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(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
268 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_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
},
328 [OVS_TUNNEL_KEY_ATTR_IPV6_SRC
] = { .len
= sizeof(struct in6_addr
) },
329 [OVS_TUNNEL_KEY_ATTR_IPV6_DST
] = { .len
= sizeof(struct in6_addr
) },
332 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
333 static const struct ovs_len_tbl ovs_key_lens
[OVS_KEY_ATTR_MAX
+ 1] = {
334 [OVS_KEY_ATTR_ENCAP
] = { .len
= OVS_ATTR_NESTED
},
335 [OVS_KEY_ATTR_PRIORITY
] = { .len
= sizeof(u32
) },
336 [OVS_KEY_ATTR_IN_PORT
] = { .len
= sizeof(u32
) },
337 [OVS_KEY_ATTR_SKB_MARK
] = { .len
= sizeof(u32
) },
338 [OVS_KEY_ATTR_ETHERNET
] = { .len
= sizeof(struct ovs_key_ethernet
) },
339 [OVS_KEY_ATTR_VLAN
] = { .len
= sizeof(__be16
) },
340 [OVS_KEY_ATTR_ETHERTYPE
] = { .len
= sizeof(__be16
) },
341 [OVS_KEY_ATTR_IPV4
] = { .len
= sizeof(struct ovs_key_ipv4
) },
342 [OVS_KEY_ATTR_IPV6
] = { .len
= sizeof(struct ovs_key_ipv6
) },
343 [OVS_KEY_ATTR_TCP
] = { .len
= sizeof(struct ovs_key_tcp
) },
344 [OVS_KEY_ATTR_TCP_FLAGS
] = { .len
= sizeof(__be16
) },
345 [OVS_KEY_ATTR_UDP
] = { .len
= sizeof(struct ovs_key_udp
) },
346 [OVS_KEY_ATTR_SCTP
] = { .len
= sizeof(struct ovs_key_sctp
) },
347 [OVS_KEY_ATTR_ICMP
] = { .len
= sizeof(struct ovs_key_icmp
) },
348 [OVS_KEY_ATTR_ICMPV6
] = { .len
= sizeof(struct ovs_key_icmpv6
) },
349 [OVS_KEY_ATTR_ARP
] = { .len
= sizeof(struct ovs_key_arp
) },
350 [OVS_KEY_ATTR_ND
] = { .len
= sizeof(struct ovs_key_nd
) },
351 [OVS_KEY_ATTR_RECIRC_ID
] = { .len
= sizeof(u32
) },
352 [OVS_KEY_ATTR_DP_HASH
] = { .len
= sizeof(u32
) },
353 [OVS_KEY_ATTR_TUNNEL
] = { .len
= OVS_ATTR_NESTED
,
354 .next
= ovs_tunnel_key_lens
, },
355 [OVS_KEY_ATTR_MPLS
] = { .len
= sizeof(struct ovs_key_mpls
) },
356 [OVS_KEY_ATTR_CT_STATE
] = { .len
= sizeof(u32
) },
357 [OVS_KEY_ATTR_CT_ZONE
] = { .len
= sizeof(u16
) },
358 [OVS_KEY_ATTR_CT_MARK
] = { .len
= sizeof(u32
) },
359 [OVS_KEY_ATTR_CT_LABELS
] = { .len
= sizeof(struct ovs_key_ct_labels
) },
362 static bool check_attr_len(unsigned int attr_len
, unsigned int expected_len
)
364 return expected_len
== attr_len
||
365 expected_len
== OVS_ATTR_NESTED
||
366 expected_len
== OVS_ATTR_VARIABLE
;
369 static bool is_all_zero(const u8
*fp
, size_t size
)
376 for (i
= 0; i
< size
; i
++)
383 static int __parse_flow_nlattrs(const struct nlattr
*attr
,
384 const struct nlattr
*a
[],
385 u64
*attrsp
, bool log
, bool nz
)
387 const struct nlattr
*nla
;
392 nla_for_each_nested(nla
, attr
, rem
) {
393 u16 type
= nla_type(nla
);
396 if (type
> OVS_KEY_ATTR_MAX
) {
397 OVS_NLERR(log
, "Key type %d is out of range max %d",
398 type
, OVS_KEY_ATTR_MAX
);
402 if (attrs
& (1ULL << type
)) {
403 OVS_NLERR(log
, "Duplicate key (type %d).", type
);
407 expected_len
= ovs_key_lens
[type
].len
;
408 if (!check_attr_len(nla_len(nla
), expected_len
)) {
409 OVS_NLERR(log
, "Key %d has unexpected len %d expected %d",
410 type
, nla_len(nla
), expected_len
);
414 if (!nz
|| !is_all_zero(nla_data(nla
), expected_len
)) {
415 attrs
|= 1ULL << type
;
420 OVS_NLERR(log
, "Message has %d unknown bytes.", rem
);
428 static int parse_flow_mask_nlattrs(const struct nlattr
*attr
,
429 const struct nlattr
*a
[], u64
*attrsp
,
432 return __parse_flow_nlattrs(attr
, a
, attrsp
, log
, true);
435 static int parse_flow_nlattrs(const struct nlattr
*attr
,
436 const struct nlattr
*a
[], u64
*attrsp
,
439 return __parse_flow_nlattrs(attr
, a
, attrsp
, log
, false);
442 static int genev_tun_opt_from_nlattr(const struct nlattr
*a
,
443 struct sw_flow_match
*match
, bool is_mask
,
446 unsigned long opt_key_offset
;
448 if (nla_len(a
) > sizeof(match
->key
->tun_opts
)) {
449 OVS_NLERR(log
, "Geneve option length err (len %d, max %zu).",
450 nla_len(a
), sizeof(match
->key
->tun_opts
));
454 if (nla_len(a
) % 4 != 0) {
455 OVS_NLERR(log
, "Geneve opt len %d is not a multiple of 4.",
460 /* We need to record the length of the options passed
461 * down, otherwise packets with the same format but
462 * additional options will be silently matched.
465 SW_FLOW_KEY_PUT(match
, tun_opts_len
, nla_len(a
),
468 /* This is somewhat unusual because it looks at
469 * both the key and mask while parsing the
470 * attributes (and by extension assumes the key
471 * is parsed first). Normally, we would verify
472 * that each is the correct length and that the
473 * attributes line up in the validate function.
474 * However, that is difficult because this is
475 * variable length and we won't have the
478 if (match
->key
->tun_opts_len
!= nla_len(a
)) {
479 OVS_NLERR(log
, "Geneve option len %d != mask len %d",
480 match
->key
->tun_opts_len
, nla_len(a
));
484 SW_FLOW_KEY_PUT(match
, tun_opts_len
, 0xff, true);
487 opt_key_offset
= TUN_METADATA_OFFSET(nla_len(a
));
488 SW_FLOW_KEY_MEMCPY_OFFSET(match
, opt_key_offset
, nla_data(a
),
489 nla_len(a
), is_mask
);
493 static int vxlan_tun_opt_from_nlattr(const struct nlattr
*attr
,
494 struct sw_flow_match
*match
, bool is_mask
,
499 unsigned long opt_key_offset
;
500 struct vxlan_metadata opts
;
502 BUILD_BUG_ON(sizeof(opts
) > sizeof(match
->key
->tun_opts
));
504 memset(&opts
, 0, sizeof(opts
));
505 nla_for_each_nested(a
, attr
, rem
) {
506 int type
= nla_type(a
);
508 if (type
> OVS_VXLAN_EXT_MAX
) {
509 OVS_NLERR(log
, "VXLAN extension %d out of range max %d",
510 type
, OVS_VXLAN_EXT_MAX
);
514 if (!check_attr_len(nla_len(a
),
515 ovs_vxlan_ext_key_lens
[type
].len
)) {
516 OVS_NLERR(log
, "VXLAN extension %d has unexpected len %d expected %d",
518 ovs_vxlan_ext_key_lens
[type
].len
);
523 case OVS_VXLAN_EXT_GBP
:
524 opts
.gbp
= nla_get_u32(a
);
527 OVS_NLERR(log
, "Unknown VXLAN extension attribute %d",
533 OVS_NLERR(log
, "VXLAN extension message has %d unknown bytes.",
539 SW_FLOW_KEY_PUT(match
, tun_opts_len
, sizeof(opts
), false);
541 SW_FLOW_KEY_PUT(match
, tun_opts_len
, 0xff, true);
543 opt_key_offset
= TUN_METADATA_OFFSET(sizeof(opts
));
544 SW_FLOW_KEY_MEMCPY_OFFSET(match
, opt_key_offset
, &opts
, sizeof(opts
),
549 static int ip_tun_from_nlattr(const struct nlattr
*attr
,
550 struct sw_flow_match
*match
, bool is_mask
,
553 bool ttl
= false, ipv4
= false, ipv6
= false;
554 __be16 tun_flags
= 0;
559 nla_for_each_nested(a
, attr
, rem
) {
560 int type
= nla_type(a
);
563 if (type
> OVS_TUNNEL_KEY_ATTR_MAX
) {
564 OVS_NLERR(log
, "Tunnel attr %d out of range max %d",
565 type
, OVS_TUNNEL_KEY_ATTR_MAX
);
569 if (!check_attr_len(nla_len(a
),
570 ovs_tunnel_key_lens
[type
].len
)) {
571 OVS_NLERR(log
, "Tunnel attr %d has unexpected len %d expected %d",
572 type
, nla_len(a
), ovs_tunnel_key_lens
[type
].len
);
577 case OVS_TUNNEL_KEY_ATTR_ID
:
578 SW_FLOW_KEY_PUT(match
, tun_key
.tun_id
,
579 nla_get_be64(a
), is_mask
);
580 tun_flags
|= TUNNEL_KEY
;
582 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC
:
583 SW_FLOW_KEY_PUT(match
, tun_key
.u
.ipv4
.src
,
584 nla_get_in_addr(a
), is_mask
);
587 case OVS_TUNNEL_KEY_ATTR_IPV4_DST
:
588 SW_FLOW_KEY_PUT(match
, tun_key
.u
.ipv4
.dst
,
589 nla_get_in_addr(a
), is_mask
);
592 case OVS_TUNNEL_KEY_ATTR_IPV6_SRC
:
593 SW_FLOW_KEY_PUT(match
, tun_key
.u
.ipv6
.dst
,
594 nla_get_in6_addr(a
), is_mask
);
597 case OVS_TUNNEL_KEY_ATTR_IPV6_DST
:
598 SW_FLOW_KEY_PUT(match
, tun_key
.u
.ipv6
.dst
,
599 nla_get_in6_addr(a
), is_mask
);
602 case OVS_TUNNEL_KEY_ATTR_TOS
:
603 SW_FLOW_KEY_PUT(match
, tun_key
.tos
,
604 nla_get_u8(a
), is_mask
);
606 case OVS_TUNNEL_KEY_ATTR_TTL
:
607 SW_FLOW_KEY_PUT(match
, tun_key
.ttl
,
608 nla_get_u8(a
), is_mask
);
611 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT
:
612 tun_flags
|= TUNNEL_DONT_FRAGMENT
;
614 case OVS_TUNNEL_KEY_ATTR_CSUM
:
615 tun_flags
|= TUNNEL_CSUM
;
617 case OVS_TUNNEL_KEY_ATTR_TP_SRC
:
618 SW_FLOW_KEY_PUT(match
, tun_key
.tp_src
,
619 nla_get_be16(a
), is_mask
);
621 case OVS_TUNNEL_KEY_ATTR_TP_DST
:
622 SW_FLOW_KEY_PUT(match
, tun_key
.tp_dst
,
623 nla_get_be16(a
), is_mask
);
625 case OVS_TUNNEL_KEY_ATTR_OAM
:
626 tun_flags
|= TUNNEL_OAM
;
628 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS
:
630 OVS_NLERR(log
, "Multiple metadata blocks provided");
634 err
= genev_tun_opt_from_nlattr(a
, match
, is_mask
, log
);
638 tun_flags
|= TUNNEL_GENEVE_OPT
;
641 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS
:
643 OVS_NLERR(log
, "Multiple metadata blocks provided");
647 err
= vxlan_tun_opt_from_nlattr(a
, match
, is_mask
, log
);
651 tun_flags
|= TUNNEL_VXLAN_OPT
;
655 OVS_NLERR(log
, "Unknown IP tunnel attribute %d",
661 SW_FLOW_KEY_PUT(match
, tun_key
.tun_flags
, tun_flags
, is_mask
);
663 SW_FLOW_KEY_MEMSET_FIELD(match
, tun_proto
, 0xff, true);
665 SW_FLOW_KEY_PUT(match
, tun_proto
, ipv6
? AF_INET6
: AF_INET
,
669 OVS_NLERR(log
, "IP tunnel attribute has %d unknown bytes.",
675 OVS_NLERR(log
, "Mixed IPv4 and IPv6 tunnel attributes");
680 if (!ipv4
&& !ipv6
) {
681 OVS_NLERR(log
, "IP tunnel dst address not specified");
684 if (ipv4
&& !match
->key
->tun_key
.u
.ipv4
.dst
) {
685 OVS_NLERR(log
, "IPv4 tunnel dst address is zero");
688 if (ipv6
&& ipv6_addr_any(&match
->key
->tun_key
.u
.ipv6
.dst
)) {
689 OVS_NLERR(log
, "IPv6 tunnel dst address is zero");
694 OVS_NLERR(log
, "IP tunnel TTL not specified.");
702 static int vxlan_opt_to_nlattr(struct sk_buff
*skb
,
703 const void *tun_opts
, int swkey_tun_opts_len
)
705 const struct vxlan_metadata
*opts
= tun_opts
;
708 nla
= nla_nest_start(skb
, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS
);
712 if (nla_put_u32(skb
, OVS_VXLAN_EXT_GBP
, opts
->gbp
) < 0)
715 nla_nest_end(skb
, nla
);
719 static int __ip_tun_to_nlattr(struct sk_buff
*skb
,
720 const struct ip_tunnel_key
*output
,
721 const void *tun_opts
, int swkey_tun_opts_len
,
722 unsigned short tun_proto
)
724 if (output
->tun_flags
& TUNNEL_KEY
&&
725 nla_put_be64(skb
, OVS_TUNNEL_KEY_ATTR_ID
, output
->tun_id
))
729 if (output
->u
.ipv4
.src
&&
730 nla_put_in_addr(skb
, OVS_TUNNEL_KEY_ATTR_IPV4_SRC
,
733 if (output
->u
.ipv4
.dst
&&
734 nla_put_in_addr(skb
, OVS_TUNNEL_KEY_ATTR_IPV4_DST
,
739 if (!ipv6_addr_any(&output
->u
.ipv6
.src
) &&
740 nla_put_in6_addr(skb
, OVS_TUNNEL_KEY_ATTR_IPV6_SRC
,
741 &output
->u
.ipv6
.src
))
743 if (!ipv6_addr_any(&output
->u
.ipv6
.dst
) &&
744 nla_put_in6_addr(skb
, OVS_TUNNEL_KEY_ATTR_IPV6_DST
,
745 &output
->u
.ipv6
.dst
))
750 nla_put_u8(skb
, OVS_TUNNEL_KEY_ATTR_TOS
, output
->tos
))
752 if (nla_put_u8(skb
, OVS_TUNNEL_KEY_ATTR_TTL
, output
->ttl
))
754 if ((output
->tun_flags
& TUNNEL_DONT_FRAGMENT
) &&
755 nla_put_flag(skb
, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT
))
757 if ((output
->tun_flags
& TUNNEL_CSUM
) &&
758 nla_put_flag(skb
, OVS_TUNNEL_KEY_ATTR_CSUM
))
760 if (output
->tp_src
&&
761 nla_put_be16(skb
, OVS_TUNNEL_KEY_ATTR_TP_SRC
, output
->tp_src
))
763 if (output
->tp_dst
&&
764 nla_put_be16(skb
, OVS_TUNNEL_KEY_ATTR_TP_DST
, output
->tp_dst
))
766 if ((output
->tun_flags
& TUNNEL_OAM
) &&
767 nla_put_flag(skb
, OVS_TUNNEL_KEY_ATTR_OAM
))
769 if (swkey_tun_opts_len
) {
770 if (output
->tun_flags
& TUNNEL_GENEVE_OPT
&&
771 nla_put(skb
, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS
,
772 swkey_tun_opts_len
, tun_opts
))
774 else if (output
->tun_flags
& TUNNEL_VXLAN_OPT
&&
775 vxlan_opt_to_nlattr(skb
, tun_opts
, swkey_tun_opts_len
))
782 static int ip_tun_to_nlattr(struct sk_buff
*skb
,
783 const struct ip_tunnel_key
*output
,
784 const void *tun_opts
, int swkey_tun_opts_len
,
785 unsigned short tun_proto
)
790 nla
= nla_nest_start(skb
, OVS_KEY_ATTR_TUNNEL
);
794 err
= __ip_tun_to_nlattr(skb
, output
, tun_opts
, swkey_tun_opts_len
,
799 nla_nest_end(skb
, nla
);
803 int ovs_nla_put_tunnel_info(struct sk_buff
*skb
,
804 struct ip_tunnel_info
*tun_info
)
806 return __ip_tun_to_nlattr(skb
, &tun_info
->key
,
807 ip_tunnel_info_opts(tun_info
),
808 tun_info
->options_len
,
809 ip_tunnel_info_af(tun_info
));
812 static int metadata_from_nlattrs(struct net
*net
, struct sw_flow_match
*match
,
813 u64
*attrs
, const struct nlattr
**a
,
814 bool is_mask
, bool log
)
816 if (*attrs
& (1ULL << OVS_KEY_ATTR_DP_HASH
)) {
817 u32 hash_val
= nla_get_u32(a
[OVS_KEY_ATTR_DP_HASH
]);
819 SW_FLOW_KEY_PUT(match
, ovs_flow_hash
, hash_val
, is_mask
);
820 *attrs
&= ~(1ULL << OVS_KEY_ATTR_DP_HASH
);
823 if (*attrs
& (1ULL << OVS_KEY_ATTR_RECIRC_ID
)) {
824 u32 recirc_id
= nla_get_u32(a
[OVS_KEY_ATTR_RECIRC_ID
]);
826 SW_FLOW_KEY_PUT(match
, recirc_id
, recirc_id
, is_mask
);
827 *attrs
&= ~(1ULL << OVS_KEY_ATTR_RECIRC_ID
);
830 if (*attrs
& (1ULL << OVS_KEY_ATTR_PRIORITY
)) {
831 SW_FLOW_KEY_PUT(match
, phy
.priority
,
832 nla_get_u32(a
[OVS_KEY_ATTR_PRIORITY
]), is_mask
);
833 *attrs
&= ~(1ULL << OVS_KEY_ATTR_PRIORITY
);
836 if (*attrs
& (1ULL << OVS_KEY_ATTR_IN_PORT
)) {
837 u32 in_port
= nla_get_u32(a
[OVS_KEY_ATTR_IN_PORT
]);
840 in_port
= 0xffffffff; /* Always exact match in_port. */
841 } else if (in_port
>= DP_MAX_PORTS
) {
842 OVS_NLERR(log
, "Port %d exceeds max allowable %d",
843 in_port
, DP_MAX_PORTS
);
847 SW_FLOW_KEY_PUT(match
, phy
.in_port
, in_port
, is_mask
);
848 *attrs
&= ~(1ULL << OVS_KEY_ATTR_IN_PORT
);
849 } else if (!is_mask
) {
850 SW_FLOW_KEY_PUT(match
, phy
.in_port
, DP_MAX_PORTS
, is_mask
);
853 if (*attrs
& (1ULL << OVS_KEY_ATTR_SKB_MARK
)) {
854 uint32_t mark
= nla_get_u32(a
[OVS_KEY_ATTR_SKB_MARK
]);
856 SW_FLOW_KEY_PUT(match
, phy
.skb_mark
, mark
, is_mask
);
857 *attrs
&= ~(1ULL << OVS_KEY_ATTR_SKB_MARK
);
859 if (*attrs
& (1ULL << OVS_KEY_ATTR_TUNNEL
)) {
860 if (ip_tun_from_nlattr(a
[OVS_KEY_ATTR_TUNNEL
], match
,
863 *attrs
&= ~(1ULL << OVS_KEY_ATTR_TUNNEL
);
866 if (*attrs
& (1 << OVS_KEY_ATTR_CT_STATE
) &&
867 ovs_ct_verify(net
, OVS_KEY_ATTR_CT_STATE
)) {
868 u32 ct_state
= nla_get_u32(a
[OVS_KEY_ATTR_CT_STATE
]);
870 if (ct_state
& ~CT_SUPPORTED_MASK
) {
871 OVS_NLERR(log
, "ct_state flags %08x unsupported",
876 SW_FLOW_KEY_PUT(match
, ct
.state
, ct_state
, is_mask
);
877 *attrs
&= ~(1ULL << OVS_KEY_ATTR_CT_STATE
);
879 if (*attrs
& (1 << OVS_KEY_ATTR_CT_ZONE
) &&
880 ovs_ct_verify(net
, OVS_KEY_ATTR_CT_ZONE
)) {
881 u16 ct_zone
= nla_get_u16(a
[OVS_KEY_ATTR_CT_ZONE
]);
883 SW_FLOW_KEY_PUT(match
, ct
.zone
, ct_zone
, is_mask
);
884 *attrs
&= ~(1ULL << OVS_KEY_ATTR_CT_ZONE
);
886 if (*attrs
& (1 << OVS_KEY_ATTR_CT_MARK
) &&
887 ovs_ct_verify(net
, OVS_KEY_ATTR_CT_MARK
)) {
888 u32 mark
= nla_get_u32(a
[OVS_KEY_ATTR_CT_MARK
]);
890 SW_FLOW_KEY_PUT(match
, ct
.mark
, mark
, is_mask
);
891 *attrs
&= ~(1ULL << OVS_KEY_ATTR_CT_MARK
);
893 if (*attrs
& (1 << OVS_KEY_ATTR_CT_LABELS
) &&
894 ovs_ct_verify(net
, OVS_KEY_ATTR_CT_LABELS
)) {
895 const struct ovs_key_ct_labels
*cl
;
897 cl
= nla_data(a
[OVS_KEY_ATTR_CT_LABELS
]);
898 SW_FLOW_KEY_MEMCPY(match
, ct
.labels
, cl
->ct_labels
,
899 sizeof(*cl
), is_mask
);
900 *attrs
&= ~(1ULL << OVS_KEY_ATTR_CT_LABELS
);
905 static int ovs_key_from_nlattrs(struct net
*net
, struct sw_flow_match
*match
,
906 u64 attrs
, const struct nlattr
**a
,
907 bool is_mask
, bool log
)
911 err
= metadata_from_nlattrs(net
, match
, &attrs
, a
, is_mask
, log
);
915 if (attrs
& (1ULL << OVS_KEY_ATTR_ETHERNET
)) {
916 const struct ovs_key_ethernet
*eth_key
;
918 eth_key
= nla_data(a
[OVS_KEY_ATTR_ETHERNET
]);
919 SW_FLOW_KEY_MEMCPY(match
, eth
.src
,
920 eth_key
->eth_src
, ETH_ALEN
, is_mask
);
921 SW_FLOW_KEY_MEMCPY(match
, eth
.dst
,
922 eth_key
->eth_dst
, ETH_ALEN
, is_mask
);
923 attrs
&= ~(1ULL << OVS_KEY_ATTR_ETHERNET
);
926 if (attrs
& (1ULL << OVS_KEY_ATTR_VLAN
)) {
929 tci
= nla_get_be16(a
[OVS_KEY_ATTR_VLAN
]);
930 if (!(tci
& htons(VLAN_TAG_PRESENT
))) {
932 OVS_NLERR(log
, "VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.");
934 OVS_NLERR(log
, "VLAN TCI does not have VLAN_TAG_PRESENT bit set.");
939 SW_FLOW_KEY_PUT(match
, eth
.tci
, tci
, is_mask
);
940 attrs
&= ~(1ULL << OVS_KEY_ATTR_VLAN
);
943 if (attrs
& (1ULL << OVS_KEY_ATTR_ETHERTYPE
)) {
946 eth_type
= nla_get_be16(a
[OVS_KEY_ATTR_ETHERTYPE
]);
948 /* Always exact match EtherType. */
949 eth_type
= htons(0xffff);
950 } else if (!eth_proto_is_802_3(eth_type
)) {
951 OVS_NLERR(log
, "EtherType %x is less than min %x",
952 ntohs(eth_type
), ETH_P_802_3_MIN
);
956 SW_FLOW_KEY_PUT(match
, eth
.type
, eth_type
, is_mask
);
957 attrs
&= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE
);
958 } else if (!is_mask
) {
959 SW_FLOW_KEY_PUT(match
, eth
.type
, htons(ETH_P_802_2
), is_mask
);
962 if (attrs
& (1 << OVS_KEY_ATTR_IPV4
)) {
963 const struct ovs_key_ipv4
*ipv4_key
;
965 ipv4_key
= nla_data(a
[OVS_KEY_ATTR_IPV4
]);
966 if (!is_mask
&& ipv4_key
->ipv4_frag
> OVS_FRAG_TYPE_MAX
) {
967 OVS_NLERR(log
, "IPv4 frag type %d is out of range max %d",
968 ipv4_key
->ipv4_frag
, OVS_FRAG_TYPE_MAX
);
971 SW_FLOW_KEY_PUT(match
, ip
.proto
,
972 ipv4_key
->ipv4_proto
, is_mask
);
973 SW_FLOW_KEY_PUT(match
, ip
.tos
,
974 ipv4_key
->ipv4_tos
, is_mask
);
975 SW_FLOW_KEY_PUT(match
, ip
.ttl
,
976 ipv4_key
->ipv4_ttl
, is_mask
);
977 SW_FLOW_KEY_PUT(match
, ip
.frag
,
978 ipv4_key
->ipv4_frag
, is_mask
);
979 SW_FLOW_KEY_PUT(match
, ipv4
.addr
.src
,
980 ipv4_key
->ipv4_src
, is_mask
);
981 SW_FLOW_KEY_PUT(match
, ipv4
.addr
.dst
,
982 ipv4_key
->ipv4_dst
, is_mask
);
983 attrs
&= ~(1 << OVS_KEY_ATTR_IPV4
);
986 if (attrs
& (1ULL << OVS_KEY_ATTR_IPV6
)) {
987 const struct ovs_key_ipv6
*ipv6_key
;
989 ipv6_key
= nla_data(a
[OVS_KEY_ATTR_IPV6
]);
990 if (!is_mask
&& ipv6_key
->ipv6_frag
> OVS_FRAG_TYPE_MAX
) {
991 OVS_NLERR(log
, "IPv6 frag type %d is out of range max %d",
992 ipv6_key
->ipv6_frag
, OVS_FRAG_TYPE_MAX
);
996 if (!is_mask
&& ipv6_key
->ipv6_label
& htonl(0xFFF00000)) {
997 OVS_NLERR(log
, "IPv6 flow label %x is out of range (max=%x).\n",
998 ntohl(ipv6_key
->ipv6_label
), (1 << 20) - 1);
1002 SW_FLOW_KEY_PUT(match
, ipv6
.label
,
1003 ipv6_key
->ipv6_label
, is_mask
);
1004 SW_FLOW_KEY_PUT(match
, ip
.proto
,
1005 ipv6_key
->ipv6_proto
, is_mask
);
1006 SW_FLOW_KEY_PUT(match
, ip
.tos
,
1007 ipv6_key
->ipv6_tclass
, is_mask
);
1008 SW_FLOW_KEY_PUT(match
, ip
.ttl
,
1009 ipv6_key
->ipv6_hlimit
, is_mask
);
1010 SW_FLOW_KEY_PUT(match
, ip
.frag
,
1011 ipv6_key
->ipv6_frag
, is_mask
);
1012 SW_FLOW_KEY_MEMCPY(match
, ipv6
.addr
.src
,
1014 sizeof(match
->key
->ipv6
.addr
.src
),
1016 SW_FLOW_KEY_MEMCPY(match
, ipv6
.addr
.dst
,
1018 sizeof(match
->key
->ipv6
.addr
.dst
),
1021 attrs
&= ~(1ULL << OVS_KEY_ATTR_IPV6
);
1024 if (attrs
& (1ULL << OVS_KEY_ATTR_ARP
)) {
1025 const struct ovs_key_arp
*arp_key
;
1027 arp_key
= nla_data(a
[OVS_KEY_ATTR_ARP
]);
1028 if (!is_mask
&& (arp_key
->arp_op
& htons(0xff00))) {
1029 OVS_NLERR(log
, "Unknown ARP opcode (opcode=%d).",
1034 SW_FLOW_KEY_PUT(match
, ipv4
.addr
.src
,
1035 arp_key
->arp_sip
, is_mask
);
1036 SW_FLOW_KEY_PUT(match
, ipv4
.addr
.dst
,
1037 arp_key
->arp_tip
, is_mask
);
1038 SW_FLOW_KEY_PUT(match
, ip
.proto
,
1039 ntohs(arp_key
->arp_op
), is_mask
);
1040 SW_FLOW_KEY_MEMCPY(match
, ipv4
.arp
.sha
,
1041 arp_key
->arp_sha
, ETH_ALEN
, is_mask
);
1042 SW_FLOW_KEY_MEMCPY(match
, ipv4
.arp
.tha
,
1043 arp_key
->arp_tha
, ETH_ALEN
, is_mask
);
1045 attrs
&= ~(1ULL << OVS_KEY_ATTR_ARP
);
1048 if (attrs
& (1ULL << OVS_KEY_ATTR_MPLS
)) {
1049 const struct ovs_key_mpls
*mpls_key
;
1051 mpls_key
= nla_data(a
[OVS_KEY_ATTR_MPLS
]);
1052 SW_FLOW_KEY_PUT(match
, mpls
.top_lse
,
1053 mpls_key
->mpls_lse
, is_mask
);
1055 attrs
&= ~(1ULL << OVS_KEY_ATTR_MPLS
);
1058 if (attrs
& (1ULL << OVS_KEY_ATTR_TCP
)) {
1059 const struct ovs_key_tcp
*tcp_key
;
1061 tcp_key
= nla_data(a
[OVS_KEY_ATTR_TCP
]);
1062 SW_FLOW_KEY_PUT(match
, tp
.src
, tcp_key
->tcp_src
, is_mask
);
1063 SW_FLOW_KEY_PUT(match
, tp
.dst
, tcp_key
->tcp_dst
, is_mask
);
1064 attrs
&= ~(1ULL << OVS_KEY_ATTR_TCP
);
1067 if (attrs
& (1ULL << OVS_KEY_ATTR_TCP_FLAGS
)) {
1068 SW_FLOW_KEY_PUT(match
, tp
.flags
,
1069 nla_get_be16(a
[OVS_KEY_ATTR_TCP_FLAGS
]),
1071 attrs
&= ~(1ULL << OVS_KEY_ATTR_TCP_FLAGS
);
1074 if (attrs
& (1ULL << OVS_KEY_ATTR_UDP
)) {
1075 const struct ovs_key_udp
*udp_key
;
1077 udp_key
= nla_data(a
[OVS_KEY_ATTR_UDP
]);
1078 SW_FLOW_KEY_PUT(match
, tp
.src
, udp_key
->udp_src
, is_mask
);
1079 SW_FLOW_KEY_PUT(match
, tp
.dst
, udp_key
->udp_dst
, is_mask
);
1080 attrs
&= ~(1ULL << OVS_KEY_ATTR_UDP
);
1083 if (attrs
& (1ULL << OVS_KEY_ATTR_SCTP
)) {
1084 const struct ovs_key_sctp
*sctp_key
;
1086 sctp_key
= nla_data(a
[OVS_KEY_ATTR_SCTP
]);
1087 SW_FLOW_KEY_PUT(match
, tp
.src
, sctp_key
->sctp_src
, is_mask
);
1088 SW_FLOW_KEY_PUT(match
, tp
.dst
, sctp_key
->sctp_dst
, is_mask
);
1089 attrs
&= ~(1ULL << OVS_KEY_ATTR_SCTP
);
1092 if (attrs
& (1ULL << OVS_KEY_ATTR_ICMP
)) {
1093 const struct ovs_key_icmp
*icmp_key
;
1095 icmp_key
= nla_data(a
[OVS_KEY_ATTR_ICMP
]);
1096 SW_FLOW_KEY_PUT(match
, tp
.src
,
1097 htons(icmp_key
->icmp_type
), is_mask
);
1098 SW_FLOW_KEY_PUT(match
, tp
.dst
,
1099 htons(icmp_key
->icmp_code
), is_mask
);
1100 attrs
&= ~(1ULL << OVS_KEY_ATTR_ICMP
);
1103 if (attrs
& (1ULL << OVS_KEY_ATTR_ICMPV6
)) {
1104 const struct ovs_key_icmpv6
*icmpv6_key
;
1106 icmpv6_key
= nla_data(a
[OVS_KEY_ATTR_ICMPV6
]);
1107 SW_FLOW_KEY_PUT(match
, tp
.src
,
1108 htons(icmpv6_key
->icmpv6_type
), is_mask
);
1109 SW_FLOW_KEY_PUT(match
, tp
.dst
,
1110 htons(icmpv6_key
->icmpv6_code
), is_mask
);
1111 attrs
&= ~(1ULL << OVS_KEY_ATTR_ICMPV6
);
1114 if (attrs
& (1ULL << OVS_KEY_ATTR_ND
)) {
1115 const struct ovs_key_nd
*nd_key
;
1117 nd_key
= nla_data(a
[OVS_KEY_ATTR_ND
]);
1118 SW_FLOW_KEY_MEMCPY(match
, ipv6
.nd
.target
,
1120 sizeof(match
->key
->ipv6
.nd
.target
),
1122 SW_FLOW_KEY_MEMCPY(match
, ipv6
.nd
.sll
,
1123 nd_key
->nd_sll
, ETH_ALEN
, is_mask
);
1124 SW_FLOW_KEY_MEMCPY(match
, ipv6
.nd
.tll
,
1125 nd_key
->nd_tll
, ETH_ALEN
, is_mask
);
1126 attrs
&= ~(1ULL << OVS_KEY_ATTR_ND
);
1130 OVS_NLERR(log
, "Unknown key attributes %llx",
1131 (unsigned long long)attrs
);
1138 static void nlattr_set(struct nlattr
*attr
, u8 val
,
1139 const struct ovs_len_tbl
*tbl
)
1144 /* The nlattr stream should already have been validated */
1145 nla_for_each_nested(nla
, attr
, rem
) {
1146 if (tbl
[nla_type(nla
)].len
== OVS_ATTR_NESTED
) {
1147 if (tbl
[nla_type(nla
)].next
)
1148 tbl
= tbl
[nla_type(nla
)].next
;
1149 nlattr_set(nla
, val
, tbl
);
1151 memset(nla_data(nla
), val
, nla_len(nla
));
1154 if (nla_type(nla
) == OVS_KEY_ATTR_CT_STATE
)
1155 *(u32
*)nla_data(nla
) &= CT_SUPPORTED_MASK
;
1159 static void mask_set_nlattr(struct nlattr
*attr
, u8 val
)
1161 nlattr_set(attr
, val
, ovs_key_lens
);
1165 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1166 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1167 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1168 * does not include any don't care bit.
1169 * @net: Used to determine per-namespace field support.
1170 * @match: receives the extracted flow match information.
1171 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1172 * sequence. The fields should of the packet that triggered the creation
1174 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1175 * attribute specifies the mask field of the wildcarded flow.
1176 * @log: Boolean to allow kernel error logging. Normally true, but when
1177 * probing for feature compatibility this should be passed in as false to
1178 * suppress unnecessary error logging.
1180 int ovs_nla_get_match(struct net
*net
, struct sw_flow_match
*match
,
1181 const struct nlattr
*nla_key
,
1182 const struct nlattr
*nla_mask
,
1185 const struct nlattr
*a
[OVS_KEY_ATTR_MAX
+ 1];
1186 const struct nlattr
*encap
;
1187 struct nlattr
*newmask
= NULL
;
1190 bool encap_valid
= false;
1193 err
= parse_flow_nlattrs(nla_key
, a
, &key_attrs
, log
);
1197 if ((key_attrs
& (1ULL << OVS_KEY_ATTR_ETHERNET
)) &&
1198 (key_attrs
& (1ULL << OVS_KEY_ATTR_ETHERTYPE
)) &&
1199 (nla_get_be16(a
[OVS_KEY_ATTR_ETHERTYPE
]) == htons(ETH_P_8021Q
))) {
1202 if (!((key_attrs
& (1ULL << OVS_KEY_ATTR_VLAN
)) &&
1203 (key_attrs
& (1ULL << OVS_KEY_ATTR_ENCAP
)))) {
1204 OVS_NLERR(log
, "Invalid Vlan frame.");
1208 key_attrs
&= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE
);
1209 tci
= nla_get_be16(a
[OVS_KEY_ATTR_VLAN
]);
1210 encap
= a
[OVS_KEY_ATTR_ENCAP
];
1211 key_attrs
&= ~(1ULL << OVS_KEY_ATTR_ENCAP
);
1214 if (tci
& htons(VLAN_TAG_PRESENT
)) {
1215 err
= parse_flow_nlattrs(encap
, a
, &key_attrs
, log
);
1219 /* Corner case for truncated 802.1Q header. */
1220 if (nla_len(encap
)) {
1221 OVS_NLERR(log
, "Truncated 802.1Q header has non-zero encap attribute.");
1225 OVS_NLERR(log
, "Encap attr is set for non-VLAN frame");
1230 err
= ovs_key_from_nlattrs(net
, match
, key_attrs
, a
, false, log
);
1236 /* Create an exact match mask. We need to set to 0xff
1237 * all the 'match->mask' fields that have been touched
1238 * in 'match->key'. We cannot simply memset
1239 * 'match->mask', because padding bytes and fields not
1240 * specified in 'match->key' should be left to 0.
1241 * Instead, we use a stream of netlink attributes,
1242 * copied from 'key' and set to 0xff.
1243 * ovs_key_from_nlattrs() will take care of filling
1244 * 'match->mask' appropriately.
1246 newmask
= kmemdup(nla_key
,
1247 nla_total_size(nla_len(nla_key
)),
1252 mask_set_nlattr(newmask
, 0xff);
1254 /* The userspace does not send tunnel attributes that
1255 * are 0, but we should not wildcard them nonetheless.
1257 if (match
->key
->tun_key
.u
.ipv4
.dst
)
1258 SW_FLOW_KEY_MEMSET_FIELD(match
, tun_key
,
1264 err
= parse_flow_mask_nlattrs(nla_mask
, a
, &mask_attrs
, log
);
1268 /* Always match on tci. */
1269 SW_FLOW_KEY_PUT(match
, eth
.tci
, htons(0xffff), true);
1271 if (mask_attrs
& 1ULL << OVS_KEY_ATTR_ENCAP
) {
1272 __be16 eth_type
= 0;
1276 OVS_NLERR(log
, "Encap mask attribute is set for non-VLAN frame.");
1281 mask_attrs
&= ~(1ULL << OVS_KEY_ATTR_ENCAP
);
1282 if (a
[OVS_KEY_ATTR_ETHERTYPE
])
1283 eth_type
= nla_get_be16(a
[OVS_KEY_ATTR_ETHERTYPE
]);
1285 if (eth_type
== htons(0xffff)) {
1286 mask_attrs
&= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE
);
1287 encap
= a
[OVS_KEY_ATTR_ENCAP
];
1288 err
= parse_flow_mask_nlattrs(encap
, a
,
1293 OVS_NLERR(log
, "VLAN frames must have an exact match on the TPID (mask=%x).",
1299 if (a
[OVS_KEY_ATTR_VLAN
])
1300 tci
= nla_get_be16(a
[OVS_KEY_ATTR_VLAN
]);
1302 if (!(tci
& htons(VLAN_TAG_PRESENT
))) {
1303 OVS_NLERR(log
, "VLAN tag present bit must have an exact match (tci_mask=%x).",
1310 err
= ovs_key_from_nlattrs(net
, match
, mask_attrs
, a
, true,
1316 if (!match_validate(match
, key_attrs
, mask_attrs
, log
))
1324 static size_t get_ufid_len(const struct nlattr
*attr
, bool log
)
1331 len
= nla_len(attr
);
1332 if (len
< 1 || len
> MAX_UFID_LENGTH
) {
1333 OVS_NLERR(log
, "ufid size %u bytes exceeds the range (1, %d)",
1334 nla_len(attr
), MAX_UFID_LENGTH
);
1341 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1342 * or false otherwise.
1344 bool ovs_nla_get_ufid(struct sw_flow_id
*sfid
, const struct nlattr
*attr
,
1347 sfid
->ufid_len
= get_ufid_len(attr
, log
);
1349 memcpy(sfid
->ufid
, nla_data(attr
), sfid
->ufid_len
);
1351 return sfid
->ufid_len
;
1354 int ovs_nla_get_identifier(struct sw_flow_id
*sfid
, const struct nlattr
*ufid
,
1355 const struct sw_flow_key
*key
, bool log
)
1357 struct sw_flow_key
*new_key
;
1359 if (ovs_nla_get_ufid(sfid
, ufid
, log
))
1362 /* If UFID was not provided, use unmasked key. */
1363 new_key
= kmalloc(sizeof(*new_key
), GFP_KERNEL
);
1366 memcpy(new_key
, key
, sizeof(*key
));
1367 sfid
->unmasked_key
= new_key
;
1372 u32
ovs_nla_get_ufid_flags(const struct nlattr
*attr
)
1374 return attr
? nla_get_u32(attr
) : 0;
1378 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1379 * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1380 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1382 * @log: Boolean to allow kernel error logging. Normally true, but when
1383 * probing for feature compatibility this should be passed in as false to
1384 * suppress unnecessary error logging.
1386 * This parses a series of Netlink attributes that form a flow key, which must
1387 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1388 * get the metadata, that is, the parts of the flow key that cannot be
1389 * extracted from the packet itself.
1392 int ovs_nla_get_flow_metadata(struct net
*net
, const struct nlattr
*attr
,
1393 struct sw_flow_key
*key
,
1396 const struct nlattr
*a
[OVS_KEY_ATTR_MAX
+ 1];
1397 struct sw_flow_match match
;
1401 err
= parse_flow_nlattrs(attr
, a
, &attrs
, log
);
1405 memset(&match
, 0, sizeof(match
));
1408 memset(key
, 0, OVS_SW_FLOW_KEY_METADATA_SIZE
);
1409 memset(&key
->ct
, 0, sizeof(key
->ct
));
1410 key
->phy
.in_port
= DP_MAX_PORTS
;
1412 return metadata_from_nlattrs(net
, &match
, &attrs
, a
, false, log
);
1415 static int __ovs_nla_put_key(const struct sw_flow_key
*swkey
,
1416 const struct sw_flow_key
*output
, bool is_mask
,
1417 struct sk_buff
*skb
)
1419 struct ovs_key_ethernet
*eth_key
;
1420 struct nlattr
*nla
, *encap
;
1422 if (nla_put_u32(skb
, OVS_KEY_ATTR_RECIRC_ID
, output
->recirc_id
))
1423 goto nla_put_failure
;
1425 if (nla_put_u32(skb
, OVS_KEY_ATTR_DP_HASH
, output
->ovs_flow_hash
))
1426 goto nla_put_failure
;
1428 if (nla_put_u32(skb
, OVS_KEY_ATTR_PRIORITY
, output
->phy
.priority
))
1429 goto nla_put_failure
;
1431 if ((swkey
->tun_proto
|| is_mask
)) {
1432 const void *opts
= NULL
;
1434 if (output
->tun_key
.tun_flags
& TUNNEL_OPTIONS_PRESENT
)
1435 opts
= TUN_METADATA_OPTS(output
, swkey
->tun_opts_len
);
1437 if (ip_tun_to_nlattr(skb
, &output
->tun_key
, opts
,
1438 swkey
->tun_opts_len
, swkey
->tun_proto
))
1439 goto nla_put_failure
;
1442 if (swkey
->phy
.in_port
== DP_MAX_PORTS
) {
1443 if (is_mask
&& (output
->phy
.in_port
== 0xffff))
1444 if (nla_put_u32(skb
, OVS_KEY_ATTR_IN_PORT
, 0xffffffff))
1445 goto nla_put_failure
;
1448 upper_u16
= !is_mask
? 0 : 0xffff;
1450 if (nla_put_u32(skb
, OVS_KEY_ATTR_IN_PORT
,
1451 (upper_u16
<< 16) | output
->phy
.in_port
))
1452 goto nla_put_failure
;
1455 if (nla_put_u32(skb
, OVS_KEY_ATTR_SKB_MARK
, output
->phy
.skb_mark
))
1456 goto nla_put_failure
;
1458 if (ovs_ct_put_key(output
, skb
))
1459 goto nla_put_failure
;
1461 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ETHERNET
, sizeof(*eth_key
));
1463 goto nla_put_failure
;
1465 eth_key
= nla_data(nla
);
1466 ether_addr_copy(eth_key
->eth_src
, output
->eth
.src
);
1467 ether_addr_copy(eth_key
->eth_dst
, output
->eth
.dst
);
1469 if (swkey
->eth
.tci
|| swkey
->eth
.type
== htons(ETH_P_8021Q
)) {
1471 eth_type
= !is_mask
? htons(ETH_P_8021Q
) : htons(0xffff);
1472 if (nla_put_be16(skb
, OVS_KEY_ATTR_ETHERTYPE
, eth_type
) ||
1473 nla_put_be16(skb
, OVS_KEY_ATTR_VLAN
, output
->eth
.tci
))
1474 goto nla_put_failure
;
1475 encap
= nla_nest_start(skb
, OVS_KEY_ATTR_ENCAP
);
1476 if (!swkey
->eth
.tci
)
1481 if (swkey
->eth
.type
== htons(ETH_P_802_2
)) {
1483 * Ethertype 802.2 is represented in the netlink with omitted
1484 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1485 * 0xffff in the mask attribute. Ethertype can also
1488 if (is_mask
&& output
->eth
.type
)
1489 if (nla_put_be16(skb
, OVS_KEY_ATTR_ETHERTYPE
,
1491 goto nla_put_failure
;
1495 if (nla_put_be16(skb
, OVS_KEY_ATTR_ETHERTYPE
, output
->eth
.type
))
1496 goto nla_put_failure
;
1498 if (swkey
->eth
.type
== htons(ETH_P_IP
)) {
1499 struct ovs_key_ipv4
*ipv4_key
;
1501 nla
= nla_reserve(skb
, OVS_KEY_ATTR_IPV4
, sizeof(*ipv4_key
));
1503 goto nla_put_failure
;
1504 ipv4_key
= nla_data(nla
);
1505 ipv4_key
->ipv4_src
= output
->ipv4
.addr
.src
;
1506 ipv4_key
->ipv4_dst
= output
->ipv4
.addr
.dst
;
1507 ipv4_key
->ipv4_proto
= output
->ip
.proto
;
1508 ipv4_key
->ipv4_tos
= output
->ip
.tos
;
1509 ipv4_key
->ipv4_ttl
= output
->ip
.ttl
;
1510 ipv4_key
->ipv4_frag
= output
->ip
.frag
;
1511 } else if (swkey
->eth
.type
== htons(ETH_P_IPV6
)) {
1512 struct ovs_key_ipv6
*ipv6_key
;
1514 nla
= nla_reserve(skb
, OVS_KEY_ATTR_IPV6
, sizeof(*ipv6_key
));
1516 goto nla_put_failure
;
1517 ipv6_key
= nla_data(nla
);
1518 memcpy(ipv6_key
->ipv6_src
, &output
->ipv6
.addr
.src
,
1519 sizeof(ipv6_key
->ipv6_src
));
1520 memcpy(ipv6_key
->ipv6_dst
, &output
->ipv6
.addr
.dst
,
1521 sizeof(ipv6_key
->ipv6_dst
));
1522 ipv6_key
->ipv6_label
= output
->ipv6
.label
;
1523 ipv6_key
->ipv6_proto
= output
->ip
.proto
;
1524 ipv6_key
->ipv6_tclass
= output
->ip
.tos
;
1525 ipv6_key
->ipv6_hlimit
= output
->ip
.ttl
;
1526 ipv6_key
->ipv6_frag
= output
->ip
.frag
;
1527 } else if (swkey
->eth
.type
== htons(ETH_P_ARP
) ||
1528 swkey
->eth
.type
== htons(ETH_P_RARP
)) {
1529 struct ovs_key_arp
*arp_key
;
1531 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ARP
, sizeof(*arp_key
));
1533 goto nla_put_failure
;
1534 arp_key
= nla_data(nla
);
1535 memset(arp_key
, 0, sizeof(struct ovs_key_arp
));
1536 arp_key
->arp_sip
= output
->ipv4
.addr
.src
;
1537 arp_key
->arp_tip
= output
->ipv4
.addr
.dst
;
1538 arp_key
->arp_op
= htons(output
->ip
.proto
);
1539 ether_addr_copy(arp_key
->arp_sha
, output
->ipv4
.arp
.sha
);
1540 ether_addr_copy(arp_key
->arp_tha
, output
->ipv4
.arp
.tha
);
1541 } else if (eth_p_mpls(swkey
->eth
.type
)) {
1542 struct ovs_key_mpls
*mpls_key
;
1544 nla
= nla_reserve(skb
, OVS_KEY_ATTR_MPLS
, sizeof(*mpls_key
));
1546 goto nla_put_failure
;
1547 mpls_key
= nla_data(nla
);
1548 mpls_key
->mpls_lse
= output
->mpls
.top_lse
;
1551 if ((swkey
->eth
.type
== htons(ETH_P_IP
) ||
1552 swkey
->eth
.type
== htons(ETH_P_IPV6
)) &&
1553 swkey
->ip
.frag
!= OVS_FRAG_TYPE_LATER
) {
1555 if (swkey
->ip
.proto
== IPPROTO_TCP
) {
1556 struct ovs_key_tcp
*tcp_key
;
1558 nla
= nla_reserve(skb
, OVS_KEY_ATTR_TCP
, sizeof(*tcp_key
));
1560 goto nla_put_failure
;
1561 tcp_key
= nla_data(nla
);
1562 tcp_key
->tcp_src
= output
->tp
.src
;
1563 tcp_key
->tcp_dst
= output
->tp
.dst
;
1564 if (nla_put_be16(skb
, OVS_KEY_ATTR_TCP_FLAGS
,
1566 goto nla_put_failure
;
1567 } else if (swkey
->ip
.proto
== IPPROTO_UDP
) {
1568 struct ovs_key_udp
*udp_key
;
1570 nla
= nla_reserve(skb
, OVS_KEY_ATTR_UDP
, sizeof(*udp_key
));
1572 goto nla_put_failure
;
1573 udp_key
= nla_data(nla
);
1574 udp_key
->udp_src
= output
->tp
.src
;
1575 udp_key
->udp_dst
= output
->tp
.dst
;
1576 } else if (swkey
->ip
.proto
== IPPROTO_SCTP
) {
1577 struct ovs_key_sctp
*sctp_key
;
1579 nla
= nla_reserve(skb
, OVS_KEY_ATTR_SCTP
, sizeof(*sctp_key
));
1581 goto nla_put_failure
;
1582 sctp_key
= nla_data(nla
);
1583 sctp_key
->sctp_src
= output
->tp
.src
;
1584 sctp_key
->sctp_dst
= output
->tp
.dst
;
1585 } else if (swkey
->eth
.type
== htons(ETH_P_IP
) &&
1586 swkey
->ip
.proto
== IPPROTO_ICMP
) {
1587 struct ovs_key_icmp
*icmp_key
;
1589 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ICMP
, sizeof(*icmp_key
));
1591 goto nla_put_failure
;
1592 icmp_key
= nla_data(nla
);
1593 icmp_key
->icmp_type
= ntohs(output
->tp
.src
);
1594 icmp_key
->icmp_code
= ntohs(output
->tp
.dst
);
1595 } else if (swkey
->eth
.type
== htons(ETH_P_IPV6
) &&
1596 swkey
->ip
.proto
== IPPROTO_ICMPV6
) {
1597 struct ovs_key_icmpv6
*icmpv6_key
;
1599 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ICMPV6
,
1600 sizeof(*icmpv6_key
));
1602 goto nla_put_failure
;
1603 icmpv6_key
= nla_data(nla
);
1604 icmpv6_key
->icmpv6_type
= ntohs(output
->tp
.src
);
1605 icmpv6_key
->icmpv6_code
= ntohs(output
->tp
.dst
);
1607 if (icmpv6_key
->icmpv6_type
== NDISC_NEIGHBOUR_SOLICITATION
||
1608 icmpv6_key
->icmpv6_type
== NDISC_NEIGHBOUR_ADVERTISEMENT
) {
1609 struct ovs_key_nd
*nd_key
;
1611 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ND
, sizeof(*nd_key
));
1613 goto nla_put_failure
;
1614 nd_key
= nla_data(nla
);
1615 memcpy(nd_key
->nd_target
, &output
->ipv6
.nd
.target
,
1616 sizeof(nd_key
->nd_target
));
1617 ether_addr_copy(nd_key
->nd_sll
, output
->ipv6
.nd
.sll
);
1618 ether_addr_copy(nd_key
->nd_tll
, output
->ipv6
.nd
.tll
);
1625 nla_nest_end(skb
, encap
);
1633 int ovs_nla_put_key(const struct sw_flow_key
*swkey
,
1634 const struct sw_flow_key
*output
, int attr
, bool is_mask
,
1635 struct sk_buff
*skb
)
1640 nla
= nla_nest_start(skb
, attr
);
1643 err
= __ovs_nla_put_key(swkey
, output
, is_mask
, skb
);
1646 nla_nest_end(skb
, nla
);
1651 /* Called with ovs_mutex or RCU read lock. */
1652 int ovs_nla_put_identifier(const struct sw_flow
*flow
, struct sk_buff
*skb
)
1654 if (ovs_identifier_is_ufid(&flow
->id
))
1655 return nla_put(skb
, OVS_FLOW_ATTR_UFID
, flow
->id
.ufid_len
,
1658 return ovs_nla_put_key(flow
->id
.unmasked_key
, flow
->id
.unmasked_key
,
1659 OVS_FLOW_ATTR_KEY
, false, skb
);
1662 /* Called with ovs_mutex or RCU read lock. */
1663 int ovs_nla_put_masked_key(const struct sw_flow
*flow
, struct sk_buff
*skb
)
1665 return ovs_nla_put_key(&flow
->key
, &flow
->key
,
1666 OVS_FLOW_ATTR_KEY
, false, skb
);
1669 /* Called with ovs_mutex or RCU read lock. */
1670 int ovs_nla_put_mask(const struct sw_flow
*flow
, struct sk_buff
*skb
)
1672 return ovs_nla_put_key(&flow
->key
, &flow
->mask
->key
,
1673 OVS_FLOW_ATTR_MASK
, true, skb
);
1676 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1678 static struct sw_flow_actions
*nla_alloc_flow_actions(int size
, bool log
)
1680 struct sw_flow_actions
*sfa
;
1682 if (size
> MAX_ACTIONS_BUFSIZE
) {
1683 OVS_NLERR(log
, "Flow action size %u bytes exceeds max", size
);
1684 return ERR_PTR(-EINVAL
);
1687 sfa
= kmalloc(sizeof(*sfa
) + size
, GFP_KERNEL
);
1689 return ERR_PTR(-ENOMEM
);
1691 sfa
->actions_len
= 0;
1695 static void ovs_nla_free_set_action(const struct nlattr
*a
)
1697 const struct nlattr
*ovs_key
= nla_data(a
);
1698 struct ovs_tunnel_info
*ovs_tun
;
1700 switch (nla_type(ovs_key
)) {
1701 case OVS_KEY_ATTR_TUNNEL_INFO
:
1702 ovs_tun
= nla_data(ovs_key
);
1703 ovs_dst_release((struct dst_entry
*)ovs_tun
->tun_dst
);
1708 void ovs_nla_free_flow_actions(struct sw_flow_actions
*sf_acts
)
1710 const struct nlattr
*a
;
1716 nla_for_each_attr(a
, sf_acts
->actions
, sf_acts
->actions_len
, rem
) {
1717 switch (nla_type(a
)) {
1718 case OVS_ACTION_ATTR_SET
:
1719 ovs_nla_free_set_action(a
);
1721 case OVS_ACTION_ATTR_CT
:
1722 ovs_ct_free_action(a
);
1730 static void __ovs_nla_free_flow_actions(struct rcu_head
*head
)
1732 ovs_nla_free_flow_actions(container_of(head
, struct sw_flow_actions
, rcu
));
1735 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1736 * The caller must hold rcu_read_lock for this to be sensible. */
1737 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions
*sf_acts
)
1739 call_rcu(&sf_acts
->rcu
, __ovs_nla_free_flow_actions
);
1742 static struct nlattr
*reserve_sfa_size(struct sw_flow_actions
**sfa
,
1743 int attr_len
, bool log
)
1746 struct sw_flow_actions
*acts
;
1748 int req_size
= NLA_ALIGN(attr_len
);
1749 int next_offset
= offsetof(struct sw_flow_actions
, actions
) +
1750 (*sfa
)->actions_len
;
1752 if (req_size
<= (ksize(*sfa
) - next_offset
))
1755 new_acts_size
= ksize(*sfa
) * 2;
1757 if (new_acts_size
> MAX_ACTIONS_BUFSIZE
) {
1758 if ((MAX_ACTIONS_BUFSIZE
- next_offset
) < req_size
)
1759 return ERR_PTR(-EMSGSIZE
);
1760 new_acts_size
= MAX_ACTIONS_BUFSIZE
;
1763 acts
= nla_alloc_flow_actions(new_acts_size
, log
);
1765 return (void *)acts
;
1767 memcpy(acts
->actions
, (*sfa
)->actions
, (*sfa
)->actions_len
);
1768 acts
->actions_len
= (*sfa
)->actions_len
;
1769 acts
->orig_len
= (*sfa
)->orig_len
;
1774 (*sfa
)->actions_len
+= req_size
;
1775 return (struct nlattr
*) ((unsigned char *)(*sfa
) + next_offset
);
1778 static struct nlattr
*__add_action(struct sw_flow_actions
**sfa
,
1779 int attrtype
, void *data
, int len
, bool log
)
1783 a
= reserve_sfa_size(sfa
, nla_attr_size(len
), log
);
1787 a
->nla_type
= attrtype
;
1788 a
->nla_len
= nla_attr_size(len
);
1791 memcpy(nla_data(a
), data
, len
);
1792 memset((unsigned char *) a
+ a
->nla_len
, 0, nla_padlen(len
));
1797 int ovs_nla_add_action(struct sw_flow_actions
**sfa
, int attrtype
, void *data
,
1802 a
= __add_action(sfa
, attrtype
, data
, len
, log
);
1809 static inline int add_nested_action_start(struct sw_flow_actions
**sfa
,
1810 int attrtype
, bool log
)
1812 int used
= (*sfa
)->actions_len
;
1815 err
= ovs_nla_add_action(sfa
, attrtype
, NULL
, 0, log
);
1822 static inline void add_nested_action_end(struct sw_flow_actions
*sfa
,
1825 struct nlattr
*a
= (struct nlattr
*) ((unsigned char *)sfa
->actions
+
1828 a
->nla_len
= sfa
->actions_len
- st_offset
;
1831 static int __ovs_nla_copy_actions(struct net
*net
, const struct nlattr
*attr
,
1832 const struct sw_flow_key
*key
,
1833 int depth
, struct sw_flow_actions
**sfa
,
1834 __be16 eth_type
, __be16 vlan_tci
, bool log
);
1836 static int validate_and_copy_sample(struct net
*net
, const struct nlattr
*attr
,
1837 const struct sw_flow_key
*key
, int depth
,
1838 struct sw_flow_actions
**sfa
,
1839 __be16 eth_type
, __be16 vlan_tci
, bool log
)
1841 const struct nlattr
*attrs
[OVS_SAMPLE_ATTR_MAX
+ 1];
1842 const struct nlattr
*probability
, *actions
;
1843 const struct nlattr
*a
;
1844 int rem
, start
, err
, st_acts
;
1846 memset(attrs
, 0, sizeof(attrs
));
1847 nla_for_each_nested(a
, attr
, rem
) {
1848 int type
= nla_type(a
);
1849 if (!type
|| type
> OVS_SAMPLE_ATTR_MAX
|| attrs
[type
])
1856 probability
= attrs
[OVS_SAMPLE_ATTR_PROBABILITY
];
1857 if (!probability
|| nla_len(probability
) != sizeof(u32
))
1860 actions
= attrs
[OVS_SAMPLE_ATTR_ACTIONS
];
1861 if (!actions
|| (nla_len(actions
) && nla_len(actions
) < NLA_HDRLEN
))
1864 /* validation done, copy sample action. */
1865 start
= add_nested_action_start(sfa
, OVS_ACTION_ATTR_SAMPLE
, log
);
1868 err
= ovs_nla_add_action(sfa
, OVS_SAMPLE_ATTR_PROBABILITY
,
1869 nla_data(probability
), sizeof(u32
), log
);
1872 st_acts
= add_nested_action_start(sfa
, OVS_SAMPLE_ATTR_ACTIONS
, log
);
1876 err
= __ovs_nla_copy_actions(net
, actions
, key
, depth
+ 1, sfa
,
1877 eth_type
, vlan_tci
, log
);
1881 add_nested_action_end(*sfa
, st_acts
);
1882 add_nested_action_end(*sfa
, start
);
1887 void ovs_match_init(struct sw_flow_match
*match
,
1888 struct sw_flow_key
*key
,
1889 struct sw_flow_mask
*mask
)
1891 memset(match
, 0, sizeof(*match
));
1895 memset(key
, 0, sizeof(*key
));
1898 memset(&mask
->key
, 0, sizeof(mask
->key
));
1899 mask
->range
.start
= mask
->range
.end
= 0;
1903 static int validate_geneve_opts(struct sw_flow_key
*key
)
1905 struct geneve_opt
*option
;
1906 int opts_len
= key
->tun_opts_len
;
1907 bool crit_opt
= false;
1909 option
= (struct geneve_opt
*)TUN_METADATA_OPTS(key
, key
->tun_opts_len
);
1910 while (opts_len
> 0) {
1913 if (opts_len
< sizeof(*option
))
1916 len
= sizeof(*option
) + option
->length
* 4;
1920 crit_opt
|= !!(option
->type
& GENEVE_CRIT_OPT_TYPE
);
1922 option
= (struct geneve_opt
*)((u8
*)option
+ len
);
1926 key
->tun_key
.tun_flags
|= crit_opt
? TUNNEL_CRIT_OPT
: 0;
1931 static int validate_and_copy_set_tun(const struct nlattr
*attr
,
1932 struct sw_flow_actions
**sfa
, bool log
)
1934 struct sw_flow_match match
;
1935 struct sw_flow_key key
;
1936 struct metadata_dst
*tun_dst
;
1937 struct ip_tunnel_info
*tun_info
;
1938 struct ovs_tunnel_info
*ovs_tun
;
1940 int err
= 0, start
, opts_type
;
1942 ovs_match_init(&match
, &key
, NULL
);
1943 opts_type
= ip_tun_from_nlattr(nla_data(attr
), &match
, false, log
);
1947 if (key
.tun_opts_len
) {
1948 switch (opts_type
) {
1949 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS
:
1950 err
= validate_geneve_opts(&key
);
1954 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS
:
1959 start
= add_nested_action_start(sfa
, OVS_ACTION_ATTR_SET
, log
);
1963 tun_dst
= metadata_dst_alloc(key
.tun_opts_len
, GFP_KERNEL
);
1967 err
= dst_cache_init(&tun_dst
->u
.tun_info
.dst_cache
, GFP_KERNEL
);
1969 dst_release((struct dst_entry
*)tun_dst
);
1972 a
= __add_action(sfa
, OVS_KEY_ATTR_TUNNEL_INFO
, NULL
,
1973 sizeof(*ovs_tun
), log
);
1975 ovs_dst_release((struct dst_entry
*)tun_dst
);
1979 ovs_tun
= nla_data(a
);
1980 ovs_tun
->tun_dst
= tun_dst
;
1982 tun_info
= &tun_dst
->u
.tun_info
;
1983 tun_info
->mode
= IP_TUNNEL_INFO_TX
;
1984 if (key
.tun_proto
== AF_INET6
)
1985 tun_info
->mode
|= IP_TUNNEL_INFO_IPV6
;
1986 tun_info
->key
= key
.tun_key
;
1988 /* We need to store the options in the action itself since
1989 * everything else will go away after flow setup. We can append
1990 * it to tun_info and then point there.
1992 ip_tunnel_info_opts_set(tun_info
,
1993 TUN_METADATA_OPTS(&key
, key
.tun_opts_len
),
1995 add_nested_action_end(*sfa
, start
);
2000 /* Return false if there are any non-masked bits set.
2001 * Mask follows data immediately, before any netlink padding.
2003 static bool validate_masked(u8
*data
, int len
)
2005 u8
*mask
= data
+ len
;
2008 if (*data
++ & ~*mask
++)
2014 static int validate_set(const struct nlattr
*a
,
2015 const struct sw_flow_key
*flow_key
,
2016 struct sw_flow_actions
**sfa
,
2017 bool *skip_copy
, __be16 eth_type
, bool masked
, bool log
)
2019 const struct nlattr
*ovs_key
= nla_data(a
);
2020 int key_type
= nla_type(ovs_key
);
2023 /* There can be only one key in a action */
2024 if (nla_total_size(nla_len(ovs_key
)) != nla_len(a
))
2027 key_len
= nla_len(ovs_key
);
2031 if (key_type
> OVS_KEY_ATTR_MAX
||
2032 !check_attr_len(key_len
, ovs_key_lens
[key_type
].len
))
2035 if (masked
&& !validate_masked(nla_data(ovs_key
), key_len
))
2039 const struct ovs_key_ipv4
*ipv4_key
;
2040 const struct ovs_key_ipv6
*ipv6_key
;
2043 case OVS_KEY_ATTR_PRIORITY
:
2044 case OVS_KEY_ATTR_SKB_MARK
:
2045 case OVS_KEY_ATTR_CT_MARK
:
2046 case OVS_KEY_ATTR_CT_LABELS
:
2047 case OVS_KEY_ATTR_ETHERNET
:
2050 case OVS_KEY_ATTR_TUNNEL
:
2051 if (eth_p_mpls(eth_type
))
2055 return -EINVAL
; /* Masked tunnel set not supported. */
2058 err
= validate_and_copy_set_tun(a
, sfa
, log
);
2063 case OVS_KEY_ATTR_IPV4
:
2064 if (eth_type
!= htons(ETH_P_IP
))
2067 ipv4_key
= nla_data(ovs_key
);
2070 const struct ovs_key_ipv4
*mask
= ipv4_key
+ 1;
2072 /* Non-writeable fields. */
2073 if (mask
->ipv4_proto
|| mask
->ipv4_frag
)
2076 if (ipv4_key
->ipv4_proto
!= flow_key
->ip
.proto
)
2079 if (ipv4_key
->ipv4_frag
!= flow_key
->ip
.frag
)
2084 case OVS_KEY_ATTR_IPV6
:
2085 if (eth_type
!= htons(ETH_P_IPV6
))
2088 ipv6_key
= nla_data(ovs_key
);
2091 const struct ovs_key_ipv6
*mask
= ipv6_key
+ 1;
2093 /* Non-writeable fields. */
2094 if (mask
->ipv6_proto
|| mask
->ipv6_frag
)
2097 /* Invalid bits in the flow label mask? */
2098 if (ntohl(mask
->ipv6_label
) & 0xFFF00000)
2101 if (ipv6_key
->ipv6_proto
!= flow_key
->ip
.proto
)
2104 if (ipv6_key
->ipv6_frag
!= flow_key
->ip
.frag
)
2107 if (ntohl(ipv6_key
->ipv6_label
) & 0xFFF00000)
2112 case OVS_KEY_ATTR_TCP
:
2113 if ((eth_type
!= htons(ETH_P_IP
) &&
2114 eth_type
!= htons(ETH_P_IPV6
)) ||
2115 flow_key
->ip
.proto
!= IPPROTO_TCP
)
2120 case OVS_KEY_ATTR_UDP
:
2121 if ((eth_type
!= htons(ETH_P_IP
) &&
2122 eth_type
!= htons(ETH_P_IPV6
)) ||
2123 flow_key
->ip
.proto
!= IPPROTO_UDP
)
2128 case OVS_KEY_ATTR_MPLS
:
2129 if (!eth_p_mpls(eth_type
))
2133 case OVS_KEY_ATTR_SCTP
:
2134 if ((eth_type
!= htons(ETH_P_IP
) &&
2135 eth_type
!= htons(ETH_P_IPV6
)) ||
2136 flow_key
->ip
.proto
!= IPPROTO_SCTP
)
2145 /* Convert non-masked non-tunnel set actions to masked set actions. */
2146 if (!masked
&& key_type
!= OVS_KEY_ATTR_TUNNEL
) {
2147 int start
, len
= key_len
* 2;
2152 start
= add_nested_action_start(sfa
,
2153 OVS_ACTION_ATTR_SET_TO_MASKED
,
2158 at
= __add_action(sfa
, key_type
, NULL
, len
, log
);
2162 memcpy(nla_data(at
), nla_data(ovs_key
), key_len
); /* Key. */
2163 memset(nla_data(at
) + key_len
, 0xff, key_len
); /* Mask. */
2164 /* Clear non-writeable bits from otherwise writeable fields. */
2165 if (key_type
== OVS_KEY_ATTR_IPV6
) {
2166 struct ovs_key_ipv6
*mask
= nla_data(at
) + key_len
;
2168 mask
->ipv6_label
&= htonl(0x000FFFFF);
2170 add_nested_action_end(*sfa
, start
);
2176 static int validate_userspace(const struct nlattr
*attr
)
2178 static const struct nla_policy userspace_policy
[OVS_USERSPACE_ATTR_MAX
+ 1] = {
2179 [OVS_USERSPACE_ATTR_PID
] = {.type
= NLA_U32
},
2180 [OVS_USERSPACE_ATTR_USERDATA
] = {.type
= NLA_UNSPEC
},
2181 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT
] = {.type
= NLA_U32
},
2183 struct nlattr
*a
[OVS_USERSPACE_ATTR_MAX
+ 1];
2186 error
= nla_parse_nested(a
, OVS_USERSPACE_ATTR_MAX
,
2187 attr
, userspace_policy
);
2191 if (!a
[OVS_USERSPACE_ATTR_PID
] ||
2192 !nla_get_u32(a
[OVS_USERSPACE_ATTR_PID
]))
2198 static int copy_action(const struct nlattr
*from
,
2199 struct sw_flow_actions
**sfa
, bool log
)
2201 int totlen
= NLA_ALIGN(from
->nla_len
);
2204 to
= reserve_sfa_size(sfa
, from
->nla_len
, log
);
2208 memcpy(to
, from
, totlen
);
2212 static int __ovs_nla_copy_actions(struct net
*net
, const struct nlattr
*attr
,
2213 const struct sw_flow_key
*key
,
2214 int depth
, struct sw_flow_actions
**sfa
,
2215 __be16 eth_type
, __be16 vlan_tci
, bool log
)
2217 const struct nlattr
*a
;
2220 if (depth
>= SAMPLE_ACTION_DEPTH
)
2223 nla_for_each_nested(a
, attr
, rem
) {
2224 /* Expected argument lengths, (u32)-1 for variable length. */
2225 static const u32 action_lens
[OVS_ACTION_ATTR_MAX
+ 1] = {
2226 [OVS_ACTION_ATTR_OUTPUT
] = sizeof(u32
),
2227 [OVS_ACTION_ATTR_RECIRC
] = sizeof(u32
),
2228 [OVS_ACTION_ATTR_USERSPACE
] = (u32
)-1,
2229 [OVS_ACTION_ATTR_PUSH_MPLS
] = sizeof(struct ovs_action_push_mpls
),
2230 [OVS_ACTION_ATTR_POP_MPLS
] = sizeof(__be16
),
2231 [OVS_ACTION_ATTR_PUSH_VLAN
] = sizeof(struct ovs_action_push_vlan
),
2232 [OVS_ACTION_ATTR_POP_VLAN
] = 0,
2233 [OVS_ACTION_ATTR_SET
] = (u32
)-1,
2234 [OVS_ACTION_ATTR_SET_MASKED
] = (u32
)-1,
2235 [OVS_ACTION_ATTR_SAMPLE
] = (u32
)-1,
2236 [OVS_ACTION_ATTR_HASH
] = sizeof(struct ovs_action_hash
),
2237 [OVS_ACTION_ATTR_CT
] = (u32
)-1,
2238 [OVS_ACTION_ATTR_TRUNC
] = sizeof(struct ovs_action_trunc
),
2240 const struct ovs_action_push_vlan
*vlan
;
2241 int type
= nla_type(a
);
2244 if (type
> OVS_ACTION_ATTR_MAX
||
2245 (action_lens
[type
] != nla_len(a
) &&
2246 action_lens
[type
] != (u32
)-1))
2251 case OVS_ACTION_ATTR_UNSPEC
:
2254 case OVS_ACTION_ATTR_USERSPACE
:
2255 err
= validate_userspace(a
);
2260 case OVS_ACTION_ATTR_OUTPUT
:
2261 if (nla_get_u32(a
) >= DP_MAX_PORTS
)
2265 case OVS_ACTION_ATTR_TRUNC
: {
2266 const struct ovs_action_trunc
*trunc
= nla_data(a
);
2268 if (trunc
->max_len
< ETH_HLEN
)
2273 case OVS_ACTION_ATTR_HASH
: {
2274 const struct ovs_action_hash
*act_hash
= nla_data(a
);
2276 switch (act_hash
->hash_alg
) {
2277 case OVS_HASH_ALG_L4
:
2286 case OVS_ACTION_ATTR_POP_VLAN
:
2287 vlan_tci
= htons(0);
2290 case OVS_ACTION_ATTR_PUSH_VLAN
:
2292 if (vlan
->vlan_tpid
!= htons(ETH_P_8021Q
))
2294 if (!(vlan
->vlan_tci
& htons(VLAN_TAG_PRESENT
)))
2296 vlan_tci
= vlan
->vlan_tci
;
2299 case OVS_ACTION_ATTR_RECIRC
:
2302 case OVS_ACTION_ATTR_PUSH_MPLS
: {
2303 const struct ovs_action_push_mpls
*mpls
= nla_data(a
);
2305 if (!eth_p_mpls(mpls
->mpls_ethertype
))
2307 /* Prohibit push MPLS other than to a white list
2308 * for packets that have a known tag order.
2310 if (vlan_tci
& htons(VLAN_TAG_PRESENT
) ||
2311 (eth_type
!= htons(ETH_P_IP
) &&
2312 eth_type
!= htons(ETH_P_IPV6
) &&
2313 eth_type
!= htons(ETH_P_ARP
) &&
2314 eth_type
!= htons(ETH_P_RARP
) &&
2315 !eth_p_mpls(eth_type
)))
2317 eth_type
= mpls
->mpls_ethertype
;
2321 case OVS_ACTION_ATTR_POP_MPLS
:
2322 if (vlan_tci
& htons(VLAN_TAG_PRESENT
) ||
2323 !eth_p_mpls(eth_type
))
2326 /* Disallow subsequent L2.5+ set and mpls_pop actions
2327 * as there is no check here to ensure that the new
2328 * eth_type is valid and thus set actions could
2329 * write off the end of the packet or otherwise
2332 * Support for these actions is planned using packet
2335 eth_type
= htons(0);
2338 case OVS_ACTION_ATTR_SET
:
2339 err
= validate_set(a
, key
, sfa
,
2340 &skip_copy
, eth_type
, false, log
);
2345 case OVS_ACTION_ATTR_SET_MASKED
:
2346 err
= validate_set(a
, key
, sfa
,
2347 &skip_copy
, eth_type
, true, log
);
2352 case OVS_ACTION_ATTR_SAMPLE
:
2353 err
= validate_and_copy_sample(net
, a
, key
, depth
, sfa
,
2354 eth_type
, vlan_tci
, log
);
2360 case OVS_ACTION_ATTR_CT
:
2361 err
= ovs_ct_copy_action(net
, a
, key
, sfa
, log
);
2368 OVS_NLERR(log
, "Unknown Action type %d", type
);
2372 err
= copy_action(a
, sfa
, log
);
2384 /* 'key' must be the masked key. */
2385 int ovs_nla_copy_actions(struct net
*net
, const struct nlattr
*attr
,
2386 const struct sw_flow_key
*key
,
2387 struct sw_flow_actions
**sfa
, bool log
)
2391 *sfa
= nla_alloc_flow_actions(nla_len(attr
), log
);
2393 return PTR_ERR(*sfa
);
2395 (*sfa
)->orig_len
= nla_len(attr
);
2396 err
= __ovs_nla_copy_actions(net
, attr
, key
, 0, sfa
, key
->eth
.type
,
2399 ovs_nla_free_flow_actions(*sfa
);
2404 static int sample_action_to_attr(const struct nlattr
*attr
, struct sk_buff
*skb
)
2406 const struct nlattr
*a
;
2407 struct nlattr
*start
;
2410 start
= nla_nest_start(skb
, OVS_ACTION_ATTR_SAMPLE
);
2414 nla_for_each_nested(a
, attr
, rem
) {
2415 int type
= nla_type(a
);
2416 struct nlattr
*st_sample
;
2419 case OVS_SAMPLE_ATTR_PROBABILITY
:
2420 if (nla_put(skb
, OVS_SAMPLE_ATTR_PROBABILITY
,
2421 sizeof(u32
), nla_data(a
)))
2424 case OVS_SAMPLE_ATTR_ACTIONS
:
2425 st_sample
= nla_nest_start(skb
, OVS_SAMPLE_ATTR_ACTIONS
);
2428 err
= ovs_nla_put_actions(nla_data(a
), nla_len(a
), skb
);
2431 nla_nest_end(skb
, st_sample
);
2436 nla_nest_end(skb
, start
);
2440 static int set_action_to_attr(const struct nlattr
*a
, struct sk_buff
*skb
)
2442 const struct nlattr
*ovs_key
= nla_data(a
);
2443 int key_type
= nla_type(ovs_key
);
2444 struct nlattr
*start
;
2448 case OVS_KEY_ATTR_TUNNEL_INFO
: {
2449 struct ovs_tunnel_info
*ovs_tun
= nla_data(ovs_key
);
2450 struct ip_tunnel_info
*tun_info
= &ovs_tun
->tun_dst
->u
.tun_info
;
2452 start
= nla_nest_start(skb
, OVS_ACTION_ATTR_SET
);
2456 err
= ovs_nla_put_tunnel_info(skb
, tun_info
);
2459 nla_nest_end(skb
, start
);
2463 if (nla_put(skb
, OVS_ACTION_ATTR_SET
, nla_len(a
), ovs_key
))
2471 static int masked_set_action_to_set_action_attr(const struct nlattr
*a
,
2472 struct sk_buff
*skb
)
2474 const struct nlattr
*ovs_key
= nla_data(a
);
2476 size_t key_len
= nla_len(ovs_key
) / 2;
2478 /* Revert the conversion we did from a non-masked set action to
2479 * masked set action.
2481 nla
= nla_nest_start(skb
, OVS_ACTION_ATTR_SET
);
2485 if (nla_put(skb
, nla_type(ovs_key
), key_len
, nla_data(ovs_key
)))
2488 nla_nest_end(skb
, nla
);
2492 int ovs_nla_put_actions(const struct nlattr
*attr
, int len
, struct sk_buff
*skb
)
2494 const struct nlattr
*a
;
2497 nla_for_each_attr(a
, attr
, len
, rem
) {
2498 int type
= nla_type(a
);
2501 case OVS_ACTION_ATTR_SET
:
2502 err
= set_action_to_attr(a
, skb
);
2507 case OVS_ACTION_ATTR_SET_TO_MASKED
:
2508 err
= masked_set_action_to_set_action_attr(a
, skb
);
2513 case OVS_ACTION_ATTR_SAMPLE
:
2514 err
= sample_action_to_attr(a
, skb
);
2519 case OVS_ACTION_ATTR_CT
:
2520 err
= ovs_ct_action_to_attr(nla_data(a
), skb
);
2526 if (nla_put(skb
, type
, nla_len(a
), nla_data(a
)))