xref: /openbmc/linux/net/ipv6/netfilter.c (revision 2cf1c348)
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 flow_keys flkeys;
28 	unsigned int hh_len;
29 	struct dst_entry *dst;
30 	int strict = (ipv6_addr_type(&iph->daddr) &
31 		      (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
32 	struct flowi6 fl6 = {
33 		.flowi6_oif = sk && sk->sk_bound_dev_if ? sk->sk_bound_dev_if :
34 			strict ? skb_dst(skb)->dev->ifindex : 0,
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 	fib6_rules_early_flow_dissect(net, skb, &fl6, &flkeys);
43 	dst = ip6_route_output(net, sk, &fl6);
44 	err = dst->error;
45 	if (err) {
46 		IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
47 		net_dbg_ratelimited("ip6_route_me_harder: No more route\n");
48 		dst_release(dst);
49 		return err;
50 	}
51 
52 	/* Drop old route. */
53 	skb_dst_drop(skb);
54 
55 	skb_dst_set(skb, dst);
56 
57 #ifdef CONFIG_XFRM
58 	if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
59 	    xfrm_decode_session(skb, flowi6_to_flowi(&fl6), AF_INET6) == 0) {
60 		skb_dst_set(skb, NULL);
61 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
62 		if (IS_ERR(dst))
63 			return PTR_ERR(dst);
64 		skb_dst_set(skb, dst);
65 	}
66 #endif
67 
68 	/* Change in oif may mean change in hh_len. */
69 	hh_len = skb_dst(skb)->dev->hard_header_len;
70 	if (skb_headroom(skb) < hh_len &&
71 	    pskb_expand_head(skb, HH_DATA_ALIGN(hh_len - skb_headroom(skb)),
72 			     0, GFP_ATOMIC))
73 		return -ENOMEM;
74 
75 	return 0;
76 }
77 EXPORT_SYMBOL(ip6_route_me_harder);
78 
79 static int nf_ip6_reroute(struct sk_buff *skb,
80 			  const struct nf_queue_entry *entry)
81 {
82 	struct ip6_rt_info *rt_info = nf_queue_entry_reroute(entry);
83 
84 	if (entry->state.hook == NF_INET_LOCAL_OUT) {
85 		const struct ipv6hdr *iph = ipv6_hdr(skb);
86 		if (!ipv6_addr_equal(&iph->daddr, &rt_info->daddr) ||
87 		    !ipv6_addr_equal(&iph->saddr, &rt_info->saddr) ||
88 		    skb->mark != rt_info->mark)
89 			return ip6_route_me_harder(entry->state.net, entry->state.sk, skb);
90 	}
91 	return 0;
92 }
93 
94 int __nf_ip6_route(struct net *net, struct dst_entry **dst,
95 		   struct flowi *fl, bool strict)
96 {
97 	static const struct ipv6_pinfo fake_pinfo;
98 	static const struct inet_sock fake_sk = {
99 		/* makes ip6_route_output set RT6_LOOKUP_F_IFACE: */
100 		.sk.sk_bound_dev_if = 1,
101 		.pinet6 = (struct ipv6_pinfo *) &fake_pinfo,
102 	};
103 	const void *sk = strict ? &fake_sk : NULL;
104 	struct dst_entry *result;
105 	int err;
106 
107 	result = ip6_route_output(net, sk, &fl->u.ip6);
108 	err = result->error;
109 	if (err)
110 		dst_release(result);
111 	else
112 		*dst = result;
113 	return err;
114 }
115 EXPORT_SYMBOL_GPL(__nf_ip6_route);
116 
117 int br_ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
118 		    struct nf_bridge_frag_data *data,
119 		    int (*output)(struct net *, struct sock *sk,
120 				  const struct nf_bridge_frag_data *data,
121 				  struct sk_buff *))
122 {
123 	int frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size;
124 	ktime_t tstamp = skb->tstamp;
125 	struct ip6_frag_state state;
126 	u8 *prevhdr, nexthdr = 0;
127 	unsigned int mtu, hlen;
128 	int hroom, err = 0;
129 	__be32 frag_id;
130 
131 	err = ip6_find_1stfragopt(skb, &prevhdr);
132 	if (err < 0)
133 		goto blackhole;
134 	hlen = err;
135 	nexthdr = *prevhdr;
136 
137 	mtu = skb->dev->mtu;
138 	if (frag_max_size > mtu ||
139 	    frag_max_size < IPV6_MIN_MTU)
140 		goto blackhole;
141 
142 	mtu = frag_max_size;
143 	if (mtu < hlen + sizeof(struct frag_hdr) + 8)
144 		goto blackhole;
145 	mtu -= hlen + sizeof(struct frag_hdr);
146 
147 	frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
148 				    &ipv6_hdr(skb)->saddr);
149 
150 	if (skb->ip_summed == CHECKSUM_PARTIAL &&
151 	    (err = skb_checksum_help(skb)))
152 		goto blackhole;
153 
154 	hroom = LL_RESERVED_SPACE(skb->dev);
155 	if (skb_has_frag_list(skb)) {
156 		unsigned int first_len = skb_pagelen(skb);
157 		struct ip6_fraglist_iter iter;
158 		struct sk_buff *frag2;
159 
160 		if (first_len - hlen > mtu ||
161 		    skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
162 			goto blackhole;
163 
164 		if (skb_cloned(skb))
165 			goto slow_path;
166 
167 		skb_walk_frags(skb, frag2) {
168 			if (frag2->len > mtu ||
169 			    skb_headroom(frag2) < (hlen + hroom + sizeof(struct frag_hdr)))
170 				goto blackhole;
171 
172 			/* Partially cloned skb? */
173 			if (skb_shared(frag2))
174 				goto slow_path;
175 		}
176 
177 		err = ip6_fraglist_init(skb, hlen, prevhdr, nexthdr, frag_id,
178 					&iter);
179 		if (err < 0)
180 			goto blackhole;
181 
182 		for (;;) {
183 			/* Prepare header of the next frame,
184 			 * before previous one went down.
185 			 */
186 			if (iter.frag)
187 				ip6_fraglist_prepare(skb, &iter);
188 
189 			skb->tstamp = tstamp;
190 			err = output(net, sk, data, skb);
191 			if (err || !iter.frag)
192 				break;
193 
194 			skb = ip6_fraglist_next(&iter);
195 		}
196 
197 		kfree(iter.tmp_hdr);
198 		if (!err)
199 			return 0;
200 
201 		kfree_skb_list(iter.frag);
202 		return err;
203 	}
204 slow_path:
205 	/* This is a linearized skbuff, the original geometry is lost for us.
206 	 * This may also be a clone skbuff, we could preserve the geometry for
207 	 * the copies but probably not worth the effort.
208 	 */
209 	ip6_frag_init(skb, hlen, mtu, skb->dev->needed_tailroom,
210 		      LL_RESERVED_SPACE(skb->dev), prevhdr, nexthdr, frag_id,
211 		      &state);
212 
213 	while (state.left > 0) {
214 		struct sk_buff *skb2;
215 
216 		skb2 = ip6_frag_next(skb, &state);
217 		if (IS_ERR(skb2)) {
218 			err = PTR_ERR(skb2);
219 			goto blackhole;
220 		}
221 
222 		skb2->tstamp = tstamp;
223 		err = output(net, sk, data, skb2);
224 		if (err)
225 			goto blackhole;
226 	}
227 	consume_skb(skb);
228 	return err;
229 
230 blackhole:
231 	kfree_skb(skb);
232 	return 0;
233 }
234 EXPORT_SYMBOL_GPL(br_ip6_fragment);
235 
236 static const struct nf_ipv6_ops ipv6ops = {
237 #if IS_MODULE(CONFIG_IPV6)
238 	.chk_addr		= ipv6_chk_addr,
239 	.route_me_harder	= ip6_route_me_harder,
240 	.dev_get_saddr		= ipv6_dev_get_saddr,
241 	.route			= __nf_ip6_route,
242 #if IS_ENABLED(CONFIG_SYN_COOKIES)
243 	.cookie_init_sequence	= __cookie_v6_init_sequence,
244 	.cookie_v6_check	= __cookie_v6_check,
245 #endif
246 #endif
247 	.route_input		= ip6_route_input,
248 	.fragment		= ip6_fragment,
249 	.reroute		= nf_ip6_reroute,
250 #if IS_MODULE(CONFIG_IPV6)
251 	.br_fragment		= br_ip6_fragment,
252 #endif
253 };
254 
255 int __init ipv6_netfilter_init(void)
256 {
257 	RCU_INIT_POINTER(nf_ipv6_ops, &ipv6ops);
258 	return 0;
259 }
260 
261 /* This can be called from inet6_init() on errors, so it cannot
262  * be marked __exit. -DaveM
263  */
264 void ipv6_netfilter_fini(void)
265 {
266 	RCU_INIT_POINTER(nf_ipv6_ops, NULL);
267 }
268