1 // SPDX-License-Identifier: GPL-2.0-only
3 * Copyright (c) 2007-2017 Nicira, Inc.
6 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
10 #include <linux/uaccess.h>
11 #include <linux/netdevice.h>
12 #include <linux/etherdevice.h>
13 #include <linux/if_ether.h>
14 #include <linux/if_vlan.h>
15 #include <net/llc_pdu.h>
16 #include <linux/kernel.h>
17 #include <linux/jhash.h>
18 #include <linux/jiffies.h>
19 #include <linux/llc.h>
20 #include <linux/module.h>
22 #include <linux/rcupdate.h>
23 #include <linux/if_arp.h>
25 #include <linux/ipv6.h>
26 #include <linux/sctp.h>
27 #include <linux/tcp.h>
28 #include <linux/udp.h>
29 #include <linux/icmp.h>
30 #include <linux/icmpv6.h>
31 #include <linux/rculist.h>
32 #include <net/geneve.h>
35 #include <net/ndisc.h>
37 #include <net/vxlan.h>
38 #include <net/tun_proto.h>
39 #include <net/erspan.h>
41 #include "flow_netlink.h"
45 const struct ovs_len_tbl *next;
48 #define OVS_ATTR_NESTED -1
49 #define OVS_ATTR_VARIABLE -2
50 #define OVS_COPY_ACTIONS_MAX_DEPTH 16
52 static bool actions_may_change_flow(const struct nlattr *actions)
57 nla_for_each_nested(nla, actions, rem) {
58 u16 action = nla_type(nla);
61 case OVS_ACTION_ATTR_OUTPUT:
62 case OVS_ACTION_ATTR_RECIRC:
63 case OVS_ACTION_ATTR_TRUNC:
64 case OVS_ACTION_ATTR_USERSPACE:
67 case OVS_ACTION_ATTR_CT:
68 case OVS_ACTION_ATTR_CT_CLEAR:
69 case OVS_ACTION_ATTR_HASH:
70 case OVS_ACTION_ATTR_POP_ETH:
71 case OVS_ACTION_ATTR_POP_MPLS:
72 case OVS_ACTION_ATTR_POP_NSH:
73 case OVS_ACTION_ATTR_POP_VLAN:
74 case OVS_ACTION_ATTR_PUSH_ETH:
75 case OVS_ACTION_ATTR_PUSH_MPLS:
76 case OVS_ACTION_ATTR_PUSH_NSH:
77 case OVS_ACTION_ATTR_PUSH_VLAN:
78 case OVS_ACTION_ATTR_SAMPLE:
79 case OVS_ACTION_ATTR_SET:
80 case OVS_ACTION_ATTR_SET_MASKED:
81 case OVS_ACTION_ATTR_METER:
82 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
83 case OVS_ACTION_ATTR_ADD_MPLS:
84 case OVS_ACTION_ATTR_DEC_TTL:
92 static void update_range(struct sw_flow_match *match,
93 size_t offset, size_t size, bool is_mask)
95 struct sw_flow_key_range *range;
96 size_t start = rounddown(offset, sizeof(long));
97 size_t end = roundup(offset + size, sizeof(long));
100 range = &match->range;
102 range = &match->mask->range;
104 if (range->start == range->end) {
105 range->start = start;
110 if (range->start > start)
111 range->start = start;
113 if (range->end < end)
117 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
119 update_range(match, offsetof(struct sw_flow_key, field), \
120 sizeof((match)->key->field), is_mask); \
122 (match)->mask->key.field = value; \
124 (match)->key->field = value; \
127 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
129 update_range(match, offset, len, is_mask); \
131 memcpy((u8 *)&(match)->mask->key + offset, value_p, \
134 memcpy((u8 *)(match)->key + offset, value_p, len); \
137 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
138 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
139 value_p, len, is_mask)
141 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
143 update_range(match, offsetof(struct sw_flow_key, field), \
144 sizeof((match)->key->field), is_mask); \
146 memset((u8 *)&(match)->mask->key.field, value, \
147 sizeof((match)->mask->key.field)); \
149 memset((u8 *)&(match)->key->field, value, \
150 sizeof((match)->key->field)); \
153 static bool match_validate(const struct sw_flow_match *match,
154 u64 key_attrs, u64 mask_attrs, bool log)
156 u64 key_expected = 0;
157 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
159 /* The following mask attributes allowed only if they
160 * pass the validation tests. */
161 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
162 | (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)
163 | (1 << OVS_KEY_ATTR_IPV6)
164 | (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)
165 | (1 << OVS_KEY_ATTR_TCP)
166 | (1 << OVS_KEY_ATTR_TCP_FLAGS)
167 | (1 << OVS_KEY_ATTR_UDP)
168 | (1 << OVS_KEY_ATTR_SCTP)
169 | (1 << OVS_KEY_ATTR_ICMP)
170 | (1 << OVS_KEY_ATTR_ICMPV6)
171 | (1 << OVS_KEY_ATTR_ARP)
172 | (1 << OVS_KEY_ATTR_ND)
173 | (1 << OVS_KEY_ATTR_MPLS)
174 | (1 << OVS_KEY_ATTR_NSH));
176 /* Always allowed mask fields. */
177 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
178 | (1 << OVS_KEY_ATTR_IN_PORT)
179 | (1 << OVS_KEY_ATTR_ETHERTYPE));
181 /* Check key attributes. */
182 if (match->key->eth.type == htons(ETH_P_ARP)
183 || match->key->eth.type == htons(ETH_P_RARP)) {
184 key_expected |= 1 << OVS_KEY_ATTR_ARP;
185 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
186 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
189 if (eth_p_mpls(match->key->eth.type)) {
190 key_expected |= 1 << OVS_KEY_ATTR_MPLS;
191 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
192 mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
195 if (match->key->eth.type == htons(ETH_P_IP)) {
196 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
197 if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
198 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
199 mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4;
202 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
203 if (match->key->ip.proto == IPPROTO_UDP) {
204 key_expected |= 1 << OVS_KEY_ATTR_UDP;
205 if (match->mask && (match->mask->key.ip.proto == 0xff))
206 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
209 if (match->key->ip.proto == IPPROTO_SCTP) {
210 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
211 if (match->mask && (match->mask->key.ip.proto == 0xff))
212 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
215 if (match->key->ip.proto == IPPROTO_TCP) {
216 key_expected |= 1 << OVS_KEY_ATTR_TCP;
217 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
218 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
219 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
220 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
224 if (match->key->ip.proto == IPPROTO_ICMP) {
225 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
226 if (match->mask && (match->mask->key.ip.proto == 0xff))
227 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
232 if (match->key->eth.type == htons(ETH_P_IPV6)) {
233 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
234 if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
235 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
236 mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6;
239 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
240 if (match->key->ip.proto == IPPROTO_UDP) {
241 key_expected |= 1 << OVS_KEY_ATTR_UDP;
242 if (match->mask && (match->mask->key.ip.proto == 0xff))
243 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
246 if (match->key->ip.proto == IPPROTO_SCTP) {
247 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
248 if (match->mask && (match->mask->key.ip.proto == 0xff))
249 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
252 if (match->key->ip.proto == IPPROTO_TCP) {
253 key_expected |= 1 << OVS_KEY_ATTR_TCP;
254 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
255 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
256 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
257 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
261 if (match->key->ip.proto == IPPROTO_ICMPV6) {
262 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
263 if (match->mask && (match->mask->key.ip.proto == 0xff))
264 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
266 if (match->key->tp.src ==
267 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
268 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
269 key_expected |= 1 << OVS_KEY_ATTR_ND;
270 /* Original direction conntrack tuple
271 * uses the same space as the ND fields
272 * in the key, so both are not allowed
275 mask_allowed &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
276 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
277 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
283 if (match->key->eth.type == htons(ETH_P_NSH)) {
284 key_expected |= 1 << OVS_KEY_ATTR_NSH;
286 match->mask->key.eth.type == htons(0xffff)) {
287 mask_allowed |= 1 << OVS_KEY_ATTR_NSH;
291 if ((key_attrs & key_expected) != key_expected) {
292 /* Key attributes check failed. */
293 OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
294 (unsigned long long)key_attrs,
295 (unsigned long long)key_expected);
299 if ((mask_attrs & mask_allowed) != mask_attrs) {
300 /* Mask attributes check failed. */
301 OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
302 (unsigned long long)mask_attrs,
303 (unsigned long long)mask_allowed);
310 size_t ovs_tun_key_attr_size(void)
312 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
313 * updating this function.
315 return nla_total_size_64bit(8) /* OVS_TUNNEL_KEY_ATTR_ID */
316 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
317 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_DST */
318 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
319 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
320 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
321 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
322 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
323 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
324 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS and
325 * OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS is mutually exclusive with
326 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
328 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
329 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
332 static size_t ovs_nsh_key_attr_size(void)
334 /* Whenever adding new OVS_NSH_KEY_ FIELDS, we should consider
335 * updating this function.
337 return nla_total_size(NSH_BASE_HDR_LEN) /* OVS_NSH_KEY_ATTR_BASE */
338 /* OVS_NSH_KEY_ATTR_MD1 and OVS_NSH_KEY_ATTR_MD2 are
339 * mutually exclusive, so the bigger one can cover
342 + nla_total_size(NSH_CTX_HDRS_MAX_LEN);
345 size_t ovs_key_attr_size(void)
347 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
348 * updating this function.
350 BUILD_BUG_ON(OVS_KEY_ATTR_MAX != 32);
352 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
353 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
354 + ovs_tun_key_attr_size()
355 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
356 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
357 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
358 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
359 + nla_total_size(4) /* OVS_KEY_ATTR_CT_STATE */
360 + nla_total_size(2) /* OVS_KEY_ATTR_CT_ZONE */
361 + nla_total_size(4) /* OVS_KEY_ATTR_CT_MARK */
362 + nla_total_size(16) /* OVS_KEY_ATTR_CT_LABELS */
363 + nla_total_size(40) /* OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6 */
364 + nla_total_size(0) /* OVS_KEY_ATTR_NSH */
365 + ovs_nsh_key_attr_size()
366 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
367 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
368 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
369 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
370 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
371 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
372 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
373 + nla_total_size(28) /* OVS_KEY_ATTR_ND */
374 + nla_total_size(2); /* OVS_KEY_ATTR_IPV6_EXTHDRS */
377 static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
378 [OVS_VXLAN_EXT_GBP] = { .len = sizeof(u32) },
381 static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
382 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
383 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
384 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
385 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
386 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
387 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
388 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
389 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
390 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
391 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
392 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_VARIABLE },
393 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED,
394 .next = ovs_vxlan_ext_key_lens },
395 [OVS_TUNNEL_KEY_ATTR_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
396 [OVS_TUNNEL_KEY_ATTR_IPV6_DST] = { .len = sizeof(struct in6_addr) },
397 [OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS] = { .len = OVS_ATTR_VARIABLE },
398 [OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE] = { .len = 0 },
401 static const struct ovs_len_tbl
402 ovs_nsh_key_attr_lens[OVS_NSH_KEY_ATTR_MAX + 1] = {
403 [OVS_NSH_KEY_ATTR_BASE] = { .len = sizeof(struct ovs_nsh_key_base) },
404 [OVS_NSH_KEY_ATTR_MD1] = { .len = sizeof(struct ovs_nsh_key_md1) },
405 [OVS_NSH_KEY_ATTR_MD2] = { .len = OVS_ATTR_VARIABLE },
408 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
409 static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
410 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
411 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
412 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
413 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
414 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
415 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
416 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
417 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
418 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
419 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
420 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
421 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
422 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
423 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
424 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
425 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
426 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
427 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
428 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
429 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
430 .next = ovs_tunnel_key_lens, },
431 [OVS_KEY_ATTR_MPLS] = { .len = OVS_ATTR_VARIABLE },
432 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u32) },
433 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
434 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
435 [OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
436 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4] = {
437 .len = sizeof(struct ovs_key_ct_tuple_ipv4) },
438 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6] = {
439 .len = sizeof(struct ovs_key_ct_tuple_ipv6) },
440 [OVS_KEY_ATTR_NSH] = { .len = OVS_ATTR_NESTED,
441 .next = ovs_nsh_key_attr_lens, },
442 [OVS_KEY_ATTR_IPV6_EXTHDRS] = {
443 .len = sizeof(struct ovs_key_ipv6_exthdrs) },
446 static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
448 return expected_len == attr_len ||
449 expected_len == OVS_ATTR_NESTED ||
450 expected_len == OVS_ATTR_VARIABLE;
453 static bool is_all_zero(const u8 *fp, size_t size)
460 for (i = 0; i < size; i++)
467 static int __parse_flow_nlattrs(const struct nlattr *attr,
468 const struct nlattr *a[],
469 u64 *attrsp, bool log, bool nz)
471 const struct nlattr *nla;
476 nla_for_each_nested(nla, attr, rem) {
477 u16 type = nla_type(nla);
480 if (type > OVS_KEY_ATTR_MAX) {
481 OVS_NLERR(log, "Key type %d is out of range max %d",
482 type, OVS_KEY_ATTR_MAX);
486 if (type == OVS_KEY_ATTR_PACKET_TYPE ||
487 type == OVS_KEY_ATTR_ND_EXTENSIONS ||
488 type == OVS_KEY_ATTR_TUNNEL_INFO) {
489 OVS_NLERR(log, "Key type %d is not supported", type);
493 if (attrs & (1ULL << type)) {
494 OVS_NLERR(log, "Duplicate key (type %d).", type);
498 expected_len = ovs_key_lens[type].len;
499 if (!check_attr_len(nla_len(nla), expected_len)) {
500 OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
501 type, nla_len(nla), expected_len);
505 if (!nz || !is_all_zero(nla_data(nla), nla_len(nla))) {
506 attrs |= 1ULL << type;
511 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
519 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
520 const struct nlattr *a[], u64 *attrsp,
523 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
526 int parse_flow_nlattrs(const struct nlattr *attr, const struct nlattr *a[],
527 u64 *attrsp, bool log)
529 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
532 static int genev_tun_opt_from_nlattr(const struct nlattr *a,
533 struct sw_flow_match *match, bool is_mask,
536 unsigned long opt_key_offset;
538 if (nla_len(a) > sizeof(match->key->tun_opts)) {
539 OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
540 nla_len(a), sizeof(match->key->tun_opts));
544 if (nla_len(a) % 4 != 0) {
545 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
550 /* We need to record the length of the options passed
551 * down, otherwise packets with the same format but
552 * additional options will be silently matched.
555 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
558 /* This is somewhat unusual because it looks at
559 * both the key and mask while parsing the
560 * attributes (and by extension assumes the key
561 * is parsed first). Normally, we would verify
562 * that each is the correct length and that the
563 * attributes line up in the validate function.
564 * However, that is difficult because this is
565 * variable length and we won't have the
568 if (match->key->tun_opts_len != nla_len(a)) {
569 OVS_NLERR(log, "Geneve option len %d != mask len %d",
570 match->key->tun_opts_len, nla_len(a));
574 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
577 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
578 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
579 nla_len(a), is_mask);
583 static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
584 struct sw_flow_match *match, bool is_mask,
589 unsigned long opt_key_offset;
590 struct vxlan_metadata opts;
592 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
594 memset(&opts, 0, sizeof(opts));
595 nla_for_each_nested(a, attr, rem) {
596 int type = nla_type(a);
598 if (type > OVS_VXLAN_EXT_MAX) {
599 OVS_NLERR(log, "VXLAN extension %d out of range max %d",
600 type, OVS_VXLAN_EXT_MAX);
604 if (!check_attr_len(nla_len(a),
605 ovs_vxlan_ext_key_lens[type].len)) {
606 OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
608 ovs_vxlan_ext_key_lens[type].len);
613 case OVS_VXLAN_EXT_GBP:
614 opts.gbp = nla_get_u32(a);
617 OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
623 OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
629 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
631 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
633 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
634 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
639 static int erspan_tun_opt_from_nlattr(const struct nlattr *a,
640 struct sw_flow_match *match, bool is_mask,
643 unsigned long opt_key_offset;
645 BUILD_BUG_ON(sizeof(struct erspan_metadata) >
646 sizeof(match->key->tun_opts));
648 if (nla_len(a) > sizeof(match->key->tun_opts)) {
649 OVS_NLERR(log, "ERSPAN option length err (len %d, max %zu).",
650 nla_len(a), sizeof(match->key->tun_opts));
655 SW_FLOW_KEY_PUT(match, tun_opts_len,
656 sizeof(struct erspan_metadata), false);
658 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
660 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
661 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
662 nla_len(a), is_mask);
666 static int ip_tun_from_nlattr(const struct nlattr *attr,
667 struct sw_flow_match *match, bool is_mask,
670 bool ttl = false, ipv4 = false, ipv6 = false;
671 bool info_bridge_mode = false;
672 __be16 tun_flags = 0;
677 nla_for_each_nested(a, attr, rem) {
678 int type = nla_type(a);
681 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
682 OVS_NLERR(log, "Tunnel attr %d out of range max %d",
683 type, OVS_TUNNEL_KEY_ATTR_MAX);
687 if (!check_attr_len(nla_len(a),
688 ovs_tunnel_key_lens[type].len)) {
689 OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
690 type, nla_len(a), ovs_tunnel_key_lens[type].len);
695 case OVS_TUNNEL_KEY_ATTR_ID:
696 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
697 nla_get_be64(a), is_mask);
698 tun_flags |= TUNNEL_KEY;
700 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
701 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
702 nla_get_in_addr(a), is_mask);
705 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
706 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
707 nla_get_in_addr(a), is_mask);
710 case OVS_TUNNEL_KEY_ATTR_IPV6_SRC:
711 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.src,
712 nla_get_in6_addr(a), is_mask);
715 case OVS_TUNNEL_KEY_ATTR_IPV6_DST:
716 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
717 nla_get_in6_addr(a), is_mask);
720 case OVS_TUNNEL_KEY_ATTR_TOS:
721 SW_FLOW_KEY_PUT(match, tun_key.tos,
722 nla_get_u8(a), is_mask);
724 case OVS_TUNNEL_KEY_ATTR_TTL:
725 SW_FLOW_KEY_PUT(match, tun_key.ttl,
726 nla_get_u8(a), is_mask);
729 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
730 tun_flags |= TUNNEL_DONT_FRAGMENT;
732 case OVS_TUNNEL_KEY_ATTR_CSUM:
733 tun_flags |= TUNNEL_CSUM;
735 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
736 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
737 nla_get_be16(a), is_mask);
739 case OVS_TUNNEL_KEY_ATTR_TP_DST:
740 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
741 nla_get_be16(a), is_mask);
743 case OVS_TUNNEL_KEY_ATTR_OAM:
744 tun_flags |= TUNNEL_OAM;
746 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
748 OVS_NLERR(log, "Multiple metadata blocks provided");
752 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
756 tun_flags |= TUNNEL_GENEVE_OPT;
759 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
761 OVS_NLERR(log, "Multiple metadata blocks provided");
765 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
769 tun_flags |= TUNNEL_VXLAN_OPT;
772 case OVS_TUNNEL_KEY_ATTR_PAD:
774 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
776 OVS_NLERR(log, "Multiple metadata blocks provided");
780 err = erspan_tun_opt_from_nlattr(a, match, is_mask,
785 tun_flags |= TUNNEL_ERSPAN_OPT;
788 case OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE:
789 info_bridge_mode = true;
793 OVS_NLERR(log, "Unknown IP tunnel attribute %d",
799 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
801 SW_FLOW_KEY_MEMSET_FIELD(match, tun_proto, 0xff, true);
803 SW_FLOW_KEY_PUT(match, tun_proto, ipv6 ? AF_INET6 : AF_INET,
807 OVS_NLERR(log, "IP tunnel attribute has %d unknown bytes.",
813 OVS_NLERR(log, "Mixed IPv4 and IPv6 tunnel attributes");
818 if (!ipv4 && !ipv6) {
819 OVS_NLERR(log, "IP tunnel dst address not specified");
823 if (info_bridge_mode) {
824 if (match->key->tun_key.u.ipv4.src ||
825 match->key->tun_key.u.ipv4.dst ||
826 match->key->tun_key.tp_src ||
827 match->key->tun_key.tp_dst ||
828 match->key->tun_key.ttl ||
829 match->key->tun_key.tos ||
830 tun_flags & ~TUNNEL_KEY) {
831 OVS_NLERR(log, "IPv4 tun info is not correct");
834 } else if (!match->key->tun_key.u.ipv4.dst) {
835 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
839 if (ipv6 && ipv6_addr_any(&match->key->tun_key.u.ipv6.dst)) {
840 OVS_NLERR(log, "IPv6 tunnel dst address is zero");
844 if (!ttl && !info_bridge_mode) {
845 OVS_NLERR(log, "IP tunnel TTL not specified.");
853 static int vxlan_opt_to_nlattr(struct sk_buff *skb,
854 const void *tun_opts, int swkey_tun_opts_len)
856 const struct vxlan_metadata *opts = tun_opts;
859 nla = nla_nest_start_noflag(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
863 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
866 nla_nest_end(skb, nla);
870 static int __ip_tun_to_nlattr(struct sk_buff *skb,
871 const struct ip_tunnel_key *output,
872 const void *tun_opts, int swkey_tun_opts_len,
873 unsigned short tun_proto, u8 mode)
875 if (output->tun_flags & TUNNEL_KEY &&
876 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id,
877 OVS_TUNNEL_KEY_ATTR_PAD))
880 if (mode & IP_TUNNEL_INFO_BRIDGE)
881 return nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE)
886 if (output->u.ipv4.src &&
887 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
890 if (output->u.ipv4.dst &&
891 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
896 if (!ipv6_addr_any(&output->u.ipv6.src) &&
897 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_SRC,
898 &output->u.ipv6.src))
900 if (!ipv6_addr_any(&output->u.ipv6.dst) &&
901 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_DST,
902 &output->u.ipv6.dst))
907 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
909 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
911 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
912 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
914 if ((output->tun_flags & TUNNEL_CSUM) &&
915 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
917 if (output->tp_src &&
918 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
920 if (output->tp_dst &&
921 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
923 if ((output->tun_flags & TUNNEL_OAM) &&
924 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
926 if (swkey_tun_opts_len) {
927 if (output->tun_flags & TUNNEL_GENEVE_OPT &&
928 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
929 swkey_tun_opts_len, tun_opts))
931 else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
932 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
934 else if (output->tun_flags & TUNNEL_ERSPAN_OPT &&
935 nla_put(skb, OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS,
936 swkey_tun_opts_len, tun_opts))
943 static int ip_tun_to_nlattr(struct sk_buff *skb,
944 const struct ip_tunnel_key *output,
945 const void *tun_opts, int swkey_tun_opts_len,
946 unsigned short tun_proto, u8 mode)
951 nla = nla_nest_start_noflag(skb, OVS_KEY_ATTR_TUNNEL);
955 err = __ip_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len,
960 nla_nest_end(skb, nla);
964 int ovs_nla_put_tunnel_info(struct sk_buff *skb,
965 struct ip_tunnel_info *tun_info)
967 return __ip_tun_to_nlattr(skb, &tun_info->key,
968 ip_tunnel_info_opts(tun_info),
969 tun_info->options_len,
970 ip_tunnel_info_af(tun_info), tun_info->mode);
973 static int encode_vlan_from_nlattrs(struct sw_flow_match *match,
974 const struct nlattr *a[],
975 bool is_mask, bool inner)
980 if (a[OVS_KEY_ATTR_VLAN])
981 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
983 if (a[OVS_KEY_ATTR_ETHERTYPE])
984 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
986 if (likely(!inner)) {
987 SW_FLOW_KEY_PUT(match, eth.vlan.tpid, tpid, is_mask);
988 SW_FLOW_KEY_PUT(match, eth.vlan.tci, tci, is_mask);
990 SW_FLOW_KEY_PUT(match, eth.cvlan.tpid, tpid, is_mask);
991 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, tci, is_mask);
996 static int validate_vlan_from_nlattrs(const struct sw_flow_match *match,
997 u64 key_attrs, bool inner,
998 const struct nlattr **a, bool log)
1002 if (!((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
1003 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
1004 eth_type_vlan(nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE])))) {
1009 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
1010 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
1011 OVS_NLERR(log, "Invalid %s frame", (inner) ? "C-VLAN" : "VLAN");
1015 if (a[OVS_KEY_ATTR_VLAN])
1016 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1018 if (!(tci & htons(VLAN_CFI_MASK))) {
1020 OVS_NLERR(log, "%s TCI does not have VLAN_CFI_MASK bit set.",
1021 (inner) ? "C-VLAN" : "VLAN");
1023 } else if (nla_len(a[OVS_KEY_ATTR_ENCAP])) {
1024 /* Corner case for truncated VLAN header. */
1025 OVS_NLERR(log, "Truncated %s header has non-zero encap attribute.",
1026 (inner) ? "C-VLAN" : "VLAN");
1034 static int validate_vlan_mask_from_nlattrs(const struct sw_flow_match *match,
1035 u64 key_attrs, bool inner,
1036 const struct nlattr **a, bool log)
1040 bool encap_valid = !!(match->key->eth.vlan.tci &
1041 htons(VLAN_CFI_MASK));
1042 bool i_encap_valid = !!(match->key->eth.cvlan.tci &
1043 htons(VLAN_CFI_MASK));
1045 if (!(key_attrs & (1 << OVS_KEY_ATTR_ENCAP))) {
1050 if ((!inner && !encap_valid) || (inner && !i_encap_valid)) {
1051 OVS_NLERR(log, "Encap mask attribute is set for non-%s frame.",
1052 (inner) ? "C-VLAN" : "VLAN");
1056 if (a[OVS_KEY_ATTR_VLAN])
1057 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1059 if (a[OVS_KEY_ATTR_ETHERTYPE])
1060 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1062 if (tpid != htons(0xffff)) {
1063 OVS_NLERR(log, "Must have an exact match on %s TPID (mask=%x).",
1064 (inner) ? "C-VLAN" : "VLAN", ntohs(tpid));
1067 if (!(tci & htons(VLAN_CFI_MASK))) {
1068 OVS_NLERR(log, "%s TCI mask does not have exact match for VLAN_CFI_MASK bit.",
1069 (inner) ? "C-VLAN" : "VLAN");
1076 static int __parse_vlan_from_nlattrs(struct sw_flow_match *match,
1077 u64 *key_attrs, bool inner,
1078 const struct nlattr **a, bool is_mask,
1082 const struct nlattr *encap;
1085 err = validate_vlan_from_nlattrs(match, *key_attrs, inner,
1088 err = validate_vlan_mask_from_nlattrs(match, *key_attrs, inner,
1093 err = encode_vlan_from_nlattrs(match, a, is_mask, inner);
1097 *key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1098 *key_attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
1099 *key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1101 encap = a[OVS_KEY_ATTR_ENCAP];
1104 err = parse_flow_nlattrs(encap, a, key_attrs, log);
1106 err = parse_flow_mask_nlattrs(encap, a, key_attrs, log);
1111 static int parse_vlan_from_nlattrs(struct sw_flow_match *match,
1112 u64 *key_attrs, const struct nlattr **a,
1113 bool is_mask, bool log)
1116 bool encap_valid = false;
1118 err = __parse_vlan_from_nlattrs(match, key_attrs, false, a,
1123 encap_valid = !!(match->key->eth.vlan.tci & htons(VLAN_CFI_MASK));
1125 err = __parse_vlan_from_nlattrs(match, key_attrs, true, a,
1134 static int parse_eth_type_from_nlattrs(struct sw_flow_match *match,
1135 u64 *attrs, const struct nlattr **a,
1136 bool is_mask, bool log)
1140 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1142 /* Always exact match EtherType. */
1143 eth_type = htons(0xffff);
1144 } else if (!eth_proto_is_802_3(eth_type)) {
1145 OVS_NLERR(log, "EtherType %x is less than min %x",
1146 ntohs(eth_type), ETH_P_802_3_MIN);
1150 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
1151 *attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1155 static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
1156 u64 *attrs, const struct nlattr **a,
1157 bool is_mask, bool log)
1159 u8 mac_proto = MAC_PROTO_ETHERNET;
1161 if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
1162 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
1164 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
1165 *attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
1168 if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
1169 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
1171 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
1172 *attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
1175 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
1176 SW_FLOW_KEY_PUT(match, phy.priority,
1177 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
1178 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
1181 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
1182 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1185 in_port = 0xffffffff; /* Always exact match in_port. */
1186 } else if (in_port >= DP_MAX_PORTS) {
1187 OVS_NLERR(log, "Port %d exceeds max allowable %d",
1188 in_port, DP_MAX_PORTS);
1192 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
1193 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1194 } else if (!is_mask) {
1195 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
1198 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1199 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1201 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1202 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1204 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1205 if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1208 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1211 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
1212 ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
1213 u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
1215 if (ct_state & ~CT_SUPPORTED_MASK) {
1216 OVS_NLERR(log, "ct_state flags %08x unsupported",
1221 SW_FLOW_KEY_PUT(match, ct_state, ct_state, is_mask);
1222 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
1224 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
1225 ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
1226 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
1228 SW_FLOW_KEY_PUT(match, ct_zone, ct_zone, is_mask);
1229 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
1231 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
1232 ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
1233 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
1235 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
1236 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
1238 if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
1239 ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
1240 const struct ovs_key_ct_labels *cl;
1242 cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
1243 SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
1244 sizeof(*cl), is_mask);
1245 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
1247 if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)) {
1248 const struct ovs_key_ct_tuple_ipv4 *ct;
1250 ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4]);
1252 SW_FLOW_KEY_PUT(match, ipv4.ct_orig.src, ct->ipv4_src, is_mask);
1253 SW_FLOW_KEY_PUT(match, ipv4.ct_orig.dst, ct->ipv4_dst, is_mask);
1254 SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1255 SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1256 SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv4_proto, is_mask);
1257 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4);
1259 if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)) {
1260 const struct ovs_key_ct_tuple_ipv6 *ct;
1262 ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6]);
1264 SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.src, &ct->ipv6_src,
1265 sizeof(match->key->ipv6.ct_orig.src),
1267 SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.dst, &ct->ipv6_dst,
1268 sizeof(match->key->ipv6.ct_orig.dst),
1270 SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1271 SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1272 SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv6_proto, is_mask);
1273 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
1276 /* For layer 3 packets the Ethernet type is provided
1277 * and treated as metadata but no MAC addresses are provided.
1279 if (!(*attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
1280 (*attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)))
1281 mac_proto = MAC_PROTO_NONE;
1283 /* Always exact match mac_proto */
1284 SW_FLOW_KEY_PUT(match, mac_proto, is_mask ? 0xff : mac_proto, is_mask);
1286 if (mac_proto == MAC_PROTO_NONE)
1287 return parse_eth_type_from_nlattrs(match, attrs, a, is_mask,
1293 int nsh_hdr_from_nlattr(const struct nlattr *attr,
1294 struct nshhdr *nh, size_t size)
1302 /* validate_nsh has check this, so we needn't do duplicate check here
1304 if (size < NSH_BASE_HDR_LEN)
1307 nla_for_each_nested(a, attr, rem) {
1308 int type = nla_type(a);
1311 case OVS_NSH_KEY_ATTR_BASE: {
1312 const struct ovs_nsh_key_base *base = nla_data(a);
1314 flags = base->flags;
1317 nh->mdtype = base->mdtype;
1318 nh->path_hdr = base->path_hdr;
1321 case OVS_NSH_KEY_ATTR_MD1:
1323 if (mdlen > size - NSH_BASE_HDR_LEN)
1325 memcpy(&nh->md1, nla_data(a), mdlen);
1328 case OVS_NSH_KEY_ATTR_MD2:
1330 if (mdlen > size - NSH_BASE_HDR_LEN)
1332 memcpy(&nh->md2, nla_data(a), mdlen);
1340 /* nsh header length = NSH_BASE_HDR_LEN + mdlen */
1341 nh->ver_flags_ttl_len = 0;
1342 nsh_set_flags_ttl_len(nh, flags, ttl, NSH_BASE_HDR_LEN + mdlen);
1347 int nsh_key_from_nlattr(const struct nlattr *attr,
1348 struct ovs_key_nsh *nsh, struct ovs_key_nsh *nsh_mask)
1353 /* validate_nsh has check this, so we needn't do duplicate check here
1355 nla_for_each_nested(a, attr, rem) {
1356 int type = nla_type(a);
1359 case OVS_NSH_KEY_ATTR_BASE: {
1360 const struct ovs_nsh_key_base *base = nla_data(a);
1361 const struct ovs_nsh_key_base *base_mask = base + 1;
1364 nsh_mask->base = *base_mask;
1367 case OVS_NSH_KEY_ATTR_MD1: {
1368 const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1369 const struct ovs_nsh_key_md1 *md1_mask = md1 + 1;
1371 memcpy(nsh->context, md1->context, sizeof(*md1));
1372 memcpy(nsh_mask->context, md1_mask->context,
1376 case OVS_NSH_KEY_ATTR_MD2:
1377 /* Not supported yet */
1387 static int nsh_key_put_from_nlattr(const struct nlattr *attr,
1388 struct sw_flow_match *match, bool is_mask,
1389 bool is_push_nsh, bool log)
1393 bool has_base = false;
1394 bool has_md1 = false;
1395 bool has_md2 = false;
1399 if (WARN_ON(is_push_nsh && is_mask))
1402 nla_for_each_nested(a, attr, rem) {
1403 int type = nla_type(a);
1406 if (type > OVS_NSH_KEY_ATTR_MAX) {
1407 OVS_NLERR(log, "nsh attr %d is out of range max %d",
1408 type, OVS_NSH_KEY_ATTR_MAX);
1412 if (!check_attr_len(nla_len(a),
1413 ovs_nsh_key_attr_lens[type].len)) {
1416 "nsh attr %d has unexpected len %d expected %d",
1419 ovs_nsh_key_attr_lens[type].len
1425 case OVS_NSH_KEY_ATTR_BASE: {
1426 const struct ovs_nsh_key_base *base = nla_data(a);
1429 mdtype = base->mdtype;
1430 SW_FLOW_KEY_PUT(match, nsh.base.flags,
1431 base->flags, is_mask);
1432 SW_FLOW_KEY_PUT(match, nsh.base.ttl,
1433 base->ttl, is_mask);
1434 SW_FLOW_KEY_PUT(match, nsh.base.mdtype,
1435 base->mdtype, is_mask);
1436 SW_FLOW_KEY_PUT(match, nsh.base.np,
1438 SW_FLOW_KEY_PUT(match, nsh.base.path_hdr,
1439 base->path_hdr, is_mask);
1442 case OVS_NSH_KEY_ATTR_MD1: {
1443 const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1446 for (i = 0; i < NSH_MD1_CONTEXT_SIZE; i++)
1447 SW_FLOW_KEY_PUT(match, nsh.context[i],
1448 md1->context[i], is_mask);
1451 case OVS_NSH_KEY_ATTR_MD2:
1452 if (!is_push_nsh) /* Not supported MD type 2 yet */
1457 if (mdlen > NSH_CTX_HDRS_MAX_LEN || mdlen <= 0) {
1460 "Invalid MD length %d for MD type %d",
1468 OVS_NLERR(log, "Unknown nsh attribute %d",
1475 OVS_NLERR(log, "nsh attribute has %d unknown bytes.", rem);
1479 if (has_md1 && has_md2) {
1482 "invalid nsh attribute: md1 and md2 are exclusive."
1488 if ((has_md1 && mdtype != NSH_M_TYPE1) ||
1489 (has_md2 && mdtype != NSH_M_TYPE2)) {
1490 OVS_NLERR(1, "nsh attribute has unmatched MD type %d.",
1496 (!has_base || (!has_md1 && !has_md2))) {
1499 "push_nsh: missing base or metadata attributes"
1508 static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
1509 u64 attrs, const struct nlattr **a,
1510 bool is_mask, bool log)
1514 err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
1518 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1519 const struct ovs_key_ethernet *eth_key;
1521 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1522 SW_FLOW_KEY_MEMCPY(match, eth.src,
1523 eth_key->eth_src, ETH_ALEN, is_mask);
1524 SW_FLOW_KEY_MEMCPY(match, eth.dst,
1525 eth_key->eth_dst, ETH_ALEN, is_mask);
1526 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1528 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1529 /* VLAN attribute is always parsed before getting here since it
1530 * may occur multiple times.
1532 OVS_NLERR(log, "VLAN attribute unexpected.");
1536 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1537 err = parse_eth_type_from_nlattrs(match, &attrs, a, is_mask,
1541 } else if (!is_mask) {
1542 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1544 } else if (!match->key->eth.type) {
1545 OVS_NLERR(log, "Either Ethernet header or EtherType is required.");
1549 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1550 const struct ovs_key_ipv4 *ipv4_key;
1552 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1553 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
1554 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
1555 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
1558 SW_FLOW_KEY_PUT(match, ip.proto,
1559 ipv4_key->ipv4_proto, is_mask);
1560 SW_FLOW_KEY_PUT(match, ip.tos,
1561 ipv4_key->ipv4_tos, is_mask);
1562 SW_FLOW_KEY_PUT(match, ip.ttl,
1563 ipv4_key->ipv4_ttl, is_mask);
1564 SW_FLOW_KEY_PUT(match, ip.frag,
1565 ipv4_key->ipv4_frag, is_mask);
1566 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1567 ipv4_key->ipv4_src, is_mask);
1568 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1569 ipv4_key->ipv4_dst, is_mask);
1570 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1573 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1574 const struct ovs_key_ipv6 *ipv6_key;
1576 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1577 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
1578 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
1579 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
1583 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
1584 OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x)",
1585 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
1589 SW_FLOW_KEY_PUT(match, ipv6.label,
1590 ipv6_key->ipv6_label, is_mask);
1591 SW_FLOW_KEY_PUT(match, ip.proto,
1592 ipv6_key->ipv6_proto, is_mask);
1593 SW_FLOW_KEY_PUT(match, ip.tos,
1594 ipv6_key->ipv6_tclass, is_mask);
1595 SW_FLOW_KEY_PUT(match, ip.ttl,
1596 ipv6_key->ipv6_hlimit, is_mask);
1597 SW_FLOW_KEY_PUT(match, ip.frag,
1598 ipv6_key->ipv6_frag, is_mask);
1599 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1601 sizeof(match->key->ipv6.addr.src),
1603 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1605 sizeof(match->key->ipv6.addr.dst),
1608 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1611 if (attrs & (1ULL << OVS_KEY_ATTR_IPV6_EXTHDRS)) {
1612 const struct ovs_key_ipv6_exthdrs *ipv6_exthdrs_key;
1614 ipv6_exthdrs_key = nla_data(a[OVS_KEY_ATTR_IPV6_EXTHDRS]);
1616 SW_FLOW_KEY_PUT(match, ipv6.exthdrs,
1617 ipv6_exthdrs_key->hdrs, is_mask);
1619 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV6_EXTHDRS);
1622 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1623 const struct ovs_key_arp *arp_key;
1625 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1626 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1627 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
1632 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1633 arp_key->arp_sip, is_mask);
1634 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1635 arp_key->arp_tip, is_mask);
1636 SW_FLOW_KEY_PUT(match, ip.proto,
1637 ntohs(arp_key->arp_op), is_mask);
1638 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1639 arp_key->arp_sha, ETH_ALEN, is_mask);
1640 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1641 arp_key->arp_tha, ETH_ALEN, is_mask);
1643 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1646 if (attrs & (1 << OVS_KEY_ATTR_NSH)) {
1647 if (nsh_key_put_from_nlattr(a[OVS_KEY_ATTR_NSH], match,
1648 is_mask, false, log) < 0)
1650 attrs &= ~(1 << OVS_KEY_ATTR_NSH);
1653 if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1654 const struct ovs_key_mpls *mpls_key;
1656 u32 label_count, label_count_mask, i;
1658 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1659 hdr_len = nla_len(a[OVS_KEY_ATTR_MPLS]);
1660 label_count = hdr_len / sizeof(struct ovs_key_mpls);
1662 if (label_count == 0 || label_count > MPLS_LABEL_DEPTH ||
1663 hdr_len % sizeof(struct ovs_key_mpls))
1666 label_count_mask = GENMASK(label_count - 1, 0);
1668 for (i = 0 ; i < label_count; i++)
1669 SW_FLOW_KEY_PUT(match, mpls.lse[i],
1670 mpls_key[i].mpls_lse, is_mask);
1672 SW_FLOW_KEY_PUT(match, mpls.num_labels_mask,
1673 label_count_mask, is_mask);
1675 attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1678 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1679 const struct ovs_key_tcp *tcp_key;
1681 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1682 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1683 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1684 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1687 if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1688 SW_FLOW_KEY_PUT(match, tp.flags,
1689 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1691 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1694 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1695 const struct ovs_key_udp *udp_key;
1697 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1698 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1699 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1700 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1703 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1704 const struct ovs_key_sctp *sctp_key;
1706 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1707 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1708 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1709 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1712 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1713 const struct ovs_key_icmp *icmp_key;
1715 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1716 SW_FLOW_KEY_PUT(match, tp.src,
1717 htons(icmp_key->icmp_type), is_mask);
1718 SW_FLOW_KEY_PUT(match, tp.dst,
1719 htons(icmp_key->icmp_code), is_mask);
1720 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1723 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1724 const struct ovs_key_icmpv6 *icmpv6_key;
1726 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1727 SW_FLOW_KEY_PUT(match, tp.src,
1728 htons(icmpv6_key->icmpv6_type), is_mask);
1729 SW_FLOW_KEY_PUT(match, tp.dst,
1730 htons(icmpv6_key->icmpv6_code), is_mask);
1731 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1734 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1735 const struct ovs_key_nd *nd_key;
1737 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1738 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1740 sizeof(match->key->ipv6.nd.target),
1742 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1743 nd_key->nd_sll, ETH_ALEN, is_mask);
1744 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1745 nd_key->nd_tll, ETH_ALEN, is_mask);
1746 attrs &= ~(1 << OVS_KEY_ATTR_ND);
1750 OVS_NLERR(log, "Unknown key attributes %llx",
1751 (unsigned long long)attrs);
1758 static void nlattr_set(struct nlattr *attr, u8 val,
1759 const struct ovs_len_tbl *tbl)
1764 /* The nlattr stream should already have been validated */
1765 nla_for_each_nested(nla, attr, rem) {
1766 if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED)
1767 nlattr_set(nla, val, tbl[nla_type(nla)].next ? : tbl);
1769 memset(nla_data(nla), val, nla_len(nla));
1771 if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1772 *(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1776 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1778 nlattr_set(attr, val, ovs_key_lens);
1782 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1783 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1784 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1785 * does not include any don't care bit.
1786 * @net: Used to determine per-namespace field support.
1787 * @match: receives the extracted flow match information.
1788 * @nla_key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1789 * sequence. The fields should of the packet that triggered the creation
1791 * @nla_mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_*
1792 * Netlink attribute specifies the mask field of the wildcarded flow.
1793 * @log: Boolean to allow kernel error logging. Normally true, but when
1794 * probing for feature compatibility this should be passed in as false to
1795 * suppress unnecessary error logging.
1797 int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1798 const struct nlattr *nla_key,
1799 const struct nlattr *nla_mask,
1802 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1803 struct nlattr *newmask = NULL;
1808 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1812 err = parse_vlan_from_nlattrs(match, &key_attrs, a, false, log);
1816 err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1822 /* Create an exact match mask. We need to set to 0xff
1823 * all the 'match->mask' fields that have been touched
1824 * in 'match->key'. We cannot simply memset
1825 * 'match->mask', because padding bytes and fields not
1826 * specified in 'match->key' should be left to 0.
1827 * Instead, we use a stream of netlink attributes,
1828 * copied from 'key' and set to 0xff.
1829 * ovs_key_from_nlattrs() will take care of filling
1830 * 'match->mask' appropriately.
1832 newmask = kmemdup(nla_key,
1833 nla_total_size(nla_len(nla_key)),
1838 mask_set_nlattr(newmask, 0xff);
1840 /* The userspace does not send tunnel attributes that
1841 * are 0, but we should not wildcard them nonetheless.
1843 if (match->key->tun_proto)
1844 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1850 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1854 /* Always match on tci. */
1855 SW_FLOW_KEY_PUT(match, eth.vlan.tci, htons(0xffff), true);
1856 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, htons(0xffff), true);
1858 err = parse_vlan_from_nlattrs(match, &mask_attrs, a, true, log);
1862 err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1868 if (!match_validate(match, key_attrs, mask_attrs, log))
1876 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1883 len = nla_len(attr);
1884 if (len < 1 || len > MAX_UFID_LENGTH) {
1885 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1886 nla_len(attr), MAX_UFID_LENGTH);
1893 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1894 * or false otherwise.
1896 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1899 sfid->ufid_len = get_ufid_len(attr, log);
1901 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1903 return sfid->ufid_len;
1906 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1907 const struct sw_flow_key *key, bool log)
1909 struct sw_flow_key *new_key;
1911 if (ovs_nla_get_ufid(sfid, ufid, log))
1914 /* If UFID was not provided, use unmasked key. */
1915 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1918 memcpy(new_key, key, sizeof(*key));
1919 sfid->unmasked_key = new_key;
1924 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1926 return attr ? nla_get_u32(attr) : 0;
1930 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1931 * @net: Network namespace.
1932 * @key: Receives extracted in_port, priority, tun_key, skb_mark and conntrack
1934 * @a: Array of netlink attributes holding parsed %OVS_KEY_ATTR_* Netlink
1936 * @attrs: Bit mask for the netlink attributes included in @a.
1937 * @log: Boolean to allow kernel error logging. Normally true, but when
1938 * probing for feature compatibility this should be passed in as false to
1939 * suppress unnecessary error logging.
1941 * This parses a series of Netlink attributes that form a flow key, which must
1942 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1943 * get the metadata, that is, the parts of the flow key that cannot be
1944 * extracted from the packet itself.
1946 * This must be called before the packet key fields are filled in 'key'.
1949 int ovs_nla_get_flow_metadata(struct net *net,
1950 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1],
1951 u64 attrs, struct sw_flow_key *key, bool log)
1953 struct sw_flow_match match;
1955 memset(&match, 0, sizeof(match));
1960 key->ct_orig_proto = 0;
1961 memset(&key->ct, 0, sizeof(key->ct));
1962 memset(&key->ipv4.ct_orig, 0, sizeof(key->ipv4.ct_orig));
1963 memset(&key->ipv6.ct_orig, 0, sizeof(key->ipv6.ct_orig));
1965 key->phy.in_port = DP_MAX_PORTS;
1967 return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1970 static int ovs_nla_put_vlan(struct sk_buff *skb, const struct vlan_head *vh,
1973 __be16 eth_type = !is_mask ? vh->tpid : htons(0xffff);
1975 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1976 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, vh->tci))
1981 static int nsh_key_to_nlattr(const struct ovs_key_nsh *nsh, bool is_mask,
1982 struct sk_buff *skb)
1984 struct nlattr *start;
1986 start = nla_nest_start_noflag(skb, OVS_KEY_ATTR_NSH);
1990 if (nla_put(skb, OVS_NSH_KEY_ATTR_BASE, sizeof(nsh->base), &nsh->base))
1991 goto nla_put_failure;
1993 if (is_mask || nsh->base.mdtype == NSH_M_TYPE1) {
1994 if (nla_put(skb, OVS_NSH_KEY_ATTR_MD1,
1995 sizeof(nsh->context), nsh->context))
1996 goto nla_put_failure;
1999 /* Don't support MD type 2 yet */
2001 nla_nest_end(skb, start);
2009 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
2010 const struct sw_flow_key *output, bool is_mask,
2011 struct sk_buff *skb)
2013 struct ovs_key_ethernet *eth_key;
2015 struct nlattr *encap = NULL;
2016 struct nlattr *in_encap = NULL;
2018 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
2019 goto nla_put_failure;
2021 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
2022 goto nla_put_failure;
2024 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
2025 goto nla_put_failure;
2027 if ((swkey->tun_proto || is_mask)) {
2028 const void *opts = NULL;
2030 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
2031 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
2033 if (ip_tun_to_nlattr(skb, &output->tun_key, opts,
2034 swkey->tun_opts_len, swkey->tun_proto, 0))
2035 goto nla_put_failure;
2038 if (swkey->phy.in_port == DP_MAX_PORTS) {
2039 if (is_mask && (output->phy.in_port == 0xffff))
2040 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
2041 goto nla_put_failure;
2044 upper_u16 = !is_mask ? 0 : 0xffff;
2046 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
2047 (upper_u16 << 16) | output->phy.in_port))
2048 goto nla_put_failure;
2051 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
2052 goto nla_put_failure;
2054 if (ovs_ct_put_key(swkey, output, skb))
2055 goto nla_put_failure;
2057 if (ovs_key_mac_proto(swkey) == MAC_PROTO_ETHERNET) {
2058 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
2060 goto nla_put_failure;
2062 eth_key = nla_data(nla);
2063 ether_addr_copy(eth_key->eth_src, output->eth.src);
2064 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
2066 if (swkey->eth.vlan.tci || eth_type_vlan(swkey->eth.type)) {
2067 if (ovs_nla_put_vlan(skb, &output->eth.vlan, is_mask))
2068 goto nla_put_failure;
2069 encap = nla_nest_start_noflag(skb, OVS_KEY_ATTR_ENCAP);
2070 if (!swkey->eth.vlan.tci)
2073 if (swkey->eth.cvlan.tci || eth_type_vlan(swkey->eth.type)) {
2074 if (ovs_nla_put_vlan(skb, &output->eth.cvlan, is_mask))
2075 goto nla_put_failure;
2076 in_encap = nla_nest_start_noflag(skb,
2077 OVS_KEY_ATTR_ENCAP);
2078 if (!swkey->eth.cvlan.tci)
2083 if (swkey->eth.type == htons(ETH_P_802_2)) {
2085 * Ethertype 802.2 is represented in the netlink with omitted
2086 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
2087 * 0xffff in the mask attribute. Ethertype can also
2090 if (is_mask && output->eth.type)
2091 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
2093 goto nla_put_failure;
2098 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
2099 goto nla_put_failure;
2101 if (eth_type_vlan(swkey->eth.type)) {
2102 /* There are 3 VLAN tags, we don't know anything about the rest
2103 * of the packet, so truncate here.
2105 WARN_ON_ONCE(!(encap && in_encap));
2109 if (swkey->eth.type == htons(ETH_P_IP)) {
2110 struct ovs_key_ipv4 *ipv4_key;
2112 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
2114 goto nla_put_failure;
2115 ipv4_key = nla_data(nla);
2116 ipv4_key->ipv4_src = output->ipv4.addr.src;
2117 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
2118 ipv4_key->ipv4_proto = output->ip.proto;
2119 ipv4_key->ipv4_tos = output->ip.tos;
2120 ipv4_key->ipv4_ttl = output->ip.ttl;
2121 ipv4_key->ipv4_frag = output->ip.frag;
2122 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
2123 struct ovs_key_ipv6 *ipv6_key;
2124 struct ovs_key_ipv6_exthdrs *ipv6_exthdrs_key;
2126 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
2128 goto nla_put_failure;
2129 ipv6_key = nla_data(nla);
2130 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
2131 sizeof(ipv6_key->ipv6_src));
2132 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
2133 sizeof(ipv6_key->ipv6_dst));
2134 ipv6_key->ipv6_label = output->ipv6.label;
2135 ipv6_key->ipv6_proto = output->ip.proto;
2136 ipv6_key->ipv6_tclass = output->ip.tos;
2137 ipv6_key->ipv6_hlimit = output->ip.ttl;
2138 ipv6_key->ipv6_frag = output->ip.frag;
2140 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6_EXTHDRS,
2141 sizeof(*ipv6_exthdrs_key));
2143 goto nla_put_failure;
2144 ipv6_exthdrs_key = nla_data(nla);
2145 ipv6_exthdrs_key->hdrs = output->ipv6.exthdrs;
2146 } else if (swkey->eth.type == htons(ETH_P_NSH)) {
2147 if (nsh_key_to_nlattr(&output->nsh, is_mask, skb))
2148 goto nla_put_failure;
2149 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
2150 swkey->eth.type == htons(ETH_P_RARP)) {
2151 struct ovs_key_arp *arp_key;
2153 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
2155 goto nla_put_failure;
2156 arp_key = nla_data(nla);
2157 memset(arp_key, 0, sizeof(struct ovs_key_arp));
2158 arp_key->arp_sip = output->ipv4.addr.src;
2159 arp_key->arp_tip = output->ipv4.addr.dst;
2160 arp_key->arp_op = htons(output->ip.proto);
2161 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
2162 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
2163 } else if (eth_p_mpls(swkey->eth.type)) {
2165 struct ovs_key_mpls *mpls_key;
2167 num_labels = hweight_long(output->mpls.num_labels_mask);
2168 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS,
2169 num_labels * sizeof(*mpls_key));
2171 goto nla_put_failure;
2173 mpls_key = nla_data(nla);
2174 for (i = 0; i < num_labels; i++)
2175 mpls_key[i].mpls_lse = output->mpls.lse[i];
2178 if ((swkey->eth.type == htons(ETH_P_IP) ||
2179 swkey->eth.type == htons(ETH_P_IPV6)) &&
2180 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
2182 if (swkey->ip.proto == IPPROTO_TCP) {
2183 struct ovs_key_tcp *tcp_key;
2185 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
2187 goto nla_put_failure;
2188 tcp_key = nla_data(nla);
2189 tcp_key->tcp_src = output->tp.src;
2190 tcp_key->tcp_dst = output->tp.dst;
2191 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
2193 goto nla_put_failure;
2194 } else if (swkey->ip.proto == IPPROTO_UDP) {
2195 struct ovs_key_udp *udp_key;
2197 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
2199 goto nla_put_failure;
2200 udp_key = nla_data(nla);
2201 udp_key->udp_src = output->tp.src;
2202 udp_key->udp_dst = output->tp.dst;
2203 } else if (swkey->ip.proto == IPPROTO_SCTP) {
2204 struct ovs_key_sctp *sctp_key;
2206 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
2208 goto nla_put_failure;
2209 sctp_key = nla_data(nla);
2210 sctp_key->sctp_src = output->tp.src;
2211 sctp_key->sctp_dst = output->tp.dst;
2212 } else if (swkey->eth.type == htons(ETH_P_IP) &&
2213 swkey->ip.proto == IPPROTO_ICMP) {
2214 struct ovs_key_icmp *icmp_key;
2216 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
2218 goto nla_put_failure;
2219 icmp_key = nla_data(nla);
2220 icmp_key->icmp_type = ntohs(output->tp.src);
2221 icmp_key->icmp_code = ntohs(output->tp.dst);
2222 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
2223 swkey->ip.proto == IPPROTO_ICMPV6) {
2224 struct ovs_key_icmpv6 *icmpv6_key;
2226 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
2227 sizeof(*icmpv6_key));
2229 goto nla_put_failure;
2230 icmpv6_key = nla_data(nla);
2231 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
2232 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
2234 if (swkey->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
2235 swkey->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
2236 struct ovs_key_nd *nd_key;
2238 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
2240 goto nla_put_failure;
2241 nd_key = nla_data(nla);
2242 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
2243 sizeof(nd_key->nd_target));
2244 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
2245 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
2252 nla_nest_end(skb, in_encap);
2254 nla_nest_end(skb, encap);
2262 int ovs_nla_put_key(const struct sw_flow_key *swkey,
2263 const struct sw_flow_key *output, int attr, bool is_mask,
2264 struct sk_buff *skb)
2269 nla = nla_nest_start_noflag(skb, attr);
2272 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
2275 nla_nest_end(skb, nla);
2280 /* Called with ovs_mutex or RCU read lock. */
2281 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
2283 if (ovs_identifier_is_ufid(&flow->id))
2284 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
2287 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
2288 OVS_FLOW_ATTR_KEY, false, skb);
2291 /* Called with ovs_mutex or RCU read lock. */
2292 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
2294 return ovs_nla_put_key(&flow->key, &flow->key,
2295 OVS_FLOW_ATTR_KEY, false, skb);
2298 /* Called with ovs_mutex or RCU read lock. */
2299 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
2301 return ovs_nla_put_key(&flow->key, &flow->mask->key,
2302 OVS_FLOW_ATTR_MASK, true, skb);
2305 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
2307 static struct sw_flow_actions *nla_alloc_flow_actions(int size)
2309 struct sw_flow_actions *sfa;
2311 WARN_ON_ONCE(size > MAX_ACTIONS_BUFSIZE);
2313 sfa = kmalloc(kmalloc_size_roundup(sizeof(*sfa) + size), GFP_KERNEL);
2315 return ERR_PTR(-ENOMEM);
2317 sfa->actions_len = 0;
2321 static void ovs_nla_free_nested_actions(const struct nlattr *actions, int len);
2323 static void ovs_nla_free_check_pkt_len_action(const struct nlattr *action)
2325 const struct nlattr *a;
2328 nla_for_each_nested(a, action, rem) {
2329 switch (nla_type(a)) {
2330 case OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL:
2331 case OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER:
2332 ovs_nla_free_nested_actions(nla_data(a), nla_len(a));
2338 static void ovs_nla_free_clone_action(const struct nlattr *action)
2340 const struct nlattr *a = nla_data(action);
2341 int rem = nla_len(action);
2343 switch (nla_type(a)) {
2344 case OVS_CLONE_ATTR_EXEC:
2345 /* The real list of actions follows this attribute. */
2346 a = nla_next(a, &rem);
2347 ovs_nla_free_nested_actions(a, rem);
2352 static void ovs_nla_free_dec_ttl_action(const struct nlattr *action)
2354 const struct nlattr *a = nla_data(action);
2356 switch (nla_type(a)) {
2357 case OVS_DEC_TTL_ATTR_ACTION:
2358 ovs_nla_free_nested_actions(nla_data(a), nla_len(a));
2363 static void ovs_nla_free_sample_action(const struct nlattr *action)
2365 const struct nlattr *a = nla_data(action);
2366 int rem = nla_len(action);
2368 switch (nla_type(a)) {
2369 case OVS_SAMPLE_ATTR_ARG:
2370 /* The real list of actions follows this attribute. */
2371 a = nla_next(a, &rem);
2372 ovs_nla_free_nested_actions(a, rem);
2377 static void ovs_nla_free_set_action(const struct nlattr *a)
2379 const struct nlattr *ovs_key = nla_data(a);
2380 struct ovs_tunnel_info *ovs_tun;
2382 switch (nla_type(ovs_key)) {
2383 case OVS_KEY_ATTR_TUNNEL_INFO:
2384 ovs_tun = nla_data(ovs_key);
2385 dst_release((struct dst_entry *)ovs_tun->tun_dst);
2390 static void ovs_nla_free_nested_actions(const struct nlattr *actions, int len)
2392 const struct nlattr *a;
2395 /* Whenever new actions are added, the need to update this
2396 * function should be considered.
2398 BUILD_BUG_ON(OVS_ACTION_ATTR_MAX != 23);
2403 nla_for_each_attr(a, actions, len, rem) {
2404 switch (nla_type(a)) {
2405 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
2406 ovs_nla_free_check_pkt_len_action(a);
2409 case OVS_ACTION_ATTR_CLONE:
2410 ovs_nla_free_clone_action(a);
2413 case OVS_ACTION_ATTR_CT:
2414 ovs_ct_free_action(a);
2417 case OVS_ACTION_ATTR_DEC_TTL:
2418 ovs_nla_free_dec_ttl_action(a);
2421 case OVS_ACTION_ATTR_SAMPLE:
2422 ovs_nla_free_sample_action(a);
2425 case OVS_ACTION_ATTR_SET:
2426 ovs_nla_free_set_action(a);
2432 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
2437 ovs_nla_free_nested_actions(sf_acts->actions, sf_acts->actions_len);
2441 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
2443 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
2446 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
2447 * The caller must hold rcu_read_lock for this to be sensible. */
2448 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
2450 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
2453 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
2454 int attr_len, bool log)
2457 struct sw_flow_actions *acts;
2459 size_t req_size = NLA_ALIGN(attr_len);
2460 int next_offset = offsetof(struct sw_flow_actions, actions) +
2461 (*sfa)->actions_len;
2463 if (req_size <= (ksize(*sfa) - next_offset))
2466 new_acts_size = max(next_offset + req_size, ksize(*sfa) * 2);
2468 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
2469 if ((next_offset + req_size) > MAX_ACTIONS_BUFSIZE) {
2470 OVS_NLERR(log, "Flow action size exceeds max %u",
2471 MAX_ACTIONS_BUFSIZE);
2472 return ERR_PTR(-EMSGSIZE);
2474 new_acts_size = MAX_ACTIONS_BUFSIZE;
2477 acts = nla_alloc_flow_actions(new_acts_size);
2479 return (void *)acts;
2481 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
2482 acts->actions_len = (*sfa)->actions_len;
2483 acts->orig_len = (*sfa)->orig_len;
2488 (*sfa)->actions_len += req_size;
2489 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
2492 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
2493 int attrtype, void *data, int len, bool log)
2497 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
2501 a->nla_type = attrtype;
2502 a->nla_len = nla_attr_size(len);
2505 memcpy(nla_data(a), data, len);
2506 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
2511 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
2516 a = __add_action(sfa, attrtype, data, len, log);
2518 return PTR_ERR_OR_ZERO(a);
2521 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
2522 int attrtype, bool log)
2524 int used = (*sfa)->actions_len;
2527 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
2534 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
2537 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
2540 a->nla_len = sfa->actions_len - st_offset;
2543 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2544 const struct sw_flow_key *key,
2545 struct sw_flow_actions **sfa,
2546 __be16 eth_type, __be16 vlan_tci,
2547 u32 mpls_label_count, bool log,
2550 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
2551 const struct sw_flow_key *key,
2552 struct sw_flow_actions **sfa,
2553 __be16 eth_type, __be16 vlan_tci,
2554 u32 mpls_label_count, bool log, bool last,
2557 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
2558 const struct nlattr *probability, *actions;
2559 const struct nlattr *a;
2560 int rem, start, err;
2561 struct sample_arg arg;
2563 memset(attrs, 0, sizeof(attrs));
2564 nla_for_each_nested(a, attr, rem) {
2565 int type = nla_type(a);
2566 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
2573 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
2574 if (!probability || nla_len(probability) != sizeof(u32))
2577 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
2578 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
2581 /* validation done, copy sample action. */
2582 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
2586 /* When both skb and flow may be changed, put the sample
2587 * into a deferred fifo. On the other hand, if only skb
2588 * may be modified, the actions can be executed in place.
2590 * Do this analysis at the flow installation time.
2591 * Set 'clone_action->exec' to true if the actions can be
2592 * executed without being deferred.
2594 * If the sample is the last action, it can always be excuted
2595 * rather than deferred.
2597 arg.exec = last || !actions_may_change_flow(actions);
2598 arg.probability = nla_get_u32(probability);
2600 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_ARG, &arg, sizeof(arg),
2605 err = __ovs_nla_copy_actions(net, actions, key, sfa,
2606 eth_type, vlan_tci, mpls_label_count, log,
2612 add_nested_action_end(*sfa, start);
2617 static int validate_and_copy_dec_ttl(struct net *net,
2618 const struct nlattr *attr,
2619 const struct sw_flow_key *key,
2620 struct sw_flow_actions **sfa,
2621 __be16 eth_type, __be16 vlan_tci,
2622 u32 mpls_label_count, bool log,
2625 const struct nlattr *attrs[OVS_DEC_TTL_ATTR_MAX + 1];
2626 int start, action_start, err, rem;
2627 const struct nlattr *a, *actions;
2629 memset(attrs, 0, sizeof(attrs));
2630 nla_for_each_nested(a, attr, rem) {
2631 int type = nla_type(a);
2633 /* Ignore unknown attributes to be future proof. */
2634 if (type > OVS_DEC_TTL_ATTR_MAX)
2637 if (!type || attrs[type]) {
2638 OVS_NLERR(log, "Duplicate or invalid key (type %d).",
2647 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
2651 actions = attrs[OVS_DEC_TTL_ATTR_ACTION];
2652 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN)) {
2653 OVS_NLERR(log, "Missing valid actions attribute.");
2657 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_DEC_TTL, log);
2661 action_start = add_nested_action_start(sfa, OVS_DEC_TTL_ATTR_ACTION, log);
2662 if (action_start < 0)
2663 return action_start;
2665 err = __ovs_nla_copy_actions(net, actions, key, sfa, eth_type,
2666 vlan_tci, mpls_label_count, log,
2671 add_nested_action_end(*sfa, action_start);
2672 add_nested_action_end(*sfa, start);
2676 static int validate_and_copy_clone(struct net *net,
2677 const struct nlattr *attr,
2678 const struct sw_flow_key *key,
2679 struct sw_flow_actions **sfa,
2680 __be16 eth_type, __be16 vlan_tci,
2681 u32 mpls_label_count, bool log, bool last,
2687 if (nla_len(attr) && nla_len(attr) < NLA_HDRLEN)
2690 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CLONE, log);
2694 exec = last || !actions_may_change_flow(attr);
2696 err = ovs_nla_add_action(sfa, OVS_CLONE_ATTR_EXEC, &exec,
2701 err = __ovs_nla_copy_actions(net, attr, key, sfa,
2702 eth_type, vlan_tci, mpls_label_count, log,
2707 add_nested_action_end(*sfa, start);
2712 void ovs_match_init(struct sw_flow_match *match,
2713 struct sw_flow_key *key,
2715 struct sw_flow_mask *mask)
2717 memset(match, 0, sizeof(*match));
2722 memset(key, 0, sizeof(*key));
2725 memset(&mask->key, 0, sizeof(mask->key));
2726 mask->range.start = mask->range.end = 0;
2730 static int validate_geneve_opts(struct sw_flow_key *key)
2732 struct geneve_opt *option;
2733 int opts_len = key->tun_opts_len;
2734 bool crit_opt = false;
2736 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2737 while (opts_len > 0) {
2740 if (opts_len < sizeof(*option))
2743 len = sizeof(*option) + option->length * 4;
2747 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2749 option = (struct geneve_opt *)((u8 *)option + len);
2753 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
2758 static int validate_and_copy_set_tun(const struct nlattr *attr,
2759 struct sw_flow_actions **sfa, bool log)
2761 struct sw_flow_match match;
2762 struct sw_flow_key key;
2763 struct metadata_dst *tun_dst;
2764 struct ip_tunnel_info *tun_info;
2765 struct ovs_tunnel_info *ovs_tun;
2767 int err = 0, start, opts_type;
2768 __be16 dst_opt_type;
2771 ovs_match_init(&match, &key, true, NULL);
2772 opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2776 if (key.tun_opts_len) {
2777 switch (opts_type) {
2778 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2779 err = validate_geneve_opts(&key);
2782 dst_opt_type = TUNNEL_GENEVE_OPT;
2784 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
2785 dst_opt_type = TUNNEL_VXLAN_OPT;
2787 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
2788 dst_opt_type = TUNNEL_ERSPAN_OPT;
2793 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2797 tun_dst = metadata_dst_alloc(key.tun_opts_len, METADATA_IP_TUNNEL,
2803 err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2805 dst_release((struct dst_entry *)tun_dst);
2809 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2810 sizeof(*ovs_tun), log);
2812 dst_release((struct dst_entry *)tun_dst);
2816 ovs_tun = nla_data(a);
2817 ovs_tun->tun_dst = tun_dst;
2819 tun_info = &tun_dst->u.tun_info;
2820 tun_info->mode = IP_TUNNEL_INFO_TX;
2821 if (key.tun_proto == AF_INET6)
2822 tun_info->mode |= IP_TUNNEL_INFO_IPV6;
2823 else if (key.tun_proto == AF_INET && key.tun_key.u.ipv4.dst == 0)
2824 tun_info->mode |= IP_TUNNEL_INFO_BRIDGE;
2825 tun_info->key = key.tun_key;
2827 /* We need to store the options in the action itself since
2828 * everything else will go away after flow setup. We can append
2829 * it to tun_info and then point there.
2831 ip_tunnel_info_opts_set(tun_info,
2832 TUN_METADATA_OPTS(&key, key.tun_opts_len),
2833 key.tun_opts_len, dst_opt_type);
2834 add_nested_action_end(*sfa, start);
2839 static bool validate_nsh(const struct nlattr *attr, bool is_mask,
2840 bool is_push_nsh, bool log)
2842 struct sw_flow_match match;
2843 struct sw_flow_key key;
2846 ovs_match_init(&match, &key, true, NULL);
2847 ret = nsh_key_put_from_nlattr(attr, &match, is_mask,
2852 /* Return false if there are any non-masked bits set.
2853 * Mask follows data immediately, before any netlink padding.
2855 static bool validate_masked(u8 *data, int len)
2857 u8 *mask = data + len;
2860 if (*data++ & ~*mask++)
2866 static int validate_set(const struct nlattr *a,
2867 const struct sw_flow_key *flow_key,
2868 struct sw_flow_actions **sfa, bool *skip_copy,
2869 u8 mac_proto, __be16 eth_type, bool masked, bool log)
2871 const struct nlattr *ovs_key = nla_data(a);
2872 int key_type = nla_type(ovs_key);
2875 /* There can be only one key in a action */
2876 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2879 key_len = nla_len(ovs_key);
2883 if (key_type > OVS_KEY_ATTR_MAX ||
2884 !check_attr_len(key_len, ovs_key_lens[key_type].len))
2887 if (masked && !validate_masked(nla_data(ovs_key), key_len))
2891 case OVS_KEY_ATTR_PRIORITY:
2892 case OVS_KEY_ATTR_SKB_MARK:
2893 case OVS_KEY_ATTR_CT_MARK:
2894 case OVS_KEY_ATTR_CT_LABELS:
2897 case OVS_KEY_ATTR_ETHERNET:
2898 if (mac_proto != MAC_PROTO_ETHERNET)
2902 case OVS_KEY_ATTR_TUNNEL: {
2906 return -EINVAL; /* Masked tunnel set not supported. */
2909 err = validate_and_copy_set_tun(a, sfa, log);
2914 case OVS_KEY_ATTR_IPV4: {
2915 const struct ovs_key_ipv4 *ipv4_key;
2917 if (eth_type != htons(ETH_P_IP))
2920 ipv4_key = nla_data(ovs_key);
2923 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2925 /* Non-writeable fields. */
2926 if (mask->ipv4_proto || mask->ipv4_frag)
2929 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2932 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2937 case OVS_KEY_ATTR_IPV6: {
2938 const struct ovs_key_ipv6 *ipv6_key;
2940 if (eth_type != htons(ETH_P_IPV6))
2943 ipv6_key = nla_data(ovs_key);
2946 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2948 /* Non-writeable fields. */
2949 if (mask->ipv6_proto || mask->ipv6_frag)
2952 /* Invalid bits in the flow label mask? */
2953 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2956 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2959 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2962 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2967 case OVS_KEY_ATTR_TCP:
2968 if ((eth_type != htons(ETH_P_IP) &&
2969 eth_type != htons(ETH_P_IPV6)) ||
2970 flow_key->ip.proto != IPPROTO_TCP)
2975 case OVS_KEY_ATTR_UDP:
2976 if ((eth_type != htons(ETH_P_IP) &&
2977 eth_type != htons(ETH_P_IPV6)) ||
2978 flow_key->ip.proto != IPPROTO_UDP)
2983 case OVS_KEY_ATTR_MPLS:
2984 if (!eth_p_mpls(eth_type))
2988 case OVS_KEY_ATTR_SCTP:
2989 if ((eth_type != htons(ETH_P_IP) &&
2990 eth_type != htons(ETH_P_IPV6)) ||
2991 flow_key->ip.proto != IPPROTO_SCTP)
2996 case OVS_KEY_ATTR_NSH:
2997 if (eth_type != htons(ETH_P_NSH))
2999 if (!validate_nsh(nla_data(a), masked, false, log))
3007 /* Convert non-masked non-tunnel set actions to masked set actions. */
3008 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
3009 int start, len = key_len * 2;
3014 start = add_nested_action_start(sfa,
3015 OVS_ACTION_ATTR_SET_TO_MASKED,
3020 at = __add_action(sfa, key_type, NULL, len, log);
3024 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
3025 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
3026 /* Clear non-writeable bits from otherwise writeable fields. */
3027 if (key_type == OVS_KEY_ATTR_IPV6) {
3028 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
3030 mask->ipv6_label &= htonl(0x000FFFFF);
3032 add_nested_action_end(*sfa, start);
3038 static int validate_userspace(const struct nlattr *attr)
3040 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
3041 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
3042 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
3043 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
3045 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
3048 error = nla_parse_nested_deprecated(a, OVS_USERSPACE_ATTR_MAX, attr,
3049 userspace_policy, NULL);
3053 if (!a[OVS_USERSPACE_ATTR_PID] ||
3054 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
3060 static const struct nla_policy cpl_policy[OVS_CHECK_PKT_LEN_ATTR_MAX + 1] = {
3061 [OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] = {.type = NLA_U16 },
3062 [OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER] = {.type = NLA_NESTED },
3063 [OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL] = {.type = NLA_NESTED },
3066 static int validate_and_copy_check_pkt_len(struct net *net,
3067 const struct nlattr *attr,
3068 const struct sw_flow_key *key,
3069 struct sw_flow_actions **sfa,
3070 __be16 eth_type, __be16 vlan_tci,
3071 u32 mpls_label_count,
3072 bool log, bool last, u32 depth)
3074 const struct nlattr *acts_if_greater, *acts_if_lesser_eq;
3075 struct nlattr *a[OVS_CHECK_PKT_LEN_ATTR_MAX + 1];
3076 struct check_pkt_len_arg arg;
3077 int nested_acts_start;
3080 err = nla_parse_deprecated_strict(a, OVS_CHECK_PKT_LEN_ATTR_MAX,
3081 nla_data(attr), nla_len(attr),
3086 if (!a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] ||
3087 !nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]))
3090 acts_if_lesser_eq = a[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL];
3091 acts_if_greater = a[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER];
3093 /* Both the nested action should be present. */
3094 if (!acts_if_greater || !acts_if_lesser_eq)
3097 /* validation done, copy the nested actions. */
3098 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CHECK_PKT_LEN,
3103 arg.pkt_len = nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]);
3104 arg.exec_for_lesser_equal =
3105 last || !actions_may_change_flow(acts_if_lesser_eq);
3106 arg.exec_for_greater =
3107 last || !actions_may_change_flow(acts_if_greater);
3109 err = ovs_nla_add_action(sfa, OVS_CHECK_PKT_LEN_ATTR_ARG, &arg,
3114 nested_acts_start = add_nested_action_start(sfa,
3115 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL, log);
3116 if (nested_acts_start < 0)
3117 return nested_acts_start;
3119 err = __ovs_nla_copy_actions(net, acts_if_lesser_eq, key, sfa,
3120 eth_type, vlan_tci, mpls_label_count, log,
3126 add_nested_action_end(*sfa, nested_acts_start);
3128 nested_acts_start = add_nested_action_start(sfa,
3129 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER, log);
3130 if (nested_acts_start < 0)
3131 return nested_acts_start;
3133 err = __ovs_nla_copy_actions(net, acts_if_greater, key, sfa,
3134 eth_type, vlan_tci, mpls_label_count, log,
3140 add_nested_action_end(*sfa, nested_acts_start);
3141 add_nested_action_end(*sfa, start);
3145 static int copy_action(const struct nlattr *from,
3146 struct sw_flow_actions **sfa, bool log)
3148 int totlen = NLA_ALIGN(from->nla_len);
3151 to = reserve_sfa_size(sfa, from->nla_len, log);
3155 memcpy(to, from, totlen);
3159 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3160 const struct sw_flow_key *key,
3161 struct sw_flow_actions **sfa,
3162 __be16 eth_type, __be16 vlan_tci,
3163 u32 mpls_label_count, bool log,
3166 u8 mac_proto = ovs_key_mac_proto(key);
3167 const struct nlattr *a;
3170 if (depth > OVS_COPY_ACTIONS_MAX_DEPTH)
3173 nla_for_each_nested(a, attr, rem) {
3174 /* Expected argument lengths, (u32)-1 for variable length. */
3175 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
3176 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
3177 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
3178 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
3179 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
3180 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
3181 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
3182 [OVS_ACTION_ATTR_POP_VLAN] = 0,
3183 [OVS_ACTION_ATTR_SET] = (u32)-1,
3184 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
3185 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
3186 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
3187 [OVS_ACTION_ATTR_CT] = (u32)-1,
3188 [OVS_ACTION_ATTR_CT_CLEAR] = 0,
3189 [OVS_ACTION_ATTR_TRUNC] = sizeof(struct ovs_action_trunc),
3190 [OVS_ACTION_ATTR_PUSH_ETH] = sizeof(struct ovs_action_push_eth),
3191 [OVS_ACTION_ATTR_POP_ETH] = 0,
3192 [OVS_ACTION_ATTR_PUSH_NSH] = (u32)-1,
3193 [OVS_ACTION_ATTR_POP_NSH] = 0,
3194 [OVS_ACTION_ATTR_METER] = sizeof(u32),
3195 [OVS_ACTION_ATTR_CLONE] = (u32)-1,
3196 [OVS_ACTION_ATTR_CHECK_PKT_LEN] = (u32)-1,
3197 [OVS_ACTION_ATTR_ADD_MPLS] = sizeof(struct ovs_action_add_mpls),
3198 [OVS_ACTION_ATTR_DEC_TTL] = (u32)-1,
3200 const struct ovs_action_push_vlan *vlan;
3201 int type = nla_type(a);
3204 if (type > OVS_ACTION_ATTR_MAX ||
3205 (action_lens[type] != nla_len(a) &&
3206 action_lens[type] != (u32)-1))
3211 case OVS_ACTION_ATTR_UNSPEC:
3214 case OVS_ACTION_ATTR_USERSPACE:
3215 err = validate_userspace(a);
3220 case OVS_ACTION_ATTR_OUTPUT:
3221 if (nla_get_u32(a) >= DP_MAX_PORTS)
3225 case OVS_ACTION_ATTR_TRUNC: {
3226 const struct ovs_action_trunc *trunc = nla_data(a);
3228 if (trunc->max_len < ETH_HLEN)
3233 case OVS_ACTION_ATTR_HASH: {
3234 const struct ovs_action_hash *act_hash = nla_data(a);
3236 switch (act_hash->hash_alg) {
3237 case OVS_HASH_ALG_L4:
3246 case OVS_ACTION_ATTR_POP_VLAN:
3247 if (mac_proto != MAC_PROTO_ETHERNET)
3249 vlan_tci = htons(0);
3252 case OVS_ACTION_ATTR_PUSH_VLAN:
3253 if (mac_proto != MAC_PROTO_ETHERNET)
3256 if (!eth_type_vlan(vlan->vlan_tpid))
3258 if (!(vlan->vlan_tci & htons(VLAN_CFI_MASK)))
3260 vlan_tci = vlan->vlan_tci;
3263 case OVS_ACTION_ATTR_RECIRC:
3266 case OVS_ACTION_ATTR_ADD_MPLS: {
3267 const struct ovs_action_add_mpls *mpls = nla_data(a);
3269 if (!eth_p_mpls(mpls->mpls_ethertype))
3272 if (mpls->tun_flags & OVS_MPLS_L3_TUNNEL_FLAG_MASK) {
3273 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3274 (eth_type != htons(ETH_P_IP) &&
3275 eth_type != htons(ETH_P_IPV6) &&
3276 eth_type != htons(ETH_P_ARP) &&
3277 eth_type != htons(ETH_P_RARP) &&
3278 !eth_p_mpls(eth_type)))
3282 if (mac_proto == MAC_PROTO_ETHERNET) {
3283 mpls_label_count = 1;
3284 mac_proto = MAC_PROTO_NONE;
3289 eth_type = mpls->mpls_ethertype;
3293 case OVS_ACTION_ATTR_PUSH_MPLS: {
3294 const struct ovs_action_push_mpls *mpls = nla_data(a);
3296 if (!eth_p_mpls(mpls->mpls_ethertype))
3298 /* Prohibit push MPLS other than to a white list
3299 * for packets that have a known tag order.
3301 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3302 (eth_type != htons(ETH_P_IP) &&
3303 eth_type != htons(ETH_P_IPV6) &&
3304 eth_type != htons(ETH_P_ARP) &&
3305 eth_type != htons(ETH_P_RARP) &&
3306 !eth_p_mpls(eth_type)))
3308 eth_type = mpls->mpls_ethertype;
3313 case OVS_ACTION_ATTR_POP_MPLS: {
3315 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3316 !eth_p_mpls(eth_type))
3319 /* Disallow subsequent L2.5+ set actions and mpls_pop
3320 * actions once the last MPLS label in the packet is
3321 * popped as there is no check here to ensure that
3322 * the new eth type is valid and thus set actions could
3323 * write off the end of the packet or otherwise corrupt
3326 * Support for these actions is planned using packet
3329 proto = nla_get_be16(a);
3331 if (proto == htons(ETH_P_TEB) &&
3332 mac_proto != MAC_PROTO_NONE)
3337 if (!eth_p_mpls(proto) || !mpls_label_count)
3338 eth_type = htons(0);
3345 case OVS_ACTION_ATTR_SET:
3346 err = validate_set(a, key, sfa,
3347 &skip_copy, mac_proto, eth_type,
3353 case OVS_ACTION_ATTR_SET_MASKED:
3354 err = validate_set(a, key, sfa,
3355 &skip_copy, mac_proto, eth_type,
3361 case OVS_ACTION_ATTR_SAMPLE: {
3362 bool last = nla_is_last(a, rem);
3364 err = validate_and_copy_sample(net, a, key, sfa,
3374 case OVS_ACTION_ATTR_CT:
3375 err = ovs_ct_copy_action(net, a, key, sfa, log);
3381 case OVS_ACTION_ATTR_CT_CLEAR:
3384 case OVS_ACTION_ATTR_PUSH_ETH:
3385 /* Disallow pushing an Ethernet header if one
3386 * is already present */
3387 if (mac_proto != MAC_PROTO_NONE)
3389 mac_proto = MAC_PROTO_ETHERNET;
3392 case OVS_ACTION_ATTR_POP_ETH:
3393 if (mac_proto != MAC_PROTO_ETHERNET)
3395 if (vlan_tci & htons(VLAN_CFI_MASK))
3397 mac_proto = MAC_PROTO_NONE;
3400 case OVS_ACTION_ATTR_PUSH_NSH:
3401 if (mac_proto != MAC_PROTO_ETHERNET) {
3404 next_proto = tun_p_from_eth_p(eth_type);
3408 mac_proto = MAC_PROTO_NONE;
3409 if (!validate_nsh(nla_data(a), false, true, true))
3413 case OVS_ACTION_ATTR_POP_NSH: {
3416 if (eth_type != htons(ETH_P_NSH))
3418 inner_proto = tun_p_to_eth_p(key->nsh.base.np);
3421 if (key->nsh.base.np == TUN_P_ETHERNET)
3422 mac_proto = MAC_PROTO_ETHERNET;
3424 mac_proto = MAC_PROTO_NONE;
3428 case OVS_ACTION_ATTR_METER:
3429 /* Non-existent meters are simply ignored. */
3432 case OVS_ACTION_ATTR_CLONE: {
3433 bool last = nla_is_last(a, rem);
3435 err = validate_and_copy_clone(net, a, key, sfa,
3445 case OVS_ACTION_ATTR_CHECK_PKT_LEN: {
3446 bool last = nla_is_last(a, rem);
3448 err = validate_and_copy_check_pkt_len(net, a, key, sfa,
3460 case OVS_ACTION_ATTR_DEC_TTL:
3461 err = validate_and_copy_dec_ttl(net, a, key, sfa,
3463 mpls_label_count, log,
3471 OVS_NLERR(log, "Unknown Action type %d", type);
3475 err = copy_action(a, sfa, log);
3487 /* 'key' must be the masked key. */
3488 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3489 const struct sw_flow_key *key,
3490 struct sw_flow_actions **sfa, bool log)
3493 u32 mpls_label_count = 0;
3495 *sfa = nla_alloc_flow_actions(min(nla_len(attr), MAX_ACTIONS_BUFSIZE));
3497 return PTR_ERR(*sfa);
3499 if (eth_p_mpls(key->eth.type))
3500 mpls_label_count = hweight_long(key->mpls.num_labels_mask);
3502 (*sfa)->orig_len = nla_len(attr);
3503 err = __ovs_nla_copy_actions(net, attr, key, sfa, key->eth.type,
3504 key->eth.vlan.tci, mpls_label_count, log,
3507 ovs_nla_free_flow_actions(*sfa);
3512 static int sample_action_to_attr(const struct nlattr *attr,
3513 struct sk_buff *skb)
3515 struct nlattr *start, *ac_start = NULL, *sample_arg;
3516 int err = 0, rem = nla_len(attr);
3517 const struct sample_arg *arg;
3518 struct nlattr *actions;
3520 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SAMPLE);
3524 sample_arg = nla_data(attr);
3525 arg = nla_data(sample_arg);
3526 actions = nla_next(sample_arg, &rem);
3528 if (nla_put_u32(skb, OVS_SAMPLE_ATTR_PROBABILITY, arg->probability)) {
3533 ac_start = nla_nest_start_noflag(skb, OVS_SAMPLE_ATTR_ACTIONS);
3539 err = ovs_nla_put_actions(actions, rem, skb);
3543 nla_nest_cancel(skb, ac_start);
3544 nla_nest_cancel(skb, start);
3546 nla_nest_end(skb, ac_start);
3547 nla_nest_end(skb, start);
3553 static int clone_action_to_attr(const struct nlattr *attr,
3554 struct sk_buff *skb)
3556 struct nlattr *start;
3557 int err = 0, rem = nla_len(attr);
3559 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CLONE);
3563 /* Skipping the OVS_CLONE_ATTR_EXEC that is always the first attribute. */
3564 attr = nla_next(nla_data(attr), &rem);
3565 err = ovs_nla_put_actions(attr, rem, skb);
3568 nla_nest_cancel(skb, start);
3570 nla_nest_end(skb, start);
3575 static int check_pkt_len_action_to_attr(const struct nlattr *attr,
3576 struct sk_buff *skb)
3578 struct nlattr *start, *ac_start = NULL;
3579 const struct check_pkt_len_arg *arg;
3580 const struct nlattr *a, *cpl_arg;
3581 int err = 0, rem = nla_len(attr);
3583 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CHECK_PKT_LEN);
3587 /* The first nested attribute in 'attr' is always
3588 * 'OVS_CHECK_PKT_LEN_ATTR_ARG'.
3590 cpl_arg = nla_data(attr);
3591 arg = nla_data(cpl_arg);
3593 if (nla_put_u16(skb, OVS_CHECK_PKT_LEN_ATTR_PKT_LEN, arg->pkt_len)) {
3598 /* Second nested attribute in 'attr' is always
3599 * 'OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL'.
3601 a = nla_next(cpl_arg, &rem);
3602 ac_start = nla_nest_start_noflag(skb,
3603 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL);
3609 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3611 nla_nest_cancel(skb, ac_start);
3614 nla_nest_end(skb, ac_start);
3617 /* Third nested attribute in 'attr' is always
3618 * OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER.
3620 a = nla_next(a, &rem);
3621 ac_start = nla_nest_start_noflag(skb,
3622 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER);
3628 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3630 nla_nest_cancel(skb, ac_start);
3633 nla_nest_end(skb, ac_start);
3636 nla_nest_end(skb, start);
3640 nla_nest_cancel(skb, start);
3644 static int dec_ttl_action_to_attr(const struct nlattr *attr,
3645 struct sk_buff *skb)
3647 struct nlattr *start, *action_start;
3648 const struct nlattr *a;
3651 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_DEC_TTL);
3655 nla_for_each_attr(a, nla_data(attr), nla_len(attr), rem) {
3656 switch (nla_type(a)) {
3657 case OVS_DEC_TTL_ATTR_ACTION:
3659 action_start = nla_nest_start_noflag(skb, OVS_DEC_TTL_ATTR_ACTION);
3660 if (!action_start) {
3665 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3669 nla_nest_end(skb, action_start);
3673 /* Ignore all other option to be future compatible */
3678 nla_nest_end(skb, start);
3682 nla_nest_cancel(skb, start);
3686 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
3688 const struct nlattr *ovs_key = nla_data(a);
3689 int key_type = nla_type(ovs_key);
3690 struct nlattr *start;
3694 case OVS_KEY_ATTR_TUNNEL_INFO: {
3695 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
3696 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
3698 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3702 err = ip_tun_to_nlattr(skb, &tun_info->key,
3703 ip_tunnel_info_opts(tun_info),
3704 tun_info->options_len,
3705 ip_tunnel_info_af(tun_info), tun_info->mode);
3708 nla_nest_end(skb, start);
3712 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
3720 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
3721 struct sk_buff *skb)
3723 const struct nlattr *ovs_key = nla_data(a);
3725 size_t key_len = nla_len(ovs_key) / 2;
3727 /* Revert the conversion we did from a non-masked set action to
3728 * masked set action.
3730 nla = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3734 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
3737 nla_nest_end(skb, nla);
3741 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
3743 const struct nlattr *a;
3746 nla_for_each_attr(a, attr, len, rem) {
3747 int type = nla_type(a);
3750 case OVS_ACTION_ATTR_SET:
3751 err = set_action_to_attr(a, skb);
3756 case OVS_ACTION_ATTR_SET_TO_MASKED:
3757 err = masked_set_action_to_set_action_attr(a, skb);
3762 case OVS_ACTION_ATTR_SAMPLE:
3763 err = sample_action_to_attr(a, skb);
3768 case OVS_ACTION_ATTR_CT:
3769 err = ovs_ct_action_to_attr(nla_data(a), skb);
3774 case OVS_ACTION_ATTR_CLONE:
3775 err = clone_action_to_attr(a, skb);
3780 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
3781 err = check_pkt_len_action_to_attr(a, skb);
3786 case OVS_ACTION_ATTR_DEC_TTL:
3787 err = dec_ttl_action_to_attr(a, skb);
3793 if (nla_put(skb, type, nla_len(a), nla_data(a)))