net: UDP tunnel encapsulation module for tunnelling different protocols like MPLS...
authorMartin Varghese <martin.varghese@nokia.com>
Mon, 24 Feb 2020 05:27:50 +0000 (10:57 +0530)
committerDavid S. Miller <davem@davemloft.net>
Mon, 24 Feb 2020 21:31:42 +0000 (13:31 -0800)
The Bareudp tunnel module provides a generic L3 encapsulation
tunnelling module for tunnelling different protocols like MPLS,
IP,NSH etc inside a UDP tunnel.

Signed-off-by: Martin Varghese <martin.varghese@nokia.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Documentation/networking/bareudp.rst [new file with mode: 0644]
Documentation/networking/index.rst
drivers/net/Kconfig
drivers/net/Makefile
drivers/net/bareudp.c [new file with mode: 0644]
include/net/bareudp.h [new file with mode: 0644]
include/net/ipv6.h
include/net/route.h
include/uapi/linux/if_link.h
net/ipv4/route.c
net/ipv6/ip6_output.c

diff --git a/Documentation/networking/bareudp.rst b/Documentation/networking/bareudp.rst
new file mode 100644 (file)
index 0000000..6ee68c1
--- /dev/null
@@ -0,0 +1,34 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+========================================
+Bare UDP Tunnelling Module Documentation
+========================================
+
+There are various L3 encapsulation standards using UDP being discussed to
+leverage the UDP based load balancing capability of different networks.
+MPLSoUDP (__ https://tools.ietf.org/html/rfc7510) is one among them.
+
+The Bareudp tunnel module provides a generic L3 encapsulation tunnelling
+support for tunnelling different L3 protocols like MPLS, IP, NSH etc. inside
+a UDP tunnel.
+
+Usage
+------
+
+1) Device creation & deletion
+
+    a) ip link add dev bareudp0 type bareudp dstport 6635 ethertype 0x8847.
+
+       This creates a bareudp tunnel device which tunnels L3 traffic with ethertype
+       0x8847 (MPLS traffic). The destination port of the UDP header will be set to
+       6635.The device will listen on UDP port 6635 to receive traffic.
+
+    b) ip link delete bareudp0
+
+2) Device Usage
+
+The bareudp device could be used along with OVS or flower filter in TC.
+The OVS or TC flower layer must set the tunnel information in SKB dst field before
+sending packet buffer to the bareudp device for transmission. On reception the
+bareudp device extracts and stores the tunnel information in SKB dst field before
+passing the packet buffer to the network stack.
index d07d9855dcd3e8b30dcdb5decb6a8e93a51a6400..3a83cfb66704ab2c2124dd35c731aa6abdebf228 100644 (file)
@@ -8,6 +8,7 @@ Contents:
 
    netdev-FAQ
    af_xdp
+   bareudp
    batman-adv
    can
    can_ucan_protocol
index 25a8f9387d5ad9dc31c3f03c78ce3ff8f433744a..66e410e58c8ee99a6a5a92ad7246eff4705b7525 100644 (file)
@@ -258,6 +258,19 @@ config GENEVE
          To compile this driver as a module, choose M here: the module
          will be called geneve.
 
+config BAREUDP
+       tristate "Bare UDP Encapsulation"
+       depends on INET
+       depends on IPV6 || !IPV6
+       select NET_UDP_TUNNEL
+       select GRO_CELLS
+       help
+          This adds a bare UDP tunnel module for tunnelling different
+          kinds of traffic like MPLS, IP, etc. inside a UDP tunnel.
+
+          To compile this driver as a module, choose M here: the module
+          will be called bareudp.
+
 config GTP
        tristate "GPRS Tunneling Protocol datapath (GTP-U)"
        depends on INET
index 71b88ffc55876952d3309ba7ccf7fad6c866f43c..65967246f24018ec765014173215532fb36edf7a 100644 (file)
@@ -29,6 +29,7 @@ obj-$(CONFIG_VETH) += veth.o
 obj-$(CONFIG_VIRTIO_NET) += virtio_net.o
 obj-$(CONFIG_VXLAN) += vxlan.o
 obj-$(CONFIG_GENEVE) += geneve.o
+obj-$(CONFIG_BAREUDP) += bareudp.o
 obj-$(CONFIG_GTP) += gtp.o
 obj-$(CONFIG_NLMON) += nlmon.o
 obj-$(CONFIG_NET_VRF) += vrf.o
diff --git a/drivers/net/bareudp.c b/drivers/net/bareudp.c
new file mode 100644 (file)
index 0000000..3251896
--- /dev/null
@@ -0,0 +1,743 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Bareudp: UDP  tunnel encasulation for different Payload types like
+ * MPLS, NSH, IP, etc.
+ * Copyright (c) 2019 Nokia, Inc.
+ * Authors:  Martin Varghese, <martin.varghese@nokia.com>
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/etherdevice.h>
+#include <linux/hash.h>
+#include <net/dst_metadata.h>
+#include <net/gro_cells.h>
+#include <net/rtnetlink.h>
+#include <net/protocol.h>
+#include <net/ip6_tunnel.h>
+#include <net/ip_tunnels.h>
+#include <net/udp_tunnel.h>
+#include <net/bareudp.h>
+
+#define BAREUDP_BASE_HLEN sizeof(struct udphdr)
+#define BAREUDP_IPV4_HLEN (sizeof(struct iphdr) + \
+                          sizeof(struct udphdr))
+#define BAREUDP_IPV6_HLEN (sizeof(struct ipv6hdr) + \
+                          sizeof(struct udphdr))
+
+static bool log_ecn_error = true;
+module_param(log_ecn_error, bool, 0644);
+MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
+
+/* per-network namespace private data for this module */
+
+static unsigned int bareudp_net_id;
+
+struct bareudp_net {
+       struct list_head        bareudp_list;
+};
+
+/* Pseudo network device */
+struct bareudp_dev {
+       struct net         *net;        /* netns for packet i/o */
+       struct net_device  *dev;        /* netdev for bareudp tunnel */
+       __be16             ethertype;
+       __be16             port;
+       u16                sport_min;
+       struct socket      __rcu *sock;
+       struct list_head   next;        /* bareudp node  on namespace list */
+       struct gro_cells   gro_cells;
+};
+
+static int bareudp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
+{
+       struct metadata_dst *tun_dst = NULL;
+       struct pcpu_sw_netstats *stats;
+       struct bareudp_dev *bareudp;
+       unsigned short family;
+       unsigned int len;
+       __be16 proto;
+       void *oiph;
+       int err;
+
+       bareudp = rcu_dereference_sk_user_data(sk);
+       if (!bareudp)
+               goto drop;
+
+       if (skb->protocol ==  htons(ETH_P_IP))
+               family = AF_INET;
+       else
+               family = AF_INET6;
+
+       proto = bareudp->ethertype;
+
+       if (iptunnel_pull_header(skb, BAREUDP_BASE_HLEN,
+                                proto,
+                                !net_eq(bareudp->net,
+                                dev_net(bareudp->dev)))) {
+               bareudp->dev->stats.rx_dropped++;
+               goto drop;
+       }
+
+       tun_dst = udp_tun_rx_dst(skb, family, TUNNEL_KEY, 0, 0);
+       if (!tun_dst) {
+               bareudp->dev->stats.rx_dropped++;
+               goto drop;
+       }
+       skb_dst_set(skb, &tun_dst->dst);
+       skb->dev = bareudp->dev;
+       oiph = skb_network_header(skb);
+       skb_reset_network_header(skb);
+
+       if (family == AF_INET)
+               err = IP_ECN_decapsulate(oiph, skb);
+#if IS_ENABLED(CONFIG_IPV6)
+       else
+               err = IP6_ECN_decapsulate(oiph, skb);
+#endif
+
+       if (unlikely(err)) {
+               if (log_ecn_error) {
+                       if  (family == AF_INET)
+                               net_info_ratelimited("non-ECT from %pI4 "
+                                                    "with TOS=%#x\n",
+                                                    &((struct iphdr *)oiph)->saddr,
+                                                    ((struct iphdr *)oiph)->tos);
+#if IS_ENABLED(CONFIG_IPV6)
+                       else
+                               net_info_ratelimited("non-ECT from %pI6\n",
+                                                    &((struct ipv6hdr *)oiph)->saddr);
+#endif
+               }
+               if (err > 1) {
+                       ++bareudp->dev->stats.rx_frame_errors;
+                       ++bareudp->dev->stats.rx_errors;
+                       goto drop;
+               }
+       }
+
+       len = skb->len;
+       err = gro_cells_receive(&bareudp->gro_cells, skb);
+       if (likely(err == NET_RX_SUCCESS)) {
+               stats = this_cpu_ptr(bareudp->dev->tstats);
+               u64_stats_update_begin(&stats->syncp);
+               stats->rx_packets++;
+               stats->rx_bytes += len;
+               u64_stats_update_end(&stats->syncp);
+       }
+       return 0;
+drop:
+       /* Consume bad packet */
+       kfree_skb(skb);
+
+       return 0;
+}
+
+static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb)
+{
+       return 0;
+}
+
+static int bareudp_init(struct net_device *dev)
+{
+       struct bareudp_dev *bareudp = netdev_priv(dev);
+       int err;
+
+       dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
+       if (!dev->tstats)
+               return -ENOMEM;
+
+       err = gro_cells_init(&bareudp->gro_cells, dev);
+       if (err) {
+               free_percpu(dev->tstats);
+               return err;
+       }
+       return 0;
+}
+
+static void bareudp_uninit(struct net_device *dev)
+{
+       struct bareudp_dev *bareudp = netdev_priv(dev);
+
+       gro_cells_destroy(&bareudp->gro_cells);
+       free_percpu(dev->tstats);
+}
+
+static struct socket *bareudp_create_sock(struct net *net, __be16 port)
+{
+       struct udp_port_cfg udp_conf;
+       struct socket *sock;
+       int err;
+
+       memset(&udp_conf, 0, sizeof(udp_conf));
+#if IS_ENABLED(CONFIG_IPV6)
+       udp_conf.family = AF_INET6;
+#else
+       udp_conf.family = AF_INET;
+#endif
+       udp_conf.local_udp_port = port;
+       /* Open UDP socket */
+       err = udp_sock_create(net, &udp_conf, &sock);
+       if (err < 0)
+               return ERR_PTR(err);
+
+       return sock;
+}
+
+/* Create new listen socket if needed */
+static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port)
+{
+       struct udp_tunnel_sock_cfg tunnel_cfg;
+       struct socket *sock;
+
+       sock = bareudp_create_sock(bareudp->net, port);
+       if (IS_ERR(sock))
+               return PTR_ERR(sock);
+
+       /* Mark socket as an encapsulation socket */
+       memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
+       tunnel_cfg.sk_user_data = bareudp;
+       tunnel_cfg.encap_type = 1;
+       tunnel_cfg.encap_rcv = bareudp_udp_encap_recv;
+       tunnel_cfg.encap_err_lookup = bareudp_err_lookup;
+       tunnel_cfg.encap_destroy = NULL;
+       setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg);
+
+       if (sock->sk->sk_family == AF_INET6)
+               udp_encap_enable();
+
+       rcu_assign_pointer(bareudp->sock, sock);
+       return 0;
+}
+
+static int bareudp_open(struct net_device *dev)
+{
+       struct bareudp_dev *bareudp = netdev_priv(dev);
+       int ret = 0;
+
+       ret =  bareudp_socket_create(bareudp, bareudp->port);
+       return ret;
+}
+
+static void bareudp_sock_release(struct bareudp_dev *bareudp)
+{
+       struct socket *sock;
+
+       sock = bareudp->sock;
+       rcu_assign_pointer(bareudp->sock, NULL);
+       synchronize_net();
+       udp_tunnel_sock_release(sock);
+}
+
+static int bareudp_stop(struct net_device *dev)
+{
+       struct bareudp_dev *bareudp = netdev_priv(dev);
+
+       bareudp_sock_release(bareudp);
+       return 0;
+}
+
+static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev,
+                           struct bareudp_dev *bareudp,
+                           const struct ip_tunnel_info *info)
+{
+       bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
+       bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
+       struct socket *sock = rcu_dereference(bareudp->sock);
+       bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
+       const struct ip_tunnel_key *key = &info->key;
+       struct rtable *rt;
+       __be16 sport, df;
+       int min_headroom;
+       __u8 tos, ttl;
+       __be32 saddr;
+       int err;
+
+       if (!sock)
+               return -ESHUTDOWN;
+
+       rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, info,
+                                   IPPROTO_UDP, use_cache);
+
+       if (IS_ERR(rt))
+               return PTR_ERR(rt);
+
+       skb_tunnel_check_pmtu(skb, &rt->dst,
+                             BAREUDP_IPV4_HLEN + info->options_len);
+
+       sport = udp_flow_src_port(bareudp->net, skb,
+                                 bareudp->sport_min, USHRT_MAX,
+                                 true);
+       tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
+       ttl = key->ttl;
+       df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
+       skb_scrub_packet(skb, xnet);
+
+       if (!skb_pull(skb, skb_network_offset(skb)))
+               goto free_dst;
+
+       min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len +
+               BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
+
+       err = skb_cow_head(skb, min_headroom);
+       if (unlikely(err))
+               goto free_dst;
+
+       err = udp_tunnel_handle_offloads(skb, udp_sum);
+       if (err)
+               goto free_dst;
+
+       skb_set_inner_protocol(skb, bareudp->ethertype);
+       udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst,
+                           tos, ttl, df, sport, bareudp->port,
+                           !net_eq(bareudp->net, dev_net(bareudp->dev)),
+                           !(info->key.tun_flags & TUNNEL_CSUM));
+       return 0;
+
+free_dst:
+       dst_release(&rt->dst);
+       return err;
+}
+
+#if IS_ENABLED(CONFIG_IPV6)
+static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
+                            struct bareudp_dev *bareudp,
+                            const struct ip_tunnel_info *info)
+{
+       bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
+       bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
+       struct socket *sock  = rcu_dereference(bareudp->sock);
+       bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
+       const struct ip_tunnel_key *key = &info->key;
+       struct dst_entry *dst = NULL;
+       struct in6_addr saddr, daddr;
+       int min_headroom;
+       __u8 prio, ttl;
+       __be16 sport;
+       int err;
+
+       if (!sock)
+               return -ESHUTDOWN;
+
+       dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info,
+                                   IPPROTO_UDP, use_cache);
+       if (IS_ERR(dst))
+               return PTR_ERR(dst);
+
+       skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len);
+
+       sport = udp_flow_src_port(bareudp->net, skb,
+                                 bareudp->sport_min, USHRT_MAX,
+                                 true);
+       prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
+       ttl = key->ttl;
+
+       skb_scrub_packet(skb, xnet);
+
+       if (!skb_pull(skb, skb_network_offset(skb)))
+               goto free_dst;
+
+       min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len +
+               BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
+
+       err = skb_cow_head(skb, min_headroom);
+       if (unlikely(err))
+               goto free_dst;
+
+       err = udp_tunnel_handle_offloads(skb, udp_sum);
+       if (err)
+               goto free_dst;
+
+       daddr = info->key.u.ipv6.dst;
+       udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev,
+                            &saddr, &daddr, prio, ttl,
+                            info->key.label, sport, bareudp->port,
+                            !(info->key.tun_flags & TUNNEL_CSUM));
+       return 0;
+
+free_dst:
+       dst_release(dst);
+       return err;
+}
+#endif
+
+static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+       struct bareudp_dev *bareudp = netdev_priv(dev);
+       struct ip_tunnel_info *info = NULL;
+       int err;
+
+       if (skb->protocol != bareudp->ethertype) {
+               err = -EINVAL;
+               goto tx_error;
+       }
+
+       info = skb_tunnel_info(skb);
+       if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
+               err = -EINVAL;
+               goto tx_error;
+       }
+
+       rcu_read_lock();
+#if IS_ENABLED(CONFIG_IPV6)
+       if (info->mode & IP_TUNNEL_INFO_IPV6)
+               err = bareudp6_xmit_skb(skb, dev, bareudp, info);
+       else
+#endif
+               err = bareudp_xmit_skb(skb, dev, bareudp, info);
+
+       rcu_read_unlock();
+
+       if (likely(!err))
+               return NETDEV_TX_OK;
+tx_error:
+       dev_kfree_skb(skb);
+
+       if (err == -ELOOP)
+               dev->stats.collisions++;
+       else if (err == -ENETUNREACH)
+               dev->stats.tx_carrier_errors++;
+
+       dev->stats.tx_errors++;
+       return NETDEV_TX_OK;
+}
+
+static int bareudp_fill_metadata_dst(struct net_device *dev,
+                                    struct sk_buff *skb)
+{
+       struct ip_tunnel_info *info = skb_tunnel_info(skb);
+       struct bareudp_dev *bareudp = netdev_priv(dev);
+       bool use_cache;
+
+       use_cache = ip_tunnel_dst_cache_usable(skb, info);
+
+       if (ip_tunnel_info_af(info) == AF_INET) {
+               struct rtable *rt;
+               __be32 saddr;
+
+               rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr,
+                                           info, IPPROTO_UDP, use_cache);
+               if (IS_ERR(rt))
+                       return PTR_ERR(rt);
+
+               ip_rt_put(rt);
+               info->key.u.ipv4.src = saddr;
+#if IS_ENABLED(CONFIG_IPV6)
+       } else if (ip_tunnel_info_af(info) == AF_INET6) {
+               struct dst_entry *dst;
+               struct in6_addr saddr;
+               struct socket *sock = rcu_dereference(bareudp->sock);
+
+               dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock,
+                                           &saddr, info, IPPROTO_UDP,
+                                           use_cache);
+               if (IS_ERR(dst))
+                       return PTR_ERR(dst);
+
+               dst_release(dst);
+               info->key.u.ipv6.src = saddr;
+#endif
+       } else {
+               return -EINVAL;
+       }
+
+       info->key.tp_src = udp_flow_src_port(bareudp->net, skb,
+                                            bareudp->sport_min,
+                       USHRT_MAX, true);
+       info->key.tp_dst = bareudp->port;
+       return 0;
+}
+
+static const struct net_device_ops bareudp_netdev_ops = {
+       .ndo_init               = bareudp_init,
+       .ndo_uninit             = bareudp_uninit,
+       .ndo_open               = bareudp_open,
+       .ndo_stop               = bareudp_stop,
+       .ndo_start_xmit         = bareudp_xmit,
+       .ndo_get_stats64        = ip_tunnel_get_stats64,
+       .ndo_fill_metadata_dst  = bareudp_fill_metadata_dst,
+};
+
+static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = {
+       [IFLA_BAREUDP_PORT]                = { .type = NLA_U16 },
+       [IFLA_BAREUDP_ETHERTYPE]           = { .type = NLA_U16 },
+       [IFLA_BAREUDP_SRCPORT_MIN]         = { .type = NLA_U16 },
+};
+
+/* Info for udev, that this is a virtual tunnel endpoint */
+static struct device_type bareudp_type = {
+       .name = "bareudp",
+};
+
+/* Initialize the device structure. */
+static void bareudp_setup(struct net_device *dev)
+{
+       dev->netdev_ops = &bareudp_netdev_ops;
+       dev->needs_free_netdev = true;
+       SET_NETDEV_DEVTYPE(dev, &bareudp_type);
+       dev->features    |= NETIF_F_SG | NETIF_F_HW_CSUM;
+       dev->features    |= NETIF_F_RXCSUM;
+       dev->features    |= NETIF_F_GSO_SOFTWARE;
+       dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
+       dev->hw_features |= NETIF_F_GSO_SOFTWARE;
+       dev->hard_header_len = 0;
+       dev->addr_len = 0;
+       dev->mtu = ETH_DATA_LEN;
+       dev->min_mtu = IPV4_MIN_MTU;
+       dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
+       dev->type = ARPHRD_NONE;
+       netif_keep_dst(dev);
+       dev->priv_flags |= IFF_NO_QUEUE;
+       dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
+}
+
+static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
+                           struct netlink_ext_ack *extack)
+{
+       if (!data) {
+               NL_SET_ERR_MSG(extack,
+                              "Not enough attributes provided to perform the operation");
+               return -EINVAL;
+       }
+       return 0;
+}
+
+static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf)
+{
+       if (!data[IFLA_BAREUDP_PORT] || !data[IFLA_BAREUDP_ETHERTYPE])
+               return -EINVAL;
+
+       if (data[IFLA_BAREUDP_PORT])
+               conf->port =  nla_get_u16(data[IFLA_BAREUDP_PORT]);
+
+       if (data[IFLA_BAREUDP_ETHERTYPE])
+               conf->ethertype =  nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]);
+
+       if (data[IFLA_BAREUDP_SRCPORT_MIN])
+               conf->sport_min =  nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]);
+
+       return 0;
+}
+
+static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
+                                           const struct bareudp_conf *conf)
+{
+       struct bareudp_dev *bareudp, *t = NULL;
+
+       list_for_each_entry(bareudp, &bn->bareudp_list, next) {
+               if (conf->port == bareudp->port)
+                       t = bareudp;
+       }
+       return t;
+}
+
+static int bareudp_configure(struct net *net, struct net_device *dev,
+                            struct bareudp_conf *conf)
+{
+       struct bareudp_net *bn = net_generic(net, bareudp_net_id);
+       struct bareudp_dev *t, *bareudp = netdev_priv(dev);
+       int err;
+
+       bareudp->net = net;
+       bareudp->dev = dev;
+       t = bareudp_find_dev(bn, conf);
+       if (t)
+               return -EBUSY;
+
+       bareudp->port = conf->port;
+       bareudp->ethertype = conf->ethertype;
+       bareudp->sport_min = conf->sport_min;
+       err = register_netdevice(dev);
+       if (err)
+               return err;
+
+       list_add(&bareudp->next, &bn->bareudp_list);
+       return 0;
+}
+
+static int bareudp_link_config(struct net_device *dev,
+                              struct nlattr *tb[])
+{
+       int err;
+
+       if (tb[IFLA_MTU]) {
+               err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
+               if (err)
+                       return err;
+       }
+       return 0;
+}
+
+static int bareudp_newlink(struct net *net, struct net_device *dev,
+                          struct nlattr *tb[], struct nlattr *data[],
+                          struct netlink_ext_ack *extack)
+{
+       struct bareudp_conf conf;
+       int err;
+
+       err = bareudp2info(data, &conf);
+       if (err)
+               return err;
+
+       err = bareudp_configure(net, dev, &conf);
+       if (err)
+               return err;
+
+       err = bareudp_link_config(dev, tb);
+       if (err)
+               return err;
+
+       return 0;
+}
+
+static void bareudp_dellink(struct net_device *dev, struct list_head *head)
+{
+       struct bareudp_dev *bareudp = netdev_priv(dev);
+
+       list_del(&bareudp->next);
+       unregister_netdevice_queue(dev, head);
+}
+
+static size_t bareudp_get_size(const struct net_device *dev)
+{
+       return  nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_PORT */
+               nla_total_size(sizeof(__be16)) +  /* IFLA_BAREUDP_ETHERTYPE */
+               nla_total_size(sizeof(__u16))  +  /* IFLA_BAREUDP_SRCPORT_MIN */
+               0;
+}
+
+static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev)
+{
+       struct bareudp_dev *bareudp = netdev_priv(dev);
+
+       if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port))
+               goto nla_put_failure;
+       if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype))
+               goto nla_put_failure;
+       if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min))
+               goto nla_put_failure;
+
+       return 0;
+
+nla_put_failure:
+       return -EMSGSIZE;
+}
+
+static struct rtnl_link_ops bareudp_link_ops __read_mostly = {
+       .kind           = "bareudp",
+       .maxtype        = IFLA_BAREUDP_MAX,
+       .policy         = bareudp_policy,
+       .priv_size      = sizeof(struct bareudp_dev),
+       .setup          = bareudp_setup,
+       .validate       = bareudp_validate,
+       .newlink        = bareudp_newlink,
+       .dellink        = bareudp_dellink,
+       .get_size       = bareudp_get_size,
+       .fill_info      = bareudp_fill_info,
+};
+
+struct net_device *bareudp_dev_create(struct net *net, const char *name,
+                                     u8 name_assign_type,
+                                     struct bareudp_conf *conf)
+{
+       struct nlattr *tb[IFLA_MAX + 1];
+       struct net_device *dev;
+       LIST_HEAD(list_kill);
+       int err;
+
+       memset(tb, 0, sizeof(tb));
+       dev = rtnl_create_link(net, name, name_assign_type,
+                              &bareudp_link_ops, tb, NULL);
+       if (IS_ERR(dev))
+               return dev;
+
+       err = bareudp_configure(net, dev, conf);
+       if (err) {
+               free_netdev(dev);
+               return ERR_PTR(err);
+       }
+       err = dev_set_mtu(dev, IP_MAX_MTU - BAREUDP_BASE_HLEN);
+       if (err)
+               goto err;
+
+       err = rtnl_configure_link(dev, NULL);
+       if (err < 0)
+               goto err;
+
+       return dev;
+err:
+       bareudp_dellink(dev, &list_kill);
+       unregister_netdevice_many(&list_kill);
+       return ERR_PTR(err);
+}
+EXPORT_SYMBOL_GPL(bareudp_dev_create);
+
+static __net_init int bareudp_init_net(struct net *net)
+{
+       struct bareudp_net *bn = net_generic(net, bareudp_net_id);
+
+       INIT_LIST_HEAD(&bn->bareudp_list);
+       return 0;
+}
+
+static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
+{
+       struct bareudp_net *bn = net_generic(net, bareudp_net_id);
+       struct bareudp_dev *bareudp, *next;
+
+       list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
+               unregister_netdevice_queue(bareudp->dev, head);
+}
+
+static void __net_exit bareudp_exit_batch_net(struct list_head *net_list)
+{
+       struct net *net;
+       LIST_HEAD(list);
+
+       rtnl_lock();
+       list_for_each_entry(net, net_list, exit_list)
+               bareudp_destroy_tunnels(net, &list);
+
+       /* unregister the devices gathered above */
+       unregister_netdevice_many(&list);
+       rtnl_unlock();
+}
+
+static struct pernet_operations bareudp_net_ops = {
+       .init = bareudp_init_net,
+       .exit_batch = bareudp_exit_batch_net,
+       .id   = &bareudp_net_id,
+       .size = sizeof(struct bareudp_net),
+};
+
+static int __init bareudp_init_module(void)
+{
+       int rc;
+
+       rc = register_pernet_subsys(&bareudp_net_ops);
+       if (rc)
+               goto out1;
+
+       rc = rtnl_link_register(&bareudp_link_ops);
+       if (rc)
+               goto out2;
+
+       return 0;
+out2:
+       unregister_pernet_subsys(&bareudp_net_ops);
+out1:
+       return rc;
+}
+late_initcall(bareudp_init_module);
+
+static void __exit bareudp_cleanup_module(void)
+{
+       rtnl_link_unregister(&bareudp_link_ops);
+       unregister_pernet_subsys(&bareudp_net_ops);
+}
+module_exit(bareudp_cleanup_module);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>");
+MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");
diff --git a/include/net/bareudp.h b/include/net/bareudp.h
new file mode 100644 (file)
index 0000000..513fae6
--- /dev/null
@@ -0,0 +1,19 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __NET_BAREUDP_H
+#define __NET_BAREUDP_H
+
+#include <linux/types.h>
+#include <linux/skbuff.h>
+
+struct bareudp_conf {
+       __be16 ethertype;
+       __be16 port;
+       u16 sport_min;
+};
+
+struct net_device *bareudp_dev_create(struct net *net, const char *name,
+                                     u8 name_assign_type,
+                                     struct bareudp_conf *info);
+
+#endif
index cec1a54401f283d8f17de8c3a2bac03dd530c606..1bf8065fe8717c31748db73a6ecdd6207476be46 100644 (file)
@@ -1027,6 +1027,12 @@ struct dst_entry *ip6_dst_lookup_flow(struct net *net, const struct sock *sk, st
 struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
                                         const struct in6_addr *final_dst,
                                         bool connected);
+struct dst_entry *ip6_dst_lookup_tunnel(struct sk_buff *skb,
+                                       struct net_device *dev,
+                                       struct net *net, struct socket *sock,
+                                       struct in6_addr *saddr,
+                                       const struct ip_tunnel_info *info,
+                                       u8 protocol, bool use_cache);
 struct dst_entry *ip6_blackhole_route(struct net *net,
                                      struct dst_entry *orig_dst);
 
index a9c60fc68e3699d4b9dcc1918b815708365aadad..81750ae5083324bee44f02d0b9186a251fab7629 100644 (file)
@@ -128,6 +128,12 @@ static inline struct rtable *__ip_route_output_key(struct net *net,
 
 struct rtable *ip_route_output_flow(struct net *, struct flowi4 *flp,
                                    const struct sock *sk);
+struct rtable *ip_route_output_tunnel(struct sk_buff *skb,
+                                     struct net_device *dev,
+                                     struct net *net, __be32 *saddr,
+                                     const struct ip_tunnel_info *info,
+                                     u8 protocol, bool use_cache);
+
 struct dst_entry *ipv4_blackhole_route(struct net *net,
                                       struct dst_entry *dst_orig);
 
index 024af2d1d0af4059cba0c40211b1095ecd55387e..fb4b33af23d592bd58cf893aeb039427f8fc1a23 100644 (file)
@@ -590,6 +590,17 @@ enum ifla_geneve_df {
        GENEVE_DF_MAX = __GENEVE_DF_END - 1,
 };
 
+/* Bareudp section  */
+enum {
+       IFLA_BAREUDP_UNSPEC,
+       IFLA_BAREUDP_PORT,
+       IFLA_BAREUDP_ETHERTYPE,
+       IFLA_BAREUDP_SRCPORT_MIN,
+       __IFLA_BAREUDP_MAX
+};
+
+#define IFLA_BAREUDP_MAX (__IFLA_BAREUDP_MAX - 1)
+
 /* PPP section */
 enum {
        IFLA_PPP_UNSPEC,
index ebe7060d0fc92c5dbaca99dd534c24e2d6ff5428..042599cc691dee5d45647f16fee73cd278db84bc 100644 (file)
@@ -2774,6 +2774,54 @@ struct rtable *ip_route_output_flow(struct net *net, struct flowi4 *flp4,
 }
 EXPORT_SYMBOL_GPL(ip_route_output_flow);
 
+struct rtable *ip_route_output_tunnel(struct sk_buff *skb,
+                                     struct net_device *dev,
+                                     struct net *net, __be32 *saddr,
+                                     const struct ip_tunnel_info *info,
+                                     u8 protocol, bool use_cache)
+{
+#ifdef CONFIG_DST_CACHE
+       struct dst_cache *dst_cache;
+#endif
+       struct rtable *rt = NULL;
+       struct flowi4 fl4;
+       __u8 tos;
+
+#ifdef CONFIG_DST_CACHE
+       dst_cache = (struct dst_cache *)&info->dst_cache;
+       if (use_cache) {
+               rt = dst_cache_get_ip4(dst_cache, saddr);
+               if (rt)
+                       return rt;
+       }
+#endif
+       memset(&fl4, 0, sizeof(fl4));
+       fl4.flowi4_mark = skb->mark;
+       fl4.flowi4_proto = protocol;
+       fl4.daddr = info->key.u.ipv4.dst;
+       fl4.saddr = info->key.u.ipv4.src;
+       tos = info->key.tos;
+       fl4.flowi4_tos = RT_TOS(tos);
+
+       rt = ip_route_output_key(net, &fl4);
+       if (IS_ERR(rt)) {
+               netdev_dbg(dev, "no route to %pI4\n", &fl4.daddr);
+               return ERR_PTR(-ENETUNREACH);
+       }
+       if (rt->dst.dev == dev) { /* is this necessary? */
+               netdev_dbg(dev, "circular route to %pI4\n", &fl4.daddr);
+               ip_rt_put(rt);
+               return ERR_PTR(-ELOOP);
+       }
+#ifdef CONFIG_DST_CACHE
+       if (use_cache)
+               dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
+#endif
+       *saddr = fl4.saddr;
+       return rt;
+}
+EXPORT_SYMBOL_GPL(ip_route_output_tunnel);
+
 /* called with rcu_read_lock held */
 static int rt_fill_info(struct net *net, __be32 dst, __be32 src,
                        struct rtable *rt, u32 table_id, struct flowi4 *fl4,
index 087304427bbb72cf2b0ccd1d46dee88fad1cebbf..8a8c2d0cfcc810e153256f63fadc9db340c349f7 100644 (file)
@@ -54,6 +54,7 @@
 #include <linux/mroute6.h>
 #include <net/l3mdev.h>
 #include <net/lwtunnel.h>
+#include <net/ip_tunnels.h>
 
 static int ip6_finish_output2(struct net *net, struct sock *sk, struct sk_buff *skb)
 {
@@ -1196,6 +1197,75 @@ struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
 }
 EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
 
+/**
+ *      ip6_dst_lookup_tunnel - perform route lookup on tunnel
+ *      @skb: Packet for which lookup is done
+ *      @dev: Tunnel device
+ *      @net: Network namespace of tunnel device
+ *      @sk: Socket which provides route info
+ *      @saddr: Memory to store the src ip address
+ *      @info: Tunnel information
+ *      @protocol: IP protocol
+ *      @use_cahce: Flag to enable cache usage
+ *      This function performs a route lookup on a tunnel
+ *
+ *      It returns a valid dst pointer and stores src address to be used in
+ *      tunnel in param saddr on success, else a pointer encoded error code.
+ */
+
+struct dst_entry *ip6_dst_lookup_tunnel(struct sk_buff *skb,
+                                       struct net_device *dev,
+                                       struct net *net,
+                                       struct socket *sock,
+                                       struct in6_addr *saddr,
+                                       const struct ip_tunnel_info *info,
+                                       u8 protocol,
+                                       bool use_cache)
+{
+       struct dst_entry *dst = NULL;
+#ifdef CONFIG_DST_CACHE
+       struct dst_cache *dst_cache;
+#endif
+       struct flowi6 fl6;
+       __u8 prio;
+
+#ifdef CONFIG_DST_CACHE
+       dst_cache = (struct dst_cache *)&info->dst_cache;
+       if (use_cache) {
+               dst = dst_cache_get_ip6(dst_cache, saddr);
+               if (dst)
+                       return dst;
+       }
+#endif
+       memset(&fl6, 0, sizeof(fl6));
+       fl6.flowi6_mark = skb->mark;
+       fl6.flowi6_proto = protocol;
+       fl6.daddr = info->key.u.ipv6.dst;
+       fl6.saddr = info->key.u.ipv6.src;
+       prio = info->key.tos;
+       fl6.flowlabel = ip6_make_flowinfo(RT_TOS(prio),
+                                         info->key.label);
+
+       dst = ipv6_stub->ipv6_dst_lookup_flow(net, sock->sk, &fl6,
+                                             NULL);
+       if (IS_ERR(dst)) {
+               netdev_dbg(dev, "no route to %pI6\n", &fl6.daddr);
+               return ERR_PTR(-ENETUNREACH);
+       }
+       if (dst->dev == dev) { /* is this necessary? */
+               netdev_dbg(dev, "circular route to %pI6\n", &fl6.daddr);
+               dst_release(dst);
+               return ERR_PTR(-ELOOP);
+       }
+#ifdef CONFIG_DST_CACHE
+       if (use_cache)
+               dst_cache_set_ip6(dst_cache, dst, &fl6.saddr);
+#endif
+       *saddr = fl6.saddr;
+       return dst;
+}
+EXPORT_SYMBOL_GPL(ip6_dst_lookup_tunnel);
+
 static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
                                               gfp_t gfp)
 {