2 * Copyright (c) 2007-2015 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
21 #include <linux/uaccess.h>
22 #include <linux/netdevice.h>
23 #include <linux/etherdevice.h>
24 #include <linux/if_ether.h>
25 #include <linux/if_vlan.h>
26 #include <net/llc_pdu.h>
27 #include <linux/kernel.h>
28 #include <linux/jhash.h>
29 #include <linux/jiffies.h>
30 #include <linux/llc.h>
31 #include <linux/module.h>
33 #include <linux/rcupdate.h>
34 #include <linux/if_arp.h>
36 #include <linux/ipv6.h>
37 #include <linux/sctp.h>
38 #include <linux/tcp.h>
39 #include <linux/udp.h>
40 #include <linux/icmp.h>
41 #include <linux/icmpv6.h>
42 #include <linux/rculist.h>
43 #include <net/geneve.h>
46 #include <net/ndisc.h>
48 #include <net/vxlan.h>
51 #include "conntrack.h"
53 #include "flow_netlink.h"
58 const struct ovs_len_tbl *next;
61 #define OVS_ATTR_NESTED -1
62 #define OVS_ATTR_VARIABLE -2
64 static void update_range(struct sw_flow_match *match,
65 size_t offset, size_t size, bool is_mask)
67 struct sw_flow_key_range *range;
68 size_t start = rounddown(offset, sizeof(long));
69 size_t end = roundup(offset + size, sizeof(long));
72 range = &match->range;
74 range = &match->mask->range;
76 if (range->start == range->end) {
82 if (range->start > start)
89 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
91 update_range(match, offsetof(struct sw_flow_key, field), \
92 sizeof((match)->key->field), is_mask); \
94 (match)->mask->key.field = value; \
96 (match)->key->field = value; \
99 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
101 update_range(match, offset, len, is_mask); \
103 memcpy((u8 *)&(match)->mask->key + offset, value_p, len);\
105 memcpy((u8 *)(match)->key + offset, value_p, len); \
108 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
109 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
110 value_p, len, is_mask)
112 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
114 update_range(match, offsetof(struct sw_flow_key, field), \
115 sizeof((match)->key->field), is_mask); \
117 memset((u8 *)&(match)->mask->key.field, value, \
118 sizeof((match)->mask->key.field)); \
120 memset((u8 *)&(match)->key->field, value, \
121 sizeof((match)->key->field)); \
124 static bool match_validate(const struct sw_flow_match *match,
125 u64 key_attrs, u64 mask_attrs, bool log)
127 u64 key_expected = 1ULL << OVS_KEY_ATTR_ETHERNET;
128 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
130 /* The following mask attributes allowed only if they
131 * pass the validation tests.
133 mask_allowed &= ~((1ULL << OVS_KEY_ATTR_IPV4)
134 | (1ULL << OVS_KEY_ATTR_IPV6)
135 | (1ULL << OVS_KEY_ATTR_TCP)
136 | (1ULL << OVS_KEY_ATTR_TCP_FLAGS)
137 | (1ULL << OVS_KEY_ATTR_UDP)
138 | (1ULL << OVS_KEY_ATTR_SCTP)
139 | (1ULL << OVS_KEY_ATTR_ICMP)
140 | (1ULL << OVS_KEY_ATTR_ICMPV6)
141 | (1ULL << OVS_KEY_ATTR_ARP)
142 | (1ULL << OVS_KEY_ATTR_ND)
143 | (1ULL << OVS_KEY_ATTR_MPLS));
145 /* Always allowed mask fields. */
146 mask_allowed |= ((1ULL << OVS_KEY_ATTR_TUNNEL)
147 | (1ULL << OVS_KEY_ATTR_IN_PORT)
148 | (1ULL << OVS_KEY_ATTR_ETHERTYPE));
150 /* Check key attributes. */
151 if (match->key->eth.type == htons(ETH_P_ARP)
152 || match->key->eth.type == htons(ETH_P_RARP)) {
153 key_expected |= 1ULL << OVS_KEY_ATTR_ARP;
154 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
155 mask_allowed |= 1ULL << OVS_KEY_ATTR_ARP;
158 if (eth_p_mpls(match->key->eth.type)) {
159 key_expected |= 1ULL << OVS_KEY_ATTR_MPLS;
160 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
161 mask_allowed |= 1ULL << OVS_KEY_ATTR_MPLS;
164 if (match->key->eth.type == htons(ETH_P_IP)) {
165 key_expected |= 1ULL << OVS_KEY_ATTR_IPV4;
166 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
167 mask_allowed |= 1ULL << OVS_KEY_ATTR_IPV4;
169 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
170 if (match->key->ip.proto == IPPROTO_UDP) {
171 key_expected |= 1ULL << OVS_KEY_ATTR_UDP;
172 if (match->mask && (match->mask->key.ip.proto == 0xff))
173 mask_allowed |= 1ULL << OVS_KEY_ATTR_UDP;
176 if (match->key->ip.proto == IPPROTO_SCTP) {
177 key_expected |= 1ULL << OVS_KEY_ATTR_SCTP;
178 if (match->mask && (match->mask->key.ip.proto == 0xff))
179 mask_allowed |= 1ULL << OVS_KEY_ATTR_SCTP;
182 if (match->key->ip.proto == IPPROTO_TCP) {
183 key_expected |= 1ULL << OVS_KEY_ATTR_TCP;
184 key_expected |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
185 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
186 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP;
187 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
191 if (match->key->ip.proto == IPPROTO_ICMP) {
192 key_expected |= 1ULL << OVS_KEY_ATTR_ICMP;
193 if (match->mask && (match->mask->key.ip.proto == 0xff))
194 mask_allowed |= 1ULL << OVS_KEY_ATTR_ICMP;
199 if (match->key->eth.type == htons(ETH_P_IPV6)) {
200 key_expected |= 1ULL << OVS_KEY_ATTR_IPV6;
201 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
202 mask_allowed |= 1ULL << OVS_KEY_ATTR_IPV6;
204 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
205 if (match->key->ip.proto == IPPROTO_UDP) {
206 key_expected |= 1ULL << OVS_KEY_ATTR_UDP;
207 if (match->mask && (match->mask->key.ip.proto == 0xff))
208 mask_allowed |= 1ULL << OVS_KEY_ATTR_UDP;
211 if (match->key->ip.proto == IPPROTO_SCTP) {
212 key_expected |= 1ULL << OVS_KEY_ATTR_SCTP;
213 if (match->mask && (match->mask->key.ip.proto == 0xff))
214 mask_allowed |= 1ULL << OVS_KEY_ATTR_SCTP;
217 if (match->key->ip.proto == IPPROTO_TCP) {
218 key_expected |= 1ULL << OVS_KEY_ATTR_TCP;
219 key_expected |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
220 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
221 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP;
222 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
226 if (match->key->ip.proto == IPPROTO_ICMPV6) {
227 key_expected |= 1ULL << OVS_KEY_ATTR_ICMPV6;
228 if (match->mask && (match->mask->key.ip.proto == 0xff))
229 mask_allowed |= 1ULL << OVS_KEY_ATTR_ICMPV6;
231 if (match->key->tp.src ==
232 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
233 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
234 key_expected |= 1ULL << OVS_KEY_ATTR_ND;
235 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
236 mask_allowed |= 1ULL << OVS_KEY_ATTR_ND;
242 if ((key_attrs & key_expected) != key_expected) {
243 /* Key attributes check failed. */
244 OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
245 (unsigned long long)key_attrs,
246 (unsigned long long)key_expected);
250 if ((mask_attrs & mask_allowed) != mask_attrs) {
251 /* Mask attributes check failed. */
252 OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
253 (unsigned long long)mask_attrs,
254 (unsigned long long)mask_allowed);
261 size_t ovs_tun_key_attr_size(void)
263 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
264 * updating this function.
266 return nla_total_size(8) /* OVS_TUNNEL_KEY_ATTR_ID */
267 + nla_total_size(4) /* OVS_TUNNEL_KEY_ATTR_IPV4_SRC */
268 + nla_total_size(4) /* OVS_TUNNEL_KEY_ATTR_IPV4_DST */
269 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
270 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
271 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
272 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
273 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
274 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
275 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS is mutually exclusive with
276 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
278 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
279 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
282 size_t ovs_key_attr_size(void)
284 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
285 * updating this function.
287 BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO != 26);
289 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
290 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
291 + ovs_tun_key_attr_size()
292 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
293 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
294 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
295 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
296 + nla_total_size(1) /* OVS_KEY_ATTR_CT_STATE */
297 + nla_total_size(2) /* OVS_KEY_ATTR_CT_ZONE */
298 + nla_total_size(4) /* OVS_KEY_ATTR_CT_MARK */
299 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
300 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
301 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
302 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
303 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
304 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
305 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
306 + nla_total_size(28); /* OVS_KEY_ATTR_ND */
309 static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
310 [OVS_VXLAN_EXT_GBP] = { .len = sizeof(u32) },
313 static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
314 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
315 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
316 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
317 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
318 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
319 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
320 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
321 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
322 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
323 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
324 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_VARIABLE },
325 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED,
326 .next = ovs_vxlan_ext_key_lens },
329 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
330 static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
331 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
332 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
333 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
334 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
335 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
336 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
337 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
338 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
339 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
340 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
341 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
342 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
343 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
344 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
345 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
346 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
347 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
348 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
349 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
350 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
351 .next = ovs_tunnel_key_lens, },
352 [OVS_KEY_ATTR_MPLS] = { .len = sizeof(struct ovs_key_mpls) },
353 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u8) },
354 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
355 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
358 static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
360 return expected_len == attr_len ||
361 expected_len == OVS_ATTR_NESTED ||
362 expected_len == OVS_ATTR_VARIABLE;
365 static bool is_all_zero(const u8 *fp, size_t size)
372 for (i = 0; i < size; i++)
379 static int __parse_flow_nlattrs(const struct nlattr *attr,
380 const struct nlattr *a[],
381 u64 *attrsp, bool log, bool nz)
383 const struct nlattr *nla;
388 nla_for_each_nested(nla, attr, rem) {
389 u16 type = nla_type(nla);
392 if (type > OVS_KEY_ATTR_MAX) {
393 OVS_NLERR(log, "Key type %d is out of range max %d",
394 type, OVS_KEY_ATTR_MAX);
398 if (attrs & (1ULL << type)) {
399 OVS_NLERR(log, "Duplicate key (type %d).", type);
403 expected_len = ovs_key_lens[type].len;
404 if (!check_attr_len(nla_len(nla), expected_len)) {
405 OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
406 type, nla_len(nla), expected_len);
410 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
411 attrs |= 1ULL << type;
416 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
424 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
425 const struct nlattr *a[], u64 *attrsp,
428 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
431 static int parse_flow_nlattrs(const struct nlattr *attr,
432 const struct nlattr *a[], u64 *attrsp,
435 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
438 static int genev_tun_opt_from_nlattr(const struct nlattr *a,
439 struct sw_flow_match *match, bool is_mask,
442 unsigned long opt_key_offset;
444 if (nla_len(a) > sizeof(match->key->tun_opts)) {
445 OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
446 nla_len(a), sizeof(match->key->tun_opts));
450 if (nla_len(a) % 4 != 0) {
451 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
456 /* We need to record the length of the options passed
457 * down, otherwise packets with the same format but
458 * additional options will be silently matched.
461 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
464 /* This is somewhat unusual because it looks at
465 * both the key and mask while parsing the
466 * attributes (and by extension assumes the key
467 * is parsed first). Normally, we would verify
468 * that each is the correct length and that the
469 * attributes line up in the validate function.
470 * However, that is difficult because this is
471 * variable length and we won't have the
474 if (match->key->tun_opts_len != nla_len(a)) {
475 OVS_NLERR(log, "Geneve option len %d != mask len %d",
476 match->key->tun_opts_len, nla_len(a));
480 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
483 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
484 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
485 nla_len(a), is_mask);
489 static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
490 struct sw_flow_match *match, bool is_mask,
495 unsigned long opt_key_offset;
496 struct vxlan_metadata opts;
498 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
500 memset(&opts, 0, sizeof(opts));
501 nla_for_each_nested(a, attr, rem) {
502 int type = nla_type(a);
504 if (type > OVS_VXLAN_EXT_MAX) {
505 OVS_NLERR(log, "VXLAN extension %d out of range max %d",
506 type, OVS_VXLAN_EXT_MAX);
510 if (!check_attr_len(nla_len(a),
511 ovs_vxlan_ext_key_lens[type].len)) {
512 OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
514 ovs_vxlan_ext_key_lens[type].len);
519 case OVS_VXLAN_EXT_GBP:
520 opts.gbp = nla_get_u32(a);
523 OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
529 OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
535 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
537 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
539 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
540 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
545 static int ipv4_tun_from_nlattr(const struct nlattr *attr,
546 struct sw_flow_match *match, bool is_mask,
552 __be16 tun_flags = 0;
555 nla_for_each_nested(a, attr, rem) {
556 int type = nla_type(a);
559 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
560 OVS_NLERR(log, "Tunnel attr %d out of range max %d",
561 type, OVS_TUNNEL_KEY_ATTR_MAX);
565 if (!check_attr_len(nla_len(a),
566 ovs_tunnel_key_lens[type].len)) {
567 OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
568 type, nla_len(a), ovs_tunnel_key_lens[type].len);
573 case OVS_TUNNEL_KEY_ATTR_ID:
574 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
575 nla_get_be64(a), is_mask);
576 tun_flags |= TUNNEL_KEY;
578 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
579 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
580 nla_get_in_addr(a), is_mask);
582 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
583 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
584 nla_get_in_addr(a), is_mask);
586 case OVS_TUNNEL_KEY_ATTR_TOS:
587 SW_FLOW_KEY_PUT(match, tun_key.tos,
588 nla_get_u8(a), is_mask);
590 case OVS_TUNNEL_KEY_ATTR_TTL:
591 SW_FLOW_KEY_PUT(match, tun_key.ttl,
592 nla_get_u8(a), is_mask);
595 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
596 tun_flags |= TUNNEL_DONT_FRAGMENT;
598 case OVS_TUNNEL_KEY_ATTR_CSUM:
599 tun_flags |= TUNNEL_CSUM;
601 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
602 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
603 nla_get_be16(a), is_mask);
605 case OVS_TUNNEL_KEY_ATTR_TP_DST:
606 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
607 nla_get_be16(a), is_mask);
609 case OVS_TUNNEL_KEY_ATTR_OAM:
610 tun_flags |= TUNNEL_OAM;
612 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
614 OVS_NLERR(log, "Multiple metadata blocks provided");
618 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
622 tun_flags |= TUNNEL_GENEVE_OPT;
625 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
627 OVS_NLERR(log, "Multiple metadata blocks provided");
631 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
635 tun_flags |= TUNNEL_VXLAN_OPT;
639 OVS_NLERR(log, "Unknown IPv4 tunnel attribute %d",
645 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
648 OVS_NLERR(log, "IPv4 tunnel attribute has %d unknown bytes.",
654 if (!match->key->tun_key.u.ipv4.dst) {
655 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
660 OVS_NLERR(log, "IPv4 tunnel TTL not specified.");
668 static int vxlan_opt_to_nlattr(struct sk_buff *skb,
669 const void *tun_opts, int swkey_tun_opts_len)
671 const struct vxlan_metadata *opts = tun_opts;
674 nla = nla_nest_start(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
678 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
681 nla_nest_end(skb, nla);
685 static int __ipv4_tun_to_nlattr(struct sk_buff *skb,
686 const struct ip_tunnel_key *output,
687 const void *tun_opts, int swkey_tun_opts_len)
689 if (output->tun_flags & TUNNEL_KEY &&
690 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
692 if (output->u.ipv4.src &&
693 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
696 if (output->u.ipv4.dst &&
697 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
701 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
703 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
705 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
706 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
708 if ((output->tun_flags & TUNNEL_CSUM) &&
709 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
711 if (output->tp_src &&
712 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
714 if (output->tp_dst &&
715 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
717 if ((output->tun_flags & TUNNEL_OAM) &&
718 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
721 if (output->tun_flags & TUNNEL_GENEVE_OPT &&
722 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
723 swkey_tun_opts_len, tun_opts))
725 else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
726 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
733 static int ipv4_tun_to_nlattr(struct sk_buff *skb,
734 const struct ip_tunnel_key *output,
735 const void *tun_opts, int swkey_tun_opts_len)
740 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
744 err = __ipv4_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len);
748 nla_nest_end(skb, nla);
752 int ovs_nla_put_egress_tunnel_key(struct sk_buff *skb,
753 const struct ip_tunnel_info *egress_tun_info,
754 const void *egress_tun_opts)
756 return __ipv4_tun_to_nlattr(skb, &egress_tun_info->key,
758 egress_tun_info->options_len);
761 static int metadata_from_nlattrs(struct sw_flow_match *match, u64 *attrs,
762 const struct nlattr **a, bool is_mask,
765 if (*attrs & (1ULL << OVS_KEY_ATTR_DP_HASH)) {
766 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
768 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
769 *attrs &= ~(1ULL << OVS_KEY_ATTR_DP_HASH);
772 if (*attrs & (1ULL << OVS_KEY_ATTR_RECIRC_ID)) {
773 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
775 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
776 *attrs &= ~(1ULL << OVS_KEY_ATTR_RECIRC_ID);
779 if (*attrs & (1ULL << OVS_KEY_ATTR_PRIORITY)) {
780 SW_FLOW_KEY_PUT(match, phy.priority,
781 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
782 *attrs &= ~(1ULL << OVS_KEY_ATTR_PRIORITY);
785 if (*attrs & (1ULL << OVS_KEY_ATTR_IN_PORT)) {
786 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
789 in_port = 0xffffffff; /* Always exact match in_port. */
790 } else if (in_port >= DP_MAX_PORTS) {
791 OVS_NLERR(log, "Port %d exceeds max allowable %d",
792 in_port, DP_MAX_PORTS);
796 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
797 *attrs &= ~(1ULL << OVS_KEY_ATTR_IN_PORT);
798 } else if (!is_mask) {
799 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
802 if (*attrs & (1ULL << OVS_KEY_ATTR_SKB_MARK)) {
803 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
805 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
806 *attrs &= ~(1ULL << OVS_KEY_ATTR_SKB_MARK);
808 if (*attrs & (1ULL << OVS_KEY_ATTR_TUNNEL)) {
809 if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
812 *attrs &= ~(1ULL << OVS_KEY_ATTR_TUNNEL);
815 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
816 ovs_ct_verify(OVS_KEY_ATTR_CT_STATE)) {
817 u8 ct_state = nla_get_u8(a[OVS_KEY_ATTR_CT_STATE]);
819 SW_FLOW_KEY_PUT(match, ct.state, ct_state, is_mask);
820 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
822 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
823 ovs_ct_verify(OVS_KEY_ATTR_CT_ZONE)) {
824 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
826 SW_FLOW_KEY_PUT(match, ct.zone, ct_zone, is_mask);
827 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
829 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
830 ovs_ct_verify(OVS_KEY_ATTR_CT_MARK)) {
831 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
833 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
834 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
839 static int ovs_key_from_nlattrs(struct sw_flow_match *match, u64 attrs,
840 const struct nlattr **a, bool is_mask,
845 err = metadata_from_nlattrs(match, &attrs, a, is_mask, log);
849 if (attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) {
850 const struct ovs_key_ethernet *eth_key;
852 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
853 SW_FLOW_KEY_MEMCPY(match, eth.src,
854 eth_key->eth_src, ETH_ALEN, is_mask);
855 SW_FLOW_KEY_MEMCPY(match, eth.dst,
856 eth_key->eth_dst, ETH_ALEN, is_mask);
857 attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERNET);
860 if (attrs & (1ULL << OVS_KEY_ATTR_VLAN)) {
863 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
864 if (!(tci & htons(VLAN_TAG_PRESENT))) {
866 OVS_NLERR(log, "VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.");
868 OVS_NLERR(log, "VLAN TCI does not have VLAN_TAG_PRESENT bit set.");
873 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
874 attrs &= ~(1ULL << OVS_KEY_ATTR_VLAN);
877 if (attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) {
880 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
882 /* Always exact match EtherType. */
883 eth_type = htons(0xffff);
884 } else if (!eth_proto_is_802_3(eth_type)) {
885 OVS_NLERR(log, "EtherType %x is less than min %x",
886 ntohs(eth_type), ETH_P_802_3_MIN);
890 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
891 attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
892 } else if (!is_mask) {
893 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
896 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
897 const struct ovs_key_ipv4 *ipv4_key;
899 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
900 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
901 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
902 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
905 SW_FLOW_KEY_PUT(match, ip.proto,
906 ipv4_key->ipv4_proto, is_mask);
907 SW_FLOW_KEY_PUT(match, ip.tos,
908 ipv4_key->ipv4_tos, is_mask);
909 SW_FLOW_KEY_PUT(match, ip.ttl,
910 ipv4_key->ipv4_ttl, is_mask);
911 SW_FLOW_KEY_PUT(match, ip.frag,
912 ipv4_key->ipv4_frag, is_mask);
913 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
914 ipv4_key->ipv4_src, is_mask);
915 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
916 ipv4_key->ipv4_dst, is_mask);
917 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
920 if (attrs & (1ULL << OVS_KEY_ATTR_IPV6)) {
921 const struct ovs_key_ipv6 *ipv6_key;
923 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
924 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
925 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
926 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
930 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
932 "Invalid IPv6 flow label value (value=%x, max=%x).",
933 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
937 SW_FLOW_KEY_PUT(match, ipv6.label,
938 ipv6_key->ipv6_label, is_mask);
939 SW_FLOW_KEY_PUT(match, ip.proto,
940 ipv6_key->ipv6_proto, is_mask);
941 SW_FLOW_KEY_PUT(match, ip.tos,
942 ipv6_key->ipv6_tclass, is_mask);
943 SW_FLOW_KEY_PUT(match, ip.ttl,
944 ipv6_key->ipv6_hlimit, is_mask);
945 SW_FLOW_KEY_PUT(match, ip.frag,
946 ipv6_key->ipv6_frag, is_mask);
947 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
949 sizeof(match->key->ipv6.addr.src),
951 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
953 sizeof(match->key->ipv6.addr.dst),
956 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV6);
959 if (attrs & (1ULL << OVS_KEY_ATTR_ARP)) {
960 const struct ovs_key_arp *arp_key;
962 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
963 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
964 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
969 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
970 arp_key->arp_sip, is_mask);
971 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
972 arp_key->arp_tip, is_mask);
973 SW_FLOW_KEY_PUT(match, ip.proto,
974 ntohs(arp_key->arp_op), is_mask);
975 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
976 arp_key->arp_sha, ETH_ALEN, is_mask);
977 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
978 arp_key->arp_tha, ETH_ALEN, is_mask);
980 attrs &= ~(1ULL << OVS_KEY_ATTR_ARP);
983 if (attrs & (1ULL << OVS_KEY_ATTR_MPLS)) {
984 const struct ovs_key_mpls *mpls_key;
986 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
987 SW_FLOW_KEY_PUT(match, mpls.top_lse,
988 mpls_key->mpls_lse, is_mask);
990 attrs &= ~(1ULL << OVS_KEY_ATTR_MPLS);
993 if (attrs & (1ULL << OVS_KEY_ATTR_TCP)) {
994 const struct ovs_key_tcp *tcp_key;
996 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
997 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
998 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
999 attrs &= ~(1ULL << OVS_KEY_ATTR_TCP);
1002 if (attrs & (1ULL << OVS_KEY_ATTR_TCP_FLAGS)) {
1003 SW_FLOW_KEY_PUT(match, tp.flags,
1004 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1006 attrs &= ~(1ULL << OVS_KEY_ATTR_TCP_FLAGS);
1009 if (attrs & (1ULL << OVS_KEY_ATTR_UDP)) {
1010 const struct ovs_key_udp *udp_key;
1012 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1013 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1014 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1015 attrs &= ~(1ULL << OVS_KEY_ATTR_UDP);
1018 if (attrs & (1ULL << OVS_KEY_ATTR_SCTP)) {
1019 const struct ovs_key_sctp *sctp_key;
1021 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1022 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1023 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1024 attrs &= ~(1ULL << OVS_KEY_ATTR_SCTP);
1027 if (attrs & (1ULL << OVS_KEY_ATTR_ICMP)) {
1028 const struct ovs_key_icmp *icmp_key;
1030 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1031 SW_FLOW_KEY_PUT(match, tp.src,
1032 htons(icmp_key->icmp_type), is_mask);
1033 SW_FLOW_KEY_PUT(match, tp.dst,
1034 htons(icmp_key->icmp_code), is_mask);
1035 attrs &= ~(1ULL << OVS_KEY_ATTR_ICMP);
1038 if (attrs & (1ULL << OVS_KEY_ATTR_ICMPV6)) {
1039 const struct ovs_key_icmpv6 *icmpv6_key;
1041 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1042 SW_FLOW_KEY_PUT(match, tp.src,
1043 htons(icmpv6_key->icmpv6_type), is_mask);
1044 SW_FLOW_KEY_PUT(match, tp.dst,
1045 htons(icmpv6_key->icmpv6_code), is_mask);
1046 attrs &= ~(1ULL << OVS_KEY_ATTR_ICMPV6);
1049 if (attrs & (1ULL << OVS_KEY_ATTR_ND)) {
1050 const struct ovs_key_nd *nd_key;
1052 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1053 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1055 sizeof(match->key->ipv6.nd.target),
1057 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1058 nd_key->nd_sll, ETH_ALEN, is_mask);
1059 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1060 nd_key->nd_tll, ETH_ALEN, is_mask);
1061 attrs &= ~(1ULL << OVS_KEY_ATTR_ND);
1065 OVS_NLERR(log, "Unknown key attributes %llx",
1066 (unsigned long long)attrs);
1073 static void nlattr_set(struct nlattr *attr, u8 val,
1074 const struct ovs_len_tbl *tbl)
1079 /* The nlattr stream should already have been validated */
1080 nla_for_each_nested(nla, attr, rem) {
1081 if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED) {
1082 if (tbl[nla_type(nla)].next)
1083 tbl = tbl[nla_type(nla)].next;
1084 nlattr_set(nla, val, tbl);
1086 memset(nla_data(nla), val, nla_len(nla));
1091 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1093 nlattr_set(attr, val, ovs_key_lens);
1097 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1098 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1099 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1100 * does not include any don't care bit.
1101 * @match: receives the extracted flow match information.
1102 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1103 * sequence. The fields should of the packet that triggered the creation
1105 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1106 * attribute specifies the mask field of the wildcarded flow.
1107 * @log: Boolean to allow kernel error logging. Normally true, but when
1108 * probing for feature compatibility this should be passed in as false to
1109 * suppress unnecessary error logging.
1111 int ovs_nla_get_match(struct sw_flow_match *match,
1112 const struct nlattr *nla_key,
1113 const struct nlattr *nla_mask,
1116 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1117 const struct nlattr *encap;
1118 struct nlattr *newmask = NULL;
1121 bool encap_valid = false;
1124 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1128 if ((key_attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
1129 (key_attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) &&
1130 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
1133 if (!((key_attrs & (1ULL << OVS_KEY_ATTR_VLAN)) &&
1134 (key_attrs & (1ULL << OVS_KEY_ATTR_ENCAP)))) {
1135 OVS_NLERR(log, "Invalid Vlan frame.");
1139 key_attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
1140 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1141 encap = a[OVS_KEY_ATTR_ENCAP];
1142 key_attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
1145 if (tci & htons(VLAN_TAG_PRESENT)) {
1146 err = parse_flow_nlattrs(encap, a, &key_attrs, log);
1150 /* Corner case for truncated 802.1Q header. */
1151 if (nla_len(encap)) {
1152 OVS_NLERR(log, "Truncated 802.1Q header has non-zero encap attribute.");
1156 OVS_NLERR(log, "Encap attr is set for non-VLAN frame");
1161 err = ovs_key_from_nlattrs(match, key_attrs, a, false, log);
1167 /* Create an exact match mask. We need to set to 0xff
1168 * all the 'match->mask' fields that have been touched
1169 * in 'match->key'. We cannot simply memset
1170 * 'match->mask', because padding bytes and fields not
1171 * specified in 'match->key' should be left to 0.
1172 * Instead, we use a stream of netlink attributes,
1173 * copied from 'key' and set to 0xff.
1174 * ovs_key_from_nlattrs() will take care of filling
1175 * 'match->mask' appropriately.
1177 newmask = kmemdup(nla_key,
1178 nla_total_size(nla_len(nla_key)),
1183 mask_set_nlattr(newmask, 0xff);
1185 /* The userspace does not send tunnel attributes that
1186 * are 0, but we should not wildcard them nonetheless.
1188 if (match->key->tun_key.u.ipv4.dst)
1189 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1195 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1199 /* Always match on tci. */
1200 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
1202 if (mask_attrs & 1ULL << OVS_KEY_ATTR_ENCAP) {
1203 __be16 eth_type = 0;
1207 OVS_NLERR(log, "Encap mask attribute is set for non-VLAN frame.");
1212 mask_attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
1213 if (a[OVS_KEY_ATTR_ETHERTYPE])
1214 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1216 if (eth_type == htons(0xffff)) {
1217 mask_attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
1218 encap = a[OVS_KEY_ATTR_ENCAP];
1219 err = parse_flow_mask_nlattrs(encap, a,
1224 OVS_NLERR(log, "VLAN frames must have an exact match on the TPID (mask=%x).",
1230 if (a[OVS_KEY_ATTR_VLAN])
1231 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1233 if (!(tci & htons(VLAN_TAG_PRESENT))) {
1234 OVS_NLERR(log, "VLAN tag present bit must have an exact match (tci_mask=%x).",
1241 err = ovs_key_from_nlattrs(match, mask_attrs, a, true, log);
1246 if (!match_validate(match, key_attrs, mask_attrs, log))
1254 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1261 len = nla_len(attr);
1262 if (len < 1 || len > MAX_UFID_LENGTH) {
1263 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1264 nla_len(attr), MAX_UFID_LENGTH);
1271 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1272 * or false otherwise.
1274 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1277 sfid->ufid_len = get_ufid_len(attr, log);
1279 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1281 return sfid->ufid_len;
1284 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1285 const struct sw_flow_key *key, bool log)
1287 struct sw_flow_key *new_key;
1289 if (ovs_nla_get_ufid(sfid, ufid, log))
1292 /* If UFID was not provided, use unmasked key. */
1293 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1296 memcpy(new_key, key, sizeof(*key));
1297 sfid->unmasked_key = new_key;
1302 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1304 return attr ? nla_get_u32(attr) : 0;
1308 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1309 * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1310 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1312 * @log: Boolean to allow kernel error logging. Normally true, but when
1313 * probing for feature compatibility this should be passed in as false to
1314 * suppress unnecessary error logging.
1316 * This parses a series of Netlink attributes that form a flow key, which must
1317 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1318 * get the metadata, that is, the parts of the flow key that cannot be
1319 * extracted from the packet itself.
1322 int ovs_nla_get_flow_metadata(const struct nlattr *attr,
1323 struct sw_flow_key *key,
1326 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1327 struct sw_flow_match match;
1331 err = parse_flow_nlattrs(attr, a, &attrs, log);
1335 memset(&match, 0, sizeof(match));
1338 memset(key, 0, OVS_SW_FLOW_KEY_METADATA_SIZE);
1339 memset(&key->ct, 0, sizeof(key->ct));
1340 key->phy.in_port = DP_MAX_PORTS;
1342 return metadata_from_nlattrs(&match, &attrs, a, false, log);
1345 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1346 const struct sw_flow_key *output, bool is_mask,
1347 struct sk_buff *skb)
1349 struct ovs_key_ethernet *eth_key;
1350 struct nlattr *nla, *encap;
1352 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1353 goto nla_put_failure;
1355 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1356 goto nla_put_failure;
1358 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1359 goto nla_put_failure;
1361 if ((swkey->tun_key.u.ipv4.dst || is_mask)) {
1362 const void *opts = NULL;
1364 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1365 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1367 if (ipv4_tun_to_nlattr(skb, &output->tun_key, opts,
1368 swkey->tun_opts_len))
1369 goto nla_put_failure;
1372 if (swkey->phy.in_port == DP_MAX_PORTS) {
1373 if (is_mask && (output->phy.in_port == 0xffff))
1374 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1375 goto nla_put_failure;
1378 upper_u16 = !is_mask ? 0 : 0xffff;
1380 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1381 (upper_u16 << 16) | output->phy.in_port))
1382 goto nla_put_failure;
1385 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1386 goto nla_put_failure;
1388 if (ovs_ct_put_key(output, skb))
1389 goto nla_put_failure;
1391 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1393 goto nla_put_failure;
1395 eth_key = nla_data(nla);
1396 ether_addr_copy(eth_key->eth_src, output->eth.src);
1397 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1399 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1401 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
1402 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1403 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
1404 goto nla_put_failure;
1405 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1406 if (!swkey->eth.tci)
1411 if (swkey->eth.type == htons(ETH_P_802_2)) {
1413 * Ethertype 802.2 is represented in the netlink with omitted
1414 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1415 * 0xffff in the mask attribute. Ethertype can also
1418 if (is_mask && output->eth.type)
1419 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1421 goto nla_put_failure;
1425 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1426 goto nla_put_failure;
1428 if (swkey->eth.type == htons(ETH_P_IP)) {
1429 struct ovs_key_ipv4 *ipv4_key;
1431 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1433 goto nla_put_failure;
1434 ipv4_key = nla_data(nla);
1435 ipv4_key->ipv4_src = output->ipv4.addr.src;
1436 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1437 ipv4_key->ipv4_proto = output->ip.proto;
1438 ipv4_key->ipv4_tos = output->ip.tos;
1439 ipv4_key->ipv4_ttl = output->ip.ttl;
1440 ipv4_key->ipv4_frag = output->ip.frag;
1441 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1442 struct ovs_key_ipv6 *ipv6_key;
1444 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1446 goto nla_put_failure;
1447 ipv6_key = nla_data(nla);
1448 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1449 sizeof(ipv6_key->ipv6_src));
1450 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1451 sizeof(ipv6_key->ipv6_dst));
1452 ipv6_key->ipv6_label = output->ipv6.label;
1453 ipv6_key->ipv6_proto = output->ip.proto;
1454 ipv6_key->ipv6_tclass = output->ip.tos;
1455 ipv6_key->ipv6_hlimit = output->ip.ttl;
1456 ipv6_key->ipv6_frag = output->ip.frag;
1457 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1458 swkey->eth.type == htons(ETH_P_RARP)) {
1459 struct ovs_key_arp *arp_key;
1461 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1463 goto nla_put_failure;
1464 arp_key = nla_data(nla);
1465 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1466 arp_key->arp_sip = output->ipv4.addr.src;
1467 arp_key->arp_tip = output->ipv4.addr.dst;
1468 arp_key->arp_op = htons(output->ip.proto);
1469 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1470 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1471 } else if (eth_p_mpls(swkey->eth.type)) {
1472 struct ovs_key_mpls *mpls_key;
1474 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1476 goto nla_put_failure;
1477 mpls_key = nla_data(nla);
1478 mpls_key->mpls_lse = output->mpls.top_lse;
1481 if ((swkey->eth.type == htons(ETH_P_IP) ||
1482 swkey->eth.type == htons(ETH_P_IPV6)) &&
1483 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1485 if (swkey->ip.proto == IPPROTO_TCP) {
1486 struct ovs_key_tcp *tcp_key;
1488 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1490 goto nla_put_failure;
1491 tcp_key = nla_data(nla);
1492 tcp_key->tcp_src = output->tp.src;
1493 tcp_key->tcp_dst = output->tp.dst;
1494 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1496 goto nla_put_failure;
1497 } else if (swkey->ip.proto == IPPROTO_UDP) {
1498 struct ovs_key_udp *udp_key;
1500 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1502 goto nla_put_failure;
1503 udp_key = nla_data(nla);
1504 udp_key->udp_src = output->tp.src;
1505 udp_key->udp_dst = output->tp.dst;
1506 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1507 struct ovs_key_sctp *sctp_key;
1509 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1511 goto nla_put_failure;
1512 sctp_key = nla_data(nla);
1513 sctp_key->sctp_src = output->tp.src;
1514 sctp_key->sctp_dst = output->tp.dst;
1515 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1516 swkey->ip.proto == IPPROTO_ICMP) {
1517 struct ovs_key_icmp *icmp_key;
1519 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1521 goto nla_put_failure;
1522 icmp_key = nla_data(nla);
1523 icmp_key->icmp_type = ntohs(output->tp.src);
1524 icmp_key->icmp_code = ntohs(output->tp.dst);
1525 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1526 swkey->ip.proto == IPPROTO_ICMPV6) {
1527 struct ovs_key_icmpv6 *icmpv6_key;
1529 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1530 sizeof(*icmpv6_key));
1532 goto nla_put_failure;
1533 icmpv6_key = nla_data(nla);
1534 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1535 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1537 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1538 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1539 struct ovs_key_nd *nd_key;
1541 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1543 goto nla_put_failure;
1544 nd_key = nla_data(nla);
1545 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1546 sizeof(nd_key->nd_target));
1547 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1548 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1555 nla_nest_end(skb, encap);
1563 int ovs_nla_put_key(const struct sw_flow_key *swkey,
1564 const struct sw_flow_key *output, int attr, bool is_mask,
1565 struct sk_buff *skb)
1570 nla = nla_nest_start(skb, attr);
1573 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
1576 nla_nest_end(skb, nla);
1581 /* Called with ovs_mutex or RCU read lock. */
1582 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
1584 if (ovs_identifier_is_ufid(&flow->id))
1585 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
1588 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
1589 OVS_FLOW_ATTR_KEY, false, skb);
1592 /* Called with ovs_mutex or RCU read lock. */
1593 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
1595 return ovs_nla_put_key(&flow->key, &flow->key,
1596 OVS_FLOW_ATTR_KEY, false, skb);
1599 /* Called with ovs_mutex or RCU read lock. */
1600 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
1602 return ovs_nla_put_key(&flow->key, &flow->mask->key,
1603 OVS_FLOW_ATTR_MASK, true, skb);
1606 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1608 static struct sw_flow_actions *nla_alloc_flow_actions(int size, bool log)
1610 struct sw_flow_actions *sfa;
1612 if (size > MAX_ACTIONS_BUFSIZE) {
1613 OVS_NLERR(log, "Flow action size %u bytes exceeds max", size);
1614 return ERR_PTR(-EINVAL);
1617 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1619 return ERR_PTR(-ENOMEM);
1621 sfa->actions_len = 0;
1625 static void ovs_nla_free_set_action(const struct nlattr *a)
1627 const struct nlattr *ovs_key = nla_data(a);
1628 struct ovs_tunnel_info *ovs_tun;
1630 switch (nla_type(ovs_key)) {
1631 case OVS_KEY_ATTR_TUNNEL_INFO:
1632 ovs_tun = nla_data(ovs_key);
1633 ovs_dst_release((struct dst_entry *)ovs_tun->tun_dst);
1638 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1640 const struct nlattr *a;
1646 nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
1647 switch (nla_type(a)) {
1648 case OVS_ACTION_ATTR_SET:
1649 ovs_nla_free_set_action(a);
1651 case OVS_ACTION_ATTR_CT:
1652 ovs_ct_free_action(a);
1660 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
1662 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
1665 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1666 * The caller must hold rcu_read_lock for this to be sensible.
1668 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
1670 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
1673 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1674 int attr_len, bool log)
1677 struct sw_flow_actions *acts;
1679 int req_size = NLA_ALIGN(attr_len);
1680 int next_offset = offsetof(struct sw_flow_actions, actions) +
1681 (*sfa)->actions_len;
1683 if (req_size <= (ksize(*sfa) - next_offset))
1686 new_acts_size = ksize(*sfa) * 2;
1688 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1689 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1690 return ERR_PTR(-EMSGSIZE);
1691 new_acts_size = MAX_ACTIONS_BUFSIZE;
1694 acts = nla_alloc_flow_actions(new_acts_size, log);
1696 return (void *)acts;
1698 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1699 acts->actions_len = (*sfa)->actions_len;
1700 acts->orig_len = (*sfa)->orig_len;
1705 (*sfa)->actions_len += req_size;
1706 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1709 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1710 int attrtype, void *data, int len, bool log)
1714 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
1718 a->nla_type = attrtype;
1719 a->nla_len = nla_attr_size(len);
1722 memcpy(nla_data(a), data, len);
1723 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1728 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
1733 a = __add_action(sfa, attrtype, data, len, log);
1740 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1741 int attrtype, bool log)
1743 int used = (*sfa)->actions_len;
1746 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
1753 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1756 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1759 a->nla_len = sfa->actions_len - st_offset;
1762 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
1763 const struct sw_flow_key *key,
1764 int depth, struct sw_flow_actions **sfa,
1765 __be16 eth_type, __be16 vlan_tci, bool log);
1767 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
1768 const struct sw_flow_key *key, int depth,
1769 struct sw_flow_actions **sfa,
1770 __be16 eth_type, __be16 vlan_tci, bool log)
1772 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1773 const struct nlattr *probability, *actions;
1774 const struct nlattr *a;
1775 int rem, start, err, st_acts;
1777 memset(attrs, 0, sizeof(attrs));
1778 nla_for_each_nested(a, attr, rem) {
1779 int type = nla_type(a);
1780 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1787 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1788 if (!probability || nla_len(probability) != sizeof(u32))
1791 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1792 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1795 /* validation done, copy sample action. */
1796 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
1799 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1800 nla_data(probability), sizeof(u32), log);
1803 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS, log);
1807 err = __ovs_nla_copy_actions(net, actions, key, depth + 1, sfa,
1808 eth_type, vlan_tci, log);
1812 add_nested_action_end(*sfa, st_acts);
1813 add_nested_action_end(*sfa, start);
1818 void ovs_match_init(struct sw_flow_match *match,
1819 struct sw_flow_key *key,
1820 struct sw_flow_mask *mask)
1822 memset(match, 0, sizeof(*match));
1826 memset(key, 0, sizeof(*key));
1829 memset(&mask->key, 0, sizeof(mask->key));
1830 mask->range.start = mask->range.end = 0;
1834 static int validate_geneve_opts(struct sw_flow_key *key)
1836 struct geneve_opt *option;
1837 int opts_len = key->tun_opts_len;
1838 bool crit_opt = false;
1840 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
1841 while (opts_len > 0) {
1844 if (opts_len < sizeof(*option))
1847 len = sizeof(*option) + option->length * 4;
1851 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
1853 option = (struct geneve_opt *)((u8 *)option + len);
1857 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
1862 static int validate_and_copy_set_tun(const struct nlattr *attr,
1863 struct sw_flow_actions **sfa, bool log)
1865 struct sw_flow_match match;
1866 struct sw_flow_key key;
1867 struct metadata_dst *tun_dst;
1868 struct ip_tunnel_info *tun_info;
1869 struct ovs_tunnel_info *ovs_tun;
1871 int err = 0, start, opts_type;
1873 ovs_match_init(&match, &key, NULL);
1874 opts_type = ipv4_tun_from_nlattr(nla_data(attr), &match, false, log);
1878 if (key.tun_opts_len) {
1879 switch (opts_type) {
1880 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
1881 err = validate_geneve_opts(&key);
1885 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
1890 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
1894 tun_dst = metadata_dst_alloc(key.tun_opts_len, GFP_KERNEL);
1898 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
1899 sizeof(*ovs_tun), log);
1901 ovs_dst_release((struct dst_entry *)tun_dst);
1905 ovs_tun = nla_data(a);
1906 ovs_tun->tun_dst = tun_dst;
1908 tun_info = &tun_dst->u.tun_info;
1909 tun_info->mode = IP_TUNNEL_INFO_TX;
1910 tun_info->key = key.tun_key;
1912 /* We need to store the options in the action itself since
1913 * everything else will go away after flow setup. We can append
1914 * it to tun_info and then point there.
1916 ip_tunnel_info_opts_set(tun_info,
1917 TUN_METADATA_OPTS(&key, key.tun_opts_len),
1919 add_nested_action_end(*sfa, start);
1924 /* Return false if there are any non-masked bits set.
1925 * Mask follows data immediately, before any netlink padding.
1927 static bool validate_masked(u8 *data, int len)
1929 u8 *mask = data + len;
1932 if (*data++ & ~*mask++)
1938 static int validate_set(const struct nlattr *a,
1939 const struct sw_flow_key *flow_key,
1940 struct sw_flow_actions **sfa,
1941 bool *skip_copy, __be16 eth_type, bool masked, bool log)
1943 const struct nlattr *ovs_key = nla_data(a);
1944 int key_type = nla_type(ovs_key);
1947 /* There can be only one key in a action */
1948 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1951 key_len = nla_len(ovs_key);
1955 if (key_type > OVS_KEY_ATTR_MAX ||
1956 !check_attr_len(key_len, ovs_key_lens[key_type].len))
1959 if (masked && !validate_masked(nla_data(ovs_key), key_len))
1963 const struct ovs_key_ipv4 *ipv4_key;
1964 const struct ovs_key_ipv6 *ipv6_key;
1967 case OVS_KEY_ATTR_PRIORITY:
1968 case OVS_KEY_ATTR_SKB_MARK:
1969 case OVS_KEY_ATTR_CT_MARK:
1970 case OVS_KEY_ATTR_ETHERNET:
1973 case OVS_KEY_ATTR_TUNNEL:
1974 if (eth_p_mpls(eth_type))
1978 return -EINVAL; /* Masked tunnel set not supported. */
1981 err = validate_and_copy_set_tun(a, sfa, log);
1986 case OVS_KEY_ATTR_IPV4:
1987 if (eth_type != htons(ETH_P_IP))
1990 ipv4_key = nla_data(ovs_key);
1993 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
1995 /* Non-writeable fields. */
1996 if (mask->ipv4_proto || mask->ipv4_frag)
1999 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2002 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2007 case OVS_KEY_ATTR_IPV6:
2008 if (eth_type != htons(ETH_P_IPV6))
2011 ipv6_key = nla_data(ovs_key);
2014 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2016 /* Non-writeable fields. */
2017 if (mask->ipv6_proto || mask->ipv6_frag)
2020 /* Invalid bits in the flow label mask? */
2021 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2024 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2027 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2030 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2035 case OVS_KEY_ATTR_TCP:
2036 if ((eth_type != htons(ETH_P_IP) &&
2037 eth_type != htons(ETH_P_IPV6)) ||
2038 flow_key->ip.proto != IPPROTO_TCP)
2043 case OVS_KEY_ATTR_UDP:
2044 if ((eth_type != htons(ETH_P_IP) &&
2045 eth_type != htons(ETH_P_IPV6)) ||
2046 flow_key->ip.proto != IPPROTO_UDP)
2051 case OVS_KEY_ATTR_MPLS:
2052 if (!eth_p_mpls(eth_type))
2056 case OVS_KEY_ATTR_SCTP:
2057 if ((eth_type != htons(ETH_P_IP) &&
2058 eth_type != htons(ETH_P_IPV6)) ||
2059 flow_key->ip.proto != IPPROTO_SCTP)
2068 /* Convert non-masked non-tunnel set actions to masked set actions. */
2069 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2070 int start, len = key_len * 2;
2075 start = add_nested_action_start(sfa,
2076 OVS_ACTION_ATTR_SET_TO_MASKED,
2081 at = __add_action(sfa, key_type, NULL, len, log);
2085 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2086 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
2087 /* Clear non-writeable bits from otherwise writeable fields. */
2088 if (key_type == OVS_KEY_ATTR_IPV6) {
2089 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2091 mask->ipv6_label &= htonl(0x000FFFFF);
2093 add_nested_action_end(*sfa, start);
2099 static int validate_userspace(const struct nlattr *attr)
2101 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2102 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2103 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2104 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2106 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2109 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
2110 attr, userspace_policy);
2114 if (!a[OVS_USERSPACE_ATTR_PID] ||
2115 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2121 static int copy_action(const struct nlattr *from,
2122 struct sw_flow_actions **sfa, bool log)
2124 int totlen = NLA_ALIGN(from->nla_len);
2127 to = reserve_sfa_size(sfa, from->nla_len, log);
2131 memcpy(to, from, totlen);
2135 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2136 const struct sw_flow_key *key,
2137 int depth, struct sw_flow_actions **sfa,
2138 __be16 eth_type, __be16 vlan_tci, bool log)
2140 const struct nlattr *a;
2143 if (depth >= SAMPLE_ACTION_DEPTH)
2146 nla_for_each_nested(a, attr, rem) {
2147 /* Expected argument lengths, (u32)-1 for variable length. */
2148 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2149 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2150 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2151 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2152 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2153 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2154 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2155 [OVS_ACTION_ATTR_POP_VLAN] = 0,
2156 [OVS_ACTION_ATTR_SET] = (u32)-1,
2157 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2158 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2159 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2160 [OVS_ACTION_ATTR_CT] = (u32)-1,
2162 const struct ovs_action_push_vlan *vlan;
2163 int type = nla_type(a);
2166 if (type > OVS_ACTION_ATTR_MAX ||
2167 (action_lens[type] != nla_len(a) &&
2168 action_lens[type] != (u32)-1))
2173 case OVS_ACTION_ATTR_UNSPEC:
2176 case OVS_ACTION_ATTR_USERSPACE:
2177 err = validate_userspace(a);
2182 case OVS_ACTION_ATTR_OUTPUT:
2183 if (nla_get_u32(a) >= DP_MAX_PORTS)
2187 case OVS_ACTION_ATTR_HASH: {
2188 const struct ovs_action_hash *act_hash = nla_data(a);
2190 switch (act_hash->hash_alg) {
2191 case OVS_HASH_ALG_L4:
2200 case OVS_ACTION_ATTR_POP_VLAN:
2201 vlan_tci = htons(0);
2204 case OVS_ACTION_ATTR_PUSH_VLAN:
2206 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
2208 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2210 vlan_tci = vlan->vlan_tci;
2213 case OVS_ACTION_ATTR_RECIRC:
2216 case OVS_ACTION_ATTR_PUSH_MPLS: {
2217 const struct ovs_action_push_mpls *mpls = nla_data(a);
2219 if (!eth_p_mpls(mpls->mpls_ethertype))
2221 /* Prohibit push MPLS other than to a white list
2222 * for packets that have a known tag order.
2224 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2225 (eth_type != htons(ETH_P_IP) &&
2226 eth_type != htons(ETH_P_IPV6) &&
2227 eth_type != htons(ETH_P_ARP) &&
2228 eth_type != htons(ETH_P_RARP) &&
2229 !eth_p_mpls(eth_type)))
2231 eth_type = mpls->mpls_ethertype;
2235 case OVS_ACTION_ATTR_POP_MPLS:
2236 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2237 !eth_p_mpls(eth_type))
2240 /* Disallow subsequent L2.5+ set and mpls_pop actions
2241 * as there is no check here to ensure that the new
2242 * eth_type is valid and thus set actions could
2243 * write off the end of the packet or otherwise
2246 * Support for these actions is planned using packet
2249 eth_type = htons(0);
2252 case OVS_ACTION_ATTR_SET:
2253 err = validate_set(a, key, sfa,
2254 &skip_copy, eth_type, false, log);
2259 case OVS_ACTION_ATTR_SET_MASKED:
2260 err = validate_set(a, key, sfa,
2261 &skip_copy, eth_type, true, log);
2266 case OVS_ACTION_ATTR_SAMPLE:
2267 err = validate_and_copy_sample(net, a, key, depth, sfa,
2268 eth_type, vlan_tci, log);
2274 case OVS_ACTION_ATTR_CT:
2275 err = ovs_ct_copy_action(net, a, key, sfa, log);
2282 OVS_NLERR(log, "Unknown Action type %d", type);
2286 err = copy_action(a, sfa, log);
2298 /* 'key' must be the masked key. */
2299 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2300 const struct sw_flow_key *key,
2301 struct sw_flow_actions **sfa, bool log)
2305 *sfa = nla_alloc_flow_actions(nla_len(attr), log);
2307 return PTR_ERR(*sfa);
2309 (*sfa)->orig_len = nla_len(attr);
2310 err = __ovs_nla_copy_actions(net, attr, key, 0, sfa, key->eth.type,
2313 ovs_nla_free_flow_actions(*sfa);
2318 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
2320 const struct nlattr *a;
2321 struct nlattr *start;
2324 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
2328 nla_for_each_nested(a, attr, rem) {
2329 int type = nla_type(a);
2330 struct nlattr *st_sample;
2333 case OVS_SAMPLE_ATTR_PROBABILITY:
2334 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
2335 sizeof(u32), nla_data(a)))
2338 case OVS_SAMPLE_ATTR_ACTIONS:
2339 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
2342 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
2345 nla_nest_end(skb, st_sample);
2350 nla_nest_end(skb, start);
2354 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
2356 const struct nlattr *ovs_key = nla_data(a);
2357 int key_type = nla_type(ovs_key);
2358 struct nlattr *start;
2362 case OVS_KEY_ATTR_TUNNEL_INFO: {
2363 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
2364 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
2366 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2370 err = ipv4_tun_to_nlattr(skb, &tun_info->key,
2371 tun_info->options_len ?
2372 ip_tunnel_info_opts(tun_info) : NULL,
2373 tun_info->options_len);
2376 nla_nest_end(skb, start);
2380 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
2388 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
2389 struct sk_buff *skb)
2391 const struct nlattr *ovs_key = nla_data(a);
2392 size_t key_len = nla_len(ovs_key) / 2;
2394 /* Revert the conversion we did from a non-masked set action to
2395 * masked set action.
2397 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a) - key_len, ovs_key))
2403 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
2405 const struct nlattr *a;
2408 nla_for_each_attr(a, attr, len, rem) {
2409 int type = nla_type(a);
2412 case OVS_ACTION_ATTR_SET:
2413 err = set_action_to_attr(a, skb);
2418 case OVS_ACTION_ATTR_SET_TO_MASKED:
2419 err = masked_set_action_to_set_action_attr(a, skb);
2424 case OVS_ACTION_ATTR_SAMPLE:
2425 err = sample_action_to_attr(a, skb);
2430 case OVS_ACTION_ATTR_CT:
2431 err = ovs_ct_action_to_attr(nla_data(a), skb);
2437 if (nla_put(skb, type, nla_len(a), nla_data(a)))