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
update_range(struct sw_flow_match * match,size_t offset,size_t size,bool is_mask)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
match_validate(const struct sw_flow_match * match,u64 key_attrs,u64 mask_attrs,bool log)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 = 1 << OVS_KEY_ATTR_ETHERNET;
127 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
128
129 /* The following mask attributes allowed only if they
130 * pass the validation tests. */
131 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
132 | (1 << OVS_KEY_ATTR_IPV6)
133 | (1 << OVS_KEY_ATTR_TCP)
134 | (1 << OVS_KEY_ATTR_TCP_FLAGS)
135 | (1 << OVS_KEY_ATTR_UDP)
136 | (1 << OVS_KEY_ATTR_SCTP)
137 | (1 << OVS_KEY_ATTR_ICMP)
138 | (1 << OVS_KEY_ATTR_ICMPV6)
139 | (1 << OVS_KEY_ATTR_ARP)
140 | (1 << OVS_KEY_ATTR_ND)
141 | (1 << OVS_KEY_ATTR_MPLS));
142
143 /* Always allowed mask fields. */
144 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
145 | (1 << OVS_KEY_ATTR_IN_PORT)
146 | (1 << OVS_KEY_ATTR_ETHERTYPE));
147
148 /* Check key attributes. */
149 if (match->key->eth.type == htons(ETH_P_ARP)
150 || match->key->eth.type == htons(ETH_P_RARP)) {
151 key_expected |= 1 << OVS_KEY_ATTR_ARP;
152 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
153 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
154 }
155
156 if (eth_p_mpls(match->key->eth.type)) {
157 key_expected |= 1 << OVS_KEY_ATTR_MPLS;
158 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
159 mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
160 }
161
162 if (match->key->eth.type == htons(ETH_P_IP)) {
163 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
164 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
165 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
166
167 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
168 if (match->key->ip.proto == IPPROTO_UDP) {
169 key_expected |= 1 << OVS_KEY_ATTR_UDP;
170 if (match->mask && (match->mask->key.ip.proto == 0xff))
171 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
172 }
173
174 if (match->key->ip.proto == IPPROTO_SCTP) {
175 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
176 if (match->mask && (match->mask->key.ip.proto == 0xff))
177 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
178 }
179
180 if (match->key->ip.proto == IPPROTO_TCP) {
181 key_expected |= 1 << OVS_KEY_ATTR_TCP;
182 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
183 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
184 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
185 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
186 }
187 }
188
189 if (match->key->ip.proto == IPPROTO_ICMP) {
190 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
191 if (match->mask && (match->mask->key.ip.proto == 0xff))
192 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
193 }
194 }
195 }
196
197 if (match->key->eth.type == htons(ETH_P_IPV6)) {
198 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
199 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
200 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
201
202 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
203 if (match->key->ip.proto == IPPROTO_UDP) {
204 key_expected |= 1 << OVS_KEY_ATTR_UDP;
205 if (match->mask && (match->mask->key.ip.proto == 0xff))
206 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
207 }
208
209 if (match->key->ip.proto == IPPROTO_SCTP) {
210 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
211 if (match->mask && (match->mask->key.ip.proto == 0xff))
212 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
213 }
214
215 if (match->key->ip.proto == IPPROTO_TCP) {
216 key_expected |= 1 << OVS_KEY_ATTR_TCP;
217 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
218 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
219 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
220 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
221 }
222 }
223
224 if (match->key->ip.proto == IPPROTO_ICMPV6) {
225 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
226 if (match->mask && (match->mask->key.ip.proto == 0xff))
227 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
228
229 if (match->key->tp.src ==
230 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
231 match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
232 key_expected |= 1 << OVS_KEY_ATTR_ND;
233 if (match->mask && (match->mask->key.tp.src == htons(0xff)))
234 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
235 }
236 }
237 }
238 }
239
240 if ((key_attrs & key_expected) != key_expected) {
241 /* Key attributes check failed. */
242 OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
243 (unsigned long long)key_attrs,
244 (unsigned long long)key_expected);
245 return false;
246 }
247
248 if ((mask_attrs & mask_allowed) != mask_attrs) {
249 /* Mask attributes check failed. */
250 OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
251 (unsigned long long)mask_attrs,
252 (unsigned long long)mask_allowed);
253 return false;
254 }
255
256 return true;
257 }
258
ovs_tun_key_attr_size(void)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
ovs_key_attr_size(void)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
check_attr_len(unsigned int attr_len,unsigned int expected_len)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
is_all_zero(const u8 * fp,size_t size)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
__parse_flow_nlattrs(const struct nlattr * attr,const struct nlattr * a[],u64 * attrsp,bool log,bool nz)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
parse_flow_mask_nlattrs(const struct nlattr * attr,const struct nlattr * a[],u64 * attrsp,bool log)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
parse_flow_nlattrs(const struct nlattr * attr,const struct nlattr * a[],u64 * attrsp,bool log)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
genev_tun_opt_from_nlattr(const struct nlattr * a,struct sw_flow_match * match,bool is_mask,bool log)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
vxlan_tun_opt_from_nlattr(const struct nlattr * attr,struct sw_flow_match * match,bool is_mask,bool log)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
ip_tun_from_nlattr(const struct nlattr * attr,struct sw_flow_match * match,bool is_mask,bool log)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
vxlan_opt_to_nlattr(struct sk_buff * skb,const void * tun_opts,int swkey_tun_opts_len)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
__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)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
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)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
ovs_nla_put_tunnel_info(struct sk_buff * skb,struct ip_tunnel_info * tun_info)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
encode_vlan_from_nlattrs(struct sw_flow_match * match,const struct nlattr * a[],bool is_mask,bool inner)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
validate_vlan_from_nlattrs(const struct sw_flow_match * match,u64 key_attrs,bool inner,const struct nlattr ** a,bool log)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
validate_vlan_mask_from_nlattrs(const struct sw_flow_match * match,u64 key_attrs,bool inner,const struct nlattr ** a,bool log)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
__parse_vlan_from_nlattrs(struct sw_flow_match * match,u64 * key_attrs,bool inner,const struct nlattr ** a,bool is_mask,bool log)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
parse_vlan_from_nlattrs(struct sw_flow_match * match,u64 * key_attrs,const struct nlattr ** a,bool is_mask,bool log)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
metadata_from_nlattrs(struct net * net,struct sw_flow_match * match,u64 * attrs,const struct nlattr ** a,bool is_mask,bool log)974 static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
975 u64 *attrs, const struct nlattr **a,
976 bool is_mask, bool log)
977 {
978 if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
979 u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
980
981 SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
982 *attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
983 }
984
985 if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
986 u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
987
988 SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
989 *attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
990 }
991
992 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
993 SW_FLOW_KEY_PUT(match, phy.priority,
994 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
995 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
996 }
997
998 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
999 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1000
1001 if (is_mask) {
1002 in_port = 0xffffffff; /* Always exact match in_port. */
1003 } else if (in_port >= DP_MAX_PORTS) {
1004 OVS_NLERR(log, "Port %d exceeds max allowable %d",
1005 in_port, DP_MAX_PORTS);
1006 return -EINVAL;
1007 }
1008
1009 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
1010 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1011 } else if (!is_mask) {
1012 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
1013 }
1014
1015 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1016 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1017
1018 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1019 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1020 }
1021 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1022 if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1023 is_mask, log) < 0)
1024 return -EINVAL;
1025 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1026 }
1027
1028 if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
1029 ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
1030 u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
1031
1032 if (ct_state & ~CT_SUPPORTED_MASK) {
1033 OVS_NLERR(log, "ct_state flags %08x unsupported",
1034 ct_state);
1035 return -EINVAL;
1036 }
1037
1038 SW_FLOW_KEY_PUT(match, ct.state, ct_state, is_mask);
1039 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
1040 }
1041 if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
1042 ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
1043 u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
1044
1045 SW_FLOW_KEY_PUT(match, ct.zone, ct_zone, is_mask);
1046 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
1047 }
1048 if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
1049 ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
1050 u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
1051
1052 SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
1053 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
1054 }
1055 if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
1056 ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
1057 const struct ovs_key_ct_labels *cl;
1058
1059 cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
1060 SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
1061 sizeof(*cl), is_mask);
1062 *attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
1063 }
1064 return 0;
1065 }
1066
ovs_key_from_nlattrs(struct net * net,struct sw_flow_match * match,u64 attrs,const struct nlattr ** a,bool is_mask,bool log)1067 static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
1068 u64 attrs, const struct nlattr **a,
1069 bool is_mask, bool log)
1070 {
1071 int err;
1072
1073 err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
1074 if (err)
1075 return err;
1076
1077 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1078 const struct ovs_key_ethernet *eth_key;
1079
1080 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1081 SW_FLOW_KEY_MEMCPY(match, eth.src,
1082 eth_key->eth_src, ETH_ALEN, is_mask);
1083 SW_FLOW_KEY_MEMCPY(match, eth.dst,
1084 eth_key->eth_dst, ETH_ALEN, is_mask);
1085 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1086 }
1087
1088 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1089 /* VLAN attribute is always parsed before getting here since it
1090 * may occur multiple times.
1091 */
1092 OVS_NLERR(log, "VLAN attribute unexpected.");
1093 return -EINVAL;
1094 }
1095
1096 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1097 __be16 eth_type;
1098
1099 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1100 if (is_mask) {
1101 /* Always exact match EtherType. */
1102 eth_type = htons(0xffff);
1103 } else if (!eth_proto_is_802_3(eth_type)) {
1104 OVS_NLERR(log, "EtherType %x is less than min %x",
1105 ntohs(eth_type), ETH_P_802_3_MIN);
1106 return -EINVAL;
1107 }
1108
1109 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
1110 attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1111 } else if (!is_mask) {
1112 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1113 }
1114
1115 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1116 const struct ovs_key_ipv4 *ipv4_key;
1117
1118 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1119 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
1120 OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
1121 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
1122 return -EINVAL;
1123 }
1124 SW_FLOW_KEY_PUT(match, ip.proto,
1125 ipv4_key->ipv4_proto, is_mask);
1126 SW_FLOW_KEY_PUT(match, ip.tos,
1127 ipv4_key->ipv4_tos, is_mask);
1128 SW_FLOW_KEY_PUT(match, ip.ttl,
1129 ipv4_key->ipv4_ttl, is_mask);
1130 SW_FLOW_KEY_PUT(match, ip.frag,
1131 ipv4_key->ipv4_frag, is_mask);
1132 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1133 ipv4_key->ipv4_src, is_mask);
1134 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1135 ipv4_key->ipv4_dst, is_mask);
1136 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1137 }
1138
1139 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1140 const struct ovs_key_ipv6 *ipv6_key;
1141
1142 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1143 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
1144 OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
1145 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
1146 return -EINVAL;
1147 }
1148
1149 if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
1150 OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x).\n",
1151 ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
1152 return -EINVAL;
1153 }
1154
1155 SW_FLOW_KEY_PUT(match, ipv6.label,
1156 ipv6_key->ipv6_label, is_mask);
1157 SW_FLOW_KEY_PUT(match, ip.proto,
1158 ipv6_key->ipv6_proto, is_mask);
1159 SW_FLOW_KEY_PUT(match, ip.tos,
1160 ipv6_key->ipv6_tclass, is_mask);
1161 SW_FLOW_KEY_PUT(match, ip.ttl,
1162 ipv6_key->ipv6_hlimit, is_mask);
1163 SW_FLOW_KEY_PUT(match, ip.frag,
1164 ipv6_key->ipv6_frag, is_mask);
1165 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1166 ipv6_key->ipv6_src,
1167 sizeof(match->key->ipv6.addr.src),
1168 is_mask);
1169 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1170 ipv6_key->ipv6_dst,
1171 sizeof(match->key->ipv6.addr.dst),
1172 is_mask);
1173
1174 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1175 }
1176
1177 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1178 const struct ovs_key_arp *arp_key;
1179
1180 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1181 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1182 OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
1183 arp_key->arp_op);
1184 return -EINVAL;
1185 }
1186
1187 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1188 arp_key->arp_sip, is_mask);
1189 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1190 arp_key->arp_tip, is_mask);
1191 SW_FLOW_KEY_PUT(match, ip.proto,
1192 ntohs(arp_key->arp_op), is_mask);
1193 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1194 arp_key->arp_sha, ETH_ALEN, is_mask);
1195 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1196 arp_key->arp_tha, ETH_ALEN, is_mask);
1197
1198 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1199 }
1200
1201 if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1202 const struct ovs_key_mpls *mpls_key;
1203
1204 mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1205 SW_FLOW_KEY_PUT(match, mpls.top_lse,
1206 mpls_key->mpls_lse, is_mask);
1207
1208 attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1209 }
1210
1211 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1212 const struct ovs_key_tcp *tcp_key;
1213
1214 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1215 SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1216 SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1217 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1218 }
1219
1220 if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1221 SW_FLOW_KEY_PUT(match, tp.flags,
1222 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1223 is_mask);
1224 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1225 }
1226
1227 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1228 const struct ovs_key_udp *udp_key;
1229
1230 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1231 SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1232 SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1233 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1234 }
1235
1236 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1237 const struct ovs_key_sctp *sctp_key;
1238
1239 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1240 SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1241 SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1242 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1243 }
1244
1245 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1246 const struct ovs_key_icmp *icmp_key;
1247
1248 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1249 SW_FLOW_KEY_PUT(match, tp.src,
1250 htons(icmp_key->icmp_type), is_mask);
1251 SW_FLOW_KEY_PUT(match, tp.dst,
1252 htons(icmp_key->icmp_code), is_mask);
1253 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1254 }
1255
1256 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1257 const struct ovs_key_icmpv6 *icmpv6_key;
1258
1259 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1260 SW_FLOW_KEY_PUT(match, tp.src,
1261 htons(icmpv6_key->icmpv6_type), is_mask);
1262 SW_FLOW_KEY_PUT(match, tp.dst,
1263 htons(icmpv6_key->icmpv6_code), is_mask);
1264 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1265 }
1266
1267 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1268 const struct ovs_key_nd *nd_key;
1269
1270 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1271 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1272 nd_key->nd_target,
1273 sizeof(match->key->ipv6.nd.target),
1274 is_mask);
1275 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1276 nd_key->nd_sll, ETH_ALEN, is_mask);
1277 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1278 nd_key->nd_tll, ETH_ALEN, is_mask);
1279 attrs &= ~(1 << OVS_KEY_ATTR_ND);
1280 }
1281
1282 if (attrs != 0) {
1283 OVS_NLERR(log, "Unknown key attributes %llx",
1284 (unsigned long long)attrs);
1285 return -EINVAL;
1286 }
1287
1288 return 0;
1289 }
1290
nlattr_set(struct nlattr * attr,u8 val,const struct ovs_len_tbl * tbl)1291 static void nlattr_set(struct nlattr *attr, u8 val,
1292 const struct ovs_len_tbl *tbl)
1293 {
1294 struct nlattr *nla;
1295 int rem;
1296
1297 /* The nlattr stream should already have been validated */
1298 nla_for_each_nested(nla, attr, rem) {
1299 if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED) {
1300 if (tbl[nla_type(nla)].next)
1301 tbl = tbl[nla_type(nla)].next;
1302 nlattr_set(nla, val, tbl);
1303 } else {
1304 memset(nla_data(nla), val, nla_len(nla));
1305 }
1306
1307 if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1308 *(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1309 }
1310 }
1311
mask_set_nlattr(struct nlattr * attr,u8 val)1312 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1313 {
1314 nlattr_set(attr, val, ovs_key_lens);
1315 }
1316
1317 /**
1318 * ovs_nla_get_match - parses Netlink attributes into a flow key and
1319 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1320 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1321 * does not include any don't care bit.
1322 * @net: Used to determine per-namespace field support.
1323 * @match: receives the extracted flow match information.
1324 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1325 * sequence. The fields should of the packet that triggered the creation
1326 * of this flow.
1327 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1328 * attribute specifies the mask field of the wildcarded flow.
1329 * @log: Boolean to allow kernel error logging. Normally true, but when
1330 * probing for feature compatibility this should be passed in as false to
1331 * suppress unnecessary error logging.
1332 */
ovs_nla_get_match(struct net * net,struct sw_flow_match * match,const struct nlattr * nla_key,const struct nlattr * nla_mask,bool log)1333 int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1334 const struct nlattr *nla_key,
1335 const struct nlattr *nla_mask,
1336 bool log)
1337 {
1338 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1339 struct nlattr *newmask = NULL;
1340 u64 key_attrs = 0;
1341 u64 mask_attrs = 0;
1342 int err;
1343
1344 err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1345 if (err)
1346 return err;
1347
1348 err = parse_vlan_from_nlattrs(match, &key_attrs, a, false, log);
1349 if (err)
1350 return err;
1351
1352 err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1353 if (err)
1354 return err;
1355
1356 if (match->mask) {
1357 if (!nla_mask) {
1358 /* Create an exact match mask. We need to set to 0xff
1359 * all the 'match->mask' fields that have been touched
1360 * in 'match->key'. We cannot simply memset
1361 * 'match->mask', because padding bytes and fields not
1362 * specified in 'match->key' should be left to 0.
1363 * Instead, we use a stream of netlink attributes,
1364 * copied from 'key' and set to 0xff.
1365 * ovs_key_from_nlattrs() will take care of filling
1366 * 'match->mask' appropriately.
1367 */
1368 newmask = kmemdup(nla_key,
1369 nla_total_size(nla_len(nla_key)),
1370 GFP_KERNEL);
1371 if (!newmask)
1372 return -ENOMEM;
1373
1374 mask_set_nlattr(newmask, 0xff);
1375
1376 /* The userspace does not send tunnel attributes that
1377 * are 0, but we should not wildcard them nonetheless.
1378 */
1379 if (match->key->tun_proto)
1380 SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1381 0xff, true);
1382
1383 nla_mask = newmask;
1384 }
1385
1386 err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1387 if (err)
1388 goto free_newmask;
1389
1390 /* Always match on tci. */
1391 SW_FLOW_KEY_PUT(match, eth.vlan.tci, htons(0xffff), true);
1392 SW_FLOW_KEY_PUT(match, eth.cvlan.tci, htons(0xffff), true);
1393
1394 err = parse_vlan_from_nlattrs(match, &mask_attrs, a, true, log);
1395 if (err)
1396 goto free_newmask;
1397
1398 err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1399 log);
1400 if (err)
1401 goto free_newmask;
1402 }
1403
1404 if (!match_validate(match, key_attrs, mask_attrs, log))
1405 err = -EINVAL;
1406
1407 free_newmask:
1408 kfree(newmask);
1409 return err;
1410 }
1411
get_ufid_len(const struct nlattr * attr,bool log)1412 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1413 {
1414 size_t len;
1415
1416 if (!attr)
1417 return 0;
1418
1419 len = nla_len(attr);
1420 if (len < 1 || len > MAX_UFID_LENGTH) {
1421 OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1422 nla_len(attr), MAX_UFID_LENGTH);
1423 return 0;
1424 }
1425
1426 return len;
1427 }
1428
1429 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1430 * or false otherwise.
1431 */
ovs_nla_get_ufid(struct sw_flow_id * sfid,const struct nlattr * attr,bool log)1432 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1433 bool log)
1434 {
1435 sfid->ufid_len = get_ufid_len(attr, log);
1436 if (sfid->ufid_len)
1437 memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1438
1439 return sfid->ufid_len;
1440 }
1441
ovs_nla_get_identifier(struct sw_flow_id * sfid,const struct nlattr * ufid,const struct sw_flow_key * key,bool log)1442 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1443 const struct sw_flow_key *key, bool log)
1444 {
1445 struct sw_flow_key *new_key;
1446
1447 if (ovs_nla_get_ufid(sfid, ufid, log))
1448 return 0;
1449
1450 /* If UFID was not provided, use unmasked key. */
1451 new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1452 if (!new_key)
1453 return -ENOMEM;
1454 memcpy(new_key, key, sizeof(*key));
1455 sfid->unmasked_key = new_key;
1456
1457 return 0;
1458 }
1459
ovs_nla_get_ufid_flags(const struct nlattr * attr)1460 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1461 {
1462 return attr ? nla_get_u32(attr) : 0;
1463 }
1464
1465 /**
1466 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1467 * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1468 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1469 * sequence.
1470 * @log: Boolean to allow kernel error logging. Normally true, but when
1471 * probing for feature compatibility this should be passed in as false to
1472 * suppress unnecessary error logging.
1473 *
1474 * This parses a series of Netlink attributes that form a flow key, which must
1475 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1476 * get the metadata, that is, the parts of the flow key that cannot be
1477 * extracted from the packet itself.
1478 */
1479
ovs_nla_get_flow_metadata(struct net * net,const struct nlattr * attr,struct sw_flow_key * key,bool log)1480 int ovs_nla_get_flow_metadata(struct net *net, const struct nlattr *attr,
1481 struct sw_flow_key *key,
1482 bool log)
1483 {
1484 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1485 struct sw_flow_match match;
1486 u64 attrs = 0;
1487 int err;
1488
1489 err = parse_flow_nlattrs(attr, a, &attrs, log);
1490 if (err)
1491 return -EINVAL;
1492
1493 memset(&match, 0, sizeof(match));
1494 match.key = key;
1495
1496 memset(&key->ct, 0, sizeof(key->ct));
1497 key->phy.in_port = DP_MAX_PORTS;
1498
1499 return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1500 }
1501
ovs_nla_put_vlan(struct sk_buff * skb,const struct vlan_head * vh,bool is_mask)1502 static int ovs_nla_put_vlan(struct sk_buff *skb, const struct vlan_head *vh,
1503 bool is_mask)
1504 {
1505 __be16 eth_type = !is_mask ? vh->tpid : htons(0xffff);
1506
1507 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1508 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, vh->tci))
1509 return -EMSGSIZE;
1510 return 0;
1511 }
1512
__ovs_nla_put_key(const struct sw_flow_key * swkey,const struct sw_flow_key * output,bool is_mask,struct sk_buff * skb)1513 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1514 const struct sw_flow_key *output, bool is_mask,
1515 struct sk_buff *skb)
1516 {
1517 struct ovs_key_ethernet *eth_key;
1518 struct nlattr *nla;
1519 struct nlattr *encap = NULL;
1520 struct nlattr *in_encap = NULL;
1521
1522 if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1523 goto nla_put_failure;
1524
1525 if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1526 goto nla_put_failure;
1527
1528 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1529 goto nla_put_failure;
1530
1531 if ((swkey->tun_proto || is_mask)) {
1532 const void *opts = NULL;
1533
1534 if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1535 opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1536
1537 if (ip_tun_to_nlattr(skb, &output->tun_key, opts,
1538 swkey->tun_opts_len, swkey->tun_proto))
1539 goto nla_put_failure;
1540 }
1541
1542 if (swkey->phy.in_port == DP_MAX_PORTS) {
1543 if (is_mask && (output->phy.in_port == 0xffff))
1544 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1545 goto nla_put_failure;
1546 } else {
1547 u16 upper_u16;
1548 upper_u16 = !is_mask ? 0 : 0xffff;
1549
1550 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1551 (upper_u16 << 16) | output->phy.in_port))
1552 goto nla_put_failure;
1553 }
1554
1555 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1556 goto nla_put_failure;
1557
1558 if (ovs_ct_put_key(output, skb))
1559 goto nla_put_failure;
1560
1561 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1562 if (!nla)
1563 goto nla_put_failure;
1564
1565 eth_key = nla_data(nla);
1566 ether_addr_copy(eth_key->eth_src, output->eth.src);
1567 ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1568
1569 if (swkey->eth.vlan.tci || eth_type_vlan(swkey->eth.type)) {
1570 if (ovs_nla_put_vlan(skb, &output->eth.vlan, is_mask))
1571 goto nla_put_failure;
1572 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1573 if (!swkey->eth.vlan.tci)
1574 goto unencap;
1575
1576 if (swkey->eth.cvlan.tci || eth_type_vlan(swkey->eth.type)) {
1577 if (ovs_nla_put_vlan(skb, &output->eth.cvlan, is_mask))
1578 goto nla_put_failure;
1579 in_encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1580 if (!swkey->eth.cvlan.tci)
1581 goto unencap;
1582 }
1583 }
1584
1585 if (swkey->eth.type == htons(ETH_P_802_2)) {
1586 /*
1587 * Ethertype 802.2 is represented in the netlink with omitted
1588 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1589 * 0xffff in the mask attribute. Ethertype can also
1590 * be wildcarded.
1591 */
1592 if (is_mask && output->eth.type)
1593 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1594 output->eth.type))
1595 goto nla_put_failure;
1596 goto unencap;
1597 }
1598
1599 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1600 goto nla_put_failure;
1601
1602 if (eth_type_vlan(swkey->eth.type)) {
1603 /* There are 3 VLAN tags, we don't know anything about the rest
1604 * of the packet, so truncate here.
1605 */
1606 WARN_ON_ONCE(!(encap && in_encap));
1607 goto unencap;
1608 }
1609
1610 if (swkey->eth.type == htons(ETH_P_IP)) {
1611 struct ovs_key_ipv4 *ipv4_key;
1612
1613 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1614 if (!nla)
1615 goto nla_put_failure;
1616 ipv4_key = nla_data(nla);
1617 ipv4_key->ipv4_src = output->ipv4.addr.src;
1618 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1619 ipv4_key->ipv4_proto = output->ip.proto;
1620 ipv4_key->ipv4_tos = output->ip.tos;
1621 ipv4_key->ipv4_ttl = output->ip.ttl;
1622 ipv4_key->ipv4_frag = output->ip.frag;
1623 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1624 struct ovs_key_ipv6 *ipv6_key;
1625
1626 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1627 if (!nla)
1628 goto nla_put_failure;
1629 ipv6_key = nla_data(nla);
1630 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1631 sizeof(ipv6_key->ipv6_src));
1632 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1633 sizeof(ipv6_key->ipv6_dst));
1634 ipv6_key->ipv6_label = output->ipv6.label;
1635 ipv6_key->ipv6_proto = output->ip.proto;
1636 ipv6_key->ipv6_tclass = output->ip.tos;
1637 ipv6_key->ipv6_hlimit = output->ip.ttl;
1638 ipv6_key->ipv6_frag = output->ip.frag;
1639 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1640 swkey->eth.type == htons(ETH_P_RARP)) {
1641 struct ovs_key_arp *arp_key;
1642
1643 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1644 if (!nla)
1645 goto nla_put_failure;
1646 arp_key = nla_data(nla);
1647 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1648 arp_key->arp_sip = output->ipv4.addr.src;
1649 arp_key->arp_tip = output->ipv4.addr.dst;
1650 arp_key->arp_op = htons(output->ip.proto);
1651 ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1652 ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1653 } else if (eth_p_mpls(swkey->eth.type)) {
1654 struct ovs_key_mpls *mpls_key;
1655
1656 nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1657 if (!nla)
1658 goto nla_put_failure;
1659 mpls_key = nla_data(nla);
1660 mpls_key->mpls_lse = output->mpls.top_lse;
1661 }
1662
1663 if ((swkey->eth.type == htons(ETH_P_IP) ||
1664 swkey->eth.type == htons(ETH_P_IPV6)) &&
1665 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1666
1667 if (swkey->ip.proto == IPPROTO_TCP) {
1668 struct ovs_key_tcp *tcp_key;
1669
1670 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1671 if (!nla)
1672 goto nla_put_failure;
1673 tcp_key = nla_data(nla);
1674 tcp_key->tcp_src = output->tp.src;
1675 tcp_key->tcp_dst = output->tp.dst;
1676 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1677 output->tp.flags))
1678 goto nla_put_failure;
1679 } else if (swkey->ip.proto == IPPROTO_UDP) {
1680 struct ovs_key_udp *udp_key;
1681
1682 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1683 if (!nla)
1684 goto nla_put_failure;
1685 udp_key = nla_data(nla);
1686 udp_key->udp_src = output->tp.src;
1687 udp_key->udp_dst = output->tp.dst;
1688 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1689 struct ovs_key_sctp *sctp_key;
1690
1691 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1692 if (!nla)
1693 goto nla_put_failure;
1694 sctp_key = nla_data(nla);
1695 sctp_key->sctp_src = output->tp.src;
1696 sctp_key->sctp_dst = output->tp.dst;
1697 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1698 swkey->ip.proto == IPPROTO_ICMP) {
1699 struct ovs_key_icmp *icmp_key;
1700
1701 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1702 if (!nla)
1703 goto nla_put_failure;
1704 icmp_key = nla_data(nla);
1705 icmp_key->icmp_type = ntohs(output->tp.src);
1706 icmp_key->icmp_code = ntohs(output->tp.dst);
1707 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1708 swkey->ip.proto == IPPROTO_ICMPV6) {
1709 struct ovs_key_icmpv6 *icmpv6_key;
1710
1711 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1712 sizeof(*icmpv6_key));
1713 if (!nla)
1714 goto nla_put_failure;
1715 icmpv6_key = nla_data(nla);
1716 icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1717 icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1718
1719 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1720 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1721 struct ovs_key_nd *nd_key;
1722
1723 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1724 if (!nla)
1725 goto nla_put_failure;
1726 nd_key = nla_data(nla);
1727 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1728 sizeof(nd_key->nd_target));
1729 ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1730 ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1731 }
1732 }
1733 }
1734
1735 unencap:
1736 if (in_encap)
1737 nla_nest_end(skb, in_encap);
1738 if (encap)
1739 nla_nest_end(skb, encap);
1740
1741 return 0;
1742
1743 nla_put_failure:
1744 return -EMSGSIZE;
1745 }
1746
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)1747 int ovs_nla_put_key(const struct sw_flow_key *swkey,
1748 const struct sw_flow_key *output, int attr, bool is_mask,
1749 struct sk_buff *skb)
1750 {
1751 int err;
1752 struct nlattr *nla;
1753
1754 nla = nla_nest_start(skb, attr);
1755 if (!nla)
1756 return -EMSGSIZE;
1757 err = __ovs_nla_put_key(swkey, output, is_mask, skb);
1758 if (err)
1759 return err;
1760 nla_nest_end(skb, nla);
1761
1762 return 0;
1763 }
1764
1765 /* Called with ovs_mutex or RCU read lock. */
ovs_nla_put_identifier(const struct sw_flow * flow,struct sk_buff * skb)1766 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
1767 {
1768 if (ovs_identifier_is_ufid(&flow->id))
1769 return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
1770 flow->id.ufid);
1771
1772 return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
1773 OVS_FLOW_ATTR_KEY, false, skb);
1774 }
1775
1776 /* Called with ovs_mutex or RCU read lock. */
ovs_nla_put_masked_key(const struct sw_flow * flow,struct sk_buff * skb)1777 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
1778 {
1779 return ovs_nla_put_key(&flow->key, &flow->key,
1780 OVS_FLOW_ATTR_KEY, false, skb);
1781 }
1782
1783 /* Called with ovs_mutex or RCU read lock. */
ovs_nla_put_mask(const struct sw_flow * flow,struct sk_buff * skb)1784 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
1785 {
1786 return ovs_nla_put_key(&flow->key, &flow->mask->key,
1787 OVS_FLOW_ATTR_MASK, true, skb);
1788 }
1789
1790 #define MAX_ACTIONS_BUFSIZE (32 * 1024)
1791
nla_alloc_flow_actions(int size)1792 static struct sw_flow_actions *nla_alloc_flow_actions(int size)
1793 {
1794 struct sw_flow_actions *sfa;
1795
1796 WARN_ON_ONCE(size > MAX_ACTIONS_BUFSIZE);
1797
1798 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1799 if (!sfa)
1800 return ERR_PTR(-ENOMEM);
1801
1802 sfa->actions_len = 0;
1803 return sfa;
1804 }
1805
ovs_nla_free_set_action(const struct nlattr * a)1806 static void ovs_nla_free_set_action(const struct nlattr *a)
1807 {
1808 const struct nlattr *ovs_key = nla_data(a);
1809 struct ovs_tunnel_info *ovs_tun;
1810
1811 switch (nla_type(ovs_key)) {
1812 case OVS_KEY_ATTR_TUNNEL_INFO:
1813 ovs_tun = nla_data(ovs_key);
1814 dst_release((struct dst_entry *)ovs_tun->tun_dst);
1815 break;
1816 }
1817 }
1818
ovs_nla_free_flow_actions(struct sw_flow_actions * sf_acts)1819 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1820 {
1821 const struct nlattr *a;
1822 int rem;
1823
1824 if (!sf_acts)
1825 return;
1826
1827 nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
1828 switch (nla_type(a)) {
1829 case OVS_ACTION_ATTR_SET:
1830 ovs_nla_free_set_action(a);
1831 break;
1832 case OVS_ACTION_ATTR_CT:
1833 ovs_ct_free_action(a);
1834 break;
1835 }
1836 }
1837
1838 kfree(sf_acts);
1839 }
1840
__ovs_nla_free_flow_actions(struct rcu_head * head)1841 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
1842 {
1843 ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
1844 }
1845
1846 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1847 * The caller must hold rcu_read_lock for this to be sensible. */
ovs_nla_free_flow_actions_rcu(struct sw_flow_actions * sf_acts)1848 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
1849 {
1850 call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
1851 }
1852
reserve_sfa_size(struct sw_flow_actions ** sfa,int attr_len,bool log)1853 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1854 int attr_len, bool log)
1855 {
1856
1857 struct sw_flow_actions *acts;
1858 int new_acts_size;
1859 int req_size = NLA_ALIGN(attr_len);
1860 int next_offset = offsetof(struct sw_flow_actions, actions) +
1861 (*sfa)->actions_len;
1862
1863 if (req_size <= (ksize(*sfa) - next_offset))
1864 goto out;
1865
1866 new_acts_size = ksize(*sfa) * 2;
1867
1868 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1869 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size) {
1870 OVS_NLERR(log, "Flow action size exceeds max %u",
1871 MAX_ACTIONS_BUFSIZE);
1872 return ERR_PTR(-EMSGSIZE);
1873 }
1874 new_acts_size = MAX_ACTIONS_BUFSIZE;
1875 }
1876
1877 acts = nla_alloc_flow_actions(new_acts_size);
1878 if (IS_ERR(acts))
1879 return (void *)acts;
1880
1881 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1882 acts->actions_len = (*sfa)->actions_len;
1883 acts->orig_len = (*sfa)->orig_len;
1884 kfree(*sfa);
1885 *sfa = acts;
1886
1887 out:
1888 (*sfa)->actions_len += req_size;
1889 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1890 }
1891
__add_action(struct sw_flow_actions ** sfa,int attrtype,void * data,int len,bool log)1892 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1893 int attrtype, void *data, int len, bool log)
1894 {
1895 struct nlattr *a;
1896
1897 a = reserve_sfa_size(sfa, nla_attr_size(len), log);
1898 if (IS_ERR(a))
1899 return a;
1900
1901 a->nla_type = attrtype;
1902 a->nla_len = nla_attr_size(len);
1903
1904 if (data)
1905 memcpy(nla_data(a), data, len);
1906 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1907
1908 return a;
1909 }
1910
ovs_nla_add_action(struct sw_flow_actions ** sfa,int attrtype,void * data,int len,bool log)1911 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
1912 int len, bool log)
1913 {
1914 struct nlattr *a;
1915
1916 a = __add_action(sfa, attrtype, data, len, log);
1917
1918 return PTR_ERR_OR_ZERO(a);
1919 }
1920
add_nested_action_start(struct sw_flow_actions ** sfa,int attrtype,bool log)1921 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1922 int attrtype, bool log)
1923 {
1924 int used = (*sfa)->actions_len;
1925 int err;
1926
1927 err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
1928 if (err)
1929 return err;
1930
1931 return used;
1932 }
1933
add_nested_action_end(struct sw_flow_actions * sfa,int st_offset)1934 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1935 int st_offset)
1936 {
1937 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1938 st_offset);
1939
1940 a->nla_len = sfa->actions_len - st_offset;
1941 }
1942
1943 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
1944 const struct sw_flow_key *key,
1945 int depth, struct sw_flow_actions **sfa,
1946 __be16 eth_type, __be16 vlan_tci, bool log);
1947
validate_and_copy_sample(struct net * net,const struct nlattr * attr,const struct sw_flow_key * key,int depth,struct sw_flow_actions ** sfa,__be16 eth_type,__be16 vlan_tci,bool log)1948 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
1949 const struct sw_flow_key *key, int depth,
1950 struct sw_flow_actions **sfa,
1951 __be16 eth_type, __be16 vlan_tci, bool log)
1952 {
1953 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1954 const struct nlattr *probability, *actions;
1955 const struct nlattr *a;
1956 int rem, start, err, st_acts;
1957
1958 memset(attrs, 0, sizeof(attrs));
1959 nla_for_each_nested(a, attr, rem) {
1960 int type = nla_type(a);
1961 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1962 return -EINVAL;
1963 attrs[type] = a;
1964 }
1965 if (rem)
1966 return -EINVAL;
1967
1968 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1969 if (!probability || nla_len(probability) != sizeof(u32))
1970 return -EINVAL;
1971
1972 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1973 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1974 return -EINVAL;
1975
1976 /* validation done, copy sample action. */
1977 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
1978 if (start < 0)
1979 return start;
1980 err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1981 nla_data(probability), sizeof(u32), log);
1982 if (err)
1983 return err;
1984 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS, log);
1985 if (st_acts < 0)
1986 return st_acts;
1987
1988 err = __ovs_nla_copy_actions(net, actions, key, depth + 1, sfa,
1989 eth_type, vlan_tci, log);
1990 if (err)
1991 return err;
1992
1993 add_nested_action_end(*sfa, st_acts);
1994 add_nested_action_end(*sfa, start);
1995
1996 return 0;
1997 }
1998
ovs_match_init(struct sw_flow_match * match,struct sw_flow_key * key,bool reset_key,struct sw_flow_mask * mask)1999 void ovs_match_init(struct sw_flow_match *match,
2000 struct sw_flow_key *key,
2001 bool reset_key,
2002 struct sw_flow_mask *mask)
2003 {
2004 memset(match, 0, sizeof(*match));
2005 match->key = key;
2006 match->mask = mask;
2007
2008 if (reset_key)
2009 memset(key, 0, sizeof(*key));
2010
2011 if (mask) {
2012 memset(&mask->key, 0, sizeof(mask->key));
2013 mask->range.start = mask->range.end = 0;
2014 }
2015 }
2016
validate_geneve_opts(struct sw_flow_key * key)2017 static int validate_geneve_opts(struct sw_flow_key *key)
2018 {
2019 struct geneve_opt *option;
2020 int opts_len = key->tun_opts_len;
2021 bool crit_opt = false;
2022
2023 option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2024 while (opts_len > 0) {
2025 int len;
2026
2027 if (opts_len < sizeof(*option))
2028 return -EINVAL;
2029
2030 len = sizeof(*option) + option->length * 4;
2031 if (len > opts_len)
2032 return -EINVAL;
2033
2034 crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2035
2036 option = (struct geneve_opt *)((u8 *)option + len);
2037 opts_len -= len;
2038 };
2039
2040 key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
2041
2042 return 0;
2043 }
2044
validate_and_copy_set_tun(const struct nlattr * attr,struct sw_flow_actions ** sfa,bool log)2045 static int validate_and_copy_set_tun(const struct nlattr *attr,
2046 struct sw_flow_actions **sfa, bool log)
2047 {
2048 struct sw_flow_match match;
2049 struct sw_flow_key key;
2050 struct metadata_dst *tun_dst;
2051 struct ip_tunnel_info *tun_info;
2052 struct ovs_tunnel_info *ovs_tun;
2053 struct nlattr *a;
2054 int err = 0, start, opts_type;
2055
2056 ovs_match_init(&match, &key, true, NULL);
2057 opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2058 if (opts_type < 0)
2059 return opts_type;
2060
2061 if (key.tun_opts_len) {
2062 switch (opts_type) {
2063 case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2064 err = validate_geneve_opts(&key);
2065 if (err < 0)
2066 return err;
2067 break;
2068 case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
2069 break;
2070 }
2071 };
2072
2073 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2074 if (start < 0)
2075 return start;
2076
2077 tun_dst = metadata_dst_alloc(key.tun_opts_len, GFP_KERNEL);
2078 if (!tun_dst)
2079 return -ENOMEM;
2080
2081 err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2082 if (err) {
2083 dst_release((struct dst_entry *)tun_dst);
2084 return err;
2085 }
2086
2087 a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2088 sizeof(*ovs_tun), log);
2089 if (IS_ERR(a)) {
2090 dst_release((struct dst_entry *)tun_dst);
2091 return PTR_ERR(a);
2092 }
2093
2094 ovs_tun = nla_data(a);
2095 ovs_tun->tun_dst = tun_dst;
2096
2097 tun_info = &tun_dst->u.tun_info;
2098 tun_info->mode = IP_TUNNEL_INFO_TX;
2099 if (key.tun_proto == AF_INET6)
2100 tun_info->mode |= IP_TUNNEL_INFO_IPV6;
2101 tun_info->key = key.tun_key;
2102
2103 /* We need to store the options in the action itself since
2104 * everything else will go away after flow setup. We can append
2105 * it to tun_info and then point there.
2106 */
2107 ip_tunnel_info_opts_set(tun_info,
2108 TUN_METADATA_OPTS(&key, key.tun_opts_len),
2109 key.tun_opts_len);
2110 add_nested_action_end(*sfa, start);
2111
2112 return err;
2113 }
2114
2115 /* Return false if there are any non-masked bits set.
2116 * Mask follows data immediately, before any netlink padding.
2117 */
validate_masked(u8 * data,int len)2118 static bool validate_masked(u8 *data, int len)
2119 {
2120 u8 *mask = data + len;
2121
2122 while (len--)
2123 if (*data++ & ~*mask++)
2124 return false;
2125
2126 return true;
2127 }
2128
validate_set(const struct nlattr * a,const struct sw_flow_key * flow_key,struct sw_flow_actions ** sfa,bool * skip_copy,__be16 eth_type,bool masked,bool log)2129 static int validate_set(const struct nlattr *a,
2130 const struct sw_flow_key *flow_key,
2131 struct sw_flow_actions **sfa,
2132 bool *skip_copy, __be16 eth_type, bool masked, bool log)
2133 {
2134 const struct nlattr *ovs_key = nla_data(a);
2135 int key_type = nla_type(ovs_key);
2136 size_t key_len;
2137
2138 /* There can be only one key in a action */
2139 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2140 return -EINVAL;
2141
2142 key_len = nla_len(ovs_key);
2143 if (masked)
2144 key_len /= 2;
2145
2146 if (key_type > OVS_KEY_ATTR_MAX ||
2147 !check_attr_len(key_len, ovs_key_lens[key_type].len))
2148 return -EINVAL;
2149
2150 if (masked && !validate_masked(nla_data(ovs_key), key_len))
2151 return -EINVAL;
2152
2153 switch (key_type) {
2154 const struct ovs_key_ipv4 *ipv4_key;
2155 const struct ovs_key_ipv6 *ipv6_key;
2156 int err;
2157
2158 case OVS_KEY_ATTR_PRIORITY:
2159 case OVS_KEY_ATTR_SKB_MARK:
2160 case OVS_KEY_ATTR_CT_MARK:
2161 case OVS_KEY_ATTR_CT_LABELS:
2162 case OVS_KEY_ATTR_ETHERNET:
2163 break;
2164
2165 case OVS_KEY_ATTR_TUNNEL:
2166 if (masked)
2167 return -EINVAL; /* Masked tunnel set not supported. */
2168
2169 *skip_copy = true;
2170 err = validate_and_copy_set_tun(a, sfa, log);
2171 if (err)
2172 return err;
2173 break;
2174
2175 case OVS_KEY_ATTR_IPV4:
2176 if (eth_type != htons(ETH_P_IP))
2177 return -EINVAL;
2178
2179 ipv4_key = nla_data(ovs_key);
2180
2181 if (masked) {
2182 const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2183
2184 /* Non-writeable fields. */
2185 if (mask->ipv4_proto || mask->ipv4_frag)
2186 return -EINVAL;
2187 } else {
2188 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2189 return -EINVAL;
2190
2191 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2192 return -EINVAL;
2193 }
2194 break;
2195
2196 case OVS_KEY_ATTR_IPV6:
2197 if (eth_type != htons(ETH_P_IPV6))
2198 return -EINVAL;
2199
2200 ipv6_key = nla_data(ovs_key);
2201
2202 if (masked) {
2203 const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2204
2205 /* Non-writeable fields. */
2206 if (mask->ipv6_proto || mask->ipv6_frag)
2207 return -EINVAL;
2208
2209 /* Invalid bits in the flow label mask? */
2210 if (ntohl(mask->ipv6_label) & 0xFFF00000)
2211 return -EINVAL;
2212 } else {
2213 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2214 return -EINVAL;
2215
2216 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2217 return -EINVAL;
2218 }
2219 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2220 return -EINVAL;
2221
2222 break;
2223
2224 case OVS_KEY_ATTR_TCP:
2225 if ((eth_type != htons(ETH_P_IP) &&
2226 eth_type != htons(ETH_P_IPV6)) ||
2227 flow_key->ip.proto != IPPROTO_TCP)
2228 return -EINVAL;
2229
2230 break;
2231
2232 case OVS_KEY_ATTR_UDP:
2233 if ((eth_type != htons(ETH_P_IP) &&
2234 eth_type != htons(ETH_P_IPV6)) ||
2235 flow_key->ip.proto != IPPROTO_UDP)
2236 return -EINVAL;
2237
2238 break;
2239
2240 case OVS_KEY_ATTR_MPLS:
2241 if (!eth_p_mpls(eth_type))
2242 return -EINVAL;
2243 break;
2244
2245 case OVS_KEY_ATTR_SCTP:
2246 if ((eth_type != htons(ETH_P_IP) &&
2247 eth_type != htons(ETH_P_IPV6)) ||
2248 flow_key->ip.proto != IPPROTO_SCTP)
2249 return -EINVAL;
2250
2251 break;
2252
2253 default:
2254 return -EINVAL;
2255 }
2256
2257 /* Convert non-masked non-tunnel set actions to masked set actions. */
2258 if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2259 int start, len = key_len * 2;
2260 struct nlattr *at;
2261
2262 *skip_copy = true;
2263
2264 start = add_nested_action_start(sfa,
2265 OVS_ACTION_ATTR_SET_TO_MASKED,
2266 log);
2267 if (start < 0)
2268 return start;
2269
2270 at = __add_action(sfa, key_type, NULL, len, log);
2271 if (IS_ERR(at))
2272 return PTR_ERR(at);
2273
2274 memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2275 memset(nla_data(at) + key_len, 0xff, key_len); /* Mask. */
2276 /* Clear non-writeable bits from otherwise writeable fields. */
2277 if (key_type == OVS_KEY_ATTR_IPV6) {
2278 struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2279
2280 mask->ipv6_label &= htonl(0x000FFFFF);
2281 }
2282 add_nested_action_end(*sfa, start);
2283 }
2284
2285 return 0;
2286 }
2287
validate_userspace(const struct nlattr * attr)2288 static int validate_userspace(const struct nlattr *attr)
2289 {
2290 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2291 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2292 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2293 [OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2294 };
2295 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2296 int error;
2297
2298 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
2299 attr, userspace_policy);
2300 if (error)
2301 return error;
2302
2303 if (!a[OVS_USERSPACE_ATTR_PID] ||
2304 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2305 return -EINVAL;
2306
2307 return 0;
2308 }
2309
copy_action(const struct nlattr * from,struct sw_flow_actions ** sfa,bool log)2310 static int copy_action(const struct nlattr *from,
2311 struct sw_flow_actions **sfa, bool log)
2312 {
2313 int totlen = NLA_ALIGN(from->nla_len);
2314 struct nlattr *to;
2315
2316 to = reserve_sfa_size(sfa, from->nla_len, log);
2317 if (IS_ERR(to))
2318 return PTR_ERR(to);
2319
2320 memcpy(to, from, totlen);
2321 return 0;
2322 }
2323
__ovs_nla_copy_actions(struct net * net,const struct nlattr * attr,const struct sw_flow_key * key,int depth,struct sw_flow_actions ** sfa,__be16 eth_type,__be16 vlan_tci,bool log)2324 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2325 const struct sw_flow_key *key,
2326 int depth, struct sw_flow_actions **sfa,
2327 __be16 eth_type, __be16 vlan_tci, bool log)
2328 {
2329 const struct nlattr *a;
2330 int rem, err;
2331
2332 if (depth >= SAMPLE_ACTION_DEPTH)
2333 return -EOVERFLOW;
2334
2335 nla_for_each_nested(a, attr, rem) {
2336 /* Expected argument lengths, (u32)-1 for variable length. */
2337 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2338 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2339 [OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2340 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2341 [OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2342 [OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2343 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2344 [OVS_ACTION_ATTR_POP_VLAN] = 0,
2345 [OVS_ACTION_ATTR_SET] = (u32)-1,
2346 [OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2347 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2348 [OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2349 [OVS_ACTION_ATTR_CT] = (u32)-1,
2350 [OVS_ACTION_ATTR_TRUNC] = sizeof(struct ovs_action_trunc),
2351 };
2352 const struct ovs_action_push_vlan *vlan;
2353 int type = nla_type(a);
2354 bool skip_copy;
2355
2356 if (type > OVS_ACTION_ATTR_MAX ||
2357 (action_lens[type] != nla_len(a) &&
2358 action_lens[type] != (u32)-1))
2359 return -EINVAL;
2360
2361 skip_copy = false;
2362 switch (type) {
2363 case OVS_ACTION_ATTR_UNSPEC:
2364 return -EINVAL;
2365
2366 case OVS_ACTION_ATTR_USERSPACE:
2367 err = validate_userspace(a);
2368 if (err)
2369 return err;
2370 break;
2371
2372 case OVS_ACTION_ATTR_OUTPUT:
2373 if (nla_get_u32(a) >= DP_MAX_PORTS)
2374 return -EINVAL;
2375 break;
2376
2377 case OVS_ACTION_ATTR_TRUNC: {
2378 const struct ovs_action_trunc *trunc = nla_data(a);
2379
2380 if (trunc->max_len < ETH_HLEN)
2381 return -EINVAL;
2382 break;
2383 }
2384
2385 case OVS_ACTION_ATTR_HASH: {
2386 const struct ovs_action_hash *act_hash = nla_data(a);
2387
2388 switch (act_hash->hash_alg) {
2389 case OVS_HASH_ALG_L4:
2390 break;
2391 default:
2392 return -EINVAL;
2393 }
2394
2395 break;
2396 }
2397
2398 case OVS_ACTION_ATTR_POP_VLAN:
2399 vlan_tci = htons(0);
2400 break;
2401
2402 case OVS_ACTION_ATTR_PUSH_VLAN:
2403 vlan = nla_data(a);
2404 if (!eth_type_vlan(vlan->vlan_tpid))
2405 return -EINVAL;
2406 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2407 return -EINVAL;
2408 vlan_tci = vlan->vlan_tci;
2409 break;
2410
2411 case OVS_ACTION_ATTR_RECIRC:
2412 break;
2413
2414 case OVS_ACTION_ATTR_PUSH_MPLS: {
2415 const struct ovs_action_push_mpls *mpls = nla_data(a);
2416
2417 if (!eth_p_mpls(mpls->mpls_ethertype))
2418 return -EINVAL;
2419 /* Prohibit push MPLS other than to a white list
2420 * for packets that have a known tag order.
2421 */
2422 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2423 (eth_type != htons(ETH_P_IP) &&
2424 eth_type != htons(ETH_P_IPV6) &&
2425 eth_type != htons(ETH_P_ARP) &&
2426 eth_type != htons(ETH_P_RARP) &&
2427 !eth_p_mpls(eth_type)))
2428 return -EINVAL;
2429 eth_type = mpls->mpls_ethertype;
2430 break;
2431 }
2432
2433 case OVS_ACTION_ATTR_POP_MPLS:
2434 if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2435 !eth_p_mpls(eth_type))
2436 return -EINVAL;
2437
2438 /* Disallow subsequent L2.5+ set and mpls_pop actions
2439 * as there is no check here to ensure that the new
2440 * eth_type is valid and thus set actions could
2441 * write off the end of the packet or otherwise
2442 * corrupt it.
2443 *
2444 * Support for these actions is planned using packet
2445 * recirculation.
2446 */
2447 eth_type = htons(0);
2448 break;
2449
2450 case OVS_ACTION_ATTR_SET:
2451 err = validate_set(a, key, sfa,
2452 &skip_copy, eth_type, false, log);
2453 if (err)
2454 return err;
2455 break;
2456
2457 case OVS_ACTION_ATTR_SET_MASKED:
2458 err = validate_set(a, key, sfa,
2459 &skip_copy, eth_type, true, log);
2460 if (err)
2461 return err;
2462 break;
2463
2464 case OVS_ACTION_ATTR_SAMPLE:
2465 err = validate_and_copy_sample(net, a, key, depth, sfa,
2466 eth_type, vlan_tci, log);
2467 if (err)
2468 return err;
2469 skip_copy = true;
2470 break;
2471
2472 case OVS_ACTION_ATTR_CT:
2473 err = ovs_ct_copy_action(net, a, key, sfa, log);
2474 if (err)
2475 return err;
2476 skip_copy = true;
2477 break;
2478
2479 default:
2480 OVS_NLERR(log, "Unknown Action type %d", type);
2481 return -EINVAL;
2482 }
2483 if (!skip_copy) {
2484 err = copy_action(a, sfa, log);
2485 if (err)
2486 return err;
2487 }
2488 }
2489
2490 if (rem > 0)
2491 return -EINVAL;
2492
2493 return 0;
2494 }
2495
2496 /* '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)2497 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2498 const struct sw_flow_key *key,
2499 struct sw_flow_actions **sfa, bool log)
2500 {
2501 int err;
2502
2503 *sfa = nla_alloc_flow_actions(min(nla_len(attr), MAX_ACTIONS_BUFSIZE));
2504 if (IS_ERR(*sfa))
2505 return PTR_ERR(*sfa);
2506
2507 (*sfa)->orig_len = nla_len(attr);
2508 err = __ovs_nla_copy_actions(net, attr, key, 0, sfa, key->eth.type,
2509 key->eth.vlan.tci, log);
2510 if (err)
2511 ovs_nla_free_flow_actions(*sfa);
2512
2513 return err;
2514 }
2515
sample_action_to_attr(const struct nlattr * attr,struct sk_buff * skb)2516 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
2517 {
2518 const struct nlattr *a;
2519 struct nlattr *start;
2520 int err = 0, rem;
2521
2522 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
2523 if (!start)
2524 return -EMSGSIZE;
2525
2526 nla_for_each_nested(a, attr, rem) {
2527 int type = nla_type(a);
2528 struct nlattr *st_sample;
2529
2530 switch (type) {
2531 case OVS_SAMPLE_ATTR_PROBABILITY:
2532 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
2533 sizeof(u32), nla_data(a)))
2534 return -EMSGSIZE;
2535 break;
2536 case OVS_SAMPLE_ATTR_ACTIONS:
2537 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
2538 if (!st_sample)
2539 return -EMSGSIZE;
2540 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
2541 if (err)
2542 return err;
2543 nla_nest_end(skb, st_sample);
2544 break;
2545 }
2546 }
2547
2548 nla_nest_end(skb, start);
2549 return err;
2550 }
2551
set_action_to_attr(const struct nlattr * a,struct sk_buff * skb)2552 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
2553 {
2554 const struct nlattr *ovs_key = nla_data(a);
2555 int key_type = nla_type(ovs_key);
2556 struct nlattr *start;
2557 int err;
2558
2559 switch (key_type) {
2560 case OVS_KEY_ATTR_TUNNEL_INFO: {
2561 struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
2562 struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
2563
2564 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2565 if (!start)
2566 return -EMSGSIZE;
2567
2568 err = ip_tun_to_nlattr(skb, &tun_info->key,
2569 ip_tunnel_info_opts(tun_info),
2570 tun_info->options_len,
2571 ip_tunnel_info_af(tun_info));
2572 if (err)
2573 return err;
2574 nla_nest_end(skb, start);
2575 break;
2576 }
2577 default:
2578 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
2579 return -EMSGSIZE;
2580 break;
2581 }
2582
2583 return 0;
2584 }
2585
masked_set_action_to_set_action_attr(const struct nlattr * a,struct sk_buff * skb)2586 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
2587 struct sk_buff *skb)
2588 {
2589 const struct nlattr *ovs_key = nla_data(a);
2590 struct nlattr *nla;
2591 size_t key_len = nla_len(ovs_key) / 2;
2592
2593 /* Revert the conversion we did from a non-masked set action to
2594 * masked set action.
2595 */
2596 nla = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2597 if (!nla)
2598 return -EMSGSIZE;
2599
2600 if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
2601 return -EMSGSIZE;
2602
2603 nla_nest_end(skb, nla);
2604 return 0;
2605 }
2606
ovs_nla_put_actions(const struct nlattr * attr,int len,struct sk_buff * skb)2607 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
2608 {
2609 const struct nlattr *a;
2610 int rem, err;
2611
2612 nla_for_each_attr(a, attr, len, rem) {
2613 int type = nla_type(a);
2614
2615 switch (type) {
2616 case OVS_ACTION_ATTR_SET:
2617 err = set_action_to_attr(a, skb);
2618 if (err)
2619 return err;
2620 break;
2621
2622 case OVS_ACTION_ATTR_SET_TO_MASKED:
2623 err = masked_set_action_to_set_action_attr(a, skb);
2624 if (err)
2625 return err;
2626 break;
2627
2628 case OVS_ACTION_ATTR_SAMPLE:
2629 err = sample_action_to_attr(a, skb);
2630 if (err)
2631 return err;
2632 break;
2633
2634 case OVS_ACTION_ATTR_CT:
2635 err = ovs_ct_action_to_attr(nla_data(a), skb);
2636 if (err)
2637 return err;
2638 break;
2639
2640 default:
2641 if (nla_put(skb, type, nla_len(a), nla_data(a)))
2642 return -EMSGSIZE;
2643 break;
2644 }
2645 }
2646
2647 return 0;
2648 }
2649