1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * Copyright (c) 2007-2017 Nicira, Inc.
4 */
5
6 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
7
8 #include "flow.h"
9 #include "datapath.h"
10 #include <linux/uaccess.h>
11 #include <linux/netdevice.h>
12 #include <linux/etherdevice.h>
13 #include <linux/if_ether.h>
14 #include <linux/if_vlan.h>
15 #include <net/llc_pdu.h>
16 #include <linux/kernel.h>
17 #include <linux/jhash.h>
18 #include <linux/jiffies.h>
19 #include <linux/llc.h>
20 #include <linux/module.h>
21 #include <linux/in.h>
22 #include <linux/rcupdate.h>
23 #include <linux/if_arp.h>
24 #include <linux/ip.h>
25 #include <linux/ipv6.h>
26 #include <linux/sctp.h>
27 #include <linux/tcp.h>
28 #include <linux/udp.h>
29 #include <linux/icmp.h>
30 #include <linux/icmpv6.h>
31 #include <linux/rculist.h>
32 #include <net/geneve.h>
33 #include <net/ip.h>
34 #include <net/ipv6.h>
35 #include <net/ndisc.h>
36 #include <net/mpls.h>
37 #include <net/vxlan.h>
38 #include <net/tun_proto.h>
39 #include <net/erspan.h>
40
41 #include "drop.h"
42 #include "flow_netlink.h"
43
44 struct ovs_len_tbl {
45 int len;
46 const struct ovs_len_tbl *next;
47 };
48
49 #define OVS_ATTR_NESTED -1
50 #define OVS_ATTR_VARIABLE -2
51 #define OVS_COPY_ACTIONS_MAX_DEPTH 16
52
actions_may_change_flow(const struct nlattr * actions)53 static bool actions_may_change_flow(const struct nlattr *actions)
54 {
55 struct nlattr *nla;
56 int rem;
57
58 nla_for_each_nested(nla, actions, rem) {
59 u16 action = nla_type(nla);
60
61 switch (action) {
62 case OVS_ACTION_ATTR_OUTPUT:
63 case OVS_ACTION_ATTR_RECIRC:
64 case OVS_ACTION_ATTR_TRUNC:
65 case OVS_ACTION_ATTR_USERSPACE:
66 case OVS_ACTION_ATTR_DROP:
67 case OVS_ACTION_ATTR_PSAMPLE:
68 break;
69
70 case OVS_ACTION_ATTR_CT:
71 case OVS_ACTION_ATTR_CT_CLEAR:
72 case OVS_ACTION_ATTR_HASH:
73 case OVS_ACTION_ATTR_POP_ETH:
74 case OVS_ACTION_ATTR_POP_MPLS:
75 case OVS_ACTION_ATTR_POP_NSH:
76 case OVS_ACTION_ATTR_POP_VLAN:
77 case OVS_ACTION_ATTR_PUSH_ETH:
78 case OVS_ACTION_ATTR_PUSH_MPLS:
79 case OVS_ACTION_ATTR_PUSH_NSH:
80 case OVS_ACTION_ATTR_PUSH_VLAN:
81 case OVS_ACTION_ATTR_SAMPLE:
82 case OVS_ACTION_ATTR_SET:
83 case OVS_ACTION_ATTR_SET_MASKED:
84 case OVS_ACTION_ATTR_METER:
85 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
86 case OVS_ACTION_ATTR_ADD_MPLS:
87 case OVS_ACTION_ATTR_DEC_TTL:
88 default:
89 return true;
90 }
91 }
92 return false;
93 }
94
update_range(struct sw_flow_match * match,size_t offset,size_t size,bool is_mask)95 static void update_range(struct sw_flow_match *match,
96 size_t offset, size_t size, bool is_mask)
97 {
98 struct sw_flow_key_range *range;
99 size_t start = rounddown(offset, sizeof(long));
100 size_t end = roundup(offset + size, sizeof(long));
101
102 if (!is_mask)
103 range = &match->range;
104 else
105 range = &match->mask->range;
106
107 if (range->start == range->end) {
108 range->start = start;
109 range->end = end;
110 return;
111 }
112
113 if (range->start > start)
114 range->start = start;
115
116 if (range->end < end)
117 range->end = end;
118 }
119
120 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
121 do { \
122 update_range(match, offsetof(struct sw_flow_key, field), \
123 sizeof((match)->key->field), is_mask); \
124 if (is_mask) \
125 (match)->mask->key.field = value; \
126 else \
127 (match)->key->field = value; \
128 } while (0)
129
130 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
131 do { \
132 update_range(match, offset, len, is_mask); \
133 if (is_mask) \
134 memcpy((u8 *)&(match)->mask->key + offset, value_p, \
135 len); \
136 else \
137 memcpy((u8 *)(match)->key + offset, value_p, len); \
138 } while (0)
139
140 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
141 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
142 value_p, len, is_mask)
143
144 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
145 do { \
146 update_range(match, offsetof(struct sw_flow_key, field), \
147 sizeof((match)->key->field), is_mask); \
148 if (is_mask) \
149 memset((u8 *)&(match)->mask->key.field, value, \
150 sizeof((match)->mask->key.field)); \
151 else \
152 memset((u8 *)&(match)->key->field, value, \
153 sizeof((match)->key->field)); \
154 } while (0)
155
156 #define SW_FLOW_KEY_BITMAP_COPY(match, field, value_p, nbits, is_mask) ({ \
157 update_range(match, offsetof(struct sw_flow_key, field), \
158 bitmap_size(nbits), is_mask); \
159 bitmap_copy(is_mask ? (match)->mask->key.field : (match)->key->field, \
160 value_p, nbits); \
161 })
162
match_validate(const struct sw_flow_match * match,u64 key_attrs,u64 mask_attrs,bool log)163 static bool match_validate(const struct sw_flow_match *match,
164 u64 key_attrs, u64 mask_attrs, bool log)
165 {
166 u64 key_expected = 0;
167 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
168
169 /* The following mask attributes allowed only if they
170 * pass the validation tests. */
171 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
172 | (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)
173 | (1 << OVS_KEY_ATTR_IPV6)
174 | (1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)
175 | (1 << OVS_KEY_ATTR_TCP)
176 | (1 << OVS_KEY_ATTR_TCP_FLAGS)
177 | (1 << OVS_KEY_ATTR_UDP)
178 | (1 << OVS_KEY_ATTR_SCTP)
179 | (1 << OVS_KEY_ATTR_ICMP)
180 | (1 << OVS_KEY_ATTR_ICMPV6)
181 | (1 << OVS_KEY_ATTR_ARP)
182 | (1 << OVS_KEY_ATTR_ND)
183 | (1 << OVS_KEY_ATTR_MPLS)
184 | (1 << OVS_KEY_ATTR_NSH));
185
186 /* Always allowed mask fields. */
187 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
188 | (1 << OVS_KEY_ATTR_IN_PORT)
189 | (1 << OVS_KEY_ATTR_ETHERTYPE));
190
191 /* Check key attributes. */
192 if (match->key->eth.type == htons(ETH_P_ARP)
193 || match->key->eth.type == htons(ETH_P_RARP)) {
194 key_expected |= 1 << OVS_KEY_ATTR_ARP;
195 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
196 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
197 }
198
199 if (eth_p_mpls(match->key->eth.type)) {
200 key_expected |= 1 << OVS_KEY_ATTR_MPLS;
201 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
202 mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
203 }
204
205 if (match->key->eth.type == htons(ETH_P_IP)) {
206 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
207 if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
208 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
209 mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4;
210 }
211
212 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
213 if (match->key->ip.proto == IPPROTO_UDP) {
214 key_expected |= 1 << OVS_KEY_ATTR_UDP;
215 if (match->mask && (match->mask->key.ip.proto == 0xff))
216 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
217 }
218
219 if (match->key->ip.proto == IPPROTO_SCTP) {
220 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
221 if (match->mask && (match->mask->key.ip.proto == 0xff))
222 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
223 }
224
225 if (match->key->ip.proto == IPPROTO_TCP) {
226 key_expected |= 1 << OVS_KEY_ATTR_TCP;
227 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
228 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
229 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
230 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
231 }
232 }
233
234 if (match->key->ip.proto == IPPROTO_ICMP) {
235 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
236 if (match->mask && (match->mask->key.ip.proto == 0xff))
237 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
238 }
239 }
240 }
241
242 if (match->key->eth.type == htons(ETH_P_IPV6)) {
243 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
244 if (match->mask && match->mask->key.eth.type == htons(0xffff)) {
245 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
246 mask_allowed |= 1 << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6;
247 }
248
249 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
250 if (match->key->ip.proto == IPPROTO_UDP) {
251 key_expected |= 1 << OVS_KEY_ATTR_UDP;
252 if (match->mask && (match->mask->key.ip.proto == 0xff))
253 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
254 }
255
256 if (match->key->ip.proto == IPPROTO_SCTP) {
257 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
258 if (match->mask && (match->mask->key.ip.proto == 0xff))
259 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
260 }
261
262 if (match->key->ip.proto == IPPROTO_TCP) {
263 key_expected |= 1 << OVS_KEY_ATTR_TCP;
264 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
265 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
266 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
267 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
268 }
269 }
270
271 if (match->key->ip.proto == IPPROTO_ICMPV6) {
272 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
273 if (match->mask && (match->mask->key.ip.proto == 0xff))
274 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
275
276 if (match->key->tp.src ==
277 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
278 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
279 key_expected |= 1 << OVS_KEY_ATTR_ND;
280 /* Original direction conntrack tuple
281 * uses the same space as the ND fields
282 * in the key, so both are not allowed
283 * at the same time.
284 */
285 mask_allowed &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
286 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
287 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
288 }
289 }
290 }
291 }
292
293 if (match->key->eth.type == htons(ETH_P_NSH)) {
294 key_expected |= 1 << OVS_KEY_ATTR_NSH;
295 if (match->mask &&
296 match->mask->key.eth.type == htons(0xffff)) {
297 mask_allowed |= 1 << OVS_KEY_ATTR_NSH;
298 }
299 }
300
301 if ((key_attrs & key_expected) != key_expected) {
302 /* Key attributes check failed. */
303 OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
304 (unsigned long long)key_attrs,
305 (unsigned long long)key_expected);
306 return false;
307 }
308
309 if ((mask_attrs & mask_allowed) != mask_attrs) {
310 /* Mask attributes check failed. */
311 OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
312 (unsigned long long)mask_attrs,
313 (unsigned long long)mask_allowed);
314 return false;
315 }
316
317 return true;
318 }
319
ovs_tun_key_attr_size(void)320 size_t ovs_tun_key_attr_size(void)
321 {
322 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
323 * updating this function.
324 */
325 return nla_total_size_64bit(8) /* OVS_TUNNEL_KEY_ATTR_ID */
326 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
327 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_DST */
328 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
329 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
330 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
331 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
332 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
333 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
334 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS and
335 * OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS is mutually exclusive with
336 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
337 */
338 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
339 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
340 }
341
ovs_nsh_key_attr_size(void)342 static size_t ovs_nsh_key_attr_size(void)
343 {
344 /* Whenever adding new OVS_NSH_KEY_ FIELDS, we should consider
345 * updating this function.
346 */
347 return nla_total_size(NSH_BASE_HDR_LEN) /* OVS_NSH_KEY_ATTR_BASE */
348 /* OVS_NSH_KEY_ATTR_MD1 and OVS_NSH_KEY_ATTR_MD2 are
349 * mutually exclusive, so the bigger one can cover
350 * the small one.
351 */
352 + nla_total_size(NSH_CTX_HDRS_MAX_LEN);
353 }
354
ovs_key_attr_size(void)355 size_t ovs_key_attr_size(void)
356 {
357 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
358 * updating this function.
359 */
360 BUILD_BUG_ON(OVS_KEY_ATTR_MAX != 32);
361
362 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
363 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
364 + ovs_tun_key_attr_size()
365 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
366 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
367 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
368 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
369 + nla_total_size(4) /* OVS_KEY_ATTR_CT_STATE */
370 + nla_total_size(2) /* OVS_KEY_ATTR_CT_ZONE */
371 + nla_total_size(4) /* OVS_KEY_ATTR_CT_MARK */
372 + nla_total_size(16) /* OVS_KEY_ATTR_CT_LABELS */
373 + nla_total_size(40) /* OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6 */
374 + nla_total_size(0) /* OVS_KEY_ATTR_NSH */
375 + ovs_nsh_key_attr_size()
376 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
377 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
378 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
379 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
380 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
381 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
382 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
383 + nla_total_size(28) /* OVS_KEY_ATTR_ND */
384 + nla_total_size(2); /* OVS_KEY_ATTR_IPV6_EXTHDRS */
385 }
386
387 static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
388 [OVS_VXLAN_EXT_GBP] = { .len = sizeof(u32) },
389 };
390
391 static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
392 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
393 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
394 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
395 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
396 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
397 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
398 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
399 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
400 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
401 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
402 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_VARIABLE },
403 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED,
404 .next = ovs_vxlan_ext_key_lens },
405 [OVS_TUNNEL_KEY_ATTR_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
406 [OVS_TUNNEL_KEY_ATTR_IPV6_DST] = { .len = sizeof(struct in6_addr) },
407 [OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS] = { .len = OVS_ATTR_VARIABLE },
408 [OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE] = { .len = 0 },
409 };
410
411 static const struct ovs_len_tbl
412 ovs_nsh_key_attr_lens[OVS_NSH_KEY_ATTR_MAX + 1] = {
413 [OVS_NSH_KEY_ATTR_BASE] = { .len = sizeof(struct ovs_nsh_key_base) },
414 [OVS_NSH_KEY_ATTR_MD1] = { .len = sizeof(struct ovs_nsh_key_md1) },
415 [OVS_NSH_KEY_ATTR_MD2] = { .len = OVS_ATTR_VARIABLE },
416 };
417
418 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
419 static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
420 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
421 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
422 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
423 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
424 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
425 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
426 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
427 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
428 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
429 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
430 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
431 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
432 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
433 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
434 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
435 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
436 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
437 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
438 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
439 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
440 .next = ovs_tunnel_key_lens, },
441 [OVS_KEY_ATTR_MPLS] = { .len = OVS_ATTR_VARIABLE },
442 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u32) },
443 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
444 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
445 [OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
446 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4] = {
447 .len = sizeof(struct ovs_key_ct_tuple_ipv4) },
448 [OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6] = {
449 .len = sizeof(struct ovs_key_ct_tuple_ipv6) },
450 [OVS_KEY_ATTR_NSH] = { .len = OVS_ATTR_NESTED,
451 .next = ovs_nsh_key_attr_lens, },
452 [OVS_KEY_ATTR_IPV6_EXTHDRS] = {
453 .len = sizeof(struct ovs_key_ipv6_exthdrs) },
454 };
455
check_attr_len(unsigned int attr_len,unsigned int expected_len)456 static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
457 {
458 return expected_len == attr_len ||
459 expected_len == OVS_ATTR_NESTED ||
460 expected_len == OVS_ATTR_VARIABLE;
461 }
462
is_all_zero(const u8 * fp,size_t size)463 static bool is_all_zero(const u8 *fp, size_t size)
464 {
465 int i;
466
467 if (!fp)
468 return false;
469
470 for (i = 0; i < size; i++)
471 if (fp[i])
472 return false;
473
474 return true;
475 }
476
__parse_flow_nlattrs(const struct nlattr * attr,const struct nlattr * a[],u64 * attrsp,bool log,bool nz)477 static int __parse_flow_nlattrs(const struct nlattr *attr,
478 const struct nlattr *a[],
479 u64 *attrsp, bool log, bool nz)
480 {
481 const struct nlattr *nla;
482 u64 attrs;
483 int rem;
484
485 attrs = *attrsp;
486 nla_for_each_nested(nla, attr, rem) {
487 u16 type = nla_type(nla);
488 int expected_len;
489
490 if (type > OVS_KEY_ATTR_MAX) {
491 OVS_NLERR(log, "Key type %d is out of range max %d",
492 type, OVS_KEY_ATTR_MAX);
493 return -EINVAL;
494 }
495
496 if (type == OVS_KEY_ATTR_PACKET_TYPE ||
497 type == OVS_KEY_ATTR_ND_EXTENSIONS ||
498 type == OVS_KEY_ATTR_TUNNEL_INFO) {
499 OVS_NLERR(log, "Key type %d is not supported", type);
500 return -EINVAL;
501 }
502
503 if (attrs & (1ULL << type)) {
504 OVS_NLERR(log, "Duplicate key (type %d).", type);
505 return -EINVAL;
506 }
507
508 expected_len = ovs_key_lens[type].len;
509 if (!check_attr_len(nla_len(nla), expected_len)) {
510 OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
511 type, nla_len(nla), expected_len);
512 return -EINVAL;
513 }
514
515 if (!nz || !is_all_zero(nla_data(nla), nla_len(nla))) {
516 attrs |= 1ULL << type;
517 a[type] = nla;
518 }
519 }
520 if (rem) {
521 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
522 return -EINVAL;
523 }
524
525 *attrsp = attrs;
526 return 0;
527 }
528
parse_flow_mask_nlattrs(const struct nlattr * attr,const struct nlattr * a[],u64 * attrsp,bool log)529 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
530 const struct nlattr *a[], u64 *attrsp,
531 bool log)
532 {
533 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
534 }
535
parse_flow_nlattrs(const struct nlattr * attr,const struct nlattr * a[],u64 * attrsp,bool log)536 int parse_flow_nlattrs(const struct nlattr *attr, const struct nlattr *a[],
537 u64 *attrsp, bool log)
538 {
539 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
540 }
541
genev_tun_opt_from_nlattr(const struct nlattr * a,struct sw_flow_match * match,bool is_mask,bool log)542 static int genev_tun_opt_from_nlattr(const struct nlattr *a,
543 struct sw_flow_match *match, bool is_mask,
544 bool log)
545 {
546 unsigned long opt_key_offset;
547
548 if (nla_len(a) > sizeof(match->key->tun_opts)) {
549 OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
550 nla_len(a), sizeof(match->key->tun_opts));
551 return -EINVAL;
552 }
553
554 if (nla_len(a) % 4 != 0) {
555 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
556 nla_len(a));
557 return -EINVAL;
558 }
559
560 /* We need to record the length of the options passed
561 * down, otherwise packets with the same format but
562 * additional options will be silently matched.
563 */
564 if (!is_mask) {
565 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
566 false);
567 } else {
568 /* This is somewhat unusual because it looks at
569 * both the key and mask while parsing the
570 * attributes (and by extension assumes the key
571 * is parsed first). Normally, we would verify
572 * that each is the correct length and that the
573 * attributes line up in the validate function.
574 * However, that is difficult because this is
575 * variable length and we won't have the
576 * information later.
577 */
578 if (match->key->tun_opts_len != nla_len(a)) {
579 OVS_NLERR(log, "Geneve option len %d != mask len %d",
580 match->key->tun_opts_len, nla_len(a));
581 return -EINVAL;
582 }
583
584 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
585 }
586
587 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
588 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
589 nla_len(a), is_mask);
590 return 0;
591 }
592
vxlan_tun_opt_from_nlattr(const struct nlattr * attr,struct sw_flow_match * match,bool is_mask,bool log)593 static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
594 struct sw_flow_match *match, bool is_mask,
595 bool log)
596 {
597 struct nlattr *a;
598 int rem;
599 unsigned long opt_key_offset;
600 struct vxlan_metadata opts;
601
602 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
603
604 memset(&opts, 0, sizeof(opts));
605 nla_for_each_nested(a, attr, rem) {
606 int type = nla_type(a);
607
608 if (type > OVS_VXLAN_EXT_MAX) {
609 OVS_NLERR(log, "VXLAN extension %d out of range max %d",
610 type, OVS_VXLAN_EXT_MAX);
611 return -EINVAL;
612 }
613
614 if (!check_attr_len(nla_len(a),
615 ovs_vxlan_ext_key_lens[type].len)) {
616 OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
617 type, nla_len(a),
618 ovs_vxlan_ext_key_lens[type].len);
619 return -EINVAL;
620 }
621
622 switch (type) {
623 case OVS_VXLAN_EXT_GBP:
624 opts.gbp = nla_get_u32(a);
625 break;
626 default:
627 OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
628 type);
629 return -EINVAL;
630 }
631 }
632 if (rem) {
633 OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
634 rem);
635 return -EINVAL;
636 }
637
638 if (!is_mask)
639 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
640 else
641 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
642
643 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
644 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
645 is_mask);
646 return 0;
647 }
648
erspan_tun_opt_from_nlattr(const struct nlattr * a,struct sw_flow_match * match,bool is_mask,bool log)649 static int erspan_tun_opt_from_nlattr(const struct nlattr *a,
650 struct sw_flow_match *match, bool is_mask,
651 bool log)
652 {
653 unsigned long opt_key_offset;
654
655 BUILD_BUG_ON(sizeof(struct erspan_metadata) >
656 sizeof(match->key->tun_opts));
657
658 if (nla_len(a) > sizeof(match->key->tun_opts)) {
659 OVS_NLERR(log, "ERSPAN option length err (len %d, max %zu).",
660 nla_len(a), sizeof(match->key->tun_opts));
661 return -EINVAL;
662 }
663
664 if (!is_mask)
665 SW_FLOW_KEY_PUT(match, tun_opts_len,
666 sizeof(struct erspan_metadata), false);
667 else
668 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
669
670 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
671 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
672 nla_len(a), is_mask);
673 return 0;
674 }
675
ip_tun_from_nlattr(const struct nlattr * attr,struct sw_flow_match * match,bool is_mask,bool log)676 static int ip_tun_from_nlattr(const struct nlattr *attr,
677 struct sw_flow_match *match, bool is_mask,
678 bool log)
679 {
680 bool ttl = false, ipv4 = false, ipv6 = false;
681 IP_TUNNEL_DECLARE_FLAGS(tun_flags) = { };
682 bool info_bridge_mode = false;
683 int opts_type = 0;
684 struct nlattr *a;
685 int rem;
686
687 nla_for_each_nested(a, attr, rem) {
688 int type = nla_type(a);
689 int err;
690
691 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
692 OVS_NLERR(log, "Tunnel attr %d out of range max %d",
693 type, OVS_TUNNEL_KEY_ATTR_MAX);
694 return -EINVAL;
695 }
696
697 if (!check_attr_len(nla_len(a),
698 ovs_tunnel_key_lens[type].len)) {
699 OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
700 type, nla_len(a), ovs_tunnel_key_lens[type].len);
701 return -EINVAL;
702 }
703
704 switch (type) {
705 case OVS_TUNNEL_KEY_ATTR_ID:
706 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
707 nla_get_be64(a), is_mask);
708 __set_bit(IP_TUNNEL_KEY_BIT, tun_flags);
709 break;
710 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
711 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
712 nla_get_in_addr(a), is_mask);
713 ipv4 = true;
714 break;
715 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
716 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
717 nla_get_in_addr(a), is_mask);
718 ipv4 = true;
719 break;
720 case OVS_TUNNEL_KEY_ATTR_IPV6_SRC:
721 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.src,
722 nla_get_in6_addr(a), is_mask);
723 ipv6 = true;
724 break;
725 case OVS_TUNNEL_KEY_ATTR_IPV6_DST:
726 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
727 nla_get_in6_addr(a), is_mask);
728 ipv6 = true;
729 break;
730 case OVS_TUNNEL_KEY_ATTR_TOS:
731 SW_FLOW_KEY_PUT(match, tun_key.tos,
732 nla_get_u8(a), is_mask);
733 break;
734 case OVS_TUNNEL_KEY_ATTR_TTL:
735 SW_FLOW_KEY_PUT(match, tun_key.ttl,
736 nla_get_u8(a), is_mask);
737 ttl = true;
738 break;
739 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
740 __set_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, tun_flags);
741 break;
742 case OVS_TUNNEL_KEY_ATTR_CSUM:
743 __set_bit(IP_TUNNEL_CSUM_BIT, tun_flags);
744 break;
745 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
746 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
747 nla_get_be16(a), is_mask);
748 break;
749 case OVS_TUNNEL_KEY_ATTR_TP_DST:
750 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
751 nla_get_be16(a), is_mask);
752 break;
753 case OVS_TUNNEL_KEY_ATTR_OAM:
754 __set_bit(IP_TUNNEL_OAM_BIT, tun_flags);
755 break;
756 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
757 if (opts_type) {
758 OVS_NLERR(log, "Multiple metadata blocks provided");
759 return -EINVAL;
760 }
761
762 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
763 if (err)
764 return err;
765
766 __set_bit(IP_TUNNEL_GENEVE_OPT_BIT, tun_flags);
767 opts_type = type;
768 break;
769 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
770 if (opts_type) {
771 OVS_NLERR(log, "Multiple metadata blocks provided");
772 return -EINVAL;
773 }
774
775 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
776 if (err)
777 return err;
778
779 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, tun_flags);
780 opts_type = type;
781 break;
782 case OVS_TUNNEL_KEY_ATTR_PAD:
783 break;
784 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
785 if (opts_type) {
786 OVS_NLERR(log, "Multiple metadata blocks provided");
787 return -EINVAL;
788 }
789
790 err = erspan_tun_opt_from_nlattr(a, match, is_mask,
791 log);
792 if (err)
793 return err;
794
795 __set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, tun_flags);
796 opts_type = type;
797 break;
798 case OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE:
799 info_bridge_mode = true;
800 ipv4 = true;
801 break;
802 default:
803 OVS_NLERR(log, "Unknown IP tunnel attribute %d",
804 type);
805 return -EINVAL;
806 }
807 }
808
809 SW_FLOW_KEY_BITMAP_COPY(match, tun_key.tun_flags, tun_flags,
810 __IP_TUNNEL_FLAG_NUM, is_mask);
811 if (is_mask)
812 SW_FLOW_KEY_MEMSET_FIELD(match, tun_proto, 0xff, true);
813 else
814 SW_FLOW_KEY_PUT(match, tun_proto, ipv6 ? AF_INET6 : AF_INET,
815 false);
816
817 if (rem > 0) {
818 OVS_NLERR(log, "IP tunnel attribute has %d unknown bytes.",
819 rem);
820 return -EINVAL;
821 }
822
823 if (ipv4 && ipv6) {
824 OVS_NLERR(log, "Mixed IPv4 and IPv6 tunnel attributes");
825 return -EINVAL;
826 }
827
828 if (!is_mask) {
829 if (!ipv4 && !ipv6) {
830 OVS_NLERR(log, "IP tunnel dst address not specified");
831 return -EINVAL;
832 }
833 if (ipv4) {
834 if (info_bridge_mode) {
835 __clear_bit(IP_TUNNEL_KEY_BIT, tun_flags);
836
837 if (match->key->tun_key.u.ipv4.src ||
838 match->key->tun_key.u.ipv4.dst ||
839 match->key->tun_key.tp_src ||
840 match->key->tun_key.tp_dst ||
841 match->key->tun_key.ttl ||
842 match->key->tun_key.tos ||
843 !ip_tunnel_flags_empty(tun_flags)) {
844 OVS_NLERR(log, "IPv4 tun info is not correct");
845 return -EINVAL;
846 }
847 } else if (!match->key->tun_key.u.ipv4.dst) {
848 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
849 return -EINVAL;
850 }
851 }
852 if (ipv6 && ipv6_addr_any(&match->key->tun_key.u.ipv6.dst)) {
853 OVS_NLERR(log, "IPv6 tunnel dst address is zero");
854 return -EINVAL;
855 }
856
857 if (!ttl && !info_bridge_mode) {
858 OVS_NLERR(log, "IP tunnel TTL not specified.");
859 return -EINVAL;
860 }
861 }
862
863 return opts_type;
864 }
865
vxlan_opt_to_nlattr(struct sk_buff * skb,const void * tun_opts,int swkey_tun_opts_len)866 static int vxlan_opt_to_nlattr(struct sk_buff *skb,
867 const void *tun_opts, int swkey_tun_opts_len)
868 {
869 const struct vxlan_metadata *opts = tun_opts;
870 struct nlattr *nla;
871
872 nla = nla_nest_start_noflag(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
873 if (!nla)
874 return -EMSGSIZE;
875
876 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
877 return -EMSGSIZE;
878
879 nla_nest_end(skb, nla);
880 return 0;
881 }
882
__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)883 static int __ip_tun_to_nlattr(struct sk_buff *skb,
884 const struct ip_tunnel_key *output,
885 const void *tun_opts, int swkey_tun_opts_len,
886 unsigned short tun_proto, u8 mode)
887 {
888 if (test_bit(IP_TUNNEL_KEY_BIT, output->tun_flags) &&
889 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id,
890 OVS_TUNNEL_KEY_ATTR_PAD))
891 return -EMSGSIZE;
892
893 if (mode & IP_TUNNEL_INFO_BRIDGE)
894 return nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_IPV4_INFO_BRIDGE)
895 ? -EMSGSIZE : 0;
896
897 switch (tun_proto) {
898 case AF_INET:
899 if (output->u.ipv4.src &&
900 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
901 output->u.ipv4.src))
902 return -EMSGSIZE;
903 if (output->u.ipv4.dst &&
904 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
905 output->u.ipv4.dst))
906 return -EMSGSIZE;
907 break;
908 case AF_INET6:
909 if (!ipv6_addr_any(&output->u.ipv6.src) &&
910 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_SRC,
911 &output->u.ipv6.src))
912 return -EMSGSIZE;
913 if (!ipv6_addr_any(&output->u.ipv6.dst) &&
914 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_DST,
915 &output->u.ipv6.dst))
916 return -EMSGSIZE;
917 break;
918 }
919 if (output->tos &&
920 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
921 return -EMSGSIZE;
922 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
923 return -EMSGSIZE;
924 if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, output->tun_flags) &&
925 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
926 return -EMSGSIZE;
927 if (test_bit(IP_TUNNEL_CSUM_BIT, output->tun_flags) &&
928 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
929 return -EMSGSIZE;
930 if (output->tp_src &&
931 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
932 return -EMSGSIZE;
933 if (output->tp_dst &&
934 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
935 return -EMSGSIZE;
936 if (test_bit(IP_TUNNEL_OAM_BIT, output->tun_flags) &&
937 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
938 return -EMSGSIZE;
939 if (swkey_tun_opts_len) {
940 if (test_bit(IP_TUNNEL_GENEVE_OPT_BIT, output->tun_flags) &&
941 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
942 swkey_tun_opts_len, tun_opts))
943 return -EMSGSIZE;
944 else if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT,
945 output->tun_flags) &&
946 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
947 return -EMSGSIZE;
948 else if (test_bit(IP_TUNNEL_ERSPAN_OPT_BIT,
949 output->tun_flags) &&
950 nla_put(skb, OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS,
951 swkey_tun_opts_len, tun_opts))
952 return -EMSGSIZE;
953 }
954
955 return 0;
956 }
957
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)958 static int ip_tun_to_nlattr(struct sk_buff *skb,
959 const struct ip_tunnel_key *output,
960 const void *tun_opts, int swkey_tun_opts_len,
961 unsigned short tun_proto, u8 mode)
962 {
963 struct nlattr *nla;
964 int err;
965
966 nla = nla_nest_start_noflag(skb, OVS_KEY_ATTR_TUNNEL);
967 if (!nla)
968 return -EMSGSIZE;
969
970 err = __ip_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len,
971 tun_proto, mode);
972 if (err)
973 return err;
974
975 nla_nest_end(skb, nla);
976 return 0;
977 }
978
ovs_nla_put_tunnel_info(struct sk_buff * skb,struct ip_tunnel_info * tun_info)979 int ovs_nla_put_tunnel_info(struct sk_buff *skb,
980 struct ip_tunnel_info *tun_info)
981 {
982 return __ip_tun_to_nlattr(skb, &tun_info->key,
983 ip_tunnel_info_opts(tun_info),
984 tun_info->options_len,
985 ip_tunnel_info_af(tun_info), tun_info->mode);
986 }
987
encode_vlan_from_nlattrs(struct sw_flow_match * match,const struct nlattr * a[],bool is_mask,bool inner)988 static int encode_vlan_from_nlattrs(struct sw_flow_match *match,
989 const struct nlattr *a[],
990 bool is_mask, bool inner)
991 {
992 __be16 tci = 0;
993 __be16 tpid = 0;
994
995 if (a[OVS_KEY_ATTR_VLAN])
996 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
997
998 if (a[OVS_KEY_ATTR_ETHERTYPE])
999 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1000
1001 if (likely(!inner)) {
1002 SW_FLOW_KEY_PUT(match, eth.vlan.tpid, tpid, is_mask);
1003 SW_FLOW_KEY_PUT(match, eth.vlan.tci, tci, is_mask);
1004 } else {
1005 SW_FLOW_KEY_PUT(match, eth.cvlan.tpid, tpid, is_mask);
1006 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, tci, is_mask);
1007 }
1008 return 0;
1009 }
1010
validate_vlan_from_nlattrs(const struct sw_flow_match * match,u64 key_attrs,bool inner,const struct nlattr ** a,bool log)1011 static int validate_vlan_from_nlattrs(const struct sw_flow_match *match,
1012 u64 key_attrs, bool inner,
1013 const struct nlattr **a, bool log)
1014 {
1015 __be16 tci = 0;
1016
1017 if (!((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
1018 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
1019 eth_type_vlan(nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE])))) {
1020 /* Not a VLAN. */
1021 return 0;
1022 }
1023
1024 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
1025 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
1026 OVS_NLERR(log, "Invalid %s frame", (inner) ? "C-VLAN" : "VLAN");
1027 return -EINVAL;
1028 }
1029
1030 if (a[OVS_KEY_ATTR_VLAN])
1031 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1032
1033 if (!(tci & htons(VLAN_CFI_MASK))) {
1034 if (tci) {
1035 OVS_NLERR(log, "%s TCI does not have VLAN_CFI_MASK bit set.",
1036 (inner) ? "C-VLAN" : "VLAN");
1037 return -EINVAL;
1038 } else if (nla_len(a[OVS_KEY_ATTR_ENCAP])) {
1039 /* Corner case for truncated VLAN header. */
1040 OVS_NLERR(log, "Truncated %s header has non-zero encap attribute.",
1041 (inner) ? "C-VLAN" : "VLAN");
1042 return -EINVAL;
1043 }
1044 }
1045
1046 return 1;
1047 }
1048
validate_vlan_mask_from_nlattrs(const struct sw_flow_match * match,u64 key_attrs,bool inner,const struct nlattr ** a,bool log)1049 static int validate_vlan_mask_from_nlattrs(const struct sw_flow_match *match,
1050 u64 key_attrs, bool inner,
1051 const struct nlattr **a, bool log)
1052 {
1053 __be16 tci = 0;
1054 __be16 tpid = 0;
1055 bool encap_valid = !!(match->key->eth.vlan.tci &
1056 htons(VLAN_CFI_MASK));
1057 bool i_encap_valid = !!(match->key->eth.cvlan.tci &
1058 htons(VLAN_CFI_MASK));
1059
1060 if (!(key_attrs & (1 << OVS_KEY_ATTR_ENCAP))) {
1061 /* Not a VLAN. */
1062 return 0;
1063 }
1064
1065 if ((!inner && !encap_valid) || (inner && !i_encap_valid)) {
1066 OVS_NLERR(log, "Encap mask attribute is set for non-%s frame.",
1067 (inner) ? "C-VLAN" : "VLAN");
1068 return -EINVAL;
1069 }
1070
1071 if (a[OVS_KEY_ATTR_VLAN])
1072 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1073
1074 if (a[OVS_KEY_ATTR_ETHERTYPE])
1075 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1076
1077 if (tpid != htons(0xffff)) {
1078 OVS_NLERR(log, "Must have an exact match on %s TPID (mask=%x).",
1079 (inner) ? "C-VLAN" : "VLAN", ntohs(tpid));
1080 return -EINVAL;
1081 }
1082 if (!(tci & htons(VLAN_CFI_MASK))) {
1083 OVS_NLERR(log, "%s TCI mask does not have exact match for VLAN_CFI_MASK bit.",
1084 (inner) ? "C-VLAN" : "VLAN");
1085 return -EINVAL;
1086 }
1087
1088 return 1;
1089 }
1090
__parse_vlan_from_nlattrs(struct sw_flow_match * match,u64 * key_attrs,bool inner,const struct nlattr ** a,bool is_mask,bool log)1091 static int __parse_vlan_from_nlattrs(struct sw_flow_match *match,
1092 u64 *key_attrs, bool inner,
1093 const struct nlattr **a, bool is_mask,
1094 bool log)
1095 {
1096 int err;
1097 const struct nlattr *encap;
1098
1099 if (!is_mask)
1100 err = validate_vlan_from_nlattrs(match, *key_attrs, inner,
1101 a, log);
1102 else
1103 err = validate_vlan_mask_from_nlattrs(match, *key_attrs, inner,
1104 a, log);
1105 if (err <= 0)
1106 return err;
1107
1108 err = encode_vlan_from_nlattrs(match, a, is_mask, inner);
1109 if (err)
1110 return err;
1111
1112 *key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1113 *key_attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
1114 *key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1115
1116 encap = a[OVS_KEY_ATTR_ENCAP];
1117
1118 if (!is_mask)
1119 err = parse_flow_nlattrs(encap, a, key_attrs, log);
1120 else
1121 err = parse_flow_mask_nlattrs(encap, a, key_attrs, log);
1122
1123 return err;
1124 }
1125
parse_vlan_from_nlattrs(struct sw_flow_match * match,u64 * key_attrs,const struct nlattr ** a,bool is_mask,bool log)1126 static int parse_vlan_from_nlattrs(struct sw_flow_match *match,
1127 u64 *key_attrs, const struct nlattr **a,
1128 bool is_mask, bool log)
1129 {
1130 int err;
1131 bool encap_valid = false;
1132
1133 err = __parse_vlan_from_nlattrs(match, key_attrs, false, a,
1134 is_mask, log);
1135 if (err)
1136 return err;
1137
1138 encap_valid = !!(match->key->eth.vlan.tci & htons(VLAN_CFI_MASK));
1139 if (encap_valid) {
1140 err = __parse_vlan_from_nlattrs(match, key_attrs, true, a,
1141 is_mask, log);
1142 if (err)
1143 return err;
1144 }
1145
1146 return 0;
1147 }
1148
parse_eth_type_from_nlattrs(struct sw_flow_match * match,u64 * attrs,const struct nlattr ** a,bool is_mask,bool log)1149 static int parse_eth_type_from_nlattrs(struct sw_flow_match *match,
1150 u64 *attrs, const struct nlattr **a,
1151 bool is_mask, bool log)
1152 {
1153 __be16 eth_type;
1154
1155 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1156 if (is_mask) {
1157 /* Always exact match EtherType. */
1158 eth_type = htons(0xffff);
1159 } else if (!eth_proto_is_802_3(eth_type)) {
1160 OVS_NLERR(log, "EtherType %x is less than min %x",
1161 ntohs(eth_type), ETH_P_802_3_MIN);
1162 return -EINVAL;
1163 }
1164
1165 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
1166 *attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1167 return 0;
1168 }
1169
metadata_from_nlattrs(struct net * net,struct sw_flow_match * match,u64 * attrs,const struct nlattr ** a,bool is_mask,bool log)1170 static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
1171 u64 *attrs, const struct nlattr **a,
1172 bool is_mask, bool log)
1173 {
1174 u8 mac_proto = MAC_PROTO_ETHERNET;
1175
1176 if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
1177 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
1178
1179 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
1180 *attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
1181 }
1182
1183 if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
1184 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
1185
1186 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
1187 *attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
1188 }
1189
1190 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
1191 SW_FLOW_KEY_PUT(match, phy.priority,
1192 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
1193 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
1194 }
1195
1196 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
1197 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1198
1199 if (is_mask) {
1200 in_port = 0xffffffff; /* Always exact match in_port. */
1201 } else if (in_port >= DP_MAX_PORTS) {
1202 OVS_NLERR(log, "Port %d exceeds max allowable %d",
1203 in_port, DP_MAX_PORTS);
1204 return -EINVAL;
1205 }
1206
1207 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
1208 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1209 } else if (!is_mask) {
1210 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
1211 }
1212
1213 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1214 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1215
1216 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1217 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1218 }
1219 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1220 if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1221 is_mask, log) < 0)
1222 return -EINVAL;
1223 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1224 }
1225
1226 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
1227 ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
1228 u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
1229
1230 if (ct_state & ~CT_SUPPORTED_MASK) {
1231 OVS_NLERR(log, "ct_state flags %08x unsupported",
1232 ct_state);
1233 return -EINVAL;
1234 }
1235
1236 SW_FLOW_KEY_PUT(match, ct_state, ct_state, is_mask);
1237 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
1238 }
1239 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
1240 ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
1241 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
1242
1243 SW_FLOW_KEY_PUT(match, ct_zone, ct_zone, is_mask);
1244 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
1245 }
1246 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
1247 ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
1248 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
1249
1250 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
1251 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
1252 }
1253 if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
1254 ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
1255 const struct ovs_key_ct_labels *cl;
1256
1257 cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
1258 SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
1259 sizeof(*cl), is_mask);
1260 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
1261 }
1262 if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4)) {
1263 const struct ovs_key_ct_tuple_ipv4 *ct;
1264
1265 ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4]);
1266
1267 SW_FLOW_KEY_PUT(match, ipv4.ct_orig.src, ct->ipv4_src, is_mask);
1268 SW_FLOW_KEY_PUT(match, ipv4.ct_orig.dst, ct->ipv4_dst, is_mask);
1269 SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1270 SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1271 SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv4_proto, is_mask);
1272 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4);
1273 }
1274 if (*attrs & (1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6)) {
1275 const struct ovs_key_ct_tuple_ipv6 *ct;
1276
1277 ct = nla_data(a[OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6]);
1278
1279 SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.src, &ct->ipv6_src,
1280 sizeof(match->key->ipv6.ct_orig.src),
1281 is_mask);
1282 SW_FLOW_KEY_MEMCPY(match, ipv6.ct_orig.dst, &ct->ipv6_dst,
1283 sizeof(match->key->ipv6.ct_orig.dst),
1284 is_mask);
1285 SW_FLOW_KEY_PUT(match, ct.orig_tp.src, ct->src_port, is_mask);
1286 SW_FLOW_KEY_PUT(match, ct.orig_tp.dst, ct->dst_port, is_mask);
1287 SW_FLOW_KEY_PUT(match, ct_orig_proto, ct->ipv6_proto, is_mask);
1288 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6);
1289 }
1290
1291 /* For layer 3 packets the Ethernet type is provided
1292 * and treated as metadata but no MAC addresses are provided.
1293 */
1294 if (!(*attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
1295 (*attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)))
1296 mac_proto = MAC_PROTO_NONE;
1297
1298 /* Always exact match mac_proto */
1299 SW_FLOW_KEY_PUT(match, mac_proto, is_mask ? 0xff : mac_proto, is_mask);
1300
1301 if (mac_proto == MAC_PROTO_NONE)
1302 return parse_eth_type_from_nlattrs(match, attrs, a, is_mask,
1303 log);
1304
1305 return 0;
1306 }
1307
nsh_hdr_from_nlattr(const struct nlattr * attr,struct nshhdr * nh,size_t size)1308 int nsh_hdr_from_nlattr(const struct nlattr *attr,
1309 struct nshhdr *nh, size_t size)
1310 {
1311 struct nlattr *a;
1312 int rem;
1313 u8 flags = 0;
1314 u8 ttl = 0;
1315 int mdlen = 0;
1316
1317 /* validate_nsh has check this, so we needn't do duplicate check here
1318 */
1319 if (size < NSH_BASE_HDR_LEN)
1320 return -ENOBUFS;
1321
1322 nla_for_each_nested(a, attr, rem) {
1323 int type = nla_type(a);
1324
1325 switch (type) {
1326 case OVS_NSH_KEY_ATTR_BASE: {
1327 const struct ovs_nsh_key_base *base = nla_data(a);
1328
1329 flags = base->flags;
1330 ttl = base->ttl;
1331 nh->np = base->np;
1332 nh->mdtype = base->mdtype;
1333 nh->path_hdr = base->path_hdr;
1334 break;
1335 }
1336 case OVS_NSH_KEY_ATTR_MD1:
1337 mdlen = nla_len(a);
1338 if (mdlen > size - NSH_BASE_HDR_LEN)
1339 return -ENOBUFS;
1340 memcpy(&nh->md1, nla_data(a), mdlen);
1341 break;
1342
1343 case OVS_NSH_KEY_ATTR_MD2:
1344 mdlen = nla_len(a);
1345 if (mdlen > size - NSH_BASE_HDR_LEN)
1346 return -ENOBUFS;
1347 memcpy(&nh->md2, nla_data(a), mdlen);
1348 break;
1349
1350 default:
1351 return -EINVAL;
1352 }
1353 }
1354
1355 /* nsh header length = NSH_BASE_HDR_LEN + mdlen */
1356 nh->ver_flags_ttl_len = 0;
1357 nsh_set_flags_ttl_len(nh, flags, ttl, NSH_BASE_HDR_LEN + mdlen);
1358
1359 return 0;
1360 }
1361
nsh_key_from_nlattr(const struct nlattr * attr,struct ovs_key_nsh * nsh,struct ovs_key_nsh * nsh_mask)1362 int nsh_key_from_nlattr(const struct nlattr *attr,
1363 struct ovs_key_nsh *nsh, struct ovs_key_nsh *nsh_mask)
1364 {
1365 struct nlattr *a;
1366 int rem;
1367
1368 /* validate_nsh has check this, so we needn't do duplicate check here
1369 */
1370 nla_for_each_nested(a, attr, rem) {
1371 int type = nla_type(a);
1372
1373 switch (type) {
1374 case OVS_NSH_KEY_ATTR_BASE: {
1375 const struct ovs_nsh_key_base *base = nla_data(a);
1376 const struct ovs_nsh_key_base *base_mask = base + 1;
1377
1378 nsh->base = *base;
1379 nsh_mask->base = *base_mask;
1380 break;
1381 }
1382 case OVS_NSH_KEY_ATTR_MD1: {
1383 const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1384 const struct ovs_nsh_key_md1 *md1_mask = md1 + 1;
1385
1386 memcpy(nsh->context, md1->context, sizeof(*md1));
1387 memcpy(nsh_mask->context, md1_mask->context,
1388 sizeof(*md1_mask));
1389 break;
1390 }
1391 case OVS_NSH_KEY_ATTR_MD2:
1392 /* Not supported yet */
1393 return -ENOTSUPP;
1394 default:
1395 return -EINVAL;
1396 }
1397 }
1398
1399 return 0;
1400 }
1401
nsh_key_put_from_nlattr(const struct nlattr * attr,struct sw_flow_match * match,bool is_mask,bool is_push_nsh,bool log)1402 static int nsh_key_put_from_nlattr(const struct nlattr *attr,
1403 struct sw_flow_match *match, bool is_mask,
1404 bool is_push_nsh, bool log)
1405 {
1406 struct nlattr *a;
1407 int rem;
1408 bool has_base = false;
1409 bool has_md1 = false;
1410 bool has_md2 = false;
1411 u8 mdtype = 0;
1412 int mdlen = 0;
1413
1414 if (WARN_ON(is_push_nsh && is_mask))
1415 return -EINVAL;
1416
1417 nla_for_each_nested(a, attr, rem) {
1418 int type = nla_type(a);
1419 int i;
1420
1421 if (type > OVS_NSH_KEY_ATTR_MAX) {
1422 OVS_NLERR(log, "nsh attr %d is out of range max %d",
1423 type, OVS_NSH_KEY_ATTR_MAX);
1424 return -EINVAL;
1425 }
1426
1427 if (!check_attr_len(nla_len(a),
1428 ovs_nsh_key_attr_lens[type].len)) {
1429 OVS_NLERR(
1430 log,
1431 "nsh attr %d has unexpected len %d expected %d",
1432 type,
1433 nla_len(a),
1434 ovs_nsh_key_attr_lens[type].len
1435 );
1436 return -EINVAL;
1437 }
1438
1439 switch (type) {
1440 case OVS_NSH_KEY_ATTR_BASE: {
1441 const struct ovs_nsh_key_base *base = nla_data(a);
1442
1443 has_base = true;
1444 mdtype = base->mdtype;
1445 SW_FLOW_KEY_PUT(match, nsh.base.flags,
1446 base->flags, is_mask);
1447 SW_FLOW_KEY_PUT(match, nsh.base.ttl,
1448 base->ttl, is_mask);
1449 SW_FLOW_KEY_PUT(match, nsh.base.mdtype,
1450 base->mdtype, is_mask);
1451 SW_FLOW_KEY_PUT(match, nsh.base.np,
1452 base->np, is_mask);
1453 SW_FLOW_KEY_PUT(match, nsh.base.path_hdr,
1454 base->path_hdr, is_mask);
1455 break;
1456 }
1457 case OVS_NSH_KEY_ATTR_MD1: {
1458 const struct ovs_nsh_key_md1 *md1 = nla_data(a);
1459
1460 has_md1 = true;
1461 for (i = 0; i < NSH_MD1_CONTEXT_SIZE; i++)
1462 SW_FLOW_KEY_PUT(match, nsh.context[i],
1463 md1->context[i], is_mask);
1464 break;
1465 }
1466 case OVS_NSH_KEY_ATTR_MD2:
1467 if (!is_push_nsh) /* Not supported MD type 2 yet */
1468 return -ENOTSUPP;
1469
1470 has_md2 = true;
1471 mdlen = nla_len(a);
1472 if (mdlen > NSH_CTX_HDRS_MAX_LEN || mdlen <= 0) {
1473 OVS_NLERR(
1474 log,
1475 "Invalid MD length %d for MD type %d",
1476 mdlen,
1477 mdtype
1478 );
1479 return -EINVAL;
1480 }
1481 break;
1482 default:
1483 OVS_NLERR(log, "Unknown nsh attribute %d",
1484 type);
1485 return -EINVAL;
1486 }
1487 }
1488
1489 if (rem > 0) {
1490 OVS_NLERR(log, "nsh attribute has %d unknown bytes.", rem);
1491 return -EINVAL;
1492 }
1493
1494 if (has_md1 && has_md2) {
1495 OVS_NLERR(
1496 1,
1497 "invalid nsh attribute: md1 and md2 are exclusive."
1498 );
1499 return -EINVAL;
1500 }
1501
1502 if (!is_mask) {
1503 if ((has_md1 && mdtype != NSH_M_TYPE1) ||
1504 (has_md2 && mdtype != NSH_M_TYPE2)) {
1505 OVS_NLERR(1, "nsh attribute has unmatched MD type %d.",
1506 mdtype);
1507 return -EINVAL;
1508 }
1509
1510 if (is_push_nsh &&
1511 (!has_base || (!has_md1 && !has_md2))) {
1512 OVS_NLERR(
1513 1,
1514 "push_nsh: missing base or metadata attributes"
1515 );
1516 return -EINVAL;
1517 }
1518 }
1519
1520 return 0;
1521 }
1522
ovs_key_from_nlattrs(struct net * net,struct sw_flow_match * match,u64 attrs,const struct nlattr ** a,bool is_mask,bool log)1523 static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
1524 u64 attrs, const struct nlattr **a,
1525 bool is_mask, bool log)
1526 {
1527 int err;
1528
1529 err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
1530 if (err)
1531 return err;
1532
1533 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1534 const struct ovs_key_ethernet *eth_key;
1535
1536 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1537 SW_FLOW_KEY_MEMCPY(match, eth.src,
1538 eth_key->eth_src, ETH_ALEN, is_mask);
1539 SW_FLOW_KEY_MEMCPY(match, eth.dst,
1540 eth_key->eth_dst, ETH_ALEN, is_mask);
1541 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1542
1543 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1544 /* VLAN attribute is always parsed before getting here since it
1545 * may occur multiple times.
1546 */
1547 OVS_NLERR(log, "VLAN attribute unexpected.");
1548 return -EINVAL;
1549 }
1550
1551 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1552 err = parse_eth_type_from_nlattrs(match, &attrs, a, is_mask,
1553 log);
1554 if (err)
1555 return err;
1556 } else if (!is_mask) {
1557 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1558 }
1559 } else if (!match->key->eth.type) {
1560 OVS_NLERR(log, "Either Ethernet header or EtherType is required.");
1561 return -EINVAL;
1562 }
1563
1564 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1565 const struct ovs_key_ipv4 *ipv4_key;
1566
1567 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1568 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
1569 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
1570 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
1571 return -EINVAL;
1572 }
1573 SW_FLOW_KEY_PUT(match, ip.proto,
1574 ipv4_key->ipv4_proto, is_mask);
1575 SW_FLOW_KEY_PUT(match, ip.tos,
1576 ipv4_key->ipv4_tos, is_mask);
1577 SW_FLOW_KEY_PUT(match, ip.ttl,
1578 ipv4_key->ipv4_ttl, is_mask);
1579 SW_FLOW_KEY_PUT(match, ip.frag,
1580 ipv4_key->ipv4_frag, is_mask);
1581 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1582 ipv4_key->ipv4_src, is_mask);
1583 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1584 ipv4_key->ipv4_dst, is_mask);
1585 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1586 }
1587
1588 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1589 const struct ovs_key_ipv6 *ipv6_key;
1590
1591 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1592 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
1593 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
1594 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
1595 return -EINVAL;
1596 }
1597
1598 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
1599 OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x)",
1600 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
1601 return -EINVAL;
1602 }
1603
1604 SW_FLOW_KEY_PUT(match, ipv6.label,
1605 ipv6_key->ipv6_label, is_mask);
1606 SW_FLOW_KEY_PUT(match, ip.proto,
1607 ipv6_key->ipv6_proto, is_mask);
1608 SW_FLOW_KEY_PUT(match, ip.tos,
1609 ipv6_key->ipv6_tclass, is_mask);
1610 SW_FLOW_KEY_PUT(match, ip.ttl,
1611 ipv6_key->ipv6_hlimit, is_mask);
1612 SW_FLOW_KEY_PUT(match, ip.frag,
1613 ipv6_key->ipv6_frag, is_mask);
1614 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1615 ipv6_key->ipv6_src,
1616 sizeof(match->key->ipv6.addr.src),
1617 is_mask);
1618 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1619 ipv6_key->ipv6_dst,
1620 sizeof(match->key->ipv6.addr.dst),
1621 is_mask);
1622
1623 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1624 }
1625
1626 if (attrs & (1ULL << OVS_KEY_ATTR_IPV6_EXTHDRS)) {
1627 const struct ovs_key_ipv6_exthdrs *ipv6_exthdrs_key;
1628
1629 ipv6_exthdrs_key = nla_data(a[OVS_KEY_ATTR_IPV6_EXTHDRS]);
1630
1631 SW_FLOW_KEY_PUT(match, ipv6.exthdrs,
1632 ipv6_exthdrs_key->hdrs, is_mask);
1633
1634 attrs &= ~(1ULL << OVS_KEY_ATTR_IPV6_EXTHDRS);
1635 }
1636
1637 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1638 const struct ovs_key_arp *arp_key;
1639
1640 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1641 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1642 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
1643 arp_key->arp_op);
1644 return -EINVAL;
1645 }
1646
1647 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1648 arp_key->arp_sip, is_mask);
1649 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1650 arp_key->arp_tip, is_mask);
1651 SW_FLOW_KEY_PUT(match, ip.proto,
1652 ntohs(arp_key->arp_op), is_mask);
1653 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1654 arp_key->arp_sha, ETH_ALEN, is_mask);
1655 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1656 arp_key->arp_tha, ETH_ALEN, is_mask);
1657
1658 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1659 }
1660
1661 if (attrs & (1 << OVS_KEY_ATTR_NSH)) {
1662 if (nsh_key_put_from_nlattr(a[OVS_KEY_ATTR_NSH], match,
1663 is_mask, false, log) < 0)
1664 return -EINVAL;
1665 attrs &= ~(1 << OVS_KEY_ATTR_NSH);
1666 }
1667
1668 if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1669 const struct ovs_key_mpls *mpls_key;
1670 u32 hdr_len;
1671 u32 label_count, label_count_mask, i;
1672
1673 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1674 hdr_len = nla_len(a[OVS_KEY_ATTR_MPLS]);
1675 label_count = hdr_len / sizeof(struct ovs_key_mpls);
1676
1677 if (label_count == 0 || label_count > MPLS_LABEL_DEPTH ||
1678 hdr_len % sizeof(struct ovs_key_mpls))
1679 return -EINVAL;
1680
1681 label_count_mask = GENMASK(label_count - 1, 0);
1682
1683 for (i = 0 ; i < label_count; i++)
1684 SW_FLOW_KEY_PUT(match, mpls.lse[i],
1685 mpls_key[i].mpls_lse, is_mask);
1686
1687 SW_FLOW_KEY_PUT(match, mpls.num_labels_mask,
1688 label_count_mask, is_mask);
1689
1690 attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1691 }
1692
1693 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1694 const struct ovs_key_tcp *tcp_key;
1695
1696 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1697 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1698 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1699 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1700 }
1701
1702 if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1703 SW_FLOW_KEY_PUT(match, tp.flags,
1704 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1705 is_mask);
1706 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1707 }
1708
1709 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1710 const struct ovs_key_udp *udp_key;
1711
1712 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1713 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1714 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1715 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1716 }
1717
1718 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1719 const struct ovs_key_sctp *sctp_key;
1720
1721 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1722 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1723 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1724 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1725 }
1726
1727 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1728 const struct ovs_key_icmp *icmp_key;
1729
1730 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1731 SW_FLOW_KEY_PUT(match, tp.src,
1732 htons(icmp_key->icmp_type), is_mask);
1733 SW_FLOW_KEY_PUT(match, tp.dst,
1734 htons(icmp_key->icmp_code), is_mask);
1735 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1736 }
1737
1738 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1739 const struct ovs_key_icmpv6 *icmpv6_key;
1740
1741 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1742 SW_FLOW_KEY_PUT(match, tp.src,
1743 htons(icmpv6_key->icmpv6_type), is_mask);
1744 SW_FLOW_KEY_PUT(match, tp.dst,
1745 htons(icmpv6_key->icmpv6_code), is_mask);
1746 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1747 }
1748
1749 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1750 const struct ovs_key_nd *nd_key;
1751
1752 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1753 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1754 nd_key->nd_target,
1755 sizeof(match->key->ipv6.nd.target),
1756 is_mask);
1757 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1758 nd_key->nd_sll, ETH_ALEN, is_mask);
1759 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1760 nd_key->nd_tll, ETH_ALEN, is_mask);
1761 attrs &= ~(1 << OVS_KEY_ATTR_ND);
1762 }
1763
1764 if (attrs != 0) {
1765 OVS_NLERR(log, "Unknown key attributes %llx",
1766 (unsigned long long)attrs);
1767 return -EINVAL;
1768 }
1769
1770 return 0;
1771 }
1772
nlattr_set(struct nlattr * attr,u8 val,const struct ovs_len_tbl * tbl)1773 static void nlattr_set(struct nlattr *attr, u8 val,
1774 const struct ovs_len_tbl *tbl)
1775 {
1776 struct nlattr *nla;
1777 int rem;
1778
1779 /* The nlattr stream should already have been validated */
1780 nla_for_each_nested(nla, attr, rem) {
1781 if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED)
1782 nlattr_set(nla, val, tbl[nla_type(nla)].next ? : tbl);
1783 else
1784 memset(nla_data(nla), val, nla_len(nla));
1785
1786 if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1787 *(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1788 }
1789 }
1790
mask_set_nlattr(struct nlattr * attr,u8 val)1791 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1792 {
1793 nlattr_set(attr, val, ovs_key_lens);
1794 }
1795
1796 /**
1797 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1798 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1799 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1800 * does not include any don't care bit.
1801 * @net: Used to determine per-namespace field support.
1802 * @match: receives the extracted flow match information.
1803 * @nla_key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1804 * sequence. The fields should of the packet that triggered the creation
1805 * of this flow.
1806 * @nla_mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_*
1807 * Netlink attribute specifies the mask field of the wildcarded flow.
1808 * @log: Boolean to allow kernel error logging. Normally true, but when
1809 * probing for feature compatibility this should be passed in as false to
1810 * suppress unnecessary error logging.
1811 */
ovs_nla_get_match(struct net * net,struct sw_flow_match * match,const struct nlattr * nla_key,const struct nlattr * nla_mask,bool log)1812 int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1813 const struct nlattr *nla_key,
1814 const struct nlattr *nla_mask,
1815 bool log)
1816 {
1817 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1818 struct nlattr *newmask = NULL;
1819 u64 key_attrs = 0;
1820 u64 mask_attrs = 0;
1821 int err;
1822
1823 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1824 if (err)
1825 return err;
1826
1827 err = parse_vlan_from_nlattrs(match, &key_attrs, a, false, log);
1828 if (err)
1829 return err;
1830
1831 err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1832 if (err)
1833 return err;
1834
1835 if (match->mask) {
1836 if (!nla_mask) {
1837 /* Create an exact match mask. We need to set to 0xff
1838 * all the 'match->mask' fields that have been touched
1839 * in 'match->key'. We cannot simply memset
1840 * 'match->mask', because padding bytes and fields not
1841 * specified in 'match->key' should be left to 0.
1842 * Instead, we use a stream of netlink attributes,
1843 * copied from 'key' and set to 0xff.
1844 * ovs_key_from_nlattrs() will take care of filling
1845 * 'match->mask' appropriately.
1846 */
1847 newmask = kmemdup(nla_key,
1848 nla_total_size(nla_len(nla_key)),
1849 GFP_KERNEL);
1850 if (!newmask)
1851 return -ENOMEM;
1852
1853 mask_set_nlattr(newmask, 0xff);
1854
1855 /* The userspace does not send tunnel attributes that
1856 * are 0, but we should not wildcard them nonetheless.
1857 */
1858 if (match->key->tun_proto)
1859 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1860 0xff, true);
1861
1862 nla_mask = newmask;
1863 }
1864
1865 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1866 if (err)
1867 goto free_newmask;
1868
1869 /* Always match on tci. */
1870 SW_FLOW_KEY_PUT(match, eth.vlan.tci, htons(0xffff), true);
1871 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, htons(0xffff), true);
1872
1873 err = parse_vlan_from_nlattrs(match, &mask_attrs, a, true, log);
1874 if (err)
1875 goto free_newmask;
1876
1877 err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1878 log);
1879 if (err)
1880 goto free_newmask;
1881 }
1882
1883 if (!match_validate(match, key_attrs, mask_attrs, log))
1884 err = -EINVAL;
1885
1886 free_newmask:
1887 kfree(newmask);
1888 return err;
1889 }
1890
get_ufid_len(const struct nlattr * attr,bool log)1891 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1892 {
1893 size_t len;
1894
1895 if (!attr)
1896 return 0;
1897
1898 len = nla_len(attr);
1899 if (len < 1 || len > MAX_UFID_LENGTH) {
1900 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1901 nla_len(attr), MAX_UFID_LENGTH);
1902 return 0;
1903 }
1904
1905 return len;
1906 }
1907
1908 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1909 * or false otherwise.
1910 */
ovs_nla_get_ufid(struct sw_flow_id * sfid,const struct nlattr * attr,bool log)1911 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1912 bool log)
1913 {
1914 sfid->ufid_len = get_ufid_len(attr, log);
1915 if (sfid->ufid_len)
1916 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1917
1918 return sfid->ufid_len;
1919 }
1920
ovs_nla_get_identifier(struct sw_flow_id * sfid,const struct nlattr * ufid,const struct sw_flow_key * key,bool log)1921 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1922 const struct sw_flow_key *key, bool log)
1923 {
1924 struct sw_flow_key *new_key;
1925
1926 if (ovs_nla_get_ufid(sfid, ufid, log))
1927 return 0;
1928
1929 /* If UFID was not provided, use unmasked key. */
1930 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1931 if (!new_key)
1932 return -ENOMEM;
1933 memcpy(new_key, key, sizeof(*key));
1934 sfid->unmasked_key = new_key;
1935
1936 return 0;
1937 }
1938
ovs_nla_get_ufid_flags(const struct nlattr * attr)1939 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1940 {
1941 return nla_get_u32_default(attr, 0);
1942 }
1943
1944 /**
1945 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1946 * @net: Network namespace.
1947 * @key: Receives extracted in_port, priority, tun_key, skb_mark and conntrack
1948 * metadata.
1949 * @a: Array of netlink attributes holding parsed %OVS_KEY_ATTR_* Netlink
1950 * attributes.
1951 * @attrs: Bit mask for the netlink attributes included in @a.
1952 * @log: Boolean to allow kernel error logging. Normally true, but when
1953 * probing for feature compatibility this should be passed in as false to
1954 * suppress unnecessary error logging.
1955 *
1956 * This parses a series of Netlink attributes that form a flow key, which must
1957 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1958 * get the metadata, that is, the parts of the flow key that cannot be
1959 * extracted from the packet itself.
1960 *
1961 * This must be called before the packet key fields are filled in 'key'.
1962 */
1963
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)1964 int ovs_nla_get_flow_metadata(struct net *net,
1965 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1],
1966 u64 attrs, struct sw_flow_key *key, bool log)
1967 {
1968 struct sw_flow_match match;
1969
1970 memset(&match, 0, sizeof(match));
1971 match.key = key;
1972
1973 key->ct_state = 0;
1974 key->ct_zone = 0;
1975 key->ct_orig_proto = 0;
1976 memset(&key->ct, 0, sizeof(key->ct));
1977 memset(&key->ipv4.ct_orig, 0, sizeof(key->ipv4.ct_orig));
1978 memset(&key->ipv6.ct_orig, 0, sizeof(key->ipv6.ct_orig));
1979
1980 key->phy.in_port = DP_MAX_PORTS;
1981
1982 return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1983 }
1984
ovs_nla_put_vlan(struct sk_buff * skb,const struct vlan_head * vh,bool is_mask)1985 static int ovs_nla_put_vlan(struct sk_buff *skb, const struct vlan_head *vh,
1986 bool is_mask)
1987 {
1988 __be16 eth_type = !is_mask ? vh->tpid : htons(0xffff);
1989
1990 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1991 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, vh->tci))
1992 return -EMSGSIZE;
1993 return 0;
1994 }
1995
nsh_key_to_nlattr(const struct ovs_key_nsh * nsh,bool is_mask,struct sk_buff * skb)1996 static int nsh_key_to_nlattr(const struct ovs_key_nsh *nsh, bool is_mask,
1997 struct sk_buff *skb)
1998 {
1999 struct nlattr *start;
2000
2001 start = nla_nest_start_noflag(skb, OVS_KEY_ATTR_NSH);
2002 if (!start)
2003 return -EMSGSIZE;
2004
2005 if (nla_put(skb, OVS_NSH_KEY_ATTR_BASE, sizeof(nsh->base), &nsh->base))
2006 goto nla_put_failure;
2007
2008 if (is_mask || nsh->base.mdtype == NSH_M_TYPE1) {
2009 if (nla_put(skb, OVS_NSH_KEY_ATTR_MD1,
2010 sizeof(nsh->context), nsh->context))
2011 goto nla_put_failure;
2012 }
2013
2014 /* Don't support MD type 2 yet */
2015
2016 nla_nest_end(skb, start);
2017
2018 return 0;
2019
2020 nla_put_failure:
2021 return -EMSGSIZE;
2022 }
2023
__ovs_nla_put_key(const struct sw_flow_key * swkey,const struct sw_flow_key * output,bool is_mask,struct sk_buff * skb)2024 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
2025 const struct sw_flow_key *output, bool is_mask,
2026 struct sk_buff *skb)
2027 {
2028 struct ovs_key_ethernet *eth_key;
2029 struct nlattr *nla;
2030 struct nlattr *encap = NULL;
2031 struct nlattr *in_encap = NULL;
2032
2033 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
2034 goto nla_put_failure;
2035
2036 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
2037 goto nla_put_failure;
2038
2039 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
2040 goto nla_put_failure;
2041
2042 if ((swkey->tun_proto || is_mask)) {
2043 const void *opts = NULL;
2044
2045 if (ip_tunnel_is_options_present(output->tun_key.tun_flags))
2046 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
2047
2048 if (ip_tun_to_nlattr(skb, &output->tun_key, opts,
2049 swkey->tun_opts_len, swkey->tun_proto, 0))
2050 goto nla_put_failure;
2051 }
2052
2053 if (swkey->phy.in_port == DP_MAX_PORTS) {
2054 if (is_mask && (output->phy.in_port == 0xffff))
2055 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
2056 goto nla_put_failure;
2057 } else {
2058 u16 upper_u16;
2059 upper_u16 = !is_mask ? 0 : 0xffff;
2060
2061 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
2062 (upper_u16 << 16) | output->phy.in_port))
2063 goto nla_put_failure;
2064 }
2065
2066 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
2067 goto nla_put_failure;
2068
2069 if (ovs_ct_put_key(swkey, output, skb))
2070 goto nla_put_failure;
2071
2072 if (ovs_key_mac_proto(swkey) == MAC_PROTO_ETHERNET) {
2073 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
2074 if (!nla)
2075 goto nla_put_failure;
2076
2077 eth_key = nla_data(nla);
2078 ether_addr_copy(eth_key->eth_src, output->eth.src);
2079 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
2080
2081 if (swkey->eth.vlan.tci || eth_type_vlan(swkey->eth.type)) {
2082 if (ovs_nla_put_vlan(skb, &output->eth.vlan, is_mask))
2083 goto nla_put_failure;
2084 encap = nla_nest_start_noflag(skb, OVS_KEY_ATTR_ENCAP);
2085 if (!swkey->eth.vlan.tci)
2086 goto unencap;
2087
2088 if (swkey->eth.cvlan.tci || eth_type_vlan(swkey->eth.type)) {
2089 if (ovs_nla_put_vlan(skb, &output->eth.cvlan, is_mask))
2090 goto nla_put_failure;
2091 in_encap = nla_nest_start_noflag(skb,
2092 OVS_KEY_ATTR_ENCAP);
2093 if (!swkey->eth.cvlan.tci)
2094 goto unencap;
2095 }
2096 }
2097
2098 if (swkey->eth.type == htons(ETH_P_802_2)) {
2099 /*
2100 * Ethertype 802.2 is represented in the netlink with omitted
2101 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
2102 * 0xffff in the mask attribute. Ethertype can also
2103 * be wildcarded.
2104 */
2105 if (is_mask && output->eth.type)
2106 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
2107 output->eth.type))
2108 goto nla_put_failure;
2109 goto unencap;
2110 }
2111 }
2112
2113 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
2114 goto nla_put_failure;
2115
2116 if (eth_type_vlan(swkey->eth.type)) {
2117 /* There are 3 VLAN tags, we don't know anything about the rest
2118 * of the packet, so truncate here.
2119 */
2120 WARN_ON_ONCE(!(encap && in_encap));
2121 goto unencap;
2122 }
2123
2124 if (swkey->eth.type == htons(ETH_P_IP)) {
2125 struct ovs_key_ipv4 *ipv4_key;
2126
2127 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
2128 if (!nla)
2129 goto nla_put_failure;
2130 ipv4_key = nla_data(nla);
2131 ipv4_key->ipv4_src = output->ipv4.addr.src;
2132 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
2133 ipv4_key->ipv4_proto = output->ip.proto;
2134 ipv4_key->ipv4_tos = output->ip.tos;
2135 ipv4_key->ipv4_ttl = output->ip.ttl;
2136 ipv4_key->ipv4_frag = output->ip.frag;
2137 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
2138 struct ovs_key_ipv6 *ipv6_key;
2139 struct ovs_key_ipv6_exthdrs *ipv6_exthdrs_key;
2140
2141 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
2142 if (!nla)
2143 goto nla_put_failure;
2144 ipv6_key = nla_data(nla);
2145 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
2146 sizeof(ipv6_key->ipv6_src));
2147 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
2148 sizeof(ipv6_key->ipv6_dst));
2149 ipv6_key->ipv6_label = output->ipv6.label;
2150 ipv6_key->ipv6_proto = output->ip.proto;
2151 ipv6_key->ipv6_tclass = output->ip.tos;
2152 ipv6_key->ipv6_hlimit = output->ip.ttl;
2153 ipv6_key->ipv6_frag = output->ip.frag;
2154
2155 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6_EXTHDRS,
2156 sizeof(*ipv6_exthdrs_key));
2157 if (!nla)
2158 goto nla_put_failure;
2159 ipv6_exthdrs_key = nla_data(nla);
2160 ipv6_exthdrs_key->hdrs = output->ipv6.exthdrs;
2161 } else if (swkey->eth.type == htons(ETH_P_NSH)) {
2162 if (nsh_key_to_nlattr(&output->nsh, is_mask, skb))
2163 goto nla_put_failure;
2164 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
2165 swkey->eth.type == htons(ETH_P_RARP)) {
2166 struct ovs_key_arp *arp_key;
2167
2168 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
2169 if (!nla)
2170 goto nla_put_failure;
2171 arp_key = nla_data(nla);
2172 memset(arp_key, 0, sizeof(struct ovs_key_arp));
2173 arp_key->arp_sip = output->ipv4.addr.src;
2174 arp_key->arp_tip = output->ipv4.addr.dst;
2175 arp_key->arp_op = htons(output->ip.proto);
2176 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
2177 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
2178 } else if (eth_p_mpls(swkey->eth.type)) {
2179 u8 i, num_labels;
2180 struct ovs_key_mpls *mpls_key;
2181
2182 num_labels = hweight_long(output->mpls.num_labels_mask);
2183 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS,
2184 num_labels * sizeof(*mpls_key));
2185 if (!nla)
2186 goto nla_put_failure;
2187
2188 mpls_key = nla_data(nla);
2189 for (i = 0; i < num_labels; i++)
2190 mpls_key[i].mpls_lse = output->mpls.lse[i];
2191 }
2192
2193 if ((swkey->eth.type == htons(ETH_P_IP) ||
2194 swkey->eth.type == htons(ETH_P_IPV6)) &&
2195 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
2196
2197 if (swkey->ip.proto == IPPROTO_TCP) {
2198 struct ovs_key_tcp *tcp_key;
2199
2200 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
2201 if (!nla)
2202 goto nla_put_failure;
2203 tcp_key = nla_data(nla);
2204 tcp_key->tcp_src = output->tp.src;
2205 tcp_key->tcp_dst = output->tp.dst;
2206 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
2207 output->tp.flags))
2208 goto nla_put_failure;
2209 } else if (swkey->ip.proto == IPPROTO_UDP) {
2210 struct ovs_key_udp *udp_key;
2211
2212 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
2213 if (!nla)
2214 goto nla_put_failure;
2215 udp_key = nla_data(nla);
2216 udp_key->udp_src = output->tp.src;
2217 udp_key->udp_dst = output->tp.dst;
2218 } else if (swkey->ip.proto == IPPROTO_SCTP) {
2219 struct ovs_key_sctp *sctp_key;
2220
2221 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
2222 if (!nla)
2223 goto nla_put_failure;
2224 sctp_key = nla_data(nla);
2225 sctp_key->sctp_src = output->tp.src;
2226 sctp_key->sctp_dst = output->tp.dst;
2227 } else if (swkey->eth.type == htons(ETH_P_IP) &&
2228 swkey->ip.proto == IPPROTO_ICMP) {
2229 struct ovs_key_icmp *icmp_key;
2230
2231 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
2232 if (!nla)
2233 goto nla_put_failure;
2234 icmp_key = nla_data(nla);
2235 icmp_key->icmp_type = ntohs(output->tp.src);
2236 icmp_key->icmp_code = ntohs(output->tp.dst);
2237 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
2238 swkey->ip.proto == IPPROTO_ICMPV6) {
2239 struct ovs_key_icmpv6 *icmpv6_key;
2240
2241 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
2242 sizeof(*icmpv6_key));
2243 if (!nla)
2244 goto nla_put_failure;
2245 icmpv6_key = nla_data(nla);
2246 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
2247 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
2248
2249 if (swkey->tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
2250 swkey->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
2251 struct ovs_key_nd *nd_key;
2252
2253 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
2254 if (!nla)
2255 goto nla_put_failure;
2256 nd_key = nla_data(nla);
2257 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
2258 sizeof(nd_key->nd_target));
2259 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
2260 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
2261 }
2262 }
2263 }
2264
2265 unencap:
2266 if (in_encap)
2267 nla_nest_end(skb, in_encap);
2268 if (encap)
2269 nla_nest_end(skb, encap);
2270
2271 return 0;
2272
2273 nla_put_failure:
2274 return -EMSGSIZE;
2275 }
2276
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)2277 int ovs_nla_put_key(const struct sw_flow_key *swkey,
2278 const struct sw_flow_key *output, int attr, bool is_mask,
2279 struct sk_buff *skb)
2280 {
2281 int err;
2282 struct nlattr *nla;
2283
2284 nla = nla_nest_start_noflag(skb, attr);
2285 if (!nla)
2286 return -EMSGSIZE;
2287 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
2288 if (err)
2289 return err;
2290 nla_nest_end(skb, nla);
2291
2292 return 0;
2293 }
2294
2295 /* Called with ovs_mutex or RCU read lock. */
ovs_nla_put_identifier(const struct sw_flow * flow,struct sk_buff * skb)2296 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
2297 {
2298 if (ovs_identifier_is_ufid(&flow->id))
2299 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
2300 flow->id.ufid);
2301
2302 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
2303 OVS_FLOW_ATTR_KEY, false, skb);
2304 }
2305
2306 /* Called with ovs_mutex or RCU read lock. */
ovs_nla_put_masked_key(const struct sw_flow * flow,struct sk_buff * skb)2307 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
2308 {
2309 return ovs_nla_put_key(&flow->key, &flow->key,
2310 OVS_FLOW_ATTR_KEY, false, skb);
2311 }
2312
2313 /* Called with ovs_mutex or RCU read lock. */
ovs_nla_put_mask(const struct sw_flow * flow,struct sk_buff * skb)2314 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
2315 {
2316 return ovs_nla_put_key(&flow->key, &flow->mask->key,
2317 OVS_FLOW_ATTR_MASK, true, skb);
2318 }
2319
2320 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
2321
nla_alloc_flow_actions(int size)2322 static struct sw_flow_actions *nla_alloc_flow_actions(int size)
2323 {
2324 struct sw_flow_actions *sfa;
2325
2326 WARN_ON_ONCE(size > MAX_ACTIONS_BUFSIZE);
2327
2328 sfa = kmalloc(kmalloc_size_roundup(sizeof(*sfa) + size), GFP_KERNEL);
2329 if (!sfa)
2330 return ERR_PTR(-ENOMEM);
2331
2332 sfa->actions_len = 0;
2333 return sfa;
2334 }
2335
2336 static void ovs_nla_free_nested_actions(const struct nlattr *actions, int len);
2337
ovs_nla_free_check_pkt_len_action(const struct nlattr * action)2338 static void ovs_nla_free_check_pkt_len_action(const struct nlattr *action)
2339 {
2340 const struct nlattr *a;
2341 int rem;
2342
2343 nla_for_each_nested(a, action, rem) {
2344 switch (nla_type(a)) {
2345 case OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL:
2346 case OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER:
2347 ovs_nla_free_nested_actions(nla_data(a), nla_len(a));
2348 break;
2349 }
2350 }
2351 }
2352
ovs_nla_free_clone_action(const struct nlattr * action)2353 static void ovs_nla_free_clone_action(const struct nlattr *action)
2354 {
2355 const struct nlattr *a = nla_data(action);
2356 int rem = nla_len(action);
2357
2358 switch (nla_type(a)) {
2359 case OVS_CLONE_ATTR_EXEC:
2360 /* The real list of actions follows this attribute. */
2361 a = nla_next(a, &rem);
2362 ovs_nla_free_nested_actions(a, rem);
2363 break;
2364 }
2365 }
2366
ovs_nla_free_dec_ttl_action(const struct nlattr * action)2367 static void ovs_nla_free_dec_ttl_action(const struct nlattr *action)
2368 {
2369 const struct nlattr *a = nla_data(action);
2370
2371 switch (nla_type(a)) {
2372 case OVS_DEC_TTL_ATTR_ACTION:
2373 ovs_nla_free_nested_actions(nla_data(a), nla_len(a));
2374 break;
2375 }
2376 }
2377
ovs_nla_free_sample_action(const struct nlattr * action)2378 static void ovs_nla_free_sample_action(const struct nlattr *action)
2379 {
2380 const struct nlattr *a = nla_data(action);
2381 int rem = nla_len(action);
2382
2383 switch (nla_type(a)) {
2384 case OVS_SAMPLE_ATTR_ARG:
2385 /* The real list of actions follows this attribute. */
2386 a = nla_next(a, &rem);
2387 ovs_nla_free_nested_actions(a, rem);
2388 break;
2389 }
2390 }
2391
ovs_nla_free_set_action(const struct nlattr * a)2392 static void ovs_nla_free_set_action(const struct nlattr *a)
2393 {
2394 const struct nlattr *ovs_key = nla_data(a);
2395 struct ovs_tunnel_info *ovs_tun;
2396
2397 switch (nla_type(ovs_key)) {
2398 case OVS_KEY_ATTR_TUNNEL_INFO:
2399 ovs_tun = nla_data(ovs_key);
2400 dst_release((struct dst_entry *)ovs_tun->tun_dst);
2401 break;
2402 }
2403 }
2404
ovs_nla_free_nested_actions(const struct nlattr * actions,int len)2405 static void ovs_nla_free_nested_actions(const struct nlattr *actions, int len)
2406 {
2407 const struct nlattr *a;
2408 int rem;
2409
2410 /* Whenever new actions are added, the need to update this
2411 * function should be considered.
2412 */
2413 BUILD_BUG_ON(OVS_ACTION_ATTR_MAX != 25);
2414
2415 if (!actions)
2416 return;
2417
2418 nla_for_each_attr(a, actions, len, rem) {
2419 switch (nla_type(a)) {
2420 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
2421 ovs_nla_free_check_pkt_len_action(a);
2422 break;
2423
2424 case OVS_ACTION_ATTR_CLONE:
2425 ovs_nla_free_clone_action(a);
2426 break;
2427
2428 case OVS_ACTION_ATTR_CT:
2429 ovs_ct_free_action(a);
2430 break;
2431
2432 case OVS_ACTION_ATTR_DEC_TTL:
2433 ovs_nla_free_dec_ttl_action(a);
2434 break;
2435
2436 case OVS_ACTION_ATTR_SAMPLE:
2437 ovs_nla_free_sample_action(a);
2438 break;
2439
2440 case OVS_ACTION_ATTR_SET:
2441 ovs_nla_free_set_action(a);
2442 break;
2443 }
2444 }
2445 }
2446
ovs_nla_free_flow_actions(struct sw_flow_actions * sf_acts)2447 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
2448 {
2449 if (!sf_acts)
2450 return;
2451
2452 ovs_nla_free_nested_actions(sf_acts->actions, sf_acts->actions_len);
2453 kfree(sf_acts);
2454 }
2455
__ovs_nla_free_flow_actions(struct rcu_head * head)2456 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
2457 {
2458 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
2459 }
2460
2461 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
2462 * The caller must hold rcu_read_lock for this to be sensible. */
ovs_nla_free_flow_actions_rcu(struct sw_flow_actions * sf_acts)2463 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
2464 {
2465 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
2466 }
2467
reserve_sfa_size(struct sw_flow_actions ** sfa,int attr_len,bool log)2468 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
2469 int attr_len, bool log)
2470 {
2471
2472 struct sw_flow_actions *acts;
2473 int new_acts_size;
2474 size_t req_size = NLA_ALIGN(attr_len);
2475 int next_offset = offsetof(struct sw_flow_actions, actions) +
2476 (*sfa)->actions_len;
2477
2478 if (req_size <= (ksize(*sfa) - next_offset))
2479 goto out;
2480
2481 new_acts_size = max(next_offset + req_size, ksize(*sfa) * 2);
2482
2483 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
2484 if ((next_offset + req_size) > MAX_ACTIONS_BUFSIZE) {
2485 OVS_NLERR(log, "Flow action size exceeds max %u",
2486 MAX_ACTIONS_BUFSIZE);
2487 return ERR_PTR(-EMSGSIZE);
2488 }
2489 new_acts_size = MAX_ACTIONS_BUFSIZE;
2490 }
2491
2492 acts = nla_alloc_flow_actions(new_acts_size);
2493 if (IS_ERR(acts))
2494 return ERR_CAST(acts);
2495
2496 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
2497 acts->actions_len = (*sfa)->actions_len;
2498 acts->orig_len = (*sfa)->orig_len;
2499 kfree(*sfa);
2500 *sfa = acts;
2501
2502 out:
2503 (*sfa)->actions_len += req_size;
2504 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
2505 }
2506
__add_action(struct sw_flow_actions ** sfa,int attrtype,void * data,int len,bool log)2507 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
2508 int attrtype, void *data, int len, bool log)
2509 {
2510 struct nlattr *a;
2511
2512 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
2513 if (IS_ERR(a))
2514 return a;
2515
2516 a->nla_type = attrtype;
2517 a->nla_len = nla_attr_size(len);
2518
2519 if (data)
2520 memcpy(nla_data(a), data, len);
2521 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
2522
2523 return a;
2524 }
2525
ovs_nla_add_action(struct sw_flow_actions ** sfa,int attrtype,void * data,int len,bool log)2526 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
2527 int len, bool log)
2528 {
2529 struct nlattr *a;
2530
2531 a = __add_action(sfa, attrtype, data, len, log);
2532
2533 return PTR_ERR_OR_ZERO(a);
2534 }
2535
add_nested_action_start(struct sw_flow_actions ** sfa,int attrtype,bool log)2536 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
2537 int attrtype, bool log)
2538 {
2539 int used = (*sfa)->actions_len;
2540 int err;
2541
2542 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
2543 if (err)
2544 return err;
2545
2546 return used;
2547 }
2548
add_nested_action_end(struct sw_flow_actions * sfa,int st_offset)2549 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
2550 int st_offset)
2551 {
2552 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
2553 st_offset);
2554
2555 a->nla_len = sfa->actions_len - st_offset;
2556 }
2557
2558 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2559 const struct sw_flow_key *key,
2560 struct sw_flow_actions **sfa,
2561 __be16 eth_type, __be16 vlan_tci,
2562 u32 mpls_label_count, bool log,
2563 u32 depth);
2564
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,u32 depth)2565 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
2566 const struct sw_flow_key *key,
2567 struct sw_flow_actions **sfa,
2568 __be16 eth_type, __be16 vlan_tci,
2569 u32 mpls_label_count, bool log, bool last,
2570 u32 depth)
2571 {
2572 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
2573 const struct nlattr *probability, *actions;
2574 const struct nlattr *a;
2575 int rem, start, err;
2576 struct sample_arg arg;
2577
2578 memset(attrs, 0, sizeof(attrs));
2579 nla_for_each_nested(a, attr, rem) {
2580 int type = nla_type(a);
2581 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
2582 return -EINVAL;
2583 attrs[type] = a;
2584 }
2585 if (rem)
2586 return -EINVAL;
2587
2588 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
2589 if (!probability || nla_len(probability) != sizeof(u32))
2590 return -EINVAL;
2591
2592 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
2593 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
2594 return -EINVAL;
2595
2596 /* validation done, copy sample action. */
2597 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
2598 if (start < 0)
2599 return start;
2600
2601 /* When both skb and flow may be changed, put the sample
2602 * into a deferred fifo. On the other hand, if only skb
2603 * may be modified, the actions can be executed in place.
2604 *
2605 * Do this analysis at the flow installation time.
2606 * Set 'clone_action->exec' to true if the actions can be
2607 * executed without being deferred.
2608 *
2609 * If the sample is the last action, it can always be excuted
2610 * rather than deferred.
2611 */
2612 arg.exec = last || !actions_may_change_flow(actions);
2613 arg.probability = nla_get_u32(probability);
2614
2615 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_ARG, &arg, sizeof(arg),
2616 log);
2617 if (err)
2618 return err;
2619
2620 err = __ovs_nla_copy_actions(net, actions, key, sfa,
2621 eth_type, vlan_tci, mpls_label_count, log,
2622 depth + 1);
2623
2624 if (err)
2625 return err;
2626
2627 add_nested_action_end(*sfa, start);
2628
2629 return 0;
2630 }
2631
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,u32 depth)2632 static int validate_and_copy_dec_ttl(struct net *net,
2633 const struct nlattr *attr,
2634 const struct sw_flow_key *key,
2635 struct sw_flow_actions **sfa,
2636 __be16 eth_type, __be16 vlan_tci,
2637 u32 mpls_label_count, bool log,
2638 u32 depth)
2639 {
2640 const struct nlattr *attrs[OVS_DEC_TTL_ATTR_MAX + 1];
2641 int start, action_start, err, rem;
2642 const struct nlattr *a, *actions;
2643
2644 memset(attrs, 0, sizeof(attrs));
2645 nla_for_each_nested(a, attr, rem) {
2646 int type = nla_type(a);
2647
2648 /* Ignore unknown attributes to be future proof. */
2649 if (type > OVS_DEC_TTL_ATTR_MAX)
2650 continue;
2651
2652 if (!type || attrs[type]) {
2653 OVS_NLERR(log, "Duplicate or invalid key (type %d).",
2654 type);
2655 return -EINVAL;
2656 }
2657
2658 attrs[type] = a;
2659 }
2660
2661 if (rem) {
2662 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
2663 return -EINVAL;
2664 }
2665
2666 actions = attrs[OVS_DEC_TTL_ATTR_ACTION];
2667 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN)) {
2668 OVS_NLERR(log, "Missing valid actions attribute.");
2669 return -EINVAL;
2670 }
2671
2672 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_DEC_TTL, log);
2673 if (start < 0)
2674 return start;
2675
2676 action_start = add_nested_action_start(sfa, OVS_DEC_TTL_ATTR_ACTION, log);
2677 if (action_start < 0)
2678 return action_start;
2679
2680 err = __ovs_nla_copy_actions(net, actions, key, sfa, eth_type,
2681 vlan_tci, mpls_label_count, log,
2682 depth + 1);
2683 if (err)
2684 return err;
2685
2686 add_nested_action_end(*sfa, action_start);
2687 add_nested_action_end(*sfa, start);
2688 return 0;
2689 }
2690
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,u32 depth)2691 static int validate_and_copy_clone(struct net *net,
2692 const struct nlattr *attr,
2693 const struct sw_flow_key *key,
2694 struct sw_flow_actions **sfa,
2695 __be16 eth_type, __be16 vlan_tci,
2696 u32 mpls_label_count, bool log, bool last,
2697 u32 depth)
2698 {
2699 int start, err;
2700 u32 exec;
2701
2702 if (nla_len(attr) && nla_len(attr) < NLA_HDRLEN)
2703 return -EINVAL;
2704
2705 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CLONE, log);
2706 if (start < 0)
2707 return start;
2708
2709 exec = last || !actions_may_change_flow(attr);
2710
2711 err = ovs_nla_add_action(sfa, OVS_CLONE_ATTR_EXEC, &exec,
2712 sizeof(exec), log);
2713 if (err)
2714 return err;
2715
2716 err = __ovs_nla_copy_actions(net, attr, key, sfa,
2717 eth_type, vlan_tci, mpls_label_count, log,
2718 depth + 1);
2719 if (err)
2720 return err;
2721
2722 add_nested_action_end(*sfa, start);
2723
2724 return 0;
2725 }
2726
ovs_match_init(struct sw_flow_match * match,struct sw_flow_key * key,bool reset_key,struct sw_flow_mask * mask)2727 void ovs_match_init(struct sw_flow_match *match,
2728 struct sw_flow_key *key,
2729 bool reset_key,
2730 struct sw_flow_mask *mask)
2731 {
2732 memset(match, 0, sizeof(*match));
2733 match->key = key;
2734 match->mask = mask;
2735
2736 if (reset_key)
2737 memset(key, 0, sizeof(*key));
2738
2739 if (mask) {
2740 memset(&mask->key, 0, sizeof(mask->key));
2741 mask->range.start = mask->range.end = 0;
2742 }
2743 }
2744
validate_geneve_opts(struct sw_flow_key * key)2745 static int validate_geneve_opts(struct sw_flow_key *key)
2746 {
2747 struct geneve_opt *option;
2748 int opts_len = key->tun_opts_len;
2749 bool crit_opt = false;
2750
2751 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2752 while (opts_len > 0) {
2753 int len;
2754
2755 if (opts_len < sizeof(*option))
2756 return -EINVAL;
2757
2758 len = sizeof(*option) + option->length * 4;
2759 if (len > opts_len)
2760 return -EINVAL;
2761
2762 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2763
2764 option = (struct geneve_opt *)((u8 *)option + len);
2765 opts_len -= len;
2766 }
2767
2768 if (crit_opt)
2769 __set_bit(IP_TUNNEL_CRIT_OPT_BIT, key->tun_key.tun_flags);
2770
2771 return 0;
2772 }
2773
validate_and_copy_set_tun(const struct nlattr * attr,struct sw_flow_actions ** sfa,bool log)2774 static int validate_and_copy_set_tun(const struct nlattr *attr,
2775 struct sw_flow_actions **sfa, bool log)
2776 {
2777 IP_TUNNEL_DECLARE_FLAGS(dst_opt_type) = { };
2778 struct sw_flow_match match;
2779 struct sw_flow_key key;
2780 struct metadata_dst *tun_dst;
2781 struct ip_tunnel_info *tun_info;
2782 struct ovs_tunnel_info *ovs_tun;
2783 struct nlattr *a;
2784 int err = 0, start, opts_type;
2785
2786 ovs_match_init(&match, &key, true, NULL);
2787 opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2788 if (opts_type < 0)
2789 return opts_type;
2790
2791 if (key.tun_opts_len) {
2792 switch (opts_type) {
2793 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2794 err = validate_geneve_opts(&key);
2795 if (err < 0)
2796 return err;
2797
2798 __set_bit(IP_TUNNEL_GENEVE_OPT_BIT, dst_opt_type);
2799 break;
2800 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
2801 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT, dst_opt_type);
2802 break;
2803 case OVS_TUNNEL_KEY_ATTR_ERSPAN_OPTS:
2804 __set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, dst_opt_type);
2805 break;
2806 }
2807 }
2808
2809 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2810 if (start < 0)
2811 return start;
2812
2813 tun_dst = metadata_dst_alloc(key.tun_opts_len, METADATA_IP_TUNNEL,
2814 GFP_KERNEL);
2815
2816 if (!tun_dst)
2817 return -ENOMEM;
2818
2819 err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2820 if (err) {
2821 dst_release((struct dst_entry *)tun_dst);
2822 return err;
2823 }
2824
2825 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2826 sizeof(*ovs_tun), log);
2827 if (IS_ERR(a)) {
2828 dst_release((struct dst_entry *)tun_dst);
2829 return PTR_ERR(a);
2830 }
2831
2832 ovs_tun = nla_data(a);
2833 ovs_tun->tun_dst = tun_dst;
2834
2835 tun_info = &tun_dst->u.tun_info;
2836 tun_info->mode = IP_TUNNEL_INFO_TX;
2837 if (key.tun_proto == AF_INET6)
2838 tun_info->mode |= IP_TUNNEL_INFO_IPV6;
2839 else if (key.tun_proto == AF_INET && key.tun_key.u.ipv4.dst == 0)
2840 tun_info->mode |= IP_TUNNEL_INFO_BRIDGE;
2841 tun_info->key = key.tun_key;
2842
2843 /* We need to store the options in the action itself since
2844 * everything else will go away after flow setup. We can append
2845 * it to tun_info and then point there.
2846 */
2847 ip_tunnel_info_opts_set(tun_info,
2848 TUN_METADATA_OPTS(&key, key.tun_opts_len),
2849 key.tun_opts_len, dst_opt_type);
2850 add_nested_action_end(*sfa, start);
2851
2852 return err;
2853 }
2854
validate_nsh(const struct nlattr * attr,bool is_mask,bool is_push_nsh,bool log)2855 static bool validate_nsh(const struct nlattr *attr, bool is_mask,
2856 bool is_push_nsh, bool log)
2857 {
2858 struct sw_flow_match match;
2859 struct sw_flow_key key;
2860 int ret = 0;
2861
2862 ovs_match_init(&match, &key, true, NULL);
2863 ret = nsh_key_put_from_nlattr(attr, &match, is_mask,
2864 is_push_nsh, log);
2865 return !ret;
2866 }
2867
2868 /* Return false if there are any non-masked bits set.
2869 * Mask follows data immediately, before any netlink padding.
2870 */
validate_masked(u8 * data,int len)2871 static bool validate_masked(u8 *data, int len)
2872 {
2873 u8 *mask = data + len;
2874
2875 while (len--)
2876 if (*data++ & ~*mask++)
2877 return false;
2878
2879 return true;
2880 }
2881
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)2882 static int validate_set(const struct nlattr *a,
2883 const struct sw_flow_key *flow_key,
2884 struct sw_flow_actions **sfa, bool *skip_copy,
2885 u8 mac_proto, __be16 eth_type, bool masked, bool log)
2886 {
2887 const struct nlattr *ovs_key = nla_data(a);
2888 int key_type = nla_type(ovs_key);
2889 size_t key_len;
2890
2891 /* There can be only one key in a action */
2892 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2893 return -EINVAL;
2894
2895 key_len = nla_len(ovs_key);
2896 if (masked)
2897 key_len /= 2;
2898
2899 if (key_type > OVS_KEY_ATTR_MAX ||
2900 !check_attr_len(key_len, ovs_key_lens[key_type].len))
2901 return -EINVAL;
2902
2903 if (masked && !validate_masked(nla_data(ovs_key), key_len))
2904 return -EINVAL;
2905
2906 switch (key_type) {
2907 case OVS_KEY_ATTR_PRIORITY:
2908 case OVS_KEY_ATTR_SKB_MARK:
2909 case OVS_KEY_ATTR_CT_MARK:
2910 case OVS_KEY_ATTR_CT_LABELS:
2911 break;
2912
2913 case OVS_KEY_ATTR_ETHERNET:
2914 if (mac_proto != MAC_PROTO_ETHERNET)
2915 return -EINVAL;
2916 break;
2917
2918 case OVS_KEY_ATTR_TUNNEL: {
2919 int err;
2920
2921 if (masked)
2922 return -EINVAL; /* Masked tunnel set not supported. */
2923
2924 *skip_copy = true;
2925 err = validate_and_copy_set_tun(a, sfa, log);
2926 if (err)
2927 return err;
2928 break;
2929 }
2930 case OVS_KEY_ATTR_IPV4: {
2931 const struct ovs_key_ipv4 *ipv4_key;
2932
2933 if (eth_type != htons(ETH_P_IP))
2934 return -EINVAL;
2935
2936 ipv4_key = nla_data(ovs_key);
2937
2938 if (masked) {
2939 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2940
2941 /* Non-writeable fields. */
2942 if (mask->ipv4_proto || mask->ipv4_frag)
2943 return -EINVAL;
2944 } else {
2945 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2946 return -EINVAL;
2947
2948 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2949 return -EINVAL;
2950 }
2951 break;
2952 }
2953 case OVS_KEY_ATTR_IPV6: {
2954 const struct ovs_key_ipv6 *ipv6_key;
2955
2956 if (eth_type != htons(ETH_P_IPV6))
2957 return -EINVAL;
2958
2959 ipv6_key = nla_data(ovs_key);
2960
2961 if (masked) {
2962 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2963
2964 /* Non-writeable fields. */
2965 if (mask->ipv6_proto || mask->ipv6_frag)
2966 return -EINVAL;
2967
2968 /* Invalid bits in the flow label mask? */
2969 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2970 return -EINVAL;
2971 } else {
2972 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2973 return -EINVAL;
2974
2975 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2976 return -EINVAL;
2977 }
2978 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2979 return -EINVAL;
2980
2981 break;
2982 }
2983 case OVS_KEY_ATTR_TCP:
2984 if ((eth_type != htons(ETH_P_IP) &&
2985 eth_type != htons(ETH_P_IPV6)) ||
2986 flow_key->ip.proto != IPPROTO_TCP)
2987 return -EINVAL;
2988
2989 break;
2990
2991 case OVS_KEY_ATTR_UDP:
2992 if ((eth_type != htons(ETH_P_IP) &&
2993 eth_type != htons(ETH_P_IPV6)) ||
2994 flow_key->ip.proto != IPPROTO_UDP)
2995 return -EINVAL;
2996
2997 break;
2998
2999 case OVS_KEY_ATTR_MPLS:
3000 if (!eth_p_mpls(eth_type))
3001 return -EINVAL;
3002 break;
3003
3004 case OVS_KEY_ATTR_SCTP:
3005 if ((eth_type != htons(ETH_P_IP) &&
3006 eth_type != htons(ETH_P_IPV6)) ||
3007 flow_key->ip.proto != IPPROTO_SCTP)
3008 return -EINVAL;
3009
3010 break;
3011
3012 case OVS_KEY_ATTR_NSH:
3013 if (eth_type != htons(ETH_P_NSH))
3014 return -EINVAL;
3015 if (!validate_nsh(nla_data(a), masked, false, log))
3016 return -EINVAL;
3017 break;
3018
3019 default:
3020 return -EINVAL;
3021 }
3022
3023 /* Convert non-masked non-tunnel set actions to masked set actions. */
3024 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
3025 int start, len = key_len * 2;
3026 struct nlattr *at;
3027
3028 *skip_copy = true;
3029
3030 start = add_nested_action_start(sfa,
3031 OVS_ACTION_ATTR_SET_TO_MASKED,
3032 log);
3033 if (start < 0)
3034 return start;
3035
3036 at = __add_action(sfa, key_type, NULL, len, log);
3037 if (IS_ERR(at))
3038 return PTR_ERR(at);
3039
3040 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
3041 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
3042 /* Clear non-writeable bits from otherwise writeable fields. */
3043 if (key_type == OVS_KEY_ATTR_IPV6) {
3044 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
3045
3046 mask->ipv6_label &= htonl(0x000FFFFF);
3047 }
3048 add_nested_action_end(*sfa, start);
3049 }
3050
3051 return 0;
3052 }
3053
validate_userspace(const struct nlattr * attr)3054 static int validate_userspace(const struct nlattr *attr)
3055 {
3056 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
3057 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
3058 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
3059 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
3060 };
3061 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
3062 int error;
3063
3064 error = nla_parse_nested_deprecated(a, OVS_USERSPACE_ATTR_MAX, attr,
3065 userspace_policy, NULL);
3066 if (error)
3067 return error;
3068
3069 if (!a[OVS_USERSPACE_ATTR_PID] ||
3070 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
3071 return -EINVAL;
3072
3073 return 0;
3074 }
3075
3076 static const struct nla_policy cpl_policy[OVS_CHECK_PKT_LEN_ATTR_MAX + 1] = {
3077 [OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] = {.type = NLA_U16 },
3078 [OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER] = {.type = NLA_NESTED },
3079 [OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL] = {.type = NLA_NESTED },
3080 };
3081
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,u32 depth)3082 static int validate_and_copy_check_pkt_len(struct net *net,
3083 const struct nlattr *attr,
3084 const struct sw_flow_key *key,
3085 struct sw_flow_actions **sfa,
3086 __be16 eth_type, __be16 vlan_tci,
3087 u32 mpls_label_count,
3088 bool log, bool last, u32 depth)
3089 {
3090 const struct nlattr *acts_if_greater, *acts_if_lesser_eq;
3091 struct nlattr *a[OVS_CHECK_PKT_LEN_ATTR_MAX + 1];
3092 struct check_pkt_len_arg arg;
3093 int nested_acts_start;
3094 int start, err;
3095
3096 err = nla_parse_deprecated_strict(a, OVS_CHECK_PKT_LEN_ATTR_MAX,
3097 nla_data(attr), nla_len(attr),
3098 cpl_policy, NULL);
3099 if (err)
3100 return err;
3101
3102 if (!a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN] ||
3103 !nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]))
3104 return -EINVAL;
3105
3106 acts_if_lesser_eq = a[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL];
3107 acts_if_greater = a[OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER];
3108
3109 /* Both the nested action should be present. */
3110 if (!acts_if_greater || !acts_if_lesser_eq)
3111 return -EINVAL;
3112
3113 /* validation done, copy the nested actions. */
3114 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_CHECK_PKT_LEN,
3115 log);
3116 if (start < 0)
3117 return start;
3118
3119 arg.pkt_len = nla_get_u16(a[OVS_CHECK_PKT_LEN_ATTR_PKT_LEN]);
3120 arg.exec_for_lesser_equal =
3121 last || !actions_may_change_flow(acts_if_lesser_eq);
3122 arg.exec_for_greater =
3123 last || !actions_may_change_flow(acts_if_greater);
3124
3125 err = ovs_nla_add_action(sfa, OVS_CHECK_PKT_LEN_ATTR_ARG, &arg,
3126 sizeof(arg), log);
3127 if (err)
3128 return err;
3129
3130 nested_acts_start = add_nested_action_start(sfa,
3131 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL, log);
3132 if (nested_acts_start < 0)
3133 return nested_acts_start;
3134
3135 err = __ovs_nla_copy_actions(net, acts_if_lesser_eq, key, sfa,
3136 eth_type, vlan_tci, mpls_label_count, log,
3137 depth + 1);
3138
3139 if (err)
3140 return err;
3141
3142 add_nested_action_end(*sfa, nested_acts_start);
3143
3144 nested_acts_start = add_nested_action_start(sfa,
3145 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER, log);
3146 if (nested_acts_start < 0)
3147 return nested_acts_start;
3148
3149 err = __ovs_nla_copy_actions(net, acts_if_greater, key, sfa,
3150 eth_type, vlan_tci, mpls_label_count, log,
3151 depth + 1);
3152
3153 if (err)
3154 return err;
3155
3156 add_nested_action_end(*sfa, nested_acts_start);
3157 add_nested_action_end(*sfa, start);
3158 return 0;
3159 }
3160
validate_psample(const struct nlattr * attr)3161 static int validate_psample(const struct nlattr *attr)
3162 {
3163 static const struct nla_policy policy[OVS_PSAMPLE_ATTR_MAX + 1] = {
3164 [OVS_PSAMPLE_ATTR_GROUP] = { .type = NLA_U32 },
3165 [OVS_PSAMPLE_ATTR_COOKIE] = {
3166 .type = NLA_BINARY,
3167 .len = OVS_PSAMPLE_COOKIE_MAX_SIZE,
3168 },
3169 };
3170 struct nlattr *a[OVS_PSAMPLE_ATTR_MAX + 1];
3171 int err;
3172
3173 if (!IS_ENABLED(CONFIG_PSAMPLE))
3174 return -EOPNOTSUPP;
3175
3176 err = nla_parse_nested(a, OVS_PSAMPLE_ATTR_MAX, attr, policy, NULL);
3177 if (err)
3178 return err;
3179
3180 return a[OVS_PSAMPLE_ATTR_GROUP] ? 0 : -EINVAL;
3181 }
3182
copy_action(const struct nlattr * from,struct sw_flow_actions ** sfa,bool log)3183 static int copy_action(const struct nlattr *from,
3184 struct sw_flow_actions **sfa, bool log)
3185 {
3186 int totlen = NLA_ALIGN(from->nla_len);
3187 struct nlattr *to;
3188
3189 to = reserve_sfa_size(sfa, from->nla_len, log);
3190 if (IS_ERR(to))
3191 return PTR_ERR(to);
3192
3193 memcpy(to, from, totlen);
3194 return 0;
3195 }
3196
__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,u32 depth)3197 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3198 const struct sw_flow_key *key,
3199 struct sw_flow_actions **sfa,
3200 __be16 eth_type, __be16 vlan_tci,
3201 u32 mpls_label_count, bool log,
3202 u32 depth)
3203 {
3204 u8 mac_proto = ovs_key_mac_proto(key);
3205 const struct nlattr *a;
3206 int rem, err;
3207
3208 if (depth > OVS_COPY_ACTIONS_MAX_DEPTH)
3209 return -EOVERFLOW;
3210
3211 nla_for_each_nested(a, attr, rem) {
3212 /* Expected argument lengths, (u32)-1 for variable length. */
3213 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
3214 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
3215 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
3216 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
3217 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
3218 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
3219 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
3220 [OVS_ACTION_ATTR_POP_VLAN] = 0,
3221 [OVS_ACTION_ATTR_SET] = (u32)-1,
3222 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
3223 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
3224 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
3225 [OVS_ACTION_ATTR_CT] = (u32)-1,
3226 [OVS_ACTION_ATTR_CT_CLEAR] = 0,
3227 [OVS_ACTION_ATTR_TRUNC] = sizeof(struct ovs_action_trunc),
3228 [OVS_ACTION_ATTR_PUSH_ETH] = sizeof(struct ovs_action_push_eth),
3229 [OVS_ACTION_ATTR_POP_ETH] = 0,
3230 [OVS_ACTION_ATTR_PUSH_NSH] = (u32)-1,
3231 [OVS_ACTION_ATTR_POP_NSH] = 0,
3232 [OVS_ACTION_ATTR_METER] = sizeof(u32),
3233 [OVS_ACTION_ATTR_CLONE] = (u32)-1,
3234 [OVS_ACTION_ATTR_CHECK_PKT_LEN] = (u32)-1,
3235 [OVS_ACTION_ATTR_ADD_MPLS] = sizeof(struct ovs_action_add_mpls),
3236 [OVS_ACTION_ATTR_DEC_TTL] = (u32)-1,
3237 [OVS_ACTION_ATTR_DROP] = sizeof(u32),
3238 [OVS_ACTION_ATTR_PSAMPLE] = (u32)-1,
3239 };
3240 const struct ovs_action_push_vlan *vlan;
3241 int type = nla_type(a);
3242 bool skip_copy;
3243
3244 if (type > OVS_ACTION_ATTR_MAX ||
3245 (action_lens[type] != nla_len(a) &&
3246 action_lens[type] != (u32)-1))
3247 return -EINVAL;
3248
3249 skip_copy = false;
3250 switch (type) {
3251 case OVS_ACTION_ATTR_UNSPEC:
3252 return -EINVAL;
3253
3254 case OVS_ACTION_ATTR_USERSPACE:
3255 err = validate_userspace(a);
3256 if (err)
3257 return err;
3258 break;
3259
3260 case OVS_ACTION_ATTR_OUTPUT:
3261 if (nla_get_u32(a) >= DP_MAX_PORTS)
3262 return -EINVAL;
3263 break;
3264
3265 case OVS_ACTION_ATTR_TRUNC: {
3266 const struct ovs_action_trunc *trunc = nla_data(a);
3267
3268 if (trunc->max_len < ETH_HLEN)
3269 return -EINVAL;
3270 break;
3271 }
3272
3273 case OVS_ACTION_ATTR_HASH: {
3274 const struct ovs_action_hash *act_hash = nla_data(a);
3275
3276 switch (act_hash->hash_alg) {
3277 case OVS_HASH_ALG_L4:
3278 fallthrough;
3279 case OVS_HASH_ALG_SYM_L4:
3280 break;
3281 default:
3282 return -EINVAL;
3283 }
3284
3285 break;
3286 }
3287
3288 case OVS_ACTION_ATTR_POP_VLAN:
3289 if (mac_proto != MAC_PROTO_ETHERNET)
3290 return -EINVAL;
3291 vlan_tci = htons(0);
3292 break;
3293
3294 case OVS_ACTION_ATTR_PUSH_VLAN:
3295 if (mac_proto != MAC_PROTO_ETHERNET)
3296 return -EINVAL;
3297 vlan = nla_data(a);
3298 if (!eth_type_vlan(vlan->vlan_tpid))
3299 return -EINVAL;
3300 if (!(vlan->vlan_tci & htons(VLAN_CFI_MASK)))
3301 return -EINVAL;
3302 vlan_tci = vlan->vlan_tci;
3303 break;
3304
3305 case OVS_ACTION_ATTR_RECIRC:
3306 break;
3307
3308 case OVS_ACTION_ATTR_ADD_MPLS: {
3309 const struct ovs_action_add_mpls *mpls = nla_data(a);
3310
3311 if (!eth_p_mpls(mpls->mpls_ethertype))
3312 return -EINVAL;
3313
3314 if (mpls->tun_flags & OVS_MPLS_L3_TUNNEL_FLAG_MASK) {
3315 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3316 (eth_type != htons(ETH_P_IP) &&
3317 eth_type != htons(ETH_P_IPV6) &&
3318 eth_type != htons(ETH_P_ARP) &&
3319 eth_type != htons(ETH_P_RARP) &&
3320 !eth_p_mpls(eth_type)))
3321 return -EINVAL;
3322 mpls_label_count++;
3323 } else {
3324 if (mac_proto == MAC_PROTO_ETHERNET) {
3325 mpls_label_count = 1;
3326 mac_proto = MAC_PROTO_NONE;
3327 } else {
3328 mpls_label_count++;
3329 }
3330 }
3331 eth_type = mpls->mpls_ethertype;
3332 break;
3333 }
3334
3335 case OVS_ACTION_ATTR_PUSH_MPLS: {
3336 const struct ovs_action_push_mpls *mpls = nla_data(a);
3337
3338 if (!eth_p_mpls(mpls->mpls_ethertype))
3339 return -EINVAL;
3340 /* Prohibit push MPLS other than to a white list
3341 * for packets that have a known tag order.
3342 */
3343 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3344 (eth_type != htons(ETH_P_IP) &&
3345 eth_type != htons(ETH_P_IPV6) &&
3346 eth_type != htons(ETH_P_ARP) &&
3347 eth_type != htons(ETH_P_RARP) &&
3348 !eth_p_mpls(eth_type)))
3349 return -EINVAL;
3350 eth_type = mpls->mpls_ethertype;
3351 mpls_label_count++;
3352 break;
3353 }
3354
3355 case OVS_ACTION_ATTR_POP_MPLS: {
3356 __be16 proto;
3357 if (vlan_tci & htons(VLAN_CFI_MASK) ||
3358 !eth_p_mpls(eth_type))
3359 return -EINVAL;
3360
3361 /* Disallow subsequent L2.5+ set actions and mpls_pop
3362 * actions once the last MPLS label in the packet is
3363 * popped as there is no check here to ensure that
3364 * the new eth type is valid and thus set actions could
3365 * write off the end of the packet or otherwise corrupt
3366 * it.
3367 *
3368 * Support for these actions is planned using packet
3369 * recirculation.
3370 */
3371 proto = nla_get_be16(a);
3372
3373 if (proto == htons(ETH_P_TEB) &&
3374 mac_proto != MAC_PROTO_NONE)
3375 return -EINVAL;
3376
3377 mpls_label_count--;
3378
3379 if (!eth_p_mpls(proto) || !mpls_label_count)
3380 eth_type = htons(0);
3381 else
3382 eth_type = proto;
3383
3384 break;
3385 }
3386
3387 case OVS_ACTION_ATTR_SET:
3388 err = validate_set(a, key, sfa,
3389 &skip_copy, mac_proto, eth_type,
3390 false, log);
3391 if (err)
3392 return err;
3393 break;
3394
3395 case OVS_ACTION_ATTR_SET_MASKED:
3396 err = validate_set(a, key, sfa,
3397 &skip_copy, mac_proto, eth_type,
3398 true, log);
3399 if (err)
3400 return err;
3401 break;
3402
3403 case OVS_ACTION_ATTR_SAMPLE: {
3404 bool last = nla_is_last(a, rem);
3405
3406 err = validate_and_copy_sample(net, a, key, sfa,
3407 eth_type, vlan_tci,
3408 mpls_label_count,
3409 log, last, depth);
3410 if (err)
3411 return err;
3412 skip_copy = true;
3413 break;
3414 }
3415
3416 case OVS_ACTION_ATTR_CT:
3417 err = ovs_ct_copy_action(net, a, key, sfa, log);
3418 if (err)
3419 return err;
3420 skip_copy = true;
3421 break;
3422
3423 case OVS_ACTION_ATTR_CT_CLEAR:
3424 break;
3425
3426 case OVS_ACTION_ATTR_PUSH_ETH:
3427 /* Disallow pushing an Ethernet header if one
3428 * is already present */
3429 if (mac_proto != MAC_PROTO_NONE)
3430 return -EINVAL;
3431 mac_proto = MAC_PROTO_ETHERNET;
3432 break;
3433
3434 case OVS_ACTION_ATTR_POP_ETH:
3435 if (mac_proto != MAC_PROTO_ETHERNET)
3436 return -EINVAL;
3437 if (vlan_tci & htons(VLAN_CFI_MASK))
3438 return -EINVAL;
3439 mac_proto = MAC_PROTO_NONE;
3440 break;
3441
3442 case OVS_ACTION_ATTR_PUSH_NSH:
3443 if (mac_proto != MAC_PROTO_ETHERNET) {
3444 u8 next_proto;
3445
3446 next_proto = tun_p_from_eth_p(eth_type);
3447 if (!next_proto)
3448 return -EINVAL;
3449 }
3450 mac_proto = MAC_PROTO_NONE;
3451 if (!validate_nsh(nla_data(a), false, true, true))
3452 return -EINVAL;
3453 break;
3454
3455 case OVS_ACTION_ATTR_POP_NSH: {
3456 __be16 inner_proto;
3457
3458 if (eth_type != htons(ETH_P_NSH))
3459 return -EINVAL;
3460 inner_proto = tun_p_to_eth_p(key->nsh.base.np);
3461 if (!inner_proto)
3462 return -EINVAL;
3463 if (key->nsh.base.np == TUN_P_ETHERNET)
3464 mac_proto = MAC_PROTO_ETHERNET;
3465 else
3466 mac_proto = MAC_PROTO_NONE;
3467 break;
3468 }
3469
3470 case OVS_ACTION_ATTR_METER:
3471 /* Non-existent meters are simply ignored. */
3472 break;
3473
3474 case OVS_ACTION_ATTR_CLONE: {
3475 bool last = nla_is_last(a, rem);
3476
3477 err = validate_and_copy_clone(net, a, key, sfa,
3478 eth_type, vlan_tci,
3479 mpls_label_count,
3480 log, last, depth);
3481 if (err)
3482 return err;
3483 skip_copy = true;
3484 break;
3485 }
3486
3487 case OVS_ACTION_ATTR_CHECK_PKT_LEN: {
3488 bool last = nla_is_last(a, rem);
3489
3490 err = validate_and_copy_check_pkt_len(net, a, key, sfa,
3491 eth_type,
3492 vlan_tci,
3493 mpls_label_count,
3494 log, last,
3495 depth);
3496 if (err)
3497 return err;
3498 skip_copy = true;
3499 break;
3500 }
3501
3502 case OVS_ACTION_ATTR_DEC_TTL:
3503 err = validate_and_copy_dec_ttl(net, a, key, sfa,
3504 eth_type, vlan_tci,
3505 mpls_label_count, log,
3506 depth);
3507 if (err)
3508 return err;
3509 skip_copy = true;
3510 break;
3511
3512 case OVS_ACTION_ATTR_DROP:
3513 if (!nla_is_last(a, rem))
3514 return -EINVAL;
3515 break;
3516
3517 case OVS_ACTION_ATTR_PSAMPLE:
3518 err = validate_psample(a);
3519 if (err)
3520 return err;
3521 break;
3522
3523 default:
3524 OVS_NLERR(log, "Unknown Action type %d", type);
3525 return -EINVAL;
3526 }
3527 if (!skip_copy) {
3528 err = copy_action(a, sfa, log);
3529 if (err)
3530 return err;
3531 }
3532 }
3533
3534 if (rem > 0)
3535 return -EINVAL;
3536
3537 return 0;
3538 }
3539
3540 /* 'key' must be the masked key. */
ovs_nla_copy_actions(struct net * net,const struct nlattr * attr,const struct sw_flow_key * key,struct sw_flow_actions ** sfa,bool log)3541 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
3542 const struct sw_flow_key *key,
3543 struct sw_flow_actions **sfa, bool log)
3544 {
3545 int err;
3546 u32 mpls_label_count = 0;
3547
3548 *sfa = nla_alloc_flow_actions(min(nla_len(attr), MAX_ACTIONS_BUFSIZE));
3549 if (IS_ERR(*sfa))
3550 return PTR_ERR(*sfa);
3551
3552 if (eth_p_mpls(key->eth.type))
3553 mpls_label_count = hweight_long(key->mpls.num_labels_mask);
3554
3555 (*sfa)->orig_len = nla_len(attr);
3556 err = __ovs_nla_copy_actions(net, attr, key, sfa, key->eth.type,
3557 key->eth.vlan.tci, mpls_label_count, log,
3558 0);
3559 if (err)
3560 ovs_nla_free_flow_actions(*sfa);
3561
3562 return err;
3563 }
3564
sample_action_to_attr(const struct nlattr * attr,struct sk_buff * skb)3565 static int sample_action_to_attr(const struct nlattr *attr,
3566 struct sk_buff *skb)
3567 {
3568 struct nlattr *start, *ac_start = NULL, *sample_arg;
3569 int err = 0, rem = nla_len(attr);
3570 const struct sample_arg *arg;
3571 struct nlattr *actions;
3572
3573 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SAMPLE);
3574 if (!start)
3575 return -EMSGSIZE;
3576
3577 sample_arg = nla_data(attr);
3578 arg = nla_data(sample_arg);
3579 actions = nla_next(sample_arg, &rem);
3580
3581 if (nla_put_u32(skb, OVS_SAMPLE_ATTR_PROBABILITY, arg->probability)) {
3582 err = -EMSGSIZE;
3583 goto out;
3584 }
3585
3586 ac_start = nla_nest_start_noflag(skb, OVS_SAMPLE_ATTR_ACTIONS);
3587 if (!ac_start) {
3588 err = -EMSGSIZE;
3589 goto out;
3590 }
3591
3592 err = ovs_nla_put_actions(actions, rem, skb);
3593
3594 out:
3595 if (err) {
3596 nla_nest_cancel(skb, ac_start);
3597 nla_nest_cancel(skb, start);
3598 } else {
3599 nla_nest_end(skb, ac_start);
3600 nla_nest_end(skb, start);
3601 }
3602
3603 return err;
3604 }
3605
clone_action_to_attr(const struct nlattr * attr,struct sk_buff * skb)3606 static int clone_action_to_attr(const struct nlattr *attr,
3607 struct sk_buff *skb)
3608 {
3609 struct nlattr *start;
3610 int err = 0, rem = nla_len(attr);
3611
3612 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CLONE);
3613 if (!start)
3614 return -EMSGSIZE;
3615
3616 /* Skipping the OVS_CLONE_ATTR_EXEC that is always the first attribute. */
3617 attr = nla_next(nla_data(attr), &rem);
3618 err = ovs_nla_put_actions(attr, rem, skb);
3619
3620 if (err)
3621 nla_nest_cancel(skb, start);
3622 else
3623 nla_nest_end(skb, start);
3624
3625 return err;
3626 }
3627
check_pkt_len_action_to_attr(const struct nlattr * attr,struct sk_buff * skb)3628 static int check_pkt_len_action_to_attr(const struct nlattr *attr,
3629 struct sk_buff *skb)
3630 {
3631 struct nlattr *start, *ac_start = NULL;
3632 const struct check_pkt_len_arg *arg;
3633 const struct nlattr *a, *cpl_arg;
3634 int err = 0, rem = nla_len(attr);
3635
3636 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CHECK_PKT_LEN);
3637 if (!start)
3638 return -EMSGSIZE;
3639
3640 /* The first nested attribute in 'attr' is always
3641 * 'OVS_CHECK_PKT_LEN_ATTR_ARG'.
3642 */
3643 cpl_arg = nla_data(attr);
3644 arg = nla_data(cpl_arg);
3645
3646 if (nla_put_u16(skb, OVS_CHECK_PKT_LEN_ATTR_PKT_LEN, arg->pkt_len)) {
3647 err = -EMSGSIZE;
3648 goto out;
3649 }
3650
3651 /* Second nested attribute in 'attr' is always
3652 * 'OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL'.
3653 */
3654 a = nla_next(cpl_arg, &rem);
3655 ac_start = nla_nest_start_noflag(skb,
3656 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL);
3657 if (!ac_start) {
3658 err = -EMSGSIZE;
3659 goto out;
3660 }
3661
3662 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3663 if (err) {
3664 nla_nest_cancel(skb, ac_start);
3665 goto out;
3666 } else {
3667 nla_nest_end(skb, ac_start);
3668 }
3669
3670 /* Third nested attribute in 'attr' is always
3671 * OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER.
3672 */
3673 a = nla_next(a, &rem);
3674 ac_start = nla_nest_start_noflag(skb,
3675 OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER);
3676 if (!ac_start) {
3677 err = -EMSGSIZE;
3678 goto out;
3679 }
3680
3681 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3682 if (err) {
3683 nla_nest_cancel(skb, ac_start);
3684 goto out;
3685 } else {
3686 nla_nest_end(skb, ac_start);
3687 }
3688
3689 nla_nest_end(skb, start);
3690 return 0;
3691
3692 out:
3693 nla_nest_cancel(skb, start);
3694 return err;
3695 }
3696
dec_ttl_action_to_attr(const struct nlattr * attr,struct sk_buff * skb)3697 static int dec_ttl_action_to_attr(const struct nlattr *attr,
3698 struct sk_buff *skb)
3699 {
3700 struct nlattr *start, *action_start;
3701 const struct nlattr *a;
3702 int err = 0, rem;
3703
3704 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_DEC_TTL);
3705 if (!start)
3706 return -EMSGSIZE;
3707
3708 nla_for_each_attr(a, nla_data(attr), nla_len(attr), rem) {
3709 switch (nla_type(a)) {
3710 case OVS_DEC_TTL_ATTR_ACTION:
3711
3712 action_start = nla_nest_start_noflag(skb, OVS_DEC_TTL_ATTR_ACTION);
3713 if (!action_start) {
3714 err = -EMSGSIZE;
3715 goto out;
3716 }
3717
3718 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
3719 if (err)
3720 goto out;
3721
3722 nla_nest_end(skb, action_start);
3723 break;
3724
3725 default:
3726 /* Ignore all other option to be future compatible */
3727 break;
3728 }
3729 }
3730
3731 nla_nest_end(skb, start);
3732 return 0;
3733
3734 out:
3735 nla_nest_cancel(skb, start);
3736 return err;
3737 }
3738
set_action_to_attr(const struct nlattr * a,struct sk_buff * skb)3739 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
3740 {
3741 const struct nlattr *ovs_key = nla_data(a);
3742 int key_type = nla_type(ovs_key);
3743 struct nlattr *start;
3744 int err;
3745
3746 switch (key_type) {
3747 case OVS_KEY_ATTR_TUNNEL_INFO: {
3748 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
3749 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
3750
3751 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3752 if (!start)
3753 return -EMSGSIZE;
3754
3755 err = ip_tun_to_nlattr(skb, &tun_info->key,
3756 ip_tunnel_info_opts(tun_info),
3757 tun_info->options_len,
3758 ip_tunnel_info_af(tun_info), tun_info->mode);
3759 if (err)
3760 return err;
3761 nla_nest_end(skb, start);
3762 break;
3763 }
3764 default:
3765 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
3766 return -EMSGSIZE;
3767 break;
3768 }
3769
3770 return 0;
3771 }
3772
masked_set_action_to_set_action_attr(const struct nlattr * a,struct sk_buff * skb)3773 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
3774 struct sk_buff *skb)
3775 {
3776 const struct nlattr *ovs_key = nla_data(a);
3777 struct nlattr *nla;
3778 size_t key_len = nla_len(ovs_key) / 2;
3779
3780 /* Revert the conversion we did from a non-masked set action to
3781 * masked set action.
3782 */
3783 nla = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_SET);
3784 if (!nla)
3785 return -EMSGSIZE;
3786
3787 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
3788 return -EMSGSIZE;
3789
3790 nla_nest_end(skb, nla);
3791 return 0;
3792 }
3793
ovs_nla_put_actions(const struct nlattr * attr,int len,struct sk_buff * skb)3794 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
3795 {
3796 const struct nlattr *a;
3797 int rem, err;
3798
3799 nla_for_each_attr(a, attr, len, rem) {
3800 int type = nla_type(a);
3801
3802 switch (type) {
3803 case OVS_ACTION_ATTR_SET:
3804 err = set_action_to_attr(a, skb);
3805 if (err)
3806 return err;
3807 break;
3808
3809 case OVS_ACTION_ATTR_SET_TO_MASKED:
3810 err = masked_set_action_to_set_action_attr(a, skb);
3811 if (err)
3812 return err;
3813 break;
3814
3815 case OVS_ACTION_ATTR_SAMPLE:
3816 err = sample_action_to_attr(a, skb);
3817 if (err)
3818 return err;
3819 break;
3820
3821 case OVS_ACTION_ATTR_CT:
3822 err = ovs_ct_action_to_attr(nla_data(a), skb);
3823 if (err)
3824 return err;
3825 break;
3826
3827 case OVS_ACTION_ATTR_CLONE:
3828 err = clone_action_to_attr(a, skb);
3829 if (err)
3830 return err;
3831 break;
3832
3833 case OVS_ACTION_ATTR_CHECK_PKT_LEN:
3834 err = check_pkt_len_action_to_attr(a, skb);
3835 if (err)
3836 return err;
3837 break;
3838
3839 case OVS_ACTION_ATTR_DEC_TTL:
3840 err = dec_ttl_action_to_attr(a, skb);
3841 if (err)
3842 return err;
3843 break;
3844
3845 default:
3846 if (nla_put(skb, type, nla_len(a), nla_data(a)))
3847 return -EMSGSIZE;
3848 break;
3849 }
3850 }
3851
3852 return 0;
3853 }
3854