xref: /linux/net/sched/cls_flower.c (revision fd639726bf15fca8ee1a00dce8e0096d0ad9bd18)
1 /*
2  * net/sched/cls_flower.c		Flower classifier
3  *
4  * Copyright (c) 2015 Jiri Pirko <jiri@resnulli.us>
5  *
6  * This program is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License as published by
8  * the Free Software Foundation; either version 2 of the License, or
9  * (at your option) any later version.
10  */
11 
12 #include <linux/kernel.h>
13 #include <linux/init.h>
14 #include <linux/module.h>
15 #include <linux/rhashtable.h>
16 #include <linux/workqueue.h>
17 
18 #include <linux/if_ether.h>
19 #include <linux/in6.h>
20 #include <linux/ip.h>
21 #include <linux/mpls.h>
22 
23 #include <net/sch_generic.h>
24 #include <net/pkt_cls.h>
25 #include <net/ip.h>
26 #include <net/flow_dissector.h>
27 
28 #include <net/dst.h>
29 #include <net/dst_metadata.h>
30 
31 struct fl_flow_key {
32 	int	indev_ifindex;
33 	struct flow_dissector_key_control control;
34 	struct flow_dissector_key_control enc_control;
35 	struct flow_dissector_key_basic basic;
36 	struct flow_dissector_key_eth_addrs eth;
37 	struct flow_dissector_key_vlan vlan;
38 	union {
39 		struct flow_dissector_key_ipv4_addrs ipv4;
40 		struct flow_dissector_key_ipv6_addrs ipv6;
41 	};
42 	struct flow_dissector_key_ports tp;
43 	struct flow_dissector_key_icmp icmp;
44 	struct flow_dissector_key_arp arp;
45 	struct flow_dissector_key_keyid enc_key_id;
46 	union {
47 		struct flow_dissector_key_ipv4_addrs enc_ipv4;
48 		struct flow_dissector_key_ipv6_addrs enc_ipv6;
49 	};
50 	struct flow_dissector_key_ports enc_tp;
51 	struct flow_dissector_key_mpls mpls;
52 	struct flow_dissector_key_tcp tcp;
53 	struct flow_dissector_key_ip ip;
54 } __aligned(BITS_PER_LONG / 8); /* Ensure that we can do comparisons as longs. */
55 
56 struct fl_flow_mask_range {
57 	unsigned short int start;
58 	unsigned short int end;
59 };
60 
61 struct fl_flow_mask {
62 	struct fl_flow_key key;
63 	struct fl_flow_mask_range range;
64 	struct rcu_head	rcu;
65 };
66 
67 struct cls_fl_head {
68 	struct rhashtable ht;
69 	struct fl_flow_mask mask;
70 	struct flow_dissector dissector;
71 	bool mask_assigned;
72 	struct list_head filters;
73 	struct rhashtable_params ht_params;
74 	union {
75 		struct work_struct work;
76 		struct rcu_head	rcu;
77 	};
78 	struct idr handle_idr;
79 };
80 
81 struct cls_fl_filter {
82 	struct rhash_head ht_node;
83 	struct fl_flow_key mkey;
84 	struct tcf_exts exts;
85 	struct tcf_result res;
86 	struct fl_flow_key key;
87 	struct list_head list;
88 	u32 handle;
89 	u32 flags;
90 	union {
91 		struct work_struct work;
92 		struct rcu_head	rcu;
93 	};
94 	struct net_device *hw_dev;
95 };
96 
97 static unsigned short int fl_mask_range(const struct fl_flow_mask *mask)
98 {
99 	return mask->range.end - mask->range.start;
100 }
101 
102 static void fl_mask_update_range(struct fl_flow_mask *mask)
103 {
104 	const u8 *bytes = (const u8 *) &mask->key;
105 	size_t size = sizeof(mask->key);
106 	size_t i, first = 0, last = size - 1;
107 
108 	for (i = 0; i < sizeof(mask->key); i++) {
109 		if (bytes[i]) {
110 			if (!first && i)
111 				first = i;
112 			last = i;
113 		}
114 	}
115 	mask->range.start = rounddown(first, sizeof(long));
116 	mask->range.end = roundup(last + 1, sizeof(long));
117 }
118 
119 static void *fl_key_get_start(struct fl_flow_key *key,
120 			      const struct fl_flow_mask *mask)
121 {
122 	return (u8 *) key + mask->range.start;
123 }
124 
125 static void fl_set_masked_key(struct fl_flow_key *mkey, struct fl_flow_key *key,
126 			      struct fl_flow_mask *mask)
127 {
128 	const long *lkey = fl_key_get_start(key, mask);
129 	const long *lmask = fl_key_get_start(&mask->key, mask);
130 	long *lmkey = fl_key_get_start(mkey, mask);
131 	int i;
132 
133 	for (i = 0; i < fl_mask_range(mask); i += sizeof(long))
134 		*lmkey++ = *lkey++ & *lmask++;
135 }
136 
137 static void fl_clear_masked_range(struct fl_flow_key *key,
138 				  struct fl_flow_mask *mask)
139 {
140 	memset(fl_key_get_start(key, mask), 0, fl_mask_range(mask));
141 }
142 
143 static struct cls_fl_filter *fl_lookup(struct cls_fl_head *head,
144 				       struct fl_flow_key *mkey)
145 {
146 	return rhashtable_lookup_fast(&head->ht,
147 				      fl_key_get_start(mkey, &head->mask),
148 				      head->ht_params);
149 }
150 
151 static int fl_classify(struct sk_buff *skb, const struct tcf_proto *tp,
152 		       struct tcf_result *res)
153 {
154 	struct cls_fl_head *head = rcu_dereference_bh(tp->root);
155 	struct cls_fl_filter *f;
156 	struct fl_flow_key skb_key;
157 	struct fl_flow_key skb_mkey;
158 
159 	if (!atomic_read(&head->ht.nelems))
160 		return -1;
161 
162 	fl_clear_masked_range(&skb_key, &head->mask);
163 
164 	skb_key.indev_ifindex = skb->skb_iif;
165 	/* skb_flow_dissect() does not set n_proto in case an unknown protocol,
166 	 * so do it rather here.
167 	 */
168 	skb_key.basic.n_proto = skb->protocol;
169 	skb_flow_dissect(skb, &head->dissector, &skb_key, 0);
170 
171 	fl_set_masked_key(&skb_mkey, &skb_key, &head->mask);
172 
173 	f = fl_lookup(head, &skb_mkey);
174 	if (f && !tc_skip_sw(f->flags)) {
175 		*res = f->res;
176 		return tcf_exts_exec(skb, &f->exts, res);
177 	}
178 	return -1;
179 }
180 
181 static int fl_init(struct tcf_proto *tp)
182 {
183 	struct cls_fl_head *head;
184 
185 	head = kzalloc(sizeof(*head), GFP_KERNEL);
186 	if (!head)
187 		return -ENOBUFS;
188 
189 	INIT_LIST_HEAD_RCU(&head->filters);
190 	rcu_assign_pointer(tp->root, head);
191 	idr_init(&head->handle_idr);
192 
193 	return 0;
194 }
195 
196 static void __fl_destroy_filter(struct cls_fl_filter *f)
197 {
198 	tcf_exts_destroy(&f->exts);
199 	tcf_exts_put_net(&f->exts);
200 	kfree(f);
201 }
202 
203 static void fl_destroy_filter_work(struct work_struct *work)
204 {
205 	struct cls_fl_filter *f = container_of(work, struct cls_fl_filter, work);
206 
207 	rtnl_lock();
208 	__fl_destroy_filter(f);
209 	rtnl_unlock();
210 }
211 
212 static void fl_destroy_filter(struct rcu_head *head)
213 {
214 	struct cls_fl_filter *f = container_of(head, struct cls_fl_filter, rcu);
215 
216 	INIT_WORK(&f->work, fl_destroy_filter_work);
217 	tcf_queue_work(&f->work);
218 }
219 
220 static void fl_hw_destroy_filter(struct tcf_proto *tp, struct cls_fl_filter *f)
221 {
222 	struct tc_cls_flower_offload cls_flower = {};
223 	struct tcf_block *block = tp->chain->block;
224 
225 	tc_cls_common_offload_init(&cls_flower.common, tp);
226 	cls_flower.command = TC_CLSFLOWER_DESTROY;
227 	cls_flower.cookie = (unsigned long) f;
228 
229 	tc_setup_cb_call(block, &f->exts, TC_SETUP_CLSFLOWER,
230 			 &cls_flower, false);
231 }
232 
233 static int fl_hw_replace_filter(struct tcf_proto *tp,
234 				struct flow_dissector *dissector,
235 				struct fl_flow_key *mask,
236 				struct cls_fl_filter *f)
237 {
238 	struct tc_cls_flower_offload cls_flower = {};
239 	struct tcf_block *block = tp->chain->block;
240 	bool skip_sw = tc_skip_sw(f->flags);
241 	int err;
242 
243 	tc_cls_common_offload_init(&cls_flower.common, tp);
244 	cls_flower.command = TC_CLSFLOWER_REPLACE;
245 	cls_flower.cookie = (unsigned long) f;
246 	cls_flower.dissector = dissector;
247 	cls_flower.mask = mask;
248 	cls_flower.key = &f->mkey;
249 	cls_flower.exts = &f->exts;
250 	cls_flower.classid = f->res.classid;
251 
252 	err = tc_setup_cb_call(block, &f->exts, TC_SETUP_CLSFLOWER,
253 			       &cls_flower, skip_sw);
254 	if (err < 0) {
255 		fl_hw_destroy_filter(tp, f);
256 		return err;
257 	} else if (err > 0) {
258 		f->flags |= TCA_CLS_FLAGS_IN_HW;
259 	}
260 
261 	if (skip_sw && !(f->flags & TCA_CLS_FLAGS_IN_HW))
262 		return -EINVAL;
263 
264 	return 0;
265 }
266 
267 static void fl_hw_update_stats(struct tcf_proto *tp, struct cls_fl_filter *f)
268 {
269 	struct tc_cls_flower_offload cls_flower = {};
270 	struct tcf_block *block = tp->chain->block;
271 
272 	tc_cls_common_offload_init(&cls_flower.common, tp);
273 	cls_flower.command = TC_CLSFLOWER_STATS;
274 	cls_flower.cookie = (unsigned long) f;
275 	cls_flower.exts = &f->exts;
276 	cls_flower.classid = f->res.classid;
277 
278 	tc_setup_cb_call(block, &f->exts, TC_SETUP_CLSFLOWER,
279 			 &cls_flower, false);
280 }
281 
282 static void __fl_delete(struct tcf_proto *tp, struct cls_fl_filter *f)
283 {
284 	struct cls_fl_head *head = rtnl_dereference(tp->root);
285 
286 	idr_remove_ext(&head->handle_idr, f->handle);
287 	list_del_rcu(&f->list);
288 	if (!tc_skip_hw(f->flags))
289 		fl_hw_destroy_filter(tp, f);
290 	tcf_unbind_filter(tp, &f->res);
291 	if (tcf_exts_get_net(&f->exts))
292 		call_rcu(&f->rcu, fl_destroy_filter);
293 	else
294 		__fl_destroy_filter(f);
295 }
296 
297 static void fl_destroy_sleepable(struct work_struct *work)
298 {
299 	struct cls_fl_head *head = container_of(work, struct cls_fl_head,
300 						work);
301 	if (head->mask_assigned)
302 		rhashtable_destroy(&head->ht);
303 	kfree(head);
304 	module_put(THIS_MODULE);
305 }
306 
307 static void fl_destroy_rcu(struct rcu_head *rcu)
308 {
309 	struct cls_fl_head *head = container_of(rcu, struct cls_fl_head, rcu);
310 
311 	INIT_WORK(&head->work, fl_destroy_sleepable);
312 	schedule_work(&head->work);
313 }
314 
315 static void fl_destroy(struct tcf_proto *tp)
316 {
317 	struct cls_fl_head *head = rtnl_dereference(tp->root);
318 	struct cls_fl_filter *f, *next;
319 
320 	list_for_each_entry_safe(f, next, &head->filters, list)
321 		__fl_delete(tp, f);
322 	idr_destroy(&head->handle_idr);
323 
324 	__module_get(THIS_MODULE);
325 	call_rcu(&head->rcu, fl_destroy_rcu);
326 }
327 
328 static void *fl_get(struct tcf_proto *tp, u32 handle)
329 {
330 	struct cls_fl_head *head = rtnl_dereference(tp->root);
331 
332 	return idr_find_ext(&head->handle_idr, handle);
333 }
334 
335 static const struct nla_policy fl_policy[TCA_FLOWER_MAX + 1] = {
336 	[TCA_FLOWER_UNSPEC]		= { .type = NLA_UNSPEC },
337 	[TCA_FLOWER_CLASSID]		= { .type = NLA_U32 },
338 	[TCA_FLOWER_INDEV]		= { .type = NLA_STRING,
339 					    .len = IFNAMSIZ },
340 	[TCA_FLOWER_KEY_ETH_DST]	= { .len = ETH_ALEN },
341 	[TCA_FLOWER_KEY_ETH_DST_MASK]	= { .len = ETH_ALEN },
342 	[TCA_FLOWER_KEY_ETH_SRC]	= { .len = ETH_ALEN },
343 	[TCA_FLOWER_KEY_ETH_SRC_MASK]	= { .len = ETH_ALEN },
344 	[TCA_FLOWER_KEY_ETH_TYPE]	= { .type = NLA_U16 },
345 	[TCA_FLOWER_KEY_IP_PROTO]	= { .type = NLA_U8 },
346 	[TCA_FLOWER_KEY_IPV4_SRC]	= { .type = NLA_U32 },
347 	[TCA_FLOWER_KEY_IPV4_SRC_MASK]	= { .type = NLA_U32 },
348 	[TCA_FLOWER_KEY_IPV4_DST]	= { .type = NLA_U32 },
349 	[TCA_FLOWER_KEY_IPV4_DST_MASK]	= { .type = NLA_U32 },
350 	[TCA_FLOWER_KEY_IPV6_SRC]	= { .len = sizeof(struct in6_addr) },
351 	[TCA_FLOWER_KEY_IPV6_SRC_MASK]	= { .len = sizeof(struct in6_addr) },
352 	[TCA_FLOWER_KEY_IPV6_DST]	= { .len = sizeof(struct in6_addr) },
353 	[TCA_FLOWER_KEY_IPV6_DST_MASK]	= { .len = sizeof(struct in6_addr) },
354 	[TCA_FLOWER_KEY_TCP_SRC]	= { .type = NLA_U16 },
355 	[TCA_FLOWER_KEY_TCP_DST]	= { .type = NLA_U16 },
356 	[TCA_FLOWER_KEY_UDP_SRC]	= { .type = NLA_U16 },
357 	[TCA_FLOWER_KEY_UDP_DST]	= { .type = NLA_U16 },
358 	[TCA_FLOWER_KEY_VLAN_ID]	= { .type = NLA_U16 },
359 	[TCA_FLOWER_KEY_VLAN_PRIO]	= { .type = NLA_U8 },
360 	[TCA_FLOWER_KEY_VLAN_ETH_TYPE]	= { .type = NLA_U16 },
361 	[TCA_FLOWER_KEY_ENC_KEY_ID]	= { .type = NLA_U32 },
362 	[TCA_FLOWER_KEY_ENC_IPV4_SRC]	= { .type = NLA_U32 },
363 	[TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK] = { .type = NLA_U32 },
364 	[TCA_FLOWER_KEY_ENC_IPV4_DST]	= { .type = NLA_U32 },
365 	[TCA_FLOWER_KEY_ENC_IPV4_DST_MASK] = { .type = NLA_U32 },
366 	[TCA_FLOWER_KEY_ENC_IPV6_SRC]	= { .len = sizeof(struct in6_addr) },
367 	[TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK] = { .len = sizeof(struct in6_addr) },
368 	[TCA_FLOWER_KEY_ENC_IPV6_DST]	= { .len = sizeof(struct in6_addr) },
369 	[TCA_FLOWER_KEY_ENC_IPV6_DST_MASK] = { .len = sizeof(struct in6_addr) },
370 	[TCA_FLOWER_KEY_TCP_SRC_MASK]	= { .type = NLA_U16 },
371 	[TCA_FLOWER_KEY_TCP_DST_MASK]	= { .type = NLA_U16 },
372 	[TCA_FLOWER_KEY_UDP_SRC_MASK]	= { .type = NLA_U16 },
373 	[TCA_FLOWER_KEY_UDP_DST_MASK]	= { .type = NLA_U16 },
374 	[TCA_FLOWER_KEY_SCTP_SRC_MASK]	= { .type = NLA_U16 },
375 	[TCA_FLOWER_KEY_SCTP_DST_MASK]	= { .type = NLA_U16 },
376 	[TCA_FLOWER_KEY_SCTP_SRC]	= { .type = NLA_U16 },
377 	[TCA_FLOWER_KEY_SCTP_DST]	= { .type = NLA_U16 },
378 	[TCA_FLOWER_KEY_ENC_UDP_SRC_PORT]	= { .type = NLA_U16 },
379 	[TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK]	= { .type = NLA_U16 },
380 	[TCA_FLOWER_KEY_ENC_UDP_DST_PORT]	= { .type = NLA_U16 },
381 	[TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK]	= { .type = NLA_U16 },
382 	[TCA_FLOWER_KEY_FLAGS]		= { .type = NLA_U32 },
383 	[TCA_FLOWER_KEY_FLAGS_MASK]	= { .type = NLA_U32 },
384 	[TCA_FLOWER_KEY_ICMPV4_TYPE]	= { .type = NLA_U8 },
385 	[TCA_FLOWER_KEY_ICMPV4_TYPE_MASK] = { .type = NLA_U8 },
386 	[TCA_FLOWER_KEY_ICMPV4_CODE]	= { .type = NLA_U8 },
387 	[TCA_FLOWER_KEY_ICMPV4_CODE_MASK] = { .type = NLA_U8 },
388 	[TCA_FLOWER_KEY_ICMPV6_TYPE]	= { .type = NLA_U8 },
389 	[TCA_FLOWER_KEY_ICMPV6_TYPE_MASK] = { .type = NLA_U8 },
390 	[TCA_FLOWER_KEY_ICMPV6_CODE]	= { .type = NLA_U8 },
391 	[TCA_FLOWER_KEY_ICMPV6_CODE_MASK] = { .type = NLA_U8 },
392 	[TCA_FLOWER_KEY_ARP_SIP]	= { .type = NLA_U32 },
393 	[TCA_FLOWER_KEY_ARP_SIP_MASK]	= { .type = NLA_U32 },
394 	[TCA_FLOWER_KEY_ARP_TIP]	= { .type = NLA_U32 },
395 	[TCA_FLOWER_KEY_ARP_TIP_MASK]	= { .type = NLA_U32 },
396 	[TCA_FLOWER_KEY_ARP_OP]		= { .type = NLA_U8 },
397 	[TCA_FLOWER_KEY_ARP_OP_MASK]	= { .type = NLA_U8 },
398 	[TCA_FLOWER_KEY_ARP_SHA]	= { .len = ETH_ALEN },
399 	[TCA_FLOWER_KEY_ARP_SHA_MASK]	= { .len = ETH_ALEN },
400 	[TCA_FLOWER_KEY_ARP_THA]	= { .len = ETH_ALEN },
401 	[TCA_FLOWER_KEY_ARP_THA_MASK]	= { .len = ETH_ALEN },
402 	[TCA_FLOWER_KEY_MPLS_TTL]	= { .type = NLA_U8 },
403 	[TCA_FLOWER_KEY_MPLS_BOS]	= { .type = NLA_U8 },
404 	[TCA_FLOWER_KEY_MPLS_TC]	= { .type = NLA_U8 },
405 	[TCA_FLOWER_KEY_MPLS_LABEL]	= { .type = NLA_U32 },
406 	[TCA_FLOWER_KEY_TCP_FLAGS]	= { .type = NLA_U16 },
407 	[TCA_FLOWER_KEY_TCP_FLAGS_MASK]	= { .type = NLA_U16 },
408 	[TCA_FLOWER_KEY_IP_TOS]		= { .type = NLA_U8 },
409 	[TCA_FLOWER_KEY_IP_TOS_MASK]	= { .type = NLA_U8 },
410 	[TCA_FLOWER_KEY_IP_TTL]		= { .type = NLA_U8 },
411 	[TCA_FLOWER_KEY_IP_TTL_MASK]	= { .type = NLA_U8 },
412 };
413 
414 static void fl_set_key_val(struct nlattr **tb,
415 			   void *val, int val_type,
416 			   void *mask, int mask_type, int len)
417 {
418 	if (!tb[val_type])
419 		return;
420 	memcpy(val, nla_data(tb[val_type]), len);
421 	if (mask_type == TCA_FLOWER_UNSPEC || !tb[mask_type])
422 		memset(mask, 0xff, len);
423 	else
424 		memcpy(mask, nla_data(tb[mask_type]), len);
425 }
426 
427 static int fl_set_key_mpls(struct nlattr **tb,
428 			   struct flow_dissector_key_mpls *key_val,
429 			   struct flow_dissector_key_mpls *key_mask)
430 {
431 	if (tb[TCA_FLOWER_KEY_MPLS_TTL]) {
432 		key_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TTL]);
433 		key_mask->mpls_ttl = MPLS_TTL_MASK;
434 	}
435 	if (tb[TCA_FLOWER_KEY_MPLS_BOS]) {
436 		u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_BOS]);
437 
438 		if (bos & ~MPLS_BOS_MASK)
439 			return -EINVAL;
440 		key_val->mpls_bos = bos;
441 		key_mask->mpls_bos = MPLS_BOS_MASK;
442 	}
443 	if (tb[TCA_FLOWER_KEY_MPLS_TC]) {
444 		u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TC]);
445 
446 		if (tc & ~MPLS_TC_MASK)
447 			return -EINVAL;
448 		key_val->mpls_tc = tc;
449 		key_mask->mpls_tc = MPLS_TC_MASK;
450 	}
451 	if (tb[TCA_FLOWER_KEY_MPLS_LABEL]) {
452 		u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_LABEL]);
453 
454 		if (label & ~MPLS_LABEL_MASK)
455 			return -EINVAL;
456 		key_val->mpls_label = label;
457 		key_mask->mpls_label = MPLS_LABEL_MASK;
458 	}
459 	return 0;
460 }
461 
462 static void fl_set_key_vlan(struct nlattr **tb,
463 			    struct flow_dissector_key_vlan *key_val,
464 			    struct flow_dissector_key_vlan *key_mask)
465 {
466 #define VLAN_PRIORITY_MASK	0x7
467 
468 	if (tb[TCA_FLOWER_KEY_VLAN_ID]) {
469 		key_val->vlan_id =
470 			nla_get_u16(tb[TCA_FLOWER_KEY_VLAN_ID]) & VLAN_VID_MASK;
471 		key_mask->vlan_id = VLAN_VID_MASK;
472 	}
473 	if (tb[TCA_FLOWER_KEY_VLAN_PRIO]) {
474 		key_val->vlan_priority =
475 			nla_get_u8(tb[TCA_FLOWER_KEY_VLAN_PRIO]) &
476 			VLAN_PRIORITY_MASK;
477 		key_mask->vlan_priority = VLAN_PRIORITY_MASK;
478 	}
479 }
480 
481 static void fl_set_key_flag(u32 flower_key, u32 flower_mask,
482 			    u32 *dissector_key, u32 *dissector_mask,
483 			    u32 flower_flag_bit, u32 dissector_flag_bit)
484 {
485 	if (flower_mask & flower_flag_bit) {
486 		*dissector_mask |= dissector_flag_bit;
487 		if (flower_key & flower_flag_bit)
488 			*dissector_key |= dissector_flag_bit;
489 	}
490 }
491 
492 static int fl_set_key_flags(struct nlattr **tb,
493 			    u32 *flags_key, u32 *flags_mask)
494 {
495 	u32 key, mask;
496 
497 	/* mask is mandatory for flags */
498 	if (!tb[TCA_FLOWER_KEY_FLAGS_MASK])
499 		return -EINVAL;
500 
501 	key = be32_to_cpu(nla_get_u32(tb[TCA_FLOWER_KEY_FLAGS]));
502 	mask = be32_to_cpu(nla_get_u32(tb[TCA_FLOWER_KEY_FLAGS_MASK]));
503 
504 	*flags_key  = 0;
505 	*flags_mask = 0;
506 
507 	fl_set_key_flag(key, mask, flags_key, flags_mask,
508 			TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
509 
510 	return 0;
511 }
512 
513 static void fl_set_key_ip(struct nlattr **tb,
514 			  struct flow_dissector_key_ip *key,
515 			  struct flow_dissector_key_ip *mask)
516 {
517 		fl_set_key_val(tb, &key->tos, TCA_FLOWER_KEY_IP_TOS,
518 			       &mask->tos, TCA_FLOWER_KEY_IP_TOS_MASK,
519 			       sizeof(key->tos));
520 
521 		fl_set_key_val(tb, &key->ttl, TCA_FLOWER_KEY_IP_TTL,
522 			       &mask->ttl, TCA_FLOWER_KEY_IP_TTL_MASK,
523 			       sizeof(key->ttl));
524 }
525 
526 static int fl_set_key(struct net *net, struct nlattr **tb,
527 		      struct fl_flow_key *key, struct fl_flow_key *mask)
528 {
529 	__be16 ethertype;
530 	int ret = 0;
531 #ifdef CONFIG_NET_CLS_IND
532 	if (tb[TCA_FLOWER_INDEV]) {
533 		int err = tcf_change_indev(net, tb[TCA_FLOWER_INDEV]);
534 		if (err < 0)
535 			return err;
536 		key->indev_ifindex = err;
537 		mask->indev_ifindex = 0xffffffff;
538 	}
539 #endif
540 
541 	fl_set_key_val(tb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
542 		       mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
543 		       sizeof(key->eth.dst));
544 	fl_set_key_val(tb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
545 		       mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
546 		       sizeof(key->eth.src));
547 
548 	if (tb[TCA_FLOWER_KEY_ETH_TYPE]) {
549 		ethertype = nla_get_be16(tb[TCA_FLOWER_KEY_ETH_TYPE]);
550 
551 		if (ethertype == htons(ETH_P_8021Q)) {
552 			fl_set_key_vlan(tb, &key->vlan, &mask->vlan);
553 			fl_set_key_val(tb, &key->basic.n_proto,
554 				       TCA_FLOWER_KEY_VLAN_ETH_TYPE,
555 				       &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
556 				       sizeof(key->basic.n_proto));
557 		} else {
558 			key->basic.n_proto = ethertype;
559 			mask->basic.n_proto = cpu_to_be16(~0);
560 		}
561 	}
562 
563 	if (key->basic.n_proto == htons(ETH_P_IP) ||
564 	    key->basic.n_proto == htons(ETH_P_IPV6)) {
565 		fl_set_key_val(tb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
566 			       &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
567 			       sizeof(key->basic.ip_proto));
568 		fl_set_key_ip(tb, &key->ip, &mask->ip);
569 	}
570 
571 	if (tb[TCA_FLOWER_KEY_IPV4_SRC] || tb[TCA_FLOWER_KEY_IPV4_DST]) {
572 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
573 		mask->control.addr_type = ~0;
574 		fl_set_key_val(tb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
575 			       &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
576 			       sizeof(key->ipv4.src));
577 		fl_set_key_val(tb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
578 			       &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
579 			       sizeof(key->ipv4.dst));
580 	} else if (tb[TCA_FLOWER_KEY_IPV6_SRC] || tb[TCA_FLOWER_KEY_IPV6_DST]) {
581 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
582 		mask->control.addr_type = ~0;
583 		fl_set_key_val(tb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
584 			       &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
585 			       sizeof(key->ipv6.src));
586 		fl_set_key_val(tb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
587 			       &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
588 			       sizeof(key->ipv6.dst));
589 	}
590 
591 	if (key->basic.ip_proto == IPPROTO_TCP) {
592 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
593 			       &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
594 			       sizeof(key->tp.src));
595 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
596 			       &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
597 			       sizeof(key->tp.dst));
598 		fl_set_key_val(tb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
599 			       &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
600 			       sizeof(key->tcp.flags));
601 	} else if (key->basic.ip_proto == IPPROTO_UDP) {
602 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
603 			       &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
604 			       sizeof(key->tp.src));
605 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
606 			       &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
607 			       sizeof(key->tp.dst));
608 	} else if (key->basic.ip_proto == IPPROTO_SCTP) {
609 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
610 			       &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
611 			       sizeof(key->tp.src));
612 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
613 			       &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
614 			       sizeof(key->tp.dst));
615 	} else if (key->basic.n_proto == htons(ETH_P_IP) &&
616 		   key->basic.ip_proto == IPPROTO_ICMP) {
617 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV4_TYPE,
618 			       &mask->icmp.type,
619 			       TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
620 			       sizeof(key->icmp.type));
621 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV4_CODE,
622 			       &mask->icmp.code,
623 			       TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
624 			       sizeof(key->icmp.code));
625 	} else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
626 		   key->basic.ip_proto == IPPROTO_ICMPV6) {
627 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV6_TYPE,
628 			       &mask->icmp.type,
629 			       TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
630 			       sizeof(key->icmp.type));
631 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV6_CODE,
632 			       &mask->icmp.code,
633 			       TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
634 			       sizeof(key->icmp.code));
635 	} else if (key->basic.n_proto == htons(ETH_P_MPLS_UC) ||
636 		   key->basic.n_proto == htons(ETH_P_MPLS_MC)) {
637 		ret = fl_set_key_mpls(tb, &key->mpls, &mask->mpls);
638 		if (ret)
639 			return ret;
640 	} else if (key->basic.n_proto == htons(ETH_P_ARP) ||
641 		   key->basic.n_proto == htons(ETH_P_RARP)) {
642 		fl_set_key_val(tb, &key->arp.sip, TCA_FLOWER_KEY_ARP_SIP,
643 			       &mask->arp.sip, TCA_FLOWER_KEY_ARP_SIP_MASK,
644 			       sizeof(key->arp.sip));
645 		fl_set_key_val(tb, &key->arp.tip, TCA_FLOWER_KEY_ARP_TIP,
646 			       &mask->arp.tip, TCA_FLOWER_KEY_ARP_TIP_MASK,
647 			       sizeof(key->arp.tip));
648 		fl_set_key_val(tb, &key->arp.op, TCA_FLOWER_KEY_ARP_OP,
649 			       &mask->arp.op, TCA_FLOWER_KEY_ARP_OP_MASK,
650 			       sizeof(key->arp.op));
651 		fl_set_key_val(tb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
652 			       mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
653 			       sizeof(key->arp.sha));
654 		fl_set_key_val(tb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
655 			       mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
656 			       sizeof(key->arp.tha));
657 	}
658 
659 	if (tb[TCA_FLOWER_KEY_ENC_IPV4_SRC] ||
660 	    tb[TCA_FLOWER_KEY_ENC_IPV4_DST]) {
661 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
662 		mask->enc_control.addr_type = ~0;
663 		fl_set_key_val(tb, &key->enc_ipv4.src,
664 			       TCA_FLOWER_KEY_ENC_IPV4_SRC,
665 			       &mask->enc_ipv4.src,
666 			       TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
667 			       sizeof(key->enc_ipv4.src));
668 		fl_set_key_val(tb, &key->enc_ipv4.dst,
669 			       TCA_FLOWER_KEY_ENC_IPV4_DST,
670 			       &mask->enc_ipv4.dst,
671 			       TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
672 			       sizeof(key->enc_ipv4.dst));
673 	}
674 
675 	if (tb[TCA_FLOWER_KEY_ENC_IPV6_SRC] ||
676 	    tb[TCA_FLOWER_KEY_ENC_IPV6_DST]) {
677 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
678 		mask->enc_control.addr_type = ~0;
679 		fl_set_key_val(tb, &key->enc_ipv6.src,
680 			       TCA_FLOWER_KEY_ENC_IPV6_SRC,
681 			       &mask->enc_ipv6.src,
682 			       TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
683 			       sizeof(key->enc_ipv6.src));
684 		fl_set_key_val(tb, &key->enc_ipv6.dst,
685 			       TCA_FLOWER_KEY_ENC_IPV6_DST,
686 			       &mask->enc_ipv6.dst,
687 			       TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
688 			       sizeof(key->enc_ipv6.dst));
689 	}
690 
691 	fl_set_key_val(tb, &key->enc_key_id.keyid, TCA_FLOWER_KEY_ENC_KEY_ID,
692 		       &mask->enc_key_id.keyid, TCA_FLOWER_UNSPEC,
693 		       sizeof(key->enc_key_id.keyid));
694 
695 	fl_set_key_val(tb, &key->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
696 		       &mask->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
697 		       sizeof(key->enc_tp.src));
698 
699 	fl_set_key_val(tb, &key->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
700 		       &mask->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
701 		       sizeof(key->enc_tp.dst));
702 
703 	if (tb[TCA_FLOWER_KEY_FLAGS])
704 		ret = fl_set_key_flags(tb, &key->control.flags, &mask->control.flags);
705 
706 	return ret;
707 }
708 
709 static bool fl_mask_eq(struct fl_flow_mask *mask1,
710 		       struct fl_flow_mask *mask2)
711 {
712 	const long *lmask1 = fl_key_get_start(&mask1->key, mask1);
713 	const long *lmask2 = fl_key_get_start(&mask2->key, mask2);
714 
715 	return !memcmp(&mask1->range, &mask2->range, sizeof(mask1->range)) &&
716 	       !memcmp(lmask1, lmask2, fl_mask_range(mask1));
717 }
718 
719 static const struct rhashtable_params fl_ht_params = {
720 	.key_offset = offsetof(struct cls_fl_filter, mkey), /* base offset */
721 	.head_offset = offsetof(struct cls_fl_filter, ht_node),
722 	.automatic_shrinking = true,
723 };
724 
725 static int fl_init_hashtable(struct cls_fl_head *head,
726 			     struct fl_flow_mask *mask)
727 {
728 	head->ht_params = fl_ht_params;
729 	head->ht_params.key_len = fl_mask_range(mask);
730 	head->ht_params.key_offset += mask->range.start;
731 
732 	return rhashtable_init(&head->ht, &head->ht_params);
733 }
734 
735 #define FL_KEY_MEMBER_OFFSET(member) offsetof(struct fl_flow_key, member)
736 #define FL_KEY_MEMBER_SIZE(member) (sizeof(((struct fl_flow_key *) 0)->member))
737 
738 #define FL_KEY_IS_MASKED(mask, member)						\
739 	memchr_inv(((char *)mask) + FL_KEY_MEMBER_OFFSET(member),		\
740 		   0, FL_KEY_MEMBER_SIZE(member))				\
741 
742 #define FL_KEY_SET(keys, cnt, id, member)					\
743 	do {									\
744 		keys[cnt].key_id = id;						\
745 		keys[cnt].offset = FL_KEY_MEMBER_OFFSET(member);		\
746 		cnt++;								\
747 	} while(0);
748 
749 #define FL_KEY_SET_IF_MASKED(mask, keys, cnt, id, member)			\
750 	do {									\
751 		if (FL_KEY_IS_MASKED(mask, member))				\
752 			FL_KEY_SET(keys, cnt, id, member);			\
753 	} while(0);
754 
755 static void fl_init_dissector(struct cls_fl_head *head,
756 			      struct fl_flow_mask *mask)
757 {
758 	struct flow_dissector_key keys[FLOW_DISSECTOR_KEY_MAX];
759 	size_t cnt = 0;
760 
761 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_CONTROL, control);
762 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_BASIC, basic);
763 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
764 			     FLOW_DISSECTOR_KEY_ETH_ADDRS, eth);
765 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
766 			     FLOW_DISSECTOR_KEY_IPV4_ADDRS, ipv4);
767 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
768 			     FLOW_DISSECTOR_KEY_IPV6_ADDRS, ipv6);
769 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
770 			     FLOW_DISSECTOR_KEY_PORTS, tp);
771 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
772 			     FLOW_DISSECTOR_KEY_IP, ip);
773 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
774 			     FLOW_DISSECTOR_KEY_TCP, tcp);
775 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
776 			     FLOW_DISSECTOR_KEY_ICMP, icmp);
777 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
778 			     FLOW_DISSECTOR_KEY_ARP, arp);
779 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
780 			     FLOW_DISSECTOR_KEY_MPLS, mpls);
781 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
782 			     FLOW_DISSECTOR_KEY_VLAN, vlan);
783 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
784 			     FLOW_DISSECTOR_KEY_ENC_KEYID, enc_key_id);
785 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
786 			     FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, enc_ipv4);
787 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
788 			     FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, enc_ipv6);
789 	if (FL_KEY_IS_MASKED(&mask->key, enc_ipv4) ||
790 	    FL_KEY_IS_MASKED(&mask->key, enc_ipv6))
791 		FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_ENC_CONTROL,
792 			   enc_control);
793 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
794 			     FLOW_DISSECTOR_KEY_ENC_PORTS, enc_tp);
795 
796 	skb_flow_dissector_init(&head->dissector, keys, cnt);
797 }
798 
799 static int fl_check_assign_mask(struct cls_fl_head *head,
800 				struct fl_flow_mask *mask)
801 {
802 	int err;
803 
804 	if (head->mask_assigned) {
805 		if (!fl_mask_eq(&head->mask, mask))
806 			return -EINVAL;
807 		else
808 			return 0;
809 	}
810 
811 	/* Mask is not assigned yet. So assign it and init hashtable
812 	 * according to that.
813 	 */
814 	err = fl_init_hashtable(head, mask);
815 	if (err)
816 		return err;
817 	memcpy(&head->mask, mask, sizeof(head->mask));
818 	head->mask_assigned = true;
819 
820 	fl_init_dissector(head, mask);
821 
822 	return 0;
823 }
824 
825 static int fl_set_parms(struct net *net, struct tcf_proto *tp,
826 			struct cls_fl_filter *f, struct fl_flow_mask *mask,
827 			unsigned long base, struct nlattr **tb,
828 			struct nlattr *est, bool ovr)
829 {
830 	int err;
831 
832 	err = tcf_exts_validate(net, tp, tb, est, &f->exts, ovr);
833 	if (err < 0)
834 		return err;
835 
836 	if (tb[TCA_FLOWER_CLASSID]) {
837 		f->res.classid = nla_get_u32(tb[TCA_FLOWER_CLASSID]);
838 		tcf_bind_filter(tp, &f->res, base);
839 	}
840 
841 	err = fl_set_key(net, tb, &f->key, &mask->key);
842 	if (err)
843 		return err;
844 
845 	fl_mask_update_range(mask);
846 	fl_set_masked_key(&f->mkey, &f->key, mask);
847 
848 	return 0;
849 }
850 
851 static int fl_change(struct net *net, struct sk_buff *in_skb,
852 		     struct tcf_proto *tp, unsigned long base,
853 		     u32 handle, struct nlattr **tca,
854 		     void **arg, bool ovr)
855 {
856 	struct cls_fl_head *head = rtnl_dereference(tp->root);
857 	struct cls_fl_filter *fold = *arg;
858 	struct cls_fl_filter *fnew;
859 	struct nlattr **tb;
860 	struct fl_flow_mask mask = {};
861 	unsigned long idr_index;
862 	int err;
863 
864 	if (!tca[TCA_OPTIONS])
865 		return -EINVAL;
866 
867 	tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
868 	if (!tb)
869 		return -ENOBUFS;
870 
871 	err = nla_parse_nested(tb, TCA_FLOWER_MAX, tca[TCA_OPTIONS],
872 			       fl_policy, NULL);
873 	if (err < 0)
874 		goto errout_tb;
875 
876 	if (fold && handle && fold->handle != handle) {
877 		err = -EINVAL;
878 		goto errout_tb;
879 	}
880 
881 	fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
882 	if (!fnew) {
883 		err = -ENOBUFS;
884 		goto errout_tb;
885 	}
886 
887 	err = tcf_exts_init(&fnew->exts, TCA_FLOWER_ACT, 0);
888 	if (err < 0)
889 		goto errout;
890 
891 	if (!handle) {
892 		err = idr_alloc_ext(&head->handle_idr, fnew, &idr_index,
893 				    1, 0x80000000, GFP_KERNEL);
894 		if (err)
895 			goto errout;
896 		fnew->handle = idr_index;
897 	}
898 
899 	/* user specifies a handle and it doesn't exist */
900 	if (handle && !fold) {
901 		err = idr_alloc_ext(&head->handle_idr, fnew, &idr_index,
902 				    handle, handle + 1, GFP_KERNEL);
903 		if (err)
904 			goto errout;
905 		fnew->handle = idr_index;
906 	}
907 
908 	if (tb[TCA_FLOWER_FLAGS]) {
909 		fnew->flags = nla_get_u32(tb[TCA_FLOWER_FLAGS]);
910 
911 		if (!tc_flags_valid(fnew->flags)) {
912 			err = -EINVAL;
913 			goto errout_idr;
914 		}
915 	}
916 
917 	err = fl_set_parms(net, tp, fnew, &mask, base, tb, tca[TCA_RATE], ovr);
918 	if (err)
919 		goto errout_idr;
920 
921 	err = fl_check_assign_mask(head, &mask);
922 	if (err)
923 		goto errout_idr;
924 
925 	if (!tc_skip_sw(fnew->flags)) {
926 		if (!fold && fl_lookup(head, &fnew->mkey)) {
927 			err = -EEXIST;
928 			goto errout_idr;
929 		}
930 
931 		err = rhashtable_insert_fast(&head->ht, &fnew->ht_node,
932 					     head->ht_params);
933 		if (err)
934 			goto errout_idr;
935 	}
936 
937 	if (!tc_skip_hw(fnew->flags)) {
938 		err = fl_hw_replace_filter(tp,
939 					   &head->dissector,
940 					   &mask.key,
941 					   fnew);
942 		if (err)
943 			goto errout_idr;
944 	}
945 
946 	if (!tc_in_hw(fnew->flags))
947 		fnew->flags |= TCA_CLS_FLAGS_NOT_IN_HW;
948 
949 	if (fold) {
950 		if (!tc_skip_sw(fold->flags))
951 			rhashtable_remove_fast(&head->ht, &fold->ht_node,
952 					       head->ht_params);
953 		if (!tc_skip_hw(fold->flags))
954 			fl_hw_destroy_filter(tp, fold);
955 	}
956 
957 	*arg = fnew;
958 
959 	if (fold) {
960 		fnew->handle = handle;
961 		idr_replace_ext(&head->handle_idr, fnew, fnew->handle);
962 		list_replace_rcu(&fold->list, &fnew->list);
963 		tcf_unbind_filter(tp, &fold->res);
964 		tcf_exts_get_net(&fold->exts);
965 		call_rcu(&fold->rcu, fl_destroy_filter);
966 	} else {
967 		list_add_tail_rcu(&fnew->list, &head->filters);
968 	}
969 
970 	kfree(tb);
971 	return 0;
972 
973 errout_idr:
974 	if (fnew->handle)
975 		idr_remove_ext(&head->handle_idr, fnew->handle);
976 errout:
977 	tcf_exts_destroy(&fnew->exts);
978 	kfree(fnew);
979 errout_tb:
980 	kfree(tb);
981 	return err;
982 }
983 
984 static int fl_delete(struct tcf_proto *tp, void *arg, bool *last)
985 {
986 	struct cls_fl_head *head = rtnl_dereference(tp->root);
987 	struct cls_fl_filter *f = arg;
988 
989 	if (!tc_skip_sw(f->flags))
990 		rhashtable_remove_fast(&head->ht, &f->ht_node,
991 				       head->ht_params);
992 	__fl_delete(tp, f);
993 	*last = list_empty(&head->filters);
994 	return 0;
995 }
996 
997 static void fl_walk(struct tcf_proto *tp, struct tcf_walker *arg)
998 {
999 	struct cls_fl_head *head = rtnl_dereference(tp->root);
1000 	struct cls_fl_filter *f;
1001 
1002 	list_for_each_entry_rcu(f, &head->filters, list) {
1003 		if (arg->count < arg->skip)
1004 			goto skip;
1005 		if (arg->fn(tp, f, arg) < 0) {
1006 			arg->stop = 1;
1007 			break;
1008 		}
1009 skip:
1010 		arg->count++;
1011 	}
1012 }
1013 
1014 static int fl_dump_key_val(struct sk_buff *skb,
1015 			   void *val, int val_type,
1016 			   void *mask, int mask_type, int len)
1017 {
1018 	int err;
1019 
1020 	if (!memchr_inv(mask, 0, len))
1021 		return 0;
1022 	err = nla_put(skb, val_type, len, val);
1023 	if (err)
1024 		return err;
1025 	if (mask_type != TCA_FLOWER_UNSPEC) {
1026 		err = nla_put(skb, mask_type, len, mask);
1027 		if (err)
1028 			return err;
1029 	}
1030 	return 0;
1031 }
1032 
1033 static int fl_dump_key_mpls(struct sk_buff *skb,
1034 			    struct flow_dissector_key_mpls *mpls_key,
1035 			    struct flow_dissector_key_mpls *mpls_mask)
1036 {
1037 	int err;
1038 
1039 	if (!memchr_inv(mpls_mask, 0, sizeof(*mpls_mask)))
1040 		return 0;
1041 	if (mpls_mask->mpls_ttl) {
1042 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TTL,
1043 				 mpls_key->mpls_ttl);
1044 		if (err)
1045 			return err;
1046 	}
1047 	if (mpls_mask->mpls_tc) {
1048 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TC,
1049 				 mpls_key->mpls_tc);
1050 		if (err)
1051 			return err;
1052 	}
1053 	if (mpls_mask->mpls_label) {
1054 		err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_LABEL,
1055 				  mpls_key->mpls_label);
1056 		if (err)
1057 			return err;
1058 	}
1059 	if (mpls_mask->mpls_bos) {
1060 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_BOS,
1061 				 mpls_key->mpls_bos);
1062 		if (err)
1063 			return err;
1064 	}
1065 	return 0;
1066 }
1067 
1068 static int fl_dump_key_ip(struct sk_buff *skb,
1069 			  struct flow_dissector_key_ip *key,
1070 			  struct flow_dissector_key_ip *mask)
1071 {
1072 	if (fl_dump_key_val(skb, &key->tos, TCA_FLOWER_KEY_IP_TOS, &mask->tos,
1073 			    TCA_FLOWER_KEY_IP_TOS_MASK, sizeof(key->tos)) ||
1074 	    fl_dump_key_val(skb, &key->ttl, TCA_FLOWER_KEY_IP_TTL, &mask->ttl,
1075 			    TCA_FLOWER_KEY_IP_TTL_MASK, sizeof(key->ttl)))
1076 		return -1;
1077 
1078 	return 0;
1079 }
1080 
1081 static int fl_dump_key_vlan(struct sk_buff *skb,
1082 			    struct flow_dissector_key_vlan *vlan_key,
1083 			    struct flow_dissector_key_vlan *vlan_mask)
1084 {
1085 	int err;
1086 
1087 	if (!memchr_inv(vlan_mask, 0, sizeof(*vlan_mask)))
1088 		return 0;
1089 	if (vlan_mask->vlan_id) {
1090 		err = nla_put_u16(skb, TCA_FLOWER_KEY_VLAN_ID,
1091 				  vlan_key->vlan_id);
1092 		if (err)
1093 			return err;
1094 	}
1095 	if (vlan_mask->vlan_priority) {
1096 		err = nla_put_u8(skb, TCA_FLOWER_KEY_VLAN_PRIO,
1097 				 vlan_key->vlan_priority);
1098 		if (err)
1099 			return err;
1100 	}
1101 	return 0;
1102 }
1103 
1104 static void fl_get_key_flag(u32 dissector_key, u32 dissector_mask,
1105 			    u32 *flower_key, u32 *flower_mask,
1106 			    u32 flower_flag_bit, u32 dissector_flag_bit)
1107 {
1108 	if (dissector_mask & dissector_flag_bit) {
1109 		*flower_mask |= flower_flag_bit;
1110 		if (dissector_key & dissector_flag_bit)
1111 			*flower_key |= flower_flag_bit;
1112 	}
1113 }
1114 
1115 static int fl_dump_key_flags(struct sk_buff *skb, u32 flags_key, u32 flags_mask)
1116 {
1117 	u32 key, mask;
1118 	__be32 _key, _mask;
1119 	int err;
1120 
1121 	if (!memchr_inv(&flags_mask, 0, sizeof(flags_mask)))
1122 		return 0;
1123 
1124 	key = 0;
1125 	mask = 0;
1126 
1127 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
1128 			TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
1129 
1130 	_key = cpu_to_be32(key);
1131 	_mask = cpu_to_be32(mask);
1132 
1133 	err = nla_put(skb, TCA_FLOWER_KEY_FLAGS, 4, &_key);
1134 	if (err)
1135 		return err;
1136 
1137 	return nla_put(skb, TCA_FLOWER_KEY_FLAGS_MASK, 4, &_mask);
1138 }
1139 
1140 static int fl_dump(struct net *net, struct tcf_proto *tp, void *fh,
1141 		   struct sk_buff *skb, struct tcmsg *t)
1142 {
1143 	struct cls_fl_head *head = rtnl_dereference(tp->root);
1144 	struct cls_fl_filter *f = fh;
1145 	struct nlattr *nest;
1146 	struct fl_flow_key *key, *mask;
1147 
1148 	if (!f)
1149 		return skb->len;
1150 
1151 	t->tcm_handle = f->handle;
1152 
1153 	nest = nla_nest_start(skb, TCA_OPTIONS);
1154 	if (!nest)
1155 		goto nla_put_failure;
1156 
1157 	if (f->res.classid &&
1158 	    nla_put_u32(skb, TCA_FLOWER_CLASSID, f->res.classid))
1159 		goto nla_put_failure;
1160 
1161 	key = &f->key;
1162 	mask = &head->mask.key;
1163 
1164 	if (mask->indev_ifindex) {
1165 		struct net_device *dev;
1166 
1167 		dev = __dev_get_by_index(net, key->indev_ifindex);
1168 		if (dev && nla_put_string(skb, TCA_FLOWER_INDEV, dev->name))
1169 			goto nla_put_failure;
1170 	}
1171 
1172 	if (!tc_skip_hw(f->flags))
1173 		fl_hw_update_stats(tp, f);
1174 
1175 	if (fl_dump_key_val(skb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
1176 			    mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
1177 			    sizeof(key->eth.dst)) ||
1178 	    fl_dump_key_val(skb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
1179 			    mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
1180 			    sizeof(key->eth.src)) ||
1181 	    fl_dump_key_val(skb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE,
1182 			    &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
1183 			    sizeof(key->basic.n_proto)))
1184 		goto nla_put_failure;
1185 
1186 	if (fl_dump_key_mpls(skb, &key->mpls, &mask->mpls))
1187 		goto nla_put_failure;
1188 
1189 	if (fl_dump_key_vlan(skb, &key->vlan, &mask->vlan))
1190 		goto nla_put_failure;
1191 
1192 	if ((key->basic.n_proto == htons(ETH_P_IP) ||
1193 	     key->basic.n_proto == htons(ETH_P_IPV6)) &&
1194 	    (fl_dump_key_val(skb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
1195 			    &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
1196 			    sizeof(key->basic.ip_proto)) ||
1197 	    fl_dump_key_ip(skb, &key->ip, &mask->ip)))
1198 		goto nla_put_failure;
1199 
1200 	if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
1201 	    (fl_dump_key_val(skb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
1202 			     &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
1203 			     sizeof(key->ipv4.src)) ||
1204 	     fl_dump_key_val(skb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
1205 			     &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
1206 			     sizeof(key->ipv4.dst))))
1207 		goto nla_put_failure;
1208 	else if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
1209 		 (fl_dump_key_val(skb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
1210 				  &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
1211 				  sizeof(key->ipv6.src)) ||
1212 		  fl_dump_key_val(skb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
1213 				  &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
1214 				  sizeof(key->ipv6.dst))))
1215 		goto nla_put_failure;
1216 
1217 	if (key->basic.ip_proto == IPPROTO_TCP &&
1218 	    (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
1219 			     &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
1220 			     sizeof(key->tp.src)) ||
1221 	     fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
1222 			     &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
1223 			     sizeof(key->tp.dst)) ||
1224 	     fl_dump_key_val(skb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
1225 			     &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
1226 			     sizeof(key->tcp.flags))))
1227 		goto nla_put_failure;
1228 	else if (key->basic.ip_proto == IPPROTO_UDP &&
1229 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
1230 				  &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
1231 				  sizeof(key->tp.src)) ||
1232 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
1233 				  &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
1234 				  sizeof(key->tp.dst))))
1235 		goto nla_put_failure;
1236 	else if (key->basic.ip_proto == IPPROTO_SCTP &&
1237 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
1238 				  &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
1239 				  sizeof(key->tp.src)) ||
1240 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
1241 				  &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
1242 				  sizeof(key->tp.dst))))
1243 		goto nla_put_failure;
1244 	else if (key->basic.n_proto == htons(ETH_P_IP) &&
1245 		 key->basic.ip_proto == IPPROTO_ICMP &&
1246 		 (fl_dump_key_val(skb, &key->icmp.type,
1247 				  TCA_FLOWER_KEY_ICMPV4_TYPE, &mask->icmp.type,
1248 				  TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
1249 				  sizeof(key->icmp.type)) ||
1250 		  fl_dump_key_val(skb, &key->icmp.code,
1251 				  TCA_FLOWER_KEY_ICMPV4_CODE, &mask->icmp.code,
1252 				  TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
1253 				  sizeof(key->icmp.code))))
1254 		goto nla_put_failure;
1255 	else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
1256 		 key->basic.ip_proto == IPPROTO_ICMPV6 &&
1257 		 (fl_dump_key_val(skb, &key->icmp.type,
1258 				  TCA_FLOWER_KEY_ICMPV6_TYPE, &mask->icmp.type,
1259 				  TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
1260 				  sizeof(key->icmp.type)) ||
1261 		  fl_dump_key_val(skb, &key->icmp.code,
1262 				  TCA_FLOWER_KEY_ICMPV6_CODE, &mask->icmp.code,
1263 				  TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
1264 				  sizeof(key->icmp.code))))
1265 		goto nla_put_failure;
1266 	else if ((key->basic.n_proto == htons(ETH_P_ARP) ||
1267 		  key->basic.n_proto == htons(ETH_P_RARP)) &&
1268 		 (fl_dump_key_val(skb, &key->arp.sip,
1269 				  TCA_FLOWER_KEY_ARP_SIP, &mask->arp.sip,
1270 				  TCA_FLOWER_KEY_ARP_SIP_MASK,
1271 				  sizeof(key->arp.sip)) ||
1272 		  fl_dump_key_val(skb, &key->arp.tip,
1273 				  TCA_FLOWER_KEY_ARP_TIP, &mask->arp.tip,
1274 				  TCA_FLOWER_KEY_ARP_TIP_MASK,
1275 				  sizeof(key->arp.tip)) ||
1276 		  fl_dump_key_val(skb, &key->arp.op,
1277 				  TCA_FLOWER_KEY_ARP_OP, &mask->arp.op,
1278 				  TCA_FLOWER_KEY_ARP_OP_MASK,
1279 				  sizeof(key->arp.op)) ||
1280 		  fl_dump_key_val(skb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
1281 				  mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
1282 				  sizeof(key->arp.sha)) ||
1283 		  fl_dump_key_val(skb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
1284 				  mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
1285 				  sizeof(key->arp.tha))))
1286 		goto nla_put_failure;
1287 
1288 	if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
1289 	    (fl_dump_key_val(skb, &key->enc_ipv4.src,
1290 			    TCA_FLOWER_KEY_ENC_IPV4_SRC, &mask->enc_ipv4.src,
1291 			    TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
1292 			    sizeof(key->enc_ipv4.src)) ||
1293 	     fl_dump_key_val(skb, &key->enc_ipv4.dst,
1294 			     TCA_FLOWER_KEY_ENC_IPV4_DST, &mask->enc_ipv4.dst,
1295 			     TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
1296 			     sizeof(key->enc_ipv4.dst))))
1297 		goto nla_put_failure;
1298 	else if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
1299 		 (fl_dump_key_val(skb, &key->enc_ipv6.src,
1300 			    TCA_FLOWER_KEY_ENC_IPV6_SRC, &mask->enc_ipv6.src,
1301 			    TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
1302 			    sizeof(key->enc_ipv6.src)) ||
1303 		 fl_dump_key_val(skb, &key->enc_ipv6.dst,
1304 				 TCA_FLOWER_KEY_ENC_IPV6_DST,
1305 				 &mask->enc_ipv6.dst,
1306 				 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
1307 			    sizeof(key->enc_ipv6.dst))))
1308 		goto nla_put_failure;
1309 
1310 	if (fl_dump_key_val(skb, &key->enc_key_id, TCA_FLOWER_KEY_ENC_KEY_ID,
1311 			    &mask->enc_key_id, TCA_FLOWER_UNSPEC,
1312 			    sizeof(key->enc_key_id)) ||
1313 	    fl_dump_key_val(skb, &key->enc_tp.src,
1314 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
1315 			    &mask->enc_tp.src,
1316 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
1317 			    sizeof(key->enc_tp.src)) ||
1318 	    fl_dump_key_val(skb, &key->enc_tp.dst,
1319 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
1320 			    &mask->enc_tp.dst,
1321 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
1322 			    sizeof(key->enc_tp.dst)))
1323 		goto nla_put_failure;
1324 
1325 	if (fl_dump_key_flags(skb, key->control.flags, mask->control.flags))
1326 		goto nla_put_failure;
1327 
1328 	if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
1329 		goto nla_put_failure;
1330 
1331 	if (tcf_exts_dump(skb, &f->exts))
1332 		goto nla_put_failure;
1333 
1334 	nla_nest_end(skb, nest);
1335 
1336 	if (tcf_exts_dump_stats(skb, &f->exts) < 0)
1337 		goto nla_put_failure;
1338 
1339 	return skb->len;
1340 
1341 nla_put_failure:
1342 	nla_nest_cancel(skb, nest);
1343 	return -1;
1344 }
1345 
1346 static void fl_bind_class(void *fh, u32 classid, unsigned long cl)
1347 {
1348 	struct cls_fl_filter *f = fh;
1349 
1350 	if (f && f->res.classid == classid)
1351 		f->res.class = cl;
1352 }
1353 
1354 static struct tcf_proto_ops cls_fl_ops __read_mostly = {
1355 	.kind		= "flower",
1356 	.classify	= fl_classify,
1357 	.init		= fl_init,
1358 	.destroy	= fl_destroy,
1359 	.get		= fl_get,
1360 	.change		= fl_change,
1361 	.delete		= fl_delete,
1362 	.walk		= fl_walk,
1363 	.dump		= fl_dump,
1364 	.bind_class	= fl_bind_class,
1365 	.owner		= THIS_MODULE,
1366 };
1367 
1368 static int __init cls_fl_init(void)
1369 {
1370 	return register_tcf_proto_ops(&cls_fl_ops);
1371 }
1372 
1373 static void __exit cls_fl_exit(void)
1374 {
1375 	unregister_tcf_proto_ops(&cls_fl_ops);
1376 }
1377 
1378 module_init(cls_fl_init);
1379 module_exit(cls_fl_exit);
1380 
1381 MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
1382 MODULE_DESCRIPTION("Flower classifier");
1383 MODULE_LICENSE("GPL v2");
1384