]> asedeno.scripts.mit.edu Git - linux.git/blobdiff - net/ipv4/ip_tunnel_core.c
Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
[linux.git] / net / ipv4 / ip_tunnel_core.c
index 1ec9d9419c34b999e8af983a996dfc1372f815c6..47f8b947eef1b763d9d9d8f487595a1ddb94739f 100644 (file)
@@ -35,6 +35,8 @@
 #include <net/rtnetlink.h>
 #include <net/dst_metadata.h>
 #include <net/geneve.h>
+#include <net/vxlan.h>
+#include <net/erspan.h>
 
 const struct ip_tunnel_encap_ops __rcu *
                iptun_encaps[MAX_IPTUN_ENCAP_OPS] __read_mostly;
@@ -213,6 +215,7 @@ void ip_tunnel_get_stats64(struct net_device *dev,
 EXPORT_SYMBOL_GPL(ip_tunnel_get_stats64);
 
 static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
+       [LWTUNNEL_IP_UNSPEC]    = { .strict_start_type = LWTUNNEL_IP_OPTS },
        [LWTUNNEL_IP_ID]        = { .type = NLA_U64 },
        [LWTUNNEL_IP_DST]       = { .type = NLA_U32 },
        [LWTUNNEL_IP_SRC]       = { .type = NLA_U32 },
@@ -224,6 +227,8 @@ static const struct nla_policy ip_tun_policy[LWTUNNEL_IP_MAX + 1] = {
 
 static const struct nla_policy ip_opts_policy[LWTUNNEL_IP_OPTS_MAX + 1] = {
        [LWTUNNEL_IP_OPTS_GENEVE]       = { .type = NLA_NESTED },
+       [LWTUNNEL_IP_OPTS_VXLAN]        = { .type = NLA_NESTED },
+       [LWTUNNEL_IP_OPTS_ERSPAN]       = { .type = NLA_NESTED },
 };
 
 static const struct nla_policy
@@ -233,15 +238,28 @@ geneve_opt_policy[LWTUNNEL_IP_OPT_GENEVE_MAX + 1] = {
        [LWTUNNEL_IP_OPT_GENEVE_DATA]   = { .type = NLA_BINARY, .len = 128 },
 };
 
+static const struct nla_policy
+vxlan_opt_policy[LWTUNNEL_IP_OPT_VXLAN_MAX + 1] = {
+       [LWTUNNEL_IP_OPT_VXLAN_GBP]     = { .type = NLA_U32 },
+};
+
+static const struct nla_policy
+erspan_opt_policy[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1] = {
+       [LWTUNNEL_IP_OPT_ERSPAN_VER]    = { .type = NLA_U8 },
+       [LWTUNNEL_IP_OPT_ERSPAN_INDEX]  = { .type = NLA_U32 },
+       [LWTUNNEL_IP_OPT_ERSPAN_DIR]    = { .type = NLA_U8 },
+       [LWTUNNEL_IP_OPT_ERSPAN_HWID]   = { .type = NLA_U8 },
+};
+
 static int ip_tun_parse_opts_geneve(struct nlattr *attr,
-                                   struct ip_tunnel_info *info,
+                                   struct ip_tunnel_info *info, int opts_len,
                                    struct netlink_ext_ack *extack)
 {
        struct nlattr *tb[LWTUNNEL_IP_OPT_GENEVE_MAX + 1];
        int data_len, err;
 
-       err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_OPT_GENEVE_MAX,
-                                         attr, geneve_opt_policy, extack);
+       err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_GENEVE_MAX, attr,
+                              geneve_opt_policy, extack);
        if (err)
                return err;
 
@@ -256,7 +274,7 @@ static int ip_tun_parse_opts_geneve(struct nlattr *attr,
                return -EINVAL;
 
        if (info) {
-               struct geneve_opt *opt = ip_tunnel_info_opts(info);
+               struct geneve_opt *opt = ip_tunnel_info_opts(info) + opts_len;
 
                memcpy(opt->opt_data, nla_data(attr), data_len);
                opt->length = data_len / 4;
@@ -270,27 +288,136 @@ static int ip_tun_parse_opts_geneve(struct nlattr *attr,
        return sizeof(struct geneve_opt) + data_len;
 }
 
+static int ip_tun_parse_opts_vxlan(struct nlattr *attr,
+                                  struct ip_tunnel_info *info, int opts_len,
+                                  struct netlink_ext_ack *extack)
+{
+       struct nlattr *tb[LWTUNNEL_IP_OPT_VXLAN_MAX + 1];
+       int err;
+
+       err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_VXLAN_MAX, attr,
+                              vxlan_opt_policy, extack);
+       if (err)
+               return err;
+
+       if (!tb[LWTUNNEL_IP_OPT_VXLAN_GBP])
+               return -EINVAL;
+
+       if (info) {
+               struct vxlan_metadata *md =
+                       ip_tunnel_info_opts(info) + opts_len;
+
+               attr = tb[LWTUNNEL_IP_OPT_VXLAN_GBP];
+               md->gbp = nla_get_u32(attr);
+               info->key.tun_flags |= TUNNEL_VXLAN_OPT;
+       }
+
+       return sizeof(struct vxlan_metadata);
+}
+
+static int ip_tun_parse_opts_erspan(struct nlattr *attr,
+                                   struct ip_tunnel_info *info, int opts_len,
+                                   struct netlink_ext_ack *extack)
+{
+       struct nlattr *tb[LWTUNNEL_IP_OPT_ERSPAN_MAX + 1];
+       int err;
+       u8 ver;
+
+       err = nla_parse_nested(tb, LWTUNNEL_IP_OPT_ERSPAN_MAX, attr,
+                              erspan_opt_policy, extack);
+       if (err)
+               return err;
+
+       if (!tb[LWTUNNEL_IP_OPT_ERSPAN_VER])
+               return -EINVAL;
+
+       ver = nla_get_u8(tb[LWTUNNEL_IP_OPT_ERSPAN_VER]);
+       if (ver == 1) {
+               if (!tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX])
+                       return -EINVAL;
+       } else if (ver == 2) {
+               if (!tb[LWTUNNEL_IP_OPT_ERSPAN_DIR] ||
+                   !tb[LWTUNNEL_IP_OPT_ERSPAN_HWID])
+                       return -EINVAL;
+       } else {
+               return -EINVAL;
+       }
+
+       if (info) {
+               struct erspan_metadata *md =
+                       ip_tunnel_info_opts(info) + opts_len;
+
+               md->version = ver;
+               if (ver == 1) {
+                       attr = tb[LWTUNNEL_IP_OPT_ERSPAN_INDEX];
+                       md->u.index = nla_get_be32(attr);
+               } else {
+                       attr = tb[LWTUNNEL_IP_OPT_ERSPAN_DIR];
+                       md->u.md2.dir = nla_get_u8(attr);
+                       attr = tb[LWTUNNEL_IP_OPT_ERSPAN_HWID];
+                       set_hwid(&md->u.md2, nla_get_u8(attr));
+               }
+
+               info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
+       }
+
+       return sizeof(struct erspan_metadata);
+}
+
 static int ip_tun_parse_opts(struct nlattr *attr, struct ip_tunnel_info *info,
                             struct netlink_ext_ack *extack)
 {
-       struct nlattr *tb[LWTUNNEL_IP_OPTS_MAX + 1];
-       int err;
+       int err, rem, opt_len, opts_len = 0, type = 0;
+       struct nlattr *nla;
 
        if (!attr)
                return 0;
 
-       err = nla_parse_nested_deprecated(tb, LWTUNNEL_IP_OPTS_MAX, attr,
-                                         ip_opts_policy, extack);
+       err = nla_validate(nla_data(attr), nla_len(attr), LWTUNNEL_IP_OPTS_MAX,
+                          ip_opts_policy, extack);
        if (err)
                return err;
 
-       if (tb[LWTUNNEL_IP_OPTS_GENEVE])
-               err = ip_tun_parse_opts_geneve(tb[LWTUNNEL_IP_OPTS_GENEVE],
-                                              info, extack);
-       else
-               err = -EINVAL;
+       nla_for_each_attr(nla, nla_data(attr), nla_len(attr), rem) {
+               switch (nla_type(nla)) {
+               case LWTUNNEL_IP_OPTS_GENEVE:
+                       if (type && type != TUNNEL_GENEVE_OPT)
+                               return -EINVAL;
+                       opt_len = ip_tun_parse_opts_geneve(nla, info, opts_len,
+                                                          extack);
+                       if (opt_len < 0)
+                               return opt_len;
+                       opts_len += opt_len;
+                       if (opts_len > IP_TUNNEL_OPTS_MAX)
+                               return -EINVAL;
+                       type = TUNNEL_GENEVE_OPT;
+                       break;
+               case LWTUNNEL_IP_OPTS_VXLAN:
+                       if (type)
+                               return -EINVAL;
+                       opt_len = ip_tun_parse_opts_vxlan(nla, info, opts_len,
+                                                         extack);
+                       if (opt_len < 0)
+                               return opt_len;
+                       opts_len += opt_len;
+                       type = TUNNEL_VXLAN_OPT;
+                       break;
+               case LWTUNNEL_IP_OPTS_ERSPAN:
+                       if (type)
+                               return -EINVAL;
+                       opt_len = ip_tun_parse_opts_erspan(nla, info, opts_len,
+                                                          extack);
+                       if (opt_len < 0)
+                               return opt_len;
+                       opts_len += opt_len;
+                       type = TUNNEL_ERSPAN_OPT;
+                       break;
+               default:
+                       return -EINVAL;
+               }
+       }
 
-       return err;
+       return opts_len;
 }
 
 static int ip_tun_get_optlen(struct nlattr *attr,
@@ -362,7 +489,9 @@ static int ip_tun_build_state(struct nlattr *attr,
                tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP_TOS]);
 
        if (tb[LWTUNNEL_IP_FLAGS])
-               tun_info->key.tun_flags |= nla_get_be16(tb[LWTUNNEL_IP_FLAGS]);
+               tun_info->key.tun_flags |=
+                               (nla_get_be16(tb[LWTUNNEL_IP_FLAGS]) &
+                                ~TUNNEL_OPTIONS_PRESENT);
 
        tun_info->mode = IP_TUNNEL_INFO_TX;
        tun_info->options_len = opt_len;
@@ -386,16 +515,41 @@ static int ip_tun_fill_encap_opts_geneve(struct sk_buff *skb,
 {
        struct geneve_opt *opt;
        struct nlattr *nest;
+       int offset = 0;
 
        nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_GENEVE);
        if (!nest)
                return -ENOMEM;
 
-       opt = ip_tunnel_info_opts(tun_info);
-       if (nla_put_be16(skb, LWTUNNEL_IP_OPT_GENEVE_CLASS, opt->opt_class) ||
-           nla_put_u8(skb, LWTUNNEL_IP_OPT_GENEVE_TYPE, opt->type) ||
-           nla_put(skb, LWTUNNEL_IP_OPT_GENEVE_DATA, opt->length * 4,
-                   opt->opt_data)) {
+       while (tun_info->options_len > offset) {
+               opt = ip_tunnel_info_opts(tun_info) + offset;
+               if (nla_put_be16(skb, LWTUNNEL_IP_OPT_GENEVE_CLASS,
+                                opt->opt_class) ||
+                   nla_put_u8(skb, LWTUNNEL_IP_OPT_GENEVE_TYPE, opt->type) ||
+                   nla_put(skb, LWTUNNEL_IP_OPT_GENEVE_DATA, opt->length * 4,
+                           opt->opt_data)) {
+                       nla_nest_cancel(skb, nest);
+                       return -ENOMEM;
+               }
+               offset += sizeof(*opt) + opt->length * 4;
+       }
+
+       nla_nest_end(skb, nest);
+       return 0;
+}
+
+static int ip_tun_fill_encap_opts_vxlan(struct sk_buff *skb,
+                                       struct ip_tunnel_info *tun_info)
+{
+       struct vxlan_metadata *md;
+       struct nlattr *nest;
+
+       nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_VXLAN);
+       if (!nest)
+               return -ENOMEM;
+
+       md = ip_tunnel_info_opts(tun_info);
+       if (nla_put_u32(skb, LWTUNNEL_IP_OPT_VXLAN_GBP, md->gbp)) {
                nla_nest_cancel(skb, nest);
                return -ENOMEM;
        }
@@ -404,13 +558,44 @@ static int ip_tun_fill_encap_opts_geneve(struct sk_buff *skb,
        return 0;
 }
 
+static int ip_tun_fill_encap_opts_erspan(struct sk_buff *skb,
+                                        struct ip_tunnel_info *tun_info)
+{
+       struct erspan_metadata *md;
+       struct nlattr *nest;
+
+       nest = nla_nest_start_noflag(skb, LWTUNNEL_IP_OPTS_ERSPAN);
+       if (!nest)
+               return -ENOMEM;
+
+       md = ip_tunnel_info_opts(tun_info);
+       if (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_VER, md->version))
+               goto err;
+
+       if (md->version == 1 &&
+           nla_put_be32(skb, LWTUNNEL_IP_OPT_ERSPAN_INDEX, md->u.index))
+               goto err;
+
+       if (md->version == 2 &&
+           (nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_DIR, md->u.md2.dir) ||
+            nla_put_u8(skb, LWTUNNEL_IP_OPT_ERSPAN_HWID,
+                       get_hwid(&md->u.md2))))
+               goto err;
+
+       nla_nest_end(skb, nest);
+       return 0;
+err:
+       nla_nest_cancel(skb, nest);
+       return -ENOMEM;
+}
+
 static int ip_tun_fill_encap_opts(struct sk_buff *skb, int type,
                                  struct ip_tunnel_info *tun_info)
 {
        struct nlattr *nest;
        int err = 0;
 
-       if (!(tun_info->key.tun_flags & TUNNEL_GENEVE_OPT))
+       if (!(tun_info->key.tun_flags & TUNNEL_OPTIONS_PRESENT))
                return 0;
 
        nest = nla_nest_start_noflag(skb, type);
@@ -419,6 +604,10 @@ static int ip_tun_fill_encap_opts(struct sk_buff *skb, int type,
 
        if (tun_info->key.tun_flags & TUNNEL_GENEVE_OPT)
                err = ip_tun_fill_encap_opts_geneve(skb, tun_info);
+       else if (tun_info->key.tun_flags & TUNNEL_VXLAN_OPT)
+               err = ip_tun_fill_encap_opts_vxlan(skb, tun_info);
+       else if (tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)
+               err = ip_tun_fill_encap_opts_erspan(skb, tun_info);
 
        if (err) {
                nla_nest_cancel(skb, nest);
@@ -451,18 +640,36 @@ static int ip_tun_opts_nlsize(struct ip_tunnel_info *info)
 {
        int opt_len;
 
-       if (!(info->key.tun_flags & TUNNEL_GENEVE_OPT))
+       if (!(info->key.tun_flags & TUNNEL_OPTIONS_PRESENT))
                return 0;
 
        opt_len = nla_total_size(0);            /* LWTUNNEL_IP_OPTS */
        if (info->key.tun_flags & TUNNEL_GENEVE_OPT) {
-               struct geneve_opt *opt = ip_tunnel_info_opts(info);
-
-               opt_len += nla_total_size(0)    /* LWTUNNEL_IP_OPTS_GENEVE */
-                          + nla_total_size(2)  /* OPT_GENEVE_CLASS */
-                          + nla_total_size(1)  /* OPT_GENEVE_TYPE */
-                          + nla_total_size(opt->length * 4);
-                                               /* OPT_GENEVE_DATA */
+               struct geneve_opt *opt;
+               int offset = 0;
+
+               opt_len += nla_total_size(0);   /* LWTUNNEL_IP_OPTS_GENEVE */
+               while (info->options_len > offset) {
+                       opt = ip_tunnel_info_opts(info) + offset;
+                       opt_len += nla_total_size(2)    /* OPT_GENEVE_CLASS */
+                                  + nla_total_size(1)  /* OPT_GENEVE_TYPE */
+                                  + nla_total_size(opt->length * 4);
+                                                       /* OPT_GENEVE_DATA */
+                       offset += sizeof(*opt) + opt->length * 4;
+               }
+       } else if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
+               opt_len += nla_total_size(0)    /* LWTUNNEL_IP_OPTS_VXLAN */
+                          + nla_total_size(4); /* OPT_VXLAN_GBP */
+       } else if (info->key.tun_flags & TUNNEL_ERSPAN_OPT) {
+               struct erspan_metadata *md = ip_tunnel_info_opts(info);
+
+               opt_len += nla_total_size(0)    /* LWTUNNEL_IP_OPTS_ERSPAN */
+                          + nla_total_size(1)  /* OPT_ERSPAN_VER */
+                          + (md->version == 1 ? nla_total_size(4)
+                                               /* OPT_ERSPAN_INDEX (v1) */
+                                              : nla_total_size(1) +
+                                                nla_total_size(1));
+                                               /* OPT_ERSPAN_DIR + HWID (v2) */
        }
 
        return opt_len;
@@ -502,12 +709,14 @@ static const struct lwtunnel_encap_ops ip_tun_lwt_ops = {
 };
 
 static const struct nla_policy ip6_tun_policy[LWTUNNEL_IP6_MAX + 1] = {
+       [LWTUNNEL_IP6_UNSPEC]   = { .strict_start_type = LWTUNNEL_IP6_OPTS },
        [LWTUNNEL_IP6_ID]               = { .type = NLA_U64 },
        [LWTUNNEL_IP6_DST]              = { .len = sizeof(struct in6_addr) },
        [LWTUNNEL_IP6_SRC]              = { .len = sizeof(struct in6_addr) },
        [LWTUNNEL_IP6_HOPLIMIT]         = { .type = NLA_U8 },
        [LWTUNNEL_IP6_TC]               = { .type = NLA_U8 },
        [LWTUNNEL_IP6_FLAGS]            = { .type = NLA_U16 },
+       [LWTUNNEL_IP6_OPTS]             = { .type = NLA_NESTED },
 };
 
 static int ip6_tun_build_state(struct nlattr *attr,
@@ -559,7 +768,9 @@ static int ip6_tun_build_state(struct nlattr *attr,
                tun_info->key.tos = nla_get_u8(tb[LWTUNNEL_IP6_TC]);
 
        if (tb[LWTUNNEL_IP6_FLAGS])
-               tun_info->key.tun_flags |= nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]);
+               tun_info->key.tun_flags |=
+                               (nla_get_be16(tb[LWTUNNEL_IP6_FLAGS]) &
+                                ~TUNNEL_OPTIONS_PRESENT);
 
        tun_info->mode = IP_TUNNEL_INFO_TX | IP_TUNNEL_INFO_IPV6;
        tun_info->options_len = opt_len;