xref: /linux/net/openvswitch/flow_netlink.c (revision 3c2f85b8ce8acee0502d61fb53015eabd7d4c8fb)
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 (tun_opts) {
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_egress_tunnel_key(struct sk_buff *skb,
753 				  const struct ip_tunnel_info *egress_tun_info,
754 				  const void *egress_tun_opts)
755 {
756 	return __ipv4_tun_to_nlattr(skb, &egress_tun_info->key,
757 				    egress_tun_opts,
758 				    egress_tun_info->options_len);
759 }
760 
761 static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
762 				 u64 *attrs, const struct nlattr **a,
763 				 bool is_mask, bool log)
764 {
765 	if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
766 		u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
767 
768 		SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
769 		*attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
770 	}
771 
772 	if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
773 		u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
774 
775 		SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
776 		*attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
777 	}
778 
779 	if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
780 		SW_FLOW_KEY_PUT(match, phy.priority,
781 			  nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
782 		*attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
783 	}
784 
785 	if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
786 		u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
787 
788 		if (is_mask) {
789 			in_port = 0xffffffff; /* Always exact match in_port. */
790 		} else if (in_port >= DP_MAX_PORTS) {
791 			OVS_NLERR(log, "Port %d exceeds max allowable %d",
792 				  in_port, DP_MAX_PORTS);
793 			return -EINVAL;
794 		}
795 
796 		SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
797 		*attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
798 	} else if (!is_mask) {
799 		SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
800 	}
801 
802 	if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
803 		uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
804 
805 		SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
806 		*attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
807 	}
808 	if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
809 		if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
810 					 is_mask, log) < 0)
811 			return -EINVAL;
812 		*attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
813 	}
814 
815 	if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
816 	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
817 		u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
818 
819 		if (!is_mask && !ovs_ct_state_supported(ct_state)) {
820 			OVS_NLERR(log, "ct_state flags %08x unsupported",
821 				  ct_state);
822 			return -EINVAL;
823 		}
824 
825 		SW_FLOW_KEY_PUT(match, ct.state, ct_state, is_mask);
826 		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
827 	}
828 	if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
829 	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
830 		u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
831 
832 		SW_FLOW_KEY_PUT(match, ct.zone, ct_zone, is_mask);
833 		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
834 	}
835 	if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
836 	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
837 		u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
838 
839 		SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
840 		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
841 	}
842 	if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
843 	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
844 		const struct ovs_key_ct_labels *cl;
845 
846 		cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
847 		SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
848 				   sizeof(*cl), is_mask);
849 		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
850 	}
851 	return 0;
852 }
853 
854 static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
855 				u64 attrs, const struct nlattr **a,
856 				bool is_mask, bool log)
857 {
858 	int err;
859 
860 	err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
861 	if (err)
862 		return err;
863 
864 	if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
865 		const struct ovs_key_ethernet *eth_key;
866 
867 		eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
868 		SW_FLOW_KEY_MEMCPY(match, eth.src,
869 				eth_key->eth_src, ETH_ALEN, is_mask);
870 		SW_FLOW_KEY_MEMCPY(match, eth.dst,
871 				eth_key->eth_dst, ETH_ALEN, is_mask);
872 		attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
873 	}
874 
875 	if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
876 		__be16 tci;
877 
878 		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
879 		if (!(tci & htons(VLAN_TAG_PRESENT))) {
880 			if (is_mask)
881 				OVS_NLERR(log, "VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.");
882 			else
883 				OVS_NLERR(log, "VLAN TCI does not have VLAN_TAG_PRESENT bit set.");
884 
885 			return -EINVAL;
886 		}
887 
888 		SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
889 		attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
890 	}
891 
892 	if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
893 		__be16 eth_type;
894 
895 		eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
896 		if (is_mask) {
897 			/* Always exact match EtherType. */
898 			eth_type = htons(0xffff);
899 		} else if (!eth_proto_is_802_3(eth_type)) {
900 			OVS_NLERR(log, "EtherType %x is less than min %x",
901 				  ntohs(eth_type), ETH_P_802_3_MIN);
902 			return -EINVAL;
903 		}
904 
905 		SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
906 		attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
907 	} else if (!is_mask) {
908 		SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
909 	}
910 
911 	if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
912 		const struct ovs_key_ipv4 *ipv4_key;
913 
914 		ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
915 		if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
916 			OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
917 				  ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
918 			return -EINVAL;
919 		}
920 		SW_FLOW_KEY_PUT(match, ip.proto,
921 				ipv4_key->ipv4_proto, is_mask);
922 		SW_FLOW_KEY_PUT(match, ip.tos,
923 				ipv4_key->ipv4_tos, is_mask);
924 		SW_FLOW_KEY_PUT(match, ip.ttl,
925 				ipv4_key->ipv4_ttl, is_mask);
926 		SW_FLOW_KEY_PUT(match, ip.frag,
927 				ipv4_key->ipv4_frag, is_mask);
928 		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
929 				ipv4_key->ipv4_src, is_mask);
930 		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
931 				ipv4_key->ipv4_dst, is_mask);
932 		attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
933 	}
934 
935 	if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
936 		const struct ovs_key_ipv6 *ipv6_key;
937 
938 		ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
939 		if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
940 			OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
941 				  ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
942 			return -EINVAL;
943 		}
944 
945 		if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
946 			OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x).\n",
947 				  ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
948 			return -EINVAL;
949 		}
950 
951 		SW_FLOW_KEY_PUT(match, ipv6.label,
952 				ipv6_key->ipv6_label, is_mask);
953 		SW_FLOW_KEY_PUT(match, ip.proto,
954 				ipv6_key->ipv6_proto, is_mask);
955 		SW_FLOW_KEY_PUT(match, ip.tos,
956 				ipv6_key->ipv6_tclass, is_mask);
957 		SW_FLOW_KEY_PUT(match, ip.ttl,
958 				ipv6_key->ipv6_hlimit, is_mask);
959 		SW_FLOW_KEY_PUT(match, ip.frag,
960 				ipv6_key->ipv6_frag, is_mask);
961 		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
962 				ipv6_key->ipv6_src,
963 				sizeof(match->key->ipv6.addr.src),
964 				is_mask);
965 		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
966 				ipv6_key->ipv6_dst,
967 				sizeof(match->key->ipv6.addr.dst),
968 				is_mask);
969 
970 		attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
971 	}
972 
973 	if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
974 		const struct ovs_key_arp *arp_key;
975 
976 		arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
977 		if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
978 			OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
979 				  arp_key->arp_op);
980 			return -EINVAL;
981 		}
982 
983 		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
984 				arp_key->arp_sip, is_mask);
985 		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
986 			arp_key->arp_tip, is_mask);
987 		SW_FLOW_KEY_PUT(match, ip.proto,
988 				ntohs(arp_key->arp_op), is_mask);
989 		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
990 				arp_key->arp_sha, ETH_ALEN, is_mask);
991 		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
992 				arp_key->arp_tha, ETH_ALEN, is_mask);
993 
994 		attrs &= ~(1 << OVS_KEY_ATTR_ARP);
995 	}
996 
997 	if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
998 		const struct ovs_key_mpls *mpls_key;
999 
1000 		mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1001 		SW_FLOW_KEY_PUT(match, mpls.top_lse,
1002 				mpls_key->mpls_lse, is_mask);
1003 
1004 		attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1005 	 }
1006 
1007 	if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1008 		const struct ovs_key_tcp *tcp_key;
1009 
1010 		tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1011 		SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1012 		SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1013 		attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1014 	}
1015 
1016 	if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1017 		SW_FLOW_KEY_PUT(match, tp.flags,
1018 				nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1019 				is_mask);
1020 		attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1021 	}
1022 
1023 	if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1024 		const struct ovs_key_udp *udp_key;
1025 
1026 		udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1027 		SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1028 		SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1029 		attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1030 	}
1031 
1032 	if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1033 		const struct ovs_key_sctp *sctp_key;
1034 
1035 		sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1036 		SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1037 		SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1038 		attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1039 	}
1040 
1041 	if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1042 		const struct ovs_key_icmp *icmp_key;
1043 
1044 		icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1045 		SW_FLOW_KEY_PUT(match, tp.src,
1046 				htons(icmp_key->icmp_type), is_mask);
1047 		SW_FLOW_KEY_PUT(match, tp.dst,
1048 				htons(icmp_key->icmp_code), is_mask);
1049 		attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1050 	}
1051 
1052 	if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1053 		const struct ovs_key_icmpv6 *icmpv6_key;
1054 
1055 		icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1056 		SW_FLOW_KEY_PUT(match, tp.src,
1057 				htons(icmpv6_key->icmpv6_type), is_mask);
1058 		SW_FLOW_KEY_PUT(match, tp.dst,
1059 				htons(icmpv6_key->icmpv6_code), is_mask);
1060 		attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1061 	}
1062 
1063 	if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1064 		const struct ovs_key_nd *nd_key;
1065 
1066 		nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1067 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1068 			nd_key->nd_target,
1069 			sizeof(match->key->ipv6.nd.target),
1070 			is_mask);
1071 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1072 			nd_key->nd_sll, ETH_ALEN, is_mask);
1073 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1074 				nd_key->nd_tll, ETH_ALEN, is_mask);
1075 		attrs &= ~(1 << OVS_KEY_ATTR_ND);
1076 	}
1077 
1078 	if (attrs != 0) {
1079 		OVS_NLERR(log, "Unknown key attributes %llx",
1080 			  (unsigned long long)attrs);
1081 		return -EINVAL;
1082 	}
1083 
1084 	return 0;
1085 }
1086 
1087 static void nlattr_set(struct nlattr *attr, u8 val,
1088 		       const struct ovs_len_tbl *tbl)
1089 {
1090 	struct nlattr *nla;
1091 	int rem;
1092 
1093 	/* The nlattr stream should already have been validated */
1094 	nla_for_each_nested(nla, attr, rem) {
1095 		if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED) {
1096 			if (tbl[nla_type(nla)].next)
1097 				tbl = tbl[nla_type(nla)].next;
1098 			nlattr_set(nla, val, tbl);
1099 		} else {
1100 			memset(nla_data(nla), val, nla_len(nla));
1101 		}
1102 	}
1103 }
1104 
1105 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1106 {
1107 	nlattr_set(attr, val, ovs_key_lens);
1108 }
1109 
1110 /**
1111  * ovs_nla_get_match - parses Netlink attributes into a flow key and
1112  * mask. In case the 'mask' is NULL, the flow is treated as exact match
1113  * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1114  * does not include any don't care bit.
1115  * @net: Used to determine per-namespace field support.
1116  * @match: receives the extracted flow match information.
1117  * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1118  * sequence. The fields should of the packet that triggered the creation
1119  * of this flow.
1120  * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1121  * attribute specifies the mask field of the wildcarded flow.
1122  * @log: Boolean to allow kernel error logging.  Normally true, but when
1123  * probing for feature compatibility this should be passed in as false to
1124  * suppress unnecessary error logging.
1125  */
1126 int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1127 		      const struct nlattr *nla_key,
1128 		      const struct nlattr *nla_mask,
1129 		      bool log)
1130 {
1131 	const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1132 	const struct nlattr *encap;
1133 	struct nlattr *newmask = NULL;
1134 	u64 key_attrs = 0;
1135 	u64 mask_attrs = 0;
1136 	bool encap_valid = false;
1137 	int err;
1138 
1139 	err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1140 	if (err)
1141 		return err;
1142 
1143 	if ((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
1144 	    (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
1145 	    (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
1146 		__be16 tci;
1147 
1148 		if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
1149 		      (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
1150 			OVS_NLERR(log, "Invalid Vlan frame.");
1151 			return -EINVAL;
1152 		}
1153 
1154 		key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1155 		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1156 		encap = a[OVS_KEY_ATTR_ENCAP];
1157 		key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1158 		encap_valid = true;
1159 
1160 		if (tci & htons(VLAN_TAG_PRESENT)) {
1161 			err = parse_flow_nlattrs(encap, a, &key_attrs, log);
1162 			if (err)
1163 				return err;
1164 		} else if (!tci) {
1165 			/* Corner case for truncated 802.1Q header. */
1166 			if (nla_len(encap)) {
1167 				OVS_NLERR(log, "Truncated 802.1Q header has non-zero encap attribute.");
1168 				return -EINVAL;
1169 			}
1170 		} else {
1171 			OVS_NLERR(log, "Encap attr is set for non-VLAN frame");
1172 			return  -EINVAL;
1173 		}
1174 	}
1175 
1176 	err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1177 	if (err)
1178 		return err;
1179 
1180 	if (match->mask) {
1181 		if (!nla_mask) {
1182 			/* Create an exact match mask. We need to set to 0xff
1183 			 * all the 'match->mask' fields that have been touched
1184 			 * in 'match->key'. We cannot simply memset
1185 			 * 'match->mask', because padding bytes and fields not
1186 			 * specified in 'match->key' should be left to 0.
1187 			 * Instead, we use a stream of netlink attributes,
1188 			 * copied from 'key' and set to 0xff.
1189 			 * ovs_key_from_nlattrs() will take care of filling
1190 			 * 'match->mask' appropriately.
1191 			 */
1192 			newmask = kmemdup(nla_key,
1193 					  nla_total_size(nla_len(nla_key)),
1194 					  GFP_KERNEL);
1195 			if (!newmask)
1196 				return -ENOMEM;
1197 
1198 			mask_set_nlattr(newmask, 0xff);
1199 
1200 			/* The userspace does not send tunnel attributes that
1201 			 * are 0, but we should not wildcard them nonetheless.
1202 			 */
1203 			if (match->key->tun_key.u.ipv4.dst)
1204 				SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1205 							 0xff, true);
1206 
1207 			nla_mask = newmask;
1208 		}
1209 
1210 		err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1211 		if (err)
1212 			goto free_newmask;
1213 
1214 		/* Always match on tci. */
1215 		SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
1216 
1217 		if (mask_attrs & 1 << OVS_KEY_ATTR_ENCAP) {
1218 			__be16 eth_type = 0;
1219 			__be16 tci = 0;
1220 
1221 			if (!encap_valid) {
1222 				OVS_NLERR(log, "Encap mask attribute is set for non-VLAN frame.");
1223 				err = -EINVAL;
1224 				goto free_newmask;
1225 			}
1226 
1227 			mask_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1228 			if (a[OVS_KEY_ATTR_ETHERTYPE])
1229 				eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1230 
1231 			if (eth_type == htons(0xffff)) {
1232 				mask_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1233 				encap = a[OVS_KEY_ATTR_ENCAP];
1234 				err = parse_flow_mask_nlattrs(encap, a,
1235 							      &mask_attrs, log);
1236 				if (err)
1237 					goto free_newmask;
1238 			} else {
1239 				OVS_NLERR(log, "VLAN frames must have an exact match on the TPID (mask=%x).",
1240 					  ntohs(eth_type));
1241 				err = -EINVAL;
1242 				goto free_newmask;
1243 			}
1244 
1245 			if (a[OVS_KEY_ATTR_VLAN])
1246 				tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1247 
1248 			if (!(tci & htons(VLAN_TAG_PRESENT))) {
1249 				OVS_NLERR(log, "VLAN tag present bit must have an exact match (tci_mask=%x).",
1250 					  ntohs(tci));
1251 				err = -EINVAL;
1252 				goto free_newmask;
1253 			}
1254 		}
1255 
1256 		err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1257 					   log);
1258 		if (err)
1259 			goto free_newmask;
1260 	}
1261 
1262 	if (!match_validate(match, key_attrs, mask_attrs, log))
1263 		err = -EINVAL;
1264 
1265 free_newmask:
1266 	kfree(newmask);
1267 	return err;
1268 }
1269 
1270 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1271 {
1272 	size_t len;
1273 
1274 	if (!attr)
1275 		return 0;
1276 
1277 	len = nla_len(attr);
1278 	if (len < 1 || len > MAX_UFID_LENGTH) {
1279 		OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1280 			  nla_len(attr), MAX_UFID_LENGTH);
1281 		return 0;
1282 	}
1283 
1284 	return len;
1285 }
1286 
1287 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1288  * or false otherwise.
1289  */
1290 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1291 		      bool log)
1292 {
1293 	sfid->ufid_len = get_ufid_len(attr, log);
1294 	if (sfid->ufid_len)
1295 		memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1296 
1297 	return sfid->ufid_len;
1298 }
1299 
1300 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1301 			   const struct sw_flow_key *key, bool log)
1302 {
1303 	struct sw_flow_key *new_key;
1304 
1305 	if (ovs_nla_get_ufid(sfid, ufid, log))
1306 		return 0;
1307 
1308 	/* If UFID was not provided, use unmasked key. */
1309 	new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1310 	if (!new_key)
1311 		return -ENOMEM;
1312 	memcpy(new_key, key, sizeof(*key));
1313 	sfid->unmasked_key = new_key;
1314 
1315 	return 0;
1316 }
1317 
1318 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1319 {
1320 	return attr ? nla_get_u32(attr) : 0;
1321 }
1322 
1323 /**
1324  * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1325  * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1326  * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1327  * sequence.
1328  * @log: Boolean to allow kernel error logging.  Normally true, but when
1329  * probing for feature compatibility this should be passed in as false to
1330  * suppress unnecessary error logging.
1331  *
1332  * This parses a series of Netlink attributes that form a flow key, which must
1333  * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1334  * get the metadata, that is, the parts of the flow key that cannot be
1335  * extracted from the packet itself.
1336  */
1337 
1338 int ovs_nla_get_flow_metadata(struct net *net, const struct nlattr *attr,
1339 			      struct sw_flow_key *key,
1340 			      bool log)
1341 {
1342 	const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1343 	struct sw_flow_match match;
1344 	u64 attrs = 0;
1345 	int err;
1346 
1347 	err = parse_flow_nlattrs(attr, a, &attrs, log);
1348 	if (err)
1349 		return -EINVAL;
1350 
1351 	memset(&match, 0, sizeof(match));
1352 	match.key = key;
1353 
1354 	memset(&key->ct, 0, sizeof(key->ct));
1355 	key->phy.in_port = DP_MAX_PORTS;
1356 
1357 	return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1358 }
1359 
1360 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1361 			     const struct sw_flow_key *output, bool is_mask,
1362 			     struct sk_buff *skb)
1363 {
1364 	struct ovs_key_ethernet *eth_key;
1365 	struct nlattr *nla, *encap;
1366 
1367 	if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1368 		goto nla_put_failure;
1369 
1370 	if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1371 		goto nla_put_failure;
1372 
1373 	if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1374 		goto nla_put_failure;
1375 
1376 	if ((swkey->tun_key.u.ipv4.dst || is_mask)) {
1377 		const void *opts = NULL;
1378 
1379 		if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1380 			opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1381 
1382 		if (ipv4_tun_to_nlattr(skb, &output->tun_key, opts,
1383 				       swkey->tun_opts_len))
1384 			goto nla_put_failure;
1385 	}
1386 
1387 	if (swkey->phy.in_port == DP_MAX_PORTS) {
1388 		if (is_mask && (output->phy.in_port == 0xffff))
1389 			if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1390 				goto nla_put_failure;
1391 	} else {
1392 		u16 upper_u16;
1393 		upper_u16 = !is_mask ? 0 : 0xffff;
1394 
1395 		if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1396 				(upper_u16 << 16) | output->phy.in_port))
1397 			goto nla_put_failure;
1398 	}
1399 
1400 	if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1401 		goto nla_put_failure;
1402 
1403 	if (ovs_ct_put_key(output, skb))
1404 		goto nla_put_failure;
1405 
1406 	nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1407 	if (!nla)
1408 		goto nla_put_failure;
1409 
1410 	eth_key = nla_data(nla);
1411 	ether_addr_copy(eth_key->eth_src, output->eth.src);
1412 	ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1413 
1414 	if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1415 		__be16 eth_type;
1416 		eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
1417 		if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1418 		    nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
1419 			goto nla_put_failure;
1420 		encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1421 		if (!swkey->eth.tci)
1422 			goto unencap;
1423 	} else
1424 		encap = NULL;
1425 
1426 	if (swkey->eth.type == htons(ETH_P_802_2)) {
1427 		/*
1428 		 * Ethertype 802.2 is represented in the netlink with omitted
1429 		 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1430 		 * 0xffff in the mask attribute.  Ethertype can also
1431 		 * be wildcarded.
1432 		 */
1433 		if (is_mask && output->eth.type)
1434 			if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1435 						output->eth.type))
1436 				goto nla_put_failure;
1437 		goto unencap;
1438 	}
1439 
1440 	if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1441 		goto nla_put_failure;
1442 
1443 	if (swkey->eth.type == htons(ETH_P_IP)) {
1444 		struct ovs_key_ipv4 *ipv4_key;
1445 
1446 		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1447 		if (!nla)
1448 			goto nla_put_failure;
1449 		ipv4_key = nla_data(nla);
1450 		ipv4_key->ipv4_src = output->ipv4.addr.src;
1451 		ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1452 		ipv4_key->ipv4_proto = output->ip.proto;
1453 		ipv4_key->ipv4_tos = output->ip.tos;
1454 		ipv4_key->ipv4_ttl = output->ip.ttl;
1455 		ipv4_key->ipv4_frag = output->ip.frag;
1456 	} else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1457 		struct ovs_key_ipv6 *ipv6_key;
1458 
1459 		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1460 		if (!nla)
1461 			goto nla_put_failure;
1462 		ipv6_key = nla_data(nla);
1463 		memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1464 				sizeof(ipv6_key->ipv6_src));
1465 		memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1466 				sizeof(ipv6_key->ipv6_dst));
1467 		ipv6_key->ipv6_label = output->ipv6.label;
1468 		ipv6_key->ipv6_proto = output->ip.proto;
1469 		ipv6_key->ipv6_tclass = output->ip.tos;
1470 		ipv6_key->ipv6_hlimit = output->ip.ttl;
1471 		ipv6_key->ipv6_frag = output->ip.frag;
1472 	} else if (swkey->eth.type == htons(ETH_P_ARP) ||
1473 		   swkey->eth.type == htons(ETH_P_RARP)) {
1474 		struct ovs_key_arp *arp_key;
1475 
1476 		nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1477 		if (!nla)
1478 			goto nla_put_failure;
1479 		arp_key = nla_data(nla);
1480 		memset(arp_key, 0, sizeof(struct ovs_key_arp));
1481 		arp_key->arp_sip = output->ipv4.addr.src;
1482 		arp_key->arp_tip = output->ipv4.addr.dst;
1483 		arp_key->arp_op = htons(output->ip.proto);
1484 		ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1485 		ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1486 	} else if (eth_p_mpls(swkey->eth.type)) {
1487 		struct ovs_key_mpls *mpls_key;
1488 
1489 		nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1490 		if (!nla)
1491 			goto nla_put_failure;
1492 		mpls_key = nla_data(nla);
1493 		mpls_key->mpls_lse = output->mpls.top_lse;
1494 	}
1495 
1496 	if ((swkey->eth.type == htons(ETH_P_IP) ||
1497 	     swkey->eth.type == htons(ETH_P_IPV6)) &&
1498 	     swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1499 
1500 		if (swkey->ip.proto == IPPROTO_TCP) {
1501 			struct ovs_key_tcp *tcp_key;
1502 
1503 			nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1504 			if (!nla)
1505 				goto nla_put_failure;
1506 			tcp_key = nla_data(nla);
1507 			tcp_key->tcp_src = output->tp.src;
1508 			tcp_key->tcp_dst = output->tp.dst;
1509 			if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1510 					 output->tp.flags))
1511 				goto nla_put_failure;
1512 		} else if (swkey->ip.proto == IPPROTO_UDP) {
1513 			struct ovs_key_udp *udp_key;
1514 
1515 			nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1516 			if (!nla)
1517 				goto nla_put_failure;
1518 			udp_key = nla_data(nla);
1519 			udp_key->udp_src = output->tp.src;
1520 			udp_key->udp_dst = output->tp.dst;
1521 		} else if (swkey->ip.proto == IPPROTO_SCTP) {
1522 			struct ovs_key_sctp *sctp_key;
1523 
1524 			nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1525 			if (!nla)
1526 				goto nla_put_failure;
1527 			sctp_key = nla_data(nla);
1528 			sctp_key->sctp_src = output->tp.src;
1529 			sctp_key->sctp_dst = output->tp.dst;
1530 		} else if (swkey->eth.type == htons(ETH_P_IP) &&
1531 			   swkey->ip.proto == IPPROTO_ICMP) {
1532 			struct ovs_key_icmp *icmp_key;
1533 
1534 			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1535 			if (!nla)
1536 				goto nla_put_failure;
1537 			icmp_key = nla_data(nla);
1538 			icmp_key->icmp_type = ntohs(output->tp.src);
1539 			icmp_key->icmp_code = ntohs(output->tp.dst);
1540 		} else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1541 			   swkey->ip.proto == IPPROTO_ICMPV6) {
1542 			struct ovs_key_icmpv6 *icmpv6_key;
1543 
1544 			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1545 						sizeof(*icmpv6_key));
1546 			if (!nla)
1547 				goto nla_put_failure;
1548 			icmpv6_key = nla_data(nla);
1549 			icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1550 			icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1551 
1552 			if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1553 			    icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1554 				struct ovs_key_nd *nd_key;
1555 
1556 				nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1557 				if (!nla)
1558 					goto nla_put_failure;
1559 				nd_key = nla_data(nla);
1560 				memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1561 							sizeof(nd_key->nd_target));
1562 				ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1563 				ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1564 			}
1565 		}
1566 	}
1567 
1568 unencap:
1569 	if (encap)
1570 		nla_nest_end(skb, encap);
1571 
1572 	return 0;
1573 
1574 nla_put_failure:
1575 	return -EMSGSIZE;
1576 }
1577 
1578 int ovs_nla_put_key(const struct sw_flow_key *swkey,
1579 		    const struct sw_flow_key *output, int attr, bool is_mask,
1580 		    struct sk_buff *skb)
1581 {
1582 	int err;
1583 	struct nlattr *nla;
1584 
1585 	nla = nla_nest_start(skb, attr);
1586 	if (!nla)
1587 		return -EMSGSIZE;
1588 	err = __ovs_nla_put_key(swkey, output, is_mask, skb);
1589 	if (err)
1590 		return err;
1591 	nla_nest_end(skb, nla);
1592 
1593 	return 0;
1594 }
1595 
1596 /* Called with ovs_mutex or RCU read lock. */
1597 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
1598 {
1599 	if (ovs_identifier_is_ufid(&flow->id))
1600 		return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
1601 			       flow->id.ufid);
1602 
1603 	return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
1604 			       OVS_FLOW_ATTR_KEY, false, skb);
1605 }
1606 
1607 /* Called with ovs_mutex or RCU read lock. */
1608 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
1609 {
1610 	return ovs_nla_put_key(&flow->key, &flow->key,
1611 				OVS_FLOW_ATTR_KEY, false, skb);
1612 }
1613 
1614 /* Called with ovs_mutex or RCU read lock. */
1615 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
1616 {
1617 	return ovs_nla_put_key(&flow->key, &flow->mask->key,
1618 				OVS_FLOW_ATTR_MASK, true, skb);
1619 }
1620 
1621 #define MAX_ACTIONS_BUFSIZE	(32 * 1024)
1622 
1623 static struct sw_flow_actions *nla_alloc_flow_actions(int size, bool log)
1624 {
1625 	struct sw_flow_actions *sfa;
1626 
1627 	if (size > MAX_ACTIONS_BUFSIZE) {
1628 		OVS_NLERR(log, "Flow action size %u bytes exceeds max", size);
1629 		return ERR_PTR(-EINVAL);
1630 	}
1631 
1632 	sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1633 	if (!sfa)
1634 		return ERR_PTR(-ENOMEM);
1635 
1636 	sfa->actions_len = 0;
1637 	return sfa;
1638 }
1639 
1640 static void ovs_nla_free_set_action(const struct nlattr *a)
1641 {
1642 	const struct nlattr *ovs_key = nla_data(a);
1643 	struct ovs_tunnel_info *ovs_tun;
1644 
1645 	switch (nla_type(ovs_key)) {
1646 	case OVS_KEY_ATTR_TUNNEL_INFO:
1647 		ovs_tun = nla_data(ovs_key);
1648 		dst_release((struct dst_entry *)ovs_tun->tun_dst);
1649 		break;
1650 	}
1651 }
1652 
1653 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1654 {
1655 	const struct nlattr *a;
1656 	int rem;
1657 
1658 	if (!sf_acts)
1659 		return;
1660 
1661 	nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
1662 		switch (nla_type(a)) {
1663 		case OVS_ACTION_ATTR_SET:
1664 			ovs_nla_free_set_action(a);
1665 			break;
1666 		case OVS_ACTION_ATTR_CT:
1667 			ovs_ct_free_action(a);
1668 			break;
1669 		}
1670 	}
1671 
1672 	kfree(sf_acts);
1673 }
1674 
1675 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
1676 {
1677 	ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
1678 }
1679 
1680 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1681  * The caller must hold rcu_read_lock for this to be sensible. */
1682 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
1683 {
1684 	call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
1685 }
1686 
1687 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1688 				       int attr_len, bool log)
1689 {
1690 
1691 	struct sw_flow_actions *acts;
1692 	int new_acts_size;
1693 	int req_size = NLA_ALIGN(attr_len);
1694 	int next_offset = offsetof(struct sw_flow_actions, actions) +
1695 					(*sfa)->actions_len;
1696 
1697 	if (req_size <= (ksize(*sfa) - next_offset))
1698 		goto out;
1699 
1700 	new_acts_size = ksize(*sfa) * 2;
1701 
1702 	if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1703 		if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1704 			return ERR_PTR(-EMSGSIZE);
1705 		new_acts_size = MAX_ACTIONS_BUFSIZE;
1706 	}
1707 
1708 	acts = nla_alloc_flow_actions(new_acts_size, log);
1709 	if (IS_ERR(acts))
1710 		return (void *)acts;
1711 
1712 	memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1713 	acts->actions_len = (*sfa)->actions_len;
1714 	acts->orig_len = (*sfa)->orig_len;
1715 	kfree(*sfa);
1716 	*sfa = acts;
1717 
1718 out:
1719 	(*sfa)->actions_len += req_size;
1720 	return  (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1721 }
1722 
1723 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1724 				   int attrtype, void *data, int len, bool log)
1725 {
1726 	struct nlattr *a;
1727 
1728 	a = reserve_sfa_size(sfa, nla_attr_size(len), log);
1729 	if (IS_ERR(a))
1730 		return a;
1731 
1732 	a->nla_type = attrtype;
1733 	a->nla_len = nla_attr_size(len);
1734 
1735 	if (data)
1736 		memcpy(nla_data(a), data, len);
1737 	memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1738 
1739 	return a;
1740 }
1741 
1742 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
1743 		       int len, bool log)
1744 {
1745 	struct nlattr *a;
1746 
1747 	a = __add_action(sfa, attrtype, data, len, log);
1748 
1749 	return PTR_ERR_OR_ZERO(a);
1750 }
1751 
1752 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1753 					  int attrtype, bool log)
1754 {
1755 	int used = (*sfa)->actions_len;
1756 	int err;
1757 
1758 	err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
1759 	if (err)
1760 		return err;
1761 
1762 	return used;
1763 }
1764 
1765 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1766 					 int st_offset)
1767 {
1768 	struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1769 							       st_offset);
1770 
1771 	a->nla_len = sfa->actions_len - st_offset;
1772 }
1773 
1774 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
1775 				  const struct sw_flow_key *key,
1776 				  int depth, struct sw_flow_actions **sfa,
1777 				  __be16 eth_type, __be16 vlan_tci, bool log);
1778 
1779 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
1780 				    const struct sw_flow_key *key, int depth,
1781 				    struct sw_flow_actions **sfa,
1782 				    __be16 eth_type, __be16 vlan_tci, bool log)
1783 {
1784 	const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1785 	const struct nlattr *probability, *actions;
1786 	const struct nlattr *a;
1787 	int rem, start, err, st_acts;
1788 
1789 	memset(attrs, 0, sizeof(attrs));
1790 	nla_for_each_nested(a, attr, rem) {
1791 		int type = nla_type(a);
1792 		if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1793 			return -EINVAL;
1794 		attrs[type] = a;
1795 	}
1796 	if (rem)
1797 		return -EINVAL;
1798 
1799 	probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1800 	if (!probability || nla_len(probability) != sizeof(u32))
1801 		return -EINVAL;
1802 
1803 	actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1804 	if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1805 		return -EINVAL;
1806 
1807 	/* validation done, copy sample action. */
1808 	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
1809 	if (start < 0)
1810 		return start;
1811 	err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1812 				 nla_data(probability), sizeof(u32), log);
1813 	if (err)
1814 		return err;
1815 	st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS, log);
1816 	if (st_acts < 0)
1817 		return st_acts;
1818 
1819 	err = __ovs_nla_copy_actions(net, actions, key, depth + 1, sfa,
1820 				     eth_type, vlan_tci, log);
1821 	if (err)
1822 		return err;
1823 
1824 	add_nested_action_end(*sfa, st_acts);
1825 	add_nested_action_end(*sfa, start);
1826 
1827 	return 0;
1828 }
1829 
1830 void ovs_match_init(struct sw_flow_match *match,
1831 		    struct sw_flow_key *key,
1832 		    struct sw_flow_mask *mask)
1833 {
1834 	memset(match, 0, sizeof(*match));
1835 	match->key = key;
1836 	match->mask = mask;
1837 
1838 	memset(key, 0, sizeof(*key));
1839 
1840 	if (mask) {
1841 		memset(&mask->key, 0, sizeof(mask->key));
1842 		mask->range.start = mask->range.end = 0;
1843 	}
1844 }
1845 
1846 static int validate_geneve_opts(struct sw_flow_key *key)
1847 {
1848 	struct geneve_opt *option;
1849 	int opts_len = key->tun_opts_len;
1850 	bool crit_opt = false;
1851 
1852 	option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
1853 	while (opts_len > 0) {
1854 		int len;
1855 
1856 		if (opts_len < sizeof(*option))
1857 			return -EINVAL;
1858 
1859 		len = sizeof(*option) + option->length * 4;
1860 		if (len > opts_len)
1861 			return -EINVAL;
1862 
1863 		crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
1864 
1865 		option = (struct geneve_opt *)((u8 *)option + len);
1866 		opts_len -= len;
1867 	};
1868 
1869 	key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
1870 
1871 	return 0;
1872 }
1873 
1874 static int validate_and_copy_set_tun(const struct nlattr *attr,
1875 				     struct sw_flow_actions **sfa, bool log)
1876 {
1877 	struct sw_flow_match match;
1878 	struct sw_flow_key key;
1879 	struct metadata_dst *tun_dst;
1880 	struct ip_tunnel_info *tun_info;
1881 	struct ovs_tunnel_info *ovs_tun;
1882 	struct nlattr *a;
1883 	int err = 0, start, opts_type;
1884 
1885 	ovs_match_init(&match, &key, NULL);
1886 	opts_type = ipv4_tun_from_nlattr(nla_data(attr), &match, false, log);
1887 	if (opts_type < 0)
1888 		return opts_type;
1889 
1890 	if (key.tun_opts_len) {
1891 		switch (opts_type) {
1892 		case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
1893 			err = validate_geneve_opts(&key);
1894 			if (err < 0)
1895 				return err;
1896 			break;
1897 		case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
1898 			break;
1899 		}
1900 	};
1901 
1902 	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
1903 	if (start < 0)
1904 		return start;
1905 
1906 	tun_dst = metadata_dst_alloc(key.tun_opts_len, GFP_KERNEL);
1907 	if (!tun_dst)
1908 		return -ENOMEM;
1909 
1910 	a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
1911 			 sizeof(*ovs_tun), log);
1912 	if (IS_ERR(a)) {
1913 		dst_release((struct dst_entry *)tun_dst);
1914 		return PTR_ERR(a);
1915 	}
1916 
1917 	ovs_tun = nla_data(a);
1918 	ovs_tun->tun_dst = tun_dst;
1919 
1920 	tun_info = &tun_dst->u.tun_info;
1921 	tun_info->mode = IP_TUNNEL_INFO_TX;
1922 	tun_info->key = key.tun_key;
1923 
1924 	/* We need to store the options in the action itself since
1925 	 * everything else will go away after flow setup. We can append
1926 	 * it to tun_info and then point there.
1927 	 */
1928 	ip_tunnel_info_opts_set(tun_info,
1929 				TUN_METADATA_OPTS(&key, key.tun_opts_len),
1930 				key.tun_opts_len);
1931 	add_nested_action_end(*sfa, start);
1932 
1933 	return err;
1934 }
1935 
1936 /* Return false if there are any non-masked bits set.
1937  * Mask follows data immediately, before any netlink padding.
1938  */
1939 static bool validate_masked(u8 *data, int len)
1940 {
1941 	u8 *mask = data + len;
1942 
1943 	while (len--)
1944 		if (*data++ & ~*mask++)
1945 			return false;
1946 
1947 	return true;
1948 }
1949 
1950 static int validate_set(const struct nlattr *a,
1951 			const struct sw_flow_key *flow_key,
1952 			struct sw_flow_actions **sfa,
1953 			bool *skip_copy, __be16 eth_type, bool masked, bool log)
1954 {
1955 	const struct nlattr *ovs_key = nla_data(a);
1956 	int key_type = nla_type(ovs_key);
1957 	size_t key_len;
1958 
1959 	/* There can be only one key in a action */
1960 	if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1961 		return -EINVAL;
1962 
1963 	key_len = nla_len(ovs_key);
1964 	if (masked)
1965 		key_len /= 2;
1966 
1967 	if (key_type > OVS_KEY_ATTR_MAX ||
1968 	    !check_attr_len(key_len, ovs_key_lens[key_type].len))
1969 		return -EINVAL;
1970 
1971 	if (masked && !validate_masked(nla_data(ovs_key), key_len))
1972 		return -EINVAL;
1973 
1974 	switch (key_type) {
1975 	const struct ovs_key_ipv4 *ipv4_key;
1976 	const struct ovs_key_ipv6 *ipv6_key;
1977 	int err;
1978 
1979 	case OVS_KEY_ATTR_PRIORITY:
1980 	case OVS_KEY_ATTR_SKB_MARK:
1981 	case OVS_KEY_ATTR_CT_MARK:
1982 	case OVS_KEY_ATTR_CT_LABELS:
1983 	case OVS_KEY_ATTR_ETHERNET:
1984 		break;
1985 
1986 	case OVS_KEY_ATTR_TUNNEL:
1987 		if (eth_p_mpls(eth_type))
1988 			return -EINVAL;
1989 
1990 		if (masked)
1991 			return -EINVAL; /* Masked tunnel set not supported. */
1992 
1993 		*skip_copy = true;
1994 		err = validate_and_copy_set_tun(a, sfa, log);
1995 		if (err)
1996 			return err;
1997 		break;
1998 
1999 	case OVS_KEY_ATTR_IPV4:
2000 		if (eth_type != htons(ETH_P_IP))
2001 			return -EINVAL;
2002 
2003 		ipv4_key = nla_data(ovs_key);
2004 
2005 		if (masked) {
2006 			const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2007 
2008 			/* Non-writeable fields. */
2009 			if (mask->ipv4_proto || mask->ipv4_frag)
2010 				return -EINVAL;
2011 		} else {
2012 			if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2013 				return -EINVAL;
2014 
2015 			if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2016 				return -EINVAL;
2017 		}
2018 		break;
2019 
2020 	case OVS_KEY_ATTR_IPV6:
2021 		if (eth_type != htons(ETH_P_IPV6))
2022 			return -EINVAL;
2023 
2024 		ipv6_key = nla_data(ovs_key);
2025 
2026 		if (masked) {
2027 			const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2028 
2029 			/* Non-writeable fields. */
2030 			if (mask->ipv6_proto || mask->ipv6_frag)
2031 				return -EINVAL;
2032 
2033 			/* Invalid bits in the flow label mask? */
2034 			if (ntohl(mask->ipv6_label) & 0xFFF00000)
2035 				return -EINVAL;
2036 		} else {
2037 			if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2038 				return -EINVAL;
2039 
2040 			if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2041 				return -EINVAL;
2042 		}
2043 		if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2044 			return -EINVAL;
2045 
2046 		break;
2047 
2048 	case OVS_KEY_ATTR_TCP:
2049 		if ((eth_type != htons(ETH_P_IP) &&
2050 		     eth_type != htons(ETH_P_IPV6)) ||
2051 		    flow_key->ip.proto != IPPROTO_TCP)
2052 			return -EINVAL;
2053 
2054 		break;
2055 
2056 	case OVS_KEY_ATTR_UDP:
2057 		if ((eth_type != htons(ETH_P_IP) &&
2058 		     eth_type != htons(ETH_P_IPV6)) ||
2059 		    flow_key->ip.proto != IPPROTO_UDP)
2060 			return -EINVAL;
2061 
2062 		break;
2063 
2064 	case OVS_KEY_ATTR_MPLS:
2065 		if (!eth_p_mpls(eth_type))
2066 			return -EINVAL;
2067 		break;
2068 
2069 	case OVS_KEY_ATTR_SCTP:
2070 		if ((eth_type != htons(ETH_P_IP) &&
2071 		     eth_type != htons(ETH_P_IPV6)) ||
2072 		    flow_key->ip.proto != IPPROTO_SCTP)
2073 			return -EINVAL;
2074 
2075 		break;
2076 
2077 	default:
2078 		return -EINVAL;
2079 	}
2080 
2081 	/* Convert non-masked non-tunnel set actions to masked set actions. */
2082 	if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2083 		int start, len = key_len * 2;
2084 		struct nlattr *at;
2085 
2086 		*skip_copy = true;
2087 
2088 		start = add_nested_action_start(sfa,
2089 						OVS_ACTION_ATTR_SET_TO_MASKED,
2090 						log);
2091 		if (start < 0)
2092 			return start;
2093 
2094 		at = __add_action(sfa, key_type, NULL, len, log);
2095 		if (IS_ERR(at))
2096 			return PTR_ERR(at);
2097 
2098 		memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2099 		memset(nla_data(at) + key_len, 0xff, key_len);    /* Mask. */
2100 		/* Clear non-writeable bits from otherwise writeable fields. */
2101 		if (key_type == OVS_KEY_ATTR_IPV6) {
2102 			struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2103 
2104 			mask->ipv6_label &= htonl(0x000FFFFF);
2105 		}
2106 		add_nested_action_end(*sfa, start);
2107 	}
2108 
2109 	return 0;
2110 }
2111 
2112 static int validate_userspace(const struct nlattr *attr)
2113 {
2114 	static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2115 		[OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2116 		[OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2117 		[OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2118 	};
2119 	struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2120 	int error;
2121 
2122 	error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
2123 				 attr, userspace_policy);
2124 	if (error)
2125 		return error;
2126 
2127 	if (!a[OVS_USERSPACE_ATTR_PID] ||
2128 	    !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2129 		return -EINVAL;
2130 
2131 	return 0;
2132 }
2133 
2134 static int copy_action(const struct nlattr *from,
2135 		       struct sw_flow_actions **sfa, bool log)
2136 {
2137 	int totlen = NLA_ALIGN(from->nla_len);
2138 	struct nlattr *to;
2139 
2140 	to = reserve_sfa_size(sfa, from->nla_len, log);
2141 	if (IS_ERR(to))
2142 		return PTR_ERR(to);
2143 
2144 	memcpy(to, from, totlen);
2145 	return 0;
2146 }
2147 
2148 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2149 				  const struct sw_flow_key *key,
2150 				  int depth, struct sw_flow_actions **sfa,
2151 				  __be16 eth_type, __be16 vlan_tci, bool log)
2152 {
2153 	const struct nlattr *a;
2154 	int rem, err;
2155 
2156 	if (depth >= SAMPLE_ACTION_DEPTH)
2157 		return -EOVERFLOW;
2158 
2159 	nla_for_each_nested(a, attr, rem) {
2160 		/* Expected argument lengths, (u32)-1 for variable length. */
2161 		static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2162 			[OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2163 			[OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2164 			[OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2165 			[OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2166 			[OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2167 			[OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2168 			[OVS_ACTION_ATTR_POP_VLAN] = 0,
2169 			[OVS_ACTION_ATTR_SET] = (u32)-1,
2170 			[OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2171 			[OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2172 			[OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2173 			[OVS_ACTION_ATTR_CT] = (u32)-1,
2174 		};
2175 		const struct ovs_action_push_vlan *vlan;
2176 		int type = nla_type(a);
2177 		bool skip_copy;
2178 
2179 		if (type > OVS_ACTION_ATTR_MAX ||
2180 		    (action_lens[type] != nla_len(a) &&
2181 		     action_lens[type] != (u32)-1))
2182 			return -EINVAL;
2183 
2184 		skip_copy = false;
2185 		switch (type) {
2186 		case OVS_ACTION_ATTR_UNSPEC:
2187 			return -EINVAL;
2188 
2189 		case OVS_ACTION_ATTR_USERSPACE:
2190 			err = validate_userspace(a);
2191 			if (err)
2192 				return err;
2193 			break;
2194 
2195 		case OVS_ACTION_ATTR_OUTPUT:
2196 			if (nla_get_u32(a) >= DP_MAX_PORTS)
2197 				return -EINVAL;
2198 			break;
2199 
2200 		case OVS_ACTION_ATTR_HASH: {
2201 			const struct ovs_action_hash *act_hash = nla_data(a);
2202 
2203 			switch (act_hash->hash_alg) {
2204 			case OVS_HASH_ALG_L4:
2205 				break;
2206 			default:
2207 				return  -EINVAL;
2208 			}
2209 
2210 			break;
2211 		}
2212 
2213 		case OVS_ACTION_ATTR_POP_VLAN:
2214 			vlan_tci = htons(0);
2215 			break;
2216 
2217 		case OVS_ACTION_ATTR_PUSH_VLAN:
2218 			vlan = nla_data(a);
2219 			if (vlan->vlan_tpid != htons(ETH_P_8021Q))
2220 				return -EINVAL;
2221 			if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2222 				return -EINVAL;
2223 			vlan_tci = vlan->vlan_tci;
2224 			break;
2225 
2226 		case OVS_ACTION_ATTR_RECIRC:
2227 			break;
2228 
2229 		case OVS_ACTION_ATTR_PUSH_MPLS: {
2230 			const struct ovs_action_push_mpls *mpls = nla_data(a);
2231 
2232 			if (!eth_p_mpls(mpls->mpls_ethertype))
2233 				return -EINVAL;
2234 			/* Prohibit push MPLS other than to a white list
2235 			 * for packets that have a known tag order.
2236 			 */
2237 			if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2238 			    (eth_type != htons(ETH_P_IP) &&
2239 			     eth_type != htons(ETH_P_IPV6) &&
2240 			     eth_type != htons(ETH_P_ARP) &&
2241 			     eth_type != htons(ETH_P_RARP) &&
2242 			     !eth_p_mpls(eth_type)))
2243 				return -EINVAL;
2244 			eth_type = mpls->mpls_ethertype;
2245 			break;
2246 		}
2247 
2248 		case OVS_ACTION_ATTR_POP_MPLS:
2249 			if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2250 			    !eth_p_mpls(eth_type))
2251 				return -EINVAL;
2252 
2253 			/* Disallow subsequent L2.5+ set and mpls_pop actions
2254 			 * as there is no check here to ensure that the new
2255 			 * eth_type is valid and thus set actions could
2256 			 * write off the end of the packet or otherwise
2257 			 * corrupt it.
2258 			 *
2259 			 * Support for these actions is planned using packet
2260 			 * recirculation.
2261 			 */
2262 			eth_type = htons(0);
2263 			break;
2264 
2265 		case OVS_ACTION_ATTR_SET:
2266 			err = validate_set(a, key, sfa,
2267 					   &skip_copy, eth_type, false, log);
2268 			if (err)
2269 				return err;
2270 			break;
2271 
2272 		case OVS_ACTION_ATTR_SET_MASKED:
2273 			err = validate_set(a, key, sfa,
2274 					   &skip_copy, eth_type, true, log);
2275 			if (err)
2276 				return err;
2277 			break;
2278 
2279 		case OVS_ACTION_ATTR_SAMPLE:
2280 			err = validate_and_copy_sample(net, a, key, depth, sfa,
2281 						       eth_type, vlan_tci, log);
2282 			if (err)
2283 				return err;
2284 			skip_copy = true;
2285 			break;
2286 
2287 		case OVS_ACTION_ATTR_CT:
2288 			err = ovs_ct_copy_action(net, a, key, sfa, log);
2289 			if (err)
2290 				return err;
2291 			skip_copy = true;
2292 			break;
2293 
2294 		default:
2295 			OVS_NLERR(log, "Unknown Action type %d", type);
2296 			return -EINVAL;
2297 		}
2298 		if (!skip_copy) {
2299 			err = copy_action(a, sfa, log);
2300 			if (err)
2301 				return err;
2302 		}
2303 	}
2304 
2305 	if (rem > 0)
2306 		return -EINVAL;
2307 
2308 	return 0;
2309 }
2310 
2311 /* 'key' must be the masked key. */
2312 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2313 			 const struct sw_flow_key *key,
2314 			 struct sw_flow_actions **sfa, bool log)
2315 {
2316 	int err;
2317 
2318 	*sfa = nla_alloc_flow_actions(nla_len(attr), log);
2319 	if (IS_ERR(*sfa))
2320 		return PTR_ERR(*sfa);
2321 
2322 	(*sfa)->orig_len = nla_len(attr);
2323 	err = __ovs_nla_copy_actions(net, attr, key, 0, sfa, key->eth.type,
2324 				     key->eth.tci, log);
2325 	if (err)
2326 		ovs_nla_free_flow_actions(*sfa);
2327 
2328 	return err;
2329 }
2330 
2331 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
2332 {
2333 	const struct nlattr *a;
2334 	struct nlattr *start;
2335 	int err = 0, rem;
2336 
2337 	start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
2338 	if (!start)
2339 		return -EMSGSIZE;
2340 
2341 	nla_for_each_nested(a, attr, rem) {
2342 		int type = nla_type(a);
2343 		struct nlattr *st_sample;
2344 
2345 		switch (type) {
2346 		case OVS_SAMPLE_ATTR_PROBABILITY:
2347 			if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
2348 				    sizeof(u32), nla_data(a)))
2349 				return -EMSGSIZE;
2350 			break;
2351 		case OVS_SAMPLE_ATTR_ACTIONS:
2352 			st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
2353 			if (!st_sample)
2354 				return -EMSGSIZE;
2355 			err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
2356 			if (err)
2357 				return err;
2358 			nla_nest_end(skb, st_sample);
2359 			break;
2360 		}
2361 	}
2362 
2363 	nla_nest_end(skb, start);
2364 	return err;
2365 }
2366 
2367 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
2368 {
2369 	const struct nlattr *ovs_key = nla_data(a);
2370 	int key_type = nla_type(ovs_key);
2371 	struct nlattr *start;
2372 	int err;
2373 
2374 	switch (key_type) {
2375 	case OVS_KEY_ATTR_TUNNEL_INFO: {
2376 		struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
2377 		struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
2378 
2379 		start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2380 		if (!start)
2381 			return -EMSGSIZE;
2382 
2383 		err = ipv4_tun_to_nlattr(skb, &tun_info->key,
2384 					 tun_info->options_len ?
2385 					     ip_tunnel_info_opts(tun_info) : NULL,
2386 					 tun_info->options_len);
2387 		if (err)
2388 			return err;
2389 		nla_nest_end(skb, start);
2390 		break;
2391 	}
2392 	default:
2393 		if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
2394 			return -EMSGSIZE;
2395 		break;
2396 	}
2397 
2398 	return 0;
2399 }
2400 
2401 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
2402 						struct sk_buff *skb)
2403 {
2404 	const struct nlattr *ovs_key = nla_data(a);
2405 	struct nlattr *nla;
2406 	size_t key_len = nla_len(ovs_key) / 2;
2407 
2408 	/* Revert the conversion we did from a non-masked set action to
2409 	 * masked set action.
2410 	 */
2411 	nla = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2412 	if (!nla)
2413 		return -EMSGSIZE;
2414 
2415 	if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
2416 		return -EMSGSIZE;
2417 
2418 	nla_nest_end(skb, nla);
2419 	return 0;
2420 }
2421 
2422 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
2423 {
2424 	const struct nlattr *a;
2425 	int rem, err;
2426 
2427 	nla_for_each_attr(a, attr, len, rem) {
2428 		int type = nla_type(a);
2429 
2430 		switch (type) {
2431 		case OVS_ACTION_ATTR_SET:
2432 			err = set_action_to_attr(a, skb);
2433 			if (err)
2434 				return err;
2435 			break;
2436 
2437 		case OVS_ACTION_ATTR_SET_TO_MASKED:
2438 			err = masked_set_action_to_set_action_attr(a, skb);
2439 			if (err)
2440 				return err;
2441 			break;
2442 
2443 		case OVS_ACTION_ATTR_SAMPLE:
2444 			err = sample_action_to_attr(a, skb);
2445 			if (err)
2446 				return err;
2447 			break;
2448 
2449 		case OVS_ACTION_ATTR_CT:
2450 			err = ovs_ct_action_to_attr(nla_data(a), skb);
2451 			if (err)
2452 				return err;
2453 			break;
2454 
2455 		default:
2456 			if (nla_put(skb, type, nla_len(a), nla_data(a)))
2457 				return -EMSGSIZE;
2458 			break;
2459 		}
2460 	}
2461 
2462 	return 0;
2463 }
2464