops;
3fb07daf	172	dst_init_metrics(dst, dst_default_metrics.metrics, true);
cf911662	173	dst->expires = 0UL;
5c1e6aa3	174	dst->path = dst;
ecd98837	175	dst->from = NULL;
cf911662 DM	176	#ifdef CONFIG_XFRM
	177	dst->xfrm = NULL;
	178	#endif
5c1e6aa3	179	dst->input = dst_discard;
ede2059d	180	dst->output = dst_discard_out;
cf911662	181	dst->error = 0;
5c1e6aa3	182	dst->obsolete = initial_obsolete;
cf911662 DM	183	dst->header_len = 0;
	184	dst->trailer_len = 0;
	185	#ifdef CONFIG_IP_ROUTE_CLASSID
	186	dst->tclassid = 0;
1da177e4	187	#endif
61adedf3	188	dst->lwtstate = NULL;
5c1e6aa3	189	atomic_set(&dst->__refcnt, initial_ref);
cf911662	190	dst->__use = 0;
5c1e6aa3 DM	191	dst->lastuse = jiffies;
5c1e6aa3 DM	192	dst->flags = flags;
cf911662	193	dst->next = NULL;
957c665f DM	194	if (!(flags & DST_NOCOUNT))
957c665f DM	195	dst_entries_add(ops, 1);
f38a9eb1 TG	196	}
	197	EXPORT_SYMBOL(dst_init);
	198
	199	void dst_alloc(struct dst_ops ops, struct net_device *dev,
	200	int initial_ref, int initial_obsolete, unsigned short flags)
	201	{
	202	struct dst_entry *dst;
	203
	204	if (ops->gc && dst_entries_get_fast(ops) > ops->gc_thresh) {
	205	if (ops->gc(ops))
	206	return NULL;
	207	}
	208
	209	dst = kmem_cache_alloc(ops->kmem_cachep, GFP_ATOMIC);
	210	if (!dst)
	211	return NULL;
	212
	213	dst_init(dst, ops, dev, initial_ref, initial_obsolete, flags);
	214
1da177e4 LT	215	return dst;
1da177e4 LT	216	}
598ed936	217	EXPORT_SYMBOL(dst_alloc);
1da177e4	218
598ed936	219	static void ___dst_free(struct dst_entry *dst)
1da177e4 LT	220	{
	221	/* The first case (dev==NULL) is required, when
	222	protocol module is unloaded.
	223	*/
aad88724 ED	224	if (dst->dev == NULL \|\| !(dst->dev->flags&IFF_UP)) {
aad88724 ED	225	dst->input = dst_discard;
ede2059d	226	dst->output = dst_discard_out;
aad88724	227	}
f5b0a874	228	dst->obsolete = DST_OBSOLETE_DEAD;
1da177e4 LT	229	}
1da177e4 LT	230
598ed936	231	void __dst_free(struct dst_entry *dst)
1da177e4	232	{
86bba269	233	spin_lock_bh(&dst_garbage.lock);
1da177e4	234	___dst_free(dst);
86bba269 ED	235	dst->next = dst_garbage.list;
	236	dst_garbage.list = dst;
	237	if (dst_garbage.timer_inc > DST_GC_INC) {
	238	dst_garbage.timer_inc = DST_GC_INC;
	239	dst_garbage.timer_expires = DST_GC_MIN;
41f63c53 TH	240	mod_delayed_work(system_wq, &dst_gc_work,
41f63c53 TH	241	dst_garbage.timer_expires);
1da177e4	242	}
86bba269	243	spin_unlock_bh(&dst_garbage.lock);
1da177e4	244	}
d79d9913	245	EXPORT_SYMBOL(__dst_free);
1da177e4 LT	246
	247	struct dst_entry dst_destroy(struct dst_entry dst)
	248	{
	249	struct dst_entry *child;
1da177e4 LT	250
	251	smp_rmb();
	252
	253	again:
1da177e4 LT	254	child = dst->child;
1da177e4 LT	255
957c665f DM	256	if (!(dst->flags & DST_NOCOUNT))
957c665f DM	257	dst_entries_add(dst->ops, -1);
1da177e4 LT	258
	259	if (dst->ops->destroy)
	260	dst->ops->destroy(dst);
	261	if (dst->dev)
	262	dev_put(dst->dev);
f38a9eb1	263
e252b3d1 WC	264	lwtstate_put(dst->lwtstate);
e252b3d1 WC	265
f38a9eb1	266	if (dst->flags & DST_METADATA)
d71785ff	267	metadata_dst_free((struct metadata_dst *)dst);
f38a9eb1 TG	268	else
f38a9eb1 TG	269	kmem_cache_free(dst->ops->kmem_cachep, dst);
1da177e4 LT	270
	271	dst = child;
	272	if (dst) {
6775cab9 HX	273	int nohash = dst->flags & DST_NOHASH;
6775cab9 HX	274
1da177e4 LT	275	if (atomic_dec_and_test(&dst->__refcnt)) {
1da177e4 LT	276	/* We were real parent of this dst, so kill child. */
6775cab9	277	if (nohash)
1da177e4 LT	278	goto again;
	279	} else {
	280	/* Child is still referenced, return it for freeing. */
6775cab9	281	if (nohash)
1da177e4 LT	282	return dst;
	283	/* Child is still in his hash table */
	284	}
	285	}
	286	return NULL;
	287	}
598ed936	288	EXPORT_SYMBOL(dst_destroy);
1da177e4	289
f8864972 ED	290	static void dst_destroy_rcu(struct rcu_head *head)
	291	{
	292	struct dst_entry *dst = container_of(head, struct dst_entry, rcu_head);
	293
	294	dst = dst_destroy(dst);
	295	if (dst)
	296	__dst_free(dst);
	297	}
	298
8d330868 IJ	299	void dst_release(struct dst_entry *dst)
	300	{
	301	if (dst) {
598ed936	302	int newrefcnt;
07a5d384	303	unsigned short nocache = dst->flags & DST_NOCACHE;
ef711cf1	304
598ed936	305	newrefcnt = atomic_dec_return(&dst->__refcnt);
8bf4ada2 KK	306	if (unlikely(newrefcnt < 0))
	307	net_warn_ratelimited("%s: dst:%p refcnt:%d\n",
	308	__func__, dst, newrefcnt);
07a5d384	309	if (!newrefcnt && unlikely(nocache))
f8864972	310	call_rcu(&dst->rcu_head, dst_destroy_rcu);
8d330868 IJ	311	}
	312	}
	313	EXPORT_SYMBOL(dst_release);
	314
62fa8a84 DM	315	u32 dst_cow_metrics_generic(struct dst_entry dst, unsigned long old)
62fa8a84 DM	316	{
3fb07daf	317	struct dst_metrics p = kmalloc(sizeof(p), GFP_ATOMIC);
62fa8a84 DM	318
62fa8a84 DM	319	if (p) {
3fb07daf	320	struct dst_metrics old_p = (struct dst_metrics )__DST_METRICS_PTR(old);
62fa8a84 DM	321	unsigned long prev, new;
62fa8a84 DM	322
3fb07daf ED	323	atomic_set(&p->refcnt, 1);
3fb07daf ED	324	memcpy(p->metrics, old_p->metrics, sizeof(p->metrics));
62fa8a84 DM	325
	326	new = (unsigned long) p;
	327	prev = cmpxchg(&dst->_metrics, old, new);
	328
	329	if (prev != old) {
	330	kfree(p);
3fb07daf	331	p = (struct dst_metrics *)__DST_METRICS_PTR(prev);
62fa8a84 DM	332	if (prev & DST_METRICS_READ_ONLY)
62fa8a84 DM	333	p = NULL;
3fb07daf ED	334	} else if (prev & DST_METRICS_REFCOUNTED) {
	335	if (atomic_dec_and_test(&old_p->refcnt))
	336	kfree(old_p);
62fa8a84 DM	337	}
62fa8a84 DM	338	}
3fb07daf ED	339	BUILD_BUG_ON(offsetof(struct dst_metrics, metrics) != 0);
3fb07daf ED	340	return (u32 *)p;
62fa8a84 DM	341	}
	342	EXPORT_SYMBOL(dst_cow_metrics_generic);
	343
	344	/* Caller asserts that dst_metrics_read_only(dst) is false. */
	345	void __dst_destroy_metrics_generic(struct dst_entry *dst, unsigned long old)
	346	{
	347	unsigned long prev, new;
	348
3fb07daf	349	new = ((unsigned long) &dst_default_metrics) \| DST_METRICS_READ_ONLY;
62fa8a84 DM	350	prev = cmpxchg(&dst->_metrics, old, new);
	351	if (prev == old)
	352	kfree(__DST_METRICS_PTR(old));
	353	}
	354	EXPORT_SYMBOL(__dst_destroy_metrics_generic);
	355
f38a9eb1 TG	356	static struct dst_ops md_dst_ops = {
	357	.family = AF_UNSPEC,
	358	};
	359
ede2059d	360	static int dst_md_discard_out(struct net net, struct sock sk, struct sk_buff *skb)
f38a9eb1 TG	361	{
	362	WARN_ONCE(1, "Attempting to call output on metadata dst\n");
	363	kfree_skb(skb);
	364	return 0;
	365	}
	366
	367	static int dst_md_discard(struct sk_buff *skb)
	368	{
	369	WARN_ONCE(1, "Attempting to call input on metadata dst\n");
	370	kfree_skb(skb);
	371	return 0;
	372	}
	373
d3aa45ce	374	static void __metadata_dst_init(struct metadata_dst *md_dst, u8 optslen)
f38a9eb1	375	{
f38a9eb1 TG	376	struct dst_entry *dst;
f38a9eb1 TG	377
f38a9eb1 TG	378	dst = &md_dst->dst;
	379	dst_init(dst, &md_dst_ops, NULL, 1, DST_OBSOLETE_NONE,
	380	DST_METADATA \| DST_NOCACHE \| DST_NOCOUNT);
	381
	382	dst->input = dst_md_discard;
ede2059d	383	dst->output = dst_md_discard_out;
f38a9eb1 TG	384
f38a9eb1 TG	385	memset(dst + 1, 0, sizeof(md_dst) + optslen - sizeof(dst));
d3aa45ce AS	386	}
	387
	388	struct metadata_dst *metadata_dst_alloc(u8 optslen, gfp_t flags)
	389	{
	390	struct metadata_dst *md_dst;
	391
	392	md_dst = kmalloc(sizeof(*md_dst) + optslen, flags);
	393	if (!md_dst)
	394	return NULL;
	395
	396	__metadata_dst_init(md_dst, optslen);
f38a9eb1 TG	397
	398	return md_dst;
	399	}
	400	EXPORT_SYMBOL_GPL(metadata_dst_alloc);
	401
d71785ff PA	402	void metadata_dst_free(struct metadata_dst *md_dst)
	403	{
	404	#ifdef CONFIG_DST_CACHE
	405	dst_cache_destroy(&md_dst->u.tun_info.dst_cache);
	406	#endif
	407	kfree(md_dst);
	408	}
	409
d3aa45ce AS	410	struct metadata_dst __percpu *metadata_dst_alloc_percpu(u8 optslen, gfp_t flags)
	411	{
	412	int cpu;
	413	struct metadata_dst __percpu *md_dst;
	414
	415	md_dst = __alloc_percpu_gfp(sizeof(struct metadata_dst) + optslen,
	416	__alignof__(struct metadata_dst), flags);
	417	if (!md_dst)
	418	return NULL;
	419
	420	for_each_possible_cpu(cpu)
	421	__metadata_dst_init(per_cpu_ptr(md_dst, cpu), optslen);
	422
	423	return md_dst;
	424	}
	425	EXPORT_SYMBOL_GPL(metadata_dst_alloc_percpu);
	426
1da177e4 LT	427	/* Dirty hack. We did it in 2.2 (in __dst_free),
	428	* we have _very_ good reasons not to repeat
	429	* this mistake in 2.3, but we have no choice
	430	* now. _It_ _is_ _explicit_ _deliberate_
	431	* _race_ _condition_.
	432	*
	433	* Commented and originally written by Alexey.
	434	*/
56115511	435	static void dst_ifdown(struct dst_entry dst, struct net_device dev,
56115511	436	int unregister)
1da177e4 LT	437	{
	438	if (dst->ops->ifdown)
	439	dst->ops->ifdown(dst, dev, unregister);
	440
	441	if (dev != dst->dev)
	442	return;
	443
	444	if (!unregister) {
aad88724	445	dst->input = dst_discard;
ede2059d	446	dst->output = dst_discard_out;
1da177e4	447	} else {
c346dca1	448	dst->dev = dev_net(dst->dev)->loopback_dev;
de3cb747	449	dev_hold(dst->dev);
1da177e4	450	dev_put(dev);
1da177e4 LT	451	}
	452	}
	453
598ed936	454	static int dst_dev_event(struct notifier_block *this, unsigned long event,
598ed936	455	void *ptr)
1da177e4	456	{
351638e7	457	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
86bba269	458	struct dst_entry dst, last = NULL;
1da177e4 LT	459
1da177e4 LT	460	switch (event) {
0115e8e3	461	case NETDEV_UNREGISTER_FINAL:
1da177e4	462	case NETDEV_DOWN:
86bba269 ED	463	mutex_lock(&dst_gc_mutex);
	464	for (dst = dst_busy_list; dst; dst = dst->next) {
	465	last = dst;
	466	dst_ifdown(dst, dev, event != NETDEV_DOWN);
	467	}
	468
	469	spin_lock_bh(&dst_garbage.lock);
	470	dst = dst_garbage.list;
	471	dst_garbage.list = NULL;
f186ce61 KJ	472	/* The code in dst_ifdown places a hold on the loopback device.
	473	* If the gc entry processing is set to expire after a lengthy
	474	* interval, this hold can cause netdev_wait_allrefs() to hang
	475	* out and wait for a long time -- until the the loopback
	476	* interface is released. If we're really unlucky, it'll emit
	477	* pr_emerg messages to console too. Reset the interval here,
	478	* so dst cleanups occur in a more timely fashion.
	479	*/
	480	if (dst_garbage.timer_inc > DST_GC_INC) {
	481	dst_garbage.timer_inc = DST_GC_INC;
	482	dst_garbage.timer_expires = DST_GC_MIN;
	483	mod_delayed_work(system_wq, &dst_gc_work,
	484	dst_garbage.timer_expires);
	485	}
86bba269 ED	486	spin_unlock_bh(&dst_garbage.lock);
	487
	488	if (last)
	489	last->next = dst;
	490	else
	491	dst_busy_list = dst;
598ed936	492	for (; dst; dst = dst->next)
1da177e4	493	dst_ifdown(dst, dev, event != NETDEV_DOWN);
86bba269	494	mutex_unlock(&dst_gc_mutex);
1da177e4 LT	495	break;
	496	}
	497	return NOTIFY_DONE;
	498	}
	499
	500	static struct notifier_block dst_dev_notifier = {
	501	.notifier_call = dst_dev_event,
332dd96f	502	.priority = -10, /* must be called after other network notifiers */
1da177e4 LT	503	};
1da177e4 LT	504
f38a9eb1	505	void __init dst_subsys_init(void)
1da177e4 LT	506	{
	507	register_netdevice_notifier(&dst_dev_notifier);
	508	}