[GitHub/mt8127/android_kernel_alcatel_ttab.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>

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_cache_percpu {
	struct hlist_head		*hash_table;
	int				hash_count;
	u32				hash_rnd;
	int				hash_rnd_recalc;
	struct tasklet_struct		flush_tasklet;
};

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

struct flow_cache {
	u32				hash_shift;
	struct flow_cache_percpu __percpu *percpu;
	struct notifier_block		hotcpu_notifier;
	int				low_watermark;
	int				high_watermark;
	struct timer_list		rnd_timer;
};

atomic_t flow_cache_genid = ATOMIC_INIT(0);
EXPORT_SYMBOL(flow_cache_genid);
static struct flow_cache flow_cache_global;
static struct kmem_cache *flow_cachep __read_mostly;

static DEFINE_SPINLOCK(flow_cache_gc_lock);
static LIST_HEAD(flow_cache_gc_list);

#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)
{
	if (atomic_read(&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)
{
	if (fle->object)
		fle->object->ops->delete(fle->object);
	kmem_cache_free(flow_cachep, fle);
}

static void flow_cache_gc_task(struct work_struct *work)
{
	struct list_head gc_list;
	struct flow_cache_entry *fce, *n;

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

	list_for_each_entry_safe(fce, n, &gc_list, u.gc_list)
		flow_entry_kill(fce);
}
static DECLARE_WORK(flow_cache_gc_work, flow_cache_gc_task);

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

	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)) {
				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);
}

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 = &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(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(&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(&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;

	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))
				continue;

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

	flow_cache_queue_garbage(fcp, deleted, &gc_list);

	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(void)
{
	struct flow_flush_info info;
	static DEFINE_MUTEX(flow_flush_sem);
	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(&flow_flush_sem);
	info.cache = &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(&flow_flush_sem);
	put_online_cpus();
	free_cpumask_var(mask);
}

static void flow_cache_flush_task(struct work_struct *work)
{
	flow_cache_flush();
}

static DECLARE_WORK(flow_cache_flush_work, flow_cache_flush_task);

void flow_cache_flush_deferred(void)
{
	schedule_work(&flow_cache_flush_work);
}

static int __cpuinit 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 __cpuinit 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;
}

static int __init flow_cache_init(struct flow_cache *fc)
{
	int i;

	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;
}

static int __init flow_cache_init_global(void)
{
	flow_cachep = kmem_cache_create("flow_cache",
					sizeof(struct flow_cache_entry),
					0, SLAB_PANIC, NULL);

	return flow_cache_init(&flow_cache_global);
}

module_init(flow_cache_init_global);
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>
1da177e4 LT	27
1da177e4 LT	28	struct flow_cache_entry {
8e479560 TT	29	union {
	30	struct hlist_node hlist;
	31	struct list_head gc_list;
	32	} u;
0542b69e	33	struct net *net;
fe1a5f03 TT	34	u16 family;
	35	u8 dir;
	36	u32 genid;
	37	struct flowi key;
	38	struct flow_cache_object *object;
1da177e4 LT	39	};
1da177e4 LT	40
d7997fe1	41	struct flow_cache_percpu {
8e479560	42	struct hlist_head *hash_table;
d7997fe1 TT	43	int hash_count;
	44	u32 hash_rnd;
	45	int hash_rnd_recalc;
	46	struct tasklet_struct flush_tasklet;
5f58a5c8	47	};
1da177e4 LT	48
1da177e4 LT	49	struct flow_flush_info {
fe1a5f03	50	struct flow_cache *cache;
d7997fe1 TT	51	atomic_t cpuleft;
d7997fe1 TT	52	struct completion completion;
1da177e4	53	};
1da177e4	54
d7997fe1 TT	55	struct flow_cache {
d7997fe1 TT	56	u32 hash_shift;
83b6b1f5	57	struct flow_cache_percpu __percpu *percpu;
d7997fe1 TT	58	struct notifier_block hotcpu_notifier;
	59	int low_watermark;
	60	int high_watermark;
	61	struct timer_list rnd_timer;
	62	};
	63
	64	atomic_t flow_cache_genid = ATOMIC_INIT(0);
9e34a5b5	65	EXPORT_SYMBOL(flow_cache_genid);
d7997fe1	66	static struct flow_cache flow_cache_global;
83b6b1f5	67	static struct kmem_cache *flow_cachep __read_mostly;
d7997fe1	68
8e479560 TT	69	static DEFINE_SPINLOCK(flow_cache_gc_lock);
	70	static LIST_HEAD(flow_cache_gc_list);
	71
d7997fe1 TT	72	#define flow_cache_hash_size(cache) (1 << (cache)->hash_shift)
d7997fe1 TT	73	#define FLOW_HASH_RND_PERIOD (10 * 60 * HZ)
1da177e4 LT	74
	75	static void flow_cache_new_hashrnd(unsigned long arg)
	76	{
d7997fe1	77	struct flow_cache fc = (void ) arg;
1da177e4 LT	78	int i;
1da177e4 LT	79
6f912042	80	for_each_possible_cpu(i)
d7997fe1	81	per_cpu_ptr(fc->percpu, i)->hash_rnd_recalc = 1;
1da177e4	82
d7997fe1 TT	83	fc->rnd_timer.expires = jiffies + FLOW_HASH_RND_PERIOD;
d7997fe1 TT	84	add_timer(&fc->rnd_timer);
1da177e4 LT	85	}
1da177e4 LT	86
fe1a5f03 TT	87	static int flow_entry_valid(struct flow_cache_entry *fle)
	88	{
	89	if (atomic_read(&flow_cache_genid) != fle->genid)
	90	return 0;
	91	if (fle->object && !fle->object->ops->check(fle->object))
	92	return 0;
	93	return 1;
	94	}
	95
8e479560	96	static void flow_entry_kill(struct flow_cache_entry *fle)
134b0fc5 JM	97	{
134b0fc5 JM	98	if (fle->object)
fe1a5f03	99	fle->object->ops->delete(fle->object);
134b0fc5	100	kmem_cache_free(flow_cachep, fle);
8e479560 TT	101	}
	102
	103	static void flow_cache_gc_task(struct work_struct *work)
	104	{
	105	struct list_head gc_list;
	106	struct flow_cache_entry fce, n;
	107
	108	INIT_LIST_HEAD(&gc_list);
	109	spin_lock_bh(&flow_cache_gc_lock);
	110	list_splice_tail_init(&flow_cache_gc_list, &gc_list);
	111	spin_unlock_bh(&flow_cache_gc_lock);
	112
	113	list_for_each_entry_safe(fce, n, &gc_list, u.gc_list)
	114	flow_entry_kill(fce);
	115	}
	116	static DECLARE_WORK(flow_cache_gc_work, flow_cache_gc_task);
	117
	118	static void flow_cache_queue_garbage(struct flow_cache_percpu *fcp,
	119	int deleted, struct list_head *gc_list)
	120	{
	121	if (deleted) {
	122	fcp->hash_count -= deleted;
	123	spin_lock_bh(&flow_cache_gc_lock);
	124	list_splice_tail(gc_list, &flow_cache_gc_list);
	125	spin_unlock_bh(&flow_cache_gc_lock);
	126	schedule_work(&flow_cache_gc_work);
	127	}
134b0fc5 JM	128	}
134b0fc5 JM	129
d7997fe1 TT	130	static void __flow_cache_shrink(struct flow_cache *fc,
	131	struct flow_cache_percpu *fcp,
	132	int shrink_to)
1da177e4	133	{
8e479560	134	struct flow_cache_entry *fle;
b67bfe0d	135	struct hlist_node *tmp;
8e479560 TT	136	LIST_HEAD(gc_list);
8e479560 TT	137	int i, deleted = 0;
1da177e4	138
d7997fe1	139	for (i = 0; i < flow_cache_hash_size(fc); i++) {
fe1a5f03	140	int saved = 0;
1da177e4	141
b67bfe0d	142	hlist_for_each_entry_safe(fle, tmp,
8e479560	143	&fcp->hash_table[i], u.hlist) {
fe1a5f03 TT	144	if (saved < shrink_to &&
	145	flow_entry_valid(fle)) {
	146	saved++;
fe1a5f03	147	} else {
8e479560 TT	148	deleted++;
	149	hlist_del(&fle->u.hlist);
	150	list_add_tail(&fle->u.gc_list, &gc_list);
fe1a5f03	151	}
1da177e4 LT	152	}
1da177e4 LT	153	}
8e479560 TT	154
8e479560 TT	155	flow_cache_queue_garbage(fcp, deleted, &gc_list);
1da177e4 LT	156	}
1da177e4 LT	157
d7997fe1 TT	158	static void flow_cache_shrink(struct flow_cache *fc,
d7997fe1 TT	159	struct flow_cache_percpu *fcp)
1da177e4	160	{
d7997fe1	161	int shrink_to = fc->low_watermark / flow_cache_hash_size(fc);
1da177e4	162
d7997fe1	163	__flow_cache_shrink(fc, fcp, shrink_to);
1da177e4 LT	164	}
1da177e4 LT	165
d7997fe1 TT	166	static void flow_new_hash_rnd(struct flow_cache *fc,
d7997fe1 TT	167	struct flow_cache_percpu *fcp)
1da177e4	168	{
d7997fe1 TT	169	get_random_bytes(&fcp->hash_rnd, sizeof(u32));
	170	fcp->hash_rnd_recalc = 0;
	171	__flow_cache_shrink(fc, fcp, 0);
1da177e4 LT	172	}
1da177e4 LT	173
d7997fe1 TT	174	static u32 flow_hash_code(struct flow_cache *fc,
d7997fe1 TT	175	struct flow_cache_percpu *fcp,
aa1c366e	176	const struct flowi *key,
aa1c366e	177	size_t keysize)
1da177e4	178	{
dee9f4bc	179	const u32 k = (const u32 ) key;
aa1c366e	180	const u32 length = keysize * sizeof(flow_compare_t) / sizeof(u32);
1da177e4	181
aa1c366e	182	return jhash2(k, length, fcp->hash_rnd)
a02cec21	183	& (flow_cache_hash_size(fc) - 1);
1da177e4 LT	184	}
1da177e4 LT	185
1da177e4	186	/* I hear what you're saying, use memcmp. But memcmp cannot make
aa1c366e	187	* important assumptions that we can here, such as alignment.
1da177e4	188	*/
aa1c366e	189	static int flow_key_compare(const struct flowi key1, const struct flowi key2,
aa1c366e	190	size_t keysize)
1da177e4	191	{
dee9f4bc	192	const flow_compare_t k1, k1_lim, *k2;
1da177e4	193
dee9f4bc	194	k1 = (const flow_compare_t *) key1;
aa1c366e	195	k1_lim = k1 + keysize;
1da177e4	196
dee9f4bc	197	k2 = (const flow_compare_t *) key2;
1da177e4 LT	198
	199	do {
	200	if (k1++ != k2++)
	201	return 1;
	202	} while (k1 < k1_lim);
	203
	204	return 0;
	205	}
	206
fe1a5f03	207	struct flow_cache_object *
dee9f4bc	208	flow_cache_lookup(struct net net, const struct flowi key, u16 family, u8 dir,
fe1a5f03	209	flow_resolve_t resolver, void *ctx)
1da177e4	210	{
d7997fe1 TT	211	struct flow_cache *fc = &flow_cache_global;
d7997fe1 TT	212	struct flow_cache_percpu *fcp;
8e479560	213	struct flow_cache_entry fle, tfle;
fe1a5f03	214	struct flow_cache_object *flo;
aa1c366e	215	size_t keysize;
1da177e4	216	unsigned int hash;
1da177e4 LT	217
1da177e4 LT	218	local_bh_disable();
7a9b2d59	219	fcp = this_cpu_ptr(fc->percpu);
1da177e4 LT	220
1da177e4 LT	221	fle = NULL;
fe1a5f03	222	flo = NULL;
aa1c366e	223
	224	keysize = flow_key_size(family);
	225	if (!keysize)
	226	goto nocache;
	227
1da177e4 LT	228	/* Packet really early in init? Making flow_cache_init a
1da177e4 LT	229	* pre-smp initcall would solve this. --RR */
d7997fe1	230	if (!fcp->hash_table)
1da177e4 LT	231	goto nocache;
1da177e4 LT	232
d7997fe1 TT	233	if (fcp->hash_rnd_recalc)
d7997fe1 TT	234	flow_new_hash_rnd(fc, fcp);
1da177e4	235
aa1c366e	236	hash = flow_hash_code(fc, fcp, key, keysize);
b67bfe0d	237	hlist_for_each_entry(tfle, &fcp->hash_table[hash], u.hlist) {
0542b69e	238	if (tfle->net == net &&
0542b69e	239	tfle->family == family &&
8e479560	240	tfle->dir == dir &&
aa1c366e	241	flow_key_compare(key, &tfle->key, keysize) == 0) {
8e479560	242	fle = tfle;
1da177e4	243	break;
8e479560	244	}
1da177e4 LT	245	}
1da177e4 LT	246
fe1a5f03	247	if (unlikely(!fle)) {
d7997fe1 TT	248	if (fcp->hash_count > fc->high_watermark)
d7997fe1 TT	249	flow_cache_shrink(fc, fcp);
1da177e4	250
54e6ecb2	251	fle = kmem_cache_alloc(flow_cachep, GFP_ATOMIC);
1da177e4	252	if (fle) {
0542b69e	253	fle->net = net;
1da177e4 LT	254	fle->family = family;
1da177e4 LT	255	fle->dir = dir;
aa1c366e	256	memcpy(&fle->key, key, keysize * sizeof(flow_compare_t));
1da177e4	257	fle->object = NULL;
8e479560	258	hlist_add_head(&fle->u.hlist, &fcp->hash_table[hash]);
d7997fe1	259	fcp->hash_count++;
1da177e4	260	}
fe1a5f03 TT	261	} else if (likely(fle->genid == atomic_read(&flow_cache_genid))) {
	262	flo = fle->object;
	263	if (!flo)
	264	goto ret_object;
	265	flo = flo->ops->get(flo);
	266	if (flo)
	267	goto ret_object;
	268	} else if (fle->object) {
	269	flo = fle->object;
	270	flo->ops->delete(flo);
	271	fle->object = NULL;
1da177e4 LT	272	}
	273
	274	nocache:
fe1a5f03 TT	275	flo = NULL;
	276	if (fle) {
	277	flo = fle->object;
	278	fle->object = NULL;
	279	}
	280	flo = resolver(net, key, family, dir, flo, ctx);
	281	if (fle) {
	282	fle->genid = atomic_read(&flow_cache_genid);
	283	if (!IS_ERR(flo))
	284	fle->object = flo;
	285	else
	286	fle->genid--;
	287	} else {
8fbcec24	288	if (!IS_ERR_OR_NULL(flo))
fe1a5f03	289	flo->ops->delete(flo);
1da177e4	290	}
fe1a5f03 TT	291	ret_object:
	292	local_bh_enable();
	293	return flo;
1da177e4	294	}
9e34a5b5	295	EXPORT_SYMBOL(flow_cache_lookup);
1da177e4 LT	296
	297	static void flow_cache_flush_tasklet(unsigned long data)
	298	{
	299	struct flow_flush_info info = (void )data;
d7997fe1 TT	300	struct flow_cache *fc = info->cache;
d7997fe1 TT	301	struct flow_cache_percpu *fcp;
8e479560	302	struct flow_cache_entry *fle;
b67bfe0d	303	struct hlist_node *tmp;
8e479560 TT	304	LIST_HEAD(gc_list);
8e479560 TT	305	int i, deleted = 0;
1da177e4	306
7a9b2d59	307	fcp = this_cpu_ptr(fc->percpu);
d7997fe1	308	for (i = 0; i < flow_cache_hash_size(fc); i++) {
b67bfe0d	309	hlist_for_each_entry_safe(fle, tmp,
8e479560	310	&fcp->hash_table[i], u.hlist) {
fe1a5f03	311	if (flow_entry_valid(fle))
1da177e4 LT	312	continue;
1da177e4 LT	313
8e479560 TT	314	deleted++;
	315	hlist_del(&fle->u.hlist);
	316	list_add_tail(&fle->u.gc_list, &gc_list);
1da177e4 LT	317	}
	318	}
	319
8e479560 TT	320	flow_cache_queue_garbage(fcp, deleted, &gc_list);
8e479560 TT	321
1da177e4 LT	322	if (atomic_dec_and_test(&info->cpuleft))
	323	complete(&info->completion);
	324	}
	325
8fdc929f CM	326	/*
	327	* Return whether a cpu needs flushing. Conservatively, we assume
	328	* the presence of any entries means the core may require flushing,
	329	* since the flow_cache_ops.check() function may assume it's running
	330	* on the same core as the per-cpu cache component.
	331	*/
	332	static int flow_cache_percpu_empty(struct flow_cache *fc, int cpu)
	333	{
	334	struct flow_cache_percpu *fcp;
	335	int i;
	336
27815032	337	fcp = per_cpu_ptr(fc->percpu, cpu);
8fdc929f CM	338	for (i = 0; i < flow_cache_hash_size(fc); i++)
	339	if (!hlist_empty(&fcp->hash_table[i]))
	340	return 0;
	341	return 1;
	342	}
	343
1da177e4 LT	344	static void flow_cache_flush_per_cpu(void *data)
	345	{
	346	struct flow_flush_info *info = data;
1da177e4 LT	347	struct tasklet_struct *tasklet;
1da177e4 LT	348
50eab050	349	tasklet = &this_cpu_ptr(info->cache->percpu)->flush_tasklet;
1da177e4 LT	350	tasklet->data = (unsigned long)info;
	351	tasklet_schedule(tasklet);
	352	}
	353
	354	void flow_cache_flush(void)
	355	{
	356	struct flow_flush_info info;
4a3e2f71	357	static DEFINE_MUTEX(flow_flush_sem);
8fdc929f CM	358	cpumask_var_t mask;
	359	int i, self;
	360
	361	/* Track which cpus need flushing to avoid disturbing all cores. */
	362	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
	363	return;
	364	cpumask_clear(mask);
1da177e4 LT	365
1da177e4 LT	366	/* Don't want cpus going down or up during this. */
86ef5c9a	367	get_online_cpus();
4a3e2f71	368	mutex_lock(&flow_flush_sem);
d7997fe1	369	info.cache = &flow_cache_global;
8fdc929f CM	370	for_each_online_cpu(i)
	371	if (!flow_cache_percpu_empty(info.cache, i))
	372	cpumask_set_cpu(i, mask);
	373	atomic_set(&info.cpuleft, cpumask_weight(mask));
	374	if (atomic_read(&info.cpuleft) == 0)
	375	goto done;
	376
1da177e4 LT	377	init_completion(&info.completion);
	378
	379	local_bh_disable();
8fdc929f CM	380	self = cpumask_test_and_clear_cpu(smp_processor_id(), mask);
	381	on_each_cpu_mask(mask, flow_cache_flush_per_cpu, &info, 0);
	382	if (self)
	383	flow_cache_flush_tasklet((unsigned long)&info);
1da177e4 LT	384	local_bh_enable();
	385
	386	wait_for_completion(&info.completion);
8fdc929f CM	387
8fdc929f CM	388	done:
4a3e2f71	389	mutex_unlock(&flow_flush_sem);
86ef5c9a	390	put_online_cpus();
8fdc929f	391	free_cpumask_var(mask);
1da177e4 LT	392	}
1da177e4 LT	393
c0ed1c14 SK	394	static void flow_cache_flush_task(struct work_struct *work)
	395	{
	396	flow_cache_flush();
	397	}
	398
	399	static DECLARE_WORK(flow_cache_flush_work, flow_cache_flush_task);
	400
	401	void flow_cache_flush_deferred(void)
	402	{
	403	schedule_work(&flow_cache_flush_work);
	404	}
	405
83b6b1f5	406	static int __cpuinit flow_cache_cpu_prepare(struct flow_cache *fc, int cpu)
1da177e4	407	{
83b6b1f5 ED	408	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, cpu);
83b6b1f5 ED	409	size_t sz = sizeof(struct hlist_head) * flow_cache_hash_size(fc);
d7997fe1	410
83b6b1f5 ED	411	if (!fcp->hash_table) {
	412	fcp->hash_table = kzalloc_node(sz, GFP_KERNEL, cpu_to_node(cpu));
	413	if (!fcp->hash_table) {
	414	pr_err("NET: failed to allocate flow cache sz %zu\n", sz);
	415	return -ENOMEM;
	416	}
	417	fcp->hash_rnd_recalc = 1;
	418	fcp->hash_count = 0;
	419	tasklet_init(&fcp->flush_tasklet, flow_cache_flush_tasklet, 0);
	420	}
	421	return 0;
1da177e4 LT	422	}
1da177e4 LT	423
83b6b1f5	424	static int __cpuinit flow_cache_cpu(struct notifier_block *nfb,
1da177e4 LT	425	unsigned long action,
	426	void *hcpu)
	427	{
d7997fe1	428	struct flow_cache *fc = container_of(nfb, struct flow_cache, hotcpu_notifier);
83b6b1f5	429	int res, cpu = (unsigned long) hcpu;
d7997fe1 TT	430	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, cpu);
d7997fe1 TT	431
83b6b1f5 ED	432	switch (action) {
	433	case CPU_UP_PREPARE:
	434	case CPU_UP_PREPARE_FROZEN:
	435	res = flow_cache_cpu_prepare(fc, cpu);
	436	if (res)
	437	return notifier_from_errno(res);
	438	break;
	439	case CPU_DEAD:
	440	case CPU_DEAD_FROZEN:
d7997fe1	441	__flow_cache_shrink(fc, fcp, 0);
83b6b1f5 ED	442	break;
83b6b1f5 ED	443	}
1da177e4 LT	444	return NOTIFY_OK;
1da177e4 LT	445	}
1da177e4	446
83b6b1f5	447	static int __init flow_cache_init(struct flow_cache *fc)
1da177e4 LT	448	{
	449	int i;
	450
d7997fe1 TT	451	fc->hash_shift = 10;
	452	fc->low_watermark = 2 * flow_cache_hash_size(fc);
	453	fc->high_watermark = 4 * flow_cache_hash_size(fc);
	454
d7997fe1	455	fc->percpu = alloc_percpu(struct flow_cache_percpu);
83b6b1f5 ED	456	if (!fc->percpu)
83b6b1f5 ED	457	return -ENOMEM;
1da177e4	458
83b6b1f5 ED	459	for_each_online_cpu(i) {
83b6b1f5 ED	460	if (flow_cache_cpu_prepare(fc, i))
6ccc3abd	461	goto err;
83b6b1f5	462	}
d7997fe1 TT	463	fc->hotcpu_notifier = (struct notifier_block){
	464	.notifier_call = flow_cache_cpu,
	465	};
	466	register_hotcpu_notifier(&fc->hotcpu_notifier);
1da177e4	467
83b6b1f5 ED	468	setup_timer(&fc->rnd_timer, flow_cache_new_hashrnd,
	469	(unsigned long) fc);
	470	fc->rnd_timer.expires = jiffies + FLOW_HASH_RND_PERIOD;
	471	add_timer(&fc->rnd_timer);
	472
1da177e4	473	return 0;
6ccc3abd	474
	475	err:
	476	for_each_possible_cpu(i) {
	477	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, i);
	478	kfree(fcp->hash_table);
	479	fcp->hash_table = NULL;
	480	}
	481
	482	free_percpu(fc->percpu);
	483	fc->percpu = NULL;
	484
	485	return -ENOMEM;
1da177e4 LT	486	}
1da177e4 LT	487
d7997fe1 TT	488	static int __init flow_cache_init_global(void)
	489	{
	490	flow_cachep = kmem_cache_create("flow_cache",
	491	sizeof(struct flow_cache_entry),
	492	0, SLAB_PANIC, NULL);
	493
	494	return flow_cache_init(&flow_cache_global);
	495	}
	496
	497	module_init(flow_cache_init_global);