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 = NULL;
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;
69 if (range->start == range->end) {
75 if (range->start > start)
82 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
84 update_range__(match, offsetof(struct sw_flow_key, field), \
85 sizeof((match)->key->field), is_mask); \
88 (match)->mask->key.field = value; \
90 (match)->key->field = value; \
94 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
96 update_range__(match, offset, len, is_mask); \
99 memcpy((u8 *)&(match)->mask->key + offset, value_p, len);\
101 memcpy((u8 *)(match)->key + offset, value_p, len); \
105 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
106 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
107 value_p, len, is_mask)
109 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
111 update_range__(match, offsetof(struct sw_flow_key, field), \
112 sizeof((match)->key->field), is_mask); \
115 memset((u8 *)&(match)->mask->key.field, value,\
116 sizeof((match)->mask->key.field)); \
118 memset((u8 *)&(match)->key->field, value, \
119 sizeof((match)->key->field)); \
123 static bool match_validate(const struct sw_flow_match *match,
124 u64 key_attrs, u64 mask_attrs)
126 u64 key_expected = 1ULL << OVS_KEY_ATTR_ETHERNET;
127 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
129 /* The following mask attributes allowed only if they
130 * pass the validation tests. */
131 mask_allowed &= ~((1ULL << OVS_KEY_ATTR_IPV4)
132 | (1ULL << OVS_KEY_ATTR_IPV6)
133 | (1ULL << OVS_KEY_ATTR_TCP)
134 | (1ULL << OVS_KEY_ATTR_TCP_FLAGS)
135 | (1ULL << OVS_KEY_ATTR_UDP)
136 | (1ULL << OVS_KEY_ATTR_SCTP)
137 | (1ULL << OVS_KEY_ATTR_ICMP)
138 | (1ULL << OVS_KEY_ATTR_ICMPV6)
139 | (1ULL << OVS_KEY_ATTR_ARP)
140 | (1ULL << OVS_KEY_ATTR_ND)
141 | (1ULL << OVS_KEY_ATTR_MPLS));
143 /* Always allowed mask fields. */
144 mask_allowed |= ((1ULL << OVS_KEY_ATTR_TUNNEL)
145 | (1ULL << OVS_KEY_ATTR_IN_PORT)
146 | (1ULL << OVS_KEY_ATTR_ETHERTYPE));
148 /* Check key attributes. */
149 if (match->key->eth.type == htons(ETH_P_ARP)
150 || match->key->eth.type == htons(ETH_P_RARP)) {
151 key_expected |= 1ULL << OVS_KEY_ATTR_ARP;
152 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
153 mask_allowed |= 1ULL << OVS_KEY_ATTR_ARP;
157 if (eth_p_mpls(match->key->eth.type)) {
158 key_expected |= 1ULL << OVS_KEY_ATTR_MPLS;
159 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
160 mask_allowed |= 1ULL << OVS_KEY_ATTR_MPLS;
163 if (match->key->eth.type == htons(ETH_P_IP)) {
164 key_expected |= 1ULL << OVS_KEY_ATTR_IPV4;
165 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
166 mask_allowed |= 1ULL << OVS_KEY_ATTR_IPV4;
168 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
169 if (match->key->ip.proto == IPPROTO_UDP) {
170 key_expected |= 1ULL << OVS_KEY_ATTR_UDP;
171 if (match->mask && (match->mask->key.ip.proto == 0xff))
172 mask_allowed |= 1ULL << OVS_KEY_ATTR_UDP;
175 if (match->key->ip.proto == IPPROTO_SCTP) {
176 key_expected |= 1ULL << OVS_KEY_ATTR_SCTP;
177 if (match->mask && (match->mask->key.ip.proto == 0xff))
178 mask_allowed |= 1ULL << OVS_KEY_ATTR_SCTP;
181 if (match->key->ip.proto == IPPROTO_TCP) {
182 key_expected |= 1ULL << OVS_KEY_ATTR_TCP;
183 key_expected |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
184 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
185 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP;
186 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
190 if (match->key->ip.proto == IPPROTO_ICMP) {
191 key_expected |= 1ULL << OVS_KEY_ATTR_ICMP;
192 if (match->mask && (match->mask->key.ip.proto == 0xff))
193 mask_allowed |= 1ULL << OVS_KEY_ATTR_ICMP;
198 if (match->key->eth.type == htons(ETH_P_IPV6)) {
199 key_expected |= 1ULL << OVS_KEY_ATTR_IPV6;
200 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
201 mask_allowed |= 1ULL << OVS_KEY_ATTR_IPV6;
203 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
204 if (match->key->ip.proto == IPPROTO_UDP) {
205 key_expected |= 1ULL << OVS_KEY_ATTR_UDP;
206 if (match->mask && (match->mask->key.ip.proto == 0xff))
207 mask_allowed |= 1ULL << OVS_KEY_ATTR_UDP;
210 if (match->key->ip.proto == IPPROTO_SCTP) {
211 key_expected |= 1ULL << OVS_KEY_ATTR_SCTP;
212 if (match->mask && (match->mask->key.ip.proto == 0xff))
213 mask_allowed |= 1ULL << OVS_KEY_ATTR_SCTP;
216 if (match->key->ip.proto == IPPROTO_TCP) {
217 key_expected |= 1ULL << OVS_KEY_ATTR_TCP;
218 key_expected |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
219 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
220 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP;
221 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
225 if (match->key->ip.proto == IPPROTO_ICMPV6) {
226 key_expected |= 1ULL << OVS_KEY_ATTR_ICMPV6;
227 if (match->mask && (match->mask->key.ip.proto == 0xff))
228 mask_allowed |= 1ULL << OVS_KEY_ATTR_ICMPV6;
230 if (match->key->tp.src ==
231 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
232 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
233 key_expected |= 1ULL << OVS_KEY_ATTR_ND;
234 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
235 mask_allowed |= 1ULL << OVS_KEY_ATTR_ND;
241 if ((key_attrs & key_expected) != key_expected) {
242 /* Key attributes check failed. */
243 OVS_NLERR("Missing expected key attributes (key_attrs=%llx, expected=%llx).\n",
244 (unsigned long long)key_attrs, (unsigned long long)key_expected);
248 if ((mask_attrs & mask_allowed) != mask_attrs) {
249 /* Mask attributes check failed. */
250 OVS_NLERR("Contain more than allowed mask fields (mask_attrs=%llx, mask_allowed=%llx).\n",
251 (unsigned long long)mask_attrs, (unsigned long long)mask_allowed);
258 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
259 static const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
260 [OVS_KEY_ATTR_ENCAP] = -1,
261 [OVS_KEY_ATTR_PRIORITY] = sizeof(u32),
262 [OVS_KEY_ATTR_IN_PORT] = sizeof(u32),
263 [OVS_KEY_ATTR_SKB_MARK] = sizeof(u32),
264 [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
265 [OVS_KEY_ATTR_VLAN] = sizeof(__be16),
266 [OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16),
267 [OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4),
268 [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
269 [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
270 [OVS_KEY_ATTR_TCP_FLAGS] = sizeof(__be16),
271 [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
272 [OVS_KEY_ATTR_SCTP] = sizeof(struct ovs_key_sctp),
273 [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
274 [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
275 [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
276 [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
277 [OVS_KEY_ATTR_DP_HASH] = sizeof(u32),
278 [OVS_KEY_ATTR_RECIRC_ID] = sizeof(u32),
279 [OVS_KEY_ATTR_TUNNEL] = -1,
280 [OVS_KEY_ATTR_MPLS] = sizeof(struct ovs_key_mpls),
283 static bool is_all_zero(const u8 *fp, size_t size)
290 for (i = 0; i < size; i++)
297 static int __parse_flow_nlattrs(const struct nlattr *attr,
298 const struct nlattr *a[],
299 u64 *attrsp, bool nz)
301 const struct nlattr *nla;
306 nla_for_each_nested(nla, attr, rem) {
307 u16 type = nla_type(nla);
310 if (type > OVS_KEY_ATTR_MAX) {
311 OVS_NLERR("Unknown key attribute (type=%d, max=%d).\n",
312 type, OVS_KEY_ATTR_MAX);
316 if (attrs & (1ULL << type)) {
317 OVS_NLERR("Duplicate key attribute (type %d).\n", type);
321 expected_len = ovs_key_lens[type];
322 if (nla_len(nla) != expected_len && expected_len != -1) {
323 OVS_NLERR("Key attribute has unexpected length (type=%d"
324 ", length=%d, expected=%d).\n", type,
325 nla_len(nla), expected_len);
329 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
330 attrs |= 1ULL << type;
335 OVS_NLERR("Message has %d unknown bytes.\n", rem);
343 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
344 const struct nlattr *a[], u64 *attrsp)
346 return __parse_flow_nlattrs(attr, a, attrsp, true);
349 static int parse_flow_nlattrs(const struct nlattr *attr,
350 const struct nlattr *a[], u64 *attrsp)
352 return __parse_flow_nlattrs(attr, a, attrsp, false);
355 static int ipv4_tun_from_nlattr(const struct nlattr *attr,
356 struct sw_flow_match *match, bool is_mask)
361 __be16 tun_flags = 0;
363 nla_for_each_nested(a, attr, rem) {
364 int type = nla_type(a);
365 static const u32 ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
366 [OVS_TUNNEL_KEY_ATTR_ID] = sizeof(u64),
367 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = sizeof(u32),
368 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = sizeof(u32),
369 [OVS_TUNNEL_KEY_ATTR_TOS] = 1,
370 [OVS_TUNNEL_KEY_ATTR_TTL] = 1,
371 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = 0,
372 [OVS_TUNNEL_KEY_ATTR_CSUM] = 0,
373 [OVS_TUNNEL_KEY_ATTR_OAM] = 0,
374 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = -1,
377 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
378 OVS_NLERR("Unknown IPv4 tunnel attribute (type=%d, max=%d).\n",
379 type, OVS_TUNNEL_KEY_ATTR_MAX);
383 if (ovs_tunnel_key_lens[type] != nla_len(a) &&
384 ovs_tunnel_key_lens[type] != -1) {
385 OVS_NLERR("IPv4 tunnel attribute type has unexpected "
386 " length (type=%d, length=%d, expected=%d).\n",
387 type, nla_len(a), ovs_tunnel_key_lens[type]);
392 case OVS_TUNNEL_KEY_ATTR_ID:
393 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
394 nla_get_be64(a), is_mask);
395 tun_flags |= TUNNEL_KEY;
397 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
398 SW_FLOW_KEY_PUT(match, tun_key.ipv4_src,
399 nla_get_be32(a), is_mask);
401 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
402 SW_FLOW_KEY_PUT(match, tun_key.ipv4_dst,
403 nla_get_be32(a), is_mask);
405 case OVS_TUNNEL_KEY_ATTR_TOS:
406 SW_FLOW_KEY_PUT(match, tun_key.ipv4_tos,
407 nla_get_u8(a), is_mask);
409 case OVS_TUNNEL_KEY_ATTR_TTL:
410 SW_FLOW_KEY_PUT(match, tun_key.ipv4_ttl,
411 nla_get_u8(a), is_mask);
414 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
415 tun_flags |= TUNNEL_DONT_FRAGMENT;
417 case OVS_TUNNEL_KEY_ATTR_CSUM:
418 tun_flags |= TUNNEL_CSUM;
420 case OVS_TUNNEL_KEY_ATTR_OAM:
421 tun_flags |= TUNNEL_OAM;
423 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
424 tun_flags |= TUNNEL_OPTIONS_PRESENT;
425 if (nla_len(a) > sizeof(match->key->tun_opts)) {
426 OVS_NLERR("Geneve option length exceeds "
427 "maximum size (len %d, max %zu).\n",
429 sizeof(match->key->tun_opts));
433 if (nla_len(a) % 4 != 0) {
434 OVS_NLERR("Geneve option length is not "
435 "a multiple of 4 (len %d).\n",
440 /* We need to record the length of the options passed
441 * down, otherwise packets with the same format but
442 * additional options will be silently matched.
445 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
448 /* This is somewhat unusual because it looks at
449 * both the key and mask while parsing the
450 * attributes (and by extension assumes the key
451 * is parsed first). Normally, we would verify
452 * that each is the correct length and that the
453 * attributes line up in the validate function.
454 * However, that is difficult because this is
455 * variable length and we won't have the
458 if (match->key->tun_opts_len != nla_len(a)) {
459 OVS_NLERR("Geneve option key length (%d)"
460 " is different from mask length (%d).",
461 match->key->tun_opts_len, nla_len(a));
465 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff,
469 SW_FLOW_KEY_MEMCPY_OFFSET(match,
470 (unsigned long)GENEVE_OPTS((struct sw_flow_key *)0,
472 nla_data(a), nla_len(a), is_mask);
475 OVS_NLERR("Unknown IPv4 tunnel attribute (%d).\n", type);
480 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
483 OVS_NLERR("IPv4 tunnel attribute has %d unknown bytes.\n", rem);
488 if (!match->key->tun_key.ipv4_dst) {
489 OVS_NLERR("IPv4 tunnel destination address is zero.\n");
494 OVS_NLERR("IPv4 tunnel TTL not specified.\n");
502 static int ipv4_tun_to_nlattr(struct sk_buff *skb,
503 const struct ovs_key_ipv4_tunnel *output,
504 const struct geneve_opt *tun_opts,
505 int swkey_tun_opts_len)
509 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
513 if (output->tun_flags & TUNNEL_KEY &&
514 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
516 if (output->ipv4_src &&
517 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC, output->ipv4_src))
519 if (output->ipv4_dst &&
520 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST, output->ipv4_dst))
522 if (output->ipv4_tos &&
523 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->ipv4_tos))
525 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ipv4_ttl))
527 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
528 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
530 if ((output->tun_flags & TUNNEL_CSUM) &&
531 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
533 if ((output->tun_flags & TUNNEL_OAM) &&
534 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
537 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
538 swkey_tun_opts_len, tun_opts))
541 nla_nest_end(skb, nla);
546 static int metadata_from_nlattrs(struct sw_flow_match *match, u64 *attrs,
547 const struct nlattr **a, bool is_mask)
549 if (*attrs & (1ULL << OVS_KEY_ATTR_DP_HASH)) {
550 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
552 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
553 *attrs &= ~(1ULL << OVS_KEY_ATTR_DP_HASH);
556 if (*attrs & (1ULL << OVS_KEY_ATTR_RECIRC_ID)) {
557 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
559 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
560 *attrs &= ~(1ULL << OVS_KEY_ATTR_RECIRC_ID);
563 if (*attrs & (1ULL << OVS_KEY_ATTR_PRIORITY)) {
564 SW_FLOW_KEY_PUT(match, phy.priority,
565 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
566 *attrs &= ~(1ULL << OVS_KEY_ATTR_PRIORITY);
569 if (*attrs & (1ULL << OVS_KEY_ATTR_IN_PORT)) {
570 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
573 in_port = 0xffffffff; /* Always exact match in_port. */
574 } else if (in_port >= DP_MAX_PORTS) {
575 OVS_NLERR("Input port (%d) exceeds maximum allowable (%d).\n",
576 in_port, DP_MAX_PORTS);
580 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
581 *attrs &= ~(1ULL << OVS_KEY_ATTR_IN_PORT);
582 } else if (!is_mask) {
583 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
586 if (*attrs & (1ULL << OVS_KEY_ATTR_SKB_MARK)) {
587 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
589 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
590 *attrs &= ~(1ULL << OVS_KEY_ATTR_SKB_MARK);
592 if (*attrs & (1ULL << OVS_KEY_ATTR_TUNNEL)) {
593 if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
596 *attrs &= ~(1ULL << OVS_KEY_ATTR_TUNNEL);
601 static int ovs_key_from_nlattrs(struct sw_flow_match *match, u64 attrs,
602 const struct nlattr **a, bool is_mask)
606 err = metadata_from_nlattrs(match, &attrs, a, is_mask);
610 if (attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) {
611 const struct ovs_key_ethernet *eth_key;
613 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
614 SW_FLOW_KEY_MEMCPY(match, eth.src,
615 eth_key->eth_src, ETH_ALEN, is_mask);
616 SW_FLOW_KEY_MEMCPY(match, eth.dst,
617 eth_key->eth_dst, ETH_ALEN, is_mask);
618 attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERNET);
621 if (attrs & (1ULL << OVS_KEY_ATTR_VLAN)) {
624 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
625 if (!(tci & htons(VLAN_TAG_PRESENT))) {
627 OVS_NLERR("VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.\n");
629 OVS_NLERR("VLAN TCI does not have VLAN_TAG_PRESENT bit set.\n");
634 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
635 attrs &= ~(1ULL << OVS_KEY_ATTR_VLAN);
637 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
639 if (attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) {
642 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
644 /* Always exact match EtherType. */
645 eth_type = htons(0xffff);
646 } else if (ntohs(eth_type) < ETH_P_802_3_MIN) {
647 OVS_NLERR("EtherType is less than minimum (type=%x, min=%x).\n",
648 ntohs(eth_type), ETH_P_802_3_MIN);
652 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
653 attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
654 } else if (!is_mask) {
655 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
658 if (attrs & (1ULL << OVS_KEY_ATTR_IPV4)) {
659 const struct ovs_key_ipv4 *ipv4_key;
661 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
662 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
663 OVS_NLERR("Unknown IPv4 fragment type (value=%d, max=%d).\n",
664 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
667 SW_FLOW_KEY_PUT(match, ip.proto,
668 ipv4_key->ipv4_proto, is_mask);
669 SW_FLOW_KEY_PUT(match, ip.tos,
670 ipv4_key->ipv4_tos, is_mask);
671 SW_FLOW_KEY_PUT(match, ip.ttl,
672 ipv4_key->ipv4_ttl, is_mask);
673 SW_FLOW_KEY_PUT(match, ip.frag,
674 ipv4_key->ipv4_frag, is_mask);
675 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
676 ipv4_key->ipv4_src, is_mask);
677 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
678 ipv4_key->ipv4_dst, is_mask);
679 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV4);
682 if (attrs & (1ULL << OVS_KEY_ATTR_IPV6)) {
683 const struct ovs_key_ipv6 *ipv6_key;
685 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
686 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
687 OVS_NLERR("Unknown IPv6 fragment type (value=%d, max=%d).\n",
688 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
691 SW_FLOW_KEY_PUT(match, ipv6.label,
692 ipv6_key->ipv6_label, is_mask);
693 SW_FLOW_KEY_PUT(match, ip.proto,
694 ipv6_key->ipv6_proto, is_mask);
695 SW_FLOW_KEY_PUT(match, ip.tos,
696 ipv6_key->ipv6_tclass, is_mask);
697 SW_FLOW_KEY_PUT(match, ip.ttl,
698 ipv6_key->ipv6_hlimit, is_mask);
699 SW_FLOW_KEY_PUT(match, ip.frag,
700 ipv6_key->ipv6_frag, is_mask);
701 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
703 sizeof(match->key->ipv6.addr.src),
705 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
707 sizeof(match->key->ipv6.addr.dst),
710 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV6);
713 if (attrs & (1ULL << OVS_KEY_ATTR_ARP)) {
714 const struct ovs_key_arp *arp_key;
716 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
717 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
718 OVS_NLERR("Unknown ARP opcode (opcode=%d).\n",
723 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
724 arp_key->arp_sip, is_mask);
725 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
726 arp_key->arp_tip, is_mask);
727 SW_FLOW_KEY_PUT(match, ip.proto,
728 ntohs(arp_key->arp_op), is_mask);
729 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
730 arp_key->arp_sha, ETH_ALEN, is_mask);
731 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
732 arp_key->arp_tha, ETH_ALEN, is_mask);
734 attrs &= ~(1ULL << OVS_KEY_ATTR_ARP);
737 if (attrs & (1ULL << OVS_KEY_ATTR_MPLS)) {
738 const struct ovs_key_mpls *mpls_key;
740 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
741 SW_FLOW_KEY_PUT(match, mpls.top_lse,
742 mpls_key->mpls_lse, is_mask);
744 attrs &= ~(1ULL << OVS_KEY_ATTR_MPLS);
747 if (attrs & (1ULL << OVS_KEY_ATTR_TCP)) {
748 const struct ovs_key_tcp *tcp_key;
750 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
751 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
752 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
753 attrs &= ~(1ULL << OVS_KEY_ATTR_TCP);
756 if (attrs & (1ULL << OVS_KEY_ATTR_TCP_FLAGS)) {
757 SW_FLOW_KEY_PUT(match, tp.flags,
758 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
760 attrs &= ~(1ULL << OVS_KEY_ATTR_TCP_FLAGS);
763 if (attrs & (1ULL << OVS_KEY_ATTR_UDP)) {
764 const struct ovs_key_udp *udp_key;
766 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
767 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
768 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
769 attrs &= ~(1ULL << OVS_KEY_ATTR_UDP);
772 if (attrs & (1ULL << OVS_KEY_ATTR_SCTP)) {
773 const struct ovs_key_sctp *sctp_key;
775 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
776 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
777 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
778 attrs &= ~(1ULL << OVS_KEY_ATTR_SCTP);
781 if (attrs & (1ULL << OVS_KEY_ATTR_ICMP)) {
782 const struct ovs_key_icmp *icmp_key;
784 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
785 SW_FLOW_KEY_PUT(match, tp.src,
786 htons(icmp_key->icmp_type), is_mask);
787 SW_FLOW_KEY_PUT(match, tp.dst,
788 htons(icmp_key->icmp_code), is_mask);
789 attrs &= ~(1ULL << OVS_KEY_ATTR_ICMP);
792 if (attrs & (1ULL << OVS_KEY_ATTR_ICMPV6)) {
793 const struct ovs_key_icmpv6 *icmpv6_key;
795 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
796 SW_FLOW_KEY_PUT(match, tp.src,
797 htons(icmpv6_key->icmpv6_type), is_mask);
798 SW_FLOW_KEY_PUT(match, tp.dst,
799 htons(icmpv6_key->icmpv6_code), is_mask);
800 attrs &= ~(1ULL << OVS_KEY_ATTR_ICMPV6);
803 if (attrs & (1ULL << OVS_KEY_ATTR_ND)) {
804 const struct ovs_key_nd *nd_key;
806 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
807 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
809 sizeof(match->key->ipv6.nd.target),
811 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
812 nd_key->nd_sll, ETH_ALEN, is_mask);
813 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
814 nd_key->nd_tll, ETH_ALEN, is_mask);
815 attrs &= ~(1ULL << OVS_KEY_ATTR_ND);
819 OVS_NLERR("Unknown key attributes (%llx).\n",
820 (unsigned long long)attrs);
827 static void nlattr_set(struct nlattr *attr, u8 val, bool is_attr_mask_key)
832 /* The nlattr stream should already have been validated */
833 nla_for_each_nested(nla, attr, rem) {
834 /* We assume that ovs_key_lens[type] == -1 means that type is a
837 if (is_attr_mask_key && ovs_key_lens[nla_type(nla)] == -1)
838 nlattr_set(nla, val, false);
840 memset(nla_data(nla), val, nla_len(nla));
844 static void mask_set_nlattr(struct nlattr *attr, u8 val)
846 nlattr_set(attr, val, true);
850 * ovs_nla_get_match - parses Netlink attributes into a flow key and
851 * mask. In case the 'mask' is NULL, the flow is treated as exact match
852 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
853 * does not include any don't care bit.
854 * @match: receives the extracted flow match information.
855 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
856 * sequence. The fields should of the packet that triggered the creation
858 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
859 * attribute specifies the mask field of the wildcarded flow.
861 int ovs_nla_get_match(struct sw_flow_match *match,
862 const struct nlattr *key,
863 const struct nlattr *mask)
865 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
866 const struct nlattr *encap;
867 struct nlattr *newmask = NULL;
870 bool encap_valid = false;
873 err = parse_flow_nlattrs(key, a, &key_attrs);
877 if ((key_attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
878 (key_attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) &&
879 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
882 if (!((key_attrs & (1ULL << OVS_KEY_ATTR_VLAN)) &&
883 (key_attrs & (1ULL << OVS_KEY_ATTR_ENCAP)))) {
884 OVS_NLERR("Invalid Vlan frame.\n");
888 key_attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
889 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
890 encap = a[OVS_KEY_ATTR_ENCAP];
891 key_attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
894 if (tci & htons(VLAN_TAG_PRESENT)) {
895 err = parse_flow_nlattrs(encap, a, &key_attrs);
899 /* Corner case for truncated 802.1Q header. */
900 if (nla_len(encap)) {
901 OVS_NLERR("Truncated 802.1Q header has non-zero encap attribute.\n");
905 OVS_NLERR("Encap attribute is set for a non-VLAN frame.\n");
910 err = ovs_key_from_nlattrs(match, key_attrs, a, false);
914 if (match->mask && !mask) {
915 /* Create an exact match mask. We need to set to 0xff all the
916 * 'match->mask' fields that have been touched in 'match->key'.
917 * We cannot simply memset 'match->mask', because padding bytes
918 * and fields not specified in 'match->key' should be left to 0.
919 * Instead, we use a stream of netlink attributes, copied from
920 * 'key' and set to 0xff: ovs_key_from_nlattrs() will take care
921 * of filling 'match->mask' appropriately.
923 newmask = kmemdup(key, nla_total_size(nla_len(key)),
928 mask_set_nlattr(newmask, 0xff);
930 /* The userspace does not send tunnel attributes that are 0,
931 * but we should not wildcard them nonetheless. */
932 if (match->key->tun_key.ipv4_dst)
933 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key, 0xff, true);
939 err = parse_flow_mask_nlattrs(mask, a, &mask_attrs);
943 if (mask_attrs & 1ULL << OVS_KEY_ATTR_ENCAP) {
948 OVS_NLERR("Encap mask attribute is set for non-VLAN frame.\n");
953 mask_attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
954 if (a[OVS_KEY_ATTR_ETHERTYPE])
955 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
957 if (eth_type == htons(0xffff)) {
958 mask_attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
959 encap = a[OVS_KEY_ATTR_ENCAP];
960 err = parse_flow_mask_nlattrs(encap, a, &mask_attrs);
964 OVS_NLERR("VLAN frames must have an exact match on the TPID (mask=%x).\n",
970 if (a[OVS_KEY_ATTR_VLAN])
971 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
973 if (!(tci & htons(VLAN_TAG_PRESENT))) {
974 OVS_NLERR("VLAN tag present bit must have an exact match (tci_mask=%x).\n", ntohs(tci));
980 err = ovs_key_from_nlattrs(match, mask_attrs, a, true);
985 if (!match_validate(match, key_attrs, mask_attrs))
994 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
995 * @key: Receives extracted in_port, priority, tun_key and skb_mark.
996 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
999 * This parses a series of Netlink attributes that form a flow key, which must
1000 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1001 * get the metadata, that is, the parts of the flow key that cannot be
1002 * extracted from the packet itself.
1004 int ovs_nla_get_flow_metadata(const struct nlattr *attr,
1005 struct sw_flow_key *key)
1007 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1008 struct sw_flow_match match;
1012 err = parse_flow_nlattrs(attr, a, &attrs);
1016 memset(&match, 0, sizeof(match));
1019 key->tun_opts_len = 0;
1020 memset(&key->tun_key, 0, sizeof(key->tun_key));
1021 key->phy.priority = 0;
1022 key->phy.skb_mark = 0;
1023 key->phy.in_port = DP_MAX_PORTS;
1024 key->ovs_flow_hash = 0;
1027 return metadata_from_nlattrs(&match, &attrs, a, false);
1030 int ovs_nla_put_flow(struct datapath *dp, const struct sw_flow_key *swkey,
1031 const struct sw_flow_key *output, struct sk_buff *skb)
1033 struct ovs_key_ethernet *eth_key;
1034 struct nlattr *nla, *encap;
1035 bool is_mask = (swkey != output);
1037 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1038 goto nla_put_failure;
1040 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1041 goto nla_put_failure;
1043 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1044 goto nla_put_failure;
1046 if ((swkey->tun_key.ipv4_dst || is_mask)) {
1047 const struct geneve_opt *opts = NULL;
1049 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1050 opts = GENEVE_OPTS(output, swkey->tun_opts_len);
1052 if (ipv4_tun_to_nlattr(skb, &output->tun_key, opts,
1053 swkey->tun_opts_len))
1054 goto nla_put_failure;
1057 if (swkey->phy.in_port == DP_MAX_PORTS) {
1058 if (is_mask && (output->phy.in_port == 0xffff))
1059 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1060 goto nla_put_failure;
1063 upper_u16 = !is_mask ? 0 : 0xffff;
1065 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1066 (upper_u16 << 16) | output->phy.in_port))
1067 goto nla_put_failure;
1070 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1071 goto nla_put_failure;
1073 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1075 goto nla_put_failure;
1077 eth_key = nla_data(nla);
1078 ether_addr_copy(eth_key->eth_src, output->eth.src);
1079 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1081 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1083 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
1084 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1085 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
1086 goto nla_put_failure;
1087 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1088 if (!swkey->eth.tci)
1093 if (swkey->eth.type == htons(ETH_P_802_2)) {
1095 * Ethertype 802.2 is represented in the netlink with omitted
1096 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1097 * 0xffff in the mask attribute. Ethertype can also
1100 if (is_mask && output->eth.type)
1101 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1103 goto nla_put_failure;
1107 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1108 goto nla_put_failure;
1110 if (swkey->eth.type == htons(ETH_P_IP)) {
1111 struct ovs_key_ipv4 *ipv4_key;
1113 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1115 goto nla_put_failure;
1116 ipv4_key = nla_data(nla);
1117 ipv4_key->ipv4_src = output->ipv4.addr.src;
1118 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1119 ipv4_key->ipv4_proto = output->ip.proto;
1120 ipv4_key->ipv4_tos = output->ip.tos;
1121 ipv4_key->ipv4_ttl = output->ip.ttl;
1122 ipv4_key->ipv4_frag = output->ip.frag;
1123 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1124 struct ovs_key_ipv6 *ipv6_key;
1126 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1128 goto nla_put_failure;
1129 ipv6_key = nla_data(nla);
1130 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1131 sizeof(ipv6_key->ipv6_src));
1132 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1133 sizeof(ipv6_key->ipv6_dst));
1134 ipv6_key->ipv6_label = output->ipv6.label;
1135 ipv6_key->ipv6_proto = output->ip.proto;
1136 ipv6_key->ipv6_tclass = output->ip.tos;
1137 ipv6_key->ipv6_hlimit = output->ip.ttl;
1138 ipv6_key->ipv6_frag = output->ip.frag;
1139 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1140 swkey->eth.type == htons(ETH_P_RARP)) {
1141 struct ovs_key_arp *arp_key;
1143 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1145 goto nla_put_failure;
1146 arp_key = nla_data(nla);
1147 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1148 arp_key->arp_sip = output->ipv4.addr.src;
1149 arp_key->arp_tip = output->ipv4.addr.dst;
1150 arp_key->arp_op = htons(output->ip.proto);
1151 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1152 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1153 } else if (eth_p_mpls(swkey->eth.type)) {
1154 struct ovs_key_mpls *mpls_key;
1156 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1158 goto nla_put_failure;
1159 mpls_key = nla_data(nla);
1160 mpls_key->mpls_lse = output->mpls.top_lse;
1163 if ((swkey->eth.type == htons(ETH_P_IP) ||
1164 swkey->eth.type == htons(ETH_P_IPV6)) &&
1165 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1167 if (swkey->ip.proto == IPPROTO_TCP) {
1168 struct ovs_key_tcp *tcp_key;
1170 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1172 goto nla_put_failure;
1173 tcp_key = nla_data(nla);
1174 tcp_key->tcp_src = output->tp.src;
1175 tcp_key->tcp_dst = output->tp.dst;
1176 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1178 goto nla_put_failure;
1179 } else if (swkey->ip.proto == IPPROTO_UDP) {
1180 struct ovs_key_udp *udp_key;
1182 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1184 goto nla_put_failure;
1185 udp_key = nla_data(nla);
1186 udp_key->udp_src = output->tp.src;
1187 udp_key->udp_dst = output->tp.dst;
1188 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1189 struct ovs_key_sctp *sctp_key;
1191 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1193 goto nla_put_failure;
1194 sctp_key = nla_data(nla);
1195 sctp_key->sctp_src = output->tp.src;
1196 sctp_key->sctp_dst = output->tp.dst;
1197 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1198 swkey->ip.proto == IPPROTO_ICMP) {
1199 struct ovs_key_icmp *icmp_key;
1201 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1203 goto nla_put_failure;
1204 icmp_key = nla_data(nla);
1205 icmp_key->icmp_type = ntohs(output->tp.src);
1206 icmp_key->icmp_code = ntohs(output->tp.dst);
1207 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1208 swkey->ip.proto == IPPROTO_ICMPV6) {
1209 struct ovs_key_icmpv6 *icmpv6_key;
1211 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1212 sizeof(*icmpv6_key));
1214 goto nla_put_failure;
1215 icmpv6_key = nla_data(nla);
1216 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1217 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1219 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1220 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1221 struct ovs_key_nd *nd_key;
1223 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1225 goto nla_put_failure;
1226 nd_key = nla_data(nla);
1227 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1228 sizeof(nd_key->nd_target));
1229 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1230 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1237 nla_nest_end(skb, encap);
1245 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1247 struct sw_flow_actions *ovs_nla_alloc_flow_actions(int size)
1249 struct sw_flow_actions *sfa;
1251 if (size > MAX_ACTIONS_BUFSIZE) {
1252 OVS_NLERR("Flow action size (%u bytes) exceeds maximum "
1253 "(%u bytes)\n", size, MAX_ACTIONS_BUFSIZE);
1254 return ERR_PTR(-EINVAL);
1257 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1259 return ERR_PTR(-ENOMEM);
1261 sfa->actions_len = 0;
1265 /* RCU callback used by ovs_nla_free_flow_actions. */
1266 static void rcu_free_acts_callback(struct rcu_head *rcu)
1268 struct sw_flow_actions *sf_acts = container_of(rcu,
1269 struct sw_flow_actions, rcu);
1273 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1274 * The caller must hold rcu_read_lock for this to be sensible. */
1275 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1277 call_rcu(&sf_acts->rcu, rcu_free_acts_callback);
1280 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1284 struct sw_flow_actions *acts;
1286 int req_size = NLA_ALIGN(attr_len);
1287 int next_offset = offsetof(struct sw_flow_actions, actions) +
1288 (*sfa)->actions_len;
1290 if (req_size <= (ksize(*sfa) - next_offset))
1293 new_acts_size = ksize(*sfa) * 2;
1295 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1296 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1297 return ERR_PTR(-EMSGSIZE);
1298 new_acts_size = MAX_ACTIONS_BUFSIZE;
1301 acts = ovs_nla_alloc_flow_actions(new_acts_size);
1303 return (void *)acts;
1305 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1306 acts->actions_len = (*sfa)->actions_len;
1311 (*sfa)->actions_len += req_size;
1312 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1315 static struct nlattr *__add_action(struct sw_flow_actions **sfa, int attrtype,
1316 void *data, int len)
1320 a = reserve_sfa_size(sfa, nla_attr_size(len));
1324 a->nla_type = attrtype;
1325 a->nla_len = nla_attr_size(len);
1328 memcpy(nla_data(a), data, len);
1329 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1334 static int add_action(struct sw_flow_actions **sfa, int attrtype,
1335 void *data, int len)
1339 a = __add_action(sfa, attrtype, data, len);
1346 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1349 int used = (*sfa)->actions_len;
1352 err = add_action(sfa, attrtype, NULL, 0);
1359 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1362 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1365 a->nla_len = sfa->actions_len - st_offset;
1368 static int ovs_nla_copy_actions__(const struct nlattr *attr,
1369 const struct sw_flow_key *key,
1370 int depth, struct sw_flow_actions **sfa,
1371 __be16 eth_type, __be16 vlan_tci);
1373 static int validate_and_copy_sample(const struct nlattr *attr,
1374 const struct sw_flow_key *key, int depth,
1375 struct sw_flow_actions **sfa,
1376 __be16 eth_type, __be16 vlan_tci)
1378 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1379 const struct nlattr *probability, *actions;
1380 const struct nlattr *a;
1381 int rem, start, err, st_acts;
1383 memset(attrs, 0, sizeof(attrs));
1384 nla_for_each_nested(a, attr, rem) {
1385 int type = nla_type(a);
1386 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1393 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1394 if (!probability || nla_len(probability) != sizeof(u32))
1397 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1398 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1401 /* validation done, copy sample action. */
1402 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE);
1405 err = add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1406 nla_data(probability), sizeof(u32));
1409 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS);
1413 err = ovs_nla_copy_actions__(actions, key, depth + 1, sfa,
1414 eth_type, vlan_tci);
1418 add_nested_action_end(*sfa, st_acts);
1419 add_nested_action_end(*sfa, start);
1424 static int validate_tp_port(const struct sw_flow_key *flow_key,
1427 if ((eth_type == htons(ETH_P_IP) || eth_type == htons(ETH_P_IPV6)) &&
1428 (flow_key->tp.src || flow_key->tp.dst))
1434 void ovs_match_init(struct sw_flow_match *match,
1435 struct sw_flow_key *key,
1436 struct sw_flow_mask *mask)
1438 memset(match, 0, sizeof(*match));
1442 memset(key, 0, sizeof(*key));
1445 memset(&mask->key, 0, sizeof(mask->key));
1446 mask->range.start = mask->range.end = 0;
1450 static int validate_and_copy_set_tun(const struct nlattr *attr,
1451 struct sw_flow_actions **sfa)
1453 struct sw_flow_match match;
1454 struct sw_flow_key key;
1455 struct ovs_tunnel_info *tun_info;
1459 ovs_match_init(&match, &key, NULL);
1460 err = ipv4_tun_from_nlattr(nla_data(attr), &match, false);
1464 if (key.tun_opts_len) {
1465 struct geneve_opt *option = GENEVE_OPTS(&key,
1467 int opts_len = key.tun_opts_len;
1468 bool crit_opt = false;
1470 while (opts_len > 0) {
1473 if (opts_len < sizeof(*option))
1476 len = sizeof(*option) + option->length * 4;
1480 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
1482 option = (struct geneve_opt *)((u8 *)option + len);
1486 key.tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
1489 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET);
1493 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
1494 sizeof(*tun_info) + key.tun_opts_len);
1498 tun_info = nla_data(a);
1499 tun_info->tunnel = key.tun_key;
1500 tun_info->options_len = key.tun_opts_len;
1502 if (tun_info->options_len) {
1503 /* We need to store the options in the action itself since
1504 * everything else will go away after flow setup. We can append
1505 * it to tun_info and then point there.
1507 tun_info->options = (struct geneve_opt *)(tun_info + 1);
1508 memcpy(tun_info->options, GENEVE_OPTS(&key, key.tun_opts_len),
1511 tun_info->options = NULL;
1514 add_nested_action_end(*sfa, start);
1519 static int validate_set(const struct nlattr *a,
1520 const struct sw_flow_key *flow_key,
1521 struct sw_flow_actions **sfa,
1522 bool *set_tun, __be16 eth_type)
1524 const struct nlattr *ovs_key = nla_data(a);
1525 int key_type = nla_type(ovs_key);
1527 /* There can be only one key in a action */
1528 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1531 if (key_type > OVS_KEY_ATTR_MAX ||
1532 (ovs_key_lens[key_type] != nla_len(ovs_key) &&
1533 ovs_key_lens[key_type] != -1))
1537 const struct ovs_key_ipv4 *ipv4_key;
1538 const struct ovs_key_ipv6 *ipv6_key;
1541 case OVS_KEY_ATTR_PRIORITY:
1542 case OVS_KEY_ATTR_SKB_MARK:
1543 case OVS_KEY_ATTR_ETHERNET:
1546 case OVS_KEY_ATTR_TUNNEL:
1548 err = validate_and_copy_set_tun(a, sfa);
1553 case OVS_KEY_ATTR_IPV4:
1554 if (eth_type != htons(ETH_P_IP))
1557 if (!flow_key->ip.proto)
1560 ipv4_key = nla_data(ovs_key);
1561 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
1564 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
1569 case OVS_KEY_ATTR_IPV6:
1570 if (eth_type != htons(ETH_P_IPV6))
1573 if (!flow_key->ip.proto)
1576 ipv6_key = nla_data(ovs_key);
1577 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
1580 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
1583 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
1588 case OVS_KEY_ATTR_TCP:
1589 if (flow_key->ip.proto != IPPROTO_TCP)
1592 return validate_tp_port(flow_key, eth_type);
1594 case OVS_KEY_ATTR_UDP:
1595 if (flow_key->ip.proto != IPPROTO_UDP)
1598 return validate_tp_port(flow_key, eth_type);
1600 case OVS_KEY_ATTR_MPLS:
1601 if (!eth_p_mpls(eth_type))
1605 case OVS_KEY_ATTR_SCTP:
1606 if (flow_key->ip.proto != IPPROTO_SCTP)
1609 return validate_tp_port(flow_key, eth_type);
1618 static int validate_userspace(const struct nlattr *attr)
1620 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
1621 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
1622 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
1624 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
1627 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
1628 attr, userspace_policy);
1632 if (!a[OVS_USERSPACE_ATTR_PID] ||
1633 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
1639 static int copy_action(const struct nlattr *from,
1640 struct sw_flow_actions **sfa)
1642 int totlen = NLA_ALIGN(from->nla_len);
1645 to = reserve_sfa_size(sfa, from->nla_len);
1649 memcpy(to, from, totlen);
1653 static int ovs_nla_copy_actions__(const struct nlattr *attr,
1654 const struct sw_flow_key *key,
1655 int depth, struct sw_flow_actions **sfa,
1656 __be16 eth_type, __be16 vlan_tci)
1658 const struct nlattr *a;
1661 if (depth >= SAMPLE_ACTION_DEPTH)
1664 nla_for_each_nested(a, attr, rem) {
1665 /* Expected argument lengths, (u32)-1 for variable length. */
1666 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
1667 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
1668 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
1669 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
1670 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
1671 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
1672 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
1673 [OVS_ACTION_ATTR_POP_VLAN] = 0,
1674 [OVS_ACTION_ATTR_SET] = (u32)-1,
1675 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
1676 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash)
1678 const struct ovs_action_push_vlan *vlan;
1679 int type = nla_type(a);
1682 if (type > OVS_ACTION_ATTR_MAX ||
1683 (action_lens[type] != nla_len(a) &&
1684 action_lens[type] != (u32)-1))
1689 case OVS_ACTION_ATTR_UNSPEC:
1692 case OVS_ACTION_ATTR_USERSPACE:
1693 err = validate_userspace(a);
1698 case OVS_ACTION_ATTR_OUTPUT:
1699 if (nla_get_u32(a) >= DP_MAX_PORTS)
1703 case OVS_ACTION_ATTR_HASH: {
1704 const struct ovs_action_hash *act_hash = nla_data(a);
1706 switch (act_hash->hash_alg) {
1707 case OVS_HASH_ALG_L4:
1716 case OVS_ACTION_ATTR_POP_VLAN:
1717 vlan_tci = htons(0);
1720 case OVS_ACTION_ATTR_PUSH_VLAN:
1722 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
1724 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
1726 vlan_tci = vlan->vlan_tci;
1729 case OVS_ACTION_ATTR_RECIRC:
1732 case OVS_ACTION_ATTR_PUSH_MPLS: {
1733 const struct ovs_action_push_mpls *mpls = nla_data(a);
1735 if (!eth_p_mpls(mpls->mpls_ethertype))
1737 /* Prohibit push MPLS other than to a white list
1738 * for packets that have a known tag order.
1740 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
1741 (eth_type != htons(ETH_P_IP) &&
1742 eth_type != htons(ETH_P_IPV6) &&
1743 eth_type != htons(ETH_P_ARP) &&
1744 eth_type != htons(ETH_P_RARP) &&
1745 !eth_p_mpls(eth_type)))
1747 eth_type = mpls->mpls_ethertype;
1751 case OVS_ACTION_ATTR_POP_MPLS:
1752 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
1753 !eth_p_mpls(eth_type))
1756 /* Disallow subsequent L2.5+ set and mpls_pop actions
1757 * as there is no check here to ensure that the new
1758 * eth_type is valid and thus set actions could
1759 * write off the end of the packet or otherwise
1762 * Support for these actions is planned using packet
1765 eth_type = htons(0);
1768 case OVS_ACTION_ATTR_SET:
1769 err = validate_set(a, key, sfa, &skip_copy, eth_type);
1774 case OVS_ACTION_ATTR_SAMPLE:
1775 err = validate_and_copy_sample(a, key, depth, sfa,
1776 eth_type, vlan_tci);
1786 err = copy_action(a, sfa);
1798 int ovs_nla_copy_actions(const struct nlattr *attr,
1799 const struct sw_flow_key *key,
1800 struct sw_flow_actions **sfa)
1802 return ovs_nla_copy_actions__(attr, key, 0, sfa, key->eth.type,
1806 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
1808 const struct nlattr *a;
1809 struct nlattr *start;
1812 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
1816 nla_for_each_nested(a, attr, rem) {
1817 int type = nla_type(a);
1818 struct nlattr *st_sample;
1821 case OVS_SAMPLE_ATTR_PROBABILITY:
1822 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
1823 sizeof(u32), nla_data(a)))
1826 case OVS_SAMPLE_ATTR_ACTIONS:
1827 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
1830 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
1833 nla_nest_end(skb, st_sample);
1838 nla_nest_end(skb, start);
1842 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
1844 const struct nlattr *ovs_key = nla_data(a);
1845 int key_type = nla_type(ovs_key);
1846 struct nlattr *start;
1850 case OVS_KEY_ATTR_TUNNEL_INFO: {
1851 struct ovs_tunnel_info *tun_info = nla_data(ovs_key);
1853 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
1857 err = ipv4_tun_to_nlattr(skb, &tun_info->tunnel,
1858 tun_info->options_len ?
1859 tun_info->options : NULL,
1860 tun_info->options_len);
1863 nla_nest_end(skb, start);
1867 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
1875 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
1877 const struct nlattr *a;
1880 nla_for_each_attr(a, attr, len, rem) {
1881 int type = nla_type(a);
1884 case OVS_ACTION_ATTR_SET:
1885 err = set_action_to_attr(a, skb);
1890 case OVS_ACTION_ATTR_SAMPLE:
1891 err = sample_action_to_attr(a, skb);
1896 if (nla_put(skb, type, nla_len(a), nla_data(a)))