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/*
2 * Copyright (c) 2007-2014 Nicira, Inc.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
16 * 02110-1301, USA
17 */
18
19#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
20
21#include "flow.h"
22#include "datapath.h"
23#include <linux/uaccess.h>
24#include <linux/netdevice.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
27#include <linux/if_vlan.h>
28#include <net/llc_pdu.h>
29#include <linux/kernel.h>
30#include <linux/jhash.h>
31#include <linux/jiffies.h>
32#include <linux/llc.h>
33#include <linux/module.h>
34#include <linux/in.h>
35#include <linux/rcupdate.h>
36#include <linux/if_arp.h>
37#include <linux/ip.h>
38#include <linux/ipv6.h>
39#include <linux/sctp.h>
40#include <linux/tcp.h>
41#include <linux/udp.h>
42#include <linux/icmp.h>
43#include <linux/icmpv6.h>
44#include <linux/rculist.h>
45#include <net/geneve.h>
46#include <net/ip.h>
47#include <net/ipv6.h>
48#include <net/ndisc.h>
49#include <net/mpls.h>
50#include <net/vxlan.h>
51
52#include "flow_netlink.h"
53
54struct ovs_len_tbl {
55 int len;
56 const struct ovs_len_tbl *next;
57};
58
59#define OVS_ATTR_NESTED -1
60#define OVS_ATTR_VARIABLE -2
61
62static void update_range(struct sw_flow_match *match,
63 size_t offset, size_t size, bool is_mask)
64{
65 struct sw_flow_key_range *range;
66 size_t start = rounddown(offset, sizeof(long));
67 size_t end = roundup(offset + size, sizeof(long));
68
69 if (!is_mask)
70 range = &match->range;
71 else
72 range = &match->mask->range;
73
74 if (range->start == range->end) {
75 range->start = start;
76 range->end = end;
77 return;
78 }
79
80 if (range->start > start)
81 range->start = start;
82
83 if (range->end < end)
84 range->end = end;
85}
86
87#define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
88 do { \
89 update_range(match, offsetof(struct sw_flow_key, field), \
90 sizeof((match)->key->field), is_mask); \
91 if (is_mask) \
92 (match)->mask->key.field = value; \
93 else \
94 (match)->key->field = value; \
95 } while (0)
96
97#define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
98 do { \
99 update_range(match, offset, len, is_mask); \
100 if (is_mask) \
101 memcpy((u8 *)&(match)->mask->key + offset, value_p, \
102 len); \
103 else \
104 memcpy((u8 *)(match)->key + offset, value_p, len); \
105 } while (0)
106
107#define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
108 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
109 value_p, len, is_mask)
110
111#define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
112 do { \
113 update_range(match, offsetof(struct sw_flow_key, field), \
114 sizeof((match)->key->field), is_mask); \
115 if (is_mask) \
116 memset((u8 *)&(match)->mask->key.field, value, \
117 sizeof((match)->mask->key.field)); \
118 else \
119 memset((u8 *)&(match)->key->field, value, \
120 sizeof((match)->key->field)); \
121 } while (0)
122
123static bool match_validate(const struct sw_flow_match *match,
124 u64 key_attrs, u64 mask_attrs, bool log)
125{
126 u64 key_expected = 1 << OVS_KEY_ATTR_ETHERNET;
127 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
128
129 /* The following mask attributes allowed only if they
130 * pass the validation tests. */
131 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
132 | (1 << OVS_KEY_ATTR_IPV6)
133 | (1 << OVS_KEY_ATTR_TCP)
134 | (1 << OVS_KEY_ATTR_TCP_FLAGS)
135 | (1 << OVS_KEY_ATTR_UDP)
136 | (1 << OVS_KEY_ATTR_SCTP)
137 | (1 << OVS_KEY_ATTR_ICMP)
138 | (1 << OVS_KEY_ATTR_ICMPV6)
139 | (1 << OVS_KEY_ATTR_ARP)
140 | (1 << OVS_KEY_ATTR_ND)
141 | (1 << OVS_KEY_ATTR_MPLS));
142
143 /* Always allowed mask fields. */
144 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
145 | (1 << OVS_KEY_ATTR_IN_PORT)
146 | (1 << OVS_KEY_ATTR_ETHERTYPE));
147
148 /* Check key attributes. */
149 if (match->key->eth.type == htons(ETH_P_ARP)
150 || match->key->eth.type == htons(ETH_P_RARP)) {
151 key_expected |= 1 << OVS_KEY_ATTR_ARP;
152 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
153 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
154 }
155
156 if (eth_p_mpls(match->key->eth.type)) {
157 key_expected |= 1 << OVS_KEY_ATTR_MPLS;
158 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
159 mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
160 }
161
162 if (match->key->eth.type == htons(ETH_P_IP)) {
163 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
164 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
165 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
166
167 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
168 if (match->key->ip.proto == IPPROTO_UDP) {
169 key_expected |= 1 << OVS_KEY_ATTR_UDP;
170 if (match->mask && (match->mask->key.ip.proto == 0xff))
171 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
172 }
173
174 if (match->key->ip.proto == IPPROTO_SCTP) {
175 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
176 if (match->mask && (match->mask->key.ip.proto == 0xff))
177 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
178 }
179
180 if (match->key->ip.proto == IPPROTO_TCP) {
181 key_expected |= 1 << OVS_KEY_ATTR_TCP;
182 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
183 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
184 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
185 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
186 }
187 }
188
189 if (match->key->ip.proto == IPPROTO_ICMP) {
190 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
191 if (match->mask && (match->mask->key.ip.proto == 0xff))
192 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
193 }
194 }
195 }
196
197 if (match->key->eth.type == htons(ETH_P_IPV6)) {
198 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
199 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
200 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
201
202 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
203 if (match->key->ip.proto == IPPROTO_UDP) {
204 key_expected |= 1 << OVS_KEY_ATTR_UDP;
205 if (match->mask && (match->mask->key.ip.proto == 0xff))
206 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
207 }
208
209 if (match->key->ip.proto == IPPROTO_SCTP) {
210 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
211 if (match->mask && (match->mask->key.ip.proto == 0xff))
212 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
213 }
214
215 if (match->key->ip.proto == IPPROTO_TCP) {
216 key_expected |= 1 << OVS_KEY_ATTR_TCP;
217 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
218 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
219 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
220 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
221 }
222 }
223
224 if (match->key->ip.proto == IPPROTO_ICMPV6) {
225 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
226 if (match->mask && (match->mask->key.ip.proto == 0xff))
227 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
228
229 if (match->key->tp.src ==
230 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
231 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
232 key_expected |= 1 << OVS_KEY_ATTR_ND;
233 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
234 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
235 }
236 }
237 }
238 }
239
240 if ((key_attrs & key_expected) != key_expected) {
241 /* Key attributes check failed. */
242 OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
243 (unsigned long long)key_attrs,
244 (unsigned long long)key_expected);
245 return false;
246 }
247
248 if ((mask_attrs & mask_allowed) != mask_attrs) {
249 /* Mask attributes check failed. */
250 OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
251 (unsigned long long)mask_attrs,
252 (unsigned long long)mask_allowed);
253 return false;
254 }
255
256 return true;
257}
258
259size_t ovs_tun_key_attr_size(void)
260{
261 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
262 * updating this function.
263 */
264 return nla_total_size(8) /* OVS_TUNNEL_KEY_ATTR_ID */
265 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
266 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_DST */
267 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
268 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
269 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
270 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
271 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
272 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
273 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS is mutually exclusive with
274 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
275 */
276 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
277 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
278}
279
280size_t ovs_key_attr_size(void)
281{
282 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
283 * updating this function.
284 */
285 BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO != 26);
286
287 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
288 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
289 + ovs_tun_key_attr_size()
290 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
291 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
292 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
293 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
294 + nla_total_size(4) /* OVS_KEY_ATTR_CT_STATE */
295 + nla_total_size(2) /* OVS_KEY_ATTR_CT_ZONE */
296 + nla_total_size(4) /* OVS_KEY_ATTR_CT_MARK */
297 + nla_total_size(16) /* OVS_KEY_ATTR_CT_LABELS */
298 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
299 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
300 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
301 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
302 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
303 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
304 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
305 + nla_total_size(28); /* OVS_KEY_ATTR_ND */
306}
307
308static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
309 [OVS_VXLAN_EXT_GBP] = { .len = sizeof(u32) },
310};
311
312static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
313 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
314 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
315 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
316 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
317 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
318 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
319 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
320 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
321 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
322 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
323 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_VARIABLE },
324 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED,
325 .next = ovs_vxlan_ext_key_lens },
326 [OVS_TUNNEL_KEY_ATTR_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
327 [OVS_TUNNEL_KEY_ATTR_IPV6_DST] = { .len = sizeof(struct in6_addr) },
328};
329
330/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
331static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
332 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
333 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
334 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
335 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
336 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
337 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
338 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
339 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
340 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
341 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
342 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
343 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
344 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
345 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
346 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
347 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
348 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
349 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
350 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
351 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
352 .next = ovs_tunnel_key_lens, },
353 [OVS_KEY_ATTR_MPLS] = { .len = sizeof(struct ovs_key_mpls) },
354 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u32) },
355 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
356 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
357 [OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
358};
359
360static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
361{
362 return expected_len == attr_len ||
363 expected_len == OVS_ATTR_NESTED ||
364 expected_len == OVS_ATTR_VARIABLE;
365}
366
367static bool is_all_zero(const u8 *fp, size_t size)
368{
369 int i;
370
371 if (!fp)
372 return false;
373
374 for (i = 0; i < size; i++)
375 if (fp[i])
376 return false;
377
378 return true;
379}
380
381static int __parse_flow_nlattrs(const struct nlattr *attr,
382 const struct nlattr *a[],
383 u64 *attrsp, bool log, bool nz)
384{
385 const struct nlattr *nla;
386 u64 attrs;
387 int rem;
388
389 attrs = *attrsp;
390 nla_for_each_nested(nla, attr, rem) {
391 u16 type = nla_type(nla);
392 int expected_len;
393
394 if (type > OVS_KEY_ATTR_MAX) {
395 OVS_NLERR(log, "Key type %d is out of range max %d",
396 type, OVS_KEY_ATTR_MAX);
397 return -EINVAL;
398 }
399
400 if (attrs & (1 << type)) {
401 OVS_NLERR(log, "Duplicate key (type %d).", type);
402 return -EINVAL;
403 }
404
405 expected_len = ovs_key_lens[type].len;
406 if (!check_attr_len(nla_len(nla), expected_len)) {
407 OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
408 type, nla_len(nla), expected_len);
409 return -EINVAL;
410 }
411
412 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
413 attrs |= 1 << type;
414 a[type] = nla;
415 }
416 }
417 if (rem) {
418 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
419 return -EINVAL;
420 }
421
422 *attrsp = attrs;
423 return 0;
424}
425
426static int parse_flow_mask_nlattrs(const struct nlattr *attr,
427 const struct nlattr *a[], u64 *attrsp,
428 bool log)
429{
430 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
431}
432
433static int parse_flow_nlattrs(const struct nlattr *attr,
434 const struct nlattr *a[], u64 *attrsp,
435 bool log)
436{
437 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
438}
439
440static int genev_tun_opt_from_nlattr(const struct nlattr *a,
441 struct sw_flow_match *match, bool is_mask,
442 bool log)
443{
444 unsigned long opt_key_offset;
445
446 if (nla_len(a) > sizeof(match->key->tun_opts)) {
447 OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
448 nla_len(a), sizeof(match->key->tun_opts));
449 return -EINVAL;
450 }
451
452 if (nla_len(a) % 4 != 0) {
453 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
454 nla_len(a));
455 return -EINVAL;
456 }
457
458 /* We need to record the length of the options passed
459 * down, otherwise packets with the same format but
460 * additional options will be silently matched.
461 */
462 if (!is_mask) {
463 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
464 false);
465 } else {
466 /* This is somewhat unusual because it looks at
467 * both the key and mask while parsing the
468 * attributes (and by extension assumes the key
469 * is parsed first). Normally, we would verify
470 * that each is the correct length and that the
471 * attributes line up in the validate function.
472 * However, that is difficult because this is
473 * variable length and we won't have the
474 * information later.
475 */
476 if (match->key->tun_opts_len != nla_len(a)) {
477 OVS_NLERR(log, "Geneve option len %d != mask len %d",
478 match->key->tun_opts_len, nla_len(a));
479 return -EINVAL;
480 }
481
482 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
483 }
484
485 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
486 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
487 nla_len(a), is_mask);
488 return 0;
489}
490
491static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
492 struct sw_flow_match *match, bool is_mask,
493 bool log)
494{
495 struct nlattr *a;
496 int rem;
497 unsigned long opt_key_offset;
498 struct vxlan_metadata opts;
499
500 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
501
502 memset(&opts, 0, sizeof(opts));
503 nla_for_each_nested(a, attr, rem) {
504 int type = nla_type(a);
505
506 if (type > OVS_VXLAN_EXT_MAX) {
507 OVS_NLERR(log, "VXLAN extension %d out of range max %d",
508 type, OVS_VXLAN_EXT_MAX);
509 return -EINVAL;
510 }
511
512 if (!check_attr_len(nla_len(a),
513 ovs_vxlan_ext_key_lens[type].len)) {
514 OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
515 type, nla_len(a),
516 ovs_vxlan_ext_key_lens[type].len);
517 return -EINVAL;
518 }
519
520 switch (type) {
521 case OVS_VXLAN_EXT_GBP:
522 opts.gbp = nla_get_u32(a);
523 break;
524 default:
525 OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
526 type);
527 return -EINVAL;
528 }
529 }
530 if (rem) {
531 OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
532 rem);
533 return -EINVAL;
534 }
535
536 if (!is_mask)
537 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
538 else
539 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
540
541 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
542 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
543 is_mask);
544 return 0;
545}
546
547static int ip_tun_from_nlattr(const struct nlattr *attr,
548 struct sw_flow_match *match, bool is_mask,
549 bool log)
550{
551 bool ttl = false, ipv4 = false, ipv6 = false;
552 __be16 tun_flags = 0;
553 int opts_type = 0;
554 struct nlattr *a;
555 int rem;
556
557 nla_for_each_nested(a, attr, rem) {
558 int type = nla_type(a);
559 int err;
560
561 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
562 OVS_NLERR(log, "Tunnel attr %d out of range max %d",
563 type, OVS_TUNNEL_KEY_ATTR_MAX);
564 return -EINVAL;
565 }
566
567 if (!check_attr_len(nla_len(a),
568 ovs_tunnel_key_lens[type].len)) {
569 OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
570 type, nla_len(a), ovs_tunnel_key_lens[type].len);
571 return -EINVAL;
572 }
573
574 switch (type) {
575 case OVS_TUNNEL_KEY_ATTR_ID:
576 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
577 nla_get_be64(a), is_mask);
578 tun_flags |= TUNNEL_KEY;
579 break;
580 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
581 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
582 nla_get_in_addr(a), is_mask);
583 ipv4 = true;
584 break;
585 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
586 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
587 nla_get_in_addr(a), is_mask);
588 ipv4 = true;
589 break;
590 case OVS_TUNNEL_KEY_ATTR_IPV6_SRC:
591 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
592 nla_get_in6_addr(a), is_mask);
593 ipv6 = true;
594 break;
595 case OVS_TUNNEL_KEY_ATTR_IPV6_DST:
596 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
597 nla_get_in6_addr(a), is_mask);
598 ipv6 = true;
599 break;
600 case OVS_TUNNEL_KEY_ATTR_TOS:
601 SW_FLOW_KEY_PUT(match, tun_key.tos,
602 nla_get_u8(a), is_mask);
603 break;
604 case OVS_TUNNEL_KEY_ATTR_TTL:
605 SW_FLOW_KEY_PUT(match, tun_key.ttl,
606 nla_get_u8(a), is_mask);
607 ttl = true;
608 break;
609 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
610 tun_flags |= TUNNEL_DONT_FRAGMENT;
611 break;
612 case OVS_TUNNEL_KEY_ATTR_CSUM:
613 tun_flags |= TUNNEL_CSUM;
614 break;
615 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
616 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
617 nla_get_be16(a), is_mask);
618 break;
619 case OVS_TUNNEL_KEY_ATTR_TP_DST:
620 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
621 nla_get_be16(a), is_mask);
622 break;
623 case OVS_TUNNEL_KEY_ATTR_OAM:
624 tun_flags |= TUNNEL_OAM;
625 break;
626 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
627 if (opts_type) {
628 OVS_NLERR(log, "Multiple metadata blocks provided");
629 return -EINVAL;
630 }
631
632 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
633 if (err)
634 return err;
635
636 tun_flags |= TUNNEL_GENEVE_OPT;
637 opts_type = type;
638 break;
639 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
640 if (opts_type) {
641 OVS_NLERR(log, "Multiple metadata blocks provided");
642 return -EINVAL;
643 }
644
645 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
646 if (err)
647 return err;
648
649 tun_flags |= TUNNEL_VXLAN_OPT;
650 opts_type = type;
651 break;
652 default:
653 OVS_NLERR(log, "Unknown IP tunnel attribute %d",
654 type);
655 return -EINVAL;
656 }
657 }
658
659 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
660 if (is_mask)
661 SW_FLOW_KEY_MEMSET_FIELD(match, tun_proto, 0xff, true);
662 else
663 SW_FLOW_KEY_PUT(match, tun_proto, ipv6 ? AF_INET6 : AF_INET,
664 false);
665
666 if (rem > 0) {
667 OVS_NLERR(log, "IP tunnel attribute has %d unknown bytes.",
668 rem);
669 return -EINVAL;
670 }
671
672 if (ipv4 && ipv6) {
673 OVS_NLERR(log, "Mixed IPv4 and IPv6 tunnel attributes");
674 return -EINVAL;
675 }
676
677 if (!is_mask) {
678 if (!ipv4 && !ipv6) {
679 OVS_NLERR(log, "IP tunnel dst address not specified");
680 return -EINVAL;
681 }
682 if (ipv4 && !match->key->tun_key.u.ipv4.dst) {
683 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
684 return -EINVAL;
685 }
686 if (ipv6 && ipv6_addr_any(&match->key->tun_key.u.ipv6.dst)) {
687 OVS_NLERR(log, "IPv6 tunnel dst address is zero");
688 return -EINVAL;
689 }
690
691 if (!ttl) {
692 OVS_NLERR(log, "IP tunnel TTL not specified.");
693 return -EINVAL;
694 }
695 }
696
697 return opts_type;
698}
699
700static int vxlan_opt_to_nlattr(struct sk_buff *skb,
701 const void *tun_opts, int swkey_tun_opts_len)
702{
703 const struct vxlan_metadata *opts = tun_opts;
704 struct nlattr *nla;
705
706 nla = nla_nest_start(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
707 if (!nla)
708 return -EMSGSIZE;
709
710 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
711 return -EMSGSIZE;
712
713 nla_nest_end(skb, nla);
714 return 0;
715}
716
717static int __ip_tun_to_nlattr(struct sk_buff *skb,
718 const struct ip_tunnel_key *output,
719 const void *tun_opts, int swkey_tun_opts_len,
720 unsigned short tun_proto)
721{
722 if (output->tun_flags & TUNNEL_KEY &&
723 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
724 return -EMSGSIZE;
725 switch (tun_proto) {
726 case AF_INET:
727 if (output->u.ipv4.src &&
728 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
729 output->u.ipv4.src))
730 return -EMSGSIZE;
731 if (output->u.ipv4.dst &&
732 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
733 output->u.ipv4.dst))
734 return -EMSGSIZE;
735 break;
736 case AF_INET6:
737 if (!ipv6_addr_any(&output->u.ipv6.src) &&
738 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_SRC,
739 &output->u.ipv6.src))
740 return -EMSGSIZE;
741 if (!ipv6_addr_any(&output->u.ipv6.dst) &&
742 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_DST,
743 &output->u.ipv6.dst))
744 return -EMSGSIZE;
745 break;
746 }
747 if (output->tos &&
748 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
749 return -EMSGSIZE;
750 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
751 return -EMSGSIZE;
752 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
753 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
754 return -EMSGSIZE;
755 if ((output->tun_flags & TUNNEL_CSUM) &&
756 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
757 return -EMSGSIZE;
758 if (output->tp_src &&
759 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
760 return -EMSGSIZE;
761 if (output->tp_dst &&
762 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
763 return -EMSGSIZE;
764 if ((output->tun_flags & TUNNEL_OAM) &&
765 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
766 return -EMSGSIZE;
767 if (swkey_tun_opts_len) {
768 if (output->tun_flags & TUNNEL_GENEVE_OPT &&
769 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
770 swkey_tun_opts_len, tun_opts))
771 return -EMSGSIZE;
772 else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
773 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
774 return -EMSGSIZE;
775 }
776
777 return 0;
778}
779
780static int ip_tun_to_nlattr(struct sk_buff *skb,
781 const struct ip_tunnel_key *output,
782 const void *tun_opts, int swkey_tun_opts_len,
783 unsigned short tun_proto)
784{
785 struct nlattr *nla;
786 int err;
787
788 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
789 if (!nla)
790 return -EMSGSIZE;
791
792 err = __ip_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len,
793 tun_proto);
794 if (err)
795 return err;
796
797 nla_nest_end(skb, nla);
798 return 0;
799}
800
801int ovs_nla_put_tunnel_info(struct sk_buff *skb,
802 struct ip_tunnel_info *tun_info)
803{
804 return __ip_tun_to_nlattr(skb, &tun_info->key,
805 ip_tunnel_info_opts(tun_info),
806 tun_info->options_len,
807 ip_tunnel_info_af(tun_info));
808}
809
810static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
811 u64 *attrs, const struct nlattr **a,
812 bool is_mask, bool log)
813{
814 if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
815 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
816
817 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
818 *attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
819 }
820
821 if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
822 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
823
824 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
825 *attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
826 }
827
828 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
829 SW_FLOW_KEY_PUT(match, phy.priority,
830 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
831 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
832 }
833
834 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
835 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
836
837 if (is_mask) {
838 in_port = 0xffffffff; /* Always exact match in_port. */
839 } else if (in_port >= DP_MAX_PORTS) {
840 OVS_NLERR(log, "Port %d exceeds max allowable %d",
841 in_port, DP_MAX_PORTS);
842 return -EINVAL;
843 }
844
845 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
846 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
847 } else if (!is_mask) {
848 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
849 }
850
851 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
852 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
853
854 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
855 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
856 }
857 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
858 if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
859 is_mask, log) < 0)
860 return -EINVAL;
861 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
862 }
863
864 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
865 ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
866 u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
867
868 if (ct_state & ~CT_SUPPORTED_MASK) {
869 OVS_NLERR(log, "ct_state flags %08x unsupported",
870 ct_state);
871 return -EINVAL;
872 }
873
874 SW_FLOW_KEY_PUT(match, ct.state, ct_state, is_mask);
875 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
876 }
877 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
878 ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
879 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
880
881 SW_FLOW_KEY_PUT(match, ct.zone, ct_zone, is_mask);
882 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
883 }
884 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
885 ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
886 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
887
888 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
889 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
890 }
891 if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
892 ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
893 const struct ovs_key_ct_labels *cl;
894
895 cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
896 SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
897 sizeof(*cl), is_mask);
898 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
899 }
900 return 0;
901}
902
903static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
904 u64 attrs, const struct nlattr **a,
905 bool is_mask, bool log)
906{
907 int err;
908
909 err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
910 if (err)
911 return err;
912
913 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
914 const struct ovs_key_ethernet *eth_key;
915
916 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
917 SW_FLOW_KEY_MEMCPY(match, eth.src,
918 eth_key->eth_src, ETH_ALEN, is_mask);
919 SW_FLOW_KEY_MEMCPY(match, eth.dst,
920 eth_key->eth_dst, ETH_ALEN, is_mask);
921 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
922 }
923
924 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
925 __be16 tci;
926
927 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
928 if (!(tci & htons(VLAN_TAG_PRESENT))) {
929 if (is_mask)
930 OVS_NLERR(log, "VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.");
931 else
932 OVS_NLERR(log, "VLAN TCI does not have VLAN_TAG_PRESENT bit set.");
933
934 return -EINVAL;
935 }
936
937 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
938 attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
939 }
940
941 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
942 __be16 eth_type;
943
944 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
945 if (is_mask) {
946 /* Always exact match EtherType. */
947 eth_type = htons(0xffff);
948 } else if (!eth_proto_is_802_3(eth_type)) {
949 OVS_NLERR(log, "EtherType %x is less than min %x",
950 ntohs(eth_type), ETH_P_802_3_MIN);
951 return -EINVAL;
952 }
953
954 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
955 attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
956 } else if (!is_mask) {
957 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
958 }
959
960 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
961 const struct ovs_key_ipv4 *ipv4_key;
962
963 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
964 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
965 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
966 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
967 return -EINVAL;
968 }
969 SW_FLOW_KEY_PUT(match, ip.proto,
970 ipv4_key->ipv4_proto, is_mask);
971 SW_FLOW_KEY_PUT(match, ip.tos,
972 ipv4_key->ipv4_tos, is_mask);
973 SW_FLOW_KEY_PUT(match, ip.ttl,
974 ipv4_key->ipv4_ttl, is_mask);
975 SW_FLOW_KEY_PUT(match, ip.frag,
976 ipv4_key->ipv4_frag, is_mask);
977 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
978 ipv4_key->ipv4_src, is_mask);
979 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
980 ipv4_key->ipv4_dst, is_mask);
981 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
982 }
983
984 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
985 const struct ovs_key_ipv6 *ipv6_key;
986
987 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
988 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
989 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
990 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
991 return -EINVAL;
992 }
993
994 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
995 OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x).\n",
996 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
997 return -EINVAL;
998 }
999
1000 SW_FLOW_KEY_PUT(match, ipv6.label,
1001 ipv6_key->ipv6_label, is_mask);
1002 SW_FLOW_KEY_PUT(match, ip.proto,
1003 ipv6_key->ipv6_proto, is_mask);
1004 SW_FLOW_KEY_PUT(match, ip.tos,
1005 ipv6_key->ipv6_tclass, is_mask);
1006 SW_FLOW_KEY_PUT(match, ip.ttl,
1007 ipv6_key->ipv6_hlimit, is_mask);
1008 SW_FLOW_KEY_PUT(match, ip.frag,
1009 ipv6_key->ipv6_frag, is_mask);
1010 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1011 ipv6_key->ipv6_src,
1012 sizeof(match->key->ipv6.addr.src),
1013 is_mask);
1014 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1015 ipv6_key->ipv6_dst,
1016 sizeof(match->key->ipv6.addr.dst),
1017 is_mask);
1018
1019 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1020 }
1021
1022 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1023 const struct ovs_key_arp *arp_key;
1024
1025 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1026 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1027 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
1028 arp_key->arp_op);
1029 return -EINVAL;
1030 }
1031
1032 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1033 arp_key->arp_sip, is_mask);
1034 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1035 arp_key->arp_tip, is_mask);
1036 SW_FLOW_KEY_PUT(match, ip.proto,
1037 ntohs(arp_key->arp_op), is_mask);
1038 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1039 arp_key->arp_sha, ETH_ALEN, is_mask);
1040 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1041 arp_key->arp_tha, ETH_ALEN, is_mask);
1042
1043 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1044 }
1045
1046 if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1047 const struct ovs_key_mpls *mpls_key;
1048
1049 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1050 SW_FLOW_KEY_PUT(match, mpls.top_lse,
1051 mpls_key->mpls_lse, is_mask);
1052
1053 attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1054 }
1055
1056 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1057 const struct ovs_key_tcp *tcp_key;
1058
1059 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1060 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1061 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1062 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1063 }
1064
1065 if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1066 SW_FLOW_KEY_PUT(match, tp.flags,
1067 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1068 is_mask);
1069 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1070 }
1071
1072 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1073 const struct ovs_key_udp *udp_key;
1074
1075 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1076 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1077 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1078 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1079 }
1080
1081 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1082 const struct ovs_key_sctp *sctp_key;
1083
1084 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1085 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1086 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1087 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1088 }
1089
1090 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1091 const struct ovs_key_icmp *icmp_key;
1092
1093 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1094 SW_FLOW_KEY_PUT(match, tp.src,
1095 htons(icmp_key->icmp_type), is_mask);
1096 SW_FLOW_KEY_PUT(match, tp.dst,
1097 htons(icmp_key->icmp_code), is_mask);
1098 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1099 }
1100
1101 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1102 const struct ovs_key_icmpv6 *icmpv6_key;
1103
1104 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1105 SW_FLOW_KEY_PUT(match, tp.src,
1106 htons(icmpv6_key->icmpv6_type), is_mask);
1107 SW_FLOW_KEY_PUT(match, tp.dst,
1108 htons(icmpv6_key->icmpv6_code), is_mask);
1109 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1110 }
1111
1112 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1113 const struct ovs_key_nd *nd_key;
1114
1115 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1116 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1117 nd_key->nd_target,
1118 sizeof(match->key->ipv6.nd.target),
1119 is_mask);
1120 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1121 nd_key->nd_sll, ETH_ALEN, is_mask);
1122 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1123 nd_key->nd_tll, ETH_ALEN, is_mask);
1124 attrs &= ~(1 << OVS_KEY_ATTR_ND);
1125 }
1126
1127 if (attrs != 0) {
1128 OVS_NLERR(log, "Unknown key attributes %llx",
1129 (unsigned long long)attrs);
1130 return -EINVAL;
1131 }
1132
1133 return 0;
1134}
1135
1136static void nlattr_set(struct nlattr *attr, u8 val,
1137 const struct ovs_len_tbl *tbl)
1138{
1139 struct nlattr *nla;
1140 int rem;
1141
1142 /* The nlattr stream should already have been validated */
1143 nla_for_each_nested(nla, attr, rem) {
1144 if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED) {
1145 if (tbl[nla_type(nla)].next)
1146 tbl = tbl[nla_type(nla)].next;
1147 nlattr_set(nla, val, tbl);
1148 } else {
1149 memset(nla_data(nla), val, nla_len(nla));
1150 }
1151
1152 if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1153 *(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1154 }
1155}
1156
1157static void mask_set_nlattr(struct nlattr *attr, u8 val)
1158{
1159 nlattr_set(attr, val, ovs_key_lens);
1160}
1161
1162/**
1163 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1164 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1165 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1166 * does not include any don't care bit.
1167 * @net: Used to determine per-namespace field support.
1168 * @match: receives the extracted flow match information.
1169 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1170 * sequence. The fields should of the packet that triggered the creation
1171 * of this flow.
1172 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1173 * attribute specifies the mask field of the wildcarded flow.
1174 * @log: Boolean to allow kernel error logging. Normally true, but when
1175 * probing for feature compatibility this should be passed in as false to
1176 * suppress unnecessary error logging.
1177 */
1178int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1179 const struct nlattr *nla_key,
1180 const struct nlattr *nla_mask,
1181 bool log)
1182{
1183 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1184 const struct nlattr *encap;
1185 struct nlattr *newmask = NULL;
1186 u64 key_attrs = 0;
1187 u64 mask_attrs = 0;
1188 bool encap_valid = false;
1189 int err;
1190
1191 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1192 if (err)
1193 return err;
1194
1195 if ((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
1196 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
1197 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
1198 __be16 tci;
1199
1200 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
1201 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
1202 OVS_NLERR(log, "Invalid Vlan frame.");
1203 return -EINVAL;
1204 }
1205
1206 key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1207 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1208 encap = a[OVS_KEY_ATTR_ENCAP];
1209 key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1210 encap_valid = true;
1211
1212 if (tci & htons(VLAN_TAG_PRESENT)) {
1213 err = parse_flow_nlattrs(encap, a, &key_attrs, log);
1214 if (err)
1215 return err;
1216 } else if (!tci) {
1217 /* Corner case for truncated 802.1Q header. */
1218 if (nla_len(encap)) {
1219 OVS_NLERR(log, "Truncated 802.1Q header has non-zero encap attribute.");
1220 return -EINVAL;
1221 }
1222 } else {
1223 OVS_NLERR(log, "Encap attr is set for non-VLAN frame");
1224 return -EINVAL;
1225 }
1226 }
1227
1228 err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1229 if (err)
1230 return err;
1231
1232 if (match->mask) {
1233 if (!nla_mask) {
1234 /* Create an exact match mask. We need to set to 0xff
1235 * all the 'match->mask' fields that have been touched
1236 * in 'match->key'. We cannot simply memset
1237 * 'match->mask', because padding bytes and fields not
1238 * specified in 'match->key' should be left to 0.
1239 * Instead, we use a stream of netlink attributes,
1240 * copied from 'key' and set to 0xff.
1241 * ovs_key_from_nlattrs() will take care of filling
1242 * 'match->mask' appropriately.
1243 */
1244 newmask = kmemdup(nla_key,
1245 nla_total_size(nla_len(nla_key)),
1246 GFP_KERNEL);
1247 if (!newmask)
1248 return -ENOMEM;
1249
1250 mask_set_nlattr(newmask, 0xff);
1251
1252 /* The userspace does not send tunnel attributes that
1253 * are 0, but we should not wildcard them nonetheless.
1254 */
1255 if (match->key->tun_proto)
1256 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1257 0xff, true);
1258
1259 nla_mask = newmask;
1260 }
1261
1262 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1263 if (err)
1264 goto free_newmask;
1265
1266 /* Always match on tci. */
1267 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
1268
1269 if (mask_attrs & 1 << OVS_KEY_ATTR_ENCAP) {
1270 __be16 eth_type = 0;
1271 __be16 tci = 0;
1272
1273 if (!encap_valid) {
1274 OVS_NLERR(log, "Encap mask attribute is set for non-VLAN frame.");
1275 err = -EINVAL;
1276 goto free_newmask;
1277 }
1278
1279 mask_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1280 if (a[OVS_KEY_ATTR_ETHERTYPE])
1281 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1282
1283 if (eth_type == htons(0xffff)) {
1284 mask_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1285 encap = a[OVS_KEY_ATTR_ENCAP];
1286 err = parse_flow_mask_nlattrs(encap, a,
1287 &mask_attrs, log);
1288 if (err)
1289 goto free_newmask;
1290 } else {
1291 OVS_NLERR(log, "VLAN frames must have an exact match on the TPID (mask=%x).",
1292 ntohs(eth_type));
1293 err = -EINVAL;
1294 goto free_newmask;
1295 }
1296
1297 if (a[OVS_KEY_ATTR_VLAN])
1298 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1299
1300 if (!(tci & htons(VLAN_TAG_PRESENT))) {
1301 OVS_NLERR(log, "VLAN tag present bit must have an exact match (tci_mask=%x).",
1302 ntohs(tci));
1303 err = -EINVAL;
1304 goto free_newmask;
1305 }
1306 }
1307
1308 err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1309 log);
1310 if (err)
1311 goto free_newmask;
1312 }
1313
1314 if (!match_validate(match, key_attrs, mask_attrs, log))
1315 err = -EINVAL;
1316
1317free_newmask:
1318 kfree(newmask);
1319 return err;
1320}
1321
1322static size_t get_ufid_len(const struct nlattr *attr, bool log)
1323{
1324 size_t len;
1325
1326 if (!attr)
1327 return 0;
1328
1329 len = nla_len(attr);
1330 if (len < 1 || len > MAX_UFID_LENGTH) {
1331 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1332 nla_len(attr), MAX_UFID_LENGTH);
1333 return 0;
1334 }
1335
1336 return len;
1337}
1338
1339/* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1340 * or false otherwise.
1341 */
1342bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1343 bool log)
1344{
1345 sfid->ufid_len = get_ufid_len(attr, log);
1346 if (sfid->ufid_len)
1347 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1348
1349 return sfid->ufid_len;
1350}
1351
1352int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1353 const struct sw_flow_key *key, bool log)
1354{
1355 struct sw_flow_key *new_key;
1356
1357 if (ovs_nla_get_ufid(sfid, ufid, log))
1358 return 0;
1359
1360 /* If UFID was not provided, use unmasked key. */
1361 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1362 if (!new_key)
1363 return -ENOMEM;
1364 memcpy(new_key, key, sizeof(*key));
1365 sfid->unmasked_key = new_key;
1366
1367 return 0;
1368}
1369
1370u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1371{
1372 return attr ? nla_get_u32(attr) : 0;
1373}
1374
1375/**
1376 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1377 * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1378 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1379 * sequence.
1380 * @log: Boolean to allow kernel error logging. Normally true, but when
1381 * probing for feature compatibility this should be passed in as false to
1382 * suppress unnecessary error logging.
1383 *
1384 * This parses a series of Netlink attributes that form a flow key, which must
1385 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1386 * get the metadata, that is, the parts of the flow key that cannot be
1387 * extracted from the packet itself.
1388 */
1389
1390int ovs_nla_get_flow_metadata(struct net *net, const struct nlattr *attr,
1391 struct sw_flow_key *key,
1392 bool log)
1393{
1394 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1395 struct sw_flow_match match;
1396 u64 attrs = 0;
1397 int err;
1398
1399 err = parse_flow_nlattrs(attr, a, &attrs, log);
1400 if (err)
1401 return -EINVAL;
1402
1403 memset(&match, 0, sizeof(match));
1404 match.key = key;
1405
1406 memset(&key->ct, 0, sizeof(key->ct));
1407 key->phy.in_port = DP_MAX_PORTS;
1408
1409 return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1410}
1411
1412static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1413 const struct sw_flow_key *output, bool is_mask,
1414 struct sk_buff *skb)
1415{
1416 struct ovs_key_ethernet *eth_key;
1417 struct nlattr *nla, *encap;
1418
1419 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1420 goto nla_put_failure;
1421
1422 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1423 goto nla_put_failure;
1424
1425 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1426 goto nla_put_failure;
1427
1428 if ((swkey->tun_proto || is_mask)) {
1429 const void *opts = NULL;
1430
1431 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1432 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1433
1434 if (ip_tun_to_nlattr(skb, &output->tun_key, opts,
1435 swkey->tun_opts_len, swkey->tun_proto))
1436 goto nla_put_failure;
1437 }
1438
1439 if (swkey->phy.in_port == DP_MAX_PORTS) {
1440 if (is_mask && (output->phy.in_port == 0xffff))
1441 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1442 goto nla_put_failure;
1443 } else {
1444 u16 upper_u16;
1445 upper_u16 = !is_mask ? 0 : 0xffff;
1446
1447 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1448 (upper_u16 << 16) | output->phy.in_port))
1449 goto nla_put_failure;
1450 }
1451
1452 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1453 goto nla_put_failure;
1454
1455 if (ovs_ct_put_key(output, skb))
1456 goto nla_put_failure;
1457
1458 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1459 if (!nla)
1460 goto nla_put_failure;
1461
1462 eth_key = nla_data(nla);
1463 ether_addr_copy(eth_key->eth_src, output->eth.src);
1464 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1465
1466 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1467 __be16 eth_type;
1468 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
1469 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1470 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
1471 goto nla_put_failure;
1472 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1473 if (!swkey->eth.tci)
1474 goto unencap;
1475 } else
1476 encap = NULL;
1477
1478 if (swkey->eth.type == htons(ETH_P_802_2)) {
1479 /*
1480 * Ethertype 802.2 is represented in the netlink with omitted
1481 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1482 * 0xffff in the mask attribute. Ethertype can also
1483 * be wildcarded.
1484 */
1485 if (is_mask && output->eth.type)
1486 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1487 output->eth.type))
1488 goto nla_put_failure;
1489 goto unencap;
1490 }
1491
1492 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1493 goto nla_put_failure;
1494
1495 if (swkey->eth.type == htons(ETH_P_IP)) {
1496 struct ovs_key_ipv4 *ipv4_key;
1497
1498 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1499 if (!nla)
1500 goto nla_put_failure;
1501 ipv4_key = nla_data(nla);
1502 ipv4_key->ipv4_src = output->ipv4.addr.src;
1503 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1504 ipv4_key->ipv4_proto = output->ip.proto;
1505 ipv4_key->ipv4_tos = output->ip.tos;
1506 ipv4_key->ipv4_ttl = output->ip.ttl;
1507 ipv4_key->ipv4_frag = output->ip.frag;
1508 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1509 struct ovs_key_ipv6 *ipv6_key;
1510
1511 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1512 if (!nla)
1513 goto nla_put_failure;
1514 ipv6_key = nla_data(nla);
1515 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1516 sizeof(ipv6_key->ipv6_src));
1517 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1518 sizeof(ipv6_key->ipv6_dst));
1519 ipv6_key->ipv6_label = output->ipv6.label;
1520 ipv6_key->ipv6_proto = output->ip.proto;
1521 ipv6_key->ipv6_tclass = output->ip.tos;
1522 ipv6_key->ipv6_hlimit = output->ip.ttl;
1523 ipv6_key->ipv6_frag = output->ip.frag;
1524 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1525 swkey->eth.type == htons(ETH_P_RARP)) {
1526 struct ovs_key_arp *arp_key;
1527
1528 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1529 if (!nla)
1530 goto nla_put_failure;
1531 arp_key = nla_data(nla);
1532 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1533 arp_key->arp_sip = output->ipv4.addr.src;
1534 arp_key->arp_tip = output->ipv4.addr.dst;
1535 arp_key->arp_op = htons(output->ip.proto);
1536 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1537 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1538 } else if (eth_p_mpls(swkey->eth.type)) {
1539 struct ovs_key_mpls *mpls_key;
1540
1541 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1542 if (!nla)
1543 goto nla_put_failure;
1544 mpls_key = nla_data(nla);
1545 mpls_key->mpls_lse = output->mpls.top_lse;
1546 }
1547
1548 if ((swkey->eth.type == htons(ETH_P_IP) ||
1549 swkey->eth.type == htons(ETH_P_IPV6)) &&
1550 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1551
1552 if (swkey->ip.proto == IPPROTO_TCP) {
1553 struct ovs_key_tcp *tcp_key;
1554
1555 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1556 if (!nla)
1557 goto nla_put_failure;
1558 tcp_key = nla_data(nla);
1559 tcp_key->tcp_src = output->tp.src;
1560 tcp_key->tcp_dst = output->tp.dst;
1561 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1562 output->tp.flags))
1563 goto nla_put_failure;
1564 } else if (swkey->ip.proto == IPPROTO_UDP) {
1565 struct ovs_key_udp *udp_key;
1566
1567 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1568 if (!nla)
1569 goto nla_put_failure;
1570 udp_key = nla_data(nla);
1571 udp_key->udp_src = output->tp.src;
1572 udp_key->udp_dst = output->tp.dst;
1573 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1574 struct ovs_key_sctp *sctp_key;
1575
1576 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1577 if (!nla)
1578 goto nla_put_failure;
1579 sctp_key = nla_data(nla);
1580 sctp_key->sctp_src = output->tp.src;
1581 sctp_key->sctp_dst = output->tp.dst;
1582 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1583 swkey->ip.proto == IPPROTO_ICMP) {
1584 struct ovs_key_icmp *icmp_key;
1585
1586 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1587 if (!nla)
1588 goto nla_put_failure;
1589 icmp_key = nla_data(nla);
1590 icmp_key->icmp_type = ntohs(output->tp.src);
1591 icmp_key->icmp_code = ntohs(output->tp.dst);
1592 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1593 swkey->ip.proto == IPPROTO_ICMPV6) {
1594 struct ovs_key_icmpv6 *icmpv6_key;
1595
1596 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1597 sizeof(*icmpv6_key));
1598 if (!nla)
1599 goto nla_put_failure;
1600 icmpv6_key = nla_data(nla);
1601 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1602 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1603
1604 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1605 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1606 struct ovs_key_nd *nd_key;
1607
1608 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1609 if (!nla)
1610 goto nla_put_failure;
1611 nd_key = nla_data(nla);
1612 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1613 sizeof(nd_key->nd_target));
1614 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1615 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1616 }
1617 }
1618 }
1619
1620unencap:
1621 if (encap)
1622 nla_nest_end(skb, encap);
1623
1624 return 0;
1625
1626nla_put_failure:
1627 return -EMSGSIZE;
1628}
1629
1630int ovs_nla_put_key(const struct sw_flow_key *swkey,
1631 const struct sw_flow_key *output, int attr, bool is_mask,
1632 struct sk_buff *skb)
1633{
1634 int err;
1635 struct nlattr *nla;
1636
1637 nla = nla_nest_start(skb, attr);
1638 if (!nla)
1639 return -EMSGSIZE;
1640 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
1641 if (err)
1642 return err;
1643 nla_nest_end(skb, nla);
1644
1645 return 0;
1646}
1647
1648/* Called with ovs_mutex or RCU read lock. */
1649int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
1650{
1651 if (ovs_identifier_is_ufid(&flow->id))
1652 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
1653 flow->id.ufid);
1654
1655 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
1656 OVS_FLOW_ATTR_KEY, false, skb);
1657}
1658
1659/* Called with ovs_mutex or RCU read lock. */
1660int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
1661{
1662 return ovs_nla_put_key(&flow->key, &flow->key,
1663 OVS_FLOW_ATTR_KEY, false, skb);
1664}
1665
1666/* Called with ovs_mutex or RCU read lock. */
1667int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
1668{
1669 return ovs_nla_put_key(&flow->key, &flow->mask->key,
1670 OVS_FLOW_ATTR_MASK, true, skb);
1671}
1672
1673#define MAX_ACTIONS_BUFSIZE (32 * 1024)
1674
1675static struct sw_flow_actions *nla_alloc_flow_actions(int size, bool log)
1676{
1677 struct sw_flow_actions *sfa;
1678
1679 if (size > MAX_ACTIONS_BUFSIZE) {
1680 OVS_NLERR(log, "Flow action size %u bytes exceeds max", size);
1681 return ERR_PTR(-EINVAL);
1682 }
1683
1684 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1685 if (!sfa)
1686 return ERR_PTR(-ENOMEM);
1687
1688 sfa->actions_len = 0;
1689 return sfa;
1690}
1691
1692static void ovs_nla_free_set_action(const struct nlattr *a)
1693{
1694 const struct nlattr *ovs_key = nla_data(a);
1695 struct ovs_tunnel_info *ovs_tun;
1696
1697 switch (nla_type(ovs_key)) {
1698 case OVS_KEY_ATTR_TUNNEL_INFO:
1699 ovs_tun = nla_data(ovs_key);
1700 dst_release((struct dst_entry *)ovs_tun->tun_dst);
1701 break;
1702 }
1703}
1704
1705void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1706{
1707 const struct nlattr *a;
1708 int rem;
1709
1710 if (!sf_acts)
1711 return;
1712
1713 nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
1714 switch (nla_type(a)) {
1715 case OVS_ACTION_ATTR_SET:
1716 ovs_nla_free_set_action(a);
1717 break;
1718 case OVS_ACTION_ATTR_CT:
1719 ovs_ct_free_action(a);
1720 break;
1721 }
1722 }
1723
1724 kfree(sf_acts);
1725}
1726
1727static void __ovs_nla_free_flow_actions(struct rcu_head *head)
1728{
1729 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
1730}
1731
1732/* Schedules 'sf_acts' to be freed after the next RCU grace period.
1733 * The caller must hold rcu_read_lock for this to be sensible. */
1734void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
1735{
1736 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
1737}
1738
1739static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1740 int attr_len, bool log)
1741{
1742
1743 struct sw_flow_actions *acts;
1744 int new_acts_size;
1745 int req_size = NLA_ALIGN(attr_len);
1746 int next_offset = offsetof(struct sw_flow_actions, actions) +
1747 (*sfa)->actions_len;
1748
1749 if (req_size <= (ksize(*sfa) - next_offset))
1750 goto out;
1751
1752 new_acts_size = ksize(*sfa) * 2;
1753
1754 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1755 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1756 return ERR_PTR(-EMSGSIZE);
1757 new_acts_size = MAX_ACTIONS_BUFSIZE;
1758 }
1759
1760 acts = nla_alloc_flow_actions(new_acts_size, log);
1761 if (IS_ERR(acts))
1762 return (void *)acts;
1763
1764 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1765 acts->actions_len = (*sfa)->actions_len;
1766 acts->orig_len = (*sfa)->orig_len;
1767 kfree(*sfa);
1768 *sfa = acts;
1769
1770out:
1771 (*sfa)->actions_len += req_size;
1772 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1773}
1774
1775static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1776 int attrtype, void *data, int len, bool log)
1777{
1778 struct nlattr *a;
1779
1780 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
1781 if (IS_ERR(a))
1782 return a;
1783
1784 a->nla_type = attrtype;
1785 a->nla_len = nla_attr_size(len);
1786
1787 if (data)
1788 memcpy(nla_data(a), data, len);
1789 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1790
1791 return a;
1792}
1793
1794int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
1795 int len, bool log)
1796{
1797 struct nlattr *a;
1798
1799 a = __add_action(sfa, attrtype, data, len, log);
1800
1801 return PTR_ERR_OR_ZERO(a);
1802}
1803
1804static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1805 int attrtype, bool log)
1806{
1807 int used = (*sfa)->actions_len;
1808 int err;
1809
1810 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
1811 if (err)
1812 return err;
1813
1814 return used;
1815}
1816
1817static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1818 int st_offset)
1819{
1820 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1821 st_offset);
1822
1823 a->nla_len = sfa->actions_len - st_offset;
1824}
1825
1826static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
1827 const struct sw_flow_key *key,
1828 int depth, struct sw_flow_actions **sfa,
1829 __be16 eth_type, __be16 vlan_tci, bool log);
1830
1831static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
1832 const struct sw_flow_key *key, int depth,
1833 struct sw_flow_actions **sfa,
1834 __be16 eth_type, __be16 vlan_tci, bool log)
1835{
1836 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1837 const struct nlattr *probability, *actions;
1838 const struct nlattr *a;
1839 int rem, start, err, st_acts;
1840
1841 memset(attrs, 0, sizeof(attrs));
1842 nla_for_each_nested(a, attr, rem) {
1843 int type = nla_type(a);
1844 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1845 return -EINVAL;
1846 attrs[type] = a;
1847 }
1848 if (rem)
1849 return -EINVAL;
1850
1851 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1852 if (!probability || nla_len(probability) != sizeof(u32))
1853 return -EINVAL;
1854
1855 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1856 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1857 return -EINVAL;
1858
1859 /* validation done, copy sample action. */
1860 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
1861 if (start < 0)
1862 return start;
1863 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1864 nla_data(probability), sizeof(u32), log);
1865 if (err)
1866 return err;
1867 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS, log);
1868 if (st_acts < 0)
1869 return st_acts;
1870
1871 err = __ovs_nla_copy_actions(net, actions, key, depth + 1, sfa,
1872 eth_type, vlan_tci, log);
1873 if (err)
1874 return err;
1875
1876 add_nested_action_end(*sfa, st_acts);
1877 add_nested_action_end(*sfa, start);
1878
1879 return 0;
1880}
1881
1882void ovs_match_init(struct sw_flow_match *match,
1883 struct sw_flow_key *key,
1884 struct sw_flow_mask *mask)
1885{
1886 memset(match, 0, sizeof(*match));
1887 match->key = key;
1888 match->mask = mask;
1889
1890 memset(key, 0, sizeof(*key));
1891
1892 if (mask) {
1893 memset(&mask->key, 0, sizeof(mask->key));
1894 mask->range.start = mask->range.end = 0;
1895 }
1896}
1897
1898static int validate_geneve_opts(struct sw_flow_key *key)
1899{
1900 struct geneve_opt *option;
1901 int opts_len = key->tun_opts_len;
1902 bool crit_opt = false;
1903
1904 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
1905 while (opts_len > 0) {
1906 int len;
1907
1908 if (opts_len < sizeof(*option))
1909 return -EINVAL;
1910
1911 len = sizeof(*option) + option->length * 4;
1912 if (len > opts_len)
1913 return -EINVAL;
1914
1915 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
1916
1917 option = (struct geneve_opt *)((u8 *)option + len);
1918 opts_len -= len;
1919 };
1920
1921 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
1922
1923 return 0;
1924}
1925
1926static int validate_and_copy_set_tun(const struct nlattr *attr,
1927 struct sw_flow_actions **sfa, bool log)
1928{
1929 struct sw_flow_match match;
1930 struct sw_flow_key key;
1931 struct metadata_dst *tun_dst;
1932 struct ip_tunnel_info *tun_info;
1933 struct ovs_tunnel_info *ovs_tun;
1934 struct nlattr *a;
1935 int err = 0, start, opts_type;
1936
1937 ovs_match_init(&match, &key, NULL);
1938 opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
1939 if (opts_type < 0)
1940 return opts_type;
1941
1942 if (key.tun_opts_len) {
1943 switch (opts_type) {
1944 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
1945 err = validate_geneve_opts(&key);
1946 if (err < 0)
1947 return err;
1948 break;
1949 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
1950 break;
1951 }
1952 };
1953
1954 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
1955 if (start < 0)
1956 return start;
1957
1958 tun_dst = metadata_dst_alloc(key.tun_opts_len, GFP_KERNEL);
1959 if (!tun_dst)
1960 return -ENOMEM;
1961
1962 err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
1963 if (err) {
1964 dst_release((struct dst_entry *)tun_dst);
1965 return err;
1966 }
1967
1968 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
1969 sizeof(*ovs_tun), log);
1970 if (IS_ERR(a)) {
1971 dst_release((struct dst_entry *)tun_dst);
1972 return PTR_ERR(a);
1973 }
1974
1975 ovs_tun = nla_data(a);
1976 ovs_tun->tun_dst = tun_dst;
1977
1978 tun_info = &tun_dst->u.tun_info;
1979 tun_info->mode = IP_TUNNEL_INFO_TX;
1980 if (key.tun_proto == AF_INET6)
1981 tun_info->mode |= IP_TUNNEL_INFO_IPV6;
1982 tun_info->key = key.tun_key;
1983
1984 /* We need to store the options in the action itself since
1985 * everything else will go away after flow setup. We can append
1986 * it to tun_info and then point there.
1987 */
1988 ip_tunnel_info_opts_set(tun_info,
1989 TUN_METADATA_OPTS(&key, key.tun_opts_len),
1990 key.tun_opts_len);
1991 add_nested_action_end(*sfa, start);
1992
1993 return err;
1994}
1995
1996/* Return false if there are any non-masked bits set.
1997 * Mask follows data immediately, before any netlink padding.
1998 */
1999static bool validate_masked(u8 *data, int len)
2000{
2001 u8 *mask = data + len;
2002
2003 while (len--)
2004 if (*data++ & ~*mask++)
2005 return false;
2006
2007 return true;
2008}
2009
2010static int validate_set(const struct nlattr *a,
2011 const struct sw_flow_key *flow_key,
2012 struct sw_flow_actions **sfa,
2013 bool *skip_copy, __be16 eth_type, bool masked, bool log)
2014{
2015 const struct nlattr *ovs_key = nla_data(a);
2016 int key_type = nla_type(ovs_key);
2017 size_t key_len;
2018
2019 /* There can be only one key in a action */
2020 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2021 return -EINVAL;
2022
2023 key_len = nla_len(ovs_key);
2024 if (masked)
2025 key_len /= 2;
2026
2027 if (key_type > OVS_KEY_ATTR_MAX ||
2028 !check_attr_len(key_len, ovs_key_lens[key_type].len))
2029 return -EINVAL;
2030
2031 if (masked && !validate_masked(nla_data(ovs_key), key_len))
2032 return -EINVAL;
2033
2034 switch (key_type) {
2035 const struct ovs_key_ipv4 *ipv4_key;
2036 const struct ovs_key_ipv6 *ipv6_key;
2037 int err;
2038
2039 case OVS_KEY_ATTR_PRIORITY:
2040 case OVS_KEY_ATTR_SKB_MARK:
2041 case OVS_KEY_ATTR_CT_MARK:
2042 case OVS_KEY_ATTR_CT_LABELS:
2043 case OVS_KEY_ATTR_ETHERNET:
2044 break;
2045
2046 case OVS_KEY_ATTR_TUNNEL:
2047 if (masked)
2048 return -EINVAL; /* Masked tunnel set not supported. */
2049
2050 *skip_copy = true;
2051 err = validate_and_copy_set_tun(a, sfa, log);
2052 if (err)
2053 return err;
2054 break;
2055
2056 case OVS_KEY_ATTR_IPV4:
2057 if (eth_type != htons(ETH_P_IP))
2058 return -EINVAL;
2059
2060 ipv4_key = nla_data(ovs_key);
2061
2062 if (masked) {
2063 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2064
2065 /* Non-writeable fields. */
2066 if (mask->ipv4_proto || mask->ipv4_frag)
2067 return -EINVAL;
2068 } else {
2069 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2070 return -EINVAL;
2071
2072 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2073 return -EINVAL;
2074 }
2075 break;
2076
2077 case OVS_KEY_ATTR_IPV6:
2078 if (eth_type != htons(ETH_P_IPV6))
2079 return -EINVAL;
2080
2081 ipv6_key = nla_data(ovs_key);
2082
2083 if (masked) {
2084 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2085
2086 /* Non-writeable fields. */
2087 if (mask->ipv6_proto || mask->ipv6_frag)
2088 return -EINVAL;
2089
2090 /* Invalid bits in the flow label mask? */
2091 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2092 return -EINVAL;
2093 } else {
2094 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2095 return -EINVAL;
2096
2097 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2098 return -EINVAL;
2099 }
2100 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2101 return -EINVAL;
2102
2103 break;
2104
2105 case OVS_KEY_ATTR_TCP:
2106 if ((eth_type != htons(ETH_P_IP) &&
2107 eth_type != htons(ETH_P_IPV6)) ||
2108 flow_key->ip.proto != IPPROTO_TCP)
2109 return -EINVAL;
2110
2111 break;
2112
2113 case OVS_KEY_ATTR_UDP:
2114 if ((eth_type != htons(ETH_P_IP) &&
2115 eth_type != htons(ETH_P_IPV6)) ||
2116 flow_key->ip.proto != IPPROTO_UDP)
2117 return -EINVAL;
2118
2119 break;
2120
2121 case OVS_KEY_ATTR_MPLS:
2122 if (!eth_p_mpls(eth_type))
2123 return -EINVAL;
2124 break;
2125
2126 case OVS_KEY_ATTR_SCTP:
2127 if ((eth_type != htons(ETH_P_IP) &&
2128 eth_type != htons(ETH_P_IPV6)) ||
2129 flow_key->ip.proto != IPPROTO_SCTP)
2130 return -EINVAL;
2131
2132 break;
2133
2134 default:
2135 return -EINVAL;
2136 }
2137
2138 /* Convert non-masked non-tunnel set actions to masked set actions. */
2139 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2140 int start, len = key_len * 2;
2141 struct nlattr *at;
2142
2143 *skip_copy = true;
2144
2145 start = add_nested_action_start(sfa,
2146 OVS_ACTION_ATTR_SET_TO_MASKED,
2147 log);
2148 if (start < 0)
2149 return start;
2150
2151 at = __add_action(sfa, key_type, NULL, len, log);
2152 if (IS_ERR(at))
2153 return PTR_ERR(at);
2154
2155 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2156 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
2157 /* Clear non-writeable bits from otherwise writeable fields. */
2158 if (key_type == OVS_KEY_ATTR_IPV6) {
2159 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2160
2161 mask->ipv6_label &= htonl(0x000FFFFF);
2162 }
2163 add_nested_action_end(*sfa, start);
2164 }
2165
2166 return 0;
2167}
2168
2169static int validate_userspace(const struct nlattr *attr)
2170{
2171 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2172 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2173 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2174 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2175 };
2176 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2177 int error;
2178
2179 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
2180 attr, userspace_policy);
2181 if (error)
2182 return error;
2183
2184 if (!a[OVS_USERSPACE_ATTR_PID] ||
2185 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2186 return -EINVAL;
2187
2188 return 0;
2189}
2190
2191static int copy_action(const struct nlattr *from,
2192 struct sw_flow_actions **sfa, bool log)
2193{
2194 int totlen = NLA_ALIGN(from->nla_len);
2195 struct nlattr *to;
2196
2197 to = reserve_sfa_size(sfa, from->nla_len, log);
2198 if (IS_ERR(to))
2199 return PTR_ERR(to);
2200
2201 memcpy(to, from, totlen);
2202 return 0;
2203}
2204
2205static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2206 const struct sw_flow_key *key,
2207 int depth, struct sw_flow_actions **sfa,
2208 __be16 eth_type, __be16 vlan_tci, bool log)
2209{
2210 const struct nlattr *a;
2211 int rem, err;
2212
2213 if (depth >= SAMPLE_ACTION_DEPTH)
2214 return -EOVERFLOW;
2215
2216 nla_for_each_nested(a, attr, rem) {
2217 /* Expected argument lengths, (u32)-1 for variable length. */
2218 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2219 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2220 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2221 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2222 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2223 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2224 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2225 [OVS_ACTION_ATTR_POP_VLAN] = 0,
2226 [OVS_ACTION_ATTR_SET] = (u32)-1,
2227 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2228 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2229 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2230 [OVS_ACTION_ATTR_CT] = (u32)-1,
2231 };
2232 const struct ovs_action_push_vlan *vlan;
2233 int type = nla_type(a);
2234 bool skip_copy;
2235
2236 if (type > OVS_ACTION_ATTR_MAX ||
2237 (action_lens[type] != nla_len(a) &&
2238 action_lens[type] != (u32)-1))
2239 return -EINVAL;
2240
2241 skip_copy = false;
2242 switch (type) {
2243 case OVS_ACTION_ATTR_UNSPEC:
2244 return -EINVAL;
2245
2246 case OVS_ACTION_ATTR_USERSPACE:
2247 err = validate_userspace(a);
2248 if (err)
2249 return err;
2250 break;
2251
2252 case OVS_ACTION_ATTR_OUTPUT:
2253 if (nla_get_u32(a) >= DP_MAX_PORTS)
2254 return -EINVAL;
2255 break;
2256
2257 case OVS_ACTION_ATTR_HASH: {
2258 const struct ovs_action_hash *act_hash = nla_data(a);
2259
2260 switch (act_hash->hash_alg) {
2261 case OVS_HASH_ALG_L4:
2262 break;
2263 default:
2264 return -EINVAL;
2265 }
2266
2267 break;
2268 }
2269
2270 case OVS_ACTION_ATTR_POP_VLAN:
2271 vlan_tci = htons(0);
2272 break;
2273
2274 case OVS_ACTION_ATTR_PUSH_VLAN:
2275 vlan = nla_data(a);
2276 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
2277 return -EINVAL;
2278 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2279 return -EINVAL;
2280 vlan_tci = vlan->vlan_tci;
2281 break;
2282
2283 case OVS_ACTION_ATTR_RECIRC:
2284 break;
2285
2286 case OVS_ACTION_ATTR_PUSH_MPLS: {
2287 const struct ovs_action_push_mpls *mpls = nla_data(a);
2288
2289 if (!eth_p_mpls(mpls->mpls_ethertype))
2290 return -EINVAL;
2291 /* Prohibit push MPLS other than to a white list
2292 * for packets that have a known tag order.
2293 */
2294 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2295 (eth_type != htons(ETH_P_IP) &&
2296 eth_type != htons(ETH_P_IPV6) &&
2297 eth_type != htons(ETH_P_ARP) &&
2298 eth_type != htons(ETH_P_RARP) &&
2299 !eth_p_mpls(eth_type)))
2300 return -EINVAL;
2301 eth_type = mpls->mpls_ethertype;
2302 break;
2303 }
2304
2305 case OVS_ACTION_ATTR_POP_MPLS:
2306 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2307 !eth_p_mpls(eth_type))
2308 return -EINVAL;
2309
2310 /* Disallow subsequent L2.5+ set and mpls_pop actions
2311 * as there is no check here to ensure that the new
2312 * eth_type is valid and thus set actions could
2313 * write off the end of the packet or otherwise
2314 * corrupt it.
2315 *
2316 * Support for these actions is planned using packet
2317 * recirculation.
2318 */
2319 eth_type = htons(0);
2320 break;
2321
2322 case OVS_ACTION_ATTR_SET:
2323 err = validate_set(a, key, sfa,
2324 &skip_copy, eth_type, false, log);
2325 if (err)
2326 return err;
2327 break;
2328
2329 case OVS_ACTION_ATTR_SET_MASKED:
2330 err = validate_set(a, key, sfa,
2331 &skip_copy, eth_type, true, log);
2332 if (err)
2333 return err;
2334 break;
2335
2336 case OVS_ACTION_ATTR_SAMPLE:
2337 err = validate_and_copy_sample(net, a, key, depth, sfa,
2338 eth_type, vlan_tci, log);
2339 if (err)
2340 return err;
2341 skip_copy = true;
2342 break;
2343
2344 case OVS_ACTION_ATTR_CT:
2345 err = ovs_ct_copy_action(net, a, key, sfa, log);
2346 if (err)
2347 return err;
2348 skip_copy = true;
2349 break;
2350
2351 default:
2352 OVS_NLERR(log, "Unknown Action type %d", type);
2353 return -EINVAL;
2354 }
2355 if (!skip_copy) {
2356 err = copy_action(a, sfa, log);
2357 if (err)
2358 return err;
2359 }
2360 }
2361
2362 if (rem > 0)
2363 return -EINVAL;
2364
2365 return 0;
2366}
2367
2368/* 'key' must be the masked key. */
2369int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2370 const struct sw_flow_key *key,
2371 struct sw_flow_actions **sfa, bool log)
2372{
2373 int err;
2374
2375 *sfa = nla_alloc_flow_actions(nla_len(attr), log);
2376 if (IS_ERR(*sfa))
2377 return PTR_ERR(*sfa);
2378
2379 (*sfa)->orig_len = nla_len(attr);
2380 err = __ovs_nla_copy_actions(net, attr, key, 0, sfa, key->eth.type,
2381 key->eth.tci, log);
2382 if (err)
2383 ovs_nla_free_flow_actions(*sfa);
2384
2385 return err;
2386}
2387
2388static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
2389{
2390 const struct nlattr *a;
2391 struct nlattr *start;
2392 int err = 0, rem;
2393
2394 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
2395 if (!start)
2396 return -EMSGSIZE;
2397
2398 nla_for_each_nested(a, attr, rem) {
2399 int type = nla_type(a);
2400 struct nlattr *st_sample;
2401
2402 switch (type) {
2403 case OVS_SAMPLE_ATTR_PROBABILITY:
2404 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
2405 sizeof(u32), nla_data(a)))
2406 return -EMSGSIZE;
2407 break;
2408 case OVS_SAMPLE_ATTR_ACTIONS:
2409 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
2410 if (!st_sample)
2411 return -EMSGSIZE;
2412 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
2413 if (err)
2414 return err;
2415 nla_nest_end(skb, st_sample);
2416 break;
2417 }
2418 }
2419
2420 nla_nest_end(skb, start);
2421 return err;
2422}
2423
2424static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
2425{
2426 const struct nlattr *ovs_key = nla_data(a);
2427 int key_type = nla_type(ovs_key);
2428 struct nlattr *start;
2429 int err;
2430
2431 switch (key_type) {
2432 case OVS_KEY_ATTR_TUNNEL_INFO: {
2433 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
2434 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
2435
2436 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2437 if (!start)
2438 return -EMSGSIZE;
2439
2440 err = ip_tun_to_nlattr(skb, &tun_info->key,
2441 ip_tunnel_info_opts(tun_info),
2442 tun_info->options_len,
2443 ip_tunnel_info_af(tun_info));
2444 if (err)
2445 return err;
2446 nla_nest_end(skb, start);
2447 break;
2448 }
2449 default:
2450 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
2451 return -EMSGSIZE;
2452 break;
2453 }
2454
2455 return 0;
2456}
2457
2458static int masked_set_action_to_set_action_attr(const struct nlattr *a,
2459 struct sk_buff *skb)
2460{
2461 const struct nlattr *ovs_key = nla_data(a);
2462 struct nlattr *nla;
2463 size_t key_len = nla_len(ovs_key) / 2;
2464
2465 /* Revert the conversion we did from a non-masked set action to
2466 * masked set action.
2467 */
2468 nla = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2469 if (!nla)
2470 return -EMSGSIZE;
2471
2472 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
2473 return -EMSGSIZE;
2474
2475 nla_nest_end(skb, nla);
2476 return 0;
2477}
2478
2479int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
2480{
2481 const struct nlattr *a;
2482 int rem, err;
2483
2484 nla_for_each_attr(a, attr, len, rem) {
2485 int type = nla_type(a);
2486
2487 switch (type) {
2488 case OVS_ACTION_ATTR_SET:
2489 err = set_action_to_attr(a, skb);
2490 if (err)
2491 return err;
2492 break;
2493
2494 case OVS_ACTION_ATTR_SET_TO_MASKED:
2495 err = masked_set_action_to_set_action_attr(a, skb);
2496 if (err)
2497 return err;
2498 break;
2499
2500 case OVS_ACTION_ATTR_SAMPLE:
2501 err = sample_action_to_attr(a, skb);
2502 if (err)
2503 return err;
2504 break;
2505
2506 case OVS_ACTION_ATTR_CT:
2507 err = ovs_ct_action_to_attr(nla_data(a), skb);
2508 if (err)
2509 return err;
2510 break;
2511
2512 default:
2513 if (nla_put(skb, type, nla_len(a), nla_data(a)))
2514 return -EMSGSIZE;
2515 break;
2516 }
2517 }
2518
2519 return 0;
2520}