Loading...
Note: File does not exist in v3.1.
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * net/sched/cls_flower.c Flower classifier
4 *
5 * Copyright (c) 2015 Jiri Pirko <jiri@resnulli.us>
6 */
7
8#include <linux/kernel.h>
9#include <linux/init.h>
10#include <linux/module.h>
11#include <linux/rhashtable.h>
12#include <linux/workqueue.h>
13#include <linux/refcount.h>
14
15#include <linux/if_ether.h>
16#include <linux/in6.h>
17#include <linux/ip.h>
18#include <linux/mpls.h>
19
20#include <net/sch_generic.h>
21#include <net/pkt_cls.h>
22#include <net/ip.h>
23#include <net/flow_dissector.h>
24#include <net/geneve.h>
25#include <net/vxlan.h>
26#include <net/erspan.h>
27
28#include <net/dst.h>
29#include <net/dst_metadata.h>
30
31#include <uapi/linux/netfilter/nf_conntrack_common.h>
32
33struct fl_flow_key {
34 struct flow_dissector_key_meta meta;
35 struct flow_dissector_key_control control;
36 struct flow_dissector_key_control enc_control;
37 struct flow_dissector_key_basic basic;
38 struct flow_dissector_key_eth_addrs eth;
39 struct flow_dissector_key_vlan vlan;
40 struct flow_dissector_key_vlan cvlan;
41 union {
42 struct flow_dissector_key_ipv4_addrs ipv4;
43 struct flow_dissector_key_ipv6_addrs ipv6;
44 };
45 struct flow_dissector_key_ports tp;
46 struct flow_dissector_key_icmp icmp;
47 struct flow_dissector_key_arp arp;
48 struct flow_dissector_key_keyid enc_key_id;
49 union {
50 struct flow_dissector_key_ipv4_addrs enc_ipv4;
51 struct flow_dissector_key_ipv6_addrs enc_ipv6;
52 };
53 struct flow_dissector_key_ports enc_tp;
54 struct flow_dissector_key_mpls mpls;
55 struct flow_dissector_key_tcp tcp;
56 struct flow_dissector_key_ip ip;
57 struct flow_dissector_key_ip enc_ip;
58 struct flow_dissector_key_enc_opts enc_opts;
59 union {
60 struct flow_dissector_key_ports tp;
61 struct {
62 struct flow_dissector_key_ports tp_min;
63 struct flow_dissector_key_ports tp_max;
64 };
65 } tp_range;
66 struct flow_dissector_key_ct ct;
67 struct flow_dissector_key_hash hash;
68} __aligned(BITS_PER_LONG / 8); /* Ensure that we can do comparisons as longs. */
69
70struct fl_flow_mask_range {
71 unsigned short int start;
72 unsigned short int end;
73};
74
75struct fl_flow_mask {
76 struct fl_flow_key key;
77 struct fl_flow_mask_range range;
78 u32 flags;
79 struct rhash_head ht_node;
80 struct rhashtable ht;
81 struct rhashtable_params filter_ht_params;
82 struct flow_dissector dissector;
83 struct list_head filters;
84 struct rcu_work rwork;
85 struct list_head list;
86 refcount_t refcnt;
87};
88
89struct fl_flow_tmplt {
90 struct fl_flow_key dummy_key;
91 struct fl_flow_key mask;
92 struct flow_dissector dissector;
93 struct tcf_chain *chain;
94};
95
96struct cls_fl_head {
97 struct rhashtable ht;
98 spinlock_t masks_lock; /* Protect masks list */
99 struct list_head masks;
100 struct list_head hw_filters;
101 struct rcu_work rwork;
102 struct idr handle_idr;
103};
104
105struct cls_fl_filter {
106 struct fl_flow_mask *mask;
107 struct rhash_head ht_node;
108 struct fl_flow_key mkey;
109 struct tcf_exts exts;
110 struct tcf_result res;
111 struct fl_flow_key key;
112 struct list_head list;
113 struct list_head hw_list;
114 u32 handle;
115 u32 flags;
116 u32 in_hw_count;
117 struct rcu_work rwork;
118 struct net_device *hw_dev;
119 /* Flower classifier is unlocked, which means that its reference counter
120 * can be changed concurrently without any kind of external
121 * synchronization. Use atomic reference counter to be concurrency-safe.
122 */
123 refcount_t refcnt;
124 bool deleted;
125};
126
127static const struct rhashtable_params mask_ht_params = {
128 .key_offset = offsetof(struct fl_flow_mask, key),
129 .key_len = sizeof(struct fl_flow_key),
130 .head_offset = offsetof(struct fl_flow_mask, ht_node),
131 .automatic_shrinking = true,
132};
133
134static unsigned short int fl_mask_range(const struct fl_flow_mask *mask)
135{
136 return mask->range.end - mask->range.start;
137}
138
139static void fl_mask_update_range(struct fl_flow_mask *mask)
140{
141 const u8 *bytes = (const u8 *) &mask->key;
142 size_t size = sizeof(mask->key);
143 size_t i, first = 0, last;
144
145 for (i = 0; i < size; i++) {
146 if (bytes[i]) {
147 first = i;
148 break;
149 }
150 }
151 last = first;
152 for (i = size - 1; i != first; i--) {
153 if (bytes[i]) {
154 last = i;
155 break;
156 }
157 }
158 mask->range.start = rounddown(first, sizeof(long));
159 mask->range.end = roundup(last + 1, sizeof(long));
160}
161
162static void *fl_key_get_start(struct fl_flow_key *key,
163 const struct fl_flow_mask *mask)
164{
165 return (u8 *) key + mask->range.start;
166}
167
168static void fl_set_masked_key(struct fl_flow_key *mkey, struct fl_flow_key *key,
169 struct fl_flow_mask *mask)
170{
171 const long *lkey = fl_key_get_start(key, mask);
172 const long *lmask = fl_key_get_start(&mask->key, mask);
173 long *lmkey = fl_key_get_start(mkey, mask);
174 int i;
175
176 for (i = 0; i < fl_mask_range(mask); i += sizeof(long))
177 *lmkey++ = *lkey++ & *lmask++;
178}
179
180static bool fl_mask_fits_tmplt(struct fl_flow_tmplt *tmplt,
181 struct fl_flow_mask *mask)
182{
183 const long *lmask = fl_key_get_start(&mask->key, mask);
184 const long *ltmplt;
185 int i;
186
187 if (!tmplt)
188 return true;
189 ltmplt = fl_key_get_start(&tmplt->mask, mask);
190 for (i = 0; i < fl_mask_range(mask); i += sizeof(long)) {
191 if (~*ltmplt++ & *lmask++)
192 return false;
193 }
194 return true;
195}
196
197static void fl_clear_masked_range(struct fl_flow_key *key,
198 struct fl_flow_mask *mask)
199{
200 memset(fl_key_get_start(key, mask), 0, fl_mask_range(mask));
201}
202
203static bool fl_range_port_dst_cmp(struct cls_fl_filter *filter,
204 struct fl_flow_key *key,
205 struct fl_flow_key *mkey)
206{
207 __be16 min_mask, max_mask, min_val, max_val;
208
209 min_mask = htons(filter->mask->key.tp_range.tp_min.dst);
210 max_mask = htons(filter->mask->key.tp_range.tp_max.dst);
211 min_val = htons(filter->key.tp_range.tp_min.dst);
212 max_val = htons(filter->key.tp_range.tp_max.dst);
213
214 if (min_mask && max_mask) {
215 if (htons(key->tp_range.tp.dst) < min_val ||
216 htons(key->tp_range.tp.dst) > max_val)
217 return false;
218
219 /* skb does not have min and max values */
220 mkey->tp_range.tp_min.dst = filter->mkey.tp_range.tp_min.dst;
221 mkey->tp_range.tp_max.dst = filter->mkey.tp_range.tp_max.dst;
222 }
223 return true;
224}
225
226static bool fl_range_port_src_cmp(struct cls_fl_filter *filter,
227 struct fl_flow_key *key,
228 struct fl_flow_key *mkey)
229{
230 __be16 min_mask, max_mask, min_val, max_val;
231
232 min_mask = htons(filter->mask->key.tp_range.tp_min.src);
233 max_mask = htons(filter->mask->key.tp_range.tp_max.src);
234 min_val = htons(filter->key.tp_range.tp_min.src);
235 max_val = htons(filter->key.tp_range.tp_max.src);
236
237 if (min_mask && max_mask) {
238 if (htons(key->tp_range.tp.src) < min_val ||
239 htons(key->tp_range.tp.src) > max_val)
240 return false;
241
242 /* skb does not have min and max values */
243 mkey->tp_range.tp_min.src = filter->mkey.tp_range.tp_min.src;
244 mkey->tp_range.tp_max.src = filter->mkey.tp_range.tp_max.src;
245 }
246 return true;
247}
248
249static struct cls_fl_filter *__fl_lookup(struct fl_flow_mask *mask,
250 struct fl_flow_key *mkey)
251{
252 return rhashtable_lookup_fast(&mask->ht, fl_key_get_start(mkey, mask),
253 mask->filter_ht_params);
254}
255
256static struct cls_fl_filter *fl_lookup_range(struct fl_flow_mask *mask,
257 struct fl_flow_key *mkey,
258 struct fl_flow_key *key)
259{
260 struct cls_fl_filter *filter, *f;
261
262 list_for_each_entry_rcu(filter, &mask->filters, list) {
263 if (!fl_range_port_dst_cmp(filter, key, mkey))
264 continue;
265
266 if (!fl_range_port_src_cmp(filter, key, mkey))
267 continue;
268
269 f = __fl_lookup(mask, mkey);
270 if (f)
271 return f;
272 }
273 return NULL;
274}
275
276static noinline_for_stack
277struct cls_fl_filter *fl_mask_lookup(struct fl_flow_mask *mask, struct fl_flow_key *key)
278{
279 struct fl_flow_key mkey;
280
281 fl_set_masked_key(&mkey, key, mask);
282 if ((mask->flags & TCA_FLOWER_MASK_FLAGS_RANGE))
283 return fl_lookup_range(mask, &mkey, key);
284
285 return __fl_lookup(mask, &mkey);
286}
287
288static u16 fl_ct_info_to_flower_map[] = {
289 [IP_CT_ESTABLISHED] = TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
290 TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED,
291 [IP_CT_RELATED] = TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
292 TCA_FLOWER_KEY_CT_FLAGS_RELATED,
293 [IP_CT_ESTABLISHED_REPLY] = TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
294 TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED,
295 [IP_CT_RELATED_REPLY] = TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
296 TCA_FLOWER_KEY_CT_FLAGS_RELATED,
297 [IP_CT_NEW] = TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
298 TCA_FLOWER_KEY_CT_FLAGS_NEW,
299};
300
301static int fl_classify(struct sk_buff *skb, const struct tcf_proto *tp,
302 struct tcf_result *res)
303{
304 struct cls_fl_head *head = rcu_dereference_bh(tp->root);
305 struct fl_flow_key skb_key;
306 struct fl_flow_mask *mask;
307 struct cls_fl_filter *f;
308
309 list_for_each_entry_rcu(mask, &head->masks, list) {
310 flow_dissector_init_keys(&skb_key.control, &skb_key.basic);
311 fl_clear_masked_range(&skb_key, mask);
312
313 skb_flow_dissect_meta(skb, &mask->dissector, &skb_key);
314 /* skb_flow_dissect() does not set n_proto in case an unknown
315 * protocol, so do it rather here.
316 */
317 skb_key.basic.n_proto = skb_protocol(skb, false);
318 skb_flow_dissect_tunnel_info(skb, &mask->dissector, &skb_key);
319 skb_flow_dissect_ct(skb, &mask->dissector, &skb_key,
320 fl_ct_info_to_flower_map,
321 ARRAY_SIZE(fl_ct_info_to_flower_map));
322 skb_flow_dissect_hash(skb, &mask->dissector, &skb_key);
323 skb_flow_dissect(skb, &mask->dissector, &skb_key, 0);
324
325 f = fl_mask_lookup(mask, &skb_key);
326 if (f && !tc_skip_sw(f->flags)) {
327 *res = f->res;
328 return tcf_exts_exec(skb, &f->exts, res);
329 }
330 }
331 return -1;
332}
333
334static int fl_init(struct tcf_proto *tp)
335{
336 struct cls_fl_head *head;
337
338 head = kzalloc(sizeof(*head), GFP_KERNEL);
339 if (!head)
340 return -ENOBUFS;
341
342 spin_lock_init(&head->masks_lock);
343 INIT_LIST_HEAD_RCU(&head->masks);
344 INIT_LIST_HEAD(&head->hw_filters);
345 rcu_assign_pointer(tp->root, head);
346 idr_init(&head->handle_idr);
347
348 return rhashtable_init(&head->ht, &mask_ht_params);
349}
350
351static void fl_mask_free(struct fl_flow_mask *mask, bool mask_init_done)
352{
353 /* temporary masks don't have their filters list and ht initialized */
354 if (mask_init_done) {
355 WARN_ON(!list_empty(&mask->filters));
356 rhashtable_destroy(&mask->ht);
357 }
358 kfree(mask);
359}
360
361static void fl_mask_free_work(struct work_struct *work)
362{
363 struct fl_flow_mask *mask = container_of(to_rcu_work(work),
364 struct fl_flow_mask, rwork);
365
366 fl_mask_free(mask, true);
367}
368
369static void fl_uninit_mask_free_work(struct work_struct *work)
370{
371 struct fl_flow_mask *mask = container_of(to_rcu_work(work),
372 struct fl_flow_mask, rwork);
373
374 fl_mask_free(mask, false);
375}
376
377static bool fl_mask_put(struct cls_fl_head *head, struct fl_flow_mask *mask)
378{
379 if (!refcount_dec_and_test(&mask->refcnt))
380 return false;
381
382 rhashtable_remove_fast(&head->ht, &mask->ht_node, mask_ht_params);
383
384 spin_lock(&head->masks_lock);
385 list_del_rcu(&mask->list);
386 spin_unlock(&head->masks_lock);
387
388 tcf_queue_work(&mask->rwork, fl_mask_free_work);
389
390 return true;
391}
392
393static struct cls_fl_head *fl_head_dereference(struct tcf_proto *tp)
394{
395 /* Flower classifier only changes root pointer during init and destroy.
396 * Users must obtain reference to tcf_proto instance before calling its
397 * API, so tp->root pointer is protected from concurrent call to
398 * fl_destroy() by reference counting.
399 */
400 return rcu_dereference_raw(tp->root);
401}
402
403static void __fl_destroy_filter(struct cls_fl_filter *f)
404{
405 tcf_exts_destroy(&f->exts);
406 tcf_exts_put_net(&f->exts);
407 kfree(f);
408}
409
410static void fl_destroy_filter_work(struct work_struct *work)
411{
412 struct cls_fl_filter *f = container_of(to_rcu_work(work),
413 struct cls_fl_filter, rwork);
414
415 __fl_destroy_filter(f);
416}
417
418static void fl_hw_destroy_filter(struct tcf_proto *tp, struct cls_fl_filter *f,
419 bool rtnl_held, struct netlink_ext_ack *extack)
420{
421 struct tcf_block *block = tp->chain->block;
422 struct flow_cls_offload cls_flower = {};
423
424 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
425 cls_flower.command = FLOW_CLS_DESTROY;
426 cls_flower.cookie = (unsigned long) f;
427
428 tc_setup_cb_destroy(block, tp, TC_SETUP_CLSFLOWER, &cls_flower, false,
429 &f->flags, &f->in_hw_count, rtnl_held);
430
431}
432
433static int fl_hw_replace_filter(struct tcf_proto *tp,
434 struct cls_fl_filter *f, bool rtnl_held,
435 struct netlink_ext_ack *extack)
436{
437 struct tcf_block *block = tp->chain->block;
438 struct flow_cls_offload cls_flower = {};
439 bool skip_sw = tc_skip_sw(f->flags);
440 int err = 0;
441
442 cls_flower.rule = flow_rule_alloc(tcf_exts_num_actions(&f->exts));
443 if (!cls_flower.rule)
444 return -ENOMEM;
445
446 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
447 cls_flower.command = FLOW_CLS_REPLACE;
448 cls_flower.cookie = (unsigned long) f;
449 cls_flower.rule->match.dissector = &f->mask->dissector;
450 cls_flower.rule->match.mask = &f->mask->key;
451 cls_flower.rule->match.key = &f->mkey;
452 cls_flower.classid = f->res.classid;
453
454 err = tc_setup_flow_action(&cls_flower.rule->action, &f->exts);
455 if (err) {
456 kfree(cls_flower.rule);
457 if (skip_sw) {
458 NL_SET_ERR_MSG_MOD(extack, "Failed to setup flow action");
459 return err;
460 }
461 return 0;
462 }
463
464 err = tc_setup_cb_add(block, tp, TC_SETUP_CLSFLOWER, &cls_flower,
465 skip_sw, &f->flags, &f->in_hw_count, rtnl_held);
466 tc_cleanup_flow_action(&cls_flower.rule->action);
467 kfree(cls_flower.rule);
468
469 if (err) {
470 fl_hw_destroy_filter(tp, f, rtnl_held, NULL);
471 return err;
472 }
473
474 if (skip_sw && !(f->flags & TCA_CLS_FLAGS_IN_HW))
475 return -EINVAL;
476
477 return 0;
478}
479
480static void fl_hw_update_stats(struct tcf_proto *tp, struct cls_fl_filter *f,
481 bool rtnl_held)
482{
483 struct tcf_block *block = tp->chain->block;
484 struct flow_cls_offload cls_flower = {};
485
486 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, NULL);
487 cls_flower.command = FLOW_CLS_STATS;
488 cls_flower.cookie = (unsigned long) f;
489 cls_flower.classid = f->res.classid;
490
491 tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false,
492 rtnl_held);
493
494 tcf_exts_stats_update(&f->exts, cls_flower.stats.bytes,
495 cls_flower.stats.pkts,
496 cls_flower.stats.drops,
497 cls_flower.stats.lastused,
498 cls_flower.stats.used_hw_stats,
499 cls_flower.stats.used_hw_stats_valid);
500}
501
502static void __fl_put(struct cls_fl_filter *f)
503{
504 if (!refcount_dec_and_test(&f->refcnt))
505 return;
506
507 if (tcf_exts_get_net(&f->exts))
508 tcf_queue_work(&f->rwork, fl_destroy_filter_work);
509 else
510 __fl_destroy_filter(f);
511}
512
513static struct cls_fl_filter *__fl_get(struct cls_fl_head *head, u32 handle)
514{
515 struct cls_fl_filter *f;
516
517 rcu_read_lock();
518 f = idr_find(&head->handle_idr, handle);
519 if (f && !refcount_inc_not_zero(&f->refcnt))
520 f = NULL;
521 rcu_read_unlock();
522
523 return f;
524}
525
526static int __fl_delete(struct tcf_proto *tp, struct cls_fl_filter *f,
527 bool *last, bool rtnl_held,
528 struct netlink_ext_ack *extack)
529{
530 struct cls_fl_head *head = fl_head_dereference(tp);
531
532 *last = false;
533
534 spin_lock(&tp->lock);
535 if (f->deleted) {
536 spin_unlock(&tp->lock);
537 return -ENOENT;
538 }
539
540 f->deleted = true;
541 rhashtable_remove_fast(&f->mask->ht, &f->ht_node,
542 f->mask->filter_ht_params);
543 idr_remove(&head->handle_idr, f->handle);
544 list_del_rcu(&f->list);
545 spin_unlock(&tp->lock);
546
547 *last = fl_mask_put(head, f->mask);
548 if (!tc_skip_hw(f->flags))
549 fl_hw_destroy_filter(tp, f, rtnl_held, extack);
550 tcf_unbind_filter(tp, &f->res);
551 __fl_put(f);
552
553 return 0;
554}
555
556static void fl_destroy_sleepable(struct work_struct *work)
557{
558 struct cls_fl_head *head = container_of(to_rcu_work(work),
559 struct cls_fl_head,
560 rwork);
561
562 rhashtable_destroy(&head->ht);
563 kfree(head);
564 module_put(THIS_MODULE);
565}
566
567static void fl_destroy(struct tcf_proto *tp, bool rtnl_held,
568 struct netlink_ext_ack *extack)
569{
570 struct cls_fl_head *head = fl_head_dereference(tp);
571 struct fl_flow_mask *mask, *next_mask;
572 struct cls_fl_filter *f, *next;
573 bool last;
574
575 list_for_each_entry_safe(mask, next_mask, &head->masks, list) {
576 list_for_each_entry_safe(f, next, &mask->filters, list) {
577 __fl_delete(tp, f, &last, rtnl_held, extack);
578 if (last)
579 break;
580 }
581 }
582 idr_destroy(&head->handle_idr);
583
584 __module_get(THIS_MODULE);
585 tcf_queue_work(&head->rwork, fl_destroy_sleepable);
586}
587
588static void fl_put(struct tcf_proto *tp, void *arg)
589{
590 struct cls_fl_filter *f = arg;
591
592 __fl_put(f);
593}
594
595static void *fl_get(struct tcf_proto *tp, u32 handle)
596{
597 struct cls_fl_head *head = fl_head_dereference(tp);
598
599 return __fl_get(head, handle);
600}
601
602static const struct nla_policy fl_policy[TCA_FLOWER_MAX + 1] = {
603 [TCA_FLOWER_UNSPEC] = { .type = NLA_UNSPEC },
604 [TCA_FLOWER_CLASSID] = { .type = NLA_U32 },
605 [TCA_FLOWER_INDEV] = { .type = NLA_STRING,
606 .len = IFNAMSIZ },
607 [TCA_FLOWER_KEY_ETH_DST] = { .len = ETH_ALEN },
608 [TCA_FLOWER_KEY_ETH_DST_MASK] = { .len = ETH_ALEN },
609 [TCA_FLOWER_KEY_ETH_SRC] = { .len = ETH_ALEN },
610 [TCA_FLOWER_KEY_ETH_SRC_MASK] = { .len = ETH_ALEN },
611 [TCA_FLOWER_KEY_ETH_TYPE] = { .type = NLA_U16 },
612 [TCA_FLOWER_KEY_IP_PROTO] = { .type = NLA_U8 },
613 [TCA_FLOWER_KEY_IPV4_SRC] = { .type = NLA_U32 },
614 [TCA_FLOWER_KEY_IPV4_SRC_MASK] = { .type = NLA_U32 },
615 [TCA_FLOWER_KEY_IPV4_DST] = { .type = NLA_U32 },
616 [TCA_FLOWER_KEY_IPV4_DST_MASK] = { .type = NLA_U32 },
617 [TCA_FLOWER_KEY_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
618 [TCA_FLOWER_KEY_IPV6_SRC_MASK] = { .len = sizeof(struct in6_addr) },
619 [TCA_FLOWER_KEY_IPV6_DST] = { .len = sizeof(struct in6_addr) },
620 [TCA_FLOWER_KEY_IPV6_DST_MASK] = { .len = sizeof(struct in6_addr) },
621 [TCA_FLOWER_KEY_TCP_SRC] = { .type = NLA_U16 },
622 [TCA_FLOWER_KEY_TCP_DST] = { .type = NLA_U16 },
623 [TCA_FLOWER_KEY_UDP_SRC] = { .type = NLA_U16 },
624 [TCA_FLOWER_KEY_UDP_DST] = { .type = NLA_U16 },
625 [TCA_FLOWER_KEY_VLAN_ID] = { .type = NLA_U16 },
626 [TCA_FLOWER_KEY_VLAN_PRIO] = { .type = NLA_U8 },
627 [TCA_FLOWER_KEY_VLAN_ETH_TYPE] = { .type = NLA_U16 },
628 [TCA_FLOWER_KEY_ENC_KEY_ID] = { .type = NLA_U32 },
629 [TCA_FLOWER_KEY_ENC_IPV4_SRC] = { .type = NLA_U32 },
630 [TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK] = { .type = NLA_U32 },
631 [TCA_FLOWER_KEY_ENC_IPV4_DST] = { .type = NLA_U32 },
632 [TCA_FLOWER_KEY_ENC_IPV4_DST_MASK] = { .type = NLA_U32 },
633 [TCA_FLOWER_KEY_ENC_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
634 [TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK] = { .len = sizeof(struct in6_addr) },
635 [TCA_FLOWER_KEY_ENC_IPV6_DST] = { .len = sizeof(struct in6_addr) },
636 [TCA_FLOWER_KEY_ENC_IPV6_DST_MASK] = { .len = sizeof(struct in6_addr) },
637 [TCA_FLOWER_KEY_TCP_SRC_MASK] = { .type = NLA_U16 },
638 [TCA_FLOWER_KEY_TCP_DST_MASK] = { .type = NLA_U16 },
639 [TCA_FLOWER_KEY_UDP_SRC_MASK] = { .type = NLA_U16 },
640 [TCA_FLOWER_KEY_UDP_DST_MASK] = { .type = NLA_U16 },
641 [TCA_FLOWER_KEY_SCTP_SRC_MASK] = { .type = NLA_U16 },
642 [TCA_FLOWER_KEY_SCTP_DST_MASK] = { .type = NLA_U16 },
643 [TCA_FLOWER_KEY_SCTP_SRC] = { .type = NLA_U16 },
644 [TCA_FLOWER_KEY_SCTP_DST] = { .type = NLA_U16 },
645 [TCA_FLOWER_KEY_ENC_UDP_SRC_PORT] = { .type = NLA_U16 },
646 [TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK] = { .type = NLA_U16 },
647 [TCA_FLOWER_KEY_ENC_UDP_DST_PORT] = { .type = NLA_U16 },
648 [TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK] = { .type = NLA_U16 },
649 [TCA_FLOWER_KEY_FLAGS] = { .type = NLA_U32 },
650 [TCA_FLOWER_KEY_FLAGS_MASK] = { .type = NLA_U32 },
651 [TCA_FLOWER_KEY_ICMPV4_TYPE] = { .type = NLA_U8 },
652 [TCA_FLOWER_KEY_ICMPV4_TYPE_MASK] = { .type = NLA_U8 },
653 [TCA_FLOWER_KEY_ICMPV4_CODE] = { .type = NLA_U8 },
654 [TCA_FLOWER_KEY_ICMPV4_CODE_MASK] = { .type = NLA_U8 },
655 [TCA_FLOWER_KEY_ICMPV6_TYPE] = { .type = NLA_U8 },
656 [TCA_FLOWER_KEY_ICMPV6_TYPE_MASK] = { .type = NLA_U8 },
657 [TCA_FLOWER_KEY_ICMPV6_CODE] = { .type = NLA_U8 },
658 [TCA_FLOWER_KEY_ICMPV6_CODE_MASK] = { .type = NLA_U8 },
659 [TCA_FLOWER_KEY_ARP_SIP] = { .type = NLA_U32 },
660 [TCA_FLOWER_KEY_ARP_SIP_MASK] = { .type = NLA_U32 },
661 [TCA_FLOWER_KEY_ARP_TIP] = { .type = NLA_U32 },
662 [TCA_FLOWER_KEY_ARP_TIP_MASK] = { .type = NLA_U32 },
663 [TCA_FLOWER_KEY_ARP_OP] = { .type = NLA_U8 },
664 [TCA_FLOWER_KEY_ARP_OP_MASK] = { .type = NLA_U8 },
665 [TCA_FLOWER_KEY_ARP_SHA] = { .len = ETH_ALEN },
666 [TCA_FLOWER_KEY_ARP_SHA_MASK] = { .len = ETH_ALEN },
667 [TCA_FLOWER_KEY_ARP_THA] = { .len = ETH_ALEN },
668 [TCA_FLOWER_KEY_ARP_THA_MASK] = { .len = ETH_ALEN },
669 [TCA_FLOWER_KEY_MPLS_TTL] = { .type = NLA_U8 },
670 [TCA_FLOWER_KEY_MPLS_BOS] = { .type = NLA_U8 },
671 [TCA_FLOWER_KEY_MPLS_TC] = { .type = NLA_U8 },
672 [TCA_FLOWER_KEY_MPLS_LABEL] = { .type = NLA_U32 },
673 [TCA_FLOWER_KEY_MPLS_OPTS] = { .type = NLA_NESTED },
674 [TCA_FLOWER_KEY_TCP_FLAGS] = { .type = NLA_U16 },
675 [TCA_FLOWER_KEY_TCP_FLAGS_MASK] = { .type = NLA_U16 },
676 [TCA_FLOWER_KEY_IP_TOS] = { .type = NLA_U8 },
677 [TCA_FLOWER_KEY_IP_TOS_MASK] = { .type = NLA_U8 },
678 [TCA_FLOWER_KEY_IP_TTL] = { .type = NLA_U8 },
679 [TCA_FLOWER_KEY_IP_TTL_MASK] = { .type = NLA_U8 },
680 [TCA_FLOWER_KEY_CVLAN_ID] = { .type = NLA_U16 },
681 [TCA_FLOWER_KEY_CVLAN_PRIO] = { .type = NLA_U8 },
682 [TCA_FLOWER_KEY_CVLAN_ETH_TYPE] = { .type = NLA_U16 },
683 [TCA_FLOWER_KEY_ENC_IP_TOS] = { .type = NLA_U8 },
684 [TCA_FLOWER_KEY_ENC_IP_TOS_MASK] = { .type = NLA_U8 },
685 [TCA_FLOWER_KEY_ENC_IP_TTL] = { .type = NLA_U8 },
686 [TCA_FLOWER_KEY_ENC_IP_TTL_MASK] = { .type = NLA_U8 },
687 [TCA_FLOWER_KEY_ENC_OPTS] = { .type = NLA_NESTED },
688 [TCA_FLOWER_KEY_ENC_OPTS_MASK] = { .type = NLA_NESTED },
689 [TCA_FLOWER_KEY_CT_STATE] = { .type = NLA_U16 },
690 [TCA_FLOWER_KEY_CT_STATE_MASK] = { .type = NLA_U16 },
691 [TCA_FLOWER_KEY_CT_ZONE] = { .type = NLA_U16 },
692 [TCA_FLOWER_KEY_CT_ZONE_MASK] = { .type = NLA_U16 },
693 [TCA_FLOWER_KEY_CT_MARK] = { .type = NLA_U32 },
694 [TCA_FLOWER_KEY_CT_MARK_MASK] = { .type = NLA_U32 },
695 [TCA_FLOWER_KEY_CT_LABELS] = { .type = NLA_BINARY,
696 .len = 128 / BITS_PER_BYTE },
697 [TCA_FLOWER_KEY_CT_LABELS_MASK] = { .type = NLA_BINARY,
698 .len = 128 / BITS_PER_BYTE },
699 [TCA_FLOWER_FLAGS] = { .type = NLA_U32 },
700 [TCA_FLOWER_KEY_HASH] = { .type = NLA_U32 },
701 [TCA_FLOWER_KEY_HASH_MASK] = { .type = NLA_U32 },
702
703};
704
705static const struct nla_policy
706enc_opts_policy[TCA_FLOWER_KEY_ENC_OPTS_MAX + 1] = {
707 [TCA_FLOWER_KEY_ENC_OPTS_UNSPEC] = {
708 .strict_start_type = TCA_FLOWER_KEY_ENC_OPTS_VXLAN },
709 [TCA_FLOWER_KEY_ENC_OPTS_GENEVE] = { .type = NLA_NESTED },
710 [TCA_FLOWER_KEY_ENC_OPTS_VXLAN] = { .type = NLA_NESTED },
711 [TCA_FLOWER_KEY_ENC_OPTS_ERSPAN] = { .type = NLA_NESTED },
712};
713
714static const struct nla_policy
715geneve_opt_policy[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1] = {
716 [TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS] = { .type = NLA_U16 },
717 [TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE] = { .type = NLA_U8 },
718 [TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA] = { .type = NLA_BINARY,
719 .len = 128 },
720};
721
722static const struct nla_policy
723vxlan_opt_policy[TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX + 1] = {
724 [TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP] = { .type = NLA_U32 },
725};
726
727static const struct nla_policy
728erspan_opt_policy[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX + 1] = {
729 [TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER] = { .type = NLA_U8 },
730 [TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX] = { .type = NLA_U32 },
731 [TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR] = { .type = NLA_U8 },
732 [TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID] = { .type = NLA_U8 },
733};
734
735static const struct nla_policy
736mpls_stack_entry_policy[TCA_FLOWER_KEY_MPLS_OPT_LSE_MAX + 1] = {
737 [TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH] = { .type = NLA_U8 },
738 [TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL] = { .type = NLA_U8 },
739 [TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS] = { .type = NLA_U8 },
740 [TCA_FLOWER_KEY_MPLS_OPT_LSE_TC] = { .type = NLA_U8 },
741 [TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL] = { .type = NLA_U32 },
742};
743
744static void fl_set_key_val(struct nlattr **tb,
745 void *val, int val_type,
746 void *mask, int mask_type, int len)
747{
748 if (!tb[val_type])
749 return;
750 nla_memcpy(val, tb[val_type], len);
751 if (mask_type == TCA_FLOWER_UNSPEC || !tb[mask_type])
752 memset(mask, 0xff, len);
753 else
754 nla_memcpy(mask, tb[mask_type], len);
755}
756
757static int fl_set_key_port_range(struct nlattr **tb, struct fl_flow_key *key,
758 struct fl_flow_key *mask,
759 struct netlink_ext_ack *extack)
760{
761 fl_set_key_val(tb, &key->tp_range.tp_min.dst,
762 TCA_FLOWER_KEY_PORT_DST_MIN, &mask->tp_range.tp_min.dst,
763 TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_min.dst));
764 fl_set_key_val(tb, &key->tp_range.tp_max.dst,
765 TCA_FLOWER_KEY_PORT_DST_MAX, &mask->tp_range.tp_max.dst,
766 TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_max.dst));
767 fl_set_key_val(tb, &key->tp_range.tp_min.src,
768 TCA_FLOWER_KEY_PORT_SRC_MIN, &mask->tp_range.tp_min.src,
769 TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_min.src));
770 fl_set_key_val(tb, &key->tp_range.tp_max.src,
771 TCA_FLOWER_KEY_PORT_SRC_MAX, &mask->tp_range.tp_max.src,
772 TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_max.src));
773
774 if (mask->tp_range.tp_min.dst && mask->tp_range.tp_max.dst &&
775 htons(key->tp_range.tp_max.dst) <=
776 htons(key->tp_range.tp_min.dst)) {
777 NL_SET_ERR_MSG_ATTR(extack,
778 tb[TCA_FLOWER_KEY_PORT_DST_MIN],
779 "Invalid destination port range (min must be strictly smaller than max)");
780 return -EINVAL;
781 }
782 if (mask->tp_range.tp_min.src && mask->tp_range.tp_max.src &&
783 htons(key->tp_range.tp_max.src) <=
784 htons(key->tp_range.tp_min.src)) {
785 NL_SET_ERR_MSG_ATTR(extack,
786 tb[TCA_FLOWER_KEY_PORT_SRC_MIN],
787 "Invalid source port range (min must be strictly smaller than max)");
788 return -EINVAL;
789 }
790
791 return 0;
792}
793
794static int fl_set_key_mpls_lse(const struct nlattr *nla_lse,
795 struct flow_dissector_key_mpls *key_val,
796 struct flow_dissector_key_mpls *key_mask,
797 struct netlink_ext_ack *extack)
798{
799 struct nlattr *tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_MAX + 1];
800 struct flow_dissector_mpls_lse *lse_mask;
801 struct flow_dissector_mpls_lse *lse_val;
802 u8 lse_index;
803 u8 depth;
804 int err;
805
806 err = nla_parse_nested(tb, TCA_FLOWER_KEY_MPLS_OPT_LSE_MAX, nla_lse,
807 mpls_stack_entry_policy, extack);
808 if (err < 0)
809 return err;
810
811 if (!tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH]) {
812 NL_SET_ERR_MSG(extack, "Missing MPLS option \"depth\"");
813 return -EINVAL;
814 }
815
816 depth = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH]);
817
818 /* LSE depth starts at 1, for consistency with terminology used by
819 * RFC 3031 (section 3.9), where depth 0 refers to unlabeled packets.
820 */
821 if (depth < 1 || depth > FLOW_DIS_MPLS_MAX) {
822 NL_SET_ERR_MSG_ATTR(extack,
823 tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH],
824 "Invalid MPLS depth");
825 return -EINVAL;
826 }
827 lse_index = depth - 1;
828
829 dissector_set_mpls_lse(key_val, lse_index);
830 dissector_set_mpls_lse(key_mask, lse_index);
831
832 lse_val = &key_val->ls[lse_index];
833 lse_mask = &key_mask->ls[lse_index];
834
835 if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL]) {
836 lse_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL]);
837 lse_mask->mpls_ttl = MPLS_TTL_MASK;
838 }
839 if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS]) {
840 u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS]);
841
842 if (bos & ~MPLS_BOS_MASK) {
843 NL_SET_ERR_MSG_ATTR(extack,
844 tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS],
845 "Bottom Of Stack (BOS) must be 0 or 1");
846 return -EINVAL;
847 }
848 lse_val->mpls_bos = bos;
849 lse_mask->mpls_bos = MPLS_BOS_MASK;
850 }
851 if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC]) {
852 u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC]);
853
854 if (tc & ~MPLS_TC_MASK) {
855 NL_SET_ERR_MSG_ATTR(extack,
856 tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC],
857 "Traffic Class (TC) must be between 0 and 7");
858 return -EINVAL;
859 }
860 lse_val->mpls_tc = tc;
861 lse_mask->mpls_tc = MPLS_TC_MASK;
862 }
863 if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL]) {
864 u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL]);
865
866 if (label & ~MPLS_LABEL_MASK) {
867 NL_SET_ERR_MSG_ATTR(extack,
868 tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL],
869 "Label must be between 0 and 1048575");
870 return -EINVAL;
871 }
872 lse_val->mpls_label = label;
873 lse_mask->mpls_label = MPLS_LABEL_MASK;
874 }
875
876 return 0;
877}
878
879static int fl_set_key_mpls_opts(const struct nlattr *nla_mpls_opts,
880 struct flow_dissector_key_mpls *key_val,
881 struct flow_dissector_key_mpls *key_mask,
882 struct netlink_ext_ack *extack)
883{
884 struct nlattr *nla_lse;
885 int rem;
886 int err;
887
888 if (!(nla_mpls_opts->nla_type & NLA_F_NESTED)) {
889 NL_SET_ERR_MSG_ATTR(extack, nla_mpls_opts,
890 "NLA_F_NESTED is missing");
891 return -EINVAL;
892 }
893
894 nla_for_each_nested(nla_lse, nla_mpls_opts, rem) {
895 if (nla_type(nla_lse) != TCA_FLOWER_KEY_MPLS_OPTS_LSE) {
896 NL_SET_ERR_MSG_ATTR(extack, nla_lse,
897 "Invalid MPLS option type");
898 return -EINVAL;
899 }
900
901 err = fl_set_key_mpls_lse(nla_lse, key_val, key_mask, extack);
902 if (err < 0)
903 return err;
904 }
905 if (rem) {
906 NL_SET_ERR_MSG(extack,
907 "Bytes leftover after parsing MPLS options");
908 return -EINVAL;
909 }
910
911 return 0;
912}
913
914static int fl_set_key_mpls(struct nlattr **tb,
915 struct flow_dissector_key_mpls *key_val,
916 struct flow_dissector_key_mpls *key_mask,
917 struct netlink_ext_ack *extack)
918{
919 struct flow_dissector_mpls_lse *lse_mask;
920 struct flow_dissector_mpls_lse *lse_val;
921
922 if (tb[TCA_FLOWER_KEY_MPLS_OPTS]) {
923 if (tb[TCA_FLOWER_KEY_MPLS_TTL] ||
924 tb[TCA_FLOWER_KEY_MPLS_BOS] ||
925 tb[TCA_FLOWER_KEY_MPLS_TC] ||
926 tb[TCA_FLOWER_KEY_MPLS_LABEL]) {
927 NL_SET_ERR_MSG_ATTR(extack,
928 tb[TCA_FLOWER_KEY_MPLS_OPTS],
929 "MPLS label, Traffic Class, Bottom Of Stack and Time To Live must be encapsulated in the MPLS options attribute");
930 return -EBADMSG;
931 }
932
933 return fl_set_key_mpls_opts(tb[TCA_FLOWER_KEY_MPLS_OPTS],
934 key_val, key_mask, extack);
935 }
936
937 lse_val = &key_val->ls[0];
938 lse_mask = &key_mask->ls[0];
939
940 if (tb[TCA_FLOWER_KEY_MPLS_TTL]) {
941 lse_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TTL]);
942 lse_mask->mpls_ttl = MPLS_TTL_MASK;
943 dissector_set_mpls_lse(key_val, 0);
944 dissector_set_mpls_lse(key_mask, 0);
945 }
946 if (tb[TCA_FLOWER_KEY_MPLS_BOS]) {
947 u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_BOS]);
948
949 if (bos & ~MPLS_BOS_MASK) {
950 NL_SET_ERR_MSG_ATTR(extack,
951 tb[TCA_FLOWER_KEY_MPLS_BOS],
952 "Bottom Of Stack (BOS) must be 0 or 1");
953 return -EINVAL;
954 }
955 lse_val->mpls_bos = bos;
956 lse_mask->mpls_bos = MPLS_BOS_MASK;
957 dissector_set_mpls_lse(key_val, 0);
958 dissector_set_mpls_lse(key_mask, 0);
959 }
960 if (tb[TCA_FLOWER_KEY_MPLS_TC]) {
961 u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TC]);
962
963 if (tc & ~MPLS_TC_MASK) {
964 NL_SET_ERR_MSG_ATTR(extack,
965 tb[TCA_FLOWER_KEY_MPLS_TC],
966 "Traffic Class (TC) must be between 0 and 7");
967 return -EINVAL;
968 }
969 lse_val->mpls_tc = tc;
970 lse_mask->mpls_tc = MPLS_TC_MASK;
971 dissector_set_mpls_lse(key_val, 0);
972 dissector_set_mpls_lse(key_mask, 0);
973 }
974 if (tb[TCA_FLOWER_KEY_MPLS_LABEL]) {
975 u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_LABEL]);
976
977 if (label & ~MPLS_LABEL_MASK) {
978 NL_SET_ERR_MSG_ATTR(extack,
979 tb[TCA_FLOWER_KEY_MPLS_LABEL],
980 "Label must be between 0 and 1048575");
981 return -EINVAL;
982 }
983 lse_val->mpls_label = label;
984 lse_mask->mpls_label = MPLS_LABEL_MASK;
985 dissector_set_mpls_lse(key_val, 0);
986 dissector_set_mpls_lse(key_mask, 0);
987 }
988 return 0;
989}
990
991static void fl_set_key_vlan(struct nlattr **tb,
992 __be16 ethertype,
993 int vlan_id_key, int vlan_prio_key,
994 struct flow_dissector_key_vlan *key_val,
995 struct flow_dissector_key_vlan *key_mask)
996{
997#define VLAN_PRIORITY_MASK 0x7
998
999 if (tb[vlan_id_key]) {
1000 key_val->vlan_id =
1001 nla_get_u16(tb[vlan_id_key]) & VLAN_VID_MASK;
1002 key_mask->vlan_id = VLAN_VID_MASK;
1003 }
1004 if (tb[vlan_prio_key]) {
1005 key_val->vlan_priority =
1006 nla_get_u8(tb[vlan_prio_key]) &
1007 VLAN_PRIORITY_MASK;
1008 key_mask->vlan_priority = VLAN_PRIORITY_MASK;
1009 }
1010 key_val->vlan_tpid = ethertype;
1011 key_mask->vlan_tpid = cpu_to_be16(~0);
1012}
1013
1014static void fl_set_key_flag(u32 flower_key, u32 flower_mask,
1015 u32 *dissector_key, u32 *dissector_mask,
1016 u32 flower_flag_bit, u32 dissector_flag_bit)
1017{
1018 if (flower_mask & flower_flag_bit) {
1019 *dissector_mask |= dissector_flag_bit;
1020 if (flower_key & flower_flag_bit)
1021 *dissector_key |= dissector_flag_bit;
1022 }
1023}
1024
1025static int fl_set_key_flags(struct nlattr **tb, u32 *flags_key,
1026 u32 *flags_mask, struct netlink_ext_ack *extack)
1027{
1028 u32 key, mask;
1029
1030 /* mask is mandatory for flags */
1031 if (!tb[TCA_FLOWER_KEY_FLAGS_MASK]) {
1032 NL_SET_ERR_MSG(extack, "Missing flags mask");
1033 return -EINVAL;
1034 }
1035
1036 key = be32_to_cpu(nla_get_u32(tb[TCA_FLOWER_KEY_FLAGS]));
1037 mask = be32_to_cpu(nla_get_u32(tb[TCA_FLOWER_KEY_FLAGS_MASK]));
1038
1039 *flags_key = 0;
1040 *flags_mask = 0;
1041
1042 fl_set_key_flag(key, mask, flags_key, flags_mask,
1043 TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
1044 fl_set_key_flag(key, mask, flags_key, flags_mask,
1045 TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
1046 FLOW_DIS_FIRST_FRAG);
1047
1048 return 0;
1049}
1050
1051static void fl_set_key_ip(struct nlattr **tb, bool encap,
1052 struct flow_dissector_key_ip *key,
1053 struct flow_dissector_key_ip *mask)
1054{
1055 int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
1056 int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
1057 int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
1058 int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
1059
1060 fl_set_key_val(tb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos));
1061 fl_set_key_val(tb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl));
1062}
1063
1064static int fl_set_geneve_opt(const struct nlattr *nla, struct fl_flow_key *key,
1065 int depth, int option_len,
1066 struct netlink_ext_ack *extack)
1067{
1068 struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1];
1069 struct nlattr *class = NULL, *type = NULL, *data = NULL;
1070 struct geneve_opt *opt;
1071 int err, data_len = 0;
1072
1073 if (option_len > sizeof(struct geneve_opt))
1074 data_len = option_len - sizeof(struct geneve_opt);
1075
1076 opt = (struct geneve_opt *)&key->enc_opts.data[key->enc_opts.len];
1077 memset(opt, 0xff, option_len);
1078 opt->length = data_len / 4;
1079 opt->r1 = 0;
1080 opt->r2 = 0;
1081 opt->r3 = 0;
1082
1083 /* If no mask has been prodived we assume an exact match. */
1084 if (!depth)
1085 return sizeof(struct geneve_opt) + data_len;
1086
1087 if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_GENEVE) {
1088 NL_SET_ERR_MSG(extack, "Non-geneve option type for mask");
1089 return -EINVAL;
1090 }
1091
1092 err = nla_parse_nested_deprecated(tb,
1093 TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX,
1094 nla, geneve_opt_policy, extack);
1095 if (err < 0)
1096 return err;
1097
1098 /* We are not allowed to omit any of CLASS, TYPE or DATA
1099 * fields from the key.
1100 */
1101 if (!option_len &&
1102 (!tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS] ||
1103 !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE] ||
1104 !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA])) {
1105 NL_SET_ERR_MSG(extack, "Missing tunnel key geneve option class, type or data");
1106 return -EINVAL;
1107 }
1108
1109 /* Omitting any of CLASS, TYPE or DATA fields is allowed
1110 * for the mask.
1111 */
1112 if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA]) {
1113 int new_len = key->enc_opts.len;
1114
1115 data = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA];
1116 data_len = nla_len(data);
1117 if (data_len < 4) {
1118 NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is less than 4 bytes long");
1119 return -ERANGE;
1120 }
1121 if (data_len % 4) {
1122 NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is not a multiple of 4 bytes long");
1123 return -ERANGE;
1124 }
1125
1126 new_len += sizeof(struct geneve_opt) + data_len;
1127 BUILD_BUG_ON(FLOW_DIS_TUN_OPTS_MAX != IP_TUNNEL_OPTS_MAX);
1128 if (new_len > FLOW_DIS_TUN_OPTS_MAX) {
1129 NL_SET_ERR_MSG(extack, "Tunnel options exceeds max size");
1130 return -ERANGE;
1131 }
1132 opt->length = data_len / 4;
1133 memcpy(opt->opt_data, nla_data(data), data_len);
1134 }
1135
1136 if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS]) {
1137 class = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS];
1138 opt->opt_class = nla_get_be16(class);
1139 }
1140
1141 if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE]) {
1142 type = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE];
1143 opt->type = nla_get_u8(type);
1144 }
1145
1146 return sizeof(struct geneve_opt) + data_len;
1147}
1148
1149static int fl_set_vxlan_opt(const struct nlattr *nla, struct fl_flow_key *key,
1150 int depth, int option_len,
1151 struct netlink_ext_ack *extack)
1152{
1153 struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX + 1];
1154 struct vxlan_metadata *md;
1155 int err;
1156
1157 md = (struct vxlan_metadata *)&key->enc_opts.data[key->enc_opts.len];
1158 memset(md, 0xff, sizeof(*md));
1159
1160 if (!depth)
1161 return sizeof(*md);
1162
1163 if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_VXLAN) {
1164 NL_SET_ERR_MSG(extack, "Non-vxlan option type for mask");
1165 return -EINVAL;
1166 }
1167
1168 err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX, nla,
1169 vxlan_opt_policy, extack);
1170 if (err < 0)
1171 return err;
1172
1173 if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]) {
1174 NL_SET_ERR_MSG(extack, "Missing tunnel key vxlan option gbp");
1175 return -EINVAL;
1176 }
1177
1178 if (tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]) {
1179 md->gbp = nla_get_u32(tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]);
1180 md->gbp &= VXLAN_GBP_MASK;
1181 }
1182
1183 return sizeof(*md);
1184}
1185
1186static int fl_set_erspan_opt(const struct nlattr *nla, struct fl_flow_key *key,
1187 int depth, int option_len,
1188 struct netlink_ext_ack *extack)
1189{
1190 struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX + 1];
1191 struct erspan_metadata *md;
1192 int err;
1193
1194 md = (struct erspan_metadata *)&key->enc_opts.data[key->enc_opts.len];
1195 memset(md, 0xff, sizeof(*md));
1196 md->version = 1;
1197
1198 if (!depth)
1199 return sizeof(*md);
1200
1201 if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_ERSPAN) {
1202 NL_SET_ERR_MSG(extack, "Non-erspan option type for mask");
1203 return -EINVAL;
1204 }
1205
1206 err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX, nla,
1207 erspan_opt_policy, extack);
1208 if (err < 0)
1209 return err;
1210
1211 if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]) {
1212 NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option ver");
1213 return -EINVAL;
1214 }
1215
1216 if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER])
1217 md->version = nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]);
1218
1219 if (md->version == 1) {
1220 if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]) {
1221 NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option index");
1222 return -EINVAL;
1223 }
1224 if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]) {
1225 nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX];
1226 memset(&md->u, 0x00, sizeof(md->u));
1227 md->u.index = nla_get_be32(nla);
1228 }
1229 } else if (md->version == 2) {
1230 if (!option_len && (!tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR] ||
1231 !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID])) {
1232 NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option dir or hwid");
1233 return -EINVAL;
1234 }
1235 if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR]) {
1236 nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR];
1237 md->u.md2.dir = nla_get_u8(nla);
1238 }
1239 if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID]) {
1240 nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID];
1241 set_hwid(&md->u.md2, nla_get_u8(nla));
1242 }
1243 } else {
1244 NL_SET_ERR_MSG(extack, "Tunnel key erspan option ver is incorrect");
1245 return -EINVAL;
1246 }
1247
1248 return sizeof(*md);
1249}
1250
1251static int fl_set_enc_opt(struct nlattr **tb, struct fl_flow_key *key,
1252 struct fl_flow_key *mask,
1253 struct netlink_ext_ack *extack)
1254{
1255 const struct nlattr *nla_enc_key, *nla_opt_key, *nla_opt_msk = NULL;
1256 int err, option_len, key_depth, msk_depth = 0;
1257
1258 err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS],
1259 TCA_FLOWER_KEY_ENC_OPTS_MAX,
1260 enc_opts_policy, extack);
1261 if (err)
1262 return err;
1263
1264 nla_enc_key = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS]);
1265
1266 if (tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]) {
1267 err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK],
1268 TCA_FLOWER_KEY_ENC_OPTS_MAX,
1269 enc_opts_policy, extack);
1270 if (err)
1271 return err;
1272
1273 nla_opt_msk = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
1274 msk_depth = nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
1275 }
1276
1277 nla_for_each_attr(nla_opt_key, nla_enc_key,
1278 nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS]), key_depth) {
1279 switch (nla_type(nla_opt_key)) {
1280 case TCA_FLOWER_KEY_ENC_OPTS_GENEVE:
1281 if (key->enc_opts.dst_opt_type &&
1282 key->enc_opts.dst_opt_type != TUNNEL_GENEVE_OPT) {
1283 NL_SET_ERR_MSG(extack, "Duplicate type for geneve options");
1284 return -EINVAL;
1285 }
1286 option_len = 0;
1287 key->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT;
1288 option_len = fl_set_geneve_opt(nla_opt_key, key,
1289 key_depth, option_len,
1290 extack);
1291 if (option_len < 0)
1292 return option_len;
1293
1294 key->enc_opts.len += option_len;
1295 /* At the same time we need to parse through the mask
1296 * in order to verify exact and mask attribute lengths.
1297 */
1298 mask->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT;
1299 option_len = fl_set_geneve_opt(nla_opt_msk, mask,
1300 msk_depth, option_len,
1301 extack);
1302 if (option_len < 0)
1303 return option_len;
1304
1305 mask->enc_opts.len += option_len;
1306 if (key->enc_opts.len != mask->enc_opts.len) {
1307 NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1308 return -EINVAL;
1309 }
1310
1311 if (msk_depth)
1312 nla_opt_msk = nla_next(nla_opt_msk, &msk_depth);
1313 break;
1314 case TCA_FLOWER_KEY_ENC_OPTS_VXLAN:
1315 if (key->enc_opts.dst_opt_type) {
1316 NL_SET_ERR_MSG(extack, "Duplicate type for vxlan options");
1317 return -EINVAL;
1318 }
1319 option_len = 0;
1320 key->enc_opts.dst_opt_type = TUNNEL_VXLAN_OPT;
1321 option_len = fl_set_vxlan_opt(nla_opt_key, key,
1322 key_depth, option_len,
1323 extack);
1324 if (option_len < 0)
1325 return option_len;
1326
1327 key->enc_opts.len += option_len;
1328 /* At the same time we need to parse through the mask
1329 * in order to verify exact and mask attribute lengths.
1330 */
1331 mask->enc_opts.dst_opt_type = TUNNEL_VXLAN_OPT;
1332 option_len = fl_set_vxlan_opt(nla_opt_msk, mask,
1333 msk_depth, option_len,
1334 extack);
1335 if (option_len < 0)
1336 return option_len;
1337
1338 mask->enc_opts.len += option_len;
1339 if (key->enc_opts.len != mask->enc_opts.len) {
1340 NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1341 return -EINVAL;
1342 }
1343
1344 if (msk_depth)
1345 nla_opt_msk = nla_next(nla_opt_msk, &msk_depth);
1346 break;
1347 case TCA_FLOWER_KEY_ENC_OPTS_ERSPAN:
1348 if (key->enc_opts.dst_opt_type) {
1349 NL_SET_ERR_MSG(extack, "Duplicate type for erspan options");
1350 return -EINVAL;
1351 }
1352 option_len = 0;
1353 key->enc_opts.dst_opt_type = TUNNEL_ERSPAN_OPT;
1354 option_len = fl_set_erspan_opt(nla_opt_key, key,
1355 key_depth, option_len,
1356 extack);
1357 if (option_len < 0)
1358 return option_len;
1359
1360 key->enc_opts.len += option_len;
1361 /* At the same time we need to parse through the mask
1362 * in order to verify exact and mask attribute lengths.
1363 */
1364 mask->enc_opts.dst_opt_type = TUNNEL_ERSPAN_OPT;
1365 option_len = fl_set_erspan_opt(nla_opt_msk, mask,
1366 msk_depth, option_len,
1367 extack);
1368 if (option_len < 0)
1369 return option_len;
1370
1371 mask->enc_opts.len += option_len;
1372 if (key->enc_opts.len != mask->enc_opts.len) {
1373 NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1374 return -EINVAL;
1375 }
1376
1377 if (msk_depth)
1378 nla_opt_msk = nla_next(nla_opt_msk, &msk_depth);
1379 break;
1380 default:
1381 NL_SET_ERR_MSG(extack, "Unknown tunnel option type");
1382 return -EINVAL;
1383 }
1384 }
1385
1386 return 0;
1387}
1388
1389static int fl_set_key_ct(struct nlattr **tb,
1390 struct flow_dissector_key_ct *key,
1391 struct flow_dissector_key_ct *mask,
1392 struct netlink_ext_ack *extack)
1393{
1394 if (tb[TCA_FLOWER_KEY_CT_STATE]) {
1395 if (!IS_ENABLED(CONFIG_NF_CONNTRACK)) {
1396 NL_SET_ERR_MSG(extack, "Conntrack isn't enabled");
1397 return -EOPNOTSUPP;
1398 }
1399 fl_set_key_val(tb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE,
1400 &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK,
1401 sizeof(key->ct_state));
1402 }
1403 if (tb[TCA_FLOWER_KEY_CT_ZONE]) {
1404 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)) {
1405 NL_SET_ERR_MSG(extack, "Conntrack zones isn't enabled");
1406 return -EOPNOTSUPP;
1407 }
1408 fl_set_key_val(tb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE,
1409 &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK,
1410 sizeof(key->ct_zone));
1411 }
1412 if (tb[TCA_FLOWER_KEY_CT_MARK]) {
1413 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)) {
1414 NL_SET_ERR_MSG(extack, "Conntrack mark isn't enabled");
1415 return -EOPNOTSUPP;
1416 }
1417 fl_set_key_val(tb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK,
1418 &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK,
1419 sizeof(key->ct_mark));
1420 }
1421 if (tb[TCA_FLOWER_KEY_CT_LABELS]) {
1422 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)) {
1423 NL_SET_ERR_MSG(extack, "Conntrack labels aren't enabled");
1424 return -EOPNOTSUPP;
1425 }
1426 fl_set_key_val(tb, key->ct_labels, TCA_FLOWER_KEY_CT_LABELS,
1427 mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK,
1428 sizeof(key->ct_labels));
1429 }
1430
1431 return 0;
1432}
1433
1434static int fl_set_key(struct net *net, struct nlattr **tb,
1435 struct fl_flow_key *key, struct fl_flow_key *mask,
1436 struct netlink_ext_ack *extack)
1437{
1438 __be16 ethertype;
1439 int ret = 0;
1440
1441 if (tb[TCA_FLOWER_INDEV]) {
1442 int err = tcf_change_indev(net, tb[TCA_FLOWER_INDEV], extack);
1443 if (err < 0)
1444 return err;
1445 key->meta.ingress_ifindex = err;
1446 mask->meta.ingress_ifindex = 0xffffffff;
1447 }
1448
1449 fl_set_key_val(tb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
1450 mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
1451 sizeof(key->eth.dst));
1452 fl_set_key_val(tb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
1453 mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
1454 sizeof(key->eth.src));
1455
1456 if (tb[TCA_FLOWER_KEY_ETH_TYPE]) {
1457 ethertype = nla_get_be16(tb[TCA_FLOWER_KEY_ETH_TYPE]);
1458
1459 if (eth_type_vlan(ethertype)) {
1460 fl_set_key_vlan(tb, ethertype, TCA_FLOWER_KEY_VLAN_ID,
1461 TCA_FLOWER_KEY_VLAN_PRIO, &key->vlan,
1462 &mask->vlan);
1463
1464 if (tb[TCA_FLOWER_KEY_VLAN_ETH_TYPE]) {
1465 ethertype = nla_get_be16(tb[TCA_FLOWER_KEY_VLAN_ETH_TYPE]);
1466 if (eth_type_vlan(ethertype)) {
1467 fl_set_key_vlan(tb, ethertype,
1468 TCA_FLOWER_KEY_CVLAN_ID,
1469 TCA_FLOWER_KEY_CVLAN_PRIO,
1470 &key->cvlan, &mask->cvlan);
1471 fl_set_key_val(tb, &key->basic.n_proto,
1472 TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
1473 &mask->basic.n_proto,
1474 TCA_FLOWER_UNSPEC,
1475 sizeof(key->basic.n_proto));
1476 } else {
1477 key->basic.n_proto = ethertype;
1478 mask->basic.n_proto = cpu_to_be16(~0);
1479 }
1480 }
1481 } else {
1482 key->basic.n_proto = ethertype;
1483 mask->basic.n_proto = cpu_to_be16(~0);
1484 }
1485 }
1486
1487 if (key->basic.n_proto == htons(ETH_P_IP) ||
1488 key->basic.n_proto == htons(ETH_P_IPV6)) {
1489 fl_set_key_val(tb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
1490 &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
1491 sizeof(key->basic.ip_proto));
1492 fl_set_key_ip(tb, false, &key->ip, &mask->ip);
1493 }
1494
1495 if (tb[TCA_FLOWER_KEY_IPV4_SRC] || tb[TCA_FLOWER_KEY_IPV4_DST]) {
1496 key->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1497 mask->control.addr_type = ~0;
1498 fl_set_key_val(tb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
1499 &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
1500 sizeof(key->ipv4.src));
1501 fl_set_key_val(tb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
1502 &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
1503 sizeof(key->ipv4.dst));
1504 } else if (tb[TCA_FLOWER_KEY_IPV6_SRC] || tb[TCA_FLOWER_KEY_IPV6_DST]) {
1505 key->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1506 mask->control.addr_type = ~0;
1507 fl_set_key_val(tb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
1508 &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
1509 sizeof(key->ipv6.src));
1510 fl_set_key_val(tb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
1511 &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
1512 sizeof(key->ipv6.dst));
1513 }
1514
1515 if (key->basic.ip_proto == IPPROTO_TCP) {
1516 fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
1517 &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
1518 sizeof(key->tp.src));
1519 fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
1520 &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
1521 sizeof(key->tp.dst));
1522 fl_set_key_val(tb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
1523 &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
1524 sizeof(key->tcp.flags));
1525 } else if (key->basic.ip_proto == IPPROTO_UDP) {
1526 fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
1527 &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
1528 sizeof(key->tp.src));
1529 fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
1530 &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
1531 sizeof(key->tp.dst));
1532 } else if (key->basic.ip_proto == IPPROTO_SCTP) {
1533 fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
1534 &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
1535 sizeof(key->tp.src));
1536 fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
1537 &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
1538 sizeof(key->tp.dst));
1539 } else if (key->basic.n_proto == htons(ETH_P_IP) &&
1540 key->basic.ip_proto == IPPROTO_ICMP) {
1541 fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV4_TYPE,
1542 &mask->icmp.type,
1543 TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
1544 sizeof(key->icmp.type));
1545 fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV4_CODE,
1546 &mask->icmp.code,
1547 TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
1548 sizeof(key->icmp.code));
1549 } else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
1550 key->basic.ip_proto == IPPROTO_ICMPV6) {
1551 fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV6_TYPE,
1552 &mask->icmp.type,
1553 TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
1554 sizeof(key->icmp.type));
1555 fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV6_CODE,
1556 &mask->icmp.code,
1557 TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
1558 sizeof(key->icmp.code));
1559 } else if (key->basic.n_proto == htons(ETH_P_MPLS_UC) ||
1560 key->basic.n_proto == htons(ETH_P_MPLS_MC)) {
1561 ret = fl_set_key_mpls(tb, &key->mpls, &mask->mpls, extack);
1562 if (ret)
1563 return ret;
1564 } else if (key->basic.n_proto == htons(ETH_P_ARP) ||
1565 key->basic.n_proto == htons(ETH_P_RARP)) {
1566 fl_set_key_val(tb, &key->arp.sip, TCA_FLOWER_KEY_ARP_SIP,
1567 &mask->arp.sip, TCA_FLOWER_KEY_ARP_SIP_MASK,
1568 sizeof(key->arp.sip));
1569 fl_set_key_val(tb, &key->arp.tip, TCA_FLOWER_KEY_ARP_TIP,
1570 &mask->arp.tip, TCA_FLOWER_KEY_ARP_TIP_MASK,
1571 sizeof(key->arp.tip));
1572 fl_set_key_val(tb, &key->arp.op, TCA_FLOWER_KEY_ARP_OP,
1573 &mask->arp.op, TCA_FLOWER_KEY_ARP_OP_MASK,
1574 sizeof(key->arp.op));
1575 fl_set_key_val(tb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
1576 mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
1577 sizeof(key->arp.sha));
1578 fl_set_key_val(tb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
1579 mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
1580 sizeof(key->arp.tha));
1581 }
1582
1583 if (key->basic.ip_proto == IPPROTO_TCP ||
1584 key->basic.ip_proto == IPPROTO_UDP ||
1585 key->basic.ip_proto == IPPROTO_SCTP) {
1586 ret = fl_set_key_port_range(tb, key, mask, extack);
1587 if (ret)
1588 return ret;
1589 }
1590
1591 if (tb[TCA_FLOWER_KEY_ENC_IPV4_SRC] ||
1592 tb[TCA_FLOWER_KEY_ENC_IPV4_DST]) {
1593 key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1594 mask->enc_control.addr_type = ~0;
1595 fl_set_key_val(tb, &key->enc_ipv4.src,
1596 TCA_FLOWER_KEY_ENC_IPV4_SRC,
1597 &mask->enc_ipv4.src,
1598 TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
1599 sizeof(key->enc_ipv4.src));
1600 fl_set_key_val(tb, &key->enc_ipv4.dst,
1601 TCA_FLOWER_KEY_ENC_IPV4_DST,
1602 &mask->enc_ipv4.dst,
1603 TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
1604 sizeof(key->enc_ipv4.dst));
1605 }
1606
1607 if (tb[TCA_FLOWER_KEY_ENC_IPV6_SRC] ||
1608 tb[TCA_FLOWER_KEY_ENC_IPV6_DST]) {
1609 key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1610 mask->enc_control.addr_type = ~0;
1611 fl_set_key_val(tb, &key->enc_ipv6.src,
1612 TCA_FLOWER_KEY_ENC_IPV6_SRC,
1613 &mask->enc_ipv6.src,
1614 TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
1615 sizeof(key->enc_ipv6.src));
1616 fl_set_key_val(tb, &key->enc_ipv6.dst,
1617 TCA_FLOWER_KEY_ENC_IPV6_DST,
1618 &mask->enc_ipv6.dst,
1619 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
1620 sizeof(key->enc_ipv6.dst));
1621 }
1622
1623 fl_set_key_val(tb, &key->enc_key_id.keyid, TCA_FLOWER_KEY_ENC_KEY_ID,
1624 &mask->enc_key_id.keyid, TCA_FLOWER_UNSPEC,
1625 sizeof(key->enc_key_id.keyid));
1626
1627 fl_set_key_val(tb, &key->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
1628 &mask->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
1629 sizeof(key->enc_tp.src));
1630
1631 fl_set_key_val(tb, &key->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
1632 &mask->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
1633 sizeof(key->enc_tp.dst));
1634
1635 fl_set_key_ip(tb, true, &key->enc_ip, &mask->enc_ip);
1636
1637 fl_set_key_val(tb, &key->hash.hash, TCA_FLOWER_KEY_HASH,
1638 &mask->hash.hash, TCA_FLOWER_KEY_HASH_MASK,
1639 sizeof(key->hash.hash));
1640
1641 if (tb[TCA_FLOWER_KEY_ENC_OPTS]) {
1642 ret = fl_set_enc_opt(tb, key, mask, extack);
1643 if (ret)
1644 return ret;
1645 }
1646
1647 ret = fl_set_key_ct(tb, &key->ct, &mask->ct, extack);
1648 if (ret)
1649 return ret;
1650
1651 if (tb[TCA_FLOWER_KEY_FLAGS])
1652 ret = fl_set_key_flags(tb, &key->control.flags,
1653 &mask->control.flags, extack);
1654
1655 return ret;
1656}
1657
1658static void fl_mask_copy(struct fl_flow_mask *dst,
1659 struct fl_flow_mask *src)
1660{
1661 const void *psrc = fl_key_get_start(&src->key, src);
1662 void *pdst = fl_key_get_start(&dst->key, src);
1663
1664 memcpy(pdst, psrc, fl_mask_range(src));
1665 dst->range = src->range;
1666}
1667
1668static const struct rhashtable_params fl_ht_params = {
1669 .key_offset = offsetof(struct cls_fl_filter, mkey), /* base offset */
1670 .head_offset = offsetof(struct cls_fl_filter, ht_node),
1671 .automatic_shrinking = true,
1672};
1673
1674static int fl_init_mask_hashtable(struct fl_flow_mask *mask)
1675{
1676 mask->filter_ht_params = fl_ht_params;
1677 mask->filter_ht_params.key_len = fl_mask_range(mask);
1678 mask->filter_ht_params.key_offset += mask->range.start;
1679
1680 return rhashtable_init(&mask->ht, &mask->filter_ht_params);
1681}
1682
1683#define FL_KEY_MEMBER_OFFSET(member) offsetof(struct fl_flow_key, member)
1684#define FL_KEY_MEMBER_SIZE(member) sizeof_field(struct fl_flow_key, member)
1685
1686#define FL_KEY_IS_MASKED(mask, member) \
1687 memchr_inv(((char *)mask) + FL_KEY_MEMBER_OFFSET(member), \
1688 0, FL_KEY_MEMBER_SIZE(member)) \
1689
1690#define FL_KEY_SET(keys, cnt, id, member) \
1691 do { \
1692 keys[cnt].key_id = id; \
1693 keys[cnt].offset = FL_KEY_MEMBER_OFFSET(member); \
1694 cnt++; \
1695 } while(0);
1696
1697#define FL_KEY_SET_IF_MASKED(mask, keys, cnt, id, member) \
1698 do { \
1699 if (FL_KEY_IS_MASKED(mask, member)) \
1700 FL_KEY_SET(keys, cnt, id, member); \
1701 } while(0);
1702
1703static void fl_init_dissector(struct flow_dissector *dissector,
1704 struct fl_flow_key *mask)
1705{
1706 struct flow_dissector_key keys[FLOW_DISSECTOR_KEY_MAX];
1707 size_t cnt = 0;
1708
1709 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1710 FLOW_DISSECTOR_KEY_META, meta);
1711 FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_CONTROL, control);
1712 FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_BASIC, basic);
1713 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1714 FLOW_DISSECTOR_KEY_ETH_ADDRS, eth);
1715 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1716 FLOW_DISSECTOR_KEY_IPV4_ADDRS, ipv4);
1717 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1718 FLOW_DISSECTOR_KEY_IPV6_ADDRS, ipv6);
1719 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1720 FLOW_DISSECTOR_KEY_PORTS, tp);
1721 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1722 FLOW_DISSECTOR_KEY_PORTS_RANGE, tp_range);
1723 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1724 FLOW_DISSECTOR_KEY_IP, ip);
1725 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1726 FLOW_DISSECTOR_KEY_TCP, tcp);
1727 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1728 FLOW_DISSECTOR_KEY_ICMP, icmp);
1729 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1730 FLOW_DISSECTOR_KEY_ARP, arp);
1731 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1732 FLOW_DISSECTOR_KEY_MPLS, mpls);
1733 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1734 FLOW_DISSECTOR_KEY_VLAN, vlan);
1735 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1736 FLOW_DISSECTOR_KEY_CVLAN, cvlan);
1737 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1738 FLOW_DISSECTOR_KEY_ENC_KEYID, enc_key_id);
1739 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1740 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, enc_ipv4);
1741 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1742 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, enc_ipv6);
1743 if (FL_KEY_IS_MASKED(mask, enc_ipv4) ||
1744 FL_KEY_IS_MASKED(mask, enc_ipv6))
1745 FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_ENC_CONTROL,
1746 enc_control);
1747 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1748 FLOW_DISSECTOR_KEY_ENC_PORTS, enc_tp);
1749 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1750 FLOW_DISSECTOR_KEY_ENC_IP, enc_ip);
1751 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1752 FLOW_DISSECTOR_KEY_ENC_OPTS, enc_opts);
1753 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1754 FLOW_DISSECTOR_KEY_CT, ct);
1755 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1756 FLOW_DISSECTOR_KEY_HASH, hash);
1757
1758 skb_flow_dissector_init(dissector, keys, cnt);
1759}
1760
1761static struct fl_flow_mask *fl_create_new_mask(struct cls_fl_head *head,
1762 struct fl_flow_mask *mask)
1763{
1764 struct fl_flow_mask *newmask;
1765 int err;
1766
1767 newmask = kzalloc(sizeof(*newmask), GFP_KERNEL);
1768 if (!newmask)
1769 return ERR_PTR(-ENOMEM);
1770
1771 fl_mask_copy(newmask, mask);
1772
1773 if ((newmask->key.tp_range.tp_min.dst &&
1774 newmask->key.tp_range.tp_max.dst) ||
1775 (newmask->key.tp_range.tp_min.src &&
1776 newmask->key.tp_range.tp_max.src))
1777 newmask->flags |= TCA_FLOWER_MASK_FLAGS_RANGE;
1778
1779 err = fl_init_mask_hashtable(newmask);
1780 if (err)
1781 goto errout_free;
1782
1783 fl_init_dissector(&newmask->dissector, &newmask->key);
1784
1785 INIT_LIST_HEAD_RCU(&newmask->filters);
1786
1787 refcount_set(&newmask->refcnt, 1);
1788 err = rhashtable_replace_fast(&head->ht, &mask->ht_node,
1789 &newmask->ht_node, mask_ht_params);
1790 if (err)
1791 goto errout_destroy;
1792
1793 spin_lock(&head->masks_lock);
1794 list_add_tail_rcu(&newmask->list, &head->masks);
1795 spin_unlock(&head->masks_lock);
1796
1797 return newmask;
1798
1799errout_destroy:
1800 rhashtable_destroy(&newmask->ht);
1801errout_free:
1802 kfree(newmask);
1803
1804 return ERR_PTR(err);
1805}
1806
1807static int fl_check_assign_mask(struct cls_fl_head *head,
1808 struct cls_fl_filter *fnew,
1809 struct cls_fl_filter *fold,
1810 struct fl_flow_mask *mask)
1811{
1812 struct fl_flow_mask *newmask;
1813 int ret = 0;
1814
1815 rcu_read_lock();
1816
1817 /* Insert mask as temporary node to prevent concurrent creation of mask
1818 * with same key. Any concurrent lookups with same key will return
1819 * -EAGAIN because mask's refcnt is zero.
1820 */
1821 fnew->mask = rhashtable_lookup_get_insert_fast(&head->ht,
1822 &mask->ht_node,
1823 mask_ht_params);
1824 if (!fnew->mask) {
1825 rcu_read_unlock();
1826
1827 if (fold) {
1828 ret = -EINVAL;
1829 goto errout_cleanup;
1830 }
1831
1832 newmask = fl_create_new_mask(head, mask);
1833 if (IS_ERR(newmask)) {
1834 ret = PTR_ERR(newmask);
1835 goto errout_cleanup;
1836 }
1837
1838 fnew->mask = newmask;
1839 return 0;
1840 } else if (IS_ERR(fnew->mask)) {
1841 ret = PTR_ERR(fnew->mask);
1842 } else if (fold && fold->mask != fnew->mask) {
1843 ret = -EINVAL;
1844 } else if (!refcount_inc_not_zero(&fnew->mask->refcnt)) {
1845 /* Mask was deleted concurrently, try again */
1846 ret = -EAGAIN;
1847 }
1848 rcu_read_unlock();
1849 return ret;
1850
1851errout_cleanup:
1852 rhashtable_remove_fast(&head->ht, &mask->ht_node,
1853 mask_ht_params);
1854 return ret;
1855}
1856
1857static int fl_set_parms(struct net *net, struct tcf_proto *tp,
1858 struct cls_fl_filter *f, struct fl_flow_mask *mask,
1859 unsigned long base, struct nlattr **tb,
1860 struct nlattr *est, bool ovr,
1861 struct fl_flow_tmplt *tmplt, bool rtnl_held,
1862 struct netlink_ext_ack *extack)
1863{
1864 int err;
1865
1866 err = tcf_exts_validate(net, tp, tb, est, &f->exts, ovr, rtnl_held,
1867 extack);
1868 if (err < 0)
1869 return err;
1870
1871 if (tb[TCA_FLOWER_CLASSID]) {
1872 f->res.classid = nla_get_u32(tb[TCA_FLOWER_CLASSID]);
1873 if (!rtnl_held)
1874 rtnl_lock();
1875 tcf_bind_filter(tp, &f->res, base);
1876 if (!rtnl_held)
1877 rtnl_unlock();
1878 }
1879
1880 err = fl_set_key(net, tb, &f->key, &mask->key, extack);
1881 if (err)
1882 return err;
1883
1884 fl_mask_update_range(mask);
1885 fl_set_masked_key(&f->mkey, &f->key, mask);
1886
1887 if (!fl_mask_fits_tmplt(tmplt, mask)) {
1888 NL_SET_ERR_MSG_MOD(extack, "Mask does not fit the template");
1889 return -EINVAL;
1890 }
1891
1892 return 0;
1893}
1894
1895static int fl_ht_insert_unique(struct cls_fl_filter *fnew,
1896 struct cls_fl_filter *fold,
1897 bool *in_ht)
1898{
1899 struct fl_flow_mask *mask = fnew->mask;
1900 int err;
1901
1902 err = rhashtable_lookup_insert_fast(&mask->ht,
1903 &fnew->ht_node,
1904 mask->filter_ht_params);
1905 if (err) {
1906 *in_ht = false;
1907 /* It is okay if filter with same key exists when
1908 * overwriting.
1909 */
1910 return fold && err == -EEXIST ? 0 : err;
1911 }
1912
1913 *in_ht = true;
1914 return 0;
1915}
1916
1917static int fl_change(struct net *net, struct sk_buff *in_skb,
1918 struct tcf_proto *tp, unsigned long base,
1919 u32 handle, struct nlattr **tca,
1920 void **arg, bool ovr, bool rtnl_held,
1921 struct netlink_ext_ack *extack)
1922{
1923 struct cls_fl_head *head = fl_head_dereference(tp);
1924 struct cls_fl_filter *fold = *arg;
1925 struct cls_fl_filter *fnew;
1926 struct fl_flow_mask *mask;
1927 struct nlattr **tb;
1928 bool in_ht;
1929 int err;
1930
1931 if (!tca[TCA_OPTIONS]) {
1932 err = -EINVAL;
1933 goto errout_fold;
1934 }
1935
1936 mask = kzalloc(sizeof(struct fl_flow_mask), GFP_KERNEL);
1937 if (!mask) {
1938 err = -ENOBUFS;
1939 goto errout_fold;
1940 }
1941
1942 tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
1943 if (!tb) {
1944 err = -ENOBUFS;
1945 goto errout_mask_alloc;
1946 }
1947
1948 err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
1949 tca[TCA_OPTIONS], fl_policy, NULL);
1950 if (err < 0)
1951 goto errout_tb;
1952
1953 if (fold && handle && fold->handle != handle) {
1954 err = -EINVAL;
1955 goto errout_tb;
1956 }
1957
1958 fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
1959 if (!fnew) {
1960 err = -ENOBUFS;
1961 goto errout_tb;
1962 }
1963 INIT_LIST_HEAD(&fnew->hw_list);
1964 refcount_set(&fnew->refcnt, 1);
1965
1966 err = tcf_exts_init(&fnew->exts, net, TCA_FLOWER_ACT, 0);
1967 if (err < 0)
1968 goto errout;
1969
1970 if (tb[TCA_FLOWER_FLAGS]) {
1971 fnew->flags = nla_get_u32(tb[TCA_FLOWER_FLAGS]);
1972
1973 if (!tc_flags_valid(fnew->flags)) {
1974 err = -EINVAL;
1975 goto errout;
1976 }
1977 }
1978
1979 err = fl_set_parms(net, tp, fnew, mask, base, tb, tca[TCA_RATE], ovr,
1980 tp->chain->tmplt_priv, rtnl_held, extack);
1981 if (err)
1982 goto errout;
1983
1984 err = fl_check_assign_mask(head, fnew, fold, mask);
1985 if (err)
1986 goto errout;
1987
1988 err = fl_ht_insert_unique(fnew, fold, &in_ht);
1989 if (err)
1990 goto errout_mask;
1991
1992 if (!tc_skip_hw(fnew->flags)) {
1993 err = fl_hw_replace_filter(tp, fnew, rtnl_held, extack);
1994 if (err)
1995 goto errout_ht;
1996 }
1997
1998 if (!tc_in_hw(fnew->flags))
1999 fnew->flags |= TCA_CLS_FLAGS_NOT_IN_HW;
2000
2001 spin_lock(&tp->lock);
2002
2003 /* tp was deleted concurrently. -EAGAIN will cause caller to lookup
2004 * proto again or create new one, if necessary.
2005 */
2006 if (tp->deleting) {
2007 err = -EAGAIN;
2008 goto errout_hw;
2009 }
2010
2011 if (fold) {
2012 /* Fold filter was deleted concurrently. Retry lookup. */
2013 if (fold->deleted) {
2014 err = -EAGAIN;
2015 goto errout_hw;
2016 }
2017
2018 fnew->handle = handle;
2019
2020 if (!in_ht) {
2021 struct rhashtable_params params =
2022 fnew->mask->filter_ht_params;
2023
2024 err = rhashtable_insert_fast(&fnew->mask->ht,
2025 &fnew->ht_node,
2026 params);
2027 if (err)
2028 goto errout_hw;
2029 in_ht = true;
2030 }
2031
2032 refcount_inc(&fnew->refcnt);
2033 rhashtable_remove_fast(&fold->mask->ht,
2034 &fold->ht_node,
2035 fold->mask->filter_ht_params);
2036 idr_replace(&head->handle_idr, fnew, fnew->handle);
2037 list_replace_rcu(&fold->list, &fnew->list);
2038 fold->deleted = true;
2039
2040 spin_unlock(&tp->lock);
2041
2042 fl_mask_put(head, fold->mask);
2043 if (!tc_skip_hw(fold->flags))
2044 fl_hw_destroy_filter(tp, fold, rtnl_held, NULL);
2045 tcf_unbind_filter(tp, &fold->res);
2046 /* Caller holds reference to fold, so refcnt is always > 0
2047 * after this.
2048 */
2049 refcount_dec(&fold->refcnt);
2050 __fl_put(fold);
2051 } else {
2052 if (handle) {
2053 /* user specifies a handle and it doesn't exist */
2054 err = idr_alloc_u32(&head->handle_idr, fnew, &handle,
2055 handle, GFP_ATOMIC);
2056
2057 /* Filter with specified handle was concurrently
2058 * inserted after initial check in cls_api. This is not
2059 * necessarily an error if NLM_F_EXCL is not set in
2060 * message flags. Returning EAGAIN will cause cls_api to
2061 * try to update concurrently inserted rule.
2062 */
2063 if (err == -ENOSPC)
2064 err = -EAGAIN;
2065 } else {
2066 handle = 1;
2067 err = idr_alloc_u32(&head->handle_idr, fnew, &handle,
2068 INT_MAX, GFP_ATOMIC);
2069 }
2070 if (err)
2071 goto errout_hw;
2072
2073 refcount_inc(&fnew->refcnt);
2074 fnew->handle = handle;
2075 list_add_tail_rcu(&fnew->list, &fnew->mask->filters);
2076 spin_unlock(&tp->lock);
2077 }
2078
2079 *arg = fnew;
2080
2081 kfree(tb);
2082 tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work);
2083 return 0;
2084
2085errout_ht:
2086 spin_lock(&tp->lock);
2087errout_hw:
2088 fnew->deleted = true;
2089 spin_unlock(&tp->lock);
2090 if (!tc_skip_hw(fnew->flags))
2091 fl_hw_destroy_filter(tp, fnew, rtnl_held, NULL);
2092 if (in_ht)
2093 rhashtable_remove_fast(&fnew->mask->ht, &fnew->ht_node,
2094 fnew->mask->filter_ht_params);
2095errout_mask:
2096 fl_mask_put(head, fnew->mask);
2097errout:
2098 __fl_put(fnew);
2099errout_tb:
2100 kfree(tb);
2101errout_mask_alloc:
2102 tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work);
2103errout_fold:
2104 if (fold)
2105 __fl_put(fold);
2106 return err;
2107}
2108
2109static int fl_delete(struct tcf_proto *tp, void *arg, bool *last,
2110 bool rtnl_held, struct netlink_ext_ack *extack)
2111{
2112 struct cls_fl_head *head = fl_head_dereference(tp);
2113 struct cls_fl_filter *f = arg;
2114 bool last_on_mask;
2115 int err = 0;
2116
2117 err = __fl_delete(tp, f, &last_on_mask, rtnl_held, extack);
2118 *last = list_empty(&head->masks);
2119 __fl_put(f);
2120
2121 return err;
2122}
2123
2124static void fl_walk(struct tcf_proto *tp, struct tcf_walker *arg,
2125 bool rtnl_held)
2126{
2127 struct cls_fl_head *head = fl_head_dereference(tp);
2128 unsigned long id = arg->cookie, tmp;
2129 struct cls_fl_filter *f;
2130
2131 arg->count = arg->skip;
2132
2133 idr_for_each_entry_continue_ul(&head->handle_idr, f, tmp, id) {
2134 /* don't return filters that are being deleted */
2135 if (!refcount_inc_not_zero(&f->refcnt))
2136 continue;
2137 if (arg->fn(tp, f, arg) < 0) {
2138 __fl_put(f);
2139 arg->stop = 1;
2140 break;
2141 }
2142 __fl_put(f);
2143 arg->count++;
2144 }
2145 arg->cookie = id;
2146}
2147
2148static struct cls_fl_filter *
2149fl_get_next_hw_filter(struct tcf_proto *tp, struct cls_fl_filter *f, bool add)
2150{
2151 struct cls_fl_head *head = fl_head_dereference(tp);
2152
2153 spin_lock(&tp->lock);
2154 if (list_empty(&head->hw_filters)) {
2155 spin_unlock(&tp->lock);
2156 return NULL;
2157 }
2158
2159 if (!f)
2160 f = list_entry(&head->hw_filters, struct cls_fl_filter,
2161 hw_list);
2162 list_for_each_entry_continue(f, &head->hw_filters, hw_list) {
2163 if (!(add && f->deleted) && refcount_inc_not_zero(&f->refcnt)) {
2164 spin_unlock(&tp->lock);
2165 return f;
2166 }
2167 }
2168
2169 spin_unlock(&tp->lock);
2170 return NULL;
2171}
2172
2173static int fl_reoffload(struct tcf_proto *tp, bool add, flow_setup_cb_t *cb,
2174 void *cb_priv, struct netlink_ext_ack *extack)
2175{
2176 struct tcf_block *block = tp->chain->block;
2177 struct flow_cls_offload cls_flower = {};
2178 struct cls_fl_filter *f = NULL;
2179 int err;
2180
2181 /* hw_filters list can only be changed by hw offload functions after
2182 * obtaining rtnl lock. Make sure it is not changed while reoffload is
2183 * iterating it.
2184 */
2185 ASSERT_RTNL();
2186
2187 while ((f = fl_get_next_hw_filter(tp, f, add))) {
2188 cls_flower.rule =
2189 flow_rule_alloc(tcf_exts_num_actions(&f->exts));
2190 if (!cls_flower.rule) {
2191 __fl_put(f);
2192 return -ENOMEM;
2193 }
2194
2195 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags,
2196 extack);
2197 cls_flower.command = add ?
2198 FLOW_CLS_REPLACE : FLOW_CLS_DESTROY;
2199 cls_flower.cookie = (unsigned long)f;
2200 cls_flower.rule->match.dissector = &f->mask->dissector;
2201 cls_flower.rule->match.mask = &f->mask->key;
2202 cls_flower.rule->match.key = &f->mkey;
2203
2204 err = tc_setup_flow_action(&cls_flower.rule->action, &f->exts);
2205 if (err) {
2206 kfree(cls_flower.rule);
2207 if (tc_skip_sw(f->flags)) {
2208 NL_SET_ERR_MSG_MOD(extack, "Failed to setup flow action");
2209 __fl_put(f);
2210 return err;
2211 }
2212 goto next_flow;
2213 }
2214
2215 cls_flower.classid = f->res.classid;
2216
2217 err = tc_setup_cb_reoffload(block, tp, add, cb,
2218 TC_SETUP_CLSFLOWER, &cls_flower,
2219 cb_priv, &f->flags,
2220 &f->in_hw_count);
2221 tc_cleanup_flow_action(&cls_flower.rule->action);
2222 kfree(cls_flower.rule);
2223
2224 if (err) {
2225 __fl_put(f);
2226 return err;
2227 }
2228next_flow:
2229 __fl_put(f);
2230 }
2231
2232 return 0;
2233}
2234
2235static void fl_hw_add(struct tcf_proto *tp, void *type_data)
2236{
2237 struct flow_cls_offload *cls_flower = type_data;
2238 struct cls_fl_filter *f =
2239 (struct cls_fl_filter *) cls_flower->cookie;
2240 struct cls_fl_head *head = fl_head_dereference(tp);
2241
2242 spin_lock(&tp->lock);
2243 list_add(&f->hw_list, &head->hw_filters);
2244 spin_unlock(&tp->lock);
2245}
2246
2247static void fl_hw_del(struct tcf_proto *tp, void *type_data)
2248{
2249 struct flow_cls_offload *cls_flower = type_data;
2250 struct cls_fl_filter *f =
2251 (struct cls_fl_filter *) cls_flower->cookie;
2252
2253 spin_lock(&tp->lock);
2254 if (!list_empty(&f->hw_list))
2255 list_del_init(&f->hw_list);
2256 spin_unlock(&tp->lock);
2257}
2258
2259static int fl_hw_create_tmplt(struct tcf_chain *chain,
2260 struct fl_flow_tmplt *tmplt)
2261{
2262 struct flow_cls_offload cls_flower = {};
2263 struct tcf_block *block = chain->block;
2264
2265 cls_flower.rule = flow_rule_alloc(0);
2266 if (!cls_flower.rule)
2267 return -ENOMEM;
2268
2269 cls_flower.common.chain_index = chain->index;
2270 cls_flower.command = FLOW_CLS_TMPLT_CREATE;
2271 cls_flower.cookie = (unsigned long) tmplt;
2272 cls_flower.rule->match.dissector = &tmplt->dissector;
2273 cls_flower.rule->match.mask = &tmplt->mask;
2274 cls_flower.rule->match.key = &tmplt->dummy_key;
2275
2276 /* We don't care if driver (any of them) fails to handle this
2277 * call. It serves just as a hint for it.
2278 */
2279 tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, true);
2280 kfree(cls_flower.rule);
2281
2282 return 0;
2283}
2284
2285static void fl_hw_destroy_tmplt(struct tcf_chain *chain,
2286 struct fl_flow_tmplt *tmplt)
2287{
2288 struct flow_cls_offload cls_flower = {};
2289 struct tcf_block *block = chain->block;
2290
2291 cls_flower.common.chain_index = chain->index;
2292 cls_flower.command = FLOW_CLS_TMPLT_DESTROY;
2293 cls_flower.cookie = (unsigned long) tmplt;
2294
2295 tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, true);
2296}
2297
2298static void *fl_tmplt_create(struct net *net, struct tcf_chain *chain,
2299 struct nlattr **tca,
2300 struct netlink_ext_ack *extack)
2301{
2302 struct fl_flow_tmplt *tmplt;
2303 struct nlattr **tb;
2304 int err;
2305
2306 if (!tca[TCA_OPTIONS])
2307 return ERR_PTR(-EINVAL);
2308
2309 tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
2310 if (!tb)
2311 return ERR_PTR(-ENOBUFS);
2312 err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
2313 tca[TCA_OPTIONS], fl_policy, NULL);
2314 if (err)
2315 goto errout_tb;
2316
2317 tmplt = kzalloc(sizeof(*tmplt), GFP_KERNEL);
2318 if (!tmplt) {
2319 err = -ENOMEM;
2320 goto errout_tb;
2321 }
2322 tmplt->chain = chain;
2323 err = fl_set_key(net, tb, &tmplt->dummy_key, &tmplt->mask, extack);
2324 if (err)
2325 goto errout_tmplt;
2326
2327 fl_init_dissector(&tmplt->dissector, &tmplt->mask);
2328
2329 err = fl_hw_create_tmplt(chain, tmplt);
2330 if (err)
2331 goto errout_tmplt;
2332
2333 kfree(tb);
2334 return tmplt;
2335
2336errout_tmplt:
2337 kfree(tmplt);
2338errout_tb:
2339 kfree(tb);
2340 return ERR_PTR(err);
2341}
2342
2343static void fl_tmplt_destroy(void *tmplt_priv)
2344{
2345 struct fl_flow_tmplt *tmplt = tmplt_priv;
2346
2347 fl_hw_destroy_tmplt(tmplt->chain, tmplt);
2348 kfree(tmplt);
2349}
2350
2351static int fl_dump_key_val(struct sk_buff *skb,
2352 void *val, int val_type,
2353 void *mask, int mask_type, int len)
2354{
2355 int err;
2356
2357 if (!memchr_inv(mask, 0, len))
2358 return 0;
2359 err = nla_put(skb, val_type, len, val);
2360 if (err)
2361 return err;
2362 if (mask_type != TCA_FLOWER_UNSPEC) {
2363 err = nla_put(skb, mask_type, len, mask);
2364 if (err)
2365 return err;
2366 }
2367 return 0;
2368}
2369
2370static int fl_dump_key_port_range(struct sk_buff *skb, struct fl_flow_key *key,
2371 struct fl_flow_key *mask)
2372{
2373 if (fl_dump_key_val(skb, &key->tp_range.tp_min.dst,
2374 TCA_FLOWER_KEY_PORT_DST_MIN,
2375 &mask->tp_range.tp_min.dst, TCA_FLOWER_UNSPEC,
2376 sizeof(key->tp_range.tp_min.dst)) ||
2377 fl_dump_key_val(skb, &key->tp_range.tp_max.dst,
2378 TCA_FLOWER_KEY_PORT_DST_MAX,
2379 &mask->tp_range.tp_max.dst, TCA_FLOWER_UNSPEC,
2380 sizeof(key->tp_range.tp_max.dst)) ||
2381 fl_dump_key_val(skb, &key->tp_range.tp_min.src,
2382 TCA_FLOWER_KEY_PORT_SRC_MIN,
2383 &mask->tp_range.tp_min.src, TCA_FLOWER_UNSPEC,
2384 sizeof(key->tp_range.tp_min.src)) ||
2385 fl_dump_key_val(skb, &key->tp_range.tp_max.src,
2386 TCA_FLOWER_KEY_PORT_SRC_MAX,
2387 &mask->tp_range.tp_max.src, TCA_FLOWER_UNSPEC,
2388 sizeof(key->tp_range.tp_max.src)))
2389 return -1;
2390
2391 return 0;
2392}
2393
2394static int fl_dump_key_mpls_opt_lse(struct sk_buff *skb,
2395 struct flow_dissector_key_mpls *mpls_key,
2396 struct flow_dissector_key_mpls *mpls_mask,
2397 u8 lse_index)
2398{
2399 struct flow_dissector_mpls_lse *lse_mask = &mpls_mask->ls[lse_index];
2400 struct flow_dissector_mpls_lse *lse_key = &mpls_key->ls[lse_index];
2401 int err;
2402
2403 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH,
2404 lse_index + 1);
2405 if (err)
2406 return err;
2407
2408 if (lse_mask->mpls_ttl) {
2409 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL,
2410 lse_key->mpls_ttl);
2411 if (err)
2412 return err;
2413 }
2414 if (lse_mask->mpls_bos) {
2415 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS,
2416 lse_key->mpls_bos);
2417 if (err)
2418 return err;
2419 }
2420 if (lse_mask->mpls_tc) {
2421 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_TC,
2422 lse_key->mpls_tc);
2423 if (err)
2424 return err;
2425 }
2426 if (lse_mask->mpls_label) {
2427 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL,
2428 lse_key->mpls_label);
2429 if (err)
2430 return err;
2431 }
2432
2433 return 0;
2434}
2435
2436static int fl_dump_key_mpls_opts(struct sk_buff *skb,
2437 struct flow_dissector_key_mpls *mpls_key,
2438 struct flow_dissector_key_mpls *mpls_mask)
2439{
2440 struct nlattr *opts;
2441 struct nlattr *lse;
2442 u8 lse_index;
2443 int err;
2444
2445 opts = nla_nest_start(skb, TCA_FLOWER_KEY_MPLS_OPTS);
2446 if (!opts)
2447 return -EMSGSIZE;
2448
2449 for (lse_index = 0; lse_index < FLOW_DIS_MPLS_MAX; lse_index++) {
2450 if (!(mpls_mask->used_lses & 1 << lse_index))
2451 continue;
2452
2453 lse = nla_nest_start(skb, TCA_FLOWER_KEY_MPLS_OPTS_LSE);
2454 if (!lse) {
2455 err = -EMSGSIZE;
2456 goto err_opts;
2457 }
2458
2459 err = fl_dump_key_mpls_opt_lse(skb, mpls_key, mpls_mask,
2460 lse_index);
2461 if (err)
2462 goto err_opts_lse;
2463 nla_nest_end(skb, lse);
2464 }
2465 nla_nest_end(skb, opts);
2466
2467 return 0;
2468
2469err_opts_lse:
2470 nla_nest_cancel(skb, lse);
2471err_opts:
2472 nla_nest_cancel(skb, opts);
2473
2474 return err;
2475}
2476
2477static int fl_dump_key_mpls(struct sk_buff *skb,
2478 struct flow_dissector_key_mpls *mpls_key,
2479 struct flow_dissector_key_mpls *mpls_mask)
2480{
2481 struct flow_dissector_mpls_lse *lse_mask;
2482 struct flow_dissector_mpls_lse *lse_key;
2483 int err;
2484
2485 if (!mpls_mask->used_lses)
2486 return 0;
2487
2488 lse_mask = &mpls_mask->ls[0];
2489 lse_key = &mpls_key->ls[0];
2490
2491 /* For backward compatibility, don't use the MPLS nested attributes if
2492 * the rule can be expressed using the old attributes.
2493 */
2494 if (mpls_mask->used_lses & ~1 ||
2495 (!lse_mask->mpls_ttl && !lse_mask->mpls_bos &&
2496 !lse_mask->mpls_tc && !lse_mask->mpls_label))
2497 return fl_dump_key_mpls_opts(skb, mpls_key, mpls_mask);
2498
2499 if (lse_mask->mpls_ttl) {
2500 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TTL,
2501 lse_key->mpls_ttl);
2502 if (err)
2503 return err;
2504 }
2505 if (lse_mask->mpls_tc) {
2506 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TC,
2507 lse_key->mpls_tc);
2508 if (err)
2509 return err;
2510 }
2511 if (lse_mask->mpls_label) {
2512 err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_LABEL,
2513 lse_key->mpls_label);
2514 if (err)
2515 return err;
2516 }
2517 if (lse_mask->mpls_bos) {
2518 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_BOS,
2519 lse_key->mpls_bos);
2520 if (err)
2521 return err;
2522 }
2523 return 0;
2524}
2525
2526static int fl_dump_key_ip(struct sk_buff *skb, bool encap,
2527 struct flow_dissector_key_ip *key,
2528 struct flow_dissector_key_ip *mask)
2529{
2530 int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
2531 int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
2532 int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
2533 int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
2534
2535 if (fl_dump_key_val(skb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos)) ||
2536 fl_dump_key_val(skb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl)))
2537 return -1;
2538
2539 return 0;
2540}
2541
2542static int fl_dump_key_vlan(struct sk_buff *skb,
2543 int vlan_id_key, int vlan_prio_key,
2544 struct flow_dissector_key_vlan *vlan_key,
2545 struct flow_dissector_key_vlan *vlan_mask)
2546{
2547 int err;
2548
2549 if (!memchr_inv(vlan_mask, 0, sizeof(*vlan_mask)))
2550 return 0;
2551 if (vlan_mask->vlan_id) {
2552 err = nla_put_u16(skb, vlan_id_key,
2553 vlan_key->vlan_id);
2554 if (err)
2555 return err;
2556 }
2557 if (vlan_mask->vlan_priority) {
2558 err = nla_put_u8(skb, vlan_prio_key,
2559 vlan_key->vlan_priority);
2560 if (err)
2561 return err;
2562 }
2563 return 0;
2564}
2565
2566static void fl_get_key_flag(u32 dissector_key, u32 dissector_mask,
2567 u32 *flower_key, u32 *flower_mask,
2568 u32 flower_flag_bit, u32 dissector_flag_bit)
2569{
2570 if (dissector_mask & dissector_flag_bit) {
2571 *flower_mask |= flower_flag_bit;
2572 if (dissector_key & dissector_flag_bit)
2573 *flower_key |= flower_flag_bit;
2574 }
2575}
2576
2577static int fl_dump_key_flags(struct sk_buff *skb, u32 flags_key, u32 flags_mask)
2578{
2579 u32 key, mask;
2580 __be32 _key, _mask;
2581 int err;
2582
2583 if (!memchr_inv(&flags_mask, 0, sizeof(flags_mask)))
2584 return 0;
2585
2586 key = 0;
2587 mask = 0;
2588
2589 fl_get_key_flag(flags_key, flags_mask, &key, &mask,
2590 TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
2591 fl_get_key_flag(flags_key, flags_mask, &key, &mask,
2592 TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
2593 FLOW_DIS_FIRST_FRAG);
2594
2595 _key = cpu_to_be32(key);
2596 _mask = cpu_to_be32(mask);
2597
2598 err = nla_put(skb, TCA_FLOWER_KEY_FLAGS, 4, &_key);
2599 if (err)
2600 return err;
2601
2602 return nla_put(skb, TCA_FLOWER_KEY_FLAGS_MASK, 4, &_mask);
2603}
2604
2605static int fl_dump_key_geneve_opt(struct sk_buff *skb,
2606 struct flow_dissector_key_enc_opts *enc_opts)
2607{
2608 struct geneve_opt *opt;
2609 struct nlattr *nest;
2610 int opt_off = 0;
2611
2612 nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_GENEVE);
2613 if (!nest)
2614 goto nla_put_failure;
2615
2616 while (enc_opts->len > opt_off) {
2617 opt = (struct geneve_opt *)&enc_opts->data[opt_off];
2618
2619 if (nla_put_be16(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS,
2620 opt->opt_class))
2621 goto nla_put_failure;
2622 if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE,
2623 opt->type))
2624 goto nla_put_failure;
2625 if (nla_put(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA,
2626 opt->length * 4, opt->opt_data))
2627 goto nla_put_failure;
2628
2629 opt_off += sizeof(struct geneve_opt) + opt->length * 4;
2630 }
2631 nla_nest_end(skb, nest);
2632 return 0;
2633
2634nla_put_failure:
2635 nla_nest_cancel(skb, nest);
2636 return -EMSGSIZE;
2637}
2638
2639static int fl_dump_key_vxlan_opt(struct sk_buff *skb,
2640 struct flow_dissector_key_enc_opts *enc_opts)
2641{
2642 struct vxlan_metadata *md;
2643 struct nlattr *nest;
2644
2645 nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_VXLAN);
2646 if (!nest)
2647 goto nla_put_failure;
2648
2649 md = (struct vxlan_metadata *)&enc_opts->data[0];
2650 if (nla_put_u32(skb, TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP, md->gbp))
2651 goto nla_put_failure;
2652
2653 nla_nest_end(skb, nest);
2654 return 0;
2655
2656nla_put_failure:
2657 nla_nest_cancel(skb, nest);
2658 return -EMSGSIZE;
2659}
2660
2661static int fl_dump_key_erspan_opt(struct sk_buff *skb,
2662 struct flow_dissector_key_enc_opts *enc_opts)
2663{
2664 struct erspan_metadata *md;
2665 struct nlattr *nest;
2666
2667 nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_ERSPAN);
2668 if (!nest)
2669 goto nla_put_failure;
2670
2671 md = (struct erspan_metadata *)&enc_opts->data[0];
2672 if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER, md->version))
2673 goto nla_put_failure;
2674
2675 if (md->version == 1 &&
2676 nla_put_be32(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX, md->u.index))
2677 goto nla_put_failure;
2678
2679 if (md->version == 2 &&
2680 (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR,
2681 md->u.md2.dir) ||
2682 nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID,
2683 get_hwid(&md->u.md2))))
2684 goto nla_put_failure;
2685
2686 nla_nest_end(skb, nest);
2687 return 0;
2688
2689nla_put_failure:
2690 nla_nest_cancel(skb, nest);
2691 return -EMSGSIZE;
2692}
2693
2694static int fl_dump_key_ct(struct sk_buff *skb,
2695 struct flow_dissector_key_ct *key,
2696 struct flow_dissector_key_ct *mask)
2697{
2698 if (IS_ENABLED(CONFIG_NF_CONNTRACK) &&
2699 fl_dump_key_val(skb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE,
2700 &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK,
2701 sizeof(key->ct_state)))
2702 goto nla_put_failure;
2703
2704 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
2705 fl_dump_key_val(skb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE,
2706 &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK,
2707 sizeof(key->ct_zone)))
2708 goto nla_put_failure;
2709
2710 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
2711 fl_dump_key_val(skb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK,
2712 &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK,
2713 sizeof(key->ct_mark)))
2714 goto nla_put_failure;
2715
2716 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
2717 fl_dump_key_val(skb, &key->ct_labels, TCA_FLOWER_KEY_CT_LABELS,
2718 &mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK,
2719 sizeof(key->ct_labels)))
2720 goto nla_put_failure;
2721
2722 return 0;
2723
2724nla_put_failure:
2725 return -EMSGSIZE;
2726}
2727
2728static int fl_dump_key_options(struct sk_buff *skb, int enc_opt_type,
2729 struct flow_dissector_key_enc_opts *enc_opts)
2730{
2731 struct nlattr *nest;
2732 int err;
2733
2734 if (!enc_opts->len)
2735 return 0;
2736
2737 nest = nla_nest_start_noflag(skb, enc_opt_type);
2738 if (!nest)
2739 goto nla_put_failure;
2740
2741 switch (enc_opts->dst_opt_type) {
2742 case TUNNEL_GENEVE_OPT:
2743 err = fl_dump_key_geneve_opt(skb, enc_opts);
2744 if (err)
2745 goto nla_put_failure;
2746 break;
2747 case TUNNEL_VXLAN_OPT:
2748 err = fl_dump_key_vxlan_opt(skb, enc_opts);
2749 if (err)
2750 goto nla_put_failure;
2751 break;
2752 case TUNNEL_ERSPAN_OPT:
2753 err = fl_dump_key_erspan_opt(skb, enc_opts);
2754 if (err)
2755 goto nla_put_failure;
2756 break;
2757 default:
2758 goto nla_put_failure;
2759 }
2760 nla_nest_end(skb, nest);
2761 return 0;
2762
2763nla_put_failure:
2764 nla_nest_cancel(skb, nest);
2765 return -EMSGSIZE;
2766}
2767
2768static int fl_dump_key_enc_opt(struct sk_buff *skb,
2769 struct flow_dissector_key_enc_opts *key_opts,
2770 struct flow_dissector_key_enc_opts *msk_opts)
2771{
2772 int err;
2773
2774 err = fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS, key_opts);
2775 if (err)
2776 return err;
2777
2778 return fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS_MASK, msk_opts);
2779}
2780
2781static int fl_dump_key(struct sk_buff *skb, struct net *net,
2782 struct fl_flow_key *key, struct fl_flow_key *mask)
2783{
2784 if (mask->meta.ingress_ifindex) {
2785 struct net_device *dev;
2786
2787 dev = __dev_get_by_index(net, key->meta.ingress_ifindex);
2788 if (dev && nla_put_string(skb, TCA_FLOWER_INDEV, dev->name))
2789 goto nla_put_failure;
2790 }
2791
2792 if (fl_dump_key_val(skb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
2793 mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
2794 sizeof(key->eth.dst)) ||
2795 fl_dump_key_val(skb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
2796 mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
2797 sizeof(key->eth.src)) ||
2798 fl_dump_key_val(skb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE,
2799 &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
2800 sizeof(key->basic.n_proto)))
2801 goto nla_put_failure;
2802
2803 if (fl_dump_key_mpls(skb, &key->mpls, &mask->mpls))
2804 goto nla_put_failure;
2805
2806 if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_VLAN_ID,
2807 TCA_FLOWER_KEY_VLAN_PRIO, &key->vlan, &mask->vlan))
2808 goto nla_put_failure;
2809
2810 if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_CVLAN_ID,
2811 TCA_FLOWER_KEY_CVLAN_PRIO,
2812 &key->cvlan, &mask->cvlan) ||
2813 (mask->cvlan.vlan_tpid &&
2814 nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
2815 key->cvlan.vlan_tpid)))
2816 goto nla_put_failure;
2817
2818 if (mask->basic.n_proto) {
2819 if (mask->cvlan.vlan_tpid) {
2820 if (nla_put_be16(skb, TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
2821 key->basic.n_proto))
2822 goto nla_put_failure;
2823 } else if (mask->vlan.vlan_tpid) {
2824 if (nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
2825 key->basic.n_proto))
2826 goto nla_put_failure;
2827 }
2828 }
2829
2830 if ((key->basic.n_proto == htons(ETH_P_IP) ||
2831 key->basic.n_proto == htons(ETH_P_IPV6)) &&
2832 (fl_dump_key_val(skb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
2833 &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
2834 sizeof(key->basic.ip_proto)) ||
2835 fl_dump_key_ip(skb, false, &key->ip, &mask->ip)))
2836 goto nla_put_failure;
2837
2838 if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
2839 (fl_dump_key_val(skb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
2840 &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
2841 sizeof(key->ipv4.src)) ||
2842 fl_dump_key_val(skb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
2843 &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
2844 sizeof(key->ipv4.dst))))
2845 goto nla_put_failure;
2846 else if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
2847 (fl_dump_key_val(skb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
2848 &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
2849 sizeof(key->ipv6.src)) ||
2850 fl_dump_key_val(skb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
2851 &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
2852 sizeof(key->ipv6.dst))))
2853 goto nla_put_failure;
2854
2855 if (key->basic.ip_proto == IPPROTO_TCP &&
2856 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
2857 &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
2858 sizeof(key->tp.src)) ||
2859 fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
2860 &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
2861 sizeof(key->tp.dst)) ||
2862 fl_dump_key_val(skb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
2863 &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
2864 sizeof(key->tcp.flags))))
2865 goto nla_put_failure;
2866 else if (key->basic.ip_proto == IPPROTO_UDP &&
2867 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
2868 &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
2869 sizeof(key->tp.src)) ||
2870 fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
2871 &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
2872 sizeof(key->tp.dst))))
2873 goto nla_put_failure;
2874 else if (key->basic.ip_proto == IPPROTO_SCTP &&
2875 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
2876 &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
2877 sizeof(key->tp.src)) ||
2878 fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
2879 &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
2880 sizeof(key->tp.dst))))
2881 goto nla_put_failure;
2882 else if (key->basic.n_proto == htons(ETH_P_IP) &&
2883 key->basic.ip_proto == IPPROTO_ICMP &&
2884 (fl_dump_key_val(skb, &key->icmp.type,
2885 TCA_FLOWER_KEY_ICMPV4_TYPE, &mask->icmp.type,
2886 TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
2887 sizeof(key->icmp.type)) ||
2888 fl_dump_key_val(skb, &key->icmp.code,
2889 TCA_FLOWER_KEY_ICMPV4_CODE, &mask->icmp.code,
2890 TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
2891 sizeof(key->icmp.code))))
2892 goto nla_put_failure;
2893 else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
2894 key->basic.ip_proto == IPPROTO_ICMPV6 &&
2895 (fl_dump_key_val(skb, &key->icmp.type,
2896 TCA_FLOWER_KEY_ICMPV6_TYPE, &mask->icmp.type,
2897 TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
2898 sizeof(key->icmp.type)) ||
2899 fl_dump_key_val(skb, &key->icmp.code,
2900 TCA_FLOWER_KEY_ICMPV6_CODE, &mask->icmp.code,
2901 TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
2902 sizeof(key->icmp.code))))
2903 goto nla_put_failure;
2904 else if ((key->basic.n_proto == htons(ETH_P_ARP) ||
2905 key->basic.n_proto == htons(ETH_P_RARP)) &&
2906 (fl_dump_key_val(skb, &key->arp.sip,
2907 TCA_FLOWER_KEY_ARP_SIP, &mask->arp.sip,
2908 TCA_FLOWER_KEY_ARP_SIP_MASK,
2909 sizeof(key->arp.sip)) ||
2910 fl_dump_key_val(skb, &key->arp.tip,
2911 TCA_FLOWER_KEY_ARP_TIP, &mask->arp.tip,
2912 TCA_FLOWER_KEY_ARP_TIP_MASK,
2913 sizeof(key->arp.tip)) ||
2914 fl_dump_key_val(skb, &key->arp.op,
2915 TCA_FLOWER_KEY_ARP_OP, &mask->arp.op,
2916 TCA_FLOWER_KEY_ARP_OP_MASK,
2917 sizeof(key->arp.op)) ||
2918 fl_dump_key_val(skb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
2919 mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
2920 sizeof(key->arp.sha)) ||
2921 fl_dump_key_val(skb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
2922 mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
2923 sizeof(key->arp.tha))))
2924 goto nla_put_failure;
2925
2926 if ((key->basic.ip_proto == IPPROTO_TCP ||
2927 key->basic.ip_proto == IPPROTO_UDP ||
2928 key->basic.ip_proto == IPPROTO_SCTP) &&
2929 fl_dump_key_port_range(skb, key, mask))
2930 goto nla_put_failure;
2931
2932 if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
2933 (fl_dump_key_val(skb, &key->enc_ipv4.src,
2934 TCA_FLOWER_KEY_ENC_IPV4_SRC, &mask->enc_ipv4.src,
2935 TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
2936 sizeof(key->enc_ipv4.src)) ||
2937 fl_dump_key_val(skb, &key->enc_ipv4.dst,
2938 TCA_FLOWER_KEY_ENC_IPV4_DST, &mask->enc_ipv4.dst,
2939 TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
2940 sizeof(key->enc_ipv4.dst))))
2941 goto nla_put_failure;
2942 else if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
2943 (fl_dump_key_val(skb, &key->enc_ipv6.src,
2944 TCA_FLOWER_KEY_ENC_IPV6_SRC, &mask->enc_ipv6.src,
2945 TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
2946 sizeof(key->enc_ipv6.src)) ||
2947 fl_dump_key_val(skb, &key->enc_ipv6.dst,
2948 TCA_FLOWER_KEY_ENC_IPV6_DST,
2949 &mask->enc_ipv6.dst,
2950 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
2951 sizeof(key->enc_ipv6.dst))))
2952 goto nla_put_failure;
2953
2954 if (fl_dump_key_val(skb, &key->enc_key_id, TCA_FLOWER_KEY_ENC_KEY_ID,
2955 &mask->enc_key_id, TCA_FLOWER_UNSPEC,
2956 sizeof(key->enc_key_id)) ||
2957 fl_dump_key_val(skb, &key->enc_tp.src,
2958 TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
2959 &mask->enc_tp.src,
2960 TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
2961 sizeof(key->enc_tp.src)) ||
2962 fl_dump_key_val(skb, &key->enc_tp.dst,
2963 TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
2964 &mask->enc_tp.dst,
2965 TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
2966 sizeof(key->enc_tp.dst)) ||
2967 fl_dump_key_ip(skb, true, &key->enc_ip, &mask->enc_ip) ||
2968 fl_dump_key_enc_opt(skb, &key->enc_opts, &mask->enc_opts))
2969 goto nla_put_failure;
2970
2971 if (fl_dump_key_ct(skb, &key->ct, &mask->ct))
2972 goto nla_put_failure;
2973
2974 if (fl_dump_key_flags(skb, key->control.flags, mask->control.flags))
2975 goto nla_put_failure;
2976
2977 if (fl_dump_key_val(skb, &key->hash.hash, TCA_FLOWER_KEY_HASH,
2978 &mask->hash.hash, TCA_FLOWER_KEY_HASH_MASK,
2979 sizeof(key->hash.hash)))
2980 goto nla_put_failure;
2981
2982 return 0;
2983
2984nla_put_failure:
2985 return -EMSGSIZE;
2986}
2987
2988static int fl_dump(struct net *net, struct tcf_proto *tp, void *fh,
2989 struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
2990{
2991 struct cls_fl_filter *f = fh;
2992 struct nlattr *nest;
2993 struct fl_flow_key *key, *mask;
2994 bool skip_hw;
2995
2996 if (!f)
2997 return skb->len;
2998
2999 t->tcm_handle = f->handle;
3000
3001 nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
3002 if (!nest)
3003 goto nla_put_failure;
3004
3005 spin_lock(&tp->lock);
3006
3007 if (f->res.classid &&
3008 nla_put_u32(skb, TCA_FLOWER_CLASSID, f->res.classid))
3009 goto nla_put_failure_locked;
3010
3011 key = &f->key;
3012 mask = &f->mask->key;
3013 skip_hw = tc_skip_hw(f->flags);
3014
3015 if (fl_dump_key(skb, net, key, mask))
3016 goto nla_put_failure_locked;
3017
3018 if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
3019 goto nla_put_failure_locked;
3020
3021 spin_unlock(&tp->lock);
3022
3023 if (!skip_hw)
3024 fl_hw_update_stats(tp, f, rtnl_held);
3025
3026 if (nla_put_u32(skb, TCA_FLOWER_IN_HW_COUNT, f->in_hw_count))
3027 goto nla_put_failure;
3028
3029 if (tcf_exts_dump(skb, &f->exts))
3030 goto nla_put_failure;
3031
3032 nla_nest_end(skb, nest);
3033
3034 if (tcf_exts_dump_stats(skb, &f->exts) < 0)
3035 goto nla_put_failure;
3036
3037 return skb->len;
3038
3039nla_put_failure_locked:
3040 spin_unlock(&tp->lock);
3041nla_put_failure:
3042 nla_nest_cancel(skb, nest);
3043 return -1;
3044}
3045
3046static int fl_terse_dump(struct net *net, struct tcf_proto *tp, void *fh,
3047 struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
3048{
3049 struct cls_fl_filter *f = fh;
3050 struct nlattr *nest;
3051 bool skip_hw;
3052
3053 if (!f)
3054 return skb->len;
3055
3056 t->tcm_handle = f->handle;
3057
3058 nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
3059 if (!nest)
3060 goto nla_put_failure;
3061
3062 spin_lock(&tp->lock);
3063
3064 skip_hw = tc_skip_hw(f->flags);
3065
3066 if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
3067 goto nla_put_failure_locked;
3068
3069 spin_unlock(&tp->lock);
3070
3071 if (!skip_hw)
3072 fl_hw_update_stats(tp, f, rtnl_held);
3073
3074 if (tcf_exts_terse_dump(skb, &f->exts))
3075 goto nla_put_failure;
3076
3077 nla_nest_end(skb, nest);
3078
3079 return skb->len;
3080
3081nla_put_failure_locked:
3082 spin_unlock(&tp->lock);
3083nla_put_failure:
3084 nla_nest_cancel(skb, nest);
3085 return -1;
3086}
3087
3088static int fl_tmplt_dump(struct sk_buff *skb, struct net *net, void *tmplt_priv)
3089{
3090 struct fl_flow_tmplt *tmplt = tmplt_priv;
3091 struct fl_flow_key *key, *mask;
3092 struct nlattr *nest;
3093
3094 nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
3095 if (!nest)
3096 goto nla_put_failure;
3097
3098 key = &tmplt->dummy_key;
3099 mask = &tmplt->mask;
3100
3101 if (fl_dump_key(skb, net, key, mask))
3102 goto nla_put_failure;
3103
3104 nla_nest_end(skb, nest);
3105
3106 return skb->len;
3107
3108nla_put_failure:
3109 nla_nest_cancel(skb, nest);
3110 return -EMSGSIZE;
3111}
3112
3113static void fl_bind_class(void *fh, u32 classid, unsigned long cl, void *q,
3114 unsigned long base)
3115{
3116 struct cls_fl_filter *f = fh;
3117
3118 if (f && f->res.classid == classid) {
3119 if (cl)
3120 __tcf_bind_filter(q, &f->res, base);
3121 else
3122 __tcf_unbind_filter(q, &f->res);
3123 }
3124}
3125
3126static bool fl_delete_empty(struct tcf_proto *tp)
3127{
3128 struct cls_fl_head *head = fl_head_dereference(tp);
3129
3130 spin_lock(&tp->lock);
3131 tp->deleting = idr_is_empty(&head->handle_idr);
3132 spin_unlock(&tp->lock);
3133
3134 return tp->deleting;
3135}
3136
3137static struct tcf_proto_ops cls_fl_ops __read_mostly = {
3138 .kind = "flower",
3139 .classify = fl_classify,
3140 .init = fl_init,
3141 .destroy = fl_destroy,
3142 .get = fl_get,
3143 .put = fl_put,
3144 .change = fl_change,
3145 .delete = fl_delete,
3146 .delete_empty = fl_delete_empty,
3147 .walk = fl_walk,
3148 .reoffload = fl_reoffload,
3149 .hw_add = fl_hw_add,
3150 .hw_del = fl_hw_del,
3151 .dump = fl_dump,
3152 .terse_dump = fl_terse_dump,
3153 .bind_class = fl_bind_class,
3154 .tmplt_create = fl_tmplt_create,
3155 .tmplt_destroy = fl_tmplt_destroy,
3156 .tmplt_dump = fl_tmplt_dump,
3157 .owner = THIS_MODULE,
3158 .flags = TCF_PROTO_OPS_DOIT_UNLOCKED,
3159};
3160
3161static int __init cls_fl_init(void)
3162{
3163 return register_tcf_proto_ops(&cls_fl_ops);
3164}
3165
3166static void __exit cls_fl_exit(void)
3167{
3168 unregister_tcf_proto_ops(&cls_fl_ops);
3169}
3170
3171module_init(cls_fl_init);
3172module_exit(cls_fl_exit);
3173
3174MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
3175MODULE_DESCRIPTION("Flower classifier");
3176MODULE_LICENSE("GPL v2");