1 /*
2  * Copyright (c) 2007-2014 Nicira, Inc.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of version 2 of the GNU General Public
6  * License as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful, but
9  * WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11  * General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public License
14  * along with this program; if not, write to the Free Software
15  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
16  * 02110-1301, USA
17  */
18 
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
20 
21 #include "flow.h"
22 #include "datapath.h"
23 #include <linux/uaccess.h>
24 #include <linux/netdevice.h>
25 #include <linux/etherdevice.h>
26 #include <linux/if_ether.h>
27 #include <linux/if_vlan.h>
28 #include <net/llc_pdu.h>
29 #include <linux/kernel.h>
30 #include <linux/jhash.h>
31 #include <linux/jiffies.h>
32 #include <linux/llc.h>
33 #include <linux/module.h>
34 #include <linux/in.h>
35 #include <linux/rcupdate.h>
36 #include <linux/if_arp.h>
37 #include <linux/ip.h>
38 #include <linux/ipv6.h>
39 #include <linux/sctp.h>
40 #include <linux/tcp.h>
41 #include <linux/udp.h>
42 #include <linux/icmp.h>
43 #include <linux/icmpv6.h>
44 #include <linux/rculist.h>
45 #include <net/geneve.h>
46 #include <net/ip.h>
47 #include <net/ipv6.h>
48 #include <net/ndisc.h>
49 
50 #include "flow_netlink.h"
51 
52 static void update_range__(struct sw_flow_match *match,
53 			   size_t offset, size_t size, bool is_mask)
54 {
55 	struct sw_flow_key_range *range = NULL;
56 	size_t start = rounddown(offset, sizeof(long));
57 	size_t end = roundup(offset + size, sizeof(long));
58 
59 	if (!is_mask)
60 		range = &match->range;
61 	else if (match->mask)
62 		range = &match->mask->range;
63 
64 	if (!range)
65 		return;
66 
67 	if (range->start == range->end) {
68 		range->start = start;
69 		range->end = end;
70 		return;
71 	}
72 
73 	if (range->start > start)
74 		range->start = start;
75 
76 	if (range->end < end)
77 		range->end = end;
78 }
79 
80 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
81 	do { \
82 		update_range__(match, offsetof(struct sw_flow_key, field),  \
83 				     sizeof((match)->key->field), is_mask); \
84 		if (is_mask) {						    \
85 			if ((match)->mask)				    \
86 				(match)->mask->key.field = value;	    \
87 		} else {                                                    \
88 			(match)->key->field = value;		            \
89 		}                                                           \
90 	} while (0)
91 
92 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask)	    \
93 	do {								    \
94 		update_range__(match, offset, len, is_mask);		    \
95 		if (is_mask)						    \
96 			memcpy((u8 *)&(match)->mask->key + offset, value_p, \
97 			       len);					    \
98 		else							    \
99 			memcpy((u8 *)(match)->key + offset, value_p, len);  \
100 	} while (0)
101 
102 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask)		      \
103 	SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
104 				  value_p, len, is_mask)
105 
106 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask) \
107 	do { \
108 		update_range__(match, offsetof(struct sw_flow_key, field),  \
109 				     sizeof((match)->key->field), is_mask); \
110 		if (is_mask) {						    \
111 			if ((match)->mask)				    \
112 				memset((u8 *)&(match)->mask->key.field, value,\
113 				       sizeof((match)->mask->key.field));   \
114 		} else {                                                    \
115 			memset((u8 *)&(match)->key->field, value,           \
116 			       sizeof((match)->key->field));                \
117 		}                                                           \
118 	} while (0)
119 
120 static bool match_validate(const struct sw_flow_match *match,
121 			   u64 key_attrs, u64 mask_attrs)
122 {
123 	u64 key_expected = 1 << OVS_KEY_ATTR_ETHERNET;
124 	u64 mask_allowed = key_attrs;  /* At most allow all key attributes */
125 
126 	/* The following mask attributes allowed only if they
127 	 * pass the validation tests. */
128 	mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
129 			| (1 << OVS_KEY_ATTR_IPV6)
130 			| (1 << OVS_KEY_ATTR_TCP)
131 			| (1 << OVS_KEY_ATTR_TCP_FLAGS)
132 			| (1 << OVS_KEY_ATTR_UDP)
133 			| (1 << OVS_KEY_ATTR_SCTP)
134 			| (1 << OVS_KEY_ATTR_ICMP)
135 			| (1 << OVS_KEY_ATTR_ICMPV6)
136 			| (1 << OVS_KEY_ATTR_ARP)
137 			| (1 << OVS_KEY_ATTR_ND));
138 
139 	/* Always allowed mask fields. */
140 	mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
141 		       | (1 << OVS_KEY_ATTR_IN_PORT)
142 		       | (1 << OVS_KEY_ATTR_ETHERTYPE));
143 
144 	/* Check key attributes. */
145 	if (match->key->eth.type == htons(ETH_P_ARP)
146 			|| match->key->eth.type == htons(ETH_P_RARP)) {
147 		key_expected |= 1 << OVS_KEY_ATTR_ARP;
148 		if (match->mask && (match->mask->key.tp.src == htons(0xff)))
149 			mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
150 	}
151 
152 	if (match->key->eth.type == htons(ETH_P_IP)) {
153 		key_expected |= 1 << OVS_KEY_ATTR_IPV4;
154 		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
155 			mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
156 
157 		if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
158 			if (match->key->ip.proto == IPPROTO_UDP) {
159 				key_expected |= 1 << OVS_KEY_ATTR_UDP;
160 				if (match->mask && (match->mask->key.ip.proto == 0xff))
161 					mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
162 			}
163 
164 			if (match->key->ip.proto == IPPROTO_SCTP) {
165 				key_expected |= 1 << OVS_KEY_ATTR_SCTP;
166 				if (match->mask && (match->mask->key.ip.proto == 0xff))
167 					mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
168 			}
169 
170 			if (match->key->ip.proto == IPPROTO_TCP) {
171 				key_expected |= 1 << OVS_KEY_ATTR_TCP;
172 				key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
173 				if (match->mask && (match->mask->key.ip.proto == 0xff)) {
174 					mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
175 					mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
176 				}
177 			}
178 
179 			if (match->key->ip.proto == IPPROTO_ICMP) {
180 				key_expected |= 1 << OVS_KEY_ATTR_ICMP;
181 				if (match->mask && (match->mask->key.ip.proto == 0xff))
182 					mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
183 			}
184 		}
185 	}
186 
187 	if (match->key->eth.type == htons(ETH_P_IPV6)) {
188 		key_expected |= 1 << OVS_KEY_ATTR_IPV6;
189 		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
190 			mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
191 
192 		if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
193 			if (match->key->ip.proto == IPPROTO_UDP) {
194 				key_expected |= 1 << OVS_KEY_ATTR_UDP;
195 				if (match->mask && (match->mask->key.ip.proto == 0xff))
196 					mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
197 			}
198 
199 			if (match->key->ip.proto == IPPROTO_SCTP) {
200 				key_expected |= 1 << OVS_KEY_ATTR_SCTP;
201 				if (match->mask && (match->mask->key.ip.proto == 0xff))
202 					mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
203 			}
204 
205 			if (match->key->ip.proto == IPPROTO_TCP) {
206 				key_expected |= 1 << OVS_KEY_ATTR_TCP;
207 				key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
208 				if (match->mask && (match->mask->key.ip.proto == 0xff)) {
209 					mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
210 					mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
211 				}
212 			}
213 
214 			if (match->key->ip.proto == IPPROTO_ICMPV6) {
215 				key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
216 				if (match->mask && (match->mask->key.ip.proto == 0xff))
217 					mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
218 
219 				if (match->key->tp.src ==
220 						htons(NDISC_NEIGHBOUR_SOLICITATION) ||
221 				    match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
222 					key_expected |= 1 << OVS_KEY_ATTR_ND;
223 					if (match->mask && (match->mask->key.tp.src == htons(0xffff)))
224 						mask_allowed |= 1 << OVS_KEY_ATTR_ND;
225 				}
226 			}
227 		}
228 	}
229 
230 	if ((key_attrs & key_expected) != key_expected) {
231 		/* Key attributes check failed. */
232 		OVS_NLERR("Missing expected key attributes (key_attrs=%llx, expected=%llx).\n",
233 				(unsigned long long)key_attrs, (unsigned long long)key_expected);
234 		return false;
235 	}
236 
237 	if ((mask_attrs & mask_allowed) != mask_attrs) {
238 		/* Mask attributes check failed. */
239 		OVS_NLERR("Contain more than allowed mask fields (mask_attrs=%llx, mask_allowed=%llx).\n",
240 				(unsigned long long)mask_attrs, (unsigned long long)mask_allowed);
241 		return false;
242 	}
243 
244 	return true;
245 }
246 
247 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute.  */
248 static const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
249 	[OVS_KEY_ATTR_ENCAP] = -1,
250 	[OVS_KEY_ATTR_PRIORITY] = sizeof(u32),
251 	[OVS_KEY_ATTR_IN_PORT] = sizeof(u32),
252 	[OVS_KEY_ATTR_SKB_MARK] = sizeof(u32),
253 	[OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
254 	[OVS_KEY_ATTR_VLAN] = sizeof(__be16),
255 	[OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16),
256 	[OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4),
257 	[OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
258 	[OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
259 	[OVS_KEY_ATTR_TCP_FLAGS] = sizeof(__be16),
260 	[OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
261 	[OVS_KEY_ATTR_SCTP] = sizeof(struct ovs_key_sctp),
262 	[OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
263 	[OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
264 	[OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
265 	[OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
266 	[OVS_KEY_ATTR_RECIRC_ID] = sizeof(u32),
267 	[OVS_KEY_ATTR_DP_HASH] = sizeof(u32),
268 	[OVS_KEY_ATTR_TUNNEL] = -1,
269 };
270 
271 static bool is_all_zero(const u8 *fp, size_t size)
272 {
273 	int i;
274 
275 	if (!fp)
276 		return false;
277 
278 	for (i = 0; i < size; i++)
279 		if (fp[i])
280 			return false;
281 
282 	return true;
283 }
284 
285 static int __parse_flow_nlattrs(const struct nlattr *attr,
286 				const struct nlattr *a[],
287 				u64 *attrsp, bool nz)
288 {
289 	const struct nlattr *nla;
290 	u64 attrs;
291 	int rem;
292 
293 	attrs = *attrsp;
294 	nla_for_each_nested(nla, attr, rem) {
295 		u16 type = nla_type(nla);
296 		int expected_len;
297 
298 		if (type > OVS_KEY_ATTR_MAX) {
299 			OVS_NLERR("Unknown key attribute (type=%d, max=%d).\n",
300 				  type, OVS_KEY_ATTR_MAX);
301 			return -EINVAL;
302 		}
303 
304 		if (attrs & (1 << type)) {
305 			OVS_NLERR("Duplicate key attribute (type %d).\n", type);
306 			return -EINVAL;
307 		}
308 
309 		expected_len = ovs_key_lens[type];
310 		if (nla_len(nla) != expected_len && expected_len != -1) {
311 			OVS_NLERR("Key attribute has unexpected length (type=%d"
312 				  ", length=%d, expected=%d).\n", type,
313 				  nla_len(nla), expected_len);
314 			return -EINVAL;
315 		}
316 
317 		if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
318 			attrs |= 1 << type;
319 			a[type] = nla;
320 		}
321 	}
322 	if (rem) {
323 		OVS_NLERR("Message has %d unknown bytes.\n", rem);
324 		return -EINVAL;
325 	}
326 
327 	*attrsp = attrs;
328 	return 0;
329 }
330 
331 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
332 				   const struct nlattr *a[], u64 *attrsp)
333 {
334 	return __parse_flow_nlattrs(attr, a, attrsp, true);
335 }
336 
337 static int parse_flow_nlattrs(const struct nlattr *attr,
338 			      const struct nlattr *a[], u64 *attrsp)
339 {
340 	return __parse_flow_nlattrs(attr, a, attrsp, false);
341 }
342 
343 static int ipv4_tun_from_nlattr(const struct nlattr *attr,
344 				struct sw_flow_match *match, bool is_mask)
345 {
346 	struct nlattr *a;
347 	int rem;
348 	bool ttl = false;
349 	__be16 tun_flags = 0;
350 	unsigned long opt_key_offset;
351 
352 	nla_for_each_nested(a, attr, rem) {
353 		int type = nla_type(a);
354 		static const u32 ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
355 			[OVS_TUNNEL_KEY_ATTR_ID] = sizeof(u64),
356 			[OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = sizeof(u32),
357 			[OVS_TUNNEL_KEY_ATTR_IPV4_DST] = sizeof(u32),
358 			[OVS_TUNNEL_KEY_ATTR_TOS] = 1,
359 			[OVS_TUNNEL_KEY_ATTR_TTL] = 1,
360 			[OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = 0,
361 			[OVS_TUNNEL_KEY_ATTR_CSUM] = 0,
362 			[OVS_TUNNEL_KEY_ATTR_OAM] = 0,
363 			[OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = -1,
364 		};
365 
366 		if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
367 			OVS_NLERR("Unknown IPv4 tunnel attribute (type=%d, max=%d).\n",
368 			type, OVS_TUNNEL_KEY_ATTR_MAX);
369 			return -EINVAL;
370 		}
371 
372 		if (ovs_tunnel_key_lens[type] != nla_len(a) &&
373 		    ovs_tunnel_key_lens[type] != -1) {
374 			OVS_NLERR("IPv4 tunnel attribute type has unexpected "
375 				  " length (type=%d, length=%d, expected=%d).\n",
376 				  type, nla_len(a), ovs_tunnel_key_lens[type]);
377 			return -EINVAL;
378 		}
379 
380 		switch (type) {
381 		case OVS_TUNNEL_KEY_ATTR_ID:
382 			SW_FLOW_KEY_PUT(match, tun_key.tun_id,
383 					nla_get_be64(a), is_mask);
384 			tun_flags |= TUNNEL_KEY;
385 			break;
386 		case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
387 			SW_FLOW_KEY_PUT(match, tun_key.ipv4_src,
388 					nla_get_be32(a), is_mask);
389 			break;
390 		case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
391 			SW_FLOW_KEY_PUT(match, tun_key.ipv4_dst,
392 					nla_get_be32(a), is_mask);
393 			break;
394 		case OVS_TUNNEL_KEY_ATTR_TOS:
395 			SW_FLOW_KEY_PUT(match, tun_key.ipv4_tos,
396 					nla_get_u8(a), is_mask);
397 			break;
398 		case OVS_TUNNEL_KEY_ATTR_TTL:
399 			SW_FLOW_KEY_PUT(match, tun_key.ipv4_ttl,
400 					nla_get_u8(a), is_mask);
401 			ttl = true;
402 			break;
403 		case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
404 			tun_flags |= TUNNEL_DONT_FRAGMENT;
405 			break;
406 		case OVS_TUNNEL_KEY_ATTR_CSUM:
407 			tun_flags |= TUNNEL_CSUM;
408 			break;
409 		case OVS_TUNNEL_KEY_ATTR_OAM:
410 			tun_flags |= TUNNEL_OAM;
411 			break;
412 		case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
413 			tun_flags |= TUNNEL_OPTIONS_PRESENT;
414 			if (nla_len(a) > sizeof(match->key->tun_opts)) {
415 				OVS_NLERR("Geneve option length exceeds maximum size (len %d, max %zu).\n",
416 					  nla_len(a),
417 					  sizeof(match->key->tun_opts));
418 				return -EINVAL;
419 			}
420 
421 			if (nla_len(a) % 4 != 0) {
422 				OVS_NLERR("Geneve option length is not a multiple of 4 (len %d).\n",
423 					  nla_len(a));
424 				return -EINVAL;
425 			}
426 
427 			/* We need to record the length of the options passed
428 			 * down, otherwise packets with the same format but
429 			 * additional options will be silently matched.
430 			 */
431 			if (!is_mask) {
432 				SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
433 						false);
434 			} else {
435 				/* This is somewhat unusual because it looks at
436 				 * both the key and mask while parsing the
437 				 * attributes (and by extension assumes the key
438 				 * is parsed first). Normally, we would verify
439 				 * that each is the correct length and that the
440 				 * attributes line up in the validate function.
441 				 * However, that is difficult because this is
442 				 * variable length and we won't have the
443 				 * information later.
444 				 */
445 				if (match->key->tun_opts_len != nla_len(a)) {
446 					OVS_NLERR("Geneve option key length (%d) is different from mask length (%d).",
447 						  match->key->tun_opts_len,
448 						  nla_len(a));
449 					return -EINVAL;
450 				}
451 
452 				SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff,
453 						true);
454 			}
455 
456 			opt_key_offset = (unsigned long)GENEVE_OPTS(
457 					  (struct sw_flow_key *)0,
458 					  nla_len(a));
459 			SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset,
460 						  nla_data(a), nla_len(a),
461 						  is_mask);
462 			break;
463 		default:
464 			OVS_NLERR("Unknown IPv4 tunnel attribute (%d).\n",
465 				  type);
466 			return -EINVAL;
467 		}
468 	}
469 
470 	SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
471 
472 	if (rem > 0) {
473 		OVS_NLERR("IPv4 tunnel attribute has %d unknown bytes.\n", rem);
474 		return -EINVAL;
475 	}
476 
477 	if (!is_mask) {
478 		if (!match->key->tun_key.ipv4_dst) {
479 			OVS_NLERR("IPv4 tunnel destination address is zero.\n");
480 			return -EINVAL;
481 		}
482 
483 		if (!ttl) {
484 			OVS_NLERR("IPv4 tunnel TTL not specified.\n");
485 			return -EINVAL;
486 		}
487 	}
488 
489 	return 0;
490 }
491 
492 static int __ipv4_tun_to_nlattr(struct sk_buff *skb,
493 				const struct ovs_key_ipv4_tunnel *output,
494 				const struct geneve_opt *tun_opts,
495 				int swkey_tun_opts_len)
496 {
497 	if (output->tun_flags & TUNNEL_KEY &&
498 	    nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
499 		return -EMSGSIZE;
500 	if (output->ipv4_src &&
501 	    nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC, output->ipv4_src))
502 		return -EMSGSIZE;
503 	if (output->ipv4_dst &&
504 	    nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST, output->ipv4_dst))
505 		return -EMSGSIZE;
506 	if (output->ipv4_tos &&
507 	    nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->ipv4_tos))
508 		return -EMSGSIZE;
509 	if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ipv4_ttl))
510 		return -EMSGSIZE;
511 	if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
512 	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
513 		return -EMSGSIZE;
514 	if ((output->tun_flags & TUNNEL_CSUM) &&
515 	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
516 		return -EMSGSIZE;
517 	if ((output->tun_flags & TUNNEL_OAM) &&
518 	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
519 		return -EMSGSIZE;
520 	if (tun_opts &&
521 	    nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
522 		    swkey_tun_opts_len, tun_opts))
523 		return -EMSGSIZE;
524 
525 	return 0;
526 }
527 
528 
529 static int ipv4_tun_to_nlattr(struct sk_buff *skb,
530 			      const struct ovs_key_ipv4_tunnel *output,
531 			      const struct geneve_opt *tun_opts,
532 			      int swkey_tun_opts_len)
533 {
534 	struct nlattr *nla;
535 	int err;
536 
537 	nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
538 	if (!nla)
539 		return -EMSGSIZE;
540 
541 	err = __ipv4_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len);
542 	if (err)
543 		return err;
544 
545 	nla_nest_end(skb, nla);
546 	return 0;
547 }
548 
549 static int metadata_from_nlattrs(struct sw_flow_match *match,  u64 *attrs,
550 				 const struct nlattr **a, bool is_mask)
551 {
552 	if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
553 		u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
554 
555 		SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
556 		*attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
557 	}
558 
559 	if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
560 		u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
561 
562 		SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
563 		*attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
564 	}
565 
566 	if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
567 		SW_FLOW_KEY_PUT(match, phy.priority,
568 			  nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
569 		*attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
570 	}
571 
572 	if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
573 		u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
574 
575 		if (is_mask)
576 			in_port = 0xffffffff; /* Always exact match in_port. */
577 		else if (in_port >= DP_MAX_PORTS)
578 			return -EINVAL;
579 
580 		SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
581 		*attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
582 	} else if (!is_mask) {
583 		SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
584 	}
585 
586 	if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
587 		uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
588 
589 		SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
590 		*attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
591 	}
592 	if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
593 		if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
594 					 is_mask))
595 			return -EINVAL;
596 		*attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
597 	}
598 	return 0;
599 }
600 
601 static int ovs_key_from_nlattrs(struct sw_flow_match *match, u64 attrs,
602 				const struct nlattr **a, bool is_mask)
603 {
604 	int err;
605 	u64 orig_attrs = attrs;
606 
607 	err = metadata_from_nlattrs(match, &attrs, a, is_mask);
608 	if (err)
609 		return err;
610 
611 	if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
612 		const struct ovs_key_ethernet *eth_key;
613 
614 		eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
615 		SW_FLOW_KEY_MEMCPY(match, eth.src,
616 				eth_key->eth_src, ETH_ALEN, is_mask);
617 		SW_FLOW_KEY_MEMCPY(match, eth.dst,
618 				eth_key->eth_dst, ETH_ALEN, is_mask);
619 		attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
620 	}
621 
622 	if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
623 		__be16 tci;
624 
625 		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
626 		if (!(tci & htons(VLAN_TAG_PRESENT))) {
627 			if (is_mask)
628 				OVS_NLERR("VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.\n");
629 			else
630 				OVS_NLERR("VLAN TCI does not have VLAN_TAG_PRESENT bit set.\n");
631 
632 			return -EINVAL;
633 		}
634 
635 		SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
636 		attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
637 	} else if (!is_mask)
638 		SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
639 
640 	if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
641 		__be16 eth_type;
642 
643 		eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
644 		if (is_mask) {
645 			/* Always exact match EtherType. */
646 			eth_type = htons(0xffff);
647 		} else if (ntohs(eth_type) < ETH_P_802_3_MIN) {
648 			OVS_NLERR("EtherType is less than minimum (type=%x, min=%x).\n",
649 					ntohs(eth_type), ETH_P_802_3_MIN);
650 			return -EINVAL;
651 		}
652 
653 		SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
654 		attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
655 	} else if (!is_mask) {
656 		SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
657 	}
658 
659 	if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
660 		const struct ovs_key_ipv4 *ipv4_key;
661 
662 		ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
663 		if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
664 			OVS_NLERR("Unknown IPv4 fragment type (value=%d, max=%d).\n",
665 				ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
666 			return -EINVAL;
667 		}
668 		SW_FLOW_KEY_PUT(match, ip.proto,
669 				ipv4_key->ipv4_proto, is_mask);
670 		SW_FLOW_KEY_PUT(match, ip.tos,
671 				ipv4_key->ipv4_tos, is_mask);
672 		SW_FLOW_KEY_PUT(match, ip.ttl,
673 				ipv4_key->ipv4_ttl, is_mask);
674 		SW_FLOW_KEY_PUT(match, ip.frag,
675 				ipv4_key->ipv4_frag, is_mask);
676 		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
677 				ipv4_key->ipv4_src, is_mask);
678 		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
679 				ipv4_key->ipv4_dst, is_mask);
680 		attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
681 	}
682 
683 	if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
684 		const struct ovs_key_ipv6 *ipv6_key;
685 
686 		ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
687 		if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
688 			OVS_NLERR("Unknown IPv6 fragment type (value=%d, max=%d).\n",
689 				ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
690 			return -EINVAL;
691 		}
692 
693 		if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
694 			OVS_NLERR("IPv6 flow label %x is out of range (max=%x).\n",
695 				  ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
696 			return -EINVAL;
697 		}
698 
699 		SW_FLOW_KEY_PUT(match, ipv6.label,
700 				ipv6_key->ipv6_label, is_mask);
701 		SW_FLOW_KEY_PUT(match, ip.proto,
702 				ipv6_key->ipv6_proto, is_mask);
703 		SW_FLOW_KEY_PUT(match, ip.tos,
704 				ipv6_key->ipv6_tclass, is_mask);
705 		SW_FLOW_KEY_PUT(match, ip.ttl,
706 				ipv6_key->ipv6_hlimit, is_mask);
707 		SW_FLOW_KEY_PUT(match, ip.frag,
708 				ipv6_key->ipv6_frag, is_mask);
709 		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
710 				ipv6_key->ipv6_src,
711 				sizeof(match->key->ipv6.addr.src),
712 				is_mask);
713 		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
714 				ipv6_key->ipv6_dst,
715 				sizeof(match->key->ipv6.addr.dst),
716 				is_mask);
717 
718 		attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
719 	}
720 
721 	if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
722 		const struct ovs_key_arp *arp_key;
723 
724 		arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
725 		if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
726 			OVS_NLERR("Unknown ARP opcode (opcode=%d).\n",
727 				  arp_key->arp_op);
728 			return -EINVAL;
729 		}
730 
731 		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
732 				arp_key->arp_sip, is_mask);
733 		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
734 			arp_key->arp_tip, is_mask);
735 		SW_FLOW_KEY_PUT(match, ip.proto,
736 				ntohs(arp_key->arp_op), is_mask);
737 		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
738 				arp_key->arp_sha, ETH_ALEN, is_mask);
739 		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
740 				arp_key->arp_tha, ETH_ALEN, is_mask);
741 
742 		attrs &= ~(1 << OVS_KEY_ATTR_ARP);
743 	}
744 
745 	if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
746 		const struct ovs_key_tcp *tcp_key;
747 
748 		tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
749 		SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
750 		SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
751 		attrs &= ~(1 << OVS_KEY_ATTR_TCP);
752 	}
753 
754 	if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
755 		if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
756 			SW_FLOW_KEY_PUT(match, tp.flags,
757 					nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
758 					is_mask);
759 		} else {
760 			SW_FLOW_KEY_PUT(match, tp.flags,
761 					nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
762 					is_mask);
763 		}
764 		attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
765 	}
766 
767 	if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
768 		const struct ovs_key_udp *udp_key;
769 
770 		udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
771 		SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
772 		SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
773 		attrs &= ~(1 << OVS_KEY_ATTR_UDP);
774 	}
775 
776 	if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
777 		const struct ovs_key_sctp *sctp_key;
778 
779 		sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
780 		SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
781 		SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
782 		attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
783 	}
784 
785 	if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
786 		const struct ovs_key_icmp *icmp_key;
787 
788 		icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
789 		SW_FLOW_KEY_PUT(match, tp.src,
790 				htons(icmp_key->icmp_type), is_mask);
791 		SW_FLOW_KEY_PUT(match, tp.dst,
792 				htons(icmp_key->icmp_code), is_mask);
793 		attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
794 	}
795 
796 	if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
797 		const struct ovs_key_icmpv6 *icmpv6_key;
798 
799 		icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
800 		SW_FLOW_KEY_PUT(match, tp.src,
801 				htons(icmpv6_key->icmpv6_type), is_mask);
802 		SW_FLOW_KEY_PUT(match, tp.dst,
803 				htons(icmpv6_key->icmpv6_code), is_mask);
804 		attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
805 	}
806 
807 	if (attrs & (1 << OVS_KEY_ATTR_ND)) {
808 		const struct ovs_key_nd *nd_key;
809 
810 		nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
811 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
812 			nd_key->nd_target,
813 			sizeof(match->key->ipv6.nd.target),
814 			is_mask);
815 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
816 			nd_key->nd_sll, ETH_ALEN, is_mask);
817 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
818 				nd_key->nd_tll, ETH_ALEN, is_mask);
819 		attrs &= ~(1 << OVS_KEY_ATTR_ND);
820 	}
821 
822 	if (attrs != 0)
823 		return -EINVAL;
824 
825 	return 0;
826 }
827 
828 static void nlattr_set(struct nlattr *attr, u8 val, bool is_attr_mask_key)
829 {
830 	struct nlattr *nla;
831 	int rem;
832 
833 	/* The nlattr stream should already have been validated */
834 	nla_for_each_nested(nla, attr, rem) {
835 		/* We assume that ovs_key_lens[type] == -1 means that type is a
836 		 * nested attribute
837 		 */
838 		if (is_attr_mask_key && ovs_key_lens[nla_type(nla)] == -1)
839 			nlattr_set(nla, val, false);
840 		else
841 			memset(nla_data(nla), val, nla_len(nla));
842 	}
843 }
844 
845 static void mask_set_nlattr(struct nlattr *attr, u8 val)
846 {
847 	nlattr_set(attr, val, true);
848 }
849 
850 /**
851  * ovs_nla_get_match - parses Netlink attributes into a flow key and
852  * mask. In case the 'mask' is NULL, the flow is treated as exact match
853  * flow. Otherwise, it is treated as a wildcarded flow, except the mask
854  * does not include any don't care bit.
855  * @match: receives the extracted flow match information.
856  * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
857  * sequence. The fields should of the packet that triggered the creation
858  * of this flow.
859  * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
860  * attribute specifies the mask field of the wildcarded flow.
861  */
862 int ovs_nla_get_match(struct sw_flow_match *match,
863 		      const struct nlattr *key,
864 		      const struct nlattr *mask)
865 {
866 	const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
867 	const struct nlattr *encap;
868 	struct nlattr *newmask = NULL;
869 	u64 key_attrs = 0;
870 	u64 mask_attrs = 0;
871 	bool encap_valid = false;
872 	int err;
873 
874 	err = parse_flow_nlattrs(key, a, &key_attrs);
875 	if (err)
876 		return err;
877 
878 	if ((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
879 	    (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
880 	    (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
881 		__be16 tci;
882 
883 		if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
884 		      (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
885 			OVS_NLERR("Invalid Vlan frame.\n");
886 			return -EINVAL;
887 		}
888 
889 		key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
890 		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
891 		encap = a[OVS_KEY_ATTR_ENCAP];
892 		key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
893 		encap_valid = true;
894 
895 		if (tci & htons(VLAN_TAG_PRESENT)) {
896 			err = parse_flow_nlattrs(encap, a, &key_attrs);
897 			if (err)
898 				return err;
899 		} else if (!tci) {
900 			/* Corner case for truncated 802.1Q header. */
901 			if (nla_len(encap)) {
902 				OVS_NLERR("Truncated 802.1Q header has non-zero encap attribute.\n");
903 				return -EINVAL;
904 			}
905 		} else {
906 			OVS_NLERR("Encap attribute is set for a non-VLAN frame.\n");
907 			return  -EINVAL;
908 		}
909 	}
910 
911 	err = ovs_key_from_nlattrs(match, key_attrs, a, false);
912 	if (err)
913 		return err;
914 
915 	if (match->mask && !mask) {
916 		/* Create an exact match mask. We need to set to 0xff all the
917 		 * 'match->mask' fields that have been touched in 'match->key'.
918 		 * We cannot simply memset 'match->mask', because padding bytes
919 		 * and fields not specified in 'match->key' should be left to 0.
920 		 * Instead, we use a stream of netlink attributes, copied from
921 		 * 'key' and set to 0xff: ovs_key_from_nlattrs() will take care
922 		 * of filling 'match->mask' appropriately.
923 		 */
924 		newmask = kmemdup(key, nla_total_size(nla_len(key)),
925 				  GFP_KERNEL);
926 		if (!newmask)
927 			return -ENOMEM;
928 
929 		mask_set_nlattr(newmask, 0xff);
930 
931 		/* The userspace does not send tunnel attributes that are 0,
932 		 * but we should not wildcard them nonetheless.
933 		 */
934 		if (match->key->tun_key.ipv4_dst)
935 			SW_FLOW_KEY_MEMSET_FIELD(match, tun_key, 0xff, true);
936 
937 		mask = newmask;
938 	}
939 
940 	if (mask) {
941 		err = parse_flow_mask_nlattrs(mask, a, &mask_attrs);
942 		if (err)
943 			goto free_newmask;
944 
945 		if (mask_attrs & 1 << OVS_KEY_ATTR_ENCAP) {
946 			__be16 eth_type = 0;
947 			__be16 tci = 0;
948 
949 			if (!encap_valid) {
950 				OVS_NLERR("Encap mask attribute is set for non-VLAN frame.\n");
951 				err = -EINVAL;
952 				goto free_newmask;
953 			}
954 
955 			mask_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
956 			if (a[OVS_KEY_ATTR_ETHERTYPE])
957 				eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
958 
959 			if (eth_type == htons(0xffff)) {
960 				mask_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
961 				encap = a[OVS_KEY_ATTR_ENCAP];
962 				err = parse_flow_mask_nlattrs(encap, a, &mask_attrs);
963 				if (err)
964 					goto free_newmask;
965 			} else {
966 				OVS_NLERR("VLAN frames must have an exact match on the TPID (mask=%x).\n",
967 						ntohs(eth_type));
968 				err = -EINVAL;
969 				goto free_newmask;
970 			}
971 
972 			if (a[OVS_KEY_ATTR_VLAN])
973 				tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
974 
975 			if (!(tci & htons(VLAN_TAG_PRESENT))) {
976 				OVS_NLERR("VLAN tag present bit must have an exact match (tci_mask=%x).\n", ntohs(tci));
977 				err = -EINVAL;
978 				goto free_newmask;
979 			}
980 		}
981 
982 		err = ovs_key_from_nlattrs(match, mask_attrs, a, true);
983 		if (err)
984 			goto free_newmask;
985 	}
986 
987 	if (!match_validate(match, key_attrs, mask_attrs))
988 		err = -EINVAL;
989 
990 free_newmask:
991 	kfree(newmask);
992 	return err;
993 }
994 
995 /**
996  * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
997  * @key: Receives extracted in_port, priority, tun_key and skb_mark.
998  * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
999  * sequence.
1000  *
1001  * This parses a series of Netlink attributes that form a flow key, which must
1002  * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1003  * get the metadata, that is, the parts of the flow key that cannot be
1004  * extracted from the packet itself.
1005  */
1006 
1007 int ovs_nla_get_flow_metadata(const struct nlattr *attr,
1008 			      struct sw_flow_key *key)
1009 {
1010 	const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1011 	struct sw_flow_match match;
1012 	u64 attrs = 0;
1013 	int err;
1014 
1015 	err = parse_flow_nlattrs(attr, a, &attrs);
1016 	if (err)
1017 		return -EINVAL;
1018 
1019 	memset(&match, 0, sizeof(match));
1020 	match.key = key;
1021 
1022 	key->phy.in_port = DP_MAX_PORTS;
1023 
1024 	return metadata_from_nlattrs(&match, &attrs, a, false);
1025 }
1026 
1027 int ovs_nla_put_flow(const struct sw_flow_key *swkey,
1028 		     const struct sw_flow_key *output, struct sk_buff *skb)
1029 {
1030 	struct ovs_key_ethernet *eth_key;
1031 	struct nlattr *nla, *encap;
1032 	bool is_mask = (swkey != output);
1033 
1034 	if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1035 		goto nla_put_failure;
1036 
1037 	if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1038 		goto nla_put_failure;
1039 
1040 	if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1041 		goto nla_put_failure;
1042 
1043 	if ((swkey->tun_key.ipv4_dst || is_mask)) {
1044 		const struct geneve_opt *opts = NULL;
1045 
1046 		if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1047 			opts = GENEVE_OPTS(output, swkey->tun_opts_len);
1048 
1049 		if (ipv4_tun_to_nlattr(skb, &output->tun_key, opts,
1050 				       swkey->tun_opts_len))
1051 			goto nla_put_failure;
1052 	}
1053 
1054 	if (swkey->phy.in_port == DP_MAX_PORTS) {
1055 		if (is_mask && (output->phy.in_port == 0xffff))
1056 			if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1057 				goto nla_put_failure;
1058 	} else {
1059 		u16 upper_u16;
1060 		upper_u16 = !is_mask ? 0 : 0xffff;
1061 
1062 		if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1063 				(upper_u16 << 16) | output->phy.in_port))
1064 			goto nla_put_failure;
1065 	}
1066 
1067 	if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1068 		goto nla_put_failure;
1069 
1070 	nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1071 	if (!nla)
1072 		goto nla_put_failure;
1073 
1074 	eth_key = nla_data(nla);
1075 	ether_addr_copy(eth_key->eth_src, output->eth.src);
1076 	ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1077 
1078 	if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1079 		__be16 eth_type;
1080 		eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
1081 		if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1082 		    nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
1083 			goto nla_put_failure;
1084 		encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1085 		if (!swkey->eth.tci)
1086 			goto unencap;
1087 	} else
1088 		encap = NULL;
1089 
1090 	if (swkey->eth.type == htons(ETH_P_802_2)) {
1091 		/*
1092 		 * Ethertype 802.2 is represented in the netlink with omitted
1093 		 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1094 		 * 0xffff in the mask attribute.  Ethertype can also
1095 		 * be wildcarded.
1096 		 */
1097 		if (is_mask && output->eth.type)
1098 			if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1099 						output->eth.type))
1100 				goto nla_put_failure;
1101 		goto unencap;
1102 	}
1103 
1104 	if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1105 		goto nla_put_failure;
1106 
1107 	if (swkey->eth.type == htons(ETH_P_IP)) {
1108 		struct ovs_key_ipv4 *ipv4_key;
1109 
1110 		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1111 		if (!nla)
1112 			goto nla_put_failure;
1113 		ipv4_key = nla_data(nla);
1114 		ipv4_key->ipv4_src = output->ipv4.addr.src;
1115 		ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1116 		ipv4_key->ipv4_proto = output->ip.proto;
1117 		ipv4_key->ipv4_tos = output->ip.tos;
1118 		ipv4_key->ipv4_ttl = output->ip.ttl;
1119 		ipv4_key->ipv4_frag = output->ip.frag;
1120 	} else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1121 		struct ovs_key_ipv6 *ipv6_key;
1122 
1123 		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1124 		if (!nla)
1125 			goto nla_put_failure;
1126 		ipv6_key = nla_data(nla);
1127 		memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1128 				sizeof(ipv6_key->ipv6_src));
1129 		memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1130 				sizeof(ipv6_key->ipv6_dst));
1131 		ipv6_key->ipv6_label = output->ipv6.label;
1132 		ipv6_key->ipv6_proto = output->ip.proto;
1133 		ipv6_key->ipv6_tclass = output->ip.tos;
1134 		ipv6_key->ipv6_hlimit = output->ip.ttl;
1135 		ipv6_key->ipv6_frag = output->ip.frag;
1136 	} else if (swkey->eth.type == htons(ETH_P_ARP) ||
1137 		   swkey->eth.type == htons(ETH_P_RARP)) {
1138 		struct ovs_key_arp *arp_key;
1139 
1140 		nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1141 		if (!nla)
1142 			goto nla_put_failure;
1143 		arp_key = nla_data(nla);
1144 		memset(arp_key, 0, sizeof(struct ovs_key_arp));
1145 		arp_key->arp_sip = output->ipv4.addr.src;
1146 		arp_key->arp_tip = output->ipv4.addr.dst;
1147 		arp_key->arp_op = htons(output->ip.proto);
1148 		ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1149 		ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1150 	}
1151 
1152 	if ((swkey->eth.type == htons(ETH_P_IP) ||
1153 	     swkey->eth.type == htons(ETH_P_IPV6)) &&
1154 	     swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1155 
1156 		if (swkey->ip.proto == IPPROTO_TCP) {
1157 			struct ovs_key_tcp *tcp_key;
1158 
1159 			nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1160 			if (!nla)
1161 				goto nla_put_failure;
1162 			tcp_key = nla_data(nla);
1163 			tcp_key->tcp_src = output->tp.src;
1164 			tcp_key->tcp_dst = output->tp.dst;
1165 			if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1166 					 output->tp.flags))
1167 				goto nla_put_failure;
1168 		} else if (swkey->ip.proto == IPPROTO_UDP) {
1169 			struct ovs_key_udp *udp_key;
1170 
1171 			nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1172 			if (!nla)
1173 				goto nla_put_failure;
1174 			udp_key = nla_data(nla);
1175 			udp_key->udp_src = output->tp.src;
1176 			udp_key->udp_dst = output->tp.dst;
1177 		} else if (swkey->ip.proto == IPPROTO_SCTP) {
1178 			struct ovs_key_sctp *sctp_key;
1179 
1180 			nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1181 			if (!nla)
1182 				goto nla_put_failure;
1183 			sctp_key = nla_data(nla);
1184 			sctp_key->sctp_src = output->tp.src;
1185 			sctp_key->sctp_dst = output->tp.dst;
1186 		} else if (swkey->eth.type == htons(ETH_P_IP) &&
1187 			   swkey->ip.proto == IPPROTO_ICMP) {
1188 			struct ovs_key_icmp *icmp_key;
1189 
1190 			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1191 			if (!nla)
1192 				goto nla_put_failure;
1193 			icmp_key = nla_data(nla);
1194 			icmp_key->icmp_type = ntohs(output->tp.src);
1195 			icmp_key->icmp_code = ntohs(output->tp.dst);
1196 		} else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1197 			   swkey->ip.proto == IPPROTO_ICMPV6) {
1198 			struct ovs_key_icmpv6 *icmpv6_key;
1199 
1200 			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1201 						sizeof(*icmpv6_key));
1202 			if (!nla)
1203 				goto nla_put_failure;
1204 			icmpv6_key = nla_data(nla);
1205 			icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1206 			icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1207 
1208 			if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1209 			    icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1210 				struct ovs_key_nd *nd_key;
1211 
1212 				nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1213 				if (!nla)
1214 					goto nla_put_failure;
1215 				nd_key = nla_data(nla);
1216 				memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1217 							sizeof(nd_key->nd_target));
1218 				ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1219 				ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1220 			}
1221 		}
1222 	}
1223 
1224 unencap:
1225 	if (encap)
1226 		nla_nest_end(skb, encap);
1227 
1228 	return 0;
1229 
1230 nla_put_failure:
1231 	return -EMSGSIZE;
1232 }
1233 
1234 #define MAX_ACTIONS_BUFSIZE	(32 * 1024)
1235 
1236 struct sw_flow_actions *ovs_nla_alloc_flow_actions(int size)
1237 {
1238 	struct sw_flow_actions *sfa;
1239 
1240 	if (size > MAX_ACTIONS_BUFSIZE)
1241 		return ERR_PTR(-EINVAL);
1242 
1243 	sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1244 	if (!sfa)
1245 		return ERR_PTR(-ENOMEM);
1246 
1247 	sfa->actions_len = 0;
1248 	return sfa;
1249 }
1250 
1251 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1252  * The caller must hold rcu_read_lock for this to be sensible. */
1253 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1254 {
1255 	kfree_rcu(sf_acts, rcu);
1256 }
1257 
1258 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1259 				       int attr_len)
1260 {
1261 
1262 	struct sw_flow_actions *acts;
1263 	int new_acts_size;
1264 	int req_size = NLA_ALIGN(attr_len);
1265 	int next_offset = offsetof(struct sw_flow_actions, actions) +
1266 					(*sfa)->actions_len;
1267 
1268 	if (req_size <= (ksize(*sfa) - next_offset))
1269 		goto out;
1270 
1271 	new_acts_size = ksize(*sfa) * 2;
1272 
1273 	if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1274 		if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1275 			return ERR_PTR(-EMSGSIZE);
1276 		new_acts_size = MAX_ACTIONS_BUFSIZE;
1277 	}
1278 
1279 	acts = ovs_nla_alloc_flow_actions(new_acts_size);
1280 	if (IS_ERR(acts))
1281 		return (void *)acts;
1282 
1283 	memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1284 	acts->actions_len = (*sfa)->actions_len;
1285 	kfree(*sfa);
1286 	*sfa = acts;
1287 
1288 out:
1289 	(*sfa)->actions_len += req_size;
1290 	return  (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1291 }
1292 
1293 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1294 				   int attrtype, void *data, int len)
1295 {
1296 	struct nlattr *a;
1297 
1298 	a = reserve_sfa_size(sfa, nla_attr_size(len));
1299 	if (IS_ERR(a))
1300 		return a;
1301 
1302 	a->nla_type = attrtype;
1303 	a->nla_len = nla_attr_size(len);
1304 
1305 	if (data)
1306 		memcpy(nla_data(a), data, len);
1307 	memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1308 
1309 	return a;
1310 }
1311 
1312 static int add_action(struct sw_flow_actions **sfa, int attrtype,
1313 		      void *data, int len)
1314 {
1315 	struct nlattr *a;
1316 
1317 	a = __add_action(sfa, attrtype, data, len);
1318 	if (IS_ERR(a))
1319 		return PTR_ERR(a);
1320 
1321 	return 0;
1322 }
1323 
1324 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1325 					  int attrtype)
1326 {
1327 	int used = (*sfa)->actions_len;
1328 	int err;
1329 
1330 	err = add_action(sfa, attrtype, NULL, 0);
1331 	if (err)
1332 		return err;
1333 
1334 	return used;
1335 }
1336 
1337 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1338 					 int st_offset)
1339 {
1340 	struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1341 							       st_offset);
1342 
1343 	a->nla_len = sfa->actions_len - st_offset;
1344 }
1345 
1346 static int validate_and_copy_sample(const struct nlattr *attr,
1347 				    const struct sw_flow_key *key, int depth,
1348 				    struct sw_flow_actions **sfa)
1349 {
1350 	const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1351 	const struct nlattr *probability, *actions;
1352 	const struct nlattr *a;
1353 	int rem, start, err, st_acts;
1354 
1355 	memset(attrs, 0, sizeof(attrs));
1356 	nla_for_each_nested(a, attr, rem) {
1357 		int type = nla_type(a);
1358 		if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1359 			return -EINVAL;
1360 		attrs[type] = a;
1361 	}
1362 	if (rem)
1363 		return -EINVAL;
1364 
1365 	probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1366 	if (!probability || nla_len(probability) != sizeof(u32))
1367 		return -EINVAL;
1368 
1369 	actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1370 	if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1371 		return -EINVAL;
1372 
1373 	/* validation done, copy sample action. */
1374 	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE);
1375 	if (start < 0)
1376 		return start;
1377 	err = add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1378 			 nla_data(probability), sizeof(u32));
1379 	if (err)
1380 		return err;
1381 	st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS);
1382 	if (st_acts < 0)
1383 		return st_acts;
1384 
1385 	err = ovs_nla_copy_actions(actions, key, depth + 1, sfa);
1386 	if (err)
1387 		return err;
1388 
1389 	add_nested_action_end(*sfa, st_acts);
1390 	add_nested_action_end(*sfa, start);
1391 
1392 	return 0;
1393 }
1394 
1395 static int validate_tp_port(const struct sw_flow_key *flow_key)
1396 {
1397 	if ((flow_key->eth.type == htons(ETH_P_IP) ||
1398 	     flow_key->eth.type == htons(ETH_P_IPV6)) &&
1399 	    (flow_key->tp.src || flow_key->tp.dst))
1400 		return 0;
1401 
1402 	return -EINVAL;
1403 }
1404 
1405 void ovs_match_init(struct sw_flow_match *match,
1406 		    struct sw_flow_key *key,
1407 		    struct sw_flow_mask *mask)
1408 {
1409 	memset(match, 0, sizeof(*match));
1410 	match->key = key;
1411 	match->mask = mask;
1412 
1413 	memset(key, 0, sizeof(*key));
1414 
1415 	if (mask) {
1416 		memset(&mask->key, 0, sizeof(mask->key));
1417 		mask->range.start = mask->range.end = 0;
1418 	}
1419 }
1420 
1421 static int validate_and_copy_set_tun(const struct nlattr *attr,
1422 				     struct sw_flow_actions **sfa)
1423 {
1424 	struct sw_flow_match match;
1425 	struct sw_flow_key key;
1426 	struct ovs_tunnel_info *tun_info;
1427 	struct nlattr *a;
1428 	int err, start;
1429 
1430 	ovs_match_init(&match, &key, NULL);
1431 	err = ipv4_tun_from_nlattr(nla_data(attr), &match, false);
1432 	if (err)
1433 		return err;
1434 
1435 	if (key.tun_opts_len) {
1436 		struct geneve_opt *option = GENEVE_OPTS(&key,
1437 							key.tun_opts_len);
1438 		int opts_len = key.tun_opts_len;
1439 		bool crit_opt = false;
1440 
1441 		while (opts_len > 0) {
1442 			int len;
1443 
1444 			if (opts_len < sizeof(*option))
1445 				return -EINVAL;
1446 
1447 			len = sizeof(*option) + option->length * 4;
1448 			if (len > opts_len)
1449 				return -EINVAL;
1450 
1451 			crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
1452 
1453 			option = (struct geneve_opt *)((u8 *)option + len);
1454 			opts_len -= len;
1455 		};
1456 
1457 		key.tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
1458 	};
1459 
1460 	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET);
1461 	if (start < 0)
1462 		return start;
1463 
1464 	a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
1465 			 sizeof(*tun_info) + key.tun_opts_len);
1466 	if (IS_ERR(a))
1467 		return PTR_ERR(a);
1468 
1469 	tun_info = nla_data(a);
1470 	tun_info->tunnel = key.tun_key;
1471 	tun_info->options_len = key.tun_opts_len;
1472 
1473 	if (tun_info->options_len) {
1474 		/* We need to store the options in the action itself since
1475 		 * everything else will go away after flow setup. We can append
1476 		 * it to tun_info and then point there.
1477 		 */
1478 		memcpy((tun_info + 1), GENEVE_OPTS(&key, key.tun_opts_len),
1479 		       key.tun_opts_len);
1480 		tun_info->options = (struct geneve_opt *)(tun_info + 1);
1481 	} else {
1482 		tun_info->options = NULL;
1483 	}
1484 
1485 	add_nested_action_end(*sfa, start);
1486 
1487 	return err;
1488 }
1489 
1490 static int validate_set(const struct nlattr *a,
1491 			const struct sw_flow_key *flow_key,
1492 			struct sw_flow_actions **sfa,
1493 			bool *set_tun)
1494 {
1495 	const struct nlattr *ovs_key = nla_data(a);
1496 	int key_type = nla_type(ovs_key);
1497 
1498 	/* There can be only one key in a action */
1499 	if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1500 		return -EINVAL;
1501 
1502 	if (key_type > OVS_KEY_ATTR_MAX ||
1503 	    (ovs_key_lens[key_type] != nla_len(ovs_key) &&
1504 	     ovs_key_lens[key_type] != -1))
1505 		return -EINVAL;
1506 
1507 	switch (key_type) {
1508 	const struct ovs_key_ipv4 *ipv4_key;
1509 	const struct ovs_key_ipv6 *ipv6_key;
1510 	int err;
1511 
1512 	case OVS_KEY_ATTR_PRIORITY:
1513 	case OVS_KEY_ATTR_SKB_MARK:
1514 	case OVS_KEY_ATTR_ETHERNET:
1515 		break;
1516 
1517 	case OVS_KEY_ATTR_TUNNEL:
1518 		*set_tun = true;
1519 		err = validate_and_copy_set_tun(a, sfa);
1520 		if (err)
1521 			return err;
1522 		break;
1523 
1524 	case OVS_KEY_ATTR_IPV4:
1525 		if (flow_key->eth.type != htons(ETH_P_IP))
1526 			return -EINVAL;
1527 
1528 		if (!flow_key->ip.proto)
1529 			return -EINVAL;
1530 
1531 		ipv4_key = nla_data(ovs_key);
1532 		if (ipv4_key->ipv4_proto != flow_key->ip.proto)
1533 			return -EINVAL;
1534 
1535 		if (ipv4_key->ipv4_frag != flow_key->ip.frag)
1536 			return -EINVAL;
1537 
1538 		break;
1539 
1540 	case OVS_KEY_ATTR_IPV6:
1541 		if (flow_key->eth.type != htons(ETH_P_IPV6))
1542 			return -EINVAL;
1543 
1544 		if (!flow_key->ip.proto)
1545 			return -EINVAL;
1546 
1547 		ipv6_key = nla_data(ovs_key);
1548 		if (ipv6_key->ipv6_proto != flow_key->ip.proto)
1549 			return -EINVAL;
1550 
1551 		if (ipv6_key->ipv6_frag != flow_key->ip.frag)
1552 			return -EINVAL;
1553 
1554 		if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
1555 			return -EINVAL;
1556 
1557 		break;
1558 
1559 	case OVS_KEY_ATTR_TCP:
1560 		if (flow_key->ip.proto != IPPROTO_TCP)
1561 			return -EINVAL;
1562 
1563 		return validate_tp_port(flow_key);
1564 
1565 	case OVS_KEY_ATTR_UDP:
1566 		if (flow_key->ip.proto != IPPROTO_UDP)
1567 			return -EINVAL;
1568 
1569 		return validate_tp_port(flow_key);
1570 
1571 	case OVS_KEY_ATTR_SCTP:
1572 		if (flow_key->ip.proto != IPPROTO_SCTP)
1573 			return -EINVAL;
1574 
1575 		return validate_tp_port(flow_key);
1576 
1577 	default:
1578 		return -EINVAL;
1579 	}
1580 
1581 	return 0;
1582 }
1583 
1584 static int validate_userspace(const struct nlattr *attr)
1585 {
1586 	static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
1587 		[OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
1588 		[OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
1589 	};
1590 	struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
1591 	int error;
1592 
1593 	error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
1594 				 attr, userspace_policy);
1595 	if (error)
1596 		return error;
1597 
1598 	if (!a[OVS_USERSPACE_ATTR_PID] ||
1599 	    !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
1600 		return -EINVAL;
1601 
1602 	return 0;
1603 }
1604 
1605 static int copy_action(const struct nlattr *from,
1606 		       struct sw_flow_actions **sfa)
1607 {
1608 	int totlen = NLA_ALIGN(from->nla_len);
1609 	struct nlattr *to;
1610 
1611 	to = reserve_sfa_size(sfa, from->nla_len);
1612 	if (IS_ERR(to))
1613 		return PTR_ERR(to);
1614 
1615 	memcpy(to, from, totlen);
1616 	return 0;
1617 }
1618 
1619 int ovs_nla_copy_actions(const struct nlattr *attr,
1620 			 const struct sw_flow_key *key,
1621 			 int depth,
1622 			 struct sw_flow_actions **sfa)
1623 {
1624 	const struct nlattr *a;
1625 	int rem, err;
1626 
1627 	if (depth >= SAMPLE_ACTION_DEPTH)
1628 		return -EOVERFLOW;
1629 
1630 	nla_for_each_nested(a, attr, rem) {
1631 		/* Expected argument lengths, (u32)-1 for variable length. */
1632 		static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
1633 			[OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
1634 			[OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
1635 			[OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
1636 			[OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
1637 			[OVS_ACTION_ATTR_POP_VLAN] = 0,
1638 			[OVS_ACTION_ATTR_SET] = (u32)-1,
1639 			[OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
1640 			[OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash)
1641 		};
1642 		const struct ovs_action_push_vlan *vlan;
1643 		int type = nla_type(a);
1644 		bool skip_copy;
1645 
1646 		if (type > OVS_ACTION_ATTR_MAX ||
1647 		    (action_lens[type] != nla_len(a) &&
1648 		     action_lens[type] != (u32)-1))
1649 			return -EINVAL;
1650 
1651 		skip_copy = false;
1652 		switch (type) {
1653 		case OVS_ACTION_ATTR_UNSPEC:
1654 			return -EINVAL;
1655 
1656 		case OVS_ACTION_ATTR_USERSPACE:
1657 			err = validate_userspace(a);
1658 			if (err)
1659 				return err;
1660 			break;
1661 
1662 		case OVS_ACTION_ATTR_OUTPUT:
1663 			if (nla_get_u32(a) >= DP_MAX_PORTS)
1664 				return -EINVAL;
1665 			break;
1666 
1667 		case OVS_ACTION_ATTR_HASH: {
1668 			const struct ovs_action_hash *act_hash = nla_data(a);
1669 
1670 			switch (act_hash->hash_alg) {
1671 			case OVS_HASH_ALG_L4:
1672 				break;
1673 			default:
1674 				return  -EINVAL;
1675 			}
1676 
1677 			break;
1678 		}
1679 
1680 		case OVS_ACTION_ATTR_POP_VLAN:
1681 			break;
1682 
1683 		case OVS_ACTION_ATTR_PUSH_VLAN:
1684 			vlan = nla_data(a);
1685 			if (vlan->vlan_tpid != htons(ETH_P_8021Q))
1686 				return -EINVAL;
1687 			if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
1688 				return -EINVAL;
1689 			break;
1690 
1691 		case OVS_ACTION_ATTR_RECIRC:
1692 			break;
1693 
1694 		case OVS_ACTION_ATTR_SET:
1695 			err = validate_set(a, key, sfa, &skip_copy);
1696 			if (err)
1697 				return err;
1698 			break;
1699 
1700 		case OVS_ACTION_ATTR_SAMPLE:
1701 			err = validate_and_copy_sample(a, key, depth, sfa);
1702 			if (err)
1703 				return err;
1704 			skip_copy = true;
1705 			break;
1706 
1707 		default:
1708 			return -EINVAL;
1709 		}
1710 		if (!skip_copy) {
1711 			err = copy_action(a, sfa);
1712 			if (err)
1713 				return err;
1714 		}
1715 	}
1716 
1717 	if (rem > 0)
1718 		return -EINVAL;
1719 
1720 	return 0;
1721 }
1722 
1723 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
1724 {
1725 	const struct nlattr *a;
1726 	struct nlattr *start;
1727 	int err = 0, rem;
1728 
1729 	start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
1730 	if (!start)
1731 		return -EMSGSIZE;
1732 
1733 	nla_for_each_nested(a, attr, rem) {
1734 		int type = nla_type(a);
1735 		struct nlattr *st_sample;
1736 
1737 		switch (type) {
1738 		case OVS_SAMPLE_ATTR_PROBABILITY:
1739 			if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
1740 				    sizeof(u32), nla_data(a)))
1741 				return -EMSGSIZE;
1742 			break;
1743 		case OVS_SAMPLE_ATTR_ACTIONS:
1744 			st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
1745 			if (!st_sample)
1746 				return -EMSGSIZE;
1747 			err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
1748 			if (err)
1749 				return err;
1750 			nla_nest_end(skb, st_sample);
1751 			break;
1752 		}
1753 	}
1754 
1755 	nla_nest_end(skb, start);
1756 	return err;
1757 }
1758 
1759 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
1760 {
1761 	const struct nlattr *ovs_key = nla_data(a);
1762 	int key_type = nla_type(ovs_key);
1763 	struct nlattr *start;
1764 	int err;
1765 
1766 	switch (key_type) {
1767 	case OVS_KEY_ATTR_TUNNEL_INFO: {
1768 		struct ovs_tunnel_info *tun_info = nla_data(ovs_key);
1769 
1770 		start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
1771 		if (!start)
1772 			return -EMSGSIZE;
1773 
1774 		err = ipv4_tun_to_nlattr(skb, &tun_info->tunnel,
1775 					 tun_info->options_len ?
1776 						tun_info->options : NULL,
1777 					 tun_info->options_len);
1778 		if (err)
1779 			return err;
1780 		nla_nest_end(skb, start);
1781 		break;
1782 	}
1783 	default:
1784 		if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
1785 			return -EMSGSIZE;
1786 		break;
1787 	}
1788 
1789 	return 0;
1790 }
1791 
1792 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
1793 {
1794 	const struct nlattr *a;
1795 	int rem, err;
1796 
1797 	nla_for_each_attr(a, attr, len, rem) {
1798 		int type = nla_type(a);
1799 
1800 		switch (type) {
1801 		case OVS_ACTION_ATTR_SET:
1802 			err = set_action_to_attr(a, skb);
1803 			if (err)
1804 				return err;
1805 			break;
1806 
1807 		case OVS_ACTION_ATTR_SAMPLE:
1808 			err = sample_action_to_attr(a, skb);
1809 			if (err)
1810 				return err;
1811 			break;
1812 		default:
1813 			if (nla_put(skb, type, nla_len(a), nla_data(a)))
1814 				return -EMSGSIZE;
1815 			break;
1816 		}
1817 	}
1818 
1819 	return 0;
1820 }
1821