netfilter.c

  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
 23int 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_dev(skb);
 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		.flowlabel = ip6_flowinfo(iph),
 40	};
 41	int err;
 42
 43	if (sk && sk->sk_bound_dev_if)
 44		fl6.flowi6_oif = sk->sk_bound_dev_if;
 45	else if (strict)
 46		fl6.flowi6_oif = dev->ifindex;
 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(net, skb, flowi6_to_flowi(&fl6), AF_INET6) == 0) {
 66		/* ignore return value from skb_dstref_steal, xfrm_lookup takes
 67		 * care of dropping the refcnt if needed.
 68		 */
 69		skb_dstref_steal(skb);
 70		dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
 71		if (IS_ERR(dst))
 72			return PTR_ERR(dst);
 73		skb_dst_set(skb, dst);
 74	}
 75#endif
 76
 77	/* Change in oif may mean change in hh_len. */
 78	hh_len = skb_dst_dev(skb)->hard_header_len;
 79	if (skb_headroom(skb) < hh_len &&
 80	    pskb_expand_head(skb, HH_DATA_ALIGN(hh_len - skb_headroom(skb)),
 81			     0, GFP_ATOMIC))
 82		return -ENOMEM;
 83
 84	return 0;
 85}
 86EXPORT_SYMBOL(ip6_route_me_harder);
 87
 88static int nf_ip6_reroute(struct sk_buff *skb,
 89			  const struct nf_queue_entry *entry)
 90{
 91	struct ip6_rt_info *rt_info = nf_queue_entry_reroute(entry);
 92
 93	if (entry->state.hook == NF_INET_LOCAL_OUT) {
 94		const struct ipv6hdr *iph = ipv6_hdr(skb);
 95		if (!ipv6_addr_equal(&iph->daddr, &rt_info->daddr) ||
 96		    !ipv6_addr_equal(&iph->saddr, &rt_info->saddr) ||
 97		    skb->mark != rt_info->mark)
 98			return ip6_route_me_harder(entry->state.net, entry->state.sk, skb);
 99	}
100	return 0;
101}
102
103int __nf_ip6_route(struct net *net, struct dst_entry **dst,
104		   struct flowi *fl, bool strict)
105{
106	static const struct ipv6_pinfo fake_pinfo;
107	static const struct inet_sock fake_sk = {
108		/* makes ip6_route_output set RT6_LOOKUP_F_IFACE: */
109		.sk.sk_bound_dev_if = 1,
110		.pinet6 = (struct ipv6_pinfo *) &fake_pinfo,
111	};
112	const void *sk = strict ? &fake_sk : NULL;
113	struct dst_entry *result;
114	int err;
115
116	result = ip6_route_output(net, sk, &fl->u.ip6);
117	err = result->error;
118	if (err)
119		dst_release(result);
120	else
121		*dst = result;
122	return err;
123}
124EXPORT_SYMBOL_GPL(__nf_ip6_route);
125
126int br_ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
127		    struct nf_bridge_frag_data *data,
128		    int (*output)(struct net *, struct sock *sk,
129				  const struct nf_bridge_frag_data *data,
130				  struct sk_buff *))
131{
132	int frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size;
133	u8 tstamp_type = skb->tstamp_type;
134	ktime_t tstamp = skb->tstamp;
135	struct ip6_frag_state state;
136	u8 *prevhdr, nexthdr = 0;
137	unsigned int mtu, hlen;
138	int hroom, err = 0;
139	__be32 frag_id;
140
141	err = ip6_find_1stfragopt(skb, &prevhdr);
142	if (err < 0)
143		goto blackhole;
144	hlen = err;
145	nexthdr = *prevhdr;
146
147	mtu = skb->dev->mtu;
148	if (frag_max_size > mtu ||
149	    frag_max_size < IPV6_MIN_MTU)
150		goto blackhole;
151
152	mtu = frag_max_size;
153	if (mtu < hlen + sizeof(struct frag_hdr) + 8)
154		goto blackhole;
155	mtu -= hlen + sizeof(struct frag_hdr);
156
157	frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
158				    &ipv6_hdr(skb)->saddr);
159
160	if (skb->ip_summed == CHECKSUM_PARTIAL &&
161	    (err = skb_checksum_help(skb)))
162		goto blackhole;
163
164	hroom = LL_RESERVED_SPACE(skb->dev);
165	if (skb_has_frag_list(skb)) {
166		unsigned int first_len = skb_pagelen(skb);
167		struct ip6_fraglist_iter iter;
168		struct sk_buff *frag2;
169
170		if (first_len - hlen > mtu)
171			goto blackhole;
172
173		if (skb_cloned(skb) ||
174		    skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
175			goto slow_path;
176
177		skb_walk_frags(skb, frag2) {
178			if (frag2->len > mtu)
179				goto blackhole;
180
181			/* Partially cloned skb? */
182			if (skb_shared(frag2) ||
183			    skb_headroom(frag2) < (hlen + hroom + sizeof(struct frag_hdr)))
184				goto slow_path;
185		}
186
187		err = ip6_fraglist_init(skb, hlen, prevhdr, nexthdr, frag_id,
188					&iter);
189		if (err < 0)
190			goto blackhole;
191
192		for (;;) {
193			/* Prepare header of the next frame,
194			 * before previous one went down.
195			 */
196			if (iter.frag)
197				ip6_fraglist_prepare(skb, &iter);
198
199			skb_set_delivery_time(skb, tstamp, tstamp_type);
200			err = output(net, sk, data, skb);
201			if (err || !iter.frag)
202				break;
203
204			skb = ip6_fraglist_next(&iter);
205		}
206
207		kfree(iter.tmp_hdr);
208		if (!err)
209			return 0;
210
211		kfree_skb_list(iter.frag);
212		return err;
213	}
214slow_path:
215	/* This is a linearized skbuff, the original geometry is lost for us.
216	 * This may also be a clone skbuff, we could preserve the geometry for
217	 * the copies but probably not worth the effort.
218	 */
219	ip6_frag_init(skb, hlen, mtu, skb->dev->needed_tailroom,
220		      LL_RESERVED_SPACE(skb->dev), prevhdr, nexthdr, frag_id,
221		      &state);
222
223	while (state.left > 0) {
224		struct sk_buff *skb2;
225
226		skb2 = ip6_frag_next(skb, &state);
227		if (IS_ERR(skb2)) {
228			err = PTR_ERR(skb2);
229			goto blackhole;
230		}
231
232		skb_set_delivery_time(skb2, tstamp, tstamp_type);
233		err = output(net, sk, data, skb2);
234		if (err)
235			goto blackhole;
236	}
237	consume_skb(skb);
238	return err;
239
240blackhole:
241	kfree_skb(skb);
242	return 0;
243}
244EXPORT_SYMBOL_GPL(br_ip6_fragment);
245
246static const struct nf_ipv6_ops ipv6ops = {
247#if IS_MODULE(CONFIG_IPV6)
248	.chk_addr		= ipv6_chk_addr,
249	.route_me_harder	= ip6_route_me_harder,
250	.dev_get_saddr		= ipv6_dev_get_saddr,
251	.route			= __nf_ip6_route,
252#if IS_ENABLED(CONFIG_SYN_COOKIES)
253	.cookie_init_sequence	= __cookie_v6_init_sequence,
254	.cookie_v6_check	= __cookie_v6_check,
255#endif
256#endif
257	.route_input		= ip6_route_input,
258	.fragment		= ip6_fragment,
259	.reroute		= nf_ip6_reroute,
260#if IS_MODULE(CONFIG_IPV6)
261	.br_fragment		= br_ip6_fragment,
262#endif
263};
264
265int __init ipv6_netfilter_init(void)
266{
267	RCU_INIT_POINTER(nf_ipv6_ops, &ipv6ops);
268	return 0;
269}
270
271/* This can be called from inet6_init() on errors, so it cannot
272 * be marked __exit. -DaveM
273 */
274void ipv6_netfilter_fini(void)
275{
276	RCU_INIT_POINTER(nf_ipv6_ops, NULL);
277}