[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>

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 *entry, *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, entry, 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 hlist_node *entry;
	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, entry, &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 (flo && !IS_ERR(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 *entry, *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, entry, 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);
}

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

	cpu = smp_processor_id();
	tasklet = &per_cpu_ptr(info->cache->percpu, cpu)->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);

	/* Don't want cpus going down or up during this. */
	get_online_cpus();
	mutex_lock(&flow_flush_sem);
	info.cache = &flow_cache_global;
	atomic_set(&info.cpuleft, num_online_cpus());
	init_completion(&info.completion);

	local_bh_disable();
	smp_call_function(flow_cache_flush_per_cpu, &info, 0);
	flow_cache_flush_tasklet((unsigned long)&info);
	local_bh_enable();

	wait_for_completion(&info.completion);
	mutex_unlock(&flow_flush_sem);
	put_online_cpus();
}

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 TT	134	struct flow_cache_entry *fle;
	135	struct hlist_node entry, tmp;
	136	LIST_HEAD(gc_list);
	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
8e479560 TT	142	hlist_for_each_entry_safe(fle, entry, tmp,
8e479560 TT	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 TT	213	struct flow_cache_entry fle, tfle;
8e479560 TT	214	struct hlist_node *entry;
fe1a5f03	215	struct flow_cache_object *flo;
aa1c366e	216	size_t keysize;
1da177e4	217	unsigned int hash;
1da177e4 LT	218
1da177e4 LT	219	local_bh_disable();
7a9b2d59	220	fcp = this_cpu_ptr(fc->percpu);
1da177e4 LT	221
1da177e4 LT	222	fle = NULL;
fe1a5f03	223	flo = NULL;
aa1c366e	224
	225	keysize = flow_key_size(family);
	226	if (!keysize)
	227	goto nocache;
	228
1da177e4 LT	229	/* Packet really early in init? Making flow_cache_init a
1da177e4 LT	230	* pre-smp initcall would solve this. --RR */
d7997fe1	231	if (!fcp->hash_table)
1da177e4 LT	232	goto nocache;
1da177e4 LT	233
d7997fe1 TT	234	if (fcp->hash_rnd_recalc)
d7997fe1 TT	235	flow_new_hash_rnd(fc, fcp);
1da177e4	236
aa1c366e	237	hash = flow_hash_code(fc, fcp, key, keysize);
8e479560	238	hlist_for_each_entry(tfle, entry, &fcp->hash_table[hash], u.hlist) {
0542b69e	239	if (tfle->net == net &&
0542b69e	240	tfle->family == family &&
8e479560	241	tfle->dir == dir &&
aa1c366e	242	flow_key_compare(key, &tfle->key, keysize) == 0) {
8e479560	243	fle = tfle;
1da177e4	244	break;
8e479560	245	}
1da177e4 LT	246	}
1da177e4 LT	247
fe1a5f03	248	if (unlikely(!fle)) {
d7997fe1 TT	249	if (fcp->hash_count > fc->high_watermark)
d7997fe1 TT	250	flow_cache_shrink(fc, fcp);
1da177e4	251
54e6ecb2	252	fle = kmem_cache_alloc(flow_cachep, GFP_ATOMIC);
1da177e4	253	if (fle) {
0542b69e	254	fle->net = net;
1da177e4 LT	255	fle->family = family;
1da177e4 LT	256	fle->dir = dir;
aa1c366e	257	memcpy(&fle->key, key, keysize * sizeof(flow_compare_t));
1da177e4	258	fle->object = NULL;
8e479560	259	hlist_add_head(&fle->u.hlist, &fcp->hash_table[hash]);
d7997fe1	260	fcp->hash_count++;
1da177e4	261	}
fe1a5f03 TT	262	} else if (likely(fle->genid == atomic_read(&flow_cache_genid))) {
	263	flo = fle->object;
	264	if (!flo)
	265	goto ret_object;
	266	flo = flo->ops->get(flo);
	267	if (flo)
	268	goto ret_object;
	269	} else if (fle->object) {
	270	flo = fle->object;
	271	flo->ops->delete(flo);
	272	fle->object = NULL;
1da177e4 LT	273	}
	274
	275	nocache:
fe1a5f03 TT	276	flo = NULL;
	277	if (fle) {
	278	flo = fle->object;
	279	fle->object = NULL;
	280	}
	281	flo = resolver(net, key, family, dir, flo, ctx);
	282	if (fle) {
	283	fle->genid = atomic_read(&flow_cache_genid);
	284	if (!IS_ERR(flo))
	285	fle->object = flo;
	286	else
	287	fle->genid--;
	288	} else {
	289	if (flo && !IS_ERR(flo))
	290	flo->ops->delete(flo);
1da177e4	291	}
fe1a5f03 TT	292	ret_object:
	293	local_bh_enable();
	294	return flo;
1da177e4	295	}
9e34a5b5	296	EXPORT_SYMBOL(flow_cache_lookup);
1da177e4 LT	297
	298	static void flow_cache_flush_tasklet(unsigned long data)
	299	{
	300	struct flow_flush_info info = (void )data;
d7997fe1 TT	301	struct flow_cache *fc = info->cache;
d7997fe1 TT	302	struct flow_cache_percpu *fcp;
8e479560 TT	303	struct flow_cache_entry *fle;
	304	struct hlist_node entry, tmp;
	305	LIST_HEAD(gc_list);
	306	int i, deleted = 0;
1da177e4	307
7a9b2d59	308	fcp = this_cpu_ptr(fc->percpu);
d7997fe1	309	for (i = 0; i < flow_cache_hash_size(fc); i++) {
8e479560 TT	310	hlist_for_each_entry_safe(fle, entry, tmp,
8e479560 TT	311	&fcp->hash_table[i], u.hlist) {
fe1a5f03	312	if (flow_entry_valid(fle))
1da177e4 LT	313	continue;
1da177e4 LT	314
8e479560 TT	315	deleted++;
	316	hlist_del(&fle->u.hlist);
	317	list_add_tail(&fle->u.gc_list, &gc_list);
1da177e4 LT	318	}
	319	}
	320
8e479560 TT	321	flow_cache_queue_garbage(fcp, deleted, &gc_list);
8e479560 TT	322
1da177e4 LT	323	if (atomic_dec_and_test(&info->cpuleft))
	324	complete(&info->completion);
	325	}
	326
1da177e4 LT	327	static void flow_cache_flush_per_cpu(void *data)
	328	{
	329	struct flow_flush_info *info = data;
	330	int cpu;
	331	struct tasklet_struct *tasklet;
	332
	333	cpu = smp_processor_id();
d7997fe1	334	tasklet = &per_cpu_ptr(info->cache->percpu, cpu)->flush_tasklet;
1da177e4 LT	335	tasklet->data = (unsigned long)info;
	336	tasklet_schedule(tasklet);
	337	}
	338
	339	void flow_cache_flush(void)
	340	{
	341	struct flow_flush_info info;
4a3e2f71	342	static DEFINE_MUTEX(flow_flush_sem);
1da177e4 LT	343
1da177e4 LT	344	/* Don't want cpus going down or up during this. */
86ef5c9a	345	get_online_cpus();
4a3e2f71	346	mutex_lock(&flow_flush_sem);
d7997fe1	347	info.cache = &flow_cache_global;
1da177e4 LT	348	atomic_set(&info.cpuleft, num_online_cpus());
	349	init_completion(&info.completion);
	350
	351	local_bh_disable();
8691e5a8	352	smp_call_function(flow_cache_flush_per_cpu, &info, 0);
1da177e4 LT	353	flow_cache_flush_tasklet((unsigned long)&info);
	354	local_bh_enable();
	355
	356	wait_for_completion(&info.completion);
4a3e2f71	357	mutex_unlock(&flow_flush_sem);
86ef5c9a	358	put_online_cpus();
1da177e4 LT	359	}
1da177e4 LT	360
c0ed1c14 SK	361	static void flow_cache_flush_task(struct work_struct *work)
	362	{
	363	flow_cache_flush();
	364	}
	365
	366	static DECLARE_WORK(flow_cache_flush_work, flow_cache_flush_task);
	367
	368	void flow_cache_flush_deferred(void)
	369	{
	370	schedule_work(&flow_cache_flush_work);
	371	}
	372
83b6b1f5	373	static int __cpuinit flow_cache_cpu_prepare(struct flow_cache *fc, int cpu)
1da177e4	374	{
83b6b1f5 ED	375	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, cpu);
83b6b1f5 ED	376	size_t sz = sizeof(struct hlist_head) * flow_cache_hash_size(fc);
d7997fe1	377
83b6b1f5 ED	378	if (!fcp->hash_table) {
	379	fcp->hash_table = kzalloc_node(sz, GFP_KERNEL, cpu_to_node(cpu));
	380	if (!fcp->hash_table) {
	381	pr_err("NET: failed to allocate flow cache sz %zu\n", sz);
	382	return -ENOMEM;
	383	}
	384	fcp->hash_rnd_recalc = 1;
	385	fcp->hash_count = 0;
	386	tasklet_init(&fcp->flush_tasklet, flow_cache_flush_tasklet, 0);
	387	}
	388	return 0;
1da177e4 LT	389	}
1da177e4 LT	390
83b6b1f5	391	static int __cpuinit flow_cache_cpu(struct notifier_block *nfb,
1da177e4 LT	392	unsigned long action,
	393	void *hcpu)
	394	{
d7997fe1	395	struct flow_cache *fc = container_of(nfb, struct flow_cache, hotcpu_notifier);
83b6b1f5	396	int res, cpu = (unsigned long) hcpu;
d7997fe1 TT	397	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, cpu);
d7997fe1 TT	398
83b6b1f5 ED	399	switch (action) {
	400	case CPU_UP_PREPARE:
	401	case CPU_UP_PREPARE_FROZEN:
	402	res = flow_cache_cpu_prepare(fc, cpu);
	403	if (res)
	404	return notifier_from_errno(res);
	405	break;
	406	case CPU_DEAD:
	407	case CPU_DEAD_FROZEN:
d7997fe1	408	__flow_cache_shrink(fc, fcp, 0);
83b6b1f5 ED	409	break;
83b6b1f5 ED	410	}
1da177e4 LT	411	return NOTIFY_OK;
1da177e4 LT	412	}
1da177e4	413
83b6b1f5	414	static int __init flow_cache_init(struct flow_cache *fc)
1da177e4 LT	415	{
	416	int i;
	417
d7997fe1 TT	418	fc->hash_shift = 10;
	419	fc->low_watermark = 2 * flow_cache_hash_size(fc);
	420	fc->high_watermark = 4 * flow_cache_hash_size(fc);
	421
d7997fe1	422	fc->percpu = alloc_percpu(struct flow_cache_percpu);
83b6b1f5 ED	423	if (!fc->percpu)
83b6b1f5 ED	424	return -ENOMEM;
1da177e4	425
83b6b1f5 ED	426	for_each_online_cpu(i) {
83b6b1f5 ED	427	if (flow_cache_cpu_prepare(fc, i))
6ccc3abd	428	goto err;
83b6b1f5	429	}
d7997fe1 TT	430	fc->hotcpu_notifier = (struct notifier_block){
	431	.notifier_call = flow_cache_cpu,
	432	};
	433	register_hotcpu_notifier(&fc->hotcpu_notifier);
1da177e4	434
83b6b1f5 ED	435	setup_timer(&fc->rnd_timer, flow_cache_new_hashrnd,
	436	(unsigned long) fc);
	437	fc->rnd_timer.expires = jiffies + FLOW_HASH_RND_PERIOD;
	438	add_timer(&fc->rnd_timer);
	439
1da177e4	440	return 0;
6ccc3abd	441
	442	err:
	443	for_each_possible_cpu(i) {
	444	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, i);
	445	kfree(fcp->hash_table);
	446	fcp->hash_table = NULL;
	447	}
	448
	449	free_percpu(fc->percpu);
	450	fc->percpu = NULL;
	451
	452	return -ENOMEM;
1da177e4 LT	453	}
1da177e4 LT	454
d7997fe1 TT	455	static int __init flow_cache_init_global(void)
	456	{
	457	flow_cachep = kmem_cache_create("flow_cache",
	458	sizeof(struct flow_cache_entry),
	459	0, SLAB_PANIC, NULL);
	460
	461	return flow_cache_init(&flow_cache_global);
	462	}
	463
	464	module_init(flow_cache_init_global);