xref: /openbmc/linux/net/sched/cls_flower.c (revision e7253313)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * net/sched/cls_flower.c		Flower classifier
4  *
5  * Copyright (c) 2015 Jiri Pirko <jiri@resnulli.us>
6  */
7 
8 #include <linux/kernel.h>
9 #include <linux/init.h>
10 #include <linux/module.h>
11 #include <linux/rhashtable.h>
12 #include <linux/workqueue.h>
13 #include <linux/refcount.h>
14 
15 #include <linux/if_ether.h>
16 #include <linux/in6.h>
17 #include <linux/ip.h>
18 #include <linux/mpls.h>
19 
20 #include <net/sch_generic.h>
21 #include <net/pkt_cls.h>
22 #include <net/ip.h>
23 #include <net/flow_dissector.h>
24 #include <net/geneve.h>
25 #include <net/vxlan.h>
26 #include <net/erspan.h>
27 
28 #include <net/dst.h>
29 #include <net/dst_metadata.h>
30 
31 #include <uapi/linux/netfilter/nf_conntrack_common.h>
32 
33 struct fl_flow_key {
34 	struct flow_dissector_key_meta meta;
35 	struct flow_dissector_key_control control;
36 	struct flow_dissector_key_control enc_control;
37 	struct flow_dissector_key_basic basic;
38 	struct flow_dissector_key_eth_addrs eth;
39 	struct flow_dissector_key_vlan vlan;
40 	struct flow_dissector_key_vlan cvlan;
41 	union {
42 		struct flow_dissector_key_ipv4_addrs ipv4;
43 		struct flow_dissector_key_ipv6_addrs ipv6;
44 	};
45 	struct flow_dissector_key_ports tp;
46 	struct flow_dissector_key_icmp icmp;
47 	struct flow_dissector_key_arp arp;
48 	struct flow_dissector_key_keyid enc_key_id;
49 	union {
50 		struct flow_dissector_key_ipv4_addrs enc_ipv4;
51 		struct flow_dissector_key_ipv6_addrs enc_ipv6;
52 	};
53 	struct flow_dissector_key_ports enc_tp;
54 	struct flow_dissector_key_mpls mpls;
55 	struct flow_dissector_key_tcp tcp;
56 	struct flow_dissector_key_ip ip;
57 	struct flow_dissector_key_ip enc_ip;
58 	struct flow_dissector_key_enc_opts enc_opts;
59 	union {
60 		struct flow_dissector_key_ports tp;
61 		struct {
62 			struct flow_dissector_key_ports tp_min;
63 			struct flow_dissector_key_ports tp_max;
64 		};
65 	} tp_range;
66 	struct flow_dissector_key_ct ct;
67 } __aligned(BITS_PER_LONG / 8); /* Ensure that we can do comparisons as longs. */
68 
69 struct fl_flow_mask_range {
70 	unsigned short int start;
71 	unsigned short int end;
72 };
73 
74 struct fl_flow_mask {
75 	struct fl_flow_key key;
76 	struct fl_flow_mask_range range;
77 	u32 flags;
78 	struct rhash_head ht_node;
79 	struct rhashtable ht;
80 	struct rhashtable_params filter_ht_params;
81 	struct flow_dissector dissector;
82 	struct list_head filters;
83 	struct rcu_work rwork;
84 	struct list_head list;
85 	refcount_t refcnt;
86 };
87 
88 struct fl_flow_tmplt {
89 	struct fl_flow_key dummy_key;
90 	struct fl_flow_key mask;
91 	struct flow_dissector dissector;
92 	struct tcf_chain *chain;
93 };
94 
95 struct cls_fl_head {
96 	struct rhashtable ht;
97 	spinlock_t masks_lock; /* Protect masks list */
98 	struct list_head masks;
99 	struct list_head hw_filters;
100 	struct rcu_work rwork;
101 	struct idr handle_idr;
102 };
103 
104 struct cls_fl_filter {
105 	struct fl_flow_mask *mask;
106 	struct rhash_head ht_node;
107 	struct fl_flow_key mkey;
108 	struct tcf_exts exts;
109 	struct tcf_result res;
110 	struct fl_flow_key key;
111 	struct list_head list;
112 	struct list_head hw_list;
113 	u32 handle;
114 	u32 flags;
115 	u32 in_hw_count;
116 	struct rcu_work rwork;
117 	struct net_device *hw_dev;
118 	/* Flower classifier is unlocked, which means that its reference counter
119 	 * can be changed concurrently without any kind of external
120 	 * synchronization. Use atomic reference counter to be concurrency-safe.
121 	 */
122 	refcount_t refcnt;
123 	bool deleted;
124 };
125 
126 static const struct rhashtable_params mask_ht_params = {
127 	.key_offset = offsetof(struct fl_flow_mask, key),
128 	.key_len = sizeof(struct fl_flow_key),
129 	.head_offset = offsetof(struct fl_flow_mask, ht_node),
130 	.automatic_shrinking = true,
131 };
132 
133 static unsigned short int fl_mask_range(const struct fl_flow_mask *mask)
134 {
135 	return mask->range.end - mask->range.start;
136 }
137 
138 static void fl_mask_update_range(struct fl_flow_mask *mask)
139 {
140 	const u8 *bytes = (const u8 *) &mask->key;
141 	size_t size = sizeof(mask->key);
142 	size_t i, first = 0, last;
143 
144 	for (i = 0; i < size; i++) {
145 		if (bytes[i]) {
146 			first = i;
147 			break;
148 		}
149 	}
150 	last = first;
151 	for (i = size - 1; i != first; i--) {
152 		if (bytes[i]) {
153 			last = i;
154 			break;
155 		}
156 	}
157 	mask->range.start = rounddown(first, sizeof(long));
158 	mask->range.end = roundup(last + 1, sizeof(long));
159 }
160 
161 static void *fl_key_get_start(struct fl_flow_key *key,
162 			      const struct fl_flow_mask *mask)
163 {
164 	return (u8 *) key + mask->range.start;
165 }
166 
167 static void fl_set_masked_key(struct fl_flow_key *mkey, struct fl_flow_key *key,
168 			      struct fl_flow_mask *mask)
169 {
170 	const long *lkey = fl_key_get_start(key, mask);
171 	const long *lmask = fl_key_get_start(&mask->key, mask);
172 	long *lmkey = fl_key_get_start(mkey, mask);
173 	int i;
174 
175 	for (i = 0; i < fl_mask_range(mask); i += sizeof(long))
176 		*lmkey++ = *lkey++ & *lmask++;
177 }
178 
179 static bool fl_mask_fits_tmplt(struct fl_flow_tmplt *tmplt,
180 			       struct fl_flow_mask *mask)
181 {
182 	const long *lmask = fl_key_get_start(&mask->key, mask);
183 	const long *ltmplt;
184 	int i;
185 
186 	if (!tmplt)
187 		return true;
188 	ltmplt = fl_key_get_start(&tmplt->mask, mask);
189 	for (i = 0; i < fl_mask_range(mask); i += sizeof(long)) {
190 		if (~*ltmplt++ & *lmask++)
191 			return false;
192 	}
193 	return true;
194 }
195 
196 static void fl_clear_masked_range(struct fl_flow_key *key,
197 				  struct fl_flow_mask *mask)
198 {
199 	memset(fl_key_get_start(key, mask), 0, fl_mask_range(mask));
200 }
201 
202 static bool fl_range_port_dst_cmp(struct cls_fl_filter *filter,
203 				  struct fl_flow_key *key,
204 				  struct fl_flow_key *mkey)
205 {
206 	__be16 min_mask, max_mask, min_val, max_val;
207 
208 	min_mask = htons(filter->mask->key.tp_range.tp_min.dst);
209 	max_mask = htons(filter->mask->key.tp_range.tp_max.dst);
210 	min_val = htons(filter->key.tp_range.tp_min.dst);
211 	max_val = htons(filter->key.tp_range.tp_max.dst);
212 
213 	if (min_mask && max_mask) {
214 		if (htons(key->tp_range.tp.dst) < min_val ||
215 		    htons(key->tp_range.tp.dst) > max_val)
216 			return false;
217 
218 		/* skb does not have min and max values */
219 		mkey->tp_range.tp_min.dst = filter->mkey.tp_range.tp_min.dst;
220 		mkey->tp_range.tp_max.dst = filter->mkey.tp_range.tp_max.dst;
221 	}
222 	return true;
223 }
224 
225 static bool fl_range_port_src_cmp(struct cls_fl_filter *filter,
226 				  struct fl_flow_key *key,
227 				  struct fl_flow_key *mkey)
228 {
229 	__be16 min_mask, max_mask, min_val, max_val;
230 
231 	min_mask = htons(filter->mask->key.tp_range.tp_min.src);
232 	max_mask = htons(filter->mask->key.tp_range.tp_max.src);
233 	min_val = htons(filter->key.tp_range.tp_min.src);
234 	max_val = htons(filter->key.tp_range.tp_max.src);
235 
236 	if (min_mask && max_mask) {
237 		if (htons(key->tp_range.tp.src) < min_val ||
238 		    htons(key->tp_range.tp.src) > max_val)
239 			return false;
240 
241 		/* skb does not have min and max values */
242 		mkey->tp_range.tp_min.src = filter->mkey.tp_range.tp_min.src;
243 		mkey->tp_range.tp_max.src = filter->mkey.tp_range.tp_max.src;
244 	}
245 	return true;
246 }
247 
248 static struct cls_fl_filter *__fl_lookup(struct fl_flow_mask *mask,
249 					 struct fl_flow_key *mkey)
250 {
251 	return rhashtable_lookup_fast(&mask->ht, fl_key_get_start(mkey, mask),
252 				      mask->filter_ht_params);
253 }
254 
255 static struct cls_fl_filter *fl_lookup_range(struct fl_flow_mask *mask,
256 					     struct fl_flow_key *mkey,
257 					     struct fl_flow_key *key)
258 {
259 	struct cls_fl_filter *filter, *f;
260 
261 	list_for_each_entry_rcu(filter, &mask->filters, list) {
262 		if (!fl_range_port_dst_cmp(filter, key, mkey))
263 			continue;
264 
265 		if (!fl_range_port_src_cmp(filter, key, mkey))
266 			continue;
267 
268 		f = __fl_lookup(mask, mkey);
269 		if (f)
270 			return f;
271 	}
272 	return NULL;
273 }
274 
275 static struct cls_fl_filter *fl_lookup(struct fl_flow_mask *mask,
276 				       struct fl_flow_key *mkey,
277 				       struct fl_flow_key *key)
278 {
279 	if ((mask->flags & TCA_FLOWER_MASK_FLAGS_RANGE))
280 		return fl_lookup_range(mask, mkey, key);
281 
282 	return __fl_lookup(mask, mkey);
283 }
284 
285 static u16 fl_ct_info_to_flower_map[] = {
286 	[IP_CT_ESTABLISHED] =		TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
287 					TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED,
288 	[IP_CT_RELATED] =		TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
289 					TCA_FLOWER_KEY_CT_FLAGS_RELATED,
290 	[IP_CT_ESTABLISHED_REPLY] =	TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
291 					TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED,
292 	[IP_CT_RELATED_REPLY] =		TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
293 					TCA_FLOWER_KEY_CT_FLAGS_RELATED,
294 	[IP_CT_NEW] =			TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
295 					TCA_FLOWER_KEY_CT_FLAGS_NEW,
296 };
297 
298 static int fl_classify(struct sk_buff *skb, const struct tcf_proto *tp,
299 		       struct tcf_result *res)
300 {
301 	struct cls_fl_head *head = rcu_dereference_bh(tp->root);
302 	struct fl_flow_key skb_mkey;
303 	struct fl_flow_key skb_key;
304 	struct fl_flow_mask *mask;
305 	struct cls_fl_filter *f;
306 
307 	list_for_each_entry_rcu(mask, &head->masks, list) {
308 		fl_clear_masked_range(&skb_key, mask);
309 
310 		skb_flow_dissect_meta(skb, &mask->dissector, &skb_key);
311 		/* skb_flow_dissect() does not set n_proto in case an unknown
312 		 * protocol, so do it rather here.
313 		 */
314 		skb_key.basic.n_proto = skb->protocol;
315 		skb_flow_dissect_tunnel_info(skb, &mask->dissector, &skb_key);
316 		skb_flow_dissect_ct(skb, &mask->dissector, &skb_key,
317 				    fl_ct_info_to_flower_map,
318 				    ARRAY_SIZE(fl_ct_info_to_flower_map));
319 		skb_flow_dissect(skb, &mask->dissector, &skb_key, 0);
320 
321 		fl_set_masked_key(&skb_mkey, &skb_key, mask);
322 
323 		f = fl_lookup(mask, &skb_mkey, &skb_key);
324 		if (f && !tc_skip_sw(f->flags)) {
325 			*res = f->res;
326 			return tcf_exts_exec(skb, &f->exts, res);
327 		}
328 	}
329 	return -1;
330 }
331 
332 static int fl_init(struct tcf_proto *tp)
333 {
334 	struct cls_fl_head *head;
335 
336 	head = kzalloc(sizeof(*head), GFP_KERNEL);
337 	if (!head)
338 		return -ENOBUFS;
339 
340 	spin_lock_init(&head->masks_lock);
341 	INIT_LIST_HEAD_RCU(&head->masks);
342 	INIT_LIST_HEAD(&head->hw_filters);
343 	rcu_assign_pointer(tp->root, head);
344 	idr_init(&head->handle_idr);
345 
346 	return rhashtable_init(&head->ht, &mask_ht_params);
347 }
348 
349 static void fl_mask_free(struct fl_flow_mask *mask, bool mask_init_done)
350 {
351 	/* temporary masks don't have their filters list and ht initialized */
352 	if (mask_init_done) {
353 		WARN_ON(!list_empty(&mask->filters));
354 		rhashtable_destroy(&mask->ht);
355 	}
356 	kfree(mask);
357 }
358 
359 static void fl_mask_free_work(struct work_struct *work)
360 {
361 	struct fl_flow_mask *mask = container_of(to_rcu_work(work),
362 						 struct fl_flow_mask, rwork);
363 
364 	fl_mask_free(mask, true);
365 }
366 
367 static void fl_uninit_mask_free_work(struct work_struct *work)
368 {
369 	struct fl_flow_mask *mask = container_of(to_rcu_work(work),
370 						 struct fl_flow_mask, rwork);
371 
372 	fl_mask_free(mask, false);
373 }
374 
375 static bool fl_mask_put(struct cls_fl_head *head, struct fl_flow_mask *mask)
376 {
377 	if (!refcount_dec_and_test(&mask->refcnt))
378 		return false;
379 
380 	rhashtable_remove_fast(&head->ht, &mask->ht_node, mask_ht_params);
381 
382 	spin_lock(&head->masks_lock);
383 	list_del_rcu(&mask->list);
384 	spin_unlock(&head->masks_lock);
385 
386 	tcf_queue_work(&mask->rwork, fl_mask_free_work);
387 
388 	return true;
389 }
390 
391 static struct cls_fl_head *fl_head_dereference(struct tcf_proto *tp)
392 {
393 	/* Flower classifier only changes root pointer during init and destroy.
394 	 * Users must obtain reference to tcf_proto instance before calling its
395 	 * API, so tp->root pointer is protected from concurrent call to
396 	 * fl_destroy() by reference counting.
397 	 */
398 	return rcu_dereference_raw(tp->root);
399 }
400 
401 static void __fl_destroy_filter(struct cls_fl_filter *f)
402 {
403 	tcf_exts_destroy(&f->exts);
404 	tcf_exts_put_net(&f->exts);
405 	kfree(f);
406 }
407 
408 static void fl_destroy_filter_work(struct work_struct *work)
409 {
410 	struct cls_fl_filter *f = container_of(to_rcu_work(work),
411 					struct cls_fl_filter, rwork);
412 
413 	__fl_destroy_filter(f);
414 }
415 
416 static void fl_hw_destroy_filter(struct tcf_proto *tp, struct cls_fl_filter *f,
417 				 bool rtnl_held, struct netlink_ext_ack *extack)
418 {
419 	struct tcf_block *block = tp->chain->block;
420 	struct flow_cls_offload cls_flower = {};
421 
422 	tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
423 	cls_flower.command = FLOW_CLS_DESTROY;
424 	cls_flower.cookie = (unsigned long) f;
425 
426 	tc_setup_cb_destroy(block, tp, TC_SETUP_CLSFLOWER, &cls_flower, false,
427 			    &f->flags, &f->in_hw_count, rtnl_held);
428 
429 }
430 
431 static int fl_hw_replace_filter(struct tcf_proto *tp,
432 				struct cls_fl_filter *f, bool rtnl_held,
433 				struct netlink_ext_ack *extack)
434 {
435 	struct tcf_block *block = tp->chain->block;
436 	struct flow_cls_offload cls_flower = {};
437 	bool skip_sw = tc_skip_sw(f->flags);
438 	int err = 0;
439 
440 	cls_flower.rule = flow_rule_alloc(tcf_exts_num_actions(&f->exts));
441 	if (!cls_flower.rule)
442 		return -ENOMEM;
443 
444 	tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
445 	cls_flower.command = FLOW_CLS_REPLACE;
446 	cls_flower.cookie = (unsigned long) f;
447 	cls_flower.rule->match.dissector = &f->mask->dissector;
448 	cls_flower.rule->match.mask = &f->mask->key;
449 	cls_flower.rule->match.key = &f->mkey;
450 	cls_flower.classid = f->res.classid;
451 
452 	err = tc_setup_flow_action(&cls_flower.rule->action, &f->exts,
453 				   rtnl_held);
454 	if (err) {
455 		kfree(cls_flower.rule);
456 		if (skip_sw) {
457 			NL_SET_ERR_MSG_MOD(extack, "Failed to setup flow action");
458 			return err;
459 		}
460 		return 0;
461 	}
462 
463 	err = tc_setup_cb_add(block, tp, TC_SETUP_CLSFLOWER, &cls_flower,
464 			      skip_sw, &f->flags, &f->in_hw_count, rtnl_held);
465 	tc_cleanup_flow_action(&cls_flower.rule->action);
466 	kfree(cls_flower.rule);
467 
468 	if (err) {
469 		fl_hw_destroy_filter(tp, f, rtnl_held, NULL);
470 		return err;
471 	}
472 
473 	if (skip_sw && !(f->flags & TCA_CLS_FLAGS_IN_HW))
474 		return -EINVAL;
475 
476 	return 0;
477 }
478 
479 static void fl_hw_update_stats(struct tcf_proto *tp, struct cls_fl_filter *f,
480 			       bool rtnl_held)
481 {
482 	struct tcf_block *block = tp->chain->block;
483 	struct flow_cls_offload cls_flower = {};
484 
485 	tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, NULL);
486 	cls_flower.command = FLOW_CLS_STATS;
487 	cls_flower.cookie = (unsigned long) f;
488 	cls_flower.classid = f->res.classid;
489 
490 	tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false,
491 			 rtnl_held);
492 
493 	tcf_exts_stats_update(&f->exts, cls_flower.stats.bytes,
494 			      cls_flower.stats.pkts,
495 			      cls_flower.stats.lastused);
496 }
497 
498 static void __fl_put(struct cls_fl_filter *f)
499 {
500 	if (!refcount_dec_and_test(&f->refcnt))
501 		return;
502 
503 	if (tcf_exts_get_net(&f->exts))
504 		tcf_queue_work(&f->rwork, fl_destroy_filter_work);
505 	else
506 		__fl_destroy_filter(f);
507 }
508 
509 static struct cls_fl_filter *__fl_get(struct cls_fl_head *head, u32 handle)
510 {
511 	struct cls_fl_filter *f;
512 
513 	rcu_read_lock();
514 	f = idr_find(&head->handle_idr, handle);
515 	if (f && !refcount_inc_not_zero(&f->refcnt))
516 		f = NULL;
517 	rcu_read_unlock();
518 
519 	return f;
520 }
521 
522 static int __fl_delete(struct tcf_proto *tp, struct cls_fl_filter *f,
523 		       bool *last, bool rtnl_held,
524 		       struct netlink_ext_ack *extack)
525 {
526 	struct cls_fl_head *head = fl_head_dereference(tp);
527 
528 	*last = false;
529 
530 	spin_lock(&tp->lock);
531 	if (f->deleted) {
532 		spin_unlock(&tp->lock);
533 		return -ENOENT;
534 	}
535 
536 	f->deleted = true;
537 	rhashtable_remove_fast(&f->mask->ht, &f->ht_node,
538 			       f->mask->filter_ht_params);
539 	idr_remove(&head->handle_idr, f->handle);
540 	list_del_rcu(&f->list);
541 	spin_unlock(&tp->lock);
542 
543 	*last = fl_mask_put(head, f->mask);
544 	if (!tc_skip_hw(f->flags))
545 		fl_hw_destroy_filter(tp, f, rtnl_held, extack);
546 	tcf_unbind_filter(tp, &f->res);
547 	__fl_put(f);
548 
549 	return 0;
550 }
551 
552 static void fl_destroy_sleepable(struct work_struct *work)
553 {
554 	struct cls_fl_head *head = container_of(to_rcu_work(work),
555 						struct cls_fl_head,
556 						rwork);
557 
558 	rhashtable_destroy(&head->ht);
559 	kfree(head);
560 	module_put(THIS_MODULE);
561 }
562 
563 static void fl_destroy(struct tcf_proto *tp, bool rtnl_held,
564 		       struct netlink_ext_ack *extack)
565 {
566 	struct cls_fl_head *head = fl_head_dereference(tp);
567 	struct fl_flow_mask *mask, *next_mask;
568 	struct cls_fl_filter *f, *next;
569 	bool last;
570 
571 	list_for_each_entry_safe(mask, next_mask, &head->masks, list) {
572 		list_for_each_entry_safe(f, next, &mask->filters, list) {
573 			__fl_delete(tp, f, &last, rtnl_held, extack);
574 			if (last)
575 				break;
576 		}
577 	}
578 	idr_destroy(&head->handle_idr);
579 
580 	__module_get(THIS_MODULE);
581 	tcf_queue_work(&head->rwork, fl_destroy_sleepable);
582 }
583 
584 static void fl_put(struct tcf_proto *tp, void *arg)
585 {
586 	struct cls_fl_filter *f = arg;
587 
588 	__fl_put(f);
589 }
590 
591 static void *fl_get(struct tcf_proto *tp, u32 handle)
592 {
593 	struct cls_fl_head *head = fl_head_dereference(tp);
594 
595 	return __fl_get(head, handle);
596 }
597 
598 static const struct nla_policy fl_policy[TCA_FLOWER_MAX + 1] = {
599 	[TCA_FLOWER_UNSPEC]		= { .type = NLA_UNSPEC },
600 	[TCA_FLOWER_CLASSID]		= { .type = NLA_U32 },
601 	[TCA_FLOWER_INDEV]		= { .type = NLA_STRING,
602 					    .len = IFNAMSIZ },
603 	[TCA_FLOWER_KEY_ETH_DST]	= { .len = ETH_ALEN },
604 	[TCA_FLOWER_KEY_ETH_DST_MASK]	= { .len = ETH_ALEN },
605 	[TCA_FLOWER_KEY_ETH_SRC]	= { .len = ETH_ALEN },
606 	[TCA_FLOWER_KEY_ETH_SRC_MASK]	= { .len = ETH_ALEN },
607 	[TCA_FLOWER_KEY_ETH_TYPE]	= { .type = NLA_U16 },
608 	[TCA_FLOWER_KEY_IP_PROTO]	= { .type = NLA_U8 },
609 	[TCA_FLOWER_KEY_IPV4_SRC]	= { .type = NLA_U32 },
610 	[TCA_FLOWER_KEY_IPV4_SRC_MASK]	= { .type = NLA_U32 },
611 	[TCA_FLOWER_KEY_IPV4_DST]	= { .type = NLA_U32 },
612 	[TCA_FLOWER_KEY_IPV4_DST_MASK]	= { .type = NLA_U32 },
613 	[TCA_FLOWER_KEY_IPV6_SRC]	= { .len = sizeof(struct in6_addr) },
614 	[TCA_FLOWER_KEY_IPV6_SRC_MASK]	= { .len = sizeof(struct in6_addr) },
615 	[TCA_FLOWER_KEY_IPV6_DST]	= { .len = sizeof(struct in6_addr) },
616 	[TCA_FLOWER_KEY_IPV6_DST_MASK]	= { .len = sizeof(struct in6_addr) },
617 	[TCA_FLOWER_KEY_TCP_SRC]	= { .type = NLA_U16 },
618 	[TCA_FLOWER_KEY_TCP_DST]	= { .type = NLA_U16 },
619 	[TCA_FLOWER_KEY_UDP_SRC]	= { .type = NLA_U16 },
620 	[TCA_FLOWER_KEY_UDP_DST]	= { .type = NLA_U16 },
621 	[TCA_FLOWER_KEY_VLAN_ID]	= { .type = NLA_U16 },
622 	[TCA_FLOWER_KEY_VLAN_PRIO]	= { .type = NLA_U8 },
623 	[TCA_FLOWER_KEY_VLAN_ETH_TYPE]	= { .type = NLA_U16 },
624 	[TCA_FLOWER_KEY_ENC_KEY_ID]	= { .type = NLA_U32 },
625 	[TCA_FLOWER_KEY_ENC_IPV4_SRC]	= { .type = NLA_U32 },
626 	[TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK] = { .type = NLA_U32 },
627 	[TCA_FLOWER_KEY_ENC_IPV4_DST]	= { .type = NLA_U32 },
628 	[TCA_FLOWER_KEY_ENC_IPV4_DST_MASK] = { .type = NLA_U32 },
629 	[TCA_FLOWER_KEY_ENC_IPV6_SRC]	= { .len = sizeof(struct in6_addr) },
630 	[TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK] = { .len = sizeof(struct in6_addr) },
631 	[TCA_FLOWER_KEY_ENC_IPV6_DST]	= { .len = sizeof(struct in6_addr) },
632 	[TCA_FLOWER_KEY_ENC_IPV6_DST_MASK] = { .len = sizeof(struct in6_addr) },
633 	[TCA_FLOWER_KEY_TCP_SRC_MASK]	= { .type = NLA_U16 },
634 	[TCA_FLOWER_KEY_TCP_DST_MASK]	= { .type = NLA_U16 },
635 	[TCA_FLOWER_KEY_UDP_SRC_MASK]	= { .type = NLA_U16 },
636 	[TCA_FLOWER_KEY_UDP_DST_MASK]	= { .type = NLA_U16 },
637 	[TCA_FLOWER_KEY_SCTP_SRC_MASK]	= { .type = NLA_U16 },
638 	[TCA_FLOWER_KEY_SCTP_DST_MASK]	= { .type = NLA_U16 },
639 	[TCA_FLOWER_KEY_SCTP_SRC]	= { .type = NLA_U16 },
640 	[TCA_FLOWER_KEY_SCTP_DST]	= { .type = NLA_U16 },
641 	[TCA_FLOWER_KEY_ENC_UDP_SRC_PORT]	= { .type = NLA_U16 },
642 	[TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK]	= { .type = NLA_U16 },
643 	[TCA_FLOWER_KEY_ENC_UDP_DST_PORT]	= { .type = NLA_U16 },
644 	[TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK]	= { .type = NLA_U16 },
645 	[TCA_FLOWER_KEY_FLAGS]		= { .type = NLA_U32 },
646 	[TCA_FLOWER_KEY_FLAGS_MASK]	= { .type = NLA_U32 },
647 	[TCA_FLOWER_KEY_ICMPV4_TYPE]	= { .type = NLA_U8 },
648 	[TCA_FLOWER_KEY_ICMPV4_TYPE_MASK] = { .type = NLA_U8 },
649 	[TCA_FLOWER_KEY_ICMPV4_CODE]	= { .type = NLA_U8 },
650 	[TCA_FLOWER_KEY_ICMPV4_CODE_MASK] = { .type = NLA_U8 },
651 	[TCA_FLOWER_KEY_ICMPV6_TYPE]	= { .type = NLA_U8 },
652 	[TCA_FLOWER_KEY_ICMPV6_TYPE_MASK] = { .type = NLA_U8 },
653 	[TCA_FLOWER_KEY_ICMPV6_CODE]	= { .type = NLA_U8 },
654 	[TCA_FLOWER_KEY_ICMPV6_CODE_MASK] = { .type = NLA_U8 },
655 	[TCA_FLOWER_KEY_ARP_SIP]	= { .type = NLA_U32 },
656 	[TCA_FLOWER_KEY_ARP_SIP_MASK]	= { .type = NLA_U32 },
657 	[TCA_FLOWER_KEY_ARP_TIP]	= { .type = NLA_U32 },
658 	[TCA_FLOWER_KEY_ARP_TIP_MASK]	= { .type = NLA_U32 },
659 	[TCA_FLOWER_KEY_ARP_OP]		= { .type = NLA_U8 },
660 	[TCA_FLOWER_KEY_ARP_OP_MASK]	= { .type = NLA_U8 },
661 	[TCA_FLOWER_KEY_ARP_SHA]	= { .len = ETH_ALEN },
662 	[TCA_FLOWER_KEY_ARP_SHA_MASK]	= { .len = ETH_ALEN },
663 	[TCA_FLOWER_KEY_ARP_THA]	= { .len = ETH_ALEN },
664 	[TCA_FLOWER_KEY_ARP_THA_MASK]	= { .len = ETH_ALEN },
665 	[TCA_FLOWER_KEY_MPLS_TTL]	= { .type = NLA_U8 },
666 	[TCA_FLOWER_KEY_MPLS_BOS]	= { .type = NLA_U8 },
667 	[TCA_FLOWER_KEY_MPLS_TC]	= { .type = NLA_U8 },
668 	[TCA_FLOWER_KEY_MPLS_LABEL]	= { .type = NLA_U32 },
669 	[TCA_FLOWER_KEY_TCP_FLAGS]	= { .type = NLA_U16 },
670 	[TCA_FLOWER_KEY_TCP_FLAGS_MASK]	= { .type = NLA_U16 },
671 	[TCA_FLOWER_KEY_IP_TOS]		= { .type = NLA_U8 },
672 	[TCA_FLOWER_KEY_IP_TOS_MASK]	= { .type = NLA_U8 },
673 	[TCA_FLOWER_KEY_IP_TTL]		= { .type = NLA_U8 },
674 	[TCA_FLOWER_KEY_IP_TTL_MASK]	= { .type = NLA_U8 },
675 	[TCA_FLOWER_KEY_CVLAN_ID]	= { .type = NLA_U16 },
676 	[TCA_FLOWER_KEY_CVLAN_PRIO]	= { .type = NLA_U8 },
677 	[TCA_FLOWER_KEY_CVLAN_ETH_TYPE]	= { .type = NLA_U16 },
678 	[TCA_FLOWER_KEY_ENC_IP_TOS]	= { .type = NLA_U8 },
679 	[TCA_FLOWER_KEY_ENC_IP_TOS_MASK] = { .type = NLA_U8 },
680 	[TCA_FLOWER_KEY_ENC_IP_TTL]	 = { .type = NLA_U8 },
681 	[TCA_FLOWER_KEY_ENC_IP_TTL_MASK] = { .type = NLA_U8 },
682 	[TCA_FLOWER_KEY_ENC_OPTS]	= { .type = NLA_NESTED },
683 	[TCA_FLOWER_KEY_ENC_OPTS_MASK]	= { .type = NLA_NESTED },
684 	[TCA_FLOWER_KEY_CT_STATE]	= { .type = NLA_U16 },
685 	[TCA_FLOWER_KEY_CT_STATE_MASK]	= { .type = NLA_U16 },
686 	[TCA_FLOWER_KEY_CT_ZONE]	= { .type = NLA_U16 },
687 	[TCA_FLOWER_KEY_CT_ZONE_MASK]	= { .type = NLA_U16 },
688 	[TCA_FLOWER_KEY_CT_MARK]	= { .type = NLA_U32 },
689 	[TCA_FLOWER_KEY_CT_MARK_MASK]	= { .type = NLA_U32 },
690 	[TCA_FLOWER_KEY_CT_LABELS]	= { .type = NLA_BINARY,
691 					    .len = 128 / BITS_PER_BYTE },
692 	[TCA_FLOWER_KEY_CT_LABELS_MASK]	= { .type = NLA_BINARY,
693 					    .len = 128 / BITS_PER_BYTE },
694 };
695 
696 static const struct nla_policy
697 enc_opts_policy[TCA_FLOWER_KEY_ENC_OPTS_MAX + 1] = {
698 	[TCA_FLOWER_KEY_ENC_OPTS_UNSPEC]        = {
699 		.strict_start_type = TCA_FLOWER_KEY_ENC_OPTS_VXLAN },
700 	[TCA_FLOWER_KEY_ENC_OPTS_GENEVE]        = { .type = NLA_NESTED },
701 	[TCA_FLOWER_KEY_ENC_OPTS_VXLAN]         = { .type = NLA_NESTED },
702 	[TCA_FLOWER_KEY_ENC_OPTS_ERSPAN]        = { .type = NLA_NESTED },
703 };
704 
705 static const struct nla_policy
706 geneve_opt_policy[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1] = {
707 	[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS]      = { .type = NLA_U16 },
708 	[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE]       = { .type = NLA_U8 },
709 	[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA]       = { .type = NLA_BINARY,
710 						       .len = 128 },
711 };
712 
713 static const struct nla_policy
714 vxlan_opt_policy[TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX + 1] = {
715 	[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]         = { .type = NLA_U32 },
716 };
717 
718 static const struct nla_policy
719 erspan_opt_policy[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX + 1] = {
720 	[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]        = { .type = NLA_U8 },
721 	[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]      = { .type = NLA_U32 },
722 	[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR]        = { .type = NLA_U8 },
723 	[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID]       = { .type = NLA_U8 },
724 };
725 
726 static void fl_set_key_val(struct nlattr **tb,
727 			   void *val, int val_type,
728 			   void *mask, int mask_type, int len)
729 {
730 	if (!tb[val_type])
731 		return;
732 	nla_memcpy(val, tb[val_type], len);
733 	if (mask_type == TCA_FLOWER_UNSPEC || !tb[mask_type])
734 		memset(mask, 0xff, len);
735 	else
736 		nla_memcpy(mask, tb[mask_type], len);
737 }
738 
739 static int fl_set_key_port_range(struct nlattr **tb, struct fl_flow_key *key,
740 				 struct fl_flow_key *mask)
741 {
742 	fl_set_key_val(tb, &key->tp_range.tp_min.dst,
743 		       TCA_FLOWER_KEY_PORT_DST_MIN, &mask->tp_range.tp_min.dst,
744 		       TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_min.dst));
745 	fl_set_key_val(tb, &key->tp_range.tp_max.dst,
746 		       TCA_FLOWER_KEY_PORT_DST_MAX, &mask->tp_range.tp_max.dst,
747 		       TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_max.dst));
748 	fl_set_key_val(tb, &key->tp_range.tp_min.src,
749 		       TCA_FLOWER_KEY_PORT_SRC_MIN, &mask->tp_range.tp_min.src,
750 		       TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_min.src));
751 	fl_set_key_val(tb, &key->tp_range.tp_max.src,
752 		       TCA_FLOWER_KEY_PORT_SRC_MAX, &mask->tp_range.tp_max.src,
753 		       TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_max.src));
754 
755 	if ((mask->tp_range.tp_min.dst && mask->tp_range.tp_max.dst &&
756 	     htons(key->tp_range.tp_max.dst) <=
757 		 htons(key->tp_range.tp_min.dst)) ||
758 	    (mask->tp_range.tp_min.src && mask->tp_range.tp_max.src &&
759 	     htons(key->tp_range.tp_max.src) <=
760 		 htons(key->tp_range.tp_min.src)))
761 		return -EINVAL;
762 
763 	return 0;
764 }
765 
766 static int fl_set_key_mpls(struct nlattr **tb,
767 			   struct flow_dissector_key_mpls *key_val,
768 			   struct flow_dissector_key_mpls *key_mask)
769 {
770 	if (tb[TCA_FLOWER_KEY_MPLS_TTL]) {
771 		key_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TTL]);
772 		key_mask->mpls_ttl = MPLS_TTL_MASK;
773 	}
774 	if (tb[TCA_FLOWER_KEY_MPLS_BOS]) {
775 		u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_BOS]);
776 
777 		if (bos & ~MPLS_BOS_MASK)
778 			return -EINVAL;
779 		key_val->mpls_bos = bos;
780 		key_mask->mpls_bos = MPLS_BOS_MASK;
781 	}
782 	if (tb[TCA_FLOWER_KEY_MPLS_TC]) {
783 		u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TC]);
784 
785 		if (tc & ~MPLS_TC_MASK)
786 			return -EINVAL;
787 		key_val->mpls_tc = tc;
788 		key_mask->mpls_tc = MPLS_TC_MASK;
789 	}
790 	if (tb[TCA_FLOWER_KEY_MPLS_LABEL]) {
791 		u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_LABEL]);
792 
793 		if (label & ~MPLS_LABEL_MASK)
794 			return -EINVAL;
795 		key_val->mpls_label = label;
796 		key_mask->mpls_label = MPLS_LABEL_MASK;
797 	}
798 	return 0;
799 }
800 
801 static void fl_set_key_vlan(struct nlattr **tb,
802 			    __be16 ethertype,
803 			    int vlan_id_key, int vlan_prio_key,
804 			    struct flow_dissector_key_vlan *key_val,
805 			    struct flow_dissector_key_vlan *key_mask)
806 {
807 #define VLAN_PRIORITY_MASK	0x7
808 
809 	if (tb[vlan_id_key]) {
810 		key_val->vlan_id =
811 			nla_get_u16(tb[vlan_id_key]) & VLAN_VID_MASK;
812 		key_mask->vlan_id = VLAN_VID_MASK;
813 	}
814 	if (tb[vlan_prio_key]) {
815 		key_val->vlan_priority =
816 			nla_get_u8(tb[vlan_prio_key]) &
817 			VLAN_PRIORITY_MASK;
818 		key_mask->vlan_priority = VLAN_PRIORITY_MASK;
819 	}
820 	key_val->vlan_tpid = ethertype;
821 	key_mask->vlan_tpid = cpu_to_be16(~0);
822 }
823 
824 static void fl_set_key_flag(u32 flower_key, u32 flower_mask,
825 			    u32 *dissector_key, u32 *dissector_mask,
826 			    u32 flower_flag_bit, u32 dissector_flag_bit)
827 {
828 	if (flower_mask & flower_flag_bit) {
829 		*dissector_mask |= dissector_flag_bit;
830 		if (flower_key & flower_flag_bit)
831 			*dissector_key |= dissector_flag_bit;
832 	}
833 }
834 
835 static int fl_set_key_flags(struct nlattr **tb,
836 			    u32 *flags_key, u32 *flags_mask)
837 {
838 	u32 key, mask;
839 
840 	/* mask is mandatory for flags */
841 	if (!tb[TCA_FLOWER_KEY_FLAGS_MASK])
842 		return -EINVAL;
843 
844 	key = be32_to_cpu(nla_get_u32(tb[TCA_FLOWER_KEY_FLAGS]));
845 	mask = be32_to_cpu(nla_get_u32(tb[TCA_FLOWER_KEY_FLAGS_MASK]));
846 
847 	*flags_key  = 0;
848 	*flags_mask = 0;
849 
850 	fl_set_key_flag(key, mask, flags_key, flags_mask,
851 			TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
852 	fl_set_key_flag(key, mask, flags_key, flags_mask,
853 			TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
854 			FLOW_DIS_FIRST_FRAG);
855 
856 	return 0;
857 }
858 
859 static void fl_set_key_ip(struct nlattr **tb, bool encap,
860 			  struct flow_dissector_key_ip *key,
861 			  struct flow_dissector_key_ip *mask)
862 {
863 	int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
864 	int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
865 	int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
866 	int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
867 
868 	fl_set_key_val(tb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos));
869 	fl_set_key_val(tb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl));
870 }
871 
872 static int fl_set_geneve_opt(const struct nlattr *nla, struct fl_flow_key *key,
873 			     int depth, int option_len,
874 			     struct netlink_ext_ack *extack)
875 {
876 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1];
877 	struct nlattr *class = NULL, *type = NULL, *data = NULL;
878 	struct geneve_opt *opt;
879 	int err, data_len = 0;
880 
881 	if (option_len > sizeof(struct geneve_opt))
882 		data_len = option_len - sizeof(struct geneve_opt);
883 
884 	opt = (struct geneve_opt *)&key->enc_opts.data[key->enc_opts.len];
885 	memset(opt, 0xff, option_len);
886 	opt->length = data_len / 4;
887 	opt->r1 = 0;
888 	opt->r2 = 0;
889 	opt->r3 = 0;
890 
891 	/* If no mask has been prodived we assume an exact match. */
892 	if (!depth)
893 		return sizeof(struct geneve_opt) + data_len;
894 
895 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_GENEVE) {
896 		NL_SET_ERR_MSG(extack, "Non-geneve option type for mask");
897 		return -EINVAL;
898 	}
899 
900 	err = nla_parse_nested_deprecated(tb,
901 					  TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX,
902 					  nla, geneve_opt_policy, extack);
903 	if (err < 0)
904 		return err;
905 
906 	/* We are not allowed to omit any of CLASS, TYPE or DATA
907 	 * fields from the key.
908 	 */
909 	if (!option_len &&
910 	    (!tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS] ||
911 	     !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE] ||
912 	     !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA])) {
913 		NL_SET_ERR_MSG(extack, "Missing tunnel key geneve option class, type or data");
914 		return -EINVAL;
915 	}
916 
917 	/* Omitting any of CLASS, TYPE or DATA fields is allowed
918 	 * for the mask.
919 	 */
920 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA]) {
921 		int new_len = key->enc_opts.len;
922 
923 		data = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA];
924 		data_len = nla_len(data);
925 		if (data_len < 4) {
926 			NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is less than 4 bytes long");
927 			return -ERANGE;
928 		}
929 		if (data_len % 4) {
930 			NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is not a multiple of 4 bytes long");
931 			return -ERANGE;
932 		}
933 
934 		new_len += sizeof(struct geneve_opt) + data_len;
935 		BUILD_BUG_ON(FLOW_DIS_TUN_OPTS_MAX != IP_TUNNEL_OPTS_MAX);
936 		if (new_len > FLOW_DIS_TUN_OPTS_MAX) {
937 			NL_SET_ERR_MSG(extack, "Tunnel options exceeds max size");
938 			return -ERANGE;
939 		}
940 		opt->length = data_len / 4;
941 		memcpy(opt->opt_data, nla_data(data), data_len);
942 	}
943 
944 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS]) {
945 		class = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS];
946 		opt->opt_class = nla_get_be16(class);
947 	}
948 
949 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE]) {
950 		type = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE];
951 		opt->type = nla_get_u8(type);
952 	}
953 
954 	return sizeof(struct geneve_opt) + data_len;
955 }
956 
957 static int fl_set_vxlan_opt(const struct nlattr *nla, struct fl_flow_key *key,
958 			    int depth, int option_len,
959 			    struct netlink_ext_ack *extack)
960 {
961 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX + 1];
962 	struct vxlan_metadata *md;
963 	int err;
964 
965 	md = (struct vxlan_metadata *)&key->enc_opts.data[key->enc_opts.len];
966 	memset(md, 0xff, sizeof(*md));
967 
968 	if (!depth)
969 		return sizeof(*md);
970 
971 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_VXLAN) {
972 		NL_SET_ERR_MSG(extack, "Non-vxlan option type for mask");
973 		return -EINVAL;
974 	}
975 
976 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX, nla,
977 			       vxlan_opt_policy, extack);
978 	if (err < 0)
979 		return err;
980 
981 	if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]) {
982 		NL_SET_ERR_MSG(extack, "Missing tunnel key vxlan option gbp");
983 		return -EINVAL;
984 	}
985 
986 	if (tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP])
987 		md->gbp = nla_get_u32(tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]);
988 
989 	return sizeof(*md);
990 }
991 
992 static int fl_set_erspan_opt(const struct nlattr *nla, struct fl_flow_key *key,
993 			     int depth, int option_len,
994 			     struct netlink_ext_ack *extack)
995 {
996 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX + 1];
997 	struct erspan_metadata *md;
998 	int err;
999 
1000 	md = (struct erspan_metadata *)&key->enc_opts.data[key->enc_opts.len];
1001 	memset(md, 0xff, sizeof(*md));
1002 	md->version = 1;
1003 
1004 	if (!depth)
1005 		return sizeof(*md);
1006 
1007 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_ERSPAN) {
1008 		NL_SET_ERR_MSG(extack, "Non-erspan option type for mask");
1009 		return -EINVAL;
1010 	}
1011 
1012 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX, nla,
1013 			       erspan_opt_policy, extack);
1014 	if (err < 0)
1015 		return err;
1016 
1017 	if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]) {
1018 		NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option ver");
1019 		return -EINVAL;
1020 	}
1021 
1022 	if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER])
1023 		md->version = nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]);
1024 
1025 	if (md->version == 1) {
1026 		if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]) {
1027 			NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option index");
1028 			return -EINVAL;
1029 		}
1030 		if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]) {
1031 			nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX];
1032 			md->u.index = nla_get_be32(nla);
1033 		}
1034 	} else if (md->version == 2) {
1035 		if (!option_len && (!tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR] ||
1036 				    !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID])) {
1037 			NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option dir or hwid");
1038 			return -EINVAL;
1039 		}
1040 		if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR]) {
1041 			nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR];
1042 			md->u.md2.dir = nla_get_u8(nla);
1043 		}
1044 		if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID]) {
1045 			nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID];
1046 			set_hwid(&md->u.md2, nla_get_u8(nla));
1047 		}
1048 	} else {
1049 		NL_SET_ERR_MSG(extack, "Tunnel key erspan option ver is incorrect");
1050 		return -EINVAL;
1051 	}
1052 
1053 	return sizeof(*md);
1054 }
1055 
1056 static int fl_set_enc_opt(struct nlattr **tb, struct fl_flow_key *key,
1057 			  struct fl_flow_key *mask,
1058 			  struct netlink_ext_ack *extack)
1059 {
1060 	const struct nlattr *nla_enc_key, *nla_opt_key, *nla_opt_msk = NULL;
1061 	int err, option_len, key_depth, msk_depth = 0;
1062 
1063 	err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS],
1064 					     TCA_FLOWER_KEY_ENC_OPTS_MAX,
1065 					     enc_opts_policy, extack);
1066 	if (err)
1067 		return err;
1068 
1069 	nla_enc_key = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS]);
1070 
1071 	if (tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]) {
1072 		err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK],
1073 						     TCA_FLOWER_KEY_ENC_OPTS_MAX,
1074 						     enc_opts_policy, extack);
1075 		if (err)
1076 			return err;
1077 
1078 		nla_opt_msk = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
1079 		msk_depth = nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
1080 	}
1081 
1082 	nla_for_each_attr(nla_opt_key, nla_enc_key,
1083 			  nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS]), key_depth) {
1084 		switch (nla_type(nla_opt_key)) {
1085 		case TCA_FLOWER_KEY_ENC_OPTS_GENEVE:
1086 			if (key->enc_opts.dst_opt_type &&
1087 			    key->enc_opts.dst_opt_type != TUNNEL_GENEVE_OPT) {
1088 				NL_SET_ERR_MSG(extack, "Duplicate type for geneve options");
1089 				return -EINVAL;
1090 			}
1091 			option_len = 0;
1092 			key->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT;
1093 			option_len = fl_set_geneve_opt(nla_opt_key, key,
1094 						       key_depth, option_len,
1095 						       extack);
1096 			if (option_len < 0)
1097 				return option_len;
1098 
1099 			key->enc_opts.len += option_len;
1100 			/* At the same time we need to parse through the mask
1101 			 * in order to verify exact and mask attribute lengths.
1102 			 */
1103 			mask->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT;
1104 			option_len = fl_set_geneve_opt(nla_opt_msk, mask,
1105 						       msk_depth, option_len,
1106 						       extack);
1107 			if (option_len < 0)
1108 				return option_len;
1109 
1110 			mask->enc_opts.len += option_len;
1111 			if (key->enc_opts.len != mask->enc_opts.len) {
1112 				NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1113 				return -EINVAL;
1114 			}
1115 
1116 			if (msk_depth)
1117 				nla_opt_msk = nla_next(nla_opt_msk, &msk_depth);
1118 			break;
1119 		case TCA_FLOWER_KEY_ENC_OPTS_VXLAN:
1120 			if (key->enc_opts.dst_opt_type) {
1121 				NL_SET_ERR_MSG(extack, "Duplicate type for vxlan options");
1122 				return -EINVAL;
1123 			}
1124 			option_len = 0;
1125 			key->enc_opts.dst_opt_type = TUNNEL_VXLAN_OPT;
1126 			option_len = fl_set_vxlan_opt(nla_opt_key, key,
1127 						      key_depth, option_len,
1128 						      extack);
1129 			if (option_len < 0)
1130 				return option_len;
1131 
1132 			key->enc_opts.len += option_len;
1133 			/* At the same time we need to parse through the mask
1134 			 * in order to verify exact and mask attribute lengths.
1135 			 */
1136 			mask->enc_opts.dst_opt_type = TUNNEL_VXLAN_OPT;
1137 			option_len = fl_set_vxlan_opt(nla_opt_msk, mask,
1138 						      msk_depth, option_len,
1139 						      extack);
1140 			if (option_len < 0)
1141 				return option_len;
1142 
1143 			mask->enc_opts.len += option_len;
1144 			if (key->enc_opts.len != mask->enc_opts.len) {
1145 				NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1146 				return -EINVAL;
1147 			}
1148 
1149 			if (msk_depth)
1150 				nla_opt_msk = nla_next(nla_opt_msk, &msk_depth);
1151 			break;
1152 		case TCA_FLOWER_KEY_ENC_OPTS_ERSPAN:
1153 			if (key->enc_opts.dst_opt_type) {
1154 				NL_SET_ERR_MSG(extack, "Duplicate type for erspan options");
1155 				return -EINVAL;
1156 			}
1157 			option_len = 0;
1158 			key->enc_opts.dst_opt_type = TUNNEL_ERSPAN_OPT;
1159 			option_len = fl_set_erspan_opt(nla_opt_key, key,
1160 						       key_depth, option_len,
1161 						       extack);
1162 			if (option_len < 0)
1163 				return option_len;
1164 
1165 			key->enc_opts.len += option_len;
1166 			/* At the same time we need to parse through the mask
1167 			 * in order to verify exact and mask attribute lengths.
1168 			 */
1169 			mask->enc_opts.dst_opt_type = TUNNEL_ERSPAN_OPT;
1170 			option_len = fl_set_erspan_opt(nla_opt_msk, mask,
1171 						       msk_depth, option_len,
1172 						       extack);
1173 			if (option_len < 0)
1174 				return option_len;
1175 
1176 			mask->enc_opts.len += option_len;
1177 			if (key->enc_opts.len != mask->enc_opts.len) {
1178 				NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1179 				return -EINVAL;
1180 			}
1181 
1182 			if (msk_depth)
1183 				nla_opt_msk = nla_next(nla_opt_msk, &msk_depth);
1184 			break;
1185 		default:
1186 			NL_SET_ERR_MSG(extack, "Unknown tunnel option type");
1187 			return -EINVAL;
1188 		}
1189 	}
1190 
1191 	return 0;
1192 }
1193 
1194 static int fl_set_key_ct(struct nlattr **tb,
1195 			 struct flow_dissector_key_ct *key,
1196 			 struct flow_dissector_key_ct *mask,
1197 			 struct netlink_ext_ack *extack)
1198 {
1199 	if (tb[TCA_FLOWER_KEY_CT_STATE]) {
1200 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK)) {
1201 			NL_SET_ERR_MSG(extack, "Conntrack isn't enabled");
1202 			return -EOPNOTSUPP;
1203 		}
1204 		fl_set_key_val(tb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE,
1205 			       &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK,
1206 			       sizeof(key->ct_state));
1207 	}
1208 	if (tb[TCA_FLOWER_KEY_CT_ZONE]) {
1209 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)) {
1210 			NL_SET_ERR_MSG(extack, "Conntrack zones isn't enabled");
1211 			return -EOPNOTSUPP;
1212 		}
1213 		fl_set_key_val(tb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE,
1214 			       &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK,
1215 			       sizeof(key->ct_zone));
1216 	}
1217 	if (tb[TCA_FLOWER_KEY_CT_MARK]) {
1218 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)) {
1219 			NL_SET_ERR_MSG(extack, "Conntrack mark isn't enabled");
1220 			return -EOPNOTSUPP;
1221 		}
1222 		fl_set_key_val(tb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK,
1223 			       &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK,
1224 			       sizeof(key->ct_mark));
1225 	}
1226 	if (tb[TCA_FLOWER_KEY_CT_LABELS]) {
1227 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)) {
1228 			NL_SET_ERR_MSG(extack, "Conntrack labels aren't enabled");
1229 			return -EOPNOTSUPP;
1230 		}
1231 		fl_set_key_val(tb, key->ct_labels, TCA_FLOWER_KEY_CT_LABELS,
1232 			       mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK,
1233 			       sizeof(key->ct_labels));
1234 	}
1235 
1236 	return 0;
1237 }
1238 
1239 static int fl_set_key(struct net *net, struct nlattr **tb,
1240 		      struct fl_flow_key *key, struct fl_flow_key *mask,
1241 		      struct netlink_ext_ack *extack)
1242 {
1243 	__be16 ethertype;
1244 	int ret = 0;
1245 
1246 	if (tb[TCA_FLOWER_INDEV]) {
1247 		int err = tcf_change_indev(net, tb[TCA_FLOWER_INDEV], extack);
1248 		if (err < 0)
1249 			return err;
1250 		key->meta.ingress_ifindex = err;
1251 		mask->meta.ingress_ifindex = 0xffffffff;
1252 	}
1253 
1254 	fl_set_key_val(tb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
1255 		       mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
1256 		       sizeof(key->eth.dst));
1257 	fl_set_key_val(tb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
1258 		       mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
1259 		       sizeof(key->eth.src));
1260 
1261 	if (tb[TCA_FLOWER_KEY_ETH_TYPE]) {
1262 		ethertype = nla_get_be16(tb[TCA_FLOWER_KEY_ETH_TYPE]);
1263 
1264 		if (eth_type_vlan(ethertype)) {
1265 			fl_set_key_vlan(tb, ethertype, TCA_FLOWER_KEY_VLAN_ID,
1266 					TCA_FLOWER_KEY_VLAN_PRIO, &key->vlan,
1267 					&mask->vlan);
1268 
1269 			if (tb[TCA_FLOWER_KEY_VLAN_ETH_TYPE]) {
1270 				ethertype = nla_get_be16(tb[TCA_FLOWER_KEY_VLAN_ETH_TYPE]);
1271 				if (eth_type_vlan(ethertype)) {
1272 					fl_set_key_vlan(tb, ethertype,
1273 							TCA_FLOWER_KEY_CVLAN_ID,
1274 							TCA_FLOWER_KEY_CVLAN_PRIO,
1275 							&key->cvlan, &mask->cvlan);
1276 					fl_set_key_val(tb, &key->basic.n_proto,
1277 						       TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
1278 						       &mask->basic.n_proto,
1279 						       TCA_FLOWER_UNSPEC,
1280 						       sizeof(key->basic.n_proto));
1281 				} else {
1282 					key->basic.n_proto = ethertype;
1283 					mask->basic.n_proto = cpu_to_be16(~0);
1284 				}
1285 			}
1286 		} else {
1287 			key->basic.n_proto = ethertype;
1288 			mask->basic.n_proto = cpu_to_be16(~0);
1289 		}
1290 	}
1291 
1292 	if (key->basic.n_proto == htons(ETH_P_IP) ||
1293 	    key->basic.n_proto == htons(ETH_P_IPV6)) {
1294 		fl_set_key_val(tb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
1295 			       &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
1296 			       sizeof(key->basic.ip_proto));
1297 		fl_set_key_ip(tb, false, &key->ip, &mask->ip);
1298 	}
1299 
1300 	if (tb[TCA_FLOWER_KEY_IPV4_SRC] || tb[TCA_FLOWER_KEY_IPV4_DST]) {
1301 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1302 		mask->control.addr_type = ~0;
1303 		fl_set_key_val(tb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
1304 			       &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
1305 			       sizeof(key->ipv4.src));
1306 		fl_set_key_val(tb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
1307 			       &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
1308 			       sizeof(key->ipv4.dst));
1309 	} else if (tb[TCA_FLOWER_KEY_IPV6_SRC] || tb[TCA_FLOWER_KEY_IPV6_DST]) {
1310 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1311 		mask->control.addr_type = ~0;
1312 		fl_set_key_val(tb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
1313 			       &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
1314 			       sizeof(key->ipv6.src));
1315 		fl_set_key_val(tb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
1316 			       &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
1317 			       sizeof(key->ipv6.dst));
1318 	}
1319 
1320 	if (key->basic.ip_proto == IPPROTO_TCP) {
1321 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
1322 			       &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
1323 			       sizeof(key->tp.src));
1324 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
1325 			       &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
1326 			       sizeof(key->tp.dst));
1327 		fl_set_key_val(tb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
1328 			       &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
1329 			       sizeof(key->tcp.flags));
1330 	} else if (key->basic.ip_proto == IPPROTO_UDP) {
1331 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
1332 			       &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
1333 			       sizeof(key->tp.src));
1334 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
1335 			       &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
1336 			       sizeof(key->tp.dst));
1337 	} else if (key->basic.ip_proto == IPPROTO_SCTP) {
1338 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
1339 			       &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
1340 			       sizeof(key->tp.src));
1341 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
1342 			       &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
1343 			       sizeof(key->tp.dst));
1344 	} else if (key->basic.n_proto == htons(ETH_P_IP) &&
1345 		   key->basic.ip_proto == IPPROTO_ICMP) {
1346 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV4_TYPE,
1347 			       &mask->icmp.type,
1348 			       TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
1349 			       sizeof(key->icmp.type));
1350 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV4_CODE,
1351 			       &mask->icmp.code,
1352 			       TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
1353 			       sizeof(key->icmp.code));
1354 	} else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
1355 		   key->basic.ip_proto == IPPROTO_ICMPV6) {
1356 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV6_TYPE,
1357 			       &mask->icmp.type,
1358 			       TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
1359 			       sizeof(key->icmp.type));
1360 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV6_CODE,
1361 			       &mask->icmp.code,
1362 			       TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
1363 			       sizeof(key->icmp.code));
1364 	} else if (key->basic.n_proto == htons(ETH_P_MPLS_UC) ||
1365 		   key->basic.n_proto == htons(ETH_P_MPLS_MC)) {
1366 		ret = fl_set_key_mpls(tb, &key->mpls, &mask->mpls);
1367 		if (ret)
1368 			return ret;
1369 	} else if (key->basic.n_proto == htons(ETH_P_ARP) ||
1370 		   key->basic.n_proto == htons(ETH_P_RARP)) {
1371 		fl_set_key_val(tb, &key->arp.sip, TCA_FLOWER_KEY_ARP_SIP,
1372 			       &mask->arp.sip, TCA_FLOWER_KEY_ARP_SIP_MASK,
1373 			       sizeof(key->arp.sip));
1374 		fl_set_key_val(tb, &key->arp.tip, TCA_FLOWER_KEY_ARP_TIP,
1375 			       &mask->arp.tip, TCA_FLOWER_KEY_ARP_TIP_MASK,
1376 			       sizeof(key->arp.tip));
1377 		fl_set_key_val(tb, &key->arp.op, TCA_FLOWER_KEY_ARP_OP,
1378 			       &mask->arp.op, TCA_FLOWER_KEY_ARP_OP_MASK,
1379 			       sizeof(key->arp.op));
1380 		fl_set_key_val(tb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
1381 			       mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
1382 			       sizeof(key->arp.sha));
1383 		fl_set_key_val(tb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
1384 			       mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
1385 			       sizeof(key->arp.tha));
1386 	}
1387 
1388 	if (key->basic.ip_proto == IPPROTO_TCP ||
1389 	    key->basic.ip_proto == IPPROTO_UDP ||
1390 	    key->basic.ip_proto == IPPROTO_SCTP) {
1391 		ret = fl_set_key_port_range(tb, key, mask);
1392 		if (ret)
1393 			return ret;
1394 	}
1395 
1396 	if (tb[TCA_FLOWER_KEY_ENC_IPV4_SRC] ||
1397 	    tb[TCA_FLOWER_KEY_ENC_IPV4_DST]) {
1398 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1399 		mask->enc_control.addr_type = ~0;
1400 		fl_set_key_val(tb, &key->enc_ipv4.src,
1401 			       TCA_FLOWER_KEY_ENC_IPV4_SRC,
1402 			       &mask->enc_ipv4.src,
1403 			       TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
1404 			       sizeof(key->enc_ipv4.src));
1405 		fl_set_key_val(tb, &key->enc_ipv4.dst,
1406 			       TCA_FLOWER_KEY_ENC_IPV4_DST,
1407 			       &mask->enc_ipv4.dst,
1408 			       TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
1409 			       sizeof(key->enc_ipv4.dst));
1410 	}
1411 
1412 	if (tb[TCA_FLOWER_KEY_ENC_IPV6_SRC] ||
1413 	    tb[TCA_FLOWER_KEY_ENC_IPV6_DST]) {
1414 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1415 		mask->enc_control.addr_type = ~0;
1416 		fl_set_key_val(tb, &key->enc_ipv6.src,
1417 			       TCA_FLOWER_KEY_ENC_IPV6_SRC,
1418 			       &mask->enc_ipv6.src,
1419 			       TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
1420 			       sizeof(key->enc_ipv6.src));
1421 		fl_set_key_val(tb, &key->enc_ipv6.dst,
1422 			       TCA_FLOWER_KEY_ENC_IPV6_DST,
1423 			       &mask->enc_ipv6.dst,
1424 			       TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
1425 			       sizeof(key->enc_ipv6.dst));
1426 	}
1427 
1428 	fl_set_key_val(tb, &key->enc_key_id.keyid, TCA_FLOWER_KEY_ENC_KEY_ID,
1429 		       &mask->enc_key_id.keyid, TCA_FLOWER_UNSPEC,
1430 		       sizeof(key->enc_key_id.keyid));
1431 
1432 	fl_set_key_val(tb, &key->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
1433 		       &mask->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
1434 		       sizeof(key->enc_tp.src));
1435 
1436 	fl_set_key_val(tb, &key->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
1437 		       &mask->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
1438 		       sizeof(key->enc_tp.dst));
1439 
1440 	fl_set_key_ip(tb, true, &key->enc_ip, &mask->enc_ip);
1441 
1442 	if (tb[TCA_FLOWER_KEY_ENC_OPTS]) {
1443 		ret = fl_set_enc_opt(tb, key, mask, extack);
1444 		if (ret)
1445 			return ret;
1446 	}
1447 
1448 	ret = fl_set_key_ct(tb, &key->ct, &mask->ct, extack);
1449 	if (ret)
1450 		return ret;
1451 
1452 	if (tb[TCA_FLOWER_KEY_FLAGS])
1453 		ret = fl_set_key_flags(tb, &key->control.flags, &mask->control.flags);
1454 
1455 	return ret;
1456 }
1457 
1458 static void fl_mask_copy(struct fl_flow_mask *dst,
1459 			 struct fl_flow_mask *src)
1460 {
1461 	const void *psrc = fl_key_get_start(&src->key, src);
1462 	void *pdst = fl_key_get_start(&dst->key, src);
1463 
1464 	memcpy(pdst, psrc, fl_mask_range(src));
1465 	dst->range = src->range;
1466 }
1467 
1468 static const struct rhashtable_params fl_ht_params = {
1469 	.key_offset = offsetof(struct cls_fl_filter, mkey), /* base offset */
1470 	.head_offset = offsetof(struct cls_fl_filter, ht_node),
1471 	.automatic_shrinking = true,
1472 };
1473 
1474 static int fl_init_mask_hashtable(struct fl_flow_mask *mask)
1475 {
1476 	mask->filter_ht_params = fl_ht_params;
1477 	mask->filter_ht_params.key_len = fl_mask_range(mask);
1478 	mask->filter_ht_params.key_offset += mask->range.start;
1479 
1480 	return rhashtable_init(&mask->ht, &mask->filter_ht_params);
1481 }
1482 
1483 #define FL_KEY_MEMBER_OFFSET(member) offsetof(struct fl_flow_key, member)
1484 #define FL_KEY_MEMBER_SIZE(member) sizeof_field(struct fl_flow_key, member)
1485 
1486 #define FL_KEY_IS_MASKED(mask, member)						\
1487 	memchr_inv(((char *)mask) + FL_KEY_MEMBER_OFFSET(member),		\
1488 		   0, FL_KEY_MEMBER_SIZE(member))				\
1489 
1490 #define FL_KEY_SET(keys, cnt, id, member)					\
1491 	do {									\
1492 		keys[cnt].key_id = id;						\
1493 		keys[cnt].offset = FL_KEY_MEMBER_OFFSET(member);		\
1494 		cnt++;								\
1495 	} while(0);
1496 
1497 #define FL_KEY_SET_IF_MASKED(mask, keys, cnt, id, member)			\
1498 	do {									\
1499 		if (FL_KEY_IS_MASKED(mask, member))				\
1500 			FL_KEY_SET(keys, cnt, id, member);			\
1501 	} while(0);
1502 
1503 static void fl_init_dissector(struct flow_dissector *dissector,
1504 			      struct fl_flow_key *mask)
1505 {
1506 	struct flow_dissector_key keys[FLOW_DISSECTOR_KEY_MAX];
1507 	size_t cnt = 0;
1508 
1509 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1510 			     FLOW_DISSECTOR_KEY_META, meta);
1511 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_CONTROL, control);
1512 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_BASIC, basic);
1513 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1514 			     FLOW_DISSECTOR_KEY_ETH_ADDRS, eth);
1515 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1516 			     FLOW_DISSECTOR_KEY_IPV4_ADDRS, ipv4);
1517 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1518 			     FLOW_DISSECTOR_KEY_IPV6_ADDRS, ipv6);
1519 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1520 			     FLOW_DISSECTOR_KEY_PORTS, tp);
1521 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1522 			     FLOW_DISSECTOR_KEY_PORTS_RANGE, tp_range);
1523 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1524 			     FLOW_DISSECTOR_KEY_IP, ip);
1525 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1526 			     FLOW_DISSECTOR_KEY_TCP, tcp);
1527 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1528 			     FLOW_DISSECTOR_KEY_ICMP, icmp);
1529 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1530 			     FLOW_DISSECTOR_KEY_ARP, arp);
1531 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1532 			     FLOW_DISSECTOR_KEY_MPLS, mpls);
1533 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1534 			     FLOW_DISSECTOR_KEY_VLAN, vlan);
1535 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1536 			     FLOW_DISSECTOR_KEY_CVLAN, cvlan);
1537 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1538 			     FLOW_DISSECTOR_KEY_ENC_KEYID, enc_key_id);
1539 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1540 			     FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, enc_ipv4);
1541 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1542 			     FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, enc_ipv6);
1543 	if (FL_KEY_IS_MASKED(mask, enc_ipv4) ||
1544 	    FL_KEY_IS_MASKED(mask, enc_ipv6))
1545 		FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_ENC_CONTROL,
1546 			   enc_control);
1547 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1548 			     FLOW_DISSECTOR_KEY_ENC_PORTS, enc_tp);
1549 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1550 			     FLOW_DISSECTOR_KEY_ENC_IP, enc_ip);
1551 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1552 			     FLOW_DISSECTOR_KEY_ENC_OPTS, enc_opts);
1553 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1554 			     FLOW_DISSECTOR_KEY_CT, ct);
1555 
1556 	skb_flow_dissector_init(dissector, keys, cnt);
1557 }
1558 
1559 static struct fl_flow_mask *fl_create_new_mask(struct cls_fl_head *head,
1560 					       struct fl_flow_mask *mask)
1561 {
1562 	struct fl_flow_mask *newmask;
1563 	int err;
1564 
1565 	newmask = kzalloc(sizeof(*newmask), GFP_KERNEL);
1566 	if (!newmask)
1567 		return ERR_PTR(-ENOMEM);
1568 
1569 	fl_mask_copy(newmask, mask);
1570 
1571 	if ((newmask->key.tp_range.tp_min.dst &&
1572 	     newmask->key.tp_range.tp_max.dst) ||
1573 	    (newmask->key.tp_range.tp_min.src &&
1574 	     newmask->key.tp_range.tp_max.src))
1575 		newmask->flags |= TCA_FLOWER_MASK_FLAGS_RANGE;
1576 
1577 	err = fl_init_mask_hashtable(newmask);
1578 	if (err)
1579 		goto errout_free;
1580 
1581 	fl_init_dissector(&newmask->dissector, &newmask->key);
1582 
1583 	INIT_LIST_HEAD_RCU(&newmask->filters);
1584 
1585 	refcount_set(&newmask->refcnt, 1);
1586 	err = rhashtable_replace_fast(&head->ht, &mask->ht_node,
1587 				      &newmask->ht_node, mask_ht_params);
1588 	if (err)
1589 		goto errout_destroy;
1590 
1591 	spin_lock(&head->masks_lock);
1592 	list_add_tail_rcu(&newmask->list, &head->masks);
1593 	spin_unlock(&head->masks_lock);
1594 
1595 	return newmask;
1596 
1597 errout_destroy:
1598 	rhashtable_destroy(&newmask->ht);
1599 errout_free:
1600 	kfree(newmask);
1601 
1602 	return ERR_PTR(err);
1603 }
1604 
1605 static int fl_check_assign_mask(struct cls_fl_head *head,
1606 				struct cls_fl_filter *fnew,
1607 				struct cls_fl_filter *fold,
1608 				struct fl_flow_mask *mask)
1609 {
1610 	struct fl_flow_mask *newmask;
1611 	int ret = 0;
1612 
1613 	rcu_read_lock();
1614 
1615 	/* Insert mask as temporary node to prevent concurrent creation of mask
1616 	 * with same key. Any concurrent lookups with same key will return
1617 	 * -EAGAIN because mask's refcnt is zero.
1618 	 */
1619 	fnew->mask = rhashtable_lookup_get_insert_fast(&head->ht,
1620 						       &mask->ht_node,
1621 						       mask_ht_params);
1622 	if (!fnew->mask) {
1623 		rcu_read_unlock();
1624 
1625 		if (fold) {
1626 			ret = -EINVAL;
1627 			goto errout_cleanup;
1628 		}
1629 
1630 		newmask = fl_create_new_mask(head, mask);
1631 		if (IS_ERR(newmask)) {
1632 			ret = PTR_ERR(newmask);
1633 			goto errout_cleanup;
1634 		}
1635 
1636 		fnew->mask = newmask;
1637 		return 0;
1638 	} else if (IS_ERR(fnew->mask)) {
1639 		ret = PTR_ERR(fnew->mask);
1640 	} else if (fold && fold->mask != fnew->mask) {
1641 		ret = -EINVAL;
1642 	} else if (!refcount_inc_not_zero(&fnew->mask->refcnt)) {
1643 		/* Mask was deleted concurrently, try again */
1644 		ret = -EAGAIN;
1645 	}
1646 	rcu_read_unlock();
1647 	return ret;
1648 
1649 errout_cleanup:
1650 	rhashtable_remove_fast(&head->ht, &mask->ht_node,
1651 			       mask_ht_params);
1652 	return ret;
1653 }
1654 
1655 static int fl_set_parms(struct net *net, struct tcf_proto *tp,
1656 			struct cls_fl_filter *f, struct fl_flow_mask *mask,
1657 			unsigned long base, struct nlattr **tb,
1658 			struct nlattr *est, bool ovr,
1659 			struct fl_flow_tmplt *tmplt, bool rtnl_held,
1660 			struct netlink_ext_ack *extack)
1661 {
1662 	int err;
1663 
1664 	err = tcf_exts_validate(net, tp, tb, est, &f->exts, ovr, rtnl_held,
1665 				extack);
1666 	if (err < 0)
1667 		return err;
1668 
1669 	if (tb[TCA_FLOWER_CLASSID]) {
1670 		f->res.classid = nla_get_u32(tb[TCA_FLOWER_CLASSID]);
1671 		if (!rtnl_held)
1672 			rtnl_lock();
1673 		tcf_bind_filter(tp, &f->res, base);
1674 		if (!rtnl_held)
1675 			rtnl_unlock();
1676 	}
1677 
1678 	err = fl_set_key(net, tb, &f->key, &mask->key, extack);
1679 	if (err)
1680 		return err;
1681 
1682 	fl_mask_update_range(mask);
1683 	fl_set_masked_key(&f->mkey, &f->key, mask);
1684 
1685 	if (!fl_mask_fits_tmplt(tmplt, mask)) {
1686 		NL_SET_ERR_MSG_MOD(extack, "Mask does not fit the template");
1687 		return -EINVAL;
1688 	}
1689 
1690 	return 0;
1691 }
1692 
1693 static int fl_ht_insert_unique(struct cls_fl_filter *fnew,
1694 			       struct cls_fl_filter *fold,
1695 			       bool *in_ht)
1696 {
1697 	struct fl_flow_mask *mask = fnew->mask;
1698 	int err;
1699 
1700 	err = rhashtable_lookup_insert_fast(&mask->ht,
1701 					    &fnew->ht_node,
1702 					    mask->filter_ht_params);
1703 	if (err) {
1704 		*in_ht = false;
1705 		/* It is okay if filter with same key exists when
1706 		 * overwriting.
1707 		 */
1708 		return fold && err == -EEXIST ? 0 : err;
1709 	}
1710 
1711 	*in_ht = true;
1712 	return 0;
1713 }
1714 
1715 static int fl_change(struct net *net, struct sk_buff *in_skb,
1716 		     struct tcf_proto *tp, unsigned long base,
1717 		     u32 handle, struct nlattr **tca,
1718 		     void **arg, bool ovr, bool rtnl_held,
1719 		     struct netlink_ext_ack *extack)
1720 {
1721 	struct cls_fl_head *head = fl_head_dereference(tp);
1722 	struct cls_fl_filter *fold = *arg;
1723 	struct cls_fl_filter *fnew;
1724 	struct fl_flow_mask *mask;
1725 	struct nlattr **tb;
1726 	bool in_ht;
1727 	int err;
1728 
1729 	if (!tca[TCA_OPTIONS]) {
1730 		err = -EINVAL;
1731 		goto errout_fold;
1732 	}
1733 
1734 	mask = kzalloc(sizeof(struct fl_flow_mask), GFP_KERNEL);
1735 	if (!mask) {
1736 		err = -ENOBUFS;
1737 		goto errout_fold;
1738 	}
1739 
1740 	tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
1741 	if (!tb) {
1742 		err = -ENOBUFS;
1743 		goto errout_mask_alloc;
1744 	}
1745 
1746 	err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
1747 					  tca[TCA_OPTIONS], fl_policy, NULL);
1748 	if (err < 0)
1749 		goto errout_tb;
1750 
1751 	if (fold && handle && fold->handle != handle) {
1752 		err = -EINVAL;
1753 		goto errout_tb;
1754 	}
1755 
1756 	fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
1757 	if (!fnew) {
1758 		err = -ENOBUFS;
1759 		goto errout_tb;
1760 	}
1761 	INIT_LIST_HEAD(&fnew->hw_list);
1762 	refcount_set(&fnew->refcnt, 1);
1763 
1764 	err = tcf_exts_init(&fnew->exts, net, TCA_FLOWER_ACT, 0);
1765 	if (err < 0)
1766 		goto errout;
1767 
1768 	if (tb[TCA_FLOWER_FLAGS]) {
1769 		fnew->flags = nla_get_u32(tb[TCA_FLOWER_FLAGS]);
1770 
1771 		if (!tc_flags_valid(fnew->flags)) {
1772 			err = -EINVAL;
1773 			goto errout;
1774 		}
1775 	}
1776 
1777 	err = fl_set_parms(net, tp, fnew, mask, base, tb, tca[TCA_RATE], ovr,
1778 			   tp->chain->tmplt_priv, rtnl_held, extack);
1779 	if (err)
1780 		goto errout;
1781 
1782 	err = fl_check_assign_mask(head, fnew, fold, mask);
1783 	if (err)
1784 		goto errout;
1785 
1786 	err = fl_ht_insert_unique(fnew, fold, &in_ht);
1787 	if (err)
1788 		goto errout_mask;
1789 
1790 	if (!tc_skip_hw(fnew->flags)) {
1791 		err = fl_hw_replace_filter(tp, fnew, rtnl_held, extack);
1792 		if (err)
1793 			goto errout_ht;
1794 	}
1795 
1796 	if (!tc_in_hw(fnew->flags))
1797 		fnew->flags |= TCA_CLS_FLAGS_NOT_IN_HW;
1798 
1799 	spin_lock(&tp->lock);
1800 
1801 	/* tp was deleted concurrently. -EAGAIN will cause caller to lookup
1802 	 * proto again or create new one, if necessary.
1803 	 */
1804 	if (tp->deleting) {
1805 		err = -EAGAIN;
1806 		goto errout_hw;
1807 	}
1808 
1809 	if (fold) {
1810 		/* Fold filter was deleted concurrently. Retry lookup. */
1811 		if (fold->deleted) {
1812 			err = -EAGAIN;
1813 			goto errout_hw;
1814 		}
1815 
1816 		fnew->handle = handle;
1817 
1818 		if (!in_ht) {
1819 			struct rhashtable_params params =
1820 				fnew->mask->filter_ht_params;
1821 
1822 			err = rhashtable_insert_fast(&fnew->mask->ht,
1823 						     &fnew->ht_node,
1824 						     params);
1825 			if (err)
1826 				goto errout_hw;
1827 			in_ht = true;
1828 		}
1829 
1830 		refcount_inc(&fnew->refcnt);
1831 		rhashtable_remove_fast(&fold->mask->ht,
1832 				       &fold->ht_node,
1833 				       fold->mask->filter_ht_params);
1834 		idr_replace(&head->handle_idr, fnew, fnew->handle);
1835 		list_replace_rcu(&fold->list, &fnew->list);
1836 		fold->deleted = true;
1837 
1838 		spin_unlock(&tp->lock);
1839 
1840 		fl_mask_put(head, fold->mask);
1841 		if (!tc_skip_hw(fold->flags))
1842 			fl_hw_destroy_filter(tp, fold, rtnl_held, NULL);
1843 		tcf_unbind_filter(tp, &fold->res);
1844 		/* Caller holds reference to fold, so refcnt is always > 0
1845 		 * after this.
1846 		 */
1847 		refcount_dec(&fold->refcnt);
1848 		__fl_put(fold);
1849 	} else {
1850 		if (handle) {
1851 			/* user specifies a handle and it doesn't exist */
1852 			err = idr_alloc_u32(&head->handle_idr, fnew, &handle,
1853 					    handle, GFP_ATOMIC);
1854 
1855 			/* Filter with specified handle was concurrently
1856 			 * inserted after initial check in cls_api. This is not
1857 			 * necessarily an error if NLM_F_EXCL is not set in
1858 			 * message flags. Returning EAGAIN will cause cls_api to
1859 			 * try to update concurrently inserted rule.
1860 			 */
1861 			if (err == -ENOSPC)
1862 				err = -EAGAIN;
1863 		} else {
1864 			handle = 1;
1865 			err = idr_alloc_u32(&head->handle_idr, fnew, &handle,
1866 					    INT_MAX, GFP_ATOMIC);
1867 		}
1868 		if (err)
1869 			goto errout_hw;
1870 
1871 		refcount_inc(&fnew->refcnt);
1872 		fnew->handle = handle;
1873 		list_add_tail_rcu(&fnew->list, &fnew->mask->filters);
1874 		spin_unlock(&tp->lock);
1875 	}
1876 
1877 	*arg = fnew;
1878 
1879 	kfree(tb);
1880 	tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work);
1881 	return 0;
1882 
1883 errout_ht:
1884 	spin_lock(&tp->lock);
1885 errout_hw:
1886 	fnew->deleted = true;
1887 	spin_unlock(&tp->lock);
1888 	if (!tc_skip_hw(fnew->flags))
1889 		fl_hw_destroy_filter(tp, fnew, rtnl_held, NULL);
1890 	if (in_ht)
1891 		rhashtable_remove_fast(&fnew->mask->ht, &fnew->ht_node,
1892 				       fnew->mask->filter_ht_params);
1893 errout_mask:
1894 	fl_mask_put(head, fnew->mask);
1895 errout:
1896 	__fl_put(fnew);
1897 errout_tb:
1898 	kfree(tb);
1899 errout_mask_alloc:
1900 	tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work);
1901 errout_fold:
1902 	if (fold)
1903 		__fl_put(fold);
1904 	return err;
1905 }
1906 
1907 static int fl_delete(struct tcf_proto *tp, void *arg, bool *last,
1908 		     bool rtnl_held, struct netlink_ext_ack *extack)
1909 {
1910 	struct cls_fl_head *head = fl_head_dereference(tp);
1911 	struct cls_fl_filter *f = arg;
1912 	bool last_on_mask;
1913 	int err = 0;
1914 
1915 	err = __fl_delete(tp, f, &last_on_mask, rtnl_held, extack);
1916 	*last = list_empty(&head->masks);
1917 	__fl_put(f);
1918 
1919 	return err;
1920 }
1921 
1922 static void fl_walk(struct tcf_proto *tp, struct tcf_walker *arg,
1923 		    bool rtnl_held)
1924 {
1925 	struct cls_fl_head *head = fl_head_dereference(tp);
1926 	unsigned long id = arg->cookie, tmp;
1927 	struct cls_fl_filter *f;
1928 
1929 	arg->count = arg->skip;
1930 
1931 	idr_for_each_entry_continue_ul(&head->handle_idr, f, tmp, id) {
1932 		/* don't return filters that are being deleted */
1933 		if (!refcount_inc_not_zero(&f->refcnt))
1934 			continue;
1935 		if (arg->fn(tp, f, arg) < 0) {
1936 			__fl_put(f);
1937 			arg->stop = 1;
1938 			break;
1939 		}
1940 		__fl_put(f);
1941 		arg->count++;
1942 	}
1943 	arg->cookie = id;
1944 }
1945 
1946 static struct cls_fl_filter *
1947 fl_get_next_hw_filter(struct tcf_proto *tp, struct cls_fl_filter *f, bool add)
1948 {
1949 	struct cls_fl_head *head = fl_head_dereference(tp);
1950 
1951 	spin_lock(&tp->lock);
1952 	if (list_empty(&head->hw_filters)) {
1953 		spin_unlock(&tp->lock);
1954 		return NULL;
1955 	}
1956 
1957 	if (!f)
1958 		f = list_entry(&head->hw_filters, struct cls_fl_filter,
1959 			       hw_list);
1960 	list_for_each_entry_continue(f, &head->hw_filters, hw_list) {
1961 		if (!(add && f->deleted) && refcount_inc_not_zero(&f->refcnt)) {
1962 			spin_unlock(&tp->lock);
1963 			return f;
1964 		}
1965 	}
1966 
1967 	spin_unlock(&tp->lock);
1968 	return NULL;
1969 }
1970 
1971 static int fl_reoffload(struct tcf_proto *tp, bool add, flow_setup_cb_t *cb,
1972 			void *cb_priv, struct netlink_ext_ack *extack)
1973 {
1974 	struct tcf_block *block = tp->chain->block;
1975 	struct flow_cls_offload cls_flower = {};
1976 	struct cls_fl_filter *f = NULL;
1977 	int err;
1978 
1979 	/* hw_filters list can only be changed by hw offload functions after
1980 	 * obtaining rtnl lock. Make sure it is not changed while reoffload is
1981 	 * iterating it.
1982 	 */
1983 	ASSERT_RTNL();
1984 
1985 	while ((f = fl_get_next_hw_filter(tp, f, add))) {
1986 		cls_flower.rule =
1987 			flow_rule_alloc(tcf_exts_num_actions(&f->exts));
1988 		if (!cls_flower.rule) {
1989 			__fl_put(f);
1990 			return -ENOMEM;
1991 		}
1992 
1993 		tc_cls_common_offload_init(&cls_flower.common, tp, f->flags,
1994 					   extack);
1995 		cls_flower.command = add ?
1996 			FLOW_CLS_REPLACE : FLOW_CLS_DESTROY;
1997 		cls_flower.cookie = (unsigned long)f;
1998 		cls_flower.rule->match.dissector = &f->mask->dissector;
1999 		cls_flower.rule->match.mask = &f->mask->key;
2000 		cls_flower.rule->match.key = &f->mkey;
2001 
2002 		err = tc_setup_flow_action(&cls_flower.rule->action, &f->exts,
2003 					   true);
2004 		if (err) {
2005 			kfree(cls_flower.rule);
2006 			if (tc_skip_sw(f->flags)) {
2007 				NL_SET_ERR_MSG_MOD(extack, "Failed to setup flow action");
2008 				__fl_put(f);
2009 				return err;
2010 			}
2011 			goto next_flow;
2012 		}
2013 
2014 		cls_flower.classid = f->res.classid;
2015 
2016 		err = tc_setup_cb_reoffload(block, tp, add, cb,
2017 					    TC_SETUP_CLSFLOWER, &cls_flower,
2018 					    cb_priv, &f->flags,
2019 					    &f->in_hw_count);
2020 		tc_cleanup_flow_action(&cls_flower.rule->action);
2021 		kfree(cls_flower.rule);
2022 
2023 		if (err) {
2024 			__fl_put(f);
2025 			return err;
2026 		}
2027 next_flow:
2028 		__fl_put(f);
2029 	}
2030 
2031 	return 0;
2032 }
2033 
2034 static void fl_hw_add(struct tcf_proto *tp, void *type_data)
2035 {
2036 	struct flow_cls_offload *cls_flower = type_data;
2037 	struct cls_fl_filter *f =
2038 		(struct cls_fl_filter *) cls_flower->cookie;
2039 	struct cls_fl_head *head = fl_head_dereference(tp);
2040 
2041 	spin_lock(&tp->lock);
2042 	list_add(&f->hw_list, &head->hw_filters);
2043 	spin_unlock(&tp->lock);
2044 }
2045 
2046 static void fl_hw_del(struct tcf_proto *tp, void *type_data)
2047 {
2048 	struct flow_cls_offload *cls_flower = type_data;
2049 	struct cls_fl_filter *f =
2050 		(struct cls_fl_filter *) cls_flower->cookie;
2051 
2052 	spin_lock(&tp->lock);
2053 	if (!list_empty(&f->hw_list))
2054 		list_del_init(&f->hw_list);
2055 	spin_unlock(&tp->lock);
2056 }
2057 
2058 static int fl_hw_create_tmplt(struct tcf_chain *chain,
2059 			      struct fl_flow_tmplt *tmplt)
2060 {
2061 	struct flow_cls_offload cls_flower = {};
2062 	struct tcf_block *block = chain->block;
2063 
2064 	cls_flower.rule = flow_rule_alloc(0);
2065 	if (!cls_flower.rule)
2066 		return -ENOMEM;
2067 
2068 	cls_flower.common.chain_index = chain->index;
2069 	cls_flower.command = FLOW_CLS_TMPLT_CREATE;
2070 	cls_flower.cookie = (unsigned long) tmplt;
2071 	cls_flower.rule->match.dissector = &tmplt->dissector;
2072 	cls_flower.rule->match.mask = &tmplt->mask;
2073 	cls_flower.rule->match.key = &tmplt->dummy_key;
2074 
2075 	/* We don't care if driver (any of them) fails to handle this
2076 	 * call. It serves just as a hint for it.
2077 	 */
2078 	tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, true);
2079 	kfree(cls_flower.rule);
2080 
2081 	return 0;
2082 }
2083 
2084 static void fl_hw_destroy_tmplt(struct tcf_chain *chain,
2085 				struct fl_flow_tmplt *tmplt)
2086 {
2087 	struct flow_cls_offload cls_flower = {};
2088 	struct tcf_block *block = chain->block;
2089 
2090 	cls_flower.common.chain_index = chain->index;
2091 	cls_flower.command = FLOW_CLS_TMPLT_DESTROY;
2092 	cls_flower.cookie = (unsigned long) tmplt;
2093 
2094 	tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, true);
2095 }
2096 
2097 static void *fl_tmplt_create(struct net *net, struct tcf_chain *chain,
2098 			     struct nlattr **tca,
2099 			     struct netlink_ext_ack *extack)
2100 {
2101 	struct fl_flow_tmplt *tmplt;
2102 	struct nlattr **tb;
2103 	int err;
2104 
2105 	if (!tca[TCA_OPTIONS])
2106 		return ERR_PTR(-EINVAL);
2107 
2108 	tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
2109 	if (!tb)
2110 		return ERR_PTR(-ENOBUFS);
2111 	err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
2112 					  tca[TCA_OPTIONS], fl_policy, NULL);
2113 	if (err)
2114 		goto errout_tb;
2115 
2116 	tmplt = kzalloc(sizeof(*tmplt), GFP_KERNEL);
2117 	if (!tmplt) {
2118 		err = -ENOMEM;
2119 		goto errout_tb;
2120 	}
2121 	tmplt->chain = chain;
2122 	err = fl_set_key(net, tb, &tmplt->dummy_key, &tmplt->mask, extack);
2123 	if (err)
2124 		goto errout_tmplt;
2125 
2126 	fl_init_dissector(&tmplt->dissector, &tmplt->mask);
2127 
2128 	err = fl_hw_create_tmplt(chain, tmplt);
2129 	if (err)
2130 		goto errout_tmplt;
2131 
2132 	kfree(tb);
2133 	return tmplt;
2134 
2135 errout_tmplt:
2136 	kfree(tmplt);
2137 errout_tb:
2138 	kfree(tb);
2139 	return ERR_PTR(err);
2140 }
2141 
2142 static void fl_tmplt_destroy(void *tmplt_priv)
2143 {
2144 	struct fl_flow_tmplt *tmplt = tmplt_priv;
2145 
2146 	fl_hw_destroy_tmplt(tmplt->chain, tmplt);
2147 	kfree(tmplt);
2148 }
2149 
2150 static int fl_dump_key_val(struct sk_buff *skb,
2151 			   void *val, int val_type,
2152 			   void *mask, int mask_type, int len)
2153 {
2154 	int err;
2155 
2156 	if (!memchr_inv(mask, 0, len))
2157 		return 0;
2158 	err = nla_put(skb, val_type, len, val);
2159 	if (err)
2160 		return err;
2161 	if (mask_type != TCA_FLOWER_UNSPEC) {
2162 		err = nla_put(skb, mask_type, len, mask);
2163 		if (err)
2164 			return err;
2165 	}
2166 	return 0;
2167 }
2168 
2169 static int fl_dump_key_port_range(struct sk_buff *skb, struct fl_flow_key *key,
2170 				  struct fl_flow_key *mask)
2171 {
2172 	if (fl_dump_key_val(skb, &key->tp_range.tp_min.dst,
2173 			    TCA_FLOWER_KEY_PORT_DST_MIN,
2174 			    &mask->tp_range.tp_min.dst, TCA_FLOWER_UNSPEC,
2175 			    sizeof(key->tp_range.tp_min.dst)) ||
2176 	    fl_dump_key_val(skb, &key->tp_range.tp_max.dst,
2177 			    TCA_FLOWER_KEY_PORT_DST_MAX,
2178 			    &mask->tp_range.tp_max.dst, TCA_FLOWER_UNSPEC,
2179 			    sizeof(key->tp_range.tp_max.dst)) ||
2180 	    fl_dump_key_val(skb, &key->tp_range.tp_min.src,
2181 			    TCA_FLOWER_KEY_PORT_SRC_MIN,
2182 			    &mask->tp_range.tp_min.src, TCA_FLOWER_UNSPEC,
2183 			    sizeof(key->tp_range.tp_min.src)) ||
2184 	    fl_dump_key_val(skb, &key->tp_range.tp_max.src,
2185 			    TCA_FLOWER_KEY_PORT_SRC_MAX,
2186 			    &mask->tp_range.tp_max.src, TCA_FLOWER_UNSPEC,
2187 			    sizeof(key->tp_range.tp_max.src)))
2188 		return -1;
2189 
2190 	return 0;
2191 }
2192 
2193 static int fl_dump_key_mpls(struct sk_buff *skb,
2194 			    struct flow_dissector_key_mpls *mpls_key,
2195 			    struct flow_dissector_key_mpls *mpls_mask)
2196 {
2197 	int err;
2198 
2199 	if (!memchr_inv(mpls_mask, 0, sizeof(*mpls_mask)))
2200 		return 0;
2201 	if (mpls_mask->mpls_ttl) {
2202 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TTL,
2203 				 mpls_key->mpls_ttl);
2204 		if (err)
2205 			return err;
2206 	}
2207 	if (mpls_mask->mpls_tc) {
2208 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TC,
2209 				 mpls_key->mpls_tc);
2210 		if (err)
2211 			return err;
2212 	}
2213 	if (mpls_mask->mpls_label) {
2214 		err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_LABEL,
2215 				  mpls_key->mpls_label);
2216 		if (err)
2217 			return err;
2218 	}
2219 	if (mpls_mask->mpls_bos) {
2220 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_BOS,
2221 				 mpls_key->mpls_bos);
2222 		if (err)
2223 			return err;
2224 	}
2225 	return 0;
2226 }
2227 
2228 static int fl_dump_key_ip(struct sk_buff *skb, bool encap,
2229 			  struct flow_dissector_key_ip *key,
2230 			  struct flow_dissector_key_ip *mask)
2231 {
2232 	int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
2233 	int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
2234 	int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
2235 	int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
2236 
2237 	if (fl_dump_key_val(skb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos)) ||
2238 	    fl_dump_key_val(skb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl)))
2239 		return -1;
2240 
2241 	return 0;
2242 }
2243 
2244 static int fl_dump_key_vlan(struct sk_buff *skb,
2245 			    int vlan_id_key, int vlan_prio_key,
2246 			    struct flow_dissector_key_vlan *vlan_key,
2247 			    struct flow_dissector_key_vlan *vlan_mask)
2248 {
2249 	int err;
2250 
2251 	if (!memchr_inv(vlan_mask, 0, sizeof(*vlan_mask)))
2252 		return 0;
2253 	if (vlan_mask->vlan_id) {
2254 		err = nla_put_u16(skb, vlan_id_key,
2255 				  vlan_key->vlan_id);
2256 		if (err)
2257 			return err;
2258 	}
2259 	if (vlan_mask->vlan_priority) {
2260 		err = nla_put_u8(skb, vlan_prio_key,
2261 				 vlan_key->vlan_priority);
2262 		if (err)
2263 			return err;
2264 	}
2265 	return 0;
2266 }
2267 
2268 static void fl_get_key_flag(u32 dissector_key, u32 dissector_mask,
2269 			    u32 *flower_key, u32 *flower_mask,
2270 			    u32 flower_flag_bit, u32 dissector_flag_bit)
2271 {
2272 	if (dissector_mask & dissector_flag_bit) {
2273 		*flower_mask |= flower_flag_bit;
2274 		if (dissector_key & dissector_flag_bit)
2275 			*flower_key |= flower_flag_bit;
2276 	}
2277 }
2278 
2279 static int fl_dump_key_flags(struct sk_buff *skb, u32 flags_key, u32 flags_mask)
2280 {
2281 	u32 key, mask;
2282 	__be32 _key, _mask;
2283 	int err;
2284 
2285 	if (!memchr_inv(&flags_mask, 0, sizeof(flags_mask)))
2286 		return 0;
2287 
2288 	key = 0;
2289 	mask = 0;
2290 
2291 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
2292 			TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
2293 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
2294 			TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
2295 			FLOW_DIS_FIRST_FRAG);
2296 
2297 	_key = cpu_to_be32(key);
2298 	_mask = cpu_to_be32(mask);
2299 
2300 	err = nla_put(skb, TCA_FLOWER_KEY_FLAGS, 4, &_key);
2301 	if (err)
2302 		return err;
2303 
2304 	return nla_put(skb, TCA_FLOWER_KEY_FLAGS_MASK, 4, &_mask);
2305 }
2306 
2307 static int fl_dump_key_geneve_opt(struct sk_buff *skb,
2308 				  struct flow_dissector_key_enc_opts *enc_opts)
2309 {
2310 	struct geneve_opt *opt;
2311 	struct nlattr *nest;
2312 	int opt_off = 0;
2313 
2314 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_GENEVE);
2315 	if (!nest)
2316 		goto nla_put_failure;
2317 
2318 	while (enc_opts->len > opt_off) {
2319 		opt = (struct geneve_opt *)&enc_opts->data[opt_off];
2320 
2321 		if (nla_put_be16(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS,
2322 				 opt->opt_class))
2323 			goto nla_put_failure;
2324 		if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE,
2325 			       opt->type))
2326 			goto nla_put_failure;
2327 		if (nla_put(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA,
2328 			    opt->length * 4, opt->opt_data))
2329 			goto nla_put_failure;
2330 
2331 		opt_off += sizeof(struct geneve_opt) + opt->length * 4;
2332 	}
2333 	nla_nest_end(skb, nest);
2334 	return 0;
2335 
2336 nla_put_failure:
2337 	nla_nest_cancel(skb, nest);
2338 	return -EMSGSIZE;
2339 }
2340 
2341 static int fl_dump_key_vxlan_opt(struct sk_buff *skb,
2342 				 struct flow_dissector_key_enc_opts *enc_opts)
2343 {
2344 	struct vxlan_metadata *md;
2345 	struct nlattr *nest;
2346 
2347 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_VXLAN);
2348 	if (!nest)
2349 		goto nla_put_failure;
2350 
2351 	md = (struct vxlan_metadata *)&enc_opts->data[0];
2352 	if (nla_put_u32(skb, TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP, md->gbp))
2353 		goto nla_put_failure;
2354 
2355 	nla_nest_end(skb, nest);
2356 	return 0;
2357 
2358 nla_put_failure:
2359 	nla_nest_cancel(skb, nest);
2360 	return -EMSGSIZE;
2361 }
2362 
2363 static int fl_dump_key_erspan_opt(struct sk_buff *skb,
2364 				  struct flow_dissector_key_enc_opts *enc_opts)
2365 {
2366 	struct erspan_metadata *md;
2367 	struct nlattr *nest;
2368 
2369 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_ERSPAN);
2370 	if (!nest)
2371 		goto nla_put_failure;
2372 
2373 	md = (struct erspan_metadata *)&enc_opts->data[0];
2374 	if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER, md->version))
2375 		goto nla_put_failure;
2376 
2377 	if (md->version == 1 &&
2378 	    nla_put_be32(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX, md->u.index))
2379 		goto nla_put_failure;
2380 
2381 	if (md->version == 2 &&
2382 	    (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR,
2383 			md->u.md2.dir) ||
2384 	     nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID,
2385 			get_hwid(&md->u.md2))))
2386 		goto nla_put_failure;
2387 
2388 	nla_nest_end(skb, nest);
2389 	return 0;
2390 
2391 nla_put_failure:
2392 	nla_nest_cancel(skb, nest);
2393 	return -EMSGSIZE;
2394 }
2395 
2396 static int fl_dump_key_ct(struct sk_buff *skb,
2397 			  struct flow_dissector_key_ct *key,
2398 			  struct flow_dissector_key_ct *mask)
2399 {
2400 	if (IS_ENABLED(CONFIG_NF_CONNTRACK) &&
2401 	    fl_dump_key_val(skb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE,
2402 			    &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK,
2403 			    sizeof(key->ct_state)))
2404 		goto nla_put_failure;
2405 
2406 	if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
2407 	    fl_dump_key_val(skb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE,
2408 			    &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK,
2409 			    sizeof(key->ct_zone)))
2410 		goto nla_put_failure;
2411 
2412 	if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
2413 	    fl_dump_key_val(skb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK,
2414 			    &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK,
2415 			    sizeof(key->ct_mark)))
2416 		goto nla_put_failure;
2417 
2418 	if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
2419 	    fl_dump_key_val(skb, &key->ct_labels, TCA_FLOWER_KEY_CT_LABELS,
2420 			    &mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK,
2421 			    sizeof(key->ct_labels)))
2422 		goto nla_put_failure;
2423 
2424 	return 0;
2425 
2426 nla_put_failure:
2427 	return -EMSGSIZE;
2428 }
2429 
2430 static int fl_dump_key_options(struct sk_buff *skb, int enc_opt_type,
2431 			       struct flow_dissector_key_enc_opts *enc_opts)
2432 {
2433 	struct nlattr *nest;
2434 	int err;
2435 
2436 	if (!enc_opts->len)
2437 		return 0;
2438 
2439 	nest = nla_nest_start_noflag(skb, enc_opt_type);
2440 	if (!nest)
2441 		goto nla_put_failure;
2442 
2443 	switch (enc_opts->dst_opt_type) {
2444 	case TUNNEL_GENEVE_OPT:
2445 		err = fl_dump_key_geneve_opt(skb, enc_opts);
2446 		if (err)
2447 			goto nla_put_failure;
2448 		break;
2449 	case TUNNEL_VXLAN_OPT:
2450 		err = fl_dump_key_vxlan_opt(skb, enc_opts);
2451 		if (err)
2452 			goto nla_put_failure;
2453 		break;
2454 	case TUNNEL_ERSPAN_OPT:
2455 		err = fl_dump_key_erspan_opt(skb, enc_opts);
2456 		if (err)
2457 			goto nla_put_failure;
2458 		break;
2459 	default:
2460 		goto nla_put_failure;
2461 	}
2462 	nla_nest_end(skb, nest);
2463 	return 0;
2464 
2465 nla_put_failure:
2466 	nla_nest_cancel(skb, nest);
2467 	return -EMSGSIZE;
2468 }
2469 
2470 static int fl_dump_key_enc_opt(struct sk_buff *skb,
2471 			       struct flow_dissector_key_enc_opts *key_opts,
2472 			       struct flow_dissector_key_enc_opts *msk_opts)
2473 {
2474 	int err;
2475 
2476 	err = fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS, key_opts);
2477 	if (err)
2478 		return err;
2479 
2480 	return fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS_MASK, msk_opts);
2481 }
2482 
2483 static int fl_dump_key(struct sk_buff *skb, struct net *net,
2484 		       struct fl_flow_key *key, struct fl_flow_key *mask)
2485 {
2486 	if (mask->meta.ingress_ifindex) {
2487 		struct net_device *dev;
2488 
2489 		dev = __dev_get_by_index(net, key->meta.ingress_ifindex);
2490 		if (dev && nla_put_string(skb, TCA_FLOWER_INDEV, dev->name))
2491 			goto nla_put_failure;
2492 	}
2493 
2494 	if (fl_dump_key_val(skb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
2495 			    mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
2496 			    sizeof(key->eth.dst)) ||
2497 	    fl_dump_key_val(skb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
2498 			    mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
2499 			    sizeof(key->eth.src)) ||
2500 	    fl_dump_key_val(skb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE,
2501 			    &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
2502 			    sizeof(key->basic.n_proto)))
2503 		goto nla_put_failure;
2504 
2505 	if (fl_dump_key_mpls(skb, &key->mpls, &mask->mpls))
2506 		goto nla_put_failure;
2507 
2508 	if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_VLAN_ID,
2509 			     TCA_FLOWER_KEY_VLAN_PRIO, &key->vlan, &mask->vlan))
2510 		goto nla_put_failure;
2511 
2512 	if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_CVLAN_ID,
2513 			     TCA_FLOWER_KEY_CVLAN_PRIO,
2514 			     &key->cvlan, &mask->cvlan) ||
2515 	    (mask->cvlan.vlan_tpid &&
2516 	     nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
2517 			  key->cvlan.vlan_tpid)))
2518 		goto nla_put_failure;
2519 
2520 	if (mask->basic.n_proto) {
2521 		if (mask->cvlan.vlan_tpid) {
2522 			if (nla_put_be16(skb, TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
2523 					 key->basic.n_proto))
2524 				goto nla_put_failure;
2525 		} else if (mask->vlan.vlan_tpid) {
2526 			if (nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
2527 					 key->basic.n_proto))
2528 				goto nla_put_failure;
2529 		}
2530 	}
2531 
2532 	if ((key->basic.n_proto == htons(ETH_P_IP) ||
2533 	     key->basic.n_proto == htons(ETH_P_IPV6)) &&
2534 	    (fl_dump_key_val(skb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
2535 			    &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
2536 			    sizeof(key->basic.ip_proto)) ||
2537 	    fl_dump_key_ip(skb, false, &key->ip, &mask->ip)))
2538 		goto nla_put_failure;
2539 
2540 	if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
2541 	    (fl_dump_key_val(skb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
2542 			     &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
2543 			     sizeof(key->ipv4.src)) ||
2544 	     fl_dump_key_val(skb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
2545 			     &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
2546 			     sizeof(key->ipv4.dst))))
2547 		goto nla_put_failure;
2548 	else if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
2549 		 (fl_dump_key_val(skb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
2550 				  &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
2551 				  sizeof(key->ipv6.src)) ||
2552 		  fl_dump_key_val(skb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
2553 				  &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
2554 				  sizeof(key->ipv6.dst))))
2555 		goto nla_put_failure;
2556 
2557 	if (key->basic.ip_proto == IPPROTO_TCP &&
2558 	    (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
2559 			     &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
2560 			     sizeof(key->tp.src)) ||
2561 	     fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
2562 			     &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
2563 			     sizeof(key->tp.dst)) ||
2564 	     fl_dump_key_val(skb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
2565 			     &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
2566 			     sizeof(key->tcp.flags))))
2567 		goto nla_put_failure;
2568 	else if (key->basic.ip_proto == IPPROTO_UDP &&
2569 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
2570 				  &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
2571 				  sizeof(key->tp.src)) ||
2572 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
2573 				  &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
2574 				  sizeof(key->tp.dst))))
2575 		goto nla_put_failure;
2576 	else if (key->basic.ip_proto == IPPROTO_SCTP &&
2577 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
2578 				  &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
2579 				  sizeof(key->tp.src)) ||
2580 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
2581 				  &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
2582 				  sizeof(key->tp.dst))))
2583 		goto nla_put_failure;
2584 	else if (key->basic.n_proto == htons(ETH_P_IP) &&
2585 		 key->basic.ip_proto == IPPROTO_ICMP &&
2586 		 (fl_dump_key_val(skb, &key->icmp.type,
2587 				  TCA_FLOWER_KEY_ICMPV4_TYPE, &mask->icmp.type,
2588 				  TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
2589 				  sizeof(key->icmp.type)) ||
2590 		  fl_dump_key_val(skb, &key->icmp.code,
2591 				  TCA_FLOWER_KEY_ICMPV4_CODE, &mask->icmp.code,
2592 				  TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
2593 				  sizeof(key->icmp.code))))
2594 		goto nla_put_failure;
2595 	else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
2596 		 key->basic.ip_proto == IPPROTO_ICMPV6 &&
2597 		 (fl_dump_key_val(skb, &key->icmp.type,
2598 				  TCA_FLOWER_KEY_ICMPV6_TYPE, &mask->icmp.type,
2599 				  TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
2600 				  sizeof(key->icmp.type)) ||
2601 		  fl_dump_key_val(skb, &key->icmp.code,
2602 				  TCA_FLOWER_KEY_ICMPV6_CODE, &mask->icmp.code,
2603 				  TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
2604 				  sizeof(key->icmp.code))))
2605 		goto nla_put_failure;
2606 	else if ((key->basic.n_proto == htons(ETH_P_ARP) ||
2607 		  key->basic.n_proto == htons(ETH_P_RARP)) &&
2608 		 (fl_dump_key_val(skb, &key->arp.sip,
2609 				  TCA_FLOWER_KEY_ARP_SIP, &mask->arp.sip,
2610 				  TCA_FLOWER_KEY_ARP_SIP_MASK,
2611 				  sizeof(key->arp.sip)) ||
2612 		  fl_dump_key_val(skb, &key->arp.tip,
2613 				  TCA_FLOWER_KEY_ARP_TIP, &mask->arp.tip,
2614 				  TCA_FLOWER_KEY_ARP_TIP_MASK,
2615 				  sizeof(key->arp.tip)) ||
2616 		  fl_dump_key_val(skb, &key->arp.op,
2617 				  TCA_FLOWER_KEY_ARP_OP, &mask->arp.op,
2618 				  TCA_FLOWER_KEY_ARP_OP_MASK,
2619 				  sizeof(key->arp.op)) ||
2620 		  fl_dump_key_val(skb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
2621 				  mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
2622 				  sizeof(key->arp.sha)) ||
2623 		  fl_dump_key_val(skb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
2624 				  mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
2625 				  sizeof(key->arp.tha))))
2626 		goto nla_put_failure;
2627 
2628 	if ((key->basic.ip_proto == IPPROTO_TCP ||
2629 	     key->basic.ip_proto == IPPROTO_UDP ||
2630 	     key->basic.ip_proto == IPPROTO_SCTP) &&
2631 	     fl_dump_key_port_range(skb, key, mask))
2632 		goto nla_put_failure;
2633 
2634 	if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
2635 	    (fl_dump_key_val(skb, &key->enc_ipv4.src,
2636 			    TCA_FLOWER_KEY_ENC_IPV4_SRC, &mask->enc_ipv4.src,
2637 			    TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
2638 			    sizeof(key->enc_ipv4.src)) ||
2639 	     fl_dump_key_val(skb, &key->enc_ipv4.dst,
2640 			     TCA_FLOWER_KEY_ENC_IPV4_DST, &mask->enc_ipv4.dst,
2641 			     TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
2642 			     sizeof(key->enc_ipv4.dst))))
2643 		goto nla_put_failure;
2644 	else if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
2645 		 (fl_dump_key_val(skb, &key->enc_ipv6.src,
2646 			    TCA_FLOWER_KEY_ENC_IPV6_SRC, &mask->enc_ipv6.src,
2647 			    TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
2648 			    sizeof(key->enc_ipv6.src)) ||
2649 		 fl_dump_key_val(skb, &key->enc_ipv6.dst,
2650 				 TCA_FLOWER_KEY_ENC_IPV6_DST,
2651 				 &mask->enc_ipv6.dst,
2652 				 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
2653 			    sizeof(key->enc_ipv6.dst))))
2654 		goto nla_put_failure;
2655 
2656 	if (fl_dump_key_val(skb, &key->enc_key_id, TCA_FLOWER_KEY_ENC_KEY_ID,
2657 			    &mask->enc_key_id, TCA_FLOWER_UNSPEC,
2658 			    sizeof(key->enc_key_id)) ||
2659 	    fl_dump_key_val(skb, &key->enc_tp.src,
2660 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
2661 			    &mask->enc_tp.src,
2662 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
2663 			    sizeof(key->enc_tp.src)) ||
2664 	    fl_dump_key_val(skb, &key->enc_tp.dst,
2665 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
2666 			    &mask->enc_tp.dst,
2667 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
2668 			    sizeof(key->enc_tp.dst)) ||
2669 	    fl_dump_key_ip(skb, true, &key->enc_ip, &mask->enc_ip) ||
2670 	    fl_dump_key_enc_opt(skb, &key->enc_opts, &mask->enc_opts))
2671 		goto nla_put_failure;
2672 
2673 	if (fl_dump_key_ct(skb, &key->ct, &mask->ct))
2674 		goto nla_put_failure;
2675 
2676 	if (fl_dump_key_flags(skb, key->control.flags, mask->control.flags))
2677 		goto nla_put_failure;
2678 
2679 	return 0;
2680 
2681 nla_put_failure:
2682 	return -EMSGSIZE;
2683 }
2684 
2685 static int fl_dump(struct net *net, struct tcf_proto *tp, void *fh,
2686 		   struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
2687 {
2688 	struct cls_fl_filter *f = fh;
2689 	struct nlattr *nest;
2690 	struct fl_flow_key *key, *mask;
2691 	bool skip_hw;
2692 
2693 	if (!f)
2694 		return skb->len;
2695 
2696 	t->tcm_handle = f->handle;
2697 
2698 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
2699 	if (!nest)
2700 		goto nla_put_failure;
2701 
2702 	spin_lock(&tp->lock);
2703 
2704 	if (f->res.classid &&
2705 	    nla_put_u32(skb, TCA_FLOWER_CLASSID, f->res.classid))
2706 		goto nla_put_failure_locked;
2707 
2708 	key = &f->key;
2709 	mask = &f->mask->key;
2710 	skip_hw = tc_skip_hw(f->flags);
2711 
2712 	if (fl_dump_key(skb, net, key, mask))
2713 		goto nla_put_failure_locked;
2714 
2715 	if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
2716 		goto nla_put_failure_locked;
2717 
2718 	spin_unlock(&tp->lock);
2719 
2720 	if (!skip_hw)
2721 		fl_hw_update_stats(tp, f, rtnl_held);
2722 
2723 	if (nla_put_u32(skb, TCA_FLOWER_IN_HW_COUNT, f->in_hw_count))
2724 		goto nla_put_failure;
2725 
2726 	if (tcf_exts_dump(skb, &f->exts))
2727 		goto nla_put_failure;
2728 
2729 	nla_nest_end(skb, nest);
2730 
2731 	if (tcf_exts_dump_stats(skb, &f->exts) < 0)
2732 		goto nla_put_failure;
2733 
2734 	return skb->len;
2735 
2736 nla_put_failure_locked:
2737 	spin_unlock(&tp->lock);
2738 nla_put_failure:
2739 	nla_nest_cancel(skb, nest);
2740 	return -1;
2741 }
2742 
2743 static int fl_tmplt_dump(struct sk_buff *skb, struct net *net, void *tmplt_priv)
2744 {
2745 	struct fl_flow_tmplt *tmplt = tmplt_priv;
2746 	struct fl_flow_key *key, *mask;
2747 	struct nlattr *nest;
2748 
2749 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
2750 	if (!nest)
2751 		goto nla_put_failure;
2752 
2753 	key = &tmplt->dummy_key;
2754 	mask = &tmplt->mask;
2755 
2756 	if (fl_dump_key(skb, net, key, mask))
2757 		goto nla_put_failure;
2758 
2759 	nla_nest_end(skb, nest);
2760 
2761 	return skb->len;
2762 
2763 nla_put_failure:
2764 	nla_nest_cancel(skb, nest);
2765 	return -EMSGSIZE;
2766 }
2767 
2768 static void fl_bind_class(void *fh, u32 classid, unsigned long cl)
2769 {
2770 	struct cls_fl_filter *f = fh;
2771 
2772 	if (f && f->res.classid == classid)
2773 		f->res.class = cl;
2774 }
2775 
2776 static bool fl_delete_empty(struct tcf_proto *tp)
2777 {
2778 	struct cls_fl_head *head = fl_head_dereference(tp);
2779 
2780 	spin_lock(&tp->lock);
2781 	tp->deleting = idr_is_empty(&head->handle_idr);
2782 	spin_unlock(&tp->lock);
2783 
2784 	return tp->deleting;
2785 }
2786 
2787 static struct tcf_proto_ops cls_fl_ops __read_mostly = {
2788 	.kind		= "flower",
2789 	.classify	= fl_classify,
2790 	.init		= fl_init,
2791 	.destroy	= fl_destroy,
2792 	.get		= fl_get,
2793 	.put		= fl_put,
2794 	.change		= fl_change,
2795 	.delete		= fl_delete,
2796 	.delete_empty	= fl_delete_empty,
2797 	.walk		= fl_walk,
2798 	.reoffload	= fl_reoffload,
2799 	.hw_add		= fl_hw_add,
2800 	.hw_del		= fl_hw_del,
2801 	.dump		= fl_dump,
2802 	.bind_class	= fl_bind_class,
2803 	.tmplt_create	= fl_tmplt_create,
2804 	.tmplt_destroy	= fl_tmplt_destroy,
2805 	.tmplt_dump	= fl_tmplt_dump,
2806 	.owner		= THIS_MODULE,
2807 	.flags		= TCF_PROTO_OPS_DOIT_UNLOCKED,
2808 };
2809 
2810 static int __init cls_fl_init(void)
2811 {
2812 	return register_tcf_proto_ops(&cls_fl_ops);
2813 }
2814 
2815 static void __exit cls_fl_exit(void)
2816 {
2817 	unregister_tcf_proto_ops(&cls_fl_ops);
2818 }
2819 
2820 module_init(cls_fl_init);
2821 module_exit(cls_fl_exit);
2822 
2823 MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
2824 MODULE_DESCRIPTION("Flower classifier");
2825 MODULE_LICENSE("GPL v2");
2826