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         [MPLS_IPTUNNEL_TTL]     = { .type = NLA_U8 },
  33 };
  34
  35 static unsigned int mpls_encap_size(struct mpls_iptunnel_encap *en)
  36 {
  37         /* The size of the layer 2.5 labels to be added for this route */
  38         return en->labels * sizeof(struct mpls_shim_hdr);
  39 }
  40
  41 static int mpls_xmit(struct sk_buff *skb)
  42 {
  43         struct mpls_iptunnel_encap *tun_encap_info;
  44         struct mpls_shim_hdr *hdr;
  45         struct net_device *out_dev;
  46         unsigned int hh_len;
  47         unsigned int new_header_size;
  48         unsigned int mtu;
  49         struct dst_entry *dst = skb_dst(skb);
  50         struct rtable *rt = NULL;
  51         struct rt6_info *rt6 = NULL;
  52         struct mpls_dev *out_mdev;
  53         struct net *net;
  54         int err = 0;
  55         bool bos;
  56         int i;
  57         unsigned int ttl;
  58
  59         /* Find the output device */
  60         out_dev = dst->dev;
  61         net = dev_net(out_dev);
  62
  63         skb_orphan(skb);
  64
  65         if (!mpls_output_possible(out_dev) ||
  66             !dst->lwtstate || skb_warn_if_lro(skb))
  67                 goto drop;
  68
  69         skb_forward_csum(skb);
  70
  71         tun_encap_info = mpls_lwtunnel_encap(dst->lwtstate);
  72
  73         /* Obtain the ttl using the following set of rules.
  74          *
  75          * LWT ttl propagation setting:
  76          *  - disabled => use default TTL value from LWT
  77          *  - enabled  => use TTL value from IPv4/IPv6 header
  78          *  - default  =>
  79          *   Global ttl propagation setting:
  80          *    - disabled => use default TTL value from global setting
  81          *    - enabled => use TTL value from IPv4/IPv6 header
  82          */
  83         if (dst->ops->family == AF_INET) {
  84                 if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DISABLED)
  85                         ttl = tun_encap_info->default_ttl;
  86                 else if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
  87                          !net->mpls.ip_ttl_propagate)
  88                         ttl = net->mpls.default_ttl;
  89                 else
  90                         ttl = ip_hdr(skb)->ttl;
  91                 rt = (struct rtable *)dst;
  92         } else if (dst->ops->family == AF_INET6) {
  93                 if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DISABLED)
  94                         ttl = tun_encap_info->default_ttl;
  95                 else if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
  96                          !net->mpls.ip_ttl_propagate)
  97                         ttl = net->mpls.default_ttl;
  98                 else
  99                         ttl = ipv6_hdr(skb)->hop_limit;
 100                 rt6 = (struct rt6_info *)dst;
 101         } else {
 102                 goto drop;
 103         }
 104
 105         /* Verify the destination can hold the packet */
 106         new_header_size = mpls_encap_size(tun_encap_info);
 107         mtu = mpls_dev_mtu(out_dev);
 108         if (mpls_pkt_too_big(skb, mtu - new_header_size))
 109                 goto drop;
 110
 111         hh_len = LL_RESERVED_SPACE(out_dev);
 112         if (!out_dev->header_ops)
 113                 hh_len = 0;
 114
 115         /* Ensure there is enough space for the headers in the skb */
 116         if (skb_cow(skb, hh_len + new_header_size))
 117                 goto drop;
 118
 119         skb_set_inner_protocol(skb, skb->protocol);
 120         skb_reset_inner_network_header(skb);
 121
 122         skb_push(skb, new_header_size);
 123
 124         skb_reset_network_header(skb);
 125
 126         skb->dev = out_dev;
 127         skb->protocol = htons(ETH_P_MPLS_UC);
 128
 129         /* Push the new labels */
 130         hdr = mpls_hdr(skb);
 131         bos = true;
 132         for (i = tun_encap_info->labels - 1; i >= 0; i--) {
 133                 hdr[i] = mpls_entry_encode(tun_encap_info->label[i],
 134                                            ttl, 0, bos);
 135                 bos = false;
 136         }
 137
 138         mpls_stats_inc_outucastpkts(out_dev, skb);
 139
 140         if (rt) {
 141                 if (rt->rt_gw_family == AF_INET)
 142                         err = neigh_xmit(NEIGH_ARP_TABLE, out_dev, &rt->rt_gw4,
 143                                          skb);
 144                 else if (rt->rt_gw_family == AF_INET6)
 145                         err = neigh_xmit(NEIGH_ND_TABLE, out_dev, &rt->rt_gw6,
 146                                          skb);
 147         } else if (rt6) {
 148                 if (ipv6_addr_v4mapped(&rt6->rt6i_gateway)) {
 149                         /* 6PE (RFC 4798) */
 150                         err = neigh_xmit(NEIGH_ARP_TABLE, out_dev, &rt6->rt6i_gateway.s6_addr32[3],
 151                                          skb);
 152                 } else
 153                         err = neigh_xmit(NEIGH_ND_TABLE, out_dev, &rt6->rt6i_gateway,
 154                                          skb);
 155         }
 156         if (err)
 157                 net_dbg_ratelimited("%s: packet transmission failed: %d\n",
 158                                     __func__, err);
 159
 160         return LWTUNNEL_XMIT_DONE;
 161
 162 drop:
 163         out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
 164         if (out_mdev)
 165                 MPLS_INC_STATS(out_mdev, tx_errors);
 166         kfree_skb(skb);
 167         return -EINVAL;
 168 }
 169
 170 static int mpls_build_state(struct nlattr *nla,
 171                             unsigned int family, const void *cfg,
 172                             struct lwtunnel_state **ts,
 173                             struct netlink_ext_ack *extack)
 174 {
 175         struct mpls_iptunnel_encap *tun_encap_info;
 176         struct nlattr *tb[MPLS_IPTUNNEL_MAX + 1];
 177         struct lwtunnel_state *newts;
 178         u8 n_labels;
 179         int ret;
 180
 181         ret = nla_parse_nested_deprecated(tb, MPLS_IPTUNNEL_MAX, nla,
 182                                           mpls_iptunnel_policy, extack);
 183         if (ret < 0)
 184                 return ret;
 185
 186         if (!tb[MPLS_IPTUNNEL_DST]) {
 187                 NL_SET_ERR_MSG(extack, "MPLS_IPTUNNEL_DST attribute is missing");
 188                 return -EINVAL;
 189         }
 190
 191         /* determine number of labels */
 192         if (nla_get_labels(tb[MPLS_IPTUNNEL_DST], MAX_NEW_LABELS,
 193                            &n_labels, NULL, extack))
 194                 return -EINVAL;
 195
 196         newts = lwtunnel_state_alloc(struct_size(tun_encap_info, label,
 197                                                  n_labels));
 198         if (!newts)
 199                 return -ENOMEM;
 200
 201         tun_encap_info = mpls_lwtunnel_encap(newts);
 202         ret = nla_get_labels(tb[MPLS_IPTUNNEL_DST], n_labels,
 203                              &tun_encap_info->labels, tun_encap_info->label,
 204                              extack);
 205         if (ret)
 206                 goto errout;
 207
 208         tun_encap_info->ttl_propagate = MPLS_TTL_PROP_DEFAULT;
 209
 210         if (tb[MPLS_IPTUNNEL_TTL]) {
 211                 tun_encap_info->default_ttl = nla_get_u8(tb[MPLS_IPTUNNEL_TTL]);
 212                 /* TTL 0 implies propagate from IP header */
 213                 tun_encap_info->ttl_propagate = tun_encap_info->default_ttl ?
 214                         MPLS_TTL_PROP_DISABLED :
 215                         MPLS_TTL_PROP_ENABLED;
 216         }
 217
 218         newts->type = LWTUNNEL_ENCAP_MPLS;
 219         newts->flags |= LWTUNNEL_STATE_XMIT_REDIRECT;
 220         newts->headroom = mpls_encap_size(tun_encap_info);
 221
 222         *ts = newts;
 223
 224         return 0;
 225
 226 errout:
 227         kfree(newts);
 228         *ts = NULL;
 229
 230         return ret;
 231 }
 232
 233 static int mpls_fill_encap_info(struct sk_buff *skb,
 234                                 struct lwtunnel_state *lwtstate)
 235 {
 236         struct mpls_iptunnel_encap *tun_encap_info;
 237
 238         tun_encap_info = mpls_lwtunnel_encap(lwtstate);
 239
 240         if (nla_put_labels(skb, MPLS_IPTUNNEL_DST, tun_encap_info->labels,
 241                            tun_encap_info->label))
 242                 goto nla_put_failure;
 243
 244         if (tun_encap_info->ttl_propagate != MPLS_TTL_PROP_DEFAULT &&
 245             nla_put_u8(skb, MPLS_IPTUNNEL_TTL, tun_encap_info->default_ttl))
 246                 goto nla_put_failure;
 247
 248         return 0;
 249
 250 nla_put_failure:
 251         return -EMSGSIZE;
 252 }
 253
 254 static int mpls_encap_nlsize(struct lwtunnel_state *lwtstate)
 255 {
 256         struct mpls_iptunnel_encap *tun_encap_info;
 257         int nlsize;
 258
 259         tun_encap_info = mpls_lwtunnel_encap(lwtstate);
 260
 261         nlsize = nla_total_size(tun_encap_info->labels * 4);
 262
 263         if (tun_encap_info->ttl_propagate != MPLS_TTL_PROP_DEFAULT)
 264                 nlsize += nla_total_size(1);
 265
 266         return nlsize;
 267 }
 268
 269 static int mpls_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
 270 {
 271         struct mpls_iptunnel_encap *a_hdr = mpls_lwtunnel_encap(a);
 272         struct mpls_iptunnel_encap *b_hdr = mpls_lwtunnel_encap(b);
 273         int l;
 274
 275         if (a_hdr->labels != b_hdr->labels ||
 276             a_hdr->ttl_propagate != b_hdr->ttl_propagate ||
 277             a_hdr->default_ttl != b_hdr->default_ttl)
 278                 return 1;
 279
 280         for (l = 0; l < a_hdr->labels; l++)
 281                 if (a_hdr->label[l] != b_hdr->label[l])
 282                         return 1;
 283         return 0;
 284 }
 285
 286 static const struct lwtunnel_encap_ops mpls_iptun_ops = {
 287         .build_state = mpls_build_state,
 288         .xmit = mpls_xmit,
 289         .fill_encap = mpls_fill_encap_info,
 290         .get_encap_size = mpls_encap_nlsize,
 291         .cmp_encap = mpls_encap_cmp,
 292         .owner = THIS_MODULE,
 293 };
 294
 295 static int __init mpls_iptunnel_init(void)
 296 {
 297         return lwtunnel_encap_add_ops(&mpls_iptun_ops, LWTUNNEL_ENCAP_MPLS);
 298 }
 299 module_init(mpls_iptunnel_init);
 300
 301 static void __exit mpls_iptunnel_exit(void)
 302 {
 303         lwtunnel_encap_del_ops(&mpls_iptun_ops, LWTUNNEL_ENCAP_MPLS);
 304 }
 305 module_exit(mpls_iptunnel_exit);
 306
 307 MODULE_ALIAS_RTNL_LWT(MPLS);
 308 MODULE_DESCRIPTION("MultiProtocol Label Switching IP Tunnels");
 309 MODULE_LICENSE("GPL v2");