2 * Copyright (c) 2007-2014 Nicira, Inc.
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.
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.
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
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
24 #include <linux/uaccess.h>
25 #include <linux/netdevice.h>
26 #include <linux/etherdevice.h>
27 #include <linux/if_ether.h>
28 #include <linux/if_vlan.h>
29 #include <net/llc_pdu.h>
30 #include <linux/kernel.h>
31 #include <linux/jhash.h>
32 #include <linux/jiffies.h>
33 #include <linux/llc.h>
34 #include <linux/module.h>
36 #include <linux/rcupdate.h>
37 #include <linux/if_arp.h>
39 #include <linux/ipv6.h>
40 #include <linux/sctp.h>
41 #include <linux/tcp.h>
42 #include <linux/udp.h>
43 #include <linux/icmp.h>
44 #include <linux/icmpv6.h>
45 #include <linux/rculist.h>
46 #include <net/geneve.h>
48 #include <net/ip_tunnels.h>
50 #include <net/ndisc.h>
52 #include "flow_netlink.h"
54 static void update_range(struct sw_flow_match *match,
55 size_t offset, size_t size, bool is_mask)
57 struct sw_flow_key_range *range;
58 size_t start = rounddown(offset, sizeof(long));
59 size_t end = roundup(offset + size, sizeof(long));
62 range = &match->range;
64 range = &match->mask->range;
66 if (range->start == range->end) {
72 if (range->start > start)
79 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
81 update_range(match, offsetof(struct sw_flow_key, field), \
82 sizeof((match)->key->field), is_mask); \
84 (match)->mask->key.field = value; \
86 (match)->key->field = value; \
89 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
91 update_range(match, offset, len, is_mask); \
93 memcpy((u8 *)&(match)->mask->key + offset, value_p, len);\
95 memcpy((u8 *)(match)->key + offset, value_p, len); \
98 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
99 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
100 value_p, len, is_mask)
102 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
104 update_range(match, offsetof(struct sw_flow_key, field), \
105 sizeof((match)->key->field), is_mask); \
107 memset((u8 *)&(match)->mask->key.field, value, \
108 sizeof((match)->mask->key.field)); \
110 memset((u8 *)&(match)->key->field, value, \
111 sizeof((match)->key->field)); \
114 static bool match_validate(const struct sw_flow_match *match,
115 u64 key_attrs, u64 mask_attrs)
117 u64 key_expected = 1ULL << OVS_KEY_ATTR_ETHERNET;
118 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
120 /* The following mask attributes allowed only if they
121 * pass the validation tests. */
122 mask_allowed &= ~((1ULL << OVS_KEY_ATTR_IPV4)
123 | (1ULL << OVS_KEY_ATTR_IPV6)
124 | (1ULL << OVS_KEY_ATTR_TCP)
125 | (1ULL << OVS_KEY_ATTR_TCP_FLAGS)
126 | (1ULL << OVS_KEY_ATTR_UDP)
127 | (1ULL << OVS_KEY_ATTR_SCTP)
128 | (1ULL << OVS_KEY_ATTR_ICMP)
129 | (1ULL << OVS_KEY_ATTR_ICMPV6)
130 | (1ULL << OVS_KEY_ATTR_ARP)
131 | (1ULL << OVS_KEY_ATTR_ND)
132 | (1ULL << OVS_KEY_ATTR_MPLS));
134 /* Always allowed mask fields. */
135 mask_allowed |= ((1ULL << OVS_KEY_ATTR_TUNNEL)
136 | (1ULL << OVS_KEY_ATTR_IN_PORT)
137 | (1ULL << OVS_KEY_ATTR_ETHERTYPE));
139 /* Check key attributes. */
140 if (match->key->eth.type == htons(ETH_P_ARP)
141 || match->key->eth.type == htons(ETH_P_RARP)) {
142 key_expected |= 1ULL << OVS_KEY_ATTR_ARP;
143 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
144 mask_allowed |= 1ULL << OVS_KEY_ATTR_ARP;
148 if (eth_p_mpls(match->key->eth.type)) {
149 key_expected |= 1ULL << OVS_KEY_ATTR_MPLS;
150 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
151 mask_allowed |= 1ULL << OVS_KEY_ATTR_MPLS;
154 if (match->key->eth.type == htons(ETH_P_IP)) {
155 key_expected |= 1ULL << OVS_KEY_ATTR_IPV4;
156 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
157 mask_allowed |= 1ULL << OVS_KEY_ATTR_IPV4;
159 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
160 if (match->key->ip.proto == IPPROTO_UDP) {
161 key_expected |= 1ULL << OVS_KEY_ATTR_UDP;
162 if (match->mask && (match->mask->key.ip.proto == 0xff))
163 mask_allowed |= 1ULL << OVS_KEY_ATTR_UDP;
166 if (match->key->ip.proto == IPPROTO_SCTP) {
167 key_expected |= 1ULL << OVS_KEY_ATTR_SCTP;
168 if (match->mask && (match->mask->key.ip.proto == 0xff))
169 mask_allowed |= 1ULL << OVS_KEY_ATTR_SCTP;
172 if (match->key->ip.proto == IPPROTO_TCP) {
173 key_expected |= 1ULL << OVS_KEY_ATTR_TCP;
174 key_expected |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
175 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
176 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP;
177 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
181 if (match->key->ip.proto == IPPROTO_ICMP) {
182 key_expected |= 1ULL << OVS_KEY_ATTR_ICMP;
183 if (match->mask && (match->mask->key.ip.proto == 0xff))
184 mask_allowed |= 1ULL << OVS_KEY_ATTR_ICMP;
189 if (match->key->eth.type == htons(ETH_P_IPV6)) {
190 key_expected |= 1ULL << OVS_KEY_ATTR_IPV6;
191 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
192 mask_allowed |= 1ULL << OVS_KEY_ATTR_IPV6;
194 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
195 if (match->key->ip.proto == IPPROTO_UDP) {
196 key_expected |= 1ULL << OVS_KEY_ATTR_UDP;
197 if (match->mask && (match->mask->key.ip.proto == 0xff))
198 mask_allowed |= 1ULL << OVS_KEY_ATTR_UDP;
201 if (match->key->ip.proto == IPPROTO_SCTP) {
202 key_expected |= 1ULL << OVS_KEY_ATTR_SCTP;
203 if (match->mask && (match->mask->key.ip.proto == 0xff))
204 mask_allowed |= 1ULL << OVS_KEY_ATTR_SCTP;
207 if (match->key->ip.proto == IPPROTO_TCP) {
208 key_expected |= 1ULL << OVS_KEY_ATTR_TCP;
209 key_expected |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
210 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
211 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP;
212 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
216 if (match->key->ip.proto == IPPROTO_ICMPV6) {
217 key_expected |= 1ULL << OVS_KEY_ATTR_ICMPV6;
218 if (match->mask && (match->mask->key.ip.proto == 0xff))
219 mask_allowed |= 1ULL << OVS_KEY_ATTR_ICMPV6;
221 if (match->key->tp.src ==
222 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
223 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
224 key_expected |= 1ULL << OVS_KEY_ATTR_ND;
225 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
226 mask_allowed |= 1ULL << OVS_KEY_ATTR_ND;
232 if ((key_attrs & key_expected) != key_expected) {
233 /* Key attributes check failed. */
234 OVS_NLERR("Missing expected key attributes (key_attrs=%llx, expected=%llx).\n",
235 (unsigned long long)key_attrs, (unsigned long long)key_expected);
239 if ((mask_attrs & mask_allowed) != mask_attrs) {
240 /* Mask attributes check failed. */
241 OVS_NLERR("Contain more than allowed mask fields (mask_attrs=%llx, mask_allowed=%llx).\n",
242 (unsigned long long)mask_attrs, (unsigned long long)mask_allowed);
249 size_t ovs_tun_key_attr_size(void)
251 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
252 * updating this function. */
253 return nla_total_size(8) /* OVS_TUNNEL_KEY_ATTR_ID */
254 + nla_total_size(4) /* OVS_TUNNEL_KEY_ATTR_IPV4_SRC */
255 + nla_total_size(4) /* OVS_TUNNEL_KEY_ATTR_IPV4_DST */
256 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
257 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
258 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
259 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
260 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
261 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
262 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
263 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
266 size_t ovs_key_attr_size(void)
268 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
269 * updating this function. */
270 BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO != 22);
272 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
273 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
274 + ovs_tun_key_attr_size()
275 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
276 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
277 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
278 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
279 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
280 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
281 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
282 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
283 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
284 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
285 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
286 + nla_total_size(28); /* OVS_KEY_ATTR_ND */
289 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
290 static const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
291 [OVS_KEY_ATTR_ENCAP] = -1,
292 [OVS_KEY_ATTR_PRIORITY] = sizeof(u32),
293 [OVS_KEY_ATTR_IN_PORT] = sizeof(u32),
294 [OVS_KEY_ATTR_SKB_MARK] = sizeof(u32),
295 [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
296 [OVS_KEY_ATTR_VLAN] = sizeof(__be16),
297 [OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16),
298 [OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4),
299 [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
300 [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
301 [OVS_KEY_ATTR_TCP_FLAGS] = sizeof(__be16),
302 [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
303 [OVS_KEY_ATTR_SCTP] = sizeof(struct ovs_key_sctp),
304 [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
305 [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
306 [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
307 [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
308 [OVS_KEY_ATTR_DP_HASH] = sizeof(u32),
309 [OVS_KEY_ATTR_RECIRC_ID] = sizeof(u32),
310 [OVS_KEY_ATTR_TUNNEL] = -1,
311 [OVS_KEY_ATTR_MPLS] = sizeof(struct ovs_key_mpls),
314 static bool is_all_zero(const u8 *fp, size_t size)
321 for (i = 0; i < size; i++)
328 static int __parse_flow_nlattrs(const struct nlattr *attr,
329 const struct nlattr *a[],
330 u64 *attrsp, bool nz)
332 const struct nlattr *nla;
337 nla_for_each_nested(nla, attr, rem) {
338 u16 type = nla_type(nla);
341 if (type > OVS_KEY_ATTR_MAX) {
342 OVS_NLERR("Unknown key attribute (type=%d, max=%d).\n",
343 type, OVS_KEY_ATTR_MAX);
347 if (attrs & (1ULL << type)) {
348 OVS_NLERR("Duplicate key attribute (type %d).\n", type);
352 expected_len = ovs_key_lens[type];
353 if (nla_len(nla) != expected_len && expected_len != -1) {
354 OVS_NLERR("Key attribute has unexpected length (type=%d"
355 ", length=%d, expected=%d).\n", type,
356 nla_len(nla), expected_len);
360 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
361 attrs |= 1ULL << type;
366 OVS_NLERR("Message has %d unknown bytes.\n", rem);
374 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
375 const struct nlattr *a[], u64 *attrsp)
377 return __parse_flow_nlattrs(attr, a, attrsp, true);
380 static int parse_flow_nlattrs(const struct nlattr *attr,
381 const struct nlattr *a[], u64 *attrsp)
383 return __parse_flow_nlattrs(attr, a, attrsp, false);
386 static int ipv4_tun_from_nlattr(const struct nlattr *attr,
387 struct sw_flow_match *match, bool is_mask)
392 __be16 tun_flags = 0;
394 nla_for_each_nested(a, attr, rem) {
395 int type = nla_type(a);
396 static const u32 ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
397 [OVS_TUNNEL_KEY_ATTR_ID] = sizeof(u64),
398 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = sizeof(u32),
399 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = sizeof(u32),
400 [OVS_TUNNEL_KEY_ATTR_TOS] = 1,
401 [OVS_TUNNEL_KEY_ATTR_TTL] = 1,
402 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = 0,
403 [OVS_TUNNEL_KEY_ATTR_CSUM] = 0,
404 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = sizeof(u16),
405 [OVS_TUNNEL_KEY_ATTR_TP_DST] = sizeof(u16),
406 [OVS_TUNNEL_KEY_ATTR_OAM] = 0,
407 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = -1,
410 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
411 OVS_NLERR("Unknown IPv4 tunnel attribute (type=%d, max=%d).\n",
412 type, OVS_TUNNEL_KEY_ATTR_MAX);
416 if (ovs_tunnel_key_lens[type] != nla_len(a) &&
417 ovs_tunnel_key_lens[type] != -1) {
418 OVS_NLERR("IPv4 tunnel attribute type has unexpected "
419 " length (type=%d, length=%d, expected=%d).\n",
420 type, nla_len(a), ovs_tunnel_key_lens[type]);
425 case OVS_TUNNEL_KEY_ATTR_ID:
426 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
427 nla_get_be64(a), is_mask);
428 tun_flags |= TUNNEL_KEY;
430 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
431 SW_FLOW_KEY_PUT(match, tun_key.ipv4_src,
432 nla_get_be32(a), is_mask);
434 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
435 SW_FLOW_KEY_PUT(match, tun_key.ipv4_dst,
436 nla_get_be32(a), is_mask);
438 case OVS_TUNNEL_KEY_ATTR_TOS:
439 SW_FLOW_KEY_PUT(match, tun_key.ipv4_tos,
440 nla_get_u8(a), is_mask);
442 case OVS_TUNNEL_KEY_ATTR_TTL:
443 SW_FLOW_KEY_PUT(match, tun_key.ipv4_ttl,
444 nla_get_u8(a), is_mask);
447 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
448 tun_flags |= TUNNEL_DONT_FRAGMENT;
450 case OVS_TUNNEL_KEY_ATTR_CSUM:
451 tun_flags |= TUNNEL_CSUM;
453 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
454 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
455 nla_get_be16(a), is_mask);
457 case OVS_TUNNEL_KEY_ATTR_TP_DST:
458 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
459 nla_get_be16(a), is_mask);
461 case OVS_TUNNEL_KEY_ATTR_OAM:
462 tun_flags |= TUNNEL_OAM;
464 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
465 tun_flags |= TUNNEL_OPTIONS_PRESENT;
466 if (nla_len(a) > sizeof(match->key->tun_opts)) {
467 OVS_NLERR("Geneve option length exceeds "
468 "maximum size (len %d, max %zu).\n",
470 sizeof(match->key->tun_opts));
474 if (nla_len(a) % 4 != 0) {
475 OVS_NLERR("Geneve option length is not "
476 "a multiple of 4 (len %d).\n",
481 /* We need to record the length of the options passed
482 * down, otherwise packets with the same format but
483 * additional options will be silently matched.
486 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
489 /* This is somewhat unusual because it looks at
490 * both the key and mask while parsing the
491 * attributes (and by extension assumes the key
492 * is parsed first). Normally, we would verify
493 * that each is the correct length and that the
494 * attributes line up in the validate function.
495 * However, that is difficult because this is
496 * variable length and we won't have the
499 if (match->key->tun_opts_len != nla_len(a)) {
500 OVS_NLERR("Geneve option key length (%d)"
501 " is different from mask length (%d).",
502 match->key->tun_opts_len, nla_len(a));
506 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff,
510 SW_FLOW_KEY_MEMCPY_OFFSET(match,
511 (unsigned long)GENEVE_OPTS((struct sw_flow_key *)0,
513 nla_data(a), nla_len(a), is_mask);
516 OVS_NLERR("Unknown IPv4 tunnel attribute (%d).\n", type);
521 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
524 OVS_NLERR("IPv4 tunnel attribute has %d unknown bytes.\n", rem);
529 if (!match->key->tun_key.ipv4_dst) {
530 OVS_NLERR("IPv4 tunnel destination address is zero.\n");
535 OVS_NLERR("IPv4 tunnel TTL not specified.\n");
543 static int __ipv4_tun_to_nlattr(struct sk_buff *skb,
544 const struct ovs_key_ipv4_tunnel *output,
545 const struct geneve_opt *tun_opts,
546 int swkey_tun_opts_len)
548 if (output->tun_flags & TUNNEL_KEY &&
549 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
551 if (output->ipv4_src &&
552 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC, output->ipv4_src))
554 if (output->ipv4_dst &&
555 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST, output->ipv4_dst))
557 if (output->ipv4_tos &&
558 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->ipv4_tos))
560 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ipv4_ttl))
562 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
563 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
565 if ((output->tun_flags & TUNNEL_CSUM) &&
566 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
568 if (output->tp_src &&
569 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
571 if (output->tp_dst &&
572 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
574 if ((output->tun_flags & TUNNEL_OAM) &&
575 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
578 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
579 swkey_tun_opts_len, tun_opts))
586 static int ipv4_tun_to_nlattr(struct sk_buff *skb,
587 const struct ovs_key_ipv4_tunnel *output,
588 const struct geneve_opt *tun_opts,
589 int swkey_tun_opts_len)
594 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
598 err = __ipv4_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len);
602 nla_nest_end(skb, nla);
606 int ovs_nla_put_egress_tunnel_key(struct sk_buff *skb,
607 const struct ovs_tunnel_info *egress_tun_info)
609 return __ipv4_tun_to_nlattr(skb, &egress_tun_info->tunnel,
610 egress_tun_info->options,
611 egress_tun_info->options_len);
614 static int metadata_from_nlattrs(struct sw_flow_match *match, u64 *attrs,
615 const struct nlattr **a, bool is_mask)
617 if (*attrs & (1ULL << OVS_KEY_ATTR_DP_HASH)) {
618 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
620 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
621 *attrs &= ~(1ULL << OVS_KEY_ATTR_DP_HASH);
624 if (*attrs & (1ULL << OVS_KEY_ATTR_RECIRC_ID)) {
625 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
627 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
628 *attrs &= ~(1ULL << OVS_KEY_ATTR_RECIRC_ID);
631 if (*attrs & (1ULL << OVS_KEY_ATTR_PRIORITY)) {
632 SW_FLOW_KEY_PUT(match, phy.priority,
633 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
634 *attrs &= ~(1ULL << OVS_KEY_ATTR_PRIORITY);
637 if (*attrs & (1ULL << OVS_KEY_ATTR_IN_PORT)) {
638 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
641 in_port = 0xffffffff; /* Always exact match in_port. */
642 } else if (in_port >= DP_MAX_PORTS) {
643 OVS_NLERR("Input port (%d) exceeds maximum allowable (%d).\n",
644 in_port, DP_MAX_PORTS);
648 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
649 *attrs &= ~(1ULL << OVS_KEY_ATTR_IN_PORT);
650 } else if (!is_mask) {
651 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
654 if (*attrs & (1ULL << OVS_KEY_ATTR_SKB_MARK)) {
655 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
657 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
658 *attrs &= ~(1ULL << OVS_KEY_ATTR_SKB_MARK);
660 if (*attrs & (1ULL << OVS_KEY_ATTR_TUNNEL)) {
661 if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
664 *attrs &= ~(1ULL << OVS_KEY_ATTR_TUNNEL);
669 static int ovs_key_from_nlattrs(struct sw_flow_match *match, u64 attrs,
670 const struct nlattr **a, bool is_mask)
674 err = metadata_from_nlattrs(match, &attrs, a, is_mask);
678 if (attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) {
679 const struct ovs_key_ethernet *eth_key;
681 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
682 SW_FLOW_KEY_MEMCPY(match, eth.src,
683 eth_key->eth_src, ETH_ALEN, is_mask);
684 SW_FLOW_KEY_MEMCPY(match, eth.dst,
685 eth_key->eth_dst, ETH_ALEN, is_mask);
686 attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERNET);
689 if (attrs & (1ULL << OVS_KEY_ATTR_VLAN)) {
692 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
693 if (!(tci & htons(VLAN_TAG_PRESENT))) {
695 OVS_NLERR("VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.\n");
697 OVS_NLERR("VLAN TCI does not have VLAN_TAG_PRESENT bit set.\n");
702 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
703 attrs &= ~(1ULL << OVS_KEY_ATTR_VLAN);
706 if (attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) {
709 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
711 /* Always exact match EtherType. */
712 eth_type = htons(0xffff);
713 } else if (ntohs(eth_type) < ETH_P_802_3_MIN) {
714 OVS_NLERR("EtherType is less than minimum (type=%x, min=%x).\n",
715 ntohs(eth_type), ETH_P_802_3_MIN);
719 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
720 attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
721 } else if (!is_mask) {
722 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
725 if (attrs & (1ULL << OVS_KEY_ATTR_IPV4)) {
726 const struct ovs_key_ipv4 *ipv4_key;
728 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
729 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
730 OVS_NLERR("Unknown IPv4 fragment type (value=%d, max=%d).\n",
731 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
734 SW_FLOW_KEY_PUT(match, ip.proto,
735 ipv4_key->ipv4_proto, is_mask);
736 SW_FLOW_KEY_PUT(match, ip.tos,
737 ipv4_key->ipv4_tos, is_mask);
738 SW_FLOW_KEY_PUT(match, ip.ttl,
739 ipv4_key->ipv4_ttl, is_mask);
740 SW_FLOW_KEY_PUT(match, ip.frag,
741 ipv4_key->ipv4_frag, is_mask);
742 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
743 ipv4_key->ipv4_src, is_mask);
744 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
745 ipv4_key->ipv4_dst, is_mask);
746 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV4);
749 if (attrs & (1ULL << OVS_KEY_ATTR_IPV6)) {
750 const struct ovs_key_ipv6 *ipv6_key;
752 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
753 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
754 OVS_NLERR("Unknown IPv6 fragment type (value=%d, max=%d).\n",
755 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
758 if (ipv6_key->ipv6_label & htonl(0xFFF00000)) {
759 OVS_NLERR("Invalid IPv6 flow label value (value=%x, max=%x).\n",
760 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
763 SW_FLOW_KEY_PUT(match, ipv6.label,
764 ipv6_key->ipv6_label, is_mask);
765 SW_FLOW_KEY_PUT(match, ip.proto,
766 ipv6_key->ipv6_proto, is_mask);
767 SW_FLOW_KEY_PUT(match, ip.tos,
768 ipv6_key->ipv6_tclass, is_mask);
769 SW_FLOW_KEY_PUT(match, ip.ttl,
770 ipv6_key->ipv6_hlimit, is_mask);
771 SW_FLOW_KEY_PUT(match, ip.frag,
772 ipv6_key->ipv6_frag, is_mask);
773 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
775 sizeof(match->key->ipv6.addr.src),
777 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
779 sizeof(match->key->ipv6.addr.dst),
782 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV6);
785 if (attrs & (1ULL << OVS_KEY_ATTR_ARP)) {
786 const struct ovs_key_arp *arp_key;
788 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
789 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
790 OVS_NLERR("Unknown ARP opcode (opcode=%d).\n",
795 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
796 arp_key->arp_sip, is_mask);
797 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
798 arp_key->arp_tip, is_mask);
799 SW_FLOW_KEY_PUT(match, ip.proto,
800 ntohs(arp_key->arp_op), is_mask);
801 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
802 arp_key->arp_sha, ETH_ALEN, is_mask);
803 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
804 arp_key->arp_tha, ETH_ALEN, is_mask);
806 attrs &= ~(1ULL << OVS_KEY_ATTR_ARP);
809 if (attrs & (1ULL << OVS_KEY_ATTR_MPLS)) {
810 const struct ovs_key_mpls *mpls_key;
812 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
813 SW_FLOW_KEY_PUT(match, mpls.top_lse,
814 mpls_key->mpls_lse, is_mask);
816 attrs &= ~(1ULL << OVS_KEY_ATTR_MPLS);
819 if (attrs & (1ULL << OVS_KEY_ATTR_TCP)) {
820 const struct ovs_key_tcp *tcp_key;
822 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
823 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
824 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
825 attrs &= ~(1ULL << OVS_KEY_ATTR_TCP);
828 if (attrs & (1ULL << OVS_KEY_ATTR_TCP_FLAGS)) {
829 SW_FLOW_KEY_PUT(match, tp.flags,
830 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
832 attrs &= ~(1ULL << OVS_KEY_ATTR_TCP_FLAGS);
835 if (attrs & (1ULL << OVS_KEY_ATTR_UDP)) {
836 const struct ovs_key_udp *udp_key;
838 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
839 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
840 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
841 attrs &= ~(1ULL << OVS_KEY_ATTR_UDP);
844 if (attrs & (1ULL << OVS_KEY_ATTR_SCTP)) {
845 const struct ovs_key_sctp *sctp_key;
847 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
848 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
849 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
850 attrs &= ~(1ULL << OVS_KEY_ATTR_SCTP);
853 if (attrs & (1ULL << OVS_KEY_ATTR_ICMP)) {
854 const struct ovs_key_icmp *icmp_key;
856 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
857 SW_FLOW_KEY_PUT(match, tp.src,
858 htons(icmp_key->icmp_type), is_mask);
859 SW_FLOW_KEY_PUT(match, tp.dst,
860 htons(icmp_key->icmp_code), is_mask);
861 attrs &= ~(1ULL << OVS_KEY_ATTR_ICMP);
864 if (attrs & (1ULL << OVS_KEY_ATTR_ICMPV6)) {
865 const struct ovs_key_icmpv6 *icmpv6_key;
867 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
868 SW_FLOW_KEY_PUT(match, tp.src,
869 htons(icmpv6_key->icmpv6_type), is_mask);
870 SW_FLOW_KEY_PUT(match, tp.dst,
871 htons(icmpv6_key->icmpv6_code), is_mask);
872 attrs &= ~(1ULL << OVS_KEY_ATTR_ICMPV6);
875 if (attrs & (1ULL << OVS_KEY_ATTR_ND)) {
876 const struct ovs_key_nd *nd_key;
878 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
879 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
881 sizeof(match->key->ipv6.nd.target),
883 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
884 nd_key->nd_sll, ETH_ALEN, is_mask);
885 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
886 nd_key->nd_tll, ETH_ALEN, is_mask);
887 attrs &= ~(1ULL << OVS_KEY_ATTR_ND);
891 OVS_NLERR("Unknown key attributes (%llx).\n",
892 (unsigned long long)attrs);
899 static void nlattr_set(struct nlattr *attr, u8 val, bool is_attr_mask_key)
904 /* The nlattr stream should already have been validated */
905 nla_for_each_nested(nla, attr, rem) {
906 /* We assume that ovs_key_lens[type] == -1 means that type is a
909 if (is_attr_mask_key && ovs_key_lens[nla_type(nla)] == -1)
910 nlattr_set(nla, val, false);
912 memset(nla_data(nla), val, nla_len(nla));
916 static void mask_set_nlattr(struct nlattr *attr, u8 val)
918 nlattr_set(attr, val, true);
922 * ovs_nla_get_match - parses Netlink attributes into a flow key and
923 * mask. In case the 'mask' is NULL, the flow is treated as exact match
924 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
925 * does not include any don't care bit.
926 * @match: receives the extracted flow match information.
927 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
928 * sequence. The fields should of the packet that triggered the creation
930 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
931 * attribute specifies the mask field of the wildcarded flow.
933 int ovs_nla_get_match(struct sw_flow_match *match,
934 const struct nlattr *nla_key,
935 const struct nlattr *nla_mask)
937 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
938 const struct nlattr *encap;
939 struct nlattr *newmask = NULL;
942 bool encap_valid = false;
945 err = parse_flow_nlattrs(nla_key, a, &key_attrs);
949 if ((key_attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
950 (key_attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) &&
951 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
954 if (!((key_attrs & (1ULL << OVS_KEY_ATTR_VLAN)) &&
955 (key_attrs & (1ULL << OVS_KEY_ATTR_ENCAP)))) {
956 OVS_NLERR("Invalid Vlan frame.\n");
960 key_attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
961 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
962 encap = a[OVS_KEY_ATTR_ENCAP];
963 key_attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
966 if (tci & htons(VLAN_TAG_PRESENT)) {
967 err = parse_flow_nlattrs(encap, a, &key_attrs);
971 /* Corner case for truncated 802.1Q header. */
972 if (nla_len(encap)) {
973 OVS_NLERR("Truncated 802.1Q header has non-zero encap attribute.\n");
977 OVS_NLERR("Encap attribute is set for a non-VLAN frame.\n");
982 err = ovs_key_from_nlattrs(match, key_attrs, a, false);
989 /* Create an exact match mask. We need to set to 0xff
990 * all the 'match->mask' fields that have been touched
991 * in 'match->key'. We cannot simply memset
992 * 'match->mask', because padding bytes and fields not
993 * specified in 'match->key' should be left to 0.
994 * Instead, we use a stream of netlink attributes,
995 * copied from 'key' and set to 0xff: ovs_key_from_nlattrs()
996 * will take care of filling 'match->mask'
999 newmask = kmemdup(nla_key,
1000 nla_total_size(nla_len(nla_key)),
1005 mask_set_nlattr(newmask, 0xff);
1007 /* The userspace does not send tunnel attributes that
1008 * are 0, but we should not wildcard them nonetheless.
1010 if (match->key->tun_key.ipv4_dst)
1011 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1017 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs);
1021 /* Always match on tci. */
1022 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
1024 if (mask_attrs & 1ULL << OVS_KEY_ATTR_ENCAP) {
1025 __be16 eth_type = 0;
1029 OVS_NLERR("Encap mask attribute is set for non-VLAN frame.\n");
1034 mask_attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
1035 if (a[OVS_KEY_ATTR_ETHERTYPE])
1036 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1038 if (eth_type == htons(0xffff)) {
1039 mask_attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
1040 encap = a[OVS_KEY_ATTR_ENCAP];
1041 err = parse_flow_mask_nlattrs(encap, a, &mask_attrs);
1045 OVS_NLERR("VLAN frames must have an exact match on the TPID (mask=%x).\n",
1051 if (a[OVS_KEY_ATTR_VLAN])
1052 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1054 if (!(tci & htons(VLAN_TAG_PRESENT))) {
1055 OVS_NLERR("VLAN tag present bit must have an exact match (tci_mask=%x).\n", ntohs(tci));
1061 err = ovs_key_from_nlattrs(match, mask_attrs, a, true);
1066 if (!match_validate(match, key_attrs, mask_attrs))
1075 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1076 * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1077 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1080 * This parses a series of Netlink attributes that form a flow key, which must
1081 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1082 * get the metadata, that is, the parts of the flow key that cannot be
1083 * extracted from the packet itself.
1085 int ovs_nla_get_flow_metadata(const struct nlattr *attr,
1086 struct sw_flow_key *key)
1088 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1089 struct sw_flow_match match;
1093 err = parse_flow_nlattrs(attr, a, &attrs);
1097 memset(&match, 0, sizeof(match));
1100 memset(key, 0, OVS_SW_FLOW_KEY_METADATA_SIZE);
1101 key->phy.in_port = DP_MAX_PORTS;
1103 return metadata_from_nlattrs(&match, &attrs, a, false);
1106 int ovs_nla_put_flow(struct datapath *dp, const struct sw_flow_key *swkey,
1107 const struct sw_flow_key *output, struct sk_buff *skb)
1109 struct ovs_key_ethernet *eth_key;
1110 struct nlattr *nla, *encap;
1111 bool is_mask = (swkey != output);
1113 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1114 goto nla_put_failure;
1116 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1117 goto nla_put_failure;
1119 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1120 goto nla_put_failure;
1122 if ((swkey->tun_key.ipv4_dst || is_mask)) {
1123 const struct geneve_opt *opts = NULL;
1125 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1126 opts = GENEVE_OPTS(output, swkey->tun_opts_len);
1128 if (ipv4_tun_to_nlattr(skb, &output->tun_key, opts,
1129 swkey->tun_opts_len))
1130 goto nla_put_failure;
1133 if (swkey->phy.in_port == DP_MAX_PORTS) {
1134 if (is_mask && (output->phy.in_port == 0xffff))
1135 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1136 goto nla_put_failure;
1139 upper_u16 = !is_mask ? 0 : 0xffff;
1141 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1142 (upper_u16 << 16) | output->phy.in_port))
1143 goto nla_put_failure;
1146 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1147 goto nla_put_failure;
1149 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1151 goto nla_put_failure;
1153 eth_key = nla_data(nla);
1154 ether_addr_copy(eth_key->eth_src, output->eth.src);
1155 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1157 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1159 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
1160 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1161 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
1162 goto nla_put_failure;
1163 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1164 if (!swkey->eth.tci)
1169 if (swkey->eth.type == htons(ETH_P_802_2)) {
1171 * Ethertype 802.2 is represented in the netlink with omitted
1172 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1173 * 0xffff in the mask attribute. Ethertype can also
1176 if (is_mask && output->eth.type)
1177 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1179 goto nla_put_failure;
1183 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1184 goto nla_put_failure;
1186 if (swkey->eth.type == htons(ETH_P_IP)) {
1187 struct ovs_key_ipv4 *ipv4_key;
1189 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1191 goto nla_put_failure;
1192 ipv4_key = nla_data(nla);
1193 ipv4_key->ipv4_src = output->ipv4.addr.src;
1194 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1195 ipv4_key->ipv4_proto = output->ip.proto;
1196 ipv4_key->ipv4_tos = output->ip.tos;
1197 ipv4_key->ipv4_ttl = output->ip.ttl;
1198 ipv4_key->ipv4_frag = output->ip.frag;
1199 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1200 struct ovs_key_ipv6 *ipv6_key;
1202 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1204 goto nla_put_failure;
1205 ipv6_key = nla_data(nla);
1206 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1207 sizeof(ipv6_key->ipv6_src));
1208 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1209 sizeof(ipv6_key->ipv6_dst));
1210 ipv6_key->ipv6_label = output->ipv6.label;
1211 ipv6_key->ipv6_proto = output->ip.proto;
1212 ipv6_key->ipv6_tclass = output->ip.tos;
1213 ipv6_key->ipv6_hlimit = output->ip.ttl;
1214 ipv6_key->ipv6_frag = output->ip.frag;
1215 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1216 swkey->eth.type == htons(ETH_P_RARP)) {
1217 struct ovs_key_arp *arp_key;
1219 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1221 goto nla_put_failure;
1222 arp_key = nla_data(nla);
1223 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1224 arp_key->arp_sip = output->ipv4.addr.src;
1225 arp_key->arp_tip = output->ipv4.addr.dst;
1226 arp_key->arp_op = htons(output->ip.proto);
1227 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1228 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1229 } else if (eth_p_mpls(swkey->eth.type)) {
1230 struct ovs_key_mpls *mpls_key;
1232 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1234 goto nla_put_failure;
1235 mpls_key = nla_data(nla);
1236 mpls_key->mpls_lse = output->mpls.top_lse;
1239 if ((swkey->eth.type == htons(ETH_P_IP) ||
1240 swkey->eth.type == htons(ETH_P_IPV6)) &&
1241 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1243 if (swkey->ip.proto == IPPROTO_TCP) {
1244 struct ovs_key_tcp *tcp_key;
1246 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1248 goto nla_put_failure;
1249 tcp_key = nla_data(nla);
1250 tcp_key->tcp_src = output->tp.src;
1251 tcp_key->tcp_dst = output->tp.dst;
1252 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1254 goto nla_put_failure;
1255 } else if (swkey->ip.proto == IPPROTO_UDP) {
1256 struct ovs_key_udp *udp_key;
1258 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1260 goto nla_put_failure;
1261 udp_key = nla_data(nla);
1262 udp_key->udp_src = output->tp.src;
1263 udp_key->udp_dst = output->tp.dst;
1264 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1265 struct ovs_key_sctp *sctp_key;
1267 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1269 goto nla_put_failure;
1270 sctp_key = nla_data(nla);
1271 sctp_key->sctp_src = output->tp.src;
1272 sctp_key->sctp_dst = output->tp.dst;
1273 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1274 swkey->ip.proto == IPPROTO_ICMP) {
1275 struct ovs_key_icmp *icmp_key;
1277 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1279 goto nla_put_failure;
1280 icmp_key = nla_data(nla);
1281 icmp_key->icmp_type = ntohs(output->tp.src);
1282 icmp_key->icmp_code = ntohs(output->tp.dst);
1283 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1284 swkey->ip.proto == IPPROTO_ICMPV6) {
1285 struct ovs_key_icmpv6 *icmpv6_key;
1287 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1288 sizeof(*icmpv6_key));
1290 goto nla_put_failure;
1291 icmpv6_key = nla_data(nla);
1292 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1293 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1295 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1296 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1297 struct ovs_key_nd *nd_key;
1299 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1301 goto nla_put_failure;
1302 nd_key = nla_data(nla);
1303 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1304 sizeof(nd_key->nd_target));
1305 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1306 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1313 nla_nest_end(skb, encap);
1321 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1323 static struct sw_flow_actions *nla_alloc_flow_actions(int size)
1325 struct sw_flow_actions *sfa;
1327 if (size > MAX_ACTIONS_BUFSIZE) {
1328 OVS_NLERR("Flow action size (%u bytes) exceeds maximum "
1329 "(%u bytes)\n", size, MAX_ACTIONS_BUFSIZE);
1330 return ERR_PTR(-EINVAL);
1333 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1335 return ERR_PTR(-ENOMEM);
1337 sfa->actions_len = 0;
1341 /* RCU callback used by ovs_nla_free_flow_actions. */
1342 static void rcu_free_acts_callback(struct rcu_head *rcu)
1344 struct sw_flow_actions *sf_acts = container_of(rcu,
1345 struct sw_flow_actions, rcu);
1349 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1350 * The caller must hold rcu_read_lock for this to be sensible. */
1351 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1353 call_rcu(&sf_acts->rcu, rcu_free_acts_callback);
1356 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1360 struct sw_flow_actions *acts;
1362 int req_size = NLA_ALIGN(attr_len);
1363 int next_offset = offsetof(struct sw_flow_actions, actions) +
1364 (*sfa)->actions_len;
1366 if (req_size <= (ksize(*sfa) - next_offset))
1369 new_acts_size = ksize(*sfa) * 2;
1371 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1372 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1373 return ERR_PTR(-EMSGSIZE);
1374 new_acts_size = MAX_ACTIONS_BUFSIZE;
1377 acts = nla_alloc_flow_actions(new_acts_size);
1379 return (void *)acts;
1381 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1382 acts->actions_len = (*sfa)->actions_len;
1387 (*sfa)->actions_len += req_size;
1388 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1391 static struct nlattr *__add_action(struct sw_flow_actions **sfa, int attrtype,
1392 void *data, int len)
1396 a = reserve_sfa_size(sfa, nla_attr_size(len));
1400 a->nla_type = attrtype;
1401 a->nla_len = nla_attr_size(len);
1404 memcpy(nla_data(a), data, len);
1405 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1410 static int add_action(struct sw_flow_actions **sfa, int attrtype,
1411 void *data, int len)
1415 a = __add_action(sfa, attrtype, data, len);
1422 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1425 int used = (*sfa)->actions_len;
1428 err = add_action(sfa, attrtype, NULL, 0);
1435 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1438 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1441 a->nla_len = sfa->actions_len - st_offset;
1444 static int __ovs_nla_copy_actions(const struct nlattr *attr,
1445 const struct sw_flow_key *key,
1446 int depth, struct sw_flow_actions **sfa,
1447 __be16 eth_type, __be16 vlan_tci);
1449 static int validate_and_copy_sample(const struct nlattr *attr,
1450 const struct sw_flow_key *key, int depth,
1451 struct sw_flow_actions **sfa,
1452 __be16 eth_type, __be16 vlan_tci)
1454 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1455 const struct nlattr *probability, *actions;
1456 const struct nlattr *a;
1457 int rem, start, err, st_acts;
1459 memset(attrs, 0, sizeof(attrs));
1460 nla_for_each_nested(a, attr, rem) {
1461 int type = nla_type(a);
1462 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1469 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1470 if (!probability || nla_len(probability) != sizeof(u32))
1473 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1474 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1477 /* validation done, copy sample action. */
1478 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE);
1481 err = add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1482 nla_data(probability), sizeof(u32));
1485 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS);
1489 err = __ovs_nla_copy_actions(actions, key, depth + 1, sfa,
1490 eth_type, vlan_tci);
1494 add_nested_action_end(*sfa, st_acts);
1495 add_nested_action_end(*sfa, start);
1500 static int validate_tp_port(const struct sw_flow_key *flow_key,
1503 if ((eth_type == htons(ETH_P_IP) || eth_type == htons(ETH_P_IPV6)) &&
1504 (flow_key->tp.src || flow_key->tp.dst))
1510 void ovs_match_init(struct sw_flow_match *match,
1511 struct sw_flow_key *key,
1512 struct sw_flow_mask *mask)
1514 memset(match, 0, sizeof(*match));
1518 memset(key, 0, sizeof(*key));
1521 memset(&mask->key, 0, sizeof(mask->key));
1522 mask->range.start = mask->range.end = 0;
1526 static int validate_and_copy_set_tun(const struct nlattr *attr,
1527 struct sw_flow_actions **sfa)
1529 struct sw_flow_match match;
1530 struct sw_flow_key key;
1531 struct ovs_tunnel_info *tun_info;
1535 ovs_match_init(&match, &key, NULL);
1536 err = ipv4_tun_from_nlattr(nla_data(attr), &match, false);
1540 if (key.tun_opts_len) {
1541 struct geneve_opt *option = GENEVE_OPTS(&key,
1543 int opts_len = key.tun_opts_len;
1544 bool crit_opt = false;
1546 while (opts_len > 0) {
1549 if (opts_len < sizeof(*option))
1552 len = sizeof(*option) + option->length * 4;
1556 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
1558 option = (struct geneve_opt *)((u8 *)option + len);
1562 key.tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
1565 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET);
1569 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
1570 sizeof(*tun_info) + key.tun_opts_len);
1574 tun_info = nla_data(a);
1575 tun_info->tunnel = key.tun_key;
1576 tun_info->options_len = key.tun_opts_len;
1578 if (tun_info->options_len) {
1579 /* We need to store the options in the action itself since
1580 * everything else will go away after flow setup. We can append
1581 * it to tun_info and then point there.
1583 tun_info->options = (struct geneve_opt *)(tun_info + 1);
1584 memcpy(tun_info->options, GENEVE_OPTS(&key, key.tun_opts_len),
1587 tun_info->options = NULL;
1590 add_nested_action_end(*sfa, start);
1595 static int validate_set(const struct nlattr *a,
1596 const struct sw_flow_key *flow_key,
1597 struct sw_flow_actions **sfa,
1598 bool *set_tun, __be16 eth_type)
1600 const struct nlattr *ovs_key = nla_data(a);
1601 int key_type = nla_type(ovs_key);
1603 /* There can be only one key in a action */
1604 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1607 if (key_type > OVS_KEY_ATTR_MAX ||
1608 (ovs_key_lens[key_type] != nla_len(ovs_key) &&
1609 ovs_key_lens[key_type] != -1))
1613 const struct ovs_key_ipv4 *ipv4_key;
1614 const struct ovs_key_ipv6 *ipv6_key;
1617 case OVS_KEY_ATTR_PRIORITY:
1618 case OVS_KEY_ATTR_SKB_MARK:
1619 case OVS_KEY_ATTR_ETHERNET:
1622 case OVS_KEY_ATTR_TUNNEL:
1624 err = validate_and_copy_set_tun(a, sfa);
1629 case OVS_KEY_ATTR_IPV4:
1630 if (eth_type != htons(ETH_P_IP))
1633 if (!flow_key->ip.proto)
1636 ipv4_key = nla_data(ovs_key);
1637 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
1640 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
1645 case OVS_KEY_ATTR_IPV6:
1646 if (eth_type != htons(ETH_P_IPV6))
1649 if (!flow_key->ip.proto)
1652 ipv6_key = nla_data(ovs_key);
1653 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
1656 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
1659 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
1664 case OVS_KEY_ATTR_TCP:
1665 if (flow_key->ip.proto != IPPROTO_TCP)
1668 return validate_tp_port(flow_key, eth_type);
1670 case OVS_KEY_ATTR_UDP:
1671 if (flow_key->ip.proto != IPPROTO_UDP)
1674 return validate_tp_port(flow_key, eth_type);
1676 case OVS_KEY_ATTR_MPLS:
1677 if (!eth_p_mpls(eth_type))
1681 case OVS_KEY_ATTR_SCTP:
1682 if (flow_key->ip.proto != IPPROTO_SCTP)
1685 return validate_tp_port(flow_key, eth_type);
1694 static int validate_userspace(const struct nlattr *attr)
1696 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
1697 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
1698 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
1699 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
1701 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
1704 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
1705 attr, userspace_policy);
1709 if (!a[OVS_USERSPACE_ATTR_PID] ||
1710 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
1716 static int copy_action(const struct nlattr *from,
1717 struct sw_flow_actions **sfa)
1719 int totlen = NLA_ALIGN(from->nla_len);
1722 to = reserve_sfa_size(sfa, from->nla_len);
1726 memcpy(to, from, totlen);
1730 static int __ovs_nla_copy_actions(const struct nlattr *attr,
1731 const struct sw_flow_key *key,
1732 int depth, struct sw_flow_actions **sfa,
1733 __be16 eth_type, __be16 vlan_tci)
1735 const struct nlattr *a;
1738 if (depth >= SAMPLE_ACTION_DEPTH)
1741 nla_for_each_nested(a, attr, rem) {
1742 /* Expected argument lengths, (u32)-1 for variable length. */
1743 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
1744 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
1745 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
1746 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
1747 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
1748 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
1749 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
1750 [OVS_ACTION_ATTR_POP_VLAN] = 0,
1751 [OVS_ACTION_ATTR_SET] = (u32)-1,
1752 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
1753 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash)
1755 const struct ovs_action_push_vlan *vlan;
1756 int type = nla_type(a);
1759 if (type > OVS_ACTION_ATTR_MAX ||
1760 (action_lens[type] != nla_len(a) &&
1761 action_lens[type] != (u32)-1))
1766 case OVS_ACTION_ATTR_UNSPEC:
1769 case OVS_ACTION_ATTR_USERSPACE:
1770 err = validate_userspace(a);
1775 case OVS_ACTION_ATTR_OUTPUT:
1776 if (nla_get_u32(a) >= DP_MAX_PORTS)
1780 case OVS_ACTION_ATTR_HASH: {
1781 const struct ovs_action_hash *act_hash = nla_data(a);
1783 switch (act_hash->hash_alg) {
1784 case OVS_HASH_ALG_L4:
1793 case OVS_ACTION_ATTR_POP_VLAN:
1794 vlan_tci = htons(0);
1797 case OVS_ACTION_ATTR_PUSH_VLAN:
1799 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
1801 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
1803 vlan_tci = vlan->vlan_tci;
1806 case OVS_ACTION_ATTR_RECIRC:
1809 case OVS_ACTION_ATTR_PUSH_MPLS: {
1810 const struct ovs_action_push_mpls *mpls = nla_data(a);
1812 if (!eth_p_mpls(mpls->mpls_ethertype))
1814 /* Prohibit push MPLS other than to a white list
1815 * for packets that have a known tag order.
1817 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
1818 (eth_type != htons(ETH_P_IP) &&
1819 eth_type != htons(ETH_P_IPV6) &&
1820 eth_type != htons(ETH_P_ARP) &&
1821 eth_type != htons(ETH_P_RARP) &&
1822 !eth_p_mpls(eth_type)))
1824 eth_type = mpls->mpls_ethertype;
1828 case OVS_ACTION_ATTR_POP_MPLS:
1829 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
1830 !eth_p_mpls(eth_type))
1833 /* Disallow subsequent L2.5+ set and mpls_pop actions
1834 * as there is no check here to ensure that the new
1835 * eth_type is valid and thus set actions could
1836 * write off the end of the packet or otherwise
1839 * Support for these actions is planned using packet
1842 eth_type = htons(0);
1845 case OVS_ACTION_ATTR_SET:
1846 err = validate_set(a, key, sfa, &skip_copy, eth_type);
1851 case OVS_ACTION_ATTR_SAMPLE:
1852 err = validate_and_copy_sample(a, key, depth, sfa,
1853 eth_type, vlan_tci);
1863 err = copy_action(a, sfa);
1875 int ovs_nla_copy_actions(const struct nlattr *attr,
1876 const struct sw_flow_key *key,
1877 struct sw_flow_actions **sfa)
1881 *sfa = nla_alloc_flow_actions(nla_len(attr));
1883 return PTR_ERR(*sfa);
1885 err = __ovs_nla_copy_actions(attr, key, 0, sfa, key->eth.type,
1893 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
1895 const struct nlattr *a;
1896 struct nlattr *start;
1899 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
1903 nla_for_each_nested(a, attr, rem) {
1904 int type = nla_type(a);
1905 struct nlattr *st_sample;
1908 case OVS_SAMPLE_ATTR_PROBABILITY:
1909 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
1910 sizeof(u32), nla_data(a)))
1913 case OVS_SAMPLE_ATTR_ACTIONS:
1914 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
1917 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
1920 nla_nest_end(skb, st_sample);
1925 nla_nest_end(skb, start);
1929 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
1931 const struct nlattr *ovs_key = nla_data(a);
1932 int key_type = nla_type(ovs_key);
1933 struct nlattr *start;
1937 case OVS_KEY_ATTR_TUNNEL_INFO: {
1938 struct ovs_tunnel_info *tun_info = nla_data(ovs_key);
1940 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
1944 err = ipv4_tun_to_nlattr(skb, &tun_info->tunnel,
1945 tun_info->options_len ?
1946 tun_info->options : NULL,
1947 tun_info->options_len);
1950 nla_nest_end(skb, start);
1954 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
1962 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
1964 const struct nlattr *a;
1967 nla_for_each_attr(a, attr, len, rem) {
1968 int type = nla_type(a);
1971 case OVS_ACTION_ATTR_SET:
1972 err = set_action_to_attr(a, skb);
1977 case OVS_ACTION_ATTR_SAMPLE:
1978 err = sample_action_to_attr(a, skb);
1983 if (nla_put(skb, type, nla_len(a), nla_data(a)))