net/ipv6/netfilter.c

0001 /*
0002  * IPv6 specific functions of netfilter core
0003  *
0004  * Rusty Russell (C) 2000 -- This code is GPL.
0005  * Patrick McHardy (C) 2006-2012
0006  */
0007 #include <linux/kernel.h>
0008 #include <linux/init.h>
0009 #include <linux/ipv6.h>
0010 #include <linux/netfilter.h>
0011 #include <linux/netfilter_ipv6.h>
0012 #include <linux/export.h>
0013 #include <net/addrconf.h>
0014 #include <net/dst.h>
0015 #include <net/ipv6.h>
0016 #include <net/ip6_route.h>
0017 #include <net/xfrm.h>
0018 #include <net/netfilter/nf_queue.h>
0019 #include <net/netfilter/nf_conntrack_bridge.h>
0020 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
0021 #include "../bridge/br_private.h"
0022
0023 int ip6_route_me_harder(struct net *net, struct sock *sk_partial, struct sk_buff *skb)
0024 {
0025     const struct ipv6hdr *iph = ipv6_hdr(skb);
0026     struct sock *sk = sk_to_full_sk(sk_partial);
0027     struct net_device *dev = skb_dst(skb)->dev;
0028     struct flow_keys flkeys;
0029     unsigned int hh_len;
0030     struct dst_entry *dst;
0031     int strict = (ipv6_addr_type(&iph->daddr) &
0032               (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
0033     struct flowi6 fl6 = {
0034         .flowi6_l3mdev = l3mdev_master_ifindex(dev),
0035         .flowi6_mark = skb->mark,
0036         .flowi6_uid = sock_net_uid(net, sk),
0037         .daddr = iph->daddr,
0038         .saddr = iph->saddr,
0039     };
0040     int err;
0041
0042     if (sk && sk->sk_bound_dev_if)
0043         fl6.flowi6_oif = sk->sk_bound_dev_if;
0044     else if (strict)
0045         fl6.flowi6_oif = dev->ifindex;
0046
0047     fib6_rules_early_flow_dissect(net, skb, &fl6, &flkeys);
0048     dst = ip6_route_output(net, sk, &fl6);
0049     err = dst->error;
0050     if (err) {
0051         IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
0052         net_dbg_ratelimited("ip6_route_me_harder: No more route\n");
0053         dst_release(dst);
0054         return err;
0055     }
0056
0057     /* Drop old route. */
0058     skb_dst_drop(skb);
0059
0060     skb_dst_set(skb, dst);
0061
0062 #ifdef CONFIG_XFRM
0063     if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
0064         xfrm_decode_session(skb, flowi6_to_flowi(&fl6), AF_INET6) == 0) {
0065         skb_dst_set(skb, NULL);
0066         dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
0067         if (IS_ERR(dst))
0068             return PTR_ERR(dst);
0069         skb_dst_set(skb, dst);
0070     }
0071 #endif
0072
0073     /* Change in oif may mean change in hh_len. */
0074     hh_len = skb_dst(skb)->dev->hard_header_len;
0075     if (skb_headroom(skb) < hh_len &&
0076         pskb_expand_head(skb, HH_DATA_ALIGN(hh_len - skb_headroom(skb)),
0077                  0, GFP_ATOMIC))
0078         return -ENOMEM;
0079
0080     return 0;
0081 }
0082 EXPORT_SYMBOL(ip6_route_me_harder);
0083
0084 static int nf_ip6_reroute(struct sk_buff *skb,
0085               const struct nf_queue_entry *entry)
0086 {
0087     struct ip6_rt_info *rt_info = nf_queue_entry_reroute(entry);
0088
0089     if (entry->state.hook == NF_INET_LOCAL_OUT) {
0090         const struct ipv6hdr *iph = ipv6_hdr(skb);
0091         if (!ipv6_addr_equal(&iph->daddr, &rt_info->daddr) ||
0092             !ipv6_addr_equal(&iph->saddr, &rt_info->saddr) ||
0093             skb->mark != rt_info->mark)
0094             return ip6_route_me_harder(entry->state.net, entry->state.sk, skb);
0095     }
0096     return 0;
0097 }
0098
0099 int __nf_ip6_route(struct net *net, struct dst_entry **dst,
0100            struct flowi *fl, bool strict)
0101 {
0102     static const struct ipv6_pinfo fake_pinfo;
0103     static const struct inet_sock fake_sk = {
0104         /* makes ip6_route_output set RT6_LOOKUP_F_IFACE: */
0105         .sk.sk_bound_dev_if = 1,
0106         .pinet6 = (struct ipv6_pinfo *) &fake_pinfo,
0107     };
0108     const void *sk = strict ? &fake_sk : NULL;
0109     struct dst_entry *result;
0110     int err;
0111
0112     result = ip6_route_output(net, sk, &fl->u.ip6);
0113     err = result->error;
0114     if (err)
0115         dst_release(result);
0116     else
0117         *dst = result;
0118     return err;
0119 }
0120 EXPORT_SYMBOL_GPL(__nf_ip6_route);
0121
0122 int br_ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
0123             struct nf_bridge_frag_data *data,
0124             int (*output)(struct net *, struct sock *sk,
0125                   const struct nf_bridge_frag_data *data,
0126                   struct sk_buff *))
0127 {
0128     int frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size;
0129     bool mono_delivery_time = skb->mono_delivery_time;
0130     ktime_t tstamp = skb->tstamp;
0131     struct ip6_frag_state state;
0132     u8 *prevhdr, nexthdr = 0;
0133     unsigned int mtu, hlen;
0134     int hroom, err = 0;
0135     __be32 frag_id;
0136
0137     err = ip6_find_1stfragopt(skb, &prevhdr);
0138     if (err < 0)
0139         goto blackhole;
0140     hlen = err;
0141     nexthdr = *prevhdr;
0142
0143     mtu = skb->dev->mtu;
0144     if (frag_max_size > mtu ||
0145         frag_max_size < IPV6_MIN_MTU)
0146         goto blackhole;
0147
0148     mtu = frag_max_size;
0149     if (mtu < hlen + sizeof(struct frag_hdr) + 8)
0150         goto blackhole;
0151     mtu -= hlen + sizeof(struct frag_hdr);
0152
0153     frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
0154                     &ipv6_hdr(skb)->saddr);
0155
0156     if (skb->ip_summed == CHECKSUM_PARTIAL &&
0157         (err = skb_checksum_help(skb)))
0158         goto blackhole;
0159
0160     hroom = LL_RESERVED_SPACE(skb->dev);
0161     if (skb_has_frag_list(skb)) {
0162         unsigned int first_len = skb_pagelen(skb);
0163         struct ip6_fraglist_iter iter;
0164         struct sk_buff *frag2;
0165
0166         if (first_len - hlen > mtu ||
0167             skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
0168             goto blackhole;
0169
0170         if (skb_cloned(skb))
0171             goto slow_path;
0172
0173         skb_walk_frags(skb, frag2) {
0174             if (frag2->len > mtu ||
0175                 skb_headroom(frag2) < (hlen + hroom + sizeof(struct frag_hdr)))
0176                 goto blackhole;
0177
0178             /* Partially cloned skb? */
0179             if (skb_shared(frag2))
0180                 goto slow_path;
0181         }
0182
0183         err = ip6_fraglist_init(skb, hlen, prevhdr, nexthdr, frag_id,
0184                     &iter);
0185         if (err < 0)
0186             goto blackhole;
0187
0188         for (;;) {
0189             /* Prepare header of the next frame,
0190              * before previous one went down.
0191              */
0192             if (iter.frag)
0193                 ip6_fraglist_prepare(skb, &iter);
0194
0195             skb_set_delivery_time(skb, tstamp, mono_delivery_time);
0196             err = output(net, sk, data, skb);
0197             if (err || !iter.frag)
0198                 break;
0199
0200             skb = ip6_fraglist_next(&iter);
0201         }
0202
0203         kfree(iter.tmp_hdr);
0204         if (!err)
0205             return 0;
0206
0207         kfree_skb_list(iter.frag);
0208         return err;
0209     }
0210 slow_path:
0211     /* This is a linearized skbuff, the original geometry is lost for us.
0212      * This may also be a clone skbuff, we could preserve the geometry for
0213      * the copies but probably not worth the effort.
0214      */
0215     ip6_frag_init(skb, hlen, mtu, skb->dev->needed_tailroom,
0216               LL_RESERVED_SPACE(skb->dev), prevhdr, nexthdr, frag_id,
0217               &state);
0218
0219     while (state.left > 0) {
0220         struct sk_buff *skb2;
0221
0222         skb2 = ip6_frag_next(skb, &state);
0223         if (IS_ERR(skb2)) {
0224             err = PTR_ERR(skb2);
0225             goto blackhole;
0226         }
0227
0228         skb_set_delivery_time(skb2, tstamp, mono_delivery_time);
0229         err = output(net, sk, data, skb2);
0230         if (err)
0231             goto blackhole;
0232     }
0233     consume_skb(skb);
0234     return err;
0235
0236 blackhole:
0237     kfree_skb(skb);
0238     return 0;
0239 }
0240 EXPORT_SYMBOL_GPL(br_ip6_fragment);
0241
0242 static const struct nf_ipv6_ops ipv6ops = {
0243 #if IS_MODULE(CONFIG_IPV6)
0244     .chk_addr       = ipv6_chk_addr,
0245     .route_me_harder    = ip6_route_me_harder,
0246     .dev_get_saddr      = ipv6_dev_get_saddr,
0247     .route          = __nf_ip6_route,
0248 #if IS_ENABLED(CONFIG_SYN_COOKIES)
0249     .cookie_init_sequence   = __cookie_v6_init_sequence,
0250     .cookie_v6_check    = __cookie_v6_check,
0251 #endif
0252 #endif
0253     .route_input        = ip6_route_input,
0254     .fragment       = ip6_fragment,
0255     .reroute        = nf_ip6_reroute,
0256 #if IS_MODULE(CONFIG_IPV6)
0257     .br_fragment        = br_ip6_fragment,
0258 #endif
0259 };
0260
0261 int __init ipv6_netfilter_init(void)
0262 {
0263     RCU_INIT_POINTER(nf_ipv6_ops, &ipv6ops);
0264     return 0;
0265 }
0266
0267 /* This can be called from inet6_init() on errors, so it cannot
0268  * be marked __exit. -DaveM
0269  */
0270 void ipv6_netfilter_fini(void)
0271 {
0272     RCU_INIT_POINTER(nf_ipv6_ops, NULL);
0273 }