xref: /openbmc/linux/net/openvswitch/flow_netlink.c (revision e983940270f10fe8551baf0098be76ea478294a3)
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_64bit(8) /* OVS_TUNNEL_KEY_ATTR_ID */
265 		+ nla_total_size(16)   /* OVS_TUNNEL_KEY_ATTR_IPV[46]_SRC */
266 		+ nla_total_size(16)   /* OVS_TUNNEL_KEY_ATTR_IPV[46]_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 	[OVS_TUNNEL_KEY_ATTR_IPV6_SRC]      = { .len = sizeof(struct in6_addr) },
327 	[OVS_TUNNEL_KEY_ATTR_IPV6_DST]      = { .len = sizeof(struct in6_addr) },
328 };
329 
330 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute.  */
331 static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
332 	[OVS_KEY_ATTR_ENCAP]	 = { .len = OVS_ATTR_NESTED },
333 	[OVS_KEY_ATTR_PRIORITY]	 = { .len = sizeof(u32) },
334 	[OVS_KEY_ATTR_IN_PORT]	 = { .len = sizeof(u32) },
335 	[OVS_KEY_ATTR_SKB_MARK]	 = { .len = sizeof(u32) },
336 	[OVS_KEY_ATTR_ETHERNET]	 = { .len = sizeof(struct ovs_key_ethernet) },
337 	[OVS_KEY_ATTR_VLAN]	 = { .len = sizeof(__be16) },
338 	[OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
339 	[OVS_KEY_ATTR_IPV4]	 = { .len = sizeof(struct ovs_key_ipv4) },
340 	[OVS_KEY_ATTR_IPV6]	 = { .len = sizeof(struct ovs_key_ipv6) },
341 	[OVS_KEY_ATTR_TCP]	 = { .len = sizeof(struct ovs_key_tcp) },
342 	[OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
343 	[OVS_KEY_ATTR_UDP]	 = { .len = sizeof(struct ovs_key_udp) },
344 	[OVS_KEY_ATTR_SCTP]	 = { .len = sizeof(struct ovs_key_sctp) },
345 	[OVS_KEY_ATTR_ICMP]	 = { .len = sizeof(struct ovs_key_icmp) },
346 	[OVS_KEY_ATTR_ICMPV6]	 = { .len = sizeof(struct ovs_key_icmpv6) },
347 	[OVS_KEY_ATTR_ARP]	 = { .len = sizeof(struct ovs_key_arp) },
348 	[OVS_KEY_ATTR_ND]	 = { .len = sizeof(struct ovs_key_nd) },
349 	[OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
350 	[OVS_KEY_ATTR_DP_HASH]	 = { .len = sizeof(u32) },
351 	[OVS_KEY_ATTR_TUNNEL]	 = { .len = OVS_ATTR_NESTED,
352 				     .next = ovs_tunnel_key_lens, },
353 	[OVS_KEY_ATTR_MPLS]	 = { .len = sizeof(struct ovs_key_mpls) },
354 	[OVS_KEY_ATTR_CT_STATE]	 = { .len = sizeof(u32) },
355 	[OVS_KEY_ATTR_CT_ZONE]	 = { .len = sizeof(u16) },
356 	[OVS_KEY_ATTR_CT_MARK]	 = { .len = sizeof(u32) },
357 	[OVS_KEY_ATTR_CT_LABELS] = { .len = sizeof(struct ovs_key_ct_labels) },
358 };
359 
360 static bool check_attr_len(unsigned int attr_len, unsigned int expected_len)
361 {
362 	return expected_len == attr_len ||
363 	       expected_len == OVS_ATTR_NESTED ||
364 	       expected_len == OVS_ATTR_VARIABLE;
365 }
366 
367 static bool is_all_zero(const u8 *fp, size_t size)
368 {
369 	int i;
370 
371 	if (!fp)
372 		return false;
373 
374 	for (i = 0; i < size; i++)
375 		if (fp[i])
376 			return false;
377 
378 	return true;
379 }
380 
381 static int __parse_flow_nlattrs(const struct nlattr *attr,
382 				const struct nlattr *a[],
383 				u64 *attrsp, bool log, bool nz)
384 {
385 	const struct nlattr *nla;
386 	u64 attrs;
387 	int rem;
388 
389 	attrs = *attrsp;
390 	nla_for_each_nested(nla, attr, rem) {
391 		u16 type = nla_type(nla);
392 		int expected_len;
393 
394 		if (type > OVS_KEY_ATTR_MAX) {
395 			OVS_NLERR(log, "Key type %d is out of range max %d",
396 				  type, OVS_KEY_ATTR_MAX);
397 			return -EINVAL;
398 		}
399 
400 		if (attrs & (1 << type)) {
401 			OVS_NLERR(log, "Duplicate key (type %d).", type);
402 			return -EINVAL;
403 		}
404 
405 		expected_len = ovs_key_lens[type].len;
406 		if (!check_attr_len(nla_len(nla), expected_len)) {
407 			OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
408 				  type, nla_len(nla), expected_len);
409 			return -EINVAL;
410 		}
411 
412 		if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
413 			attrs |= 1 << type;
414 			a[type] = nla;
415 		}
416 	}
417 	if (rem) {
418 		OVS_NLERR(log, "Message has %d unknown bytes.", rem);
419 		return -EINVAL;
420 	}
421 
422 	*attrsp = attrs;
423 	return 0;
424 }
425 
426 static int parse_flow_mask_nlattrs(const struct nlattr *attr,
427 				   const struct nlattr *a[], u64 *attrsp,
428 				   bool log)
429 {
430 	return __parse_flow_nlattrs(attr, a, attrsp, log, true);
431 }
432 
433 static int parse_flow_nlattrs(const struct nlattr *attr,
434 			      const struct nlattr *a[], u64 *attrsp,
435 			      bool log)
436 {
437 	return __parse_flow_nlattrs(attr, a, attrsp, log, false);
438 }
439 
440 static int genev_tun_opt_from_nlattr(const struct nlattr *a,
441 				     struct sw_flow_match *match, bool is_mask,
442 				     bool log)
443 {
444 	unsigned long opt_key_offset;
445 
446 	if (nla_len(a) > sizeof(match->key->tun_opts)) {
447 		OVS_NLERR(log, "Geneve option length err (len %d, max %zu).",
448 			  nla_len(a), sizeof(match->key->tun_opts));
449 		return -EINVAL;
450 	}
451 
452 	if (nla_len(a) % 4 != 0) {
453 		OVS_NLERR(log, "Geneve opt len %d is not a multiple of 4.",
454 			  nla_len(a));
455 		return -EINVAL;
456 	}
457 
458 	/* We need to record the length of the options passed
459 	 * down, otherwise packets with the same format but
460 	 * additional options will be silently matched.
461 	 */
462 	if (!is_mask) {
463 		SW_FLOW_KEY_PUT(match, tun_opts_len, nla_len(a),
464 				false);
465 	} else {
466 		/* This is somewhat unusual because it looks at
467 		 * both the key and mask while parsing the
468 		 * attributes (and by extension assumes the key
469 		 * is parsed first). Normally, we would verify
470 		 * that each is the correct length and that the
471 		 * attributes line up in the validate function.
472 		 * However, that is difficult because this is
473 		 * variable length and we won't have the
474 		 * information later.
475 		 */
476 		if (match->key->tun_opts_len != nla_len(a)) {
477 			OVS_NLERR(log, "Geneve option len %d != mask len %d",
478 				  match->key->tun_opts_len, nla_len(a));
479 			return -EINVAL;
480 		}
481 
482 		SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
483 	}
484 
485 	opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
486 	SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
487 				  nla_len(a), is_mask);
488 	return 0;
489 }
490 
491 static int vxlan_tun_opt_from_nlattr(const struct nlattr *attr,
492 				     struct sw_flow_match *match, bool is_mask,
493 				     bool log)
494 {
495 	struct nlattr *a;
496 	int rem;
497 	unsigned long opt_key_offset;
498 	struct vxlan_metadata opts;
499 
500 	BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
501 
502 	memset(&opts, 0, sizeof(opts));
503 	nla_for_each_nested(a, attr, rem) {
504 		int type = nla_type(a);
505 
506 		if (type > OVS_VXLAN_EXT_MAX) {
507 			OVS_NLERR(log, "VXLAN extension %d out of range max %d",
508 				  type, OVS_VXLAN_EXT_MAX);
509 			return -EINVAL;
510 		}
511 
512 		if (!check_attr_len(nla_len(a),
513 				    ovs_vxlan_ext_key_lens[type].len)) {
514 			OVS_NLERR(log, "VXLAN extension %d has unexpected len %d expected %d",
515 				  type, nla_len(a),
516 				  ovs_vxlan_ext_key_lens[type].len);
517 			return -EINVAL;
518 		}
519 
520 		switch (type) {
521 		case OVS_VXLAN_EXT_GBP:
522 			opts.gbp = nla_get_u32(a);
523 			break;
524 		default:
525 			OVS_NLERR(log, "Unknown VXLAN extension attribute %d",
526 				  type);
527 			return -EINVAL;
528 		}
529 	}
530 	if (rem) {
531 		OVS_NLERR(log, "VXLAN extension message has %d unknown bytes.",
532 			  rem);
533 		return -EINVAL;
534 	}
535 
536 	if (!is_mask)
537 		SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
538 	else
539 		SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
540 
541 	opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
542 	SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
543 				  is_mask);
544 	return 0;
545 }
546 
547 static int ip_tun_from_nlattr(const struct nlattr *attr,
548 			      struct sw_flow_match *match, bool is_mask,
549 			      bool log)
550 {
551 	bool ttl = false, ipv4 = false, ipv6 = false;
552 	__be16 tun_flags = 0;
553 	int opts_type = 0;
554 	struct nlattr *a;
555 	int rem;
556 
557 	nla_for_each_nested(a, attr, rem) {
558 		int type = nla_type(a);
559 		int err;
560 
561 		if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
562 			OVS_NLERR(log, "Tunnel attr %d out of range max %d",
563 				  type, OVS_TUNNEL_KEY_ATTR_MAX);
564 			return -EINVAL;
565 		}
566 
567 		if (!check_attr_len(nla_len(a),
568 				    ovs_tunnel_key_lens[type].len)) {
569 			OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
570 				  type, nla_len(a), ovs_tunnel_key_lens[type].len);
571 			return -EINVAL;
572 		}
573 
574 		switch (type) {
575 		case OVS_TUNNEL_KEY_ATTR_ID:
576 			SW_FLOW_KEY_PUT(match, tun_key.tun_id,
577 					nla_get_be64(a), is_mask);
578 			tun_flags |= TUNNEL_KEY;
579 			break;
580 		case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
581 			SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.src,
582 					nla_get_in_addr(a), is_mask);
583 			ipv4 = true;
584 			break;
585 		case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
586 			SW_FLOW_KEY_PUT(match, tun_key.u.ipv4.dst,
587 					nla_get_in_addr(a), is_mask);
588 			ipv4 = true;
589 			break;
590 		case OVS_TUNNEL_KEY_ATTR_IPV6_SRC:
591 			SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
592 					nla_get_in6_addr(a), is_mask);
593 			ipv6 = true;
594 			break;
595 		case OVS_TUNNEL_KEY_ATTR_IPV6_DST:
596 			SW_FLOW_KEY_PUT(match, tun_key.u.ipv6.dst,
597 					nla_get_in6_addr(a), is_mask);
598 			ipv6 = true;
599 			break;
600 		case OVS_TUNNEL_KEY_ATTR_TOS:
601 			SW_FLOW_KEY_PUT(match, tun_key.tos,
602 					nla_get_u8(a), is_mask);
603 			break;
604 		case OVS_TUNNEL_KEY_ATTR_TTL:
605 			SW_FLOW_KEY_PUT(match, tun_key.ttl,
606 					nla_get_u8(a), is_mask);
607 			ttl = true;
608 			break;
609 		case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
610 			tun_flags |= TUNNEL_DONT_FRAGMENT;
611 			break;
612 		case OVS_TUNNEL_KEY_ATTR_CSUM:
613 			tun_flags |= TUNNEL_CSUM;
614 			break;
615 		case OVS_TUNNEL_KEY_ATTR_TP_SRC:
616 			SW_FLOW_KEY_PUT(match, tun_key.tp_src,
617 					nla_get_be16(a), is_mask);
618 			break;
619 		case OVS_TUNNEL_KEY_ATTR_TP_DST:
620 			SW_FLOW_KEY_PUT(match, tun_key.tp_dst,
621 					nla_get_be16(a), is_mask);
622 			break;
623 		case OVS_TUNNEL_KEY_ATTR_OAM:
624 			tun_flags |= TUNNEL_OAM;
625 			break;
626 		case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
627 			if (opts_type) {
628 				OVS_NLERR(log, "Multiple metadata blocks provided");
629 				return -EINVAL;
630 			}
631 
632 			err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
633 			if (err)
634 				return err;
635 
636 			tun_flags |= TUNNEL_GENEVE_OPT;
637 			opts_type = type;
638 			break;
639 		case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
640 			if (opts_type) {
641 				OVS_NLERR(log, "Multiple metadata blocks provided");
642 				return -EINVAL;
643 			}
644 
645 			err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
646 			if (err)
647 				return err;
648 
649 			tun_flags |= TUNNEL_VXLAN_OPT;
650 			opts_type = type;
651 			break;
652 		default:
653 			OVS_NLERR(log, "Unknown IP tunnel attribute %d",
654 				  type);
655 			return -EINVAL;
656 		}
657 	}
658 
659 	SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
660 	if (is_mask)
661 		SW_FLOW_KEY_MEMSET_FIELD(match, tun_proto, 0xff, true);
662 	else
663 		SW_FLOW_KEY_PUT(match, tun_proto, ipv6 ? AF_INET6 : AF_INET,
664 				false);
665 
666 	if (rem > 0) {
667 		OVS_NLERR(log, "IP tunnel attribute has %d unknown bytes.",
668 			  rem);
669 		return -EINVAL;
670 	}
671 
672 	if (ipv4 && ipv6) {
673 		OVS_NLERR(log, "Mixed IPv4 and IPv6 tunnel attributes");
674 		return -EINVAL;
675 	}
676 
677 	if (!is_mask) {
678 		if (!ipv4 && !ipv6) {
679 			OVS_NLERR(log, "IP tunnel dst address not specified");
680 			return -EINVAL;
681 		}
682 		if (ipv4 && !match->key->tun_key.u.ipv4.dst) {
683 			OVS_NLERR(log, "IPv4 tunnel dst address is zero");
684 			return -EINVAL;
685 		}
686 		if (ipv6 && ipv6_addr_any(&match->key->tun_key.u.ipv6.dst)) {
687 			OVS_NLERR(log, "IPv6 tunnel dst address is zero");
688 			return -EINVAL;
689 		}
690 
691 		if (!ttl) {
692 			OVS_NLERR(log, "IP tunnel TTL not specified.");
693 			return -EINVAL;
694 		}
695 	}
696 
697 	return opts_type;
698 }
699 
700 static int vxlan_opt_to_nlattr(struct sk_buff *skb,
701 			       const void *tun_opts, int swkey_tun_opts_len)
702 {
703 	const struct vxlan_metadata *opts = tun_opts;
704 	struct nlattr *nla;
705 
706 	nla = nla_nest_start(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
707 	if (!nla)
708 		return -EMSGSIZE;
709 
710 	if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
711 		return -EMSGSIZE;
712 
713 	nla_nest_end(skb, nla);
714 	return 0;
715 }
716 
717 static int __ip_tun_to_nlattr(struct sk_buff *skb,
718 			      const struct ip_tunnel_key *output,
719 			      const void *tun_opts, int swkey_tun_opts_len,
720 			      unsigned short tun_proto)
721 {
722 	if (output->tun_flags & TUNNEL_KEY &&
723 	    nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id,
724 			 OVS_TUNNEL_KEY_ATTR_PAD))
725 		return -EMSGSIZE;
726 	switch (tun_proto) {
727 	case AF_INET:
728 		if (output->u.ipv4.src &&
729 		    nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC,
730 				    output->u.ipv4.src))
731 			return -EMSGSIZE;
732 		if (output->u.ipv4.dst &&
733 		    nla_put_in_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST,
734 				    output->u.ipv4.dst))
735 			return -EMSGSIZE;
736 		break;
737 	case AF_INET6:
738 		if (!ipv6_addr_any(&output->u.ipv6.src) &&
739 		    nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_SRC,
740 				     &output->u.ipv6.src))
741 			return -EMSGSIZE;
742 		if (!ipv6_addr_any(&output->u.ipv6.dst) &&
743 		    nla_put_in6_addr(skb, OVS_TUNNEL_KEY_ATTR_IPV6_DST,
744 				     &output->u.ipv6.dst))
745 			return -EMSGSIZE;
746 		break;
747 	}
748 	if (output->tos &&
749 	    nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->tos))
750 		return -EMSGSIZE;
751 	if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ttl))
752 		return -EMSGSIZE;
753 	if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
754 	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
755 		return -EMSGSIZE;
756 	if ((output->tun_flags & TUNNEL_CSUM) &&
757 	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
758 		return -EMSGSIZE;
759 	if (output->tp_src &&
760 	    nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_SRC, output->tp_src))
761 		return -EMSGSIZE;
762 	if (output->tp_dst &&
763 	    nla_put_be16(skb, OVS_TUNNEL_KEY_ATTR_TP_DST, output->tp_dst))
764 		return -EMSGSIZE;
765 	if ((output->tun_flags & TUNNEL_OAM) &&
766 	    nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
767 		return -EMSGSIZE;
768 	if (swkey_tun_opts_len) {
769 		if (output->tun_flags & TUNNEL_GENEVE_OPT &&
770 		    nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
771 			    swkey_tun_opts_len, tun_opts))
772 			return -EMSGSIZE;
773 		else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
774 			 vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
775 			return -EMSGSIZE;
776 	}
777 
778 	return 0;
779 }
780 
781 static int ip_tun_to_nlattr(struct sk_buff *skb,
782 			    const struct ip_tunnel_key *output,
783 			    const void *tun_opts, int swkey_tun_opts_len,
784 			    unsigned short tun_proto)
785 {
786 	struct nlattr *nla;
787 	int err;
788 
789 	nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
790 	if (!nla)
791 		return -EMSGSIZE;
792 
793 	err = __ip_tun_to_nlattr(skb, output, tun_opts, swkey_tun_opts_len,
794 				 tun_proto);
795 	if (err)
796 		return err;
797 
798 	nla_nest_end(skb, nla);
799 	return 0;
800 }
801 
802 int ovs_nla_put_tunnel_info(struct sk_buff *skb,
803 			    struct ip_tunnel_info *tun_info)
804 {
805 	return __ip_tun_to_nlattr(skb, &tun_info->key,
806 				  ip_tunnel_info_opts(tun_info),
807 				  tun_info->options_len,
808 				  ip_tunnel_info_af(tun_info));
809 }
810 
811 static int encode_vlan_from_nlattrs(struct sw_flow_match *match,
812 				    const struct nlattr *a[],
813 				    bool is_mask, bool inner)
814 {
815 	__be16 tci = 0;
816 	__be16 tpid = 0;
817 
818 	if (a[OVS_KEY_ATTR_VLAN])
819 		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
820 
821 	if (a[OVS_KEY_ATTR_ETHERTYPE])
822 		tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
823 
824 	if (likely(!inner)) {
825 		SW_FLOW_KEY_PUT(match, eth.vlan.tpid, tpid, is_mask);
826 		SW_FLOW_KEY_PUT(match, eth.vlan.tci, tci, is_mask);
827 	} else {
828 		SW_FLOW_KEY_PUT(match, eth.cvlan.tpid, tpid, is_mask);
829 		SW_FLOW_KEY_PUT(match, eth.cvlan.tci, tci, is_mask);
830 	}
831 	return 0;
832 }
833 
834 static int validate_vlan_from_nlattrs(const struct sw_flow_match *match,
835 				      u64 key_attrs, bool inner,
836 				      const struct nlattr **a, bool log)
837 {
838 	__be16 tci = 0;
839 
840 	if (!((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
841 	      (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
842 	       eth_type_vlan(nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE])))) {
843 		/* Not a VLAN. */
844 		return 0;
845 	}
846 
847 	if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
848 	      (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
849 		OVS_NLERR(log, "Invalid %s frame", (inner) ? "C-VLAN" : "VLAN");
850 		return -EINVAL;
851 	}
852 
853 	if (a[OVS_KEY_ATTR_VLAN])
854 		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
855 
856 	if (!(tci & htons(VLAN_TAG_PRESENT))) {
857 		if (tci) {
858 			OVS_NLERR(log, "%s TCI does not have VLAN_TAG_PRESENT bit set.",
859 				  (inner) ? "C-VLAN" : "VLAN");
860 			return -EINVAL;
861 		} else if (nla_len(a[OVS_KEY_ATTR_ENCAP])) {
862 			/* Corner case for truncated VLAN header. */
863 			OVS_NLERR(log, "Truncated %s header has non-zero encap attribute.",
864 				  (inner) ? "C-VLAN" : "VLAN");
865 			return -EINVAL;
866 		}
867 	}
868 
869 	return 1;
870 }
871 
872 static int validate_vlan_mask_from_nlattrs(const struct sw_flow_match *match,
873 					   u64 key_attrs, bool inner,
874 					   const struct nlattr **a, bool log)
875 {
876 	__be16 tci = 0;
877 	__be16 tpid = 0;
878 	bool encap_valid = !!(match->key->eth.vlan.tci &
879 			      htons(VLAN_TAG_PRESENT));
880 	bool i_encap_valid = !!(match->key->eth.cvlan.tci &
881 				htons(VLAN_TAG_PRESENT));
882 
883 	if (!(key_attrs & (1 << OVS_KEY_ATTR_ENCAP))) {
884 		/* Not a VLAN. */
885 		return 0;
886 	}
887 
888 	if ((!inner && !encap_valid) || (inner && !i_encap_valid)) {
889 		OVS_NLERR(log, "Encap mask attribute is set for non-%s frame.",
890 			  (inner) ? "C-VLAN" : "VLAN");
891 		return -EINVAL;
892 	}
893 
894 	if (a[OVS_KEY_ATTR_VLAN])
895 		tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
896 
897 	if (a[OVS_KEY_ATTR_ETHERTYPE])
898 		tpid = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
899 
900 	if (tpid != htons(0xffff)) {
901 		OVS_NLERR(log, "Must have an exact match on %s TPID (mask=%x).",
902 			  (inner) ? "C-VLAN" : "VLAN", ntohs(tpid));
903 		return -EINVAL;
904 	}
905 	if (!(tci & htons(VLAN_TAG_PRESENT))) {
906 		OVS_NLERR(log, "%s TCI mask does not have exact match for VLAN_TAG_PRESENT bit.",
907 			  (inner) ? "C-VLAN" : "VLAN");
908 		return -EINVAL;
909 	}
910 
911 	return 1;
912 }
913 
914 static int __parse_vlan_from_nlattrs(struct sw_flow_match *match,
915 				     u64 *key_attrs, bool inner,
916 				     const struct nlattr **a, bool is_mask,
917 				     bool log)
918 {
919 	int err;
920 	const struct nlattr *encap;
921 
922 	if (!is_mask)
923 		err = validate_vlan_from_nlattrs(match, *key_attrs, inner,
924 						 a, log);
925 	else
926 		err = validate_vlan_mask_from_nlattrs(match, *key_attrs, inner,
927 						      a, log);
928 	if (err <= 0)
929 		return err;
930 
931 	err = encode_vlan_from_nlattrs(match, a, is_mask, inner);
932 	if (err)
933 		return err;
934 
935 	*key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
936 	*key_attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
937 	*key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
938 
939 	encap = a[OVS_KEY_ATTR_ENCAP];
940 
941 	if (!is_mask)
942 		err = parse_flow_nlattrs(encap, a, key_attrs, log);
943 	else
944 		err = parse_flow_mask_nlattrs(encap, a, key_attrs, log);
945 
946 	return err;
947 }
948 
949 static int parse_vlan_from_nlattrs(struct sw_flow_match *match,
950 				   u64 *key_attrs, const struct nlattr **a,
951 				   bool is_mask, bool log)
952 {
953 	int err;
954 	bool encap_valid = false;
955 
956 	err = __parse_vlan_from_nlattrs(match, key_attrs, false, a,
957 					is_mask, log);
958 	if (err)
959 		return err;
960 
961 	encap_valid = !!(match->key->eth.vlan.tci & htons(VLAN_TAG_PRESENT));
962 	if (encap_valid) {
963 		err = __parse_vlan_from_nlattrs(match, key_attrs, true, a,
964 						is_mask, log);
965 		if (err)
966 			return err;
967 	}
968 
969 	return 0;
970 }
971 
972 static int metadata_from_nlattrs(struct net *net, struct sw_flow_match *match,
973 				 u64 *attrs, const struct nlattr **a,
974 				 bool is_mask, bool log)
975 {
976 	if (*attrs & (1 << OVS_KEY_ATTR_DP_HASH)) {
977 		u32 hash_val = nla_get_u32(a[OVS_KEY_ATTR_DP_HASH]);
978 
979 		SW_FLOW_KEY_PUT(match, ovs_flow_hash, hash_val, is_mask);
980 		*attrs &= ~(1 << OVS_KEY_ATTR_DP_HASH);
981 	}
982 
983 	if (*attrs & (1 << OVS_KEY_ATTR_RECIRC_ID)) {
984 		u32 recirc_id = nla_get_u32(a[OVS_KEY_ATTR_RECIRC_ID]);
985 
986 		SW_FLOW_KEY_PUT(match, recirc_id, recirc_id, is_mask);
987 		*attrs &= ~(1 << OVS_KEY_ATTR_RECIRC_ID);
988 	}
989 
990 	if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
991 		SW_FLOW_KEY_PUT(match, phy.priority,
992 			  nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
993 		*attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
994 	}
995 
996 	if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
997 		u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
998 
999 		if (is_mask) {
1000 			in_port = 0xffffffff; /* Always exact match in_port. */
1001 		} else if (in_port >= DP_MAX_PORTS) {
1002 			OVS_NLERR(log, "Port %d exceeds max allowable %d",
1003 				  in_port, DP_MAX_PORTS);
1004 			return -EINVAL;
1005 		}
1006 
1007 		SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
1008 		*attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1009 	} else if (!is_mask) {
1010 		SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
1011 	}
1012 
1013 	if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1014 		uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1015 
1016 		SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1017 		*attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1018 	}
1019 	if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1020 		if (ip_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1021 				       is_mask, log) < 0)
1022 			return -EINVAL;
1023 		*attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1024 	}
1025 
1026 	if (*attrs & (1 << OVS_KEY_ATTR_CT_STATE) &&
1027 	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_STATE)) {
1028 		u32 ct_state = nla_get_u32(a[OVS_KEY_ATTR_CT_STATE]);
1029 
1030 		if (ct_state & ~CT_SUPPORTED_MASK) {
1031 			OVS_NLERR(log, "ct_state flags %08x unsupported",
1032 				  ct_state);
1033 			return -EINVAL;
1034 		}
1035 
1036 		SW_FLOW_KEY_PUT(match, ct.state, ct_state, is_mask);
1037 		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_STATE);
1038 	}
1039 	if (*attrs & (1 << OVS_KEY_ATTR_CT_ZONE) &&
1040 	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_ZONE)) {
1041 		u16 ct_zone = nla_get_u16(a[OVS_KEY_ATTR_CT_ZONE]);
1042 
1043 		SW_FLOW_KEY_PUT(match, ct.zone, ct_zone, is_mask);
1044 		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_ZONE);
1045 	}
1046 	if (*attrs & (1 << OVS_KEY_ATTR_CT_MARK) &&
1047 	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_MARK)) {
1048 		u32 mark = nla_get_u32(a[OVS_KEY_ATTR_CT_MARK]);
1049 
1050 		SW_FLOW_KEY_PUT(match, ct.mark, mark, is_mask);
1051 		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_MARK);
1052 	}
1053 	if (*attrs & (1 << OVS_KEY_ATTR_CT_LABELS) &&
1054 	    ovs_ct_verify(net, OVS_KEY_ATTR_CT_LABELS)) {
1055 		const struct ovs_key_ct_labels *cl;
1056 
1057 		cl = nla_data(a[OVS_KEY_ATTR_CT_LABELS]);
1058 		SW_FLOW_KEY_MEMCPY(match, ct.labels, cl->ct_labels,
1059 				   sizeof(*cl), is_mask);
1060 		*attrs &= ~(1ULL << OVS_KEY_ATTR_CT_LABELS);
1061 	}
1062 	return 0;
1063 }
1064 
1065 static int ovs_key_from_nlattrs(struct net *net, struct sw_flow_match *match,
1066 				u64 attrs, const struct nlattr **a,
1067 				bool is_mask, bool log)
1068 {
1069 	int err;
1070 
1071 	err = metadata_from_nlattrs(net, match, &attrs, a, is_mask, log);
1072 	if (err)
1073 		return err;
1074 
1075 	if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1076 		const struct ovs_key_ethernet *eth_key;
1077 
1078 		eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1079 		SW_FLOW_KEY_MEMCPY(match, eth.src,
1080 				eth_key->eth_src, ETH_ALEN, is_mask);
1081 		SW_FLOW_KEY_MEMCPY(match, eth.dst,
1082 				eth_key->eth_dst, ETH_ALEN, is_mask);
1083 		attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1084 	}
1085 
1086 	if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1087 		/* VLAN attribute is always parsed before getting here since it
1088 		 * may occur multiple times.
1089 		 */
1090 		OVS_NLERR(log, "VLAN attribute unexpected.");
1091 		return -EINVAL;
1092 	}
1093 
1094 	if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1095 		__be16 eth_type;
1096 
1097 		eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1098 		if (is_mask) {
1099 			/* Always exact match EtherType. */
1100 			eth_type = htons(0xffff);
1101 		} else if (!eth_proto_is_802_3(eth_type)) {
1102 			OVS_NLERR(log, "EtherType %x is less than min %x",
1103 				  ntohs(eth_type), ETH_P_802_3_MIN);
1104 			return -EINVAL;
1105 		}
1106 
1107 		SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
1108 		attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1109 	} else if (!is_mask) {
1110 		SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1111 	}
1112 
1113 	if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1114 		const struct ovs_key_ipv4 *ipv4_key;
1115 
1116 		ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1117 		if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
1118 			OVS_NLERR(log, "IPv4 frag type %d is out of range max %d",
1119 				  ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
1120 			return -EINVAL;
1121 		}
1122 		SW_FLOW_KEY_PUT(match, ip.proto,
1123 				ipv4_key->ipv4_proto, is_mask);
1124 		SW_FLOW_KEY_PUT(match, ip.tos,
1125 				ipv4_key->ipv4_tos, is_mask);
1126 		SW_FLOW_KEY_PUT(match, ip.ttl,
1127 				ipv4_key->ipv4_ttl, is_mask);
1128 		SW_FLOW_KEY_PUT(match, ip.frag,
1129 				ipv4_key->ipv4_frag, is_mask);
1130 		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1131 				ipv4_key->ipv4_src, is_mask);
1132 		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1133 				ipv4_key->ipv4_dst, is_mask);
1134 		attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1135 	}
1136 
1137 	if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1138 		const struct ovs_key_ipv6 *ipv6_key;
1139 
1140 		ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1141 		if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
1142 			OVS_NLERR(log, "IPv6 frag type %d is out of range max %d",
1143 				  ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
1144 			return -EINVAL;
1145 		}
1146 
1147 		if (!is_mask && ipv6_key->ipv6_label & htonl(0xFFF00000)) {
1148 			OVS_NLERR(log, "IPv6 flow label %x is out of range (max=%x).\n",
1149 				  ntohl(ipv6_key->ipv6_label), (1 << 20) - 1);
1150 			return -EINVAL;
1151 		}
1152 
1153 		SW_FLOW_KEY_PUT(match, ipv6.label,
1154 				ipv6_key->ipv6_label, is_mask);
1155 		SW_FLOW_KEY_PUT(match, ip.proto,
1156 				ipv6_key->ipv6_proto, is_mask);
1157 		SW_FLOW_KEY_PUT(match, ip.tos,
1158 				ipv6_key->ipv6_tclass, is_mask);
1159 		SW_FLOW_KEY_PUT(match, ip.ttl,
1160 				ipv6_key->ipv6_hlimit, is_mask);
1161 		SW_FLOW_KEY_PUT(match, ip.frag,
1162 				ipv6_key->ipv6_frag, is_mask);
1163 		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1164 				ipv6_key->ipv6_src,
1165 				sizeof(match->key->ipv6.addr.src),
1166 				is_mask);
1167 		SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1168 				ipv6_key->ipv6_dst,
1169 				sizeof(match->key->ipv6.addr.dst),
1170 				is_mask);
1171 
1172 		attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1173 	}
1174 
1175 	if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1176 		const struct ovs_key_arp *arp_key;
1177 
1178 		arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1179 		if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1180 			OVS_NLERR(log, "Unknown ARP opcode (opcode=%d).",
1181 				  arp_key->arp_op);
1182 			return -EINVAL;
1183 		}
1184 
1185 		SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1186 				arp_key->arp_sip, is_mask);
1187 		SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1188 			arp_key->arp_tip, is_mask);
1189 		SW_FLOW_KEY_PUT(match, ip.proto,
1190 				ntohs(arp_key->arp_op), is_mask);
1191 		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1192 				arp_key->arp_sha, ETH_ALEN, is_mask);
1193 		SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1194 				arp_key->arp_tha, ETH_ALEN, is_mask);
1195 
1196 		attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1197 	}
1198 
1199 	if (attrs & (1 << OVS_KEY_ATTR_MPLS)) {
1200 		const struct ovs_key_mpls *mpls_key;
1201 
1202 		mpls_key = nla_data(a[OVS_KEY_ATTR_MPLS]);
1203 		SW_FLOW_KEY_PUT(match, mpls.top_lse,
1204 				mpls_key->mpls_lse, is_mask);
1205 
1206 		attrs &= ~(1 << OVS_KEY_ATTR_MPLS);
1207 	 }
1208 
1209 	if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1210 		const struct ovs_key_tcp *tcp_key;
1211 
1212 		tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1213 		SW_FLOW_KEY_PUT(match, tp.src, tcp_key->tcp_src, is_mask);
1214 		SW_FLOW_KEY_PUT(match, tp.dst, tcp_key->tcp_dst, is_mask);
1215 		attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1216 	}
1217 
1218 	if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
1219 		SW_FLOW_KEY_PUT(match, tp.flags,
1220 				nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
1221 				is_mask);
1222 		attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
1223 	}
1224 
1225 	if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1226 		const struct ovs_key_udp *udp_key;
1227 
1228 		udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1229 		SW_FLOW_KEY_PUT(match, tp.src, udp_key->udp_src, is_mask);
1230 		SW_FLOW_KEY_PUT(match, tp.dst, udp_key->udp_dst, is_mask);
1231 		attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1232 	}
1233 
1234 	if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1235 		const struct ovs_key_sctp *sctp_key;
1236 
1237 		sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1238 		SW_FLOW_KEY_PUT(match, tp.src, sctp_key->sctp_src, is_mask);
1239 		SW_FLOW_KEY_PUT(match, tp.dst, sctp_key->sctp_dst, is_mask);
1240 		attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1241 	}
1242 
1243 	if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1244 		const struct ovs_key_icmp *icmp_key;
1245 
1246 		icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1247 		SW_FLOW_KEY_PUT(match, tp.src,
1248 				htons(icmp_key->icmp_type), is_mask);
1249 		SW_FLOW_KEY_PUT(match, tp.dst,
1250 				htons(icmp_key->icmp_code), is_mask);
1251 		attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1252 	}
1253 
1254 	if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1255 		const struct ovs_key_icmpv6 *icmpv6_key;
1256 
1257 		icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1258 		SW_FLOW_KEY_PUT(match, tp.src,
1259 				htons(icmpv6_key->icmpv6_type), is_mask);
1260 		SW_FLOW_KEY_PUT(match, tp.dst,
1261 				htons(icmpv6_key->icmpv6_code), is_mask);
1262 		attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1263 	}
1264 
1265 	if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1266 		const struct ovs_key_nd *nd_key;
1267 
1268 		nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1269 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1270 			nd_key->nd_target,
1271 			sizeof(match->key->ipv6.nd.target),
1272 			is_mask);
1273 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1274 			nd_key->nd_sll, ETH_ALEN, is_mask);
1275 		SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1276 				nd_key->nd_tll, ETH_ALEN, is_mask);
1277 		attrs &= ~(1 << OVS_KEY_ATTR_ND);
1278 	}
1279 
1280 	if (attrs != 0) {
1281 		OVS_NLERR(log, "Unknown key attributes %llx",
1282 			  (unsigned long long)attrs);
1283 		return -EINVAL;
1284 	}
1285 
1286 	return 0;
1287 }
1288 
1289 static void nlattr_set(struct nlattr *attr, u8 val,
1290 		       const struct ovs_len_tbl *tbl)
1291 {
1292 	struct nlattr *nla;
1293 	int rem;
1294 
1295 	/* The nlattr stream should already have been validated */
1296 	nla_for_each_nested(nla, attr, rem) {
1297 		if (tbl[nla_type(nla)].len == OVS_ATTR_NESTED) {
1298 			if (tbl[nla_type(nla)].next)
1299 				tbl = tbl[nla_type(nla)].next;
1300 			nlattr_set(nla, val, tbl);
1301 		} else {
1302 			memset(nla_data(nla), val, nla_len(nla));
1303 		}
1304 
1305 		if (nla_type(nla) == OVS_KEY_ATTR_CT_STATE)
1306 			*(u32 *)nla_data(nla) &= CT_SUPPORTED_MASK;
1307 	}
1308 }
1309 
1310 static void mask_set_nlattr(struct nlattr *attr, u8 val)
1311 {
1312 	nlattr_set(attr, val, ovs_key_lens);
1313 }
1314 
1315 /**
1316  * ovs_nla_get_match - parses Netlink attributes into a flow key and
1317  * mask. In case the 'mask' is NULL, the flow is treated as exact match
1318  * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1319  * does not include any don't care bit.
1320  * @net: Used to determine per-namespace field support.
1321  * @match: receives the extracted flow match information.
1322  * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1323  * sequence. The fields should of the packet that triggered the creation
1324  * of this flow.
1325  * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1326  * attribute specifies the mask field of the wildcarded flow.
1327  * @log: Boolean to allow kernel error logging.  Normally true, but when
1328  * probing for feature compatibility this should be passed in as false to
1329  * suppress unnecessary error logging.
1330  */
1331 int ovs_nla_get_match(struct net *net, struct sw_flow_match *match,
1332 		      const struct nlattr *nla_key,
1333 		      const struct nlattr *nla_mask,
1334 		      bool log)
1335 {
1336 	const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1337 	struct nlattr *newmask = NULL;
1338 	u64 key_attrs = 0;
1339 	u64 mask_attrs = 0;
1340 	int err;
1341 
1342 	err = parse_flow_nlattrs(nla_key, a, &key_attrs, log);
1343 	if (err)
1344 		return err;
1345 
1346 	err = parse_vlan_from_nlattrs(match, &key_attrs, a, false, log);
1347 	if (err)
1348 		return err;
1349 
1350 	err = ovs_key_from_nlattrs(net, match, key_attrs, a, false, log);
1351 	if (err)
1352 		return err;
1353 
1354 	if (match->mask) {
1355 		if (!nla_mask) {
1356 			/* Create an exact match mask. We need to set to 0xff
1357 			 * all the 'match->mask' fields that have been touched
1358 			 * in 'match->key'. We cannot simply memset
1359 			 * 'match->mask', because padding bytes and fields not
1360 			 * specified in 'match->key' should be left to 0.
1361 			 * Instead, we use a stream of netlink attributes,
1362 			 * copied from 'key' and set to 0xff.
1363 			 * ovs_key_from_nlattrs() will take care of filling
1364 			 * 'match->mask' appropriately.
1365 			 */
1366 			newmask = kmemdup(nla_key,
1367 					  nla_total_size(nla_len(nla_key)),
1368 					  GFP_KERNEL);
1369 			if (!newmask)
1370 				return -ENOMEM;
1371 
1372 			mask_set_nlattr(newmask, 0xff);
1373 
1374 			/* The userspace does not send tunnel attributes that
1375 			 * are 0, but we should not wildcard them nonetheless.
1376 			 */
1377 			if (match->key->tun_proto)
1378 				SW_FLOW_KEY_MEMSET_FIELD(match, tun_key,
1379 							 0xff, true);
1380 
1381 			nla_mask = newmask;
1382 		}
1383 
1384 		err = parse_flow_mask_nlattrs(nla_mask, a, &mask_attrs, log);
1385 		if (err)
1386 			goto free_newmask;
1387 
1388 		/* Always match on tci. */
1389 		SW_FLOW_KEY_PUT(match, eth.vlan.tci, htons(0xffff), true);
1390 		SW_FLOW_KEY_PUT(match, eth.cvlan.tci, htons(0xffff), true);
1391 
1392 		err = parse_vlan_from_nlattrs(match, &mask_attrs, a, true, log);
1393 		if (err)
1394 			goto free_newmask;
1395 
1396 		err = ovs_key_from_nlattrs(net, match, mask_attrs, a, true,
1397 					   log);
1398 		if (err)
1399 			goto free_newmask;
1400 	}
1401 
1402 	if (!match_validate(match, key_attrs, mask_attrs, log))
1403 		err = -EINVAL;
1404 
1405 free_newmask:
1406 	kfree(newmask);
1407 	return err;
1408 }
1409 
1410 static size_t get_ufid_len(const struct nlattr *attr, bool log)
1411 {
1412 	size_t len;
1413 
1414 	if (!attr)
1415 		return 0;
1416 
1417 	len = nla_len(attr);
1418 	if (len < 1 || len > MAX_UFID_LENGTH) {
1419 		OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
1420 			  nla_len(attr), MAX_UFID_LENGTH);
1421 		return 0;
1422 	}
1423 
1424 	return len;
1425 }
1426 
1427 /* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
1428  * or false otherwise.
1429  */
1430 bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
1431 		      bool log)
1432 {
1433 	sfid->ufid_len = get_ufid_len(attr, log);
1434 	if (sfid->ufid_len)
1435 		memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
1436 
1437 	return sfid->ufid_len;
1438 }
1439 
1440 int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
1441 			   const struct sw_flow_key *key, bool log)
1442 {
1443 	struct sw_flow_key *new_key;
1444 
1445 	if (ovs_nla_get_ufid(sfid, ufid, log))
1446 		return 0;
1447 
1448 	/* If UFID was not provided, use unmasked key. */
1449 	new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
1450 	if (!new_key)
1451 		return -ENOMEM;
1452 	memcpy(new_key, key, sizeof(*key));
1453 	sfid->unmasked_key = new_key;
1454 
1455 	return 0;
1456 }
1457 
1458 u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
1459 {
1460 	return attr ? nla_get_u32(attr) : 0;
1461 }
1462 
1463 /**
1464  * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
1465  * @key: Receives extracted in_port, priority, tun_key and skb_mark.
1466  * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1467  * sequence.
1468  * @log: Boolean to allow kernel error logging.  Normally true, but when
1469  * probing for feature compatibility this should be passed in as false to
1470  * suppress unnecessary error logging.
1471  *
1472  * This parses a series of Netlink attributes that form a flow key, which must
1473  * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1474  * get the metadata, that is, the parts of the flow key that cannot be
1475  * extracted from the packet itself.
1476  */
1477 
1478 int ovs_nla_get_flow_metadata(struct net *net, const struct nlattr *attr,
1479 			      struct sw_flow_key *key,
1480 			      bool log)
1481 {
1482 	const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1483 	struct sw_flow_match match;
1484 	u64 attrs = 0;
1485 	int err;
1486 
1487 	err = parse_flow_nlattrs(attr, a, &attrs, log);
1488 	if (err)
1489 		return -EINVAL;
1490 
1491 	memset(&match, 0, sizeof(match));
1492 	match.key = key;
1493 
1494 	memset(&key->ct, 0, sizeof(key->ct));
1495 	key->phy.in_port = DP_MAX_PORTS;
1496 
1497 	return metadata_from_nlattrs(net, &match, &attrs, a, false, log);
1498 }
1499 
1500 static int ovs_nla_put_vlan(struct sk_buff *skb, const struct vlan_head *vh,
1501 			    bool is_mask)
1502 {
1503 	__be16 eth_type = !is_mask ? vh->tpid : htons(0xffff);
1504 
1505 	if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1506 	    nla_put_be16(skb, OVS_KEY_ATTR_VLAN, vh->tci))
1507 		return -EMSGSIZE;
1508 	return 0;
1509 }
1510 
1511 static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
1512 			     const struct sw_flow_key *output, bool is_mask,
1513 			     struct sk_buff *skb)
1514 {
1515 	struct ovs_key_ethernet *eth_key;
1516 	struct nlattr *nla;
1517 	struct nlattr *encap = NULL;
1518 	struct nlattr *in_encap = NULL;
1519 
1520 	if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
1521 		goto nla_put_failure;
1522 
1523 	if (nla_put_u32(skb, OVS_KEY_ATTR_DP_HASH, output->ovs_flow_hash))
1524 		goto nla_put_failure;
1525 
1526 	if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1527 		goto nla_put_failure;
1528 
1529 	if ((swkey->tun_proto || is_mask)) {
1530 		const void *opts = NULL;
1531 
1532 		if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
1533 			opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
1534 
1535 		if (ip_tun_to_nlattr(skb, &output->tun_key, opts,
1536 				     swkey->tun_opts_len, swkey->tun_proto))
1537 			goto nla_put_failure;
1538 	}
1539 
1540 	if (swkey->phy.in_port == DP_MAX_PORTS) {
1541 		if (is_mask && (output->phy.in_port == 0xffff))
1542 			if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1543 				goto nla_put_failure;
1544 	} else {
1545 		u16 upper_u16;
1546 		upper_u16 = !is_mask ? 0 : 0xffff;
1547 
1548 		if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1549 				(upper_u16 << 16) | output->phy.in_port))
1550 			goto nla_put_failure;
1551 	}
1552 
1553 	if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1554 		goto nla_put_failure;
1555 
1556 	if (ovs_ct_put_key(output, skb))
1557 		goto nla_put_failure;
1558 
1559 	nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1560 	if (!nla)
1561 		goto nla_put_failure;
1562 
1563 	eth_key = nla_data(nla);
1564 	ether_addr_copy(eth_key->eth_src, output->eth.src);
1565 	ether_addr_copy(eth_key->eth_dst, output->eth.dst);
1566 
1567 	if (swkey->eth.vlan.tci || eth_type_vlan(swkey->eth.type)) {
1568 		if (ovs_nla_put_vlan(skb, &output->eth.vlan, is_mask))
1569 			goto nla_put_failure;
1570 		encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1571 		if (!swkey->eth.vlan.tci)
1572 			goto unencap;
1573 
1574 		if (swkey->eth.cvlan.tci || eth_type_vlan(swkey->eth.type)) {
1575 			if (ovs_nla_put_vlan(skb, &output->eth.cvlan, is_mask))
1576 				goto nla_put_failure;
1577 			in_encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1578 			if (!swkey->eth.cvlan.tci)
1579 				goto unencap;
1580 		}
1581 	}
1582 
1583 	if (swkey->eth.type == htons(ETH_P_802_2)) {
1584 		/*
1585 		 * Ethertype 802.2 is represented in the netlink with omitted
1586 		 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1587 		 * 0xffff in the mask attribute.  Ethertype can also
1588 		 * be wildcarded.
1589 		 */
1590 		if (is_mask && output->eth.type)
1591 			if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1592 						output->eth.type))
1593 				goto nla_put_failure;
1594 		goto unencap;
1595 	}
1596 
1597 	if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1598 		goto nla_put_failure;
1599 
1600 	if (eth_type_vlan(swkey->eth.type)) {
1601 		/* There are 3 VLAN tags, we don't know anything about the rest
1602 		 * of the packet, so truncate here.
1603 		 */
1604 		WARN_ON_ONCE(!(encap && in_encap));
1605 		goto unencap;
1606 	}
1607 
1608 	if (swkey->eth.type == htons(ETH_P_IP)) {
1609 		struct ovs_key_ipv4 *ipv4_key;
1610 
1611 		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1612 		if (!nla)
1613 			goto nla_put_failure;
1614 		ipv4_key = nla_data(nla);
1615 		ipv4_key->ipv4_src = output->ipv4.addr.src;
1616 		ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1617 		ipv4_key->ipv4_proto = output->ip.proto;
1618 		ipv4_key->ipv4_tos = output->ip.tos;
1619 		ipv4_key->ipv4_ttl = output->ip.ttl;
1620 		ipv4_key->ipv4_frag = output->ip.frag;
1621 	} else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1622 		struct ovs_key_ipv6 *ipv6_key;
1623 
1624 		nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1625 		if (!nla)
1626 			goto nla_put_failure;
1627 		ipv6_key = nla_data(nla);
1628 		memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1629 				sizeof(ipv6_key->ipv6_src));
1630 		memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1631 				sizeof(ipv6_key->ipv6_dst));
1632 		ipv6_key->ipv6_label = output->ipv6.label;
1633 		ipv6_key->ipv6_proto = output->ip.proto;
1634 		ipv6_key->ipv6_tclass = output->ip.tos;
1635 		ipv6_key->ipv6_hlimit = output->ip.ttl;
1636 		ipv6_key->ipv6_frag = output->ip.frag;
1637 	} else if (swkey->eth.type == htons(ETH_P_ARP) ||
1638 		   swkey->eth.type == htons(ETH_P_RARP)) {
1639 		struct ovs_key_arp *arp_key;
1640 
1641 		nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1642 		if (!nla)
1643 			goto nla_put_failure;
1644 		arp_key = nla_data(nla);
1645 		memset(arp_key, 0, sizeof(struct ovs_key_arp));
1646 		arp_key->arp_sip = output->ipv4.addr.src;
1647 		arp_key->arp_tip = output->ipv4.addr.dst;
1648 		arp_key->arp_op = htons(output->ip.proto);
1649 		ether_addr_copy(arp_key->arp_sha, output->ipv4.arp.sha);
1650 		ether_addr_copy(arp_key->arp_tha, output->ipv4.arp.tha);
1651 	} else if (eth_p_mpls(swkey->eth.type)) {
1652 		struct ovs_key_mpls *mpls_key;
1653 
1654 		nla = nla_reserve(skb, OVS_KEY_ATTR_MPLS, sizeof(*mpls_key));
1655 		if (!nla)
1656 			goto nla_put_failure;
1657 		mpls_key = nla_data(nla);
1658 		mpls_key->mpls_lse = output->mpls.top_lse;
1659 	}
1660 
1661 	if ((swkey->eth.type == htons(ETH_P_IP) ||
1662 	     swkey->eth.type == htons(ETH_P_IPV6)) &&
1663 	     swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1664 
1665 		if (swkey->ip.proto == IPPROTO_TCP) {
1666 			struct ovs_key_tcp *tcp_key;
1667 
1668 			nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1669 			if (!nla)
1670 				goto nla_put_failure;
1671 			tcp_key = nla_data(nla);
1672 			tcp_key->tcp_src = output->tp.src;
1673 			tcp_key->tcp_dst = output->tp.dst;
1674 			if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1675 					 output->tp.flags))
1676 				goto nla_put_failure;
1677 		} else if (swkey->ip.proto == IPPROTO_UDP) {
1678 			struct ovs_key_udp *udp_key;
1679 
1680 			nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1681 			if (!nla)
1682 				goto nla_put_failure;
1683 			udp_key = nla_data(nla);
1684 			udp_key->udp_src = output->tp.src;
1685 			udp_key->udp_dst = output->tp.dst;
1686 		} else if (swkey->ip.proto == IPPROTO_SCTP) {
1687 			struct ovs_key_sctp *sctp_key;
1688 
1689 			nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1690 			if (!nla)
1691 				goto nla_put_failure;
1692 			sctp_key = nla_data(nla);
1693 			sctp_key->sctp_src = output->tp.src;
1694 			sctp_key->sctp_dst = output->tp.dst;
1695 		} else if (swkey->eth.type == htons(ETH_P_IP) &&
1696 			   swkey->ip.proto == IPPROTO_ICMP) {
1697 			struct ovs_key_icmp *icmp_key;
1698 
1699 			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1700 			if (!nla)
1701 				goto nla_put_failure;
1702 			icmp_key = nla_data(nla);
1703 			icmp_key->icmp_type = ntohs(output->tp.src);
1704 			icmp_key->icmp_code = ntohs(output->tp.dst);
1705 		} else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1706 			   swkey->ip.proto == IPPROTO_ICMPV6) {
1707 			struct ovs_key_icmpv6 *icmpv6_key;
1708 
1709 			nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1710 						sizeof(*icmpv6_key));
1711 			if (!nla)
1712 				goto nla_put_failure;
1713 			icmpv6_key = nla_data(nla);
1714 			icmpv6_key->icmpv6_type = ntohs(output->tp.src);
1715 			icmpv6_key->icmpv6_code = ntohs(output->tp.dst);
1716 
1717 			if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1718 			    icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1719 				struct ovs_key_nd *nd_key;
1720 
1721 				nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1722 				if (!nla)
1723 					goto nla_put_failure;
1724 				nd_key = nla_data(nla);
1725 				memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1726 							sizeof(nd_key->nd_target));
1727 				ether_addr_copy(nd_key->nd_sll, output->ipv6.nd.sll);
1728 				ether_addr_copy(nd_key->nd_tll, output->ipv6.nd.tll);
1729 			}
1730 		}
1731 	}
1732 
1733 unencap:
1734 	if (in_encap)
1735 		nla_nest_end(skb, in_encap);
1736 	if (encap)
1737 		nla_nest_end(skb, encap);
1738 
1739 	return 0;
1740 
1741 nla_put_failure:
1742 	return -EMSGSIZE;
1743 }
1744 
1745 int ovs_nla_put_key(const struct sw_flow_key *swkey,
1746 		    const struct sw_flow_key *output, int attr, bool is_mask,
1747 		    struct sk_buff *skb)
1748 {
1749 	int err;
1750 	struct nlattr *nla;
1751 
1752 	nla = nla_nest_start(skb, attr);
1753 	if (!nla)
1754 		return -EMSGSIZE;
1755 	err = __ovs_nla_put_key(swkey, output, is_mask, skb);
1756 	if (err)
1757 		return err;
1758 	nla_nest_end(skb, nla);
1759 
1760 	return 0;
1761 }
1762 
1763 /* Called with ovs_mutex or RCU read lock. */
1764 int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
1765 {
1766 	if (ovs_identifier_is_ufid(&flow->id))
1767 		return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
1768 			       flow->id.ufid);
1769 
1770 	return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
1771 			       OVS_FLOW_ATTR_KEY, false, skb);
1772 }
1773 
1774 /* Called with ovs_mutex or RCU read lock. */
1775 int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
1776 {
1777 	return ovs_nla_put_key(&flow->key, &flow->key,
1778 				OVS_FLOW_ATTR_KEY, false, skb);
1779 }
1780 
1781 /* Called with ovs_mutex or RCU read lock. */
1782 int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
1783 {
1784 	return ovs_nla_put_key(&flow->key, &flow->mask->key,
1785 				OVS_FLOW_ATTR_MASK, true, skb);
1786 }
1787 
1788 #define MAX_ACTIONS_BUFSIZE	(32 * 1024)
1789 
1790 static struct sw_flow_actions *nla_alloc_flow_actions(int size, bool log)
1791 {
1792 	struct sw_flow_actions *sfa;
1793 
1794 	if (size > MAX_ACTIONS_BUFSIZE) {
1795 		OVS_NLERR(log, "Flow action size %u bytes exceeds max", size);
1796 		return ERR_PTR(-EINVAL);
1797 	}
1798 
1799 	sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1800 	if (!sfa)
1801 		return ERR_PTR(-ENOMEM);
1802 
1803 	sfa->actions_len = 0;
1804 	return sfa;
1805 }
1806 
1807 static void ovs_nla_free_set_action(const struct nlattr *a)
1808 {
1809 	const struct nlattr *ovs_key = nla_data(a);
1810 	struct ovs_tunnel_info *ovs_tun;
1811 
1812 	switch (nla_type(ovs_key)) {
1813 	case OVS_KEY_ATTR_TUNNEL_INFO:
1814 		ovs_tun = nla_data(ovs_key);
1815 		dst_release((struct dst_entry *)ovs_tun->tun_dst);
1816 		break;
1817 	}
1818 }
1819 
1820 void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1821 {
1822 	const struct nlattr *a;
1823 	int rem;
1824 
1825 	if (!sf_acts)
1826 		return;
1827 
1828 	nla_for_each_attr(a, sf_acts->actions, sf_acts->actions_len, rem) {
1829 		switch (nla_type(a)) {
1830 		case OVS_ACTION_ATTR_SET:
1831 			ovs_nla_free_set_action(a);
1832 			break;
1833 		case OVS_ACTION_ATTR_CT:
1834 			ovs_ct_free_action(a);
1835 			break;
1836 		}
1837 	}
1838 
1839 	kfree(sf_acts);
1840 }
1841 
1842 static void __ovs_nla_free_flow_actions(struct rcu_head *head)
1843 {
1844 	ovs_nla_free_flow_actions(container_of(head, struct sw_flow_actions, rcu));
1845 }
1846 
1847 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
1848  * The caller must hold rcu_read_lock for this to be sensible. */
1849 void ovs_nla_free_flow_actions_rcu(struct sw_flow_actions *sf_acts)
1850 {
1851 	call_rcu(&sf_acts->rcu, __ovs_nla_free_flow_actions);
1852 }
1853 
1854 static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1855 				       int attr_len, bool log)
1856 {
1857 
1858 	struct sw_flow_actions *acts;
1859 	int new_acts_size;
1860 	int req_size = NLA_ALIGN(attr_len);
1861 	int next_offset = offsetof(struct sw_flow_actions, actions) +
1862 					(*sfa)->actions_len;
1863 
1864 	if (req_size <= (ksize(*sfa) - next_offset))
1865 		goto out;
1866 
1867 	new_acts_size = ksize(*sfa) * 2;
1868 
1869 	if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1870 		if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1871 			return ERR_PTR(-EMSGSIZE);
1872 		new_acts_size = MAX_ACTIONS_BUFSIZE;
1873 	}
1874 
1875 	acts = nla_alloc_flow_actions(new_acts_size, log);
1876 	if (IS_ERR(acts))
1877 		return (void *)acts;
1878 
1879 	memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1880 	acts->actions_len = (*sfa)->actions_len;
1881 	acts->orig_len = (*sfa)->orig_len;
1882 	kfree(*sfa);
1883 	*sfa = acts;
1884 
1885 out:
1886 	(*sfa)->actions_len += req_size;
1887 	return  (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1888 }
1889 
1890 static struct nlattr *__add_action(struct sw_flow_actions **sfa,
1891 				   int attrtype, void *data, int len, bool log)
1892 {
1893 	struct nlattr *a;
1894 
1895 	a = reserve_sfa_size(sfa, nla_attr_size(len), log);
1896 	if (IS_ERR(a))
1897 		return a;
1898 
1899 	a->nla_type = attrtype;
1900 	a->nla_len = nla_attr_size(len);
1901 
1902 	if (data)
1903 		memcpy(nla_data(a), data, len);
1904 	memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1905 
1906 	return a;
1907 }
1908 
1909 int ovs_nla_add_action(struct sw_flow_actions **sfa, int attrtype, void *data,
1910 		       int len, bool log)
1911 {
1912 	struct nlattr *a;
1913 
1914 	a = __add_action(sfa, attrtype, data, len, log);
1915 
1916 	return PTR_ERR_OR_ZERO(a);
1917 }
1918 
1919 static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1920 					  int attrtype, bool log)
1921 {
1922 	int used = (*sfa)->actions_len;
1923 	int err;
1924 
1925 	err = ovs_nla_add_action(sfa, attrtype, NULL, 0, log);
1926 	if (err)
1927 		return err;
1928 
1929 	return used;
1930 }
1931 
1932 static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1933 					 int st_offset)
1934 {
1935 	struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1936 							       st_offset);
1937 
1938 	a->nla_len = sfa->actions_len - st_offset;
1939 }
1940 
1941 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
1942 				  const struct sw_flow_key *key,
1943 				  int depth, struct sw_flow_actions **sfa,
1944 				  __be16 eth_type, __be16 vlan_tci, bool log);
1945 
1946 static int validate_and_copy_sample(struct net *net, const struct nlattr *attr,
1947 				    const struct sw_flow_key *key, int depth,
1948 				    struct sw_flow_actions **sfa,
1949 				    __be16 eth_type, __be16 vlan_tci, bool log)
1950 {
1951 	const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1952 	const struct nlattr *probability, *actions;
1953 	const struct nlattr *a;
1954 	int rem, start, err, st_acts;
1955 
1956 	memset(attrs, 0, sizeof(attrs));
1957 	nla_for_each_nested(a, attr, rem) {
1958 		int type = nla_type(a);
1959 		if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1960 			return -EINVAL;
1961 		attrs[type] = a;
1962 	}
1963 	if (rem)
1964 		return -EINVAL;
1965 
1966 	probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1967 	if (!probability || nla_len(probability) != sizeof(u32))
1968 		return -EINVAL;
1969 
1970 	actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1971 	if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1972 		return -EINVAL;
1973 
1974 	/* validation done, copy sample action. */
1975 	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE, log);
1976 	if (start < 0)
1977 		return start;
1978 	err = ovs_nla_add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1979 				 nla_data(probability), sizeof(u32), log);
1980 	if (err)
1981 		return err;
1982 	st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS, log);
1983 	if (st_acts < 0)
1984 		return st_acts;
1985 
1986 	err = __ovs_nla_copy_actions(net, actions, key, depth + 1, sfa,
1987 				     eth_type, vlan_tci, log);
1988 	if (err)
1989 		return err;
1990 
1991 	add_nested_action_end(*sfa, st_acts);
1992 	add_nested_action_end(*sfa, start);
1993 
1994 	return 0;
1995 }
1996 
1997 void ovs_match_init(struct sw_flow_match *match,
1998 		    struct sw_flow_key *key,
1999 		    bool reset_key,
2000 		    struct sw_flow_mask *mask)
2001 {
2002 	memset(match, 0, sizeof(*match));
2003 	match->key = key;
2004 	match->mask = mask;
2005 
2006 	if (reset_key)
2007 		memset(key, 0, sizeof(*key));
2008 
2009 	if (mask) {
2010 		memset(&mask->key, 0, sizeof(mask->key));
2011 		mask->range.start = mask->range.end = 0;
2012 	}
2013 }
2014 
2015 static int validate_geneve_opts(struct sw_flow_key *key)
2016 {
2017 	struct geneve_opt *option;
2018 	int opts_len = key->tun_opts_len;
2019 	bool crit_opt = false;
2020 
2021 	option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
2022 	while (opts_len > 0) {
2023 		int len;
2024 
2025 		if (opts_len < sizeof(*option))
2026 			return -EINVAL;
2027 
2028 		len = sizeof(*option) + option->length * 4;
2029 		if (len > opts_len)
2030 			return -EINVAL;
2031 
2032 		crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
2033 
2034 		option = (struct geneve_opt *)((u8 *)option + len);
2035 		opts_len -= len;
2036 	};
2037 
2038 	key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
2039 
2040 	return 0;
2041 }
2042 
2043 static int validate_and_copy_set_tun(const struct nlattr *attr,
2044 				     struct sw_flow_actions **sfa, bool log)
2045 {
2046 	struct sw_flow_match match;
2047 	struct sw_flow_key key;
2048 	struct metadata_dst *tun_dst;
2049 	struct ip_tunnel_info *tun_info;
2050 	struct ovs_tunnel_info *ovs_tun;
2051 	struct nlattr *a;
2052 	int err = 0, start, opts_type;
2053 
2054 	ovs_match_init(&match, &key, true, NULL);
2055 	opts_type = ip_tun_from_nlattr(nla_data(attr), &match, false, log);
2056 	if (opts_type < 0)
2057 		return opts_type;
2058 
2059 	if (key.tun_opts_len) {
2060 		switch (opts_type) {
2061 		case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
2062 			err = validate_geneve_opts(&key);
2063 			if (err < 0)
2064 				return err;
2065 			break;
2066 		case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
2067 			break;
2068 		}
2069 	};
2070 
2071 	start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
2072 	if (start < 0)
2073 		return start;
2074 
2075 	tun_dst = metadata_dst_alloc(key.tun_opts_len, GFP_KERNEL);
2076 	if (!tun_dst)
2077 		return -ENOMEM;
2078 
2079 	err = dst_cache_init(&tun_dst->u.tun_info.dst_cache, GFP_KERNEL);
2080 	if (err) {
2081 		dst_release((struct dst_entry *)tun_dst);
2082 		return err;
2083 	}
2084 
2085 	a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
2086 			 sizeof(*ovs_tun), log);
2087 	if (IS_ERR(a)) {
2088 		dst_release((struct dst_entry *)tun_dst);
2089 		return PTR_ERR(a);
2090 	}
2091 
2092 	ovs_tun = nla_data(a);
2093 	ovs_tun->tun_dst = tun_dst;
2094 
2095 	tun_info = &tun_dst->u.tun_info;
2096 	tun_info->mode = IP_TUNNEL_INFO_TX;
2097 	if (key.tun_proto == AF_INET6)
2098 		tun_info->mode |= IP_TUNNEL_INFO_IPV6;
2099 	tun_info->key = key.tun_key;
2100 
2101 	/* We need to store the options in the action itself since
2102 	 * everything else will go away after flow setup. We can append
2103 	 * it to tun_info and then point there.
2104 	 */
2105 	ip_tunnel_info_opts_set(tun_info,
2106 				TUN_METADATA_OPTS(&key, key.tun_opts_len),
2107 				key.tun_opts_len);
2108 	add_nested_action_end(*sfa, start);
2109 
2110 	return err;
2111 }
2112 
2113 /* Return false if there are any non-masked bits set.
2114  * Mask follows data immediately, before any netlink padding.
2115  */
2116 static bool validate_masked(u8 *data, int len)
2117 {
2118 	u8 *mask = data + len;
2119 
2120 	while (len--)
2121 		if (*data++ & ~*mask++)
2122 			return false;
2123 
2124 	return true;
2125 }
2126 
2127 static int validate_set(const struct nlattr *a,
2128 			const struct sw_flow_key *flow_key,
2129 			struct sw_flow_actions **sfa,
2130 			bool *skip_copy, __be16 eth_type, bool masked, bool log)
2131 {
2132 	const struct nlattr *ovs_key = nla_data(a);
2133 	int key_type = nla_type(ovs_key);
2134 	size_t key_len;
2135 
2136 	/* There can be only one key in a action */
2137 	if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
2138 		return -EINVAL;
2139 
2140 	key_len = nla_len(ovs_key);
2141 	if (masked)
2142 		key_len /= 2;
2143 
2144 	if (key_type > OVS_KEY_ATTR_MAX ||
2145 	    !check_attr_len(key_len, ovs_key_lens[key_type].len))
2146 		return -EINVAL;
2147 
2148 	if (masked && !validate_masked(nla_data(ovs_key), key_len))
2149 		return -EINVAL;
2150 
2151 	switch (key_type) {
2152 	const struct ovs_key_ipv4 *ipv4_key;
2153 	const struct ovs_key_ipv6 *ipv6_key;
2154 	int err;
2155 
2156 	case OVS_KEY_ATTR_PRIORITY:
2157 	case OVS_KEY_ATTR_SKB_MARK:
2158 	case OVS_KEY_ATTR_CT_MARK:
2159 	case OVS_KEY_ATTR_CT_LABELS:
2160 	case OVS_KEY_ATTR_ETHERNET:
2161 		break;
2162 
2163 	case OVS_KEY_ATTR_TUNNEL:
2164 		if (masked)
2165 			return -EINVAL; /* Masked tunnel set not supported. */
2166 
2167 		*skip_copy = true;
2168 		err = validate_and_copy_set_tun(a, sfa, log);
2169 		if (err)
2170 			return err;
2171 		break;
2172 
2173 	case OVS_KEY_ATTR_IPV4:
2174 		if (eth_type != htons(ETH_P_IP))
2175 			return -EINVAL;
2176 
2177 		ipv4_key = nla_data(ovs_key);
2178 
2179 		if (masked) {
2180 			const struct ovs_key_ipv4 *mask = ipv4_key + 1;
2181 
2182 			/* Non-writeable fields. */
2183 			if (mask->ipv4_proto || mask->ipv4_frag)
2184 				return -EINVAL;
2185 		} else {
2186 			if (ipv4_key->ipv4_proto != flow_key->ip.proto)
2187 				return -EINVAL;
2188 
2189 			if (ipv4_key->ipv4_frag != flow_key->ip.frag)
2190 				return -EINVAL;
2191 		}
2192 		break;
2193 
2194 	case OVS_KEY_ATTR_IPV6:
2195 		if (eth_type != htons(ETH_P_IPV6))
2196 			return -EINVAL;
2197 
2198 		ipv6_key = nla_data(ovs_key);
2199 
2200 		if (masked) {
2201 			const struct ovs_key_ipv6 *mask = ipv6_key + 1;
2202 
2203 			/* Non-writeable fields. */
2204 			if (mask->ipv6_proto || mask->ipv6_frag)
2205 				return -EINVAL;
2206 
2207 			/* Invalid bits in the flow label mask? */
2208 			if (ntohl(mask->ipv6_label) & 0xFFF00000)
2209 				return -EINVAL;
2210 		} else {
2211 			if (ipv6_key->ipv6_proto != flow_key->ip.proto)
2212 				return -EINVAL;
2213 
2214 			if (ipv6_key->ipv6_frag != flow_key->ip.frag)
2215 				return -EINVAL;
2216 		}
2217 		if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
2218 			return -EINVAL;
2219 
2220 		break;
2221 
2222 	case OVS_KEY_ATTR_TCP:
2223 		if ((eth_type != htons(ETH_P_IP) &&
2224 		     eth_type != htons(ETH_P_IPV6)) ||
2225 		    flow_key->ip.proto != IPPROTO_TCP)
2226 			return -EINVAL;
2227 
2228 		break;
2229 
2230 	case OVS_KEY_ATTR_UDP:
2231 		if ((eth_type != htons(ETH_P_IP) &&
2232 		     eth_type != htons(ETH_P_IPV6)) ||
2233 		    flow_key->ip.proto != IPPROTO_UDP)
2234 			return -EINVAL;
2235 
2236 		break;
2237 
2238 	case OVS_KEY_ATTR_MPLS:
2239 		if (!eth_p_mpls(eth_type))
2240 			return -EINVAL;
2241 		break;
2242 
2243 	case OVS_KEY_ATTR_SCTP:
2244 		if ((eth_type != htons(ETH_P_IP) &&
2245 		     eth_type != htons(ETH_P_IPV6)) ||
2246 		    flow_key->ip.proto != IPPROTO_SCTP)
2247 			return -EINVAL;
2248 
2249 		break;
2250 
2251 	default:
2252 		return -EINVAL;
2253 	}
2254 
2255 	/* Convert non-masked non-tunnel set actions to masked set actions. */
2256 	if (!masked && key_type != OVS_KEY_ATTR_TUNNEL) {
2257 		int start, len = key_len * 2;
2258 		struct nlattr *at;
2259 
2260 		*skip_copy = true;
2261 
2262 		start = add_nested_action_start(sfa,
2263 						OVS_ACTION_ATTR_SET_TO_MASKED,
2264 						log);
2265 		if (start < 0)
2266 			return start;
2267 
2268 		at = __add_action(sfa, key_type, NULL, len, log);
2269 		if (IS_ERR(at))
2270 			return PTR_ERR(at);
2271 
2272 		memcpy(nla_data(at), nla_data(ovs_key), key_len); /* Key. */
2273 		memset(nla_data(at) + key_len, 0xff, key_len);    /* Mask. */
2274 		/* Clear non-writeable bits from otherwise writeable fields. */
2275 		if (key_type == OVS_KEY_ATTR_IPV6) {
2276 			struct ovs_key_ipv6 *mask = nla_data(at) + key_len;
2277 
2278 			mask->ipv6_label &= htonl(0x000FFFFF);
2279 		}
2280 		add_nested_action_end(*sfa, start);
2281 	}
2282 
2283 	return 0;
2284 }
2285 
2286 static int validate_userspace(const struct nlattr *attr)
2287 {
2288 	static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
2289 		[OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
2290 		[OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
2291 		[OVS_USERSPACE_ATTR_EGRESS_TUN_PORT] = {.type = NLA_U32 },
2292 	};
2293 	struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
2294 	int error;
2295 
2296 	error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
2297 				 attr, userspace_policy);
2298 	if (error)
2299 		return error;
2300 
2301 	if (!a[OVS_USERSPACE_ATTR_PID] ||
2302 	    !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
2303 		return -EINVAL;
2304 
2305 	return 0;
2306 }
2307 
2308 static int copy_action(const struct nlattr *from,
2309 		       struct sw_flow_actions **sfa, bool log)
2310 {
2311 	int totlen = NLA_ALIGN(from->nla_len);
2312 	struct nlattr *to;
2313 
2314 	to = reserve_sfa_size(sfa, from->nla_len, log);
2315 	if (IS_ERR(to))
2316 		return PTR_ERR(to);
2317 
2318 	memcpy(to, from, totlen);
2319 	return 0;
2320 }
2321 
2322 static int __ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2323 				  const struct sw_flow_key *key,
2324 				  int depth, struct sw_flow_actions **sfa,
2325 				  __be16 eth_type, __be16 vlan_tci, bool log)
2326 {
2327 	const struct nlattr *a;
2328 	int rem, err;
2329 
2330 	if (depth >= SAMPLE_ACTION_DEPTH)
2331 		return -EOVERFLOW;
2332 
2333 	nla_for_each_nested(a, attr, rem) {
2334 		/* Expected argument lengths, (u32)-1 for variable length. */
2335 		static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
2336 			[OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
2337 			[OVS_ACTION_ATTR_RECIRC] = sizeof(u32),
2338 			[OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
2339 			[OVS_ACTION_ATTR_PUSH_MPLS] = sizeof(struct ovs_action_push_mpls),
2340 			[OVS_ACTION_ATTR_POP_MPLS] = sizeof(__be16),
2341 			[OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
2342 			[OVS_ACTION_ATTR_POP_VLAN] = 0,
2343 			[OVS_ACTION_ATTR_SET] = (u32)-1,
2344 			[OVS_ACTION_ATTR_SET_MASKED] = (u32)-1,
2345 			[OVS_ACTION_ATTR_SAMPLE] = (u32)-1,
2346 			[OVS_ACTION_ATTR_HASH] = sizeof(struct ovs_action_hash),
2347 			[OVS_ACTION_ATTR_CT] = (u32)-1,
2348 			[OVS_ACTION_ATTR_TRUNC] = sizeof(struct ovs_action_trunc),
2349 		};
2350 		const struct ovs_action_push_vlan *vlan;
2351 		int type = nla_type(a);
2352 		bool skip_copy;
2353 
2354 		if (type > OVS_ACTION_ATTR_MAX ||
2355 		    (action_lens[type] != nla_len(a) &&
2356 		     action_lens[type] != (u32)-1))
2357 			return -EINVAL;
2358 
2359 		skip_copy = false;
2360 		switch (type) {
2361 		case OVS_ACTION_ATTR_UNSPEC:
2362 			return -EINVAL;
2363 
2364 		case OVS_ACTION_ATTR_USERSPACE:
2365 			err = validate_userspace(a);
2366 			if (err)
2367 				return err;
2368 			break;
2369 
2370 		case OVS_ACTION_ATTR_OUTPUT:
2371 			if (nla_get_u32(a) >= DP_MAX_PORTS)
2372 				return -EINVAL;
2373 			break;
2374 
2375 		case OVS_ACTION_ATTR_TRUNC: {
2376 			const struct ovs_action_trunc *trunc = nla_data(a);
2377 
2378 			if (trunc->max_len < ETH_HLEN)
2379 				return -EINVAL;
2380 			break;
2381 		}
2382 
2383 		case OVS_ACTION_ATTR_HASH: {
2384 			const struct ovs_action_hash *act_hash = nla_data(a);
2385 
2386 			switch (act_hash->hash_alg) {
2387 			case OVS_HASH_ALG_L4:
2388 				break;
2389 			default:
2390 				return  -EINVAL;
2391 			}
2392 
2393 			break;
2394 		}
2395 
2396 		case OVS_ACTION_ATTR_POP_VLAN:
2397 			vlan_tci = htons(0);
2398 			break;
2399 
2400 		case OVS_ACTION_ATTR_PUSH_VLAN:
2401 			vlan = nla_data(a);
2402 			if (!eth_type_vlan(vlan->vlan_tpid))
2403 				return -EINVAL;
2404 			if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
2405 				return -EINVAL;
2406 			vlan_tci = vlan->vlan_tci;
2407 			break;
2408 
2409 		case OVS_ACTION_ATTR_RECIRC:
2410 			break;
2411 
2412 		case OVS_ACTION_ATTR_PUSH_MPLS: {
2413 			const struct ovs_action_push_mpls *mpls = nla_data(a);
2414 
2415 			if (!eth_p_mpls(mpls->mpls_ethertype))
2416 				return -EINVAL;
2417 			/* Prohibit push MPLS other than to a white list
2418 			 * for packets that have a known tag order.
2419 			 */
2420 			if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2421 			    (eth_type != htons(ETH_P_IP) &&
2422 			     eth_type != htons(ETH_P_IPV6) &&
2423 			     eth_type != htons(ETH_P_ARP) &&
2424 			     eth_type != htons(ETH_P_RARP) &&
2425 			     !eth_p_mpls(eth_type)))
2426 				return -EINVAL;
2427 			eth_type = mpls->mpls_ethertype;
2428 			break;
2429 		}
2430 
2431 		case OVS_ACTION_ATTR_POP_MPLS:
2432 			if (vlan_tci & htons(VLAN_TAG_PRESENT) ||
2433 			    !eth_p_mpls(eth_type))
2434 				return -EINVAL;
2435 
2436 			/* Disallow subsequent L2.5+ set and mpls_pop actions
2437 			 * as there is no check here to ensure that the new
2438 			 * eth_type is valid and thus set actions could
2439 			 * write off the end of the packet or otherwise
2440 			 * corrupt it.
2441 			 *
2442 			 * Support for these actions is planned using packet
2443 			 * recirculation.
2444 			 */
2445 			eth_type = htons(0);
2446 			break;
2447 
2448 		case OVS_ACTION_ATTR_SET:
2449 			err = validate_set(a, key, sfa,
2450 					   &skip_copy, eth_type, false, log);
2451 			if (err)
2452 				return err;
2453 			break;
2454 
2455 		case OVS_ACTION_ATTR_SET_MASKED:
2456 			err = validate_set(a, key, sfa,
2457 					   &skip_copy, eth_type, true, log);
2458 			if (err)
2459 				return err;
2460 			break;
2461 
2462 		case OVS_ACTION_ATTR_SAMPLE:
2463 			err = validate_and_copy_sample(net, a, key, depth, sfa,
2464 						       eth_type, vlan_tci, log);
2465 			if (err)
2466 				return err;
2467 			skip_copy = true;
2468 			break;
2469 
2470 		case OVS_ACTION_ATTR_CT:
2471 			err = ovs_ct_copy_action(net, a, key, sfa, log);
2472 			if (err)
2473 				return err;
2474 			skip_copy = true;
2475 			break;
2476 
2477 		default:
2478 			OVS_NLERR(log, "Unknown Action type %d", type);
2479 			return -EINVAL;
2480 		}
2481 		if (!skip_copy) {
2482 			err = copy_action(a, sfa, log);
2483 			if (err)
2484 				return err;
2485 		}
2486 	}
2487 
2488 	if (rem > 0)
2489 		return -EINVAL;
2490 
2491 	return 0;
2492 }
2493 
2494 /* 'key' must be the masked key. */
2495 int ovs_nla_copy_actions(struct net *net, const struct nlattr *attr,
2496 			 const struct sw_flow_key *key,
2497 			 struct sw_flow_actions **sfa, bool log)
2498 {
2499 	int err;
2500 
2501 	*sfa = nla_alloc_flow_actions(nla_len(attr), log);
2502 	if (IS_ERR(*sfa))
2503 		return PTR_ERR(*sfa);
2504 
2505 	(*sfa)->orig_len = nla_len(attr);
2506 	err = __ovs_nla_copy_actions(net, attr, key, 0, sfa, key->eth.type,
2507 				     key->eth.vlan.tci, log);
2508 	if (err)
2509 		ovs_nla_free_flow_actions(*sfa);
2510 
2511 	return err;
2512 }
2513 
2514 static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
2515 {
2516 	const struct nlattr *a;
2517 	struct nlattr *start;
2518 	int err = 0, rem;
2519 
2520 	start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
2521 	if (!start)
2522 		return -EMSGSIZE;
2523 
2524 	nla_for_each_nested(a, attr, rem) {
2525 		int type = nla_type(a);
2526 		struct nlattr *st_sample;
2527 
2528 		switch (type) {
2529 		case OVS_SAMPLE_ATTR_PROBABILITY:
2530 			if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
2531 				    sizeof(u32), nla_data(a)))
2532 				return -EMSGSIZE;
2533 			break;
2534 		case OVS_SAMPLE_ATTR_ACTIONS:
2535 			st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
2536 			if (!st_sample)
2537 				return -EMSGSIZE;
2538 			err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
2539 			if (err)
2540 				return err;
2541 			nla_nest_end(skb, st_sample);
2542 			break;
2543 		}
2544 	}
2545 
2546 	nla_nest_end(skb, start);
2547 	return err;
2548 }
2549 
2550 static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
2551 {
2552 	const struct nlattr *ovs_key = nla_data(a);
2553 	int key_type = nla_type(ovs_key);
2554 	struct nlattr *start;
2555 	int err;
2556 
2557 	switch (key_type) {
2558 	case OVS_KEY_ATTR_TUNNEL_INFO: {
2559 		struct ovs_tunnel_info *ovs_tun = nla_data(ovs_key);
2560 		struct ip_tunnel_info *tun_info = &ovs_tun->tun_dst->u.tun_info;
2561 
2562 		start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2563 		if (!start)
2564 			return -EMSGSIZE;
2565 
2566 		err =  ip_tun_to_nlattr(skb, &tun_info->key,
2567 					ip_tunnel_info_opts(tun_info),
2568 					tun_info->options_len,
2569 					ip_tunnel_info_af(tun_info));
2570 		if (err)
2571 			return err;
2572 		nla_nest_end(skb, start);
2573 		break;
2574 	}
2575 	default:
2576 		if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
2577 			return -EMSGSIZE;
2578 		break;
2579 	}
2580 
2581 	return 0;
2582 }
2583 
2584 static int masked_set_action_to_set_action_attr(const struct nlattr *a,
2585 						struct sk_buff *skb)
2586 {
2587 	const struct nlattr *ovs_key = nla_data(a);
2588 	struct nlattr *nla;
2589 	size_t key_len = nla_len(ovs_key) / 2;
2590 
2591 	/* Revert the conversion we did from a non-masked set action to
2592 	 * masked set action.
2593 	 */
2594 	nla = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
2595 	if (!nla)
2596 		return -EMSGSIZE;
2597 
2598 	if (nla_put(skb, nla_type(ovs_key), key_len, nla_data(ovs_key)))
2599 		return -EMSGSIZE;
2600 
2601 	nla_nest_end(skb, nla);
2602 	return 0;
2603 }
2604 
2605 int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
2606 {
2607 	const struct nlattr *a;
2608 	int rem, err;
2609 
2610 	nla_for_each_attr(a, attr, len, rem) {
2611 		int type = nla_type(a);
2612 
2613 		switch (type) {
2614 		case OVS_ACTION_ATTR_SET:
2615 			err = set_action_to_attr(a, skb);
2616 			if (err)
2617 				return err;
2618 			break;
2619 
2620 		case OVS_ACTION_ATTR_SET_TO_MASKED:
2621 			err = masked_set_action_to_set_action_attr(a, skb);
2622 			if (err)
2623 				return err;
2624 			break;
2625 
2626 		case OVS_ACTION_ATTR_SAMPLE:
2627 			err = sample_action_to_attr(a, skb);
2628 			if (err)
2629 				return err;
2630 			break;
2631 
2632 		case OVS_ACTION_ATTR_CT:
2633 			err = ovs_ct_action_to_attr(nla_data(a), skb);
2634 			if (err)
2635 				return err;
2636 			break;
2637 
2638 		default:
2639 			if (nla_put(skb, type, nla_len(a), nla_data(a)))
2640 				return -EMSGSIZE;
2641 			break;
2642 		}
2643 	}
2644 
2645 	return 0;
2646 }
2647