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
43 struct 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
actions_may_change_flow(const struct nlattr * actions)51 static 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
update_range(struct sw_flow_match * match,size_t offset,size_t size,bool is_mask)91 static 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
match_validate(const struct sw_flow_match * match,u64 key_attrs,u64 mask_attrs,bool log)152 static 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
ovs_tun_key_attr_size(void)309 size_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
ovs_nsh_key_attr_size(void)331 static 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
ovs_key_attr_size(void)344 size_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
375 static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
376 [OVS_VXLAN_EXT_GBP] = { .len = sizeof(u32) },
377 };
378
379 static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
380 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
381 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
382 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
383 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
384 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
385 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
386 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
387 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
388 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
389 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
390 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_VARIABLE },
391 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED,
392 .next = ovs_vxlan_ext_key_lens },
393 [OVS_TUNNEL_KEY_ATTR_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
394 [OVS_TUNNEL_KEY_ATTR_IPV6_DST] = { .len = sizeof(struct in6_addr) },
395 [OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS] = { .len = OVS_ATTR_VARIABLE },
396 [OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE] = { .len = 0 },
397 };
398
399 static const struct ovs_len_tbl
400 ovs_nsh_key_attr_lens[OVS_NSH_KEY_ATTR_MAX + 1] = {
401 [OVS_NSH_KEY_ATTR_BASE] = { .len = sizeof(struct ovs_nsh_key_base) },
402 [OVS_NSH_KEY_ATTR_MD1] = { .len = sizeof(struct ovs_nsh_key_md1) },
403 [OVS_NSH_KEY_ATTR_MD2] = { .len = OVS_ATTR_VARIABLE },
404 };
405
406 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
407 static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
408 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
409 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
410 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
411 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
412 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
413 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
414 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
415 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
416 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
417 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
418 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
419 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
420 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
421 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
422 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
423 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
424 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
425 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
426 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
427 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
428 .next = ovs_tunnel_key_lens, },
429 [OVS_KEY_ATTR_MPLS] = { .len = OVS_ATTR_VARIABLE },
430 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u32) },
431 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
432 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
433 [OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
434 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4] = {
435 .len = sizeof(struct ovs_key_ct_tuple_ipv4) },
436 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6] = {
437 .len = sizeof(struct ovs_key_ct_tuple_ipv6) },
438 [OVS_KEY_ATTR_NSH] = { .len = OVS_ATTR_NESTED,
439 .next = ovs_nsh_key_attr_lens, },
440 };
441
check_attr_len(unsigned int attr_len,unsigned int expected_len)442 static 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
is_all_zero(const u8 * fp,size_t size)449 static 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
__parse_flow_nlattrs(const struct nlattr * attr,const struct nlattr * a[],u64 * attrsp,bool log,bool nz)463 static 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
parse_flow_mask_nlattrs(const struct nlattr * attr,const struct nlattr * a[],u64 * attrsp,bool log)508 static 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
parse_flow_nlattrs(const struct nlattr * attr,const struct nlattr * a[],u64 * attrsp,bool log)515 int 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
genev_tun_opt_from_nlattr(const struct nlattr * a,struct sw_flow_match * match,bool is_mask,bool log)521 static 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
vxlan_tun_opt_from_nlattr(const struct nlattr * attr,struct sw_flow_match * match,bool is_mask,bool log)572 static 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
erspan_tun_opt_from_nlattr(const struct nlattr * a,struct sw_flow_match * match,bool is_mask,bool log)628 static 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
ip_tun_from_nlattr(const struct nlattr * attr,struct sw_flow_match * match,bool is_mask,bool log)655 static 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
vxlan_opt_to_nlattr(struct sk_buff * skb,const void * tun_opts,int swkey_tun_opts_len)842 static 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
__ip_tun_to_nlattr(struct sk_buff * skb,const struct ip_tunnel_key * output,const void * tun_opts,int swkey_tun_opts_len,unsigned short tun_proto,u8 mode)859 static int __ip_tun_to_nlattr(struct sk_buff *skb,
860 const struct ip_tunnel_key *output,
861 const void *tun_opts, int swkey_tun_opts_len,
862 unsigned short tun_proto, u8 mode)
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
ip_tun_to_nlattr(struct sk_buff * skb,const struct ip_tunnel_key * output,const void * tun_opts,int swkey_tun_opts_len,unsigned short tun_proto,u8 mode)932 static int ip_tun_to_nlattr(struct sk_buff *skb,
933 const struct ip_tunnel_key *output,
934 const void *tun_opts, int swkey_tun_opts_len,
935 unsigned short tun_proto, u8 mode)
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
ovs_nla_put_tunnel_info(struct sk_buff * skb,struct ip_tunnel_info * tun_info)953 int 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
encode_vlan_from_nlattrs(struct sw_flow_match * match,const struct nlattr * a[],bool is_mask,bool inner)962 static 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
validate_vlan_from_nlattrs(const struct sw_flow_match * match,u64 key_attrs,bool inner,const struct nlattr ** a,bool log)985 static int validate_vlan_from_nlattrs(const struct sw_flow_match *match,
986 u64 key_attrs, bool inner,
987 const struct nlattr **a, bool log)
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
validate_vlan_mask_from_nlattrs(const struct sw_flow_match * match,u64 key_attrs,bool inner,const struct nlattr ** a,bool log)1023 static int validate_vlan_mask_from_nlattrs(const struct sw_flow_match *match,
1024 u64 key_attrs, bool inner,
1025 const struct nlattr **a, bool log)
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
__parse_vlan_from_nlattrs(struct sw_flow_match * match,u64 * key_attrs,bool inner,const struct nlattr ** a,bool is_mask,bool log)1065 static int __parse_vlan_from_nlattrs(struct sw_flow_match *match,
1066 u64 *key_attrs, bool inner,
1067 const struct nlattr **a, bool is_mask,
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
parse_vlan_from_nlattrs(struct sw_flow_match * match,u64 * key_attrs,const struct nlattr ** a,bool is_mask,bool log)1100 static int parse_vlan_from_nlattrs(struct sw_flow_match *match,
1101 u64 *key_attrs, const struct nlattr **a,
1102 bool is_mask, bool log)
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
parse_eth_type_from_nlattrs(struct sw_flow_match * match,u64 * attrs,const struct nlattr ** a,bool is_mask,bool log)1123 static int parse_eth_type_from_nlattrs(struct sw_flow_match *match,
1124 u64 *attrs, const struct nlattr **a,
1125 bool is_mask, bool log)
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
metadata_from_nlattrs(struct net * net,struct sw_flow_match * match,u64 * attrs,const struct nlattr ** a,bool is_mask,bool log)1144 static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
1145 u64 *attrs, const struct nlattr **a,
1146 bool is_mask, bool log)
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
nsh_hdr_from_nlattr(const struct nlattr * attr,struct nshhdr * nh,size_t size)1282 int 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
nsh_key_from_nlattr(const struct nlattr * attr,struct ovs_key_nsh * nsh,struct ovs_key_nsh * nsh_mask)1336 int 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
nsh_key_put_from_nlattr(const struct nlattr * attr,struct sw_flow_match * match,bool is_mask,bool is_push_nsh,bool log)1376 static int nsh_key_put_from_nlattr(const struct nlattr *attr,
1377 struct sw_flow_match *match, bool is_mask,
1378 bool is_push_nsh, bool log)
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
ovs_key_from_nlattrs(struct net * net,struct sw_flow_match * match,u64 attrs,const struct nlattr ** a,bool is_mask,bool log)1497 static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
1498 u64 attrs, const struct nlattr **a,
1499 bool is_mask, bool log)
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
nlattr_set(struct nlattr * attr,u8 val,const struct ovs_len_tbl * tbl)1736 static 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
mask_set_nlattr(struct nlattr * attr,u8 val)1754 static 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 */
ovs_nla_get_match(struct net * net,struct sw_flow_match * match,const struct nlattr * nla_key,const struct nlattr * nla_mask,bool log)1775 int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1776 const struct nlattr *nla_key,
1777 const struct nlattr *nla_mask,
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
1849 free_newmask:
1850 kfree(newmask);
1851 return err;
1852 }
1853
get_ufid_len(const struct nlattr * attr,bool log)1854 static 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 */
ovs_nla_get_ufid(struct sw_flow_id * sfid,const struct nlattr * attr,bool log)1874 bool 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
ovs_nla_get_identifier(struct sw_flow_id * sfid,const struct nlattr * ufid,const struct sw_flow_key * key,bool log)1884 int 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
ovs_nla_get_ufid_flags(const struct nlattr * attr)1902 u32 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
ovs_nla_get_flow_metadata(struct net * net,const struct nlattr * a[OVS_KEY_ATTR_MAX+1],u64 attrs,struct sw_flow_key * key,bool log)1927 int ovs_nla_get_flow_metadata(struct net *net,
1928 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1],
1929 u64 attrs, struct sw_flow_key *key, bool log)
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
ovs_nla_put_vlan(struct sk_buff * skb,const struct vlan_head * vh,bool is_mask)1948 static 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
nsh_key_to_nlattr(const struct ovs_key_nsh * nsh,bool is_mask,struct sk_buff * skb)1959 static 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
1983 nla_put_failure:
1984 return -EMSGSIZE;
1985 }
1986
__ovs_nla_put_key(const struct sw_flow_key * swkey,const struct sw_flow_key * output,bool is_mask,struct sk_buff * skb)1987 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1988 const struct sw_flow_key *output, bool is_mask,
1989 struct sk_buff *skb)
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 (swkey->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
2205 swkey->tp.src == htons(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
2220 unencap:
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
2228 nla_put_failure:
2229 return -EMSGSIZE;
2230 }
2231
ovs_nla_put_key(const struct sw_flow_key * swkey,const struct sw_flow_key * output,int attr,bool is_mask,struct sk_buff * skb)2232 int ovs_nla_put_key(const struct sw_flow_key *swkey,
2233 const struct sw_flow_key *output, int attr, bool is_mask,
2234 struct sk_buff *skb)
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. */
ovs_nla_put_identifier(const struct sw_flow * flow,struct sk_buff * skb)2251 int 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. */
ovs_nla_put_masked_key(const struct sw_flow * flow,struct sk_buff * skb)2262 int 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. */
ovs_nla_put_mask(const struct sw_flow * flow,struct sk_buff * skb)2269 int 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
nla_alloc_flow_actions(int size)2277 static 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
2291 static void ovs_nla_free_nested_actions(const struct nlattr *actions, int len);
2292
ovs_nla_free_check_pkt_len_action(const struct nlattr * action)2293 static void ovs_nla_free_check_pkt_len_action(const struct nlattr *action)
2294 {
2295 const struct nlattr *a;
2296 int rem;
2297
2298 nla_for_each_nested(a, action, rem) {
2299 switch (nla_type(a)) {
2300 case OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL:
2301 case OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER:
2302 ovs_nla_free_nested_actions(nla_data(a), nla_len(a));
2303 break;
2304 }
2305 }
2306 }
2307
ovs_nla_free_clone_action(const struct nlattr * action)2308 static void ovs_nla_free_clone_action(const struct nlattr *action)
2309 {
2310 const struct nlattr *a = nla_data(action);
2311 int rem = nla_len(action);
2312
2313 switch (nla_type(a)) {
2314 case OVS_CLONE_ATTR_EXEC:
2315 /* The real list of actions follows this attribute. */
2316 a = nla_next(a, &rem);
2317 ovs_nla_free_nested_actions(a, rem);
2318 break;
2319 }
2320 }
2321
ovs_nla_free_dec_ttl_action(const struct nlattr * action)2322 static void ovs_nla_free_dec_ttl_action(const struct nlattr *action)
2323 {
2324 const struct nlattr *a = nla_data(action);
2325
2326 switch (nla_type(a)) {
2327 case OVS_DEC_TTL_ATTR_ACTION:
2328 ovs_nla_free_nested_actions(nla_data(a), nla_len(a));
2329 break;
2330 }
2331 }
2332
ovs_nla_free_sample_action(const struct nlattr * action)2333 static void ovs_nla_free_sample_action(const struct nlattr *action)
2334 {
2335 const struct nlattr *a = nla_data(action);
2336 int rem = nla_len(action);
2337
2338 switch (nla_type(a)) {
2339 case OVS_SAMPLE_ATTR_ARG:
2340 /* The real list of actions follows this attribute. */
2341 a = nla_next(a, &rem);
2342 ovs_nla_free_nested_actions(a, rem);
2343 break;
2344 }
2345 }
2346
ovs_nla_free_set_action(const struct nlattr * a)2347 static void ovs_nla_free_set_action(const struct nlattr *a)
2348 {
2349 const struct nlattr *ovs_key = nla_data(a);
2350 struct ovs_tunnel_info *ovs_tun;
2351
2352 switch (nla_type(ovs_key)) {
2353 case OVS_KEY_ATTR_TUNNEL_INFO:
2354 ovs_tun = nla_data(ovs_key);
2355 dst_release((struct dst_entry *)ovs_tun->tun_dst);
2356 break;
2357 }
2358 }
2359
ovs_nla_free_nested_actions(const struct nlattr * actions,int len)2360 static void ovs_nla_free_nested_actions(const struct nlattr *actions, int len)
2361 {
2362 const struct nlattr *a;
2363 int rem;
2364
2365 /* Whenever new actions are added, the need to update this
2366 * function should be considered.
2367 */
2368 BUILD_BUG_ON(OVS_ACTION_ATTR_MAX != 23);
2369
2370 if (!actions)
2371 return;
2372
2373 nla_for_each_attr(a, actions, len, rem) {
2374 switch (nla_type(a)) {
2375 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
2376 ovs_nla_free_check_pkt_len_action(a);
2377 break;
2378
2379 case OVS_ACTION_ATTR_CLONE:
2380 ovs_nla_free_clone_action(a);
2381 break;
2382
2383 case OVS_ACTION_ATTR_CT:
2384 ovs_ct_free_action(a);
2385 break;
2386
2387 case OVS_ACTION_ATTR_DEC_TTL:
2388 ovs_nla_free_dec_ttl_action(a);
2389 break;
2390
2391 case OVS_ACTION_ATTR_SAMPLE:
2392 ovs_nla_free_sample_action(a);
2393 break;
2394
2395 case OVS_ACTION_ATTR_SET:
2396 ovs_nla_free_set_action(a);
2397 break;
2398 }
2399 }
2400 }
2401
ovs_nla_free_flow_actions(struct sw_flow_actions * sf_acts)2402 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
2403 {
2404 if (!sf_acts)
2405 return;
2406
2407 ovs_nla_free_nested_actions(sf_acts->actions, sf_acts->actions_len);
2408 kfree(sf_acts);
2409 }
2410
__ovs_nla_free_flow_actions(struct rcu_head * head)2411 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
2412 {
2413 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
2414 }
2415
2416 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
2417 * The caller must hold rcu_read_lock for this to be sensible. */
ovs_nla_free_flow_actions_rcu(struct sw_flow_actions * sf_acts)2418 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
2419 {
2420 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
2421 }
2422
reserve_sfa_size(struct sw_flow_actions ** sfa,int attr_len,bool log)2423 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
2424 int attr_len, bool log)
2425 {
2426
2427 struct sw_flow_actions *acts;
2428 int new_acts_size;
2429 size_t req_size = NLA_ALIGN(attr_len);
2430 int next_offset = offsetof(struct sw_flow_actions, actions) +
2431 (*sfa)->actions_len;
2432
2433 if (req_size <= (ksize(*sfa) - next_offset))
2434 goto out;
2435
2436 new_acts_size = max(next_offset + req_size, ksize(*sfa) * 2);
2437
2438 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
2439 if ((next_offset + req_size) > MAX_ACTIONS_BUFSIZE) {
2440 OVS_NLERR(log, "Flow action size exceeds max %u",
2441 MAX_ACTIONS_BUFSIZE);
2442 return ERR_PTR(-EMSGSIZE);
2443 }
2444 new_acts_size = MAX_ACTIONS_BUFSIZE;
2445 }
2446
2447 acts = nla_alloc_flow_actions(new_acts_size);
2448 if (IS_ERR(acts))
2449 return (void *)acts;
2450
2451 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
2452 acts->actions_len = (*sfa)->actions_len;
2453 acts->orig_len = (*sfa)->orig_len;
2454 kfree(*sfa);
2455 *sfa = acts;
2456
2457 out:
2458 (*sfa)->actions_len += req_size;
2459 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
2460 }
2461
__add_action(struct sw_flow_actions ** sfa,int attrtype,void * data,int len,bool log)2462 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
2463 int attrtype, void *data, int len, bool log)
2464 {
2465 struct nlattr *a;
2466
2467 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
2468 if (IS_ERR(a))
2469 return a;
2470
2471 a->nla_type = attrtype;
2472 a->nla_len = nla_attr_size(len);
2473
2474 if (data)
2475 memcpy(nla_data(a), data, len);
2476 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
2477
2478 return a;
2479 }
2480
ovs_nla_add_action(struct sw_flow_actions ** sfa,int attrtype,void * data,int len,bool log)2481 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
2482 int len, bool log)
2483 {
2484 struct nlattr *a;
2485
2486 a = __add_action(sfa, attrtype, data, len, log);
2487
2488 return PTR_ERR_OR_ZERO(a);
2489 }
2490
add_nested_action_start(struct sw_flow_actions ** sfa,int attrtype,bool log)2491 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
2492 int attrtype, bool log)
2493 {
2494 int used = (*sfa)->actions_len;
2495 int err;
2496
2497 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
2498 if (err)
2499 return err;
2500
2501 return used;
2502 }
2503
add_nested_action_end(struct sw_flow_actions * sfa,int st_offset)2504 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
2505 int st_offset)
2506 {
2507 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
2508 st_offset);
2509
2510 a->nla_len = sfa->actions_len - st_offset;
2511 }
2512
2513 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2514 const struct sw_flow_key *key,
2515 struct sw_flow_actions **sfa,
2516 __be16 eth_type, __be16 vlan_tci,
2517 u32 mpls_label_count, bool log);
2518
validate_and_copy_sample(struct net * net,const struct nlattr * attr,const struct sw_flow_key * key,struct sw_flow_actions ** sfa,__be16 eth_type,__be16 vlan_tci,u32 mpls_label_count,bool log,bool last)2519 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
2520 const struct sw_flow_key *key,
2521 struct sw_flow_actions **sfa,
2522 __be16 eth_type, __be16 vlan_tci,
2523 u32 mpls_label_count, bool log, bool last)
2524 {
2525 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
2526 const struct nlattr *probability, *actions;
2527 const struct nlattr *a;
2528 int rem, start, err;
2529 struct sample_arg arg;
2530
2531 memset(attrs, 0, sizeof(attrs));
2532 nla_for_each_nested(a, attr, rem) {
2533 int type = nla_type(a);
2534 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
2535 return -EINVAL;
2536 attrs[type] = a;
2537 }
2538 if (rem)
2539 return -EINVAL;
2540
2541 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
2542 if (!probability || nla_len(probability) != sizeof(u32))
2543 return -EINVAL;
2544
2545 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
2546 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
2547 return -EINVAL;
2548
2549 /* validation done, copy sample action. */
2550 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
2551 if (start < 0)
2552 return start;
2553
2554 /* When both skb and flow may be changed, put the sample
2555 * into a deferred fifo. On the other hand, if only skb
2556 * may be modified, the actions can be executed in place.
2557 *
2558 * Do this analysis at the flow installation time.
2559 * Set 'clone_action->exec' to true if the actions can be
2560 * executed without being deferred.
2561 *
2562 * If the sample is the last action, it can always be excuted
2563 * rather than deferred.
2564 */
2565 arg.exec = last || !actions_may_change_flow(actions);
2566 arg.probability = nla_get_u32(probability);
2567
2568 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_ARG, &arg, sizeof(arg),
2569 log);
2570 if (err)
2571 return err;
2572
2573 err = __ovs_nla_copy_actions(net, actions, key, sfa,
2574 eth_type, vlan_tci, mpls_label_count, log);
2575
2576 if (err)
2577 return err;
2578
2579 add_nested_action_end(*sfa, start);
2580
2581 return 0;
2582 }
2583
validate_and_copy_dec_ttl(struct net * net,const struct nlattr * attr,const struct sw_flow_key * key,struct sw_flow_actions ** sfa,__be16 eth_type,__be16 vlan_tci,u32 mpls_label_count,bool log)2584 static int validate_and_copy_dec_ttl(struct net *net,
2585 const struct nlattr *attr,
2586 const struct sw_flow_key *key,
2587 struct sw_flow_actions **sfa,
2588 __be16 eth_type, __be16 vlan_tci,
2589 u32 mpls_label_count, bool log)
2590 {
2591 const struct nlattr *attrs[OVS_DEC_TTL_ATTR_MAX + 1];
2592 int start, action_start, err, rem;
2593 const struct nlattr *a, *actions;
2594
2595 memset(attrs, 0, sizeof(attrs));
2596 nla_for_each_nested(a, attr, rem) {
2597 int type = nla_type(a);
2598
2599 /* Ignore unknown attributes to be future proof. */
2600 if (type > OVS_DEC_TTL_ATTR_MAX)
2601 continue;
2602
2603 if (!type || attrs[type])
2604 return -EINVAL;
2605
2606 attrs[type] = a;
2607 }
2608
2609 actions = attrs[OVS_DEC_TTL_ATTR_ACTION];
2610 if (rem || !actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
2611 return -EINVAL;
2612
2613 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_DEC_TTL, log);
2614 if (start < 0)
2615 return start;
2616
2617 action_start = add_nested_action_start(sfa, OVS_DEC_TTL_ATTR_ACTION, log);
2618 if (action_start < 0)
2619 return action_start;
2620
2621 err = __ovs_nla_copy_actions(net, actions, key, sfa, eth_type,
2622 vlan_tci, mpls_label_count, log);
2623 if (err)
2624 return err;
2625
2626 add_nested_action_end(*sfa, action_start);
2627 add_nested_action_end(*sfa, start);
2628 return 0;
2629 }
2630
validate_and_copy_clone(struct net * net,const struct nlattr * attr,const struct sw_flow_key * key,struct sw_flow_actions ** sfa,__be16 eth_type,__be16 vlan_tci,u32 mpls_label_count,bool log,bool last)2631 static int validate_and_copy_clone(struct net *net,
2632 const struct nlattr *attr,
2633 const struct sw_flow_key *key,
2634 struct sw_flow_actions **sfa,
2635 __be16 eth_type, __be16 vlan_tci,
2636 u32 mpls_label_count, bool log, bool last)
2637 {
2638 int start, err;
2639 u32 exec;
2640
2641 if (nla_len(attr) && nla_len(attr) < NLA_HDRLEN)
2642 return -EINVAL;
2643
2644 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CLONE, log);
2645 if (start < 0)
2646 return start;
2647
2648 exec = last || !actions_may_change_flow(attr);
2649
2650 err = ovs_nla_add_action(sfa, OVS_CLONE_ATTR_EXEC, &exec,
2651 sizeof(exec), log);
2652 if (err)
2653 return err;
2654
2655 err = __ovs_nla_copy_actions(net, attr, key, sfa,
2656 eth_type, vlan_tci, mpls_label_count, log);
2657 if (err)
2658 return err;
2659
2660 add_nested_action_end(*sfa, start);
2661
2662 return 0;
2663 }
2664
ovs_match_init(struct sw_flow_match * match,struct sw_flow_key * key,bool reset_key,struct sw_flow_mask * mask)2665 void ovs_match_init(struct sw_flow_match *match,
2666 struct sw_flow_key *key,
2667 bool reset_key,
2668 struct sw_flow_mask *mask)
2669 {
2670 memset(match, 0, sizeof(*match));
2671 match->key = key;
2672 match->mask = mask;
2673
2674 if (reset_key)
2675 memset(key, 0, sizeof(*key));
2676
2677 if (mask) {
2678 memset(&mask->key, 0, sizeof(mask->key));
2679 mask->range.start = mask->range.end = 0;
2680 }
2681 }
2682
validate_geneve_opts(struct sw_flow_key * key)2683 static int validate_geneve_opts(struct sw_flow_key *key)
2684 {
2685 struct geneve_opt *option;
2686 int opts_len = key->tun_opts_len;
2687 bool crit_opt = false;
2688
2689 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2690 while (opts_len > 0) {
2691 int len;
2692
2693 if (opts_len < sizeof(*option))
2694 return -EINVAL;
2695
2696 len = sizeof(*option) + option->length * 4;
2697 if (len > opts_len)
2698 return -EINVAL;
2699
2700 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2701
2702 option = (struct geneve_opt *)((u8 *)option + len);
2703 opts_len -= len;
2704 }
2705
2706 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
2707
2708 return 0;
2709 }
2710
validate_and_copy_set_tun(const struct nlattr * attr,struct sw_flow_actions ** sfa,bool log)2711 static int validate_and_copy_set_tun(const struct nlattr *attr,
2712 struct sw_flow_actions **sfa, bool log)
2713 {
2714 struct sw_flow_match match;
2715 struct sw_flow_key key;
2716 struct metadata_dst *tun_dst;
2717 struct ip_tunnel_info *tun_info;
2718 struct ovs_tunnel_info *ovs_tun;
2719 struct nlattr *a;
2720 int err = 0, start, opts_type;
2721 __be16 dst_opt_type;
2722
2723 dst_opt_type = 0;
2724 ovs_match_init(&match, &key, true, NULL);
2725 opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2726 if (opts_type < 0)
2727 return opts_type;
2728
2729 if (key.tun_opts_len) {
2730 switch (opts_type) {
2731 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2732 err = validate_geneve_opts(&key);
2733 if (err < 0)
2734 return err;
2735 dst_opt_type = TUNNEL_GENEVE_OPT;
2736 break;
2737 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
2738 dst_opt_type = TUNNEL_VXLAN_OPT;
2739 break;
2740 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
2741 dst_opt_type = TUNNEL_ERSPAN_OPT;
2742 break;
2743 }
2744 }
2745
2746 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2747 if (start < 0)
2748 return start;
2749
2750 tun_dst = metadata_dst_alloc(key.tun_opts_len, METADATA_IP_TUNNEL,
2751 GFP_KERNEL);
2752
2753 if (!tun_dst)
2754 return -ENOMEM;
2755
2756 err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2757 if (err) {
2758 dst_release((struct dst_entry *)tun_dst);
2759 return err;
2760 }
2761
2762 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2763 sizeof(*ovs_tun), log);
2764 if (IS_ERR(a)) {
2765 dst_release((struct dst_entry *)tun_dst);
2766 return PTR_ERR(a);
2767 }
2768
2769 ovs_tun = nla_data(a);
2770 ovs_tun->tun_dst = tun_dst;
2771
2772 tun_info = &tun_dst->u.tun_info;
2773 tun_info->mode = IP_TUNNEL_INFO_TX;
2774 if (key.tun_proto == AF_INET6)
2775 tun_info->mode |= IP_TUNNEL_INFO_IPV6;
2776 else if (key.tun_proto == AF_INET && key.tun_key.u.ipv4.dst == 0)
2777 tun_info->mode |= IP_TUNNEL_INFO_BRIDGE;
2778 tun_info->key = key.tun_key;
2779
2780 /* We need to store the options in the action itself since
2781 * everything else will go away after flow setup. We can append
2782 * it to tun_info and then point there.
2783 */
2784 ip_tunnel_info_opts_set(tun_info,
2785 TUN_METADATA_OPTS(&key, key.tun_opts_len),
2786 key.tun_opts_len, dst_opt_type);
2787 add_nested_action_end(*sfa, start);
2788
2789 return err;
2790 }
2791
validate_nsh(const struct nlattr * attr,bool is_mask,bool is_push_nsh,bool log)2792 static bool validate_nsh(const struct nlattr *attr, bool is_mask,
2793 bool is_push_nsh, bool log)
2794 {
2795 struct sw_flow_match match;
2796 struct sw_flow_key key;
2797 int ret = 0;
2798
2799 ovs_match_init(&match, &key, true, NULL);
2800 ret = nsh_key_put_from_nlattr(attr, &match, is_mask,
2801 is_push_nsh, log);
2802 return !ret;
2803 }
2804
2805 /* Return false if there are any non-masked bits set.
2806 * Mask follows data immediately, before any netlink padding.
2807 */
validate_masked(u8 * data,int len)2808 static bool validate_masked(u8 *data, int len)
2809 {
2810 u8 *mask = data + len;
2811
2812 while (len--)
2813 if (*data++ & ~*mask++)
2814 return false;
2815
2816 return true;
2817 }
2818
validate_set(const struct nlattr * a,const struct sw_flow_key * flow_key,struct sw_flow_actions ** sfa,bool * skip_copy,u8 mac_proto,__be16 eth_type,bool masked,bool log)2819 static int validate_set(const struct nlattr *a,
2820 const struct sw_flow_key *flow_key,
2821 struct sw_flow_actions **sfa, bool *skip_copy,
2822 u8 mac_proto, __be16 eth_type, bool masked, bool log)
2823 {
2824 const struct nlattr *ovs_key = nla_data(a);
2825 int key_type = nla_type(ovs_key);
2826 size_t key_len;
2827
2828 /* There can be only one key in a action */
2829 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2830 return -EINVAL;
2831
2832 key_len = nla_len(ovs_key);
2833 if (masked)
2834 key_len /= 2;
2835
2836 if (key_type > OVS_KEY_ATTR_MAX ||
2837 !check_attr_len(key_len, ovs_key_lens[key_type].len))
2838 return -EINVAL;
2839
2840 if (masked && !validate_masked(nla_data(ovs_key), key_len))
2841 return -EINVAL;
2842
2843 switch (key_type) {
2844 case OVS_KEY_ATTR_PRIORITY:
2845 case OVS_KEY_ATTR_SKB_MARK:
2846 case OVS_KEY_ATTR_CT_MARK:
2847 case OVS_KEY_ATTR_CT_LABELS:
2848 break;
2849
2850 case OVS_KEY_ATTR_ETHERNET:
2851 if (mac_proto != MAC_PROTO_ETHERNET)
2852 return -EINVAL;
2853 break;
2854
2855 case OVS_KEY_ATTR_TUNNEL: {
2856 int err;
2857
2858 if (masked)
2859 return -EINVAL; /* Masked tunnel set not supported. */
2860
2861 *skip_copy = true;
2862 err = validate_and_copy_set_tun(a, sfa, log);
2863 if (err)
2864 return err;
2865 break;
2866 }
2867 case OVS_KEY_ATTR_IPV4: {
2868 const struct ovs_key_ipv4 *ipv4_key;
2869
2870 if (eth_type != htons(ETH_P_IP))
2871 return -EINVAL;
2872
2873 ipv4_key = nla_data(ovs_key);
2874
2875 if (masked) {
2876 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2877
2878 /* Non-writeable fields. */
2879 if (mask->ipv4_proto || mask->ipv4_frag)
2880 return -EINVAL;
2881 } else {
2882 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2883 return -EINVAL;
2884
2885 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2886 return -EINVAL;
2887 }
2888 break;
2889 }
2890 case OVS_KEY_ATTR_IPV6: {
2891 const struct ovs_key_ipv6 *ipv6_key;
2892
2893 if (eth_type != htons(ETH_P_IPV6))
2894 return -EINVAL;
2895
2896 ipv6_key = nla_data(ovs_key);
2897
2898 if (masked) {
2899 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2900
2901 /* Non-writeable fields. */
2902 if (mask->ipv6_proto || mask->ipv6_frag)
2903 return -EINVAL;
2904
2905 /* Invalid bits in the flow label mask? */
2906 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2907 return -EINVAL;
2908 } else {
2909 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2910 return -EINVAL;
2911
2912 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2913 return -EINVAL;
2914 }
2915 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2916 return -EINVAL;
2917
2918 break;
2919 }
2920 case OVS_KEY_ATTR_TCP:
2921 if ((eth_type != htons(ETH_P_IP) &&
2922 eth_type != htons(ETH_P_IPV6)) ||
2923 flow_key->ip.proto != IPPROTO_TCP)
2924 return -EINVAL;
2925
2926 break;
2927
2928 case OVS_KEY_ATTR_UDP:
2929 if ((eth_type != htons(ETH_P_IP) &&
2930 eth_type != htons(ETH_P_IPV6)) ||
2931 flow_key->ip.proto != IPPROTO_UDP)
2932 return -EINVAL;
2933
2934 break;
2935
2936 case OVS_KEY_ATTR_MPLS:
2937 if (!eth_p_mpls(eth_type))
2938 return -EINVAL;
2939 break;
2940
2941 case OVS_KEY_ATTR_SCTP:
2942 if ((eth_type != htons(ETH_P_IP) &&
2943 eth_type != htons(ETH_P_IPV6)) ||
2944 flow_key->ip.proto != IPPROTO_SCTP)
2945 return -EINVAL;
2946
2947 break;
2948
2949 case OVS_KEY_ATTR_NSH:
2950 if (eth_type != htons(ETH_P_NSH))
2951 return -EINVAL;
2952 if (!validate_nsh(nla_data(a), masked, false, log))
2953 return -EINVAL;
2954 break;
2955
2956 default:
2957 return -EINVAL;
2958 }
2959
2960 /* Convert non-masked non-tunnel set actions to masked set actions. */
2961 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2962 int start, len = key_len * 2;
2963 struct nlattr *at;
2964
2965 *skip_copy = true;
2966
2967 start = add_nested_action_start(sfa,
2968 OVS_ACTION_ATTR_SET_TO_MASKED,
2969 log);
2970 if (start < 0)
2971 return start;
2972
2973 at = __add_action(sfa, key_type, NULL, len, log);
2974 if (IS_ERR(at))
2975 return PTR_ERR(at);
2976
2977 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2978 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
2979 /* Clear non-writeable bits from otherwise writeable fields. */
2980 if (key_type == OVS_KEY_ATTR_IPV6) {
2981 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2982
2983 mask->ipv6_label &= htonl(0x000FFFFF);
2984 }
2985 add_nested_action_end(*sfa, start);
2986 }
2987
2988 return 0;
2989 }
2990
validate_userspace(const struct nlattr * attr)2991 static int validate_userspace(const struct nlattr *attr)
2992 {
2993 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2994 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2995 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2996 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2997 };
2998 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2999 int error;
3000
3001 error = nla_parse_nested_deprecated(a, OVS_USERSPACE_ATTR_MAX, attr,
3002 userspace_policy, NULL);
3003 if (error)
3004 return error;
3005
3006 if (!a[OVS_USERSPACE_ATTR_PID] ||
3007 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
3008 return -EINVAL;
3009
3010 return 0;
3011 }
3012
3013 static const struct nla_policy cpl_policy[OVS_CHECK_PKT_LEN_ATTR_MAX + 1] = {
3014 [OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] = {.type = NLA_U16 },
3015 [OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER] = {.type = NLA_NESTED },
3016 [OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL] = {.type = NLA_NESTED },
3017 };
3018
validate_and_copy_check_pkt_len(struct net * net,const struct nlattr * attr,const struct sw_flow_key * key,struct sw_flow_actions ** sfa,__be16 eth_type,__be16 vlan_tci,u32 mpls_label_count,bool log,bool last)3019 static int validate_and_copy_check_pkt_len(struct net *net,
3020 const struct nlattr *attr,
3021 const struct sw_flow_key *key,
3022 struct sw_flow_actions **sfa,
3023 __be16 eth_type, __be16 vlan_tci,
3024 u32 mpls_label_count,
3025 bool log, bool last)
3026 {
3027 const struct nlattr *acts_if_greater, *acts_if_lesser_eq;
3028 struct nlattr *a[OVS_CHECK_PKT_LEN_ATTR_MAX + 1];
3029 struct check_pkt_len_arg arg;
3030 int nested_acts_start;
3031 int start, err;
3032
3033 err = nla_parse_deprecated_strict(a, OVS_CHECK_PKT_LEN_ATTR_MAX,
3034 nla_data(attr), nla_len(attr),
3035 cpl_policy, NULL);
3036 if (err)
3037 return err;
3038
3039 if (!a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] ||
3040 !nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]))
3041 return -EINVAL;
3042
3043 acts_if_lesser_eq = a[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL];
3044 acts_if_greater = a[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER];
3045
3046 /* Both the nested action should be present. */
3047 if (!acts_if_greater || !acts_if_lesser_eq)
3048 return -EINVAL;
3049
3050 /* validation done, copy the nested actions. */
3051 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CHECK_PKT_LEN,
3052 log);
3053 if (start < 0)
3054 return start;
3055
3056 arg.pkt_len = nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]);
3057 arg.exec_for_lesser_equal =
3058 last || !actions_may_change_flow(acts_if_lesser_eq);
3059 arg.exec_for_greater =
3060 last || !actions_may_change_flow(acts_if_greater);
3061
3062 err = ovs_nla_add_action(sfa, OVS_CHECK_PKT_LEN_ATTR_ARG, &arg,
3063 sizeof(arg), log);
3064 if (err)
3065 return err;
3066
3067 nested_acts_start = add_nested_action_start(sfa,
3068 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL, log);
3069 if (nested_acts_start < 0)
3070 return nested_acts_start;
3071
3072 err = __ovs_nla_copy_actions(net, acts_if_lesser_eq, key, sfa,
3073 eth_type, vlan_tci, mpls_label_count, log);
3074
3075 if (err)
3076 return err;
3077
3078 add_nested_action_end(*sfa, nested_acts_start);
3079
3080 nested_acts_start = add_nested_action_start(sfa,
3081 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER, log);
3082 if (nested_acts_start < 0)
3083 return nested_acts_start;
3084
3085 err = __ovs_nla_copy_actions(net, acts_if_greater, key, sfa,
3086 eth_type, vlan_tci, mpls_label_count, log);
3087
3088 if (err)
3089 return err;
3090
3091 add_nested_action_end(*sfa, nested_acts_start);
3092 add_nested_action_end(*sfa, start);
3093 return 0;
3094 }
3095
copy_action(const struct nlattr * from,struct sw_flow_actions ** sfa,bool log)3096 static int copy_action(const struct nlattr *from,
3097 struct sw_flow_actions **sfa, bool log)
3098 {
3099 int totlen = NLA_ALIGN(from->nla_len);
3100 struct nlattr *to;
3101
3102 to = reserve_sfa_size(sfa, from->nla_len, log);
3103 if (IS_ERR(to))
3104 return PTR_ERR(to);
3105
3106 memcpy(to, from, totlen);
3107 return 0;
3108 }
3109
__ovs_nla_copy_actions(struct net * net,const struct nlattr * attr,const struct sw_flow_key * key,struct sw_flow_actions ** sfa,__be16 eth_type,__be16 vlan_tci,u32 mpls_label_count,bool log)3110 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3111 const struct sw_flow_key *key,
3112 struct sw_flow_actions **sfa,
3113 __be16 eth_type, __be16 vlan_tci,
3114 u32 mpls_label_count, bool log)
3115 {
3116 u8 mac_proto = ovs_key_mac_proto(key);
3117 const struct nlattr *a;
3118 int rem, err;
3119
3120 nla_for_each_nested(a, attr, rem) {
3121 /* Expected argument lengths, (u32)-1 for variable length. */
3122 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
3123 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
3124 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
3125 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
3126 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
3127 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
3128 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
3129 [OVS_ACTION_ATTR_POP_VLAN] = 0,
3130 [OVS_ACTION_ATTR_SET] = (u32)-1,
3131 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
3132 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
3133 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
3134 [OVS_ACTION_ATTR_CT] = (u32)-1,
3135 [OVS_ACTION_ATTR_CT_CLEAR] = 0,
3136 [OVS_ACTION_ATTR_TRUNC] = sizeof(struct ovs_action_trunc),
3137 [OVS_ACTION_ATTR_PUSH_ETH] = sizeof(struct ovs_action_push_eth),
3138 [OVS_ACTION_ATTR_POP_ETH] = 0,
3139 [OVS_ACTION_ATTR_PUSH_NSH] = (u32)-1,
3140 [OVS_ACTION_ATTR_POP_NSH] = 0,
3141 [OVS_ACTION_ATTR_METER] = sizeof(u32),
3142 [OVS_ACTION_ATTR_CLONE] = (u32)-1,
3143 [OVS_ACTION_ATTR_CHECK_PKT_LEN] = (u32)-1,
3144 [OVS_ACTION_ATTR_ADD_MPLS] = sizeof(struct ovs_action_add_mpls),
3145 [OVS_ACTION_ATTR_DEC_TTL] = (u32)-1,
3146 };
3147 const struct ovs_action_push_vlan *vlan;
3148 int type = nla_type(a);
3149 bool skip_copy;
3150
3151 if (type > OVS_ACTION_ATTR_MAX ||
3152 (action_lens[type] != nla_len(a) &&
3153 action_lens[type] != (u32)-1))
3154 return -EINVAL;
3155
3156 skip_copy = false;
3157 switch (type) {
3158 case OVS_ACTION_ATTR_UNSPEC:
3159 return -EINVAL;
3160
3161 case OVS_ACTION_ATTR_USERSPACE:
3162 err = validate_userspace(a);
3163 if (err)
3164 return err;
3165 break;
3166
3167 case OVS_ACTION_ATTR_OUTPUT:
3168 if (nla_get_u32(a) >= DP_MAX_PORTS)
3169 return -EINVAL;
3170 break;
3171
3172 case OVS_ACTION_ATTR_TRUNC: {
3173 const struct ovs_action_trunc *trunc = nla_data(a);
3174
3175 if (trunc->max_len < ETH_HLEN)
3176 return -EINVAL;
3177 break;
3178 }
3179
3180 case OVS_ACTION_ATTR_HASH: {
3181 const struct ovs_action_hash *act_hash = nla_data(a);
3182
3183 switch (act_hash->hash_alg) {
3184 case OVS_HASH_ALG_L4:
3185 break;
3186 default:
3187 return -EINVAL;
3188 }
3189
3190 break;
3191 }
3192
3193 case OVS_ACTION_ATTR_POP_VLAN:
3194 if (mac_proto != MAC_PROTO_ETHERNET)
3195 return -EINVAL;
3196 vlan_tci = htons(0);
3197 break;
3198
3199 case OVS_ACTION_ATTR_PUSH_VLAN:
3200 if (mac_proto != MAC_PROTO_ETHERNET)
3201 return -EINVAL;
3202 vlan = nla_data(a);
3203 if (!eth_type_vlan(vlan->vlan_tpid))
3204 return -EINVAL;
3205 if (!(vlan->vlan_tci & htons(VLAN_CFI_MASK)))
3206 return -EINVAL;
3207 vlan_tci = vlan->vlan_tci;
3208 break;
3209
3210 case OVS_ACTION_ATTR_RECIRC:
3211 break;
3212
3213 case OVS_ACTION_ATTR_ADD_MPLS: {
3214 const struct ovs_action_add_mpls *mpls = nla_data(a);
3215
3216 if (!eth_p_mpls(mpls->mpls_ethertype))
3217 return -EINVAL;
3218
3219 if (mpls->tun_flags & OVS_MPLS_L3_TUNNEL_FLAG_MASK) {
3220 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3221 (eth_type != htons(ETH_P_IP) &&
3222 eth_type != htons(ETH_P_IPV6) &&
3223 eth_type != htons(ETH_P_ARP) &&
3224 eth_type != htons(ETH_P_RARP) &&
3225 !eth_p_mpls(eth_type)))
3226 return -EINVAL;
3227 mpls_label_count++;
3228 } else {
3229 if (mac_proto == MAC_PROTO_ETHERNET) {
3230 mpls_label_count = 1;
3231 mac_proto = MAC_PROTO_NONE;
3232 } else {
3233 mpls_label_count++;
3234 }
3235 }
3236 eth_type = mpls->mpls_ethertype;
3237 break;
3238 }
3239
3240 case OVS_ACTION_ATTR_PUSH_MPLS: {
3241 const struct ovs_action_push_mpls *mpls = nla_data(a);
3242
3243 if (!eth_p_mpls(mpls->mpls_ethertype))
3244 return -EINVAL;
3245 /* Prohibit push MPLS other than to a white list
3246 * for packets that have a known tag order.
3247 */
3248 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3249 (eth_type != htons(ETH_P_IP) &&
3250 eth_type != htons(ETH_P_IPV6) &&
3251 eth_type != htons(ETH_P_ARP) &&
3252 eth_type != htons(ETH_P_RARP) &&
3253 !eth_p_mpls(eth_type)))
3254 return -EINVAL;
3255 eth_type = mpls->mpls_ethertype;
3256 mpls_label_count++;
3257 break;
3258 }
3259
3260 case OVS_ACTION_ATTR_POP_MPLS: {
3261 __be16 proto;
3262 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3263 !eth_p_mpls(eth_type))
3264 return -EINVAL;
3265
3266 /* Disallow subsequent L2.5+ set actions and mpls_pop
3267 * actions once the last MPLS label in the packet is
3268 * is popped as there is no check here to ensure that
3269 * the new eth type is valid and thus set actions could
3270 * write off the end of the packet or otherwise corrupt
3271 * it.
3272 *
3273 * Support for these actions is planned using packet
3274 * recirculation.
3275 */
3276 proto = nla_get_be16(a);
3277
3278 if (proto == htons(ETH_P_TEB) &&
3279 mac_proto != MAC_PROTO_NONE)
3280 return -EINVAL;
3281
3282 mpls_label_count--;
3283
3284 if (!eth_p_mpls(proto) || !mpls_label_count)
3285 eth_type = htons(0);
3286 else
3287 eth_type = proto;
3288
3289 break;
3290 }
3291
3292 case OVS_ACTION_ATTR_SET:
3293 err = validate_set(a, key, sfa,
3294 &skip_copy, mac_proto, eth_type,
3295 false, log);
3296 if (err)
3297 return err;
3298 break;
3299
3300 case OVS_ACTION_ATTR_SET_MASKED:
3301 err = validate_set(a, key, sfa,
3302 &skip_copy, mac_proto, eth_type,
3303 true, log);
3304 if (err)
3305 return err;
3306 break;
3307
3308 case OVS_ACTION_ATTR_SAMPLE: {
3309 bool last = nla_is_last(a, rem);
3310
3311 err = validate_and_copy_sample(net, a, key, sfa,
3312 eth_type, vlan_tci,
3313 mpls_label_count,
3314 log, last);
3315 if (err)
3316 return err;
3317 skip_copy = true;
3318 break;
3319 }
3320
3321 case OVS_ACTION_ATTR_CT:
3322 err = ovs_ct_copy_action(net, a, key, sfa, log);
3323 if (err)
3324 return err;
3325 skip_copy = true;
3326 break;
3327
3328 case OVS_ACTION_ATTR_CT_CLEAR:
3329 break;
3330
3331 case OVS_ACTION_ATTR_PUSH_ETH:
3332 /* Disallow pushing an Ethernet header if one
3333 * is already present */
3334 if (mac_proto != MAC_PROTO_NONE)
3335 return -EINVAL;
3336 mac_proto = MAC_PROTO_ETHERNET;
3337 break;
3338
3339 case OVS_ACTION_ATTR_POP_ETH:
3340 if (mac_proto != MAC_PROTO_ETHERNET)
3341 return -EINVAL;
3342 if (vlan_tci & htons(VLAN_CFI_MASK))
3343 return -EINVAL;
3344 mac_proto = MAC_PROTO_NONE;
3345 break;
3346
3347 case OVS_ACTION_ATTR_PUSH_NSH:
3348 if (mac_proto != MAC_PROTO_ETHERNET) {
3349 u8 next_proto;
3350
3351 next_proto = tun_p_from_eth_p(eth_type);
3352 if (!next_proto)
3353 return -EINVAL;
3354 }
3355 mac_proto = MAC_PROTO_NONE;
3356 if (!validate_nsh(nla_data(a), false, true, true))
3357 return -EINVAL;
3358 break;
3359
3360 case OVS_ACTION_ATTR_POP_NSH: {
3361 __be16 inner_proto;
3362
3363 if (eth_type != htons(ETH_P_NSH))
3364 return -EINVAL;
3365 inner_proto = tun_p_to_eth_p(key->nsh.base.np);
3366 if (!inner_proto)
3367 return -EINVAL;
3368 if (key->nsh.base.np == TUN_P_ETHERNET)
3369 mac_proto = MAC_PROTO_ETHERNET;
3370 else
3371 mac_proto = MAC_PROTO_NONE;
3372 break;
3373 }
3374
3375 case OVS_ACTION_ATTR_METER:
3376 /* Non-existent meters are simply ignored. */
3377 break;
3378
3379 case OVS_ACTION_ATTR_CLONE: {
3380 bool last = nla_is_last(a, rem);
3381
3382 err = validate_and_copy_clone(net, a, key, sfa,
3383 eth_type, vlan_tci,
3384 mpls_label_count,
3385 log, last);
3386 if (err)
3387 return err;
3388 skip_copy = true;
3389 break;
3390 }
3391
3392 case OVS_ACTION_ATTR_CHECK_PKT_LEN: {
3393 bool last = nla_is_last(a, rem);
3394
3395 err = validate_and_copy_check_pkt_len(net, a, key, sfa,
3396 eth_type,
3397 vlan_tci,
3398 mpls_label_count,
3399 log, last);
3400 if (err)
3401 return err;
3402 skip_copy = true;
3403 break;
3404 }
3405
3406 case OVS_ACTION_ATTR_DEC_TTL:
3407 err = validate_and_copy_dec_ttl(net, a, key, sfa,
3408 eth_type, vlan_tci,
3409 mpls_label_count, log);
3410 if (err)
3411 return err;
3412 skip_copy = true;
3413 break;
3414
3415 default:
3416 OVS_NLERR(log, "Unknown Action type %d", type);
3417 return -EINVAL;
3418 }
3419 if (!skip_copy) {
3420 err = copy_action(a, sfa, log);
3421 if (err)
3422 return err;
3423 }
3424 }
3425
3426 if (rem > 0)
3427 return -EINVAL;
3428
3429 return 0;
3430 }
3431
3432 /* 'key' must be the masked key. */
ovs_nla_copy_actions(struct net * net,const struct nlattr * attr,const struct sw_flow_key * key,struct sw_flow_actions ** sfa,bool log)3433 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3434 const struct sw_flow_key *key,
3435 struct sw_flow_actions **sfa, bool log)
3436 {
3437 int err;
3438 u32 mpls_label_count = 0;
3439
3440 *sfa = nla_alloc_flow_actions(min(nla_len(attr), MAX_ACTIONS_BUFSIZE));
3441 if (IS_ERR(*sfa))
3442 return PTR_ERR(*sfa);
3443
3444 if (eth_p_mpls(key->eth.type))
3445 mpls_label_count = hweight_long(key->mpls.num_labels_mask);
3446
3447 (*sfa)->orig_len = nla_len(attr);
3448 err = __ovs_nla_copy_actions(net, attr, key, sfa, key->eth.type,
3449 key->eth.vlan.tci, mpls_label_count, log);
3450 if (err)
3451 ovs_nla_free_flow_actions(*sfa);
3452
3453 return err;
3454 }
3455
sample_action_to_attr(const struct nlattr * attr,struct sk_buff * skb)3456 static int sample_action_to_attr(const struct nlattr *attr,
3457 struct sk_buff *skb)
3458 {
3459 struct nlattr *start, *ac_start = NULL, *sample_arg;
3460 int err = 0, rem = nla_len(attr);
3461 const struct sample_arg *arg;
3462 struct nlattr *actions;
3463
3464 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SAMPLE);
3465 if (!start)
3466 return -EMSGSIZE;
3467
3468 sample_arg = nla_data(attr);
3469 arg = nla_data(sample_arg);
3470 actions = nla_next(sample_arg, &rem);
3471
3472 if (nla_put_u32(skb, OVS_SAMPLE_ATTR_PROBABILITY, arg->probability)) {
3473 err = -EMSGSIZE;
3474 goto out;
3475 }
3476
3477 ac_start = nla_nest_start_noflag(skb, OVS_SAMPLE_ATTR_ACTIONS);
3478 if (!ac_start) {
3479 err = -EMSGSIZE;
3480 goto out;
3481 }
3482
3483 err = ovs_nla_put_actions(actions, rem, skb);
3484
3485 out:
3486 if (err) {
3487 nla_nest_cancel(skb, ac_start);
3488 nla_nest_cancel(skb, start);
3489 } else {
3490 nla_nest_end(skb, ac_start);
3491 nla_nest_end(skb, start);
3492 }
3493
3494 return err;
3495 }
3496
clone_action_to_attr(const struct nlattr * attr,struct sk_buff * skb)3497 static int clone_action_to_attr(const struct nlattr *attr,
3498 struct sk_buff *skb)
3499 {
3500 struct nlattr *start;
3501 int err = 0, rem = nla_len(attr);
3502
3503 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CLONE);
3504 if (!start)
3505 return -EMSGSIZE;
3506
3507 /* Skipping the OVS_CLONE_ATTR_EXEC that is always the first attribute. */
3508 attr = nla_next(nla_data(attr), &rem);
3509 err = ovs_nla_put_actions(attr, rem, skb);
3510
3511 if (err)
3512 nla_nest_cancel(skb, start);
3513 else
3514 nla_nest_end(skb, start);
3515
3516 return err;
3517 }
3518
check_pkt_len_action_to_attr(const struct nlattr * attr,struct sk_buff * skb)3519 static int check_pkt_len_action_to_attr(const struct nlattr *attr,
3520 struct sk_buff *skb)
3521 {
3522 struct nlattr *start, *ac_start = NULL;
3523 const struct check_pkt_len_arg *arg;
3524 const struct nlattr *a, *cpl_arg;
3525 int err = 0, rem = nla_len(attr);
3526
3527 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CHECK_PKT_LEN);
3528 if (!start)
3529 return -EMSGSIZE;
3530
3531 /* The first nested attribute in 'attr' is always
3532 * 'OVS_CHECK_PKT_LEN_ATTR_ARG'.
3533 */
3534 cpl_arg = nla_data(attr);
3535 arg = nla_data(cpl_arg);
3536
3537 if (nla_put_u16(skb, OVS_CHECK_PKT_LEN_ATTR_PKT_LEN, arg->pkt_len)) {
3538 err = -EMSGSIZE;
3539 goto out;
3540 }
3541
3542 /* Second nested attribute in 'attr' is always
3543 * 'OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL'.
3544 */
3545 a = nla_next(cpl_arg, &rem);
3546 ac_start = nla_nest_start_noflag(skb,
3547 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL);
3548 if (!ac_start) {
3549 err = -EMSGSIZE;
3550 goto out;
3551 }
3552
3553 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3554 if (err) {
3555 nla_nest_cancel(skb, ac_start);
3556 goto out;
3557 } else {
3558 nla_nest_end(skb, ac_start);
3559 }
3560
3561 /* Third nested attribute in 'attr' is always
3562 * OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER.
3563 */
3564 a = nla_next(a, &rem);
3565 ac_start = nla_nest_start_noflag(skb,
3566 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER);
3567 if (!ac_start) {
3568 err = -EMSGSIZE;
3569 goto out;
3570 }
3571
3572 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3573 if (err) {
3574 nla_nest_cancel(skb, ac_start);
3575 goto out;
3576 } else {
3577 nla_nest_end(skb, ac_start);
3578 }
3579
3580 nla_nest_end(skb, start);
3581 return 0;
3582
3583 out:
3584 nla_nest_cancel(skb, start);
3585 return err;
3586 }
3587
dec_ttl_action_to_attr(const struct nlattr * attr,struct sk_buff * skb)3588 static int dec_ttl_action_to_attr(const struct nlattr *attr,
3589 struct sk_buff *skb)
3590 {
3591 struct nlattr *start, *action_start;
3592 const struct nlattr *a;
3593 int err = 0, rem;
3594
3595 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_DEC_TTL);
3596 if (!start)
3597 return -EMSGSIZE;
3598
3599 nla_for_each_attr(a, nla_data(attr), nla_len(attr), rem) {
3600 switch (nla_type(a)) {
3601 case OVS_DEC_TTL_ATTR_ACTION:
3602
3603 action_start = nla_nest_start_noflag(skb, OVS_DEC_TTL_ATTR_ACTION);
3604 if (!action_start) {
3605 err = -EMSGSIZE;
3606 goto out;
3607 }
3608
3609 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3610 if (err)
3611 goto out;
3612
3613 nla_nest_end(skb, action_start);
3614 break;
3615
3616 default:
3617 /* Ignore all other option to be future compatible */
3618 break;
3619 }
3620 }
3621
3622 nla_nest_end(skb, start);
3623 return 0;
3624
3625 out:
3626 nla_nest_cancel(skb, start);
3627 return err;
3628 }
3629
set_action_to_attr(const struct nlattr * a,struct sk_buff * skb)3630 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
3631 {
3632 const struct nlattr *ovs_key = nla_data(a);
3633 int key_type = nla_type(ovs_key);
3634 struct nlattr *start;
3635 int err;
3636
3637 switch (key_type) {
3638 case OVS_KEY_ATTR_TUNNEL_INFO: {
3639 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
3640 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
3641
3642 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3643 if (!start)
3644 return -EMSGSIZE;
3645
3646 err = ip_tun_to_nlattr(skb, &tun_info->key,
3647 ip_tunnel_info_opts(tun_info),
3648 tun_info->options_len,
3649 ip_tunnel_info_af(tun_info), tun_info->mode);
3650 if (err)
3651 return err;
3652 nla_nest_end(skb, start);
3653 break;
3654 }
3655 default:
3656 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
3657 return -EMSGSIZE;
3658 break;
3659 }
3660
3661 return 0;
3662 }
3663
masked_set_action_to_set_action_attr(const struct nlattr * a,struct sk_buff * skb)3664 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
3665 struct sk_buff *skb)
3666 {
3667 const struct nlattr *ovs_key = nla_data(a);
3668 struct nlattr *nla;
3669 size_t key_len = nla_len(ovs_key) / 2;
3670
3671 /* Revert the conversion we did from a non-masked set action to
3672 * masked set action.
3673 */
3674 nla = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3675 if (!nla)
3676 return -EMSGSIZE;
3677
3678 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
3679 return -EMSGSIZE;
3680
3681 nla_nest_end(skb, nla);
3682 return 0;
3683 }
3684
ovs_nla_put_actions(const struct nlattr * attr,int len,struct sk_buff * skb)3685 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
3686 {
3687 const struct nlattr *a;
3688 int rem, err;
3689
3690 nla_for_each_attr(a, attr, len, rem) {
3691 int type = nla_type(a);
3692
3693 switch (type) {
3694 case OVS_ACTION_ATTR_SET:
3695 err = set_action_to_attr(a, skb);
3696 if (err)
3697 return err;
3698 break;
3699
3700 case OVS_ACTION_ATTR_SET_TO_MASKED:
3701 err = masked_set_action_to_set_action_attr(a, skb);
3702 if (err)
3703 return err;
3704 break;
3705
3706 case OVS_ACTION_ATTR_SAMPLE:
3707 err = sample_action_to_attr(a, skb);
3708 if (err)
3709 return err;
3710 break;
3711
3712 case OVS_ACTION_ATTR_CT:
3713 err = ovs_ct_action_to_attr(nla_data(a), skb);
3714 if (err)
3715 return err;
3716 break;
3717
3718 case OVS_ACTION_ATTR_CLONE:
3719 err = clone_action_to_attr(a, skb);
3720 if (err)
3721 return err;
3722 break;
3723
3724 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
3725 err = check_pkt_len_action_to_attr(a, skb);
3726 if (err)
3727 return err;
3728 break;
3729
3730 case OVS_ACTION_ATTR_DEC_TTL:
3731 err = dec_ttl_action_to_attr(a, skb);
3732 if (err)
3733 return err;
3734 break;
3735
3736 default:
3737 if (nla_put(skb, type, nla_len(a), nla_data(a)))
3738 return -EMSGSIZE;
3739 break;
3740 }
3741 }
3742
3743 return 0;
3744 }
3745