xref: /linux/net/sched/cls_flower.c (revision b48543c451c30387b53ee6e202dda8d5303f6268)
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/pfcp.h>
32 #include <net/tc_wrapper.h>
33 
34 #include <net/dst.h>
35 #include <net/dst_metadata.h>
36 
37 #include <uapi/linux/netfilter/nf_conntrack_common.h>
38 
39 #define TCA_FLOWER_KEY_CT_FLAGS_MAX \
40 		((__TCA_FLOWER_KEY_CT_FLAGS_MAX - 1) << 1)
41 #define TCA_FLOWER_KEY_CT_FLAGS_MASK \
42 		(TCA_FLOWER_KEY_CT_FLAGS_MAX - 1)
43 
44 #define TCA_FLOWER_KEY_FLAGS_POLICY_MASK \
45 		(TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT | \
46 		TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST)
47 
48 #define TCA_FLOWER_KEY_ENC_FLAGS_POLICY_MASK \
49 		(TCA_FLOWER_KEY_FLAGS_TUNNEL_CSUM | \
50 		TCA_FLOWER_KEY_FLAGS_TUNNEL_DONT_FRAGMENT | \
51 		TCA_FLOWER_KEY_FLAGS_TUNNEL_OAM | \
52 		TCA_FLOWER_KEY_FLAGS_TUNNEL_CRIT_OPT)
53 
54 struct fl_flow_key {
55 	struct flow_dissector_key_meta meta;
56 	struct flow_dissector_key_control control;
57 	struct flow_dissector_key_control enc_control;
58 	struct flow_dissector_key_basic basic;
59 	struct flow_dissector_key_eth_addrs eth;
60 	struct flow_dissector_key_vlan vlan;
61 	struct flow_dissector_key_vlan cvlan;
62 	union {
63 		struct flow_dissector_key_ipv4_addrs ipv4;
64 		struct flow_dissector_key_ipv6_addrs ipv6;
65 	};
66 	struct flow_dissector_key_ports tp;
67 	struct flow_dissector_key_icmp icmp;
68 	struct flow_dissector_key_arp arp;
69 	struct flow_dissector_key_keyid enc_key_id;
70 	union {
71 		struct flow_dissector_key_ipv4_addrs enc_ipv4;
72 		struct flow_dissector_key_ipv6_addrs enc_ipv6;
73 	};
74 	struct flow_dissector_key_ports enc_tp;
75 	struct flow_dissector_key_mpls mpls;
76 	struct flow_dissector_key_tcp tcp;
77 	struct flow_dissector_key_ip ip;
78 	struct flow_dissector_key_ip enc_ip;
79 	struct flow_dissector_key_enc_opts enc_opts;
80 	struct flow_dissector_key_ports_range tp_range;
81 	struct flow_dissector_key_ct ct;
82 	struct flow_dissector_key_hash hash;
83 	struct flow_dissector_key_num_of_vlans num_of_vlans;
84 	struct flow_dissector_key_pppoe pppoe;
85 	struct flow_dissector_key_l2tpv3 l2tpv3;
86 	struct flow_dissector_key_ipsec ipsec;
87 	struct flow_dissector_key_cfm cfm;
88 } __aligned(BITS_PER_LONG / 8); /* Ensure that we can do comparisons as longs. */
89 
90 struct fl_flow_mask_range {
91 	unsigned short int start;
92 	unsigned short int end;
93 };
94 
95 struct fl_flow_mask {
96 	struct fl_flow_key key;
97 	struct fl_flow_mask_range range;
98 	u32 flags;
99 	struct rhash_head ht_node;
100 	struct rhashtable ht;
101 	struct rhashtable_params filter_ht_params;
102 	struct flow_dissector dissector;
103 	struct list_head filters;
104 	struct rcu_work rwork;
105 	struct list_head list;
106 	refcount_t refcnt;
107 };
108 
109 struct fl_flow_tmplt {
110 	struct fl_flow_key dummy_key;
111 	struct fl_flow_key mask;
112 	struct flow_dissector dissector;
113 	struct tcf_chain *chain;
114 };
115 
116 struct cls_fl_head {
117 	struct rhashtable ht;
118 	spinlock_t masks_lock; /* Protect masks list */
119 	struct list_head masks;
120 	struct list_head hw_filters;
121 	struct rcu_work rwork;
122 	struct idr handle_idr;
123 };
124 
125 struct cls_fl_filter {
126 	struct fl_flow_mask *mask;
127 	struct rhash_head ht_node;
128 	struct fl_flow_key mkey;
129 	struct tcf_exts exts;
130 	struct tcf_result res;
131 	struct fl_flow_key key;
132 	struct list_head list;
133 	struct list_head hw_list;
134 	u32 handle;
135 	u32 flags;
136 	u32 in_hw_count;
137 	u8 needs_tc_skb_ext:1;
138 	struct rcu_work rwork;
139 	struct net_device *hw_dev;
140 	/* Flower classifier is unlocked, which means that its reference counter
141 	 * can be changed concurrently without any kind of external
142 	 * synchronization. Use atomic reference counter to be concurrency-safe.
143 	 */
144 	refcount_t refcnt;
145 	bool deleted;
146 };
147 
148 static const struct rhashtable_params mask_ht_params = {
149 	.key_offset = offsetof(struct fl_flow_mask, key),
150 	.key_len = sizeof(struct fl_flow_key),
151 	.head_offset = offsetof(struct fl_flow_mask, ht_node),
152 	.automatic_shrinking = true,
153 };
154 
155 static unsigned short int fl_mask_range(const struct fl_flow_mask *mask)
156 {
157 	return mask->range.end - mask->range.start;
158 }
159 
160 static void fl_mask_update_range(struct fl_flow_mask *mask)
161 {
162 	const u8 *bytes = (const u8 *) &mask->key;
163 	size_t size = sizeof(mask->key);
164 	size_t i, first = 0, last;
165 
166 	for (i = 0; i < size; i++) {
167 		if (bytes[i]) {
168 			first = i;
169 			break;
170 		}
171 	}
172 	last = first;
173 	for (i = size - 1; i != first; i--) {
174 		if (bytes[i]) {
175 			last = i;
176 			break;
177 		}
178 	}
179 	mask->range.start = rounddown(first, sizeof(long));
180 	mask->range.end = roundup(last + 1, sizeof(long));
181 }
182 
183 static void *fl_key_get_start(struct fl_flow_key *key,
184 			      const struct fl_flow_mask *mask)
185 {
186 	return (u8 *) key + mask->range.start;
187 }
188 
189 static void fl_set_masked_key(struct fl_flow_key *mkey, struct fl_flow_key *key,
190 			      struct fl_flow_mask *mask)
191 {
192 	const long *lkey = fl_key_get_start(key, mask);
193 	const long *lmask = fl_key_get_start(&mask->key, mask);
194 	long *lmkey = fl_key_get_start(mkey, mask);
195 	int i;
196 
197 	for (i = 0; i < fl_mask_range(mask); i += sizeof(long))
198 		*lmkey++ = *lkey++ & *lmask++;
199 }
200 
201 static bool fl_mask_fits_tmplt(struct fl_flow_tmplt *tmplt,
202 			       struct fl_flow_mask *mask)
203 {
204 	const long *lmask = fl_key_get_start(&mask->key, mask);
205 	const long *ltmplt;
206 	int i;
207 
208 	if (!tmplt)
209 		return true;
210 	ltmplt = fl_key_get_start(&tmplt->mask, mask);
211 	for (i = 0; i < fl_mask_range(mask); i += sizeof(long)) {
212 		if (~*ltmplt++ & *lmask++)
213 			return false;
214 	}
215 	return true;
216 }
217 
218 static void fl_clear_masked_range(struct fl_flow_key *key,
219 				  struct fl_flow_mask *mask)
220 {
221 	memset(fl_key_get_start(key, mask), 0, fl_mask_range(mask));
222 }
223 
224 static bool fl_range_port_dst_cmp(struct cls_fl_filter *filter,
225 				  struct fl_flow_key *key,
226 				  struct fl_flow_key *mkey)
227 {
228 	u16 min_mask, max_mask, min_val, max_val;
229 
230 	min_mask = ntohs(filter->mask->key.tp_range.tp_min.dst);
231 	max_mask = ntohs(filter->mask->key.tp_range.tp_max.dst);
232 	min_val = ntohs(filter->key.tp_range.tp_min.dst);
233 	max_val = ntohs(filter->key.tp_range.tp_max.dst);
234 
235 	if (min_mask && max_mask) {
236 		if (ntohs(key->tp_range.tp.dst) < min_val ||
237 		    ntohs(key->tp_range.tp.dst) > max_val)
238 			return false;
239 
240 		/* skb does not have min and max values */
241 		mkey->tp_range.tp_min.dst = filter->mkey.tp_range.tp_min.dst;
242 		mkey->tp_range.tp_max.dst = filter->mkey.tp_range.tp_max.dst;
243 	}
244 	return true;
245 }
246 
247 static bool fl_range_port_src_cmp(struct cls_fl_filter *filter,
248 				  struct fl_flow_key *key,
249 				  struct fl_flow_key *mkey)
250 {
251 	u16 min_mask, max_mask, min_val, max_val;
252 
253 	min_mask = ntohs(filter->mask->key.tp_range.tp_min.src);
254 	max_mask = ntohs(filter->mask->key.tp_range.tp_max.src);
255 	min_val = ntohs(filter->key.tp_range.tp_min.src);
256 	max_val = ntohs(filter->key.tp_range.tp_max.src);
257 
258 	if (min_mask && max_mask) {
259 		if (ntohs(key->tp_range.tp.src) < min_val ||
260 		    ntohs(key->tp_range.tp.src) > max_val)
261 			return false;
262 
263 		/* skb does not have min and max values */
264 		mkey->tp_range.tp_min.src = filter->mkey.tp_range.tp_min.src;
265 		mkey->tp_range.tp_max.src = filter->mkey.tp_range.tp_max.src;
266 	}
267 	return true;
268 }
269 
270 static struct cls_fl_filter *__fl_lookup(struct fl_flow_mask *mask,
271 					 struct fl_flow_key *mkey)
272 {
273 	return rhashtable_lookup_fast(&mask->ht, fl_key_get_start(mkey, mask),
274 				      mask->filter_ht_params);
275 }
276 
277 static struct cls_fl_filter *fl_lookup_range(struct fl_flow_mask *mask,
278 					     struct fl_flow_key *mkey,
279 					     struct fl_flow_key *key)
280 {
281 	struct cls_fl_filter *filter, *f;
282 
283 	list_for_each_entry_rcu(filter, &mask->filters, list) {
284 		if (!fl_range_port_dst_cmp(filter, key, mkey))
285 			continue;
286 
287 		if (!fl_range_port_src_cmp(filter, key, mkey))
288 			continue;
289 
290 		f = __fl_lookup(mask, mkey);
291 		if (f)
292 			return f;
293 	}
294 	return NULL;
295 }
296 
297 static noinline_for_stack
298 struct cls_fl_filter *fl_mask_lookup(struct fl_flow_mask *mask, struct fl_flow_key *key)
299 {
300 	struct fl_flow_key mkey;
301 
302 	fl_set_masked_key(&mkey, key, mask);
303 	if ((mask->flags & TCA_FLOWER_MASK_FLAGS_RANGE))
304 		return fl_lookup_range(mask, &mkey, key);
305 
306 	return __fl_lookup(mask, &mkey);
307 }
308 
309 static u16 fl_ct_info_to_flower_map[] = {
310 	[IP_CT_ESTABLISHED] =		TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
311 					TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED,
312 	[IP_CT_RELATED] =		TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
313 					TCA_FLOWER_KEY_CT_FLAGS_RELATED,
314 	[IP_CT_ESTABLISHED_REPLY] =	TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
315 					TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED |
316 					TCA_FLOWER_KEY_CT_FLAGS_REPLY,
317 	[IP_CT_RELATED_REPLY] =		TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
318 					TCA_FLOWER_KEY_CT_FLAGS_RELATED |
319 					TCA_FLOWER_KEY_CT_FLAGS_REPLY,
320 	[IP_CT_NEW] =			TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
321 					TCA_FLOWER_KEY_CT_FLAGS_NEW,
322 };
323 
324 TC_INDIRECT_SCOPE int fl_classify(struct sk_buff *skb,
325 				  const struct tcf_proto *tp,
326 				  struct tcf_result *res)
327 {
328 	struct cls_fl_head *head = rcu_dereference_bh(tp->root);
329 	bool post_ct = tc_skb_cb(skb)->post_ct;
330 	u16 zone = tc_skb_cb(skb)->zone;
331 	struct fl_flow_key skb_key;
332 	struct fl_flow_mask *mask;
333 	struct cls_fl_filter *f;
334 
335 	list_for_each_entry_rcu(mask, &head->masks, list) {
336 		flow_dissector_init_keys(&skb_key.control, &skb_key.basic);
337 		fl_clear_masked_range(&skb_key, mask);
338 
339 		skb_flow_dissect_meta(skb, &mask->dissector, &skb_key);
340 		/* skb_flow_dissect() does not set n_proto in case an unknown
341 		 * protocol, so do it rather here.
342 		 */
343 		skb_key.basic.n_proto = skb_protocol(skb, false);
344 		skb_flow_dissect_tunnel_info(skb, &mask->dissector, &skb_key);
345 		skb_flow_dissect_ct(skb, &mask->dissector, &skb_key,
346 				    fl_ct_info_to_flower_map,
347 				    ARRAY_SIZE(fl_ct_info_to_flower_map),
348 				    post_ct, zone);
349 		skb_flow_dissect_hash(skb, &mask->dissector, &skb_key);
350 		skb_flow_dissect(skb, &mask->dissector, &skb_key,
351 				 FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP);
352 
353 		f = fl_mask_lookup(mask, &skb_key);
354 		if (f && !tc_skip_sw(f->flags)) {
355 			*res = f->res;
356 			return tcf_exts_exec(skb, &f->exts, res);
357 		}
358 	}
359 	return -1;
360 }
361 
362 static int fl_init(struct tcf_proto *tp)
363 {
364 	struct cls_fl_head *head;
365 
366 	head = kzalloc(sizeof(*head), GFP_KERNEL);
367 	if (!head)
368 		return -ENOBUFS;
369 
370 	spin_lock_init(&head->masks_lock);
371 	INIT_LIST_HEAD_RCU(&head->masks);
372 	INIT_LIST_HEAD(&head->hw_filters);
373 	rcu_assign_pointer(tp->root, head);
374 	idr_init(&head->handle_idr);
375 
376 	return rhashtable_init(&head->ht, &mask_ht_params);
377 }
378 
379 static void fl_mask_free(struct fl_flow_mask *mask, bool mask_init_done)
380 {
381 	/* temporary masks don't have their filters list and ht initialized */
382 	if (mask_init_done) {
383 		WARN_ON(!list_empty(&mask->filters));
384 		rhashtable_destroy(&mask->ht);
385 	}
386 	kfree(mask);
387 }
388 
389 static void fl_mask_free_work(struct work_struct *work)
390 {
391 	struct fl_flow_mask *mask = container_of(to_rcu_work(work),
392 						 struct fl_flow_mask, rwork);
393 
394 	fl_mask_free(mask, true);
395 }
396 
397 static void fl_uninit_mask_free_work(struct work_struct *work)
398 {
399 	struct fl_flow_mask *mask = container_of(to_rcu_work(work),
400 						 struct fl_flow_mask, rwork);
401 
402 	fl_mask_free(mask, false);
403 }
404 
405 static bool fl_mask_put(struct cls_fl_head *head, struct fl_flow_mask *mask)
406 {
407 	if (!refcount_dec_and_test(&mask->refcnt))
408 		return false;
409 
410 	rhashtable_remove_fast(&head->ht, &mask->ht_node, mask_ht_params);
411 
412 	spin_lock(&head->masks_lock);
413 	list_del_rcu(&mask->list);
414 	spin_unlock(&head->masks_lock);
415 
416 	tcf_queue_work(&mask->rwork, fl_mask_free_work);
417 
418 	return true;
419 }
420 
421 static struct cls_fl_head *fl_head_dereference(struct tcf_proto *tp)
422 {
423 	/* Flower classifier only changes root pointer during init and destroy.
424 	 * Users must obtain reference to tcf_proto instance before calling its
425 	 * API, so tp->root pointer is protected from concurrent call to
426 	 * fl_destroy() by reference counting.
427 	 */
428 	return rcu_dereference_raw(tp->root);
429 }
430 
431 static void __fl_destroy_filter(struct cls_fl_filter *f)
432 {
433 	if (f->needs_tc_skb_ext)
434 		tc_skb_ext_tc_disable();
435 	tcf_exts_destroy(&f->exts);
436 	tcf_exts_put_net(&f->exts);
437 	kfree(f);
438 }
439 
440 static void fl_destroy_filter_work(struct work_struct *work)
441 {
442 	struct cls_fl_filter *f = container_of(to_rcu_work(work),
443 					struct cls_fl_filter, rwork);
444 
445 	__fl_destroy_filter(f);
446 }
447 
448 static void fl_hw_destroy_filter(struct tcf_proto *tp, struct cls_fl_filter *f,
449 				 bool rtnl_held, struct netlink_ext_ack *extack)
450 {
451 	struct tcf_block *block = tp->chain->block;
452 	struct flow_cls_offload cls_flower = {};
453 
454 	tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
455 	cls_flower.command = FLOW_CLS_DESTROY;
456 	cls_flower.cookie = (unsigned long) f;
457 
458 	tc_setup_cb_destroy(block, tp, TC_SETUP_CLSFLOWER, &cls_flower, false,
459 			    &f->flags, &f->in_hw_count, rtnl_held);
460 
461 }
462 
463 static int fl_hw_replace_filter(struct tcf_proto *tp,
464 				struct cls_fl_filter *f, bool rtnl_held,
465 				struct netlink_ext_ack *extack)
466 {
467 	struct tcf_block *block = tp->chain->block;
468 	struct flow_cls_offload cls_flower = {};
469 	bool skip_sw = tc_skip_sw(f->flags);
470 	int err = 0;
471 
472 	cls_flower.rule = flow_rule_alloc(tcf_exts_num_actions(&f->exts));
473 	if (!cls_flower.rule)
474 		return -ENOMEM;
475 
476 	tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
477 	cls_flower.command = FLOW_CLS_REPLACE;
478 	cls_flower.cookie = (unsigned long) f;
479 	cls_flower.rule->match.dissector = &f->mask->dissector;
480 	cls_flower.rule->match.mask = &f->mask->key;
481 	cls_flower.rule->match.key = &f->mkey;
482 	cls_flower.classid = f->res.classid;
483 
484 	err = tc_setup_offload_action(&cls_flower.rule->action, &f->exts,
485 				      cls_flower.common.extack);
486 	if (err) {
487 		kfree(cls_flower.rule);
488 
489 		return skip_sw ? err : 0;
490 	}
491 
492 	err = tc_setup_cb_add(block, tp, TC_SETUP_CLSFLOWER, &cls_flower,
493 			      skip_sw, &f->flags, &f->in_hw_count, rtnl_held);
494 	tc_cleanup_offload_action(&cls_flower.rule->action);
495 	kfree(cls_flower.rule);
496 
497 	if (err) {
498 		fl_hw_destroy_filter(tp, f, rtnl_held, NULL);
499 		return err;
500 	}
501 
502 	if (skip_sw && !(f->flags & TCA_CLS_FLAGS_IN_HW))
503 		return -EINVAL;
504 
505 	return 0;
506 }
507 
508 static void fl_hw_update_stats(struct tcf_proto *tp, struct cls_fl_filter *f,
509 			       bool rtnl_held)
510 {
511 	struct tcf_block *block = tp->chain->block;
512 	struct flow_cls_offload cls_flower = {};
513 
514 	tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, NULL);
515 	cls_flower.command = FLOW_CLS_STATS;
516 	cls_flower.cookie = (unsigned long) f;
517 	cls_flower.classid = f->res.classid;
518 
519 	tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false,
520 			 rtnl_held);
521 
522 	tcf_exts_hw_stats_update(&f->exts, &cls_flower.stats, cls_flower.use_act_stats);
523 }
524 
525 static void __fl_put(struct cls_fl_filter *f)
526 {
527 	if (!refcount_dec_and_test(&f->refcnt))
528 		return;
529 
530 	if (tcf_exts_get_net(&f->exts))
531 		tcf_queue_work(&f->rwork, fl_destroy_filter_work);
532 	else
533 		__fl_destroy_filter(f);
534 }
535 
536 static struct cls_fl_filter *__fl_get(struct cls_fl_head *head, u32 handle)
537 {
538 	struct cls_fl_filter *f;
539 
540 	rcu_read_lock();
541 	f = idr_find(&head->handle_idr, handle);
542 	if (f && !refcount_inc_not_zero(&f->refcnt))
543 		f = NULL;
544 	rcu_read_unlock();
545 
546 	return f;
547 }
548 
549 static struct tcf_exts *fl_get_exts(const struct tcf_proto *tp, u32 handle)
550 {
551 	struct cls_fl_head *head = rcu_dereference_bh(tp->root);
552 	struct cls_fl_filter *f;
553 
554 	f = idr_find(&head->handle_idr, handle);
555 	return f ? &f->exts : NULL;
556 }
557 
558 static int __fl_delete(struct tcf_proto *tp, struct cls_fl_filter *f,
559 		       bool *last, bool rtnl_held,
560 		       struct netlink_ext_ack *extack)
561 {
562 	struct cls_fl_head *head = fl_head_dereference(tp);
563 
564 	*last = false;
565 
566 	spin_lock(&tp->lock);
567 	if (f->deleted) {
568 		spin_unlock(&tp->lock);
569 		return -ENOENT;
570 	}
571 
572 	f->deleted = true;
573 	rhashtable_remove_fast(&f->mask->ht, &f->ht_node,
574 			       f->mask->filter_ht_params);
575 	idr_remove(&head->handle_idr, f->handle);
576 	list_del_rcu(&f->list);
577 	spin_unlock(&tp->lock);
578 
579 	*last = fl_mask_put(head, f->mask);
580 	if (!tc_skip_hw(f->flags))
581 		fl_hw_destroy_filter(tp, f, rtnl_held, extack);
582 	tcf_unbind_filter(tp, &f->res);
583 	__fl_put(f);
584 
585 	return 0;
586 }
587 
588 static void fl_destroy_sleepable(struct work_struct *work)
589 {
590 	struct cls_fl_head *head = container_of(to_rcu_work(work),
591 						struct cls_fl_head,
592 						rwork);
593 
594 	rhashtable_destroy(&head->ht);
595 	kfree(head);
596 	module_put(THIS_MODULE);
597 }
598 
599 static void fl_destroy(struct tcf_proto *tp, bool rtnl_held,
600 		       struct netlink_ext_ack *extack)
601 {
602 	struct cls_fl_head *head = fl_head_dereference(tp);
603 	struct fl_flow_mask *mask, *next_mask;
604 	struct cls_fl_filter *f, *next;
605 	bool last;
606 
607 	list_for_each_entry_safe(mask, next_mask, &head->masks, list) {
608 		list_for_each_entry_safe(f, next, &mask->filters, list) {
609 			__fl_delete(tp, f, &last, rtnl_held, extack);
610 			if (last)
611 				break;
612 		}
613 	}
614 	idr_destroy(&head->handle_idr);
615 
616 	__module_get(THIS_MODULE);
617 	tcf_queue_work(&head->rwork, fl_destroy_sleepable);
618 }
619 
620 static void fl_put(struct tcf_proto *tp, void *arg)
621 {
622 	struct cls_fl_filter *f = arg;
623 
624 	__fl_put(f);
625 }
626 
627 static void *fl_get(struct tcf_proto *tp, u32 handle)
628 {
629 	struct cls_fl_head *head = fl_head_dereference(tp);
630 
631 	return __fl_get(head, handle);
632 }
633 
634 static const struct nla_policy fl_policy[TCA_FLOWER_MAX + 1] = {
635 	[TCA_FLOWER_UNSPEC]		= { .strict_start_type =
636 						TCA_FLOWER_L2_MISS },
637 	[TCA_FLOWER_CLASSID]		= { .type = NLA_U32 },
638 	[TCA_FLOWER_INDEV]		= { .type = NLA_STRING,
639 					    .len = IFNAMSIZ },
640 	[TCA_FLOWER_KEY_ETH_DST]	= { .len = ETH_ALEN },
641 	[TCA_FLOWER_KEY_ETH_DST_MASK]	= { .len = ETH_ALEN },
642 	[TCA_FLOWER_KEY_ETH_SRC]	= { .len = ETH_ALEN },
643 	[TCA_FLOWER_KEY_ETH_SRC_MASK]	= { .len = ETH_ALEN },
644 	[TCA_FLOWER_KEY_ETH_TYPE]	= { .type = NLA_U16 },
645 	[TCA_FLOWER_KEY_IP_PROTO]	= { .type = NLA_U8 },
646 	[TCA_FLOWER_KEY_IPV4_SRC]	= { .type = NLA_U32 },
647 	[TCA_FLOWER_KEY_IPV4_SRC_MASK]	= { .type = NLA_U32 },
648 	[TCA_FLOWER_KEY_IPV4_DST]	= { .type = NLA_U32 },
649 	[TCA_FLOWER_KEY_IPV4_DST_MASK]	= { .type = NLA_U32 },
650 	[TCA_FLOWER_KEY_IPV6_SRC]	= { .len = sizeof(struct in6_addr) },
651 	[TCA_FLOWER_KEY_IPV6_SRC_MASK]	= { .len = sizeof(struct in6_addr) },
652 	[TCA_FLOWER_KEY_IPV6_DST]	= { .len = sizeof(struct in6_addr) },
653 	[TCA_FLOWER_KEY_IPV6_DST_MASK]	= { .len = sizeof(struct in6_addr) },
654 	[TCA_FLOWER_KEY_TCP_SRC]	= { .type = NLA_U16 },
655 	[TCA_FLOWER_KEY_TCP_DST]	= { .type = NLA_U16 },
656 	[TCA_FLOWER_KEY_UDP_SRC]	= { .type = NLA_U16 },
657 	[TCA_FLOWER_KEY_UDP_DST]	= { .type = NLA_U16 },
658 	[TCA_FLOWER_KEY_VLAN_ID]	= { .type = NLA_U16 },
659 	[TCA_FLOWER_KEY_VLAN_PRIO]	= { .type = NLA_U8 },
660 	[TCA_FLOWER_KEY_VLAN_ETH_TYPE]	= { .type = NLA_U16 },
661 	[TCA_FLOWER_KEY_ENC_KEY_ID]	= { .type = NLA_U32 },
662 	[TCA_FLOWER_KEY_ENC_IPV4_SRC]	= { .type = NLA_U32 },
663 	[TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK] = { .type = NLA_U32 },
664 	[TCA_FLOWER_KEY_ENC_IPV4_DST]	= { .type = NLA_U32 },
665 	[TCA_FLOWER_KEY_ENC_IPV4_DST_MASK] = { .type = NLA_U32 },
666 	[TCA_FLOWER_KEY_ENC_IPV6_SRC]	= { .len = sizeof(struct in6_addr) },
667 	[TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK] = { .len = sizeof(struct in6_addr) },
668 	[TCA_FLOWER_KEY_ENC_IPV6_DST]	= { .len = sizeof(struct in6_addr) },
669 	[TCA_FLOWER_KEY_ENC_IPV6_DST_MASK] = { .len = sizeof(struct in6_addr) },
670 	[TCA_FLOWER_KEY_TCP_SRC_MASK]	= { .type = NLA_U16 },
671 	[TCA_FLOWER_KEY_TCP_DST_MASK]	= { .type = NLA_U16 },
672 	[TCA_FLOWER_KEY_UDP_SRC_MASK]	= { .type = NLA_U16 },
673 	[TCA_FLOWER_KEY_UDP_DST_MASK]	= { .type = NLA_U16 },
674 	[TCA_FLOWER_KEY_SCTP_SRC_MASK]	= { .type = NLA_U16 },
675 	[TCA_FLOWER_KEY_SCTP_DST_MASK]	= { .type = NLA_U16 },
676 	[TCA_FLOWER_KEY_SCTP_SRC]	= { .type = NLA_U16 },
677 	[TCA_FLOWER_KEY_SCTP_DST]	= { .type = NLA_U16 },
678 	[TCA_FLOWER_KEY_ENC_UDP_SRC_PORT]	= { .type = NLA_U16 },
679 	[TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK]	= { .type = NLA_U16 },
680 	[TCA_FLOWER_KEY_ENC_UDP_DST_PORT]	= { .type = NLA_U16 },
681 	[TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK]	= { .type = NLA_U16 },
682 	[TCA_FLOWER_KEY_FLAGS]		= NLA_POLICY_MASK(NLA_BE32,
683 							  TCA_FLOWER_KEY_FLAGS_POLICY_MASK),
684 	[TCA_FLOWER_KEY_FLAGS_MASK]	= NLA_POLICY_MASK(NLA_BE32,
685 							  TCA_FLOWER_KEY_FLAGS_POLICY_MASK),
686 	[TCA_FLOWER_KEY_ICMPV4_TYPE]	= { .type = NLA_U8 },
687 	[TCA_FLOWER_KEY_ICMPV4_TYPE_MASK] = { .type = NLA_U8 },
688 	[TCA_FLOWER_KEY_ICMPV4_CODE]	= { .type = NLA_U8 },
689 	[TCA_FLOWER_KEY_ICMPV4_CODE_MASK] = { .type = NLA_U8 },
690 	[TCA_FLOWER_KEY_ICMPV6_TYPE]	= { .type = NLA_U8 },
691 	[TCA_FLOWER_KEY_ICMPV6_TYPE_MASK] = { .type = NLA_U8 },
692 	[TCA_FLOWER_KEY_ICMPV6_CODE]	= { .type = NLA_U8 },
693 	[TCA_FLOWER_KEY_ICMPV6_CODE_MASK] = { .type = NLA_U8 },
694 	[TCA_FLOWER_KEY_ARP_SIP]	= { .type = NLA_U32 },
695 	[TCA_FLOWER_KEY_ARP_SIP_MASK]	= { .type = NLA_U32 },
696 	[TCA_FLOWER_KEY_ARP_TIP]	= { .type = NLA_U32 },
697 	[TCA_FLOWER_KEY_ARP_TIP_MASK]	= { .type = NLA_U32 },
698 	[TCA_FLOWER_KEY_ARP_OP]		= { .type = NLA_U8 },
699 	[TCA_FLOWER_KEY_ARP_OP_MASK]	= { .type = NLA_U8 },
700 	[TCA_FLOWER_KEY_ARP_SHA]	= { .len = ETH_ALEN },
701 	[TCA_FLOWER_KEY_ARP_SHA_MASK]	= { .len = ETH_ALEN },
702 	[TCA_FLOWER_KEY_ARP_THA]	= { .len = ETH_ALEN },
703 	[TCA_FLOWER_KEY_ARP_THA_MASK]	= { .len = ETH_ALEN },
704 	[TCA_FLOWER_KEY_MPLS_TTL]	= { .type = NLA_U8 },
705 	[TCA_FLOWER_KEY_MPLS_BOS]	= { .type = NLA_U8 },
706 	[TCA_FLOWER_KEY_MPLS_TC]	= { .type = NLA_U8 },
707 	[TCA_FLOWER_KEY_MPLS_LABEL]	= { .type = NLA_U32 },
708 	[TCA_FLOWER_KEY_MPLS_OPTS]	= { .type = NLA_NESTED },
709 	[TCA_FLOWER_KEY_TCP_FLAGS]	= { .type = NLA_U16 },
710 	[TCA_FLOWER_KEY_TCP_FLAGS_MASK]	= { .type = NLA_U16 },
711 	[TCA_FLOWER_KEY_IP_TOS]		= { .type = NLA_U8 },
712 	[TCA_FLOWER_KEY_IP_TOS_MASK]	= { .type = NLA_U8 },
713 	[TCA_FLOWER_KEY_IP_TTL]		= { .type = NLA_U8 },
714 	[TCA_FLOWER_KEY_IP_TTL_MASK]	= { .type = NLA_U8 },
715 	[TCA_FLOWER_KEY_CVLAN_ID]	= { .type = NLA_U16 },
716 	[TCA_FLOWER_KEY_CVLAN_PRIO]	= { .type = NLA_U8 },
717 	[TCA_FLOWER_KEY_CVLAN_ETH_TYPE]	= { .type = NLA_U16 },
718 	[TCA_FLOWER_KEY_ENC_IP_TOS]	= { .type = NLA_U8 },
719 	[TCA_FLOWER_KEY_ENC_IP_TOS_MASK] = { .type = NLA_U8 },
720 	[TCA_FLOWER_KEY_ENC_IP_TTL]	 = { .type = NLA_U8 },
721 	[TCA_FLOWER_KEY_ENC_IP_TTL_MASK] = { .type = NLA_U8 },
722 	[TCA_FLOWER_KEY_ENC_OPTS]	= { .type = NLA_NESTED },
723 	[TCA_FLOWER_KEY_ENC_OPTS_MASK]	= { .type = NLA_NESTED },
724 	[TCA_FLOWER_KEY_CT_STATE]	=
725 		NLA_POLICY_MASK(NLA_U16, TCA_FLOWER_KEY_CT_FLAGS_MASK),
726 	[TCA_FLOWER_KEY_CT_STATE_MASK]	=
727 		NLA_POLICY_MASK(NLA_U16, TCA_FLOWER_KEY_CT_FLAGS_MASK),
728 	[TCA_FLOWER_KEY_CT_ZONE]	= { .type = NLA_U16 },
729 	[TCA_FLOWER_KEY_CT_ZONE_MASK]	= { .type = NLA_U16 },
730 	[TCA_FLOWER_KEY_CT_MARK]	= { .type = NLA_U32 },
731 	[TCA_FLOWER_KEY_CT_MARK_MASK]	= { .type = NLA_U32 },
732 	[TCA_FLOWER_KEY_CT_LABELS]	= { .type = NLA_BINARY,
733 					    .len = 128 / BITS_PER_BYTE },
734 	[TCA_FLOWER_KEY_CT_LABELS_MASK]	= { .type = NLA_BINARY,
735 					    .len = 128 / BITS_PER_BYTE },
736 	[TCA_FLOWER_FLAGS]		= { .type = NLA_U32 },
737 	[TCA_FLOWER_KEY_HASH]		= { .type = NLA_U32 },
738 	[TCA_FLOWER_KEY_HASH_MASK]	= { .type = NLA_U32 },
739 	[TCA_FLOWER_KEY_NUM_OF_VLANS]	= { .type = NLA_U8 },
740 	[TCA_FLOWER_KEY_PPPOE_SID]	= { .type = NLA_U16 },
741 	[TCA_FLOWER_KEY_PPP_PROTO]	= { .type = NLA_U16 },
742 	[TCA_FLOWER_KEY_L2TPV3_SID]	= { .type = NLA_U32 },
743 	[TCA_FLOWER_KEY_SPI]		= { .type = NLA_U32 },
744 	[TCA_FLOWER_KEY_SPI_MASK]	= { .type = NLA_U32 },
745 	[TCA_FLOWER_L2_MISS]		= NLA_POLICY_MAX(NLA_U8, 1),
746 	[TCA_FLOWER_KEY_CFM]		= { .type = NLA_NESTED },
747 	[TCA_FLOWER_KEY_ENC_FLAGS]	= NLA_POLICY_MASK(NLA_BE32,
748 							  TCA_FLOWER_KEY_ENC_FLAGS_POLICY_MASK),
749 	[TCA_FLOWER_KEY_ENC_FLAGS_MASK]	= NLA_POLICY_MASK(NLA_BE32,
750 							  TCA_FLOWER_KEY_ENC_FLAGS_POLICY_MASK),
751 };
752 
753 static const struct nla_policy
754 enc_opts_policy[TCA_FLOWER_KEY_ENC_OPTS_MAX + 1] = {
755 	[TCA_FLOWER_KEY_ENC_OPTS_UNSPEC]        = {
756 		.strict_start_type = TCA_FLOWER_KEY_ENC_OPTS_VXLAN },
757 	[TCA_FLOWER_KEY_ENC_OPTS_GENEVE]        = { .type = NLA_NESTED },
758 	[TCA_FLOWER_KEY_ENC_OPTS_VXLAN]         = { .type = NLA_NESTED },
759 	[TCA_FLOWER_KEY_ENC_OPTS_ERSPAN]        = { .type = NLA_NESTED },
760 	[TCA_FLOWER_KEY_ENC_OPTS_GTP]		= { .type = NLA_NESTED },
761 	[TCA_FLOWER_KEY_ENC_OPTS_PFCP]		= { .type = NLA_NESTED },
762 };
763 
764 static const struct nla_policy
765 geneve_opt_policy[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1] = {
766 	[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS]      = { .type = NLA_U16 },
767 	[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE]       = { .type = NLA_U8 },
768 	[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA]       = { .type = NLA_BINARY,
769 						       .len = 128 },
770 };
771 
772 static const struct nla_policy
773 vxlan_opt_policy[TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX + 1] = {
774 	[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]         = { .type = NLA_U32 },
775 };
776 
777 static const struct nla_policy
778 erspan_opt_policy[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX + 1] = {
779 	[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]        = { .type = NLA_U8 },
780 	[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]      = { .type = NLA_U32 },
781 	[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR]        = { .type = NLA_U8 },
782 	[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID]       = { .type = NLA_U8 },
783 };
784 
785 static const struct nla_policy
786 gtp_opt_policy[TCA_FLOWER_KEY_ENC_OPT_GTP_MAX + 1] = {
787 	[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE]	   = { .type = NLA_U8 },
788 	[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI]	   = { .type = NLA_U8 },
789 };
790 
791 static const struct nla_policy
792 pfcp_opt_policy[TCA_FLOWER_KEY_ENC_OPT_PFCP_MAX + 1] = {
793 	[TCA_FLOWER_KEY_ENC_OPT_PFCP_TYPE]	   = { .type = NLA_U8 },
794 	[TCA_FLOWER_KEY_ENC_OPT_PFCP_SEID]	   = { .type = NLA_U64 },
795 };
796 
797 static const struct nla_policy
798 mpls_stack_entry_policy[TCA_FLOWER_KEY_MPLS_OPT_LSE_MAX + 1] = {
799 	[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH]    = { .type = NLA_U8 },
800 	[TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL]      = { .type = NLA_U8 },
801 	[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS]      = { .type = NLA_U8 },
802 	[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC]       = { .type = NLA_U8 },
803 	[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL]    = { .type = NLA_U32 },
804 };
805 
806 static const struct nla_policy
807 cfm_opt_policy[TCA_FLOWER_KEY_CFM_OPT_MAX + 1] = {
808 	[TCA_FLOWER_KEY_CFM_MD_LEVEL]	= NLA_POLICY_MAX(NLA_U8,
809 						FLOW_DIS_CFM_MDL_MAX),
810 	[TCA_FLOWER_KEY_CFM_OPCODE]	= { .type = NLA_U8 },
811 };
812 
813 static void fl_set_key_val(struct nlattr **tb,
814 			   void *val, int val_type,
815 			   void *mask, int mask_type, int len)
816 {
817 	if (!tb[val_type])
818 		return;
819 	nla_memcpy(val, tb[val_type], len);
820 	if (mask_type == TCA_FLOWER_UNSPEC || !tb[mask_type])
821 		memset(mask, 0xff, len);
822 	else
823 		nla_memcpy(mask, tb[mask_type], len);
824 }
825 
826 static int fl_set_key_spi(struct nlattr **tb, struct fl_flow_key *key,
827 			  struct fl_flow_key *mask,
828 			  struct netlink_ext_ack *extack)
829 {
830 	if (key->basic.ip_proto != IPPROTO_ESP &&
831 	    key->basic.ip_proto != IPPROTO_AH) {
832 		NL_SET_ERR_MSG(extack,
833 			       "Protocol must be either ESP or AH");
834 		return -EINVAL;
835 	}
836 
837 	fl_set_key_val(tb, &key->ipsec.spi,
838 		       TCA_FLOWER_KEY_SPI,
839 		       &mask->ipsec.spi, TCA_FLOWER_KEY_SPI_MASK,
840 		       sizeof(key->ipsec.spi));
841 	return 0;
842 }
843 
844 static int fl_set_key_port_range(struct nlattr **tb, struct fl_flow_key *key,
845 				 struct fl_flow_key *mask,
846 				 struct netlink_ext_ack *extack)
847 {
848 	fl_set_key_val(tb, &key->tp_range.tp_min.dst,
849 		       TCA_FLOWER_KEY_PORT_DST_MIN, &mask->tp_range.tp_min.dst,
850 		       TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_min.dst));
851 	fl_set_key_val(tb, &key->tp_range.tp_max.dst,
852 		       TCA_FLOWER_KEY_PORT_DST_MAX, &mask->tp_range.tp_max.dst,
853 		       TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_max.dst));
854 	fl_set_key_val(tb, &key->tp_range.tp_min.src,
855 		       TCA_FLOWER_KEY_PORT_SRC_MIN, &mask->tp_range.tp_min.src,
856 		       TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_min.src));
857 	fl_set_key_val(tb, &key->tp_range.tp_max.src,
858 		       TCA_FLOWER_KEY_PORT_SRC_MAX, &mask->tp_range.tp_max.src,
859 		       TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_max.src));
860 
861 	if (mask->tp_range.tp_min.dst != mask->tp_range.tp_max.dst) {
862 		NL_SET_ERR_MSG(extack,
863 			       "Both min and max destination ports must be specified");
864 		return -EINVAL;
865 	}
866 	if (mask->tp_range.tp_min.src != mask->tp_range.tp_max.src) {
867 		NL_SET_ERR_MSG(extack,
868 			       "Both min and max source ports must be specified");
869 		return -EINVAL;
870 	}
871 	if (mask->tp_range.tp_min.dst && mask->tp_range.tp_max.dst &&
872 	    ntohs(key->tp_range.tp_max.dst) <=
873 	    ntohs(key->tp_range.tp_min.dst)) {
874 		NL_SET_ERR_MSG_ATTR(extack,
875 				    tb[TCA_FLOWER_KEY_PORT_DST_MIN],
876 				    "Invalid destination port range (min must be strictly smaller than max)");
877 		return -EINVAL;
878 	}
879 	if (mask->tp_range.tp_min.src && mask->tp_range.tp_max.src &&
880 	    ntohs(key->tp_range.tp_max.src) <=
881 	    ntohs(key->tp_range.tp_min.src)) {
882 		NL_SET_ERR_MSG_ATTR(extack,
883 				    tb[TCA_FLOWER_KEY_PORT_SRC_MIN],
884 				    "Invalid source port range (min must be strictly smaller than max)");
885 		return -EINVAL;
886 	}
887 
888 	return 0;
889 }
890 
891 static int fl_set_key_mpls_lse(const struct nlattr *nla_lse,
892 			       struct flow_dissector_key_mpls *key_val,
893 			       struct flow_dissector_key_mpls *key_mask,
894 			       struct netlink_ext_ack *extack)
895 {
896 	struct nlattr *tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_MAX + 1];
897 	struct flow_dissector_mpls_lse *lse_mask;
898 	struct flow_dissector_mpls_lse *lse_val;
899 	u8 lse_index;
900 	u8 depth;
901 	int err;
902 
903 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_MPLS_OPT_LSE_MAX, nla_lse,
904 			       mpls_stack_entry_policy, extack);
905 	if (err < 0)
906 		return err;
907 
908 	if (!tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH]) {
909 		NL_SET_ERR_MSG(extack, "Missing MPLS option \"depth\"");
910 		return -EINVAL;
911 	}
912 
913 	depth = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH]);
914 
915 	/* LSE depth starts at 1, for consistency with terminology used by
916 	 * RFC 3031 (section 3.9), where depth 0 refers to unlabeled packets.
917 	 */
918 	if (depth < 1 || depth > FLOW_DIS_MPLS_MAX) {
919 		NL_SET_ERR_MSG_ATTR(extack,
920 				    tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH],
921 				    "Invalid MPLS depth");
922 		return -EINVAL;
923 	}
924 	lse_index = depth - 1;
925 
926 	dissector_set_mpls_lse(key_val, lse_index);
927 	dissector_set_mpls_lse(key_mask, lse_index);
928 
929 	lse_val = &key_val->ls[lse_index];
930 	lse_mask = &key_mask->ls[lse_index];
931 
932 	if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL]) {
933 		lse_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL]);
934 		lse_mask->mpls_ttl = MPLS_TTL_MASK;
935 	}
936 	if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS]) {
937 		u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS]);
938 
939 		if (bos & ~MPLS_BOS_MASK) {
940 			NL_SET_ERR_MSG_ATTR(extack,
941 					    tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS],
942 					    "Bottom Of Stack (BOS) must be 0 or 1");
943 			return -EINVAL;
944 		}
945 		lse_val->mpls_bos = bos;
946 		lse_mask->mpls_bos = MPLS_BOS_MASK;
947 	}
948 	if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC]) {
949 		u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC]);
950 
951 		if (tc & ~MPLS_TC_MASK) {
952 			NL_SET_ERR_MSG_ATTR(extack,
953 					    tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC],
954 					    "Traffic Class (TC) must be between 0 and 7");
955 			return -EINVAL;
956 		}
957 		lse_val->mpls_tc = tc;
958 		lse_mask->mpls_tc = MPLS_TC_MASK;
959 	}
960 	if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL]) {
961 		u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL]);
962 
963 		if (label & ~MPLS_LABEL_MASK) {
964 			NL_SET_ERR_MSG_ATTR(extack,
965 					    tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL],
966 					    "Label must be between 0 and 1048575");
967 			return -EINVAL;
968 		}
969 		lse_val->mpls_label = label;
970 		lse_mask->mpls_label = MPLS_LABEL_MASK;
971 	}
972 
973 	return 0;
974 }
975 
976 static int fl_set_key_mpls_opts(const struct nlattr *nla_mpls_opts,
977 				struct flow_dissector_key_mpls *key_val,
978 				struct flow_dissector_key_mpls *key_mask,
979 				struct netlink_ext_ack *extack)
980 {
981 	struct nlattr *nla_lse;
982 	int rem;
983 	int err;
984 
985 	if (!(nla_mpls_opts->nla_type & NLA_F_NESTED)) {
986 		NL_SET_ERR_MSG_ATTR(extack, nla_mpls_opts,
987 				    "NLA_F_NESTED is missing");
988 		return -EINVAL;
989 	}
990 
991 	nla_for_each_nested(nla_lse, nla_mpls_opts, rem) {
992 		if (nla_type(nla_lse) != TCA_FLOWER_KEY_MPLS_OPTS_LSE) {
993 			NL_SET_ERR_MSG_ATTR(extack, nla_lse,
994 					    "Invalid MPLS option type");
995 			return -EINVAL;
996 		}
997 
998 		err = fl_set_key_mpls_lse(nla_lse, key_val, key_mask, extack);
999 		if (err < 0)
1000 			return err;
1001 	}
1002 	if (rem) {
1003 		NL_SET_ERR_MSG(extack,
1004 			       "Bytes leftover after parsing MPLS options");
1005 		return -EINVAL;
1006 	}
1007 
1008 	return 0;
1009 }
1010 
1011 static int fl_set_key_mpls(struct nlattr **tb,
1012 			   struct flow_dissector_key_mpls *key_val,
1013 			   struct flow_dissector_key_mpls *key_mask,
1014 			   struct netlink_ext_ack *extack)
1015 {
1016 	struct flow_dissector_mpls_lse *lse_mask;
1017 	struct flow_dissector_mpls_lse *lse_val;
1018 
1019 	if (tb[TCA_FLOWER_KEY_MPLS_OPTS]) {
1020 		if (tb[TCA_FLOWER_KEY_MPLS_TTL] ||
1021 		    tb[TCA_FLOWER_KEY_MPLS_BOS] ||
1022 		    tb[TCA_FLOWER_KEY_MPLS_TC] ||
1023 		    tb[TCA_FLOWER_KEY_MPLS_LABEL]) {
1024 			NL_SET_ERR_MSG_ATTR(extack,
1025 					    tb[TCA_FLOWER_KEY_MPLS_OPTS],
1026 					    "MPLS label, Traffic Class, Bottom Of Stack and Time To Live must be encapsulated in the MPLS options attribute");
1027 			return -EBADMSG;
1028 		}
1029 
1030 		return fl_set_key_mpls_opts(tb[TCA_FLOWER_KEY_MPLS_OPTS],
1031 					    key_val, key_mask, extack);
1032 	}
1033 
1034 	lse_val = &key_val->ls[0];
1035 	lse_mask = &key_mask->ls[0];
1036 
1037 	if (tb[TCA_FLOWER_KEY_MPLS_TTL]) {
1038 		lse_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TTL]);
1039 		lse_mask->mpls_ttl = MPLS_TTL_MASK;
1040 		dissector_set_mpls_lse(key_val, 0);
1041 		dissector_set_mpls_lse(key_mask, 0);
1042 	}
1043 	if (tb[TCA_FLOWER_KEY_MPLS_BOS]) {
1044 		u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_BOS]);
1045 
1046 		if (bos & ~MPLS_BOS_MASK) {
1047 			NL_SET_ERR_MSG_ATTR(extack,
1048 					    tb[TCA_FLOWER_KEY_MPLS_BOS],
1049 					    "Bottom Of Stack (BOS) must be 0 or 1");
1050 			return -EINVAL;
1051 		}
1052 		lse_val->mpls_bos = bos;
1053 		lse_mask->mpls_bos = MPLS_BOS_MASK;
1054 		dissector_set_mpls_lse(key_val, 0);
1055 		dissector_set_mpls_lse(key_mask, 0);
1056 	}
1057 	if (tb[TCA_FLOWER_KEY_MPLS_TC]) {
1058 		u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TC]);
1059 
1060 		if (tc & ~MPLS_TC_MASK) {
1061 			NL_SET_ERR_MSG_ATTR(extack,
1062 					    tb[TCA_FLOWER_KEY_MPLS_TC],
1063 					    "Traffic Class (TC) must be between 0 and 7");
1064 			return -EINVAL;
1065 		}
1066 		lse_val->mpls_tc = tc;
1067 		lse_mask->mpls_tc = MPLS_TC_MASK;
1068 		dissector_set_mpls_lse(key_val, 0);
1069 		dissector_set_mpls_lse(key_mask, 0);
1070 	}
1071 	if (tb[TCA_FLOWER_KEY_MPLS_LABEL]) {
1072 		u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_LABEL]);
1073 
1074 		if (label & ~MPLS_LABEL_MASK) {
1075 			NL_SET_ERR_MSG_ATTR(extack,
1076 					    tb[TCA_FLOWER_KEY_MPLS_LABEL],
1077 					    "Label must be between 0 and 1048575");
1078 			return -EINVAL;
1079 		}
1080 		lse_val->mpls_label = label;
1081 		lse_mask->mpls_label = MPLS_LABEL_MASK;
1082 		dissector_set_mpls_lse(key_val, 0);
1083 		dissector_set_mpls_lse(key_mask, 0);
1084 	}
1085 	return 0;
1086 }
1087 
1088 static void fl_set_key_vlan(struct nlattr **tb,
1089 			    __be16 ethertype,
1090 			    int vlan_id_key, int vlan_prio_key,
1091 			    int vlan_next_eth_type_key,
1092 			    struct flow_dissector_key_vlan *key_val,
1093 			    struct flow_dissector_key_vlan *key_mask)
1094 {
1095 #define VLAN_PRIORITY_MASK	0x7
1096 
1097 	if (tb[vlan_id_key]) {
1098 		key_val->vlan_id =
1099 			nla_get_u16(tb[vlan_id_key]) & VLAN_VID_MASK;
1100 		key_mask->vlan_id = VLAN_VID_MASK;
1101 	}
1102 	if (tb[vlan_prio_key]) {
1103 		key_val->vlan_priority =
1104 			nla_get_u8(tb[vlan_prio_key]) &
1105 			VLAN_PRIORITY_MASK;
1106 		key_mask->vlan_priority = VLAN_PRIORITY_MASK;
1107 	}
1108 	if (ethertype) {
1109 		key_val->vlan_tpid = ethertype;
1110 		key_mask->vlan_tpid = cpu_to_be16(~0);
1111 	}
1112 	if (tb[vlan_next_eth_type_key]) {
1113 		key_val->vlan_eth_type =
1114 			nla_get_be16(tb[vlan_next_eth_type_key]);
1115 		key_mask->vlan_eth_type = cpu_to_be16(~0);
1116 	}
1117 }
1118 
1119 static void fl_set_key_pppoe(struct nlattr **tb,
1120 			     struct flow_dissector_key_pppoe *key_val,
1121 			     struct flow_dissector_key_pppoe *key_mask,
1122 			     struct fl_flow_key *key,
1123 			     struct fl_flow_key *mask)
1124 {
1125 	/* key_val::type must be set to ETH_P_PPP_SES
1126 	 * because ETH_P_PPP_SES was stored in basic.n_proto
1127 	 * which might get overwritten by ppp_proto
1128 	 * or might be set to 0, the role of key_val::type
1129 	 * is similar to vlan_key::tpid
1130 	 */
1131 	key_val->type = htons(ETH_P_PPP_SES);
1132 	key_mask->type = cpu_to_be16(~0);
1133 
1134 	if (tb[TCA_FLOWER_KEY_PPPOE_SID]) {
1135 		key_val->session_id =
1136 			nla_get_be16(tb[TCA_FLOWER_KEY_PPPOE_SID]);
1137 		key_mask->session_id = cpu_to_be16(~0);
1138 	}
1139 	if (tb[TCA_FLOWER_KEY_PPP_PROTO]) {
1140 		key_val->ppp_proto =
1141 			nla_get_be16(tb[TCA_FLOWER_KEY_PPP_PROTO]);
1142 		key_mask->ppp_proto = cpu_to_be16(~0);
1143 
1144 		if (key_val->ppp_proto == htons(PPP_IP)) {
1145 			key->basic.n_proto = htons(ETH_P_IP);
1146 			mask->basic.n_proto = cpu_to_be16(~0);
1147 		} else if (key_val->ppp_proto == htons(PPP_IPV6)) {
1148 			key->basic.n_proto = htons(ETH_P_IPV6);
1149 			mask->basic.n_proto = cpu_to_be16(~0);
1150 		} else if (key_val->ppp_proto == htons(PPP_MPLS_UC)) {
1151 			key->basic.n_proto = htons(ETH_P_MPLS_UC);
1152 			mask->basic.n_proto = cpu_to_be16(~0);
1153 		} else if (key_val->ppp_proto == htons(PPP_MPLS_MC)) {
1154 			key->basic.n_proto = htons(ETH_P_MPLS_MC);
1155 			mask->basic.n_proto = cpu_to_be16(~0);
1156 		}
1157 	} else {
1158 		key->basic.n_proto = 0;
1159 		mask->basic.n_proto = cpu_to_be16(0);
1160 	}
1161 }
1162 
1163 static void fl_set_key_flag(u32 flower_key, u32 flower_mask,
1164 			    u32 *dissector_key, u32 *dissector_mask,
1165 			    u32 flower_flag_bit, u32 dissector_flag_bit)
1166 {
1167 	if (flower_mask & flower_flag_bit) {
1168 		*dissector_mask |= dissector_flag_bit;
1169 		if (flower_key & flower_flag_bit)
1170 			*dissector_key |= dissector_flag_bit;
1171 	}
1172 }
1173 
1174 static int fl_set_key_flags(struct nlattr *tca_opts, struct nlattr **tb,
1175 			    bool encap, u32 *flags_key, u32 *flags_mask,
1176 			    struct netlink_ext_ack *extack)
1177 {
1178 	int fl_key, fl_mask;
1179 	u32 key, mask;
1180 
1181 	if (encap) {
1182 		fl_key = TCA_FLOWER_KEY_ENC_FLAGS;
1183 		fl_mask = TCA_FLOWER_KEY_ENC_FLAGS_MASK;
1184 	} else {
1185 		fl_key = TCA_FLOWER_KEY_FLAGS;
1186 		fl_mask = TCA_FLOWER_KEY_FLAGS_MASK;
1187 	}
1188 
1189 	/* mask is mandatory for flags */
1190 	if (NL_REQ_ATTR_CHECK(extack, tca_opts, tb, fl_mask)) {
1191 		NL_SET_ERR_MSG(extack, "Missing flags mask");
1192 		return -EINVAL;
1193 	}
1194 
1195 	key = be32_to_cpu(nla_get_be32(tb[fl_key]));
1196 	mask = be32_to_cpu(nla_get_be32(tb[fl_mask]));
1197 
1198 	*flags_key  = 0;
1199 	*flags_mask = 0;
1200 
1201 	fl_set_key_flag(key, mask, flags_key, flags_mask,
1202 			TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
1203 	fl_set_key_flag(key, mask, flags_key, flags_mask,
1204 			TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
1205 			FLOW_DIS_FIRST_FRAG);
1206 
1207 	fl_set_key_flag(key, mask, flags_key, flags_mask,
1208 			TCA_FLOWER_KEY_FLAGS_TUNNEL_CSUM,
1209 			FLOW_DIS_F_TUNNEL_CSUM);
1210 
1211 	fl_set_key_flag(key, mask, flags_key, flags_mask,
1212 			TCA_FLOWER_KEY_FLAGS_TUNNEL_DONT_FRAGMENT,
1213 			FLOW_DIS_F_TUNNEL_DONT_FRAGMENT);
1214 
1215 	fl_set_key_flag(key, mask, flags_key, flags_mask,
1216 			TCA_FLOWER_KEY_FLAGS_TUNNEL_OAM, FLOW_DIS_F_TUNNEL_OAM);
1217 
1218 	fl_set_key_flag(key, mask, flags_key, flags_mask,
1219 			TCA_FLOWER_KEY_FLAGS_TUNNEL_CRIT_OPT,
1220 			FLOW_DIS_F_TUNNEL_CRIT_OPT);
1221 
1222 	return 0;
1223 }
1224 
1225 static void fl_set_key_ip(struct nlattr **tb, bool encap,
1226 			  struct flow_dissector_key_ip *key,
1227 			  struct flow_dissector_key_ip *mask)
1228 {
1229 	int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
1230 	int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
1231 	int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
1232 	int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
1233 
1234 	fl_set_key_val(tb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos));
1235 	fl_set_key_val(tb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl));
1236 }
1237 
1238 static int fl_set_geneve_opt(const struct nlattr *nla, struct fl_flow_key *key,
1239 			     int depth, int option_len,
1240 			     struct netlink_ext_ack *extack)
1241 {
1242 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1];
1243 	struct nlattr *class = NULL, *type = NULL, *data = NULL;
1244 	struct geneve_opt *opt;
1245 	int err, data_len = 0;
1246 
1247 	if (option_len > sizeof(struct geneve_opt))
1248 		data_len = option_len - sizeof(struct geneve_opt);
1249 
1250 	if (key->enc_opts.len > FLOW_DIS_TUN_OPTS_MAX - 4)
1251 		return -ERANGE;
1252 
1253 	opt = (struct geneve_opt *)&key->enc_opts.data[key->enc_opts.len];
1254 	memset(opt, 0xff, option_len);
1255 	opt->length = data_len / 4;
1256 	opt->r1 = 0;
1257 	opt->r2 = 0;
1258 	opt->r3 = 0;
1259 
1260 	/* If no mask has been prodived we assume an exact match. */
1261 	if (!depth)
1262 		return sizeof(struct geneve_opt) + data_len;
1263 
1264 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_GENEVE) {
1265 		NL_SET_ERR_MSG(extack, "Non-geneve option type for mask");
1266 		return -EINVAL;
1267 	}
1268 
1269 	err = nla_parse_nested_deprecated(tb,
1270 					  TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX,
1271 					  nla, geneve_opt_policy, extack);
1272 	if (err < 0)
1273 		return err;
1274 
1275 	/* We are not allowed to omit any of CLASS, TYPE or DATA
1276 	 * fields from the key.
1277 	 */
1278 	if (!option_len &&
1279 	    (!tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS] ||
1280 	     !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE] ||
1281 	     !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA])) {
1282 		NL_SET_ERR_MSG(extack, "Missing tunnel key geneve option class, type or data");
1283 		return -EINVAL;
1284 	}
1285 
1286 	/* Omitting any of CLASS, TYPE or DATA fields is allowed
1287 	 * for the mask.
1288 	 */
1289 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA]) {
1290 		int new_len = key->enc_opts.len;
1291 
1292 		data = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA];
1293 		data_len = nla_len(data);
1294 		if (data_len < 4) {
1295 			NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is less than 4 bytes long");
1296 			return -ERANGE;
1297 		}
1298 		if (data_len % 4) {
1299 			NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is not a multiple of 4 bytes long");
1300 			return -ERANGE;
1301 		}
1302 
1303 		new_len += sizeof(struct geneve_opt) + data_len;
1304 		BUILD_BUG_ON(FLOW_DIS_TUN_OPTS_MAX != IP_TUNNEL_OPTS_MAX);
1305 		if (new_len > FLOW_DIS_TUN_OPTS_MAX) {
1306 			NL_SET_ERR_MSG(extack, "Tunnel options exceeds max size");
1307 			return -ERANGE;
1308 		}
1309 		opt->length = data_len / 4;
1310 		memcpy(opt->opt_data, nla_data(data), data_len);
1311 	}
1312 
1313 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS]) {
1314 		class = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS];
1315 		opt->opt_class = nla_get_be16(class);
1316 	}
1317 
1318 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE]) {
1319 		type = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE];
1320 		opt->type = nla_get_u8(type);
1321 	}
1322 
1323 	return sizeof(struct geneve_opt) + data_len;
1324 }
1325 
1326 static int fl_set_vxlan_opt(const struct nlattr *nla, struct fl_flow_key *key,
1327 			    int depth, int option_len,
1328 			    struct netlink_ext_ack *extack)
1329 {
1330 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX + 1];
1331 	struct vxlan_metadata *md;
1332 	int err;
1333 
1334 	md = (struct vxlan_metadata *)&key->enc_opts.data[key->enc_opts.len];
1335 	memset(md, 0xff, sizeof(*md));
1336 
1337 	if (!depth)
1338 		return sizeof(*md);
1339 
1340 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_VXLAN) {
1341 		NL_SET_ERR_MSG(extack, "Non-vxlan option type for mask");
1342 		return -EINVAL;
1343 	}
1344 
1345 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX, nla,
1346 			       vxlan_opt_policy, extack);
1347 	if (err < 0)
1348 		return err;
1349 
1350 	if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]) {
1351 		NL_SET_ERR_MSG(extack, "Missing tunnel key vxlan option gbp");
1352 		return -EINVAL;
1353 	}
1354 
1355 	if (tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]) {
1356 		md->gbp = nla_get_u32(tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]);
1357 		md->gbp &= VXLAN_GBP_MASK;
1358 	}
1359 
1360 	return sizeof(*md);
1361 }
1362 
1363 static int fl_set_erspan_opt(const struct nlattr *nla, struct fl_flow_key *key,
1364 			     int depth, int option_len,
1365 			     struct netlink_ext_ack *extack)
1366 {
1367 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX + 1];
1368 	struct erspan_metadata *md;
1369 	int err;
1370 
1371 	md = (struct erspan_metadata *)&key->enc_opts.data[key->enc_opts.len];
1372 	memset(md, 0xff, sizeof(*md));
1373 	md->version = 1;
1374 
1375 	if (!depth)
1376 		return sizeof(*md);
1377 
1378 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_ERSPAN) {
1379 		NL_SET_ERR_MSG(extack, "Non-erspan option type for mask");
1380 		return -EINVAL;
1381 	}
1382 
1383 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX, nla,
1384 			       erspan_opt_policy, extack);
1385 	if (err < 0)
1386 		return err;
1387 
1388 	if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]) {
1389 		NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option ver");
1390 		return -EINVAL;
1391 	}
1392 
1393 	if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER])
1394 		md->version = nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]);
1395 
1396 	if (md->version == 1) {
1397 		if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]) {
1398 			NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option index");
1399 			return -EINVAL;
1400 		}
1401 		if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]) {
1402 			nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX];
1403 			memset(&md->u, 0x00, sizeof(md->u));
1404 			md->u.index = nla_get_be32(nla);
1405 		}
1406 	} else if (md->version == 2) {
1407 		if (!option_len && (!tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR] ||
1408 				    !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID])) {
1409 			NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option dir or hwid");
1410 			return -EINVAL;
1411 		}
1412 		if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR]) {
1413 			nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR];
1414 			md->u.md2.dir = nla_get_u8(nla);
1415 		}
1416 		if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID]) {
1417 			nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID];
1418 			set_hwid(&md->u.md2, nla_get_u8(nla));
1419 		}
1420 	} else {
1421 		NL_SET_ERR_MSG(extack, "Tunnel key erspan option ver is incorrect");
1422 		return -EINVAL;
1423 	}
1424 
1425 	return sizeof(*md);
1426 }
1427 
1428 static int fl_set_gtp_opt(const struct nlattr *nla, struct fl_flow_key *key,
1429 			  int depth, int option_len,
1430 			  struct netlink_ext_ack *extack)
1431 {
1432 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_GTP_MAX + 1];
1433 	struct gtp_pdu_session_info *sinfo;
1434 	u8 len = key->enc_opts.len;
1435 	int err;
1436 
1437 	sinfo = (struct gtp_pdu_session_info *)&key->enc_opts.data[len];
1438 	memset(sinfo, 0xff, option_len);
1439 
1440 	if (!depth)
1441 		return sizeof(*sinfo);
1442 
1443 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_GTP) {
1444 		NL_SET_ERR_MSG_MOD(extack, "Non-gtp option type for mask");
1445 		return -EINVAL;
1446 	}
1447 
1448 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_GTP_MAX, nla,
1449 			       gtp_opt_policy, extack);
1450 	if (err < 0)
1451 		return err;
1452 
1453 	if (!option_len &&
1454 	    (!tb[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE] ||
1455 	     !tb[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI])) {
1456 		NL_SET_ERR_MSG_MOD(extack,
1457 				   "Missing tunnel key gtp option pdu type or qfi");
1458 		return -EINVAL;
1459 	}
1460 
1461 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE])
1462 		sinfo->pdu_type =
1463 			nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE]);
1464 
1465 	if (tb[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI])
1466 		sinfo->qfi = nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI]);
1467 
1468 	return sizeof(*sinfo);
1469 }
1470 
1471 static int fl_set_pfcp_opt(const struct nlattr *nla, struct fl_flow_key *key,
1472 			   int depth, int option_len,
1473 			   struct netlink_ext_ack *extack)
1474 {
1475 	struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_PFCP_MAX + 1];
1476 	struct pfcp_metadata *md;
1477 	int err;
1478 
1479 	md = (struct pfcp_metadata *)&key->enc_opts.data[key->enc_opts.len];
1480 	memset(md, 0xff, sizeof(*md));
1481 
1482 	if (!depth)
1483 		return sizeof(*md);
1484 
1485 	if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_PFCP) {
1486 		NL_SET_ERR_MSG_MOD(extack, "Non-pfcp option type for mask");
1487 		return -EINVAL;
1488 	}
1489 
1490 	err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_PFCP_MAX, nla,
1491 			       pfcp_opt_policy, extack);
1492 	if (err < 0)
1493 		return err;
1494 
1495 	if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_PFCP_TYPE]) {
1496 		NL_SET_ERR_MSG_MOD(extack, "Missing tunnel key pfcp option type");
1497 		return -EINVAL;
1498 	}
1499 
1500 	if (tb[TCA_FLOWER_KEY_ENC_OPT_PFCP_TYPE])
1501 		md->type = nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_PFCP_TYPE]);
1502 
1503 	if (tb[TCA_FLOWER_KEY_ENC_OPT_PFCP_SEID])
1504 		md->seid = nla_get_be64(tb[TCA_FLOWER_KEY_ENC_OPT_PFCP_SEID]);
1505 
1506 	return sizeof(*md);
1507 }
1508 
1509 static int fl_set_enc_opt(struct nlattr **tb, struct fl_flow_key *key,
1510 			  struct fl_flow_key *mask,
1511 			  struct netlink_ext_ack *extack)
1512 {
1513 	const struct nlattr *nla_enc_key, *nla_opt_key, *nla_opt_msk = NULL;
1514 	int err, option_len, key_depth, msk_depth = 0;
1515 
1516 	err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS],
1517 					     TCA_FLOWER_KEY_ENC_OPTS_MAX,
1518 					     enc_opts_policy, extack);
1519 	if (err)
1520 		return err;
1521 
1522 	nla_enc_key = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS]);
1523 
1524 	if (tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]) {
1525 		err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK],
1526 						     TCA_FLOWER_KEY_ENC_OPTS_MAX,
1527 						     enc_opts_policy, extack);
1528 		if (err)
1529 			return err;
1530 
1531 		nla_opt_msk = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
1532 		msk_depth = nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
1533 		if (!nla_ok(nla_opt_msk, msk_depth)) {
1534 			NL_SET_ERR_MSG(extack, "Invalid nested attribute for masks");
1535 			return -EINVAL;
1536 		}
1537 	}
1538 
1539 	nla_for_each_attr(nla_opt_key, nla_enc_key,
1540 			  nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS]), key_depth) {
1541 		switch (nla_type(nla_opt_key)) {
1542 		case TCA_FLOWER_KEY_ENC_OPTS_GENEVE:
1543 			if (key->enc_opts.dst_opt_type &&
1544 			    key->enc_opts.dst_opt_type !=
1545 			    IP_TUNNEL_GENEVE_OPT_BIT) {
1546 				NL_SET_ERR_MSG(extack, "Duplicate type for geneve options");
1547 				return -EINVAL;
1548 			}
1549 			option_len = 0;
1550 			key->enc_opts.dst_opt_type = IP_TUNNEL_GENEVE_OPT_BIT;
1551 			option_len = fl_set_geneve_opt(nla_opt_key, key,
1552 						       key_depth, option_len,
1553 						       extack);
1554 			if (option_len < 0)
1555 				return option_len;
1556 
1557 			key->enc_opts.len += option_len;
1558 			/* At the same time we need to parse through the mask
1559 			 * in order to verify exact and mask attribute lengths.
1560 			 */
1561 			mask->enc_opts.dst_opt_type = IP_TUNNEL_GENEVE_OPT_BIT;
1562 			option_len = fl_set_geneve_opt(nla_opt_msk, mask,
1563 						       msk_depth, option_len,
1564 						       extack);
1565 			if (option_len < 0)
1566 				return option_len;
1567 
1568 			mask->enc_opts.len += option_len;
1569 			if (key->enc_opts.len != mask->enc_opts.len) {
1570 				NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1571 				return -EINVAL;
1572 			}
1573 			break;
1574 		case TCA_FLOWER_KEY_ENC_OPTS_VXLAN:
1575 			if (key->enc_opts.dst_opt_type) {
1576 				NL_SET_ERR_MSG(extack, "Duplicate type for vxlan options");
1577 				return -EINVAL;
1578 			}
1579 			option_len = 0;
1580 			key->enc_opts.dst_opt_type = IP_TUNNEL_VXLAN_OPT_BIT;
1581 			option_len = fl_set_vxlan_opt(nla_opt_key, key,
1582 						      key_depth, option_len,
1583 						      extack);
1584 			if (option_len < 0)
1585 				return option_len;
1586 
1587 			key->enc_opts.len += option_len;
1588 			/* At the same time we need to parse through the mask
1589 			 * in order to verify exact and mask attribute lengths.
1590 			 */
1591 			mask->enc_opts.dst_opt_type = IP_TUNNEL_VXLAN_OPT_BIT;
1592 			option_len = fl_set_vxlan_opt(nla_opt_msk, mask,
1593 						      msk_depth, option_len,
1594 						      extack);
1595 			if (option_len < 0)
1596 				return option_len;
1597 
1598 			mask->enc_opts.len += option_len;
1599 			if (key->enc_opts.len != mask->enc_opts.len) {
1600 				NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1601 				return -EINVAL;
1602 			}
1603 			break;
1604 		case TCA_FLOWER_KEY_ENC_OPTS_ERSPAN:
1605 			if (key->enc_opts.dst_opt_type) {
1606 				NL_SET_ERR_MSG(extack, "Duplicate type for erspan options");
1607 				return -EINVAL;
1608 			}
1609 			option_len = 0;
1610 			key->enc_opts.dst_opt_type = IP_TUNNEL_ERSPAN_OPT_BIT;
1611 			option_len = fl_set_erspan_opt(nla_opt_key, key,
1612 						       key_depth, option_len,
1613 						       extack);
1614 			if (option_len < 0)
1615 				return option_len;
1616 
1617 			key->enc_opts.len += option_len;
1618 			/* At the same time we need to parse through the mask
1619 			 * in order to verify exact and mask attribute lengths.
1620 			 */
1621 			mask->enc_opts.dst_opt_type = IP_TUNNEL_ERSPAN_OPT_BIT;
1622 			option_len = fl_set_erspan_opt(nla_opt_msk, mask,
1623 						       msk_depth, option_len,
1624 						       extack);
1625 			if (option_len < 0)
1626 				return option_len;
1627 
1628 			mask->enc_opts.len += option_len;
1629 			if (key->enc_opts.len != mask->enc_opts.len) {
1630 				NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1631 				return -EINVAL;
1632 			}
1633 			break;
1634 		case TCA_FLOWER_KEY_ENC_OPTS_GTP:
1635 			if (key->enc_opts.dst_opt_type) {
1636 				NL_SET_ERR_MSG_MOD(extack,
1637 						   "Duplicate type for gtp options");
1638 				return -EINVAL;
1639 			}
1640 			option_len = 0;
1641 			key->enc_opts.dst_opt_type = IP_TUNNEL_GTP_OPT_BIT;
1642 			option_len = fl_set_gtp_opt(nla_opt_key, key,
1643 						    key_depth, option_len,
1644 						    extack);
1645 			if (option_len < 0)
1646 				return option_len;
1647 
1648 			key->enc_opts.len += option_len;
1649 			/* At the same time we need to parse through the mask
1650 			 * in order to verify exact and mask attribute lengths.
1651 			 */
1652 			mask->enc_opts.dst_opt_type = IP_TUNNEL_GTP_OPT_BIT;
1653 			option_len = fl_set_gtp_opt(nla_opt_msk, mask,
1654 						    msk_depth, option_len,
1655 						    extack);
1656 			if (option_len < 0)
1657 				return option_len;
1658 
1659 			mask->enc_opts.len += option_len;
1660 			if (key->enc_opts.len != mask->enc_opts.len) {
1661 				NL_SET_ERR_MSG_MOD(extack,
1662 						   "Key and mask miss aligned");
1663 				return -EINVAL;
1664 			}
1665 			break;
1666 		case TCA_FLOWER_KEY_ENC_OPTS_PFCP:
1667 			if (key->enc_opts.dst_opt_type) {
1668 				NL_SET_ERR_MSG_MOD(extack, "Duplicate type for pfcp options");
1669 				return -EINVAL;
1670 			}
1671 			option_len = 0;
1672 			key->enc_opts.dst_opt_type = IP_TUNNEL_PFCP_OPT_BIT;
1673 			option_len = fl_set_pfcp_opt(nla_opt_key, key,
1674 						     key_depth, option_len,
1675 						     extack);
1676 			if (option_len < 0)
1677 				return option_len;
1678 
1679 			key->enc_opts.len += option_len;
1680 			/* At the same time we need to parse through the mask
1681 			 * in order to verify exact and mask attribute lengths.
1682 			 */
1683 			mask->enc_opts.dst_opt_type = IP_TUNNEL_PFCP_OPT_BIT;
1684 			option_len = fl_set_pfcp_opt(nla_opt_msk, mask,
1685 						     msk_depth, option_len,
1686 						     extack);
1687 			if (option_len < 0)
1688 				return option_len;
1689 
1690 			mask->enc_opts.len += option_len;
1691 			if (key->enc_opts.len != mask->enc_opts.len) {
1692 				NL_SET_ERR_MSG_MOD(extack, "Key and mask miss aligned");
1693 				return -EINVAL;
1694 			}
1695 			break;
1696 		default:
1697 			NL_SET_ERR_MSG(extack, "Unknown tunnel option type");
1698 			return -EINVAL;
1699 		}
1700 
1701 		if (!msk_depth)
1702 			continue;
1703 
1704 		if (!nla_ok(nla_opt_msk, msk_depth)) {
1705 			NL_SET_ERR_MSG(extack, "A mask attribute is invalid");
1706 			return -EINVAL;
1707 		}
1708 		nla_opt_msk = nla_next(nla_opt_msk, &msk_depth);
1709 	}
1710 
1711 	return 0;
1712 }
1713 
1714 static int fl_validate_ct_state(u16 state, struct nlattr *tb,
1715 				struct netlink_ext_ack *extack)
1716 {
1717 	if (state && !(state & TCA_FLOWER_KEY_CT_FLAGS_TRACKED)) {
1718 		NL_SET_ERR_MSG_ATTR(extack, tb,
1719 				    "no trk, so no other flag can be set");
1720 		return -EINVAL;
1721 	}
1722 
1723 	if (state & TCA_FLOWER_KEY_CT_FLAGS_NEW &&
1724 	    state & TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED) {
1725 		NL_SET_ERR_MSG_ATTR(extack, tb,
1726 				    "new and est are mutually exclusive");
1727 		return -EINVAL;
1728 	}
1729 
1730 	if (state & TCA_FLOWER_KEY_CT_FLAGS_INVALID &&
1731 	    state & ~(TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
1732 		      TCA_FLOWER_KEY_CT_FLAGS_INVALID)) {
1733 		NL_SET_ERR_MSG_ATTR(extack, tb,
1734 				    "when inv is set, only trk may be set");
1735 		return -EINVAL;
1736 	}
1737 
1738 	if (state & TCA_FLOWER_KEY_CT_FLAGS_NEW &&
1739 	    state & TCA_FLOWER_KEY_CT_FLAGS_REPLY) {
1740 		NL_SET_ERR_MSG_ATTR(extack, tb,
1741 				    "new and rpl are mutually exclusive");
1742 		return -EINVAL;
1743 	}
1744 
1745 	return 0;
1746 }
1747 
1748 static int fl_set_key_ct(struct nlattr **tb,
1749 			 struct flow_dissector_key_ct *key,
1750 			 struct flow_dissector_key_ct *mask,
1751 			 struct netlink_ext_ack *extack)
1752 {
1753 	if (tb[TCA_FLOWER_KEY_CT_STATE]) {
1754 		int err;
1755 
1756 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK)) {
1757 			NL_SET_ERR_MSG(extack, "Conntrack isn't enabled");
1758 			return -EOPNOTSUPP;
1759 		}
1760 		fl_set_key_val(tb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE,
1761 			       &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK,
1762 			       sizeof(key->ct_state));
1763 
1764 		err = fl_validate_ct_state(key->ct_state & mask->ct_state,
1765 					   tb[TCA_FLOWER_KEY_CT_STATE_MASK],
1766 					   extack);
1767 		if (err)
1768 			return err;
1769 
1770 	}
1771 	if (tb[TCA_FLOWER_KEY_CT_ZONE]) {
1772 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)) {
1773 			NL_SET_ERR_MSG(extack, "Conntrack zones isn't enabled");
1774 			return -EOPNOTSUPP;
1775 		}
1776 		fl_set_key_val(tb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE,
1777 			       &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK,
1778 			       sizeof(key->ct_zone));
1779 	}
1780 	if (tb[TCA_FLOWER_KEY_CT_MARK]) {
1781 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)) {
1782 			NL_SET_ERR_MSG(extack, "Conntrack mark isn't enabled");
1783 			return -EOPNOTSUPP;
1784 		}
1785 		fl_set_key_val(tb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK,
1786 			       &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK,
1787 			       sizeof(key->ct_mark));
1788 	}
1789 	if (tb[TCA_FLOWER_KEY_CT_LABELS]) {
1790 		if (!IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)) {
1791 			NL_SET_ERR_MSG(extack, "Conntrack labels aren't enabled");
1792 			return -EOPNOTSUPP;
1793 		}
1794 		fl_set_key_val(tb, key->ct_labels, TCA_FLOWER_KEY_CT_LABELS,
1795 			       mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK,
1796 			       sizeof(key->ct_labels));
1797 	}
1798 
1799 	return 0;
1800 }
1801 
1802 static bool is_vlan_key(struct nlattr *tb, __be16 *ethertype,
1803 			struct fl_flow_key *key, struct fl_flow_key *mask,
1804 			int vthresh)
1805 {
1806 	const bool good_num_of_vlans = key->num_of_vlans.num_of_vlans > vthresh;
1807 
1808 	if (!tb) {
1809 		*ethertype = 0;
1810 		return good_num_of_vlans;
1811 	}
1812 
1813 	*ethertype = nla_get_be16(tb);
1814 	if (good_num_of_vlans || eth_type_vlan(*ethertype))
1815 		return true;
1816 
1817 	key->basic.n_proto = *ethertype;
1818 	mask->basic.n_proto = cpu_to_be16(~0);
1819 	return false;
1820 }
1821 
1822 static void fl_set_key_cfm_md_level(struct nlattr **tb,
1823 				    struct fl_flow_key *key,
1824 				    struct fl_flow_key *mask,
1825 				    struct netlink_ext_ack *extack)
1826 {
1827 	u8 level;
1828 
1829 	if (!tb[TCA_FLOWER_KEY_CFM_MD_LEVEL])
1830 		return;
1831 
1832 	level = nla_get_u8(tb[TCA_FLOWER_KEY_CFM_MD_LEVEL]);
1833 	key->cfm.mdl_ver = FIELD_PREP(FLOW_DIS_CFM_MDL_MASK, level);
1834 	mask->cfm.mdl_ver = FLOW_DIS_CFM_MDL_MASK;
1835 }
1836 
1837 static void fl_set_key_cfm_opcode(struct nlattr **tb,
1838 				  struct fl_flow_key *key,
1839 				  struct fl_flow_key *mask,
1840 				  struct netlink_ext_ack *extack)
1841 {
1842 	fl_set_key_val(tb, &key->cfm.opcode, TCA_FLOWER_KEY_CFM_OPCODE,
1843 		       &mask->cfm.opcode, TCA_FLOWER_UNSPEC,
1844 		       sizeof(key->cfm.opcode));
1845 }
1846 
1847 static int fl_set_key_cfm(struct nlattr **tb,
1848 			  struct fl_flow_key *key,
1849 			  struct fl_flow_key *mask,
1850 			  struct netlink_ext_ack *extack)
1851 {
1852 	struct nlattr *nla_cfm_opt[TCA_FLOWER_KEY_CFM_OPT_MAX + 1];
1853 	int err;
1854 
1855 	if (!tb[TCA_FLOWER_KEY_CFM])
1856 		return 0;
1857 
1858 	err = nla_parse_nested(nla_cfm_opt, TCA_FLOWER_KEY_CFM_OPT_MAX,
1859 			       tb[TCA_FLOWER_KEY_CFM], cfm_opt_policy, extack);
1860 	if (err < 0)
1861 		return err;
1862 
1863 	fl_set_key_cfm_opcode(nla_cfm_opt, key, mask, extack);
1864 	fl_set_key_cfm_md_level(nla_cfm_opt, key, mask, extack);
1865 
1866 	return 0;
1867 }
1868 
1869 static int fl_set_key(struct net *net, struct nlattr *tca_opts,
1870 		      struct nlattr **tb, struct fl_flow_key *key,
1871 		      struct fl_flow_key *mask, struct netlink_ext_ack *extack)
1872 {
1873 	__be16 ethertype;
1874 	int ret = 0;
1875 
1876 	if (tb[TCA_FLOWER_INDEV]) {
1877 		int err = tcf_change_indev(net, tb[TCA_FLOWER_INDEV], extack);
1878 		if (err < 0)
1879 			return err;
1880 		key->meta.ingress_ifindex = err;
1881 		mask->meta.ingress_ifindex = 0xffffffff;
1882 	}
1883 
1884 	fl_set_key_val(tb, &key->meta.l2_miss, TCA_FLOWER_L2_MISS,
1885 		       &mask->meta.l2_miss, TCA_FLOWER_UNSPEC,
1886 		       sizeof(key->meta.l2_miss));
1887 
1888 	fl_set_key_val(tb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
1889 		       mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
1890 		       sizeof(key->eth.dst));
1891 	fl_set_key_val(tb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
1892 		       mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
1893 		       sizeof(key->eth.src));
1894 	fl_set_key_val(tb, &key->num_of_vlans,
1895 		       TCA_FLOWER_KEY_NUM_OF_VLANS,
1896 		       &mask->num_of_vlans,
1897 		       TCA_FLOWER_UNSPEC,
1898 		       sizeof(key->num_of_vlans));
1899 
1900 	if (is_vlan_key(tb[TCA_FLOWER_KEY_ETH_TYPE], &ethertype, key, mask, 0)) {
1901 		fl_set_key_vlan(tb, ethertype, TCA_FLOWER_KEY_VLAN_ID,
1902 				TCA_FLOWER_KEY_VLAN_PRIO,
1903 				TCA_FLOWER_KEY_VLAN_ETH_TYPE,
1904 				&key->vlan, &mask->vlan);
1905 
1906 		if (is_vlan_key(tb[TCA_FLOWER_KEY_VLAN_ETH_TYPE],
1907 				&ethertype, key, mask, 1)) {
1908 			fl_set_key_vlan(tb, ethertype,
1909 					TCA_FLOWER_KEY_CVLAN_ID,
1910 					TCA_FLOWER_KEY_CVLAN_PRIO,
1911 					TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
1912 					&key->cvlan, &mask->cvlan);
1913 			fl_set_key_val(tb, &key->basic.n_proto,
1914 				       TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
1915 				       &mask->basic.n_proto,
1916 				       TCA_FLOWER_UNSPEC,
1917 				       sizeof(key->basic.n_proto));
1918 		}
1919 	}
1920 
1921 	if (key->basic.n_proto == htons(ETH_P_PPP_SES))
1922 		fl_set_key_pppoe(tb, &key->pppoe, &mask->pppoe, key, mask);
1923 
1924 	if (key->basic.n_proto == htons(ETH_P_IP) ||
1925 	    key->basic.n_proto == htons(ETH_P_IPV6)) {
1926 		fl_set_key_val(tb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
1927 			       &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
1928 			       sizeof(key->basic.ip_proto));
1929 		fl_set_key_ip(tb, false, &key->ip, &mask->ip);
1930 	}
1931 
1932 	if (tb[TCA_FLOWER_KEY_IPV4_SRC] || tb[TCA_FLOWER_KEY_IPV4_DST]) {
1933 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1934 		mask->control.addr_type = ~0;
1935 		fl_set_key_val(tb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
1936 			       &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
1937 			       sizeof(key->ipv4.src));
1938 		fl_set_key_val(tb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
1939 			       &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
1940 			       sizeof(key->ipv4.dst));
1941 	} else if (tb[TCA_FLOWER_KEY_IPV6_SRC] || tb[TCA_FLOWER_KEY_IPV6_DST]) {
1942 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1943 		mask->control.addr_type = ~0;
1944 		fl_set_key_val(tb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
1945 			       &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
1946 			       sizeof(key->ipv6.src));
1947 		fl_set_key_val(tb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
1948 			       &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
1949 			       sizeof(key->ipv6.dst));
1950 	}
1951 
1952 	if (key->basic.ip_proto == IPPROTO_TCP) {
1953 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
1954 			       &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
1955 			       sizeof(key->tp.src));
1956 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
1957 			       &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
1958 			       sizeof(key->tp.dst));
1959 		fl_set_key_val(tb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
1960 			       &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
1961 			       sizeof(key->tcp.flags));
1962 	} else if (key->basic.ip_proto == IPPROTO_UDP) {
1963 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
1964 			       &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
1965 			       sizeof(key->tp.src));
1966 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
1967 			       &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
1968 			       sizeof(key->tp.dst));
1969 	} else if (key->basic.ip_proto == IPPROTO_SCTP) {
1970 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
1971 			       &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
1972 			       sizeof(key->tp.src));
1973 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
1974 			       &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
1975 			       sizeof(key->tp.dst));
1976 	} else if (key->basic.n_proto == htons(ETH_P_IP) &&
1977 		   key->basic.ip_proto == IPPROTO_ICMP) {
1978 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV4_TYPE,
1979 			       &mask->icmp.type,
1980 			       TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
1981 			       sizeof(key->icmp.type));
1982 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV4_CODE,
1983 			       &mask->icmp.code,
1984 			       TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
1985 			       sizeof(key->icmp.code));
1986 	} else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
1987 		   key->basic.ip_proto == IPPROTO_ICMPV6) {
1988 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV6_TYPE,
1989 			       &mask->icmp.type,
1990 			       TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
1991 			       sizeof(key->icmp.type));
1992 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV6_CODE,
1993 			       &mask->icmp.code,
1994 			       TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
1995 			       sizeof(key->icmp.code));
1996 	} else if (key->basic.n_proto == htons(ETH_P_MPLS_UC) ||
1997 		   key->basic.n_proto == htons(ETH_P_MPLS_MC)) {
1998 		ret = fl_set_key_mpls(tb, &key->mpls, &mask->mpls, extack);
1999 		if (ret)
2000 			return ret;
2001 	} else if (key->basic.n_proto == htons(ETH_P_ARP) ||
2002 		   key->basic.n_proto == htons(ETH_P_RARP)) {
2003 		fl_set_key_val(tb, &key->arp.sip, TCA_FLOWER_KEY_ARP_SIP,
2004 			       &mask->arp.sip, TCA_FLOWER_KEY_ARP_SIP_MASK,
2005 			       sizeof(key->arp.sip));
2006 		fl_set_key_val(tb, &key->arp.tip, TCA_FLOWER_KEY_ARP_TIP,
2007 			       &mask->arp.tip, TCA_FLOWER_KEY_ARP_TIP_MASK,
2008 			       sizeof(key->arp.tip));
2009 		fl_set_key_val(tb, &key->arp.op, TCA_FLOWER_KEY_ARP_OP,
2010 			       &mask->arp.op, TCA_FLOWER_KEY_ARP_OP_MASK,
2011 			       sizeof(key->arp.op));
2012 		fl_set_key_val(tb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
2013 			       mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
2014 			       sizeof(key->arp.sha));
2015 		fl_set_key_val(tb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
2016 			       mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
2017 			       sizeof(key->arp.tha));
2018 	} else if (key->basic.ip_proto == IPPROTO_L2TP) {
2019 		fl_set_key_val(tb, &key->l2tpv3.session_id,
2020 			       TCA_FLOWER_KEY_L2TPV3_SID,
2021 			       &mask->l2tpv3.session_id, TCA_FLOWER_UNSPEC,
2022 			       sizeof(key->l2tpv3.session_id));
2023 	} else if (key->basic.n_proto  == htons(ETH_P_CFM)) {
2024 		ret = fl_set_key_cfm(tb, key, mask, extack);
2025 		if (ret)
2026 			return ret;
2027 	}
2028 
2029 	if (key->basic.ip_proto == IPPROTO_TCP ||
2030 	    key->basic.ip_proto == IPPROTO_UDP ||
2031 	    key->basic.ip_proto == IPPROTO_SCTP) {
2032 		ret = fl_set_key_port_range(tb, key, mask, extack);
2033 		if (ret)
2034 			return ret;
2035 	}
2036 
2037 	if (tb[TCA_FLOWER_KEY_SPI]) {
2038 		ret = fl_set_key_spi(tb, key, mask, extack);
2039 		if (ret)
2040 			return ret;
2041 	}
2042 
2043 	if (tb[TCA_FLOWER_KEY_ENC_IPV4_SRC] ||
2044 	    tb[TCA_FLOWER_KEY_ENC_IPV4_DST]) {
2045 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2046 		mask->enc_control.addr_type = ~0;
2047 		fl_set_key_val(tb, &key->enc_ipv4.src,
2048 			       TCA_FLOWER_KEY_ENC_IPV4_SRC,
2049 			       &mask->enc_ipv4.src,
2050 			       TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
2051 			       sizeof(key->enc_ipv4.src));
2052 		fl_set_key_val(tb, &key->enc_ipv4.dst,
2053 			       TCA_FLOWER_KEY_ENC_IPV4_DST,
2054 			       &mask->enc_ipv4.dst,
2055 			       TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
2056 			       sizeof(key->enc_ipv4.dst));
2057 	}
2058 
2059 	if (tb[TCA_FLOWER_KEY_ENC_IPV6_SRC] ||
2060 	    tb[TCA_FLOWER_KEY_ENC_IPV6_DST]) {
2061 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2062 		mask->enc_control.addr_type = ~0;
2063 		fl_set_key_val(tb, &key->enc_ipv6.src,
2064 			       TCA_FLOWER_KEY_ENC_IPV6_SRC,
2065 			       &mask->enc_ipv6.src,
2066 			       TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
2067 			       sizeof(key->enc_ipv6.src));
2068 		fl_set_key_val(tb, &key->enc_ipv6.dst,
2069 			       TCA_FLOWER_KEY_ENC_IPV6_DST,
2070 			       &mask->enc_ipv6.dst,
2071 			       TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
2072 			       sizeof(key->enc_ipv6.dst));
2073 	}
2074 
2075 	fl_set_key_val(tb, &key->enc_key_id.keyid, TCA_FLOWER_KEY_ENC_KEY_ID,
2076 		       &mask->enc_key_id.keyid, TCA_FLOWER_UNSPEC,
2077 		       sizeof(key->enc_key_id.keyid));
2078 
2079 	fl_set_key_val(tb, &key->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
2080 		       &mask->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
2081 		       sizeof(key->enc_tp.src));
2082 
2083 	fl_set_key_val(tb, &key->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
2084 		       &mask->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
2085 		       sizeof(key->enc_tp.dst));
2086 
2087 	fl_set_key_ip(tb, true, &key->enc_ip, &mask->enc_ip);
2088 
2089 	fl_set_key_val(tb, &key->hash.hash, TCA_FLOWER_KEY_HASH,
2090 		       &mask->hash.hash, TCA_FLOWER_KEY_HASH_MASK,
2091 		       sizeof(key->hash.hash));
2092 
2093 	if (tb[TCA_FLOWER_KEY_ENC_OPTS]) {
2094 		ret = fl_set_enc_opt(tb, key, mask, extack);
2095 		if (ret)
2096 			return ret;
2097 	}
2098 
2099 	ret = fl_set_key_ct(tb, &key->ct, &mask->ct, extack);
2100 	if (ret)
2101 		return ret;
2102 
2103 	if (tb[TCA_FLOWER_KEY_FLAGS]) {
2104 		ret = fl_set_key_flags(tca_opts, tb, false,
2105 				       &key->control.flags,
2106 				       &mask->control.flags, extack);
2107 		if (ret)
2108 			return ret;
2109 	}
2110 
2111 	if (tb[TCA_FLOWER_KEY_ENC_FLAGS])
2112 		ret = fl_set_key_flags(tca_opts, tb, true,
2113 				       &key->enc_control.flags,
2114 				       &mask->enc_control.flags, extack);
2115 
2116 	return ret;
2117 }
2118 
2119 static void fl_mask_copy(struct fl_flow_mask *dst,
2120 			 struct fl_flow_mask *src)
2121 {
2122 	const void *psrc = fl_key_get_start(&src->key, src);
2123 	void *pdst = fl_key_get_start(&dst->key, src);
2124 
2125 	memcpy(pdst, psrc, fl_mask_range(src));
2126 	dst->range = src->range;
2127 }
2128 
2129 static const struct rhashtable_params fl_ht_params = {
2130 	.key_offset = offsetof(struct cls_fl_filter, mkey), /* base offset */
2131 	.head_offset = offsetof(struct cls_fl_filter, ht_node),
2132 	.automatic_shrinking = true,
2133 };
2134 
2135 static int fl_init_mask_hashtable(struct fl_flow_mask *mask)
2136 {
2137 	mask->filter_ht_params = fl_ht_params;
2138 	mask->filter_ht_params.key_len = fl_mask_range(mask);
2139 	mask->filter_ht_params.key_offset += mask->range.start;
2140 
2141 	return rhashtable_init(&mask->ht, &mask->filter_ht_params);
2142 }
2143 
2144 #define FL_KEY_MEMBER_OFFSET(member) offsetof(struct fl_flow_key, member)
2145 #define FL_KEY_MEMBER_SIZE(member) sizeof_field(struct fl_flow_key, member)
2146 
2147 #define FL_KEY_IS_MASKED(mask, member)						\
2148 	memchr_inv(((char *)mask) + FL_KEY_MEMBER_OFFSET(member),		\
2149 		   0, FL_KEY_MEMBER_SIZE(member))				\
2150 
2151 #define FL_KEY_SET(keys, cnt, id, member)					\
2152 	do {									\
2153 		keys[cnt].key_id = id;						\
2154 		keys[cnt].offset = FL_KEY_MEMBER_OFFSET(member);		\
2155 		cnt++;								\
2156 	} while(0);
2157 
2158 #define FL_KEY_SET_IF_MASKED(mask, keys, cnt, id, member)			\
2159 	do {									\
2160 		if (FL_KEY_IS_MASKED(mask, member))				\
2161 			FL_KEY_SET(keys, cnt, id, member);			\
2162 	} while(0);
2163 
2164 static void fl_init_dissector(struct flow_dissector *dissector,
2165 			      struct fl_flow_key *mask)
2166 {
2167 	struct flow_dissector_key keys[FLOW_DISSECTOR_KEY_MAX];
2168 	size_t cnt = 0;
2169 
2170 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2171 			     FLOW_DISSECTOR_KEY_META, meta);
2172 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_CONTROL, control);
2173 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_BASIC, basic);
2174 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2175 			     FLOW_DISSECTOR_KEY_ETH_ADDRS, eth);
2176 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2177 			     FLOW_DISSECTOR_KEY_IPV4_ADDRS, ipv4);
2178 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2179 			     FLOW_DISSECTOR_KEY_IPV6_ADDRS, ipv6);
2180 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2181 			     FLOW_DISSECTOR_KEY_PORTS, tp);
2182 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2183 			     FLOW_DISSECTOR_KEY_PORTS_RANGE, tp_range);
2184 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2185 			     FLOW_DISSECTOR_KEY_IP, ip);
2186 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2187 			     FLOW_DISSECTOR_KEY_TCP, tcp);
2188 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2189 			     FLOW_DISSECTOR_KEY_ICMP, icmp);
2190 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2191 			     FLOW_DISSECTOR_KEY_ARP, arp);
2192 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2193 			     FLOW_DISSECTOR_KEY_MPLS, mpls);
2194 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2195 			     FLOW_DISSECTOR_KEY_VLAN, vlan);
2196 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2197 			     FLOW_DISSECTOR_KEY_CVLAN, cvlan);
2198 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2199 			     FLOW_DISSECTOR_KEY_ENC_KEYID, enc_key_id);
2200 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2201 			     FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, enc_ipv4);
2202 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2203 			     FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, enc_ipv6);
2204 	if (FL_KEY_IS_MASKED(mask, enc_ipv4) ||
2205 	    FL_KEY_IS_MASKED(mask, enc_ipv6) ||
2206 	    FL_KEY_IS_MASKED(mask, enc_control))
2207 		FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_ENC_CONTROL,
2208 			   enc_control);
2209 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2210 			     FLOW_DISSECTOR_KEY_ENC_PORTS, enc_tp);
2211 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2212 			     FLOW_DISSECTOR_KEY_ENC_IP, enc_ip);
2213 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2214 			     FLOW_DISSECTOR_KEY_ENC_OPTS, enc_opts);
2215 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2216 			     FLOW_DISSECTOR_KEY_CT, ct);
2217 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2218 			     FLOW_DISSECTOR_KEY_HASH, hash);
2219 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2220 			     FLOW_DISSECTOR_KEY_NUM_OF_VLANS, num_of_vlans);
2221 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2222 			     FLOW_DISSECTOR_KEY_PPPOE, pppoe);
2223 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2224 			     FLOW_DISSECTOR_KEY_L2TPV3, l2tpv3);
2225 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2226 			     FLOW_DISSECTOR_KEY_IPSEC, ipsec);
2227 	FL_KEY_SET_IF_MASKED(mask, keys, cnt,
2228 			     FLOW_DISSECTOR_KEY_CFM, cfm);
2229 
2230 	skb_flow_dissector_init(dissector, keys, cnt);
2231 }
2232 
2233 static struct fl_flow_mask *fl_create_new_mask(struct cls_fl_head *head,
2234 					       struct fl_flow_mask *mask)
2235 {
2236 	struct fl_flow_mask *newmask;
2237 	int err;
2238 
2239 	newmask = kzalloc(sizeof(*newmask), GFP_KERNEL);
2240 	if (!newmask)
2241 		return ERR_PTR(-ENOMEM);
2242 
2243 	fl_mask_copy(newmask, mask);
2244 
2245 	if ((newmask->key.tp_range.tp_min.dst &&
2246 	     newmask->key.tp_range.tp_max.dst) ||
2247 	    (newmask->key.tp_range.tp_min.src &&
2248 	     newmask->key.tp_range.tp_max.src))
2249 		newmask->flags |= TCA_FLOWER_MASK_FLAGS_RANGE;
2250 
2251 	err = fl_init_mask_hashtable(newmask);
2252 	if (err)
2253 		goto errout_free;
2254 
2255 	fl_init_dissector(&newmask->dissector, &newmask->key);
2256 
2257 	INIT_LIST_HEAD_RCU(&newmask->filters);
2258 
2259 	refcount_set(&newmask->refcnt, 1);
2260 	err = rhashtable_replace_fast(&head->ht, &mask->ht_node,
2261 				      &newmask->ht_node, mask_ht_params);
2262 	if (err)
2263 		goto errout_destroy;
2264 
2265 	spin_lock(&head->masks_lock);
2266 	list_add_tail_rcu(&newmask->list, &head->masks);
2267 	spin_unlock(&head->masks_lock);
2268 
2269 	return newmask;
2270 
2271 errout_destroy:
2272 	rhashtable_destroy(&newmask->ht);
2273 errout_free:
2274 	kfree(newmask);
2275 
2276 	return ERR_PTR(err);
2277 }
2278 
2279 static int fl_check_assign_mask(struct cls_fl_head *head,
2280 				struct cls_fl_filter *fnew,
2281 				struct cls_fl_filter *fold,
2282 				struct fl_flow_mask *mask)
2283 {
2284 	struct fl_flow_mask *newmask;
2285 	int ret = 0;
2286 
2287 	rcu_read_lock();
2288 
2289 	/* Insert mask as temporary node to prevent concurrent creation of mask
2290 	 * with same key. Any concurrent lookups with same key will return
2291 	 * -EAGAIN because mask's refcnt is zero.
2292 	 */
2293 	fnew->mask = rhashtable_lookup_get_insert_fast(&head->ht,
2294 						       &mask->ht_node,
2295 						       mask_ht_params);
2296 	if (!fnew->mask) {
2297 		rcu_read_unlock();
2298 
2299 		if (fold) {
2300 			ret = -EINVAL;
2301 			goto errout_cleanup;
2302 		}
2303 
2304 		newmask = fl_create_new_mask(head, mask);
2305 		if (IS_ERR(newmask)) {
2306 			ret = PTR_ERR(newmask);
2307 			goto errout_cleanup;
2308 		}
2309 
2310 		fnew->mask = newmask;
2311 		return 0;
2312 	} else if (IS_ERR(fnew->mask)) {
2313 		ret = PTR_ERR(fnew->mask);
2314 	} else if (fold && fold->mask != fnew->mask) {
2315 		ret = -EINVAL;
2316 	} else if (!refcount_inc_not_zero(&fnew->mask->refcnt)) {
2317 		/* Mask was deleted concurrently, try again */
2318 		ret = -EAGAIN;
2319 	}
2320 	rcu_read_unlock();
2321 	return ret;
2322 
2323 errout_cleanup:
2324 	rhashtable_remove_fast(&head->ht, &mask->ht_node,
2325 			       mask_ht_params);
2326 	return ret;
2327 }
2328 
2329 static bool fl_needs_tc_skb_ext(const struct fl_flow_key *mask)
2330 {
2331 	return mask->meta.l2_miss;
2332 }
2333 
2334 static int fl_ht_insert_unique(struct cls_fl_filter *fnew,
2335 			       struct cls_fl_filter *fold,
2336 			       bool *in_ht)
2337 {
2338 	struct fl_flow_mask *mask = fnew->mask;
2339 	int err;
2340 
2341 	err = rhashtable_lookup_insert_fast(&mask->ht,
2342 					    &fnew->ht_node,
2343 					    mask->filter_ht_params);
2344 	if (err) {
2345 		*in_ht = false;
2346 		/* It is okay if filter with same key exists when
2347 		 * overwriting.
2348 		 */
2349 		return fold && err == -EEXIST ? 0 : err;
2350 	}
2351 
2352 	*in_ht = true;
2353 	return 0;
2354 }
2355 
2356 static int fl_change(struct net *net, struct sk_buff *in_skb,
2357 		     struct tcf_proto *tp, unsigned long base,
2358 		     u32 handle, struct nlattr **tca,
2359 		     void **arg, u32 flags,
2360 		     struct netlink_ext_ack *extack)
2361 {
2362 	struct cls_fl_head *head = fl_head_dereference(tp);
2363 	bool rtnl_held = !(flags & TCA_ACT_FLAGS_NO_RTNL);
2364 	struct nlattr *tca_opts = tca[TCA_OPTIONS];
2365 	struct cls_fl_filter *fold = *arg;
2366 	bool bound_to_filter = false;
2367 	struct cls_fl_filter *fnew;
2368 	struct fl_flow_mask *mask;
2369 	struct nlattr **tb;
2370 	bool in_ht;
2371 	int err;
2372 
2373 	if (!tca_opts) {
2374 		err = -EINVAL;
2375 		goto errout_fold;
2376 	}
2377 
2378 	mask = kzalloc(sizeof(struct fl_flow_mask), GFP_KERNEL);
2379 	if (!mask) {
2380 		err = -ENOBUFS;
2381 		goto errout_fold;
2382 	}
2383 
2384 	tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
2385 	if (!tb) {
2386 		err = -ENOBUFS;
2387 		goto errout_mask_alloc;
2388 	}
2389 
2390 	err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
2391 					  tca_opts, fl_policy, NULL);
2392 	if (err < 0)
2393 		goto errout_tb;
2394 
2395 	if (fold && handle && fold->handle != handle) {
2396 		err = -EINVAL;
2397 		goto errout_tb;
2398 	}
2399 
2400 	fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
2401 	if (!fnew) {
2402 		err = -ENOBUFS;
2403 		goto errout_tb;
2404 	}
2405 	INIT_LIST_HEAD(&fnew->hw_list);
2406 	refcount_set(&fnew->refcnt, 1);
2407 
2408 	if (tb[TCA_FLOWER_FLAGS]) {
2409 		fnew->flags = nla_get_u32(tb[TCA_FLOWER_FLAGS]);
2410 
2411 		if (!tc_flags_valid(fnew->flags)) {
2412 			kfree(fnew);
2413 			err = -EINVAL;
2414 			goto errout_tb;
2415 		}
2416 	}
2417 
2418 	if (!fold) {
2419 		spin_lock(&tp->lock);
2420 		if (!handle) {
2421 			handle = 1;
2422 			err = idr_alloc_u32(&head->handle_idr, NULL, &handle,
2423 					    INT_MAX, GFP_ATOMIC);
2424 		} else {
2425 			err = idr_alloc_u32(&head->handle_idr, NULL, &handle,
2426 					    handle, GFP_ATOMIC);
2427 
2428 			/* Filter with specified handle was concurrently
2429 			 * inserted after initial check in cls_api. This is not
2430 			 * necessarily an error if NLM_F_EXCL is not set in
2431 			 * message flags. Returning EAGAIN will cause cls_api to
2432 			 * try to update concurrently inserted rule.
2433 			 */
2434 			if (err == -ENOSPC)
2435 				err = -EAGAIN;
2436 		}
2437 		spin_unlock(&tp->lock);
2438 
2439 		if (err) {
2440 			kfree(fnew);
2441 			goto errout_tb;
2442 		}
2443 	}
2444 	fnew->handle = handle;
2445 
2446 	err = tcf_exts_init_ex(&fnew->exts, net, TCA_FLOWER_ACT, 0, tp, handle,
2447 			       !tc_skip_hw(fnew->flags));
2448 	if (err < 0)
2449 		goto errout_idr;
2450 
2451 	err = tcf_exts_validate_ex(net, tp, tb, tca[TCA_RATE],
2452 				   &fnew->exts, flags, fnew->flags,
2453 				   extack);
2454 	if (err < 0)
2455 		goto errout_idr;
2456 
2457 	if (tb[TCA_FLOWER_CLASSID]) {
2458 		fnew->res.classid = nla_get_u32(tb[TCA_FLOWER_CLASSID]);
2459 		if (flags & TCA_ACT_FLAGS_NO_RTNL)
2460 			rtnl_lock();
2461 		tcf_bind_filter(tp, &fnew->res, base);
2462 		if (flags & TCA_ACT_FLAGS_NO_RTNL)
2463 			rtnl_unlock();
2464 		bound_to_filter = true;
2465 	}
2466 
2467 	err = fl_set_key(net, tca_opts, tb, &fnew->key, &mask->key, extack);
2468 	if (err)
2469 		goto unbind_filter;
2470 
2471 	fl_mask_update_range(mask);
2472 	fl_set_masked_key(&fnew->mkey, &fnew->key, mask);
2473 
2474 	if (!fl_mask_fits_tmplt(tp->chain->tmplt_priv, mask)) {
2475 		NL_SET_ERR_MSG_MOD(extack, "Mask does not fit the template");
2476 		err = -EINVAL;
2477 		goto unbind_filter;
2478 	}
2479 
2480 	/* Enable tc skb extension if filter matches on data extracted from
2481 	 * this extension.
2482 	 */
2483 	if (fl_needs_tc_skb_ext(&mask->key)) {
2484 		fnew->needs_tc_skb_ext = 1;
2485 		tc_skb_ext_tc_enable();
2486 	}
2487 
2488 	err = fl_check_assign_mask(head, fnew, fold, mask);
2489 	if (err)
2490 		goto unbind_filter;
2491 
2492 	err = fl_ht_insert_unique(fnew, fold, &in_ht);
2493 	if (err)
2494 		goto errout_mask;
2495 
2496 	if (!tc_skip_hw(fnew->flags)) {
2497 		err = fl_hw_replace_filter(tp, fnew, rtnl_held, extack);
2498 		if (err)
2499 			goto errout_ht;
2500 	}
2501 
2502 	if (!tc_in_hw(fnew->flags))
2503 		fnew->flags |= TCA_CLS_FLAGS_NOT_IN_HW;
2504 
2505 	spin_lock(&tp->lock);
2506 
2507 	/* tp was deleted concurrently. -EAGAIN will cause caller to lookup
2508 	 * proto again or create new one, if necessary.
2509 	 */
2510 	if (tp->deleting) {
2511 		err = -EAGAIN;
2512 		goto errout_hw;
2513 	}
2514 
2515 	if (fold) {
2516 		/* Fold filter was deleted concurrently. Retry lookup. */
2517 		if (fold->deleted) {
2518 			err = -EAGAIN;
2519 			goto errout_hw;
2520 		}
2521 
2522 		fnew->handle = handle;
2523 
2524 		if (!in_ht) {
2525 			struct rhashtable_params params =
2526 				fnew->mask->filter_ht_params;
2527 
2528 			err = rhashtable_insert_fast(&fnew->mask->ht,
2529 						     &fnew->ht_node,
2530 						     params);
2531 			if (err)
2532 				goto errout_hw;
2533 			in_ht = true;
2534 		}
2535 
2536 		refcount_inc(&fnew->refcnt);
2537 		rhashtable_remove_fast(&fold->mask->ht,
2538 				       &fold->ht_node,
2539 				       fold->mask->filter_ht_params);
2540 		idr_replace(&head->handle_idr, fnew, fnew->handle);
2541 		list_replace_rcu(&fold->list, &fnew->list);
2542 		fold->deleted = true;
2543 
2544 		spin_unlock(&tp->lock);
2545 
2546 		fl_mask_put(head, fold->mask);
2547 		if (!tc_skip_hw(fold->flags))
2548 			fl_hw_destroy_filter(tp, fold, rtnl_held, NULL);
2549 		tcf_unbind_filter(tp, &fold->res);
2550 		/* Caller holds reference to fold, so refcnt is always > 0
2551 		 * after this.
2552 		 */
2553 		refcount_dec(&fold->refcnt);
2554 		__fl_put(fold);
2555 	} else {
2556 		idr_replace(&head->handle_idr, fnew, fnew->handle);
2557 
2558 		refcount_inc(&fnew->refcnt);
2559 		list_add_tail_rcu(&fnew->list, &fnew->mask->filters);
2560 		spin_unlock(&tp->lock);
2561 	}
2562 
2563 	*arg = fnew;
2564 
2565 	kfree(tb);
2566 	tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work);
2567 	return 0;
2568 
2569 errout_ht:
2570 	spin_lock(&tp->lock);
2571 errout_hw:
2572 	fnew->deleted = true;
2573 	spin_unlock(&tp->lock);
2574 	if (!tc_skip_hw(fnew->flags))
2575 		fl_hw_destroy_filter(tp, fnew, rtnl_held, NULL);
2576 	if (in_ht)
2577 		rhashtable_remove_fast(&fnew->mask->ht, &fnew->ht_node,
2578 				       fnew->mask->filter_ht_params);
2579 errout_mask:
2580 	fl_mask_put(head, fnew->mask);
2581 
2582 unbind_filter:
2583 	if (bound_to_filter) {
2584 		if (flags & TCA_ACT_FLAGS_NO_RTNL)
2585 			rtnl_lock();
2586 		tcf_unbind_filter(tp, &fnew->res);
2587 		if (flags & TCA_ACT_FLAGS_NO_RTNL)
2588 			rtnl_unlock();
2589 	}
2590 
2591 errout_idr:
2592 	if (!fold) {
2593 		spin_lock(&tp->lock);
2594 		idr_remove(&head->handle_idr, fnew->handle);
2595 		spin_unlock(&tp->lock);
2596 	}
2597 	__fl_put(fnew);
2598 errout_tb:
2599 	kfree(tb);
2600 errout_mask_alloc:
2601 	tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work);
2602 errout_fold:
2603 	if (fold)
2604 		__fl_put(fold);
2605 	return err;
2606 }
2607 
2608 static int fl_delete(struct tcf_proto *tp, void *arg, bool *last,
2609 		     bool rtnl_held, struct netlink_ext_ack *extack)
2610 {
2611 	struct cls_fl_head *head = fl_head_dereference(tp);
2612 	struct cls_fl_filter *f = arg;
2613 	bool last_on_mask;
2614 	int err = 0;
2615 
2616 	err = __fl_delete(tp, f, &last_on_mask, rtnl_held, extack);
2617 	*last = list_empty(&head->masks);
2618 	__fl_put(f);
2619 
2620 	return err;
2621 }
2622 
2623 static void fl_walk(struct tcf_proto *tp, struct tcf_walker *arg,
2624 		    bool rtnl_held)
2625 {
2626 	struct cls_fl_head *head = fl_head_dereference(tp);
2627 	unsigned long id = arg->cookie, tmp;
2628 	struct cls_fl_filter *f;
2629 
2630 	arg->count = arg->skip;
2631 
2632 	rcu_read_lock();
2633 	idr_for_each_entry_continue_ul(&head->handle_idr, f, tmp, id) {
2634 		/* don't return filters that are being deleted */
2635 		if (!f || !refcount_inc_not_zero(&f->refcnt))
2636 			continue;
2637 		rcu_read_unlock();
2638 
2639 		if (arg->fn(tp, f, arg) < 0) {
2640 			__fl_put(f);
2641 			arg->stop = 1;
2642 			rcu_read_lock();
2643 			break;
2644 		}
2645 		__fl_put(f);
2646 		arg->count++;
2647 		rcu_read_lock();
2648 	}
2649 	rcu_read_unlock();
2650 	arg->cookie = id;
2651 }
2652 
2653 static struct cls_fl_filter *
2654 fl_get_next_hw_filter(struct tcf_proto *tp, struct cls_fl_filter *f, bool add)
2655 {
2656 	struct cls_fl_head *head = fl_head_dereference(tp);
2657 
2658 	spin_lock(&tp->lock);
2659 	if (list_empty(&head->hw_filters)) {
2660 		spin_unlock(&tp->lock);
2661 		return NULL;
2662 	}
2663 
2664 	if (!f)
2665 		f = list_entry(&head->hw_filters, struct cls_fl_filter,
2666 			       hw_list);
2667 	list_for_each_entry_continue(f, &head->hw_filters, hw_list) {
2668 		if (!(add && f->deleted) && refcount_inc_not_zero(&f->refcnt)) {
2669 			spin_unlock(&tp->lock);
2670 			return f;
2671 		}
2672 	}
2673 
2674 	spin_unlock(&tp->lock);
2675 	return NULL;
2676 }
2677 
2678 static int fl_reoffload(struct tcf_proto *tp, bool add, flow_setup_cb_t *cb,
2679 			void *cb_priv, struct netlink_ext_ack *extack)
2680 {
2681 	struct tcf_block *block = tp->chain->block;
2682 	struct flow_cls_offload cls_flower = {};
2683 	struct cls_fl_filter *f = NULL;
2684 	int err;
2685 
2686 	/* hw_filters list can only be changed by hw offload functions after
2687 	 * obtaining rtnl lock. Make sure it is not changed while reoffload is
2688 	 * iterating it.
2689 	 */
2690 	ASSERT_RTNL();
2691 
2692 	while ((f = fl_get_next_hw_filter(tp, f, add))) {
2693 		cls_flower.rule =
2694 			flow_rule_alloc(tcf_exts_num_actions(&f->exts));
2695 		if (!cls_flower.rule) {
2696 			__fl_put(f);
2697 			return -ENOMEM;
2698 		}
2699 
2700 		tc_cls_common_offload_init(&cls_flower.common, tp, f->flags,
2701 					   extack);
2702 		cls_flower.command = add ?
2703 			FLOW_CLS_REPLACE : FLOW_CLS_DESTROY;
2704 		cls_flower.cookie = (unsigned long)f;
2705 		cls_flower.rule->match.dissector = &f->mask->dissector;
2706 		cls_flower.rule->match.mask = &f->mask->key;
2707 		cls_flower.rule->match.key = &f->mkey;
2708 
2709 		err = tc_setup_offload_action(&cls_flower.rule->action, &f->exts,
2710 					      cls_flower.common.extack);
2711 		if (err) {
2712 			kfree(cls_flower.rule);
2713 			if (tc_skip_sw(f->flags)) {
2714 				__fl_put(f);
2715 				return err;
2716 			}
2717 			goto next_flow;
2718 		}
2719 
2720 		cls_flower.classid = f->res.classid;
2721 
2722 		err = tc_setup_cb_reoffload(block, tp, add, cb,
2723 					    TC_SETUP_CLSFLOWER, &cls_flower,
2724 					    cb_priv, &f->flags,
2725 					    &f->in_hw_count);
2726 		tc_cleanup_offload_action(&cls_flower.rule->action);
2727 		kfree(cls_flower.rule);
2728 
2729 		if (err) {
2730 			__fl_put(f);
2731 			return err;
2732 		}
2733 next_flow:
2734 		__fl_put(f);
2735 	}
2736 
2737 	return 0;
2738 }
2739 
2740 static void fl_hw_add(struct tcf_proto *tp, void *type_data)
2741 {
2742 	struct flow_cls_offload *cls_flower = type_data;
2743 	struct cls_fl_filter *f =
2744 		(struct cls_fl_filter *) cls_flower->cookie;
2745 	struct cls_fl_head *head = fl_head_dereference(tp);
2746 
2747 	spin_lock(&tp->lock);
2748 	list_add(&f->hw_list, &head->hw_filters);
2749 	spin_unlock(&tp->lock);
2750 }
2751 
2752 static void fl_hw_del(struct tcf_proto *tp, void *type_data)
2753 {
2754 	struct flow_cls_offload *cls_flower = type_data;
2755 	struct cls_fl_filter *f =
2756 		(struct cls_fl_filter *) cls_flower->cookie;
2757 
2758 	spin_lock(&tp->lock);
2759 	if (!list_empty(&f->hw_list))
2760 		list_del_init(&f->hw_list);
2761 	spin_unlock(&tp->lock);
2762 }
2763 
2764 static int fl_hw_create_tmplt(struct tcf_chain *chain,
2765 			      struct fl_flow_tmplt *tmplt)
2766 {
2767 	struct flow_cls_offload cls_flower = {};
2768 	struct tcf_block *block = chain->block;
2769 
2770 	cls_flower.rule = flow_rule_alloc(0);
2771 	if (!cls_flower.rule)
2772 		return -ENOMEM;
2773 
2774 	cls_flower.common.chain_index = chain->index;
2775 	cls_flower.command = FLOW_CLS_TMPLT_CREATE;
2776 	cls_flower.cookie = (unsigned long) tmplt;
2777 	cls_flower.rule->match.dissector = &tmplt->dissector;
2778 	cls_flower.rule->match.mask = &tmplt->mask;
2779 	cls_flower.rule->match.key = &tmplt->dummy_key;
2780 
2781 	/* We don't care if driver (any of them) fails to handle this
2782 	 * call. It serves just as a hint for it.
2783 	 */
2784 	tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, true);
2785 	kfree(cls_flower.rule);
2786 
2787 	return 0;
2788 }
2789 
2790 static void fl_hw_destroy_tmplt(struct tcf_chain *chain,
2791 				struct fl_flow_tmplt *tmplt)
2792 {
2793 	struct flow_cls_offload cls_flower = {};
2794 	struct tcf_block *block = chain->block;
2795 
2796 	cls_flower.common.chain_index = chain->index;
2797 	cls_flower.command = FLOW_CLS_TMPLT_DESTROY;
2798 	cls_flower.cookie = (unsigned long) tmplt;
2799 
2800 	tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, true);
2801 }
2802 
2803 static void *fl_tmplt_create(struct net *net, struct tcf_chain *chain,
2804 			     struct nlattr **tca,
2805 			     struct netlink_ext_ack *extack)
2806 {
2807 	struct nlattr *tca_opts = tca[TCA_OPTIONS];
2808 	struct fl_flow_tmplt *tmplt;
2809 	struct nlattr **tb;
2810 	int err;
2811 
2812 	if (!tca_opts)
2813 		return ERR_PTR(-EINVAL);
2814 
2815 	tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
2816 	if (!tb)
2817 		return ERR_PTR(-ENOBUFS);
2818 	err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
2819 					  tca_opts, fl_policy, NULL);
2820 	if (err)
2821 		goto errout_tb;
2822 
2823 	tmplt = kzalloc(sizeof(*tmplt), GFP_KERNEL);
2824 	if (!tmplt) {
2825 		err = -ENOMEM;
2826 		goto errout_tb;
2827 	}
2828 	tmplt->chain = chain;
2829 	err = fl_set_key(net, tca_opts, tb, &tmplt->dummy_key,
2830 			 &tmplt->mask, extack);
2831 	if (err)
2832 		goto errout_tmplt;
2833 
2834 	fl_init_dissector(&tmplt->dissector, &tmplt->mask);
2835 
2836 	err = fl_hw_create_tmplt(chain, tmplt);
2837 	if (err)
2838 		goto errout_tmplt;
2839 
2840 	kfree(tb);
2841 	return tmplt;
2842 
2843 errout_tmplt:
2844 	kfree(tmplt);
2845 errout_tb:
2846 	kfree(tb);
2847 	return ERR_PTR(err);
2848 }
2849 
2850 static void fl_tmplt_destroy(void *tmplt_priv)
2851 {
2852 	struct fl_flow_tmplt *tmplt = tmplt_priv;
2853 
2854 	fl_hw_destroy_tmplt(tmplt->chain, tmplt);
2855 	kfree(tmplt);
2856 }
2857 
2858 static void fl_tmplt_reoffload(struct tcf_chain *chain, bool add,
2859 			       flow_setup_cb_t *cb, void *cb_priv)
2860 {
2861 	struct fl_flow_tmplt *tmplt = chain->tmplt_priv;
2862 	struct flow_cls_offload cls_flower = {};
2863 
2864 	cls_flower.rule = flow_rule_alloc(0);
2865 	if (!cls_flower.rule)
2866 		return;
2867 
2868 	cls_flower.common.chain_index = chain->index;
2869 	cls_flower.command = add ? FLOW_CLS_TMPLT_CREATE :
2870 				   FLOW_CLS_TMPLT_DESTROY;
2871 	cls_flower.cookie = (unsigned long) tmplt;
2872 	cls_flower.rule->match.dissector = &tmplt->dissector;
2873 	cls_flower.rule->match.mask = &tmplt->mask;
2874 	cls_flower.rule->match.key = &tmplt->dummy_key;
2875 
2876 	cb(TC_SETUP_CLSFLOWER, &cls_flower, cb_priv);
2877 	kfree(cls_flower.rule);
2878 }
2879 
2880 static int fl_dump_key_val(struct sk_buff *skb,
2881 			   void *val, int val_type,
2882 			   void *mask, int mask_type, int len)
2883 {
2884 	int err;
2885 
2886 	if (!memchr_inv(mask, 0, len))
2887 		return 0;
2888 	err = nla_put(skb, val_type, len, val);
2889 	if (err)
2890 		return err;
2891 	if (mask_type != TCA_FLOWER_UNSPEC) {
2892 		err = nla_put(skb, mask_type, len, mask);
2893 		if (err)
2894 			return err;
2895 	}
2896 	return 0;
2897 }
2898 
2899 static int fl_dump_key_port_range(struct sk_buff *skb, struct fl_flow_key *key,
2900 				  struct fl_flow_key *mask)
2901 {
2902 	if (fl_dump_key_val(skb, &key->tp_range.tp_min.dst,
2903 			    TCA_FLOWER_KEY_PORT_DST_MIN,
2904 			    &mask->tp_range.tp_min.dst, TCA_FLOWER_UNSPEC,
2905 			    sizeof(key->tp_range.tp_min.dst)) ||
2906 	    fl_dump_key_val(skb, &key->tp_range.tp_max.dst,
2907 			    TCA_FLOWER_KEY_PORT_DST_MAX,
2908 			    &mask->tp_range.tp_max.dst, TCA_FLOWER_UNSPEC,
2909 			    sizeof(key->tp_range.tp_max.dst)) ||
2910 	    fl_dump_key_val(skb, &key->tp_range.tp_min.src,
2911 			    TCA_FLOWER_KEY_PORT_SRC_MIN,
2912 			    &mask->tp_range.tp_min.src, TCA_FLOWER_UNSPEC,
2913 			    sizeof(key->tp_range.tp_min.src)) ||
2914 	    fl_dump_key_val(skb, &key->tp_range.tp_max.src,
2915 			    TCA_FLOWER_KEY_PORT_SRC_MAX,
2916 			    &mask->tp_range.tp_max.src, TCA_FLOWER_UNSPEC,
2917 			    sizeof(key->tp_range.tp_max.src)))
2918 		return -1;
2919 
2920 	return 0;
2921 }
2922 
2923 static int fl_dump_key_mpls_opt_lse(struct sk_buff *skb,
2924 				    struct flow_dissector_key_mpls *mpls_key,
2925 				    struct flow_dissector_key_mpls *mpls_mask,
2926 				    u8 lse_index)
2927 {
2928 	struct flow_dissector_mpls_lse *lse_mask = &mpls_mask->ls[lse_index];
2929 	struct flow_dissector_mpls_lse *lse_key = &mpls_key->ls[lse_index];
2930 	int err;
2931 
2932 	err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH,
2933 			 lse_index + 1);
2934 	if (err)
2935 		return err;
2936 
2937 	if (lse_mask->mpls_ttl) {
2938 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL,
2939 				 lse_key->mpls_ttl);
2940 		if (err)
2941 			return err;
2942 	}
2943 	if (lse_mask->mpls_bos) {
2944 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS,
2945 				 lse_key->mpls_bos);
2946 		if (err)
2947 			return err;
2948 	}
2949 	if (lse_mask->mpls_tc) {
2950 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_TC,
2951 				 lse_key->mpls_tc);
2952 		if (err)
2953 			return err;
2954 	}
2955 	if (lse_mask->mpls_label) {
2956 		err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL,
2957 				  lse_key->mpls_label);
2958 		if (err)
2959 			return err;
2960 	}
2961 
2962 	return 0;
2963 }
2964 
2965 static int fl_dump_key_mpls_opts(struct sk_buff *skb,
2966 				 struct flow_dissector_key_mpls *mpls_key,
2967 				 struct flow_dissector_key_mpls *mpls_mask)
2968 {
2969 	struct nlattr *opts;
2970 	struct nlattr *lse;
2971 	u8 lse_index;
2972 	int err;
2973 
2974 	opts = nla_nest_start(skb, TCA_FLOWER_KEY_MPLS_OPTS);
2975 	if (!opts)
2976 		return -EMSGSIZE;
2977 
2978 	for (lse_index = 0; lse_index < FLOW_DIS_MPLS_MAX; lse_index++) {
2979 		if (!(mpls_mask->used_lses & 1 << lse_index))
2980 			continue;
2981 
2982 		lse = nla_nest_start(skb, TCA_FLOWER_KEY_MPLS_OPTS_LSE);
2983 		if (!lse) {
2984 			err = -EMSGSIZE;
2985 			goto err_opts;
2986 		}
2987 
2988 		err = fl_dump_key_mpls_opt_lse(skb, mpls_key, mpls_mask,
2989 					       lse_index);
2990 		if (err)
2991 			goto err_opts_lse;
2992 		nla_nest_end(skb, lse);
2993 	}
2994 	nla_nest_end(skb, opts);
2995 
2996 	return 0;
2997 
2998 err_opts_lse:
2999 	nla_nest_cancel(skb, lse);
3000 err_opts:
3001 	nla_nest_cancel(skb, opts);
3002 
3003 	return err;
3004 }
3005 
3006 static int fl_dump_key_mpls(struct sk_buff *skb,
3007 			    struct flow_dissector_key_mpls *mpls_key,
3008 			    struct flow_dissector_key_mpls *mpls_mask)
3009 {
3010 	struct flow_dissector_mpls_lse *lse_mask;
3011 	struct flow_dissector_mpls_lse *lse_key;
3012 	int err;
3013 
3014 	if (!mpls_mask->used_lses)
3015 		return 0;
3016 
3017 	lse_mask = &mpls_mask->ls[0];
3018 	lse_key = &mpls_key->ls[0];
3019 
3020 	/* For backward compatibility, don't use the MPLS nested attributes if
3021 	 * the rule can be expressed using the old attributes.
3022 	 */
3023 	if (mpls_mask->used_lses & ~1 ||
3024 	    (!lse_mask->mpls_ttl && !lse_mask->mpls_bos &&
3025 	     !lse_mask->mpls_tc && !lse_mask->mpls_label))
3026 		return fl_dump_key_mpls_opts(skb, mpls_key, mpls_mask);
3027 
3028 	if (lse_mask->mpls_ttl) {
3029 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TTL,
3030 				 lse_key->mpls_ttl);
3031 		if (err)
3032 			return err;
3033 	}
3034 	if (lse_mask->mpls_tc) {
3035 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TC,
3036 				 lse_key->mpls_tc);
3037 		if (err)
3038 			return err;
3039 	}
3040 	if (lse_mask->mpls_label) {
3041 		err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_LABEL,
3042 				  lse_key->mpls_label);
3043 		if (err)
3044 			return err;
3045 	}
3046 	if (lse_mask->mpls_bos) {
3047 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_BOS,
3048 				 lse_key->mpls_bos);
3049 		if (err)
3050 			return err;
3051 	}
3052 	return 0;
3053 }
3054 
3055 static int fl_dump_key_ip(struct sk_buff *skb, bool encap,
3056 			  struct flow_dissector_key_ip *key,
3057 			  struct flow_dissector_key_ip *mask)
3058 {
3059 	int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
3060 	int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
3061 	int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
3062 	int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
3063 
3064 	if (fl_dump_key_val(skb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos)) ||
3065 	    fl_dump_key_val(skb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl)))
3066 		return -1;
3067 
3068 	return 0;
3069 }
3070 
3071 static int fl_dump_key_vlan(struct sk_buff *skb,
3072 			    int vlan_id_key, int vlan_prio_key,
3073 			    struct flow_dissector_key_vlan *vlan_key,
3074 			    struct flow_dissector_key_vlan *vlan_mask)
3075 {
3076 	int err;
3077 
3078 	if (!memchr_inv(vlan_mask, 0, sizeof(*vlan_mask)))
3079 		return 0;
3080 	if (vlan_mask->vlan_id) {
3081 		err = nla_put_u16(skb, vlan_id_key,
3082 				  vlan_key->vlan_id);
3083 		if (err)
3084 			return err;
3085 	}
3086 	if (vlan_mask->vlan_priority) {
3087 		err = nla_put_u8(skb, vlan_prio_key,
3088 				 vlan_key->vlan_priority);
3089 		if (err)
3090 			return err;
3091 	}
3092 	return 0;
3093 }
3094 
3095 static void fl_get_key_flag(u32 dissector_key, u32 dissector_mask,
3096 			    u32 *flower_key, u32 *flower_mask,
3097 			    u32 flower_flag_bit, u32 dissector_flag_bit)
3098 {
3099 	if (dissector_mask & dissector_flag_bit) {
3100 		*flower_mask |= flower_flag_bit;
3101 		if (dissector_key & dissector_flag_bit)
3102 			*flower_key |= flower_flag_bit;
3103 	}
3104 }
3105 
3106 static int fl_dump_key_flags(struct sk_buff *skb, bool encap,
3107 			     u32 flags_key, u32 flags_mask)
3108 {
3109 	int fl_key, fl_mask;
3110 	__be32 _key, _mask;
3111 	u32 key, mask;
3112 	int err;
3113 
3114 	if (encap) {
3115 		fl_key = TCA_FLOWER_KEY_ENC_FLAGS;
3116 		fl_mask = TCA_FLOWER_KEY_ENC_FLAGS_MASK;
3117 	} else {
3118 		fl_key = TCA_FLOWER_KEY_FLAGS;
3119 		fl_mask = TCA_FLOWER_KEY_FLAGS_MASK;
3120 	}
3121 
3122 	if (!memchr_inv(&flags_mask, 0, sizeof(flags_mask)))
3123 		return 0;
3124 
3125 	key = 0;
3126 	mask = 0;
3127 
3128 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
3129 			TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
3130 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
3131 			TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
3132 			FLOW_DIS_FIRST_FRAG);
3133 
3134 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
3135 			TCA_FLOWER_KEY_FLAGS_TUNNEL_CSUM,
3136 			FLOW_DIS_F_TUNNEL_CSUM);
3137 
3138 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
3139 			TCA_FLOWER_KEY_FLAGS_TUNNEL_DONT_FRAGMENT,
3140 			FLOW_DIS_F_TUNNEL_DONT_FRAGMENT);
3141 
3142 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
3143 			TCA_FLOWER_KEY_FLAGS_TUNNEL_OAM, FLOW_DIS_F_TUNNEL_OAM);
3144 
3145 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
3146 			TCA_FLOWER_KEY_FLAGS_TUNNEL_CRIT_OPT,
3147 			FLOW_DIS_F_TUNNEL_CRIT_OPT);
3148 
3149 	_key = cpu_to_be32(key);
3150 	_mask = cpu_to_be32(mask);
3151 
3152 	err = nla_put(skb, fl_key, 4, &_key);
3153 	if (err)
3154 		return err;
3155 
3156 	return nla_put(skb, fl_mask, 4, &_mask);
3157 }
3158 
3159 static int fl_dump_key_geneve_opt(struct sk_buff *skb,
3160 				  struct flow_dissector_key_enc_opts *enc_opts)
3161 {
3162 	struct geneve_opt *opt;
3163 	struct nlattr *nest;
3164 	int opt_off = 0;
3165 
3166 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_GENEVE);
3167 	if (!nest)
3168 		goto nla_put_failure;
3169 
3170 	while (enc_opts->len > opt_off) {
3171 		opt = (struct geneve_opt *)&enc_opts->data[opt_off];
3172 
3173 		if (nla_put_be16(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS,
3174 				 opt->opt_class))
3175 			goto nla_put_failure;
3176 		if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE,
3177 			       opt->type))
3178 			goto nla_put_failure;
3179 		if (nla_put(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA,
3180 			    opt->length * 4, opt->opt_data))
3181 			goto nla_put_failure;
3182 
3183 		opt_off += sizeof(struct geneve_opt) + opt->length * 4;
3184 	}
3185 	nla_nest_end(skb, nest);
3186 	return 0;
3187 
3188 nla_put_failure:
3189 	nla_nest_cancel(skb, nest);
3190 	return -EMSGSIZE;
3191 }
3192 
3193 static int fl_dump_key_vxlan_opt(struct sk_buff *skb,
3194 				 struct flow_dissector_key_enc_opts *enc_opts)
3195 {
3196 	struct vxlan_metadata *md;
3197 	struct nlattr *nest;
3198 
3199 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_VXLAN);
3200 	if (!nest)
3201 		goto nla_put_failure;
3202 
3203 	md = (struct vxlan_metadata *)&enc_opts->data[0];
3204 	if (nla_put_u32(skb, TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP, md->gbp))
3205 		goto nla_put_failure;
3206 
3207 	nla_nest_end(skb, nest);
3208 	return 0;
3209 
3210 nla_put_failure:
3211 	nla_nest_cancel(skb, nest);
3212 	return -EMSGSIZE;
3213 }
3214 
3215 static int fl_dump_key_erspan_opt(struct sk_buff *skb,
3216 				  struct flow_dissector_key_enc_opts *enc_opts)
3217 {
3218 	struct erspan_metadata *md;
3219 	struct nlattr *nest;
3220 
3221 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_ERSPAN);
3222 	if (!nest)
3223 		goto nla_put_failure;
3224 
3225 	md = (struct erspan_metadata *)&enc_opts->data[0];
3226 	if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER, md->version))
3227 		goto nla_put_failure;
3228 
3229 	if (md->version == 1 &&
3230 	    nla_put_be32(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX, md->u.index))
3231 		goto nla_put_failure;
3232 
3233 	if (md->version == 2 &&
3234 	    (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR,
3235 			md->u.md2.dir) ||
3236 	     nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID,
3237 			get_hwid(&md->u.md2))))
3238 		goto nla_put_failure;
3239 
3240 	nla_nest_end(skb, nest);
3241 	return 0;
3242 
3243 nla_put_failure:
3244 	nla_nest_cancel(skb, nest);
3245 	return -EMSGSIZE;
3246 }
3247 
3248 static int fl_dump_key_gtp_opt(struct sk_buff *skb,
3249 			       struct flow_dissector_key_enc_opts *enc_opts)
3250 
3251 {
3252 	struct gtp_pdu_session_info *session_info;
3253 	struct nlattr *nest;
3254 
3255 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_GTP);
3256 	if (!nest)
3257 		goto nla_put_failure;
3258 
3259 	session_info = (struct gtp_pdu_session_info *)&enc_opts->data[0];
3260 
3261 	if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE,
3262 		       session_info->pdu_type))
3263 		goto nla_put_failure;
3264 
3265 	if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GTP_QFI, session_info->qfi))
3266 		goto nla_put_failure;
3267 
3268 	nla_nest_end(skb, nest);
3269 	return 0;
3270 
3271 nla_put_failure:
3272 	nla_nest_cancel(skb, nest);
3273 	return -EMSGSIZE;
3274 }
3275 
3276 static int fl_dump_key_pfcp_opt(struct sk_buff *skb,
3277 				struct flow_dissector_key_enc_opts *enc_opts)
3278 {
3279 	struct pfcp_metadata *md;
3280 	struct nlattr *nest;
3281 
3282 	nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_PFCP);
3283 	if (!nest)
3284 		goto nla_put_failure;
3285 
3286 	md = (struct pfcp_metadata *)&enc_opts->data[0];
3287 	if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_PFCP_TYPE, md->type))
3288 		goto nla_put_failure;
3289 
3290 	if (nla_put_be64(skb, TCA_FLOWER_KEY_ENC_OPT_PFCP_SEID,
3291 			 md->seid, 0))
3292 		goto nla_put_failure;
3293 
3294 	nla_nest_end(skb, nest);
3295 	return 0;
3296 
3297 nla_put_failure:
3298 	nla_nest_cancel(skb, nest);
3299 	return -EMSGSIZE;
3300 }
3301 
3302 static int fl_dump_key_ct(struct sk_buff *skb,
3303 			  struct flow_dissector_key_ct *key,
3304 			  struct flow_dissector_key_ct *mask)
3305 {
3306 	if (IS_ENABLED(CONFIG_NF_CONNTRACK) &&
3307 	    fl_dump_key_val(skb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE,
3308 			    &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK,
3309 			    sizeof(key->ct_state)))
3310 		goto nla_put_failure;
3311 
3312 	if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
3313 	    fl_dump_key_val(skb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE,
3314 			    &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK,
3315 			    sizeof(key->ct_zone)))
3316 		goto nla_put_failure;
3317 
3318 	if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
3319 	    fl_dump_key_val(skb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK,
3320 			    &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK,
3321 			    sizeof(key->ct_mark)))
3322 		goto nla_put_failure;
3323 
3324 	if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
3325 	    fl_dump_key_val(skb, &key->ct_labels, TCA_FLOWER_KEY_CT_LABELS,
3326 			    &mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK,
3327 			    sizeof(key->ct_labels)))
3328 		goto nla_put_failure;
3329 
3330 	return 0;
3331 
3332 nla_put_failure:
3333 	return -EMSGSIZE;
3334 }
3335 
3336 static int fl_dump_key_cfm(struct sk_buff *skb,
3337 			   struct flow_dissector_key_cfm *key,
3338 			   struct flow_dissector_key_cfm *mask)
3339 {
3340 	struct nlattr *opts;
3341 	int err;
3342 	u8 mdl;
3343 
3344 	if (!memchr_inv(mask, 0, sizeof(*mask)))
3345 		return 0;
3346 
3347 	opts = nla_nest_start(skb, TCA_FLOWER_KEY_CFM);
3348 	if (!opts)
3349 		return -EMSGSIZE;
3350 
3351 	if (FIELD_GET(FLOW_DIS_CFM_MDL_MASK, mask->mdl_ver)) {
3352 		mdl = FIELD_GET(FLOW_DIS_CFM_MDL_MASK, key->mdl_ver);
3353 		err = nla_put_u8(skb, TCA_FLOWER_KEY_CFM_MD_LEVEL, mdl);
3354 		if (err)
3355 			goto err_cfm_opts;
3356 	}
3357 
3358 	if (mask->opcode) {
3359 		err = nla_put_u8(skb, TCA_FLOWER_KEY_CFM_OPCODE, key->opcode);
3360 		if (err)
3361 			goto err_cfm_opts;
3362 	}
3363 
3364 	nla_nest_end(skb, opts);
3365 
3366 	return 0;
3367 
3368 err_cfm_opts:
3369 	nla_nest_cancel(skb, opts);
3370 	return err;
3371 }
3372 
3373 static int fl_dump_key_options(struct sk_buff *skb, int enc_opt_type,
3374 			       struct flow_dissector_key_enc_opts *enc_opts)
3375 {
3376 	struct nlattr *nest;
3377 	int err;
3378 
3379 	if (!enc_opts->len)
3380 		return 0;
3381 
3382 	nest = nla_nest_start_noflag(skb, enc_opt_type);
3383 	if (!nest)
3384 		goto nla_put_failure;
3385 
3386 	switch (enc_opts->dst_opt_type) {
3387 	case IP_TUNNEL_GENEVE_OPT_BIT:
3388 		err = fl_dump_key_geneve_opt(skb, enc_opts);
3389 		if (err)
3390 			goto nla_put_failure;
3391 		break;
3392 	case IP_TUNNEL_VXLAN_OPT_BIT:
3393 		err = fl_dump_key_vxlan_opt(skb, enc_opts);
3394 		if (err)
3395 			goto nla_put_failure;
3396 		break;
3397 	case IP_TUNNEL_ERSPAN_OPT_BIT:
3398 		err = fl_dump_key_erspan_opt(skb, enc_opts);
3399 		if (err)
3400 			goto nla_put_failure;
3401 		break;
3402 	case IP_TUNNEL_GTP_OPT_BIT:
3403 		err = fl_dump_key_gtp_opt(skb, enc_opts);
3404 		if (err)
3405 			goto nla_put_failure;
3406 		break;
3407 	case IP_TUNNEL_PFCP_OPT_BIT:
3408 		err = fl_dump_key_pfcp_opt(skb, enc_opts);
3409 		if (err)
3410 			goto nla_put_failure;
3411 		break;
3412 	default:
3413 		goto nla_put_failure;
3414 	}
3415 	nla_nest_end(skb, nest);
3416 	return 0;
3417 
3418 nla_put_failure:
3419 	nla_nest_cancel(skb, nest);
3420 	return -EMSGSIZE;
3421 }
3422 
3423 static int fl_dump_key_enc_opt(struct sk_buff *skb,
3424 			       struct flow_dissector_key_enc_opts *key_opts,
3425 			       struct flow_dissector_key_enc_opts *msk_opts)
3426 {
3427 	int err;
3428 
3429 	err = fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS, key_opts);
3430 	if (err)
3431 		return err;
3432 
3433 	return fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS_MASK, msk_opts);
3434 }
3435 
3436 static int fl_dump_key(struct sk_buff *skb, struct net *net,
3437 		       struct fl_flow_key *key, struct fl_flow_key *mask)
3438 {
3439 	if (mask->meta.ingress_ifindex) {
3440 		struct net_device *dev;
3441 
3442 		dev = __dev_get_by_index(net, key->meta.ingress_ifindex);
3443 		if (dev && nla_put_string(skb, TCA_FLOWER_INDEV, dev->name))
3444 			goto nla_put_failure;
3445 	}
3446 
3447 	if (fl_dump_key_val(skb, &key->meta.l2_miss,
3448 			    TCA_FLOWER_L2_MISS, &mask->meta.l2_miss,
3449 			    TCA_FLOWER_UNSPEC, sizeof(key->meta.l2_miss)))
3450 		goto nla_put_failure;
3451 
3452 	if (fl_dump_key_val(skb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
3453 			    mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
3454 			    sizeof(key->eth.dst)) ||
3455 	    fl_dump_key_val(skb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
3456 			    mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
3457 			    sizeof(key->eth.src)) ||
3458 	    fl_dump_key_val(skb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE,
3459 			    &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
3460 			    sizeof(key->basic.n_proto)))
3461 		goto nla_put_failure;
3462 
3463 	if (mask->num_of_vlans.num_of_vlans) {
3464 		if (nla_put_u8(skb, TCA_FLOWER_KEY_NUM_OF_VLANS, key->num_of_vlans.num_of_vlans))
3465 			goto nla_put_failure;
3466 	}
3467 
3468 	if (fl_dump_key_mpls(skb, &key->mpls, &mask->mpls))
3469 		goto nla_put_failure;
3470 
3471 	if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_VLAN_ID,
3472 			     TCA_FLOWER_KEY_VLAN_PRIO, &key->vlan, &mask->vlan))
3473 		goto nla_put_failure;
3474 
3475 	if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_CVLAN_ID,
3476 			     TCA_FLOWER_KEY_CVLAN_PRIO,
3477 			     &key->cvlan, &mask->cvlan) ||
3478 	    (mask->cvlan.vlan_tpid &&
3479 	     nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
3480 			  key->cvlan.vlan_tpid)))
3481 		goto nla_put_failure;
3482 
3483 	if (mask->basic.n_proto) {
3484 		if (mask->cvlan.vlan_eth_type) {
3485 			if (nla_put_be16(skb, TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
3486 					 key->basic.n_proto))
3487 				goto nla_put_failure;
3488 		} else if (mask->vlan.vlan_eth_type) {
3489 			if (nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
3490 					 key->vlan.vlan_eth_type))
3491 				goto nla_put_failure;
3492 		}
3493 	}
3494 
3495 	if ((key->basic.n_proto == htons(ETH_P_IP) ||
3496 	     key->basic.n_proto == htons(ETH_P_IPV6)) &&
3497 	    (fl_dump_key_val(skb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
3498 			    &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
3499 			    sizeof(key->basic.ip_proto)) ||
3500 	    fl_dump_key_ip(skb, false, &key->ip, &mask->ip)))
3501 		goto nla_put_failure;
3502 
3503 	if (mask->pppoe.session_id) {
3504 		if (nla_put_be16(skb, TCA_FLOWER_KEY_PPPOE_SID,
3505 				 key->pppoe.session_id))
3506 			goto nla_put_failure;
3507 	}
3508 	if (mask->basic.n_proto && mask->pppoe.ppp_proto) {
3509 		if (nla_put_be16(skb, TCA_FLOWER_KEY_PPP_PROTO,
3510 				 key->pppoe.ppp_proto))
3511 			goto nla_put_failure;
3512 	}
3513 
3514 	if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
3515 	    (fl_dump_key_val(skb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
3516 			     &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
3517 			     sizeof(key->ipv4.src)) ||
3518 	     fl_dump_key_val(skb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
3519 			     &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
3520 			     sizeof(key->ipv4.dst))))
3521 		goto nla_put_failure;
3522 	else if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
3523 		 (fl_dump_key_val(skb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
3524 				  &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
3525 				  sizeof(key->ipv6.src)) ||
3526 		  fl_dump_key_val(skb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
3527 				  &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
3528 				  sizeof(key->ipv6.dst))))
3529 		goto nla_put_failure;
3530 
3531 	if (key->basic.ip_proto == IPPROTO_TCP &&
3532 	    (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
3533 			     &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
3534 			     sizeof(key->tp.src)) ||
3535 	     fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
3536 			     &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
3537 			     sizeof(key->tp.dst)) ||
3538 	     fl_dump_key_val(skb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
3539 			     &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
3540 			     sizeof(key->tcp.flags))))
3541 		goto nla_put_failure;
3542 	else if (key->basic.ip_proto == IPPROTO_UDP &&
3543 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
3544 				  &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
3545 				  sizeof(key->tp.src)) ||
3546 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
3547 				  &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
3548 				  sizeof(key->tp.dst))))
3549 		goto nla_put_failure;
3550 	else if (key->basic.ip_proto == IPPROTO_SCTP &&
3551 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
3552 				  &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
3553 				  sizeof(key->tp.src)) ||
3554 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
3555 				  &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
3556 				  sizeof(key->tp.dst))))
3557 		goto nla_put_failure;
3558 	else if (key->basic.n_proto == htons(ETH_P_IP) &&
3559 		 key->basic.ip_proto == IPPROTO_ICMP &&
3560 		 (fl_dump_key_val(skb, &key->icmp.type,
3561 				  TCA_FLOWER_KEY_ICMPV4_TYPE, &mask->icmp.type,
3562 				  TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
3563 				  sizeof(key->icmp.type)) ||
3564 		  fl_dump_key_val(skb, &key->icmp.code,
3565 				  TCA_FLOWER_KEY_ICMPV4_CODE, &mask->icmp.code,
3566 				  TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
3567 				  sizeof(key->icmp.code))))
3568 		goto nla_put_failure;
3569 	else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
3570 		 key->basic.ip_proto == IPPROTO_ICMPV6 &&
3571 		 (fl_dump_key_val(skb, &key->icmp.type,
3572 				  TCA_FLOWER_KEY_ICMPV6_TYPE, &mask->icmp.type,
3573 				  TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
3574 				  sizeof(key->icmp.type)) ||
3575 		  fl_dump_key_val(skb, &key->icmp.code,
3576 				  TCA_FLOWER_KEY_ICMPV6_CODE, &mask->icmp.code,
3577 				  TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
3578 				  sizeof(key->icmp.code))))
3579 		goto nla_put_failure;
3580 	else if ((key->basic.n_proto == htons(ETH_P_ARP) ||
3581 		  key->basic.n_proto == htons(ETH_P_RARP)) &&
3582 		 (fl_dump_key_val(skb, &key->arp.sip,
3583 				  TCA_FLOWER_KEY_ARP_SIP, &mask->arp.sip,
3584 				  TCA_FLOWER_KEY_ARP_SIP_MASK,
3585 				  sizeof(key->arp.sip)) ||
3586 		  fl_dump_key_val(skb, &key->arp.tip,
3587 				  TCA_FLOWER_KEY_ARP_TIP, &mask->arp.tip,
3588 				  TCA_FLOWER_KEY_ARP_TIP_MASK,
3589 				  sizeof(key->arp.tip)) ||
3590 		  fl_dump_key_val(skb, &key->arp.op,
3591 				  TCA_FLOWER_KEY_ARP_OP, &mask->arp.op,
3592 				  TCA_FLOWER_KEY_ARP_OP_MASK,
3593 				  sizeof(key->arp.op)) ||
3594 		  fl_dump_key_val(skb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
3595 				  mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
3596 				  sizeof(key->arp.sha)) ||
3597 		  fl_dump_key_val(skb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
3598 				  mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
3599 				  sizeof(key->arp.tha))))
3600 		goto nla_put_failure;
3601 	else if (key->basic.ip_proto == IPPROTO_L2TP &&
3602 		 fl_dump_key_val(skb, &key->l2tpv3.session_id,
3603 				 TCA_FLOWER_KEY_L2TPV3_SID,
3604 				 &mask->l2tpv3.session_id,
3605 				 TCA_FLOWER_UNSPEC,
3606 				 sizeof(key->l2tpv3.session_id)))
3607 		goto nla_put_failure;
3608 
3609 	if (key->ipsec.spi &&
3610 	    fl_dump_key_val(skb, &key->ipsec.spi, TCA_FLOWER_KEY_SPI,
3611 			    &mask->ipsec.spi, TCA_FLOWER_KEY_SPI_MASK,
3612 			    sizeof(key->ipsec.spi)))
3613 		goto nla_put_failure;
3614 
3615 	if ((key->basic.ip_proto == IPPROTO_TCP ||
3616 	     key->basic.ip_proto == IPPROTO_UDP ||
3617 	     key->basic.ip_proto == IPPROTO_SCTP) &&
3618 	     fl_dump_key_port_range(skb, key, mask))
3619 		goto nla_put_failure;
3620 
3621 	if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
3622 	    (fl_dump_key_val(skb, &key->enc_ipv4.src,
3623 			    TCA_FLOWER_KEY_ENC_IPV4_SRC, &mask->enc_ipv4.src,
3624 			    TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
3625 			    sizeof(key->enc_ipv4.src)) ||
3626 	     fl_dump_key_val(skb, &key->enc_ipv4.dst,
3627 			     TCA_FLOWER_KEY_ENC_IPV4_DST, &mask->enc_ipv4.dst,
3628 			     TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
3629 			     sizeof(key->enc_ipv4.dst))))
3630 		goto nla_put_failure;
3631 	else if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
3632 		 (fl_dump_key_val(skb, &key->enc_ipv6.src,
3633 			    TCA_FLOWER_KEY_ENC_IPV6_SRC, &mask->enc_ipv6.src,
3634 			    TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
3635 			    sizeof(key->enc_ipv6.src)) ||
3636 		 fl_dump_key_val(skb, &key->enc_ipv6.dst,
3637 				 TCA_FLOWER_KEY_ENC_IPV6_DST,
3638 				 &mask->enc_ipv6.dst,
3639 				 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
3640 			    sizeof(key->enc_ipv6.dst))))
3641 		goto nla_put_failure;
3642 
3643 	if (fl_dump_key_val(skb, &key->enc_key_id, TCA_FLOWER_KEY_ENC_KEY_ID,
3644 			    &mask->enc_key_id, TCA_FLOWER_UNSPEC,
3645 			    sizeof(key->enc_key_id)) ||
3646 	    fl_dump_key_val(skb, &key->enc_tp.src,
3647 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
3648 			    &mask->enc_tp.src,
3649 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
3650 			    sizeof(key->enc_tp.src)) ||
3651 	    fl_dump_key_val(skb, &key->enc_tp.dst,
3652 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
3653 			    &mask->enc_tp.dst,
3654 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
3655 			    sizeof(key->enc_tp.dst)) ||
3656 	    fl_dump_key_ip(skb, true, &key->enc_ip, &mask->enc_ip) ||
3657 	    fl_dump_key_enc_opt(skb, &key->enc_opts, &mask->enc_opts))
3658 		goto nla_put_failure;
3659 
3660 	if (fl_dump_key_ct(skb, &key->ct, &mask->ct))
3661 		goto nla_put_failure;
3662 
3663 	if (fl_dump_key_flags(skb, false, key->control.flags,
3664 			      mask->control.flags))
3665 		goto nla_put_failure;
3666 
3667 	if (fl_dump_key_val(skb, &key->hash.hash, TCA_FLOWER_KEY_HASH,
3668 			     &mask->hash.hash, TCA_FLOWER_KEY_HASH_MASK,
3669 			     sizeof(key->hash.hash)))
3670 		goto nla_put_failure;
3671 
3672 	if (fl_dump_key_cfm(skb, &key->cfm, &mask->cfm))
3673 		goto nla_put_failure;
3674 
3675 	if (fl_dump_key_flags(skb, true, key->enc_control.flags,
3676 			      mask->enc_control.flags))
3677 		goto nla_put_failure;
3678 
3679 	return 0;
3680 
3681 nla_put_failure:
3682 	return -EMSGSIZE;
3683 }
3684 
3685 static int fl_dump(struct net *net, struct tcf_proto *tp, void *fh,
3686 		   struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
3687 {
3688 	struct cls_fl_filter *f = fh;
3689 	struct nlattr *nest;
3690 	struct fl_flow_key *key, *mask;
3691 	bool skip_hw;
3692 
3693 	if (!f)
3694 		return skb->len;
3695 
3696 	t->tcm_handle = f->handle;
3697 
3698 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
3699 	if (!nest)
3700 		goto nla_put_failure;
3701 
3702 	spin_lock(&tp->lock);
3703 
3704 	if (f->res.classid &&
3705 	    nla_put_u32(skb, TCA_FLOWER_CLASSID, f->res.classid))
3706 		goto nla_put_failure_locked;
3707 
3708 	key = &f->key;
3709 	mask = &f->mask->key;
3710 	skip_hw = tc_skip_hw(f->flags);
3711 
3712 	if (fl_dump_key(skb, net, key, mask))
3713 		goto nla_put_failure_locked;
3714 
3715 	if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
3716 		goto nla_put_failure_locked;
3717 
3718 	spin_unlock(&tp->lock);
3719 
3720 	if (!skip_hw)
3721 		fl_hw_update_stats(tp, f, rtnl_held);
3722 
3723 	if (nla_put_u32(skb, TCA_FLOWER_IN_HW_COUNT, f->in_hw_count))
3724 		goto nla_put_failure;
3725 
3726 	if (tcf_exts_dump(skb, &f->exts))
3727 		goto nla_put_failure;
3728 
3729 	nla_nest_end(skb, nest);
3730 
3731 	if (tcf_exts_dump_stats(skb, &f->exts) < 0)
3732 		goto nla_put_failure;
3733 
3734 	return skb->len;
3735 
3736 nla_put_failure_locked:
3737 	spin_unlock(&tp->lock);
3738 nla_put_failure:
3739 	nla_nest_cancel(skb, nest);
3740 	return -1;
3741 }
3742 
3743 static int fl_terse_dump(struct net *net, struct tcf_proto *tp, void *fh,
3744 			 struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
3745 {
3746 	struct cls_fl_filter *f = fh;
3747 	struct nlattr *nest;
3748 	bool skip_hw;
3749 
3750 	if (!f)
3751 		return skb->len;
3752 
3753 	t->tcm_handle = f->handle;
3754 
3755 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
3756 	if (!nest)
3757 		goto nla_put_failure;
3758 
3759 	spin_lock(&tp->lock);
3760 
3761 	skip_hw = tc_skip_hw(f->flags);
3762 
3763 	if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
3764 		goto nla_put_failure_locked;
3765 
3766 	spin_unlock(&tp->lock);
3767 
3768 	if (!skip_hw)
3769 		fl_hw_update_stats(tp, f, rtnl_held);
3770 
3771 	if (tcf_exts_terse_dump(skb, &f->exts))
3772 		goto nla_put_failure;
3773 
3774 	nla_nest_end(skb, nest);
3775 
3776 	return skb->len;
3777 
3778 nla_put_failure_locked:
3779 	spin_unlock(&tp->lock);
3780 nla_put_failure:
3781 	nla_nest_cancel(skb, nest);
3782 	return -1;
3783 }
3784 
3785 static int fl_tmplt_dump(struct sk_buff *skb, struct net *net, void *tmplt_priv)
3786 {
3787 	struct fl_flow_tmplt *tmplt = tmplt_priv;
3788 	struct fl_flow_key *key, *mask;
3789 	struct nlattr *nest;
3790 
3791 	nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
3792 	if (!nest)
3793 		goto nla_put_failure;
3794 
3795 	key = &tmplt->dummy_key;
3796 	mask = &tmplt->mask;
3797 
3798 	if (fl_dump_key(skb, net, key, mask))
3799 		goto nla_put_failure;
3800 
3801 	nla_nest_end(skb, nest);
3802 
3803 	return skb->len;
3804 
3805 nla_put_failure:
3806 	nla_nest_cancel(skb, nest);
3807 	return -EMSGSIZE;
3808 }
3809 
3810 static void fl_bind_class(void *fh, u32 classid, unsigned long cl, void *q,
3811 			  unsigned long base)
3812 {
3813 	struct cls_fl_filter *f = fh;
3814 
3815 	tc_cls_bind_class(classid, cl, q, &f->res, base);
3816 }
3817 
3818 static bool fl_delete_empty(struct tcf_proto *tp)
3819 {
3820 	struct cls_fl_head *head = fl_head_dereference(tp);
3821 
3822 	spin_lock(&tp->lock);
3823 	tp->deleting = idr_is_empty(&head->handle_idr);
3824 	spin_unlock(&tp->lock);
3825 
3826 	return tp->deleting;
3827 }
3828 
3829 static struct tcf_proto_ops cls_fl_ops __read_mostly = {
3830 	.kind		= "flower",
3831 	.classify	= fl_classify,
3832 	.init		= fl_init,
3833 	.destroy	= fl_destroy,
3834 	.get		= fl_get,
3835 	.put		= fl_put,
3836 	.change		= fl_change,
3837 	.delete		= fl_delete,
3838 	.delete_empty	= fl_delete_empty,
3839 	.walk		= fl_walk,
3840 	.reoffload	= fl_reoffload,
3841 	.hw_add		= fl_hw_add,
3842 	.hw_del		= fl_hw_del,
3843 	.dump		= fl_dump,
3844 	.terse_dump	= fl_terse_dump,
3845 	.bind_class	= fl_bind_class,
3846 	.tmplt_create	= fl_tmplt_create,
3847 	.tmplt_destroy	= fl_tmplt_destroy,
3848 	.tmplt_reoffload = fl_tmplt_reoffload,
3849 	.tmplt_dump	= fl_tmplt_dump,
3850 	.get_exts	= fl_get_exts,
3851 	.owner		= THIS_MODULE,
3852 	.flags		= TCF_PROTO_OPS_DOIT_UNLOCKED,
3853 };
3854 MODULE_ALIAS_NET_CLS("flower");
3855 
3856 static int __init cls_fl_init(void)
3857 {
3858 	return register_tcf_proto_ops(&cls_fl_ops);
3859 }
3860 
3861 static void __exit cls_fl_exit(void)
3862 {
3863 	unregister_tcf_proto_ops(&cls_fl_ops);
3864 }
3865 
3866 module_init(cls_fl_init);
3867 module_exit(cls_fl_exit);
3868 
3869 MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
3870 MODULE_DESCRIPTION("Flower classifier");
3871 MODULE_LICENSE("GPL v2");
3872