1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * net/sched/cls_flower.c Flower classifier
5 * Copyright (c) 2015 Jiri Pirko <jiri@resnulli.us>
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>
15 #include <linux/if_ether.h>
16 #include <linux/in6.h>
18 #include <linux/mpls.h>
20 #include <net/sch_generic.h>
21 #include <net/pkt_cls.h>
23 #include <net/flow_dissector.h>
24 #include <net/geneve.h>
27 #include <net/dst_metadata.h>
31 struct flow_dissector_key_control control;
32 struct flow_dissector_key_control enc_control;
33 struct flow_dissector_key_basic basic;
34 struct flow_dissector_key_eth_addrs eth;
35 struct flow_dissector_key_vlan vlan;
36 struct flow_dissector_key_vlan cvlan;
38 struct flow_dissector_key_ipv4_addrs ipv4;
39 struct flow_dissector_key_ipv6_addrs ipv6;
41 struct flow_dissector_key_ports tp;
42 struct flow_dissector_key_icmp icmp;
43 struct flow_dissector_key_arp arp;
44 struct flow_dissector_key_keyid enc_key_id;
46 struct flow_dissector_key_ipv4_addrs enc_ipv4;
47 struct flow_dissector_key_ipv6_addrs enc_ipv6;
49 struct flow_dissector_key_ports enc_tp;
50 struct flow_dissector_key_mpls mpls;
51 struct flow_dissector_key_tcp tcp;
52 struct flow_dissector_key_ip ip;
53 struct flow_dissector_key_ip enc_ip;
54 struct flow_dissector_key_enc_opts enc_opts;
55 struct flow_dissector_key_ports tp_min;
56 struct flow_dissector_key_ports tp_max;
57 } __aligned(BITS_PER_LONG / 8); /* Ensure that we can do comparisons as longs. */
59 struct fl_flow_mask_range {
60 unsigned short int start;
61 unsigned short int end;
65 struct fl_flow_key key;
66 struct fl_flow_mask_range range;
68 struct rhash_head ht_node;
70 struct rhashtable_params filter_ht_params;
71 struct flow_dissector dissector;
72 struct list_head filters;
73 struct rcu_work rwork;
74 struct list_head list;
78 struct fl_flow_tmplt {
79 struct fl_flow_key dummy_key;
80 struct fl_flow_key mask;
81 struct flow_dissector dissector;
82 struct tcf_chain *chain;
87 spinlock_t masks_lock; /* Protect masks list */
88 struct list_head masks;
89 struct list_head hw_filters;
90 struct rcu_work rwork;
91 struct idr handle_idr;
94 struct cls_fl_filter {
95 struct fl_flow_mask *mask;
96 struct rhash_head ht_node;
97 struct fl_flow_key mkey;
99 struct tcf_result res;
100 struct fl_flow_key key;
101 struct list_head list;
102 struct list_head hw_list;
106 struct rcu_work rwork;
107 struct net_device *hw_dev;
108 /* Flower classifier is unlocked, which means that its reference counter
109 * can be changed concurrently without any kind of external
110 * synchronization. Use atomic reference counter to be concurrency-safe.
116 static const struct rhashtable_params mask_ht_params = {
117 .key_offset = offsetof(struct fl_flow_mask, key),
118 .key_len = sizeof(struct fl_flow_key),
119 .head_offset = offsetof(struct fl_flow_mask, ht_node),
120 .automatic_shrinking = true,
123 static unsigned short int fl_mask_range(const struct fl_flow_mask *mask)
125 return mask->range.end - mask->range.start;
128 static void fl_mask_update_range(struct fl_flow_mask *mask)
130 const u8 *bytes = (const u8 *) &mask->key;
131 size_t size = sizeof(mask->key);
132 size_t i, first = 0, last;
134 for (i = 0; i < size; i++) {
141 for (i = size - 1; i != first; i--) {
147 mask->range.start = rounddown(first, sizeof(long));
148 mask->range.end = roundup(last + 1, sizeof(long));
151 static void *fl_key_get_start(struct fl_flow_key *key,
152 const struct fl_flow_mask *mask)
154 return (u8 *) key + mask->range.start;
157 static void fl_set_masked_key(struct fl_flow_key *mkey, struct fl_flow_key *key,
158 struct fl_flow_mask *mask)
160 const long *lkey = fl_key_get_start(key, mask);
161 const long *lmask = fl_key_get_start(&mask->key, mask);
162 long *lmkey = fl_key_get_start(mkey, mask);
165 for (i = 0; i < fl_mask_range(mask); i += sizeof(long))
166 *lmkey++ = *lkey++ & *lmask++;
169 static bool fl_mask_fits_tmplt(struct fl_flow_tmplt *tmplt,
170 struct fl_flow_mask *mask)
172 const long *lmask = fl_key_get_start(&mask->key, mask);
178 ltmplt = fl_key_get_start(&tmplt->mask, mask);
179 for (i = 0; i < fl_mask_range(mask); i += sizeof(long)) {
180 if (~*ltmplt++ & *lmask++)
186 static void fl_clear_masked_range(struct fl_flow_key *key,
187 struct fl_flow_mask *mask)
189 memset(fl_key_get_start(key, mask), 0, fl_mask_range(mask));
192 static bool fl_range_port_dst_cmp(struct cls_fl_filter *filter,
193 struct fl_flow_key *key,
194 struct fl_flow_key *mkey)
196 __be16 min_mask, max_mask, min_val, max_val;
198 min_mask = htons(filter->mask->key.tp_min.dst);
199 max_mask = htons(filter->mask->key.tp_max.dst);
200 min_val = htons(filter->key.tp_min.dst);
201 max_val = htons(filter->key.tp_max.dst);
203 if (min_mask && max_mask) {
204 if (htons(key->tp.dst) < min_val ||
205 htons(key->tp.dst) > max_val)
208 /* skb does not have min and max values */
209 mkey->tp_min.dst = filter->mkey.tp_min.dst;
210 mkey->tp_max.dst = filter->mkey.tp_max.dst;
215 static bool fl_range_port_src_cmp(struct cls_fl_filter *filter,
216 struct fl_flow_key *key,
217 struct fl_flow_key *mkey)
219 __be16 min_mask, max_mask, min_val, max_val;
221 min_mask = htons(filter->mask->key.tp_min.src);
222 max_mask = htons(filter->mask->key.tp_max.src);
223 min_val = htons(filter->key.tp_min.src);
224 max_val = htons(filter->key.tp_max.src);
226 if (min_mask && max_mask) {
227 if (htons(key->tp.src) < min_val ||
228 htons(key->tp.src) > max_val)
231 /* skb does not have min and max values */
232 mkey->tp_min.src = filter->mkey.tp_min.src;
233 mkey->tp_max.src = filter->mkey.tp_max.src;
238 static struct cls_fl_filter *__fl_lookup(struct fl_flow_mask *mask,
239 struct fl_flow_key *mkey)
241 return rhashtable_lookup_fast(&mask->ht, fl_key_get_start(mkey, mask),
242 mask->filter_ht_params);
245 static struct cls_fl_filter *fl_lookup_range(struct fl_flow_mask *mask,
246 struct fl_flow_key *mkey,
247 struct fl_flow_key *key)
249 struct cls_fl_filter *filter, *f;
251 list_for_each_entry_rcu(filter, &mask->filters, list) {
252 if (!fl_range_port_dst_cmp(filter, key, mkey))
255 if (!fl_range_port_src_cmp(filter, key, mkey))
258 f = __fl_lookup(mask, mkey);
265 static struct cls_fl_filter *fl_lookup(struct fl_flow_mask *mask,
266 struct fl_flow_key *mkey,
267 struct fl_flow_key *key)
269 if ((mask->flags & TCA_FLOWER_MASK_FLAGS_RANGE))
270 return fl_lookup_range(mask, mkey, key);
272 return __fl_lookup(mask, mkey);
275 static int fl_classify(struct sk_buff *skb, const struct tcf_proto *tp,
276 struct tcf_result *res)
278 struct cls_fl_head *head = rcu_dereference_bh(tp->root);
279 struct cls_fl_filter *f;
280 struct fl_flow_mask *mask;
281 struct fl_flow_key skb_key;
282 struct fl_flow_key skb_mkey;
284 list_for_each_entry_rcu(mask, &head->masks, list) {
285 fl_clear_masked_range(&skb_key, mask);
287 skb_key.indev_ifindex = skb->skb_iif;
288 /* skb_flow_dissect() does not set n_proto in case an unknown
289 * protocol, so do it rather here.
291 skb_key.basic.n_proto = skb->protocol;
292 skb_flow_dissect_tunnel_info(skb, &mask->dissector, &skb_key);
293 skb_flow_dissect(skb, &mask->dissector, &skb_key, 0);
295 fl_set_masked_key(&skb_mkey, &skb_key, mask);
297 f = fl_lookup(mask, &skb_mkey, &skb_key);
298 if (f && !tc_skip_sw(f->flags)) {
300 return tcf_exts_exec(skb, &f->exts, res);
306 static int fl_init(struct tcf_proto *tp)
308 struct cls_fl_head *head;
310 head = kzalloc(sizeof(*head), GFP_KERNEL);
314 spin_lock_init(&head->masks_lock);
315 INIT_LIST_HEAD_RCU(&head->masks);
316 INIT_LIST_HEAD(&head->hw_filters);
317 rcu_assign_pointer(tp->root, head);
318 idr_init(&head->handle_idr);
320 return rhashtable_init(&head->ht, &mask_ht_params);
323 static void fl_mask_free(struct fl_flow_mask *mask)
325 WARN_ON(!list_empty(&mask->filters));
326 rhashtable_destroy(&mask->ht);
330 static void fl_mask_free_work(struct work_struct *work)
332 struct fl_flow_mask *mask = container_of(to_rcu_work(work),
333 struct fl_flow_mask, rwork);
338 static bool fl_mask_put(struct cls_fl_head *head, struct fl_flow_mask *mask)
340 if (!refcount_dec_and_test(&mask->refcnt))
343 rhashtable_remove_fast(&head->ht, &mask->ht_node, mask_ht_params);
345 spin_lock(&head->masks_lock);
346 list_del_rcu(&mask->list);
347 spin_unlock(&head->masks_lock);
349 tcf_queue_work(&mask->rwork, fl_mask_free_work);
354 static struct cls_fl_head *fl_head_dereference(struct tcf_proto *tp)
356 /* Flower classifier only changes root pointer during init and destroy.
357 * Users must obtain reference to tcf_proto instance before calling its
358 * API, so tp->root pointer is protected from concurrent call to
359 * fl_destroy() by reference counting.
361 return rcu_dereference_raw(tp->root);
364 static void __fl_destroy_filter(struct cls_fl_filter *f)
366 tcf_exts_destroy(&f->exts);
367 tcf_exts_put_net(&f->exts);
371 static void fl_destroy_filter_work(struct work_struct *work)
373 struct cls_fl_filter *f = container_of(to_rcu_work(work),
374 struct cls_fl_filter, rwork);
376 __fl_destroy_filter(f);
379 static void fl_hw_destroy_filter(struct tcf_proto *tp, struct cls_fl_filter *f,
380 bool rtnl_held, struct netlink_ext_ack *extack)
382 struct tc_cls_flower_offload cls_flower = {};
383 struct tcf_block *block = tp->chain->block;
388 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
389 cls_flower.command = TC_CLSFLOWER_DESTROY;
390 cls_flower.cookie = (unsigned long) f;
392 tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false);
393 spin_lock(&tp->lock);
394 list_del_init(&f->hw_list);
395 tcf_block_offload_dec(block, &f->flags);
396 spin_unlock(&tp->lock);
402 static int fl_hw_replace_filter(struct tcf_proto *tp,
403 struct cls_fl_filter *f, bool rtnl_held,
404 struct netlink_ext_ack *extack)
406 struct cls_fl_head *head = fl_head_dereference(tp);
407 struct tc_cls_flower_offload cls_flower = {};
408 struct tcf_block *block = tp->chain->block;
409 bool skip_sw = tc_skip_sw(f->flags);
415 cls_flower.rule = flow_rule_alloc(tcf_exts_num_actions(&f->exts));
416 if (!cls_flower.rule) {
421 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
422 cls_flower.command = TC_CLSFLOWER_REPLACE;
423 cls_flower.cookie = (unsigned long) f;
424 cls_flower.rule->match.dissector = &f->mask->dissector;
425 cls_flower.rule->match.mask = &f->mask->key;
426 cls_flower.rule->match.key = &f->mkey;
427 cls_flower.classid = f->res.classid;
429 err = tc_setup_flow_action(&cls_flower.rule->action, &f->exts);
431 kfree(cls_flower.rule);
433 NL_SET_ERR_MSG_MOD(extack, "Failed to setup flow action");
439 err = tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, skip_sw);
440 kfree(cls_flower.rule);
443 fl_hw_destroy_filter(tp, f, true, NULL);
445 } else if (err > 0) {
446 f->in_hw_count = err;
448 spin_lock(&tp->lock);
449 tcf_block_offload_inc(block, &f->flags);
450 spin_unlock(&tp->lock);
453 if (skip_sw && !(f->flags & TCA_CLS_FLAGS_IN_HW)) {
458 spin_lock(&tp->lock);
459 list_add(&f->hw_list, &head->hw_filters);
460 spin_unlock(&tp->lock);
468 static void fl_hw_update_stats(struct tcf_proto *tp, struct cls_fl_filter *f,
471 struct tc_cls_flower_offload cls_flower = {};
472 struct tcf_block *block = tp->chain->block;
477 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, NULL);
478 cls_flower.command = TC_CLSFLOWER_STATS;
479 cls_flower.cookie = (unsigned long) f;
480 cls_flower.classid = f->res.classid;
482 tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false);
484 tcf_exts_stats_update(&f->exts, cls_flower.stats.bytes,
485 cls_flower.stats.pkts,
486 cls_flower.stats.lastused);
492 static void __fl_put(struct cls_fl_filter *f)
494 if (!refcount_dec_and_test(&f->refcnt))
497 if (tcf_exts_get_net(&f->exts))
498 tcf_queue_work(&f->rwork, fl_destroy_filter_work);
500 __fl_destroy_filter(f);
503 static struct cls_fl_filter *__fl_get(struct cls_fl_head *head, u32 handle)
505 struct cls_fl_filter *f;
508 f = idr_find(&head->handle_idr, handle);
509 if (f && !refcount_inc_not_zero(&f->refcnt))
516 static struct cls_fl_filter *fl_get_next_filter(struct tcf_proto *tp,
517 unsigned long *handle)
519 struct cls_fl_head *head = fl_head_dereference(tp);
520 struct cls_fl_filter *f;
523 while ((f = idr_get_next_ul(&head->handle_idr, handle))) {
524 /* don't return filters that are being deleted */
525 if (refcount_inc_not_zero(&f->refcnt))
534 static int __fl_delete(struct tcf_proto *tp, struct cls_fl_filter *f,
535 bool *last, bool rtnl_held,
536 struct netlink_ext_ack *extack)
538 struct cls_fl_head *head = fl_head_dereference(tp);
542 spin_lock(&tp->lock);
544 spin_unlock(&tp->lock);
549 rhashtable_remove_fast(&f->mask->ht, &f->ht_node,
550 f->mask->filter_ht_params);
551 idr_remove(&head->handle_idr, f->handle);
552 list_del_rcu(&f->list);
553 spin_unlock(&tp->lock);
555 *last = fl_mask_put(head, f->mask);
556 if (!tc_skip_hw(f->flags))
557 fl_hw_destroy_filter(tp, f, rtnl_held, extack);
558 tcf_unbind_filter(tp, &f->res);
564 static void fl_destroy_sleepable(struct work_struct *work)
566 struct cls_fl_head *head = container_of(to_rcu_work(work),
570 rhashtable_destroy(&head->ht);
572 module_put(THIS_MODULE);
575 static void fl_destroy(struct tcf_proto *tp, bool rtnl_held,
576 struct netlink_ext_ack *extack)
578 struct cls_fl_head *head = fl_head_dereference(tp);
579 struct fl_flow_mask *mask, *next_mask;
580 struct cls_fl_filter *f, *next;
583 list_for_each_entry_safe(mask, next_mask, &head->masks, list) {
584 list_for_each_entry_safe(f, next, &mask->filters, list) {
585 __fl_delete(tp, f, &last, rtnl_held, extack);
590 idr_destroy(&head->handle_idr);
592 __module_get(THIS_MODULE);
593 tcf_queue_work(&head->rwork, fl_destroy_sleepable);
596 static void fl_put(struct tcf_proto *tp, void *arg)
598 struct cls_fl_filter *f = arg;
603 static void *fl_get(struct tcf_proto *tp, u32 handle)
605 struct cls_fl_head *head = fl_head_dereference(tp);
607 return __fl_get(head, handle);
610 static const struct nla_policy fl_policy[TCA_FLOWER_MAX + 1] = {
611 [TCA_FLOWER_UNSPEC] = { .type = NLA_UNSPEC },
612 [TCA_FLOWER_CLASSID] = { .type = NLA_U32 },
613 [TCA_FLOWER_INDEV] = { .type = NLA_STRING,
615 [TCA_FLOWER_KEY_ETH_DST] = { .len = ETH_ALEN },
616 [TCA_FLOWER_KEY_ETH_DST_MASK] = { .len = ETH_ALEN },
617 [TCA_FLOWER_KEY_ETH_SRC] = { .len = ETH_ALEN },
618 [TCA_FLOWER_KEY_ETH_SRC_MASK] = { .len = ETH_ALEN },
619 [TCA_FLOWER_KEY_ETH_TYPE] = { .type = NLA_U16 },
620 [TCA_FLOWER_KEY_IP_PROTO] = { .type = NLA_U8 },
621 [TCA_FLOWER_KEY_IPV4_SRC] = { .type = NLA_U32 },
622 [TCA_FLOWER_KEY_IPV4_SRC_MASK] = { .type = NLA_U32 },
623 [TCA_FLOWER_KEY_IPV4_DST] = { .type = NLA_U32 },
624 [TCA_FLOWER_KEY_IPV4_DST_MASK] = { .type = NLA_U32 },
625 [TCA_FLOWER_KEY_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
626 [TCA_FLOWER_KEY_IPV6_SRC_MASK] = { .len = sizeof(struct in6_addr) },
627 [TCA_FLOWER_KEY_IPV6_DST] = { .len = sizeof(struct in6_addr) },
628 [TCA_FLOWER_KEY_IPV6_DST_MASK] = { .len = sizeof(struct in6_addr) },
629 [TCA_FLOWER_KEY_TCP_SRC] = { .type = NLA_U16 },
630 [TCA_FLOWER_KEY_TCP_DST] = { .type = NLA_U16 },
631 [TCA_FLOWER_KEY_UDP_SRC] = { .type = NLA_U16 },
632 [TCA_FLOWER_KEY_UDP_DST] = { .type = NLA_U16 },
633 [TCA_FLOWER_KEY_VLAN_ID] = { .type = NLA_U16 },
634 [TCA_FLOWER_KEY_VLAN_PRIO] = { .type = NLA_U8 },
635 [TCA_FLOWER_KEY_VLAN_ETH_TYPE] = { .type = NLA_U16 },
636 [TCA_FLOWER_KEY_ENC_KEY_ID] = { .type = NLA_U32 },
637 [TCA_FLOWER_KEY_ENC_IPV4_SRC] = { .type = NLA_U32 },
638 [TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK] = { .type = NLA_U32 },
639 [TCA_FLOWER_KEY_ENC_IPV4_DST] = { .type = NLA_U32 },
640 [TCA_FLOWER_KEY_ENC_IPV4_DST_MASK] = { .type = NLA_U32 },
641 [TCA_FLOWER_KEY_ENC_IPV6_SRC] = { .len = sizeof(struct in6_addr) },
642 [TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK] = { .len = sizeof(struct in6_addr) },
643 [TCA_FLOWER_KEY_ENC_IPV6_DST] = { .len = sizeof(struct in6_addr) },
644 [TCA_FLOWER_KEY_ENC_IPV6_DST_MASK] = { .len = sizeof(struct in6_addr) },
645 [TCA_FLOWER_KEY_TCP_SRC_MASK] = { .type = NLA_U16 },
646 [TCA_FLOWER_KEY_TCP_DST_MASK] = { .type = NLA_U16 },
647 [TCA_FLOWER_KEY_UDP_SRC_MASK] = { .type = NLA_U16 },
648 [TCA_FLOWER_KEY_UDP_DST_MASK] = { .type = NLA_U16 },
649 [TCA_FLOWER_KEY_SCTP_SRC_MASK] = { .type = NLA_U16 },
650 [TCA_FLOWER_KEY_SCTP_DST_MASK] = { .type = NLA_U16 },
651 [TCA_FLOWER_KEY_SCTP_SRC] = { .type = NLA_U16 },
652 [TCA_FLOWER_KEY_SCTP_DST] = { .type = NLA_U16 },
653 [TCA_FLOWER_KEY_ENC_UDP_SRC_PORT] = { .type = NLA_U16 },
654 [TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK] = { .type = NLA_U16 },
655 [TCA_FLOWER_KEY_ENC_UDP_DST_PORT] = { .type = NLA_U16 },
656 [TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK] = { .type = NLA_U16 },
657 [TCA_FLOWER_KEY_FLAGS] = { .type = NLA_U32 },
658 [TCA_FLOWER_KEY_FLAGS_MASK] = { .type = NLA_U32 },
659 [TCA_FLOWER_KEY_ICMPV4_TYPE] = { .type = NLA_U8 },
660 [TCA_FLOWER_KEY_ICMPV4_TYPE_MASK] = { .type = NLA_U8 },
661 [TCA_FLOWER_KEY_ICMPV4_CODE] = { .type = NLA_U8 },
662 [TCA_FLOWER_KEY_ICMPV4_CODE_MASK] = { .type = NLA_U8 },
663 [TCA_FLOWER_KEY_ICMPV6_TYPE] = { .type = NLA_U8 },
664 [TCA_FLOWER_KEY_ICMPV6_TYPE_MASK] = { .type = NLA_U8 },
665 [TCA_FLOWER_KEY_ICMPV6_CODE] = { .type = NLA_U8 },
666 [TCA_FLOWER_KEY_ICMPV6_CODE_MASK] = { .type = NLA_U8 },
667 [TCA_FLOWER_KEY_ARP_SIP] = { .type = NLA_U32 },
668 [TCA_FLOWER_KEY_ARP_SIP_MASK] = { .type = NLA_U32 },
669 [TCA_FLOWER_KEY_ARP_TIP] = { .type = NLA_U32 },
670 [TCA_FLOWER_KEY_ARP_TIP_MASK] = { .type = NLA_U32 },
671 [TCA_FLOWER_KEY_ARP_OP] = { .type = NLA_U8 },
672 [TCA_FLOWER_KEY_ARP_OP_MASK] = { .type = NLA_U8 },
673 [TCA_FLOWER_KEY_ARP_SHA] = { .len = ETH_ALEN },
674 [TCA_FLOWER_KEY_ARP_SHA_MASK] = { .len = ETH_ALEN },
675 [TCA_FLOWER_KEY_ARP_THA] = { .len = ETH_ALEN },
676 [TCA_FLOWER_KEY_ARP_THA_MASK] = { .len = ETH_ALEN },
677 [TCA_FLOWER_KEY_MPLS_TTL] = { .type = NLA_U8 },
678 [TCA_FLOWER_KEY_MPLS_BOS] = { .type = NLA_U8 },
679 [TCA_FLOWER_KEY_MPLS_TC] = { .type = NLA_U8 },
680 [TCA_FLOWER_KEY_MPLS_LABEL] = { .type = NLA_U32 },
681 [TCA_FLOWER_KEY_TCP_FLAGS] = { .type = NLA_U16 },
682 [TCA_FLOWER_KEY_TCP_FLAGS_MASK] = { .type = NLA_U16 },
683 [TCA_FLOWER_KEY_IP_TOS] = { .type = NLA_U8 },
684 [TCA_FLOWER_KEY_IP_TOS_MASK] = { .type = NLA_U8 },
685 [TCA_FLOWER_KEY_IP_TTL] = { .type = NLA_U8 },
686 [TCA_FLOWER_KEY_IP_TTL_MASK] = { .type = NLA_U8 },
687 [TCA_FLOWER_KEY_CVLAN_ID] = { .type = NLA_U16 },
688 [TCA_FLOWER_KEY_CVLAN_PRIO] = { .type = NLA_U8 },
689 [TCA_FLOWER_KEY_CVLAN_ETH_TYPE] = { .type = NLA_U16 },
690 [TCA_FLOWER_KEY_ENC_IP_TOS] = { .type = NLA_U8 },
691 [TCA_FLOWER_KEY_ENC_IP_TOS_MASK] = { .type = NLA_U8 },
692 [TCA_FLOWER_KEY_ENC_IP_TTL] = { .type = NLA_U8 },
693 [TCA_FLOWER_KEY_ENC_IP_TTL_MASK] = { .type = NLA_U8 },
694 [TCA_FLOWER_KEY_ENC_OPTS] = { .type = NLA_NESTED },
695 [TCA_FLOWER_KEY_ENC_OPTS_MASK] = { .type = NLA_NESTED },
698 static const struct nla_policy
699 enc_opts_policy[TCA_FLOWER_KEY_ENC_OPTS_MAX + 1] = {
700 [TCA_FLOWER_KEY_ENC_OPTS_GENEVE] = { .type = NLA_NESTED },
703 static const struct nla_policy
704 geneve_opt_policy[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1] = {
705 [TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS] = { .type = NLA_U16 },
706 [TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE] = { .type = NLA_U8 },
707 [TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA] = { .type = NLA_BINARY,
711 static void fl_set_key_val(struct nlattr **tb,
712 void *val, int val_type,
713 void *mask, int mask_type, int len)
717 memcpy(val, nla_data(tb[val_type]), len);
718 if (mask_type == TCA_FLOWER_UNSPEC || !tb[mask_type])
719 memset(mask, 0xff, len);
721 memcpy(mask, nla_data(tb[mask_type]), len);
724 static int fl_set_key_port_range(struct nlattr **tb, struct fl_flow_key *key,
725 struct fl_flow_key *mask)
727 fl_set_key_val(tb, &key->tp_min.dst,
728 TCA_FLOWER_KEY_PORT_DST_MIN, &mask->tp_min.dst,
729 TCA_FLOWER_UNSPEC, sizeof(key->tp_min.dst));
730 fl_set_key_val(tb, &key->tp_max.dst,
731 TCA_FLOWER_KEY_PORT_DST_MAX, &mask->tp_max.dst,
732 TCA_FLOWER_UNSPEC, sizeof(key->tp_max.dst));
733 fl_set_key_val(tb, &key->tp_min.src,
734 TCA_FLOWER_KEY_PORT_SRC_MIN, &mask->tp_min.src,
735 TCA_FLOWER_UNSPEC, sizeof(key->tp_min.src));
736 fl_set_key_val(tb, &key->tp_max.src,
737 TCA_FLOWER_KEY_PORT_SRC_MAX, &mask->tp_max.src,
738 TCA_FLOWER_UNSPEC, sizeof(key->tp_max.src));
740 if ((mask->tp_min.dst && mask->tp_max.dst &&
741 htons(key->tp_max.dst) <= htons(key->tp_min.dst)) ||
742 (mask->tp_min.src && mask->tp_max.src &&
743 htons(key->tp_max.src) <= htons(key->tp_min.src)))
749 static int fl_set_key_mpls(struct nlattr **tb,
750 struct flow_dissector_key_mpls *key_val,
751 struct flow_dissector_key_mpls *key_mask)
753 if (tb[TCA_FLOWER_KEY_MPLS_TTL]) {
754 key_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TTL]);
755 key_mask->mpls_ttl = MPLS_TTL_MASK;
757 if (tb[TCA_FLOWER_KEY_MPLS_BOS]) {
758 u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_BOS]);
760 if (bos & ~MPLS_BOS_MASK)
762 key_val->mpls_bos = bos;
763 key_mask->mpls_bos = MPLS_BOS_MASK;
765 if (tb[TCA_FLOWER_KEY_MPLS_TC]) {
766 u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TC]);
768 if (tc & ~MPLS_TC_MASK)
770 key_val->mpls_tc = tc;
771 key_mask->mpls_tc = MPLS_TC_MASK;
773 if (tb[TCA_FLOWER_KEY_MPLS_LABEL]) {
774 u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_LABEL]);
776 if (label & ~MPLS_LABEL_MASK)
778 key_val->mpls_label = label;
779 key_mask->mpls_label = MPLS_LABEL_MASK;
784 static void fl_set_key_vlan(struct nlattr **tb,
786 int vlan_id_key, int vlan_prio_key,
787 struct flow_dissector_key_vlan *key_val,
788 struct flow_dissector_key_vlan *key_mask)
790 #define VLAN_PRIORITY_MASK 0x7
792 if (tb[vlan_id_key]) {
794 nla_get_u16(tb[vlan_id_key]) & VLAN_VID_MASK;
795 key_mask->vlan_id = VLAN_VID_MASK;
797 if (tb[vlan_prio_key]) {
798 key_val->vlan_priority =
799 nla_get_u8(tb[vlan_prio_key]) &
801 key_mask->vlan_priority = VLAN_PRIORITY_MASK;
803 key_val->vlan_tpid = ethertype;
804 key_mask->vlan_tpid = cpu_to_be16(~0);
807 static void fl_set_key_flag(u32 flower_key, u32 flower_mask,
808 u32 *dissector_key, u32 *dissector_mask,
809 u32 flower_flag_bit, u32 dissector_flag_bit)
811 if (flower_mask & flower_flag_bit) {
812 *dissector_mask |= dissector_flag_bit;
813 if (flower_key & flower_flag_bit)
814 *dissector_key |= dissector_flag_bit;
818 static int fl_set_key_flags(struct nlattr **tb,
819 u32 *flags_key, u32 *flags_mask)
823 /* mask is mandatory for flags */
824 if (!tb[TCA_FLOWER_KEY_FLAGS_MASK])
827 key = be32_to_cpu(nla_get_u32(tb[TCA_FLOWER_KEY_FLAGS]));
828 mask = be32_to_cpu(nla_get_u32(tb[TCA_FLOWER_KEY_FLAGS_MASK]));
833 fl_set_key_flag(key, mask, flags_key, flags_mask,
834 TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
835 fl_set_key_flag(key, mask, flags_key, flags_mask,
836 TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
837 FLOW_DIS_FIRST_FRAG);
842 static void fl_set_key_ip(struct nlattr **tb, bool encap,
843 struct flow_dissector_key_ip *key,
844 struct flow_dissector_key_ip *mask)
846 int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
847 int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
848 int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
849 int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
851 fl_set_key_val(tb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos));
852 fl_set_key_val(tb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl));
855 static int fl_set_geneve_opt(const struct nlattr *nla, struct fl_flow_key *key,
856 int depth, int option_len,
857 struct netlink_ext_ack *extack)
859 struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1];
860 struct nlattr *class = NULL, *type = NULL, *data = NULL;
861 struct geneve_opt *opt;
862 int err, data_len = 0;
864 if (option_len > sizeof(struct geneve_opt))
865 data_len = option_len - sizeof(struct geneve_opt);
867 opt = (struct geneve_opt *)&key->enc_opts.data[key->enc_opts.len];
868 memset(opt, 0xff, option_len);
869 opt->length = data_len / 4;
874 /* If no mask has been prodived we assume an exact match. */
876 return sizeof(struct geneve_opt) + data_len;
878 if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_GENEVE) {
879 NL_SET_ERR_MSG(extack, "Non-geneve option type for mask");
883 err = nla_parse_nested_deprecated(tb,
884 TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX,
885 nla, geneve_opt_policy, extack);
889 /* We are not allowed to omit any of CLASS, TYPE or DATA
890 * fields from the key.
893 (!tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS] ||
894 !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE] ||
895 !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA])) {
896 NL_SET_ERR_MSG(extack, "Missing tunnel key geneve option class, type or data");
900 /* Omitting any of CLASS, TYPE or DATA fields is allowed
903 if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA]) {
904 int new_len = key->enc_opts.len;
906 data = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA];
907 data_len = nla_len(data);
909 NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is less than 4 bytes long");
913 NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is not a multiple of 4 bytes long");
917 new_len += sizeof(struct geneve_opt) + data_len;
918 BUILD_BUG_ON(FLOW_DIS_TUN_OPTS_MAX != IP_TUNNEL_OPTS_MAX);
919 if (new_len > FLOW_DIS_TUN_OPTS_MAX) {
920 NL_SET_ERR_MSG(extack, "Tunnel options exceeds max size");
923 opt->length = data_len / 4;
924 memcpy(opt->opt_data, nla_data(data), data_len);
927 if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS]) {
928 class = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS];
929 opt->opt_class = nla_get_be16(class);
932 if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE]) {
933 type = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE];
934 opt->type = nla_get_u8(type);
937 return sizeof(struct geneve_opt) + data_len;
940 static int fl_set_enc_opt(struct nlattr **tb, struct fl_flow_key *key,
941 struct fl_flow_key *mask,
942 struct netlink_ext_ack *extack)
944 const struct nlattr *nla_enc_key, *nla_opt_key, *nla_opt_msk = NULL;
945 int err, option_len, key_depth, msk_depth = 0;
947 err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS],
948 TCA_FLOWER_KEY_ENC_OPTS_MAX,
949 enc_opts_policy, extack);
953 nla_enc_key = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS]);
955 if (tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]) {
956 err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK],
957 TCA_FLOWER_KEY_ENC_OPTS_MAX,
958 enc_opts_policy, extack);
962 nla_opt_msk = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
963 msk_depth = nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
966 nla_for_each_attr(nla_opt_key, nla_enc_key,
967 nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS]), key_depth) {
968 switch (nla_type(nla_opt_key)) {
969 case TCA_FLOWER_KEY_ENC_OPTS_GENEVE:
971 key->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT;
972 option_len = fl_set_geneve_opt(nla_opt_key, key,
973 key_depth, option_len,
978 key->enc_opts.len += option_len;
979 /* At the same time we need to parse through the mask
980 * in order to verify exact and mask attribute lengths.
982 mask->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT;
983 option_len = fl_set_geneve_opt(nla_opt_msk, mask,
984 msk_depth, option_len,
989 mask->enc_opts.len += option_len;
990 if (key->enc_opts.len != mask->enc_opts.len) {
991 NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
996 nla_opt_msk = nla_next(nla_opt_msk, &msk_depth);
999 NL_SET_ERR_MSG(extack, "Unknown tunnel option type");
1007 static int fl_set_key(struct net *net, struct nlattr **tb,
1008 struct fl_flow_key *key, struct fl_flow_key *mask,
1009 struct netlink_ext_ack *extack)
1013 #ifdef CONFIG_NET_CLS_IND
1014 if (tb[TCA_FLOWER_INDEV]) {
1015 int err = tcf_change_indev(net, tb[TCA_FLOWER_INDEV], extack);
1018 key->indev_ifindex = err;
1019 mask->indev_ifindex = 0xffffffff;
1023 fl_set_key_val(tb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
1024 mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
1025 sizeof(key->eth.dst));
1026 fl_set_key_val(tb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
1027 mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
1028 sizeof(key->eth.src));
1030 if (tb[TCA_FLOWER_KEY_ETH_TYPE]) {
1031 ethertype = nla_get_be16(tb[TCA_FLOWER_KEY_ETH_TYPE]);
1033 if (eth_type_vlan(ethertype)) {
1034 fl_set_key_vlan(tb, ethertype, TCA_FLOWER_KEY_VLAN_ID,
1035 TCA_FLOWER_KEY_VLAN_PRIO, &key->vlan,
1038 if (tb[TCA_FLOWER_KEY_VLAN_ETH_TYPE]) {
1039 ethertype = nla_get_be16(tb[TCA_FLOWER_KEY_VLAN_ETH_TYPE]);
1040 if (eth_type_vlan(ethertype)) {
1041 fl_set_key_vlan(tb, ethertype,
1042 TCA_FLOWER_KEY_CVLAN_ID,
1043 TCA_FLOWER_KEY_CVLAN_PRIO,
1044 &key->cvlan, &mask->cvlan);
1045 fl_set_key_val(tb, &key->basic.n_proto,
1046 TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
1047 &mask->basic.n_proto,
1049 sizeof(key->basic.n_proto));
1051 key->basic.n_proto = ethertype;
1052 mask->basic.n_proto = cpu_to_be16(~0);
1056 key->basic.n_proto = ethertype;
1057 mask->basic.n_proto = cpu_to_be16(~0);
1061 if (key->basic.n_proto == htons(ETH_P_IP) ||
1062 key->basic.n_proto == htons(ETH_P_IPV6)) {
1063 fl_set_key_val(tb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
1064 &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
1065 sizeof(key->basic.ip_proto));
1066 fl_set_key_ip(tb, false, &key->ip, &mask->ip);
1069 if (tb[TCA_FLOWER_KEY_IPV4_SRC] || tb[TCA_FLOWER_KEY_IPV4_DST]) {
1070 key->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1071 mask->control.addr_type = ~0;
1072 fl_set_key_val(tb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
1073 &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
1074 sizeof(key->ipv4.src));
1075 fl_set_key_val(tb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
1076 &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
1077 sizeof(key->ipv4.dst));
1078 } else if (tb[TCA_FLOWER_KEY_IPV6_SRC] || tb[TCA_FLOWER_KEY_IPV6_DST]) {
1079 key->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1080 mask->control.addr_type = ~0;
1081 fl_set_key_val(tb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
1082 &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
1083 sizeof(key->ipv6.src));
1084 fl_set_key_val(tb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
1085 &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
1086 sizeof(key->ipv6.dst));
1089 if (key->basic.ip_proto == IPPROTO_TCP) {
1090 fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
1091 &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
1092 sizeof(key->tp.src));
1093 fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
1094 &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
1095 sizeof(key->tp.dst));
1096 fl_set_key_val(tb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
1097 &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
1098 sizeof(key->tcp.flags));
1099 } else if (key->basic.ip_proto == IPPROTO_UDP) {
1100 fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
1101 &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
1102 sizeof(key->tp.src));
1103 fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
1104 &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
1105 sizeof(key->tp.dst));
1106 } else if (key->basic.ip_proto == IPPROTO_SCTP) {
1107 fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
1108 &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
1109 sizeof(key->tp.src));
1110 fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
1111 &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
1112 sizeof(key->tp.dst));
1113 } else if (key->basic.n_proto == htons(ETH_P_IP) &&
1114 key->basic.ip_proto == IPPROTO_ICMP) {
1115 fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV4_TYPE,
1117 TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
1118 sizeof(key->icmp.type));
1119 fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV4_CODE,
1121 TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
1122 sizeof(key->icmp.code));
1123 } else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
1124 key->basic.ip_proto == IPPROTO_ICMPV6) {
1125 fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV6_TYPE,
1127 TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
1128 sizeof(key->icmp.type));
1129 fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV6_CODE,
1131 TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
1132 sizeof(key->icmp.code));
1133 } else if (key->basic.n_proto == htons(ETH_P_MPLS_UC) ||
1134 key->basic.n_proto == htons(ETH_P_MPLS_MC)) {
1135 ret = fl_set_key_mpls(tb, &key->mpls, &mask->mpls);
1138 } else if (key->basic.n_proto == htons(ETH_P_ARP) ||
1139 key->basic.n_proto == htons(ETH_P_RARP)) {
1140 fl_set_key_val(tb, &key->arp.sip, TCA_FLOWER_KEY_ARP_SIP,
1141 &mask->arp.sip, TCA_FLOWER_KEY_ARP_SIP_MASK,
1142 sizeof(key->arp.sip));
1143 fl_set_key_val(tb, &key->arp.tip, TCA_FLOWER_KEY_ARP_TIP,
1144 &mask->arp.tip, TCA_FLOWER_KEY_ARP_TIP_MASK,
1145 sizeof(key->arp.tip));
1146 fl_set_key_val(tb, &key->arp.op, TCA_FLOWER_KEY_ARP_OP,
1147 &mask->arp.op, TCA_FLOWER_KEY_ARP_OP_MASK,
1148 sizeof(key->arp.op));
1149 fl_set_key_val(tb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
1150 mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
1151 sizeof(key->arp.sha));
1152 fl_set_key_val(tb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
1153 mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
1154 sizeof(key->arp.tha));
1157 if (key->basic.ip_proto == IPPROTO_TCP ||
1158 key->basic.ip_proto == IPPROTO_UDP ||
1159 key->basic.ip_proto == IPPROTO_SCTP) {
1160 ret = fl_set_key_port_range(tb, key, mask);
1165 if (tb[TCA_FLOWER_KEY_ENC_IPV4_SRC] ||
1166 tb[TCA_FLOWER_KEY_ENC_IPV4_DST]) {
1167 key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1168 mask->enc_control.addr_type = ~0;
1169 fl_set_key_val(tb, &key->enc_ipv4.src,
1170 TCA_FLOWER_KEY_ENC_IPV4_SRC,
1171 &mask->enc_ipv4.src,
1172 TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
1173 sizeof(key->enc_ipv4.src));
1174 fl_set_key_val(tb, &key->enc_ipv4.dst,
1175 TCA_FLOWER_KEY_ENC_IPV4_DST,
1176 &mask->enc_ipv4.dst,
1177 TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
1178 sizeof(key->enc_ipv4.dst));
1181 if (tb[TCA_FLOWER_KEY_ENC_IPV6_SRC] ||
1182 tb[TCA_FLOWER_KEY_ENC_IPV6_DST]) {
1183 key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1184 mask->enc_control.addr_type = ~0;
1185 fl_set_key_val(tb, &key->enc_ipv6.src,
1186 TCA_FLOWER_KEY_ENC_IPV6_SRC,
1187 &mask->enc_ipv6.src,
1188 TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
1189 sizeof(key->enc_ipv6.src));
1190 fl_set_key_val(tb, &key->enc_ipv6.dst,
1191 TCA_FLOWER_KEY_ENC_IPV6_DST,
1192 &mask->enc_ipv6.dst,
1193 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
1194 sizeof(key->enc_ipv6.dst));
1197 fl_set_key_val(tb, &key->enc_key_id.keyid, TCA_FLOWER_KEY_ENC_KEY_ID,
1198 &mask->enc_key_id.keyid, TCA_FLOWER_UNSPEC,
1199 sizeof(key->enc_key_id.keyid));
1201 fl_set_key_val(tb, &key->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
1202 &mask->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
1203 sizeof(key->enc_tp.src));
1205 fl_set_key_val(tb, &key->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
1206 &mask->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
1207 sizeof(key->enc_tp.dst));
1209 fl_set_key_ip(tb, true, &key->enc_ip, &mask->enc_ip);
1211 if (tb[TCA_FLOWER_KEY_ENC_OPTS]) {
1212 ret = fl_set_enc_opt(tb, key, mask, extack);
1217 if (tb[TCA_FLOWER_KEY_FLAGS])
1218 ret = fl_set_key_flags(tb, &key->control.flags, &mask->control.flags);
1223 static void fl_mask_copy(struct fl_flow_mask *dst,
1224 struct fl_flow_mask *src)
1226 const void *psrc = fl_key_get_start(&src->key, src);
1227 void *pdst = fl_key_get_start(&dst->key, src);
1229 memcpy(pdst, psrc, fl_mask_range(src));
1230 dst->range = src->range;
1233 static const struct rhashtable_params fl_ht_params = {
1234 .key_offset = offsetof(struct cls_fl_filter, mkey), /* base offset */
1235 .head_offset = offsetof(struct cls_fl_filter, ht_node),
1236 .automatic_shrinking = true,
1239 static int fl_init_mask_hashtable(struct fl_flow_mask *mask)
1241 mask->filter_ht_params = fl_ht_params;
1242 mask->filter_ht_params.key_len = fl_mask_range(mask);
1243 mask->filter_ht_params.key_offset += mask->range.start;
1245 return rhashtable_init(&mask->ht, &mask->filter_ht_params);
1248 #define FL_KEY_MEMBER_OFFSET(member) offsetof(struct fl_flow_key, member)
1249 #define FL_KEY_MEMBER_SIZE(member) FIELD_SIZEOF(struct fl_flow_key, member)
1251 #define FL_KEY_IS_MASKED(mask, member) \
1252 memchr_inv(((char *)mask) + FL_KEY_MEMBER_OFFSET(member), \
1253 0, FL_KEY_MEMBER_SIZE(member)) \
1255 #define FL_KEY_SET(keys, cnt, id, member) \
1257 keys[cnt].key_id = id; \
1258 keys[cnt].offset = FL_KEY_MEMBER_OFFSET(member); \
1262 #define FL_KEY_SET_IF_MASKED(mask, keys, cnt, id, member) \
1264 if (FL_KEY_IS_MASKED(mask, member)) \
1265 FL_KEY_SET(keys, cnt, id, member); \
1268 static void fl_init_dissector(struct flow_dissector *dissector,
1269 struct fl_flow_key *mask)
1271 struct flow_dissector_key keys[FLOW_DISSECTOR_KEY_MAX];
1274 FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_CONTROL, control);
1275 FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_BASIC, basic);
1276 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1277 FLOW_DISSECTOR_KEY_ETH_ADDRS, eth);
1278 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1279 FLOW_DISSECTOR_KEY_IPV4_ADDRS, ipv4);
1280 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1281 FLOW_DISSECTOR_KEY_IPV6_ADDRS, ipv6);
1282 if (FL_KEY_IS_MASKED(mask, tp) ||
1283 FL_KEY_IS_MASKED(mask, tp_min) || FL_KEY_IS_MASKED(mask, tp_max))
1284 FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_PORTS, tp);
1285 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1286 FLOW_DISSECTOR_KEY_IP, ip);
1287 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1288 FLOW_DISSECTOR_KEY_TCP, tcp);
1289 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1290 FLOW_DISSECTOR_KEY_ICMP, icmp);
1291 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1292 FLOW_DISSECTOR_KEY_ARP, arp);
1293 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1294 FLOW_DISSECTOR_KEY_MPLS, mpls);
1295 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1296 FLOW_DISSECTOR_KEY_VLAN, vlan);
1297 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1298 FLOW_DISSECTOR_KEY_CVLAN, cvlan);
1299 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1300 FLOW_DISSECTOR_KEY_ENC_KEYID, enc_key_id);
1301 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1302 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, enc_ipv4);
1303 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1304 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, enc_ipv6);
1305 if (FL_KEY_IS_MASKED(mask, enc_ipv4) ||
1306 FL_KEY_IS_MASKED(mask, enc_ipv6))
1307 FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_ENC_CONTROL,
1309 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1310 FLOW_DISSECTOR_KEY_ENC_PORTS, enc_tp);
1311 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1312 FLOW_DISSECTOR_KEY_ENC_IP, enc_ip);
1313 FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1314 FLOW_DISSECTOR_KEY_ENC_OPTS, enc_opts);
1316 skb_flow_dissector_init(dissector, keys, cnt);
1319 static struct fl_flow_mask *fl_create_new_mask(struct cls_fl_head *head,
1320 struct fl_flow_mask *mask)
1322 struct fl_flow_mask *newmask;
1325 newmask = kzalloc(sizeof(*newmask), GFP_KERNEL);
1327 return ERR_PTR(-ENOMEM);
1329 fl_mask_copy(newmask, mask);
1331 if ((newmask->key.tp_min.dst && newmask->key.tp_max.dst) ||
1332 (newmask->key.tp_min.src && newmask->key.tp_max.src))
1333 newmask->flags |= TCA_FLOWER_MASK_FLAGS_RANGE;
1335 err = fl_init_mask_hashtable(newmask);
1339 fl_init_dissector(&newmask->dissector, &newmask->key);
1341 INIT_LIST_HEAD_RCU(&newmask->filters);
1343 refcount_set(&newmask->refcnt, 1);
1344 err = rhashtable_replace_fast(&head->ht, &mask->ht_node,
1345 &newmask->ht_node, mask_ht_params);
1347 goto errout_destroy;
1349 /* Wait until any potential concurrent users of mask are finished */
1352 spin_lock(&head->masks_lock);
1353 list_add_tail_rcu(&newmask->list, &head->masks);
1354 spin_unlock(&head->masks_lock);
1359 rhashtable_destroy(&newmask->ht);
1363 return ERR_PTR(err);
1366 static int fl_check_assign_mask(struct cls_fl_head *head,
1367 struct cls_fl_filter *fnew,
1368 struct cls_fl_filter *fold,
1369 struct fl_flow_mask *mask)
1371 struct fl_flow_mask *newmask;
1376 /* Insert mask as temporary node to prevent concurrent creation of mask
1377 * with same key. Any concurrent lookups with same key will return
1378 * -EAGAIN because mask's refcnt is zero. It is safe to insert
1379 * stack-allocated 'mask' to masks hash table because we call
1380 * synchronize_rcu() before returning from this function (either in case
1381 * of error or after replacing it with heap-allocated mask in
1382 * fl_create_new_mask()).
1384 fnew->mask = rhashtable_lookup_get_insert_fast(&head->ht,
1392 goto errout_cleanup;
1395 newmask = fl_create_new_mask(head, mask);
1396 if (IS_ERR(newmask)) {
1397 ret = PTR_ERR(newmask);
1398 goto errout_cleanup;
1401 fnew->mask = newmask;
1403 } else if (IS_ERR(fnew->mask)) {
1404 ret = PTR_ERR(fnew->mask);
1405 } else if (fold && fold->mask != fnew->mask) {
1407 } else if (!refcount_inc_not_zero(&fnew->mask->refcnt)) {
1408 /* Mask was deleted concurrently, try again */
1415 rhashtable_remove_fast(&head->ht, &mask->ht_node,
1417 /* Wait until any potential concurrent users of mask are finished */
1422 static int fl_set_parms(struct net *net, struct tcf_proto *tp,
1423 struct cls_fl_filter *f, struct fl_flow_mask *mask,
1424 unsigned long base, struct nlattr **tb,
1425 struct nlattr *est, bool ovr,
1426 struct fl_flow_tmplt *tmplt, bool rtnl_held,
1427 struct netlink_ext_ack *extack)
1431 err = tcf_exts_validate(net, tp, tb, est, &f->exts, ovr, rtnl_held,
1436 if (tb[TCA_FLOWER_CLASSID]) {
1437 f->res.classid = nla_get_u32(tb[TCA_FLOWER_CLASSID]);
1440 tcf_bind_filter(tp, &f->res, base);
1445 err = fl_set_key(net, tb, &f->key, &mask->key, extack);
1449 fl_mask_update_range(mask);
1450 fl_set_masked_key(&f->mkey, &f->key, mask);
1452 if (!fl_mask_fits_tmplt(tmplt, mask)) {
1453 NL_SET_ERR_MSG_MOD(extack, "Mask does not fit the template");
1460 static int fl_ht_insert_unique(struct cls_fl_filter *fnew,
1461 struct cls_fl_filter *fold,
1464 struct fl_flow_mask *mask = fnew->mask;
1467 err = rhashtable_lookup_insert_fast(&mask->ht,
1469 mask->filter_ht_params);
1472 /* It is okay if filter with same key exists when
1475 return fold && err == -EEXIST ? 0 : err;
1482 static int fl_change(struct net *net, struct sk_buff *in_skb,
1483 struct tcf_proto *tp, unsigned long base,
1484 u32 handle, struct nlattr **tca,
1485 void **arg, bool ovr, bool rtnl_held,
1486 struct netlink_ext_ack *extack)
1488 struct cls_fl_head *head = fl_head_dereference(tp);
1489 struct cls_fl_filter *fold = *arg;
1490 struct cls_fl_filter *fnew;
1491 struct fl_flow_mask *mask;
1496 if (!tca[TCA_OPTIONS]) {
1501 mask = kzalloc(sizeof(struct fl_flow_mask), GFP_KERNEL);
1507 tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
1510 goto errout_mask_alloc;
1513 err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
1514 tca[TCA_OPTIONS], fl_policy, NULL);
1518 if (fold && handle && fold->handle != handle) {
1523 fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
1528 INIT_LIST_HEAD(&fnew->hw_list);
1529 refcount_set(&fnew->refcnt, 1);
1531 err = tcf_exts_init(&fnew->exts, net, TCA_FLOWER_ACT, 0);
1535 if (tb[TCA_FLOWER_FLAGS]) {
1536 fnew->flags = nla_get_u32(tb[TCA_FLOWER_FLAGS]);
1538 if (!tc_flags_valid(fnew->flags)) {
1544 err = fl_set_parms(net, tp, fnew, mask, base, tb, tca[TCA_RATE], ovr,
1545 tp->chain->tmplt_priv, rtnl_held, extack);
1549 err = fl_check_assign_mask(head, fnew, fold, mask);
1553 err = fl_ht_insert_unique(fnew, fold, &in_ht);
1557 if (!tc_skip_hw(fnew->flags)) {
1558 err = fl_hw_replace_filter(tp, fnew, rtnl_held, extack);
1563 if (!tc_in_hw(fnew->flags))
1564 fnew->flags |= TCA_CLS_FLAGS_NOT_IN_HW;
1566 spin_lock(&tp->lock);
1568 /* tp was deleted concurrently. -EAGAIN will cause caller to lookup
1569 * proto again or create new one, if necessary.
1577 /* Fold filter was deleted concurrently. Retry lookup. */
1578 if (fold->deleted) {
1583 fnew->handle = handle;
1586 struct rhashtable_params params =
1587 fnew->mask->filter_ht_params;
1589 err = rhashtable_insert_fast(&fnew->mask->ht,
1597 refcount_inc(&fnew->refcnt);
1598 rhashtable_remove_fast(&fold->mask->ht,
1600 fold->mask->filter_ht_params);
1601 idr_replace(&head->handle_idr, fnew, fnew->handle);
1602 list_replace_rcu(&fold->list, &fnew->list);
1603 fold->deleted = true;
1605 spin_unlock(&tp->lock);
1607 fl_mask_put(head, fold->mask);
1608 if (!tc_skip_hw(fold->flags))
1609 fl_hw_destroy_filter(tp, fold, rtnl_held, NULL);
1610 tcf_unbind_filter(tp, &fold->res);
1611 /* Caller holds reference to fold, so refcnt is always > 0
1614 refcount_dec(&fold->refcnt);
1618 /* user specifies a handle and it doesn't exist */
1619 err = idr_alloc_u32(&head->handle_idr, fnew, &handle,
1620 handle, GFP_ATOMIC);
1622 /* Filter with specified handle was concurrently
1623 * inserted after initial check in cls_api. This is not
1624 * necessarily an error if NLM_F_EXCL is not set in
1625 * message flags. Returning EAGAIN will cause cls_api to
1626 * try to update concurrently inserted rule.
1632 err = idr_alloc_u32(&head->handle_idr, fnew, &handle,
1633 INT_MAX, GFP_ATOMIC);
1638 refcount_inc(&fnew->refcnt);
1639 fnew->handle = handle;
1640 list_add_tail_rcu(&fnew->list, &fnew->mask->filters);
1641 spin_unlock(&tp->lock);
1651 spin_lock(&tp->lock);
1653 fnew->deleted = true;
1654 spin_unlock(&tp->lock);
1655 if (!tc_skip_hw(fnew->flags))
1656 fl_hw_destroy_filter(tp, fnew, rtnl_held, NULL);
1658 rhashtable_remove_fast(&fnew->mask->ht, &fnew->ht_node,
1659 fnew->mask->filter_ht_params);
1661 fl_mask_put(head, fnew->mask);
1674 static int fl_delete(struct tcf_proto *tp, void *arg, bool *last,
1675 bool rtnl_held, struct netlink_ext_ack *extack)
1677 struct cls_fl_head *head = fl_head_dereference(tp);
1678 struct cls_fl_filter *f = arg;
1682 err = __fl_delete(tp, f, &last_on_mask, rtnl_held, extack);
1683 *last = list_empty(&head->masks);
1689 static void fl_walk(struct tcf_proto *tp, struct tcf_walker *arg,
1692 struct cls_fl_filter *f;
1694 arg->count = arg->skip;
1696 while ((f = fl_get_next_filter(tp, &arg->cookie)) != NULL) {
1697 if (arg->fn(tp, f, arg) < 0) {
1708 static struct cls_fl_filter *
1709 fl_get_next_hw_filter(struct tcf_proto *tp, struct cls_fl_filter *f, bool add)
1711 struct cls_fl_head *head = fl_head_dereference(tp);
1713 spin_lock(&tp->lock);
1714 if (list_empty(&head->hw_filters)) {
1715 spin_unlock(&tp->lock);
1720 f = list_entry(&head->hw_filters, struct cls_fl_filter,
1722 list_for_each_entry_continue(f, &head->hw_filters, hw_list) {
1723 if (!(add && f->deleted) && refcount_inc_not_zero(&f->refcnt)) {
1724 spin_unlock(&tp->lock);
1729 spin_unlock(&tp->lock);
1733 static int fl_reoffload(struct tcf_proto *tp, bool add, tc_setup_cb_t *cb,
1734 void *cb_priv, struct netlink_ext_ack *extack)
1736 struct tc_cls_flower_offload cls_flower = {};
1737 struct tcf_block *block = tp->chain->block;
1738 struct cls_fl_filter *f = NULL;
1741 /* hw_filters list can only be changed by hw offload functions after
1742 * obtaining rtnl lock. Make sure it is not changed while reoffload is
1747 while ((f = fl_get_next_hw_filter(tp, f, add))) {
1749 flow_rule_alloc(tcf_exts_num_actions(&f->exts));
1750 if (!cls_flower.rule) {
1755 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags,
1757 cls_flower.command = add ?
1758 TC_CLSFLOWER_REPLACE : TC_CLSFLOWER_DESTROY;
1759 cls_flower.cookie = (unsigned long)f;
1760 cls_flower.rule->match.dissector = &f->mask->dissector;
1761 cls_flower.rule->match.mask = &f->mask->key;
1762 cls_flower.rule->match.key = &f->mkey;
1764 err = tc_setup_flow_action(&cls_flower.rule->action, &f->exts);
1766 kfree(cls_flower.rule);
1767 if (tc_skip_sw(f->flags)) {
1768 NL_SET_ERR_MSG_MOD(extack, "Failed to setup flow action");
1775 cls_flower.classid = f->res.classid;
1777 err = cb(TC_SETUP_CLSFLOWER, &cls_flower, cb_priv);
1778 kfree(cls_flower.rule);
1781 if (add && tc_skip_sw(f->flags)) {
1788 spin_lock(&tp->lock);
1789 tc_cls_offload_cnt_update(block, &f->in_hw_count, &f->flags,
1791 spin_unlock(&tp->lock);
1799 static int fl_hw_create_tmplt(struct tcf_chain *chain,
1800 struct fl_flow_tmplt *tmplt)
1802 struct tc_cls_flower_offload cls_flower = {};
1803 struct tcf_block *block = chain->block;
1805 cls_flower.rule = flow_rule_alloc(0);
1806 if (!cls_flower.rule)
1809 cls_flower.common.chain_index = chain->index;
1810 cls_flower.command = TC_CLSFLOWER_TMPLT_CREATE;
1811 cls_flower.cookie = (unsigned long) tmplt;
1812 cls_flower.rule->match.dissector = &tmplt->dissector;
1813 cls_flower.rule->match.mask = &tmplt->mask;
1814 cls_flower.rule->match.key = &tmplt->dummy_key;
1816 /* We don't care if driver (any of them) fails to handle this
1817 * call. It serves just as a hint for it.
1819 tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false);
1820 kfree(cls_flower.rule);
1825 static void fl_hw_destroy_tmplt(struct tcf_chain *chain,
1826 struct fl_flow_tmplt *tmplt)
1828 struct tc_cls_flower_offload cls_flower = {};
1829 struct tcf_block *block = chain->block;
1831 cls_flower.common.chain_index = chain->index;
1832 cls_flower.command = TC_CLSFLOWER_TMPLT_DESTROY;
1833 cls_flower.cookie = (unsigned long) tmplt;
1835 tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false);
1838 static void *fl_tmplt_create(struct net *net, struct tcf_chain *chain,
1839 struct nlattr **tca,
1840 struct netlink_ext_ack *extack)
1842 struct fl_flow_tmplt *tmplt;
1846 if (!tca[TCA_OPTIONS])
1847 return ERR_PTR(-EINVAL);
1849 tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
1851 return ERR_PTR(-ENOBUFS);
1852 err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
1853 tca[TCA_OPTIONS], fl_policy, NULL);
1857 tmplt = kzalloc(sizeof(*tmplt), GFP_KERNEL);
1862 tmplt->chain = chain;
1863 err = fl_set_key(net, tb, &tmplt->dummy_key, &tmplt->mask, extack);
1867 fl_init_dissector(&tmplt->dissector, &tmplt->mask);
1869 err = fl_hw_create_tmplt(chain, tmplt);
1880 return ERR_PTR(err);
1883 static void fl_tmplt_destroy(void *tmplt_priv)
1885 struct fl_flow_tmplt *tmplt = tmplt_priv;
1887 fl_hw_destroy_tmplt(tmplt->chain, tmplt);
1891 static int fl_dump_key_val(struct sk_buff *skb,
1892 void *val, int val_type,
1893 void *mask, int mask_type, int len)
1897 if (!memchr_inv(mask, 0, len))
1899 err = nla_put(skb, val_type, len, val);
1902 if (mask_type != TCA_FLOWER_UNSPEC) {
1903 err = nla_put(skb, mask_type, len, mask);
1910 static int fl_dump_key_port_range(struct sk_buff *skb, struct fl_flow_key *key,
1911 struct fl_flow_key *mask)
1913 if (fl_dump_key_val(skb, &key->tp_min.dst, TCA_FLOWER_KEY_PORT_DST_MIN,
1914 &mask->tp_min.dst, TCA_FLOWER_UNSPEC,
1915 sizeof(key->tp_min.dst)) ||
1916 fl_dump_key_val(skb, &key->tp_max.dst, TCA_FLOWER_KEY_PORT_DST_MAX,
1917 &mask->tp_max.dst, TCA_FLOWER_UNSPEC,
1918 sizeof(key->tp_max.dst)) ||
1919 fl_dump_key_val(skb, &key->tp_min.src, TCA_FLOWER_KEY_PORT_SRC_MIN,
1920 &mask->tp_min.src, TCA_FLOWER_UNSPEC,
1921 sizeof(key->tp_min.src)) ||
1922 fl_dump_key_val(skb, &key->tp_max.src, TCA_FLOWER_KEY_PORT_SRC_MAX,
1923 &mask->tp_max.src, TCA_FLOWER_UNSPEC,
1924 sizeof(key->tp_max.src)))
1930 static int fl_dump_key_mpls(struct sk_buff *skb,
1931 struct flow_dissector_key_mpls *mpls_key,
1932 struct flow_dissector_key_mpls *mpls_mask)
1936 if (!memchr_inv(mpls_mask, 0, sizeof(*mpls_mask)))
1938 if (mpls_mask->mpls_ttl) {
1939 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TTL,
1940 mpls_key->mpls_ttl);
1944 if (mpls_mask->mpls_tc) {
1945 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TC,
1950 if (mpls_mask->mpls_label) {
1951 err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_LABEL,
1952 mpls_key->mpls_label);
1956 if (mpls_mask->mpls_bos) {
1957 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_BOS,
1958 mpls_key->mpls_bos);
1965 static int fl_dump_key_ip(struct sk_buff *skb, bool encap,
1966 struct flow_dissector_key_ip *key,
1967 struct flow_dissector_key_ip *mask)
1969 int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
1970 int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
1971 int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
1972 int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
1974 if (fl_dump_key_val(skb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos)) ||
1975 fl_dump_key_val(skb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl)))
1981 static int fl_dump_key_vlan(struct sk_buff *skb,
1982 int vlan_id_key, int vlan_prio_key,
1983 struct flow_dissector_key_vlan *vlan_key,
1984 struct flow_dissector_key_vlan *vlan_mask)
1988 if (!memchr_inv(vlan_mask, 0, sizeof(*vlan_mask)))
1990 if (vlan_mask->vlan_id) {
1991 err = nla_put_u16(skb, vlan_id_key,
1996 if (vlan_mask->vlan_priority) {
1997 err = nla_put_u8(skb, vlan_prio_key,
1998 vlan_key->vlan_priority);
2005 static void fl_get_key_flag(u32 dissector_key, u32 dissector_mask,
2006 u32 *flower_key, u32 *flower_mask,
2007 u32 flower_flag_bit, u32 dissector_flag_bit)
2009 if (dissector_mask & dissector_flag_bit) {
2010 *flower_mask |= flower_flag_bit;
2011 if (dissector_key & dissector_flag_bit)
2012 *flower_key |= flower_flag_bit;
2016 static int fl_dump_key_flags(struct sk_buff *skb, u32 flags_key, u32 flags_mask)
2022 if (!memchr_inv(&flags_mask, 0, sizeof(flags_mask)))
2028 fl_get_key_flag(flags_key, flags_mask, &key, &mask,
2029 TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
2030 fl_get_key_flag(flags_key, flags_mask, &key, &mask,
2031 TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
2032 FLOW_DIS_FIRST_FRAG);
2034 _key = cpu_to_be32(key);
2035 _mask = cpu_to_be32(mask);
2037 err = nla_put(skb, TCA_FLOWER_KEY_FLAGS, 4, &_key);
2041 return nla_put(skb, TCA_FLOWER_KEY_FLAGS_MASK, 4, &_mask);
2044 static int fl_dump_key_geneve_opt(struct sk_buff *skb,
2045 struct flow_dissector_key_enc_opts *enc_opts)
2047 struct geneve_opt *opt;
2048 struct nlattr *nest;
2051 nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_GENEVE);
2053 goto nla_put_failure;
2055 while (enc_opts->len > opt_off) {
2056 opt = (struct geneve_opt *)&enc_opts->data[opt_off];
2058 if (nla_put_be16(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS,
2060 goto nla_put_failure;
2061 if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE,
2063 goto nla_put_failure;
2064 if (nla_put(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA,
2065 opt->length * 4, opt->opt_data))
2066 goto nla_put_failure;
2068 opt_off += sizeof(struct geneve_opt) + opt->length * 4;
2070 nla_nest_end(skb, nest);
2074 nla_nest_cancel(skb, nest);
2078 static int fl_dump_key_options(struct sk_buff *skb, int enc_opt_type,
2079 struct flow_dissector_key_enc_opts *enc_opts)
2081 struct nlattr *nest;
2087 nest = nla_nest_start_noflag(skb, enc_opt_type);
2089 goto nla_put_failure;
2091 switch (enc_opts->dst_opt_type) {
2092 case TUNNEL_GENEVE_OPT:
2093 err = fl_dump_key_geneve_opt(skb, enc_opts);
2095 goto nla_put_failure;
2098 goto nla_put_failure;
2100 nla_nest_end(skb, nest);
2104 nla_nest_cancel(skb, nest);
2108 static int fl_dump_key_enc_opt(struct sk_buff *skb,
2109 struct flow_dissector_key_enc_opts *key_opts,
2110 struct flow_dissector_key_enc_opts *msk_opts)
2114 err = fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS, key_opts);
2118 return fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS_MASK, msk_opts);
2121 static int fl_dump_key(struct sk_buff *skb, struct net *net,
2122 struct fl_flow_key *key, struct fl_flow_key *mask)
2124 if (mask->indev_ifindex) {
2125 struct net_device *dev;
2127 dev = __dev_get_by_index(net, key->indev_ifindex);
2128 if (dev && nla_put_string(skb, TCA_FLOWER_INDEV, dev->name))
2129 goto nla_put_failure;
2132 if (fl_dump_key_val(skb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
2133 mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
2134 sizeof(key->eth.dst)) ||
2135 fl_dump_key_val(skb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
2136 mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
2137 sizeof(key->eth.src)) ||
2138 fl_dump_key_val(skb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE,
2139 &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
2140 sizeof(key->basic.n_proto)))
2141 goto nla_put_failure;
2143 if (fl_dump_key_mpls(skb, &key->mpls, &mask->mpls))
2144 goto nla_put_failure;
2146 if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_VLAN_ID,
2147 TCA_FLOWER_KEY_VLAN_PRIO, &key->vlan, &mask->vlan))
2148 goto nla_put_failure;
2150 if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_CVLAN_ID,
2151 TCA_FLOWER_KEY_CVLAN_PRIO,
2152 &key->cvlan, &mask->cvlan) ||
2153 (mask->cvlan.vlan_tpid &&
2154 nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
2155 key->cvlan.vlan_tpid)))
2156 goto nla_put_failure;
2158 if (mask->basic.n_proto) {
2159 if (mask->cvlan.vlan_tpid) {
2160 if (nla_put_be16(skb, TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
2161 key->basic.n_proto))
2162 goto nla_put_failure;
2163 } else if (mask->vlan.vlan_tpid) {
2164 if (nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
2165 key->basic.n_proto))
2166 goto nla_put_failure;
2170 if ((key->basic.n_proto == htons(ETH_P_IP) ||
2171 key->basic.n_proto == htons(ETH_P_IPV6)) &&
2172 (fl_dump_key_val(skb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
2173 &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
2174 sizeof(key->basic.ip_proto)) ||
2175 fl_dump_key_ip(skb, false, &key->ip, &mask->ip)))
2176 goto nla_put_failure;
2178 if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
2179 (fl_dump_key_val(skb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
2180 &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
2181 sizeof(key->ipv4.src)) ||
2182 fl_dump_key_val(skb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
2183 &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
2184 sizeof(key->ipv4.dst))))
2185 goto nla_put_failure;
2186 else if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
2187 (fl_dump_key_val(skb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
2188 &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
2189 sizeof(key->ipv6.src)) ||
2190 fl_dump_key_val(skb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
2191 &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
2192 sizeof(key->ipv6.dst))))
2193 goto nla_put_failure;
2195 if (key->basic.ip_proto == IPPROTO_TCP &&
2196 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
2197 &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
2198 sizeof(key->tp.src)) ||
2199 fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
2200 &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
2201 sizeof(key->tp.dst)) ||
2202 fl_dump_key_val(skb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
2203 &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
2204 sizeof(key->tcp.flags))))
2205 goto nla_put_failure;
2206 else if (key->basic.ip_proto == IPPROTO_UDP &&
2207 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
2208 &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
2209 sizeof(key->tp.src)) ||
2210 fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
2211 &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
2212 sizeof(key->tp.dst))))
2213 goto nla_put_failure;
2214 else if (key->basic.ip_proto == IPPROTO_SCTP &&
2215 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
2216 &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
2217 sizeof(key->tp.src)) ||
2218 fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
2219 &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
2220 sizeof(key->tp.dst))))
2221 goto nla_put_failure;
2222 else if (key->basic.n_proto == htons(ETH_P_IP) &&
2223 key->basic.ip_proto == IPPROTO_ICMP &&
2224 (fl_dump_key_val(skb, &key->icmp.type,
2225 TCA_FLOWER_KEY_ICMPV4_TYPE, &mask->icmp.type,
2226 TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
2227 sizeof(key->icmp.type)) ||
2228 fl_dump_key_val(skb, &key->icmp.code,
2229 TCA_FLOWER_KEY_ICMPV4_CODE, &mask->icmp.code,
2230 TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
2231 sizeof(key->icmp.code))))
2232 goto nla_put_failure;
2233 else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
2234 key->basic.ip_proto == IPPROTO_ICMPV6 &&
2235 (fl_dump_key_val(skb, &key->icmp.type,
2236 TCA_FLOWER_KEY_ICMPV6_TYPE, &mask->icmp.type,
2237 TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
2238 sizeof(key->icmp.type)) ||
2239 fl_dump_key_val(skb, &key->icmp.code,
2240 TCA_FLOWER_KEY_ICMPV6_CODE, &mask->icmp.code,
2241 TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
2242 sizeof(key->icmp.code))))
2243 goto nla_put_failure;
2244 else if ((key->basic.n_proto == htons(ETH_P_ARP) ||
2245 key->basic.n_proto == htons(ETH_P_RARP)) &&
2246 (fl_dump_key_val(skb, &key->arp.sip,
2247 TCA_FLOWER_KEY_ARP_SIP, &mask->arp.sip,
2248 TCA_FLOWER_KEY_ARP_SIP_MASK,
2249 sizeof(key->arp.sip)) ||
2250 fl_dump_key_val(skb, &key->arp.tip,
2251 TCA_FLOWER_KEY_ARP_TIP, &mask->arp.tip,
2252 TCA_FLOWER_KEY_ARP_TIP_MASK,
2253 sizeof(key->arp.tip)) ||
2254 fl_dump_key_val(skb, &key->arp.op,
2255 TCA_FLOWER_KEY_ARP_OP, &mask->arp.op,
2256 TCA_FLOWER_KEY_ARP_OP_MASK,
2257 sizeof(key->arp.op)) ||
2258 fl_dump_key_val(skb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
2259 mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
2260 sizeof(key->arp.sha)) ||
2261 fl_dump_key_val(skb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
2262 mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
2263 sizeof(key->arp.tha))))
2264 goto nla_put_failure;
2266 if ((key->basic.ip_proto == IPPROTO_TCP ||
2267 key->basic.ip_proto == IPPROTO_UDP ||
2268 key->basic.ip_proto == IPPROTO_SCTP) &&
2269 fl_dump_key_port_range(skb, key, mask))
2270 goto nla_put_failure;
2272 if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
2273 (fl_dump_key_val(skb, &key->enc_ipv4.src,
2274 TCA_FLOWER_KEY_ENC_IPV4_SRC, &mask->enc_ipv4.src,
2275 TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
2276 sizeof(key->enc_ipv4.src)) ||
2277 fl_dump_key_val(skb, &key->enc_ipv4.dst,
2278 TCA_FLOWER_KEY_ENC_IPV4_DST, &mask->enc_ipv4.dst,
2279 TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
2280 sizeof(key->enc_ipv4.dst))))
2281 goto nla_put_failure;
2282 else if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
2283 (fl_dump_key_val(skb, &key->enc_ipv6.src,
2284 TCA_FLOWER_KEY_ENC_IPV6_SRC, &mask->enc_ipv6.src,
2285 TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
2286 sizeof(key->enc_ipv6.src)) ||
2287 fl_dump_key_val(skb, &key->enc_ipv6.dst,
2288 TCA_FLOWER_KEY_ENC_IPV6_DST,
2289 &mask->enc_ipv6.dst,
2290 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
2291 sizeof(key->enc_ipv6.dst))))
2292 goto nla_put_failure;
2294 if (fl_dump_key_val(skb, &key->enc_key_id, TCA_FLOWER_KEY_ENC_KEY_ID,
2295 &mask->enc_key_id, TCA_FLOWER_UNSPEC,
2296 sizeof(key->enc_key_id)) ||
2297 fl_dump_key_val(skb, &key->enc_tp.src,
2298 TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
2300 TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
2301 sizeof(key->enc_tp.src)) ||
2302 fl_dump_key_val(skb, &key->enc_tp.dst,
2303 TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
2305 TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
2306 sizeof(key->enc_tp.dst)) ||
2307 fl_dump_key_ip(skb, true, &key->enc_ip, &mask->enc_ip) ||
2308 fl_dump_key_enc_opt(skb, &key->enc_opts, &mask->enc_opts))
2309 goto nla_put_failure;
2311 if (fl_dump_key_flags(skb, key->control.flags, mask->control.flags))
2312 goto nla_put_failure;
2320 static int fl_dump(struct net *net, struct tcf_proto *tp, void *fh,
2321 struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
2323 struct cls_fl_filter *f = fh;
2324 struct nlattr *nest;
2325 struct fl_flow_key *key, *mask;
2331 t->tcm_handle = f->handle;
2333 nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
2335 goto nla_put_failure;
2337 spin_lock(&tp->lock);
2339 if (f->res.classid &&
2340 nla_put_u32(skb, TCA_FLOWER_CLASSID, f->res.classid))
2341 goto nla_put_failure_locked;
2344 mask = &f->mask->key;
2345 skip_hw = tc_skip_hw(f->flags);
2347 if (fl_dump_key(skb, net, key, mask))
2348 goto nla_put_failure_locked;
2350 if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
2351 goto nla_put_failure_locked;
2353 spin_unlock(&tp->lock);
2356 fl_hw_update_stats(tp, f, rtnl_held);
2358 if (nla_put_u32(skb, TCA_FLOWER_IN_HW_COUNT, f->in_hw_count))
2359 goto nla_put_failure;
2361 if (tcf_exts_dump(skb, &f->exts))
2362 goto nla_put_failure;
2364 nla_nest_end(skb, nest);
2366 if (tcf_exts_dump_stats(skb, &f->exts) < 0)
2367 goto nla_put_failure;
2371 nla_put_failure_locked:
2372 spin_unlock(&tp->lock);
2374 nla_nest_cancel(skb, nest);
2378 static int fl_tmplt_dump(struct sk_buff *skb, struct net *net, void *tmplt_priv)
2380 struct fl_flow_tmplt *tmplt = tmplt_priv;
2381 struct fl_flow_key *key, *mask;
2382 struct nlattr *nest;
2384 nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
2386 goto nla_put_failure;
2388 key = &tmplt->dummy_key;
2389 mask = &tmplt->mask;
2391 if (fl_dump_key(skb, net, key, mask))
2392 goto nla_put_failure;
2394 nla_nest_end(skb, nest);
2399 nla_nest_cancel(skb, nest);
2403 static void fl_bind_class(void *fh, u32 classid, unsigned long cl)
2405 struct cls_fl_filter *f = fh;
2407 if (f && f->res.classid == classid)
2411 static struct tcf_proto_ops cls_fl_ops __read_mostly = {
2413 .classify = fl_classify,
2415 .destroy = fl_destroy,
2418 .change = fl_change,
2419 .delete = fl_delete,
2421 .reoffload = fl_reoffload,
2423 .bind_class = fl_bind_class,
2424 .tmplt_create = fl_tmplt_create,
2425 .tmplt_destroy = fl_tmplt_destroy,
2426 .tmplt_dump = fl_tmplt_dump,
2427 .owner = THIS_MODULE,
2428 .flags = TCF_PROTO_OPS_DOIT_UNLOCKED,
2431 static int __init cls_fl_init(void)
2433 return register_tcf_proto_ops(&cls_fl_ops);
2436 static void __exit cls_fl_exit(void)
2438 unregister_tcf_proto_ops(&cls_fl_ops);
2441 module_init(cls_fl_init);
2442 module_exit(cls_fl_exit);
2444 MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
2445 MODULE_DESCRIPTION("Flower classifier");
2446 MODULE_LICENSE("GPL v2");