Linux-libre 5.3.12-gnu
[librecmc/linux-libre.git] / net / sched / cls_flower.c
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
26 #include <net/dst.h>
27 #include <net/dst_metadata.h>
28
29 #include <uapi/linux/netfilter/nf_conntrack_common.h>
30
31 struct fl_flow_key {
32         struct flow_dissector_key_meta meta;
33         struct flow_dissector_key_control control;
34         struct flow_dissector_key_control enc_control;
35         struct flow_dissector_key_basic basic;
36         struct flow_dissector_key_eth_addrs eth;
37         struct flow_dissector_key_vlan vlan;
38         struct flow_dissector_key_vlan cvlan;
39         union {
40                 struct flow_dissector_key_ipv4_addrs ipv4;
41                 struct flow_dissector_key_ipv6_addrs ipv6;
42         };
43         struct flow_dissector_key_ports tp;
44         struct flow_dissector_key_icmp icmp;
45         struct flow_dissector_key_arp arp;
46         struct flow_dissector_key_keyid enc_key_id;
47         union {
48                 struct flow_dissector_key_ipv4_addrs enc_ipv4;
49                 struct flow_dissector_key_ipv6_addrs enc_ipv6;
50         };
51         struct flow_dissector_key_ports enc_tp;
52         struct flow_dissector_key_mpls mpls;
53         struct flow_dissector_key_tcp tcp;
54         struct flow_dissector_key_ip ip;
55         struct flow_dissector_key_ip enc_ip;
56         struct flow_dissector_key_enc_opts enc_opts;
57         struct flow_dissector_key_ports tp_min;
58         struct flow_dissector_key_ports tp_max;
59         struct flow_dissector_key_ct ct;
60 } __aligned(BITS_PER_LONG / 8); /* Ensure that we can do comparisons as longs. */
61
62 struct fl_flow_mask_range {
63         unsigned short int start;
64         unsigned short int end;
65 };
66
67 struct fl_flow_mask {
68         struct fl_flow_key key;
69         struct fl_flow_mask_range range;
70         u32 flags;
71         struct rhash_head ht_node;
72         struct rhashtable ht;
73         struct rhashtable_params filter_ht_params;
74         struct flow_dissector dissector;
75         struct list_head filters;
76         struct rcu_work rwork;
77         struct list_head list;
78         refcount_t refcnt;
79 };
80
81 struct fl_flow_tmplt {
82         struct fl_flow_key dummy_key;
83         struct fl_flow_key mask;
84         struct flow_dissector dissector;
85         struct tcf_chain *chain;
86 };
87
88 struct cls_fl_head {
89         struct rhashtable ht;
90         spinlock_t masks_lock; /* Protect masks list */
91         struct list_head masks;
92         struct list_head hw_filters;
93         struct rcu_work rwork;
94         struct idr handle_idr;
95 };
96
97 struct cls_fl_filter {
98         struct fl_flow_mask *mask;
99         struct rhash_head ht_node;
100         struct fl_flow_key mkey;
101         struct tcf_exts exts;
102         struct tcf_result res;
103         struct fl_flow_key key;
104         struct list_head list;
105         struct list_head hw_list;
106         u32 handle;
107         u32 flags;
108         u32 in_hw_count;
109         struct rcu_work rwork;
110         struct net_device *hw_dev;
111         /* Flower classifier is unlocked, which means that its reference counter
112          * can be changed concurrently without any kind of external
113          * synchronization. Use atomic reference counter to be concurrency-safe.
114          */
115         refcount_t refcnt;
116         bool deleted;
117 };
118
119 static const struct rhashtable_params mask_ht_params = {
120         .key_offset = offsetof(struct fl_flow_mask, key),
121         .key_len = sizeof(struct fl_flow_key),
122         .head_offset = offsetof(struct fl_flow_mask, ht_node),
123         .automatic_shrinking = true,
124 };
125
126 static unsigned short int fl_mask_range(const struct fl_flow_mask *mask)
127 {
128         return mask->range.end - mask->range.start;
129 }
130
131 static void fl_mask_update_range(struct fl_flow_mask *mask)
132 {
133         const u8 *bytes = (const u8 *) &mask->key;
134         size_t size = sizeof(mask->key);
135         size_t i, first = 0, last;
136
137         for (i = 0; i < size; i++) {
138                 if (bytes[i]) {
139                         first = i;
140                         break;
141                 }
142         }
143         last = first;
144         for (i = size - 1; i != first; i--) {
145                 if (bytes[i]) {
146                         last = i;
147                         break;
148                 }
149         }
150         mask->range.start = rounddown(first, sizeof(long));
151         mask->range.end = roundup(last + 1, sizeof(long));
152 }
153
154 static void *fl_key_get_start(struct fl_flow_key *key,
155                               const struct fl_flow_mask *mask)
156 {
157         return (u8 *) key + mask->range.start;
158 }
159
160 static void fl_set_masked_key(struct fl_flow_key *mkey, struct fl_flow_key *key,
161                               struct fl_flow_mask *mask)
162 {
163         const long *lkey = fl_key_get_start(key, mask);
164         const long *lmask = fl_key_get_start(&mask->key, mask);
165         long *lmkey = fl_key_get_start(mkey, mask);
166         int i;
167
168         for (i = 0; i < fl_mask_range(mask); i += sizeof(long))
169                 *lmkey++ = *lkey++ & *lmask++;
170 }
171
172 static bool fl_mask_fits_tmplt(struct fl_flow_tmplt *tmplt,
173                                struct fl_flow_mask *mask)
174 {
175         const long *lmask = fl_key_get_start(&mask->key, mask);
176         const long *ltmplt;
177         int i;
178
179         if (!tmplt)
180                 return true;
181         ltmplt = fl_key_get_start(&tmplt->mask, mask);
182         for (i = 0; i < fl_mask_range(mask); i += sizeof(long)) {
183                 if (~*ltmplt++ & *lmask++)
184                         return false;
185         }
186         return true;
187 }
188
189 static void fl_clear_masked_range(struct fl_flow_key *key,
190                                   struct fl_flow_mask *mask)
191 {
192         memset(fl_key_get_start(key, mask), 0, fl_mask_range(mask));
193 }
194
195 static bool fl_range_port_dst_cmp(struct cls_fl_filter *filter,
196                                   struct fl_flow_key *key,
197                                   struct fl_flow_key *mkey)
198 {
199         __be16 min_mask, max_mask, min_val, max_val;
200
201         min_mask = htons(filter->mask->key.tp_min.dst);
202         max_mask = htons(filter->mask->key.tp_max.dst);
203         min_val = htons(filter->key.tp_min.dst);
204         max_val = htons(filter->key.tp_max.dst);
205
206         if (min_mask && max_mask) {
207                 if (htons(key->tp.dst) < min_val ||
208                     htons(key->tp.dst) > max_val)
209                         return false;
210
211                 /* skb does not have min and max values */
212                 mkey->tp_min.dst = filter->mkey.tp_min.dst;
213                 mkey->tp_max.dst = filter->mkey.tp_max.dst;
214         }
215         return true;
216 }
217
218 static bool fl_range_port_src_cmp(struct cls_fl_filter *filter,
219                                   struct fl_flow_key *key,
220                                   struct fl_flow_key *mkey)
221 {
222         __be16 min_mask, max_mask, min_val, max_val;
223
224         min_mask = htons(filter->mask->key.tp_min.src);
225         max_mask = htons(filter->mask->key.tp_max.src);
226         min_val = htons(filter->key.tp_min.src);
227         max_val = htons(filter->key.tp_max.src);
228
229         if (min_mask && max_mask) {
230                 if (htons(key->tp.src) < min_val ||
231                     htons(key->tp.src) > max_val)
232                         return false;
233
234                 /* skb does not have min and max values */
235                 mkey->tp_min.src = filter->mkey.tp_min.src;
236                 mkey->tp_max.src = filter->mkey.tp_max.src;
237         }
238         return true;
239 }
240
241 static struct cls_fl_filter *__fl_lookup(struct fl_flow_mask *mask,
242                                          struct fl_flow_key *mkey)
243 {
244         return rhashtable_lookup_fast(&mask->ht, fl_key_get_start(mkey, mask),
245                                       mask->filter_ht_params);
246 }
247
248 static struct cls_fl_filter *fl_lookup_range(struct fl_flow_mask *mask,
249                                              struct fl_flow_key *mkey,
250                                              struct fl_flow_key *key)
251 {
252         struct cls_fl_filter *filter, *f;
253
254         list_for_each_entry_rcu(filter, &mask->filters, list) {
255                 if (!fl_range_port_dst_cmp(filter, key, mkey))
256                         continue;
257
258                 if (!fl_range_port_src_cmp(filter, key, mkey))
259                         continue;
260
261                 f = __fl_lookup(mask, mkey);
262                 if (f)
263                         return f;
264         }
265         return NULL;
266 }
267
268 static struct cls_fl_filter *fl_lookup(struct fl_flow_mask *mask,
269                                        struct fl_flow_key *mkey,
270                                        struct fl_flow_key *key)
271 {
272         if ((mask->flags & TCA_FLOWER_MASK_FLAGS_RANGE))
273                 return fl_lookup_range(mask, mkey, key);
274
275         return __fl_lookup(mask, mkey);
276 }
277
278 static u16 fl_ct_info_to_flower_map[] = {
279         [IP_CT_ESTABLISHED] =           TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
280                                         TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED,
281         [IP_CT_RELATED] =               TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
282                                         TCA_FLOWER_KEY_CT_FLAGS_RELATED,
283         [IP_CT_ESTABLISHED_REPLY] =     TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
284                                         TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED,
285         [IP_CT_RELATED_REPLY] =         TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
286                                         TCA_FLOWER_KEY_CT_FLAGS_RELATED,
287         [IP_CT_NEW] =                   TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
288                                         TCA_FLOWER_KEY_CT_FLAGS_NEW,
289 };
290
291 static int fl_classify(struct sk_buff *skb, const struct tcf_proto *tp,
292                        struct tcf_result *res)
293 {
294         struct cls_fl_head *head = rcu_dereference_bh(tp->root);
295         struct fl_flow_key skb_mkey;
296         struct fl_flow_key skb_key;
297         struct fl_flow_mask *mask;
298         struct cls_fl_filter *f;
299
300         list_for_each_entry_rcu(mask, &head->masks, list) {
301                 fl_clear_masked_range(&skb_key, mask);
302
303                 skb_flow_dissect_meta(skb, &mask->dissector, &skb_key);
304                 /* skb_flow_dissect() does not set n_proto in case an unknown
305                  * protocol, so do it rather here.
306                  */
307                 skb_key.basic.n_proto = skb->protocol;
308                 skb_flow_dissect_tunnel_info(skb, &mask->dissector, &skb_key);
309                 skb_flow_dissect_ct(skb, &mask->dissector, &skb_key,
310                                     fl_ct_info_to_flower_map,
311                                     ARRAY_SIZE(fl_ct_info_to_flower_map));
312                 skb_flow_dissect(skb, &mask->dissector, &skb_key, 0);
313
314                 fl_set_masked_key(&skb_mkey, &skb_key, mask);
315
316                 f = fl_lookup(mask, &skb_mkey, &skb_key);
317                 if (f && !tc_skip_sw(f->flags)) {
318                         *res = f->res;
319                         return tcf_exts_exec(skb, &f->exts, res);
320                 }
321         }
322         return -1;
323 }
324
325 static int fl_init(struct tcf_proto *tp)
326 {
327         struct cls_fl_head *head;
328
329         head = kzalloc(sizeof(*head), GFP_KERNEL);
330         if (!head)
331                 return -ENOBUFS;
332
333         spin_lock_init(&head->masks_lock);
334         INIT_LIST_HEAD_RCU(&head->masks);
335         INIT_LIST_HEAD(&head->hw_filters);
336         rcu_assign_pointer(tp->root, head);
337         idr_init(&head->handle_idr);
338
339         return rhashtable_init(&head->ht, &mask_ht_params);
340 }
341
342 static void fl_mask_free(struct fl_flow_mask *mask, bool mask_init_done)
343 {
344         /* temporary masks don't have their filters list and ht initialized */
345         if (mask_init_done) {
346                 WARN_ON(!list_empty(&mask->filters));
347                 rhashtable_destroy(&mask->ht);
348         }
349         kfree(mask);
350 }
351
352 static void fl_mask_free_work(struct work_struct *work)
353 {
354         struct fl_flow_mask *mask = container_of(to_rcu_work(work),
355                                                  struct fl_flow_mask, rwork);
356
357         fl_mask_free(mask, true);
358 }
359
360 static void fl_uninit_mask_free_work(struct work_struct *work)
361 {
362         struct fl_flow_mask *mask = container_of(to_rcu_work(work),
363                                                  struct fl_flow_mask, rwork);
364
365         fl_mask_free(mask, false);
366 }
367
368 static bool fl_mask_put(struct cls_fl_head *head, struct fl_flow_mask *mask)
369 {
370         if (!refcount_dec_and_test(&mask->refcnt))
371                 return false;
372
373         rhashtable_remove_fast(&head->ht, &mask->ht_node, mask_ht_params);
374
375         spin_lock(&head->masks_lock);
376         list_del_rcu(&mask->list);
377         spin_unlock(&head->masks_lock);
378
379         tcf_queue_work(&mask->rwork, fl_mask_free_work);
380
381         return true;
382 }
383
384 static struct cls_fl_head *fl_head_dereference(struct tcf_proto *tp)
385 {
386         /* Flower classifier only changes root pointer during init and destroy.
387          * Users must obtain reference to tcf_proto instance before calling its
388          * API, so tp->root pointer is protected from concurrent call to
389          * fl_destroy() by reference counting.
390          */
391         return rcu_dereference_raw(tp->root);
392 }
393
394 static void __fl_destroy_filter(struct cls_fl_filter *f)
395 {
396         tcf_exts_destroy(&f->exts);
397         tcf_exts_put_net(&f->exts);
398         kfree(f);
399 }
400
401 static void fl_destroy_filter_work(struct work_struct *work)
402 {
403         struct cls_fl_filter *f = container_of(to_rcu_work(work),
404                                         struct cls_fl_filter, rwork);
405
406         __fl_destroy_filter(f);
407 }
408
409 static void fl_hw_destroy_filter(struct tcf_proto *tp, struct cls_fl_filter *f,
410                                  bool rtnl_held, struct netlink_ext_ack *extack)
411 {
412         struct tcf_block *block = tp->chain->block;
413         struct flow_cls_offload cls_flower = {};
414
415         if (!rtnl_held)
416                 rtnl_lock();
417
418         tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
419         cls_flower.command = FLOW_CLS_DESTROY;
420         cls_flower.cookie = (unsigned long) f;
421
422         tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false);
423         spin_lock(&tp->lock);
424         list_del_init(&f->hw_list);
425         tcf_block_offload_dec(block, &f->flags);
426         spin_unlock(&tp->lock);
427
428         if (!rtnl_held)
429                 rtnl_unlock();
430 }
431
432 static int fl_hw_replace_filter(struct tcf_proto *tp,
433                                 struct cls_fl_filter *f, bool rtnl_held,
434                                 struct netlink_ext_ack *extack)
435 {
436         struct cls_fl_head *head = fl_head_dereference(tp);
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         if (!rtnl_held)
443                 rtnl_lock();
444
445         cls_flower.rule = flow_rule_alloc(tcf_exts_num_actions(&f->exts));
446         if (!cls_flower.rule) {
447                 err = -ENOMEM;
448                 goto errout;
449         }
450
451         tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack);
452         cls_flower.command = FLOW_CLS_REPLACE;
453         cls_flower.cookie = (unsigned long) f;
454         cls_flower.rule->match.dissector = &f->mask->dissector;
455         cls_flower.rule->match.mask = &f->mask->key;
456         cls_flower.rule->match.key = &f->mkey;
457         cls_flower.classid = f->res.classid;
458
459         err = tc_setup_flow_action(&cls_flower.rule->action, &f->exts);
460         if (err) {
461                 kfree(cls_flower.rule);
462                 if (skip_sw)
463                         NL_SET_ERR_MSG_MOD(extack, "Failed to setup flow action");
464                 else
465                         err = 0;
466                 goto errout;
467         }
468
469         err = tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, skip_sw);
470         kfree(cls_flower.rule);
471
472         if (err < 0) {
473                 fl_hw_destroy_filter(tp, f, true, NULL);
474                 goto errout;
475         } else if (err > 0) {
476                 f->in_hw_count = err;
477                 err = 0;
478                 spin_lock(&tp->lock);
479                 tcf_block_offload_inc(block, &f->flags);
480                 spin_unlock(&tp->lock);
481         }
482
483         if (skip_sw && !(f->flags & TCA_CLS_FLAGS_IN_HW)) {
484                 err = -EINVAL;
485                 goto errout;
486         }
487
488         spin_lock(&tp->lock);
489         list_add(&f->hw_list, &head->hw_filters);
490         spin_unlock(&tp->lock);
491 errout:
492         if (!rtnl_held)
493                 rtnl_unlock();
494
495         return err;
496 }
497
498 static void fl_hw_update_stats(struct tcf_proto *tp, struct cls_fl_filter *f,
499                                bool rtnl_held)
500 {
501         struct tcf_block *block = tp->chain->block;
502         struct flow_cls_offload cls_flower = {};
503
504         if (!rtnl_held)
505                 rtnl_lock();
506
507         tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, NULL);
508         cls_flower.command = FLOW_CLS_STATS;
509         cls_flower.cookie = (unsigned long) f;
510         cls_flower.classid = f->res.classid;
511
512         tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false);
513
514         tcf_exts_stats_update(&f->exts, cls_flower.stats.bytes,
515                               cls_flower.stats.pkts,
516                               cls_flower.stats.lastused);
517
518         if (!rtnl_held)
519                 rtnl_unlock();
520 }
521
522 static void __fl_put(struct cls_fl_filter *f)
523 {
524         if (!refcount_dec_and_test(&f->refcnt))
525                 return;
526
527         if (tcf_exts_get_net(&f->exts))
528                 tcf_queue_work(&f->rwork, fl_destroy_filter_work);
529         else
530                 __fl_destroy_filter(f);
531 }
532
533 static struct cls_fl_filter *__fl_get(struct cls_fl_head *head, u32 handle)
534 {
535         struct cls_fl_filter *f;
536
537         rcu_read_lock();
538         f = idr_find(&head->handle_idr, handle);
539         if (f && !refcount_inc_not_zero(&f->refcnt))
540                 f = NULL;
541         rcu_read_unlock();
542
543         return f;
544 }
545
546 static int __fl_delete(struct tcf_proto *tp, struct cls_fl_filter *f,
547                        bool *last, bool rtnl_held,
548                        struct netlink_ext_ack *extack)
549 {
550         struct cls_fl_head *head = fl_head_dereference(tp);
551
552         *last = false;
553
554         spin_lock(&tp->lock);
555         if (f->deleted) {
556                 spin_unlock(&tp->lock);
557                 return -ENOENT;
558         }
559
560         f->deleted = true;
561         rhashtable_remove_fast(&f->mask->ht, &f->ht_node,
562                                f->mask->filter_ht_params);
563         idr_remove(&head->handle_idr, f->handle);
564         list_del_rcu(&f->list);
565         spin_unlock(&tp->lock);
566
567         *last = fl_mask_put(head, f->mask);
568         if (!tc_skip_hw(f->flags))
569                 fl_hw_destroy_filter(tp, f, rtnl_held, extack);
570         tcf_unbind_filter(tp, &f->res);
571         __fl_put(f);
572
573         return 0;
574 }
575
576 static void fl_destroy_sleepable(struct work_struct *work)
577 {
578         struct cls_fl_head *head = container_of(to_rcu_work(work),
579                                                 struct cls_fl_head,
580                                                 rwork);
581
582         rhashtable_destroy(&head->ht);
583         kfree(head);
584         module_put(THIS_MODULE);
585 }
586
587 static void fl_destroy(struct tcf_proto *tp, bool rtnl_held,
588                        struct netlink_ext_ack *extack)
589 {
590         struct cls_fl_head *head = fl_head_dereference(tp);
591         struct fl_flow_mask *mask, *next_mask;
592         struct cls_fl_filter *f, *next;
593         bool last;
594
595         list_for_each_entry_safe(mask, next_mask, &head->masks, list) {
596                 list_for_each_entry_safe(f, next, &mask->filters, list) {
597                         __fl_delete(tp, f, &last, rtnl_held, extack);
598                         if (last)
599                                 break;
600                 }
601         }
602         idr_destroy(&head->handle_idr);
603
604         __module_get(THIS_MODULE);
605         tcf_queue_work(&head->rwork, fl_destroy_sleepable);
606 }
607
608 static void fl_put(struct tcf_proto *tp, void *arg)
609 {
610         struct cls_fl_filter *f = arg;
611
612         __fl_put(f);
613 }
614
615 static void *fl_get(struct tcf_proto *tp, u32 handle)
616 {
617         struct cls_fl_head *head = fl_head_dereference(tp);
618
619         return __fl_get(head, handle);
620 }
621
622 static const struct nla_policy fl_policy[TCA_FLOWER_MAX + 1] = {
623         [TCA_FLOWER_UNSPEC]             = { .type = NLA_UNSPEC },
624         [TCA_FLOWER_CLASSID]            = { .type = NLA_U32 },
625         [TCA_FLOWER_INDEV]              = { .type = NLA_STRING,
626                                             .len = IFNAMSIZ },
627         [TCA_FLOWER_KEY_ETH_DST]        = { .len = ETH_ALEN },
628         [TCA_FLOWER_KEY_ETH_DST_MASK]   = { .len = ETH_ALEN },
629         [TCA_FLOWER_KEY_ETH_SRC]        = { .len = ETH_ALEN },
630         [TCA_FLOWER_KEY_ETH_SRC_MASK]   = { .len = ETH_ALEN },
631         [TCA_FLOWER_KEY_ETH_TYPE]       = { .type = NLA_U16 },
632         [TCA_FLOWER_KEY_IP_PROTO]       = { .type = NLA_U8 },
633         [TCA_FLOWER_KEY_IPV4_SRC]       = { .type = NLA_U32 },
634         [TCA_FLOWER_KEY_IPV4_SRC_MASK]  = { .type = NLA_U32 },
635         [TCA_FLOWER_KEY_IPV4_DST]       = { .type = NLA_U32 },
636         [TCA_FLOWER_KEY_IPV4_DST_MASK]  = { .type = NLA_U32 },
637         [TCA_FLOWER_KEY_IPV6_SRC]       = { .len = sizeof(struct in6_addr) },
638         [TCA_FLOWER_KEY_IPV6_SRC_MASK]  = { .len = sizeof(struct in6_addr) },
639         [TCA_FLOWER_KEY_IPV6_DST]       = { .len = sizeof(struct in6_addr) },
640         [TCA_FLOWER_KEY_IPV6_DST_MASK]  = { .len = sizeof(struct in6_addr) },
641         [TCA_FLOWER_KEY_TCP_SRC]        = { .type = NLA_U16 },
642         [TCA_FLOWER_KEY_TCP_DST]        = { .type = NLA_U16 },
643         [TCA_FLOWER_KEY_UDP_SRC]        = { .type = NLA_U16 },
644         [TCA_FLOWER_KEY_UDP_DST]        = { .type = NLA_U16 },
645         [TCA_FLOWER_KEY_VLAN_ID]        = { .type = NLA_U16 },
646         [TCA_FLOWER_KEY_VLAN_PRIO]      = { .type = NLA_U8 },
647         [TCA_FLOWER_KEY_VLAN_ETH_TYPE]  = { .type = NLA_U16 },
648         [TCA_FLOWER_KEY_ENC_KEY_ID]     = { .type = NLA_U32 },
649         [TCA_FLOWER_KEY_ENC_IPV4_SRC]   = { .type = NLA_U32 },
650         [TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK] = { .type = NLA_U32 },
651         [TCA_FLOWER_KEY_ENC_IPV4_DST]   = { .type = NLA_U32 },
652         [TCA_FLOWER_KEY_ENC_IPV4_DST_MASK] = { .type = NLA_U32 },
653         [TCA_FLOWER_KEY_ENC_IPV6_SRC]   = { .len = sizeof(struct in6_addr) },
654         [TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK] = { .len = sizeof(struct in6_addr) },
655         [TCA_FLOWER_KEY_ENC_IPV6_DST]   = { .len = sizeof(struct in6_addr) },
656         [TCA_FLOWER_KEY_ENC_IPV6_DST_MASK] = { .len = sizeof(struct in6_addr) },
657         [TCA_FLOWER_KEY_TCP_SRC_MASK]   = { .type = NLA_U16 },
658         [TCA_FLOWER_KEY_TCP_DST_MASK]   = { .type = NLA_U16 },
659         [TCA_FLOWER_KEY_UDP_SRC_MASK]   = { .type = NLA_U16 },
660         [TCA_FLOWER_KEY_UDP_DST_MASK]   = { .type = NLA_U16 },
661         [TCA_FLOWER_KEY_SCTP_SRC_MASK]  = { .type = NLA_U16 },
662         [TCA_FLOWER_KEY_SCTP_DST_MASK]  = { .type = NLA_U16 },
663         [TCA_FLOWER_KEY_SCTP_SRC]       = { .type = NLA_U16 },
664         [TCA_FLOWER_KEY_SCTP_DST]       = { .type = NLA_U16 },
665         [TCA_FLOWER_KEY_ENC_UDP_SRC_PORT]       = { .type = NLA_U16 },
666         [TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK]  = { .type = NLA_U16 },
667         [TCA_FLOWER_KEY_ENC_UDP_DST_PORT]       = { .type = NLA_U16 },
668         [TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK]  = { .type = NLA_U16 },
669         [TCA_FLOWER_KEY_FLAGS]          = { .type = NLA_U32 },
670         [TCA_FLOWER_KEY_FLAGS_MASK]     = { .type = NLA_U32 },
671         [TCA_FLOWER_KEY_ICMPV4_TYPE]    = { .type = NLA_U8 },
672         [TCA_FLOWER_KEY_ICMPV4_TYPE_MASK] = { .type = NLA_U8 },
673         [TCA_FLOWER_KEY_ICMPV4_CODE]    = { .type = NLA_U8 },
674         [TCA_FLOWER_KEY_ICMPV4_CODE_MASK] = { .type = NLA_U8 },
675         [TCA_FLOWER_KEY_ICMPV6_TYPE]    = { .type = NLA_U8 },
676         [TCA_FLOWER_KEY_ICMPV6_TYPE_MASK] = { .type = NLA_U8 },
677         [TCA_FLOWER_KEY_ICMPV6_CODE]    = { .type = NLA_U8 },
678         [TCA_FLOWER_KEY_ICMPV6_CODE_MASK] = { .type = NLA_U8 },
679         [TCA_FLOWER_KEY_ARP_SIP]        = { .type = NLA_U32 },
680         [TCA_FLOWER_KEY_ARP_SIP_MASK]   = { .type = NLA_U32 },
681         [TCA_FLOWER_KEY_ARP_TIP]        = { .type = NLA_U32 },
682         [TCA_FLOWER_KEY_ARP_TIP_MASK]   = { .type = NLA_U32 },
683         [TCA_FLOWER_KEY_ARP_OP]         = { .type = NLA_U8 },
684         [TCA_FLOWER_KEY_ARP_OP_MASK]    = { .type = NLA_U8 },
685         [TCA_FLOWER_KEY_ARP_SHA]        = { .len = ETH_ALEN },
686         [TCA_FLOWER_KEY_ARP_SHA_MASK]   = { .len = ETH_ALEN },
687         [TCA_FLOWER_KEY_ARP_THA]        = { .len = ETH_ALEN },
688         [TCA_FLOWER_KEY_ARP_THA_MASK]   = { .len = ETH_ALEN },
689         [TCA_FLOWER_KEY_MPLS_TTL]       = { .type = NLA_U8 },
690         [TCA_FLOWER_KEY_MPLS_BOS]       = { .type = NLA_U8 },
691         [TCA_FLOWER_KEY_MPLS_TC]        = { .type = NLA_U8 },
692         [TCA_FLOWER_KEY_MPLS_LABEL]     = { .type = NLA_U32 },
693         [TCA_FLOWER_KEY_TCP_FLAGS]      = { .type = NLA_U16 },
694         [TCA_FLOWER_KEY_TCP_FLAGS_MASK] = { .type = NLA_U16 },
695         [TCA_FLOWER_KEY_IP_TOS]         = { .type = NLA_U8 },
696         [TCA_FLOWER_KEY_IP_TOS_MASK]    = { .type = NLA_U8 },
697         [TCA_FLOWER_KEY_IP_TTL]         = { .type = NLA_U8 },
698         [TCA_FLOWER_KEY_IP_TTL_MASK]    = { .type = NLA_U8 },
699         [TCA_FLOWER_KEY_CVLAN_ID]       = { .type = NLA_U16 },
700         [TCA_FLOWER_KEY_CVLAN_PRIO]     = { .type = NLA_U8 },
701         [TCA_FLOWER_KEY_CVLAN_ETH_TYPE] = { .type = NLA_U16 },
702         [TCA_FLOWER_KEY_ENC_IP_TOS]     = { .type = NLA_U8 },
703         [TCA_FLOWER_KEY_ENC_IP_TOS_MASK] = { .type = NLA_U8 },
704         [TCA_FLOWER_KEY_ENC_IP_TTL]      = { .type = NLA_U8 },
705         [TCA_FLOWER_KEY_ENC_IP_TTL_MASK] = { .type = NLA_U8 },
706         [TCA_FLOWER_KEY_ENC_OPTS]       = { .type = NLA_NESTED },
707         [TCA_FLOWER_KEY_ENC_OPTS_MASK]  = { .type = NLA_NESTED },
708         [TCA_FLOWER_KEY_CT_STATE]       = { .type = NLA_U16 },
709         [TCA_FLOWER_KEY_CT_STATE_MASK]  = { .type = NLA_U16 },
710         [TCA_FLOWER_KEY_CT_ZONE]        = { .type = NLA_U16 },
711         [TCA_FLOWER_KEY_CT_ZONE_MASK]   = { .type = NLA_U16 },
712         [TCA_FLOWER_KEY_CT_MARK]        = { .type = NLA_U32 },
713         [TCA_FLOWER_KEY_CT_MARK_MASK]   = { .type = NLA_U32 },
714         [TCA_FLOWER_KEY_CT_LABELS]      = { .type = NLA_BINARY,
715                                             .len = 128 / BITS_PER_BYTE },
716         [TCA_FLOWER_KEY_CT_LABELS_MASK] = { .type = NLA_BINARY,
717                                             .len = 128 / BITS_PER_BYTE },
718 };
719
720 static const struct nla_policy
721 enc_opts_policy[TCA_FLOWER_KEY_ENC_OPTS_MAX + 1] = {
722         [TCA_FLOWER_KEY_ENC_OPTS_GENEVE]        = { .type = NLA_NESTED },
723 };
724
725 static const struct nla_policy
726 geneve_opt_policy[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1] = {
727         [TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS]      = { .type = NLA_U16 },
728         [TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE]       = { .type = NLA_U8 },
729         [TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA]       = { .type = NLA_BINARY,
730                                                        .len = 128 },
731 };
732
733 static void fl_set_key_val(struct nlattr **tb,
734                            void *val, int val_type,
735                            void *mask, int mask_type, int len)
736 {
737         if (!tb[val_type])
738                 return;
739         nla_memcpy(val, tb[val_type], len);
740         if (mask_type == TCA_FLOWER_UNSPEC || !tb[mask_type])
741                 memset(mask, 0xff, len);
742         else
743                 nla_memcpy(mask, tb[mask_type], len);
744 }
745
746 static int fl_set_key_port_range(struct nlattr **tb, struct fl_flow_key *key,
747                                  struct fl_flow_key *mask)
748 {
749         fl_set_key_val(tb, &key->tp_min.dst,
750                        TCA_FLOWER_KEY_PORT_DST_MIN, &mask->tp_min.dst,
751                        TCA_FLOWER_UNSPEC, sizeof(key->tp_min.dst));
752         fl_set_key_val(tb, &key->tp_max.dst,
753                        TCA_FLOWER_KEY_PORT_DST_MAX, &mask->tp_max.dst,
754                        TCA_FLOWER_UNSPEC, sizeof(key->tp_max.dst));
755         fl_set_key_val(tb, &key->tp_min.src,
756                        TCA_FLOWER_KEY_PORT_SRC_MIN, &mask->tp_min.src,
757                        TCA_FLOWER_UNSPEC, sizeof(key->tp_min.src));
758         fl_set_key_val(tb, &key->tp_max.src,
759                        TCA_FLOWER_KEY_PORT_SRC_MAX, &mask->tp_max.src,
760                        TCA_FLOWER_UNSPEC, sizeof(key->tp_max.src));
761
762         if ((mask->tp_min.dst && mask->tp_max.dst &&
763              htons(key->tp_max.dst) <= htons(key->tp_min.dst)) ||
764              (mask->tp_min.src && mask->tp_max.src &&
765               htons(key->tp_max.src) <= htons(key->tp_min.src)))
766                 return -EINVAL;
767
768         return 0;
769 }
770
771 static int fl_set_key_mpls(struct nlattr **tb,
772                            struct flow_dissector_key_mpls *key_val,
773                            struct flow_dissector_key_mpls *key_mask)
774 {
775         if (tb[TCA_FLOWER_KEY_MPLS_TTL]) {
776                 key_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TTL]);
777                 key_mask->mpls_ttl = MPLS_TTL_MASK;
778         }
779         if (tb[TCA_FLOWER_KEY_MPLS_BOS]) {
780                 u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_BOS]);
781
782                 if (bos & ~MPLS_BOS_MASK)
783                         return -EINVAL;
784                 key_val->mpls_bos = bos;
785                 key_mask->mpls_bos = MPLS_BOS_MASK;
786         }
787         if (tb[TCA_FLOWER_KEY_MPLS_TC]) {
788                 u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TC]);
789
790                 if (tc & ~MPLS_TC_MASK)
791                         return -EINVAL;
792                 key_val->mpls_tc = tc;
793                 key_mask->mpls_tc = MPLS_TC_MASK;
794         }
795         if (tb[TCA_FLOWER_KEY_MPLS_LABEL]) {
796                 u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_LABEL]);
797
798                 if (label & ~MPLS_LABEL_MASK)
799                         return -EINVAL;
800                 key_val->mpls_label = label;
801                 key_mask->mpls_label = MPLS_LABEL_MASK;
802         }
803         return 0;
804 }
805
806 static void fl_set_key_vlan(struct nlattr **tb,
807                             __be16 ethertype,
808                             int vlan_id_key, int vlan_prio_key,
809                             struct flow_dissector_key_vlan *key_val,
810                             struct flow_dissector_key_vlan *key_mask)
811 {
812 #define VLAN_PRIORITY_MASK      0x7
813
814         if (tb[vlan_id_key]) {
815                 key_val->vlan_id =
816                         nla_get_u16(tb[vlan_id_key]) & VLAN_VID_MASK;
817                 key_mask->vlan_id = VLAN_VID_MASK;
818         }
819         if (tb[vlan_prio_key]) {
820                 key_val->vlan_priority =
821                         nla_get_u8(tb[vlan_prio_key]) &
822                         VLAN_PRIORITY_MASK;
823                 key_mask->vlan_priority = VLAN_PRIORITY_MASK;
824         }
825         key_val->vlan_tpid = ethertype;
826         key_mask->vlan_tpid = cpu_to_be16(~0);
827 }
828
829 static void fl_set_key_flag(u32 flower_key, u32 flower_mask,
830                             u32 *dissector_key, u32 *dissector_mask,
831                             u32 flower_flag_bit, u32 dissector_flag_bit)
832 {
833         if (flower_mask & flower_flag_bit) {
834                 *dissector_mask |= dissector_flag_bit;
835                 if (flower_key & flower_flag_bit)
836                         *dissector_key |= dissector_flag_bit;
837         }
838 }
839
840 static int fl_set_key_flags(struct nlattr **tb,
841                             u32 *flags_key, u32 *flags_mask)
842 {
843         u32 key, mask;
844
845         /* mask is mandatory for flags */
846         if (!tb[TCA_FLOWER_KEY_FLAGS_MASK])
847                 return -EINVAL;
848
849         key = be32_to_cpu(nla_get_u32(tb[TCA_FLOWER_KEY_FLAGS]));
850         mask = be32_to_cpu(nla_get_u32(tb[TCA_FLOWER_KEY_FLAGS_MASK]));
851
852         *flags_key  = 0;
853         *flags_mask = 0;
854
855         fl_set_key_flag(key, mask, flags_key, flags_mask,
856                         TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
857         fl_set_key_flag(key, mask, flags_key, flags_mask,
858                         TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
859                         FLOW_DIS_FIRST_FRAG);
860
861         return 0;
862 }
863
864 static void fl_set_key_ip(struct nlattr **tb, bool encap,
865                           struct flow_dissector_key_ip *key,
866                           struct flow_dissector_key_ip *mask)
867 {
868         int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
869         int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
870         int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
871         int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
872
873         fl_set_key_val(tb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos));
874         fl_set_key_val(tb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl));
875 }
876
877 static int fl_set_geneve_opt(const struct nlattr *nla, struct fl_flow_key *key,
878                              int depth, int option_len,
879                              struct netlink_ext_ack *extack)
880 {
881         struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1];
882         struct nlattr *class = NULL, *type = NULL, *data = NULL;
883         struct geneve_opt *opt;
884         int err, data_len = 0;
885
886         if (option_len > sizeof(struct geneve_opt))
887                 data_len = option_len - sizeof(struct geneve_opt);
888
889         opt = (struct geneve_opt *)&key->enc_opts.data[key->enc_opts.len];
890         memset(opt, 0xff, option_len);
891         opt->length = data_len / 4;
892         opt->r1 = 0;
893         opt->r2 = 0;
894         opt->r3 = 0;
895
896         /* If no mask has been prodived we assume an exact match. */
897         if (!depth)
898                 return sizeof(struct geneve_opt) + data_len;
899
900         if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_GENEVE) {
901                 NL_SET_ERR_MSG(extack, "Non-geneve option type for mask");
902                 return -EINVAL;
903         }
904
905         err = nla_parse_nested_deprecated(tb,
906                                           TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX,
907                                           nla, geneve_opt_policy, extack);
908         if (err < 0)
909                 return err;
910
911         /* We are not allowed to omit any of CLASS, TYPE or DATA
912          * fields from the key.
913          */
914         if (!option_len &&
915             (!tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS] ||
916              !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE] ||
917              !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA])) {
918                 NL_SET_ERR_MSG(extack, "Missing tunnel key geneve option class, type or data");
919                 return -EINVAL;
920         }
921
922         /* Omitting any of CLASS, TYPE or DATA fields is allowed
923          * for the mask.
924          */
925         if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA]) {
926                 int new_len = key->enc_opts.len;
927
928                 data = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA];
929                 data_len = nla_len(data);
930                 if (data_len < 4) {
931                         NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is less than 4 bytes long");
932                         return -ERANGE;
933                 }
934                 if (data_len % 4) {
935                         NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is not a multiple of 4 bytes long");
936                         return -ERANGE;
937                 }
938
939                 new_len += sizeof(struct geneve_opt) + data_len;
940                 BUILD_BUG_ON(FLOW_DIS_TUN_OPTS_MAX != IP_TUNNEL_OPTS_MAX);
941                 if (new_len > FLOW_DIS_TUN_OPTS_MAX) {
942                         NL_SET_ERR_MSG(extack, "Tunnel options exceeds max size");
943                         return -ERANGE;
944                 }
945                 opt->length = data_len / 4;
946                 memcpy(opt->opt_data, nla_data(data), data_len);
947         }
948
949         if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS]) {
950                 class = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS];
951                 opt->opt_class = nla_get_be16(class);
952         }
953
954         if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE]) {
955                 type = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE];
956                 opt->type = nla_get_u8(type);
957         }
958
959         return sizeof(struct geneve_opt) + data_len;
960 }
961
962 static int fl_set_enc_opt(struct nlattr **tb, struct fl_flow_key *key,
963                           struct fl_flow_key *mask,
964                           struct netlink_ext_ack *extack)
965 {
966         const struct nlattr *nla_enc_key, *nla_opt_key, *nla_opt_msk = NULL;
967         int err, option_len, key_depth, msk_depth = 0;
968
969         err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS],
970                                              TCA_FLOWER_KEY_ENC_OPTS_MAX,
971                                              enc_opts_policy, extack);
972         if (err)
973                 return err;
974
975         nla_enc_key = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS]);
976
977         if (tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]) {
978                 err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK],
979                                                      TCA_FLOWER_KEY_ENC_OPTS_MAX,
980                                                      enc_opts_policy, extack);
981                 if (err)
982                         return err;
983
984                 nla_opt_msk = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
985                 msk_depth = nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]);
986         }
987
988         nla_for_each_attr(nla_opt_key, nla_enc_key,
989                           nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS]), key_depth) {
990                 switch (nla_type(nla_opt_key)) {
991                 case TCA_FLOWER_KEY_ENC_OPTS_GENEVE:
992                         option_len = 0;
993                         key->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT;
994                         option_len = fl_set_geneve_opt(nla_opt_key, key,
995                                                        key_depth, option_len,
996                                                        extack);
997                         if (option_len < 0)
998                                 return option_len;
999
1000                         key->enc_opts.len += option_len;
1001                         /* At the same time we need to parse through the mask
1002                          * in order to verify exact and mask attribute lengths.
1003                          */
1004                         mask->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT;
1005                         option_len = fl_set_geneve_opt(nla_opt_msk, mask,
1006                                                        msk_depth, option_len,
1007                                                        extack);
1008                         if (option_len < 0)
1009                                 return option_len;
1010
1011                         mask->enc_opts.len += option_len;
1012                         if (key->enc_opts.len != mask->enc_opts.len) {
1013                                 NL_SET_ERR_MSG(extack, "Key and mask miss aligned");
1014                                 return -EINVAL;
1015                         }
1016
1017                         if (msk_depth)
1018                                 nla_opt_msk = nla_next(nla_opt_msk, &msk_depth);
1019                         break;
1020                 default:
1021                         NL_SET_ERR_MSG(extack, "Unknown tunnel option type");
1022                         return -EINVAL;
1023                 }
1024         }
1025
1026         return 0;
1027 }
1028
1029 static int fl_set_key_ct(struct nlattr **tb,
1030                          struct flow_dissector_key_ct *key,
1031                          struct flow_dissector_key_ct *mask,
1032                          struct netlink_ext_ack *extack)
1033 {
1034         if (tb[TCA_FLOWER_KEY_CT_STATE]) {
1035                 if (!IS_ENABLED(CONFIG_NF_CONNTRACK)) {
1036                         NL_SET_ERR_MSG(extack, "Conntrack isn't enabled");
1037                         return -EOPNOTSUPP;
1038                 }
1039                 fl_set_key_val(tb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE,
1040                                &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK,
1041                                sizeof(key->ct_state));
1042         }
1043         if (tb[TCA_FLOWER_KEY_CT_ZONE]) {
1044                 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)) {
1045                         NL_SET_ERR_MSG(extack, "Conntrack zones isn't enabled");
1046                         return -EOPNOTSUPP;
1047                 }
1048                 fl_set_key_val(tb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE,
1049                                &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK,
1050                                sizeof(key->ct_zone));
1051         }
1052         if (tb[TCA_FLOWER_KEY_CT_MARK]) {
1053                 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)) {
1054                         NL_SET_ERR_MSG(extack, "Conntrack mark isn't enabled");
1055                         return -EOPNOTSUPP;
1056                 }
1057                 fl_set_key_val(tb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK,
1058                                &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK,
1059                                sizeof(key->ct_mark));
1060         }
1061         if (tb[TCA_FLOWER_KEY_CT_LABELS]) {
1062                 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)) {
1063                         NL_SET_ERR_MSG(extack, "Conntrack labels aren't enabled");
1064                         return -EOPNOTSUPP;
1065                 }
1066                 fl_set_key_val(tb, key->ct_labels, TCA_FLOWER_KEY_CT_LABELS,
1067                                mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK,
1068                                sizeof(key->ct_labels));
1069         }
1070
1071         return 0;
1072 }
1073
1074 static int fl_set_key(struct net *net, struct nlattr **tb,
1075                       struct fl_flow_key *key, struct fl_flow_key *mask,
1076                       struct netlink_ext_ack *extack)
1077 {
1078         __be16 ethertype;
1079         int ret = 0;
1080
1081         if (tb[TCA_FLOWER_INDEV]) {
1082                 int err = tcf_change_indev(net, tb[TCA_FLOWER_INDEV], extack);
1083                 if (err < 0)
1084                         return err;
1085                 key->meta.ingress_ifindex = err;
1086                 mask->meta.ingress_ifindex = 0xffffffff;
1087         }
1088
1089         fl_set_key_val(tb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
1090                        mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
1091                        sizeof(key->eth.dst));
1092         fl_set_key_val(tb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
1093                        mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
1094                        sizeof(key->eth.src));
1095
1096         if (tb[TCA_FLOWER_KEY_ETH_TYPE]) {
1097                 ethertype = nla_get_be16(tb[TCA_FLOWER_KEY_ETH_TYPE]);
1098
1099                 if (eth_type_vlan(ethertype)) {
1100                         fl_set_key_vlan(tb, ethertype, TCA_FLOWER_KEY_VLAN_ID,
1101                                         TCA_FLOWER_KEY_VLAN_PRIO, &key->vlan,
1102                                         &mask->vlan);
1103
1104                         if (tb[TCA_FLOWER_KEY_VLAN_ETH_TYPE]) {
1105                                 ethertype = nla_get_be16(tb[TCA_FLOWER_KEY_VLAN_ETH_TYPE]);
1106                                 if (eth_type_vlan(ethertype)) {
1107                                         fl_set_key_vlan(tb, ethertype,
1108                                                         TCA_FLOWER_KEY_CVLAN_ID,
1109                                                         TCA_FLOWER_KEY_CVLAN_PRIO,
1110                                                         &key->cvlan, &mask->cvlan);
1111                                         fl_set_key_val(tb, &key->basic.n_proto,
1112                                                        TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
1113                                                        &mask->basic.n_proto,
1114                                                        TCA_FLOWER_UNSPEC,
1115                                                        sizeof(key->basic.n_proto));
1116                                 } else {
1117                                         key->basic.n_proto = ethertype;
1118                                         mask->basic.n_proto = cpu_to_be16(~0);
1119                                 }
1120                         }
1121                 } else {
1122                         key->basic.n_proto = ethertype;
1123                         mask->basic.n_proto = cpu_to_be16(~0);
1124                 }
1125         }
1126
1127         if (key->basic.n_proto == htons(ETH_P_IP) ||
1128             key->basic.n_proto == htons(ETH_P_IPV6)) {
1129                 fl_set_key_val(tb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
1130                                &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
1131                                sizeof(key->basic.ip_proto));
1132                 fl_set_key_ip(tb, false, &key->ip, &mask->ip);
1133         }
1134
1135         if (tb[TCA_FLOWER_KEY_IPV4_SRC] || tb[TCA_FLOWER_KEY_IPV4_DST]) {
1136                 key->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1137                 mask->control.addr_type = ~0;
1138                 fl_set_key_val(tb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
1139                                &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
1140                                sizeof(key->ipv4.src));
1141                 fl_set_key_val(tb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
1142                                &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
1143                                sizeof(key->ipv4.dst));
1144         } else if (tb[TCA_FLOWER_KEY_IPV6_SRC] || tb[TCA_FLOWER_KEY_IPV6_DST]) {
1145                 key->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1146                 mask->control.addr_type = ~0;
1147                 fl_set_key_val(tb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
1148                                &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
1149                                sizeof(key->ipv6.src));
1150                 fl_set_key_val(tb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
1151                                &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
1152                                sizeof(key->ipv6.dst));
1153         }
1154
1155         if (key->basic.ip_proto == IPPROTO_TCP) {
1156                 fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
1157                                &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
1158                                sizeof(key->tp.src));
1159                 fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
1160                                &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
1161                                sizeof(key->tp.dst));
1162                 fl_set_key_val(tb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
1163                                &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
1164                                sizeof(key->tcp.flags));
1165         } else if (key->basic.ip_proto == IPPROTO_UDP) {
1166                 fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
1167                                &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
1168                                sizeof(key->tp.src));
1169                 fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
1170                                &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
1171                                sizeof(key->tp.dst));
1172         } else if (key->basic.ip_proto == IPPROTO_SCTP) {
1173                 fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
1174                                &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
1175                                sizeof(key->tp.src));
1176                 fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
1177                                &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
1178                                sizeof(key->tp.dst));
1179         } else if (key->basic.n_proto == htons(ETH_P_IP) &&
1180                    key->basic.ip_proto == IPPROTO_ICMP) {
1181                 fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV4_TYPE,
1182                                &mask->icmp.type,
1183                                TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
1184                                sizeof(key->icmp.type));
1185                 fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV4_CODE,
1186                                &mask->icmp.code,
1187                                TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
1188                                sizeof(key->icmp.code));
1189         } else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
1190                    key->basic.ip_proto == IPPROTO_ICMPV6) {
1191                 fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV6_TYPE,
1192                                &mask->icmp.type,
1193                                TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
1194                                sizeof(key->icmp.type));
1195                 fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV6_CODE,
1196                                &mask->icmp.code,
1197                                TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
1198                                sizeof(key->icmp.code));
1199         } else if (key->basic.n_proto == htons(ETH_P_MPLS_UC) ||
1200                    key->basic.n_proto == htons(ETH_P_MPLS_MC)) {
1201                 ret = fl_set_key_mpls(tb, &key->mpls, &mask->mpls);
1202                 if (ret)
1203                         return ret;
1204         } else if (key->basic.n_proto == htons(ETH_P_ARP) ||
1205                    key->basic.n_proto == htons(ETH_P_RARP)) {
1206                 fl_set_key_val(tb, &key->arp.sip, TCA_FLOWER_KEY_ARP_SIP,
1207                                &mask->arp.sip, TCA_FLOWER_KEY_ARP_SIP_MASK,
1208                                sizeof(key->arp.sip));
1209                 fl_set_key_val(tb, &key->arp.tip, TCA_FLOWER_KEY_ARP_TIP,
1210                                &mask->arp.tip, TCA_FLOWER_KEY_ARP_TIP_MASK,
1211                                sizeof(key->arp.tip));
1212                 fl_set_key_val(tb, &key->arp.op, TCA_FLOWER_KEY_ARP_OP,
1213                                &mask->arp.op, TCA_FLOWER_KEY_ARP_OP_MASK,
1214                                sizeof(key->arp.op));
1215                 fl_set_key_val(tb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
1216                                mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
1217                                sizeof(key->arp.sha));
1218                 fl_set_key_val(tb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
1219                                mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
1220                                sizeof(key->arp.tha));
1221         }
1222
1223         if (key->basic.ip_proto == IPPROTO_TCP ||
1224             key->basic.ip_proto == IPPROTO_UDP ||
1225             key->basic.ip_proto == IPPROTO_SCTP) {
1226                 ret = fl_set_key_port_range(tb, key, mask);
1227                 if (ret)
1228                         return ret;
1229         }
1230
1231         if (tb[TCA_FLOWER_KEY_ENC_IPV4_SRC] ||
1232             tb[TCA_FLOWER_KEY_ENC_IPV4_DST]) {
1233                 key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1234                 mask->enc_control.addr_type = ~0;
1235                 fl_set_key_val(tb, &key->enc_ipv4.src,
1236                                TCA_FLOWER_KEY_ENC_IPV4_SRC,
1237                                &mask->enc_ipv4.src,
1238                                TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
1239                                sizeof(key->enc_ipv4.src));
1240                 fl_set_key_val(tb, &key->enc_ipv4.dst,
1241                                TCA_FLOWER_KEY_ENC_IPV4_DST,
1242                                &mask->enc_ipv4.dst,
1243                                TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
1244                                sizeof(key->enc_ipv4.dst));
1245         }
1246
1247         if (tb[TCA_FLOWER_KEY_ENC_IPV6_SRC] ||
1248             tb[TCA_FLOWER_KEY_ENC_IPV6_DST]) {
1249                 key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1250                 mask->enc_control.addr_type = ~0;
1251                 fl_set_key_val(tb, &key->enc_ipv6.src,
1252                                TCA_FLOWER_KEY_ENC_IPV6_SRC,
1253                                &mask->enc_ipv6.src,
1254                                TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
1255                                sizeof(key->enc_ipv6.src));
1256                 fl_set_key_val(tb, &key->enc_ipv6.dst,
1257                                TCA_FLOWER_KEY_ENC_IPV6_DST,
1258                                &mask->enc_ipv6.dst,
1259                                TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
1260                                sizeof(key->enc_ipv6.dst));
1261         }
1262
1263         fl_set_key_val(tb, &key->enc_key_id.keyid, TCA_FLOWER_KEY_ENC_KEY_ID,
1264                        &mask->enc_key_id.keyid, TCA_FLOWER_UNSPEC,
1265                        sizeof(key->enc_key_id.keyid));
1266
1267         fl_set_key_val(tb, &key->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
1268                        &mask->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
1269                        sizeof(key->enc_tp.src));
1270
1271         fl_set_key_val(tb, &key->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
1272                        &mask->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
1273                        sizeof(key->enc_tp.dst));
1274
1275         fl_set_key_ip(tb, true, &key->enc_ip, &mask->enc_ip);
1276
1277         if (tb[TCA_FLOWER_KEY_ENC_OPTS]) {
1278                 ret = fl_set_enc_opt(tb, key, mask, extack);
1279                 if (ret)
1280                         return ret;
1281         }
1282
1283         ret = fl_set_key_ct(tb, &key->ct, &mask->ct, extack);
1284         if (ret)
1285                 return ret;
1286
1287         if (tb[TCA_FLOWER_KEY_FLAGS])
1288                 ret = fl_set_key_flags(tb, &key->control.flags, &mask->control.flags);
1289
1290         return ret;
1291 }
1292
1293 static void fl_mask_copy(struct fl_flow_mask *dst,
1294                          struct fl_flow_mask *src)
1295 {
1296         const void *psrc = fl_key_get_start(&src->key, src);
1297         void *pdst = fl_key_get_start(&dst->key, src);
1298
1299         memcpy(pdst, psrc, fl_mask_range(src));
1300         dst->range = src->range;
1301 }
1302
1303 static const struct rhashtable_params fl_ht_params = {
1304         .key_offset = offsetof(struct cls_fl_filter, mkey), /* base offset */
1305         .head_offset = offsetof(struct cls_fl_filter, ht_node),
1306         .automatic_shrinking = true,
1307 };
1308
1309 static int fl_init_mask_hashtable(struct fl_flow_mask *mask)
1310 {
1311         mask->filter_ht_params = fl_ht_params;
1312         mask->filter_ht_params.key_len = fl_mask_range(mask);
1313         mask->filter_ht_params.key_offset += mask->range.start;
1314
1315         return rhashtable_init(&mask->ht, &mask->filter_ht_params);
1316 }
1317
1318 #define FL_KEY_MEMBER_OFFSET(member) offsetof(struct fl_flow_key, member)
1319 #define FL_KEY_MEMBER_SIZE(member) FIELD_SIZEOF(struct fl_flow_key, member)
1320
1321 #define FL_KEY_IS_MASKED(mask, member)                                          \
1322         memchr_inv(((char *)mask) + FL_KEY_MEMBER_OFFSET(member),               \
1323                    0, FL_KEY_MEMBER_SIZE(member))                               \
1324
1325 #define FL_KEY_SET(keys, cnt, id, member)                                       \
1326         do {                                                                    \
1327                 keys[cnt].key_id = id;                                          \
1328                 keys[cnt].offset = FL_KEY_MEMBER_OFFSET(member);                \
1329                 cnt++;                                                          \
1330         } while(0);
1331
1332 #define FL_KEY_SET_IF_MASKED(mask, keys, cnt, id, member)                       \
1333         do {                                                                    \
1334                 if (FL_KEY_IS_MASKED(mask, member))                             \
1335                         FL_KEY_SET(keys, cnt, id, member);                      \
1336         } while(0);
1337
1338 static void fl_init_dissector(struct flow_dissector *dissector,
1339                               struct fl_flow_key *mask)
1340 {
1341         struct flow_dissector_key keys[FLOW_DISSECTOR_KEY_MAX];
1342         size_t cnt = 0;
1343
1344         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1345                              FLOW_DISSECTOR_KEY_META, meta);
1346         FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_CONTROL, control);
1347         FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_BASIC, basic);
1348         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1349                              FLOW_DISSECTOR_KEY_ETH_ADDRS, eth);
1350         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1351                              FLOW_DISSECTOR_KEY_IPV4_ADDRS, ipv4);
1352         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1353                              FLOW_DISSECTOR_KEY_IPV6_ADDRS, ipv6);
1354         if (FL_KEY_IS_MASKED(mask, tp) ||
1355             FL_KEY_IS_MASKED(mask, tp_min) || FL_KEY_IS_MASKED(mask, tp_max))
1356                 FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_PORTS, tp);
1357         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1358                              FLOW_DISSECTOR_KEY_IP, ip);
1359         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1360                              FLOW_DISSECTOR_KEY_TCP, tcp);
1361         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1362                              FLOW_DISSECTOR_KEY_ICMP, icmp);
1363         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1364                              FLOW_DISSECTOR_KEY_ARP, arp);
1365         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1366                              FLOW_DISSECTOR_KEY_MPLS, mpls);
1367         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1368                              FLOW_DISSECTOR_KEY_VLAN, vlan);
1369         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1370                              FLOW_DISSECTOR_KEY_CVLAN, cvlan);
1371         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1372                              FLOW_DISSECTOR_KEY_ENC_KEYID, enc_key_id);
1373         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1374                              FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, enc_ipv4);
1375         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1376                              FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, enc_ipv6);
1377         if (FL_KEY_IS_MASKED(mask, enc_ipv4) ||
1378             FL_KEY_IS_MASKED(mask, enc_ipv6))
1379                 FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_ENC_CONTROL,
1380                            enc_control);
1381         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1382                              FLOW_DISSECTOR_KEY_ENC_PORTS, enc_tp);
1383         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1384                              FLOW_DISSECTOR_KEY_ENC_IP, enc_ip);
1385         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1386                              FLOW_DISSECTOR_KEY_ENC_OPTS, enc_opts);
1387         FL_KEY_SET_IF_MASKED(mask, keys, cnt,
1388                              FLOW_DISSECTOR_KEY_CT, ct);
1389
1390         skb_flow_dissector_init(dissector, keys, cnt);
1391 }
1392
1393 static struct fl_flow_mask *fl_create_new_mask(struct cls_fl_head *head,
1394                                                struct fl_flow_mask *mask)
1395 {
1396         struct fl_flow_mask *newmask;
1397         int err;
1398
1399         newmask = kzalloc(sizeof(*newmask), GFP_KERNEL);
1400         if (!newmask)
1401                 return ERR_PTR(-ENOMEM);
1402
1403         fl_mask_copy(newmask, mask);
1404
1405         if ((newmask->key.tp_min.dst && newmask->key.tp_max.dst) ||
1406             (newmask->key.tp_min.src && newmask->key.tp_max.src))
1407                 newmask->flags |= TCA_FLOWER_MASK_FLAGS_RANGE;
1408
1409         err = fl_init_mask_hashtable(newmask);
1410         if (err)
1411                 goto errout_free;
1412
1413         fl_init_dissector(&newmask->dissector, &newmask->key);
1414
1415         INIT_LIST_HEAD_RCU(&newmask->filters);
1416
1417         refcount_set(&newmask->refcnt, 1);
1418         err = rhashtable_replace_fast(&head->ht, &mask->ht_node,
1419                                       &newmask->ht_node, mask_ht_params);
1420         if (err)
1421                 goto errout_destroy;
1422
1423         spin_lock(&head->masks_lock);
1424         list_add_tail_rcu(&newmask->list, &head->masks);
1425         spin_unlock(&head->masks_lock);
1426
1427         return newmask;
1428
1429 errout_destroy:
1430         rhashtable_destroy(&newmask->ht);
1431 errout_free:
1432         kfree(newmask);
1433
1434         return ERR_PTR(err);
1435 }
1436
1437 static int fl_check_assign_mask(struct cls_fl_head *head,
1438                                 struct cls_fl_filter *fnew,
1439                                 struct cls_fl_filter *fold,
1440                                 struct fl_flow_mask *mask)
1441 {
1442         struct fl_flow_mask *newmask;
1443         int ret = 0;
1444
1445         rcu_read_lock();
1446
1447         /* Insert mask as temporary node to prevent concurrent creation of mask
1448          * with same key. Any concurrent lookups with same key will return
1449          * -EAGAIN because mask's refcnt is zero.
1450          */
1451         fnew->mask = rhashtable_lookup_get_insert_fast(&head->ht,
1452                                                        &mask->ht_node,
1453                                                        mask_ht_params);
1454         if (!fnew->mask) {
1455                 rcu_read_unlock();
1456
1457                 if (fold) {
1458                         ret = -EINVAL;
1459                         goto errout_cleanup;
1460                 }
1461
1462                 newmask = fl_create_new_mask(head, mask);
1463                 if (IS_ERR(newmask)) {
1464                         ret = PTR_ERR(newmask);
1465                         goto errout_cleanup;
1466                 }
1467
1468                 fnew->mask = newmask;
1469                 return 0;
1470         } else if (IS_ERR(fnew->mask)) {
1471                 ret = PTR_ERR(fnew->mask);
1472         } else if (fold && fold->mask != fnew->mask) {
1473                 ret = -EINVAL;
1474         } else if (!refcount_inc_not_zero(&fnew->mask->refcnt)) {
1475                 /* Mask was deleted concurrently, try again */
1476                 ret = -EAGAIN;
1477         }
1478         rcu_read_unlock();
1479         return ret;
1480
1481 errout_cleanup:
1482         rhashtable_remove_fast(&head->ht, &mask->ht_node,
1483                                mask_ht_params);
1484         return ret;
1485 }
1486
1487 static int fl_set_parms(struct net *net, struct tcf_proto *tp,
1488                         struct cls_fl_filter *f, struct fl_flow_mask *mask,
1489                         unsigned long base, struct nlattr **tb,
1490                         struct nlattr *est, bool ovr,
1491                         struct fl_flow_tmplt *tmplt, bool rtnl_held,
1492                         struct netlink_ext_ack *extack)
1493 {
1494         int err;
1495
1496         err = tcf_exts_validate(net, tp, tb, est, &f->exts, ovr, rtnl_held,
1497                                 extack);
1498         if (err < 0)
1499                 return err;
1500
1501         if (tb[TCA_FLOWER_CLASSID]) {
1502                 f->res.classid = nla_get_u32(tb[TCA_FLOWER_CLASSID]);
1503                 if (!rtnl_held)
1504                         rtnl_lock();
1505                 tcf_bind_filter(tp, &f->res, base);
1506                 if (!rtnl_held)
1507                         rtnl_unlock();
1508         }
1509
1510         err = fl_set_key(net, tb, &f->key, &mask->key, extack);
1511         if (err)
1512                 return err;
1513
1514         fl_mask_update_range(mask);
1515         fl_set_masked_key(&f->mkey, &f->key, mask);
1516
1517         if (!fl_mask_fits_tmplt(tmplt, mask)) {
1518                 NL_SET_ERR_MSG_MOD(extack, "Mask does not fit the template");
1519                 return -EINVAL;
1520         }
1521
1522         return 0;
1523 }
1524
1525 static int fl_ht_insert_unique(struct cls_fl_filter *fnew,
1526                                struct cls_fl_filter *fold,
1527                                bool *in_ht)
1528 {
1529         struct fl_flow_mask *mask = fnew->mask;
1530         int err;
1531
1532         err = rhashtable_lookup_insert_fast(&mask->ht,
1533                                             &fnew->ht_node,
1534                                             mask->filter_ht_params);
1535         if (err) {
1536                 *in_ht = false;
1537                 /* It is okay if filter with same key exists when
1538                  * overwriting.
1539                  */
1540                 return fold && err == -EEXIST ? 0 : err;
1541         }
1542
1543         *in_ht = true;
1544         return 0;
1545 }
1546
1547 static int fl_change(struct net *net, struct sk_buff *in_skb,
1548                      struct tcf_proto *tp, unsigned long base,
1549                      u32 handle, struct nlattr **tca,
1550                      void **arg, bool ovr, bool rtnl_held,
1551                      struct netlink_ext_ack *extack)
1552 {
1553         struct cls_fl_head *head = fl_head_dereference(tp);
1554         struct cls_fl_filter *fold = *arg;
1555         struct cls_fl_filter *fnew;
1556         struct fl_flow_mask *mask;
1557         struct nlattr **tb;
1558         bool in_ht;
1559         int err;
1560
1561         if (!tca[TCA_OPTIONS]) {
1562                 err = -EINVAL;
1563                 goto errout_fold;
1564         }
1565
1566         mask = kzalloc(sizeof(struct fl_flow_mask), GFP_KERNEL);
1567         if (!mask) {
1568                 err = -ENOBUFS;
1569                 goto errout_fold;
1570         }
1571
1572         tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
1573         if (!tb) {
1574                 err = -ENOBUFS;
1575                 goto errout_mask_alloc;
1576         }
1577
1578         err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
1579                                           tca[TCA_OPTIONS], fl_policy, NULL);
1580         if (err < 0)
1581                 goto errout_tb;
1582
1583         if (fold && handle && fold->handle != handle) {
1584                 err = -EINVAL;
1585                 goto errout_tb;
1586         }
1587
1588         fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
1589         if (!fnew) {
1590                 err = -ENOBUFS;
1591                 goto errout_tb;
1592         }
1593         INIT_LIST_HEAD(&fnew->hw_list);
1594         refcount_set(&fnew->refcnt, 1);
1595
1596         err = tcf_exts_init(&fnew->exts, net, TCA_FLOWER_ACT, 0);
1597         if (err < 0)
1598                 goto errout;
1599
1600         if (tb[TCA_FLOWER_FLAGS]) {
1601                 fnew->flags = nla_get_u32(tb[TCA_FLOWER_FLAGS]);
1602
1603                 if (!tc_flags_valid(fnew->flags)) {
1604                         err = -EINVAL;
1605                         goto errout;
1606                 }
1607         }
1608
1609         err = fl_set_parms(net, tp, fnew, mask, base, tb, tca[TCA_RATE], ovr,
1610                            tp->chain->tmplt_priv, rtnl_held, extack);
1611         if (err)
1612                 goto errout;
1613
1614         err = fl_check_assign_mask(head, fnew, fold, mask);
1615         if (err)
1616                 goto errout;
1617
1618         err = fl_ht_insert_unique(fnew, fold, &in_ht);
1619         if (err)
1620                 goto errout_mask;
1621
1622         if (!tc_skip_hw(fnew->flags)) {
1623                 err = fl_hw_replace_filter(tp, fnew, rtnl_held, extack);
1624                 if (err)
1625                         goto errout_ht;
1626         }
1627
1628         if (!tc_in_hw(fnew->flags))
1629                 fnew->flags |= TCA_CLS_FLAGS_NOT_IN_HW;
1630
1631         spin_lock(&tp->lock);
1632
1633         /* tp was deleted concurrently. -EAGAIN will cause caller to lookup
1634          * proto again or create new one, if necessary.
1635          */
1636         if (tp->deleting) {
1637                 err = -EAGAIN;
1638                 goto errout_hw;
1639         }
1640
1641         if (fold) {
1642                 /* Fold filter was deleted concurrently. Retry lookup. */
1643                 if (fold->deleted) {
1644                         err = -EAGAIN;
1645                         goto errout_hw;
1646                 }
1647
1648                 fnew->handle = handle;
1649
1650                 if (!in_ht) {
1651                         struct rhashtable_params params =
1652                                 fnew->mask->filter_ht_params;
1653
1654                         err = rhashtable_insert_fast(&fnew->mask->ht,
1655                                                      &fnew->ht_node,
1656                                                      params);
1657                         if (err)
1658                                 goto errout_hw;
1659                         in_ht = true;
1660                 }
1661
1662                 refcount_inc(&fnew->refcnt);
1663                 rhashtable_remove_fast(&fold->mask->ht,
1664                                        &fold->ht_node,
1665                                        fold->mask->filter_ht_params);
1666                 idr_replace(&head->handle_idr, fnew, fnew->handle);
1667                 list_replace_rcu(&fold->list, &fnew->list);
1668                 fold->deleted = true;
1669
1670                 spin_unlock(&tp->lock);
1671
1672                 fl_mask_put(head, fold->mask);
1673                 if (!tc_skip_hw(fold->flags))
1674                         fl_hw_destroy_filter(tp, fold, rtnl_held, NULL);
1675                 tcf_unbind_filter(tp, &fold->res);
1676                 /* Caller holds reference to fold, so refcnt is always > 0
1677                  * after this.
1678                  */
1679                 refcount_dec(&fold->refcnt);
1680                 __fl_put(fold);
1681         } else {
1682                 if (handle) {
1683                         /* user specifies a handle and it doesn't exist */
1684                         err = idr_alloc_u32(&head->handle_idr, fnew, &handle,
1685                                             handle, GFP_ATOMIC);
1686
1687                         /* Filter with specified handle was concurrently
1688                          * inserted after initial check in cls_api. This is not
1689                          * necessarily an error if NLM_F_EXCL is not set in
1690                          * message flags. Returning EAGAIN will cause cls_api to
1691                          * try to update concurrently inserted rule.
1692                          */
1693                         if (err == -ENOSPC)
1694                                 err = -EAGAIN;
1695                 } else {
1696                         handle = 1;
1697                         err = idr_alloc_u32(&head->handle_idr, fnew, &handle,
1698                                             INT_MAX, GFP_ATOMIC);
1699                 }
1700                 if (err)
1701                         goto errout_hw;
1702
1703                 refcount_inc(&fnew->refcnt);
1704                 fnew->handle = handle;
1705                 list_add_tail_rcu(&fnew->list, &fnew->mask->filters);
1706                 spin_unlock(&tp->lock);
1707         }
1708
1709         *arg = fnew;
1710
1711         kfree(tb);
1712         tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work);
1713         return 0;
1714
1715 errout_ht:
1716         spin_lock(&tp->lock);
1717 errout_hw:
1718         fnew->deleted = true;
1719         spin_unlock(&tp->lock);
1720         if (!tc_skip_hw(fnew->flags))
1721                 fl_hw_destroy_filter(tp, fnew, rtnl_held, NULL);
1722         if (in_ht)
1723                 rhashtable_remove_fast(&fnew->mask->ht, &fnew->ht_node,
1724                                        fnew->mask->filter_ht_params);
1725 errout_mask:
1726         fl_mask_put(head, fnew->mask);
1727 errout:
1728         __fl_put(fnew);
1729 errout_tb:
1730         kfree(tb);
1731 errout_mask_alloc:
1732         tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work);
1733 errout_fold:
1734         if (fold)
1735                 __fl_put(fold);
1736         return err;
1737 }
1738
1739 static int fl_delete(struct tcf_proto *tp, void *arg, bool *last,
1740                      bool rtnl_held, struct netlink_ext_ack *extack)
1741 {
1742         struct cls_fl_head *head = fl_head_dereference(tp);
1743         struct cls_fl_filter *f = arg;
1744         bool last_on_mask;
1745         int err = 0;
1746
1747         err = __fl_delete(tp, f, &last_on_mask, rtnl_held, extack);
1748         *last = list_empty(&head->masks);
1749         __fl_put(f);
1750
1751         return err;
1752 }
1753
1754 static void fl_walk(struct tcf_proto *tp, struct tcf_walker *arg,
1755                     bool rtnl_held)
1756 {
1757         struct cls_fl_head *head = fl_head_dereference(tp);
1758         unsigned long id = arg->cookie, tmp;
1759         struct cls_fl_filter *f;
1760
1761         arg->count = arg->skip;
1762
1763         idr_for_each_entry_continue_ul(&head->handle_idr, f, tmp, id) {
1764                 /* don't return filters that are being deleted */
1765                 if (!refcount_inc_not_zero(&f->refcnt))
1766                         continue;
1767                 if (arg->fn(tp, f, arg) < 0) {
1768                         __fl_put(f);
1769                         arg->stop = 1;
1770                         break;
1771                 }
1772                 __fl_put(f);
1773                 arg->count++;
1774         }
1775         arg->cookie = id;
1776 }
1777
1778 static struct cls_fl_filter *
1779 fl_get_next_hw_filter(struct tcf_proto *tp, struct cls_fl_filter *f, bool add)
1780 {
1781         struct cls_fl_head *head = fl_head_dereference(tp);
1782
1783         spin_lock(&tp->lock);
1784         if (list_empty(&head->hw_filters)) {
1785                 spin_unlock(&tp->lock);
1786                 return NULL;
1787         }
1788
1789         if (!f)
1790                 f = list_entry(&head->hw_filters, struct cls_fl_filter,
1791                                hw_list);
1792         list_for_each_entry_continue(f, &head->hw_filters, hw_list) {
1793                 if (!(add && f->deleted) && refcount_inc_not_zero(&f->refcnt)) {
1794                         spin_unlock(&tp->lock);
1795                         return f;
1796                 }
1797         }
1798
1799         spin_unlock(&tp->lock);
1800         return NULL;
1801 }
1802
1803 static int fl_reoffload(struct tcf_proto *tp, bool add, flow_setup_cb_t *cb,
1804                         void *cb_priv, struct netlink_ext_ack *extack)
1805 {
1806         struct tcf_block *block = tp->chain->block;
1807         struct flow_cls_offload cls_flower = {};
1808         struct cls_fl_filter *f = NULL;
1809         int err;
1810
1811         /* hw_filters list can only be changed by hw offload functions after
1812          * obtaining rtnl lock. Make sure it is not changed while reoffload is
1813          * iterating it.
1814          */
1815         ASSERT_RTNL();
1816
1817         while ((f = fl_get_next_hw_filter(tp, f, add))) {
1818                 cls_flower.rule =
1819                         flow_rule_alloc(tcf_exts_num_actions(&f->exts));
1820                 if (!cls_flower.rule) {
1821                         __fl_put(f);
1822                         return -ENOMEM;
1823                 }
1824
1825                 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags,
1826                                            extack);
1827                 cls_flower.command = add ?
1828                         FLOW_CLS_REPLACE : FLOW_CLS_DESTROY;
1829                 cls_flower.cookie = (unsigned long)f;
1830                 cls_flower.rule->match.dissector = &f->mask->dissector;
1831                 cls_flower.rule->match.mask = &f->mask->key;
1832                 cls_flower.rule->match.key = &f->mkey;
1833
1834                 err = tc_setup_flow_action(&cls_flower.rule->action, &f->exts);
1835                 if (err) {
1836                         kfree(cls_flower.rule);
1837                         if (tc_skip_sw(f->flags)) {
1838                                 NL_SET_ERR_MSG_MOD(extack, "Failed to setup flow action");
1839                                 __fl_put(f);
1840                                 return err;
1841                         }
1842                         goto next_flow;
1843                 }
1844
1845                 cls_flower.classid = f->res.classid;
1846
1847                 err = cb(TC_SETUP_CLSFLOWER, &cls_flower, cb_priv);
1848                 kfree(cls_flower.rule);
1849
1850                 if (err) {
1851                         if (add && tc_skip_sw(f->flags)) {
1852                                 __fl_put(f);
1853                                 return err;
1854                         }
1855                         goto next_flow;
1856                 }
1857
1858                 spin_lock(&tp->lock);
1859                 tc_cls_offload_cnt_update(block, &f->in_hw_count, &f->flags,
1860                                           add);
1861                 spin_unlock(&tp->lock);
1862 next_flow:
1863                 __fl_put(f);
1864         }
1865
1866         return 0;
1867 }
1868
1869 static int fl_hw_create_tmplt(struct tcf_chain *chain,
1870                               struct fl_flow_tmplt *tmplt)
1871 {
1872         struct flow_cls_offload cls_flower = {};
1873         struct tcf_block *block = chain->block;
1874
1875         cls_flower.rule = flow_rule_alloc(0);
1876         if (!cls_flower.rule)
1877                 return -ENOMEM;
1878
1879         cls_flower.common.chain_index = chain->index;
1880         cls_flower.command = FLOW_CLS_TMPLT_CREATE;
1881         cls_flower.cookie = (unsigned long) tmplt;
1882         cls_flower.rule->match.dissector = &tmplt->dissector;
1883         cls_flower.rule->match.mask = &tmplt->mask;
1884         cls_flower.rule->match.key = &tmplt->dummy_key;
1885
1886         /* We don't care if driver (any of them) fails to handle this
1887          * call. It serves just as a hint for it.
1888          */
1889         tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false);
1890         kfree(cls_flower.rule);
1891
1892         return 0;
1893 }
1894
1895 static void fl_hw_destroy_tmplt(struct tcf_chain *chain,
1896                                 struct fl_flow_tmplt *tmplt)
1897 {
1898         struct flow_cls_offload cls_flower = {};
1899         struct tcf_block *block = chain->block;
1900
1901         cls_flower.common.chain_index = chain->index;
1902         cls_flower.command = FLOW_CLS_TMPLT_DESTROY;
1903         cls_flower.cookie = (unsigned long) tmplt;
1904
1905         tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false);
1906 }
1907
1908 static void *fl_tmplt_create(struct net *net, struct tcf_chain *chain,
1909                              struct nlattr **tca,
1910                              struct netlink_ext_ack *extack)
1911 {
1912         struct fl_flow_tmplt *tmplt;
1913         struct nlattr **tb;
1914         int err;
1915
1916         if (!tca[TCA_OPTIONS])
1917                 return ERR_PTR(-EINVAL);
1918
1919         tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL);
1920         if (!tb)
1921                 return ERR_PTR(-ENOBUFS);
1922         err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX,
1923                                           tca[TCA_OPTIONS], fl_policy, NULL);
1924         if (err)
1925                 goto errout_tb;
1926
1927         tmplt = kzalloc(sizeof(*tmplt), GFP_KERNEL);
1928         if (!tmplt) {
1929                 err = -ENOMEM;
1930                 goto errout_tb;
1931         }
1932         tmplt->chain = chain;
1933         err = fl_set_key(net, tb, &tmplt->dummy_key, &tmplt->mask, extack);
1934         if (err)
1935                 goto errout_tmplt;
1936
1937         fl_init_dissector(&tmplt->dissector, &tmplt->mask);
1938
1939         err = fl_hw_create_tmplt(chain, tmplt);
1940         if (err)
1941                 goto errout_tmplt;
1942
1943         kfree(tb);
1944         return tmplt;
1945
1946 errout_tmplt:
1947         kfree(tmplt);
1948 errout_tb:
1949         kfree(tb);
1950         return ERR_PTR(err);
1951 }
1952
1953 static void fl_tmplt_destroy(void *tmplt_priv)
1954 {
1955         struct fl_flow_tmplt *tmplt = tmplt_priv;
1956
1957         fl_hw_destroy_tmplt(tmplt->chain, tmplt);
1958         kfree(tmplt);
1959 }
1960
1961 static int fl_dump_key_val(struct sk_buff *skb,
1962                            void *val, int val_type,
1963                            void *mask, int mask_type, int len)
1964 {
1965         int err;
1966
1967         if (!memchr_inv(mask, 0, len))
1968                 return 0;
1969         err = nla_put(skb, val_type, len, val);
1970         if (err)
1971                 return err;
1972         if (mask_type != TCA_FLOWER_UNSPEC) {
1973                 err = nla_put(skb, mask_type, len, mask);
1974                 if (err)
1975                         return err;
1976         }
1977         return 0;
1978 }
1979
1980 static int fl_dump_key_port_range(struct sk_buff *skb, struct fl_flow_key *key,
1981                                   struct fl_flow_key *mask)
1982 {
1983         if (fl_dump_key_val(skb, &key->tp_min.dst, TCA_FLOWER_KEY_PORT_DST_MIN,
1984                             &mask->tp_min.dst, TCA_FLOWER_UNSPEC,
1985                             sizeof(key->tp_min.dst)) ||
1986             fl_dump_key_val(skb, &key->tp_max.dst, TCA_FLOWER_KEY_PORT_DST_MAX,
1987                             &mask->tp_max.dst, TCA_FLOWER_UNSPEC,
1988                             sizeof(key->tp_max.dst)) ||
1989             fl_dump_key_val(skb, &key->tp_min.src, TCA_FLOWER_KEY_PORT_SRC_MIN,
1990                             &mask->tp_min.src, TCA_FLOWER_UNSPEC,
1991                             sizeof(key->tp_min.src)) ||
1992             fl_dump_key_val(skb, &key->tp_max.src, TCA_FLOWER_KEY_PORT_SRC_MAX,
1993                             &mask->tp_max.src, TCA_FLOWER_UNSPEC,
1994                             sizeof(key->tp_max.src)))
1995                 return -1;
1996
1997         return 0;
1998 }
1999
2000 static int fl_dump_key_mpls(struct sk_buff *skb,
2001                             struct flow_dissector_key_mpls *mpls_key,
2002                             struct flow_dissector_key_mpls *mpls_mask)
2003 {
2004         int err;
2005
2006         if (!memchr_inv(mpls_mask, 0, sizeof(*mpls_mask)))
2007                 return 0;
2008         if (mpls_mask->mpls_ttl) {
2009                 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TTL,
2010                                  mpls_key->mpls_ttl);
2011                 if (err)
2012                         return err;
2013         }
2014         if (mpls_mask->mpls_tc) {
2015                 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TC,
2016                                  mpls_key->mpls_tc);
2017                 if (err)
2018                         return err;
2019         }
2020         if (mpls_mask->mpls_label) {
2021                 err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_LABEL,
2022                                   mpls_key->mpls_label);
2023                 if (err)
2024                         return err;
2025         }
2026         if (mpls_mask->mpls_bos) {
2027                 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_BOS,
2028                                  mpls_key->mpls_bos);
2029                 if (err)
2030                         return err;
2031         }
2032         return 0;
2033 }
2034
2035 static int fl_dump_key_ip(struct sk_buff *skb, bool encap,
2036                           struct flow_dissector_key_ip *key,
2037                           struct flow_dissector_key_ip *mask)
2038 {
2039         int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS;
2040         int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL;
2041         int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK;
2042         int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK;
2043
2044         if (fl_dump_key_val(skb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos)) ||
2045             fl_dump_key_val(skb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl)))
2046                 return -1;
2047
2048         return 0;
2049 }
2050
2051 static int fl_dump_key_vlan(struct sk_buff *skb,
2052                             int vlan_id_key, int vlan_prio_key,
2053                             struct flow_dissector_key_vlan *vlan_key,
2054                             struct flow_dissector_key_vlan *vlan_mask)
2055 {
2056         int err;
2057
2058         if (!memchr_inv(vlan_mask, 0, sizeof(*vlan_mask)))
2059                 return 0;
2060         if (vlan_mask->vlan_id) {
2061                 err = nla_put_u16(skb, vlan_id_key,
2062                                   vlan_key->vlan_id);
2063                 if (err)
2064                         return err;
2065         }
2066         if (vlan_mask->vlan_priority) {
2067                 err = nla_put_u8(skb, vlan_prio_key,
2068                                  vlan_key->vlan_priority);
2069                 if (err)
2070                         return err;
2071         }
2072         return 0;
2073 }
2074
2075 static void fl_get_key_flag(u32 dissector_key, u32 dissector_mask,
2076                             u32 *flower_key, u32 *flower_mask,
2077                             u32 flower_flag_bit, u32 dissector_flag_bit)
2078 {
2079         if (dissector_mask & dissector_flag_bit) {
2080                 *flower_mask |= flower_flag_bit;
2081                 if (dissector_key & dissector_flag_bit)
2082                         *flower_key |= flower_flag_bit;
2083         }
2084 }
2085
2086 static int fl_dump_key_flags(struct sk_buff *skb, u32 flags_key, u32 flags_mask)
2087 {
2088         u32 key, mask;
2089         __be32 _key, _mask;
2090         int err;
2091
2092         if (!memchr_inv(&flags_mask, 0, sizeof(flags_mask)))
2093                 return 0;
2094
2095         key = 0;
2096         mask = 0;
2097
2098         fl_get_key_flag(flags_key, flags_mask, &key, &mask,
2099                         TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT);
2100         fl_get_key_flag(flags_key, flags_mask, &key, &mask,
2101                         TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST,
2102                         FLOW_DIS_FIRST_FRAG);
2103
2104         _key = cpu_to_be32(key);
2105         _mask = cpu_to_be32(mask);
2106
2107         err = nla_put(skb, TCA_FLOWER_KEY_FLAGS, 4, &_key);
2108         if (err)
2109                 return err;
2110
2111         return nla_put(skb, TCA_FLOWER_KEY_FLAGS_MASK, 4, &_mask);
2112 }
2113
2114 static int fl_dump_key_geneve_opt(struct sk_buff *skb,
2115                                   struct flow_dissector_key_enc_opts *enc_opts)
2116 {
2117         struct geneve_opt *opt;
2118         struct nlattr *nest;
2119         int opt_off = 0;
2120
2121         nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_GENEVE);
2122         if (!nest)
2123                 goto nla_put_failure;
2124
2125         while (enc_opts->len > opt_off) {
2126                 opt = (struct geneve_opt *)&enc_opts->data[opt_off];
2127
2128                 if (nla_put_be16(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS,
2129                                  opt->opt_class))
2130                         goto nla_put_failure;
2131                 if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE,
2132                                opt->type))
2133                         goto nla_put_failure;
2134                 if (nla_put(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA,
2135                             opt->length * 4, opt->opt_data))
2136                         goto nla_put_failure;
2137
2138                 opt_off += sizeof(struct geneve_opt) + opt->length * 4;
2139         }
2140         nla_nest_end(skb, nest);
2141         return 0;
2142
2143 nla_put_failure:
2144         nla_nest_cancel(skb, nest);
2145         return -EMSGSIZE;
2146 }
2147
2148 static int fl_dump_key_ct(struct sk_buff *skb,
2149                           struct flow_dissector_key_ct *key,
2150                           struct flow_dissector_key_ct *mask)
2151 {
2152         if (IS_ENABLED(CONFIG_NF_CONNTRACK) &&
2153             fl_dump_key_val(skb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE,
2154                             &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK,
2155                             sizeof(key->ct_state)))
2156                 goto nla_put_failure;
2157
2158         if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
2159             fl_dump_key_val(skb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE,
2160                             &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK,
2161                             sizeof(key->ct_zone)))
2162                 goto nla_put_failure;
2163
2164         if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
2165             fl_dump_key_val(skb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK,
2166                             &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK,
2167                             sizeof(key->ct_mark)))
2168                 goto nla_put_failure;
2169
2170         if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
2171             fl_dump_key_val(skb, &key->ct_labels, TCA_FLOWER_KEY_CT_LABELS,
2172                             &mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK,
2173                             sizeof(key->ct_labels)))
2174                 goto nla_put_failure;
2175
2176         return 0;
2177
2178 nla_put_failure:
2179         return -EMSGSIZE;
2180 }
2181
2182 static int fl_dump_key_options(struct sk_buff *skb, int enc_opt_type,
2183                                struct flow_dissector_key_enc_opts *enc_opts)
2184 {
2185         struct nlattr *nest;
2186         int err;
2187
2188         if (!enc_opts->len)
2189                 return 0;
2190
2191         nest = nla_nest_start_noflag(skb, enc_opt_type);
2192         if (!nest)
2193                 goto nla_put_failure;
2194
2195         switch (enc_opts->dst_opt_type) {
2196         case TUNNEL_GENEVE_OPT:
2197                 err = fl_dump_key_geneve_opt(skb, enc_opts);
2198                 if (err)
2199                         goto nla_put_failure;
2200                 break;
2201         default:
2202                 goto nla_put_failure;
2203         }
2204         nla_nest_end(skb, nest);
2205         return 0;
2206
2207 nla_put_failure:
2208         nla_nest_cancel(skb, nest);
2209         return -EMSGSIZE;
2210 }
2211
2212 static int fl_dump_key_enc_opt(struct sk_buff *skb,
2213                                struct flow_dissector_key_enc_opts *key_opts,
2214                                struct flow_dissector_key_enc_opts *msk_opts)
2215 {
2216         int err;
2217
2218         err = fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS, key_opts);
2219         if (err)
2220                 return err;
2221
2222         return fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS_MASK, msk_opts);
2223 }
2224
2225 static int fl_dump_key(struct sk_buff *skb, struct net *net,
2226                        struct fl_flow_key *key, struct fl_flow_key *mask)
2227 {
2228         if (mask->meta.ingress_ifindex) {
2229                 struct net_device *dev;
2230
2231                 dev = __dev_get_by_index(net, key->meta.ingress_ifindex);
2232                 if (dev && nla_put_string(skb, TCA_FLOWER_INDEV, dev->name))
2233                         goto nla_put_failure;
2234         }
2235
2236         if (fl_dump_key_val(skb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
2237                             mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
2238                             sizeof(key->eth.dst)) ||
2239             fl_dump_key_val(skb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
2240                             mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
2241                             sizeof(key->eth.src)) ||
2242             fl_dump_key_val(skb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE,
2243                             &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
2244                             sizeof(key->basic.n_proto)))
2245                 goto nla_put_failure;
2246
2247         if (fl_dump_key_mpls(skb, &key->mpls, &mask->mpls))
2248                 goto nla_put_failure;
2249
2250         if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_VLAN_ID,
2251                              TCA_FLOWER_KEY_VLAN_PRIO, &key->vlan, &mask->vlan))
2252                 goto nla_put_failure;
2253
2254         if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_CVLAN_ID,
2255                              TCA_FLOWER_KEY_CVLAN_PRIO,
2256                              &key->cvlan, &mask->cvlan) ||
2257             (mask->cvlan.vlan_tpid &&
2258              nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
2259                           key->cvlan.vlan_tpid)))
2260                 goto nla_put_failure;
2261
2262         if (mask->basic.n_proto) {
2263                 if (mask->cvlan.vlan_tpid) {
2264                         if (nla_put_be16(skb, TCA_FLOWER_KEY_CVLAN_ETH_TYPE,
2265                                          key->basic.n_proto))
2266                                 goto nla_put_failure;
2267                 } else if (mask->vlan.vlan_tpid) {
2268                         if (nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE,
2269                                          key->basic.n_proto))
2270                                 goto nla_put_failure;
2271                 }
2272         }
2273
2274         if ((key->basic.n_proto == htons(ETH_P_IP) ||
2275              key->basic.n_proto == htons(ETH_P_IPV6)) &&
2276             (fl_dump_key_val(skb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
2277                             &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
2278                             sizeof(key->basic.ip_proto)) ||
2279             fl_dump_key_ip(skb, false, &key->ip, &mask->ip)))
2280                 goto nla_put_failure;
2281
2282         if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
2283             (fl_dump_key_val(skb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
2284                              &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
2285                              sizeof(key->ipv4.src)) ||
2286              fl_dump_key_val(skb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
2287                              &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
2288                              sizeof(key->ipv4.dst))))
2289                 goto nla_put_failure;
2290         else if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
2291                  (fl_dump_key_val(skb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
2292                                   &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
2293                                   sizeof(key->ipv6.src)) ||
2294                   fl_dump_key_val(skb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
2295                                   &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
2296                                   sizeof(key->ipv6.dst))))
2297                 goto nla_put_failure;
2298
2299         if (key->basic.ip_proto == IPPROTO_TCP &&
2300             (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
2301                              &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK,
2302                              sizeof(key->tp.src)) ||
2303              fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
2304                              &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK,
2305                              sizeof(key->tp.dst)) ||
2306              fl_dump_key_val(skb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS,
2307                              &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK,
2308                              sizeof(key->tcp.flags))))
2309                 goto nla_put_failure;
2310         else if (key->basic.ip_proto == IPPROTO_UDP &&
2311                  (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
2312                                   &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK,
2313                                   sizeof(key->tp.src)) ||
2314                   fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
2315                                   &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK,
2316                                   sizeof(key->tp.dst))))
2317                 goto nla_put_failure;
2318         else if (key->basic.ip_proto == IPPROTO_SCTP &&
2319                  (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC,
2320                                   &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK,
2321                                   sizeof(key->tp.src)) ||
2322                   fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST,
2323                                   &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK,
2324                                   sizeof(key->tp.dst))))
2325                 goto nla_put_failure;
2326         else if (key->basic.n_proto == htons(ETH_P_IP) &&
2327                  key->basic.ip_proto == IPPROTO_ICMP &&
2328                  (fl_dump_key_val(skb, &key->icmp.type,
2329                                   TCA_FLOWER_KEY_ICMPV4_TYPE, &mask->icmp.type,
2330                                   TCA_FLOWER_KEY_ICMPV4_TYPE_MASK,
2331                                   sizeof(key->icmp.type)) ||
2332                   fl_dump_key_val(skb, &key->icmp.code,
2333                                   TCA_FLOWER_KEY_ICMPV4_CODE, &mask->icmp.code,
2334                                   TCA_FLOWER_KEY_ICMPV4_CODE_MASK,
2335                                   sizeof(key->icmp.code))))
2336                 goto nla_put_failure;
2337         else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
2338                  key->basic.ip_proto == IPPROTO_ICMPV6 &&
2339                  (fl_dump_key_val(skb, &key->icmp.type,
2340                                   TCA_FLOWER_KEY_ICMPV6_TYPE, &mask->icmp.type,
2341                                   TCA_FLOWER_KEY_ICMPV6_TYPE_MASK,
2342                                   sizeof(key->icmp.type)) ||
2343                   fl_dump_key_val(skb, &key->icmp.code,
2344                                   TCA_FLOWER_KEY_ICMPV6_CODE, &mask->icmp.code,
2345                                   TCA_FLOWER_KEY_ICMPV6_CODE_MASK,
2346                                   sizeof(key->icmp.code))))
2347                 goto nla_put_failure;
2348         else if ((key->basic.n_proto == htons(ETH_P_ARP) ||
2349                   key->basic.n_proto == htons(ETH_P_RARP)) &&
2350                  (fl_dump_key_val(skb, &key->arp.sip,
2351                                   TCA_FLOWER_KEY_ARP_SIP, &mask->arp.sip,
2352                                   TCA_FLOWER_KEY_ARP_SIP_MASK,
2353                                   sizeof(key->arp.sip)) ||
2354                   fl_dump_key_val(skb, &key->arp.tip,
2355                                   TCA_FLOWER_KEY_ARP_TIP, &mask->arp.tip,
2356                                   TCA_FLOWER_KEY_ARP_TIP_MASK,
2357                                   sizeof(key->arp.tip)) ||
2358                   fl_dump_key_val(skb, &key->arp.op,
2359                                   TCA_FLOWER_KEY_ARP_OP, &mask->arp.op,
2360                                   TCA_FLOWER_KEY_ARP_OP_MASK,
2361                                   sizeof(key->arp.op)) ||
2362                   fl_dump_key_val(skb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA,
2363                                   mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK,
2364                                   sizeof(key->arp.sha)) ||
2365                   fl_dump_key_val(skb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA,
2366                                   mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK,
2367                                   sizeof(key->arp.tha))))
2368                 goto nla_put_failure;
2369
2370         if ((key->basic.ip_proto == IPPROTO_TCP ||
2371              key->basic.ip_proto == IPPROTO_UDP ||
2372              key->basic.ip_proto == IPPROTO_SCTP) &&
2373              fl_dump_key_port_range(skb, key, mask))
2374                 goto nla_put_failure;
2375
2376         if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS &&
2377             (fl_dump_key_val(skb, &key->enc_ipv4.src,
2378                             TCA_FLOWER_KEY_ENC_IPV4_SRC, &mask->enc_ipv4.src,
2379                             TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK,
2380                             sizeof(key->enc_ipv4.src)) ||
2381              fl_dump_key_val(skb, &key->enc_ipv4.dst,
2382                              TCA_FLOWER_KEY_ENC_IPV4_DST, &mask->enc_ipv4.dst,
2383                              TCA_FLOWER_KEY_ENC_IPV4_DST_MASK,
2384                              sizeof(key->enc_ipv4.dst))))
2385                 goto nla_put_failure;
2386         else if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS &&
2387                  (fl_dump_key_val(skb, &key->enc_ipv6.src,
2388                             TCA_FLOWER_KEY_ENC_IPV6_SRC, &mask->enc_ipv6.src,
2389                             TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK,
2390                             sizeof(key->enc_ipv6.src)) ||
2391                  fl_dump_key_val(skb, &key->enc_ipv6.dst,
2392                                  TCA_FLOWER_KEY_ENC_IPV6_DST,
2393                                  &mask->enc_ipv6.dst,
2394                                  TCA_FLOWER_KEY_ENC_IPV6_DST_MASK,
2395                             sizeof(key->enc_ipv6.dst))))
2396                 goto nla_put_failure;
2397
2398         if (fl_dump_key_val(skb, &key->enc_key_id, TCA_FLOWER_KEY_ENC_KEY_ID,
2399                             &mask->enc_key_id, TCA_FLOWER_UNSPEC,
2400                             sizeof(key->enc_key_id)) ||
2401             fl_dump_key_val(skb, &key->enc_tp.src,
2402                             TCA_FLOWER_KEY_ENC_UDP_SRC_PORT,
2403                             &mask->enc_tp.src,
2404                             TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK,
2405                             sizeof(key->enc_tp.src)) ||
2406             fl_dump_key_val(skb, &key->enc_tp.dst,
2407                             TCA_FLOWER_KEY_ENC_UDP_DST_PORT,
2408                             &mask->enc_tp.dst,
2409                             TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK,
2410                             sizeof(key->enc_tp.dst)) ||
2411             fl_dump_key_ip(skb, true, &key->enc_ip, &mask->enc_ip) ||
2412             fl_dump_key_enc_opt(skb, &key->enc_opts, &mask->enc_opts))
2413                 goto nla_put_failure;
2414
2415         if (fl_dump_key_ct(skb, &key->ct, &mask->ct))
2416                 goto nla_put_failure;
2417
2418         if (fl_dump_key_flags(skb, key->control.flags, mask->control.flags))
2419                 goto nla_put_failure;
2420
2421         return 0;
2422
2423 nla_put_failure:
2424         return -EMSGSIZE;
2425 }
2426
2427 static int fl_dump(struct net *net, struct tcf_proto *tp, void *fh,
2428                    struct sk_buff *skb, struct tcmsg *t, bool rtnl_held)
2429 {
2430         struct cls_fl_filter *f = fh;
2431         struct nlattr *nest;
2432         struct fl_flow_key *key, *mask;
2433         bool skip_hw;
2434
2435         if (!f)
2436                 return skb->len;
2437
2438         t->tcm_handle = f->handle;
2439
2440         nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
2441         if (!nest)
2442                 goto nla_put_failure;
2443
2444         spin_lock(&tp->lock);
2445
2446         if (f->res.classid &&
2447             nla_put_u32(skb, TCA_FLOWER_CLASSID, f->res.classid))
2448                 goto nla_put_failure_locked;
2449
2450         key = &f->key;
2451         mask = &f->mask->key;
2452         skip_hw = tc_skip_hw(f->flags);
2453
2454         if (fl_dump_key(skb, net, key, mask))
2455                 goto nla_put_failure_locked;
2456
2457         if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags))
2458                 goto nla_put_failure_locked;
2459
2460         spin_unlock(&tp->lock);
2461
2462         if (!skip_hw)
2463                 fl_hw_update_stats(tp, f, rtnl_held);
2464
2465         if (nla_put_u32(skb, TCA_FLOWER_IN_HW_COUNT, f->in_hw_count))
2466                 goto nla_put_failure;
2467
2468         if (tcf_exts_dump(skb, &f->exts))
2469                 goto nla_put_failure;
2470
2471         nla_nest_end(skb, nest);
2472
2473         if (tcf_exts_dump_stats(skb, &f->exts) < 0)
2474                 goto nla_put_failure;
2475
2476         return skb->len;
2477
2478 nla_put_failure_locked:
2479         spin_unlock(&tp->lock);
2480 nla_put_failure:
2481         nla_nest_cancel(skb, nest);
2482         return -1;
2483 }
2484
2485 static int fl_tmplt_dump(struct sk_buff *skb, struct net *net, void *tmplt_priv)
2486 {
2487         struct fl_flow_tmplt *tmplt = tmplt_priv;
2488         struct fl_flow_key *key, *mask;
2489         struct nlattr *nest;
2490
2491         nest = nla_nest_start_noflag(skb, TCA_OPTIONS);
2492         if (!nest)
2493                 goto nla_put_failure;
2494
2495         key = &tmplt->dummy_key;
2496         mask = &tmplt->mask;
2497
2498         if (fl_dump_key(skb, net, key, mask))
2499                 goto nla_put_failure;
2500
2501         nla_nest_end(skb, nest);
2502
2503         return skb->len;
2504
2505 nla_put_failure:
2506         nla_nest_cancel(skb, nest);
2507         return -EMSGSIZE;
2508 }
2509
2510 static void fl_bind_class(void *fh, u32 classid, unsigned long cl)
2511 {
2512         struct cls_fl_filter *f = fh;
2513
2514         if (f && f->res.classid == classid)
2515                 f->res.class = cl;
2516 }
2517
2518 static struct tcf_proto_ops cls_fl_ops __read_mostly = {
2519         .kind           = "flower",
2520         .classify       = fl_classify,
2521         .init           = fl_init,
2522         .destroy        = fl_destroy,
2523         .get            = fl_get,
2524         .put            = fl_put,
2525         .change         = fl_change,
2526         .delete         = fl_delete,
2527         .walk           = fl_walk,
2528         .reoffload      = fl_reoffload,
2529         .dump           = fl_dump,
2530         .bind_class     = fl_bind_class,
2531         .tmplt_create   = fl_tmplt_create,
2532         .tmplt_destroy  = fl_tmplt_destroy,
2533         .tmplt_dump     = fl_tmplt_dump,
2534         .owner          = THIS_MODULE,
2535         .flags          = TCF_PROTO_OPS_DOIT_UNLOCKED,
2536 };
2537
2538 static int __init cls_fl_init(void)
2539 {
2540         return register_tcf_proto_ops(&cls_fl_ops);
2541 }
2542
2543 static void __exit cls_fl_exit(void)
2544 {
2545         unregister_tcf_proto_ops(&cls_fl_ops);
2546 }
2547
2548 module_init(cls_fl_init);
2549 module_exit(cls_fl_exit);
2550
2551 MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
2552 MODULE_DESCRIPTION("Flower classifier");
2553 MODULE_LICENSE("GPL v2");