xref: /linux/net/sched/cls_flower.c (revision ae22a94997b8a03dcb3c922857c203246711f9d4)
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 	memset(md, 0xff, sizeof(*md));
1324 	md->version = 1;
1325 
1326 	if (!depth)
1327 		return sizeof(*md);
1328 
1329 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_ERSPAN) {
1330 		NL_SET_ERR_MSG(extack, "Non-erspan option type for mask");
1331 		return -EINVAL;
1332 	}
1333 
1334 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX, nla,
1335 			       erspan_opt_policy, extack);
1336 	if (err < 0)
1337 		return err;
1338 
1339 	if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]) {
1340 		NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option ver");
1341 		return -EINVAL;
1342 	}
1343 
1344 	if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER])
1345 		md->version = nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]);
1346 
1347 	if (md->version == 1) {
1348 		if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]) {
1349 			NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option index");
1350 			return -EINVAL;
1351 		}
1352 		if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]) {
1353 			nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX];
1354 			memset(&md->u, 0x00, sizeof(md->u));
1355 			md->u.index = nla_get_be32(nla);
1356 		}
1357 	} else if (md->version == 2) {
1358 		if (!option_len && (!tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR] ||
1359 				    !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID])) {
1360 			NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option dir or hwid");
1361 			return -EINVAL;
1362 		}
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 		if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID]) {
1368 			nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID];
1369 			set_hwid(&md->u.md2, nla_get_u8(nla));
1370 		}
1371 	} else {
1372 		NL_SET_ERR_MSG(extack, "Tunnel key erspan option ver is incorrect");
1373 		return -EINVAL;
1374 	}
1375 
1376 	return sizeof(*md);
1377 }
1378 
1379 static int fl_set_gtp_opt(const struct nlattr *nla, struct fl_flow_key *key,
1380 			  int depth, int option_len,
1381 			  struct netlink_ext_ack *extack)
1382 {
1383 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_GTP_MAX + 1];
1384 	struct gtp_pdu_session_info *sinfo;
1385 	u8 len = key->enc_opts.len;
1386 	int err;
1387 
1388 	sinfo = (struct gtp_pdu_session_info *)&key->enc_opts.data[len];
1389 	memset(sinfo, 0xff, option_len);
1390 
1391 	if (!depth)
1392 		return sizeof(*sinfo);
1393 
1394 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_GTP) {
1395 		NL_SET_ERR_MSG_MOD(extack, "Non-gtp option type for mask");
1396 		return -EINVAL;
1397 	}
1398 
1399 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_GTP_MAX, nla,
1400 			       gtp_opt_policy, extack);
1401 	if (err < 0)
1402 		return err;
1403 
1404 	if (!option_len &&
1405 	    (!tb[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE] ||
1406 	     !tb[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI])) {
1407 		NL_SET_ERR_MSG_MOD(extack,
1408 				   "Missing tunnel key gtp option pdu type or qfi");
1409 		return -EINVAL;
1410 	}
1411 
1412 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE])
1413 		sinfo->pdu_type =
1414 			nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE]);
1415 
1416 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI])
1417 		sinfo->qfi = nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI]);
1418 
1419 	return sizeof(*sinfo);
1420 }
1421 
1422 static int fl_set_enc_opt(struct nlattr **tb, struct fl_flow_key *key,
1423 			  struct fl_flow_key *mask,
1424 			  struct netlink_ext_ack *extack)
1425 {
1426 	const struct nlattr *nla_enc_key, *nla_opt_key, *nla_opt_msk = NULL;
1427 	int err, option_len, key_depth, msk_depth = 0;
1428 
1429 	err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS],
1430 					     TCA_FLOWER_KEY_ENC_OPTS_MAX,
1431 					     enc_opts_policy, extack);
1432 	if (err)
1433 		return err;
1434 
1435 	nla_enc_key = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS]);
1436 
1437 	if (tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]) {
1438 		err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK],
1439 						     TCA_FLOWER_KEY_ENC_OPTS_MAX,
1440 						     enc_opts_policy, extack);
1441 		if (err)
1442 			return err;
1443 
1444 		nla_opt_msk = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
1445 		msk_depth = nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
1446 		if (!nla_ok(nla_opt_msk, msk_depth)) {
1447 			NL_SET_ERR_MSG(extack, "Invalid nested attribute for masks");
1448 			return -EINVAL;
1449 		}
1450 	}
1451 
1452 	nla_for_each_attr(nla_opt_key, nla_enc_key,
1453 			  nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS]), key_depth) {
1454 		switch (nla_type(nla_opt_key)) {
1455 		case TCA_FLOWER_KEY_ENC_OPTS_GENEVE:
1456 			if (key->enc_opts.dst_opt_type &&
1457 			    key->enc_opts.dst_opt_type != TUNNEL_GENEVE_OPT) {
1458 				NL_SET_ERR_MSG(extack, "Duplicate type for geneve options");
1459 				return -EINVAL;
1460 			}
1461 			option_len = 0;
1462 			key->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT;
1463 			option_len = fl_set_geneve_opt(nla_opt_key, key,
1464 						       key_depth, option_len,
1465 						       extack);
1466 			if (option_len < 0)
1467 				return option_len;
1468 
1469 			key->enc_opts.len += option_len;
1470 			/* At the same time we need to parse through the mask
1471 			 * in order to verify exact and mask attribute lengths.
1472 			 */
1473 			mask->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT;
1474 			option_len = fl_set_geneve_opt(nla_opt_msk, mask,
1475 						       msk_depth, option_len,
1476 						       extack);
1477 			if (option_len < 0)
1478 				return option_len;
1479 
1480 			mask->enc_opts.len += option_len;
1481 			if (key->enc_opts.len != mask->enc_opts.len) {
1482 				NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1483 				return -EINVAL;
1484 			}
1485 			break;
1486 		case TCA_FLOWER_KEY_ENC_OPTS_VXLAN:
1487 			if (key->enc_opts.dst_opt_type) {
1488 				NL_SET_ERR_MSG(extack, "Duplicate type for vxlan options");
1489 				return -EINVAL;
1490 			}
1491 			option_len = 0;
1492 			key->enc_opts.dst_opt_type = TUNNEL_VXLAN_OPT;
1493 			option_len = fl_set_vxlan_opt(nla_opt_key, key,
1494 						      key_depth, option_len,
1495 						      extack);
1496 			if (option_len < 0)
1497 				return option_len;
1498 
1499 			key->enc_opts.len += option_len;
1500 			/* At the same time we need to parse through the mask
1501 			 * in order to verify exact and mask attribute lengths.
1502 			 */
1503 			mask->enc_opts.dst_opt_type = TUNNEL_VXLAN_OPT;
1504 			option_len = fl_set_vxlan_opt(nla_opt_msk, mask,
1505 						      msk_depth, option_len,
1506 						      extack);
1507 			if (option_len < 0)
1508 				return option_len;
1509 
1510 			mask->enc_opts.len += option_len;
1511 			if (key->enc_opts.len != mask->enc_opts.len) {
1512 				NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1513 				return -EINVAL;
1514 			}
1515 			break;
1516 		case TCA_FLOWER_KEY_ENC_OPTS_ERSPAN:
1517 			if (key->enc_opts.dst_opt_type) {
1518 				NL_SET_ERR_MSG(extack, "Duplicate type for erspan options");
1519 				return -EINVAL;
1520 			}
1521 			option_len = 0;
1522 			key->enc_opts.dst_opt_type = TUNNEL_ERSPAN_OPT;
1523 			option_len = fl_set_erspan_opt(nla_opt_key, key,
1524 						       key_depth, option_len,
1525 						       extack);
1526 			if (option_len < 0)
1527 				return option_len;
1528 
1529 			key->enc_opts.len += option_len;
1530 			/* At the same time we need to parse through the mask
1531 			 * in order to verify exact and mask attribute lengths.
1532 			 */
1533 			mask->enc_opts.dst_opt_type = TUNNEL_ERSPAN_OPT;
1534 			option_len = fl_set_erspan_opt(nla_opt_msk, mask,
1535 						       msk_depth, option_len,
1536 						       extack);
1537 			if (option_len < 0)
1538 				return option_len;
1539 
1540 			mask->enc_opts.len += option_len;
1541 			if (key->enc_opts.len != mask->enc_opts.len) {
1542 				NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1543 				return -EINVAL;
1544 			}
1545 			break;
1546 		case TCA_FLOWER_KEY_ENC_OPTS_GTP:
1547 			if (key->enc_opts.dst_opt_type) {
1548 				NL_SET_ERR_MSG_MOD(extack,
1549 						   "Duplicate type for gtp options");
1550 				return -EINVAL;
1551 			}
1552 			option_len = 0;
1553 			key->enc_opts.dst_opt_type = TUNNEL_GTP_OPT;
1554 			option_len = fl_set_gtp_opt(nla_opt_key, key,
1555 						    key_depth, option_len,
1556 						    extack);
1557 			if (option_len < 0)
1558 				return option_len;
1559 
1560 			key->enc_opts.len += option_len;
1561 			/* At the same time we need to parse through the mask
1562 			 * in order to verify exact and mask attribute lengths.
1563 			 */
1564 			mask->enc_opts.dst_opt_type = TUNNEL_GTP_OPT;
1565 			option_len = fl_set_gtp_opt(nla_opt_msk, mask,
1566 						    msk_depth, option_len,
1567 						    extack);
1568 			if (option_len < 0)
1569 				return option_len;
1570 
1571 			mask->enc_opts.len += option_len;
1572 			if (key->enc_opts.len != mask->enc_opts.len) {
1573 				NL_SET_ERR_MSG_MOD(extack,
1574 						   "Key and mask miss aligned");
1575 				return -EINVAL;
1576 			}
1577 			break;
1578 		default:
1579 			NL_SET_ERR_MSG(extack, "Unknown tunnel option type");
1580 			return -EINVAL;
1581 		}
1582 
1583 		if (!msk_depth)
1584 			continue;
1585 
1586 		if (!nla_ok(nla_opt_msk, msk_depth)) {
1587 			NL_SET_ERR_MSG(extack, "A mask attribute is invalid");
1588 			return -EINVAL;
1589 		}
1590 		nla_opt_msk = nla_next(nla_opt_msk, &msk_depth);
1591 	}
1592 
1593 	return 0;
1594 }
1595 
1596 static int fl_validate_ct_state(u16 state, struct nlattr *tb,
1597 				struct netlink_ext_ack *extack)
1598 {
1599 	if (state && !(state & TCA_FLOWER_KEY_CT_FLAGS_TRACKED)) {
1600 		NL_SET_ERR_MSG_ATTR(extack, tb,
1601 				    "no trk, so no other flag can be set");
1602 		return -EINVAL;
1603 	}
1604 
1605 	if (state & TCA_FLOWER_KEY_CT_FLAGS_NEW &&
1606 	    state & TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED) {
1607 		NL_SET_ERR_MSG_ATTR(extack, tb,
1608 				    "new and est are mutually exclusive");
1609 		return -EINVAL;
1610 	}
1611 
1612 	if (state & TCA_FLOWER_KEY_CT_FLAGS_INVALID &&
1613 	    state & ~(TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
1614 		      TCA_FLOWER_KEY_CT_FLAGS_INVALID)) {
1615 		NL_SET_ERR_MSG_ATTR(extack, tb,
1616 				    "when inv is set, only trk may be set");
1617 		return -EINVAL;
1618 	}
1619 
1620 	if (state & TCA_FLOWER_KEY_CT_FLAGS_NEW &&
1621 	    state & TCA_FLOWER_KEY_CT_FLAGS_REPLY) {
1622 		NL_SET_ERR_MSG_ATTR(extack, tb,
1623 				    "new and rpl are mutually exclusive");
1624 		return -EINVAL;
1625 	}
1626 
1627 	return 0;
1628 }
1629 
1630 static int fl_set_key_ct(struct nlattr **tb,
1631 			 struct flow_dissector_key_ct *key,
1632 			 struct flow_dissector_key_ct *mask,
1633 			 struct netlink_ext_ack *extack)
1634 {
1635 	if (tb[TCA_FLOWER_KEY_CT_STATE]) {
1636 		int err;
1637 
1638 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK)) {
1639 			NL_SET_ERR_MSG(extack, "Conntrack isn't enabled");
1640 			return -EOPNOTSUPP;
1641 		}
1642 		fl_set_key_val(tb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE,
1643 			       &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK,
1644 			       sizeof(key->ct_state));
1645 
1646 		err = fl_validate_ct_state(key->ct_state & mask->ct_state,
1647 					   tb[TCA_FLOWER_KEY_CT_STATE_MASK],
1648 					   extack);
1649 		if (err)
1650 			return err;
1651 
1652 	}
1653 	if (tb[TCA_FLOWER_KEY_CT_ZONE]) {
1654 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)) {
1655 			NL_SET_ERR_MSG(extack, "Conntrack zones isn't enabled");
1656 			return -EOPNOTSUPP;
1657 		}
1658 		fl_set_key_val(tb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE,
1659 			       &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK,
1660 			       sizeof(key->ct_zone));
1661 	}
1662 	if (tb[TCA_FLOWER_KEY_CT_MARK]) {
1663 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)) {
1664 			NL_SET_ERR_MSG(extack, "Conntrack mark isn't enabled");
1665 			return -EOPNOTSUPP;
1666 		}
1667 		fl_set_key_val(tb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK,
1668 			       &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK,
1669 			       sizeof(key->ct_mark));
1670 	}
1671 	if (tb[TCA_FLOWER_KEY_CT_LABELS]) {
1672 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)) {
1673 			NL_SET_ERR_MSG(extack, "Conntrack labels aren't enabled");
1674 			return -EOPNOTSUPP;
1675 		}
1676 		fl_set_key_val(tb, key->ct_labels, TCA_FLOWER_KEY_CT_LABELS,
1677 			       mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK,
1678 			       sizeof(key->ct_labels));
1679 	}
1680 
1681 	return 0;
1682 }
1683 
1684 static bool is_vlan_key(struct nlattr *tb, __be16 *ethertype,
1685 			struct fl_flow_key *key, struct fl_flow_key *mask,
1686 			int vthresh)
1687 {
1688 	const bool good_num_of_vlans = key->num_of_vlans.num_of_vlans > vthresh;
1689 
1690 	if (!tb) {
1691 		*ethertype = 0;
1692 		return good_num_of_vlans;
1693 	}
1694 
1695 	*ethertype = nla_get_be16(tb);
1696 	if (good_num_of_vlans || eth_type_vlan(*ethertype))
1697 		return true;
1698 
1699 	key->basic.n_proto = *ethertype;
1700 	mask->basic.n_proto = cpu_to_be16(~0);
1701 	return false;
1702 }
1703 
1704 static void fl_set_key_cfm_md_level(struct nlattr **tb,
1705 				    struct fl_flow_key *key,
1706 				    struct fl_flow_key *mask,
1707 				    struct netlink_ext_ack *extack)
1708 {
1709 	u8 level;
1710 
1711 	if (!tb[TCA_FLOWER_KEY_CFM_MD_LEVEL])
1712 		return;
1713 
1714 	level = nla_get_u8(tb[TCA_FLOWER_KEY_CFM_MD_LEVEL]);
1715 	key->cfm.mdl_ver = FIELD_PREP(FLOW_DIS_CFM_MDL_MASK, level);
1716 	mask->cfm.mdl_ver = FLOW_DIS_CFM_MDL_MASK;
1717 }
1718 
1719 static void fl_set_key_cfm_opcode(struct nlattr **tb,
1720 				  struct fl_flow_key *key,
1721 				  struct fl_flow_key *mask,
1722 				  struct netlink_ext_ack *extack)
1723 {
1724 	fl_set_key_val(tb, &key->cfm.opcode, TCA_FLOWER_KEY_CFM_OPCODE,
1725 		       &mask->cfm.opcode, TCA_FLOWER_UNSPEC,
1726 		       sizeof(key->cfm.opcode));
1727 }
1728 
1729 static int fl_set_key_cfm(struct nlattr **tb,
1730 			  struct fl_flow_key *key,
1731 			  struct fl_flow_key *mask,
1732 			  struct netlink_ext_ack *extack)
1733 {
1734 	struct nlattr *nla_cfm_opt[TCA_FLOWER_KEY_CFM_OPT_MAX + 1];
1735 	int err;
1736 
1737 	if (!tb[TCA_FLOWER_KEY_CFM])
1738 		return 0;
1739 
1740 	err = nla_parse_nested(nla_cfm_opt, TCA_FLOWER_KEY_CFM_OPT_MAX,
1741 			       tb[TCA_FLOWER_KEY_CFM], cfm_opt_policy, extack);
1742 	if (err < 0)
1743 		return err;
1744 
1745 	fl_set_key_cfm_opcode(nla_cfm_opt, key, mask, extack);
1746 	fl_set_key_cfm_md_level(nla_cfm_opt, key, mask, extack);
1747 
1748 	return 0;
1749 }
1750 
1751 static int fl_set_key(struct net *net, struct nlattr **tb,
1752 		      struct fl_flow_key *key, struct fl_flow_key *mask,
1753 		      struct netlink_ext_ack *extack)
1754 {
1755 	__be16 ethertype;
1756 	int ret = 0;
1757 
1758 	if (tb[TCA_FLOWER_INDEV]) {
1759 		int err = tcf_change_indev(net, tb[TCA_FLOWER_INDEV], extack);
1760 		if (err < 0)
1761 			return err;
1762 		key->meta.ingress_ifindex = err;
1763 		mask->meta.ingress_ifindex = 0xffffffff;
1764 	}
1765 
1766 	fl_set_key_val(tb, &key->meta.l2_miss, TCA_FLOWER_L2_MISS,
1767 		       &mask->meta.l2_miss, TCA_FLOWER_UNSPEC,
1768 		       sizeof(key->meta.l2_miss));
1769 
1770 	fl_set_key_val(tb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
1771 		       mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
1772 		       sizeof(key->eth.dst));
1773 	fl_set_key_val(tb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
1774 		       mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
1775 		       sizeof(key->eth.src));
1776 	fl_set_key_val(tb, &key->num_of_vlans,
1777 		       TCA_FLOWER_KEY_NUM_OF_VLANS,
1778 		       &mask->num_of_vlans,
1779 		       TCA_FLOWER_UNSPEC,
1780 		       sizeof(key->num_of_vlans));
1781 
1782 	if (is_vlan_key(tb[TCA_FLOWER_KEY_ETH_TYPE], &ethertype, key, mask, 0)) {
1783 		fl_set_key_vlan(tb, ethertype, TCA_FLOWER_KEY_VLAN_ID,
1784 				TCA_FLOWER_KEY_VLAN_PRIO,
1785 				TCA_FLOWER_KEY_VLAN_ETH_TYPE,
1786 				&key->vlan, &mask->vlan);
1787 
1788 		if (is_vlan_key(tb[TCA_FLOWER_KEY_VLAN_ETH_TYPE],
1789 				&ethertype, key, mask, 1)) {
1790 			fl_set_key_vlan(tb, ethertype,
1791 					TCA_FLOWER_KEY_CVLAN_ID,
1792 					TCA_FLOWER_KEY_CVLAN_PRIO,
1793 					TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
1794 					&key->cvlan, &mask->cvlan);
1795 			fl_set_key_val(tb, &key->basic.n_proto,
1796 				       TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
1797 				       &mask->basic.n_proto,
1798 				       TCA_FLOWER_UNSPEC,
1799 				       sizeof(key->basic.n_proto));
1800 		}
1801 	}
1802 
1803 	if (key->basic.n_proto == htons(ETH_P_PPP_SES))
1804 		fl_set_key_pppoe(tb, &key->pppoe, &mask->pppoe, key, mask);
1805 
1806 	if (key->basic.n_proto == htons(ETH_P_IP) ||
1807 	    key->basic.n_proto == htons(ETH_P_IPV6)) {
1808 		fl_set_key_val(tb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
1809 			       &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
1810 			       sizeof(key->basic.ip_proto));
1811 		fl_set_key_ip(tb, false, &key->ip, &mask->ip);
1812 	}
1813 
1814 	if (tb[TCA_FLOWER_KEY_IPV4_SRC] || tb[TCA_FLOWER_KEY_IPV4_DST]) {
1815 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1816 		mask->control.addr_type = ~0;
1817 		fl_set_key_val(tb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
1818 			       &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
1819 			       sizeof(key->ipv4.src));
1820 		fl_set_key_val(tb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
1821 			       &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
1822 			       sizeof(key->ipv4.dst));
1823 	} else if (tb[TCA_FLOWER_KEY_IPV6_SRC] || tb[TCA_FLOWER_KEY_IPV6_DST]) {
1824 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1825 		mask->control.addr_type = ~0;
1826 		fl_set_key_val(tb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
1827 			       &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
1828 			       sizeof(key->ipv6.src));
1829 		fl_set_key_val(tb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
1830 			       &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
1831 			       sizeof(key->ipv6.dst));
1832 	}
1833 
1834 	if (key->basic.ip_proto == IPPROTO_TCP) {
1835 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
1836 			       &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
1837 			       sizeof(key->tp.src));
1838 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
1839 			       &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
1840 			       sizeof(key->tp.dst));
1841 		fl_set_key_val(tb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
1842 			       &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
1843 			       sizeof(key->tcp.flags));
1844 	} else if (key->basic.ip_proto == IPPROTO_UDP) {
1845 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
1846 			       &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
1847 			       sizeof(key->tp.src));
1848 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
1849 			       &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
1850 			       sizeof(key->tp.dst));
1851 	} else if (key->basic.ip_proto == IPPROTO_SCTP) {
1852 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
1853 			       &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
1854 			       sizeof(key->tp.src));
1855 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
1856 			       &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
1857 			       sizeof(key->tp.dst));
1858 	} else if (key->basic.n_proto == htons(ETH_P_IP) &&
1859 		   key->basic.ip_proto == IPPROTO_ICMP) {
1860 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV4_TYPE,
1861 			       &mask->icmp.type,
1862 			       TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
1863 			       sizeof(key->icmp.type));
1864 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV4_CODE,
1865 			       &mask->icmp.code,
1866 			       TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
1867 			       sizeof(key->icmp.code));
1868 	} else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
1869 		   key->basic.ip_proto == IPPROTO_ICMPV6) {
1870 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV6_TYPE,
1871 			       &mask->icmp.type,
1872 			       TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
1873 			       sizeof(key->icmp.type));
1874 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV6_CODE,
1875 			       &mask->icmp.code,
1876 			       TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
1877 			       sizeof(key->icmp.code));
1878 	} else if (key->basic.n_proto == htons(ETH_P_MPLS_UC) ||
1879 		   key->basic.n_proto == htons(ETH_P_MPLS_MC)) {
1880 		ret = fl_set_key_mpls(tb, &key->mpls, &mask->mpls, extack);
1881 		if (ret)
1882 			return ret;
1883 	} else if (key->basic.n_proto == htons(ETH_P_ARP) ||
1884 		   key->basic.n_proto == htons(ETH_P_RARP)) {
1885 		fl_set_key_val(tb, &key->arp.sip, TCA_FLOWER_KEY_ARP_SIP,
1886 			       &mask->arp.sip, TCA_FLOWER_KEY_ARP_SIP_MASK,
1887 			       sizeof(key->arp.sip));
1888 		fl_set_key_val(tb, &key->arp.tip, TCA_FLOWER_KEY_ARP_TIP,
1889 			       &mask->arp.tip, TCA_FLOWER_KEY_ARP_TIP_MASK,
1890 			       sizeof(key->arp.tip));
1891 		fl_set_key_val(tb, &key->arp.op, TCA_FLOWER_KEY_ARP_OP,
1892 			       &mask->arp.op, TCA_FLOWER_KEY_ARP_OP_MASK,
1893 			       sizeof(key->arp.op));
1894 		fl_set_key_val(tb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
1895 			       mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
1896 			       sizeof(key->arp.sha));
1897 		fl_set_key_val(tb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
1898 			       mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
1899 			       sizeof(key->arp.tha));
1900 	} else if (key->basic.ip_proto == IPPROTO_L2TP) {
1901 		fl_set_key_val(tb, &key->l2tpv3.session_id,
1902 			       TCA_FLOWER_KEY_L2TPV3_SID,
1903 			       &mask->l2tpv3.session_id, TCA_FLOWER_UNSPEC,
1904 			       sizeof(key->l2tpv3.session_id));
1905 	} else if (key->basic.n_proto  == htons(ETH_P_CFM)) {
1906 		ret = fl_set_key_cfm(tb, key, mask, extack);
1907 		if (ret)
1908 			return ret;
1909 	}
1910 
1911 	if (key->basic.ip_proto == IPPROTO_TCP ||
1912 	    key->basic.ip_proto == IPPROTO_UDP ||
1913 	    key->basic.ip_proto == IPPROTO_SCTP) {
1914 		ret = fl_set_key_port_range(tb, key, mask, extack);
1915 		if (ret)
1916 			return ret;
1917 	}
1918 
1919 	if (tb[TCA_FLOWER_KEY_SPI]) {
1920 		ret = fl_set_key_spi(tb, key, mask, extack);
1921 		if (ret)
1922 			return ret;
1923 	}
1924 
1925 	if (tb[TCA_FLOWER_KEY_ENC_IPV4_SRC] ||
1926 	    tb[TCA_FLOWER_KEY_ENC_IPV4_DST]) {
1927 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1928 		mask->enc_control.addr_type = ~0;
1929 		fl_set_key_val(tb, &key->enc_ipv4.src,
1930 			       TCA_FLOWER_KEY_ENC_IPV4_SRC,
1931 			       &mask->enc_ipv4.src,
1932 			       TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
1933 			       sizeof(key->enc_ipv4.src));
1934 		fl_set_key_val(tb, &key->enc_ipv4.dst,
1935 			       TCA_FLOWER_KEY_ENC_IPV4_DST,
1936 			       &mask->enc_ipv4.dst,
1937 			       TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
1938 			       sizeof(key->enc_ipv4.dst));
1939 	}
1940 
1941 	if (tb[TCA_FLOWER_KEY_ENC_IPV6_SRC] ||
1942 	    tb[TCA_FLOWER_KEY_ENC_IPV6_DST]) {
1943 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1944 		mask->enc_control.addr_type = ~0;
1945 		fl_set_key_val(tb, &key->enc_ipv6.src,
1946 			       TCA_FLOWER_KEY_ENC_IPV6_SRC,
1947 			       &mask->enc_ipv6.src,
1948 			       TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
1949 			       sizeof(key->enc_ipv6.src));
1950 		fl_set_key_val(tb, &key->enc_ipv6.dst,
1951 			       TCA_FLOWER_KEY_ENC_IPV6_DST,
1952 			       &mask->enc_ipv6.dst,
1953 			       TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
1954 			       sizeof(key->enc_ipv6.dst));
1955 	}
1956 
1957 	fl_set_key_val(tb, &key->enc_key_id.keyid, TCA_FLOWER_KEY_ENC_KEY_ID,
1958 		       &mask->enc_key_id.keyid, TCA_FLOWER_UNSPEC,
1959 		       sizeof(key->enc_key_id.keyid));
1960 
1961 	fl_set_key_val(tb, &key->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
1962 		       &mask->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
1963 		       sizeof(key->enc_tp.src));
1964 
1965 	fl_set_key_val(tb, &key->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
1966 		       &mask->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
1967 		       sizeof(key->enc_tp.dst));
1968 
1969 	fl_set_key_ip(tb, true, &key->enc_ip, &mask->enc_ip);
1970 
1971 	fl_set_key_val(tb, &key->hash.hash, TCA_FLOWER_KEY_HASH,
1972 		       &mask->hash.hash, TCA_FLOWER_KEY_HASH_MASK,
1973 		       sizeof(key->hash.hash));
1974 
1975 	if (tb[TCA_FLOWER_KEY_ENC_OPTS]) {
1976 		ret = fl_set_enc_opt(tb, key, mask, extack);
1977 		if (ret)
1978 			return ret;
1979 	}
1980 
1981 	ret = fl_set_key_ct(tb, &key->ct, &mask->ct, extack);
1982 	if (ret)
1983 		return ret;
1984 
1985 	if (tb[TCA_FLOWER_KEY_FLAGS])
1986 		ret = fl_set_key_flags(tb, &key->control.flags,
1987 				       &mask->control.flags, extack);
1988 
1989 	return ret;
1990 }
1991 
1992 static void fl_mask_copy(struct fl_flow_mask *dst,
1993 			 struct fl_flow_mask *src)
1994 {
1995 	const void *psrc = fl_key_get_start(&src->key, src);
1996 	void *pdst = fl_key_get_start(&dst->key, src);
1997 
1998 	memcpy(pdst, psrc, fl_mask_range(src));
1999 	dst->range = src->range;
2000 }
2001 
2002 static const struct rhashtable_params fl_ht_params = {
2003 	.key_offset = offsetof(struct cls_fl_filter, mkey), /* base offset */
2004 	.head_offset = offsetof(struct cls_fl_filter, ht_node),
2005 	.automatic_shrinking = true,
2006 };
2007 
2008 static int fl_init_mask_hashtable(struct fl_flow_mask *mask)
2009 {
2010 	mask->filter_ht_params = fl_ht_params;
2011 	mask->filter_ht_params.key_len = fl_mask_range(mask);
2012 	mask->filter_ht_params.key_offset += mask->range.start;
2013 
2014 	return rhashtable_init(&mask->ht, &mask->filter_ht_params);
2015 }
2016 
2017 #define FL_KEY_MEMBER_OFFSET(member) offsetof(struct fl_flow_key, member)
2018 #define FL_KEY_MEMBER_SIZE(member) sizeof_field(struct fl_flow_key, member)
2019 
2020 #define FL_KEY_IS_MASKED(mask, member)						\
2021 	memchr_inv(((char *)mask) + FL_KEY_MEMBER_OFFSET(member),		\
2022 		   0, FL_KEY_MEMBER_SIZE(member))				\
2023 
2024 #define FL_KEY_SET(keys, cnt, id, member)					\
2025 	do {									\
2026 		keys[cnt].key_id = id;						\
2027 		keys[cnt].offset = FL_KEY_MEMBER_OFFSET(member);		\
2028 		cnt++;								\
2029 	} while(0);
2030 
2031 #define FL_KEY_SET_IF_MASKED(mask, keys, cnt, id, member)			\
2032 	do {									\
2033 		if (FL_KEY_IS_MASKED(mask, member))				\
2034 			FL_KEY_SET(keys, cnt, id, member);			\
2035 	} while(0);
2036 
2037 static void fl_init_dissector(struct flow_dissector *dissector,
2038 			      struct fl_flow_key *mask)
2039 {
2040 	struct flow_dissector_key keys[FLOW_DISSECTOR_KEY_MAX];
2041 	size_t cnt = 0;
2042 
2043 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2044 			     FLOW_DISSECTOR_KEY_META, meta);
2045 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_CONTROL, control);
2046 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_BASIC, basic);
2047 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2048 			     FLOW_DISSECTOR_KEY_ETH_ADDRS, eth);
2049 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2050 			     FLOW_DISSECTOR_KEY_IPV4_ADDRS, ipv4);
2051 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2052 			     FLOW_DISSECTOR_KEY_IPV6_ADDRS, ipv6);
2053 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2054 			     FLOW_DISSECTOR_KEY_PORTS, tp);
2055 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2056 			     FLOW_DISSECTOR_KEY_PORTS_RANGE, tp_range);
2057 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2058 			     FLOW_DISSECTOR_KEY_IP, ip);
2059 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2060 			     FLOW_DISSECTOR_KEY_TCP, tcp);
2061 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2062 			     FLOW_DISSECTOR_KEY_ICMP, icmp);
2063 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2064 			     FLOW_DISSECTOR_KEY_ARP, arp);
2065 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2066 			     FLOW_DISSECTOR_KEY_MPLS, mpls);
2067 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2068 			     FLOW_DISSECTOR_KEY_VLAN, vlan);
2069 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2070 			     FLOW_DISSECTOR_KEY_CVLAN, cvlan);
2071 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2072 			     FLOW_DISSECTOR_KEY_ENC_KEYID, enc_key_id);
2073 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2074 			     FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, enc_ipv4);
2075 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2076 			     FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, enc_ipv6);
2077 	if (FL_KEY_IS_MASKED(mask, enc_ipv4) ||
2078 	    FL_KEY_IS_MASKED(mask, enc_ipv6))
2079 		FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_ENC_CONTROL,
2080 			   enc_control);
2081 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2082 			     FLOW_DISSECTOR_KEY_ENC_PORTS, enc_tp);
2083 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2084 			     FLOW_DISSECTOR_KEY_ENC_IP, enc_ip);
2085 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2086 			     FLOW_DISSECTOR_KEY_ENC_OPTS, enc_opts);
2087 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2088 			     FLOW_DISSECTOR_KEY_CT, ct);
2089 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2090 			     FLOW_DISSECTOR_KEY_HASH, hash);
2091 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2092 			     FLOW_DISSECTOR_KEY_NUM_OF_VLANS, num_of_vlans);
2093 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2094 			     FLOW_DISSECTOR_KEY_PPPOE, pppoe);
2095 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2096 			     FLOW_DISSECTOR_KEY_L2TPV3, l2tpv3);
2097 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2098 			     FLOW_DISSECTOR_KEY_IPSEC, ipsec);
2099 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2100 			     FLOW_DISSECTOR_KEY_CFM, cfm);
2101 
2102 	skb_flow_dissector_init(dissector, keys, cnt);
2103 }
2104 
2105 static struct fl_flow_mask *fl_create_new_mask(struct cls_fl_head *head,
2106 					       struct fl_flow_mask *mask)
2107 {
2108 	struct fl_flow_mask *newmask;
2109 	int err;
2110 
2111 	newmask = kzalloc(sizeof(*newmask), GFP_KERNEL);
2112 	if (!newmask)
2113 		return ERR_PTR(-ENOMEM);
2114 
2115 	fl_mask_copy(newmask, mask);
2116 
2117 	if ((newmask->key.tp_range.tp_min.dst &&
2118 	     newmask->key.tp_range.tp_max.dst) ||
2119 	    (newmask->key.tp_range.tp_min.src &&
2120 	     newmask->key.tp_range.tp_max.src))
2121 		newmask->flags |= TCA_FLOWER_MASK_FLAGS_RANGE;
2122 
2123 	err = fl_init_mask_hashtable(newmask);
2124 	if (err)
2125 		goto errout_free;
2126 
2127 	fl_init_dissector(&newmask->dissector, &newmask->key);
2128 
2129 	INIT_LIST_HEAD_RCU(&newmask->filters);
2130 
2131 	refcount_set(&newmask->refcnt, 1);
2132 	err = rhashtable_replace_fast(&head->ht, &mask->ht_node,
2133 				      &newmask->ht_node, mask_ht_params);
2134 	if (err)
2135 		goto errout_destroy;
2136 
2137 	spin_lock(&head->masks_lock);
2138 	list_add_tail_rcu(&newmask->list, &head->masks);
2139 	spin_unlock(&head->masks_lock);
2140 
2141 	return newmask;
2142 
2143 errout_destroy:
2144 	rhashtable_destroy(&newmask->ht);
2145 errout_free:
2146 	kfree(newmask);
2147 
2148 	return ERR_PTR(err);
2149 }
2150 
2151 static int fl_check_assign_mask(struct cls_fl_head *head,
2152 				struct cls_fl_filter *fnew,
2153 				struct cls_fl_filter *fold,
2154 				struct fl_flow_mask *mask)
2155 {
2156 	struct fl_flow_mask *newmask;
2157 	int ret = 0;
2158 
2159 	rcu_read_lock();
2160 
2161 	/* Insert mask as temporary node to prevent concurrent creation of mask
2162 	 * with same key. Any concurrent lookups with same key will return
2163 	 * -EAGAIN because mask's refcnt is zero.
2164 	 */
2165 	fnew->mask = rhashtable_lookup_get_insert_fast(&head->ht,
2166 						       &mask->ht_node,
2167 						       mask_ht_params);
2168 	if (!fnew->mask) {
2169 		rcu_read_unlock();
2170 
2171 		if (fold) {
2172 			ret = -EINVAL;
2173 			goto errout_cleanup;
2174 		}
2175 
2176 		newmask = fl_create_new_mask(head, mask);
2177 		if (IS_ERR(newmask)) {
2178 			ret = PTR_ERR(newmask);
2179 			goto errout_cleanup;
2180 		}
2181 
2182 		fnew->mask = newmask;
2183 		return 0;
2184 	} else if (IS_ERR(fnew->mask)) {
2185 		ret = PTR_ERR(fnew->mask);
2186 	} else if (fold && fold->mask != fnew->mask) {
2187 		ret = -EINVAL;
2188 	} else if (!refcount_inc_not_zero(&fnew->mask->refcnt)) {
2189 		/* Mask was deleted concurrently, try again */
2190 		ret = -EAGAIN;
2191 	}
2192 	rcu_read_unlock();
2193 	return ret;
2194 
2195 errout_cleanup:
2196 	rhashtable_remove_fast(&head->ht, &mask->ht_node,
2197 			       mask_ht_params);
2198 	return ret;
2199 }
2200 
2201 static bool fl_needs_tc_skb_ext(const struct fl_flow_key *mask)
2202 {
2203 	return mask->meta.l2_miss;
2204 }
2205 
2206 static int fl_ht_insert_unique(struct cls_fl_filter *fnew,
2207 			       struct cls_fl_filter *fold,
2208 			       bool *in_ht)
2209 {
2210 	struct fl_flow_mask *mask = fnew->mask;
2211 	int err;
2212 
2213 	err = rhashtable_lookup_insert_fast(&mask->ht,
2214 					    &fnew->ht_node,
2215 					    mask->filter_ht_params);
2216 	if (err) {
2217 		*in_ht = false;
2218 		/* It is okay if filter with same key exists when
2219 		 * overwriting.
2220 		 */
2221 		return fold && err == -EEXIST ? 0 : err;
2222 	}
2223 
2224 	*in_ht = true;
2225 	return 0;
2226 }
2227 
2228 static int fl_change(struct net *net, struct sk_buff *in_skb,
2229 		     struct tcf_proto *tp, unsigned long base,
2230 		     u32 handle, struct nlattr **tca,
2231 		     void **arg, u32 flags,
2232 		     struct netlink_ext_ack *extack)
2233 {
2234 	struct cls_fl_head *head = fl_head_dereference(tp);
2235 	bool rtnl_held = !(flags & TCA_ACT_FLAGS_NO_RTNL);
2236 	struct cls_fl_filter *fold = *arg;
2237 	bool bound_to_filter = false;
2238 	struct cls_fl_filter *fnew;
2239 	struct fl_flow_mask *mask;
2240 	struct nlattr **tb;
2241 	bool in_ht;
2242 	int err;
2243 
2244 	if (!tca[TCA_OPTIONS]) {
2245 		err = -EINVAL;
2246 		goto errout_fold;
2247 	}
2248 
2249 	mask = kzalloc(sizeof(struct fl_flow_mask), GFP_KERNEL);
2250 	if (!mask) {
2251 		err = -ENOBUFS;
2252 		goto errout_fold;
2253 	}
2254 
2255 	tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
2256 	if (!tb) {
2257 		err = -ENOBUFS;
2258 		goto errout_mask_alloc;
2259 	}
2260 
2261 	err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
2262 					  tca[TCA_OPTIONS], fl_policy, NULL);
2263 	if (err < 0)
2264 		goto errout_tb;
2265 
2266 	if (fold && handle && fold->handle != handle) {
2267 		err = -EINVAL;
2268 		goto errout_tb;
2269 	}
2270 
2271 	fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
2272 	if (!fnew) {
2273 		err = -ENOBUFS;
2274 		goto errout_tb;
2275 	}
2276 	INIT_LIST_HEAD(&fnew->hw_list);
2277 	refcount_set(&fnew->refcnt, 1);
2278 
2279 	if (tb[TCA_FLOWER_FLAGS]) {
2280 		fnew->flags = nla_get_u32(tb[TCA_FLOWER_FLAGS]);
2281 
2282 		if (!tc_flags_valid(fnew->flags)) {
2283 			kfree(fnew);
2284 			err = -EINVAL;
2285 			goto errout_tb;
2286 		}
2287 	}
2288 
2289 	if (!fold) {
2290 		spin_lock(&tp->lock);
2291 		if (!handle) {
2292 			handle = 1;
2293 			err = idr_alloc_u32(&head->handle_idr, NULL, &handle,
2294 					    INT_MAX, GFP_ATOMIC);
2295 		} else {
2296 			err = idr_alloc_u32(&head->handle_idr, NULL, &handle,
2297 					    handle, GFP_ATOMIC);
2298 
2299 			/* Filter with specified handle was concurrently
2300 			 * inserted after initial check in cls_api. This is not
2301 			 * necessarily an error if NLM_F_EXCL is not set in
2302 			 * message flags. Returning EAGAIN will cause cls_api to
2303 			 * try to update concurrently inserted rule.
2304 			 */
2305 			if (err == -ENOSPC)
2306 				err = -EAGAIN;
2307 		}
2308 		spin_unlock(&tp->lock);
2309 
2310 		if (err) {
2311 			kfree(fnew);
2312 			goto errout_tb;
2313 		}
2314 	}
2315 	fnew->handle = handle;
2316 
2317 	err = tcf_exts_init_ex(&fnew->exts, net, TCA_FLOWER_ACT, 0, tp, handle,
2318 			       !tc_skip_hw(fnew->flags));
2319 	if (err < 0)
2320 		goto errout_idr;
2321 
2322 	err = tcf_exts_validate_ex(net, tp, tb, tca[TCA_RATE],
2323 				   &fnew->exts, flags, fnew->flags,
2324 				   extack);
2325 	if (err < 0)
2326 		goto errout_idr;
2327 
2328 	if (tb[TCA_FLOWER_CLASSID]) {
2329 		fnew->res.classid = nla_get_u32(tb[TCA_FLOWER_CLASSID]);
2330 		if (flags & TCA_ACT_FLAGS_NO_RTNL)
2331 			rtnl_lock();
2332 		tcf_bind_filter(tp, &fnew->res, base);
2333 		if (flags & TCA_ACT_FLAGS_NO_RTNL)
2334 			rtnl_unlock();
2335 		bound_to_filter = true;
2336 	}
2337 
2338 	err = fl_set_key(net, tb, &fnew->key, &mask->key, extack);
2339 	if (err)
2340 		goto unbind_filter;
2341 
2342 	fl_mask_update_range(mask);
2343 	fl_set_masked_key(&fnew->mkey, &fnew->key, mask);
2344 
2345 	if (!fl_mask_fits_tmplt(tp->chain->tmplt_priv, mask)) {
2346 		NL_SET_ERR_MSG_MOD(extack, "Mask does not fit the template");
2347 		err = -EINVAL;
2348 		goto unbind_filter;
2349 	}
2350 
2351 	/* Enable tc skb extension if filter matches on data extracted from
2352 	 * this extension.
2353 	 */
2354 	if (fl_needs_tc_skb_ext(&mask->key)) {
2355 		fnew->needs_tc_skb_ext = 1;
2356 		tc_skb_ext_tc_enable();
2357 	}
2358 
2359 	err = fl_check_assign_mask(head, fnew, fold, mask);
2360 	if (err)
2361 		goto unbind_filter;
2362 
2363 	err = fl_ht_insert_unique(fnew, fold, &in_ht);
2364 	if (err)
2365 		goto errout_mask;
2366 
2367 	if (!tc_skip_hw(fnew->flags)) {
2368 		err = fl_hw_replace_filter(tp, fnew, rtnl_held, extack);
2369 		if (err)
2370 			goto errout_ht;
2371 	}
2372 
2373 	if (!tc_in_hw(fnew->flags))
2374 		fnew->flags |= TCA_CLS_FLAGS_NOT_IN_HW;
2375 
2376 	spin_lock(&tp->lock);
2377 
2378 	/* tp was deleted concurrently. -EAGAIN will cause caller to lookup
2379 	 * proto again or create new one, if necessary.
2380 	 */
2381 	if (tp->deleting) {
2382 		err = -EAGAIN;
2383 		goto errout_hw;
2384 	}
2385 
2386 	if (fold) {
2387 		/* Fold filter was deleted concurrently. Retry lookup. */
2388 		if (fold->deleted) {
2389 			err = -EAGAIN;
2390 			goto errout_hw;
2391 		}
2392 
2393 		fnew->handle = handle;
2394 
2395 		if (!in_ht) {
2396 			struct rhashtable_params params =
2397 				fnew->mask->filter_ht_params;
2398 
2399 			err = rhashtable_insert_fast(&fnew->mask->ht,
2400 						     &fnew->ht_node,
2401 						     params);
2402 			if (err)
2403 				goto errout_hw;
2404 			in_ht = true;
2405 		}
2406 
2407 		refcount_inc(&fnew->refcnt);
2408 		rhashtable_remove_fast(&fold->mask->ht,
2409 				       &fold->ht_node,
2410 				       fold->mask->filter_ht_params);
2411 		idr_replace(&head->handle_idr, fnew, fnew->handle);
2412 		list_replace_rcu(&fold->list, &fnew->list);
2413 		fold->deleted = true;
2414 
2415 		spin_unlock(&tp->lock);
2416 
2417 		fl_mask_put(head, fold->mask);
2418 		if (!tc_skip_hw(fold->flags))
2419 			fl_hw_destroy_filter(tp, fold, rtnl_held, NULL);
2420 		tcf_unbind_filter(tp, &fold->res);
2421 		/* Caller holds reference to fold, so refcnt is always > 0
2422 		 * after this.
2423 		 */
2424 		refcount_dec(&fold->refcnt);
2425 		__fl_put(fold);
2426 	} else {
2427 		idr_replace(&head->handle_idr, fnew, fnew->handle);
2428 
2429 		refcount_inc(&fnew->refcnt);
2430 		list_add_tail_rcu(&fnew->list, &fnew->mask->filters);
2431 		spin_unlock(&tp->lock);
2432 	}
2433 
2434 	*arg = fnew;
2435 
2436 	kfree(tb);
2437 	tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work);
2438 	return 0;
2439 
2440 errout_ht:
2441 	spin_lock(&tp->lock);
2442 errout_hw:
2443 	fnew->deleted = true;
2444 	spin_unlock(&tp->lock);
2445 	if (!tc_skip_hw(fnew->flags))
2446 		fl_hw_destroy_filter(tp, fnew, rtnl_held, NULL);
2447 	if (in_ht)
2448 		rhashtable_remove_fast(&fnew->mask->ht, &fnew->ht_node,
2449 				       fnew->mask->filter_ht_params);
2450 errout_mask:
2451 	fl_mask_put(head, fnew->mask);
2452 
2453 unbind_filter:
2454 	if (bound_to_filter) {
2455 		if (flags & TCA_ACT_FLAGS_NO_RTNL)
2456 			rtnl_lock();
2457 		tcf_unbind_filter(tp, &fnew->res);
2458 		if (flags & TCA_ACT_FLAGS_NO_RTNL)
2459 			rtnl_unlock();
2460 	}
2461 
2462 errout_idr:
2463 	if (!fold) {
2464 		spin_lock(&tp->lock);
2465 		idr_remove(&head->handle_idr, fnew->handle);
2466 		spin_unlock(&tp->lock);
2467 	}
2468 	__fl_put(fnew);
2469 errout_tb:
2470 	kfree(tb);
2471 errout_mask_alloc:
2472 	tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work);
2473 errout_fold:
2474 	if (fold)
2475 		__fl_put(fold);
2476 	return err;
2477 }
2478 
2479 static int fl_delete(struct tcf_proto *tp, void *arg, bool *last,
2480 		     bool rtnl_held, struct netlink_ext_ack *extack)
2481 {
2482 	struct cls_fl_head *head = fl_head_dereference(tp);
2483 	struct cls_fl_filter *f = arg;
2484 	bool last_on_mask;
2485 	int err = 0;
2486 
2487 	err = __fl_delete(tp, f, &last_on_mask, rtnl_held, extack);
2488 	*last = list_empty(&head->masks);
2489 	__fl_put(f);
2490 
2491 	return err;
2492 }
2493 
2494 static void fl_walk(struct tcf_proto *tp, struct tcf_walker *arg,
2495 		    bool rtnl_held)
2496 {
2497 	struct cls_fl_head *head = fl_head_dereference(tp);
2498 	unsigned long id = arg->cookie, tmp;
2499 	struct cls_fl_filter *f;
2500 
2501 	arg->count = arg->skip;
2502 
2503 	rcu_read_lock();
2504 	idr_for_each_entry_continue_ul(&head->handle_idr, f, tmp, id) {
2505 		/* don't return filters that are being deleted */
2506 		if (!f || !refcount_inc_not_zero(&f->refcnt))
2507 			continue;
2508 		rcu_read_unlock();
2509 
2510 		if (arg->fn(tp, f, arg) < 0) {
2511 			__fl_put(f);
2512 			arg->stop = 1;
2513 			rcu_read_lock();
2514 			break;
2515 		}
2516 		__fl_put(f);
2517 		arg->count++;
2518 		rcu_read_lock();
2519 	}
2520 	rcu_read_unlock();
2521 	arg->cookie = id;
2522 }
2523 
2524 static struct cls_fl_filter *
2525 fl_get_next_hw_filter(struct tcf_proto *tp, struct cls_fl_filter *f, bool add)
2526 {
2527 	struct cls_fl_head *head = fl_head_dereference(tp);
2528 
2529 	spin_lock(&tp->lock);
2530 	if (list_empty(&head->hw_filters)) {
2531 		spin_unlock(&tp->lock);
2532 		return NULL;
2533 	}
2534 
2535 	if (!f)
2536 		f = list_entry(&head->hw_filters, struct cls_fl_filter,
2537 			       hw_list);
2538 	list_for_each_entry_continue(f, &head->hw_filters, hw_list) {
2539 		if (!(add && f->deleted) && refcount_inc_not_zero(&f->refcnt)) {
2540 			spin_unlock(&tp->lock);
2541 			return f;
2542 		}
2543 	}
2544 
2545 	spin_unlock(&tp->lock);
2546 	return NULL;
2547 }
2548 
2549 static int fl_reoffload(struct tcf_proto *tp, bool add, flow_setup_cb_t *cb,
2550 			void *cb_priv, struct netlink_ext_ack *extack)
2551 {
2552 	struct tcf_block *block = tp->chain->block;
2553 	struct flow_cls_offload cls_flower = {};
2554 	struct cls_fl_filter *f = NULL;
2555 	int err;
2556 
2557 	/* hw_filters list can only be changed by hw offload functions after
2558 	 * obtaining rtnl lock. Make sure it is not changed while reoffload is
2559 	 * iterating it.
2560 	 */
2561 	ASSERT_RTNL();
2562 
2563 	while ((f = fl_get_next_hw_filter(tp, f, add))) {
2564 		cls_flower.rule =
2565 			flow_rule_alloc(tcf_exts_num_actions(&f->exts));
2566 		if (!cls_flower.rule) {
2567 			__fl_put(f);
2568 			return -ENOMEM;
2569 		}
2570 
2571 		tc_cls_common_offload_init(&cls_flower.common, tp, f->flags,
2572 					   extack);
2573 		cls_flower.command = add ?
2574 			FLOW_CLS_REPLACE : FLOW_CLS_DESTROY;
2575 		cls_flower.cookie = (unsigned long)f;
2576 		cls_flower.rule->match.dissector = &f->mask->dissector;
2577 		cls_flower.rule->match.mask = &f->mask->key;
2578 		cls_flower.rule->match.key = &f->mkey;
2579 
2580 		err = tc_setup_offload_action(&cls_flower.rule->action, &f->exts,
2581 					      cls_flower.common.extack);
2582 		if (err) {
2583 			kfree(cls_flower.rule);
2584 			if (tc_skip_sw(f->flags)) {
2585 				__fl_put(f);
2586 				return err;
2587 			}
2588 			goto next_flow;
2589 		}
2590 
2591 		cls_flower.classid = f->res.classid;
2592 
2593 		err = tc_setup_cb_reoffload(block, tp, add, cb,
2594 					    TC_SETUP_CLSFLOWER, &cls_flower,
2595 					    cb_priv, &f->flags,
2596 					    &f->in_hw_count);
2597 		tc_cleanup_offload_action(&cls_flower.rule->action);
2598 		kfree(cls_flower.rule);
2599 
2600 		if (err) {
2601 			__fl_put(f);
2602 			return err;
2603 		}
2604 next_flow:
2605 		__fl_put(f);
2606 	}
2607 
2608 	return 0;
2609 }
2610 
2611 static void fl_hw_add(struct tcf_proto *tp, void *type_data)
2612 {
2613 	struct flow_cls_offload *cls_flower = type_data;
2614 	struct cls_fl_filter *f =
2615 		(struct cls_fl_filter *) cls_flower->cookie;
2616 	struct cls_fl_head *head = fl_head_dereference(tp);
2617 
2618 	spin_lock(&tp->lock);
2619 	list_add(&f->hw_list, &head->hw_filters);
2620 	spin_unlock(&tp->lock);
2621 }
2622 
2623 static void fl_hw_del(struct tcf_proto *tp, void *type_data)
2624 {
2625 	struct flow_cls_offload *cls_flower = type_data;
2626 	struct cls_fl_filter *f =
2627 		(struct cls_fl_filter *) cls_flower->cookie;
2628 
2629 	spin_lock(&tp->lock);
2630 	if (!list_empty(&f->hw_list))
2631 		list_del_init(&f->hw_list);
2632 	spin_unlock(&tp->lock);
2633 }
2634 
2635 static int fl_hw_create_tmplt(struct tcf_chain *chain,
2636 			      struct fl_flow_tmplt *tmplt)
2637 {
2638 	struct flow_cls_offload cls_flower = {};
2639 	struct tcf_block *block = chain->block;
2640 
2641 	cls_flower.rule = flow_rule_alloc(0);
2642 	if (!cls_flower.rule)
2643 		return -ENOMEM;
2644 
2645 	cls_flower.common.chain_index = chain->index;
2646 	cls_flower.command = FLOW_CLS_TMPLT_CREATE;
2647 	cls_flower.cookie = (unsigned long) tmplt;
2648 	cls_flower.rule->match.dissector = &tmplt->dissector;
2649 	cls_flower.rule->match.mask = &tmplt->mask;
2650 	cls_flower.rule->match.key = &tmplt->dummy_key;
2651 
2652 	/* We don't care if driver (any of them) fails to handle this
2653 	 * call. It serves just as a hint for it.
2654 	 */
2655 	tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, true);
2656 	kfree(cls_flower.rule);
2657 
2658 	return 0;
2659 }
2660 
2661 static void fl_hw_destroy_tmplt(struct tcf_chain *chain,
2662 				struct fl_flow_tmplt *tmplt)
2663 {
2664 	struct flow_cls_offload cls_flower = {};
2665 	struct tcf_block *block = chain->block;
2666 
2667 	cls_flower.common.chain_index = chain->index;
2668 	cls_flower.command = FLOW_CLS_TMPLT_DESTROY;
2669 	cls_flower.cookie = (unsigned long) tmplt;
2670 
2671 	tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, true);
2672 }
2673 
2674 static void *fl_tmplt_create(struct net *net, struct tcf_chain *chain,
2675 			     struct nlattr **tca,
2676 			     struct netlink_ext_ack *extack)
2677 {
2678 	struct fl_flow_tmplt *tmplt;
2679 	struct nlattr **tb;
2680 	int err;
2681 
2682 	if (!tca[TCA_OPTIONS])
2683 		return ERR_PTR(-EINVAL);
2684 
2685 	tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
2686 	if (!tb)
2687 		return ERR_PTR(-ENOBUFS);
2688 	err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
2689 					  tca[TCA_OPTIONS], fl_policy, NULL);
2690 	if (err)
2691 		goto errout_tb;
2692 
2693 	tmplt = kzalloc(sizeof(*tmplt), GFP_KERNEL);
2694 	if (!tmplt) {
2695 		err = -ENOMEM;
2696 		goto errout_tb;
2697 	}
2698 	tmplt->chain = chain;
2699 	err = fl_set_key(net, tb, &tmplt->dummy_key, &tmplt->mask, extack);
2700 	if (err)
2701 		goto errout_tmplt;
2702 
2703 	fl_init_dissector(&tmplt->dissector, &tmplt->mask);
2704 
2705 	err = fl_hw_create_tmplt(chain, tmplt);
2706 	if (err)
2707 		goto errout_tmplt;
2708 
2709 	kfree(tb);
2710 	return tmplt;
2711 
2712 errout_tmplt:
2713 	kfree(tmplt);
2714 errout_tb:
2715 	kfree(tb);
2716 	return ERR_PTR(err);
2717 }
2718 
2719 static void fl_tmplt_destroy(void *tmplt_priv)
2720 {
2721 	struct fl_flow_tmplt *tmplt = tmplt_priv;
2722 
2723 	fl_hw_destroy_tmplt(tmplt->chain, tmplt);
2724 	kfree(tmplt);
2725 }
2726 
2727 static void fl_tmplt_reoffload(struct tcf_chain *chain, bool add,
2728 			       flow_setup_cb_t *cb, void *cb_priv)
2729 {
2730 	struct fl_flow_tmplt *tmplt = chain->tmplt_priv;
2731 	struct flow_cls_offload cls_flower = {};
2732 
2733 	cls_flower.rule = flow_rule_alloc(0);
2734 	if (!cls_flower.rule)
2735 		return;
2736 
2737 	cls_flower.common.chain_index = chain->index;
2738 	cls_flower.command = add ? FLOW_CLS_TMPLT_CREATE :
2739 				   FLOW_CLS_TMPLT_DESTROY;
2740 	cls_flower.cookie = (unsigned long) tmplt;
2741 	cls_flower.rule->match.dissector = &tmplt->dissector;
2742 	cls_flower.rule->match.mask = &tmplt->mask;
2743 	cls_flower.rule->match.key = &tmplt->dummy_key;
2744 
2745 	cb(TC_SETUP_CLSFLOWER, &cls_flower, cb_priv);
2746 	kfree(cls_flower.rule);
2747 }
2748 
2749 static int fl_dump_key_val(struct sk_buff *skb,
2750 			   void *val, int val_type,
2751 			   void *mask, int mask_type, int len)
2752 {
2753 	int err;
2754 
2755 	if (!memchr_inv(mask, 0, len))
2756 		return 0;
2757 	err = nla_put(skb, val_type, len, val);
2758 	if (err)
2759 		return err;
2760 	if (mask_type != TCA_FLOWER_UNSPEC) {
2761 		err = nla_put(skb, mask_type, len, mask);
2762 		if (err)
2763 			return err;
2764 	}
2765 	return 0;
2766 }
2767 
2768 static int fl_dump_key_port_range(struct sk_buff *skb, struct fl_flow_key *key,
2769 				  struct fl_flow_key *mask)
2770 {
2771 	if (fl_dump_key_val(skb, &key->tp_range.tp_min.dst,
2772 			    TCA_FLOWER_KEY_PORT_DST_MIN,
2773 			    &mask->tp_range.tp_min.dst, TCA_FLOWER_UNSPEC,
2774 			    sizeof(key->tp_range.tp_min.dst)) ||
2775 	    fl_dump_key_val(skb, &key->tp_range.tp_max.dst,
2776 			    TCA_FLOWER_KEY_PORT_DST_MAX,
2777 			    &mask->tp_range.tp_max.dst, TCA_FLOWER_UNSPEC,
2778 			    sizeof(key->tp_range.tp_max.dst)) ||
2779 	    fl_dump_key_val(skb, &key->tp_range.tp_min.src,
2780 			    TCA_FLOWER_KEY_PORT_SRC_MIN,
2781 			    &mask->tp_range.tp_min.src, TCA_FLOWER_UNSPEC,
2782 			    sizeof(key->tp_range.tp_min.src)) ||
2783 	    fl_dump_key_val(skb, &key->tp_range.tp_max.src,
2784 			    TCA_FLOWER_KEY_PORT_SRC_MAX,
2785 			    &mask->tp_range.tp_max.src, TCA_FLOWER_UNSPEC,
2786 			    sizeof(key->tp_range.tp_max.src)))
2787 		return -1;
2788 
2789 	return 0;
2790 }
2791 
2792 static int fl_dump_key_mpls_opt_lse(struct sk_buff *skb,
2793 				    struct flow_dissector_key_mpls *mpls_key,
2794 				    struct flow_dissector_key_mpls *mpls_mask,
2795 				    u8 lse_index)
2796 {
2797 	struct flow_dissector_mpls_lse *lse_mask = &mpls_mask->ls[lse_index];
2798 	struct flow_dissector_mpls_lse *lse_key = &mpls_key->ls[lse_index];
2799 	int err;
2800 
2801 	err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH,
2802 			 lse_index + 1);
2803 	if (err)
2804 		return err;
2805 
2806 	if (lse_mask->mpls_ttl) {
2807 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL,
2808 				 lse_key->mpls_ttl);
2809 		if (err)
2810 			return err;
2811 	}
2812 	if (lse_mask->mpls_bos) {
2813 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS,
2814 				 lse_key->mpls_bos);
2815 		if (err)
2816 			return err;
2817 	}
2818 	if (lse_mask->mpls_tc) {
2819 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_TC,
2820 				 lse_key->mpls_tc);
2821 		if (err)
2822 			return err;
2823 	}
2824 	if (lse_mask->mpls_label) {
2825 		err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL,
2826 				  lse_key->mpls_label);
2827 		if (err)
2828 			return err;
2829 	}
2830 
2831 	return 0;
2832 }
2833 
2834 static int fl_dump_key_mpls_opts(struct sk_buff *skb,
2835 				 struct flow_dissector_key_mpls *mpls_key,
2836 				 struct flow_dissector_key_mpls *mpls_mask)
2837 {
2838 	struct nlattr *opts;
2839 	struct nlattr *lse;
2840 	u8 lse_index;
2841 	int err;
2842 
2843 	opts = nla_nest_start(skb, TCA_FLOWER_KEY_MPLS_OPTS);
2844 	if (!opts)
2845 		return -EMSGSIZE;
2846 
2847 	for (lse_index = 0; lse_index < FLOW_DIS_MPLS_MAX; lse_index++) {
2848 		if (!(mpls_mask->used_lses & 1 << lse_index))
2849 			continue;
2850 
2851 		lse = nla_nest_start(skb, TCA_FLOWER_KEY_MPLS_OPTS_LSE);
2852 		if (!lse) {
2853 			err = -EMSGSIZE;
2854 			goto err_opts;
2855 		}
2856 
2857 		err = fl_dump_key_mpls_opt_lse(skb, mpls_key, mpls_mask,
2858 					       lse_index);
2859 		if (err)
2860 			goto err_opts_lse;
2861 		nla_nest_end(skb, lse);
2862 	}
2863 	nla_nest_end(skb, opts);
2864 
2865 	return 0;
2866 
2867 err_opts_lse:
2868 	nla_nest_cancel(skb, lse);
2869 err_opts:
2870 	nla_nest_cancel(skb, opts);
2871 
2872 	return err;
2873 }
2874 
2875 static int fl_dump_key_mpls(struct sk_buff *skb,
2876 			    struct flow_dissector_key_mpls *mpls_key,
2877 			    struct flow_dissector_key_mpls *mpls_mask)
2878 {
2879 	struct flow_dissector_mpls_lse *lse_mask;
2880 	struct flow_dissector_mpls_lse *lse_key;
2881 	int err;
2882 
2883 	if (!mpls_mask->used_lses)
2884 		return 0;
2885 
2886 	lse_mask = &mpls_mask->ls[0];
2887 	lse_key = &mpls_key->ls[0];
2888 
2889 	/* For backward compatibility, don't use the MPLS nested attributes if
2890 	 * the rule can be expressed using the old attributes.
2891 	 */
2892 	if (mpls_mask->used_lses & ~1 ||
2893 	    (!lse_mask->mpls_ttl && !lse_mask->mpls_bos &&
2894 	     !lse_mask->mpls_tc && !lse_mask->mpls_label))
2895 		return fl_dump_key_mpls_opts(skb, mpls_key, mpls_mask);
2896 
2897 	if (lse_mask->mpls_ttl) {
2898 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TTL,
2899 				 lse_key->mpls_ttl);
2900 		if (err)
2901 			return err;
2902 	}
2903 	if (lse_mask->mpls_tc) {
2904 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TC,
2905 				 lse_key->mpls_tc);
2906 		if (err)
2907 			return err;
2908 	}
2909 	if (lse_mask->mpls_label) {
2910 		err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_LABEL,
2911 				  lse_key->mpls_label);
2912 		if (err)
2913 			return err;
2914 	}
2915 	if (lse_mask->mpls_bos) {
2916 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_BOS,
2917 				 lse_key->mpls_bos);
2918 		if (err)
2919 			return err;
2920 	}
2921 	return 0;
2922 }
2923 
2924 static int fl_dump_key_ip(struct sk_buff *skb, bool encap,
2925 			  struct flow_dissector_key_ip *key,
2926 			  struct flow_dissector_key_ip *mask)
2927 {
2928 	int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
2929 	int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
2930 	int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
2931 	int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
2932 
2933 	if (fl_dump_key_val(skb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos)) ||
2934 	    fl_dump_key_val(skb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl)))
2935 		return -1;
2936 
2937 	return 0;
2938 }
2939 
2940 static int fl_dump_key_vlan(struct sk_buff *skb,
2941 			    int vlan_id_key, int vlan_prio_key,
2942 			    struct flow_dissector_key_vlan *vlan_key,
2943 			    struct flow_dissector_key_vlan *vlan_mask)
2944 {
2945 	int err;
2946 
2947 	if (!memchr_inv(vlan_mask, 0, sizeof(*vlan_mask)))
2948 		return 0;
2949 	if (vlan_mask->vlan_id) {
2950 		err = nla_put_u16(skb, vlan_id_key,
2951 				  vlan_key->vlan_id);
2952 		if (err)
2953 			return err;
2954 	}
2955 	if (vlan_mask->vlan_priority) {
2956 		err = nla_put_u8(skb, vlan_prio_key,
2957 				 vlan_key->vlan_priority);
2958 		if (err)
2959 			return err;
2960 	}
2961 	return 0;
2962 }
2963 
2964 static void fl_get_key_flag(u32 dissector_key, u32 dissector_mask,
2965 			    u32 *flower_key, u32 *flower_mask,
2966 			    u32 flower_flag_bit, u32 dissector_flag_bit)
2967 {
2968 	if (dissector_mask & dissector_flag_bit) {
2969 		*flower_mask |= flower_flag_bit;
2970 		if (dissector_key & dissector_flag_bit)
2971 			*flower_key |= flower_flag_bit;
2972 	}
2973 }
2974 
2975 static int fl_dump_key_flags(struct sk_buff *skb, u32 flags_key, u32 flags_mask)
2976 {
2977 	u32 key, mask;
2978 	__be32 _key, _mask;
2979 	int err;
2980 
2981 	if (!memchr_inv(&flags_mask, 0, sizeof(flags_mask)))
2982 		return 0;
2983 
2984 	key = 0;
2985 	mask = 0;
2986 
2987 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
2988 			TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
2989 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
2990 			TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
2991 			FLOW_DIS_FIRST_FRAG);
2992 
2993 	_key = cpu_to_be32(key);
2994 	_mask = cpu_to_be32(mask);
2995 
2996 	err = nla_put(skb, TCA_FLOWER_KEY_FLAGS, 4, &_key);
2997 	if (err)
2998 		return err;
2999 
3000 	return nla_put(skb, TCA_FLOWER_KEY_FLAGS_MASK, 4, &_mask);
3001 }
3002 
3003 static int fl_dump_key_geneve_opt(struct sk_buff *skb,
3004 				  struct flow_dissector_key_enc_opts *enc_opts)
3005 {
3006 	struct geneve_opt *opt;
3007 	struct nlattr *nest;
3008 	int opt_off = 0;
3009 
3010 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_GENEVE);
3011 	if (!nest)
3012 		goto nla_put_failure;
3013 
3014 	while (enc_opts->len > opt_off) {
3015 		opt = (struct geneve_opt *)&enc_opts->data[opt_off];
3016 
3017 		if (nla_put_be16(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS,
3018 				 opt->opt_class))
3019 			goto nla_put_failure;
3020 		if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE,
3021 			       opt->type))
3022 			goto nla_put_failure;
3023 		if (nla_put(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA,
3024 			    opt->length * 4, opt->opt_data))
3025 			goto nla_put_failure;
3026 
3027 		opt_off += sizeof(struct geneve_opt) + opt->length * 4;
3028 	}
3029 	nla_nest_end(skb, nest);
3030 	return 0;
3031 
3032 nla_put_failure:
3033 	nla_nest_cancel(skb, nest);
3034 	return -EMSGSIZE;
3035 }
3036 
3037 static int fl_dump_key_vxlan_opt(struct sk_buff *skb,
3038 				 struct flow_dissector_key_enc_opts *enc_opts)
3039 {
3040 	struct vxlan_metadata *md;
3041 	struct nlattr *nest;
3042 
3043 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_VXLAN);
3044 	if (!nest)
3045 		goto nla_put_failure;
3046 
3047 	md = (struct vxlan_metadata *)&enc_opts->data[0];
3048 	if (nla_put_u32(skb, TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP, md->gbp))
3049 		goto nla_put_failure;
3050 
3051 	nla_nest_end(skb, nest);
3052 	return 0;
3053 
3054 nla_put_failure:
3055 	nla_nest_cancel(skb, nest);
3056 	return -EMSGSIZE;
3057 }
3058 
3059 static int fl_dump_key_erspan_opt(struct sk_buff *skb,
3060 				  struct flow_dissector_key_enc_opts *enc_opts)
3061 {
3062 	struct erspan_metadata *md;
3063 	struct nlattr *nest;
3064 
3065 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_ERSPAN);
3066 	if (!nest)
3067 		goto nla_put_failure;
3068 
3069 	md = (struct erspan_metadata *)&enc_opts->data[0];
3070 	if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER, md->version))
3071 		goto nla_put_failure;
3072 
3073 	if (md->version == 1 &&
3074 	    nla_put_be32(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX, md->u.index))
3075 		goto nla_put_failure;
3076 
3077 	if (md->version == 2 &&
3078 	    (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR,
3079 			md->u.md2.dir) ||
3080 	     nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID,
3081 			get_hwid(&md->u.md2))))
3082 		goto nla_put_failure;
3083 
3084 	nla_nest_end(skb, nest);
3085 	return 0;
3086 
3087 nla_put_failure:
3088 	nla_nest_cancel(skb, nest);
3089 	return -EMSGSIZE;
3090 }
3091 
3092 static int fl_dump_key_gtp_opt(struct sk_buff *skb,
3093 			       struct flow_dissector_key_enc_opts *enc_opts)
3094 
3095 {
3096 	struct gtp_pdu_session_info *session_info;
3097 	struct nlattr *nest;
3098 
3099 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_GTP);
3100 	if (!nest)
3101 		goto nla_put_failure;
3102 
3103 	session_info = (struct gtp_pdu_session_info *)&enc_opts->data[0];
3104 
3105 	if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE,
3106 		       session_info->pdu_type))
3107 		goto nla_put_failure;
3108 
3109 	if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GTP_QFI, session_info->qfi))
3110 		goto nla_put_failure;
3111 
3112 	nla_nest_end(skb, nest);
3113 	return 0;
3114 
3115 nla_put_failure:
3116 	nla_nest_cancel(skb, nest);
3117 	return -EMSGSIZE;
3118 }
3119 
3120 static int fl_dump_key_ct(struct sk_buff *skb,
3121 			  struct flow_dissector_key_ct *key,
3122 			  struct flow_dissector_key_ct *mask)
3123 {
3124 	if (IS_ENABLED(CONFIG_NF_CONNTRACK) &&
3125 	    fl_dump_key_val(skb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE,
3126 			    &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK,
3127 			    sizeof(key->ct_state)))
3128 		goto nla_put_failure;
3129 
3130 	if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
3131 	    fl_dump_key_val(skb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE,
3132 			    &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK,
3133 			    sizeof(key->ct_zone)))
3134 		goto nla_put_failure;
3135 
3136 	if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
3137 	    fl_dump_key_val(skb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK,
3138 			    &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK,
3139 			    sizeof(key->ct_mark)))
3140 		goto nla_put_failure;
3141 
3142 	if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
3143 	    fl_dump_key_val(skb, &key->ct_labels, TCA_FLOWER_KEY_CT_LABELS,
3144 			    &mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK,
3145 			    sizeof(key->ct_labels)))
3146 		goto nla_put_failure;
3147 
3148 	return 0;
3149 
3150 nla_put_failure:
3151 	return -EMSGSIZE;
3152 }
3153 
3154 static int fl_dump_key_cfm(struct sk_buff *skb,
3155 			   struct flow_dissector_key_cfm *key,
3156 			   struct flow_dissector_key_cfm *mask)
3157 {
3158 	struct nlattr *opts;
3159 	int err;
3160 	u8 mdl;
3161 
3162 	if (!memchr_inv(mask, 0, sizeof(*mask)))
3163 		return 0;
3164 
3165 	opts = nla_nest_start(skb, TCA_FLOWER_KEY_CFM);
3166 	if (!opts)
3167 		return -EMSGSIZE;
3168 
3169 	if (FIELD_GET(FLOW_DIS_CFM_MDL_MASK, mask->mdl_ver)) {
3170 		mdl = FIELD_GET(FLOW_DIS_CFM_MDL_MASK, key->mdl_ver);
3171 		err = nla_put_u8(skb, TCA_FLOWER_KEY_CFM_MD_LEVEL, mdl);
3172 		if (err)
3173 			goto err_cfm_opts;
3174 	}
3175 
3176 	if (mask->opcode) {
3177 		err = nla_put_u8(skb, TCA_FLOWER_KEY_CFM_OPCODE, key->opcode);
3178 		if (err)
3179 			goto err_cfm_opts;
3180 	}
3181 
3182 	nla_nest_end(skb, opts);
3183 
3184 	return 0;
3185 
3186 err_cfm_opts:
3187 	nla_nest_cancel(skb, opts);
3188 	return err;
3189 }
3190 
3191 static int fl_dump_key_options(struct sk_buff *skb, int enc_opt_type,
3192 			       struct flow_dissector_key_enc_opts *enc_opts)
3193 {
3194 	struct nlattr *nest;
3195 	int err;
3196 
3197 	if (!enc_opts->len)
3198 		return 0;
3199 
3200 	nest = nla_nest_start_noflag(skb, enc_opt_type);
3201 	if (!nest)
3202 		goto nla_put_failure;
3203 
3204 	switch (enc_opts->dst_opt_type) {
3205 	case TUNNEL_GENEVE_OPT:
3206 		err = fl_dump_key_geneve_opt(skb, enc_opts);
3207 		if (err)
3208 			goto nla_put_failure;
3209 		break;
3210 	case TUNNEL_VXLAN_OPT:
3211 		err = fl_dump_key_vxlan_opt(skb, enc_opts);
3212 		if (err)
3213 			goto nla_put_failure;
3214 		break;
3215 	case TUNNEL_ERSPAN_OPT:
3216 		err = fl_dump_key_erspan_opt(skb, enc_opts);
3217 		if (err)
3218 			goto nla_put_failure;
3219 		break;
3220 	case TUNNEL_GTP_OPT:
3221 		err = fl_dump_key_gtp_opt(skb, enc_opts);
3222 		if (err)
3223 			goto nla_put_failure;
3224 		break;
3225 	default:
3226 		goto nla_put_failure;
3227 	}
3228 	nla_nest_end(skb, nest);
3229 	return 0;
3230 
3231 nla_put_failure:
3232 	nla_nest_cancel(skb, nest);
3233 	return -EMSGSIZE;
3234 }
3235 
3236 static int fl_dump_key_enc_opt(struct sk_buff *skb,
3237 			       struct flow_dissector_key_enc_opts *key_opts,
3238 			       struct flow_dissector_key_enc_opts *msk_opts)
3239 {
3240 	int err;
3241 
3242 	err = fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS, key_opts);
3243 	if (err)
3244 		return err;
3245 
3246 	return fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS_MASK, msk_opts);
3247 }
3248 
3249 static int fl_dump_key(struct sk_buff *skb, struct net *net,
3250 		       struct fl_flow_key *key, struct fl_flow_key *mask)
3251 {
3252 	if (mask->meta.ingress_ifindex) {
3253 		struct net_device *dev;
3254 
3255 		dev = __dev_get_by_index(net, key->meta.ingress_ifindex);
3256 		if (dev && nla_put_string(skb, TCA_FLOWER_INDEV, dev->name))
3257 			goto nla_put_failure;
3258 	}
3259 
3260 	if (fl_dump_key_val(skb, &key->meta.l2_miss,
3261 			    TCA_FLOWER_L2_MISS, &mask->meta.l2_miss,
3262 			    TCA_FLOWER_UNSPEC, sizeof(key->meta.l2_miss)))
3263 		goto nla_put_failure;
3264 
3265 	if (fl_dump_key_val(skb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
3266 			    mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
3267 			    sizeof(key->eth.dst)) ||
3268 	    fl_dump_key_val(skb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
3269 			    mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
3270 			    sizeof(key->eth.src)) ||
3271 	    fl_dump_key_val(skb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE,
3272 			    &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
3273 			    sizeof(key->basic.n_proto)))
3274 		goto nla_put_failure;
3275 
3276 	if (mask->num_of_vlans.num_of_vlans) {
3277 		if (nla_put_u8(skb, TCA_FLOWER_KEY_NUM_OF_VLANS, key->num_of_vlans.num_of_vlans))
3278 			goto nla_put_failure;
3279 	}
3280 
3281 	if (fl_dump_key_mpls(skb, &key->mpls, &mask->mpls))
3282 		goto nla_put_failure;
3283 
3284 	if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_VLAN_ID,
3285 			     TCA_FLOWER_KEY_VLAN_PRIO, &key->vlan, &mask->vlan))
3286 		goto nla_put_failure;
3287 
3288 	if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_CVLAN_ID,
3289 			     TCA_FLOWER_KEY_CVLAN_PRIO,
3290 			     &key->cvlan, &mask->cvlan) ||
3291 	    (mask->cvlan.vlan_tpid &&
3292 	     nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
3293 			  key->cvlan.vlan_tpid)))
3294 		goto nla_put_failure;
3295 
3296 	if (mask->basic.n_proto) {
3297 		if (mask->cvlan.vlan_eth_type) {
3298 			if (nla_put_be16(skb, TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
3299 					 key->basic.n_proto))
3300 				goto nla_put_failure;
3301 		} else if (mask->vlan.vlan_eth_type) {
3302 			if (nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
3303 					 key->vlan.vlan_eth_type))
3304 				goto nla_put_failure;
3305 		}
3306 	}
3307 
3308 	if ((key->basic.n_proto == htons(ETH_P_IP) ||
3309 	     key->basic.n_proto == htons(ETH_P_IPV6)) &&
3310 	    (fl_dump_key_val(skb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
3311 			    &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
3312 			    sizeof(key->basic.ip_proto)) ||
3313 	    fl_dump_key_ip(skb, false, &key->ip, &mask->ip)))
3314 		goto nla_put_failure;
3315 
3316 	if (mask->pppoe.session_id) {
3317 		if (nla_put_be16(skb, TCA_FLOWER_KEY_PPPOE_SID,
3318 				 key->pppoe.session_id))
3319 			goto nla_put_failure;
3320 	}
3321 	if (mask->basic.n_proto && mask->pppoe.ppp_proto) {
3322 		if (nla_put_be16(skb, TCA_FLOWER_KEY_PPP_PROTO,
3323 				 key->pppoe.ppp_proto))
3324 			goto nla_put_failure;
3325 	}
3326 
3327 	if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
3328 	    (fl_dump_key_val(skb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
3329 			     &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
3330 			     sizeof(key->ipv4.src)) ||
3331 	     fl_dump_key_val(skb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
3332 			     &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
3333 			     sizeof(key->ipv4.dst))))
3334 		goto nla_put_failure;
3335 	else if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
3336 		 (fl_dump_key_val(skb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
3337 				  &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
3338 				  sizeof(key->ipv6.src)) ||
3339 		  fl_dump_key_val(skb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
3340 				  &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
3341 				  sizeof(key->ipv6.dst))))
3342 		goto nla_put_failure;
3343 
3344 	if (key->basic.ip_proto == IPPROTO_TCP &&
3345 	    (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
3346 			     &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
3347 			     sizeof(key->tp.src)) ||
3348 	     fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
3349 			     &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
3350 			     sizeof(key->tp.dst)) ||
3351 	     fl_dump_key_val(skb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
3352 			     &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
3353 			     sizeof(key->tcp.flags))))
3354 		goto nla_put_failure;
3355 	else if (key->basic.ip_proto == IPPROTO_UDP &&
3356 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
3357 				  &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
3358 				  sizeof(key->tp.src)) ||
3359 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
3360 				  &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
3361 				  sizeof(key->tp.dst))))
3362 		goto nla_put_failure;
3363 	else if (key->basic.ip_proto == IPPROTO_SCTP &&
3364 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
3365 				  &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
3366 				  sizeof(key->tp.src)) ||
3367 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
3368 				  &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
3369 				  sizeof(key->tp.dst))))
3370 		goto nla_put_failure;
3371 	else if (key->basic.n_proto == htons(ETH_P_IP) &&
3372 		 key->basic.ip_proto == IPPROTO_ICMP &&
3373 		 (fl_dump_key_val(skb, &key->icmp.type,
3374 				  TCA_FLOWER_KEY_ICMPV4_TYPE, &mask->icmp.type,
3375 				  TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
3376 				  sizeof(key->icmp.type)) ||
3377 		  fl_dump_key_val(skb, &key->icmp.code,
3378 				  TCA_FLOWER_KEY_ICMPV4_CODE, &mask->icmp.code,
3379 				  TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
3380 				  sizeof(key->icmp.code))))
3381 		goto nla_put_failure;
3382 	else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
3383 		 key->basic.ip_proto == IPPROTO_ICMPV6 &&
3384 		 (fl_dump_key_val(skb, &key->icmp.type,
3385 				  TCA_FLOWER_KEY_ICMPV6_TYPE, &mask->icmp.type,
3386 				  TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
3387 				  sizeof(key->icmp.type)) ||
3388 		  fl_dump_key_val(skb, &key->icmp.code,
3389 				  TCA_FLOWER_KEY_ICMPV6_CODE, &mask->icmp.code,
3390 				  TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
3391 				  sizeof(key->icmp.code))))
3392 		goto nla_put_failure;
3393 	else if ((key->basic.n_proto == htons(ETH_P_ARP) ||
3394 		  key->basic.n_proto == htons(ETH_P_RARP)) &&
3395 		 (fl_dump_key_val(skb, &key->arp.sip,
3396 				  TCA_FLOWER_KEY_ARP_SIP, &mask->arp.sip,
3397 				  TCA_FLOWER_KEY_ARP_SIP_MASK,
3398 				  sizeof(key->arp.sip)) ||
3399 		  fl_dump_key_val(skb, &key->arp.tip,
3400 				  TCA_FLOWER_KEY_ARP_TIP, &mask->arp.tip,
3401 				  TCA_FLOWER_KEY_ARP_TIP_MASK,
3402 				  sizeof(key->arp.tip)) ||
3403 		  fl_dump_key_val(skb, &key->arp.op,
3404 				  TCA_FLOWER_KEY_ARP_OP, &mask->arp.op,
3405 				  TCA_FLOWER_KEY_ARP_OP_MASK,
3406 				  sizeof(key->arp.op)) ||
3407 		  fl_dump_key_val(skb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
3408 				  mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
3409 				  sizeof(key->arp.sha)) ||
3410 		  fl_dump_key_val(skb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
3411 				  mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
3412 				  sizeof(key->arp.tha))))
3413 		goto nla_put_failure;
3414 	else if (key->basic.ip_proto == IPPROTO_L2TP &&
3415 		 fl_dump_key_val(skb, &key->l2tpv3.session_id,
3416 				 TCA_FLOWER_KEY_L2TPV3_SID,
3417 				 &mask->l2tpv3.session_id,
3418 				 TCA_FLOWER_UNSPEC,
3419 				 sizeof(key->l2tpv3.session_id)))
3420 		goto nla_put_failure;
3421 
3422 	if (key->ipsec.spi &&
3423 	    fl_dump_key_val(skb, &key->ipsec.spi, TCA_FLOWER_KEY_SPI,
3424 			    &mask->ipsec.spi, TCA_FLOWER_KEY_SPI_MASK,
3425 			    sizeof(key->ipsec.spi)))
3426 		goto nla_put_failure;
3427 
3428 	if ((key->basic.ip_proto == IPPROTO_TCP ||
3429 	     key->basic.ip_proto == IPPROTO_UDP ||
3430 	     key->basic.ip_proto == IPPROTO_SCTP) &&
3431 	     fl_dump_key_port_range(skb, key, mask))
3432 		goto nla_put_failure;
3433 
3434 	if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
3435 	    (fl_dump_key_val(skb, &key->enc_ipv4.src,
3436 			    TCA_FLOWER_KEY_ENC_IPV4_SRC, &mask->enc_ipv4.src,
3437 			    TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
3438 			    sizeof(key->enc_ipv4.src)) ||
3439 	     fl_dump_key_val(skb, &key->enc_ipv4.dst,
3440 			     TCA_FLOWER_KEY_ENC_IPV4_DST, &mask->enc_ipv4.dst,
3441 			     TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
3442 			     sizeof(key->enc_ipv4.dst))))
3443 		goto nla_put_failure;
3444 	else if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
3445 		 (fl_dump_key_val(skb, &key->enc_ipv6.src,
3446 			    TCA_FLOWER_KEY_ENC_IPV6_SRC, &mask->enc_ipv6.src,
3447 			    TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
3448 			    sizeof(key->enc_ipv6.src)) ||
3449 		 fl_dump_key_val(skb, &key->enc_ipv6.dst,
3450 				 TCA_FLOWER_KEY_ENC_IPV6_DST,
3451 				 &mask->enc_ipv6.dst,
3452 				 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
3453 			    sizeof(key->enc_ipv6.dst))))
3454 		goto nla_put_failure;
3455 
3456 	if (fl_dump_key_val(skb, &key->enc_key_id, TCA_FLOWER_KEY_ENC_KEY_ID,
3457 			    &mask->enc_key_id, TCA_FLOWER_UNSPEC,
3458 			    sizeof(key->enc_key_id)) ||
3459 	    fl_dump_key_val(skb, &key->enc_tp.src,
3460 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
3461 			    &mask->enc_tp.src,
3462 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
3463 			    sizeof(key->enc_tp.src)) ||
3464 	    fl_dump_key_val(skb, &key->enc_tp.dst,
3465 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
3466 			    &mask->enc_tp.dst,
3467 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
3468 			    sizeof(key->enc_tp.dst)) ||
3469 	    fl_dump_key_ip(skb, true, &key->enc_ip, &mask->enc_ip) ||
3470 	    fl_dump_key_enc_opt(skb, &key->enc_opts, &mask->enc_opts))
3471 		goto nla_put_failure;
3472 
3473 	if (fl_dump_key_ct(skb, &key->ct, &mask->ct))
3474 		goto nla_put_failure;
3475 
3476 	if (fl_dump_key_flags(skb, key->control.flags, mask->control.flags))
3477 		goto nla_put_failure;
3478 
3479 	if (fl_dump_key_val(skb, &key->hash.hash, TCA_FLOWER_KEY_HASH,
3480 			     &mask->hash.hash, TCA_FLOWER_KEY_HASH_MASK,
3481 			     sizeof(key->hash.hash)))
3482 		goto nla_put_failure;
3483 
3484 	if (fl_dump_key_cfm(skb, &key->cfm, &mask->cfm))
3485 		goto nla_put_failure;
3486 
3487 	return 0;
3488 
3489 nla_put_failure:
3490 	return -EMSGSIZE;
3491 }
3492 
3493 static int fl_dump(struct net *net, struct tcf_proto *tp, void *fh,
3494 		   struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
3495 {
3496 	struct cls_fl_filter *f = fh;
3497 	struct nlattr *nest;
3498 	struct fl_flow_key *key, *mask;
3499 	bool skip_hw;
3500 
3501 	if (!f)
3502 		return skb->len;
3503 
3504 	t->tcm_handle = f->handle;
3505 
3506 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
3507 	if (!nest)
3508 		goto nla_put_failure;
3509 
3510 	spin_lock(&tp->lock);
3511 
3512 	if (f->res.classid &&
3513 	    nla_put_u32(skb, TCA_FLOWER_CLASSID, f->res.classid))
3514 		goto nla_put_failure_locked;
3515 
3516 	key = &f->key;
3517 	mask = &f->mask->key;
3518 	skip_hw = tc_skip_hw(f->flags);
3519 
3520 	if (fl_dump_key(skb, net, key, mask))
3521 		goto nla_put_failure_locked;
3522 
3523 	if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
3524 		goto nla_put_failure_locked;
3525 
3526 	spin_unlock(&tp->lock);
3527 
3528 	if (!skip_hw)
3529 		fl_hw_update_stats(tp, f, rtnl_held);
3530 
3531 	if (nla_put_u32(skb, TCA_FLOWER_IN_HW_COUNT, f->in_hw_count))
3532 		goto nla_put_failure;
3533 
3534 	if (tcf_exts_dump(skb, &f->exts))
3535 		goto nla_put_failure;
3536 
3537 	nla_nest_end(skb, nest);
3538 
3539 	if (tcf_exts_dump_stats(skb, &f->exts) < 0)
3540 		goto nla_put_failure;
3541 
3542 	return skb->len;
3543 
3544 nla_put_failure_locked:
3545 	spin_unlock(&tp->lock);
3546 nla_put_failure:
3547 	nla_nest_cancel(skb, nest);
3548 	return -1;
3549 }
3550 
3551 static int fl_terse_dump(struct net *net, struct tcf_proto *tp, void *fh,
3552 			 struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
3553 {
3554 	struct cls_fl_filter *f = fh;
3555 	struct nlattr *nest;
3556 	bool skip_hw;
3557 
3558 	if (!f)
3559 		return skb->len;
3560 
3561 	t->tcm_handle = f->handle;
3562 
3563 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
3564 	if (!nest)
3565 		goto nla_put_failure;
3566 
3567 	spin_lock(&tp->lock);
3568 
3569 	skip_hw = tc_skip_hw(f->flags);
3570 
3571 	if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
3572 		goto nla_put_failure_locked;
3573 
3574 	spin_unlock(&tp->lock);
3575 
3576 	if (!skip_hw)
3577 		fl_hw_update_stats(tp, f, rtnl_held);
3578 
3579 	if (tcf_exts_terse_dump(skb, &f->exts))
3580 		goto nla_put_failure;
3581 
3582 	nla_nest_end(skb, nest);
3583 
3584 	return skb->len;
3585 
3586 nla_put_failure_locked:
3587 	spin_unlock(&tp->lock);
3588 nla_put_failure:
3589 	nla_nest_cancel(skb, nest);
3590 	return -1;
3591 }
3592 
3593 static int fl_tmplt_dump(struct sk_buff *skb, struct net *net, void *tmplt_priv)
3594 {
3595 	struct fl_flow_tmplt *tmplt = tmplt_priv;
3596 	struct fl_flow_key *key, *mask;
3597 	struct nlattr *nest;
3598 
3599 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
3600 	if (!nest)
3601 		goto nla_put_failure;
3602 
3603 	key = &tmplt->dummy_key;
3604 	mask = &tmplt->mask;
3605 
3606 	if (fl_dump_key(skb, net, key, mask))
3607 		goto nla_put_failure;
3608 
3609 	nla_nest_end(skb, nest);
3610 
3611 	return skb->len;
3612 
3613 nla_put_failure:
3614 	nla_nest_cancel(skb, nest);
3615 	return -EMSGSIZE;
3616 }
3617 
3618 static void fl_bind_class(void *fh, u32 classid, unsigned long cl, void *q,
3619 			  unsigned long base)
3620 {
3621 	struct cls_fl_filter *f = fh;
3622 
3623 	tc_cls_bind_class(classid, cl, q, &f->res, base);
3624 }
3625 
3626 static bool fl_delete_empty(struct tcf_proto *tp)
3627 {
3628 	struct cls_fl_head *head = fl_head_dereference(tp);
3629 
3630 	spin_lock(&tp->lock);
3631 	tp->deleting = idr_is_empty(&head->handle_idr);
3632 	spin_unlock(&tp->lock);
3633 
3634 	return tp->deleting;
3635 }
3636 
3637 static struct tcf_proto_ops cls_fl_ops __read_mostly = {
3638 	.kind		= "flower",
3639 	.classify	= fl_classify,
3640 	.init		= fl_init,
3641 	.destroy	= fl_destroy,
3642 	.get		= fl_get,
3643 	.put		= fl_put,
3644 	.change		= fl_change,
3645 	.delete		= fl_delete,
3646 	.delete_empty	= fl_delete_empty,
3647 	.walk		= fl_walk,
3648 	.reoffload	= fl_reoffload,
3649 	.hw_add		= fl_hw_add,
3650 	.hw_del		= fl_hw_del,
3651 	.dump		= fl_dump,
3652 	.terse_dump	= fl_terse_dump,
3653 	.bind_class	= fl_bind_class,
3654 	.tmplt_create	= fl_tmplt_create,
3655 	.tmplt_destroy	= fl_tmplt_destroy,
3656 	.tmplt_reoffload = fl_tmplt_reoffload,
3657 	.tmplt_dump	= fl_tmplt_dump,
3658 	.get_exts	= fl_get_exts,
3659 	.owner		= THIS_MODULE,
3660 	.flags		= TCF_PROTO_OPS_DOIT_UNLOCKED,
3661 };
3662 MODULE_ALIAS_NET_CLS("flower");
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