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 != 25);
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(12) /* OVS_KEY_ATTR_ETHERNET */
297 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
298 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
299 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
300 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
301 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
302 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
303 + nla_total_size(28); /* OVS_KEY_ATTR_ND */
306 static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
307 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
308 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
309 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
310 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
311 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
312 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
313 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
314 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
315 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
316 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
317 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_NESTED },
318 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED },
321 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
322 static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
323 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
324 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
325 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
326 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
327 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
328 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
329 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
330 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
331 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
332 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
333 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
334 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
335 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
336 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
337 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
338 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
339 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
340 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
341 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
342 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
343 .next = ovs_tunnel_key_lens, },
344 [OVS_KEY_ATTR_MPLS] = { .len = sizeof(struct ovs_key_mpls) },
345 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u8) },
346 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
347 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
350 static bool is_all_zero(const u8 *fp, size_t size)
357 for (i = 0; i < size; i++)
364 static int __parse_flow_nlattrs(const struct nlattr *attr,
365 const struct nlattr *a[],
366 u64 *attrsp, bool log, bool nz)
368 const struct nlattr *nla;
373 nla_for_each_nested(nla, attr, rem) {
374 u16 type = nla_type(nla);
377 if (type > OVS_KEY_ATTR_MAX) {
378 OVS_NLERR(log, "Key type %d is out of range max %d",
379 type, OVS_KEY_ATTR_MAX);
383 if (attrs & (1 << type)) {
384 OVS_NLERR(log, "Duplicate key (type %d).", type);
388 expected_len = ovs_key_lens[type].len;
389 if (nla_len(nla) != expected_len && expected_len != OVS_ATTR_NESTED) {
390 OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
391 type, nla_len(nla), expected_len);
395 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
401 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
409 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
410 const struct nlattr *a[], u64 *attrsp,
413 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
416 static int parse_flow_nlattrs(const struct nlattr *attr,
417 const struct nlattr *a[], u64 *attrsp,
420 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
423 static int genev_tun_opt_from_nlattr(const struct nlattr *a,
424 struct sw_flow_match *match, bool is_mask,
427 unsigned long opt_key_offset;
429 if (nla_len(a) > sizeof(match->key->tun_opts)) {
430 OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
431 nla_len(a), sizeof(match->key->tun_opts));
435 if (nla_len(a) % 4 != 0) {
436 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
441 /* We need to record the length of the options passed
442 * down, otherwise packets with the same format but
443 * additional options will be silently matched.
446 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
449 /* This is somewhat unusual because it looks at
450 * both the key and mask while parsing the
451 * attributes (and by extension assumes the key
452 * is parsed first). Normally, we would verify
453 * that each is the correct length and that the
454 * attributes line up in the validate function.
455 * However, that is difficult because this is
456 * variable length and we won't have the
459 if (match->key->tun_opts_len != nla_len(a)) {
460 OVS_NLERR(log, "Geneve option len %d != mask len %d",
461 match->key->tun_opts_len, nla_len(a));
465 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
468 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
469 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
470 nla_len(a), is_mask);
474 static const struct nla_policy vxlan_opt_policy[OVS_VXLAN_EXT_MAX + 1] = {
475 [OVS_VXLAN_EXT_GBP] = { .type = NLA_U32 },
478 static int vxlan_tun_opt_from_nlattr(const struct nlattr *a,
479 struct sw_flow_match *match, bool is_mask,
482 struct nlattr *tb[OVS_VXLAN_EXT_MAX+1];
483 unsigned long opt_key_offset;
484 struct vxlan_metadata opts;
487 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
489 err = nla_parse_nested(tb, OVS_VXLAN_EXT_MAX, a, vxlan_opt_policy);
493 memset(&opts, 0, sizeof(opts));
495 if (tb[OVS_VXLAN_EXT_GBP])
496 opts.gbp = nla_get_u32(tb[OVS_VXLAN_EXT_GBP]);
499 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
501 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
503 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
504 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
509 static int ipv4_tun_from_nlattr(const struct nlattr *attr,
510 struct sw_flow_match *match, bool is_mask,
516 __be16 tun_flags = 0;
519 nla_for_each_nested(a, attr, rem) {
520 int type = nla_type(a);
523 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
524 OVS_NLERR(log, "Tunnel attr %d out of range max %d",
525 type, OVS_TUNNEL_KEY_ATTR_MAX);
529 if (ovs_tunnel_key_lens[type].len != nla_len(a) &&
530 ovs_tunnel_key_lens[type].len != OVS_ATTR_NESTED) {
531 OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
532 type, nla_len(a), ovs_tunnel_key_lens[type].len);
537 case OVS_TUNNEL_KEY_ATTR_ID:
538 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
539 nla_get_be64(a), is_mask);
540 tun_flags |= TUNNEL_KEY;
542 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
543 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
544 nla_get_in_addr(a), is_mask);
546 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
547 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
548 nla_get_in_addr(a), is_mask);
550 case OVS_TUNNEL_KEY_ATTR_TOS:
551 SW_FLOW_KEY_PUT(match, tun_key.tos,
552 nla_get_u8(a), is_mask);
554 case OVS_TUNNEL_KEY_ATTR_TTL:
555 SW_FLOW_KEY_PUT(match, tun_key.ttl,
556 nla_get_u8(a), is_mask);
559 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
560 tun_flags |= TUNNEL_DONT_FRAGMENT;
562 case OVS_TUNNEL_KEY_ATTR_CSUM:
563 tun_flags |= TUNNEL_CSUM;
565 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
566 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
567 nla_get_be16(a), is_mask);
569 case OVS_TUNNEL_KEY_ATTR_TP_DST:
570 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
571 nla_get_be16(a), is_mask);
573 case OVS_TUNNEL_KEY_ATTR_OAM:
574 tun_flags |= TUNNEL_OAM;
576 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
578 OVS_NLERR(log, "Multiple metadata blocks provided");
582 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
586 tun_flags |= TUNNEL_GENEVE_OPT;
589 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
591 OVS_NLERR(log, "Multiple metadata blocks provided");
595 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
599 tun_flags |= TUNNEL_VXLAN_OPT;
603 OVS_NLERR(log, "Unknown IPv4 tunnel attribute %d",
609 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
612 OVS_NLERR(log, "IPv4 tunnel attribute has %d unknown bytes.",
618 if (!match->key->tun_key.u.ipv4.dst) {
619 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
624 OVS_NLERR(log, "IPv4 tunnel TTL not specified.");
632 static int vxlan_opt_to_nlattr(struct sk_buff *skb,
633 const void *tun_opts, int swkey_tun_opts_len)
635 const struct vxlan_metadata *opts = tun_opts;
638 nla = nla_nest_start(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
642 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
645 nla_nest_end(skb, nla);
649 static int __ipv4_tun_to_nlattr(struct sk_buff *skb,
650 const struct ip_tunnel_key *output,
651 const void *tun_opts, int swkey_tun_opts_len)
653 if (output->tun_flags & TUNNEL_KEY &&
654 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
656 if (output->u.ipv4.src &&
657 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
660 if (output->u.ipv4.dst &&
661 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
665 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
667 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
669 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
670 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
672 if ((output->tun_flags & TUNNEL_CSUM) &&
673 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
675 if (output->tp_src &&
676 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
678 if (output->tp_dst &&
679 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
681 if ((output->tun_flags & TUNNEL_OAM) &&
682 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
685 if (output->tun_flags & TUNNEL_GENEVE_OPT &&
686 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
687 swkey_tun_opts_len, tun_opts))
689 else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
690 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
697 static int ipv4_tun_to_nlattr(struct sk_buff *skb,
698 const struct ip_tunnel_key *output,
699 const void *tun_opts, int swkey_tun_opts_len)
704 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
708 err = __ipv4_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len);
712 nla_nest_end(skb, nla);
716 int ovs_nla_put_egress_tunnel_key(struct sk_buff *skb,
717 const struct ip_tunnel_info *egress_tun_info)
719 return __ipv4_tun_to_nlattr(skb, &egress_tun_info->key,
720 egress_tun_info->options,
721 egress_tun_info->options_len);
724 static int metadata_from_nlattrs(struct sw_flow_match *match, u64 *attrs,
725 const struct nlattr **a, bool is_mask,
728 if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
729 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
731 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
732 *attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
735 if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
736 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
738 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
739 *attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
742 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
743 SW_FLOW_KEY_PUT(match, phy.priority,
744 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
745 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
748 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
749 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
752 in_port = 0xffffffff; /* Always exact match in_port. */
753 } else if (in_port >= DP_MAX_PORTS) {
754 OVS_NLERR(log, "Port %d exceeds max allowable %d",
755 in_port, DP_MAX_PORTS);
759 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
760 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
761 } else if (!is_mask) {
762 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
765 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
766 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
768 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
769 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
771 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
772 if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
775 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
778 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
779 ovs_ct_verify(OVS_KEY_ATTR_CT_STATE)) {
780 u8 ct_state = nla_get_u8(a[OVS_KEY_ATTR_CT_STATE]);
782 SW_FLOW_KEY_PUT(match, ct.state, ct_state, is_mask);
783 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
785 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
786 ovs_ct_verify(OVS_KEY_ATTR_CT_ZONE)) {
787 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
789 SW_FLOW_KEY_PUT(match, ct.zone, ct_zone, is_mask);
790 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
792 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
793 ovs_ct_verify(OVS_KEY_ATTR_CT_MARK)) {
794 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
796 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
797 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
802 static int ovs_key_from_nlattrs(struct sw_flow_match *match, u64 attrs,
803 const struct nlattr **a, bool is_mask,
808 err = metadata_from_nlattrs(match, &attrs, a, is_mask, log);
812 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
813 const struct ovs_key_ethernet *eth_key;
815 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
816 SW_FLOW_KEY_MEMCPY(match, eth.src,
817 eth_key->eth_src, ETH_ALEN, is_mask);
818 SW_FLOW_KEY_MEMCPY(match, eth.dst,
819 eth_key->eth_dst, ETH_ALEN, is_mask);
820 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
823 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
826 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
827 if (!(tci & htons(VLAN_TAG_PRESENT))) {
829 OVS_NLERR(log, "VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.");
831 OVS_NLERR(log, "VLAN TCI does not have VLAN_TAG_PRESENT bit set.");
836 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
837 attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
840 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
843 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
845 /* Always exact match EtherType. */
846 eth_type = htons(0xffff);
847 } else if (!eth_proto_is_802_3(eth_type)) {
848 OVS_NLERR(log, "EtherType %x is less than min %x",
849 ntohs(eth_type), ETH_P_802_3_MIN);
853 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
854 attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
855 } else if (!is_mask) {
856 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
859 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
860 const struct ovs_key_ipv4 *ipv4_key;
862 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
863 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
864 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
865 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
868 SW_FLOW_KEY_PUT(match, ip.proto,
869 ipv4_key->ipv4_proto, is_mask);
870 SW_FLOW_KEY_PUT(match, ip.tos,
871 ipv4_key->ipv4_tos, is_mask);
872 SW_FLOW_KEY_PUT(match, ip.ttl,
873 ipv4_key->ipv4_ttl, is_mask);
874 SW_FLOW_KEY_PUT(match, ip.frag,
875 ipv4_key->ipv4_frag, is_mask);
876 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
877 ipv4_key->ipv4_src, is_mask);
878 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
879 ipv4_key->ipv4_dst, is_mask);
880 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
883 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
884 const struct ovs_key_ipv6 *ipv6_key;
886 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
887 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
888 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
889 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
893 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
894 OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x).\n",
895 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
899 SW_FLOW_KEY_PUT(match, ipv6.label,
900 ipv6_key->ipv6_label, is_mask);
901 SW_FLOW_KEY_PUT(match, ip.proto,
902 ipv6_key->ipv6_proto, is_mask);
903 SW_FLOW_KEY_PUT(match, ip.tos,
904 ipv6_key->ipv6_tclass, is_mask);
905 SW_FLOW_KEY_PUT(match, ip.ttl,
906 ipv6_key->ipv6_hlimit, is_mask);
907 SW_FLOW_KEY_PUT(match, ip.frag,
908 ipv6_key->ipv6_frag, is_mask);
909 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
911 sizeof(match->key->ipv6.addr.src),
913 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
915 sizeof(match->key->ipv6.addr.dst),
918 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
921 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
922 const struct ovs_key_arp *arp_key;
924 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
925 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
926 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
931 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
932 arp_key->arp_sip, is_mask);
933 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
934 arp_key->arp_tip, is_mask);
935 SW_FLOW_KEY_PUT(match, ip.proto,
936 ntohs(arp_key->arp_op), is_mask);
937 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
938 arp_key->arp_sha, ETH_ALEN, is_mask);
939 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
940 arp_key->arp_tha, ETH_ALEN, is_mask);
942 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
945 if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
946 const struct ovs_key_mpls *mpls_key;
948 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
949 SW_FLOW_KEY_PUT(match, mpls.top_lse,
950 mpls_key->mpls_lse, is_mask);
952 attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
955 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
956 const struct ovs_key_tcp *tcp_key;
958 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
959 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
960 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
961 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
964 if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
965 SW_FLOW_KEY_PUT(match, tp.flags,
966 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
968 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
971 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
972 const struct ovs_key_udp *udp_key;
974 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
975 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
976 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
977 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
980 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
981 const struct ovs_key_sctp *sctp_key;
983 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
984 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
985 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
986 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
989 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
990 const struct ovs_key_icmp *icmp_key;
992 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
993 SW_FLOW_KEY_PUT(match, tp.src,
994 htons(icmp_key->icmp_type), is_mask);
995 SW_FLOW_KEY_PUT(match, tp.dst,
996 htons(icmp_key->icmp_code), is_mask);
997 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1000 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1001 const struct ovs_key_icmpv6 *icmpv6_key;
1003 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1004 SW_FLOW_KEY_PUT(match, tp.src,
1005 htons(icmpv6_key->icmpv6_type), is_mask);
1006 SW_FLOW_KEY_PUT(match, tp.dst,
1007 htons(icmpv6_key->icmpv6_code), is_mask);
1008 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1011 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1012 const struct ovs_key_nd *nd_key;
1014 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1015 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1017 sizeof(match->key->ipv6.nd.target),
1019 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1020 nd_key->nd_sll, ETH_ALEN, is_mask);
1021 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1022 nd_key->nd_tll, ETH_ALEN, is_mask);
1023 attrs &= ~(1 << OVS_KEY_ATTR_ND);
1027 OVS_NLERR(log, "Unknown key attributes %llx",
1028 (unsigned long long)attrs);
1035 static void nlattr_set(struct nlattr *attr, u8 val,
1036 const struct ovs_len_tbl *tbl)
1041 /* The nlattr stream should already have been validated */
1042 nla_for_each_nested(nla, attr, rem) {
1043 if (tbl && tbl[nla_type(nla)].len == OVS_ATTR_NESTED)
1044 nlattr_set(nla, val, tbl[nla_type(nla)].next);
1046 memset(nla_data(nla), val, nla_len(nla));
1050 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1052 nlattr_set(attr, val, ovs_key_lens);
1056 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1057 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1058 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1059 * does not include any don't care bit.
1060 * @match: receives the extracted flow match information.
1061 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1062 * sequence. The fields should of the packet that triggered the creation
1064 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1065 * attribute specifies the mask field of the wildcarded flow.
1066 * @log: Boolean to allow kernel error logging. Normally true, but when
1067 * probing for feature compatibility this should be passed in as false to
1068 * suppress unnecessary error logging.
1070 int ovs_nla_get_match(struct sw_flow_match *match,
1071 const struct nlattr *nla_key,
1072 const struct nlattr *nla_mask,
1075 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1076 const struct nlattr *encap;
1077 struct nlattr *newmask = NULL;
1080 bool encap_valid = false;
1083 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1087 if ((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
1088 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
1089 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
1092 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
1093 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
1094 OVS_NLERR(log, "Invalid Vlan frame.");
1098 key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1099 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1100 encap = a[OVS_KEY_ATTR_ENCAP];
1101 key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1104 if (tci & htons(VLAN_TAG_PRESENT)) {
1105 err = parse_flow_nlattrs(encap, a, &key_attrs, log);
1109 /* Corner case for truncated 802.1Q header. */
1110 if (nla_len(encap)) {
1111 OVS_NLERR(log, "Truncated 802.1Q header has non-zero encap attribute.");
1115 OVS_NLERR(log, "Encap attr is set for non-VLAN frame");
1120 err = ovs_key_from_nlattrs(match, key_attrs, a, false, log);
1126 /* Create an exact match mask. We need to set to 0xff
1127 * all the 'match->mask' fields that have been touched
1128 * in 'match->key'. We cannot simply memset
1129 * 'match->mask', because padding bytes and fields not
1130 * specified in 'match->key' should be left to 0.
1131 * Instead, we use a stream of netlink attributes,
1132 * copied from 'key' and set to 0xff.
1133 * ovs_key_from_nlattrs() will take care of filling
1134 * 'match->mask' appropriately.
1136 newmask = kmemdup(nla_key,
1137 nla_total_size(nla_len(nla_key)),
1142 mask_set_nlattr(newmask, 0xff);
1144 /* The userspace does not send tunnel attributes that
1145 * are 0, but we should not wildcard them nonetheless.
1147 if (match->key->tun_key.u.ipv4.dst)
1148 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1154 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1158 /* Always match on tci. */
1159 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
1161 if (mask_attrs & 1 << OVS_KEY_ATTR_ENCAP) {
1162 __be16 eth_type = 0;
1166 OVS_NLERR(log, "Encap mask attribute is set for non-VLAN frame.");
1171 mask_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1172 if (a[OVS_KEY_ATTR_ETHERTYPE])
1173 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1175 if (eth_type == htons(0xffff)) {
1176 mask_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1177 encap = a[OVS_KEY_ATTR_ENCAP];
1178 err = parse_flow_mask_nlattrs(encap, a,
1183 OVS_NLERR(log, "VLAN frames must have an exact match on the TPID (mask=%x).",
1189 if (a[OVS_KEY_ATTR_VLAN])
1190 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1192 if (!(tci & htons(VLAN_TAG_PRESENT))) {
1193 OVS_NLERR(log, "VLAN tag present bit must have an exact match (tci_mask=%x).",
1200 err = ovs_key_from_nlattrs(match, mask_attrs, a, true, log);
1205 if (!match_validate(match, key_attrs, mask_attrs, log))
1213 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1220 len = nla_len(attr);
1221 if (len < 1 || len > MAX_UFID_LENGTH) {
1222 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1223 nla_len(attr), MAX_UFID_LENGTH);
1230 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1231 * or false otherwise.
1233 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1236 sfid->ufid_len = get_ufid_len(attr, log);
1238 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1240 return sfid->ufid_len;
1243 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1244 const struct sw_flow_key *key, bool log)
1246 struct sw_flow_key *new_key;
1248 if (ovs_nla_get_ufid(sfid, ufid, log))
1251 /* If UFID was not provided, use unmasked key. */
1252 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1255 memcpy(new_key, key, sizeof(*key));
1256 sfid->unmasked_key = new_key;
1261 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1263 return attr ? nla_get_u32(attr) : 0;
1267 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1268 * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1269 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1271 * @log: Boolean to allow kernel error logging. Normally true, but when
1272 * probing for feature compatibility this should be passed in as false to
1273 * suppress unnecessary error logging.
1275 * This parses a series of Netlink attributes that form a flow key, which must
1276 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1277 * get the metadata, that is, the parts of the flow key that cannot be
1278 * extracted from the packet itself.
1281 int ovs_nla_get_flow_metadata(const struct nlattr *attr,
1282 struct sw_flow_key *key,
1285 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1286 struct sw_flow_match match;
1290 err = parse_flow_nlattrs(attr, a, &attrs, log);
1294 memset(&match, 0, sizeof(match));
1297 memset(&key->ct, 0, sizeof(key->ct));
1298 key->phy.in_port = DP_MAX_PORTS;
1300 return metadata_from_nlattrs(&match, &attrs, a, false, log);
1303 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1304 const struct sw_flow_key *output, bool is_mask,
1305 struct sk_buff *skb)
1307 struct ovs_key_ethernet *eth_key;
1308 struct nlattr *nla, *encap;
1310 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1311 goto nla_put_failure;
1313 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1314 goto nla_put_failure;
1316 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1317 goto nla_put_failure;
1319 if ((swkey->tun_key.u.ipv4.dst || is_mask)) {
1320 const void *opts = NULL;
1322 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1323 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1325 if (ipv4_tun_to_nlattr(skb, &output->tun_key, opts,
1326 swkey->tun_opts_len))
1327 goto nla_put_failure;
1330 if (swkey->phy.in_port == DP_MAX_PORTS) {
1331 if (is_mask && (output->phy.in_port == 0xffff))
1332 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1333 goto nla_put_failure;
1336 upper_u16 = !is_mask ? 0 : 0xffff;
1338 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1339 (upper_u16 << 16) | output->phy.in_port))
1340 goto nla_put_failure;
1343 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1344 goto nla_put_failure;
1346 if (ovs_ct_put_key(output, skb))
1347 goto nla_put_failure;
1349 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1351 goto nla_put_failure;
1353 eth_key = nla_data(nla);
1354 ether_addr_copy(eth_key->eth_src, output->eth.src);
1355 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1357 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1359 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
1360 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1361 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
1362 goto nla_put_failure;
1363 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1364 if (!swkey->eth.tci)
1369 if (swkey->eth.type == htons(ETH_P_802_2)) {
1371 * Ethertype 802.2 is represented in the netlink with omitted
1372 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1373 * 0xffff in the mask attribute. Ethertype can also
1376 if (is_mask && output->eth.type)
1377 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1379 goto nla_put_failure;
1383 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1384 goto nla_put_failure;
1386 if (swkey->eth.type == htons(ETH_P_IP)) {
1387 struct ovs_key_ipv4 *ipv4_key;
1389 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1391 goto nla_put_failure;
1392 ipv4_key = nla_data(nla);
1393 ipv4_key->ipv4_src = output->ipv4.addr.src;
1394 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1395 ipv4_key->ipv4_proto = output->ip.proto;
1396 ipv4_key->ipv4_tos = output->ip.tos;
1397 ipv4_key->ipv4_ttl = output->ip.ttl;
1398 ipv4_key->ipv4_frag = output->ip.frag;
1399 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1400 struct ovs_key_ipv6 *ipv6_key;
1402 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1404 goto nla_put_failure;
1405 ipv6_key = nla_data(nla);
1406 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1407 sizeof(ipv6_key->ipv6_src));
1408 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1409 sizeof(ipv6_key->ipv6_dst));
1410 ipv6_key->ipv6_label = output->ipv6.label;
1411 ipv6_key->ipv6_proto = output->ip.proto;
1412 ipv6_key->ipv6_tclass = output->ip.tos;
1413 ipv6_key->ipv6_hlimit = output->ip.ttl;
1414 ipv6_key->ipv6_frag = output->ip.frag;
1415 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1416 swkey->eth.type == htons(ETH_P_RARP)) {
1417 struct ovs_key_arp *arp_key;
1419 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1421 goto nla_put_failure;
1422 arp_key = nla_data(nla);
1423 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1424 arp_key->arp_sip = output->ipv4.addr.src;
1425 arp_key->arp_tip = output->ipv4.addr.dst;
1426 arp_key->arp_op = htons(output->ip.proto);
1427 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1428 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1429 } else if (eth_p_mpls(swkey->eth.type)) {
1430 struct ovs_key_mpls *mpls_key;
1432 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1434 goto nla_put_failure;
1435 mpls_key = nla_data(nla);
1436 mpls_key->mpls_lse = output->mpls.top_lse;
1439 if ((swkey->eth.type == htons(ETH_P_IP) ||
1440 swkey->eth.type == htons(ETH_P_IPV6)) &&
1441 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1443 if (swkey->ip.proto == IPPROTO_TCP) {
1444 struct ovs_key_tcp *tcp_key;
1446 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1448 goto nla_put_failure;
1449 tcp_key = nla_data(nla);
1450 tcp_key->tcp_src = output->tp.src;
1451 tcp_key->tcp_dst = output->tp.dst;
1452 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1454 goto nla_put_failure;
1455 } else if (swkey->ip.proto == IPPROTO_UDP) {
1456 struct ovs_key_udp *udp_key;
1458 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1460 goto nla_put_failure;
1461 udp_key = nla_data(nla);
1462 udp_key->udp_src = output->tp.src;
1463 udp_key->udp_dst = output->tp.dst;
1464 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1465 struct ovs_key_sctp *sctp_key;
1467 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1469 goto nla_put_failure;
1470 sctp_key = nla_data(nla);
1471 sctp_key->sctp_src = output->tp.src;
1472 sctp_key->sctp_dst = output->tp.dst;
1473 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1474 swkey->ip.proto == IPPROTO_ICMP) {
1475 struct ovs_key_icmp *icmp_key;
1477 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1479 goto nla_put_failure;
1480 icmp_key = nla_data(nla);
1481 icmp_key->icmp_type = ntohs(output->tp.src);
1482 icmp_key->icmp_code = ntohs(output->tp.dst);
1483 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1484 swkey->ip.proto == IPPROTO_ICMPV6) {
1485 struct ovs_key_icmpv6 *icmpv6_key;
1487 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1488 sizeof(*icmpv6_key));
1490 goto nla_put_failure;
1491 icmpv6_key = nla_data(nla);
1492 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1493 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1495 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1496 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1497 struct ovs_key_nd *nd_key;
1499 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1501 goto nla_put_failure;
1502 nd_key = nla_data(nla);
1503 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1504 sizeof(nd_key->nd_target));
1505 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1506 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1513 nla_nest_end(skb, encap);
1521 int ovs_nla_put_key(const struct sw_flow_key *swkey,
1522 const struct sw_flow_key *output, int attr, bool is_mask,
1523 struct sk_buff *skb)
1528 nla = nla_nest_start(skb, attr);
1531 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
1534 nla_nest_end(skb, nla);
1539 /* Called with ovs_mutex or RCU read lock. */
1540 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
1542 if (ovs_identifier_is_ufid(&flow->id))
1543 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
1546 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
1547 OVS_FLOW_ATTR_KEY, false, skb);
1550 /* Called with ovs_mutex or RCU read lock. */
1551 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
1553 return ovs_nla_put_key(&flow->key, &flow->key,
1554 OVS_FLOW_ATTR_KEY, false, skb);
1557 /* Called with ovs_mutex or RCU read lock. */
1558 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
1560 return ovs_nla_put_key(&flow->key, &flow->mask->key,
1561 OVS_FLOW_ATTR_MASK, true, skb);
1564 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1566 static struct sw_flow_actions *nla_alloc_flow_actions(int size, bool log)
1568 struct sw_flow_actions *sfa;
1570 if (size > MAX_ACTIONS_BUFSIZE) {
1571 OVS_NLERR(log, "Flow action size %u bytes exceeds max", size);
1572 return ERR_PTR(-EINVAL);
1575 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1577 return ERR_PTR(-ENOMEM);
1579 sfa->actions_len = 0;
1583 static void ovs_nla_free_set_action(const struct nlattr *a)
1585 const struct nlattr *ovs_key = nla_data(a);
1586 struct ovs_tunnel_info *ovs_tun;
1588 switch (nla_type(ovs_key)) {
1589 case OVS_KEY_ATTR_TUNNEL_INFO:
1590 ovs_tun = nla_data(ovs_key);
1591 dst_release((struct dst_entry *)ovs_tun->tun_dst);
1596 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1598 const struct nlattr *a;
1604 nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
1605 switch (nla_type(a)) {
1606 case OVS_ACTION_ATTR_SET:
1607 ovs_nla_free_set_action(a);
1609 case OVS_ACTION_ATTR_CT:
1610 ovs_ct_free_action(a);
1618 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
1620 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
1623 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1624 * The caller must hold rcu_read_lock for this to be sensible. */
1625 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
1627 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
1630 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1631 int attr_len, bool log)
1634 struct sw_flow_actions *acts;
1636 int req_size = NLA_ALIGN(attr_len);
1637 int next_offset = offsetof(struct sw_flow_actions, actions) +
1638 (*sfa)->actions_len;
1640 if (req_size <= (ksize(*sfa) - next_offset))
1643 new_acts_size = ksize(*sfa) * 2;
1645 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1646 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1647 return ERR_PTR(-EMSGSIZE);
1648 new_acts_size = MAX_ACTIONS_BUFSIZE;
1651 acts = nla_alloc_flow_actions(new_acts_size, log);
1653 return (void *)acts;
1655 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1656 acts->actions_len = (*sfa)->actions_len;
1657 acts->orig_len = (*sfa)->orig_len;
1662 (*sfa)->actions_len += req_size;
1663 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1666 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1667 int attrtype, void *data, int len, bool log)
1671 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
1675 a->nla_type = attrtype;
1676 a->nla_len = nla_attr_size(len);
1679 memcpy(nla_data(a), data, len);
1680 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1685 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
1690 a = __add_action(sfa, attrtype, data, len, log);
1692 return PTR_ERR_OR_ZERO(a);
1695 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1696 int attrtype, bool log)
1698 int used = (*sfa)->actions_len;
1701 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
1708 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1711 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1714 a->nla_len = sfa->actions_len - st_offset;
1717 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
1718 const struct sw_flow_key *key,
1719 int depth, struct sw_flow_actions **sfa,
1720 __be16 eth_type, __be16 vlan_tci, bool log);
1722 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
1723 const struct sw_flow_key *key, int depth,
1724 struct sw_flow_actions **sfa,
1725 __be16 eth_type, __be16 vlan_tci, bool log)
1727 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1728 const struct nlattr *probability, *actions;
1729 const struct nlattr *a;
1730 int rem, start, err, st_acts;
1732 memset(attrs, 0, sizeof(attrs));
1733 nla_for_each_nested(a, attr, rem) {
1734 int type = nla_type(a);
1735 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1742 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1743 if (!probability || nla_len(probability) != sizeof(u32))
1746 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1747 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1750 /* validation done, copy sample action. */
1751 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
1754 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1755 nla_data(probability), sizeof(u32), log);
1758 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS, log);
1762 err = __ovs_nla_copy_actions(net, actions, key, depth + 1, sfa,
1763 eth_type, vlan_tci, log);
1767 add_nested_action_end(*sfa, st_acts);
1768 add_nested_action_end(*sfa, start);
1773 void ovs_match_init(struct sw_flow_match *match,
1774 struct sw_flow_key *key,
1775 struct sw_flow_mask *mask)
1777 memset(match, 0, sizeof(*match));
1781 memset(key, 0, sizeof(*key));
1784 memset(&mask->key, 0, sizeof(mask->key));
1785 mask->range.start = mask->range.end = 0;
1789 static int validate_geneve_opts(struct sw_flow_key *key)
1791 struct geneve_opt *option;
1792 int opts_len = key->tun_opts_len;
1793 bool crit_opt = false;
1795 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
1796 while (opts_len > 0) {
1799 if (opts_len < sizeof(*option))
1802 len = sizeof(*option) + option->length * 4;
1806 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
1808 option = (struct geneve_opt *)((u8 *)option + len);
1812 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
1817 static int validate_and_copy_set_tun(const struct nlattr *attr,
1818 struct sw_flow_actions **sfa, bool log)
1820 struct sw_flow_match match;
1821 struct sw_flow_key key;
1822 struct metadata_dst *tun_dst;
1823 struct ip_tunnel_info *tun_info;
1824 struct ovs_tunnel_info *ovs_tun;
1826 int err = 0, start, opts_type;
1828 ovs_match_init(&match, &key, NULL);
1829 opts_type = ipv4_tun_from_nlattr(nla_data(attr), &match, false, log);
1833 if (key.tun_opts_len) {
1834 switch (opts_type) {
1835 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
1836 err = validate_geneve_opts(&key);
1840 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
1845 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
1849 tun_dst = metadata_dst_alloc(key.tun_opts_len, GFP_KERNEL);
1853 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
1854 sizeof(*ovs_tun), log);
1856 dst_release((struct dst_entry *)tun_dst);
1860 ovs_tun = nla_data(a);
1861 ovs_tun->tun_dst = tun_dst;
1863 tun_info = &tun_dst->u.tun_info;
1864 tun_info->mode = IP_TUNNEL_INFO_TX;
1865 tun_info->key = key.tun_key;
1866 tun_info->options_len = key.tun_opts_len;
1868 if (tun_info->options_len) {
1869 /* We need to store the options in the action itself since
1870 * everything else will go away after flow setup. We can append
1871 * it to tun_info and then point there.
1873 memcpy((tun_info + 1),
1874 TUN_METADATA_OPTS(&key, key.tun_opts_len), key.tun_opts_len);
1875 tun_info->options = (tun_info + 1);
1877 tun_info->options = NULL;
1880 add_nested_action_end(*sfa, start);
1885 /* Return false if there are any non-masked bits set.
1886 * Mask follows data immediately, before any netlink padding.
1888 static bool validate_masked(u8 *data, int len)
1890 u8 *mask = data + len;
1893 if (*data++ & ~*mask++)
1899 static int validate_set(const struct nlattr *a,
1900 const struct sw_flow_key *flow_key,
1901 struct sw_flow_actions **sfa,
1902 bool *skip_copy, __be16 eth_type, bool masked, bool log)
1904 const struct nlattr *ovs_key = nla_data(a);
1905 int key_type = nla_type(ovs_key);
1908 /* There can be only one key in a action */
1909 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1912 key_len = nla_len(ovs_key);
1916 if (key_type > OVS_KEY_ATTR_MAX ||
1917 (ovs_key_lens[key_type].len != key_len &&
1918 ovs_key_lens[key_type].len != OVS_ATTR_NESTED))
1921 if (masked && !validate_masked(nla_data(ovs_key), key_len))
1925 const struct ovs_key_ipv4 *ipv4_key;
1926 const struct ovs_key_ipv6 *ipv6_key;
1929 case OVS_KEY_ATTR_PRIORITY:
1930 case OVS_KEY_ATTR_SKB_MARK:
1931 case OVS_KEY_ATTR_CT_MARK:
1932 case OVS_KEY_ATTR_ETHERNET:
1935 case OVS_KEY_ATTR_TUNNEL:
1936 if (eth_p_mpls(eth_type))
1940 return -EINVAL; /* Masked tunnel set not supported. */
1943 err = validate_and_copy_set_tun(a, sfa, log);
1948 case OVS_KEY_ATTR_IPV4:
1949 if (eth_type != htons(ETH_P_IP))
1952 ipv4_key = nla_data(ovs_key);
1955 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
1957 /* Non-writeable fields. */
1958 if (mask->ipv4_proto || mask->ipv4_frag)
1961 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
1964 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
1969 case OVS_KEY_ATTR_IPV6:
1970 if (eth_type != htons(ETH_P_IPV6))
1973 ipv6_key = nla_data(ovs_key);
1976 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
1978 /* Non-writeable fields. */
1979 if (mask->ipv6_proto || mask->ipv6_frag)
1982 /* Invalid bits in the flow label mask? */
1983 if (ntohl(mask->ipv6_label) & 0xFFF00000)
1986 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
1989 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
1992 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
1997 case OVS_KEY_ATTR_TCP:
1998 if ((eth_type != htons(ETH_P_IP) &&
1999 eth_type != htons(ETH_P_IPV6)) ||
2000 flow_key->ip.proto != IPPROTO_TCP)
2005 case OVS_KEY_ATTR_UDP:
2006 if ((eth_type != htons(ETH_P_IP) &&
2007 eth_type != htons(ETH_P_IPV6)) ||
2008 flow_key->ip.proto != IPPROTO_UDP)
2013 case OVS_KEY_ATTR_MPLS:
2014 if (!eth_p_mpls(eth_type))
2018 case OVS_KEY_ATTR_SCTP:
2019 if ((eth_type != htons(ETH_P_IP) &&
2020 eth_type != htons(ETH_P_IPV6)) ||
2021 flow_key->ip.proto != IPPROTO_SCTP)
2030 /* Convert non-masked non-tunnel set actions to masked set actions. */
2031 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2032 int start, len = key_len * 2;
2037 start = add_nested_action_start(sfa,
2038 OVS_ACTION_ATTR_SET_TO_MASKED,
2043 at = __add_action(sfa, key_type, NULL, len, log);
2047 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2048 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
2049 /* Clear non-writeable bits from otherwise writeable fields. */
2050 if (key_type == OVS_KEY_ATTR_IPV6) {
2051 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2053 mask->ipv6_label &= htonl(0x000FFFFF);
2055 add_nested_action_end(*sfa, start);
2061 static int validate_userspace(const struct nlattr *attr)
2063 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2064 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2065 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2066 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2068 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2071 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
2072 attr, userspace_policy);
2076 if (!a[OVS_USERSPACE_ATTR_PID] ||
2077 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2083 static int copy_action(const struct nlattr *from,
2084 struct sw_flow_actions **sfa, bool log)
2086 int totlen = NLA_ALIGN(from->nla_len);
2089 to = reserve_sfa_size(sfa, from->nla_len, log);
2093 memcpy(to, from, totlen);
2097 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2098 const struct sw_flow_key *key,
2099 int depth, struct sw_flow_actions **sfa,
2100 __be16 eth_type, __be16 vlan_tci, bool log)
2102 const struct nlattr *a;
2105 if (depth >= SAMPLE_ACTION_DEPTH)
2108 nla_for_each_nested(a, attr, rem) {
2109 /* Expected argument lengths, (u32)-1 for variable length. */
2110 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2111 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2112 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2113 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2114 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2115 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2116 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2117 [OVS_ACTION_ATTR_POP_VLAN] = 0,
2118 [OVS_ACTION_ATTR_SET] = (u32)-1,
2119 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2120 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2121 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2122 [OVS_ACTION_ATTR_CT] = (u32)-1,
2124 const struct ovs_action_push_vlan *vlan;
2125 int type = nla_type(a);
2128 if (type > OVS_ACTION_ATTR_MAX ||
2129 (action_lens[type] != nla_len(a) &&
2130 action_lens[type] != (u32)-1))
2135 case OVS_ACTION_ATTR_UNSPEC:
2138 case OVS_ACTION_ATTR_USERSPACE:
2139 err = validate_userspace(a);
2144 case OVS_ACTION_ATTR_OUTPUT:
2145 if (nla_get_u32(a) >= DP_MAX_PORTS)
2149 case OVS_ACTION_ATTR_HASH: {
2150 const struct ovs_action_hash *act_hash = nla_data(a);
2152 switch (act_hash->hash_alg) {
2153 case OVS_HASH_ALG_L4:
2162 case OVS_ACTION_ATTR_POP_VLAN:
2163 vlan_tci = htons(0);
2166 case OVS_ACTION_ATTR_PUSH_VLAN:
2168 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
2170 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2172 vlan_tci = vlan->vlan_tci;
2175 case OVS_ACTION_ATTR_RECIRC:
2178 case OVS_ACTION_ATTR_PUSH_MPLS: {
2179 const struct ovs_action_push_mpls *mpls = nla_data(a);
2181 if (!eth_p_mpls(mpls->mpls_ethertype))
2183 /* Prohibit push MPLS other than to a white list
2184 * for packets that have a known tag order.
2186 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2187 (eth_type != htons(ETH_P_IP) &&
2188 eth_type != htons(ETH_P_IPV6) &&
2189 eth_type != htons(ETH_P_ARP) &&
2190 eth_type != htons(ETH_P_RARP) &&
2191 !eth_p_mpls(eth_type)))
2193 eth_type = mpls->mpls_ethertype;
2197 case OVS_ACTION_ATTR_POP_MPLS:
2198 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2199 !eth_p_mpls(eth_type))
2202 /* Disallow subsequent L2.5+ set and mpls_pop actions
2203 * as there is no check here to ensure that the new
2204 * eth_type is valid and thus set actions could
2205 * write off the end of the packet or otherwise
2208 * Support for these actions is planned using packet
2211 eth_type = htons(0);
2214 case OVS_ACTION_ATTR_SET:
2215 err = validate_set(a, key, sfa,
2216 &skip_copy, eth_type, false, log);
2221 case OVS_ACTION_ATTR_SET_MASKED:
2222 err = validate_set(a, key, sfa,
2223 &skip_copy, eth_type, true, log);
2228 case OVS_ACTION_ATTR_SAMPLE:
2229 err = validate_and_copy_sample(net, a, key, depth, sfa,
2230 eth_type, vlan_tci, log);
2236 case OVS_ACTION_ATTR_CT:
2237 err = ovs_ct_copy_action(net, a, key, sfa, log);
2244 OVS_NLERR(log, "Unknown Action type %d", type);
2248 err = copy_action(a, sfa, log);
2260 /* 'key' must be the masked key. */
2261 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2262 const struct sw_flow_key *key,
2263 struct sw_flow_actions **sfa, bool log)
2267 *sfa = nla_alloc_flow_actions(nla_len(attr), log);
2269 return PTR_ERR(*sfa);
2271 (*sfa)->orig_len = nla_len(attr);
2272 err = __ovs_nla_copy_actions(net, attr, key, 0, sfa, key->eth.type,
2275 ovs_nla_free_flow_actions(*sfa);
2280 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
2282 const struct nlattr *a;
2283 struct nlattr *start;
2286 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
2290 nla_for_each_nested(a, attr, rem) {
2291 int type = nla_type(a);
2292 struct nlattr *st_sample;
2295 case OVS_SAMPLE_ATTR_PROBABILITY:
2296 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
2297 sizeof(u32), nla_data(a)))
2300 case OVS_SAMPLE_ATTR_ACTIONS:
2301 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
2304 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
2307 nla_nest_end(skb, st_sample);
2312 nla_nest_end(skb, start);
2316 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
2318 const struct nlattr *ovs_key = nla_data(a);
2319 int key_type = nla_type(ovs_key);
2320 struct nlattr *start;
2324 case OVS_KEY_ATTR_TUNNEL_INFO: {
2325 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
2326 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
2328 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2332 err = ipv4_tun_to_nlattr(skb, &tun_info->key,
2333 tun_info->options_len ?
2334 tun_info->options : NULL,
2335 tun_info->options_len);
2338 nla_nest_end(skb, start);
2342 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
2350 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
2351 struct sk_buff *skb)
2353 const struct nlattr *ovs_key = nla_data(a);
2355 size_t key_len = nla_len(ovs_key) / 2;
2357 /* Revert the conversion we did from a non-masked set action to
2358 * masked set action.
2360 nla = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2364 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
2367 nla_nest_end(skb, nla);
2371 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
2373 const struct nlattr *a;
2376 nla_for_each_attr(a, attr, len, rem) {
2377 int type = nla_type(a);
2380 case OVS_ACTION_ATTR_SET:
2381 err = set_action_to_attr(a, skb);
2386 case OVS_ACTION_ATTR_SET_TO_MASKED:
2387 err = masked_set_action_to_set_action_attr(a, skb);
2392 case OVS_ACTION_ATTR_SAMPLE:
2393 err = sample_action_to_attr(a, skb);
2398 case OVS_ACTION_ATTR_CT:
2399 err = ovs_ct_action_to_attr(nla_data(a), skb);
2405 if (nla_put(skb, type, nla_len(a), nla_data(a)))