ipv6: Cleanup IPv6 tunnel receive path
authorTom Herbert <tom@herbertland.com>
Sat, 30 Apr 2016 00:12:15 +0000 (17:12 -0700)
committerDavid S. Miller <davem@davemloft.net>
Mon, 2 May 2016 23:23:31 +0000 (19:23 -0400)
Some basic changes to make IPv6 tunnel receive path look more like
IPv4 path:
  - Make ip6_tnl_rcv non-static so that GREv6 and others can call it
  - Make ip6_tnl_rcv look like ip_tunnel_rcv
  - Switch to gro_cells_receive
  - Make ip6_tnl_rcv non-static and export it

Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/net/ip6_tunnel.h
net/ipv6/ip6_tunnel.c

index 499a707765ea68fa010e03716319dcef57de49fe..eab3a9b19ae0925d7d8ae311256c1258a7a80746 100644 (file)
@@ -42,6 +42,7 @@ struct ip6_tnl {
        struct __ip6_tnl_parm parms;    /* tunnel configuration parameters */
        struct flowi fl;        /* flowi template for xmit */
        struct dst_cache dst_cache;     /* cached dst */
+       struct gro_cells gro_cells;
 
        int err_count;
        unsigned long err_time;
@@ -63,6 +64,9 @@ struct ipv6_tlv_tnl_enc_lim {
 
 int ip6_tnl_rcv_ctl(struct ip6_tnl *t, const struct in6_addr *laddr,
                const struct in6_addr *raddr);
+int ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
+               const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
+               bool log_ecn_error);
 int ip6_tnl_xmit_ctl(struct ip6_tnl *t, const struct in6_addr *laddr,
                     const struct in6_addr *raddr);
 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw);
index 1f20345cbc970e88dd3b1ace2cc111cb56e5f73b..94ed065eff7f3be5ecad31e7f26ae82a1ea235cb 100644 (file)
@@ -238,6 +238,7 @@ static void ip6_dev_free(struct net_device *dev)
 {
        struct ip6_tnl *t = netdev_priv(dev);
 
+       gro_cells_destroy(&t->gro_cells);
        dst_cache_destroy(&t->dst_cache);
        free_percpu(dev->tstats);
        free_netdev(dev);
@@ -753,97 +754,157 @@ int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 }
 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 
-/**
- * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
- *   @skb: received socket buffer
- *   @protocol: ethernet protocol ID
- *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
- *
- * Return: 0
- **/
-
-static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
-                      __u8 ipproto,
-                      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
-                                                  const struct ipv6hdr *ipv6h,
-                                                  struct sk_buff *skb))
+static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
+                        const struct tnl_ptk_info *tpi,
+                        struct metadata_dst *tun_dst,
+                        int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
+                                               const struct ipv6hdr *ipv6h,
+                                               struct sk_buff *skb),
+                        bool log_ecn_err)
 {
-       struct ip6_tnl *t;
+       struct pcpu_sw_netstats *tstats;
        const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
-       u8 tproto;
        int err;
 
-       rcu_read_lock();
-       t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
-       if (t) {
-               struct pcpu_sw_netstats *tstats;
+       if ((!(tpi->flags & TUNNEL_CSUM) &&
+            (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
+           ((tpi->flags & TUNNEL_CSUM) &&
+            !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
+               tunnel->dev->stats.rx_crc_errors++;
+               tunnel->dev->stats.rx_errors++;
+               goto drop;
+       }
 
-               tproto = ACCESS_ONCE(t->parms.proto);
-               if (tproto != ipproto && tproto != 0) {
-                       rcu_read_unlock();
-                       goto discard;
+       if (tunnel->parms.i_flags & TUNNEL_SEQ) {
+               if (!(tpi->flags & TUNNEL_SEQ) ||
+                   (tunnel->i_seqno &&
+                    (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
+                       tunnel->dev->stats.rx_fifo_errors++;
+                       tunnel->dev->stats.rx_errors++;
+                       goto drop;
                }
+               tunnel->i_seqno = ntohl(tpi->seq) + 1;
+       }
 
-               if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
-                       rcu_read_unlock();
-                       goto discard;
-               }
+       skb->protocol = tpi->proto;
 
-               if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
-                       t->dev->stats.rx_dropped++;
-                       rcu_read_unlock();
-                       goto discard;
+       /* Warning: All skb pointers will be invalidated! */
+       if (tunnel->dev->type == ARPHRD_ETHER) {
+               if (!pskb_may_pull(skb, ETH_HLEN)) {
+                       tunnel->dev->stats.rx_length_errors++;
+                       tunnel->dev->stats.rx_errors++;
+                       goto drop;
                }
-               skb->mac_header = skb->network_header;
-               skb_reset_network_header(skb);
-               skb->protocol = htons(protocol);
-               memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
-
-               __skb_tunnel_rx(skb, t->dev, t->net);
-
-               err = dscp_ecn_decapsulate(t, ipv6h, skb);
-               if (unlikely(err)) {
-                       if (log_ecn_error)
-                               net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
-                                                    &ipv6h->saddr,
-                                                    ipv6_get_dsfield(ipv6h));
-                       if (err > 1) {
-                               ++t->dev->stats.rx_frame_errors;
-                               ++t->dev->stats.rx_errors;
-                               rcu_read_unlock();
-                               goto discard;
-                       }
+
+               ipv6h = ipv6_hdr(skb);
+               skb->protocol = eth_type_trans(skb, tunnel->dev);
+               skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
+       } else {
+               skb->dev = tunnel->dev;
+       }
+
+       skb_reset_network_header(skb);
+       memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
+
+       __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
+
+       err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
+       if (unlikely(err)) {
+               if (log_ecn_err)
+                       net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
+                                            &ipv6h->saddr,
+                                            ipv6_get_dsfield(ipv6h));
+               if (err > 1) {
+                       ++tunnel->dev->stats.rx_frame_errors;
+                       ++tunnel->dev->stats.rx_errors;
+                       goto drop;
                }
+       }
 
-               tstats = this_cpu_ptr(t->dev->tstats);
-               u64_stats_update_begin(&tstats->syncp);
-               tstats->rx_packets++;
-               tstats->rx_bytes += skb->len;
-               u64_stats_update_end(&tstats->syncp);
+       tstats = this_cpu_ptr(tunnel->dev->tstats);
+       u64_stats_update_begin(&tstats->syncp);
+       tstats->rx_packets++;
+       tstats->rx_bytes += skb->len;
+       u64_stats_update_end(&tstats->syncp);
 
-               netif_rx(skb);
+       skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 
-               rcu_read_unlock();
-               return 0;
+       gro_cells_receive(&tunnel->gro_cells, skb);
+       return 0;
+
+drop:
+       kfree_skb(skb);
+       return 0;
+}
+
+int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
+               const struct tnl_ptk_info *tpi,
+               struct metadata_dst *tun_dst,
+               bool log_ecn_err)
+{
+       return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
+                            log_ecn_err);
+}
+EXPORT_SYMBOL(ip6_tnl_rcv);
+
+static const struct tnl_ptk_info tpi_v6 = {
+       /* no tunnel info required for ipxip6. */
+       .proto = htons(ETH_P_IPV6),
+};
+
+static const struct tnl_ptk_info tpi_v4 = {
+       /* no tunnel info required for ipxip6. */
+       .proto = htons(ETH_P_IP),
+};
+
+static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
+                     const struct tnl_ptk_info *tpi,
+                     int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
+                                                 const struct ipv6hdr *ipv6h,
+                                                 struct sk_buff *skb))
+{
+       struct ip6_tnl *t;
+       const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
+       int ret = -1;
+
+       rcu_read_lock();
+       t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
+
+       if (t) {
+               u8 tproto = ACCESS_ONCE(t->parms.proto);
+
+               if (tproto != ipproto && tproto != 0)
+                       goto drop;
+               if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
+                       goto drop;
+               if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
+                       goto drop;
+               if (iptunnel_pull_header(skb, 0, tpi->proto, false))
+                       goto drop;
+               ret = __ip6_tnl_rcv(t, skb, tpi, NULL, dscp_ecn_decapsulate,
+                                   log_ecn_error);
        }
+
        rcu_read_unlock();
-       return 1;
 
-discard:
+       return ret;
+
+drop:
+       rcu_read_unlock();
        kfree_skb(skb);
        return 0;
 }
 
 static int ip4ip6_rcv(struct sk_buff *skb)
 {
-       return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
-                          ip4ip6_dscp_ecn_decapsulate);
+       return ipxip6_rcv(skb, IPPROTO_IP, &tpi_v4,
+                         ip4ip6_dscp_ecn_decapsulate);
 }
 
 static int ip6ip6_rcv(struct sk_buff *skb)
 {
-       return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
-                          ip6ip6_dscp_ecn_decapsulate);
+       return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
+                         ip6ip6_dscp_ecn_decapsulate);
 }
 
 struct ipv6_tel_txoption {
@@ -1370,6 +1431,8 @@ ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
        struct net *net = t->net;
        struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 
+       memset(&p1, 0, sizeof(p1));
+
        switch (cmd) {
        case SIOCGETTUNNEL:
                if (dev == ip6n->fb_tnl_dev) {
@@ -1549,13 +1612,22 @@ ip6_tnl_dev_init_gen(struct net_device *dev)
                return -ENOMEM;
 
        ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
-       if (ret) {
-               free_percpu(dev->tstats);
-               dev->tstats = NULL;
-               return ret;
-       }
+       if (ret)
+               goto free_stats;
+
+       ret = gro_cells_init(&t->gro_cells, dev);
+       if (ret)
+               goto destroy_dst;
 
        return 0;
+
+destroy_dst:
+       dst_cache_destroy(&t->dst_cache);
+free_stats:
+       free_percpu(dev->tstats);
+       dev->tstats = NULL;
+
+       return ret;
 }
 
 /**