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 
ip6_route_me_harder(struct net * net,struct sock * sk_partial,struct sk_buff * skb)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_l3mdev = l3mdev_master_ifindex(dev),
35 		.flowi6_mark = skb->mark,
36 		.flowi6_uid = sock_net_uid(net, sk),
37 		.daddr = iph->daddr,
38 		.saddr = iph->saddr,
39 	};
40 	int err;
41 
42 	if (sk && sk->sk_bound_dev_if)
43 		fl6.flowi6_oif = sk->sk_bound_dev_if;
44 	else if (strict)
45 		fl6.flowi6_oif = dev->ifindex;
46 
47 	fib6_rules_early_flow_dissect(net, skb, &fl6, &flkeys);
48 	dst = ip6_route_output(net, sk, &fl6);
49 	err = dst->error;
50 	if (err) {
51 		IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
52 		net_dbg_ratelimited("ip6_route_me_harder: No more route\n");
53 		dst_release(dst);
54 		return err;
55 	}
56 
57 	/* Drop old route. */
58 	skb_dst_drop(skb);
59 
60 	skb_dst_set(skb, dst);
61 
62 #ifdef CONFIG_XFRM
63 	if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
64 	    xfrm_decode_session(skb, flowi6_to_flowi(&fl6), AF_INET6) == 0) {
65 		skb_dst_set(skb, NULL);
66 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
67 		if (IS_ERR(dst))
68 			return PTR_ERR(dst);
69 		skb_dst_set(skb, dst);
70 	}
71 #endif
72 
73 	/* Change in oif may mean change in hh_len. */
74 	hh_len = skb_dst(skb)->dev->hard_header_len;
75 	if (skb_headroom(skb) < hh_len &&
76 	    pskb_expand_head(skb, HH_DATA_ALIGN(hh_len - skb_headroom(skb)),
77 			     0, GFP_ATOMIC))
78 		return -ENOMEM;
79 
80 	return 0;
81 }
82 EXPORT_SYMBOL(ip6_route_me_harder);
83 
nf_ip6_reroute(struct sk_buff * skb,const struct nf_queue_entry * entry)84 static int nf_ip6_reroute(struct sk_buff *skb,
85 			  const struct nf_queue_entry *entry)
86 {
87 	struct ip6_rt_info *rt_info = nf_queue_entry_reroute(entry);
88 
89 	if (entry->state.hook == NF_INET_LOCAL_OUT) {
90 		const struct ipv6hdr *iph = ipv6_hdr(skb);
91 		if (!ipv6_addr_equal(&iph->daddr, &rt_info->daddr) ||
92 		    !ipv6_addr_equal(&iph->saddr, &rt_info->saddr) ||
93 		    skb->mark != rt_info->mark)
94 			return ip6_route_me_harder(entry->state.net, entry->state.sk, skb);
95 	}
96 	return 0;
97 }
98 
__nf_ip6_route(struct net * net,struct dst_entry ** dst,struct flowi * fl,bool strict)99 int __nf_ip6_route(struct net *net, struct dst_entry **dst,
100 		   struct flowi *fl, bool strict)
101 {
102 	static const struct ipv6_pinfo fake_pinfo;
103 	static const struct inet_sock fake_sk = {
104 		/* makes ip6_route_output set RT6_LOOKUP_F_IFACE: */
105 		.sk.sk_bound_dev_if = 1,
106 		.pinet6 = (struct ipv6_pinfo *) &fake_pinfo,
107 	};
108 	const void *sk = strict ? &fake_sk : NULL;
109 	struct dst_entry *result;
110 	int err;
111 
112 	result = ip6_route_output(net, sk, &fl->u.ip6);
113 	err = result->error;
114 	if (err)
115 		dst_release(result);
116 	else
117 		*dst = result;
118 	return err;
119 }
120 EXPORT_SYMBOL_GPL(__nf_ip6_route);
121 
br_ip6_fragment(struct net * net,struct sock * sk,struct sk_buff * skb,struct nf_bridge_frag_data * data,int (* output)(struct net *,struct sock * sk,const struct nf_bridge_frag_data * data,struct sk_buff *))122 int br_ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
123 		    struct nf_bridge_frag_data *data,
124 		    int (*output)(struct net *, struct sock *sk,
125 				  const struct nf_bridge_frag_data *data,
126 				  struct sk_buff *))
127 {
128 	int frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size;
129 	bool mono_delivery_time = skb->mono_delivery_time;
130 	ktime_t tstamp = skb->tstamp;
131 	struct ip6_frag_state state;
132 	u8 *prevhdr, nexthdr = 0;
133 	unsigned int mtu, hlen;
134 	int hroom, err = 0;
135 	__be32 frag_id;
136 
137 	err = ip6_find_1stfragopt(skb, &prevhdr);
138 	if (err < 0)
139 		goto blackhole;
140 	hlen = err;
141 	nexthdr = *prevhdr;
142 
143 	mtu = skb->dev->mtu;
144 	if (frag_max_size > mtu ||
145 	    frag_max_size < IPV6_MIN_MTU)
146 		goto blackhole;
147 
148 	mtu = frag_max_size;
149 	if (mtu < hlen + sizeof(struct frag_hdr) + 8)
150 		goto blackhole;
151 	mtu -= hlen + sizeof(struct frag_hdr);
152 
153 	frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
154 				    &ipv6_hdr(skb)->saddr);
155 
156 	if (skb->ip_summed == CHECKSUM_PARTIAL &&
157 	    (err = skb_checksum_help(skb)))
158 		goto blackhole;
159 
160 	hroom = LL_RESERVED_SPACE(skb->dev);
161 	if (skb_has_frag_list(skb)) {
162 		unsigned int first_len = skb_pagelen(skb);
163 		struct ip6_fraglist_iter iter;
164 		struct sk_buff *frag2;
165 
166 		if (first_len - hlen > mtu ||
167 		    skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
168 			goto blackhole;
169 
170 		if (skb_cloned(skb))
171 			goto slow_path;
172 
173 		skb_walk_frags(skb, frag2) {
174 			if (frag2->len > mtu ||
175 			    skb_headroom(frag2) < (hlen + hroom + sizeof(struct frag_hdr)))
176 				goto blackhole;
177 
178 			/* Partially cloned skb? */
179 			if (skb_shared(frag2))
180 				goto slow_path;
181 		}
182 
183 		err = ip6_fraglist_init(skb, hlen, prevhdr, nexthdr, frag_id,
184 					&iter);
185 		if (err < 0)
186 			goto blackhole;
187 
188 		for (;;) {
189 			/* Prepare header of the next frame,
190 			 * before previous one went down.
191 			 */
192 			if (iter.frag)
193 				ip6_fraglist_prepare(skb, &iter);
194 
195 			skb_set_delivery_time(skb, tstamp, mono_delivery_time);
196 			err = output(net, sk, data, skb);
197 			if (err || !iter.frag)
198 				break;
199 
200 			skb = ip6_fraglist_next(&iter);
201 		}
202 
203 		kfree(iter.tmp_hdr);
204 		if (!err)
205 			return 0;
206 
207 		kfree_skb_list(iter.frag);
208 		return err;
209 	}
210 slow_path:
211 	/* This is a linearized skbuff, the original geometry is lost for us.
212 	 * This may also be a clone skbuff, we could preserve the geometry for
213 	 * the copies but probably not worth the effort.
214 	 */
215 	ip6_frag_init(skb, hlen, mtu, skb->dev->needed_tailroom,
216 		      LL_RESERVED_SPACE(skb->dev), prevhdr, nexthdr, frag_id,
217 		      &state);
218 
219 	while (state.left > 0) {
220 		struct sk_buff *skb2;
221 
222 		skb2 = ip6_frag_next(skb, &state);
223 		if (IS_ERR(skb2)) {
224 			err = PTR_ERR(skb2);
225 			goto blackhole;
226 		}
227 
228 		skb_set_delivery_time(skb2, tstamp, mono_delivery_time);
229 		err = output(net, sk, data, skb2);
230 		if (err)
231 			goto blackhole;
232 	}
233 	consume_skb(skb);
234 	return err;
235 
236 blackhole:
237 	kfree_skb(skb);
238 	return 0;
239 }
240 EXPORT_SYMBOL_GPL(br_ip6_fragment);
241 
242 static const struct nf_ipv6_ops ipv6ops = {
243 #if IS_MODULE(CONFIG_IPV6)
244 	.chk_addr		= ipv6_chk_addr,
245 	.route_me_harder	= ip6_route_me_harder,
246 	.dev_get_saddr		= ipv6_dev_get_saddr,
247 	.route			= __nf_ip6_route,
248 #if IS_ENABLED(CONFIG_SYN_COOKIES)
249 	.cookie_init_sequence	= __cookie_v6_init_sequence,
250 	.cookie_v6_check	= __cookie_v6_check,
251 #endif
252 #endif
253 	.route_input		= ip6_route_input,
254 	.fragment		= ip6_fragment,
255 	.reroute		= nf_ip6_reroute,
256 #if IS_MODULE(CONFIG_IPV6)
257 	.br_fragment		= br_ip6_fragment,
258 #endif
259 };
260 
ipv6_netfilter_init(void)261 int __init ipv6_netfilter_init(void)
262 {
263 	RCU_INIT_POINTER(nf_ipv6_ops, &ipv6ops);
264 	return 0;
265 }
266 
267 /* This can be called from inet6_init() on errors, so it cannot
268  * be marked __exit. -DaveM
269  */
ipv6_netfilter_fini(void)270 void ipv6_netfilter_fini(void)
271 {
272 	RCU_INIT_POINTER(nf_ipv6_ops, NULL);
273 }
274