2 * Copyright (c) 2007-2014 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
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>
51 #include "flow_netlink.h"
52 #include "vport-vxlan.h"
56 const struct ovs_len_tbl
*next
;
59 #define OVS_ATTR_NESTED -1
61 static void update_range(struct sw_flow_match
*match
,
62 size_t offset
, size_t size
, bool is_mask
)
64 struct sw_flow_key_range
*range
;
65 size_t start
= rounddown(offset
, sizeof(long));
66 size_t end
= roundup(offset
+ size
, sizeof(long));
69 range
= &match
->range
;
71 range
= &match
->mask
->range
;
73 if (range
->start
== range
->end
) {
79 if (range
->start
> start
)
86 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
88 update_range(match, offsetof(struct sw_flow_key, field), \
89 sizeof((match)->key->field), is_mask); \
91 (match)->mask->key.field = value; \
93 (match)->key->field = value; \
96 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
98 update_range(match, offset, len, is_mask); \
100 memcpy((u8 *)&(match)->mask->key + offset, value_p, len);\
102 memcpy((u8 *)(match)->key + offset, value_p, len); \
105 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
106 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
107 value_p, len, is_mask)
109 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
111 update_range(match, offsetof(struct sw_flow_key, field), \
112 sizeof((match)->key->field), is_mask); \
114 memset((u8 *)&(match)->mask->key.field, value, \
115 sizeof((match)->mask->key.field)); \
117 memset((u8 *)&(match)->key->field, value, \
118 sizeof((match)->key->field)); \
121 static bool match_validate(const struct sw_flow_match
*match
,
122 u64 key_attrs
, u64 mask_attrs
, bool log
)
124 u64 key_expected
= 1ULL << OVS_KEY_ATTR_ETHERNET
;
125 u64 mask_allowed
= key_attrs
; /* At most allow all key attributes */
127 /* The following mask attributes allowed only if they
128 * pass the validation tests.
130 mask_allowed
&= ~((1ULL << OVS_KEY_ATTR_IPV4
)
131 | (1ULL << OVS_KEY_ATTR_IPV6
)
132 | (1ULL << OVS_KEY_ATTR_TCP
)
133 | (1ULL << OVS_KEY_ATTR_TCP_FLAGS
)
134 | (1ULL << OVS_KEY_ATTR_UDP
)
135 | (1ULL << OVS_KEY_ATTR_SCTP
)
136 | (1ULL << OVS_KEY_ATTR_ICMP
)
137 | (1ULL << OVS_KEY_ATTR_ICMPV6
)
138 | (1ULL << OVS_KEY_ATTR_ARP
)
139 | (1ULL << OVS_KEY_ATTR_ND
)
140 | (1ULL << OVS_KEY_ATTR_MPLS
));
142 /* Always allowed mask fields. */
143 mask_allowed
|= ((1ULL << OVS_KEY_ATTR_TUNNEL
)
144 | (1ULL << OVS_KEY_ATTR_IN_PORT
)
145 | (1ULL << OVS_KEY_ATTR_ETHERTYPE
));
147 /* Check key attributes. */
148 if (match
->key
->eth
.type
== htons(ETH_P_ARP
)
149 || match
->key
->eth
.type
== htons(ETH_P_RARP
)) {
150 key_expected
|= 1ULL << OVS_KEY_ATTR_ARP
;
151 if (match
->mask
&& (match
->mask
->key
.eth
.type
== htons(0xffff)))
152 mask_allowed
|= 1ULL << OVS_KEY_ATTR_ARP
;
155 if (eth_p_mpls(match
->key
->eth
.type
)) {
156 key_expected
|= 1ULL << OVS_KEY_ATTR_MPLS
;
157 if (match
->mask
&& (match
->mask
->key
.eth
.type
== htons(0xffff)))
158 mask_allowed
|= 1ULL << OVS_KEY_ATTR_MPLS
;
161 if (match
->key
->eth
.type
== htons(ETH_P_IP
)) {
162 key_expected
|= 1ULL << OVS_KEY_ATTR_IPV4
;
163 if (match
->mask
&& (match
->mask
->key
.eth
.type
== htons(0xffff)))
164 mask_allowed
|= 1ULL << OVS_KEY_ATTR_IPV4
;
166 if (match
->key
->ip
.frag
!= OVS_FRAG_TYPE_LATER
) {
167 if (match
->key
->ip
.proto
== IPPROTO_UDP
) {
168 key_expected
|= 1ULL << OVS_KEY_ATTR_UDP
;
169 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff))
170 mask_allowed
|= 1ULL << OVS_KEY_ATTR_UDP
;
173 if (match
->key
->ip
.proto
== IPPROTO_SCTP
) {
174 key_expected
|= 1ULL << OVS_KEY_ATTR_SCTP
;
175 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff))
176 mask_allowed
|= 1ULL << OVS_KEY_ATTR_SCTP
;
179 if (match
->key
->ip
.proto
== IPPROTO_TCP
) {
180 key_expected
|= 1ULL << OVS_KEY_ATTR_TCP
;
181 key_expected
|= 1ULL << OVS_KEY_ATTR_TCP_FLAGS
;
182 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff)) {
183 mask_allowed
|= 1ULL << OVS_KEY_ATTR_TCP
;
184 mask_allowed
|= 1ULL << OVS_KEY_ATTR_TCP_FLAGS
;
188 if (match
->key
->ip
.proto
== IPPROTO_ICMP
) {
189 key_expected
|= 1ULL << OVS_KEY_ATTR_ICMP
;
190 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff))
191 mask_allowed
|= 1ULL << OVS_KEY_ATTR_ICMP
;
196 if (match
->key
->eth
.type
== htons(ETH_P_IPV6
)) {
197 key_expected
|= 1ULL << OVS_KEY_ATTR_IPV6
;
198 if (match
->mask
&& (match
->mask
->key
.eth
.type
== htons(0xffff)))
199 mask_allowed
|= 1ULL << OVS_KEY_ATTR_IPV6
;
201 if (match
->key
->ip
.frag
!= OVS_FRAG_TYPE_LATER
) {
202 if (match
->key
->ip
.proto
== IPPROTO_UDP
) {
203 key_expected
|= 1ULL << OVS_KEY_ATTR_UDP
;
204 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff))
205 mask_allowed
|= 1ULL << OVS_KEY_ATTR_UDP
;
208 if (match
->key
->ip
.proto
== IPPROTO_SCTP
) {
209 key_expected
|= 1ULL << OVS_KEY_ATTR_SCTP
;
210 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff))
211 mask_allowed
|= 1ULL << OVS_KEY_ATTR_SCTP
;
214 if (match
->key
->ip
.proto
== IPPROTO_TCP
) {
215 key_expected
|= 1ULL << OVS_KEY_ATTR_TCP
;
216 key_expected
|= 1ULL << OVS_KEY_ATTR_TCP_FLAGS
;
217 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff)) {
218 mask_allowed
|= 1ULL << OVS_KEY_ATTR_TCP
;
219 mask_allowed
|= 1ULL << OVS_KEY_ATTR_TCP_FLAGS
;
223 if (match
->key
->ip
.proto
== IPPROTO_ICMPV6
) {
224 key_expected
|= 1ULL << OVS_KEY_ATTR_ICMPV6
;
225 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff))
226 mask_allowed
|= 1ULL << OVS_KEY_ATTR_ICMPV6
;
228 if (match
->key
->tp
.src
==
229 htons(NDISC_NEIGHBOUR_SOLICITATION
) ||
230 match
->key
->tp
.src
== htons(NDISC_NEIGHBOUR_ADVERTISEMENT
)) {
231 key_expected
|= 1ULL << OVS_KEY_ATTR_ND
;
232 if (match
->mask
&& (match
->mask
->key
.tp
.src
== htons(0xff)))
233 mask_allowed
|= 1ULL << OVS_KEY_ATTR_ND
;
239 if ((key_attrs
& key_expected
) != key_expected
) {
240 /* Key attributes check failed. */
241 OVS_NLERR(log
, "Missing key (keys=%llx, expected=%llx)",
242 (unsigned long long)key_attrs
,
243 (unsigned long long)key_expected
);
247 if ((mask_attrs
& mask_allowed
) != mask_attrs
) {
248 /* Mask attributes check failed. */
249 OVS_NLERR(log
, "Unexpected mask (mask=%llx, allowed=%llx)",
250 (unsigned long long)mask_attrs
,
251 (unsigned long long)mask_allowed
);
258 size_t ovs_tun_key_attr_size(void)
260 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
261 * updating this function.
263 return nla_total_size(8) /* OVS_TUNNEL_KEY_ATTR_ID */
264 + nla_total_size(4) /* OVS_TUNNEL_KEY_ATTR_IPV4_SRC */
265 + nla_total_size(4) /* OVS_TUNNEL_KEY_ATTR_IPV4_DST */
266 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
267 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
268 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
269 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
270 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
271 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
272 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS is mutually exclusive with
273 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
275 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
276 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
279 size_t ovs_key_attr_size(void)
281 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
282 * updating this function.
284 BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO
!= 22);
286 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
287 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
288 + ovs_tun_key_attr_size()
289 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
290 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
291 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
292 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
293 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
294 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
295 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
296 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
297 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
298 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
299 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
300 + nla_total_size(28); /* OVS_KEY_ATTR_ND */
303 static const struct ovs_len_tbl ovs_tunnel_key_lens
[OVS_TUNNEL_KEY_ATTR_MAX
+ 1] = {
304 [OVS_TUNNEL_KEY_ATTR_ID
] = { .len
= sizeof(u64
) },
305 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC
] = { .len
= sizeof(u32
) },
306 [OVS_TUNNEL_KEY_ATTR_IPV4_DST
] = { .len
= sizeof(u32
) },
307 [OVS_TUNNEL_KEY_ATTR_TOS
] = { .len
= 1 },
308 [OVS_TUNNEL_KEY_ATTR_TTL
] = { .len
= 1 },
309 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT
] = { .len
= 0 },
310 [OVS_TUNNEL_KEY_ATTR_CSUM
] = { .len
= 0 },
311 [OVS_TUNNEL_KEY_ATTR_TP_SRC
] = { .len
= sizeof(u16
) },
312 [OVS_TUNNEL_KEY_ATTR_TP_DST
] = { .len
= sizeof(u16
) },
313 [OVS_TUNNEL_KEY_ATTR_OAM
] = { .len
= 0 },
314 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS
] = { .len
= OVS_ATTR_NESTED
},
315 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS
] = { .len
= OVS_ATTR_NESTED
},
318 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
319 static const struct ovs_len_tbl ovs_key_lens
[OVS_KEY_ATTR_MAX
+ 1] = {
320 [OVS_KEY_ATTR_ENCAP
] = { .len
= OVS_ATTR_NESTED
},
321 [OVS_KEY_ATTR_PRIORITY
] = { .len
= sizeof(u32
) },
322 [OVS_KEY_ATTR_IN_PORT
] = { .len
= sizeof(u32
) },
323 [OVS_KEY_ATTR_SKB_MARK
] = { .len
= sizeof(u32
) },
324 [OVS_KEY_ATTR_ETHERNET
] = { .len
= sizeof(struct ovs_key_ethernet
) },
325 [OVS_KEY_ATTR_VLAN
] = { .len
= sizeof(__be16
) },
326 [OVS_KEY_ATTR_ETHERTYPE
] = { .len
= sizeof(__be16
) },
327 [OVS_KEY_ATTR_IPV4
] = { .len
= sizeof(struct ovs_key_ipv4
) },
328 [OVS_KEY_ATTR_IPV6
] = { .len
= sizeof(struct ovs_key_ipv6
) },
329 [OVS_KEY_ATTR_TCP
] = { .len
= sizeof(struct ovs_key_tcp
) },
330 [OVS_KEY_ATTR_TCP_FLAGS
] = { .len
= sizeof(__be16
) },
331 [OVS_KEY_ATTR_UDP
] = { .len
= sizeof(struct ovs_key_udp
) },
332 [OVS_KEY_ATTR_SCTP
] = { .len
= sizeof(struct ovs_key_sctp
) },
333 [OVS_KEY_ATTR_ICMP
] = { .len
= sizeof(struct ovs_key_icmp
) },
334 [OVS_KEY_ATTR_ICMPV6
] = { .len
= sizeof(struct ovs_key_icmpv6
) },
335 [OVS_KEY_ATTR_ARP
] = { .len
= sizeof(struct ovs_key_arp
) },
336 [OVS_KEY_ATTR_ND
] = { .len
= sizeof(struct ovs_key_nd
) },
337 [OVS_KEY_ATTR_RECIRC_ID
] = { .len
= sizeof(u32
) },
338 [OVS_KEY_ATTR_DP_HASH
] = { .len
= sizeof(u32
) },
339 [OVS_KEY_ATTR_TUNNEL
] = { .len
= OVS_ATTR_NESTED
,
340 .next
= ovs_tunnel_key_lens
, },
341 [OVS_KEY_ATTR_MPLS
] = { .len
= sizeof(struct ovs_key_mpls
) },
343 static bool is_all_zero(const u8
*fp
, size_t size
)
350 for (i
= 0; i
< size
; i
++)
357 static int __parse_flow_nlattrs(const struct nlattr
*attr
,
358 const struct nlattr
*a
[],
359 u64
*attrsp
, bool log
, bool nz
)
361 const struct nlattr
*nla
;
366 nla_for_each_nested(nla
, attr
, rem
) {
367 u16 type
= nla_type(nla
);
370 if (type
> OVS_KEY_ATTR_MAX
) {
371 OVS_NLERR(log
, "Key type %d is out of range max %d",
372 type
, OVS_KEY_ATTR_MAX
);
376 if (attrs
& (1ULL << type
)) {
377 OVS_NLERR(log
, "Duplicate key (type %d).", type
);
381 expected_len
= ovs_key_lens
[type
].len
;
382 if (nla_len(nla
) != expected_len
&& expected_len
!= OVS_ATTR_NESTED
) {
383 OVS_NLERR(log
, "Key %d has unexpected len %d expected %d",
384 type
, nla_len(nla
), expected_len
);
388 if (!nz
|| !is_all_zero(nla_data(nla
), expected_len
)) {
389 attrs
|= 1ULL << type
;
394 OVS_NLERR(log
, "Message has %d unknown bytes.", rem
);
402 static int parse_flow_mask_nlattrs(const struct nlattr
*attr
,
403 const struct nlattr
*a
[], u64
*attrsp
,
406 return __parse_flow_nlattrs(attr
, a
, attrsp
, log
, true);
409 static int parse_flow_nlattrs(const struct nlattr
*attr
,
410 const struct nlattr
*a
[], u64
*attrsp
,
413 return __parse_flow_nlattrs(attr
, a
, attrsp
, log
, false);
416 static int genev_tun_opt_from_nlattr(const struct nlattr
*a
,
417 struct sw_flow_match
*match
, bool is_mask
,
420 unsigned long opt_key_offset
;
422 if (nla_len(a
) > sizeof(match
->key
->tun_opts
)) {
423 OVS_NLERR(log
, "Geneve option length err (len %d, max %zu).",
424 nla_len(a
), sizeof(match
->key
->tun_opts
));
428 if (nla_len(a
) % 4 != 0) {
429 OVS_NLERR(log
, "Geneve opt len %d is not a multiple of 4.",
434 /* We need to record the length of the options passed
435 * down, otherwise packets with the same format but
436 * additional options will be silently matched.
439 SW_FLOW_KEY_PUT(match
, tun_opts_len
, nla_len(a
),
442 /* This is somewhat unusual because it looks at
443 * both the key and mask while parsing the
444 * attributes (and by extension assumes the key
445 * is parsed first). Normally, we would verify
446 * that each is the correct length and that the
447 * attributes line up in the validate function.
448 * However, that is difficult because this is
449 * variable length and we won't have the
452 if (match
->key
->tun_opts_len
!= nla_len(a
)) {
453 OVS_NLERR(log
, "Geneve option len %d != mask len %d",
454 match
->key
->tun_opts_len
, nla_len(a
));
458 SW_FLOW_KEY_PUT(match
, tun_opts_len
, 0xff, true);
461 opt_key_offset
= TUN_METADATA_OFFSET(nla_len(a
));
462 SW_FLOW_KEY_MEMCPY_OFFSET(match
, opt_key_offset
, nla_data(a
),
463 nla_len(a
), is_mask
);
467 static const struct nla_policy vxlan_opt_policy
[OVS_VXLAN_EXT_MAX
+ 1] = {
468 [OVS_VXLAN_EXT_GBP
] = { .type
= NLA_U32
},
471 static int vxlan_tun_opt_from_nlattr(const struct nlattr
*a
,
472 struct sw_flow_match
*match
, bool is_mask
,
475 struct nlattr
*tb
[OVS_VXLAN_EXT_MAX
+1];
476 unsigned long opt_key_offset
;
477 struct ovs_vxlan_opts opts
;
480 BUILD_BUG_ON(sizeof(opts
) > sizeof(match
->key
->tun_opts
));
482 err
= nla_parse_nested(tb
, OVS_VXLAN_EXT_MAX
, a
, vxlan_opt_policy
);
486 memset(&opts
, 0, sizeof(opts
));
488 if (tb
[OVS_VXLAN_EXT_GBP
])
489 opts
.gbp
= nla_get_u32(tb
[OVS_VXLAN_EXT_GBP
]);
492 SW_FLOW_KEY_PUT(match
, tun_opts_len
, sizeof(opts
), false);
494 SW_FLOW_KEY_PUT(match
, tun_opts_len
, 0xff, true);
496 opt_key_offset
= TUN_METADATA_OFFSET(sizeof(opts
));
497 SW_FLOW_KEY_MEMCPY_OFFSET(match
, opt_key_offset
, &opts
, sizeof(opts
),
502 static int ipv4_tun_from_nlattr(const struct nlattr
*attr
,
503 struct sw_flow_match
*match
, bool is_mask
,
509 __be16 tun_flags
= 0;
512 nla_for_each_nested(a
, attr
, rem
) {
513 int type
= nla_type(a
);
516 if (type
> OVS_TUNNEL_KEY_ATTR_MAX
) {
517 OVS_NLERR(log
, "Tunnel attr %d out of range max %d",
518 type
, OVS_TUNNEL_KEY_ATTR_MAX
);
522 if (ovs_tunnel_key_lens
[type
].len
!= nla_len(a
) &&
523 ovs_tunnel_key_lens
[type
].len
!= OVS_ATTR_NESTED
) {
524 OVS_NLERR(log
, "Tunnel attr %d has unexpected len %d expected %d",
525 type
, nla_len(a
), ovs_tunnel_key_lens
[type
].len
);
530 case OVS_TUNNEL_KEY_ATTR_ID
:
531 SW_FLOW_KEY_PUT(match
, tun_key
.tun_id
,
532 nla_get_be64(a
), is_mask
);
533 tun_flags
|= TUNNEL_KEY
;
535 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC
:
536 SW_FLOW_KEY_PUT(match
, tun_key
.ipv4_src
,
537 nla_get_be32(a
), is_mask
);
539 case OVS_TUNNEL_KEY_ATTR_IPV4_DST
:
540 SW_FLOW_KEY_PUT(match
, tun_key
.ipv4_dst
,
541 nla_get_be32(a
), is_mask
);
543 case OVS_TUNNEL_KEY_ATTR_TOS
:
544 SW_FLOW_KEY_PUT(match
, tun_key
.ipv4_tos
,
545 nla_get_u8(a
), is_mask
);
547 case OVS_TUNNEL_KEY_ATTR_TTL
:
548 SW_FLOW_KEY_PUT(match
, tun_key
.ipv4_ttl
,
549 nla_get_u8(a
), is_mask
);
552 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT
:
553 tun_flags
|= TUNNEL_DONT_FRAGMENT
;
555 case OVS_TUNNEL_KEY_ATTR_CSUM
:
556 tun_flags
|= TUNNEL_CSUM
;
558 case OVS_TUNNEL_KEY_ATTR_TP_SRC
:
559 SW_FLOW_KEY_PUT(match
, tun_key
.tp_src
,
560 nla_get_be16(a
), is_mask
);
562 case OVS_TUNNEL_KEY_ATTR_TP_DST
:
563 SW_FLOW_KEY_PUT(match
, tun_key
.tp_dst
,
564 nla_get_be16(a
), is_mask
);
566 case OVS_TUNNEL_KEY_ATTR_OAM
:
567 tun_flags
|= TUNNEL_OAM
;
569 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS
:
571 OVS_NLERR(log
, "Multiple metadata blocks provided");
575 err
= genev_tun_opt_from_nlattr(a
, match
, is_mask
, log
);
579 tun_flags
|= TUNNEL_GENEVE_OPT
;
582 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS
:
584 OVS_NLERR(log
, "Multiple metadata blocks provided");
588 err
= vxlan_tun_opt_from_nlattr(a
, match
, is_mask
, log
);
592 tun_flags
|= TUNNEL_VXLAN_OPT
;
596 OVS_NLERR(log
, "Unknown IPv4 tunnel attribute %d",
602 SW_FLOW_KEY_PUT(match
, tun_key
.tun_flags
, tun_flags
, is_mask
);
605 OVS_NLERR(log
, "IPv4 tunnel attribute has %d unknown bytes.",
611 if (!match
->key
->tun_key
.ipv4_dst
) {
612 OVS_NLERR(log
, "IPv4 tunnel dst address is zero");
617 OVS_NLERR(log
, "IPv4 tunnel TTL not specified.");
625 static int vxlan_opt_to_nlattr(struct sk_buff
*skb
,
626 const void *tun_opts
, int swkey_tun_opts_len
)
628 const struct ovs_vxlan_opts
*opts
= tun_opts
;
631 nla
= nla_nest_start(skb
, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS
);
635 if (nla_put_u32(skb
, OVS_VXLAN_EXT_GBP
, opts
->gbp
) < 0)
638 nla_nest_end(skb
, nla
);
642 static int __ipv4_tun_to_nlattr(struct sk_buff
*skb
,
643 const struct ovs_key_ipv4_tunnel
*output
,
644 const void *tun_opts
, int swkey_tun_opts_len
)
646 if (output
->tun_flags
& TUNNEL_KEY
&&
647 nla_put_be64(skb
, OVS_TUNNEL_KEY_ATTR_ID
, output
->tun_id
))
649 if (output
->ipv4_src
&&
650 nla_put_be32(skb
, OVS_TUNNEL_KEY_ATTR_IPV4_SRC
, output
->ipv4_src
))
652 if (output
->ipv4_dst
&&
653 nla_put_be32(skb
, OVS_TUNNEL_KEY_ATTR_IPV4_DST
, output
->ipv4_dst
))
655 if (output
->ipv4_tos
&&
656 nla_put_u8(skb
, OVS_TUNNEL_KEY_ATTR_TOS
, output
->ipv4_tos
))
658 if (nla_put_u8(skb
, OVS_TUNNEL_KEY_ATTR_TTL
, output
->ipv4_ttl
))
660 if ((output
->tun_flags
& TUNNEL_DONT_FRAGMENT
) &&
661 nla_put_flag(skb
, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT
))
663 if ((output
->tun_flags
& TUNNEL_CSUM
) &&
664 nla_put_flag(skb
, OVS_TUNNEL_KEY_ATTR_CSUM
))
666 if (output
->tp_src
&&
667 nla_put_be16(skb
, OVS_TUNNEL_KEY_ATTR_TP_SRC
, output
->tp_src
))
669 if (output
->tp_dst
&&
670 nla_put_be16(skb
, OVS_TUNNEL_KEY_ATTR_TP_DST
, output
->tp_dst
))
672 if ((output
->tun_flags
& TUNNEL_OAM
) &&
673 nla_put_flag(skb
, OVS_TUNNEL_KEY_ATTR_OAM
))
676 if (output
->tun_flags
& TUNNEL_GENEVE_OPT
&&
677 nla_put(skb
, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS
,
678 swkey_tun_opts_len
, tun_opts
))
680 else if (output
->tun_flags
& TUNNEL_VXLAN_OPT
&&
681 vxlan_opt_to_nlattr(skb
, tun_opts
, swkey_tun_opts_len
))
688 static int ipv4_tun_to_nlattr(struct sk_buff
*skb
,
689 const struct ovs_key_ipv4_tunnel
*output
,
690 const void *tun_opts
, int swkey_tun_opts_len
)
695 nla
= nla_nest_start(skb
, OVS_KEY_ATTR_TUNNEL
);
699 err
= __ipv4_tun_to_nlattr(skb
, output
, tun_opts
, swkey_tun_opts_len
);
703 nla_nest_end(skb
, nla
);
707 int ovs_nla_put_egress_tunnel_key(struct sk_buff
*skb
,
708 const struct ovs_tunnel_info
*egress_tun_info
)
710 return __ipv4_tun_to_nlattr(skb
, &egress_tun_info
->tunnel
,
711 egress_tun_info
->options
,
712 egress_tun_info
->options_len
);
715 static int metadata_from_nlattrs(struct sw_flow_match
*match
, u64
*attrs
,
716 const struct nlattr
**a
, bool is_mask
,
719 if (*attrs
& (1ULL << OVS_KEY_ATTR_DP_HASH
)) {
720 u32 hash_val
= nla_get_u32(a
[OVS_KEY_ATTR_DP_HASH
]);
722 SW_FLOW_KEY_PUT(match
, ovs_flow_hash
, hash_val
, is_mask
);
723 *attrs
&= ~(1ULL << OVS_KEY_ATTR_DP_HASH
);
726 if (*attrs
& (1ULL << OVS_KEY_ATTR_RECIRC_ID
)) {
727 u32 recirc_id
= nla_get_u32(a
[OVS_KEY_ATTR_RECIRC_ID
]);
729 SW_FLOW_KEY_PUT(match
, recirc_id
, recirc_id
, is_mask
);
730 *attrs
&= ~(1ULL << OVS_KEY_ATTR_RECIRC_ID
);
733 if (*attrs
& (1ULL << OVS_KEY_ATTR_PRIORITY
)) {
734 SW_FLOW_KEY_PUT(match
, phy
.priority
,
735 nla_get_u32(a
[OVS_KEY_ATTR_PRIORITY
]), is_mask
);
736 *attrs
&= ~(1ULL << OVS_KEY_ATTR_PRIORITY
);
739 if (*attrs
& (1ULL << OVS_KEY_ATTR_IN_PORT
)) {
740 u32 in_port
= nla_get_u32(a
[OVS_KEY_ATTR_IN_PORT
]);
743 in_port
= 0xffffffff; /* Always exact match in_port. */
744 } else if (in_port
>= DP_MAX_PORTS
) {
745 OVS_NLERR(log
, "Port %d exceeds max allowable %d",
746 in_port
, DP_MAX_PORTS
);
750 SW_FLOW_KEY_PUT(match
, phy
.in_port
, in_port
, is_mask
);
751 *attrs
&= ~(1ULL << OVS_KEY_ATTR_IN_PORT
);
752 } else if (!is_mask
) {
753 SW_FLOW_KEY_PUT(match
, phy
.in_port
, DP_MAX_PORTS
, is_mask
);
756 if (*attrs
& (1ULL << OVS_KEY_ATTR_SKB_MARK
)) {
757 uint32_t mark
= nla_get_u32(a
[OVS_KEY_ATTR_SKB_MARK
]);
759 SW_FLOW_KEY_PUT(match
, phy
.skb_mark
, mark
, is_mask
);
760 *attrs
&= ~(1ULL << OVS_KEY_ATTR_SKB_MARK
);
762 if (*attrs
& (1ULL << OVS_KEY_ATTR_TUNNEL
)) {
763 if (ipv4_tun_from_nlattr(a
[OVS_KEY_ATTR_TUNNEL
], match
,
766 *attrs
&= ~(1ULL << OVS_KEY_ATTR_TUNNEL
);
771 static int ovs_key_from_nlattrs(struct sw_flow_match
*match
, u64 attrs
,
772 const struct nlattr
**a
, bool is_mask
,
777 err
= metadata_from_nlattrs(match
, &attrs
, a
, is_mask
, log
);
781 if (attrs
& (1ULL << OVS_KEY_ATTR_ETHERNET
)) {
782 const struct ovs_key_ethernet
*eth_key
;
784 eth_key
= nla_data(a
[OVS_KEY_ATTR_ETHERNET
]);
785 SW_FLOW_KEY_MEMCPY(match
, eth
.src
,
786 eth_key
->eth_src
, ETH_ALEN
, is_mask
);
787 SW_FLOW_KEY_MEMCPY(match
, eth
.dst
,
788 eth_key
->eth_dst
, ETH_ALEN
, is_mask
);
789 attrs
&= ~(1ULL << OVS_KEY_ATTR_ETHERNET
);
792 if (attrs
& (1ULL << OVS_KEY_ATTR_VLAN
)) {
795 tci
= nla_get_be16(a
[OVS_KEY_ATTR_VLAN
]);
796 if (!(tci
& htons(VLAN_TAG_PRESENT
))) {
798 OVS_NLERR(log
, "VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.");
800 OVS_NLERR(log
, "VLAN TCI does not have VLAN_TAG_PRESENT bit set.");
805 SW_FLOW_KEY_PUT(match
, eth
.tci
, tci
, is_mask
);
806 attrs
&= ~(1ULL << OVS_KEY_ATTR_VLAN
);
809 if (attrs
& (1ULL << OVS_KEY_ATTR_ETHERTYPE
)) {
812 eth_type
= nla_get_be16(a
[OVS_KEY_ATTR_ETHERTYPE
]);
814 /* Always exact match EtherType. */
815 eth_type
= htons(0xffff);
816 } else if (ntohs(eth_type
) < ETH_P_802_3_MIN
) {
817 OVS_NLERR(log
, "EtherType %x is less than min %x",
818 ntohs(eth_type
), ETH_P_802_3_MIN
);
822 SW_FLOW_KEY_PUT(match
, eth
.type
, eth_type
, is_mask
);
823 attrs
&= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE
);
824 } else if (!is_mask
) {
825 SW_FLOW_KEY_PUT(match
, eth
.type
, htons(ETH_P_802_2
), is_mask
);
828 if (attrs
& (1ULL << OVS_KEY_ATTR_IPV4
)) {
829 const struct ovs_key_ipv4
*ipv4_key
;
831 ipv4_key
= nla_data(a
[OVS_KEY_ATTR_IPV4
]);
832 if (!is_mask
&& ipv4_key
->ipv4_frag
> OVS_FRAG_TYPE_MAX
) {
833 OVS_NLERR(log
, "IPv4 frag type %d is out of range max %d",
834 ipv4_key
->ipv4_frag
, OVS_FRAG_TYPE_MAX
);
837 SW_FLOW_KEY_PUT(match
, ip
.proto
,
838 ipv4_key
->ipv4_proto
, is_mask
);
839 SW_FLOW_KEY_PUT(match
, ip
.tos
,
840 ipv4_key
->ipv4_tos
, is_mask
);
841 SW_FLOW_KEY_PUT(match
, ip
.ttl
,
842 ipv4_key
->ipv4_ttl
, is_mask
);
843 SW_FLOW_KEY_PUT(match
, ip
.frag
,
844 ipv4_key
->ipv4_frag
, is_mask
);
845 SW_FLOW_KEY_PUT(match
, ipv4
.addr
.src
,
846 ipv4_key
->ipv4_src
, is_mask
);
847 SW_FLOW_KEY_PUT(match
, ipv4
.addr
.dst
,
848 ipv4_key
->ipv4_dst
, is_mask
);
849 attrs
&= ~(1ULL << OVS_KEY_ATTR_IPV4
);
852 if (attrs
& (1ULL << OVS_KEY_ATTR_IPV6
)) {
853 const struct ovs_key_ipv6
*ipv6_key
;
855 ipv6_key
= nla_data(a
[OVS_KEY_ATTR_IPV6
]);
856 if (!is_mask
&& ipv6_key
->ipv6_frag
> OVS_FRAG_TYPE_MAX
) {
857 OVS_NLERR(log
, "IPv6 frag type %d is out of range max %d",
858 ipv6_key
->ipv6_frag
, OVS_FRAG_TYPE_MAX
);
861 if (!is_mask
&& ipv6_key
->ipv6_label
& htonl(0xFFF00000)) {
863 "Invalid IPv6 flow label value (value=%x, max=%x).",
864 ntohl(ipv6_key
->ipv6_label
), (1 << 20) - 1);
867 SW_FLOW_KEY_PUT(match
, ipv6
.label
,
868 ipv6_key
->ipv6_label
, is_mask
);
869 SW_FLOW_KEY_PUT(match
, ip
.proto
,
870 ipv6_key
->ipv6_proto
, is_mask
);
871 SW_FLOW_KEY_PUT(match
, ip
.tos
,
872 ipv6_key
->ipv6_tclass
, is_mask
);
873 SW_FLOW_KEY_PUT(match
, ip
.ttl
,
874 ipv6_key
->ipv6_hlimit
, is_mask
);
875 SW_FLOW_KEY_PUT(match
, ip
.frag
,
876 ipv6_key
->ipv6_frag
, is_mask
);
877 SW_FLOW_KEY_MEMCPY(match
, ipv6
.addr
.src
,
879 sizeof(match
->key
->ipv6
.addr
.src
),
881 SW_FLOW_KEY_MEMCPY(match
, ipv6
.addr
.dst
,
883 sizeof(match
->key
->ipv6
.addr
.dst
),
886 attrs
&= ~(1ULL << OVS_KEY_ATTR_IPV6
);
889 if (attrs
& (1ULL << OVS_KEY_ATTR_ARP
)) {
890 const struct ovs_key_arp
*arp_key
;
892 arp_key
= nla_data(a
[OVS_KEY_ATTR_ARP
]);
893 if (!is_mask
&& (arp_key
->arp_op
& htons(0xff00))) {
894 OVS_NLERR(log
, "Unknown ARP opcode (opcode=%d).",
899 SW_FLOW_KEY_PUT(match
, ipv4
.addr
.src
,
900 arp_key
->arp_sip
, is_mask
);
901 SW_FLOW_KEY_PUT(match
, ipv4
.addr
.dst
,
902 arp_key
->arp_tip
, is_mask
);
903 SW_FLOW_KEY_PUT(match
, ip
.proto
,
904 ntohs(arp_key
->arp_op
), is_mask
);
905 SW_FLOW_KEY_MEMCPY(match
, ipv4
.arp
.sha
,
906 arp_key
->arp_sha
, ETH_ALEN
, is_mask
);
907 SW_FLOW_KEY_MEMCPY(match
, ipv4
.arp
.tha
,
908 arp_key
->arp_tha
, ETH_ALEN
, is_mask
);
910 attrs
&= ~(1ULL << OVS_KEY_ATTR_ARP
);
913 if (attrs
& (1ULL << OVS_KEY_ATTR_MPLS
)) {
914 const struct ovs_key_mpls
*mpls_key
;
916 mpls_key
= nla_data(a
[OVS_KEY_ATTR_MPLS
]);
917 SW_FLOW_KEY_PUT(match
, mpls
.top_lse
,
918 mpls_key
->mpls_lse
, is_mask
);
920 attrs
&= ~(1ULL << OVS_KEY_ATTR_MPLS
);
923 if (attrs
& (1ULL << OVS_KEY_ATTR_TCP
)) {
924 const struct ovs_key_tcp
*tcp_key
;
926 tcp_key
= nla_data(a
[OVS_KEY_ATTR_TCP
]);
927 SW_FLOW_KEY_PUT(match
, tp
.src
, tcp_key
->tcp_src
, is_mask
);
928 SW_FLOW_KEY_PUT(match
, tp
.dst
, tcp_key
->tcp_dst
, is_mask
);
929 attrs
&= ~(1ULL << OVS_KEY_ATTR_TCP
);
932 if (attrs
& (1ULL << OVS_KEY_ATTR_TCP_FLAGS
)) {
933 SW_FLOW_KEY_PUT(match
, tp
.flags
,
934 nla_get_be16(a
[OVS_KEY_ATTR_TCP_FLAGS
]),
936 attrs
&= ~(1ULL << OVS_KEY_ATTR_TCP_FLAGS
);
939 if (attrs
& (1ULL << OVS_KEY_ATTR_UDP
)) {
940 const struct ovs_key_udp
*udp_key
;
942 udp_key
= nla_data(a
[OVS_KEY_ATTR_UDP
]);
943 SW_FLOW_KEY_PUT(match
, tp
.src
, udp_key
->udp_src
, is_mask
);
944 SW_FLOW_KEY_PUT(match
, tp
.dst
, udp_key
->udp_dst
, is_mask
);
945 attrs
&= ~(1ULL << OVS_KEY_ATTR_UDP
);
948 if (attrs
& (1ULL << OVS_KEY_ATTR_SCTP
)) {
949 const struct ovs_key_sctp
*sctp_key
;
951 sctp_key
= nla_data(a
[OVS_KEY_ATTR_SCTP
]);
952 SW_FLOW_KEY_PUT(match
, tp
.src
, sctp_key
->sctp_src
, is_mask
);
953 SW_FLOW_KEY_PUT(match
, tp
.dst
, sctp_key
->sctp_dst
, is_mask
);
954 attrs
&= ~(1ULL << OVS_KEY_ATTR_SCTP
);
957 if (attrs
& (1ULL << OVS_KEY_ATTR_ICMP
)) {
958 const struct ovs_key_icmp
*icmp_key
;
960 icmp_key
= nla_data(a
[OVS_KEY_ATTR_ICMP
]);
961 SW_FLOW_KEY_PUT(match
, tp
.src
,
962 htons(icmp_key
->icmp_type
), is_mask
);
963 SW_FLOW_KEY_PUT(match
, tp
.dst
,
964 htons(icmp_key
->icmp_code
), is_mask
);
965 attrs
&= ~(1ULL << OVS_KEY_ATTR_ICMP
);
968 if (attrs
& (1ULL << OVS_KEY_ATTR_ICMPV6
)) {
969 const struct ovs_key_icmpv6
*icmpv6_key
;
971 icmpv6_key
= nla_data(a
[OVS_KEY_ATTR_ICMPV6
]);
972 SW_FLOW_KEY_PUT(match
, tp
.src
,
973 htons(icmpv6_key
->icmpv6_type
), is_mask
);
974 SW_FLOW_KEY_PUT(match
, tp
.dst
,
975 htons(icmpv6_key
->icmpv6_code
), is_mask
);
976 attrs
&= ~(1ULL << OVS_KEY_ATTR_ICMPV6
);
979 if (attrs
& (1ULL << OVS_KEY_ATTR_ND
)) {
980 const struct ovs_key_nd
*nd_key
;
982 nd_key
= nla_data(a
[OVS_KEY_ATTR_ND
]);
983 SW_FLOW_KEY_MEMCPY(match
, ipv6
.nd
.target
,
985 sizeof(match
->key
->ipv6
.nd
.target
),
987 SW_FLOW_KEY_MEMCPY(match
, ipv6
.nd
.sll
,
988 nd_key
->nd_sll
, ETH_ALEN
, is_mask
);
989 SW_FLOW_KEY_MEMCPY(match
, ipv6
.nd
.tll
,
990 nd_key
->nd_tll
, ETH_ALEN
, is_mask
);
991 attrs
&= ~(1ULL << OVS_KEY_ATTR_ND
);
995 OVS_NLERR(log
, "Unknown key attributes %llx",
996 (unsigned long long)attrs
);
1003 static void nlattr_set(struct nlattr
*attr
, u8 val
,
1004 const struct ovs_len_tbl
*tbl
)
1009 /* The nlattr stream should already have been validated */
1010 nla_for_each_nested(nla
, attr
, rem
) {
1011 if (tbl
&& tbl
[nla_type(nla
)].len
== OVS_ATTR_NESTED
)
1012 nlattr_set(nla
, val
, tbl
[nla_type(nla
)].next
);
1014 memset(nla_data(nla
), val
, nla_len(nla
));
1018 static void mask_set_nlattr(struct nlattr
*attr
, u8 val
)
1020 nlattr_set(attr
, val
, ovs_key_lens
);
1024 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1025 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1026 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1027 * does not include any don't care bit.
1028 * @match: receives the extracted flow match information.
1029 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1030 * sequence. The fields should of the packet that triggered the creation
1032 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1033 * attribute specifies the mask field of the wildcarded flow.
1034 * @log: Boolean to allow kernel error logging. Normally true, but when
1035 * probing for feature compatibility this should be passed in as false to
1036 * suppress unnecessary error logging.
1038 int ovs_nla_get_match(struct sw_flow_match
*match
,
1039 const struct nlattr
*nla_key
,
1040 const struct nlattr
*nla_mask
,
1043 const struct nlattr
*a
[OVS_KEY_ATTR_MAX
+ 1];
1044 const struct nlattr
*encap
;
1045 struct nlattr
*newmask
= NULL
;
1048 bool encap_valid
= false;
1051 err
= parse_flow_nlattrs(nla_key
, a
, &key_attrs
, log
);
1055 if ((key_attrs
& (1ULL << OVS_KEY_ATTR_ETHERNET
)) &&
1056 (key_attrs
& (1ULL << OVS_KEY_ATTR_ETHERTYPE
)) &&
1057 (nla_get_be16(a
[OVS_KEY_ATTR_ETHERTYPE
]) == htons(ETH_P_8021Q
))) {
1060 if (!((key_attrs
& (1ULL << OVS_KEY_ATTR_VLAN
)) &&
1061 (key_attrs
& (1ULL << OVS_KEY_ATTR_ENCAP
)))) {
1062 OVS_NLERR(log
, "Invalid Vlan frame.");
1066 key_attrs
&= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE
);
1067 tci
= nla_get_be16(a
[OVS_KEY_ATTR_VLAN
]);
1068 encap
= a
[OVS_KEY_ATTR_ENCAP
];
1069 key_attrs
&= ~(1ULL << OVS_KEY_ATTR_ENCAP
);
1072 if (tci
& htons(VLAN_TAG_PRESENT
)) {
1073 err
= parse_flow_nlattrs(encap
, a
, &key_attrs
, log
);
1077 /* Corner case for truncated 802.1Q header. */
1078 if (nla_len(encap
)) {
1079 OVS_NLERR(log
, "Truncated 802.1Q header has non-zero encap attribute.");
1083 OVS_NLERR(log
, "Encap attr is set for non-VLAN frame");
1088 err
= ovs_key_from_nlattrs(match
, key_attrs
, a
, false, log
);
1094 /* Create an exact match mask. We need to set to 0xff
1095 * all the 'match->mask' fields that have been touched
1096 * in 'match->key'. We cannot simply memset
1097 * 'match->mask', because padding bytes and fields not
1098 * specified in 'match->key' should be left to 0.
1099 * Instead, we use a stream of netlink attributes,
1100 * copied from 'key' and set to 0xff.
1101 * ovs_key_from_nlattrs() will take care of filling
1102 * 'match->mask' appropriately.
1104 newmask
= kmemdup(nla_key
,
1105 nla_total_size(nla_len(nla_key
)),
1110 mask_set_nlattr(newmask
, 0xff);
1112 /* The userspace does not send tunnel attributes that
1113 * are 0, but we should not wildcard them nonetheless.
1115 if (match
->key
->tun_key
.ipv4_dst
)
1116 SW_FLOW_KEY_MEMSET_FIELD(match
, tun_key
,
1122 err
= parse_flow_mask_nlattrs(nla_mask
, a
, &mask_attrs
, log
);
1126 /* Always match on tci. */
1127 SW_FLOW_KEY_PUT(match
, eth
.tci
, htons(0xffff), true);
1129 if (mask_attrs
& 1ULL << OVS_KEY_ATTR_ENCAP
) {
1130 __be16 eth_type
= 0;
1134 OVS_NLERR(log
, "Encap mask attribute is set for non-VLAN frame.");
1139 mask_attrs
&= ~(1ULL << OVS_KEY_ATTR_ENCAP
);
1140 if (a
[OVS_KEY_ATTR_ETHERTYPE
])
1141 eth_type
= nla_get_be16(a
[OVS_KEY_ATTR_ETHERTYPE
]);
1143 if (eth_type
== htons(0xffff)) {
1144 mask_attrs
&= ~(1ULL << OVS_KEY_ATTR_ETHERTYPE
);
1145 encap
= a
[OVS_KEY_ATTR_ENCAP
];
1146 err
= parse_flow_mask_nlattrs(encap
, a
,
1151 OVS_NLERR(log
, "VLAN frames must have an exact match on the TPID (mask=%x).",
1157 if (a
[OVS_KEY_ATTR_VLAN
])
1158 tci
= nla_get_be16(a
[OVS_KEY_ATTR_VLAN
]);
1160 if (!(tci
& htons(VLAN_TAG_PRESENT
))) {
1161 OVS_NLERR(log
, "VLAN tag present bit must have an exact match (tci_mask=%x).",
1168 err
= ovs_key_from_nlattrs(match
, mask_attrs
, a
, true, log
);
1173 if (!match_validate(match
, key_attrs
, mask_attrs
, log
))
1181 static size_t get_ufid_len(const struct nlattr
*attr
, bool log
)
1188 len
= nla_len(attr
);
1189 if (len
< 1 || len
> MAX_UFID_LENGTH
) {
1190 OVS_NLERR(log
, "ufid size %u bytes exceeds the range (1, %d)",
1191 nla_len(attr
), MAX_UFID_LENGTH
);
1198 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1199 * or false otherwise.
1201 bool ovs_nla_get_ufid(struct sw_flow_id
*sfid
, const struct nlattr
*attr
,
1204 sfid
->ufid_len
= get_ufid_len(attr
, log
);
1206 memcpy(sfid
->ufid
, nla_data(attr
), sfid
->ufid_len
);
1208 return sfid
->ufid_len
;
1211 int ovs_nla_get_identifier(struct sw_flow_id
*sfid
, const struct nlattr
*ufid
,
1212 const struct sw_flow_key
*key
, bool log
)
1214 struct sw_flow_key
*new_key
;
1216 if (ovs_nla_get_ufid(sfid
, ufid
, log
))
1219 /* If UFID was not provided, use unmasked key. */
1220 new_key
= kmalloc(sizeof(*new_key
), GFP_KERNEL
);
1223 memcpy(new_key
, key
, sizeof(*key
));
1224 sfid
->unmasked_key
= new_key
;
1229 u32
ovs_nla_get_ufid_flags(const struct nlattr
*attr
)
1231 return attr
? nla_get_u32(attr
) : 0;
1235 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1236 * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1237 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1239 * @log: Boolean to allow kernel error logging. Normally true, but when
1240 * probing for feature compatibility this should be passed in as false to
1241 * suppress unnecessary error logging.
1243 * This parses a series of Netlink attributes that form a flow key, which must
1244 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1245 * get the metadata, that is, the parts of the flow key that cannot be
1246 * extracted from the packet itself.
1249 int ovs_nla_get_flow_metadata(const struct nlattr
*attr
,
1250 struct sw_flow_key
*key
,
1253 const struct nlattr
*a
[OVS_KEY_ATTR_MAX
+ 1];
1254 struct sw_flow_match match
;
1258 err
= parse_flow_nlattrs(attr
, a
, &attrs
, log
);
1262 memset(&match
, 0, sizeof(match
));
1265 memset(key
, 0, OVS_SW_FLOW_KEY_METADATA_SIZE
);
1266 key
->phy
.in_port
= DP_MAX_PORTS
;
1268 return metadata_from_nlattrs(&match
, &attrs
, a
, false, log
);
1271 static int __ovs_nla_put_key(const struct sw_flow_key
*swkey
,
1272 const struct sw_flow_key
*output
, bool is_mask
,
1273 struct sk_buff
*skb
)
1275 struct ovs_key_ethernet
*eth_key
;
1276 struct nlattr
*nla
, *encap
;
1278 if (nla_put_u32(skb
, OVS_KEY_ATTR_RECIRC_ID
, output
->recirc_id
))
1279 goto nla_put_failure
;
1281 if (nla_put_u32(skb
, OVS_KEY_ATTR_DP_HASH
, output
->ovs_flow_hash
))
1282 goto nla_put_failure
;
1284 if (nla_put_u32(skb
, OVS_KEY_ATTR_PRIORITY
, output
->phy
.priority
))
1285 goto nla_put_failure
;
1287 if ((swkey
->tun_key
.ipv4_dst
|| is_mask
)) {
1288 const void *opts
= NULL
;
1290 if (output
->tun_key
.tun_flags
& TUNNEL_OPTIONS_PRESENT
)
1291 opts
= TUN_METADATA_OPTS(output
, swkey
->tun_opts_len
);
1293 if (ipv4_tun_to_nlattr(skb
, &output
->tun_key
, opts
,
1294 swkey
->tun_opts_len
))
1295 goto nla_put_failure
;
1298 if (swkey
->phy
.in_port
== DP_MAX_PORTS
) {
1299 if (is_mask
&& (output
->phy
.in_port
== 0xffff))
1300 if (nla_put_u32(skb
, OVS_KEY_ATTR_IN_PORT
, 0xffffffff))
1301 goto nla_put_failure
;
1304 upper_u16
= !is_mask
? 0 : 0xffff;
1306 if (nla_put_u32(skb
, OVS_KEY_ATTR_IN_PORT
,
1307 (upper_u16
<< 16) | output
->phy
.in_port
))
1308 goto nla_put_failure
;
1311 if (nla_put_u32(skb
, OVS_KEY_ATTR_SKB_MARK
, output
->phy
.skb_mark
))
1312 goto nla_put_failure
;
1314 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ETHERNET
, sizeof(*eth_key
));
1316 goto nla_put_failure
;
1318 eth_key
= nla_data(nla
);
1319 ether_addr_copy(eth_key
->eth_src
, output
->eth
.src
);
1320 ether_addr_copy(eth_key
->eth_dst
, output
->eth
.dst
);
1322 if (swkey
->eth
.tci
|| swkey
->eth
.type
== htons(ETH_P_8021Q
)) {
1324 eth_type
= !is_mask
? htons(ETH_P_8021Q
) : htons(0xffff);
1325 if (nla_put_be16(skb
, OVS_KEY_ATTR_ETHERTYPE
, eth_type
) ||
1326 nla_put_be16(skb
, OVS_KEY_ATTR_VLAN
, output
->eth
.tci
))
1327 goto nla_put_failure
;
1328 encap
= nla_nest_start(skb
, OVS_KEY_ATTR_ENCAP
);
1329 if (!swkey
->eth
.tci
)
1334 if (swkey
->eth
.type
== htons(ETH_P_802_2
)) {
1336 * Ethertype 802.2 is represented in the netlink with omitted
1337 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1338 * 0xffff in the mask attribute. Ethertype can also
1341 if (is_mask
&& output
->eth
.type
)
1342 if (nla_put_be16(skb
, OVS_KEY_ATTR_ETHERTYPE
,
1344 goto nla_put_failure
;
1348 if (nla_put_be16(skb
, OVS_KEY_ATTR_ETHERTYPE
, output
->eth
.type
))
1349 goto nla_put_failure
;
1351 if (swkey
->eth
.type
== htons(ETH_P_IP
)) {
1352 struct ovs_key_ipv4
*ipv4_key
;
1354 nla
= nla_reserve(skb
, OVS_KEY_ATTR_IPV4
, sizeof(*ipv4_key
));
1356 goto nla_put_failure
;
1357 ipv4_key
= nla_data(nla
);
1358 ipv4_key
->ipv4_src
= output
->ipv4
.addr
.src
;
1359 ipv4_key
->ipv4_dst
= output
->ipv4
.addr
.dst
;
1360 ipv4_key
->ipv4_proto
= output
->ip
.proto
;
1361 ipv4_key
->ipv4_tos
= output
->ip
.tos
;
1362 ipv4_key
->ipv4_ttl
= output
->ip
.ttl
;
1363 ipv4_key
->ipv4_frag
= output
->ip
.frag
;
1364 } else if (swkey
->eth
.type
== htons(ETH_P_IPV6
)) {
1365 struct ovs_key_ipv6
*ipv6_key
;
1367 nla
= nla_reserve(skb
, OVS_KEY_ATTR_IPV6
, sizeof(*ipv6_key
));
1369 goto nla_put_failure
;
1370 ipv6_key
= nla_data(nla
);
1371 memcpy(ipv6_key
->ipv6_src
, &output
->ipv6
.addr
.src
,
1372 sizeof(ipv6_key
->ipv6_src
));
1373 memcpy(ipv6_key
->ipv6_dst
, &output
->ipv6
.addr
.dst
,
1374 sizeof(ipv6_key
->ipv6_dst
));
1375 ipv6_key
->ipv6_label
= output
->ipv6
.label
;
1376 ipv6_key
->ipv6_proto
= output
->ip
.proto
;
1377 ipv6_key
->ipv6_tclass
= output
->ip
.tos
;
1378 ipv6_key
->ipv6_hlimit
= output
->ip
.ttl
;
1379 ipv6_key
->ipv6_frag
= output
->ip
.frag
;
1380 } else if (swkey
->eth
.type
== htons(ETH_P_ARP
) ||
1381 swkey
->eth
.type
== htons(ETH_P_RARP
)) {
1382 struct ovs_key_arp
*arp_key
;
1384 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ARP
, sizeof(*arp_key
));
1386 goto nla_put_failure
;
1387 arp_key
= nla_data(nla
);
1388 memset(arp_key
, 0, sizeof(struct ovs_key_arp
));
1389 arp_key
->arp_sip
= output
->ipv4
.addr
.src
;
1390 arp_key
->arp_tip
= output
->ipv4
.addr
.dst
;
1391 arp_key
->arp_op
= htons(output
->ip
.proto
);
1392 ether_addr_copy(arp_key
->arp_sha
, output
->ipv4
.arp
.sha
);
1393 ether_addr_copy(arp_key
->arp_tha
, output
->ipv4
.arp
.tha
);
1394 } else if (eth_p_mpls(swkey
->eth
.type
)) {
1395 struct ovs_key_mpls
*mpls_key
;
1397 nla
= nla_reserve(skb
, OVS_KEY_ATTR_MPLS
, sizeof(*mpls_key
));
1399 goto nla_put_failure
;
1400 mpls_key
= nla_data(nla
);
1401 mpls_key
->mpls_lse
= output
->mpls
.top_lse
;
1404 if ((swkey
->eth
.type
== htons(ETH_P_IP
) ||
1405 swkey
->eth
.type
== htons(ETH_P_IPV6
)) &&
1406 swkey
->ip
.frag
!= OVS_FRAG_TYPE_LATER
) {
1408 if (swkey
->ip
.proto
== IPPROTO_TCP
) {
1409 struct ovs_key_tcp
*tcp_key
;
1411 nla
= nla_reserve(skb
, OVS_KEY_ATTR_TCP
, sizeof(*tcp_key
));
1413 goto nla_put_failure
;
1414 tcp_key
= nla_data(nla
);
1415 tcp_key
->tcp_src
= output
->tp
.src
;
1416 tcp_key
->tcp_dst
= output
->tp
.dst
;
1417 if (nla_put_be16(skb
, OVS_KEY_ATTR_TCP_FLAGS
,
1419 goto nla_put_failure
;
1420 } else if (swkey
->ip
.proto
== IPPROTO_UDP
) {
1421 struct ovs_key_udp
*udp_key
;
1423 nla
= nla_reserve(skb
, OVS_KEY_ATTR_UDP
, sizeof(*udp_key
));
1425 goto nla_put_failure
;
1426 udp_key
= nla_data(nla
);
1427 udp_key
->udp_src
= output
->tp
.src
;
1428 udp_key
->udp_dst
= output
->tp
.dst
;
1429 } else if (swkey
->ip
.proto
== IPPROTO_SCTP
) {
1430 struct ovs_key_sctp
*sctp_key
;
1432 nla
= nla_reserve(skb
, OVS_KEY_ATTR_SCTP
, sizeof(*sctp_key
));
1434 goto nla_put_failure
;
1435 sctp_key
= nla_data(nla
);
1436 sctp_key
->sctp_src
= output
->tp
.src
;
1437 sctp_key
->sctp_dst
= output
->tp
.dst
;
1438 } else if (swkey
->eth
.type
== htons(ETH_P_IP
) &&
1439 swkey
->ip
.proto
== IPPROTO_ICMP
) {
1440 struct ovs_key_icmp
*icmp_key
;
1442 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ICMP
, sizeof(*icmp_key
));
1444 goto nla_put_failure
;
1445 icmp_key
= nla_data(nla
);
1446 icmp_key
->icmp_type
= ntohs(output
->tp
.src
);
1447 icmp_key
->icmp_code
= ntohs(output
->tp
.dst
);
1448 } else if (swkey
->eth
.type
== htons(ETH_P_IPV6
) &&
1449 swkey
->ip
.proto
== IPPROTO_ICMPV6
) {
1450 struct ovs_key_icmpv6
*icmpv6_key
;
1452 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ICMPV6
,
1453 sizeof(*icmpv6_key
));
1455 goto nla_put_failure
;
1456 icmpv6_key
= nla_data(nla
);
1457 icmpv6_key
->icmpv6_type
= ntohs(output
->tp
.src
);
1458 icmpv6_key
->icmpv6_code
= ntohs(output
->tp
.dst
);
1460 if (icmpv6_key
->icmpv6_type
== NDISC_NEIGHBOUR_SOLICITATION
||
1461 icmpv6_key
->icmpv6_type
== NDISC_NEIGHBOUR_ADVERTISEMENT
) {
1462 struct ovs_key_nd
*nd_key
;
1464 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ND
, sizeof(*nd_key
));
1466 goto nla_put_failure
;
1467 nd_key
= nla_data(nla
);
1468 memcpy(nd_key
->nd_target
, &output
->ipv6
.nd
.target
,
1469 sizeof(nd_key
->nd_target
));
1470 ether_addr_copy(nd_key
->nd_sll
, output
->ipv6
.nd
.sll
);
1471 ether_addr_copy(nd_key
->nd_tll
, output
->ipv6
.nd
.tll
);
1478 nla_nest_end(skb
, encap
);
1486 int ovs_nla_put_key(const struct sw_flow_key
*swkey
,
1487 const struct sw_flow_key
*output
, int attr
, bool is_mask
,
1488 struct sk_buff
*skb
)
1493 nla
= nla_nest_start(skb
, attr
);
1496 err
= __ovs_nla_put_key(swkey
, output
, is_mask
, skb
);
1499 nla_nest_end(skb
, nla
);
1504 /* Called with ovs_mutex or RCU read lock. */
1505 int ovs_nla_put_identifier(const struct sw_flow
*flow
, struct sk_buff
*skb
)
1507 if (ovs_identifier_is_ufid(&flow
->id
))
1508 return nla_put(skb
, OVS_FLOW_ATTR_UFID
, flow
->id
.ufid_len
,
1511 return ovs_nla_put_key(flow
->id
.unmasked_key
, flow
->id
.unmasked_key
,
1512 OVS_FLOW_ATTR_KEY
, false, skb
);
1515 /* Called with ovs_mutex or RCU read lock. */
1516 int ovs_nla_put_masked_key(const struct sw_flow
*flow
, struct sk_buff
*skb
)
1518 return ovs_nla_put_key(&flow
->key
, &flow
->key
,
1519 OVS_FLOW_ATTR_KEY
, false, skb
);
1522 /* Called with ovs_mutex or RCU read lock. */
1523 int ovs_nla_put_mask(const struct sw_flow
*flow
, struct sk_buff
*skb
)
1525 return ovs_nla_put_key(&flow
->key
, &flow
->mask
->key
,
1526 OVS_FLOW_ATTR_MASK
, true, skb
);
1529 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1531 static struct sw_flow_actions
*nla_alloc_flow_actions(int size
, bool log
)
1533 struct sw_flow_actions
*sfa
;
1535 if (size
> MAX_ACTIONS_BUFSIZE
) {
1536 OVS_NLERR(log
, "Flow action size %u bytes exceeds max", size
);
1537 return ERR_PTR(-EINVAL
);
1540 sfa
= kmalloc(sizeof(*sfa
) + size
, GFP_KERNEL
);
1542 return ERR_PTR(-ENOMEM
);
1544 sfa
->actions_len
= 0;
1548 /* RCU callback used by ovs_nla_free_flow_actions. */
1549 static void rcu_free_acts_callback(struct rcu_head
*rcu
)
1551 struct sw_flow_actions
*sf_acts
= container_of(rcu
,
1552 struct sw_flow_actions
, rcu
);
1556 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1557 * The caller must hold rcu_read_lock for this to be sensible.
1559 void ovs_nla_free_flow_actions(struct sw_flow_actions
*sf_acts
)
1561 call_rcu(&sf_acts
->rcu
, rcu_free_acts_callback
);
1564 static struct nlattr
*reserve_sfa_size(struct sw_flow_actions
**sfa
,
1565 int attr_len
, bool log
)
1568 struct sw_flow_actions
*acts
;
1570 int req_size
= NLA_ALIGN(attr_len
);
1571 int next_offset
= offsetof(struct sw_flow_actions
, actions
) +
1572 (*sfa
)->actions_len
;
1574 if (req_size
<= (ksize(*sfa
) - next_offset
))
1577 new_acts_size
= ksize(*sfa
) * 2;
1579 if (new_acts_size
> MAX_ACTIONS_BUFSIZE
) {
1580 if ((MAX_ACTIONS_BUFSIZE
- next_offset
) < req_size
)
1581 return ERR_PTR(-EMSGSIZE
);
1582 new_acts_size
= MAX_ACTIONS_BUFSIZE
;
1585 acts
= nla_alloc_flow_actions(new_acts_size
, log
);
1587 return (void *)acts
;
1589 memcpy(acts
->actions
, (*sfa
)->actions
, (*sfa
)->actions_len
);
1590 acts
->actions_len
= (*sfa
)->actions_len
;
1595 (*sfa
)->actions_len
+= req_size
;
1596 return (struct nlattr
*) ((unsigned char *)(*sfa
) + next_offset
);
1599 static struct nlattr
*__add_action(struct sw_flow_actions
**sfa
,
1600 int attrtype
, void *data
, int len
, bool log
)
1604 a
= reserve_sfa_size(sfa
, nla_attr_size(len
), log
);
1608 a
->nla_type
= attrtype
;
1609 a
->nla_len
= nla_attr_size(len
);
1612 memcpy(nla_data(a
), data
, len
);
1613 memset((unsigned char *) a
+ a
->nla_len
, 0, nla_padlen(len
));
1618 static int add_action(struct sw_flow_actions
**sfa
, int attrtype
,
1619 void *data
, int len
, bool log
)
1623 a
= __add_action(sfa
, attrtype
, data
, len
, log
);
1630 static inline int add_nested_action_start(struct sw_flow_actions
**sfa
,
1631 int attrtype
, bool log
)
1633 int used
= (*sfa
)->actions_len
;
1636 err
= add_action(sfa
, attrtype
, NULL
, 0, log
);
1643 static inline void add_nested_action_end(struct sw_flow_actions
*sfa
,
1646 struct nlattr
*a
= (struct nlattr
*) ((unsigned char *)sfa
->actions
+
1649 a
->nla_len
= sfa
->actions_len
- st_offset
;
1652 static int __ovs_nla_copy_actions(const struct nlattr
*attr
,
1653 const struct sw_flow_key
*key
,
1654 int depth
, struct sw_flow_actions
**sfa
,
1655 __be16 eth_type
, __be16 vlan_tci
, bool log
);
1657 static int validate_and_copy_sample(const struct nlattr
*attr
,
1658 const struct sw_flow_key
*key
, int depth
,
1659 struct sw_flow_actions
**sfa
,
1660 __be16 eth_type
, __be16 vlan_tci
, bool log
)
1662 const struct nlattr
*attrs
[OVS_SAMPLE_ATTR_MAX
+ 1];
1663 const struct nlattr
*probability
, *actions
;
1664 const struct nlattr
*a
;
1665 int rem
, start
, err
, st_acts
;
1667 memset(attrs
, 0, sizeof(attrs
));
1668 nla_for_each_nested(a
, attr
, rem
) {
1669 int type
= nla_type(a
);
1670 if (!type
|| type
> OVS_SAMPLE_ATTR_MAX
|| attrs
[type
])
1677 probability
= attrs
[OVS_SAMPLE_ATTR_PROBABILITY
];
1678 if (!probability
|| nla_len(probability
) != sizeof(u32
))
1681 actions
= attrs
[OVS_SAMPLE_ATTR_ACTIONS
];
1682 if (!actions
|| (nla_len(actions
) && nla_len(actions
) < NLA_HDRLEN
))
1685 /* validation done, copy sample action. */
1686 start
= add_nested_action_start(sfa
, OVS_ACTION_ATTR_SAMPLE
, log
);
1689 err
= add_action(sfa
, OVS_SAMPLE_ATTR_PROBABILITY
,
1690 nla_data(probability
), sizeof(u32
), log
);
1693 st_acts
= add_nested_action_start(sfa
, OVS_SAMPLE_ATTR_ACTIONS
, log
);
1697 err
= __ovs_nla_copy_actions(actions
, key
, depth
+ 1, sfa
,
1698 eth_type
, vlan_tci
, log
);
1702 add_nested_action_end(*sfa
, st_acts
);
1703 add_nested_action_end(*sfa
, start
);
1708 static int validate_tp_port(const struct sw_flow_key
*flow_key
,
1711 if ((eth_type
== htons(ETH_P_IP
) || eth_type
== htons(ETH_P_IPV6
)) &&
1712 (flow_key
->tp
.src
|| flow_key
->tp
.dst
))
1718 void ovs_match_init(struct sw_flow_match
*match
,
1719 struct sw_flow_key
*key
,
1720 struct sw_flow_mask
*mask
)
1722 memset(match
, 0, sizeof(*match
));
1726 memset(key
, 0, sizeof(*key
));
1729 memset(&mask
->key
, 0, sizeof(mask
->key
));
1730 mask
->range
.start
= mask
->range
.end
= 0;
1734 static int validate_geneve_opts(struct sw_flow_key
*key
)
1736 struct geneve_opt
*option
;
1737 int opts_len
= key
->tun_opts_len
;
1738 bool crit_opt
= false;
1740 option
= (struct geneve_opt
*)TUN_METADATA_OPTS(key
, key
->tun_opts_len
);
1741 while (opts_len
> 0) {
1744 if (opts_len
< sizeof(*option
))
1747 len
= sizeof(*option
) + option
->length
* 4;
1751 crit_opt
|= !!(option
->type
& GENEVE_CRIT_OPT_TYPE
);
1753 option
= (struct geneve_opt
*)((u8
*)option
+ len
);
1757 key
->tun_key
.tun_flags
|= crit_opt
? TUNNEL_CRIT_OPT
: 0;
1762 static int validate_and_copy_set_tun(const struct nlattr
*attr
,
1763 struct sw_flow_actions
**sfa
, bool log
)
1765 struct sw_flow_match match
;
1766 struct sw_flow_key key
;
1767 struct ovs_tunnel_info
*tun_info
;
1769 int start
, opts_type
;
1772 ovs_match_init(&match
, &key
, NULL
);
1773 opts_type
= ipv4_tun_from_nlattr(nla_data(attr
), &match
, false, log
);
1777 if (key
.tun_opts_len
) {
1778 switch (opts_type
) {
1779 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS
:
1780 err
= validate_geneve_opts(&key
);
1784 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS
:
1789 start
= add_nested_action_start(sfa
, OVS_ACTION_ATTR_SET
, log
);
1793 a
= __add_action(sfa
, OVS_KEY_ATTR_TUNNEL_INFO
, NULL
,
1794 sizeof(*tun_info
) + key
.tun_opts_len
, log
);
1798 tun_info
= nla_data(a
);
1799 tun_info
->tunnel
= key
.tun_key
;
1800 tun_info
->options_len
= key
.tun_opts_len
;
1802 if (tun_info
->options_len
) {
1803 /* We need to store the options in the action itself since
1804 * everything else will go away after flow setup. We can append
1805 * it to tun_info and then point there.
1807 memcpy((tun_info
+ 1),
1808 TUN_METADATA_OPTS(&key
, key
.tun_opts_len
), key
.tun_opts_len
);
1809 tun_info
->options
= (tun_info
+ 1);
1812 tun_info
->options
= NULL
;
1815 add_nested_action_end(*sfa
, start
);
1820 static int validate_set(const struct nlattr
*a
,
1821 const struct sw_flow_key
*flow_key
,
1822 struct sw_flow_actions
**sfa
,
1823 bool *set_tun
, __be16 eth_type
, bool log
)
1825 const struct nlattr
*ovs_key
= nla_data(a
);
1826 int key_type
= nla_type(ovs_key
);
1828 /* There can be only one key in a action */
1829 if (nla_total_size(nla_len(ovs_key
)) != nla_len(a
))
1832 if (key_type
> OVS_KEY_ATTR_MAX
||
1833 (ovs_key_lens
[key_type
].len
!= nla_len(ovs_key
) &&
1834 ovs_key_lens
[key_type
].len
!= OVS_ATTR_NESTED
))
1838 const struct ovs_key_ipv4
*ipv4_key
;
1839 const struct ovs_key_ipv6
*ipv6_key
;
1842 case OVS_KEY_ATTR_PRIORITY
:
1843 case OVS_KEY_ATTR_SKB_MARK
:
1844 case OVS_KEY_ATTR_ETHERNET
:
1847 case OVS_KEY_ATTR_TUNNEL
:
1848 if (eth_p_mpls(eth_type
))
1852 err
= validate_and_copy_set_tun(a
, sfa
, log
);
1857 case OVS_KEY_ATTR_IPV4
:
1858 if (eth_type
!= htons(ETH_P_IP
))
1861 if (!flow_key
->ip
.proto
)
1864 ipv4_key
= nla_data(ovs_key
);
1865 if (ipv4_key
->ipv4_proto
!= flow_key
->ip
.proto
)
1868 if (ipv4_key
->ipv4_frag
!= flow_key
->ip
.frag
)
1873 case OVS_KEY_ATTR_IPV6
:
1874 if (eth_type
!= htons(ETH_P_IPV6
))
1877 if (!flow_key
->ip
.proto
)
1880 ipv6_key
= nla_data(ovs_key
);
1881 if (ipv6_key
->ipv6_proto
!= flow_key
->ip
.proto
)
1884 if (ipv6_key
->ipv6_frag
!= flow_key
->ip
.frag
)
1887 if (ntohl(ipv6_key
->ipv6_label
) & 0xFFF00000)
1892 case OVS_KEY_ATTR_TCP
:
1893 if (flow_key
->ip
.proto
!= IPPROTO_TCP
)
1896 return validate_tp_port(flow_key
, eth_type
);
1898 case OVS_KEY_ATTR_UDP
:
1899 if (flow_key
->ip
.proto
!= IPPROTO_UDP
)
1902 return validate_tp_port(flow_key
, eth_type
);
1904 case OVS_KEY_ATTR_MPLS
:
1905 if (!eth_p_mpls(eth_type
))
1909 case OVS_KEY_ATTR_SCTP
:
1910 if (flow_key
->ip
.proto
!= IPPROTO_SCTP
)
1913 return validate_tp_port(flow_key
, eth_type
);
1922 static int validate_userspace(const struct nlattr
*attr
)
1924 static const struct nla_policy userspace_policy
[OVS_USERSPACE_ATTR_MAX
+ 1] = {
1925 [OVS_USERSPACE_ATTR_PID
] = {.type
= NLA_U32
},
1926 [OVS_USERSPACE_ATTR_USERDATA
] = {.type
= NLA_UNSPEC
},
1927 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT
] = {.type
= NLA_U32
},
1929 struct nlattr
*a
[OVS_USERSPACE_ATTR_MAX
+ 1];
1932 error
= nla_parse_nested(a
, OVS_USERSPACE_ATTR_MAX
,
1933 attr
, userspace_policy
);
1937 if (!a
[OVS_USERSPACE_ATTR_PID
] ||
1938 !nla_get_u32(a
[OVS_USERSPACE_ATTR_PID
]))
1944 static int copy_action(const struct nlattr
*from
,
1945 struct sw_flow_actions
**sfa
, bool log
)
1947 int totlen
= NLA_ALIGN(from
->nla_len
);
1950 to
= reserve_sfa_size(sfa
, from
->nla_len
, log
);
1954 memcpy(to
, from
, totlen
);
1958 static int __ovs_nla_copy_actions(const struct nlattr
*attr
,
1959 const struct sw_flow_key
*key
,
1960 int depth
, struct sw_flow_actions
**sfa
,
1961 __be16 eth_type
, __be16 vlan_tci
, bool log
)
1963 const struct nlattr
*a
;
1966 if (depth
>= SAMPLE_ACTION_DEPTH
)
1969 nla_for_each_nested(a
, attr
, rem
) {
1970 /* Expected argument lengths, (u32)-1 for variable length. */
1971 static const u32 action_lens
[OVS_ACTION_ATTR_MAX
+ 1] = {
1972 [OVS_ACTION_ATTR_OUTPUT
] = sizeof(u32
),
1973 [OVS_ACTION_ATTR_RECIRC
] = sizeof(u32
),
1974 [OVS_ACTION_ATTR_USERSPACE
] = (u32
)-1,
1975 [OVS_ACTION_ATTR_PUSH_MPLS
] = sizeof(struct ovs_action_push_mpls
),
1976 [OVS_ACTION_ATTR_POP_MPLS
] = sizeof(__be16
),
1977 [OVS_ACTION_ATTR_PUSH_VLAN
] = sizeof(struct ovs_action_push_vlan
),
1978 [OVS_ACTION_ATTR_POP_VLAN
] = 0,
1979 [OVS_ACTION_ATTR_SET
] = (u32
)-1,
1980 [OVS_ACTION_ATTR_SAMPLE
] = (u32
)-1,
1981 [OVS_ACTION_ATTR_HASH
] = sizeof(struct ovs_action_hash
)
1983 const struct ovs_action_push_vlan
*vlan
;
1984 int type
= nla_type(a
);
1987 if (type
> OVS_ACTION_ATTR_MAX
||
1988 (action_lens
[type
] != nla_len(a
) &&
1989 action_lens
[type
] != (u32
)-1))
1994 case OVS_ACTION_ATTR_UNSPEC
:
1997 case OVS_ACTION_ATTR_USERSPACE
:
1998 err
= validate_userspace(a
);
2003 case OVS_ACTION_ATTR_OUTPUT
:
2004 if (nla_get_u32(a
) >= DP_MAX_PORTS
)
2009 case OVS_ACTION_ATTR_HASH
: {
2010 const struct ovs_action_hash
*act_hash
= nla_data(a
);
2012 switch (act_hash
->hash_alg
) {
2013 case OVS_HASH_ALG_L4
:
2022 case OVS_ACTION_ATTR_POP_VLAN
:
2023 vlan_tci
= htons(0);
2026 case OVS_ACTION_ATTR_PUSH_VLAN
:
2028 if (vlan
->vlan_tpid
!= htons(ETH_P_8021Q
))
2030 if (!(vlan
->vlan_tci
& htons(VLAN_TAG_PRESENT
)))
2032 vlan_tci
= vlan
->vlan_tci
;
2035 case OVS_ACTION_ATTR_RECIRC
:
2038 case OVS_ACTION_ATTR_PUSH_MPLS
: {
2039 const struct ovs_action_push_mpls
*mpls
= nla_data(a
);
2041 if (!eth_p_mpls(mpls
->mpls_ethertype
))
2044 /* Prohibit push MPLS other than to a white list
2045 * for packets that have a known tag order.
2047 if (vlan_tci
& htons(VLAN_TAG_PRESENT
) ||
2048 (eth_type
!= htons(ETH_P_IP
) &&
2049 eth_type
!= htons(ETH_P_IPV6
) &&
2050 eth_type
!= htons(ETH_P_ARP
) &&
2051 eth_type
!= htons(ETH_P_RARP
) &&
2052 !eth_p_mpls(eth_type
)))
2054 eth_type
= mpls
->mpls_ethertype
;
2058 case OVS_ACTION_ATTR_POP_MPLS
:
2059 if (vlan_tci
& htons(VLAN_TAG_PRESENT
) ||
2060 !eth_p_mpls(eth_type
))
2063 /* Disallow subsequent L2.5+ set and mpls_pop actions
2064 * as there is no check here to ensure that the new
2065 * eth_type is valid and thus set actions could
2066 * write off the end of the packet or otherwise
2069 * Support for these actions is planned using packet
2072 eth_type
= htons(0);
2075 case OVS_ACTION_ATTR_SET
:
2076 err
= validate_set(a
, key
, sfa
,
2077 &skip_copy
, eth_type
, log
);
2082 case OVS_ACTION_ATTR_SAMPLE
:
2083 err
= validate_and_copy_sample(a
, key
, depth
, sfa
,
2084 eth_type
, vlan_tci
, log
);
2091 OVS_NLERR(log
, "Unknown Action type %d", type
);
2095 err
= copy_action(a
, sfa
, log
);
2107 int ovs_nla_copy_actions(const struct nlattr
*attr
,
2108 const struct sw_flow_key
*key
,
2109 struct sw_flow_actions
**sfa
, bool log
)
2113 *sfa
= nla_alloc_flow_actions(nla_len(attr
), log
);
2115 return PTR_ERR(*sfa
);
2117 err
= __ovs_nla_copy_actions(attr
, key
, 0, sfa
, key
->eth
.type
,
2125 static int sample_action_to_attr(const struct nlattr
*attr
, struct sk_buff
*skb
)
2127 const struct nlattr
*a
;
2128 struct nlattr
*start
;
2131 start
= nla_nest_start(skb
, OVS_ACTION_ATTR_SAMPLE
);
2135 nla_for_each_nested(a
, attr
, rem
) {
2136 int type
= nla_type(a
);
2137 struct nlattr
*st_sample
;
2140 case OVS_SAMPLE_ATTR_PROBABILITY
:
2141 if (nla_put(skb
, OVS_SAMPLE_ATTR_PROBABILITY
,
2142 sizeof(u32
), nla_data(a
)))
2145 case OVS_SAMPLE_ATTR_ACTIONS
:
2146 st_sample
= nla_nest_start(skb
, OVS_SAMPLE_ATTR_ACTIONS
);
2149 err
= ovs_nla_put_actions(nla_data(a
), nla_len(a
), skb
);
2152 nla_nest_end(skb
, st_sample
);
2157 nla_nest_end(skb
, start
);
2161 static int set_action_to_attr(const struct nlattr
*a
, struct sk_buff
*skb
)
2163 const struct nlattr
*ovs_key
= nla_data(a
);
2164 int key_type
= nla_type(ovs_key
);
2165 struct nlattr
*start
;
2169 case OVS_KEY_ATTR_TUNNEL_INFO
: {
2170 struct ovs_tunnel_info
*tun_info
= nla_data(ovs_key
);
2172 start
= nla_nest_start(skb
, OVS_ACTION_ATTR_SET
);
2176 err
= ipv4_tun_to_nlattr(skb
, &tun_info
->tunnel
,
2177 tun_info
->options_len
?
2178 tun_info
->options
: NULL
,
2179 tun_info
->options_len
);
2182 nla_nest_end(skb
, start
);
2186 if (nla_put(skb
, OVS_ACTION_ATTR_SET
, nla_len(a
), ovs_key
))
2194 int ovs_nla_put_actions(const struct nlattr
*attr
, int len
, struct sk_buff
*skb
)
2196 const struct nlattr
*a
;
2199 nla_for_each_attr(a
, attr
, len
, rem
) {
2200 int type
= nla_type(a
);
2203 case OVS_ACTION_ATTR_SET
:
2204 err
= set_action_to_attr(a
, skb
);
2209 case OVS_ACTION_ATTR_SAMPLE
:
2210 err
= sample_action_to_attr(a
, skb
);
2215 if (nla_put(skb
, type
, nla_len(a
), nla_data(a
)))