2 * inet fragments management
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
9 * Authors: Pavel Emelyanov <xemul@openvz.org>
10 * Started as consolidation of ipv4/ip_fragment.c,
11 * ipv6/reassembly. and ipv6 nf conntrack reassembly
14 #include <linux/list.h>
15 #include <linux/spinlock.h>
16 #include <linux/module.h>
17 #include <linux/timer.h>
19 #include <linux/random.h>
20 #include <linux/skbuff.h>
21 #include <linux/rtnetlink.h>
22 #include <linux/slab.h>
25 #include <net/inet_frag.h>
26 #include <net/inet_ecn.h>
28 /* Given the OR values of all fragments, apply RFC 3168 5.3 requirements
29 * Value : 0xff if frame should be dropped.
30 * 0 or INET_ECN_CE value, to be ORed in to final iph->tos field
32 const u8 ip_frag_ecn_table[16] = {
33 /* at least one fragment had CE, and others ECT_0 or ECT_1 */
34 [IPFRAG_ECN_CE | IPFRAG_ECN_ECT_0] = INET_ECN_CE,
35 [IPFRAG_ECN_CE | IPFRAG_ECN_ECT_1] = INET_ECN_CE,
36 [IPFRAG_ECN_CE | IPFRAG_ECN_ECT_0 | IPFRAG_ECN_ECT_1] = INET_ECN_CE,
38 /* invalid combinations : drop frame */
39 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_CE] = 0xff,
40 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_ECT_0] = 0xff,
41 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_ECT_1] = 0xff,
42 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_ECT_0 | IPFRAG_ECN_ECT_1] = 0xff,
43 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_CE | IPFRAG_ECN_ECT_0] = 0xff,
44 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_CE | IPFRAG_ECN_ECT_1] = 0xff,
45 [IPFRAG_ECN_NOT_ECT | IPFRAG_ECN_CE | IPFRAG_ECN_ECT_0 | IPFRAG_ECN_ECT_1] = 0xff,
47 EXPORT_SYMBOL(ip_frag_ecn_table);
49 int inet_frags_init(struct inet_frags *f)
51 f->frags_cachep = kmem_cache_create(f->frags_cache_name, f->qsize, 0, 0,
58 EXPORT_SYMBOL(inet_frags_init);
60 void inet_frags_fini(struct inet_frags *f)
62 /* We must wait that all inet_frag_destroy_rcu() have completed. */
65 kmem_cache_destroy(f->frags_cachep);
66 f->frags_cachep = NULL;
68 EXPORT_SYMBOL(inet_frags_fini);
70 static void inet_frags_free_cb(void *ptr, void *arg)
72 struct inet_frag_queue *fq = ptr;
74 /* If we can not cancel the timer, it means this frag_queue
75 * is already disappearing, we have nothing to do.
76 * Otherwise, we own a refcount until the end of this function.
78 if (!del_timer(&fq->timer))
81 spin_lock_bh(&fq->lock);
82 if (!(fq->flags & INET_FRAG_COMPLETE)) {
83 fq->flags |= INET_FRAG_COMPLETE;
84 atomic_dec(&fq->refcnt);
86 spin_unlock_bh(&fq->lock);
91 void inet_frags_exit_net(struct netns_frags *nf)
93 nf->high_thresh = 0; /* prevent creation of new frags */
95 rhashtable_free_and_destroy(&nf->rhashtable, inet_frags_free_cb, NULL);
97 EXPORT_SYMBOL(inet_frags_exit_net);
99 void inet_frag_kill(struct inet_frag_queue *fq)
101 if (del_timer(&fq->timer))
102 atomic_dec(&fq->refcnt);
104 if (!(fq->flags & INET_FRAG_COMPLETE)) {
105 struct netns_frags *nf = fq->net;
107 fq->flags |= INET_FRAG_COMPLETE;
108 rhashtable_remove_fast(&nf->rhashtable, &fq->node, nf->f->rhash_params);
109 atomic_dec(&fq->refcnt);
112 EXPORT_SYMBOL(inet_frag_kill);
114 static inline void frag_kfree_skb(struct netns_frags *nf, struct inet_frags *f,
122 static void inet_frag_destroy_rcu(struct rcu_head *head)
124 struct inet_frag_queue *q = container_of(head, struct inet_frag_queue,
126 struct inet_frags *f = q->net->f;
130 kmem_cache_free(f->frags_cachep, q);
133 void inet_frag_destroy(struct inet_frag_queue *q)
136 struct netns_frags *nf;
137 unsigned int sum, sum_truesize = 0;
138 struct inet_frags *f;
140 WARN_ON(!(q->flags & INET_FRAG_COMPLETE));
141 WARN_ON(del_timer(&q->timer) != 0);
143 /* Release all fragment data. */
149 struct sk_buff *xp = fp->next;
151 sum_truesize += fp->truesize;
152 frag_kfree_skb(nf, f, fp);
156 sum_truesize = inet_frag_rbtree_purge(&q->rb_fragments);
158 sum = sum_truesize + f->qsize;
160 call_rcu(&q->rcu, inet_frag_destroy_rcu);
162 sub_frag_mem_limit(nf, sum);
164 EXPORT_SYMBOL(inet_frag_destroy);
166 static struct inet_frag_queue *inet_frag_alloc(struct netns_frags *nf,
167 struct inet_frags *f,
170 struct inet_frag_queue *q;
172 if (!nf->high_thresh || frag_mem_limit(nf) > nf->high_thresh)
175 q = kmem_cache_zalloc(f->frags_cachep, GFP_ATOMIC);
180 f->constructor(q, arg);
181 add_frag_mem_limit(nf, f->qsize);
183 setup_timer(&q->timer, f->frag_expire, (unsigned long)q);
184 spin_lock_init(&q->lock);
185 atomic_set(&q->refcnt, 3);
190 static struct inet_frag_queue *inet_frag_create(struct netns_frags *nf,
192 struct inet_frag_queue **prev)
194 struct inet_frags *f = nf->f;
195 struct inet_frag_queue *q;
197 q = inet_frag_alloc(nf, f, arg);
199 *prev = ERR_PTR(-ENOMEM);
202 mod_timer(&q->timer, jiffies + nf->timeout);
204 *prev = rhashtable_lookup_get_insert_key(&nf->rhashtable, &q->key,
205 &q->node, f->rhash_params);
207 q->flags |= INET_FRAG_COMPLETE;
209 inet_frag_destroy(q);
214 EXPORT_SYMBOL(inet_frag_create);
216 /* TODO : call from rcu_read_lock() and no longer use refcount_inc_not_zero() */
217 struct inet_frag_queue *inet_frag_find(struct netns_frags *nf, void *key)
219 struct inet_frag_queue *fq = NULL, *prev;
222 prev = rhashtable_lookup(&nf->rhashtable, key, nf->f->rhash_params);
224 fq = inet_frag_create(nf, key, &prev);
225 if (prev && !IS_ERR(prev)) {
227 if (!atomic_inc_not_zero(&fq->refcnt))
233 EXPORT_SYMBOL(inet_frag_find);