]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - net/openvswitch/flow_netlink.c
projects / mirror_ubuntu-zesty-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
UBUNTU: Ubuntu-4.10.0-37.41
[mirror_ubuntu-zesty-kernel.git] / net / openvswitch / flow_netlink.c
1 /*
2 * Copyright (c) 2007-2014 Nicira, Inc.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
16 * 02110-1301, USA
17 */
18
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
20
21 #include "flow.h"
22 #include "datapath.h"
23 #include <linux/uaccess.h>
24 #include <linux/netdevice.h>
25 #include <linux/etherdevice.h>
26 #include <linux/if_ether.h>
27 #include <linux/if_vlan.h>
28 #include <net/llc_pdu.h>
29 #include <linux/kernel.h>
30 #include <linux/jhash.h>
31 #include <linux/jiffies.h>
32 #include <linux/llc.h>
33 #include <linux/module.h>
34 #include <linux/in.h>
35 #include <linux/rcupdate.h>
36 #include <linux/if_arp.h>
37 #include <linux/ip.h>
38 #include <linux/ipv6.h>
39 #include <linux/sctp.h>
40 #include <linux/tcp.h>
41 #include <linux/udp.h>
42 #include <linux/icmp.h>
43 #include <linux/icmpv6.h>
44 #include <linux/rculist.h>
45 #include <net/geneve.h>
46 #include <net/ip.h>
47 #include <net/ipv6.h>
48 #include <net/ndisc.h>
49 #include <net/mpls.h>
50 #include <net/vxlan.h>
51
52 #include "flow_netlink.h"
53
54 struct ovs_len_tbl {
55 int len;
56 const struct ovs_len_tbl *next;
57 };
58
59 #define OVS_ATTR_NESTED -1
60 #define OVS_ATTR_VARIABLE -2
61
62 static void update_range(struct sw_flow_match *match,
63 size_t offset, size_t size, bool is_mask)
64 {
65 struct sw_flow_key_range *range;
66 size_t start = rounddown(offset, sizeof(long));
67 size_t end = roundup(offset + size, sizeof(long));
68
69 if (!is_mask)
70 range = &match->range;
71 else
72 range = &match->mask->range;
73
74 if (range->start == range->end) {
75 range->start = start;
76 range->end = end;
77 return;
78 }
79
80 if (range->start > start)
81 range->start = start;
82
83 if (range->end < end)
84 range->end = end;
85 }
86
87 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
88 do { \
89 update_range(match, offsetof(struct sw_flow_key, field), \
90 sizeof((match)->key->field), is_mask); \
91 if (is_mask) \
92 (match)->mask->key.field = value; \
93 else \
94 (match)->key->field = value; \
95 } while (0)
96
97 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask) \
98 do { \
99 update_range(match, offset, len, is_mask); \
100 if (is_mask) \
101 memcpy((u8 *)&(match)->mask->key + offset, value_p, \
102 len); \
103 else \
104 memcpy((u8 *)(match)->key + offset, value_p, len); \
105 } while (0)
106
107 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
108 SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
109 value_p, len, is_mask)
110
111 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
112 do { \
113 update_range(match, offsetof(struct sw_flow_key, field), \
114 sizeof((match)->key->field), is_mask); \
115 if (is_mask) \
116 memset((u8 *)&(match)->mask->key.field, value, \
117 sizeof((match)->mask->key.field)); \
118 else \
119 memset((u8 *)&(match)->key->field, value, \
120 sizeof((match)->key->field)); \
121 } while (0)
122
123 static bool match_validate(const struct sw_flow_match *match,
124 u64 key_attrs, u64 mask_attrs, bool log)
125 {
126 u64 key_expected = 0;
127 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
128
129 /* The following mask attributes allowed only if they
130 * pass the validation tests. */
131 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
132 | (1 << OVS_KEY_ATTR_IPV6)
133 | (1 << OVS_KEY_ATTR_TCP)
134 | (1 << OVS_KEY_ATTR_TCP_FLAGS)
135 | (1 << OVS_KEY_ATTR_UDP)
136 | (1 << OVS_KEY_ATTR_SCTP)
137 | (1 << OVS_KEY_ATTR_ICMP)
138 | (1 << OVS_KEY_ATTR_ICMPV6)
139 | (1 << OVS_KEY_ATTR_ARP)
140 | (1 << OVS_KEY_ATTR_ND)
141 | (1 << OVS_KEY_ATTR_MPLS));
142
143 /* Always allowed mask fields. */
144 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
145 | (1 << OVS_KEY_ATTR_IN_PORT)
146 | (1 << OVS_KEY_ATTR_ETHERTYPE));
147
148 /* Check key attributes. */
149 if (match->key->eth.type == htons(ETH_P_ARP)
150 || match->key->eth.type == htons(ETH_P_RARP)) {
151 key_expected |= 1 << OVS_KEY_ATTR_ARP;
152 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
153 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
154 }
155
156 if (eth_p_mpls(match->key->eth.type)) {
157 key_expected |= 1 << OVS_KEY_ATTR_MPLS;
158 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
159 mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
160 }
161
162 if (match->key->eth.type == htons(ETH_P_IP)) {
163 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
164 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
165 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
166
167 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
168 if (match->key->ip.proto == IPPROTO_UDP) {
169 key_expected |= 1 << OVS_KEY_ATTR_UDP;
170 if (match->mask && (match->mask->key.ip.proto == 0xff))
171 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
172 }
173
174 if (match->key->ip.proto == IPPROTO_SCTP) {
175 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
176 if (match->mask && (match->mask->key.ip.proto == 0xff))
177 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
178 }
179
180 if (match->key->ip.proto == IPPROTO_TCP) {
181 key_expected |= 1 << OVS_KEY_ATTR_TCP;
182 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
183 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
184 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
185 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
186 }
187 }
188
189 if (match->key->ip.proto == IPPROTO_ICMP) {
190 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
191 if (match->mask && (match->mask->key.ip.proto == 0xff))
192 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
193 }
194 }
195 }
196
197 if (match->key->eth.type == htons(ETH_P_IPV6)) {
198 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
199 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
200 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
201
202 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
203 if (match->key->ip.proto == IPPROTO_UDP) {
204 key_expected |= 1 << OVS_KEY_ATTR_UDP;
205 if (match->mask && (match->mask->key.ip.proto == 0xff))
206 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
207 }
208
209 if (match->key->ip.proto == IPPROTO_SCTP) {
210 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
211 if (match->mask && (match->mask->key.ip.proto == 0xff))
212 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
213 }
214
215 if (match->key->ip.proto == IPPROTO_TCP) {
216 key_expected |= 1 << OVS_KEY_ATTR_TCP;
217 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
218 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
219 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
220 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
221 }
222 }
223
224 if (match->key->ip.proto == IPPROTO_ICMPV6) {
225 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
226 if (match->mask && (match->mask->key.ip.proto == 0xff))
227 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
228
229 if (match->key->tp.src ==
230 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
231 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
232 key_expected |= 1 << OVS_KEY_ATTR_ND;
233 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
234 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
235 }
236 }
237 }
238 }
239
240 if ((key_attrs & key_expected) != key_expected) {
241 /* Key attributes check failed. */
242 OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
243 (unsigned long long)key_attrs,
244 (unsigned long long)key_expected);
245 return false;
246 }
247
248 if ((mask_attrs & mask_allowed) != mask_attrs) {
249 /* Mask attributes check failed. */
250 OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
251 (unsigned long long)mask_attrs,
252 (unsigned long long)mask_allowed);
253 return false;
254 }
255
256 return true;
257 }
258
259 size_t ovs_tun_key_attr_size(void)
260 {
261 /* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
262 * updating this function.
263 */
264 return nla_total_size_64bit(8) /* OVS_TUNNEL_KEY_ATTR_ID */
265 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
266 + nla_total_size(16) /* OVS_TUNNEL_KEY_ATTR_IPV[46]_DST */
267 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TOS */
268 + nla_total_size(1) /* OVS_TUNNEL_KEY_ATTR_TTL */
269 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
270 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
271 + nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
272 + nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
273 /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS is mutually exclusive with
274 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
275 */
276 + nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
277 + nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
278 }
279
280 size_t ovs_key_attr_size(void)
281 {
282 /* Whenever adding new OVS_KEY_ FIELDS, we should consider
283 * updating this function.
284 */
285 BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO != 26);
286
287 return nla_total_size(4) /* OVS_KEY_ATTR_PRIORITY */
288 + nla_total_size(0) /* OVS_KEY_ATTR_TUNNEL */
289 + ovs_tun_key_attr_size()
290 + nla_total_size(4) /* OVS_KEY_ATTR_IN_PORT */
291 + nla_total_size(4) /* OVS_KEY_ATTR_SKB_MARK */
292 + nla_total_size(4) /* OVS_KEY_ATTR_DP_HASH */
293 + nla_total_size(4) /* OVS_KEY_ATTR_RECIRC_ID */
294 + nla_total_size(4) /* OVS_KEY_ATTR_CT_STATE */
295 + nla_total_size(2) /* OVS_KEY_ATTR_CT_ZONE */
296 + nla_total_size(4) /* OVS_KEY_ATTR_CT_MARK */
297 + nla_total_size(16) /* OVS_KEY_ATTR_CT_LABELS */
298 + nla_total_size(12) /* OVS_KEY_ATTR_ETHERNET */
299 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
300 + nla_total_size(4) /* OVS_KEY_ATTR_VLAN */
301 + nla_total_size(0) /* OVS_KEY_ATTR_ENCAP */
302 + nla_total_size(2) /* OVS_KEY_ATTR_ETHERTYPE */
303 + nla_total_size(40) /* OVS_KEY_ATTR_IPV6 */
304 + nla_total_size(2) /* OVS_KEY_ATTR_ICMPV6 */
305 + nla_total_size(28); /* OVS_KEY_ATTR_ND */
306 }
307
308 static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
309 [OVS_VXLAN_EXT_GBP] = { .len = sizeof(u32) },
310 };
311
312 static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
313 [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
314 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
315 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
316 [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
317 [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
318 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
319 [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
320 [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
321 [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
322 [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
323 [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_VARIABLE },
324 [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED,
325 .next = ovs_vxlan_ext_key_lens },
326 [OVS_TUNNEL_KEY_ATTR_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
327 [OVS_TUNNEL_KEY_ATTR_IPV6_DST] = { .len = sizeof(struct in6_addr) },
328 };
329
330 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
331 static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
332 [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
333 [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
334 [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
335 [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
336 [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
337 [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
338 [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
339 [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
340 [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
341 [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
342 [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
343 [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
344 [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
345 [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
346 [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
347 [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
348 [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
349 [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
350 [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
351 [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
352 .next = ovs_tunnel_key_lens, },
353 [OVS_KEY_ATTR_MPLS] = { .len = sizeof(struct ovs_key_mpls) },
354 [OVS_KEY_ATTR_CT_STATE] = { .len = sizeof(u32) },
355 [OVS_KEY_ATTR_CT_ZONE] = { .len = sizeof(u16) },
356 [OVS_KEY_ATTR_CT_MARK] = { .len = sizeof(u32) },
357 [OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
358 };
359
360 static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
361 {
362 return expected_len == attr_len ||
363 expected_len == OVS_ATTR_NESTED ||
364 expected_len == OVS_ATTR_VARIABLE;
365 }
366
367 static bool is_all_zero(const u8 *fp, size_t size)
368 {
369 int i;
370
371 if (!fp)
372 return false;
373
374 for (i = 0; i < size; i++)
375 if (fp[i])
376 return false;
377
378 return true;
379 }
380
381 static int __parse_flow_nlattrs(const struct nlattr *attr,
382 const struct nlattr *a[],
383 u64 *attrsp, bool log, bool nz)
384 {
385 const struct nlattr *nla;
386 u64 attrs;
387 int rem;
388
389 attrs = *attrsp;
390 nla_for_each_nested(nla, attr, rem) {
391 u16 type = nla_type(nla);
392 int expected_len;
393
394 if (type > OVS_KEY_ATTR_MAX) {
395 OVS_NLERR(log, "Key type %d is out of range max %d",
396 type, OVS_KEY_ATTR_MAX);
397 return -EINVAL;
398 }
399
400 if (attrs & (1 << type)) {
401 OVS_NLERR(log, "Duplicate key (type %d).", type);
402 return -EINVAL;
403 }
404
405 expected_len = ovs_key_lens[type].len;
406 if (!check_attr_len(nla_len(nla), expected_len)) {
407 OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
408 type, nla_len(nla), expected_len);
409 return -EINVAL;
410 }
411
412 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
413 attrs |= 1 << type;
414 a[type] = nla;
415 }
416 }
417 if (rem) {
418 OVS_NLERR(log, "Message has %d unknown bytes.", rem);
419 return -EINVAL;
420 }
421
422 *attrsp = attrs;
423 return 0;
424 }
425
426 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
427 const struct nlattr *a[], u64 *attrsp,
428 bool log)
429 {
430 return __parse_flow_nlattrs(attr, a, attrsp, log, true);
431 }
432
433 static int parse_flow_nlattrs(const struct nlattr *attr,
434 const struct nlattr *a[], u64 *attrsp,
435 bool log)
436 {
437 return __parse_flow_nlattrs(attr, a, attrsp, log, false);
438 }
439
440 static int genev_tun_opt_from_nlattr(const struct nlattr *a,
441 struct sw_flow_match *match, bool is_mask,
442 bool log)
443 {
444 unsigned long opt_key_offset;
445
446 if (nla_len(a) > sizeof(match->key->tun_opts)) {
447 OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
448 nla_len(a), sizeof(match->key->tun_opts));
449 return -EINVAL;
450 }
451
452 if (nla_len(a) % 4 != 0) {
453 OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
454 nla_len(a));
455 return -EINVAL;
456 }
457
458 /* We need to record the length of the options passed
459 * down, otherwise packets with the same format but
460 * additional options will be silently matched.
461 */
462 if (!is_mask) {
463 SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
464 false);
465 } else {
466 /* This is somewhat unusual because it looks at
467 * both the key and mask while parsing the
468 * attributes (and by extension assumes the key
469 * is parsed first). Normally, we would verify
470 * that each is the correct length and that the
471 * attributes line up in the validate function.
472 * However, that is difficult because this is
473 * variable length and we won't have the
474 * information later.
475 */
476 if (match->key->tun_opts_len != nla_len(a)) {
477 OVS_NLERR(log, "Geneve option len %d != mask len %d",
478 match->key->tun_opts_len, nla_len(a));
479 return -EINVAL;
480 }
481
482 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
483 }
484
485 opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
486 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
487 nla_len(a), is_mask);
488 return 0;
489 }
490
491 static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
492 struct sw_flow_match *match, bool is_mask,
493 bool log)
494 {
495 struct nlattr *a;
496 int rem;
497 unsigned long opt_key_offset;
498 struct vxlan_metadata opts;
499
500 BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
501
502 memset(&opts, 0, sizeof(opts));
503 nla_for_each_nested(a, attr, rem) {
504 int type = nla_type(a);
505
506 if (type > OVS_VXLAN_EXT_MAX) {
507 OVS_NLERR(log, "VXLAN extension %d out of range max %d",
508 type, OVS_VXLAN_EXT_MAX);
509 return -EINVAL;
510 }
511
512 if (!check_attr_len(nla_len(a),
513 ovs_vxlan_ext_key_lens[type].len)) {
514 OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
515 type, nla_len(a),
516 ovs_vxlan_ext_key_lens[type].len);
517 return -EINVAL;
518 }
519
520 switch (type) {
521 case OVS_VXLAN_EXT_GBP:
522 opts.gbp = nla_get_u32(a);
523 break;
524 default:
525 OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
526 type);
527 return -EINVAL;
528 }
529 }
530 if (rem) {
531 OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
532 rem);
533 return -EINVAL;
534 }
535
536 if (!is_mask)
537 SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
538 else
539 SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
540
541 opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
542 SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
543 is_mask);
544 return 0;
545 }
546
547 static int ip_tun_from_nlattr(const struct nlattr *attr,
548 struct sw_flow_match *match, bool is_mask,
549 bool log)
550 {
551 bool ttl = false, ipv4 = false, ipv6 = false;
552 __be16 tun_flags = 0;
553 int opts_type = 0;
554 struct nlattr *a;
555 int rem;
556
557 nla_for_each_nested(a, attr, rem) {
558 int type = nla_type(a);
559 int err;
560
561 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
562 OVS_NLERR(log, "Tunnel attr %d out of range max %d",
563 type, OVS_TUNNEL_KEY_ATTR_MAX);
564 return -EINVAL;
565 }
566
567 if (!check_attr_len(nla_len(a),
568 ovs_tunnel_key_lens[type].len)) {
569 OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
570 type, nla_len(a), ovs_tunnel_key_lens[type].len);
571 return -EINVAL;
572 }
573
574 switch (type) {
575 case OVS_TUNNEL_KEY_ATTR_ID:
576 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
577 nla_get_be64(a), is_mask);
578 tun_flags |= TUNNEL_KEY;
579 break;
580 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
581 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
582 nla_get_in_addr(a), is_mask);
583 ipv4 = true;
584 break;
585 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
586 SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
587 nla_get_in_addr(a), is_mask);
588 ipv4 = true;
589 break;
590 case OVS_TUNNEL_KEY_ATTR_IPV6_SRC:
591 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.src,
592 nla_get_in6_addr(a), is_mask);
593 ipv6 = true;
594 break;
595 case OVS_TUNNEL_KEY_ATTR_IPV6_DST:
596 SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
597 nla_get_in6_addr(a), is_mask);
598 ipv6 = true;
599 break;
600 case OVS_TUNNEL_KEY_ATTR_TOS:
601 SW_FLOW_KEY_PUT(match, tun_key.tos,
602 nla_get_u8(a), is_mask);
603 break;
604 case OVS_TUNNEL_KEY_ATTR_TTL:
605 SW_FLOW_KEY_PUT(match, tun_key.ttl,
606 nla_get_u8(a), is_mask);
607 ttl = true;
608 break;
609 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
610 tun_flags |= TUNNEL_DONT_FRAGMENT;
611 break;
612 case OVS_TUNNEL_KEY_ATTR_CSUM:
613 tun_flags |= TUNNEL_CSUM;
614 break;
615 case OVS_TUNNEL_KEY_ATTR_TP_SRC:
616 SW_FLOW_KEY_PUT(match, tun_key.tp_src,
617 nla_get_be16(a), is_mask);
618 break;
619 case OVS_TUNNEL_KEY_ATTR_TP_DST:
620 SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
621 nla_get_be16(a), is_mask);
622 break;
623 case OVS_TUNNEL_KEY_ATTR_OAM:
624 tun_flags |= TUNNEL_OAM;
625 break;
626 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
627 if (opts_type) {
628 OVS_NLERR(log, "Multiple metadata blocks provided");
629 return -EINVAL;
630 }
631
632 err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
633 if (err)
634 return err;
635
636 tun_flags |= TUNNEL_GENEVE_OPT;
637 opts_type = type;
638 break;
639 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
640 if (opts_type) {
641 OVS_NLERR(log, "Multiple metadata blocks provided");
642 return -EINVAL;
643 }
644
645 err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
646 if (err)
647 return err;
648
649 tun_flags |= TUNNEL_VXLAN_OPT;
650 opts_type = type;
651 break;
652 case OVS_TUNNEL_KEY_ATTR_PAD:
653 break;
654 default:
655 OVS_NLERR(log, "Unknown IP tunnel attribute %d",
656 type);
657 return -EINVAL;
658 }
659 }
660
661 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
662 if (is_mask)
663 SW_FLOW_KEY_MEMSET_FIELD(match, tun_proto, 0xff, true);
664 else
665 SW_FLOW_KEY_PUT(match, tun_proto, ipv6 ? AF_INET6 : AF_INET,
666 false);
667
668 if (rem > 0) {
669 OVS_NLERR(log, "IP tunnel attribute has %d unknown bytes.",
670 rem);
671 return -EINVAL;
672 }
673
674 if (ipv4 && ipv6) {
675 OVS_NLERR(log, "Mixed IPv4 and IPv6 tunnel attributes");
676 return -EINVAL;
677 }
678
679 if (!is_mask) {
680 if (!ipv4 && !ipv6) {
681 OVS_NLERR(log, "IP tunnel dst address not specified");
682 return -EINVAL;
683 }
684 if (ipv4 && !match->key->tun_key.u.ipv4.dst) {
685 OVS_NLERR(log, "IPv4 tunnel dst address is zero");
686 return -EINVAL;
687 }
688 if (ipv6 && ipv6_addr_any(&match->key->tun_key.u.ipv6.dst)) {
689 OVS_NLERR(log, "IPv6 tunnel dst address is zero");
690 return -EINVAL;
691 }
692
693 if (!ttl) {
694 OVS_NLERR(log, "IP tunnel TTL not specified.");
695 return -EINVAL;
696 }
697 }
698
699 return opts_type;
700 }
701
702 static int vxlan_opt_to_nlattr(struct sk_buff *skb,
703 const void *tun_opts, int swkey_tun_opts_len)
704 {
705 const struct vxlan_metadata *opts = tun_opts;
706 struct nlattr *nla;
707
708 nla = nla_nest_start(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
709 if (!nla)
710 return -EMSGSIZE;
711
712 if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
713 return -EMSGSIZE;
714
715 nla_nest_end(skb, nla);
716 return 0;
717 }
718
719 static int __ip_tun_to_nlattr(struct sk_buff *skb,
720 const struct ip_tunnel_key *output,
721 const void *tun_opts, int swkey_tun_opts_len,
722 unsigned short tun_proto)
723 {
724 if (output->tun_flags & TUNNEL_KEY &&
725 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id,
726 OVS_TUNNEL_KEY_ATTR_PAD))
727 return -EMSGSIZE;
728 switch (tun_proto) {
729 case AF_INET:
730 if (output->u.ipv4.src &&
731 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
732 output->u.ipv4.src))
733 return -EMSGSIZE;
734 if (output->u.ipv4.dst &&
735 nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
736 output->u.ipv4.dst))
737 return -EMSGSIZE;
738 break;
739 case AF_INET6:
740 if (!ipv6_addr_any(&output->u.ipv6.src) &&
741 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_SRC,
742 &output->u.ipv6.src))
743 return -EMSGSIZE;
744 if (!ipv6_addr_any(&output->u.ipv6.dst) &&
745 nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_DST,
746 &output->u.ipv6.dst))
747 return -EMSGSIZE;
748 break;
749 }
750 if (output->tos &&
751 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
752 return -EMSGSIZE;
753 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
754 return -EMSGSIZE;
755 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
756 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
757 return -EMSGSIZE;
758 if ((output->tun_flags & TUNNEL_CSUM) &&
759 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
760 return -EMSGSIZE;
761 if (output->tp_src &&
762 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
763 return -EMSGSIZE;
764 if (output->tp_dst &&
765 nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
766 return -EMSGSIZE;
767 if ((output->tun_flags & TUNNEL_OAM) &&
768 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
769 return -EMSGSIZE;
770 if (swkey_tun_opts_len) {
771 if (output->tun_flags & TUNNEL_GENEVE_OPT &&
772 nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
773 swkey_tun_opts_len, tun_opts))
774 return -EMSGSIZE;
775 else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
776 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
777 return -EMSGSIZE;
778 }
779
780 return 0;
781 }
782
783 static int ip_tun_to_nlattr(struct sk_buff *skb,
784 const struct ip_tunnel_key *output,
785 const void *tun_opts, int swkey_tun_opts_len,
786 unsigned short tun_proto)
787 {
788 struct nlattr *nla;
789 int err;
790
791 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
792 if (!nla)
793 return -EMSGSIZE;
794
795 err = __ip_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len,
796 tun_proto);
797 if (err)
798 return err;
799
800 nla_nest_end(skb, nla);
801 return 0;
802 }
803
804 int ovs_nla_put_tunnel_info(struct sk_buff *skb,
805 struct ip_tunnel_info *tun_info)
806 {
807 return __ip_tun_to_nlattr(skb, &tun_info->key,
808 ip_tunnel_info_opts(tun_info),
809 tun_info->options_len,
810 ip_tunnel_info_af(tun_info));
811 }
812
813 static int encode_vlan_from_nlattrs(struct sw_flow_match *match,
814 const struct nlattr *a[],
815 bool is_mask, bool inner)
816 {
817 __be16 tci = 0;
818 __be16 tpid = 0;
819
820 if (a[OVS_KEY_ATTR_VLAN])
821 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
822
823 if (a[OVS_KEY_ATTR_ETHERTYPE])
824 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
825
826 if (likely(!inner)) {
827 SW_FLOW_KEY_PUT(match, eth.vlan.tpid, tpid, is_mask);
828 SW_FLOW_KEY_PUT(match, eth.vlan.tci, tci, is_mask);
829 } else {
830 SW_FLOW_KEY_PUT(match, eth.cvlan.tpid, tpid, is_mask);
831 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, tci, is_mask);
832 }
833 return 0;
834 }
835
836 static int validate_vlan_from_nlattrs(const struct sw_flow_match *match,
837 u64 key_attrs, bool inner,
838 const struct nlattr **a, bool log)
839 {
840 __be16 tci = 0;
841
842 if (!((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
843 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
844 eth_type_vlan(nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE])))) {
845 /* Not a VLAN. */
846 return 0;
847 }
848
849 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
850 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
851 OVS_NLERR(log, "Invalid %s frame", (inner) ? "C-VLAN" : "VLAN");
852 return -EINVAL;
853 }
854
855 if (a[OVS_KEY_ATTR_VLAN])
856 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
857
858 if (!(tci & htons(VLAN_TAG_PRESENT))) {
859 if (tci) {
860 OVS_NLERR(log, "%s TCI does not have VLAN_TAG_PRESENT bit set.",
861 (inner) ? "C-VLAN" : "VLAN");
862 return -EINVAL;
863 } else if (nla_len(a[OVS_KEY_ATTR_ENCAP])) {
864 /* Corner case for truncated VLAN header. */
865 OVS_NLERR(log, "Truncated %s header has non-zero encap attribute.",
866 (inner) ? "C-VLAN" : "VLAN");
867 return -EINVAL;
868 }
869 }
870
871 return 1;
872 }
873
874 static int validate_vlan_mask_from_nlattrs(const struct sw_flow_match *match,
875 u64 key_attrs, bool inner,
876 const struct nlattr **a, bool log)
877 {
878 __be16 tci = 0;
879 __be16 tpid = 0;
880 bool encap_valid = !!(match->key->eth.vlan.tci &
881 htons(VLAN_TAG_PRESENT));
882 bool i_encap_valid = !!(match->key->eth.cvlan.tci &
883 htons(VLAN_TAG_PRESENT));
884
885 if (!(key_attrs & (1 << OVS_KEY_ATTR_ENCAP))) {
886 /* Not a VLAN. */
887 return 0;
888 }
889
890 if ((!inner && !encap_valid) || (inner && !i_encap_valid)) {
891 OVS_NLERR(log, "Encap mask attribute is set for non-%s frame.",
892 (inner) ? "C-VLAN" : "VLAN");
893 return -EINVAL;
894 }
895
896 if (a[OVS_KEY_ATTR_VLAN])
897 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
898
899 if (a[OVS_KEY_ATTR_ETHERTYPE])
900 tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
901
902 if (tpid != htons(0xffff)) {
903 OVS_NLERR(log, "Must have an exact match on %s TPID (mask=%x).",
904 (inner) ? "C-VLAN" : "VLAN", ntohs(tpid));
905 return -EINVAL;
906 }
907 if (!(tci & htons(VLAN_TAG_PRESENT))) {
908 OVS_NLERR(log, "%s TCI mask does not have exact match for VLAN_TAG_PRESENT bit.",
909 (inner) ? "C-VLAN" : "VLAN");
910 return -EINVAL;
911 }
912
913 return 1;
914 }
915
916 static int __parse_vlan_from_nlattrs(struct sw_flow_match *match,
917 u64 *key_attrs, bool inner,
918 const struct nlattr **a, bool is_mask,
919 bool log)
920 {
921 int err;
922 const struct nlattr *encap;
923
924 if (!is_mask)
925 err = validate_vlan_from_nlattrs(match, *key_attrs, inner,
926 a, log);
927 else
928 err = validate_vlan_mask_from_nlattrs(match, *key_attrs, inner,
929 a, log);
930 if (err <= 0)
931 return err;
932
933 err = encode_vlan_from_nlattrs(match, a, is_mask, inner);
934 if (err)
935 return err;
936
937 *key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
938 *key_attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
939 *key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
940
941 encap = a[OVS_KEY_ATTR_ENCAP];
942
943 if (!is_mask)
944 err = parse_flow_nlattrs(encap, a, key_attrs, log);
945 else
946 err = parse_flow_mask_nlattrs(encap, a, key_attrs, log);
947
948 return err;
949 }
950
951 static int parse_vlan_from_nlattrs(struct sw_flow_match *match,
952 u64 *key_attrs, const struct nlattr **a,
953 bool is_mask, bool log)
954 {
955 int err;
956 bool encap_valid = false;
957
958 err = __parse_vlan_from_nlattrs(match, key_attrs, false, a,
959 is_mask, log);
960 if (err)
961 return err;
962
963 encap_valid = !!(match->key->eth.vlan.tci & htons(VLAN_TAG_PRESENT));
964 if (encap_valid) {
965 err = __parse_vlan_from_nlattrs(match, key_attrs, true, a,
966 is_mask, log);
967 if (err)
968 return err;
969 }
970
971 return 0;
972 }
973
974 static int parse_eth_type_from_nlattrs(struct sw_flow_match *match,
975 u64 *attrs, const struct nlattr **a,
976 bool is_mask, bool log)
977 {
978 __be16 eth_type;
979
980 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
981 if (is_mask) {
982 /* Always exact match EtherType. */
983 eth_type = htons(0xffff);
984 } else if (!eth_proto_is_802_3(eth_type)) {
985 OVS_NLERR(log, "EtherType %x is less than min %x",
986 ntohs(eth_type), ETH_P_802_3_MIN);
987 return -EINVAL;
988 }
989
990 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
991 *attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
992 return 0;
993 }
994
995 static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
996 u64 *attrs, const struct nlattr **a,
997 bool is_mask, bool log)
998 {
999 u8 mac_proto = MAC_PROTO_ETHERNET;
1000
1001 if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
1002 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
1003
1004 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
1005 *attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
1006 }
1007
1008 if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
1009 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
1010
1011 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
1012 *attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
1013 }
1014
1015 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
1016 SW_FLOW_KEY_PUT(match, phy.priority,
1017 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
1018 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
1019 }
1020
1021 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
1022 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1023
1024 if (is_mask) {
1025 in_port = 0xffffffff; /* Always exact match in_port. */
1026 } else if (in_port >= DP_MAX_PORTS) {
1027 OVS_NLERR(log, "Port %d exceeds max allowable %d",
1028 in_port, DP_MAX_PORTS);
1029 return -EINVAL;
1030 }
1031
1032 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
1033 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1034 } else if (!is_mask) {
1035 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
1036 }
1037
1038 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1039 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1040
1041 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1042 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1043 }
1044 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1045 if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1046 is_mask, log) < 0)
1047 return -EINVAL;
1048 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1049 }
1050
1051 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
1052 ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
1053 u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
1054
1055 if (ct_state & ~CT_SUPPORTED_MASK) {
1056 OVS_NLERR(log, "ct_state flags %08x unsupported",
1057 ct_state);
1058 return -EINVAL;
1059 }
1060
1061 SW_FLOW_KEY_PUT(match, ct.state, ct_state, is_mask);
1062 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
1063 }
1064 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
1065 ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
1066 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
1067
1068 SW_FLOW_KEY_PUT(match, ct.zone, ct_zone, is_mask);
1069 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
1070 }
1071 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
1072 ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
1073 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
1074
1075 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
1076 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
1077 }
1078 if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
1079 ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
1080 const struct ovs_key_ct_labels *cl;
1081
1082 cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
1083 SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
1084 sizeof(*cl), is_mask);
1085 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
1086 }
1087
1088 /* For layer 3 packets the Ethernet type is provided
1089 * and treated as metadata but no MAC addresses are provided.
1090 */
1091 if (!(*attrs & (1ULL << OVS_KEY_ATTR_ETHERNET)) &&
1092 (*attrs & (1ULL << OVS_KEY_ATTR_ETHERTYPE)))
1093 mac_proto = MAC_PROTO_NONE;
1094
1095 /* Always exact match mac_proto */
1096 SW_FLOW_KEY_PUT(match, mac_proto, is_mask ? 0xff : mac_proto, is_mask);
1097
1098 if (mac_proto == MAC_PROTO_NONE)
1099 return parse_eth_type_from_nlattrs(match, attrs, a, is_mask,
1100 log);
1101
1102 return 0;
1103 }
1104
1105 static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
1106 u64 attrs, const struct nlattr **a,
1107 bool is_mask, bool log)
1108 {
1109 int err;
1110
1111 err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
1112 if (err)
1113 return err;
1114
1115 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1116 const struct ovs_key_ethernet *eth_key;
1117
1118 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1119 SW_FLOW_KEY_MEMCPY(match, eth.src,
1120 eth_key->eth_src, ETH_ALEN, is_mask);
1121 SW_FLOW_KEY_MEMCPY(match, eth.dst,
1122 eth_key->eth_dst, ETH_ALEN, is_mask);
1123 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1124
1125 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1126 /* VLAN attribute is always parsed before getting here since it
1127 * may occur multiple times.
1128 */
1129 OVS_NLERR(log, "VLAN attribute unexpected.");
1130 return -EINVAL;
1131 }
1132
1133 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1134 err = parse_eth_type_from_nlattrs(match, &attrs, a, is_mask,
1135 log);
1136 if (err)
1137 return err;
1138 } else if (!is_mask) {
1139 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1140 }
1141 } else if (!match->key->eth.type) {
1142 OVS_NLERR(log, "Either Ethernet header or EtherType is required.");
1143 return -EINVAL;
1144 }
1145
1146 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1147 const struct ovs_key_ipv4 *ipv4_key;
1148
1149 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1150 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
1151 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
1152 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
1153 return -EINVAL;
1154 }
1155 SW_FLOW_KEY_PUT(match, ip.proto,
1156 ipv4_key->ipv4_proto, is_mask);
1157 SW_FLOW_KEY_PUT(match, ip.tos,
1158 ipv4_key->ipv4_tos, is_mask);
1159 SW_FLOW_KEY_PUT(match, ip.ttl,
1160 ipv4_key->ipv4_ttl, is_mask);
1161 SW_FLOW_KEY_PUT(match, ip.frag,
1162 ipv4_key->ipv4_frag, is_mask);
1163 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1164 ipv4_key->ipv4_src, is_mask);
1165 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1166 ipv4_key->ipv4_dst, is_mask);
1167 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1168 }
1169
1170 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1171 const struct ovs_key_ipv6 *ipv6_key;
1172
1173 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1174 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
1175 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
1176 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
1177 return -EINVAL;
1178 }
1179
1180 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
1181 OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x).\n",
1182 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
1183 return -EINVAL;
1184 }
1185
1186 SW_FLOW_KEY_PUT(match, ipv6.label,
1187 ipv6_key->ipv6_label, is_mask);
1188 SW_FLOW_KEY_PUT(match, ip.proto,
1189 ipv6_key->ipv6_proto, is_mask);
1190 SW_FLOW_KEY_PUT(match, ip.tos,
1191 ipv6_key->ipv6_tclass, is_mask);
1192 SW_FLOW_KEY_PUT(match, ip.ttl,
1193 ipv6_key->ipv6_hlimit, is_mask);
1194 SW_FLOW_KEY_PUT(match, ip.frag,
1195 ipv6_key->ipv6_frag, is_mask);
1196 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1197 ipv6_key->ipv6_src,
1198 sizeof(match->key->ipv6.addr.src),
1199 is_mask);
1200 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1201 ipv6_key->ipv6_dst,
1202 sizeof(match->key->ipv6.addr.dst),
1203 is_mask);
1204
1205 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1206 }
1207
1208 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1209 const struct ovs_key_arp *arp_key;
1210
1211 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1212 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1213 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
1214 arp_key->arp_op);
1215 return -EINVAL;
1216 }
1217
1218 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1219 arp_key->arp_sip, is_mask);
1220 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1221 arp_key->arp_tip, is_mask);
1222 SW_FLOW_KEY_PUT(match, ip.proto,
1223 ntohs(arp_key->arp_op), is_mask);
1224 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1225 arp_key->arp_sha, ETH_ALEN, is_mask);
1226 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1227 arp_key->arp_tha, ETH_ALEN, is_mask);
1228
1229 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1230 }
1231
1232 if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1233 const struct ovs_key_mpls *mpls_key;
1234
1235 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1236 SW_FLOW_KEY_PUT(match, mpls.top_lse,
1237 mpls_key->mpls_lse, is_mask);
1238
1239 attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1240 }
1241
1242 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1243 const struct ovs_key_tcp *tcp_key;
1244
1245 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1246 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1247 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1248 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1249 }
1250
1251 if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1252 SW_FLOW_KEY_PUT(match, tp.flags,
1253 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1254 is_mask);
1255 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1256 }
1257
1258 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1259 const struct ovs_key_udp *udp_key;
1260
1261 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1262 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1263 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1264 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1265 }
1266
1267 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1268 const struct ovs_key_sctp *sctp_key;
1269
1270 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1271 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1272 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1273 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1274 }
1275
1276 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1277 const struct ovs_key_icmp *icmp_key;
1278
1279 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1280 SW_FLOW_KEY_PUT(match, tp.src,
1281 htons(icmp_key->icmp_type), is_mask);
1282 SW_FLOW_KEY_PUT(match, tp.dst,
1283 htons(icmp_key->icmp_code), is_mask);
1284 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1285 }
1286
1287 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1288 const struct ovs_key_icmpv6 *icmpv6_key;
1289
1290 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1291 SW_FLOW_KEY_PUT(match, tp.src,
1292 htons(icmpv6_key->icmpv6_type), is_mask);
1293 SW_FLOW_KEY_PUT(match, tp.dst,
1294 htons(icmpv6_key->icmpv6_code), is_mask);
1295 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1296 }
1297
1298 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1299 const struct ovs_key_nd *nd_key;
1300
1301 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1302 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1303 nd_key->nd_target,
1304 sizeof(match->key->ipv6.nd.target),
1305 is_mask);
1306 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1307 nd_key->nd_sll, ETH_ALEN, is_mask);
1308 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1309 nd_key->nd_tll, ETH_ALEN, is_mask);
1310 attrs &= ~(1 << OVS_KEY_ATTR_ND);
1311 }
1312
1313 if (attrs != 0) {
1314 OVS_NLERR(log, "Unknown key attributes %llx",
1315 (unsigned long long)attrs);
1316 return -EINVAL;
1317 }
1318
1319 return 0;
1320 }
1321
1322 static void nlattr_set(struct nlattr *attr, u8 val,
1323 const struct ovs_len_tbl *tbl)
1324 {
1325 struct nlattr *nla;
1326 int rem;
1327
1328 /* The nlattr stream should already have been validated */
1329 nla_for_each_nested(nla, attr, rem) {
1330 if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED) {
1331 if (tbl[nla_type(nla)].next)
1332 tbl = tbl[nla_type(nla)].next;
1333 nlattr_set(nla, val, tbl);
1334 } else {
1335 memset(nla_data(nla), val, nla_len(nla));
1336 }
1337
1338 if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1339 *(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1340 }
1341 }
1342
1343 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1344 {
1345 nlattr_set(attr, val, ovs_key_lens);
1346 }
1347
1348 /**
1349 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1350 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1351 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1352 * does not include any don't care bit.
1353 * @net: Used to determine per-namespace field support.
1354 * @match: receives the extracted flow match information.
1355 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1356 * sequence. The fields should of the packet that triggered the creation
1357 * of this flow.
1358 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1359 * attribute specifies the mask field of the wildcarded flow.
1360 * @log: Boolean to allow kernel error logging. Normally true, but when
1361 * probing for feature compatibility this should be passed in as false to
1362 * suppress unnecessary error logging.
1363 */
1364 int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1365 const struct nlattr *nla_key,
1366 const struct nlattr *nla_mask,
1367 bool log)
1368 {
1369 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1370 struct nlattr *newmask = NULL;
1371 u64 key_attrs = 0;
1372 u64 mask_attrs = 0;
1373 int err;
1374
1375 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1376 if (err)
1377 return err;
1378
1379 err = parse_vlan_from_nlattrs(match, &key_attrs, a, false, log);
1380 if (err)
1381 return err;
1382
1383 err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1384 if (err)
1385 return err;
1386
1387 if (match->mask) {
1388 if (!nla_mask) {
1389 /* Create an exact match mask. We need to set to 0xff
1390 * all the 'match->mask' fields that have been touched
1391 * in 'match->key'. We cannot simply memset
1392 * 'match->mask', because padding bytes and fields not
1393 * specified in 'match->key' should be left to 0.
1394 * Instead, we use a stream of netlink attributes,
1395 * copied from 'key' and set to 0xff.
1396 * ovs_key_from_nlattrs() will take care of filling
1397 * 'match->mask' appropriately.
1398 */
1399 newmask = kmemdup(nla_key,
1400 nla_total_size(nla_len(nla_key)),
1401 GFP_KERNEL);
1402 if (!newmask)
1403 return -ENOMEM;
1404
1405 mask_set_nlattr(newmask, 0xff);
1406
1407 /* The userspace does not send tunnel attributes that
1408 * are 0, but we should not wildcard them nonetheless.
1409 */
1410 if (match->key->tun_proto)
1411 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1412 0xff, true);
1413
1414 nla_mask = newmask;
1415 }
1416
1417 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1418 if (err)
1419 goto free_newmask;
1420
1421 /* Always match on tci. */
1422 SW_FLOW_KEY_PUT(match, eth.vlan.tci, htons(0xffff), true);
1423 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, htons(0xffff), true);
1424
1425 err = parse_vlan_from_nlattrs(match, &mask_attrs, a, true, log);
1426 if (err)
1427 goto free_newmask;
1428
1429 err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1430 log);
1431 if (err)
1432 goto free_newmask;
1433 }
1434
1435 if (!match_validate(match, key_attrs, mask_attrs, log))
1436 err = -EINVAL;
1437
1438 free_newmask:
1439 kfree(newmask);
1440 return err;
1441 }
1442
1443 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1444 {
1445 size_t len;
1446
1447 if (!attr)
1448 return 0;
1449
1450 len = nla_len(attr);
1451 if (len < 1 || len > MAX_UFID_LENGTH) {
1452 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1453 nla_len(attr), MAX_UFID_LENGTH);
1454 return 0;
1455 }
1456
1457 return len;
1458 }
1459
1460 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1461 * or false otherwise.
1462 */
1463 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1464 bool log)
1465 {
1466 sfid->ufid_len = get_ufid_len(attr, log);
1467 if (sfid->ufid_len)
1468 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1469
1470 return sfid->ufid_len;
1471 }
1472
1473 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1474 const struct sw_flow_key *key, bool log)
1475 {
1476 struct sw_flow_key *new_key;
1477
1478 if (ovs_nla_get_ufid(sfid, ufid, log))
1479 return 0;
1480
1481 /* If UFID was not provided, use unmasked key. */
1482 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1483 if (!new_key)
1484 return -ENOMEM;
1485 memcpy(new_key, key, sizeof(*key));
1486 sfid->unmasked_key = new_key;
1487
1488 return 0;
1489 }
1490
1491 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1492 {
1493 return attr ? nla_get_u32(attr) : 0;
1494 }
1495
1496 /**
1497 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1498 * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1499 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1500 * sequence.
1501 * @log: Boolean to allow kernel error logging. Normally true, but when
1502 * probing for feature compatibility this should be passed in as false to
1503 * suppress unnecessary error logging.
1504 *
1505 * This parses a series of Netlink attributes that form a flow key, which must
1506 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1507 * get the metadata, that is, the parts of the flow key that cannot be
1508 * extracted from the packet itself.
1509 */
1510
1511 int ovs_nla_get_flow_metadata(struct net *net, const struct nlattr *attr,
1512 struct sw_flow_key *key,
1513 bool log)
1514 {
1515 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1516 struct sw_flow_match match;
1517 u64 attrs = 0;
1518 int err;
1519
1520 err = parse_flow_nlattrs(attr, a, &attrs, log);
1521 if (err)
1522 return -EINVAL;
1523
1524 memset(&match, 0, sizeof(match));
1525 match.key = key;
1526
1527 memset(&key->ct, 0, sizeof(key->ct));
1528 key->phy.in_port = DP_MAX_PORTS;
1529
1530 return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1531 }
1532
1533 static int ovs_nla_put_vlan(struct sk_buff *skb, const struct vlan_head *vh,
1534 bool is_mask)
1535 {
1536 __be16 eth_type = !is_mask ? vh->tpid : htons(0xffff);
1537
1538 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1539 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, vh->tci))
1540 return -EMSGSIZE;
1541 return 0;
1542 }
1543
1544 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1545 const struct sw_flow_key *output, bool is_mask,
1546 struct sk_buff *skb)
1547 {
1548 struct ovs_key_ethernet *eth_key;
1549 struct nlattr *nla;
1550 struct nlattr *encap = NULL;
1551 struct nlattr *in_encap = NULL;
1552
1553 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1554 goto nla_put_failure;
1555
1556 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1557 goto nla_put_failure;
1558
1559 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1560 goto nla_put_failure;
1561
1562 if ((swkey->tun_proto || is_mask)) {
1563 const void *opts = NULL;
1564
1565 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1566 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1567
1568 if (ip_tun_to_nlattr(skb, &output->tun_key, opts,
1569 swkey->tun_opts_len, swkey->tun_proto))
1570 goto nla_put_failure;
1571 }
1572
1573 if (swkey->phy.in_port == DP_MAX_PORTS) {
1574 if (is_mask && (output->phy.in_port == 0xffff))
1575 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1576 goto nla_put_failure;
1577 } else {
1578 u16 upper_u16;
1579 upper_u16 = !is_mask ? 0 : 0xffff;
1580
1581 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1582 (upper_u16 << 16) | output->phy.in_port))
1583 goto nla_put_failure;
1584 }
1585
1586 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1587 goto nla_put_failure;
1588
1589 if (ovs_ct_put_key(output, skb))
1590 goto nla_put_failure;
1591
1592 if (ovs_key_mac_proto(swkey) == MAC_PROTO_ETHERNET) {
1593 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1594 if (!nla)
1595 goto nla_put_failure;
1596
1597 eth_key = nla_data(nla);
1598 ether_addr_copy(eth_key->eth_src, output->eth.src);
1599 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1600
1601 if (swkey->eth.vlan.tci || eth_type_vlan(swkey->eth.type)) {
1602 if (ovs_nla_put_vlan(skb, &output->eth.vlan, is_mask))
1603 goto nla_put_failure;
1604 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1605 if (!swkey->eth.vlan.tci)
1606 goto unencap;
1607
1608 if (swkey->eth.cvlan.tci || eth_type_vlan(swkey->eth.type)) {
1609 if (ovs_nla_put_vlan(skb, &output->eth.cvlan, is_mask))
1610 goto nla_put_failure;
1611 in_encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1612 if (!swkey->eth.cvlan.tci)
1613 goto unencap;
1614 }
1615 }
1616
1617 if (swkey->eth.type == htons(ETH_P_802_2)) {
1618 /*
1619 * Ethertype 802.2 is represented in the netlink with omitted
1620 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1621 * 0xffff in the mask attribute. Ethertype can also
1622 * be wildcarded.
1623 */
1624 if (is_mask && output->eth.type)
1625 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1626 output->eth.type))
1627 goto nla_put_failure;
1628 goto unencap;
1629 }
1630 }
1631
1632 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1633 goto nla_put_failure;
1634
1635 if (eth_type_vlan(swkey->eth.type)) {
1636 /* There are 3 VLAN tags, we don't know anything about the rest
1637 * of the packet, so truncate here.
1638 */
1639 WARN_ON_ONCE(!(encap && in_encap));
1640 goto unencap;
1641 }
1642
1643 if (swkey->eth.type == htons(ETH_P_IP)) {
1644 struct ovs_key_ipv4 *ipv4_key;
1645
1646 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1647 if (!nla)
1648 goto nla_put_failure;
1649 ipv4_key = nla_data(nla);
1650 ipv4_key->ipv4_src = output->ipv4.addr.src;
1651 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1652 ipv4_key->ipv4_proto = output->ip.proto;
1653 ipv4_key->ipv4_tos = output->ip.tos;
1654 ipv4_key->ipv4_ttl = output->ip.ttl;
1655 ipv4_key->ipv4_frag = output->ip.frag;
1656 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1657 struct ovs_key_ipv6 *ipv6_key;
1658
1659 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1660 if (!nla)
1661 goto nla_put_failure;
1662 ipv6_key = nla_data(nla);
1663 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1664 sizeof(ipv6_key->ipv6_src));
1665 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1666 sizeof(ipv6_key->ipv6_dst));
1667 ipv6_key->ipv6_label = output->ipv6.label;
1668 ipv6_key->ipv6_proto = output->ip.proto;
1669 ipv6_key->ipv6_tclass = output->ip.tos;
1670 ipv6_key->ipv6_hlimit = output->ip.ttl;
1671 ipv6_key->ipv6_frag = output->ip.frag;
1672 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1673 swkey->eth.type == htons(ETH_P_RARP)) {
1674 struct ovs_key_arp *arp_key;
1675
1676 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1677 if (!nla)
1678 goto nla_put_failure;
1679 arp_key = nla_data(nla);
1680 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1681 arp_key->arp_sip = output->ipv4.addr.src;
1682 arp_key->arp_tip = output->ipv4.addr.dst;
1683 arp_key->arp_op = htons(output->ip.proto);
1684 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1685 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1686 } else if (eth_p_mpls(swkey->eth.type)) {
1687 struct ovs_key_mpls *mpls_key;
1688
1689 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1690 if (!nla)
1691 goto nla_put_failure;
1692 mpls_key = nla_data(nla);
1693 mpls_key->mpls_lse = output->mpls.top_lse;
1694 }
1695
1696 if ((swkey->eth.type == htons(ETH_P_IP) ||
1697 swkey->eth.type == htons(ETH_P_IPV6)) &&
1698 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1699
1700 if (swkey->ip.proto == IPPROTO_TCP) {
1701 struct ovs_key_tcp *tcp_key;
1702
1703 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1704 if (!nla)
1705 goto nla_put_failure;
1706 tcp_key = nla_data(nla);
1707 tcp_key->tcp_src = output->tp.src;
1708 tcp_key->tcp_dst = output->tp.dst;
1709 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1710 output->tp.flags))
1711 goto nla_put_failure;
1712 } else if (swkey->ip.proto == IPPROTO_UDP) {
1713 struct ovs_key_udp *udp_key;
1714
1715 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1716 if (!nla)
1717 goto nla_put_failure;
1718 udp_key = nla_data(nla);
1719 udp_key->udp_src = output->tp.src;
1720 udp_key->udp_dst = output->tp.dst;
1721 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1722 struct ovs_key_sctp *sctp_key;
1723
1724 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1725 if (!nla)
1726 goto nla_put_failure;
1727 sctp_key = nla_data(nla);
1728 sctp_key->sctp_src = output->tp.src;
1729 sctp_key->sctp_dst = output->tp.dst;
1730 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1731 swkey->ip.proto == IPPROTO_ICMP) {
1732 struct ovs_key_icmp *icmp_key;
1733
1734 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1735 if (!nla)
1736 goto nla_put_failure;
1737 icmp_key = nla_data(nla);
1738 icmp_key->icmp_type = ntohs(output->tp.src);
1739 icmp_key->icmp_code = ntohs(output->tp.dst);
1740 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1741 swkey->ip.proto == IPPROTO_ICMPV6) {
1742 struct ovs_key_icmpv6 *icmpv6_key;
1743
1744 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1745 sizeof(*icmpv6_key));
1746 if (!nla)
1747 goto nla_put_failure;
1748 icmpv6_key = nla_data(nla);
1749 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1750 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1751
1752 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1753 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1754 struct ovs_key_nd *nd_key;
1755
1756 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1757 if (!nla)
1758 goto nla_put_failure;
1759 nd_key = nla_data(nla);
1760 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1761 sizeof(nd_key->nd_target));
1762 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1763 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1764 }
1765 }
1766 }
1767
1768 unencap:
1769 if (in_encap)
1770 nla_nest_end(skb, in_encap);
1771 if (encap)
1772 nla_nest_end(skb, encap);
1773
1774 return 0;
1775
1776 nla_put_failure:
1777 return -EMSGSIZE;
1778 }
1779
1780 int ovs_nla_put_key(const struct sw_flow_key *swkey,
1781 const struct sw_flow_key *output, int attr, bool is_mask,
1782 struct sk_buff *skb)
1783 {
1784 int err;
1785 struct nlattr *nla;
1786
1787 nla = nla_nest_start(skb, attr);
1788 if (!nla)
1789 return -EMSGSIZE;
1790 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
1791 if (err)
1792 return err;
1793 nla_nest_end(skb, nla);
1794
1795 return 0;
1796 }
1797
1798 /* Called with ovs_mutex or RCU read lock. */
1799 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
1800 {
1801 if (ovs_identifier_is_ufid(&flow->id))
1802 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
1803 flow->id.ufid);
1804
1805 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
1806 OVS_FLOW_ATTR_KEY, false, skb);
1807 }
1808
1809 /* Called with ovs_mutex or RCU read lock. */
1810 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
1811 {
1812 return ovs_nla_put_key(&flow->key, &flow->key,
1813 OVS_FLOW_ATTR_KEY, false, skb);
1814 }
1815
1816 /* Called with ovs_mutex or RCU read lock. */
1817 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
1818 {
1819 return ovs_nla_put_key(&flow->key, &flow->mask->key,
1820 OVS_FLOW_ATTR_MASK, true, skb);
1821 }
1822
1823 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1824
1825 static struct sw_flow_actions *nla_alloc_flow_actions(int size, bool log)
1826 {
1827 struct sw_flow_actions *sfa;
1828
1829 if (size > MAX_ACTIONS_BUFSIZE) {
1830 OVS_NLERR(log, "Flow action size %u bytes exceeds max", size);
1831 return ERR_PTR(-EINVAL);
1832 }
1833
1834 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1835 if (!sfa)
1836 return ERR_PTR(-ENOMEM);
1837
1838 sfa->actions_len = 0;
1839 return sfa;
1840 }
1841
1842 static void ovs_nla_free_set_action(const struct nlattr *a)
1843 {
1844 const struct nlattr *ovs_key = nla_data(a);
1845 struct ovs_tunnel_info *ovs_tun;
1846
1847 switch (nla_type(ovs_key)) {
1848 case OVS_KEY_ATTR_TUNNEL_INFO:
1849 ovs_tun = nla_data(ovs_key);
1850 dst_release((struct dst_entry *)ovs_tun->tun_dst);
1851 break;
1852 }
1853 }
1854
1855 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1856 {
1857 const struct nlattr *a;
1858 int rem;
1859
1860 if (!sf_acts)
1861 return;
1862
1863 nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
1864 switch (nla_type(a)) {
1865 case OVS_ACTION_ATTR_SET:
1866 ovs_nla_free_set_action(a);
1867 break;
1868 case OVS_ACTION_ATTR_CT:
1869 ovs_ct_free_action(a);
1870 break;
1871 }
1872 }
1873
1874 kfree(sf_acts);
1875 }
1876
1877 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
1878 {
1879 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
1880 }
1881
1882 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1883 * The caller must hold rcu_read_lock for this to be sensible. */
1884 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
1885 {
1886 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
1887 }
1888
1889 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1890 int attr_len, bool log)
1891 {
1892
1893 struct sw_flow_actions *acts;
1894 int new_acts_size;
1895 int req_size = NLA_ALIGN(attr_len);
1896 int next_offset = offsetof(struct sw_flow_actions, actions) +
1897 (*sfa)->actions_len;
1898
1899 if (req_size <= (ksize(*sfa) - next_offset))
1900 goto out;
1901
1902 new_acts_size = ksize(*sfa) * 2;
1903
1904 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1905 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1906 return ERR_PTR(-EMSGSIZE);
1907 new_acts_size = MAX_ACTIONS_BUFSIZE;
1908 }
1909
1910 acts = nla_alloc_flow_actions(new_acts_size, log);
1911 if (IS_ERR(acts))
1912 return (void *)acts;
1913
1914 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1915 acts->actions_len = (*sfa)->actions_len;
1916 acts->orig_len = (*sfa)->orig_len;
1917 kfree(*sfa);
1918 *sfa = acts;
1919
1920 out:
1921 (*sfa)->actions_len += req_size;
1922 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1923 }
1924
1925 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1926 int attrtype, void *data, int len, bool log)
1927 {
1928 struct nlattr *a;
1929
1930 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
1931 if (IS_ERR(a))
1932 return a;
1933
1934 a->nla_type = attrtype;
1935 a->nla_len = nla_attr_size(len);
1936
1937 if (data)
1938 memcpy(nla_data(a), data, len);
1939 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1940
1941 return a;
1942 }
1943
1944 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
1945 int len, bool log)
1946 {
1947 struct nlattr *a;
1948
1949 a = __add_action(sfa, attrtype, data, len, log);
1950
1951 return PTR_ERR_OR_ZERO(a);
1952 }
1953
1954 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1955 int attrtype, bool log)
1956 {
1957 int used = (*sfa)->actions_len;
1958 int err;
1959
1960 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
1961 if (err)
1962 return err;
1963
1964 return used;
1965 }
1966
1967 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1968 int st_offset)
1969 {
1970 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1971 st_offset);
1972
1973 a->nla_len = sfa->actions_len - st_offset;
1974 }
1975
1976 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
1977 const struct sw_flow_key *key,
1978 int depth, struct sw_flow_actions **sfa,
1979 __be16 eth_type, __be16 vlan_tci, bool log);
1980
1981 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
1982 const struct sw_flow_key *key, int depth,
1983 struct sw_flow_actions **sfa,
1984 __be16 eth_type, __be16 vlan_tci, bool log)
1985 {
1986 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1987 const struct nlattr *probability, *actions;
1988 const struct nlattr *a;
1989 int rem, start, err, st_acts;
1990
1991 memset(attrs, 0, sizeof(attrs));
1992 nla_for_each_nested(a, attr, rem) {
1993 int type = nla_type(a);
1994 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1995 return -EINVAL;
1996 attrs[type] = a;
1997 }
1998 if (rem)
1999 return -EINVAL;
2000
2001 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
2002 if (!probability || nla_len(probability) != sizeof(u32))
2003 return -EINVAL;
2004
2005 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
2006 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
2007 return -EINVAL;
2008
2009 /* validation done, copy sample action. */
2010 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
2011 if (start < 0)
2012 return start;
2013 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
2014 nla_data(probability), sizeof(u32), log);
2015 if (err)
2016 return err;
2017 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS, log);
2018 if (st_acts < 0)
2019 return st_acts;
2020
2021 err = __ovs_nla_copy_actions(net, actions, key, depth + 1, sfa,
2022 eth_type, vlan_tci, log);
2023 if (err)
2024 return err;
2025
2026 add_nested_action_end(*sfa, st_acts);
2027 add_nested_action_end(*sfa, start);
2028
2029 return 0;
2030 }
2031
2032 void ovs_match_init(struct sw_flow_match *match,
2033 struct sw_flow_key *key,
2034 bool reset_key,
2035 struct sw_flow_mask *mask)
2036 {
2037 memset(match, 0, sizeof(*match));
2038 match->key = key;
2039 match->mask = mask;
2040
2041 if (reset_key)
2042 memset(key, 0, sizeof(*key));
2043
2044 if (mask) {
2045 memset(&mask->key, 0, sizeof(mask->key));
2046 mask->range.start = mask->range.end = 0;
2047 }
2048 }
2049
2050 static int validate_geneve_opts(struct sw_flow_key *key)
2051 {
2052 struct geneve_opt *option;
2053 int opts_len = key->tun_opts_len;
2054 bool crit_opt = false;
2055
2056 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2057 while (opts_len > 0) {
2058 int len;
2059
2060 if (opts_len < sizeof(*option))
2061 return -EINVAL;
2062
2063 len = sizeof(*option) + option->length * 4;
2064 if (len > opts_len)
2065 return -EINVAL;
2066
2067 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2068
2069 option = (struct geneve_opt *)((u8 *)option + len);
2070 opts_len -= len;
2071 };
2072
2073 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
2074
2075 return 0;
2076 }
2077
2078 static int validate_and_copy_set_tun(const struct nlattr *attr,
2079 struct sw_flow_actions **sfa, bool log)
2080 {
2081 struct sw_flow_match match;
2082 struct sw_flow_key key;
2083 struct metadata_dst *tun_dst;
2084 struct ip_tunnel_info *tun_info;
2085 struct ovs_tunnel_info *ovs_tun;
2086 struct nlattr *a;
2087 int err = 0, start, opts_type;
2088
2089 ovs_match_init(&match, &key, true, NULL);
2090 opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2091 if (opts_type < 0)
2092 return opts_type;
2093
2094 if (key.tun_opts_len) {
2095 switch (opts_type) {
2096 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2097 err = validate_geneve_opts(&key);
2098 if (err < 0)
2099 return err;
2100 break;
2101 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
2102 break;
2103 }
2104 };
2105
2106 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2107 if (start < 0)
2108 return start;
2109
2110 tun_dst = metadata_dst_alloc(key.tun_opts_len, GFP_KERNEL);
2111 if (!tun_dst)
2112 return -ENOMEM;
2113
2114 err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2115 if (err) {
2116 dst_release((struct dst_entry *)tun_dst);
2117 return err;
2118 }
2119
2120 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2121 sizeof(*ovs_tun), log);
2122 if (IS_ERR(a)) {
2123 dst_release((struct dst_entry *)tun_dst);
2124 return PTR_ERR(a);
2125 }
2126
2127 ovs_tun = nla_data(a);
2128 ovs_tun->tun_dst = tun_dst;
2129
2130 tun_info = &tun_dst->u.tun_info;
2131 tun_info->mode = IP_TUNNEL_INFO_TX;
2132 if (key.tun_proto == AF_INET6)
2133 tun_info->mode |= IP_TUNNEL_INFO_IPV6;
2134 tun_info->key = key.tun_key;
2135
2136 /* We need to store the options in the action itself since
2137 * everything else will go away after flow setup. We can append
2138 * it to tun_info and then point there.
2139 */
2140 ip_tunnel_info_opts_set(tun_info,
2141 TUN_METADATA_OPTS(&key, key.tun_opts_len),
2142 key.tun_opts_len);
2143 add_nested_action_end(*sfa, start);
2144
2145 return err;
2146 }
2147
2148 /* Return false if there are any non-masked bits set.
2149 * Mask follows data immediately, before any netlink padding.
2150 */
2151 static bool validate_masked(u8 *data, int len)
2152 {
2153 u8 *mask = data + len;
2154
2155 while (len--)
2156 if (*data++ & ~*mask++)
2157 return false;
2158
2159 return true;
2160 }
2161
2162 static int validate_set(const struct nlattr *a,
2163 const struct sw_flow_key *flow_key,
2164 struct sw_flow_actions **sfa, bool *skip_copy,
2165 u8 mac_proto, __be16 eth_type, bool masked, bool log)
2166 {
2167 const struct nlattr *ovs_key = nla_data(a);
2168 int key_type = nla_type(ovs_key);
2169 size_t key_len;
2170
2171 /* There can be only one key in a action */
2172 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2173 return -EINVAL;
2174
2175 key_len = nla_len(ovs_key);
2176 if (masked)
2177 key_len /= 2;
2178
2179 if (key_type > OVS_KEY_ATTR_MAX ||
2180 !check_attr_len(key_len, ovs_key_lens[key_type].len))
2181 return -EINVAL;
2182
2183 if (masked && !validate_masked(nla_data(ovs_key), key_len))
2184 return -EINVAL;
2185
2186 switch (key_type) {
2187 const struct ovs_key_ipv4 *ipv4_key;
2188 const struct ovs_key_ipv6 *ipv6_key;
2189 int err;
2190
2191 case OVS_KEY_ATTR_PRIORITY:
2192 case OVS_KEY_ATTR_SKB_MARK:
2193 case OVS_KEY_ATTR_CT_MARK:
2194 case OVS_KEY_ATTR_CT_LABELS:
2195 break;
2196
2197 case OVS_KEY_ATTR_ETHERNET:
2198 if (mac_proto != MAC_PROTO_ETHERNET)
2199 return -EINVAL;
2200 break;
2201
2202 case OVS_KEY_ATTR_TUNNEL:
2203 if (masked)
2204 return -EINVAL; /* Masked tunnel set not supported. */
2205
2206 *skip_copy = true;
2207 err = validate_and_copy_set_tun(a, sfa, log);
2208 if (err)
2209 return err;
2210 break;
2211
2212 case OVS_KEY_ATTR_IPV4:
2213 if (eth_type != htons(ETH_P_IP))
2214 return -EINVAL;
2215
2216 ipv4_key = nla_data(ovs_key);
2217
2218 if (masked) {
2219 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2220
2221 /* Non-writeable fields. */
2222 if (mask->ipv4_proto || mask->ipv4_frag)
2223 return -EINVAL;
2224 } else {
2225 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2226 return -EINVAL;
2227
2228 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2229 return -EINVAL;
2230 }
2231 break;
2232
2233 case OVS_KEY_ATTR_IPV6:
2234 if (eth_type != htons(ETH_P_IPV6))
2235 return -EINVAL;
2236
2237 ipv6_key = nla_data(ovs_key);
2238
2239 if (masked) {
2240 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2241
2242 /* Non-writeable fields. */
2243 if (mask->ipv6_proto || mask->ipv6_frag)
2244 return -EINVAL;
2245
2246 /* Invalid bits in the flow label mask? */
2247 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2248 return -EINVAL;
2249 } else {
2250 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2251 return -EINVAL;
2252
2253 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2254 return -EINVAL;
2255 }
2256 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2257 return -EINVAL;
2258
2259 break;
2260
2261 case OVS_KEY_ATTR_TCP:
2262 if ((eth_type != htons(ETH_P_IP) &&
2263 eth_type != htons(ETH_P_IPV6)) ||
2264 flow_key->ip.proto != IPPROTO_TCP)
2265 return -EINVAL;
2266
2267 break;
2268
2269 case OVS_KEY_ATTR_UDP:
2270 if ((eth_type != htons(ETH_P_IP) &&
2271 eth_type != htons(ETH_P_IPV6)) ||
2272 flow_key->ip.proto != IPPROTO_UDP)
2273 return -EINVAL;
2274
2275 break;
2276
2277 case OVS_KEY_ATTR_MPLS:
2278 if (!eth_p_mpls(eth_type))
2279 return -EINVAL;
2280 break;
2281
2282 case OVS_KEY_ATTR_SCTP:
2283 if ((eth_type != htons(ETH_P_IP) &&
2284 eth_type != htons(ETH_P_IPV6)) ||
2285 flow_key->ip.proto != IPPROTO_SCTP)
2286 return -EINVAL;
2287
2288 break;
2289
2290 default:
2291 return -EINVAL;
2292 }
2293
2294 /* Convert non-masked non-tunnel set actions to masked set actions. */
2295 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2296 int start, len = key_len * 2;
2297 struct nlattr *at;
2298
2299 *skip_copy = true;
2300
2301 start = add_nested_action_start(sfa,
2302 OVS_ACTION_ATTR_SET_TO_MASKED,
2303 log);
2304 if (start < 0)
2305 return start;
2306
2307 at = __add_action(sfa, key_type, NULL, len, log);
2308 if (IS_ERR(at))
2309 return PTR_ERR(at);
2310
2311 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2312 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
2313 /* Clear non-writeable bits from otherwise writeable fields. */
2314 if (key_type == OVS_KEY_ATTR_IPV6) {
2315 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2316
2317 mask->ipv6_label &= htonl(0x000FFFFF);
2318 }
2319 add_nested_action_end(*sfa, start);
2320 }
2321
2322 return 0;
2323 }
2324
2325 static int validate_userspace(const struct nlattr *attr)
2326 {
2327 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2328 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2329 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2330 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2331 };
2332 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2333 int error;
2334
2335 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
2336 attr, userspace_policy);
2337 if (error)
2338 return error;
2339
2340 if (!a[OVS_USERSPACE_ATTR_PID] ||
2341 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2342 return -EINVAL;
2343
2344 return 0;
2345 }
2346
2347 static int copy_action(const struct nlattr *from,
2348 struct sw_flow_actions **sfa, bool log)
2349 {
2350 int totlen = NLA_ALIGN(from->nla_len);
2351 struct nlattr *to;
2352
2353 to = reserve_sfa_size(sfa, from->nla_len, log);
2354 if (IS_ERR(to))
2355 return PTR_ERR(to);
2356
2357 memcpy(to, from, totlen);
2358 return 0;
2359 }
2360
2361 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2362 const struct sw_flow_key *key,
2363 int depth, struct sw_flow_actions **sfa,
2364 __be16 eth_type, __be16 vlan_tci, bool log)
2365 {
2366 u8 mac_proto = ovs_key_mac_proto(key);
2367 const struct nlattr *a;
2368 int rem, err;
2369
2370 if (depth >= SAMPLE_ACTION_DEPTH)
2371 return -EOVERFLOW;
2372
2373 nla_for_each_nested(a, attr, rem) {
2374 /* Expected argument lengths, (u32)-1 for variable length. */
2375 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2376 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2377 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2378 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2379 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2380 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2381 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2382 [OVS_ACTION_ATTR_POP_VLAN] = 0,
2383 [OVS_ACTION_ATTR_SET] = (u32)-1,
2384 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2385 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2386 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2387 [OVS_ACTION_ATTR_CT] = (u32)-1,
2388 [OVS_ACTION_ATTR_TRUNC] = sizeof(struct ovs_action_trunc),
2389 [OVS_ACTION_ATTR_PUSH_ETH] = sizeof(struct ovs_action_push_eth),
2390 [OVS_ACTION_ATTR_POP_ETH] = 0,
2391 };
2392 const struct ovs_action_push_vlan *vlan;
2393 int type = nla_type(a);
2394 bool skip_copy;
2395
2396 if (type > OVS_ACTION_ATTR_MAX ||
2397 (action_lens[type] != nla_len(a) &&
2398 action_lens[type] != (u32)-1))
2399 return -EINVAL;
2400
2401 skip_copy = false;
2402 switch (type) {
2403 case OVS_ACTION_ATTR_UNSPEC:
2404 return -EINVAL;
2405
2406 case OVS_ACTION_ATTR_USERSPACE:
2407 err = validate_userspace(a);
2408 if (err)
2409 return err;
2410 break;
2411
2412 case OVS_ACTION_ATTR_OUTPUT:
2413 if (nla_get_u32(a) >= DP_MAX_PORTS)
2414 return -EINVAL;
2415 break;
2416
2417 case OVS_ACTION_ATTR_TRUNC: {
2418 const struct ovs_action_trunc *trunc = nla_data(a);
2419
2420 if (trunc->max_len < ETH_HLEN)
2421 return -EINVAL;
2422 break;
2423 }
2424
2425 case OVS_ACTION_ATTR_HASH: {
2426 const struct ovs_action_hash *act_hash = nla_data(a);
2427
2428 switch (act_hash->hash_alg) {
2429 case OVS_HASH_ALG_L4:
2430 break;
2431 default:
2432 return -EINVAL;
2433 }
2434
2435 break;
2436 }
2437
2438 case OVS_ACTION_ATTR_POP_VLAN:
2439 if (mac_proto != MAC_PROTO_ETHERNET)
2440 return -EINVAL;
2441 vlan_tci = htons(0);
2442 break;
2443
2444 case OVS_ACTION_ATTR_PUSH_VLAN:
2445 if (mac_proto != MAC_PROTO_ETHERNET)
2446 return -EINVAL;
2447 vlan = nla_data(a);
2448 if (!eth_type_vlan(vlan->vlan_tpid))
2449 return -EINVAL;
2450 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2451 return -EINVAL;
2452 vlan_tci = vlan->vlan_tci;
2453 break;
2454
2455 case OVS_ACTION_ATTR_RECIRC:
2456 break;
2457
2458 case OVS_ACTION_ATTR_PUSH_MPLS: {
2459 const struct ovs_action_push_mpls *mpls = nla_data(a);
2460
2461 if (!eth_p_mpls(mpls->mpls_ethertype))
2462 return -EINVAL;
2463 /* Prohibit push MPLS other than to a white list
2464 * for packets that have a known tag order.
2465 */
2466 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2467 (eth_type != htons(ETH_P_IP) &&
2468 eth_type != htons(ETH_P_IPV6) &&
2469 eth_type != htons(ETH_P_ARP) &&
2470 eth_type != htons(ETH_P_RARP) &&
2471 !eth_p_mpls(eth_type)))
2472 return -EINVAL;
2473 eth_type = mpls->mpls_ethertype;
2474 break;
2475 }
2476
2477 case OVS_ACTION_ATTR_POP_MPLS:
2478 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2479 !eth_p_mpls(eth_type))
2480 return -EINVAL;
2481
2482 /* Disallow subsequent L2.5+ set and mpls_pop actions
2483 * as there is no check here to ensure that the new
2484 * eth_type is valid and thus set actions could
2485 * write off the end of the packet or otherwise
2486 * corrupt it.
2487 *
2488 * Support for these actions is planned using packet
2489 * recirculation.
2490 */
2491 eth_type = htons(0);
2492 break;
2493
2494 case OVS_ACTION_ATTR_SET:
2495 err = validate_set(a, key, sfa,
2496 &skip_copy, mac_proto, eth_type,
2497 false, log);
2498 if (err)
2499 return err;
2500 break;
2501
2502 case OVS_ACTION_ATTR_SET_MASKED:
2503 err = validate_set(a, key, sfa,
2504 &skip_copy, mac_proto, eth_type,
2505 true, log);
2506 if (err)
2507 return err;
2508 break;
2509
2510 case OVS_ACTION_ATTR_SAMPLE:
2511 err = validate_and_copy_sample(net, a, key, depth, sfa,
2512 eth_type, vlan_tci, log);
2513 if (err)
2514 return err;
2515 skip_copy = true;
2516 break;
2517
2518 case OVS_ACTION_ATTR_CT:
2519 err = ovs_ct_copy_action(net, a, key, sfa, log);
2520 if (err)
2521 return err;
2522 skip_copy = true;
2523 break;
2524
2525 case OVS_ACTION_ATTR_PUSH_ETH:
2526 /* Disallow pushing an Ethernet header if one
2527 * is already present */
2528 if (mac_proto != MAC_PROTO_NONE)
2529 return -EINVAL;
2530 mac_proto = MAC_PROTO_NONE;
2531 break;
2532
2533 case OVS_ACTION_ATTR_POP_ETH:
2534 if (mac_proto != MAC_PROTO_ETHERNET)
2535 return -EINVAL;
2536 if (vlan_tci & htons(VLAN_TAG_PRESENT))
2537 return -EINVAL;
2538 mac_proto = MAC_PROTO_ETHERNET;
2539 break;
2540
2541 default:
2542 OVS_NLERR(log, "Unknown Action type %d", type);
2543 return -EINVAL;
2544 }
2545 if (!skip_copy) {
2546 err = copy_action(a, sfa, log);
2547 if (err)
2548 return err;
2549 }
2550 }
2551
2552 if (rem > 0)
2553 return -EINVAL;
2554
2555 return 0;
2556 }
2557
2558 /* 'key' must be the masked key. */
2559 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2560 const struct sw_flow_key *key,
2561 struct sw_flow_actions **sfa, bool log)
2562 {
2563 int err;
2564
2565 *sfa = nla_alloc_flow_actions(nla_len(attr), log);
2566 if (IS_ERR(*sfa))
2567 return PTR_ERR(*sfa);
2568
2569 (*sfa)->orig_len = nla_len(attr);
2570 err = __ovs_nla_copy_actions(net, attr, key, 0, sfa, key->eth.type,
2571 key->eth.vlan.tci, log);
2572 if (err)
2573 ovs_nla_free_flow_actions(*sfa);
2574
2575 return err;
2576 }
2577
2578 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
2579 {
2580 const struct nlattr *a;
2581 struct nlattr *start;
2582 int err = 0, rem;
2583
2584 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
2585 if (!start)
2586 return -EMSGSIZE;
2587
2588 nla_for_each_nested(a, attr, rem) {
2589 int type = nla_type(a);
2590 struct nlattr *st_sample;
2591
2592 switch (type) {
2593 case OVS_SAMPLE_ATTR_PROBABILITY:
2594 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
2595 sizeof(u32), nla_data(a)))
2596 return -EMSGSIZE;
2597 break;
2598 case OVS_SAMPLE_ATTR_ACTIONS:
2599 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
2600 if (!st_sample)
2601 return -EMSGSIZE;
2602 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
2603 if (err)
2604 return err;
2605 nla_nest_end(skb, st_sample);
2606 break;
2607 }
2608 }
2609
2610 nla_nest_end(skb, start);
2611 return err;
2612 }
2613
2614 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
2615 {
2616 const struct nlattr *ovs_key = nla_data(a);
2617 int key_type = nla_type(ovs_key);
2618 struct nlattr *start;
2619 int err;
2620
2621 switch (key_type) {
2622 case OVS_KEY_ATTR_TUNNEL_INFO: {
2623 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
2624 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
2625
2626 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2627 if (!start)
2628 return -EMSGSIZE;
2629
2630 err = ip_tun_to_nlattr(skb, &tun_info->key,
2631 ip_tunnel_info_opts(tun_info),
2632 tun_info->options_len,
2633 ip_tunnel_info_af(tun_info));
2634 if (err)
2635 return err;
2636 nla_nest_end(skb, start);
2637 break;
2638 }
2639 default:
2640 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
2641 return -EMSGSIZE;
2642 break;
2643 }
2644
2645 return 0;
2646 }
2647
2648 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
2649 struct sk_buff *skb)
2650 {
2651 const struct nlattr *ovs_key = nla_data(a);
2652 struct nlattr *nla;
2653 size_t key_len = nla_len(ovs_key) / 2;
2654
2655 /* Revert the conversion we did from a non-masked set action to
2656 * masked set action.
2657 */
2658 nla = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2659 if (!nla)
2660 return -EMSGSIZE;
2661
2662 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
2663 return -EMSGSIZE;
2664
2665 nla_nest_end(skb, nla);
2666 return 0;
2667 }
2668
2669 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
2670 {
2671 const struct nlattr *a;
2672 int rem, err;
2673
2674 nla_for_each_attr(a, attr, len, rem) {
2675 int type = nla_type(a);
2676
2677 switch (type) {
2678 case OVS_ACTION_ATTR_SET:
2679 err = set_action_to_attr(a, skb);
2680 if (err)
2681 return err;
2682 break;
2683
2684 case OVS_ACTION_ATTR_SET_TO_MASKED:
2685 err = masked_set_action_to_set_action_attr(a, skb);
2686 if (err)
2687 return err;
2688 break;
2689
2690 case OVS_ACTION_ATTR_SAMPLE:
2691 err = sample_action_to_attr(a, skb);
2692 if (err)
2693 return err;
2694 break;
2695
2696 case OVS_ACTION_ATTR_CT:
2697 err = ovs_ct_action_to_attr(nla_data(a), skb);
2698 if (err)
2699 return err;
2700 break;
2701
2702 default:
2703 if (nla_put(skb, type, nla_len(a), nla_data(a)))
2704 return -EMSGSIZE;
2705 break;
2706 }
2707 }
2708
2709 return 0;
2710 }
ubuntu-zesty-kernel mirror