xref: /openbmc/linux/net/sched/cls_flower.c (revision e5242c5f)
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 #include <linux/bitfield.h>
15 
16 #include <linux/if_ether.h>
17 #include <linux/in6.h>
18 #include <linux/ip.h>
19 #include <linux/mpls.h>
20 #include <linux/ppp_defs.h>
21 
22 #include <net/sch_generic.h>
23 #include <net/pkt_cls.h>
24 #include <net/pkt_sched.h>
25 #include <net/ip.h>
26 #include <net/flow_dissector.h>
27 #include <net/geneve.h>
28 #include <net/vxlan.h>
29 #include <net/erspan.h>
30 #include <net/gtp.h>
31 #include <net/tc_wrapper.h>
32 
33 #include <net/dst.h>
34 #include <net/dst_metadata.h>
35 
36 #include <uapi/linux/netfilter/nf_conntrack_common.h>
37 
38 #define TCA_FLOWER_KEY_CT_FLAGS_MAX \
39 		((__TCA_FLOWER_KEY_CT_FLAGS_MAX - 1) << 1)
40 #define TCA_FLOWER_KEY_CT_FLAGS_MASK \
41 		(TCA_FLOWER_KEY_CT_FLAGS_MAX - 1)
42 
43 struct fl_flow_key {
44 	struct flow_dissector_key_meta meta;
45 	struct flow_dissector_key_control control;
46 	struct flow_dissector_key_control enc_control;
47 	struct flow_dissector_key_basic basic;
48 	struct flow_dissector_key_eth_addrs eth;
49 	struct flow_dissector_key_vlan vlan;
50 	struct flow_dissector_key_vlan cvlan;
51 	union {
52 		struct flow_dissector_key_ipv4_addrs ipv4;
53 		struct flow_dissector_key_ipv6_addrs ipv6;
54 	};
55 	struct flow_dissector_key_ports tp;
56 	struct flow_dissector_key_icmp icmp;
57 	struct flow_dissector_key_arp arp;
58 	struct flow_dissector_key_keyid enc_key_id;
59 	union {
60 		struct flow_dissector_key_ipv4_addrs enc_ipv4;
61 		struct flow_dissector_key_ipv6_addrs enc_ipv6;
62 	};
63 	struct flow_dissector_key_ports enc_tp;
64 	struct flow_dissector_key_mpls mpls;
65 	struct flow_dissector_key_tcp tcp;
66 	struct flow_dissector_key_ip ip;
67 	struct flow_dissector_key_ip enc_ip;
68 	struct flow_dissector_key_enc_opts enc_opts;
69 	struct flow_dissector_key_ports_range tp_range;
70 	struct flow_dissector_key_ct ct;
71 	struct flow_dissector_key_hash hash;
72 	struct flow_dissector_key_num_of_vlans num_of_vlans;
73 	struct flow_dissector_key_pppoe pppoe;
74 	struct flow_dissector_key_l2tpv3 l2tpv3;
75 	struct flow_dissector_key_ipsec ipsec;
76 	struct flow_dissector_key_cfm cfm;
77 } __aligned(BITS_PER_LONG / 8); /* Ensure that we can do comparisons as longs. */
78 
79 struct fl_flow_mask_range {
80 	unsigned short int start;
81 	unsigned short int end;
82 };
83 
84 struct fl_flow_mask {
85 	struct fl_flow_key key;
86 	struct fl_flow_mask_range range;
87 	u32 flags;
88 	struct rhash_head ht_node;
89 	struct rhashtable ht;
90 	struct rhashtable_params filter_ht_params;
91 	struct flow_dissector dissector;
92 	struct list_head filters;
93 	struct rcu_work rwork;
94 	struct list_head list;
95 	refcount_t refcnt;
96 };
97 
98 struct fl_flow_tmplt {
99 	struct fl_flow_key dummy_key;
100 	struct fl_flow_key mask;
101 	struct flow_dissector dissector;
102 	struct tcf_chain *chain;
103 };
104 
105 struct cls_fl_head {
106 	struct rhashtable ht;
107 	spinlock_t masks_lock; /* Protect masks list */
108 	struct list_head masks;
109 	struct list_head hw_filters;
110 	struct rcu_work rwork;
111 	struct idr handle_idr;
112 };
113 
114 struct cls_fl_filter {
115 	struct fl_flow_mask *mask;
116 	struct rhash_head ht_node;
117 	struct fl_flow_key mkey;
118 	struct tcf_exts exts;
119 	struct tcf_result res;
120 	struct fl_flow_key key;
121 	struct list_head list;
122 	struct list_head hw_list;
123 	u32 handle;
124 	u32 flags;
125 	u32 in_hw_count;
126 	u8 needs_tc_skb_ext:1;
127 	struct rcu_work rwork;
128 	struct net_device *hw_dev;
129 	/* Flower classifier is unlocked, which means that its reference counter
130 	 * can be changed concurrently without any kind of external
131 	 * synchronization. Use atomic reference counter to be concurrency-safe.
132 	 */
133 	refcount_t refcnt;
134 	bool deleted;
135 };
136 
137 static const struct rhashtable_params mask_ht_params = {
138 	.key_offset = offsetof(struct fl_flow_mask, key),
139 	.key_len = sizeof(struct fl_flow_key),
140 	.head_offset = offsetof(struct fl_flow_mask, ht_node),
141 	.automatic_shrinking = true,
142 };
143 
144 static unsigned short int fl_mask_range(const struct fl_flow_mask *mask)
145 {
146 	return mask->range.end - mask->range.start;
147 }
148 
149 static void fl_mask_update_range(struct fl_flow_mask *mask)
150 {
151 	const u8 *bytes = (const u8 *) &mask->key;
152 	size_t size = sizeof(mask->key);
153 	size_t i, first = 0, last;
154 
155 	for (i = 0; i < size; i++) {
156 		if (bytes[i]) {
157 			first = i;
158 			break;
159 		}
160 	}
161 	last = first;
162 	for (i = size - 1; i != first; i--) {
163 		if (bytes[i]) {
164 			last = i;
165 			break;
166 		}
167 	}
168 	mask->range.start = rounddown(first, sizeof(long));
169 	mask->range.end = roundup(last + 1, sizeof(long));
170 }
171 
172 static void *fl_key_get_start(struct fl_flow_key *key,
173 			      const struct fl_flow_mask *mask)
174 {
175 	return (u8 *) key + mask->range.start;
176 }
177 
178 static void fl_set_masked_key(struct fl_flow_key *mkey, struct fl_flow_key *key,
179 			      struct fl_flow_mask *mask)
180 {
181 	const long *lkey = fl_key_get_start(key, mask);
182 	const long *lmask = fl_key_get_start(&mask->key, mask);
183 	long *lmkey = fl_key_get_start(mkey, mask);
184 	int i;
185 
186 	for (i = 0; i < fl_mask_range(mask); i += sizeof(long))
187 		*lmkey++ = *lkey++ & *lmask++;
188 }
189 
190 static bool fl_mask_fits_tmplt(struct fl_flow_tmplt *tmplt,
191 			       struct fl_flow_mask *mask)
192 {
193 	const long *lmask = fl_key_get_start(&mask->key, mask);
194 	const long *ltmplt;
195 	int i;
196 
197 	if (!tmplt)
198 		return true;
199 	ltmplt = fl_key_get_start(&tmplt->mask, mask);
200 	for (i = 0; i < fl_mask_range(mask); i += sizeof(long)) {
201 		if (~*ltmplt++ & *lmask++)
202 			return false;
203 	}
204 	return true;
205 }
206 
207 static void fl_clear_masked_range(struct fl_flow_key *key,
208 				  struct fl_flow_mask *mask)
209 {
210 	memset(fl_key_get_start(key, mask), 0, fl_mask_range(mask));
211 }
212 
213 static bool fl_range_port_dst_cmp(struct cls_fl_filter *filter,
214 				  struct fl_flow_key *key,
215 				  struct fl_flow_key *mkey)
216 {
217 	u16 min_mask, max_mask, min_val, max_val;
218 
219 	min_mask = ntohs(filter->mask->key.tp_range.tp_min.dst);
220 	max_mask = ntohs(filter->mask->key.tp_range.tp_max.dst);
221 	min_val = ntohs(filter->key.tp_range.tp_min.dst);
222 	max_val = ntohs(filter->key.tp_range.tp_max.dst);
223 
224 	if (min_mask && max_mask) {
225 		if (ntohs(key->tp_range.tp.dst) < min_val ||
226 		    ntohs(key->tp_range.tp.dst) > max_val)
227 			return false;
228 
229 		/* skb does not have min and max values */
230 		mkey->tp_range.tp_min.dst = filter->mkey.tp_range.tp_min.dst;
231 		mkey->tp_range.tp_max.dst = filter->mkey.tp_range.tp_max.dst;
232 	}
233 	return true;
234 }
235 
236 static bool fl_range_port_src_cmp(struct cls_fl_filter *filter,
237 				  struct fl_flow_key *key,
238 				  struct fl_flow_key *mkey)
239 {
240 	u16 min_mask, max_mask, min_val, max_val;
241 
242 	min_mask = ntohs(filter->mask->key.tp_range.tp_min.src);
243 	max_mask = ntohs(filter->mask->key.tp_range.tp_max.src);
244 	min_val = ntohs(filter->key.tp_range.tp_min.src);
245 	max_val = ntohs(filter->key.tp_range.tp_max.src);
246 
247 	if (min_mask && max_mask) {
248 		if (ntohs(key->tp_range.tp.src) < min_val ||
249 		    ntohs(key->tp_range.tp.src) > max_val)
250 			return false;
251 
252 		/* skb does not have min and max values */
253 		mkey->tp_range.tp_min.src = filter->mkey.tp_range.tp_min.src;
254 		mkey->tp_range.tp_max.src = filter->mkey.tp_range.tp_max.src;
255 	}
256 	return true;
257 }
258 
259 static struct cls_fl_filter *__fl_lookup(struct fl_flow_mask *mask,
260 					 struct fl_flow_key *mkey)
261 {
262 	return rhashtable_lookup_fast(&mask->ht, fl_key_get_start(mkey, mask),
263 				      mask->filter_ht_params);
264 }
265 
266 static struct cls_fl_filter *fl_lookup_range(struct fl_flow_mask *mask,
267 					     struct fl_flow_key *mkey,
268 					     struct fl_flow_key *key)
269 {
270 	struct cls_fl_filter *filter, *f;
271 
272 	list_for_each_entry_rcu(filter, &mask->filters, list) {
273 		if (!fl_range_port_dst_cmp(filter, key, mkey))
274 			continue;
275 
276 		if (!fl_range_port_src_cmp(filter, key, mkey))
277 			continue;
278 
279 		f = __fl_lookup(mask, mkey);
280 		if (f)
281 			return f;
282 	}
283 	return NULL;
284 }
285 
286 static noinline_for_stack
287 struct cls_fl_filter *fl_mask_lookup(struct fl_flow_mask *mask, struct fl_flow_key *key)
288 {
289 	struct fl_flow_key mkey;
290 
291 	fl_set_masked_key(&mkey, key, mask);
292 	if ((mask->flags & TCA_FLOWER_MASK_FLAGS_RANGE))
293 		return fl_lookup_range(mask, &mkey, key);
294 
295 	return __fl_lookup(mask, &mkey);
296 }
297 
298 static u16 fl_ct_info_to_flower_map[] = {
299 	[IP_CT_ESTABLISHED] =		TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
300 					TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED,
301 	[IP_CT_RELATED] =		TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
302 					TCA_FLOWER_KEY_CT_FLAGS_RELATED,
303 	[IP_CT_ESTABLISHED_REPLY] =	TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
304 					TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED |
305 					TCA_FLOWER_KEY_CT_FLAGS_REPLY,
306 	[IP_CT_RELATED_REPLY] =		TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
307 					TCA_FLOWER_KEY_CT_FLAGS_RELATED |
308 					TCA_FLOWER_KEY_CT_FLAGS_REPLY,
309 	[IP_CT_NEW] =			TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
310 					TCA_FLOWER_KEY_CT_FLAGS_NEW,
311 };
312 
313 TC_INDIRECT_SCOPE int fl_classify(struct sk_buff *skb,
314 				  const struct tcf_proto *tp,
315 				  struct tcf_result *res)
316 {
317 	struct cls_fl_head *head = rcu_dereference_bh(tp->root);
318 	bool post_ct = tc_skb_cb(skb)->post_ct;
319 	u16 zone = tc_skb_cb(skb)->zone;
320 	struct fl_flow_key skb_key;
321 	struct fl_flow_mask *mask;
322 	struct cls_fl_filter *f;
323 
324 	list_for_each_entry_rcu(mask, &head->masks, list) {
325 		flow_dissector_init_keys(&skb_key.control, &skb_key.basic);
326 		fl_clear_masked_range(&skb_key, mask);
327 
328 		skb_flow_dissect_meta(skb, &mask->dissector, &skb_key);
329 		/* skb_flow_dissect() does not set n_proto in case an unknown
330 		 * protocol, so do it rather here.
331 		 */
332 		skb_key.basic.n_proto = skb_protocol(skb, false);
333 		skb_flow_dissect_tunnel_info(skb, &mask->dissector, &skb_key);
334 		skb_flow_dissect_ct(skb, &mask->dissector, &skb_key,
335 				    fl_ct_info_to_flower_map,
336 				    ARRAY_SIZE(fl_ct_info_to_flower_map),
337 				    post_ct, zone);
338 		skb_flow_dissect_hash(skb, &mask->dissector, &skb_key);
339 		skb_flow_dissect(skb, &mask->dissector, &skb_key,
340 				 FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP);
341 
342 		f = fl_mask_lookup(mask, &skb_key);
343 		if (f && !tc_skip_sw(f->flags)) {
344 			*res = f->res;
345 			return tcf_exts_exec(skb, &f->exts, res);
346 		}
347 	}
348 	return -1;
349 }
350 
351 static int fl_init(struct tcf_proto *tp)
352 {
353 	struct cls_fl_head *head;
354 
355 	head = kzalloc(sizeof(*head), GFP_KERNEL);
356 	if (!head)
357 		return -ENOBUFS;
358 
359 	spin_lock_init(&head->masks_lock);
360 	INIT_LIST_HEAD_RCU(&head->masks);
361 	INIT_LIST_HEAD(&head->hw_filters);
362 	rcu_assign_pointer(tp->root, head);
363 	idr_init(&head->handle_idr);
364 
365 	return rhashtable_init(&head->ht, &mask_ht_params);
366 }
367 
368 static void fl_mask_free(struct fl_flow_mask *mask, bool mask_init_done)
369 {
370 	/* temporary masks don't have their filters list and ht initialized */
371 	if (mask_init_done) {
372 		WARN_ON(!list_empty(&mask->filters));
373 		rhashtable_destroy(&mask->ht);
374 	}
375 	kfree(mask);
376 }
377 
378 static void fl_mask_free_work(struct work_struct *work)
379 {
380 	struct fl_flow_mask *mask = container_of(to_rcu_work(work),
381 						 struct fl_flow_mask, rwork);
382 
383 	fl_mask_free(mask, true);
384 }
385 
386 static void fl_uninit_mask_free_work(struct work_struct *work)
387 {
388 	struct fl_flow_mask *mask = container_of(to_rcu_work(work),
389 						 struct fl_flow_mask, rwork);
390 
391 	fl_mask_free(mask, false);
392 }
393 
394 static bool fl_mask_put(struct cls_fl_head *head, struct fl_flow_mask *mask)
395 {
396 	if (!refcount_dec_and_test(&mask->refcnt))
397 		return false;
398 
399 	rhashtable_remove_fast(&head->ht, &mask->ht_node, mask_ht_params);
400 
401 	spin_lock(&head->masks_lock);
402 	list_del_rcu(&mask->list);
403 	spin_unlock(&head->masks_lock);
404 
405 	tcf_queue_work(&mask->rwork, fl_mask_free_work);
406 
407 	return true;
408 }
409 
410 static struct cls_fl_head *fl_head_dereference(struct tcf_proto *tp)
411 {
412 	/* Flower classifier only changes root pointer during init and destroy.
413 	 * Users must obtain reference to tcf_proto instance before calling its
414 	 * API, so tp->root pointer is protected from concurrent call to
415 	 * fl_destroy() by reference counting.
416 	 */
417 	return rcu_dereference_raw(tp->root);
418 }
419 
420 static void __fl_destroy_filter(struct cls_fl_filter *f)
421 {
422 	if (f->needs_tc_skb_ext)
423 		tc_skb_ext_tc_disable();
424 	tcf_exts_destroy(&f->exts);
425 	tcf_exts_put_net(&f->exts);
426 	kfree(f);
427 }
428 
429 static void fl_destroy_filter_work(struct work_struct *work)
430 {
431 	struct cls_fl_filter *f = container_of(to_rcu_work(work),
432 					struct cls_fl_filter, rwork);
433 
434 	__fl_destroy_filter(f);
435 }
436 
437 static void fl_hw_destroy_filter(struct tcf_proto *tp, struct cls_fl_filter *f,
438 				 bool rtnl_held, struct netlink_ext_ack *extack)
439 {
440 	struct tcf_block *block = tp->chain->block;
441 	struct flow_cls_offload cls_flower = {};
442 
443 	tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
444 	cls_flower.command = FLOW_CLS_DESTROY;
445 	cls_flower.cookie = (unsigned long) f;
446 
447 	tc_setup_cb_destroy(block, tp, TC_SETUP_CLSFLOWER, &cls_flower, false,
448 			    &f->flags, &f->in_hw_count, rtnl_held);
449 
450 }
451 
452 static int fl_hw_replace_filter(struct tcf_proto *tp,
453 				struct cls_fl_filter *f, bool rtnl_held,
454 				struct netlink_ext_ack *extack)
455 {
456 	struct tcf_block *block = tp->chain->block;
457 	struct flow_cls_offload cls_flower = {};
458 	bool skip_sw = tc_skip_sw(f->flags);
459 	int err = 0;
460 
461 	cls_flower.rule = flow_rule_alloc(tcf_exts_num_actions(&f->exts));
462 	if (!cls_flower.rule)
463 		return -ENOMEM;
464 
465 	tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
466 	cls_flower.command = FLOW_CLS_REPLACE;
467 	cls_flower.cookie = (unsigned long) f;
468 	cls_flower.rule->match.dissector = &f->mask->dissector;
469 	cls_flower.rule->match.mask = &f->mask->key;
470 	cls_flower.rule->match.key = &f->mkey;
471 	cls_flower.classid = f->res.classid;
472 
473 	err = tc_setup_offload_action(&cls_flower.rule->action, &f->exts,
474 				      cls_flower.common.extack);
475 	if (err) {
476 		kfree(cls_flower.rule);
477 
478 		return skip_sw ? err : 0;
479 	}
480 
481 	err = tc_setup_cb_add(block, tp, TC_SETUP_CLSFLOWER, &cls_flower,
482 			      skip_sw, &f->flags, &f->in_hw_count, rtnl_held);
483 	tc_cleanup_offload_action(&cls_flower.rule->action);
484 	kfree(cls_flower.rule);
485 
486 	if (err) {
487 		fl_hw_destroy_filter(tp, f, rtnl_held, NULL);
488 		return err;
489 	}
490 
491 	if (skip_sw && !(f->flags & TCA_CLS_FLAGS_IN_HW))
492 		return -EINVAL;
493 
494 	return 0;
495 }
496 
497 static void fl_hw_update_stats(struct tcf_proto *tp, struct cls_fl_filter *f,
498 			       bool rtnl_held)
499 {
500 	struct tcf_block *block = tp->chain->block;
501 	struct flow_cls_offload cls_flower = {};
502 
503 	tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, NULL);
504 	cls_flower.command = FLOW_CLS_STATS;
505 	cls_flower.cookie = (unsigned long) f;
506 	cls_flower.classid = f->res.classid;
507 
508 	tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false,
509 			 rtnl_held);
510 
511 	tcf_exts_hw_stats_update(&f->exts, &cls_flower.stats, cls_flower.use_act_stats);
512 }
513 
514 static void __fl_put(struct cls_fl_filter *f)
515 {
516 	if (!refcount_dec_and_test(&f->refcnt))
517 		return;
518 
519 	if (tcf_exts_get_net(&f->exts))
520 		tcf_queue_work(&f->rwork, fl_destroy_filter_work);
521 	else
522 		__fl_destroy_filter(f);
523 }
524 
525 static struct cls_fl_filter *__fl_get(struct cls_fl_head *head, u32 handle)
526 {
527 	struct cls_fl_filter *f;
528 
529 	rcu_read_lock();
530 	f = idr_find(&head->handle_idr, handle);
531 	if (f && !refcount_inc_not_zero(&f->refcnt))
532 		f = NULL;
533 	rcu_read_unlock();
534 
535 	return f;
536 }
537 
538 static struct tcf_exts *fl_get_exts(const struct tcf_proto *tp, u32 handle)
539 {
540 	struct cls_fl_head *head = rcu_dereference_bh(tp->root);
541 	struct cls_fl_filter *f;
542 
543 	f = idr_find(&head->handle_idr, handle);
544 	return f ? &f->exts : NULL;
545 }
546 
547 static int __fl_delete(struct tcf_proto *tp, struct cls_fl_filter *f,
548 		       bool *last, bool rtnl_held,
549 		       struct netlink_ext_ack *extack)
550 {
551 	struct cls_fl_head *head = fl_head_dereference(tp);
552 
553 	*last = false;
554 
555 	spin_lock(&tp->lock);
556 	if (f->deleted) {
557 		spin_unlock(&tp->lock);
558 		return -ENOENT;
559 	}
560 
561 	f->deleted = true;
562 	rhashtable_remove_fast(&f->mask->ht, &f->ht_node,
563 			       f->mask->filter_ht_params);
564 	idr_remove(&head->handle_idr, f->handle);
565 	list_del_rcu(&f->list);
566 	spin_unlock(&tp->lock);
567 
568 	*last = fl_mask_put(head, f->mask);
569 	if (!tc_skip_hw(f->flags))
570 		fl_hw_destroy_filter(tp, f, rtnl_held, extack);
571 	tcf_unbind_filter(tp, &f->res);
572 	__fl_put(f);
573 
574 	return 0;
575 }
576 
577 static void fl_destroy_sleepable(struct work_struct *work)
578 {
579 	struct cls_fl_head *head = container_of(to_rcu_work(work),
580 						struct cls_fl_head,
581 						rwork);
582 
583 	rhashtable_destroy(&head->ht);
584 	kfree(head);
585 	module_put(THIS_MODULE);
586 }
587 
588 static void fl_destroy(struct tcf_proto *tp, bool rtnl_held,
589 		       struct netlink_ext_ack *extack)
590 {
591 	struct cls_fl_head *head = fl_head_dereference(tp);
592 	struct fl_flow_mask *mask, *next_mask;
593 	struct cls_fl_filter *f, *next;
594 	bool last;
595 
596 	list_for_each_entry_safe(mask, next_mask, &head->masks, list) {
597 		list_for_each_entry_safe(f, next, &mask->filters, list) {
598 			__fl_delete(tp, f, &last, rtnl_held, extack);
599 			if (last)
600 				break;
601 		}
602 	}
603 	idr_destroy(&head->handle_idr);
604 
605 	__module_get(THIS_MODULE);
606 	tcf_queue_work(&head->rwork, fl_destroy_sleepable);
607 }
608 
609 static void fl_put(struct tcf_proto *tp, void *arg)
610 {
611 	struct cls_fl_filter *f = arg;
612 
613 	__fl_put(f);
614 }
615 
616 static void *fl_get(struct tcf_proto *tp, u32 handle)
617 {
618 	struct cls_fl_head *head = fl_head_dereference(tp);
619 
620 	return __fl_get(head, handle);
621 }
622 
623 static const struct nla_policy fl_policy[TCA_FLOWER_MAX + 1] = {
624 	[TCA_FLOWER_UNSPEC]		= { .strict_start_type =
625 						TCA_FLOWER_L2_MISS },
626 	[TCA_FLOWER_CLASSID]		= { .type = NLA_U32 },
627 	[TCA_FLOWER_INDEV]		= { .type = NLA_STRING,
628 					    .len = IFNAMSIZ },
629 	[TCA_FLOWER_KEY_ETH_DST]	= { .len = ETH_ALEN },
630 	[TCA_FLOWER_KEY_ETH_DST_MASK]	= { .len = ETH_ALEN },
631 	[TCA_FLOWER_KEY_ETH_SRC]	= { .len = ETH_ALEN },
632 	[TCA_FLOWER_KEY_ETH_SRC_MASK]	= { .len = ETH_ALEN },
633 	[TCA_FLOWER_KEY_ETH_TYPE]	= { .type = NLA_U16 },
634 	[TCA_FLOWER_KEY_IP_PROTO]	= { .type = NLA_U8 },
635 	[TCA_FLOWER_KEY_IPV4_SRC]	= { .type = NLA_U32 },
636 	[TCA_FLOWER_KEY_IPV4_SRC_MASK]	= { .type = NLA_U32 },
637 	[TCA_FLOWER_KEY_IPV4_DST]	= { .type = NLA_U32 },
638 	[TCA_FLOWER_KEY_IPV4_DST_MASK]	= { .type = NLA_U32 },
639 	[TCA_FLOWER_KEY_IPV6_SRC]	= { .len = sizeof(struct in6_addr) },
640 	[TCA_FLOWER_KEY_IPV6_SRC_MASK]	= { .len = sizeof(struct in6_addr) },
641 	[TCA_FLOWER_KEY_IPV6_DST]	= { .len = sizeof(struct in6_addr) },
642 	[TCA_FLOWER_KEY_IPV6_DST_MASK]	= { .len = sizeof(struct in6_addr) },
643 	[TCA_FLOWER_KEY_TCP_SRC]	= { .type = NLA_U16 },
644 	[TCA_FLOWER_KEY_TCP_DST]	= { .type = NLA_U16 },
645 	[TCA_FLOWER_KEY_UDP_SRC]	= { .type = NLA_U16 },
646 	[TCA_FLOWER_KEY_UDP_DST]	= { .type = NLA_U16 },
647 	[TCA_FLOWER_KEY_VLAN_ID]	= { .type = NLA_U16 },
648 	[TCA_FLOWER_KEY_VLAN_PRIO]	= { .type = NLA_U8 },
649 	[TCA_FLOWER_KEY_VLAN_ETH_TYPE]	= { .type = NLA_U16 },
650 	[TCA_FLOWER_KEY_ENC_KEY_ID]	= { .type = NLA_U32 },
651 	[TCA_FLOWER_KEY_ENC_IPV4_SRC]	= { .type = NLA_U32 },
652 	[TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK] = { .type = NLA_U32 },
653 	[TCA_FLOWER_KEY_ENC_IPV4_DST]	= { .type = NLA_U32 },
654 	[TCA_FLOWER_KEY_ENC_IPV4_DST_MASK] = { .type = NLA_U32 },
655 	[TCA_FLOWER_KEY_ENC_IPV6_SRC]	= { .len = sizeof(struct in6_addr) },
656 	[TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK] = { .len = sizeof(struct in6_addr) },
657 	[TCA_FLOWER_KEY_ENC_IPV6_DST]	= { .len = sizeof(struct in6_addr) },
658 	[TCA_FLOWER_KEY_ENC_IPV6_DST_MASK] = { .len = sizeof(struct in6_addr) },
659 	[TCA_FLOWER_KEY_TCP_SRC_MASK]	= { .type = NLA_U16 },
660 	[TCA_FLOWER_KEY_TCP_DST_MASK]	= { .type = NLA_U16 },
661 	[TCA_FLOWER_KEY_UDP_SRC_MASK]	= { .type = NLA_U16 },
662 	[TCA_FLOWER_KEY_UDP_DST_MASK]	= { .type = NLA_U16 },
663 	[TCA_FLOWER_KEY_SCTP_SRC_MASK]	= { .type = NLA_U16 },
664 	[TCA_FLOWER_KEY_SCTP_DST_MASK]	= { .type = NLA_U16 },
665 	[TCA_FLOWER_KEY_SCTP_SRC]	= { .type = NLA_U16 },
666 	[TCA_FLOWER_KEY_SCTP_DST]	= { .type = NLA_U16 },
667 	[TCA_FLOWER_KEY_ENC_UDP_SRC_PORT]	= { .type = NLA_U16 },
668 	[TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK]	= { .type = NLA_U16 },
669 	[TCA_FLOWER_KEY_ENC_UDP_DST_PORT]	= { .type = NLA_U16 },
670 	[TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK]	= { .type = NLA_U16 },
671 	[TCA_FLOWER_KEY_FLAGS]		= { .type = NLA_U32 },
672 	[TCA_FLOWER_KEY_FLAGS_MASK]	= { .type = NLA_U32 },
673 	[TCA_FLOWER_KEY_ICMPV4_TYPE]	= { .type = NLA_U8 },
674 	[TCA_FLOWER_KEY_ICMPV4_TYPE_MASK] = { .type = NLA_U8 },
675 	[TCA_FLOWER_KEY_ICMPV4_CODE]	= { .type = NLA_U8 },
676 	[TCA_FLOWER_KEY_ICMPV4_CODE_MASK] = { .type = NLA_U8 },
677 	[TCA_FLOWER_KEY_ICMPV6_TYPE]	= { .type = NLA_U8 },
678 	[TCA_FLOWER_KEY_ICMPV6_TYPE_MASK] = { .type = NLA_U8 },
679 	[TCA_FLOWER_KEY_ICMPV6_CODE]	= { .type = NLA_U8 },
680 	[TCA_FLOWER_KEY_ICMPV6_CODE_MASK] = { .type = NLA_U8 },
681 	[TCA_FLOWER_KEY_ARP_SIP]	= { .type = NLA_U32 },
682 	[TCA_FLOWER_KEY_ARP_SIP_MASK]	= { .type = NLA_U32 },
683 	[TCA_FLOWER_KEY_ARP_TIP]	= { .type = NLA_U32 },
684 	[TCA_FLOWER_KEY_ARP_TIP_MASK]	= { .type = NLA_U32 },
685 	[TCA_FLOWER_KEY_ARP_OP]		= { .type = NLA_U8 },
686 	[TCA_FLOWER_KEY_ARP_OP_MASK]	= { .type = NLA_U8 },
687 	[TCA_FLOWER_KEY_ARP_SHA]	= { .len = ETH_ALEN },
688 	[TCA_FLOWER_KEY_ARP_SHA_MASK]	= { .len = ETH_ALEN },
689 	[TCA_FLOWER_KEY_ARP_THA]	= { .len = ETH_ALEN },
690 	[TCA_FLOWER_KEY_ARP_THA_MASK]	= { .len = ETH_ALEN },
691 	[TCA_FLOWER_KEY_MPLS_TTL]	= { .type = NLA_U8 },
692 	[TCA_FLOWER_KEY_MPLS_BOS]	= { .type = NLA_U8 },
693 	[TCA_FLOWER_KEY_MPLS_TC]	= { .type = NLA_U8 },
694 	[TCA_FLOWER_KEY_MPLS_LABEL]	= { .type = NLA_U32 },
695 	[TCA_FLOWER_KEY_MPLS_OPTS]	= { .type = NLA_NESTED },
696 	[TCA_FLOWER_KEY_TCP_FLAGS]	= { .type = NLA_U16 },
697 	[TCA_FLOWER_KEY_TCP_FLAGS_MASK]	= { .type = NLA_U16 },
698 	[TCA_FLOWER_KEY_IP_TOS]		= { .type = NLA_U8 },
699 	[TCA_FLOWER_KEY_IP_TOS_MASK]	= { .type = NLA_U8 },
700 	[TCA_FLOWER_KEY_IP_TTL]		= { .type = NLA_U8 },
701 	[TCA_FLOWER_KEY_IP_TTL_MASK]	= { .type = NLA_U8 },
702 	[TCA_FLOWER_KEY_CVLAN_ID]	= { .type = NLA_U16 },
703 	[TCA_FLOWER_KEY_CVLAN_PRIO]	= { .type = NLA_U8 },
704 	[TCA_FLOWER_KEY_CVLAN_ETH_TYPE]	= { .type = NLA_U16 },
705 	[TCA_FLOWER_KEY_ENC_IP_TOS]	= { .type = NLA_U8 },
706 	[TCA_FLOWER_KEY_ENC_IP_TOS_MASK] = { .type = NLA_U8 },
707 	[TCA_FLOWER_KEY_ENC_IP_TTL]	 = { .type = NLA_U8 },
708 	[TCA_FLOWER_KEY_ENC_IP_TTL_MASK] = { .type = NLA_U8 },
709 	[TCA_FLOWER_KEY_ENC_OPTS]	= { .type = NLA_NESTED },
710 	[TCA_FLOWER_KEY_ENC_OPTS_MASK]	= { .type = NLA_NESTED },
711 	[TCA_FLOWER_KEY_CT_STATE]	=
712 		NLA_POLICY_MASK(NLA_U16, TCA_FLOWER_KEY_CT_FLAGS_MASK),
713 	[TCA_FLOWER_KEY_CT_STATE_MASK]	=
714 		NLA_POLICY_MASK(NLA_U16, TCA_FLOWER_KEY_CT_FLAGS_MASK),
715 	[TCA_FLOWER_KEY_CT_ZONE]	= { .type = NLA_U16 },
716 	[TCA_FLOWER_KEY_CT_ZONE_MASK]	= { .type = NLA_U16 },
717 	[TCA_FLOWER_KEY_CT_MARK]	= { .type = NLA_U32 },
718 	[TCA_FLOWER_KEY_CT_MARK_MASK]	= { .type = NLA_U32 },
719 	[TCA_FLOWER_KEY_CT_LABELS]	= { .type = NLA_BINARY,
720 					    .len = 128 / BITS_PER_BYTE },
721 	[TCA_FLOWER_KEY_CT_LABELS_MASK]	= { .type = NLA_BINARY,
722 					    .len = 128 / BITS_PER_BYTE },
723 	[TCA_FLOWER_FLAGS]		= { .type = NLA_U32 },
724 	[TCA_FLOWER_KEY_HASH]		= { .type = NLA_U32 },
725 	[TCA_FLOWER_KEY_HASH_MASK]	= { .type = NLA_U32 },
726 	[TCA_FLOWER_KEY_NUM_OF_VLANS]	= { .type = NLA_U8 },
727 	[TCA_FLOWER_KEY_PPPOE_SID]	= { .type = NLA_U16 },
728 	[TCA_FLOWER_KEY_PPP_PROTO]	= { .type = NLA_U16 },
729 	[TCA_FLOWER_KEY_L2TPV3_SID]	= { .type = NLA_U32 },
730 	[TCA_FLOWER_KEY_SPI]		= { .type = NLA_U32 },
731 	[TCA_FLOWER_KEY_SPI_MASK]	= { .type = NLA_U32 },
732 	[TCA_FLOWER_L2_MISS]		= NLA_POLICY_MAX(NLA_U8, 1),
733 	[TCA_FLOWER_KEY_CFM]		= { .type = NLA_NESTED },
734 };
735 
736 static const struct nla_policy
737 enc_opts_policy[TCA_FLOWER_KEY_ENC_OPTS_MAX + 1] = {
738 	[TCA_FLOWER_KEY_ENC_OPTS_UNSPEC]        = {
739 		.strict_start_type = TCA_FLOWER_KEY_ENC_OPTS_VXLAN },
740 	[TCA_FLOWER_KEY_ENC_OPTS_GENEVE]        = { .type = NLA_NESTED },
741 	[TCA_FLOWER_KEY_ENC_OPTS_VXLAN]         = { .type = NLA_NESTED },
742 	[TCA_FLOWER_KEY_ENC_OPTS_ERSPAN]        = { .type = NLA_NESTED },
743 	[TCA_FLOWER_KEY_ENC_OPTS_GTP]		= { .type = NLA_NESTED },
744 };
745 
746 static const struct nla_policy
747 geneve_opt_policy[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1] = {
748 	[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS]      = { .type = NLA_U16 },
749 	[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE]       = { .type = NLA_U8 },
750 	[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA]       = { .type = NLA_BINARY,
751 						       .len = 128 },
752 };
753 
754 static const struct nla_policy
755 vxlan_opt_policy[TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX + 1] = {
756 	[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]         = { .type = NLA_U32 },
757 };
758 
759 static const struct nla_policy
760 erspan_opt_policy[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX + 1] = {
761 	[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]        = { .type = NLA_U8 },
762 	[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]      = { .type = NLA_U32 },
763 	[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR]        = { .type = NLA_U8 },
764 	[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID]       = { .type = NLA_U8 },
765 };
766 
767 static const struct nla_policy
768 gtp_opt_policy[TCA_FLOWER_KEY_ENC_OPT_GTP_MAX + 1] = {
769 	[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE]	   = { .type = NLA_U8 },
770 	[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI]	   = { .type = NLA_U8 },
771 };
772 
773 static const struct nla_policy
774 mpls_stack_entry_policy[TCA_FLOWER_KEY_MPLS_OPT_LSE_MAX + 1] = {
775 	[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH]    = { .type = NLA_U8 },
776 	[TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL]      = { .type = NLA_U8 },
777 	[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS]      = { .type = NLA_U8 },
778 	[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC]       = { .type = NLA_U8 },
779 	[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL]    = { .type = NLA_U32 },
780 };
781 
782 static const struct nla_policy
783 cfm_opt_policy[TCA_FLOWER_KEY_CFM_OPT_MAX + 1] = {
784 	[TCA_FLOWER_KEY_CFM_MD_LEVEL]	= NLA_POLICY_MAX(NLA_U8,
785 						FLOW_DIS_CFM_MDL_MAX),
786 	[TCA_FLOWER_KEY_CFM_OPCODE]	= { .type = NLA_U8 },
787 };
788 
789 static void fl_set_key_val(struct nlattr **tb,
790 			   void *val, int val_type,
791 			   void *mask, int mask_type, int len)
792 {
793 	if (!tb[val_type])
794 		return;
795 	nla_memcpy(val, tb[val_type], len);
796 	if (mask_type == TCA_FLOWER_UNSPEC || !tb[mask_type])
797 		memset(mask, 0xff, len);
798 	else
799 		nla_memcpy(mask, tb[mask_type], len);
800 }
801 
802 static int fl_set_key_spi(struct nlattr **tb, struct fl_flow_key *key,
803 			  struct fl_flow_key *mask,
804 			  struct netlink_ext_ack *extack)
805 {
806 	if (key->basic.ip_proto != IPPROTO_ESP &&
807 	    key->basic.ip_proto != IPPROTO_AH) {
808 		NL_SET_ERR_MSG(extack,
809 			       "Protocol must be either ESP or AH");
810 		return -EINVAL;
811 	}
812 
813 	fl_set_key_val(tb, &key->ipsec.spi,
814 		       TCA_FLOWER_KEY_SPI,
815 		       &mask->ipsec.spi, TCA_FLOWER_KEY_SPI_MASK,
816 		       sizeof(key->ipsec.spi));
817 	return 0;
818 }
819 
820 static int fl_set_key_port_range(struct nlattr **tb, struct fl_flow_key *key,
821 				 struct fl_flow_key *mask,
822 				 struct netlink_ext_ack *extack)
823 {
824 	fl_set_key_val(tb, &key->tp_range.tp_min.dst,
825 		       TCA_FLOWER_KEY_PORT_DST_MIN, &mask->tp_range.tp_min.dst,
826 		       TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_min.dst));
827 	fl_set_key_val(tb, &key->tp_range.tp_max.dst,
828 		       TCA_FLOWER_KEY_PORT_DST_MAX, &mask->tp_range.tp_max.dst,
829 		       TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_max.dst));
830 	fl_set_key_val(tb, &key->tp_range.tp_min.src,
831 		       TCA_FLOWER_KEY_PORT_SRC_MIN, &mask->tp_range.tp_min.src,
832 		       TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_min.src));
833 	fl_set_key_val(tb, &key->tp_range.tp_max.src,
834 		       TCA_FLOWER_KEY_PORT_SRC_MAX, &mask->tp_range.tp_max.src,
835 		       TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_max.src));
836 
837 	if (mask->tp_range.tp_min.dst != mask->tp_range.tp_max.dst) {
838 		NL_SET_ERR_MSG(extack,
839 			       "Both min and max destination ports must be specified");
840 		return -EINVAL;
841 	}
842 	if (mask->tp_range.tp_min.src != mask->tp_range.tp_max.src) {
843 		NL_SET_ERR_MSG(extack,
844 			       "Both min and max source ports must be specified");
845 		return -EINVAL;
846 	}
847 	if (mask->tp_range.tp_min.dst && mask->tp_range.tp_max.dst &&
848 	    ntohs(key->tp_range.tp_max.dst) <=
849 	    ntohs(key->tp_range.tp_min.dst)) {
850 		NL_SET_ERR_MSG_ATTR(extack,
851 				    tb[TCA_FLOWER_KEY_PORT_DST_MIN],
852 				    "Invalid destination port range (min must be strictly smaller than max)");
853 		return -EINVAL;
854 	}
855 	if (mask->tp_range.tp_min.src && mask->tp_range.tp_max.src &&
856 	    ntohs(key->tp_range.tp_max.src) <=
857 	    ntohs(key->tp_range.tp_min.src)) {
858 		NL_SET_ERR_MSG_ATTR(extack,
859 				    tb[TCA_FLOWER_KEY_PORT_SRC_MIN],
860 				    "Invalid source port range (min must be strictly smaller than max)");
861 		return -EINVAL;
862 	}
863 
864 	return 0;
865 }
866 
867 static int fl_set_key_mpls_lse(const struct nlattr *nla_lse,
868 			       struct flow_dissector_key_mpls *key_val,
869 			       struct flow_dissector_key_mpls *key_mask,
870 			       struct netlink_ext_ack *extack)
871 {
872 	struct nlattr *tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_MAX + 1];
873 	struct flow_dissector_mpls_lse *lse_mask;
874 	struct flow_dissector_mpls_lse *lse_val;
875 	u8 lse_index;
876 	u8 depth;
877 	int err;
878 
879 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_MPLS_OPT_LSE_MAX, nla_lse,
880 			       mpls_stack_entry_policy, extack);
881 	if (err < 0)
882 		return err;
883 
884 	if (!tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH]) {
885 		NL_SET_ERR_MSG(extack, "Missing MPLS option \"depth\"");
886 		return -EINVAL;
887 	}
888 
889 	depth = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH]);
890 
891 	/* LSE depth starts at 1, for consistency with terminology used by
892 	 * RFC 3031 (section 3.9), where depth 0 refers to unlabeled packets.
893 	 */
894 	if (depth < 1 || depth > FLOW_DIS_MPLS_MAX) {
895 		NL_SET_ERR_MSG_ATTR(extack,
896 				    tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH],
897 				    "Invalid MPLS depth");
898 		return -EINVAL;
899 	}
900 	lse_index = depth - 1;
901 
902 	dissector_set_mpls_lse(key_val, lse_index);
903 	dissector_set_mpls_lse(key_mask, lse_index);
904 
905 	lse_val = &key_val->ls[lse_index];
906 	lse_mask = &key_mask->ls[lse_index];
907 
908 	if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL]) {
909 		lse_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL]);
910 		lse_mask->mpls_ttl = MPLS_TTL_MASK;
911 	}
912 	if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS]) {
913 		u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS]);
914 
915 		if (bos & ~MPLS_BOS_MASK) {
916 			NL_SET_ERR_MSG_ATTR(extack,
917 					    tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS],
918 					    "Bottom Of Stack (BOS) must be 0 or 1");
919 			return -EINVAL;
920 		}
921 		lse_val->mpls_bos = bos;
922 		lse_mask->mpls_bos = MPLS_BOS_MASK;
923 	}
924 	if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC]) {
925 		u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC]);
926 
927 		if (tc & ~MPLS_TC_MASK) {
928 			NL_SET_ERR_MSG_ATTR(extack,
929 					    tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC],
930 					    "Traffic Class (TC) must be between 0 and 7");
931 			return -EINVAL;
932 		}
933 		lse_val->mpls_tc = tc;
934 		lse_mask->mpls_tc = MPLS_TC_MASK;
935 	}
936 	if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL]) {
937 		u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL]);
938 
939 		if (label & ~MPLS_LABEL_MASK) {
940 			NL_SET_ERR_MSG_ATTR(extack,
941 					    tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL],
942 					    "Label must be between 0 and 1048575");
943 			return -EINVAL;
944 		}
945 		lse_val->mpls_label = label;
946 		lse_mask->mpls_label = MPLS_LABEL_MASK;
947 	}
948 
949 	return 0;
950 }
951 
952 static int fl_set_key_mpls_opts(const struct nlattr *nla_mpls_opts,
953 				struct flow_dissector_key_mpls *key_val,
954 				struct flow_dissector_key_mpls *key_mask,
955 				struct netlink_ext_ack *extack)
956 {
957 	struct nlattr *nla_lse;
958 	int rem;
959 	int err;
960 
961 	if (!(nla_mpls_opts->nla_type & NLA_F_NESTED)) {
962 		NL_SET_ERR_MSG_ATTR(extack, nla_mpls_opts,
963 				    "NLA_F_NESTED is missing");
964 		return -EINVAL;
965 	}
966 
967 	nla_for_each_nested(nla_lse, nla_mpls_opts, rem) {
968 		if (nla_type(nla_lse) != TCA_FLOWER_KEY_MPLS_OPTS_LSE) {
969 			NL_SET_ERR_MSG_ATTR(extack, nla_lse,
970 					    "Invalid MPLS option type");
971 			return -EINVAL;
972 		}
973 
974 		err = fl_set_key_mpls_lse(nla_lse, key_val, key_mask, extack);
975 		if (err < 0)
976 			return err;
977 	}
978 	if (rem) {
979 		NL_SET_ERR_MSG(extack,
980 			       "Bytes leftover after parsing MPLS options");
981 		return -EINVAL;
982 	}
983 
984 	return 0;
985 }
986 
987 static int fl_set_key_mpls(struct nlattr **tb,
988 			   struct flow_dissector_key_mpls *key_val,
989 			   struct flow_dissector_key_mpls *key_mask,
990 			   struct netlink_ext_ack *extack)
991 {
992 	struct flow_dissector_mpls_lse *lse_mask;
993 	struct flow_dissector_mpls_lse *lse_val;
994 
995 	if (tb[TCA_FLOWER_KEY_MPLS_OPTS]) {
996 		if (tb[TCA_FLOWER_KEY_MPLS_TTL] ||
997 		    tb[TCA_FLOWER_KEY_MPLS_BOS] ||
998 		    tb[TCA_FLOWER_KEY_MPLS_TC] ||
999 		    tb[TCA_FLOWER_KEY_MPLS_LABEL]) {
1000 			NL_SET_ERR_MSG_ATTR(extack,
1001 					    tb[TCA_FLOWER_KEY_MPLS_OPTS],
1002 					    "MPLS label, Traffic Class, Bottom Of Stack and Time To Live must be encapsulated in the MPLS options attribute");
1003 			return -EBADMSG;
1004 		}
1005 
1006 		return fl_set_key_mpls_opts(tb[TCA_FLOWER_KEY_MPLS_OPTS],
1007 					    key_val, key_mask, extack);
1008 	}
1009 
1010 	lse_val = &key_val->ls[0];
1011 	lse_mask = &key_mask->ls[0];
1012 
1013 	if (tb[TCA_FLOWER_KEY_MPLS_TTL]) {
1014 		lse_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TTL]);
1015 		lse_mask->mpls_ttl = MPLS_TTL_MASK;
1016 		dissector_set_mpls_lse(key_val, 0);
1017 		dissector_set_mpls_lse(key_mask, 0);
1018 	}
1019 	if (tb[TCA_FLOWER_KEY_MPLS_BOS]) {
1020 		u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_BOS]);
1021 
1022 		if (bos & ~MPLS_BOS_MASK) {
1023 			NL_SET_ERR_MSG_ATTR(extack,
1024 					    tb[TCA_FLOWER_KEY_MPLS_BOS],
1025 					    "Bottom Of Stack (BOS) must be 0 or 1");
1026 			return -EINVAL;
1027 		}
1028 		lse_val->mpls_bos = bos;
1029 		lse_mask->mpls_bos = MPLS_BOS_MASK;
1030 		dissector_set_mpls_lse(key_val, 0);
1031 		dissector_set_mpls_lse(key_mask, 0);
1032 	}
1033 	if (tb[TCA_FLOWER_KEY_MPLS_TC]) {
1034 		u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TC]);
1035 
1036 		if (tc & ~MPLS_TC_MASK) {
1037 			NL_SET_ERR_MSG_ATTR(extack,
1038 					    tb[TCA_FLOWER_KEY_MPLS_TC],
1039 					    "Traffic Class (TC) must be between 0 and 7");
1040 			return -EINVAL;
1041 		}
1042 		lse_val->mpls_tc = tc;
1043 		lse_mask->mpls_tc = MPLS_TC_MASK;
1044 		dissector_set_mpls_lse(key_val, 0);
1045 		dissector_set_mpls_lse(key_mask, 0);
1046 	}
1047 	if (tb[TCA_FLOWER_KEY_MPLS_LABEL]) {
1048 		u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_LABEL]);
1049 
1050 		if (label & ~MPLS_LABEL_MASK) {
1051 			NL_SET_ERR_MSG_ATTR(extack,
1052 					    tb[TCA_FLOWER_KEY_MPLS_LABEL],
1053 					    "Label must be between 0 and 1048575");
1054 			return -EINVAL;
1055 		}
1056 		lse_val->mpls_label = label;
1057 		lse_mask->mpls_label = MPLS_LABEL_MASK;
1058 		dissector_set_mpls_lse(key_val, 0);
1059 		dissector_set_mpls_lse(key_mask, 0);
1060 	}
1061 	return 0;
1062 }
1063 
1064 static void fl_set_key_vlan(struct nlattr **tb,
1065 			    __be16 ethertype,
1066 			    int vlan_id_key, int vlan_prio_key,
1067 			    int vlan_next_eth_type_key,
1068 			    struct flow_dissector_key_vlan *key_val,
1069 			    struct flow_dissector_key_vlan *key_mask)
1070 {
1071 #define VLAN_PRIORITY_MASK	0x7
1072 
1073 	if (tb[vlan_id_key]) {
1074 		key_val->vlan_id =
1075 			nla_get_u16(tb[vlan_id_key]) & VLAN_VID_MASK;
1076 		key_mask->vlan_id = VLAN_VID_MASK;
1077 	}
1078 	if (tb[vlan_prio_key]) {
1079 		key_val->vlan_priority =
1080 			nla_get_u8(tb[vlan_prio_key]) &
1081 			VLAN_PRIORITY_MASK;
1082 		key_mask->vlan_priority = VLAN_PRIORITY_MASK;
1083 	}
1084 	if (ethertype) {
1085 		key_val->vlan_tpid = ethertype;
1086 		key_mask->vlan_tpid = cpu_to_be16(~0);
1087 	}
1088 	if (tb[vlan_next_eth_type_key]) {
1089 		key_val->vlan_eth_type =
1090 			nla_get_be16(tb[vlan_next_eth_type_key]);
1091 		key_mask->vlan_eth_type = cpu_to_be16(~0);
1092 	}
1093 }
1094 
1095 static void fl_set_key_pppoe(struct nlattr **tb,
1096 			     struct flow_dissector_key_pppoe *key_val,
1097 			     struct flow_dissector_key_pppoe *key_mask,
1098 			     struct fl_flow_key *key,
1099 			     struct fl_flow_key *mask)
1100 {
1101 	/* key_val::type must be set to ETH_P_PPP_SES
1102 	 * because ETH_P_PPP_SES was stored in basic.n_proto
1103 	 * which might get overwritten by ppp_proto
1104 	 * or might be set to 0, the role of key_val::type
1105 	 * is similar to vlan_key::tpid
1106 	 */
1107 	key_val->type = htons(ETH_P_PPP_SES);
1108 	key_mask->type = cpu_to_be16(~0);
1109 
1110 	if (tb[TCA_FLOWER_KEY_PPPOE_SID]) {
1111 		key_val->session_id =
1112 			nla_get_be16(tb[TCA_FLOWER_KEY_PPPOE_SID]);
1113 		key_mask->session_id = cpu_to_be16(~0);
1114 	}
1115 	if (tb[TCA_FLOWER_KEY_PPP_PROTO]) {
1116 		key_val->ppp_proto =
1117 			nla_get_be16(tb[TCA_FLOWER_KEY_PPP_PROTO]);
1118 		key_mask->ppp_proto = cpu_to_be16(~0);
1119 
1120 		if (key_val->ppp_proto == htons(PPP_IP)) {
1121 			key->basic.n_proto = htons(ETH_P_IP);
1122 			mask->basic.n_proto = cpu_to_be16(~0);
1123 		} else if (key_val->ppp_proto == htons(PPP_IPV6)) {
1124 			key->basic.n_proto = htons(ETH_P_IPV6);
1125 			mask->basic.n_proto = cpu_to_be16(~0);
1126 		} else if (key_val->ppp_proto == htons(PPP_MPLS_UC)) {
1127 			key->basic.n_proto = htons(ETH_P_MPLS_UC);
1128 			mask->basic.n_proto = cpu_to_be16(~0);
1129 		} else if (key_val->ppp_proto == htons(PPP_MPLS_MC)) {
1130 			key->basic.n_proto = htons(ETH_P_MPLS_MC);
1131 			mask->basic.n_proto = cpu_to_be16(~0);
1132 		}
1133 	} else {
1134 		key->basic.n_proto = 0;
1135 		mask->basic.n_proto = cpu_to_be16(0);
1136 	}
1137 }
1138 
1139 static void fl_set_key_flag(u32 flower_key, u32 flower_mask,
1140 			    u32 *dissector_key, u32 *dissector_mask,
1141 			    u32 flower_flag_bit, u32 dissector_flag_bit)
1142 {
1143 	if (flower_mask & flower_flag_bit) {
1144 		*dissector_mask |= dissector_flag_bit;
1145 		if (flower_key & flower_flag_bit)
1146 			*dissector_key |= dissector_flag_bit;
1147 	}
1148 }
1149 
1150 static int fl_set_key_flags(struct nlattr **tb, u32 *flags_key,
1151 			    u32 *flags_mask, struct netlink_ext_ack *extack)
1152 {
1153 	u32 key, mask;
1154 
1155 	/* mask is mandatory for flags */
1156 	if (!tb[TCA_FLOWER_KEY_FLAGS_MASK]) {
1157 		NL_SET_ERR_MSG(extack, "Missing flags mask");
1158 		return -EINVAL;
1159 	}
1160 
1161 	key = be32_to_cpu(nla_get_be32(tb[TCA_FLOWER_KEY_FLAGS]));
1162 	mask = be32_to_cpu(nla_get_be32(tb[TCA_FLOWER_KEY_FLAGS_MASK]));
1163 
1164 	*flags_key  = 0;
1165 	*flags_mask = 0;
1166 
1167 	fl_set_key_flag(key, mask, flags_key, flags_mask,
1168 			TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
1169 	fl_set_key_flag(key, mask, flags_key, flags_mask,
1170 			TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
1171 			FLOW_DIS_FIRST_FRAG);
1172 
1173 	return 0;
1174 }
1175 
1176 static void fl_set_key_ip(struct nlattr **tb, bool encap,
1177 			  struct flow_dissector_key_ip *key,
1178 			  struct flow_dissector_key_ip *mask)
1179 {
1180 	int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
1181 	int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
1182 	int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
1183 	int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
1184 
1185 	fl_set_key_val(tb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos));
1186 	fl_set_key_val(tb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl));
1187 }
1188 
1189 static int fl_set_geneve_opt(const struct nlattr *nla, struct fl_flow_key *key,
1190 			     int depth, int option_len,
1191 			     struct netlink_ext_ack *extack)
1192 {
1193 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1];
1194 	struct nlattr *class = NULL, *type = NULL, *data = NULL;
1195 	struct geneve_opt *opt;
1196 	int err, data_len = 0;
1197 
1198 	if (option_len > sizeof(struct geneve_opt))
1199 		data_len = option_len - sizeof(struct geneve_opt);
1200 
1201 	if (key->enc_opts.len > FLOW_DIS_TUN_OPTS_MAX - 4)
1202 		return -ERANGE;
1203 
1204 	opt = (struct geneve_opt *)&key->enc_opts.data[key->enc_opts.len];
1205 	memset(opt, 0xff, option_len);
1206 	opt->length = data_len / 4;
1207 	opt->r1 = 0;
1208 	opt->r2 = 0;
1209 	opt->r3 = 0;
1210 
1211 	/* If no mask has been prodived we assume an exact match. */
1212 	if (!depth)
1213 		return sizeof(struct geneve_opt) + data_len;
1214 
1215 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_GENEVE) {
1216 		NL_SET_ERR_MSG(extack, "Non-geneve option type for mask");
1217 		return -EINVAL;
1218 	}
1219 
1220 	err = nla_parse_nested_deprecated(tb,
1221 					  TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX,
1222 					  nla, geneve_opt_policy, extack);
1223 	if (err < 0)
1224 		return err;
1225 
1226 	/* We are not allowed to omit any of CLASS, TYPE or DATA
1227 	 * fields from the key.
1228 	 */
1229 	if (!option_len &&
1230 	    (!tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS] ||
1231 	     !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE] ||
1232 	     !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA])) {
1233 		NL_SET_ERR_MSG(extack, "Missing tunnel key geneve option class, type or data");
1234 		return -EINVAL;
1235 	}
1236 
1237 	/* Omitting any of CLASS, TYPE or DATA fields is allowed
1238 	 * for the mask.
1239 	 */
1240 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA]) {
1241 		int new_len = key->enc_opts.len;
1242 
1243 		data = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA];
1244 		data_len = nla_len(data);
1245 		if (data_len < 4) {
1246 			NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is less than 4 bytes long");
1247 			return -ERANGE;
1248 		}
1249 		if (data_len % 4) {
1250 			NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is not a multiple of 4 bytes long");
1251 			return -ERANGE;
1252 		}
1253 
1254 		new_len += sizeof(struct geneve_opt) + data_len;
1255 		BUILD_BUG_ON(FLOW_DIS_TUN_OPTS_MAX != IP_TUNNEL_OPTS_MAX);
1256 		if (new_len > FLOW_DIS_TUN_OPTS_MAX) {
1257 			NL_SET_ERR_MSG(extack, "Tunnel options exceeds max size");
1258 			return -ERANGE;
1259 		}
1260 		opt->length = data_len / 4;
1261 		memcpy(opt->opt_data, nla_data(data), data_len);
1262 	}
1263 
1264 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS]) {
1265 		class = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS];
1266 		opt->opt_class = nla_get_be16(class);
1267 	}
1268 
1269 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE]) {
1270 		type = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE];
1271 		opt->type = nla_get_u8(type);
1272 	}
1273 
1274 	return sizeof(struct geneve_opt) + data_len;
1275 }
1276 
1277 static int fl_set_vxlan_opt(const struct nlattr *nla, struct fl_flow_key *key,
1278 			    int depth, int option_len,
1279 			    struct netlink_ext_ack *extack)
1280 {
1281 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX + 1];
1282 	struct vxlan_metadata *md;
1283 	int err;
1284 
1285 	md = (struct vxlan_metadata *)&key->enc_opts.data[key->enc_opts.len];
1286 	memset(md, 0xff, sizeof(*md));
1287 
1288 	if (!depth)
1289 		return sizeof(*md);
1290 
1291 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_VXLAN) {
1292 		NL_SET_ERR_MSG(extack, "Non-vxlan option type for mask");
1293 		return -EINVAL;
1294 	}
1295 
1296 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX, nla,
1297 			       vxlan_opt_policy, extack);
1298 	if (err < 0)
1299 		return err;
1300 
1301 	if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]) {
1302 		NL_SET_ERR_MSG(extack, "Missing tunnel key vxlan option gbp");
1303 		return -EINVAL;
1304 	}
1305 
1306 	if (tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]) {
1307 		md->gbp = nla_get_u32(tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]);
1308 		md->gbp &= VXLAN_GBP_MASK;
1309 	}
1310 
1311 	return sizeof(*md);
1312 }
1313 
1314 static int fl_set_erspan_opt(const struct nlattr *nla, struct fl_flow_key *key,
1315 			     int depth, int option_len,
1316 			     struct netlink_ext_ack *extack)
1317 {
1318 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX + 1];
1319 	struct erspan_metadata *md;
1320 	int err;
1321 
1322 	md = (struct erspan_metadata *)&key->enc_opts.data[key->enc_opts.len];
1323 	md->version = 1;
1324 
1325 	if (!depth)
1326 		return sizeof(*md);
1327 
1328 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_ERSPAN) {
1329 		NL_SET_ERR_MSG(extack, "Non-erspan option type for mask");
1330 		return -EINVAL;
1331 	}
1332 
1333 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX, nla,
1334 			       erspan_opt_policy, extack);
1335 	if (err < 0)
1336 		return err;
1337 
1338 	if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]) {
1339 		NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option ver");
1340 		return -EINVAL;
1341 	}
1342 
1343 	if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER])
1344 		md->version = nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]);
1345 
1346 	if (md->version == 1) {
1347 		if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]) {
1348 			NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option index");
1349 			return -EINVAL;
1350 		}
1351 		memset(&md->u.index, 0xff, sizeof(md->u.index));
1352 		if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]) {
1353 			nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX];
1354 			md->u.index = nla_get_be32(nla);
1355 		}
1356 	} else if (md->version == 2) {
1357 		if (!option_len && (!tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR] ||
1358 				    !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID])) {
1359 			NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option dir or hwid");
1360 			return -EINVAL;
1361 		}
1362 		md->u.md2.dir = 1;
1363 		if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR]) {
1364 			nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR];
1365 			md->u.md2.dir = nla_get_u8(nla);
1366 		}
1367 		set_hwid(&md->u.md2, 0xff);
1368 		if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID]) {
1369 			nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID];
1370 			set_hwid(&md->u.md2, nla_get_u8(nla));
1371 		}
1372 	} else {
1373 		NL_SET_ERR_MSG(extack, "Tunnel key erspan option ver is incorrect");
1374 		return -EINVAL;
1375 	}
1376 
1377 	return sizeof(*md);
1378 }
1379 
1380 static int fl_set_gtp_opt(const struct nlattr *nla, struct fl_flow_key *key,
1381 			  int depth, int option_len,
1382 			  struct netlink_ext_ack *extack)
1383 {
1384 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_GTP_MAX + 1];
1385 	struct gtp_pdu_session_info *sinfo;
1386 	u8 len = key->enc_opts.len;
1387 	int err;
1388 
1389 	sinfo = (struct gtp_pdu_session_info *)&key->enc_opts.data[len];
1390 	memset(sinfo, 0xff, option_len);
1391 
1392 	if (!depth)
1393 		return sizeof(*sinfo);
1394 
1395 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_GTP) {
1396 		NL_SET_ERR_MSG_MOD(extack, "Non-gtp option type for mask");
1397 		return -EINVAL;
1398 	}
1399 
1400 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_GTP_MAX, nla,
1401 			       gtp_opt_policy, extack);
1402 	if (err < 0)
1403 		return err;
1404 
1405 	if (!option_len &&
1406 	    (!tb[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE] ||
1407 	     !tb[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI])) {
1408 		NL_SET_ERR_MSG_MOD(extack,
1409 				   "Missing tunnel key gtp option pdu type or qfi");
1410 		return -EINVAL;
1411 	}
1412 
1413 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE])
1414 		sinfo->pdu_type =
1415 			nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE]);
1416 
1417 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI])
1418 		sinfo->qfi = nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI]);
1419 
1420 	return sizeof(*sinfo);
1421 }
1422 
1423 static int fl_set_enc_opt(struct nlattr **tb, struct fl_flow_key *key,
1424 			  struct fl_flow_key *mask,
1425 			  struct netlink_ext_ack *extack)
1426 {
1427 	const struct nlattr *nla_enc_key, *nla_opt_key, *nla_opt_msk = NULL;
1428 	int err, option_len, key_depth, msk_depth = 0;
1429 
1430 	err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS],
1431 					     TCA_FLOWER_KEY_ENC_OPTS_MAX,
1432 					     enc_opts_policy, extack);
1433 	if (err)
1434 		return err;
1435 
1436 	nla_enc_key = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS]);
1437 
1438 	if (tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]) {
1439 		err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK],
1440 						     TCA_FLOWER_KEY_ENC_OPTS_MAX,
1441 						     enc_opts_policy, extack);
1442 		if (err)
1443 			return err;
1444 
1445 		nla_opt_msk = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
1446 		msk_depth = nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
1447 		if (!nla_ok(nla_opt_msk, msk_depth)) {
1448 			NL_SET_ERR_MSG(extack, "Invalid nested attribute for masks");
1449 			return -EINVAL;
1450 		}
1451 	}
1452 
1453 	nla_for_each_attr(nla_opt_key, nla_enc_key,
1454 			  nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS]), key_depth) {
1455 		switch (nla_type(nla_opt_key)) {
1456 		case TCA_FLOWER_KEY_ENC_OPTS_GENEVE:
1457 			if (key->enc_opts.dst_opt_type &&
1458 			    key->enc_opts.dst_opt_type != TUNNEL_GENEVE_OPT) {
1459 				NL_SET_ERR_MSG(extack, "Duplicate type for geneve options");
1460 				return -EINVAL;
1461 			}
1462 			option_len = 0;
1463 			key->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT;
1464 			option_len = fl_set_geneve_opt(nla_opt_key, key,
1465 						       key_depth, option_len,
1466 						       extack);
1467 			if (option_len < 0)
1468 				return option_len;
1469 
1470 			key->enc_opts.len += option_len;
1471 			/* At the same time we need to parse through the mask
1472 			 * in order to verify exact and mask attribute lengths.
1473 			 */
1474 			mask->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT;
1475 			option_len = fl_set_geneve_opt(nla_opt_msk, mask,
1476 						       msk_depth, option_len,
1477 						       extack);
1478 			if (option_len < 0)
1479 				return option_len;
1480 
1481 			mask->enc_opts.len += option_len;
1482 			if (key->enc_opts.len != mask->enc_opts.len) {
1483 				NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1484 				return -EINVAL;
1485 			}
1486 			break;
1487 		case TCA_FLOWER_KEY_ENC_OPTS_VXLAN:
1488 			if (key->enc_opts.dst_opt_type) {
1489 				NL_SET_ERR_MSG(extack, "Duplicate type for vxlan options");
1490 				return -EINVAL;
1491 			}
1492 			option_len = 0;
1493 			key->enc_opts.dst_opt_type = TUNNEL_VXLAN_OPT;
1494 			option_len = fl_set_vxlan_opt(nla_opt_key, key,
1495 						      key_depth, option_len,
1496 						      extack);
1497 			if (option_len < 0)
1498 				return option_len;
1499 
1500 			key->enc_opts.len += option_len;
1501 			/* At the same time we need to parse through the mask
1502 			 * in order to verify exact and mask attribute lengths.
1503 			 */
1504 			mask->enc_opts.dst_opt_type = TUNNEL_VXLAN_OPT;
1505 			option_len = fl_set_vxlan_opt(nla_opt_msk, mask,
1506 						      msk_depth, option_len,
1507 						      extack);
1508 			if (option_len < 0)
1509 				return option_len;
1510 
1511 			mask->enc_opts.len += option_len;
1512 			if (key->enc_opts.len != mask->enc_opts.len) {
1513 				NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1514 				return -EINVAL;
1515 			}
1516 			break;
1517 		case TCA_FLOWER_KEY_ENC_OPTS_ERSPAN:
1518 			if (key->enc_opts.dst_opt_type) {
1519 				NL_SET_ERR_MSG(extack, "Duplicate type for erspan options");
1520 				return -EINVAL;
1521 			}
1522 			option_len = 0;
1523 			key->enc_opts.dst_opt_type = TUNNEL_ERSPAN_OPT;
1524 			option_len = fl_set_erspan_opt(nla_opt_key, key,
1525 						       key_depth, option_len,
1526 						       extack);
1527 			if (option_len < 0)
1528 				return option_len;
1529 
1530 			key->enc_opts.len += option_len;
1531 			/* At the same time we need to parse through the mask
1532 			 * in order to verify exact and mask attribute lengths.
1533 			 */
1534 			mask->enc_opts.dst_opt_type = TUNNEL_ERSPAN_OPT;
1535 			option_len = fl_set_erspan_opt(nla_opt_msk, mask,
1536 						       msk_depth, option_len,
1537 						       extack);
1538 			if (option_len < 0)
1539 				return option_len;
1540 
1541 			mask->enc_opts.len += option_len;
1542 			if (key->enc_opts.len != mask->enc_opts.len) {
1543 				NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1544 				return -EINVAL;
1545 			}
1546 			break;
1547 		case TCA_FLOWER_KEY_ENC_OPTS_GTP:
1548 			if (key->enc_opts.dst_opt_type) {
1549 				NL_SET_ERR_MSG_MOD(extack,
1550 						   "Duplicate type for gtp options");
1551 				return -EINVAL;
1552 			}
1553 			option_len = 0;
1554 			key->enc_opts.dst_opt_type = TUNNEL_GTP_OPT;
1555 			option_len = fl_set_gtp_opt(nla_opt_key, key,
1556 						    key_depth, option_len,
1557 						    extack);
1558 			if (option_len < 0)
1559 				return option_len;
1560 
1561 			key->enc_opts.len += option_len;
1562 			/* At the same time we need to parse through the mask
1563 			 * in order to verify exact and mask attribute lengths.
1564 			 */
1565 			mask->enc_opts.dst_opt_type = TUNNEL_GTP_OPT;
1566 			option_len = fl_set_gtp_opt(nla_opt_msk, mask,
1567 						    msk_depth, option_len,
1568 						    extack);
1569 			if (option_len < 0)
1570 				return option_len;
1571 
1572 			mask->enc_opts.len += option_len;
1573 			if (key->enc_opts.len != mask->enc_opts.len) {
1574 				NL_SET_ERR_MSG_MOD(extack,
1575 						   "Key and mask miss aligned");
1576 				return -EINVAL;
1577 			}
1578 			break;
1579 		default:
1580 			NL_SET_ERR_MSG(extack, "Unknown tunnel option type");
1581 			return -EINVAL;
1582 		}
1583 
1584 		if (!msk_depth)
1585 			continue;
1586 
1587 		if (!nla_ok(nla_opt_msk, msk_depth)) {
1588 			NL_SET_ERR_MSG(extack, "A mask attribute is invalid");
1589 			return -EINVAL;
1590 		}
1591 		nla_opt_msk = nla_next(nla_opt_msk, &msk_depth);
1592 	}
1593 
1594 	return 0;
1595 }
1596 
1597 static int fl_validate_ct_state(u16 state, struct nlattr *tb,
1598 				struct netlink_ext_ack *extack)
1599 {
1600 	if (state && !(state & TCA_FLOWER_KEY_CT_FLAGS_TRACKED)) {
1601 		NL_SET_ERR_MSG_ATTR(extack, tb,
1602 				    "no trk, so no other flag can be set");
1603 		return -EINVAL;
1604 	}
1605 
1606 	if (state & TCA_FLOWER_KEY_CT_FLAGS_NEW &&
1607 	    state & TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED) {
1608 		NL_SET_ERR_MSG_ATTR(extack, tb,
1609 				    "new and est are mutually exclusive");
1610 		return -EINVAL;
1611 	}
1612 
1613 	if (state & TCA_FLOWER_KEY_CT_FLAGS_INVALID &&
1614 	    state & ~(TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
1615 		      TCA_FLOWER_KEY_CT_FLAGS_INVALID)) {
1616 		NL_SET_ERR_MSG_ATTR(extack, tb,
1617 				    "when inv is set, only trk may be set");
1618 		return -EINVAL;
1619 	}
1620 
1621 	if (state & TCA_FLOWER_KEY_CT_FLAGS_NEW &&
1622 	    state & TCA_FLOWER_KEY_CT_FLAGS_REPLY) {
1623 		NL_SET_ERR_MSG_ATTR(extack, tb,
1624 				    "new and rpl are mutually exclusive");
1625 		return -EINVAL;
1626 	}
1627 
1628 	return 0;
1629 }
1630 
1631 static int fl_set_key_ct(struct nlattr **tb,
1632 			 struct flow_dissector_key_ct *key,
1633 			 struct flow_dissector_key_ct *mask,
1634 			 struct netlink_ext_ack *extack)
1635 {
1636 	if (tb[TCA_FLOWER_KEY_CT_STATE]) {
1637 		int err;
1638 
1639 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK)) {
1640 			NL_SET_ERR_MSG(extack, "Conntrack isn't enabled");
1641 			return -EOPNOTSUPP;
1642 		}
1643 		fl_set_key_val(tb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE,
1644 			       &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK,
1645 			       sizeof(key->ct_state));
1646 
1647 		err = fl_validate_ct_state(key->ct_state & mask->ct_state,
1648 					   tb[TCA_FLOWER_KEY_CT_STATE_MASK],
1649 					   extack);
1650 		if (err)
1651 			return err;
1652 
1653 	}
1654 	if (tb[TCA_FLOWER_KEY_CT_ZONE]) {
1655 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)) {
1656 			NL_SET_ERR_MSG(extack, "Conntrack zones isn't enabled");
1657 			return -EOPNOTSUPP;
1658 		}
1659 		fl_set_key_val(tb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE,
1660 			       &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK,
1661 			       sizeof(key->ct_zone));
1662 	}
1663 	if (tb[TCA_FLOWER_KEY_CT_MARK]) {
1664 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)) {
1665 			NL_SET_ERR_MSG(extack, "Conntrack mark isn't enabled");
1666 			return -EOPNOTSUPP;
1667 		}
1668 		fl_set_key_val(tb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK,
1669 			       &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK,
1670 			       sizeof(key->ct_mark));
1671 	}
1672 	if (tb[TCA_FLOWER_KEY_CT_LABELS]) {
1673 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)) {
1674 			NL_SET_ERR_MSG(extack, "Conntrack labels aren't enabled");
1675 			return -EOPNOTSUPP;
1676 		}
1677 		fl_set_key_val(tb, key->ct_labels, TCA_FLOWER_KEY_CT_LABELS,
1678 			       mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK,
1679 			       sizeof(key->ct_labels));
1680 	}
1681 
1682 	return 0;
1683 }
1684 
1685 static bool is_vlan_key(struct nlattr *tb, __be16 *ethertype,
1686 			struct fl_flow_key *key, struct fl_flow_key *mask,
1687 			int vthresh)
1688 {
1689 	const bool good_num_of_vlans = key->num_of_vlans.num_of_vlans > vthresh;
1690 
1691 	if (!tb) {
1692 		*ethertype = 0;
1693 		return good_num_of_vlans;
1694 	}
1695 
1696 	*ethertype = nla_get_be16(tb);
1697 	if (good_num_of_vlans || eth_type_vlan(*ethertype))
1698 		return true;
1699 
1700 	key->basic.n_proto = *ethertype;
1701 	mask->basic.n_proto = cpu_to_be16(~0);
1702 	return false;
1703 }
1704 
1705 static void fl_set_key_cfm_md_level(struct nlattr **tb,
1706 				    struct fl_flow_key *key,
1707 				    struct fl_flow_key *mask,
1708 				    struct netlink_ext_ack *extack)
1709 {
1710 	u8 level;
1711 
1712 	if (!tb[TCA_FLOWER_KEY_CFM_MD_LEVEL])
1713 		return;
1714 
1715 	level = nla_get_u8(tb[TCA_FLOWER_KEY_CFM_MD_LEVEL]);
1716 	key->cfm.mdl_ver = FIELD_PREP(FLOW_DIS_CFM_MDL_MASK, level);
1717 	mask->cfm.mdl_ver = FLOW_DIS_CFM_MDL_MASK;
1718 }
1719 
1720 static void fl_set_key_cfm_opcode(struct nlattr **tb,
1721 				  struct fl_flow_key *key,
1722 				  struct fl_flow_key *mask,
1723 				  struct netlink_ext_ack *extack)
1724 {
1725 	fl_set_key_val(tb, &key->cfm.opcode, TCA_FLOWER_KEY_CFM_OPCODE,
1726 		       &mask->cfm.opcode, TCA_FLOWER_UNSPEC,
1727 		       sizeof(key->cfm.opcode));
1728 }
1729 
1730 static int fl_set_key_cfm(struct nlattr **tb,
1731 			  struct fl_flow_key *key,
1732 			  struct fl_flow_key *mask,
1733 			  struct netlink_ext_ack *extack)
1734 {
1735 	struct nlattr *nla_cfm_opt[TCA_FLOWER_KEY_CFM_OPT_MAX + 1];
1736 	int err;
1737 
1738 	if (!tb[TCA_FLOWER_KEY_CFM])
1739 		return 0;
1740 
1741 	err = nla_parse_nested(nla_cfm_opt, TCA_FLOWER_KEY_CFM_OPT_MAX,
1742 			       tb[TCA_FLOWER_KEY_CFM], cfm_opt_policy, extack);
1743 	if (err < 0)
1744 		return err;
1745 
1746 	fl_set_key_cfm_opcode(nla_cfm_opt, key, mask, extack);
1747 	fl_set_key_cfm_md_level(nla_cfm_opt, key, mask, extack);
1748 
1749 	return 0;
1750 }
1751 
1752 static int fl_set_key(struct net *net, struct nlattr **tb,
1753 		      struct fl_flow_key *key, struct fl_flow_key *mask,
1754 		      struct netlink_ext_ack *extack)
1755 {
1756 	__be16 ethertype;
1757 	int ret = 0;
1758 
1759 	if (tb[TCA_FLOWER_INDEV]) {
1760 		int err = tcf_change_indev(net, tb[TCA_FLOWER_INDEV], extack);
1761 		if (err < 0)
1762 			return err;
1763 		key->meta.ingress_ifindex = err;
1764 		mask->meta.ingress_ifindex = 0xffffffff;
1765 	}
1766 
1767 	fl_set_key_val(tb, &key->meta.l2_miss, TCA_FLOWER_L2_MISS,
1768 		       &mask->meta.l2_miss, TCA_FLOWER_UNSPEC,
1769 		       sizeof(key->meta.l2_miss));
1770 
1771 	fl_set_key_val(tb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
1772 		       mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
1773 		       sizeof(key->eth.dst));
1774 	fl_set_key_val(tb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
1775 		       mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
1776 		       sizeof(key->eth.src));
1777 	fl_set_key_val(tb, &key->num_of_vlans,
1778 		       TCA_FLOWER_KEY_NUM_OF_VLANS,
1779 		       &mask->num_of_vlans,
1780 		       TCA_FLOWER_UNSPEC,
1781 		       sizeof(key->num_of_vlans));
1782 
1783 	if (is_vlan_key(tb[TCA_FLOWER_KEY_ETH_TYPE], &ethertype, key, mask, 0)) {
1784 		fl_set_key_vlan(tb, ethertype, TCA_FLOWER_KEY_VLAN_ID,
1785 				TCA_FLOWER_KEY_VLAN_PRIO,
1786 				TCA_FLOWER_KEY_VLAN_ETH_TYPE,
1787 				&key->vlan, &mask->vlan);
1788 
1789 		if (is_vlan_key(tb[TCA_FLOWER_KEY_VLAN_ETH_TYPE],
1790 				&ethertype, key, mask, 1)) {
1791 			fl_set_key_vlan(tb, ethertype,
1792 					TCA_FLOWER_KEY_CVLAN_ID,
1793 					TCA_FLOWER_KEY_CVLAN_PRIO,
1794 					TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
1795 					&key->cvlan, &mask->cvlan);
1796 			fl_set_key_val(tb, &key->basic.n_proto,
1797 				       TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
1798 				       &mask->basic.n_proto,
1799 				       TCA_FLOWER_UNSPEC,
1800 				       sizeof(key->basic.n_proto));
1801 		}
1802 	}
1803 
1804 	if (key->basic.n_proto == htons(ETH_P_PPP_SES))
1805 		fl_set_key_pppoe(tb, &key->pppoe, &mask->pppoe, key, mask);
1806 
1807 	if (key->basic.n_proto == htons(ETH_P_IP) ||
1808 	    key->basic.n_proto == htons(ETH_P_IPV6)) {
1809 		fl_set_key_val(tb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
1810 			       &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
1811 			       sizeof(key->basic.ip_proto));
1812 		fl_set_key_ip(tb, false, &key->ip, &mask->ip);
1813 	}
1814 
1815 	if (tb[TCA_FLOWER_KEY_IPV4_SRC] || tb[TCA_FLOWER_KEY_IPV4_DST]) {
1816 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1817 		mask->control.addr_type = ~0;
1818 		fl_set_key_val(tb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
1819 			       &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
1820 			       sizeof(key->ipv4.src));
1821 		fl_set_key_val(tb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
1822 			       &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
1823 			       sizeof(key->ipv4.dst));
1824 	} else if (tb[TCA_FLOWER_KEY_IPV6_SRC] || tb[TCA_FLOWER_KEY_IPV6_DST]) {
1825 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1826 		mask->control.addr_type = ~0;
1827 		fl_set_key_val(tb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
1828 			       &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
1829 			       sizeof(key->ipv6.src));
1830 		fl_set_key_val(tb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
1831 			       &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
1832 			       sizeof(key->ipv6.dst));
1833 	}
1834 
1835 	if (key->basic.ip_proto == IPPROTO_TCP) {
1836 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
1837 			       &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
1838 			       sizeof(key->tp.src));
1839 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
1840 			       &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
1841 			       sizeof(key->tp.dst));
1842 		fl_set_key_val(tb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
1843 			       &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
1844 			       sizeof(key->tcp.flags));
1845 	} else if (key->basic.ip_proto == IPPROTO_UDP) {
1846 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
1847 			       &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
1848 			       sizeof(key->tp.src));
1849 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
1850 			       &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
1851 			       sizeof(key->tp.dst));
1852 	} else if (key->basic.ip_proto == IPPROTO_SCTP) {
1853 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
1854 			       &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
1855 			       sizeof(key->tp.src));
1856 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
1857 			       &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
1858 			       sizeof(key->tp.dst));
1859 	} else if (key->basic.n_proto == htons(ETH_P_IP) &&
1860 		   key->basic.ip_proto == IPPROTO_ICMP) {
1861 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV4_TYPE,
1862 			       &mask->icmp.type,
1863 			       TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
1864 			       sizeof(key->icmp.type));
1865 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV4_CODE,
1866 			       &mask->icmp.code,
1867 			       TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
1868 			       sizeof(key->icmp.code));
1869 	} else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
1870 		   key->basic.ip_proto == IPPROTO_ICMPV6) {
1871 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV6_TYPE,
1872 			       &mask->icmp.type,
1873 			       TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
1874 			       sizeof(key->icmp.type));
1875 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV6_CODE,
1876 			       &mask->icmp.code,
1877 			       TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
1878 			       sizeof(key->icmp.code));
1879 	} else if (key->basic.n_proto == htons(ETH_P_MPLS_UC) ||
1880 		   key->basic.n_proto == htons(ETH_P_MPLS_MC)) {
1881 		ret = fl_set_key_mpls(tb, &key->mpls, &mask->mpls, extack);
1882 		if (ret)
1883 			return ret;
1884 	} else if (key->basic.n_proto == htons(ETH_P_ARP) ||
1885 		   key->basic.n_proto == htons(ETH_P_RARP)) {
1886 		fl_set_key_val(tb, &key->arp.sip, TCA_FLOWER_KEY_ARP_SIP,
1887 			       &mask->arp.sip, TCA_FLOWER_KEY_ARP_SIP_MASK,
1888 			       sizeof(key->arp.sip));
1889 		fl_set_key_val(tb, &key->arp.tip, TCA_FLOWER_KEY_ARP_TIP,
1890 			       &mask->arp.tip, TCA_FLOWER_KEY_ARP_TIP_MASK,
1891 			       sizeof(key->arp.tip));
1892 		fl_set_key_val(tb, &key->arp.op, TCA_FLOWER_KEY_ARP_OP,
1893 			       &mask->arp.op, TCA_FLOWER_KEY_ARP_OP_MASK,
1894 			       sizeof(key->arp.op));
1895 		fl_set_key_val(tb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
1896 			       mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
1897 			       sizeof(key->arp.sha));
1898 		fl_set_key_val(tb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
1899 			       mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
1900 			       sizeof(key->arp.tha));
1901 	} else if (key->basic.ip_proto == IPPROTO_L2TP) {
1902 		fl_set_key_val(tb, &key->l2tpv3.session_id,
1903 			       TCA_FLOWER_KEY_L2TPV3_SID,
1904 			       &mask->l2tpv3.session_id, TCA_FLOWER_UNSPEC,
1905 			       sizeof(key->l2tpv3.session_id));
1906 	} else if (key->basic.n_proto  == htons(ETH_P_CFM)) {
1907 		ret = fl_set_key_cfm(tb, key, mask, extack);
1908 		if (ret)
1909 			return ret;
1910 	}
1911 
1912 	if (key->basic.ip_proto == IPPROTO_TCP ||
1913 	    key->basic.ip_proto == IPPROTO_UDP ||
1914 	    key->basic.ip_proto == IPPROTO_SCTP) {
1915 		ret = fl_set_key_port_range(tb, key, mask, extack);
1916 		if (ret)
1917 			return ret;
1918 	}
1919 
1920 	if (tb[TCA_FLOWER_KEY_SPI]) {
1921 		ret = fl_set_key_spi(tb, key, mask, extack);
1922 		if (ret)
1923 			return ret;
1924 	}
1925 
1926 	if (tb[TCA_FLOWER_KEY_ENC_IPV4_SRC] ||
1927 	    tb[TCA_FLOWER_KEY_ENC_IPV4_DST]) {
1928 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1929 		mask->enc_control.addr_type = ~0;
1930 		fl_set_key_val(tb, &key->enc_ipv4.src,
1931 			       TCA_FLOWER_KEY_ENC_IPV4_SRC,
1932 			       &mask->enc_ipv4.src,
1933 			       TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
1934 			       sizeof(key->enc_ipv4.src));
1935 		fl_set_key_val(tb, &key->enc_ipv4.dst,
1936 			       TCA_FLOWER_KEY_ENC_IPV4_DST,
1937 			       &mask->enc_ipv4.dst,
1938 			       TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
1939 			       sizeof(key->enc_ipv4.dst));
1940 	}
1941 
1942 	if (tb[TCA_FLOWER_KEY_ENC_IPV6_SRC] ||
1943 	    tb[TCA_FLOWER_KEY_ENC_IPV6_DST]) {
1944 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1945 		mask->enc_control.addr_type = ~0;
1946 		fl_set_key_val(tb, &key->enc_ipv6.src,
1947 			       TCA_FLOWER_KEY_ENC_IPV6_SRC,
1948 			       &mask->enc_ipv6.src,
1949 			       TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
1950 			       sizeof(key->enc_ipv6.src));
1951 		fl_set_key_val(tb, &key->enc_ipv6.dst,
1952 			       TCA_FLOWER_KEY_ENC_IPV6_DST,
1953 			       &mask->enc_ipv6.dst,
1954 			       TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
1955 			       sizeof(key->enc_ipv6.dst));
1956 	}
1957 
1958 	fl_set_key_val(tb, &key->enc_key_id.keyid, TCA_FLOWER_KEY_ENC_KEY_ID,
1959 		       &mask->enc_key_id.keyid, TCA_FLOWER_UNSPEC,
1960 		       sizeof(key->enc_key_id.keyid));
1961 
1962 	fl_set_key_val(tb, &key->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
1963 		       &mask->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
1964 		       sizeof(key->enc_tp.src));
1965 
1966 	fl_set_key_val(tb, &key->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
1967 		       &mask->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
1968 		       sizeof(key->enc_tp.dst));
1969 
1970 	fl_set_key_ip(tb, true, &key->enc_ip, &mask->enc_ip);
1971 
1972 	fl_set_key_val(tb, &key->hash.hash, TCA_FLOWER_KEY_HASH,
1973 		       &mask->hash.hash, TCA_FLOWER_KEY_HASH_MASK,
1974 		       sizeof(key->hash.hash));
1975 
1976 	if (tb[TCA_FLOWER_KEY_ENC_OPTS]) {
1977 		ret = fl_set_enc_opt(tb, key, mask, extack);
1978 		if (ret)
1979 			return ret;
1980 	}
1981 
1982 	ret = fl_set_key_ct(tb, &key->ct, &mask->ct, extack);
1983 	if (ret)
1984 		return ret;
1985 
1986 	if (tb[TCA_FLOWER_KEY_FLAGS])
1987 		ret = fl_set_key_flags(tb, &key->control.flags,
1988 				       &mask->control.flags, extack);
1989 
1990 	return ret;
1991 }
1992 
1993 static void fl_mask_copy(struct fl_flow_mask *dst,
1994 			 struct fl_flow_mask *src)
1995 {
1996 	const void *psrc = fl_key_get_start(&src->key, src);
1997 	void *pdst = fl_key_get_start(&dst->key, src);
1998 
1999 	memcpy(pdst, psrc, fl_mask_range(src));
2000 	dst->range = src->range;
2001 }
2002 
2003 static const struct rhashtable_params fl_ht_params = {
2004 	.key_offset = offsetof(struct cls_fl_filter, mkey), /* base offset */
2005 	.head_offset = offsetof(struct cls_fl_filter, ht_node),
2006 	.automatic_shrinking = true,
2007 };
2008 
2009 static int fl_init_mask_hashtable(struct fl_flow_mask *mask)
2010 {
2011 	mask->filter_ht_params = fl_ht_params;
2012 	mask->filter_ht_params.key_len = fl_mask_range(mask);
2013 	mask->filter_ht_params.key_offset += mask->range.start;
2014 
2015 	return rhashtable_init(&mask->ht, &mask->filter_ht_params);
2016 }
2017 
2018 #define FL_KEY_MEMBER_OFFSET(member) offsetof(struct fl_flow_key, member)
2019 #define FL_KEY_MEMBER_SIZE(member) sizeof_field(struct fl_flow_key, member)
2020 
2021 #define FL_KEY_IS_MASKED(mask, member)						\
2022 	memchr_inv(((char *)mask) + FL_KEY_MEMBER_OFFSET(member),		\
2023 		   0, FL_KEY_MEMBER_SIZE(member))				\
2024 
2025 #define FL_KEY_SET(keys, cnt, id, member)					\
2026 	do {									\
2027 		keys[cnt].key_id = id;						\
2028 		keys[cnt].offset = FL_KEY_MEMBER_OFFSET(member);		\
2029 		cnt++;								\
2030 	} while(0);
2031 
2032 #define FL_KEY_SET_IF_MASKED(mask, keys, cnt, id, member)			\
2033 	do {									\
2034 		if (FL_KEY_IS_MASKED(mask, member))				\
2035 			FL_KEY_SET(keys, cnt, id, member);			\
2036 	} while(0);
2037 
2038 static void fl_init_dissector(struct flow_dissector *dissector,
2039 			      struct fl_flow_key *mask)
2040 {
2041 	struct flow_dissector_key keys[FLOW_DISSECTOR_KEY_MAX];
2042 	size_t cnt = 0;
2043 
2044 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2045 			     FLOW_DISSECTOR_KEY_META, meta);
2046 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_CONTROL, control);
2047 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_BASIC, basic);
2048 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2049 			     FLOW_DISSECTOR_KEY_ETH_ADDRS, eth);
2050 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2051 			     FLOW_DISSECTOR_KEY_IPV4_ADDRS, ipv4);
2052 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2053 			     FLOW_DISSECTOR_KEY_IPV6_ADDRS, ipv6);
2054 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2055 			     FLOW_DISSECTOR_KEY_PORTS, tp);
2056 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2057 			     FLOW_DISSECTOR_KEY_PORTS_RANGE, tp_range);
2058 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2059 			     FLOW_DISSECTOR_KEY_IP, ip);
2060 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2061 			     FLOW_DISSECTOR_KEY_TCP, tcp);
2062 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2063 			     FLOW_DISSECTOR_KEY_ICMP, icmp);
2064 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2065 			     FLOW_DISSECTOR_KEY_ARP, arp);
2066 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2067 			     FLOW_DISSECTOR_KEY_MPLS, mpls);
2068 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2069 			     FLOW_DISSECTOR_KEY_VLAN, vlan);
2070 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2071 			     FLOW_DISSECTOR_KEY_CVLAN, cvlan);
2072 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2073 			     FLOW_DISSECTOR_KEY_ENC_KEYID, enc_key_id);
2074 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2075 			     FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, enc_ipv4);
2076 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2077 			     FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, enc_ipv6);
2078 	if (FL_KEY_IS_MASKED(mask, enc_ipv4) ||
2079 	    FL_KEY_IS_MASKED(mask, enc_ipv6))
2080 		FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_ENC_CONTROL,
2081 			   enc_control);
2082 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2083 			     FLOW_DISSECTOR_KEY_ENC_PORTS, enc_tp);
2084 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2085 			     FLOW_DISSECTOR_KEY_ENC_IP, enc_ip);
2086 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2087 			     FLOW_DISSECTOR_KEY_ENC_OPTS, enc_opts);
2088 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2089 			     FLOW_DISSECTOR_KEY_CT, ct);
2090 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2091 			     FLOW_DISSECTOR_KEY_HASH, hash);
2092 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2093 			     FLOW_DISSECTOR_KEY_NUM_OF_VLANS, num_of_vlans);
2094 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2095 			     FLOW_DISSECTOR_KEY_PPPOE, pppoe);
2096 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2097 			     FLOW_DISSECTOR_KEY_L2TPV3, l2tpv3);
2098 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2099 			     FLOW_DISSECTOR_KEY_IPSEC, ipsec);
2100 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2101 			     FLOW_DISSECTOR_KEY_CFM, cfm);
2102 
2103 	skb_flow_dissector_init(dissector, keys, cnt);
2104 }
2105 
2106 static struct fl_flow_mask *fl_create_new_mask(struct cls_fl_head *head,
2107 					       struct fl_flow_mask *mask)
2108 {
2109 	struct fl_flow_mask *newmask;
2110 	int err;
2111 
2112 	newmask = kzalloc(sizeof(*newmask), GFP_KERNEL);
2113 	if (!newmask)
2114 		return ERR_PTR(-ENOMEM);
2115 
2116 	fl_mask_copy(newmask, mask);
2117 
2118 	if ((newmask->key.tp_range.tp_min.dst &&
2119 	     newmask->key.tp_range.tp_max.dst) ||
2120 	    (newmask->key.tp_range.tp_min.src &&
2121 	     newmask->key.tp_range.tp_max.src))
2122 		newmask->flags |= TCA_FLOWER_MASK_FLAGS_RANGE;
2123 
2124 	err = fl_init_mask_hashtable(newmask);
2125 	if (err)
2126 		goto errout_free;
2127 
2128 	fl_init_dissector(&newmask->dissector, &newmask->key);
2129 
2130 	INIT_LIST_HEAD_RCU(&newmask->filters);
2131 
2132 	refcount_set(&newmask->refcnt, 1);
2133 	err = rhashtable_replace_fast(&head->ht, &mask->ht_node,
2134 				      &newmask->ht_node, mask_ht_params);
2135 	if (err)
2136 		goto errout_destroy;
2137 
2138 	spin_lock(&head->masks_lock);
2139 	list_add_tail_rcu(&newmask->list, &head->masks);
2140 	spin_unlock(&head->masks_lock);
2141 
2142 	return newmask;
2143 
2144 errout_destroy:
2145 	rhashtable_destroy(&newmask->ht);
2146 errout_free:
2147 	kfree(newmask);
2148 
2149 	return ERR_PTR(err);
2150 }
2151 
2152 static int fl_check_assign_mask(struct cls_fl_head *head,
2153 				struct cls_fl_filter *fnew,
2154 				struct cls_fl_filter *fold,
2155 				struct fl_flow_mask *mask)
2156 {
2157 	struct fl_flow_mask *newmask;
2158 	int ret = 0;
2159 
2160 	rcu_read_lock();
2161 
2162 	/* Insert mask as temporary node to prevent concurrent creation of mask
2163 	 * with same key. Any concurrent lookups with same key will return
2164 	 * -EAGAIN because mask's refcnt is zero.
2165 	 */
2166 	fnew->mask = rhashtable_lookup_get_insert_fast(&head->ht,
2167 						       &mask->ht_node,
2168 						       mask_ht_params);
2169 	if (!fnew->mask) {
2170 		rcu_read_unlock();
2171 
2172 		if (fold) {
2173 			ret = -EINVAL;
2174 			goto errout_cleanup;
2175 		}
2176 
2177 		newmask = fl_create_new_mask(head, mask);
2178 		if (IS_ERR(newmask)) {
2179 			ret = PTR_ERR(newmask);
2180 			goto errout_cleanup;
2181 		}
2182 
2183 		fnew->mask = newmask;
2184 		return 0;
2185 	} else if (IS_ERR(fnew->mask)) {
2186 		ret = PTR_ERR(fnew->mask);
2187 	} else if (fold && fold->mask != fnew->mask) {
2188 		ret = -EINVAL;
2189 	} else if (!refcount_inc_not_zero(&fnew->mask->refcnt)) {
2190 		/* Mask was deleted concurrently, try again */
2191 		ret = -EAGAIN;
2192 	}
2193 	rcu_read_unlock();
2194 	return ret;
2195 
2196 errout_cleanup:
2197 	rhashtable_remove_fast(&head->ht, &mask->ht_node,
2198 			       mask_ht_params);
2199 	return ret;
2200 }
2201 
2202 static bool fl_needs_tc_skb_ext(const struct fl_flow_key *mask)
2203 {
2204 	return mask->meta.l2_miss;
2205 }
2206 
2207 static int fl_ht_insert_unique(struct cls_fl_filter *fnew,
2208 			       struct cls_fl_filter *fold,
2209 			       bool *in_ht)
2210 {
2211 	struct fl_flow_mask *mask = fnew->mask;
2212 	int err;
2213 
2214 	err = rhashtable_lookup_insert_fast(&mask->ht,
2215 					    &fnew->ht_node,
2216 					    mask->filter_ht_params);
2217 	if (err) {
2218 		*in_ht = false;
2219 		/* It is okay if filter with same key exists when
2220 		 * overwriting.
2221 		 */
2222 		return fold && err == -EEXIST ? 0 : err;
2223 	}
2224 
2225 	*in_ht = true;
2226 	return 0;
2227 }
2228 
2229 static int fl_change(struct net *net, struct sk_buff *in_skb,
2230 		     struct tcf_proto *tp, unsigned long base,
2231 		     u32 handle, struct nlattr **tca,
2232 		     void **arg, u32 flags,
2233 		     struct netlink_ext_ack *extack)
2234 {
2235 	struct cls_fl_head *head = fl_head_dereference(tp);
2236 	bool rtnl_held = !(flags & TCA_ACT_FLAGS_NO_RTNL);
2237 	struct cls_fl_filter *fold = *arg;
2238 	bool bound_to_filter = false;
2239 	struct cls_fl_filter *fnew;
2240 	struct fl_flow_mask *mask;
2241 	struct nlattr **tb;
2242 	bool in_ht;
2243 	int err;
2244 
2245 	if (!tca[TCA_OPTIONS]) {
2246 		err = -EINVAL;
2247 		goto errout_fold;
2248 	}
2249 
2250 	mask = kzalloc(sizeof(struct fl_flow_mask), GFP_KERNEL);
2251 	if (!mask) {
2252 		err = -ENOBUFS;
2253 		goto errout_fold;
2254 	}
2255 
2256 	tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
2257 	if (!tb) {
2258 		err = -ENOBUFS;
2259 		goto errout_mask_alloc;
2260 	}
2261 
2262 	err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
2263 					  tca[TCA_OPTIONS], fl_policy, NULL);
2264 	if (err < 0)
2265 		goto errout_tb;
2266 
2267 	if (fold && handle && fold->handle != handle) {
2268 		err = -EINVAL;
2269 		goto errout_tb;
2270 	}
2271 
2272 	fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
2273 	if (!fnew) {
2274 		err = -ENOBUFS;
2275 		goto errout_tb;
2276 	}
2277 	INIT_LIST_HEAD(&fnew->hw_list);
2278 	refcount_set(&fnew->refcnt, 1);
2279 
2280 	if (tb[TCA_FLOWER_FLAGS]) {
2281 		fnew->flags = nla_get_u32(tb[TCA_FLOWER_FLAGS]);
2282 
2283 		if (!tc_flags_valid(fnew->flags)) {
2284 			kfree(fnew);
2285 			err = -EINVAL;
2286 			goto errout_tb;
2287 		}
2288 	}
2289 
2290 	if (!fold) {
2291 		spin_lock(&tp->lock);
2292 		if (!handle) {
2293 			handle = 1;
2294 			err = idr_alloc_u32(&head->handle_idr, NULL, &handle,
2295 					    INT_MAX, GFP_ATOMIC);
2296 		} else {
2297 			err = idr_alloc_u32(&head->handle_idr, NULL, &handle,
2298 					    handle, GFP_ATOMIC);
2299 
2300 			/* Filter with specified handle was concurrently
2301 			 * inserted after initial check in cls_api. This is not
2302 			 * necessarily an error if NLM_F_EXCL is not set in
2303 			 * message flags. Returning EAGAIN will cause cls_api to
2304 			 * try to update concurrently inserted rule.
2305 			 */
2306 			if (err == -ENOSPC)
2307 				err = -EAGAIN;
2308 		}
2309 		spin_unlock(&tp->lock);
2310 
2311 		if (err) {
2312 			kfree(fnew);
2313 			goto errout_tb;
2314 		}
2315 	}
2316 	fnew->handle = handle;
2317 
2318 	err = tcf_exts_init_ex(&fnew->exts, net, TCA_FLOWER_ACT, 0, tp, handle,
2319 			       !tc_skip_hw(fnew->flags));
2320 	if (err < 0)
2321 		goto errout_idr;
2322 
2323 	err = tcf_exts_validate_ex(net, tp, tb, tca[TCA_RATE],
2324 				   &fnew->exts, flags, fnew->flags,
2325 				   extack);
2326 	if (err < 0)
2327 		goto errout_idr;
2328 
2329 	if (tb[TCA_FLOWER_CLASSID]) {
2330 		fnew->res.classid = nla_get_u32(tb[TCA_FLOWER_CLASSID]);
2331 		if (flags & TCA_ACT_FLAGS_NO_RTNL)
2332 			rtnl_lock();
2333 		tcf_bind_filter(tp, &fnew->res, base);
2334 		if (flags & TCA_ACT_FLAGS_NO_RTNL)
2335 			rtnl_unlock();
2336 		bound_to_filter = true;
2337 	}
2338 
2339 	err = fl_set_key(net, tb, &fnew->key, &mask->key, extack);
2340 	if (err)
2341 		goto unbind_filter;
2342 
2343 	fl_mask_update_range(mask);
2344 	fl_set_masked_key(&fnew->mkey, &fnew->key, mask);
2345 
2346 	if (!fl_mask_fits_tmplt(tp->chain->tmplt_priv, mask)) {
2347 		NL_SET_ERR_MSG_MOD(extack, "Mask does not fit the template");
2348 		err = -EINVAL;
2349 		goto unbind_filter;
2350 	}
2351 
2352 	/* Enable tc skb extension if filter matches on data extracted from
2353 	 * this extension.
2354 	 */
2355 	if (fl_needs_tc_skb_ext(&mask->key)) {
2356 		fnew->needs_tc_skb_ext = 1;
2357 		tc_skb_ext_tc_enable();
2358 	}
2359 
2360 	err = fl_check_assign_mask(head, fnew, fold, mask);
2361 	if (err)
2362 		goto unbind_filter;
2363 
2364 	err = fl_ht_insert_unique(fnew, fold, &in_ht);
2365 	if (err)
2366 		goto errout_mask;
2367 
2368 	if (!tc_skip_hw(fnew->flags)) {
2369 		err = fl_hw_replace_filter(tp, fnew, rtnl_held, extack);
2370 		if (err)
2371 			goto errout_ht;
2372 	}
2373 
2374 	if (!tc_in_hw(fnew->flags))
2375 		fnew->flags |= TCA_CLS_FLAGS_NOT_IN_HW;
2376 
2377 	spin_lock(&tp->lock);
2378 
2379 	/* tp was deleted concurrently. -EAGAIN will cause caller to lookup
2380 	 * proto again or create new one, if necessary.
2381 	 */
2382 	if (tp->deleting) {
2383 		err = -EAGAIN;
2384 		goto errout_hw;
2385 	}
2386 
2387 	if (fold) {
2388 		/* Fold filter was deleted concurrently. Retry lookup. */
2389 		if (fold->deleted) {
2390 			err = -EAGAIN;
2391 			goto errout_hw;
2392 		}
2393 
2394 		fnew->handle = handle;
2395 
2396 		if (!in_ht) {
2397 			struct rhashtable_params params =
2398 				fnew->mask->filter_ht_params;
2399 
2400 			err = rhashtable_insert_fast(&fnew->mask->ht,
2401 						     &fnew->ht_node,
2402 						     params);
2403 			if (err)
2404 				goto errout_hw;
2405 			in_ht = true;
2406 		}
2407 
2408 		refcount_inc(&fnew->refcnt);
2409 		rhashtable_remove_fast(&fold->mask->ht,
2410 				       &fold->ht_node,
2411 				       fold->mask->filter_ht_params);
2412 		idr_replace(&head->handle_idr, fnew, fnew->handle);
2413 		list_replace_rcu(&fold->list, &fnew->list);
2414 		fold->deleted = true;
2415 
2416 		spin_unlock(&tp->lock);
2417 
2418 		fl_mask_put(head, fold->mask);
2419 		if (!tc_skip_hw(fold->flags))
2420 			fl_hw_destroy_filter(tp, fold, rtnl_held, NULL);
2421 		tcf_unbind_filter(tp, &fold->res);
2422 		/* Caller holds reference to fold, so refcnt is always > 0
2423 		 * after this.
2424 		 */
2425 		refcount_dec(&fold->refcnt);
2426 		__fl_put(fold);
2427 	} else {
2428 		idr_replace(&head->handle_idr, fnew, fnew->handle);
2429 
2430 		refcount_inc(&fnew->refcnt);
2431 		list_add_tail_rcu(&fnew->list, &fnew->mask->filters);
2432 		spin_unlock(&tp->lock);
2433 	}
2434 
2435 	*arg = fnew;
2436 
2437 	kfree(tb);
2438 	tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work);
2439 	return 0;
2440 
2441 errout_ht:
2442 	spin_lock(&tp->lock);
2443 errout_hw:
2444 	fnew->deleted = true;
2445 	spin_unlock(&tp->lock);
2446 	if (!tc_skip_hw(fnew->flags))
2447 		fl_hw_destroy_filter(tp, fnew, rtnl_held, NULL);
2448 	if (in_ht)
2449 		rhashtable_remove_fast(&fnew->mask->ht, &fnew->ht_node,
2450 				       fnew->mask->filter_ht_params);
2451 errout_mask:
2452 	fl_mask_put(head, fnew->mask);
2453 
2454 unbind_filter:
2455 	if (bound_to_filter) {
2456 		if (flags & TCA_ACT_FLAGS_NO_RTNL)
2457 			rtnl_lock();
2458 		tcf_unbind_filter(tp, &fnew->res);
2459 		if (flags & TCA_ACT_FLAGS_NO_RTNL)
2460 			rtnl_unlock();
2461 	}
2462 
2463 errout_idr:
2464 	if (!fold) {
2465 		spin_lock(&tp->lock);
2466 		idr_remove(&head->handle_idr, fnew->handle);
2467 		spin_unlock(&tp->lock);
2468 	}
2469 	__fl_put(fnew);
2470 errout_tb:
2471 	kfree(tb);
2472 errout_mask_alloc:
2473 	tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work);
2474 errout_fold:
2475 	if (fold)
2476 		__fl_put(fold);
2477 	return err;
2478 }
2479 
2480 static int fl_delete(struct tcf_proto *tp, void *arg, bool *last,
2481 		     bool rtnl_held, struct netlink_ext_ack *extack)
2482 {
2483 	struct cls_fl_head *head = fl_head_dereference(tp);
2484 	struct cls_fl_filter *f = arg;
2485 	bool last_on_mask;
2486 	int err = 0;
2487 
2488 	err = __fl_delete(tp, f, &last_on_mask, rtnl_held, extack);
2489 	*last = list_empty(&head->masks);
2490 	__fl_put(f);
2491 
2492 	return err;
2493 }
2494 
2495 static void fl_walk(struct tcf_proto *tp, struct tcf_walker *arg,
2496 		    bool rtnl_held)
2497 {
2498 	struct cls_fl_head *head = fl_head_dereference(tp);
2499 	unsigned long id = arg->cookie, tmp;
2500 	struct cls_fl_filter *f;
2501 
2502 	arg->count = arg->skip;
2503 
2504 	rcu_read_lock();
2505 	idr_for_each_entry_continue_ul(&head->handle_idr, f, tmp, id) {
2506 		/* don't return filters that are being deleted */
2507 		if (!f || !refcount_inc_not_zero(&f->refcnt))
2508 			continue;
2509 		rcu_read_unlock();
2510 
2511 		if (arg->fn(tp, f, arg) < 0) {
2512 			__fl_put(f);
2513 			arg->stop = 1;
2514 			rcu_read_lock();
2515 			break;
2516 		}
2517 		__fl_put(f);
2518 		arg->count++;
2519 		rcu_read_lock();
2520 	}
2521 	rcu_read_unlock();
2522 	arg->cookie = id;
2523 }
2524 
2525 static struct cls_fl_filter *
2526 fl_get_next_hw_filter(struct tcf_proto *tp, struct cls_fl_filter *f, bool add)
2527 {
2528 	struct cls_fl_head *head = fl_head_dereference(tp);
2529 
2530 	spin_lock(&tp->lock);
2531 	if (list_empty(&head->hw_filters)) {
2532 		spin_unlock(&tp->lock);
2533 		return NULL;
2534 	}
2535 
2536 	if (!f)
2537 		f = list_entry(&head->hw_filters, struct cls_fl_filter,
2538 			       hw_list);
2539 	list_for_each_entry_continue(f, &head->hw_filters, hw_list) {
2540 		if (!(add && f->deleted) && refcount_inc_not_zero(&f->refcnt)) {
2541 			spin_unlock(&tp->lock);
2542 			return f;
2543 		}
2544 	}
2545 
2546 	spin_unlock(&tp->lock);
2547 	return NULL;
2548 }
2549 
2550 static int fl_reoffload(struct tcf_proto *tp, bool add, flow_setup_cb_t *cb,
2551 			void *cb_priv, struct netlink_ext_ack *extack)
2552 {
2553 	struct tcf_block *block = tp->chain->block;
2554 	struct flow_cls_offload cls_flower = {};
2555 	struct cls_fl_filter *f = NULL;
2556 	int err;
2557 
2558 	/* hw_filters list can only be changed by hw offload functions after
2559 	 * obtaining rtnl lock. Make sure it is not changed while reoffload is
2560 	 * iterating it.
2561 	 */
2562 	ASSERT_RTNL();
2563 
2564 	while ((f = fl_get_next_hw_filter(tp, f, add))) {
2565 		cls_flower.rule =
2566 			flow_rule_alloc(tcf_exts_num_actions(&f->exts));
2567 		if (!cls_flower.rule) {
2568 			__fl_put(f);
2569 			return -ENOMEM;
2570 		}
2571 
2572 		tc_cls_common_offload_init(&cls_flower.common, tp, f->flags,
2573 					   extack);
2574 		cls_flower.command = add ?
2575 			FLOW_CLS_REPLACE : FLOW_CLS_DESTROY;
2576 		cls_flower.cookie = (unsigned long)f;
2577 		cls_flower.rule->match.dissector = &f->mask->dissector;
2578 		cls_flower.rule->match.mask = &f->mask->key;
2579 		cls_flower.rule->match.key = &f->mkey;
2580 
2581 		err = tc_setup_offload_action(&cls_flower.rule->action, &f->exts,
2582 					      cls_flower.common.extack);
2583 		if (err) {
2584 			kfree(cls_flower.rule);
2585 			if (tc_skip_sw(f->flags)) {
2586 				__fl_put(f);
2587 				return err;
2588 			}
2589 			goto next_flow;
2590 		}
2591 
2592 		cls_flower.classid = f->res.classid;
2593 
2594 		err = tc_setup_cb_reoffload(block, tp, add, cb,
2595 					    TC_SETUP_CLSFLOWER, &cls_flower,
2596 					    cb_priv, &f->flags,
2597 					    &f->in_hw_count);
2598 		tc_cleanup_offload_action(&cls_flower.rule->action);
2599 		kfree(cls_flower.rule);
2600 
2601 		if (err) {
2602 			__fl_put(f);
2603 			return err;
2604 		}
2605 next_flow:
2606 		__fl_put(f);
2607 	}
2608 
2609 	return 0;
2610 }
2611 
2612 static void fl_hw_add(struct tcf_proto *tp, void *type_data)
2613 {
2614 	struct flow_cls_offload *cls_flower = type_data;
2615 	struct cls_fl_filter *f =
2616 		(struct cls_fl_filter *) cls_flower->cookie;
2617 	struct cls_fl_head *head = fl_head_dereference(tp);
2618 
2619 	spin_lock(&tp->lock);
2620 	list_add(&f->hw_list, &head->hw_filters);
2621 	spin_unlock(&tp->lock);
2622 }
2623 
2624 static void fl_hw_del(struct tcf_proto *tp, void *type_data)
2625 {
2626 	struct flow_cls_offload *cls_flower = type_data;
2627 	struct cls_fl_filter *f =
2628 		(struct cls_fl_filter *) cls_flower->cookie;
2629 
2630 	spin_lock(&tp->lock);
2631 	if (!list_empty(&f->hw_list))
2632 		list_del_init(&f->hw_list);
2633 	spin_unlock(&tp->lock);
2634 }
2635 
2636 static int fl_hw_create_tmplt(struct tcf_chain *chain,
2637 			      struct fl_flow_tmplt *tmplt)
2638 {
2639 	struct flow_cls_offload cls_flower = {};
2640 	struct tcf_block *block = chain->block;
2641 
2642 	cls_flower.rule = flow_rule_alloc(0);
2643 	if (!cls_flower.rule)
2644 		return -ENOMEM;
2645 
2646 	cls_flower.common.chain_index = chain->index;
2647 	cls_flower.command = FLOW_CLS_TMPLT_CREATE;
2648 	cls_flower.cookie = (unsigned long) tmplt;
2649 	cls_flower.rule->match.dissector = &tmplt->dissector;
2650 	cls_flower.rule->match.mask = &tmplt->mask;
2651 	cls_flower.rule->match.key = &tmplt->dummy_key;
2652 
2653 	/* We don't care if driver (any of them) fails to handle this
2654 	 * call. It serves just as a hint for it.
2655 	 */
2656 	tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, true);
2657 	kfree(cls_flower.rule);
2658 
2659 	return 0;
2660 }
2661 
2662 static void fl_hw_destroy_tmplt(struct tcf_chain *chain,
2663 				struct fl_flow_tmplt *tmplt)
2664 {
2665 	struct flow_cls_offload cls_flower = {};
2666 	struct tcf_block *block = chain->block;
2667 
2668 	cls_flower.common.chain_index = chain->index;
2669 	cls_flower.command = FLOW_CLS_TMPLT_DESTROY;
2670 	cls_flower.cookie = (unsigned long) tmplt;
2671 
2672 	tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, true);
2673 }
2674 
2675 static void *fl_tmplt_create(struct net *net, struct tcf_chain *chain,
2676 			     struct nlattr **tca,
2677 			     struct netlink_ext_ack *extack)
2678 {
2679 	struct fl_flow_tmplt *tmplt;
2680 	struct nlattr **tb;
2681 	int err;
2682 
2683 	if (!tca[TCA_OPTIONS])
2684 		return ERR_PTR(-EINVAL);
2685 
2686 	tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
2687 	if (!tb)
2688 		return ERR_PTR(-ENOBUFS);
2689 	err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
2690 					  tca[TCA_OPTIONS], fl_policy, NULL);
2691 	if (err)
2692 		goto errout_tb;
2693 
2694 	tmplt = kzalloc(sizeof(*tmplt), GFP_KERNEL);
2695 	if (!tmplt) {
2696 		err = -ENOMEM;
2697 		goto errout_tb;
2698 	}
2699 	tmplt->chain = chain;
2700 	err = fl_set_key(net, tb, &tmplt->dummy_key, &tmplt->mask, extack);
2701 	if (err)
2702 		goto errout_tmplt;
2703 
2704 	fl_init_dissector(&tmplt->dissector, &tmplt->mask);
2705 
2706 	err = fl_hw_create_tmplt(chain, tmplt);
2707 	if (err)
2708 		goto errout_tmplt;
2709 
2710 	kfree(tb);
2711 	return tmplt;
2712 
2713 errout_tmplt:
2714 	kfree(tmplt);
2715 errout_tb:
2716 	kfree(tb);
2717 	return ERR_PTR(err);
2718 }
2719 
2720 static void fl_tmplt_destroy(void *tmplt_priv)
2721 {
2722 	struct fl_flow_tmplt *tmplt = tmplt_priv;
2723 
2724 	fl_hw_destroy_tmplt(tmplt->chain, tmplt);
2725 	kfree(tmplt);
2726 }
2727 
2728 static void fl_tmplt_reoffload(struct tcf_chain *chain, bool add,
2729 			       flow_setup_cb_t *cb, void *cb_priv)
2730 {
2731 	struct fl_flow_tmplt *tmplt = chain->tmplt_priv;
2732 	struct flow_cls_offload cls_flower = {};
2733 
2734 	cls_flower.rule = flow_rule_alloc(0);
2735 	if (!cls_flower.rule)
2736 		return;
2737 
2738 	cls_flower.common.chain_index = chain->index;
2739 	cls_flower.command = add ? FLOW_CLS_TMPLT_CREATE :
2740 				   FLOW_CLS_TMPLT_DESTROY;
2741 	cls_flower.cookie = (unsigned long) tmplt;
2742 	cls_flower.rule->match.dissector = &tmplt->dissector;
2743 	cls_flower.rule->match.mask = &tmplt->mask;
2744 	cls_flower.rule->match.key = &tmplt->dummy_key;
2745 
2746 	cb(TC_SETUP_CLSFLOWER, &cls_flower, cb_priv);
2747 	kfree(cls_flower.rule);
2748 }
2749 
2750 static int fl_dump_key_val(struct sk_buff *skb,
2751 			   void *val, int val_type,
2752 			   void *mask, int mask_type, int len)
2753 {
2754 	int err;
2755 
2756 	if (!memchr_inv(mask, 0, len))
2757 		return 0;
2758 	err = nla_put(skb, val_type, len, val);
2759 	if (err)
2760 		return err;
2761 	if (mask_type != TCA_FLOWER_UNSPEC) {
2762 		err = nla_put(skb, mask_type, len, mask);
2763 		if (err)
2764 			return err;
2765 	}
2766 	return 0;
2767 }
2768 
2769 static int fl_dump_key_port_range(struct sk_buff *skb, struct fl_flow_key *key,
2770 				  struct fl_flow_key *mask)
2771 {
2772 	if (fl_dump_key_val(skb, &key->tp_range.tp_min.dst,
2773 			    TCA_FLOWER_KEY_PORT_DST_MIN,
2774 			    &mask->tp_range.tp_min.dst, TCA_FLOWER_UNSPEC,
2775 			    sizeof(key->tp_range.tp_min.dst)) ||
2776 	    fl_dump_key_val(skb, &key->tp_range.tp_max.dst,
2777 			    TCA_FLOWER_KEY_PORT_DST_MAX,
2778 			    &mask->tp_range.tp_max.dst, TCA_FLOWER_UNSPEC,
2779 			    sizeof(key->tp_range.tp_max.dst)) ||
2780 	    fl_dump_key_val(skb, &key->tp_range.tp_min.src,
2781 			    TCA_FLOWER_KEY_PORT_SRC_MIN,
2782 			    &mask->tp_range.tp_min.src, TCA_FLOWER_UNSPEC,
2783 			    sizeof(key->tp_range.tp_min.src)) ||
2784 	    fl_dump_key_val(skb, &key->tp_range.tp_max.src,
2785 			    TCA_FLOWER_KEY_PORT_SRC_MAX,
2786 			    &mask->tp_range.tp_max.src, TCA_FLOWER_UNSPEC,
2787 			    sizeof(key->tp_range.tp_max.src)))
2788 		return -1;
2789 
2790 	return 0;
2791 }
2792 
2793 static int fl_dump_key_mpls_opt_lse(struct sk_buff *skb,
2794 				    struct flow_dissector_key_mpls *mpls_key,
2795 				    struct flow_dissector_key_mpls *mpls_mask,
2796 				    u8 lse_index)
2797 {
2798 	struct flow_dissector_mpls_lse *lse_mask = &mpls_mask->ls[lse_index];
2799 	struct flow_dissector_mpls_lse *lse_key = &mpls_key->ls[lse_index];
2800 	int err;
2801 
2802 	err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH,
2803 			 lse_index + 1);
2804 	if (err)
2805 		return err;
2806 
2807 	if (lse_mask->mpls_ttl) {
2808 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL,
2809 				 lse_key->mpls_ttl);
2810 		if (err)
2811 			return err;
2812 	}
2813 	if (lse_mask->mpls_bos) {
2814 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS,
2815 				 lse_key->mpls_bos);
2816 		if (err)
2817 			return err;
2818 	}
2819 	if (lse_mask->mpls_tc) {
2820 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_TC,
2821 				 lse_key->mpls_tc);
2822 		if (err)
2823 			return err;
2824 	}
2825 	if (lse_mask->mpls_label) {
2826 		err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL,
2827 				  lse_key->mpls_label);
2828 		if (err)
2829 			return err;
2830 	}
2831 
2832 	return 0;
2833 }
2834 
2835 static int fl_dump_key_mpls_opts(struct sk_buff *skb,
2836 				 struct flow_dissector_key_mpls *mpls_key,
2837 				 struct flow_dissector_key_mpls *mpls_mask)
2838 {
2839 	struct nlattr *opts;
2840 	struct nlattr *lse;
2841 	u8 lse_index;
2842 	int err;
2843 
2844 	opts = nla_nest_start(skb, TCA_FLOWER_KEY_MPLS_OPTS);
2845 	if (!opts)
2846 		return -EMSGSIZE;
2847 
2848 	for (lse_index = 0; lse_index < FLOW_DIS_MPLS_MAX; lse_index++) {
2849 		if (!(mpls_mask->used_lses & 1 << lse_index))
2850 			continue;
2851 
2852 		lse = nla_nest_start(skb, TCA_FLOWER_KEY_MPLS_OPTS_LSE);
2853 		if (!lse) {
2854 			err = -EMSGSIZE;
2855 			goto err_opts;
2856 		}
2857 
2858 		err = fl_dump_key_mpls_opt_lse(skb, mpls_key, mpls_mask,
2859 					       lse_index);
2860 		if (err)
2861 			goto err_opts_lse;
2862 		nla_nest_end(skb, lse);
2863 	}
2864 	nla_nest_end(skb, opts);
2865 
2866 	return 0;
2867 
2868 err_opts_lse:
2869 	nla_nest_cancel(skb, lse);
2870 err_opts:
2871 	nla_nest_cancel(skb, opts);
2872 
2873 	return err;
2874 }
2875 
2876 static int fl_dump_key_mpls(struct sk_buff *skb,
2877 			    struct flow_dissector_key_mpls *mpls_key,
2878 			    struct flow_dissector_key_mpls *mpls_mask)
2879 {
2880 	struct flow_dissector_mpls_lse *lse_mask;
2881 	struct flow_dissector_mpls_lse *lse_key;
2882 	int err;
2883 
2884 	if (!mpls_mask->used_lses)
2885 		return 0;
2886 
2887 	lse_mask = &mpls_mask->ls[0];
2888 	lse_key = &mpls_key->ls[0];
2889 
2890 	/* For backward compatibility, don't use the MPLS nested attributes if
2891 	 * the rule can be expressed using the old attributes.
2892 	 */
2893 	if (mpls_mask->used_lses & ~1 ||
2894 	    (!lse_mask->mpls_ttl && !lse_mask->mpls_bos &&
2895 	     !lse_mask->mpls_tc && !lse_mask->mpls_label))
2896 		return fl_dump_key_mpls_opts(skb, mpls_key, mpls_mask);
2897 
2898 	if (lse_mask->mpls_ttl) {
2899 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TTL,
2900 				 lse_key->mpls_ttl);
2901 		if (err)
2902 			return err;
2903 	}
2904 	if (lse_mask->mpls_tc) {
2905 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TC,
2906 				 lse_key->mpls_tc);
2907 		if (err)
2908 			return err;
2909 	}
2910 	if (lse_mask->mpls_label) {
2911 		err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_LABEL,
2912 				  lse_key->mpls_label);
2913 		if (err)
2914 			return err;
2915 	}
2916 	if (lse_mask->mpls_bos) {
2917 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_BOS,
2918 				 lse_key->mpls_bos);
2919 		if (err)
2920 			return err;
2921 	}
2922 	return 0;
2923 }
2924 
2925 static int fl_dump_key_ip(struct sk_buff *skb, bool encap,
2926 			  struct flow_dissector_key_ip *key,
2927 			  struct flow_dissector_key_ip *mask)
2928 {
2929 	int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
2930 	int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
2931 	int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
2932 	int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
2933 
2934 	if (fl_dump_key_val(skb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos)) ||
2935 	    fl_dump_key_val(skb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl)))
2936 		return -1;
2937 
2938 	return 0;
2939 }
2940 
2941 static int fl_dump_key_vlan(struct sk_buff *skb,
2942 			    int vlan_id_key, int vlan_prio_key,
2943 			    struct flow_dissector_key_vlan *vlan_key,
2944 			    struct flow_dissector_key_vlan *vlan_mask)
2945 {
2946 	int err;
2947 
2948 	if (!memchr_inv(vlan_mask, 0, sizeof(*vlan_mask)))
2949 		return 0;
2950 	if (vlan_mask->vlan_id) {
2951 		err = nla_put_u16(skb, vlan_id_key,
2952 				  vlan_key->vlan_id);
2953 		if (err)
2954 			return err;
2955 	}
2956 	if (vlan_mask->vlan_priority) {
2957 		err = nla_put_u8(skb, vlan_prio_key,
2958 				 vlan_key->vlan_priority);
2959 		if (err)
2960 			return err;
2961 	}
2962 	return 0;
2963 }
2964 
2965 static void fl_get_key_flag(u32 dissector_key, u32 dissector_mask,
2966 			    u32 *flower_key, u32 *flower_mask,
2967 			    u32 flower_flag_bit, u32 dissector_flag_bit)
2968 {
2969 	if (dissector_mask & dissector_flag_bit) {
2970 		*flower_mask |= flower_flag_bit;
2971 		if (dissector_key & dissector_flag_bit)
2972 			*flower_key |= flower_flag_bit;
2973 	}
2974 }
2975 
2976 static int fl_dump_key_flags(struct sk_buff *skb, u32 flags_key, u32 flags_mask)
2977 {
2978 	u32 key, mask;
2979 	__be32 _key, _mask;
2980 	int err;
2981 
2982 	if (!memchr_inv(&flags_mask, 0, sizeof(flags_mask)))
2983 		return 0;
2984 
2985 	key = 0;
2986 	mask = 0;
2987 
2988 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
2989 			TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
2990 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
2991 			TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
2992 			FLOW_DIS_FIRST_FRAG);
2993 
2994 	_key = cpu_to_be32(key);
2995 	_mask = cpu_to_be32(mask);
2996 
2997 	err = nla_put(skb, TCA_FLOWER_KEY_FLAGS, 4, &_key);
2998 	if (err)
2999 		return err;
3000 
3001 	return nla_put(skb, TCA_FLOWER_KEY_FLAGS_MASK, 4, &_mask);
3002 }
3003 
3004 static int fl_dump_key_geneve_opt(struct sk_buff *skb,
3005 				  struct flow_dissector_key_enc_opts *enc_opts)
3006 {
3007 	struct geneve_opt *opt;
3008 	struct nlattr *nest;
3009 	int opt_off = 0;
3010 
3011 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_GENEVE);
3012 	if (!nest)
3013 		goto nla_put_failure;
3014 
3015 	while (enc_opts->len > opt_off) {
3016 		opt = (struct geneve_opt *)&enc_opts->data[opt_off];
3017 
3018 		if (nla_put_be16(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS,
3019 				 opt->opt_class))
3020 			goto nla_put_failure;
3021 		if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE,
3022 			       opt->type))
3023 			goto nla_put_failure;
3024 		if (nla_put(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA,
3025 			    opt->length * 4, opt->opt_data))
3026 			goto nla_put_failure;
3027 
3028 		opt_off += sizeof(struct geneve_opt) + opt->length * 4;
3029 	}
3030 	nla_nest_end(skb, nest);
3031 	return 0;
3032 
3033 nla_put_failure:
3034 	nla_nest_cancel(skb, nest);
3035 	return -EMSGSIZE;
3036 }
3037 
3038 static int fl_dump_key_vxlan_opt(struct sk_buff *skb,
3039 				 struct flow_dissector_key_enc_opts *enc_opts)
3040 {
3041 	struct vxlan_metadata *md;
3042 	struct nlattr *nest;
3043 
3044 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_VXLAN);
3045 	if (!nest)
3046 		goto nla_put_failure;
3047 
3048 	md = (struct vxlan_metadata *)&enc_opts->data[0];
3049 	if (nla_put_u32(skb, TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP, md->gbp))
3050 		goto nla_put_failure;
3051 
3052 	nla_nest_end(skb, nest);
3053 	return 0;
3054 
3055 nla_put_failure:
3056 	nla_nest_cancel(skb, nest);
3057 	return -EMSGSIZE;
3058 }
3059 
3060 static int fl_dump_key_erspan_opt(struct sk_buff *skb,
3061 				  struct flow_dissector_key_enc_opts *enc_opts)
3062 {
3063 	struct erspan_metadata *md;
3064 	struct nlattr *nest;
3065 
3066 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_ERSPAN);
3067 	if (!nest)
3068 		goto nla_put_failure;
3069 
3070 	md = (struct erspan_metadata *)&enc_opts->data[0];
3071 	if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER, md->version))
3072 		goto nla_put_failure;
3073 
3074 	if (md->version == 1 &&
3075 	    nla_put_be32(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX, md->u.index))
3076 		goto nla_put_failure;
3077 
3078 	if (md->version == 2 &&
3079 	    (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR,
3080 			md->u.md2.dir) ||
3081 	     nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID,
3082 			get_hwid(&md->u.md2))))
3083 		goto nla_put_failure;
3084 
3085 	nla_nest_end(skb, nest);
3086 	return 0;
3087 
3088 nla_put_failure:
3089 	nla_nest_cancel(skb, nest);
3090 	return -EMSGSIZE;
3091 }
3092 
3093 static int fl_dump_key_gtp_opt(struct sk_buff *skb,
3094 			       struct flow_dissector_key_enc_opts *enc_opts)
3095 
3096 {
3097 	struct gtp_pdu_session_info *session_info;
3098 	struct nlattr *nest;
3099 
3100 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_GTP);
3101 	if (!nest)
3102 		goto nla_put_failure;
3103 
3104 	session_info = (struct gtp_pdu_session_info *)&enc_opts->data[0];
3105 
3106 	if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE,
3107 		       session_info->pdu_type))
3108 		goto nla_put_failure;
3109 
3110 	if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GTP_QFI, session_info->qfi))
3111 		goto nla_put_failure;
3112 
3113 	nla_nest_end(skb, nest);
3114 	return 0;
3115 
3116 nla_put_failure:
3117 	nla_nest_cancel(skb, nest);
3118 	return -EMSGSIZE;
3119 }
3120 
3121 static int fl_dump_key_ct(struct sk_buff *skb,
3122 			  struct flow_dissector_key_ct *key,
3123 			  struct flow_dissector_key_ct *mask)
3124 {
3125 	if (IS_ENABLED(CONFIG_NF_CONNTRACK) &&
3126 	    fl_dump_key_val(skb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE,
3127 			    &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK,
3128 			    sizeof(key->ct_state)))
3129 		goto nla_put_failure;
3130 
3131 	if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
3132 	    fl_dump_key_val(skb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE,
3133 			    &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK,
3134 			    sizeof(key->ct_zone)))
3135 		goto nla_put_failure;
3136 
3137 	if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
3138 	    fl_dump_key_val(skb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK,
3139 			    &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK,
3140 			    sizeof(key->ct_mark)))
3141 		goto nla_put_failure;
3142 
3143 	if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
3144 	    fl_dump_key_val(skb, &key->ct_labels, TCA_FLOWER_KEY_CT_LABELS,
3145 			    &mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK,
3146 			    sizeof(key->ct_labels)))
3147 		goto nla_put_failure;
3148 
3149 	return 0;
3150 
3151 nla_put_failure:
3152 	return -EMSGSIZE;
3153 }
3154 
3155 static int fl_dump_key_cfm(struct sk_buff *skb,
3156 			   struct flow_dissector_key_cfm *key,
3157 			   struct flow_dissector_key_cfm *mask)
3158 {
3159 	struct nlattr *opts;
3160 	int err;
3161 	u8 mdl;
3162 
3163 	if (!memchr_inv(mask, 0, sizeof(*mask)))
3164 		return 0;
3165 
3166 	opts = nla_nest_start(skb, TCA_FLOWER_KEY_CFM);
3167 	if (!opts)
3168 		return -EMSGSIZE;
3169 
3170 	if (FIELD_GET(FLOW_DIS_CFM_MDL_MASK, mask->mdl_ver)) {
3171 		mdl = FIELD_GET(FLOW_DIS_CFM_MDL_MASK, key->mdl_ver);
3172 		err = nla_put_u8(skb, TCA_FLOWER_KEY_CFM_MD_LEVEL, mdl);
3173 		if (err)
3174 			goto err_cfm_opts;
3175 	}
3176 
3177 	if (mask->opcode) {
3178 		err = nla_put_u8(skb, TCA_FLOWER_KEY_CFM_OPCODE, key->opcode);
3179 		if (err)
3180 			goto err_cfm_opts;
3181 	}
3182 
3183 	nla_nest_end(skb, opts);
3184 
3185 	return 0;
3186 
3187 err_cfm_opts:
3188 	nla_nest_cancel(skb, opts);
3189 	return err;
3190 }
3191 
3192 static int fl_dump_key_options(struct sk_buff *skb, int enc_opt_type,
3193 			       struct flow_dissector_key_enc_opts *enc_opts)
3194 {
3195 	struct nlattr *nest;
3196 	int err;
3197 
3198 	if (!enc_opts->len)
3199 		return 0;
3200 
3201 	nest = nla_nest_start_noflag(skb, enc_opt_type);
3202 	if (!nest)
3203 		goto nla_put_failure;
3204 
3205 	switch (enc_opts->dst_opt_type) {
3206 	case TUNNEL_GENEVE_OPT:
3207 		err = fl_dump_key_geneve_opt(skb, enc_opts);
3208 		if (err)
3209 			goto nla_put_failure;
3210 		break;
3211 	case TUNNEL_VXLAN_OPT:
3212 		err = fl_dump_key_vxlan_opt(skb, enc_opts);
3213 		if (err)
3214 			goto nla_put_failure;
3215 		break;
3216 	case TUNNEL_ERSPAN_OPT:
3217 		err = fl_dump_key_erspan_opt(skb, enc_opts);
3218 		if (err)
3219 			goto nla_put_failure;
3220 		break;
3221 	case TUNNEL_GTP_OPT:
3222 		err = fl_dump_key_gtp_opt(skb, enc_opts);
3223 		if (err)
3224 			goto nla_put_failure;
3225 		break;
3226 	default:
3227 		goto nla_put_failure;
3228 	}
3229 	nla_nest_end(skb, nest);
3230 	return 0;
3231 
3232 nla_put_failure:
3233 	nla_nest_cancel(skb, nest);
3234 	return -EMSGSIZE;
3235 }
3236 
3237 static int fl_dump_key_enc_opt(struct sk_buff *skb,
3238 			       struct flow_dissector_key_enc_opts *key_opts,
3239 			       struct flow_dissector_key_enc_opts *msk_opts)
3240 {
3241 	int err;
3242 
3243 	err = fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS, key_opts);
3244 	if (err)
3245 		return err;
3246 
3247 	return fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS_MASK, msk_opts);
3248 }
3249 
3250 static int fl_dump_key(struct sk_buff *skb, struct net *net,
3251 		       struct fl_flow_key *key, struct fl_flow_key *mask)
3252 {
3253 	if (mask->meta.ingress_ifindex) {
3254 		struct net_device *dev;
3255 
3256 		dev = __dev_get_by_index(net, key->meta.ingress_ifindex);
3257 		if (dev && nla_put_string(skb, TCA_FLOWER_INDEV, dev->name))
3258 			goto nla_put_failure;
3259 	}
3260 
3261 	if (fl_dump_key_val(skb, &key->meta.l2_miss,
3262 			    TCA_FLOWER_L2_MISS, &mask->meta.l2_miss,
3263 			    TCA_FLOWER_UNSPEC, sizeof(key->meta.l2_miss)))
3264 		goto nla_put_failure;
3265 
3266 	if (fl_dump_key_val(skb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
3267 			    mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
3268 			    sizeof(key->eth.dst)) ||
3269 	    fl_dump_key_val(skb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
3270 			    mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
3271 			    sizeof(key->eth.src)) ||
3272 	    fl_dump_key_val(skb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE,
3273 			    &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
3274 			    sizeof(key->basic.n_proto)))
3275 		goto nla_put_failure;
3276 
3277 	if (mask->num_of_vlans.num_of_vlans) {
3278 		if (nla_put_u8(skb, TCA_FLOWER_KEY_NUM_OF_VLANS, key->num_of_vlans.num_of_vlans))
3279 			goto nla_put_failure;
3280 	}
3281 
3282 	if (fl_dump_key_mpls(skb, &key->mpls, &mask->mpls))
3283 		goto nla_put_failure;
3284 
3285 	if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_VLAN_ID,
3286 			     TCA_FLOWER_KEY_VLAN_PRIO, &key->vlan, &mask->vlan))
3287 		goto nla_put_failure;
3288 
3289 	if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_CVLAN_ID,
3290 			     TCA_FLOWER_KEY_CVLAN_PRIO,
3291 			     &key->cvlan, &mask->cvlan) ||
3292 	    (mask->cvlan.vlan_tpid &&
3293 	     nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
3294 			  key->cvlan.vlan_tpid)))
3295 		goto nla_put_failure;
3296 
3297 	if (mask->basic.n_proto) {
3298 		if (mask->cvlan.vlan_eth_type) {
3299 			if (nla_put_be16(skb, TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
3300 					 key->basic.n_proto))
3301 				goto nla_put_failure;
3302 		} else if (mask->vlan.vlan_eth_type) {
3303 			if (nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
3304 					 key->vlan.vlan_eth_type))
3305 				goto nla_put_failure;
3306 		}
3307 	}
3308 
3309 	if ((key->basic.n_proto == htons(ETH_P_IP) ||
3310 	     key->basic.n_proto == htons(ETH_P_IPV6)) &&
3311 	    (fl_dump_key_val(skb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
3312 			    &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
3313 			    sizeof(key->basic.ip_proto)) ||
3314 	    fl_dump_key_ip(skb, false, &key->ip, &mask->ip)))
3315 		goto nla_put_failure;
3316 
3317 	if (mask->pppoe.session_id) {
3318 		if (nla_put_be16(skb, TCA_FLOWER_KEY_PPPOE_SID,
3319 				 key->pppoe.session_id))
3320 			goto nla_put_failure;
3321 	}
3322 	if (mask->basic.n_proto && mask->pppoe.ppp_proto) {
3323 		if (nla_put_be16(skb, TCA_FLOWER_KEY_PPP_PROTO,
3324 				 key->pppoe.ppp_proto))
3325 			goto nla_put_failure;
3326 	}
3327 
3328 	if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
3329 	    (fl_dump_key_val(skb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
3330 			     &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
3331 			     sizeof(key->ipv4.src)) ||
3332 	     fl_dump_key_val(skb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
3333 			     &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
3334 			     sizeof(key->ipv4.dst))))
3335 		goto nla_put_failure;
3336 	else if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
3337 		 (fl_dump_key_val(skb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
3338 				  &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
3339 				  sizeof(key->ipv6.src)) ||
3340 		  fl_dump_key_val(skb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
3341 				  &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
3342 				  sizeof(key->ipv6.dst))))
3343 		goto nla_put_failure;
3344 
3345 	if (key->basic.ip_proto == IPPROTO_TCP &&
3346 	    (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
3347 			     &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
3348 			     sizeof(key->tp.src)) ||
3349 	     fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
3350 			     &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
3351 			     sizeof(key->tp.dst)) ||
3352 	     fl_dump_key_val(skb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
3353 			     &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
3354 			     sizeof(key->tcp.flags))))
3355 		goto nla_put_failure;
3356 	else if (key->basic.ip_proto == IPPROTO_UDP &&
3357 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
3358 				  &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
3359 				  sizeof(key->tp.src)) ||
3360 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
3361 				  &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
3362 				  sizeof(key->tp.dst))))
3363 		goto nla_put_failure;
3364 	else if (key->basic.ip_proto == IPPROTO_SCTP &&
3365 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
3366 				  &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
3367 				  sizeof(key->tp.src)) ||
3368 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
3369 				  &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
3370 				  sizeof(key->tp.dst))))
3371 		goto nla_put_failure;
3372 	else if (key->basic.n_proto == htons(ETH_P_IP) &&
3373 		 key->basic.ip_proto == IPPROTO_ICMP &&
3374 		 (fl_dump_key_val(skb, &key->icmp.type,
3375 				  TCA_FLOWER_KEY_ICMPV4_TYPE, &mask->icmp.type,
3376 				  TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
3377 				  sizeof(key->icmp.type)) ||
3378 		  fl_dump_key_val(skb, &key->icmp.code,
3379 				  TCA_FLOWER_KEY_ICMPV4_CODE, &mask->icmp.code,
3380 				  TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
3381 				  sizeof(key->icmp.code))))
3382 		goto nla_put_failure;
3383 	else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
3384 		 key->basic.ip_proto == IPPROTO_ICMPV6 &&
3385 		 (fl_dump_key_val(skb, &key->icmp.type,
3386 				  TCA_FLOWER_KEY_ICMPV6_TYPE, &mask->icmp.type,
3387 				  TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
3388 				  sizeof(key->icmp.type)) ||
3389 		  fl_dump_key_val(skb, &key->icmp.code,
3390 				  TCA_FLOWER_KEY_ICMPV6_CODE, &mask->icmp.code,
3391 				  TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
3392 				  sizeof(key->icmp.code))))
3393 		goto nla_put_failure;
3394 	else if ((key->basic.n_proto == htons(ETH_P_ARP) ||
3395 		  key->basic.n_proto == htons(ETH_P_RARP)) &&
3396 		 (fl_dump_key_val(skb, &key->arp.sip,
3397 				  TCA_FLOWER_KEY_ARP_SIP, &mask->arp.sip,
3398 				  TCA_FLOWER_KEY_ARP_SIP_MASK,
3399 				  sizeof(key->arp.sip)) ||
3400 		  fl_dump_key_val(skb, &key->arp.tip,
3401 				  TCA_FLOWER_KEY_ARP_TIP, &mask->arp.tip,
3402 				  TCA_FLOWER_KEY_ARP_TIP_MASK,
3403 				  sizeof(key->arp.tip)) ||
3404 		  fl_dump_key_val(skb, &key->arp.op,
3405 				  TCA_FLOWER_KEY_ARP_OP, &mask->arp.op,
3406 				  TCA_FLOWER_KEY_ARP_OP_MASK,
3407 				  sizeof(key->arp.op)) ||
3408 		  fl_dump_key_val(skb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
3409 				  mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
3410 				  sizeof(key->arp.sha)) ||
3411 		  fl_dump_key_val(skb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
3412 				  mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
3413 				  sizeof(key->arp.tha))))
3414 		goto nla_put_failure;
3415 	else if (key->basic.ip_proto == IPPROTO_L2TP &&
3416 		 fl_dump_key_val(skb, &key->l2tpv3.session_id,
3417 				 TCA_FLOWER_KEY_L2TPV3_SID,
3418 				 &mask->l2tpv3.session_id,
3419 				 TCA_FLOWER_UNSPEC,
3420 				 sizeof(key->l2tpv3.session_id)))
3421 		goto nla_put_failure;
3422 
3423 	if (key->ipsec.spi &&
3424 	    fl_dump_key_val(skb, &key->ipsec.spi, TCA_FLOWER_KEY_SPI,
3425 			    &mask->ipsec.spi, TCA_FLOWER_KEY_SPI_MASK,
3426 			    sizeof(key->ipsec.spi)))
3427 		goto nla_put_failure;
3428 
3429 	if ((key->basic.ip_proto == IPPROTO_TCP ||
3430 	     key->basic.ip_proto == IPPROTO_UDP ||
3431 	     key->basic.ip_proto == IPPROTO_SCTP) &&
3432 	     fl_dump_key_port_range(skb, key, mask))
3433 		goto nla_put_failure;
3434 
3435 	if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
3436 	    (fl_dump_key_val(skb, &key->enc_ipv4.src,
3437 			    TCA_FLOWER_KEY_ENC_IPV4_SRC, &mask->enc_ipv4.src,
3438 			    TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
3439 			    sizeof(key->enc_ipv4.src)) ||
3440 	     fl_dump_key_val(skb, &key->enc_ipv4.dst,
3441 			     TCA_FLOWER_KEY_ENC_IPV4_DST, &mask->enc_ipv4.dst,
3442 			     TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
3443 			     sizeof(key->enc_ipv4.dst))))
3444 		goto nla_put_failure;
3445 	else if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
3446 		 (fl_dump_key_val(skb, &key->enc_ipv6.src,
3447 			    TCA_FLOWER_KEY_ENC_IPV6_SRC, &mask->enc_ipv6.src,
3448 			    TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
3449 			    sizeof(key->enc_ipv6.src)) ||
3450 		 fl_dump_key_val(skb, &key->enc_ipv6.dst,
3451 				 TCA_FLOWER_KEY_ENC_IPV6_DST,
3452 				 &mask->enc_ipv6.dst,
3453 				 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
3454 			    sizeof(key->enc_ipv6.dst))))
3455 		goto nla_put_failure;
3456 
3457 	if (fl_dump_key_val(skb, &key->enc_key_id, TCA_FLOWER_KEY_ENC_KEY_ID,
3458 			    &mask->enc_key_id, TCA_FLOWER_UNSPEC,
3459 			    sizeof(key->enc_key_id)) ||
3460 	    fl_dump_key_val(skb, &key->enc_tp.src,
3461 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
3462 			    &mask->enc_tp.src,
3463 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
3464 			    sizeof(key->enc_tp.src)) ||
3465 	    fl_dump_key_val(skb, &key->enc_tp.dst,
3466 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
3467 			    &mask->enc_tp.dst,
3468 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
3469 			    sizeof(key->enc_tp.dst)) ||
3470 	    fl_dump_key_ip(skb, true, &key->enc_ip, &mask->enc_ip) ||
3471 	    fl_dump_key_enc_opt(skb, &key->enc_opts, &mask->enc_opts))
3472 		goto nla_put_failure;
3473 
3474 	if (fl_dump_key_ct(skb, &key->ct, &mask->ct))
3475 		goto nla_put_failure;
3476 
3477 	if (fl_dump_key_flags(skb, key->control.flags, mask->control.flags))
3478 		goto nla_put_failure;
3479 
3480 	if (fl_dump_key_val(skb, &key->hash.hash, TCA_FLOWER_KEY_HASH,
3481 			     &mask->hash.hash, TCA_FLOWER_KEY_HASH_MASK,
3482 			     sizeof(key->hash.hash)))
3483 		goto nla_put_failure;
3484 
3485 	if (fl_dump_key_cfm(skb, &key->cfm, &mask->cfm))
3486 		goto nla_put_failure;
3487 
3488 	return 0;
3489 
3490 nla_put_failure:
3491 	return -EMSGSIZE;
3492 }
3493 
3494 static int fl_dump(struct net *net, struct tcf_proto *tp, void *fh,
3495 		   struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
3496 {
3497 	struct cls_fl_filter *f = fh;
3498 	struct nlattr *nest;
3499 	struct fl_flow_key *key, *mask;
3500 	bool skip_hw;
3501 
3502 	if (!f)
3503 		return skb->len;
3504 
3505 	t->tcm_handle = f->handle;
3506 
3507 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
3508 	if (!nest)
3509 		goto nla_put_failure;
3510 
3511 	spin_lock(&tp->lock);
3512 
3513 	if (f->res.classid &&
3514 	    nla_put_u32(skb, TCA_FLOWER_CLASSID, f->res.classid))
3515 		goto nla_put_failure_locked;
3516 
3517 	key = &f->key;
3518 	mask = &f->mask->key;
3519 	skip_hw = tc_skip_hw(f->flags);
3520 
3521 	if (fl_dump_key(skb, net, key, mask))
3522 		goto nla_put_failure_locked;
3523 
3524 	if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
3525 		goto nla_put_failure_locked;
3526 
3527 	spin_unlock(&tp->lock);
3528 
3529 	if (!skip_hw)
3530 		fl_hw_update_stats(tp, f, rtnl_held);
3531 
3532 	if (nla_put_u32(skb, TCA_FLOWER_IN_HW_COUNT, f->in_hw_count))
3533 		goto nla_put_failure;
3534 
3535 	if (tcf_exts_dump(skb, &f->exts))
3536 		goto nla_put_failure;
3537 
3538 	nla_nest_end(skb, nest);
3539 
3540 	if (tcf_exts_dump_stats(skb, &f->exts) < 0)
3541 		goto nla_put_failure;
3542 
3543 	return skb->len;
3544 
3545 nla_put_failure_locked:
3546 	spin_unlock(&tp->lock);
3547 nla_put_failure:
3548 	nla_nest_cancel(skb, nest);
3549 	return -1;
3550 }
3551 
3552 static int fl_terse_dump(struct net *net, struct tcf_proto *tp, void *fh,
3553 			 struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
3554 {
3555 	struct cls_fl_filter *f = fh;
3556 	struct nlattr *nest;
3557 	bool skip_hw;
3558 
3559 	if (!f)
3560 		return skb->len;
3561 
3562 	t->tcm_handle = f->handle;
3563 
3564 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
3565 	if (!nest)
3566 		goto nla_put_failure;
3567 
3568 	spin_lock(&tp->lock);
3569 
3570 	skip_hw = tc_skip_hw(f->flags);
3571 
3572 	if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
3573 		goto nla_put_failure_locked;
3574 
3575 	spin_unlock(&tp->lock);
3576 
3577 	if (!skip_hw)
3578 		fl_hw_update_stats(tp, f, rtnl_held);
3579 
3580 	if (tcf_exts_terse_dump(skb, &f->exts))
3581 		goto nla_put_failure;
3582 
3583 	nla_nest_end(skb, nest);
3584 
3585 	return skb->len;
3586 
3587 nla_put_failure_locked:
3588 	spin_unlock(&tp->lock);
3589 nla_put_failure:
3590 	nla_nest_cancel(skb, nest);
3591 	return -1;
3592 }
3593 
3594 static int fl_tmplt_dump(struct sk_buff *skb, struct net *net, void *tmplt_priv)
3595 {
3596 	struct fl_flow_tmplt *tmplt = tmplt_priv;
3597 	struct fl_flow_key *key, *mask;
3598 	struct nlattr *nest;
3599 
3600 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
3601 	if (!nest)
3602 		goto nla_put_failure;
3603 
3604 	key = &tmplt->dummy_key;
3605 	mask = &tmplt->mask;
3606 
3607 	if (fl_dump_key(skb, net, key, mask))
3608 		goto nla_put_failure;
3609 
3610 	nla_nest_end(skb, nest);
3611 
3612 	return skb->len;
3613 
3614 nla_put_failure:
3615 	nla_nest_cancel(skb, nest);
3616 	return -EMSGSIZE;
3617 }
3618 
3619 static void fl_bind_class(void *fh, u32 classid, unsigned long cl, void *q,
3620 			  unsigned long base)
3621 {
3622 	struct cls_fl_filter *f = fh;
3623 
3624 	tc_cls_bind_class(classid, cl, q, &f->res, base);
3625 }
3626 
3627 static bool fl_delete_empty(struct tcf_proto *tp)
3628 {
3629 	struct cls_fl_head *head = fl_head_dereference(tp);
3630 
3631 	spin_lock(&tp->lock);
3632 	tp->deleting = idr_is_empty(&head->handle_idr);
3633 	spin_unlock(&tp->lock);
3634 
3635 	return tp->deleting;
3636 }
3637 
3638 static struct tcf_proto_ops cls_fl_ops __read_mostly = {
3639 	.kind		= "flower",
3640 	.classify	= fl_classify,
3641 	.init		= fl_init,
3642 	.destroy	= fl_destroy,
3643 	.get		= fl_get,
3644 	.put		= fl_put,
3645 	.change		= fl_change,
3646 	.delete		= fl_delete,
3647 	.delete_empty	= fl_delete_empty,
3648 	.walk		= fl_walk,
3649 	.reoffload	= fl_reoffload,
3650 	.hw_add		= fl_hw_add,
3651 	.hw_del		= fl_hw_del,
3652 	.dump		= fl_dump,
3653 	.terse_dump	= fl_terse_dump,
3654 	.bind_class	= fl_bind_class,
3655 	.tmplt_create	= fl_tmplt_create,
3656 	.tmplt_destroy	= fl_tmplt_destroy,
3657 	.tmplt_reoffload = fl_tmplt_reoffload,
3658 	.tmplt_dump	= fl_tmplt_dump,
3659 	.get_exts	= fl_get_exts,
3660 	.owner		= THIS_MODULE,
3661 	.flags		= TCF_PROTO_OPS_DOIT_UNLOCKED,
3662 };
3663 
3664 static int __init cls_fl_init(void)
3665 {
3666 	return register_tcf_proto_ops(&cls_fl_ops);
3667 }
3668 
3669 static void __exit cls_fl_exit(void)
3670 {
3671 	unregister_tcf_proto_ops(&cls_fl_ops);
3672 }
3673 
3674 module_init(cls_fl_init);
3675 module_exit(cls_fl_exit);
3676 
3677 MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
3678 MODULE_DESCRIPTION("Flower classifier");
3679 MODULE_LICENSE("GPL v2");
3680