1 // SPDX-License-Identifier: GPL-2.0-only
3 * Copyright (c) 2007-2017 Nicira, Inc.
6 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
10 #include <linux/uaccess.h>
11 #include <linux/netdevice.h>
12 #include <linux/etherdevice.h>
13 #include <linux/if_ether.h>
14 #include <linux/if_vlan.h>
15 #include <net/llc_pdu.h>
16 #include <linux/kernel.h>
17 #include <linux/jhash.h>
18 #include <linux/jiffies.h>
19 #include <linux/llc.h>
20 #include <linux/module.h>
22 #include <linux/rcupdate.h>
23 #include <linux/if_arp.h>
25 #include <linux/ipv6.h>
26 #include <linux/sctp.h>
27 #include <linux/tcp.h>
28 #include <linux/udp.h>
29 #include <linux/icmp.h>
30 #include <linux/icmpv6.h>
31 #include <linux/rculist.h>
32 #include <net/geneve.h>
35 #include <net/ndisc.h>
37 #include <net/vxlan.h>
38 #include <net/tun_proto.h>
39 #include <net/erspan.h>
41 #include "flow_netlink.h"
45 const struct ovs_len_tbl
*next
;
48 #define OVS_ATTR_NESTED -1
49 #define OVS_ATTR_VARIABLE -2
51 static bool actions_may_change_flow(const struct nlattr
*actions
)
56 nla_for_each_nested(nla
, actions
, rem
) {
57 u16 action
= nla_type(nla
);
60 case OVS_ACTION_ATTR_OUTPUT
:
61 case OVS_ACTION_ATTR_RECIRC
:
62 case OVS_ACTION_ATTR_TRUNC
:
63 case OVS_ACTION_ATTR_USERSPACE
:
66 case OVS_ACTION_ATTR_CT
:
67 case OVS_ACTION_ATTR_CT_CLEAR
:
68 case OVS_ACTION_ATTR_HASH
:
69 case OVS_ACTION_ATTR_POP_ETH
:
70 case OVS_ACTION_ATTR_POP_MPLS
:
71 case OVS_ACTION_ATTR_POP_NSH
:
72 case OVS_ACTION_ATTR_POP_VLAN
:
73 case OVS_ACTION_ATTR_PUSH_ETH
:
74 case OVS_ACTION_ATTR_PUSH_MPLS
:
75 case OVS_ACTION_ATTR_PUSH_NSH
:
76 case OVS_ACTION_ATTR_PUSH_VLAN
:
77 case OVS_ACTION_ATTR_SAMPLE
:
78 case OVS_ACTION_ATTR_SET
:
79 case OVS_ACTION_ATTR_SET_MASKED
:
80 case OVS_ACTION_ATTR_METER
:
81 case OVS_ACTION_ATTR_CHECK_PKT_LEN
:
82 case OVS_ACTION_ATTR_ADD_MPLS
:
83 case OVS_ACTION_ATTR_DEC_TTL
:
91 static void update_range(struct sw_flow_match
*match
,
92 size_t offset
, size_t size
, bool is_mask
)
94 struct sw_flow_key_range
*range
;
95 size_t start
= rounddown(offset
, sizeof(long));
96 size_t end
= roundup(offset
+ size
, sizeof(long));
99 range
= &match
->range
;
101 range
= &match
->mask
->range
;
103 if (range
->start
== range
->end
) {
104 range
->start
= start
;
109 if (range
->start
> start
)
110 range
->start
= start
;
112 if (range
->end
< end
)
116 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
118 update_range(match, offsetof(struct sw_flow_key, field), \
119 sizeof((match)->key->field), is_mask); \
121 (match)->mask->key.field = value; \
123 (match)->key->field = value; \
126 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
128 update_range(match, offset, len, is_mask); \
130 memcpy((u8 *)&(match)->mask->key + offset, value_p, \
133 memcpy((u8 *)(match)->key + offset, value_p, len); \
136 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
137 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
138 value_p, len, is_mask)
140 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
142 update_range(match, offsetof(struct sw_flow_key, field), \
143 sizeof((match)->key->field), is_mask); \
145 memset((u8 *)&(match)->mask->key.field, value, \
146 sizeof((match)->mask->key.field)); \
148 memset((u8 *)&(match)->key->field, value, \
149 sizeof((match)->key->field)); \
152 static bool match_validate(const struct sw_flow_match
*match
,
153 u64 key_attrs
, u64 mask_attrs
, bool log
)
155 u64 key_expected
= 0;
156 u64 mask_allowed
= key_attrs
; /* At most allow all key attributes */
158 /* The following mask attributes allowed only if they
159 * pass the validation tests. */
160 mask_allowed
&= ~((1 << OVS_KEY_ATTR_IPV4
)
161 | (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4
)
162 | (1 << OVS_KEY_ATTR_IPV6
)
163 | (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6
)
164 | (1 << OVS_KEY_ATTR_TCP
)
165 | (1 << OVS_KEY_ATTR_TCP_FLAGS
)
166 | (1 << OVS_KEY_ATTR_UDP
)
167 | (1 << OVS_KEY_ATTR_SCTP
)
168 | (1 << OVS_KEY_ATTR_ICMP
)
169 | (1 << OVS_KEY_ATTR_ICMPV6
)
170 | (1 << OVS_KEY_ATTR_ARP
)
171 | (1 << OVS_KEY_ATTR_ND
)
172 | (1 << OVS_KEY_ATTR_MPLS
)
173 | (1 << OVS_KEY_ATTR_NSH
));
175 /* Always allowed mask fields. */
176 mask_allowed
|= ((1 << OVS_KEY_ATTR_TUNNEL
)
177 | (1 << OVS_KEY_ATTR_IN_PORT
)
178 | (1 << OVS_KEY_ATTR_ETHERTYPE
));
180 /* Check key attributes. */
181 if (match
->key
->eth
.type
== htons(ETH_P_ARP
)
182 || match
->key
->eth
.type
== htons(ETH_P_RARP
)) {
183 key_expected
|= 1 << OVS_KEY_ATTR_ARP
;
184 if (match
->mask
&& (match
->mask
->key
.eth
.type
== htons(0xffff)))
185 mask_allowed
|= 1 << OVS_KEY_ATTR_ARP
;
188 if (eth_p_mpls(match
->key
->eth
.type
)) {
189 key_expected
|= 1 << OVS_KEY_ATTR_MPLS
;
190 if (match
->mask
&& (match
->mask
->key
.eth
.type
== htons(0xffff)))
191 mask_allowed
|= 1 << OVS_KEY_ATTR_MPLS
;
194 if (match
->key
->eth
.type
== htons(ETH_P_IP
)) {
195 key_expected
|= 1 << OVS_KEY_ATTR_IPV4
;
196 if (match
->mask
&& match
->mask
->key
.eth
.type
== htons(0xffff)) {
197 mask_allowed
|= 1 << OVS_KEY_ATTR_IPV4
;
198 mask_allowed
|= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4
;
201 if (match
->key
->ip
.frag
!= OVS_FRAG_TYPE_LATER
) {
202 if (match
->key
->ip
.proto
== IPPROTO_UDP
) {
203 key_expected
|= 1 << OVS_KEY_ATTR_UDP
;
204 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff))
205 mask_allowed
|= 1 << OVS_KEY_ATTR_UDP
;
208 if (match
->key
->ip
.proto
== IPPROTO_SCTP
) {
209 key_expected
|= 1 << OVS_KEY_ATTR_SCTP
;
210 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff))
211 mask_allowed
|= 1 << OVS_KEY_ATTR_SCTP
;
214 if (match
->key
->ip
.proto
== IPPROTO_TCP
) {
215 key_expected
|= 1 << OVS_KEY_ATTR_TCP
;
216 key_expected
|= 1 << OVS_KEY_ATTR_TCP_FLAGS
;
217 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff)) {
218 mask_allowed
|= 1 << OVS_KEY_ATTR_TCP
;
219 mask_allowed
|= 1 << OVS_KEY_ATTR_TCP_FLAGS
;
223 if (match
->key
->ip
.proto
== IPPROTO_ICMP
) {
224 key_expected
|= 1 << OVS_KEY_ATTR_ICMP
;
225 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff))
226 mask_allowed
|= 1 << OVS_KEY_ATTR_ICMP
;
231 if (match
->key
->eth
.type
== htons(ETH_P_IPV6
)) {
232 key_expected
|= 1 << OVS_KEY_ATTR_IPV6
;
233 if (match
->mask
&& match
->mask
->key
.eth
.type
== htons(0xffff)) {
234 mask_allowed
|= 1 << OVS_KEY_ATTR_IPV6
;
235 mask_allowed
|= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6
;
238 if (match
->key
->ip
.frag
!= OVS_FRAG_TYPE_LATER
) {
239 if (match
->key
->ip
.proto
== IPPROTO_UDP
) {
240 key_expected
|= 1 << OVS_KEY_ATTR_UDP
;
241 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff))
242 mask_allowed
|= 1 << OVS_KEY_ATTR_UDP
;
245 if (match
->key
->ip
.proto
== IPPROTO_SCTP
) {
246 key_expected
|= 1 << OVS_KEY_ATTR_SCTP
;
247 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff))
248 mask_allowed
|= 1 << OVS_KEY_ATTR_SCTP
;
251 if (match
->key
->ip
.proto
== IPPROTO_TCP
) {
252 key_expected
|= 1 << OVS_KEY_ATTR_TCP
;
253 key_expected
|= 1 << OVS_KEY_ATTR_TCP_FLAGS
;
254 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff)) {
255 mask_allowed
|= 1 << OVS_KEY_ATTR_TCP
;
256 mask_allowed
|= 1 << OVS_KEY_ATTR_TCP_FLAGS
;
260 if (match
->key
->ip
.proto
== IPPROTO_ICMPV6
) {
261 key_expected
|= 1 << OVS_KEY_ATTR_ICMPV6
;
262 if (match
->mask
&& (match
->mask
->key
.ip
.proto
== 0xff))
263 mask_allowed
|= 1 << OVS_KEY_ATTR_ICMPV6
;
265 if (match
->key
->tp
.src
==
266 htons(NDISC_NEIGHBOUR_SOLICITATION
) ||
267 match
->key
->tp
.src
== htons(NDISC_NEIGHBOUR_ADVERTISEMENT
)) {
268 key_expected
|= 1 << OVS_KEY_ATTR_ND
;
269 /* Original direction conntrack tuple
270 * uses the same space as the ND fields
271 * in the key, so both are not allowed
274 mask_allowed
&= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6
);
275 if (match
->mask
&& (match
->mask
->key
.tp
.src
== htons(0xff)))
276 mask_allowed
|= 1 << OVS_KEY_ATTR_ND
;
282 if (match
->key
->eth
.type
== htons(ETH_P_NSH
)) {
283 key_expected
|= 1 << OVS_KEY_ATTR_NSH
;
285 match
->mask
->key
.eth
.type
== htons(0xffff)) {
286 mask_allowed
|= 1 << OVS_KEY_ATTR_NSH
;
290 if ((key_attrs
& key_expected
) != key_expected
) {
291 /* Key attributes check failed. */
292 OVS_NLERR(log
, "Missing key (keys=%llx, expected=%llx)",
293 (unsigned long long)key_attrs
,
294 (unsigned long long)key_expected
);
298 if ((mask_attrs
& mask_allowed
) != mask_attrs
) {
299 /* Mask attributes check failed. */
300 OVS_NLERR(log
, "Unexpected mask (mask=%llx, allowed=%llx)",
301 (unsigned long long)mask_attrs
,
302 (unsigned long long)mask_allowed
);
309 size_t ovs_tun_key_attr_size(void)
311 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
312 * updating this function.
314 return nla_total_size_64bit(8) /* OVS_TUNNEL_KEY_ATTR_ID */
315 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
316 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_DST */
317 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
318 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
319 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
320 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
321 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
322 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
323 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS and
324 * OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS is mutually exclusive with
325 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
327 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
328 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
331 static size_t ovs_nsh_key_attr_size(void)
333 /* Whenever adding new OVS_NSH_KEY_ FIELDS, we should consider
334 * updating this function.
336 return nla_total_size(NSH_BASE_HDR_LEN
) /* OVS_NSH_KEY_ATTR_BASE */
337 /* OVS_NSH_KEY_ATTR_MD1 and OVS_NSH_KEY_ATTR_MD2 are
338 * mutually exclusive, so the bigger one can cover
341 + nla_total_size(NSH_CTX_HDRS_MAX_LEN
);
344 size_t ovs_key_attr_size(void)
346 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
347 * updating this function.
349 BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO
!= 29);
351 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
352 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
353 + ovs_tun_key_attr_size()
354 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
355 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
356 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
357 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
358 + nla_total_size(4) /* OVS_KEY_ATTR_CT_STATE */
359 + nla_total_size(2) /* OVS_KEY_ATTR_CT_ZONE */
360 + nla_total_size(4) /* OVS_KEY_ATTR_CT_MARK */
361 + nla_total_size(16) /* OVS_KEY_ATTR_CT_LABELS */
362 + nla_total_size(40) /* OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6 */
363 + nla_total_size(0) /* OVS_KEY_ATTR_NSH */
364 + ovs_nsh_key_attr_size()
365 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
366 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
367 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
368 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
369 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
370 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
371 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
372 + nla_total_size(28); /* OVS_KEY_ATTR_ND */
375 static const struct ovs_len_tbl ovs_vxlan_ext_key_lens
[OVS_VXLAN_EXT_MAX
+ 1] = {
376 [OVS_VXLAN_EXT_GBP
] = { .len
= sizeof(u32
) },
379 static const struct ovs_len_tbl ovs_tunnel_key_lens
[OVS_TUNNEL_KEY_ATTR_MAX
+ 1] = {
380 [OVS_TUNNEL_KEY_ATTR_ID
] = { .len
= sizeof(u64
) },
381 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC
] = { .len
= sizeof(u32
) },
382 [OVS_TUNNEL_KEY_ATTR_IPV4_DST
] = { .len
= sizeof(u32
) },
383 [OVS_TUNNEL_KEY_ATTR_TOS
] = { .len
= 1 },
384 [OVS_TUNNEL_KEY_ATTR_TTL
] = { .len
= 1 },
385 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT
] = { .len
= 0 },
386 [OVS_TUNNEL_KEY_ATTR_CSUM
] = { .len
= 0 },
387 [OVS_TUNNEL_KEY_ATTR_TP_SRC
] = { .len
= sizeof(u16
) },
388 [OVS_TUNNEL_KEY_ATTR_TP_DST
] = { .len
= sizeof(u16
) },
389 [OVS_TUNNEL_KEY_ATTR_OAM
] = { .len
= 0 },
390 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS
] = { .len
= OVS_ATTR_VARIABLE
},
391 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS
] = { .len
= OVS_ATTR_NESTED
,
392 .next
= ovs_vxlan_ext_key_lens
},
393 [OVS_TUNNEL_KEY_ATTR_IPV6_SRC
] = { .len
= sizeof(struct in6_addr
) },
394 [OVS_TUNNEL_KEY_ATTR_IPV6_DST
] = { .len
= sizeof(struct in6_addr
) },
395 [OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS
] = { .len
= OVS_ATTR_VARIABLE
},
396 [OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE
] = { .len
= 0 },
399 static const struct ovs_len_tbl
400 ovs_nsh_key_attr_lens
[OVS_NSH_KEY_ATTR_MAX
+ 1] = {
401 [OVS_NSH_KEY_ATTR_BASE
] = { .len
= sizeof(struct ovs_nsh_key_base
) },
402 [OVS_NSH_KEY_ATTR_MD1
] = { .len
= sizeof(struct ovs_nsh_key_md1
) },
403 [OVS_NSH_KEY_ATTR_MD2
] = { .len
= OVS_ATTR_VARIABLE
},
406 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
407 static const struct ovs_len_tbl ovs_key_lens
[OVS_KEY_ATTR_MAX
+ 1] = {
408 [OVS_KEY_ATTR_ENCAP
] = { .len
= OVS_ATTR_NESTED
},
409 [OVS_KEY_ATTR_PRIORITY
] = { .len
= sizeof(u32
) },
410 [OVS_KEY_ATTR_IN_PORT
] = { .len
= sizeof(u32
) },
411 [OVS_KEY_ATTR_SKB_MARK
] = { .len
= sizeof(u32
) },
412 [OVS_KEY_ATTR_ETHERNET
] = { .len
= sizeof(struct ovs_key_ethernet
) },
413 [OVS_KEY_ATTR_VLAN
] = { .len
= sizeof(__be16
) },
414 [OVS_KEY_ATTR_ETHERTYPE
] = { .len
= sizeof(__be16
) },
415 [OVS_KEY_ATTR_IPV4
] = { .len
= sizeof(struct ovs_key_ipv4
) },
416 [OVS_KEY_ATTR_IPV6
] = { .len
= sizeof(struct ovs_key_ipv6
) },
417 [OVS_KEY_ATTR_TCP
] = { .len
= sizeof(struct ovs_key_tcp
) },
418 [OVS_KEY_ATTR_TCP_FLAGS
] = { .len
= sizeof(__be16
) },
419 [OVS_KEY_ATTR_UDP
] = { .len
= sizeof(struct ovs_key_udp
) },
420 [OVS_KEY_ATTR_SCTP
] = { .len
= sizeof(struct ovs_key_sctp
) },
421 [OVS_KEY_ATTR_ICMP
] = { .len
= sizeof(struct ovs_key_icmp
) },
422 [OVS_KEY_ATTR_ICMPV6
] = { .len
= sizeof(struct ovs_key_icmpv6
) },
423 [OVS_KEY_ATTR_ARP
] = { .len
= sizeof(struct ovs_key_arp
) },
424 [OVS_KEY_ATTR_ND
] = { .len
= sizeof(struct ovs_key_nd
) },
425 [OVS_KEY_ATTR_RECIRC_ID
] = { .len
= sizeof(u32
) },
426 [OVS_KEY_ATTR_DP_HASH
] = { .len
= sizeof(u32
) },
427 [OVS_KEY_ATTR_TUNNEL
] = { .len
= OVS_ATTR_NESTED
,
428 .next
= ovs_tunnel_key_lens
, },
429 [OVS_KEY_ATTR_MPLS
] = { .len
= OVS_ATTR_VARIABLE
},
430 [OVS_KEY_ATTR_CT_STATE
] = { .len
= sizeof(u32
) },
431 [OVS_KEY_ATTR_CT_ZONE
] = { .len
= sizeof(u16
) },
432 [OVS_KEY_ATTR_CT_MARK
] = { .len
= sizeof(u32
) },
433 [OVS_KEY_ATTR_CT_LABELS
] = { .len
= sizeof(struct ovs_key_ct_labels
) },
434 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4
] = {
435 .len
= sizeof(struct ovs_key_ct_tuple_ipv4
) },
436 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6
] = {
437 .len
= sizeof(struct ovs_key_ct_tuple_ipv6
) },
438 [OVS_KEY_ATTR_NSH
] = { .len
= OVS_ATTR_NESTED
,
439 .next
= ovs_nsh_key_attr_lens
, },
442 static bool check_attr_len(unsigned int attr_len
, unsigned int expected_len
)
444 return expected_len
== attr_len
||
445 expected_len
== OVS_ATTR_NESTED
||
446 expected_len
== OVS_ATTR_VARIABLE
;
449 static bool is_all_zero(const u8
*fp
, size_t size
)
456 for (i
= 0; i
< size
; i
++)
463 static int __parse_flow_nlattrs(const struct nlattr
*attr
,
464 const struct nlattr
*a
[],
465 u64
*attrsp
, bool log
, bool nz
)
467 const struct nlattr
*nla
;
472 nla_for_each_nested(nla
, attr
, rem
) {
473 u16 type
= nla_type(nla
);
476 if (type
> OVS_KEY_ATTR_MAX
) {
477 OVS_NLERR(log
, "Key type %d is out of range max %d",
478 type
, OVS_KEY_ATTR_MAX
);
482 if (attrs
& (1 << type
)) {
483 OVS_NLERR(log
, "Duplicate key (type %d).", type
);
487 expected_len
= ovs_key_lens
[type
].len
;
488 if (!check_attr_len(nla_len(nla
), expected_len
)) {
489 OVS_NLERR(log
, "Key %d has unexpected len %d expected %d",
490 type
, nla_len(nla
), expected_len
);
494 if (!nz
|| !is_all_zero(nla_data(nla
), nla_len(nla
))) {
500 OVS_NLERR(log
, "Message has %d unknown bytes.", rem
);
508 static int parse_flow_mask_nlattrs(const struct nlattr
*attr
,
509 const struct nlattr
*a
[], u64
*attrsp
,
512 return __parse_flow_nlattrs(attr
, a
, attrsp
, log
, true);
515 int parse_flow_nlattrs(const struct nlattr
*attr
, const struct nlattr
*a
[],
516 u64
*attrsp
, bool log
)
518 return __parse_flow_nlattrs(attr
, a
, attrsp
, log
, false);
521 static int genev_tun_opt_from_nlattr(const struct nlattr
*a
,
522 struct sw_flow_match
*match
, bool is_mask
,
525 unsigned long opt_key_offset
;
527 if (nla_len(a
) > sizeof(match
->key
->tun_opts
)) {
528 OVS_NLERR(log
, "Geneve option length err (len %d, max %zu).",
529 nla_len(a
), sizeof(match
->key
->tun_opts
));
533 if (nla_len(a
) % 4 != 0) {
534 OVS_NLERR(log
, "Geneve opt len %d is not a multiple of 4.",
539 /* We need to record the length of the options passed
540 * down, otherwise packets with the same format but
541 * additional options will be silently matched.
544 SW_FLOW_KEY_PUT(match
, tun_opts_len
, nla_len(a
),
547 /* This is somewhat unusual because it looks at
548 * both the key and mask while parsing the
549 * attributes (and by extension assumes the key
550 * is parsed first). Normally, we would verify
551 * that each is the correct length and that the
552 * attributes line up in the validate function.
553 * However, that is difficult because this is
554 * variable length and we won't have the
557 if (match
->key
->tun_opts_len
!= nla_len(a
)) {
558 OVS_NLERR(log
, "Geneve option len %d != mask len %d",
559 match
->key
->tun_opts_len
, nla_len(a
));
563 SW_FLOW_KEY_PUT(match
, tun_opts_len
, 0xff, true);
566 opt_key_offset
= TUN_METADATA_OFFSET(nla_len(a
));
567 SW_FLOW_KEY_MEMCPY_OFFSET(match
, opt_key_offset
, nla_data(a
),
568 nla_len(a
), is_mask
);
572 static int vxlan_tun_opt_from_nlattr(const struct nlattr
*attr
,
573 struct sw_flow_match
*match
, bool is_mask
,
578 unsigned long opt_key_offset
;
579 struct vxlan_metadata opts
;
581 BUILD_BUG_ON(sizeof(opts
) > sizeof(match
->key
->tun_opts
));
583 memset(&opts
, 0, sizeof(opts
));
584 nla_for_each_nested(a
, attr
, rem
) {
585 int type
= nla_type(a
);
587 if (type
> OVS_VXLAN_EXT_MAX
) {
588 OVS_NLERR(log
, "VXLAN extension %d out of range max %d",
589 type
, OVS_VXLAN_EXT_MAX
);
593 if (!check_attr_len(nla_len(a
),
594 ovs_vxlan_ext_key_lens
[type
].len
)) {
595 OVS_NLERR(log
, "VXLAN extension %d has unexpected len %d expected %d",
597 ovs_vxlan_ext_key_lens
[type
].len
);
602 case OVS_VXLAN_EXT_GBP
:
603 opts
.gbp
= nla_get_u32(a
);
606 OVS_NLERR(log
, "Unknown VXLAN extension attribute %d",
612 OVS_NLERR(log
, "VXLAN extension message has %d unknown bytes.",
618 SW_FLOW_KEY_PUT(match
, tun_opts_len
, sizeof(opts
), false);
620 SW_FLOW_KEY_PUT(match
, tun_opts_len
, 0xff, true);
622 opt_key_offset
= TUN_METADATA_OFFSET(sizeof(opts
));
623 SW_FLOW_KEY_MEMCPY_OFFSET(match
, opt_key_offset
, &opts
, sizeof(opts
),
628 static int erspan_tun_opt_from_nlattr(const struct nlattr
*a
,
629 struct sw_flow_match
*match
, bool is_mask
,
632 unsigned long opt_key_offset
;
634 BUILD_BUG_ON(sizeof(struct erspan_metadata
) >
635 sizeof(match
->key
->tun_opts
));
637 if (nla_len(a
) > sizeof(match
->key
->tun_opts
)) {
638 OVS_NLERR(log
, "ERSPAN option length err (len %d, max %zu).",
639 nla_len(a
), sizeof(match
->key
->tun_opts
));
644 SW_FLOW_KEY_PUT(match
, tun_opts_len
,
645 sizeof(struct erspan_metadata
), false);
647 SW_FLOW_KEY_PUT(match
, tun_opts_len
, 0xff, true);
649 opt_key_offset
= TUN_METADATA_OFFSET(nla_len(a
));
650 SW_FLOW_KEY_MEMCPY_OFFSET(match
, opt_key_offset
, nla_data(a
),
651 nla_len(a
), is_mask
);
655 static int ip_tun_from_nlattr(const struct nlattr
*attr
,
656 struct sw_flow_match
*match
, bool is_mask
,
659 bool ttl
= false, ipv4
= false, ipv6
= false;
660 bool info_bridge_mode
= false;
661 __be16 tun_flags
= 0;
666 nla_for_each_nested(a
, attr
, rem
) {
667 int type
= nla_type(a
);
670 if (type
> OVS_TUNNEL_KEY_ATTR_MAX
) {
671 OVS_NLERR(log
, "Tunnel attr %d out of range max %d",
672 type
, OVS_TUNNEL_KEY_ATTR_MAX
);
676 if (!check_attr_len(nla_len(a
),
677 ovs_tunnel_key_lens
[type
].len
)) {
678 OVS_NLERR(log
, "Tunnel attr %d has unexpected len %d expected %d",
679 type
, nla_len(a
), ovs_tunnel_key_lens
[type
].len
);
684 case OVS_TUNNEL_KEY_ATTR_ID
:
685 SW_FLOW_KEY_PUT(match
, tun_key
.tun_id
,
686 nla_get_be64(a
), is_mask
);
687 tun_flags
|= TUNNEL_KEY
;
689 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC
:
690 SW_FLOW_KEY_PUT(match
, tun_key
.u
.ipv4
.src
,
691 nla_get_in_addr(a
), is_mask
);
694 case OVS_TUNNEL_KEY_ATTR_IPV4_DST
:
695 SW_FLOW_KEY_PUT(match
, tun_key
.u
.ipv4
.dst
,
696 nla_get_in_addr(a
), is_mask
);
699 case OVS_TUNNEL_KEY_ATTR_IPV6_SRC
:
700 SW_FLOW_KEY_PUT(match
, tun_key
.u
.ipv6
.src
,
701 nla_get_in6_addr(a
), is_mask
);
704 case OVS_TUNNEL_KEY_ATTR_IPV6_DST
:
705 SW_FLOW_KEY_PUT(match
, tun_key
.u
.ipv6
.dst
,
706 nla_get_in6_addr(a
), is_mask
);
709 case OVS_TUNNEL_KEY_ATTR_TOS
:
710 SW_FLOW_KEY_PUT(match
, tun_key
.tos
,
711 nla_get_u8(a
), is_mask
);
713 case OVS_TUNNEL_KEY_ATTR_TTL
:
714 SW_FLOW_KEY_PUT(match
, tun_key
.ttl
,
715 nla_get_u8(a
), is_mask
);
718 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT
:
719 tun_flags
|= TUNNEL_DONT_FRAGMENT
;
721 case OVS_TUNNEL_KEY_ATTR_CSUM
:
722 tun_flags
|= TUNNEL_CSUM
;
724 case OVS_TUNNEL_KEY_ATTR_TP_SRC
:
725 SW_FLOW_KEY_PUT(match
, tun_key
.tp_src
,
726 nla_get_be16(a
), is_mask
);
728 case OVS_TUNNEL_KEY_ATTR_TP_DST
:
729 SW_FLOW_KEY_PUT(match
, tun_key
.tp_dst
,
730 nla_get_be16(a
), is_mask
);
732 case OVS_TUNNEL_KEY_ATTR_OAM
:
733 tun_flags
|= TUNNEL_OAM
;
735 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS
:
737 OVS_NLERR(log
, "Multiple metadata blocks provided");
741 err
= genev_tun_opt_from_nlattr(a
, match
, is_mask
, log
);
745 tun_flags
|= TUNNEL_GENEVE_OPT
;
748 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS
:
750 OVS_NLERR(log
, "Multiple metadata blocks provided");
754 err
= vxlan_tun_opt_from_nlattr(a
, match
, is_mask
, log
);
758 tun_flags
|= TUNNEL_VXLAN_OPT
;
761 case OVS_TUNNEL_KEY_ATTR_PAD
:
763 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS
:
765 OVS_NLERR(log
, "Multiple metadata blocks provided");
769 err
= erspan_tun_opt_from_nlattr(a
, match
, is_mask
,
774 tun_flags
|= TUNNEL_ERSPAN_OPT
;
777 case OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE
:
778 info_bridge_mode
= true;
782 OVS_NLERR(log
, "Unknown IP tunnel attribute %d",
788 SW_FLOW_KEY_PUT(match
, tun_key
.tun_flags
, tun_flags
, is_mask
);
790 SW_FLOW_KEY_MEMSET_FIELD(match
, tun_proto
, 0xff, true);
792 SW_FLOW_KEY_PUT(match
, tun_proto
, ipv6
? AF_INET6
: AF_INET
,
796 OVS_NLERR(log
, "IP tunnel attribute has %d unknown bytes.",
802 OVS_NLERR(log
, "Mixed IPv4 and IPv6 tunnel attributes");
807 if (!ipv4
&& !ipv6
) {
808 OVS_NLERR(log
, "IP tunnel dst address not specified");
812 if (info_bridge_mode
) {
813 if (match
->key
->tun_key
.u
.ipv4
.src
||
814 match
->key
->tun_key
.u
.ipv4
.dst
||
815 match
->key
->tun_key
.tp_src
||
816 match
->key
->tun_key
.tp_dst
||
817 match
->key
->tun_key
.ttl
||
818 match
->key
->tun_key
.tos
||
819 tun_flags
& ~TUNNEL_KEY
) {
820 OVS_NLERR(log
, "IPv4 tun info is not correct");
823 } else if (!match
->key
->tun_key
.u
.ipv4
.dst
) {
824 OVS_NLERR(log
, "IPv4 tunnel dst address is zero");
828 if (ipv6
&& ipv6_addr_any(&match
->key
->tun_key
.u
.ipv6
.dst
)) {
829 OVS_NLERR(log
, "IPv6 tunnel dst address is zero");
833 if (!ttl
&& !info_bridge_mode
) {
834 OVS_NLERR(log
, "IP tunnel TTL not specified.");
842 static int vxlan_opt_to_nlattr(struct sk_buff
*skb
,
843 const void *tun_opts
, int swkey_tun_opts_len
)
845 const struct vxlan_metadata
*opts
= tun_opts
;
848 nla
= nla_nest_start_noflag(skb
, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS
);
852 if (nla_put_u32(skb
, OVS_VXLAN_EXT_GBP
, opts
->gbp
) < 0)
855 nla_nest_end(skb
, nla
);
859 static int __ip_tun_to_nlattr(struct sk_buff
*skb
,
860 const struct ip_tunnel_key
*output
,
861 const void *tun_opts
, int swkey_tun_opts_len
,
862 unsigned short tun_proto
, u8 mode
)
864 if (output
->tun_flags
& TUNNEL_KEY
&&
865 nla_put_be64(skb
, OVS_TUNNEL_KEY_ATTR_ID
, output
->tun_id
,
866 OVS_TUNNEL_KEY_ATTR_PAD
))
869 if (mode
& IP_TUNNEL_INFO_BRIDGE
)
870 return nla_put_flag(skb
, OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE
)
875 if (output
->u
.ipv4
.src
&&
876 nla_put_in_addr(skb
, OVS_TUNNEL_KEY_ATTR_IPV4_SRC
,
879 if (output
->u
.ipv4
.dst
&&
880 nla_put_in_addr(skb
, OVS_TUNNEL_KEY_ATTR_IPV4_DST
,
885 if (!ipv6_addr_any(&output
->u
.ipv6
.src
) &&
886 nla_put_in6_addr(skb
, OVS_TUNNEL_KEY_ATTR_IPV6_SRC
,
887 &output
->u
.ipv6
.src
))
889 if (!ipv6_addr_any(&output
->u
.ipv6
.dst
) &&
890 nla_put_in6_addr(skb
, OVS_TUNNEL_KEY_ATTR_IPV6_DST
,
891 &output
->u
.ipv6
.dst
))
896 nla_put_u8(skb
, OVS_TUNNEL_KEY_ATTR_TOS
, output
->tos
))
898 if (nla_put_u8(skb
, OVS_TUNNEL_KEY_ATTR_TTL
, output
->ttl
))
900 if ((output
->tun_flags
& TUNNEL_DONT_FRAGMENT
) &&
901 nla_put_flag(skb
, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT
))
903 if ((output
->tun_flags
& TUNNEL_CSUM
) &&
904 nla_put_flag(skb
, OVS_TUNNEL_KEY_ATTR_CSUM
))
906 if (output
->tp_src
&&
907 nla_put_be16(skb
, OVS_TUNNEL_KEY_ATTR_TP_SRC
, output
->tp_src
))
909 if (output
->tp_dst
&&
910 nla_put_be16(skb
, OVS_TUNNEL_KEY_ATTR_TP_DST
, output
->tp_dst
))
912 if ((output
->tun_flags
& TUNNEL_OAM
) &&
913 nla_put_flag(skb
, OVS_TUNNEL_KEY_ATTR_OAM
))
915 if (swkey_tun_opts_len
) {
916 if (output
->tun_flags
& TUNNEL_GENEVE_OPT
&&
917 nla_put(skb
, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS
,
918 swkey_tun_opts_len
, tun_opts
))
920 else if (output
->tun_flags
& TUNNEL_VXLAN_OPT
&&
921 vxlan_opt_to_nlattr(skb
, tun_opts
, swkey_tun_opts_len
))
923 else if (output
->tun_flags
& TUNNEL_ERSPAN_OPT
&&
924 nla_put(skb
, OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS
,
925 swkey_tun_opts_len
, tun_opts
))
932 static int ip_tun_to_nlattr(struct sk_buff
*skb
,
933 const struct ip_tunnel_key
*output
,
934 const void *tun_opts
, int swkey_tun_opts_len
,
935 unsigned short tun_proto
, u8 mode
)
940 nla
= nla_nest_start_noflag(skb
, OVS_KEY_ATTR_TUNNEL
);
944 err
= __ip_tun_to_nlattr(skb
, output
, tun_opts
, swkey_tun_opts_len
,
949 nla_nest_end(skb
, nla
);
953 int ovs_nla_put_tunnel_info(struct sk_buff
*skb
,
954 struct ip_tunnel_info
*tun_info
)
956 return __ip_tun_to_nlattr(skb
, &tun_info
->key
,
957 ip_tunnel_info_opts(tun_info
),
958 tun_info
->options_len
,
959 ip_tunnel_info_af(tun_info
), tun_info
->mode
);
962 static int encode_vlan_from_nlattrs(struct sw_flow_match
*match
,
963 const struct nlattr
*a
[],
964 bool is_mask
, bool inner
)
969 if (a
[OVS_KEY_ATTR_VLAN
])
970 tci
= nla_get_be16(a
[OVS_KEY_ATTR_VLAN
]);
972 if (a
[OVS_KEY_ATTR_ETHERTYPE
])
973 tpid
= nla_get_be16(a
[OVS_KEY_ATTR_ETHERTYPE
]);
975 if (likely(!inner
)) {
976 SW_FLOW_KEY_PUT(match
, eth
.vlan
.tpid
, tpid
, is_mask
);
977 SW_FLOW_KEY_PUT(match
, eth
.vlan
.tci
, tci
, is_mask
);
979 SW_FLOW_KEY_PUT(match
, eth
.cvlan
.tpid
, tpid
, is_mask
);
980 SW_FLOW_KEY_PUT(match
, eth
.cvlan
.tci
, tci
, is_mask
);
985 static int validate_vlan_from_nlattrs(const struct sw_flow_match
*match
,
986 u64 key_attrs
, bool inner
,
987 const struct nlattr
**a
, bool log
)
991 if (!((key_attrs
& (1 << OVS_KEY_ATTR_ETHERNET
)) &&
992 (key_attrs
& (1 << OVS_KEY_ATTR_ETHERTYPE
)) &&
993 eth_type_vlan(nla_get_be16(a
[OVS_KEY_ATTR_ETHERTYPE
])))) {
998 if (!((key_attrs
& (1 << OVS_KEY_ATTR_VLAN
)) &&
999 (key_attrs
& (1 << OVS_KEY_ATTR_ENCAP
)))) {
1000 OVS_NLERR(log
, "Invalid %s frame", (inner
) ? "C-VLAN" : "VLAN");
1004 if (a
[OVS_KEY_ATTR_VLAN
])
1005 tci
= nla_get_be16(a
[OVS_KEY_ATTR_VLAN
]);
1007 if (!(tci
& htons(VLAN_CFI_MASK
))) {
1009 OVS_NLERR(log
, "%s TCI does not have VLAN_CFI_MASK bit set.",
1010 (inner
) ? "C-VLAN" : "VLAN");
1012 } else if (nla_len(a
[OVS_KEY_ATTR_ENCAP
])) {
1013 /* Corner case for truncated VLAN header. */
1014 OVS_NLERR(log
, "Truncated %s header has non-zero encap attribute.",
1015 (inner
) ? "C-VLAN" : "VLAN");
1023 static int validate_vlan_mask_from_nlattrs(const struct sw_flow_match
*match
,
1024 u64 key_attrs
, bool inner
,
1025 const struct nlattr
**a
, bool log
)
1029 bool encap_valid
= !!(match
->key
->eth
.vlan
.tci
&
1030 htons(VLAN_CFI_MASK
));
1031 bool i_encap_valid
= !!(match
->key
->eth
.cvlan
.tci
&
1032 htons(VLAN_CFI_MASK
));
1034 if (!(key_attrs
& (1 << OVS_KEY_ATTR_ENCAP
))) {
1039 if ((!inner
&& !encap_valid
) || (inner
&& !i_encap_valid
)) {
1040 OVS_NLERR(log
, "Encap mask attribute is set for non-%s frame.",
1041 (inner
) ? "C-VLAN" : "VLAN");
1045 if (a
[OVS_KEY_ATTR_VLAN
])
1046 tci
= nla_get_be16(a
[OVS_KEY_ATTR_VLAN
]);
1048 if (a
[OVS_KEY_ATTR_ETHERTYPE
])
1049 tpid
= nla_get_be16(a
[OVS_KEY_ATTR_ETHERTYPE
]);
1051 if (tpid
!= htons(0xffff)) {
1052 OVS_NLERR(log
, "Must have an exact match on %s TPID (mask=%x).",
1053 (inner
) ? "C-VLAN" : "VLAN", ntohs(tpid
));
1056 if (!(tci
& htons(VLAN_CFI_MASK
))) {
1057 OVS_NLERR(log
, "%s TCI mask does not have exact match for VLAN_CFI_MASK bit.",
1058 (inner
) ? "C-VLAN" : "VLAN");
1065 static int __parse_vlan_from_nlattrs(struct sw_flow_match
*match
,
1066 u64
*key_attrs
, bool inner
,
1067 const struct nlattr
**a
, bool is_mask
,
1071 const struct nlattr
*encap
;
1074 err
= validate_vlan_from_nlattrs(match
, *key_attrs
, inner
,
1077 err
= validate_vlan_mask_from_nlattrs(match
, *key_attrs
, inner
,
1082 err
= encode_vlan_from_nlattrs(match
, a
, is_mask
, inner
);
1086 *key_attrs
&= ~(1 << OVS_KEY_ATTR_ENCAP
);
1087 *key_attrs
&= ~(1 << OVS_KEY_ATTR_VLAN
);
1088 *key_attrs
&= ~(1 << OVS_KEY_ATTR_ETHERTYPE
);
1090 encap
= a
[OVS_KEY_ATTR_ENCAP
];
1093 err
= parse_flow_nlattrs(encap
, a
, key_attrs
, log
);
1095 err
= parse_flow_mask_nlattrs(encap
, a
, key_attrs
, log
);
1100 static int parse_vlan_from_nlattrs(struct sw_flow_match
*match
,
1101 u64
*key_attrs
, const struct nlattr
**a
,
1102 bool is_mask
, bool log
)
1105 bool encap_valid
= false;
1107 err
= __parse_vlan_from_nlattrs(match
, key_attrs
, false, a
,
1112 encap_valid
= !!(match
->key
->eth
.vlan
.tci
& htons(VLAN_CFI_MASK
));
1114 err
= __parse_vlan_from_nlattrs(match
, key_attrs
, true, a
,
1123 static int parse_eth_type_from_nlattrs(struct sw_flow_match
*match
,
1124 u64
*attrs
, const struct nlattr
**a
,
1125 bool is_mask
, bool log
)
1129 eth_type
= nla_get_be16(a
[OVS_KEY_ATTR_ETHERTYPE
]);
1131 /* Always exact match EtherType. */
1132 eth_type
= htons(0xffff);
1133 } else if (!eth_proto_is_802_3(eth_type
)) {
1134 OVS_NLERR(log
, "EtherType %x is less than min %x",
1135 ntohs(eth_type
), ETH_P_802_3_MIN
);
1139 SW_FLOW_KEY_PUT(match
, eth
.type
, eth_type
, is_mask
);
1140 *attrs
&= ~(1 << OVS_KEY_ATTR_ETHERTYPE
);
1144 static int metadata_from_nlattrs(struct net
*net
, struct sw_flow_match
*match
,
1145 u64
*attrs
, const struct nlattr
**a
,
1146 bool is_mask
, bool log
)
1148 u8 mac_proto
= MAC_PROTO_ETHERNET
;
1150 if (*attrs
& (1 << OVS_KEY_ATTR_DP_HASH
)) {
1151 u32 hash_val
= nla_get_u32(a
[OVS_KEY_ATTR_DP_HASH
]);
1153 SW_FLOW_KEY_PUT(match
, ovs_flow_hash
, hash_val
, is_mask
);
1154 *attrs
&= ~(1 << OVS_KEY_ATTR_DP_HASH
);
1157 if (*attrs
& (1 << OVS_KEY_ATTR_RECIRC_ID
)) {
1158 u32 recirc_id
= nla_get_u32(a
[OVS_KEY_ATTR_RECIRC_ID
]);
1160 SW_FLOW_KEY_PUT(match
, recirc_id
, recirc_id
, is_mask
);
1161 *attrs
&= ~(1 << OVS_KEY_ATTR_RECIRC_ID
);
1164 if (*attrs
& (1 << OVS_KEY_ATTR_PRIORITY
)) {
1165 SW_FLOW_KEY_PUT(match
, phy
.priority
,
1166 nla_get_u32(a
[OVS_KEY_ATTR_PRIORITY
]), is_mask
);
1167 *attrs
&= ~(1 << OVS_KEY_ATTR_PRIORITY
);
1170 if (*attrs
& (1 << OVS_KEY_ATTR_IN_PORT
)) {
1171 u32 in_port
= nla_get_u32(a
[OVS_KEY_ATTR_IN_PORT
]);
1174 in_port
= 0xffffffff; /* Always exact match in_port. */
1175 } else if (in_port
>= DP_MAX_PORTS
) {
1176 OVS_NLERR(log
, "Port %d exceeds max allowable %d",
1177 in_port
, DP_MAX_PORTS
);
1181 SW_FLOW_KEY_PUT(match
, phy
.in_port
, in_port
, is_mask
);
1182 *attrs
&= ~(1 << OVS_KEY_ATTR_IN_PORT
);
1183 } else if (!is_mask
) {
1184 SW_FLOW_KEY_PUT(match
, phy
.in_port
, DP_MAX_PORTS
, is_mask
);
1187 if (*attrs
& (1 << OVS_KEY_ATTR_SKB_MARK
)) {
1188 uint32_t mark
= nla_get_u32(a
[OVS_KEY_ATTR_SKB_MARK
]);
1190 SW_FLOW_KEY_PUT(match
, phy
.skb_mark
, mark
, is_mask
);
1191 *attrs
&= ~(1 << OVS_KEY_ATTR_SKB_MARK
);
1193 if (*attrs
& (1 << OVS_KEY_ATTR_TUNNEL
)) {
1194 if (ip_tun_from_nlattr(a
[OVS_KEY_ATTR_TUNNEL
], match
,
1197 *attrs
&= ~(1 << OVS_KEY_ATTR_TUNNEL
);
1200 if (*attrs
& (1 << OVS_KEY_ATTR_CT_STATE
) &&
1201 ovs_ct_verify(net
, OVS_KEY_ATTR_CT_STATE
)) {
1202 u32 ct_state
= nla_get_u32(a
[OVS_KEY_ATTR_CT_STATE
]);
1204 if (ct_state
& ~CT_SUPPORTED_MASK
) {
1205 OVS_NLERR(log
, "ct_state flags %08x unsupported",
1210 SW_FLOW_KEY_PUT(match
, ct_state
, ct_state
, is_mask
);
1211 *attrs
&= ~(1ULL << OVS_KEY_ATTR_CT_STATE
);
1213 if (*attrs
& (1 << OVS_KEY_ATTR_CT_ZONE
) &&
1214 ovs_ct_verify(net
, OVS_KEY_ATTR_CT_ZONE
)) {
1215 u16 ct_zone
= nla_get_u16(a
[OVS_KEY_ATTR_CT_ZONE
]);
1217 SW_FLOW_KEY_PUT(match
, ct_zone
, ct_zone
, is_mask
);
1218 *attrs
&= ~(1ULL << OVS_KEY_ATTR_CT_ZONE
);
1220 if (*attrs
& (1 << OVS_KEY_ATTR_CT_MARK
) &&
1221 ovs_ct_verify(net
, OVS_KEY_ATTR_CT_MARK
)) {
1222 u32 mark
= nla_get_u32(a
[OVS_KEY_ATTR_CT_MARK
]);
1224 SW_FLOW_KEY_PUT(match
, ct
.mark
, mark
, is_mask
);
1225 *attrs
&= ~(1ULL << OVS_KEY_ATTR_CT_MARK
);
1227 if (*attrs
& (1 << OVS_KEY_ATTR_CT_LABELS
) &&
1228 ovs_ct_verify(net
, OVS_KEY_ATTR_CT_LABELS
)) {
1229 const struct ovs_key_ct_labels
*cl
;
1231 cl
= nla_data(a
[OVS_KEY_ATTR_CT_LABELS
]);
1232 SW_FLOW_KEY_MEMCPY(match
, ct
.labels
, cl
->ct_labels
,
1233 sizeof(*cl
), is_mask
);
1234 *attrs
&= ~(1ULL << OVS_KEY_ATTR_CT_LABELS
);
1236 if (*attrs
& (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4
)) {
1237 const struct ovs_key_ct_tuple_ipv4
*ct
;
1239 ct
= nla_data(a
[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4
]);
1241 SW_FLOW_KEY_PUT(match
, ipv4
.ct_orig
.src
, ct
->ipv4_src
, is_mask
);
1242 SW_FLOW_KEY_PUT(match
, ipv4
.ct_orig
.dst
, ct
->ipv4_dst
, is_mask
);
1243 SW_FLOW_KEY_PUT(match
, ct
.orig_tp
.src
, ct
->src_port
, is_mask
);
1244 SW_FLOW_KEY_PUT(match
, ct
.orig_tp
.dst
, ct
->dst_port
, is_mask
);
1245 SW_FLOW_KEY_PUT(match
, ct_orig_proto
, ct
->ipv4_proto
, is_mask
);
1246 *attrs
&= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4
);
1248 if (*attrs
& (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6
)) {
1249 const struct ovs_key_ct_tuple_ipv6
*ct
;
1251 ct
= nla_data(a
[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6
]);
1253 SW_FLOW_KEY_MEMCPY(match
, ipv6
.ct_orig
.src
, &ct
->ipv6_src
,
1254 sizeof(match
->key
->ipv6
.ct_orig
.src
),
1256 SW_FLOW_KEY_MEMCPY(match
, ipv6
.ct_orig
.dst
, &ct
->ipv6_dst
,
1257 sizeof(match
->key
->ipv6
.ct_orig
.dst
),
1259 SW_FLOW_KEY_PUT(match
, ct
.orig_tp
.src
, ct
->src_port
, is_mask
);
1260 SW_FLOW_KEY_PUT(match
, ct
.orig_tp
.dst
, ct
->dst_port
, is_mask
);
1261 SW_FLOW_KEY_PUT(match
, ct_orig_proto
, ct
->ipv6_proto
, is_mask
);
1262 *attrs
&= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6
);
1265 /* For layer 3 packets the Ethernet type is provided
1266 * and treated as metadata but no MAC addresses are provided.
1268 if (!(*attrs
& (1ULL << OVS_KEY_ATTR_ETHERNET
)) &&
1269 (*attrs
& (1ULL << OVS_KEY_ATTR_ETHERTYPE
)))
1270 mac_proto
= MAC_PROTO_NONE
;
1272 /* Always exact match mac_proto */
1273 SW_FLOW_KEY_PUT(match
, mac_proto
, is_mask
? 0xff : mac_proto
, is_mask
);
1275 if (mac_proto
== MAC_PROTO_NONE
)
1276 return parse_eth_type_from_nlattrs(match
, attrs
, a
, is_mask
,
1282 int nsh_hdr_from_nlattr(const struct nlattr
*attr
,
1283 struct nshhdr
*nh
, size_t size
)
1291 /* validate_nsh has check this, so we needn't do duplicate check here
1293 if (size
< NSH_BASE_HDR_LEN
)
1296 nla_for_each_nested(a
, attr
, rem
) {
1297 int type
= nla_type(a
);
1300 case OVS_NSH_KEY_ATTR_BASE
: {
1301 const struct ovs_nsh_key_base
*base
= nla_data(a
);
1303 flags
= base
->flags
;
1306 nh
->mdtype
= base
->mdtype
;
1307 nh
->path_hdr
= base
->path_hdr
;
1310 case OVS_NSH_KEY_ATTR_MD1
:
1312 if (mdlen
> size
- NSH_BASE_HDR_LEN
)
1314 memcpy(&nh
->md1
, nla_data(a
), mdlen
);
1317 case OVS_NSH_KEY_ATTR_MD2
:
1319 if (mdlen
> size
- NSH_BASE_HDR_LEN
)
1321 memcpy(&nh
->md2
, nla_data(a
), mdlen
);
1329 /* nsh header length = NSH_BASE_HDR_LEN + mdlen */
1330 nh
->ver_flags_ttl_len
= 0;
1331 nsh_set_flags_ttl_len(nh
, flags
, ttl
, NSH_BASE_HDR_LEN
+ mdlen
);
1336 int nsh_key_from_nlattr(const struct nlattr
*attr
,
1337 struct ovs_key_nsh
*nsh
, struct ovs_key_nsh
*nsh_mask
)
1342 /* validate_nsh has check this, so we needn't do duplicate check here
1344 nla_for_each_nested(a
, attr
, rem
) {
1345 int type
= nla_type(a
);
1348 case OVS_NSH_KEY_ATTR_BASE
: {
1349 const struct ovs_nsh_key_base
*base
= nla_data(a
);
1350 const struct ovs_nsh_key_base
*base_mask
= base
+ 1;
1353 nsh_mask
->base
= *base_mask
;
1356 case OVS_NSH_KEY_ATTR_MD1
: {
1357 const struct ovs_nsh_key_md1
*md1
= nla_data(a
);
1358 const struct ovs_nsh_key_md1
*md1_mask
= md1
+ 1;
1360 memcpy(nsh
->context
, md1
->context
, sizeof(*md1
));
1361 memcpy(nsh_mask
->context
, md1_mask
->context
,
1365 case OVS_NSH_KEY_ATTR_MD2
:
1366 /* Not supported yet */
1376 static int nsh_key_put_from_nlattr(const struct nlattr
*attr
,
1377 struct sw_flow_match
*match
, bool is_mask
,
1378 bool is_push_nsh
, bool log
)
1382 bool has_base
= false;
1383 bool has_md1
= false;
1384 bool has_md2
= false;
1388 if (WARN_ON(is_push_nsh
&& is_mask
))
1391 nla_for_each_nested(a
, attr
, rem
) {
1392 int type
= nla_type(a
);
1395 if (type
> OVS_NSH_KEY_ATTR_MAX
) {
1396 OVS_NLERR(log
, "nsh attr %d is out of range max %d",
1397 type
, OVS_NSH_KEY_ATTR_MAX
);
1401 if (!check_attr_len(nla_len(a
),
1402 ovs_nsh_key_attr_lens
[type
].len
)) {
1405 "nsh attr %d has unexpected len %d expected %d",
1408 ovs_nsh_key_attr_lens
[type
].len
1414 case OVS_NSH_KEY_ATTR_BASE
: {
1415 const struct ovs_nsh_key_base
*base
= nla_data(a
);
1418 mdtype
= base
->mdtype
;
1419 SW_FLOW_KEY_PUT(match
, nsh
.base
.flags
,
1420 base
->flags
, is_mask
);
1421 SW_FLOW_KEY_PUT(match
, nsh
.base
.ttl
,
1422 base
->ttl
, is_mask
);
1423 SW_FLOW_KEY_PUT(match
, nsh
.base
.mdtype
,
1424 base
->mdtype
, is_mask
);
1425 SW_FLOW_KEY_PUT(match
, nsh
.base
.np
,
1427 SW_FLOW_KEY_PUT(match
, nsh
.base
.path_hdr
,
1428 base
->path_hdr
, is_mask
);
1431 case OVS_NSH_KEY_ATTR_MD1
: {
1432 const struct ovs_nsh_key_md1
*md1
= nla_data(a
);
1435 for (i
= 0; i
< NSH_MD1_CONTEXT_SIZE
; i
++)
1436 SW_FLOW_KEY_PUT(match
, nsh
.context
[i
],
1437 md1
->context
[i
], is_mask
);
1440 case OVS_NSH_KEY_ATTR_MD2
:
1441 if (!is_push_nsh
) /* Not supported MD type 2 yet */
1446 if (mdlen
> NSH_CTX_HDRS_MAX_LEN
|| mdlen
<= 0) {
1449 "Invalid MD length %d for MD type %d",
1457 OVS_NLERR(log
, "Unknown nsh attribute %d",
1464 OVS_NLERR(log
, "nsh attribute has %d unknown bytes.", rem
);
1468 if (has_md1
&& has_md2
) {
1471 "invalid nsh attribute: md1 and md2 are exclusive."
1477 if ((has_md1
&& mdtype
!= NSH_M_TYPE1
) ||
1478 (has_md2
&& mdtype
!= NSH_M_TYPE2
)) {
1479 OVS_NLERR(1, "nsh attribute has unmatched MD type %d.",
1485 (!has_base
|| (!has_md1
&& !has_md2
))) {
1488 "push_nsh: missing base or metadata attributes"
1497 static int ovs_key_from_nlattrs(struct net
*net
, struct sw_flow_match
*match
,
1498 u64 attrs
, const struct nlattr
**a
,
1499 bool is_mask
, bool log
)
1503 err
= metadata_from_nlattrs(net
, match
, &attrs
, a
, is_mask
, log
);
1507 if (attrs
& (1 << OVS_KEY_ATTR_ETHERNET
)) {
1508 const struct ovs_key_ethernet
*eth_key
;
1510 eth_key
= nla_data(a
[OVS_KEY_ATTR_ETHERNET
]);
1511 SW_FLOW_KEY_MEMCPY(match
, eth
.src
,
1512 eth_key
->eth_src
, ETH_ALEN
, is_mask
);
1513 SW_FLOW_KEY_MEMCPY(match
, eth
.dst
,
1514 eth_key
->eth_dst
, ETH_ALEN
, is_mask
);
1515 attrs
&= ~(1 << OVS_KEY_ATTR_ETHERNET
);
1517 if (attrs
& (1 << OVS_KEY_ATTR_VLAN
)) {
1518 /* VLAN attribute is always parsed before getting here since it
1519 * may occur multiple times.
1521 OVS_NLERR(log
, "VLAN attribute unexpected.");
1525 if (attrs
& (1 << OVS_KEY_ATTR_ETHERTYPE
)) {
1526 err
= parse_eth_type_from_nlattrs(match
, &attrs
, a
, is_mask
,
1530 } else if (!is_mask
) {
1531 SW_FLOW_KEY_PUT(match
, eth
.type
, htons(ETH_P_802_2
), is_mask
);
1533 } else if (!match
->key
->eth
.type
) {
1534 OVS_NLERR(log
, "Either Ethernet header or EtherType is required.");
1538 if (attrs
& (1 << OVS_KEY_ATTR_IPV4
)) {
1539 const struct ovs_key_ipv4
*ipv4_key
;
1541 ipv4_key
= nla_data(a
[OVS_KEY_ATTR_IPV4
]);
1542 if (!is_mask
&& ipv4_key
->ipv4_frag
> OVS_FRAG_TYPE_MAX
) {
1543 OVS_NLERR(log
, "IPv4 frag type %d is out of range max %d",
1544 ipv4_key
->ipv4_frag
, OVS_FRAG_TYPE_MAX
);
1547 SW_FLOW_KEY_PUT(match
, ip
.proto
,
1548 ipv4_key
->ipv4_proto
, is_mask
);
1549 SW_FLOW_KEY_PUT(match
, ip
.tos
,
1550 ipv4_key
->ipv4_tos
, is_mask
);
1551 SW_FLOW_KEY_PUT(match
, ip
.ttl
,
1552 ipv4_key
->ipv4_ttl
, is_mask
);
1553 SW_FLOW_KEY_PUT(match
, ip
.frag
,
1554 ipv4_key
->ipv4_frag
, is_mask
);
1555 SW_FLOW_KEY_PUT(match
, ipv4
.addr
.src
,
1556 ipv4_key
->ipv4_src
, is_mask
);
1557 SW_FLOW_KEY_PUT(match
, ipv4
.addr
.dst
,
1558 ipv4_key
->ipv4_dst
, is_mask
);
1559 attrs
&= ~(1 << OVS_KEY_ATTR_IPV4
);
1562 if (attrs
& (1 << OVS_KEY_ATTR_IPV6
)) {
1563 const struct ovs_key_ipv6
*ipv6_key
;
1565 ipv6_key
= nla_data(a
[OVS_KEY_ATTR_IPV6
]);
1566 if (!is_mask
&& ipv6_key
->ipv6_frag
> OVS_FRAG_TYPE_MAX
) {
1567 OVS_NLERR(log
, "IPv6 frag type %d is out of range max %d",
1568 ipv6_key
->ipv6_frag
, OVS_FRAG_TYPE_MAX
);
1572 if (!is_mask
&& ipv6_key
->ipv6_label
& htonl(0xFFF00000)) {
1573 OVS_NLERR(log
, "IPv6 flow label %x is out of range (max=%x)",
1574 ntohl(ipv6_key
->ipv6_label
), (1 << 20) - 1);
1578 SW_FLOW_KEY_PUT(match
, ipv6
.label
,
1579 ipv6_key
->ipv6_label
, is_mask
);
1580 SW_FLOW_KEY_PUT(match
, ip
.proto
,
1581 ipv6_key
->ipv6_proto
, is_mask
);
1582 SW_FLOW_KEY_PUT(match
, ip
.tos
,
1583 ipv6_key
->ipv6_tclass
, is_mask
);
1584 SW_FLOW_KEY_PUT(match
, ip
.ttl
,
1585 ipv6_key
->ipv6_hlimit
, is_mask
);
1586 SW_FLOW_KEY_PUT(match
, ip
.frag
,
1587 ipv6_key
->ipv6_frag
, is_mask
);
1588 SW_FLOW_KEY_MEMCPY(match
, ipv6
.addr
.src
,
1590 sizeof(match
->key
->ipv6
.addr
.src
),
1592 SW_FLOW_KEY_MEMCPY(match
, ipv6
.addr
.dst
,
1594 sizeof(match
->key
->ipv6
.addr
.dst
),
1597 attrs
&= ~(1 << OVS_KEY_ATTR_IPV6
);
1600 if (attrs
& (1 << OVS_KEY_ATTR_ARP
)) {
1601 const struct ovs_key_arp
*arp_key
;
1603 arp_key
= nla_data(a
[OVS_KEY_ATTR_ARP
]);
1604 if (!is_mask
&& (arp_key
->arp_op
& htons(0xff00))) {
1605 OVS_NLERR(log
, "Unknown ARP opcode (opcode=%d).",
1610 SW_FLOW_KEY_PUT(match
, ipv4
.addr
.src
,
1611 arp_key
->arp_sip
, is_mask
);
1612 SW_FLOW_KEY_PUT(match
, ipv4
.addr
.dst
,
1613 arp_key
->arp_tip
, is_mask
);
1614 SW_FLOW_KEY_PUT(match
, ip
.proto
,
1615 ntohs(arp_key
->arp_op
), is_mask
);
1616 SW_FLOW_KEY_MEMCPY(match
, ipv4
.arp
.sha
,
1617 arp_key
->arp_sha
, ETH_ALEN
, is_mask
);
1618 SW_FLOW_KEY_MEMCPY(match
, ipv4
.arp
.tha
,
1619 arp_key
->arp_tha
, ETH_ALEN
, is_mask
);
1621 attrs
&= ~(1 << OVS_KEY_ATTR_ARP
);
1624 if (attrs
& (1 << OVS_KEY_ATTR_NSH
)) {
1625 if (nsh_key_put_from_nlattr(a
[OVS_KEY_ATTR_NSH
], match
,
1626 is_mask
, false, log
) < 0)
1628 attrs
&= ~(1 << OVS_KEY_ATTR_NSH
);
1631 if (attrs
& (1 << OVS_KEY_ATTR_MPLS
)) {
1632 const struct ovs_key_mpls
*mpls_key
;
1634 u32 label_count
, label_count_mask
, i
;
1636 mpls_key
= nla_data(a
[OVS_KEY_ATTR_MPLS
]);
1637 hdr_len
= nla_len(a
[OVS_KEY_ATTR_MPLS
]);
1638 label_count
= hdr_len
/ sizeof(struct ovs_key_mpls
);
1640 if (label_count
== 0 || label_count
> MPLS_LABEL_DEPTH
||
1641 hdr_len
% sizeof(struct ovs_key_mpls
))
1644 label_count_mask
= GENMASK(label_count
- 1, 0);
1646 for (i
= 0 ; i
< label_count
; i
++)
1647 SW_FLOW_KEY_PUT(match
, mpls
.lse
[i
],
1648 mpls_key
[i
].mpls_lse
, is_mask
);
1650 SW_FLOW_KEY_PUT(match
, mpls
.num_labels_mask
,
1651 label_count_mask
, is_mask
);
1653 attrs
&= ~(1 << OVS_KEY_ATTR_MPLS
);
1656 if (attrs
& (1 << OVS_KEY_ATTR_TCP
)) {
1657 const struct ovs_key_tcp
*tcp_key
;
1659 tcp_key
= nla_data(a
[OVS_KEY_ATTR_TCP
]);
1660 SW_FLOW_KEY_PUT(match
, tp
.src
, tcp_key
->tcp_src
, is_mask
);
1661 SW_FLOW_KEY_PUT(match
, tp
.dst
, tcp_key
->tcp_dst
, is_mask
);
1662 attrs
&= ~(1 << OVS_KEY_ATTR_TCP
);
1665 if (attrs
& (1 << OVS_KEY_ATTR_TCP_FLAGS
)) {
1666 SW_FLOW_KEY_PUT(match
, tp
.flags
,
1667 nla_get_be16(a
[OVS_KEY_ATTR_TCP_FLAGS
]),
1669 attrs
&= ~(1 << OVS_KEY_ATTR_TCP_FLAGS
);
1672 if (attrs
& (1 << OVS_KEY_ATTR_UDP
)) {
1673 const struct ovs_key_udp
*udp_key
;
1675 udp_key
= nla_data(a
[OVS_KEY_ATTR_UDP
]);
1676 SW_FLOW_KEY_PUT(match
, tp
.src
, udp_key
->udp_src
, is_mask
);
1677 SW_FLOW_KEY_PUT(match
, tp
.dst
, udp_key
->udp_dst
, is_mask
);
1678 attrs
&= ~(1 << OVS_KEY_ATTR_UDP
);
1681 if (attrs
& (1 << OVS_KEY_ATTR_SCTP
)) {
1682 const struct ovs_key_sctp
*sctp_key
;
1684 sctp_key
= nla_data(a
[OVS_KEY_ATTR_SCTP
]);
1685 SW_FLOW_KEY_PUT(match
, tp
.src
, sctp_key
->sctp_src
, is_mask
);
1686 SW_FLOW_KEY_PUT(match
, tp
.dst
, sctp_key
->sctp_dst
, is_mask
);
1687 attrs
&= ~(1 << OVS_KEY_ATTR_SCTP
);
1690 if (attrs
& (1 << OVS_KEY_ATTR_ICMP
)) {
1691 const struct ovs_key_icmp
*icmp_key
;
1693 icmp_key
= nla_data(a
[OVS_KEY_ATTR_ICMP
]);
1694 SW_FLOW_KEY_PUT(match
, tp
.src
,
1695 htons(icmp_key
->icmp_type
), is_mask
);
1696 SW_FLOW_KEY_PUT(match
, tp
.dst
,
1697 htons(icmp_key
->icmp_code
), is_mask
);
1698 attrs
&= ~(1 << OVS_KEY_ATTR_ICMP
);
1701 if (attrs
& (1 << OVS_KEY_ATTR_ICMPV6
)) {
1702 const struct ovs_key_icmpv6
*icmpv6_key
;
1704 icmpv6_key
= nla_data(a
[OVS_KEY_ATTR_ICMPV6
]);
1705 SW_FLOW_KEY_PUT(match
, tp
.src
,
1706 htons(icmpv6_key
->icmpv6_type
), is_mask
);
1707 SW_FLOW_KEY_PUT(match
, tp
.dst
,
1708 htons(icmpv6_key
->icmpv6_code
), is_mask
);
1709 attrs
&= ~(1 << OVS_KEY_ATTR_ICMPV6
);
1712 if (attrs
& (1 << OVS_KEY_ATTR_ND
)) {
1713 const struct ovs_key_nd
*nd_key
;
1715 nd_key
= nla_data(a
[OVS_KEY_ATTR_ND
]);
1716 SW_FLOW_KEY_MEMCPY(match
, ipv6
.nd
.target
,
1718 sizeof(match
->key
->ipv6
.nd
.target
),
1720 SW_FLOW_KEY_MEMCPY(match
, ipv6
.nd
.sll
,
1721 nd_key
->nd_sll
, ETH_ALEN
, is_mask
);
1722 SW_FLOW_KEY_MEMCPY(match
, ipv6
.nd
.tll
,
1723 nd_key
->nd_tll
, ETH_ALEN
, is_mask
);
1724 attrs
&= ~(1 << OVS_KEY_ATTR_ND
);
1728 OVS_NLERR(log
, "Unknown key attributes %llx",
1729 (unsigned long long)attrs
);
1736 static void nlattr_set(struct nlattr
*attr
, u8 val
,
1737 const struct ovs_len_tbl
*tbl
)
1742 /* The nlattr stream should already have been validated */
1743 nla_for_each_nested(nla
, attr
, rem
) {
1744 if (tbl
[nla_type(nla
)].len
== OVS_ATTR_NESTED
)
1745 nlattr_set(nla
, val
, tbl
[nla_type(nla
)].next
? : tbl
);
1747 memset(nla_data(nla
), val
, nla_len(nla
));
1749 if (nla_type(nla
) == OVS_KEY_ATTR_CT_STATE
)
1750 *(u32
*)nla_data(nla
) &= CT_SUPPORTED_MASK
;
1754 static void mask_set_nlattr(struct nlattr
*attr
, u8 val
)
1756 nlattr_set(attr
, val
, ovs_key_lens
);
1760 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1761 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1762 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1763 * does not include any don't care bit.
1764 * @net: Used to determine per-namespace field support.
1765 * @match: receives the extracted flow match information.
1766 * @nla_key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1767 * sequence. The fields should of the packet that triggered the creation
1769 * @nla_mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_*
1770 * Netlink attribute specifies the mask field of the wildcarded flow.
1771 * @log: Boolean to allow kernel error logging. Normally true, but when
1772 * probing for feature compatibility this should be passed in as false to
1773 * suppress unnecessary error logging.
1775 int ovs_nla_get_match(struct net
*net
, struct sw_flow_match
*match
,
1776 const struct nlattr
*nla_key
,
1777 const struct nlattr
*nla_mask
,
1780 const struct nlattr
*a
[OVS_KEY_ATTR_MAX
+ 1];
1781 struct nlattr
*newmask
= NULL
;
1786 err
= parse_flow_nlattrs(nla_key
, a
, &key_attrs
, log
);
1790 err
= parse_vlan_from_nlattrs(match
, &key_attrs
, a
, false, log
);
1794 err
= ovs_key_from_nlattrs(net
, match
, key_attrs
, a
, false, log
);
1800 /* Create an exact match mask. We need to set to 0xff
1801 * all the 'match->mask' fields that have been touched
1802 * in 'match->key'. We cannot simply memset
1803 * 'match->mask', because padding bytes and fields not
1804 * specified in 'match->key' should be left to 0.
1805 * Instead, we use a stream of netlink attributes,
1806 * copied from 'key' and set to 0xff.
1807 * ovs_key_from_nlattrs() will take care of filling
1808 * 'match->mask' appropriately.
1810 newmask
= kmemdup(nla_key
,
1811 nla_total_size(nla_len(nla_key
)),
1816 mask_set_nlattr(newmask
, 0xff);
1818 /* The userspace does not send tunnel attributes that
1819 * are 0, but we should not wildcard them nonetheless.
1821 if (match
->key
->tun_proto
)
1822 SW_FLOW_KEY_MEMSET_FIELD(match
, tun_key
,
1828 err
= parse_flow_mask_nlattrs(nla_mask
, a
, &mask_attrs
, log
);
1832 /* Always match on tci. */
1833 SW_FLOW_KEY_PUT(match
, eth
.vlan
.tci
, htons(0xffff), true);
1834 SW_FLOW_KEY_PUT(match
, eth
.cvlan
.tci
, htons(0xffff), true);
1836 err
= parse_vlan_from_nlattrs(match
, &mask_attrs
, a
, true, log
);
1840 err
= ovs_key_from_nlattrs(net
, match
, mask_attrs
, a
, true,
1846 if (!match_validate(match
, key_attrs
, mask_attrs
, log
))
1854 static size_t get_ufid_len(const struct nlattr
*attr
, bool log
)
1861 len
= nla_len(attr
);
1862 if (len
< 1 || len
> MAX_UFID_LENGTH
) {
1863 OVS_NLERR(log
, "ufid size %u bytes exceeds the range (1, %d)",
1864 nla_len(attr
), MAX_UFID_LENGTH
);
1871 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1872 * or false otherwise.
1874 bool ovs_nla_get_ufid(struct sw_flow_id
*sfid
, const struct nlattr
*attr
,
1877 sfid
->ufid_len
= get_ufid_len(attr
, log
);
1879 memcpy(sfid
->ufid
, nla_data(attr
), sfid
->ufid_len
);
1881 return sfid
->ufid_len
;
1884 int ovs_nla_get_identifier(struct sw_flow_id
*sfid
, const struct nlattr
*ufid
,
1885 const struct sw_flow_key
*key
, bool log
)
1887 struct sw_flow_key
*new_key
;
1889 if (ovs_nla_get_ufid(sfid
, ufid
, log
))
1892 /* If UFID was not provided, use unmasked key. */
1893 new_key
= kmalloc(sizeof(*new_key
), GFP_KERNEL
);
1896 memcpy(new_key
, key
, sizeof(*key
));
1897 sfid
->unmasked_key
= new_key
;
1902 u32
ovs_nla_get_ufid_flags(const struct nlattr
*attr
)
1904 return attr
? nla_get_u32(attr
) : 0;
1908 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1909 * @net: Network namespace.
1910 * @key: Receives extracted in_port, priority, tun_key, skb_mark and conntrack
1912 * @a: Array of netlink attributes holding parsed %OVS_KEY_ATTR_* Netlink
1914 * @attrs: Bit mask for the netlink attributes included in @a.
1915 * @log: Boolean to allow kernel error logging. Normally true, but when
1916 * probing for feature compatibility this should be passed in as false to
1917 * suppress unnecessary error logging.
1919 * This parses a series of Netlink attributes that form a flow key, which must
1920 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1921 * get the metadata, that is, the parts of the flow key that cannot be
1922 * extracted from the packet itself.
1924 * This must be called before the packet key fields are filled in 'key'.
1927 int ovs_nla_get_flow_metadata(struct net
*net
,
1928 const struct nlattr
*a
[OVS_KEY_ATTR_MAX
+ 1],
1929 u64 attrs
, struct sw_flow_key
*key
, bool log
)
1931 struct sw_flow_match match
;
1933 memset(&match
, 0, sizeof(match
));
1938 key
->ct_orig_proto
= 0;
1939 memset(&key
->ct
, 0, sizeof(key
->ct
));
1940 memset(&key
->ipv4
.ct_orig
, 0, sizeof(key
->ipv4
.ct_orig
));
1941 memset(&key
->ipv6
.ct_orig
, 0, sizeof(key
->ipv6
.ct_orig
));
1943 key
->phy
.in_port
= DP_MAX_PORTS
;
1945 return metadata_from_nlattrs(net
, &match
, &attrs
, a
, false, log
);
1948 static int ovs_nla_put_vlan(struct sk_buff
*skb
, const struct vlan_head
*vh
,
1951 __be16 eth_type
= !is_mask
? vh
->tpid
: htons(0xffff);
1953 if (nla_put_be16(skb
, OVS_KEY_ATTR_ETHERTYPE
, eth_type
) ||
1954 nla_put_be16(skb
, OVS_KEY_ATTR_VLAN
, vh
->tci
))
1959 static int nsh_key_to_nlattr(const struct ovs_key_nsh
*nsh
, bool is_mask
,
1960 struct sk_buff
*skb
)
1962 struct nlattr
*start
;
1964 start
= nla_nest_start_noflag(skb
, OVS_KEY_ATTR_NSH
);
1968 if (nla_put(skb
, OVS_NSH_KEY_ATTR_BASE
, sizeof(nsh
->base
), &nsh
->base
))
1969 goto nla_put_failure
;
1971 if (is_mask
|| nsh
->base
.mdtype
== NSH_M_TYPE1
) {
1972 if (nla_put(skb
, OVS_NSH_KEY_ATTR_MD1
,
1973 sizeof(nsh
->context
), nsh
->context
))
1974 goto nla_put_failure
;
1977 /* Don't support MD type 2 yet */
1979 nla_nest_end(skb
, start
);
1987 static int __ovs_nla_put_key(const struct sw_flow_key
*swkey
,
1988 const struct sw_flow_key
*output
, bool is_mask
,
1989 struct sk_buff
*skb
)
1991 struct ovs_key_ethernet
*eth_key
;
1993 struct nlattr
*encap
= NULL
;
1994 struct nlattr
*in_encap
= NULL
;
1996 if (nla_put_u32(skb
, OVS_KEY_ATTR_RECIRC_ID
, output
->recirc_id
))
1997 goto nla_put_failure
;
1999 if (nla_put_u32(skb
, OVS_KEY_ATTR_DP_HASH
, output
->ovs_flow_hash
))
2000 goto nla_put_failure
;
2002 if (nla_put_u32(skb
, OVS_KEY_ATTR_PRIORITY
, output
->phy
.priority
))
2003 goto nla_put_failure
;
2005 if ((swkey
->tun_proto
|| is_mask
)) {
2006 const void *opts
= NULL
;
2008 if (output
->tun_key
.tun_flags
& TUNNEL_OPTIONS_PRESENT
)
2009 opts
= TUN_METADATA_OPTS(output
, swkey
->tun_opts_len
);
2011 if (ip_tun_to_nlattr(skb
, &output
->tun_key
, opts
,
2012 swkey
->tun_opts_len
, swkey
->tun_proto
, 0))
2013 goto nla_put_failure
;
2016 if (swkey
->phy
.in_port
== DP_MAX_PORTS
) {
2017 if (is_mask
&& (output
->phy
.in_port
== 0xffff))
2018 if (nla_put_u32(skb
, OVS_KEY_ATTR_IN_PORT
, 0xffffffff))
2019 goto nla_put_failure
;
2022 upper_u16
= !is_mask
? 0 : 0xffff;
2024 if (nla_put_u32(skb
, OVS_KEY_ATTR_IN_PORT
,
2025 (upper_u16
<< 16) | output
->phy
.in_port
))
2026 goto nla_put_failure
;
2029 if (nla_put_u32(skb
, OVS_KEY_ATTR_SKB_MARK
, output
->phy
.skb_mark
))
2030 goto nla_put_failure
;
2032 if (ovs_ct_put_key(swkey
, output
, skb
))
2033 goto nla_put_failure
;
2035 if (ovs_key_mac_proto(swkey
) == MAC_PROTO_ETHERNET
) {
2036 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ETHERNET
, sizeof(*eth_key
));
2038 goto nla_put_failure
;
2040 eth_key
= nla_data(nla
);
2041 ether_addr_copy(eth_key
->eth_src
, output
->eth
.src
);
2042 ether_addr_copy(eth_key
->eth_dst
, output
->eth
.dst
);
2044 if (swkey
->eth
.vlan
.tci
|| eth_type_vlan(swkey
->eth
.type
)) {
2045 if (ovs_nla_put_vlan(skb
, &output
->eth
.vlan
, is_mask
))
2046 goto nla_put_failure
;
2047 encap
= nla_nest_start_noflag(skb
, OVS_KEY_ATTR_ENCAP
);
2048 if (!swkey
->eth
.vlan
.tci
)
2051 if (swkey
->eth
.cvlan
.tci
|| eth_type_vlan(swkey
->eth
.type
)) {
2052 if (ovs_nla_put_vlan(skb
, &output
->eth
.cvlan
, is_mask
))
2053 goto nla_put_failure
;
2054 in_encap
= nla_nest_start_noflag(skb
,
2055 OVS_KEY_ATTR_ENCAP
);
2056 if (!swkey
->eth
.cvlan
.tci
)
2061 if (swkey
->eth
.type
== htons(ETH_P_802_2
)) {
2063 * Ethertype 802.2 is represented in the netlink with omitted
2064 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
2065 * 0xffff in the mask attribute. Ethertype can also
2068 if (is_mask
&& output
->eth
.type
)
2069 if (nla_put_be16(skb
, OVS_KEY_ATTR_ETHERTYPE
,
2071 goto nla_put_failure
;
2076 if (nla_put_be16(skb
, OVS_KEY_ATTR_ETHERTYPE
, output
->eth
.type
))
2077 goto nla_put_failure
;
2079 if (eth_type_vlan(swkey
->eth
.type
)) {
2080 /* There are 3 VLAN tags, we don't know anything about the rest
2081 * of the packet, so truncate here.
2083 WARN_ON_ONCE(!(encap
&& in_encap
));
2087 if (swkey
->eth
.type
== htons(ETH_P_IP
)) {
2088 struct ovs_key_ipv4
*ipv4_key
;
2090 nla
= nla_reserve(skb
, OVS_KEY_ATTR_IPV4
, sizeof(*ipv4_key
));
2092 goto nla_put_failure
;
2093 ipv4_key
= nla_data(nla
);
2094 ipv4_key
->ipv4_src
= output
->ipv4
.addr
.src
;
2095 ipv4_key
->ipv4_dst
= output
->ipv4
.addr
.dst
;
2096 ipv4_key
->ipv4_proto
= output
->ip
.proto
;
2097 ipv4_key
->ipv4_tos
= output
->ip
.tos
;
2098 ipv4_key
->ipv4_ttl
= output
->ip
.ttl
;
2099 ipv4_key
->ipv4_frag
= output
->ip
.frag
;
2100 } else if (swkey
->eth
.type
== htons(ETH_P_IPV6
)) {
2101 struct ovs_key_ipv6
*ipv6_key
;
2103 nla
= nla_reserve(skb
, OVS_KEY_ATTR_IPV6
, sizeof(*ipv6_key
));
2105 goto nla_put_failure
;
2106 ipv6_key
= nla_data(nla
);
2107 memcpy(ipv6_key
->ipv6_src
, &output
->ipv6
.addr
.src
,
2108 sizeof(ipv6_key
->ipv6_src
));
2109 memcpy(ipv6_key
->ipv6_dst
, &output
->ipv6
.addr
.dst
,
2110 sizeof(ipv6_key
->ipv6_dst
));
2111 ipv6_key
->ipv6_label
= output
->ipv6
.label
;
2112 ipv6_key
->ipv6_proto
= output
->ip
.proto
;
2113 ipv6_key
->ipv6_tclass
= output
->ip
.tos
;
2114 ipv6_key
->ipv6_hlimit
= output
->ip
.ttl
;
2115 ipv6_key
->ipv6_frag
= output
->ip
.frag
;
2116 } else if (swkey
->eth
.type
== htons(ETH_P_NSH
)) {
2117 if (nsh_key_to_nlattr(&output
->nsh
, is_mask
, skb
))
2118 goto nla_put_failure
;
2119 } else if (swkey
->eth
.type
== htons(ETH_P_ARP
) ||
2120 swkey
->eth
.type
== htons(ETH_P_RARP
)) {
2121 struct ovs_key_arp
*arp_key
;
2123 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ARP
, sizeof(*arp_key
));
2125 goto nla_put_failure
;
2126 arp_key
= nla_data(nla
);
2127 memset(arp_key
, 0, sizeof(struct ovs_key_arp
));
2128 arp_key
->arp_sip
= output
->ipv4
.addr
.src
;
2129 arp_key
->arp_tip
= output
->ipv4
.addr
.dst
;
2130 arp_key
->arp_op
= htons(output
->ip
.proto
);
2131 ether_addr_copy(arp_key
->arp_sha
, output
->ipv4
.arp
.sha
);
2132 ether_addr_copy(arp_key
->arp_tha
, output
->ipv4
.arp
.tha
);
2133 } else if (eth_p_mpls(swkey
->eth
.type
)) {
2135 struct ovs_key_mpls
*mpls_key
;
2137 num_labels
= hweight_long(output
->mpls
.num_labels_mask
);
2138 nla
= nla_reserve(skb
, OVS_KEY_ATTR_MPLS
,
2139 num_labels
* sizeof(*mpls_key
));
2141 goto nla_put_failure
;
2143 mpls_key
= nla_data(nla
);
2144 for (i
= 0; i
< num_labels
; i
++)
2145 mpls_key
[i
].mpls_lse
= output
->mpls
.lse
[i
];
2148 if ((swkey
->eth
.type
== htons(ETH_P_IP
) ||
2149 swkey
->eth
.type
== htons(ETH_P_IPV6
)) &&
2150 swkey
->ip
.frag
!= OVS_FRAG_TYPE_LATER
) {
2152 if (swkey
->ip
.proto
== IPPROTO_TCP
) {
2153 struct ovs_key_tcp
*tcp_key
;
2155 nla
= nla_reserve(skb
, OVS_KEY_ATTR_TCP
, sizeof(*tcp_key
));
2157 goto nla_put_failure
;
2158 tcp_key
= nla_data(nla
);
2159 tcp_key
->tcp_src
= output
->tp
.src
;
2160 tcp_key
->tcp_dst
= output
->tp
.dst
;
2161 if (nla_put_be16(skb
, OVS_KEY_ATTR_TCP_FLAGS
,
2163 goto nla_put_failure
;
2164 } else if (swkey
->ip
.proto
== IPPROTO_UDP
) {
2165 struct ovs_key_udp
*udp_key
;
2167 nla
= nla_reserve(skb
, OVS_KEY_ATTR_UDP
, sizeof(*udp_key
));
2169 goto nla_put_failure
;
2170 udp_key
= nla_data(nla
);
2171 udp_key
->udp_src
= output
->tp
.src
;
2172 udp_key
->udp_dst
= output
->tp
.dst
;
2173 } else if (swkey
->ip
.proto
== IPPROTO_SCTP
) {
2174 struct ovs_key_sctp
*sctp_key
;
2176 nla
= nla_reserve(skb
, OVS_KEY_ATTR_SCTP
, sizeof(*sctp_key
));
2178 goto nla_put_failure
;
2179 sctp_key
= nla_data(nla
);
2180 sctp_key
->sctp_src
= output
->tp
.src
;
2181 sctp_key
->sctp_dst
= output
->tp
.dst
;
2182 } else if (swkey
->eth
.type
== htons(ETH_P_IP
) &&
2183 swkey
->ip
.proto
== IPPROTO_ICMP
) {
2184 struct ovs_key_icmp
*icmp_key
;
2186 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ICMP
, sizeof(*icmp_key
));
2188 goto nla_put_failure
;
2189 icmp_key
= nla_data(nla
);
2190 icmp_key
->icmp_type
= ntohs(output
->tp
.src
);
2191 icmp_key
->icmp_code
= ntohs(output
->tp
.dst
);
2192 } else if (swkey
->eth
.type
== htons(ETH_P_IPV6
) &&
2193 swkey
->ip
.proto
== IPPROTO_ICMPV6
) {
2194 struct ovs_key_icmpv6
*icmpv6_key
;
2196 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ICMPV6
,
2197 sizeof(*icmpv6_key
));
2199 goto nla_put_failure
;
2200 icmpv6_key
= nla_data(nla
);
2201 icmpv6_key
->icmpv6_type
= ntohs(output
->tp
.src
);
2202 icmpv6_key
->icmpv6_code
= ntohs(output
->tp
.dst
);
2204 if (icmpv6_key
->icmpv6_type
== NDISC_NEIGHBOUR_SOLICITATION
||
2205 icmpv6_key
->icmpv6_type
== NDISC_NEIGHBOUR_ADVERTISEMENT
) {
2206 struct ovs_key_nd
*nd_key
;
2208 nla
= nla_reserve(skb
, OVS_KEY_ATTR_ND
, sizeof(*nd_key
));
2210 goto nla_put_failure
;
2211 nd_key
= nla_data(nla
);
2212 memcpy(nd_key
->nd_target
, &output
->ipv6
.nd
.target
,
2213 sizeof(nd_key
->nd_target
));
2214 ether_addr_copy(nd_key
->nd_sll
, output
->ipv6
.nd
.sll
);
2215 ether_addr_copy(nd_key
->nd_tll
, output
->ipv6
.nd
.tll
);
2222 nla_nest_end(skb
, in_encap
);
2224 nla_nest_end(skb
, encap
);
2232 int ovs_nla_put_key(const struct sw_flow_key
*swkey
,
2233 const struct sw_flow_key
*output
, int attr
, bool is_mask
,
2234 struct sk_buff
*skb
)
2239 nla
= nla_nest_start_noflag(skb
, attr
);
2242 err
= __ovs_nla_put_key(swkey
, output
, is_mask
, skb
);
2245 nla_nest_end(skb
, nla
);
2250 /* Called with ovs_mutex or RCU read lock. */
2251 int ovs_nla_put_identifier(const struct sw_flow
*flow
, struct sk_buff
*skb
)
2253 if (ovs_identifier_is_ufid(&flow
->id
))
2254 return nla_put(skb
, OVS_FLOW_ATTR_UFID
, flow
->id
.ufid_len
,
2257 return ovs_nla_put_key(flow
->id
.unmasked_key
, flow
->id
.unmasked_key
,
2258 OVS_FLOW_ATTR_KEY
, false, skb
);
2261 /* Called with ovs_mutex or RCU read lock. */
2262 int ovs_nla_put_masked_key(const struct sw_flow
*flow
, struct sk_buff
*skb
)
2264 return ovs_nla_put_key(&flow
->key
, &flow
->key
,
2265 OVS_FLOW_ATTR_KEY
, false, skb
);
2268 /* Called with ovs_mutex or RCU read lock. */
2269 int ovs_nla_put_mask(const struct sw_flow
*flow
, struct sk_buff
*skb
)
2271 return ovs_nla_put_key(&flow
->key
, &flow
->mask
->key
,
2272 OVS_FLOW_ATTR_MASK
, true, skb
);
2275 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
2277 static struct sw_flow_actions
*nla_alloc_flow_actions(int size
)
2279 struct sw_flow_actions
*sfa
;
2281 WARN_ON_ONCE(size
> MAX_ACTIONS_BUFSIZE
);
2283 sfa
= kmalloc(sizeof(*sfa
) + size
, GFP_KERNEL
);
2285 return ERR_PTR(-ENOMEM
);
2287 sfa
->actions_len
= 0;
2291 static void ovs_nla_free_set_action(const struct nlattr
*a
)
2293 const struct nlattr
*ovs_key
= nla_data(a
);
2294 struct ovs_tunnel_info
*ovs_tun
;
2296 switch (nla_type(ovs_key
)) {
2297 case OVS_KEY_ATTR_TUNNEL_INFO
:
2298 ovs_tun
= nla_data(ovs_key
);
2299 dst_release((struct dst_entry
*)ovs_tun
->tun_dst
);
2304 void ovs_nla_free_flow_actions(struct sw_flow_actions
*sf_acts
)
2306 const struct nlattr
*a
;
2312 nla_for_each_attr(a
, sf_acts
->actions
, sf_acts
->actions_len
, rem
) {
2313 switch (nla_type(a
)) {
2314 case OVS_ACTION_ATTR_SET
:
2315 ovs_nla_free_set_action(a
);
2317 case OVS_ACTION_ATTR_CT
:
2318 ovs_ct_free_action(a
);
2326 static void __ovs_nla_free_flow_actions(struct rcu_head
*head
)
2328 ovs_nla_free_flow_actions(container_of(head
, struct sw_flow_actions
, rcu
));
2331 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
2332 * The caller must hold rcu_read_lock for this to be sensible. */
2333 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions
*sf_acts
)
2335 call_rcu(&sf_acts
->rcu
, __ovs_nla_free_flow_actions
);
2338 static struct nlattr
*reserve_sfa_size(struct sw_flow_actions
**sfa
,
2339 int attr_len
, bool log
)
2342 struct sw_flow_actions
*acts
;
2344 size_t req_size
= NLA_ALIGN(attr_len
);
2345 int next_offset
= offsetof(struct sw_flow_actions
, actions
) +
2346 (*sfa
)->actions_len
;
2348 if (req_size
<= (ksize(*sfa
) - next_offset
))
2351 new_acts_size
= max(next_offset
+ req_size
, ksize(*sfa
) * 2);
2353 if (new_acts_size
> MAX_ACTIONS_BUFSIZE
) {
2354 if ((MAX_ACTIONS_BUFSIZE
- next_offset
) < req_size
) {
2355 OVS_NLERR(log
, "Flow action size exceeds max %u",
2356 MAX_ACTIONS_BUFSIZE
);
2357 return ERR_PTR(-EMSGSIZE
);
2359 new_acts_size
= MAX_ACTIONS_BUFSIZE
;
2362 acts
= nla_alloc_flow_actions(new_acts_size
);
2364 return (void *)acts
;
2366 memcpy(acts
->actions
, (*sfa
)->actions
, (*sfa
)->actions_len
);
2367 acts
->actions_len
= (*sfa
)->actions_len
;
2368 acts
->orig_len
= (*sfa
)->orig_len
;
2373 (*sfa
)->actions_len
+= req_size
;
2374 return (struct nlattr
*) ((unsigned char *)(*sfa
) + next_offset
);
2377 static struct nlattr
*__add_action(struct sw_flow_actions
**sfa
,
2378 int attrtype
, void *data
, int len
, bool log
)
2382 a
= reserve_sfa_size(sfa
, nla_attr_size(len
), log
);
2386 a
->nla_type
= attrtype
;
2387 a
->nla_len
= nla_attr_size(len
);
2390 memcpy(nla_data(a
), data
, len
);
2391 memset((unsigned char *) a
+ a
->nla_len
, 0, nla_padlen(len
));
2396 int ovs_nla_add_action(struct sw_flow_actions
**sfa
, int attrtype
, void *data
,
2401 a
= __add_action(sfa
, attrtype
, data
, len
, log
);
2403 return PTR_ERR_OR_ZERO(a
);
2406 static inline int add_nested_action_start(struct sw_flow_actions
**sfa
,
2407 int attrtype
, bool log
)
2409 int used
= (*sfa
)->actions_len
;
2412 err
= ovs_nla_add_action(sfa
, attrtype
, NULL
, 0, log
);
2419 static inline void add_nested_action_end(struct sw_flow_actions
*sfa
,
2422 struct nlattr
*a
= (struct nlattr
*) ((unsigned char *)sfa
->actions
+
2425 a
->nla_len
= sfa
->actions_len
- st_offset
;
2428 static int __ovs_nla_copy_actions(struct net
*net
, const struct nlattr
*attr
,
2429 const struct sw_flow_key
*key
,
2430 struct sw_flow_actions
**sfa
,
2431 __be16 eth_type
, __be16 vlan_tci
,
2432 u32 mpls_label_count
, bool log
);
2434 static int validate_and_copy_sample(struct net
*net
, const struct nlattr
*attr
,
2435 const struct sw_flow_key
*key
,
2436 struct sw_flow_actions
**sfa
,
2437 __be16 eth_type
, __be16 vlan_tci
,
2438 u32 mpls_label_count
, bool log
, bool last
)
2440 const struct nlattr
*attrs
[OVS_SAMPLE_ATTR_MAX
+ 1];
2441 const struct nlattr
*probability
, *actions
;
2442 const struct nlattr
*a
;
2443 int rem
, start
, err
;
2444 struct sample_arg arg
;
2446 memset(attrs
, 0, sizeof(attrs
));
2447 nla_for_each_nested(a
, attr
, rem
) {
2448 int type
= nla_type(a
);
2449 if (!type
|| type
> OVS_SAMPLE_ATTR_MAX
|| attrs
[type
])
2456 probability
= attrs
[OVS_SAMPLE_ATTR_PROBABILITY
];
2457 if (!probability
|| nla_len(probability
) != sizeof(u32
))
2460 actions
= attrs
[OVS_SAMPLE_ATTR_ACTIONS
];
2461 if (!actions
|| (nla_len(actions
) && nla_len(actions
) < NLA_HDRLEN
))
2464 /* validation done, copy sample action. */
2465 start
= add_nested_action_start(sfa
, OVS_ACTION_ATTR_SAMPLE
, log
);
2469 /* When both skb and flow may be changed, put the sample
2470 * into a deferred fifo. On the other hand, if only skb
2471 * may be modified, the actions can be executed in place.
2473 * Do this analysis at the flow installation time.
2474 * Set 'clone_action->exec' to true if the actions can be
2475 * executed without being deferred.
2477 * If the sample is the last action, it can always be excuted
2478 * rather than deferred.
2480 arg
.exec
= last
|| !actions_may_change_flow(actions
);
2481 arg
.probability
= nla_get_u32(probability
);
2483 err
= ovs_nla_add_action(sfa
, OVS_SAMPLE_ATTR_ARG
, &arg
, sizeof(arg
),
2488 err
= __ovs_nla_copy_actions(net
, actions
, key
, sfa
,
2489 eth_type
, vlan_tci
, mpls_label_count
, log
);
2494 add_nested_action_end(*sfa
, start
);
2499 static int validate_and_copy_dec_ttl(struct net
*net
,
2500 const struct nlattr
*attr
,
2501 const struct sw_flow_key
*key
,
2502 struct sw_flow_actions
**sfa
,
2503 __be16 eth_type
, __be16 vlan_tci
,
2504 u32 mpls_label_count
, bool log
)
2506 const struct nlattr
*attrs
[OVS_DEC_TTL_ATTR_MAX
+ 1];
2507 int start
, action_start
, err
, rem
;
2508 const struct nlattr
*a
, *actions
;
2510 memset(attrs
, 0, sizeof(attrs
));
2511 nla_for_each_nested(a
, attr
, rem
) {
2512 int type
= nla_type(a
);
2514 /* Ignore unknown attributes to be future proof. */
2515 if (type
> OVS_DEC_TTL_ATTR_MAX
)
2518 if (!type
|| attrs
[type
]) {
2519 OVS_NLERR(log
, "Duplicate or invalid key (type %d).",
2528 OVS_NLERR(log
, "Message has %d unknown bytes.", rem
);
2532 actions
= attrs
[OVS_DEC_TTL_ATTR_ACTION
];
2533 if (!actions
|| (nla_len(actions
) && nla_len(actions
) < NLA_HDRLEN
)) {
2534 OVS_NLERR(log
, "Missing valid actions attribute.");
2538 start
= add_nested_action_start(sfa
, OVS_ACTION_ATTR_DEC_TTL
, log
);
2542 action_start
= add_nested_action_start(sfa
, OVS_DEC_TTL_ATTR_ACTION
, log
);
2543 if (action_start
< 0)
2544 return action_start
;
2546 err
= __ovs_nla_copy_actions(net
, actions
, key
, sfa
, eth_type
,
2547 vlan_tci
, mpls_label_count
, log
);
2551 add_nested_action_end(*sfa
, action_start
);
2552 add_nested_action_end(*sfa
, start
);
2556 static int validate_and_copy_clone(struct net
*net
,
2557 const struct nlattr
*attr
,
2558 const struct sw_flow_key
*key
,
2559 struct sw_flow_actions
**sfa
,
2560 __be16 eth_type
, __be16 vlan_tci
,
2561 u32 mpls_label_count
, bool log
, bool last
)
2566 if (nla_len(attr
) && nla_len(attr
) < NLA_HDRLEN
)
2569 start
= add_nested_action_start(sfa
, OVS_ACTION_ATTR_CLONE
, log
);
2573 exec
= last
|| !actions_may_change_flow(attr
);
2575 err
= ovs_nla_add_action(sfa
, OVS_CLONE_ATTR_EXEC
, &exec
,
2580 err
= __ovs_nla_copy_actions(net
, attr
, key
, sfa
,
2581 eth_type
, vlan_tci
, mpls_label_count
, log
);
2585 add_nested_action_end(*sfa
, start
);
2590 void ovs_match_init(struct sw_flow_match
*match
,
2591 struct sw_flow_key
*key
,
2593 struct sw_flow_mask
*mask
)
2595 memset(match
, 0, sizeof(*match
));
2600 memset(key
, 0, sizeof(*key
));
2603 memset(&mask
->key
, 0, sizeof(mask
->key
));
2604 mask
->range
.start
= mask
->range
.end
= 0;
2608 static int validate_geneve_opts(struct sw_flow_key
*key
)
2610 struct geneve_opt
*option
;
2611 int opts_len
= key
->tun_opts_len
;
2612 bool crit_opt
= false;
2614 option
= (struct geneve_opt
*)TUN_METADATA_OPTS(key
, key
->tun_opts_len
);
2615 while (opts_len
> 0) {
2618 if (opts_len
< sizeof(*option
))
2621 len
= sizeof(*option
) + option
->length
* 4;
2625 crit_opt
|= !!(option
->type
& GENEVE_CRIT_OPT_TYPE
);
2627 option
= (struct geneve_opt
*)((u8
*)option
+ len
);
2631 key
->tun_key
.tun_flags
|= crit_opt
? TUNNEL_CRIT_OPT
: 0;
2636 static int validate_and_copy_set_tun(const struct nlattr
*attr
,
2637 struct sw_flow_actions
**sfa
, bool log
)
2639 struct sw_flow_match match
;
2640 struct sw_flow_key key
;
2641 struct metadata_dst
*tun_dst
;
2642 struct ip_tunnel_info
*tun_info
;
2643 struct ovs_tunnel_info
*ovs_tun
;
2645 int err
= 0, start
, opts_type
;
2646 __be16 dst_opt_type
;
2649 ovs_match_init(&match
, &key
, true, NULL
);
2650 opts_type
= ip_tun_from_nlattr(nla_data(attr
), &match
, false, log
);
2654 if (key
.tun_opts_len
) {
2655 switch (opts_type
) {
2656 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS
:
2657 err
= validate_geneve_opts(&key
);
2660 dst_opt_type
= TUNNEL_GENEVE_OPT
;
2662 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS
:
2663 dst_opt_type
= TUNNEL_VXLAN_OPT
;
2665 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS
:
2666 dst_opt_type
= TUNNEL_ERSPAN_OPT
;
2671 start
= add_nested_action_start(sfa
, OVS_ACTION_ATTR_SET
, log
);
2675 tun_dst
= metadata_dst_alloc(key
.tun_opts_len
, METADATA_IP_TUNNEL
,
2681 err
= dst_cache_init(&tun_dst
->u
.tun_info
.dst_cache
, GFP_KERNEL
);
2683 dst_release((struct dst_entry
*)tun_dst
);
2687 a
= __add_action(sfa
, OVS_KEY_ATTR_TUNNEL_INFO
, NULL
,
2688 sizeof(*ovs_tun
), log
);
2690 dst_release((struct dst_entry
*)tun_dst
);
2694 ovs_tun
= nla_data(a
);
2695 ovs_tun
->tun_dst
= tun_dst
;
2697 tun_info
= &tun_dst
->u
.tun_info
;
2698 tun_info
->mode
= IP_TUNNEL_INFO_TX
;
2699 if (key
.tun_proto
== AF_INET6
)
2700 tun_info
->mode
|= IP_TUNNEL_INFO_IPV6
;
2701 else if (key
.tun_proto
== AF_INET
&& key
.tun_key
.u
.ipv4
.dst
== 0)
2702 tun_info
->mode
|= IP_TUNNEL_INFO_BRIDGE
;
2703 tun_info
->key
= key
.tun_key
;
2705 /* We need to store the options in the action itself since
2706 * everything else will go away after flow setup. We can append
2707 * it to tun_info and then point there.
2709 ip_tunnel_info_opts_set(tun_info
,
2710 TUN_METADATA_OPTS(&key
, key
.tun_opts_len
),
2711 key
.tun_opts_len
, dst_opt_type
);
2712 add_nested_action_end(*sfa
, start
);
2717 static bool validate_nsh(const struct nlattr
*attr
, bool is_mask
,
2718 bool is_push_nsh
, bool log
)
2720 struct sw_flow_match match
;
2721 struct sw_flow_key key
;
2724 ovs_match_init(&match
, &key
, true, NULL
);
2725 ret
= nsh_key_put_from_nlattr(attr
, &match
, is_mask
,
2730 /* Return false if there are any non-masked bits set.
2731 * Mask follows data immediately, before any netlink padding.
2733 static bool validate_masked(u8
*data
, int len
)
2735 u8
*mask
= data
+ len
;
2738 if (*data
++ & ~*mask
++)
2744 static int validate_set(const struct nlattr
*a
,
2745 const struct sw_flow_key
*flow_key
,
2746 struct sw_flow_actions
**sfa
, bool *skip_copy
,
2747 u8 mac_proto
, __be16 eth_type
, bool masked
, bool log
)
2749 const struct nlattr
*ovs_key
= nla_data(a
);
2750 int key_type
= nla_type(ovs_key
);
2753 /* There can be only one key in a action */
2754 if (nla_total_size(nla_len(ovs_key
)) != nla_len(a
))
2757 key_len
= nla_len(ovs_key
);
2761 if (key_type
> OVS_KEY_ATTR_MAX
||
2762 !check_attr_len(key_len
, ovs_key_lens
[key_type
].len
))
2765 if (masked
&& !validate_masked(nla_data(ovs_key
), key_len
))
2769 case OVS_KEY_ATTR_PRIORITY
:
2770 case OVS_KEY_ATTR_SKB_MARK
:
2771 case OVS_KEY_ATTR_CT_MARK
:
2772 case OVS_KEY_ATTR_CT_LABELS
:
2775 case OVS_KEY_ATTR_ETHERNET
:
2776 if (mac_proto
!= MAC_PROTO_ETHERNET
)
2780 case OVS_KEY_ATTR_TUNNEL
: {
2784 return -EINVAL
; /* Masked tunnel set not supported. */
2787 err
= validate_and_copy_set_tun(a
, sfa
, log
);
2792 case OVS_KEY_ATTR_IPV4
: {
2793 const struct ovs_key_ipv4
*ipv4_key
;
2795 if (eth_type
!= htons(ETH_P_IP
))
2798 ipv4_key
= nla_data(ovs_key
);
2801 const struct ovs_key_ipv4
*mask
= ipv4_key
+ 1;
2803 /* Non-writeable fields. */
2804 if (mask
->ipv4_proto
|| mask
->ipv4_frag
)
2807 if (ipv4_key
->ipv4_proto
!= flow_key
->ip
.proto
)
2810 if (ipv4_key
->ipv4_frag
!= flow_key
->ip
.frag
)
2815 case OVS_KEY_ATTR_IPV6
: {
2816 const struct ovs_key_ipv6
*ipv6_key
;
2818 if (eth_type
!= htons(ETH_P_IPV6
))
2821 ipv6_key
= nla_data(ovs_key
);
2824 const struct ovs_key_ipv6
*mask
= ipv6_key
+ 1;
2826 /* Non-writeable fields. */
2827 if (mask
->ipv6_proto
|| mask
->ipv6_frag
)
2830 /* Invalid bits in the flow label mask? */
2831 if (ntohl(mask
->ipv6_label
) & 0xFFF00000)
2834 if (ipv6_key
->ipv6_proto
!= flow_key
->ip
.proto
)
2837 if (ipv6_key
->ipv6_frag
!= flow_key
->ip
.frag
)
2840 if (ntohl(ipv6_key
->ipv6_label
) & 0xFFF00000)
2845 case OVS_KEY_ATTR_TCP
:
2846 if ((eth_type
!= htons(ETH_P_IP
) &&
2847 eth_type
!= htons(ETH_P_IPV6
)) ||
2848 flow_key
->ip
.proto
!= IPPROTO_TCP
)
2853 case OVS_KEY_ATTR_UDP
:
2854 if ((eth_type
!= htons(ETH_P_IP
) &&
2855 eth_type
!= htons(ETH_P_IPV6
)) ||
2856 flow_key
->ip
.proto
!= IPPROTO_UDP
)
2861 case OVS_KEY_ATTR_MPLS
:
2862 if (!eth_p_mpls(eth_type
))
2866 case OVS_KEY_ATTR_SCTP
:
2867 if ((eth_type
!= htons(ETH_P_IP
) &&
2868 eth_type
!= htons(ETH_P_IPV6
)) ||
2869 flow_key
->ip
.proto
!= IPPROTO_SCTP
)
2874 case OVS_KEY_ATTR_NSH
:
2875 if (eth_type
!= htons(ETH_P_NSH
))
2877 if (!validate_nsh(nla_data(a
), masked
, false, log
))
2885 /* Convert non-masked non-tunnel set actions to masked set actions. */
2886 if (!masked
&& key_type
!= OVS_KEY_ATTR_TUNNEL
) {
2887 int start
, len
= key_len
* 2;
2892 start
= add_nested_action_start(sfa
,
2893 OVS_ACTION_ATTR_SET_TO_MASKED
,
2898 at
= __add_action(sfa
, key_type
, NULL
, len
, log
);
2902 memcpy(nla_data(at
), nla_data(ovs_key
), key_len
); /* Key. */
2903 memset(nla_data(at
) + key_len
, 0xff, key_len
); /* Mask. */
2904 /* Clear non-writeable bits from otherwise writeable fields. */
2905 if (key_type
== OVS_KEY_ATTR_IPV6
) {
2906 struct ovs_key_ipv6
*mask
= nla_data(at
) + key_len
;
2908 mask
->ipv6_label
&= htonl(0x000FFFFF);
2910 add_nested_action_end(*sfa
, start
);
2916 static int validate_userspace(const struct nlattr
*attr
)
2918 static const struct nla_policy userspace_policy
[OVS_USERSPACE_ATTR_MAX
+ 1] = {
2919 [OVS_USERSPACE_ATTR_PID
] = {.type
= NLA_U32
},
2920 [OVS_USERSPACE_ATTR_USERDATA
] = {.type
= NLA_UNSPEC
},
2921 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT
] = {.type
= NLA_U32
},
2923 struct nlattr
*a
[OVS_USERSPACE_ATTR_MAX
+ 1];
2926 error
= nla_parse_nested_deprecated(a
, OVS_USERSPACE_ATTR_MAX
, attr
,
2927 userspace_policy
, NULL
);
2931 if (!a
[OVS_USERSPACE_ATTR_PID
] ||
2932 !nla_get_u32(a
[OVS_USERSPACE_ATTR_PID
]))
2938 static const struct nla_policy cpl_policy
[OVS_CHECK_PKT_LEN_ATTR_MAX
+ 1] = {
2939 [OVS_CHECK_PKT_LEN_ATTR_PKT_LEN
] = {.type
= NLA_U16
},
2940 [OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER
] = {.type
= NLA_NESTED
},
2941 [OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL
] = {.type
= NLA_NESTED
},
2944 static int validate_and_copy_check_pkt_len(struct net
*net
,
2945 const struct nlattr
*attr
,
2946 const struct sw_flow_key
*key
,
2947 struct sw_flow_actions
**sfa
,
2948 __be16 eth_type
, __be16 vlan_tci
,
2949 u32 mpls_label_count
,
2950 bool log
, bool last
)
2952 const struct nlattr
*acts_if_greater
, *acts_if_lesser_eq
;
2953 struct nlattr
*a
[OVS_CHECK_PKT_LEN_ATTR_MAX
+ 1];
2954 struct check_pkt_len_arg arg
;
2955 int nested_acts_start
;
2958 err
= nla_parse_deprecated_strict(a
, OVS_CHECK_PKT_LEN_ATTR_MAX
,
2959 nla_data(attr
), nla_len(attr
),
2964 if (!a
[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN
] ||
2965 !nla_get_u16(a
[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN
]))
2968 acts_if_lesser_eq
= a
[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL
];
2969 acts_if_greater
= a
[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER
];
2971 /* Both the nested action should be present. */
2972 if (!acts_if_greater
|| !acts_if_lesser_eq
)
2975 /* validation done, copy the nested actions. */
2976 start
= add_nested_action_start(sfa
, OVS_ACTION_ATTR_CHECK_PKT_LEN
,
2981 arg
.pkt_len
= nla_get_u16(a
[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN
]);
2982 arg
.exec_for_lesser_equal
=
2983 last
|| !actions_may_change_flow(acts_if_lesser_eq
);
2984 arg
.exec_for_greater
=
2985 last
|| !actions_may_change_flow(acts_if_greater
);
2987 err
= ovs_nla_add_action(sfa
, OVS_CHECK_PKT_LEN_ATTR_ARG
, &arg
,
2992 nested_acts_start
= add_nested_action_start(sfa
,
2993 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL
, log
);
2994 if (nested_acts_start
< 0)
2995 return nested_acts_start
;
2997 err
= __ovs_nla_copy_actions(net
, acts_if_lesser_eq
, key
, sfa
,
2998 eth_type
, vlan_tci
, mpls_label_count
, log
);
3003 add_nested_action_end(*sfa
, nested_acts_start
);
3005 nested_acts_start
= add_nested_action_start(sfa
,
3006 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER
, log
);
3007 if (nested_acts_start
< 0)
3008 return nested_acts_start
;
3010 err
= __ovs_nla_copy_actions(net
, acts_if_greater
, key
, sfa
,
3011 eth_type
, vlan_tci
, mpls_label_count
, log
);
3016 add_nested_action_end(*sfa
, nested_acts_start
);
3017 add_nested_action_end(*sfa
, start
);
3021 static int copy_action(const struct nlattr
*from
,
3022 struct sw_flow_actions
**sfa
, bool log
)
3024 int totlen
= NLA_ALIGN(from
->nla_len
);
3027 to
= reserve_sfa_size(sfa
, from
->nla_len
, log
);
3031 memcpy(to
, from
, totlen
);
3035 static int __ovs_nla_copy_actions(struct net
*net
, const struct nlattr
*attr
,
3036 const struct sw_flow_key
*key
,
3037 struct sw_flow_actions
**sfa
,
3038 __be16 eth_type
, __be16 vlan_tci
,
3039 u32 mpls_label_count
, bool log
)
3041 u8 mac_proto
= ovs_key_mac_proto(key
);
3042 const struct nlattr
*a
;
3045 nla_for_each_nested(a
, attr
, rem
) {
3046 /* Expected argument lengths, (u32)-1 for variable length. */
3047 static const u32 action_lens
[OVS_ACTION_ATTR_MAX
+ 1] = {
3048 [OVS_ACTION_ATTR_OUTPUT
] = sizeof(u32
),
3049 [OVS_ACTION_ATTR_RECIRC
] = sizeof(u32
),
3050 [OVS_ACTION_ATTR_USERSPACE
] = (u32
)-1,
3051 [OVS_ACTION_ATTR_PUSH_MPLS
] = sizeof(struct ovs_action_push_mpls
),
3052 [OVS_ACTION_ATTR_POP_MPLS
] = sizeof(__be16
),
3053 [OVS_ACTION_ATTR_PUSH_VLAN
] = sizeof(struct ovs_action_push_vlan
),
3054 [OVS_ACTION_ATTR_POP_VLAN
] = 0,
3055 [OVS_ACTION_ATTR_SET
] = (u32
)-1,
3056 [OVS_ACTION_ATTR_SET_MASKED
] = (u32
)-1,
3057 [OVS_ACTION_ATTR_SAMPLE
] = (u32
)-1,
3058 [OVS_ACTION_ATTR_HASH
] = sizeof(struct ovs_action_hash
),
3059 [OVS_ACTION_ATTR_CT
] = (u32
)-1,
3060 [OVS_ACTION_ATTR_CT_CLEAR
] = 0,
3061 [OVS_ACTION_ATTR_TRUNC
] = sizeof(struct ovs_action_trunc
),
3062 [OVS_ACTION_ATTR_PUSH_ETH
] = sizeof(struct ovs_action_push_eth
),
3063 [OVS_ACTION_ATTR_POP_ETH
] = 0,
3064 [OVS_ACTION_ATTR_PUSH_NSH
] = (u32
)-1,
3065 [OVS_ACTION_ATTR_POP_NSH
] = 0,
3066 [OVS_ACTION_ATTR_METER
] = sizeof(u32
),
3067 [OVS_ACTION_ATTR_CLONE
] = (u32
)-1,
3068 [OVS_ACTION_ATTR_CHECK_PKT_LEN
] = (u32
)-1,
3069 [OVS_ACTION_ATTR_ADD_MPLS
] = sizeof(struct ovs_action_add_mpls
),
3070 [OVS_ACTION_ATTR_DEC_TTL
] = (u32
)-1,
3072 const struct ovs_action_push_vlan
*vlan
;
3073 int type
= nla_type(a
);
3076 if (type
> OVS_ACTION_ATTR_MAX
||
3077 (action_lens
[type
] != nla_len(a
) &&
3078 action_lens
[type
] != (u32
)-1))
3083 case OVS_ACTION_ATTR_UNSPEC
:
3086 case OVS_ACTION_ATTR_USERSPACE
:
3087 err
= validate_userspace(a
);
3092 case OVS_ACTION_ATTR_OUTPUT
:
3093 if (nla_get_u32(a
) >= DP_MAX_PORTS
)
3097 case OVS_ACTION_ATTR_TRUNC
: {
3098 const struct ovs_action_trunc
*trunc
= nla_data(a
);
3100 if (trunc
->max_len
< ETH_HLEN
)
3105 case OVS_ACTION_ATTR_HASH
: {
3106 const struct ovs_action_hash
*act_hash
= nla_data(a
);
3108 switch (act_hash
->hash_alg
) {
3109 case OVS_HASH_ALG_L4
:
3118 case OVS_ACTION_ATTR_POP_VLAN
:
3119 if (mac_proto
!= MAC_PROTO_ETHERNET
)
3121 vlan_tci
= htons(0);
3124 case OVS_ACTION_ATTR_PUSH_VLAN
:
3125 if (mac_proto
!= MAC_PROTO_ETHERNET
)
3128 if (!eth_type_vlan(vlan
->vlan_tpid
))
3130 if (!(vlan
->vlan_tci
& htons(VLAN_CFI_MASK
)))
3132 vlan_tci
= vlan
->vlan_tci
;
3135 case OVS_ACTION_ATTR_RECIRC
:
3138 case OVS_ACTION_ATTR_ADD_MPLS
: {
3139 const struct ovs_action_add_mpls
*mpls
= nla_data(a
);
3141 if (!eth_p_mpls(mpls
->mpls_ethertype
))
3144 if (mpls
->tun_flags
& OVS_MPLS_L3_TUNNEL_FLAG_MASK
) {
3145 if (vlan_tci
& htons(VLAN_CFI_MASK
) ||
3146 (eth_type
!= htons(ETH_P_IP
) &&
3147 eth_type
!= htons(ETH_P_IPV6
) &&
3148 eth_type
!= htons(ETH_P_ARP
) &&
3149 eth_type
!= htons(ETH_P_RARP
) &&
3150 !eth_p_mpls(eth_type
)))
3154 if (mac_proto
== MAC_PROTO_ETHERNET
) {
3155 mpls_label_count
= 1;
3156 mac_proto
= MAC_PROTO_NONE
;
3161 eth_type
= mpls
->mpls_ethertype
;
3165 case OVS_ACTION_ATTR_PUSH_MPLS
: {
3166 const struct ovs_action_push_mpls
*mpls
= nla_data(a
);
3168 if (!eth_p_mpls(mpls
->mpls_ethertype
))
3170 /* Prohibit push MPLS other than to a white list
3171 * for packets that have a known tag order.
3173 if (vlan_tci
& htons(VLAN_CFI_MASK
) ||
3174 (eth_type
!= htons(ETH_P_IP
) &&
3175 eth_type
!= htons(ETH_P_IPV6
) &&
3176 eth_type
!= htons(ETH_P_ARP
) &&
3177 eth_type
!= htons(ETH_P_RARP
) &&
3178 !eth_p_mpls(eth_type
)))
3180 eth_type
= mpls
->mpls_ethertype
;
3185 case OVS_ACTION_ATTR_POP_MPLS
: {
3187 if (vlan_tci
& htons(VLAN_CFI_MASK
) ||
3188 !eth_p_mpls(eth_type
))
3191 /* Disallow subsequent L2.5+ set actions and mpls_pop
3192 * actions once the last MPLS label in the packet is
3193 * is popped as there is no check here to ensure that
3194 * the new eth type is valid and thus set actions could
3195 * write off the end of the packet or otherwise corrupt
3198 * Support for these actions is planned using packet
3201 proto
= nla_get_be16(a
);
3203 if (proto
== htons(ETH_P_TEB
) &&
3204 mac_proto
!= MAC_PROTO_NONE
)
3209 if (!eth_p_mpls(proto
) || !mpls_label_count
)
3210 eth_type
= htons(0);
3217 case OVS_ACTION_ATTR_SET
:
3218 err
= validate_set(a
, key
, sfa
,
3219 &skip_copy
, mac_proto
, eth_type
,
3225 case OVS_ACTION_ATTR_SET_MASKED
:
3226 err
= validate_set(a
, key
, sfa
,
3227 &skip_copy
, mac_proto
, eth_type
,
3233 case OVS_ACTION_ATTR_SAMPLE
: {
3234 bool last
= nla_is_last(a
, rem
);
3236 err
= validate_and_copy_sample(net
, a
, key
, sfa
,
3246 case OVS_ACTION_ATTR_CT
:
3247 err
= ovs_ct_copy_action(net
, a
, key
, sfa
, log
);
3253 case OVS_ACTION_ATTR_CT_CLEAR
:
3256 case OVS_ACTION_ATTR_PUSH_ETH
:
3257 /* Disallow pushing an Ethernet header if one
3258 * is already present */
3259 if (mac_proto
!= MAC_PROTO_NONE
)
3261 mac_proto
= MAC_PROTO_ETHERNET
;
3264 case OVS_ACTION_ATTR_POP_ETH
:
3265 if (mac_proto
!= MAC_PROTO_ETHERNET
)
3267 if (vlan_tci
& htons(VLAN_CFI_MASK
))
3269 mac_proto
= MAC_PROTO_NONE
;
3272 case OVS_ACTION_ATTR_PUSH_NSH
:
3273 if (mac_proto
!= MAC_PROTO_ETHERNET
) {
3276 next_proto
= tun_p_from_eth_p(eth_type
);
3280 mac_proto
= MAC_PROTO_NONE
;
3281 if (!validate_nsh(nla_data(a
), false, true, true))
3285 case OVS_ACTION_ATTR_POP_NSH
: {
3288 if (eth_type
!= htons(ETH_P_NSH
))
3290 inner_proto
= tun_p_to_eth_p(key
->nsh
.base
.np
);
3293 if (key
->nsh
.base
.np
== TUN_P_ETHERNET
)
3294 mac_proto
= MAC_PROTO_ETHERNET
;
3296 mac_proto
= MAC_PROTO_NONE
;
3300 case OVS_ACTION_ATTR_METER
:
3301 /* Non-existent meters are simply ignored. */
3304 case OVS_ACTION_ATTR_CLONE
: {
3305 bool last
= nla_is_last(a
, rem
);
3307 err
= validate_and_copy_clone(net
, a
, key
, sfa
,
3317 case OVS_ACTION_ATTR_CHECK_PKT_LEN
: {
3318 bool last
= nla_is_last(a
, rem
);
3320 err
= validate_and_copy_check_pkt_len(net
, a
, key
, sfa
,
3331 case OVS_ACTION_ATTR_DEC_TTL
:
3332 err
= validate_and_copy_dec_ttl(net
, a
, key
, sfa
,
3334 mpls_label_count
, log
);
3341 OVS_NLERR(log
, "Unknown Action type %d", type
);
3345 err
= copy_action(a
, sfa
, log
);
3357 /* 'key' must be the masked key. */
3358 int ovs_nla_copy_actions(struct net
*net
, const struct nlattr
*attr
,
3359 const struct sw_flow_key
*key
,
3360 struct sw_flow_actions
**sfa
, bool log
)
3363 u32 mpls_label_count
= 0;
3365 *sfa
= nla_alloc_flow_actions(min(nla_len(attr
), MAX_ACTIONS_BUFSIZE
));
3367 return PTR_ERR(*sfa
);
3369 if (eth_p_mpls(key
->eth
.type
))
3370 mpls_label_count
= hweight_long(key
->mpls
.num_labels_mask
);
3372 (*sfa
)->orig_len
= nla_len(attr
);
3373 err
= __ovs_nla_copy_actions(net
, attr
, key
, sfa
, key
->eth
.type
,
3374 key
->eth
.vlan
.tci
, mpls_label_count
, log
);
3376 ovs_nla_free_flow_actions(*sfa
);
3381 static int sample_action_to_attr(const struct nlattr
*attr
,
3382 struct sk_buff
*skb
)
3384 struct nlattr
*start
, *ac_start
= NULL
, *sample_arg
;
3385 int err
= 0, rem
= nla_len(attr
);
3386 const struct sample_arg
*arg
;
3387 struct nlattr
*actions
;
3389 start
= nla_nest_start_noflag(skb
, OVS_ACTION_ATTR_SAMPLE
);
3393 sample_arg
= nla_data(attr
);
3394 arg
= nla_data(sample_arg
);
3395 actions
= nla_next(sample_arg
, &rem
);
3397 if (nla_put_u32(skb
, OVS_SAMPLE_ATTR_PROBABILITY
, arg
->probability
)) {
3402 ac_start
= nla_nest_start_noflag(skb
, OVS_SAMPLE_ATTR_ACTIONS
);
3408 err
= ovs_nla_put_actions(actions
, rem
, skb
);
3412 nla_nest_cancel(skb
, ac_start
);
3413 nla_nest_cancel(skb
, start
);
3415 nla_nest_end(skb
, ac_start
);
3416 nla_nest_end(skb
, start
);
3422 static int clone_action_to_attr(const struct nlattr
*attr
,
3423 struct sk_buff
*skb
)
3425 struct nlattr
*start
;
3426 int err
= 0, rem
= nla_len(attr
);
3428 start
= nla_nest_start_noflag(skb
, OVS_ACTION_ATTR_CLONE
);
3432 err
= ovs_nla_put_actions(nla_data(attr
), rem
, skb
);
3435 nla_nest_cancel(skb
, start
);
3437 nla_nest_end(skb
, start
);
3442 static int check_pkt_len_action_to_attr(const struct nlattr
*attr
,
3443 struct sk_buff
*skb
)
3445 struct nlattr
*start
, *ac_start
= NULL
;
3446 const struct check_pkt_len_arg
*arg
;
3447 const struct nlattr
*a
, *cpl_arg
;
3448 int err
= 0, rem
= nla_len(attr
);
3450 start
= nla_nest_start_noflag(skb
, OVS_ACTION_ATTR_CHECK_PKT_LEN
);
3454 /* The first nested attribute in 'attr' is always
3455 * 'OVS_CHECK_PKT_LEN_ATTR_ARG'.
3457 cpl_arg
= nla_data(attr
);
3458 arg
= nla_data(cpl_arg
);
3460 if (nla_put_u16(skb
, OVS_CHECK_PKT_LEN_ATTR_PKT_LEN
, arg
->pkt_len
)) {
3465 /* Second nested attribute in 'attr' is always
3466 * 'OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL'.
3468 a
= nla_next(cpl_arg
, &rem
);
3469 ac_start
= nla_nest_start_noflag(skb
,
3470 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL
);
3476 err
= ovs_nla_put_actions(nla_data(a
), nla_len(a
), skb
);
3478 nla_nest_cancel(skb
, ac_start
);
3481 nla_nest_end(skb
, ac_start
);
3484 /* Third nested attribute in 'attr' is always
3485 * OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER.
3487 a
= nla_next(a
, &rem
);
3488 ac_start
= nla_nest_start_noflag(skb
,
3489 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER
);
3495 err
= ovs_nla_put_actions(nla_data(a
), nla_len(a
), skb
);
3497 nla_nest_cancel(skb
, ac_start
);
3500 nla_nest_end(skb
, ac_start
);
3503 nla_nest_end(skb
, start
);
3507 nla_nest_cancel(skb
, start
);
3511 static int dec_ttl_action_to_attr(const struct nlattr
*attr
,
3512 struct sk_buff
*skb
)
3514 struct nlattr
*start
, *action_start
;
3515 const struct nlattr
*a
;
3518 start
= nla_nest_start_noflag(skb
, OVS_ACTION_ATTR_DEC_TTL
);
3522 nla_for_each_attr(a
, nla_data(attr
), nla_len(attr
), rem
) {
3523 switch (nla_type(a
)) {
3524 case OVS_DEC_TTL_ATTR_ACTION
:
3526 action_start
= nla_nest_start_noflag(skb
, OVS_DEC_TTL_ATTR_ACTION
);
3527 if (!action_start
) {
3532 err
= ovs_nla_put_actions(nla_data(a
), nla_len(a
), skb
);
3536 nla_nest_end(skb
, action_start
);
3540 /* Ignore all other option to be future compatible */
3545 nla_nest_end(skb
, start
);
3549 nla_nest_cancel(skb
, start
);
3553 static int set_action_to_attr(const struct nlattr
*a
, struct sk_buff
*skb
)
3555 const struct nlattr
*ovs_key
= nla_data(a
);
3556 int key_type
= nla_type(ovs_key
);
3557 struct nlattr
*start
;
3561 case OVS_KEY_ATTR_TUNNEL_INFO
: {
3562 struct ovs_tunnel_info
*ovs_tun
= nla_data(ovs_key
);
3563 struct ip_tunnel_info
*tun_info
= &ovs_tun
->tun_dst
->u
.tun_info
;
3565 start
= nla_nest_start_noflag(skb
, OVS_ACTION_ATTR_SET
);
3569 err
= ip_tun_to_nlattr(skb
, &tun_info
->key
,
3570 ip_tunnel_info_opts(tun_info
),
3571 tun_info
->options_len
,
3572 ip_tunnel_info_af(tun_info
), tun_info
->mode
);
3575 nla_nest_end(skb
, start
);
3579 if (nla_put(skb
, OVS_ACTION_ATTR_SET
, nla_len(a
), ovs_key
))
3587 static int masked_set_action_to_set_action_attr(const struct nlattr
*a
,
3588 struct sk_buff
*skb
)
3590 const struct nlattr
*ovs_key
= nla_data(a
);
3592 size_t key_len
= nla_len(ovs_key
) / 2;
3594 /* Revert the conversion we did from a non-masked set action to
3595 * masked set action.
3597 nla
= nla_nest_start_noflag(skb
, OVS_ACTION_ATTR_SET
);
3601 if (nla_put(skb
, nla_type(ovs_key
), key_len
, nla_data(ovs_key
)))
3604 nla_nest_end(skb
, nla
);
3608 int ovs_nla_put_actions(const struct nlattr
*attr
, int len
, struct sk_buff
*skb
)
3610 const struct nlattr
*a
;
3613 nla_for_each_attr(a
, attr
, len
, rem
) {
3614 int type
= nla_type(a
);
3617 case OVS_ACTION_ATTR_SET
:
3618 err
= set_action_to_attr(a
, skb
);
3623 case OVS_ACTION_ATTR_SET_TO_MASKED
:
3624 err
= masked_set_action_to_set_action_attr(a
, skb
);
3629 case OVS_ACTION_ATTR_SAMPLE
:
3630 err
= sample_action_to_attr(a
, skb
);
3635 case OVS_ACTION_ATTR_CT
:
3636 err
= ovs_ct_action_to_attr(nla_data(a
), skb
);
3641 case OVS_ACTION_ATTR_CLONE
:
3642 err
= clone_action_to_attr(a
, skb
);
3647 case OVS_ACTION_ATTR_CHECK_PKT_LEN
:
3648 err
= check_pkt_len_action_to_attr(a
, skb
);
3653 case OVS_ACTION_ATTR_DEC_TTL
:
3654 err
= dec_ttl_action_to_attr(a
, skb
);
3660 if (nla_put(skb
, type
, nla_len(a
), nla_data(a
)))