linux/net/openvswitch/flow.c
<<
>>
Prefs
   1/*
   2 * Copyright (c) 2007-2011 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#include "flow.h"
  20#include "datapath.h"
  21#include <linux/uaccess.h>
  22#include <linux/netdevice.h>
  23#include <linux/etherdevice.h>
  24#include <linux/if_ether.h>
  25#include <linux/if_vlan.h>
  26#include <net/llc_pdu.h>
  27#include <linux/kernel.h>
  28#include <linux/jhash.h>
  29#include <linux/jiffies.h>
  30#include <linux/llc.h>
  31#include <linux/module.h>
  32#include <linux/in.h>
  33#include <linux/rcupdate.h>
  34#include <linux/if_arp.h>
  35#include <linux/ip.h>
  36#include <linux/ipv6.h>
  37#include <linux/tcp.h>
  38#include <linux/udp.h>
  39#include <linux/icmp.h>
  40#include <linux/icmpv6.h>
  41#include <linux/rculist.h>
  42#include <net/ip.h>
  43#include <net/ipv6.h>
  44#include <net/ndisc.h>
  45
  46static struct kmem_cache *flow_cache;
  47
  48static int check_header(struct sk_buff *skb, int len)
  49{
  50        if (unlikely(skb->len < len))
  51                return -EINVAL;
  52        if (unlikely(!pskb_may_pull(skb, len)))
  53                return -ENOMEM;
  54        return 0;
  55}
  56
  57static bool arphdr_ok(struct sk_buff *skb)
  58{
  59        return pskb_may_pull(skb, skb_network_offset(skb) +
  60                                  sizeof(struct arp_eth_header));
  61}
  62
  63static int check_iphdr(struct sk_buff *skb)
  64{
  65        unsigned int nh_ofs = skb_network_offset(skb);
  66        unsigned int ip_len;
  67        int err;
  68
  69        err = check_header(skb, nh_ofs + sizeof(struct iphdr));
  70        if (unlikely(err))
  71                return err;
  72
  73        ip_len = ip_hdrlen(skb);
  74        if (unlikely(ip_len < sizeof(struct iphdr) ||
  75                     skb->len < nh_ofs + ip_len))
  76                return -EINVAL;
  77
  78        skb_set_transport_header(skb, nh_ofs + ip_len);
  79        return 0;
  80}
  81
  82static bool tcphdr_ok(struct sk_buff *skb)
  83{
  84        int th_ofs = skb_transport_offset(skb);
  85        int tcp_len;
  86
  87        if (unlikely(!pskb_may_pull(skb, th_ofs + sizeof(struct tcphdr))))
  88                return false;
  89
  90        tcp_len = tcp_hdrlen(skb);
  91        if (unlikely(tcp_len < sizeof(struct tcphdr) ||
  92                     skb->len < th_ofs + tcp_len))
  93                return false;
  94
  95        return true;
  96}
  97
  98static bool udphdr_ok(struct sk_buff *skb)
  99{
 100        return pskb_may_pull(skb, skb_transport_offset(skb) +
 101                                  sizeof(struct udphdr));
 102}
 103
 104static bool icmphdr_ok(struct sk_buff *skb)
 105{
 106        return pskb_may_pull(skb, skb_transport_offset(skb) +
 107                                  sizeof(struct icmphdr));
 108}
 109
 110u64 ovs_flow_used_time(unsigned long flow_jiffies)
 111{
 112        struct timespec cur_ts;
 113        u64 cur_ms, idle_ms;
 114
 115        ktime_get_ts(&cur_ts);
 116        idle_ms = jiffies_to_msecs(jiffies - flow_jiffies);
 117        cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC +
 118                 cur_ts.tv_nsec / NSEC_PER_MSEC;
 119
 120        return cur_ms - idle_ms;
 121}
 122
 123#define SW_FLOW_KEY_OFFSET(field)               \
 124        (offsetof(struct sw_flow_key, field) +  \
 125         FIELD_SIZEOF(struct sw_flow_key, field))
 126
 127static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key,
 128                         int *key_lenp)
 129{
 130        unsigned int nh_ofs = skb_network_offset(skb);
 131        unsigned int nh_len;
 132        int payload_ofs;
 133        struct ipv6hdr *nh;
 134        uint8_t nexthdr;
 135        __be16 frag_off;
 136        int err;
 137
 138        *key_lenp = SW_FLOW_KEY_OFFSET(ipv6.label);
 139
 140        err = check_header(skb, nh_ofs + sizeof(*nh));
 141        if (unlikely(err))
 142                return err;
 143
 144        nh = ipv6_hdr(skb);
 145        nexthdr = nh->nexthdr;
 146        payload_ofs = (u8 *)(nh + 1) - skb->data;
 147
 148        key->ip.proto = NEXTHDR_NONE;
 149        key->ip.tos = ipv6_get_dsfield(nh);
 150        key->ip.ttl = nh->hop_limit;
 151        key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL);
 152        key->ipv6.addr.src = nh->saddr;
 153        key->ipv6.addr.dst = nh->daddr;
 154
 155        payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr, &frag_off);
 156        if (unlikely(payload_ofs < 0))
 157                return -EINVAL;
 158
 159        if (frag_off) {
 160                if (frag_off & htons(~0x7))
 161                        key->ip.frag = OVS_FRAG_TYPE_LATER;
 162                else
 163                        key->ip.frag = OVS_FRAG_TYPE_FIRST;
 164        }
 165
 166        nh_len = payload_ofs - nh_ofs;
 167        skb_set_transport_header(skb, nh_ofs + nh_len);
 168        key->ip.proto = nexthdr;
 169        return nh_len;
 170}
 171
 172static bool icmp6hdr_ok(struct sk_buff *skb)
 173{
 174        return pskb_may_pull(skb, skb_transport_offset(skb) +
 175                                  sizeof(struct icmp6hdr));
 176}
 177
 178#define TCP_FLAGS_OFFSET 13
 179#define TCP_FLAG_MASK 0x3f
 180
 181void ovs_flow_used(struct sw_flow *flow, struct sk_buff *skb)
 182{
 183        u8 tcp_flags = 0;
 184
 185        if ((flow->key.eth.type == htons(ETH_P_IP) ||
 186             flow->key.eth.type == htons(ETH_P_IPV6)) &&
 187            flow->key.ip.proto == IPPROTO_TCP &&
 188            likely(skb->len >= skb_transport_offset(skb) + sizeof(struct tcphdr))) {
 189                u8 *tcp = (u8 *)tcp_hdr(skb);
 190                tcp_flags = *(tcp + TCP_FLAGS_OFFSET) & TCP_FLAG_MASK;
 191        }
 192
 193        spin_lock(&flow->lock);
 194        flow->used = jiffies;
 195        flow->packet_count++;
 196        flow->byte_count += skb->len;
 197        flow->tcp_flags |= tcp_flags;
 198        spin_unlock(&flow->lock);
 199}
 200
 201struct sw_flow_actions *ovs_flow_actions_alloc(const struct nlattr *actions)
 202{
 203        int actions_len = nla_len(actions);
 204        struct sw_flow_actions *sfa;
 205
 206        if (actions_len > MAX_ACTIONS_BUFSIZE)
 207                return ERR_PTR(-EINVAL);
 208
 209        sfa = kmalloc(sizeof(*sfa) + actions_len, GFP_KERNEL);
 210        if (!sfa)
 211                return ERR_PTR(-ENOMEM);
 212
 213        sfa->actions_len = actions_len;
 214        memcpy(sfa->actions, nla_data(actions), actions_len);
 215        return sfa;
 216}
 217
 218struct sw_flow *ovs_flow_alloc(void)
 219{
 220        struct sw_flow *flow;
 221
 222        flow = kmem_cache_alloc(flow_cache, GFP_KERNEL);
 223        if (!flow)
 224                return ERR_PTR(-ENOMEM);
 225
 226        spin_lock_init(&flow->lock);
 227        flow->sf_acts = NULL;
 228
 229        return flow;
 230}
 231
 232static struct hlist_head *find_bucket(struct flow_table *table, u32 hash)
 233{
 234        hash = jhash_1word(hash, table->hash_seed);
 235        return flex_array_get(table->buckets,
 236                                (hash & (table->n_buckets - 1)));
 237}
 238
 239static struct flex_array *alloc_buckets(unsigned int n_buckets)
 240{
 241        struct flex_array *buckets;
 242        int i, err;
 243
 244        buckets = flex_array_alloc(sizeof(struct hlist_head *),
 245                                   n_buckets, GFP_KERNEL);
 246        if (!buckets)
 247                return NULL;
 248
 249        err = flex_array_prealloc(buckets, 0, n_buckets, GFP_KERNEL);
 250        if (err) {
 251                flex_array_free(buckets);
 252                return NULL;
 253        }
 254
 255        for (i = 0; i < n_buckets; i++)
 256                INIT_HLIST_HEAD((struct hlist_head *)
 257                                        flex_array_get(buckets, i));
 258
 259        return buckets;
 260}
 261
 262static void free_buckets(struct flex_array *buckets)
 263{
 264        flex_array_free(buckets);
 265}
 266
 267struct flow_table *ovs_flow_tbl_alloc(int new_size)
 268{
 269        struct flow_table *table = kmalloc(sizeof(*table), GFP_KERNEL);
 270
 271        if (!table)
 272                return NULL;
 273
 274        table->buckets = alloc_buckets(new_size);
 275
 276        if (!table->buckets) {
 277                kfree(table);
 278                return NULL;
 279        }
 280        table->n_buckets = new_size;
 281        table->count = 0;
 282        table->node_ver = 0;
 283        table->keep_flows = false;
 284        get_random_bytes(&table->hash_seed, sizeof(u32));
 285
 286        return table;
 287}
 288
 289void ovs_flow_tbl_destroy(struct flow_table *table)
 290{
 291        int i;
 292
 293        if (!table)
 294                return;
 295
 296        if (table->keep_flows)
 297                goto skip_flows;
 298
 299        for (i = 0; i < table->n_buckets; i++) {
 300                struct sw_flow *flow;
 301                struct hlist_head *head = flex_array_get(table->buckets, i);
 302                struct hlist_node *node, *n;
 303                int ver = table->node_ver;
 304
 305                hlist_for_each_entry_safe(flow, node, n, head, hash_node[ver]) {
 306                        hlist_del_rcu(&flow->hash_node[ver]);
 307                        ovs_flow_free(flow);
 308                }
 309        }
 310
 311skip_flows:
 312        free_buckets(table->buckets);
 313        kfree(table);
 314}
 315
 316static void flow_tbl_destroy_rcu_cb(struct rcu_head *rcu)
 317{
 318        struct flow_table *table = container_of(rcu, struct flow_table, rcu);
 319
 320        ovs_flow_tbl_destroy(table);
 321}
 322
 323void ovs_flow_tbl_deferred_destroy(struct flow_table *table)
 324{
 325        if (!table)
 326                return;
 327
 328        call_rcu(&table->rcu, flow_tbl_destroy_rcu_cb);
 329}
 330
 331struct sw_flow *ovs_flow_tbl_next(struct flow_table *table, u32 *bucket, u32 *last)
 332{
 333        struct sw_flow *flow;
 334        struct hlist_head *head;
 335        struct hlist_node *n;
 336        int ver;
 337        int i;
 338
 339        ver = table->node_ver;
 340        while (*bucket < table->n_buckets) {
 341                i = 0;
 342                head = flex_array_get(table->buckets, *bucket);
 343                hlist_for_each_entry_rcu(flow, n, head, hash_node[ver]) {
 344                        if (i < *last) {
 345                                i++;
 346                                continue;
 347                        }
 348                        *last = i + 1;
 349                        return flow;
 350                }
 351                (*bucket)++;
 352                *last = 0;
 353        }
 354
 355        return NULL;
 356}
 357
 358static void flow_table_copy_flows(struct flow_table *old, struct flow_table *new)
 359{
 360        int old_ver;
 361        int i;
 362
 363        old_ver = old->node_ver;
 364        new->node_ver = !old_ver;
 365
 366        /* Insert in new table. */
 367        for (i = 0; i < old->n_buckets; i++) {
 368                struct sw_flow *flow;
 369                struct hlist_head *head;
 370                struct hlist_node *n;
 371
 372                head = flex_array_get(old->buckets, i);
 373
 374                hlist_for_each_entry(flow, n, head, hash_node[old_ver])
 375                        ovs_flow_tbl_insert(new, flow);
 376        }
 377        old->keep_flows = true;
 378}
 379
 380static struct flow_table *__flow_tbl_rehash(struct flow_table *table, int n_buckets)
 381{
 382        struct flow_table *new_table;
 383
 384        new_table = ovs_flow_tbl_alloc(n_buckets);
 385        if (!new_table)
 386                return ERR_PTR(-ENOMEM);
 387
 388        flow_table_copy_flows(table, new_table);
 389
 390        return new_table;
 391}
 392
 393struct flow_table *ovs_flow_tbl_rehash(struct flow_table *table)
 394{
 395        return __flow_tbl_rehash(table, table->n_buckets);
 396}
 397
 398struct flow_table *ovs_flow_tbl_expand(struct flow_table *table)
 399{
 400        return __flow_tbl_rehash(table, table->n_buckets * 2);
 401}
 402
 403void ovs_flow_free(struct sw_flow *flow)
 404{
 405        if (unlikely(!flow))
 406                return;
 407
 408        kfree((struct sf_flow_acts __force *)flow->sf_acts);
 409        kmem_cache_free(flow_cache, flow);
 410}
 411
 412/* RCU callback used by ovs_flow_deferred_free. */
 413static void rcu_free_flow_callback(struct rcu_head *rcu)
 414{
 415        struct sw_flow *flow = container_of(rcu, struct sw_flow, rcu);
 416
 417        ovs_flow_free(flow);
 418}
 419
 420/* Schedules 'flow' to be freed after the next RCU grace period.
 421 * The caller must hold rcu_read_lock for this to be sensible. */
 422void ovs_flow_deferred_free(struct sw_flow *flow)
 423{
 424        call_rcu(&flow->rcu, rcu_free_flow_callback);
 425}
 426
 427/* Schedules 'sf_acts' to be freed after the next RCU grace period.
 428 * The caller must hold rcu_read_lock for this to be sensible. */
 429void ovs_flow_deferred_free_acts(struct sw_flow_actions *sf_acts)
 430{
 431        kfree_rcu(sf_acts, rcu);
 432}
 433
 434static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key)
 435{
 436        struct qtag_prefix {
 437                __be16 eth_type; /* ETH_P_8021Q */
 438                __be16 tci;
 439        };
 440        struct qtag_prefix *qp;
 441
 442        if (unlikely(skb->len < sizeof(struct qtag_prefix) + sizeof(__be16)))
 443                return 0;
 444
 445        if (unlikely(!pskb_may_pull(skb, sizeof(struct qtag_prefix) +
 446                                         sizeof(__be16))))
 447                return -ENOMEM;
 448
 449        qp = (struct qtag_prefix *) skb->data;
 450        key->eth.tci = qp->tci | htons(VLAN_TAG_PRESENT);
 451        __skb_pull(skb, sizeof(struct qtag_prefix));
 452
 453        return 0;
 454}
 455
 456static __be16 parse_ethertype(struct sk_buff *skb)
 457{
 458        struct llc_snap_hdr {
 459                u8  dsap;  /* Always 0xAA */
 460                u8  ssap;  /* Always 0xAA */
 461                u8  ctrl;
 462                u8  oui[3];
 463                __be16 ethertype;
 464        };
 465        struct llc_snap_hdr *llc;
 466        __be16 proto;
 467
 468        proto = *(__be16 *) skb->data;
 469        __skb_pull(skb, sizeof(__be16));
 470
 471        if (ntohs(proto) >= 1536)
 472                return proto;
 473
 474        if (skb->len < sizeof(struct llc_snap_hdr))
 475                return htons(ETH_P_802_2);
 476
 477        if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr))))
 478                return htons(0);
 479
 480        llc = (struct llc_snap_hdr *) skb->data;
 481        if (llc->dsap != LLC_SAP_SNAP ||
 482            llc->ssap != LLC_SAP_SNAP ||
 483            (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0)
 484                return htons(ETH_P_802_2);
 485
 486        __skb_pull(skb, sizeof(struct llc_snap_hdr));
 487        return llc->ethertype;
 488}
 489
 490static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
 491                        int *key_lenp, int nh_len)
 492{
 493        struct icmp6hdr *icmp = icmp6_hdr(skb);
 494        int error = 0;
 495        int key_len;
 496
 497        /* The ICMPv6 type and code fields use the 16-bit transport port
 498         * fields, so we need to store them in 16-bit network byte order.
 499         */
 500        key->ipv6.tp.src = htons(icmp->icmp6_type);
 501        key->ipv6.tp.dst = htons(icmp->icmp6_code);
 502        key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
 503
 504        if (icmp->icmp6_code == 0 &&
 505            (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION ||
 506             icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) {
 507                int icmp_len = skb->len - skb_transport_offset(skb);
 508                struct nd_msg *nd;
 509                int offset;
 510
 511                key_len = SW_FLOW_KEY_OFFSET(ipv6.nd);
 512
 513                /* In order to process neighbor discovery options, we need the
 514                 * entire packet.
 515                 */
 516                if (unlikely(icmp_len < sizeof(*nd)))
 517                        goto out;
 518                if (unlikely(skb_linearize(skb))) {
 519                        error = -ENOMEM;
 520                        goto out;
 521                }
 522
 523                nd = (struct nd_msg *)skb_transport_header(skb);
 524                key->ipv6.nd.target = nd->target;
 525                key_len = SW_FLOW_KEY_OFFSET(ipv6.nd);
 526
 527                icmp_len -= sizeof(*nd);
 528                offset = 0;
 529                while (icmp_len >= 8) {
 530                        struct nd_opt_hdr *nd_opt =
 531                                 (struct nd_opt_hdr *)(nd->opt + offset);
 532                        int opt_len = nd_opt->nd_opt_len * 8;
 533
 534                        if (unlikely(!opt_len || opt_len > icmp_len))
 535                                goto invalid;
 536
 537                        /* Store the link layer address if the appropriate
 538                         * option is provided.  It is considered an error if
 539                         * the same link layer option is specified twice.
 540                         */
 541                        if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR
 542                            && opt_len == 8) {
 543                                if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll)))
 544                                        goto invalid;
 545                                memcpy(key->ipv6.nd.sll,
 546                                    &nd->opt[offset+sizeof(*nd_opt)], ETH_ALEN);
 547                        } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR
 548                                   && opt_len == 8) {
 549                                if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll)))
 550                                        goto invalid;
 551                                memcpy(key->ipv6.nd.tll,
 552                                    &nd->opt[offset+sizeof(*nd_opt)], ETH_ALEN);
 553                        }
 554
 555                        icmp_len -= opt_len;
 556                        offset += opt_len;
 557                }
 558        }
 559
 560        goto out;
 561
 562invalid:
 563        memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target));
 564        memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll));
 565        memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll));
 566
 567out:
 568        *key_lenp = key_len;
 569        return error;
 570}
 571
 572/**
 573 * ovs_flow_extract - extracts a flow key from an Ethernet frame.
 574 * @skb: sk_buff that contains the frame, with skb->data pointing to the
 575 * Ethernet header
 576 * @in_port: port number on which @skb was received.
 577 * @key: output flow key
 578 * @key_lenp: length of output flow key
 579 *
 580 * The caller must ensure that skb->len >= ETH_HLEN.
 581 *
 582 * Returns 0 if successful, otherwise a negative errno value.
 583 *
 584 * Initializes @skb header pointers as follows:
 585 *
 586 *    - skb->mac_header: the Ethernet header.
 587 *
 588 *    - skb->network_header: just past the Ethernet header, or just past the
 589 *      VLAN header, to the first byte of the Ethernet payload.
 590 *
 591 *    - skb->transport_header: If key->dl_type is ETH_P_IP or ETH_P_IPV6
 592 *      on output, then just past the IP header, if one is present and
 593 *      of a correct length, otherwise the same as skb->network_header.
 594 *      For other key->dl_type values it is left untouched.
 595 */
 596int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key,
 597                 int *key_lenp)
 598{
 599        int error = 0;
 600        int key_len = SW_FLOW_KEY_OFFSET(eth);
 601        struct ethhdr *eth;
 602
 603        memset(key, 0, sizeof(*key));
 604
 605        key->phy.priority = skb->priority;
 606        key->phy.in_port = in_port;
 607
 608        skb_reset_mac_header(skb);
 609
 610        /* Link layer.  We are guaranteed to have at least the 14 byte Ethernet
 611         * header in the linear data area.
 612         */
 613        eth = eth_hdr(skb);
 614        memcpy(key->eth.src, eth->h_source, ETH_ALEN);
 615        memcpy(key->eth.dst, eth->h_dest, ETH_ALEN);
 616
 617        __skb_pull(skb, 2 * ETH_ALEN);
 618
 619        if (vlan_tx_tag_present(skb))
 620                key->eth.tci = htons(skb->vlan_tci);
 621        else if (eth->h_proto == htons(ETH_P_8021Q))
 622                if (unlikely(parse_vlan(skb, key)))
 623                        return -ENOMEM;
 624
 625        key->eth.type = parse_ethertype(skb);
 626        if (unlikely(key->eth.type == htons(0)))
 627                return -ENOMEM;
 628
 629        skb_reset_network_header(skb);
 630        __skb_push(skb, skb->data - skb_mac_header(skb));
 631
 632        /* Network layer. */
 633        if (key->eth.type == htons(ETH_P_IP)) {
 634                struct iphdr *nh;
 635                __be16 offset;
 636
 637                key_len = SW_FLOW_KEY_OFFSET(ipv4.addr);
 638
 639                error = check_iphdr(skb);
 640                if (unlikely(error)) {
 641                        if (error == -EINVAL) {
 642                                skb->transport_header = skb->network_header;
 643                                error = 0;
 644                        }
 645                        goto out;
 646                }
 647
 648                nh = ip_hdr(skb);
 649                key->ipv4.addr.src = nh->saddr;
 650                key->ipv4.addr.dst = nh->daddr;
 651
 652                key->ip.proto = nh->protocol;
 653                key->ip.tos = nh->tos;
 654                key->ip.ttl = nh->ttl;
 655
 656                offset = nh->frag_off & htons(IP_OFFSET);
 657                if (offset) {
 658                        key->ip.frag = OVS_FRAG_TYPE_LATER;
 659                        goto out;
 660                }
 661                if (nh->frag_off & htons(IP_MF) ||
 662                         skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
 663                        key->ip.frag = OVS_FRAG_TYPE_FIRST;
 664
 665                /* Transport layer. */
 666                if (key->ip.proto == IPPROTO_TCP) {
 667                        key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
 668                        if (tcphdr_ok(skb)) {
 669                                struct tcphdr *tcp = tcp_hdr(skb);
 670                                key->ipv4.tp.src = tcp->source;
 671                                key->ipv4.tp.dst = tcp->dest;
 672                        }
 673                } else if (key->ip.proto == IPPROTO_UDP) {
 674                        key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
 675                        if (udphdr_ok(skb)) {
 676                                struct udphdr *udp = udp_hdr(skb);
 677                                key->ipv4.tp.src = udp->source;
 678                                key->ipv4.tp.dst = udp->dest;
 679                        }
 680                } else if (key->ip.proto == IPPROTO_ICMP) {
 681                        key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
 682                        if (icmphdr_ok(skb)) {
 683                                struct icmphdr *icmp = icmp_hdr(skb);
 684                                /* The ICMP type and code fields use the 16-bit
 685                                 * transport port fields, so we need to store
 686                                 * them in 16-bit network byte order. */
 687                                key->ipv4.tp.src = htons(icmp->type);
 688                                key->ipv4.tp.dst = htons(icmp->code);
 689                        }
 690                }
 691
 692        } else if (key->eth.type == htons(ETH_P_ARP) && arphdr_ok(skb)) {
 693                struct arp_eth_header *arp;
 694
 695                arp = (struct arp_eth_header *)skb_network_header(skb);
 696
 697                if (arp->ar_hrd == htons(ARPHRD_ETHER)
 698                                && arp->ar_pro == htons(ETH_P_IP)
 699                                && arp->ar_hln == ETH_ALEN
 700                                && arp->ar_pln == 4) {
 701
 702                        /* We only match on the lower 8 bits of the opcode. */
 703                        if (ntohs(arp->ar_op) <= 0xff)
 704                                key->ip.proto = ntohs(arp->ar_op);
 705                        memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src));
 706                        memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst));
 707                        memcpy(key->ipv4.arp.sha, arp->ar_sha, ETH_ALEN);
 708                        memcpy(key->ipv4.arp.tha, arp->ar_tha, ETH_ALEN);
 709                        key_len = SW_FLOW_KEY_OFFSET(ipv4.arp);
 710                }
 711        } else if (key->eth.type == htons(ETH_P_IPV6)) {
 712                int nh_len;             /* IPv6 Header + Extensions */
 713
 714                nh_len = parse_ipv6hdr(skb, key, &key_len);
 715                if (unlikely(nh_len < 0)) {
 716                        if (nh_len == -EINVAL)
 717                                skb->transport_header = skb->network_header;
 718                        else
 719                                error = nh_len;
 720                        goto out;
 721                }
 722
 723                if (key->ip.frag == OVS_FRAG_TYPE_LATER)
 724                        goto out;
 725                if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
 726                        key->ip.frag = OVS_FRAG_TYPE_FIRST;
 727
 728                /* Transport layer. */
 729                if (key->ip.proto == NEXTHDR_TCP) {
 730                        key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
 731                        if (tcphdr_ok(skb)) {
 732                                struct tcphdr *tcp = tcp_hdr(skb);
 733                                key->ipv6.tp.src = tcp->source;
 734                                key->ipv6.tp.dst = tcp->dest;
 735                        }
 736                } else if (key->ip.proto == NEXTHDR_UDP) {
 737                        key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
 738                        if (udphdr_ok(skb)) {
 739                                struct udphdr *udp = udp_hdr(skb);
 740                                key->ipv6.tp.src = udp->source;
 741                                key->ipv6.tp.dst = udp->dest;
 742                        }
 743                } else if (key->ip.proto == NEXTHDR_ICMP) {
 744                        key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
 745                        if (icmp6hdr_ok(skb)) {
 746                                error = parse_icmpv6(skb, key, &key_len, nh_len);
 747                                if (error < 0)
 748                                        goto out;
 749                        }
 750                }
 751        }
 752
 753out:
 754        *key_lenp = key_len;
 755        return error;
 756}
 757
 758u32 ovs_flow_hash(const struct sw_flow_key *key, int key_len)
 759{
 760        return jhash2((u32 *)key, DIV_ROUND_UP(key_len, sizeof(u32)), 0);
 761}
 762
 763struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *table,
 764                                struct sw_flow_key *key, int key_len)
 765{
 766        struct sw_flow *flow;
 767        struct hlist_node *n;
 768        struct hlist_head *head;
 769        u32 hash;
 770
 771        hash = ovs_flow_hash(key, key_len);
 772
 773        head = find_bucket(table, hash);
 774        hlist_for_each_entry_rcu(flow, n, head, hash_node[table->node_ver]) {
 775
 776                if (flow->hash == hash &&
 777                    !memcmp(&flow->key, key, key_len)) {
 778                        return flow;
 779                }
 780        }
 781        return NULL;
 782}
 783
 784void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow)
 785{
 786        struct hlist_head *head;
 787
 788        head = find_bucket(table, flow->hash);
 789        hlist_add_head_rcu(&flow->hash_node[table->node_ver], head);
 790        table->count++;
 791}
 792
 793void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow)
 794{
 795        hlist_del_rcu(&flow->hash_node[table->node_ver]);
 796        table->count--;
 797        BUG_ON(table->count < 0);
 798}
 799
 800/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute.  */
 801const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
 802        [OVS_KEY_ATTR_ENCAP] = -1,
 803        [OVS_KEY_ATTR_PRIORITY] = sizeof(u32),
 804        [OVS_KEY_ATTR_IN_PORT] = sizeof(u32),
 805        [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
 806        [OVS_KEY_ATTR_VLAN] = sizeof(__be16),
 807        [OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16),
 808        [OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4),
 809        [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
 810        [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
 811        [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
 812        [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
 813        [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
 814        [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
 815        [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
 816};
 817
 818static int ipv4_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_len,
 819                                  const struct nlattr *a[], u32 *attrs)
 820{
 821        const struct ovs_key_icmp *icmp_key;
 822        const struct ovs_key_tcp *tcp_key;
 823        const struct ovs_key_udp *udp_key;
 824
 825        switch (swkey->ip.proto) {
 826        case IPPROTO_TCP:
 827                if (!(*attrs & (1 << OVS_KEY_ATTR_TCP)))
 828                        return -EINVAL;
 829                *attrs &= ~(1 << OVS_KEY_ATTR_TCP);
 830
 831                *key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
 832                tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
 833                swkey->ipv4.tp.src = tcp_key->tcp_src;
 834                swkey->ipv4.tp.dst = tcp_key->tcp_dst;
 835                break;
 836
 837        case IPPROTO_UDP:
 838                if (!(*attrs & (1 << OVS_KEY_ATTR_UDP)))
 839                        return -EINVAL;
 840                *attrs &= ~(1 << OVS_KEY_ATTR_UDP);
 841
 842                *key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
 843                udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
 844                swkey->ipv4.tp.src = udp_key->udp_src;
 845                swkey->ipv4.tp.dst = udp_key->udp_dst;
 846                break;
 847
 848        case IPPROTO_ICMP:
 849                if (!(*attrs & (1 << OVS_KEY_ATTR_ICMP)))
 850                        return -EINVAL;
 851                *attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
 852
 853                *key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
 854                icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
 855                swkey->ipv4.tp.src = htons(icmp_key->icmp_type);
 856                swkey->ipv4.tp.dst = htons(icmp_key->icmp_code);
 857                break;
 858        }
 859
 860        return 0;
 861}
 862
 863static int ipv6_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_len,
 864                                  const struct nlattr *a[], u32 *attrs)
 865{
 866        const struct ovs_key_icmpv6 *icmpv6_key;
 867        const struct ovs_key_tcp *tcp_key;
 868        const struct ovs_key_udp *udp_key;
 869
 870        switch (swkey->ip.proto) {
 871        case IPPROTO_TCP:
 872                if (!(*attrs & (1 << OVS_KEY_ATTR_TCP)))
 873                        return -EINVAL;
 874                *attrs &= ~(1 << OVS_KEY_ATTR_TCP);
 875
 876                *key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
 877                tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
 878                swkey->ipv6.tp.src = tcp_key->tcp_src;
 879                swkey->ipv6.tp.dst = tcp_key->tcp_dst;
 880                break;
 881
 882        case IPPROTO_UDP:
 883                if (!(*attrs & (1 << OVS_KEY_ATTR_UDP)))
 884                        return -EINVAL;
 885                *attrs &= ~(1 << OVS_KEY_ATTR_UDP);
 886
 887                *key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
 888                udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
 889                swkey->ipv6.tp.src = udp_key->udp_src;
 890                swkey->ipv6.tp.dst = udp_key->udp_dst;
 891                break;
 892
 893        case IPPROTO_ICMPV6:
 894                if (!(*attrs & (1 << OVS_KEY_ATTR_ICMPV6)))
 895                        return -EINVAL;
 896                *attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
 897
 898                *key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
 899                icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
 900                swkey->ipv6.tp.src = htons(icmpv6_key->icmpv6_type);
 901                swkey->ipv6.tp.dst = htons(icmpv6_key->icmpv6_code);
 902
 903                if (swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
 904                    swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
 905                        const struct ovs_key_nd *nd_key;
 906
 907                        if (!(*attrs & (1 << OVS_KEY_ATTR_ND)))
 908                                return -EINVAL;
 909                        *attrs &= ~(1 << OVS_KEY_ATTR_ND);
 910
 911                        *key_len = SW_FLOW_KEY_OFFSET(ipv6.nd);
 912                        nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
 913                        memcpy(&swkey->ipv6.nd.target, nd_key->nd_target,
 914                               sizeof(swkey->ipv6.nd.target));
 915                        memcpy(swkey->ipv6.nd.sll, nd_key->nd_sll, ETH_ALEN);
 916                        memcpy(swkey->ipv6.nd.tll, nd_key->nd_tll, ETH_ALEN);
 917                }
 918                break;
 919        }
 920
 921        return 0;
 922}
 923
 924static int parse_flow_nlattrs(const struct nlattr *attr,
 925                              const struct nlattr *a[], u32 *attrsp)
 926{
 927        const struct nlattr *nla;
 928        u32 attrs;
 929        int rem;
 930
 931        attrs = 0;
 932        nla_for_each_nested(nla, attr, rem) {
 933                u16 type = nla_type(nla);
 934                int expected_len;
 935
 936                if (type > OVS_KEY_ATTR_MAX || attrs & (1 << type))
 937                        return -EINVAL;
 938
 939                expected_len = ovs_key_lens[type];
 940                if (nla_len(nla) != expected_len && expected_len != -1)
 941                        return -EINVAL;
 942
 943                attrs |= 1 << type;
 944                a[type] = nla;
 945        }
 946        if (rem)
 947                return -EINVAL;
 948
 949        *attrsp = attrs;
 950        return 0;
 951}
 952
 953/**
 954 * ovs_flow_from_nlattrs - parses Netlink attributes into a flow key.
 955 * @swkey: receives the extracted flow key.
 956 * @key_lenp: number of bytes used in @swkey.
 957 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
 958 * sequence.
 959 */
 960int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp,
 961                      const struct nlattr *attr)
 962{
 963        const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
 964        const struct ovs_key_ethernet *eth_key;
 965        int key_len;
 966        u32 attrs;
 967        int err;
 968
 969        memset(swkey, 0, sizeof(struct sw_flow_key));
 970        key_len = SW_FLOW_KEY_OFFSET(eth);
 971
 972        err = parse_flow_nlattrs(attr, a, &attrs);
 973        if (err)
 974                return err;
 975
 976        /* Metadata attributes. */
 977        if (attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
 978                swkey->phy.priority = nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]);
 979                attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
 980        }
 981        if (attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
 982                u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
 983                if (in_port >= DP_MAX_PORTS)
 984                        return -EINVAL;
 985                swkey->phy.in_port = in_port;
 986                attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
 987        } else {
 988                swkey->phy.in_port = DP_MAX_PORTS;
 989        }
 990
 991        /* Data attributes. */
 992        if (!(attrs & (1 << OVS_KEY_ATTR_ETHERNET)))
 993                return -EINVAL;
 994        attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
 995
 996        eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
 997        memcpy(swkey->eth.src, eth_key->eth_src, ETH_ALEN);
 998        memcpy(swkey->eth.dst, eth_key->eth_dst, ETH_ALEN);
 999
1000        if (attrs & (1u << OVS_KEY_ATTR_ETHERTYPE) &&
1001            nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q)) {
1002                const struct nlattr *encap;
1003                __be16 tci;
1004
1005                if (attrs != ((1 << OVS_KEY_ATTR_VLAN) |
1006                              (1 << OVS_KEY_ATTR_ETHERTYPE) |
1007                              (1 << OVS_KEY_ATTR_ENCAP)))
1008                        return -EINVAL;
1009
1010                encap = a[OVS_KEY_ATTR_ENCAP];
1011                tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1012                if (tci & htons(VLAN_TAG_PRESENT)) {
1013                        swkey->eth.tci = tci;
1014
1015                        err = parse_flow_nlattrs(encap, a, &attrs);
1016                        if (err)
1017                                return err;
1018                } else if (!tci) {
1019                        /* Corner case for truncated 802.1Q header. */
1020                        if (nla_len(encap))
1021                                return -EINVAL;
1022
1023                        swkey->eth.type = htons(ETH_P_8021Q);
1024                        *key_lenp = key_len;
1025                        return 0;
1026                } else {
1027                        return -EINVAL;
1028                }
1029        }
1030
1031        if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1032                swkey->eth.type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1033                if (ntohs(swkey->eth.type) < 1536)
1034                        return -EINVAL;
1035                attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1036        } else {
1037                swkey->eth.type = htons(ETH_P_802_2);
1038        }
1039
1040        if (swkey->eth.type == htons(ETH_P_IP)) {
1041                const struct ovs_key_ipv4 *ipv4_key;
1042
1043                if (!(attrs & (1 << OVS_KEY_ATTR_IPV4)))
1044                        return -EINVAL;
1045                attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1046
1047                key_len = SW_FLOW_KEY_OFFSET(ipv4.addr);
1048                ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1049                if (ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX)
1050                        return -EINVAL;
1051                swkey->ip.proto = ipv4_key->ipv4_proto;
1052                swkey->ip.tos = ipv4_key->ipv4_tos;
1053                swkey->ip.ttl = ipv4_key->ipv4_ttl;
1054                swkey->ip.frag = ipv4_key->ipv4_frag;
1055                swkey->ipv4.addr.src = ipv4_key->ipv4_src;
1056                swkey->ipv4.addr.dst = ipv4_key->ipv4_dst;
1057
1058                if (swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1059                        err = ipv4_flow_from_nlattrs(swkey, &key_len, a, &attrs);
1060                        if (err)
1061                                return err;
1062                }
1063        } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1064                const struct ovs_key_ipv6 *ipv6_key;
1065
1066                if (!(attrs & (1 << OVS_KEY_ATTR_IPV6)))
1067                        return -EINVAL;
1068                attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1069
1070                key_len = SW_FLOW_KEY_OFFSET(ipv6.label);
1071                ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1072                if (ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX)
1073                        return -EINVAL;
1074                swkey->ipv6.label = ipv6_key->ipv6_label;
1075                swkey->ip.proto = ipv6_key->ipv6_proto;
1076                swkey->ip.tos = ipv6_key->ipv6_tclass;
1077                swkey->ip.ttl = ipv6_key->ipv6_hlimit;
1078                swkey->ip.frag = ipv6_key->ipv6_frag;
1079                memcpy(&swkey->ipv6.addr.src, ipv6_key->ipv6_src,
1080                       sizeof(swkey->ipv6.addr.src));
1081                memcpy(&swkey->ipv6.addr.dst, ipv6_key->ipv6_dst,
1082                       sizeof(swkey->ipv6.addr.dst));
1083
1084                if (swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1085                        err = ipv6_flow_from_nlattrs(swkey, &key_len, a, &attrs);
1086                        if (err)
1087                                return err;
1088                }
1089        } else if (swkey->eth.type == htons(ETH_P_ARP)) {
1090                const struct ovs_key_arp *arp_key;
1091
1092                if (!(attrs & (1 << OVS_KEY_ATTR_ARP)))
1093                        return -EINVAL;
1094                attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1095
1096                key_len = SW_FLOW_KEY_OFFSET(ipv4.arp);
1097                arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1098                swkey->ipv4.addr.src = arp_key->arp_sip;
1099                swkey->ipv4.addr.dst = arp_key->arp_tip;
1100                if (arp_key->arp_op & htons(0xff00))
1101                        return -EINVAL;
1102                swkey->ip.proto = ntohs(arp_key->arp_op);
1103                memcpy(swkey->ipv4.arp.sha, arp_key->arp_sha, ETH_ALEN);
1104                memcpy(swkey->ipv4.arp.tha, arp_key->arp_tha, ETH_ALEN);
1105        }
1106
1107        if (attrs)
1108                return -EINVAL;
1109        *key_lenp = key_len;
1110
1111        return 0;
1112}
1113
1114/**
1115 * ovs_flow_metadata_from_nlattrs - parses Netlink attributes into a flow key.
1116 * @in_port: receives the extracted input port.
1117 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1118 * sequence.
1119 *
1120 * This parses a series of Netlink attributes that form a flow key, which must
1121 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1122 * get the metadata, that is, the parts of the flow key that cannot be
1123 * extracted from the packet itself.
1124 */
1125int ovs_flow_metadata_from_nlattrs(u32 *priority, u16 *in_port,
1126                               const struct nlattr *attr)
1127{
1128        const struct nlattr *nla;
1129        int rem;
1130
1131        *in_port = DP_MAX_PORTS;
1132        *priority = 0;
1133
1134        nla_for_each_nested(nla, attr, rem) {
1135                int type = nla_type(nla);
1136
1137                if (type <= OVS_KEY_ATTR_MAX && ovs_key_lens[type] > 0) {
1138                        if (nla_len(nla) != ovs_key_lens[type])
1139                                return -EINVAL;
1140
1141                        switch (type) {
1142                        case OVS_KEY_ATTR_PRIORITY:
1143                                *priority = nla_get_u32(nla);
1144                                break;
1145
1146                        case OVS_KEY_ATTR_IN_PORT:
1147                                if (nla_get_u32(nla) >= DP_MAX_PORTS)
1148                                        return -EINVAL;
1149                                *in_port = nla_get_u32(nla);
1150                                break;
1151                        }
1152                }
1153        }
1154        if (rem)
1155                return -EINVAL;
1156        return 0;
1157}
1158
1159int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb)
1160{
1161        struct ovs_key_ethernet *eth_key;
1162        struct nlattr *nla, *encap;
1163
1164        if (swkey->phy.priority &&
1165            nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, swkey->phy.priority))
1166                goto nla_put_failure;
1167
1168        if (swkey->phy.in_port != DP_MAX_PORTS &&
1169            nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, swkey->phy.in_port))
1170                goto nla_put_failure;
1171
1172        nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1173        if (!nla)
1174                goto nla_put_failure;
1175        eth_key = nla_data(nla);
1176        memcpy(eth_key->eth_src, swkey->eth.src, ETH_ALEN);
1177        memcpy(eth_key->eth_dst, swkey->eth.dst, ETH_ALEN);
1178
1179        if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1180                if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, htons(ETH_P_8021Q)) ||
1181                    nla_put_be16(skb, OVS_KEY_ATTR_VLAN, swkey->eth.tci))
1182                        goto nla_put_failure;
1183                encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1184                if (!swkey->eth.tci)
1185                        goto unencap;
1186        } else {
1187                encap = NULL;
1188        }
1189
1190        if (swkey->eth.type == htons(ETH_P_802_2))
1191                goto unencap;
1192
1193        if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, swkey->eth.type))
1194                goto nla_put_failure;
1195
1196        if (swkey->eth.type == htons(ETH_P_IP)) {
1197                struct ovs_key_ipv4 *ipv4_key;
1198
1199                nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1200                if (!nla)
1201                        goto nla_put_failure;
1202                ipv4_key = nla_data(nla);
1203                ipv4_key->ipv4_src = swkey->ipv4.addr.src;
1204                ipv4_key->ipv4_dst = swkey->ipv4.addr.dst;
1205                ipv4_key->ipv4_proto = swkey->ip.proto;
1206                ipv4_key->ipv4_tos = swkey->ip.tos;
1207                ipv4_key->ipv4_ttl = swkey->ip.ttl;
1208                ipv4_key->ipv4_frag = swkey->ip.frag;
1209        } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1210                struct ovs_key_ipv6 *ipv6_key;
1211
1212                nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1213                if (!nla)
1214                        goto nla_put_failure;
1215                ipv6_key = nla_data(nla);
1216                memcpy(ipv6_key->ipv6_src, &swkey->ipv6.addr.src,
1217                                sizeof(ipv6_key->ipv6_src));
1218                memcpy(ipv6_key->ipv6_dst, &swkey->ipv6.addr.dst,
1219                                sizeof(ipv6_key->ipv6_dst));
1220                ipv6_key->ipv6_label = swkey->ipv6.label;
1221                ipv6_key->ipv6_proto = swkey->ip.proto;
1222                ipv6_key->ipv6_tclass = swkey->ip.tos;
1223                ipv6_key->ipv6_hlimit = swkey->ip.ttl;
1224                ipv6_key->ipv6_frag = swkey->ip.frag;
1225        } else if (swkey->eth.type == htons(ETH_P_ARP)) {
1226                struct ovs_key_arp *arp_key;
1227
1228                nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1229                if (!nla)
1230                        goto nla_put_failure;
1231                arp_key = nla_data(nla);
1232                memset(arp_key, 0, sizeof(struct ovs_key_arp));
1233                arp_key->arp_sip = swkey->ipv4.addr.src;
1234                arp_key->arp_tip = swkey->ipv4.addr.dst;
1235                arp_key->arp_op = htons(swkey->ip.proto);
1236                memcpy(arp_key->arp_sha, swkey->ipv4.arp.sha, ETH_ALEN);
1237                memcpy(arp_key->arp_tha, swkey->ipv4.arp.tha, ETH_ALEN);
1238        }
1239
1240        if ((swkey->eth.type == htons(ETH_P_IP) ||
1241             swkey->eth.type == htons(ETH_P_IPV6)) &&
1242             swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1243
1244                if (swkey->ip.proto == IPPROTO_TCP) {
1245                        struct ovs_key_tcp *tcp_key;
1246
1247                        nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1248                        if (!nla)
1249                                goto nla_put_failure;
1250                        tcp_key = nla_data(nla);
1251                        if (swkey->eth.type == htons(ETH_P_IP)) {
1252                                tcp_key->tcp_src = swkey->ipv4.tp.src;
1253                                tcp_key->tcp_dst = swkey->ipv4.tp.dst;
1254                        } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1255                                tcp_key->tcp_src = swkey->ipv6.tp.src;
1256                                tcp_key->tcp_dst = swkey->ipv6.tp.dst;
1257                        }
1258                } else if (swkey->ip.proto == IPPROTO_UDP) {
1259                        struct ovs_key_udp *udp_key;
1260
1261                        nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1262                        if (!nla)
1263                                goto nla_put_failure;
1264                        udp_key = nla_data(nla);
1265                        if (swkey->eth.type == htons(ETH_P_IP)) {
1266                                udp_key->udp_src = swkey->ipv4.tp.src;
1267                                udp_key->udp_dst = swkey->ipv4.tp.dst;
1268                        } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1269                                udp_key->udp_src = swkey->ipv6.tp.src;
1270                                udp_key->udp_dst = swkey->ipv6.tp.dst;
1271                        }
1272                } else if (swkey->eth.type == htons(ETH_P_IP) &&
1273                           swkey->ip.proto == IPPROTO_ICMP) {
1274                        struct ovs_key_icmp *icmp_key;
1275
1276                        nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1277                        if (!nla)
1278                                goto nla_put_failure;
1279                        icmp_key = nla_data(nla);
1280                        icmp_key->icmp_type = ntohs(swkey->ipv4.tp.src);
1281                        icmp_key->icmp_code = ntohs(swkey->ipv4.tp.dst);
1282                } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1283                           swkey->ip.proto == IPPROTO_ICMPV6) {
1284                        struct ovs_key_icmpv6 *icmpv6_key;
1285
1286                        nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1287                                                sizeof(*icmpv6_key));
1288                        if (!nla)
1289                                goto nla_put_failure;
1290                        icmpv6_key = nla_data(nla);
1291                        icmpv6_key->icmpv6_type = ntohs(swkey->ipv6.tp.src);
1292                        icmpv6_key->icmpv6_code = ntohs(swkey->ipv6.tp.dst);
1293
1294                        if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1295                            icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1296                                struct ovs_key_nd *nd_key;
1297
1298                                nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1299                                if (!nla)
1300                                        goto nla_put_failure;
1301                                nd_key = nla_data(nla);
1302                                memcpy(nd_key->nd_target, &swkey->ipv6.nd.target,
1303                                                        sizeof(nd_key->nd_target));
1304                                memcpy(nd_key->nd_sll, swkey->ipv6.nd.sll, ETH_ALEN);
1305                                memcpy(nd_key->nd_tll, swkey->ipv6.nd.tll, ETH_ALEN);
1306                        }
1307                }
1308        }
1309
1310unencap:
1311        if (encap)
1312                nla_nest_end(skb, encap);
1313
1314        return 0;
1315
1316nla_put_failure:
1317        return -EMSGSIZE;
1318}
1319
1320/* Initializes the flow module.
1321 * Returns zero if successful or a negative error code. */
1322int ovs_flow_init(void)
1323{
1324        flow_cache = kmem_cache_create("sw_flow", sizeof(struct sw_flow), 0,
1325                                        0, NULL);
1326        if (flow_cache == NULL)
1327                return -ENOMEM;
1328
1329        return 0;
1330}
1331
1332/* Uninitializes the flow module. */
1333void ovs_flow_exit(void)
1334{
1335        kmem_cache_destroy(flow_cache);
1336}
1337
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.