net/mpls/mpls_iptunnel.c

   1 /*
   2  * mpls tunnels An implementation mpls tunnels using the light weight tunnel
   3  *              infrastructure
   4  *
   5  * Authors:     Roopa Prabhu, <roopa@cumulusnetworks.com>
   6  *
   7  *              This program is free software; you can redistribute it and/or
   8  *              modify it under the terms of the GNU General Public License
   9  *              as published by the Free Software Foundation; either version
  10  *              2 of the License, or (at your option) any later version.
  11  *
  12  */
  13 #include <linux/types.h>
  14 #include <linux/skbuff.h>
  15 #include <linux/net.h>
  16 #include <linux/module.h>
  17 #include <linux/mpls.h>
  18 #include <linux/vmalloc.h>
  19 #include <net/ip.h>
  20 #include <net/dst.h>
  21 #include <net/lwtunnel.h>
  22 #include <net/netevent.h>
  23 #include <net/netns/generic.h>
  24 #include <net/ip6_fib.h>
  25 #include <net/route.h>
  26 #include <net/mpls_iptunnel.h>
  27 #include <linux/mpls_iptunnel.h>
  28 #include "internal.h"
  29
  30 static const struct nla_policy mpls_iptunnel_policy[MPLS_IPTUNNEL_MAX + 1] = {
  31         [MPLS_IPTUNNEL_DST]     = { .type = NLA_U32 },
  32 };
  33
  34 static unsigned int mpls_encap_size(struct mpls_iptunnel_encap *en)
  35 {
  36         /* The size of the layer 2.5 labels to be added for this route */
  37         return en->labels * sizeof(struct mpls_shim_hdr);
  38 }
  39
  40 static int mpls_xmit(struct sk_buff *skb)
  41 {
  42         struct mpls_iptunnel_encap *tun_encap_info;
  43         struct mpls_shim_hdr *hdr;
  44         struct net_device *out_dev;
  45         unsigned int hh_len;
  46         unsigned int new_header_size;
  47         unsigned int mtu;
  48         struct dst_entry *dst = skb_dst(skb);
  49         struct rtable *rt = NULL;
  50         struct rt6_info *rt6 = NULL;
  51         int err = 0;
  52         bool bos;
  53         int i;
  54         unsigned int ttl;
  55
  56         /* Obtain the ttl */
  57         if (dst->ops->family == AF_INET) {
  58                 ttl = ip_hdr(skb)->ttl;
  59                 rt = (struct rtable *)dst;
  60         } else if (dst->ops->family == AF_INET6) {
  61                 ttl = ipv6_hdr(skb)->hop_limit;
  62                 rt6 = (struct rt6_info *)dst;
  63         } else {
  64                 goto drop;
  65         }
  66
  67         skb_orphan(skb);
  68
  69         /* Find the output device */
  70         out_dev = dst->dev;
  71         if (!mpls_output_possible(out_dev) ||
  72             !dst->lwtstate || skb_warn_if_lro(skb))
  73                 goto drop;
  74
  75         skb_forward_csum(skb);
  76
  77         tun_encap_info = mpls_lwtunnel_encap(dst->lwtstate);
  78
  79         /* Verify the destination can hold the packet */
  80         new_header_size = mpls_encap_size(tun_encap_info);
  81         mtu = mpls_dev_mtu(out_dev);
  82         if (mpls_pkt_too_big(skb, mtu - new_header_size))
  83                 goto drop;
  84
  85         hh_len = LL_RESERVED_SPACE(out_dev);
  86         if (!out_dev->header_ops)
  87                 hh_len = 0;
  88
  89         /* Ensure there is enough space for the headers in the skb */
  90         if (skb_cow(skb, hh_len + new_header_size))
  91                 goto drop;
  92
  93         skb_set_inner_protocol(skb, skb->protocol);
  94         skb_reset_inner_network_header(skb);
  95
  96         skb_push(skb, new_header_size);
  97
  98         skb_reset_network_header(skb);
  99
 100         skb->dev = out_dev;
 101         skb->protocol = htons(ETH_P_MPLS_UC);
 102
 103         /* Push the new labels */
 104         hdr = mpls_hdr(skb);
 105         bos = true;
 106         for (i = tun_encap_info->labels - 1; i >= 0; i--) {
 107                 hdr[i] = mpls_entry_encode(tun_encap_info->label[i],
 108                                            ttl, 0, bos);
 109                 bos = false;
 110         }
 111
 112         if (rt)
 113                 err = neigh_xmit(NEIGH_ARP_TABLE, out_dev, &rt->rt_gateway,
 114                                  skb);
 115         else if (rt6)
 116                 err = neigh_xmit(NEIGH_ND_TABLE, out_dev, &rt6->rt6i_gateway,
 117                                  skb);
 118         if (err)
 119                 net_dbg_ratelimited("%s: packet transmission failed: %d\n",
 120                                     __func__, err);
 121
 122         return LWTUNNEL_XMIT_DONE;
 123
 124 drop:
 125         kfree_skb(skb);
 126         return -EINVAL;
 127 }
 128
 129 static int mpls_build_state(struct net_device *dev, struct nlattr *nla,
 130                             unsigned int family, const void *cfg,
 131                             struct lwtunnel_state **ts)
 132 {
 133         struct mpls_iptunnel_encap *tun_encap_info;
 134         struct nlattr *tb[MPLS_IPTUNNEL_MAX + 1];
 135         struct lwtunnel_state *newts;
 136         int ret;
 137
 138         ret = nla_parse_nested(tb, MPLS_IPTUNNEL_MAX, nla,
 139                                mpls_iptunnel_policy);
 140         if (ret < 0)
 141                 return ret;
 142
 143         if (!tb[MPLS_IPTUNNEL_DST])
 144                 return -EINVAL;
 145
 146
 147         newts = lwtunnel_state_alloc(sizeof(*tun_encap_info));
 148         if (!newts)
 149                 return -ENOMEM;
 150
 151         tun_encap_info = mpls_lwtunnel_encap(newts);
 152         ret = nla_get_labels(tb[MPLS_IPTUNNEL_DST], MAX_NEW_LABELS,
 153                              &tun_encap_info->labels, tun_encap_info->label);
 154         if (ret)
 155                 goto errout;
 156         newts->type = LWTUNNEL_ENCAP_MPLS;
 157         newts->flags |= LWTUNNEL_STATE_XMIT_REDIRECT;
 158         newts->headroom = mpls_encap_size(tun_encap_info);
 159
 160         *ts = newts;
 161
 162         return 0;
 163
 164 errout:
 165         kfree(newts);
 166         *ts = NULL;
 167
 168         return ret;
 169 }
 170
 171 static int mpls_fill_encap_info(struct sk_buff *skb,
 172                                 struct lwtunnel_state *lwtstate)
 173 {
 174         struct mpls_iptunnel_encap *tun_encap_info;
 175
 176         tun_encap_info = mpls_lwtunnel_encap(lwtstate);
 177
 178         if (nla_put_labels(skb, MPLS_IPTUNNEL_DST, tun_encap_info->labels,
 179                            tun_encap_info->label))
 180                 goto nla_put_failure;
 181
 182         return 0;
 183
 184 nla_put_failure:
 185         return -EMSGSIZE;
 186 }
 187
 188 static int mpls_encap_nlsize(struct lwtunnel_state *lwtstate)
 189 {
 190         struct mpls_iptunnel_encap *tun_encap_info;
 191
 192         tun_encap_info = mpls_lwtunnel_encap(lwtstate);
 193
 194         return nla_total_size(tun_encap_info->labels * 4);
 195 }
 196
 197 static int mpls_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
 198 {
 199         struct mpls_iptunnel_encap *a_hdr = mpls_lwtunnel_encap(a);
 200         struct mpls_iptunnel_encap *b_hdr = mpls_lwtunnel_encap(b);
 201         int l;
 202
 203         if (a_hdr->labels != b_hdr->labels)
 204                 return 1;
 205
 206         for (l = 0; l < MAX_NEW_LABELS; l++)
 207                 if (a_hdr->label[l] != b_hdr->label[l])
 208                         return 1;
 209         return 0;
 210 }
 211
 212 static const struct lwtunnel_encap_ops mpls_iptun_ops = {
 213         .build_state = mpls_build_state,
 214         .xmit = mpls_xmit,
 215         .fill_encap = mpls_fill_encap_info,
 216         .get_encap_size = mpls_encap_nlsize,
 217         .cmp_encap = mpls_encap_cmp,
 218 };
 219
 220 static int __init mpls_iptunnel_init(void)
 221 {
 222         return lwtunnel_encap_add_ops(&mpls_iptun_ops, LWTUNNEL_ENCAP_MPLS);
 223 }
 224 module_init(mpls_iptunnel_init);
 225
 226 static void __exit mpls_iptunnel_exit(void)
 227 {
 228         lwtunnel_encap_del_ops(&mpls_iptun_ops, LWTUNNEL_ENCAP_MPLS);
 229 }
 230 module_exit(mpls_iptunnel_exit);
 231
 232 MODULE_ALIAS_RTNL_LWT(MPLS);
 233 MODULE_DESCRIPTION("MultiProtocol Label Switching IP Tunnels");
 234 MODULE_LICENSE("GPL v2");