xref: /openbmc/linux/net/sched/cls_flower.c (revision cfbb9be8)
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 
515 	return 0;
516 }
517 
518 static void fl_set_key_ip(struct nlattr **tb,
519 			  struct flow_dissector_key_ip *key,
520 			  struct flow_dissector_key_ip *mask)
521 {
522 		fl_set_key_val(tb, &key->tos, TCA_FLOWER_KEY_IP_TOS,
523 			       &mask->tos, TCA_FLOWER_KEY_IP_TOS_MASK,
524 			       sizeof(key->tos));
525 
526 		fl_set_key_val(tb, &key->ttl, TCA_FLOWER_KEY_IP_TTL,
527 			       &mask->ttl, TCA_FLOWER_KEY_IP_TTL_MASK,
528 			       sizeof(key->ttl));
529 }
530 
531 static int fl_set_key(struct net *net, struct nlattr **tb,
532 		      struct fl_flow_key *key, struct fl_flow_key *mask,
533 		      struct netlink_ext_ack *extack)
534 {
535 	__be16 ethertype;
536 	int ret = 0;
537 #ifdef CONFIG_NET_CLS_IND
538 	if (tb[TCA_FLOWER_INDEV]) {
539 		int err = tcf_change_indev(net, tb[TCA_FLOWER_INDEV], extack);
540 		if (err < 0)
541 			return err;
542 		key->indev_ifindex = err;
543 		mask->indev_ifindex = 0xffffffff;
544 	}
545 #endif
546 
547 	fl_set_key_val(tb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
548 		       mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
549 		       sizeof(key->eth.dst));
550 	fl_set_key_val(tb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
551 		       mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
552 		       sizeof(key->eth.src));
553 
554 	if (tb[TCA_FLOWER_KEY_ETH_TYPE]) {
555 		ethertype = nla_get_be16(tb[TCA_FLOWER_KEY_ETH_TYPE]);
556 
557 		if (ethertype == htons(ETH_P_8021Q)) {
558 			fl_set_key_vlan(tb, &key->vlan, &mask->vlan);
559 			fl_set_key_val(tb, &key->basic.n_proto,
560 				       TCA_FLOWER_KEY_VLAN_ETH_TYPE,
561 				       &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
562 				       sizeof(key->basic.n_proto));
563 		} else {
564 			key->basic.n_proto = ethertype;
565 			mask->basic.n_proto = cpu_to_be16(~0);
566 		}
567 	}
568 
569 	if (key->basic.n_proto == htons(ETH_P_IP) ||
570 	    key->basic.n_proto == htons(ETH_P_IPV6)) {
571 		fl_set_key_val(tb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
572 			       &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
573 			       sizeof(key->basic.ip_proto));
574 		fl_set_key_ip(tb, &key->ip, &mask->ip);
575 	}
576 
577 	if (tb[TCA_FLOWER_KEY_IPV4_SRC] || tb[TCA_FLOWER_KEY_IPV4_DST]) {
578 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
579 		mask->control.addr_type = ~0;
580 		fl_set_key_val(tb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
581 			       &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
582 			       sizeof(key->ipv4.src));
583 		fl_set_key_val(tb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
584 			       &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
585 			       sizeof(key->ipv4.dst));
586 	} else if (tb[TCA_FLOWER_KEY_IPV6_SRC] || tb[TCA_FLOWER_KEY_IPV6_DST]) {
587 		key->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
588 		mask->control.addr_type = ~0;
589 		fl_set_key_val(tb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
590 			       &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
591 			       sizeof(key->ipv6.src));
592 		fl_set_key_val(tb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
593 			       &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
594 			       sizeof(key->ipv6.dst));
595 	}
596 
597 	if (key->basic.ip_proto == IPPROTO_TCP) {
598 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
599 			       &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
600 			       sizeof(key->tp.src));
601 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
602 			       &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
603 			       sizeof(key->tp.dst));
604 		fl_set_key_val(tb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
605 			       &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
606 			       sizeof(key->tcp.flags));
607 	} else if (key->basic.ip_proto == IPPROTO_UDP) {
608 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
609 			       &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
610 			       sizeof(key->tp.src));
611 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
612 			       &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
613 			       sizeof(key->tp.dst));
614 	} else if (key->basic.ip_proto == IPPROTO_SCTP) {
615 		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
616 			       &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
617 			       sizeof(key->tp.src));
618 		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
619 			       &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
620 			       sizeof(key->tp.dst));
621 	} else if (key->basic.n_proto == htons(ETH_P_IP) &&
622 		   key->basic.ip_proto == IPPROTO_ICMP) {
623 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV4_TYPE,
624 			       &mask->icmp.type,
625 			       TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
626 			       sizeof(key->icmp.type));
627 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV4_CODE,
628 			       &mask->icmp.code,
629 			       TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
630 			       sizeof(key->icmp.code));
631 	} else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
632 		   key->basic.ip_proto == IPPROTO_ICMPV6) {
633 		fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV6_TYPE,
634 			       &mask->icmp.type,
635 			       TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
636 			       sizeof(key->icmp.type));
637 		fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV6_CODE,
638 			       &mask->icmp.code,
639 			       TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
640 			       sizeof(key->icmp.code));
641 	} else if (key->basic.n_proto == htons(ETH_P_MPLS_UC) ||
642 		   key->basic.n_proto == htons(ETH_P_MPLS_MC)) {
643 		ret = fl_set_key_mpls(tb, &key->mpls, &mask->mpls);
644 		if (ret)
645 			return ret;
646 	} else if (key->basic.n_proto == htons(ETH_P_ARP) ||
647 		   key->basic.n_proto == htons(ETH_P_RARP)) {
648 		fl_set_key_val(tb, &key->arp.sip, TCA_FLOWER_KEY_ARP_SIP,
649 			       &mask->arp.sip, TCA_FLOWER_KEY_ARP_SIP_MASK,
650 			       sizeof(key->arp.sip));
651 		fl_set_key_val(tb, &key->arp.tip, TCA_FLOWER_KEY_ARP_TIP,
652 			       &mask->arp.tip, TCA_FLOWER_KEY_ARP_TIP_MASK,
653 			       sizeof(key->arp.tip));
654 		fl_set_key_val(tb, &key->arp.op, TCA_FLOWER_KEY_ARP_OP,
655 			       &mask->arp.op, TCA_FLOWER_KEY_ARP_OP_MASK,
656 			       sizeof(key->arp.op));
657 		fl_set_key_val(tb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
658 			       mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
659 			       sizeof(key->arp.sha));
660 		fl_set_key_val(tb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
661 			       mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
662 			       sizeof(key->arp.tha));
663 	}
664 
665 	if (tb[TCA_FLOWER_KEY_ENC_IPV4_SRC] ||
666 	    tb[TCA_FLOWER_KEY_ENC_IPV4_DST]) {
667 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
668 		mask->enc_control.addr_type = ~0;
669 		fl_set_key_val(tb, &key->enc_ipv4.src,
670 			       TCA_FLOWER_KEY_ENC_IPV4_SRC,
671 			       &mask->enc_ipv4.src,
672 			       TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
673 			       sizeof(key->enc_ipv4.src));
674 		fl_set_key_val(tb, &key->enc_ipv4.dst,
675 			       TCA_FLOWER_KEY_ENC_IPV4_DST,
676 			       &mask->enc_ipv4.dst,
677 			       TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
678 			       sizeof(key->enc_ipv4.dst));
679 	}
680 
681 	if (tb[TCA_FLOWER_KEY_ENC_IPV6_SRC] ||
682 	    tb[TCA_FLOWER_KEY_ENC_IPV6_DST]) {
683 		key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
684 		mask->enc_control.addr_type = ~0;
685 		fl_set_key_val(tb, &key->enc_ipv6.src,
686 			       TCA_FLOWER_KEY_ENC_IPV6_SRC,
687 			       &mask->enc_ipv6.src,
688 			       TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
689 			       sizeof(key->enc_ipv6.src));
690 		fl_set_key_val(tb, &key->enc_ipv6.dst,
691 			       TCA_FLOWER_KEY_ENC_IPV6_DST,
692 			       &mask->enc_ipv6.dst,
693 			       TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
694 			       sizeof(key->enc_ipv6.dst));
695 	}
696 
697 	fl_set_key_val(tb, &key->enc_key_id.keyid, TCA_FLOWER_KEY_ENC_KEY_ID,
698 		       &mask->enc_key_id.keyid, TCA_FLOWER_UNSPEC,
699 		       sizeof(key->enc_key_id.keyid));
700 
701 	fl_set_key_val(tb, &key->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
702 		       &mask->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
703 		       sizeof(key->enc_tp.src));
704 
705 	fl_set_key_val(tb, &key->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
706 		       &mask->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
707 		       sizeof(key->enc_tp.dst));
708 
709 	if (tb[TCA_FLOWER_KEY_FLAGS])
710 		ret = fl_set_key_flags(tb, &key->control.flags, &mask->control.flags);
711 
712 	return ret;
713 }
714 
715 static bool fl_mask_eq(struct fl_flow_mask *mask1,
716 		       struct fl_flow_mask *mask2)
717 {
718 	const long *lmask1 = fl_key_get_start(&mask1->key, mask1);
719 	const long *lmask2 = fl_key_get_start(&mask2->key, mask2);
720 
721 	return !memcmp(&mask1->range, &mask2->range, sizeof(mask1->range)) &&
722 	       !memcmp(lmask1, lmask2, fl_mask_range(mask1));
723 }
724 
725 static const struct rhashtable_params fl_ht_params = {
726 	.key_offset = offsetof(struct cls_fl_filter, mkey), /* base offset */
727 	.head_offset = offsetof(struct cls_fl_filter, ht_node),
728 	.automatic_shrinking = true,
729 };
730 
731 static int fl_init_hashtable(struct cls_fl_head *head,
732 			     struct fl_flow_mask *mask)
733 {
734 	head->ht_params = fl_ht_params;
735 	head->ht_params.key_len = fl_mask_range(mask);
736 	head->ht_params.key_offset += mask->range.start;
737 
738 	return rhashtable_init(&head->ht, &head->ht_params);
739 }
740 
741 #define FL_KEY_MEMBER_OFFSET(member) offsetof(struct fl_flow_key, member)
742 #define FL_KEY_MEMBER_SIZE(member) (sizeof(((struct fl_flow_key *) 0)->member))
743 
744 #define FL_KEY_IS_MASKED(mask, member)						\
745 	memchr_inv(((char *)mask) + FL_KEY_MEMBER_OFFSET(member),		\
746 		   0, FL_KEY_MEMBER_SIZE(member))				\
747 
748 #define FL_KEY_SET(keys, cnt, id, member)					\
749 	do {									\
750 		keys[cnt].key_id = id;						\
751 		keys[cnt].offset = FL_KEY_MEMBER_OFFSET(member);		\
752 		cnt++;								\
753 	} while(0);
754 
755 #define FL_KEY_SET_IF_MASKED(mask, keys, cnt, id, member)			\
756 	do {									\
757 		if (FL_KEY_IS_MASKED(mask, member))				\
758 			FL_KEY_SET(keys, cnt, id, member);			\
759 	} while(0);
760 
761 static void fl_init_dissector(struct cls_fl_head *head,
762 			      struct fl_flow_mask *mask)
763 {
764 	struct flow_dissector_key keys[FLOW_DISSECTOR_KEY_MAX];
765 	size_t cnt = 0;
766 
767 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_CONTROL, control);
768 	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_BASIC, basic);
769 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
770 			     FLOW_DISSECTOR_KEY_ETH_ADDRS, eth);
771 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
772 			     FLOW_DISSECTOR_KEY_IPV4_ADDRS, ipv4);
773 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
774 			     FLOW_DISSECTOR_KEY_IPV6_ADDRS, ipv6);
775 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
776 			     FLOW_DISSECTOR_KEY_PORTS, tp);
777 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
778 			     FLOW_DISSECTOR_KEY_IP, ip);
779 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
780 			     FLOW_DISSECTOR_KEY_TCP, tcp);
781 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
782 			     FLOW_DISSECTOR_KEY_ICMP, icmp);
783 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
784 			     FLOW_DISSECTOR_KEY_ARP, arp);
785 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
786 			     FLOW_DISSECTOR_KEY_MPLS, mpls);
787 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
788 			     FLOW_DISSECTOR_KEY_VLAN, vlan);
789 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
790 			     FLOW_DISSECTOR_KEY_ENC_KEYID, enc_key_id);
791 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
792 			     FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, enc_ipv4);
793 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
794 			     FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, enc_ipv6);
795 	if (FL_KEY_IS_MASKED(&mask->key, enc_ipv4) ||
796 	    FL_KEY_IS_MASKED(&mask->key, enc_ipv6))
797 		FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_ENC_CONTROL,
798 			   enc_control);
799 	FL_KEY_SET_IF_MASKED(&mask->key, keys, cnt,
800 			     FLOW_DISSECTOR_KEY_ENC_PORTS, enc_tp);
801 
802 	skb_flow_dissector_init(&head->dissector, keys, cnt);
803 }
804 
805 static int fl_check_assign_mask(struct cls_fl_head *head,
806 				struct fl_flow_mask *mask)
807 {
808 	int err;
809 
810 	if (head->mask_assigned) {
811 		if (!fl_mask_eq(&head->mask, mask))
812 			return -EINVAL;
813 		else
814 			return 0;
815 	}
816 
817 	/* Mask is not assigned yet. So assign it and init hashtable
818 	 * according to that.
819 	 */
820 	err = fl_init_hashtable(head, mask);
821 	if (err)
822 		return err;
823 	memcpy(&head->mask, mask, sizeof(head->mask));
824 	head->mask_assigned = true;
825 
826 	fl_init_dissector(head, mask);
827 
828 	return 0;
829 }
830 
831 static int fl_set_parms(struct net *net, struct tcf_proto *tp,
832 			struct cls_fl_filter *f, struct fl_flow_mask *mask,
833 			unsigned long base, struct nlattr **tb,
834 			struct nlattr *est, bool ovr,
835 			struct netlink_ext_ack *extack)
836 {
837 	int err;
838 
839 	err = tcf_exts_validate(net, tp, tb, est, &f->exts, ovr, extack);
840 	if (err < 0)
841 		return err;
842 
843 	if (tb[TCA_FLOWER_CLASSID]) {
844 		f->res.classid = nla_get_u32(tb[TCA_FLOWER_CLASSID]);
845 		tcf_bind_filter(tp, &f->res, base);
846 	}
847 
848 	err = fl_set_key(net, tb, &f->key, &mask->key, extack);
849 	if (err)
850 		return err;
851 
852 	fl_mask_update_range(mask);
853 	fl_set_masked_key(&f->mkey, &f->key, mask);
854 
855 	return 0;
856 }
857 
858 static int fl_change(struct net *net, struct sk_buff *in_skb,
859 		     struct tcf_proto *tp, unsigned long base,
860 		     u32 handle, struct nlattr **tca,
861 		     void **arg, bool ovr, struct netlink_ext_ack *extack)
862 {
863 	struct cls_fl_head *head = rtnl_dereference(tp->root);
864 	struct cls_fl_filter *fold = *arg;
865 	struct cls_fl_filter *fnew;
866 	struct nlattr **tb;
867 	struct fl_flow_mask mask = {};
868 	int err;
869 
870 	if (!tca[TCA_OPTIONS])
871 		return -EINVAL;
872 
873 	tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
874 	if (!tb)
875 		return -ENOBUFS;
876 
877 	err = nla_parse_nested(tb, TCA_FLOWER_MAX, tca[TCA_OPTIONS],
878 			       fl_policy, NULL);
879 	if (err < 0)
880 		goto errout_tb;
881 
882 	if (fold && handle && fold->handle != handle) {
883 		err = -EINVAL;
884 		goto errout_tb;
885 	}
886 
887 	fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
888 	if (!fnew) {
889 		err = -ENOBUFS;
890 		goto errout_tb;
891 	}
892 
893 	err = tcf_exts_init(&fnew->exts, TCA_FLOWER_ACT, 0);
894 	if (err < 0)
895 		goto errout;
896 
897 	if (!handle) {
898 		handle = 1;
899 		err = idr_alloc_u32(&head->handle_idr, fnew, &handle,
900 				    INT_MAX, GFP_KERNEL);
901 	} else if (!fold) {
902 		/* user specifies a handle and it doesn't exist */
903 		err = idr_alloc_u32(&head->handle_idr, fnew, &handle,
904 				    handle, GFP_KERNEL);
905 	}
906 	if (err)
907 		goto errout;
908 	fnew->handle = handle;
909 
910 	if (tb[TCA_FLOWER_FLAGS]) {
911 		fnew->flags = nla_get_u32(tb[TCA_FLOWER_FLAGS]);
912 
913 		if (!tc_flags_valid(fnew->flags)) {
914 			err = -EINVAL;
915 			goto errout_idr;
916 		}
917 	}
918 
919 	err = fl_set_parms(net, tp, fnew, &mask, base, tb, tca[TCA_RATE], ovr,
920 			   extack);
921 	if (err)
922 		goto errout_idr;
923 
924 	err = fl_check_assign_mask(head, &mask);
925 	if (err)
926 		goto errout_idr;
927 
928 	if (!tc_skip_sw(fnew->flags)) {
929 		if (!fold && fl_lookup(head, &fnew->mkey)) {
930 			err = -EEXIST;
931 			goto errout_idr;
932 		}
933 
934 		err = rhashtable_insert_fast(&head->ht, &fnew->ht_node,
935 					     head->ht_params);
936 		if (err)
937 			goto errout_idr;
938 	}
939 
940 	if (!tc_skip_hw(fnew->flags)) {
941 		err = fl_hw_replace_filter(tp,
942 					   &head->dissector,
943 					   &mask.key,
944 					   fnew,
945 					   extack);
946 		if (err)
947 			goto errout_idr;
948 	}
949 
950 	if (!tc_in_hw(fnew->flags))
951 		fnew->flags |= TCA_CLS_FLAGS_NOT_IN_HW;
952 
953 	if (fold) {
954 		if (!tc_skip_sw(fold->flags))
955 			rhashtable_remove_fast(&head->ht, &fold->ht_node,
956 					       head->ht_params);
957 		if (!tc_skip_hw(fold->flags))
958 			fl_hw_destroy_filter(tp, fold, NULL);
959 	}
960 
961 	*arg = fnew;
962 
963 	if (fold) {
964 		idr_replace(&head->handle_idr, fnew, fnew->handle);
965 		list_replace_rcu(&fold->list, &fnew->list);
966 		tcf_unbind_filter(tp, &fold->res);
967 		tcf_exts_get_net(&fold->exts);
968 		call_rcu(&fold->rcu, fl_destroy_filter);
969 	} else {
970 		list_add_tail_rcu(&fnew->list, &head->filters);
971 	}
972 
973 	kfree(tb);
974 	return 0;
975 
976 errout_idr:
977 	if (fnew->handle)
978 		idr_remove(&head->handle_idr, fnew->handle);
979 errout:
980 	tcf_exts_destroy(&fnew->exts);
981 	kfree(fnew);
982 errout_tb:
983 	kfree(tb);
984 	return err;
985 }
986 
987 static int fl_delete(struct tcf_proto *tp, void *arg, bool *last,
988 		     struct netlink_ext_ack *extack)
989 {
990 	struct cls_fl_head *head = rtnl_dereference(tp->root);
991 	struct cls_fl_filter *f = arg;
992 
993 	if (!tc_skip_sw(f->flags))
994 		rhashtable_remove_fast(&head->ht, &f->ht_node,
995 				       head->ht_params);
996 	__fl_delete(tp, f, extack);
997 	*last = list_empty(&head->filters);
998 	return 0;
999 }
1000 
1001 static void fl_walk(struct tcf_proto *tp, struct tcf_walker *arg)
1002 {
1003 	struct cls_fl_head *head = rtnl_dereference(tp->root);
1004 	struct cls_fl_filter *f;
1005 
1006 	list_for_each_entry_rcu(f, &head->filters, list) {
1007 		if (arg->count < arg->skip)
1008 			goto skip;
1009 		if (arg->fn(tp, f, arg) < 0) {
1010 			arg->stop = 1;
1011 			break;
1012 		}
1013 skip:
1014 		arg->count++;
1015 	}
1016 }
1017 
1018 static int fl_dump_key_val(struct sk_buff *skb,
1019 			   void *val, int val_type,
1020 			   void *mask, int mask_type, int len)
1021 {
1022 	int err;
1023 
1024 	if (!memchr_inv(mask, 0, len))
1025 		return 0;
1026 	err = nla_put(skb, val_type, len, val);
1027 	if (err)
1028 		return err;
1029 	if (mask_type != TCA_FLOWER_UNSPEC) {
1030 		err = nla_put(skb, mask_type, len, mask);
1031 		if (err)
1032 			return err;
1033 	}
1034 	return 0;
1035 }
1036 
1037 static int fl_dump_key_mpls(struct sk_buff *skb,
1038 			    struct flow_dissector_key_mpls *mpls_key,
1039 			    struct flow_dissector_key_mpls *mpls_mask)
1040 {
1041 	int err;
1042 
1043 	if (!memchr_inv(mpls_mask, 0, sizeof(*mpls_mask)))
1044 		return 0;
1045 	if (mpls_mask->mpls_ttl) {
1046 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TTL,
1047 				 mpls_key->mpls_ttl);
1048 		if (err)
1049 			return err;
1050 	}
1051 	if (mpls_mask->mpls_tc) {
1052 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TC,
1053 				 mpls_key->mpls_tc);
1054 		if (err)
1055 			return err;
1056 	}
1057 	if (mpls_mask->mpls_label) {
1058 		err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_LABEL,
1059 				  mpls_key->mpls_label);
1060 		if (err)
1061 			return err;
1062 	}
1063 	if (mpls_mask->mpls_bos) {
1064 		err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_BOS,
1065 				 mpls_key->mpls_bos);
1066 		if (err)
1067 			return err;
1068 	}
1069 	return 0;
1070 }
1071 
1072 static int fl_dump_key_ip(struct sk_buff *skb,
1073 			  struct flow_dissector_key_ip *key,
1074 			  struct flow_dissector_key_ip *mask)
1075 {
1076 	if (fl_dump_key_val(skb, &key->tos, TCA_FLOWER_KEY_IP_TOS, &mask->tos,
1077 			    TCA_FLOWER_KEY_IP_TOS_MASK, sizeof(key->tos)) ||
1078 	    fl_dump_key_val(skb, &key->ttl, TCA_FLOWER_KEY_IP_TTL, &mask->ttl,
1079 			    TCA_FLOWER_KEY_IP_TTL_MASK, sizeof(key->ttl)))
1080 		return -1;
1081 
1082 	return 0;
1083 }
1084 
1085 static int fl_dump_key_vlan(struct sk_buff *skb,
1086 			    struct flow_dissector_key_vlan *vlan_key,
1087 			    struct flow_dissector_key_vlan *vlan_mask)
1088 {
1089 	int err;
1090 
1091 	if (!memchr_inv(vlan_mask, 0, sizeof(*vlan_mask)))
1092 		return 0;
1093 	if (vlan_mask->vlan_id) {
1094 		err = nla_put_u16(skb, TCA_FLOWER_KEY_VLAN_ID,
1095 				  vlan_key->vlan_id);
1096 		if (err)
1097 			return err;
1098 	}
1099 	if (vlan_mask->vlan_priority) {
1100 		err = nla_put_u8(skb, TCA_FLOWER_KEY_VLAN_PRIO,
1101 				 vlan_key->vlan_priority);
1102 		if (err)
1103 			return err;
1104 	}
1105 	return 0;
1106 }
1107 
1108 static void fl_get_key_flag(u32 dissector_key, u32 dissector_mask,
1109 			    u32 *flower_key, u32 *flower_mask,
1110 			    u32 flower_flag_bit, u32 dissector_flag_bit)
1111 {
1112 	if (dissector_mask & dissector_flag_bit) {
1113 		*flower_mask |= flower_flag_bit;
1114 		if (dissector_key & dissector_flag_bit)
1115 			*flower_key |= flower_flag_bit;
1116 	}
1117 }
1118 
1119 static int fl_dump_key_flags(struct sk_buff *skb, u32 flags_key, u32 flags_mask)
1120 {
1121 	u32 key, mask;
1122 	__be32 _key, _mask;
1123 	int err;
1124 
1125 	if (!memchr_inv(&flags_mask, 0, sizeof(flags_mask)))
1126 		return 0;
1127 
1128 	key = 0;
1129 	mask = 0;
1130 
1131 	fl_get_key_flag(flags_key, flags_mask, &key, &mask,
1132 			TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
1133 
1134 	_key = cpu_to_be32(key);
1135 	_mask = cpu_to_be32(mask);
1136 
1137 	err = nla_put(skb, TCA_FLOWER_KEY_FLAGS, 4, &_key);
1138 	if (err)
1139 		return err;
1140 
1141 	return nla_put(skb, TCA_FLOWER_KEY_FLAGS_MASK, 4, &_mask);
1142 }
1143 
1144 static int fl_dump(struct net *net, struct tcf_proto *tp, void *fh,
1145 		   struct sk_buff *skb, struct tcmsg *t)
1146 {
1147 	struct cls_fl_head *head = rtnl_dereference(tp->root);
1148 	struct cls_fl_filter *f = fh;
1149 	struct nlattr *nest;
1150 	struct fl_flow_key *key, *mask;
1151 
1152 	if (!f)
1153 		return skb->len;
1154 
1155 	t->tcm_handle = f->handle;
1156 
1157 	nest = nla_nest_start(skb, TCA_OPTIONS);
1158 	if (!nest)
1159 		goto nla_put_failure;
1160 
1161 	if (f->res.classid &&
1162 	    nla_put_u32(skb, TCA_FLOWER_CLASSID, f->res.classid))
1163 		goto nla_put_failure;
1164 
1165 	key = &f->key;
1166 	mask = &head->mask.key;
1167 
1168 	if (mask->indev_ifindex) {
1169 		struct net_device *dev;
1170 
1171 		dev = __dev_get_by_index(net, key->indev_ifindex);
1172 		if (dev && nla_put_string(skb, TCA_FLOWER_INDEV, dev->name))
1173 			goto nla_put_failure;
1174 	}
1175 
1176 	if (!tc_skip_hw(f->flags))
1177 		fl_hw_update_stats(tp, f);
1178 
1179 	if (fl_dump_key_val(skb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
1180 			    mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
1181 			    sizeof(key->eth.dst)) ||
1182 	    fl_dump_key_val(skb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
1183 			    mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
1184 			    sizeof(key->eth.src)) ||
1185 	    fl_dump_key_val(skb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE,
1186 			    &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
1187 			    sizeof(key->basic.n_proto)))
1188 		goto nla_put_failure;
1189 
1190 	if (fl_dump_key_mpls(skb, &key->mpls, &mask->mpls))
1191 		goto nla_put_failure;
1192 
1193 	if (fl_dump_key_vlan(skb, &key->vlan, &mask->vlan))
1194 		goto nla_put_failure;
1195 
1196 	if ((key->basic.n_proto == htons(ETH_P_IP) ||
1197 	     key->basic.n_proto == htons(ETH_P_IPV6)) &&
1198 	    (fl_dump_key_val(skb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
1199 			    &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
1200 			    sizeof(key->basic.ip_proto)) ||
1201 	    fl_dump_key_ip(skb, &key->ip, &mask->ip)))
1202 		goto nla_put_failure;
1203 
1204 	if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
1205 	    (fl_dump_key_val(skb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
1206 			     &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
1207 			     sizeof(key->ipv4.src)) ||
1208 	     fl_dump_key_val(skb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
1209 			     &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
1210 			     sizeof(key->ipv4.dst))))
1211 		goto nla_put_failure;
1212 	else if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
1213 		 (fl_dump_key_val(skb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
1214 				  &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
1215 				  sizeof(key->ipv6.src)) ||
1216 		  fl_dump_key_val(skb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
1217 				  &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
1218 				  sizeof(key->ipv6.dst))))
1219 		goto nla_put_failure;
1220 
1221 	if (key->basic.ip_proto == IPPROTO_TCP &&
1222 	    (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
1223 			     &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
1224 			     sizeof(key->tp.src)) ||
1225 	     fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
1226 			     &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
1227 			     sizeof(key->tp.dst)) ||
1228 	     fl_dump_key_val(skb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
1229 			     &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
1230 			     sizeof(key->tcp.flags))))
1231 		goto nla_put_failure;
1232 	else if (key->basic.ip_proto == IPPROTO_UDP &&
1233 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
1234 				  &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
1235 				  sizeof(key->tp.src)) ||
1236 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
1237 				  &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
1238 				  sizeof(key->tp.dst))))
1239 		goto nla_put_failure;
1240 	else if (key->basic.ip_proto == IPPROTO_SCTP &&
1241 		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
1242 				  &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
1243 				  sizeof(key->tp.src)) ||
1244 		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
1245 				  &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
1246 				  sizeof(key->tp.dst))))
1247 		goto nla_put_failure;
1248 	else if (key->basic.n_proto == htons(ETH_P_IP) &&
1249 		 key->basic.ip_proto == IPPROTO_ICMP &&
1250 		 (fl_dump_key_val(skb, &key->icmp.type,
1251 				  TCA_FLOWER_KEY_ICMPV4_TYPE, &mask->icmp.type,
1252 				  TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
1253 				  sizeof(key->icmp.type)) ||
1254 		  fl_dump_key_val(skb, &key->icmp.code,
1255 				  TCA_FLOWER_KEY_ICMPV4_CODE, &mask->icmp.code,
1256 				  TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
1257 				  sizeof(key->icmp.code))))
1258 		goto nla_put_failure;
1259 	else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
1260 		 key->basic.ip_proto == IPPROTO_ICMPV6 &&
1261 		 (fl_dump_key_val(skb, &key->icmp.type,
1262 				  TCA_FLOWER_KEY_ICMPV6_TYPE, &mask->icmp.type,
1263 				  TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
1264 				  sizeof(key->icmp.type)) ||
1265 		  fl_dump_key_val(skb, &key->icmp.code,
1266 				  TCA_FLOWER_KEY_ICMPV6_CODE, &mask->icmp.code,
1267 				  TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
1268 				  sizeof(key->icmp.code))))
1269 		goto nla_put_failure;
1270 	else if ((key->basic.n_proto == htons(ETH_P_ARP) ||
1271 		  key->basic.n_proto == htons(ETH_P_RARP)) &&
1272 		 (fl_dump_key_val(skb, &key->arp.sip,
1273 				  TCA_FLOWER_KEY_ARP_SIP, &mask->arp.sip,
1274 				  TCA_FLOWER_KEY_ARP_SIP_MASK,
1275 				  sizeof(key->arp.sip)) ||
1276 		  fl_dump_key_val(skb, &key->arp.tip,
1277 				  TCA_FLOWER_KEY_ARP_TIP, &mask->arp.tip,
1278 				  TCA_FLOWER_KEY_ARP_TIP_MASK,
1279 				  sizeof(key->arp.tip)) ||
1280 		  fl_dump_key_val(skb, &key->arp.op,
1281 				  TCA_FLOWER_KEY_ARP_OP, &mask->arp.op,
1282 				  TCA_FLOWER_KEY_ARP_OP_MASK,
1283 				  sizeof(key->arp.op)) ||
1284 		  fl_dump_key_val(skb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
1285 				  mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
1286 				  sizeof(key->arp.sha)) ||
1287 		  fl_dump_key_val(skb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
1288 				  mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
1289 				  sizeof(key->arp.tha))))
1290 		goto nla_put_failure;
1291 
1292 	if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
1293 	    (fl_dump_key_val(skb, &key->enc_ipv4.src,
1294 			    TCA_FLOWER_KEY_ENC_IPV4_SRC, &mask->enc_ipv4.src,
1295 			    TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
1296 			    sizeof(key->enc_ipv4.src)) ||
1297 	     fl_dump_key_val(skb, &key->enc_ipv4.dst,
1298 			     TCA_FLOWER_KEY_ENC_IPV4_DST, &mask->enc_ipv4.dst,
1299 			     TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
1300 			     sizeof(key->enc_ipv4.dst))))
1301 		goto nla_put_failure;
1302 	else if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
1303 		 (fl_dump_key_val(skb, &key->enc_ipv6.src,
1304 			    TCA_FLOWER_KEY_ENC_IPV6_SRC, &mask->enc_ipv6.src,
1305 			    TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
1306 			    sizeof(key->enc_ipv6.src)) ||
1307 		 fl_dump_key_val(skb, &key->enc_ipv6.dst,
1308 				 TCA_FLOWER_KEY_ENC_IPV6_DST,
1309 				 &mask->enc_ipv6.dst,
1310 				 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
1311 			    sizeof(key->enc_ipv6.dst))))
1312 		goto nla_put_failure;
1313 
1314 	if (fl_dump_key_val(skb, &key->enc_key_id, TCA_FLOWER_KEY_ENC_KEY_ID,
1315 			    &mask->enc_key_id, TCA_FLOWER_UNSPEC,
1316 			    sizeof(key->enc_key_id)) ||
1317 	    fl_dump_key_val(skb, &key->enc_tp.src,
1318 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
1319 			    &mask->enc_tp.src,
1320 			    TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
1321 			    sizeof(key->enc_tp.src)) ||
1322 	    fl_dump_key_val(skb, &key->enc_tp.dst,
1323 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
1324 			    &mask->enc_tp.dst,
1325 			    TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
1326 			    sizeof(key->enc_tp.dst)))
1327 		goto nla_put_failure;
1328 
1329 	if (fl_dump_key_flags(skb, key->control.flags, mask->control.flags))
1330 		goto nla_put_failure;
1331 
1332 	if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
1333 		goto nla_put_failure;
1334 
1335 	if (tcf_exts_dump(skb, &f->exts))
1336 		goto nla_put_failure;
1337 
1338 	nla_nest_end(skb, nest);
1339 
1340 	if (tcf_exts_dump_stats(skb, &f->exts) < 0)
1341 		goto nla_put_failure;
1342 
1343 	return skb->len;
1344 
1345 nla_put_failure:
1346 	nla_nest_cancel(skb, nest);
1347 	return -1;
1348 }
1349 
1350 static void fl_bind_class(void *fh, u32 classid, unsigned long cl)
1351 {
1352 	struct cls_fl_filter *f = fh;
1353 
1354 	if (f && f->res.classid == classid)
1355 		f->res.class = cl;
1356 }
1357 
1358 static struct tcf_proto_ops cls_fl_ops __read_mostly = {
1359 	.kind		= "flower",
1360 	.classify	= fl_classify,
1361 	.init		= fl_init,
1362 	.destroy	= fl_destroy,
1363 	.get		= fl_get,
1364 	.change		= fl_change,
1365 	.delete		= fl_delete,
1366 	.walk		= fl_walk,
1367 	.dump		= fl_dump,
1368 	.bind_class	= fl_bind_class,
1369 	.owner		= THIS_MODULE,
1370 };
1371 
1372 static int __init cls_fl_init(void)
1373 {
1374 	return register_tcf_proto_ops(&cls_fl_ops);
1375 }
1376 
1377 static void __exit cls_fl_exit(void)
1378 {
1379 	unregister_tcf_proto_ops(&cls_fl_ops);
1380 }
1381 
1382 module_init(cls_fl_init);
1383 module_exit(cls_fl_exit);
1384 
1385 MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
1386 MODULE_DESCRIPTION("Flower classifier");
1387 MODULE_LICENSE("GPL v2");
1388