2 * Copyright (c) 2007-2015 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
21 #include <linux/uaccess.h>
22 #include <linux/netdevice.h>
23 #include <linux/etherdevice.h>
24 #include <linux/if_ether.h>
25 #include <linux/if_vlan.h>
26 #include <net/llc_pdu.h>
27 #include <linux/kernel.h>
28 #include <linux/jhash.h>
29 #include <linux/jiffies.h>
30 #include <linux/llc.h>
31 #include <linux/module.h>
33 #include <linux/rcupdate.h>
34 #include <linux/if_arp.h>
36 #include <linux/ipv6.h>
37 #include <linux/sctp.h>
38 #include <linux/tcp.h>
39 #include <linux/udp.h>
40 #include <linux/icmp.h>
41 #include <linux/icmpv6.h>
42 #include <linux/rculist.h>
43 #include <net/geneve.h>
46 #include <net/ndisc.h>
48 #include <net/vxlan.h>
51 #include "conntrack.h"
53 #include "flow_netlink.h"
58 const struct ovs_len_tbl *next;
61 #define OVS_ATTR_NESTED -1
62 #define OVS_ATTR_VARIABLE -2
64 static void update_range(struct sw_flow_match *match,
65 size_t offset, size_t size, bool is_mask)
67 struct sw_flow_key_range *range;
68 size_t start = rounddown(offset, sizeof(long));
69 size_t end = roundup(offset + size, sizeof(long));
72 range = &match->range;
74 range = &match->mask->range;
76 if (range->start == range->end) {
82 if (range->start > start)
89 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
91 update_range(match, offsetof(struct sw_flow_key, field), \
92 sizeof((match)->key->field), is_mask); \
94 (match)->mask->key.field = value; \
96 (match)->key->field = value; \
99 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
101 update_range(match, offset, len, is_mask); \
103 memcpy((u8 *)&(match)->mask->key + offset, value_p, len);\
105 memcpy((u8 *)(match)->key + offset, value_p, len); \
108 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
109 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
110 value_p, len, is_mask)
112 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
114 update_range(match, offsetof(struct sw_flow_key, field), \
115 sizeof((match)->key->field), is_mask); \
117 memset((u8 *)&(match)->mask->key.field, value, \
118 sizeof((match)->mask->key.field)); \
120 memset((u8 *)&(match)->key->field, value, \
121 sizeof((match)->key->field)); \
124 static bool match_validate(const struct sw_flow_match *match,
125 u64 key_attrs, u64 mask_attrs, bool log)
127 u64 key_expected = 1ULL << OVS_KEY_ATTR_ETHERNET;
128 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
130 /* The following mask attributes allowed only if they
131 * pass the validation tests.
133 mask_allowed &= ~((1ULL << OVS_KEY_ATTR_IPV4)
134 | (1ULL << OVS_KEY_ATTR_IPV6)
135 | (1ULL << OVS_KEY_ATTR_TCP)
136 | (1ULL << OVS_KEY_ATTR_TCP_FLAGS)
137 | (1ULL << OVS_KEY_ATTR_UDP)
138 | (1ULL << OVS_KEY_ATTR_SCTP)
139 | (1ULL << OVS_KEY_ATTR_ICMP)
140 | (1ULL << OVS_KEY_ATTR_ICMPV6)
141 | (1ULL << OVS_KEY_ATTR_ARP)
142 | (1ULL << OVS_KEY_ATTR_ND)
143 | (1ULL << OVS_KEY_ATTR_MPLS));
145 /* Always allowed mask fields. */
146 mask_allowed |= ((1ULL << OVS_KEY_ATTR_TUNNEL)
147 | (1ULL << OVS_KEY_ATTR_IN_PORT)
148 | (1ULL << OVS_KEY_ATTR_ETHERTYPE));
150 /* Check key attributes. */
151 if (match->key->eth.type == htons(ETH_P_ARP)
152 || match->key->eth.type == htons(ETH_P_RARP)) {
153 key_expected |= 1ULL << OVS_KEY_ATTR_ARP;
154 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
155 mask_allowed |= 1ULL << OVS_KEY_ATTR_ARP;
158 if (eth_p_mpls(match->key->eth.type)) {
159 key_expected |= 1ULL << OVS_KEY_ATTR_MPLS;
160 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
161 mask_allowed |= 1ULL << OVS_KEY_ATTR_MPLS;
164 if (match->key->eth.type == htons(ETH_P_IP)) {
165 key_expected |= 1ULL << OVS_KEY_ATTR_IPV4;
166 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
167 mask_allowed |= 1ULL << OVS_KEY_ATTR_IPV4;
169 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
170 if (match->key->ip.proto == IPPROTO_UDP) {
171 key_expected |= 1ULL << OVS_KEY_ATTR_UDP;
172 if (match->mask && (match->mask->key.ip.proto == 0xff))
173 mask_allowed |= 1ULL << OVS_KEY_ATTR_UDP;
176 if (match->key->ip.proto == IPPROTO_SCTP) {
177 key_expected |= 1ULL << OVS_KEY_ATTR_SCTP;
178 if (match->mask && (match->mask->key.ip.proto == 0xff))
179 mask_allowed |= 1ULL << OVS_KEY_ATTR_SCTP;
182 if (match->key->ip.proto == IPPROTO_TCP) {
183 key_expected |= 1ULL << OVS_KEY_ATTR_TCP;
184 key_expected |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
185 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
186 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP;
187 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
191 if (match->key->ip.proto == IPPROTO_ICMP) {
192 key_expected |= 1ULL << OVS_KEY_ATTR_ICMP;
193 if (match->mask && (match->mask->key.ip.proto == 0xff))
194 mask_allowed |= 1ULL << OVS_KEY_ATTR_ICMP;
199 if (match->key->eth.type == htons(ETH_P_IPV6)) {
200 key_expected |= 1ULL << OVS_KEY_ATTR_IPV6;
201 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
202 mask_allowed |= 1ULL << OVS_KEY_ATTR_IPV6;
204 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
205 if (match->key->ip.proto == IPPROTO_UDP) {
206 key_expected |= 1ULL << OVS_KEY_ATTR_UDP;
207 if (match->mask && (match->mask->key.ip.proto == 0xff))
208 mask_allowed |= 1ULL << OVS_KEY_ATTR_UDP;
211 if (match->key->ip.proto == IPPROTO_SCTP) {
212 key_expected |= 1ULL << OVS_KEY_ATTR_SCTP;
213 if (match->mask && (match->mask->key.ip.proto == 0xff))
214 mask_allowed |= 1ULL << OVS_KEY_ATTR_SCTP;
217 if (match->key->ip.proto == IPPROTO_TCP) {
218 key_expected |= 1ULL << OVS_KEY_ATTR_TCP;
219 key_expected |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
220 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
221 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP;
222 mask_allowed |= 1ULL << OVS_KEY_ATTR_TCP_FLAGS;
226 if (match->key->ip.proto == IPPROTO_ICMPV6) {
227 key_expected |= 1ULL << OVS_KEY_ATTR_ICMPV6;
228 if (match->mask && (match->mask->key.ip.proto == 0xff))
229 mask_allowed |= 1ULL << OVS_KEY_ATTR_ICMPV6;
231 if (match->key->tp.src ==
232 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
233 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
234 key_expected |= 1ULL << OVS_KEY_ATTR_ND;
235 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
236 mask_allowed |= 1ULL << OVS_KEY_ATTR_ND;
242 if ((key_attrs & key_expected) != key_expected) {
243 /* Key attributes check failed. */
244 OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
245 (unsigned long long)key_attrs,
246 (unsigned long long)key_expected);
250 if ((mask_attrs & mask_allowed) != mask_attrs) {
251 /* Mask attributes check failed. */
252 OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
253 (unsigned long long)mask_attrs,
254 (unsigned long long)mask_allowed);
261 size_t ovs_tun_key_attr_size(void)
263 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
264 * updating this function.
266 return nla_total_size(8) /* OVS_TUNNEL_KEY_ATTR_ID */
267 + nla_total_size(4) /* OVS_TUNNEL_KEY_ATTR_IPV4_SRC */
268 + nla_total_size(4) /* OVS_TUNNEL_KEY_ATTR_IPV4_DST */
269 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
270 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
271 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
272 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
273 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
274 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
275 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS is mutually exclusive with
276 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
278 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
279 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
282 size_t ovs_key_attr_size(void)
284 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
285 * updating this function.
287 BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO != 26);
289 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
290 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
291 + ovs_tun_key_attr_size()
292 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
293 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
294 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
295 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
296 + nla_total_size(4) /* OVS_KEY_ATTR_CT_STATE */
297 + nla_total_size(2) /* OVS_KEY_ATTR_CT_ZONE */
298 + nla_total_size(4) /* OVS_KEY_ATTR_CT_MARK */
299 + nla_total_size(16) /* OVS_KEY_ATTR_CT_LABELS */
300 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
301 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
302 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
303 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
304 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
305 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
306 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
307 + nla_total_size(28); /* OVS_KEY_ATTR_ND */
310 static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
311 [OVS_VXLAN_EXT_GBP] = { .len = sizeof(u32) },
314 static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
315 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
316 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
317 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
318 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
319 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
320 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
321 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
322 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
323 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
324 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
325 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_VARIABLE },
326 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED,
327 .next = ovs_vxlan_ext_key_lens },
330 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
331 static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
332 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
333 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
334 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
335 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
336 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
337 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
338 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
339 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
340 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
341 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
342 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
343 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
344 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
345 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
346 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
347 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
348 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
349 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
350 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
351 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
352 .next = ovs_tunnel_key_lens, },
353 [OVS_KEY_ATTR_MPLS] = { .len = sizeof(struct ovs_key_mpls) },
354 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u32) },
355 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
356 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
357 [OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
360 static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
362 return expected_len == attr_len ||
363 expected_len == OVS_ATTR_NESTED ||
364 expected_len == OVS_ATTR_VARIABLE;
367 static bool is_all_zero(const u8 *fp, size_t size)
374 for (i = 0; i < size; i++)
381 static int __parse_flow_nlattrs(const struct nlattr *attr,
382 const struct nlattr *a[],
383 u64 *attrsp, bool log, bool nz)
385 const struct nlattr *nla;
390 nla_for_each_nested(nla, attr, rem) {
391 u16 type = nla_type(nla);
394 if (type > OVS_KEY_ATTR_MAX) {
395 OVS_NLERR(log, "Key type %d is out of range max %d",
396 type, OVS_KEY_ATTR_MAX);
400 if (attrs & (1ULL << type)) {
401 OVS_NLERR(log, "Duplicate key (type %d).", type);
405 expected_len = ovs_key_lens[type].len;
406 if (!check_attr_len(nla_len(nla), expected_len)) {
407 OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
408 type, nla_len(nla), expected_len);
412 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
413 attrs |= 1ULL << type;
418 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
426 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
427 const struct nlattr *a[], u64 *attrsp,
430 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
433 static int parse_flow_nlattrs(const struct nlattr *attr,
434 const struct nlattr *a[], u64 *attrsp,
437 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
440 static int genev_tun_opt_from_nlattr(const struct nlattr *a,
441 struct sw_flow_match *match, bool is_mask,
444 unsigned long opt_key_offset;
446 if (nla_len(a) > sizeof(match->key->tun_opts)) {
447 OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
448 nla_len(a), sizeof(match->key->tun_opts));
452 if (nla_len(a) % 4 != 0) {
453 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
458 /* We need to record the length of the options passed
459 * down, otherwise packets with the same format but
460 * additional options will be silently matched.
463 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
466 /* This is somewhat unusual because it looks at
467 * both the key and mask while parsing the
468 * attributes (and by extension assumes the key
469 * is parsed first). Normally, we would verify
470 * that each is the correct length and that the
471 * attributes line up in the validate function.
472 * However, that is difficult because this is
473 * variable length and we won't have the
476 if (match->key->tun_opts_len != nla_len(a)) {
477 OVS_NLERR(log, "Geneve option len %d != mask len %d",
478 match->key->tun_opts_len, nla_len(a));
482 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
485 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
486 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
487 nla_len(a), is_mask);
491 static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
492 struct sw_flow_match *match, bool is_mask,
497 unsigned long opt_key_offset;
498 struct vxlan_metadata opts;
500 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
502 memset(&opts, 0, sizeof(opts));
503 nla_for_each_nested(a, attr, rem) {
504 int type = nla_type(a);
506 if (type > OVS_VXLAN_EXT_MAX) {
507 OVS_NLERR(log, "VXLAN extension %d out of range max %d",
508 type, OVS_VXLAN_EXT_MAX);
512 if (!check_attr_len(nla_len(a),
513 ovs_vxlan_ext_key_lens[type].len)) {
514 OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
516 ovs_vxlan_ext_key_lens[type].len);
521 case OVS_VXLAN_EXT_GBP:
522 opts.gbp = nla_get_u32(a);
525 OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
531 OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
537 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
539 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
541 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
542 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
547 static int ipv4_tun_from_nlattr(const struct nlattr *attr,
548 struct sw_flow_match *match, bool is_mask,
554 __be16 tun_flags = 0;
557 nla_for_each_nested(a, attr, rem) {
558 int type = nla_type(a);
561 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
562 OVS_NLERR(log, "Tunnel attr %d out of range max %d",
563 type, OVS_TUNNEL_KEY_ATTR_MAX);
567 if (!check_attr_len(nla_len(a),
568 ovs_tunnel_key_lens[type].len)) {
569 OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
570 type, nla_len(a), ovs_tunnel_key_lens[type].len);
575 case OVS_TUNNEL_KEY_ATTR_ID:
576 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
577 nla_get_be64(a), is_mask);
578 tun_flags |= TUNNEL_KEY;
580 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
581 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
582 nla_get_in_addr(a), is_mask);
584 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
585 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
586 nla_get_in_addr(a), is_mask);
588 case OVS_TUNNEL_KEY_ATTR_TOS:
589 SW_FLOW_KEY_PUT(match, tun_key.tos,
590 nla_get_u8(a), is_mask);
592 case OVS_TUNNEL_KEY_ATTR_TTL:
593 SW_FLOW_KEY_PUT(match, tun_key.ttl,
594 nla_get_u8(a), is_mask);
597 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
598 tun_flags |= TUNNEL_DONT_FRAGMENT;
600 case OVS_TUNNEL_KEY_ATTR_CSUM:
601 tun_flags |= TUNNEL_CSUM;
603 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
604 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
605 nla_get_be16(a), is_mask);
607 case OVS_TUNNEL_KEY_ATTR_TP_DST:
608 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
609 nla_get_be16(a), is_mask);
611 case OVS_TUNNEL_KEY_ATTR_OAM:
612 tun_flags |= TUNNEL_OAM;
614 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
616 OVS_NLERR(log, "Multiple metadata blocks provided");
620 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
624 tun_flags |= TUNNEL_GENEVE_OPT;
627 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
629 OVS_NLERR(log, "Multiple metadata blocks provided");
633 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
637 tun_flags |= TUNNEL_VXLAN_OPT;
641 OVS_NLERR(log, "Unknown IPv4 tunnel attribute %d",
647 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
650 OVS_NLERR(log, "IPv4 tunnel attribute has %d unknown bytes.",
656 if (!match->key->tun_key.u.ipv4.dst) {
657 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
662 OVS_NLERR(log, "IPv4 tunnel TTL not specified.");
670 static int vxlan_opt_to_nlattr(struct sk_buff *skb,
671 const void *tun_opts, int swkey_tun_opts_len)
673 const struct vxlan_metadata *opts = tun_opts;
676 nla = nla_nest_start(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
680 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
683 nla_nest_end(skb, nla);
687 static int __ipv4_tun_to_nlattr(struct sk_buff *skb,
688 const struct ip_tunnel_key *output,
689 const void *tun_opts, int swkey_tun_opts_len)
691 if (output->tun_flags & TUNNEL_KEY &&
692 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
694 if (output->u.ipv4.src &&
695 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
698 if (output->u.ipv4.dst &&
699 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
703 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
705 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
707 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
708 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
710 if ((output->tun_flags & TUNNEL_CSUM) &&
711 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
713 if (output->tp_src &&
714 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
716 if (output->tp_dst &&
717 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
719 if ((output->tun_flags & TUNNEL_OAM) &&
720 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
723 if (output->tun_flags & TUNNEL_GENEVE_OPT &&
724 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
725 swkey_tun_opts_len, tun_opts))
727 else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
728 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
735 static int ipv4_tun_to_nlattr(struct sk_buff *skb,
736 const struct ip_tunnel_key *output,
737 const void *tun_opts, int swkey_tun_opts_len)
742 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
746 err = __ipv4_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len);
750 nla_nest_end(skb, nla);
754 int ovs_nla_put_egress_tunnel_key(struct sk_buff *skb,
755 const struct ip_tunnel_info *egress_tun_info,
756 const void *egress_tun_opts)
758 return __ipv4_tun_to_nlattr(skb, &egress_tun_info->key,
760 egress_tun_info->options_len);
763 static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
764 u64 *attrs, const struct nlattr **a,
765 bool is_mask, bool log)
767 if (*attrs & (1ULL << OVS_KEY_ATTR_DP_HASH)) {
768 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
770 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
771 *attrs &= ~(1ULL << OVS_KEY_ATTR_DP_HASH);
774 if (*attrs & (1ULL << OVS_KEY_ATTR_RECIRC_ID)) {
775 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
777 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
778 *attrs &= ~(1ULL << OVS_KEY_ATTR_RECIRC_ID);
781 if (*attrs & (1ULL << OVS_KEY_ATTR_PRIORITY)) {
782 SW_FLOW_KEY_PUT(match, phy.priority,
783 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
784 *attrs &= ~(1ULL << OVS_KEY_ATTR_PRIORITY);
787 if (*attrs & (1ULL << OVS_KEY_ATTR_IN_PORT)) {
788 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
791 in_port = 0xffffffff; /* Always exact match in_port. */
792 } else if (in_port >= DP_MAX_PORTS) {
793 OVS_NLERR(log, "Port %d exceeds max allowable %d",
794 in_port, DP_MAX_PORTS);
798 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
799 *attrs &= ~(1ULL << OVS_KEY_ATTR_IN_PORT);
800 } else if (!is_mask) {
801 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
804 if (*attrs & (1ULL << OVS_KEY_ATTR_SKB_MARK)) {
805 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
807 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
808 *attrs &= ~(1ULL << OVS_KEY_ATTR_SKB_MARK);
810 if (*attrs & (1ULL << OVS_KEY_ATTR_TUNNEL)) {
811 if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
814 *attrs &= ~(1ULL << OVS_KEY_ATTR_TUNNEL);
817 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
818 ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
819 u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
821 if (ct_state & ~CT_SUPPORTED_MASK) {
822 OVS_NLERR(log, "ct_state flags %08x unsupported",
827 SW_FLOW_KEY_PUT(match, ct.state, ct_state, is_mask);
828 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
830 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
831 ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
832 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
834 SW_FLOW_KEY_PUT(match, ct.zone, ct_zone, is_mask);
835 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
837 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
838 ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
839 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
841 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
842 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
844 if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
845 ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
846 const struct ovs_key_ct_labels *cl;
848 cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
849 SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
850 sizeof(*cl), is_mask);
851 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
856 static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
857 u64 attrs, const struct nlattr **a,
858 bool is_mask, bool log)
862 err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
866 if (attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) {
867 const struct ovs_key_ethernet *eth_key;
869 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
870 SW_FLOW_KEY_MEMCPY(match, eth.src,
871 eth_key->eth_src, ETH_ALEN, is_mask);
872 SW_FLOW_KEY_MEMCPY(match, eth.dst,
873 eth_key->eth_dst, ETH_ALEN, is_mask);
874 attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERNET);
877 if (attrs & (1ULL << OVS_KEY_ATTR_VLAN)) {
880 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
881 if (!(tci & htons(VLAN_TAG_PRESENT))) {
883 OVS_NLERR(log, "VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.");
885 OVS_NLERR(log, "VLAN TCI does not have VLAN_TAG_PRESENT bit set.");
890 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
891 attrs &= ~(1ULL << OVS_KEY_ATTR_VLAN);
894 if (attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) {
897 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
899 /* Always exact match EtherType. */
900 eth_type = htons(0xffff);
901 } else if (!eth_proto_is_802_3(eth_type)) {
902 OVS_NLERR(log, "EtherType %x is less than min %x",
903 ntohs(eth_type), ETH_P_802_3_MIN);
907 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
908 attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
909 } else if (!is_mask) {
910 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
913 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
914 const struct ovs_key_ipv4 *ipv4_key;
916 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
917 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
918 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
919 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
922 SW_FLOW_KEY_PUT(match, ip.proto,
923 ipv4_key->ipv4_proto, is_mask);
924 SW_FLOW_KEY_PUT(match, ip.tos,
925 ipv4_key->ipv4_tos, is_mask);
926 SW_FLOW_KEY_PUT(match, ip.ttl,
927 ipv4_key->ipv4_ttl, is_mask);
928 SW_FLOW_KEY_PUT(match, ip.frag,
929 ipv4_key->ipv4_frag, is_mask);
930 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
931 ipv4_key->ipv4_src, is_mask);
932 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
933 ipv4_key->ipv4_dst, is_mask);
934 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
937 if (attrs & (1ULL << OVS_KEY_ATTR_IPV6)) {
938 const struct ovs_key_ipv6 *ipv6_key;
940 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
941 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
942 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
943 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
947 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
949 "Invalid IPv6 flow label value (value=%x, max=%x).",
950 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
954 SW_FLOW_KEY_PUT(match, ipv6.label,
955 ipv6_key->ipv6_label, is_mask);
956 SW_FLOW_KEY_PUT(match, ip.proto,
957 ipv6_key->ipv6_proto, is_mask);
958 SW_FLOW_KEY_PUT(match, ip.tos,
959 ipv6_key->ipv6_tclass, is_mask);
960 SW_FLOW_KEY_PUT(match, ip.ttl,
961 ipv6_key->ipv6_hlimit, is_mask);
962 SW_FLOW_KEY_PUT(match, ip.frag,
963 ipv6_key->ipv6_frag, is_mask);
964 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
966 sizeof(match->key->ipv6.addr.src),
968 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
970 sizeof(match->key->ipv6.addr.dst),
973 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV6);
976 if (attrs & (1ULL << OVS_KEY_ATTR_ARP)) {
977 const struct ovs_key_arp *arp_key;
979 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
980 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
981 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
986 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
987 arp_key->arp_sip, is_mask);
988 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
989 arp_key->arp_tip, is_mask);
990 SW_FLOW_KEY_PUT(match, ip.proto,
991 ntohs(arp_key->arp_op), is_mask);
992 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
993 arp_key->arp_sha, ETH_ALEN, is_mask);
994 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
995 arp_key->arp_tha, ETH_ALEN, is_mask);
997 attrs &= ~(1ULL << OVS_KEY_ATTR_ARP);
1000 if (attrs & (1ULL << OVS_KEY_ATTR_MPLS)) {
1001 const struct ovs_key_mpls *mpls_key;
1003 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1004 SW_FLOW_KEY_PUT(match, mpls.top_lse,
1005 mpls_key->mpls_lse, is_mask);
1007 attrs &= ~(1ULL << OVS_KEY_ATTR_MPLS);
1010 if (attrs & (1ULL << OVS_KEY_ATTR_TCP)) {
1011 const struct ovs_key_tcp *tcp_key;
1013 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1014 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1015 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1016 attrs &= ~(1ULL << OVS_KEY_ATTR_TCP);
1019 if (attrs & (1ULL << OVS_KEY_ATTR_TCP_FLAGS)) {
1020 SW_FLOW_KEY_PUT(match, tp.flags,
1021 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1023 attrs &= ~(1ULL << OVS_KEY_ATTR_TCP_FLAGS);
1026 if (attrs & (1ULL << OVS_KEY_ATTR_UDP)) {
1027 const struct ovs_key_udp *udp_key;
1029 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1030 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1031 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1032 attrs &= ~(1ULL << OVS_KEY_ATTR_UDP);
1035 if (attrs & (1ULL << OVS_KEY_ATTR_SCTP)) {
1036 const struct ovs_key_sctp *sctp_key;
1038 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1039 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1040 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1041 attrs &= ~(1ULL << OVS_KEY_ATTR_SCTP);
1044 if (attrs & (1ULL << OVS_KEY_ATTR_ICMP)) {
1045 const struct ovs_key_icmp *icmp_key;
1047 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1048 SW_FLOW_KEY_PUT(match, tp.src,
1049 htons(icmp_key->icmp_type), is_mask);
1050 SW_FLOW_KEY_PUT(match, tp.dst,
1051 htons(icmp_key->icmp_code), is_mask);
1052 attrs &= ~(1ULL << OVS_KEY_ATTR_ICMP);
1055 if (attrs & (1ULL << OVS_KEY_ATTR_ICMPV6)) {
1056 const struct ovs_key_icmpv6 *icmpv6_key;
1058 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1059 SW_FLOW_KEY_PUT(match, tp.src,
1060 htons(icmpv6_key->icmpv6_type), is_mask);
1061 SW_FLOW_KEY_PUT(match, tp.dst,
1062 htons(icmpv6_key->icmpv6_code), is_mask);
1063 attrs &= ~(1ULL << OVS_KEY_ATTR_ICMPV6);
1066 if (attrs & (1ULL << OVS_KEY_ATTR_ND)) {
1067 const struct ovs_key_nd *nd_key;
1069 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1070 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1072 sizeof(match->key->ipv6.nd.target),
1074 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1075 nd_key->nd_sll, ETH_ALEN, is_mask);
1076 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1077 nd_key->nd_tll, ETH_ALEN, is_mask);
1078 attrs &= ~(1ULL << OVS_KEY_ATTR_ND);
1082 OVS_NLERR(log, "Unknown key attributes %llx",
1083 (unsigned long long)attrs);
1090 static void nlattr_set(struct nlattr *attr, u8 val,
1091 const struct ovs_len_tbl *tbl)
1096 /* The nlattr stream should already have been validated */
1097 nla_for_each_nested(nla, attr, rem) {
1098 if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED) {
1099 if (tbl[nla_type(nla)].next)
1100 tbl = tbl[nla_type(nla)].next;
1101 nlattr_set(nla, val, tbl);
1103 memset(nla_data(nla), val, nla_len(nla));
1106 if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1107 *(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1111 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1113 nlattr_set(attr, val, ovs_key_lens);
1117 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1118 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1119 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1120 * does not include any don't care bit.
1121 * @net: Used to determine per-namespace field support.
1122 * @match: receives the extracted flow match information.
1123 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1124 * sequence. The fields should of the packet that triggered the creation
1126 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1127 * attribute specifies the mask field of the wildcarded flow.
1128 * @log: Boolean to allow kernel error logging. Normally true, but when
1129 * probing for feature compatibility this should be passed in as false to
1130 * suppress unnecessary error logging.
1132 int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1133 const struct nlattr *nla_key,
1134 const struct nlattr *nla_mask,
1137 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1138 const struct nlattr *encap;
1139 struct nlattr *newmask = NULL;
1142 bool encap_valid = false;
1145 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1149 if ((key_attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
1150 (key_attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)) &&
1151 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
1154 if (!((key_attrs & (1ULL << OVS_KEY_ATTR_VLAN)) &&
1155 (key_attrs & (1ULL << OVS_KEY_ATTR_ENCAP)))) {
1156 OVS_NLERR(log, "Invalid Vlan frame.");
1160 key_attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
1161 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1162 encap = a[OVS_KEY_ATTR_ENCAP];
1163 key_attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
1166 if (tci & htons(VLAN_TAG_PRESENT)) {
1167 err = parse_flow_nlattrs(encap, a, &key_attrs, log);
1171 /* Corner case for truncated 802.1Q header. */
1172 if (nla_len(encap)) {
1173 OVS_NLERR(log, "Truncated 802.1Q header has non-zero encap attribute.");
1177 OVS_NLERR(log, "Encap attr is set for non-VLAN frame");
1182 err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1188 /* Create an exact match mask. We need to set to 0xff
1189 * all the 'match->mask' fields that have been touched
1190 * in 'match->key'. We cannot simply memset
1191 * 'match->mask', because padding bytes and fields not
1192 * specified in 'match->key' should be left to 0.
1193 * Instead, we use a stream of netlink attributes,
1194 * copied from 'key' and set to 0xff.
1195 * ovs_key_from_nlattrs() will take care of filling
1196 * 'match->mask' appropriately.
1198 newmask = kmemdup(nla_key,
1199 nla_total_size(nla_len(nla_key)),
1204 mask_set_nlattr(newmask, 0xff);
1206 /* The userspace does not send tunnel attributes that
1207 * are 0, but we should not wildcard them nonetheless.
1209 if (match->key->tun_key.u.ipv4.dst)
1210 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1216 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1220 /* Always match on tci. */
1221 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
1223 if (mask_attrs & 1ULL << OVS_KEY_ATTR_ENCAP) {
1224 __be16 eth_type = 0;
1228 OVS_NLERR(log, "Encap mask attribute is set for non-VLAN frame.");
1233 mask_attrs &= ~(1ULL << OVS_KEY_ATTR_ENCAP);
1234 if (a[OVS_KEY_ATTR_ETHERTYPE])
1235 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1237 if (eth_type == htons(0xffff)) {
1238 mask_attrs &= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE);
1239 encap = a[OVS_KEY_ATTR_ENCAP];
1240 err = parse_flow_mask_nlattrs(encap, a,
1245 OVS_NLERR(log, "VLAN frames must have an exact match on the TPID (mask=%x).",
1251 if (a[OVS_KEY_ATTR_VLAN])
1252 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1254 if (!(tci & htons(VLAN_TAG_PRESENT))) {
1255 OVS_NLERR(log, "VLAN tag present bit must have an exact match (tci_mask=%x).",
1262 err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1268 if (!match_validate(match, key_attrs, mask_attrs, log))
1276 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1283 len = nla_len(attr);
1284 if (len < 1 || len > MAX_UFID_LENGTH) {
1285 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1286 nla_len(attr), MAX_UFID_LENGTH);
1293 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1294 * or false otherwise.
1296 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1299 sfid->ufid_len = get_ufid_len(attr, log);
1301 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1303 return sfid->ufid_len;
1306 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1307 const struct sw_flow_key *key, bool log)
1309 struct sw_flow_key *new_key;
1311 if (ovs_nla_get_ufid(sfid, ufid, log))
1314 /* If UFID was not provided, use unmasked key. */
1315 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1318 memcpy(new_key, key, sizeof(*key));
1319 sfid->unmasked_key = new_key;
1324 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1326 return attr ? nla_get_u32(attr) : 0;
1330 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1331 * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1332 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1334 * @log: Boolean to allow kernel error logging. Normally true, but when
1335 * probing for feature compatibility this should be passed in as false to
1336 * suppress unnecessary error logging.
1338 * This parses a series of Netlink attributes that form a flow key, which must
1339 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1340 * get the metadata, that is, the parts of the flow key that cannot be
1341 * extracted from the packet itself.
1344 int ovs_nla_get_flow_metadata(struct net *net, const struct nlattr *attr,
1345 struct sw_flow_key *key,
1348 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1349 struct sw_flow_match match;
1353 err = parse_flow_nlattrs(attr, a, &attrs, log);
1357 memset(&match, 0, sizeof(match));
1360 memset(key, 0, OVS_SW_FLOW_KEY_METADATA_SIZE);
1361 memset(&key->ct, 0, sizeof(key->ct));
1362 key->phy.in_port = DP_MAX_PORTS;
1364 return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1367 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1368 const struct sw_flow_key *output, bool is_mask,
1369 struct sk_buff *skb)
1371 struct ovs_key_ethernet *eth_key;
1372 struct nlattr *nla, *encap;
1374 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1375 goto nla_put_failure;
1377 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1378 goto nla_put_failure;
1380 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1381 goto nla_put_failure;
1383 if ((swkey->tun_key.u.ipv4.dst || is_mask)) {
1384 const void *opts = NULL;
1386 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1387 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1389 if (ipv4_tun_to_nlattr(skb, &output->tun_key, opts,
1390 swkey->tun_opts_len))
1391 goto nla_put_failure;
1394 if (swkey->phy.in_port == DP_MAX_PORTS) {
1395 if (is_mask && (output->phy.in_port == 0xffff))
1396 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1397 goto nla_put_failure;
1400 upper_u16 = !is_mask ? 0 : 0xffff;
1402 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1403 (upper_u16 << 16) | output->phy.in_port))
1404 goto nla_put_failure;
1407 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1408 goto nla_put_failure;
1410 if (ovs_ct_put_key(output, skb))
1411 goto nla_put_failure;
1413 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1415 goto nla_put_failure;
1417 eth_key = nla_data(nla);
1418 ether_addr_copy(eth_key->eth_src, output->eth.src);
1419 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1421 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1423 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
1424 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1425 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
1426 goto nla_put_failure;
1427 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1428 if (!swkey->eth.tci)
1433 if (swkey->eth.type == htons(ETH_P_802_2)) {
1435 * Ethertype 802.2 is represented in the netlink with omitted
1436 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1437 * 0xffff in the mask attribute. Ethertype can also
1440 if (is_mask && output->eth.type)
1441 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1443 goto nla_put_failure;
1447 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1448 goto nla_put_failure;
1450 if (swkey->eth.type == htons(ETH_P_IP)) {
1451 struct ovs_key_ipv4 *ipv4_key;
1453 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1455 goto nla_put_failure;
1456 ipv4_key = nla_data(nla);
1457 ipv4_key->ipv4_src = output->ipv4.addr.src;
1458 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1459 ipv4_key->ipv4_proto = output->ip.proto;
1460 ipv4_key->ipv4_tos = output->ip.tos;
1461 ipv4_key->ipv4_ttl = output->ip.ttl;
1462 ipv4_key->ipv4_frag = output->ip.frag;
1463 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1464 struct ovs_key_ipv6 *ipv6_key;
1466 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1468 goto nla_put_failure;
1469 ipv6_key = nla_data(nla);
1470 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1471 sizeof(ipv6_key->ipv6_src));
1472 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1473 sizeof(ipv6_key->ipv6_dst));
1474 ipv6_key->ipv6_label = output->ipv6.label;
1475 ipv6_key->ipv6_proto = output->ip.proto;
1476 ipv6_key->ipv6_tclass = output->ip.tos;
1477 ipv6_key->ipv6_hlimit = output->ip.ttl;
1478 ipv6_key->ipv6_frag = output->ip.frag;
1479 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1480 swkey->eth.type == htons(ETH_P_RARP)) {
1481 struct ovs_key_arp *arp_key;
1483 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1485 goto nla_put_failure;
1486 arp_key = nla_data(nla);
1487 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1488 arp_key->arp_sip = output->ipv4.addr.src;
1489 arp_key->arp_tip = output->ipv4.addr.dst;
1490 arp_key->arp_op = htons(output->ip.proto);
1491 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1492 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1493 } else if (eth_p_mpls(swkey->eth.type)) {
1494 struct ovs_key_mpls *mpls_key;
1496 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1498 goto nla_put_failure;
1499 mpls_key = nla_data(nla);
1500 mpls_key->mpls_lse = output->mpls.top_lse;
1503 if ((swkey->eth.type == htons(ETH_P_IP) ||
1504 swkey->eth.type == htons(ETH_P_IPV6)) &&
1505 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1507 if (swkey->ip.proto == IPPROTO_TCP) {
1508 struct ovs_key_tcp *tcp_key;
1510 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1512 goto nla_put_failure;
1513 tcp_key = nla_data(nla);
1514 tcp_key->tcp_src = output->tp.src;
1515 tcp_key->tcp_dst = output->tp.dst;
1516 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1518 goto nla_put_failure;
1519 } else if (swkey->ip.proto == IPPROTO_UDP) {
1520 struct ovs_key_udp *udp_key;
1522 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1524 goto nla_put_failure;
1525 udp_key = nla_data(nla);
1526 udp_key->udp_src = output->tp.src;
1527 udp_key->udp_dst = output->tp.dst;
1528 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1529 struct ovs_key_sctp *sctp_key;
1531 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1533 goto nla_put_failure;
1534 sctp_key = nla_data(nla);
1535 sctp_key->sctp_src = output->tp.src;
1536 sctp_key->sctp_dst = output->tp.dst;
1537 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1538 swkey->ip.proto == IPPROTO_ICMP) {
1539 struct ovs_key_icmp *icmp_key;
1541 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1543 goto nla_put_failure;
1544 icmp_key = nla_data(nla);
1545 icmp_key->icmp_type = ntohs(output->tp.src);
1546 icmp_key->icmp_code = ntohs(output->tp.dst);
1547 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1548 swkey->ip.proto == IPPROTO_ICMPV6) {
1549 struct ovs_key_icmpv6 *icmpv6_key;
1551 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1552 sizeof(*icmpv6_key));
1554 goto nla_put_failure;
1555 icmpv6_key = nla_data(nla);
1556 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1557 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1559 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1560 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1561 struct ovs_key_nd *nd_key;
1563 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1565 goto nla_put_failure;
1566 nd_key = nla_data(nla);
1567 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1568 sizeof(nd_key->nd_target));
1569 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1570 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1577 nla_nest_end(skb, encap);
1585 int ovs_nla_put_key(const struct sw_flow_key *swkey,
1586 const struct sw_flow_key *output, int attr, bool is_mask,
1587 struct sk_buff *skb)
1592 nla = nla_nest_start(skb, attr);
1595 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
1598 nla_nest_end(skb, nla);
1603 /* Called with ovs_mutex or RCU read lock. */
1604 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
1606 if (ovs_identifier_is_ufid(&flow->id))
1607 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
1610 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
1611 OVS_FLOW_ATTR_KEY, false, skb);
1614 /* Called with ovs_mutex or RCU read lock. */
1615 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
1617 return ovs_nla_put_key(&flow->key, &flow->key,
1618 OVS_FLOW_ATTR_KEY, false, skb);
1621 /* Called with ovs_mutex or RCU read lock. */
1622 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
1624 return ovs_nla_put_key(&flow->key, &flow->mask->key,
1625 OVS_FLOW_ATTR_MASK, true, skb);
1628 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1630 static struct sw_flow_actions *nla_alloc_flow_actions(int size, bool log)
1632 struct sw_flow_actions *sfa;
1634 if (size > MAX_ACTIONS_BUFSIZE) {
1635 OVS_NLERR(log, "Flow action size %u bytes exceeds max", size);
1636 return ERR_PTR(-EINVAL);
1639 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1641 return ERR_PTR(-ENOMEM);
1643 sfa->actions_len = 0;
1647 static void ovs_nla_free_set_action(const struct nlattr *a)
1649 const struct nlattr *ovs_key = nla_data(a);
1650 struct ovs_tunnel_info *ovs_tun;
1652 switch (nla_type(ovs_key)) {
1653 case OVS_KEY_ATTR_TUNNEL_INFO:
1654 ovs_tun = nla_data(ovs_key);
1655 ovs_dst_release((struct dst_entry *)ovs_tun->tun_dst);
1660 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1662 const struct nlattr *a;
1668 nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
1669 switch (nla_type(a)) {
1670 case OVS_ACTION_ATTR_SET:
1671 ovs_nla_free_set_action(a);
1673 case OVS_ACTION_ATTR_CT:
1674 ovs_ct_free_action(a);
1682 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
1684 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
1687 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1688 * The caller must hold rcu_read_lock for this to be sensible.
1690 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
1692 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
1695 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1696 int attr_len, bool log)
1699 struct sw_flow_actions *acts;
1701 int req_size = NLA_ALIGN(attr_len);
1702 int next_offset = offsetof(struct sw_flow_actions, actions) +
1703 (*sfa)->actions_len;
1705 if (req_size <= (ksize(*sfa) - next_offset))
1708 new_acts_size = ksize(*sfa) * 2;
1710 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1711 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1712 return ERR_PTR(-EMSGSIZE);
1713 new_acts_size = MAX_ACTIONS_BUFSIZE;
1716 acts = nla_alloc_flow_actions(new_acts_size, log);
1718 return (void *)acts;
1720 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1721 acts->actions_len = (*sfa)->actions_len;
1722 acts->orig_len = (*sfa)->orig_len;
1727 (*sfa)->actions_len += req_size;
1728 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1731 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1732 int attrtype, void *data, int len, bool log)
1736 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
1740 a->nla_type = attrtype;
1741 a->nla_len = nla_attr_size(len);
1744 memcpy(nla_data(a), data, len);
1745 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1750 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
1755 a = __add_action(sfa, attrtype, data, len, log);
1762 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1763 int attrtype, bool log)
1765 int used = (*sfa)->actions_len;
1768 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
1775 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1778 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1781 a->nla_len = sfa->actions_len - st_offset;
1784 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
1785 const struct sw_flow_key *key,
1786 int depth, struct sw_flow_actions **sfa,
1787 __be16 eth_type, __be16 vlan_tci, bool log);
1789 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
1790 const struct sw_flow_key *key, int depth,
1791 struct sw_flow_actions **sfa,
1792 __be16 eth_type, __be16 vlan_tci, bool log)
1794 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1795 const struct nlattr *probability, *actions;
1796 const struct nlattr *a;
1797 int rem, start, err, st_acts;
1799 memset(attrs, 0, sizeof(attrs));
1800 nla_for_each_nested(a, attr, rem) {
1801 int type = nla_type(a);
1802 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1809 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1810 if (!probability || nla_len(probability) != sizeof(u32))
1813 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1814 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1817 /* validation done, copy sample action. */
1818 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
1821 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1822 nla_data(probability), sizeof(u32), log);
1825 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS, log);
1829 err = __ovs_nla_copy_actions(net, actions, key, depth + 1, sfa,
1830 eth_type, vlan_tci, log);
1834 add_nested_action_end(*sfa, st_acts);
1835 add_nested_action_end(*sfa, start);
1840 void ovs_match_init(struct sw_flow_match *match,
1841 struct sw_flow_key *key,
1842 struct sw_flow_mask *mask)
1844 memset(match, 0, sizeof(*match));
1848 memset(key, 0, sizeof(*key));
1851 memset(&mask->key, 0, sizeof(mask->key));
1852 mask->range.start = mask->range.end = 0;
1856 static int validate_geneve_opts(struct sw_flow_key *key)
1858 struct geneve_opt *option;
1859 int opts_len = key->tun_opts_len;
1860 bool crit_opt = false;
1862 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
1863 while (opts_len > 0) {
1866 if (opts_len < sizeof(*option))
1869 len = sizeof(*option) + option->length * 4;
1873 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
1875 option = (struct geneve_opt *)((u8 *)option + len);
1879 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
1884 static int validate_and_copy_set_tun(const struct nlattr *attr,
1885 struct sw_flow_actions **sfa, bool log)
1887 struct sw_flow_match match;
1888 struct sw_flow_key key;
1889 struct metadata_dst *tun_dst;
1890 struct ip_tunnel_info *tun_info;
1891 struct ovs_tunnel_info *ovs_tun;
1893 int err = 0, start, opts_type;
1895 ovs_match_init(&match, &key, NULL);
1896 opts_type = ipv4_tun_from_nlattr(nla_data(attr), &match, false, log);
1900 if (key.tun_opts_len) {
1901 switch (opts_type) {
1902 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
1903 err = validate_geneve_opts(&key);
1907 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
1912 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
1916 tun_dst = metadata_dst_alloc(key.tun_opts_len, GFP_KERNEL);
1920 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
1921 sizeof(*ovs_tun), log);
1923 ovs_dst_release((struct dst_entry *)tun_dst);
1927 ovs_tun = nla_data(a);
1928 ovs_tun->tun_dst = tun_dst;
1930 tun_info = &tun_dst->u.tun_info;
1931 tun_info->mode = IP_TUNNEL_INFO_TX;
1932 tun_info->key = key.tun_key;
1934 /* We need to store the options in the action itself since
1935 * everything else will go away after flow setup. We can append
1936 * it to tun_info and then point there.
1938 ip_tunnel_info_opts_set(tun_info,
1939 TUN_METADATA_OPTS(&key, key.tun_opts_len),
1941 add_nested_action_end(*sfa, start);
1946 /* Return false if there are any non-masked bits set.
1947 * Mask follows data immediately, before any netlink padding.
1949 static bool validate_masked(u8 *data, int len)
1951 u8 *mask = data + len;
1954 if (*data++ & ~*mask++)
1960 static int validate_set(const struct nlattr *a,
1961 const struct sw_flow_key *flow_key,
1962 struct sw_flow_actions **sfa,
1963 bool *skip_copy, __be16 eth_type, bool masked, bool log)
1965 const struct nlattr *ovs_key = nla_data(a);
1966 int key_type = nla_type(ovs_key);
1969 /* There can be only one key in a action */
1970 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1973 key_len = nla_len(ovs_key);
1977 if (key_type > OVS_KEY_ATTR_MAX ||
1978 !check_attr_len(key_len, ovs_key_lens[key_type].len))
1981 if (masked && !validate_masked(nla_data(ovs_key), key_len))
1985 const struct ovs_key_ipv4 *ipv4_key;
1986 const struct ovs_key_ipv6 *ipv6_key;
1989 case OVS_KEY_ATTR_PRIORITY:
1990 case OVS_KEY_ATTR_SKB_MARK:
1991 case OVS_KEY_ATTR_CT_MARK:
1992 case OVS_KEY_ATTR_CT_LABELS:
1993 case OVS_KEY_ATTR_ETHERNET:
1996 case OVS_KEY_ATTR_TUNNEL:
1997 if (eth_p_mpls(eth_type))
2001 return -EINVAL; /* Masked tunnel set not supported. */
2004 err = validate_and_copy_set_tun(a, sfa, log);
2009 case OVS_KEY_ATTR_IPV4:
2010 if (eth_type != htons(ETH_P_IP))
2013 ipv4_key = nla_data(ovs_key);
2016 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2018 /* Non-writeable fields. */
2019 if (mask->ipv4_proto || mask->ipv4_frag)
2022 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2025 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2030 case OVS_KEY_ATTR_IPV6:
2031 if (eth_type != htons(ETH_P_IPV6))
2034 ipv6_key = nla_data(ovs_key);
2037 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2039 /* Non-writeable fields. */
2040 if (mask->ipv6_proto || mask->ipv6_frag)
2043 /* Invalid bits in the flow label mask? */
2044 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2047 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2050 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2053 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2058 case OVS_KEY_ATTR_TCP:
2059 if ((eth_type != htons(ETH_P_IP) &&
2060 eth_type != htons(ETH_P_IPV6)) ||
2061 flow_key->ip.proto != IPPROTO_TCP)
2066 case OVS_KEY_ATTR_UDP:
2067 if ((eth_type != htons(ETH_P_IP) &&
2068 eth_type != htons(ETH_P_IPV6)) ||
2069 flow_key->ip.proto != IPPROTO_UDP)
2074 case OVS_KEY_ATTR_MPLS:
2075 if (!eth_p_mpls(eth_type))
2079 case OVS_KEY_ATTR_SCTP:
2080 if ((eth_type != htons(ETH_P_IP) &&
2081 eth_type != htons(ETH_P_IPV6)) ||
2082 flow_key->ip.proto != IPPROTO_SCTP)
2091 /* Convert non-masked non-tunnel set actions to masked set actions. */
2092 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2093 int start, len = key_len * 2;
2098 start = add_nested_action_start(sfa,
2099 OVS_ACTION_ATTR_SET_TO_MASKED,
2104 at = __add_action(sfa, key_type, NULL, len, log);
2108 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2109 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
2110 /* Clear non-writeable bits from otherwise writeable fields. */
2111 if (key_type == OVS_KEY_ATTR_IPV6) {
2112 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2114 mask->ipv6_label &= htonl(0x000FFFFF);
2116 add_nested_action_end(*sfa, start);
2122 static int validate_userspace(const struct nlattr *attr)
2124 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2125 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2126 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2127 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2129 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2132 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
2133 attr, userspace_policy);
2137 if (!a[OVS_USERSPACE_ATTR_PID] ||
2138 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2144 static int copy_action(const struct nlattr *from,
2145 struct sw_flow_actions **sfa, bool log)
2147 int totlen = NLA_ALIGN(from->nla_len);
2150 to = reserve_sfa_size(sfa, from->nla_len, log);
2154 memcpy(to, from, totlen);
2158 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2159 const struct sw_flow_key *key,
2160 int depth, struct sw_flow_actions **sfa,
2161 __be16 eth_type, __be16 vlan_tci, bool log)
2163 const struct nlattr *a;
2166 if (depth >= SAMPLE_ACTION_DEPTH)
2169 nla_for_each_nested(a, attr, rem) {
2170 /* Expected argument lengths, (u32)-1 for variable length. */
2171 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2172 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2173 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2174 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2175 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2176 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2177 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2178 [OVS_ACTION_ATTR_POP_VLAN] = 0,
2179 [OVS_ACTION_ATTR_SET] = (u32)-1,
2180 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2181 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2182 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2183 [OVS_ACTION_ATTR_CT] = (u32)-1,
2185 const struct ovs_action_push_vlan *vlan;
2186 int type = nla_type(a);
2189 if (type > OVS_ACTION_ATTR_MAX ||
2190 (action_lens[type] != nla_len(a) &&
2191 action_lens[type] != (u32)-1))
2196 case OVS_ACTION_ATTR_UNSPEC:
2199 case OVS_ACTION_ATTR_USERSPACE:
2200 err = validate_userspace(a);
2205 case OVS_ACTION_ATTR_OUTPUT:
2206 if (nla_get_u32(a) >= DP_MAX_PORTS)
2210 case OVS_ACTION_ATTR_HASH: {
2211 const struct ovs_action_hash *act_hash = nla_data(a);
2213 switch (act_hash->hash_alg) {
2214 case OVS_HASH_ALG_L4:
2223 case OVS_ACTION_ATTR_POP_VLAN:
2224 vlan_tci = htons(0);
2227 case OVS_ACTION_ATTR_PUSH_VLAN:
2229 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
2231 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2233 vlan_tci = vlan->vlan_tci;
2236 case OVS_ACTION_ATTR_RECIRC:
2239 case OVS_ACTION_ATTR_PUSH_MPLS: {
2240 const struct ovs_action_push_mpls *mpls = nla_data(a);
2242 if (!eth_p_mpls(mpls->mpls_ethertype))
2244 /* Prohibit push MPLS other than to a white list
2245 * for packets that have a known tag order.
2247 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2248 (eth_type != htons(ETH_P_IP) &&
2249 eth_type != htons(ETH_P_IPV6) &&
2250 eth_type != htons(ETH_P_ARP) &&
2251 eth_type != htons(ETH_P_RARP) &&
2252 !eth_p_mpls(eth_type)))
2254 eth_type = mpls->mpls_ethertype;
2258 case OVS_ACTION_ATTR_POP_MPLS:
2259 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2260 !eth_p_mpls(eth_type))
2263 /* Disallow subsequent L2.5+ set and mpls_pop actions
2264 * as there is no check here to ensure that the new
2265 * eth_type is valid and thus set actions could
2266 * write off the end of the packet or otherwise
2269 * Support for these actions is planned using packet
2272 eth_type = htons(0);
2275 case OVS_ACTION_ATTR_SET:
2276 err = validate_set(a, key, sfa,
2277 &skip_copy, eth_type, false, log);
2282 case OVS_ACTION_ATTR_SET_MASKED:
2283 err = validate_set(a, key, sfa,
2284 &skip_copy, eth_type, true, log);
2289 case OVS_ACTION_ATTR_SAMPLE:
2290 err = validate_and_copy_sample(net, a, key, depth, sfa,
2291 eth_type, vlan_tci, log);
2297 case OVS_ACTION_ATTR_CT:
2298 err = ovs_ct_copy_action(net, a, key, sfa, log);
2305 OVS_NLERR(log, "Unknown Action type %d", type);
2309 err = copy_action(a, sfa, log);
2321 /* 'key' must be the masked key. */
2322 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2323 const struct sw_flow_key *key,
2324 struct sw_flow_actions **sfa, bool log)
2328 *sfa = nla_alloc_flow_actions(nla_len(attr), log);
2330 return PTR_ERR(*sfa);
2332 (*sfa)->orig_len = nla_len(attr);
2333 err = __ovs_nla_copy_actions(net, attr, key, 0, sfa, key->eth.type,
2336 ovs_nla_free_flow_actions(*sfa);
2341 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
2343 const struct nlattr *a;
2344 struct nlattr *start;
2347 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
2351 nla_for_each_nested(a, attr, rem) {
2352 int type = nla_type(a);
2353 struct nlattr *st_sample;
2356 case OVS_SAMPLE_ATTR_PROBABILITY:
2357 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
2358 sizeof(u32), nla_data(a)))
2361 case OVS_SAMPLE_ATTR_ACTIONS:
2362 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
2365 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
2368 nla_nest_end(skb, st_sample);
2373 nla_nest_end(skb, start);
2377 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
2379 const struct nlattr *ovs_key = nla_data(a);
2380 int key_type = nla_type(ovs_key);
2381 struct nlattr *start;
2385 case OVS_KEY_ATTR_TUNNEL_INFO: {
2386 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
2387 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
2389 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2393 err = ipv4_tun_to_nlattr(skb, &tun_info->key,
2394 tun_info->options_len ?
2395 ip_tunnel_info_opts(tun_info) : NULL,
2396 tun_info->options_len);
2399 nla_nest_end(skb, start);
2403 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
2411 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
2412 struct sk_buff *skb)
2414 const struct nlattr *ovs_key = nla_data(a);
2415 size_t key_len = nla_len(ovs_key) / 2;
2417 /* Revert the conversion we did from a non-masked set action to
2418 * masked set action.
2420 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a) - key_len, ovs_key))
2426 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
2428 const struct nlattr *a;
2431 nla_for_each_attr(a, attr, len, rem) {
2432 int type = nla_type(a);
2435 case OVS_ACTION_ATTR_SET:
2436 err = set_action_to_attr(a, skb);
2441 case OVS_ACTION_ATTR_SET_TO_MASKED:
2442 err = masked_set_action_to_set_action_attr(a, skb);
2447 case OVS_ACTION_ATTR_SAMPLE:
2448 err = sample_action_to_attr(a, skb);
2453 case OVS_ACTION_ATTR_CT:
2454 err = ovs_ct_action_to_attr(nla_data(a), skb);
2460 if (nla_put(skb, type, nla_len(a), nla_data(a)))