[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;
};

static struct kmem_cache *flow_cachep __read_mostly;

#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(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);
		atomic_dec(&xfrm->flow_cache_gc_count);
		WARN_ON(atomic_read(&xfrm->flow_cache_gc_count) < 0);
	}
}

static void flow_cache_queue_garbage(struct flow_cache_percpu *fcp,
				     int deleted, struct list_head *gc_list,
				     struct netns_xfrm *xfrm)
{
	if (deleted) {
		atomic_add(deleted, &xfrm->flow_cache_gc_count);
		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);

		if (fcp->hash_count > 2 * fc->high_watermark ||
		    atomic_read(&net->xfrm.flow_cache_gc_count) > fc->high_watermark) {
			atomic_inc(&net->xfrm.flow_cache_genid);
			flo = ERR_PTR(-ENOBUFS);
			goto ret_object;
		}

		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(&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_flush_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;

	if (!flow_cachep)
		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);
	atomic_set(&net->xfrm.flow_cache_gc_count, 0);

	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;

	cpu_notifier_register_begin();

	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);

	cpu_notifier_register_done();

	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;
	}

	cpu_notifier_register_done();

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

	return -ENOMEM;
}
EXPORT_SYMBOL(flow_cache_init);

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

	del_timer_sync(&fc->rnd_timer);
	unregister_hotcpu_notifier(&fc->hotcpu_notifier);

	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;
}
EXPORT_SYMBOL(flow_cache_fini);
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
d32d9bb8 ED	48	static struct kmem_cache *flow_cachep __read_mostly;
d32d9bb8 ED	49
d7997fe1 TT	50	#define flow_cache_hash_size(cache) (1 << (cache)->hash_shift)
d7997fe1 TT	51	#define FLOW_HASH_RND_PERIOD (10 * 60 * HZ)
1da177e4 LT	52
	53	static void flow_cache_new_hashrnd(unsigned long arg)
	54	{
d7997fe1	55	struct flow_cache fc = (void ) arg;
1da177e4 LT	56	int i;
1da177e4 LT	57
6f912042	58	for_each_possible_cpu(i)
d7997fe1	59	per_cpu_ptr(fc->percpu, i)->hash_rnd_recalc = 1;
1da177e4	60
d7997fe1 TT	61	fc->rnd_timer.expires = jiffies + FLOW_HASH_RND_PERIOD;
d7997fe1 TT	62	add_timer(&fc->rnd_timer);
1da177e4 LT	63	}
1da177e4 LT	64
ca925cf1 FD	65	static int flow_entry_valid(struct flow_cache_entry *fle,
ca925cf1 FD	66	struct netns_xfrm *xfrm)
fe1a5f03	67	{
ca925cf1	68	if (atomic_read(&xfrm->flow_cache_genid) != fle->genid)
fe1a5f03 TT	69	return 0;
	70	if (fle->object && !fle->object->ops->check(fle->object))
	71	return 0;
	72	return 1;
	73	}
	74
ca925cf1 FD	75	static void flow_entry_kill(struct flow_cache_entry *fle,
ca925cf1 FD	76	struct netns_xfrm *xfrm)
134b0fc5 JM	77	{
134b0fc5 JM	78	if (fle->object)
fe1a5f03	79	fle->object->ops->delete(fle->object);
d32d9bb8	80	kmem_cache_free(flow_cachep, fle);
8e479560 TT	81	}
	82
	83	static void flow_cache_gc_task(struct work_struct *work)
	84	{
	85	struct list_head gc_list;
	86	struct flow_cache_entry fce, n;
ca925cf1 FD	87	struct netns_xfrm *xfrm = container_of(work, struct netns_xfrm,
ca925cf1 FD	88	flow_cache_gc_work);
8e479560 TT	89
8e479560 TT	90	INIT_LIST_HEAD(&gc_list);
ca925cf1 FD	91	spin_lock_bh(&xfrm->flow_cache_gc_lock);
	92	list_splice_tail_init(&xfrm->flow_cache_gc_list, &gc_list);
	93	spin_unlock_bh(&xfrm->flow_cache_gc_lock);
8e479560	94
6ad3122a	95	list_for_each_entry_safe(fce, n, &gc_list, u.gc_list) {
ca925cf1	96	flow_entry_kill(fce, xfrm);
6ad3122a SK	97	atomic_dec(&xfrm->flow_cache_gc_count);
	98	WARN_ON(atomic_read(&xfrm->flow_cache_gc_count) < 0);
	99	}
8e479560	100	}
8e479560 TT	101
8e479560 TT	102	static void flow_cache_queue_garbage(struct flow_cache_percpu *fcp,
ca925cf1 FD	103	int deleted, struct list_head *gc_list,
ca925cf1 FD	104	struct netns_xfrm *xfrm)
8e479560 TT	105	{
8e479560 TT	106	if (deleted) {
6ad3122a	107	atomic_add(deleted, &xfrm->flow_cache_gc_count);
8e479560	108	fcp->hash_count -= deleted;
ca925cf1 FD	109	spin_lock_bh(&xfrm->flow_cache_gc_lock);
	110	list_splice_tail(gc_list, &xfrm->flow_cache_gc_list);
	111	spin_unlock_bh(&xfrm->flow_cache_gc_lock);
	112	schedule_work(&xfrm->flow_cache_gc_work);
8e479560	113	}
134b0fc5 JM	114	}
134b0fc5 JM	115
d7997fe1 TT	116	static void __flow_cache_shrink(struct flow_cache *fc,
	117	struct flow_cache_percpu *fcp,
	118	int shrink_to)
1da177e4	119	{
8e479560	120	struct flow_cache_entry *fle;
b67bfe0d	121	struct hlist_node *tmp;
8e479560 TT	122	LIST_HEAD(gc_list);
8e479560 TT	123	int i, deleted = 0;
ca925cf1 FD	124	struct netns_xfrm *xfrm = container_of(fc, struct netns_xfrm,
ca925cf1 FD	125	flow_cache_global);
1da177e4	126
d7997fe1	127	for (i = 0; i < flow_cache_hash_size(fc); i++) {
fe1a5f03	128	int saved = 0;
1da177e4	129
b67bfe0d	130	hlist_for_each_entry_safe(fle, tmp,
8e479560	131	&fcp->hash_table[i], u.hlist) {
fe1a5f03	132	if (saved < shrink_to &&
ca925cf1	133	flow_entry_valid(fle, xfrm)) {
fe1a5f03	134	saved++;
fe1a5f03	135	} else {
8e479560 TT	136	deleted++;
	137	hlist_del(&fle->u.hlist);
	138	list_add_tail(&fle->u.gc_list, &gc_list);
fe1a5f03	139	}
1da177e4 LT	140	}
1da177e4 LT	141	}
8e479560	142
ca925cf1	143	flow_cache_queue_garbage(fcp, deleted, &gc_list, xfrm);
1da177e4 LT	144	}
1da177e4 LT	145
d7997fe1 TT	146	static void flow_cache_shrink(struct flow_cache *fc,
d7997fe1 TT	147	struct flow_cache_percpu *fcp)
1da177e4	148	{
d7997fe1	149	int shrink_to = fc->low_watermark / flow_cache_hash_size(fc);
1da177e4	150
d7997fe1	151	__flow_cache_shrink(fc, fcp, shrink_to);
1da177e4 LT	152	}
1da177e4 LT	153
d7997fe1 TT	154	static void flow_new_hash_rnd(struct flow_cache *fc,
d7997fe1 TT	155	struct flow_cache_percpu *fcp)
1da177e4	156	{
d7997fe1 TT	157	get_random_bytes(&fcp->hash_rnd, sizeof(u32));
	158	fcp->hash_rnd_recalc = 0;
	159	__flow_cache_shrink(fc, fcp, 0);
1da177e4 LT	160	}
1da177e4 LT	161
d7997fe1 TT	162	static u32 flow_hash_code(struct flow_cache *fc,
d7997fe1 TT	163	struct flow_cache_percpu *fcp,
aa1c366e	164	const struct flowi *key,
aa1c366e	165	size_t keysize)
1da177e4	166	{
dee9f4bc	167	const u32 k = (const u32 ) key;
aa1c366e	168	const u32 length = keysize * sizeof(flow_compare_t) / sizeof(u32);
1da177e4	169
aa1c366e	170	return jhash2(k, length, fcp->hash_rnd)
a02cec21	171	& (flow_cache_hash_size(fc) - 1);
1da177e4 LT	172	}
1da177e4 LT	173
1da177e4	174	/* I hear what you're saying, use memcmp. But memcmp cannot make
aa1c366e	175	* important assumptions that we can here, such as alignment.
1da177e4	176	*/
aa1c366e	177	static int flow_key_compare(const struct flowi key1, const struct flowi key2,
aa1c366e	178	size_t keysize)
1da177e4	179	{
dee9f4bc	180	const flow_compare_t k1, k1_lim, *k2;
1da177e4	181
dee9f4bc	182	k1 = (const flow_compare_t *) key1;
aa1c366e	183	k1_lim = k1 + keysize;
1da177e4	184
dee9f4bc	185	k2 = (const flow_compare_t *) key2;
1da177e4 LT	186
	187	do {
	188	if (k1++ != k2++)
	189	return 1;
	190	} while (k1 < k1_lim);
	191
	192	return 0;
	193	}
	194
fe1a5f03	195	struct flow_cache_object *
dee9f4bc	196	flow_cache_lookup(struct net net, const struct flowi key, u16 family, u8 dir,
fe1a5f03	197	flow_resolve_t resolver, void *ctx)
1da177e4	198	{
ca925cf1	199	struct flow_cache *fc = &net->xfrm.flow_cache_global;
d7997fe1	200	struct flow_cache_percpu *fcp;
8e479560	201	struct flow_cache_entry fle, tfle;
fe1a5f03	202	struct flow_cache_object *flo;
aa1c366e	203	size_t keysize;
1da177e4	204	unsigned int hash;
1da177e4 LT	205
1da177e4 LT	206	local_bh_disable();
7a9b2d59	207	fcp = this_cpu_ptr(fc->percpu);
1da177e4 LT	208
1da177e4 LT	209	fle = NULL;
fe1a5f03	210	flo = NULL;
aa1c366e	211
	212	keysize = flow_key_size(family);
	213	if (!keysize)
	214	goto nocache;
	215
1da177e4 LT	216	/* Packet really early in init? Making flow_cache_init a
1da177e4 LT	217	* pre-smp initcall would solve this. --RR */
d7997fe1	218	if (!fcp->hash_table)
1da177e4 LT	219	goto nocache;
1da177e4 LT	220
d7997fe1 TT	221	if (fcp->hash_rnd_recalc)
d7997fe1 TT	222	flow_new_hash_rnd(fc, fcp);
1da177e4	223
aa1c366e	224	hash = flow_hash_code(fc, fcp, key, keysize);
b67bfe0d	225	hlist_for_each_entry(tfle, &fcp->hash_table[hash], u.hlist) {
0542b69e	226	if (tfle->net == net &&
0542b69e	227	tfle->family == family &&
8e479560	228	tfle->dir == dir &&
aa1c366e	229	flow_key_compare(key, &tfle->key, keysize) == 0) {
8e479560	230	fle = tfle;
1da177e4	231	break;
8e479560	232	}
1da177e4 LT	233	}
1da177e4 LT	234
fe1a5f03	235	if (unlikely(!fle)) {
d7997fe1 TT	236	if (fcp->hash_count > fc->high_watermark)
d7997fe1 TT	237	flow_cache_shrink(fc, fcp);
1da177e4	238
6ad3122a SK	239	if (fcp->hash_count > 2 * fc->high_watermark \|\|
	240	atomic_read(&net->xfrm.flow_cache_gc_count) > fc->high_watermark) {
	241	atomic_inc(&net->xfrm.flow_cache_genid);
	242	flo = ERR_PTR(-ENOBUFS);
	243	goto ret_object;
	244	}
	245
d32d9bb8	246	fle = kmem_cache_alloc(flow_cachep, GFP_ATOMIC);
1da177e4	247	if (fle) {
0542b69e	248	fle->net = net;
1da177e4 LT	249	fle->family = family;
1da177e4 LT	250	fle->dir = dir;
aa1c366e	251	memcpy(&fle->key, key, keysize * sizeof(flow_compare_t));
1da177e4	252	fle->object = NULL;
8e479560	253	hlist_add_head(&fle->u.hlist, &fcp->hash_table[hash]);
d7997fe1	254	fcp->hash_count++;
1da177e4	255	}
ca925cf1	256	} else if (likely(fle->genid == atomic_read(&net->xfrm.flow_cache_genid))) {
fe1a5f03 TT	257	flo = fle->object;
	258	if (!flo)
	259	goto ret_object;
	260	flo = flo->ops->get(flo);
	261	if (flo)
	262	goto ret_object;
	263	} else if (fle->object) {
	264	flo = fle->object;
	265	flo->ops->delete(flo);
	266	fle->object = NULL;
1da177e4 LT	267	}
	268
	269	nocache:
fe1a5f03 TT	270	flo = NULL;
	271	if (fle) {
	272	flo = fle->object;
	273	fle->object = NULL;
	274	}
	275	flo = resolver(net, key, family, dir, flo, ctx);
	276	if (fle) {
ca925cf1	277	fle->genid = atomic_read(&net->xfrm.flow_cache_genid);
fe1a5f03 TT	278	if (!IS_ERR(flo))
	279	fle->object = flo;
	280	else
	281	fle->genid--;
	282	} else {
8fbcec24	283	if (!IS_ERR_OR_NULL(flo))
fe1a5f03	284	flo->ops->delete(flo);
1da177e4	285	}
fe1a5f03 TT	286	ret_object:
	287	local_bh_enable();
	288	return flo;
1da177e4	289	}
9e34a5b5	290	EXPORT_SYMBOL(flow_cache_lookup);
1da177e4 LT	291
	292	static void flow_cache_flush_tasklet(unsigned long data)
	293	{
	294	struct flow_flush_info info = (void )data;
d7997fe1 TT	295	struct flow_cache *fc = info->cache;
d7997fe1 TT	296	struct flow_cache_percpu *fcp;
8e479560	297	struct flow_cache_entry *fle;
b67bfe0d	298	struct hlist_node *tmp;
8e479560 TT	299	LIST_HEAD(gc_list);
8e479560 TT	300	int i, deleted = 0;
ca925cf1 FD	301	struct netns_xfrm *xfrm = container_of(fc, struct netns_xfrm,
ca925cf1 FD	302	flow_cache_global);
1da177e4	303
7a9b2d59	304	fcp = this_cpu_ptr(fc->percpu);
d7997fe1	305	for (i = 0; i < flow_cache_hash_size(fc); i++) {
b67bfe0d	306	hlist_for_each_entry_safe(fle, tmp,
8e479560	307	&fcp->hash_table[i], u.hlist) {
ca925cf1	308	if (flow_entry_valid(fle, xfrm))
1da177e4 LT	309	continue;
1da177e4 LT	310
8e479560 TT	311	deleted++;
	312	hlist_del(&fle->u.hlist);
	313	list_add_tail(&fle->u.gc_list, &gc_list);
1da177e4 LT	314	}
	315	}
	316
ca925cf1	317	flow_cache_queue_garbage(fcp, deleted, &gc_list, xfrm);
8e479560	318
1da177e4 LT	319	if (atomic_dec_and_test(&info->cpuleft))
	320	complete(&info->completion);
	321	}
	322
8fdc929f CM	323	/*
	324	* Return whether a cpu needs flushing. Conservatively, we assume
	325	* the presence of any entries means the core may require flushing,
	326	* since the flow_cache_ops.check() function may assume it's running
	327	* on the same core as the per-cpu cache component.
	328	*/
	329	static int flow_cache_percpu_empty(struct flow_cache *fc, int cpu)
	330	{
	331	struct flow_cache_percpu *fcp;
	332	int i;
	333
27815032	334	fcp = per_cpu_ptr(fc->percpu, cpu);
8fdc929f CM	335	for (i = 0; i < flow_cache_hash_size(fc); i++)
	336	if (!hlist_empty(&fcp->hash_table[i]))
	337	return 0;
	338	return 1;
	339	}
	340
1da177e4 LT	341	static void flow_cache_flush_per_cpu(void *data)
	342	{
	343	struct flow_flush_info *info = data;
1da177e4 LT	344	struct tasklet_struct *tasklet;
1da177e4 LT	345
50eab050	346	tasklet = &this_cpu_ptr(info->cache->percpu)->flush_tasklet;
1da177e4 LT	347	tasklet->data = (unsigned long)info;
	348	tasklet_schedule(tasklet);
	349	}
	350
ca925cf1	351	void flow_cache_flush(struct net *net)
1da177e4 LT	352	{
1da177e4 LT	353	struct flow_flush_info info;
8fdc929f CM	354	cpumask_var_t mask;
	355	int i, self;
	356
	357	/* Track which cpus need flushing to avoid disturbing all cores. */
	358	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
	359	return;
	360	cpumask_clear(mask);
1da177e4 LT	361
1da177e4 LT	362	/* Don't want cpus going down or up during this. */
86ef5c9a	363	get_online_cpus();
ca925cf1 FD	364	mutex_lock(&net->xfrm.flow_flush_sem);
ca925cf1 FD	365	info.cache = &net->xfrm.flow_cache_global;
8fdc929f CM	366	for_each_online_cpu(i)
	367	if (!flow_cache_percpu_empty(info.cache, i))
	368	cpumask_set_cpu(i, mask);
	369	atomic_set(&info.cpuleft, cpumask_weight(mask));
	370	if (atomic_read(&info.cpuleft) == 0)
	371	goto done;
	372
1da177e4 LT	373	init_completion(&info.completion);
	374
	375	local_bh_disable();
8fdc929f CM	376	self = cpumask_test_and_clear_cpu(smp_processor_id(), mask);
	377	on_each_cpu_mask(mask, flow_cache_flush_per_cpu, &info, 0);
	378	if (self)
	379	flow_cache_flush_tasklet((unsigned long)&info);
1da177e4 LT	380	local_bh_enable();
	381
	382	wait_for_completion(&info.completion);
8fdc929f CM	383
8fdc929f CM	384	done:
ca925cf1	385	mutex_unlock(&net->xfrm.flow_flush_sem);
86ef5c9a	386	put_online_cpus();
8fdc929f	387	free_cpumask_var(mask);
1da177e4 LT	388	}
1da177e4 LT	389
c0ed1c14 SK	390	static void flow_cache_flush_task(struct work_struct *work)
c0ed1c14 SK	391	{
ca925cf1	392	struct netns_xfrm *xfrm = container_of(work, struct netns_xfrm,
233c96fc	393	flow_cache_flush_work);
ca925cf1	394	struct net *net = container_of(xfrm, struct net, xfrm);
c0ed1c14	395
ca925cf1 FD	396	flow_cache_flush(net);
ca925cf1 FD	397	}
c0ed1c14	398
ca925cf1	399	void flow_cache_flush_deferred(struct net *net)
c0ed1c14	400	{
ca925cf1	401	schedule_work(&net->xfrm.flow_cache_flush_work);
c0ed1c14 SK	402	}
c0ed1c14 SK	403
013dbb32	404	static int flow_cache_cpu_prepare(struct flow_cache *fc, int cpu)
1da177e4	405	{
83b6b1f5 ED	406	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, cpu);
83b6b1f5 ED	407	size_t sz = sizeof(struct hlist_head) * flow_cache_hash_size(fc);
d7997fe1	408
83b6b1f5 ED	409	if (!fcp->hash_table) {
	410	fcp->hash_table = kzalloc_node(sz, GFP_KERNEL, cpu_to_node(cpu));
	411	if (!fcp->hash_table) {
	412	pr_err("NET: failed to allocate flow cache sz %zu\n", sz);
	413	return -ENOMEM;
	414	}
	415	fcp->hash_rnd_recalc = 1;
	416	fcp->hash_count = 0;
	417	tasklet_init(&fcp->flush_tasklet, flow_cache_flush_tasklet, 0);
	418	}
	419	return 0;
1da177e4 LT	420	}
1da177e4 LT	421
013dbb32	422	static int flow_cache_cpu(struct notifier_block *nfb,
1da177e4 LT	423	unsigned long action,
	424	void *hcpu)
	425	{
ca925cf1 FD	426	struct flow_cache *fc = container_of(nfb, struct flow_cache,
ca925cf1 FD	427	hotcpu_notifier);
83b6b1f5	428	int res, cpu = (unsigned long) hcpu;
d7997fe1 TT	429	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, cpu);
d7997fe1 TT	430
83b6b1f5 ED	431	switch (action) {
	432	case CPU_UP_PREPARE:
	433	case CPU_UP_PREPARE_FROZEN:
	434	res = flow_cache_cpu_prepare(fc, cpu);
	435	if (res)
	436	return notifier_from_errno(res);
	437	break;
	438	case CPU_DEAD:
	439	case CPU_DEAD_FROZEN:
d7997fe1	440	__flow_cache_shrink(fc, fcp, 0);
83b6b1f5 ED	441	break;
83b6b1f5 ED	442	}
1da177e4 LT	443	return NOTIFY_OK;
1da177e4 LT	444	}
1da177e4	445
ca925cf1	446	int flow_cache_init(struct net *net)
1da177e4 LT	447	{
1da177e4 LT	448	int i;
ca925cf1 FD	449	struct flow_cache *fc = &net->xfrm.flow_cache_global;
ca925cf1 FD	450
d32d9bb8 ED	451	if (!flow_cachep)
	452	flow_cachep = kmem_cache_create("flow_cache",
	453	sizeof(struct flow_cache_entry),
	454	0, SLAB_PANIC, NULL);
ca925cf1 FD	455	spin_lock_init(&net->xfrm.flow_cache_gc_lock);
	456	INIT_LIST_HEAD(&net->xfrm.flow_cache_gc_list);
	457	INIT_WORK(&net->xfrm.flow_cache_gc_work, flow_cache_gc_task);
	458	INIT_WORK(&net->xfrm.flow_cache_flush_work, flow_cache_flush_task);
	459	mutex_init(&net->xfrm.flow_flush_sem);
6ad3122a	460	atomic_set(&net->xfrm.flow_cache_gc_count, 0);
1da177e4	461
d7997fe1 TT	462	fc->hash_shift = 10;
	463	fc->low_watermark = 2 * flow_cache_hash_size(fc);
	464	fc->high_watermark = 4 * flow_cache_hash_size(fc);
	465
d7997fe1	466	fc->percpu = alloc_percpu(struct flow_cache_percpu);
83b6b1f5 ED	467	if (!fc->percpu)
83b6b1f5 ED	468	return -ENOMEM;
1da177e4	469
e30a293e SB	470	cpu_notifier_register_begin();
e30a293e SB	471
83b6b1f5 ED	472	for_each_online_cpu(i) {
83b6b1f5 ED	473	if (flow_cache_cpu_prepare(fc, i))
6ccc3abd	474	goto err;
83b6b1f5	475	}
d7997fe1 TT	476	fc->hotcpu_notifier = (struct notifier_block){
	477	.notifier_call = flow_cache_cpu,
	478	};
e30a293e SB	479	__register_hotcpu_notifier(&fc->hotcpu_notifier);
	480
	481	cpu_notifier_register_done();
1da177e4	482
83b6b1f5 ED	483	setup_timer(&fc->rnd_timer, flow_cache_new_hashrnd,
	484	(unsigned long) fc);
	485	fc->rnd_timer.expires = jiffies + FLOW_HASH_RND_PERIOD;
	486	add_timer(&fc->rnd_timer);
	487
1da177e4	488	return 0;
6ccc3abd	489
	490	err:
	491	for_each_possible_cpu(i) {
	492	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, i);
	493	kfree(fcp->hash_table);
	494	fcp->hash_table = NULL;
	495	}
	496
e30a293e SB	497	cpu_notifier_register_done();
e30a293e SB	498
6ccc3abd	499	free_percpu(fc->percpu);
	500	fc->percpu = NULL;
	501
	502	return -ENOMEM;
1da177e4	503	}
ca925cf1	504	EXPORT_SYMBOL(flow_cache_init);
4a93f509 SK	505
	506	void flow_cache_fini(struct net *net)
	507	{
	508	int i;
	509	struct flow_cache *fc = &net->xfrm.flow_cache_global;
	510
	511	del_timer_sync(&fc->rnd_timer);
	512	unregister_hotcpu_notifier(&fc->hotcpu_notifier);
	513
	514	for_each_possible_cpu(i) {
	515	struct flow_cache_percpu *fcp = per_cpu_ptr(fc->percpu, i);
	516	kfree(fcp->hash_table);
	517	fcp->hash_table = NULL;
	518	}
	519
	520	free_percpu(fc->percpu);
	521	fc->percpu = NULL;
	522	}
	523	EXPORT_SYMBOL(flow_cache_fini);