ip_gre: Add support to collect tunnel metadata.
authorPravin B Shelar <pshelar@nicira.com>
Sat, 8 Aug 2015 06:51:42 +0000 (23:51 -0700)
committerDavid S. Miller <davem@davemloft.net>
Mon, 10 Aug 2015 21:03:54 +0000 (14:03 -0700)
Following patch create new tunnel flag which enable
tunnel metadata collection on given device.

Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/net/ip_tunnels.h
include/uapi/linux/if_tunnel.h
net/ipv4/ip_gre.c
net/ipv4/ip_tunnel.c
net/ipv4/ipip.c
net/ipv6/sit.c

index 47984415f5d1e7758d6a1c8b0feaeb55f902a7dd..984dbfa15e13f199682825fa10e0f295be9102ad 100644 (file)
@@ -82,6 +82,8 @@ struct ip_tunnel_dst {
        __be32                           saddr;
 };
 
+struct metadata_dst;
+
 struct ip_tunnel {
        struct ip_tunnel __rcu  *next;
        struct hlist_node hash_node;
@@ -115,6 +117,7 @@ struct ip_tunnel {
        unsigned int            prl_count;      /* # of entries in PRL */
        int                     ip_tnl_net_id;
        struct gro_cells        gro_cells;
+       bool                    collect_md;
 };
 
 #define TUNNEL_CSUM            __cpu_to_be16(0x01)
@@ -149,6 +152,7 @@ struct tnl_ptk_info {
 struct ip_tunnel_net {
        struct net_device *fb_tunnel_dev;
        struct hlist_head tunnels[IP_TNL_HASH_SIZE];
+       struct ip_tunnel __rcu *collect_md_tun;
 };
 
 struct ip_tunnel_encap_ops {
@@ -235,7 +239,8 @@ struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
                                   __be32 key);
 
 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
-                 const struct tnl_ptk_info *tpi, bool log_ecn_error);
+                 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
+                 bool log_ecn_error);
 int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
                         struct ip_tunnel_parm *p);
 int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
index bd3cc11a431f9e5bb01f18d84d3f388915043367..af4de90ba27d18c76944b062d1db4e4267dcd844 100644 (file)
@@ -112,6 +112,7 @@ enum {
        IFLA_GRE_ENCAP_FLAGS,
        IFLA_GRE_ENCAP_SPORT,
        IFLA_GRE_ENCAP_DPORT,
+       IFLA_GRE_COLLECT_METADATA,
        __IFLA_GRE_MAX,
 };
 
index 5fd706473c733402c9aad9c6c30466549de8c54d..554a760c2cd071dba453df301190c6c50c7ee76e 100644 (file)
@@ -25,6 +25,7 @@
 #include <linux/udp.h>
 #include <linux/if_arp.h>
 #include <linux/mroute.h>
+#include <linux/if_vlan.h>
 #include <linux/init.h>
 #include <linux/in6.h>
 #include <linux/inetdevice.h>
@@ -47,6 +48,7 @@
 #include <net/netns/generic.h>
 #include <net/rtnetlink.h>
 #include <net/gre.h>
+#include <net/dst_metadata.h>
 
 #if IS_ENABLED(CONFIG_IPV6)
 #include <net/ipv6.h>
@@ -200,9 +202,29 @@ static int ipgre_err(struct sk_buff *skb, u32 info,
        return PACKET_RCVD;
 }
 
+static __be64 key_to_tunnel_id(__be32 key)
+{
+#ifdef __BIG_ENDIAN
+       return (__force __be64)((__force u32)key);
+#else
+       return (__force __be64)((__force u64)key << 32);
+#endif
+}
+
+/* Returns the least-significant 32 bits of a __be64. */
+static __be32 tunnel_id_to_key(__be64 x)
+{
+#ifdef __BIG_ENDIAN
+       return (__force __be32)x;
+#else
+       return (__force __be32)((__force u64)x >> 32);
+#endif
+}
+
 static int ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
 {
        struct net *net = dev_net(skb->dev);
+       struct metadata_dst *tun_dst = NULL;
        struct ip_tunnel_net *itn;
        const struct iphdr *iph;
        struct ip_tunnel *tunnel;
@@ -218,40 +240,162 @@ static int ipgre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
 
        if (tunnel) {
                skb_pop_mac_header(skb);
-               ip_tunnel_rcv(tunnel, skb, tpi, log_ecn_error);
+               if (tunnel->collect_md) {
+                       struct ip_tunnel_info *info;
+
+                       tun_dst = metadata_dst_alloc(0, GFP_ATOMIC);
+                       if (!tun_dst)
+                               return PACKET_REJECT;
+
+                       info = &tun_dst->u.tun_info;
+                       info->key.ipv4_src = iph->saddr;
+                       info->key.ipv4_dst = iph->daddr;
+                       info->key.ipv4_tos = iph->tos;
+                       info->key.ipv4_ttl = iph->ttl;
+
+                       info->mode = IP_TUNNEL_INFO_RX;
+                       info->key.tun_flags = tpi->flags &
+                                             (TUNNEL_CSUM | TUNNEL_KEY);
+                       info->key.tun_id = key_to_tunnel_id(tpi->key);
+
+                       info->key.tp_src = 0;
+                       info->key.tp_dst = 0;
+               }
+
+               ip_tunnel_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
                return PACKET_RCVD;
        }
        return PACKET_REJECT;
 }
 
+static void build_header(struct sk_buff *skb, int hdr_len, __be16 flags,
+                        __be16 proto, __be32 key, __be32 seq)
+{
+       struct gre_base_hdr *greh;
+
+       skb_push(skb, hdr_len);
+
+       skb_reset_transport_header(skb);
+       greh = (struct gre_base_hdr *)skb->data;
+       greh->flags = tnl_flags_to_gre_flags(flags);
+       greh->protocol = proto;
+
+       if (flags & (TUNNEL_KEY | TUNNEL_CSUM | TUNNEL_SEQ)) {
+               __be32 *ptr = (__be32 *)(((u8 *)greh) + hdr_len - 4);
+
+               if (flags & TUNNEL_SEQ) {
+                       *ptr = seq;
+                       ptr--;
+               }
+               if (flags & TUNNEL_KEY) {
+                       *ptr = key;
+                       ptr--;
+               }
+               if (flags & TUNNEL_CSUM &&
+                   !(skb_shinfo(skb)->gso_type &
+                     (SKB_GSO_GRE | SKB_GSO_GRE_CSUM))) {
+                       *ptr = 0;
+                       *(__sum16 *)ptr = csum_fold(skb_checksum(skb, 0,
+                                                                skb->len, 0));
+               }
+       }
+}
+
 static void __gre_xmit(struct sk_buff *skb, struct net_device *dev,
                       const struct iphdr *tnl_params,
                       __be16 proto)
 {
        struct ip_tunnel *tunnel = netdev_priv(dev);
-       struct tnl_ptk_info tpi;
 
-       tpi.flags = tunnel->parms.o_flags;
-       tpi.proto = proto;
-       tpi.key = tunnel->parms.o_key;
        if (tunnel->parms.o_flags & TUNNEL_SEQ)
                tunnel->o_seqno++;
-       tpi.seq = htonl(tunnel->o_seqno);
 
        /* Push GRE header. */
-       gre_build_header(skb, &tpi, tunnel->tun_hlen);
-
-       skb_set_inner_protocol(skb, tpi.proto);
+       build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
+                    proto, tunnel->parms.o_key, htonl(tunnel->o_seqno));
 
+       skb_set_inner_protocol(skb, proto);
        ip_tunnel_xmit(skb, dev, tnl_params, tnl_params->protocol);
 }
 
+static void gre_fb_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+       struct ip_tunnel_info *tun_info;
+       struct net *net = dev_net(dev);
+       const struct ip_tunnel_key *key;
+       struct flowi4 fl;
+       struct rtable *rt;
+       int min_headroom;
+       int tunnel_hlen;
+       __be16 df, flags;
+       int err;
+
+       tun_info = skb_tunnel_info(skb, AF_INET);
+       if (unlikely(!tun_info || tun_info->mode != IP_TUNNEL_INFO_TX))
+               goto err_free_skb;
+
+       key = &tun_info->key;
+       memset(&fl, 0, sizeof(fl));
+       fl.daddr = key->ipv4_dst;
+       fl.saddr = key->ipv4_src;
+       fl.flowi4_tos = RT_TOS(key->ipv4_tos);
+       fl.flowi4_mark = skb->mark;
+       fl.flowi4_proto = IPPROTO_GRE;
+
+       rt = ip_route_output_key(net, &fl);
+       if (IS_ERR(rt))
+               goto err_free_skb;
+
+       tunnel_hlen = ip_gre_calc_hlen(key->tun_flags);
+
+       min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
+                       + tunnel_hlen + sizeof(struct iphdr);
+       if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
+               int head_delta = SKB_DATA_ALIGN(min_headroom -
+                                               skb_headroom(skb) +
+                                               16);
+               err = pskb_expand_head(skb, max_t(int, head_delta, 0),
+                                      0, GFP_ATOMIC);
+               if (unlikely(err))
+                       goto err_free_rt;
+       }
+
+       /* Push Tunnel header. */
+       skb = gre_handle_offloads(skb, !!(tun_info->key.tun_flags & TUNNEL_CSUM));
+       if (IS_ERR(skb)) {
+               skb = NULL;
+               goto err_free_rt;
+       }
+
+       flags = tun_info->key.tun_flags & (TUNNEL_CSUM | TUNNEL_KEY);
+       build_header(skb, tunnel_hlen, flags, htons(ETH_P_TEB),
+                    tunnel_id_to_key(tun_info->key.tun_id), 0);
+
+       df = key->tun_flags & TUNNEL_DONT_FRAGMENT ?  htons(IP_DF) : 0;
+       err = iptunnel_xmit(skb->sk, rt, skb, fl.saddr,
+                           key->ipv4_dst, IPPROTO_GRE,
+                           key->ipv4_tos, key->ipv4_ttl, df, false);
+       iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
+       return;
+
+err_free_rt:
+       ip_rt_put(rt);
+err_free_skb:
+       kfree_skb(skb);
+       dev->stats.tx_dropped++;
+}
+
 static netdev_tx_t ipgre_xmit(struct sk_buff *skb,
                              struct net_device *dev)
 {
        struct ip_tunnel *tunnel = netdev_priv(dev);
        const struct iphdr *tnl_params;
 
+       if (tunnel->collect_md) {
+               gre_fb_xmit(skb, dev);
+               return NETDEV_TX_OK;
+       }
+
        if (dev->header_ops) {
                /* Need space for new headers */
                if (skb_cow_head(skb, dev->needed_headroom -
@@ -277,7 +421,6 @@ static netdev_tx_t ipgre_xmit(struct sk_buff *skb,
                goto out;
 
        __gre_xmit(skb, dev, tnl_params, skb->protocol);
-
        return NETDEV_TX_OK;
 
 free_skb:
@@ -292,6 +435,11 @@ static netdev_tx_t gre_tap_xmit(struct sk_buff *skb,
 {
        struct ip_tunnel *tunnel = netdev_priv(dev);
 
+       if (tunnel->collect_md) {
+               gre_fb_xmit(skb, dev);
+               return NETDEV_TX_OK;
+       }
+
        skb = gre_handle_offloads(skb, !!(tunnel->parms.o_flags&TUNNEL_CSUM));
        if (IS_ERR(skb))
                goto out;
@@ -300,7 +448,6 @@ static netdev_tx_t gre_tap_xmit(struct sk_buff *skb,
                goto free_skb;
 
        __gre_xmit(skb, dev, &tunnel->parms.iph, htons(ETH_P_TEB));
-
        return NETDEV_TX_OK;
 
 free_skb:
@@ -596,8 +743,10 @@ out:
        return ipgre_tunnel_validate(tb, data);
 }
 
-static void ipgre_netlink_parms(struct nlattr *data[], struct nlattr *tb[],
-                              struct ip_tunnel_parm *parms)
+static void ipgre_netlink_parms(struct net_device *dev,
+                               struct nlattr *data[],
+                               struct nlattr *tb[],
+                               struct ip_tunnel_parm *parms)
 {
        memset(parms, 0, sizeof(*parms));
 
@@ -635,6 +784,12 @@ static void ipgre_netlink_parms(struct nlattr *data[], struct nlattr *tb[],
 
        if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC]))
                parms->iph.frag_off = htons(IP_DF);
+
+       if (data[IFLA_GRE_COLLECT_METADATA]) {
+               struct ip_tunnel *t = netdev_priv(dev);
+
+               t->collect_md = true;
+       }
 }
 
 /* This function returns true when ENCAP attributes are present in the nl msg */
@@ -712,7 +867,7 @@ static int ipgre_newlink(struct net *src_net, struct net_device *dev,
                        return err;
        }
 
-       ipgre_netlink_parms(data, tb, &p);
+       ipgre_netlink_parms(dev, data, tb, &p);
        return ip_tunnel_newlink(dev, tb, &p);
 }
 
@@ -730,7 +885,7 @@ static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[],
                        return err;
        }
 
-       ipgre_netlink_parms(data, tb, &p);
+       ipgre_netlink_parms(dev, data, tb, &p);
        return ip_tunnel_changelink(dev, tb, &p);
 }
 
@@ -765,6 +920,8 @@ static size_t ipgre_get_size(const struct net_device *dev)
                nla_total_size(2) +
                /* IFLA_GRE_ENCAP_DPORT */
                nla_total_size(2) +
+               /* IFLA_GRE_COLLECT_METADATA */
+               nla_total_size(0) +
                0;
 }
 
@@ -796,6 +953,11 @@ static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev)
                        t->encap.flags))
                goto nla_put_failure;
 
+       if (t->collect_md) {
+               if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
+                       goto nla_put_failure;
+       }
+
        return 0;
 
 nla_put_failure:
@@ -817,6 +979,7 @@ static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = {
        [IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
        [IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
        [IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
+       [IFLA_GRE_COLLECT_METADATA]     = { .type = NLA_FLAG },
 };
 
 static struct rtnl_link_ops ipgre_link_ops __read_mostly = {
@@ -851,7 +1014,7 @@ static struct rtnl_link_ops ipgre_tap_ops __read_mostly = {
 
 static int __net_init ipgre_tap_init_net(struct net *net)
 {
-       return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, NULL);
+       return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, "gretap0");
 }
 
 static void __net_exit ipgre_tap_exit_net(struct net *net)
index 626d9e56a6bd2671611f3dde4f8090b76645301c..cbb51f3fac06c66e76204957b3243f50a58384b4 100644 (file)
@@ -230,10 +230,13 @@ skip_key_lookup:
        if (cand)
                return cand;
 
+       t = rcu_dereference(itn->collect_md_tun);
+       if (t)
+               return t;
+
        if (itn->fb_tunnel_dev && itn->fb_tunnel_dev->flags & IFF_UP)
                return netdev_priv(itn->fb_tunnel_dev);
 
-
        return NULL;
 }
 EXPORT_SYMBOL_GPL(ip_tunnel_lookup);
@@ -261,11 +264,15 @@ static void ip_tunnel_add(struct ip_tunnel_net *itn, struct ip_tunnel *t)
 {
        struct hlist_head *head = ip_bucket(itn, &t->parms);
 
+       if (t->collect_md)
+               rcu_assign_pointer(itn->collect_md_tun, t);
        hlist_add_head_rcu(&t->hash_node, head);
 }
 
-static void ip_tunnel_del(struct ip_tunnel *t)
+static void ip_tunnel_del(struct ip_tunnel_net *itn, struct ip_tunnel *t)
 {
+       if (t->collect_md)
+               rcu_assign_pointer(itn->collect_md_tun, NULL);
        hlist_del_init_rcu(&t->hash_node);
 }
 
@@ -419,7 +426,8 @@ static struct ip_tunnel *ip_tunnel_create(struct net *net,
 }
 
 int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
-                 const struct tnl_ptk_info *tpi, bool log_ecn_error)
+                 const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
+                 bool log_ecn_error)
 {
        struct pcpu_sw_netstats *tstats;
        const struct iphdr *iph = ip_hdr(skb);
@@ -478,6 +486,9 @@ int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
                skb->dev = tunnel->dev;
        }
 
+       if (tun_dst)
+               skb_dst_set(skb, (struct dst_entry *)tun_dst);
+
        gro_cells_receive(&tunnel->gro_cells, skb);
        return 0;
 
@@ -806,7 +817,7 @@ static void ip_tunnel_update(struct ip_tunnel_net *itn,
                             struct ip_tunnel_parm *p,
                             bool set_mtu)
 {
-       ip_tunnel_del(t);
+       ip_tunnel_del(itn, t);
        t->parms.iph.saddr = p->iph.saddr;
        t->parms.iph.daddr = p->iph.daddr;
        t->parms.i_key = p->i_key;
@@ -967,7 +978,7 @@ void ip_tunnel_dellink(struct net_device *dev, struct list_head *head)
        itn = net_generic(tunnel->net, tunnel->ip_tnl_net_id);
 
        if (itn->fb_tunnel_dev != dev) {
-               ip_tunnel_del(netdev_priv(dev));
+               ip_tunnel_del(itn, netdev_priv(dev));
                unregister_netdevice_queue(dev, head);
        }
 }
@@ -1072,8 +1083,13 @@ int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
        nt = netdev_priv(dev);
        itn = net_generic(net, nt->ip_tnl_net_id);
 
-       if (ip_tunnel_find(itn, p, dev->type))
-               return -EEXIST;
+       if (nt->collect_md) {
+               if (rtnl_dereference(itn->collect_md_tun))
+                       return -EEXIST;
+       } else {
+               if (ip_tunnel_find(itn, p, dev->type))
+                       return -EEXIST;
+       }
 
        nt->net = net;
        nt->parms = *p;
@@ -1089,7 +1105,6 @@ int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
                dev->mtu = mtu;
 
        ip_tunnel_add(itn, nt);
-
 out:
        return err;
 }
@@ -1163,6 +1178,10 @@ int ip_tunnel_init(struct net_device *dev)
        iph->version            = 4;
        iph->ihl                = 5;
 
+       if (tunnel->collect_md) {
+               dev->features |= NETIF_F_NETNS_LOCAL;
+               netif_keep_dst(dev);
+       }
        return 0;
 }
 EXPORT_SYMBOL_GPL(ip_tunnel_init);
@@ -1176,7 +1195,7 @@ void ip_tunnel_uninit(struct net_device *dev)
        itn = net_generic(net, tunnel->ip_tnl_net_id);
        /* fb_tunnel_dev will be unregisted in net-exit call. */
        if (itn->fb_tunnel_dev != dev)
-               ip_tunnel_del(netdev_priv(dev));
+               ip_tunnel_del(itn, netdev_priv(dev));
 
        ip_tunnel_dst_reset_all(tunnel);
 }
index 254238daf58bd9f6e398609b67ca6a0b946600d2..f34c31defafe083fcc9146affff5c39745150d04 100644 (file)
@@ -198,7 +198,7 @@ static int ipip_rcv(struct sk_buff *skb)
                        goto drop;
                if (iptunnel_pull_header(skb, 0, tpi.proto))
                        goto drop;
-               return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
+               return ip_tunnel_rcv(tunnel, skb, &tpi, NULL, log_ecn_error);
        }
 
        return -1;
index ac35a28599be557cac1114cad35e3c47c41436c1..94428fd85b2f9543c0e3015f785306ce203e73dc 100644 (file)
@@ -742,7 +742,7 @@ static int ipip_rcv(struct sk_buff *skb)
                        goto drop;
                if (iptunnel_pull_header(skb, 0, tpi.proto))
                        goto drop;
-               return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
+               return ip_tunnel_rcv(tunnel, skb, &tpi, NULL, log_ecn_error);
        }
 
        return 1;