2 * Copyright (c) 2007-2014 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>
52 #include "flow_netlink.h"
56 const struct ovs_len_tbl *next;
59 #define OVS_ATTR_NESTED -1
60 #define OVS_ATTR_VARIABLE -2
62 static void update_range(struct sw_flow_match *match,
63 size_t offset, size_t size, bool is_mask)
65 struct sw_flow_key_range *range;
66 size_t start = rounddown(offset, sizeof(long));
67 size_t end = roundup(offset + size, sizeof(long));
70 range = &match->range;
72 range = &match->mask->range;
74 if (range->start == range->end) {
80 if (range->start > start)
87 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
89 update_range(match, offsetof(struct sw_flow_key, field), \
90 sizeof((match)->key->field), is_mask); \
92 (match)->mask->key.field = value; \
94 (match)->key->field = value; \
97 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
99 update_range(match, offset, len, is_mask); \
101 memcpy((u8 *)&(match)->mask->key + offset, value_p, \
104 memcpy((u8 *)(match)->key + offset, value_p, len); \
107 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
108 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
109 value_p, len, is_mask)
111 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
113 update_range(match, offsetof(struct sw_flow_key, field), \
114 sizeof((match)->key->field), is_mask); \
116 memset((u8 *)&(match)->mask->key.field, value, \
117 sizeof((match)->mask->key.field)); \
119 memset((u8 *)&(match)->key->field, value, \
120 sizeof((match)->key->field)); \
123 static bool match_validate(const struct sw_flow_match *match,
124 u64 key_attrs, u64 mask_attrs, bool log)
126 u64 key_expected = 0;
127 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
129 /* The following mask attributes allowed only if they
130 * pass the validation tests. */
131 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
132 | (1 << OVS_KEY_ATTR_IPV6)
133 | (1 << OVS_KEY_ATTR_TCP)
134 | (1 << OVS_KEY_ATTR_TCP_FLAGS)
135 | (1 << OVS_KEY_ATTR_UDP)
136 | (1 << OVS_KEY_ATTR_SCTP)
137 | (1 << OVS_KEY_ATTR_ICMP)
138 | (1 << OVS_KEY_ATTR_ICMPV6)
139 | (1 << OVS_KEY_ATTR_ARP)
140 | (1 << OVS_KEY_ATTR_ND)
141 | (1 << OVS_KEY_ATTR_MPLS));
143 /* Always allowed mask fields. */
144 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
145 | (1 << OVS_KEY_ATTR_IN_PORT)
146 | (1 << OVS_KEY_ATTR_ETHERTYPE));
148 /* Check key attributes. */
149 if (match->key->eth.type == htons(ETH_P_ARP)
150 || match->key->eth.type == htons(ETH_P_RARP)) {
151 key_expected |= 1 << OVS_KEY_ATTR_ARP;
152 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
153 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
156 if (eth_p_mpls(match->key->eth.type)) {
157 key_expected |= 1 << OVS_KEY_ATTR_MPLS;
158 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
159 mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
162 if (match->key->eth.type == htons(ETH_P_IP)) {
163 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
164 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
165 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
167 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
168 if (match->key->ip.proto == IPPROTO_UDP) {
169 key_expected |= 1 << OVS_KEY_ATTR_UDP;
170 if (match->mask && (match->mask->key.ip.proto == 0xff))
171 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
174 if (match->key->ip.proto == IPPROTO_SCTP) {
175 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
176 if (match->mask && (match->mask->key.ip.proto == 0xff))
177 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
180 if (match->key->ip.proto == IPPROTO_TCP) {
181 key_expected |= 1 << OVS_KEY_ATTR_TCP;
182 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
183 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
184 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
185 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
189 if (match->key->ip.proto == IPPROTO_ICMP) {
190 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
191 if (match->mask && (match->mask->key.ip.proto == 0xff))
192 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
197 if (match->key->eth.type == htons(ETH_P_IPV6)) {
198 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
199 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
200 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
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_ICMPV6) {
225 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
226 if (match->mask && (match->mask->key.ip.proto == 0xff))
227 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
229 if (match->key->tp.src ==
230 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
231 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
232 key_expected |= 1 << OVS_KEY_ATTR_ND;
233 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
234 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
240 if ((key_attrs & key_expected) != key_expected) {
241 /* Key attributes check failed. */
242 OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
243 (unsigned long long)key_attrs,
244 (unsigned long long)key_expected);
248 if ((mask_attrs & mask_allowed) != mask_attrs) {
249 /* Mask attributes check failed. */
250 OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
251 (unsigned long long)mask_attrs,
252 (unsigned long long)mask_allowed);
259 size_t ovs_tun_key_attr_size(void)
261 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
262 * updating this function.
264 return nla_total_size_64bit(8) /* OVS_TUNNEL_KEY_ATTR_ID */
265 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
266 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_DST */
267 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
268 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
269 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
270 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
271 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
272 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
273 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS is mutually exclusive with
274 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
276 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
277 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
280 size_t ovs_key_attr_size(void)
282 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
283 * updating this function.
285 BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO != 26);
287 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
288 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
289 + ovs_tun_key_attr_size()
290 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
291 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
292 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
293 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
294 + nla_total_size(4) /* OVS_KEY_ATTR_CT_STATE */
295 + nla_total_size(2) /* OVS_KEY_ATTR_CT_ZONE */
296 + nla_total_size(4) /* OVS_KEY_ATTR_CT_MARK */
297 + nla_total_size(16) /* OVS_KEY_ATTR_CT_LABELS */
298 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
299 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
300 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
301 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
302 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
303 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
304 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
305 + nla_total_size(28); /* OVS_KEY_ATTR_ND */
308 static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
309 [OVS_VXLAN_EXT_GBP] = { .len = sizeof(u32) },
312 static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
313 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
314 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
315 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
316 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
317 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
318 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
319 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
320 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
321 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
322 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
323 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_VARIABLE },
324 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED,
325 .next = ovs_vxlan_ext_key_lens },
326 [OVS_TUNNEL_KEY_ATTR_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
327 [OVS_TUNNEL_KEY_ATTR_IPV6_DST] = { .len = sizeof(struct in6_addr) },
330 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
331 static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
332 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
333 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
334 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
335 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
336 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
337 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
338 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
339 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
340 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
341 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
342 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
343 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
344 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
345 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
346 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
347 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
348 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
349 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
350 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
351 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
352 .next = ovs_tunnel_key_lens, },
353 [OVS_KEY_ATTR_MPLS] = { .len = sizeof(struct ovs_key_mpls) },
354 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u32) },
355 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
356 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
357 [OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
360 static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
362 return expected_len == attr_len ||
363 expected_len == OVS_ATTR_NESTED ||
364 expected_len == OVS_ATTR_VARIABLE;
367 static bool is_all_zero(const u8 *fp, size_t size)
374 for (i = 0; i < size; i++)
381 static int __parse_flow_nlattrs(const struct nlattr *attr,
382 const struct nlattr *a[],
383 u64 *attrsp, bool log, bool nz)
385 const struct nlattr *nla;
390 nla_for_each_nested(nla, attr, rem) {
391 u16 type = nla_type(nla);
394 if (type > OVS_KEY_ATTR_MAX) {
395 OVS_NLERR(log, "Key type %d is out of range max %d",
396 type, OVS_KEY_ATTR_MAX);
400 if (attrs & (1 << type)) {
401 OVS_NLERR(log, "Duplicate key (type %d).", type);
405 expected_len = ovs_key_lens[type].len;
406 if (!check_attr_len(nla_len(nla), expected_len)) {
407 OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
408 type, nla_len(nla), expected_len);
412 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
418 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
426 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
427 const struct nlattr *a[], u64 *attrsp,
430 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
433 static int parse_flow_nlattrs(const struct nlattr *attr,
434 const struct nlattr *a[], u64 *attrsp,
437 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
440 static int genev_tun_opt_from_nlattr(const struct nlattr *a,
441 struct sw_flow_match *match, bool is_mask,
444 unsigned long opt_key_offset;
446 if (nla_len(a) > sizeof(match->key->tun_opts)) {
447 OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
448 nla_len(a), sizeof(match->key->tun_opts));
452 if (nla_len(a) % 4 != 0) {
453 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
458 /* We need to record the length of the options passed
459 * down, otherwise packets with the same format but
460 * additional options will be silently matched.
463 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
466 /* This is somewhat unusual because it looks at
467 * both the key and mask while parsing the
468 * attributes (and by extension assumes the key
469 * is parsed first). Normally, we would verify
470 * that each is the correct length and that the
471 * attributes line up in the validate function.
472 * However, that is difficult because this is
473 * variable length and we won't have the
476 if (match->key->tun_opts_len != nla_len(a)) {
477 OVS_NLERR(log, "Geneve option len %d != mask len %d",
478 match->key->tun_opts_len, nla_len(a));
482 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
485 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
486 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
487 nla_len(a), is_mask);
491 static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
492 struct sw_flow_match *match, bool is_mask,
497 unsigned long opt_key_offset;
498 struct vxlan_metadata opts;
500 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
502 memset(&opts, 0, sizeof(opts));
503 nla_for_each_nested(a, attr, rem) {
504 int type = nla_type(a);
506 if (type > OVS_VXLAN_EXT_MAX) {
507 OVS_NLERR(log, "VXLAN extension %d out of range max %d",
508 type, OVS_VXLAN_EXT_MAX);
512 if (!check_attr_len(nla_len(a),
513 ovs_vxlan_ext_key_lens[type].len)) {
514 OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
516 ovs_vxlan_ext_key_lens[type].len);
521 case OVS_VXLAN_EXT_GBP:
522 opts.gbp = nla_get_u32(a);
525 OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
531 OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
537 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
539 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
541 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
542 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
547 static int ip_tun_from_nlattr(const struct nlattr *attr,
548 struct sw_flow_match *match, bool is_mask,
551 bool ttl = false, ipv4 = false, ipv6 = false;
552 __be16 tun_flags = 0;
557 nla_for_each_nested(a, attr, rem) {
558 int type = nla_type(a);
561 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
562 OVS_NLERR(log, "Tunnel attr %d out of range max %d",
563 type, OVS_TUNNEL_KEY_ATTR_MAX);
567 if (!check_attr_len(nla_len(a),
568 ovs_tunnel_key_lens[type].len)) {
569 OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
570 type, nla_len(a), ovs_tunnel_key_lens[type].len);
575 case OVS_TUNNEL_KEY_ATTR_ID:
576 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
577 nla_get_be64(a), is_mask);
578 tun_flags |= TUNNEL_KEY;
580 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
581 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
582 nla_get_in_addr(a), is_mask);
585 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
586 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
587 nla_get_in_addr(a), is_mask);
590 case OVS_TUNNEL_KEY_ATTR_IPV6_SRC:
591 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
592 nla_get_in6_addr(a), is_mask);
595 case OVS_TUNNEL_KEY_ATTR_IPV6_DST:
596 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
597 nla_get_in6_addr(a), is_mask);
600 case OVS_TUNNEL_KEY_ATTR_TOS:
601 SW_FLOW_KEY_PUT(match, tun_key.tos,
602 nla_get_u8(a), is_mask);
604 case OVS_TUNNEL_KEY_ATTR_TTL:
605 SW_FLOW_KEY_PUT(match, tun_key.ttl,
606 nla_get_u8(a), is_mask);
609 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
610 tun_flags |= TUNNEL_DONT_FRAGMENT;
612 case OVS_TUNNEL_KEY_ATTR_CSUM:
613 tun_flags |= TUNNEL_CSUM;
615 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
616 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
617 nla_get_be16(a), is_mask);
619 case OVS_TUNNEL_KEY_ATTR_TP_DST:
620 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
621 nla_get_be16(a), is_mask);
623 case OVS_TUNNEL_KEY_ATTR_OAM:
624 tun_flags |= TUNNEL_OAM;
626 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
628 OVS_NLERR(log, "Multiple metadata blocks provided");
632 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
636 tun_flags |= TUNNEL_GENEVE_OPT;
639 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
641 OVS_NLERR(log, "Multiple metadata blocks provided");
645 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
649 tun_flags |= TUNNEL_VXLAN_OPT;
653 OVS_NLERR(log, "Unknown IP tunnel attribute %d",
659 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
661 SW_FLOW_KEY_MEMSET_FIELD(match, tun_proto, 0xff, true);
663 SW_FLOW_KEY_PUT(match, tun_proto, ipv6 ? AF_INET6 : AF_INET,
667 OVS_NLERR(log, "IP tunnel attribute has %d unknown bytes.",
673 OVS_NLERR(log, "Mixed IPv4 and IPv6 tunnel attributes");
678 if (!ipv4 && !ipv6) {
679 OVS_NLERR(log, "IP tunnel dst address not specified");
682 if (ipv4 && !match->key->tun_key.u.ipv4.dst) {
683 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
686 if (ipv6 && ipv6_addr_any(&match->key->tun_key.u.ipv6.dst)) {
687 OVS_NLERR(log, "IPv6 tunnel dst address is zero");
692 OVS_NLERR(log, "IP tunnel TTL not specified.");
700 static int vxlan_opt_to_nlattr(struct sk_buff *skb,
701 const void *tun_opts, int swkey_tun_opts_len)
703 const struct vxlan_metadata *opts = tun_opts;
706 nla = nla_nest_start(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
710 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
713 nla_nest_end(skb, nla);
717 static int __ip_tun_to_nlattr(struct sk_buff *skb,
718 const struct ip_tunnel_key *output,
719 const void *tun_opts, int swkey_tun_opts_len,
720 unsigned short tun_proto)
722 if (output->tun_flags & TUNNEL_KEY &&
723 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id,
724 OVS_TUNNEL_KEY_ATTR_PAD))
728 if (output->u.ipv4.src &&
729 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
732 if (output->u.ipv4.dst &&
733 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
738 if (!ipv6_addr_any(&output->u.ipv6.src) &&
739 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_SRC,
740 &output->u.ipv6.src))
742 if (!ipv6_addr_any(&output->u.ipv6.dst) &&
743 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_DST,
744 &output->u.ipv6.dst))
749 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
751 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
753 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
754 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
756 if ((output->tun_flags & TUNNEL_CSUM) &&
757 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
759 if (output->tp_src &&
760 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
762 if (output->tp_dst &&
763 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
765 if ((output->tun_flags & TUNNEL_OAM) &&
766 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
768 if (swkey_tun_opts_len) {
769 if (output->tun_flags & TUNNEL_GENEVE_OPT &&
770 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
771 swkey_tun_opts_len, tun_opts))
773 else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
774 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
781 static int ip_tun_to_nlattr(struct sk_buff *skb,
782 const struct ip_tunnel_key *output,
783 const void *tun_opts, int swkey_tun_opts_len,
784 unsigned short tun_proto)
789 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
793 err = __ip_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len,
798 nla_nest_end(skb, nla);
802 int ovs_nla_put_tunnel_info(struct sk_buff *skb,
803 struct ip_tunnel_info *tun_info)
805 return __ip_tun_to_nlattr(skb, &tun_info->key,
806 ip_tunnel_info_opts(tun_info),
807 tun_info->options_len,
808 ip_tunnel_info_af(tun_info));
811 static int encode_vlan_from_nlattrs(struct sw_flow_match *match,
812 const struct nlattr *a[],
813 bool is_mask, bool inner)
818 if (a[OVS_KEY_ATTR_VLAN])
819 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
821 if (a[OVS_KEY_ATTR_ETHERTYPE])
822 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
824 if (likely(!inner)) {
825 SW_FLOW_KEY_PUT(match, eth.vlan.tpid, tpid, is_mask);
826 SW_FLOW_KEY_PUT(match, eth.vlan.tci, tci, is_mask);
828 SW_FLOW_KEY_PUT(match, eth.cvlan.tpid, tpid, is_mask);
829 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, tci, is_mask);
834 static int validate_vlan_from_nlattrs(const struct sw_flow_match *match,
835 u64 key_attrs, bool inner,
836 const struct nlattr **a, bool log)
840 if (!((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
841 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
842 eth_type_vlan(nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE])))) {
847 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
848 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
849 OVS_NLERR(log, "Invalid %s frame", (inner) ? "C-VLAN" : "VLAN");
853 if (a[OVS_KEY_ATTR_VLAN])
854 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
856 if (!(tci & htons(VLAN_TAG_PRESENT))) {
858 OVS_NLERR(log, "%s TCI does not have VLAN_TAG_PRESENT bit set.",
859 (inner) ? "C-VLAN" : "VLAN");
861 } else if (nla_len(a[OVS_KEY_ATTR_ENCAP])) {
862 /* Corner case for truncated VLAN header. */
863 OVS_NLERR(log, "Truncated %s header has non-zero encap attribute.",
864 (inner) ? "C-VLAN" : "VLAN");
872 static int validate_vlan_mask_from_nlattrs(const struct sw_flow_match *match,
873 u64 key_attrs, bool inner,
874 const struct nlattr **a, bool log)
878 bool encap_valid = !!(match->key->eth.vlan.tci &
879 htons(VLAN_TAG_PRESENT));
880 bool i_encap_valid = !!(match->key->eth.cvlan.tci &
881 htons(VLAN_TAG_PRESENT));
883 if (!(key_attrs & (1 << OVS_KEY_ATTR_ENCAP))) {
888 if ((!inner && !encap_valid) || (inner && !i_encap_valid)) {
889 OVS_NLERR(log, "Encap mask attribute is set for non-%s frame.",
890 (inner) ? "C-VLAN" : "VLAN");
894 if (a[OVS_KEY_ATTR_VLAN])
895 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
897 if (a[OVS_KEY_ATTR_ETHERTYPE])
898 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
900 if (tpid != htons(0xffff)) {
901 OVS_NLERR(log, "Must have an exact match on %s TPID (mask=%x).",
902 (inner) ? "C-VLAN" : "VLAN", ntohs(tpid));
905 if (!(tci & htons(VLAN_TAG_PRESENT))) {
906 OVS_NLERR(log, "%s TCI mask does not have exact match for VLAN_TAG_PRESENT bit.",
907 (inner) ? "C-VLAN" : "VLAN");
914 static int __parse_vlan_from_nlattrs(struct sw_flow_match *match,
915 u64 *key_attrs, bool inner,
916 const struct nlattr **a, bool is_mask,
920 const struct nlattr *encap;
923 err = validate_vlan_from_nlattrs(match, *key_attrs, inner,
926 err = validate_vlan_mask_from_nlattrs(match, *key_attrs, inner,
931 err = encode_vlan_from_nlattrs(match, a, is_mask, inner);
935 *key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
936 *key_attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
937 *key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
939 encap = a[OVS_KEY_ATTR_ENCAP];
942 err = parse_flow_nlattrs(encap, a, key_attrs, log);
944 err = parse_flow_mask_nlattrs(encap, a, key_attrs, log);
949 static int parse_vlan_from_nlattrs(struct sw_flow_match *match,
950 u64 *key_attrs, const struct nlattr **a,
951 bool is_mask, bool log)
954 bool encap_valid = false;
956 err = __parse_vlan_from_nlattrs(match, key_attrs, false, a,
961 encap_valid = !!(match->key->eth.vlan.tci & htons(VLAN_TAG_PRESENT));
963 err = __parse_vlan_from_nlattrs(match, key_attrs, true, a,
972 static int parse_eth_type_from_nlattrs(struct sw_flow_match *match,
973 u64 *attrs, const struct nlattr **a,
974 bool is_mask, bool log)
978 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
980 /* Always exact match EtherType. */
981 eth_type = htons(0xffff);
982 } else if (!eth_proto_is_802_3(eth_type)) {
983 OVS_NLERR(log, "EtherType %x is less than min %x",
984 ntohs(eth_type), ETH_P_802_3_MIN);
988 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
989 *attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
993 static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
994 u64 *attrs, const struct nlattr **a,
995 bool is_mask, bool log)
997 u8 mac_proto = MAC_PROTO_ETHERNET;
999 if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
1000 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
1002 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
1003 *attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
1006 if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
1007 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
1009 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
1010 *attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
1013 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
1014 SW_FLOW_KEY_PUT(match, phy.priority,
1015 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
1016 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
1019 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
1020 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1023 in_port = 0xffffffff; /* Always exact match in_port. */
1024 } else if (in_port >= DP_MAX_PORTS) {
1025 OVS_NLERR(log, "Port %d exceeds max allowable %d",
1026 in_port, DP_MAX_PORTS);
1030 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
1031 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1032 } else if (!is_mask) {
1033 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
1036 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1037 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1039 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1040 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1042 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1043 if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1046 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1049 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
1050 ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
1051 u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
1053 if (ct_state & ~CT_SUPPORTED_MASK) {
1054 OVS_NLERR(log, "ct_state flags %08x unsupported",
1059 SW_FLOW_KEY_PUT(match, ct.state, ct_state, is_mask);
1060 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
1062 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
1063 ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
1064 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
1066 SW_FLOW_KEY_PUT(match, ct.zone, ct_zone, is_mask);
1067 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
1069 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
1070 ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
1071 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
1073 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
1074 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
1076 if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
1077 ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
1078 const struct ovs_key_ct_labels *cl;
1080 cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
1081 SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
1082 sizeof(*cl), is_mask);
1083 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
1086 /* For layer 3 packets the Ethernet type is provided
1087 * and treated as metadata but no MAC addresses are provided.
1089 if (!(*attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
1090 (*attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)))
1091 mac_proto = MAC_PROTO_NONE;
1093 /* Always exact match mac_proto */
1094 SW_FLOW_KEY_PUT(match, mac_proto, is_mask ? 0xff : mac_proto, is_mask);
1096 if (mac_proto == MAC_PROTO_NONE)
1097 return parse_eth_type_from_nlattrs(match, attrs, a, is_mask,
1103 static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
1104 u64 attrs, const struct nlattr **a,
1105 bool is_mask, bool log)
1109 err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
1113 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1114 const struct ovs_key_ethernet *eth_key;
1116 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1117 SW_FLOW_KEY_MEMCPY(match, eth.src,
1118 eth_key->eth_src, ETH_ALEN, is_mask);
1119 SW_FLOW_KEY_MEMCPY(match, eth.dst,
1120 eth_key->eth_dst, ETH_ALEN, is_mask);
1121 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1123 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1124 /* VLAN attribute is always parsed before getting here since it
1125 * may occur multiple times.
1127 OVS_NLERR(log, "VLAN attribute unexpected.");
1131 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1132 err = parse_eth_type_from_nlattrs(match, &attrs, a, is_mask,
1136 } else if (!is_mask) {
1137 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1139 } else if (!match->key->eth.type) {
1140 OVS_NLERR(log, "Either Ethernet header or EtherType is required.");
1144 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1145 const struct ovs_key_ipv4 *ipv4_key;
1147 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1148 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
1149 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
1150 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
1153 SW_FLOW_KEY_PUT(match, ip.proto,
1154 ipv4_key->ipv4_proto, is_mask);
1155 SW_FLOW_KEY_PUT(match, ip.tos,
1156 ipv4_key->ipv4_tos, is_mask);
1157 SW_FLOW_KEY_PUT(match, ip.ttl,
1158 ipv4_key->ipv4_ttl, is_mask);
1159 SW_FLOW_KEY_PUT(match, ip.frag,
1160 ipv4_key->ipv4_frag, is_mask);
1161 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1162 ipv4_key->ipv4_src, is_mask);
1163 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1164 ipv4_key->ipv4_dst, is_mask);
1165 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1168 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1169 const struct ovs_key_ipv6 *ipv6_key;
1171 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1172 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
1173 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
1174 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
1178 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
1179 OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x).\n",
1180 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
1184 SW_FLOW_KEY_PUT(match, ipv6.label,
1185 ipv6_key->ipv6_label, is_mask);
1186 SW_FLOW_KEY_PUT(match, ip.proto,
1187 ipv6_key->ipv6_proto, is_mask);
1188 SW_FLOW_KEY_PUT(match, ip.tos,
1189 ipv6_key->ipv6_tclass, is_mask);
1190 SW_FLOW_KEY_PUT(match, ip.ttl,
1191 ipv6_key->ipv6_hlimit, is_mask);
1192 SW_FLOW_KEY_PUT(match, ip.frag,
1193 ipv6_key->ipv6_frag, is_mask);
1194 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1196 sizeof(match->key->ipv6.addr.src),
1198 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1200 sizeof(match->key->ipv6.addr.dst),
1203 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1206 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1207 const struct ovs_key_arp *arp_key;
1209 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1210 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1211 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
1216 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1217 arp_key->arp_sip, is_mask);
1218 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1219 arp_key->arp_tip, is_mask);
1220 SW_FLOW_KEY_PUT(match, ip.proto,
1221 ntohs(arp_key->arp_op), is_mask);
1222 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1223 arp_key->arp_sha, ETH_ALEN, is_mask);
1224 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1225 arp_key->arp_tha, ETH_ALEN, is_mask);
1227 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1230 if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1231 const struct ovs_key_mpls *mpls_key;
1233 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1234 SW_FLOW_KEY_PUT(match, mpls.top_lse,
1235 mpls_key->mpls_lse, is_mask);
1237 attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1240 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1241 const struct ovs_key_tcp *tcp_key;
1243 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1244 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1245 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1246 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1249 if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1250 SW_FLOW_KEY_PUT(match, tp.flags,
1251 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1253 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1256 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1257 const struct ovs_key_udp *udp_key;
1259 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1260 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1261 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1262 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1265 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1266 const struct ovs_key_sctp *sctp_key;
1268 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1269 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1270 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1271 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1274 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1275 const struct ovs_key_icmp *icmp_key;
1277 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1278 SW_FLOW_KEY_PUT(match, tp.src,
1279 htons(icmp_key->icmp_type), is_mask);
1280 SW_FLOW_KEY_PUT(match, tp.dst,
1281 htons(icmp_key->icmp_code), is_mask);
1282 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1285 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1286 const struct ovs_key_icmpv6 *icmpv6_key;
1288 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1289 SW_FLOW_KEY_PUT(match, tp.src,
1290 htons(icmpv6_key->icmpv6_type), is_mask);
1291 SW_FLOW_KEY_PUT(match, tp.dst,
1292 htons(icmpv6_key->icmpv6_code), is_mask);
1293 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1296 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1297 const struct ovs_key_nd *nd_key;
1299 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1300 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1302 sizeof(match->key->ipv6.nd.target),
1304 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1305 nd_key->nd_sll, ETH_ALEN, is_mask);
1306 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1307 nd_key->nd_tll, ETH_ALEN, is_mask);
1308 attrs &= ~(1 << OVS_KEY_ATTR_ND);
1312 OVS_NLERR(log, "Unknown key attributes %llx",
1313 (unsigned long long)attrs);
1320 static void nlattr_set(struct nlattr *attr, u8 val,
1321 const struct ovs_len_tbl *tbl)
1326 /* The nlattr stream should already have been validated */
1327 nla_for_each_nested(nla, attr, rem) {
1328 if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED) {
1329 if (tbl[nla_type(nla)].next)
1330 tbl = tbl[nla_type(nla)].next;
1331 nlattr_set(nla, val, tbl);
1333 memset(nla_data(nla), val, nla_len(nla));
1336 if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1337 *(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1341 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1343 nlattr_set(attr, val, ovs_key_lens);
1347 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1348 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1349 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1350 * does not include any don't care bit.
1351 * @net: Used to determine per-namespace field support.
1352 * @match: receives the extracted flow match information.
1353 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1354 * sequence. The fields should of the packet that triggered the creation
1356 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1357 * attribute specifies the mask field of the wildcarded flow.
1358 * @log: Boolean to allow kernel error logging. Normally true, but when
1359 * probing for feature compatibility this should be passed in as false to
1360 * suppress unnecessary error logging.
1362 int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1363 const struct nlattr *nla_key,
1364 const struct nlattr *nla_mask,
1367 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1368 struct nlattr *newmask = NULL;
1373 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1377 err = parse_vlan_from_nlattrs(match, &key_attrs, a, false, log);
1381 err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1387 /* Create an exact match mask. We need to set to 0xff
1388 * all the 'match->mask' fields that have been touched
1389 * in 'match->key'. We cannot simply memset
1390 * 'match->mask', because padding bytes and fields not
1391 * specified in 'match->key' should be left to 0.
1392 * Instead, we use a stream of netlink attributes,
1393 * copied from 'key' and set to 0xff.
1394 * ovs_key_from_nlattrs() will take care of filling
1395 * 'match->mask' appropriately.
1397 newmask = kmemdup(nla_key,
1398 nla_total_size(nla_len(nla_key)),
1403 mask_set_nlattr(newmask, 0xff);
1405 /* The userspace does not send tunnel attributes that
1406 * are 0, but we should not wildcard them nonetheless.
1408 if (match->key->tun_proto)
1409 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1415 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1419 /* Always match on tci. */
1420 SW_FLOW_KEY_PUT(match, eth.vlan.tci, htons(0xffff), true);
1421 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, htons(0xffff), true);
1423 err = parse_vlan_from_nlattrs(match, &mask_attrs, a, true, log);
1427 err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1433 if (!match_validate(match, key_attrs, mask_attrs, log))
1441 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1448 len = nla_len(attr);
1449 if (len < 1 || len > MAX_UFID_LENGTH) {
1450 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1451 nla_len(attr), MAX_UFID_LENGTH);
1458 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1459 * or false otherwise.
1461 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1464 sfid->ufid_len = get_ufid_len(attr, log);
1466 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1468 return sfid->ufid_len;
1471 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1472 const struct sw_flow_key *key, bool log)
1474 struct sw_flow_key *new_key;
1476 if (ovs_nla_get_ufid(sfid, ufid, log))
1479 /* If UFID was not provided, use unmasked key. */
1480 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1483 memcpy(new_key, key, sizeof(*key));
1484 sfid->unmasked_key = new_key;
1489 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1491 return attr ? nla_get_u32(attr) : 0;
1495 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1496 * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1497 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1499 * @log: Boolean to allow kernel error logging. Normally true, but when
1500 * probing for feature compatibility this should be passed in as false to
1501 * suppress unnecessary error logging.
1503 * This parses a series of Netlink attributes that form a flow key, which must
1504 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1505 * get the metadata, that is, the parts of the flow key that cannot be
1506 * extracted from the packet itself.
1509 int ovs_nla_get_flow_metadata(struct net *net, const struct nlattr *attr,
1510 struct sw_flow_key *key,
1513 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1514 struct sw_flow_match match;
1518 err = parse_flow_nlattrs(attr, a, &attrs, log);
1522 memset(&match, 0, sizeof(match));
1525 memset(&key->ct, 0, sizeof(key->ct));
1526 key->phy.in_port = DP_MAX_PORTS;
1528 return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1531 static int ovs_nla_put_vlan(struct sk_buff *skb, const struct vlan_head *vh,
1534 __be16 eth_type = !is_mask ? vh->tpid : htons(0xffff);
1536 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1537 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, vh->tci))
1542 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1543 const struct sw_flow_key *output, bool is_mask,
1544 struct sk_buff *skb)
1546 struct ovs_key_ethernet *eth_key;
1548 struct nlattr *encap = NULL;
1549 struct nlattr *in_encap = NULL;
1551 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1552 goto nla_put_failure;
1554 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1555 goto nla_put_failure;
1557 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1558 goto nla_put_failure;
1560 if ((swkey->tun_proto || is_mask)) {
1561 const void *opts = NULL;
1563 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1564 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1566 if (ip_tun_to_nlattr(skb, &output->tun_key, opts,
1567 swkey->tun_opts_len, swkey->tun_proto))
1568 goto nla_put_failure;
1571 if (swkey->phy.in_port == DP_MAX_PORTS) {
1572 if (is_mask && (output->phy.in_port == 0xffff))
1573 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1574 goto nla_put_failure;
1577 upper_u16 = !is_mask ? 0 : 0xffff;
1579 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1580 (upper_u16 << 16) | output->phy.in_port))
1581 goto nla_put_failure;
1584 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1585 goto nla_put_failure;
1587 if (ovs_ct_put_key(output, skb))
1588 goto nla_put_failure;
1590 if (ovs_key_mac_proto(swkey) == MAC_PROTO_ETHERNET) {
1591 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1593 goto nla_put_failure;
1595 eth_key = nla_data(nla);
1596 ether_addr_copy(eth_key->eth_src, output->eth.src);
1597 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1599 if (swkey->eth.vlan.tci || eth_type_vlan(swkey->eth.type)) {
1600 if (ovs_nla_put_vlan(skb, &output->eth.vlan, is_mask))
1601 goto nla_put_failure;
1602 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1603 if (!swkey->eth.vlan.tci)
1606 if (swkey->eth.cvlan.tci || eth_type_vlan(swkey->eth.type)) {
1607 if (ovs_nla_put_vlan(skb, &output->eth.cvlan, is_mask))
1608 goto nla_put_failure;
1609 in_encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1610 if (!swkey->eth.cvlan.tci)
1615 if (swkey->eth.type == htons(ETH_P_802_2)) {
1617 * Ethertype 802.2 is represented in the netlink with omitted
1618 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1619 * 0xffff in the mask attribute. Ethertype can also
1622 if (is_mask && output->eth.type)
1623 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1625 goto nla_put_failure;
1630 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1631 goto nla_put_failure;
1633 if (eth_type_vlan(swkey->eth.type)) {
1634 /* There are 3 VLAN tags, we don't know anything about the rest
1635 * of the packet, so truncate here.
1637 WARN_ON_ONCE(!(encap && in_encap));
1641 if (swkey->eth.type == htons(ETH_P_IP)) {
1642 struct ovs_key_ipv4 *ipv4_key;
1644 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1646 goto nla_put_failure;
1647 ipv4_key = nla_data(nla);
1648 ipv4_key->ipv4_src = output->ipv4.addr.src;
1649 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1650 ipv4_key->ipv4_proto = output->ip.proto;
1651 ipv4_key->ipv4_tos = output->ip.tos;
1652 ipv4_key->ipv4_ttl = output->ip.ttl;
1653 ipv4_key->ipv4_frag = output->ip.frag;
1654 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1655 struct ovs_key_ipv6 *ipv6_key;
1657 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1659 goto nla_put_failure;
1660 ipv6_key = nla_data(nla);
1661 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1662 sizeof(ipv6_key->ipv6_src));
1663 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1664 sizeof(ipv6_key->ipv6_dst));
1665 ipv6_key->ipv6_label = output->ipv6.label;
1666 ipv6_key->ipv6_proto = output->ip.proto;
1667 ipv6_key->ipv6_tclass = output->ip.tos;
1668 ipv6_key->ipv6_hlimit = output->ip.ttl;
1669 ipv6_key->ipv6_frag = output->ip.frag;
1670 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1671 swkey->eth.type == htons(ETH_P_RARP)) {
1672 struct ovs_key_arp *arp_key;
1674 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1676 goto nla_put_failure;
1677 arp_key = nla_data(nla);
1678 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1679 arp_key->arp_sip = output->ipv4.addr.src;
1680 arp_key->arp_tip = output->ipv4.addr.dst;
1681 arp_key->arp_op = htons(output->ip.proto);
1682 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1683 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1684 } else if (eth_p_mpls(swkey->eth.type)) {
1685 struct ovs_key_mpls *mpls_key;
1687 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1689 goto nla_put_failure;
1690 mpls_key = nla_data(nla);
1691 mpls_key->mpls_lse = output->mpls.top_lse;
1694 if ((swkey->eth.type == htons(ETH_P_IP) ||
1695 swkey->eth.type == htons(ETH_P_IPV6)) &&
1696 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1698 if (swkey->ip.proto == IPPROTO_TCP) {
1699 struct ovs_key_tcp *tcp_key;
1701 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1703 goto nla_put_failure;
1704 tcp_key = nla_data(nla);
1705 tcp_key->tcp_src = output->tp.src;
1706 tcp_key->tcp_dst = output->tp.dst;
1707 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1709 goto nla_put_failure;
1710 } else if (swkey->ip.proto == IPPROTO_UDP) {
1711 struct ovs_key_udp *udp_key;
1713 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1715 goto nla_put_failure;
1716 udp_key = nla_data(nla);
1717 udp_key->udp_src = output->tp.src;
1718 udp_key->udp_dst = output->tp.dst;
1719 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1720 struct ovs_key_sctp *sctp_key;
1722 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1724 goto nla_put_failure;
1725 sctp_key = nla_data(nla);
1726 sctp_key->sctp_src = output->tp.src;
1727 sctp_key->sctp_dst = output->tp.dst;
1728 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1729 swkey->ip.proto == IPPROTO_ICMP) {
1730 struct ovs_key_icmp *icmp_key;
1732 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1734 goto nla_put_failure;
1735 icmp_key = nla_data(nla);
1736 icmp_key->icmp_type = ntohs(output->tp.src);
1737 icmp_key->icmp_code = ntohs(output->tp.dst);
1738 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1739 swkey->ip.proto == IPPROTO_ICMPV6) {
1740 struct ovs_key_icmpv6 *icmpv6_key;
1742 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1743 sizeof(*icmpv6_key));
1745 goto nla_put_failure;
1746 icmpv6_key = nla_data(nla);
1747 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1748 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1750 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1751 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1752 struct ovs_key_nd *nd_key;
1754 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1756 goto nla_put_failure;
1757 nd_key = nla_data(nla);
1758 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1759 sizeof(nd_key->nd_target));
1760 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1761 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1768 nla_nest_end(skb, in_encap);
1770 nla_nest_end(skb, encap);
1778 int ovs_nla_put_key(const struct sw_flow_key *swkey,
1779 const struct sw_flow_key *output, int attr, bool is_mask,
1780 struct sk_buff *skb)
1785 nla = nla_nest_start(skb, attr);
1788 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
1791 nla_nest_end(skb, nla);
1796 /* Called with ovs_mutex or RCU read lock. */
1797 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
1799 if (ovs_identifier_is_ufid(&flow->id))
1800 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
1803 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
1804 OVS_FLOW_ATTR_KEY, false, skb);
1807 /* Called with ovs_mutex or RCU read lock. */
1808 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
1810 return ovs_nla_put_key(&flow->key, &flow->key,
1811 OVS_FLOW_ATTR_KEY, false, skb);
1814 /* Called with ovs_mutex or RCU read lock. */
1815 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
1817 return ovs_nla_put_key(&flow->key, &flow->mask->key,
1818 OVS_FLOW_ATTR_MASK, true, skb);
1821 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1823 static struct sw_flow_actions *nla_alloc_flow_actions(int size, bool log)
1825 struct sw_flow_actions *sfa;
1827 if (size > MAX_ACTIONS_BUFSIZE) {
1828 OVS_NLERR(log, "Flow action size %u bytes exceeds max", size);
1829 return ERR_PTR(-EINVAL);
1832 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1834 return ERR_PTR(-ENOMEM);
1836 sfa->actions_len = 0;
1840 static void ovs_nla_free_set_action(const struct nlattr *a)
1842 const struct nlattr *ovs_key = nla_data(a);
1843 struct ovs_tunnel_info *ovs_tun;
1845 switch (nla_type(ovs_key)) {
1846 case OVS_KEY_ATTR_TUNNEL_INFO:
1847 ovs_tun = nla_data(ovs_key);
1848 dst_release((struct dst_entry *)ovs_tun->tun_dst);
1853 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1855 const struct nlattr *a;
1861 nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
1862 switch (nla_type(a)) {
1863 case OVS_ACTION_ATTR_SET:
1864 ovs_nla_free_set_action(a);
1866 case OVS_ACTION_ATTR_CT:
1867 ovs_ct_free_action(a);
1875 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
1877 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
1880 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1881 * The caller must hold rcu_read_lock for this to be sensible. */
1882 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
1884 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
1887 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1888 int attr_len, bool log)
1891 struct sw_flow_actions *acts;
1893 int req_size = NLA_ALIGN(attr_len);
1894 int next_offset = offsetof(struct sw_flow_actions, actions) +
1895 (*sfa)->actions_len;
1897 if (req_size <= (ksize(*sfa) - next_offset))
1900 new_acts_size = ksize(*sfa) * 2;
1902 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1903 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1904 return ERR_PTR(-EMSGSIZE);
1905 new_acts_size = MAX_ACTIONS_BUFSIZE;
1908 acts = nla_alloc_flow_actions(new_acts_size, log);
1910 return (void *)acts;
1912 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1913 acts->actions_len = (*sfa)->actions_len;
1914 acts->orig_len = (*sfa)->orig_len;
1919 (*sfa)->actions_len += req_size;
1920 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1923 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1924 int attrtype, void *data, int len, bool log)
1928 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
1932 a->nla_type = attrtype;
1933 a->nla_len = nla_attr_size(len);
1936 memcpy(nla_data(a), data, len);
1937 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1942 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
1947 a = __add_action(sfa, attrtype, data, len, log);
1949 return PTR_ERR_OR_ZERO(a);
1952 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1953 int attrtype, bool log)
1955 int used = (*sfa)->actions_len;
1958 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
1965 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1968 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1971 a->nla_len = sfa->actions_len - st_offset;
1974 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
1975 const struct sw_flow_key *key,
1976 int depth, struct sw_flow_actions **sfa,
1977 __be16 eth_type, __be16 vlan_tci, bool log);
1979 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
1980 const struct sw_flow_key *key, int depth,
1981 struct sw_flow_actions **sfa,
1982 __be16 eth_type, __be16 vlan_tci, bool log)
1984 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1985 const struct nlattr *probability, *actions;
1986 const struct nlattr *a;
1987 int rem, start, err, st_acts;
1989 memset(attrs, 0, sizeof(attrs));
1990 nla_for_each_nested(a, attr, rem) {
1991 int type = nla_type(a);
1992 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1999 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
2000 if (!probability || nla_len(probability) != sizeof(u32))
2003 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
2004 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
2007 /* validation done, copy sample action. */
2008 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
2011 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
2012 nla_data(probability), sizeof(u32), log);
2015 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS, log);
2019 err = __ovs_nla_copy_actions(net, actions, key, depth + 1, sfa,
2020 eth_type, vlan_tci, log);
2024 add_nested_action_end(*sfa, st_acts);
2025 add_nested_action_end(*sfa, start);
2030 void ovs_match_init(struct sw_flow_match *match,
2031 struct sw_flow_key *key,
2033 struct sw_flow_mask *mask)
2035 memset(match, 0, sizeof(*match));
2040 memset(key, 0, sizeof(*key));
2043 memset(&mask->key, 0, sizeof(mask->key));
2044 mask->range.start = mask->range.end = 0;
2048 static int validate_geneve_opts(struct sw_flow_key *key)
2050 struct geneve_opt *option;
2051 int opts_len = key->tun_opts_len;
2052 bool crit_opt = false;
2054 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2055 while (opts_len > 0) {
2058 if (opts_len < sizeof(*option))
2061 len = sizeof(*option) + option->length * 4;
2065 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2067 option = (struct geneve_opt *)((u8 *)option + len);
2071 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
2076 static int validate_and_copy_set_tun(const struct nlattr *attr,
2077 struct sw_flow_actions **sfa, bool log)
2079 struct sw_flow_match match;
2080 struct sw_flow_key key;
2081 struct metadata_dst *tun_dst;
2082 struct ip_tunnel_info *tun_info;
2083 struct ovs_tunnel_info *ovs_tun;
2085 int err = 0, start, opts_type;
2087 ovs_match_init(&match, &key, true, NULL);
2088 opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2092 if (key.tun_opts_len) {
2093 switch (opts_type) {
2094 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2095 err = validate_geneve_opts(&key);
2099 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
2104 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2108 tun_dst = metadata_dst_alloc(key.tun_opts_len, GFP_KERNEL);
2112 err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2114 dst_release((struct dst_entry *)tun_dst);
2118 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2119 sizeof(*ovs_tun), log);
2121 dst_release((struct dst_entry *)tun_dst);
2125 ovs_tun = nla_data(a);
2126 ovs_tun->tun_dst = tun_dst;
2128 tun_info = &tun_dst->u.tun_info;
2129 tun_info->mode = IP_TUNNEL_INFO_TX;
2130 if (key.tun_proto == AF_INET6)
2131 tun_info->mode |= IP_TUNNEL_INFO_IPV6;
2132 tun_info->key = key.tun_key;
2134 /* We need to store the options in the action itself since
2135 * everything else will go away after flow setup. We can append
2136 * it to tun_info and then point there.
2138 ip_tunnel_info_opts_set(tun_info,
2139 TUN_METADATA_OPTS(&key, key.tun_opts_len),
2141 add_nested_action_end(*sfa, start);
2146 /* Return false if there are any non-masked bits set.
2147 * Mask follows data immediately, before any netlink padding.
2149 static bool validate_masked(u8 *data, int len)
2151 u8 *mask = data + len;
2154 if (*data++ & ~*mask++)
2160 static int validate_set(const struct nlattr *a,
2161 const struct sw_flow_key *flow_key,
2162 struct sw_flow_actions **sfa, bool *skip_copy,
2163 u8 mac_proto, __be16 eth_type, bool masked, bool log)
2165 const struct nlattr *ovs_key = nla_data(a);
2166 int key_type = nla_type(ovs_key);
2169 /* There can be only one key in a action */
2170 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2173 key_len = nla_len(ovs_key);
2177 if (key_type > OVS_KEY_ATTR_MAX ||
2178 !check_attr_len(key_len, ovs_key_lens[key_type].len))
2181 if (masked && !validate_masked(nla_data(ovs_key), key_len))
2185 const struct ovs_key_ipv4 *ipv4_key;
2186 const struct ovs_key_ipv6 *ipv6_key;
2189 case OVS_KEY_ATTR_PRIORITY:
2190 case OVS_KEY_ATTR_SKB_MARK:
2191 case OVS_KEY_ATTR_CT_MARK:
2192 case OVS_KEY_ATTR_CT_LABELS:
2195 case OVS_KEY_ATTR_ETHERNET:
2196 if (mac_proto != MAC_PROTO_ETHERNET)
2200 case OVS_KEY_ATTR_TUNNEL:
2202 return -EINVAL; /* Masked tunnel set not supported. */
2205 err = validate_and_copy_set_tun(a, sfa, log);
2210 case OVS_KEY_ATTR_IPV4:
2211 if (eth_type != htons(ETH_P_IP))
2214 ipv4_key = nla_data(ovs_key);
2217 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2219 /* Non-writeable fields. */
2220 if (mask->ipv4_proto || mask->ipv4_frag)
2223 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2226 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2231 case OVS_KEY_ATTR_IPV6:
2232 if (eth_type != htons(ETH_P_IPV6))
2235 ipv6_key = nla_data(ovs_key);
2238 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2240 /* Non-writeable fields. */
2241 if (mask->ipv6_proto || mask->ipv6_frag)
2244 /* Invalid bits in the flow label mask? */
2245 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2248 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2251 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2254 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2259 case OVS_KEY_ATTR_TCP:
2260 if ((eth_type != htons(ETH_P_IP) &&
2261 eth_type != htons(ETH_P_IPV6)) ||
2262 flow_key->ip.proto != IPPROTO_TCP)
2267 case OVS_KEY_ATTR_UDP:
2268 if ((eth_type != htons(ETH_P_IP) &&
2269 eth_type != htons(ETH_P_IPV6)) ||
2270 flow_key->ip.proto != IPPROTO_UDP)
2275 case OVS_KEY_ATTR_MPLS:
2276 if (!eth_p_mpls(eth_type))
2280 case OVS_KEY_ATTR_SCTP:
2281 if ((eth_type != htons(ETH_P_IP) &&
2282 eth_type != htons(ETH_P_IPV6)) ||
2283 flow_key->ip.proto != IPPROTO_SCTP)
2292 /* Convert non-masked non-tunnel set actions to masked set actions. */
2293 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2294 int start, len = key_len * 2;
2299 start = add_nested_action_start(sfa,
2300 OVS_ACTION_ATTR_SET_TO_MASKED,
2305 at = __add_action(sfa, key_type, NULL, len, log);
2309 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2310 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
2311 /* Clear non-writeable bits from otherwise writeable fields. */
2312 if (key_type == OVS_KEY_ATTR_IPV6) {
2313 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2315 mask->ipv6_label &= htonl(0x000FFFFF);
2317 add_nested_action_end(*sfa, start);
2323 static int validate_userspace(const struct nlattr *attr)
2325 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2326 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2327 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2328 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2330 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2333 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
2334 attr, userspace_policy);
2338 if (!a[OVS_USERSPACE_ATTR_PID] ||
2339 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2345 static int copy_action(const struct nlattr *from,
2346 struct sw_flow_actions **sfa, bool log)
2348 int totlen = NLA_ALIGN(from->nla_len);
2351 to = reserve_sfa_size(sfa, from->nla_len, log);
2355 memcpy(to, from, totlen);
2359 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2360 const struct sw_flow_key *key,
2361 int depth, struct sw_flow_actions **sfa,
2362 __be16 eth_type, __be16 vlan_tci, bool log)
2364 u8 mac_proto = ovs_key_mac_proto(key);
2365 const struct nlattr *a;
2368 if (depth >= SAMPLE_ACTION_DEPTH)
2371 nla_for_each_nested(a, attr, rem) {
2372 /* Expected argument lengths, (u32)-1 for variable length. */
2373 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2374 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2375 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2376 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2377 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2378 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2379 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2380 [OVS_ACTION_ATTR_POP_VLAN] = 0,
2381 [OVS_ACTION_ATTR_SET] = (u32)-1,
2382 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2383 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2384 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2385 [OVS_ACTION_ATTR_CT] = (u32)-1,
2386 [OVS_ACTION_ATTR_TRUNC] = sizeof(struct ovs_action_trunc),
2387 [OVS_ACTION_ATTR_PUSH_ETH] = sizeof(struct ovs_action_push_eth),
2388 [OVS_ACTION_ATTR_POP_ETH] = 0,
2390 const struct ovs_action_push_vlan *vlan;
2391 int type = nla_type(a);
2394 if (type > OVS_ACTION_ATTR_MAX ||
2395 (action_lens[type] != nla_len(a) &&
2396 action_lens[type] != (u32)-1))
2401 case OVS_ACTION_ATTR_UNSPEC:
2404 case OVS_ACTION_ATTR_USERSPACE:
2405 err = validate_userspace(a);
2410 case OVS_ACTION_ATTR_OUTPUT:
2411 if (nla_get_u32(a) >= DP_MAX_PORTS)
2415 case OVS_ACTION_ATTR_TRUNC: {
2416 const struct ovs_action_trunc *trunc = nla_data(a);
2418 if (trunc->max_len < ETH_HLEN)
2423 case OVS_ACTION_ATTR_HASH: {
2424 const struct ovs_action_hash *act_hash = nla_data(a);
2426 switch (act_hash->hash_alg) {
2427 case OVS_HASH_ALG_L4:
2436 case OVS_ACTION_ATTR_POP_VLAN:
2437 if (mac_proto != MAC_PROTO_ETHERNET)
2439 vlan_tci = htons(0);
2442 case OVS_ACTION_ATTR_PUSH_VLAN:
2443 if (mac_proto != MAC_PROTO_ETHERNET)
2446 if (!eth_type_vlan(vlan->vlan_tpid))
2448 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2450 vlan_tci = vlan->vlan_tci;
2453 case OVS_ACTION_ATTR_RECIRC:
2456 case OVS_ACTION_ATTR_PUSH_MPLS: {
2457 const struct ovs_action_push_mpls *mpls = nla_data(a);
2459 if (!eth_p_mpls(mpls->mpls_ethertype))
2461 /* Prohibit push MPLS other than to a white list
2462 * for packets that have a known tag order.
2464 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2465 (eth_type != htons(ETH_P_IP) &&
2466 eth_type != htons(ETH_P_IPV6) &&
2467 eth_type != htons(ETH_P_ARP) &&
2468 eth_type != htons(ETH_P_RARP) &&
2469 !eth_p_mpls(eth_type)))
2471 eth_type = mpls->mpls_ethertype;
2475 case OVS_ACTION_ATTR_POP_MPLS:
2476 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2477 !eth_p_mpls(eth_type))
2480 /* Disallow subsequent L2.5+ set and mpls_pop actions
2481 * as there is no check here to ensure that the new
2482 * eth_type is valid and thus set actions could
2483 * write off the end of the packet or otherwise
2486 * Support for these actions is planned using packet
2489 eth_type = htons(0);
2492 case OVS_ACTION_ATTR_SET:
2493 err = validate_set(a, key, sfa,
2494 &skip_copy, mac_proto, eth_type,
2500 case OVS_ACTION_ATTR_SET_MASKED:
2501 err = validate_set(a, key, sfa,
2502 &skip_copy, mac_proto, eth_type,
2508 case OVS_ACTION_ATTR_SAMPLE:
2509 err = validate_and_copy_sample(net, a, key, depth, sfa,
2510 eth_type, vlan_tci, log);
2516 case OVS_ACTION_ATTR_CT:
2517 err = ovs_ct_copy_action(net, a, key, sfa, log);
2523 case OVS_ACTION_ATTR_PUSH_ETH:
2524 /* Disallow pushing an Ethernet header if one
2525 * is already present */
2526 if (mac_proto != MAC_PROTO_NONE)
2528 mac_proto = MAC_PROTO_NONE;
2531 case OVS_ACTION_ATTR_POP_ETH:
2532 if (mac_proto != MAC_PROTO_ETHERNET)
2534 if (vlan_tci & htons(VLAN_TAG_PRESENT))
2536 mac_proto = MAC_PROTO_ETHERNET;
2540 OVS_NLERR(log, "Unknown Action type %d", type);
2544 err = copy_action(a, sfa, log);
2556 /* 'key' must be the masked key. */
2557 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2558 const struct sw_flow_key *key,
2559 struct sw_flow_actions **sfa, bool log)
2563 *sfa = nla_alloc_flow_actions(nla_len(attr), log);
2565 return PTR_ERR(*sfa);
2567 (*sfa)->orig_len = nla_len(attr);
2568 err = __ovs_nla_copy_actions(net, attr, key, 0, sfa, key->eth.type,
2569 key->eth.vlan.tci, log);
2571 ovs_nla_free_flow_actions(*sfa);
2576 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
2578 const struct nlattr *a;
2579 struct nlattr *start;
2582 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
2586 nla_for_each_nested(a, attr, rem) {
2587 int type = nla_type(a);
2588 struct nlattr *st_sample;
2591 case OVS_SAMPLE_ATTR_PROBABILITY:
2592 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
2593 sizeof(u32), nla_data(a)))
2596 case OVS_SAMPLE_ATTR_ACTIONS:
2597 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
2600 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
2603 nla_nest_end(skb, st_sample);
2608 nla_nest_end(skb, start);
2612 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
2614 const struct nlattr *ovs_key = nla_data(a);
2615 int key_type = nla_type(ovs_key);
2616 struct nlattr *start;
2620 case OVS_KEY_ATTR_TUNNEL_INFO: {
2621 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
2622 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
2624 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2628 err = ip_tun_to_nlattr(skb, &tun_info->key,
2629 ip_tunnel_info_opts(tun_info),
2630 tun_info->options_len,
2631 ip_tunnel_info_af(tun_info));
2634 nla_nest_end(skb, start);
2638 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
2646 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
2647 struct sk_buff *skb)
2649 const struct nlattr *ovs_key = nla_data(a);
2651 size_t key_len = nla_len(ovs_key) / 2;
2653 /* Revert the conversion we did from a non-masked set action to
2654 * masked set action.
2656 nla = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2660 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
2663 nla_nest_end(skb, nla);
2667 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
2669 const struct nlattr *a;
2672 nla_for_each_attr(a, attr, len, rem) {
2673 int type = nla_type(a);
2676 case OVS_ACTION_ATTR_SET:
2677 err = set_action_to_attr(a, skb);
2682 case OVS_ACTION_ATTR_SET_TO_MASKED:
2683 err = masked_set_action_to_set_action_attr(a, skb);
2688 case OVS_ACTION_ATTR_SAMPLE:
2689 err = sample_action_to_attr(a, skb);
2694 case OVS_ACTION_ATTR_CT:
2695 err = ovs_ct_action_to_attr(nla_data(a), skb);
2701 if (nla_put(skb, type, nla_len(a), nla_data(a)))