aboutsummaryrefslogtreecommitdiffstats
path: root/net/openvswitch/flow_netlink.c
diff options
context:
space:
mode:
Diffstat (limited to 'net/openvswitch/flow_netlink.c')
-rw-r--r--net/openvswitch/flow_netlink.c1603
1 files changed, 1603 insertions, 0 deletions
diff --git a/net/openvswitch/flow_netlink.c b/net/openvswitch/flow_netlink.c
new file mode 100644
index 000000000000..e04649c56a96
--- /dev/null
+++ b/net/openvswitch/flow_netlink.c
@@ -0,0 +1,1603 @@
1/*
2 * Copyright (c) 2007-2013 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#include "flow.h"
20#include "datapath.h"
21#include <linux/uaccess.h>
22#include <linux/netdevice.h>
23#include <linux/etherdevice.h>
24#include <linux/if_ether.h>
25#include <linux/if_vlan.h>
26#include <net/llc_pdu.h>
27#include <linux/kernel.h>
28#include <linux/jhash.h>
29#include <linux/jiffies.h>
30#include <linux/llc.h>
31#include <linux/module.h>
32#include <linux/in.h>
33#include <linux/rcupdate.h>
34#include <linux/if_arp.h>
35#include <linux/ip.h>
36#include <linux/ipv6.h>
37#include <linux/sctp.h>
38#include <linux/tcp.h>
39#include <linux/udp.h>
40#include <linux/icmp.h>
41#include <linux/icmpv6.h>
42#include <linux/rculist.h>
43#include <net/ip.h>
44#include <net/ipv6.h>
45#include <net/ndisc.h>
46
47#include "flow_netlink.h"
48
49static void update_range__(struct sw_flow_match *match,
50 size_t offset, size_t size, bool is_mask)
51{
52 struct sw_flow_key_range *range = NULL;
53 size_t start = rounddown(offset, sizeof(long));
54 size_t end = roundup(offset + size, sizeof(long));
55
56 if (!is_mask)
57 range = &match->range;
58 else if (match->mask)
59 range = &match->mask->range;
60
61 if (!range)
62 return;
63
64 if (range->start == range->end) {
65 range->start = start;
66 range->end = end;
67 return;
68 }
69
70 if (range->start > start)
71 range->start = start;
72
73 if (range->end < end)
74 range->end = end;
75}
76
77#define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
78 do { \
79 update_range__(match, offsetof(struct sw_flow_key, field), \
80 sizeof((match)->key->field), is_mask); \
81 if (is_mask) { \
82 if ((match)->mask) \
83 (match)->mask->key.field = value; \
84 } else { \
85 (match)->key->field = value; \
86 } \
87 } while (0)
88
89#define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
90 do { \
91 update_range__(match, offsetof(struct sw_flow_key, field), \
92 len, is_mask); \
93 if (is_mask) { \
94 if ((match)->mask) \
95 memcpy(&(match)->mask->key.field, value_p, len);\
96 } else { \
97 memcpy(&(match)->key->field, value_p, len); \
98 } \
99 } while (0)
100
101static u16 range_n_bytes(const struct sw_flow_key_range *range)
102{
103 return range->end - range->start;
104}
105
106static bool match_validate(const struct sw_flow_match *match,
107 u64 key_attrs, u64 mask_attrs)
108{
109 u64 key_expected = 1 << OVS_KEY_ATTR_ETHERNET;
110 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
111
112 /* The following mask attributes allowed only if they
113 * pass the validation tests. */
114 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
115 | (1 << OVS_KEY_ATTR_IPV6)
116 | (1 << OVS_KEY_ATTR_TCP)
117 | (1 << OVS_KEY_ATTR_UDP)
118 | (1 << OVS_KEY_ATTR_SCTP)
119 | (1 << OVS_KEY_ATTR_ICMP)
120 | (1 << OVS_KEY_ATTR_ICMPV6)
121 | (1 << OVS_KEY_ATTR_ARP)
122 | (1 << OVS_KEY_ATTR_ND));
123
124 /* Always allowed mask fields. */
125 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
126 | (1 << OVS_KEY_ATTR_IN_PORT)
127 | (1 << OVS_KEY_ATTR_ETHERTYPE));
128
129 /* Check key attributes. */
130 if (match->key->eth.type == htons(ETH_P_ARP)
131 || match->key->eth.type == htons(ETH_P_RARP)) {
132 key_expected |= 1 << OVS_KEY_ATTR_ARP;
133 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
134 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
135 }
136
137 if (match->key->eth.type == htons(ETH_P_IP)) {
138 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
139 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
140 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
141
142 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
143 if (match->key->ip.proto == IPPROTO_UDP) {
144 key_expected |= 1 << OVS_KEY_ATTR_UDP;
145 if (match->mask && (match->mask->key.ip.proto == 0xff))
146 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
147 }
148
149 if (match->key->ip.proto == IPPROTO_SCTP) {
150 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
151 if (match->mask && (match->mask->key.ip.proto == 0xff))
152 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
153 }
154
155 if (match->key->ip.proto == IPPROTO_TCP) {
156 key_expected |= 1 << OVS_KEY_ATTR_TCP;
157 if (match->mask && (match->mask->key.ip.proto == 0xff))
158 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
159 }
160
161 if (match->key->ip.proto == IPPROTO_ICMP) {
162 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
163 if (match->mask && (match->mask->key.ip.proto == 0xff))
164 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
165 }
166 }
167 }
168
169 if (match->key->eth.type == htons(ETH_P_IPV6)) {
170 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
171 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
172 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
173
174 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
175 if (match->key->ip.proto == IPPROTO_UDP) {
176 key_expected |= 1 << OVS_KEY_ATTR_UDP;
177 if (match->mask && (match->mask->key.ip.proto == 0xff))
178 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
179 }
180
181 if (match->key->ip.proto == IPPROTO_SCTP) {
182 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
183 if (match->mask && (match->mask->key.ip.proto == 0xff))
184 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
185 }
186
187 if (match->key->ip.proto == IPPROTO_TCP) {
188 key_expected |= 1 << OVS_KEY_ATTR_TCP;
189 if (match->mask && (match->mask->key.ip.proto == 0xff))
190 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
191 }
192
193 if (match->key->ip.proto == IPPROTO_ICMPV6) {
194 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
195 if (match->mask && (match->mask->key.ip.proto == 0xff))
196 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
197
198 if (match->key->ipv6.tp.src ==
199 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
200 match->key->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
201 key_expected |= 1 << OVS_KEY_ATTR_ND;
202 if (match->mask && (match->mask->key.ipv6.tp.src == htons(0xffff)))
203 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
204 }
205 }
206 }
207 }
208
209 if ((key_attrs & key_expected) != key_expected) {
210 /* Key attributes check failed. */
211 OVS_NLERR("Missing expected key attributes (key_attrs=%llx, expected=%llx).\n",
212 key_attrs, key_expected);
213 return false;
214 }
215
216 if ((mask_attrs & mask_allowed) != mask_attrs) {
217 /* Mask attributes check failed. */
218 OVS_NLERR("Contain more than allowed mask fields (mask_attrs=%llx, mask_allowed=%llx).\n",
219 mask_attrs, mask_allowed);
220 return false;
221 }
222
223 return true;
224}
225
226/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
227static const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
228 [OVS_KEY_ATTR_ENCAP] = -1,
229 [OVS_KEY_ATTR_PRIORITY] = sizeof(u32),
230 [OVS_KEY_ATTR_IN_PORT] = sizeof(u32),
231 [OVS_KEY_ATTR_SKB_MARK] = sizeof(u32),
232 [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
233 [OVS_KEY_ATTR_VLAN] = sizeof(__be16),
234 [OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16),
235 [OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4),
236 [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
237 [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
238 [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
239 [OVS_KEY_ATTR_SCTP] = sizeof(struct ovs_key_sctp),
240 [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
241 [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
242 [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
243 [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
244 [OVS_KEY_ATTR_TUNNEL] = -1,
245};
246
247static bool is_all_zero(const u8 *fp, size_t size)
248{
249 int i;
250
251 if (!fp)
252 return false;
253
254 for (i = 0; i < size; i++)
255 if (fp[i])
256 return false;
257
258 return true;
259}
260
261static int __parse_flow_nlattrs(const struct nlattr *attr,
262 const struct nlattr *a[],
263 u64 *attrsp, bool nz)
264{
265 const struct nlattr *nla;
266 u64 attrs;
267 int rem;
268
269 attrs = *attrsp;
270 nla_for_each_nested(nla, attr, rem) {
271 u16 type = nla_type(nla);
272 int expected_len;
273
274 if (type > OVS_KEY_ATTR_MAX) {
275 OVS_NLERR("Unknown key attribute (type=%d, max=%d).\n",
276 type, OVS_KEY_ATTR_MAX);
277 return -EINVAL;
278 }
279
280 if (attrs & (1 << type)) {
281 OVS_NLERR("Duplicate key attribute (type %d).\n", type);
282 return -EINVAL;
283 }
284
285 expected_len = ovs_key_lens[type];
286 if (nla_len(nla) != expected_len && expected_len != -1) {
287 OVS_NLERR("Key attribute has unexpected length (type=%d"
288 ", length=%d, expected=%d).\n", type,
289 nla_len(nla), expected_len);
290 return -EINVAL;
291 }
292
293 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
294 attrs |= 1 << type;
295 a[type] = nla;
296 }
297 }
298 if (rem) {
299 OVS_NLERR("Message has %d unknown bytes.\n", rem);
300 return -EINVAL;
301 }
302
303 *attrsp = attrs;
304 return 0;
305}
306
307static int parse_flow_mask_nlattrs(const struct nlattr *attr,
308 const struct nlattr *a[], u64 *attrsp)
309{
310 return __parse_flow_nlattrs(attr, a, attrsp, true);
311}
312
313static int parse_flow_nlattrs(const struct nlattr *attr,
314 const struct nlattr *a[], u64 *attrsp)
315{
316 return __parse_flow_nlattrs(attr, a, attrsp, false);
317}
318
319static int ipv4_tun_from_nlattr(const struct nlattr *attr,
320 struct sw_flow_match *match, bool is_mask)
321{
322 struct nlattr *a;
323 int rem;
324 bool ttl = false;
325 __be16 tun_flags = 0;
326
327 nla_for_each_nested(a, attr, rem) {
328 int type = nla_type(a);
329 static const u32 ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
330 [OVS_TUNNEL_KEY_ATTR_ID] = sizeof(u64),
331 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = sizeof(u32),
332 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = sizeof(u32),
333 [OVS_TUNNEL_KEY_ATTR_TOS] = 1,
334 [OVS_TUNNEL_KEY_ATTR_TTL] = 1,
335 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = 0,
336 [OVS_TUNNEL_KEY_ATTR_CSUM] = 0,
337 };
338
339 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
340 OVS_NLERR("Unknown IPv4 tunnel attribute (type=%d, max=%d).\n",
341 type, OVS_TUNNEL_KEY_ATTR_MAX);
342 return -EINVAL;
343 }
344
345 if (ovs_tunnel_key_lens[type] != nla_len(a)) {
346 OVS_NLERR("IPv4 tunnel attribute type has unexpected "
347 " length (type=%d, length=%d, expected=%d).\n",
348 type, nla_len(a), ovs_tunnel_key_lens[type]);
349 return -EINVAL;
350 }
351
352 switch (type) {
353 case OVS_TUNNEL_KEY_ATTR_ID:
354 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
355 nla_get_be64(a), is_mask);
356 tun_flags |= TUNNEL_KEY;
357 break;
358 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
359 SW_FLOW_KEY_PUT(match, tun_key.ipv4_src,
360 nla_get_be32(a), is_mask);
361 break;
362 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
363 SW_FLOW_KEY_PUT(match, tun_key.ipv4_dst,
364 nla_get_be32(a), is_mask);
365 break;
366 case OVS_TUNNEL_KEY_ATTR_TOS:
367 SW_FLOW_KEY_PUT(match, tun_key.ipv4_tos,
368 nla_get_u8(a), is_mask);
369 break;
370 case OVS_TUNNEL_KEY_ATTR_TTL:
371 SW_FLOW_KEY_PUT(match, tun_key.ipv4_ttl,
372 nla_get_u8(a), is_mask);
373 ttl = true;
374 break;
375 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
376 tun_flags |= TUNNEL_DONT_FRAGMENT;
377 break;
378 case OVS_TUNNEL_KEY_ATTR_CSUM:
379 tun_flags |= TUNNEL_CSUM;
380 break;
381 default:
382 return -EINVAL;
383 }
384 }
385
386 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
387
388 if (rem > 0) {
389 OVS_NLERR("IPv4 tunnel attribute has %d unknown bytes.\n", rem);
390 return -EINVAL;
391 }
392
393 if (!is_mask) {
394 if (!match->key->tun_key.ipv4_dst) {
395 OVS_NLERR("IPv4 tunnel destination address is zero.\n");
396 return -EINVAL;
397 }
398
399 if (!ttl) {
400 OVS_NLERR("IPv4 tunnel TTL not specified.\n");
401 return -EINVAL;
402 }
403 }
404
405 return 0;
406}
407
408static int ipv4_tun_to_nlattr(struct sk_buff *skb,
409 const struct ovs_key_ipv4_tunnel *tun_key,
410 const struct ovs_key_ipv4_tunnel *output)
411{
412 struct nlattr *nla;
413
414 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
415 if (!nla)
416 return -EMSGSIZE;
417
418 if (output->tun_flags & TUNNEL_KEY &&
419 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
420 return -EMSGSIZE;
421 if (output->ipv4_src &&
422 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC, output->ipv4_src))
423 return -EMSGSIZE;
424 if (output->ipv4_dst &&
425 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST, output->ipv4_dst))
426 return -EMSGSIZE;
427 if (output->ipv4_tos &&
428 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->ipv4_tos))
429 return -EMSGSIZE;
430 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ipv4_ttl))
431 return -EMSGSIZE;
432 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
433 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
434 return -EMSGSIZE;
435 if ((output->tun_flags & TUNNEL_CSUM) &&
436 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
437 return -EMSGSIZE;
438
439 nla_nest_end(skb, nla);
440 return 0;
441}
442
443
444static int metadata_from_nlattrs(struct sw_flow_match *match, u64 *attrs,
445 const struct nlattr **a, bool is_mask)
446{
447 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
448 SW_FLOW_KEY_PUT(match, phy.priority,
449 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
450 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
451 }
452
453 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
454 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
455
456 if (is_mask)
457 in_port = 0xffffffff; /* Always exact match in_port. */
458 else if (in_port >= DP_MAX_PORTS)
459 return -EINVAL;
460
461 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
462 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
463 } else if (!is_mask) {
464 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
465 }
466
467 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
468 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
469
470 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
471 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
472 }
473 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
474 if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
475 is_mask))
476 return -EINVAL;
477 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
478 }
479 return 0;
480}
481
482static int ovs_key_from_nlattrs(struct sw_flow_match *match, u64 attrs,
483 const struct nlattr **a, bool is_mask)
484{
485 int err;
486 u64 orig_attrs = attrs;
487
488 err = metadata_from_nlattrs(match, &attrs, a, is_mask);
489 if (err)
490 return err;
491
492 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
493 const struct ovs_key_ethernet *eth_key;
494
495 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
496 SW_FLOW_KEY_MEMCPY(match, eth.src,
497 eth_key->eth_src, ETH_ALEN, is_mask);
498 SW_FLOW_KEY_MEMCPY(match, eth.dst,
499 eth_key->eth_dst, ETH_ALEN, is_mask);
500 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
501 }
502
503 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
504 __be16 tci;
505
506 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
507 if (!(tci & htons(VLAN_TAG_PRESENT))) {
508 if (is_mask)
509 OVS_NLERR("VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.\n");
510 else
511 OVS_NLERR("VLAN TCI does not have VLAN_TAG_PRESENT bit set.\n");
512
513 return -EINVAL;
514 }
515
516 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
517 attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
518 } else if (!is_mask)
519 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
520
521 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
522 __be16 eth_type;
523
524 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
525 if (is_mask) {
526 /* Always exact match EtherType. */
527 eth_type = htons(0xffff);
528 } else if (ntohs(eth_type) < ETH_P_802_3_MIN) {
529 OVS_NLERR("EtherType is less than minimum (type=%x, min=%x).\n",
530 ntohs(eth_type), ETH_P_802_3_MIN);
531 return -EINVAL;
532 }
533
534 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
535 attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
536 } else if (!is_mask) {
537 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
538 }
539
540 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
541 const struct ovs_key_ipv4 *ipv4_key;
542
543 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
544 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
545 OVS_NLERR("Unknown IPv4 fragment type (value=%d, max=%d).\n",
546 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
547 return -EINVAL;
548 }
549 SW_FLOW_KEY_PUT(match, ip.proto,
550 ipv4_key->ipv4_proto, is_mask);
551 SW_FLOW_KEY_PUT(match, ip.tos,
552 ipv4_key->ipv4_tos, is_mask);
553 SW_FLOW_KEY_PUT(match, ip.ttl,
554 ipv4_key->ipv4_ttl, is_mask);
555 SW_FLOW_KEY_PUT(match, ip.frag,
556 ipv4_key->ipv4_frag, is_mask);
557 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
558 ipv4_key->ipv4_src, is_mask);
559 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
560 ipv4_key->ipv4_dst, is_mask);
561 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
562 }
563
564 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
565 const struct ovs_key_ipv6 *ipv6_key;
566
567 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
568 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
569 OVS_NLERR("Unknown IPv6 fragment type (value=%d, max=%d).\n",
570 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
571 return -EINVAL;
572 }
573 SW_FLOW_KEY_PUT(match, ipv6.label,
574 ipv6_key->ipv6_label, is_mask);
575 SW_FLOW_KEY_PUT(match, ip.proto,
576 ipv6_key->ipv6_proto, is_mask);
577 SW_FLOW_KEY_PUT(match, ip.tos,
578 ipv6_key->ipv6_tclass, is_mask);
579 SW_FLOW_KEY_PUT(match, ip.ttl,
580 ipv6_key->ipv6_hlimit, is_mask);
581 SW_FLOW_KEY_PUT(match, ip.frag,
582 ipv6_key->ipv6_frag, is_mask);
583 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
584 ipv6_key->ipv6_src,
585 sizeof(match->key->ipv6.addr.src),
586 is_mask);
587 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
588 ipv6_key->ipv6_dst,
589 sizeof(match->key->ipv6.addr.dst),
590 is_mask);
591
592 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
593 }
594
595 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
596 const struct ovs_key_arp *arp_key;
597
598 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
599 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
600 OVS_NLERR("Unknown ARP opcode (opcode=%d).\n",
601 arp_key->arp_op);
602 return -EINVAL;
603 }
604
605 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
606 arp_key->arp_sip, is_mask);
607 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
608 arp_key->arp_tip, is_mask);
609 SW_FLOW_KEY_PUT(match, ip.proto,
610 ntohs(arp_key->arp_op), is_mask);
611 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
612 arp_key->arp_sha, ETH_ALEN, is_mask);
613 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
614 arp_key->arp_tha, ETH_ALEN, is_mask);
615
616 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
617 }
618
619 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
620 const struct ovs_key_tcp *tcp_key;
621
622 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
623 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
624 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
625 tcp_key->tcp_src, is_mask);
626 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
627 tcp_key->tcp_dst, is_mask);
628 } else {
629 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
630 tcp_key->tcp_src, is_mask);
631 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
632 tcp_key->tcp_dst, is_mask);
633 }
634 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
635 }
636
637 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
638 const struct ovs_key_udp *udp_key;
639
640 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
641 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
642 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
643 udp_key->udp_src, is_mask);
644 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
645 udp_key->udp_dst, is_mask);
646 } else {
647 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
648 udp_key->udp_src, is_mask);
649 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
650 udp_key->udp_dst, is_mask);
651 }
652 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
653 }
654
655 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
656 const struct ovs_key_sctp *sctp_key;
657
658 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
659 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
660 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
661 sctp_key->sctp_src, is_mask);
662 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
663 sctp_key->sctp_dst, is_mask);
664 } else {
665 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
666 sctp_key->sctp_src, is_mask);
667 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
668 sctp_key->sctp_dst, is_mask);
669 }
670 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
671 }
672
673 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
674 const struct ovs_key_icmp *icmp_key;
675
676 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
677 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
678 htons(icmp_key->icmp_type), is_mask);
679 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
680 htons(icmp_key->icmp_code), is_mask);
681 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
682 }
683
684 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
685 const struct ovs_key_icmpv6 *icmpv6_key;
686
687 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
688 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
689 htons(icmpv6_key->icmpv6_type), is_mask);
690 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
691 htons(icmpv6_key->icmpv6_code), is_mask);
692 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
693 }
694
695 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
696 const struct ovs_key_nd *nd_key;
697
698 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
699 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
700 nd_key->nd_target,
701 sizeof(match->key->ipv6.nd.target),
702 is_mask);
703 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
704 nd_key->nd_sll, ETH_ALEN, is_mask);
705 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
706 nd_key->nd_tll, ETH_ALEN, is_mask);
707 attrs &= ~(1 << OVS_KEY_ATTR_ND);
708 }
709
710 if (attrs != 0)
711 return -EINVAL;
712
713 return 0;
714}
715
716static void sw_flow_mask_set(struct sw_flow_mask *mask,
717 struct sw_flow_key_range *range, u8 val)
718{
719 u8 *m = (u8 *)&mask->key + range->start;
720
721 mask->range = *range;
722 memset(m, val, range_n_bytes(range));
723}
724
725/**
726 * ovs_nla_get_match - parses Netlink attributes into a flow key and
727 * mask. In case the 'mask' is NULL, the flow is treated as exact match
728 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
729 * does not include any don't care bit.
730 * @match: receives the extracted flow match information.
731 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
732 * sequence. The fields should of the packet that triggered the creation
733 * of this flow.
734 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
735 * attribute specifies the mask field of the wildcarded flow.
736 */
737int ovs_nla_get_match(struct sw_flow_match *match,
738 const struct nlattr *key,
739 const struct nlattr *mask)
740{
741 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
742 const struct nlattr *encap;
743 u64 key_attrs = 0;
744 u64 mask_attrs = 0;
745 bool encap_valid = false;
746 int err;
747
748 err = parse_flow_nlattrs(key, a, &key_attrs);
749 if (err)
750 return err;
751
752 if ((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
753 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
754 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
755 __be16 tci;
756
757 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
758 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
759 OVS_NLERR("Invalid Vlan frame.\n");
760 return -EINVAL;
761 }
762
763 key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
764 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
765 encap = a[OVS_KEY_ATTR_ENCAP];
766 key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
767 encap_valid = true;
768
769 if (tci & htons(VLAN_TAG_PRESENT)) {
770 err = parse_flow_nlattrs(encap, a, &key_attrs);
771 if (err)
772 return err;
773 } else if (!tci) {
774 /* Corner case for truncated 802.1Q header. */
775 if (nla_len(encap)) {
776 OVS_NLERR("Truncated 802.1Q header has non-zero encap attribute.\n");
777 return -EINVAL;
778 }
779 } else {
780 OVS_NLERR("Encap attribute is set for a non-VLAN frame.\n");
781 return -EINVAL;
782 }
783 }
784
785 err = ovs_key_from_nlattrs(match, key_attrs, a, false);
786 if (err)
787 return err;
788
789 if (mask) {
790 err = parse_flow_mask_nlattrs(mask, a, &mask_attrs);
791 if (err)
792 return err;
793
794 if (mask_attrs & 1 << OVS_KEY_ATTR_ENCAP) {
795 __be16 eth_type = 0;
796 __be16 tci = 0;
797
798 if (!encap_valid) {
799 OVS_NLERR("Encap mask attribute is set for non-VLAN frame.\n");
800 return -EINVAL;
801 }
802
803 mask_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
804 if (a[OVS_KEY_ATTR_ETHERTYPE])
805 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
806
807 if (eth_type == htons(0xffff)) {
808 mask_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
809 encap = a[OVS_KEY_ATTR_ENCAP];
810 err = parse_flow_mask_nlattrs(encap, a, &mask_attrs);
811 } else {
812 OVS_NLERR("VLAN frames must have an exact match on the TPID (mask=%x).\n",
813 ntohs(eth_type));
814 return -EINVAL;
815 }
816
817 if (a[OVS_KEY_ATTR_VLAN])
818 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
819
820 if (!(tci & htons(VLAN_TAG_PRESENT))) {
821 OVS_NLERR("VLAN tag present bit must have an exact match (tci_mask=%x).\n", ntohs(tci));
822 return -EINVAL;
823 }
824 }
825
826 err = ovs_key_from_nlattrs(match, mask_attrs, a, true);
827 if (err)
828 return err;
829 } else {
830 /* Populate exact match flow's key mask. */
831 if (match->mask)
832 sw_flow_mask_set(match->mask, &match->range, 0xff);
833 }
834
835 if (!match_validate(match, key_attrs, mask_attrs))
836 return -EINVAL;
837
838 return 0;
839}
840
841/**
842 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
843 * @flow: Receives extracted in_port, priority, tun_key and skb_mark.
844 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
845 * sequence.
846 *
847 * This parses a series of Netlink attributes that form a flow key, which must
848 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
849 * get the metadata, that is, the parts of the flow key that cannot be
850 * extracted from the packet itself.
851 */
852
853int ovs_nla_get_flow_metadata(struct sw_flow *flow,
854 const struct nlattr *attr)
855{
856 struct ovs_key_ipv4_tunnel *tun_key = &flow->key.tun_key;
857 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
858 u64 attrs = 0;
859 int err;
860 struct sw_flow_match match;
861
862 flow->key.phy.in_port = DP_MAX_PORTS;
863 flow->key.phy.priority = 0;
864 flow->key.phy.skb_mark = 0;
865 memset(tun_key, 0, sizeof(flow->key.tun_key));
866
867 err = parse_flow_nlattrs(attr, a, &attrs);
868 if (err)
869 return -EINVAL;
870
871 memset(&match, 0, sizeof(match));
872 match.key = &flow->key;
873
874 err = metadata_from_nlattrs(&match, &attrs, a, false);
875 if (err)
876 return err;
877
878 return 0;
879}
880
881int ovs_nla_put_flow(const struct sw_flow_key *swkey,
882 const struct sw_flow_key *output, struct sk_buff *skb)
883{
884 struct ovs_key_ethernet *eth_key;
885 struct nlattr *nla, *encap;
886 bool is_mask = (swkey != output);
887
888 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
889 goto nla_put_failure;
890
891 if ((swkey->tun_key.ipv4_dst || is_mask) &&
892 ipv4_tun_to_nlattr(skb, &swkey->tun_key, &output->tun_key))
893 goto nla_put_failure;
894
895 if (swkey->phy.in_port == DP_MAX_PORTS) {
896 if (is_mask && (output->phy.in_port == 0xffff))
897 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
898 goto nla_put_failure;
899 } else {
900 u16 upper_u16;
901 upper_u16 = !is_mask ? 0 : 0xffff;
902
903 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
904 (upper_u16 << 16) | output->phy.in_port))
905 goto nla_put_failure;
906 }
907
908 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
909 goto nla_put_failure;
910
911 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
912 if (!nla)
913 goto nla_put_failure;
914
915 eth_key = nla_data(nla);
916 memcpy(eth_key->eth_src, output->eth.src, ETH_ALEN);
917 memcpy(eth_key->eth_dst, output->eth.dst, ETH_ALEN);
918
919 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
920 __be16 eth_type;
921 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
922 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
923 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
924 goto nla_put_failure;
925 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
926 if (!swkey->eth.tci)
927 goto unencap;
928 } else
929 encap = NULL;
930
931 if (swkey->eth.type == htons(ETH_P_802_2)) {
932 /*
933 * Ethertype 802.2 is represented in the netlink with omitted
934 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
935 * 0xffff in the mask attribute. Ethertype can also
936 * be wildcarded.
937 */
938 if (is_mask && output->eth.type)
939 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
940 output->eth.type))
941 goto nla_put_failure;
942 goto unencap;
943 }
944
945 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
946 goto nla_put_failure;
947
948 if (swkey->eth.type == htons(ETH_P_IP)) {
949 struct ovs_key_ipv4 *ipv4_key;
950
951 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
952 if (!nla)
953 goto nla_put_failure;
954 ipv4_key = nla_data(nla);
955 ipv4_key->ipv4_src = output->ipv4.addr.src;
956 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
957 ipv4_key->ipv4_proto = output->ip.proto;
958 ipv4_key->ipv4_tos = output->ip.tos;
959 ipv4_key->ipv4_ttl = output->ip.ttl;
960 ipv4_key->ipv4_frag = output->ip.frag;
961 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
962 struct ovs_key_ipv6 *ipv6_key;
963
964 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
965 if (!nla)
966 goto nla_put_failure;
967 ipv6_key = nla_data(nla);
968 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
969 sizeof(ipv6_key->ipv6_src));
970 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
971 sizeof(ipv6_key->ipv6_dst));
972 ipv6_key->ipv6_label = output->ipv6.label;
973 ipv6_key->ipv6_proto = output->ip.proto;
974 ipv6_key->ipv6_tclass = output->ip.tos;
975 ipv6_key->ipv6_hlimit = output->ip.ttl;
976 ipv6_key->ipv6_frag = output->ip.frag;
977 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
978 swkey->eth.type == htons(ETH_P_RARP)) {
979 struct ovs_key_arp *arp_key;
980
981 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
982 if (!nla)
983 goto nla_put_failure;
984 arp_key = nla_data(nla);
985 memset(arp_key, 0, sizeof(struct ovs_key_arp));
986 arp_key->arp_sip = output->ipv4.addr.src;
987 arp_key->arp_tip = output->ipv4.addr.dst;
988 arp_key->arp_op = htons(output->ip.proto);
989 memcpy(arp_key->arp_sha, output->ipv4.arp.sha, ETH_ALEN);
990 memcpy(arp_key->arp_tha, output->ipv4.arp.tha, ETH_ALEN);
991 }
992
993 if ((swkey->eth.type == htons(ETH_P_IP) ||
994 swkey->eth.type == htons(ETH_P_IPV6)) &&
995 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
996
997 if (swkey->ip.proto == IPPROTO_TCP) {
998 struct ovs_key_tcp *tcp_key;
999
1000 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1001 if (!nla)
1002 goto nla_put_failure;
1003 tcp_key = nla_data(nla);
1004 if (swkey->eth.type == htons(ETH_P_IP)) {
1005 tcp_key->tcp_src = output->ipv4.tp.src;
1006 tcp_key->tcp_dst = output->ipv4.tp.dst;
1007 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1008 tcp_key->tcp_src = output->ipv6.tp.src;
1009 tcp_key->tcp_dst = output->ipv6.tp.dst;
1010 }
1011 } else if (swkey->ip.proto == IPPROTO_UDP) {
1012 struct ovs_key_udp *udp_key;
1013
1014 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1015 if (!nla)
1016 goto nla_put_failure;
1017 udp_key = nla_data(nla);
1018 if (swkey->eth.type == htons(ETH_P_IP)) {
1019 udp_key->udp_src = output->ipv4.tp.src;
1020 udp_key->udp_dst = output->ipv4.tp.dst;
1021 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1022 udp_key->udp_src = output->ipv6.tp.src;
1023 udp_key->udp_dst = output->ipv6.tp.dst;
1024 }
1025 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1026 struct ovs_key_sctp *sctp_key;
1027
1028 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1029 if (!nla)
1030 goto nla_put_failure;
1031 sctp_key = nla_data(nla);
1032 if (swkey->eth.type == htons(ETH_P_IP)) {
1033 sctp_key->sctp_src = swkey->ipv4.tp.src;
1034 sctp_key->sctp_dst = swkey->ipv4.tp.dst;
1035 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1036 sctp_key->sctp_src = swkey->ipv6.tp.src;
1037 sctp_key->sctp_dst = swkey->ipv6.tp.dst;
1038 }
1039 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1040 swkey->ip.proto == IPPROTO_ICMP) {
1041 struct ovs_key_icmp *icmp_key;
1042
1043 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1044 if (!nla)
1045 goto nla_put_failure;
1046 icmp_key = nla_data(nla);
1047 icmp_key->icmp_type = ntohs(output->ipv4.tp.src);
1048 icmp_key->icmp_code = ntohs(output->ipv4.tp.dst);
1049 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1050 swkey->ip.proto == IPPROTO_ICMPV6) {
1051 struct ovs_key_icmpv6 *icmpv6_key;
1052
1053 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1054 sizeof(*icmpv6_key));
1055 if (!nla)
1056 goto nla_put_failure;
1057 icmpv6_key = nla_data(nla);
1058 icmpv6_key->icmpv6_type = ntohs(output->ipv6.tp.src);
1059 icmpv6_key->icmpv6_code = ntohs(output->ipv6.tp.dst);
1060
1061 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1062 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1063 struct ovs_key_nd *nd_key;
1064
1065 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1066 if (!nla)
1067 goto nla_put_failure;
1068 nd_key = nla_data(nla);
1069 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1070 sizeof(nd_key->nd_target));
1071 memcpy(nd_key->nd_sll, output->ipv6.nd.sll, ETH_ALEN);
1072 memcpy(nd_key->nd_tll, output->ipv6.nd.tll, ETH_ALEN);
1073 }
1074 }
1075 }
1076
1077unencap:
1078 if (encap)
1079 nla_nest_end(skb, encap);
1080
1081 return 0;
1082
1083nla_put_failure:
1084 return -EMSGSIZE;
1085}
1086
1087#define MAX_ACTIONS_BUFSIZE (32 * 1024)
1088
1089struct sw_flow_actions *ovs_nla_alloc_flow_actions(int size)
1090{
1091 struct sw_flow_actions *sfa;
1092
1093 if (size > MAX_ACTIONS_BUFSIZE)
1094 return ERR_PTR(-EINVAL);
1095
1096 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1097 if (!sfa)
1098 return ERR_PTR(-ENOMEM);
1099
1100 sfa->actions_len = 0;
1101 return sfa;
1102}
1103
1104/* RCU callback used by ovs_nla_free_flow_actions. */
1105static void rcu_free_acts_callback(struct rcu_head *rcu)
1106{
1107 struct sw_flow_actions *sf_acts = container_of(rcu,
1108 struct sw_flow_actions, rcu);
1109 kfree(sf_acts);
1110}
1111
1112/* Schedules 'sf_acts' to be freed after the next RCU grace period.
1113 * The caller must hold rcu_read_lock for this to be sensible. */
1114void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1115{
1116 call_rcu(&sf_acts->rcu, rcu_free_acts_callback);
1117}
1118
1119static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1120 int attr_len)
1121{
1122
1123 struct sw_flow_actions *acts;
1124 int new_acts_size;
1125 int req_size = NLA_ALIGN(attr_len);
1126 int next_offset = offsetof(struct sw_flow_actions, actions) +
1127 (*sfa)->actions_len;
1128
1129 if (req_size <= (ksize(*sfa) - next_offset))
1130 goto out;
1131
1132 new_acts_size = ksize(*sfa) * 2;
1133
1134 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1135 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1136 return ERR_PTR(-EMSGSIZE);
1137 new_acts_size = MAX_ACTIONS_BUFSIZE;
1138 }
1139
1140 acts = ovs_nla_alloc_flow_actions(new_acts_size);
1141 if (IS_ERR(acts))
1142 return (void *)acts;
1143
1144 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1145 acts->actions_len = (*sfa)->actions_len;
1146 kfree(*sfa);
1147 *sfa = acts;
1148
1149out:
1150 (*sfa)->actions_len += req_size;
1151 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1152}
1153
1154static int add_action(struct sw_flow_actions **sfa, int attrtype, void *data, int len)
1155{
1156 struct nlattr *a;
1157
1158 a = reserve_sfa_size(sfa, nla_attr_size(len));
1159 if (IS_ERR(a))
1160 return PTR_ERR(a);
1161
1162 a->nla_type = attrtype;
1163 a->nla_len = nla_attr_size(len);
1164
1165 if (data)
1166 memcpy(nla_data(a), data, len);
1167 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1168
1169 return 0;
1170}
1171
1172static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1173 int attrtype)
1174{
1175 int used = (*sfa)->actions_len;
1176 int err;
1177
1178 err = add_action(sfa, attrtype, NULL, 0);
1179 if (err)
1180 return err;
1181
1182 return used;
1183}
1184
1185static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1186 int st_offset)
1187{
1188 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1189 st_offset);
1190
1191 a->nla_len = sfa->actions_len - st_offset;
1192}
1193
1194static int validate_and_copy_sample(const struct nlattr *attr,
1195 const struct sw_flow_key *key, int depth,
1196 struct sw_flow_actions **sfa)
1197{
1198 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1199 const struct nlattr *probability, *actions;
1200 const struct nlattr *a;
1201 int rem, start, err, st_acts;
1202
1203 memset(attrs, 0, sizeof(attrs));
1204 nla_for_each_nested(a, attr, rem) {
1205 int type = nla_type(a);
1206 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1207 return -EINVAL;
1208 attrs[type] = a;
1209 }
1210 if (rem)
1211 return -EINVAL;
1212
1213 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1214 if (!probability || nla_len(probability) != sizeof(u32))
1215 return -EINVAL;
1216
1217 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1218 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1219 return -EINVAL;
1220
1221 /* validation done, copy sample action. */
1222 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE);
1223 if (start < 0)
1224 return start;
1225 err = add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1226 nla_data(probability), sizeof(u32));
1227 if (err)
1228 return err;
1229 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS);
1230 if (st_acts < 0)
1231 return st_acts;
1232
1233 err = ovs_nla_copy_actions(actions, key, depth + 1, sfa);
1234 if (err)
1235 return err;
1236
1237 add_nested_action_end(*sfa, st_acts);
1238 add_nested_action_end(*sfa, start);
1239
1240 return 0;
1241}
1242
1243static int validate_tp_port(const struct sw_flow_key *flow_key)
1244{
1245 if (flow_key->eth.type == htons(ETH_P_IP)) {
1246 if (flow_key->ipv4.tp.src || flow_key->ipv4.tp.dst)
1247 return 0;
1248 } else if (flow_key->eth.type == htons(ETH_P_IPV6)) {
1249 if (flow_key->ipv6.tp.src || flow_key->ipv6.tp.dst)
1250 return 0;
1251 }
1252
1253 return -EINVAL;
1254}
1255
1256void ovs_match_init(struct sw_flow_match *match,
1257 struct sw_flow_key *key,
1258 struct sw_flow_mask *mask)
1259{
1260 memset(match, 0, sizeof(*match));
1261 match->key = key;
1262 match->mask = mask;
1263
1264 memset(key, 0, sizeof(*key));
1265
1266 if (mask) {
1267 memset(&mask->key, 0, sizeof(mask->key));
1268 mask->range.start = mask->range.end = 0;
1269 }
1270}
1271
1272static int validate_and_copy_set_tun(const struct nlattr *attr,
1273 struct sw_flow_actions **sfa)
1274{
1275 struct sw_flow_match match;
1276 struct sw_flow_key key;
1277 int err, start;
1278
1279 ovs_match_init(&match, &key, NULL);
1280 err = ipv4_tun_from_nlattr(nla_data(attr), &match, false);
1281 if (err)
1282 return err;
1283
1284 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET);
1285 if (start < 0)
1286 return start;
1287
1288 err = add_action(sfa, OVS_KEY_ATTR_IPV4_TUNNEL, &match.key->tun_key,
1289 sizeof(match.key->tun_key));
1290 add_nested_action_end(*sfa, start);
1291
1292 return err;
1293}
1294
1295static int validate_set(const struct nlattr *a,
1296 const struct sw_flow_key *flow_key,
1297 struct sw_flow_actions **sfa,
1298 bool *set_tun)
1299{
1300 const struct nlattr *ovs_key = nla_data(a);
1301 int key_type = nla_type(ovs_key);
1302
1303 /* There can be only one key in a action */
1304 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1305 return -EINVAL;
1306
1307 if (key_type > OVS_KEY_ATTR_MAX ||
1308 (ovs_key_lens[key_type] != nla_len(ovs_key) &&
1309 ovs_key_lens[key_type] != -1))
1310 return -EINVAL;
1311
1312 switch (key_type) {
1313 const struct ovs_key_ipv4 *ipv4_key;
1314 const struct ovs_key_ipv6 *ipv6_key;
1315 int err;
1316
1317 case OVS_KEY_ATTR_PRIORITY:
1318 case OVS_KEY_ATTR_SKB_MARK:
1319 case OVS_KEY_ATTR_ETHERNET:
1320 break;
1321
1322 case OVS_KEY_ATTR_TUNNEL:
1323 *set_tun = true;
1324 err = validate_and_copy_set_tun(a, sfa);
1325 if (err)
1326 return err;
1327 break;
1328
1329 case OVS_KEY_ATTR_IPV4:
1330 if (flow_key->eth.type != htons(ETH_P_IP))
1331 return -EINVAL;
1332
1333 if (!flow_key->ip.proto)
1334 return -EINVAL;
1335
1336 ipv4_key = nla_data(ovs_key);
1337 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
1338 return -EINVAL;
1339
1340 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
1341 return -EINVAL;
1342
1343 break;
1344
1345 case OVS_KEY_ATTR_IPV6:
1346 if (flow_key->eth.type != htons(ETH_P_IPV6))
1347 return -EINVAL;
1348
1349 if (!flow_key->ip.proto)
1350 return -EINVAL;
1351
1352 ipv6_key = nla_data(ovs_key);
1353 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
1354 return -EINVAL;
1355
1356 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
1357 return -EINVAL;
1358
1359 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
1360 return -EINVAL;
1361
1362 break;
1363
1364 case OVS_KEY_ATTR_TCP:
1365 if (flow_key->ip.proto != IPPROTO_TCP)
1366 return -EINVAL;
1367
1368 return validate_tp_port(flow_key);
1369
1370 case OVS_KEY_ATTR_UDP:
1371 if (flow_key->ip.proto != IPPROTO_UDP)
1372 return -EINVAL;
1373
1374 return validate_tp_port(flow_key);
1375
1376 case OVS_KEY_ATTR_SCTP:
1377 if (flow_key->ip.proto != IPPROTO_SCTP)
1378 return -EINVAL;
1379
1380 return validate_tp_port(flow_key);
1381
1382 default:
1383 return -EINVAL;
1384 }
1385
1386 return 0;
1387}
1388
1389static int validate_userspace(const struct nlattr *attr)
1390{
1391 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
1392 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
1393 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
1394 };
1395 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
1396 int error;
1397
1398 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
1399 attr, userspace_policy);
1400 if (error)
1401 return error;
1402
1403 if (!a[OVS_USERSPACE_ATTR_PID] ||
1404 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
1405 return -EINVAL;
1406
1407 return 0;
1408}
1409
1410static int copy_action(const struct nlattr *from,
1411 struct sw_flow_actions **sfa)
1412{
1413 int totlen = NLA_ALIGN(from->nla_len);
1414 struct nlattr *to;
1415
1416 to = reserve_sfa_size(sfa, from->nla_len);
1417 if (IS_ERR(to))
1418 return PTR_ERR(to);
1419
1420 memcpy(to, from, totlen);
1421 return 0;
1422}
1423
1424int ovs_nla_copy_actions(const struct nlattr *attr,
1425 const struct sw_flow_key *key,
1426 int depth,
1427 struct sw_flow_actions **sfa)
1428{
1429 const struct nlattr *a;
1430 int rem, err;
1431
1432 if (depth >= SAMPLE_ACTION_DEPTH)
1433 return -EOVERFLOW;
1434
1435 nla_for_each_nested(a, attr, rem) {
1436 /* Expected argument lengths, (u32)-1 for variable length. */
1437 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
1438 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
1439 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
1440 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
1441 [OVS_ACTION_ATTR_POP_VLAN] = 0,
1442 [OVS_ACTION_ATTR_SET] = (u32)-1,
1443 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1
1444 };
1445 const struct ovs_action_push_vlan *vlan;
1446 int type = nla_type(a);
1447 bool skip_copy;
1448
1449 if (type > OVS_ACTION_ATTR_MAX ||
1450 (action_lens[type] != nla_len(a) &&
1451 action_lens[type] != (u32)-1))
1452 return -EINVAL;
1453
1454 skip_copy = false;
1455 switch (type) {
1456 case OVS_ACTION_ATTR_UNSPEC:
1457 return -EINVAL;
1458
1459 case OVS_ACTION_ATTR_USERSPACE:
1460 err = validate_userspace(a);
1461 if (err)
1462 return err;
1463 break;
1464
1465 case OVS_ACTION_ATTR_OUTPUT:
1466 if (nla_get_u32(a) >= DP_MAX_PORTS)
1467 return -EINVAL;
1468 break;
1469
1470
1471 case OVS_ACTION_ATTR_POP_VLAN:
1472 break;
1473
1474 case OVS_ACTION_ATTR_PUSH_VLAN:
1475 vlan = nla_data(a);
1476 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
1477 return -EINVAL;
1478 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
1479 return -EINVAL;
1480 break;
1481
1482 case OVS_ACTION_ATTR_SET:
1483 err = validate_set(a, key, sfa, &skip_copy);
1484 if (err)
1485 return err;
1486 break;
1487
1488 case OVS_ACTION_ATTR_SAMPLE:
1489 err = validate_and_copy_sample(a, key, depth, sfa);
1490 if (err)
1491 return err;
1492 skip_copy = true;
1493 break;
1494
1495 default:
1496 return -EINVAL;
1497 }
1498 if (!skip_copy) {
1499 err = copy_action(a, sfa);
1500 if (err)
1501 return err;
1502 }
1503 }
1504
1505 if (rem > 0)
1506 return -EINVAL;
1507
1508 return 0;
1509}
1510
1511static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
1512{
1513 const struct nlattr *a;
1514 struct nlattr *start;
1515 int err = 0, rem;
1516
1517 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
1518 if (!start)
1519 return -EMSGSIZE;
1520
1521 nla_for_each_nested(a, attr, rem) {
1522 int type = nla_type(a);
1523 struct nlattr *st_sample;
1524
1525 switch (type) {
1526 case OVS_SAMPLE_ATTR_PROBABILITY:
1527 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
1528 sizeof(u32), nla_data(a)))
1529 return -EMSGSIZE;
1530 break;
1531 case OVS_SAMPLE_ATTR_ACTIONS:
1532 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
1533 if (!st_sample)
1534 return -EMSGSIZE;
1535 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
1536 if (err)
1537 return err;
1538 nla_nest_end(skb, st_sample);
1539 break;
1540 }
1541 }
1542
1543 nla_nest_end(skb, start);
1544 return err;
1545}
1546
1547static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
1548{
1549 const struct nlattr *ovs_key = nla_data(a);
1550 int key_type = nla_type(ovs_key);
1551 struct nlattr *start;
1552 int err;
1553
1554 switch (key_type) {
1555 case OVS_KEY_ATTR_IPV4_TUNNEL:
1556 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
1557 if (!start)
1558 return -EMSGSIZE;
1559
1560 err = ipv4_tun_to_nlattr(skb, nla_data(ovs_key),
1561 nla_data(ovs_key));
1562 if (err)
1563 return err;
1564 nla_nest_end(skb, start);
1565 break;
1566 default:
1567 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
1568 return -EMSGSIZE;
1569 break;
1570 }
1571
1572 return 0;
1573}
1574
1575int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
1576{
1577 const struct nlattr *a;
1578 int rem, err;
1579
1580 nla_for_each_attr(a, attr, len, rem) {
1581 int type = nla_type(a);
1582
1583 switch (type) {
1584 case OVS_ACTION_ATTR_SET:
1585 err = set_action_to_attr(a, skb);
1586 if (err)
1587 return err;
1588 break;
1589
1590 case OVS_ACTION_ATTR_SAMPLE:
1591 err = sample_action_to_attr(a, skb);
1592 if (err)
1593 return err;
1594 break;
1595 default:
1596 if (nla_put(skb, type, nla_len(a), nla_data(a)))
1597 return -EMSGSIZE;
1598 break;
1599 }
1600 }
1601
1602 return 0;
1603}
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-20 00:14:44 -0500