2 * Copyright (c) 2007-2017 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
23 #include <linux/uaccess.h>
24 #include <linux/netdevice.h>
25 #include <linux/etherdevice.h>
26 #include <linux/if_ether.h>
27 #include <linux/if_vlan.h>
28 #include <net/llc_pdu.h>
29 #include <linux/kernel.h>
30 #include <linux/jhash.h>
31 #include <linux/jiffies.h>
32 #include <linux/llc.h>
33 #include <linux/module.h>
35 #include <linux/rcupdate.h>
36 #include <linux/if_arp.h>
38 #include <linux/ipv6.h>
39 #include <linux/sctp.h>
40 #include <linux/tcp.h>
41 #include <linux/udp.h>
42 #include <linux/icmp.h>
43 #include <linux/icmpv6.h>
44 #include <linux/rculist.h>
45 #include <net/geneve.h>
48 #include <net/ndisc.h>
50 #include <net/vxlan.h>
51 #include <net/tun_proto.h>
53 #include "flow_netlink.h"
57 const struct ovs_len_tbl *next;
60 #define OVS_ATTR_NESTED -1
61 #define OVS_ATTR_VARIABLE -2
63 static bool actions_may_change_flow(const struct nlattr *actions)
68 nla_for_each_nested(nla, actions, rem) {
69 u16 action = nla_type(nla);
72 case OVS_ACTION_ATTR_OUTPUT:
73 case OVS_ACTION_ATTR_RECIRC:
74 case OVS_ACTION_ATTR_TRUNC:
75 case OVS_ACTION_ATTR_USERSPACE:
78 case OVS_ACTION_ATTR_CT:
79 case OVS_ACTION_ATTR_CT_CLEAR:
80 case OVS_ACTION_ATTR_HASH:
81 case OVS_ACTION_ATTR_POP_ETH:
82 case OVS_ACTION_ATTR_POP_MPLS:
83 case OVS_ACTION_ATTR_POP_NSH:
84 case OVS_ACTION_ATTR_POP_VLAN:
85 case OVS_ACTION_ATTR_PUSH_ETH:
86 case OVS_ACTION_ATTR_PUSH_MPLS:
87 case OVS_ACTION_ATTR_PUSH_NSH:
88 case OVS_ACTION_ATTR_PUSH_VLAN:
89 case OVS_ACTION_ATTR_SAMPLE:
90 case OVS_ACTION_ATTR_SET:
91 case OVS_ACTION_ATTR_SET_MASKED:
92 case OVS_ACTION_ATTR_METER:
100 static void update_range(struct sw_flow_match *match,
101 size_t offset, size_t size, bool is_mask)
103 struct sw_flow_key_range *range;
104 size_t start = rounddown(offset, sizeof(long));
105 size_t end = roundup(offset + size, sizeof(long));
108 range = &match->range;
110 range = &match->mask->range;
112 if (range->start == range->end) {
113 range->start = start;
118 if (range->start > start)
119 range->start = start;
121 if (range->end < end)
125 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
127 update_range(match, offsetof(struct sw_flow_key, field), \
128 sizeof((match)->key->field), is_mask); \
130 (match)->mask->key.field = value; \
132 (match)->key->field = value; \
135 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
137 update_range(match, offset, len, is_mask); \
139 memcpy((u8 *)&(match)->mask->key + offset, value_p, \
142 memcpy((u8 *)(match)->key + offset, value_p, len); \
145 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
146 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
147 value_p, len, is_mask)
149 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
151 update_range(match, offsetof(struct sw_flow_key, field), \
152 sizeof((match)->key->field), is_mask); \
154 memset((u8 *)&(match)->mask->key.field, value, \
155 sizeof((match)->mask->key.field)); \
157 memset((u8 *)&(match)->key->field, value, \
158 sizeof((match)->key->field)); \
161 static bool match_validate(const struct sw_flow_match *match,
162 u64 key_attrs, u64 mask_attrs, bool log)
164 u64 key_expected = 0;
165 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
167 /* The following mask attributes allowed only if they
168 * pass the validation tests. */
169 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
170 | (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)
171 | (1 << OVS_KEY_ATTR_IPV6)
172 | (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)
173 | (1 << OVS_KEY_ATTR_TCP)
174 | (1 << OVS_KEY_ATTR_TCP_FLAGS)
175 | (1 << OVS_KEY_ATTR_UDP)
176 | (1 << OVS_KEY_ATTR_SCTP)
177 | (1 << OVS_KEY_ATTR_ICMP)
178 | (1 << OVS_KEY_ATTR_ICMPV6)
179 | (1 << OVS_KEY_ATTR_ARP)
180 | (1 << OVS_KEY_ATTR_ND)
181 | (1 << OVS_KEY_ATTR_MPLS)
182 | (1 << OVS_KEY_ATTR_NSH));
184 /* Always allowed mask fields. */
185 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
186 | (1 << OVS_KEY_ATTR_IN_PORT)
187 | (1 << OVS_KEY_ATTR_ETHERTYPE));
189 /* Check key attributes. */
190 if (match->key->eth.type == htons(ETH_P_ARP)
191 || match->key->eth.type == htons(ETH_P_RARP)) {
192 key_expected |= 1 << OVS_KEY_ATTR_ARP;
193 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
194 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
197 if (eth_p_mpls(match->key->eth.type)) {
198 key_expected |= 1 << OVS_KEY_ATTR_MPLS;
199 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
200 mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
203 if (match->key->eth.type == htons(ETH_P_IP)) {
204 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
205 if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
206 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
207 mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4;
210 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
211 if (match->key->ip.proto == IPPROTO_UDP) {
212 key_expected |= 1 << OVS_KEY_ATTR_UDP;
213 if (match->mask && (match->mask->key.ip.proto == 0xff))
214 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
217 if (match->key->ip.proto == IPPROTO_SCTP) {
218 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
219 if (match->mask && (match->mask->key.ip.proto == 0xff))
220 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
223 if (match->key->ip.proto == IPPROTO_TCP) {
224 key_expected |= 1 << OVS_KEY_ATTR_TCP;
225 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
226 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
227 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
228 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
232 if (match->key->ip.proto == IPPROTO_ICMP) {
233 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
234 if (match->mask && (match->mask->key.ip.proto == 0xff))
235 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
240 if (match->key->eth.type == htons(ETH_P_IPV6)) {
241 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
242 if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
243 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
244 mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6;
247 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
248 if (match->key->ip.proto == IPPROTO_UDP) {
249 key_expected |= 1 << OVS_KEY_ATTR_UDP;
250 if (match->mask && (match->mask->key.ip.proto == 0xff))
251 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
254 if (match->key->ip.proto == IPPROTO_SCTP) {
255 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
256 if (match->mask && (match->mask->key.ip.proto == 0xff))
257 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
260 if (match->key->ip.proto == IPPROTO_TCP) {
261 key_expected |= 1 << OVS_KEY_ATTR_TCP;
262 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
263 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
264 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
265 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
269 if (match->key->ip.proto == IPPROTO_ICMPV6) {
270 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
271 if (match->mask && (match->mask->key.ip.proto == 0xff))
272 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
274 if (match->key->tp.src ==
275 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
276 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
277 key_expected |= 1 << OVS_KEY_ATTR_ND;
278 /* Original direction conntrack tuple
279 * uses the same space as the ND fields
280 * in the key, so both are not allowed
283 mask_allowed &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
284 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
285 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
291 if (match->key->eth.type == htons(ETH_P_NSH)) {
292 key_expected |= 1 << OVS_KEY_ATTR_NSH;
294 match->mask->key.eth.type == htons(0xffff)) {
295 mask_allowed |= 1 << OVS_KEY_ATTR_NSH;
299 if ((key_attrs & key_expected) != key_expected) {
300 /* Key attributes check failed. */
301 OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
302 (unsigned long long)key_attrs,
303 (unsigned long long)key_expected);
307 if ((mask_attrs & mask_allowed) != mask_attrs) {
308 /* Mask attributes check failed. */
309 OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
310 (unsigned long long)mask_attrs,
311 (unsigned long long)mask_allowed);
318 size_t ovs_tun_key_attr_size(void)
320 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
321 * updating this function.
323 return nla_total_size_64bit(8) /* OVS_TUNNEL_KEY_ATTR_ID */
324 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
325 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_DST */
326 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
327 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
328 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
329 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
330 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
331 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
332 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS is mutually exclusive with
333 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
335 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
336 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
339 static size_t ovs_nsh_key_attr_size(void)
341 /* Whenever adding new OVS_NSH_KEY_ FIELDS, we should consider
342 * updating this function.
344 return nla_total_size(NSH_BASE_HDR_LEN) /* OVS_NSH_KEY_ATTR_BASE */
345 /* OVS_NSH_KEY_ATTR_MD1 and OVS_NSH_KEY_ATTR_MD2 are
346 * mutually exclusive, so the bigger one can cover
349 + nla_total_size(NSH_CTX_HDRS_MAX_LEN);
352 size_t ovs_key_attr_size(void)
354 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
355 * updating this function.
357 BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO != 29);
359 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
360 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
361 + ovs_tun_key_attr_size()
362 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
363 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
364 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
365 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
366 + nla_total_size(4) /* OVS_KEY_ATTR_CT_STATE */
367 + nla_total_size(2) /* OVS_KEY_ATTR_CT_ZONE */
368 + nla_total_size(4) /* OVS_KEY_ATTR_CT_MARK */
369 + nla_total_size(16) /* OVS_KEY_ATTR_CT_LABELS */
370 + nla_total_size(40) /* OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6 */
371 + nla_total_size(0) /* OVS_KEY_ATTR_NSH */
372 + ovs_nsh_key_attr_size()
373 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
374 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
375 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
376 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
377 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
378 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
379 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
380 + nla_total_size(28); /* OVS_KEY_ATTR_ND */
383 static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
384 [OVS_VXLAN_EXT_GBP] = { .len = sizeof(u32) },
387 static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
388 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
389 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
390 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
391 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
392 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
393 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
394 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
395 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
396 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
397 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
398 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_VARIABLE },
399 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED,
400 .next = ovs_vxlan_ext_key_lens },
401 [OVS_TUNNEL_KEY_ATTR_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
402 [OVS_TUNNEL_KEY_ATTR_IPV6_DST] = { .len = sizeof(struct in6_addr) },
405 static const struct ovs_len_tbl
406 ovs_nsh_key_attr_lens[OVS_NSH_KEY_ATTR_MAX + 1] = {
407 [OVS_NSH_KEY_ATTR_BASE] = { .len = sizeof(struct ovs_nsh_key_base) },
408 [OVS_NSH_KEY_ATTR_MD1] = { .len = sizeof(struct ovs_nsh_key_md1) },
409 [OVS_NSH_KEY_ATTR_MD2] = { .len = OVS_ATTR_VARIABLE },
412 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
413 static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
414 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
415 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
416 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
417 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
418 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
419 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
420 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
421 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
422 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
423 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
424 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
425 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
426 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
427 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
428 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
429 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
430 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
431 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
432 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
433 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
434 .next = ovs_tunnel_key_lens, },
435 [OVS_KEY_ATTR_MPLS] = { .len = sizeof(struct ovs_key_mpls) },
436 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u32) },
437 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
438 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
439 [OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
440 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4] = {
441 .len = sizeof(struct ovs_key_ct_tuple_ipv4) },
442 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6] = {
443 .len = sizeof(struct ovs_key_ct_tuple_ipv6) },
444 [OVS_KEY_ATTR_NSH] = { .len = OVS_ATTR_NESTED,
445 .next = ovs_nsh_key_attr_lens, },
448 static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
450 return expected_len == attr_len ||
451 expected_len == OVS_ATTR_NESTED ||
452 expected_len == OVS_ATTR_VARIABLE;
455 static bool is_all_zero(const u8 *fp, size_t size)
462 for (i = 0; i < size; i++)
469 static int __parse_flow_nlattrs(const struct nlattr *attr,
470 const struct nlattr *a[],
471 u64 *attrsp, bool log, bool nz)
473 const struct nlattr *nla;
478 nla_for_each_nested(nla, attr, rem) {
479 u16 type = nla_type(nla);
482 if (type > OVS_KEY_ATTR_MAX) {
483 OVS_NLERR(log, "Key type %d is out of range max %d",
484 type, OVS_KEY_ATTR_MAX);
488 if (attrs & (1 << type)) {
489 OVS_NLERR(log, "Duplicate key (type %d).", type);
493 expected_len = ovs_key_lens[type].len;
494 if (!check_attr_len(nla_len(nla), expected_len)) {
495 OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
496 type, nla_len(nla), expected_len);
500 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
506 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
514 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
515 const struct nlattr *a[], u64 *attrsp,
518 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
521 int parse_flow_nlattrs(const struct nlattr *attr, const struct nlattr *a[],
522 u64 *attrsp, bool log)
524 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
527 static int genev_tun_opt_from_nlattr(const struct nlattr *a,
528 struct sw_flow_match *match, bool is_mask,
531 unsigned long opt_key_offset;
533 if (nla_len(a) > sizeof(match->key->tun_opts)) {
534 OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
535 nla_len(a), sizeof(match->key->tun_opts));
539 if (nla_len(a) % 4 != 0) {
540 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
545 /* We need to record the length of the options passed
546 * down, otherwise packets with the same format but
547 * additional options will be silently matched.
550 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
553 /* This is somewhat unusual because it looks at
554 * both the key and mask while parsing the
555 * attributes (and by extension assumes the key
556 * is parsed first). Normally, we would verify
557 * that each is the correct length and that the
558 * attributes line up in the validate function.
559 * However, that is difficult because this is
560 * variable length and we won't have the
563 if (match->key->tun_opts_len != nla_len(a)) {
564 OVS_NLERR(log, "Geneve option len %d != mask len %d",
565 match->key->tun_opts_len, nla_len(a));
569 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
572 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
573 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
574 nla_len(a), is_mask);
578 static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
579 struct sw_flow_match *match, bool is_mask,
584 unsigned long opt_key_offset;
585 struct vxlan_metadata opts;
587 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
589 memset(&opts, 0, sizeof(opts));
590 nla_for_each_nested(a, attr, rem) {
591 int type = nla_type(a);
593 if (type > OVS_VXLAN_EXT_MAX) {
594 OVS_NLERR(log, "VXLAN extension %d out of range max %d",
595 type, OVS_VXLAN_EXT_MAX);
599 if (!check_attr_len(nla_len(a),
600 ovs_vxlan_ext_key_lens[type].len)) {
601 OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
603 ovs_vxlan_ext_key_lens[type].len);
608 case OVS_VXLAN_EXT_GBP:
609 opts.gbp = nla_get_u32(a);
612 OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
618 OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
624 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
626 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
628 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
629 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
634 static int ip_tun_from_nlattr(const struct nlattr *attr,
635 struct sw_flow_match *match, bool is_mask,
638 bool ttl = false, ipv4 = false, ipv6 = false;
639 __be16 tun_flags = 0;
644 nla_for_each_nested(a, attr, rem) {
645 int type = nla_type(a);
648 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
649 OVS_NLERR(log, "Tunnel attr %d out of range max %d",
650 type, OVS_TUNNEL_KEY_ATTR_MAX);
654 if (!check_attr_len(nla_len(a),
655 ovs_tunnel_key_lens[type].len)) {
656 OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
657 type, nla_len(a), ovs_tunnel_key_lens[type].len);
662 case OVS_TUNNEL_KEY_ATTR_ID:
663 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
664 nla_get_be64(a), is_mask);
665 tun_flags |= TUNNEL_KEY;
667 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
668 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
669 nla_get_in_addr(a), is_mask);
672 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
673 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
674 nla_get_in_addr(a), is_mask);
677 case OVS_TUNNEL_KEY_ATTR_IPV6_SRC:
678 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.src,
679 nla_get_in6_addr(a), is_mask);
682 case OVS_TUNNEL_KEY_ATTR_IPV6_DST:
683 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
684 nla_get_in6_addr(a), is_mask);
687 case OVS_TUNNEL_KEY_ATTR_TOS:
688 SW_FLOW_KEY_PUT(match, tun_key.tos,
689 nla_get_u8(a), is_mask);
691 case OVS_TUNNEL_KEY_ATTR_TTL:
692 SW_FLOW_KEY_PUT(match, tun_key.ttl,
693 nla_get_u8(a), is_mask);
696 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
697 tun_flags |= TUNNEL_DONT_FRAGMENT;
699 case OVS_TUNNEL_KEY_ATTR_CSUM:
700 tun_flags |= TUNNEL_CSUM;
702 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
703 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
704 nla_get_be16(a), is_mask);
706 case OVS_TUNNEL_KEY_ATTR_TP_DST:
707 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
708 nla_get_be16(a), is_mask);
710 case OVS_TUNNEL_KEY_ATTR_OAM:
711 tun_flags |= TUNNEL_OAM;
713 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
715 OVS_NLERR(log, "Multiple metadata blocks provided");
719 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
723 tun_flags |= TUNNEL_GENEVE_OPT;
726 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
728 OVS_NLERR(log, "Multiple metadata blocks provided");
732 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
736 tun_flags |= TUNNEL_VXLAN_OPT;
739 case OVS_TUNNEL_KEY_ATTR_PAD:
742 OVS_NLERR(log, "Unknown IP tunnel attribute %d",
748 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
750 SW_FLOW_KEY_MEMSET_FIELD(match, tun_proto, 0xff, true);
752 SW_FLOW_KEY_PUT(match, tun_proto, ipv6 ? AF_INET6 : AF_INET,
756 OVS_NLERR(log, "IP tunnel attribute has %d unknown bytes.",
762 OVS_NLERR(log, "Mixed IPv4 and IPv6 tunnel attributes");
767 if (!ipv4 && !ipv6) {
768 OVS_NLERR(log, "IP tunnel dst address not specified");
771 if (ipv4 && !match->key->tun_key.u.ipv4.dst) {
772 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
775 if (ipv6 && ipv6_addr_any(&match->key->tun_key.u.ipv6.dst)) {
776 OVS_NLERR(log, "IPv6 tunnel dst address is zero");
781 OVS_NLERR(log, "IP tunnel TTL not specified.");
789 static int vxlan_opt_to_nlattr(struct sk_buff *skb,
790 const void *tun_opts, int swkey_tun_opts_len)
792 const struct vxlan_metadata *opts = tun_opts;
795 nla = nla_nest_start(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
799 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
802 nla_nest_end(skb, nla);
806 static int __ip_tun_to_nlattr(struct sk_buff *skb,
807 const struct ip_tunnel_key *output,
808 const void *tun_opts, int swkey_tun_opts_len,
809 unsigned short tun_proto)
811 if (output->tun_flags & TUNNEL_KEY &&
812 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id,
813 OVS_TUNNEL_KEY_ATTR_PAD))
817 if (output->u.ipv4.src &&
818 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
821 if (output->u.ipv4.dst &&
822 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
827 if (!ipv6_addr_any(&output->u.ipv6.src) &&
828 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_SRC,
829 &output->u.ipv6.src))
831 if (!ipv6_addr_any(&output->u.ipv6.dst) &&
832 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_DST,
833 &output->u.ipv6.dst))
838 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
840 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
842 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
843 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
845 if ((output->tun_flags & TUNNEL_CSUM) &&
846 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
848 if (output->tp_src &&
849 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
851 if (output->tp_dst &&
852 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
854 if ((output->tun_flags & TUNNEL_OAM) &&
855 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
857 if (swkey_tun_opts_len) {
858 if (output->tun_flags & TUNNEL_GENEVE_OPT &&
859 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
860 swkey_tun_opts_len, tun_opts))
862 else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
863 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
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)
878 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
882 err = __ip_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len,
887 nla_nest_end(skb, nla);
891 int ovs_nla_put_tunnel_info(struct sk_buff *skb,
892 struct ip_tunnel_info *tun_info)
894 return __ip_tun_to_nlattr(skb, &tun_info->key,
895 ip_tunnel_info_opts(tun_info),
896 tun_info->options_len,
897 ip_tunnel_info_af(tun_info));
900 static int encode_vlan_from_nlattrs(struct sw_flow_match *match,
901 const struct nlattr *a[],
902 bool is_mask, bool inner)
907 if (a[OVS_KEY_ATTR_VLAN])
908 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
910 if (a[OVS_KEY_ATTR_ETHERTYPE])
911 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
913 if (likely(!inner)) {
914 SW_FLOW_KEY_PUT(match, eth.vlan.tpid, tpid, is_mask);
915 SW_FLOW_KEY_PUT(match, eth.vlan.tci, tci, is_mask);
917 SW_FLOW_KEY_PUT(match, eth.cvlan.tpid, tpid, is_mask);
918 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, tci, is_mask);
923 static int validate_vlan_from_nlattrs(const struct sw_flow_match *match,
924 u64 key_attrs, bool inner,
925 const struct nlattr **a, bool log)
929 if (!((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
930 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
931 eth_type_vlan(nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE])))) {
936 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
937 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
938 OVS_NLERR(log, "Invalid %s frame", (inner) ? "C-VLAN" : "VLAN");
942 if (a[OVS_KEY_ATTR_VLAN])
943 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
945 if (!(tci & htons(VLAN_TAG_PRESENT))) {
947 OVS_NLERR(log, "%s TCI does not have VLAN_TAG_PRESENT bit set.",
948 (inner) ? "C-VLAN" : "VLAN");
950 } else if (nla_len(a[OVS_KEY_ATTR_ENCAP])) {
951 /* Corner case for truncated VLAN header. */
952 OVS_NLERR(log, "Truncated %s header has non-zero encap attribute.",
953 (inner) ? "C-VLAN" : "VLAN");
961 static int validate_vlan_mask_from_nlattrs(const struct sw_flow_match *match,
962 u64 key_attrs, bool inner,
963 const struct nlattr **a, bool log)
967 bool encap_valid = !!(match->key->eth.vlan.tci &
968 htons(VLAN_TAG_PRESENT));
969 bool i_encap_valid = !!(match->key->eth.cvlan.tci &
970 htons(VLAN_TAG_PRESENT));
972 if (!(key_attrs & (1 << OVS_KEY_ATTR_ENCAP))) {
977 if ((!inner && !encap_valid) || (inner && !i_encap_valid)) {
978 OVS_NLERR(log, "Encap mask attribute is set for non-%s frame.",
979 (inner) ? "C-VLAN" : "VLAN");
983 if (a[OVS_KEY_ATTR_VLAN])
984 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
986 if (a[OVS_KEY_ATTR_ETHERTYPE])
987 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
989 if (tpid != htons(0xffff)) {
990 OVS_NLERR(log, "Must have an exact match on %s TPID (mask=%x).",
991 (inner) ? "C-VLAN" : "VLAN", ntohs(tpid));
994 if (!(tci & htons(VLAN_TAG_PRESENT))) {
995 OVS_NLERR(log, "%s TCI mask does not have exact match for VLAN_TAG_PRESENT bit.",
996 (inner) ? "C-VLAN" : "VLAN");
1003 static int __parse_vlan_from_nlattrs(struct sw_flow_match *match,
1004 u64 *key_attrs, bool inner,
1005 const struct nlattr **a, bool is_mask,
1009 const struct nlattr *encap;
1012 err = validate_vlan_from_nlattrs(match, *key_attrs, inner,
1015 err = validate_vlan_mask_from_nlattrs(match, *key_attrs, inner,
1020 err = encode_vlan_from_nlattrs(match, a, is_mask, inner);
1024 *key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1025 *key_attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
1026 *key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1028 encap = a[OVS_KEY_ATTR_ENCAP];
1031 err = parse_flow_nlattrs(encap, a, key_attrs, log);
1033 err = parse_flow_mask_nlattrs(encap, a, key_attrs, log);
1038 static int parse_vlan_from_nlattrs(struct sw_flow_match *match,
1039 u64 *key_attrs, const struct nlattr **a,
1040 bool is_mask, bool log)
1043 bool encap_valid = false;
1045 err = __parse_vlan_from_nlattrs(match, key_attrs, false, a,
1050 encap_valid = !!(match->key->eth.vlan.tci & htons(VLAN_TAG_PRESENT));
1052 err = __parse_vlan_from_nlattrs(match, key_attrs, true, a,
1061 static int parse_eth_type_from_nlattrs(struct sw_flow_match *match,
1062 u64 *attrs, const struct nlattr **a,
1063 bool is_mask, bool log)
1067 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1069 /* Always exact match EtherType. */
1070 eth_type = htons(0xffff);
1071 } else if (!eth_proto_is_802_3(eth_type)) {
1072 OVS_NLERR(log, "EtherType %x is less than min %x",
1073 ntohs(eth_type), ETH_P_802_3_MIN);
1077 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
1078 *attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1082 static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
1083 u64 *attrs, const struct nlattr **a,
1084 bool is_mask, bool log)
1086 u8 mac_proto = MAC_PROTO_ETHERNET;
1088 if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
1089 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
1091 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
1092 *attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
1095 if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
1096 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
1098 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
1099 *attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
1102 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
1103 SW_FLOW_KEY_PUT(match, phy.priority,
1104 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
1105 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
1108 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
1109 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1112 in_port = 0xffffffff; /* Always exact match in_port. */
1113 } else if (in_port >= DP_MAX_PORTS) {
1114 OVS_NLERR(log, "Port %d exceeds max allowable %d",
1115 in_port, DP_MAX_PORTS);
1119 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
1120 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1121 } else if (!is_mask) {
1122 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
1125 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1126 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1128 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1129 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1131 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1132 if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1135 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1138 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
1139 ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
1140 u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
1142 if (ct_state & ~CT_SUPPORTED_MASK) {
1143 OVS_NLERR(log, "ct_state flags %08x unsupported",
1148 SW_FLOW_KEY_PUT(match, ct_state, ct_state, is_mask);
1149 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
1151 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
1152 ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
1153 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
1155 SW_FLOW_KEY_PUT(match, ct_zone, ct_zone, is_mask);
1156 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
1158 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
1159 ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
1160 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
1162 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
1163 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
1165 if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
1166 ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
1167 const struct ovs_key_ct_labels *cl;
1169 cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
1170 SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
1171 sizeof(*cl), is_mask);
1172 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
1174 if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)) {
1175 const struct ovs_key_ct_tuple_ipv4 *ct;
1177 ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4]);
1179 SW_FLOW_KEY_PUT(match, ipv4.ct_orig.src, ct->ipv4_src, is_mask);
1180 SW_FLOW_KEY_PUT(match, ipv4.ct_orig.dst, ct->ipv4_dst, is_mask);
1181 SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1182 SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1183 SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv4_proto, is_mask);
1184 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4);
1186 if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)) {
1187 const struct ovs_key_ct_tuple_ipv6 *ct;
1189 ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6]);
1191 SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.src, &ct->ipv6_src,
1192 sizeof(match->key->ipv6.ct_orig.src),
1194 SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.dst, &ct->ipv6_dst,
1195 sizeof(match->key->ipv6.ct_orig.dst),
1197 SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1198 SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1199 SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv6_proto, is_mask);
1200 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
1203 /* For layer 3 packets the Ethernet type is provided
1204 * and treated as metadata but no MAC addresses are provided.
1206 if (!(*attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
1207 (*attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)))
1208 mac_proto = MAC_PROTO_NONE;
1210 /* Always exact match mac_proto */
1211 SW_FLOW_KEY_PUT(match, mac_proto, is_mask ? 0xff : mac_proto, is_mask);
1213 if (mac_proto == MAC_PROTO_NONE)
1214 return parse_eth_type_from_nlattrs(match, attrs, a, is_mask,
1220 int nsh_hdr_from_nlattr(const struct nlattr *attr,
1221 struct nshhdr *nh, size_t size)
1229 /* validate_nsh has check this, so we needn't do duplicate check here
1231 if (size < NSH_BASE_HDR_LEN)
1234 nla_for_each_nested(a, attr, rem) {
1235 int type = nla_type(a);
1238 case OVS_NSH_KEY_ATTR_BASE: {
1239 const struct ovs_nsh_key_base *base = nla_data(a);
1241 flags = base->flags;
1244 nh->mdtype = base->mdtype;
1245 nh->path_hdr = base->path_hdr;
1248 case OVS_NSH_KEY_ATTR_MD1:
1250 if (mdlen > size - NSH_BASE_HDR_LEN)
1252 memcpy(&nh->md1, nla_data(a), mdlen);
1255 case OVS_NSH_KEY_ATTR_MD2:
1257 if (mdlen > size - NSH_BASE_HDR_LEN)
1259 memcpy(&nh->md2, nla_data(a), mdlen);
1267 /* nsh header length = NSH_BASE_HDR_LEN + mdlen */
1268 nh->ver_flags_ttl_len = 0;
1269 nsh_set_flags_ttl_len(nh, flags, ttl, NSH_BASE_HDR_LEN + mdlen);
1274 int nsh_key_from_nlattr(const struct nlattr *attr,
1275 struct ovs_key_nsh *nsh, struct ovs_key_nsh *nsh_mask)
1280 /* validate_nsh has check this, so we needn't do duplicate check here
1282 nla_for_each_nested(a, attr, rem) {
1283 int type = nla_type(a);
1286 case OVS_NSH_KEY_ATTR_BASE: {
1287 const struct ovs_nsh_key_base *base = nla_data(a);
1288 const struct ovs_nsh_key_base *base_mask = base + 1;
1291 nsh_mask->base = *base_mask;
1294 case OVS_NSH_KEY_ATTR_MD1: {
1295 const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1296 const struct ovs_nsh_key_md1 *md1_mask = md1 + 1;
1298 memcpy(nsh->context, md1->context, sizeof(*md1));
1299 memcpy(nsh_mask->context, md1_mask->context,
1303 case OVS_NSH_KEY_ATTR_MD2:
1304 /* Not supported yet */
1314 static int nsh_key_put_from_nlattr(const struct nlattr *attr,
1315 struct sw_flow_match *match, bool is_mask,
1316 bool is_push_nsh, bool log)
1320 bool has_base = false;
1321 bool has_md1 = false;
1322 bool has_md2 = false;
1326 if (WARN_ON(is_push_nsh && is_mask))
1329 nla_for_each_nested(a, attr, rem) {
1330 int type = nla_type(a);
1333 if (type > OVS_NSH_KEY_ATTR_MAX) {
1334 OVS_NLERR(log, "nsh attr %d is out of range max %d",
1335 type, OVS_NSH_KEY_ATTR_MAX);
1339 if (!check_attr_len(nla_len(a),
1340 ovs_nsh_key_attr_lens[type].len)) {
1343 "nsh attr %d has unexpected len %d expected %d",
1346 ovs_nsh_key_attr_lens[type].len
1352 case OVS_NSH_KEY_ATTR_BASE: {
1353 const struct ovs_nsh_key_base *base = nla_data(a);
1356 mdtype = base->mdtype;
1357 SW_FLOW_KEY_PUT(match, nsh.base.flags,
1358 base->flags, is_mask);
1359 SW_FLOW_KEY_PUT(match, nsh.base.ttl,
1360 base->ttl, is_mask);
1361 SW_FLOW_KEY_PUT(match, nsh.base.mdtype,
1362 base->mdtype, is_mask);
1363 SW_FLOW_KEY_PUT(match, nsh.base.np,
1365 SW_FLOW_KEY_PUT(match, nsh.base.path_hdr,
1366 base->path_hdr, is_mask);
1369 case OVS_NSH_KEY_ATTR_MD1: {
1370 const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1373 for (i = 0; i < NSH_MD1_CONTEXT_SIZE; i++)
1374 SW_FLOW_KEY_PUT(match, nsh.context[i],
1375 md1->context[i], is_mask);
1378 case OVS_NSH_KEY_ATTR_MD2:
1379 if (!is_push_nsh) /* Not supported MD type 2 yet */
1384 if (mdlen > NSH_CTX_HDRS_MAX_LEN || mdlen <= 0) {
1387 "Invalid MD length %d for MD type %d",
1395 OVS_NLERR(log, "Unknown nsh attribute %d",
1402 OVS_NLERR(log, "nsh attribute has %d unknown bytes.", rem);
1406 if (has_md1 && has_md2) {
1409 "invalid nsh attribute: md1 and md2 are exclusive."
1415 if ((has_md1 && mdtype != NSH_M_TYPE1) ||
1416 (has_md2 && mdtype != NSH_M_TYPE2)) {
1417 OVS_NLERR(1, "nsh attribute has unmatched MD type %d.",
1423 (!has_base || (!has_md1 && !has_md2))) {
1426 "push_nsh: missing base or metadata attributes"
1435 static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
1436 u64 attrs, const struct nlattr **a,
1437 bool is_mask, bool log)
1441 err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
1445 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1446 const struct ovs_key_ethernet *eth_key;
1448 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1449 SW_FLOW_KEY_MEMCPY(match, eth.src,
1450 eth_key->eth_src, ETH_ALEN, is_mask);
1451 SW_FLOW_KEY_MEMCPY(match, eth.dst,
1452 eth_key->eth_dst, ETH_ALEN, is_mask);
1453 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1455 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1456 /* VLAN attribute is always parsed before getting here since it
1457 * may occur multiple times.
1459 OVS_NLERR(log, "VLAN attribute unexpected.");
1463 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1464 err = parse_eth_type_from_nlattrs(match, &attrs, a, is_mask,
1468 } else if (!is_mask) {
1469 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1471 } else if (!match->key->eth.type) {
1472 OVS_NLERR(log, "Either Ethernet header or EtherType is required.");
1476 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1477 const struct ovs_key_ipv4 *ipv4_key;
1479 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1480 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
1481 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
1482 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
1485 SW_FLOW_KEY_PUT(match, ip.proto,
1486 ipv4_key->ipv4_proto, is_mask);
1487 SW_FLOW_KEY_PUT(match, ip.tos,
1488 ipv4_key->ipv4_tos, is_mask);
1489 SW_FLOW_KEY_PUT(match, ip.ttl,
1490 ipv4_key->ipv4_ttl, is_mask);
1491 SW_FLOW_KEY_PUT(match, ip.frag,
1492 ipv4_key->ipv4_frag, is_mask);
1493 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1494 ipv4_key->ipv4_src, is_mask);
1495 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1496 ipv4_key->ipv4_dst, is_mask);
1497 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1500 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1501 const struct ovs_key_ipv6 *ipv6_key;
1503 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1504 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
1505 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
1506 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
1510 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
1511 OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x)",
1512 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
1516 SW_FLOW_KEY_PUT(match, ipv6.label,
1517 ipv6_key->ipv6_label, is_mask);
1518 SW_FLOW_KEY_PUT(match, ip.proto,
1519 ipv6_key->ipv6_proto, is_mask);
1520 SW_FLOW_KEY_PUT(match, ip.tos,
1521 ipv6_key->ipv6_tclass, is_mask);
1522 SW_FLOW_KEY_PUT(match, ip.ttl,
1523 ipv6_key->ipv6_hlimit, is_mask);
1524 SW_FLOW_KEY_PUT(match, ip.frag,
1525 ipv6_key->ipv6_frag, is_mask);
1526 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1528 sizeof(match->key->ipv6.addr.src),
1530 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1532 sizeof(match->key->ipv6.addr.dst),
1535 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1538 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1539 const struct ovs_key_arp *arp_key;
1541 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1542 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1543 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
1548 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1549 arp_key->arp_sip, is_mask);
1550 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1551 arp_key->arp_tip, is_mask);
1552 SW_FLOW_KEY_PUT(match, ip.proto,
1553 ntohs(arp_key->arp_op), is_mask);
1554 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1555 arp_key->arp_sha, ETH_ALEN, is_mask);
1556 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1557 arp_key->arp_tha, ETH_ALEN, is_mask);
1559 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1562 if (attrs & (1 << OVS_KEY_ATTR_NSH)) {
1563 if (nsh_key_put_from_nlattr(a[OVS_KEY_ATTR_NSH], match,
1564 is_mask, false, log) < 0)
1566 attrs &= ~(1 << OVS_KEY_ATTR_NSH);
1569 if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1570 const struct ovs_key_mpls *mpls_key;
1572 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1573 SW_FLOW_KEY_PUT(match, mpls.top_lse,
1574 mpls_key->mpls_lse, is_mask);
1576 attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1579 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1580 const struct ovs_key_tcp *tcp_key;
1582 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1583 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1584 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1585 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1588 if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1589 SW_FLOW_KEY_PUT(match, tp.flags,
1590 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1592 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1595 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1596 const struct ovs_key_udp *udp_key;
1598 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1599 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1600 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1601 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1604 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1605 const struct ovs_key_sctp *sctp_key;
1607 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1608 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1609 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1610 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1613 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1614 const struct ovs_key_icmp *icmp_key;
1616 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1617 SW_FLOW_KEY_PUT(match, tp.src,
1618 htons(icmp_key->icmp_type), is_mask);
1619 SW_FLOW_KEY_PUT(match, tp.dst,
1620 htons(icmp_key->icmp_code), is_mask);
1621 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1624 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1625 const struct ovs_key_icmpv6 *icmpv6_key;
1627 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1628 SW_FLOW_KEY_PUT(match, tp.src,
1629 htons(icmpv6_key->icmpv6_type), is_mask);
1630 SW_FLOW_KEY_PUT(match, tp.dst,
1631 htons(icmpv6_key->icmpv6_code), is_mask);
1632 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1635 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1636 const struct ovs_key_nd *nd_key;
1638 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1639 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1641 sizeof(match->key->ipv6.nd.target),
1643 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1644 nd_key->nd_sll, ETH_ALEN, is_mask);
1645 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1646 nd_key->nd_tll, ETH_ALEN, is_mask);
1647 attrs &= ~(1 << OVS_KEY_ATTR_ND);
1651 OVS_NLERR(log, "Unknown key attributes %llx",
1652 (unsigned long long)attrs);
1659 static void nlattr_set(struct nlattr *attr, u8 val,
1660 const struct ovs_len_tbl *tbl)
1665 /* The nlattr stream should already have been validated */
1666 nla_for_each_nested(nla, attr, rem) {
1667 if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED) {
1668 if (tbl[nla_type(nla)].next)
1669 tbl = tbl[nla_type(nla)].next;
1670 nlattr_set(nla, val, tbl);
1672 memset(nla_data(nla), val, nla_len(nla));
1675 if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1676 *(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1680 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1682 nlattr_set(attr, val, ovs_key_lens);
1686 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1687 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1688 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1689 * does not include any don't care bit.
1690 * @net: Used to determine per-namespace field support.
1691 * @match: receives the extracted flow match information.
1692 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1693 * sequence. The fields should of the packet that triggered the creation
1695 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1696 * attribute specifies the mask field of the wildcarded flow.
1697 * @log: Boolean to allow kernel error logging. Normally true, but when
1698 * probing for feature compatibility this should be passed in as false to
1699 * suppress unnecessary error logging.
1701 int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1702 const struct nlattr *nla_key,
1703 const struct nlattr *nla_mask,
1706 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1707 struct nlattr *newmask = NULL;
1712 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1716 err = parse_vlan_from_nlattrs(match, &key_attrs, a, false, log);
1720 err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1726 /* Create an exact match mask. We need to set to 0xff
1727 * all the 'match->mask' fields that have been touched
1728 * in 'match->key'. We cannot simply memset
1729 * 'match->mask', because padding bytes and fields not
1730 * specified in 'match->key' should be left to 0.
1731 * Instead, we use a stream of netlink attributes,
1732 * copied from 'key' and set to 0xff.
1733 * ovs_key_from_nlattrs() will take care of filling
1734 * 'match->mask' appropriately.
1736 newmask = kmemdup(nla_key,
1737 nla_total_size(nla_len(nla_key)),
1742 mask_set_nlattr(newmask, 0xff);
1744 /* The userspace does not send tunnel attributes that
1745 * are 0, but we should not wildcard them nonetheless.
1747 if (match->key->tun_proto)
1748 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1754 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1758 /* Always match on tci. */
1759 SW_FLOW_KEY_PUT(match, eth.vlan.tci, htons(0xffff), true);
1760 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, htons(0xffff), true);
1762 err = parse_vlan_from_nlattrs(match, &mask_attrs, a, true, log);
1766 err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1772 if (!match_validate(match, key_attrs, mask_attrs, log))
1780 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1787 len = nla_len(attr);
1788 if (len < 1 || len > MAX_UFID_LENGTH) {
1789 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1790 nla_len(attr), MAX_UFID_LENGTH);
1797 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1798 * or false otherwise.
1800 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1803 sfid->ufid_len = get_ufid_len(attr, log);
1805 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1807 return sfid->ufid_len;
1810 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1811 const struct sw_flow_key *key, bool log)
1813 struct sw_flow_key *new_key;
1815 if (ovs_nla_get_ufid(sfid, ufid, log))
1818 /* If UFID was not provided, use unmasked key. */
1819 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1822 memcpy(new_key, key, sizeof(*key));
1823 sfid->unmasked_key = new_key;
1828 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1830 return attr ? nla_get_u32(attr) : 0;
1834 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1835 * @net: Network namespace.
1836 * @key: Receives extracted in_port, priority, tun_key, skb_mark and conntrack
1838 * @a: Array of netlink attributes holding parsed %OVS_KEY_ATTR_* Netlink
1840 * @attrs: Bit mask for the netlink attributes included in @a.
1841 * @log: Boolean to allow kernel error logging. Normally true, but when
1842 * probing for feature compatibility this should be passed in as false to
1843 * suppress unnecessary error logging.
1845 * This parses a series of Netlink attributes that form a flow key, which must
1846 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1847 * get the metadata, that is, the parts of the flow key that cannot be
1848 * extracted from the packet itself.
1850 * This must be called before the packet key fields are filled in 'key'.
1853 int ovs_nla_get_flow_metadata(struct net *net,
1854 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1],
1855 u64 attrs, struct sw_flow_key *key, bool log)
1857 struct sw_flow_match match;
1859 memset(&match, 0, sizeof(match));
1864 key->ct_orig_proto = 0;
1865 memset(&key->ct, 0, sizeof(key->ct));
1866 memset(&key->ipv4.ct_orig, 0, sizeof(key->ipv4.ct_orig));
1867 memset(&key->ipv6.ct_orig, 0, sizeof(key->ipv6.ct_orig));
1869 key->phy.in_port = DP_MAX_PORTS;
1871 return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1874 static int ovs_nla_put_vlan(struct sk_buff *skb, const struct vlan_head *vh,
1877 __be16 eth_type = !is_mask ? vh->tpid : htons(0xffff);
1879 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1880 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, vh->tci))
1885 static int nsh_key_to_nlattr(const struct ovs_key_nsh *nsh, bool is_mask,
1886 struct sk_buff *skb)
1888 struct nlattr *start;
1890 start = nla_nest_start(skb, OVS_KEY_ATTR_NSH);
1894 if (nla_put(skb, OVS_NSH_KEY_ATTR_BASE, sizeof(nsh->base), &nsh->base))
1895 goto nla_put_failure;
1897 if (is_mask || nsh->base.mdtype == NSH_M_TYPE1) {
1898 if (nla_put(skb, OVS_NSH_KEY_ATTR_MD1,
1899 sizeof(nsh->context), nsh->context))
1900 goto nla_put_failure;
1903 /* Don't support MD type 2 yet */
1905 nla_nest_end(skb, start);
1913 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1914 const struct sw_flow_key *output, bool is_mask,
1915 struct sk_buff *skb)
1917 struct ovs_key_ethernet *eth_key;
1919 struct nlattr *encap = NULL;
1920 struct nlattr *in_encap = NULL;
1922 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1923 goto nla_put_failure;
1925 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1926 goto nla_put_failure;
1928 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1929 goto nla_put_failure;
1931 if ((swkey->tun_proto || is_mask)) {
1932 const void *opts = NULL;
1934 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1935 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1937 if (ip_tun_to_nlattr(skb, &output->tun_key, opts,
1938 swkey->tun_opts_len, swkey->tun_proto))
1939 goto nla_put_failure;
1942 if (swkey->phy.in_port == DP_MAX_PORTS) {
1943 if (is_mask && (output->phy.in_port == 0xffff))
1944 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1945 goto nla_put_failure;
1948 upper_u16 = !is_mask ? 0 : 0xffff;
1950 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1951 (upper_u16 << 16) | output->phy.in_port))
1952 goto nla_put_failure;
1955 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1956 goto nla_put_failure;
1958 if (ovs_ct_put_key(swkey, output, skb))
1959 goto nla_put_failure;
1961 if (ovs_key_mac_proto(swkey) == MAC_PROTO_ETHERNET) {
1962 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1964 goto nla_put_failure;
1966 eth_key = nla_data(nla);
1967 ether_addr_copy(eth_key->eth_src, output->eth.src);
1968 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1970 if (swkey->eth.vlan.tci || eth_type_vlan(swkey->eth.type)) {
1971 if (ovs_nla_put_vlan(skb, &output->eth.vlan, is_mask))
1972 goto nla_put_failure;
1973 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1974 if (!swkey->eth.vlan.tci)
1977 if (swkey->eth.cvlan.tci || eth_type_vlan(swkey->eth.type)) {
1978 if (ovs_nla_put_vlan(skb, &output->eth.cvlan, is_mask))
1979 goto nla_put_failure;
1980 in_encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1981 if (!swkey->eth.cvlan.tci)
1986 if (swkey->eth.type == htons(ETH_P_802_2)) {
1988 * Ethertype 802.2 is represented in the netlink with omitted
1989 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1990 * 0xffff in the mask attribute. Ethertype can also
1993 if (is_mask && output->eth.type)
1994 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1996 goto nla_put_failure;
2001 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
2002 goto nla_put_failure;
2004 if (eth_type_vlan(swkey->eth.type)) {
2005 /* There are 3 VLAN tags, we don't know anything about the rest
2006 * of the packet, so truncate here.
2008 WARN_ON_ONCE(!(encap && in_encap));
2012 if (swkey->eth.type == htons(ETH_P_IP)) {
2013 struct ovs_key_ipv4 *ipv4_key;
2015 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
2017 goto nla_put_failure;
2018 ipv4_key = nla_data(nla);
2019 ipv4_key->ipv4_src = output->ipv4.addr.src;
2020 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
2021 ipv4_key->ipv4_proto = output->ip.proto;
2022 ipv4_key->ipv4_tos = output->ip.tos;
2023 ipv4_key->ipv4_ttl = output->ip.ttl;
2024 ipv4_key->ipv4_frag = output->ip.frag;
2025 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
2026 struct ovs_key_ipv6 *ipv6_key;
2028 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
2030 goto nla_put_failure;
2031 ipv6_key = nla_data(nla);
2032 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
2033 sizeof(ipv6_key->ipv6_src));
2034 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
2035 sizeof(ipv6_key->ipv6_dst));
2036 ipv6_key->ipv6_label = output->ipv6.label;
2037 ipv6_key->ipv6_proto = output->ip.proto;
2038 ipv6_key->ipv6_tclass = output->ip.tos;
2039 ipv6_key->ipv6_hlimit = output->ip.ttl;
2040 ipv6_key->ipv6_frag = output->ip.frag;
2041 } else if (swkey->eth.type == htons(ETH_P_NSH)) {
2042 if (nsh_key_to_nlattr(&output->nsh, is_mask, skb))
2043 goto nla_put_failure;
2044 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
2045 swkey->eth.type == htons(ETH_P_RARP)) {
2046 struct ovs_key_arp *arp_key;
2048 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
2050 goto nla_put_failure;
2051 arp_key = nla_data(nla);
2052 memset(arp_key, 0, sizeof(struct ovs_key_arp));
2053 arp_key->arp_sip = output->ipv4.addr.src;
2054 arp_key->arp_tip = output->ipv4.addr.dst;
2055 arp_key->arp_op = htons(output->ip.proto);
2056 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
2057 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
2058 } else if (eth_p_mpls(swkey->eth.type)) {
2059 struct ovs_key_mpls *mpls_key;
2061 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
2063 goto nla_put_failure;
2064 mpls_key = nla_data(nla);
2065 mpls_key->mpls_lse = output->mpls.top_lse;
2068 if ((swkey->eth.type == htons(ETH_P_IP) ||
2069 swkey->eth.type == htons(ETH_P_IPV6)) &&
2070 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
2072 if (swkey->ip.proto == IPPROTO_TCP) {
2073 struct ovs_key_tcp *tcp_key;
2075 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
2077 goto nla_put_failure;
2078 tcp_key = nla_data(nla);
2079 tcp_key->tcp_src = output->tp.src;
2080 tcp_key->tcp_dst = output->tp.dst;
2081 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
2083 goto nla_put_failure;
2084 } else if (swkey->ip.proto == IPPROTO_UDP) {
2085 struct ovs_key_udp *udp_key;
2087 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
2089 goto nla_put_failure;
2090 udp_key = nla_data(nla);
2091 udp_key->udp_src = output->tp.src;
2092 udp_key->udp_dst = output->tp.dst;
2093 } else if (swkey->ip.proto == IPPROTO_SCTP) {
2094 struct ovs_key_sctp *sctp_key;
2096 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
2098 goto nla_put_failure;
2099 sctp_key = nla_data(nla);
2100 sctp_key->sctp_src = output->tp.src;
2101 sctp_key->sctp_dst = output->tp.dst;
2102 } else if (swkey->eth.type == htons(ETH_P_IP) &&
2103 swkey->ip.proto == IPPROTO_ICMP) {
2104 struct ovs_key_icmp *icmp_key;
2106 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
2108 goto nla_put_failure;
2109 icmp_key = nla_data(nla);
2110 icmp_key->icmp_type = ntohs(output->tp.src);
2111 icmp_key->icmp_code = ntohs(output->tp.dst);
2112 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
2113 swkey->ip.proto == IPPROTO_ICMPV6) {
2114 struct ovs_key_icmpv6 *icmpv6_key;
2116 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
2117 sizeof(*icmpv6_key));
2119 goto nla_put_failure;
2120 icmpv6_key = nla_data(nla);
2121 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
2122 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
2124 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
2125 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
2126 struct ovs_key_nd *nd_key;
2128 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
2130 goto nla_put_failure;
2131 nd_key = nla_data(nla);
2132 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
2133 sizeof(nd_key->nd_target));
2134 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
2135 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
2142 nla_nest_end(skb, in_encap);
2144 nla_nest_end(skb, encap);
2152 int ovs_nla_put_key(const struct sw_flow_key *swkey,
2153 const struct sw_flow_key *output, int attr, bool is_mask,
2154 struct sk_buff *skb)
2159 nla = nla_nest_start(skb, attr);
2162 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
2165 nla_nest_end(skb, nla);
2170 /* Called with ovs_mutex or RCU read lock. */
2171 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
2173 if (ovs_identifier_is_ufid(&flow->id))
2174 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
2177 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
2178 OVS_FLOW_ATTR_KEY, false, skb);
2181 /* Called with ovs_mutex or RCU read lock. */
2182 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
2184 return ovs_nla_put_key(&flow->key, &flow->key,
2185 OVS_FLOW_ATTR_KEY, false, skb);
2188 /* Called with ovs_mutex or RCU read lock. */
2189 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
2191 return ovs_nla_put_key(&flow->key, &flow->mask->key,
2192 OVS_FLOW_ATTR_MASK, true, skb);
2195 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
2197 static struct sw_flow_actions *nla_alloc_flow_actions(int size)
2199 struct sw_flow_actions *sfa;
2201 WARN_ON_ONCE(size > MAX_ACTIONS_BUFSIZE);
2203 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
2205 return ERR_PTR(-ENOMEM);
2207 sfa->actions_len = 0;
2211 static void ovs_nla_free_set_action(const struct nlattr *a)
2213 const struct nlattr *ovs_key = nla_data(a);
2214 struct ovs_tunnel_info *ovs_tun;
2216 switch (nla_type(ovs_key)) {
2217 case OVS_KEY_ATTR_TUNNEL_INFO:
2218 ovs_tun = nla_data(ovs_key);
2219 dst_release((struct dst_entry *)ovs_tun->tun_dst);
2224 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
2226 const struct nlattr *a;
2232 nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
2233 switch (nla_type(a)) {
2234 case OVS_ACTION_ATTR_SET:
2235 ovs_nla_free_set_action(a);
2237 case OVS_ACTION_ATTR_CT:
2238 ovs_ct_free_action(a);
2246 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
2248 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
2251 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
2252 * The caller must hold rcu_read_lock for this to be sensible. */
2253 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
2255 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
2258 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
2259 int attr_len, bool log)
2262 struct sw_flow_actions *acts;
2264 int req_size = NLA_ALIGN(attr_len);
2265 int next_offset = offsetof(struct sw_flow_actions, actions) +
2266 (*sfa)->actions_len;
2268 if (req_size <= (ksize(*sfa) - next_offset))
2271 new_acts_size = ksize(*sfa) * 2;
2273 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
2274 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size) {
2275 OVS_NLERR(log, "Flow action size exceeds max %u",
2276 MAX_ACTIONS_BUFSIZE);
2277 return ERR_PTR(-EMSGSIZE);
2279 new_acts_size = MAX_ACTIONS_BUFSIZE;
2282 acts = nla_alloc_flow_actions(new_acts_size);
2284 return (void *)acts;
2286 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
2287 acts->actions_len = (*sfa)->actions_len;
2288 acts->orig_len = (*sfa)->orig_len;
2293 (*sfa)->actions_len += req_size;
2294 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
2297 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
2298 int attrtype, void *data, int len, bool log)
2302 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
2306 a->nla_type = attrtype;
2307 a->nla_len = nla_attr_size(len);
2310 memcpy(nla_data(a), data, len);
2311 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
2316 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
2321 a = __add_action(sfa, attrtype, data, len, log);
2323 return PTR_ERR_OR_ZERO(a);
2326 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
2327 int attrtype, bool log)
2329 int used = (*sfa)->actions_len;
2332 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
2339 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
2342 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
2345 a->nla_len = sfa->actions_len - st_offset;
2348 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2349 const struct sw_flow_key *key,
2350 struct sw_flow_actions **sfa,
2351 __be16 eth_type, __be16 vlan_tci, bool log);
2353 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
2354 const struct sw_flow_key *key,
2355 struct sw_flow_actions **sfa,
2356 __be16 eth_type, __be16 vlan_tci,
2357 bool log, bool last)
2359 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
2360 const struct nlattr *probability, *actions;
2361 const struct nlattr *a;
2362 int rem, start, err;
2363 struct sample_arg arg;
2365 memset(attrs, 0, sizeof(attrs));
2366 nla_for_each_nested(a, attr, rem) {
2367 int type = nla_type(a);
2368 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
2375 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
2376 if (!probability || nla_len(probability) != sizeof(u32))
2379 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
2380 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
2383 /* validation done, copy sample action. */
2384 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
2388 /* When both skb and flow may be changed, put the sample
2389 * into a deferred fifo. On the other hand, if only skb
2390 * may be modified, the actions can be executed in place.
2392 * Do this analysis at the flow installation time.
2393 * Set 'clone_action->exec' to true if the actions can be
2394 * executed without being deferred.
2396 * If the sample is the last action, it can always be excuted
2397 * rather than deferred.
2399 arg.exec = last || !actions_may_change_flow(actions);
2400 arg.probability = nla_get_u32(probability);
2402 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_ARG, &arg, sizeof(arg),
2407 err = __ovs_nla_copy_actions(net, actions, key, sfa,
2408 eth_type, vlan_tci, log);
2413 add_nested_action_end(*sfa, start);
2418 void ovs_match_init(struct sw_flow_match *match,
2419 struct sw_flow_key *key,
2421 struct sw_flow_mask *mask)
2423 memset(match, 0, sizeof(*match));
2428 memset(key, 0, sizeof(*key));
2431 memset(&mask->key, 0, sizeof(mask->key));
2432 mask->range.start = mask->range.end = 0;
2436 static int validate_geneve_opts(struct sw_flow_key *key)
2438 struct geneve_opt *option;
2439 int opts_len = key->tun_opts_len;
2440 bool crit_opt = false;
2442 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2443 while (opts_len > 0) {
2446 if (opts_len < sizeof(*option))
2449 len = sizeof(*option) + option->length * 4;
2453 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2455 option = (struct geneve_opt *)((u8 *)option + len);
2459 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
2464 static int validate_and_copy_set_tun(const struct nlattr *attr,
2465 struct sw_flow_actions **sfa, bool log)
2467 struct sw_flow_match match;
2468 struct sw_flow_key key;
2469 struct metadata_dst *tun_dst;
2470 struct ip_tunnel_info *tun_info;
2471 struct ovs_tunnel_info *ovs_tun;
2473 int err = 0, start, opts_type;
2475 ovs_match_init(&match, &key, true, NULL);
2476 opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2480 if (key.tun_opts_len) {
2481 switch (opts_type) {
2482 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2483 err = validate_geneve_opts(&key);
2487 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
2492 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2496 tun_dst = metadata_dst_alloc(key.tun_opts_len, METADATA_IP_TUNNEL,
2502 err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2504 dst_release((struct dst_entry *)tun_dst);
2508 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2509 sizeof(*ovs_tun), log);
2511 dst_release((struct dst_entry *)tun_dst);
2515 ovs_tun = nla_data(a);
2516 ovs_tun->tun_dst = tun_dst;
2518 tun_info = &tun_dst->u.tun_info;
2519 tun_info->mode = IP_TUNNEL_INFO_TX;
2520 if (key.tun_proto == AF_INET6)
2521 tun_info->mode |= IP_TUNNEL_INFO_IPV6;
2522 tun_info->key = key.tun_key;
2524 /* We need to store the options in the action itself since
2525 * everything else will go away after flow setup. We can append
2526 * it to tun_info and then point there.
2528 ip_tunnel_info_opts_set(tun_info,
2529 TUN_METADATA_OPTS(&key, key.tun_opts_len),
2531 add_nested_action_end(*sfa, start);
2536 static bool validate_nsh(const struct nlattr *attr, bool is_mask,
2537 bool is_push_nsh, bool log)
2539 struct sw_flow_match match;
2540 struct sw_flow_key key;
2543 ovs_match_init(&match, &key, true, NULL);
2544 ret = nsh_key_put_from_nlattr(attr, &match, is_mask,
2549 /* Return false if there are any non-masked bits set.
2550 * Mask follows data immediately, before any netlink padding.
2552 static bool validate_masked(u8 *data, int len)
2554 u8 *mask = data + len;
2557 if (*data++ & ~*mask++)
2563 static int validate_set(const struct nlattr *a,
2564 const struct sw_flow_key *flow_key,
2565 struct sw_flow_actions **sfa, bool *skip_copy,
2566 u8 mac_proto, __be16 eth_type, bool masked, bool log)
2568 const struct nlattr *ovs_key = nla_data(a);
2569 int key_type = nla_type(ovs_key);
2572 /* There can be only one key in a action */
2573 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2576 key_len = nla_len(ovs_key);
2580 if (key_type > OVS_KEY_ATTR_MAX ||
2581 !check_attr_len(key_len, ovs_key_lens[key_type].len))
2584 if (masked && !validate_masked(nla_data(ovs_key), key_len))
2588 const struct ovs_key_ipv4 *ipv4_key;
2589 const struct ovs_key_ipv6 *ipv6_key;
2592 case OVS_KEY_ATTR_PRIORITY:
2593 case OVS_KEY_ATTR_SKB_MARK:
2594 case OVS_KEY_ATTR_CT_MARK:
2595 case OVS_KEY_ATTR_CT_LABELS:
2598 case OVS_KEY_ATTR_ETHERNET:
2599 if (mac_proto != MAC_PROTO_ETHERNET)
2603 case OVS_KEY_ATTR_TUNNEL:
2605 return -EINVAL; /* Masked tunnel set not supported. */
2608 err = validate_and_copy_set_tun(a, sfa, log);
2613 case OVS_KEY_ATTR_IPV4:
2614 if (eth_type != htons(ETH_P_IP))
2617 ipv4_key = nla_data(ovs_key);
2620 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2622 /* Non-writeable fields. */
2623 if (mask->ipv4_proto || mask->ipv4_frag)
2626 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2629 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2634 case OVS_KEY_ATTR_IPV6:
2635 if (eth_type != htons(ETH_P_IPV6))
2638 ipv6_key = nla_data(ovs_key);
2641 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2643 /* Non-writeable fields. */
2644 if (mask->ipv6_proto || mask->ipv6_frag)
2647 /* Invalid bits in the flow label mask? */
2648 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2651 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2654 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2657 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2662 case OVS_KEY_ATTR_TCP:
2663 if ((eth_type != htons(ETH_P_IP) &&
2664 eth_type != htons(ETH_P_IPV6)) ||
2665 flow_key->ip.proto != IPPROTO_TCP)
2670 case OVS_KEY_ATTR_UDP:
2671 if ((eth_type != htons(ETH_P_IP) &&
2672 eth_type != htons(ETH_P_IPV6)) ||
2673 flow_key->ip.proto != IPPROTO_UDP)
2678 case OVS_KEY_ATTR_MPLS:
2679 if (!eth_p_mpls(eth_type))
2683 case OVS_KEY_ATTR_SCTP:
2684 if ((eth_type != htons(ETH_P_IP) &&
2685 eth_type != htons(ETH_P_IPV6)) ||
2686 flow_key->ip.proto != IPPROTO_SCTP)
2691 case OVS_KEY_ATTR_NSH:
2692 if (eth_type != htons(ETH_P_NSH))
2694 if (!validate_nsh(nla_data(a), masked, false, log))
2702 /* Convert non-masked non-tunnel set actions to masked set actions. */
2703 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2704 int start, len = key_len * 2;
2709 start = add_nested_action_start(sfa,
2710 OVS_ACTION_ATTR_SET_TO_MASKED,
2715 at = __add_action(sfa, key_type, NULL, len, log);
2719 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2720 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
2721 /* Clear non-writeable bits from otherwise writeable fields. */
2722 if (key_type == OVS_KEY_ATTR_IPV6) {
2723 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2725 mask->ipv6_label &= htonl(0x000FFFFF);
2727 add_nested_action_end(*sfa, start);
2733 static int validate_userspace(const struct nlattr *attr)
2735 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2736 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2737 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2738 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2740 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2743 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX, attr,
2744 userspace_policy, NULL);
2748 if (!a[OVS_USERSPACE_ATTR_PID] ||
2749 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2755 static int copy_action(const struct nlattr *from,
2756 struct sw_flow_actions **sfa, bool log)
2758 int totlen = NLA_ALIGN(from->nla_len);
2761 to = reserve_sfa_size(sfa, from->nla_len, log);
2765 memcpy(to, from, totlen);
2769 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2770 const struct sw_flow_key *key,
2771 struct sw_flow_actions **sfa,
2772 __be16 eth_type, __be16 vlan_tci, bool log)
2774 u8 mac_proto = ovs_key_mac_proto(key);
2775 const struct nlattr *a;
2778 nla_for_each_nested(a, attr, rem) {
2779 /* Expected argument lengths, (u32)-1 for variable length. */
2780 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2781 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2782 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2783 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2784 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2785 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2786 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2787 [OVS_ACTION_ATTR_POP_VLAN] = 0,
2788 [OVS_ACTION_ATTR_SET] = (u32)-1,
2789 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2790 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2791 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2792 [OVS_ACTION_ATTR_CT] = (u32)-1,
2793 [OVS_ACTION_ATTR_CT_CLEAR] = 0,
2794 [OVS_ACTION_ATTR_TRUNC] = sizeof(struct ovs_action_trunc),
2795 [OVS_ACTION_ATTR_PUSH_ETH] = sizeof(struct ovs_action_push_eth),
2796 [OVS_ACTION_ATTR_POP_ETH] = 0,
2797 [OVS_ACTION_ATTR_PUSH_NSH] = (u32)-1,
2798 [OVS_ACTION_ATTR_POP_NSH] = 0,
2799 [OVS_ACTION_ATTR_METER] = sizeof(u32),
2801 const struct ovs_action_push_vlan *vlan;
2802 int type = nla_type(a);
2805 if (type > OVS_ACTION_ATTR_MAX ||
2806 (action_lens[type] != nla_len(a) &&
2807 action_lens[type] != (u32)-1))
2812 case OVS_ACTION_ATTR_UNSPEC:
2815 case OVS_ACTION_ATTR_USERSPACE:
2816 err = validate_userspace(a);
2821 case OVS_ACTION_ATTR_OUTPUT:
2822 if (nla_get_u32(a) >= DP_MAX_PORTS)
2826 case OVS_ACTION_ATTR_TRUNC: {
2827 const struct ovs_action_trunc *trunc = nla_data(a);
2829 if (trunc->max_len < ETH_HLEN)
2834 case OVS_ACTION_ATTR_HASH: {
2835 const struct ovs_action_hash *act_hash = nla_data(a);
2837 switch (act_hash->hash_alg) {
2838 case OVS_HASH_ALG_L4:
2847 case OVS_ACTION_ATTR_POP_VLAN:
2848 if (mac_proto != MAC_PROTO_ETHERNET)
2850 vlan_tci = htons(0);
2853 case OVS_ACTION_ATTR_PUSH_VLAN:
2854 if (mac_proto != MAC_PROTO_ETHERNET)
2857 if (!eth_type_vlan(vlan->vlan_tpid))
2859 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2861 vlan_tci = vlan->vlan_tci;
2864 case OVS_ACTION_ATTR_RECIRC:
2867 case OVS_ACTION_ATTR_PUSH_MPLS: {
2868 const struct ovs_action_push_mpls *mpls = nla_data(a);
2870 if (!eth_p_mpls(mpls->mpls_ethertype))
2872 /* Prohibit push MPLS other than to a white list
2873 * for packets that have a known tag order.
2875 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2876 (eth_type != htons(ETH_P_IP) &&
2877 eth_type != htons(ETH_P_IPV6) &&
2878 eth_type != htons(ETH_P_ARP) &&
2879 eth_type != htons(ETH_P_RARP) &&
2880 !eth_p_mpls(eth_type)))
2882 eth_type = mpls->mpls_ethertype;
2886 case OVS_ACTION_ATTR_POP_MPLS:
2887 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2888 !eth_p_mpls(eth_type))
2891 /* Disallow subsequent L2.5+ set and mpls_pop actions
2892 * as there is no check here to ensure that the new
2893 * eth_type is valid and thus set actions could
2894 * write off the end of the packet or otherwise
2897 * Support for these actions is planned using packet
2900 eth_type = htons(0);
2903 case OVS_ACTION_ATTR_SET:
2904 err = validate_set(a, key, sfa,
2905 &skip_copy, mac_proto, eth_type,
2911 case OVS_ACTION_ATTR_SET_MASKED:
2912 err = validate_set(a, key, sfa,
2913 &skip_copy, mac_proto, eth_type,
2919 case OVS_ACTION_ATTR_SAMPLE: {
2920 bool last = nla_is_last(a, rem);
2922 err = validate_and_copy_sample(net, a, key, sfa,
2931 case OVS_ACTION_ATTR_CT:
2932 err = ovs_ct_copy_action(net, a, key, sfa, log);
2938 case OVS_ACTION_ATTR_CT_CLEAR:
2941 case OVS_ACTION_ATTR_PUSH_ETH:
2942 /* Disallow pushing an Ethernet header if one
2943 * is already present */
2944 if (mac_proto != MAC_PROTO_NONE)
2946 mac_proto = MAC_PROTO_NONE;
2949 case OVS_ACTION_ATTR_POP_ETH:
2950 if (mac_proto != MAC_PROTO_ETHERNET)
2952 if (vlan_tci & htons(VLAN_TAG_PRESENT))
2954 mac_proto = MAC_PROTO_ETHERNET;
2957 case OVS_ACTION_ATTR_PUSH_NSH:
2958 if (mac_proto != MAC_PROTO_ETHERNET) {
2961 next_proto = tun_p_from_eth_p(eth_type);
2965 mac_proto = MAC_PROTO_NONE;
2966 if (!validate_nsh(nla_data(a), false, true, true))
2970 case OVS_ACTION_ATTR_POP_NSH: {
2973 if (eth_type != htons(ETH_P_NSH))
2975 inner_proto = tun_p_to_eth_p(key->nsh.base.np);
2978 if (key->nsh.base.np == TUN_P_ETHERNET)
2979 mac_proto = MAC_PROTO_ETHERNET;
2981 mac_proto = MAC_PROTO_NONE;
2985 case OVS_ACTION_ATTR_METER:
2986 /* Non-existent meters are simply ignored. */
2990 OVS_NLERR(log, "Unknown Action type %d", type);
2994 err = copy_action(a, sfa, log);
3006 /* 'key' must be the masked key. */
3007 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3008 const struct sw_flow_key *key,
3009 struct sw_flow_actions **sfa, bool log)
3013 *sfa = nla_alloc_flow_actions(min(nla_len(attr), MAX_ACTIONS_BUFSIZE));
3015 return PTR_ERR(*sfa);
3017 (*sfa)->orig_len = nla_len(attr);
3018 err = __ovs_nla_copy_actions(net, attr, key, sfa, key->eth.type,
3019 key->eth.vlan.tci, log);
3021 ovs_nla_free_flow_actions(*sfa);
3026 static int sample_action_to_attr(const struct nlattr *attr,
3027 struct sk_buff *skb)
3029 struct nlattr *start, *ac_start = NULL, *sample_arg;
3030 int err = 0, rem = nla_len(attr);
3031 const struct sample_arg *arg;
3032 struct nlattr *actions;
3034 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
3038 sample_arg = nla_data(attr);
3039 arg = nla_data(sample_arg);
3040 actions = nla_next(sample_arg, &rem);
3042 if (nla_put_u32(skb, OVS_SAMPLE_ATTR_PROBABILITY, arg->probability)) {
3047 ac_start = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
3053 err = ovs_nla_put_actions(actions, rem, skb);
3057 nla_nest_cancel(skb, ac_start);
3058 nla_nest_cancel(skb, start);
3060 nla_nest_end(skb, ac_start);
3061 nla_nest_end(skb, start);
3067 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
3069 const struct nlattr *ovs_key = nla_data(a);
3070 int key_type = nla_type(ovs_key);
3071 struct nlattr *start;
3075 case OVS_KEY_ATTR_TUNNEL_INFO: {
3076 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
3077 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
3079 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
3083 err = ip_tun_to_nlattr(skb, &tun_info->key,
3084 ip_tunnel_info_opts(tun_info),
3085 tun_info->options_len,
3086 ip_tunnel_info_af(tun_info));
3089 nla_nest_end(skb, start);
3093 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
3101 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
3102 struct sk_buff *skb)
3104 const struct nlattr *ovs_key = nla_data(a);
3106 size_t key_len = nla_len(ovs_key) / 2;
3108 /* Revert the conversion we did from a non-masked set action to
3109 * masked set action.
3111 nla = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
3115 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
3118 nla_nest_end(skb, nla);
3122 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
3124 const struct nlattr *a;
3127 nla_for_each_attr(a, attr, len, rem) {
3128 int type = nla_type(a);
3131 case OVS_ACTION_ATTR_SET:
3132 err = set_action_to_attr(a, skb);
3137 case OVS_ACTION_ATTR_SET_TO_MASKED:
3138 err = masked_set_action_to_set_action_attr(a, skb);
3143 case OVS_ACTION_ATTR_SAMPLE:
3144 err = sample_action_to_attr(a, skb);
3149 case OVS_ACTION_ATTR_CT:
3150 err = ovs_ct_action_to_attr(nla_data(a), skb);
3156 if (nla_put(skb, type, nla_len(a), nla_data(a)))