datapath/linux/compat/ip_tunnels_core.c

   1 /*
   2  * Copyright (c) 2007-2013 Nicira, Inc.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of version 2 of the GNU General Public
   6  * License as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful, but
   9  * WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public License
  14  * along with this program; if not, write to the Free Software
  15  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
  16  * 02110-1301, USA
  17  */
  18
  19
  20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  21
  22 #include <linux/if_vlan.h>
  23 #include <linux/in.h>
  24 #include <linux/in_route.h>
  25 #include <linux/inetdevice.h>
  26 #include <linux/jhash.h>
  27 #include <linux/list.h>
  28 #include <linux/kernel.h>
  29 #include <linux/version.h>
  30 #include <linux/workqueue.h>
  31 #include <linux/rculist.h>
  32 #include <net/ip_tunnels.h>
  33 #include <net/ip6_tunnel.h>
  34 #include <net/route.h>
  35 #include <net/xfrm.h>
  36
  37 #include "compat.h"
  38 #include "gso.h"
  39 #include "vport-netdev.h"
  40
  41 #ifndef USE_UPSTREAM_TUNNEL
  42 void rpl_iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb,
  43                       __be32 src, __be32 dst, __u8 proto, __u8 tos, __u8 ttl,
  44                       __be16 df, bool xnet)
  45 {
  46         struct net_device *dev = skb->dev;
  47         int pkt_len = skb->len - skb_inner_network_offset(skb);
  48         struct iphdr *iph;
  49         int err;
  50
  51         skb_scrub_packet(skb, xnet);
  52
  53         skb_clear_hash(skb);
  54         skb_dst_set(skb, &rt->dst);
  55
  56 #if 0
  57         /* Do not clear ovs_skb_cb.  It will be done in gso code. */
  58         memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
  59 #endif
  60
  61         /* Push down and install the IP header. */
  62         __skb_push(skb, sizeof(struct iphdr));
  63         skb_reset_network_header(skb);
  64
  65         iph = ip_hdr(skb);
  66
  67         iph->version    =       4;
  68         iph->ihl        =       sizeof(struct iphdr) >> 2;
  69         iph->frag_off   =       df;
  70         iph->protocol   =       proto;
  71         iph->tos        =       tos;
  72         iph->daddr      =       dst;
  73         iph->saddr      =       src;
  74         iph->ttl        =       ttl;
  75
  76 #ifdef HAVE_IP_SELECT_IDENT_USING_DST_ENTRY
  77         __ip_select_ident(iph, &rt->dst, (skb_shinfo(skb)->gso_segs ?: 1) - 1);
  78 #elif defined(HAVE_IP_SELECT_IDENT_USING_NET)
  79         __ip_select_ident(dev_net(rt->dst.dev), iph,
  80                           skb_shinfo(skb)->gso_segs ?: 1);
  81 #else
  82         __ip_select_ident(iph, skb_shinfo(skb)->gso_segs ?: 1);
  83 #endif
  84
  85         err = ip_local_out(dev_net(rt->dst.dev), sk, skb);
  86         if (unlikely(net_xmit_eval(err)))
  87                 pkt_len = 0;
  88         iptunnel_xmit_stats(dev, pkt_len);
  89 }
  90 EXPORT_SYMBOL_GPL(rpl_iptunnel_xmit);
  91
  92 int ovs_iptunnel_handle_offloads(struct sk_buff *skb,
  93                                  int gso_type_mask,
  94                                  void (*fix_segment)(struct sk_buff *))
  95 {
  96         int err;
  97
  98         if (likely(!skb_is_encapsulated(skb))) {
  99                 skb_reset_inner_headers(skb);
 100                 skb->encapsulation = 1;
 101         } else if (skb_is_gso(skb)) {
 102                 err = -ENOSYS;
 103                 goto error;
 104         }
 105
 106         if (skb_is_gso(skb)) {
 107                 err = skb_unclone(skb, GFP_ATOMIC);
 108                 if (unlikely(err))
 109                         goto error;
 110                 skb_shinfo(skb)->gso_type |= gso_type_mask;
 111
 112 #ifndef USE_UPSTREAM_TUNNEL_GSO
 113                 if (gso_type_mask)
 114                         fix_segment = NULL;
 115
 116                 OVS_GSO_CB(skb)->fix_segment = fix_segment;
 117 #endif
 118                 return 0;
 119         }
 120
 121         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 122                 skb->ip_summed = CHECKSUM_NONE;
 123                 skb->encapsulation = 0;
 124         }
 125
 126         return 0;
 127 error:
 128         return err;
 129 }
 130 EXPORT_SYMBOL_GPL(ovs_iptunnel_handle_offloads);
 131
 132
 133 #if LINUX_VERSION_CODE < KERNEL_VERSION(4,7,0)
 134 struct sk_buff *rpl_iptunnel_handle_offloads(struct sk_buff *skb,
 135                                              bool csum_help,
 136                                              int gso_type_mask)
 137 #else
 138 int rpl_iptunnel_handle_offloads(struct sk_buff *skb,
 139                                  bool csum_help,
 140                                  int gso_type_mask)
 141 #endif
 142 {
 143         int err;
 144
 145         if (likely(!skb->encapsulation)) {
 146                 skb_reset_inner_headers(skb);
 147                 skb->encapsulation = 1;
 148         }
 149
 150         if (skb_is_gso(skb)) {
 151                 err = skb_unclone(skb, GFP_ATOMIC);
 152                 if (unlikely(err))
 153                         goto error;
 154                 skb_shinfo(skb)->gso_type |= gso_type_mask;
 155                 goto out;
 156         }
 157
 158         /* If packet is not gso and we are resolving any partial checksum,
 159          * clear encapsulation flag. This allows setting CHECKSUM_PARTIAL
 160          * on the outer header without confusing devices that implement
 161          * NETIF_F_IP_CSUM with encapsulation.
 162          */
 163         if (csum_help)
 164                 skb->encapsulation = 0;
 165
 166         if (skb->ip_summed == CHECKSUM_PARTIAL && csum_help) {
 167                 err = skb_checksum_help(skb);
 168                 if (unlikely(err))
 169                         goto error;
 170         } else if (skb->ip_summed != CHECKSUM_PARTIAL)
 171                 skb->ip_summed = CHECKSUM_NONE;
 172
 173 #if LINUX_VERSION_CODE < KERNEL_VERSION(4,7,0)
 174 out:
 175         return skb;
 176 error:
 177         kfree_skb(skb);
 178         return ERR_PTR(err);
 179 #else
 180 out:
 181 error:
 182         return 0;
 183 #endif
 184 }
 185 EXPORT_SYMBOL_GPL(rpl_iptunnel_handle_offloads);
 186
 187 int rpl___iptunnel_pull_header(struct sk_buff *skb, int hdr_len,
 188                                __be16 inner_proto, bool raw_proto, bool xnet)
 189 {
 190         if (unlikely(!pskb_may_pull(skb, hdr_len)))
 191                 return -ENOMEM;
 192
 193         skb_pull_rcsum(skb, hdr_len);
 194
 195         if (!raw_proto && inner_proto == htons(ETH_P_TEB)) {
 196                 struct ethhdr *eh;
 197
 198                 if (unlikely(!pskb_may_pull(skb, ETH_HLEN)))
 199                         return -ENOMEM;
 200
 201                 eh = (struct ethhdr *)skb->data;
 202                 if (likely(eth_proto_is_802_3(eh->h_proto)))
 203                         skb->protocol = eh->h_proto;
 204                 else
 205                         skb->protocol = htons(ETH_P_802_2);
 206
 207         } else {
 208                 skb->protocol = inner_proto;
 209         }
 210
 211         skb_clear_hash_if_not_l4(skb);
 212         skb->vlan_tci = 0;
 213         skb_set_queue_mapping(skb, 0);
 214         skb_scrub_packet(skb, xnet);
 215
 216         return iptunnel_pull_offloads(skb);
 217 }
 218 EXPORT_SYMBOL_GPL(rpl___iptunnel_pull_header);
 219 #endif /* USE_UPSTREAM_TUNNEL */
 220
 221 bool ovs_skb_is_encapsulated(struct sk_buff *skb)
 222 {
 223         /* checking for inner protocol should be sufficient on newer kernel, but
 224          * old kernel just set encapsulation bit.
 225          */
 226         return ovs_skb_get_inner_protocol(skb) || skb->encapsulation;
 227 }
 228 EXPORT_SYMBOL_GPL(ovs_skb_is_encapsulated);
 229
 230 /* derived from ip_tunnel_rcv(). */
 231 void ovs_ip_tunnel_rcv(struct net_device *dev, struct sk_buff *skb,
 232                        struct metadata_dst *tun_dst)
 233 {
 234         struct pcpu_sw_netstats *tstats;
 235
 236         tstats = this_cpu_ptr((struct pcpu_sw_netstats __percpu *)dev->tstats);
 237         u64_stats_update_begin(&tstats->syncp);
 238         tstats->rx_packets++;
 239         tstats->rx_bytes += skb->len;
 240         u64_stats_update_end(&tstats->syncp);
 241
 242         skb_reset_mac_header(skb);
 243         skb_scrub_packet(skb, false);
 244         skb->protocol = eth_type_trans(skb, dev);
 245         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 246
 247         ovs_skb_dst_set(skb, (struct dst_entry *)tun_dst);
 248
 249 #ifndef USE_UPSTREAM_TUNNEL
 250         netdev_port_receive(skb, &tun_dst->u.tun_info);
 251 #else
 252         netif_rx(skb);
 253 #endif
 254 }
 255
 256 #ifndef HAVE_PCPU_SW_NETSTATS
 257 #define netdev_stats_to_stats64 rpl_netdev_stats_to_stats64
 258 static void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
 259                                     const struct net_device_stats *netdev_stats)
 260 {
 261 #if BITS_PER_LONG == 64
 262         BUILD_BUG_ON(sizeof(*stats64) != sizeof(*netdev_stats));
 263         memcpy(stats64, netdev_stats, sizeof(*stats64));
 264 #else
 265         size_t i, n = sizeof(*stats64) / sizeof(u64);
 266         const unsigned long *src = (const unsigned long *)netdev_stats;
 267         u64 *dst = (u64 *)stats64;
 268
 269         BUILD_BUG_ON(sizeof(*netdev_stats) / sizeof(unsigned long) !=
 270                      sizeof(*stats64) / sizeof(u64));
 271         for (i = 0; i < n; i++)
 272                 dst[i] = src[i];
 273 #endif
 274 }
 275 #endif
 276
 277 #if LINUX_VERSION_CODE < KERNEL_VERSION(4,11,0)
 278 struct rtnl_link_stats64 *rpl_ip_tunnel_get_stats64(struct net_device *dev,
 279                                                 struct rtnl_link_stats64 *tot)
 280 #else
 281 void rpl_ip_tunnel_get_stats64(struct net_device *dev,
 282                                                 struct rtnl_link_stats64 *tot)
 283 #endif
 284 {
 285         int i;
 286
 287         netdev_stats_to_stats64(tot, &dev->stats);
 288
 289         for_each_possible_cpu(i) {
 290                 const struct pcpu_sw_netstats *tstats =
 291                                                    per_cpu_ptr((struct pcpu_sw_netstats __percpu *)dev->tstats, i);
 292                 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
 293                 unsigned int start;
 294
 295                 do {
 296                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
 297                         rx_packets = tstats->rx_packets;
 298                         tx_packets = tstats->tx_packets;
 299                         rx_bytes = tstats->rx_bytes;
 300                         tx_bytes = tstats->tx_bytes;
 301                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
 302
 303                 tot->rx_packets += rx_packets;
 304                 tot->tx_packets += tx_packets;
 305                 tot->rx_bytes   += rx_bytes;
 306                 tot->tx_bytes   += tx_bytes;
 307         }
 308
 309 #if LINUX_VERSION_CODE < KERNEL_VERSION(4,11,0)
 310         return tot;
 311 #endif
 312 }
 313
 314 void rpl_ip6tunnel_xmit(struct sock *sk, struct sk_buff *skb,
 315                     struct net_device *dev)
 316 {
 317         int pkt_len, err;
 318
 319         pkt_len = skb->len - skb_inner_network_offset(skb);
 320 #ifdef HAVE_IP6_LOCAL_OUT_SK
 321         err = ip6_local_out_sk(sk, skb);
 322 #else
 323         err = ip6_local_out(dev_net(skb_dst(skb)->dev), sk, skb);
 324 #endif
 325         if (net_xmit_eval(err))
 326                 pkt_len = -1;
 327
 328         iptunnel_xmit_stats(dev, pkt_len);
 329 }
 330 EXPORT_SYMBOL_GPL(rpl_ip6tunnel_xmit);