xref: /linux/net/ipv6/netfilter.c (revision f69e98a91a01fd7c5755dd710e94a17d6e9f583f)
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 sock *sk_partial, struct sk_buff *skb)
24 {
25 	const struct ipv6hdr *iph = ipv6_hdr(skb);
26 	struct sock *sk = sk_to_full_sk(sk_partial);
27 	struct net_device *dev = skb_dst(skb)->dev;
28 	struct flow_keys flkeys;
29 	unsigned int hh_len;
30 	struct dst_entry *dst;
31 	int strict = (ipv6_addr_type(&iph->daddr) &
32 		      (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
33 	struct flowi6 fl6 = {
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 	if (sk && sk->sk_bound_dev_if)
42 		fl6.flowi6_oif = sk->sk_bound_dev_if;
43 	else if (strict)
44 		fl6.flowi6_oif = dev->ifindex;
45 	else
46 		fl6.flowi6_oif = l3mdev_master_ifindex(dev);
47 
48 	fib6_rules_early_flow_dissect(net, skb, &fl6, &flkeys);
49 	dst = ip6_route_output(net, sk, &fl6);
50 	err = dst->error;
51 	if (err) {
52 		IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
53 		net_dbg_ratelimited("ip6_route_me_harder: No more route\n");
54 		dst_release(dst);
55 		return err;
56 	}
57 
58 	/* Drop old route. */
59 	skb_dst_drop(skb);
60 
61 	skb_dst_set(skb, dst);
62 
63 #ifdef CONFIG_XFRM
64 	if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
65 	    xfrm_decode_session(skb, flowi6_to_flowi(&fl6), AF_INET6) == 0) {
66 		skb_dst_set(skb, NULL);
67 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
68 		if (IS_ERR(dst))
69 			return PTR_ERR(dst);
70 		skb_dst_set(skb, dst);
71 	}
72 #endif
73 
74 	/* Change in oif may mean change in hh_len. */
75 	hh_len = skb_dst(skb)->dev->hard_header_len;
76 	if (skb_headroom(skb) < hh_len &&
77 	    pskb_expand_head(skb, HH_DATA_ALIGN(hh_len - skb_headroom(skb)),
78 			     0, GFP_ATOMIC))
79 		return -ENOMEM;
80 
81 	return 0;
82 }
83 EXPORT_SYMBOL(ip6_route_me_harder);
84 
85 static int nf_ip6_reroute(struct sk_buff *skb,
86 			  const struct nf_queue_entry *entry)
87 {
88 	struct ip6_rt_info *rt_info = nf_queue_entry_reroute(entry);
89 
90 	if (entry->state.hook == NF_INET_LOCAL_OUT) {
91 		const struct ipv6hdr *iph = ipv6_hdr(skb);
92 		if (!ipv6_addr_equal(&iph->daddr, &rt_info->daddr) ||
93 		    !ipv6_addr_equal(&iph->saddr, &rt_info->saddr) ||
94 		    skb->mark != rt_info->mark)
95 			return ip6_route_me_harder(entry->state.net, entry->state.sk, skb);
96 	}
97 	return 0;
98 }
99 
100 int __nf_ip6_route(struct net *net, struct dst_entry **dst,
101 		   struct flowi *fl, bool strict)
102 {
103 	static const struct ipv6_pinfo fake_pinfo;
104 	static const struct inet_sock fake_sk = {
105 		/* makes ip6_route_output set RT6_LOOKUP_F_IFACE: */
106 		.sk.sk_bound_dev_if = 1,
107 		.pinet6 = (struct ipv6_pinfo *) &fake_pinfo,
108 	};
109 	const void *sk = strict ? &fake_sk : NULL;
110 	struct dst_entry *result;
111 	int err;
112 
113 	result = ip6_route_output(net, sk, &fl->u.ip6);
114 	err = result->error;
115 	if (err)
116 		dst_release(result);
117 	else
118 		*dst = result;
119 	return err;
120 }
121 EXPORT_SYMBOL_GPL(__nf_ip6_route);
122 
123 int br_ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
124 		    struct nf_bridge_frag_data *data,
125 		    int (*output)(struct net *, struct sock *sk,
126 				  const struct nf_bridge_frag_data *data,
127 				  struct sk_buff *))
128 {
129 	int frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size;
130 	bool mono_delivery_time = skb->mono_delivery_time;
131 	ktime_t tstamp = skb->tstamp;
132 	struct ip6_frag_state state;
133 	u8 *prevhdr, nexthdr = 0;
134 	unsigned int mtu, hlen;
135 	int hroom, err = 0;
136 	__be32 frag_id;
137 
138 	err = ip6_find_1stfragopt(skb, &prevhdr);
139 	if (err < 0)
140 		goto blackhole;
141 	hlen = err;
142 	nexthdr = *prevhdr;
143 
144 	mtu = skb->dev->mtu;
145 	if (frag_max_size > mtu ||
146 	    frag_max_size < IPV6_MIN_MTU)
147 		goto blackhole;
148 
149 	mtu = frag_max_size;
150 	if (mtu < hlen + sizeof(struct frag_hdr) + 8)
151 		goto blackhole;
152 	mtu -= hlen + sizeof(struct frag_hdr);
153 
154 	frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
155 				    &ipv6_hdr(skb)->saddr);
156 
157 	if (skb->ip_summed == CHECKSUM_PARTIAL &&
158 	    (err = skb_checksum_help(skb)))
159 		goto blackhole;
160 
161 	hroom = LL_RESERVED_SPACE(skb->dev);
162 	if (skb_has_frag_list(skb)) {
163 		unsigned int first_len = skb_pagelen(skb);
164 		struct ip6_fraglist_iter iter;
165 		struct sk_buff *frag2;
166 
167 		if (first_len - hlen > mtu ||
168 		    skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
169 			goto blackhole;
170 
171 		if (skb_cloned(skb))
172 			goto slow_path;
173 
174 		skb_walk_frags(skb, frag2) {
175 			if (frag2->len > mtu ||
176 			    skb_headroom(frag2) < (hlen + hroom + sizeof(struct frag_hdr)))
177 				goto blackhole;
178 
179 			/* Partially cloned skb? */
180 			if (skb_shared(frag2))
181 				goto slow_path;
182 		}
183 
184 		err = ip6_fraglist_init(skb, hlen, prevhdr, nexthdr, frag_id,
185 					&iter);
186 		if (err < 0)
187 			goto blackhole;
188 
189 		for (;;) {
190 			/* Prepare header of the next frame,
191 			 * before previous one went down.
192 			 */
193 			if (iter.frag)
194 				ip6_fraglist_prepare(skb, &iter);
195 
196 			skb_set_delivery_time(skb, tstamp, mono_delivery_time);
197 			err = output(net, sk, data, skb);
198 			if (err || !iter.frag)
199 				break;
200 
201 			skb = ip6_fraglist_next(&iter);
202 		}
203 
204 		kfree(iter.tmp_hdr);
205 		if (!err)
206 			return 0;
207 
208 		kfree_skb_list(iter.frag);
209 		return err;
210 	}
211 slow_path:
212 	/* This is a linearized skbuff, the original geometry is lost for us.
213 	 * This may also be a clone skbuff, we could preserve the geometry for
214 	 * the copies but probably not worth the effort.
215 	 */
216 	ip6_frag_init(skb, hlen, mtu, skb->dev->needed_tailroom,
217 		      LL_RESERVED_SPACE(skb->dev), prevhdr, nexthdr, frag_id,
218 		      &state);
219 
220 	while (state.left > 0) {
221 		struct sk_buff *skb2;
222 
223 		skb2 = ip6_frag_next(skb, &state);
224 		if (IS_ERR(skb2)) {
225 			err = PTR_ERR(skb2);
226 			goto blackhole;
227 		}
228 
229 		skb_set_delivery_time(skb2, tstamp, mono_delivery_time);
230 		err = output(net, sk, data, skb2);
231 		if (err)
232 			goto blackhole;
233 	}
234 	consume_skb(skb);
235 	return err;
236 
237 blackhole:
238 	kfree_skb(skb);
239 	return 0;
240 }
241 EXPORT_SYMBOL_GPL(br_ip6_fragment);
242 
243 static const struct nf_ipv6_ops ipv6ops = {
244 #if IS_MODULE(CONFIG_IPV6)
245 	.chk_addr		= ipv6_chk_addr,
246 	.route_me_harder	= ip6_route_me_harder,
247 	.dev_get_saddr		= ipv6_dev_get_saddr,
248 	.route			= __nf_ip6_route,
249 #if IS_ENABLED(CONFIG_SYN_COOKIES)
250 	.cookie_init_sequence	= __cookie_v6_init_sequence,
251 	.cookie_v6_check	= __cookie_v6_check,
252 #endif
253 #endif
254 	.route_input		= ip6_route_input,
255 	.fragment		= ip6_fragment,
256 	.reroute		= nf_ip6_reroute,
257 #if IS_MODULE(CONFIG_IPV6)
258 	.br_fragment		= br_ip6_fragment,
259 #endif
260 };
261 
262 int __init ipv6_netfilter_init(void)
263 {
264 	RCU_INIT_POINTER(nf_ipv6_ops, &ipv6ops);
265 	return 0;
266 }
267 
268 /* This can be called from inet6_init() on errors, so it cannot
269  * be marked __exit. -DaveM
270  */
271 void ipv6_netfilter_fini(void)
272 {
273 	RCU_INIT_POINTER(nf_ipv6_ops, NULL);
274 }
275