netfilter: Dissect flow after packet mangling
[linux-block.git] / net / ipv6 / netfilter.c
CommitLineData
f229f6ce
PM
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 */
020b4c12 7#include <linux/kernel.h>
bb94aa16 8#include <linux/init.h>
020b4c12 9#include <linux/ipv6.h>
2cc7d573
HW
10#include <linux/netfilter.h>
11#include <linux/netfilter_ipv6.h>
bc3b2d7f 12#include <linux/export.h>
2a7851bf 13#include <net/addrconf.h>
020b4c12
HW
14#include <net/dst.h>
15#include <net/ipv6.h>
16#include <net/ip6_route.h>
3e3850e9 17#include <net/xfrm.h>
c01cd429 18#include <net/netfilter/nf_queue.h>
764dd163
PNA
19#include <net/netfilter/nf_conntrack_bridge.h>
20#include <net/netfilter/ipv6/nf_defrag_ipv6.h>
21#include "../bridge/br_private.h"
020b4c12 22
46d6c5ae 23int ip6_route_me_harder(struct net *net, struct sock *sk_partial, struct sk_buff *skb)
020b4c12 24{
b71d1d42 25 const struct ipv6hdr *iph = ipv6_hdr(skb);
46d6c5ae 26 struct sock *sk = sk_to_full_sk(sk_partial);
812fa71f 27 struct flow_keys flkeys;
0ad352cb 28 unsigned int hh_len;
020b4c12 29 struct dst_entry *dst;
15df03c6
EC
30 int strict = (ipv6_addr_type(&iph->daddr) &
31 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
4c9483b2 32 struct flowi6 fl6 = {
508b0904 33 .flowi6_oif = sk && sk->sk_bound_dev_if ? sk->sk_bound_dev_if :
15df03c6 34 strict ? skb_dst(skb)->dev->ifindex : 0,
4c9483b2 35 .flowi6_mark = skb->mark,
7d98386d 36 .flowi6_uid = sock_net_uid(net, sk),
4c9483b2
DM
37 .daddr = iph->daddr,
38 .saddr = iph->saddr,
020b4c12 39 };
a8951d58 40 int err;
020b4c12 41
812fa71f 42 fib6_rules_early_flow_dissect(net, skb, &fl6, &flkeys);
7d98386d 43 dst = ip6_route_output(net, sk, &fl6);
a8951d58
SP
44 err = dst->error;
45 if (err) {
eef9d90d 46 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
ba7a46f1 47 net_dbg_ratelimited("ip6_route_me_harder: No more route\n");
020b4c12 48 dst_release(dst);
a8951d58 49 return err;
020b4c12
HW
50 }
51
52 /* Drop old route. */
adf30907 53 skb_dst_drop(skb);
020b4c12 54
adf30907 55 skb_dst_set(skb, dst);
90348e0e
UW
56
57#ifdef CONFIG_XFRM
58 if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
4c9483b2 59 xfrm_decode_session(skb, flowi6_to_flowi(&fl6), AF_INET6) == 0) {
90348e0e 60 skb_dst_set(skb, NULL);
7d98386d 61 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
452edd59 62 if (IS_ERR(dst))
58e35d14 63 return PTR_ERR(dst);
90348e0e
UW
64 skb_dst_set(skb, dst);
65 }
66#endif
67
0ad352cb
PM
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))
58e35d14 73 return -ENOMEM;
0ad352cb 74
020b4c12
HW
75 return 0;
76}
77EXPORT_SYMBOL(ip6_route_me_harder);
78
ce388f45 79static int nf_ip6_reroute(struct sk_buff *skb,
02f014d8 80 const struct nf_queue_entry *entry)
2cc7d573 81{
02f014d8 82 struct ip6_rt_info *rt_info = nf_queue_entry_reroute(entry);
2cc7d573 83
1d1de89b 84 if (entry->state.hook == NF_INET_LOCAL_OUT) {
b71d1d42 85 const struct ipv6hdr *iph = ipv6_hdr(skb);
2cc7d573 86 if (!ipv6_addr_equal(&iph->daddr, &rt_info->daddr) ||
9f40ac71
EL
87 !ipv6_addr_equal(&iph->saddr, &rt_info->saddr) ||
88 skb->mark != rt_info->mark)
46d6c5ae 89 return ip6_route_me_harder(entry->state.net, entry->state.sk, skb);
2cc7d573
HW
90 }
91 return 0;
92}
93
ac02bcf9
FW
94int __nf_ip6_route(struct net *net, struct dst_entry **dst,
95 struct flowi *fl, bool strict)
1841a4c7 96{
0fae2e77
FW
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;
2dad81ad
FW
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;
1841a4c7 114}
ac02bcf9 115EXPORT_SYMBOL_GPL(__nf_ip6_route);
1841a4c7 116
764dd163 117int br_ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
46705b07 118 struct nf_bridge_frag_data *data,
764dd163 119 int (*output)(struct net *, struct sock *sk,
46705b07 120 const struct nf_bridge_frag_data *data,
764dd163
PNA
121 struct sk_buff *))
122{
123 int frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size;
9669fffc 124 ktime_t tstamp = skb->tstamp;
764dd163
PNA
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
9669fffc 189 skb->tstamp = tstamp;
764dd163
PNA
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
b7034146 201 kfree_skb_list(iter.frag);
764dd163
PNA
202 return err;
203 }
204slow_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
9669fffc 222 skb2->tstamp = tstamp;
764dd163
PNA
223 err = output(net, sk, data, skb2);
224 if (err)
225 goto blackhole;
226 }
227 consume_skb(skb);
228 return err;
229
230blackhole:
231 kfree_skb(skb);
232 return 0;
233}
234EXPORT_SYMBOL_GPL(br_ip6_fragment);
235
2a7851bf 236static const struct nf_ipv6_ops ipv6ops = {
96058728 237#if IS_MODULE(CONFIG_IPV6)
f7dcbe2f 238 .chk_addr = ipv6_chk_addr,
96058728
FW
239 .route_me_harder = ip6_route_me_harder,
240 .dev_get_saddr = ipv6_dev_get_saddr,
ac02bcf9 241 .route = __nf_ip6_route,
8527fa6c 242#if IS_ENABLED(CONFIG_SYN_COOKIES)
3006a522
FFM
243 .cookie_init_sequence = __cookie_v6_init_sequence,
244 .cookie_v6_check = __cookie_v6_check,
8527fa6c 245#endif
96058728 246#endif
ac02bcf9 247 .route_input = ip6_route_input,
f7dcbe2f 248 .fragment = ip6_fragment,
ce388f45 249 .reroute = nf_ip6_reroute,
16e6427c 250#if IS_MODULE(CONFIG_IPV6)
764dd163
PNA
251 .br_fragment = br_ip6_fragment,
252#endif
2a7851bf
FW
253};
254
2cc7d573
HW
255int __init ipv6_netfilter_init(void)
256{
2a7851bf 257 RCU_INIT_POINTER(nf_ipv6_ops, &ipv6ops);
b3a61254 258 return 0;
2cc7d573
HW
259}
260
5bf887f2
DM
261/* This can be called from inet6_init() on errors, so it cannot
262 * be marked __exit. -DaveM
263 */
264void ipv6_netfilter_fini(void)
2cc7d573 265{
2a7851bf 266 RCU_INIT_POINTER(nf_ipv6_ops, NULL);
2cc7d573 267}