xref: /linux/net/ipv6/netfilter.c (revision 0889d07f3e4b171c453b2aaf2b257f9074cdf624)
1 /*
2  * IPv6 specific functions of netfilter core
3  *
4  * Rusty Russell (C) 2000 -- This code is GPL.
5  * Patrick McHardy (C) 2006-2012
6  */
7 #include <linux/kernel.h>
8 #include <linux/init.h>
9 #include <linux/ipv6.h>
10 #include <linux/netfilter.h>
11 #include <linux/netfilter_ipv6.h>
12 #include <linux/export.h>
13 #include <net/addrconf.h>
14 #include <net/dst.h>
15 #include <net/ipv6.h>
16 #include <net/ip6_route.h>
17 #include <net/xfrm.h>
18 #include <net/netfilter/nf_queue.h>
19 #include <net/netfilter/nf_conntrack_bridge.h>
20 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
21 #include "../bridge/br_private.h"
22 
23 int ip6_route_me_harder(struct net *net, struct sk_buff *skb)
24 {
25 	const struct ipv6hdr *iph = ipv6_hdr(skb);
26 	struct sock *sk = sk_to_full_sk(skb->sk);
27 	unsigned int hh_len;
28 	struct dst_entry *dst;
29 	int strict = (ipv6_addr_type(&iph->daddr) &
30 		      (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
31 	struct flowi6 fl6 = {
32 		.flowi6_oif = sk && sk->sk_bound_dev_if ? sk->sk_bound_dev_if :
33 			strict ? skb_dst(skb)->dev->ifindex : 0,
34 		.flowi6_mark = skb->mark,
35 		.flowi6_uid = sock_net_uid(net, sk),
36 		.daddr = iph->daddr,
37 		.saddr = iph->saddr,
38 	};
39 	int err;
40 
41 	dst = ip6_route_output(net, sk, &fl6);
42 	err = dst->error;
43 	if (err) {
44 		IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
45 		net_dbg_ratelimited("ip6_route_me_harder: No more route\n");
46 		dst_release(dst);
47 		return err;
48 	}
49 
50 	/* Drop old route. */
51 	skb_dst_drop(skb);
52 
53 	skb_dst_set(skb, dst);
54 
55 #ifdef CONFIG_XFRM
56 	if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
57 	    xfrm_decode_session(skb, flowi6_to_flowi(&fl6), AF_INET6) == 0) {
58 		skb_dst_set(skb, NULL);
59 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
60 		if (IS_ERR(dst))
61 			return PTR_ERR(dst);
62 		skb_dst_set(skb, dst);
63 	}
64 #endif
65 
66 	/* Change in oif may mean change in hh_len. */
67 	hh_len = skb_dst(skb)->dev->hard_header_len;
68 	if (skb_headroom(skb) < hh_len &&
69 	    pskb_expand_head(skb, HH_DATA_ALIGN(hh_len - skb_headroom(skb)),
70 			     0, GFP_ATOMIC))
71 		return -ENOMEM;
72 
73 	return 0;
74 }
75 EXPORT_SYMBOL(ip6_route_me_harder);
76 
77 static int nf_ip6_reroute(struct sk_buff *skb,
78 			  const struct nf_queue_entry *entry)
79 {
80 	struct ip6_rt_info *rt_info = nf_queue_entry_reroute(entry);
81 
82 	if (entry->state.hook == NF_INET_LOCAL_OUT) {
83 		const struct ipv6hdr *iph = ipv6_hdr(skb);
84 		if (!ipv6_addr_equal(&iph->daddr, &rt_info->daddr) ||
85 		    !ipv6_addr_equal(&iph->saddr, &rt_info->saddr) ||
86 		    skb->mark != rt_info->mark)
87 			return ip6_route_me_harder(entry->state.net, skb);
88 	}
89 	return 0;
90 }
91 
92 int __nf_ip6_route(struct net *net, struct dst_entry **dst,
93 		   struct flowi *fl, bool strict)
94 {
95 	static const struct ipv6_pinfo fake_pinfo;
96 	static const struct inet_sock fake_sk = {
97 		/* makes ip6_route_output set RT6_LOOKUP_F_IFACE: */
98 		.sk.sk_bound_dev_if = 1,
99 		.pinet6 = (struct ipv6_pinfo *) &fake_pinfo,
100 	};
101 	const void *sk = strict ? &fake_sk : NULL;
102 	struct dst_entry *result;
103 	int err;
104 
105 	result = ip6_route_output(net, sk, &fl->u.ip6);
106 	err = result->error;
107 	if (err)
108 		dst_release(result);
109 	else
110 		*dst = result;
111 	return err;
112 }
113 EXPORT_SYMBOL_GPL(__nf_ip6_route);
114 
115 int br_ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
116 		    struct nf_bridge_frag_data *data,
117 		    int (*output)(struct net *, struct sock *sk,
118 				  const struct nf_bridge_frag_data *data,
119 				  struct sk_buff *))
120 {
121 	int frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size;
122 	struct ip6_frag_state state;
123 	u8 *prevhdr, nexthdr = 0;
124 	unsigned int mtu, hlen;
125 	int hroom, err = 0;
126 	__be32 frag_id;
127 
128 	err = ip6_find_1stfragopt(skb, &prevhdr);
129 	if (err < 0)
130 		goto blackhole;
131 	hlen = err;
132 	nexthdr = *prevhdr;
133 
134 	mtu = skb->dev->mtu;
135 	if (frag_max_size > mtu ||
136 	    frag_max_size < IPV6_MIN_MTU)
137 		goto blackhole;
138 
139 	mtu = frag_max_size;
140 	if (mtu < hlen + sizeof(struct frag_hdr) + 8)
141 		goto blackhole;
142 	mtu -= hlen + sizeof(struct frag_hdr);
143 
144 	frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
145 				    &ipv6_hdr(skb)->saddr);
146 
147 	if (skb->ip_summed == CHECKSUM_PARTIAL &&
148 	    (err = skb_checksum_help(skb)))
149 		goto blackhole;
150 
151 	hroom = LL_RESERVED_SPACE(skb->dev);
152 	if (skb_has_frag_list(skb)) {
153 		unsigned int first_len = skb_pagelen(skb);
154 		struct ip6_fraglist_iter iter;
155 		struct sk_buff *frag2;
156 
157 		if (first_len - hlen > mtu ||
158 		    skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
159 			goto blackhole;
160 
161 		if (skb_cloned(skb))
162 			goto slow_path;
163 
164 		skb_walk_frags(skb, frag2) {
165 			if (frag2->len > mtu ||
166 			    skb_headroom(frag2) < (hlen + hroom + sizeof(struct frag_hdr)))
167 				goto blackhole;
168 
169 			/* Partially cloned skb? */
170 			if (skb_shared(frag2))
171 				goto slow_path;
172 		}
173 
174 		err = ip6_fraglist_init(skb, hlen, prevhdr, nexthdr, frag_id,
175 					&iter);
176 		if (err < 0)
177 			goto blackhole;
178 
179 		for (;;) {
180 			/* Prepare header of the next frame,
181 			 * before previous one went down.
182 			 */
183 			if (iter.frag)
184 				ip6_fraglist_prepare(skb, &iter);
185 
186 			err = output(net, sk, data, skb);
187 			if (err || !iter.frag)
188 				break;
189 
190 			skb = ip6_fraglist_next(&iter);
191 		}
192 
193 		kfree(iter.tmp_hdr);
194 		if (!err)
195 			return 0;
196 
197 		kfree_skb_list(iter.frag);
198 		return err;
199 	}
200 slow_path:
201 	/* This is a linearized skbuff, the original geometry is lost for us.
202 	 * This may also be a clone skbuff, we could preserve the geometry for
203 	 * the copies but probably not worth the effort.
204 	 */
205 	ip6_frag_init(skb, hlen, mtu, skb->dev->needed_tailroom,
206 		      LL_RESERVED_SPACE(skb->dev), prevhdr, nexthdr, frag_id,
207 		      &state);
208 
209 	while (state.left > 0) {
210 		struct sk_buff *skb2;
211 
212 		skb2 = ip6_frag_next(skb, &state);
213 		if (IS_ERR(skb2)) {
214 			err = PTR_ERR(skb2);
215 			goto blackhole;
216 		}
217 
218 		err = output(net, sk, data, skb2);
219 		if (err)
220 			goto blackhole;
221 	}
222 	consume_skb(skb);
223 	return err;
224 
225 blackhole:
226 	kfree_skb(skb);
227 	return 0;
228 }
229 EXPORT_SYMBOL_GPL(br_ip6_fragment);
230 
231 static const struct nf_ipv6_ops ipv6ops = {
232 #if IS_MODULE(CONFIG_IPV6)
233 	.chk_addr		= ipv6_chk_addr,
234 	.route_me_harder	= ip6_route_me_harder,
235 	.dev_get_saddr		= ipv6_dev_get_saddr,
236 	.route			= __nf_ip6_route,
237 #if IS_ENABLED(CONFIG_SYN_COOKIES)
238 	.cookie_init_sequence	= __cookie_v6_init_sequence,
239 	.cookie_v6_check	= __cookie_v6_check,
240 #endif
241 #endif
242 	.route_input		= ip6_route_input,
243 	.fragment		= ip6_fragment,
244 	.reroute		= nf_ip6_reroute,
245 #if IS_MODULE(CONFIG_IPV6) && IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
246 	.br_defrag		= nf_ct_frag6_gather,
247 #endif
248 #if IS_MODULE(CONFIG_IPV6)
249 	.br_fragment		= br_ip6_fragment,
250 #endif
251 };
252 
253 int __init ipv6_netfilter_init(void)
254 {
255 	RCU_INIT_POINTER(nf_ipv6_ops, &ipv6ops);
256 	return 0;
257 }
258 
259 /* This can be called from inet6_init() on errors, so it cannot
260  * be marked __exit. -DaveM
261  */
262 void ipv6_netfilter_fini(void)
263 {
264 	RCU_INIT_POINTER(nf_ipv6_ops, NULL);
265 }
266