net/ipv4/netfilter.c

   1 /*
   2  * IPv4 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/netfilter.h>
   9 #include <linux/netfilter_ipv4.h>
  10 #include <linux/ip.h>
  11 #include <linux/skbuff.h>
  12 #include <linux/gfp.h>
  13 #include <linux/export.h>
  14 #include <net/route.h>
  15 #include <net/xfrm.h>
  16 #include <net/ip.h>
  17 #include <net/netfilter/nf_queue.h>
  18
  19 /* route_me_harder function, used by iptable_nat, iptable_mangle + ip_queue */
  20 int ip_route_me_harder(struct net *net, struct sk_buff *skb, unsigned int addr_type)
  21 {
  22         const struct iphdr *iph = ip_hdr(skb);
  23         struct rtable *rt;
  24         struct flowi4 fl4 = {};
  25         __be32 saddr = iph->saddr;
  26         __u8 flags = skb->sk ? inet_sk_flowi_flags(skb->sk) : 0;
  27         unsigned int hh_len;
  28
  29         if (addr_type == RTN_UNSPEC)
  30                 addr_type = inet_addr_type(net, saddr);
  31         if (addr_type == RTN_LOCAL || addr_type == RTN_UNICAST)
  32                 flags |= FLOWI_FLAG_ANYSRC;
  33         else
  34                 saddr = 0;
  35
  36         /* some non-standard hacks like ipt_REJECT.c:send_reset() can cause
  37          * packets with foreign saddr to appear on the NF_INET_LOCAL_OUT hook.
  38          */
  39         fl4.daddr = iph->daddr;
  40         fl4.saddr = saddr;
  41         fl4.flowi4_tos = RT_TOS(iph->tos);
  42         fl4.flowi4_oif = skb->sk ? skb->sk->sk_bound_dev_if : 0;
  43         fl4.flowi4_mark = skb->mark;
  44         fl4.flowi4_flags = flags;
  45         rt = ip_route_output_key(net, &fl4);
  46         if (IS_ERR(rt))
  47                 return PTR_ERR(rt);
  48
  49         /* Drop old route. */
  50         skb_dst_drop(skb);
  51         skb_dst_set(skb, &rt->dst);
  52
  53         if (skb_dst(skb)->error)
  54                 return skb_dst(skb)->error;
  55
  56 #ifdef CONFIG_XFRM
  57         if (!(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
  58             xfrm_decode_session(skb, flowi4_to_flowi(&fl4), AF_INET) == 0) {
  59                 struct dst_entry *dst = skb_dst(skb);
  60                 skb_dst_set(skb, NULL);
  61                 dst = xfrm_lookup(net, dst, flowi4_to_flowi(&fl4), skb->sk, 0);
  62                 if (IS_ERR(dst))
  63                         return PTR_ERR(dst);
  64                 skb_dst_set(skb, dst);
  65         }
  66 #endif
  67
  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))
  73                 return -ENOMEM;
  74
  75         return 0;
  76 }
  77 EXPORT_SYMBOL(ip_route_me_harder);
  78
  79 /*
  80  * Extra routing may needed on local out, as the QUEUE target never
  81  * returns control to the table.
  82  */
  83
  84 struct ip_rt_info {
  85         __be32 daddr;
  86         __be32 saddr;
  87         u_int8_t tos;
  88         u_int32_t mark;
  89 };
  90
  91 static void nf_ip_saveroute(const struct sk_buff *skb,
  92                             struct nf_queue_entry *entry)
  93 {
  94         struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
  95
  96         if (entry->state.hook == NF_INET_LOCAL_OUT) {
  97                 const struct iphdr *iph = ip_hdr(skb);
  98
  99                 rt_info->tos = iph->tos;
 100                 rt_info->daddr = iph->daddr;
 101                 rt_info->saddr = iph->saddr;
 102                 rt_info->mark = skb->mark;
 103         }
 104 }
 105
 106 static int nf_ip_reroute(struct net *net, struct sk_buff *skb,
 107                          const struct nf_queue_entry *entry)
 108 {
 109         const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
 110
 111         if (entry->state.hook == NF_INET_LOCAL_OUT) {
 112                 const struct iphdr *iph = ip_hdr(skb);
 113
 114                 if (!(iph->tos == rt_info->tos &&
 115                       skb->mark == rt_info->mark &&
 116                       iph->daddr == rt_info->daddr &&
 117                       iph->saddr == rt_info->saddr))
 118                         return ip_route_me_harder(net, skb, RTN_UNSPEC);
 119         }
 120         return 0;
 121 }
 122
 123 __sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
 124                             unsigned int dataoff, u_int8_t protocol)
 125 {
 126         const struct iphdr *iph = ip_hdr(skb);
 127         __sum16 csum = 0;
 128
 129         switch (skb->ip_summed) {
 130         case CHECKSUM_COMPLETE:
 131                 if (hook != NF_INET_PRE_ROUTING && hook != NF_INET_LOCAL_IN)
 132                         break;
 133                 if ((protocol == 0 && !csum_fold(skb->csum)) ||
 134                     !csum_tcpudp_magic(iph->saddr, iph->daddr,
 135                                        skb->len - dataoff, protocol,
 136                                        skb->csum)) {
 137                         skb->ip_summed = CHECKSUM_UNNECESSARY;
 138                         break;
 139                 }
 140                 /* fall through */
 141         case CHECKSUM_NONE:
 142                 if (protocol == 0)
 143                         skb->csum = 0;
 144                 else
 145                         skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
 146                                                        skb->len - dataoff,
 147                                                        protocol, 0);
 148                 csum = __skb_checksum_complete(skb);
 149         }
 150         return csum;
 151 }
 152 EXPORT_SYMBOL(nf_ip_checksum);
 153
 154 static __sum16 nf_ip_checksum_partial(struct sk_buff *skb, unsigned int hook,
 155                                       unsigned int dataoff, unsigned int len,
 156                                       u_int8_t protocol)
 157 {
 158         const struct iphdr *iph = ip_hdr(skb);
 159         __sum16 csum = 0;
 160
 161         switch (skb->ip_summed) {
 162         case CHECKSUM_COMPLETE:
 163                 if (len == skb->len - dataoff)
 164                         return nf_ip_checksum(skb, hook, dataoff, protocol);
 165                 /* fall through */
 166         case CHECKSUM_NONE:
 167                 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, protocol,
 168                                                skb->len - dataoff, 0);
 169                 skb->ip_summed = CHECKSUM_NONE;
 170                 return __skb_checksum_complete_head(skb, dataoff + len);
 171         }
 172         return csum;
 173 }
 174
 175 static int nf_ip_route(struct net *net, struct dst_entry **dst,
 176                        struct flowi *fl, bool strict __always_unused)
 177 {
 178         struct rtable *rt = ip_route_output_key(net, &fl->u.ip4);
 179         if (IS_ERR(rt))
 180                 return PTR_ERR(rt);
 181         *dst = &rt->dst;
 182         return 0;
 183 }
 184
 185 static const struct nf_afinfo nf_ip_afinfo = {
 186         .family                 = AF_INET,
 187         .checksum               = nf_ip_checksum,
 188         .checksum_partial       = nf_ip_checksum_partial,
 189         .route                  = nf_ip_route,
 190         .saveroute              = nf_ip_saveroute,
 191         .reroute                = nf_ip_reroute,
 192         .route_key_size         = sizeof(struct ip_rt_info),
 193 };
 194
 195 static int __init ipv4_netfilter_init(void)
 196 {
 197         return nf_register_afinfo(&nf_ip_afinfo);
 198 }
 199 subsys_initcall(ipv4_netfilter_init);