Loading...
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * Copyright (c) 2007-2017 Nicira, Inc.
4 */
5
6#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
7
8#include "flow.h"
9#include "datapath.h"
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>
21#include <linux/in.h>
22#include <linux/rcupdate.h>
23#include <linux/if_arp.h>
24#include <linux/ip.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>
33#include <net/ip.h>
34#include <net/ipv6.h>
35#include <net/ndisc.h>
36#include <net/mpls.h>
37#include <net/vxlan.h>
38#include <net/tun_proto.h>
39#include <net/erspan.h>
40
41#include "flow_netlink.h"
42
43struct ovs_len_tbl {
44 int len;
45 const struct ovs_len_tbl *next;
46};
47
48#define OVS_ATTR_NESTED -1
49#define OVS_ATTR_VARIABLE -2
50
51static bool actions_may_change_flow(const struct nlattr *actions)
52{
53 struct nlattr *nla;
54 int rem;
55
56 nla_for_each_nested(nla, actions, rem) {
57 u16 action = nla_type(nla);
58
59 switch (action) {
60 case OVS_ACTION_ATTR_OUTPUT:
61 case OVS_ACTION_ATTR_RECIRC:
62 case OVS_ACTION_ATTR_TRUNC:
63 case OVS_ACTION_ATTR_USERSPACE:
64 break;
65
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:
84 default:
85 return true;
86 }
87 }
88 return false;
89}
90
91static void update_range(struct sw_flow_match *match,
92 size_t offset, size_t size, bool is_mask)
93{
94 struct sw_flow_key_range *range;
95 size_t start = rounddown(offset, sizeof(long));
96 size_t end = roundup(offset + size, sizeof(long));
97
98 if (!is_mask)
99 range = &match->range;
100 else
101 range = &match->mask->range;
102
103 if (range->start == range->end) {
104 range->start = start;
105 range->end = end;
106 return;
107 }
108
109 if (range->start > start)
110 range->start = start;
111
112 if (range->end < end)
113 range->end = end;
114}
115
116#define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
117 do { \
118 update_range(match, offsetof(struct sw_flow_key, field), \
119 sizeof((match)->key->field), is_mask); \
120 if (is_mask) \
121 (match)->mask->key.field = value; \
122 else \
123 (match)->key->field = value; \
124 } while (0)
125
126#define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
127 do { \
128 update_range(match, offset, len, is_mask); \
129 if (is_mask) \
130 memcpy((u8 *)&(match)->mask->key + offset, value_p, \
131 len); \
132 else \
133 memcpy((u8 *)(match)->key + offset, value_p, len); \
134 } while (0)
135
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)
139
140#define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
141 do { \
142 update_range(match, offsetof(struct sw_flow_key, field), \
143 sizeof((match)->key->field), is_mask); \
144 if (is_mask) \
145 memset((u8 *)&(match)->mask->key.field, value, \
146 sizeof((match)->mask->key.field)); \
147 else \
148 memset((u8 *)&(match)->key->field, value, \
149 sizeof((match)->key->field)); \
150 } while (0)
151
152static bool match_validate(const struct sw_flow_match *match,
153 u64 key_attrs, u64 mask_attrs, bool log)
154{
155 u64 key_expected = 0;
156 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
157
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));
174
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));
179
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;
186 }
187
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;
192 }
193
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;
199 }
200
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;
206 }
207
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;
212 }
213
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;
220 }
221 }
222
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;
227 }
228 }
229 }
230
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;
236 }
237
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;
243 }
244
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;
249 }
250
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;
257 }
258 }
259
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;
264
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
272 * at the same time.
273 */
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;
277 }
278 }
279 }
280 }
281
282 if (match->key->eth.type == htons(ETH_P_NSH)) {
283 key_expected |= 1 << OVS_KEY_ATTR_NSH;
284 if (match->mask &&
285 match->mask->key.eth.type == htons(0xffff)) {
286 mask_allowed |= 1 << OVS_KEY_ATTR_NSH;
287 }
288 }
289
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);
295 return false;
296 }
297
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);
303 return false;
304 }
305
306 return true;
307}
308
309size_t ovs_tun_key_attr_size(void)
310{
311 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
312 * updating this function.
313 */
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.
326 */
327 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
328 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
329}
330
331static size_t ovs_nsh_key_attr_size(void)
332{
333 /* Whenever adding new OVS_NSH_KEY_ FIELDS, we should consider
334 * updating this function.
335 */
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
339 * the small one.
340 */
341 + nla_total_size(NSH_CTX_HDRS_MAX_LEN);
342}
343
344size_t ovs_key_attr_size(void)
345{
346 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
347 * updating this function.
348 */
349 BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO != 29);
350
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 */
373}
374
375static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
376 [OVS_VXLAN_EXT_GBP] = { .len = sizeof(u32) },
377};
378
379static 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 },
397};
398
399static const struct ovs_len_tbl
400ovs_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 },
404};
405
406/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
407static 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, },
440};
441
442static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
443{
444 return expected_len == attr_len ||
445 expected_len == OVS_ATTR_NESTED ||
446 expected_len == OVS_ATTR_VARIABLE;
447}
448
449static bool is_all_zero(const u8 *fp, size_t size)
450{
451 int i;
452
453 if (!fp)
454 return false;
455
456 for (i = 0; i < size; i++)
457 if (fp[i])
458 return false;
459
460 return true;
461}
462
463static int __parse_flow_nlattrs(const struct nlattr *attr,
464 const struct nlattr *a[],
465 u64 *attrsp, bool log, bool nz)
466{
467 const struct nlattr *nla;
468 u64 attrs;
469 int rem;
470
471 attrs = *attrsp;
472 nla_for_each_nested(nla, attr, rem) {
473 u16 type = nla_type(nla);
474 int expected_len;
475
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);
479 return -EINVAL;
480 }
481
482 if (attrs & (1 << type)) {
483 OVS_NLERR(log, "Duplicate key (type %d).", type);
484 return -EINVAL;
485 }
486
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);
491 return -EINVAL;
492 }
493
494 if (!nz || !is_all_zero(nla_data(nla), nla_len(nla))) {
495 attrs |= 1 << type;
496 a[type] = nla;
497 }
498 }
499 if (rem) {
500 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
501 return -EINVAL;
502 }
503
504 *attrsp = attrs;
505 return 0;
506}
507
508static int parse_flow_mask_nlattrs(const struct nlattr *attr,
509 const struct nlattr *a[], u64 *attrsp,
510 bool log)
511{
512 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
513}
514
515int parse_flow_nlattrs(const struct nlattr *attr, const struct nlattr *a[],
516 u64 *attrsp, bool log)
517{
518 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
519}
520
521static int genev_tun_opt_from_nlattr(const struct nlattr *a,
522 struct sw_flow_match *match, bool is_mask,
523 bool log)
524{
525 unsigned long opt_key_offset;
526
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));
530 return -EINVAL;
531 }
532
533 if (nla_len(a) % 4 != 0) {
534 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
535 nla_len(a));
536 return -EINVAL;
537 }
538
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.
542 */
543 if (!is_mask) {
544 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
545 false);
546 } else {
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
555 * information later.
556 */
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));
560 return -EINVAL;
561 }
562
563 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
564 }
565
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);
569 return 0;
570}
571
572static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
573 struct sw_flow_match *match, bool is_mask,
574 bool log)
575{
576 struct nlattr *a;
577 int rem;
578 unsigned long opt_key_offset;
579 struct vxlan_metadata opts;
580
581 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
582
583 memset(&opts, 0, sizeof(opts));
584 nla_for_each_nested(a, attr, rem) {
585 int type = nla_type(a);
586
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);
590 return -EINVAL;
591 }
592
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",
596 type, nla_len(a),
597 ovs_vxlan_ext_key_lens[type].len);
598 return -EINVAL;
599 }
600
601 switch (type) {
602 case OVS_VXLAN_EXT_GBP:
603 opts.gbp = nla_get_u32(a);
604 break;
605 default:
606 OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
607 type);
608 return -EINVAL;
609 }
610 }
611 if (rem) {
612 OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
613 rem);
614 return -EINVAL;
615 }
616
617 if (!is_mask)
618 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
619 else
620 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
621
622 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
623 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
624 is_mask);
625 return 0;
626}
627
628static int erspan_tun_opt_from_nlattr(const struct nlattr *a,
629 struct sw_flow_match *match, bool is_mask,
630 bool log)
631{
632 unsigned long opt_key_offset;
633
634 BUILD_BUG_ON(sizeof(struct erspan_metadata) >
635 sizeof(match->key->tun_opts));
636
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));
640 return -EINVAL;
641 }
642
643 if (!is_mask)
644 SW_FLOW_KEY_PUT(match, tun_opts_len,
645 sizeof(struct erspan_metadata), false);
646 else
647 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
648
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);
652 return 0;
653}
654
655static int ip_tun_from_nlattr(const struct nlattr *attr,
656 struct sw_flow_match *match, bool is_mask,
657 bool log)
658{
659 bool ttl = false, ipv4 = false, ipv6 = false;
660 bool info_bridge_mode = false;
661 __be16 tun_flags = 0;
662 int opts_type = 0;
663 struct nlattr *a;
664 int rem;
665
666 nla_for_each_nested(a, attr, rem) {
667 int type = nla_type(a);
668 int err;
669
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);
673 return -EINVAL;
674 }
675
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);
680 return -EINVAL;
681 }
682
683 switch (type) {
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;
688 break;
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);
692 ipv4 = true;
693 break;
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);
697 ipv4 = true;
698 break;
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);
702 ipv6 = true;
703 break;
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);
707 ipv6 = true;
708 break;
709 case OVS_TUNNEL_KEY_ATTR_TOS:
710 SW_FLOW_KEY_PUT(match, tun_key.tos,
711 nla_get_u8(a), is_mask);
712 break;
713 case OVS_TUNNEL_KEY_ATTR_TTL:
714 SW_FLOW_KEY_PUT(match, tun_key.ttl,
715 nla_get_u8(a), is_mask);
716 ttl = true;
717 break;
718 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
719 tun_flags |= TUNNEL_DONT_FRAGMENT;
720 break;
721 case OVS_TUNNEL_KEY_ATTR_CSUM:
722 tun_flags |= TUNNEL_CSUM;
723 break;
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);
727 break;
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);
731 break;
732 case OVS_TUNNEL_KEY_ATTR_OAM:
733 tun_flags |= TUNNEL_OAM;
734 break;
735 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
736 if (opts_type) {
737 OVS_NLERR(log, "Multiple metadata blocks provided");
738 return -EINVAL;
739 }
740
741 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
742 if (err)
743 return err;
744
745 tun_flags |= TUNNEL_GENEVE_OPT;
746 opts_type = type;
747 break;
748 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
749 if (opts_type) {
750 OVS_NLERR(log, "Multiple metadata blocks provided");
751 return -EINVAL;
752 }
753
754 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
755 if (err)
756 return err;
757
758 tun_flags |= TUNNEL_VXLAN_OPT;
759 opts_type = type;
760 break;
761 case OVS_TUNNEL_KEY_ATTR_PAD:
762 break;
763 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
764 if (opts_type) {
765 OVS_NLERR(log, "Multiple metadata blocks provided");
766 return -EINVAL;
767 }
768
769 err = erspan_tun_opt_from_nlattr(a, match, is_mask,
770 log);
771 if (err)
772 return err;
773
774 tun_flags |= TUNNEL_ERSPAN_OPT;
775 opts_type = type;
776 break;
777 case OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE:
778 info_bridge_mode = true;
779 ipv4 = true;
780 break;
781 default:
782 OVS_NLERR(log, "Unknown IP tunnel attribute %d",
783 type);
784 return -EINVAL;
785 }
786 }
787
788 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
789 if (is_mask)
790 SW_FLOW_KEY_MEMSET_FIELD(match, tun_proto, 0xff, true);
791 else
792 SW_FLOW_KEY_PUT(match, tun_proto, ipv6 ? AF_INET6 : AF_INET,
793 false);
794
795 if (rem > 0) {
796 OVS_NLERR(log, "IP tunnel attribute has %d unknown bytes.",
797 rem);
798 return -EINVAL;
799 }
800
801 if (ipv4 && ipv6) {
802 OVS_NLERR(log, "Mixed IPv4 and IPv6 tunnel attributes");
803 return -EINVAL;
804 }
805
806 if (!is_mask) {
807 if (!ipv4 && !ipv6) {
808 OVS_NLERR(log, "IP tunnel dst address not specified");
809 return -EINVAL;
810 }
811 if (ipv4) {
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");
821 return -EINVAL;
822 }
823 } else if (!match->key->tun_key.u.ipv4.dst) {
824 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
825 return -EINVAL;
826 }
827 }
828 if (ipv6 && ipv6_addr_any(&match->key->tun_key.u.ipv6.dst)) {
829 OVS_NLERR(log, "IPv6 tunnel dst address is zero");
830 return -EINVAL;
831 }
832
833 if (!ttl && !info_bridge_mode) {
834 OVS_NLERR(log, "IP tunnel TTL not specified.");
835 return -EINVAL;
836 }
837 }
838
839 return opts_type;
840}
841
842static int vxlan_opt_to_nlattr(struct sk_buff *skb,
843 const void *tun_opts, int swkey_tun_opts_len)
844{
845 const struct vxlan_metadata *opts = tun_opts;
846 struct nlattr *nla;
847
848 nla = nla_nest_start_noflag(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
849 if (!nla)
850 return -EMSGSIZE;
851
852 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
853 return -EMSGSIZE;
854
855 nla_nest_end(skb, nla);
856 return 0;
857}
858
859static 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)
863{
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))
867 return -EMSGSIZE;
868
869 if (mode & IP_TUNNEL_INFO_BRIDGE)
870 return nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE)
871 ? -EMSGSIZE : 0;
872
873 switch (tun_proto) {
874 case AF_INET:
875 if (output->u.ipv4.src &&
876 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
877 output->u.ipv4.src))
878 return -EMSGSIZE;
879 if (output->u.ipv4.dst &&
880 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
881 output->u.ipv4.dst))
882 return -EMSGSIZE;
883 break;
884 case AF_INET6:
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))
888 return -EMSGSIZE;
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))
892 return -EMSGSIZE;
893 break;
894 }
895 if (output->tos &&
896 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
897 return -EMSGSIZE;
898 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
899 return -EMSGSIZE;
900 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
901 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
902 return -EMSGSIZE;
903 if ((output->tun_flags & TUNNEL_CSUM) &&
904 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
905 return -EMSGSIZE;
906 if (output->tp_src &&
907 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
908 return -EMSGSIZE;
909 if (output->tp_dst &&
910 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
911 return -EMSGSIZE;
912 if ((output->tun_flags & TUNNEL_OAM) &&
913 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
914 return -EMSGSIZE;
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))
919 return -EMSGSIZE;
920 else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
921 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
922 return -EMSGSIZE;
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))
926 return -EMSGSIZE;
927 }
928
929 return 0;
930}
931
932static 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)
936{
937 struct nlattr *nla;
938 int err;
939
940 nla = nla_nest_start_noflag(skb, OVS_KEY_ATTR_TUNNEL);
941 if (!nla)
942 return -EMSGSIZE;
943
944 err = __ip_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len,
945 tun_proto, mode);
946 if (err)
947 return err;
948
949 nla_nest_end(skb, nla);
950 return 0;
951}
952
953int ovs_nla_put_tunnel_info(struct sk_buff *skb,
954 struct ip_tunnel_info *tun_info)
955{
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);
960}
961
962static int encode_vlan_from_nlattrs(struct sw_flow_match *match,
963 const struct nlattr *a[],
964 bool is_mask, bool inner)
965{
966 __be16 tci = 0;
967 __be16 tpid = 0;
968
969 if (a[OVS_KEY_ATTR_VLAN])
970 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
971
972 if (a[OVS_KEY_ATTR_ETHERTYPE])
973 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
974
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);
978 } else {
979 SW_FLOW_KEY_PUT(match, eth.cvlan.tpid, tpid, is_mask);
980 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, tci, is_mask);
981 }
982 return 0;
983}
984
985static int validate_vlan_from_nlattrs(const struct sw_flow_match *match,
986 u64 key_attrs, bool inner,
987 const struct nlattr **a, bool log)
988{
989 __be16 tci = 0;
990
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])))) {
994 /* Not a VLAN. */
995 return 0;
996 }
997
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");
1001 return -EINVAL;
1002 }
1003
1004 if (a[OVS_KEY_ATTR_VLAN])
1005 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1006
1007 if (!(tci & htons(VLAN_CFI_MASK))) {
1008 if (tci) {
1009 OVS_NLERR(log, "%s TCI does not have VLAN_CFI_MASK bit set.",
1010 (inner) ? "C-VLAN" : "VLAN");
1011 return -EINVAL;
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");
1016 return -EINVAL;
1017 }
1018 }
1019
1020 return 1;
1021}
1022
1023static 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)
1026{
1027 __be16 tci = 0;
1028 __be16 tpid = 0;
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));
1033
1034 if (!(key_attrs & (1 << OVS_KEY_ATTR_ENCAP))) {
1035 /* Not a VLAN. */
1036 return 0;
1037 }
1038
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");
1042 return -EINVAL;
1043 }
1044
1045 if (a[OVS_KEY_ATTR_VLAN])
1046 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1047
1048 if (a[OVS_KEY_ATTR_ETHERTYPE])
1049 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1050
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));
1054 return -EINVAL;
1055 }
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");
1059 return -EINVAL;
1060 }
1061
1062 return 1;
1063}
1064
1065static int __parse_vlan_from_nlattrs(struct sw_flow_match *match,
1066 u64 *key_attrs, bool inner,
1067 const struct nlattr **a, bool is_mask,
1068 bool log)
1069{
1070 int err;
1071 const struct nlattr *encap;
1072
1073 if (!is_mask)
1074 err = validate_vlan_from_nlattrs(match, *key_attrs, inner,
1075 a, log);
1076 else
1077 err = validate_vlan_mask_from_nlattrs(match, *key_attrs, inner,
1078 a, log);
1079 if (err <= 0)
1080 return err;
1081
1082 err = encode_vlan_from_nlattrs(match, a, is_mask, inner);
1083 if (err)
1084 return err;
1085
1086 *key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1087 *key_attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
1088 *key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1089
1090 encap = a[OVS_KEY_ATTR_ENCAP];
1091
1092 if (!is_mask)
1093 err = parse_flow_nlattrs(encap, a, key_attrs, log);
1094 else
1095 err = parse_flow_mask_nlattrs(encap, a, key_attrs, log);
1096
1097 return err;
1098}
1099
1100static int parse_vlan_from_nlattrs(struct sw_flow_match *match,
1101 u64 *key_attrs, const struct nlattr **a,
1102 bool is_mask, bool log)
1103{
1104 int err;
1105 bool encap_valid = false;
1106
1107 err = __parse_vlan_from_nlattrs(match, key_attrs, false, a,
1108 is_mask, log);
1109 if (err)
1110 return err;
1111
1112 encap_valid = !!(match->key->eth.vlan.tci & htons(VLAN_CFI_MASK));
1113 if (encap_valid) {
1114 err = __parse_vlan_from_nlattrs(match, key_attrs, true, a,
1115 is_mask, log);
1116 if (err)
1117 return err;
1118 }
1119
1120 return 0;
1121}
1122
1123static int parse_eth_type_from_nlattrs(struct sw_flow_match *match,
1124 u64 *attrs, const struct nlattr **a,
1125 bool is_mask, bool log)
1126{
1127 __be16 eth_type;
1128
1129 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1130 if (is_mask) {
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);
1136 return -EINVAL;
1137 }
1138
1139 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
1140 *attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1141 return 0;
1142}
1143
1144static 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)
1147{
1148 u8 mac_proto = MAC_PROTO_ETHERNET;
1149
1150 if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
1151 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
1152
1153 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
1154 *attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
1155 }
1156
1157 if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
1158 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
1159
1160 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
1161 *attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
1162 }
1163
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);
1168 }
1169
1170 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
1171 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1172
1173 if (is_mask) {
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);
1178 return -EINVAL;
1179 }
1180
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);
1185 }
1186
1187 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1188 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1189
1190 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1191 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1192 }
1193 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1194 if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1195 is_mask, log) < 0)
1196 return -EINVAL;
1197 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1198 }
1199
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]);
1203
1204 if (ct_state & ~CT_SUPPORTED_MASK) {
1205 OVS_NLERR(log, "ct_state flags %08x unsupported",
1206 ct_state);
1207 return -EINVAL;
1208 }
1209
1210 SW_FLOW_KEY_PUT(match, ct_state, ct_state, is_mask);
1211 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
1212 }
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]);
1216
1217 SW_FLOW_KEY_PUT(match, ct_zone, ct_zone, is_mask);
1218 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
1219 }
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]);
1223
1224 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
1225 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
1226 }
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;
1230
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);
1235 }
1236 if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)) {
1237 const struct ovs_key_ct_tuple_ipv4 *ct;
1238
1239 ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4]);
1240
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);
1247 }
1248 if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)) {
1249 const struct ovs_key_ct_tuple_ipv6 *ct;
1250
1251 ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6]);
1252
1253 SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.src, &ct->ipv6_src,
1254 sizeof(match->key->ipv6.ct_orig.src),
1255 is_mask);
1256 SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.dst, &ct->ipv6_dst,
1257 sizeof(match->key->ipv6.ct_orig.dst),
1258 is_mask);
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);
1263 }
1264
1265 /* For layer 3 packets the Ethernet type is provided
1266 * and treated as metadata but no MAC addresses are provided.
1267 */
1268 if (!(*attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
1269 (*attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)))
1270 mac_proto = MAC_PROTO_NONE;
1271
1272 /* Always exact match mac_proto */
1273 SW_FLOW_KEY_PUT(match, mac_proto, is_mask ? 0xff : mac_proto, is_mask);
1274
1275 if (mac_proto == MAC_PROTO_NONE)
1276 return parse_eth_type_from_nlattrs(match, attrs, a, is_mask,
1277 log);
1278
1279 return 0;
1280}
1281
1282int nsh_hdr_from_nlattr(const struct nlattr *attr,
1283 struct nshhdr *nh, size_t size)
1284{
1285 struct nlattr *a;
1286 int rem;
1287 u8 flags = 0;
1288 u8 ttl = 0;
1289 int mdlen = 0;
1290
1291 /* validate_nsh has check this, so we needn't do duplicate check here
1292 */
1293 if (size < NSH_BASE_HDR_LEN)
1294 return -ENOBUFS;
1295
1296 nla_for_each_nested(a, attr, rem) {
1297 int type = nla_type(a);
1298
1299 switch (type) {
1300 case OVS_NSH_KEY_ATTR_BASE: {
1301 const struct ovs_nsh_key_base *base = nla_data(a);
1302
1303 flags = base->flags;
1304 ttl = base->ttl;
1305 nh->np = base->np;
1306 nh->mdtype = base->mdtype;
1307 nh->path_hdr = base->path_hdr;
1308 break;
1309 }
1310 case OVS_NSH_KEY_ATTR_MD1:
1311 mdlen = nla_len(a);
1312 if (mdlen > size - NSH_BASE_HDR_LEN)
1313 return -ENOBUFS;
1314 memcpy(&nh->md1, nla_data(a), mdlen);
1315 break;
1316
1317 case OVS_NSH_KEY_ATTR_MD2:
1318 mdlen = nla_len(a);
1319 if (mdlen > size - NSH_BASE_HDR_LEN)
1320 return -ENOBUFS;
1321 memcpy(&nh->md2, nla_data(a), mdlen);
1322 break;
1323
1324 default:
1325 return -EINVAL;
1326 }
1327 }
1328
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);
1332
1333 return 0;
1334}
1335
1336int nsh_key_from_nlattr(const struct nlattr *attr,
1337 struct ovs_key_nsh *nsh, struct ovs_key_nsh *nsh_mask)
1338{
1339 struct nlattr *a;
1340 int rem;
1341
1342 /* validate_nsh has check this, so we needn't do duplicate check here
1343 */
1344 nla_for_each_nested(a, attr, rem) {
1345 int type = nla_type(a);
1346
1347 switch (type) {
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;
1351
1352 nsh->base = *base;
1353 nsh_mask->base = *base_mask;
1354 break;
1355 }
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;
1359
1360 memcpy(nsh->context, md1->context, sizeof(*md1));
1361 memcpy(nsh_mask->context, md1_mask->context,
1362 sizeof(*md1_mask));
1363 break;
1364 }
1365 case OVS_NSH_KEY_ATTR_MD2:
1366 /* Not supported yet */
1367 return -ENOTSUPP;
1368 default:
1369 return -EINVAL;
1370 }
1371 }
1372
1373 return 0;
1374}
1375
1376static 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)
1379{
1380 struct nlattr *a;
1381 int rem;
1382 bool has_base = false;
1383 bool has_md1 = false;
1384 bool has_md2 = false;
1385 u8 mdtype = 0;
1386 int mdlen = 0;
1387
1388 if (WARN_ON(is_push_nsh && is_mask))
1389 return -EINVAL;
1390
1391 nla_for_each_nested(a, attr, rem) {
1392 int type = nla_type(a);
1393 int i;
1394
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);
1398 return -EINVAL;
1399 }
1400
1401 if (!check_attr_len(nla_len(a),
1402 ovs_nsh_key_attr_lens[type].len)) {
1403 OVS_NLERR(
1404 log,
1405 "nsh attr %d has unexpected len %d expected %d",
1406 type,
1407 nla_len(a),
1408 ovs_nsh_key_attr_lens[type].len
1409 );
1410 return -EINVAL;
1411 }
1412
1413 switch (type) {
1414 case OVS_NSH_KEY_ATTR_BASE: {
1415 const struct ovs_nsh_key_base *base = nla_data(a);
1416
1417 has_base = true;
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,
1426 base->np, is_mask);
1427 SW_FLOW_KEY_PUT(match, nsh.base.path_hdr,
1428 base->path_hdr, is_mask);
1429 break;
1430 }
1431 case OVS_NSH_KEY_ATTR_MD1: {
1432 const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1433
1434 has_md1 = true;
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);
1438 break;
1439 }
1440 case OVS_NSH_KEY_ATTR_MD2:
1441 if (!is_push_nsh) /* Not supported MD type 2 yet */
1442 return -ENOTSUPP;
1443
1444 has_md2 = true;
1445 mdlen = nla_len(a);
1446 if (mdlen > NSH_CTX_HDRS_MAX_LEN || mdlen <= 0) {
1447 OVS_NLERR(
1448 log,
1449 "Invalid MD length %d for MD type %d",
1450 mdlen,
1451 mdtype
1452 );
1453 return -EINVAL;
1454 }
1455 break;
1456 default:
1457 OVS_NLERR(log, "Unknown nsh attribute %d",
1458 type);
1459 return -EINVAL;
1460 }
1461 }
1462
1463 if (rem > 0) {
1464 OVS_NLERR(log, "nsh attribute has %d unknown bytes.", rem);
1465 return -EINVAL;
1466 }
1467
1468 if (has_md1 && has_md2) {
1469 OVS_NLERR(
1470 1,
1471 "invalid nsh attribute: md1 and md2 are exclusive."
1472 );
1473 return -EINVAL;
1474 }
1475
1476 if (!is_mask) {
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.",
1480 mdtype);
1481 return -EINVAL;
1482 }
1483
1484 if (is_push_nsh &&
1485 (!has_base || (!has_md1 && !has_md2))) {
1486 OVS_NLERR(
1487 1,
1488 "push_nsh: missing base or metadata attributes"
1489 );
1490 return -EINVAL;
1491 }
1492 }
1493
1494 return 0;
1495}
1496
1497static 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)
1500{
1501 int err;
1502
1503 err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
1504 if (err)
1505 return err;
1506
1507 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1508 const struct ovs_key_ethernet *eth_key;
1509
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);
1516
1517 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1518 /* VLAN attribute is always parsed before getting here since it
1519 * may occur multiple times.
1520 */
1521 OVS_NLERR(log, "VLAN attribute unexpected.");
1522 return -EINVAL;
1523 }
1524
1525 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1526 err = parse_eth_type_from_nlattrs(match, &attrs, a, is_mask,
1527 log);
1528 if (err)
1529 return err;
1530 } else if (!is_mask) {
1531 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1532 }
1533 } else if (!match->key->eth.type) {
1534 OVS_NLERR(log, "Either Ethernet header or EtherType is required.");
1535 return -EINVAL;
1536 }
1537
1538 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1539 const struct ovs_key_ipv4 *ipv4_key;
1540
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);
1545 return -EINVAL;
1546 }
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);
1560 }
1561
1562 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1563 const struct ovs_key_ipv6 *ipv6_key;
1564
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);
1569 return -EINVAL;
1570 }
1571
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);
1575 return -EINVAL;
1576 }
1577
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,
1589 ipv6_key->ipv6_src,
1590 sizeof(match->key->ipv6.addr.src),
1591 is_mask);
1592 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1593 ipv6_key->ipv6_dst,
1594 sizeof(match->key->ipv6.addr.dst),
1595 is_mask);
1596
1597 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1598 }
1599
1600 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1601 const struct ovs_key_arp *arp_key;
1602
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).",
1606 arp_key->arp_op);
1607 return -EINVAL;
1608 }
1609
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);
1620
1621 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1622 }
1623
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)
1627 return -EINVAL;
1628 attrs &= ~(1 << OVS_KEY_ATTR_NSH);
1629 }
1630
1631 if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1632 const struct ovs_key_mpls *mpls_key;
1633 u32 hdr_len;
1634 u32 label_count, label_count_mask, i;
1635
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);
1639
1640 if (label_count == 0 || label_count > MPLS_LABEL_DEPTH ||
1641 hdr_len % sizeof(struct ovs_key_mpls))
1642 return -EINVAL;
1643
1644 label_count_mask = GENMASK(label_count - 1, 0);
1645
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);
1649
1650 SW_FLOW_KEY_PUT(match, mpls.num_labels_mask,
1651 label_count_mask, is_mask);
1652
1653 attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1654 }
1655
1656 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1657 const struct ovs_key_tcp *tcp_key;
1658
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);
1663 }
1664
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]),
1668 is_mask);
1669 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1670 }
1671
1672 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1673 const struct ovs_key_udp *udp_key;
1674
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);
1679 }
1680
1681 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1682 const struct ovs_key_sctp *sctp_key;
1683
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);
1688 }
1689
1690 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1691 const struct ovs_key_icmp *icmp_key;
1692
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);
1699 }
1700
1701 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1702 const struct ovs_key_icmpv6 *icmpv6_key;
1703
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);
1710 }
1711
1712 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1713 const struct ovs_key_nd *nd_key;
1714
1715 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1716 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1717 nd_key->nd_target,
1718 sizeof(match->key->ipv6.nd.target),
1719 is_mask);
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);
1725 }
1726
1727 if (attrs != 0) {
1728 OVS_NLERR(log, "Unknown key attributes %llx",
1729 (unsigned long long)attrs);
1730 return -EINVAL;
1731 }
1732
1733 return 0;
1734}
1735
1736static void nlattr_set(struct nlattr *attr, u8 val,
1737 const struct ovs_len_tbl *tbl)
1738{
1739 struct nlattr *nla;
1740 int rem;
1741
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);
1746 else
1747 memset(nla_data(nla), val, nla_len(nla));
1748
1749 if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1750 *(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1751 }
1752}
1753
1754static void mask_set_nlattr(struct nlattr *attr, u8 val)
1755{
1756 nlattr_set(attr, val, ovs_key_lens);
1757}
1758
1759/**
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
1768 * of this flow.
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.
1774 */
1775int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1776 const struct nlattr *nla_key,
1777 const struct nlattr *nla_mask,
1778 bool log)
1779{
1780 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1781 struct nlattr *newmask = NULL;
1782 u64 key_attrs = 0;
1783 u64 mask_attrs = 0;
1784 int err;
1785
1786 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1787 if (err)
1788 return err;
1789
1790 err = parse_vlan_from_nlattrs(match, &key_attrs, a, false, log);
1791 if (err)
1792 return err;
1793
1794 err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1795 if (err)
1796 return err;
1797
1798 if (match->mask) {
1799 if (!nla_mask) {
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.
1809 */
1810 newmask = kmemdup(nla_key,
1811 nla_total_size(nla_len(nla_key)),
1812 GFP_KERNEL);
1813 if (!newmask)
1814 return -ENOMEM;
1815
1816 mask_set_nlattr(newmask, 0xff);
1817
1818 /* The userspace does not send tunnel attributes that
1819 * are 0, but we should not wildcard them nonetheless.
1820 */
1821 if (match->key->tun_proto)
1822 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1823 0xff, true);
1824
1825 nla_mask = newmask;
1826 }
1827
1828 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1829 if (err)
1830 goto free_newmask;
1831
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);
1835
1836 err = parse_vlan_from_nlattrs(match, &mask_attrs, a, true, log);
1837 if (err)
1838 goto free_newmask;
1839
1840 err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1841 log);
1842 if (err)
1843 goto free_newmask;
1844 }
1845
1846 if (!match_validate(match, key_attrs, mask_attrs, log))
1847 err = -EINVAL;
1848
1849free_newmask:
1850 kfree(newmask);
1851 return err;
1852}
1853
1854static size_t get_ufid_len(const struct nlattr *attr, bool log)
1855{
1856 size_t len;
1857
1858 if (!attr)
1859 return 0;
1860
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);
1865 return 0;
1866 }
1867
1868 return len;
1869}
1870
1871/* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1872 * or false otherwise.
1873 */
1874bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1875 bool log)
1876{
1877 sfid->ufid_len = get_ufid_len(attr, log);
1878 if (sfid->ufid_len)
1879 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1880
1881 return sfid->ufid_len;
1882}
1883
1884int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1885 const struct sw_flow_key *key, bool log)
1886{
1887 struct sw_flow_key *new_key;
1888
1889 if (ovs_nla_get_ufid(sfid, ufid, log))
1890 return 0;
1891
1892 /* If UFID was not provided, use unmasked key. */
1893 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1894 if (!new_key)
1895 return -ENOMEM;
1896 memcpy(new_key, key, sizeof(*key));
1897 sfid->unmasked_key = new_key;
1898
1899 return 0;
1900}
1901
1902u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1903{
1904 return attr ? nla_get_u32(attr) : 0;
1905}
1906
1907/**
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
1911 * metadata.
1912 * @a: Array of netlink attributes holding parsed %OVS_KEY_ATTR_* Netlink
1913 * attributes.
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.
1918 *
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.
1923 *
1924 * This must be called before the packet key fields are filled in 'key'.
1925 */
1926
1927int 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)
1930{
1931 struct sw_flow_match match;
1932
1933 memset(&match, 0, sizeof(match));
1934 match.key = key;
1935
1936 key->ct_state = 0;
1937 key->ct_zone = 0;
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));
1942
1943 key->phy.in_port = DP_MAX_PORTS;
1944
1945 return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1946}
1947
1948static int ovs_nla_put_vlan(struct sk_buff *skb, const struct vlan_head *vh,
1949 bool is_mask)
1950{
1951 __be16 eth_type = !is_mask ? vh->tpid : htons(0xffff);
1952
1953 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1954 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, vh->tci))
1955 return -EMSGSIZE;
1956 return 0;
1957}
1958
1959static int nsh_key_to_nlattr(const struct ovs_key_nsh *nsh, bool is_mask,
1960 struct sk_buff *skb)
1961{
1962 struct nlattr *start;
1963
1964 start = nla_nest_start_noflag(skb, OVS_KEY_ATTR_NSH);
1965 if (!start)
1966 return -EMSGSIZE;
1967
1968 if (nla_put(skb, OVS_NSH_KEY_ATTR_BASE, sizeof(nsh->base), &nsh->base))
1969 goto nla_put_failure;
1970
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;
1975 }
1976
1977 /* Don't support MD type 2 yet */
1978
1979 nla_nest_end(skb, start);
1980
1981 return 0;
1982
1983nla_put_failure:
1984 return -EMSGSIZE;
1985}
1986
1987static 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)
1990{
1991 struct ovs_key_ethernet *eth_key;
1992 struct nlattr *nla;
1993 struct nlattr *encap = NULL;
1994 struct nlattr *in_encap = NULL;
1995
1996 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1997 goto nla_put_failure;
1998
1999 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
2000 goto nla_put_failure;
2001
2002 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
2003 goto nla_put_failure;
2004
2005 if ((swkey->tun_proto || is_mask)) {
2006 const void *opts = NULL;
2007
2008 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
2009 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
2010
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;
2014 }
2015
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;
2020 } else {
2021 u16 upper_u16;
2022 upper_u16 = !is_mask ? 0 : 0xffff;
2023
2024 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
2025 (upper_u16 << 16) | output->phy.in_port))
2026 goto nla_put_failure;
2027 }
2028
2029 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
2030 goto nla_put_failure;
2031
2032 if (ovs_ct_put_key(swkey, output, skb))
2033 goto nla_put_failure;
2034
2035 if (ovs_key_mac_proto(swkey) == MAC_PROTO_ETHERNET) {
2036 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
2037 if (!nla)
2038 goto nla_put_failure;
2039
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);
2043
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)
2049 goto unencap;
2050
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)
2057 goto unencap;
2058 }
2059 }
2060
2061 if (swkey->eth.type == htons(ETH_P_802_2)) {
2062 /*
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
2066 * be wildcarded.
2067 */
2068 if (is_mask && output->eth.type)
2069 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
2070 output->eth.type))
2071 goto nla_put_failure;
2072 goto unencap;
2073 }
2074 }
2075
2076 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
2077 goto nla_put_failure;
2078
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.
2082 */
2083 WARN_ON_ONCE(!(encap && in_encap));
2084 goto unencap;
2085 }
2086
2087 if (swkey->eth.type == htons(ETH_P_IP)) {
2088 struct ovs_key_ipv4 *ipv4_key;
2089
2090 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
2091 if (!nla)
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;
2102
2103 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
2104 if (!nla)
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;
2122
2123 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
2124 if (!nla)
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)) {
2134 u8 i, num_labels;
2135 struct ovs_key_mpls *mpls_key;
2136
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));
2140 if (!nla)
2141 goto nla_put_failure;
2142
2143 mpls_key = nla_data(nla);
2144 for (i = 0; i < num_labels; i++)
2145 mpls_key[i].mpls_lse = output->mpls.lse[i];
2146 }
2147
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) {
2151
2152 if (swkey->ip.proto == IPPROTO_TCP) {
2153 struct ovs_key_tcp *tcp_key;
2154
2155 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
2156 if (!nla)
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,
2162 output->tp.flags))
2163 goto nla_put_failure;
2164 } else if (swkey->ip.proto == IPPROTO_UDP) {
2165 struct ovs_key_udp *udp_key;
2166
2167 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
2168 if (!nla)
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;
2175
2176 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
2177 if (!nla)
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;
2185
2186 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
2187 if (!nla)
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;
2195
2196 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
2197 sizeof(*icmpv6_key));
2198 if (!nla)
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);
2203
2204 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
2205 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
2206 struct ovs_key_nd *nd_key;
2207
2208 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
2209 if (!nla)
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);
2216 }
2217 }
2218 }
2219
2220unencap:
2221 if (in_encap)
2222 nla_nest_end(skb, in_encap);
2223 if (encap)
2224 nla_nest_end(skb, encap);
2225
2226 return 0;
2227
2228nla_put_failure:
2229 return -EMSGSIZE;
2230}
2231
2232int 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)
2235{
2236 int err;
2237 struct nlattr *nla;
2238
2239 nla = nla_nest_start_noflag(skb, attr);
2240 if (!nla)
2241 return -EMSGSIZE;
2242 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
2243 if (err)
2244 return err;
2245 nla_nest_end(skb, nla);
2246
2247 return 0;
2248}
2249
2250/* Called with ovs_mutex or RCU read lock. */
2251int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
2252{
2253 if (ovs_identifier_is_ufid(&flow->id))
2254 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
2255 flow->id.ufid);
2256
2257 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
2258 OVS_FLOW_ATTR_KEY, false, skb);
2259}
2260
2261/* Called with ovs_mutex or RCU read lock. */
2262int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
2263{
2264 return ovs_nla_put_key(&flow->key, &flow->key,
2265 OVS_FLOW_ATTR_KEY, false, skb);
2266}
2267
2268/* Called with ovs_mutex or RCU read lock. */
2269int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
2270{
2271 return ovs_nla_put_key(&flow->key, &flow->mask->key,
2272 OVS_FLOW_ATTR_MASK, true, skb);
2273}
2274
2275#define MAX_ACTIONS_BUFSIZE (32 * 1024)
2276
2277static struct sw_flow_actions *nla_alloc_flow_actions(int size)
2278{
2279 struct sw_flow_actions *sfa;
2280
2281 WARN_ON_ONCE(size > MAX_ACTIONS_BUFSIZE);
2282
2283 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
2284 if (!sfa)
2285 return ERR_PTR(-ENOMEM);
2286
2287 sfa->actions_len = 0;
2288 return sfa;
2289}
2290
2291static void ovs_nla_free_set_action(const struct nlattr *a)
2292{
2293 const struct nlattr *ovs_key = nla_data(a);
2294 struct ovs_tunnel_info *ovs_tun;
2295
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);
2300 break;
2301 }
2302}
2303
2304void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
2305{
2306 const struct nlattr *a;
2307 int rem;
2308
2309 if (!sf_acts)
2310 return;
2311
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);
2316 break;
2317 case OVS_ACTION_ATTR_CT:
2318 ovs_ct_free_action(a);
2319 break;
2320 }
2321 }
2322
2323 kfree(sf_acts);
2324}
2325
2326static void __ovs_nla_free_flow_actions(struct rcu_head *head)
2327{
2328 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
2329}
2330
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. */
2333void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
2334{
2335 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
2336}
2337
2338static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
2339 int attr_len, bool log)
2340{
2341
2342 struct sw_flow_actions *acts;
2343 int new_acts_size;
2344 size_t req_size = NLA_ALIGN(attr_len);
2345 int next_offset = offsetof(struct sw_flow_actions, actions) +
2346 (*sfa)->actions_len;
2347
2348 if (req_size <= (ksize(*sfa) - next_offset))
2349 goto out;
2350
2351 new_acts_size = max(next_offset + req_size, ksize(*sfa) * 2);
2352
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);
2358 }
2359 new_acts_size = MAX_ACTIONS_BUFSIZE;
2360 }
2361
2362 acts = nla_alloc_flow_actions(new_acts_size);
2363 if (IS_ERR(acts))
2364 return (void *)acts;
2365
2366 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
2367 acts->actions_len = (*sfa)->actions_len;
2368 acts->orig_len = (*sfa)->orig_len;
2369 kfree(*sfa);
2370 *sfa = acts;
2371
2372out:
2373 (*sfa)->actions_len += req_size;
2374 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
2375}
2376
2377static struct nlattr *__add_action(struct sw_flow_actions **sfa,
2378 int attrtype, void *data, int len, bool log)
2379{
2380 struct nlattr *a;
2381
2382 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
2383 if (IS_ERR(a))
2384 return a;
2385
2386 a->nla_type = attrtype;
2387 a->nla_len = nla_attr_size(len);
2388
2389 if (data)
2390 memcpy(nla_data(a), data, len);
2391 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
2392
2393 return a;
2394}
2395
2396int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
2397 int len, bool log)
2398{
2399 struct nlattr *a;
2400
2401 a = __add_action(sfa, attrtype, data, len, log);
2402
2403 return PTR_ERR_OR_ZERO(a);
2404}
2405
2406static inline int add_nested_action_start(struct sw_flow_actions **sfa,
2407 int attrtype, bool log)
2408{
2409 int used = (*sfa)->actions_len;
2410 int err;
2411
2412 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
2413 if (err)
2414 return err;
2415
2416 return used;
2417}
2418
2419static inline void add_nested_action_end(struct sw_flow_actions *sfa,
2420 int st_offset)
2421{
2422 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
2423 st_offset);
2424
2425 a->nla_len = sfa->actions_len - st_offset;
2426}
2427
2428static 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);
2433
2434static 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)
2439{
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;
2445
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])
2450 return -EINVAL;
2451 attrs[type] = a;
2452 }
2453 if (rem)
2454 return -EINVAL;
2455
2456 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
2457 if (!probability || nla_len(probability) != sizeof(u32))
2458 return -EINVAL;
2459
2460 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
2461 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
2462 return -EINVAL;
2463
2464 /* validation done, copy sample action. */
2465 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
2466 if (start < 0)
2467 return start;
2468
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.
2472 *
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.
2476 *
2477 * If the sample is the last action, it can always be excuted
2478 * rather than deferred.
2479 */
2480 arg.exec = last || !actions_may_change_flow(actions);
2481 arg.probability = nla_get_u32(probability);
2482
2483 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_ARG, &arg, sizeof(arg),
2484 log);
2485 if (err)
2486 return err;
2487
2488 err = __ovs_nla_copy_actions(net, actions, key, sfa,
2489 eth_type, vlan_tci, mpls_label_count, log);
2490
2491 if (err)
2492 return err;
2493
2494 add_nested_action_end(*sfa, start);
2495
2496 return 0;
2497}
2498
2499static 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)
2505{
2506 const struct nlattr *attrs[OVS_DEC_TTL_ATTR_MAX + 1];
2507 int start, action_start, err, rem;
2508 const struct nlattr *a, *actions;
2509
2510 memset(attrs, 0, sizeof(attrs));
2511 nla_for_each_nested(a, attr, rem) {
2512 int type = nla_type(a);
2513
2514 /* Ignore unknown attributes to be future proof. */
2515 if (type > OVS_DEC_TTL_ATTR_MAX)
2516 continue;
2517
2518 if (!type || attrs[type]) {
2519 OVS_NLERR(log, "Duplicate or invalid key (type %d).",
2520 type);
2521 return -EINVAL;
2522 }
2523
2524 attrs[type] = a;
2525 }
2526
2527 if (rem) {
2528 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
2529 return -EINVAL;
2530 }
2531
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.");
2535 return -EINVAL;
2536 }
2537
2538 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_DEC_TTL, log);
2539 if (start < 0)
2540 return start;
2541
2542 action_start = add_nested_action_start(sfa, OVS_DEC_TTL_ATTR_ACTION, log);
2543 if (action_start < 0)
2544 return action_start;
2545
2546 err = __ovs_nla_copy_actions(net, actions, key, sfa, eth_type,
2547 vlan_tci, mpls_label_count, log);
2548 if (err)
2549 return err;
2550
2551 add_nested_action_end(*sfa, action_start);
2552 add_nested_action_end(*sfa, start);
2553 return 0;
2554}
2555
2556static 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)
2562{
2563 int start, err;
2564 u32 exec;
2565
2566 if (nla_len(attr) && nla_len(attr) < NLA_HDRLEN)
2567 return -EINVAL;
2568
2569 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CLONE, log);
2570 if (start < 0)
2571 return start;
2572
2573 exec = last || !actions_may_change_flow(attr);
2574
2575 err = ovs_nla_add_action(sfa, OVS_CLONE_ATTR_EXEC, &exec,
2576 sizeof(exec), log);
2577 if (err)
2578 return err;
2579
2580 err = __ovs_nla_copy_actions(net, attr, key, sfa,
2581 eth_type, vlan_tci, mpls_label_count, log);
2582 if (err)
2583 return err;
2584
2585 add_nested_action_end(*sfa, start);
2586
2587 return 0;
2588}
2589
2590void ovs_match_init(struct sw_flow_match *match,
2591 struct sw_flow_key *key,
2592 bool reset_key,
2593 struct sw_flow_mask *mask)
2594{
2595 memset(match, 0, sizeof(*match));
2596 match->key = key;
2597 match->mask = mask;
2598
2599 if (reset_key)
2600 memset(key, 0, sizeof(*key));
2601
2602 if (mask) {
2603 memset(&mask->key, 0, sizeof(mask->key));
2604 mask->range.start = mask->range.end = 0;
2605 }
2606}
2607
2608static int validate_geneve_opts(struct sw_flow_key *key)
2609{
2610 struct geneve_opt *option;
2611 int opts_len = key->tun_opts_len;
2612 bool crit_opt = false;
2613
2614 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2615 while (opts_len > 0) {
2616 int len;
2617
2618 if (opts_len < sizeof(*option))
2619 return -EINVAL;
2620
2621 len = sizeof(*option) + option->length * 4;
2622 if (len > opts_len)
2623 return -EINVAL;
2624
2625 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2626
2627 option = (struct geneve_opt *)((u8 *)option + len);
2628 opts_len -= len;
2629 }
2630
2631 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
2632
2633 return 0;
2634}
2635
2636static int validate_and_copy_set_tun(const struct nlattr *attr,
2637 struct sw_flow_actions **sfa, bool log)
2638{
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;
2644 struct nlattr *a;
2645 int err = 0, start, opts_type;
2646 __be16 dst_opt_type;
2647
2648 dst_opt_type = 0;
2649 ovs_match_init(&match, &key, true, NULL);
2650 opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2651 if (opts_type < 0)
2652 return opts_type;
2653
2654 if (key.tun_opts_len) {
2655 switch (opts_type) {
2656 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2657 err = validate_geneve_opts(&key);
2658 if (err < 0)
2659 return err;
2660 dst_opt_type = TUNNEL_GENEVE_OPT;
2661 break;
2662 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
2663 dst_opt_type = TUNNEL_VXLAN_OPT;
2664 break;
2665 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
2666 dst_opt_type = TUNNEL_ERSPAN_OPT;
2667 break;
2668 }
2669 }
2670
2671 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2672 if (start < 0)
2673 return start;
2674
2675 tun_dst = metadata_dst_alloc(key.tun_opts_len, METADATA_IP_TUNNEL,
2676 GFP_KERNEL);
2677
2678 if (!tun_dst)
2679 return -ENOMEM;
2680
2681 err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2682 if (err) {
2683 dst_release((struct dst_entry *)tun_dst);
2684 return err;
2685 }
2686
2687 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2688 sizeof(*ovs_tun), log);
2689 if (IS_ERR(a)) {
2690 dst_release((struct dst_entry *)tun_dst);
2691 return PTR_ERR(a);
2692 }
2693
2694 ovs_tun = nla_data(a);
2695 ovs_tun->tun_dst = tun_dst;
2696
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;
2704
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.
2708 */
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);
2713
2714 return err;
2715}
2716
2717static bool validate_nsh(const struct nlattr *attr, bool is_mask,
2718 bool is_push_nsh, bool log)
2719{
2720 struct sw_flow_match match;
2721 struct sw_flow_key key;
2722 int ret = 0;
2723
2724 ovs_match_init(&match, &key, true, NULL);
2725 ret = nsh_key_put_from_nlattr(attr, &match, is_mask,
2726 is_push_nsh, log);
2727 return !ret;
2728}
2729
2730/* Return false if there are any non-masked bits set.
2731 * Mask follows data immediately, before any netlink padding.
2732 */
2733static bool validate_masked(u8 *data, int len)
2734{
2735 u8 *mask = data + len;
2736
2737 while (len--)
2738 if (*data++ & ~*mask++)
2739 return false;
2740
2741 return true;
2742}
2743
2744static 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)
2748{
2749 const struct nlattr *ovs_key = nla_data(a);
2750 int key_type = nla_type(ovs_key);
2751 size_t key_len;
2752
2753 /* There can be only one key in a action */
2754 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2755 return -EINVAL;
2756
2757 key_len = nla_len(ovs_key);
2758 if (masked)
2759 key_len /= 2;
2760
2761 if (key_type > OVS_KEY_ATTR_MAX ||
2762 !check_attr_len(key_len, ovs_key_lens[key_type].len))
2763 return -EINVAL;
2764
2765 if (masked && !validate_masked(nla_data(ovs_key), key_len))
2766 return -EINVAL;
2767
2768 switch (key_type) {
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:
2773 break;
2774
2775 case OVS_KEY_ATTR_ETHERNET:
2776 if (mac_proto != MAC_PROTO_ETHERNET)
2777 return -EINVAL;
2778 break;
2779
2780 case OVS_KEY_ATTR_TUNNEL: {
2781 int err;
2782
2783 if (masked)
2784 return -EINVAL; /* Masked tunnel set not supported. */
2785
2786 *skip_copy = true;
2787 err = validate_and_copy_set_tun(a, sfa, log);
2788 if (err)
2789 return err;
2790 break;
2791 }
2792 case OVS_KEY_ATTR_IPV4: {
2793 const struct ovs_key_ipv4 *ipv4_key;
2794
2795 if (eth_type != htons(ETH_P_IP))
2796 return -EINVAL;
2797
2798 ipv4_key = nla_data(ovs_key);
2799
2800 if (masked) {
2801 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2802
2803 /* Non-writeable fields. */
2804 if (mask->ipv4_proto || mask->ipv4_frag)
2805 return -EINVAL;
2806 } else {
2807 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2808 return -EINVAL;
2809
2810 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2811 return -EINVAL;
2812 }
2813 break;
2814 }
2815 case OVS_KEY_ATTR_IPV6: {
2816 const struct ovs_key_ipv6 *ipv6_key;
2817
2818 if (eth_type != htons(ETH_P_IPV6))
2819 return -EINVAL;
2820
2821 ipv6_key = nla_data(ovs_key);
2822
2823 if (masked) {
2824 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2825
2826 /* Non-writeable fields. */
2827 if (mask->ipv6_proto || mask->ipv6_frag)
2828 return -EINVAL;
2829
2830 /* Invalid bits in the flow label mask? */
2831 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2832 return -EINVAL;
2833 } else {
2834 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2835 return -EINVAL;
2836
2837 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2838 return -EINVAL;
2839 }
2840 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2841 return -EINVAL;
2842
2843 break;
2844 }
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)
2849 return -EINVAL;
2850
2851 break;
2852
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)
2857 return -EINVAL;
2858
2859 break;
2860
2861 case OVS_KEY_ATTR_MPLS:
2862 if (!eth_p_mpls(eth_type))
2863 return -EINVAL;
2864 break;
2865
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)
2870 return -EINVAL;
2871
2872 break;
2873
2874 case OVS_KEY_ATTR_NSH:
2875 if (eth_type != htons(ETH_P_NSH))
2876 return -EINVAL;
2877 if (!validate_nsh(nla_data(a), masked, false, log))
2878 return -EINVAL;
2879 break;
2880
2881 default:
2882 return -EINVAL;
2883 }
2884
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;
2888 struct nlattr *at;
2889
2890 *skip_copy = true;
2891
2892 start = add_nested_action_start(sfa,
2893 OVS_ACTION_ATTR_SET_TO_MASKED,
2894 log);
2895 if (start < 0)
2896 return start;
2897
2898 at = __add_action(sfa, key_type, NULL, len, log);
2899 if (IS_ERR(at))
2900 return PTR_ERR(at);
2901
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;
2907
2908 mask->ipv6_label &= htonl(0x000FFFFF);
2909 }
2910 add_nested_action_end(*sfa, start);
2911 }
2912
2913 return 0;
2914}
2915
2916static int validate_userspace(const struct nlattr *attr)
2917{
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 },
2922 };
2923 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2924 int error;
2925
2926 error = nla_parse_nested_deprecated(a, OVS_USERSPACE_ATTR_MAX, attr,
2927 userspace_policy, NULL);
2928 if (error)
2929 return error;
2930
2931 if (!a[OVS_USERSPACE_ATTR_PID] ||
2932 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2933 return -EINVAL;
2934
2935 return 0;
2936}
2937
2938static 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 },
2942};
2943
2944static 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)
2951{
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;
2956 int start, err;
2957
2958 err = nla_parse_deprecated_strict(a, OVS_CHECK_PKT_LEN_ATTR_MAX,
2959 nla_data(attr), nla_len(attr),
2960 cpl_policy, NULL);
2961 if (err)
2962 return err;
2963
2964 if (!a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] ||
2965 !nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]))
2966 return -EINVAL;
2967
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];
2970
2971 /* Both the nested action should be present. */
2972 if (!acts_if_greater || !acts_if_lesser_eq)
2973 return -EINVAL;
2974
2975 /* validation done, copy the nested actions. */
2976 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CHECK_PKT_LEN,
2977 log);
2978 if (start < 0)
2979 return start;
2980
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);
2986
2987 err = ovs_nla_add_action(sfa, OVS_CHECK_PKT_LEN_ATTR_ARG, &arg,
2988 sizeof(arg), log);
2989 if (err)
2990 return err;
2991
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;
2996
2997 err = __ovs_nla_copy_actions(net, acts_if_lesser_eq, key, sfa,
2998 eth_type, vlan_tci, mpls_label_count, log);
2999
3000 if (err)
3001 return err;
3002
3003 add_nested_action_end(*sfa, nested_acts_start);
3004
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;
3009
3010 err = __ovs_nla_copy_actions(net, acts_if_greater, key, sfa,
3011 eth_type, vlan_tci, mpls_label_count, log);
3012
3013 if (err)
3014 return err;
3015
3016 add_nested_action_end(*sfa, nested_acts_start);
3017 add_nested_action_end(*sfa, start);
3018 return 0;
3019}
3020
3021static int copy_action(const struct nlattr *from,
3022 struct sw_flow_actions **sfa, bool log)
3023{
3024 int totlen = NLA_ALIGN(from->nla_len);
3025 struct nlattr *to;
3026
3027 to = reserve_sfa_size(sfa, from->nla_len, log);
3028 if (IS_ERR(to))
3029 return PTR_ERR(to);
3030
3031 memcpy(to, from, totlen);
3032 return 0;
3033}
3034
3035static 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)
3040{
3041 u8 mac_proto = ovs_key_mac_proto(key);
3042 const struct nlattr *a;
3043 int rem, err;
3044
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,
3071 };
3072 const struct ovs_action_push_vlan *vlan;
3073 int type = nla_type(a);
3074 bool skip_copy;
3075
3076 if (type > OVS_ACTION_ATTR_MAX ||
3077 (action_lens[type] != nla_len(a) &&
3078 action_lens[type] != (u32)-1))
3079 return -EINVAL;
3080
3081 skip_copy = false;
3082 switch (type) {
3083 case OVS_ACTION_ATTR_UNSPEC:
3084 return -EINVAL;
3085
3086 case OVS_ACTION_ATTR_USERSPACE:
3087 err = validate_userspace(a);
3088 if (err)
3089 return err;
3090 break;
3091
3092 case OVS_ACTION_ATTR_OUTPUT:
3093 if (nla_get_u32(a) >= DP_MAX_PORTS)
3094 return -EINVAL;
3095 break;
3096
3097 case OVS_ACTION_ATTR_TRUNC: {
3098 const struct ovs_action_trunc *trunc = nla_data(a);
3099
3100 if (trunc->max_len < ETH_HLEN)
3101 return -EINVAL;
3102 break;
3103 }
3104
3105 case OVS_ACTION_ATTR_HASH: {
3106 const struct ovs_action_hash *act_hash = nla_data(a);
3107
3108 switch (act_hash->hash_alg) {
3109 case OVS_HASH_ALG_L4:
3110 break;
3111 default:
3112 return -EINVAL;
3113 }
3114
3115 break;
3116 }
3117
3118 case OVS_ACTION_ATTR_POP_VLAN:
3119 if (mac_proto != MAC_PROTO_ETHERNET)
3120 return -EINVAL;
3121 vlan_tci = htons(0);
3122 break;
3123
3124 case OVS_ACTION_ATTR_PUSH_VLAN:
3125 if (mac_proto != MAC_PROTO_ETHERNET)
3126 return -EINVAL;
3127 vlan = nla_data(a);
3128 if (!eth_type_vlan(vlan->vlan_tpid))
3129 return -EINVAL;
3130 if (!(vlan->vlan_tci & htons(VLAN_CFI_MASK)))
3131 return -EINVAL;
3132 vlan_tci = vlan->vlan_tci;
3133 break;
3134
3135 case OVS_ACTION_ATTR_RECIRC:
3136 break;
3137
3138 case OVS_ACTION_ATTR_ADD_MPLS: {
3139 const struct ovs_action_add_mpls *mpls = nla_data(a);
3140
3141 if (!eth_p_mpls(mpls->mpls_ethertype))
3142 return -EINVAL;
3143
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)))
3151 return -EINVAL;
3152 mpls_label_count++;
3153 } else {
3154 if (mac_proto == MAC_PROTO_ETHERNET) {
3155 mpls_label_count = 1;
3156 mac_proto = MAC_PROTO_NONE;
3157 } else {
3158 mpls_label_count++;
3159 }
3160 }
3161 eth_type = mpls->mpls_ethertype;
3162 break;
3163 }
3164
3165 case OVS_ACTION_ATTR_PUSH_MPLS: {
3166 const struct ovs_action_push_mpls *mpls = nla_data(a);
3167
3168 if (!eth_p_mpls(mpls->mpls_ethertype))
3169 return -EINVAL;
3170 /* Prohibit push MPLS other than to a white list
3171 * for packets that have a known tag order.
3172 */
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)))
3179 return -EINVAL;
3180 eth_type = mpls->mpls_ethertype;
3181 mpls_label_count++;
3182 break;
3183 }
3184
3185 case OVS_ACTION_ATTR_POP_MPLS: {
3186 __be16 proto;
3187 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3188 !eth_p_mpls(eth_type))
3189 return -EINVAL;
3190
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
3196 * it.
3197 *
3198 * Support for these actions is planned using packet
3199 * recirculation.
3200 */
3201 proto = nla_get_be16(a);
3202
3203 if (proto == htons(ETH_P_TEB) &&
3204 mac_proto != MAC_PROTO_NONE)
3205 return -EINVAL;
3206
3207 mpls_label_count--;
3208
3209 if (!eth_p_mpls(proto) || !mpls_label_count)
3210 eth_type = htons(0);
3211 else
3212 eth_type = proto;
3213
3214 break;
3215 }
3216
3217 case OVS_ACTION_ATTR_SET:
3218 err = validate_set(a, key, sfa,
3219 &skip_copy, mac_proto, eth_type,
3220 false, log);
3221 if (err)
3222 return err;
3223 break;
3224
3225 case OVS_ACTION_ATTR_SET_MASKED:
3226 err = validate_set(a, key, sfa,
3227 &skip_copy, mac_proto, eth_type,
3228 true, log);
3229 if (err)
3230 return err;
3231 break;
3232
3233 case OVS_ACTION_ATTR_SAMPLE: {
3234 bool last = nla_is_last(a, rem);
3235
3236 err = validate_and_copy_sample(net, a, key, sfa,
3237 eth_type, vlan_tci,
3238 mpls_label_count,
3239 log, last);
3240 if (err)
3241 return err;
3242 skip_copy = true;
3243 break;
3244 }
3245
3246 case OVS_ACTION_ATTR_CT:
3247 err = ovs_ct_copy_action(net, a, key, sfa, log);
3248 if (err)
3249 return err;
3250 skip_copy = true;
3251 break;
3252
3253 case OVS_ACTION_ATTR_CT_CLEAR:
3254 break;
3255
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)
3260 return -EINVAL;
3261 mac_proto = MAC_PROTO_ETHERNET;
3262 break;
3263
3264 case OVS_ACTION_ATTR_POP_ETH:
3265 if (mac_proto != MAC_PROTO_ETHERNET)
3266 return -EINVAL;
3267 if (vlan_tci & htons(VLAN_CFI_MASK))
3268 return -EINVAL;
3269 mac_proto = MAC_PROTO_NONE;
3270 break;
3271
3272 case OVS_ACTION_ATTR_PUSH_NSH:
3273 if (mac_proto != MAC_PROTO_ETHERNET) {
3274 u8 next_proto;
3275
3276 next_proto = tun_p_from_eth_p(eth_type);
3277 if (!next_proto)
3278 return -EINVAL;
3279 }
3280 mac_proto = MAC_PROTO_NONE;
3281 if (!validate_nsh(nla_data(a), false, true, true))
3282 return -EINVAL;
3283 break;
3284
3285 case OVS_ACTION_ATTR_POP_NSH: {
3286 __be16 inner_proto;
3287
3288 if (eth_type != htons(ETH_P_NSH))
3289 return -EINVAL;
3290 inner_proto = tun_p_to_eth_p(key->nsh.base.np);
3291 if (!inner_proto)
3292 return -EINVAL;
3293 if (key->nsh.base.np == TUN_P_ETHERNET)
3294 mac_proto = MAC_PROTO_ETHERNET;
3295 else
3296 mac_proto = MAC_PROTO_NONE;
3297 break;
3298 }
3299
3300 case OVS_ACTION_ATTR_METER:
3301 /* Non-existent meters are simply ignored. */
3302 break;
3303
3304 case OVS_ACTION_ATTR_CLONE: {
3305 bool last = nla_is_last(a, rem);
3306
3307 err = validate_and_copy_clone(net, a, key, sfa,
3308 eth_type, vlan_tci,
3309 mpls_label_count,
3310 log, last);
3311 if (err)
3312 return err;
3313 skip_copy = true;
3314 break;
3315 }
3316
3317 case OVS_ACTION_ATTR_CHECK_PKT_LEN: {
3318 bool last = nla_is_last(a, rem);
3319
3320 err = validate_and_copy_check_pkt_len(net, a, key, sfa,
3321 eth_type,
3322 vlan_tci,
3323 mpls_label_count,
3324 log, last);
3325 if (err)
3326 return err;
3327 skip_copy = true;
3328 break;
3329 }
3330
3331 case OVS_ACTION_ATTR_DEC_TTL:
3332 err = validate_and_copy_dec_ttl(net, a, key, sfa,
3333 eth_type, vlan_tci,
3334 mpls_label_count, log);
3335 if (err)
3336 return err;
3337 skip_copy = true;
3338 break;
3339
3340 default:
3341 OVS_NLERR(log, "Unknown Action type %d", type);
3342 return -EINVAL;
3343 }
3344 if (!skip_copy) {
3345 err = copy_action(a, sfa, log);
3346 if (err)
3347 return err;
3348 }
3349 }
3350
3351 if (rem > 0)
3352 return -EINVAL;
3353
3354 return 0;
3355}
3356
3357/* 'key' must be the masked key. */
3358int 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)
3361{
3362 int err;
3363 u32 mpls_label_count = 0;
3364
3365 *sfa = nla_alloc_flow_actions(min(nla_len(attr), MAX_ACTIONS_BUFSIZE));
3366 if (IS_ERR(*sfa))
3367 return PTR_ERR(*sfa);
3368
3369 if (eth_p_mpls(key->eth.type))
3370 mpls_label_count = hweight_long(key->mpls.num_labels_mask);
3371
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);
3375 if (err)
3376 ovs_nla_free_flow_actions(*sfa);
3377
3378 return err;
3379}
3380
3381static int sample_action_to_attr(const struct nlattr *attr,
3382 struct sk_buff *skb)
3383{
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;
3388
3389 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SAMPLE);
3390 if (!start)
3391 return -EMSGSIZE;
3392
3393 sample_arg = nla_data(attr);
3394 arg = nla_data(sample_arg);
3395 actions = nla_next(sample_arg, &rem);
3396
3397 if (nla_put_u32(skb, OVS_SAMPLE_ATTR_PROBABILITY, arg->probability)) {
3398 err = -EMSGSIZE;
3399 goto out;
3400 }
3401
3402 ac_start = nla_nest_start_noflag(skb, OVS_SAMPLE_ATTR_ACTIONS);
3403 if (!ac_start) {
3404 err = -EMSGSIZE;
3405 goto out;
3406 }
3407
3408 err = ovs_nla_put_actions(actions, rem, skb);
3409
3410out:
3411 if (err) {
3412 nla_nest_cancel(skb, ac_start);
3413 nla_nest_cancel(skb, start);
3414 } else {
3415 nla_nest_end(skb, ac_start);
3416 nla_nest_end(skb, start);
3417 }
3418
3419 return err;
3420}
3421
3422static int clone_action_to_attr(const struct nlattr *attr,
3423 struct sk_buff *skb)
3424{
3425 struct nlattr *start;
3426 int err = 0, rem = nla_len(attr);
3427
3428 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CLONE);
3429 if (!start)
3430 return -EMSGSIZE;
3431
3432 err = ovs_nla_put_actions(nla_data(attr), rem, skb);
3433
3434 if (err)
3435 nla_nest_cancel(skb, start);
3436 else
3437 nla_nest_end(skb, start);
3438
3439 return err;
3440}
3441
3442static int check_pkt_len_action_to_attr(const struct nlattr *attr,
3443 struct sk_buff *skb)
3444{
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);
3449
3450 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CHECK_PKT_LEN);
3451 if (!start)
3452 return -EMSGSIZE;
3453
3454 /* The first nested attribute in 'attr' is always
3455 * 'OVS_CHECK_PKT_LEN_ATTR_ARG'.
3456 */
3457 cpl_arg = nla_data(attr);
3458 arg = nla_data(cpl_arg);
3459
3460 if (nla_put_u16(skb, OVS_CHECK_PKT_LEN_ATTR_PKT_LEN, arg->pkt_len)) {
3461 err = -EMSGSIZE;
3462 goto out;
3463 }
3464
3465 /* Second nested attribute in 'attr' is always
3466 * 'OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL'.
3467 */
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);
3471 if (!ac_start) {
3472 err = -EMSGSIZE;
3473 goto out;
3474 }
3475
3476 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3477 if (err) {
3478 nla_nest_cancel(skb, ac_start);
3479 goto out;
3480 } else {
3481 nla_nest_end(skb, ac_start);
3482 }
3483
3484 /* Third nested attribute in 'attr' is always
3485 * OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER.
3486 */
3487 a = nla_next(a, &rem);
3488 ac_start = nla_nest_start_noflag(skb,
3489 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER);
3490 if (!ac_start) {
3491 err = -EMSGSIZE;
3492 goto out;
3493 }
3494
3495 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3496 if (err) {
3497 nla_nest_cancel(skb, ac_start);
3498 goto out;
3499 } else {
3500 nla_nest_end(skb, ac_start);
3501 }
3502
3503 nla_nest_end(skb, start);
3504 return 0;
3505
3506out:
3507 nla_nest_cancel(skb, start);
3508 return err;
3509}
3510
3511static int dec_ttl_action_to_attr(const struct nlattr *attr,
3512 struct sk_buff *skb)
3513{
3514 struct nlattr *start, *action_start;
3515 const struct nlattr *a;
3516 int err = 0, rem;
3517
3518 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_DEC_TTL);
3519 if (!start)
3520 return -EMSGSIZE;
3521
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:
3525
3526 action_start = nla_nest_start_noflag(skb, OVS_DEC_TTL_ATTR_ACTION);
3527 if (!action_start) {
3528 err = -EMSGSIZE;
3529 goto out;
3530 }
3531
3532 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3533 if (err)
3534 goto out;
3535
3536 nla_nest_end(skb, action_start);
3537 break;
3538
3539 default:
3540 /* Ignore all other option to be future compatible */
3541 break;
3542 }
3543 }
3544
3545 nla_nest_end(skb, start);
3546 return 0;
3547
3548out:
3549 nla_nest_cancel(skb, start);
3550 return err;
3551}
3552
3553static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
3554{
3555 const struct nlattr *ovs_key = nla_data(a);
3556 int key_type = nla_type(ovs_key);
3557 struct nlattr *start;
3558 int err;
3559
3560 switch (key_type) {
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;
3564
3565 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3566 if (!start)
3567 return -EMSGSIZE;
3568
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);
3573 if (err)
3574 return err;
3575 nla_nest_end(skb, start);
3576 break;
3577 }
3578 default:
3579 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
3580 return -EMSGSIZE;
3581 break;
3582 }
3583
3584 return 0;
3585}
3586
3587static int masked_set_action_to_set_action_attr(const struct nlattr *a,
3588 struct sk_buff *skb)
3589{
3590 const struct nlattr *ovs_key = nla_data(a);
3591 struct nlattr *nla;
3592 size_t key_len = nla_len(ovs_key) / 2;
3593
3594 /* Revert the conversion we did from a non-masked set action to
3595 * masked set action.
3596 */
3597 nla = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3598 if (!nla)
3599 return -EMSGSIZE;
3600
3601 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
3602 return -EMSGSIZE;
3603
3604 nla_nest_end(skb, nla);
3605 return 0;
3606}
3607
3608int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
3609{
3610 const struct nlattr *a;
3611 int rem, err;
3612
3613 nla_for_each_attr(a, attr, len, rem) {
3614 int type = nla_type(a);
3615
3616 switch (type) {
3617 case OVS_ACTION_ATTR_SET:
3618 err = set_action_to_attr(a, skb);
3619 if (err)
3620 return err;
3621 break;
3622
3623 case OVS_ACTION_ATTR_SET_TO_MASKED:
3624 err = masked_set_action_to_set_action_attr(a, skb);
3625 if (err)
3626 return err;
3627 break;
3628
3629 case OVS_ACTION_ATTR_SAMPLE:
3630 err = sample_action_to_attr(a, skb);
3631 if (err)
3632 return err;
3633 break;
3634
3635 case OVS_ACTION_ATTR_CT:
3636 err = ovs_ct_action_to_attr(nla_data(a), skb);
3637 if (err)
3638 return err;
3639 break;
3640
3641 case OVS_ACTION_ATTR_CLONE:
3642 err = clone_action_to_attr(a, skb);
3643 if (err)
3644 return err;
3645 break;
3646
3647 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
3648 err = check_pkt_len_action_to_attr(a, skb);
3649 if (err)
3650 return err;
3651 break;
3652
3653 case OVS_ACTION_ATTR_DEC_TTL:
3654 err = dec_ttl_action_to_attr(a, skb);
3655 if (err)
3656 return err;
3657 break;
3658
3659 default:
3660 if (nla_put(skb, type, nla_len(a), nla_data(a)))
3661 return -EMSGSIZE;
3662 break;
3663 }
3664 }
3665
3666 return 0;
3667}
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * Copyright (c) 2007-2017 Nicira, Inc.
4 */
5
6#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
7
8#include "flow.h"
9#include "datapath.h"
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>
21#include <linux/in.h>
22#include <linux/rcupdate.h>
23#include <linux/if_arp.h>
24#include <linux/ip.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>
33#include <net/ip.h>
34#include <net/ipv6.h>
35#include <net/ndisc.h>
36#include <net/mpls.h>
37#include <net/vxlan.h>
38#include <net/tun_proto.h>
39#include <net/erspan.h>
40
41#include "drop.h"
42#include "flow_netlink.h"
43
44struct ovs_len_tbl {
45 int len;
46 const struct ovs_len_tbl *next;
47};
48
49#define OVS_ATTR_NESTED -1
50#define OVS_ATTR_VARIABLE -2
51#define OVS_COPY_ACTIONS_MAX_DEPTH 16
52
53static bool actions_may_change_flow(const struct nlattr *actions)
54{
55 struct nlattr *nla;
56 int rem;
57
58 nla_for_each_nested(nla, actions, rem) {
59 u16 action = nla_type(nla);
60
61 switch (action) {
62 case OVS_ACTION_ATTR_OUTPUT:
63 case OVS_ACTION_ATTR_RECIRC:
64 case OVS_ACTION_ATTR_TRUNC:
65 case OVS_ACTION_ATTR_USERSPACE:
66 case OVS_ACTION_ATTR_DROP:
67 break;
68
69 case OVS_ACTION_ATTR_CT:
70 case OVS_ACTION_ATTR_CT_CLEAR:
71 case OVS_ACTION_ATTR_HASH:
72 case OVS_ACTION_ATTR_POP_ETH:
73 case OVS_ACTION_ATTR_POP_MPLS:
74 case OVS_ACTION_ATTR_POP_NSH:
75 case OVS_ACTION_ATTR_POP_VLAN:
76 case OVS_ACTION_ATTR_PUSH_ETH:
77 case OVS_ACTION_ATTR_PUSH_MPLS:
78 case OVS_ACTION_ATTR_PUSH_NSH:
79 case OVS_ACTION_ATTR_PUSH_VLAN:
80 case OVS_ACTION_ATTR_SAMPLE:
81 case OVS_ACTION_ATTR_SET:
82 case OVS_ACTION_ATTR_SET_MASKED:
83 case OVS_ACTION_ATTR_METER:
84 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
85 case OVS_ACTION_ATTR_ADD_MPLS:
86 case OVS_ACTION_ATTR_DEC_TTL:
87 default:
88 return true;
89 }
90 }
91 return false;
92}
93
94static void update_range(struct sw_flow_match *match,
95 size_t offset, size_t size, bool is_mask)
96{
97 struct sw_flow_key_range *range;
98 size_t start = rounddown(offset, sizeof(long));
99 size_t end = roundup(offset + size, sizeof(long));
100
101 if (!is_mask)
102 range = &match->range;
103 else
104 range = &match->mask->range;
105
106 if (range->start == range->end) {
107 range->start = start;
108 range->end = end;
109 return;
110 }
111
112 if (range->start > start)
113 range->start = start;
114
115 if (range->end < end)
116 range->end = end;
117}
118
119#define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
120 do { \
121 update_range(match, offsetof(struct sw_flow_key, field), \
122 sizeof((match)->key->field), is_mask); \
123 if (is_mask) \
124 (match)->mask->key.field = value; \
125 else \
126 (match)->key->field = value; \
127 } while (0)
128
129#define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
130 do { \
131 update_range(match, offset, len, is_mask); \
132 if (is_mask) \
133 memcpy((u8 *)&(match)->mask->key + offset, value_p, \
134 len); \
135 else \
136 memcpy((u8 *)(match)->key + offset, value_p, len); \
137 } while (0)
138
139#define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
140 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
141 value_p, len, is_mask)
142
143#define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
144 do { \
145 update_range(match, offsetof(struct sw_flow_key, field), \
146 sizeof((match)->key->field), is_mask); \
147 if (is_mask) \
148 memset((u8 *)&(match)->mask->key.field, value, \
149 sizeof((match)->mask->key.field)); \
150 else \
151 memset((u8 *)&(match)->key->field, value, \
152 sizeof((match)->key->field)); \
153 } while (0)
154
155static bool match_validate(const struct sw_flow_match *match,
156 u64 key_attrs, u64 mask_attrs, bool log)
157{
158 u64 key_expected = 0;
159 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
160
161 /* The following mask attributes allowed only if they
162 * pass the validation tests. */
163 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
164 | (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)
165 | (1 << OVS_KEY_ATTR_IPV6)
166 | (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)
167 | (1 << OVS_KEY_ATTR_TCP)
168 | (1 << OVS_KEY_ATTR_TCP_FLAGS)
169 | (1 << OVS_KEY_ATTR_UDP)
170 | (1 << OVS_KEY_ATTR_SCTP)
171 | (1 << OVS_KEY_ATTR_ICMP)
172 | (1 << OVS_KEY_ATTR_ICMPV6)
173 | (1 << OVS_KEY_ATTR_ARP)
174 | (1 << OVS_KEY_ATTR_ND)
175 | (1 << OVS_KEY_ATTR_MPLS)
176 | (1 << OVS_KEY_ATTR_NSH));
177
178 /* Always allowed mask fields. */
179 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
180 | (1 << OVS_KEY_ATTR_IN_PORT)
181 | (1 << OVS_KEY_ATTR_ETHERTYPE));
182
183 /* Check key attributes. */
184 if (match->key->eth.type == htons(ETH_P_ARP)
185 || match->key->eth.type == htons(ETH_P_RARP)) {
186 key_expected |= 1 << OVS_KEY_ATTR_ARP;
187 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
188 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
189 }
190
191 if (eth_p_mpls(match->key->eth.type)) {
192 key_expected |= 1 << OVS_KEY_ATTR_MPLS;
193 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
194 mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
195 }
196
197 if (match->key->eth.type == htons(ETH_P_IP)) {
198 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
199 if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
200 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
201 mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4;
202 }
203
204 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
205 if (match->key->ip.proto == IPPROTO_UDP) {
206 key_expected |= 1 << OVS_KEY_ATTR_UDP;
207 if (match->mask && (match->mask->key.ip.proto == 0xff))
208 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
209 }
210
211 if (match->key->ip.proto == IPPROTO_SCTP) {
212 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
213 if (match->mask && (match->mask->key.ip.proto == 0xff))
214 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
215 }
216
217 if (match->key->ip.proto == IPPROTO_TCP) {
218 key_expected |= 1 << OVS_KEY_ATTR_TCP;
219 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
220 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
221 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
222 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
223 }
224 }
225
226 if (match->key->ip.proto == IPPROTO_ICMP) {
227 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
228 if (match->mask && (match->mask->key.ip.proto == 0xff))
229 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
230 }
231 }
232 }
233
234 if (match->key->eth.type == htons(ETH_P_IPV6)) {
235 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
236 if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
237 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
238 mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6;
239 }
240
241 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
242 if (match->key->ip.proto == IPPROTO_UDP) {
243 key_expected |= 1 << OVS_KEY_ATTR_UDP;
244 if (match->mask && (match->mask->key.ip.proto == 0xff))
245 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
246 }
247
248 if (match->key->ip.proto == IPPROTO_SCTP) {
249 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
250 if (match->mask && (match->mask->key.ip.proto == 0xff))
251 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
252 }
253
254 if (match->key->ip.proto == IPPROTO_TCP) {
255 key_expected |= 1 << OVS_KEY_ATTR_TCP;
256 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
257 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
258 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
259 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
260 }
261 }
262
263 if (match->key->ip.proto == IPPROTO_ICMPV6) {
264 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
265 if (match->mask && (match->mask->key.ip.proto == 0xff))
266 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
267
268 if (match->key->tp.src ==
269 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
270 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
271 key_expected |= 1 << OVS_KEY_ATTR_ND;
272 /* Original direction conntrack tuple
273 * uses the same space as the ND fields
274 * in the key, so both are not allowed
275 * at the same time.
276 */
277 mask_allowed &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
278 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
279 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
280 }
281 }
282 }
283 }
284
285 if (match->key->eth.type == htons(ETH_P_NSH)) {
286 key_expected |= 1 << OVS_KEY_ATTR_NSH;
287 if (match->mask &&
288 match->mask->key.eth.type == htons(0xffff)) {
289 mask_allowed |= 1 << OVS_KEY_ATTR_NSH;
290 }
291 }
292
293 if ((key_attrs & key_expected) != key_expected) {
294 /* Key attributes check failed. */
295 OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
296 (unsigned long long)key_attrs,
297 (unsigned long long)key_expected);
298 return false;
299 }
300
301 if ((mask_attrs & mask_allowed) != mask_attrs) {
302 /* Mask attributes check failed. */
303 OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
304 (unsigned long long)mask_attrs,
305 (unsigned long long)mask_allowed);
306 return false;
307 }
308
309 return true;
310}
311
312size_t ovs_tun_key_attr_size(void)
313{
314 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
315 * updating this function.
316 */
317 return nla_total_size_64bit(8) /* OVS_TUNNEL_KEY_ATTR_ID */
318 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
319 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_DST */
320 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
321 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
322 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
323 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
324 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
325 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
326 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS and
327 * OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS is mutually exclusive with
328 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
329 */
330 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
331 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
332}
333
334static size_t ovs_nsh_key_attr_size(void)
335{
336 /* Whenever adding new OVS_NSH_KEY_ FIELDS, we should consider
337 * updating this function.
338 */
339 return nla_total_size(NSH_BASE_HDR_LEN) /* OVS_NSH_KEY_ATTR_BASE */
340 /* OVS_NSH_KEY_ATTR_MD1 and OVS_NSH_KEY_ATTR_MD2 are
341 * mutually exclusive, so the bigger one can cover
342 * the small one.
343 */
344 + nla_total_size(NSH_CTX_HDRS_MAX_LEN);
345}
346
347size_t ovs_key_attr_size(void)
348{
349 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
350 * updating this function.
351 */
352 BUILD_BUG_ON(OVS_KEY_ATTR_MAX != 32);
353
354 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
355 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
356 + ovs_tun_key_attr_size()
357 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
358 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
359 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
360 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
361 + nla_total_size(4) /* OVS_KEY_ATTR_CT_STATE */
362 + nla_total_size(2) /* OVS_KEY_ATTR_CT_ZONE */
363 + nla_total_size(4) /* OVS_KEY_ATTR_CT_MARK */
364 + nla_total_size(16) /* OVS_KEY_ATTR_CT_LABELS */
365 + nla_total_size(40) /* OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6 */
366 + nla_total_size(0) /* OVS_KEY_ATTR_NSH */
367 + ovs_nsh_key_attr_size()
368 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
369 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
370 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
371 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
372 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
373 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
374 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
375 + nla_total_size(28) /* OVS_KEY_ATTR_ND */
376 + nla_total_size(2); /* OVS_KEY_ATTR_IPV6_EXTHDRS */
377}
378
379static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
380 [OVS_VXLAN_EXT_GBP] = { .len = sizeof(u32) },
381};
382
383static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
384 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
385 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
386 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
387 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
388 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
389 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
390 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
391 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
392 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
393 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
394 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_VARIABLE },
395 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED,
396 .next = ovs_vxlan_ext_key_lens },
397 [OVS_TUNNEL_KEY_ATTR_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
398 [OVS_TUNNEL_KEY_ATTR_IPV6_DST] = { .len = sizeof(struct in6_addr) },
399 [OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS] = { .len = OVS_ATTR_VARIABLE },
400 [OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE] = { .len = 0 },
401};
402
403static const struct ovs_len_tbl
404ovs_nsh_key_attr_lens[OVS_NSH_KEY_ATTR_MAX + 1] = {
405 [OVS_NSH_KEY_ATTR_BASE] = { .len = sizeof(struct ovs_nsh_key_base) },
406 [OVS_NSH_KEY_ATTR_MD1] = { .len = sizeof(struct ovs_nsh_key_md1) },
407 [OVS_NSH_KEY_ATTR_MD2] = { .len = OVS_ATTR_VARIABLE },
408};
409
410/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
411static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
412 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
413 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
414 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
415 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
416 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
417 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
418 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
419 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
420 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
421 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
422 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
423 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
424 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
425 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
426 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
427 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
428 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
429 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
430 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
431 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
432 .next = ovs_tunnel_key_lens, },
433 [OVS_KEY_ATTR_MPLS] = { .len = OVS_ATTR_VARIABLE },
434 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u32) },
435 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
436 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
437 [OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
438 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4] = {
439 .len = sizeof(struct ovs_key_ct_tuple_ipv4) },
440 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6] = {
441 .len = sizeof(struct ovs_key_ct_tuple_ipv6) },
442 [OVS_KEY_ATTR_NSH] = { .len = OVS_ATTR_NESTED,
443 .next = ovs_nsh_key_attr_lens, },
444 [OVS_KEY_ATTR_IPV6_EXTHDRS] = {
445 .len = sizeof(struct ovs_key_ipv6_exthdrs) },
446};
447
448static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
449{
450 return expected_len == attr_len ||
451 expected_len == OVS_ATTR_NESTED ||
452 expected_len == OVS_ATTR_VARIABLE;
453}
454
455static bool is_all_zero(const u8 *fp, size_t size)
456{
457 int i;
458
459 if (!fp)
460 return false;
461
462 for (i = 0; i < size; i++)
463 if (fp[i])
464 return false;
465
466 return true;
467}
468
469static int __parse_flow_nlattrs(const struct nlattr *attr,
470 const struct nlattr *a[],
471 u64 *attrsp, bool log, bool nz)
472{
473 const struct nlattr *nla;
474 u64 attrs;
475 int rem;
476
477 attrs = *attrsp;
478 nla_for_each_nested(nla, attr, rem) {
479 u16 type = nla_type(nla);
480 int expected_len;
481
482 if (type > OVS_KEY_ATTR_MAX) {
483 OVS_NLERR(log, "Key type %d is out of range max %d",
484 type, OVS_KEY_ATTR_MAX);
485 return -EINVAL;
486 }
487
488 if (type == OVS_KEY_ATTR_PACKET_TYPE ||
489 type == OVS_KEY_ATTR_ND_EXTENSIONS ||
490 type == OVS_KEY_ATTR_TUNNEL_INFO) {
491 OVS_NLERR(log, "Key type %d is not supported", type);
492 return -EINVAL;
493 }
494
495 if (attrs & (1ULL << type)) {
496 OVS_NLERR(log, "Duplicate key (type %d).", type);
497 return -EINVAL;
498 }
499
500 expected_len = ovs_key_lens[type].len;
501 if (!check_attr_len(nla_len(nla), expected_len)) {
502 OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
503 type, nla_len(nla), expected_len);
504 return -EINVAL;
505 }
506
507 if (!nz || !is_all_zero(nla_data(nla), nla_len(nla))) {
508 attrs |= 1ULL << type;
509 a[type] = nla;
510 }
511 }
512 if (rem) {
513 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
514 return -EINVAL;
515 }
516
517 *attrsp = attrs;
518 return 0;
519}
520
521static int parse_flow_mask_nlattrs(const struct nlattr *attr,
522 const struct nlattr *a[], u64 *attrsp,
523 bool log)
524{
525 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
526}
527
528int parse_flow_nlattrs(const struct nlattr *attr, const struct nlattr *a[],
529 u64 *attrsp, bool log)
530{
531 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
532}
533
534static int genev_tun_opt_from_nlattr(const struct nlattr *a,
535 struct sw_flow_match *match, bool is_mask,
536 bool log)
537{
538 unsigned long opt_key_offset;
539
540 if (nla_len(a) > sizeof(match->key->tun_opts)) {
541 OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
542 nla_len(a), sizeof(match->key->tun_opts));
543 return -EINVAL;
544 }
545
546 if (nla_len(a) % 4 != 0) {
547 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
548 nla_len(a));
549 return -EINVAL;
550 }
551
552 /* We need to record the length of the options passed
553 * down, otherwise packets with the same format but
554 * additional options will be silently matched.
555 */
556 if (!is_mask) {
557 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
558 false);
559 } else {
560 /* This is somewhat unusual because it looks at
561 * both the key and mask while parsing the
562 * attributes (and by extension assumes the key
563 * is parsed first). Normally, we would verify
564 * that each is the correct length and that the
565 * attributes line up in the validate function.
566 * However, that is difficult because this is
567 * variable length and we won't have the
568 * information later.
569 */
570 if (match->key->tun_opts_len != nla_len(a)) {
571 OVS_NLERR(log, "Geneve option len %d != mask len %d",
572 match->key->tun_opts_len, nla_len(a));
573 return -EINVAL;
574 }
575
576 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
577 }
578
579 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
580 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
581 nla_len(a), is_mask);
582 return 0;
583}
584
585static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
586 struct sw_flow_match *match, bool is_mask,
587 bool log)
588{
589 struct nlattr *a;
590 int rem;
591 unsigned long opt_key_offset;
592 struct vxlan_metadata opts;
593
594 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
595
596 memset(&opts, 0, sizeof(opts));
597 nla_for_each_nested(a, attr, rem) {
598 int type = nla_type(a);
599
600 if (type > OVS_VXLAN_EXT_MAX) {
601 OVS_NLERR(log, "VXLAN extension %d out of range max %d",
602 type, OVS_VXLAN_EXT_MAX);
603 return -EINVAL;
604 }
605
606 if (!check_attr_len(nla_len(a),
607 ovs_vxlan_ext_key_lens[type].len)) {
608 OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
609 type, nla_len(a),
610 ovs_vxlan_ext_key_lens[type].len);
611 return -EINVAL;
612 }
613
614 switch (type) {
615 case OVS_VXLAN_EXT_GBP:
616 opts.gbp = nla_get_u32(a);
617 break;
618 default:
619 OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
620 type);
621 return -EINVAL;
622 }
623 }
624 if (rem) {
625 OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
626 rem);
627 return -EINVAL;
628 }
629
630 if (!is_mask)
631 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
632 else
633 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
634
635 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
636 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
637 is_mask);
638 return 0;
639}
640
641static int erspan_tun_opt_from_nlattr(const struct nlattr *a,
642 struct sw_flow_match *match, bool is_mask,
643 bool log)
644{
645 unsigned long opt_key_offset;
646
647 BUILD_BUG_ON(sizeof(struct erspan_metadata) >
648 sizeof(match->key->tun_opts));
649
650 if (nla_len(a) > sizeof(match->key->tun_opts)) {
651 OVS_NLERR(log, "ERSPAN option length err (len %d, max %zu).",
652 nla_len(a), sizeof(match->key->tun_opts));
653 return -EINVAL;
654 }
655
656 if (!is_mask)
657 SW_FLOW_KEY_PUT(match, tun_opts_len,
658 sizeof(struct erspan_metadata), false);
659 else
660 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
661
662 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
663 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
664 nla_len(a), is_mask);
665 return 0;
666}
667
668static int ip_tun_from_nlattr(const struct nlattr *attr,
669 struct sw_flow_match *match, bool is_mask,
670 bool log)
671{
672 bool ttl = false, ipv4 = false, ipv6 = false;
673 bool info_bridge_mode = false;
674 __be16 tun_flags = 0;
675 int opts_type = 0;
676 struct nlattr *a;
677 int rem;
678
679 nla_for_each_nested(a, attr, rem) {
680 int type = nla_type(a);
681 int err;
682
683 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
684 OVS_NLERR(log, "Tunnel attr %d out of range max %d",
685 type, OVS_TUNNEL_KEY_ATTR_MAX);
686 return -EINVAL;
687 }
688
689 if (!check_attr_len(nla_len(a),
690 ovs_tunnel_key_lens[type].len)) {
691 OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
692 type, nla_len(a), ovs_tunnel_key_lens[type].len);
693 return -EINVAL;
694 }
695
696 switch (type) {
697 case OVS_TUNNEL_KEY_ATTR_ID:
698 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
699 nla_get_be64(a), is_mask);
700 tun_flags |= TUNNEL_KEY;
701 break;
702 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
703 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
704 nla_get_in_addr(a), is_mask);
705 ipv4 = true;
706 break;
707 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
708 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
709 nla_get_in_addr(a), is_mask);
710 ipv4 = true;
711 break;
712 case OVS_TUNNEL_KEY_ATTR_IPV6_SRC:
713 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.src,
714 nla_get_in6_addr(a), is_mask);
715 ipv6 = true;
716 break;
717 case OVS_TUNNEL_KEY_ATTR_IPV6_DST:
718 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
719 nla_get_in6_addr(a), is_mask);
720 ipv6 = true;
721 break;
722 case OVS_TUNNEL_KEY_ATTR_TOS:
723 SW_FLOW_KEY_PUT(match, tun_key.tos,
724 nla_get_u8(a), is_mask);
725 break;
726 case OVS_TUNNEL_KEY_ATTR_TTL:
727 SW_FLOW_KEY_PUT(match, tun_key.ttl,
728 nla_get_u8(a), is_mask);
729 ttl = true;
730 break;
731 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
732 tun_flags |= TUNNEL_DONT_FRAGMENT;
733 break;
734 case OVS_TUNNEL_KEY_ATTR_CSUM:
735 tun_flags |= TUNNEL_CSUM;
736 break;
737 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
738 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
739 nla_get_be16(a), is_mask);
740 break;
741 case OVS_TUNNEL_KEY_ATTR_TP_DST:
742 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
743 nla_get_be16(a), is_mask);
744 break;
745 case OVS_TUNNEL_KEY_ATTR_OAM:
746 tun_flags |= TUNNEL_OAM;
747 break;
748 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
749 if (opts_type) {
750 OVS_NLERR(log, "Multiple metadata blocks provided");
751 return -EINVAL;
752 }
753
754 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
755 if (err)
756 return err;
757
758 tun_flags |= TUNNEL_GENEVE_OPT;
759 opts_type = type;
760 break;
761 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
762 if (opts_type) {
763 OVS_NLERR(log, "Multiple metadata blocks provided");
764 return -EINVAL;
765 }
766
767 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
768 if (err)
769 return err;
770
771 tun_flags |= TUNNEL_VXLAN_OPT;
772 opts_type = type;
773 break;
774 case OVS_TUNNEL_KEY_ATTR_PAD:
775 break;
776 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
777 if (opts_type) {
778 OVS_NLERR(log, "Multiple metadata blocks provided");
779 return -EINVAL;
780 }
781
782 err = erspan_tun_opt_from_nlattr(a, match, is_mask,
783 log);
784 if (err)
785 return err;
786
787 tun_flags |= TUNNEL_ERSPAN_OPT;
788 opts_type = type;
789 break;
790 case OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE:
791 info_bridge_mode = true;
792 ipv4 = true;
793 break;
794 default:
795 OVS_NLERR(log, "Unknown IP tunnel attribute %d",
796 type);
797 return -EINVAL;
798 }
799 }
800
801 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
802 if (is_mask)
803 SW_FLOW_KEY_MEMSET_FIELD(match, tun_proto, 0xff, true);
804 else
805 SW_FLOW_KEY_PUT(match, tun_proto, ipv6 ? AF_INET6 : AF_INET,
806 false);
807
808 if (rem > 0) {
809 OVS_NLERR(log, "IP tunnel attribute has %d unknown bytes.",
810 rem);
811 return -EINVAL;
812 }
813
814 if (ipv4 && ipv6) {
815 OVS_NLERR(log, "Mixed IPv4 and IPv6 tunnel attributes");
816 return -EINVAL;
817 }
818
819 if (!is_mask) {
820 if (!ipv4 && !ipv6) {
821 OVS_NLERR(log, "IP tunnel dst address not specified");
822 return -EINVAL;
823 }
824 if (ipv4) {
825 if (info_bridge_mode) {
826 if (match->key->tun_key.u.ipv4.src ||
827 match->key->tun_key.u.ipv4.dst ||
828 match->key->tun_key.tp_src ||
829 match->key->tun_key.tp_dst ||
830 match->key->tun_key.ttl ||
831 match->key->tun_key.tos ||
832 tun_flags & ~TUNNEL_KEY) {
833 OVS_NLERR(log, "IPv4 tun info is not correct");
834 return -EINVAL;
835 }
836 } else if (!match->key->tun_key.u.ipv4.dst) {
837 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
838 return -EINVAL;
839 }
840 }
841 if (ipv6 && ipv6_addr_any(&match->key->tun_key.u.ipv6.dst)) {
842 OVS_NLERR(log, "IPv6 tunnel dst address is zero");
843 return -EINVAL;
844 }
845
846 if (!ttl && !info_bridge_mode) {
847 OVS_NLERR(log, "IP tunnel TTL not specified.");
848 return -EINVAL;
849 }
850 }
851
852 return opts_type;
853}
854
855static int vxlan_opt_to_nlattr(struct sk_buff *skb,
856 const void *tun_opts, int swkey_tun_opts_len)
857{
858 const struct vxlan_metadata *opts = tun_opts;
859 struct nlattr *nla;
860
861 nla = nla_nest_start_noflag(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
862 if (!nla)
863 return -EMSGSIZE;
864
865 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
866 return -EMSGSIZE;
867
868 nla_nest_end(skb, nla);
869 return 0;
870}
871
872static int __ip_tun_to_nlattr(struct sk_buff *skb,
873 const struct ip_tunnel_key *output,
874 const void *tun_opts, int swkey_tun_opts_len,
875 unsigned short tun_proto, u8 mode)
876{
877 if (output->tun_flags & TUNNEL_KEY &&
878 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id,
879 OVS_TUNNEL_KEY_ATTR_PAD))
880 return -EMSGSIZE;
881
882 if (mode & IP_TUNNEL_INFO_BRIDGE)
883 return nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE)
884 ? -EMSGSIZE : 0;
885
886 switch (tun_proto) {
887 case AF_INET:
888 if (output->u.ipv4.src &&
889 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
890 output->u.ipv4.src))
891 return -EMSGSIZE;
892 if (output->u.ipv4.dst &&
893 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
894 output->u.ipv4.dst))
895 return -EMSGSIZE;
896 break;
897 case AF_INET6:
898 if (!ipv6_addr_any(&output->u.ipv6.src) &&
899 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_SRC,
900 &output->u.ipv6.src))
901 return -EMSGSIZE;
902 if (!ipv6_addr_any(&output->u.ipv6.dst) &&
903 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_DST,
904 &output->u.ipv6.dst))
905 return -EMSGSIZE;
906 break;
907 }
908 if (output->tos &&
909 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
910 return -EMSGSIZE;
911 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
912 return -EMSGSIZE;
913 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
914 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
915 return -EMSGSIZE;
916 if ((output->tun_flags & TUNNEL_CSUM) &&
917 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
918 return -EMSGSIZE;
919 if (output->tp_src &&
920 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
921 return -EMSGSIZE;
922 if (output->tp_dst &&
923 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
924 return -EMSGSIZE;
925 if ((output->tun_flags & TUNNEL_OAM) &&
926 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
927 return -EMSGSIZE;
928 if (swkey_tun_opts_len) {
929 if (output->tun_flags & TUNNEL_GENEVE_OPT &&
930 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
931 swkey_tun_opts_len, tun_opts))
932 return -EMSGSIZE;
933 else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
934 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
935 return -EMSGSIZE;
936 else if (output->tun_flags & TUNNEL_ERSPAN_OPT &&
937 nla_put(skb, OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS,
938 swkey_tun_opts_len, tun_opts))
939 return -EMSGSIZE;
940 }
941
942 return 0;
943}
944
945static int ip_tun_to_nlattr(struct sk_buff *skb,
946 const struct ip_tunnel_key *output,
947 const void *tun_opts, int swkey_tun_opts_len,
948 unsigned short tun_proto, u8 mode)
949{
950 struct nlattr *nla;
951 int err;
952
953 nla = nla_nest_start_noflag(skb, OVS_KEY_ATTR_TUNNEL);
954 if (!nla)
955 return -EMSGSIZE;
956
957 err = __ip_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len,
958 tun_proto, mode);
959 if (err)
960 return err;
961
962 nla_nest_end(skb, nla);
963 return 0;
964}
965
966int ovs_nla_put_tunnel_info(struct sk_buff *skb,
967 struct ip_tunnel_info *tun_info)
968{
969 return __ip_tun_to_nlattr(skb, &tun_info->key,
970 ip_tunnel_info_opts(tun_info),
971 tun_info->options_len,
972 ip_tunnel_info_af(tun_info), tun_info->mode);
973}
974
975static int encode_vlan_from_nlattrs(struct sw_flow_match *match,
976 const struct nlattr *a[],
977 bool is_mask, bool inner)
978{
979 __be16 tci = 0;
980 __be16 tpid = 0;
981
982 if (a[OVS_KEY_ATTR_VLAN])
983 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
984
985 if (a[OVS_KEY_ATTR_ETHERTYPE])
986 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
987
988 if (likely(!inner)) {
989 SW_FLOW_KEY_PUT(match, eth.vlan.tpid, tpid, is_mask);
990 SW_FLOW_KEY_PUT(match, eth.vlan.tci, tci, is_mask);
991 } else {
992 SW_FLOW_KEY_PUT(match, eth.cvlan.tpid, tpid, is_mask);
993 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, tci, is_mask);
994 }
995 return 0;
996}
997
998static int validate_vlan_from_nlattrs(const struct sw_flow_match *match,
999 u64 key_attrs, bool inner,
1000 const struct nlattr **a, bool log)
1001{
1002 __be16 tci = 0;
1003
1004 if (!((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
1005 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
1006 eth_type_vlan(nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE])))) {
1007 /* Not a VLAN. */
1008 return 0;
1009 }
1010
1011 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
1012 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
1013 OVS_NLERR(log, "Invalid %s frame", (inner) ? "C-VLAN" : "VLAN");
1014 return -EINVAL;
1015 }
1016
1017 if (a[OVS_KEY_ATTR_VLAN])
1018 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1019
1020 if (!(tci & htons(VLAN_CFI_MASK))) {
1021 if (tci) {
1022 OVS_NLERR(log, "%s TCI does not have VLAN_CFI_MASK bit set.",
1023 (inner) ? "C-VLAN" : "VLAN");
1024 return -EINVAL;
1025 } else if (nla_len(a[OVS_KEY_ATTR_ENCAP])) {
1026 /* Corner case for truncated VLAN header. */
1027 OVS_NLERR(log, "Truncated %s header has non-zero encap attribute.",
1028 (inner) ? "C-VLAN" : "VLAN");
1029 return -EINVAL;
1030 }
1031 }
1032
1033 return 1;
1034}
1035
1036static int validate_vlan_mask_from_nlattrs(const struct sw_flow_match *match,
1037 u64 key_attrs, bool inner,
1038 const struct nlattr **a, bool log)
1039{
1040 __be16 tci = 0;
1041 __be16 tpid = 0;
1042 bool encap_valid = !!(match->key->eth.vlan.tci &
1043 htons(VLAN_CFI_MASK));
1044 bool i_encap_valid = !!(match->key->eth.cvlan.tci &
1045 htons(VLAN_CFI_MASK));
1046
1047 if (!(key_attrs & (1 << OVS_KEY_ATTR_ENCAP))) {
1048 /* Not a VLAN. */
1049 return 0;
1050 }
1051
1052 if ((!inner && !encap_valid) || (inner && !i_encap_valid)) {
1053 OVS_NLERR(log, "Encap mask attribute is set for non-%s frame.",
1054 (inner) ? "C-VLAN" : "VLAN");
1055 return -EINVAL;
1056 }
1057
1058 if (a[OVS_KEY_ATTR_VLAN])
1059 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1060
1061 if (a[OVS_KEY_ATTR_ETHERTYPE])
1062 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1063
1064 if (tpid != htons(0xffff)) {
1065 OVS_NLERR(log, "Must have an exact match on %s TPID (mask=%x).",
1066 (inner) ? "C-VLAN" : "VLAN", ntohs(tpid));
1067 return -EINVAL;
1068 }
1069 if (!(tci & htons(VLAN_CFI_MASK))) {
1070 OVS_NLERR(log, "%s TCI mask does not have exact match for VLAN_CFI_MASK bit.",
1071 (inner) ? "C-VLAN" : "VLAN");
1072 return -EINVAL;
1073 }
1074
1075 return 1;
1076}
1077
1078static int __parse_vlan_from_nlattrs(struct sw_flow_match *match,
1079 u64 *key_attrs, bool inner,
1080 const struct nlattr **a, bool is_mask,
1081 bool log)
1082{
1083 int err;
1084 const struct nlattr *encap;
1085
1086 if (!is_mask)
1087 err = validate_vlan_from_nlattrs(match, *key_attrs, inner,
1088 a, log);
1089 else
1090 err = validate_vlan_mask_from_nlattrs(match, *key_attrs, inner,
1091 a, log);
1092 if (err <= 0)
1093 return err;
1094
1095 err = encode_vlan_from_nlattrs(match, a, is_mask, inner);
1096 if (err)
1097 return err;
1098
1099 *key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1100 *key_attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
1101 *key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1102
1103 encap = a[OVS_KEY_ATTR_ENCAP];
1104
1105 if (!is_mask)
1106 err = parse_flow_nlattrs(encap, a, key_attrs, log);
1107 else
1108 err = parse_flow_mask_nlattrs(encap, a, key_attrs, log);
1109
1110 return err;
1111}
1112
1113static int parse_vlan_from_nlattrs(struct sw_flow_match *match,
1114 u64 *key_attrs, const struct nlattr **a,
1115 bool is_mask, bool log)
1116{
1117 int err;
1118 bool encap_valid = false;
1119
1120 err = __parse_vlan_from_nlattrs(match, key_attrs, false, a,
1121 is_mask, log);
1122 if (err)
1123 return err;
1124
1125 encap_valid = !!(match->key->eth.vlan.tci & htons(VLAN_CFI_MASK));
1126 if (encap_valid) {
1127 err = __parse_vlan_from_nlattrs(match, key_attrs, true, a,
1128 is_mask, log);
1129 if (err)
1130 return err;
1131 }
1132
1133 return 0;
1134}
1135
1136static int parse_eth_type_from_nlattrs(struct sw_flow_match *match,
1137 u64 *attrs, const struct nlattr **a,
1138 bool is_mask, bool log)
1139{
1140 __be16 eth_type;
1141
1142 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1143 if (is_mask) {
1144 /* Always exact match EtherType. */
1145 eth_type = htons(0xffff);
1146 } else if (!eth_proto_is_802_3(eth_type)) {
1147 OVS_NLERR(log, "EtherType %x is less than min %x",
1148 ntohs(eth_type), ETH_P_802_3_MIN);
1149 return -EINVAL;
1150 }
1151
1152 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
1153 *attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1154 return 0;
1155}
1156
1157static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
1158 u64 *attrs, const struct nlattr **a,
1159 bool is_mask, bool log)
1160{
1161 u8 mac_proto = MAC_PROTO_ETHERNET;
1162
1163 if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
1164 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
1165
1166 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
1167 *attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
1168 }
1169
1170 if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
1171 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
1172
1173 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
1174 *attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
1175 }
1176
1177 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
1178 SW_FLOW_KEY_PUT(match, phy.priority,
1179 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
1180 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
1181 }
1182
1183 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
1184 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1185
1186 if (is_mask) {
1187 in_port = 0xffffffff; /* Always exact match in_port. */
1188 } else if (in_port >= DP_MAX_PORTS) {
1189 OVS_NLERR(log, "Port %d exceeds max allowable %d",
1190 in_port, DP_MAX_PORTS);
1191 return -EINVAL;
1192 }
1193
1194 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
1195 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1196 } else if (!is_mask) {
1197 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
1198 }
1199
1200 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1201 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1202
1203 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1204 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1205 }
1206 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1207 if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1208 is_mask, log) < 0)
1209 return -EINVAL;
1210 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1211 }
1212
1213 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
1214 ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
1215 u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
1216
1217 if (ct_state & ~CT_SUPPORTED_MASK) {
1218 OVS_NLERR(log, "ct_state flags %08x unsupported",
1219 ct_state);
1220 return -EINVAL;
1221 }
1222
1223 SW_FLOW_KEY_PUT(match, ct_state, ct_state, is_mask);
1224 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
1225 }
1226 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
1227 ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
1228 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
1229
1230 SW_FLOW_KEY_PUT(match, ct_zone, ct_zone, is_mask);
1231 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
1232 }
1233 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
1234 ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
1235 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
1236
1237 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
1238 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
1239 }
1240 if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
1241 ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
1242 const struct ovs_key_ct_labels *cl;
1243
1244 cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
1245 SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
1246 sizeof(*cl), is_mask);
1247 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
1248 }
1249 if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)) {
1250 const struct ovs_key_ct_tuple_ipv4 *ct;
1251
1252 ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4]);
1253
1254 SW_FLOW_KEY_PUT(match, ipv4.ct_orig.src, ct->ipv4_src, is_mask);
1255 SW_FLOW_KEY_PUT(match, ipv4.ct_orig.dst, ct->ipv4_dst, is_mask);
1256 SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1257 SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1258 SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv4_proto, is_mask);
1259 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4);
1260 }
1261 if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)) {
1262 const struct ovs_key_ct_tuple_ipv6 *ct;
1263
1264 ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6]);
1265
1266 SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.src, &ct->ipv6_src,
1267 sizeof(match->key->ipv6.ct_orig.src),
1268 is_mask);
1269 SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.dst, &ct->ipv6_dst,
1270 sizeof(match->key->ipv6.ct_orig.dst),
1271 is_mask);
1272 SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1273 SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1274 SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv6_proto, is_mask);
1275 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
1276 }
1277
1278 /* For layer 3 packets the Ethernet type is provided
1279 * and treated as metadata but no MAC addresses are provided.
1280 */
1281 if (!(*attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
1282 (*attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)))
1283 mac_proto = MAC_PROTO_NONE;
1284
1285 /* Always exact match mac_proto */
1286 SW_FLOW_KEY_PUT(match, mac_proto, is_mask ? 0xff : mac_proto, is_mask);
1287
1288 if (mac_proto == MAC_PROTO_NONE)
1289 return parse_eth_type_from_nlattrs(match, attrs, a, is_mask,
1290 log);
1291
1292 return 0;
1293}
1294
1295int nsh_hdr_from_nlattr(const struct nlattr *attr,
1296 struct nshhdr *nh, size_t size)
1297{
1298 struct nlattr *a;
1299 int rem;
1300 u8 flags = 0;
1301 u8 ttl = 0;
1302 int mdlen = 0;
1303
1304 /* validate_nsh has check this, so we needn't do duplicate check here
1305 */
1306 if (size < NSH_BASE_HDR_LEN)
1307 return -ENOBUFS;
1308
1309 nla_for_each_nested(a, attr, rem) {
1310 int type = nla_type(a);
1311
1312 switch (type) {
1313 case OVS_NSH_KEY_ATTR_BASE: {
1314 const struct ovs_nsh_key_base *base = nla_data(a);
1315
1316 flags = base->flags;
1317 ttl = base->ttl;
1318 nh->np = base->np;
1319 nh->mdtype = base->mdtype;
1320 nh->path_hdr = base->path_hdr;
1321 break;
1322 }
1323 case OVS_NSH_KEY_ATTR_MD1:
1324 mdlen = nla_len(a);
1325 if (mdlen > size - NSH_BASE_HDR_LEN)
1326 return -ENOBUFS;
1327 memcpy(&nh->md1, nla_data(a), mdlen);
1328 break;
1329
1330 case OVS_NSH_KEY_ATTR_MD2:
1331 mdlen = nla_len(a);
1332 if (mdlen > size - NSH_BASE_HDR_LEN)
1333 return -ENOBUFS;
1334 memcpy(&nh->md2, nla_data(a), mdlen);
1335 break;
1336
1337 default:
1338 return -EINVAL;
1339 }
1340 }
1341
1342 /* nsh header length = NSH_BASE_HDR_LEN + mdlen */
1343 nh->ver_flags_ttl_len = 0;
1344 nsh_set_flags_ttl_len(nh, flags, ttl, NSH_BASE_HDR_LEN + mdlen);
1345
1346 return 0;
1347}
1348
1349int nsh_key_from_nlattr(const struct nlattr *attr,
1350 struct ovs_key_nsh *nsh, struct ovs_key_nsh *nsh_mask)
1351{
1352 struct nlattr *a;
1353 int rem;
1354
1355 /* validate_nsh has check this, so we needn't do duplicate check here
1356 */
1357 nla_for_each_nested(a, attr, rem) {
1358 int type = nla_type(a);
1359
1360 switch (type) {
1361 case OVS_NSH_KEY_ATTR_BASE: {
1362 const struct ovs_nsh_key_base *base = nla_data(a);
1363 const struct ovs_nsh_key_base *base_mask = base + 1;
1364
1365 nsh->base = *base;
1366 nsh_mask->base = *base_mask;
1367 break;
1368 }
1369 case OVS_NSH_KEY_ATTR_MD1: {
1370 const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1371 const struct ovs_nsh_key_md1 *md1_mask = md1 + 1;
1372
1373 memcpy(nsh->context, md1->context, sizeof(*md1));
1374 memcpy(nsh_mask->context, md1_mask->context,
1375 sizeof(*md1_mask));
1376 break;
1377 }
1378 case OVS_NSH_KEY_ATTR_MD2:
1379 /* Not supported yet */
1380 return -ENOTSUPP;
1381 default:
1382 return -EINVAL;
1383 }
1384 }
1385
1386 return 0;
1387}
1388
1389static int nsh_key_put_from_nlattr(const struct nlattr *attr,
1390 struct sw_flow_match *match, bool is_mask,
1391 bool is_push_nsh, bool log)
1392{
1393 struct nlattr *a;
1394 int rem;
1395 bool has_base = false;
1396 bool has_md1 = false;
1397 bool has_md2 = false;
1398 u8 mdtype = 0;
1399 int mdlen = 0;
1400
1401 if (WARN_ON(is_push_nsh && is_mask))
1402 return -EINVAL;
1403
1404 nla_for_each_nested(a, attr, rem) {
1405 int type = nla_type(a);
1406 int i;
1407
1408 if (type > OVS_NSH_KEY_ATTR_MAX) {
1409 OVS_NLERR(log, "nsh attr %d is out of range max %d",
1410 type, OVS_NSH_KEY_ATTR_MAX);
1411 return -EINVAL;
1412 }
1413
1414 if (!check_attr_len(nla_len(a),
1415 ovs_nsh_key_attr_lens[type].len)) {
1416 OVS_NLERR(
1417 log,
1418 "nsh attr %d has unexpected len %d expected %d",
1419 type,
1420 nla_len(a),
1421 ovs_nsh_key_attr_lens[type].len
1422 );
1423 return -EINVAL;
1424 }
1425
1426 switch (type) {
1427 case OVS_NSH_KEY_ATTR_BASE: {
1428 const struct ovs_nsh_key_base *base = nla_data(a);
1429
1430 has_base = true;
1431 mdtype = base->mdtype;
1432 SW_FLOW_KEY_PUT(match, nsh.base.flags,
1433 base->flags, is_mask);
1434 SW_FLOW_KEY_PUT(match, nsh.base.ttl,
1435 base->ttl, is_mask);
1436 SW_FLOW_KEY_PUT(match, nsh.base.mdtype,
1437 base->mdtype, is_mask);
1438 SW_FLOW_KEY_PUT(match, nsh.base.np,
1439 base->np, is_mask);
1440 SW_FLOW_KEY_PUT(match, nsh.base.path_hdr,
1441 base->path_hdr, is_mask);
1442 break;
1443 }
1444 case OVS_NSH_KEY_ATTR_MD1: {
1445 const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1446
1447 has_md1 = true;
1448 for (i = 0; i < NSH_MD1_CONTEXT_SIZE; i++)
1449 SW_FLOW_KEY_PUT(match, nsh.context[i],
1450 md1->context[i], is_mask);
1451 break;
1452 }
1453 case OVS_NSH_KEY_ATTR_MD2:
1454 if (!is_push_nsh) /* Not supported MD type 2 yet */
1455 return -ENOTSUPP;
1456
1457 has_md2 = true;
1458 mdlen = nla_len(a);
1459 if (mdlen > NSH_CTX_HDRS_MAX_LEN || mdlen <= 0) {
1460 OVS_NLERR(
1461 log,
1462 "Invalid MD length %d for MD type %d",
1463 mdlen,
1464 mdtype
1465 );
1466 return -EINVAL;
1467 }
1468 break;
1469 default:
1470 OVS_NLERR(log, "Unknown nsh attribute %d",
1471 type);
1472 return -EINVAL;
1473 }
1474 }
1475
1476 if (rem > 0) {
1477 OVS_NLERR(log, "nsh attribute has %d unknown bytes.", rem);
1478 return -EINVAL;
1479 }
1480
1481 if (has_md1 && has_md2) {
1482 OVS_NLERR(
1483 1,
1484 "invalid nsh attribute: md1 and md2 are exclusive."
1485 );
1486 return -EINVAL;
1487 }
1488
1489 if (!is_mask) {
1490 if ((has_md1 && mdtype != NSH_M_TYPE1) ||
1491 (has_md2 && mdtype != NSH_M_TYPE2)) {
1492 OVS_NLERR(1, "nsh attribute has unmatched MD type %d.",
1493 mdtype);
1494 return -EINVAL;
1495 }
1496
1497 if (is_push_nsh &&
1498 (!has_base || (!has_md1 && !has_md2))) {
1499 OVS_NLERR(
1500 1,
1501 "push_nsh: missing base or metadata attributes"
1502 );
1503 return -EINVAL;
1504 }
1505 }
1506
1507 return 0;
1508}
1509
1510static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
1511 u64 attrs, const struct nlattr **a,
1512 bool is_mask, bool log)
1513{
1514 int err;
1515
1516 err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
1517 if (err)
1518 return err;
1519
1520 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1521 const struct ovs_key_ethernet *eth_key;
1522
1523 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1524 SW_FLOW_KEY_MEMCPY(match, eth.src,
1525 eth_key->eth_src, ETH_ALEN, is_mask);
1526 SW_FLOW_KEY_MEMCPY(match, eth.dst,
1527 eth_key->eth_dst, ETH_ALEN, is_mask);
1528 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1529
1530 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1531 /* VLAN attribute is always parsed before getting here since it
1532 * may occur multiple times.
1533 */
1534 OVS_NLERR(log, "VLAN attribute unexpected.");
1535 return -EINVAL;
1536 }
1537
1538 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1539 err = parse_eth_type_from_nlattrs(match, &attrs, a, is_mask,
1540 log);
1541 if (err)
1542 return err;
1543 } else if (!is_mask) {
1544 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1545 }
1546 } else if (!match->key->eth.type) {
1547 OVS_NLERR(log, "Either Ethernet header or EtherType is required.");
1548 return -EINVAL;
1549 }
1550
1551 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1552 const struct ovs_key_ipv4 *ipv4_key;
1553
1554 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1555 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
1556 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
1557 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
1558 return -EINVAL;
1559 }
1560 SW_FLOW_KEY_PUT(match, ip.proto,
1561 ipv4_key->ipv4_proto, is_mask);
1562 SW_FLOW_KEY_PUT(match, ip.tos,
1563 ipv4_key->ipv4_tos, is_mask);
1564 SW_FLOW_KEY_PUT(match, ip.ttl,
1565 ipv4_key->ipv4_ttl, is_mask);
1566 SW_FLOW_KEY_PUT(match, ip.frag,
1567 ipv4_key->ipv4_frag, is_mask);
1568 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1569 ipv4_key->ipv4_src, is_mask);
1570 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1571 ipv4_key->ipv4_dst, is_mask);
1572 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1573 }
1574
1575 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1576 const struct ovs_key_ipv6 *ipv6_key;
1577
1578 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1579 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
1580 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
1581 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
1582 return -EINVAL;
1583 }
1584
1585 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
1586 OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x)",
1587 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
1588 return -EINVAL;
1589 }
1590
1591 SW_FLOW_KEY_PUT(match, ipv6.label,
1592 ipv6_key->ipv6_label, is_mask);
1593 SW_FLOW_KEY_PUT(match, ip.proto,
1594 ipv6_key->ipv6_proto, is_mask);
1595 SW_FLOW_KEY_PUT(match, ip.tos,
1596 ipv6_key->ipv6_tclass, is_mask);
1597 SW_FLOW_KEY_PUT(match, ip.ttl,
1598 ipv6_key->ipv6_hlimit, is_mask);
1599 SW_FLOW_KEY_PUT(match, ip.frag,
1600 ipv6_key->ipv6_frag, is_mask);
1601 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1602 ipv6_key->ipv6_src,
1603 sizeof(match->key->ipv6.addr.src),
1604 is_mask);
1605 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1606 ipv6_key->ipv6_dst,
1607 sizeof(match->key->ipv6.addr.dst),
1608 is_mask);
1609
1610 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1611 }
1612
1613 if (attrs & (1ULL << OVS_KEY_ATTR_IPV6_EXTHDRS)) {
1614 const struct ovs_key_ipv6_exthdrs *ipv6_exthdrs_key;
1615
1616 ipv6_exthdrs_key = nla_data(a[OVS_KEY_ATTR_IPV6_EXTHDRS]);
1617
1618 SW_FLOW_KEY_PUT(match, ipv6.exthdrs,
1619 ipv6_exthdrs_key->hdrs, is_mask);
1620
1621 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV6_EXTHDRS);
1622 }
1623
1624 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1625 const struct ovs_key_arp *arp_key;
1626
1627 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1628 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1629 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
1630 arp_key->arp_op);
1631 return -EINVAL;
1632 }
1633
1634 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1635 arp_key->arp_sip, is_mask);
1636 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1637 arp_key->arp_tip, is_mask);
1638 SW_FLOW_KEY_PUT(match, ip.proto,
1639 ntohs(arp_key->arp_op), is_mask);
1640 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1641 arp_key->arp_sha, ETH_ALEN, is_mask);
1642 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1643 arp_key->arp_tha, ETH_ALEN, is_mask);
1644
1645 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1646 }
1647
1648 if (attrs & (1 << OVS_KEY_ATTR_NSH)) {
1649 if (nsh_key_put_from_nlattr(a[OVS_KEY_ATTR_NSH], match,
1650 is_mask, false, log) < 0)
1651 return -EINVAL;
1652 attrs &= ~(1 << OVS_KEY_ATTR_NSH);
1653 }
1654
1655 if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1656 const struct ovs_key_mpls *mpls_key;
1657 u32 hdr_len;
1658 u32 label_count, label_count_mask, i;
1659
1660 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1661 hdr_len = nla_len(a[OVS_KEY_ATTR_MPLS]);
1662 label_count = hdr_len / sizeof(struct ovs_key_mpls);
1663
1664 if (label_count == 0 || label_count > MPLS_LABEL_DEPTH ||
1665 hdr_len % sizeof(struct ovs_key_mpls))
1666 return -EINVAL;
1667
1668 label_count_mask = GENMASK(label_count - 1, 0);
1669
1670 for (i = 0 ; i < label_count; i++)
1671 SW_FLOW_KEY_PUT(match, mpls.lse[i],
1672 mpls_key[i].mpls_lse, is_mask);
1673
1674 SW_FLOW_KEY_PUT(match, mpls.num_labels_mask,
1675 label_count_mask, is_mask);
1676
1677 attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1678 }
1679
1680 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1681 const struct ovs_key_tcp *tcp_key;
1682
1683 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1684 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1685 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1686 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1687 }
1688
1689 if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1690 SW_FLOW_KEY_PUT(match, tp.flags,
1691 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1692 is_mask);
1693 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1694 }
1695
1696 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1697 const struct ovs_key_udp *udp_key;
1698
1699 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1700 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1701 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1702 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1703 }
1704
1705 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1706 const struct ovs_key_sctp *sctp_key;
1707
1708 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1709 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1710 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1711 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1712 }
1713
1714 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1715 const struct ovs_key_icmp *icmp_key;
1716
1717 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1718 SW_FLOW_KEY_PUT(match, tp.src,
1719 htons(icmp_key->icmp_type), is_mask);
1720 SW_FLOW_KEY_PUT(match, tp.dst,
1721 htons(icmp_key->icmp_code), is_mask);
1722 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1723 }
1724
1725 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1726 const struct ovs_key_icmpv6 *icmpv6_key;
1727
1728 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1729 SW_FLOW_KEY_PUT(match, tp.src,
1730 htons(icmpv6_key->icmpv6_type), is_mask);
1731 SW_FLOW_KEY_PUT(match, tp.dst,
1732 htons(icmpv6_key->icmpv6_code), is_mask);
1733 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1734 }
1735
1736 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1737 const struct ovs_key_nd *nd_key;
1738
1739 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1740 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1741 nd_key->nd_target,
1742 sizeof(match->key->ipv6.nd.target),
1743 is_mask);
1744 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1745 nd_key->nd_sll, ETH_ALEN, is_mask);
1746 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1747 nd_key->nd_tll, ETH_ALEN, is_mask);
1748 attrs &= ~(1 << OVS_KEY_ATTR_ND);
1749 }
1750
1751 if (attrs != 0) {
1752 OVS_NLERR(log, "Unknown key attributes %llx",
1753 (unsigned long long)attrs);
1754 return -EINVAL;
1755 }
1756
1757 return 0;
1758}
1759
1760static void nlattr_set(struct nlattr *attr, u8 val,
1761 const struct ovs_len_tbl *tbl)
1762{
1763 struct nlattr *nla;
1764 int rem;
1765
1766 /* The nlattr stream should already have been validated */
1767 nla_for_each_nested(nla, attr, rem) {
1768 if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED)
1769 nlattr_set(nla, val, tbl[nla_type(nla)].next ? : tbl);
1770 else
1771 memset(nla_data(nla), val, nla_len(nla));
1772
1773 if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1774 *(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1775 }
1776}
1777
1778static void mask_set_nlattr(struct nlattr *attr, u8 val)
1779{
1780 nlattr_set(attr, val, ovs_key_lens);
1781}
1782
1783/**
1784 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1785 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1786 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1787 * does not include any don't care bit.
1788 * @net: Used to determine per-namespace field support.
1789 * @match: receives the extracted flow match information.
1790 * @nla_key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1791 * sequence. The fields should of the packet that triggered the creation
1792 * of this flow.
1793 * @nla_mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_*
1794 * Netlink attribute specifies the mask field of the wildcarded flow.
1795 * @log: Boolean to allow kernel error logging. Normally true, but when
1796 * probing for feature compatibility this should be passed in as false to
1797 * suppress unnecessary error logging.
1798 */
1799int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1800 const struct nlattr *nla_key,
1801 const struct nlattr *nla_mask,
1802 bool log)
1803{
1804 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1805 struct nlattr *newmask = NULL;
1806 u64 key_attrs = 0;
1807 u64 mask_attrs = 0;
1808 int err;
1809
1810 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1811 if (err)
1812 return err;
1813
1814 err = parse_vlan_from_nlattrs(match, &key_attrs, a, false, log);
1815 if (err)
1816 return err;
1817
1818 err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1819 if (err)
1820 return err;
1821
1822 if (match->mask) {
1823 if (!nla_mask) {
1824 /* Create an exact match mask. We need to set to 0xff
1825 * all the 'match->mask' fields that have been touched
1826 * in 'match->key'. We cannot simply memset
1827 * 'match->mask', because padding bytes and fields not
1828 * specified in 'match->key' should be left to 0.
1829 * Instead, we use a stream of netlink attributes,
1830 * copied from 'key' and set to 0xff.
1831 * ovs_key_from_nlattrs() will take care of filling
1832 * 'match->mask' appropriately.
1833 */
1834 newmask = kmemdup(nla_key,
1835 nla_total_size(nla_len(nla_key)),
1836 GFP_KERNEL);
1837 if (!newmask)
1838 return -ENOMEM;
1839
1840 mask_set_nlattr(newmask, 0xff);
1841
1842 /* The userspace does not send tunnel attributes that
1843 * are 0, but we should not wildcard them nonetheless.
1844 */
1845 if (match->key->tun_proto)
1846 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1847 0xff, true);
1848
1849 nla_mask = newmask;
1850 }
1851
1852 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1853 if (err)
1854 goto free_newmask;
1855
1856 /* Always match on tci. */
1857 SW_FLOW_KEY_PUT(match, eth.vlan.tci, htons(0xffff), true);
1858 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, htons(0xffff), true);
1859
1860 err = parse_vlan_from_nlattrs(match, &mask_attrs, a, true, log);
1861 if (err)
1862 goto free_newmask;
1863
1864 err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1865 log);
1866 if (err)
1867 goto free_newmask;
1868 }
1869
1870 if (!match_validate(match, key_attrs, mask_attrs, log))
1871 err = -EINVAL;
1872
1873free_newmask:
1874 kfree(newmask);
1875 return err;
1876}
1877
1878static size_t get_ufid_len(const struct nlattr *attr, bool log)
1879{
1880 size_t len;
1881
1882 if (!attr)
1883 return 0;
1884
1885 len = nla_len(attr);
1886 if (len < 1 || len > MAX_UFID_LENGTH) {
1887 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1888 nla_len(attr), MAX_UFID_LENGTH);
1889 return 0;
1890 }
1891
1892 return len;
1893}
1894
1895/* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1896 * or false otherwise.
1897 */
1898bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1899 bool log)
1900{
1901 sfid->ufid_len = get_ufid_len(attr, log);
1902 if (sfid->ufid_len)
1903 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1904
1905 return sfid->ufid_len;
1906}
1907
1908int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1909 const struct sw_flow_key *key, bool log)
1910{
1911 struct sw_flow_key *new_key;
1912
1913 if (ovs_nla_get_ufid(sfid, ufid, log))
1914 return 0;
1915
1916 /* If UFID was not provided, use unmasked key. */
1917 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1918 if (!new_key)
1919 return -ENOMEM;
1920 memcpy(new_key, key, sizeof(*key));
1921 sfid->unmasked_key = new_key;
1922
1923 return 0;
1924}
1925
1926u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1927{
1928 return attr ? nla_get_u32(attr) : 0;
1929}
1930
1931/**
1932 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1933 * @net: Network namespace.
1934 * @key: Receives extracted in_port, priority, tun_key, skb_mark and conntrack
1935 * metadata.
1936 * @a: Array of netlink attributes holding parsed %OVS_KEY_ATTR_* Netlink
1937 * attributes.
1938 * @attrs: Bit mask for the netlink attributes included in @a.
1939 * @log: Boolean to allow kernel error logging. Normally true, but when
1940 * probing for feature compatibility this should be passed in as false to
1941 * suppress unnecessary error logging.
1942 *
1943 * This parses a series of Netlink attributes that form a flow key, which must
1944 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1945 * get the metadata, that is, the parts of the flow key that cannot be
1946 * extracted from the packet itself.
1947 *
1948 * This must be called before the packet key fields are filled in 'key'.
1949 */
1950
1951int ovs_nla_get_flow_metadata(struct net *net,
1952 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1],
1953 u64 attrs, struct sw_flow_key *key, bool log)
1954{
1955 struct sw_flow_match match;
1956
1957 memset(&match, 0, sizeof(match));
1958 match.key = key;
1959
1960 key->ct_state = 0;
1961 key->ct_zone = 0;
1962 key->ct_orig_proto = 0;
1963 memset(&key->ct, 0, sizeof(key->ct));
1964 memset(&key->ipv4.ct_orig, 0, sizeof(key->ipv4.ct_orig));
1965 memset(&key->ipv6.ct_orig, 0, sizeof(key->ipv6.ct_orig));
1966
1967 key->phy.in_port = DP_MAX_PORTS;
1968
1969 return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1970}
1971
1972static int ovs_nla_put_vlan(struct sk_buff *skb, const struct vlan_head *vh,
1973 bool is_mask)
1974{
1975 __be16 eth_type = !is_mask ? vh->tpid : htons(0xffff);
1976
1977 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1978 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, vh->tci))
1979 return -EMSGSIZE;
1980 return 0;
1981}
1982
1983static int nsh_key_to_nlattr(const struct ovs_key_nsh *nsh, bool is_mask,
1984 struct sk_buff *skb)
1985{
1986 struct nlattr *start;
1987
1988 start = nla_nest_start_noflag(skb, OVS_KEY_ATTR_NSH);
1989 if (!start)
1990 return -EMSGSIZE;
1991
1992 if (nla_put(skb, OVS_NSH_KEY_ATTR_BASE, sizeof(nsh->base), &nsh->base))
1993 goto nla_put_failure;
1994
1995 if (is_mask || nsh->base.mdtype == NSH_M_TYPE1) {
1996 if (nla_put(skb, OVS_NSH_KEY_ATTR_MD1,
1997 sizeof(nsh->context), nsh->context))
1998 goto nla_put_failure;
1999 }
2000
2001 /* Don't support MD type 2 yet */
2002
2003 nla_nest_end(skb, start);
2004
2005 return 0;
2006
2007nla_put_failure:
2008 return -EMSGSIZE;
2009}
2010
2011static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
2012 const struct sw_flow_key *output, bool is_mask,
2013 struct sk_buff *skb)
2014{
2015 struct ovs_key_ethernet *eth_key;
2016 struct nlattr *nla;
2017 struct nlattr *encap = NULL;
2018 struct nlattr *in_encap = NULL;
2019
2020 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
2021 goto nla_put_failure;
2022
2023 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
2024 goto nla_put_failure;
2025
2026 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
2027 goto nla_put_failure;
2028
2029 if ((swkey->tun_proto || is_mask)) {
2030 const void *opts = NULL;
2031
2032 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
2033 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
2034
2035 if (ip_tun_to_nlattr(skb, &output->tun_key, opts,
2036 swkey->tun_opts_len, swkey->tun_proto, 0))
2037 goto nla_put_failure;
2038 }
2039
2040 if (swkey->phy.in_port == DP_MAX_PORTS) {
2041 if (is_mask && (output->phy.in_port == 0xffff))
2042 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
2043 goto nla_put_failure;
2044 } else {
2045 u16 upper_u16;
2046 upper_u16 = !is_mask ? 0 : 0xffff;
2047
2048 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
2049 (upper_u16 << 16) | output->phy.in_port))
2050 goto nla_put_failure;
2051 }
2052
2053 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
2054 goto nla_put_failure;
2055
2056 if (ovs_ct_put_key(swkey, output, skb))
2057 goto nla_put_failure;
2058
2059 if (ovs_key_mac_proto(swkey) == MAC_PROTO_ETHERNET) {
2060 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
2061 if (!nla)
2062 goto nla_put_failure;
2063
2064 eth_key = nla_data(nla);
2065 ether_addr_copy(eth_key->eth_src, output->eth.src);
2066 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
2067
2068 if (swkey->eth.vlan.tci || eth_type_vlan(swkey->eth.type)) {
2069 if (ovs_nla_put_vlan(skb, &output->eth.vlan, is_mask))
2070 goto nla_put_failure;
2071 encap = nla_nest_start_noflag(skb, OVS_KEY_ATTR_ENCAP);
2072 if (!swkey->eth.vlan.tci)
2073 goto unencap;
2074
2075 if (swkey->eth.cvlan.tci || eth_type_vlan(swkey->eth.type)) {
2076 if (ovs_nla_put_vlan(skb, &output->eth.cvlan, is_mask))
2077 goto nla_put_failure;
2078 in_encap = nla_nest_start_noflag(skb,
2079 OVS_KEY_ATTR_ENCAP);
2080 if (!swkey->eth.cvlan.tci)
2081 goto unencap;
2082 }
2083 }
2084
2085 if (swkey->eth.type == htons(ETH_P_802_2)) {
2086 /*
2087 * Ethertype 802.2 is represented in the netlink with omitted
2088 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
2089 * 0xffff in the mask attribute. Ethertype can also
2090 * be wildcarded.
2091 */
2092 if (is_mask && output->eth.type)
2093 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
2094 output->eth.type))
2095 goto nla_put_failure;
2096 goto unencap;
2097 }
2098 }
2099
2100 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
2101 goto nla_put_failure;
2102
2103 if (eth_type_vlan(swkey->eth.type)) {
2104 /* There are 3 VLAN tags, we don't know anything about the rest
2105 * of the packet, so truncate here.
2106 */
2107 WARN_ON_ONCE(!(encap && in_encap));
2108 goto unencap;
2109 }
2110
2111 if (swkey->eth.type == htons(ETH_P_IP)) {
2112 struct ovs_key_ipv4 *ipv4_key;
2113
2114 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
2115 if (!nla)
2116 goto nla_put_failure;
2117 ipv4_key = nla_data(nla);
2118 ipv4_key->ipv4_src = output->ipv4.addr.src;
2119 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
2120 ipv4_key->ipv4_proto = output->ip.proto;
2121 ipv4_key->ipv4_tos = output->ip.tos;
2122 ipv4_key->ipv4_ttl = output->ip.ttl;
2123 ipv4_key->ipv4_frag = output->ip.frag;
2124 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
2125 struct ovs_key_ipv6 *ipv6_key;
2126 struct ovs_key_ipv6_exthdrs *ipv6_exthdrs_key;
2127
2128 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
2129 if (!nla)
2130 goto nla_put_failure;
2131 ipv6_key = nla_data(nla);
2132 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
2133 sizeof(ipv6_key->ipv6_src));
2134 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
2135 sizeof(ipv6_key->ipv6_dst));
2136 ipv6_key->ipv6_label = output->ipv6.label;
2137 ipv6_key->ipv6_proto = output->ip.proto;
2138 ipv6_key->ipv6_tclass = output->ip.tos;
2139 ipv6_key->ipv6_hlimit = output->ip.ttl;
2140 ipv6_key->ipv6_frag = output->ip.frag;
2141
2142 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6_EXTHDRS,
2143 sizeof(*ipv6_exthdrs_key));
2144 if (!nla)
2145 goto nla_put_failure;
2146 ipv6_exthdrs_key = nla_data(nla);
2147 ipv6_exthdrs_key->hdrs = output->ipv6.exthdrs;
2148 } else if (swkey->eth.type == htons(ETH_P_NSH)) {
2149 if (nsh_key_to_nlattr(&output->nsh, is_mask, skb))
2150 goto nla_put_failure;
2151 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
2152 swkey->eth.type == htons(ETH_P_RARP)) {
2153 struct ovs_key_arp *arp_key;
2154
2155 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
2156 if (!nla)
2157 goto nla_put_failure;
2158 arp_key = nla_data(nla);
2159 memset(arp_key, 0, sizeof(struct ovs_key_arp));
2160 arp_key->arp_sip = output->ipv4.addr.src;
2161 arp_key->arp_tip = output->ipv4.addr.dst;
2162 arp_key->arp_op = htons(output->ip.proto);
2163 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
2164 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
2165 } else if (eth_p_mpls(swkey->eth.type)) {
2166 u8 i, num_labels;
2167 struct ovs_key_mpls *mpls_key;
2168
2169 num_labels = hweight_long(output->mpls.num_labels_mask);
2170 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS,
2171 num_labels * sizeof(*mpls_key));
2172 if (!nla)
2173 goto nla_put_failure;
2174
2175 mpls_key = nla_data(nla);
2176 for (i = 0; i < num_labels; i++)
2177 mpls_key[i].mpls_lse = output->mpls.lse[i];
2178 }
2179
2180 if ((swkey->eth.type == htons(ETH_P_IP) ||
2181 swkey->eth.type == htons(ETH_P_IPV6)) &&
2182 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
2183
2184 if (swkey->ip.proto == IPPROTO_TCP) {
2185 struct ovs_key_tcp *tcp_key;
2186
2187 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
2188 if (!nla)
2189 goto nla_put_failure;
2190 tcp_key = nla_data(nla);
2191 tcp_key->tcp_src = output->tp.src;
2192 tcp_key->tcp_dst = output->tp.dst;
2193 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
2194 output->tp.flags))
2195 goto nla_put_failure;
2196 } else if (swkey->ip.proto == IPPROTO_UDP) {
2197 struct ovs_key_udp *udp_key;
2198
2199 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
2200 if (!nla)
2201 goto nla_put_failure;
2202 udp_key = nla_data(nla);
2203 udp_key->udp_src = output->tp.src;
2204 udp_key->udp_dst = output->tp.dst;
2205 } else if (swkey->ip.proto == IPPROTO_SCTP) {
2206 struct ovs_key_sctp *sctp_key;
2207
2208 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
2209 if (!nla)
2210 goto nla_put_failure;
2211 sctp_key = nla_data(nla);
2212 sctp_key->sctp_src = output->tp.src;
2213 sctp_key->sctp_dst = output->tp.dst;
2214 } else if (swkey->eth.type == htons(ETH_P_IP) &&
2215 swkey->ip.proto == IPPROTO_ICMP) {
2216 struct ovs_key_icmp *icmp_key;
2217
2218 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
2219 if (!nla)
2220 goto nla_put_failure;
2221 icmp_key = nla_data(nla);
2222 icmp_key->icmp_type = ntohs(output->tp.src);
2223 icmp_key->icmp_code = ntohs(output->tp.dst);
2224 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
2225 swkey->ip.proto == IPPROTO_ICMPV6) {
2226 struct ovs_key_icmpv6 *icmpv6_key;
2227
2228 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
2229 sizeof(*icmpv6_key));
2230 if (!nla)
2231 goto nla_put_failure;
2232 icmpv6_key = nla_data(nla);
2233 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
2234 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
2235
2236 if (swkey->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
2237 swkey->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
2238 struct ovs_key_nd *nd_key;
2239
2240 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
2241 if (!nla)
2242 goto nla_put_failure;
2243 nd_key = nla_data(nla);
2244 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
2245 sizeof(nd_key->nd_target));
2246 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
2247 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
2248 }
2249 }
2250 }
2251
2252unencap:
2253 if (in_encap)
2254 nla_nest_end(skb, in_encap);
2255 if (encap)
2256 nla_nest_end(skb, encap);
2257
2258 return 0;
2259
2260nla_put_failure:
2261 return -EMSGSIZE;
2262}
2263
2264int ovs_nla_put_key(const struct sw_flow_key *swkey,
2265 const struct sw_flow_key *output, int attr, bool is_mask,
2266 struct sk_buff *skb)
2267{
2268 int err;
2269 struct nlattr *nla;
2270
2271 nla = nla_nest_start_noflag(skb, attr);
2272 if (!nla)
2273 return -EMSGSIZE;
2274 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
2275 if (err)
2276 return err;
2277 nla_nest_end(skb, nla);
2278
2279 return 0;
2280}
2281
2282/* Called with ovs_mutex or RCU read lock. */
2283int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
2284{
2285 if (ovs_identifier_is_ufid(&flow->id))
2286 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
2287 flow->id.ufid);
2288
2289 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
2290 OVS_FLOW_ATTR_KEY, false, skb);
2291}
2292
2293/* Called with ovs_mutex or RCU read lock. */
2294int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
2295{
2296 return ovs_nla_put_key(&flow->key, &flow->key,
2297 OVS_FLOW_ATTR_KEY, false, skb);
2298}
2299
2300/* Called with ovs_mutex or RCU read lock. */
2301int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
2302{
2303 return ovs_nla_put_key(&flow->key, &flow->mask->key,
2304 OVS_FLOW_ATTR_MASK, true, skb);
2305}
2306
2307#define MAX_ACTIONS_BUFSIZE (32 * 1024)
2308
2309static struct sw_flow_actions *nla_alloc_flow_actions(int size)
2310{
2311 struct sw_flow_actions *sfa;
2312
2313 WARN_ON_ONCE(size > MAX_ACTIONS_BUFSIZE);
2314
2315 sfa = kmalloc(kmalloc_size_roundup(sizeof(*sfa) + size), GFP_KERNEL);
2316 if (!sfa)
2317 return ERR_PTR(-ENOMEM);
2318
2319 sfa->actions_len = 0;
2320 return sfa;
2321}
2322
2323static void ovs_nla_free_nested_actions(const struct nlattr *actions, int len);
2324
2325static void ovs_nla_free_check_pkt_len_action(const struct nlattr *action)
2326{
2327 const struct nlattr *a;
2328 int rem;
2329
2330 nla_for_each_nested(a, action, rem) {
2331 switch (nla_type(a)) {
2332 case OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL:
2333 case OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER:
2334 ovs_nla_free_nested_actions(nla_data(a), nla_len(a));
2335 break;
2336 }
2337 }
2338}
2339
2340static void ovs_nla_free_clone_action(const struct nlattr *action)
2341{
2342 const struct nlattr *a = nla_data(action);
2343 int rem = nla_len(action);
2344
2345 switch (nla_type(a)) {
2346 case OVS_CLONE_ATTR_EXEC:
2347 /* The real list of actions follows this attribute. */
2348 a = nla_next(a, &rem);
2349 ovs_nla_free_nested_actions(a, rem);
2350 break;
2351 }
2352}
2353
2354static void ovs_nla_free_dec_ttl_action(const struct nlattr *action)
2355{
2356 const struct nlattr *a = nla_data(action);
2357
2358 switch (nla_type(a)) {
2359 case OVS_DEC_TTL_ATTR_ACTION:
2360 ovs_nla_free_nested_actions(nla_data(a), nla_len(a));
2361 break;
2362 }
2363}
2364
2365static void ovs_nla_free_sample_action(const struct nlattr *action)
2366{
2367 const struct nlattr *a = nla_data(action);
2368 int rem = nla_len(action);
2369
2370 switch (nla_type(a)) {
2371 case OVS_SAMPLE_ATTR_ARG:
2372 /* The real list of actions follows this attribute. */
2373 a = nla_next(a, &rem);
2374 ovs_nla_free_nested_actions(a, rem);
2375 break;
2376 }
2377}
2378
2379static void ovs_nla_free_set_action(const struct nlattr *a)
2380{
2381 const struct nlattr *ovs_key = nla_data(a);
2382 struct ovs_tunnel_info *ovs_tun;
2383
2384 switch (nla_type(ovs_key)) {
2385 case OVS_KEY_ATTR_TUNNEL_INFO:
2386 ovs_tun = nla_data(ovs_key);
2387 dst_release((struct dst_entry *)ovs_tun->tun_dst);
2388 break;
2389 }
2390}
2391
2392static void ovs_nla_free_nested_actions(const struct nlattr *actions, int len)
2393{
2394 const struct nlattr *a;
2395 int rem;
2396
2397 /* Whenever new actions are added, the need to update this
2398 * function should be considered.
2399 */
2400 BUILD_BUG_ON(OVS_ACTION_ATTR_MAX != 24);
2401
2402 if (!actions)
2403 return;
2404
2405 nla_for_each_attr(a, actions, len, rem) {
2406 switch (nla_type(a)) {
2407 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
2408 ovs_nla_free_check_pkt_len_action(a);
2409 break;
2410
2411 case OVS_ACTION_ATTR_CLONE:
2412 ovs_nla_free_clone_action(a);
2413 break;
2414
2415 case OVS_ACTION_ATTR_CT:
2416 ovs_ct_free_action(a);
2417 break;
2418
2419 case OVS_ACTION_ATTR_DEC_TTL:
2420 ovs_nla_free_dec_ttl_action(a);
2421 break;
2422
2423 case OVS_ACTION_ATTR_SAMPLE:
2424 ovs_nla_free_sample_action(a);
2425 break;
2426
2427 case OVS_ACTION_ATTR_SET:
2428 ovs_nla_free_set_action(a);
2429 break;
2430 }
2431 }
2432}
2433
2434void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
2435{
2436 if (!sf_acts)
2437 return;
2438
2439 ovs_nla_free_nested_actions(sf_acts->actions, sf_acts->actions_len);
2440 kfree(sf_acts);
2441}
2442
2443static void __ovs_nla_free_flow_actions(struct rcu_head *head)
2444{
2445 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
2446}
2447
2448/* Schedules 'sf_acts' to be freed after the next RCU grace period.
2449 * The caller must hold rcu_read_lock for this to be sensible. */
2450void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
2451{
2452 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
2453}
2454
2455static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
2456 int attr_len, bool log)
2457{
2458
2459 struct sw_flow_actions *acts;
2460 int new_acts_size;
2461 size_t req_size = NLA_ALIGN(attr_len);
2462 int next_offset = offsetof(struct sw_flow_actions, actions) +
2463 (*sfa)->actions_len;
2464
2465 if (req_size <= (ksize(*sfa) - next_offset))
2466 goto out;
2467
2468 new_acts_size = max(next_offset + req_size, ksize(*sfa) * 2);
2469
2470 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
2471 if ((next_offset + req_size) > MAX_ACTIONS_BUFSIZE) {
2472 OVS_NLERR(log, "Flow action size exceeds max %u",
2473 MAX_ACTIONS_BUFSIZE);
2474 return ERR_PTR(-EMSGSIZE);
2475 }
2476 new_acts_size = MAX_ACTIONS_BUFSIZE;
2477 }
2478
2479 acts = nla_alloc_flow_actions(new_acts_size);
2480 if (IS_ERR(acts))
2481 return (void *)acts;
2482
2483 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
2484 acts->actions_len = (*sfa)->actions_len;
2485 acts->orig_len = (*sfa)->orig_len;
2486 kfree(*sfa);
2487 *sfa = acts;
2488
2489out:
2490 (*sfa)->actions_len += req_size;
2491 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
2492}
2493
2494static struct nlattr *__add_action(struct sw_flow_actions **sfa,
2495 int attrtype, void *data, int len, bool log)
2496{
2497 struct nlattr *a;
2498
2499 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
2500 if (IS_ERR(a))
2501 return a;
2502
2503 a->nla_type = attrtype;
2504 a->nla_len = nla_attr_size(len);
2505
2506 if (data)
2507 memcpy(nla_data(a), data, len);
2508 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
2509
2510 return a;
2511}
2512
2513int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
2514 int len, bool log)
2515{
2516 struct nlattr *a;
2517
2518 a = __add_action(sfa, attrtype, data, len, log);
2519
2520 return PTR_ERR_OR_ZERO(a);
2521}
2522
2523static inline int add_nested_action_start(struct sw_flow_actions **sfa,
2524 int attrtype, bool log)
2525{
2526 int used = (*sfa)->actions_len;
2527 int err;
2528
2529 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
2530 if (err)
2531 return err;
2532
2533 return used;
2534}
2535
2536static inline void add_nested_action_end(struct sw_flow_actions *sfa,
2537 int st_offset)
2538{
2539 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
2540 st_offset);
2541
2542 a->nla_len = sfa->actions_len - st_offset;
2543}
2544
2545static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2546 const struct sw_flow_key *key,
2547 struct sw_flow_actions **sfa,
2548 __be16 eth_type, __be16 vlan_tci,
2549 u32 mpls_label_count, bool log,
2550 u32 depth);
2551
2552static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
2553 const struct sw_flow_key *key,
2554 struct sw_flow_actions **sfa,
2555 __be16 eth_type, __be16 vlan_tci,
2556 u32 mpls_label_count, bool log, bool last,
2557 u32 depth)
2558{
2559 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
2560 const struct nlattr *probability, *actions;
2561 const struct nlattr *a;
2562 int rem, start, err;
2563 struct sample_arg arg;
2564
2565 memset(attrs, 0, sizeof(attrs));
2566 nla_for_each_nested(a, attr, rem) {
2567 int type = nla_type(a);
2568 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
2569 return -EINVAL;
2570 attrs[type] = a;
2571 }
2572 if (rem)
2573 return -EINVAL;
2574
2575 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
2576 if (!probability || nla_len(probability) != sizeof(u32))
2577 return -EINVAL;
2578
2579 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
2580 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
2581 return -EINVAL;
2582
2583 /* validation done, copy sample action. */
2584 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
2585 if (start < 0)
2586 return start;
2587
2588 /* When both skb and flow may be changed, put the sample
2589 * into a deferred fifo. On the other hand, if only skb
2590 * may be modified, the actions can be executed in place.
2591 *
2592 * Do this analysis at the flow installation time.
2593 * Set 'clone_action->exec' to true if the actions can be
2594 * executed without being deferred.
2595 *
2596 * If the sample is the last action, it can always be excuted
2597 * rather than deferred.
2598 */
2599 arg.exec = last || !actions_may_change_flow(actions);
2600 arg.probability = nla_get_u32(probability);
2601
2602 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_ARG, &arg, sizeof(arg),
2603 log);
2604 if (err)
2605 return err;
2606
2607 err = __ovs_nla_copy_actions(net, actions, key, sfa,
2608 eth_type, vlan_tci, mpls_label_count, log,
2609 depth + 1);
2610
2611 if (err)
2612 return err;
2613
2614 add_nested_action_end(*sfa, start);
2615
2616 return 0;
2617}
2618
2619static int validate_and_copy_dec_ttl(struct net *net,
2620 const struct nlattr *attr,
2621 const struct sw_flow_key *key,
2622 struct sw_flow_actions **sfa,
2623 __be16 eth_type, __be16 vlan_tci,
2624 u32 mpls_label_count, bool log,
2625 u32 depth)
2626{
2627 const struct nlattr *attrs[OVS_DEC_TTL_ATTR_MAX + 1];
2628 int start, action_start, err, rem;
2629 const struct nlattr *a, *actions;
2630
2631 memset(attrs, 0, sizeof(attrs));
2632 nla_for_each_nested(a, attr, rem) {
2633 int type = nla_type(a);
2634
2635 /* Ignore unknown attributes to be future proof. */
2636 if (type > OVS_DEC_TTL_ATTR_MAX)
2637 continue;
2638
2639 if (!type || attrs[type]) {
2640 OVS_NLERR(log, "Duplicate or invalid key (type %d).",
2641 type);
2642 return -EINVAL;
2643 }
2644
2645 attrs[type] = a;
2646 }
2647
2648 if (rem) {
2649 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
2650 return -EINVAL;
2651 }
2652
2653 actions = attrs[OVS_DEC_TTL_ATTR_ACTION];
2654 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN)) {
2655 OVS_NLERR(log, "Missing valid actions attribute.");
2656 return -EINVAL;
2657 }
2658
2659 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_DEC_TTL, log);
2660 if (start < 0)
2661 return start;
2662
2663 action_start = add_nested_action_start(sfa, OVS_DEC_TTL_ATTR_ACTION, log);
2664 if (action_start < 0)
2665 return action_start;
2666
2667 err = __ovs_nla_copy_actions(net, actions, key, sfa, eth_type,
2668 vlan_tci, mpls_label_count, log,
2669 depth + 1);
2670 if (err)
2671 return err;
2672
2673 add_nested_action_end(*sfa, action_start);
2674 add_nested_action_end(*sfa, start);
2675 return 0;
2676}
2677
2678static int validate_and_copy_clone(struct net *net,
2679 const struct nlattr *attr,
2680 const struct sw_flow_key *key,
2681 struct sw_flow_actions **sfa,
2682 __be16 eth_type, __be16 vlan_tci,
2683 u32 mpls_label_count, bool log, bool last,
2684 u32 depth)
2685{
2686 int start, err;
2687 u32 exec;
2688
2689 if (nla_len(attr) && nla_len(attr) < NLA_HDRLEN)
2690 return -EINVAL;
2691
2692 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CLONE, log);
2693 if (start < 0)
2694 return start;
2695
2696 exec = last || !actions_may_change_flow(attr);
2697
2698 err = ovs_nla_add_action(sfa, OVS_CLONE_ATTR_EXEC, &exec,
2699 sizeof(exec), log);
2700 if (err)
2701 return err;
2702
2703 err = __ovs_nla_copy_actions(net, attr, key, sfa,
2704 eth_type, vlan_tci, mpls_label_count, log,
2705 depth + 1);
2706 if (err)
2707 return err;
2708
2709 add_nested_action_end(*sfa, start);
2710
2711 return 0;
2712}
2713
2714void ovs_match_init(struct sw_flow_match *match,
2715 struct sw_flow_key *key,
2716 bool reset_key,
2717 struct sw_flow_mask *mask)
2718{
2719 memset(match, 0, sizeof(*match));
2720 match->key = key;
2721 match->mask = mask;
2722
2723 if (reset_key)
2724 memset(key, 0, sizeof(*key));
2725
2726 if (mask) {
2727 memset(&mask->key, 0, sizeof(mask->key));
2728 mask->range.start = mask->range.end = 0;
2729 }
2730}
2731
2732static int validate_geneve_opts(struct sw_flow_key *key)
2733{
2734 struct geneve_opt *option;
2735 int opts_len = key->tun_opts_len;
2736 bool crit_opt = false;
2737
2738 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2739 while (opts_len > 0) {
2740 int len;
2741
2742 if (opts_len < sizeof(*option))
2743 return -EINVAL;
2744
2745 len = sizeof(*option) + option->length * 4;
2746 if (len > opts_len)
2747 return -EINVAL;
2748
2749 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2750
2751 option = (struct geneve_opt *)((u8 *)option + len);
2752 opts_len -= len;
2753 }
2754
2755 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
2756
2757 return 0;
2758}
2759
2760static int validate_and_copy_set_tun(const struct nlattr *attr,
2761 struct sw_flow_actions **sfa, bool log)
2762{
2763 struct sw_flow_match match;
2764 struct sw_flow_key key;
2765 struct metadata_dst *tun_dst;
2766 struct ip_tunnel_info *tun_info;
2767 struct ovs_tunnel_info *ovs_tun;
2768 struct nlattr *a;
2769 int err = 0, start, opts_type;
2770 __be16 dst_opt_type;
2771
2772 dst_opt_type = 0;
2773 ovs_match_init(&match, &key, true, NULL);
2774 opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2775 if (opts_type < 0)
2776 return opts_type;
2777
2778 if (key.tun_opts_len) {
2779 switch (opts_type) {
2780 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2781 err = validate_geneve_opts(&key);
2782 if (err < 0)
2783 return err;
2784 dst_opt_type = TUNNEL_GENEVE_OPT;
2785 break;
2786 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
2787 dst_opt_type = TUNNEL_VXLAN_OPT;
2788 break;
2789 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
2790 dst_opt_type = TUNNEL_ERSPAN_OPT;
2791 break;
2792 }
2793 }
2794
2795 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2796 if (start < 0)
2797 return start;
2798
2799 tun_dst = metadata_dst_alloc(key.tun_opts_len, METADATA_IP_TUNNEL,
2800 GFP_KERNEL);
2801
2802 if (!tun_dst)
2803 return -ENOMEM;
2804
2805 err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2806 if (err) {
2807 dst_release((struct dst_entry *)tun_dst);
2808 return err;
2809 }
2810
2811 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2812 sizeof(*ovs_tun), log);
2813 if (IS_ERR(a)) {
2814 dst_release((struct dst_entry *)tun_dst);
2815 return PTR_ERR(a);
2816 }
2817
2818 ovs_tun = nla_data(a);
2819 ovs_tun->tun_dst = tun_dst;
2820
2821 tun_info = &tun_dst->u.tun_info;
2822 tun_info->mode = IP_TUNNEL_INFO_TX;
2823 if (key.tun_proto == AF_INET6)
2824 tun_info->mode |= IP_TUNNEL_INFO_IPV6;
2825 else if (key.tun_proto == AF_INET && key.tun_key.u.ipv4.dst == 0)
2826 tun_info->mode |= IP_TUNNEL_INFO_BRIDGE;
2827 tun_info->key = key.tun_key;
2828
2829 /* We need to store the options in the action itself since
2830 * everything else will go away after flow setup. We can append
2831 * it to tun_info and then point there.
2832 */
2833 ip_tunnel_info_opts_set(tun_info,
2834 TUN_METADATA_OPTS(&key, key.tun_opts_len),
2835 key.tun_opts_len, dst_opt_type);
2836 add_nested_action_end(*sfa, start);
2837
2838 return err;
2839}
2840
2841static bool validate_nsh(const struct nlattr *attr, bool is_mask,
2842 bool is_push_nsh, bool log)
2843{
2844 struct sw_flow_match match;
2845 struct sw_flow_key key;
2846 int ret = 0;
2847
2848 ovs_match_init(&match, &key, true, NULL);
2849 ret = nsh_key_put_from_nlattr(attr, &match, is_mask,
2850 is_push_nsh, log);
2851 return !ret;
2852}
2853
2854/* Return false if there are any non-masked bits set.
2855 * Mask follows data immediately, before any netlink padding.
2856 */
2857static bool validate_masked(u8 *data, int len)
2858{
2859 u8 *mask = data + len;
2860
2861 while (len--)
2862 if (*data++ & ~*mask++)
2863 return false;
2864
2865 return true;
2866}
2867
2868static int validate_set(const struct nlattr *a,
2869 const struct sw_flow_key *flow_key,
2870 struct sw_flow_actions **sfa, bool *skip_copy,
2871 u8 mac_proto, __be16 eth_type, bool masked, bool log)
2872{
2873 const struct nlattr *ovs_key = nla_data(a);
2874 int key_type = nla_type(ovs_key);
2875 size_t key_len;
2876
2877 /* There can be only one key in a action */
2878 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2879 return -EINVAL;
2880
2881 key_len = nla_len(ovs_key);
2882 if (masked)
2883 key_len /= 2;
2884
2885 if (key_type > OVS_KEY_ATTR_MAX ||
2886 !check_attr_len(key_len, ovs_key_lens[key_type].len))
2887 return -EINVAL;
2888
2889 if (masked && !validate_masked(nla_data(ovs_key), key_len))
2890 return -EINVAL;
2891
2892 switch (key_type) {
2893 case OVS_KEY_ATTR_PRIORITY:
2894 case OVS_KEY_ATTR_SKB_MARK:
2895 case OVS_KEY_ATTR_CT_MARK:
2896 case OVS_KEY_ATTR_CT_LABELS:
2897 break;
2898
2899 case OVS_KEY_ATTR_ETHERNET:
2900 if (mac_proto != MAC_PROTO_ETHERNET)
2901 return -EINVAL;
2902 break;
2903
2904 case OVS_KEY_ATTR_TUNNEL: {
2905 int err;
2906
2907 if (masked)
2908 return -EINVAL; /* Masked tunnel set not supported. */
2909
2910 *skip_copy = true;
2911 err = validate_and_copy_set_tun(a, sfa, log);
2912 if (err)
2913 return err;
2914 break;
2915 }
2916 case OVS_KEY_ATTR_IPV4: {
2917 const struct ovs_key_ipv4 *ipv4_key;
2918
2919 if (eth_type != htons(ETH_P_IP))
2920 return -EINVAL;
2921
2922 ipv4_key = nla_data(ovs_key);
2923
2924 if (masked) {
2925 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2926
2927 /* Non-writeable fields. */
2928 if (mask->ipv4_proto || mask->ipv4_frag)
2929 return -EINVAL;
2930 } else {
2931 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2932 return -EINVAL;
2933
2934 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2935 return -EINVAL;
2936 }
2937 break;
2938 }
2939 case OVS_KEY_ATTR_IPV6: {
2940 const struct ovs_key_ipv6 *ipv6_key;
2941
2942 if (eth_type != htons(ETH_P_IPV6))
2943 return -EINVAL;
2944
2945 ipv6_key = nla_data(ovs_key);
2946
2947 if (masked) {
2948 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2949
2950 /* Non-writeable fields. */
2951 if (mask->ipv6_proto || mask->ipv6_frag)
2952 return -EINVAL;
2953
2954 /* Invalid bits in the flow label mask? */
2955 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2956 return -EINVAL;
2957 } else {
2958 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2959 return -EINVAL;
2960
2961 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2962 return -EINVAL;
2963 }
2964 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2965 return -EINVAL;
2966
2967 break;
2968 }
2969 case OVS_KEY_ATTR_TCP:
2970 if ((eth_type != htons(ETH_P_IP) &&
2971 eth_type != htons(ETH_P_IPV6)) ||
2972 flow_key->ip.proto != IPPROTO_TCP)
2973 return -EINVAL;
2974
2975 break;
2976
2977 case OVS_KEY_ATTR_UDP:
2978 if ((eth_type != htons(ETH_P_IP) &&
2979 eth_type != htons(ETH_P_IPV6)) ||
2980 flow_key->ip.proto != IPPROTO_UDP)
2981 return -EINVAL;
2982
2983 break;
2984
2985 case OVS_KEY_ATTR_MPLS:
2986 if (!eth_p_mpls(eth_type))
2987 return -EINVAL;
2988 break;
2989
2990 case OVS_KEY_ATTR_SCTP:
2991 if ((eth_type != htons(ETH_P_IP) &&
2992 eth_type != htons(ETH_P_IPV6)) ||
2993 flow_key->ip.proto != IPPROTO_SCTP)
2994 return -EINVAL;
2995
2996 break;
2997
2998 case OVS_KEY_ATTR_NSH:
2999 if (eth_type != htons(ETH_P_NSH))
3000 return -EINVAL;
3001 if (!validate_nsh(nla_data(a), masked, false, log))
3002 return -EINVAL;
3003 break;
3004
3005 default:
3006 return -EINVAL;
3007 }
3008
3009 /* Convert non-masked non-tunnel set actions to masked set actions. */
3010 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
3011 int start, len = key_len * 2;
3012 struct nlattr *at;
3013
3014 *skip_copy = true;
3015
3016 start = add_nested_action_start(sfa,
3017 OVS_ACTION_ATTR_SET_TO_MASKED,
3018 log);
3019 if (start < 0)
3020 return start;
3021
3022 at = __add_action(sfa, key_type, NULL, len, log);
3023 if (IS_ERR(at))
3024 return PTR_ERR(at);
3025
3026 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
3027 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
3028 /* Clear non-writeable bits from otherwise writeable fields. */
3029 if (key_type == OVS_KEY_ATTR_IPV6) {
3030 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
3031
3032 mask->ipv6_label &= htonl(0x000FFFFF);
3033 }
3034 add_nested_action_end(*sfa, start);
3035 }
3036
3037 return 0;
3038}
3039
3040static int validate_userspace(const struct nlattr *attr)
3041{
3042 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
3043 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
3044 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
3045 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
3046 };
3047 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
3048 int error;
3049
3050 error = nla_parse_nested_deprecated(a, OVS_USERSPACE_ATTR_MAX, attr,
3051 userspace_policy, NULL);
3052 if (error)
3053 return error;
3054
3055 if (!a[OVS_USERSPACE_ATTR_PID] ||
3056 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
3057 return -EINVAL;
3058
3059 return 0;
3060}
3061
3062static const struct nla_policy cpl_policy[OVS_CHECK_PKT_LEN_ATTR_MAX + 1] = {
3063 [OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] = {.type = NLA_U16 },
3064 [OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER] = {.type = NLA_NESTED },
3065 [OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL] = {.type = NLA_NESTED },
3066};
3067
3068static int validate_and_copy_check_pkt_len(struct net *net,
3069 const struct nlattr *attr,
3070 const struct sw_flow_key *key,
3071 struct sw_flow_actions **sfa,
3072 __be16 eth_type, __be16 vlan_tci,
3073 u32 mpls_label_count,
3074 bool log, bool last, u32 depth)
3075{
3076 const struct nlattr *acts_if_greater, *acts_if_lesser_eq;
3077 struct nlattr *a[OVS_CHECK_PKT_LEN_ATTR_MAX + 1];
3078 struct check_pkt_len_arg arg;
3079 int nested_acts_start;
3080 int start, err;
3081
3082 err = nla_parse_deprecated_strict(a, OVS_CHECK_PKT_LEN_ATTR_MAX,
3083 nla_data(attr), nla_len(attr),
3084 cpl_policy, NULL);
3085 if (err)
3086 return err;
3087
3088 if (!a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] ||
3089 !nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]))
3090 return -EINVAL;
3091
3092 acts_if_lesser_eq = a[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL];
3093 acts_if_greater = a[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER];
3094
3095 /* Both the nested action should be present. */
3096 if (!acts_if_greater || !acts_if_lesser_eq)
3097 return -EINVAL;
3098
3099 /* validation done, copy the nested actions. */
3100 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CHECK_PKT_LEN,
3101 log);
3102 if (start < 0)
3103 return start;
3104
3105 arg.pkt_len = nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]);
3106 arg.exec_for_lesser_equal =
3107 last || !actions_may_change_flow(acts_if_lesser_eq);
3108 arg.exec_for_greater =
3109 last || !actions_may_change_flow(acts_if_greater);
3110
3111 err = ovs_nla_add_action(sfa, OVS_CHECK_PKT_LEN_ATTR_ARG, &arg,
3112 sizeof(arg), log);
3113 if (err)
3114 return err;
3115
3116 nested_acts_start = add_nested_action_start(sfa,
3117 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL, log);
3118 if (nested_acts_start < 0)
3119 return nested_acts_start;
3120
3121 err = __ovs_nla_copy_actions(net, acts_if_lesser_eq, key, sfa,
3122 eth_type, vlan_tci, mpls_label_count, log,
3123 depth + 1);
3124
3125 if (err)
3126 return err;
3127
3128 add_nested_action_end(*sfa, nested_acts_start);
3129
3130 nested_acts_start = add_nested_action_start(sfa,
3131 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER, log);
3132 if (nested_acts_start < 0)
3133 return nested_acts_start;
3134
3135 err = __ovs_nla_copy_actions(net, acts_if_greater, key, sfa,
3136 eth_type, vlan_tci, mpls_label_count, log,
3137 depth + 1);
3138
3139 if (err)
3140 return err;
3141
3142 add_nested_action_end(*sfa, nested_acts_start);
3143 add_nested_action_end(*sfa, start);
3144 return 0;
3145}
3146
3147static int copy_action(const struct nlattr *from,
3148 struct sw_flow_actions **sfa, bool log)
3149{
3150 int totlen = NLA_ALIGN(from->nla_len);
3151 struct nlattr *to;
3152
3153 to = reserve_sfa_size(sfa, from->nla_len, log);
3154 if (IS_ERR(to))
3155 return PTR_ERR(to);
3156
3157 memcpy(to, from, totlen);
3158 return 0;
3159}
3160
3161static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3162 const struct sw_flow_key *key,
3163 struct sw_flow_actions **sfa,
3164 __be16 eth_type, __be16 vlan_tci,
3165 u32 mpls_label_count, bool log,
3166 u32 depth)
3167{
3168 u8 mac_proto = ovs_key_mac_proto(key);
3169 const struct nlattr *a;
3170 int rem, err;
3171
3172 if (depth > OVS_COPY_ACTIONS_MAX_DEPTH)
3173 return -EOVERFLOW;
3174
3175 nla_for_each_nested(a, attr, rem) {
3176 /* Expected argument lengths, (u32)-1 for variable length. */
3177 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
3178 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
3179 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
3180 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
3181 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
3182 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
3183 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
3184 [OVS_ACTION_ATTR_POP_VLAN] = 0,
3185 [OVS_ACTION_ATTR_SET] = (u32)-1,
3186 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
3187 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
3188 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
3189 [OVS_ACTION_ATTR_CT] = (u32)-1,
3190 [OVS_ACTION_ATTR_CT_CLEAR] = 0,
3191 [OVS_ACTION_ATTR_TRUNC] = sizeof(struct ovs_action_trunc),
3192 [OVS_ACTION_ATTR_PUSH_ETH] = sizeof(struct ovs_action_push_eth),
3193 [OVS_ACTION_ATTR_POP_ETH] = 0,
3194 [OVS_ACTION_ATTR_PUSH_NSH] = (u32)-1,
3195 [OVS_ACTION_ATTR_POP_NSH] = 0,
3196 [OVS_ACTION_ATTR_METER] = sizeof(u32),
3197 [OVS_ACTION_ATTR_CLONE] = (u32)-1,
3198 [OVS_ACTION_ATTR_CHECK_PKT_LEN] = (u32)-1,
3199 [OVS_ACTION_ATTR_ADD_MPLS] = sizeof(struct ovs_action_add_mpls),
3200 [OVS_ACTION_ATTR_DEC_TTL] = (u32)-1,
3201 [OVS_ACTION_ATTR_DROP] = sizeof(u32),
3202 };
3203 const struct ovs_action_push_vlan *vlan;
3204 int type = nla_type(a);
3205 bool skip_copy;
3206
3207 if (type > OVS_ACTION_ATTR_MAX ||
3208 (action_lens[type] != nla_len(a) &&
3209 action_lens[type] != (u32)-1))
3210 return -EINVAL;
3211
3212 skip_copy = false;
3213 switch (type) {
3214 case OVS_ACTION_ATTR_UNSPEC:
3215 return -EINVAL;
3216
3217 case OVS_ACTION_ATTR_USERSPACE:
3218 err = validate_userspace(a);
3219 if (err)
3220 return err;
3221 break;
3222
3223 case OVS_ACTION_ATTR_OUTPUT:
3224 if (nla_get_u32(a) >= DP_MAX_PORTS)
3225 return -EINVAL;
3226 break;
3227
3228 case OVS_ACTION_ATTR_TRUNC: {
3229 const struct ovs_action_trunc *trunc = nla_data(a);
3230
3231 if (trunc->max_len < ETH_HLEN)
3232 return -EINVAL;
3233 break;
3234 }
3235
3236 case OVS_ACTION_ATTR_HASH: {
3237 const struct ovs_action_hash *act_hash = nla_data(a);
3238
3239 switch (act_hash->hash_alg) {
3240 case OVS_HASH_ALG_L4:
3241 fallthrough;
3242 case OVS_HASH_ALG_SYM_L4:
3243 break;
3244 default:
3245 return -EINVAL;
3246 }
3247
3248 break;
3249 }
3250
3251 case OVS_ACTION_ATTR_POP_VLAN:
3252 if (mac_proto != MAC_PROTO_ETHERNET)
3253 return -EINVAL;
3254 vlan_tci = htons(0);
3255 break;
3256
3257 case OVS_ACTION_ATTR_PUSH_VLAN:
3258 if (mac_proto != MAC_PROTO_ETHERNET)
3259 return -EINVAL;
3260 vlan = nla_data(a);
3261 if (!eth_type_vlan(vlan->vlan_tpid))
3262 return -EINVAL;
3263 if (!(vlan->vlan_tci & htons(VLAN_CFI_MASK)))
3264 return -EINVAL;
3265 vlan_tci = vlan->vlan_tci;
3266 break;
3267
3268 case OVS_ACTION_ATTR_RECIRC:
3269 break;
3270
3271 case OVS_ACTION_ATTR_ADD_MPLS: {
3272 const struct ovs_action_add_mpls *mpls = nla_data(a);
3273
3274 if (!eth_p_mpls(mpls->mpls_ethertype))
3275 return -EINVAL;
3276
3277 if (mpls->tun_flags & OVS_MPLS_L3_TUNNEL_FLAG_MASK) {
3278 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3279 (eth_type != htons(ETH_P_IP) &&
3280 eth_type != htons(ETH_P_IPV6) &&
3281 eth_type != htons(ETH_P_ARP) &&
3282 eth_type != htons(ETH_P_RARP) &&
3283 !eth_p_mpls(eth_type)))
3284 return -EINVAL;
3285 mpls_label_count++;
3286 } else {
3287 if (mac_proto == MAC_PROTO_ETHERNET) {
3288 mpls_label_count = 1;
3289 mac_proto = MAC_PROTO_NONE;
3290 } else {
3291 mpls_label_count++;
3292 }
3293 }
3294 eth_type = mpls->mpls_ethertype;
3295 break;
3296 }
3297
3298 case OVS_ACTION_ATTR_PUSH_MPLS: {
3299 const struct ovs_action_push_mpls *mpls = nla_data(a);
3300
3301 if (!eth_p_mpls(mpls->mpls_ethertype))
3302 return -EINVAL;
3303 /* Prohibit push MPLS other than to a white list
3304 * for packets that have a known tag order.
3305 */
3306 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3307 (eth_type != htons(ETH_P_IP) &&
3308 eth_type != htons(ETH_P_IPV6) &&
3309 eth_type != htons(ETH_P_ARP) &&
3310 eth_type != htons(ETH_P_RARP) &&
3311 !eth_p_mpls(eth_type)))
3312 return -EINVAL;
3313 eth_type = mpls->mpls_ethertype;
3314 mpls_label_count++;
3315 break;
3316 }
3317
3318 case OVS_ACTION_ATTR_POP_MPLS: {
3319 __be16 proto;
3320 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3321 !eth_p_mpls(eth_type))
3322 return -EINVAL;
3323
3324 /* Disallow subsequent L2.5+ set actions and mpls_pop
3325 * actions once the last MPLS label in the packet is
3326 * popped as there is no check here to ensure that
3327 * the new eth type is valid and thus set actions could
3328 * write off the end of the packet or otherwise corrupt
3329 * it.
3330 *
3331 * Support for these actions is planned using packet
3332 * recirculation.
3333 */
3334 proto = nla_get_be16(a);
3335
3336 if (proto == htons(ETH_P_TEB) &&
3337 mac_proto != MAC_PROTO_NONE)
3338 return -EINVAL;
3339
3340 mpls_label_count--;
3341
3342 if (!eth_p_mpls(proto) || !mpls_label_count)
3343 eth_type = htons(0);
3344 else
3345 eth_type = proto;
3346
3347 break;
3348 }
3349
3350 case OVS_ACTION_ATTR_SET:
3351 err = validate_set(a, key, sfa,
3352 &skip_copy, mac_proto, eth_type,
3353 false, log);
3354 if (err)
3355 return err;
3356 break;
3357
3358 case OVS_ACTION_ATTR_SET_MASKED:
3359 err = validate_set(a, key, sfa,
3360 &skip_copy, mac_proto, eth_type,
3361 true, log);
3362 if (err)
3363 return err;
3364 break;
3365
3366 case OVS_ACTION_ATTR_SAMPLE: {
3367 bool last = nla_is_last(a, rem);
3368
3369 err = validate_and_copy_sample(net, a, key, sfa,
3370 eth_type, vlan_tci,
3371 mpls_label_count,
3372 log, last, depth);
3373 if (err)
3374 return err;
3375 skip_copy = true;
3376 break;
3377 }
3378
3379 case OVS_ACTION_ATTR_CT:
3380 err = ovs_ct_copy_action(net, a, key, sfa, log);
3381 if (err)
3382 return err;
3383 skip_copy = true;
3384 break;
3385
3386 case OVS_ACTION_ATTR_CT_CLEAR:
3387 break;
3388
3389 case OVS_ACTION_ATTR_PUSH_ETH:
3390 /* Disallow pushing an Ethernet header if one
3391 * is already present */
3392 if (mac_proto != MAC_PROTO_NONE)
3393 return -EINVAL;
3394 mac_proto = MAC_PROTO_ETHERNET;
3395 break;
3396
3397 case OVS_ACTION_ATTR_POP_ETH:
3398 if (mac_proto != MAC_PROTO_ETHERNET)
3399 return -EINVAL;
3400 if (vlan_tci & htons(VLAN_CFI_MASK))
3401 return -EINVAL;
3402 mac_proto = MAC_PROTO_NONE;
3403 break;
3404
3405 case OVS_ACTION_ATTR_PUSH_NSH:
3406 if (mac_proto != MAC_PROTO_ETHERNET) {
3407 u8 next_proto;
3408
3409 next_proto = tun_p_from_eth_p(eth_type);
3410 if (!next_proto)
3411 return -EINVAL;
3412 }
3413 mac_proto = MAC_PROTO_NONE;
3414 if (!validate_nsh(nla_data(a), false, true, true))
3415 return -EINVAL;
3416 break;
3417
3418 case OVS_ACTION_ATTR_POP_NSH: {
3419 __be16 inner_proto;
3420
3421 if (eth_type != htons(ETH_P_NSH))
3422 return -EINVAL;
3423 inner_proto = tun_p_to_eth_p(key->nsh.base.np);
3424 if (!inner_proto)
3425 return -EINVAL;
3426 if (key->nsh.base.np == TUN_P_ETHERNET)
3427 mac_proto = MAC_PROTO_ETHERNET;
3428 else
3429 mac_proto = MAC_PROTO_NONE;
3430 break;
3431 }
3432
3433 case OVS_ACTION_ATTR_METER:
3434 /* Non-existent meters are simply ignored. */
3435 break;
3436
3437 case OVS_ACTION_ATTR_CLONE: {
3438 bool last = nla_is_last(a, rem);
3439
3440 err = validate_and_copy_clone(net, a, key, sfa,
3441 eth_type, vlan_tci,
3442 mpls_label_count,
3443 log, last, depth);
3444 if (err)
3445 return err;
3446 skip_copy = true;
3447 break;
3448 }
3449
3450 case OVS_ACTION_ATTR_CHECK_PKT_LEN: {
3451 bool last = nla_is_last(a, rem);
3452
3453 err = validate_and_copy_check_pkt_len(net, a, key, sfa,
3454 eth_type,
3455 vlan_tci,
3456 mpls_label_count,
3457 log, last,
3458 depth);
3459 if (err)
3460 return err;
3461 skip_copy = true;
3462 break;
3463 }
3464
3465 case OVS_ACTION_ATTR_DEC_TTL:
3466 err = validate_and_copy_dec_ttl(net, a, key, sfa,
3467 eth_type, vlan_tci,
3468 mpls_label_count, log,
3469 depth);
3470 if (err)
3471 return err;
3472 skip_copy = true;
3473 break;
3474
3475 case OVS_ACTION_ATTR_DROP:
3476 if (!nla_is_last(a, rem))
3477 return -EINVAL;
3478 break;
3479
3480 default:
3481 OVS_NLERR(log, "Unknown Action type %d", type);
3482 return -EINVAL;
3483 }
3484 if (!skip_copy) {
3485 err = copy_action(a, sfa, log);
3486 if (err)
3487 return err;
3488 }
3489 }
3490
3491 if (rem > 0)
3492 return -EINVAL;
3493
3494 return 0;
3495}
3496
3497/* 'key' must be the masked key. */
3498int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3499 const struct sw_flow_key *key,
3500 struct sw_flow_actions **sfa, bool log)
3501{
3502 int err;
3503 u32 mpls_label_count = 0;
3504
3505 *sfa = nla_alloc_flow_actions(min(nla_len(attr), MAX_ACTIONS_BUFSIZE));
3506 if (IS_ERR(*sfa))
3507 return PTR_ERR(*sfa);
3508
3509 if (eth_p_mpls(key->eth.type))
3510 mpls_label_count = hweight_long(key->mpls.num_labels_mask);
3511
3512 (*sfa)->orig_len = nla_len(attr);
3513 err = __ovs_nla_copy_actions(net, attr, key, sfa, key->eth.type,
3514 key->eth.vlan.tci, mpls_label_count, log,
3515 0);
3516 if (err)
3517 ovs_nla_free_flow_actions(*sfa);
3518
3519 return err;
3520}
3521
3522static int sample_action_to_attr(const struct nlattr *attr,
3523 struct sk_buff *skb)
3524{
3525 struct nlattr *start, *ac_start = NULL, *sample_arg;
3526 int err = 0, rem = nla_len(attr);
3527 const struct sample_arg *arg;
3528 struct nlattr *actions;
3529
3530 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SAMPLE);
3531 if (!start)
3532 return -EMSGSIZE;
3533
3534 sample_arg = nla_data(attr);
3535 arg = nla_data(sample_arg);
3536 actions = nla_next(sample_arg, &rem);
3537
3538 if (nla_put_u32(skb, OVS_SAMPLE_ATTR_PROBABILITY, arg->probability)) {
3539 err = -EMSGSIZE;
3540 goto out;
3541 }
3542
3543 ac_start = nla_nest_start_noflag(skb, OVS_SAMPLE_ATTR_ACTIONS);
3544 if (!ac_start) {
3545 err = -EMSGSIZE;
3546 goto out;
3547 }
3548
3549 err = ovs_nla_put_actions(actions, rem, skb);
3550
3551out:
3552 if (err) {
3553 nla_nest_cancel(skb, ac_start);
3554 nla_nest_cancel(skb, start);
3555 } else {
3556 nla_nest_end(skb, ac_start);
3557 nla_nest_end(skb, start);
3558 }
3559
3560 return err;
3561}
3562
3563static int clone_action_to_attr(const struct nlattr *attr,
3564 struct sk_buff *skb)
3565{
3566 struct nlattr *start;
3567 int err = 0, rem = nla_len(attr);
3568
3569 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CLONE);
3570 if (!start)
3571 return -EMSGSIZE;
3572
3573 /* Skipping the OVS_CLONE_ATTR_EXEC that is always the first attribute. */
3574 attr = nla_next(nla_data(attr), &rem);
3575 err = ovs_nla_put_actions(attr, rem, skb);
3576
3577 if (err)
3578 nla_nest_cancel(skb, start);
3579 else
3580 nla_nest_end(skb, start);
3581
3582 return err;
3583}
3584
3585static int check_pkt_len_action_to_attr(const struct nlattr *attr,
3586 struct sk_buff *skb)
3587{
3588 struct nlattr *start, *ac_start = NULL;
3589 const struct check_pkt_len_arg *arg;
3590 const struct nlattr *a, *cpl_arg;
3591 int err = 0, rem = nla_len(attr);
3592
3593 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CHECK_PKT_LEN);
3594 if (!start)
3595 return -EMSGSIZE;
3596
3597 /* The first nested attribute in 'attr' is always
3598 * 'OVS_CHECK_PKT_LEN_ATTR_ARG'.
3599 */
3600 cpl_arg = nla_data(attr);
3601 arg = nla_data(cpl_arg);
3602
3603 if (nla_put_u16(skb, OVS_CHECK_PKT_LEN_ATTR_PKT_LEN, arg->pkt_len)) {
3604 err = -EMSGSIZE;
3605 goto out;
3606 }
3607
3608 /* Second nested attribute in 'attr' is always
3609 * 'OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL'.
3610 */
3611 a = nla_next(cpl_arg, &rem);
3612 ac_start = nla_nest_start_noflag(skb,
3613 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL);
3614 if (!ac_start) {
3615 err = -EMSGSIZE;
3616 goto out;
3617 }
3618
3619 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3620 if (err) {
3621 nla_nest_cancel(skb, ac_start);
3622 goto out;
3623 } else {
3624 nla_nest_end(skb, ac_start);
3625 }
3626
3627 /* Third nested attribute in 'attr' is always
3628 * OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER.
3629 */
3630 a = nla_next(a, &rem);
3631 ac_start = nla_nest_start_noflag(skb,
3632 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER);
3633 if (!ac_start) {
3634 err = -EMSGSIZE;
3635 goto out;
3636 }
3637
3638 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3639 if (err) {
3640 nla_nest_cancel(skb, ac_start);
3641 goto out;
3642 } else {
3643 nla_nest_end(skb, ac_start);
3644 }
3645
3646 nla_nest_end(skb, start);
3647 return 0;
3648
3649out:
3650 nla_nest_cancel(skb, start);
3651 return err;
3652}
3653
3654static int dec_ttl_action_to_attr(const struct nlattr *attr,
3655 struct sk_buff *skb)
3656{
3657 struct nlattr *start, *action_start;
3658 const struct nlattr *a;
3659 int err = 0, rem;
3660
3661 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_DEC_TTL);
3662 if (!start)
3663 return -EMSGSIZE;
3664
3665 nla_for_each_attr(a, nla_data(attr), nla_len(attr), rem) {
3666 switch (nla_type(a)) {
3667 case OVS_DEC_TTL_ATTR_ACTION:
3668
3669 action_start = nla_nest_start_noflag(skb, OVS_DEC_TTL_ATTR_ACTION);
3670 if (!action_start) {
3671 err = -EMSGSIZE;
3672 goto out;
3673 }
3674
3675 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3676 if (err)
3677 goto out;
3678
3679 nla_nest_end(skb, action_start);
3680 break;
3681
3682 default:
3683 /* Ignore all other option to be future compatible */
3684 break;
3685 }
3686 }
3687
3688 nla_nest_end(skb, start);
3689 return 0;
3690
3691out:
3692 nla_nest_cancel(skb, start);
3693 return err;
3694}
3695
3696static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
3697{
3698 const struct nlattr *ovs_key = nla_data(a);
3699 int key_type = nla_type(ovs_key);
3700 struct nlattr *start;
3701 int err;
3702
3703 switch (key_type) {
3704 case OVS_KEY_ATTR_TUNNEL_INFO: {
3705 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
3706 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
3707
3708 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3709 if (!start)
3710 return -EMSGSIZE;
3711
3712 err = ip_tun_to_nlattr(skb, &tun_info->key,
3713 ip_tunnel_info_opts(tun_info),
3714 tun_info->options_len,
3715 ip_tunnel_info_af(tun_info), tun_info->mode);
3716 if (err)
3717 return err;
3718 nla_nest_end(skb, start);
3719 break;
3720 }
3721 default:
3722 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
3723 return -EMSGSIZE;
3724 break;
3725 }
3726
3727 return 0;
3728}
3729
3730static int masked_set_action_to_set_action_attr(const struct nlattr *a,
3731 struct sk_buff *skb)
3732{
3733 const struct nlattr *ovs_key = nla_data(a);
3734 struct nlattr *nla;
3735 size_t key_len = nla_len(ovs_key) / 2;
3736
3737 /* Revert the conversion we did from a non-masked set action to
3738 * masked set action.
3739 */
3740 nla = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3741 if (!nla)
3742 return -EMSGSIZE;
3743
3744 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
3745 return -EMSGSIZE;
3746
3747 nla_nest_end(skb, nla);
3748 return 0;
3749}
3750
3751int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
3752{
3753 const struct nlattr *a;
3754 int rem, err;
3755
3756 nla_for_each_attr(a, attr, len, rem) {
3757 int type = nla_type(a);
3758
3759 switch (type) {
3760 case OVS_ACTION_ATTR_SET:
3761 err = set_action_to_attr(a, skb);
3762 if (err)
3763 return err;
3764 break;
3765
3766 case OVS_ACTION_ATTR_SET_TO_MASKED:
3767 err = masked_set_action_to_set_action_attr(a, skb);
3768 if (err)
3769 return err;
3770 break;
3771
3772 case OVS_ACTION_ATTR_SAMPLE:
3773 err = sample_action_to_attr(a, skb);
3774 if (err)
3775 return err;
3776 break;
3777
3778 case OVS_ACTION_ATTR_CT:
3779 err = ovs_ct_action_to_attr(nla_data(a), skb);
3780 if (err)
3781 return err;
3782 break;
3783
3784 case OVS_ACTION_ATTR_CLONE:
3785 err = clone_action_to_attr(a, skb);
3786 if (err)
3787 return err;
3788 break;
3789
3790 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
3791 err = check_pkt_len_action_to_attr(a, skb);
3792 if (err)
3793 return err;
3794 break;
3795
3796 case OVS_ACTION_ATTR_DEC_TTL:
3797 err = dec_ttl_action_to_attr(a, skb);
3798 if (err)
3799 return err;
3800 break;
3801
3802 default:
3803 if (nla_put(skb, type, nla_len(a), nla_data(a)))
3804 return -EMSGSIZE;
3805 break;
3806 }
3807 }
3808
3809 return 0;
3810}