1/*
  2 * Copyright (c) 2008-2009 Patrick McHardy <kaber@trash.net>
  3 *
  4 * This program is free software; you can redistribute it and/or modify
  5 * it under the terms of the GNU General Public License version 2 as
  6 * published by the Free Software Foundation.
  7 *
  8 * Development of this code funded by Astaro AG (http://www.astaro.com/)
  9 */
 10
 11#include <linux/kernel.h>
 12#include <linux/init.h>
 13#include <linux/module.h>
 14#include <linux/list.h>
 15#include <linux/rbtree.h>
 16#include <linux/netlink.h>
 17#include <linux/netfilter.h>
 18#include <linux/netfilter/nf_tables.h>
 19#include <net/netfilter/nf_tables.h>
 20
 21static DEFINE_SPINLOCK(nft_rbtree_lock);
 22
 23struct nft_rbtree {
 24	struct rb_root		root;
 25};
 26
 27struct nft_rbtree_elem {
 28	struct rb_node		node;
 29	struct nft_set_ext	ext;
 30};
 31
 32
 33static bool nft_rbtree_lookup(const struct nft_set *set, const u32 *key,
 34			      const struct nft_set_ext **ext)
 35{
 36	const struct nft_rbtree *priv = nft_set_priv(set);
 37	const struct nft_rbtree_elem *rbe, *interval = NULL;
 38	const struct rb_node *parent;
 39	u8 genmask = nft_genmask_cur(read_pnet(&set->pnet));
 40	int d;
 41
 42	spin_lock_bh(&nft_rbtree_lock);
 43	parent = priv->root.rb_node;
 44	while (parent != NULL) {
 45		rbe = rb_entry(parent, struct nft_rbtree_elem, node);
 46
 47		d = memcmp(nft_set_ext_key(&rbe->ext), key, set->klen);
 48		if (d < 0) {
 49			parent = parent->rb_left;
 50			interval = rbe;
 51		} else if (d > 0)
 52			parent = parent->rb_right;
 53		else {
 54found:
 55			if (!nft_set_elem_active(&rbe->ext, genmask)) {
 56				parent = parent->rb_left;
 57				continue;
 58			}
 59			if (nft_set_ext_exists(&rbe->ext, NFT_SET_EXT_FLAGS) &&
 60			    *nft_set_ext_flags(&rbe->ext) &
 61			    NFT_SET_ELEM_INTERVAL_END)
 62				goto out;
 63			spin_unlock_bh(&nft_rbtree_lock);
 64
 65			*ext = &rbe->ext;
 66			return true;
 67		}
 68	}
 69
 70	if (set->flags & NFT_SET_INTERVAL && interval != NULL) {
 71		rbe = interval;
 72		goto found;
 73	}
 74out:
 75	spin_unlock_bh(&nft_rbtree_lock);
 76	return false;
 77}
 78
 79static int __nft_rbtree_insert(const struct nft_set *set,
 80			       struct nft_rbtree_elem *new)
 81{
 82	struct nft_rbtree *priv = nft_set_priv(set);
 83	struct nft_rbtree_elem *rbe;
 84	struct rb_node *parent, **p;
 85	u8 genmask = nft_genmask_next(read_pnet(&set->pnet));
 86	int d;
 87
 88	parent = NULL;
 89	p = &priv->root.rb_node;
 90	while (*p != NULL) {
 91		parent = *p;
 92		rbe = rb_entry(parent, struct nft_rbtree_elem, node);
 93		d = memcmp(nft_set_ext_key(&rbe->ext),
 94			   nft_set_ext_key(&new->ext),
 95			   set->klen);
 96		if (d < 0)
 97			p = &parent->rb_left;
 98		else if (d > 0)
 99			p = &parent->rb_right;
100		else {
101			if (nft_set_elem_active(&rbe->ext, genmask))
102				return -EEXIST;
103			p = &parent->rb_left;
104		}
105	}
106	rb_link_node(&new->node, parent, p);
107	rb_insert_color(&new->node, &priv->root);
108	return 0;
109}
110
111static int nft_rbtree_insert(const struct nft_set *set,
112			     const struct nft_set_elem *elem)
113{
114	struct nft_rbtree_elem *rbe = elem->priv;
115	int err;
116
117	spin_lock_bh(&nft_rbtree_lock);
118	err = __nft_rbtree_insert(set, rbe);
119	spin_unlock_bh(&nft_rbtree_lock);
120
121	return err;
122}
123
124static void nft_rbtree_remove(const struct nft_set *set,
125			      const struct nft_set_elem *elem)
126{
127	struct nft_rbtree *priv = nft_set_priv(set);
128	struct nft_rbtree_elem *rbe = elem->priv;
129
130	spin_lock_bh(&nft_rbtree_lock);
131	rb_erase(&rbe->node, &priv->root);
132	spin_unlock_bh(&nft_rbtree_lock);
133}
134
135static void nft_rbtree_activate(const struct nft_set *set,
136				const struct nft_set_elem *elem)
137{
138	struct nft_rbtree_elem *rbe = elem->priv;
139
140	nft_set_elem_change_active(set, &rbe->ext);
141}
142
143static void *nft_rbtree_deactivate(const struct nft_set *set,
144				   const struct nft_set_elem *elem)
145{
146	const struct nft_rbtree *priv = nft_set_priv(set);
147	const struct rb_node *parent = priv->root.rb_node;
148	struct nft_rbtree_elem *rbe;
149	u8 genmask = nft_genmask_cur(read_pnet(&set->pnet));
150	int d;
151
152	while (parent != NULL) {
153		rbe = rb_entry(parent, struct nft_rbtree_elem, node);
154
155		d = memcmp(nft_set_ext_key(&rbe->ext), &elem->key.val,
156					   set->klen);
157		if (d < 0)
158			parent = parent->rb_left;
159		else if (d > 0)
160			parent = parent->rb_right;
161		else {
162			if (!nft_set_elem_active(&rbe->ext, genmask)) {
163				parent = parent->rb_left;
164				continue;
165			}
166			nft_set_elem_change_active(set, &rbe->ext);
167			return rbe;
168		}
169	}
170	return NULL;
171}
172
173static void nft_rbtree_walk(const struct nft_ctx *ctx,
174			    const struct nft_set *set,
175			    struct nft_set_iter *iter)
176{
177	const struct nft_rbtree *priv = nft_set_priv(set);
178	struct nft_rbtree_elem *rbe;
179	struct nft_set_elem elem;
180	struct rb_node *node;
181	u8 genmask = nft_genmask_cur(read_pnet(&set->pnet));
182
183	spin_lock_bh(&nft_rbtree_lock);
184	for (node = rb_first(&priv->root); node != NULL; node = rb_next(node)) {
185		rbe = rb_entry(node, struct nft_rbtree_elem, node);
186
187		if (iter->count < iter->skip)
188			goto cont;
189		if (!nft_set_elem_active(&rbe->ext, genmask))
190			goto cont;
191
192		elem.priv = rbe;
193
194		iter->err = iter->fn(ctx, set, iter, &elem);
195		if (iter->err < 0) {
196			spin_unlock_bh(&nft_rbtree_lock);
197			return;
198		}
199cont:
200		iter->count++;
201	}
202	spin_unlock_bh(&nft_rbtree_lock);
203}
204
205static unsigned int nft_rbtree_privsize(const struct nlattr * const nla[])
206{
207	return sizeof(struct nft_rbtree);
208}
209
210static int nft_rbtree_init(const struct nft_set *set,
211			   const struct nft_set_desc *desc,
212			   const struct nlattr * const nla[])
213{
214	struct nft_rbtree *priv = nft_set_priv(set);
215
216	priv->root = RB_ROOT;
217	return 0;
218}
219
220static void nft_rbtree_destroy(const struct nft_set *set)
221{
222	struct nft_rbtree *priv = nft_set_priv(set);
223	struct nft_rbtree_elem *rbe;
224	struct rb_node *node;
225
226	while ((node = priv->root.rb_node) != NULL) {
227		rb_erase(node, &priv->root);
228		rbe = rb_entry(node, struct nft_rbtree_elem, node);
229		nft_set_elem_destroy(set, rbe);
230	}
231}
232
233static bool nft_rbtree_estimate(const struct nft_set_desc *desc, u32 features,
234				struct nft_set_estimate *est)
235{
236	unsigned int nsize;
237
238	nsize = sizeof(struct nft_rbtree_elem);
239	if (desc->size)
240		est->size = sizeof(struct nft_rbtree) + desc->size * nsize;
241	else
242		est->size = nsize;
243
244	est->class = NFT_SET_CLASS_O_LOG_N;
245
246	return true;
247}
248
249static struct nft_set_ops nft_rbtree_ops __read_mostly = {
250	.privsize	= nft_rbtree_privsize,
251	.elemsize	= offsetof(struct nft_rbtree_elem, ext),
252	.estimate	= nft_rbtree_estimate,
253	.init		= nft_rbtree_init,
254	.destroy	= nft_rbtree_destroy,
255	.insert		= nft_rbtree_insert,
256	.remove		= nft_rbtree_remove,
257	.deactivate	= nft_rbtree_deactivate,
258	.activate	= nft_rbtree_activate,
259	.lookup		= nft_rbtree_lookup,
260	.walk		= nft_rbtree_walk,
261	.features	= NFT_SET_INTERVAL | NFT_SET_MAP,
262	.owner		= THIS_MODULE,
263};
264
265static int __init nft_rbtree_module_init(void)
266{
267	return nft_register_set(&nft_rbtree_ops);
268}
269
270static void __exit nft_rbtree_module_exit(void)
271{
272	nft_unregister_set(&nft_rbtree_ops);
273}
274
275module_init(nft_rbtree_module_init);
276module_exit(nft_rbtree_module_exit);
277
278MODULE_LICENSE("GPL");
279MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
280MODULE_ALIAS_NFT_SET();