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