2 * Copyright (c) 2007-2013 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
23 #include <linux/uaccess.h>
24 #include <linux/netdevice.h>
25 #include <linux/etherdevice.h>
26 #include <linux/if_ether.h>
27 #include <linux/if_vlan.h>
28 #include <net/llc_pdu.h>
29 #include <linux/kernel.h>
30 #include <linux/jhash.h>
31 #include <linux/jiffies.h>
32 #include <linux/llc.h>
33 #include <linux/module.h>
35 #include <linux/rcupdate.h>
36 #include <linux/if_arp.h>
38 #include <linux/ipv6.h>
39 #include <linux/sctp.h>
40 #include <linux/tcp.h>
41 #include <linux/udp.h>
42 #include <linux/icmp.h>
43 #include <linux/icmpv6.h>
44 #include <linux/rculist.h>
46 #include <net/ip_tunnels.h>
48 #include <net/ndisc.h>
50 #include "flow_netlink.h"
52 static void update_range__(struct sw_flow_match *match,
53 size_t offset, size_t size, bool is_mask)
55 struct sw_flow_key_range *range = NULL;
56 size_t start = rounddown(offset, sizeof(long));
57 size_t end = roundup(offset + size, sizeof(long));
60 range = &match->range;
62 range = &match->mask->range;
67 if (range->start == range->end) {
73 if (range->start > start)
80 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
82 update_range__(match, offsetof(struct sw_flow_key, field), \
83 sizeof((match)->key->field), is_mask); \
86 (match)->mask->key.field = value; \
88 (match)->key->field = value; \
92 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
94 update_range__(match, offsetof(struct sw_flow_key, field), \
98 memcpy(&(match)->mask->key.field, value_p, len);\
100 memcpy(&(match)->key->field, value_p, len); \
104 static u16 range_n_bytes(const struct sw_flow_key_range *range)
106 return range->end - range->start;
109 static bool match_validate(const struct sw_flow_match *match,
110 u64 key_attrs, u64 mask_attrs)
112 u64 key_expected = 1ULL << OVS_KEY_ATTR_ETHERNET;
113 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
115 /* The following mask attributes allowed only if they
116 * pass the validation tests. */
117 mask_allowed &= ~((1ULL << OVS_KEY_ATTR_IPV4)
118 | (1ULL << OVS_KEY_ATTR_IPV6)
119 | (1ULL << OVS_KEY_ATTR_TCP)
120 | (1ULL << OVS_KEY_ATTR_TCP_FLAGS)
121 | (1ULL << OVS_KEY_ATTR_UDP)
122 | (1ULL << OVS_KEY_ATTR_SCTP)
123 | (1ULL << OVS_KEY_ATTR_ICMP)
124 | (1ULL << OVS_KEY_ATTR_ICMPV6)
125 | (1ULL << OVS_KEY_ATTR_ARP)
126 | (1ULL << OVS_KEY_ATTR_ND));
128 /* Always allowed mask fields. */
129 mask_allowed |= ((1ULL << OVS_KEY_ATTR_TUNNEL)
130 | (1ULL << OVS_KEY_ATTR_IN_PORT)
131 | (1ULL << OVS_KEY_ATTR_ETHERTYPE));
133 /* Check key attributes. */
134 if (match->key->eth.type == htons(ETH_P_ARP)
135 || match->key->eth.type == htons(ETH_P_RARP)) {
136 key_expected |= 1ULL << OVS_KEY_ATTR_ARP;
137 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
138 mask_allowed |= 1ULL << OVS_KEY_ATTR_ARP;
141 if (match->key->eth.type == htons(ETH_P_IP)) {
142 key_expected |= 1ULL << OVS_KEY_ATTR_IPV4;
143 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
144 mask_allowed |= 1ULL << OVS_KEY_ATTR_IPV4;
146 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
147 if (match->key->ip.proto == IPPROTO_UDP) {
148 key_expected |= 1ULL << OVS_KEY_ATTR_UDP;
149 if (match->mask && (match->mask->key.ip.proto == 0xff))
150 mask_allowed |= 1ULL << OVS_KEY_ATTR_UDP;
153 if (match->key->ip.proto == IPPROTO_SCTP) {
154 key_expected |= 1ULL << OVS_KEY_ATTR_SCTP;
155 if (match->mask && (match->mask->key.ip.proto == 0xff))
156 mask_allowed |= 1ULL << OVS_KEY_ATTR_SCTP;
159 if (match->key->ip.proto == IPPROTO_TCP) {
160 key_expected |= 1ULL << OVS_KEY_ATTR_TCP;
161 key_expected |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
162 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
163 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP;
164 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
168 if (match->key->ip.proto == IPPROTO_ICMP) {
169 key_expected |= 1ULL << OVS_KEY_ATTR_ICMP;
170 if (match->mask && (match->mask->key.ip.proto == 0xff))
171 mask_allowed |= 1ULL << OVS_KEY_ATTR_ICMP;
176 if (match->key->eth.type == htons(ETH_P_IPV6)) {
177 key_expected |= 1ULL << OVS_KEY_ATTR_IPV6;
178 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
179 mask_allowed |= 1ULL << OVS_KEY_ATTR_IPV6;
181 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
182 if (match->key->ip.proto == IPPROTO_UDP) {
183 key_expected |= 1ULL << OVS_KEY_ATTR_UDP;
184 if (match->mask && (match->mask->key.ip.proto == 0xff))
185 mask_allowed |= 1ULL << OVS_KEY_ATTR_UDP;
188 if (match->key->ip.proto == IPPROTO_SCTP) {
189 key_expected |= 1ULL << OVS_KEY_ATTR_SCTP;
190 if (match->mask && (match->mask->key.ip.proto == 0xff))
191 mask_allowed |= 1ULL << OVS_KEY_ATTR_SCTP;
194 if (match->key->ip.proto == IPPROTO_TCP) {
195 key_expected |= 1ULL << OVS_KEY_ATTR_TCP;
196 key_expected |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
197 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
198 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP;
199 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
203 if (match->key->ip.proto == IPPROTO_ICMPV6) {
204 key_expected |= 1ULL << OVS_KEY_ATTR_ICMPV6;
205 if (match->mask && (match->mask->key.ip.proto == 0xff))
206 mask_allowed |= 1ULL << OVS_KEY_ATTR_ICMPV6;
208 if (match->key->tp.src ==
209 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
210 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
211 key_expected |= 1ULL << OVS_KEY_ATTR_ND;
212 if (match->mask && (match->mask->key.tp.src == htons(0xffff)))
213 mask_allowed |= 1ULL << OVS_KEY_ATTR_ND;
219 if ((key_attrs & key_expected) != key_expected) {
220 /* Key attributes check failed. */
221 OVS_NLERR("Missing expected key attributes (key_attrs=%llx, expected=%llx).\n",
222 (unsigned long long)key_attrs, (unsigned long long)key_expected);
226 if ((mask_attrs & mask_allowed) != mask_attrs) {
227 /* Mask attributes check failed. */
228 OVS_NLERR("Contain more than allowed mask fields (mask_attrs=%llx, mask_allowed=%llx).\n",
229 (unsigned long long)mask_attrs, (unsigned long long)mask_allowed);
236 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
237 static const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
238 [OVS_KEY_ATTR_ENCAP] = -1,
239 [OVS_KEY_ATTR_PRIORITY] = sizeof(u32),
240 [OVS_KEY_ATTR_IN_PORT] = sizeof(u32),
241 [OVS_KEY_ATTR_SKB_MARK] = sizeof(u32),
242 [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
243 [OVS_KEY_ATTR_VLAN] = sizeof(__be16),
244 [OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16),
245 [OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4),
246 [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
247 [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
248 [OVS_KEY_ATTR_TCP_FLAGS] = sizeof(__be16),
249 [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
250 [OVS_KEY_ATTR_SCTP] = sizeof(struct ovs_key_sctp),
251 [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
252 [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
253 [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
254 [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
255 [OVS_KEY_ATTR_DP_HASH] = sizeof(u32),
256 [OVS_KEY_ATTR_TUNNEL] = -1,
259 static bool is_all_zero(const u8 *fp, size_t size)
266 for (i = 0; i < size; i++)
273 static int __parse_flow_nlattrs(const struct nlattr *attr,
274 const struct nlattr *a[],
275 u64 *attrsp, bool nz)
277 const struct nlattr *nla;
282 nla_for_each_nested(nla, attr, rem) {
283 u16 type = nla_type(nla);
286 if (type > OVS_KEY_ATTR_MAX) {
287 OVS_NLERR("Unknown key attribute (type=%d, max=%d).\n",
288 type, OVS_KEY_ATTR_MAX);
292 if (attrs & (1ULL << type)) {
293 OVS_NLERR("Duplicate key attribute (type %d).\n", type);
297 expected_len = ovs_key_lens[type];
298 if (nla_len(nla) != expected_len && expected_len != -1) {
299 OVS_NLERR("Key attribute has unexpected length (type=%d"
300 ", length=%d, expected=%d).\n", type,
301 nla_len(nla), expected_len);
305 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
306 attrs |= 1ULL << type;
311 OVS_NLERR("Message has %d unknown bytes.\n", rem);
319 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
320 const struct nlattr *a[], u64 *attrsp)
322 return __parse_flow_nlattrs(attr, a, attrsp, true);
325 static int parse_flow_nlattrs(const struct nlattr *attr,
326 const struct nlattr *a[], u64 *attrsp)
328 return __parse_flow_nlattrs(attr, a, attrsp, false);
331 static int ipv4_tun_from_nlattr(const struct nlattr *attr,
332 struct sw_flow_match *match, bool is_mask)
337 __be16 tun_flags = 0;
339 nla_for_each_nested(a, attr, rem) {
340 int type = nla_type(a);
341 static const u32 ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
342 [OVS_TUNNEL_KEY_ATTR_ID] = sizeof(u64),
343 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = sizeof(u32),
344 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = sizeof(u32),
345 [OVS_TUNNEL_KEY_ATTR_TOS] = 1,
346 [OVS_TUNNEL_KEY_ATTR_TTL] = 1,
347 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = 0,
348 [OVS_TUNNEL_KEY_ATTR_CSUM] = 0,
351 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
352 OVS_NLERR("Unknown IPv4 tunnel attribute (type=%d, max=%d).\n",
353 type, OVS_TUNNEL_KEY_ATTR_MAX);
357 if (ovs_tunnel_key_lens[type] != nla_len(a)) {
358 OVS_NLERR("IPv4 tunnel attribute type has unexpected "
359 " length (type=%d, length=%d, expected=%d).\n",
360 type, nla_len(a), ovs_tunnel_key_lens[type]);
365 case OVS_TUNNEL_KEY_ATTR_ID:
366 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
367 nla_get_be64(a), is_mask);
368 tun_flags |= TUNNEL_KEY;
370 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
371 SW_FLOW_KEY_PUT(match, tun_key.ipv4_src,
372 nla_get_be32(a), is_mask);
374 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
375 SW_FLOW_KEY_PUT(match, tun_key.ipv4_dst,
376 nla_get_be32(a), is_mask);
378 case OVS_TUNNEL_KEY_ATTR_TOS:
379 SW_FLOW_KEY_PUT(match, tun_key.ipv4_tos,
380 nla_get_u8(a), is_mask);
382 case OVS_TUNNEL_KEY_ATTR_TTL:
383 SW_FLOW_KEY_PUT(match, tun_key.ipv4_ttl,
384 nla_get_u8(a), is_mask);
387 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
388 tun_flags |= TUNNEL_DONT_FRAGMENT;
390 case OVS_TUNNEL_KEY_ATTR_CSUM:
391 tun_flags |= TUNNEL_CSUM;
398 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
401 OVS_NLERR("IPv4 tunnel attribute has %d unknown bytes.\n", rem);
406 if (!match->key->tun_key.ipv4_dst) {
407 OVS_NLERR("IPv4 tunnel destination address is zero.\n");
412 OVS_NLERR("IPv4 tunnel TTL not specified.\n");
420 static int ipv4_tun_to_nlattr(struct sk_buff *skb,
421 const struct ovs_key_ipv4_tunnel *tun_key,
422 const struct ovs_key_ipv4_tunnel *output)
426 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
430 if (output->tun_flags & TUNNEL_KEY &&
431 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
433 if (output->ipv4_src &&
434 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC, output->ipv4_src))
436 if (output->ipv4_dst &&
437 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST, output->ipv4_dst))
439 if (output->ipv4_tos &&
440 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->ipv4_tos))
442 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ipv4_ttl))
444 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
445 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
447 if ((output->tun_flags & TUNNEL_CSUM) &&
448 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
451 nla_nest_end(skb, nla);
456 static int metadata_from_nlattrs(struct sw_flow_match *match, u64 *attrs,
457 const struct nlattr **a, bool is_mask)
459 if (*attrs & (1ULL << OVS_KEY_ATTR_DP_HASH)) {
460 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
462 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
463 *attrs &= ~(1ULL << OVS_KEY_ATTR_DP_HASH);
466 if (*attrs & (1ULL << OVS_KEY_ATTR_PRIORITY)) {
467 SW_FLOW_KEY_PUT(match, phy.priority,
468 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
469 *attrs &= ~(1ULL << OVS_KEY_ATTR_PRIORITY);
472 if (*attrs & (1ULL << OVS_KEY_ATTR_IN_PORT)) {
473 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
476 in_port = 0xffffffff; /* Always exact match in_port. */
477 else if (in_port >= DP_MAX_PORTS)
480 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
481 *attrs &= ~(1ULL << OVS_KEY_ATTR_IN_PORT);
482 } else if (!is_mask) {
483 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
486 if (*attrs & (1ULL << OVS_KEY_ATTR_SKB_MARK)) {
487 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
489 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
490 *attrs &= ~(1ULL << OVS_KEY_ATTR_SKB_MARK);
492 if (*attrs & (1ULL << OVS_KEY_ATTR_TUNNEL)) {
493 if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
496 *attrs &= ~(1ULL << OVS_KEY_ATTR_TUNNEL);
501 static int ovs_key_from_nlattrs(struct sw_flow_match *match, u64 attrs,
502 const struct nlattr **a, bool is_mask)
505 u64 orig_attrs = attrs;
507 err = metadata_from_nlattrs(match, &attrs, a, is_mask);
511 if (attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) {
512 const struct ovs_key_ethernet *eth_key;
514 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
515 SW_FLOW_KEY_MEMCPY(match, eth.src,
516 eth_key->eth_src, ETH_ALEN, is_mask);
517 SW_FLOW_KEY_MEMCPY(match, eth.dst,
518 eth_key->eth_dst, ETH_ALEN, is_mask);
519 attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERNET);
522 if (attrs & (1ULL << OVS_KEY_ATTR_VLAN)) {
525 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
526 if (!(tci & htons(VLAN_TAG_PRESENT))) {
528 OVS_NLERR("VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.\n");
530 OVS_NLERR("VLAN TCI does not have VLAN_TAG_PRESENT bit set.\n");
535 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
536 attrs &= ~(1ULL << OVS_KEY_ATTR_VLAN);
538 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
540 if (attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) {
543 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
545 /* Always exact match EtherType. */
546 eth_type = htons(0xffff);
547 } else if (ntohs(eth_type) < ETH_P_802_3_MIN) {
548 OVS_NLERR("EtherType is less than minimum (type=%x, min=%x).\n",
549 ntohs(eth_type), ETH_P_802_3_MIN);
553 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
554 attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
555 } else if (!is_mask) {
556 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
559 if (attrs & (1ULL << OVS_KEY_ATTR_IPV4)) {
560 const struct ovs_key_ipv4 *ipv4_key;
562 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
563 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
564 OVS_NLERR("Unknown IPv4 fragment type (value=%d, max=%d).\n",
565 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
568 SW_FLOW_KEY_PUT(match, ip.proto,
569 ipv4_key->ipv4_proto, is_mask);
570 SW_FLOW_KEY_PUT(match, ip.tos,
571 ipv4_key->ipv4_tos, is_mask);
572 SW_FLOW_KEY_PUT(match, ip.ttl,
573 ipv4_key->ipv4_ttl, is_mask);
574 SW_FLOW_KEY_PUT(match, ip.frag,
575 ipv4_key->ipv4_frag, is_mask);
576 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
577 ipv4_key->ipv4_src, is_mask);
578 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
579 ipv4_key->ipv4_dst, is_mask);
580 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV4);
583 if (attrs & (1ULL << OVS_KEY_ATTR_IPV6)) {
584 const struct ovs_key_ipv6 *ipv6_key;
586 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
587 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
588 OVS_NLERR("Unknown IPv6 fragment type (value=%d, max=%d).\n",
589 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
592 SW_FLOW_KEY_PUT(match, ipv6.label,
593 ipv6_key->ipv6_label, is_mask);
594 SW_FLOW_KEY_PUT(match, ip.proto,
595 ipv6_key->ipv6_proto, is_mask);
596 SW_FLOW_KEY_PUT(match, ip.tos,
597 ipv6_key->ipv6_tclass, is_mask);
598 SW_FLOW_KEY_PUT(match, ip.ttl,
599 ipv6_key->ipv6_hlimit, is_mask);
600 SW_FLOW_KEY_PUT(match, ip.frag,
601 ipv6_key->ipv6_frag, is_mask);
602 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
604 sizeof(match->key->ipv6.addr.src),
606 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
608 sizeof(match->key->ipv6.addr.dst),
611 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV6);
614 if (attrs & (1ULL << OVS_KEY_ATTR_ARP)) {
615 const struct ovs_key_arp *arp_key;
617 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
618 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
619 OVS_NLERR("Unknown ARP opcode (opcode=%d).\n",
624 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
625 arp_key->arp_sip, is_mask);
626 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
627 arp_key->arp_tip, is_mask);
628 SW_FLOW_KEY_PUT(match, ip.proto,
629 ntohs(arp_key->arp_op), is_mask);
630 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
631 arp_key->arp_sha, ETH_ALEN, is_mask);
632 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
633 arp_key->arp_tha, ETH_ALEN, is_mask);
635 attrs &= ~(1ULL << OVS_KEY_ATTR_ARP);
638 if (attrs & (1ULL << OVS_KEY_ATTR_TCP)) {
639 const struct ovs_key_tcp *tcp_key;
641 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
642 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
643 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
644 attrs &= ~(1ULL << OVS_KEY_ATTR_TCP);
647 if (attrs & (1ULL << OVS_KEY_ATTR_TCP_FLAGS)) {
648 if (orig_attrs & (1ULL << OVS_KEY_ATTR_IPV4)) {
649 SW_FLOW_KEY_PUT(match, tp.flags,
650 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
653 SW_FLOW_KEY_PUT(match, tp.flags,
654 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
657 attrs &= ~(1ULL << OVS_KEY_ATTR_TCP_FLAGS);
660 if (attrs & (1ULL << OVS_KEY_ATTR_UDP)) {
661 const struct ovs_key_udp *udp_key;
663 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
664 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
665 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
666 attrs &= ~(1ULL << OVS_KEY_ATTR_UDP);
669 if (attrs & (1ULL << OVS_KEY_ATTR_SCTP)) {
670 const struct ovs_key_sctp *sctp_key;
672 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
673 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
674 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
675 attrs &= ~(1ULL << OVS_KEY_ATTR_SCTP);
678 if (attrs & (1ULL << OVS_KEY_ATTR_ICMP)) {
679 const struct ovs_key_icmp *icmp_key;
681 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
682 SW_FLOW_KEY_PUT(match, tp.src,
683 htons(icmp_key->icmp_type), is_mask);
684 SW_FLOW_KEY_PUT(match, tp.dst,
685 htons(icmp_key->icmp_code), is_mask);
686 attrs &= ~(1ULL << OVS_KEY_ATTR_ICMP);
689 if (attrs & (1ULL << OVS_KEY_ATTR_ICMPV6)) {
690 const struct ovs_key_icmpv6 *icmpv6_key;
692 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
693 SW_FLOW_KEY_PUT(match, tp.src,
694 htons(icmpv6_key->icmpv6_type), is_mask);
695 SW_FLOW_KEY_PUT(match, tp.dst,
696 htons(icmpv6_key->icmpv6_code), is_mask);
697 attrs &= ~(1ULL << OVS_KEY_ATTR_ICMPV6);
700 if (attrs & (1ULL << OVS_KEY_ATTR_ND)) {
701 const struct ovs_key_nd *nd_key;
703 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
704 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
706 sizeof(match->key->ipv6.nd.target),
708 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
709 nd_key->nd_sll, ETH_ALEN, is_mask);
710 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
711 nd_key->nd_tll, ETH_ALEN, is_mask);
712 attrs &= ~(1ULL << OVS_KEY_ATTR_ND);
721 static void sw_flow_mask_set(struct sw_flow_mask *mask,
722 struct sw_flow_key_range *range, u8 val)
724 u8 *m = (u8 *)&mask->key + range->start;
726 mask->range = *range;
727 memset(m, val, range_n_bytes(range));
731 * ovs_nla_get_match - parses Netlink attributes into a flow key and
732 * mask. In case the 'mask' is NULL, the flow is treated as exact match
733 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
734 * does not include any don't care bit.
735 * @match: receives the extracted flow match information.
736 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
737 * sequence. The fields should of the packet that triggered the creation
739 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
740 * attribute specifies the mask field of the wildcarded flow.
742 int ovs_nla_get_match(struct sw_flow_match *match,
743 const struct nlattr *key,
744 const struct nlattr *mask)
746 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
747 const struct nlattr *encap;
750 bool encap_valid = false;
753 err = parse_flow_nlattrs(key, a, &key_attrs);
757 if ((key_attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
758 (key_attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) &&
759 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
762 if (!((key_attrs & (1ULL << OVS_KEY_ATTR_VLAN)) &&
763 (key_attrs & (1ULL << OVS_KEY_ATTR_ENCAP)))) {
764 OVS_NLERR("Invalid Vlan frame.\n");
768 key_attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
769 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
770 encap = a[OVS_KEY_ATTR_ENCAP];
771 key_attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
774 if (tci & htons(VLAN_TAG_PRESENT)) {
775 err = parse_flow_nlattrs(encap, a, &key_attrs);
779 /* Corner case for truncated 802.1Q header. */
780 if (nla_len(encap)) {
781 OVS_NLERR("Truncated 802.1Q header has non-zero encap attribute.\n");
785 OVS_NLERR("Encap attribute is set for a non-VLAN frame.\n");
790 err = ovs_key_from_nlattrs(match, key_attrs, a, false);
795 err = parse_flow_mask_nlattrs(mask, a, &mask_attrs);
799 if (mask_attrs & 1ULL << OVS_KEY_ATTR_ENCAP) {
804 OVS_NLERR("Encap mask attribute is set for non-VLAN frame.\n");
808 mask_attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
809 if (a[OVS_KEY_ATTR_ETHERTYPE])
810 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
812 if (eth_type == htons(0xffff)) {
813 mask_attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
814 encap = a[OVS_KEY_ATTR_ENCAP];
815 err = parse_flow_mask_nlattrs(encap, a, &mask_attrs);
817 OVS_NLERR("VLAN frames must have an exact match on the TPID (mask=%x).\n",
822 if (a[OVS_KEY_ATTR_VLAN])
823 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
825 if (!(tci & htons(VLAN_TAG_PRESENT))) {
826 OVS_NLERR("VLAN tag present bit must have an exact match (tci_mask=%x).\n", ntohs(tci));
831 err = ovs_key_from_nlattrs(match, mask_attrs, a, true);
835 /* Populate exact match flow's key mask. */
837 sw_flow_mask_set(match->mask, &match->range, 0xff);
840 if (!match_validate(match, key_attrs, mask_attrs))
847 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
848 * @flow: Receives extracted in_port, priority, tun_key and skb_mark.
849 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
852 * This parses a series of Netlink attributes that form a flow key, which must
853 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
854 * get the metadata, that is, the parts of the flow key that cannot be
855 * extracted from the packet itself.
858 int ovs_nla_get_flow_metadata(struct sw_flow *flow,
859 const struct nlattr *attr)
861 struct ovs_key_ipv4_tunnel *tun_key = &flow->key.tun_key;
862 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
865 struct sw_flow_match match;
867 flow->key.phy.in_port = DP_MAX_PORTS;
868 flow->key.phy.priority = 0;
869 flow->key.phy.skb_mark = 0;
870 flow->key.ovs_flow_hash = 0;
871 memset(tun_key, 0, sizeof(flow->key.tun_key));
873 err = parse_flow_nlattrs(attr, a, &attrs);
877 memset(&match, 0, sizeof(match));
878 match.key = &flow->key;
880 err = metadata_from_nlattrs(&match, &attrs, a, false);
887 int ovs_nla_put_flow(const struct sw_flow_key *swkey,
888 const struct sw_flow_key *output, struct sk_buff *skb)
890 struct ovs_key_ethernet *eth_key;
891 struct nlattr *nla, *encap;
892 bool is_mask = (swkey != output);
894 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
895 goto nla_put_failure;
897 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
898 goto nla_put_failure;
900 if ((swkey->tun_key.ipv4_dst || is_mask) &&
901 ipv4_tun_to_nlattr(skb, &swkey->tun_key, &output->tun_key))
902 goto nla_put_failure;
904 if (swkey->phy.in_port == DP_MAX_PORTS) {
905 if (is_mask && (output->phy.in_port == 0xffff))
906 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
907 goto nla_put_failure;
910 upper_u16 = !is_mask ? 0 : 0xffff;
912 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
913 (upper_u16 << 16) | output->phy.in_port))
914 goto nla_put_failure;
917 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
918 goto nla_put_failure;
920 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
922 goto nla_put_failure;
924 eth_key = nla_data(nla);
925 ether_addr_copy(eth_key->eth_src, output->eth.src);
926 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
928 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
930 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
931 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
932 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
933 goto nla_put_failure;
934 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
940 if (swkey->eth.type == htons(ETH_P_802_2)) {
942 * Ethertype 802.2 is represented in the netlink with omitted
943 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
944 * 0xffff in the mask attribute. Ethertype can also
947 if (is_mask && output->eth.type)
948 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
950 goto nla_put_failure;
954 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
955 goto nla_put_failure;
957 if (swkey->eth.type == htons(ETH_P_IP)) {
958 struct ovs_key_ipv4 *ipv4_key;
960 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
962 goto nla_put_failure;
963 ipv4_key = nla_data(nla);
964 ipv4_key->ipv4_src = output->ipv4.addr.src;
965 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
966 ipv4_key->ipv4_proto = output->ip.proto;
967 ipv4_key->ipv4_tos = output->ip.tos;
968 ipv4_key->ipv4_ttl = output->ip.ttl;
969 ipv4_key->ipv4_frag = output->ip.frag;
970 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
971 struct ovs_key_ipv6 *ipv6_key;
973 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
975 goto nla_put_failure;
976 ipv6_key = nla_data(nla);
977 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
978 sizeof(ipv6_key->ipv6_src));
979 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
980 sizeof(ipv6_key->ipv6_dst));
981 ipv6_key->ipv6_label = output->ipv6.label;
982 ipv6_key->ipv6_proto = output->ip.proto;
983 ipv6_key->ipv6_tclass = output->ip.tos;
984 ipv6_key->ipv6_hlimit = output->ip.ttl;
985 ipv6_key->ipv6_frag = output->ip.frag;
986 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
987 swkey->eth.type == htons(ETH_P_RARP)) {
988 struct ovs_key_arp *arp_key;
990 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
992 goto nla_put_failure;
993 arp_key = nla_data(nla);
994 memset(arp_key, 0, sizeof(struct ovs_key_arp));
995 arp_key->arp_sip = output->ipv4.addr.src;
996 arp_key->arp_tip = output->ipv4.addr.dst;
997 arp_key->arp_op = htons(output->ip.proto);
998 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
999 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1002 if ((swkey->eth.type == htons(ETH_P_IP) ||
1003 swkey->eth.type == htons(ETH_P_IPV6)) &&
1004 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1006 if (swkey->ip.proto == IPPROTO_TCP) {
1007 struct ovs_key_tcp *tcp_key;
1009 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1011 goto nla_put_failure;
1012 tcp_key = nla_data(nla);
1013 tcp_key->tcp_src = output->tp.src;
1014 tcp_key->tcp_dst = output->tp.dst;
1015 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1017 goto nla_put_failure;
1018 } else if (swkey->ip.proto == IPPROTO_UDP) {
1019 struct ovs_key_udp *udp_key;
1021 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1023 goto nla_put_failure;
1024 udp_key = nla_data(nla);
1025 udp_key->udp_src = output->tp.src;
1026 udp_key->udp_dst = output->tp.dst;
1027 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1028 struct ovs_key_sctp *sctp_key;
1030 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1032 goto nla_put_failure;
1033 sctp_key = nla_data(nla);
1034 sctp_key->sctp_src = output->tp.src;
1035 sctp_key->sctp_dst = output->tp.dst;
1036 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1037 swkey->ip.proto == IPPROTO_ICMP) {
1038 struct ovs_key_icmp *icmp_key;
1040 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1042 goto nla_put_failure;
1043 icmp_key = nla_data(nla);
1044 icmp_key->icmp_type = ntohs(output->tp.src);
1045 icmp_key->icmp_code = ntohs(output->tp.dst);
1046 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1047 swkey->ip.proto == IPPROTO_ICMPV6) {
1048 struct ovs_key_icmpv6 *icmpv6_key;
1050 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1051 sizeof(*icmpv6_key));
1053 goto nla_put_failure;
1054 icmpv6_key = nla_data(nla);
1055 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1056 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1058 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1059 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1060 struct ovs_key_nd *nd_key;
1062 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1064 goto nla_put_failure;
1065 nd_key = nla_data(nla);
1066 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1067 sizeof(nd_key->nd_target));
1068 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1069 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1076 nla_nest_end(skb, encap);
1084 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1086 struct sw_flow_actions *ovs_nla_alloc_flow_actions(int size)
1088 struct sw_flow_actions *sfa;
1090 if (size > MAX_ACTIONS_BUFSIZE)
1091 return ERR_PTR(-EINVAL);
1093 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1095 return ERR_PTR(-ENOMEM);
1097 sfa->actions_len = 0;
1101 /* RCU callback used by ovs_nla_free_flow_actions. */
1102 static void rcu_free_acts_callback(struct rcu_head *rcu)
1104 struct sw_flow_actions *sf_acts = container_of(rcu,
1105 struct sw_flow_actions, rcu);
1109 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1110 * The caller must hold rcu_read_lock for this to be sensible. */
1111 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1113 call_rcu(&sf_acts->rcu, rcu_free_acts_callback);
1116 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1120 struct sw_flow_actions *acts;
1122 int req_size = NLA_ALIGN(attr_len);
1123 int next_offset = offsetof(struct sw_flow_actions, actions) +
1124 (*sfa)->actions_len;
1126 if (req_size <= (ksize(*sfa) - next_offset))
1129 new_acts_size = ksize(*sfa) * 2;
1131 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1132 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1133 return ERR_PTR(-EMSGSIZE);
1134 new_acts_size = MAX_ACTIONS_BUFSIZE;
1137 acts = ovs_nla_alloc_flow_actions(new_acts_size);
1139 return (void *)acts;
1141 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1142 acts->actions_len = (*sfa)->actions_len;
1147 (*sfa)->actions_len += req_size;
1148 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1151 static int add_action(struct sw_flow_actions **sfa, int attrtype, void *data, int len)
1155 a = reserve_sfa_size(sfa, nla_attr_size(len));
1159 a->nla_type = attrtype;
1160 a->nla_len = nla_attr_size(len);
1163 memcpy(nla_data(a), data, len);
1164 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1169 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1172 int used = (*sfa)->actions_len;
1175 err = add_action(sfa, attrtype, NULL, 0);
1182 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1185 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1188 a->nla_len = sfa->actions_len - st_offset;
1191 static int validate_and_copy_sample(const struct nlattr *attr,
1192 const struct sw_flow_key *key, int depth,
1193 struct sw_flow_actions **sfa)
1195 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1196 const struct nlattr *probability, *actions;
1197 const struct nlattr *a;
1198 int rem, start, err, st_acts;
1200 memset(attrs, 0, sizeof(attrs));
1201 nla_for_each_nested(a, attr, rem) {
1202 int type = nla_type(a);
1203 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1210 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1211 if (!probability || nla_len(probability) != sizeof(u32))
1214 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1215 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1218 /* validation done, copy sample action. */
1219 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE);
1222 err = add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1223 nla_data(probability), sizeof(u32));
1226 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS);
1230 err = ovs_nla_copy_actions(actions, key, depth + 1, sfa);
1234 add_nested_action_end(*sfa, st_acts);
1235 add_nested_action_end(*sfa, start);
1240 static int validate_tp_port(const struct sw_flow_key *flow_key)
1242 if ((flow_key->eth.type == htons(ETH_P_IP) ||
1243 flow_key->eth.type == htons(ETH_P_IPV6)) &&
1244 (flow_key->tp.src || flow_key->tp.dst))
1250 void ovs_match_init(struct sw_flow_match *match,
1251 struct sw_flow_key *key,
1252 struct sw_flow_mask *mask)
1254 memset(match, 0, sizeof(*match));
1258 memset(key, 0, sizeof(*key));
1261 memset(&mask->key, 0, sizeof(mask->key));
1262 mask->range.start = mask->range.end = 0;
1266 static int validate_and_copy_set_tun(const struct nlattr *attr,
1267 struct sw_flow_actions **sfa)
1269 struct sw_flow_match match;
1270 struct sw_flow_key key;
1273 ovs_match_init(&match, &key, NULL);
1274 err = ipv4_tun_from_nlattr(nla_data(attr), &match, false);
1278 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET);
1282 err = add_action(sfa, OVS_KEY_ATTR_IPV4_TUNNEL, &match.key->tun_key,
1283 sizeof(match.key->tun_key));
1284 add_nested_action_end(*sfa, start);
1289 static int validate_set(const struct nlattr *a,
1290 const struct sw_flow_key *flow_key,
1291 struct sw_flow_actions **sfa,
1294 const struct nlattr *ovs_key = nla_data(a);
1295 int key_type = nla_type(ovs_key);
1297 /* There can be only one key in a action */
1298 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1301 if (key_type > OVS_KEY_ATTR_MAX ||
1302 (ovs_key_lens[key_type] != nla_len(ovs_key) &&
1303 ovs_key_lens[key_type] != -1))
1307 const struct ovs_key_ipv4 *ipv4_key;
1308 const struct ovs_key_ipv6 *ipv6_key;
1311 case OVS_KEY_ATTR_PRIORITY:
1312 case OVS_KEY_ATTR_SKB_MARK:
1313 case OVS_KEY_ATTR_ETHERNET:
1316 case OVS_KEY_ATTR_TUNNEL:
1318 err = validate_and_copy_set_tun(a, sfa);
1323 case OVS_KEY_ATTR_IPV4:
1324 if (flow_key->eth.type != htons(ETH_P_IP))
1327 if (!flow_key->ip.proto)
1330 ipv4_key = nla_data(ovs_key);
1331 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
1334 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
1339 case OVS_KEY_ATTR_IPV6:
1340 if (flow_key->eth.type != htons(ETH_P_IPV6))
1343 if (!flow_key->ip.proto)
1346 ipv6_key = nla_data(ovs_key);
1347 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
1350 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
1353 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
1358 case OVS_KEY_ATTR_TCP:
1359 if (flow_key->ip.proto != IPPROTO_TCP)
1362 return validate_tp_port(flow_key);
1364 case OVS_KEY_ATTR_UDP:
1365 if (flow_key->ip.proto != IPPROTO_UDP)
1368 return validate_tp_port(flow_key);
1370 case OVS_KEY_ATTR_SCTP:
1371 if (flow_key->ip.proto != IPPROTO_SCTP)
1374 return validate_tp_port(flow_key);
1383 static int validate_userspace(const struct nlattr *attr)
1385 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
1386 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
1387 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
1389 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
1392 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
1393 attr, userspace_policy);
1397 if (!a[OVS_USERSPACE_ATTR_PID] ||
1398 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
1404 static int copy_action(const struct nlattr *from,
1405 struct sw_flow_actions **sfa)
1407 int totlen = NLA_ALIGN(from->nla_len);
1410 to = reserve_sfa_size(sfa, from->nla_len);
1414 memcpy(to, from, totlen);
1418 int ovs_nla_copy_actions(const struct nlattr *attr,
1419 const struct sw_flow_key *key,
1421 struct sw_flow_actions **sfa)
1423 const struct nlattr *a;
1426 if (depth >= SAMPLE_ACTION_DEPTH)
1429 nla_for_each_nested(a, attr, rem) {
1430 /* Expected argument lengths, (u32)-1 for variable length. */
1431 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
1432 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
1433 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
1434 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
1435 [OVS_ACTION_ATTR_POP_VLAN] = 0,
1436 [OVS_ACTION_ATTR_SET] = (u32)-1,
1437 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
1438 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash)
1440 const struct ovs_action_push_vlan *vlan;
1441 int type = nla_type(a);
1444 if (type > OVS_ACTION_ATTR_MAX ||
1445 (action_lens[type] != nla_len(a) &&
1446 action_lens[type] != (u32)-1))
1451 case OVS_ACTION_ATTR_UNSPEC:
1454 case OVS_ACTION_ATTR_USERSPACE:
1455 err = validate_userspace(a);
1460 case OVS_ACTION_ATTR_OUTPUT:
1461 if (nla_get_u32(a) >= DP_MAX_PORTS)
1465 case OVS_ACTION_ATTR_HASH: {
1466 const struct ovs_action_hash *act_hash = nla_data(a);
1468 switch (act_hash->hash_alg) {
1469 case OVS_HASH_ALG_L4:
1478 case OVS_ACTION_ATTR_POP_VLAN:
1481 case OVS_ACTION_ATTR_PUSH_VLAN:
1483 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
1485 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
1489 case OVS_ACTION_ATTR_SET:
1490 err = validate_set(a, key, sfa, &skip_copy);
1495 case OVS_ACTION_ATTR_SAMPLE:
1496 err = validate_and_copy_sample(a, key, depth, sfa);
1506 err = copy_action(a, sfa);
1518 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
1520 const struct nlattr *a;
1521 struct nlattr *start;
1524 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
1528 nla_for_each_nested(a, attr, rem) {
1529 int type = nla_type(a);
1530 struct nlattr *st_sample;
1533 case OVS_SAMPLE_ATTR_PROBABILITY:
1534 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
1535 sizeof(u32), nla_data(a)))
1538 case OVS_SAMPLE_ATTR_ACTIONS:
1539 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
1542 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
1545 nla_nest_end(skb, st_sample);
1550 nla_nest_end(skb, start);
1554 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
1556 const struct nlattr *ovs_key = nla_data(a);
1557 int key_type = nla_type(ovs_key);
1558 struct nlattr *start;
1562 case OVS_KEY_ATTR_IPV4_TUNNEL:
1563 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
1567 err = ipv4_tun_to_nlattr(skb, nla_data(ovs_key),
1571 nla_nest_end(skb, start);
1574 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
1582 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
1584 const struct nlattr *a;
1587 nla_for_each_attr(a, attr, len, rem) {
1588 int type = nla_type(a);
1591 case OVS_ACTION_ATTR_SET:
1592 err = set_action_to_attr(a, skb);
1597 case OVS_ACTION_ATTR_SAMPLE:
1598 err = sample_action_to_attr(a, skb);
1603 if (nla_put(skb, type, nla_len(a), nla_data(a)))