Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 1 | /* |
Raju Subramanian | caf2ee1 | 2012-05-03 18:55:23 -0700 | [diff] [blame] | 2 | * Copyright (c) 2007-2011 Nicira, Inc. |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 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> |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 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 | |
| 46 | static struct kmem_cache *flow_cache; |
| 47 | |
| 48 | static 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 | |
| 57 | static 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 | |
| 63 | static 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 | |
| 82 | static 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 | |
| 98 | static bool udphdr_ok(struct sk_buff *skb) |
| 99 | { |
| 100 | return pskb_may_pull(skb, skb_transport_offset(skb) + |
| 101 | sizeof(struct udphdr)); |
| 102 | } |
| 103 | |
| 104 | static bool icmphdr_ok(struct sk_buff *skb) |
| 105 | { |
| 106 | return pskb_may_pull(skb, skb_transport_offset(skb) + |
| 107 | sizeof(struct icmphdr)); |
| 108 | } |
| 109 | |
| 110 | u64 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 | |
| 127 | static 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 | |
| 172 | static 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 | |
| 181 | void ovs_flow_used(struct sw_flow *flow, struct sk_buff *skb) |
| 182 | { |
| 183 | u8 tcp_flags = 0; |
| 184 | |
Jesse Gross | c55177e | 2012-04-02 15:13:36 -0700 | [diff] [blame] | 185 | if ((flow->key.eth.type == htons(ETH_P_IP) || |
| 186 | flow->key.eth.type == htons(ETH_P_IPV6)) && |
Jesse Gross | bf32fec | 2012-04-02 14:26:27 -0700 | [diff] [blame] | 187 | flow->key.ip.proto == IPPROTO_TCP && |
| 188 | likely(skb->len >= skb_transport_offset(skb) + sizeof(struct tcphdr))) { |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 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 | |
| 201 | struct 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 | |
Pravin B Shelar | 15eac2a | 2012-08-23 12:40:54 -0700 | [diff] [blame] | 206 | if (actions_len > MAX_ACTIONS_BUFSIZE) |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 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 | |
| 218 | struct 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 | |
| 232 | static 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 | |
| 239 | static 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 | |
| 262 | static void free_buckets(struct flex_array *buckets) |
| 263 | { |
| 264 | flex_array_free(buckets); |
| 265 | } |
| 266 | |
| 267 | struct 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 | |
| 289 | void 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 | |
| 311 | skip_flows: |
| 312 | free_buckets(table->buckets); |
| 313 | kfree(table); |
| 314 | } |
| 315 | |
| 316 | static 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 | |
| 323 | void 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 | |
| 331 | struct 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 | |
| 358 | static 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 | |
| 380 | static 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 | |
| 393 | struct flow_table *ovs_flow_tbl_rehash(struct flow_table *table) |
| 394 | { |
| 395 | return __flow_tbl_rehash(table, table->n_buckets); |
| 396 | } |
| 397 | |
| 398 | struct flow_table *ovs_flow_tbl_expand(struct flow_table *table) |
| 399 | { |
| 400 | return __flow_tbl_rehash(table, table->n_buckets * 2); |
| 401 | } |
| 402 | |
| 403 | void 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. */ |
| 413 | static 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. */ |
| 422 | void ovs_flow_deferred_free(struct sw_flow *flow) |
| 423 | { |
| 424 | call_rcu(&flow->rcu, rcu_free_flow_callback); |
| 425 | } |
| 426 | |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 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. */ |
| 429 | void ovs_flow_deferred_free_acts(struct sw_flow_actions *sf_acts) |
| 430 | { |
Wei Yongjun | 80f0fd8 | 2012-08-26 18:20:45 +0000 | [diff] [blame] | 431 | kfree_rcu(sf_acts, rcu); |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 432 | } |
| 433 | |
| 434 | static 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 | |
| 456 | static __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 | |
| 490 | static 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 | |
| 562 | invalid: |
| 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 | |
| 567 | out: |
| 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 | */ |
| 596 | int 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 | |
| 706 | if (key->ip.proto == ARPOP_REQUEST |
| 707 | || key->ip.proto == ARPOP_REPLY) { |
| 708 | memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src)); |
| 709 | memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst)); |
| 710 | memcpy(key->ipv4.arp.sha, arp->ar_sha, ETH_ALEN); |
| 711 | memcpy(key->ipv4.arp.tha, arp->ar_tha, ETH_ALEN); |
| 712 | key_len = SW_FLOW_KEY_OFFSET(ipv4.arp); |
| 713 | } |
| 714 | } |
| 715 | } else if (key->eth.type == htons(ETH_P_IPV6)) { |
| 716 | int nh_len; /* IPv6 Header + Extensions */ |
| 717 | |
| 718 | nh_len = parse_ipv6hdr(skb, key, &key_len); |
| 719 | if (unlikely(nh_len < 0)) { |
| 720 | if (nh_len == -EINVAL) |
| 721 | skb->transport_header = skb->network_header; |
| 722 | else |
| 723 | error = nh_len; |
| 724 | goto out; |
| 725 | } |
| 726 | |
| 727 | if (key->ip.frag == OVS_FRAG_TYPE_LATER) |
| 728 | goto out; |
| 729 | if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
| 730 | key->ip.frag = OVS_FRAG_TYPE_FIRST; |
| 731 | |
| 732 | /* Transport layer. */ |
| 733 | if (key->ip.proto == NEXTHDR_TCP) { |
| 734 | key_len = SW_FLOW_KEY_OFFSET(ipv6.tp); |
| 735 | if (tcphdr_ok(skb)) { |
| 736 | struct tcphdr *tcp = tcp_hdr(skb); |
| 737 | key->ipv6.tp.src = tcp->source; |
| 738 | key->ipv6.tp.dst = tcp->dest; |
| 739 | } |
| 740 | } else if (key->ip.proto == NEXTHDR_UDP) { |
| 741 | key_len = SW_FLOW_KEY_OFFSET(ipv6.tp); |
| 742 | if (udphdr_ok(skb)) { |
| 743 | struct udphdr *udp = udp_hdr(skb); |
| 744 | key->ipv6.tp.src = udp->source; |
| 745 | key->ipv6.tp.dst = udp->dest; |
| 746 | } |
| 747 | } else if (key->ip.proto == NEXTHDR_ICMP) { |
| 748 | key_len = SW_FLOW_KEY_OFFSET(ipv6.tp); |
| 749 | if (icmp6hdr_ok(skb)) { |
| 750 | error = parse_icmpv6(skb, key, &key_len, nh_len); |
| 751 | if (error < 0) |
| 752 | goto out; |
| 753 | } |
| 754 | } |
| 755 | } |
| 756 | |
| 757 | out: |
| 758 | *key_lenp = key_len; |
| 759 | return error; |
| 760 | } |
| 761 | |
| 762 | u32 ovs_flow_hash(const struct sw_flow_key *key, int key_len) |
| 763 | { |
| 764 | return jhash2((u32 *)key, DIV_ROUND_UP(key_len, sizeof(u32)), 0); |
| 765 | } |
| 766 | |
| 767 | struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *table, |
| 768 | struct sw_flow_key *key, int key_len) |
| 769 | { |
| 770 | struct sw_flow *flow; |
| 771 | struct hlist_node *n; |
| 772 | struct hlist_head *head; |
| 773 | u32 hash; |
| 774 | |
| 775 | hash = ovs_flow_hash(key, key_len); |
| 776 | |
| 777 | head = find_bucket(table, hash); |
| 778 | hlist_for_each_entry_rcu(flow, n, head, hash_node[table->node_ver]) { |
| 779 | |
| 780 | if (flow->hash == hash && |
| 781 | !memcmp(&flow->key, key, key_len)) { |
| 782 | return flow; |
| 783 | } |
| 784 | } |
| 785 | return NULL; |
| 786 | } |
| 787 | |
| 788 | void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow) |
| 789 | { |
| 790 | struct hlist_head *head; |
| 791 | |
| 792 | head = find_bucket(table, flow->hash); |
| 793 | hlist_add_head_rcu(&flow->hash_node[table->node_ver], head); |
| 794 | table->count++; |
| 795 | } |
| 796 | |
| 797 | void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow) |
| 798 | { |
| 799 | hlist_del_rcu(&flow->hash_node[table->node_ver]); |
| 800 | table->count--; |
| 801 | BUG_ON(table->count < 0); |
| 802 | } |
| 803 | |
| 804 | /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */ |
| 805 | const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = { |
| 806 | [OVS_KEY_ATTR_ENCAP] = -1, |
| 807 | [OVS_KEY_ATTR_PRIORITY] = sizeof(u32), |
| 808 | [OVS_KEY_ATTR_IN_PORT] = sizeof(u32), |
| 809 | [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet), |
| 810 | [OVS_KEY_ATTR_VLAN] = sizeof(__be16), |
| 811 | [OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16), |
| 812 | [OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4), |
| 813 | [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6), |
| 814 | [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp), |
| 815 | [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp), |
| 816 | [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp), |
| 817 | [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6), |
| 818 | [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp), |
| 819 | [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd), |
| 820 | }; |
| 821 | |
| 822 | static int ipv4_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_len, |
| 823 | const struct nlattr *a[], u32 *attrs) |
| 824 | { |
| 825 | const struct ovs_key_icmp *icmp_key; |
| 826 | const struct ovs_key_tcp *tcp_key; |
| 827 | const struct ovs_key_udp *udp_key; |
| 828 | |
| 829 | switch (swkey->ip.proto) { |
| 830 | case IPPROTO_TCP: |
| 831 | if (!(*attrs & (1 << OVS_KEY_ATTR_TCP))) |
| 832 | return -EINVAL; |
| 833 | *attrs &= ~(1 << OVS_KEY_ATTR_TCP); |
| 834 | |
| 835 | *key_len = SW_FLOW_KEY_OFFSET(ipv4.tp); |
| 836 | tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]); |
| 837 | swkey->ipv4.tp.src = tcp_key->tcp_src; |
| 838 | swkey->ipv4.tp.dst = tcp_key->tcp_dst; |
| 839 | break; |
| 840 | |
| 841 | case IPPROTO_UDP: |
| 842 | if (!(*attrs & (1 << OVS_KEY_ATTR_UDP))) |
| 843 | return -EINVAL; |
| 844 | *attrs &= ~(1 << OVS_KEY_ATTR_UDP); |
| 845 | |
| 846 | *key_len = SW_FLOW_KEY_OFFSET(ipv4.tp); |
| 847 | udp_key = nla_data(a[OVS_KEY_ATTR_UDP]); |
| 848 | swkey->ipv4.tp.src = udp_key->udp_src; |
| 849 | swkey->ipv4.tp.dst = udp_key->udp_dst; |
| 850 | break; |
| 851 | |
| 852 | case IPPROTO_ICMP: |
| 853 | if (!(*attrs & (1 << OVS_KEY_ATTR_ICMP))) |
| 854 | return -EINVAL; |
| 855 | *attrs &= ~(1 << OVS_KEY_ATTR_ICMP); |
| 856 | |
| 857 | *key_len = SW_FLOW_KEY_OFFSET(ipv4.tp); |
| 858 | icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]); |
| 859 | swkey->ipv4.tp.src = htons(icmp_key->icmp_type); |
| 860 | swkey->ipv4.tp.dst = htons(icmp_key->icmp_code); |
| 861 | break; |
| 862 | } |
| 863 | |
| 864 | return 0; |
| 865 | } |
| 866 | |
| 867 | static int ipv6_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_len, |
| 868 | const struct nlattr *a[], u32 *attrs) |
| 869 | { |
| 870 | const struct ovs_key_icmpv6 *icmpv6_key; |
| 871 | const struct ovs_key_tcp *tcp_key; |
| 872 | const struct ovs_key_udp *udp_key; |
| 873 | |
| 874 | switch (swkey->ip.proto) { |
| 875 | case IPPROTO_TCP: |
| 876 | if (!(*attrs & (1 << OVS_KEY_ATTR_TCP))) |
| 877 | return -EINVAL; |
| 878 | *attrs &= ~(1 << OVS_KEY_ATTR_TCP); |
| 879 | |
| 880 | *key_len = SW_FLOW_KEY_OFFSET(ipv6.tp); |
| 881 | tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]); |
| 882 | swkey->ipv6.tp.src = tcp_key->tcp_src; |
| 883 | swkey->ipv6.tp.dst = tcp_key->tcp_dst; |
| 884 | break; |
| 885 | |
| 886 | case IPPROTO_UDP: |
| 887 | if (!(*attrs & (1 << OVS_KEY_ATTR_UDP))) |
| 888 | return -EINVAL; |
| 889 | *attrs &= ~(1 << OVS_KEY_ATTR_UDP); |
| 890 | |
| 891 | *key_len = SW_FLOW_KEY_OFFSET(ipv6.tp); |
| 892 | udp_key = nla_data(a[OVS_KEY_ATTR_UDP]); |
| 893 | swkey->ipv6.tp.src = udp_key->udp_src; |
| 894 | swkey->ipv6.tp.dst = udp_key->udp_dst; |
| 895 | break; |
| 896 | |
| 897 | case IPPROTO_ICMPV6: |
| 898 | if (!(*attrs & (1 << OVS_KEY_ATTR_ICMPV6))) |
| 899 | return -EINVAL; |
| 900 | *attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6); |
| 901 | |
| 902 | *key_len = SW_FLOW_KEY_OFFSET(ipv6.tp); |
| 903 | icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]); |
| 904 | swkey->ipv6.tp.src = htons(icmpv6_key->icmpv6_type); |
| 905 | swkey->ipv6.tp.dst = htons(icmpv6_key->icmpv6_code); |
| 906 | |
| 907 | if (swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) || |
| 908 | swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) { |
| 909 | const struct ovs_key_nd *nd_key; |
| 910 | |
| 911 | if (!(*attrs & (1 << OVS_KEY_ATTR_ND))) |
| 912 | return -EINVAL; |
| 913 | *attrs &= ~(1 << OVS_KEY_ATTR_ND); |
| 914 | |
| 915 | *key_len = SW_FLOW_KEY_OFFSET(ipv6.nd); |
| 916 | nd_key = nla_data(a[OVS_KEY_ATTR_ND]); |
| 917 | memcpy(&swkey->ipv6.nd.target, nd_key->nd_target, |
| 918 | sizeof(swkey->ipv6.nd.target)); |
| 919 | memcpy(swkey->ipv6.nd.sll, nd_key->nd_sll, ETH_ALEN); |
| 920 | memcpy(swkey->ipv6.nd.tll, nd_key->nd_tll, ETH_ALEN); |
| 921 | } |
| 922 | break; |
| 923 | } |
| 924 | |
| 925 | return 0; |
| 926 | } |
| 927 | |
| 928 | static int parse_flow_nlattrs(const struct nlattr *attr, |
| 929 | const struct nlattr *a[], u32 *attrsp) |
| 930 | { |
| 931 | const struct nlattr *nla; |
| 932 | u32 attrs; |
| 933 | int rem; |
| 934 | |
| 935 | attrs = 0; |
| 936 | nla_for_each_nested(nla, attr, rem) { |
| 937 | u16 type = nla_type(nla); |
| 938 | int expected_len; |
| 939 | |
| 940 | if (type > OVS_KEY_ATTR_MAX || attrs & (1 << type)) |
| 941 | return -EINVAL; |
| 942 | |
| 943 | expected_len = ovs_key_lens[type]; |
| 944 | if (nla_len(nla) != expected_len && expected_len != -1) |
| 945 | return -EINVAL; |
| 946 | |
| 947 | attrs |= 1 << type; |
| 948 | a[type] = nla; |
| 949 | } |
| 950 | if (rem) |
| 951 | return -EINVAL; |
| 952 | |
| 953 | *attrsp = attrs; |
| 954 | return 0; |
| 955 | } |
| 956 | |
| 957 | /** |
| 958 | * ovs_flow_from_nlattrs - parses Netlink attributes into a flow key. |
| 959 | * @swkey: receives the extracted flow key. |
| 960 | * @key_lenp: number of bytes used in @swkey. |
| 961 | * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute |
| 962 | * sequence. |
| 963 | */ |
| 964 | int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp, |
| 965 | const struct nlattr *attr) |
| 966 | { |
| 967 | const struct nlattr *a[OVS_KEY_ATTR_MAX + 1]; |
| 968 | const struct ovs_key_ethernet *eth_key; |
| 969 | int key_len; |
| 970 | u32 attrs; |
| 971 | int err; |
| 972 | |
| 973 | memset(swkey, 0, sizeof(struct sw_flow_key)); |
| 974 | key_len = SW_FLOW_KEY_OFFSET(eth); |
| 975 | |
| 976 | err = parse_flow_nlattrs(attr, a, &attrs); |
| 977 | if (err) |
| 978 | return err; |
| 979 | |
| 980 | /* Metadata attributes. */ |
| 981 | if (attrs & (1 << OVS_KEY_ATTR_PRIORITY)) { |
| 982 | swkey->phy.priority = nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]); |
| 983 | attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY); |
| 984 | } |
| 985 | if (attrs & (1 << OVS_KEY_ATTR_IN_PORT)) { |
| 986 | u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]); |
| 987 | if (in_port >= DP_MAX_PORTS) |
| 988 | return -EINVAL; |
| 989 | swkey->phy.in_port = in_port; |
| 990 | attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT); |
| 991 | } else { |
Pravin B Shelar | 15eac2a | 2012-08-23 12:40:54 -0700 | [diff] [blame] | 992 | swkey->phy.in_port = DP_MAX_PORTS; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 993 | } |
| 994 | |
| 995 | /* Data attributes. */ |
| 996 | if (!(attrs & (1 << OVS_KEY_ATTR_ETHERNET))) |
| 997 | return -EINVAL; |
| 998 | attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET); |
| 999 | |
| 1000 | eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]); |
| 1001 | memcpy(swkey->eth.src, eth_key->eth_src, ETH_ALEN); |
| 1002 | memcpy(swkey->eth.dst, eth_key->eth_dst, ETH_ALEN); |
| 1003 | |
| 1004 | if (attrs & (1u << OVS_KEY_ATTR_ETHERTYPE) && |
| 1005 | nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q)) { |
| 1006 | const struct nlattr *encap; |
| 1007 | __be16 tci; |
| 1008 | |
| 1009 | if (attrs != ((1 << OVS_KEY_ATTR_VLAN) | |
| 1010 | (1 << OVS_KEY_ATTR_ETHERTYPE) | |
| 1011 | (1 << OVS_KEY_ATTR_ENCAP))) |
| 1012 | return -EINVAL; |
| 1013 | |
| 1014 | encap = a[OVS_KEY_ATTR_ENCAP]; |
| 1015 | tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]); |
| 1016 | if (tci & htons(VLAN_TAG_PRESENT)) { |
| 1017 | swkey->eth.tci = tci; |
| 1018 | |
| 1019 | err = parse_flow_nlattrs(encap, a, &attrs); |
| 1020 | if (err) |
| 1021 | return err; |
| 1022 | } else if (!tci) { |
| 1023 | /* Corner case for truncated 802.1Q header. */ |
| 1024 | if (nla_len(encap)) |
| 1025 | return -EINVAL; |
| 1026 | |
| 1027 | swkey->eth.type = htons(ETH_P_8021Q); |
| 1028 | *key_lenp = key_len; |
| 1029 | return 0; |
| 1030 | } else { |
| 1031 | return -EINVAL; |
| 1032 | } |
| 1033 | } |
| 1034 | |
| 1035 | if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) { |
| 1036 | swkey->eth.type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]); |
| 1037 | if (ntohs(swkey->eth.type) < 1536) |
| 1038 | return -EINVAL; |
| 1039 | attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE); |
| 1040 | } else { |
| 1041 | swkey->eth.type = htons(ETH_P_802_2); |
| 1042 | } |
| 1043 | |
| 1044 | if (swkey->eth.type == htons(ETH_P_IP)) { |
| 1045 | const struct ovs_key_ipv4 *ipv4_key; |
| 1046 | |
| 1047 | if (!(attrs & (1 << OVS_KEY_ATTR_IPV4))) |
| 1048 | return -EINVAL; |
| 1049 | attrs &= ~(1 << OVS_KEY_ATTR_IPV4); |
| 1050 | |
| 1051 | key_len = SW_FLOW_KEY_OFFSET(ipv4.addr); |
| 1052 | ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]); |
| 1053 | if (ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) |
| 1054 | return -EINVAL; |
| 1055 | swkey->ip.proto = ipv4_key->ipv4_proto; |
| 1056 | swkey->ip.tos = ipv4_key->ipv4_tos; |
| 1057 | swkey->ip.ttl = ipv4_key->ipv4_ttl; |
| 1058 | swkey->ip.frag = ipv4_key->ipv4_frag; |
| 1059 | swkey->ipv4.addr.src = ipv4_key->ipv4_src; |
| 1060 | swkey->ipv4.addr.dst = ipv4_key->ipv4_dst; |
| 1061 | |
| 1062 | if (swkey->ip.frag != OVS_FRAG_TYPE_LATER) { |
| 1063 | err = ipv4_flow_from_nlattrs(swkey, &key_len, a, &attrs); |
| 1064 | if (err) |
| 1065 | return err; |
| 1066 | } |
| 1067 | } else if (swkey->eth.type == htons(ETH_P_IPV6)) { |
| 1068 | const struct ovs_key_ipv6 *ipv6_key; |
| 1069 | |
| 1070 | if (!(attrs & (1 << OVS_KEY_ATTR_IPV6))) |
| 1071 | return -EINVAL; |
| 1072 | attrs &= ~(1 << OVS_KEY_ATTR_IPV6); |
| 1073 | |
| 1074 | key_len = SW_FLOW_KEY_OFFSET(ipv6.label); |
| 1075 | ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]); |
| 1076 | if (ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) |
| 1077 | return -EINVAL; |
| 1078 | swkey->ipv6.label = ipv6_key->ipv6_label; |
| 1079 | swkey->ip.proto = ipv6_key->ipv6_proto; |
| 1080 | swkey->ip.tos = ipv6_key->ipv6_tclass; |
| 1081 | swkey->ip.ttl = ipv6_key->ipv6_hlimit; |
| 1082 | swkey->ip.frag = ipv6_key->ipv6_frag; |
| 1083 | memcpy(&swkey->ipv6.addr.src, ipv6_key->ipv6_src, |
| 1084 | sizeof(swkey->ipv6.addr.src)); |
| 1085 | memcpy(&swkey->ipv6.addr.dst, ipv6_key->ipv6_dst, |
| 1086 | sizeof(swkey->ipv6.addr.dst)); |
| 1087 | |
| 1088 | if (swkey->ip.frag != OVS_FRAG_TYPE_LATER) { |
| 1089 | err = ipv6_flow_from_nlattrs(swkey, &key_len, a, &attrs); |
| 1090 | if (err) |
| 1091 | return err; |
| 1092 | } |
| 1093 | } else if (swkey->eth.type == htons(ETH_P_ARP)) { |
| 1094 | const struct ovs_key_arp *arp_key; |
| 1095 | |
| 1096 | if (!(attrs & (1 << OVS_KEY_ATTR_ARP))) |
| 1097 | return -EINVAL; |
| 1098 | attrs &= ~(1 << OVS_KEY_ATTR_ARP); |
| 1099 | |
| 1100 | key_len = SW_FLOW_KEY_OFFSET(ipv4.arp); |
| 1101 | arp_key = nla_data(a[OVS_KEY_ATTR_ARP]); |
| 1102 | swkey->ipv4.addr.src = arp_key->arp_sip; |
| 1103 | swkey->ipv4.addr.dst = arp_key->arp_tip; |
| 1104 | if (arp_key->arp_op & htons(0xff00)) |
| 1105 | return -EINVAL; |
| 1106 | swkey->ip.proto = ntohs(arp_key->arp_op); |
| 1107 | memcpy(swkey->ipv4.arp.sha, arp_key->arp_sha, ETH_ALEN); |
| 1108 | memcpy(swkey->ipv4.arp.tha, arp_key->arp_tha, ETH_ALEN); |
| 1109 | } |
| 1110 | |
| 1111 | if (attrs) |
| 1112 | return -EINVAL; |
| 1113 | *key_lenp = key_len; |
| 1114 | |
| 1115 | return 0; |
| 1116 | } |
| 1117 | |
| 1118 | /** |
| 1119 | * ovs_flow_metadata_from_nlattrs - parses Netlink attributes into a flow key. |
| 1120 | * @in_port: receives the extracted input port. |
| 1121 | * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute |
| 1122 | * sequence. |
| 1123 | * |
| 1124 | * This parses a series of Netlink attributes that form a flow key, which must |
| 1125 | * take the same form accepted by flow_from_nlattrs(), but only enough of it to |
| 1126 | * get the metadata, that is, the parts of the flow key that cannot be |
| 1127 | * extracted from the packet itself. |
| 1128 | */ |
| 1129 | int ovs_flow_metadata_from_nlattrs(u32 *priority, u16 *in_port, |
| 1130 | const struct nlattr *attr) |
| 1131 | { |
| 1132 | const struct nlattr *nla; |
| 1133 | int rem; |
| 1134 | |
Pravin B Shelar | 15eac2a | 2012-08-23 12:40:54 -0700 | [diff] [blame] | 1135 | *in_port = DP_MAX_PORTS; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 1136 | *priority = 0; |
| 1137 | |
| 1138 | nla_for_each_nested(nla, attr, rem) { |
| 1139 | int type = nla_type(nla); |
| 1140 | |
| 1141 | if (type <= OVS_KEY_ATTR_MAX && ovs_key_lens[type] > 0) { |
| 1142 | if (nla_len(nla) != ovs_key_lens[type]) |
| 1143 | return -EINVAL; |
| 1144 | |
| 1145 | switch (type) { |
| 1146 | case OVS_KEY_ATTR_PRIORITY: |
| 1147 | *priority = nla_get_u32(nla); |
| 1148 | break; |
| 1149 | |
| 1150 | case OVS_KEY_ATTR_IN_PORT: |
| 1151 | if (nla_get_u32(nla) >= DP_MAX_PORTS) |
| 1152 | return -EINVAL; |
| 1153 | *in_port = nla_get_u32(nla); |
| 1154 | break; |
| 1155 | } |
| 1156 | } |
| 1157 | } |
| 1158 | if (rem) |
| 1159 | return -EINVAL; |
| 1160 | return 0; |
| 1161 | } |
| 1162 | |
| 1163 | int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb) |
| 1164 | { |
| 1165 | struct ovs_key_ethernet *eth_key; |
| 1166 | struct nlattr *nla, *encap; |
| 1167 | |
David S. Miller | 028d6a6 | 2012-03-29 23:20:48 -0400 | [diff] [blame] | 1168 | if (swkey->phy.priority && |
| 1169 | nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, swkey->phy.priority)) |
| 1170 | goto nla_put_failure; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 1171 | |
Pravin B Shelar | 15eac2a | 2012-08-23 12:40:54 -0700 | [diff] [blame] | 1172 | if (swkey->phy.in_port != DP_MAX_PORTS && |
David S. Miller | 028d6a6 | 2012-03-29 23:20:48 -0400 | [diff] [blame] | 1173 | nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, swkey->phy.in_port)) |
| 1174 | goto nla_put_failure; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 1175 | |
| 1176 | nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key)); |
| 1177 | if (!nla) |
| 1178 | goto nla_put_failure; |
| 1179 | eth_key = nla_data(nla); |
| 1180 | memcpy(eth_key->eth_src, swkey->eth.src, ETH_ALEN); |
| 1181 | memcpy(eth_key->eth_dst, swkey->eth.dst, ETH_ALEN); |
| 1182 | |
| 1183 | if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) { |
David S. Miller | 028d6a6 | 2012-03-29 23:20:48 -0400 | [diff] [blame] | 1184 | if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, htons(ETH_P_8021Q)) || |
| 1185 | nla_put_be16(skb, OVS_KEY_ATTR_VLAN, swkey->eth.tci)) |
| 1186 | goto nla_put_failure; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 1187 | encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP); |
| 1188 | if (!swkey->eth.tci) |
| 1189 | goto unencap; |
| 1190 | } else { |
| 1191 | encap = NULL; |
| 1192 | } |
| 1193 | |
| 1194 | if (swkey->eth.type == htons(ETH_P_802_2)) |
| 1195 | goto unencap; |
| 1196 | |
David S. Miller | 028d6a6 | 2012-03-29 23:20:48 -0400 | [diff] [blame] | 1197 | if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, swkey->eth.type)) |
| 1198 | goto nla_put_failure; |
Jesse Gross | ccb1352 | 2011-10-25 19:26:31 -0700 | [diff] [blame] | 1199 | |
| 1200 | if (swkey->eth.type == htons(ETH_P_IP)) { |
| 1201 | struct ovs_key_ipv4 *ipv4_key; |
| 1202 | |
| 1203 | nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key)); |
| 1204 | if (!nla) |
| 1205 | goto nla_put_failure; |
| 1206 | ipv4_key = nla_data(nla); |
| 1207 | ipv4_key->ipv4_src = swkey->ipv4.addr.src; |
| 1208 | ipv4_key->ipv4_dst = swkey->ipv4.addr.dst; |
| 1209 | ipv4_key->ipv4_proto = swkey->ip.proto; |
| 1210 | ipv4_key->ipv4_tos = swkey->ip.tos; |
| 1211 | ipv4_key->ipv4_ttl = swkey->ip.ttl; |
| 1212 | ipv4_key->ipv4_frag = swkey->ip.frag; |
| 1213 | } else if (swkey->eth.type == htons(ETH_P_IPV6)) { |
| 1214 | struct ovs_key_ipv6 *ipv6_key; |
| 1215 | |
| 1216 | nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key)); |
| 1217 | if (!nla) |
| 1218 | goto nla_put_failure; |
| 1219 | ipv6_key = nla_data(nla); |
| 1220 | memcpy(ipv6_key->ipv6_src, &swkey->ipv6.addr.src, |
| 1221 | sizeof(ipv6_key->ipv6_src)); |
| 1222 | memcpy(ipv6_key->ipv6_dst, &swkey->ipv6.addr.dst, |
| 1223 | sizeof(ipv6_key->ipv6_dst)); |
| 1224 | ipv6_key->ipv6_label = swkey->ipv6.label; |
| 1225 | ipv6_key->ipv6_proto = swkey->ip.proto; |
| 1226 | ipv6_key->ipv6_tclass = swkey->ip.tos; |
| 1227 | ipv6_key->ipv6_hlimit = swkey->ip.ttl; |
| 1228 | ipv6_key->ipv6_frag = swkey->ip.frag; |
| 1229 | } else if (swkey->eth.type == htons(ETH_P_ARP)) { |
| 1230 | struct ovs_key_arp *arp_key; |
| 1231 | |
| 1232 | nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key)); |
| 1233 | if (!nla) |
| 1234 | goto nla_put_failure; |
| 1235 | arp_key = nla_data(nla); |
| 1236 | memset(arp_key, 0, sizeof(struct ovs_key_arp)); |
| 1237 | arp_key->arp_sip = swkey->ipv4.addr.src; |
| 1238 | arp_key->arp_tip = swkey->ipv4.addr.dst; |
| 1239 | arp_key->arp_op = htons(swkey->ip.proto); |
| 1240 | memcpy(arp_key->arp_sha, swkey->ipv4.arp.sha, ETH_ALEN); |
| 1241 | memcpy(arp_key->arp_tha, swkey->ipv4.arp.tha, ETH_ALEN); |
| 1242 | } |
| 1243 | |
| 1244 | if ((swkey->eth.type == htons(ETH_P_IP) || |
| 1245 | swkey->eth.type == htons(ETH_P_IPV6)) && |
| 1246 | swkey->ip.frag != OVS_FRAG_TYPE_LATER) { |
| 1247 | |
| 1248 | if (swkey->ip.proto == IPPROTO_TCP) { |
| 1249 | struct ovs_key_tcp *tcp_key; |
| 1250 | |
| 1251 | nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key)); |
| 1252 | if (!nla) |
| 1253 | goto nla_put_failure; |
| 1254 | tcp_key = nla_data(nla); |
| 1255 | if (swkey->eth.type == htons(ETH_P_IP)) { |
| 1256 | tcp_key->tcp_src = swkey->ipv4.tp.src; |
| 1257 | tcp_key->tcp_dst = swkey->ipv4.tp.dst; |
| 1258 | } else if (swkey->eth.type == htons(ETH_P_IPV6)) { |
| 1259 | tcp_key->tcp_src = swkey->ipv6.tp.src; |
| 1260 | tcp_key->tcp_dst = swkey->ipv6.tp.dst; |
| 1261 | } |
| 1262 | } else if (swkey->ip.proto == IPPROTO_UDP) { |
| 1263 | struct ovs_key_udp *udp_key; |
| 1264 | |
| 1265 | nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key)); |
| 1266 | if (!nla) |
| 1267 | goto nla_put_failure; |
| 1268 | udp_key = nla_data(nla); |
| 1269 | if (swkey->eth.type == htons(ETH_P_IP)) { |
| 1270 | udp_key->udp_src = swkey->ipv4.tp.src; |
| 1271 | udp_key->udp_dst = swkey->ipv4.tp.dst; |
| 1272 | } else if (swkey->eth.type == htons(ETH_P_IPV6)) { |
| 1273 | udp_key->udp_src = swkey->ipv6.tp.src; |
| 1274 | udp_key->udp_dst = swkey->ipv6.tp.dst; |
| 1275 | } |
| 1276 | } else if (swkey->eth.type == htons(ETH_P_IP) && |
| 1277 | swkey->ip.proto == IPPROTO_ICMP) { |
| 1278 | struct ovs_key_icmp *icmp_key; |
| 1279 | |
| 1280 | nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key)); |
| 1281 | if (!nla) |
| 1282 | goto nla_put_failure; |
| 1283 | icmp_key = nla_data(nla); |
| 1284 | icmp_key->icmp_type = ntohs(swkey->ipv4.tp.src); |
| 1285 | icmp_key->icmp_code = ntohs(swkey->ipv4.tp.dst); |
| 1286 | } else if (swkey->eth.type == htons(ETH_P_IPV6) && |
| 1287 | swkey->ip.proto == IPPROTO_ICMPV6) { |
| 1288 | struct ovs_key_icmpv6 *icmpv6_key; |
| 1289 | |
| 1290 | nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6, |
| 1291 | sizeof(*icmpv6_key)); |
| 1292 | if (!nla) |
| 1293 | goto nla_put_failure; |
| 1294 | icmpv6_key = nla_data(nla); |
| 1295 | icmpv6_key->icmpv6_type = ntohs(swkey->ipv6.tp.src); |
| 1296 | icmpv6_key->icmpv6_code = ntohs(swkey->ipv6.tp.dst); |
| 1297 | |
| 1298 | if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION || |
| 1299 | icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) { |
| 1300 | struct ovs_key_nd *nd_key; |
| 1301 | |
| 1302 | nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key)); |
| 1303 | if (!nla) |
| 1304 | goto nla_put_failure; |
| 1305 | nd_key = nla_data(nla); |
| 1306 | memcpy(nd_key->nd_target, &swkey->ipv6.nd.target, |
| 1307 | sizeof(nd_key->nd_target)); |
| 1308 | memcpy(nd_key->nd_sll, swkey->ipv6.nd.sll, ETH_ALEN); |
| 1309 | memcpy(nd_key->nd_tll, swkey->ipv6.nd.tll, ETH_ALEN); |
| 1310 | } |
| 1311 | } |
| 1312 | } |
| 1313 | |
| 1314 | unencap: |
| 1315 | if (encap) |
| 1316 | nla_nest_end(skb, encap); |
| 1317 | |
| 1318 | return 0; |
| 1319 | |
| 1320 | nla_put_failure: |
| 1321 | return -EMSGSIZE; |
| 1322 | } |
| 1323 | |
| 1324 | /* Initializes the flow module. |
| 1325 | * Returns zero if successful or a negative error code. */ |
| 1326 | int ovs_flow_init(void) |
| 1327 | { |
| 1328 | flow_cache = kmem_cache_create("sw_flow", sizeof(struct sw_flow), 0, |
| 1329 | 0, NULL); |
| 1330 | if (flow_cache == NULL) |
| 1331 | return -ENOMEM; |
| 1332 | |
| 1333 | return 0; |
| 1334 | } |
| 1335 | |
| 1336 | /* Uninitializes the flow module. */ |
| 1337 | void ovs_flow_exit(void) |
| 1338 | { |
| 1339 | kmem_cache_destroy(flow_cache); |
| 1340 | } |