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->ipv6.tp.src ==
209 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
210 match->key->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
211 key_expected |= 1ULL << OVS_KEY_ATTR_ND;
212 if (match->mask && (match->mask->key.ipv6.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_TUNNEL] = -1,
258 static bool is_all_zero(const u8 *fp, size_t size)
265 for (i = 0; i < size; i++)
272 static int __parse_flow_nlattrs(const struct nlattr *attr,
273 const struct nlattr *a[],
274 u64 *attrsp, bool nz)
276 const struct nlattr *nla;
281 nla_for_each_nested(nla, attr, rem) {
282 u16 type = nla_type(nla);
285 if (type > OVS_KEY_ATTR_MAX) {
286 OVS_NLERR("Unknown key attribute (type=%d, max=%d).\n",
287 type, OVS_KEY_ATTR_MAX);
291 if (attrs & (1ULL << type)) {
292 OVS_NLERR("Duplicate key attribute (type %d).\n", type);
296 expected_len = ovs_key_lens[type];
297 if (nla_len(nla) != expected_len && expected_len != -1) {
298 OVS_NLERR("Key attribute has unexpected length (type=%d"
299 ", length=%d, expected=%d).\n", type,
300 nla_len(nla), expected_len);
304 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
305 attrs |= 1ULL << type;
310 OVS_NLERR("Message has %d unknown bytes.\n", rem);
318 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
319 const struct nlattr *a[], u64 *attrsp)
321 return __parse_flow_nlattrs(attr, a, attrsp, true);
324 static int parse_flow_nlattrs(const struct nlattr *attr,
325 const struct nlattr *a[], u64 *attrsp)
327 return __parse_flow_nlattrs(attr, a, attrsp, false);
330 static int ipv4_tun_from_nlattr(const struct nlattr *attr,
331 struct sw_flow_match *match, bool is_mask)
336 __be16 tun_flags = 0;
338 nla_for_each_nested(a, attr, rem) {
339 int type = nla_type(a);
340 static const u32 ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
341 [OVS_TUNNEL_KEY_ATTR_ID] = sizeof(u64),
342 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = sizeof(u32),
343 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = sizeof(u32),
344 [OVS_TUNNEL_KEY_ATTR_TOS] = 1,
345 [OVS_TUNNEL_KEY_ATTR_TTL] = 1,
346 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = 0,
347 [OVS_TUNNEL_KEY_ATTR_CSUM] = 0,
350 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
351 OVS_NLERR("Unknown IPv4 tunnel attribute (type=%d, max=%d).\n",
352 type, OVS_TUNNEL_KEY_ATTR_MAX);
356 if (ovs_tunnel_key_lens[type] != nla_len(a)) {
357 OVS_NLERR("IPv4 tunnel attribute type has unexpected "
358 " length (type=%d, length=%d, expected=%d).\n",
359 type, nla_len(a), ovs_tunnel_key_lens[type]);
364 case OVS_TUNNEL_KEY_ATTR_ID:
365 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
366 nla_get_be64(a), is_mask);
367 tun_flags |= TUNNEL_KEY;
369 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
370 SW_FLOW_KEY_PUT(match, tun_key.ipv4_src,
371 nla_get_be32(a), is_mask);
373 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
374 SW_FLOW_KEY_PUT(match, tun_key.ipv4_dst,
375 nla_get_be32(a), is_mask);
377 case OVS_TUNNEL_KEY_ATTR_TOS:
378 SW_FLOW_KEY_PUT(match, tun_key.ipv4_tos,
379 nla_get_u8(a), is_mask);
381 case OVS_TUNNEL_KEY_ATTR_TTL:
382 SW_FLOW_KEY_PUT(match, tun_key.ipv4_ttl,
383 nla_get_u8(a), is_mask);
386 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
387 tun_flags |= TUNNEL_DONT_FRAGMENT;
389 case OVS_TUNNEL_KEY_ATTR_CSUM:
390 tun_flags |= TUNNEL_CSUM;
397 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
400 OVS_NLERR("IPv4 tunnel attribute has %d unknown bytes.\n", rem);
405 if (!match->key->tun_key.ipv4_dst) {
406 OVS_NLERR("IPv4 tunnel destination address is zero.\n");
411 OVS_NLERR("IPv4 tunnel TTL not specified.\n");
419 static int ipv4_tun_to_nlattr(struct sk_buff *skb,
420 const struct ovs_key_ipv4_tunnel *tun_key,
421 const struct ovs_key_ipv4_tunnel *output)
425 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
429 if (output->tun_flags & TUNNEL_KEY &&
430 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
432 if (output->ipv4_src &&
433 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC, output->ipv4_src))
435 if (output->ipv4_dst &&
436 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST, output->ipv4_dst))
438 if (output->ipv4_tos &&
439 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->ipv4_tos))
441 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ipv4_ttl))
443 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
444 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
446 if ((output->tun_flags & TUNNEL_CSUM) &&
447 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
450 nla_nest_end(skb, nla);
455 static int metadata_from_nlattrs(struct sw_flow_match *match, u64 *attrs,
456 const struct nlattr **a, bool is_mask)
458 if (*attrs & (1ULL << OVS_KEY_ATTR_PRIORITY)) {
459 SW_FLOW_KEY_PUT(match, phy.priority,
460 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
461 *attrs &= ~(1ULL << OVS_KEY_ATTR_PRIORITY);
464 if (*attrs & (1ULL << OVS_KEY_ATTR_IN_PORT)) {
465 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
468 in_port = 0xffffffff; /* Always exact match in_port. */
469 else if (in_port >= DP_MAX_PORTS)
472 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
473 *attrs &= ~(1ULL << OVS_KEY_ATTR_IN_PORT);
474 } else if (!is_mask) {
475 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
478 if (*attrs & (1ULL << OVS_KEY_ATTR_SKB_MARK)) {
479 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
481 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
482 *attrs &= ~(1ULL << OVS_KEY_ATTR_SKB_MARK);
484 if (*attrs & (1ULL << OVS_KEY_ATTR_TUNNEL)) {
485 if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
488 *attrs &= ~(1ULL << OVS_KEY_ATTR_TUNNEL);
493 static int ovs_key_from_nlattrs(struct sw_flow_match *match, u64 attrs,
494 const struct nlattr **a, bool is_mask)
497 u64 orig_attrs = attrs;
499 err = metadata_from_nlattrs(match, &attrs, a, is_mask);
503 if (attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) {
504 const struct ovs_key_ethernet *eth_key;
506 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
507 SW_FLOW_KEY_MEMCPY(match, eth.src,
508 eth_key->eth_src, ETH_ALEN, is_mask);
509 SW_FLOW_KEY_MEMCPY(match, eth.dst,
510 eth_key->eth_dst, ETH_ALEN, is_mask);
511 attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERNET);
514 if (attrs & (1ULL << OVS_KEY_ATTR_VLAN)) {
517 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
518 if (!(tci & htons(VLAN_TAG_PRESENT))) {
520 OVS_NLERR("VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.\n");
522 OVS_NLERR("VLAN TCI does not have VLAN_TAG_PRESENT bit set.\n");
527 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
528 attrs &= ~(1ULL << OVS_KEY_ATTR_VLAN);
530 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
532 if (attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) {
535 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
537 /* Always exact match EtherType. */
538 eth_type = htons(0xffff);
539 } else if (ntohs(eth_type) < ETH_P_802_3_MIN) {
540 OVS_NLERR("EtherType is less than minimum (type=%x, min=%x).\n",
541 ntohs(eth_type), ETH_P_802_3_MIN);
545 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
546 attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
547 } else if (!is_mask) {
548 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
551 if (attrs & (1ULL << OVS_KEY_ATTR_IPV4)) {
552 const struct ovs_key_ipv4 *ipv4_key;
554 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
555 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
556 OVS_NLERR("Unknown IPv4 fragment type (value=%d, max=%d).\n",
557 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
560 SW_FLOW_KEY_PUT(match, ip.proto,
561 ipv4_key->ipv4_proto, is_mask);
562 SW_FLOW_KEY_PUT(match, ip.tos,
563 ipv4_key->ipv4_tos, is_mask);
564 SW_FLOW_KEY_PUT(match, ip.ttl,
565 ipv4_key->ipv4_ttl, is_mask);
566 SW_FLOW_KEY_PUT(match, ip.frag,
567 ipv4_key->ipv4_frag, is_mask);
568 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
569 ipv4_key->ipv4_src, is_mask);
570 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
571 ipv4_key->ipv4_dst, is_mask);
572 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV4);
575 if (attrs & (1ULL << OVS_KEY_ATTR_IPV6)) {
576 const struct ovs_key_ipv6 *ipv6_key;
578 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
579 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
580 OVS_NLERR("Unknown IPv6 fragment type (value=%d, max=%d).\n",
581 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
584 SW_FLOW_KEY_PUT(match, ipv6.label,
585 ipv6_key->ipv6_label, is_mask);
586 SW_FLOW_KEY_PUT(match, ip.proto,
587 ipv6_key->ipv6_proto, is_mask);
588 SW_FLOW_KEY_PUT(match, ip.tos,
589 ipv6_key->ipv6_tclass, is_mask);
590 SW_FLOW_KEY_PUT(match, ip.ttl,
591 ipv6_key->ipv6_hlimit, is_mask);
592 SW_FLOW_KEY_PUT(match, ip.frag,
593 ipv6_key->ipv6_frag, is_mask);
594 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
596 sizeof(match->key->ipv6.addr.src),
598 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
600 sizeof(match->key->ipv6.addr.dst),
603 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV6);
606 if (attrs & (1ULL << OVS_KEY_ATTR_ARP)) {
607 const struct ovs_key_arp *arp_key;
609 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
610 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
611 OVS_NLERR("Unknown ARP opcode (opcode=%d).\n",
616 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
617 arp_key->arp_sip, is_mask);
618 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
619 arp_key->arp_tip, is_mask);
620 SW_FLOW_KEY_PUT(match, ip.proto,
621 ntohs(arp_key->arp_op), is_mask);
622 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
623 arp_key->arp_sha, ETH_ALEN, is_mask);
624 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
625 arp_key->arp_tha, ETH_ALEN, is_mask);
627 attrs &= ~(1ULL << OVS_KEY_ATTR_ARP);
630 if (attrs & (1ULL << OVS_KEY_ATTR_TCP)) {
631 const struct ovs_key_tcp *tcp_key;
633 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
634 if (orig_attrs & (1ULL << OVS_KEY_ATTR_IPV4)) {
635 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
636 tcp_key->tcp_src, is_mask);
637 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
638 tcp_key->tcp_dst, is_mask);
640 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
641 tcp_key->tcp_src, is_mask);
642 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
643 tcp_key->tcp_dst, is_mask);
645 attrs &= ~(1ULL << OVS_KEY_ATTR_TCP);
648 if (attrs & (1ULL << OVS_KEY_ATTR_TCP_FLAGS)) {
649 if (orig_attrs & (1ULL << OVS_KEY_ATTR_IPV4)) {
650 SW_FLOW_KEY_PUT(match, ipv4.tp.flags,
651 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
654 SW_FLOW_KEY_PUT(match, ipv6.tp.flags,
655 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
658 attrs &= ~(1ULL << OVS_KEY_ATTR_TCP_FLAGS);
661 if (attrs & (1ULL << OVS_KEY_ATTR_UDP)) {
662 const struct ovs_key_udp *udp_key;
664 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
665 if (orig_attrs & (1ULL << OVS_KEY_ATTR_IPV4)) {
666 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
667 udp_key->udp_src, is_mask);
668 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
669 udp_key->udp_dst, is_mask);
671 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
672 udp_key->udp_src, is_mask);
673 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
674 udp_key->udp_dst, is_mask);
676 attrs &= ~(1ULL << OVS_KEY_ATTR_UDP);
679 if (attrs & (1ULL << OVS_KEY_ATTR_SCTP)) {
680 const struct ovs_key_sctp *sctp_key;
682 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
683 if (orig_attrs & (1ULL << OVS_KEY_ATTR_IPV4)) {
684 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
685 sctp_key->sctp_src, is_mask);
686 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
687 sctp_key->sctp_dst, is_mask);
689 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
690 sctp_key->sctp_src, is_mask);
691 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
692 sctp_key->sctp_dst, is_mask);
694 attrs &= ~(1ULL << OVS_KEY_ATTR_SCTP);
697 if (attrs & (1ULL << OVS_KEY_ATTR_ICMP)) {
698 const struct ovs_key_icmp *icmp_key;
700 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
701 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
702 htons(icmp_key->icmp_type), is_mask);
703 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
704 htons(icmp_key->icmp_code), is_mask);
705 attrs &= ~(1ULL << OVS_KEY_ATTR_ICMP);
708 if (attrs & (1ULL << OVS_KEY_ATTR_ICMPV6)) {
709 const struct ovs_key_icmpv6 *icmpv6_key;
711 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
712 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
713 htons(icmpv6_key->icmpv6_type), is_mask);
714 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
715 htons(icmpv6_key->icmpv6_code), is_mask);
716 attrs &= ~(1ULL << OVS_KEY_ATTR_ICMPV6);
719 if (attrs & (1ULL << OVS_KEY_ATTR_ND)) {
720 const struct ovs_key_nd *nd_key;
722 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
723 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
725 sizeof(match->key->ipv6.nd.target),
727 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
728 nd_key->nd_sll, ETH_ALEN, is_mask);
729 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
730 nd_key->nd_tll, ETH_ALEN, is_mask);
731 attrs &= ~(1ULL << OVS_KEY_ATTR_ND);
740 static void sw_flow_mask_set(struct sw_flow_mask *mask,
741 struct sw_flow_key_range *range, u8 val)
743 u8 *m = (u8 *)&mask->key + range->start;
745 mask->range = *range;
746 memset(m, val, range_n_bytes(range));
750 * ovs_nla_get_match - parses Netlink attributes into a flow key and
751 * mask. In case the 'mask' is NULL, the flow is treated as exact match
752 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
753 * does not include any don't care bit.
754 * @match: receives the extracted flow match information.
755 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
756 * sequence. The fields should of the packet that triggered the creation
758 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
759 * attribute specifies the mask field of the wildcarded flow.
761 int ovs_nla_get_match(struct sw_flow_match *match,
762 const struct nlattr *key,
763 const struct nlattr *mask)
765 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
766 const struct nlattr *encap;
769 bool encap_valid = false;
772 err = parse_flow_nlattrs(key, a, &key_attrs);
776 if ((key_attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
777 (key_attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) &&
778 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
781 if (!((key_attrs & (1ULL << OVS_KEY_ATTR_VLAN)) &&
782 (key_attrs & (1ULL << OVS_KEY_ATTR_ENCAP)))) {
783 OVS_NLERR("Invalid Vlan frame.\n");
787 key_attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
788 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
789 encap = a[OVS_KEY_ATTR_ENCAP];
790 key_attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
793 if (tci & htons(VLAN_TAG_PRESENT)) {
794 err = parse_flow_nlattrs(encap, a, &key_attrs);
798 /* Corner case for truncated 802.1Q header. */
799 if (nla_len(encap)) {
800 OVS_NLERR("Truncated 802.1Q header has non-zero encap attribute.\n");
804 OVS_NLERR("Encap attribute is set for a non-VLAN frame.\n");
809 err = ovs_key_from_nlattrs(match, key_attrs, a, false);
814 err = parse_flow_mask_nlattrs(mask, a, &mask_attrs);
818 if (mask_attrs & 1ULL << OVS_KEY_ATTR_ENCAP) {
823 OVS_NLERR("Encap mask attribute is set for non-VLAN frame.\n");
827 mask_attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
828 if (a[OVS_KEY_ATTR_ETHERTYPE])
829 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
831 if (eth_type == htons(0xffff)) {
832 mask_attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
833 encap = a[OVS_KEY_ATTR_ENCAP];
834 err = parse_flow_mask_nlattrs(encap, a, &mask_attrs);
836 OVS_NLERR("VLAN frames must have an exact match on the TPID (mask=%x).\n",
841 if (a[OVS_KEY_ATTR_VLAN])
842 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
844 if (!(tci & htons(VLAN_TAG_PRESENT))) {
845 OVS_NLERR("VLAN tag present bit must have an exact match (tci_mask=%x).\n", ntohs(tci));
850 err = ovs_key_from_nlattrs(match, mask_attrs, a, true);
854 /* Populate exact match flow's key mask. */
856 sw_flow_mask_set(match->mask, &match->range, 0xff);
859 if (!match_validate(match, key_attrs, mask_attrs))
866 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
867 * @flow: Receives extracted in_port, priority, tun_key and skb_mark.
868 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
871 * This parses a series of Netlink attributes that form a flow key, which must
872 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
873 * get the metadata, that is, the parts of the flow key that cannot be
874 * extracted from the packet itself.
877 int ovs_nla_get_flow_metadata(struct sw_flow *flow,
878 const struct nlattr *attr)
880 struct ovs_key_ipv4_tunnel *tun_key = &flow->key.tun_key;
881 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
884 struct sw_flow_match match;
886 flow->key.phy.in_port = DP_MAX_PORTS;
887 flow->key.phy.priority = 0;
888 flow->key.phy.skb_mark = 0;
889 memset(tun_key, 0, sizeof(flow->key.tun_key));
891 err = parse_flow_nlattrs(attr, a, &attrs);
895 memset(&match, 0, sizeof(match));
896 match.key = &flow->key;
898 err = metadata_from_nlattrs(&match, &attrs, a, false);
905 int ovs_nla_put_flow(const struct sw_flow_key *swkey,
906 const struct sw_flow_key *output, struct sk_buff *skb)
908 struct ovs_key_ethernet *eth_key;
909 struct nlattr *nla, *encap;
910 bool is_mask = (swkey != output);
912 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
913 goto nla_put_failure;
915 if ((swkey->tun_key.ipv4_dst || is_mask) &&
916 ipv4_tun_to_nlattr(skb, &swkey->tun_key, &output->tun_key))
917 goto nla_put_failure;
919 if (swkey->phy.in_port == DP_MAX_PORTS) {
920 if (is_mask && (output->phy.in_port == 0xffff))
921 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
922 goto nla_put_failure;
925 upper_u16 = !is_mask ? 0 : 0xffff;
927 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
928 (upper_u16 << 16) | output->phy.in_port))
929 goto nla_put_failure;
932 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
933 goto nla_put_failure;
935 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
937 goto nla_put_failure;
939 eth_key = nla_data(nla);
940 ether_addr_copy(eth_key->eth_src, output->eth.src);
941 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
943 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
945 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
946 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
947 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
948 goto nla_put_failure;
949 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
955 if (swkey->eth.type == htons(ETH_P_802_2)) {
957 * Ethertype 802.2 is represented in the netlink with omitted
958 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
959 * 0xffff in the mask attribute. Ethertype can also
962 if (is_mask && output->eth.type)
963 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
965 goto nla_put_failure;
969 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
970 goto nla_put_failure;
972 if (swkey->eth.type == htons(ETH_P_IP)) {
973 struct ovs_key_ipv4 *ipv4_key;
975 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
977 goto nla_put_failure;
978 ipv4_key = nla_data(nla);
979 ipv4_key->ipv4_src = output->ipv4.addr.src;
980 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
981 ipv4_key->ipv4_proto = output->ip.proto;
982 ipv4_key->ipv4_tos = output->ip.tos;
983 ipv4_key->ipv4_ttl = output->ip.ttl;
984 ipv4_key->ipv4_frag = output->ip.frag;
985 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
986 struct ovs_key_ipv6 *ipv6_key;
988 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
990 goto nla_put_failure;
991 ipv6_key = nla_data(nla);
992 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
993 sizeof(ipv6_key->ipv6_src));
994 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
995 sizeof(ipv6_key->ipv6_dst));
996 ipv6_key->ipv6_label = output->ipv6.label;
997 ipv6_key->ipv6_proto = output->ip.proto;
998 ipv6_key->ipv6_tclass = output->ip.tos;
999 ipv6_key->ipv6_hlimit = output->ip.ttl;
1000 ipv6_key->ipv6_frag = output->ip.frag;
1001 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1002 swkey->eth.type == htons(ETH_P_RARP)) {
1003 struct ovs_key_arp *arp_key;
1005 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1007 goto nla_put_failure;
1008 arp_key = nla_data(nla);
1009 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1010 arp_key->arp_sip = output->ipv4.addr.src;
1011 arp_key->arp_tip = output->ipv4.addr.dst;
1012 arp_key->arp_op = htons(output->ip.proto);
1013 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1014 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1017 if ((swkey->eth.type == htons(ETH_P_IP) ||
1018 swkey->eth.type == htons(ETH_P_IPV6)) &&
1019 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1021 if (swkey->ip.proto == IPPROTO_TCP) {
1022 struct ovs_key_tcp *tcp_key;
1024 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1026 goto nla_put_failure;
1027 tcp_key = nla_data(nla);
1028 if (swkey->eth.type == htons(ETH_P_IP)) {
1029 tcp_key->tcp_src = output->ipv4.tp.src;
1030 tcp_key->tcp_dst = output->ipv4.tp.dst;
1031 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1032 output->ipv4.tp.flags))
1033 goto nla_put_failure;
1034 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1035 tcp_key->tcp_src = output->ipv6.tp.src;
1036 tcp_key->tcp_dst = output->ipv6.tp.dst;
1037 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1038 output->ipv6.tp.flags))
1039 goto nla_put_failure;
1041 } else if (swkey->ip.proto == IPPROTO_UDP) {
1042 struct ovs_key_udp *udp_key;
1044 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1046 goto nla_put_failure;
1047 udp_key = nla_data(nla);
1048 if (swkey->eth.type == htons(ETH_P_IP)) {
1049 udp_key->udp_src = output->ipv4.tp.src;
1050 udp_key->udp_dst = output->ipv4.tp.dst;
1051 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1052 udp_key->udp_src = output->ipv6.tp.src;
1053 udp_key->udp_dst = output->ipv6.tp.dst;
1055 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1056 struct ovs_key_sctp *sctp_key;
1058 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1060 goto nla_put_failure;
1061 sctp_key = nla_data(nla);
1062 if (swkey->eth.type == htons(ETH_P_IP)) {
1063 sctp_key->sctp_src = swkey->ipv4.tp.src;
1064 sctp_key->sctp_dst = swkey->ipv4.tp.dst;
1065 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1066 sctp_key->sctp_src = swkey->ipv6.tp.src;
1067 sctp_key->sctp_dst = swkey->ipv6.tp.dst;
1069 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1070 swkey->ip.proto == IPPROTO_ICMP) {
1071 struct ovs_key_icmp *icmp_key;
1073 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1075 goto nla_put_failure;
1076 icmp_key = nla_data(nla);
1077 icmp_key->icmp_type = ntohs(output->ipv4.tp.src);
1078 icmp_key->icmp_code = ntohs(output->ipv4.tp.dst);
1079 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1080 swkey->ip.proto == IPPROTO_ICMPV6) {
1081 struct ovs_key_icmpv6 *icmpv6_key;
1083 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1084 sizeof(*icmpv6_key));
1086 goto nla_put_failure;
1087 icmpv6_key = nla_data(nla);
1088 icmpv6_key->icmpv6_type = ntohs(output->ipv6.tp.src);
1089 icmpv6_key->icmpv6_code = ntohs(output->ipv6.tp.dst);
1091 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1092 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1093 struct ovs_key_nd *nd_key;
1095 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1097 goto nla_put_failure;
1098 nd_key = nla_data(nla);
1099 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1100 sizeof(nd_key->nd_target));
1101 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1102 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1109 nla_nest_end(skb, encap);
1117 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1119 struct sw_flow_actions *ovs_nla_alloc_flow_actions(int size)
1121 struct sw_flow_actions *sfa;
1123 if (size > MAX_ACTIONS_BUFSIZE)
1124 return ERR_PTR(-EINVAL);
1126 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1128 return ERR_PTR(-ENOMEM);
1130 sfa->actions_len = 0;
1134 /* RCU callback used by ovs_nla_free_flow_actions. */
1135 static void rcu_free_acts_callback(struct rcu_head *rcu)
1137 struct sw_flow_actions *sf_acts = container_of(rcu,
1138 struct sw_flow_actions, rcu);
1142 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1143 * The caller must hold rcu_read_lock for this to be sensible. */
1144 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1146 call_rcu(&sf_acts->rcu, rcu_free_acts_callback);
1149 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1153 struct sw_flow_actions *acts;
1155 int req_size = NLA_ALIGN(attr_len);
1156 int next_offset = offsetof(struct sw_flow_actions, actions) +
1157 (*sfa)->actions_len;
1159 if (req_size <= (ksize(*sfa) - next_offset))
1162 new_acts_size = ksize(*sfa) * 2;
1164 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1165 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1166 return ERR_PTR(-EMSGSIZE);
1167 new_acts_size = MAX_ACTIONS_BUFSIZE;
1170 acts = ovs_nla_alloc_flow_actions(new_acts_size);
1172 return (void *)acts;
1174 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1175 acts->actions_len = (*sfa)->actions_len;
1180 (*sfa)->actions_len += req_size;
1181 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1184 static int add_action(struct sw_flow_actions **sfa, int attrtype, void *data, int len)
1188 a = reserve_sfa_size(sfa, nla_attr_size(len));
1192 a->nla_type = attrtype;
1193 a->nla_len = nla_attr_size(len);
1196 memcpy(nla_data(a), data, len);
1197 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1202 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1205 int used = (*sfa)->actions_len;
1208 err = add_action(sfa, attrtype, NULL, 0);
1215 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1218 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1221 a->nla_len = sfa->actions_len - st_offset;
1224 static int validate_and_copy_sample(const struct nlattr *attr,
1225 const struct sw_flow_key *key, int depth,
1226 struct sw_flow_actions **sfa)
1228 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1229 const struct nlattr *probability, *actions;
1230 const struct nlattr *a;
1231 int rem, start, err, st_acts;
1233 memset(attrs, 0, sizeof(attrs));
1234 nla_for_each_nested(a, attr, rem) {
1235 int type = nla_type(a);
1236 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1243 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1244 if (!probability || nla_len(probability) != sizeof(u32))
1247 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1248 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1251 /* validation done, copy sample action. */
1252 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE);
1255 err = add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1256 nla_data(probability), sizeof(u32));
1259 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS);
1263 err = ovs_nla_copy_actions(actions, key, depth + 1, sfa);
1267 add_nested_action_end(*sfa, st_acts);
1268 add_nested_action_end(*sfa, start);
1273 static int validate_tp_port(const struct sw_flow_key *flow_key)
1275 if (flow_key->eth.type == htons(ETH_P_IP)) {
1276 if (flow_key->ipv4.tp.src || flow_key->ipv4.tp.dst)
1278 } else if (flow_key->eth.type == htons(ETH_P_IPV6)) {
1279 if (flow_key->ipv6.tp.src || flow_key->ipv6.tp.dst)
1286 void ovs_match_init(struct sw_flow_match *match,
1287 struct sw_flow_key *key,
1288 struct sw_flow_mask *mask)
1290 memset(match, 0, sizeof(*match));
1294 memset(key, 0, sizeof(*key));
1297 memset(&mask->key, 0, sizeof(mask->key));
1298 mask->range.start = mask->range.end = 0;
1302 static int validate_and_copy_set_tun(const struct nlattr *attr,
1303 struct sw_flow_actions **sfa)
1305 struct sw_flow_match match;
1306 struct sw_flow_key key;
1309 ovs_match_init(&match, &key, NULL);
1310 err = ipv4_tun_from_nlattr(nla_data(attr), &match, false);
1314 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET);
1318 err = add_action(sfa, OVS_KEY_ATTR_IPV4_TUNNEL, &match.key->tun_key,
1319 sizeof(match.key->tun_key));
1320 add_nested_action_end(*sfa, start);
1325 static int validate_set(const struct nlattr *a,
1326 const struct sw_flow_key *flow_key,
1327 struct sw_flow_actions **sfa,
1330 const struct nlattr *ovs_key = nla_data(a);
1331 int key_type = nla_type(ovs_key);
1333 /* There can be only one key in a action */
1334 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1337 if (key_type > OVS_KEY_ATTR_MAX ||
1338 (ovs_key_lens[key_type] != nla_len(ovs_key) &&
1339 ovs_key_lens[key_type] != -1))
1343 const struct ovs_key_ipv4 *ipv4_key;
1344 const struct ovs_key_ipv6 *ipv6_key;
1347 case OVS_KEY_ATTR_PRIORITY:
1348 case OVS_KEY_ATTR_SKB_MARK:
1349 case OVS_KEY_ATTR_ETHERNET:
1352 case OVS_KEY_ATTR_TUNNEL:
1354 err = validate_and_copy_set_tun(a, sfa);
1359 case OVS_KEY_ATTR_IPV4:
1360 if (flow_key->eth.type != htons(ETH_P_IP))
1363 if (!flow_key->ip.proto)
1366 ipv4_key = nla_data(ovs_key);
1367 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
1370 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
1375 case OVS_KEY_ATTR_IPV6:
1376 if (flow_key->eth.type != htons(ETH_P_IPV6))
1379 if (!flow_key->ip.proto)
1382 ipv6_key = nla_data(ovs_key);
1383 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
1386 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
1389 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
1394 case OVS_KEY_ATTR_TCP:
1395 if (flow_key->ip.proto != IPPROTO_TCP)
1398 return validate_tp_port(flow_key);
1400 case OVS_KEY_ATTR_UDP:
1401 if (flow_key->ip.proto != IPPROTO_UDP)
1404 return validate_tp_port(flow_key);
1406 case OVS_KEY_ATTR_SCTP:
1407 if (flow_key->ip.proto != IPPROTO_SCTP)
1410 return validate_tp_port(flow_key);
1419 static int validate_userspace(const struct nlattr *attr)
1421 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
1422 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
1423 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
1425 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
1428 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
1429 attr, userspace_policy);
1433 if (!a[OVS_USERSPACE_ATTR_PID] ||
1434 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
1440 static int copy_action(const struct nlattr *from,
1441 struct sw_flow_actions **sfa)
1443 int totlen = NLA_ALIGN(from->nla_len);
1446 to = reserve_sfa_size(sfa, from->nla_len);
1450 memcpy(to, from, totlen);
1454 int ovs_nla_copy_actions(const struct nlattr *attr,
1455 const struct sw_flow_key *key,
1457 struct sw_flow_actions **sfa)
1459 const struct nlattr *a;
1462 if (depth >= SAMPLE_ACTION_DEPTH)
1465 nla_for_each_nested(a, attr, rem) {
1466 /* Expected argument lengths, (u32)-1 for variable length. */
1467 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
1468 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
1469 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
1470 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
1471 [OVS_ACTION_ATTR_POP_VLAN] = 0,
1472 [OVS_ACTION_ATTR_SET] = (u32)-1,
1473 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1
1475 const struct ovs_action_push_vlan *vlan;
1476 int type = nla_type(a);
1479 if (type > OVS_ACTION_ATTR_MAX ||
1480 (action_lens[type] != nla_len(a) &&
1481 action_lens[type] != (u32)-1))
1486 case OVS_ACTION_ATTR_UNSPEC:
1489 case OVS_ACTION_ATTR_USERSPACE:
1490 err = validate_userspace(a);
1495 case OVS_ACTION_ATTR_OUTPUT:
1496 if (nla_get_u32(a) >= DP_MAX_PORTS)
1501 case OVS_ACTION_ATTR_POP_VLAN:
1504 case OVS_ACTION_ATTR_PUSH_VLAN:
1506 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
1508 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
1512 case OVS_ACTION_ATTR_SET:
1513 err = validate_set(a, key, sfa, &skip_copy);
1518 case OVS_ACTION_ATTR_SAMPLE:
1519 err = validate_and_copy_sample(a, key, depth, sfa);
1529 err = copy_action(a, sfa);
1541 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
1543 const struct nlattr *a;
1544 struct nlattr *start;
1547 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
1551 nla_for_each_nested(a, attr, rem) {
1552 int type = nla_type(a);
1553 struct nlattr *st_sample;
1556 case OVS_SAMPLE_ATTR_PROBABILITY:
1557 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
1558 sizeof(u32), nla_data(a)))
1561 case OVS_SAMPLE_ATTR_ACTIONS:
1562 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
1565 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
1568 nla_nest_end(skb, st_sample);
1573 nla_nest_end(skb, start);
1577 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
1579 const struct nlattr *ovs_key = nla_data(a);
1580 int key_type = nla_type(ovs_key);
1581 struct nlattr *start;
1585 case OVS_KEY_ATTR_IPV4_TUNNEL:
1586 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
1590 err = ipv4_tun_to_nlattr(skb, nla_data(ovs_key),
1594 nla_nest_end(skb, start);
1597 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
1605 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
1607 const struct nlattr *a;
1610 nla_for_each_attr(a, attr, len, rem) {
1611 int type = nla_type(a);
1614 case OVS_ACTION_ATTR_SET:
1615 err = set_action_to_attr(a, skb);
1620 case OVS_ACTION_ATTR_SAMPLE:
1621 err = sample_action_to_attr(a, skb);
1626 if (nla_put(skb, type, nla_len(a), nla_data(a)))