xref: /linux/net/sched/cls_flower.c (revision a8b70ccf10e38775785d9cb12ead916474549f99)
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_tunnel_info(skb, &head->dissector, &skb_key);
170 	skb_flow_dissect(skb, &head->dissector, &skb_key, 0);
171 
172 	fl_set_masked_key(&skb_mkey, &skb_key, &head->mask);
173 
174 	f = fl_lookup(head, &skb_mkey);
175 	if (f && !tc_skip_sw(f->flags)) {
176 		*res = f->res;
177 		return tcf_exts_exec(skb, &f->exts, res);
178 	}
179 	return -1;
180 }
181 
182 static int fl_init(struct tcf_proto *tp)
183 {
184 	struct cls_fl_head *head;
185 
186 	head = kzalloc(sizeof(*head), GFP_KERNEL);
187 	if (!head)
188 		return -ENOBUFS;
189 
190 	INIT_LIST_HEAD_RCU(&head->filters);
191 	rcu_assign_pointer(tp->root, head);
192 	idr_init(&head->handle_idr);
193 
194 	return 0;
195 }
196 
197 static void __fl_destroy_filter(struct cls_fl_filter *f)
198 {
199 	tcf_exts_destroy(&f->exts);
200 	tcf_exts_put_net(&f->exts);
201 	kfree(f);
202 }
203 
204 static void fl_destroy_filter_work(struct work_struct *work)
205 {
206 	struct cls_fl_filter *f = container_of(work, struct cls_fl_filter, work);
207 
208 	rtnl_lock();
209 	__fl_destroy_filter(f);
210 	rtnl_unlock();
211 }
212 
213 static void fl_destroy_filter(struct rcu_head *head)
214 {
215 	struct cls_fl_filter *f = container_of(head, struct cls_fl_filter, rcu);
216 
217 	INIT_WORK(&f->work, fl_destroy_filter_work);
218 	tcf_queue_work(&f->work);
219 }
220 
221 static void fl_hw_destroy_filter(struct tcf_proto *tp, struct cls_fl_filter *f,
222 				 struct netlink_ext_ack *extack)
223 {
224 	struct tc_cls_flower_offload cls_flower = {};
225 	struct tcf_block *block = tp->chain->block;
226 
227 	tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
228 	cls_flower.command = TC_CLSFLOWER_DESTROY;
229 	cls_flower.cookie = (unsigned long) f;
230 
231 	tc_setup_cb_call(block, &f->exts, TC_SETUP_CLSFLOWER,
232 			 &cls_flower, false);
233 	tcf_block_offload_dec(block, &f->flags);
234 }
235 
236 static int fl_hw_replace_filter(struct tcf_proto *tp,
237 				struct flow_dissector *dissector,
238 				struct fl_flow_key *mask,
239 				struct cls_fl_filter *f,
240 				struct netlink_ext_ack *extack)
241 {
242 	struct tc_cls_flower_offload cls_flower = {};
243 	struct tcf_block *block = tp->chain->block;
244 	bool skip_sw = tc_skip_sw(f->flags);
245 	int err;
246 
247 	tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
248 	cls_flower.command = TC_CLSFLOWER_REPLACE;
249 	cls_flower.cookie = (unsigned long) f;
250 	cls_flower.dissector = dissector;
251 	cls_flower.mask = mask;
252 	cls_flower.key = &f->mkey;
253 	cls_flower.exts = &f->exts;
254 	cls_flower.classid = f->res.classid;
255 
256 	err = tc_setup_cb_call(block, &f->exts, TC_SETUP_CLSFLOWER,
257 			       &cls_flower, skip_sw);
258 	if (err < 0) {
259 		fl_hw_destroy_filter(tp, f, NULL);
260 		return err;
261 	} else if (err > 0) {
262 		tcf_block_offload_inc(block, &f->flags);
263 	}
264 
265 	if (skip_sw && !(f->flags & TCA_CLS_FLAGS_IN_HW))
266 		return -EINVAL;
267 
268 	return 0;
269 }
270 
271 static void fl_hw_update_stats(struct tcf_proto *tp, struct cls_fl_filter *f)
272 {
273 	struct tc_cls_flower_offload cls_flower = {};
274 	struct tcf_block *block = tp->chain->block;
275 
276 	tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, NULL);
277 	cls_flower.command = TC_CLSFLOWER_STATS;
278 	cls_flower.cookie = (unsigned long) f;
279 	cls_flower.exts = &f->exts;
280 	cls_flower.classid = f->res.classid;
281 
282 	tc_setup_cb_call(block, &f->exts, TC_SETUP_CLSFLOWER,
283 			 &cls_flower, false);
284 }
285 
286 static void __fl_delete(struct tcf_proto *tp, struct cls_fl_filter *f,
287 			struct netlink_ext_ack *extack)
288 {
289 	struct cls_fl_head *head = rtnl_dereference(tp->root);
290 
291 	idr_remove(&head->handle_idr, f->handle);
292 	list_del_rcu(&f->list);
293 	if (!tc_skip_hw(f->flags))
294 		fl_hw_destroy_filter(tp, f, extack);
295 	tcf_unbind_filter(tp, &f->res);
296 	if (tcf_exts_get_net(&f->exts))
297 		call_rcu(&f->rcu, fl_destroy_filter);
298 	else
299 		__fl_destroy_filter(f);
300 }
301 
302 static void fl_destroy_sleepable(struct work_struct *work)
303 {
304 	struct cls_fl_head *head = container_of(work, struct cls_fl_head,
305 						work);
306 	if (head->mask_assigned)
307 		rhashtable_destroy(&head->ht);
308 	kfree(head);
309 	module_put(THIS_MODULE);
310 }
311 
312 static void fl_destroy_rcu(struct rcu_head *rcu)
313 {
314 	struct cls_fl_head *head = container_of(rcu, struct cls_fl_head, rcu);
315 
316 	INIT_WORK(&head->work, fl_destroy_sleepable);
317 	schedule_work(&head->work);
318 }
319 
320 static void fl_destroy(struct tcf_proto *tp, struct netlink_ext_ack *extack)
321 {
322 	struct cls_fl_head *head = rtnl_dereference(tp->root);
323 	struct cls_fl_filter *f, *next;
324 
325 	list_for_each_entry_safe(f, next, &head->filters, list)
326 		__fl_delete(tp, f, extack);
327 	idr_destroy(&head->handle_idr);
328 
329 	__module_get(THIS_MODULE);
330 	call_rcu(&head->rcu, fl_destroy_rcu);
331 }
332 
333 static void *fl_get(struct tcf_proto *tp, u32 handle)
334 {
335 	struct cls_fl_head *head = rtnl_dereference(tp->root);
336 
337 	return idr_find(&head->handle_idr, handle);
338 }
339 
340 static const struct nla_policy fl_policy[TCA_FLOWER_MAX + 1] = {
341 	[TCA_FLOWER_UNSPEC]		= { .type = NLA_UNSPEC },
342 	[TCA_FLOWER_CLASSID]		= { .type = NLA_U32 },
343 	[TCA_FLOWER_INDEV]		= { .type = NLA_STRING,
344 					    .len = IFNAMSIZ },
345 	[TCA_FLOWER_KEY_ETH_DST]	= { .len = ETH_ALEN },
346 	[TCA_FLOWER_KEY_ETH_DST_MASK]	= { .len = ETH_ALEN },
347 	[TCA_FLOWER_KEY_ETH_SRC]	= { .len = ETH_ALEN },
348 	[TCA_FLOWER_KEY_ETH_SRC_MASK]	= { .len = ETH_ALEN },
349 	[TCA_FLOWER_KEY_ETH_TYPE]	= { .type = NLA_U16 },
350 	[TCA_FLOWER_KEY_IP_PROTO]	= { .type = NLA_U8 },
351 	[TCA_FLOWER_KEY_IPV4_SRC]	= { .type = NLA_U32 },
352 	[TCA_FLOWER_KEY_IPV4_SRC_MASK]	= { .type = NLA_U32 },
353 	[TCA_FLOWER_KEY_IPV4_DST]	= { .type = NLA_U32 },
354 	[TCA_FLOWER_KEY_IPV4_DST_MASK]	= { .type = NLA_U32 },
355 	[TCA_FLOWER_KEY_IPV6_SRC]	= { .len = sizeof(struct in6_addr) },
356 	[TCA_FLOWER_KEY_IPV6_SRC_MASK]	= { .len = sizeof(struct in6_addr) },
357 	[TCA_FLOWER_KEY_IPV6_DST]	= { .len = sizeof(struct in6_addr) },
358 	[TCA_FLOWER_KEY_IPV6_DST_MASK]	= { .len = sizeof(struct in6_addr) },
359 	[TCA_FLOWER_KEY_TCP_SRC]	= { .type = NLA_U16 },
360 	[TCA_FLOWER_KEY_TCP_DST]	= { .type = NLA_U16 },
361 	[TCA_FLOWER_KEY_UDP_SRC]	= { .type = NLA_U16 },
362 	[TCA_FLOWER_KEY_UDP_DST]	= { .type = NLA_U16 },
363 	[TCA_FLOWER_KEY_VLAN_ID]	= { .type = NLA_U16 },
364 	[TCA_FLOWER_KEY_VLAN_PRIO]	= { .type = NLA_U8 },
365 	[TCA_FLOWER_KEY_VLAN_ETH_TYPE]	= { .type = NLA_U16 },
366 	[TCA_FLOWER_KEY_ENC_KEY_ID]	= { .type = NLA_U32 },
367 	[TCA_FLOWER_KEY_ENC_IPV4_SRC]	= { .type = NLA_U32 },
368 	[TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK] = { .type = NLA_U32 },
369 	[TCA_FLOWER_KEY_ENC_IPV4_DST]	= { .type = NLA_U32 },
370 	[TCA_FLOWER_KEY_ENC_IPV4_DST_MASK] = { .type = NLA_U32 },
371 	[TCA_FLOWER_KEY_ENC_IPV6_SRC]	= { .len = sizeof(struct in6_addr) },
372 	[TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK] = { .len = sizeof(struct in6_addr) },
373 	[TCA_FLOWER_KEY_ENC_IPV6_DST]	= { .len = sizeof(struct in6_addr) },
374 	[TCA_FLOWER_KEY_ENC_IPV6_DST_MASK] = { .len = sizeof(struct in6_addr) },
375 	[TCA_FLOWER_KEY_TCP_SRC_MASK]	= { .type = NLA_U16 },
376 	[TCA_FLOWER_KEY_TCP_DST_MASK]	= { .type = NLA_U16 },
377 	[TCA_FLOWER_KEY_UDP_SRC_MASK]	= { .type = NLA_U16 },
378 	[TCA_FLOWER_KEY_UDP_DST_MASK]	= { .type = NLA_U16 },
379 	[TCA_FLOWER_KEY_SCTP_SRC_MASK]	= { .type = NLA_U16 },
380 	[TCA_FLOWER_KEY_SCTP_DST_MASK]	= { .type = NLA_U16 },
381 	[TCA_FLOWER_KEY_SCTP_SRC]	= { .type = NLA_U16 },
382 	[TCA_FLOWER_KEY_SCTP_DST]	= { .type = NLA_U16 },
383 	[TCA_FLOWER_KEY_ENC_UDP_SRC_PORT]	= { .type = NLA_U16 },
384 	[TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK]	= { .type = NLA_U16 },
385 	[TCA_FLOWER_KEY_ENC_UDP_DST_PORT]	= { .type = NLA_U16 },
386 	[TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK]	= { .type = NLA_U16 },
387 	[TCA_FLOWER_KEY_FLAGS]		= { .type = NLA_U32 },
388 	[TCA_FLOWER_KEY_FLAGS_MASK]	= { .type = NLA_U32 },
389 	[TCA_FLOWER_KEY_ICMPV4_TYPE]	= { .type = NLA_U8 },
390 	[TCA_FLOWER_KEY_ICMPV4_TYPE_MASK] = { .type = NLA_U8 },
391 	[TCA_FLOWER_KEY_ICMPV4_CODE]	= { .type = NLA_U8 },
392 	[TCA_FLOWER_KEY_ICMPV4_CODE_MASK] = { .type = NLA_U8 },
393 	[TCA_FLOWER_KEY_ICMPV6_TYPE]	= { .type = NLA_U8 },
394 	[TCA_FLOWER_KEY_ICMPV6_TYPE_MASK] = { .type = NLA_U8 },
395 	[TCA_FLOWER_KEY_ICMPV6_CODE]	= { .type = NLA_U8 },
396 	[TCA_FLOWER_KEY_ICMPV6_CODE_MASK] = { .type = NLA_U8 },
397 	[TCA_FLOWER_KEY_ARP_SIP]	= { .type = NLA_U32 },
398 	[TCA_FLOWER_KEY_ARP_SIP_MASK]	= { .type = NLA_U32 },
399 	[TCA_FLOWER_KEY_ARP_TIP]	= { .type = NLA_U32 },
400 	[TCA_FLOWER_KEY_ARP_TIP_MASK]	= { .type = NLA_U32 },
401 	[TCA_FLOWER_KEY_ARP_OP]		= { .type = NLA_U8 },
402 	[TCA_FLOWER_KEY_ARP_OP_MASK]	= { .type = NLA_U8 },
403 	[TCA_FLOWER_KEY_ARP_SHA]	= { .len = ETH_ALEN },
404 	[TCA_FLOWER_KEY_ARP_SHA_MASK]	= { .len = ETH_ALEN },
405 	[TCA_FLOWER_KEY_ARP_THA]	= { .len = ETH_ALEN },
406 	[TCA_FLOWER_KEY_ARP_THA_MASK]	= { .len = ETH_ALEN },
407 	[TCA_FLOWER_KEY_MPLS_TTL]	= { .type = NLA_U8 },
408 	[TCA_FLOWER_KEY_MPLS_BOS]	= { .type = NLA_U8 },
409 	[TCA_FLOWER_KEY_MPLS_TC]	= { .type = NLA_U8 },
410 	[TCA_FLOWER_KEY_MPLS_LABEL]	= { .type = NLA_U32 },
411 	[TCA_FLOWER_KEY_TCP_FLAGS]	= { .type = NLA_U16 },
412 	[TCA_FLOWER_KEY_TCP_FLAGS_MASK]	= { .type = NLA_U16 },
413 	[TCA_FLOWER_KEY_IP_TOS]		= { .type = NLA_U8 },
414 	[TCA_FLOWER_KEY_IP_TOS_MASK]	= { .type = NLA_U8 },
415 	[TCA_FLOWER_KEY_IP_TTL]		= { .type = NLA_U8 },
416 	[TCA_FLOWER_KEY_IP_TTL_MASK]	= { .type = NLA_U8 },
417 };
418 
419 static void fl_set_key_val(struct nlattr **tb,
420 			   void *val, int val_type,
421 			   void *mask, int mask_type, int len)
422 {
423 	if (!tb[val_type])
424 		return;
425 	memcpy(val, nla_data(tb[val_type]), len);
426 	if (mask_type == TCA_FLOWER_UNSPEC || !tb[mask_type])
427 		memset(mask, 0xff, len);
428 	else
429 		memcpy(mask, nla_data(tb[mask_type]), len);
430 }
431 
432 static int fl_set_key_mpls(struct nlattr **tb,
433 			   struct flow_dissector_key_mpls *key_val,
434 			   struct flow_dissector_key_mpls *key_mask)
435 {
436 	if (tb[TCA_FLOWER_KEY_MPLS_TTL]) {
437 		key_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TTL]);
438 		key_mask->mpls_ttl = MPLS_TTL_MASK;
439 	}
440 	if (tb[TCA_FLOWER_KEY_MPLS_BOS]) {
441 		u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_BOS]);
442 
443 		if (bos & ~MPLS_BOS_MASK)
444 			return -EINVAL;
445 		key_val->mpls_bos = bos;
446 		key_mask->mpls_bos = MPLS_BOS_MASK;
447 	}
448 	if (tb[TCA_FLOWER_KEY_MPLS_TC]) {
449 		u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TC]);
450 
451 		if (tc & ~MPLS_TC_MASK)
452 			return -EINVAL;
453 		key_val->mpls_tc = tc;
454 		key_mask->mpls_tc = MPLS_TC_MASK;
455 	}
456 	if (tb[TCA_FLOWER_KEY_MPLS_LABEL]) {
457 		u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_LABEL]);
458 
459 		if (label & ~MPLS_LABEL_MASK)
460 			return -EINVAL;
461 		key_val->mpls_label = label;
462 		key_mask->mpls_label = MPLS_LABEL_MASK;
463 	}
464 	return 0;
465 }
466 
467 static void fl_set_key_vlan(struct nlattr **tb,
468 			    struct flow_dissector_key_vlan *key_val,
469 			    struct flow_dissector_key_vlan *key_mask)
470 {
471 #define VLAN_PRIORITY_MASK	0x7
472 
473 	if (tb[TCA_FLOWER_KEY_VLAN_ID]) {
474 		key_val->vlan_id =
475 			nla_get_u16(tb[TCA_FLOWER_KEY_VLAN_ID]) & VLAN_VID_MASK;
476 		key_mask->vlan_id = VLAN_VID_MASK;
477 	}
478 	if (tb[TCA_FLOWER_KEY_VLAN_PRIO]) {
479 		key_val->vlan_priority =
480 			nla_get_u8(tb[TCA_FLOWER_KEY_VLAN_PRIO]) &
481 			VLAN_PRIORITY_MASK;
482 		key_mask->vlan_priority = VLAN_PRIORITY_MASK;
483 	}
484 }
485 
486 static void fl_set_key_flag(u32 flower_key, u32 flower_mask,
487 			    u32 *dissector_key, u32 *dissector_mask,
488 			    u32 flower_flag_bit, u32 dissector_flag_bit)
489 {
490 	if (flower_mask & flower_flag_bit) {
491 		*dissector_mask |= dissector_flag_bit;
492 		if (flower_key & flower_flag_bit)
493 			*dissector_key |= dissector_flag_bit;
494 	}
495 }
496 
497 static int fl_set_key_flags(struct nlattr **tb,
498 			    u32 *flags_key, u32 *flags_mask)
499 {
500 	u32 key, mask;
501 
502 	/* mask is mandatory for flags */
503 	if (!tb[TCA_FLOWER_KEY_FLAGS_MASK])
504 		return -EINVAL;
505 
506 	key = be32_to_cpu(nla_get_u32(tb[TCA_FLOWER_KEY_FLAGS]));
507 	mask = be32_to_cpu(nla_get_u32(tb[TCA_FLOWER_KEY_FLAGS_MASK]));
508 
509 	*flags_key  = 0;
510 	*flags_mask = 0;
511 
512 	fl_set_key_flag(key, mask, flags_key, flags_mask,
513 			TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
514 	fl_set_key_flag(key, mask, flags_key, flags_mask,
515 			TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
516 			FLOW_DIS_FIRST_FRAG);
517 
518 	return 0;
519 }
520 
521 static void fl_set_key_ip(struct nlattr **tb,
522 			  struct flow_dissector_key_ip *key,
523 			  struct flow_dissector_key_ip *mask)
524 {
525 		fl_set_key_val(tb, &key->tos, TCA_FLOWER_KEY_IP_TOS,
526 			       &mask->tos, TCA_FLOWER_KEY_IP_TOS_MASK,
527 			       sizeof(key->tos));
528 
529 		fl_set_key_val(tb, &key->ttl, TCA_FLOWER_KEY_IP_TTL,
530 			       &mask->ttl, TCA_FLOWER_KEY_IP_TTL_MASK,
531 			       sizeof(key->ttl));
532 }
533 
534 static int fl_set_key(struct net *net, struct nlattr **tb,
535 		      struct fl_flow_key *key, struct fl_flow_key *mask,
536 		      struct netlink_ext_ack *extack)
537 {
538 	__be16 ethertype;
539 	int ret = 0;
540 #ifdef CONFIG_NET_CLS_IND
541 	if (tb[TCA_FLOWER_INDEV]) {
542 		int err = tcf_change_indev(net, tb[TCA_FLOWER_INDEV], extack);
543 		if (err < 0)
544 			return err;
545 		key->indev_ifindex = err;
546 		mask->indev_ifindex = 0xffffffff;
547 	}
548 #endif
549 
550 	fl_set_key_val(tb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
551 		       mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
552 		       sizeof(key->eth.dst));
553 	fl_set_key_val(tb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
554 		       mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
555 		       sizeof(key->eth.src));
556 
557 	if (tb[TCA_FLOWER_KEY_ETH_TYPE]) {
558 		ethertype = nla_get_be16(tb[TCA_FLOWER_KEY_ETH_TYPE]);
559 
560 		if (ethertype == htons(ETH_P_8021Q)) {
561 			fl_set_key_vlan(tb, &key->vlan, &mask->vlan);
562 			fl_set_key_val(tb, &key->basic.n_proto,
563 				       TCA_FLOWER_KEY_VLAN_ETH_TYPE,
564 				       &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
565 				       sizeof(key->basic.n_proto));
566 		} else {
567 			key->basic.n_proto = ethertype;
568 			mask->basic.n_proto = cpu_to_be16(~0);
569 		}
570 	}
571 
572 	if (key->basic.n_proto == htons(ETH_P_IP) ||
573 	    key->basic.n_proto == htons(ETH_P_IPV6)) {
574 		fl_set_key_val(tb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
575 			       &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
576 			       sizeof(key->basic.ip_proto));
577 		fl_set_key_ip(tb, &key->ip, &mask->ip);
578 	}
579 
580 	if (tb[TCA_FLOWER_KEY_IPV4_SRC] || tb[TCA_FLOWER_KEY_IPV4_DST]) {
581 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
582 		mask->control.addr_type = ~0;
583 		fl_set_key_val(tb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
584 			       &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
585 			       sizeof(key->ipv4.src));
586 		fl_set_key_val(tb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
587 			       &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
588 			       sizeof(key->ipv4.dst));
589 	} else if (tb[TCA_FLOWER_KEY_IPV6_SRC] || tb[TCA_FLOWER_KEY_IPV6_DST]) {
590 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
591 		mask->control.addr_type = ~0;
592 		fl_set_key_val(tb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
593 			       &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
594 			       sizeof(key->ipv6.src));
595 		fl_set_key_val(tb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
596 			       &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
597 			       sizeof(key->ipv6.dst));
598 	}
599 
600 	if (key->basic.ip_proto == IPPROTO_TCP) {
601 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
602 			       &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
603 			       sizeof(key->tp.src));
604 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
605 			       &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
606 			       sizeof(key->tp.dst));
607 		fl_set_key_val(tb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
608 			       &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
609 			       sizeof(key->tcp.flags));
610 	} else if (key->basic.ip_proto == IPPROTO_UDP) {
611 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
612 			       &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
613 			       sizeof(key->tp.src));
614 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
615 			       &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
616 			       sizeof(key->tp.dst));
617 	} else if (key->basic.ip_proto == IPPROTO_SCTP) {
618 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
619 			       &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
620 			       sizeof(key->tp.src));
621 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
622 			       &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
623 			       sizeof(key->tp.dst));
624 	} else if (key->basic.n_proto == htons(ETH_P_IP) &&
625 		   key->basic.ip_proto == IPPROTO_ICMP) {
626 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV4_TYPE,
627 			       &mask->icmp.type,
628 			       TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
629 			       sizeof(key->icmp.type));
630 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV4_CODE,
631 			       &mask->icmp.code,
632 			       TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
633 			       sizeof(key->icmp.code));
634 	} else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
635 		   key->basic.ip_proto == IPPROTO_ICMPV6) {
636 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV6_TYPE,
637 			       &mask->icmp.type,
638 			       TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
639 			       sizeof(key->icmp.type));
640 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV6_CODE,
641 			       &mask->icmp.code,
642 			       TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
643 			       sizeof(key->icmp.code));
644 	} else if (key->basic.n_proto == htons(ETH_P_MPLS_UC) ||
645 		   key->basic.n_proto == htons(ETH_P_MPLS_MC)) {
646 		ret = fl_set_key_mpls(tb, &key->mpls, &mask->mpls);
647 		if (ret)
648 			return ret;
649 	} else if (key->basic.n_proto == htons(ETH_P_ARP) ||
650 		   key->basic.n_proto == htons(ETH_P_RARP)) {
651 		fl_set_key_val(tb, &key->arp.sip, TCA_FLOWER_KEY_ARP_SIP,
652 			       &mask->arp.sip, TCA_FLOWER_KEY_ARP_SIP_MASK,
653 			       sizeof(key->arp.sip));
654 		fl_set_key_val(tb, &key->arp.tip, TCA_FLOWER_KEY_ARP_TIP,
655 			       &mask->arp.tip, TCA_FLOWER_KEY_ARP_TIP_MASK,
656 			       sizeof(key->arp.tip));
657 		fl_set_key_val(tb, &key->arp.op, TCA_FLOWER_KEY_ARP_OP,
658 			       &mask->arp.op, TCA_FLOWER_KEY_ARP_OP_MASK,
659 			       sizeof(key->arp.op));
660 		fl_set_key_val(tb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
661 			       mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
662 			       sizeof(key->arp.sha));
663 		fl_set_key_val(tb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
664 			       mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
665 			       sizeof(key->arp.tha));
666 	}
667 
668 	if (tb[TCA_FLOWER_KEY_ENC_IPV4_SRC] ||
669 	    tb[TCA_FLOWER_KEY_ENC_IPV4_DST]) {
670 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
671 		mask->enc_control.addr_type = ~0;
672 		fl_set_key_val(tb, &key->enc_ipv4.src,
673 			       TCA_FLOWER_KEY_ENC_IPV4_SRC,
674 			       &mask->enc_ipv4.src,
675 			       TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
676 			       sizeof(key->enc_ipv4.src));
677 		fl_set_key_val(tb, &key->enc_ipv4.dst,
678 			       TCA_FLOWER_KEY_ENC_IPV4_DST,
679 			       &mask->enc_ipv4.dst,
680 			       TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
681 			       sizeof(key->enc_ipv4.dst));
682 	}
683 
684 	if (tb[TCA_FLOWER_KEY_ENC_IPV6_SRC] ||
685 	    tb[TCA_FLOWER_KEY_ENC_IPV6_DST]) {
686 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
687 		mask->enc_control.addr_type = ~0;
688 		fl_set_key_val(tb, &key->enc_ipv6.src,
689 			       TCA_FLOWER_KEY_ENC_IPV6_SRC,
690 			       &mask->enc_ipv6.src,
691 			       TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
692 			       sizeof(key->enc_ipv6.src));
693 		fl_set_key_val(tb, &key->enc_ipv6.dst,
694 			       TCA_FLOWER_KEY_ENC_IPV6_DST,
695 			       &mask->enc_ipv6.dst,
696 			       TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
697 			       sizeof(key->enc_ipv6.dst));
698 	}
699 
700 	fl_set_key_val(tb, &key->enc_key_id.keyid, TCA_FLOWER_KEY_ENC_KEY_ID,
701 		       &mask->enc_key_id.keyid, TCA_FLOWER_UNSPEC,
702 		       sizeof(key->enc_key_id.keyid));
703 
704 	fl_set_key_val(tb, &key->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
705 		       &mask->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
706 		       sizeof(key->enc_tp.src));
707 
708 	fl_set_key_val(tb, &key->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
709 		       &mask->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
710 		       sizeof(key->enc_tp.dst));
711 
712 	if (tb[TCA_FLOWER_KEY_FLAGS])
713 		ret = fl_set_key_flags(tb, &key->control.flags, &mask->control.flags);
714 
715 	return ret;
716 }
717 
718 static bool fl_mask_eq(struct fl_flow_mask *mask1,
719 		       struct fl_flow_mask *mask2)
720 {
721 	const long *lmask1 = fl_key_get_start(&mask1->key, mask1);
722 	const long *lmask2 = fl_key_get_start(&mask2->key, mask2);
723 
724 	return !memcmp(&mask1->range, &mask2->range, sizeof(mask1->range)) &&
725 	       !memcmp(lmask1, lmask2, fl_mask_range(mask1));
726 }
727 
728 static const struct rhashtable_params fl_ht_params = {
729 	.key_offset = offsetof(struct cls_fl_filter, mkey), /* base offset */
730 	.head_offset = offsetof(struct cls_fl_filter, ht_node),
731 	.automatic_shrinking = true,
732 };
733 
734 static int fl_init_hashtable(struct cls_fl_head *head,
735 			     struct fl_flow_mask *mask)
736 {
737 	head->ht_params = fl_ht_params;
738 	head->ht_params.key_len = fl_mask_range(mask);
739 	head->ht_params.key_offset += mask->range.start;
740 
741 	return rhashtable_init(&head->ht, &head->ht_params);
742 }
743 
744 #define FL_KEY_MEMBER_OFFSET(member) offsetof(struct fl_flow_key, member)
745 #define FL_KEY_MEMBER_SIZE(member) (sizeof(((struct fl_flow_key *) 0)->member))
746 
747 #define FL_KEY_IS_MASKED(mask, member)						\
748 	memchr_inv(((char *)mask) + FL_KEY_MEMBER_OFFSET(member),		\
749 		   0, FL_KEY_MEMBER_SIZE(member))				\
750 
751 #define FL_KEY_SET(keys, cnt, id, member)					\
752 	do {									\
753 		keys[cnt].key_id = id;						\
754 		keys[cnt].offset = FL_KEY_MEMBER_OFFSET(member);		\
755 		cnt++;								\
756 	} while(0);
757 
758 #define FL_KEY_SET_IF_MASKED(mask, keys, cnt, id, member)			\
759 	do {									\
760 		if (FL_KEY_IS_MASKED(mask, member))				\
761 			FL_KEY_SET(keys, cnt, id, member);			\
762 	} while(0);
763 
764 static void fl_init_dissector(struct cls_fl_head *head,
765 			      struct fl_flow_mask *mask)
766 {
767 	struct flow_dissector_key keys[FLOW_DISSECTOR_KEY_MAX];
768 	size_t cnt = 0;
769 
770 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_CONTROL, control);
771 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_BASIC, basic);
772 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
773 			     FLOW_DISSECTOR_KEY_ETH_ADDRS, eth);
774 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
775 			     FLOW_DISSECTOR_KEY_IPV4_ADDRS, ipv4);
776 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
777 			     FLOW_DISSECTOR_KEY_IPV6_ADDRS, ipv6);
778 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
779 			     FLOW_DISSECTOR_KEY_PORTS, tp);
780 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
781 			     FLOW_DISSECTOR_KEY_IP, ip);
782 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
783 			     FLOW_DISSECTOR_KEY_TCP, tcp);
784 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
785 			     FLOW_DISSECTOR_KEY_ICMP, icmp);
786 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
787 			     FLOW_DISSECTOR_KEY_ARP, arp);
788 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
789 			     FLOW_DISSECTOR_KEY_MPLS, mpls);
790 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
791 			     FLOW_DISSECTOR_KEY_VLAN, vlan);
792 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
793 			     FLOW_DISSECTOR_KEY_ENC_KEYID, enc_key_id);
794 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
795 			     FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, enc_ipv4);
796 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
797 			     FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, enc_ipv6);
798 	if (FL_KEY_IS_MASKED(&mask->key, enc_ipv4) ||
799 	    FL_KEY_IS_MASKED(&mask->key, enc_ipv6))
800 		FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_ENC_CONTROL,
801 			   enc_control);
802 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
803 			     FLOW_DISSECTOR_KEY_ENC_PORTS, enc_tp);
804 
805 	skb_flow_dissector_init(&head->dissector, keys, cnt);
806 }
807 
808 static int fl_check_assign_mask(struct cls_fl_head *head,
809 				struct fl_flow_mask *mask)
810 {
811 	int err;
812 
813 	if (head->mask_assigned) {
814 		if (!fl_mask_eq(&head->mask, mask))
815 			return -EINVAL;
816 		else
817 			return 0;
818 	}
819 
820 	/* Mask is not assigned yet. So assign it and init hashtable
821 	 * according to that.
822 	 */
823 	err = fl_init_hashtable(head, mask);
824 	if (err)
825 		return err;
826 	memcpy(&head->mask, mask, sizeof(head->mask));
827 	head->mask_assigned = true;
828 
829 	fl_init_dissector(head, mask);
830 
831 	return 0;
832 }
833 
834 static int fl_set_parms(struct net *net, struct tcf_proto *tp,
835 			struct cls_fl_filter *f, struct fl_flow_mask *mask,
836 			unsigned long base, struct nlattr **tb,
837 			struct nlattr *est, bool ovr,
838 			struct netlink_ext_ack *extack)
839 {
840 	int err;
841 
842 	err = tcf_exts_validate(net, tp, tb, est, &f->exts, ovr, extack);
843 	if (err < 0)
844 		return err;
845 
846 	if (tb[TCA_FLOWER_CLASSID]) {
847 		f->res.classid = nla_get_u32(tb[TCA_FLOWER_CLASSID]);
848 		tcf_bind_filter(tp, &f->res, base);
849 	}
850 
851 	err = fl_set_key(net, tb, &f->key, &mask->key, extack);
852 	if (err)
853 		return err;
854 
855 	fl_mask_update_range(mask);
856 	fl_set_masked_key(&f->mkey, &f->key, mask);
857 
858 	return 0;
859 }
860 
861 static int fl_change(struct net *net, struct sk_buff *in_skb,
862 		     struct tcf_proto *tp, unsigned long base,
863 		     u32 handle, struct nlattr **tca,
864 		     void **arg, bool ovr, struct netlink_ext_ack *extack)
865 {
866 	struct cls_fl_head *head = rtnl_dereference(tp->root);
867 	struct cls_fl_filter *fold = *arg;
868 	struct cls_fl_filter *fnew;
869 	struct nlattr **tb;
870 	struct fl_flow_mask mask = {};
871 	int err;
872 
873 	if (!tca[TCA_OPTIONS])
874 		return -EINVAL;
875 
876 	tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
877 	if (!tb)
878 		return -ENOBUFS;
879 
880 	err = nla_parse_nested(tb, TCA_FLOWER_MAX, tca[TCA_OPTIONS],
881 			       fl_policy, NULL);
882 	if (err < 0)
883 		goto errout_tb;
884 
885 	if (fold && handle && fold->handle != handle) {
886 		err = -EINVAL;
887 		goto errout_tb;
888 	}
889 
890 	fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
891 	if (!fnew) {
892 		err = -ENOBUFS;
893 		goto errout_tb;
894 	}
895 
896 	err = tcf_exts_init(&fnew->exts, TCA_FLOWER_ACT, 0);
897 	if (err < 0)
898 		goto errout;
899 
900 	if (!handle) {
901 		handle = 1;
902 		err = idr_alloc_u32(&head->handle_idr, fnew, &handle,
903 				    INT_MAX, GFP_KERNEL);
904 	} else if (!fold) {
905 		/* user specifies a handle and it doesn't exist */
906 		err = idr_alloc_u32(&head->handle_idr, fnew, &handle,
907 				    handle, GFP_KERNEL);
908 	}
909 	if (err)
910 		goto errout;
911 	fnew->handle = handle;
912 
913 	if (tb[TCA_FLOWER_FLAGS]) {
914 		fnew->flags = nla_get_u32(tb[TCA_FLOWER_FLAGS]);
915 
916 		if (!tc_flags_valid(fnew->flags)) {
917 			err = -EINVAL;
918 			goto errout_idr;
919 		}
920 	}
921 
922 	err = fl_set_parms(net, tp, fnew, &mask, base, tb, tca[TCA_RATE], ovr,
923 			   extack);
924 	if (err)
925 		goto errout_idr;
926 
927 	err = fl_check_assign_mask(head, &mask);
928 	if (err)
929 		goto errout_idr;
930 
931 	if (!tc_skip_sw(fnew->flags)) {
932 		if (!fold && fl_lookup(head, &fnew->mkey)) {
933 			err = -EEXIST;
934 			goto errout_idr;
935 		}
936 
937 		err = rhashtable_insert_fast(&head->ht, &fnew->ht_node,
938 					     head->ht_params);
939 		if (err)
940 			goto errout_idr;
941 	}
942 
943 	if (!tc_skip_hw(fnew->flags)) {
944 		err = fl_hw_replace_filter(tp,
945 					   &head->dissector,
946 					   &mask.key,
947 					   fnew,
948 					   extack);
949 		if (err)
950 			goto errout_idr;
951 	}
952 
953 	if (!tc_in_hw(fnew->flags))
954 		fnew->flags |= TCA_CLS_FLAGS_NOT_IN_HW;
955 
956 	if (fold) {
957 		if (!tc_skip_sw(fold->flags))
958 			rhashtable_remove_fast(&head->ht, &fold->ht_node,
959 					       head->ht_params);
960 		if (!tc_skip_hw(fold->flags))
961 			fl_hw_destroy_filter(tp, fold, NULL);
962 	}
963 
964 	*arg = fnew;
965 
966 	if (fold) {
967 		idr_replace(&head->handle_idr, fnew, fnew->handle);
968 		list_replace_rcu(&fold->list, &fnew->list);
969 		tcf_unbind_filter(tp, &fold->res);
970 		tcf_exts_get_net(&fold->exts);
971 		call_rcu(&fold->rcu, fl_destroy_filter);
972 	} else {
973 		list_add_tail_rcu(&fnew->list, &head->filters);
974 	}
975 
976 	kfree(tb);
977 	return 0;
978 
979 errout_idr:
980 	if (fnew->handle)
981 		idr_remove(&head->handle_idr, fnew->handle);
982 errout:
983 	tcf_exts_destroy(&fnew->exts);
984 	kfree(fnew);
985 errout_tb:
986 	kfree(tb);
987 	return err;
988 }
989 
990 static int fl_delete(struct tcf_proto *tp, void *arg, bool *last,
991 		     struct netlink_ext_ack *extack)
992 {
993 	struct cls_fl_head *head = rtnl_dereference(tp->root);
994 	struct cls_fl_filter *f = arg;
995 
996 	if (!tc_skip_sw(f->flags))
997 		rhashtable_remove_fast(&head->ht, &f->ht_node,
998 				       head->ht_params);
999 	__fl_delete(tp, f, extack);
1000 	*last = list_empty(&head->filters);
1001 	return 0;
1002 }
1003 
1004 static void fl_walk(struct tcf_proto *tp, struct tcf_walker *arg)
1005 {
1006 	struct cls_fl_head *head = rtnl_dereference(tp->root);
1007 	struct cls_fl_filter *f;
1008 
1009 	list_for_each_entry_rcu(f, &head->filters, list) {
1010 		if (arg->count < arg->skip)
1011 			goto skip;
1012 		if (arg->fn(tp, f, arg) < 0) {
1013 			arg->stop = 1;
1014 			break;
1015 		}
1016 skip:
1017 		arg->count++;
1018 	}
1019 }
1020 
1021 static int fl_dump_key_val(struct sk_buff *skb,
1022 			   void *val, int val_type,
1023 			   void *mask, int mask_type, int len)
1024 {
1025 	int err;
1026 
1027 	if (!memchr_inv(mask, 0, len))
1028 		return 0;
1029 	err = nla_put(skb, val_type, len, val);
1030 	if (err)
1031 		return err;
1032 	if (mask_type != TCA_FLOWER_UNSPEC) {
1033 		err = nla_put(skb, mask_type, len, mask);
1034 		if (err)
1035 			return err;
1036 	}
1037 	return 0;
1038 }
1039 
1040 static int fl_dump_key_mpls(struct sk_buff *skb,
1041 			    struct flow_dissector_key_mpls *mpls_key,
1042 			    struct flow_dissector_key_mpls *mpls_mask)
1043 {
1044 	int err;
1045 
1046 	if (!memchr_inv(mpls_mask, 0, sizeof(*mpls_mask)))
1047 		return 0;
1048 	if (mpls_mask->mpls_ttl) {
1049 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TTL,
1050 				 mpls_key->mpls_ttl);
1051 		if (err)
1052 			return err;
1053 	}
1054 	if (mpls_mask->mpls_tc) {
1055 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TC,
1056 				 mpls_key->mpls_tc);
1057 		if (err)
1058 			return err;
1059 	}
1060 	if (mpls_mask->mpls_label) {
1061 		err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_LABEL,
1062 				  mpls_key->mpls_label);
1063 		if (err)
1064 			return err;
1065 	}
1066 	if (mpls_mask->mpls_bos) {
1067 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_BOS,
1068 				 mpls_key->mpls_bos);
1069 		if (err)
1070 			return err;
1071 	}
1072 	return 0;
1073 }
1074 
1075 static int fl_dump_key_ip(struct sk_buff *skb,
1076 			  struct flow_dissector_key_ip *key,
1077 			  struct flow_dissector_key_ip *mask)
1078 {
1079 	if (fl_dump_key_val(skb, &key->tos, TCA_FLOWER_KEY_IP_TOS, &mask->tos,
1080 			    TCA_FLOWER_KEY_IP_TOS_MASK, sizeof(key->tos)) ||
1081 	    fl_dump_key_val(skb, &key->ttl, TCA_FLOWER_KEY_IP_TTL, &mask->ttl,
1082 			    TCA_FLOWER_KEY_IP_TTL_MASK, sizeof(key->ttl)))
1083 		return -1;
1084 
1085 	return 0;
1086 }
1087 
1088 static int fl_dump_key_vlan(struct sk_buff *skb,
1089 			    struct flow_dissector_key_vlan *vlan_key,
1090 			    struct flow_dissector_key_vlan *vlan_mask)
1091 {
1092 	int err;
1093 
1094 	if (!memchr_inv(vlan_mask, 0, sizeof(*vlan_mask)))
1095 		return 0;
1096 	if (vlan_mask->vlan_id) {
1097 		err = nla_put_u16(skb, TCA_FLOWER_KEY_VLAN_ID,
1098 				  vlan_key->vlan_id);
1099 		if (err)
1100 			return err;
1101 	}
1102 	if (vlan_mask->vlan_priority) {
1103 		err = nla_put_u8(skb, TCA_FLOWER_KEY_VLAN_PRIO,
1104 				 vlan_key->vlan_priority);
1105 		if (err)
1106 			return err;
1107 	}
1108 	return 0;
1109 }
1110 
1111 static void fl_get_key_flag(u32 dissector_key, u32 dissector_mask,
1112 			    u32 *flower_key, u32 *flower_mask,
1113 			    u32 flower_flag_bit, u32 dissector_flag_bit)
1114 {
1115 	if (dissector_mask & dissector_flag_bit) {
1116 		*flower_mask |= flower_flag_bit;
1117 		if (dissector_key & dissector_flag_bit)
1118 			*flower_key |= flower_flag_bit;
1119 	}
1120 }
1121 
1122 static int fl_dump_key_flags(struct sk_buff *skb, u32 flags_key, u32 flags_mask)
1123 {
1124 	u32 key, mask;
1125 	__be32 _key, _mask;
1126 	int err;
1127 
1128 	if (!memchr_inv(&flags_mask, 0, sizeof(flags_mask)))
1129 		return 0;
1130 
1131 	key = 0;
1132 	mask = 0;
1133 
1134 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
1135 			TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
1136 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
1137 			TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
1138 			FLOW_DIS_FIRST_FRAG);
1139 
1140 	_key = cpu_to_be32(key);
1141 	_mask = cpu_to_be32(mask);
1142 
1143 	err = nla_put(skb, TCA_FLOWER_KEY_FLAGS, 4, &_key);
1144 	if (err)
1145 		return err;
1146 
1147 	return nla_put(skb, TCA_FLOWER_KEY_FLAGS_MASK, 4, &_mask);
1148 }
1149 
1150 static int fl_dump(struct net *net, struct tcf_proto *tp, void *fh,
1151 		   struct sk_buff *skb, struct tcmsg *t)
1152 {
1153 	struct cls_fl_head *head = rtnl_dereference(tp->root);
1154 	struct cls_fl_filter *f = fh;
1155 	struct nlattr *nest;
1156 	struct fl_flow_key *key, *mask;
1157 
1158 	if (!f)
1159 		return skb->len;
1160 
1161 	t->tcm_handle = f->handle;
1162 
1163 	nest = nla_nest_start(skb, TCA_OPTIONS);
1164 	if (!nest)
1165 		goto nla_put_failure;
1166 
1167 	if (f->res.classid &&
1168 	    nla_put_u32(skb, TCA_FLOWER_CLASSID, f->res.classid))
1169 		goto nla_put_failure;
1170 
1171 	key = &f->key;
1172 	mask = &head->mask.key;
1173 
1174 	if (mask->indev_ifindex) {
1175 		struct net_device *dev;
1176 
1177 		dev = __dev_get_by_index(net, key->indev_ifindex);
1178 		if (dev && nla_put_string(skb, TCA_FLOWER_INDEV, dev->name))
1179 			goto nla_put_failure;
1180 	}
1181 
1182 	if (!tc_skip_hw(f->flags))
1183 		fl_hw_update_stats(tp, f);
1184 
1185 	if (fl_dump_key_val(skb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
1186 			    mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
1187 			    sizeof(key->eth.dst)) ||
1188 	    fl_dump_key_val(skb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
1189 			    mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
1190 			    sizeof(key->eth.src)) ||
1191 	    fl_dump_key_val(skb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE,
1192 			    &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
1193 			    sizeof(key->basic.n_proto)))
1194 		goto nla_put_failure;
1195 
1196 	if (fl_dump_key_mpls(skb, &key->mpls, &mask->mpls))
1197 		goto nla_put_failure;
1198 
1199 	if (fl_dump_key_vlan(skb, &key->vlan, &mask->vlan))
1200 		goto nla_put_failure;
1201 
1202 	if ((key->basic.n_proto == htons(ETH_P_IP) ||
1203 	     key->basic.n_proto == htons(ETH_P_IPV6)) &&
1204 	    (fl_dump_key_val(skb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
1205 			    &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
1206 			    sizeof(key->basic.ip_proto)) ||
1207 	    fl_dump_key_ip(skb, &key->ip, &mask->ip)))
1208 		goto nla_put_failure;
1209 
1210 	if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
1211 	    (fl_dump_key_val(skb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
1212 			     &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
1213 			     sizeof(key->ipv4.src)) ||
1214 	     fl_dump_key_val(skb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
1215 			     &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
1216 			     sizeof(key->ipv4.dst))))
1217 		goto nla_put_failure;
1218 	else if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
1219 		 (fl_dump_key_val(skb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
1220 				  &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
1221 				  sizeof(key->ipv6.src)) ||
1222 		  fl_dump_key_val(skb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
1223 				  &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
1224 				  sizeof(key->ipv6.dst))))
1225 		goto nla_put_failure;
1226 
1227 	if (key->basic.ip_proto == IPPROTO_TCP &&
1228 	    (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
1229 			     &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
1230 			     sizeof(key->tp.src)) ||
1231 	     fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
1232 			     &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
1233 			     sizeof(key->tp.dst)) ||
1234 	     fl_dump_key_val(skb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
1235 			     &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
1236 			     sizeof(key->tcp.flags))))
1237 		goto nla_put_failure;
1238 	else if (key->basic.ip_proto == IPPROTO_UDP &&
1239 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
1240 				  &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
1241 				  sizeof(key->tp.src)) ||
1242 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
1243 				  &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
1244 				  sizeof(key->tp.dst))))
1245 		goto nla_put_failure;
1246 	else if (key->basic.ip_proto == IPPROTO_SCTP &&
1247 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
1248 				  &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
1249 				  sizeof(key->tp.src)) ||
1250 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
1251 				  &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
1252 				  sizeof(key->tp.dst))))
1253 		goto nla_put_failure;
1254 	else if (key->basic.n_proto == htons(ETH_P_IP) &&
1255 		 key->basic.ip_proto == IPPROTO_ICMP &&
1256 		 (fl_dump_key_val(skb, &key->icmp.type,
1257 				  TCA_FLOWER_KEY_ICMPV4_TYPE, &mask->icmp.type,
1258 				  TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
1259 				  sizeof(key->icmp.type)) ||
1260 		  fl_dump_key_val(skb, &key->icmp.code,
1261 				  TCA_FLOWER_KEY_ICMPV4_CODE, &mask->icmp.code,
1262 				  TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
1263 				  sizeof(key->icmp.code))))
1264 		goto nla_put_failure;
1265 	else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
1266 		 key->basic.ip_proto == IPPROTO_ICMPV6 &&
1267 		 (fl_dump_key_val(skb, &key->icmp.type,
1268 				  TCA_FLOWER_KEY_ICMPV6_TYPE, &mask->icmp.type,
1269 				  TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
1270 				  sizeof(key->icmp.type)) ||
1271 		  fl_dump_key_val(skb, &key->icmp.code,
1272 				  TCA_FLOWER_KEY_ICMPV6_CODE, &mask->icmp.code,
1273 				  TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
1274 				  sizeof(key->icmp.code))))
1275 		goto nla_put_failure;
1276 	else if ((key->basic.n_proto == htons(ETH_P_ARP) ||
1277 		  key->basic.n_proto == htons(ETH_P_RARP)) &&
1278 		 (fl_dump_key_val(skb, &key->arp.sip,
1279 				  TCA_FLOWER_KEY_ARP_SIP, &mask->arp.sip,
1280 				  TCA_FLOWER_KEY_ARP_SIP_MASK,
1281 				  sizeof(key->arp.sip)) ||
1282 		  fl_dump_key_val(skb, &key->arp.tip,
1283 				  TCA_FLOWER_KEY_ARP_TIP, &mask->arp.tip,
1284 				  TCA_FLOWER_KEY_ARP_TIP_MASK,
1285 				  sizeof(key->arp.tip)) ||
1286 		  fl_dump_key_val(skb, &key->arp.op,
1287 				  TCA_FLOWER_KEY_ARP_OP, &mask->arp.op,
1288 				  TCA_FLOWER_KEY_ARP_OP_MASK,
1289 				  sizeof(key->arp.op)) ||
1290 		  fl_dump_key_val(skb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
1291 				  mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
1292 				  sizeof(key->arp.sha)) ||
1293 		  fl_dump_key_val(skb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
1294 				  mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
1295 				  sizeof(key->arp.tha))))
1296 		goto nla_put_failure;
1297 
1298 	if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
1299 	    (fl_dump_key_val(skb, &key->enc_ipv4.src,
1300 			    TCA_FLOWER_KEY_ENC_IPV4_SRC, &mask->enc_ipv4.src,
1301 			    TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
1302 			    sizeof(key->enc_ipv4.src)) ||
1303 	     fl_dump_key_val(skb, &key->enc_ipv4.dst,
1304 			     TCA_FLOWER_KEY_ENC_IPV4_DST, &mask->enc_ipv4.dst,
1305 			     TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
1306 			     sizeof(key->enc_ipv4.dst))))
1307 		goto nla_put_failure;
1308 	else if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
1309 		 (fl_dump_key_val(skb, &key->enc_ipv6.src,
1310 			    TCA_FLOWER_KEY_ENC_IPV6_SRC, &mask->enc_ipv6.src,
1311 			    TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
1312 			    sizeof(key->enc_ipv6.src)) ||
1313 		 fl_dump_key_val(skb, &key->enc_ipv6.dst,
1314 				 TCA_FLOWER_KEY_ENC_IPV6_DST,
1315 				 &mask->enc_ipv6.dst,
1316 				 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
1317 			    sizeof(key->enc_ipv6.dst))))
1318 		goto nla_put_failure;
1319 
1320 	if (fl_dump_key_val(skb, &key->enc_key_id, TCA_FLOWER_KEY_ENC_KEY_ID,
1321 			    &mask->enc_key_id, TCA_FLOWER_UNSPEC,
1322 			    sizeof(key->enc_key_id)) ||
1323 	    fl_dump_key_val(skb, &key->enc_tp.src,
1324 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
1325 			    &mask->enc_tp.src,
1326 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
1327 			    sizeof(key->enc_tp.src)) ||
1328 	    fl_dump_key_val(skb, &key->enc_tp.dst,
1329 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
1330 			    &mask->enc_tp.dst,
1331 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
1332 			    sizeof(key->enc_tp.dst)))
1333 		goto nla_put_failure;
1334 
1335 	if (fl_dump_key_flags(skb, key->control.flags, mask->control.flags))
1336 		goto nla_put_failure;
1337 
1338 	if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
1339 		goto nla_put_failure;
1340 
1341 	if (tcf_exts_dump(skb, &f->exts))
1342 		goto nla_put_failure;
1343 
1344 	nla_nest_end(skb, nest);
1345 
1346 	if (tcf_exts_dump_stats(skb, &f->exts) < 0)
1347 		goto nla_put_failure;
1348 
1349 	return skb->len;
1350 
1351 nla_put_failure:
1352 	nla_nest_cancel(skb, nest);
1353 	return -1;
1354 }
1355 
1356 static void fl_bind_class(void *fh, u32 classid, unsigned long cl)
1357 {
1358 	struct cls_fl_filter *f = fh;
1359 
1360 	if (f && f->res.classid == classid)
1361 		f->res.class = cl;
1362 }
1363 
1364 static struct tcf_proto_ops cls_fl_ops __read_mostly = {
1365 	.kind		= "flower",
1366 	.classify	= fl_classify,
1367 	.init		= fl_init,
1368 	.destroy	= fl_destroy,
1369 	.get		= fl_get,
1370 	.change		= fl_change,
1371 	.delete		= fl_delete,
1372 	.walk		= fl_walk,
1373 	.dump		= fl_dump,
1374 	.bind_class	= fl_bind_class,
1375 	.owner		= THIS_MODULE,
1376 };
1377 
1378 static int __init cls_fl_init(void)
1379 {
1380 	return register_tcf_proto_ops(&cls_fl_ops);
1381 }
1382 
1383 static void __exit cls_fl_exit(void)
1384 {
1385 	unregister_tcf_proto_ops(&cls_fl_ops);
1386 }
1387 
1388 module_init(cls_fl_init);
1389 module_exit(cls_fl_exit);
1390 
1391 MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
1392 MODULE_DESCRIPTION("Flower classifier");
1393 MODULE_LICENSE("GPL v2");
1394