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 case OVS_ACTION_ATTR_PSAMPLE:
68 break;
69
70 case OVS_ACTION_ATTR_CT:
71 case OVS_ACTION_ATTR_CT_CLEAR:
72 case OVS_ACTION_ATTR_HASH:
73 case OVS_ACTION_ATTR_POP_ETH:
74 case OVS_ACTION_ATTR_POP_MPLS:
75 case OVS_ACTION_ATTR_POP_NSH:
76 case OVS_ACTION_ATTR_POP_VLAN:
77 case OVS_ACTION_ATTR_PUSH_ETH:
78 case OVS_ACTION_ATTR_PUSH_MPLS:
79 case OVS_ACTION_ATTR_PUSH_NSH:
80 case OVS_ACTION_ATTR_PUSH_VLAN:
81 case OVS_ACTION_ATTR_SAMPLE:
82 case OVS_ACTION_ATTR_SET:
83 case OVS_ACTION_ATTR_SET_MASKED:
84 case OVS_ACTION_ATTR_METER:
85 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
86 case OVS_ACTION_ATTR_ADD_MPLS:
87 case OVS_ACTION_ATTR_DEC_TTL:
88 default:
89 return true;
90 }
91 }
92 return false;
93}
94
95static void update_range(struct sw_flow_match *match,
96 size_t offset, size_t size, bool is_mask)
97{
98 struct sw_flow_key_range *range;
99 size_t start = rounddown(offset, sizeof(long));
100 size_t end = roundup(offset + size, sizeof(long));
101
102 if (!is_mask)
103 range = &match->range;
104 else
105 range = &match->mask->range;
106
107 if (range->start == range->end) {
108 range->start = start;
109 range->end = end;
110 return;
111 }
112
113 if (range->start > start)
114 range->start = start;
115
116 if (range->end < end)
117 range->end = end;
118}
119
120#define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
121 do { \
122 update_range(match, offsetof(struct sw_flow_key, field), \
123 sizeof((match)->key->field), is_mask); \
124 if (is_mask) \
125 (match)->mask->key.field = value; \
126 else \
127 (match)->key->field = value; \
128 } while (0)
129
130#define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
131 do { \
132 update_range(match, offset, len, is_mask); \
133 if (is_mask) \
134 memcpy((u8 *)&(match)->mask->key + offset, value_p, \
135 len); \
136 else \
137 memcpy((u8 *)(match)->key + offset, value_p, len); \
138 } while (0)
139
140#define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
141 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
142 value_p, len, is_mask)
143
144#define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
145 do { \
146 update_range(match, offsetof(struct sw_flow_key, field), \
147 sizeof((match)->key->field), is_mask); \
148 if (is_mask) \
149 memset((u8 *)&(match)->mask->key.field, value, \
150 sizeof((match)->mask->key.field)); \
151 else \
152 memset((u8 *)&(match)->key->field, value, \
153 sizeof((match)->key->field)); \
154 } while (0)
155
156#define SW_FLOW_KEY_BITMAP_COPY(match, field, value_p, nbits, is_mask) ({ \
157 update_range(match, offsetof(struct sw_flow_key, field), \
158 bitmap_size(nbits), is_mask); \
159 bitmap_copy(is_mask ? (match)->mask->key.field : (match)->key->field, \
160 value_p, nbits); \
161})
162
163static bool match_validate(const struct sw_flow_match *match,
164 u64 key_attrs, u64 mask_attrs, bool log)
165{
166 u64 key_expected = 0;
167 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
168
169 /* The following mask attributes allowed only if they
170 * pass the validation tests. */
171 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
172 | (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)
173 | (1 << OVS_KEY_ATTR_IPV6)
174 | (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)
175 | (1 << OVS_KEY_ATTR_TCP)
176 | (1 << OVS_KEY_ATTR_TCP_FLAGS)
177 | (1 << OVS_KEY_ATTR_UDP)
178 | (1 << OVS_KEY_ATTR_SCTP)
179 | (1 << OVS_KEY_ATTR_ICMP)
180 | (1 << OVS_KEY_ATTR_ICMPV6)
181 | (1 << OVS_KEY_ATTR_ARP)
182 | (1 << OVS_KEY_ATTR_ND)
183 | (1 << OVS_KEY_ATTR_MPLS)
184 | (1 << OVS_KEY_ATTR_NSH));
185
186 /* Always allowed mask fields. */
187 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
188 | (1 << OVS_KEY_ATTR_IN_PORT)
189 | (1 << OVS_KEY_ATTR_ETHERTYPE));
190
191 /* Check key attributes. */
192 if (match->key->eth.type == htons(ETH_P_ARP)
193 || match->key->eth.type == htons(ETH_P_RARP)) {
194 key_expected |= 1 << OVS_KEY_ATTR_ARP;
195 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
196 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
197 }
198
199 if (eth_p_mpls(match->key->eth.type)) {
200 key_expected |= 1 << OVS_KEY_ATTR_MPLS;
201 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
202 mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
203 }
204
205 if (match->key->eth.type == htons(ETH_P_IP)) {
206 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
207 if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
208 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
209 mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4;
210 }
211
212 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
213 if (match->key->ip.proto == IPPROTO_UDP) {
214 key_expected |= 1 << OVS_KEY_ATTR_UDP;
215 if (match->mask && (match->mask->key.ip.proto == 0xff))
216 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
217 }
218
219 if (match->key->ip.proto == IPPROTO_SCTP) {
220 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
221 if (match->mask && (match->mask->key.ip.proto == 0xff))
222 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
223 }
224
225 if (match->key->ip.proto == IPPROTO_TCP) {
226 key_expected |= 1 << OVS_KEY_ATTR_TCP;
227 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
228 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
229 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
230 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
231 }
232 }
233
234 if (match->key->ip.proto == IPPROTO_ICMP) {
235 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
236 if (match->mask && (match->mask->key.ip.proto == 0xff))
237 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
238 }
239 }
240 }
241
242 if (match->key->eth.type == htons(ETH_P_IPV6)) {
243 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
244 if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
245 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
246 mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6;
247 }
248
249 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
250 if (match->key->ip.proto == IPPROTO_UDP) {
251 key_expected |= 1 << OVS_KEY_ATTR_UDP;
252 if (match->mask && (match->mask->key.ip.proto == 0xff))
253 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
254 }
255
256 if (match->key->ip.proto == IPPROTO_SCTP) {
257 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
258 if (match->mask && (match->mask->key.ip.proto == 0xff))
259 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
260 }
261
262 if (match->key->ip.proto == IPPROTO_TCP) {
263 key_expected |= 1 << OVS_KEY_ATTR_TCP;
264 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
265 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
266 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
267 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
268 }
269 }
270
271 if (match->key->ip.proto == IPPROTO_ICMPV6) {
272 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
273 if (match->mask && (match->mask->key.ip.proto == 0xff))
274 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
275
276 if (match->key->tp.src ==
277 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
278 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
279 key_expected |= 1 << OVS_KEY_ATTR_ND;
280 /* Original direction conntrack tuple
281 * uses the same space as the ND fields
282 * in the key, so both are not allowed
283 * at the same time.
284 */
285 mask_allowed &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
286 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
287 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
288 }
289 }
290 }
291 }
292
293 if (match->key->eth.type == htons(ETH_P_NSH)) {
294 key_expected |= 1 << OVS_KEY_ATTR_NSH;
295 if (match->mask &&
296 match->mask->key.eth.type == htons(0xffff)) {
297 mask_allowed |= 1 << OVS_KEY_ATTR_NSH;
298 }
299 }
300
301 if ((key_attrs & key_expected) != key_expected) {
302 /* Key attributes check failed. */
303 OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
304 (unsigned long long)key_attrs,
305 (unsigned long long)key_expected);
306 return false;
307 }
308
309 if ((mask_attrs & mask_allowed) != mask_attrs) {
310 /* Mask attributes check failed. */
311 OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
312 (unsigned long long)mask_attrs,
313 (unsigned long long)mask_allowed);
314 return false;
315 }
316
317 return true;
318}
319
320size_t ovs_tun_key_attr_size(void)
321{
322 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
323 * updating this function.
324 */
325 return nla_total_size_64bit(8) /* OVS_TUNNEL_KEY_ATTR_ID */
326 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
327 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_DST */
328 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
329 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
330 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
331 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
332 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
333 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
334 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS and
335 * OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS is mutually exclusive with
336 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
337 */
338 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
339 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
340}
341
342static size_t ovs_nsh_key_attr_size(void)
343{
344 /* Whenever adding new OVS_NSH_KEY_ FIELDS, we should consider
345 * updating this function.
346 */
347 return nla_total_size(NSH_BASE_HDR_LEN) /* OVS_NSH_KEY_ATTR_BASE */
348 /* OVS_NSH_KEY_ATTR_MD1 and OVS_NSH_KEY_ATTR_MD2 are
349 * mutually exclusive, so the bigger one can cover
350 * the small one.
351 */
352 + nla_total_size(NSH_CTX_HDRS_MAX_LEN);
353}
354
355size_t ovs_key_attr_size(void)
356{
357 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
358 * updating this function.
359 */
360 BUILD_BUG_ON(OVS_KEY_ATTR_MAX != 32);
361
362 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
363 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
364 + ovs_tun_key_attr_size()
365 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
366 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
367 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
368 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
369 + nla_total_size(4) /* OVS_KEY_ATTR_CT_STATE */
370 + nla_total_size(2) /* OVS_KEY_ATTR_CT_ZONE */
371 + nla_total_size(4) /* OVS_KEY_ATTR_CT_MARK */
372 + nla_total_size(16) /* OVS_KEY_ATTR_CT_LABELS */
373 + nla_total_size(40) /* OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6 */
374 + nla_total_size(0) /* OVS_KEY_ATTR_NSH */
375 + ovs_nsh_key_attr_size()
376 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
377 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
378 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
379 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
380 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
381 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
382 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
383 + nla_total_size(28) /* OVS_KEY_ATTR_ND */
384 + nla_total_size(2); /* OVS_KEY_ATTR_IPV6_EXTHDRS */
385}
386
387static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
388 [OVS_VXLAN_EXT_GBP] = { .len = sizeof(u32) },
389};
390
391static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
392 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
393 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
394 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
395 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
396 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
397 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
398 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
399 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
400 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
401 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
402 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_VARIABLE },
403 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED,
404 .next = ovs_vxlan_ext_key_lens },
405 [OVS_TUNNEL_KEY_ATTR_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
406 [OVS_TUNNEL_KEY_ATTR_IPV6_DST] = { .len = sizeof(struct in6_addr) },
407 [OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS] = { .len = OVS_ATTR_VARIABLE },
408 [OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE] = { .len = 0 },
409};
410
411static const struct ovs_len_tbl
412ovs_nsh_key_attr_lens[OVS_NSH_KEY_ATTR_MAX + 1] = {
413 [OVS_NSH_KEY_ATTR_BASE] = { .len = sizeof(struct ovs_nsh_key_base) },
414 [OVS_NSH_KEY_ATTR_MD1] = { .len = sizeof(struct ovs_nsh_key_md1) },
415 [OVS_NSH_KEY_ATTR_MD2] = { .len = OVS_ATTR_VARIABLE },
416};
417
418/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
419static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
420 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
421 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
422 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
423 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
424 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
425 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
426 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
427 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
428 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
429 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
430 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
431 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
432 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
433 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
434 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
435 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
436 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
437 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
438 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
439 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
440 .next = ovs_tunnel_key_lens, },
441 [OVS_KEY_ATTR_MPLS] = { .len = OVS_ATTR_VARIABLE },
442 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u32) },
443 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
444 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
445 [OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
446 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4] = {
447 .len = sizeof(struct ovs_key_ct_tuple_ipv4) },
448 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6] = {
449 .len = sizeof(struct ovs_key_ct_tuple_ipv6) },
450 [OVS_KEY_ATTR_NSH] = { .len = OVS_ATTR_NESTED,
451 .next = ovs_nsh_key_attr_lens, },
452 [OVS_KEY_ATTR_IPV6_EXTHDRS] = {
453 .len = sizeof(struct ovs_key_ipv6_exthdrs) },
454};
455
456static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
457{
458 return expected_len == attr_len ||
459 expected_len == OVS_ATTR_NESTED ||
460 expected_len == OVS_ATTR_VARIABLE;
461}
462
463static bool is_all_zero(const u8 *fp, size_t size)
464{
465 int i;
466
467 if (!fp)
468 return false;
469
470 for (i = 0; i < size; i++)
471 if (fp[i])
472 return false;
473
474 return true;
475}
476
477static int __parse_flow_nlattrs(const struct nlattr *attr,
478 const struct nlattr *a[],
479 u64 *attrsp, bool log, bool nz)
480{
481 const struct nlattr *nla;
482 u64 attrs;
483 int rem;
484
485 attrs = *attrsp;
486 nla_for_each_nested(nla, attr, rem) {
487 u16 type = nla_type(nla);
488 int expected_len;
489
490 if (type > OVS_KEY_ATTR_MAX) {
491 OVS_NLERR(log, "Key type %d is out of range max %d",
492 type, OVS_KEY_ATTR_MAX);
493 return -EINVAL;
494 }
495
496 if (type == OVS_KEY_ATTR_PACKET_TYPE ||
497 type == OVS_KEY_ATTR_ND_EXTENSIONS ||
498 type == OVS_KEY_ATTR_TUNNEL_INFO) {
499 OVS_NLERR(log, "Key type %d is not supported", type);
500 return -EINVAL;
501 }
502
503 if (attrs & (1ULL << type)) {
504 OVS_NLERR(log, "Duplicate key (type %d).", type);
505 return -EINVAL;
506 }
507
508 expected_len = ovs_key_lens[type].len;
509 if (!check_attr_len(nla_len(nla), expected_len)) {
510 OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
511 type, nla_len(nla), expected_len);
512 return -EINVAL;
513 }
514
515 if (!nz || !is_all_zero(nla_data(nla), nla_len(nla))) {
516 attrs |= 1ULL << type;
517 a[type] = nla;
518 }
519 }
520 if (rem) {
521 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
522 return -EINVAL;
523 }
524
525 *attrsp = attrs;
526 return 0;
527}
528
529static int parse_flow_mask_nlattrs(const struct nlattr *attr,
530 const struct nlattr *a[], u64 *attrsp,
531 bool log)
532{
533 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
534}
535
536int parse_flow_nlattrs(const struct nlattr *attr, const struct nlattr *a[],
537 u64 *attrsp, bool log)
538{
539 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
540}
541
542static int genev_tun_opt_from_nlattr(const struct nlattr *a,
543 struct sw_flow_match *match, bool is_mask,
544 bool log)
545{
546 unsigned long opt_key_offset;
547
548 if (nla_len(a) > sizeof(match->key->tun_opts)) {
549 OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
550 nla_len(a), sizeof(match->key->tun_opts));
551 return -EINVAL;
552 }
553
554 if (nla_len(a) % 4 != 0) {
555 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
556 nla_len(a));
557 return -EINVAL;
558 }
559
560 /* We need to record the length of the options passed
561 * down, otherwise packets with the same format but
562 * additional options will be silently matched.
563 */
564 if (!is_mask) {
565 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
566 false);
567 } else {
568 /* This is somewhat unusual because it looks at
569 * both the key and mask while parsing the
570 * attributes (and by extension assumes the key
571 * is parsed first). Normally, we would verify
572 * that each is the correct length and that the
573 * attributes line up in the validate function.
574 * However, that is difficult because this is
575 * variable length and we won't have the
576 * information later.
577 */
578 if (match->key->tun_opts_len != nla_len(a)) {
579 OVS_NLERR(log, "Geneve option len %d != mask len %d",
580 match->key->tun_opts_len, nla_len(a));
581 return -EINVAL;
582 }
583
584 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
585 }
586
587 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
588 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
589 nla_len(a), is_mask);
590 return 0;
591}
592
593static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
594 struct sw_flow_match *match, bool is_mask,
595 bool log)
596{
597 struct nlattr *a;
598 int rem;
599 unsigned long opt_key_offset;
600 struct vxlan_metadata opts;
601
602 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
603
604 memset(&opts, 0, sizeof(opts));
605 nla_for_each_nested(a, attr, rem) {
606 int type = nla_type(a);
607
608 if (type > OVS_VXLAN_EXT_MAX) {
609 OVS_NLERR(log, "VXLAN extension %d out of range max %d",
610 type, OVS_VXLAN_EXT_MAX);
611 return -EINVAL;
612 }
613
614 if (!check_attr_len(nla_len(a),
615 ovs_vxlan_ext_key_lens[type].len)) {
616 OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
617 type, nla_len(a),
618 ovs_vxlan_ext_key_lens[type].len);
619 return -EINVAL;
620 }
621
622 switch (type) {
623 case OVS_VXLAN_EXT_GBP:
624 opts.gbp = nla_get_u32(a);
625 break;
626 default:
627 OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
628 type);
629 return -EINVAL;
630 }
631 }
632 if (rem) {
633 OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
634 rem);
635 return -EINVAL;
636 }
637
638 if (!is_mask)
639 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
640 else
641 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
642
643 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
644 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
645 is_mask);
646 return 0;
647}
648
649static int erspan_tun_opt_from_nlattr(const struct nlattr *a,
650 struct sw_flow_match *match, bool is_mask,
651 bool log)
652{
653 unsigned long opt_key_offset;
654
655 BUILD_BUG_ON(sizeof(struct erspan_metadata) >
656 sizeof(match->key->tun_opts));
657
658 if (nla_len(a) > sizeof(match->key->tun_opts)) {
659 OVS_NLERR(log, "ERSPAN option length err (len %d, max %zu).",
660 nla_len(a), sizeof(match->key->tun_opts));
661 return -EINVAL;
662 }
663
664 if (!is_mask)
665 SW_FLOW_KEY_PUT(match, tun_opts_len,
666 sizeof(struct erspan_metadata), false);
667 else
668 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
669
670 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
671 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
672 nla_len(a), is_mask);
673 return 0;
674}
675
676static int ip_tun_from_nlattr(const struct nlattr *attr,
677 struct sw_flow_match *match, bool is_mask,
678 bool log)
679{
680 bool ttl = false, ipv4 = false, ipv6 = false;
681 IP_TUNNEL_DECLARE_FLAGS(tun_flags) = { };
682 bool info_bridge_mode = false;
683 int opts_type = 0;
684 struct nlattr *a;
685 int rem;
686
687 nla_for_each_nested(a, attr, rem) {
688 int type = nla_type(a);
689 int err;
690
691 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
692 OVS_NLERR(log, "Tunnel attr %d out of range max %d",
693 type, OVS_TUNNEL_KEY_ATTR_MAX);
694 return -EINVAL;
695 }
696
697 if (!check_attr_len(nla_len(a),
698 ovs_tunnel_key_lens[type].len)) {
699 OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
700 type, nla_len(a), ovs_tunnel_key_lens[type].len);
701 return -EINVAL;
702 }
703
704 switch (type) {
705 case OVS_TUNNEL_KEY_ATTR_ID:
706 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
707 nla_get_be64(a), is_mask);
708 __set_bit(IP_TUNNEL_KEY_BIT, tun_flags);
709 break;
710 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
711 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
712 nla_get_in_addr(a), is_mask);
713 ipv4 = true;
714 break;
715 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
716 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
717 nla_get_in_addr(a), is_mask);
718 ipv4 = true;
719 break;
720 case OVS_TUNNEL_KEY_ATTR_IPV6_SRC:
721 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.src,
722 nla_get_in6_addr(a), is_mask);
723 ipv6 = true;
724 break;
725 case OVS_TUNNEL_KEY_ATTR_IPV6_DST:
726 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
727 nla_get_in6_addr(a), is_mask);
728 ipv6 = true;
729 break;
730 case OVS_TUNNEL_KEY_ATTR_TOS:
731 SW_FLOW_KEY_PUT(match, tun_key.tos,
732 nla_get_u8(a), is_mask);
733 break;
734 case OVS_TUNNEL_KEY_ATTR_TTL:
735 SW_FLOW_KEY_PUT(match, tun_key.ttl,
736 nla_get_u8(a), is_mask);
737 ttl = true;
738 break;
739 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
740 __set_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, tun_flags);
741 break;
742 case OVS_TUNNEL_KEY_ATTR_CSUM:
743 __set_bit(IP_TUNNEL_CSUM_BIT, tun_flags);
744 break;
745 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
746 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
747 nla_get_be16(a), is_mask);
748 break;
749 case OVS_TUNNEL_KEY_ATTR_TP_DST:
750 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
751 nla_get_be16(a), is_mask);
752 break;
753 case OVS_TUNNEL_KEY_ATTR_OAM:
754 __set_bit(IP_TUNNEL_OAM_BIT, tun_flags);
755 break;
756 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
757 if (opts_type) {
758 OVS_NLERR(log, "Multiple metadata blocks provided");
759 return -EINVAL;
760 }
761
762 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
763 if (err)
764 return err;
765
766 __set_bit(IP_TUNNEL_GENEVE_OPT_BIT, tun_flags);
767 opts_type = type;
768 break;
769 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
770 if (opts_type) {
771 OVS_NLERR(log, "Multiple metadata blocks provided");
772 return -EINVAL;
773 }
774
775 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
776 if (err)
777 return err;
778
779 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, tun_flags);
780 opts_type = type;
781 break;
782 case OVS_TUNNEL_KEY_ATTR_PAD:
783 break;
784 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
785 if (opts_type) {
786 OVS_NLERR(log, "Multiple metadata blocks provided");
787 return -EINVAL;
788 }
789
790 err = erspan_tun_opt_from_nlattr(a, match, is_mask,
791 log);
792 if (err)
793 return err;
794
795 __set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, tun_flags);
796 opts_type = type;
797 break;
798 case OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE:
799 info_bridge_mode = true;
800 ipv4 = true;
801 break;
802 default:
803 OVS_NLERR(log, "Unknown IP tunnel attribute %d",
804 type);
805 return -EINVAL;
806 }
807 }
808
809 SW_FLOW_KEY_BITMAP_COPY(match, tun_key.tun_flags, tun_flags,
810 __IP_TUNNEL_FLAG_NUM, is_mask);
811 if (is_mask)
812 SW_FLOW_KEY_MEMSET_FIELD(match, tun_proto, 0xff, true);
813 else
814 SW_FLOW_KEY_PUT(match, tun_proto, ipv6 ? AF_INET6 : AF_INET,
815 false);
816
817 if (rem > 0) {
818 OVS_NLERR(log, "IP tunnel attribute has %d unknown bytes.",
819 rem);
820 return -EINVAL;
821 }
822
823 if (ipv4 && ipv6) {
824 OVS_NLERR(log, "Mixed IPv4 and IPv6 tunnel attributes");
825 return -EINVAL;
826 }
827
828 if (!is_mask) {
829 if (!ipv4 && !ipv6) {
830 OVS_NLERR(log, "IP tunnel dst address not specified");
831 return -EINVAL;
832 }
833 if (ipv4) {
834 if (info_bridge_mode) {
835 __clear_bit(IP_TUNNEL_KEY_BIT, tun_flags);
836
837 if (match->key->tun_key.u.ipv4.src ||
838 match->key->tun_key.u.ipv4.dst ||
839 match->key->tun_key.tp_src ||
840 match->key->tun_key.tp_dst ||
841 match->key->tun_key.ttl ||
842 match->key->tun_key.tos ||
843 !ip_tunnel_flags_empty(tun_flags)) {
844 OVS_NLERR(log, "IPv4 tun info is not correct");
845 return -EINVAL;
846 }
847 } else if (!match->key->tun_key.u.ipv4.dst) {
848 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
849 return -EINVAL;
850 }
851 }
852 if (ipv6 && ipv6_addr_any(&match->key->tun_key.u.ipv6.dst)) {
853 OVS_NLERR(log, "IPv6 tunnel dst address is zero");
854 return -EINVAL;
855 }
856
857 if (!ttl && !info_bridge_mode) {
858 OVS_NLERR(log, "IP tunnel TTL not specified.");
859 return -EINVAL;
860 }
861 }
862
863 return opts_type;
864}
865
866static int vxlan_opt_to_nlattr(struct sk_buff *skb,
867 const void *tun_opts, int swkey_tun_opts_len)
868{
869 const struct vxlan_metadata *opts = tun_opts;
870 struct nlattr *nla;
871
872 nla = nla_nest_start_noflag(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
873 if (!nla)
874 return -EMSGSIZE;
875
876 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
877 return -EMSGSIZE;
878
879 nla_nest_end(skb, nla);
880 return 0;
881}
882
883static int __ip_tun_to_nlattr(struct sk_buff *skb,
884 const struct ip_tunnel_key *output,
885 const void *tun_opts, int swkey_tun_opts_len,
886 unsigned short tun_proto, u8 mode)
887{
888 if (test_bit(IP_TUNNEL_KEY_BIT, output->tun_flags) &&
889 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id,
890 OVS_TUNNEL_KEY_ATTR_PAD))
891 return -EMSGSIZE;
892
893 if (mode & IP_TUNNEL_INFO_BRIDGE)
894 return nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE)
895 ? -EMSGSIZE : 0;
896
897 switch (tun_proto) {
898 case AF_INET:
899 if (output->u.ipv4.src &&
900 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
901 output->u.ipv4.src))
902 return -EMSGSIZE;
903 if (output->u.ipv4.dst &&
904 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
905 output->u.ipv4.dst))
906 return -EMSGSIZE;
907 break;
908 case AF_INET6:
909 if (!ipv6_addr_any(&output->u.ipv6.src) &&
910 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_SRC,
911 &output->u.ipv6.src))
912 return -EMSGSIZE;
913 if (!ipv6_addr_any(&output->u.ipv6.dst) &&
914 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_DST,
915 &output->u.ipv6.dst))
916 return -EMSGSIZE;
917 break;
918 }
919 if (output->tos &&
920 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
921 return -EMSGSIZE;
922 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
923 return -EMSGSIZE;
924 if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, output->tun_flags) &&
925 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
926 return -EMSGSIZE;
927 if (test_bit(IP_TUNNEL_CSUM_BIT, output->tun_flags) &&
928 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
929 return -EMSGSIZE;
930 if (output->tp_src &&
931 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
932 return -EMSGSIZE;
933 if (output->tp_dst &&
934 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
935 return -EMSGSIZE;
936 if (test_bit(IP_TUNNEL_OAM_BIT, output->tun_flags) &&
937 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
938 return -EMSGSIZE;
939 if (swkey_tun_opts_len) {
940 if (test_bit(IP_TUNNEL_GENEVE_OPT_BIT, output->tun_flags) &&
941 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
942 swkey_tun_opts_len, tun_opts))
943 return -EMSGSIZE;
944 else if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT,
945 output->tun_flags) &&
946 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
947 return -EMSGSIZE;
948 else if (test_bit(IP_TUNNEL_ERSPAN_OPT_BIT,
949 output->tun_flags) &&
950 nla_put(skb, OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS,
951 swkey_tun_opts_len, tun_opts))
952 return -EMSGSIZE;
953 }
954
955 return 0;
956}
957
958static int ip_tun_to_nlattr(struct sk_buff *skb,
959 const struct ip_tunnel_key *output,
960 const void *tun_opts, int swkey_tun_opts_len,
961 unsigned short tun_proto, u8 mode)
962{
963 struct nlattr *nla;
964 int err;
965
966 nla = nla_nest_start_noflag(skb, OVS_KEY_ATTR_TUNNEL);
967 if (!nla)
968 return -EMSGSIZE;
969
970 err = __ip_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len,
971 tun_proto, mode);
972 if (err)
973 return err;
974
975 nla_nest_end(skb, nla);
976 return 0;
977}
978
979int ovs_nla_put_tunnel_info(struct sk_buff *skb,
980 struct ip_tunnel_info *tun_info)
981{
982 return __ip_tun_to_nlattr(skb, &tun_info->key,
983 ip_tunnel_info_opts(tun_info),
984 tun_info->options_len,
985 ip_tunnel_info_af(tun_info), tun_info->mode);
986}
987
988static int encode_vlan_from_nlattrs(struct sw_flow_match *match,
989 const struct nlattr *a[],
990 bool is_mask, bool inner)
991{
992 __be16 tci = 0;
993 __be16 tpid = 0;
994
995 if (a[OVS_KEY_ATTR_VLAN])
996 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
997
998 if (a[OVS_KEY_ATTR_ETHERTYPE])
999 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1000
1001 if (likely(!inner)) {
1002 SW_FLOW_KEY_PUT(match, eth.vlan.tpid, tpid, is_mask);
1003 SW_FLOW_KEY_PUT(match, eth.vlan.tci, tci, is_mask);
1004 } else {
1005 SW_FLOW_KEY_PUT(match, eth.cvlan.tpid, tpid, is_mask);
1006 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, tci, is_mask);
1007 }
1008 return 0;
1009}
1010
1011static int validate_vlan_from_nlattrs(const struct sw_flow_match *match,
1012 u64 key_attrs, bool inner,
1013 const struct nlattr **a, bool log)
1014{
1015 __be16 tci = 0;
1016
1017 if (!((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
1018 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
1019 eth_type_vlan(nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE])))) {
1020 /* Not a VLAN. */
1021 return 0;
1022 }
1023
1024 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
1025 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
1026 OVS_NLERR(log, "Invalid %s frame", (inner) ? "C-VLAN" : "VLAN");
1027 return -EINVAL;
1028 }
1029
1030 if (a[OVS_KEY_ATTR_VLAN])
1031 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1032
1033 if (!(tci & htons(VLAN_CFI_MASK))) {
1034 if (tci) {
1035 OVS_NLERR(log, "%s TCI does not have VLAN_CFI_MASK bit set.",
1036 (inner) ? "C-VLAN" : "VLAN");
1037 return -EINVAL;
1038 } else if (nla_len(a[OVS_KEY_ATTR_ENCAP])) {
1039 /* Corner case for truncated VLAN header. */
1040 OVS_NLERR(log, "Truncated %s header has non-zero encap attribute.",
1041 (inner) ? "C-VLAN" : "VLAN");
1042 return -EINVAL;
1043 }
1044 }
1045
1046 return 1;
1047}
1048
1049static int validate_vlan_mask_from_nlattrs(const struct sw_flow_match *match,
1050 u64 key_attrs, bool inner,
1051 const struct nlattr **a, bool log)
1052{
1053 __be16 tci = 0;
1054 __be16 tpid = 0;
1055 bool encap_valid = !!(match->key->eth.vlan.tci &
1056 htons(VLAN_CFI_MASK));
1057 bool i_encap_valid = !!(match->key->eth.cvlan.tci &
1058 htons(VLAN_CFI_MASK));
1059
1060 if (!(key_attrs & (1 << OVS_KEY_ATTR_ENCAP))) {
1061 /* Not a VLAN. */
1062 return 0;
1063 }
1064
1065 if ((!inner && !encap_valid) || (inner && !i_encap_valid)) {
1066 OVS_NLERR(log, "Encap mask attribute is set for non-%s frame.",
1067 (inner) ? "C-VLAN" : "VLAN");
1068 return -EINVAL;
1069 }
1070
1071 if (a[OVS_KEY_ATTR_VLAN])
1072 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1073
1074 if (a[OVS_KEY_ATTR_ETHERTYPE])
1075 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1076
1077 if (tpid != htons(0xffff)) {
1078 OVS_NLERR(log, "Must have an exact match on %s TPID (mask=%x).",
1079 (inner) ? "C-VLAN" : "VLAN", ntohs(tpid));
1080 return -EINVAL;
1081 }
1082 if (!(tci & htons(VLAN_CFI_MASK))) {
1083 OVS_NLERR(log, "%s TCI mask does not have exact match for VLAN_CFI_MASK bit.",
1084 (inner) ? "C-VLAN" : "VLAN");
1085 return -EINVAL;
1086 }
1087
1088 return 1;
1089}
1090
1091static int __parse_vlan_from_nlattrs(struct sw_flow_match *match,
1092 u64 *key_attrs, bool inner,
1093 const struct nlattr **a, bool is_mask,
1094 bool log)
1095{
1096 int err;
1097 const struct nlattr *encap;
1098
1099 if (!is_mask)
1100 err = validate_vlan_from_nlattrs(match, *key_attrs, inner,
1101 a, log);
1102 else
1103 err = validate_vlan_mask_from_nlattrs(match, *key_attrs, inner,
1104 a, log);
1105 if (err <= 0)
1106 return err;
1107
1108 err = encode_vlan_from_nlattrs(match, a, is_mask, inner);
1109 if (err)
1110 return err;
1111
1112 *key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1113 *key_attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
1114 *key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1115
1116 encap = a[OVS_KEY_ATTR_ENCAP];
1117
1118 if (!is_mask)
1119 err = parse_flow_nlattrs(encap, a, key_attrs, log);
1120 else
1121 err = parse_flow_mask_nlattrs(encap, a, key_attrs, log);
1122
1123 return err;
1124}
1125
1126static int parse_vlan_from_nlattrs(struct sw_flow_match *match,
1127 u64 *key_attrs, const struct nlattr **a,
1128 bool is_mask, bool log)
1129{
1130 int err;
1131 bool encap_valid = false;
1132
1133 err = __parse_vlan_from_nlattrs(match, key_attrs, false, a,
1134 is_mask, log);
1135 if (err)
1136 return err;
1137
1138 encap_valid = !!(match->key->eth.vlan.tci & htons(VLAN_CFI_MASK));
1139 if (encap_valid) {
1140 err = __parse_vlan_from_nlattrs(match, key_attrs, true, a,
1141 is_mask, log);
1142 if (err)
1143 return err;
1144 }
1145
1146 return 0;
1147}
1148
1149static int parse_eth_type_from_nlattrs(struct sw_flow_match *match,
1150 u64 *attrs, const struct nlattr **a,
1151 bool is_mask, bool log)
1152{
1153 __be16 eth_type;
1154
1155 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1156 if (is_mask) {
1157 /* Always exact match EtherType. */
1158 eth_type = htons(0xffff);
1159 } else if (!eth_proto_is_802_3(eth_type)) {
1160 OVS_NLERR(log, "EtherType %x is less than min %x",
1161 ntohs(eth_type), ETH_P_802_3_MIN);
1162 return -EINVAL;
1163 }
1164
1165 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
1166 *attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1167 return 0;
1168}
1169
1170static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
1171 u64 *attrs, const struct nlattr **a,
1172 bool is_mask, bool log)
1173{
1174 u8 mac_proto = MAC_PROTO_ETHERNET;
1175
1176 if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
1177 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
1178
1179 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
1180 *attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
1181 }
1182
1183 if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
1184 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
1185
1186 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
1187 *attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
1188 }
1189
1190 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
1191 SW_FLOW_KEY_PUT(match, phy.priority,
1192 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
1193 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
1194 }
1195
1196 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
1197 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1198
1199 if (is_mask) {
1200 in_port = 0xffffffff; /* Always exact match in_port. */
1201 } else if (in_port >= DP_MAX_PORTS) {
1202 OVS_NLERR(log, "Port %d exceeds max allowable %d",
1203 in_port, DP_MAX_PORTS);
1204 return -EINVAL;
1205 }
1206
1207 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
1208 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1209 } else if (!is_mask) {
1210 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
1211 }
1212
1213 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1214 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1215
1216 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1217 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1218 }
1219 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1220 if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1221 is_mask, log) < 0)
1222 return -EINVAL;
1223 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1224 }
1225
1226 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
1227 ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
1228 u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
1229
1230 if (ct_state & ~CT_SUPPORTED_MASK) {
1231 OVS_NLERR(log, "ct_state flags %08x unsupported",
1232 ct_state);
1233 return -EINVAL;
1234 }
1235
1236 SW_FLOW_KEY_PUT(match, ct_state, ct_state, is_mask);
1237 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
1238 }
1239 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
1240 ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
1241 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
1242
1243 SW_FLOW_KEY_PUT(match, ct_zone, ct_zone, is_mask);
1244 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
1245 }
1246 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
1247 ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
1248 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
1249
1250 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
1251 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
1252 }
1253 if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
1254 ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
1255 const struct ovs_key_ct_labels *cl;
1256
1257 cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
1258 SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
1259 sizeof(*cl), is_mask);
1260 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
1261 }
1262 if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)) {
1263 const struct ovs_key_ct_tuple_ipv4 *ct;
1264
1265 ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4]);
1266
1267 SW_FLOW_KEY_PUT(match, ipv4.ct_orig.src, ct->ipv4_src, is_mask);
1268 SW_FLOW_KEY_PUT(match, ipv4.ct_orig.dst, ct->ipv4_dst, is_mask);
1269 SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1270 SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1271 SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv4_proto, is_mask);
1272 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4);
1273 }
1274 if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)) {
1275 const struct ovs_key_ct_tuple_ipv6 *ct;
1276
1277 ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6]);
1278
1279 SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.src, &ct->ipv6_src,
1280 sizeof(match->key->ipv6.ct_orig.src),
1281 is_mask);
1282 SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.dst, &ct->ipv6_dst,
1283 sizeof(match->key->ipv6.ct_orig.dst),
1284 is_mask);
1285 SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1286 SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1287 SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv6_proto, is_mask);
1288 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
1289 }
1290
1291 /* For layer 3 packets the Ethernet type is provided
1292 * and treated as metadata but no MAC addresses are provided.
1293 */
1294 if (!(*attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
1295 (*attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)))
1296 mac_proto = MAC_PROTO_NONE;
1297
1298 /* Always exact match mac_proto */
1299 SW_FLOW_KEY_PUT(match, mac_proto, is_mask ? 0xff : mac_proto, is_mask);
1300
1301 if (mac_proto == MAC_PROTO_NONE)
1302 return parse_eth_type_from_nlattrs(match, attrs, a, is_mask,
1303 log);
1304
1305 return 0;
1306}
1307
1308int nsh_hdr_from_nlattr(const struct nlattr *attr,
1309 struct nshhdr *nh, size_t size)
1310{
1311 struct nlattr *a;
1312 int rem;
1313 u8 flags = 0;
1314 u8 ttl = 0;
1315 int mdlen = 0;
1316
1317 /* validate_nsh has check this, so we needn't do duplicate check here
1318 */
1319 if (size < NSH_BASE_HDR_LEN)
1320 return -ENOBUFS;
1321
1322 nla_for_each_nested(a, attr, rem) {
1323 int type = nla_type(a);
1324
1325 switch (type) {
1326 case OVS_NSH_KEY_ATTR_BASE: {
1327 const struct ovs_nsh_key_base *base = nla_data(a);
1328
1329 flags = base->flags;
1330 ttl = base->ttl;
1331 nh->np = base->np;
1332 nh->mdtype = base->mdtype;
1333 nh->path_hdr = base->path_hdr;
1334 break;
1335 }
1336 case OVS_NSH_KEY_ATTR_MD1:
1337 mdlen = nla_len(a);
1338 if (mdlen > size - NSH_BASE_HDR_LEN)
1339 return -ENOBUFS;
1340 memcpy(&nh->md1, nla_data(a), mdlen);
1341 break;
1342
1343 case OVS_NSH_KEY_ATTR_MD2:
1344 mdlen = nla_len(a);
1345 if (mdlen > size - NSH_BASE_HDR_LEN)
1346 return -ENOBUFS;
1347 memcpy(&nh->md2, nla_data(a), mdlen);
1348 break;
1349
1350 default:
1351 return -EINVAL;
1352 }
1353 }
1354
1355 /* nsh header length = NSH_BASE_HDR_LEN + mdlen */
1356 nh->ver_flags_ttl_len = 0;
1357 nsh_set_flags_ttl_len(nh, flags, ttl, NSH_BASE_HDR_LEN + mdlen);
1358
1359 return 0;
1360}
1361
1362int nsh_key_from_nlattr(const struct nlattr *attr,
1363 struct ovs_key_nsh *nsh, struct ovs_key_nsh *nsh_mask)
1364{
1365 struct nlattr *a;
1366 int rem;
1367
1368 /* validate_nsh has check this, so we needn't do duplicate check here
1369 */
1370 nla_for_each_nested(a, attr, rem) {
1371 int type = nla_type(a);
1372
1373 switch (type) {
1374 case OVS_NSH_KEY_ATTR_BASE: {
1375 const struct ovs_nsh_key_base *base = nla_data(a);
1376 const struct ovs_nsh_key_base *base_mask = base + 1;
1377
1378 nsh->base = *base;
1379 nsh_mask->base = *base_mask;
1380 break;
1381 }
1382 case OVS_NSH_KEY_ATTR_MD1: {
1383 const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1384 const struct ovs_nsh_key_md1 *md1_mask = md1 + 1;
1385
1386 memcpy(nsh->context, md1->context, sizeof(*md1));
1387 memcpy(nsh_mask->context, md1_mask->context,
1388 sizeof(*md1_mask));
1389 break;
1390 }
1391 case OVS_NSH_KEY_ATTR_MD2:
1392 /* Not supported yet */
1393 return -ENOTSUPP;
1394 default:
1395 return -EINVAL;
1396 }
1397 }
1398
1399 return 0;
1400}
1401
1402static int nsh_key_put_from_nlattr(const struct nlattr *attr,
1403 struct sw_flow_match *match, bool is_mask,
1404 bool is_push_nsh, bool log)
1405{
1406 struct nlattr *a;
1407 int rem;
1408 bool has_base = false;
1409 bool has_md1 = false;
1410 bool has_md2 = false;
1411 u8 mdtype = 0;
1412 int mdlen = 0;
1413
1414 if (WARN_ON(is_push_nsh && is_mask))
1415 return -EINVAL;
1416
1417 nla_for_each_nested(a, attr, rem) {
1418 int type = nla_type(a);
1419 int i;
1420
1421 if (type > OVS_NSH_KEY_ATTR_MAX) {
1422 OVS_NLERR(log, "nsh attr %d is out of range max %d",
1423 type, OVS_NSH_KEY_ATTR_MAX);
1424 return -EINVAL;
1425 }
1426
1427 if (!check_attr_len(nla_len(a),
1428 ovs_nsh_key_attr_lens[type].len)) {
1429 OVS_NLERR(
1430 log,
1431 "nsh attr %d has unexpected len %d expected %d",
1432 type,
1433 nla_len(a),
1434 ovs_nsh_key_attr_lens[type].len
1435 );
1436 return -EINVAL;
1437 }
1438
1439 switch (type) {
1440 case OVS_NSH_KEY_ATTR_BASE: {
1441 const struct ovs_nsh_key_base *base = nla_data(a);
1442
1443 has_base = true;
1444 mdtype = base->mdtype;
1445 SW_FLOW_KEY_PUT(match, nsh.base.flags,
1446 base->flags, is_mask);
1447 SW_FLOW_KEY_PUT(match, nsh.base.ttl,
1448 base->ttl, is_mask);
1449 SW_FLOW_KEY_PUT(match, nsh.base.mdtype,
1450 base->mdtype, is_mask);
1451 SW_FLOW_KEY_PUT(match, nsh.base.np,
1452 base->np, is_mask);
1453 SW_FLOW_KEY_PUT(match, nsh.base.path_hdr,
1454 base->path_hdr, is_mask);
1455 break;
1456 }
1457 case OVS_NSH_KEY_ATTR_MD1: {
1458 const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1459
1460 has_md1 = true;
1461 for (i = 0; i < NSH_MD1_CONTEXT_SIZE; i++)
1462 SW_FLOW_KEY_PUT(match, nsh.context[i],
1463 md1->context[i], is_mask);
1464 break;
1465 }
1466 case OVS_NSH_KEY_ATTR_MD2:
1467 if (!is_push_nsh) /* Not supported MD type 2 yet */
1468 return -ENOTSUPP;
1469
1470 has_md2 = true;
1471 mdlen = nla_len(a);
1472 if (mdlen > NSH_CTX_HDRS_MAX_LEN || mdlen <= 0) {
1473 OVS_NLERR(
1474 log,
1475 "Invalid MD length %d for MD type %d",
1476 mdlen,
1477 mdtype
1478 );
1479 return -EINVAL;
1480 }
1481 break;
1482 default:
1483 OVS_NLERR(log, "Unknown nsh attribute %d",
1484 type);
1485 return -EINVAL;
1486 }
1487 }
1488
1489 if (rem > 0) {
1490 OVS_NLERR(log, "nsh attribute has %d unknown bytes.", rem);
1491 return -EINVAL;
1492 }
1493
1494 if (has_md1 && has_md2) {
1495 OVS_NLERR(
1496 1,
1497 "invalid nsh attribute: md1 and md2 are exclusive."
1498 );
1499 return -EINVAL;
1500 }
1501
1502 if (!is_mask) {
1503 if ((has_md1 && mdtype != NSH_M_TYPE1) ||
1504 (has_md2 && mdtype != NSH_M_TYPE2)) {
1505 OVS_NLERR(1, "nsh attribute has unmatched MD type %d.",
1506 mdtype);
1507 return -EINVAL;
1508 }
1509
1510 if (is_push_nsh &&
1511 (!has_base || (!has_md1 && !has_md2))) {
1512 OVS_NLERR(
1513 1,
1514 "push_nsh: missing base or metadata attributes"
1515 );
1516 return -EINVAL;
1517 }
1518 }
1519
1520 return 0;
1521}
1522
1523static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
1524 u64 attrs, const struct nlattr **a,
1525 bool is_mask, bool log)
1526{
1527 int err;
1528
1529 err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
1530 if (err)
1531 return err;
1532
1533 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1534 const struct ovs_key_ethernet *eth_key;
1535
1536 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1537 SW_FLOW_KEY_MEMCPY(match, eth.src,
1538 eth_key->eth_src, ETH_ALEN, is_mask);
1539 SW_FLOW_KEY_MEMCPY(match, eth.dst,
1540 eth_key->eth_dst, ETH_ALEN, is_mask);
1541 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1542
1543 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1544 /* VLAN attribute is always parsed before getting here since it
1545 * may occur multiple times.
1546 */
1547 OVS_NLERR(log, "VLAN attribute unexpected.");
1548 return -EINVAL;
1549 }
1550
1551 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1552 err = parse_eth_type_from_nlattrs(match, &attrs, a, is_mask,
1553 log);
1554 if (err)
1555 return err;
1556 } else if (!is_mask) {
1557 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1558 }
1559 } else if (!match->key->eth.type) {
1560 OVS_NLERR(log, "Either Ethernet header or EtherType is required.");
1561 return -EINVAL;
1562 }
1563
1564 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1565 const struct ovs_key_ipv4 *ipv4_key;
1566
1567 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1568 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
1569 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
1570 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
1571 return -EINVAL;
1572 }
1573 SW_FLOW_KEY_PUT(match, ip.proto,
1574 ipv4_key->ipv4_proto, is_mask);
1575 SW_FLOW_KEY_PUT(match, ip.tos,
1576 ipv4_key->ipv4_tos, is_mask);
1577 SW_FLOW_KEY_PUT(match, ip.ttl,
1578 ipv4_key->ipv4_ttl, is_mask);
1579 SW_FLOW_KEY_PUT(match, ip.frag,
1580 ipv4_key->ipv4_frag, is_mask);
1581 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1582 ipv4_key->ipv4_src, is_mask);
1583 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1584 ipv4_key->ipv4_dst, is_mask);
1585 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1586 }
1587
1588 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1589 const struct ovs_key_ipv6 *ipv6_key;
1590
1591 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1592 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
1593 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
1594 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
1595 return -EINVAL;
1596 }
1597
1598 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
1599 OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x)",
1600 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
1601 return -EINVAL;
1602 }
1603
1604 SW_FLOW_KEY_PUT(match, ipv6.label,
1605 ipv6_key->ipv6_label, is_mask);
1606 SW_FLOW_KEY_PUT(match, ip.proto,
1607 ipv6_key->ipv6_proto, is_mask);
1608 SW_FLOW_KEY_PUT(match, ip.tos,
1609 ipv6_key->ipv6_tclass, is_mask);
1610 SW_FLOW_KEY_PUT(match, ip.ttl,
1611 ipv6_key->ipv6_hlimit, is_mask);
1612 SW_FLOW_KEY_PUT(match, ip.frag,
1613 ipv6_key->ipv6_frag, is_mask);
1614 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1615 ipv6_key->ipv6_src,
1616 sizeof(match->key->ipv6.addr.src),
1617 is_mask);
1618 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1619 ipv6_key->ipv6_dst,
1620 sizeof(match->key->ipv6.addr.dst),
1621 is_mask);
1622
1623 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1624 }
1625
1626 if (attrs & (1ULL << OVS_KEY_ATTR_IPV6_EXTHDRS)) {
1627 const struct ovs_key_ipv6_exthdrs *ipv6_exthdrs_key;
1628
1629 ipv6_exthdrs_key = nla_data(a[OVS_KEY_ATTR_IPV6_EXTHDRS]);
1630
1631 SW_FLOW_KEY_PUT(match, ipv6.exthdrs,
1632 ipv6_exthdrs_key->hdrs, is_mask);
1633
1634 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV6_EXTHDRS);
1635 }
1636
1637 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1638 const struct ovs_key_arp *arp_key;
1639
1640 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1641 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1642 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
1643 arp_key->arp_op);
1644 return -EINVAL;
1645 }
1646
1647 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1648 arp_key->arp_sip, is_mask);
1649 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1650 arp_key->arp_tip, is_mask);
1651 SW_FLOW_KEY_PUT(match, ip.proto,
1652 ntohs(arp_key->arp_op), is_mask);
1653 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1654 arp_key->arp_sha, ETH_ALEN, is_mask);
1655 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1656 arp_key->arp_tha, ETH_ALEN, is_mask);
1657
1658 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1659 }
1660
1661 if (attrs & (1 << OVS_KEY_ATTR_NSH)) {
1662 if (nsh_key_put_from_nlattr(a[OVS_KEY_ATTR_NSH], match,
1663 is_mask, false, log) < 0)
1664 return -EINVAL;
1665 attrs &= ~(1 << OVS_KEY_ATTR_NSH);
1666 }
1667
1668 if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1669 const struct ovs_key_mpls *mpls_key;
1670 u32 hdr_len;
1671 u32 label_count, label_count_mask, i;
1672
1673 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1674 hdr_len = nla_len(a[OVS_KEY_ATTR_MPLS]);
1675 label_count = hdr_len / sizeof(struct ovs_key_mpls);
1676
1677 if (label_count == 0 || label_count > MPLS_LABEL_DEPTH ||
1678 hdr_len % sizeof(struct ovs_key_mpls))
1679 return -EINVAL;
1680
1681 label_count_mask = GENMASK(label_count - 1, 0);
1682
1683 for (i = 0 ; i < label_count; i++)
1684 SW_FLOW_KEY_PUT(match, mpls.lse[i],
1685 mpls_key[i].mpls_lse, is_mask);
1686
1687 SW_FLOW_KEY_PUT(match, mpls.num_labels_mask,
1688 label_count_mask, is_mask);
1689
1690 attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1691 }
1692
1693 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1694 const struct ovs_key_tcp *tcp_key;
1695
1696 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1697 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1698 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1699 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1700 }
1701
1702 if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1703 SW_FLOW_KEY_PUT(match, tp.flags,
1704 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1705 is_mask);
1706 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1707 }
1708
1709 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1710 const struct ovs_key_udp *udp_key;
1711
1712 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1713 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1714 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1715 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1716 }
1717
1718 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1719 const struct ovs_key_sctp *sctp_key;
1720
1721 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1722 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1723 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1724 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1725 }
1726
1727 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1728 const struct ovs_key_icmp *icmp_key;
1729
1730 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1731 SW_FLOW_KEY_PUT(match, tp.src,
1732 htons(icmp_key->icmp_type), is_mask);
1733 SW_FLOW_KEY_PUT(match, tp.dst,
1734 htons(icmp_key->icmp_code), is_mask);
1735 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1736 }
1737
1738 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1739 const struct ovs_key_icmpv6 *icmpv6_key;
1740
1741 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1742 SW_FLOW_KEY_PUT(match, tp.src,
1743 htons(icmpv6_key->icmpv6_type), is_mask);
1744 SW_FLOW_KEY_PUT(match, tp.dst,
1745 htons(icmpv6_key->icmpv6_code), is_mask);
1746 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1747 }
1748
1749 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1750 const struct ovs_key_nd *nd_key;
1751
1752 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1753 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1754 nd_key->nd_target,
1755 sizeof(match->key->ipv6.nd.target),
1756 is_mask);
1757 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1758 nd_key->nd_sll, ETH_ALEN, is_mask);
1759 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1760 nd_key->nd_tll, ETH_ALEN, is_mask);
1761 attrs &= ~(1 << OVS_KEY_ATTR_ND);
1762 }
1763
1764 if (attrs != 0) {
1765 OVS_NLERR(log, "Unknown key attributes %llx",
1766 (unsigned long long)attrs);
1767 return -EINVAL;
1768 }
1769
1770 return 0;
1771}
1772
1773static void nlattr_set(struct nlattr *attr, u8 val,
1774 const struct ovs_len_tbl *tbl)
1775{
1776 struct nlattr *nla;
1777 int rem;
1778
1779 /* The nlattr stream should already have been validated */
1780 nla_for_each_nested(nla, attr, rem) {
1781 if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED)
1782 nlattr_set(nla, val, tbl[nla_type(nla)].next ? : tbl);
1783 else
1784 memset(nla_data(nla), val, nla_len(nla));
1785
1786 if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1787 *(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1788 }
1789}
1790
1791static void mask_set_nlattr(struct nlattr *attr, u8 val)
1792{
1793 nlattr_set(attr, val, ovs_key_lens);
1794}
1795
1796/**
1797 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1798 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1799 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1800 * does not include any don't care bit.
1801 * @net: Used to determine per-namespace field support.
1802 * @match: receives the extracted flow match information.
1803 * @nla_key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1804 * sequence. The fields should of the packet that triggered the creation
1805 * of this flow.
1806 * @nla_mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_*
1807 * Netlink attribute specifies the mask field of the wildcarded flow.
1808 * @log: Boolean to allow kernel error logging. Normally true, but when
1809 * probing for feature compatibility this should be passed in as false to
1810 * suppress unnecessary error logging.
1811 */
1812int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1813 const struct nlattr *nla_key,
1814 const struct nlattr *nla_mask,
1815 bool log)
1816{
1817 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1818 struct nlattr *newmask = NULL;
1819 u64 key_attrs = 0;
1820 u64 mask_attrs = 0;
1821 int err;
1822
1823 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1824 if (err)
1825 return err;
1826
1827 err = parse_vlan_from_nlattrs(match, &key_attrs, a, false, log);
1828 if (err)
1829 return err;
1830
1831 err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1832 if (err)
1833 return err;
1834
1835 if (match->mask) {
1836 if (!nla_mask) {
1837 /* Create an exact match mask. We need to set to 0xff
1838 * all the 'match->mask' fields that have been touched
1839 * in 'match->key'. We cannot simply memset
1840 * 'match->mask', because padding bytes and fields not
1841 * specified in 'match->key' should be left to 0.
1842 * Instead, we use a stream of netlink attributes,
1843 * copied from 'key' and set to 0xff.
1844 * ovs_key_from_nlattrs() will take care of filling
1845 * 'match->mask' appropriately.
1846 */
1847 newmask = kmemdup(nla_key,
1848 nla_total_size(nla_len(nla_key)),
1849 GFP_KERNEL);
1850 if (!newmask)
1851 return -ENOMEM;
1852
1853 mask_set_nlattr(newmask, 0xff);
1854
1855 /* The userspace does not send tunnel attributes that
1856 * are 0, but we should not wildcard them nonetheless.
1857 */
1858 if (match->key->tun_proto)
1859 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1860 0xff, true);
1861
1862 nla_mask = newmask;
1863 }
1864
1865 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1866 if (err)
1867 goto free_newmask;
1868
1869 /* Always match on tci. */
1870 SW_FLOW_KEY_PUT(match, eth.vlan.tci, htons(0xffff), true);
1871 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, htons(0xffff), true);
1872
1873 err = parse_vlan_from_nlattrs(match, &mask_attrs, a, true, log);
1874 if (err)
1875 goto free_newmask;
1876
1877 err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1878 log);
1879 if (err)
1880 goto free_newmask;
1881 }
1882
1883 if (!match_validate(match, key_attrs, mask_attrs, log))
1884 err = -EINVAL;
1885
1886free_newmask:
1887 kfree(newmask);
1888 return err;
1889}
1890
1891static size_t get_ufid_len(const struct nlattr *attr, bool log)
1892{
1893 size_t len;
1894
1895 if (!attr)
1896 return 0;
1897
1898 len = nla_len(attr);
1899 if (len < 1 || len > MAX_UFID_LENGTH) {
1900 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1901 nla_len(attr), MAX_UFID_LENGTH);
1902 return 0;
1903 }
1904
1905 return len;
1906}
1907
1908/* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1909 * or false otherwise.
1910 */
1911bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1912 bool log)
1913{
1914 sfid->ufid_len = get_ufid_len(attr, log);
1915 if (sfid->ufid_len)
1916 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1917
1918 return sfid->ufid_len;
1919}
1920
1921int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1922 const struct sw_flow_key *key, bool log)
1923{
1924 struct sw_flow_key *new_key;
1925
1926 if (ovs_nla_get_ufid(sfid, ufid, log))
1927 return 0;
1928
1929 /* If UFID was not provided, use unmasked key. */
1930 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1931 if (!new_key)
1932 return -ENOMEM;
1933 memcpy(new_key, key, sizeof(*key));
1934 sfid->unmasked_key = new_key;
1935
1936 return 0;
1937}
1938
1939u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1940{
1941 return nla_get_u32_default(attr, 0);
1942}
1943
1944/**
1945 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1946 * @net: Network namespace.
1947 * @key: Receives extracted in_port, priority, tun_key, skb_mark and conntrack
1948 * metadata.
1949 * @a: Array of netlink attributes holding parsed %OVS_KEY_ATTR_* Netlink
1950 * attributes.
1951 * @attrs: Bit mask for the netlink attributes included in @a.
1952 * @log: Boolean to allow kernel error logging. Normally true, but when
1953 * probing for feature compatibility this should be passed in as false to
1954 * suppress unnecessary error logging.
1955 *
1956 * This parses a series of Netlink attributes that form a flow key, which must
1957 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1958 * get the metadata, that is, the parts of the flow key that cannot be
1959 * extracted from the packet itself.
1960 *
1961 * This must be called before the packet key fields are filled in 'key'.
1962 */
1963
1964int ovs_nla_get_flow_metadata(struct net *net,
1965 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1],
1966 u64 attrs, struct sw_flow_key *key, bool log)
1967{
1968 struct sw_flow_match match;
1969
1970 memset(&match, 0, sizeof(match));
1971 match.key = key;
1972
1973 key->ct_state = 0;
1974 key->ct_zone = 0;
1975 key->ct_orig_proto = 0;
1976 memset(&key->ct, 0, sizeof(key->ct));
1977 memset(&key->ipv4.ct_orig, 0, sizeof(key->ipv4.ct_orig));
1978 memset(&key->ipv6.ct_orig, 0, sizeof(key->ipv6.ct_orig));
1979
1980 key->phy.in_port = DP_MAX_PORTS;
1981
1982 return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1983}
1984
1985static int ovs_nla_put_vlan(struct sk_buff *skb, const struct vlan_head *vh,
1986 bool is_mask)
1987{
1988 __be16 eth_type = !is_mask ? vh->tpid : htons(0xffff);
1989
1990 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1991 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, vh->tci))
1992 return -EMSGSIZE;
1993 return 0;
1994}
1995
1996static int nsh_key_to_nlattr(const struct ovs_key_nsh *nsh, bool is_mask,
1997 struct sk_buff *skb)
1998{
1999 struct nlattr *start;
2000
2001 start = nla_nest_start_noflag(skb, OVS_KEY_ATTR_NSH);
2002 if (!start)
2003 return -EMSGSIZE;
2004
2005 if (nla_put(skb, OVS_NSH_KEY_ATTR_BASE, sizeof(nsh->base), &nsh->base))
2006 goto nla_put_failure;
2007
2008 if (is_mask || nsh->base.mdtype == NSH_M_TYPE1) {
2009 if (nla_put(skb, OVS_NSH_KEY_ATTR_MD1,
2010 sizeof(nsh->context), nsh->context))
2011 goto nla_put_failure;
2012 }
2013
2014 /* Don't support MD type 2 yet */
2015
2016 nla_nest_end(skb, start);
2017
2018 return 0;
2019
2020nla_put_failure:
2021 return -EMSGSIZE;
2022}
2023
2024static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
2025 const struct sw_flow_key *output, bool is_mask,
2026 struct sk_buff *skb)
2027{
2028 struct ovs_key_ethernet *eth_key;
2029 struct nlattr *nla;
2030 struct nlattr *encap = NULL;
2031 struct nlattr *in_encap = NULL;
2032
2033 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
2034 goto nla_put_failure;
2035
2036 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
2037 goto nla_put_failure;
2038
2039 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
2040 goto nla_put_failure;
2041
2042 if ((swkey->tun_proto || is_mask)) {
2043 const void *opts = NULL;
2044
2045 if (ip_tunnel_is_options_present(output->tun_key.tun_flags))
2046 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
2047
2048 if (ip_tun_to_nlattr(skb, &output->tun_key, opts,
2049 swkey->tun_opts_len, swkey->tun_proto, 0))
2050 goto nla_put_failure;
2051 }
2052
2053 if (swkey->phy.in_port == DP_MAX_PORTS) {
2054 if (is_mask && (output->phy.in_port == 0xffff))
2055 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
2056 goto nla_put_failure;
2057 } else {
2058 u16 upper_u16;
2059 upper_u16 = !is_mask ? 0 : 0xffff;
2060
2061 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
2062 (upper_u16 << 16) | output->phy.in_port))
2063 goto nla_put_failure;
2064 }
2065
2066 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
2067 goto nla_put_failure;
2068
2069 if (ovs_ct_put_key(swkey, output, skb))
2070 goto nla_put_failure;
2071
2072 if (ovs_key_mac_proto(swkey) == MAC_PROTO_ETHERNET) {
2073 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
2074 if (!nla)
2075 goto nla_put_failure;
2076
2077 eth_key = nla_data(nla);
2078 ether_addr_copy(eth_key->eth_src, output->eth.src);
2079 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
2080
2081 if (swkey->eth.vlan.tci || eth_type_vlan(swkey->eth.type)) {
2082 if (ovs_nla_put_vlan(skb, &output->eth.vlan, is_mask))
2083 goto nla_put_failure;
2084 encap = nla_nest_start_noflag(skb, OVS_KEY_ATTR_ENCAP);
2085 if (!swkey->eth.vlan.tci)
2086 goto unencap;
2087
2088 if (swkey->eth.cvlan.tci || eth_type_vlan(swkey->eth.type)) {
2089 if (ovs_nla_put_vlan(skb, &output->eth.cvlan, is_mask))
2090 goto nla_put_failure;
2091 in_encap = nla_nest_start_noflag(skb,
2092 OVS_KEY_ATTR_ENCAP);
2093 if (!swkey->eth.cvlan.tci)
2094 goto unencap;
2095 }
2096 }
2097
2098 if (swkey->eth.type == htons(ETH_P_802_2)) {
2099 /*
2100 * Ethertype 802.2 is represented in the netlink with omitted
2101 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
2102 * 0xffff in the mask attribute. Ethertype can also
2103 * be wildcarded.
2104 */
2105 if (is_mask && output->eth.type)
2106 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
2107 output->eth.type))
2108 goto nla_put_failure;
2109 goto unencap;
2110 }
2111 }
2112
2113 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
2114 goto nla_put_failure;
2115
2116 if (eth_type_vlan(swkey->eth.type)) {
2117 /* There are 3 VLAN tags, we don't know anything about the rest
2118 * of the packet, so truncate here.
2119 */
2120 WARN_ON_ONCE(!(encap && in_encap));
2121 goto unencap;
2122 }
2123
2124 if (swkey->eth.type == htons(ETH_P_IP)) {
2125 struct ovs_key_ipv4 *ipv4_key;
2126
2127 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
2128 if (!nla)
2129 goto nla_put_failure;
2130 ipv4_key = nla_data(nla);
2131 ipv4_key->ipv4_src = output->ipv4.addr.src;
2132 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
2133 ipv4_key->ipv4_proto = output->ip.proto;
2134 ipv4_key->ipv4_tos = output->ip.tos;
2135 ipv4_key->ipv4_ttl = output->ip.ttl;
2136 ipv4_key->ipv4_frag = output->ip.frag;
2137 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
2138 struct ovs_key_ipv6 *ipv6_key;
2139 struct ovs_key_ipv6_exthdrs *ipv6_exthdrs_key;
2140
2141 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
2142 if (!nla)
2143 goto nla_put_failure;
2144 ipv6_key = nla_data(nla);
2145 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
2146 sizeof(ipv6_key->ipv6_src));
2147 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
2148 sizeof(ipv6_key->ipv6_dst));
2149 ipv6_key->ipv6_label = output->ipv6.label;
2150 ipv6_key->ipv6_proto = output->ip.proto;
2151 ipv6_key->ipv6_tclass = output->ip.tos;
2152 ipv6_key->ipv6_hlimit = output->ip.ttl;
2153 ipv6_key->ipv6_frag = output->ip.frag;
2154
2155 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6_EXTHDRS,
2156 sizeof(*ipv6_exthdrs_key));
2157 if (!nla)
2158 goto nla_put_failure;
2159 ipv6_exthdrs_key = nla_data(nla);
2160 ipv6_exthdrs_key->hdrs = output->ipv6.exthdrs;
2161 } else if (swkey->eth.type == htons(ETH_P_NSH)) {
2162 if (nsh_key_to_nlattr(&output->nsh, is_mask, skb))
2163 goto nla_put_failure;
2164 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
2165 swkey->eth.type == htons(ETH_P_RARP)) {
2166 struct ovs_key_arp *arp_key;
2167
2168 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
2169 if (!nla)
2170 goto nla_put_failure;
2171 arp_key = nla_data(nla);
2172 memset(arp_key, 0, sizeof(struct ovs_key_arp));
2173 arp_key->arp_sip = output->ipv4.addr.src;
2174 arp_key->arp_tip = output->ipv4.addr.dst;
2175 arp_key->arp_op = htons(output->ip.proto);
2176 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
2177 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
2178 } else if (eth_p_mpls(swkey->eth.type)) {
2179 u8 i, num_labels;
2180 struct ovs_key_mpls *mpls_key;
2181
2182 num_labels = hweight_long(output->mpls.num_labels_mask);
2183 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS,
2184 num_labels * sizeof(*mpls_key));
2185 if (!nla)
2186 goto nla_put_failure;
2187
2188 mpls_key = nla_data(nla);
2189 for (i = 0; i < num_labels; i++)
2190 mpls_key[i].mpls_lse = output->mpls.lse[i];
2191 }
2192
2193 if ((swkey->eth.type == htons(ETH_P_IP) ||
2194 swkey->eth.type == htons(ETH_P_IPV6)) &&
2195 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
2196
2197 if (swkey->ip.proto == IPPROTO_TCP) {
2198 struct ovs_key_tcp *tcp_key;
2199
2200 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
2201 if (!nla)
2202 goto nla_put_failure;
2203 tcp_key = nla_data(nla);
2204 tcp_key->tcp_src = output->tp.src;
2205 tcp_key->tcp_dst = output->tp.dst;
2206 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
2207 output->tp.flags))
2208 goto nla_put_failure;
2209 } else if (swkey->ip.proto == IPPROTO_UDP) {
2210 struct ovs_key_udp *udp_key;
2211
2212 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
2213 if (!nla)
2214 goto nla_put_failure;
2215 udp_key = nla_data(nla);
2216 udp_key->udp_src = output->tp.src;
2217 udp_key->udp_dst = output->tp.dst;
2218 } else if (swkey->ip.proto == IPPROTO_SCTP) {
2219 struct ovs_key_sctp *sctp_key;
2220
2221 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
2222 if (!nla)
2223 goto nla_put_failure;
2224 sctp_key = nla_data(nla);
2225 sctp_key->sctp_src = output->tp.src;
2226 sctp_key->sctp_dst = output->tp.dst;
2227 } else if (swkey->eth.type == htons(ETH_P_IP) &&
2228 swkey->ip.proto == IPPROTO_ICMP) {
2229 struct ovs_key_icmp *icmp_key;
2230
2231 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
2232 if (!nla)
2233 goto nla_put_failure;
2234 icmp_key = nla_data(nla);
2235 icmp_key->icmp_type = ntohs(output->tp.src);
2236 icmp_key->icmp_code = ntohs(output->tp.dst);
2237 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
2238 swkey->ip.proto == IPPROTO_ICMPV6) {
2239 struct ovs_key_icmpv6 *icmpv6_key;
2240
2241 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
2242 sizeof(*icmpv6_key));
2243 if (!nla)
2244 goto nla_put_failure;
2245 icmpv6_key = nla_data(nla);
2246 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
2247 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
2248
2249 if (swkey->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
2250 swkey->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
2251 struct ovs_key_nd *nd_key;
2252
2253 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
2254 if (!nla)
2255 goto nla_put_failure;
2256 nd_key = nla_data(nla);
2257 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
2258 sizeof(nd_key->nd_target));
2259 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
2260 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
2261 }
2262 }
2263 }
2264
2265unencap:
2266 if (in_encap)
2267 nla_nest_end(skb, in_encap);
2268 if (encap)
2269 nla_nest_end(skb, encap);
2270
2271 return 0;
2272
2273nla_put_failure:
2274 return -EMSGSIZE;
2275}
2276
2277int ovs_nla_put_key(const struct sw_flow_key *swkey,
2278 const struct sw_flow_key *output, int attr, bool is_mask,
2279 struct sk_buff *skb)
2280{
2281 int err;
2282 struct nlattr *nla;
2283
2284 nla = nla_nest_start_noflag(skb, attr);
2285 if (!nla)
2286 return -EMSGSIZE;
2287 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
2288 if (err)
2289 return err;
2290 nla_nest_end(skb, nla);
2291
2292 return 0;
2293}
2294
2295/* Called with ovs_mutex or RCU read lock. */
2296int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
2297{
2298 if (ovs_identifier_is_ufid(&flow->id))
2299 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
2300 flow->id.ufid);
2301
2302 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
2303 OVS_FLOW_ATTR_KEY, false, skb);
2304}
2305
2306/* Called with ovs_mutex or RCU read lock. */
2307int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
2308{
2309 return ovs_nla_put_key(&flow->key, &flow->key,
2310 OVS_FLOW_ATTR_KEY, false, skb);
2311}
2312
2313/* Called with ovs_mutex or RCU read lock. */
2314int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
2315{
2316 return ovs_nla_put_key(&flow->key, &flow->mask->key,
2317 OVS_FLOW_ATTR_MASK, true, skb);
2318}
2319
2320#define MAX_ACTIONS_BUFSIZE (32 * 1024)
2321
2322static struct sw_flow_actions *nla_alloc_flow_actions(int size)
2323{
2324 struct sw_flow_actions *sfa;
2325
2326 WARN_ON_ONCE(size > MAX_ACTIONS_BUFSIZE);
2327
2328 sfa = kmalloc(kmalloc_size_roundup(sizeof(*sfa) + size), GFP_KERNEL);
2329 if (!sfa)
2330 return ERR_PTR(-ENOMEM);
2331
2332 sfa->actions_len = 0;
2333 return sfa;
2334}
2335
2336static void ovs_nla_free_nested_actions(const struct nlattr *actions, int len);
2337
2338static void ovs_nla_free_check_pkt_len_action(const struct nlattr *action)
2339{
2340 const struct nlattr *a;
2341 int rem;
2342
2343 nla_for_each_nested(a, action, rem) {
2344 switch (nla_type(a)) {
2345 case OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL:
2346 case OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER:
2347 ovs_nla_free_nested_actions(nla_data(a), nla_len(a));
2348 break;
2349 }
2350 }
2351}
2352
2353static void ovs_nla_free_clone_action(const struct nlattr *action)
2354{
2355 const struct nlattr *a = nla_data(action);
2356 int rem = nla_len(action);
2357
2358 switch (nla_type(a)) {
2359 case OVS_CLONE_ATTR_EXEC:
2360 /* The real list of actions follows this attribute. */
2361 a = nla_next(a, &rem);
2362 ovs_nla_free_nested_actions(a, rem);
2363 break;
2364 }
2365}
2366
2367static void ovs_nla_free_dec_ttl_action(const struct nlattr *action)
2368{
2369 const struct nlattr *a = nla_data(action);
2370
2371 switch (nla_type(a)) {
2372 case OVS_DEC_TTL_ATTR_ACTION:
2373 ovs_nla_free_nested_actions(nla_data(a), nla_len(a));
2374 break;
2375 }
2376}
2377
2378static void ovs_nla_free_sample_action(const struct nlattr *action)
2379{
2380 const struct nlattr *a = nla_data(action);
2381 int rem = nla_len(action);
2382
2383 switch (nla_type(a)) {
2384 case OVS_SAMPLE_ATTR_ARG:
2385 /* The real list of actions follows this attribute. */
2386 a = nla_next(a, &rem);
2387 ovs_nla_free_nested_actions(a, rem);
2388 break;
2389 }
2390}
2391
2392static void ovs_nla_free_set_action(const struct nlattr *a)
2393{
2394 const struct nlattr *ovs_key = nla_data(a);
2395 struct ovs_tunnel_info *ovs_tun;
2396
2397 switch (nla_type(ovs_key)) {
2398 case OVS_KEY_ATTR_TUNNEL_INFO:
2399 ovs_tun = nla_data(ovs_key);
2400 dst_release((struct dst_entry *)ovs_tun->tun_dst);
2401 break;
2402 }
2403}
2404
2405static void ovs_nla_free_nested_actions(const struct nlattr *actions, int len)
2406{
2407 const struct nlattr *a;
2408 int rem;
2409
2410 /* Whenever new actions are added, the need to update this
2411 * function should be considered.
2412 */
2413 BUILD_BUG_ON(OVS_ACTION_ATTR_MAX != 25);
2414
2415 if (!actions)
2416 return;
2417
2418 nla_for_each_attr(a, actions, len, rem) {
2419 switch (nla_type(a)) {
2420 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
2421 ovs_nla_free_check_pkt_len_action(a);
2422 break;
2423
2424 case OVS_ACTION_ATTR_CLONE:
2425 ovs_nla_free_clone_action(a);
2426 break;
2427
2428 case OVS_ACTION_ATTR_CT:
2429 ovs_ct_free_action(a);
2430 break;
2431
2432 case OVS_ACTION_ATTR_DEC_TTL:
2433 ovs_nla_free_dec_ttl_action(a);
2434 break;
2435
2436 case OVS_ACTION_ATTR_SAMPLE:
2437 ovs_nla_free_sample_action(a);
2438 break;
2439
2440 case OVS_ACTION_ATTR_SET:
2441 ovs_nla_free_set_action(a);
2442 break;
2443 }
2444 }
2445}
2446
2447void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
2448{
2449 if (!sf_acts)
2450 return;
2451
2452 ovs_nla_free_nested_actions(sf_acts->actions, sf_acts->actions_len);
2453 kfree(sf_acts);
2454}
2455
2456static void __ovs_nla_free_flow_actions(struct rcu_head *head)
2457{
2458 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
2459}
2460
2461/* Schedules 'sf_acts' to be freed after the next RCU grace period.
2462 * The caller must hold rcu_read_lock for this to be sensible. */
2463void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
2464{
2465 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
2466}
2467
2468static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
2469 int attr_len, bool log)
2470{
2471
2472 struct sw_flow_actions *acts;
2473 int new_acts_size;
2474 size_t req_size = NLA_ALIGN(attr_len);
2475 int next_offset = offsetof(struct sw_flow_actions, actions) +
2476 (*sfa)->actions_len;
2477
2478 if (req_size <= (ksize(*sfa) - next_offset))
2479 goto out;
2480
2481 new_acts_size = max(next_offset + req_size, ksize(*sfa) * 2);
2482
2483 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
2484 if ((next_offset + req_size) > MAX_ACTIONS_BUFSIZE) {
2485 OVS_NLERR(log, "Flow action size exceeds max %u",
2486 MAX_ACTIONS_BUFSIZE);
2487 return ERR_PTR(-EMSGSIZE);
2488 }
2489 new_acts_size = MAX_ACTIONS_BUFSIZE;
2490 }
2491
2492 acts = nla_alloc_flow_actions(new_acts_size);
2493 if (IS_ERR(acts))
2494 return ERR_CAST(acts);
2495
2496 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
2497 acts->actions_len = (*sfa)->actions_len;
2498 acts->orig_len = (*sfa)->orig_len;
2499 kfree(*sfa);
2500 *sfa = acts;
2501
2502out:
2503 (*sfa)->actions_len += req_size;
2504 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
2505}
2506
2507static struct nlattr *__add_action(struct sw_flow_actions **sfa,
2508 int attrtype, void *data, int len, bool log)
2509{
2510 struct nlattr *a;
2511
2512 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
2513 if (IS_ERR(a))
2514 return a;
2515
2516 a->nla_type = attrtype;
2517 a->nla_len = nla_attr_size(len);
2518
2519 if (data)
2520 memcpy(nla_data(a), data, len);
2521 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
2522
2523 return a;
2524}
2525
2526int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
2527 int len, bool log)
2528{
2529 struct nlattr *a;
2530
2531 a = __add_action(sfa, attrtype, data, len, log);
2532
2533 return PTR_ERR_OR_ZERO(a);
2534}
2535
2536static inline int add_nested_action_start(struct sw_flow_actions **sfa,
2537 int attrtype, bool log)
2538{
2539 int used = (*sfa)->actions_len;
2540 int err;
2541
2542 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
2543 if (err)
2544 return err;
2545
2546 return used;
2547}
2548
2549static inline void add_nested_action_end(struct sw_flow_actions *sfa,
2550 int st_offset)
2551{
2552 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
2553 st_offset);
2554
2555 a->nla_len = sfa->actions_len - st_offset;
2556}
2557
2558static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2559 const struct sw_flow_key *key,
2560 struct sw_flow_actions **sfa,
2561 __be16 eth_type, __be16 vlan_tci,
2562 u32 mpls_label_count, bool log,
2563 u32 depth);
2564
2565static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
2566 const struct sw_flow_key *key,
2567 struct sw_flow_actions **sfa,
2568 __be16 eth_type, __be16 vlan_tci,
2569 u32 mpls_label_count, bool log, bool last,
2570 u32 depth)
2571{
2572 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
2573 const struct nlattr *probability, *actions;
2574 const struct nlattr *a;
2575 int rem, start, err;
2576 struct sample_arg arg;
2577
2578 memset(attrs, 0, sizeof(attrs));
2579 nla_for_each_nested(a, attr, rem) {
2580 int type = nla_type(a);
2581 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
2582 return -EINVAL;
2583 attrs[type] = a;
2584 }
2585 if (rem)
2586 return -EINVAL;
2587
2588 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
2589 if (!probability || nla_len(probability) != sizeof(u32))
2590 return -EINVAL;
2591
2592 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
2593 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
2594 return -EINVAL;
2595
2596 /* validation done, copy sample action. */
2597 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
2598 if (start < 0)
2599 return start;
2600
2601 /* When both skb and flow may be changed, put the sample
2602 * into a deferred fifo. On the other hand, if only skb
2603 * may be modified, the actions can be executed in place.
2604 *
2605 * Do this analysis at the flow installation time.
2606 * Set 'clone_action->exec' to true if the actions can be
2607 * executed without being deferred.
2608 *
2609 * If the sample is the last action, it can always be excuted
2610 * rather than deferred.
2611 */
2612 arg.exec = last || !actions_may_change_flow(actions);
2613 arg.probability = nla_get_u32(probability);
2614
2615 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_ARG, &arg, sizeof(arg),
2616 log);
2617 if (err)
2618 return err;
2619
2620 err = __ovs_nla_copy_actions(net, actions, key, sfa,
2621 eth_type, vlan_tci, mpls_label_count, log,
2622 depth + 1);
2623
2624 if (err)
2625 return err;
2626
2627 add_nested_action_end(*sfa, start);
2628
2629 return 0;
2630}
2631
2632static int validate_and_copy_dec_ttl(struct net *net,
2633 const struct nlattr *attr,
2634 const struct sw_flow_key *key,
2635 struct sw_flow_actions **sfa,
2636 __be16 eth_type, __be16 vlan_tci,
2637 u32 mpls_label_count, bool log,
2638 u32 depth)
2639{
2640 const struct nlattr *attrs[OVS_DEC_TTL_ATTR_MAX + 1];
2641 int start, action_start, err, rem;
2642 const struct nlattr *a, *actions;
2643
2644 memset(attrs, 0, sizeof(attrs));
2645 nla_for_each_nested(a, attr, rem) {
2646 int type = nla_type(a);
2647
2648 /* Ignore unknown attributes to be future proof. */
2649 if (type > OVS_DEC_TTL_ATTR_MAX)
2650 continue;
2651
2652 if (!type || attrs[type]) {
2653 OVS_NLERR(log, "Duplicate or invalid key (type %d).",
2654 type);
2655 return -EINVAL;
2656 }
2657
2658 attrs[type] = a;
2659 }
2660
2661 if (rem) {
2662 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
2663 return -EINVAL;
2664 }
2665
2666 actions = attrs[OVS_DEC_TTL_ATTR_ACTION];
2667 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN)) {
2668 OVS_NLERR(log, "Missing valid actions attribute.");
2669 return -EINVAL;
2670 }
2671
2672 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_DEC_TTL, log);
2673 if (start < 0)
2674 return start;
2675
2676 action_start = add_nested_action_start(sfa, OVS_DEC_TTL_ATTR_ACTION, log);
2677 if (action_start < 0)
2678 return action_start;
2679
2680 err = __ovs_nla_copy_actions(net, actions, key, sfa, eth_type,
2681 vlan_tci, mpls_label_count, log,
2682 depth + 1);
2683 if (err)
2684 return err;
2685
2686 add_nested_action_end(*sfa, action_start);
2687 add_nested_action_end(*sfa, start);
2688 return 0;
2689}
2690
2691static int validate_and_copy_clone(struct net *net,
2692 const struct nlattr *attr,
2693 const struct sw_flow_key *key,
2694 struct sw_flow_actions **sfa,
2695 __be16 eth_type, __be16 vlan_tci,
2696 u32 mpls_label_count, bool log, bool last,
2697 u32 depth)
2698{
2699 int start, err;
2700 u32 exec;
2701
2702 if (nla_len(attr) && nla_len(attr) < NLA_HDRLEN)
2703 return -EINVAL;
2704
2705 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CLONE, log);
2706 if (start < 0)
2707 return start;
2708
2709 exec = last || !actions_may_change_flow(attr);
2710
2711 err = ovs_nla_add_action(sfa, OVS_CLONE_ATTR_EXEC, &exec,
2712 sizeof(exec), log);
2713 if (err)
2714 return err;
2715
2716 err = __ovs_nla_copy_actions(net, attr, key, sfa,
2717 eth_type, vlan_tci, mpls_label_count, log,
2718 depth + 1);
2719 if (err)
2720 return err;
2721
2722 add_nested_action_end(*sfa, start);
2723
2724 return 0;
2725}
2726
2727void ovs_match_init(struct sw_flow_match *match,
2728 struct sw_flow_key *key,
2729 bool reset_key,
2730 struct sw_flow_mask *mask)
2731{
2732 memset(match, 0, sizeof(*match));
2733 match->key = key;
2734 match->mask = mask;
2735
2736 if (reset_key)
2737 memset(key, 0, sizeof(*key));
2738
2739 if (mask) {
2740 memset(&mask->key, 0, sizeof(mask->key));
2741 mask->range.start = mask->range.end = 0;
2742 }
2743}
2744
2745static int validate_geneve_opts(struct sw_flow_key *key)
2746{
2747 struct geneve_opt *option;
2748 int opts_len = key->tun_opts_len;
2749 bool crit_opt = false;
2750
2751 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2752 while (opts_len > 0) {
2753 int len;
2754
2755 if (opts_len < sizeof(*option))
2756 return -EINVAL;
2757
2758 len = sizeof(*option) + option->length * 4;
2759 if (len > opts_len)
2760 return -EINVAL;
2761
2762 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2763
2764 option = (struct geneve_opt *)((u8 *)option + len);
2765 opts_len -= len;
2766 }
2767
2768 if (crit_opt)
2769 __set_bit(IP_TUNNEL_CRIT_OPT_BIT, key->tun_key.tun_flags);
2770
2771 return 0;
2772}
2773
2774static int validate_and_copy_set_tun(const struct nlattr *attr,
2775 struct sw_flow_actions **sfa, bool log)
2776{
2777 IP_TUNNEL_DECLARE_FLAGS(dst_opt_type) = { };
2778 struct sw_flow_match match;
2779 struct sw_flow_key key;
2780 struct metadata_dst *tun_dst;
2781 struct ip_tunnel_info *tun_info;
2782 struct ovs_tunnel_info *ovs_tun;
2783 struct nlattr *a;
2784 int err = 0, start, opts_type;
2785
2786 ovs_match_init(&match, &key, true, NULL);
2787 opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2788 if (opts_type < 0)
2789 return opts_type;
2790
2791 if (key.tun_opts_len) {
2792 switch (opts_type) {
2793 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2794 err = validate_geneve_opts(&key);
2795 if (err < 0)
2796 return err;
2797
2798 __set_bit(IP_TUNNEL_GENEVE_OPT_BIT, dst_opt_type);
2799 break;
2800 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
2801 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, dst_opt_type);
2802 break;
2803 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
2804 __set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, dst_opt_type);
2805 break;
2806 }
2807 }
2808
2809 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2810 if (start < 0)
2811 return start;
2812
2813 tun_dst = metadata_dst_alloc(key.tun_opts_len, METADATA_IP_TUNNEL,
2814 GFP_KERNEL);
2815
2816 if (!tun_dst)
2817 return -ENOMEM;
2818
2819 err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2820 if (err) {
2821 dst_release((struct dst_entry *)tun_dst);
2822 return err;
2823 }
2824
2825 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2826 sizeof(*ovs_tun), log);
2827 if (IS_ERR(a)) {
2828 dst_release((struct dst_entry *)tun_dst);
2829 return PTR_ERR(a);
2830 }
2831
2832 ovs_tun = nla_data(a);
2833 ovs_tun->tun_dst = tun_dst;
2834
2835 tun_info = &tun_dst->u.tun_info;
2836 tun_info->mode = IP_TUNNEL_INFO_TX;
2837 if (key.tun_proto == AF_INET6)
2838 tun_info->mode |= IP_TUNNEL_INFO_IPV6;
2839 else if (key.tun_proto == AF_INET && key.tun_key.u.ipv4.dst == 0)
2840 tun_info->mode |= IP_TUNNEL_INFO_BRIDGE;
2841 tun_info->key = key.tun_key;
2842
2843 /* We need to store the options in the action itself since
2844 * everything else will go away after flow setup. We can append
2845 * it to tun_info and then point there.
2846 */
2847 ip_tunnel_info_opts_set(tun_info,
2848 TUN_METADATA_OPTS(&key, key.tun_opts_len),
2849 key.tun_opts_len, dst_opt_type);
2850 add_nested_action_end(*sfa, start);
2851
2852 return err;
2853}
2854
2855static bool validate_nsh(const struct nlattr *attr, bool is_mask,
2856 bool is_push_nsh, bool log)
2857{
2858 struct sw_flow_match match;
2859 struct sw_flow_key key;
2860 int ret = 0;
2861
2862 ovs_match_init(&match, &key, true, NULL);
2863 ret = nsh_key_put_from_nlattr(attr, &match, is_mask,
2864 is_push_nsh, log);
2865 return !ret;
2866}
2867
2868/* Return false if there are any non-masked bits set.
2869 * Mask follows data immediately, before any netlink padding.
2870 */
2871static bool validate_masked(u8 *data, int len)
2872{
2873 u8 *mask = data + len;
2874
2875 while (len--)
2876 if (*data++ & ~*mask++)
2877 return false;
2878
2879 return true;
2880}
2881
2882static int validate_set(const struct nlattr *a,
2883 const struct sw_flow_key *flow_key,
2884 struct sw_flow_actions **sfa, bool *skip_copy,
2885 u8 mac_proto, __be16 eth_type, bool masked, bool log)
2886{
2887 const struct nlattr *ovs_key = nla_data(a);
2888 int key_type = nla_type(ovs_key);
2889 size_t key_len;
2890
2891 /* There can be only one key in a action */
2892 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2893 return -EINVAL;
2894
2895 key_len = nla_len(ovs_key);
2896 if (masked)
2897 key_len /= 2;
2898
2899 if (key_type > OVS_KEY_ATTR_MAX ||
2900 !check_attr_len(key_len, ovs_key_lens[key_type].len))
2901 return -EINVAL;
2902
2903 if (masked && !validate_masked(nla_data(ovs_key), key_len))
2904 return -EINVAL;
2905
2906 switch (key_type) {
2907 case OVS_KEY_ATTR_PRIORITY:
2908 case OVS_KEY_ATTR_SKB_MARK:
2909 case OVS_KEY_ATTR_CT_MARK:
2910 case OVS_KEY_ATTR_CT_LABELS:
2911 break;
2912
2913 case OVS_KEY_ATTR_ETHERNET:
2914 if (mac_proto != MAC_PROTO_ETHERNET)
2915 return -EINVAL;
2916 break;
2917
2918 case OVS_KEY_ATTR_TUNNEL: {
2919 int err;
2920
2921 if (masked)
2922 return -EINVAL; /* Masked tunnel set not supported. */
2923
2924 *skip_copy = true;
2925 err = validate_and_copy_set_tun(a, sfa, log);
2926 if (err)
2927 return err;
2928 break;
2929 }
2930 case OVS_KEY_ATTR_IPV4: {
2931 const struct ovs_key_ipv4 *ipv4_key;
2932
2933 if (eth_type != htons(ETH_P_IP))
2934 return -EINVAL;
2935
2936 ipv4_key = nla_data(ovs_key);
2937
2938 if (masked) {
2939 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2940
2941 /* Non-writeable fields. */
2942 if (mask->ipv4_proto || mask->ipv4_frag)
2943 return -EINVAL;
2944 } else {
2945 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2946 return -EINVAL;
2947
2948 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2949 return -EINVAL;
2950 }
2951 break;
2952 }
2953 case OVS_KEY_ATTR_IPV6: {
2954 const struct ovs_key_ipv6 *ipv6_key;
2955
2956 if (eth_type != htons(ETH_P_IPV6))
2957 return -EINVAL;
2958
2959 ipv6_key = nla_data(ovs_key);
2960
2961 if (masked) {
2962 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2963
2964 /* Non-writeable fields. */
2965 if (mask->ipv6_proto || mask->ipv6_frag)
2966 return -EINVAL;
2967
2968 /* Invalid bits in the flow label mask? */
2969 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2970 return -EINVAL;
2971 } else {
2972 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2973 return -EINVAL;
2974
2975 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2976 return -EINVAL;
2977 }
2978 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2979 return -EINVAL;
2980
2981 break;
2982 }
2983 case OVS_KEY_ATTR_TCP:
2984 if ((eth_type != htons(ETH_P_IP) &&
2985 eth_type != htons(ETH_P_IPV6)) ||
2986 flow_key->ip.proto != IPPROTO_TCP)
2987 return -EINVAL;
2988
2989 break;
2990
2991 case OVS_KEY_ATTR_UDP:
2992 if ((eth_type != htons(ETH_P_IP) &&
2993 eth_type != htons(ETH_P_IPV6)) ||
2994 flow_key->ip.proto != IPPROTO_UDP)
2995 return -EINVAL;
2996
2997 break;
2998
2999 case OVS_KEY_ATTR_MPLS:
3000 if (!eth_p_mpls(eth_type))
3001 return -EINVAL;
3002 break;
3003
3004 case OVS_KEY_ATTR_SCTP:
3005 if ((eth_type != htons(ETH_P_IP) &&
3006 eth_type != htons(ETH_P_IPV6)) ||
3007 flow_key->ip.proto != IPPROTO_SCTP)
3008 return -EINVAL;
3009
3010 break;
3011
3012 case OVS_KEY_ATTR_NSH:
3013 if (eth_type != htons(ETH_P_NSH))
3014 return -EINVAL;
3015 if (!validate_nsh(nla_data(a), masked, false, log))
3016 return -EINVAL;
3017 break;
3018
3019 default:
3020 return -EINVAL;
3021 }
3022
3023 /* Convert non-masked non-tunnel set actions to masked set actions. */
3024 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
3025 int start, len = key_len * 2;
3026 struct nlattr *at;
3027
3028 *skip_copy = true;
3029
3030 start = add_nested_action_start(sfa,
3031 OVS_ACTION_ATTR_SET_TO_MASKED,
3032 log);
3033 if (start < 0)
3034 return start;
3035
3036 at = __add_action(sfa, key_type, NULL, len, log);
3037 if (IS_ERR(at))
3038 return PTR_ERR(at);
3039
3040 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
3041 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
3042 /* Clear non-writeable bits from otherwise writeable fields. */
3043 if (key_type == OVS_KEY_ATTR_IPV6) {
3044 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
3045
3046 mask->ipv6_label &= htonl(0x000FFFFF);
3047 }
3048 add_nested_action_end(*sfa, start);
3049 }
3050
3051 return 0;
3052}
3053
3054static int validate_userspace(const struct nlattr *attr)
3055{
3056 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
3057 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
3058 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
3059 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
3060 };
3061 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
3062 int error;
3063
3064 error = nla_parse_nested_deprecated(a, OVS_USERSPACE_ATTR_MAX, attr,
3065 userspace_policy, NULL);
3066 if (error)
3067 return error;
3068
3069 if (!a[OVS_USERSPACE_ATTR_PID] ||
3070 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
3071 return -EINVAL;
3072
3073 return 0;
3074}
3075
3076static const struct nla_policy cpl_policy[OVS_CHECK_PKT_LEN_ATTR_MAX + 1] = {
3077 [OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] = {.type = NLA_U16 },
3078 [OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER] = {.type = NLA_NESTED },
3079 [OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL] = {.type = NLA_NESTED },
3080};
3081
3082static int validate_and_copy_check_pkt_len(struct net *net,
3083 const struct nlattr *attr,
3084 const struct sw_flow_key *key,
3085 struct sw_flow_actions **sfa,
3086 __be16 eth_type, __be16 vlan_tci,
3087 u32 mpls_label_count,
3088 bool log, bool last, u32 depth)
3089{
3090 const struct nlattr *acts_if_greater, *acts_if_lesser_eq;
3091 struct nlattr *a[OVS_CHECK_PKT_LEN_ATTR_MAX + 1];
3092 struct check_pkt_len_arg arg;
3093 int nested_acts_start;
3094 int start, err;
3095
3096 err = nla_parse_deprecated_strict(a, OVS_CHECK_PKT_LEN_ATTR_MAX,
3097 nla_data(attr), nla_len(attr),
3098 cpl_policy, NULL);
3099 if (err)
3100 return err;
3101
3102 if (!a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] ||
3103 !nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]))
3104 return -EINVAL;
3105
3106 acts_if_lesser_eq = a[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL];
3107 acts_if_greater = a[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER];
3108
3109 /* Both the nested action should be present. */
3110 if (!acts_if_greater || !acts_if_lesser_eq)
3111 return -EINVAL;
3112
3113 /* validation done, copy the nested actions. */
3114 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CHECK_PKT_LEN,
3115 log);
3116 if (start < 0)
3117 return start;
3118
3119 arg.pkt_len = nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]);
3120 arg.exec_for_lesser_equal =
3121 last || !actions_may_change_flow(acts_if_lesser_eq);
3122 arg.exec_for_greater =
3123 last || !actions_may_change_flow(acts_if_greater);
3124
3125 err = ovs_nla_add_action(sfa, OVS_CHECK_PKT_LEN_ATTR_ARG, &arg,
3126 sizeof(arg), log);
3127 if (err)
3128 return err;
3129
3130 nested_acts_start = add_nested_action_start(sfa,
3131 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL, log);
3132 if (nested_acts_start < 0)
3133 return nested_acts_start;
3134
3135 err = __ovs_nla_copy_actions(net, acts_if_lesser_eq, 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
3144 nested_acts_start = add_nested_action_start(sfa,
3145 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER, log);
3146 if (nested_acts_start < 0)
3147 return nested_acts_start;
3148
3149 err = __ovs_nla_copy_actions(net, acts_if_greater, key, sfa,
3150 eth_type, vlan_tci, mpls_label_count, log,
3151 depth + 1);
3152
3153 if (err)
3154 return err;
3155
3156 add_nested_action_end(*sfa, nested_acts_start);
3157 add_nested_action_end(*sfa, start);
3158 return 0;
3159}
3160
3161static int validate_psample(const struct nlattr *attr)
3162{
3163 static const struct nla_policy policy[OVS_PSAMPLE_ATTR_MAX + 1] = {
3164 [OVS_PSAMPLE_ATTR_GROUP] = { .type = NLA_U32 },
3165 [OVS_PSAMPLE_ATTR_COOKIE] = {
3166 .type = NLA_BINARY,
3167 .len = OVS_PSAMPLE_COOKIE_MAX_SIZE,
3168 },
3169 };
3170 struct nlattr *a[OVS_PSAMPLE_ATTR_MAX + 1];
3171 int err;
3172
3173 if (!IS_ENABLED(CONFIG_PSAMPLE))
3174 return -EOPNOTSUPP;
3175
3176 err = nla_parse_nested(a, OVS_PSAMPLE_ATTR_MAX, attr, policy, NULL);
3177 if (err)
3178 return err;
3179
3180 return a[OVS_PSAMPLE_ATTR_GROUP] ? 0 : -EINVAL;
3181}
3182
3183static int copy_action(const struct nlattr *from,
3184 struct sw_flow_actions **sfa, bool log)
3185{
3186 int totlen = NLA_ALIGN(from->nla_len);
3187 struct nlattr *to;
3188
3189 to = reserve_sfa_size(sfa, from->nla_len, log);
3190 if (IS_ERR(to))
3191 return PTR_ERR(to);
3192
3193 memcpy(to, from, totlen);
3194 return 0;
3195}
3196
3197static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3198 const struct sw_flow_key *key,
3199 struct sw_flow_actions **sfa,
3200 __be16 eth_type, __be16 vlan_tci,
3201 u32 mpls_label_count, bool log,
3202 u32 depth)
3203{
3204 u8 mac_proto = ovs_key_mac_proto(key);
3205 const struct nlattr *a;
3206 int rem, err;
3207
3208 if (depth > OVS_COPY_ACTIONS_MAX_DEPTH)
3209 return -EOVERFLOW;
3210
3211 nla_for_each_nested(a, attr, rem) {
3212 /* Expected argument lengths, (u32)-1 for variable length. */
3213 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
3214 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
3215 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
3216 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
3217 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
3218 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
3219 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
3220 [OVS_ACTION_ATTR_POP_VLAN] = 0,
3221 [OVS_ACTION_ATTR_SET] = (u32)-1,
3222 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
3223 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
3224 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
3225 [OVS_ACTION_ATTR_CT] = (u32)-1,
3226 [OVS_ACTION_ATTR_CT_CLEAR] = 0,
3227 [OVS_ACTION_ATTR_TRUNC] = sizeof(struct ovs_action_trunc),
3228 [OVS_ACTION_ATTR_PUSH_ETH] = sizeof(struct ovs_action_push_eth),
3229 [OVS_ACTION_ATTR_POP_ETH] = 0,
3230 [OVS_ACTION_ATTR_PUSH_NSH] = (u32)-1,
3231 [OVS_ACTION_ATTR_POP_NSH] = 0,
3232 [OVS_ACTION_ATTR_METER] = sizeof(u32),
3233 [OVS_ACTION_ATTR_CLONE] = (u32)-1,
3234 [OVS_ACTION_ATTR_CHECK_PKT_LEN] = (u32)-1,
3235 [OVS_ACTION_ATTR_ADD_MPLS] = sizeof(struct ovs_action_add_mpls),
3236 [OVS_ACTION_ATTR_DEC_TTL] = (u32)-1,
3237 [OVS_ACTION_ATTR_DROP] = sizeof(u32),
3238 [OVS_ACTION_ATTR_PSAMPLE] = (u32)-1,
3239 };
3240 const struct ovs_action_push_vlan *vlan;
3241 int type = nla_type(a);
3242 bool skip_copy;
3243
3244 if (type > OVS_ACTION_ATTR_MAX ||
3245 (action_lens[type] != nla_len(a) &&
3246 action_lens[type] != (u32)-1))
3247 return -EINVAL;
3248
3249 skip_copy = false;
3250 switch (type) {
3251 case OVS_ACTION_ATTR_UNSPEC:
3252 return -EINVAL;
3253
3254 case OVS_ACTION_ATTR_USERSPACE:
3255 err = validate_userspace(a);
3256 if (err)
3257 return err;
3258 break;
3259
3260 case OVS_ACTION_ATTR_OUTPUT:
3261 if (nla_get_u32(a) >= DP_MAX_PORTS)
3262 return -EINVAL;
3263 break;
3264
3265 case OVS_ACTION_ATTR_TRUNC: {
3266 const struct ovs_action_trunc *trunc = nla_data(a);
3267
3268 if (trunc->max_len < ETH_HLEN)
3269 return -EINVAL;
3270 break;
3271 }
3272
3273 case OVS_ACTION_ATTR_HASH: {
3274 const struct ovs_action_hash *act_hash = nla_data(a);
3275
3276 switch (act_hash->hash_alg) {
3277 case OVS_HASH_ALG_L4:
3278 fallthrough;
3279 case OVS_HASH_ALG_SYM_L4:
3280 break;
3281 default:
3282 return -EINVAL;
3283 }
3284
3285 break;
3286 }
3287
3288 case OVS_ACTION_ATTR_POP_VLAN:
3289 if (mac_proto != MAC_PROTO_ETHERNET)
3290 return -EINVAL;
3291 vlan_tci = htons(0);
3292 break;
3293
3294 case OVS_ACTION_ATTR_PUSH_VLAN:
3295 if (mac_proto != MAC_PROTO_ETHERNET)
3296 return -EINVAL;
3297 vlan = nla_data(a);
3298 if (!eth_type_vlan(vlan->vlan_tpid))
3299 return -EINVAL;
3300 if (!(vlan->vlan_tci & htons(VLAN_CFI_MASK)))
3301 return -EINVAL;
3302 vlan_tci = vlan->vlan_tci;
3303 break;
3304
3305 case OVS_ACTION_ATTR_RECIRC:
3306 break;
3307
3308 case OVS_ACTION_ATTR_ADD_MPLS: {
3309 const struct ovs_action_add_mpls *mpls = nla_data(a);
3310
3311 if (!eth_p_mpls(mpls->mpls_ethertype))
3312 return -EINVAL;
3313
3314 if (mpls->tun_flags & OVS_MPLS_L3_TUNNEL_FLAG_MASK) {
3315 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3316 (eth_type != htons(ETH_P_IP) &&
3317 eth_type != htons(ETH_P_IPV6) &&
3318 eth_type != htons(ETH_P_ARP) &&
3319 eth_type != htons(ETH_P_RARP) &&
3320 !eth_p_mpls(eth_type)))
3321 return -EINVAL;
3322 mpls_label_count++;
3323 } else {
3324 if (mac_proto == MAC_PROTO_ETHERNET) {
3325 mpls_label_count = 1;
3326 mac_proto = MAC_PROTO_NONE;
3327 } else {
3328 mpls_label_count++;
3329 }
3330 }
3331 eth_type = mpls->mpls_ethertype;
3332 break;
3333 }
3334
3335 case OVS_ACTION_ATTR_PUSH_MPLS: {
3336 const struct ovs_action_push_mpls *mpls = nla_data(a);
3337
3338 if (!eth_p_mpls(mpls->mpls_ethertype))
3339 return -EINVAL;
3340 /* Prohibit push MPLS other than to a white list
3341 * for packets that have a known tag order.
3342 */
3343 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3344 (eth_type != htons(ETH_P_IP) &&
3345 eth_type != htons(ETH_P_IPV6) &&
3346 eth_type != htons(ETH_P_ARP) &&
3347 eth_type != htons(ETH_P_RARP) &&
3348 !eth_p_mpls(eth_type)))
3349 return -EINVAL;
3350 eth_type = mpls->mpls_ethertype;
3351 mpls_label_count++;
3352 break;
3353 }
3354
3355 case OVS_ACTION_ATTR_POP_MPLS: {
3356 __be16 proto;
3357 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3358 !eth_p_mpls(eth_type))
3359 return -EINVAL;
3360
3361 /* Disallow subsequent L2.5+ set actions and mpls_pop
3362 * actions once the last MPLS label in the packet is
3363 * popped as there is no check here to ensure that
3364 * the new eth type is valid and thus set actions could
3365 * write off the end of the packet or otherwise corrupt
3366 * it.
3367 *
3368 * Support for these actions is planned using packet
3369 * recirculation.
3370 */
3371 proto = nla_get_be16(a);
3372
3373 if (proto == htons(ETH_P_TEB) &&
3374 mac_proto != MAC_PROTO_NONE)
3375 return -EINVAL;
3376
3377 mpls_label_count--;
3378
3379 if (!eth_p_mpls(proto) || !mpls_label_count)
3380 eth_type = htons(0);
3381 else
3382 eth_type = proto;
3383
3384 break;
3385 }
3386
3387 case OVS_ACTION_ATTR_SET:
3388 err = validate_set(a, key, sfa,
3389 &skip_copy, mac_proto, eth_type,
3390 false, log);
3391 if (err)
3392 return err;
3393 break;
3394
3395 case OVS_ACTION_ATTR_SET_MASKED:
3396 err = validate_set(a, key, sfa,
3397 &skip_copy, mac_proto, eth_type,
3398 true, log);
3399 if (err)
3400 return err;
3401 break;
3402
3403 case OVS_ACTION_ATTR_SAMPLE: {
3404 bool last = nla_is_last(a, rem);
3405
3406 err = validate_and_copy_sample(net, a, key, sfa,
3407 eth_type, vlan_tci,
3408 mpls_label_count,
3409 log, last, depth);
3410 if (err)
3411 return err;
3412 skip_copy = true;
3413 break;
3414 }
3415
3416 case OVS_ACTION_ATTR_CT:
3417 err = ovs_ct_copy_action(net, a, key, sfa, log);
3418 if (err)
3419 return err;
3420 skip_copy = true;
3421 break;
3422
3423 case OVS_ACTION_ATTR_CT_CLEAR:
3424 break;
3425
3426 case OVS_ACTION_ATTR_PUSH_ETH:
3427 /* Disallow pushing an Ethernet header if one
3428 * is already present */
3429 if (mac_proto != MAC_PROTO_NONE)
3430 return -EINVAL;
3431 mac_proto = MAC_PROTO_ETHERNET;
3432 break;
3433
3434 case OVS_ACTION_ATTR_POP_ETH:
3435 if (mac_proto != MAC_PROTO_ETHERNET)
3436 return -EINVAL;
3437 if (vlan_tci & htons(VLAN_CFI_MASK))
3438 return -EINVAL;
3439 mac_proto = MAC_PROTO_NONE;
3440 break;
3441
3442 case OVS_ACTION_ATTR_PUSH_NSH:
3443 if (mac_proto != MAC_PROTO_ETHERNET) {
3444 u8 next_proto;
3445
3446 next_proto = tun_p_from_eth_p(eth_type);
3447 if (!next_proto)
3448 return -EINVAL;
3449 }
3450 mac_proto = MAC_PROTO_NONE;
3451 if (!validate_nsh(nla_data(a), false, true, true))
3452 return -EINVAL;
3453 break;
3454
3455 case OVS_ACTION_ATTR_POP_NSH: {
3456 __be16 inner_proto;
3457
3458 if (eth_type != htons(ETH_P_NSH))
3459 return -EINVAL;
3460 inner_proto = tun_p_to_eth_p(key->nsh.base.np);
3461 if (!inner_proto)
3462 return -EINVAL;
3463 if (key->nsh.base.np == TUN_P_ETHERNET)
3464 mac_proto = MAC_PROTO_ETHERNET;
3465 else
3466 mac_proto = MAC_PROTO_NONE;
3467 break;
3468 }
3469
3470 case OVS_ACTION_ATTR_METER:
3471 /* Non-existent meters are simply ignored. */
3472 break;
3473
3474 case OVS_ACTION_ATTR_CLONE: {
3475 bool last = nla_is_last(a, rem);
3476
3477 err = validate_and_copy_clone(net, a, key, sfa,
3478 eth_type, vlan_tci,
3479 mpls_label_count,
3480 log, last, depth);
3481 if (err)
3482 return err;
3483 skip_copy = true;
3484 break;
3485 }
3486
3487 case OVS_ACTION_ATTR_CHECK_PKT_LEN: {
3488 bool last = nla_is_last(a, rem);
3489
3490 err = validate_and_copy_check_pkt_len(net, a, key, sfa,
3491 eth_type,
3492 vlan_tci,
3493 mpls_label_count,
3494 log, last,
3495 depth);
3496 if (err)
3497 return err;
3498 skip_copy = true;
3499 break;
3500 }
3501
3502 case OVS_ACTION_ATTR_DEC_TTL:
3503 err = validate_and_copy_dec_ttl(net, a, key, sfa,
3504 eth_type, vlan_tci,
3505 mpls_label_count, log,
3506 depth);
3507 if (err)
3508 return err;
3509 skip_copy = true;
3510 break;
3511
3512 case OVS_ACTION_ATTR_DROP:
3513 if (!nla_is_last(a, rem))
3514 return -EINVAL;
3515 break;
3516
3517 case OVS_ACTION_ATTR_PSAMPLE:
3518 err = validate_psample(a);
3519 if (err)
3520 return err;
3521 break;
3522
3523 default:
3524 OVS_NLERR(log, "Unknown Action type %d", type);
3525 return -EINVAL;
3526 }
3527 if (!skip_copy) {
3528 err = copy_action(a, sfa, log);
3529 if (err)
3530 return err;
3531 }
3532 }
3533
3534 if (rem > 0)
3535 return -EINVAL;
3536
3537 return 0;
3538}
3539
3540/* 'key' must be the masked key. */
3541int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3542 const struct sw_flow_key *key,
3543 struct sw_flow_actions **sfa, bool log)
3544{
3545 int err;
3546 u32 mpls_label_count = 0;
3547
3548 *sfa = nla_alloc_flow_actions(min(nla_len(attr), MAX_ACTIONS_BUFSIZE));
3549 if (IS_ERR(*sfa))
3550 return PTR_ERR(*sfa);
3551
3552 if (eth_p_mpls(key->eth.type))
3553 mpls_label_count = hweight_long(key->mpls.num_labels_mask);
3554
3555 (*sfa)->orig_len = nla_len(attr);
3556 err = __ovs_nla_copy_actions(net, attr, key, sfa, key->eth.type,
3557 key->eth.vlan.tci, mpls_label_count, log,
3558 0);
3559 if (err)
3560 ovs_nla_free_flow_actions(*sfa);
3561
3562 return err;
3563}
3564
3565static int sample_action_to_attr(const struct nlattr *attr,
3566 struct sk_buff *skb)
3567{
3568 struct nlattr *start, *ac_start = NULL, *sample_arg;
3569 int err = 0, rem = nla_len(attr);
3570 const struct sample_arg *arg;
3571 struct nlattr *actions;
3572
3573 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SAMPLE);
3574 if (!start)
3575 return -EMSGSIZE;
3576
3577 sample_arg = nla_data(attr);
3578 arg = nla_data(sample_arg);
3579 actions = nla_next(sample_arg, &rem);
3580
3581 if (nla_put_u32(skb, OVS_SAMPLE_ATTR_PROBABILITY, arg->probability)) {
3582 err = -EMSGSIZE;
3583 goto out;
3584 }
3585
3586 ac_start = nla_nest_start_noflag(skb, OVS_SAMPLE_ATTR_ACTIONS);
3587 if (!ac_start) {
3588 err = -EMSGSIZE;
3589 goto out;
3590 }
3591
3592 err = ovs_nla_put_actions(actions, rem, skb);
3593
3594out:
3595 if (err) {
3596 nla_nest_cancel(skb, ac_start);
3597 nla_nest_cancel(skb, start);
3598 } else {
3599 nla_nest_end(skb, ac_start);
3600 nla_nest_end(skb, start);
3601 }
3602
3603 return err;
3604}
3605
3606static int clone_action_to_attr(const struct nlattr *attr,
3607 struct sk_buff *skb)
3608{
3609 struct nlattr *start;
3610 int err = 0, rem = nla_len(attr);
3611
3612 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CLONE);
3613 if (!start)
3614 return -EMSGSIZE;
3615
3616 /* Skipping the OVS_CLONE_ATTR_EXEC that is always the first attribute. */
3617 attr = nla_next(nla_data(attr), &rem);
3618 err = ovs_nla_put_actions(attr, rem, skb);
3619
3620 if (err)
3621 nla_nest_cancel(skb, start);
3622 else
3623 nla_nest_end(skb, start);
3624
3625 return err;
3626}
3627
3628static int check_pkt_len_action_to_attr(const struct nlattr *attr,
3629 struct sk_buff *skb)
3630{
3631 struct nlattr *start, *ac_start = NULL;
3632 const struct check_pkt_len_arg *arg;
3633 const struct nlattr *a, *cpl_arg;
3634 int err = 0, rem = nla_len(attr);
3635
3636 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CHECK_PKT_LEN);
3637 if (!start)
3638 return -EMSGSIZE;
3639
3640 /* The first nested attribute in 'attr' is always
3641 * 'OVS_CHECK_PKT_LEN_ATTR_ARG'.
3642 */
3643 cpl_arg = nla_data(attr);
3644 arg = nla_data(cpl_arg);
3645
3646 if (nla_put_u16(skb, OVS_CHECK_PKT_LEN_ATTR_PKT_LEN, arg->pkt_len)) {
3647 err = -EMSGSIZE;
3648 goto out;
3649 }
3650
3651 /* Second nested attribute in 'attr' is always
3652 * 'OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL'.
3653 */
3654 a = nla_next(cpl_arg, &rem);
3655 ac_start = nla_nest_start_noflag(skb,
3656 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL);
3657 if (!ac_start) {
3658 err = -EMSGSIZE;
3659 goto out;
3660 }
3661
3662 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3663 if (err) {
3664 nla_nest_cancel(skb, ac_start);
3665 goto out;
3666 } else {
3667 nla_nest_end(skb, ac_start);
3668 }
3669
3670 /* Third nested attribute in 'attr' is always
3671 * OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER.
3672 */
3673 a = nla_next(a, &rem);
3674 ac_start = nla_nest_start_noflag(skb,
3675 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER);
3676 if (!ac_start) {
3677 err = -EMSGSIZE;
3678 goto out;
3679 }
3680
3681 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3682 if (err) {
3683 nla_nest_cancel(skb, ac_start);
3684 goto out;
3685 } else {
3686 nla_nest_end(skb, ac_start);
3687 }
3688
3689 nla_nest_end(skb, start);
3690 return 0;
3691
3692out:
3693 nla_nest_cancel(skb, start);
3694 return err;
3695}
3696
3697static int dec_ttl_action_to_attr(const struct nlattr *attr,
3698 struct sk_buff *skb)
3699{
3700 struct nlattr *start, *action_start;
3701 const struct nlattr *a;
3702 int err = 0, rem;
3703
3704 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_DEC_TTL);
3705 if (!start)
3706 return -EMSGSIZE;
3707
3708 nla_for_each_attr(a, nla_data(attr), nla_len(attr), rem) {
3709 switch (nla_type(a)) {
3710 case OVS_DEC_TTL_ATTR_ACTION:
3711
3712 action_start = nla_nest_start_noflag(skb, OVS_DEC_TTL_ATTR_ACTION);
3713 if (!action_start) {
3714 err = -EMSGSIZE;
3715 goto out;
3716 }
3717
3718 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3719 if (err)
3720 goto out;
3721
3722 nla_nest_end(skb, action_start);
3723 break;
3724
3725 default:
3726 /* Ignore all other option to be future compatible */
3727 break;
3728 }
3729 }
3730
3731 nla_nest_end(skb, start);
3732 return 0;
3733
3734out:
3735 nla_nest_cancel(skb, start);
3736 return err;
3737}
3738
3739static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
3740{
3741 const struct nlattr *ovs_key = nla_data(a);
3742 int key_type = nla_type(ovs_key);
3743 struct nlattr *start;
3744 int err;
3745
3746 switch (key_type) {
3747 case OVS_KEY_ATTR_TUNNEL_INFO: {
3748 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
3749 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
3750
3751 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3752 if (!start)
3753 return -EMSGSIZE;
3754
3755 err = ip_tun_to_nlattr(skb, &tun_info->key,
3756 ip_tunnel_info_opts(tun_info),
3757 tun_info->options_len,
3758 ip_tunnel_info_af(tun_info), tun_info->mode);
3759 if (err)
3760 return err;
3761 nla_nest_end(skb, start);
3762 break;
3763 }
3764 default:
3765 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
3766 return -EMSGSIZE;
3767 break;
3768 }
3769
3770 return 0;
3771}
3772
3773static int masked_set_action_to_set_action_attr(const struct nlattr *a,
3774 struct sk_buff *skb)
3775{
3776 const struct nlattr *ovs_key = nla_data(a);
3777 struct nlattr *nla;
3778 size_t key_len = nla_len(ovs_key) / 2;
3779
3780 /* Revert the conversion we did from a non-masked set action to
3781 * masked set action.
3782 */
3783 nla = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3784 if (!nla)
3785 return -EMSGSIZE;
3786
3787 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
3788 return -EMSGSIZE;
3789
3790 nla_nest_end(skb, nla);
3791 return 0;
3792}
3793
3794int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
3795{
3796 const struct nlattr *a;
3797 int rem, err;
3798
3799 nla_for_each_attr(a, attr, len, rem) {
3800 int type = nla_type(a);
3801
3802 switch (type) {
3803 case OVS_ACTION_ATTR_SET:
3804 err = set_action_to_attr(a, skb);
3805 if (err)
3806 return err;
3807 break;
3808
3809 case OVS_ACTION_ATTR_SET_TO_MASKED:
3810 err = masked_set_action_to_set_action_attr(a, skb);
3811 if (err)
3812 return err;
3813 break;
3814
3815 case OVS_ACTION_ATTR_SAMPLE:
3816 err = sample_action_to_attr(a, skb);
3817 if (err)
3818 return err;
3819 break;
3820
3821 case OVS_ACTION_ATTR_CT:
3822 err = ovs_ct_action_to_attr(nla_data(a), skb);
3823 if (err)
3824 return err;
3825 break;
3826
3827 case OVS_ACTION_ATTR_CLONE:
3828 err = clone_action_to_attr(a, skb);
3829 if (err)
3830 return err;
3831 break;
3832
3833 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
3834 err = check_pkt_len_action_to_attr(a, skb);
3835 if (err)
3836 return err;
3837 break;
3838
3839 case OVS_ACTION_ATTR_DEC_TTL:
3840 err = dec_ttl_action_to_attr(a, skb);
3841 if (err)
3842 return err;
3843 break;
3844
3845 default:
3846 if (nla_put(skb, type, nla_len(a), nla_data(a)))
3847 return -EMSGSIZE;
3848 break;
3849 }
3850 }
3851
3852 return 0;
3853}