inet_fragment.c 6.3 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 * inet fragments management
 *
 *		This program is free software; you can redistribute it and/or
 *		modify it under the terms of the GNU General Public License
 *		as published by the Free Software Foundation; either version
 *		2 of the License, or (at your option) any later version.
 *
 * 		Authors:	Pavel Emelyanov <xemul@openvz.org>
 *				Started as consolidation of ipv4/ip_fragment.c,
 *				ipv6/reassembly. and ipv6 nf conntrack reassembly
 */

#include <linux/list.h>
#include <linux/spinlock.h>
#include <linux/module.h>
#include <linux/timer.h>
#include <linux/mm.h>
19
#include <linux/random.h>
20 21
#include <linux/skbuff.h>
#include <linux/rtnetlink.h>
22 23 24

#include <net/inet_frag.h>

25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49
static void inet_frag_secret_rebuild(unsigned long dummy)
{
	struct inet_frags *f = (struct inet_frags *)dummy;
	unsigned long now = jiffies;
	int i;

	write_lock(&f->lock);
	get_random_bytes(&f->rnd, sizeof(u32));
	for (i = 0; i < INETFRAGS_HASHSZ; i++) {
		struct inet_frag_queue *q;
		struct hlist_node *p, *n;

		hlist_for_each_entry_safe(q, p, n, &f->hash[i], list) {
			unsigned int hval = f->hashfn(q);

			if (hval != i) {
				hlist_del(&q->list);

				/* Relink to new hash chain. */
				hlist_add_head(&q->list, &f->hash[hval]);
			}
		}
	}
	write_unlock(&f->lock);

50
	mod_timer(&f->secret_timer, now + f->secret_interval);
51 52
}

53 54 55 56 57 58 59 60 61 62 63 64
void inet_frags_init(struct inet_frags *f)
{
	int i;

	for (i = 0; i < INETFRAGS_HASHSZ; i++)
		INIT_HLIST_HEAD(&f->hash[i]);

	rwlock_init(&f->lock);

	f->rnd = (u32) ((num_physpages ^ (num_physpages>>7)) ^
				   (jiffies ^ (jiffies >> 6)));

65 66
	setup_timer(&f->secret_timer, inet_frag_secret_rebuild,
			(unsigned long)f);
67
	f->secret_timer.expires = jiffies + f->secret_interval;
68
	add_timer(&f->secret_timer);
69 70 71
}
EXPORT_SYMBOL(inet_frags_init);

72 73 74
void inet_frags_init_net(struct netns_frags *nf)
{
	nf->nqueues = 0;
75
	atomic_set(&nf->mem, 0);
76
	INIT_LIST_HEAD(&nf->lru_list);
77 78 79
}
EXPORT_SYMBOL(inet_frags_init_net);

80 81
void inet_frags_fini(struct inet_frags *f)
{
82
	del_timer(&f->secret_timer);
83 84
}
EXPORT_SYMBOL(inet_frags_fini);
85

86 87 88
void inet_frags_exit_net(struct netns_frags *nf, struct inet_frags *f)
{
	nf->low_thresh = 0;
89 90

	local_bh_disable();
91
	inet_frag_evictor(nf, f);
92
	local_bh_enable();
93 94 95
}
EXPORT_SYMBOL(inet_frags_exit_net);

96 97 98 99 100
static inline void fq_unlink(struct inet_frag_queue *fq, struct inet_frags *f)
{
	write_lock(&f->lock);
	hlist_del(&fq->list);
	list_del(&fq->lru_list);
101
	fq->net->nqueues--;
102 103 104 105 106 107 108 109
	write_unlock(&f->lock);
}

void inet_frag_kill(struct inet_frag_queue *fq, struct inet_frags *f)
{
	if (del_timer(&fq->timer))
		atomic_dec(&fq->refcnt);

110
	if (!(fq->last_in & INET_FRAG_COMPLETE)) {
111 112
		fq_unlink(fq, f);
		atomic_dec(&fq->refcnt);
113
		fq->last_in |= INET_FRAG_COMPLETE;
114 115 116 117
	}
}

EXPORT_SYMBOL(inet_frag_kill);
118

119 120
static inline void frag_kfree_skb(struct netns_frags *nf, struct inet_frags *f,
		struct sk_buff *skb, int *work)
121 122 123 124
{
	if (work)
		*work -= skb->truesize;

125
	atomic_sub(skb->truesize, &nf->mem);
126 127 128 129 130 131 132 133 134
	if (f->skb_free)
		f->skb_free(skb);
	kfree_skb(skb);
}

void inet_frag_destroy(struct inet_frag_queue *q, struct inet_frags *f,
					int *work)
{
	struct sk_buff *fp;
135
	struct netns_frags *nf;
136

137 138
	WARN_ON(!(q->last_in & INET_FRAG_COMPLETE));
	WARN_ON(del_timer(&q->timer) != 0);
139 140 141

	/* Release all fragment data. */
	fp = q->fragments;
142
	nf = q->net;
143 144 145
	while (fp) {
		struct sk_buff *xp = fp->next;

146
		frag_kfree_skb(nf, f, fp, work);
147 148 149 150 151
		fp = xp;
	}

	if (work)
		*work -= f->qsize;
152
	atomic_sub(f->qsize, &nf->mem);
153

154 155 156
	if (f->destructor)
		f->destructor(q);
	kfree(q);
157 158 159

}
EXPORT_SYMBOL(inet_frag_destroy);
160

161
int inet_frag_evictor(struct netns_frags *nf, struct inet_frags *f)
162 163 164 165
{
	struct inet_frag_queue *q;
	int work, evicted = 0;

166
	work = atomic_read(&nf->mem) - nf->low_thresh;
167 168
	while (work > 0) {
		read_lock(&f->lock);
169
		if (list_empty(&nf->lru_list)) {
170 171 172 173
			read_unlock(&f->lock);
			break;
		}

174
		q = list_first_entry(&nf->lru_list,
175 176 177 178 179
				struct inet_frag_queue, lru_list);
		atomic_inc(&q->refcnt);
		read_unlock(&f->lock);

		spin_lock(&q->lock);
180
		if (!(q->last_in & INET_FRAG_COMPLETE))
181 182 183 184 185 186 187 188 189 190 191
			inet_frag_kill(q, f);
		spin_unlock(&q->lock);

		if (atomic_dec_and_test(&q->refcnt))
			inet_frag_destroy(q, f, &work);
		evicted++;
	}

	return evicted;
}
EXPORT_SYMBOL(inet_frag_evictor);
192

193 194
static struct inet_frag_queue *inet_frag_intern(struct netns_frags *nf,
		struct inet_frag_queue *qp_in, struct inet_frags *f,
195
		void *arg)
196 197 198 199 200
{
	struct inet_frag_queue *qp;
#ifdef CONFIG_SMP
	struct hlist_node *n;
#endif
201
	unsigned int hash;
202 203

	write_lock(&f->lock);
204 205 206 207 208 209
	/*
	 * While we stayed w/o the lock other CPU could update
	 * the rnd seed, so we need to re-calculate the hash
	 * chain. Fortunatelly the qp_in can be used to get one.
	 */
	hash = f->hashfn(qp_in);
210 211 212 213 214 215
#ifdef CONFIG_SMP
	/* With SMP race we have to recheck hash table, because
	 * such entry could be created on other cpu, while we
	 * promoted read lock to write lock.
	 */
	hlist_for_each_entry(qp, n, &f->hash[hash], list) {
216
		if (qp->net == nf && f->match(qp, arg)) {
217 218
			atomic_inc(&qp->refcnt);
			write_unlock(&f->lock);
219
			qp_in->last_in |= INET_FRAG_COMPLETE;
220 221 222 223 224 225
			inet_frag_put(qp_in, f);
			return qp;
		}
	}
#endif
	qp = qp_in;
226
	if (!mod_timer(&qp->timer, jiffies + nf->timeout))
227 228 229 230
		atomic_inc(&qp->refcnt);

	atomic_inc(&qp->refcnt);
	hlist_add_head(&qp->list, &f->hash[hash]);
231
	list_add_tail(&qp->lru_list, &nf->lru_list);
232
	nf->nqueues++;
233 234 235
	write_unlock(&f->lock);
	return qp;
}
236

237 238
static struct inet_frag_queue *inet_frag_alloc(struct netns_frags *nf,
		struct inet_frags *f, void *arg)
239 240 241 242 243 244 245
{
	struct inet_frag_queue *q;

	q = kzalloc(f->qsize, GFP_ATOMIC);
	if (q == NULL)
		return NULL;

246
	f->constructor(q, arg);
247
	atomic_add(f->qsize, &nf->mem);
248 249 250
	setup_timer(&q->timer, f->frag_expire, (unsigned long)q);
	spin_lock_init(&q->lock);
	atomic_set(&q->refcnt, 1);
251
	q->net = nf;
252 253 254

	return q;
}
255

256
static struct inet_frag_queue *inet_frag_create(struct netns_frags *nf,
257
		struct inet_frags *f, void *arg)
258 259 260
{
	struct inet_frag_queue *q;

261
	q = inet_frag_alloc(nf, f, arg);
262 263 264
	if (q == NULL)
		return NULL;

265
	return inet_frag_intern(nf, q, f, arg);
266
}
267

268 269
struct inet_frag_queue *inet_frag_find(struct netns_frags *nf,
		struct inet_frags *f, void *key, unsigned int hash)
270 271 272 273 274
{
	struct inet_frag_queue *q;
	struct hlist_node *n;

	hlist_for_each_entry(q, n, &f->hash[hash], list) {
275
		if (q->net == nf && f->match(q, key)) {
276 277 278 279 280 281 282
			atomic_inc(&q->refcnt);
			read_unlock(&f->lock);
			return q;
		}
	}
	read_unlock(&f->lock);

283
	return inet_frag_create(nf, f, key);
284 285
}
EXPORT_SYMBOL(inet_frag_find);