xref: /linux/net/openvswitch/flow_netlink.c (revision ca55b2fef3a9373fcfc30f82fd26bc7fccbda732)
1 /*
2  * Copyright (c) 2007-2014 Nicira, Inc.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of version 2 of the GNU General Public
6  * License as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful, but
9  * WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11  * General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public License
14  * along with this program; if not, write to the Free Software
15  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
16  * 02110-1301, USA
17  */
18 
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
20 
21 #include "flow.h"
22 #include "datapath.h"
23 #include <linux/uaccess.h>
24 #include <linux/netdevice.h>
25 #include <linux/etherdevice.h>
26 #include <linux/if_ether.h>
27 #include <linux/if_vlan.h>
28 #include <net/llc_pdu.h>
29 #include <linux/kernel.h>
30 #include <linux/jhash.h>
31 #include <linux/jiffies.h>
32 #include <linux/llc.h>
33 #include <linux/module.h>
34 #include <linux/in.h>
35 #include <linux/rcupdate.h>
36 #include <linux/if_arp.h>
37 #include <linux/ip.h>
38 #include <linux/ipv6.h>
39 #include <linux/sctp.h>
40 #include <linux/tcp.h>
41 #include <linux/udp.h>
42 #include <linux/icmp.h>
43 #include <linux/icmpv6.h>
44 #include <linux/rculist.h>
45 #include <net/geneve.h>
46 #include <net/ip.h>
47 #include <net/ipv6.h>
48 #include <net/ndisc.h>
49 #include <net/mpls.h>
50 #include <net/vxlan.h>
51 
52 #include "flow_netlink.h"
53 
54 struct ovs_len_tbl {
55 	int len;
56 	const struct ovs_len_tbl *next;
57 };
58 
59 #define OVS_ATTR_NESTED -1
60 #define OVS_ATTR_VARIABLE -2
61 
62 static void update_range(struct sw_flow_match *match,
63 			 size_t offset, size_t size, bool is_mask)
64 {
65 	struct sw_flow_key_range *range;
66 	size_t start = rounddown(offset, sizeof(long));
67 	size_t end = roundup(offset + size, sizeof(long));
68 
69 	if (!is_mask)
70 		range = &match->range;
71 	else
72 		range = &match->mask->range;
73 
74 	if (range->start == range->end) {
75 		range->start = start;
76 		range->end = end;
77 		return;
78 	}
79 
80 	if (range->start > start)
81 		range->start = start;
82 
83 	if (range->end < end)
84 		range->end = end;
85 }
86 
87 #define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
88 	do { \
89 		update_range(match, offsetof(struct sw_flow_key, field),    \
90 			     sizeof((match)->key->field), is_mask);	    \
91 		if (is_mask)						    \
92 			(match)->mask->key.field = value;		    \
93 		else							    \
94 			(match)->key->field = value;		            \
95 	} while (0)
96 
97 #define SW_FLOW_KEY_MEMCPY_OFFSET(match, offset, value_p, len, is_mask)	    \
98 	do {								    \
99 		update_range(match, offset, len, is_mask);		    \
100 		if (is_mask)						    \
101 			memcpy((u8 *)&(match)->mask->key + offset, value_p, \
102 			       len);					   \
103 		else							    \
104 			memcpy((u8 *)(match)->key + offset, value_p, len);  \
105 	} while (0)
106 
107 #define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask)		      \
108 	SW_FLOW_KEY_MEMCPY_OFFSET(match, offsetof(struct sw_flow_key, field), \
109 				  value_p, len, is_mask)
110 
111 #define SW_FLOW_KEY_MEMSET_FIELD(match, field, value, is_mask)		    \
112 	do {								    \
113 		update_range(match, offsetof(struct sw_flow_key, field),    \
114 			     sizeof((match)->key->field), is_mask);	    \
115 		if (is_mask)						    \
116 			memset((u8 *)&(match)->mask->key.field, value,      \
117 			       sizeof((match)->mask->key.field));	    \
118 		else							    \
119 			memset((u8 *)&(match)->key->field, value,           \
120 			       sizeof((match)->key->field));                \
121 	} while (0)
122 
123 static bool match_validate(const struct sw_flow_match *match,
124 			   u64 key_attrs, u64 mask_attrs, bool log)
125 {
126 	u64 key_expected = 1 << OVS_KEY_ATTR_ETHERNET;
127 	u64 mask_allowed = key_attrs;  /* At most allow all key attributes */
128 
129 	/* The following mask attributes allowed only if they
130 	 * pass the validation tests. */
131 	mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
132 			| (1 << OVS_KEY_ATTR_IPV6)
133 			| (1 << OVS_KEY_ATTR_TCP)
134 			| (1 << OVS_KEY_ATTR_TCP_FLAGS)
135 			| (1 << OVS_KEY_ATTR_UDP)
136 			| (1 << OVS_KEY_ATTR_SCTP)
137 			| (1 << OVS_KEY_ATTR_ICMP)
138 			| (1 << OVS_KEY_ATTR_ICMPV6)
139 			| (1 << OVS_KEY_ATTR_ARP)
140 			| (1 << OVS_KEY_ATTR_ND)
141 			| (1 << OVS_KEY_ATTR_MPLS));
142 
143 	/* Always allowed mask fields. */
144 	mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
145 		       | (1 << OVS_KEY_ATTR_IN_PORT)
146 		       | (1 << OVS_KEY_ATTR_ETHERTYPE));
147 
148 	/* Check key attributes. */
149 	if (match->key->eth.type == htons(ETH_P_ARP)
150 			|| match->key->eth.type == htons(ETH_P_RARP)) {
151 		key_expected |= 1 << OVS_KEY_ATTR_ARP;
152 		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
153 			mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
154 	}
155 
156 	if (eth_p_mpls(match->key->eth.type)) {
157 		key_expected |= 1 << OVS_KEY_ATTR_MPLS;
158 		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
159 			mask_allowed |= 1 << OVS_KEY_ATTR_MPLS;
160 	}
161 
162 	if (match->key->eth.type == htons(ETH_P_IP)) {
163 		key_expected |= 1 << OVS_KEY_ATTR_IPV4;
164 		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
165 			mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
166 
167 		if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
168 			if (match->key->ip.proto == IPPROTO_UDP) {
169 				key_expected |= 1 << OVS_KEY_ATTR_UDP;
170 				if (match->mask && (match->mask->key.ip.proto == 0xff))
171 					mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
172 			}
173 
174 			if (match->key->ip.proto == IPPROTO_SCTP) {
175 				key_expected |= 1 << OVS_KEY_ATTR_SCTP;
176 				if (match->mask && (match->mask->key.ip.proto == 0xff))
177 					mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
178 			}
179 
180 			if (match->key->ip.proto == IPPROTO_TCP) {
181 				key_expected |= 1 << OVS_KEY_ATTR_TCP;
182 				key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
183 				if (match->mask && (match->mask->key.ip.proto == 0xff)) {
184 					mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
185 					mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
186 				}
187 			}
188 
189 			if (match->key->ip.proto == IPPROTO_ICMP) {
190 				key_expected |= 1 << OVS_KEY_ATTR_ICMP;
191 				if (match->mask && (match->mask->key.ip.proto == 0xff))
192 					mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
193 			}
194 		}
195 	}
196 
197 	if (match->key->eth.type == htons(ETH_P_IPV6)) {
198 		key_expected |= 1 << OVS_KEY_ATTR_IPV6;
199 		if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
200 			mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
201 
202 		if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
203 			if (match->key->ip.proto == IPPROTO_UDP) {
204 				key_expected |= 1 << OVS_KEY_ATTR_UDP;
205 				if (match->mask && (match->mask->key.ip.proto == 0xff))
206 					mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
207 			}
208 
209 			if (match->key->ip.proto == IPPROTO_SCTP) {
210 				key_expected |= 1 << OVS_KEY_ATTR_SCTP;
211 				if (match->mask && (match->mask->key.ip.proto == 0xff))
212 					mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
213 			}
214 
215 			if (match->key->ip.proto == IPPROTO_TCP) {
216 				key_expected |= 1 << OVS_KEY_ATTR_TCP;
217 				key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
218 				if (match->mask && (match->mask->key.ip.proto == 0xff)) {
219 					mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
220 					mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
221 				}
222 			}
223 
224 			if (match->key->ip.proto == IPPROTO_ICMPV6) {
225 				key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
226 				if (match->mask && (match->mask->key.ip.proto == 0xff))
227 					mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
228 
229 				if (match->key->tp.src ==
230 						htons(NDISC_NEIGHBOUR_SOLICITATION) ||
231 				    match->key->tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
232 					key_expected |= 1 << OVS_KEY_ATTR_ND;
233 					if (match->mask && (match->mask->key.tp.src == htons(0xff)))
234 						mask_allowed |= 1 << OVS_KEY_ATTR_ND;
235 				}
236 			}
237 		}
238 	}
239 
240 	if ((key_attrs & key_expected) != key_expected) {
241 		/* Key attributes check failed. */
242 		OVS_NLERR(log, "Missing key (keys=%llx, expected=%llx)",
243 			  (unsigned long long)key_attrs,
244 			  (unsigned long long)key_expected);
245 		return false;
246 	}
247 
248 	if ((mask_attrs & mask_allowed) != mask_attrs) {
249 		/* Mask attributes check failed. */
250 		OVS_NLERR(log, "Unexpected mask (mask=%llx, allowed=%llx)",
251 			  (unsigned long long)mask_attrs,
252 			  (unsigned long long)mask_allowed);
253 		return false;
254 	}
255 
256 	return true;
257 }
258 
259 size_t ovs_tun_key_attr_size(void)
260 {
261 	/* Whenever adding new OVS_TUNNEL_KEY_ FIELDS, we should consider
262 	 * updating this function.
263 	 */
264 	return    nla_total_size(8)    /* OVS_TUNNEL_KEY_ATTR_ID */
265 		+ nla_total_size(4)    /* OVS_TUNNEL_KEY_ATTR_IPV4_SRC */
266 		+ nla_total_size(4)    /* OVS_TUNNEL_KEY_ATTR_IPV4_DST */
267 		+ nla_total_size(1)    /* OVS_TUNNEL_KEY_ATTR_TOS */
268 		+ nla_total_size(1)    /* OVS_TUNNEL_KEY_ATTR_TTL */
269 		+ nla_total_size(0)    /* OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT */
270 		+ nla_total_size(0)    /* OVS_TUNNEL_KEY_ATTR_CSUM */
271 		+ nla_total_size(0)    /* OVS_TUNNEL_KEY_ATTR_OAM */
272 		+ nla_total_size(256)  /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
273 		/* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS is mutually exclusive with
274 		 * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
275 		 */
276 		+ nla_total_size(2)    /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
277 		+ nla_total_size(2);   /* OVS_TUNNEL_KEY_ATTR_TP_DST */
278 }
279 
280 size_t ovs_key_attr_size(void)
281 {
282 	/* Whenever adding new OVS_KEY_ FIELDS, we should consider
283 	 * updating this function.
284 	 */
285 	BUILD_BUG_ON(OVS_KEY_ATTR_TUNNEL_INFO != 26);
286 
287 	return    nla_total_size(4)   /* OVS_KEY_ATTR_PRIORITY */
288 		+ nla_total_size(0)   /* OVS_KEY_ATTR_TUNNEL */
289 		  + ovs_tun_key_attr_size()
290 		+ nla_total_size(4)   /* OVS_KEY_ATTR_IN_PORT */
291 		+ nla_total_size(4)   /* OVS_KEY_ATTR_SKB_MARK */
292 		+ nla_total_size(4)   /* OVS_KEY_ATTR_DP_HASH */
293 		+ nla_total_size(4)   /* OVS_KEY_ATTR_RECIRC_ID */
294 		+ nla_total_size(4)   /* OVS_KEY_ATTR_CT_STATE */
295 		+ nla_total_size(2)   /* OVS_KEY_ATTR_CT_ZONE */
296 		+ nla_total_size(4)   /* OVS_KEY_ATTR_CT_MARK */
297 		+ nla_total_size(16)  /* OVS_KEY_ATTR_CT_LABELS */
298 		+ nla_total_size(12)  /* OVS_KEY_ATTR_ETHERNET */
299 		+ nla_total_size(2)   /* OVS_KEY_ATTR_ETHERTYPE */
300 		+ nla_total_size(4)   /* OVS_KEY_ATTR_VLAN */
301 		+ nla_total_size(0)   /* OVS_KEY_ATTR_ENCAP */
302 		+ nla_total_size(2)   /* OVS_KEY_ATTR_ETHERTYPE */
303 		+ nla_total_size(40)  /* OVS_KEY_ATTR_IPV6 */
304 		+ nla_total_size(2)   /* OVS_KEY_ATTR_ICMPV6 */
305 		+ nla_total_size(28); /* OVS_KEY_ATTR_ND */
306 }
307 
308 static const struct ovs_len_tbl ovs_vxlan_ext_key_lens[OVS_VXLAN_EXT_MAX + 1] = {
309 	[OVS_VXLAN_EXT_GBP]	    = { .len = sizeof(u32) },
310 };
311 
312 static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
313 	[OVS_TUNNEL_KEY_ATTR_ID]	    = { .len = sizeof(u64) },
314 	[OVS_TUNNEL_KEY_ATTR_IPV4_SRC]	    = { .len = sizeof(u32) },
315 	[OVS_TUNNEL_KEY_ATTR_IPV4_DST]	    = { .len = sizeof(u32) },
316 	[OVS_TUNNEL_KEY_ATTR_TOS]	    = { .len = 1 },
317 	[OVS_TUNNEL_KEY_ATTR_TTL]	    = { .len = 1 },
318 	[OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
319 	[OVS_TUNNEL_KEY_ATTR_CSUM]	    = { .len = 0 },
320 	[OVS_TUNNEL_KEY_ATTR_TP_SRC]	    = { .len = sizeof(u16) },
321 	[OVS_TUNNEL_KEY_ATTR_TP_DST]	    = { .len = sizeof(u16) },
322 	[OVS_TUNNEL_KEY_ATTR_OAM]	    = { .len = 0 },
323 	[OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS]   = { .len = OVS_ATTR_VARIABLE },
324 	[OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS]    = { .len = OVS_ATTR_NESTED,
325 						.next = ovs_vxlan_ext_key_lens },
326 };
327 
328 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute.  */
329 static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
330 	[OVS_KEY_ATTR_ENCAP]	 = { .len = OVS_ATTR_NESTED },
331 	[OVS_KEY_ATTR_PRIORITY]	 = { .len = sizeof(u32) },
332 	[OVS_KEY_ATTR_IN_PORT]	 = { .len = sizeof(u32) },
333 	[OVS_KEY_ATTR_SKB_MARK]	 = { .len = sizeof(u32) },
334 	[OVS_KEY_ATTR_ETHERNET]	 = { .len = sizeof(struct ovs_key_ethernet) },
335 	[OVS_KEY_ATTR_VLAN]	 = { .len = sizeof(__be16) },
336 	[OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
337 	[OVS_KEY_ATTR_IPV4]	 = { .len = sizeof(struct ovs_key_ipv4) },
338 	[OVS_KEY_ATTR_IPV6]	 = { .len = sizeof(struct ovs_key_ipv6) },
339 	[OVS_KEY_ATTR_TCP]	 = { .len = sizeof(struct ovs_key_tcp) },
340 	[OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
341 	[OVS_KEY_ATTR_UDP]	 = { .len = sizeof(struct ovs_key_udp) },
342 	[OVS_KEY_ATTR_SCTP]	 = { .len = sizeof(struct ovs_key_sctp) },
343 	[OVS_KEY_ATTR_ICMP]	 = { .len = sizeof(struct ovs_key_icmp) },
344 	[OVS_KEY_ATTR_ICMPV6]	 = { .len = sizeof(struct ovs_key_icmpv6) },
345 	[OVS_KEY_ATTR_ARP]	 = { .len = sizeof(struct ovs_key_arp) },
346 	[OVS_KEY_ATTR_ND]	 = { .len = sizeof(struct ovs_key_nd) },
347 	[OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
348 	[OVS_KEY_ATTR_DP_HASH]	 = { .len = sizeof(u32) },
349 	[OVS_KEY_ATTR_TUNNEL]	 = { .len = OVS_ATTR_NESTED,
350 				     .next = ovs_tunnel_key_lens, },
351 	[OVS_KEY_ATTR_MPLS]	 = { .len = sizeof(struct ovs_key_mpls) },
352 	[OVS_KEY_ATTR_CT_STATE]	 = { .len = sizeof(u32) },
353 	[OVS_KEY_ATTR_CT_ZONE]	 = { .len = sizeof(u16) },
354 	[OVS_KEY_ATTR_CT_MARK]	 = { .len = sizeof(u32) },
355 	[OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
356 };
357 
358 static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
359 {
360 	return expected_len == attr_len ||
361 	       expected_len == OVS_ATTR_NESTED ||
362 	       expected_len == OVS_ATTR_VARIABLE;
363 }
364 
365 static bool is_all_zero(const u8 *fp, size_t size)
366 {
367 	int i;
368 
369 	if (!fp)
370 		return false;
371 
372 	for (i = 0; i < size; i++)
373 		if (fp[i])
374 			return false;
375 
376 	return true;
377 }
378 
379 static int __parse_flow_nlattrs(const struct nlattr *attr,
380 				const struct nlattr *a[],
381 				u64 *attrsp, bool log, bool nz)
382 {
383 	const struct nlattr *nla;
384 	u64 attrs;
385 	int rem;
386 
387 	attrs = *attrsp;
388 	nla_for_each_nested(nla, attr, rem) {
389 		u16 type = nla_type(nla);
390 		int expected_len;
391 
392 		if (type > OVS_KEY_ATTR_MAX) {
393 			OVS_NLERR(log, "Key type %d is out of range max %d",
394 				  type, OVS_KEY_ATTR_MAX);
395 			return -EINVAL;
396 		}
397 
398 		if (attrs & (1 << type)) {
399 			OVS_NLERR(log, "Duplicate key (type %d).", type);
400 			return -EINVAL;
401 		}
402 
403 		expected_len = ovs_key_lens[type].len;
404 		if (!check_attr_len(nla_len(nla), expected_len)) {
405 			OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
406 				  type, nla_len(nla), expected_len);
407 			return -EINVAL;
408 		}
409 
410 		if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
411 			attrs |= 1 << type;
412 			a[type] = nla;
413 		}
414 	}
415 	if (rem) {
416 		OVS_NLERR(log, "Message has %d unknown bytes.", rem);
417 		return -EINVAL;
418 	}
419 
420 	*attrsp = attrs;
421 	return 0;
422 }
423 
424 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
425 				   const struct nlattr *a[], u64 *attrsp,
426 				   bool log)
427 {
428 	return __parse_flow_nlattrs(attr, a, attrsp, log, true);
429 }
430 
431 static int parse_flow_nlattrs(const struct nlattr *attr,
432 			      const struct nlattr *a[], u64 *attrsp,
433 			      bool log)
434 {
435 	return __parse_flow_nlattrs(attr, a, attrsp, log, false);
436 }
437 
438 static int genev_tun_opt_from_nlattr(const struct nlattr *a,
439 				     struct sw_flow_match *match, bool is_mask,
440 				     bool log)
441 {
442 	unsigned long opt_key_offset;
443 
444 	if (nla_len(a) > sizeof(match->key->tun_opts)) {
445 		OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
446 			  nla_len(a), sizeof(match->key->tun_opts));
447 		return -EINVAL;
448 	}
449 
450 	if (nla_len(a) % 4 != 0) {
451 		OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
452 			  nla_len(a));
453 		return -EINVAL;
454 	}
455 
456 	/* We need to record the length of the options passed
457 	 * down, otherwise packets with the same format but
458 	 * additional options will be silently matched.
459 	 */
460 	if (!is_mask) {
461 		SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
462 				false);
463 	} else {
464 		/* This is somewhat unusual because it looks at
465 		 * both the key and mask while parsing the
466 		 * attributes (and by extension assumes the key
467 		 * is parsed first). Normally, we would verify
468 		 * that each is the correct length and that the
469 		 * attributes line up in the validate function.
470 		 * However, that is difficult because this is
471 		 * variable length and we won't have the
472 		 * information later.
473 		 */
474 		if (match->key->tun_opts_len != nla_len(a)) {
475 			OVS_NLERR(log, "Geneve option len %d != mask len %d",
476 				  match->key->tun_opts_len, nla_len(a));
477 			return -EINVAL;
478 		}
479 
480 		SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
481 	}
482 
483 	opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
484 	SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
485 				  nla_len(a), is_mask);
486 	return 0;
487 }
488 
489 static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
490 				     struct sw_flow_match *match, bool is_mask,
491 				     bool log)
492 {
493 	struct nlattr *a;
494 	int rem;
495 	unsigned long opt_key_offset;
496 	struct vxlan_metadata opts;
497 
498 	BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
499 
500 	memset(&opts, 0, sizeof(opts));
501 	nla_for_each_nested(a, attr, rem) {
502 		int type = nla_type(a);
503 
504 		if (type > OVS_VXLAN_EXT_MAX) {
505 			OVS_NLERR(log, "VXLAN extension %d out of range max %d",
506 				  type, OVS_VXLAN_EXT_MAX);
507 			return -EINVAL;
508 		}
509 
510 		if (!check_attr_len(nla_len(a),
511 				    ovs_vxlan_ext_key_lens[type].len)) {
512 			OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
513 				  type, nla_len(a),
514 				  ovs_vxlan_ext_key_lens[type].len);
515 			return -EINVAL;
516 		}
517 
518 		switch (type) {
519 		case OVS_VXLAN_EXT_GBP:
520 			opts.gbp = nla_get_u32(a);
521 			break;
522 		default:
523 			OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
524 				  type);
525 			return -EINVAL;
526 		}
527 	}
528 	if (rem) {
529 		OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
530 			  rem);
531 		return -EINVAL;
532 	}
533 
534 	if (!is_mask)
535 		SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
536 	else
537 		SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
538 
539 	opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
540 	SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
541 				  is_mask);
542 	return 0;
543 }
544 
545 static int ipv4_tun_from_nlattr(const struct nlattr *attr,
546 				struct sw_flow_match *match, bool is_mask,
547 				bool log)
548 {
549 	struct nlattr *a;
550 	int rem;
551 	bool ttl = false;
552 	__be16 tun_flags = 0;
553 	int opts_type = 0;
554 
555 	nla_for_each_nested(a, attr, rem) {
556 		int type = nla_type(a);
557 		int err;
558 
559 		if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
560 			OVS_NLERR(log, "Tunnel attr %d out of range max %d",
561 				  type, OVS_TUNNEL_KEY_ATTR_MAX);
562 			return -EINVAL;
563 		}
564 
565 		if (!check_attr_len(nla_len(a),
566 				    ovs_tunnel_key_lens[type].len)) {
567 			OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
568 				  type, nla_len(a), ovs_tunnel_key_lens[type].len);
569 			return -EINVAL;
570 		}
571 
572 		switch (type) {
573 		case OVS_TUNNEL_KEY_ATTR_ID:
574 			SW_FLOW_KEY_PUT(match, tun_key.tun_id,
575 					nla_get_be64(a), is_mask);
576 			tun_flags |= TUNNEL_KEY;
577 			break;
578 		case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
579 			SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
580 					nla_get_in_addr(a), is_mask);
581 			break;
582 		case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
583 			SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
584 					nla_get_in_addr(a), is_mask);
585 			break;
586 		case OVS_TUNNEL_KEY_ATTR_TOS:
587 			SW_FLOW_KEY_PUT(match, tun_key.tos,
588 					nla_get_u8(a), is_mask);
589 			break;
590 		case OVS_TUNNEL_KEY_ATTR_TTL:
591 			SW_FLOW_KEY_PUT(match, tun_key.ttl,
592 					nla_get_u8(a), is_mask);
593 			ttl = true;
594 			break;
595 		case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
596 			tun_flags |= TUNNEL_DONT_FRAGMENT;
597 			break;
598 		case OVS_TUNNEL_KEY_ATTR_CSUM:
599 			tun_flags |= TUNNEL_CSUM;
600 			break;
601 		case OVS_TUNNEL_KEY_ATTR_TP_SRC:
602 			SW_FLOW_KEY_PUT(match, tun_key.tp_src,
603 					nla_get_be16(a), is_mask);
604 			break;
605 		case OVS_TUNNEL_KEY_ATTR_TP_DST:
606 			SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
607 					nla_get_be16(a), is_mask);
608 			break;
609 		case OVS_TUNNEL_KEY_ATTR_OAM:
610 			tun_flags |= TUNNEL_OAM;
611 			break;
612 		case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
613 			if (opts_type) {
614 				OVS_NLERR(log, "Multiple metadata blocks provided");
615 				return -EINVAL;
616 			}
617 
618 			err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
619 			if (err)
620 				return err;
621 
622 			tun_flags |= TUNNEL_GENEVE_OPT;
623 			opts_type = type;
624 			break;
625 		case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
626 			if (opts_type) {
627 				OVS_NLERR(log, "Multiple metadata blocks provided");
628 				return -EINVAL;
629 			}
630 
631 			err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
632 			if (err)
633 				return err;
634 
635 			tun_flags |= TUNNEL_VXLAN_OPT;
636 			opts_type = type;
637 			break;
638 		default:
639 			OVS_NLERR(log, "Unknown IPv4 tunnel attribute %d",
640 				  type);
641 			return -EINVAL;
642 		}
643 	}
644 
645 	SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
646 
647 	if (rem > 0) {
648 		OVS_NLERR(log, "IPv4 tunnel attribute has %d unknown bytes.",
649 			  rem);
650 		return -EINVAL;
651 	}
652 
653 	if (!is_mask) {
654 		if (!match->key->tun_key.u.ipv4.dst) {
655 			OVS_NLERR(log, "IPv4 tunnel dst address is zero");
656 			return -EINVAL;
657 		}
658 
659 		if (!ttl) {
660 			OVS_NLERR(log, "IPv4 tunnel TTL not specified.");
661 			return -EINVAL;
662 		}
663 	}
664 
665 	return opts_type;
666 }
667 
668 static int vxlan_opt_to_nlattr(struct sk_buff *skb,
669 			       const void *tun_opts, int swkey_tun_opts_len)
670 {
671 	const struct vxlan_metadata *opts = tun_opts;
672 	struct nlattr *nla;
673 
674 	nla = nla_nest_start(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
675 	if (!nla)
676 		return -EMSGSIZE;
677 
678 	if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
679 		return -EMSGSIZE;
680 
681 	nla_nest_end(skb, nla);
682 	return 0;
683 }
684 
685 static int __ipv4_tun_to_nlattr(struct sk_buff *skb,
686 				const struct ip_tunnel_key *output,
687 				const void *tun_opts, int swkey_tun_opts_len)
688 {
689 	if (output->tun_flags & TUNNEL_KEY &&
690 	    nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
691 		return -EMSGSIZE;
692 	if (output->u.ipv4.src &&
693 	    nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
694 			    output->u.ipv4.src))
695 		return -EMSGSIZE;
696 	if (output->u.ipv4.dst &&
697 	    nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
698 			    output->u.ipv4.dst))
699 		return -EMSGSIZE;
700 	if (output->tos &&
701 	    nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
702 		return -EMSGSIZE;
703 	if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
704 		return -EMSGSIZE;
705 	if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
706 	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
707 		return -EMSGSIZE;
708 	if ((output->tun_flags & TUNNEL_CSUM) &&
709 	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
710 		return -EMSGSIZE;
711 	if (output->tp_src &&
712 	    nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
713 		return -EMSGSIZE;
714 	if (output->tp_dst &&
715 	    nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
716 		return -EMSGSIZE;
717 	if ((output->tun_flags & TUNNEL_OAM) &&
718 	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
719 		return -EMSGSIZE;
720 	if (swkey_tun_opts_len) {
721 		if (output->tun_flags & TUNNEL_GENEVE_OPT &&
722 		    nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
723 			    swkey_tun_opts_len, tun_opts))
724 			return -EMSGSIZE;
725 		else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
726 			 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
727 			return -EMSGSIZE;
728 	}
729 
730 	return 0;
731 }
732 
733 static int ipv4_tun_to_nlattr(struct sk_buff *skb,
734 			      const struct ip_tunnel_key *output,
735 			      const void *tun_opts, int swkey_tun_opts_len)
736 {
737 	struct nlattr *nla;
738 	int err;
739 
740 	nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
741 	if (!nla)
742 		return -EMSGSIZE;
743 
744 	err = __ipv4_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len);
745 	if (err)
746 		return err;
747 
748 	nla_nest_end(skb, nla);
749 	return 0;
750 }
751 
752 int ovs_nla_put_tunnel_info(struct sk_buff *skb,
753 			    struct ip_tunnel_info *tun_info)
754 {
755 	return __ipv4_tun_to_nlattr(skb, &tun_info->key,
756 				    ip_tunnel_info_opts(tun_info),
757 				    tun_info->options_len);
758 }
759 
760 static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
761 				 u64 *attrs, const struct nlattr **a,
762 				 bool is_mask, bool log)
763 {
764 	if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
765 		u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
766 
767 		SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
768 		*attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
769 	}
770 
771 	if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
772 		u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
773 
774 		SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
775 		*attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
776 	}
777 
778 	if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
779 		SW_FLOW_KEY_PUT(match, phy.priority,
780 			  nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
781 		*attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
782 	}
783 
784 	if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
785 		u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
786 
787 		if (is_mask) {
788 			in_port = 0xffffffff; /* Always exact match in_port. */
789 		} else if (in_port >= DP_MAX_PORTS) {
790 			OVS_NLERR(log, "Port %d exceeds max allowable %d",
791 				  in_port, DP_MAX_PORTS);
792 			return -EINVAL;
793 		}
794 
795 		SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
796 		*attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
797 	} else if (!is_mask) {
798 		SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
799 	}
800 
801 	if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
802 		uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
803 
804 		SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
805 		*attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
806 	}
807 	if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
808 		if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
809 					 is_mask, log) < 0)
810 			return -EINVAL;
811 		*attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
812 	}
813 
814 	if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
815 	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
816 		u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
817 
818 		if (ct_state & ~CT_SUPPORTED_MASK) {
819 			OVS_NLERR(log, "ct_state flags %08x unsupported",
820 				  ct_state);
821 			return -EINVAL;
822 		}
823 
824 		SW_FLOW_KEY_PUT(match, ct.state, ct_state, is_mask);
825 		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
826 	}
827 	if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
828 	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
829 		u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
830 
831 		SW_FLOW_KEY_PUT(match, ct.zone, ct_zone, is_mask);
832 		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
833 	}
834 	if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
835 	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
836 		u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
837 
838 		SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
839 		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
840 	}
841 	if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
842 	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
843 		const struct ovs_key_ct_labels *cl;
844 
845 		cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
846 		SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
847 				   sizeof(*cl), is_mask);
848 		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
849 	}
850 	return 0;
851 }
852 
853 static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
854 				u64 attrs, const struct nlattr **a,
855 				bool is_mask, bool log)
856 {
857 	int err;
858 
859 	err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
860 	if (err)
861 		return err;
862 
863 	if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
864 		const struct ovs_key_ethernet *eth_key;
865 
866 		eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
867 		SW_FLOW_KEY_MEMCPY(match, eth.src,
868 				eth_key->eth_src, ETH_ALEN, is_mask);
869 		SW_FLOW_KEY_MEMCPY(match, eth.dst,
870 				eth_key->eth_dst, ETH_ALEN, is_mask);
871 		attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
872 	}
873 
874 	if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
875 		__be16 tci;
876 
877 		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
878 		if (!(tci & htons(VLAN_TAG_PRESENT))) {
879 			if (is_mask)
880 				OVS_NLERR(log, "VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.");
881 			else
882 				OVS_NLERR(log, "VLAN TCI does not have VLAN_TAG_PRESENT bit set.");
883 
884 			return -EINVAL;
885 		}
886 
887 		SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
888 		attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
889 	}
890 
891 	if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
892 		__be16 eth_type;
893 
894 		eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
895 		if (is_mask) {
896 			/* Always exact match EtherType. */
897 			eth_type = htons(0xffff);
898 		} else if (!eth_proto_is_802_3(eth_type)) {
899 			OVS_NLERR(log, "EtherType %x is less than min %x",
900 				  ntohs(eth_type), ETH_P_802_3_MIN);
901 			return -EINVAL;
902 		}
903 
904 		SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
905 		attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
906 	} else if (!is_mask) {
907 		SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
908 	}
909 
910 	if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
911 		const struct ovs_key_ipv4 *ipv4_key;
912 
913 		ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
914 		if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
915 			OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
916 				  ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
917 			return -EINVAL;
918 		}
919 		SW_FLOW_KEY_PUT(match, ip.proto,
920 				ipv4_key->ipv4_proto, is_mask);
921 		SW_FLOW_KEY_PUT(match, ip.tos,
922 				ipv4_key->ipv4_tos, is_mask);
923 		SW_FLOW_KEY_PUT(match, ip.ttl,
924 				ipv4_key->ipv4_ttl, is_mask);
925 		SW_FLOW_KEY_PUT(match, ip.frag,
926 				ipv4_key->ipv4_frag, is_mask);
927 		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
928 				ipv4_key->ipv4_src, is_mask);
929 		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
930 				ipv4_key->ipv4_dst, is_mask);
931 		attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
932 	}
933 
934 	if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
935 		const struct ovs_key_ipv6 *ipv6_key;
936 
937 		ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
938 		if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
939 			OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
940 				  ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
941 			return -EINVAL;
942 		}
943 
944 		if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
945 			OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x).\n",
946 				  ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
947 			return -EINVAL;
948 		}
949 
950 		SW_FLOW_KEY_PUT(match, ipv6.label,
951 				ipv6_key->ipv6_label, is_mask);
952 		SW_FLOW_KEY_PUT(match, ip.proto,
953 				ipv6_key->ipv6_proto, is_mask);
954 		SW_FLOW_KEY_PUT(match, ip.tos,
955 				ipv6_key->ipv6_tclass, is_mask);
956 		SW_FLOW_KEY_PUT(match, ip.ttl,
957 				ipv6_key->ipv6_hlimit, is_mask);
958 		SW_FLOW_KEY_PUT(match, ip.frag,
959 				ipv6_key->ipv6_frag, is_mask);
960 		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
961 				ipv6_key->ipv6_src,
962 				sizeof(match->key->ipv6.addr.src),
963 				is_mask);
964 		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
965 				ipv6_key->ipv6_dst,
966 				sizeof(match->key->ipv6.addr.dst),
967 				is_mask);
968 
969 		attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
970 	}
971 
972 	if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
973 		const struct ovs_key_arp *arp_key;
974 
975 		arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
976 		if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
977 			OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
978 				  arp_key->arp_op);
979 			return -EINVAL;
980 		}
981 
982 		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
983 				arp_key->arp_sip, is_mask);
984 		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
985 			arp_key->arp_tip, is_mask);
986 		SW_FLOW_KEY_PUT(match, ip.proto,
987 				ntohs(arp_key->arp_op), is_mask);
988 		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
989 				arp_key->arp_sha, ETH_ALEN, is_mask);
990 		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
991 				arp_key->arp_tha, ETH_ALEN, is_mask);
992 
993 		attrs &= ~(1 << OVS_KEY_ATTR_ARP);
994 	}
995 
996 	if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
997 		const struct ovs_key_mpls *mpls_key;
998 
999 		mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1000 		SW_FLOW_KEY_PUT(match, mpls.top_lse,
1001 				mpls_key->mpls_lse, is_mask);
1002 
1003 		attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1004 	 }
1005 
1006 	if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1007 		const struct ovs_key_tcp *tcp_key;
1008 
1009 		tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1010 		SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1011 		SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1012 		attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1013 	}
1014 
1015 	if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1016 		SW_FLOW_KEY_PUT(match, tp.flags,
1017 				nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1018 				is_mask);
1019 		attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1020 	}
1021 
1022 	if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1023 		const struct ovs_key_udp *udp_key;
1024 
1025 		udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1026 		SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1027 		SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1028 		attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1029 	}
1030 
1031 	if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1032 		const struct ovs_key_sctp *sctp_key;
1033 
1034 		sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1035 		SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1036 		SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1037 		attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1038 	}
1039 
1040 	if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1041 		const struct ovs_key_icmp *icmp_key;
1042 
1043 		icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1044 		SW_FLOW_KEY_PUT(match, tp.src,
1045 				htons(icmp_key->icmp_type), is_mask);
1046 		SW_FLOW_KEY_PUT(match, tp.dst,
1047 				htons(icmp_key->icmp_code), is_mask);
1048 		attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1049 	}
1050 
1051 	if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1052 		const struct ovs_key_icmpv6 *icmpv6_key;
1053 
1054 		icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1055 		SW_FLOW_KEY_PUT(match, tp.src,
1056 				htons(icmpv6_key->icmpv6_type), is_mask);
1057 		SW_FLOW_KEY_PUT(match, tp.dst,
1058 				htons(icmpv6_key->icmpv6_code), is_mask);
1059 		attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1060 	}
1061 
1062 	if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1063 		const struct ovs_key_nd *nd_key;
1064 
1065 		nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1066 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1067 			nd_key->nd_target,
1068 			sizeof(match->key->ipv6.nd.target),
1069 			is_mask);
1070 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1071 			nd_key->nd_sll, ETH_ALEN, is_mask);
1072 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1073 				nd_key->nd_tll, ETH_ALEN, is_mask);
1074 		attrs &= ~(1 << OVS_KEY_ATTR_ND);
1075 	}
1076 
1077 	if (attrs != 0) {
1078 		OVS_NLERR(log, "Unknown key attributes %llx",
1079 			  (unsigned long long)attrs);
1080 		return -EINVAL;
1081 	}
1082 
1083 	return 0;
1084 }
1085 
1086 static void nlattr_set(struct nlattr *attr, u8 val,
1087 		       const struct ovs_len_tbl *tbl)
1088 {
1089 	struct nlattr *nla;
1090 	int rem;
1091 
1092 	/* The nlattr stream should already have been validated */
1093 	nla_for_each_nested(nla, attr, rem) {
1094 		if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED) {
1095 			if (tbl[nla_type(nla)].next)
1096 				tbl = tbl[nla_type(nla)].next;
1097 			nlattr_set(nla, val, tbl);
1098 		} else {
1099 			memset(nla_data(nla), val, nla_len(nla));
1100 		}
1101 
1102 		if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1103 			*(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1104 	}
1105 }
1106 
1107 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1108 {
1109 	nlattr_set(attr, val, ovs_key_lens);
1110 }
1111 
1112 /**
1113  * ovs_nla_get_match - parses Netlink attributes into a flow key and
1114  * mask. In case the 'mask' is NULL, the flow is treated as exact match
1115  * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1116  * does not include any don't care bit.
1117  * @net: Used to determine per-namespace field support.
1118  * @match: receives the extracted flow match information.
1119  * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1120  * sequence. The fields should of the packet that triggered the creation
1121  * of this flow.
1122  * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1123  * attribute specifies the mask field of the wildcarded flow.
1124  * @log: Boolean to allow kernel error logging.  Normally true, but when
1125  * probing for feature compatibility this should be passed in as false to
1126  * suppress unnecessary error logging.
1127  */
1128 int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1129 		      const struct nlattr *nla_key,
1130 		      const struct nlattr *nla_mask,
1131 		      bool log)
1132 {
1133 	const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1134 	const struct nlattr *encap;
1135 	struct nlattr *newmask = NULL;
1136 	u64 key_attrs = 0;
1137 	u64 mask_attrs = 0;
1138 	bool encap_valid = false;
1139 	int err;
1140 
1141 	err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1142 	if (err)
1143 		return err;
1144 
1145 	if ((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
1146 	    (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
1147 	    (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
1148 		__be16 tci;
1149 
1150 		if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
1151 		      (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
1152 			OVS_NLERR(log, "Invalid Vlan frame.");
1153 			return -EINVAL;
1154 		}
1155 
1156 		key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1157 		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1158 		encap = a[OVS_KEY_ATTR_ENCAP];
1159 		key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1160 		encap_valid = true;
1161 
1162 		if (tci & htons(VLAN_TAG_PRESENT)) {
1163 			err = parse_flow_nlattrs(encap, a, &key_attrs, log);
1164 			if (err)
1165 				return err;
1166 		} else if (!tci) {
1167 			/* Corner case for truncated 802.1Q header. */
1168 			if (nla_len(encap)) {
1169 				OVS_NLERR(log, "Truncated 802.1Q header has non-zero encap attribute.");
1170 				return -EINVAL;
1171 			}
1172 		} else {
1173 			OVS_NLERR(log, "Encap attr is set for non-VLAN frame");
1174 			return  -EINVAL;
1175 		}
1176 	}
1177 
1178 	err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1179 	if (err)
1180 		return err;
1181 
1182 	if (match->mask) {
1183 		if (!nla_mask) {
1184 			/* Create an exact match mask. We need to set to 0xff
1185 			 * all the 'match->mask' fields that have been touched
1186 			 * in 'match->key'. We cannot simply memset
1187 			 * 'match->mask', because padding bytes and fields not
1188 			 * specified in 'match->key' should be left to 0.
1189 			 * Instead, we use a stream of netlink attributes,
1190 			 * copied from 'key' and set to 0xff.
1191 			 * ovs_key_from_nlattrs() will take care of filling
1192 			 * 'match->mask' appropriately.
1193 			 */
1194 			newmask = kmemdup(nla_key,
1195 					  nla_total_size(nla_len(nla_key)),
1196 					  GFP_KERNEL);
1197 			if (!newmask)
1198 				return -ENOMEM;
1199 
1200 			mask_set_nlattr(newmask, 0xff);
1201 
1202 			/* The userspace does not send tunnel attributes that
1203 			 * are 0, but we should not wildcard them nonetheless.
1204 			 */
1205 			if (match->key->tun_key.u.ipv4.dst)
1206 				SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1207 							 0xff, true);
1208 
1209 			nla_mask = newmask;
1210 		}
1211 
1212 		err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1213 		if (err)
1214 			goto free_newmask;
1215 
1216 		/* Always match on tci. */
1217 		SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
1218 
1219 		if (mask_attrs & 1 << OVS_KEY_ATTR_ENCAP) {
1220 			__be16 eth_type = 0;
1221 			__be16 tci = 0;
1222 
1223 			if (!encap_valid) {
1224 				OVS_NLERR(log, "Encap mask attribute is set for non-VLAN frame.");
1225 				err = -EINVAL;
1226 				goto free_newmask;
1227 			}
1228 
1229 			mask_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1230 			if (a[OVS_KEY_ATTR_ETHERTYPE])
1231 				eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1232 
1233 			if (eth_type == htons(0xffff)) {
1234 				mask_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1235 				encap = a[OVS_KEY_ATTR_ENCAP];
1236 				err = parse_flow_mask_nlattrs(encap, a,
1237 							      &mask_attrs, log);
1238 				if (err)
1239 					goto free_newmask;
1240 			} else {
1241 				OVS_NLERR(log, "VLAN frames must have an exact match on the TPID (mask=%x).",
1242 					  ntohs(eth_type));
1243 				err = -EINVAL;
1244 				goto free_newmask;
1245 			}
1246 
1247 			if (a[OVS_KEY_ATTR_VLAN])
1248 				tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1249 
1250 			if (!(tci & htons(VLAN_TAG_PRESENT))) {
1251 				OVS_NLERR(log, "VLAN tag present bit must have an exact match (tci_mask=%x).",
1252 					  ntohs(tci));
1253 				err = -EINVAL;
1254 				goto free_newmask;
1255 			}
1256 		}
1257 
1258 		err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1259 					   log);
1260 		if (err)
1261 			goto free_newmask;
1262 	}
1263 
1264 	if (!match_validate(match, key_attrs, mask_attrs, log))
1265 		err = -EINVAL;
1266 
1267 free_newmask:
1268 	kfree(newmask);
1269 	return err;
1270 }
1271 
1272 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1273 {
1274 	size_t len;
1275 
1276 	if (!attr)
1277 		return 0;
1278 
1279 	len = nla_len(attr);
1280 	if (len < 1 || len > MAX_UFID_LENGTH) {
1281 		OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1282 			  nla_len(attr), MAX_UFID_LENGTH);
1283 		return 0;
1284 	}
1285 
1286 	return len;
1287 }
1288 
1289 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1290  * or false otherwise.
1291  */
1292 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1293 		      bool log)
1294 {
1295 	sfid->ufid_len = get_ufid_len(attr, log);
1296 	if (sfid->ufid_len)
1297 		memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1298 
1299 	return sfid->ufid_len;
1300 }
1301 
1302 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1303 			   const struct sw_flow_key *key, bool log)
1304 {
1305 	struct sw_flow_key *new_key;
1306 
1307 	if (ovs_nla_get_ufid(sfid, ufid, log))
1308 		return 0;
1309 
1310 	/* If UFID was not provided, use unmasked key. */
1311 	new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1312 	if (!new_key)
1313 		return -ENOMEM;
1314 	memcpy(new_key, key, sizeof(*key));
1315 	sfid->unmasked_key = new_key;
1316 
1317 	return 0;
1318 }
1319 
1320 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1321 {
1322 	return attr ? nla_get_u32(attr) : 0;
1323 }
1324 
1325 /**
1326  * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1327  * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1328  * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1329  * sequence.
1330  * @log: Boolean to allow kernel error logging.  Normally true, but when
1331  * probing for feature compatibility this should be passed in as false to
1332  * suppress unnecessary error logging.
1333  *
1334  * This parses a series of Netlink attributes that form a flow key, which must
1335  * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1336  * get the metadata, that is, the parts of the flow key that cannot be
1337  * extracted from the packet itself.
1338  */
1339 
1340 int ovs_nla_get_flow_metadata(struct net *net, const struct nlattr *attr,
1341 			      struct sw_flow_key *key,
1342 			      bool log)
1343 {
1344 	const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1345 	struct sw_flow_match match;
1346 	u64 attrs = 0;
1347 	int err;
1348 
1349 	err = parse_flow_nlattrs(attr, a, &attrs, log);
1350 	if (err)
1351 		return -EINVAL;
1352 
1353 	memset(&match, 0, sizeof(match));
1354 	match.key = key;
1355 
1356 	memset(&key->ct, 0, sizeof(key->ct));
1357 	key->phy.in_port = DP_MAX_PORTS;
1358 
1359 	return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1360 }
1361 
1362 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1363 			     const struct sw_flow_key *output, bool is_mask,
1364 			     struct sk_buff *skb)
1365 {
1366 	struct ovs_key_ethernet *eth_key;
1367 	struct nlattr *nla, *encap;
1368 
1369 	if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1370 		goto nla_put_failure;
1371 
1372 	if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1373 		goto nla_put_failure;
1374 
1375 	if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1376 		goto nla_put_failure;
1377 
1378 	if ((swkey->tun_key.u.ipv4.dst || is_mask)) {
1379 		const void *opts = NULL;
1380 
1381 		if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1382 			opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1383 
1384 		if (ipv4_tun_to_nlattr(skb, &output->tun_key, opts,
1385 				       swkey->tun_opts_len))
1386 			goto nla_put_failure;
1387 	}
1388 
1389 	if (swkey->phy.in_port == DP_MAX_PORTS) {
1390 		if (is_mask && (output->phy.in_port == 0xffff))
1391 			if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1392 				goto nla_put_failure;
1393 	} else {
1394 		u16 upper_u16;
1395 		upper_u16 = !is_mask ? 0 : 0xffff;
1396 
1397 		if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1398 				(upper_u16 << 16) | output->phy.in_port))
1399 			goto nla_put_failure;
1400 	}
1401 
1402 	if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1403 		goto nla_put_failure;
1404 
1405 	if (ovs_ct_put_key(output, skb))
1406 		goto nla_put_failure;
1407 
1408 	nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1409 	if (!nla)
1410 		goto nla_put_failure;
1411 
1412 	eth_key = nla_data(nla);
1413 	ether_addr_copy(eth_key->eth_src, output->eth.src);
1414 	ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1415 
1416 	if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1417 		__be16 eth_type;
1418 		eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
1419 		if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1420 		    nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
1421 			goto nla_put_failure;
1422 		encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1423 		if (!swkey->eth.tci)
1424 			goto unencap;
1425 	} else
1426 		encap = NULL;
1427 
1428 	if (swkey->eth.type == htons(ETH_P_802_2)) {
1429 		/*
1430 		 * Ethertype 802.2 is represented in the netlink with omitted
1431 		 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1432 		 * 0xffff in the mask attribute.  Ethertype can also
1433 		 * be wildcarded.
1434 		 */
1435 		if (is_mask && output->eth.type)
1436 			if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1437 						output->eth.type))
1438 				goto nla_put_failure;
1439 		goto unencap;
1440 	}
1441 
1442 	if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1443 		goto nla_put_failure;
1444 
1445 	if (swkey->eth.type == htons(ETH_P_IP)) {
1446 		struct ovs_key_ipv4 *ipv4_key;
1447 
1448 		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1449 		if (!nla)
1450 			goto nla_put_failure;
1451 		ipv4_key = nla_data(nla);
1452 		ipv4_key->ipv4_src = output->ipv4.addr.src;
1453 		ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1454 		ipv4_key->ipv4_proto = output->ip.proto;
1455 		ipv4_key->ipv4_tos = output->ip.tos;
1456 		ipv4_key->ipv4_ttl = output->ip.ttl;
1457 		ipv4_key->ipv4_frag = output->ip.frag;
1458 	} else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1459 		struct ovs_key_ipv6 *ipv6_key;
1460 
1461 		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1462 		if (!nla)
1463 			goto nla_put_failure;
1464 		ipv6_key = nla_data(nla);
1465 		memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1466 				sizeof(ipv6_key->ipv6_src));
1467 		memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1468 				sizeof(ipv6_key->ipv6_dst));
1469 		ipv6_key->ipv6_label = output->ipv6.label;
1470 		ipv6_key->ipv6_proto = output->ip.proto;
1471 		ipv6_key->ipv6_tclass = output->ip.tos;
1472 		ipv6_key->ipv6_hlimit = output->ip.ttl;
1473 		ipv6_key->ipv6_frag = output->ip.frag;
1474 	} else if (swkey->eth.type == htons(ETH_P_ARP) ||
1475 		   swkey->eth.type == htons(ETH_P_RARP)) {
1476 		struct ovs_key_arp *arp_key;
1477 
1478 		nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1479 		if (!nla)
1480 			goto nla_put_failure;
1481 		arp_key = nla_data(nla);
1482 		memset(arp_key, 0, sizeof(struct ovs_key_arp));
1483 		arp_key->arp_sip = output->ipv4.addr.src;
1484 		arp_key->arp_tip = output->ipv4.addr.dst;
1485 		arp_key->arp_op = htons(output->ip.proto);
1486 		ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1487 		ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1488 	} else if (eth_p_mpls(swkey->eth.type)) {
1489 		struct ovs_key_mpls *mpls_key;
1490 
1491 		nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1492 		if (!nla)
1493 			goto nla_put_failure;
1494 		mpls_key = nla_data(nla);
1495 		mpls_key->mpls_lse = output->mpls.top_lse;
1496 	}
1497 
1498 	if ((swkey->eth.type == htons(ETH_P_IP) ||
1499 	     swkey->eth.type == htons(ETH_P_IPV6)) &&
1500 	     swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1501 
1502 		if (swkey->ip.proto == IPPROTO_TCP) {
1503 			struct ovs_key_tcp *tcp_key;
1504 
1505 			nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1506 			if (!nla)
1507 				goto nla_put_failure;
1508 			tcp_key = nla_data(nla);
1509 			tcp_key->tcp_src = output->tp.src;
1510 			tcp_key->tcp_dst = output->tp.dst;
1511 			if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1512 					 output->tp.flags))
1513 				goto nla_put_failure;
1514 		} else if (swkey->ip.proto == IPPROTO_UDP) {
1515 			struct ovs_key_udp *udp_key;
1516 
1517 			nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1518 			if (!nla)
1519 				goto nla_put_failure;
1520 			udp_key = nla_data(nla);
1521 			udp_key->udp_src = output->tp.src;
1522 			udp_key->udp_dst = output->tp.dst;
1523 		} else if (swkey->ip.proto == IPPROTO_SCTP) {
1524 			struct ovs_key_sctp *sctp_key;
1525 
1526 			nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1527 			if (!nla)
1528 				goto nla_put_failure;
1529 			sctp_key = nla_data(nla);
1530 			sctp_key->sctp_src = output->tp.src;
1531 			sctp_key->sctp_dst = output->tp.dst;
1532 		} else if (swkey->eth.type == htons(ETH_P_IP) &&
1533 			   swkey->ip.proto == IPPROTO_ICMP) {
1534 			struct ovs_key_icmp *icmp_key;
1535 
1536 			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1537 			if (!nla)
1538 				goto nla_put_failure;
1539 			icmp_key = nla_data(nla);
1540 			icmp_key->icmp_type = ntohs(output->tp.src);
1541 			icmp_key->icmp_code = ntohs(output->tp.dst);
1542 		} else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1543 			   swkey->ip.proto == IPPROTO_ICMPV6) {
1544 			struct ovs_key_icmpv6 *icmpv6_key;
1545 
1546 			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1547 						sizeof(*icmpv6_key));
1548 			if (!nla)
1549 				goto nla_put_failure;
1550 			icmpv6_key = nla_data(nla);
1551 			icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1552 			icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1553 
1554 			if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1555 			    icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1556 				struct ovs_key_nd *nd_key;
1557 
1558 				nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1559 				if (!nla)
1560 					goto nla_put_failure;
1561 				nd_key = nla_data(nla);
1562 				memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1563 							sizeof(nd_key->nd_target));
1564 				ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1565 				ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1566 			}
1567 		}
1568 	}
1569 
1570 unencap:
1571 	if (encap)
1572 		nla_nest_end(skb, encap);
1573 
1574 	return 0;
1575 
1576 nla_put_failure:
1577 	return -EMSGSIZE;
1578 }
1579 
1580 int ovs_nla_put_key(const struct sw_flow_key *swkey,
1581 		    const struct sw_flow_key *output, int attr, bool is_mask,
1582 		    struct sk_buff *skb)
1583 {
1584 	int err;
1585 	struct nlattr *nla;
1586 
1587 	nla = nla_nest_start(skb, attr);
1588 	if (!nla)
1589 		return -EMSGSIZE;
1590 	err = __ovs_nla_put_key(swkey, output, is_mask, skb);
1591 	if (err)
1592 		return err;
1593 	nla_nest_end(skb, nla);
1594 
1595 	return 0;
1596 }
1597 
1598 /* Called with ovs_mutex or RCU read lock. */
1599 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
1600 {
1601 	if (ovs_identifier_is_ufid(&flow->id))
1602 		return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
1603 			       flow->id.ufid);
1604 
1605 	return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
1606 			       OVS_FLOW_ATTR_KEY, false, skb);
1607 }
1608 
1609 /* Called with ovs_mutex or RCU read lock. */
1610 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
1611 {
1612 	return ovs_nla_put_key(&flow->key, &flow->key,
1613 				OVS_FLOW_ATTR_KEY, false, skb);
1614 }
1615 
1616 /* Called with ovs_mutex or RCU read lock. */
1617 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
1618 {
1619 	return ovs_nla_put_key(&flow->key, &flow->mask->key,
1620 				OVS_FLOW_ATTR_MASK, true, skb);
1621 }
1622 
1623 #define MAX_ACTIONS_BUFSIZE	(32 * 1024)
1624 
1625 static struct sw_flow_actions *nla_alloc_flow_actions(int size, bool log)
1626 {
1627 	struct sw_flow_actions *sfa;
1628 
1629 	if (size > MAX_ACTIONS_BUFSIZE) {
1630 		OVS_NLERR(log, "Flow action size %u bytes exceeds max", size);
1631 		return ERR_PTR(-EINVAL);
1632 	}
1633 
1634 	sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1635 	if (!sfa)
1636 		return ERR_PTR(-ENOMEM);
1637 
1638 	sfa->actions_len = 0;
1639 	return sfa;
1640 }
1641 
1642 static void ovs_nla_free_set_action(const struct nlattr *a)
1643 {
1644 	const struct nlattr *ovs_key = nla_data(a);
1645 	struct ovs_tunnel_info *ovs_tun;
1646 
1647 	switch (nla_type(ovs_key)) {
1648 	case OVS_KEY_ATTR_TUNNEL_INFO:
1649 		ovs_tun = nla_data(ovs_key);
1650 		dst_release((struct dst_entry *)ovs_tun->tun_dst);
1651 		break;
1652 	}
1653 }
1654 
1655 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1656 {
1657 	const struct nlattr *a;
1658 	int rem;
1659 
1660 	if (!sf_acts)
1661 		return;
1662 
1663 	nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
1664 		switch (nla_type(a)) {
1665 		case OVS_ACTION_ATTR_SET:
1666 			ovs_nla_free_set_action(a);
1667 			break;
1668 		case OVS_ACTION_ATTR_CT:
1669 			ovs_ct_free_action(a);
1670 			break;
1671 		}
1672 	}
1673 
1674 	kfree(sf_acts);
1675 }
1676 
1677 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
1678 {
1679 	ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
1680 }
1681 
1682 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1683  * The caller must hold rcu_read_lock for this to be sensible. */
1684 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
1685 {
1686 	call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
1687 }
1688 
1689 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1690 				       int attr_len, bool log)
1691 {
1692 
1693 	struct sw_flow_actions *acts;
1694 	int new_acts_size;
1695 	int req_size = NLA_ALIGN(attr_len);
1696 	int next_offset = offsetof(struct sw_flow_actions, actions) +
1697 					(*sfa)->actions_len;
1698 
1699 	if (req_size <= (ksize(*sfa) - next_offset))
1700 		goto out;
1701 
1702 	new_acts_size = ksize(*sfa) * 2;
1703 
1704 	if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1705 		if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1706 			return ERR_PTR(-EMSGSIZE);
1707 		new_acts_size = MAX_ACTIONS_BUFSIZE;
1708 	}
1709 
1710 	acts = nla_alloc_flow_actions(new_acts_size, log);
1711 	if (IS_ERR(acts))
1712 		return (void *)acts;
1713 
1714 	memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1715 	acts->actions_len = (*sfa)->actions_len;
1716 	acts->orig_len = (*sfa)->orig_len;
1717 	kfree(*sfa);
1718 	*sfa = acts;
1719 
1720 out:
1721 	(*sfa)->actions_len += req_size;
1722 	return  (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1723 }
1724 
1725 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1726 				   int attrtype, void *data, int len, bool log)
1727 {
1728 	struct nlattr *a;
1729 
1730 	a = reserve_sfa_size(sfa, nla_attr_size(len), log);
1731 	if (IS_ERR(a))
1732 		return a;
1733 
1734 	a->nla_type = attrtype;
1735 	a->nla_len = nla_attr_size(len);
1736 
1737 	if (data)
1738 		memcpy(nla_data(a), data, len);
1739 	memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1740 
1741 	return a;
1742 }
1743 
1744 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
1745 		       int len, bool log)
1746 {
1747 	struct nlattr *a;
1748 
1749 	a = __add_action(sfa, attrtype, data, len, log);
1750 
1751 	return PTR_ERR_OR_ZERO(a);
1752 }
1753 
1754 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1755 					  int attrtype, bool log)
1756 {
1757 	int used = (*sfa)->actions_len;
1758 	int err;
1759 
1760 	err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
1761 	if (err)
1762 		return err;
1763 
1764 	return used;
1765 }
1766 
1767 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1768 					 int st_offset)
1769 {
1770 	struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1771 							       st_offset);
1772 
1773 	a->nla_len = sfa->actions_len - st_offset;
1774 }
1775 
1776 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
1777 				  const struct sw_flow_key *key,
1778 				  int depth, struct sw_flow_actions **sfa,
1779 				  __be16 eth_type, __be16 vlan_tci, bool log);
1780 
1781 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
1782 				    const struct sw_flow_key *key, int depth,
1783 				    struct sw_flow_actions **sfa,
1784 				    __be16 eth_type, __be16 vlan_tci, bool log)
1785 {
1786 	const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1787 	const struct nlattr *probability, *actions;
1788 	const struct nlattr *a;
1789 	int rem, start, err, st_acts;
1790 
1791 	memset(attrs, 0, sizeof(attrs));
1792 	nla_for_each_nested(a, attr, rem) {
1793 		int type = nla_type(a);
1794 		if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1795 			return -EINVAL;
1796 		attrs[type] = a;
1797 	}
1798 	if (rem)
1799 		return -EINVAL;
1800 
1801 	probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1802 	if (!probability || nla_len(probability) != sizeof(u32))
1803 		return -EINVAL;
1804 
1805 	actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1806 	if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1807 		return -EINVAL;
1808 
1809 	/* validation done, copy sample action. */
1810 	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
1811 	if (start < 0)
1812 		return start;
1813 	err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1814 				 nla_data(probability), sizeof(u32), log);
1815 	if (err)
1816 		return err;
1817 	st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS, log);
1818 	if (st_acts < 0)
1819 		return st_acts;
1820 
1821 	err = __ovs_nla_copy_actions(net, actions, key, depth + 1, sfa,
1822 				     eth_type, vlan_tci, log);
1823 	if (err)
1824 		return err;
1825 
1826 	add_nested_action_end(*sfa, st_acts);
1827 	add_nested_action_end(*sfa, start);
1828 
1829 	return 0;
1830 }
1831 
1832 void ovs_match_init(struct sw_flow_match *match,
1833 		    struct sw_flow_key *key,
1834 		    struct sw_flow_mask *mask)
1835 {
1836 	memset(match, 0, sizeof(*match));
1837 	match->key = key;
1838 	match->mask = mask;
1839 
1840 	memset(key, 0, sizeof(*key));
1841 
1842 	if (mask) {
1843 		memset(&mask->key, 0, sizeof(mask->key));
1844 		mask->range.start = mask->range.end = 0;
1845 	}
1846 }
1847 
1848 static int validate_geneve_opts(struct sw_flow_key *key)
1849 {
1850 	struct geneve_opt *option;
1851 	int opts_len = key->tun_opts_len;
1852 	bool crit_opt = false;
1853 
1854 	option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
1855 	while (opts_len > 0) {
1856 		int len;
1857 
1858 		if (opts_len < sizeof(*option))
1859 			return -EINVAL;
1860 
1861 		len = sizeof(*option) + option->length * 4;
1862 		if (len > opts_len)
1863 			return -EINVAL;
1864 
1865 		crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
1866 
1867 		option = (struct geneve_opt *)((u8 *)option + len);
1868 		opts_len -= len;
1869 	};
1870 
1871 	key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
1872 
1873 	return 0;
1874 }
1875 
1876 static int validate_and_copy_set_tun(const struct nlattr *attr,
1877 				     struct sw_flow_actions **sfa, bool log)
1878 {
1879 	struct sw_flow_match match;
1880 	struct sw_flow_key key;
1881 	struct metadata_dst *tun_dst;
1882 	struct ip_tunnel_info *tun_info;
1883 	struct ovs_tunnel_info *ovs_tun;
1884 	struct nlattr *a;
1885 	int err = 0, start, opts_type;
1886 
1887 	ovs_match_init(&match, &key, NULL);
1888 	opts_type = ipv4_tun_from_nlattr(nla_data(attr), &match, false, log);
1889 	if (opts_type < 0)
1890 		return opts_type;
1891 
1892 	if (key.tun_opts_len) {
1893 		switch (opts_type) {
1894 		case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
1895 			err = validate_geneve_opts(&key);
1896 			if (err < 0)
1897 				return err;
1898 			break;
1899 		case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
1900 			break;
1901 		}
1902 	};
1903 
1904 	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
1905 	if (start < 0)
1906 		return start;
1907 
1908 	tun_dst = metadata_dst_alloc(key.tun_opts_len, GFP_KERNEL);
1909 	if (!tun_dst)
1910 		return -ENOMEM;
1911 
1912 	a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
1913 			 sizeof(*ovs_tun), log);
1914 	if (IS_ERR(a)) {
1915 		dst_release((struct dst_entry *)tun_dst);
1916 		return PTR_ERR(a);
1917 	}
1918 
1919 	ovs_tun = nla_data(a);
1920 	ovs_tun->tun_dst = tun_dst;
1921 
1922 	tun_info = &tun_dst->u.tun_info;
1923 	tun_info->mode = IP_TUNNEL_INFO_TX;
1924 	tun_info->key = key.tun_key;
1925 
1926 	/* We need to store the options in the action itself since
1927 	 * everything else will go away after flow setup. We can append
1928 	 * it to tun_info and then point there.
1929 	 */
1930 	ip_tunnel_info_opts_set(tun_info,
1931 				TUN_METADATA_OPTS(&key, key.tun_opts_len),
1932 				key.tun_opts_len);
1933 	add_nested_action_end(*sfa, start);
1934 
1935 	return err;
1936 }
1937 
1938 /* Return false if there are any non-masked bits set.
1939  * Mask follows data immediately, before any netlink padding.
1940  */
1941 static bool validate_masked(u8 *data, int len)
1942 {
1943 	u8 *mask = data + len;
1944 
1945 	while (len--)
1946 		if (*data++ & ~*mask++)
1947 			return false;
1948 
1949 	return true;
1950 }
1951 
1952 static int validate_set(const struct nlattr *a,
1953 			const struct sw_flow_key *flow_key,
1954 			struct sw_flow_actions **sfa,
1955 			bool *skip_copy, __be16 eth_type, bool masked, bool log)
1956 {
1957 	const struct nlattr *ovs_key = nla_data(a);
1958 	int key_type = nla_type(ovs_key);
1959 	size_t key_len;
1960 
1961 	/* There can be only one key in a action */
1962 	if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1963 		return -EINVAL;
1964 
1965 	key_len = nla_len(ovs_key);
1966 	if (masked)
1967 		key_len /= 2;
1968 
1969 	if (key_type > OVS_KEY_ATTR_MAX ||
1970 	    !check_attr_len(key_len, ovs_key_lens[key_type].len))
1971 		return -EINVAL;
1972 
1973 	if (masked && !validate_masked(nla_data(ovs_key), key_len))
1974 		return -EINVAL;
1975 
1976 	switch (key_type) {
1977 	const struct ovs_key_ipv4 *ipv4_key;
1978 	const struct ovs_key_ipv6 *ipv6_key;
1979 	int err;
1980 
1981 	case OVS_KEY_ATTR_PRIORITY:
1982 	case OVS_KEY_ATTR_SKB_MARK:
1983 	case OVS_KEY_ATTR_CT_MARK:
1984 	case OVS_KEY_ATTR_CT_LABELS:
1985 	case OVS_KEY_ATTR_ETHERNET:
1986 		break;
1987 
1988 	case OVS_KEY_ATTR_TUNNEL:
1989 		if (eth_p_mpls(eth_type))
1990 			return -EINVAL;
1991 
1992 		if (masked)
1993 			return -EINVAL; /* Masked tunnel set not supported. */
1994 
1995 		*skip_copy = true;
1996 		err = validate_and_copy_set_tun(a, sfa, log);
1997 		if (err)
1998 			return err;
1999 		break;
2000 
2001 	case OVS_KEY_ATTR_IPV4:
2002 		if (eth_type != htons(ETH_P_IP))
2003 			return -EINVAL;
2004 
2005 		ipv4_key = nla_data(ovs_key);
2006 
2007 		if (masked) {
2008 			const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2009 
2010 			/* Non-writeable fields. */
2011 			if (mask->ipv4_proto || mask->ipv4_frag)
2012 				return -EINVAL;
2013 		} else {
2014 			if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2015 				return -EINVAL;
2016 
2017 			if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2018 				return -EINVAL;
2019 		}
2020 		break;
2021 
2022 	case OVS_KEY_ATTR_IPV6:
2023 		if (eth_type != htons(ETH_P_IPV6))
2024 			return -EINVAL;
2025 
2026 		ipv6_key = nla_data(ovs_key);
2027 
2028 		if (masked) {
2029 			const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2030 
2031 			/* Non-writeable fields. */
2032 			if (mask->ipv6_proto || mask->ipv6_frag)
2033 				return -EINVAL;
2034 
2035 			/* Invalid bits in the flow label mask? */
2036 			if (ntohl(mask->ipv6_label) & 0xFFF00000)
2037 				return -EINVAL;
2038 		} else {
2039 			if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2040 				return -EINVAL;
2041 
2042 			if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2043 				return -EINVAL;
2044 		}
2045 		if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2046 			return -EINVAL;
2047 
2048 		break;
2049 
2050 	case OVS_KEY_ATTR_TCP:
2051 		if ((eth_type != htons(ETH_P_IP) &&
2052 		     eth_type != htons(ETH_P_IPV6)) ||
2053 		    flow_key->ip.proto != IPPROTO_TCP)
2054 			return -EINVAL;
2055 
2056 		break;
2057 
2058 	case OVS_KEY_ATTR_UDP:
2059 		if ((eth_type != htons(ETH_P_IP) &&
2060 		     eth_type != htons(ETH_P_IPV6)) ||
2061 		    flow_key->ip.proto != IPPROTO_UDP)
2062 			return -EINVAL;
2063 
2064 		break;
2065 
2066 	case OVS_KEY_ATTR_MPLS:
2067 		if (!eth_p_mpls(eth_type))
2068 			return -EINVAL;
2069 		break;
2070 
2071 	case OVS_KEY_ATTR_SCTP:
2072 		if ((eth_type != htons(ETH_P_IP) &&
2073 		     eth_type != htons(ETH_P_IPV6)) ||
2074 		    flow_key->ip.proto != IPPROTO_SCTP)
2075 			return -EINVAL;
2076 
2077 		break;
2078 
2079 	default:
2080 		return -EINVAL;
2081 	}
2082 
2083 	/* Convert non-masked non-tunnel set actions to masked set actions. */
2084 	if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2085 		int start, len = key_len * 2;
2086 		struct nlattr *at;
2087 
2088 		*skip_copy = true;
2089 
2090 		start = add_nested_action_start(sfa,
2091 						OVS_ACTION_ATTR_SET_TO_MASKED,
2092 						log);
2093 		if (start < 0)
2094 			return start;
2095 
2096 		at = __add_action(sfa, key_type, NULL, len, log);
2097 		if (IS_ERR(at))
2098 			return PTR_ERR(at);
2099 
2100 		memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2101 		memset(nla_data(at) + key_len, 0xff, key_len);    /* Mask. */
2102 		/* Clear non-writeable bits from otherwise writeable fields. */
2103 		if (key_type == OVS_KEY_ATTR_IPV6) {
2104 			struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2105 
2106 			mask->ipv6_label &= htonl(0x000FFFFF);
2107 		}
2108 		add_nested_action_end(*sfa, start);
2109 	}
2110 
2111 	return 0;
2112 }
2113 
2114 static int validate_userspace(const struct nlattr *attr)
2115 {
2116 	static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2117 		[OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2118 		[OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2119 		[OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2120 	};
2121 	struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2122 	int error;
2123 
2124 	error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
2125 				 attr, userspace_policy);
2126 	if (error)
2127 		return error;
2128 
2129 	if (!a[OVS_USERSPACE_ATTR_PID] ||
2130 	    !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2131 		return -EINVAL;
2132 
2133 	return 0;
2134 }
2135 
2136 static int copy_action(const struct nlattr *from,
2137 		       struct sw_flow_actions **sfa, bool log)
2138 {
2139 	int totlen = NLA_ALIGN(from->nla_len);
2140 	struct nlattr *to;
2141 
2142 	to = reserve_sfa_size(sfa, from->nla_len, log);
2143 	if (IS_ERR(to))
2144 		return PTR_ERR(to);
2145 
2146 	memcpy(to, from, totlen);
2147 	return 0;
2148 }
2149 
2150 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2151 				  const struct sw_flow_key *key,
2152 				  int depth, struct sw_flow_actions **sfa,
2153 				  __be16 eth_type, __be16 vlan_tci, bool log)
2154 {
2155 	const struct nlattr *a;
2156 	int rem, err;
2157 
2158 	if (depth >= SAMPLE_ACTION_DEPTH)
2159 		return -EOVERFLOW;
2160 
2161 	nla_for_each_nested(a, attr, rem) {
2162 		/* Expected argument lengths, (u32)-1 for variable length. */
2163 		static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2164 			[OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2165 			[OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2166 			[OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2167 			[OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2168 			[OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2169 			[OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2170 			[OVS_ACTION_ATTR_POP_VLAN] = 0,
2171 			[OVS_ACTION_ATTR_SET] = (u32)-1,
2172 			[OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2173 			[OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2174 			[OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2175 			[OVS_ACTION_ATTR_CT] = (u32)-1,
2176 		};
2177 		const struct ovs_action_push_vlan *vlan;
2178 		int type = nla_type(a);
2179 		bool skip_copy;
2180 
2181 		if (type > OVS_ACTION_ATTR_MAX ||
2182 		    (action_lens[type] != nla_len(a) &&
2183 		     action_lens[type] != (u32)-1))
2184 			return -EINVAL;
2185 
2186 		skip_copy = false;
2187 		switch (type) {
2188 		case OVS_ACTION_ATTR_UNSPEC:
2189 			return -EINVAL;
2190 
2191 		case OVS_ACTION_ATTR_USERSPACE:
2192 			err = validate_userspace(a);
2193 			if (err)
2194 				return err;
2195 			break;
2196 
2197 		case OVS_ACTION_ATTR_OUTPUT:
2198 			if (nla_get_u32(a) >= DP_MAX_PORTS)
2199 				return -EINVAL;
2200 			break;
2201 
2202 		case OVS_ACTION_ATTR_HASH: {
2203 			const struct ovs_action_hash *act_hash = nla_data(a);
2204 
2205 			switch (act_hash->hash_alg) {
2206 			case OVS_HASH_ALG_L4:
2207 				break;
2208 			default:
2209 				return  -EINVAL;
2210 			}
2211 
2212 			break;
2213 		}
2214 
2215 		case OVS_ACTION_ATTR_POP_VLAN:
2216 			vlan_tci = htons(0);
2217 			break;
2218 
2219 		case OVS_ACTION_ATTR_PUSH_VLAN:
2220 			vlan = nla_data(a);
2221 			if (vlan->vlan_tpid != htons(ETH_P_8021Q))
2222 				return -EINVAL;
2223 			if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2224 				return -EINVAL;
2225 			vlan_tci = vlan->vlan_tci;
2226 			break;
2227 
2228 		case OVS_ACTION_ATTR_RECIRC:
2229 			break;
2230 
2231 		case OVS_ACTION_ATTR_PUSH_MPLS: {
2232 			const struct ovs_action_push_mpls *mpls = nla_data(a);
2233 
2234 			if (!eth_p_mpls(mpls->mpls_ethertype))
2235 				return -EINVAL;
2236 			/* Prohibit push MPLS other than to a white list
2237 			 * for packets that have a known tag order.
2238 			 */
2239 			if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2240 			    (eth_type != htons(ETH_P_IP) &&
2241 			     eth_type != htons(ETH_P_IPV6) &&
2242 			     eth_type != htons(ETH_P_ARP) &&
2243 			     eth_type != htons(ETH_P_RARP) &&
2244 			     !eth_p_mpls(eth_type)))
2245 				return -EINVAL;
2246 			eth_type = mpls->mpls_ethertype;
2247 			break;
2248 		}
2249 
2250 		case OVS_ACTION_ATTR_POP_MPLS:
2251 			if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2252 			    !eth_p_mpls(eth_type))
2253 				return -EINVAL;
2254 
2255 			/* Disallow subsequent L2.5+ set and mpls_pop actions
2256 			 * as there is no check here to ensure that the new
2257 			 * eth_type is valid and thus set actions could
2258 			 * write off the end of the packet or otherwise
2259 			 * corrupt it.
2260 			 *
2261 			 * Support for these actions is planned using packet
2262 			 * recirculation.
2263 			 */
2264 			eth_type = htons(0);
2265 			break;
2266 
2267 		case OVS_ACTION_ATTR_SET:
2268 			err = validate_set(a, key, sfa,
2269 					   &skip_copy, eth_type, false, log);
2270 			if (err)
2271 				return err;
2272 			break;
2273 
2274 		case OVS_ACTION_ATTR_SET_MASKED:
2275 			err = validate_set(a, key, sfa,
2276 					   &skip_copy, eth_type, true, log);
2277 			if (err)
2278 				return err;
2279 			break;
2280 
2281 		case OVS_ACTION_ATTR_SAMPLE:
2282 			err = validate_and_copy_sample(net, a, key, depth, sfa,
2283 						       eth_type, vlan_tci, log);
2284 			if (err)
2285 				return err;
2286 			skip_copy = true;
2287 			break;
2288 
2289 		case OVS_ACTION_ATTR_CT:
2290 			err = ovs_ct_copy_action(net, a, key, sfa, log);
2291 			if (err)
2292 				return err;
2293 			skip_copy = true;
2294 			break;
2295 
2296 		default:
2297 			OVS_NLERR(log, "Unknown Action type %d", type);
2298 			return -EINVAL;
2299 		}
2300 		if (!skip_copy) {
2301 			err = copy_action(a, sfa, log);
2302 			if (err)
2303 				return err;
2304 		}
2305 	}
2306 
2307 	if (rem > 0)
2308 		return -EINVAL;
2309 
2310 	return 0;
2311 }
2312 
2313 /* 'key' must be the masked key. */
2314 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2315 			 const struct sw_flow_key *key,
2316 			 struct sw_flow_actions **sfa, bool log)
2317 {
2318 	int err;
2319 
2320 	*sfa = nla_alloc_flow_actions(nla_len(attr), log);
2321 	if (IS_ERR(*sfa))
2322 		return PTR_ERR(*sfa);
2323 
2324 	(*sfa)->orig_len = nla_len(attr);
2325 	err = __ovs_nla_copy_actions(net, attr, key, 0, sfa, key->eth.type,
2326 				     key->eth.tci, log);
2327 	if (err)
2328 		ovs_nla_free_flow_actions(*sfa);
2329 
2330 	return err;
2331 }
2332 
2333 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
2334 {
2335 	const struct nlattr *a;
2336 	struct nlattr *start;
2337 	int err = 0, rem;
2338 
2339 	start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
2340 	if (!start)
2341 		return -EMSGSIZE;
2342 
2343 	nla_for_each_nested(a, attr, rem) {
2344 		int type = nla_type(a);
2345 		struct nlattr *st_sample;
2346 
2347 		switch (type) {
2348 		case OVS_SAMPLE_ATTR_PROBABILITY:
2349 			if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
2350 				    sizeof(u32), nla_data(a)))
2351 				return -EMSGSIZE;
2352 			break;
2353 		case OVS_SAMPLE_ATTR_ACTIONS:
2354 			st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
2355 			if (!st_sample)
2356 				return -EMSGSIZE;
2357 			err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
2358 			if (err)
2359 				return err;
2360 			nla_nest_end(skb, st_sample);
2361 			break;
2362 		}
2363 	}
2364 
2365 	nla_nest_end(skb, start);
2366 	return err;
2367 }
2368 
2369 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
2370 {
2371 	const struct nlattr *ovs_key = nla_data(a);
2372 	int key_type = nla_type(ovs_key);
2373 	struct nlattr *start;
2374 	int err;
2375 
2376 	switch (key_type) {
2377 	case OVS_KEY_ATTR_TUNNEL_INFO: {
2378 		struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
2379 		struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
2380 
2381 		start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2382 		if (!start)
2383 			return -EMSGSIZE;
2384 
2385 		err = ovs_nla_put_tunnel_info(skb, tun_info);
2386 		if (err)
2387 			return err;
2388 		nla_nest_end(skb, start);
2389 		break;
2390 	}
2391 	default:
2392 		if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
2393 			return -EMSGSIZE;
2394 		break;
2395 	}
2396 
2397 	return 0;
2398 }
2399 
2400 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
2401 						struct sk_buff *skb)
2402 {
2403 	const struct nlattr *ovs_key = nla_data(a);
2404 	struct nlattr *nla;
2405 	size_t key_len = nla_len(ovs_key) / 2;
2406 
2407 	/* Revert the conversion we did from a non-masked set action to
2408 	 * masked set action.
2409 	 */
2410 	nla = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2411 	if (!nla)
2412 		return -EMSGSIZE;
2413 
2414 	if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
2415 		return -EMSGSIZE;
2416 
2417 	nla_nest_end(skb, nla);
2418 	return 0;
2419 }
2420 
2421 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
2422 {
2423 	const struct nlattr *a;
2424 	int rem, err;
2425 
2426 	nla_for_each_attr(a, attr, len, rem) {
2427 		int type = nla_type(a);
2428 
2429 		switch (type) {
2430 		case OVS_ACTION_ATTR_SET:
2431 			err = set_action_to_attr(a, skb);
2432 			if (err)
2433 				return err;
2434 			break;
2435 
2436 		case OVS_ACTION_ATTR_SET_TO_MASKED:
2437 			err = masked_set_action_to_set_action_attr(a, skb);
2438 			if (err)
2439 				return err;
2440 			break;
2441 
2442 		case OVS_ACTION_ATTR_SAMPLE:
2443 			err = sample_action_to_attr(a, skb);
2444 			if (err)
2445 				return err;
2446 			break;
2447 
2448 		case OVS_ACTION_ATTR_CT:
2449 			err = ovs_ct_action_to_attr(nla_data(a), skb);
2450 			if (err)
2451 				return err;
2452 			break;
2453 
2454 		default:
2455 			if (nla_put(skb, type, nla_len(a), nla_data(a)))
2456 				return -EMSGSIZE;
2457 			break;
2458 		}
2459 	}
2460 
2461 	return 0;
2462 }
2463