[deliverable/linux.git] / net / core / flow.c

/* flow.c: Generic flow cache.
 *
 * Copyright (C) 2003 Alexey N. Kuznetsov (kuznet@ms2.inr.ac.ru)
 * Copyright (C) 2003 David S. Miller (davem@redhat.com)
 */

#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/list.h>
#include <linux/jhash.h>
#include <linux/interrupt.h>
#include <linux/mm.h>
#include <linux/random.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/smp.h>
#include <linux/completion.h>
#include <linux/percpu.h>
#include <linux/bitops.h>
#include <linux/notifier.h>
#include <linux/cpu.h>
#include <linux/cpumask.h>
#include <linux/mutex.h>
#include <net/flow.h>
#include <linux/atomic.h>
#include <linux/security.h>
#include <net/net_namespace.h>

struct flow_cache_entry {
	union {
		struct hlist_node	hlist;
		struct list_head	gc_list;
	} u;
	struct net			*net;
	u16				family;
	u8				dir;
	u32				genid;
	struct flowi			key;
	struct flow_cache_object	*object;
};

struct flow_flush_info {
	struct flow_cache		*cache;
	atomic_t			cpuleft;
	struct completion		completion;
};

#define flow_cache_hash_size(cache)	(1 << (cache)->hash_shift)
#define FLOW_HASH_RND_PERIOD		(10 * 60 * HZ)

static void flow_cache_new_hashrnd(unsigned long arg)
{
	struct flow_cache *fc = (void *) arg;
	int i;

	for_each_possible_cpu(i)
		per_cpu_ptr(fc->percpu, i)->hash_rnd_recalc = 1;

	fc->rnd_timer.expires = jiffies + FLOW_HASH_RND_PERIOD;
	add_timer(&fc->rnd_timer);
}

static int flow_entry_valid(struct flow_cache_entry *fle,
				struct netns_xfrm *xfrm)
{
	if (atomic_read(&xfrm->flow_cache_genid) != fle->genid)
		return 0;
	if (fle->object && !fle->object->ops->check(fle->object))
		return 0;
	return 1;
}

static void flow_entry_kill(struct flow_cache_entry *fle,
				struct netns_xfrm *xfrm)
{
	if (fle->object)
		fle->object->ops->delete(fle->object);
	kmem_cache_free(xfrm->flow_cachep, fle);
}

static void flow_cache_gc_task(struct work_struct *work)
{
	struct list_head gc_list;
	struct flow_cache_entry *fce, *n;
	struct netns_xfrm *xfrm = container_of(work, struct netns_xfrm,
						flow_cache_gc_work);

	INIT_LIST_HEAD(&gc_list);
	spin_lock_bh(&xfrm->flow_cache_gc_lock);
	list_splice_tail_init(&xfrm->flow_cache_gc_list, &gc_list);
	spin_unlock_bh(&xfrm->flow_cache_gc_lock);

	list_for_each_entry_safe(fce, n, &gc_list, u.gc_list)
		flow_entry_kill(fce, xfrm);
}

static void flow_cache_queue_garbage(struct flow_cache_percpu *fcp,
				     int deleted, struct list_head *gc_list,
				     struct netns_xfrm *xfrm)
{
	if (deleted) {
		fcp->hash_count -= deleted;
		spin_lock_bh(&xfrm->flow_cache_gc_lock);
		list_splice_tail(gc_list, &xfrm->flow_cache_gc_list);
		spin_unlock_bh(&xfrm->flow_cache_gc_lock);
		schedule_work(&xfrm->flow_cache_gc_work);
	}
}

static void __flow_cache_shrink(struct flow_cache *fc,
				struct flow_cache_percpu *fcp,
				int shrink_to)
{
	struct flow_cache_entry *fle;
	struct hlist_node *tmp;
	LIST_HEAD(gc_list);
	int i, deleted = 0;
	struct netns_xfrm *xfrm = container_of(fc, struct netns_xfrm,
						flow_cache_global);

	for (i = 0; i < flow_cache_hash_size(fc); i++) {
		int saved = 0;

		hlist_for_each_entry_safe(fle, tmp,
					  &fcp->hash_table[i], u.hlist) {
			if (saved < shrink_to &&
			    flow_entry_valid(fle, xfrm)) {
				saved++;
			} else {
				deleted++;
				hlist_del(&fle->u.hlist);
				list_add_tail(&fle->u.gc_list, &gc_list);
			}
		}
	}

	flow_cache_queue_garbage(fcp, deleted, &gc_list, xfrm);
}

static void flow_cache_shrink(struct flow_cache *fc,
			      struct flow_cache_percpu *fcp)
{
	int shrink_to = fc->low_watermark / flow_cache_hash_size(fc);

	__flow_cache_shrink(fc, fcp, shrink_to);
}

static void flow_new_hash_rnd(struct flow_cache *fc,
			      struct flow_cache_percpu *fcp)
{
	get_random_bytes(&fcp->hash_rnd, sizeof(u32));
	fcp->hash_rnd_recalc = 0;
	__flow_cache_shrink(fc, fcp, 0);
}

static u32 flow_hash_code(struct flow_cache *fc,
			  struct flow_cache_percpu *fcp,
			  const struct flowi *key,
			  size_t keysize)
{
	const u32 *k = (const u32 *) key;
	const u32 length = keysize * sizeof(flow_compare_t) / sizeof(u32);

	return jhash2(k, length, fcp->hash_rnd)
		& (flow_cache_hash_size(fc) - 1);
}

/* I hear what you're saying, use memcmp.  But memcmp cannot make
 * important assumptions that we can here, such as alignment.
 */
static int flow_key_compare(const struct flowi *key1, const struct flowi *key2,
			    size_t keysize)
{
	const flow_compare_t *k1, *k1_lim, *k2;

	k1 = (const flow_compare_t *) key1;
	k1_lim = k1 + keysize;

	k2 = (const flow_compare_t *) key2;

	do {
		if (*k1++ != *k2++)
			return 1;
	} while (k1 < k1_lim);

	return 0;
}

struct flow_cache_object *
flow_cache_lookup(struct net *net, const struct flowi *key, u16 family, u8 dir,
		  flow_resolve_t resolver, void *ctx)
{
	struct flow_cache *fc = &net->xfrm.flow_cache_global;
	struct flow_cache_percpu *fcp;
	struct flow_cache_entry *fle, *tfle;
	struct flow_cache_object *flo;
	size_t keysize;
	unsigned int hash;

	local_bh_disable();
	fcp = this_cpu_ptr(fc->percpu);

	fle = NULL;
	flo = NULL;

	keysize = flow_key_size(family);
	if (!keysize)
		goto nocache;

	/* Packet really early in init?  Making flow_cache_init a
	 * pre-smp initcall would solve this.  --RR */
	if (!fcp->hash_table)
		goto nocache;

	if (fcp->hash_rnd_recalc)
		flow_new_hash_rnd(fc, fcp);

	hash = flow_hash_code(fc, fcp, key, keysize);
	hlist_for_each_entry(tfle, &fcp->hash_table[hash], u.hlist) {
		if (tfle->net == net &&
		    tfle->family == family &&
		    tfle->dir == dir &&
		    flow_key_compare(key, &tfle->key, keysize) == 0) {
			fle = tfle;
			break;
		}
	}

	if (unlikely(!fle)) {
		if (fcp->hash_count > fc->high_watermark)
			flow_cache_shrink(fc, fcp);

		fle = kmem_cache_alloc(net->xfrm.flow_cachep, GFP_ATOMIC);
		if (fle) {
			fle->net = net;
			fle->family = family;
			fle->dir = dir;
			memcpy(&fle->key, key, keysize * sizeof(flow_compare_t));
			fle->object = NULL;
			hlist_add_head(&fle->u.hlist, &fcp->hash_table[hash]);
			fcp->hash_count++;
		}
	} else if (likely(fle->genid == atomic_read(&net->xfrm.flow_cache_genid))) {
		flo = fle->object;
		if (!flo)
			goto ret_object;
		flo = flo->ops->get(flo);
		if (flo)
			goto ret_object;
	} else if (fle->object) {
	        flo = fle->object;
	        flo->ops->delete(flo);
	        fle->object = NULL;
	}

nocache:
	flo = NULL;
	if (fle) {
		flo = fle->object;
		fle->object = NULL;
	}
	flo = resolver(net, key, family, dir, flo, ctx);
	if (fle) {
		fle->genid = atomic_read(&net->xfrm.flow_cache_genid);
		if (!IS_ERR(flo))
			fle->object = flo;
		else
			fle->genid--;
	} else {
		if (!IS_ERR_OR_NULL(flo))
			flo->ops->delete(flo);
	}
ret_object:
	local_bh_enable();
	return flo;
}
EXPORT_SYMBOL(flow_cache_lookup);

static void flow_cache_flush_tasklet(unsigned long data)
{
	struct flow_flush_info *info = (void *)data;
	struct flow_cache *fc = info->cache;
	struct flow_cache_percpu *fcp;
	struct flow_cache_entry *fle;
	struct hlist_node *tmp;
	LIST_HEAD(gc_list);
	int i, deleted = 0;
	struct netns_xfrm *xfrm = container_of(fc, struct netns_xfrm,
						flow_cache_global);

	fcp = this_cpu_ptr(fc->percpu);
	for (i = 0; i < flow_cache_hash_size(fc); i++) {
		hlist_for_each_entry_safe(fle, tmp,
					  &fcp->hash_table[i], u.hlist) {
			if (flow_entry_valid(fle, xfrm))
				continue;

			deleted++;
			hlist_del(&fle->u.hlist);
			list_add_tail(&fle->u.gc_list, &gc_list);
		}
	}

	flow_cache_queue_garbage(fcp, deleted, &gc_list, xfrm);

	if (atomic_dec_and_test(&info->cpuleft))
		complete(&info->completion);
}

/*
 * Return whether a cpu needs flushing.  Conservatively, we assume
 * the presence of any entries means the core may require flushing,
 * since the flow_cache_ops.check() function may assume it's running
 * on the same core as the per-cpu cache component.
 */
static int flow_cache_percpu_empty(struct flow_cache *fc, int cpu)
{
	struct flow_cache_percpu *fcp;
	int i;

	fcp = per_cpu_ptr(fc->percpu, cpu);
	for (i = 0; i < flow_cache_hash_size(fc); i++)
		if (!hlist_empty(&fcp->hash_table[i]))
			return 0;
	return 1;
}

static void flow_cache_flush_per_cpu(void *data)
{
	struct flow_flush_info *info = data;
	struct tasklet_struct *tasklet;

	tasklet = &this_cpu_ptr(info->cache->percpu)->flush_tasklet;
	tasklet->data = (unsigned long)info;
	tasklet_schedule(tasklet);
}

void flow_cache_flush(struct net *net)
{
	struct flow_flush_info info;
	cpumask_var_t mask;
	int i, self;

	/* Track which cpus need flushing to avoid disturbing all cores. */
	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
		return;
	cpumask_clear(mask);

	/* Don't want cpus going down or up during this. */
	get_online_cpus();
	mutex_lock(&net->xfrm.flow_flush_sem);
	info.cache = &net->xfrm.flow_cache_global;
	for_each_online_cpu(i)
		if (!flow_cache_percpu_empty(info.cache, i))
			cpumask_set_cpu(i, mask);
	atomic_set(&info.cpuleft, cpumask_weight(mask));
	if (atomic_read(&info.cpuleft) == 0)
		goto done;

	init_completion(&info.completion);

	local_bh_disable();
	self = cpumask_test_and_clear_cpu(smp_processor_id(), mask);
	on_each_cpu_mask(mask, flow_cache_flush_per_cpu, &info, 0);
	if (self)
		flow_cache_flush_tasklet((unsigned long)&info);
	local_bh_enable();

	wait_for_completion(&info.completion);

done:
	mutex_unlock(&net->xfrm.flow_flush_sem);
	put_online_cpus();
	free_cpumask_var(mask);
}

static void flow_cache_flush_task(struct work_struct *work)
{
	struct netns_xfrm *xfrm = container_of(work, struct netns_xfrm,
						flow_cache_gc_work);
	struct net *net = container_of(xfrm, struct net, xfrm);

	flow_cache_flush(net);
}

void flow_cache_flush_deferred(struct net *net)
{
	schedule_work(&net->xfrm.flow_cache_flush_work);
}

static int flow_cache_cpu_prepare(struct flow_cache *fc, int cpu)
{
	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, cpu);
	size_t sz = sizeof(struct hlist_head) * flow_cache_hash_size(fc);

	if (!fcp->hash_table) {
		fcp->hash_table = kzalloc_node(sz, GFP_KERNEL, cpu_to_node(cpu));
		if (!fcp->hash_table) {
			pr_err("NET: failed to allocate flow cache sz %zu\n", sz);
			return -ENOMEM;
		}
		fcp->hash_rnd_recalc = 1;
		fcp->hash_count = 0;
		tasklet_init(&fcp->flush_tasklet, flow_cache_flush_tasklet, 0);
	}
	return 0;
}

static int flow_cache_cpu(struct notifier_block *nfb,
			  unsigned long action,
			  void *hcpu)
{
	struct flow_cache *fc = container_of(nfb, struct flow_cache,
						hotcpu_notifier);
	int res, cpu = (unsigned long) hcpu;
	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, cpu);

	switch (action) {
	case CPU_UP_PREPARE:
	case CPU_UP_PREPARE_FROZEN:
		res = flow_cache_cpu_prepare(fc, cpu);
		if (res)
			return notifier_from_errno(res);
		break;
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		__flow_cache_shrink(fc, fcp, 0);
		break;
	}
	return NOTIFY_OK;
}

int flow_cache_init(struct net *net)
{
	int i;
	struct flow_cache *fc = &net->xfrm.flow_cache_global;

	/* Initialize per-net flow cache global variables here */
	net->xfrm.flow_cachep = kmem_cache_create("flow_cache",
					sizeof(struct flow_cache_entry),
					0, SLAB_PANIC, NULL);
	spin_lock_init(&net->xfrm.flow_cache_gc_lock);
	INIT_LIST_HEAD(&net->xfrm.flow_cache_gc_list);
	INIT_WORK(&net->xfrm.flow_cache_gc_work, flow_cache_gc_task);
	INIT_WORK(&net->xfrm.flow_cache_flush_work, flow_cache_flush_task);
	mutex_init(&net->xfrm.flow_flush_sem);

	fc->hash_shift = 10;
	fc->low_watermark = 2 * flow_cache_hash_size(fc);
	fc->high_watermark = 4 * flow_cache_hash_size(fc);

	fc->percpu = alloc_percpu(struct flow_cache_percpu);
	if (!fc->percpu)
		return -ENOMEM;

	for_each_online_cpu(i) {
		if (flow_cache_cpu_prepare(fc, i))
			goto err;
	}
	fc->hotcpu_notifier = (struct notifier_block){
		.notifier_call = flow_cache_cpu,
	};
	register_hotcpu_notifier(&fc->hotcpu_notifier);

	setup_timer(&fc->rnd_timer, flow_cache_new_hashrnd,
		    (unsigned long) fc);
	fc->rnd_timer.expires = jiffies + FLOW_HASH_RND_PERIOD;
	add_timer(&fc->rnd_timer);

	return 0;

err:
	for_each_possible_cpu(i) {
		struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, i);
		kfree(fcp->hash_table);
		fcp->hash_table = NULL;
	}

	free_percpu(fc->percpu);
	fc->percpu = NULL;

	return -ENOMEM;
}
EXPORT_SYMBOL(flow_cache_init);
Commit	Line	Data
1da177e4 LT	1	/* flow.c: Generic flow cache.
	2	*
	3	* Copyright (C) 2003 Alexey N. Kuznetsov (kuznet@ms2.inr.ac.ru)
	4	* Copyright (C) 2003 David S. Miller (davem@redhat.com)
	5	*/
	6
	7	#include <linux/kernel.h>
	8	#include <linux/module.h>
	9	#include <linux/list.h>
	10	#include <linux/jhash.h>
	11	#include <linux/interrupt.h>
	12	#include <linux/mm.h>
	13	#include <linux/random.h>
	14	#include <linux/init.h>
	15	#include <linux/slab.h>
	16	#include <linux/smp.h>
	17	#include <linux/completion.h>
	18	#include <linux/percpu.h>
	19	#include <linux/bitops.h>
	20	#include <linux/notifier.h>
	21	#include <linux/cpu.h>
	22	#include <linux/cpumask.h>
4a3e2f71	23	#include <linux/mutex.h>
1da177e4	24	#include <net/flow.h>
60063497	25	#include <linux/atomic.h>
df71837d	26	#include <linux/security.h>
ca925cf1	27	#include <net/net_namespace.h>
1da177e4 LT	28
1da177e4 LT	29	struct flow_cache_entry {
8e479560 TT	30	union {
	31	struct hlist_node hlist;
	32	struct list_head gc_list;
	33	} u;
0542b69e	34	struct net *net;
fe1a5f03 TT	35	u16 family;
	36	u8 dir;
	37	u32 genid;
	38	struct flowi key;
	39	struct flow_cache_object *object;
1da177e4 LT	40	};
1da177e4 LT	41
1da177e4	42	struct flow_flush_info {
fe1a5f03	43	struct flow_cache *cache;
d7997fe1 TT	44	atomic_t cpuleft;
d7997fe1 TT	45	struct completion completion;
1da177e4	46	};
1da177e4	47
d7997fe1 TT	48	#define flow_cache_hash_size(cache) (1 << (cache)->hash_shift)
d7997fe1 TT	49	#define FLOW_HASH_RND_PERIOD (10 * 60 * HZ)
1da177e4 LT	50
	51	static void flow_cache_new_hashrnd(unsigned long arg)
	52	{
d7997fe1	53	struct flow_cache fc = (void ) arg;
1da177e4 LT	54	int i;
1da177e4 LT	55
6f912042	56	for_each_possible_cpu(i)
d7997fe1	57	per_cpu_ptr(fc->percpu, i)->hash_rnd_recalc = 1;
1da177e4	58
d7997fe1 TT	59	fc->rnd_timer.expires = jiffies + FLOW_HASH_RND_PERIOD;
d7997fe1 TT	60	add_timer(&fc->rnd_timer);
1da177e4 LT	61	}
1da177e4 LT	62
ca925cf1 FD	63	static int flow_entry_valid(struct flow_cache_entry *fle,
ca925cf1 FD	64	struct netns_xfrm *xfrm)
fe1a5f03	65	{
ca925cf1	66	if (atomic_read(&xfrm->flow_cache_genid) != fle->genid)
fe1a5f03 TT	67	return 0;
	68	if (fle->object && !fle->object->ops->check(fle->object))
	69	return 0;
	70	return 1;
	71	}
	72
ca925cf1 FD	73	static void flow_entry_kill(struct flow_cache_entry *fle,
ca925cf1 FD	74	struct netns_xfrm *xfrm)
134b0fc5 JM	75	{
134b0fc5 JM	76	if (fle->object)
fe1a5f03	77	fle->object->ops->delete(fle->object);
ca925cf1	78	kmem_cache_free(xfrm->flow_cachep, fle);
8e479560 TT	79	}
	80
	81	static void flow_cache_gc_task(struct work_struct *work)
	82	{
	83	struct list_head gc_list;
	84	struct flow_cache_entry fce, n;
ca925cf1 FD	85	struct netns_xfrm *xfrm = container_of(work, struct netns_xfrm,
ca925cf1 FD	86	flow_cache_gc_work);
8e479560 TT	87
8e479560 TT	88	INIT_LIST_HEAD(&gc_list);
ca925cf1 FD	89	spin_lock_bh(&xfrm->flow_cache_gc_lock);
	90	list_splice_tail_init(&xfrm->flow_cache_gc_list, &gc_list);
	91	spin_unlock_bh(&xfrm->flow_cache_gc_lock);
8e479560 TT	92
8e479560 TT	93	list_for_each_entry_safe(fce, n, &gc_list, u.gc_list)
ca925cf1	94	flow_entry_kill(fce, xfrm);
8e479560	95	}
8e479560 TT	96
8e479560 TT	97	static void flow_cache_queue_garbage(struct flow_cache_percpu *fcp,
ca925cf1 FD	98	int deleted, struct list_head *gc_list,
ca925cf1 FD	99	struct netns_xfrm *xfrm)
8e479560 TT	100	{
	101	if (deleted) {
	102	fcp->hash_count -= deleted;
ca925cf1 FD	103	spin_lock_bh(&xfrm->flow_cache_gc_lock);
	104	list_splice_tail(gc_list, &xfrm->flow_cache_gc_list);
	105	spin_unlock_bh(&xfrm->flow_cache_gc_lock);
	106	schedule_work(&xfrm->flow_cache_gc_work);
8e479560	107	}
134b0fc5 JM	108	}
134b0fc5 JM	109
d7997fe1 TT	110	static void __flow_cache_shrink(struct flow_cache *fc,
	111	struct flow_cache_percpu *fcp,
	112	int shrink_to)
1da177e4	113	{
8e479560	114	struct flow_cache_entry *fle;
b67bfe0d	115	struct hlist_node *tmp;
8e479560 TT	116	LIST_HEAD(gc_list);
8e479560 TT	117	int i, deleted = 0;
ca925cf1 FD	118	struct netns_xfrm *xfrm = container_of(fc, struct netns_xfrm,
ca925cf1 FD	119	flow_cache_global);
1da177e4	120
d7997fe1	121	for (i = 0; i < flow_cache_hash_size(fc); i++) {
fe1a5f03	122	int saved = 0;
1da177e4	123
b67bfe0d	124	hlist_for_each_entry_safe(fle, tmp,
8e479560	125	&fcp->hash_table[i], u.hlist) {
fe1a5f03	126	if (saved < shrink_to &&
ca925cf1	127	flow_entry_valid(fle, xfrm)) {
fe1a5f03	128	saved++;
fe1a5f03	129	} else {
8e479560 TT	130	deleted++;
	131	hlist_del(&fle->u.hlist);
	132	list_add_tail(&fle->u.gc_list, &gc_list);
fe1a5f03	133	}
1da177e4 LT	134	}
1da177e4 LT	135	}
8e479560	136
ca925cf1	137	flow_cache_queue_garbage(fcp, deleted, &gc_list, xfrm);
1da177e4 LT	138	}
1da177e4 LT	139
d7997fe1 TT	140	static void flow_cache_shrink(struct flow_cache *fc,
d7997fe1 TT	141	struct flow_cache_percpu *fcp)
1da177e4	142	{
d7997fe1	143	int shrink_to = fc->low_watermark / flow_cache_hash_size(fc);
1da177e4	144
d7997fe1	145	__flow_cache_shrink(fc, fcp, shrink_to);
1da177e4 LT	146	}
1da177e4 LT	147
d7997fe1 TT	148	static void flow_new_hash_rnd(struct flow_cache *fc,
d7997fe1 TT	149	struct flow_cache_percpu *fcp)
1da177e4	150	{
d7997fe1 TT	151	get_random_bytes(&fcp->hash_rnd, sizeof(u32));
	152	fcp->hash_rnd_recalc = 0;
	153	__flow_cache_shrink(fc, fcp, 0);
1da177e4 LT	154	}
1da177e4 LT	155
d7997fe1 TT	156	static u32 flow_hash_code(struct flow_cache *fc,
d7997fe1 TT	157	struct flow_cache_percpu *fcp,
aa1c366e	158	const struct flowi *key,
aa1c366e	159	size_t keysize)
1da177e4	160	{
dee9f4bc	161	const u32 k = (const u32 ) key;
aa1c366e	162	const u32 length = keysize * sizeof(flow_compare_t) / sizeof(u32);
1da177e4	163
aa1c366e	164	return jhash2(k, length, fcp->hash_rnd)
a02cec21	165	& (flow_cache_hash_size(fc) - 1);
1da177e4 LT	166	}
1da177e4 LT	167
1da177e4	168	/* I hear what you're saying, use memcmp. But memcmp cannot make
aa1c366e	169	* important assumptions that we can here, such as alignment.
1da177e4	170	*/
aa1c366e	171	static int flow_key_compare(const struct flowi key1, const struct flowi key2,
aa1c366e	172	size_t keysize)
1da177e4	173	{
dee9f4bc	174	const flow_compare_t k1, k1_lim, *k2;
1da177e4	175
dee9f4bc	176	k1 = (const flow_compare_t *) key1;
aa1c366e	177	k1_lim = k1 + keysize;
1da177e4	178
dee9f4bc	179	k2 = (const flow_compare_t *) key2;
1da177e4 LT	180
	181	do {
	182	if (k1++ != k2++)
	183	return 1;
	184	} while (k1 < k1_lim);
	185
	186	return 0;
	187	}
	188
fe1a5f03	189	struct flow_cache_object *
dee9f4bc	190	flow_cache_lookup(struct net net, const struct flowi key, u16 family, u8 dir,
fe1a5f03	191	flow_resolve_t resolver, void *ctx)
1da177e4	192	{
ca925cf1	193	struct flow_cache *fc = &net->xfrm.flow_cache_global;
d7997fe1	194	struct flow_cache_percpu *fcp;
8e479560	195	struct flow_cache_entry fle, tfle;
fe1a5f03	196	struct flow_cache_object *flo;
aa1c366e	197	size_t keysize;
1da177e4	198	unsigned int hash;
1da177e4 LT	199
1da177e4 LT	200	local_bh_disable();
7a9b2d59	201	fcp = this_cpu_ptr(fc->percpu);
1da177e4 LT	202
1da177e4 LT	203	fle = NULL;
fe1a5f03	204	flo = NULL;
aa1c366e	205
	206	keysize = flow_key_size(family);
	207	if (!keysize)
	208	goto nocache;
	209
1da177e4 LT	210	/* Packet really early in init? Making flow_cache_init a
1da177e4 LT	211	* pre-smp initcall would solve this. --RR */
d7997fe1	212	if (!fcp->hash_table)
1da177e4 LT	213	goto nocache;
1da177e4 LT	214
d7997fe1 TT	215	if (fcp->hash_rnd_recalc)
d7997fe1 TT	216	flow_new_hash_rnd(fc, fcp);
1da177e4	217
aa1c366e	218	hash = flow_hash_code(fc, fcp, key, keysize);
b67bfe0d	219	hlist_for_each_entry(tfle, &fcp->hash_table[hash], u.hlist) {
0542b69e	220	if (tfle->net == net &&
0542b69e	221	tfle->family == family &&
8e479560	222	tfle->dir == dir &&
aa1c366e	223	flow_key_compare(key, &tfle->key, keysize) == 0) {
8e479560	224	fle = tfle;
1da177e4	225	break;
8e479560	226	}
1da177e4 LT	227	}
1da177e4 LT	228
fe1a5f03	229	if (unlikely(!fle)) {
d7997fe1 TT	230	if (fcp->hash_count > fc->high_watermark)
d7997fe1 TT	231	flow_cache_shrink(fc, fcp);
1da177e4	232
ca925cf1	233	fle = kmem_cache_alloc(net->xfrm.flow_cachep, GFP_ATOMIC);
1da177e4	234	if (fle) {
0542b69e	235	fle->net = net;
1da177e4 LT	236	fle->family = family;
1da177e4 LT	237	fle->dir = dir;
aa1c366e	238	memcpy(&fle->key, key, keysize * sizeof(flow_compare_t));
1da177e4	239	fle->object = NULL;
8e479560	240	hlist_add_head(&fle->u.hlist, &fcp->hash_table[hash]);
d7997fe1	241	fcp->hash_count++;
1da177e4	242	}
ca925cf1	243	} else if (likely(fle->genid == atomic_read(&net->xfrm.flow_cache_genid))) {
fe1a5f03 TT	244	flo = fle->object;
	245	if (!flo)
	246	goto ret_object;
	247	flo = flo->ops->get(flo);
	248	if (flo)
	249	goto ret_object;
	250	} else if (fle->object) {
	251	flo = fle->object;
	252	flo->ops->delete(flo);
	253	fle->object = NULL;
1da177e4 LT	254	}
	255
	256	nocache:
fe1a5f03 TT	257	flo = NULL;
	258	if (fle) {
	259	flo = fle->object;
	260	fle->object = NULL;
	261	}
	262	flo = resolver(net, key, family, dir, flo, ctx);
	263	if (fle) {
ca925cf1	264	fle->genid = atomic_read(&net->xfrm.flow_cache_genid);
fe1a5f03 TT	265	if (!IS_ERR(flo))
	266	fle->object = flo;
	267	else
	268	fle->genid--;
	269	} else {
8fbcec24	270	if (!IS_ERR_OR_NULL(flo))
fe1a5f03	271	flo->ops->delete(flo);
1da177e4	272	}
fe1a5f03 TT	273	ret_object:
	274	local_bh_enable();
	275	return flo;
1da177e4	276	}
9e34a5b5	277	EXPORT_SYMBOL(flow_cache_lookup);
1da177e4 LT	278
	279	static void flow_cache_flush_tasklet(unsigned long data)
	280	{
	281	struct flow_flush_info info = (void )data;
d7997fe1 TT	282	struct flow_cache *fc = info->cache;
d7997fe1 TT	283	struct flow_cache_percpu *fcp;
8e479560	284	struct flow_cache_entry *fle;
b67bfe0d	285	struct hlist_node *tmp;
8e479560 TT	286	LIST_HEAD(gc_list);
8e479560 TT	287	int i, deleted = 0;
ca925cf1 FD	288	struct netns_xfrm *xfrm = container_of(fc, struct netns_xfrm,
ca925cf1 FD	289	flow_cache_global);
1da177e4	290
7a9b2d59	291	fcp = this_cpu_ptr(fc->percpu);
d7997fe1	292	for (i = 0; i < flow_cache_hash_size(fc); i++) {
b67bfe0d	293	hlist_for_each_entry_safe(fle, tmp,
8e479560	294	&fcp->hash_table[i], u.hlist) {
ca925cf1	295	if (flow_entry_valid(fle, xfrm))
1da177e4 LT	296	continue;
1da177e4 LT	297
8e479560 TT	298	deleted++;
	299	hlist_del(&fle->u.hlist);
	300	list_add_tail(&fle->u.gc_list, &gc_list);
1da177e4 LT	301	}
	302	}
	303
ca925cf1	304	flow_cache_queue_garbage(fcp, deleted, &gc_list, xfrm);
8e479560	305
1da177e4 LT	306	if (atomic_dec_and_test(&info->cpuleft))
	307	complete(&info->completion);
	308	}
	309
8fdc929f CM	310	/*
	311	* Return whether a cpu needs flushing. Conservatively, we assume
	312	* the presence of any entries means the core may require flushing,
	313	* since the flow_cache_ops.check() function may assume it's running
	314	* on the same core as the per-cpu cache component.
	315	*/
	316	static int flow_cache_percpu_empty(struct flow_cache *fc, int cpu)
	317	{
	318	struct flow_cache_percpu *fcp;
	319	int i;
	320
27815032	321	fcp = per_cpu_ptr(fc->percpu, cpu);
8fdc929f CM	322	for (i = 0; i < flow_cache_hash_size(fc); i++)
	323	if (!hlist_empty(&fcp->hash_table[i]))
	324	return 0;
	325	return 1;
	326	}
	327
1da177e4 LT	328	static void flow_cache_flush_per_cpu(void *data)
	329	{
	330	struct flow_flush_info *info = data;
1da177e4 LT	331	struct tasklet_struct *tasklet;
1da177e4 LT	332
50eab050	333	tasklet = &this_cpu_ptr(info->cache->percpu)->flush_tasklet;
1da177e4 LT	334	tasklet->data = (unsigned long)info;
	335	tasklet_schedule(tasklet);
	336	}
	337
ca925cf1	338	void flow_cache_flush(struct net *net)
1da177e4 LT	339	{
1da177e4 LT	340	struct flow_flush_info info;
8fdc929f CM	341	cpumask_var_t mask;
	342	int i, self;
	343
	344	/* Track which cpus need flushing to avoid disturbing all cores. */
	345	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
	346	return;
	347	cpumask_clear(mask);
1da177e4 LT	348
1da177e4 LT	349	/* Don't want cpus going down or up during this. */
86ef5c9a	350	get_online_cpus();
ca925cf1 FD	351	mutex_lock(&net->xfrm.flow_flush_sem);
ca925cf1 FD	352	info.cache = &net->xfrm.flow_cache_global;
8fdc929f CM	353	for_each_online_cpu(i)
	354	if (!flow_cache_percpu_empty(info.cache, i))
	355	cpumask_set_cpu(i, mask);
	356	atomic_set(&info.cpuleft, cpumask_weight(mask));
	357	if (atomic_read(&info.cpuleft) == 0)
	358	goto done;
	359
1da177e4 LT	360	init_completion(&info.completion);
	361
	362	local_bh_disable();
8fdc929f CM	363	self = cpumask_test_and_clear_cpu(smp_processor_id(), mask);
	364	on_each_cpu_mask(mask, flow_cache_flush_per_cpu, &info, 0);
	365	if (self)
	366	flow_cache_flush_tasklet((unsigned long)&info);
1da177e4 LT	367	local_bh_enable();
	368
	369	wait_for_completion(&info.completion);
8fdc929f CM	370
8fdc929f CM	371	done:
ca925cf1	372	mutex_unlock(&net->xfrm.flow_flush_sem);
86ef5c9a	373	put_online_cpus();
8fdc929f	374	free_cpumask_var(mask);
1da177e4 LT	375	}
1da177e4 LT	376
c0ed1c14 SK	377	static void flow_cache_flush_task(struct work_struct *work)
c0ed1c14 SK	378	{
ca925cf1 FD	379	struct netns_xfrm *xfrm = container_of(work, struct netns_xfrm,
	380	flow_cache_gc_work);
	381	struct net *net = container_of(xfrm, struct net, xfrm);
c0ed1c14	382
ca925cf1 FD	383	flow_cache_flush(net);
ca925cf1 FD	384	}
c0ed1c14	385
ca925cf1	386	void flow_cache_flush_deferred(struct net *net)
c0ed1c14	387	{
ca925cf1	388	schedule_work(&net->xfrm.flow_cache_flush_work);
c0ed1c14 SK	389	}
c0ed1c14 SK	390
013dbb32	391	static int flow_cache_cpu_prepare(struct flow_cache *fc, int cpu)
1da177e4	392	{
83b6b1f5 ED	393	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, cpu);
83b6b1f5 ED	394	size_t sz = sizeof(struct hlist_head) * flow_cache_hash_size(fc);
d7997fe1	395
83b6b1f5 ED	396	if (!fcp->hash_table) {
	397	fcp->hash_table = kzalloc_node(sz, GFP_KERNEL, cpu_to_node(cpu));
	398	if (!fcp->hash_table) {
	399	pr_err("NET: failed to allocate flow cache sz %zu\n", sz);
	400	return -ENOMEM;
	401	}
	402	fcp->hash_rnd_recalc = 1;
	403	fcp->hash_count = 0;
	404	tasklet_init(&fcp->flush_tasklet, flow_cache_flush_tasklet, 0);
	405	}
	406	return 0;
1da177e4 LT	407	}
1da177e4 LT	408
013dbb32	409	static int flow_cache_cpu(struct notifier_block *nfb,
1da177e4 LT	410	unsigned long action,
	411	void *hcpu)
	412	{
ca925cf1 FD	413	struct flow_cache *fc = container_of(nfb, struct flow_cache,
ca925cf1 FD	414	hotcpu_notifier);
83b6b1f5	415	int res, cpu = (unsigned long) hcpu;
d7997fe1 TT	416	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, cpu);
d7997fe1 TT	417
83b6b1f5 ED	418	switch (action) {
	419	case CPU_UP_PREPARE:
	420	case CPU_UP_PREPARE_FROZEN:
	421	res = flow_cache_cpu_prepare(fc, cpu);
	422	if (res)
	423	return notifier_from_errno(res);
	424	break;
	425	case CPU_DEAD:
	426	case CPU_DEAD_FROZEN:
d7997fe1	427	__flow_cache_shrink(fc, fcp, 0);
83b6b1f5 ED	428	break;
83b6b1f5 ED	429	}
1da177e4 LT	430	return NOTIFY_OK;
1da177e4 LT	431	}
1da177e4	432
ca925cf1	433	int flow_cache_init(struct net *net)
1da177e4 LT	434	{
1da177e4 LT	435	int i;
ca925cf1 FD	436	struct flow_cache *fc = &net->xfrm.flow_cache_global;
	437
	438	/* Initialize per-net flow cache global variables here */
	439	net->xfrm.flow_cachep = kmem_cache_create("flow_cache",
	440	sizeof(struct flow_cache_entry),
	441	0, SLAB_PANIC, NULL);
	442	spin_lock_init(&net->xfrm.flow_cache_gc_lock);
	443	INIT_LIST_HEAD(&net->xfrm.flow_cache_gc_list);
	444	INIT_WORK(&net->xfrm.flow_cache_gc_work, flow_cache_gc_task);
	445	INIT_WORK(&net->xfrm.flow_cache_flush_work, flow_cache_flush_task);
	446	mutex_init(&net->xfrm.flow_flush_sem);
1da177e4	447
d7997fe1 TT	448	fc->hash_shift = 10;
	449	fc->low_watermark = 2 * flow_cache_hash_size(fc);
	450	fc->high_watermark = 4 * flow_cache_hash_size(fc);
	451
d7997fe1	452	fc->percpu = alloc_percpu(struct flow_cache_percpu);
83b6b1f5 ED	453	if (!fc->percpu)
83b6b1f5 ED	454	return -ENOMEM;
1da177e4	455
83b6b1f5 ED	456	for_each_online_cpu(i) {
83b6b1f5 ED	457	if (flow_cache_cpu_prepare(fc, i))
6ccc3abd	458	goto err;
83b6b1f5	459	}
d7997fe1 TT	460	fc->hotcpu_notifier = (struct notifier_block){
	461	.notifier_call = flow_cache_cpu,
	462	};
	463	register_hotcpu_notifier(&fc->hotcpu_notifier);
1da177e4	464
83b6b1f5 ED	465	setup_timer(&fc->rnd_timer, flow_cache_new_hashrnd,
	466	(unsigned long) fc);
	467	fc->rnd_timer.expires = jiffies + FLOW_HASH_RND_PERIOD;
	468	add_timer(&fc->rnd_timer);
	469
1da177e4	470	return 0;
6ccc3abd	471
	472	err:
	473	for_each_possible_cpu(i) {
	474	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, i);
	475	kfree(fcp->hash_table);
	476	fcp->hash_table = NULL;
	477	}
	478
	479	free_percpu(fc->percpu);
	480	fc->percpu = NULL;
	481
	482	return -ENOMEM;
1da177e4	483	}
ca925cf1	484	EXPORT_SYMBOL(flow_cache_init);