net/ipv4/netfilter.c

   1 /* IPv4 specific functions of netfilter core */
   2 #include <linux/kernel.h>
   3 #include <linux/netfilter.h>
   4 #include <linux/netfilter_ipv4.h>
   5 #include <linux/ip.h>
   6 #include <linux/skbuff.h>
   7 #include <net/route.h>
   8 #include <net/xfrm.h>
   9 #include <net/ip.h>
  10 #include <net/netfilter/nf_queue.h>
  11
  12 /* route_me_harder function, used by iptable_nat, iptable_mangle + ip_queue */
  13 int ip_route_me_harder(struct sk_buff *skb, unsigned addr_type)
  14 {
  15         const struct iphdr *iph = ip_hdr(skb);
  16         struct rtable *rt;
  17         struct flowi fl = {};
  18         struct dst_entry *odst;
  19         unsigned int hh_len;
  20         unsigned int type;
  21
  22         type = inet_addr_type(&init_net, iph->saddr);
  23         if (addr_type == RTN_UNSPEC)
  24                 addr_type = type;
  25
  26         /* some non-standard hacks like ipt_REJECT.c:send_reset() can cause
  27          * packets with foreign saddr to appear on the NF_INET_LOCAL_OUT hook.
  28          */
  29         if (addr_type == RTN_LOCAL) {
  30                 fl.nl_u.ip4_u.daddr = iph->daddr;
  31                 if (type == RTN_LOCAL)
  32                         fl.nl_u.ip4_u.saddr = iph->saddr;
  33                 fl.nl_u.ip4_u.tos = RT_TOS(iph->tos);
  34                 fl.oif = skb->sk ? skb->sk->sk_bound_dev_if : 0;
  35                 fl.mark = skb->mark;
  36                 if (ip_route_output_key(&init_net, &rt, &fl) != 0)
  37                         return -1;
  38
  39                 /* Drop old route. */
  40                 dst_release(skb->dst);
  41                 skb->dst = &rt->u.dst;
  42         } else {
  43                 /* non-local src, find valid iif to satisfy
  44                  * rp-filter when calling ip_route_input. */
  45                 fl.nl_u.ip4_u.daddr = iph->saddr;
  46                 if (ip_route_output_key(&init_net, &rt, &fl) != 0)
  47                         return -1;
  48
  49                 odst = skb->dst;
  50                 if (ip_route_input(skb, iph->daddr, iph->saddr,
  51                                    RT_TOS(iph->tos), rt->u.dst.dev) != 0) {
  52                         dst_release(&rt->u.dst);
  53                         return -1;
  54                 }
  55                 dst_release(&rt->u.dst);
  56                 dst_release(odst);
  57         }
  58
  59         if (skb->dst->error)
  60                 return -1;
  61
  62 #ifdef CONFIG_XFRM
  63         if (!(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
  64             xfrm_decode_session(skb, &fl, AF_INET) == 0)
  65                 if (xfrm_lookup(&skb->dst, &fl, skb->sk, 0))
  66                         return -1;
  67 #endif
  68
  69         /* Change in oif may mean change in hh_len. */
  70         hh_len = skb->dst->dev->hard_header_len;
  71         if (skb_headroom(skb) < hh_len &&
  72             pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
  73                 return -1;
  74
  75         return 0;
  76 }
  77 EXPORT_SYMBOL(ip_route_me_harder);
  78
  79 #ifdef CONFIG_XFRM
  80 int ip_xfrm_me_harder(struct sk_buff *skb)
  81 {
  82         struct flowi fl;
  83         unsigned int hh_len;
  84         struct dst_entry *dst;
  85
  86         if (IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED)
  87                 return 0;
  88         if (xfrm_decode_session(skb, &fl, AF_INET) < 0)
  89                 return -1;
  90
  91         dst = skb->dst;
  92         if (dst->xfrm)
  93                 dst = ((struct xfrm_dst *)dst)->route;
  94         dst_hold(dst);
  95
  96         if (xfrm_lookup(&dst, &fl, skb->sk, 0) < 0)
  97                 return -1;
  98
  99         dst_release(skb->dst);
 100         skb->dst = dst;
 101
 102         /* Change in oif may mean change in hh_len. */
 103         hh_len = skb->dst->dev->hard_header_len;
 104         if (skb_headroom(skb) < hh_len &&
 105             pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
 106                 return -1;
 107         return 0;
 108 }
 109 EXPORT_SYMBOL(ip_xfrm_me_harder);
 110 #endif
 111
 112 void (*ip_nat_decode_session)(struct sk_buff *, struct flowi *);
 113 EXPORT_SYMBOL(ip_nat_decode_session);
 114
 115 /*
 116  * Extra routing may needed on local out, as the QUEUE target never
 117  * returns control to the table.
 118  */
 119
 120 struct ip_rt_info {
 121         __be32 daddr;
 122         __be32 saddr;
 123         u_int8_t tos;
 124 };
 125
 126 static void nf_ip_saveroute(const struct sk_buff *skb,
 127                             struct nf_queue_entry *entry)
 128 {
 129         struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
 130
 131         if (entry->hook == NF_INET_LOCAL_OUT) {
 132                 const struct iphdr *iph = ip_hdr(skb);
 133
 134                 rt_info->tos = iph->tos;
 135                 rt_info->daddr = iph->daddr;
 136                 rt_info->saddr = iph->saddr;
 137         }
 138 }
 139
 140 static int nf_ip_reroute(struct sk_buff *skb,
 141                          const struct nf_queue_entry *entry)
 142 {
 143         const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
 144
 145         if (entry->hook == NF_INET_LOCAL_OUT) {
 146                 const struct iphdr *iph = ip_hdr(skb);
 147
 148                 if (!(iph->tos == rt_info->tos
 149                       && iph->daddr == rt_info->daddr
 150                       && iph->saddr == rt_info->saddr))
 151                         return ip_route_me_harder(skb, RTN_UNSPEC);
 152         }
 153         return 0;
 154 }
 155
 156 __sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
 157                             unsigned int dataoff, u_int8_t protocol)
 158 {
 159         const struct iphdr *iph = ip_hdr(skb);
 160         __sum16 csum = 0;
 161
 162         switch (skb->ip_summed) {
 163         case CHECKSUM_COMPLETE:
 164                 if (hook != NF_INET_PRE_ROUTING && hook != NF_INET_LOCAL_IN)
 165                         break;
 166                 if ((protocol == 0 && !csum_fold(skb->csum)) ||
 167                     !csum_tcpudp_magic(iph->saddr, iph->daddr,
 168                                        skb->len - dataoff, protocol,
 169                                        skb->csum)) {
 170                         skb->ip_summed = CHECKSUM_UNNECESSARY;
 171                         break;
 172                 }
 173                 /* fall through */
 174         case CHECKSUM_NONE:
 175                 if (protocol == 0)
 176                         skb->csum = 0;
 177                 else
 178                         skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
 179                                                        skb->len - dataoff,
 180                                                        protocol, 0);
 181                 csum = __skb_checksum_complete(skb);
 182         }
 183         return csum;
 184 }
 185 EXPORT_SYMBOL(nf_ip_checksum);
 186
 187 static __sum16 nf_ip_checksum_partial(struct sk_buff *skb, unsigned int hook,
 188                                       unsigned int dataoff, unsigned int len,
 189                                       u_int8_t protocol)
 190 {
 191         const struct iphdr *iph = ip_hdr(skb);
 192         __sum16 csum = 0;
 193
 194         switch (skb->ip_summed) {
 195         case CHECKSUM_COMPLETE:
 196                 if (len == skb->len - dataoff)
 197                         return nf_ip_checksum(skb, hook, dataoff, protocol);
 198                 /* fall through */
 199         case CHECKSUM_NONE:
 200                 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, protocol,
 201                                                skb->len - dataoff, 0);
 202                 skb->ip_summed = CHECKSUM_NONE;
 203                 csum = __skb_checksum_complete_head(skb, dataoff + len);
 204                 if (!csum)
 205                         skb->ip_summed = CHECKSUM_UNNECESSARY;
 206         }
 207         return csum;
 208 }
 209
 210 static int nf_ip_route(struct dst_entry **dst, struct flowi *fl)
 211 {
 212         return ip_route_output_key(&init_net, (struct rtable **)dst, fl);
 213 }
 214
 215 static const struct nf_afinfo nf_ip_afinfo = {
 216         .family                 = AF_INET,
 217         .checksum               = nf_ip_checksum,
 218         .checksum_partial       = nf_ip_checksum_partial,
 219         .route                  = nf_ip_route,
 220         .saveroute              = nf_ip_saveroute,
 221         .reroute                = nf_ip_reroute,
 222         .route_key_size         = sizeof(struct ip_rt_info),
 223 };
 224
 225 static int ipv4_netfilter_init(void)
 226 {
 227         return nf_register_afinfo(&nf_ip_afinfo);
 228 }
 229
 230 static void ipv4_netfilter_fini(void)
 231 {
 232         nf_unregister_afinfo(&nf_ip_afinfo);
 233 }
 234
 235 module_init(ipv4_netfilter_init);
 236 module_exit(ipv4_netfilter_fini);
 237
 238 #ifdef CONFIG_SYSCTL
 239 struct ctl_path nf_net_ipv4_netfilter_sysctl_path[] = {
 240         { .procname = "net", .ctl_name = CTL_NET, },
 241         { .procname = "ipv4", .ctl_name = NET_IPV4, },
 242         { .procname = "netfilter", .ctl_name = NET_IPV4_NETFILTER, },
 243         { }
 244 };
 245 EXPORT_SYMBOL_GPL(nf_net_ipv4_netfilter_sysctl_path);
 246 #endif /* CONFIG_SYSCTL */