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.eth.type == htons(0xffff)))
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 		SW_FLOW_KEY_PUT(match, ipv6.label,
693 				ipv6_key->ipv6_label, is_mask);
694 		SW_FLOW_KEY_PUT(match, ip.proto,
695 				ipv6_key->ipv6_proto, is_mask);
696 		SW_FLOW_KEY_PUT(match, ip.tos,
697 				ipv6_key->ipv6_tclass, is_mask);
698 		SW_FLOW_KEY_PUT(match, ip.ttl,
699 				ipv6_key->ipv6_hlimit, is_mask);
700 		SW_FLOW_KEY_PUT(match, ip.frag,
701 				ipv6_key->ipv6_frag, is_mask);
702 		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
703 				ipv6_key->ipv6_src,
704 				sizeof(match->key->ipv6.addr.src),
705 				is_mask);
706 		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
707 				ipv6_key->ipv6_dst,
708 				sizeof(match->key->ipv6.addr.dst),
709 				is_mask);
710 
711 		attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
712 	}
713 
714 	if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
715 		const struct ovs_key_arp *arp_key;
716 
717 		arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
718 		if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
719 			OVS_NLERR("Unknown ARP opcode (opcode=%d).\n",
720 				  arp_key->arp_op);
721 			return -EINVAL;
722 		}
723 
724 		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
725 				arp_key->arp_sip, is_mask);
726 		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
727 			arp_key->arp_tip, is_mask);
728 		SW_FLOW_KEY_PUT(match, ip.proto,
729 				ntohs(arp_key->arp_op), is_mask);
730 		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
731 				arp_key->arp_sha, ETH_ALEN, is_mask);
732 		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
733 				arp_key->arp_tha, ETH_ALEN, is_mask);
734 
735 		attrs &= ~(1 << OVS_KEY_ATTR_ARP);
736 	}
737 
738 	if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
739 		const struct ovs_key_tcp *tcp_key;
740 
741 		tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
742 		SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
743 		SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
744 		attrs &= ~(1 << OVS_KEY_ATTR_TCP);
745 	}
746 
747 	if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
748 		if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
749 			SW_FLOW_KEY_PUT(match, tp.flags,
750 					nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
751 					is_mask);
752 		} else {
753 			SW_FLOW_KEY_PUT(match, tp.flags,
754 					nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
755 					is_mask);
756 		}
757 		attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
758 	}
759 
760 	if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
761 		const struct ovs_key_udp *udp_key;
762 
763 		udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
764 		SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
765 		SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
766 		attrs &= ~(1 << OVS_KEY_ATTR_UDP);
767 	}
768 
769 	if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
770 		const struct ovs_key_sctp *sctp_key;
771 
772 		sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
773 		SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
774 		SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
775 		attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
776 	}
777 
778 	if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
779 		const struct ovs_key_icmp *icmp_key;
780 
781 		icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
782 		SW_FLOW_KEY_PUT(match, tp.src,
783 				htons(icmp_key->icmp_type), is_mask);
784 		SW_FLOW_KEY_PUT(match, tp.dst,
785 				htons(icmp_key->icmp_code), is_mask);
786 		attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
787 	}
788 
789 	if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
790 		const struct ovs_key_icmpv6 *icmpv6_key;
791 
792 		icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
793 		SW_FLOW_KEY_PUT(match, tp.src,
794 				htons(icmpv6_key->icmpv6_type), is_mask);
795 		SW_FLOW_KEY_PUT(match, tp.dst,
796 				htons(icmpv6_key->icmpv6_code), is_mask);
797 		attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
798 	}
799 
800 	if (attrs & (1 << OVS_KEY_ATTR_ND)) {
801 		const struct ovs_key_nd *nd_key;
802 
803 		nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
804 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
805 			nd_key->nd_target,
806 			sizeof(match->key->ipv6.nd.target),
807 			is_mask);
808 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
809 			nd_key->nd_sll, ETH_ALEN, is_mask);
810 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
811 				nd_key->nd_tll, ETH_ALEN, is_mask);
812 		attrs &= ~(1 << OVS_KEY_ATTR_ND);
813 	}
814 
815 	if (attrs != 0)
816 		return -EINVAL;
817 
818 	return 0;
819 }
820 
821 static void nlattr_set(struct nlattr *attr, u8 val, bool is_attr_mask_key)
822 {
823 	struct nlattr *nla;
824 	int rem;
825 
826 	/* The nlattr stream should already have been validated */
827 	nla_for_each_nested(nla, attr, rem) {
828 		/* We assume that ovs_key_lens[type] == -1 means that type is a
829 		 * nested attribute
830 		 */
831 		if (is_attr_mask_key && ovs_key_lens[nla_type(nla)] == -1)
832 			nlattr_set(nla, val, false);
833 		else
834 			memset(nla_data(nla), val, nla_len(nla));
835 	}
836 }
837 
838 static void mask_set_nlattr(struct nlattr *attr, u8 val)
839 {
840 	nlattr_set(attr, val, true);
841 }
842 
843 /**
844  * ovs_nla_get_match - parses Netlink attributes into a flow key and
845  * mask. In case the 'mask' is NULL, the flow is treated as exact match
846  * flow. Otherwise, it is treated as a wildcarded flow, except the mask
847  * does not include any don't care bit.
848  * @match: receives the extracted flow match information.
849  * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
850  * sequence. The fields should of the packet that triggered the creation
851  * of this flow.
852  * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
853  * attribute specifies the mask field of the wildcarded flow.
854  */
855 int ovs_nla_get_match(struct sw_flow_match *match,
856 		      const struct nlattr *key,
857 		      const struct nlattr *mask)
858 {
859 	const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
860 	const struct nlattr *encap;
861 	struct nlattr *newmask = NULL;
862 	u64 key_attrs = 0;
863 	u64 mask_attrs = 0;
864 	bool encap_valid = false;
865 	int err;
866 
867 	err = parse_flow_nlattrs(key, a, &key_attrs);
868 	if (err)
869 		return err;
870 
871 	if ((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
872 	    (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
873 	    (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
874 		__be16 tci;
875 
876 		if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
877 		      (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
878 			OVS_NLERR("Invalid Vlan frame.\n");
879 			return -EINVAL;
880 		}
881 
882 		key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
883 		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
884 		encap = a[OVS_KEY_ATTR_ENCAP];
885 		key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
886 		encap_valid = true;
887 
888 		if (tci & htons(VLAN_TAG_PRESENT)) {
889 			err = parse_flow_nlattrs(encap, a, &key_attrs);
890 			if (err)
891 				return err;
892 		} else if (!tci) {
893 			/* Corner case for truncated 802.1Q header. */
894 			if (nla_len(encap)) {
895 				OVS_NLERR("Truncated 802.1Q header has non-zero encap attribute.\n");
896 				return -EINVAL;
897 			}
898 		} else {
899 			OVS_NLERR("Encap attribute is set for a non-VLAN frame.\n");
900 			return  -EINVAL;
901 		}
902 	}
903 
904 	err = ovs_key_from_nlattrs(match, key_attrs, a, false);
905 	if (err)
906 		return err;
907 
908 	if (match->mask && !mask) {
909 		/* Create an exact match mask. We need to set to 0xff all the
910 		 * 'match->mask' fields that have been touched in 'match->key'.
911 		 * We cannot simply memset 'match->mask', because padding bytes
912 		 * and fields not specified in 'match->key' should be left to 0.
913 		 * Instead, we use a stream of netlink attributes, copied from
914 		 * 'key' and set to 0xff: ovs_key_from_nlattrs() will take care
915 		 * of filling 'match->mask' appropriately.
916 		 */
917 		newmask = kmemdup(key, nla_total_size(nla_len(key)),
918 				  GFP_KERNEL);
919 		if (!newmask)
920 			return -ENOMEM;
921 
922 		mask_set_nlattr(newmask, 0xff);
923 
924 		/* The userspace does not send tunnel attributes that are 0,
925 		 * but we should not wildcard them nonetheless.
926 		 */
927 		if (match->key->tun_key.ipv4_dst)
928 			SW_FLOW_KEY_MEMSET_FIELD(match, tun_key, 0xff, true);
929 
930 		mask = newmask;
931 	}
932 
933 	if (mask) {
934 		err = parse_flow_mask_nlattrs(mask, a, &mask_attrs);
935 		if (err)
936 			goto free_newmask;
937 
938 		if (mask_attrs & 1 << OVS_KEY_ATTR_ENCAP) {
939 			__be16 eth_type = 0;
940 			__be16 tci = 0;
941 
942 			if (!encap_valid) {
943 				OVS_NLERR("Encap mask attribute is set for non-VLAN frame.\n");
944 				err = -EINVAL;
945 				goto free_newmask;
946 			}
947 
948 			mask_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
949 			if (a[OVS_KEY_ATTR_ETHERTYPE])
950 				eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
951 
952 			if (eth_type == htons(0xffff)) {
953 				mask_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
954 				encap = a[OVS_KEY_ATTR_ENCAP];
955 				err = parse_flow_mask_nlattrs(encap, a, &mask_attrs);
956 				if (err)
957 					goto free_newmask;
958 			} else {
959 				OVS_NLERR("VLAN frames must have an exact match on the TPID (mask=%x).\n",
960 						ntohs(eth_type));
961 				err = -EINVAL;
962 				goto free_newmask;
963 			}
964 
965 			if (a[OVS_KEY_ATTR_VLAN])
966 				tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
967 
968 			if (!(tci & htons(VLAN_TAG_PRESENT))) {
969 				OVS_NLERR("VLAN tag present bit must have an exact match (tci_mask=%x).\n", ntohs(tci));
970 				err = -EINVAL;
971 				goto free_newmask;
972 			}
973 		}
974 
975 		err = ovs_key_from_nlattrs(match, mask_attrs, a, true);
976 		if (err)
977 			goto free_newmask;
978 	}
979 
980 	if (!match_validate(match, key_attrs, mask_attrs))
981 		err = -EINVAL;
982 
983 free_newmask:
984 	kfree(newmask);
985 	return err;
986 }
987 
988 /**
989  * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
990  * @key: Receives extracted in_port, priority, tun_key and skb_mark.
991  * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
992  * sequence.
993  *
994  * This parses a series of Netlink attributes that form a flow key, which must
995  * take the same form accepted by flow_from_nlattrs(), but only enough of it to
996  * get the metadata, that is, the parts of the flow key that cannot be
997  * extracted from the packet itself.
998  */
999 
1000 int ovs_nla_get_flow_metadata(const struct nlattr *attr,
1001 			      struct sw_flow_key *key)
1002 {
1003 	const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1004 	struct sw_flow_match match;
1005 	u64 attrs = 0;
1006 	int err;
1007 
1008 	err = parse_flow_nlattrs(attr, a, &attrs);
1009 	if (err)
1010 		return -EINVAL;
1011 
1012 	memset(&match, 0, sizeof(match));
1013 	match.key = key;
1014 
1015 	key->phy.in_port = DP_MAX_PORTS;
1016 
1017 	return metadata_from_nlattrs(&match, &attrs, a, false);
1018 }
1019 
1020 int ovs_nla_put_flow(const struct sw_flow_key *swkey,
1021 		     const struct sw_flow_key *output, struct sk_buff *skb)
1022 {
1023 	struct ovs_key_ethernet *eth_key;
1024 	struct nlattr *nla, *encap;
1025 	bool is_mask = (swkey != output);
1026 
1027 	if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1028 		goto nla_put_failure;
1029 
1030 	if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1031 		goto nla_put_failure;
1032 
1033 	if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1034 		goto nla_put_failure;
1035 
1036 	if ((swkey->tun_key.ipv4_dst || is_mask)) {
1037 		const struct geneve_opt *opts = NULL;
1038 
1039 		if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1040 			opts = GENEVE_OPTS(output, swkey->tun_opts_len);
1041 
1042 		if (ipv4_tun_to_nlattr(skb, &output->tun_key, opts,
1043 				       swkey->tun_opts_len))
1044 			goto nla_put_failure;
1045 	}
1046 
1047 	if (swkey->phy.in_port == DP_MAX_PORTS) {
1048 		if (is_mask && (output->phy.in_port == 0xffff))
1049 			if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1050 				goto nla_put_failure;
1051 	} else {
1052 		u16 upper_u16;
1053 		upper_u16 = !is_mask ? 0 : 0xffff;
1054 
1055 		if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1056 				(upper_u16 << 16) | output->phy.in_port))
1057 			goto nla_put_failure;
1058 	}
1059 
1060 	if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1061 		goto nla_put_failure;
1062 
1063 	nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1064 	if (!nla)
1065 		goto nla_put_failure;
1066 
1067 	eth_key = nla_data(nla);
1068 	ether_addr_copy(eth_key->eth_src, output->eth.src);
1069 	ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1070 
1071 	if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1072 		__be16 eth_type;
1073 		eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
1074 		if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1075 		    nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
1076 			goto nla_put_failure;
1077 		encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1078 		if (!swkey->eth.tci)
1079 			goto unencap;
1080 	} else
1081 		encap = NULL;
1082 
1083 	if (swkey->eth.type == htons(ETH_P_802_2)) {
1084 		/*
1085 		 * Ethertype 802.2 is represented in the netlink with omitted
1086 		 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1087 		 * 0xffff in the mask attribute.  Ethertype can also
1088 		 * be wildcarded.
1089 		 */
1090 		if (is_mask && output->eth.type)
1091 			if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1092 						output->eth.type))
1093 				goto nla_put_failure;
1094 		goto unencap;
1095 	}
1096 
1097 	if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1098 		goto nla_put_failure;
1099 
1100 	if (swkey->eth.type == htons(ETH_P_IP)) {
1101 		struct ovs_key_ipv4 *ipv4_key;
1102 
1103 		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1104 		if (!nla)
1105 			goto nla_put_failure;
1106 		ipv4_key = nla_data(nla);
1107 		ipv4_key->ipv4_src = output->ipv4.addr.src;
1108 		ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1109 		ipv4_key->ipv4_proto = output->ip.proto;
1110 		ipv4_key->ipv4_tos = output->ip.tos;
1111 		ipv4_key->ipv4_ttl = output->ip.ttl;
1112 		ipv4_key->ipv4_frag = output->ip.frag;
1113 	} else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1114 		struct ovs_key_ipv6 *ipv6_key;
1115 
1116 		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1117 		if (!nla)
1118 			goto nla_put_failure;
1119 		ipv6_key = nla_data(nla);
1120 		memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1121 				sizeof(ipv6_key->ipv6_src));
1122 		memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1123 				sizeof(ipv6_key->ipv6_dst));
1124 		ipv6_key->ipv6_label = output->ipv6.label;
1125 		ipv6_key->ipv6_proto = output->ip.proto;
1126 		ipv6_key->ipv6_tclass = output->ip.tos;
1127 		ipv6_key->ipv6_hlimit = output->ip.ttl;
1128 		ipv6_key->ipv6_frag = output->ip.frag;
1129 	} else if (swkey->eth.type == htons(ETH_P_ARP) ||
1130 		   swkey->eth.type == htons(ETH_P_RARP)) {
1131 		struct ovs_key_arp *arp_key;
1132 
1133 		nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1134 		if (!nla)
1135 			goto nla_put_failure;
1136 		arp_key = nla_data(nla);
1137 		memset(arp_key, 0, sizeof(struct ovs_key_arp));
1138 		arp_key->arp_sip = output->ipv4.addr.src;
1139 		arp_key->arp_tip = output->ipv4.addr.dst;
1140 		arp_key->arp_op = htons(output->ip.proto);
1141 		ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1142 		ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1143 	}
1144 
1145 	if ((swkey->eth.type == htons(ETH_P_IP) ||
1146 	     swkey->eth.type == htons(ETH_P_IPV6)) &&
1147 	     swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1148 
1149 		if (swkey->ip.proto == IPPROTO_TCP) {
1150 			struct ovs_key_tcp *tcp_key;
1151 
1152 			nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1153 			if (!nla)
1154 				goto nla_put_failure;
1155 			tcp_key = nla_data(nla);
1156 			tcp_key->tcp_src = output->tp.src;
1157 			tcp_key->tcp_dst = output->tp.dst;
1158 			if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1159 					 output->tp.flags))
1160 				goto nla_put_failure;
1161 		} else if (swkey->ip.proto == IPPROTO_UDP) {
1162 			struct ovs_key_udp *udp_key;
1163 
1164 			nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1165 			if (!nla)
1166 				goto nla_put_failure;
1167 			udp_key = nla_data(nla);
1168 			udp_key->udp_src = output->tp.src;
1169 			udp_key->udp_dst = output->tp.dst;
1170 		} else if (swkey->ip.proto == IPPROTO_SCTP) {
1171 			struct ovs_key_sctp *sctp_key;
1172 
1173 			nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1174 			if (!nla)
1175 				goto nla_put_failure;
1176 			sctp_key = nla_data(nla);
1177 			sctp_key->sctp_src = output->tp.src;
1178 			sctp_key->sctp_dst = output->tp.dst;
1179 		} else if (swkey->eth.type == htons(ETH_P_IP) &&
1180 			   swkey->ip.proto == IPPROTO_ICMP) {
1181 			struct ovs_key_icmp *icmp_key;
1182 
1183 			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1184 			if (!nla)
1185 				goto nla_put_failure;
1186 			icmp_key = nla_data(nla);
1187 			icmp_key->icmp_type = ntohs(output->tp.src);
1188 			icmp_key->icmp_code = ntohs(output->tp.dst);
1189 		} else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1190 			   swkey->ip.proto == IPPROTO_ICMPV6) {
1191 			struct ovs_key_icmpv6 *icmpv6_key;
1192 
1193 			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1194 						sizeof(*icmpv6_key));
1195 			if (!nla)
1196 				goto nla_put_failure;
1197 			icmpv6_key = nla_data(nla);
1198 			icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1199 			icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1200 
1201 			if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1202 			    icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1203 				struct ovs_key_nd *nd_key;
1204 
1205 				nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1206 				if (!nla)
1207 					goto nla_put_failure;
1208 				nd_key = nla_data(nla);
1209 				memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1210 							sizeof(nd_key->nd_target));
1211 				ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1212 				ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1213 			}
1214 		}
1215 	}
1216 
1217 unencap:
1218 	if (encap)
1219 		nla_nest_end(skb, encap);
1220 
1221 	return 0;
1222 
1223 nla_put_failure:
1224 	return -EMSGSIZE;
1225 }
1226 
1227 #define MAX_ACTIONS_BUFSIZE	(32 * 1024)
1228 
1229 struct sw_flow_actions *ovs_nla_alloc_flow_actions(int size)
1230 {
1231 	struct sw_flow_actions *sfa;
1232 
1233 	if (size > MAX_ACTIONS_BUFSIZE)
1234 		return ERR_PTR(-EINVAL);
1235 
1236 	sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1237 	if (!sfa)
1238 		return ERR_PTR(-ENOMEM);
1239 
1240 	sfa->actions_len = 0;
1241 	return sfa;
1242 }
1243 
1244 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1245  * The caller must hold rcu_read_lock for this to be sensible. */
1246 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1247 {
1248 	kfree_rcu(sf_acts, rcu);
1249 }
1250 
1251 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1252 				       int attr_len)
1253 {
1254 
1255 	struct sw_flow_actions *acts;
1256 	int new_acts_size;
1257 	int req_size = NLA_ALIGN(attr_len);
1258 	int next_offset = offsetof(struct sw_flow_actions, actions) +
1259 					(*sfa)->actions_len;
1260 
1261 	if (req_size <= (ksize(*sfa) - next_offset))
1262 		goto out;
1263 
1264 	new_acts_size = ksize(*sfa) * 2;
1265 
1266 	if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1267 		if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1268 			return ERR_PTR(-EMSGSIZE);
1269 		new_acts_size = MAX_ACTIONS_BUFSIZE;
1270 	}
1271 
1272 	acts = ovs_nla_alloc_flow_actions(new_acts_size);
1273 	if (IS_ERR(acts))
1274 		return (void *)acts;
1275 
1276 	memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1277 	acts->actions_len = (*sfa)->actions_len;
1278 	kfree(*sfa);
1279 	*sfa = acts;
1280 
1281 out:
1282 	(*sfa)->actions_len += req_size;
1283 	return  (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1284 }
1285 
1286 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1287 				   int attrtype, void *data, int len)
1288 {
1289 	struct nlattr *a;
1290 
1291 	a = reserve_sfa_size(sfa, nla_attr_size(len));
1292 	if (IS_ERR(a))
1293 		return a;
1294 
1295 	a->nla_type = attrtype;
1296 	a->nla_len = nla_attr_size(len);
1297 
1298 	if (data)
1299 		memcpy(nla_data(a), data, len);
1300 	memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1301 
1302 	return a;
1303 }
1304 
1305 static int add_action(struct sw_flow_actions **sfa, int attrtype,
1306 		      void *data, int len)
1307 {
1308 	struct nlattr *a;
1309 
1310 	a = __add_action(sfa, attrtype, data, len);
1311 	if (IS_ERR(a))
1312 		return PTR_ERR(a);
1313 
1314 	return 0;
1315 }
1316 
1317 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1318 					  int attrtype)
1319 {
1320 	int used = (*sfa)->actions_len;
1321 	int err;
1322 
1323 	err = add_action(sfa, attrtype, NULL, 0);
1324 	if (err)
1325 		return err;
1326 
1327 	return used;
1328 }
1329 
1330 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1331 					 int st_offset)
1332 {
1333 	struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1334 							       st_offset);
1335 
1336 	a->nla_len = sfa->actions_len - st_offset;
1337 }
1338 
1339 static int validate_and_copy_sample(const struct nlattr *attr,
1340 				    const struct sw_flow_key *key, int depth,
1341 				    struct sw_flow_actions **sfa)
1342 {
1343 	const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1344 	const struct nlattr *probability, *actions;
1345 	const struct nlattr *a;
1346 	int rem, start, err, st_acts;
1347 
1348 	memset(attrs, 0, sizeof(attrs));
1349 	nla_for_each_nested(a, attr, rem) {
1350 		int type = nla_type(a);
1351 		if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1352 			return -EINVAL;
1353 		attrs[type] = a;
1354 	}
1355 	if (rem)
1356 		return -EINVAL;
1357 
1358 	probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1359 	if (!probability || nla_len(probability) != sizeof(u32))
1360 		return -EINVAL;
1361 
1362 	actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1363 	if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1364 		return -EINVAL;
1365 
1366 	/* validation done, copy sample action. */
1367 	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE);
1368 	if (start < 0)
1369 		return start;
1370 	err = add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1371 			 nla_data(probability), sizeof(u32));
1372 	if (err)
1373 		return err;
1374 	st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS);
1375 	if (st_acts < 0)
1376 		return st_acts;
1377 
1378 	err = ovs_nla_copy_actions(actions, key, depth + 1, sfa);
1379 	if (err)
1380 		return err;
1381 
1382 	add_nested_action_end(*sfa, st_acts);
1383 	add_nested_action_end(*sfa, start);
1384 
1385 	return 0;
1386 }
1387 
1388 static int validate_tp_port(const struct sw_flow_key *flow_key)
1389 {
1390 	if ((flow_key->eth.type == htons(ETH_P_IP) ||
1391 	     flow_key->eth.type == htons(ETH_P_IPV6)) &&
1392 	    (flow_key->tp.src || flow_key->tp.dst))
1393 		return 0;
1394 
1395 	return -EINVAL;
1396 }
1397 
1398 void ovs_match_init(struct sw_flow_match *match,
1399 		    struct sw_flow_key *key,
1400 		    struct sw_flow_mask *mask)
1401 {
1402 	memset(match, 0, sizeof(*match));
1403 	match->key = key;
1404 	match->mask = mask;
1405 
1406 	memset(key, 0, sizeof(*key));
1407 
1408 	if (mask) {
1409 		memset(&mask->key, 0, sizeof(mask->key));
1410 		mask->range.start = mask->range.end = 0;
1411 	}
1412 }
1413 
1414 static int validate_and_copy_set_tun(const struct nlattr *attr,
1415 				     struct sw_flow_actions **sfa)
1416 {
1417 	struct sw_flow_match match;
1418 	struct sw_flow_key key;
1419 	struct ovs_tunnel_info *tun_info;
1420 	struct nlattr *a;
1421 	int err, start;
1422 
1423 	ovs_match_init(&match, &key, NULL);
1424 	err = ipv4_tun_from_nlattr(nla_data(attr), &match, false);
1425 	if (err)
1426 		return err;
1427 
1428 	if (key.tun_opts_len) {
1429 		struct geneve_opt *option = GENEVE_OPTS(&key,
1430 							key.tun_opts_len);
1431 		int opts_len = key.tun_opts_len;
1432 		bool crit_opt = false;
1433 
1434 		while (opts_len > 0) {
1435 			int len;
1436 
1437 			if (opts_len < sizeof(*option))
1438 				return -EINVAL;
1439 
1440 			len = sizeof(*option) + option->length * 4;
1441 			if (len > opts_len)
1442 				return -EINVAL;
1443 
1444 			crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
1445 
1446 			option = (struct geneve_opt *)((u8 *)option + len);
1447 			opts_len -= len;
1448 		};
1449 
1450 		key.tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
1451 	};
1452 
1453 	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET);
1454 	if (start < 0)
1455 		return start;
1456 
1457 	a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
1458 			 sizeof(*tun_info) + key.tun_opts_len);
1459 	if (IS_ERR(a))
1460 		return PTR_ERR(a);
1461 
1462 	tun_info = nla_data(a);
1463 	tun_info->tunnel = key.tun_key;
1464 	tun_info->options_len = key.tun_opts_len;
1465 
1466 	if (tun_info->options_len) {
1467 		/* We need to store the options in the action itself since
1468 		 * everything else will go away after flow setup. We can append
1469 		 * it to tun_info and then point there.
1470 		 */
1471 		memcpy((tun_info + 1), GENEVE_OPTS(&key, key.tun_opts_len),
1472 		       key.tun_opts_len);
1473 		tun_info->options = (struct geneve_opt *)(tun_info + 1);
1474 	} else {
1475 		tun_info->options = NULL;
1476 	}
1477 
1478 	add_nested_action_end(*sfa, start);
1479 
1480 	return err;
1481 }
1482 
1483 static int validate_set(const struct nlattr *a,
1484 			const struct sw_flow_key *flow_key,
1485 			struct sw_flow_actions **sfa,
1486 			bool *set_tun)
1487 {
1488 	const struct nlattr *ovs_key = nla_data(a);
1489 	int key_type = nla_type(ovs_key);
1490 
1491 	/* There can be only one key in a action */
1492 	if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1493 		return -EINVAL;
1494 
1495 	if (key_type > OVS_KEY_ATTR_MAX ||
1496 	    (ovs_key_lens[key_type] != nla_len(ovs_key) &&
1497 	     ovs_key_lens[key_type] != -1))
1498 		return -EINVAL;
1499 
1500 	switch (key_type) {
1501 	const struct ovs_key_ipv4 *ipv4_key;
1502 	const struct ovs_key_ipv6 *ipv6_key;
1503 	int err;
1504 
1505 	case OVS_KEY_ATTR_PRIORITY:
1506 	case OVS_KEY_ATTR_SKB_MARK:
1507 	case OVS_KEY_ATTR_ETHERNET:
1508 		break;
1509 
1510 	case OVS_KEY_ATTR_TUNNEL:
1511 		*set_tun = true;
1512 		err = validate_and_copy_set_tun(a, sfa);
1513 		if (err)
1514 			return err;
1515 		break;
1516 
1517 	case OVS_KEY_ATTR_IPV4:
1518 		if (flow_key->eth.type != htons(ETH_P_IP))
1519 			return -EINVAL;
1520 
1521 		if (!flow_key->ip.proto)
1522 			return -EINVAL;
1523 
1524 		ipv4_key = nla_data(ovs_key);
1525 		if (ipv4_key->ipv4_proto != flow_key->ip.proto)
1526 			return -EINVAL;
1527 
1528 		if (ipv4_key->ipv4_frag != flow_key->ip.frag)
1529 			return -EINVAL;
1530 
1531 		break;
1532 
1533 	case OVS_KEY_ATTR_IPV6:
1534 		if (flow_key->eth.type != htons(ETH_P_IPV6))
1535 			return -EINVAL;
1536 
1537 		if (!flow_key->ip.proto)
1538 			return -EINVAL;
1539 
1540 		ipv6_key = nla_data(ovs_key);
1541 		if (ipv6_key->ipv6_proto != flow_key->ip.proto)
1542 			return -EINVAL;
1543 
1544 		if (ipv6_key->ipv6_frag != flow_key->ip.frag)
1545 			return -EINVAL;
1546 
1547 		if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
1548 			return -EINVAL;
1549 
1550 		break;
1551 
1552 	case OVS_KEY_ATTR_TCP:
1553 		if (flow_key->ip.proto != IPPROTO_TCP)
1554 			return -EINVAL;
1555 
1556 		return validate_tp_port(flow_key);
1557 
1558 	case OVS_KEY_ATTR_UDP:
1559 		if (flow_key->ip.proto != IPPROTO_UDP)
1560 			return -EINVAL;
1561 
1562 		return validate_tp_port(flow_key);
1563 
1564 	case OVS_KEY_ATTR_SCTP:
1565 		if (flow_key->ip.proto != IPPROTO_SCTP)
1566 			return -EINVAL;
1567 
1568 		return validate_tp_port(flow_key);
1569 
1570 	default:
1571 		return -EINVAL;
1572 	}
1573 
1574 	return 0;
1575 }
1576 
1577 static int validate_userspace(const struct nlattr *attr)
1578 {
1579 	static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
1580 		[OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
1581 		[OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
1582 	};
1583 	struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
1584 	int error;
1585 
1586 	error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
1587 				 attr, userspace_policy);
1588 	if (error)
1589 		return error;
1590 
1591 	if (!a[OVS_USERSPACE_ATTR_PID] ||
1592 	    !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
1593 		return -EINVAL;
1594 
1595 	return 0;
1596 }
1597 
1598 static int copy_action(const struct nlattr *from,
1599 		       struct sw_flow_actions **sfa)
1600 {
1601 	int totlen = NLA_ALIGN(from->nla_len);
1602 	struct nlattr *to;
1603 
1604 	to = reserve_sfa_size(sfa, from->nla_len);
1605 	if (IS_ERR(to))
1606 		return PTR_ERR(to);
1607 
1608 	memcpy(to, from, totlen);
1609 	return 0;
1610 }
1611 
1612 int ovs_nla_copy_actions(const struct nlattr *attr,
1613 			 const struct sw_flow_key *key,
1614 			 int depth,
1615 			 struct sw_flow_actions **sfa)
1616 {
1617 	const struct nlattr *a;
1618 	int rem, err;
1619 
1620 	if (depth >= SAMPLE_ACTION_DEPTH)
1621 		return -EOVERFLOW;
1622 
1623 	nla_for_each_nested(a, attr, rem) {
1624 		/* Expected argument lengths, (u32)-1 for variable length. */
1625 		static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
1626 			[OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
1627 			[OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
1628 			[OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
1629 			[OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
1630 			[OVS_ACTION_ATTR_POP_VLAN] = 0,
1631 			[OVS_ACTION_ATTR_SET] = (u32)-1,
1632 			[OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
1633 			[OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash)
1634 		};
1635 		const struct ovs_action_push_vlan *vlan;
1636 		int type = nla_type(a);
1637 		bool skip_copy;
1638 
1639 		if (type > OVS_ACTION_ATTR_MAX ||
1640 		    (action_lens[type] != nla_len(a) &&
1641 		     action_lens[type] != (u32)-1))
1642 			return -EINVAL;
1643 
1644 		skip_copy = false;
1645 		switch (type) {
1646 		case OVS_ACTION_ATTR_UNSPEC:
1647 			return -EINVAL;
1648 
1649 		case OVS_ACTION_ATTR_USERSPACE:
1650 			err = validate_userspace(a);
1651 			if (err)
1652 				return err;
1653 			break;
1654 
1655 		case OVS_ACTION_ATTR_OUTPUT:
1656 			if (nla_get_u32(a) >= DP_MAX_PORTS)
1657 				return -EINVAL;
1658 			break;
1659 
1660 		case OVS_ACTION_ATTR_HASH: {
1661 			const struct ovs_action_hash *act_hash = nla_data(a);
1662 
1663 			switch (act_hash->hash_alg) {
1664 			case OVS_HASH_ALG_L4:
1665 				break;
1666 			default:
1667 				return  -EINVAL;
1668 			}
1669 
1670 			break;
1671 		}
1672 
1673 		case OVS_ACTION_ATTR_POP_VLAN:
1674 			break;
1675 
1676 		case OVS_ACTION_ATTR_PUSH_VLAN:
1677 			vlan = nla_data(a);
1678 			if (vlan->vlan_tpid != htons(ETH_P_8021Q))
1679 				return -EINVAL;
1680 			if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
1681 				return -EINVAL;
1682 			break;
1683 
1684 		case OVS_ACTION_ATTR_RECIRC:
1685 			break;
1686 
1687 		case OVS_ACTION_ATTR_SET:
1688 			err = validate_set(a, key, sfa, &skip_copy);
1689 			if (err)
1690 				return err;
1691 			break;
1692 
1693 		case OVS_ACTION_ATTR_SAMPLE:
1694 			err = validate_and_copy_sample(a, key, depth, sfa);
1695 			if (err)
1696 				return err;
1697 			skip_copy = true;
1698 			break;
1699 
1700 		default:
1701 			return -EINVAL;
1702 		}
1703 		if (!skip_copy) {
1704 			err = copy_action(a, sfa);
1705 			if (err)
1706 				return err;
1707 		}
1708 	}
1709 
1710 	if (rem > 0)
1711 		return -EINVAL;
1712 
1713 	return 0;
1714 }
1715 
1716 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
1717 {
1718 	const struct nlattr *a;
1719 	struct nlattr *start;
1720 	int err = 0, rem;
1721 
1722 	start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
1723 	if (!start)
1724 		return -EMSGSIZE;
1725 
1726 	nla_for_each_nested(a, attr, rem) {
1727 		int type = nla_type(a);
1728 		struct nlattr *st_sample;
1729 
1730 		switch (type) {
1731 		case OVS_SAMPLE_ATTR_PROBABILITY:
1732 			if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
1733 				    sizeof(u32), nla_data(a)))
1734 				return -EMSGSIZE;
1735 			break;
1736 		case OVS_SAMPLE_ATTR_ACTIONS:
1737 			st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
1738 			if (!st_sample)
1739 				return -EMSGSIZE;
1740 			err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
1741 			if (err)
1742 				return err;
1743 			nla_nest_end(skb, st_sample);
1744 			break;
1745 		}
1746 	}
1747 
1748 	nla_nest_end(skb, start);
1749 	return err;
1750 }
1751 
1752 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
1753 {
1754 	const struct nlattr *ovs_key = nla_data(a);
1755 	int key_type = nla_type(ovs_key);
1756 	struct nlattr *start;
1757 	int err;
1758 
1759 	switch (key_type) {
1760 	case OVS_KEY_ATTR_TUNNEL_INFO: {
1761 		struct ovs_tunnel_info *tun_info = nla_data(ovs_key);
1762 
1763 		start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
1764 		if (!start)
1765 			return -EMSGSIZE;
1766 
1767 		err = ipv4_tun_to_nlattr(skb, &tun_info->tunnel,
1768 					 tun_info->options_len ?
1769 						tun_info->options : NULL,
1770 					 tun_info->options_len);
1771 		if (err)
1772 			return err;
1773 		nla_nest_end(skb, start);
1774 		break;
1775 	}
1776 	default:
1777 		if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
1778 			return -EMSGSIZE;
1779 		break;
1780 	}
1781 
1782 	return 0;
1783 }
1784 
1785 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
1786 {
1787 	const struct nlattr *a;
1788 	int rem, err;
1789 
1790 	nla_for_each_attr(a, attr, len, rem) {
1791 		int type = nla_type(a);
1792 
1793 		switch (type) {
1794 		case OVS_ACTION_ATTR_SET:
1795 			err = set_action_to_attr(a, skb);
1796 			if (err)
1797 				return err;
1798 			break;
1799 
1800 		case OVS_ACTION_ATTR_SAMPLE:
1801 			err = sample_action_to_attr(a, skb);
1802 			if (err)
1803 				return err;
1804 			break;
1805 		default:
1806 			if (nla_put(skb, type, nla_len(a), nla_data(a)))
1807 				return -EMSGSIZE;
1808 			break;
1809 		}
1810 	}
1811 
1812 	return 0;
1813 }
1814