inet_fragment.c 6.4 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
#include <linux/slab.h>
23 24 25

#include <net/inet_frag.h>

26 27 28 29 30 31 32 33 34 35
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;
36
		struct hlist_node *n;
37

38
		hlist_for_each_entry_safe(q, n, &f->hash[i], list) {
39 40 41 42 43 44 45 46 47 48 49 50
			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);

51
	mod_timer(&f->secret_timer, now + f->secret_interval);
52 53
}

54 55 56 57 58 59 60 61 62 63 64 65
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)));

66 67
	setup_timer(&f->secret_timer, inet_frag_secret_rebuild,
			(unsigned long)f);
68
	f->secret_timer.expires = jiffies + f->secret_interval;
69
	add_timer(&f->secret_timer);
70 71 72
}
EXPORT_SYMBOL(inet_frags_init);

73 74 75
void inet_frags_init_net(struct netns_frags *nf)
{
	nf->nqueues = 0;
76
	init_frag_mem_limit(nf);
77
	INIT_LIST_HEAD(&nf->lru_list);
78
	spin_lock_init(&nf->lru_lock);
79 80 81
}
EXPORT_SYMBOL(inet_frags_init_net);

82 83
void inet_frags_fini(struct inet_frags *f)
{
84
	del_timer(&f->secret_timer);
85 86
}
EXPORT_SYMBOL(inet_frags_fini);
87

88 89 90
void inet_frags_exit_net(struct netns_frags *nf, struct inet_frags *f)
{
	nf->low_thresh = 0;
91 92

	local_bh_disable();
93
	inet_frag_evictor(nf, f, true);
94
	local_bh_enable();
95 96

	percpu_counter_destroy(&nf->mem);
97 98 99
}
EXPORT_SYMBOL(inet_frags_exit_net);

100 101 102 103
static inline void fq_unlink(struct inet_frag_queue *fq, struct inet_frags *f)
{
	write_lock(&f->lock);
	hlist_del(&fq->list);
104
	fq->net->nqueues--;
105
	write_unlock(&f->lock);
106
	inet_frag_lru_del(fq);
107 108 109 110 111 112 113
}

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

114
	if (!(fq->last_in & INET_FRAG_COMPLETE)) {
115 116
		fq_unlink(fq, f);
		atomic_dec(&fq->refcnt);
117
		fq->last_in |= INET_FRAG_COMPLETE;
118 119 120
	}
}
EXPORT_SYMBOL(inet_frag_kill);
121

122
static inline void frag_kfree_skb(struct netns_frags *nf, struct inet_frags *f,
123
		struct sk_buff *skb)
124 125 126 127 128 129 130 131 132 133
{
	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;
134
	struct netns_frags *nf;
135
	unsigned int sum, sum_truesize = 0;
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;

146 147
		sum_truesize += fp->truesize;
		frag_kfree_skb(nf, f, fp);
148 149
		fp = xp;
	}
150
	sum = sum_truesize + f->qsize;
151
	if (work)
152 153
		*work -= sum;
	sub_frag_mem_limit(q, sum);
154

155 156 157
	if (f->destructor)
		f->destructor(q);
	kfree(q);
158 159 160

}
EXPORT_SYMBOL(inet_frag_destroy);
161

162
int inet_frag_evictor(struct netns_frags *nf, struct inet_frags *f, bool force)
163 164 165 166
{
	struct inet_frag_queue *q;
	int work, evicted = 0;

167
	if (!force) {
168
		if (frag_mem_limit(nf) <= nf->high_thresh)
169 170 171
			return 0;
	}

172
	work = frag_mem_limit(nf) - nf->low_thresh;
173
	while (work > 0) {
174 175
		spin_lock(&nf->lru_lock);

176
		if (list_empty(&nf->lru_list)) {
177
			spin_unlock(&nf->lru_lock);
178 179 180
			break;
		}

181
		q = list_first_entry(&nf->lru_list,
182 183
				struct inet_frag_queue, lru_list);
		atomic_inc(&q->refcnt);
184
		spin_unlock(&nf->lru_lock);
185 186

		spin_lock(&q->lock);
187
		if (!(q->last_in & INET_FRAG_COMPLETE))
188 189 190 191 192 193 194 195 196 197 198
			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);
199

200 201
static struct inet_frag_queue *inet_frag_intern(struct netns_frags *nf,
		struct inet_frag_queue *qp_in, struct inet_frags *f,
202
		void *arg)
203 204 205 206
{
	struct inet_frag_queue *qp;
#ifdef CONFIG_SMP
#endif
207
	unsigned int hash;
208 209

	write_lock(&f->lock);
210 211 212 213 214 215
	/*
	 * 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);
216 217 218 219 220
#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.
	 */
221
	hlist_for_each_entry(qp, &f->hash[hash], list) {
222
		if (qp->net == nf && f->match(qp, arg)) {
223 224
			atomic_inc(&qp->refcnt);
			write_unlock(&f->lock);
225
			qp_in->last_in |= INET_FRAG_COMPLETE;
226 227 228 229 230 231
			inet_frag_put(qp_in, f);
			return qp;
		}
	}
#endif
	qp = qp_in;
232
	if (!mod_timer(&qp->timer, jiffies + nf->timeout))
233 234 235 236
		atomic_inc(&qp->refcnt);

	atomic_inc(&qp->refcnt);
	hlist_add_head(&qp->list, &f->hash[hash]);
237
	nf->nqueues++;
238
	write_unlock(&f->lock);
239
	inet_frag_lru_add(nf, qp);
240 241
	return qp;
}
242

243 244
static struct inet_frag_queue *inet_frag_alloc(struct netns_frags *nf,
		struct inet_frags *f, void *arg)
245 246 247 248 249 250 251
{
	struct inet_frag_queue *q;

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

252
	q->net = nf;
253
	f->constructor(q, arg);
254 255
	add_frag_mem_limit(q, f->qsize);

256 257 258 259 260 261
	setup_timer(&q->timer, f->frag_expire, (unsigned long)q);
	spin_lock_init(&q->lock);
	atomic_set(&q->refcnt, 1);

	return q;
}
262

263
static struct inet_frag_queue *inet_frag_create(struct netns_frags *nf,
264
		struct inet_frags *f, void *arg)
265 266 267
{
	struct inet_frag_queue *q;

268
	q = inet_frag_alloc(nf, f, arg);
269 270 271
	if (q == NULL)
		return NULL;

272
	return inet_frag_intern(nf, q, f, arg);
273
}
274

275 276
struct inet_frag_queue *inet_frag_find(struct netns_frags *nf,
		struct inet_frags *f, void *key, unsigned int hash)
277
	__releases(&f->lock)
278 279 280
{
	struct inet_frag_queue *q;

281
	hlist_for_each_entry(q, &f->hash[hash], list) {
282
		if (q->net == nf && f->match(q, key)) {
283 284 285 286 287 288 289
			atomic_inc(&q->refcnt);
			read_unlock(&f->lock);
			return q;
		}
	}
	read_unlock(&f->lock);

290
	return inet_frag_create(nf, f, key);
291 292
}
EXPORT_SYMBOL(inet_frag_find);