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
61 static void update_range(struct sw_flow_match *match,
62 size_t offset, size_t size, bool is_mask)
64 struct sw_flow_key_range *range;
65 size_t start = rounddown(offset, sizeof(long));
66 size_t end = roundup(offset + size, sizeof(long));
69 range = &match->range;
71 range = &match->mask->range;
73 if (range->start == range->end) {
79 if (range->start > start)
86 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
88 update_range(match, offsetof(struct sw_flow_key, field), \
89 sizeof((match)->key->field), is_mask); \
91 (match)->mask->key.field = value; \
93 (match)->key->field = value; \
96 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
98 update_range(match, offset, len, is_mask); \
100 memcpy((u8 *)&(match)->mask->key + offset, value_p, \
103 memcpy((u8 *)(match)->key + offset, value_p, len); \
106 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
107 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
108 value_p, len, is_mask)
110 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
112 update_range(match, offsetof(struct sw_flow_key, field), \
113 sizeof((match)->key->field), is_mask); \
115 memset((u8 *)&(match)->mask->key.field, value, \
116 sizeof((match)->mask->key.field)); \
118 memset((u8 *)&(match)->key->field, value, \
119 sizeof((match)->key->field)); \
122 static bool match_validate(const struct sw_flow_match *match,
123 u64 key_attrs, u64 mask_attrs, bool log)
125 u64 key_expected = 1 << OVS_KEY_ATTR_ETHERNET;
126 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
128 /* The following mask attributes allowed only if they
129 * pass the validation tests. */
130 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
131 | (1 << OVS_KEY_ATTR_IPV6)
132 | (1 << OVS_KEY_ATTR_TCP)
133 | (1 << OVS_KEY_ATTR_TCP_FLAGS)
134 | (1 << OVS_KEY_ATTR_UDP)
135 | (1 << OVS_KEY_ATTR_SCTP)
136 | (1 << OVS_KEY_ATTR_ICMP)
137 | (1 << OVS_KEY_ATTR_ICMPV6)
138 | (1 << OVS_KEY_ATTR_ARP)
139 | (1 << OVS_KEY_ATTR_ND)
140 | (1 << OVS_KEY_ATTR_MPLS));
142 /* Always allowed mask fields. */
143 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
144 | (1 << OVS_KEY_ATTR_IN_PORT)
145 | (1 << OVS_KEY_ATTR_ETHERTYPE));
147 /* Check key attributes. */
148 if (match->key->eth.type == htons(ETH_P_ARP)
149 || match->key->eth.type == htons(ETH_P_RARP)) {
150 key_expected |= 1 << OVS_KEY_ATTR_ARP;
151 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
152 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
155 if (eth_p_mpls(match->key->eth.type)) {
156 key_expected |= 1 << OVS_KEY_ATTR_MPLS;
157 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
158 mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
161 if (match->key->eth.type == htons(ETH_P_IP)) {
162 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
163 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
164 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
166 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
167 if (match->key->ip.proto == IPPROTO_UDP) {
168 key_expected |= 1 << OVS_KEY_ATTR_UDP;
169 if (match->mask && (match->mask->key.ip.proto == 0xff))
170 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
173 if (match->key->ip.proto == IPPROTO_SCTP) {
174 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
175 if (match->mask && (match->mask->key.ip.proto == 0xff))
176 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
179 if (match->key->ip.proto == IPPROTO_TCP) {
180 key_expected |= 1 << OVS_KEY_ATTR_TCP;
181 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
182 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
183 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
184 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
188 if (match->key->ip.proto == IPPROTO_ICMP) {
189 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
190 if (match->mask && (match->mask->key.ip.proto == 0xff))
191 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
196 if (match->key->eth.type == htons(ETH_P_IPV6)) {
197 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
198 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
199 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
201 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
202 if (match->key->ip.proto == IPPROTO_UDP) {
203 key_expected |= 1 << OVS_KEY_ATTR_UDP;
204 if (match->mask && (match->mask->key.ip.proto == 0xff))
205 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
208 if (match->key->ip.proto == IPPROTO_SCTP) {
209 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
210 if (match->mask && (match->mask->key.ip.proto == 0xff))
211 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
214 if (match->key->ip.proto == IPPROTO_TCP) {
215 key_expected |= 1 << OVS_KEY_ATTR_TCP;
216 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
217 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
218 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
219 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
223 if (match->key->ip.proto == IPPROTO_ICMPV6) {
224 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
225 if (match->mask && (match->mask->key.ip.proto == 0xff))
226 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
228 if (match->key->tp.src ==
229 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
230 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
231 key_expected |= 1 << OVS_KEY_ATTR_ND;
232 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
233 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
239 if ((key_attrs & key_expected) != key_expected) {
240 /* Key attributes check failed. */
241 OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
242 (unsigned long long)key_attrs,
243 (unsigned long long)key_expected);
247 if ((mask_attrs & mask_allowed) != mask_attrs) {
248 /* Mask attributes check failed. */
249 OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
250 (unsigned long long)mask_attrs,
251 (unsigned long long)mask_allowed);
258 size_t ovs_tun_key_attr_size(void)
260 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
261 * updating this function.
263 return nla_total_size(8) /* OVS_TUNNEL_KEY_ATTR_ID */
264 + nla_total_size(4) /* OVS_TUNNEL_KEY_ATTR_IPV4_SRC */
265 + nla_total_size(4) /* OVS_TUNNEL_KEY_ATTR_IPV4_DST */
266 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
267 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
268 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
269 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
270 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
271 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
272 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS is mutually exclusive with
273 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
275 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
276 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
279 size_t ovs_key_attr_size(void)
281 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
282 * updating this function.
284 BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO != 26);
286 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
287 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
288 + ovs_tun_key_attr_size()
289 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
290 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
291 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
292 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
293 + nla_total_size(1) /* OVS_KEY_ATTR_CT_STATE */
294 + nla_total_size(2) /* OVS_KEY_ATTR_CT_ZONE */
295 + nla_total_size(4) /* OVS_KEY_ATTR_CT_MARK */
296 + nla_total_size(16) /* OVS_KEY_ATTR_CT_LABEL */
297 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
298 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
299 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
300 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
301 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
302 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
303 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
304 + nla_total_size(28); /* OVS_KEY_ATTR_ND */
307 static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
308 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
309 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
310 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
311 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
312 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
313 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
314 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
315 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
316 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
317 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
318 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_NESTED },
319 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED },
322 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
323 static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
324 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
325 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
326 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
327 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
328 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
329 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
330 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
331 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
332 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
333 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
334 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
335 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
336 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
337 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
338 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
339 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
340 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
341 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
342 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
343 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
344 .next = ovs_tunnel_key_lens, },
345 [OVS_KEY_ATTR_MPLS] = { .len = sizeof(struct ovs_key_mpls) },
346 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u8) },
347 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
348 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
349 [OVS_KEY_ATTR_CT_LABEL] = { .len = sizeof(struct ovs_key_ct_label) },
352 static bool is_all_zero(const u8 *fp, size_t size)
359 for (i = 0; i < size; i++)
366 static int __parse_flow_nlattrs(const struct nlattr *attr,
367 const struct nlattr *a[],
368 u64 *attrsp, bool log, bool nz)
370 const struct nlattr *nla;
375 nla_for_each_nested(nla, attr, rem) {
376 u16 type = nla_type(nla);
379 if (type > OVS_KEY_ATTR_MAX) {
380 OVS_NLERR(log, "Key type %d is out of range max %d",
381 type, OVS_KEY_ATTR_MAX);
385 if (attrs & (1 << type)) {
386 OVS_NLERR(log, "Duplicate key (type %d).", type);
390 expected_len = ovs_key_lens[type].len;
391 if (nla_len(nla) != expected_len && expected_len != OVS_ATTR_NESTED) {
392 OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
393 type, nla_len(nla), expected_len);
397 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
403 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
411 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
412 const struct nlattr *a[], u64 *attrsp,
415 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
418 static int parse_flow_nlattrs(const struct nlattr *attr,
419 const struct nlattr *a[], u64 *attrsp,
422 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
425 static int genev_tun_opt_from_nlattr(const struct nlattr *a,
426 struct sw_flow_match *match, bool is_mask,
429 unsigned long opt_key_offset;
431 if (nla_len(a) > sizeof(match->key->tun_opts)) {
432 OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
433 nla_len(a), sizeof(match->key->tun_opts));
437 if (nla_len(a) % 4 != 0) {
438 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
443 /* We need to record the length of the options passed
444 * down, otherwise packets with the same format but
445 * additional options will be silently matched.
448 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
451 /* This is somewhat unusual because it looks at
452 * both the key and mask while parsing the
453 * attributes (and by extension assumes the key
454 * is parsed first). Normally, we would verify
455 * that each is the correct length and that the
456 * attributes line up in the validate function.
457 * However, that is difficult because this is
458 * variable length and we won't have the
461 if (match->key->tun_opts_len != nla_len(a)) {
462 OVS_NLERR(log, "Geneve option len %d != mask len %d",
463 match->key->tun_opts_len, nla_len(a));
467 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
470 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
471 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
472 nla_len(a), is_mask);
476 static const struct nla_policy vxlan_opt_policy[OVS_VXLAN_EXT_MAX + 1] = {
477 [OVS_VXLAN_EXT_GBP] = { .type = NLA_U32 },
480 static int vxlan_tun_opt_from_nlattr(const struct nlattr *a,
481 struct sw_flow_match *match, bool is_mask,
484 struct nlattr *tb[OVS_VXLAN_EXT_MAX+1];
485 unsigned long opt_key_offset;
486 struct vxlan_metadata opts;
489 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
491 err = nla_parse_nested(tb, OVS_VXLAN_EXT_MAX, a, vxlan_opt_policy);
495 memset(&opts, 0, sizeof(opts));
497 if (tb[OVS_VXLAN_EXT_GBP])
498 opts.gbp = nla_get_u32(tb[OVS_VXLAN_EXT_GBP]);
501 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
503 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
505 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
506 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
511 static int ipv4_tun_from_nlattr(const struct nlattr *attr,
512 struct sw_flow_match *match, bool is_mask,
518 __be16 tun_flags = 0;
521 nla_for_each_nested(a, attr, rem) {
522 int type = nla_type(a);
525 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
526 OVS_NLERR(log, "Tunnel attr %d out of range max %d",
527 type, OVS_TUNNEL_KEY_ATTR_MAX);
531 if (ovs_tunnel_key_lens[type].len != nla_len(a) &&
532 ovs_tunnel_key_lens[type].len != OVS_ATTR_NESTED) {
533 OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
534 type, nla_len(a), ovs_tunnel_key_lens[type].len);
539 case OVS_TUNNEL_KEY_ATTR_ID:
540 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
541 nla_get_be64(a), is_mask);
542 tun_flags |= TUNNEL_KEY;
544 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
545 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
546 nla_get_in_addr(a), is_mask);
548 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
549 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
550 nla_get_in_addr(a), is_mask);
552 case OVS_TUNNEL_KEY_ATTR_TOS:
553 SW_FLOW_KEY_PUT(match, tun_key.tos,
554 nla_get_u8(a), is_mask);
556 case OVS_TUNNEL_KEY_ATTR_TTL:
557 SW_FLOW_KEY_PUT(match, tun_key.ttl,
558 nla_get_u8(a), is_mask);
561 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
562 tun_flags |= TUNNEL_DONT_FRAGMENT;
564 case OVS_TUNNEL_KEY_ATTR_CSUM:
565 tun_flags |= TUNNEL_CSUM;
567 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
568 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
569 nla_get_be16(a), is_mask);
571 case OVS_TUNNEL_KEY_ATTR_TP_DST:
572 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
573 nla_get_be16(a), is_mask);
575 case OVS_TUNNEL_KEY_ATTR_OAM:
576 tun_flags |= TUNNEL_OAM;
578 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
580 OVS_NLERR(log, "Multiple metadata blocks provided");
584 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
588 tun_flags |= TUNNEL_GENEVE_OPT;
591 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
593 OVS_NLERR(log, "Multiple metadata blocks provided");
597 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
601 tun_flags |= TUNNEL_VXLAN_OPT;
605 OVS_NLERR(log, "Unknown IPv4 tunnel attribute %d",
611 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
614 OVS_NLERR(log, "IPv4 tunnel attribute has %d unknown bytes.",
620 if (!match->key->tun_key.u.ipv4.dst) {
621 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
626 OVS_NLERR(log, "IPv4 tunnel TTL not specified.");
634 static int vxlan_opt_to_nlattr(struct sk_buff *skb,
635 const void *tun_opts, int swkey_tun_opts_len)
637 const struct vxlan_metadata *opts = tun_opts;
640 nla = nla_nest_start(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
644 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
647 nla_nest_end(skb, nla);
651 static int __ipv4_tun_to_nlattr(struct sk_buff *skb,
652 const struct ip_tunnel_key *output,
653 const void *tun_opts, int swkey_tun_opts_len)
655 if (output->tun_flags & TUNNEL_KEY &&
656 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
658 if (output->u.ipv4.src &&
659 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
662 if (output->u.ipv4.dst &&
663 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
667 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
669 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
671 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
672 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
674 if ((output->tun_flags & TUNNEL_CSUM) &&
675 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
677 if (output->tp_src &&
678 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
680 if (output->tp_dst &&
681 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
683 if ((output->tun_flags & TUNNEL_OAM) &&
684 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
687 if (output->tun_flags & TUNNEL_GENEVE_OPT &&
688 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
689 swkey_tun_opts_len, tun_opts))
691 else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
692 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
699 static int ipv4_tun_to_nlattr(struct sk_buff *skb,
700 const struct ip_tunnel_key *output,
701 const void *tun_opts, int swkey_tun_opts_len)
706 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
710 err = __ipv4_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len);
714 nla_nest_end(skb, nla);
718 int ovs_nla_put_egress_tunnel_key(struct sk_buff *skb,
719 const struct ip_tunnel_info *egress_tun_info)
721 return __ipv4_tun_to_nlattr(skb, &egress_tun_info->key,
722 egress_tun_info->options,
723 egress_tun_info->options_len);
726 static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
727 u64 *attrs, const struct nlattr **a,
728 bool is_mask, bool log)
730 if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
731 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
733 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
734 *attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
737 if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
738 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
740 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
741 *attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
744 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
745 SW_FLOW_KEY_PUT(match, phy.priority,
746 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
747 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
750 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
751 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
754 in_port = 0xffffffff; /* Always exact match in_port. */
755 } else if (in_port >= DP_MAX_PORTS) {
756 OVS_NLERR(log, "Port %d exceeds max allowable %d",
757 in_port, DP_MAX_PORTS);
761 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
762 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
763 } else if (!is_mask) {
764 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
767 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
768 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
770 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
771 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
773 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
774 if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
777 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
780 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
781 ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
782 u8 ct_state = nla_get_u8(a[OVS_KEY_ATTR_CT_STATE]);
784 SW_FLOW_KEY_PUT(match, ct.state, ct_state, is_mask);
785 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
787 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
788 ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
789 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
791 SW_FLOW_KEY_PUT(match, ct.zone, ct_zone, is_mask);
792 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
794 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
795 ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
796 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
798 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
799 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
801 if (*attrs & (1 << OVS_KEY_ATTR_CT_LABEL) &&
802 ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABEL)) {
803 const struct ovs_key_ct_label *cl;
805 cl = nla_data(a[OVS_KEY_ATTR_CT_LABEL]);
806 SW_FLOW_KEY_MEMCPY(match, ct.label, cl->ct_label,
807 sizeof(*cl), is_mask);
808 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABEL);
813 static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
814 u64 attrs, const struct nlattr **a,
815 bool is_mask, bool log)
819 err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
823 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
824 const struct ovs_key_ethernet *eth_key;
826 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
827 SW_FLOW_KEY_MEMCPY(match, eth.src,
828 eth_key->eth_src, ETH_ALEN, is_mask);
829 SW_FLOW_KEY_MEMCPY(match, eth.dst,
830 eth_key->eth_dst, ETH_ALEN, is_mask);
831 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
834 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
837 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
838 if (!(tci & htons(VLAN_TAG_PRESENT))) {
840 OVS_NLERR(log, "VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.");
842 OVS_NLERR(log, "VLAN TCI does not have VLAN_TAG_PRESENT bit set.");
847 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
848 attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
851 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
854 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
856 /* Always exact match EtherType. */
857 eth_type = htons(0xffff);
858 } else if (!eth_proto_is_802_3(eth_type)) {
859 OVS_NLERR(log, "EtherType %x is less than min %x",
860 ntohs(eth_type), ETH_P_802_3_MIN);
864 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
865 attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
866 } else if (!is_mask) {
867 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
870 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
871 const struct ovs_key_ipv4 *ipv4_key;
873 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
874 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
875 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
876 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
879 SW_FLOW_KEY_PUT(match, ip.proto,
880 ipv4_key->ipv4_proto, is_mask);
881 SW_FLOW_KEY_PUT(match, ip.tos,
882 ipv4_key->ipv4_tos, is_mask);
883 SW_FLOW_KEY_PUT(match, ip.ttl,
884 ipv4_key->ipv4_ttl, is_mask);
885 SW_FLOW_KEY_PUT(match, ip.frag,
886 ipv4_key->ipv4_frag, is_mask);
887 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
888 ipv4_key->ipv4_src, is_mask);
889 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
890 ipv4_key->ipv4_dst, is_mask);
891 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
894 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
895 const struct ovs_key_ipv6 *ipv6_key;
897 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
898 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
899 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
900 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
904 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
905 OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x).\n",
906 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
910 SW_FLOW_KEY_PUT(match, ipv6.label,
911 ipv6_key->ipv6_label, is_mask);
912 SW_FLOW_KEY_PUT(match, ip.proto,
913 ipv6_key->ipv6_proto, is_mask);
914 SW_FLOW_KEY_PUT(match, ip.tos,
915 ipv6_key->ipv6_tclass, is_mask);
916 SW_FLOW_KEY_PUT(match, ip.ttl,
917 ipv6_key->ipv6_hlimit, is_mask);
918 SW_FLOW_KEY_PUT(match, ip.frag,
919 ipv6_key->ipv6_frag, is_mask);
920 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
922 sizeof(match->key->ipv6.addr.src),
924 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
926 sizeof(match->key->ipv6.addr.dst),
929 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
932 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
933 const struct ovs_key_arp *arp_key;
935 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
936 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
937 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
942 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
943 arp_key->arp_sip, is_mask);
944 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
945 arp_key->arp_tip, is_mask);
946 SW_FLOW_KEY_PUT(match, ip.proto,
947 ntohs(arp_key->arp_op), is_mask);
948 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
949 arp_key->arp_sha, ETH_ALEN, is_mask);
950 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
951 arp_key->arp_tha, ETH_ALEN, is_mask);
953 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
956 if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
957 const struct ovs_key_mpls *mpls_key;
959 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
960 SW_FLOW_KEY_PUT(match, mpls.top_lse,
961 mpls_key->mpls_lse, is_mask);
963 attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
966 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
967 const struct ovs_key_tcp *tcp_key;
969 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
970 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
971 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
972 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
975 if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
976 SW_FLOW_KEY_PUT(match, tp.flags,
977 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
979 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
982 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
983 const struct ovs_key_udp *udp_key;
985 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
986 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
987 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
988 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
991 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
992 const struct ovs_key_sctp *sctp_key;
994 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
995 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
996 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
997 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1000 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1001 const struct ovs_key_icmp *icmp_key;
1003 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1004 SW_FLOW_KEY_PUT(match, tp.src,
1005 htons(icmp_key->icmp_type), is_mask);
1006 SW_FLOW_KEY_PUT(match, tp.dst,
1007 htons(icmp_key->icmp_code), is_mask);
1008 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1011 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1012 const struct ovs_key_icmpv6 *icmpv6_key;
1014 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1015 SW_FLOW_KEY_PUT(match, tp.src,
1016 htons(icmpv6_key->icmpv6_type), is_mask);
1017 SW_FLOW_KEY_PUT(match, tp.dst,
1018 htons(icmpv6_key->icmpv6_code), is_mask);
1019 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1022 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1023 const struct ovs_key_nd *nd_key;
1025 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1026 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1028 sizeof(match->key->ipv6.nd.target),
1030 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1031 nd_key->nd_sll, ETH_ALEN, is_mask);
1032 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1033 nd_key->nd_tll, ETH_ALEN, is_mask);
1034 attrs &= ~(1 << OVS_KEY_ATTR_ND);
1038 OVS_NLERR(log, "Unknown key attributes %llx",
1039 (unsigned long long)attrs);
1046 static void nlattr_set(struct nlattr *attr, u8 val,
1047 const struct ovs_len_tbl *tbl)
1052 /* The nlattr stream should already have been validated */
1053 nla_for_each_nested(nla, attr, rem) {
1054 if (tbl && tbl[nla_type(nla)].len == OVS_ATTR_NESTED)
1055 nlattr_set(nla, val, tbl[nla_type(nla)].next);
1057 memset(nla_data(nla), val, nla_len(nla));
1061 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1063 nlattr_set(attr, val, ovs_key_lens);
1067 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1068 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1069 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1070 * does not include any don't care bit.
1071 * @net: Used to determine per-namespace field support.
1072 * @match: receives the extracted flow match information.
1073 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1074 * sequence. The fields should of the packet that triggered the creation
1076 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1077 * attribute specifies the mask field of the wildcarded flow.
1078 * @log: Boolean to allow kernel error logging. Normally true, but when
1079 * probing for feature compatibility this should be passed in as false to
1080 * suppress unnecessary error logging.
1082 int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1083 const struct nlattr *nla_key,
1084 const struct nlattr *nla_mask,
1087 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1088 const struct nlattr *encap;
1089 struct nlattr *newmask = NULL;
1092 bool encap_valid = false;
1095 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1099 if ((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
1100 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
1101 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
1104 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
1105 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
1106 OVS_NLERR(log, "Invalid Vlan frame.");
1110 key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1111 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1112 encap = a[OVS_KEY_ATTR_ENCAP];
1113 key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1116 if (tci & htons(VLAN_TAG_PRESENT)) {
1117 err = parse_flow_nlattrs(encap, a, &key_attrs, log);
1121 /* Corner case for truncated 802.1Q header. */
1122 if (nla_len(encap)) {
1123 OVS_NLERR(log, "Truncated 802.1Q header has non-zero encap attribute.");
1127 OVS_NLERR(log, "Encap attr is set for non-VLAN frame");
1132 err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1138 /* Create an exact match mask. We need to set to 0xff
1139 * all the 'match->mask' fields that have been touched
1140 * in 'match->key'. We cannot simply memset
1141 * 'match->mask', because padding bytes and fields not
1142 * specified in 'match->key' should be left to 0.
1143 * Instead, we use a stream of netlink attributes,
1144 * copied from 'key' and set to 0xff.
1145 * ovs_key_from_nlattrs() will take care of filling
1146 * 'match->mask' appropriately.
1148 newmask = kmemdup(nla_key,
1149 nla_total_size(nla_len(nla_key)),
1154 mask_set_nlattr(newmask, 0xff);
1156 /* The userspace does not send tunnel attributes that
1157 * are 0, but we should not wildcard them nonetheless.
1159 if (match->key->tun_key.u.ipv4.dst)
1160 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1166 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1170 /* Always match on tci. */
1171 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
1173 if (mask_attrs & 1 << OVS_KEY_ATTR_ENCAP) {
1174 __be16 eth_type = 0;
1178 OVS_NLERR(log, "Encap mask attribute is set for non-VLAN frame.");
1183 mask_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1184 if (a[OVS_KEY_ATTR_ETHERTYPE])
1185 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1187 if (eth_type == htons(0xffff)) {
1188 mask_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1189 encap = a[OVS_KEY_ATTR_ENCAP];
1190 err = parse_flow_mask_nlattrs(encap, a,
1195 OVS_NLERR(log, "VLAN frames must have an exact match on the TPID (mask=%x).",
1201 if (a[OVS_KEY_ATTR_VLAN])
1202 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1204 if (!(tci & htons(VLAN_TAG_PRESENT))) {
1205 OVS_NLERR(log, "VLAN tag present bit must have an exact match (tci_mask=%x).",
1212 err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1218 if (!match_validate(match, key_attrs, mask_attrs, log))
1226 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1233 len = nla_len(attr);
1234 if (len < 1 || len > MAX_UFID_LENGTH) {
1235 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1236 nla_len(attr), MAX_UFID_LENGTH);
1243 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1244 * or false otherwise.
1246 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1249 sfid->ufid_len = get_ufid_len(attr, log);
1251 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1253 return sfid->ufid_len;
1256 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1257 const struct sw_flow_key *key, bool log)
1259 struct sw_flow_key *new_key;
1261 if (ovs_nla_get_ufid(sfid, ufid, log))
1264 /* If UFID was not provided, use unmasked key. */
1265 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1268 memcpy(new_key, key, sizeof(*key));
1269 sfid->unmasked_key = new_key;
1274 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1276 return attr ? nla_get_u32(attr) : 0;
1280 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1281 * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1282 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1284 * @log: Boolean to allow kernel error logging. Normally true, but when
1285 * probing for feature compatibility this should be passed in as false to
1286 * suppress unnecessary error logging.
1288 * This parses a series of Netlink attributes that form a flow key, which must
1289 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1290 * get the metadata, that is, the parts of the flow key that cannot be
1291 * extracted from the packet itself.
1294 int ovs_nla_get_flow_metadata(struct net *net, const struct nlattr *attr,
1295 struct sw_flow_key *key,
1298 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1299 struct sw_flow_match match;
1303 err = parse_flow_nlattrs(attr, a, &attrs, log);
1307 memset(&match, 0, sizeof(match));
1310 memset(&key->ct, 0, sizeof(key->ct));
1311 key->phy.in_port = DP_MAX_PORTS;
1313 return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1316 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1317 const struct sw_flow_key *output, bool is_mask,
1318 struct sk_buff *skb)
1320 struct ovs_key_ethernet *eth_key;
1321 struct nlattr *nla, *encap;
1323 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1324 goto nla_put_failure;
1326 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1327 goto nla_put_failure;
1329 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1330 goto nla_put_failure;
1332 if ((swkey->tun_key.u.ipv4.dst || is_mask)) {
1333 const void *opts = NULL;
1335 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1336 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1338 if (ipv4_tun_to_nlattr(skb, &output->tun_key, opts,
1339 swkey->tun_opts_len))
1340 goto nla_put_failure;
1343 if (swkey->phy.in_port == DP_MAX_PORTS) {
1344 if (is_mask && (output->phy.in_port == 0xffff))
1345 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1346 goto nla_put_failure;
1349 upper_u16 = !is_mask ? 0 : 0xffff;
1351 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1352 (upper_u16 << 16) | output->phy.in_port))
1353 goto nla_put_failure;
1356 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1357 goto nla_put_failure;
1359 if (ovs_ct_put_key(output, skb))
1360 goto nla_put_failure;
1362 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1364 goto nla_put_failure;
1366 eth_key = nla_data(nla);
1367 ether_addr_copy(eth_key->eth_src, output->eth.src);
1368 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1370 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1372 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
1373 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1374 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
1375 goto nla_put_failure;
1376 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1377 if (!swkey->eth.tci)
1382 if (swkey->eth.type == htons(ETH_P_802_2)) {
1384 * Ethertype 802.2 is represented in the netlink with omitted
1385 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1386 * 0xffff in the mask attribute. Ethertype can also
1389 if (is_mask && output->eth.type)
1390 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1392 goto nla_put_failure;
1396 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1397 goto nla_put_failure;
1399 if (swkey->eth.type == htons(ETH_P_IP)) {
1400 struct ovs_key_ipv4 *ipv4_key;
1402 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1404 goto nla_put_failure;
1405 ipv4_key = nla_data(nla);
1406 ipv4_key->ipv4_src = output->ipv4.addr.src;
1407 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1408 ipv4_key->ipv4_proto = output->ip.proto;
1409 ipv4_key->ipv4_tos = output->ip.tos;
1410 ipv4_key->ipv4_ttl = output->ip.ttl;
1411 ipv4_key->ipv4_frag = output->ip.frag;
1412 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1413 struct ovs_key_ipv6 *ipv6_key;
1415 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1417 goto nla_put_failure;
1418 ipv6_key = nla_data(nla);
1419 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1420 sizeof(ipv6_key->ipv6_src));
1421 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1422 sizeof(ipv6_key->ipv6_dst));
1423 ipv6_key->ipv6_label = output->ipv6.label;
1424 ipv6_key->ipv6_proto = output->ip.proto;
1425 ipv6_key->ipv6_tclass = output->ip.tos;
1426 ipv6_key->ipv6_hlimit = output->ip.ttl;
1427 ipv6_key->ipv6_frag = output->ip.frag;
1428 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1429 swkey->eth.type == htons(ETH_P_RARP)) {
1430 struct ovs_key_arp *arp_key;
1432 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1434 goto nla_put_failure;
1435 arp_key = nla_data(nla);
1436 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1437 arp_key->arp_sip = output->ipv4.addr.src;
1438 arp_key->arp_tip = output->ipv4.addr.dst;
1439 arp_key->arp_op = htons(output->ip.proto);
1440 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1441 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1442 } else if (eth_p_mpls(swkey->eth.type)) {
1443 struct ovs_key_mpls *mpls_key;
1445 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1447 goto nla_put_failure;
1448 mpls_key = nla_data(nla);
1449 mpls_key->mpls_lse = output->mpls.top_lse;
1452 if ((swkey->eth.type == htons(ETH_P_IP) ||
1453 swkey->eth.type == htons(ETH_P_IPV6)) &&
1454 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1456 if (swkey->ip.proto == IPPROTO_TCP) {
1457 struct ovs_key_tcp *tcp_key;
1459 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1461 goto nla_put_failure;
1462 tcp_key = nla_data(nla);
1463 tcp_key->tcp_src = output->tp.src;
1464 tcp_key->tcp_dst = output->tp.dst;
1465 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1467 goto nla_put_failure;
1468 } else if (swkey->ip.proto == IPPROTO_UDP) {
1469 struct ovs_key_udp *udp_key;
1471 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1473 goto nla_put_failure;
1474 udp_key = nla_data(nla);
1475 udp_key->udp_src = output->tp.src;
1476 udp_key->udp_dst = output->tp.dst;
1477 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1478 struct ovs_key_sctp *sctp_key;
1480 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1482 goto nla_put_failure;
1483 sctp_key = nla_data(nla);
1484 sctp_key->sctp_src = output->tp.src;
1485 sctp_key->sctp_dst = output->tp.dst;
1486 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1487 swkey->ip.proto == IPPROTO_ICMP) {
1488 struct ovs_key_icmp *icmp_key;
1490 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1492 goto nla_put_failure;
1493 icmp_key = nla_data(nla);
1494 icmp_key->icmp_type = ntohs(output->tp.src);
1495 icmp_key->icmp_code = ntohs(output->tp.dst);
1496 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1497 swkey->ip.proto == IPPROTO_ICMPV6) {
1498 struct ovs_key_icmpv6 *icmpv6_key;
1500 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1501 sizeof(*icmpv6_key));
1503 goto nla_put_failure;
1504 icmpv6_key = nla_data(nla);
1505 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1506 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1508 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1509 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1510 struct ovs_key_nd *nd_key;
1512 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1514 goto nla_put_failure;
1515 nd_key = nla_data(nla);
1516 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1517 sizeof(nd_key->nd_target));
1518 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1519 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1526 nla_nest_end(skb, encap);
1534 int ovs_nla_put_key(const struct sw_flow_key *swkey,
1535 const struct sw_flow_key *output, int attr, bool is_mask,
1536 struct sk_buff *skb)
1541 nla = nla_nest_start(skb, attr);
1544 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
1547 nla_nest_end(skb, nla);
1552 /* Called with ovs_mutex or RCU read lock. */
1553 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
1555 if (ovs_identifier_is_ufid(&flow->id))
1556 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
1559 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
1560 OVS_FLOW_ATTR_KEY, false, skb);
1563 /* Called with ovs_mutex or RCU read lock. */
1564 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
1566 return ovs_nla_put_key(&flow->key, &flow->key,
1567 OVS_FLOW_ATTR_KEY, false, skb);
1570 /* Called with ovs_mutex or RCU read lock. */
1571 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
1573 return ovs_nla_put_key(&flow->key, &flow->mask->key,
1574 OVS_FLOW_ATTR_MASK, true, skb);
1577 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1579 static struct sw_flow_actions *nla_alloc_flow_actions(int size, bool log)
1581 struct sw_flow_actions *sfa;
1583 if (size > MAX_ACTIONS_BUFSIZE) {
1584 OVS_NLERR(log, "Flow action size %u bytes exceeds max", size);
1585 return ERR_PTR(-EINVAL);
1588 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1590 return ERR_PTR(-ENOMEM);
1592 sfa->actions_len = 0;
1596 static void ovs_nla_free_set_action(const struct nlattr *a)
1598 const struct nlattr *ovs_key = nla_data(a);
1599 struct ovs_tunnel_info *ovs_tun;
1601 switch (nla_type(ovs_key)) {
1602 case OVS_KEY_ATTR_TUNNEL_INFO:
1603 ovs_tun = nla_data(ovs_key);
1604 dst_release((struct dst_entry *)ovs_tun->tun_dst);
1609 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1611 const struct nlattr *a;
1617 nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
1618 switch (nla_type(a)) {
1619 case OVS_ACTION_ATTR_SET:
1620 ovs_nla_free_set_action(a);
1622 case OVS_ACTION_ATTR_CT:
1623 ovs_ct_free_action(a);
1631 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
1633 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
1636 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1637 * The caller must hold rcu_read_lock for this to be sensible. */
1638 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
1640 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
1643 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1644 int attr_len, bool log)
1647 struct sw_flow_actions *acts;
1649 int req_size = NLA_ALIGN(attr_len);
1650 int next_offset = offsetof(struct sw_flow_actions, actions) +
1651 (*sfa)->actions_len;
1653 if (req_size <= (ksize(*sfa) - next_offset))
1656 new_acts_size = ksize(*sfa) * 2;
1658 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1659 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1660 return ERR_PTR(-EMSGSIZE);
1661 new_acts_size = MAX_ACTIONS_BUFSIZE;
1664 acts = nla_alloc_flow_actions(new_acts_size, log);
1666 return (void *)acts;
1668 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1669 acts->actions_len = (*sfa)->actions_len;
1670 acts->orig_len = (*sfa)->orig_len;
1675 (*sfa)->actions_len += req_size;
1676 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1679 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1680 int attrtype, void *data, int len, bool log)
1684 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
1688 a->nla_type = attrtype;
1689 a->nla_len = nla_attr_size(len);
1692 memcpy(nla_data(a), data, len);
1693 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1698 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
1703 a = __add_action(sfa, attrtype, data, len, log);
1705 return PTR_ERR_OR_ZERO(a);
1708 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1709 int attrtype, bool log)
1711 int used = (*sfa)->actions_len;
1714 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
1721 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1724 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1727 a->nla_len = sfa->actions_len - st_offset;
1730 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
1731 const struct sw_flow_key *key,
1732 int depth, struct sw_flow_actions **sfa,
1733 __be16 eth_type, __be16 vlan_tci, bool log);
1735 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
1736 const struct sw_flow_key *key, int depth,
1737 struct sw_flow_actions **sfa,
1738 __be16 eth_type, __be16 vlan_tci, bool log)
1740 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1741 const struct nlattr *probability, *actions;
1742 const struct nlattr *a;
1743 int rem, start, err, st_acts;
1745 memset(attrs, 0, sizeof(attrs));
1746 nla_for_each_nested(a, attr, rem) {
1747 int type = nla_type(a);
1748 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1755 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1756 if (!probability || nla_len(probability) != sizeof(u32))
1759 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1760 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1763 /* validation done, copy sample action. */
1764 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
1767 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1768 nla_data(probability), sizeof(u32), log);
1771 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS, log);
1775 err = __ovs_nla_copy_actions(net, actions, key, depth + 1, sfa,
1776 eth_type, vlan_tci, log);
1780 add_nested_action_end(*sfa, st_acts);
1781 add_nested_action_end(*sfa, start);
1786 void ovs_match_init(struct sw_flow_match *match,
1787 struct sw_flow_key *key,
1788 struct sw_flow_mask *mask)
1790 memset(match, 0, sizeof(*match));
1794 memset(key, 0, sizeof(*key));
1797 memset(&mask->key, 0, sizeof(mask->key));
1798 mask->range.start = mask->range.end = 0;
1802 static int validate_geneve_opts(struct sw_flow_key *key)
1804 struct geneve_opt *option;
1805 int opts_len = key->tun_opts_len;
1806 bool crit_opt = false;
1808 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
1809 while (opts_len > 0) {
1812 if (opts_len < sizeof(*option))
1815 len = sizeof(*option) + option->length * 4;
1819 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
1821 option = (struct geneve_opt *)((u8 *)option + len);
1825 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
1830 static int validate_and_copy_set_tun(const struct nlattr *attr,
1831 struct sw_flow_actions **sfa, bool log)
1833 struct sw_flow_match match;
1834 struct sw_flow_key key;
1835 struct metadata_dst *tun_dst;
1836 struct ip_tunnel_info *tun_info;
1837 struct ovs_tunnel_info *ovs_tun;
1839 int err = 0, start, opts_type;
1841 ovs_match_init(&match, &key, NULL);
1842 opts_type = ipv4_tun_from_nlattr(nla_data(attr), &match, false, log);
1846 if (key.tun_opts_len) {
1847 switch (opts_type) {
1848 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
1849 err = validate_geneve_opts(&key);
1853 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
1858 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
1862 tun_dst = metadata_dst_alloc(key.tun_opts_len, GFP_KERNEL);
1866 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
1867 sizeof(*ovs_tun), log);
1869 dst_release((struct dst_entry *)tun_dst);
1873 ovs_tun = nla_data(a);
1874 ovs_tun->tun_dst = tun_dst;
1876 tun_info = &tun_dst->u.tun_info;
1877 tun_info->mode = IP_TUNNEL_INFO_TX;
1878 tun_info->key = key.tun_key;
1879 tun_info->options_len = key.tun_opts_len;
1881 if (tun_info->options_len) {
1882 /* We need to store the options in the action itself since
1883 * everything else will go away after flow setup. We can append
1884 * it to tun_info and then point there.
1886 memcpy((tun_info + 1),
1887 TUN_METADATA_OPTS(&key, key.tun_opts_len), key.tun_opts_len);
1888 tun_info->options = (tun_info + 1);
1890 tun_info->options = NULL;
1893 add_nested_action_end(*sfa, start);
1898 /* Return false if there are any non-masked bits set.
1899 * Mask follows data immediately, before any netlink padding.
1901 static bool validate_masked(u8 *data, int len)
1903 u8 *mask = data + len;
1906 if (*data++ & ~*mask++)
1912 static int validate_set(const struct nlattr *a,
1913 const struct sw_flow_key *flow_key,
1914 struct sw_flow_actions **sfa,
1915 bool *skip_copy, __be16 eth_type, bool masked, bool log)
1917 const struct nlattr *ovs_key = nla_data(a);
1918 int key_type = nla_type(ovs_key);
1921 /* There can be only one key in a action */
1922 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1925 key_len = nla_len(ovs_key);
1929 if (key_type > OVS_KEY_ATTR_MAX ||
1930 (ovs_key_lens[key_type].len != key_len &&
1931 ovs_key_lens[key_type].len != OVS_ATTR_NESTED))
1934 if (masked && !validate_masked(nla_data(ovs_key), key_len))
1938 const struct ovs_key_ipv4 *ipv4_key;
1939 const struct ovs_key_ipv6 *ipv6_key;
1942 case OVS_KEY_ATTR_PRIORITY:
1943 case OVS_KEY_ATTR_SKB_MARK:
1944 case OVS_KEY_ATTR_CT_MARK:
1945 case OVS_KEY_ATTR_CT_LABEL:
1946 case OVS_KEY_ATTR_ETHERNET:
1949 case OVS_KEY_ATTR_TUNNEL:
1950 if (eth_p_mpls(eth_type))
1954 return -EINVAL; /* Masked tunnel set not supported. */
1957 err = validate_and_copy_set_tun(a, sfa, log);
1962 case OVS_KEY_ATTR_IPV4:
1963 if (eth_type != htons(ETH_P_IP))
1966 ipv4_key = nla_data(ovs_key);
1969 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
1971 /* Non-writeable fields. */
1972 if (mask->ipv4_proto || mask->ipv4_frag)
1975 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
1978 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
1983 case OVS_KEY_ATTR_IPV6:
1984 if (eth_type != htons(ETH_P_IPV6))
1987 ipv6_key = nla_data(ovs_key);
1990 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
1992 /* Non-writeable fields. */
1993 if (mask->ipv6_proto || mask->ipv6_frag)
1996 /* Invalid bits in the flow label mask? */
1997 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2000 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2003 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2006 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2011 case OVS_KEY_ATTR_TCP:
2012 if ((eth_type != htons(ETH_P_IP) &&
2013 eth_type != htons(ETH_P_IPV6)) ||
2014 flow_key->ip.proto != IPPROTO_TCP)
2019 case OVS_KEY_ATTR_UDP:
2020 if ((eth_type != htons(ETH_P_IP) &&
2021 eth_type != htons(ETH_P_IPV6)) ||
2022 flow_key->ip.proto != IPPROTO_UDP)
2027 case OVS_KEY_ATTR_MPLS:
2028 if (!eth_p_mpls(eth_type))
2032 case OVS_KEY_ATTR_SCTP:
2033 if ((eth_type != htons(ETH_P_IP) &&
2034 eth_type != htons(ETH_P_IPV6)) ||
2035 flow_key->ip.proto != IPPROTO_SCTP)
2044 /* Convert non-masked non-tunnel set actions to masked set actions. */
2045 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2046 int start, len = key_len * 2;
2051 start = add_nested_action_start(sfa,
2052 OVS_ACTION_ATTR_SET_TO_MASKED,
2057 at = __add_action(sfa, key_type, NULL, len, log);
2061 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2062 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
2063 /* Clear non-writeable bits from otherwise writeable fields. */
2064 if (key_type == OVS_KEY_ATTR_IPV6) {
2065 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2067 mask->ipv6_label &= htonl(0x000FFFFF);
2069 add_nested_action_end(*sfa, start);
2075 static int validate_userspace(const struct nlattr *attr)
2077 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2078 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2079 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2080 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2082 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2085 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
2086 attr, userspace_policy);
2090 if (!a[OVS_USERSPACE_ATTR_PID] ||
2091 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2097 static int copy_action(const struct nlattr *from,
2098 struct sw_flow_actions **sfa, bool log)
2100 int totlen = NLA_ALIGN(from->nla_len);
2103 to = reserve_sfa_size(sfa, from->nla_len, log);
2107 memcpy(to, from, totlen);
2111 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2112 const struct sw_flow_key *key,
2113 int depth, struct sw_flow_actions **sfa,
2114 __be16 eth_type, __be16 vlan_tci, bool log)
2116 const struct nlattr *a;
2119 if (depth >= SAMPLE_ACTION_DEPTH)
2122 nla_for_each_nested(a, attr, rem) {
2123 /* Expected argument lengths, (u32)-1 for variable length. */
2124 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2125 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2126 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2127 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2128 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2129 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2130 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2131 [OVS_ACTION_ATTR_POP_VLAN] = 0,
2132 [OVS_ACTION_ATTR_SET] = (u32)-1,
2133 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2134 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2135 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2136 [OVS_ACTION_ATTR_CT] = (u32)-1,
2138 const struct ovs_action_push_vlan *vlan;
2139 int type = nla_type(a);
2142 if (type > OVS_ACTION_ATTR_MAX ||
2143 (action_lens[type] != nla_len(a) &&
2144 action_lens[type] != (u32)-1))
2149 case OVS_ACTION_ATTR_UNSPEC:
2152 case OVS_ACTION_ATTR_USERSPACE:
2153 err = validate_userspace(a);
2158 case OVS_ACTION_ATTR_OUTPUT:
2159 if (nla_get_u32(a) >= DP_MAX_PORTS)
2163 case OVS_ACTION_ATTR_HASH: {
2164 const struct ovs_action_hash *act_hash = nla_data(a);
2166 switch (act_hash->hash_alg) {
2167 case OVS_HASH_ALG_L4:
2176 case OVS_ACTION_ATTR_POP_VLAN:
2177 vlan_tci = htons(0);
2180 case OVS_ACTION_ATTR_PUSH_VLAN:
2182 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
2184 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2186 vlan_tci = vlan->vlan_tci;
2189 case OVS_ACTION_ATTR_RECIRC:
2192 case OVS_ACTION_ATTR_PUSH_MPLS: {
2193 const struct ovs_action_push_mpls *mpls = nla_data(a);
2195 if (!eth_p_mpls(mpls->mpls_ethertype))
2197 /* Prohibit push MPLS other than to a white list
2198 * for packets that have a known tag order.
2200 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2201 (eth_type != htons(ETH_P_IP) &&
2202 eth_type != htons(ETH_P_IPV6) &&
2203 eth_type != htons(ETH_P_ARP) &&
2204 eth_type != htons(ETH_P_RARP) &&
2205 !eth_p_mpls(eth_type)))
2207 eth_type = mpls->mpls_ethertype;
2211 case OVS_ACTION_ATTR_POP_MPLS:
2212 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2213 !eth_p_mpls(eth_type))
2216 /* Disallow subsequent L2.5+ set and mpls_pop actions
2217 * as there is no check here to ensure that the new
2218 * eth_type is valid and thus set actions could
2219 * write off the end of the packet or otherwise
2222 * Support for these actions is planned using packet
2225 eth_type = htons(0);
2228 case OVS_ACTION_ATTR_SET:
2229 err = validate_set(a, key, sfa,
2230 &skip_copy, eth_type, false, log);
2235 case OVS_ACTION_ATTR_SET_MASKED:
2236 err = validate_set(a, key, sfa,
2237 &skip_copy, eth_type, true, log);
2242 case OVS_ACTION_ATTR_SAMPLE:
2243 err = validate_and_copy_sample(net, a, key, depth, sfa,
2244 eth_type, vlan_tci, log);
2250 case OVS_ACTION_ATTR_CT:
2251 err = ovs_ct_copy_action(net, a, key, sfa, log);
2258 OVS_NLERR(log, "Unknown Action type %d", type);
2262 err = copy_action(a, sfa, log);
2274 /* 'key' must be the masked key. */
2275 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2276 const struct sw_flow_key *key,
2277 struct sw_flow_actions **sfa, bool log)
2281 *sfa = nla_alloc_flow_actions(nla_len(attr), log);
2283 return PTR_ERR(*sfa);
2285 (*sfa)->orig_len = nla_len(attr);
2286 err = __ovs_nla_copy_actions(net, attr, key, 0, sfa, key->eth.type,
2289 ovs_nla_free_flow_actions(*sfa);
2294 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
2296 const struct nlattr *a;
2297 struct nlattr *start;
2300 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
2304 nla_for_each_nested(a, attr, rem) {
2305 int type = nla_type(a);
2306 struct nlattr *st_sample;
2309 case OVS_SAMPLE_ATTR_PROBABILITY:
2310 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
2311 sizeof(u32), nla_data(a)))
2314 case OVS_SAMPLE_ATTR_ACTIONS:
2315 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
2318 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
2321 nla_nest_end(skb, st_sample);
2326 nla_nest_end(skb, start);
2330 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
2332 const struct nlattr *ovs_key = nla_data(a);
2333 int key_type = nla_type(ovs_key);
2334 struct nlattr *start;
2338 case OVS_KEY_ATTR_TUNNEL_INFO: {
2339 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
2340 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
2342 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2346 err = ipv4_tun_to_nlattr(skb, &tun_info->key,
2347 tun_info->options_len ?
2348 tun_info->options : NULL,
2349 tun_info->options_len);
2352 nla_nest_end(skb, start);
2356 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
2364 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
2365 struct sk_buff *skb)
2367 const struct nlattr *ovs_key = nla_data(a);
2369 size_t key_len = nla_len(ovs_key) / 2;
2371 /* Revert the conversion we did from a non-masked set action to
2372 * masked set action.
2374 nla = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2378 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
2381 nla_nest_end(skb, nla);
2385 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
2387 const struct nlattr *a;
2390 nla_for_each_attr(a, attr, len, rem) {
2391 int type = nla_type(a);
2394 case OVS_ACTION_ATTR_SET:
2395 err = set_action_to_attr(a, skb);
2400 case OVS_ACTION_ATTR_SET_TO_MASKED:
2401 err = masked_set_action_to_set_action_attr(a, skb);
2406 case OVS_ACTION_ATTR_SAMPLE:
2407 err = sample_action_to_attr(a, skb);
2412 case OVS_ACTION_ATTR_CT:
2413 err = ovs_ct_action_to_attr(nla_data(a), skb);
2419 if (nla_put(skb, type, nla_len(a), nla_data(a)))