1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * IPv6 specific functions of netfilter core 4 * 5 * Rusty Russell (C) 2000 6 * Patrick McHardy (C) 2006-2012 7 */ 8 #include <linux/kernel.h> 9 #include <linux/init.h> 10 #include <linux/ipv6.h> 11 #include <linux/netfilter.h> 12 #include <linux/netfilter_ipv6.h> 13 #include <linux/export.h> 14 #include <net/addrconf.h> 15 #include <net/dst.h> 16 #include <net/ipv6.h> 17 #include <net/ip6_route.h> 18 #include <net/xfrm.h> 19 #include <net/netfilter/nf_queue.h> 20 #include <net/netfilter/nf_conntrack_bridge.h> 21 #include <net/netfilter/ipv6/nf_defrag_ipv6.h> 22 #include "../bridge/br_private.h" 23 24 int ip6_route_me_harder(struct net *net, struct sock *sk_partial, struct sk_buff *skb) 25 { 26 const struct ipv6hdr *iph = ipv6_hdr(skb); 27 struct sock *sk = sk_to_full_sk(sk_partial); 28 struct net_device *dev = skb_dst_dev(skb); 29 struct flow_keys flkeys; 30 unsigned int hh_len; 31 struct dst_entry *dst; 32 int strict = (ipv6_addr_type(&iph->daddr) & 33 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)); 34 struct flowi6 fl6 = { 35 .flowi6_l3mdev = l3mdev_master_ifindex(dev), 36 .flowi6_mark = skb->mark, 37 .flowi6_uid = sock_net_uid(net, sk), 38 .daddr = iph->daddr, 39 .saddr = iph->saddr, 40 .flowlabel = ip6_flowinfo(iph), 41 }; 42 int err; 43 44 if (sk && sk->sk_bound_dev_if) 45 fl6.flowi6_oif = sk->sk_bound_dev_if; 46 else if (strict) 47 fl6.flowi6_oif = dev->ifindex; 48 49 fib6_rules_early_flow_dissect(net, skb, &fl6, &flkeys); 50 dst = ip6_route_output(net, sk, &fl6); 51 err = dst->error; 52 if (err) { 53 IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES); 54 net_dbg_ratelimited("ip6_route_me_harder: No more route\n"); 55 dst_release(dst); 56 return err; 57 } 58 59 /* Drop old route. */ 60 skb_dst_drop(skb); 61 62 skb_dst_set(skb, dst); 63 64 #ifdef CONFIG_XFRM 65 if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) && 66 xfrm_decode_session(net, skb, flowi6_to_flowi(&fl6), AF_INET6) == 0) { 67 /* ignore return value from skb_dstref_steal, xfrm_lookup takes 68 * care of dropping the refcnt if needed. 69 */ 70 skb_dstref_steal(skb); 71 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0); 72 if (IS_ERR(dst)) 73 return PTR_ERR(dst); 74 skb_dst_set(skb, dst); 75 } 76 #endif 77 78 /* Change in oif may mean change in hh_len. */ 79 hh_len = skb_dst_dev(skb)->hard_header_len; 80 if (skb_headroom(skb) < hh_len && 81 pskb_expand_head(skb, HH_DATA_ALIGN(hh_len - skb_headroom(skb)), 82 0, GFP_ATOMIC)) 83 return -ENOMEM; 84 85 return 0; 86 } 87 EXPORT_SYMBOL(ip6_route_me_harder); 88 89 static int nf_ip6_reroute(struct sk_buff *skb, 90 const struct nf_queue_entry *entry) 91 { 92 struct ip6_rt_info *rt_info = nf_queue_entry_reroute(entry); 93 94 if (entry->state.hook == NF_INET_LOCAL_OUT) { 95 const struct ipv6hdr *iph = ipv6_hdr(skb); 96 if (!ipv6_addr_equal(&iph->daddr, &rt_info->daddr) || 97 !ipv6_addr_equal(&iph->saddr, &rt_info->saddr) || 98 skb->mark != rt_info->mark) 99 return ip6_route_me_harder(entry->state.net, entry->state.sk, skb); 100 } 101 return 0; 102 } 103 104 int __nf_ip6_route(struct net *net, struct dst_entry **dst, 105 struct flowi *fl, bool strict) 106 { 107 static const struct ipv6_pinfo fake_pinfo; 108 static const struct inet_sock fake_sk = { 109 /* makes ip6_route_output set RT6_LOOKUP_F_IFACE: */ 110 .sk.sk_bound_dev_if = 1, 111 .pinet6 = (struct ipv6_pinfo *) &fake_pinfo, 112 }; 113 const void *sk = strict ? &fake_sk : NULL; 114 struct dst_entry *result; 115 int err; 116 117 result = ip6_route_output(net, sk, &fl->u.ip6); 118 err = result->error; 119 if (err) 120 dst_release(result); 121 else 122 *dst = result; 123 return err; 124 } 125 EXPORT_SYMBOL_GPL(__nf_ip6_route); 126 127 int br_ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb, 128 struct nf_bridge_frag_data *data, 129 int (*output)(struct net *, struct sock *sk, 130 const struct nf_bridge_frag_data *data, 131 struct sk_buff *)) 132 { 133 int frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size; 134 u8 tstamp_type = skb->tstamp_type; 135 ktime_t tstamp = skb->tstamp; 136 struct ip6_frag_state state; 137 u8 *prevhdr, nexthdr = 0; 138 unsigned int mtu, hlen; 139 int hroom, err = 0; 140 __be32 frag_id; 141 142 err = ip6_find_1stfragopt(skb, &prevhdr); 143 if (err < 0) 144 goto blackhole; 145 hlen = err; 146 nexthdr = *prevhdr; 147 148 mtu = skb->dev->mtu; 149 if (frag_max_size > mtu || 150 frag_max_size < IPV6_MIN_MTU) 151 goto blackhole; 152 153 mtu = frag_max_size; 154 if (mtu < hlen + sizeof(struct frag_hdr) + 8) 155 goto blackhole; 156 mtu -= hlen + sizeof(struct frag_hdr); 157 158 frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr, 159 &ipv6_hdr(skb)->saddr); 160 161 if (skb->ip_summed == CHECKSUM_PARTIAL && 162 (err = skb_checksum_help(skb))) 163 goto blackhole; 164 165 hroom = LL_RESERVED_SPACE(skb->dev); 166 if (skb_has_frag_list(skb)) { 167 unsigned int first_len = skb_pagelen(skb); 168 struct ip6_fraglist_iter iter; 169 struct sk_buff *frag2; 170 171 if (first_len - hlen > mtu) 172 goto blackhole; 173 174 if (skb_cloned(skb) || 175 skb_headroom(skb) < (hroom + sizeof(struct frag_hdr))) 176 goto slow_path; 177 178 skb_walk_frags(skb, frag2) { 179 if (frag2->len > mtu) 180 goto blackhole; 181 182 /* Partially cloned skb? */ 183 if (skb_shared(frag2) || 184 skb_headroom(frag2) < (hlen + hroom + sizeof(struct frag_hdr))) 185 goto slow_path; 186 } 187 188 err = ip6_fraglist_init(skb, hlen, prevhdr, nexthdr, frag_id, 189 &iter); 190 if (err < 0) 191 goto blackhole; 192 193 for (;;) { 194 /* Prepare header of the next frame, 195 * before previous one went down. 196 */ 197 if (iter.frag) 198 ip6_fraglist_prepare(skb, &iter); 199 200 skb_set_delivery_time(skb, tstamp, tstamp_type); 201 err = output(net, sk, data, skb); 202 if (err || !iter.frag) 203 break; 204 205 skb = ip6_fraglist_next(&iter); 206 } 207 208 kfree(iter.tmp_hdr); 209 if (!err) 210 return 0; 211 212 kfree_skb_list(iter.frag); 213 return err; 214 } 215 slow_path: 216 /* This is a linearized skbuff, the original geometry is lost for us. 217 * This may also be a clone skbuff, we could preserve the geometry for 218 * the copies but probably not worth the effort. 219 */ 220 ip6_frag_init(skb, hlen, mtu, skb->dev->needed_tailroom, 221 LL_RESERVED_SPACE(skb->dev), prevhdr, nexthdr, frag_id, 222 &state); 223 224 while (state.left > 0) { 225 struct sk_buff *skb2; 226 227 skb2 = ip6_frag_next(skb, &state); 228 if (IS_ERR(skb2)) { 229 err = PTR_ERR(skb2); 230 goto blackhole; 231 } 232 233 skb_set_delivery_time(skb2, tstamp, tstamp_type); 234 err = output(net, sk, data, skb2); 235 if (err) 236 goto blackhole; 237 } 238 consume_skb(skb); 239 return err; 240 241 blackhole: 242 kfree_skb(skb); 243 return 0; 244 } 245 EXPORT_SYMBOL_GPL(br_ip6_fragment); 246 247 static const struct nf_ipv6_ops ipv6ops = { 248 #if IS_MODULE(CONFIG_IPV6) 249 .chk_addr = ipv6_chk_addr, 250 .route_me_harder = ip6_route_me_harder, 251 .dev_get_saddr = ipv6_dev_get_saddr, 252 .route = __nf_ip6_route, 253 #if IS_ENABLED(CONFIG_SYN_COOKIES) 254 .cookie_init_sequence = __cookie_v6_init_sequence, 255 .cookie_v6_check = __cookie_v6_check, 256 #endif 257 #endif 258 .route_input = ip6_route_input, 259 .fragment = ip6_fragment, 260 .reroute = nf_ip6_reroute, 261 #if IS_MODULE(CONFIG_IPV6) 262 .br_fragment = br_ip6_fragment, 263 #endif 264 }; 265 266 int __init ipv6_netfilter_init(void) 267 { 268 RCU_INIT_POINTER(nf_ipv6_ops, &ipv6ops); 269 return 0; 270 } 271 272 /* This can be called from inet6_init() on errors, so it cannot 273 * be marked __exit. -DaveM 274 */ 275 void ipv6_netfilter_fini(void) 276 { 277 RCU_INIT_POINTER(nf_ipv6_ops, NULL); 278 } 279