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 #include <linux/bitfield.h> 15 16 #include <linux/if_ether.h> 17 #include <linux/in6.h> 18 #include <linux/ip.h> 19 #include <linux/mpls.h> 20 #include <linux/ppp_defs.h> 21 22 #include <net/sch_generic.h> 23 #include <net/pkt_cls.h> 24 #include <net/pkt_sched.h> 25 #include <net/ip.h> 26 #include <net/flow_dissector.h> 27 #include <net/geneve.h> 28 #include <net/vxlan.h> 29 #include <net/erspan.h> 30 #include <net/gtp.h> 31 #include <net/tc_wrapper.h> 32 33 #include <net/dst.h> 34 #include <net/dst_metadata.h> 35 36 #include <uapi/linux/netfilter/nf_conntrack_common.h> 37 38 #define TCA_FLOWER_KEY_CT_FLAGS_MAX \ 39 ((__TCA_FLOWER_KEY_CT_FLAGS_MAX - 1) << 1) 40 #define TCA_FLOWER_KEY_CT_FLAGS_MASK \ 41 (TCA_FLOWER_KEY_CT_FLAGS_MAX - 1) 42 43 struct fl_flow_key { 44 struct flow_dissector_key_meta meta; 45 struct flow_dissector_key_control control; 46 struct flow_dissector_key_control enc_control; 47 struct flow_dissector_key_basic basic; 48 struct flow_dissector_key_eth_addrs eth; 49 struct flow_dissector_key_vlan vlan; 50 struct flow_dissector_key_vlan cvlan; 51 union { 52 struct flow_dissector_key_ipv4_addrs ipv4; 53 struct flow_dissector_key_ipv6_addrs ipv6; 54 }; 55 struct flow_dissector_key_ports tp; 56 struct flow_dissector_key_icmp icmp; 57 struct flow_dissector_key_arp arp; 58 struct flow_dissector_key_keyid enc_key_id; 59 union { 60 struct flow_dissector_key_ipv4_addrs enc_ipv4; 61 struct flow_dissector_key_ipv6_addrs enc_ipv6; 62 }; 63 struct flow_dissector_key_ports enc_tp; 64 struct flow_dissector_key_mpls mpls; 65 struct flow_dissector_key_tcp tcp; 66 struct flow_dissector_key_ip ip; 67 struct flow_dissector_key_ip enc_ip; 68 struct flow_dissector_key_enc_opts enc_opts; 69 struct flow_dissector_key_ports_range tp_range; 70 struct flow_dissector_key_ct ct; 71 struct flow_dissector_key_hash hash; 72 struct flow_dissector_key_num_of_vlans num_of_vlans; 73 struct flow_dissector_key_pppoe pppoe; 74 struct flow_dissector_key_l2tpv3 l2tpv3; 75 struct flow_dissector_key_cfm cfm; 76 } __aligned(BITS_PER_LONG / 8); /* Ensure that we can do comparisons as longs. */ 77 78 struct fl_flow_mask_range { 79 unsigned short int start; 80 unsigned short int end; 81 }; 82 83 struct fl_flow_mask { 84 struct fl_flow_key key; 85 struct fl_flow_mask_range range; 86 u32 flags; 87 struct rhash_head ht_node; 88 struct rhashtable ht; 89 struct rhashtable_params filter_ht_params; 90 struct flow_dissector dissector; 91 struct list_head filters; 92 struct rcu_work rwork; 93 struct list_head list; 94 refcount_t refcnt; 95 }; 96 97 struct fl_flow_tmplt { 98 struct fl_flow_key dummy_key; 99 struct fl_flow_key mask; 100 struct flow_dissector dissector; 101 struct tcf_chain *chain; 102 }; 103 104 struct cls_fl_head { 105 struct rhashtable ht; 106 spinlock_t masks_lock; /* Protect masks list */ 107 struct list_head masks; 108 struct list_head hw_filters; 109 struct rcu_work rwork; 110 struct idr handle_idr; 111 }; 112 113 struct cls_fl_filter { 114 struct fl_flow_mask *mask; 115 struct rhash_head ht_node; 116 struct fl_flow_key mkey; 117 struct tcf_exts exts; 118 struct tcf_result res; 119 struct fl_flow_key key; 120 struct list_head list; 121 struct list_head hw_list; 122 u32 handle; 123 u32 flags; 124 u32 in_hw_count; 125 u8 needs_tc_skb_ext:1; 126 struct rcu_work rwork; 127 struct net_device *hw_dev; 128 /* Flower classifier is unlocked, which means that its reference counter 129 * can be changed concurrently without any kind of external 130 * synchronization. Use atomic reference counter to be concurrency-safe. 131 */ 132 refcount_t refcnt; 133 bool deleted; 134 }; 135 136 static const struct rhashtable_params mask_ht_params = { 137 .key_offset = offsetof(struct fl_flow_mask, key), 138 .key_len = sizeof(struct fl_flow_key), 139 .head_offset = offsetof(struct fl_flow_mask, ht_node), 140 .automatic_shrinking = true, 141 }; 142 143 static unsigned short int fl_mask_range(const struct fl_flow_mask *mask) 144 { 145 return mask->range.end - mask->range.start; 146 } 147 148 static void fl_mask_update_range(struct fl_flow_mask *mask) 149 { 150 const u8 *bytes = (const u8 *) &mask->key; 151 size_t size = sizeof(mask->key); 152 size_t i, first = 0, last; 153 154 for (i = 0; i < size; i++) { 155 if (bytes[i]) { 156 first = i; 157 break; 158 } 159 } 160 last = first; 161 for (i = size - 1; i != first; i--) { 162 if (bytes[i]) { 163 last = i; 164 break; 165 } 166 } 167 mask->range.start = rounddown(first, sizeof(long)); 168 mask->range.end = roundup(last + 1, sizeof(long)); 169 } 170 171 static void *fl_key_get_start(struct fl_flow_key *key, 172 const struct fl_flow_mask *mask) 173 { 174 return (u8 *) key + mask->range.start; 175 } 176 177 static void fl_set_masked_key(struct fl_flow_key *mkey, struct fl_flow_key *key, 178 struct fl_flow_mask *mask) 179 { 180 const long *lkey = fl_key_get_start(key, mask); 181 const long *lmask = fl_key_get_start(&mask->key, mask); 182 long *lmkey = fl_key_get_start(mkey, mask); 183 int i; 184 185 for (i = 0; i < fl_mask_range(mask); i += sizeof(long)) 186 *lmkey++ = *lkey++ & *lmask++; 187 } 188 189 static bool fl_mask_fits_tmplt(struct fl_flow_tmplt *tmplt, 190 struct fl_flow_mask *mask) 191 { 192 const long *lmask = fl_key_get_start(&mask->key, mask); 193 const long *ltmplt; 194 int i; 195 196 if (!tmplt) 197 return true; 198 ltmplt = fl_key_get_start(&tmplt->mask, mask); 199 for (i = 0; i < fl_mask_range(mask); i += sizeof(long)) { 200 if (~*ltmplt++ & *lmask++) 201 return false; 202 } 203 return true; 204 } 205 206 static void fl_clear_masked_range(struct fl_flow_key *key, 207 struct fl_flow_mask *mask) 208 { 209 memset(fl_key_get_start(key, mask), 0, fl_mask_range(mask)); 210 } 211 212 static bool fl_range_port_dst_cmp(struct cls_fl_filter *filter, 213 struct fl_flow_key *key, 214 struct fl_flow_key *mkey) 215 { 216 u16 min_mask, max_mask, min_val, max_val; 217 218 min_mask = ntohs(filter->mask->key.tp_range.tp_min.dst); 219 max_mask = ntohs(filter->mask->key.tp_range.tp_max.dst); 220 min_val = ntohs(filter->key.tp_range.tp_min.dst); 221 max_val = ntohs(filter->key.tp_range.tp_max.dst); 222 223 if (min_mask && max_mask) { 224 if (ntohs(key->tp_range.tp.dst) < min_val || 225 ntohs(key->tp_range.tp.dst) > max_val) 226 return false; 227 228 /* skb does not have min and max values */ 229 mkey->tp_range.tp_min.dst = filter->mkey.tp_range.tp_min.dst; 230 mkey->tp_range.tp_max.dst = filter->mkey.tp_range.tp_max.dst; 231 } 232 return true; 233 } 234 235 static bool fl_range_port_src_cmp(struct cls_fl_filter *filter, 236 struct fl_flow_key *key, 237 struct fl_flow_key *mkey) 238 { 239 u16 min_mask, max_mask, min_val, max_val; 240 241 min_mask = ntohs(filter->mask->key.tp_range.tp_min.src); 242 max_mask = ntohs(filter->mask->key.tp_range.tp_max.src); 243 min_val = ntohs(filter->key.tp_range.tp_min.src); 244 max_val = ntohs(filter->key.tp_range.tp_max.src); 245 246 if (min_mask && max_mask) { 247 if (ntohs(key->tp_range.tp.src) < min_val || 248 ntohs(key->tp_range.tp.src) > max_val) 249 return false; 250 251 /* skb does not have min and max values */ 252 mkey->tp_range.tp_min.src = filter->mkey.tp_range.tp_min.src; 253 mkey->tp_range.tp_max.src = filter->mkey.tp_range.tp_max.src; 254 } 255 return true; 256 } 257 258 static struct cls_fl_filter *__fl_lookup(struct fl_flow_mask *mask, 259 struct fl_flow_key *mkey) 260 { 261 return rhashtable_lookup_fast(&mask->ht, fl_key_get_start(mkey, mask), 262 mask->filter_ht_params); 263 } 264 265 static struct cls_fl_filter *fl_lookup_range(struct fl_flow_mask *mask, 266 struct fl_flow_key *mkey, 267 struct fl_flow_key *key) 268 { 269 struct cls_fl_filter *filter, *f; 270 271 list_for_each_entry_rcu(filter, &mask->filters, list) { 272 if (!fl_range_port_dst_cmp(filter, key, mkey)) 273 continue; 274 275 if (!fl_range_port_src_cmp(filter, key, mkey)) 276 continue; 277 278 f = __fl_lookup(mask, mkey); 279 if (f) 280 return f; 281 } 282 return NULL; 283 } 284 285 static noinline_for_stack 286 struct cls_fl_filter *fl_mask_lookup(struct fl_flow_mask *mask, struct fl_flow_key *key) 287 { 288 struct fl_flow_key mkey; 289 290 fl_set_masked_key(&mkey, key, mask); 291 if ((mask->flags & TCA_FLOWER_MASK_FLAGS_RANGE)) 292 return fl_lookup_range(mask, &mkey, key); 293 294 return __fl_lookup(mask, &mkey); 295 } 296 297 static u16 fl_ct_info_to_flower_map[] = { 298 [IP_CT_ESTABLISHED] = TCA_FLOWER_KEY_CT_FLAGS_TRACKED | 299 TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED, 300 [IP_CT_RELATED] = TCA_FLOWER_KEY_CT_FLAGS_TRACKED | 301 TCA_FLOWER_KEY_CT_FLAGS_RELATED, 302 [IP_CT_ESTABLISHED_REPLY] = TCA_FLOWER_KEY_CT_FLAGS_TRACKED | 303 TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED | 304 TCA_FLOWER_KEY_CT_FLAGS_REPLY, 305 [IP_CT_RELATED_REPLY] = TCA_FLOWER_KEY_CT_FLAGS_TRACKED | 306 TCA_FLOWER_KEY_CT_FLAGS_RELATED | 307 TCA_FLOWER_KEY_CT_FLAGS_REPLY, 308 [IP_CT_NEW] = TCA_FLOWER_KEY_CT_FLAGS_TRACKED | 309 TCA_FLOWER_KEY_CT_FLAGS_NEW, 310 }; 311 312 TC_INDIRECT_SCOPE int fl_classify(struct sk_buff *skb, 313 const struct tcf_proto *tp, 314 struct tcf_result *res) 315 { 316 struct cls_fl_head *head = rcu_dereference_bh(tp->root); 317 bool post_ct = tc_skb_cb(skb)->post_ct; 318 u16 zone = tc_skb_cb(skb)->zone; 319 struct fl_flow_key skb_key; 320 struct fl_flow_mask *mask; 321 struct cls_fl_filter *f; 322 323 list_for_each_entry_rcu(mask, &head->masks, list) { 324 flow_dissector_init_keys(&skb_key.control, &skb_key.basic); 325 fl_clear_masked_range(&skb_key, mask); 326 327 skb_flow_dissect_meta(skb, &mask->dissector, &skb_key); 328 /* skb_flow_dissect() does not set n_proto in case an unknown 329 * protocol, so do it rather here. 330 */ 331 skb_key.basic.n_proto = skb_protocol(skb, false); 332 skb_flow_dissect_tunnel_info(skb, &mask->dissector, &skb_key); 333 skb_flow_dissect_ct(skb, &mask->dissector, &skb_key, 334 fl_ct_info_to_flower_map, 335 ARRAY_SIZE(fl_ct_info_to_flower_map), 336 post_ct, zone); 337 skb_flow_dissect_hash(skb, &mask->dissector, &skb_key); 338 skb_flow_dissect(skb, &mask->dissector, &skb_key, 339 FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP); 340 341 f = fl_mask_lookup(mask, &skb_key); 342 if (f && !tc_skip_sw(f->flags)) { 343 *res = f->res; 344 return tcf_exts_exec(skb, &f->exts, res); 345 } 346 } 347 return -1; 348 } 349 350 static int fl_init(struct tcf_proto *tp) 351 { 352 struct cls_fl_head *head; 353 354 head = kzalloc(sizeof(*head), GFP_KERNEL); 355 if (!head) 356 return -ENOBUFS; 357 358 spin_lock_init(&head->masks_lock); 359 INIT_LIST_HEAD_RCU(&head->masks); 360 INIT_LIST_HEAD(&head->hw_filters); 361 rcu_assign_pointer(tp->root, head); 362 idr_init(&head->handle_idr); 363 364 return rhashtable_init(&head->ht, &mask_ht_params); 365 } 366 367 static void fl_mask_free(struct fl_flow_mask *mask, bool mask_init_done) 368 { 369 /* temporary masks don't have their filters list and ht initialized */ 370 if (mask_init_done) { 371 WARN_ON(!list_empty(&mask->filters)); 372 rhashtable_destroy(&mask->ht); 373 } 374 kfree(mask); 375 } 376 377 static void fl_mask_free_work(struct work_struct *work) 378 { 379 struct fl_flow_mask *mask = container_of(to_rcu_work(work), 380 struct fl_flow_mask, rwork); 381 382 fl_mask_free(mask, true); 383 } 384 385 static void fl_uninit_mask_free_work(struct work_struct *work) 386 { 387 struct fl_flow_mask *mask = container_of(to_rcu_work(work), 388 struct fl_flow_mask, rwork); 389 390 fl_mask_free(mask, false); 391 } 392 393 static bool fl_mask_put(struct cls_fl_head *head, struct fl_flow_mask *mask) 394 { 395 if (!refcount_dec_and_test(&mask->refcnt)) 396 return false; 397 398 rhashtable_remove_fast(&head->ht, &mask->ht_node, mask_ht_params); 399 400 spin_lock(&head->masks_lock); 401 list_del_rcu(&mask->list); 402 spin_unlock(&head->masks_lock); 403 404 tcf_queue_work(&mask->rwork, fl_mask_free_work); 405 406 return true; 407 } 408 409 static struct cls_fl_head *fl_head_dereference(struct tcf_proto *tp) 410 { 411 /* Flower classifier only changes root pointer during init and destroy. 412 * Users must obtain reference to tcf_proto instance before calling its 413 * API, so tp->root pointer is protected from concurrent call to 414 * fl_destroy() by reference counting. 415 */ 416 return rcu_dereference_raw(tp->root); 417 } 418 419 static void __fl_destroy_filter(struct cls_fl_filter *f) 420 { 421 if (f->needs_tc_skb_ext) 422 tc_skb_ext_tc_disable(); 423 tcf_exts_destroy(&f->exts); 424 tcf_exts_put_net(&f->exts); 425 kfree(f); 426 } 427 428 static void fl_destroy_filter_work(struct work_struct *work) 429 { 430 struct cls_fl_filter *f = container_of(to_rcu_work(work), 431 struct cls_fl_filter, rwork); 432 433 __fl_destroy_filter(f); 434 } 435 436 static void fl_hw_destroy_filter(struct tcf_proto *tp, struct cls_fl_filter *f, 437 bool rtnl_held, struct netlink_ext_ack *extack) 438 { 439 struct tcf_block *block = tp->chain->block; 440 struct flow_cls_offload cls_flower = {}; 441 442 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack); 443 cls_flower.command = FLOW_CLS_DESTROY; 444 cls_flower.cookie = (unsigned long) f; 445 446 tc_setup_cb_destroy(block, tp, TC_SETUP_CLSFLOWER, &cls_flower, false, 447 &f->flags, &f->in_hw_count, rtnl_held); 448 449 } 450 451 static int fl_hw_replace_filter(struct tcf_proto *tp, 452 struct cls_fl_filter *f, bool rtnl_held, 453 struct netlink_ext_ack *extack) 454 { 455 struct tcf_block *block = tp->chain->block; 456 struct flow_cls_offload cls_flower = {}; 457 bool skip_sw = tc_skip_sw(f->flags); 458 int err = 0; 459 460 cls_flower.rule = flow_rule_alloc(tcf_exts_num_actions(&f->exts)); 461 if (!cls_flower.rule) 462 return -ENOMEM; 463 464 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, extack); 465 cls_flower.command = FLOW_CLS_REPLACE; 466 cls_flower.cookie = (unsigned long) f; 467 cls_flower.rule->match.dissector = &f->mask->dissector; 468 cls_flower.rule->match.mask = &f->mask->key; 469 cls_flower.rule->match.key = &f->mkey; 470 cls_flower.classid = f->res.classid; 471 472 err = tc_setup_offload_action(&cls_flower.rule->action, &f->exts, 473 cls_flower.common.extack); 474 if (err) { 475 kfree(cls_flower.rule); 476 477 return skip_sw ? err : 0; 478 } 479 480 err = tc_setup_cb_add(block, tp, TC_SETUP_CLSFLOWER, &cls_flower, 481 skip_sw, &f->flags, &f->in_hw_count, rtnl_held); 482 tc_cleanup_offload_action(&cls_flower.rule->action); 483 kfree(cls_flower.rule); 484 485 if (err) { 486 fl_hw_destroy_filter(tp, f, rtnl_held, NULL); 487 return err; 488 } 489 490 if (skip_sw && !(f->flags & TCA_CLS_FLAGS_IN_HW)) 491 return -EINVAL; 492 493 return 0; 494 } 495 496 static void fl_hw_update_stats(struct tcf_proto *tp, struct cls_fl_filter *f, 497 bool rtnl_held) 498 { 499 struct tcf_block *block = tp->chain->block; 500 struct flow_cls_offload cls_flower = {}; 501 502 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, NULL); 503 cls_flower.command = FLOW_CLS_STATS; 504 cls_flower.cookie = (unsigned long) f; 505 cls_flower.classid = f->res.classid; 506 507 tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, 508 rtnl_held); 509 510 tcf_exts_hw_stats_update(&f->exts, &cls_flower.stats, cls_flower.use_act_stats); 511 } 512 513 static void __fl_put(struct cls_fl_filter *f) 514 { 515 if (!refcount_dec_and_test(&f->refcnt)) 516 return; 517 518 if (tcf_exts_get_net(&f->exts)) 519 tcf_queue_work(&f->rwork, fl_destroy_filter_work); 520 else 521 __fl_destroy_filter(f); 522 } 523 524 static struct cls_fl_filter *__fl_get(struct cls_fl_head *head, u32 handle) 525 { 526 struct cls_fl_filter *f; 527 528 rcu_read_lock(); 529 f = idr_find(&head->handle_idr, handle); 530 if (f && !refcount_inc_not_zero(&f->refcnt)) 531 f = NULL; 532 rcu_read_unlock(); 533 534 return f; 535 } 536 537 static struct tcf_exts *fl_get_exts(const struct tcf_proto *tp, u32 handle) 538 { 539 struct cls_fl_head *head = rcu_dereference_bh(tp->root); 540 struct cls_fl_filter *f; 541 542 f = idr_find(&head->handle_idr, handle); 543 return f ? &f->exts : NULL; 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] = { .strict_start_type = 624 TCA_FLOWER_L2_MISS }, 625 [TCA_FLOWER_CLASSID] = { .type = NLA_U32 }, 626 [TCA_FLOWER_INDEV] = { .type = NLA_STRING, 627 .len = IFNAMSIZ }, 628 [TCA_FLOWER_KEY_ETH_DST] = { .len = ETH_ALEN }, 629 [TCA_FLOWER_KEY_ETH_DST_MASK] = { .len = ETH_ALEN }, 630 [TCA_FLOWER_KEY_ETH_SRC] = { .len = ETH_ALEN }, 631 [TCA_FLOWER_KEY_ETH_SRC_MASK] = { .len = ETH_ALEN }, 632 [TCA_FLOWER_KEY_ETH_TYPE] = { .type = NLA_U16 }, 633 [TCA_FLOWER_KEY_IP_PROTO] = { .type = NLA_U8 }, 634 [TCA_FLOWER_KEY_IPV4_SRC] = { .type = NLA_U32 }, 635 [TCA_FLOWER_KEY_IPV4_SRC_MASK] = { .type = NLA_U32 }, 636 [TCA_FLOWER_KEY_IPV4_DST] = { .type = NLA_U32 }, 637 [TCA_FLOWER_KEY_IPV4_DST_MASK] = { .type = NLA_U32 }, 638 [TCA_FLOWER_KEY_IPV6_SRC] = { .len = sizeof(struct in6_addr) }, 639 [TCA_FLOWER_KEY_IPV6_SRC_MASK] = { .len = sizeof(struct in6_addr) }, 640 [TCA_FLOWER_KEY_IPV6_DST] = { .len = sizeof(struct in6_addr) }, 641 [TCA_FLOWER_KEY_IPV6_DST_MASK] = { .len = sizeof(struct in6_addr) }, 642 [TCA_FLOWER_KEY_TCP_SRC] = { .type = NLA_U16 }, 643 [TCA_FLOWER_KEY_TCP_DST] = { .type = NLA_U16 }, 644 [TCA_FLOWER_KEY_UDP_SRC] = { .type = NLA_U16 }, 645 [TCA_FLOWER_KEY_UDP_DST] = { .type = NLA_U16 }, 646 [TCA_FLOWER_KEY_VLAN_ID] = { .type = NLA_U16 }, 647 [TCA_FLOWER_KEY_VLAN_PRIO] = { .type = NLA_U8 }, 648 [TCA_FLOWER_KEY_VLAN_ETH_TYPE] = { .type = NLA_U16 }, 649 [TCA_FLOWER_KEY_ENC_KEY_ID] = { .type = NLA_U32 }, 650 [TCA_FLOWER_KEY_ENC_IPV4_SRC] = { .type = NLA_U32 }, 651 [TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK] = { .type = NLA_U32 }, 652 [TCA_FLOWER_KEY_ENC_IPV4_DST] = { .type = NLA_U32 }, 653 [TCA_FLOWER_KEY_ENC_IPV4_DST_MASK] = { .type = NLA_U32 }, 654 [TCA_FLOWER_KEY_ENC_IPV6_SRC] = { .len = sizeof(struct in6_addr) }, 655 [TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK] = { .len = sizeof(struct in6_addr) }, 656 [TCA_FLOWER_KEY_ENC_IPV6_DST] = { .len = sizeof(struct in6_addr) }, 657 [TCA_FLOWER_KEY_ENC_IPV6_DST_MASK] = { .len = sizeof(struct in6_addr) }, 658 [TCA_FLOWER_KEY_TCP_SRC_MASK] = { .type = NLA_U16 }, 659 [TCA_FLOWER_KEY_TCP_DST_MASK] = { .type = NLA_U16 }, 660 [TCA_FLOWER_KEY_UDP_SRC_MASK] = { .type = NLA_U16 }, 661 [TCA_FLOWER_KEY_UDP_DST_MASK] = { .type = NLA_U16 }, 662 [TCA_FLOWER_KEY_SCTP_SRC_MASK] = { .type = NLA_U16 }, 663 [TCA_FLOWER_KEY_SCTP_DST_MASK] = { .type = NLA_U16 }, 664 [TCA_FLOWER_KEY_SCTP_SRC] = { .type = NLA_U16 }, 665 [TCA_FLOWER_KEY_SCTP_DST] = { .type = NLA_U16 }, 666 [TCA_FLOWER_KEY_ENC_UDP_SRC_PORT] = { .type = NLA_U16 }, 667 [TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK] = { .type = NLA_U16 }, 668 [TCA_FLOWER_KEY_ENC_UDP_DST_PORT] = { .type = NLA_U16 }, 669 [TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK] = { .type = NLA_U16 }, 670 [TCA_FLOWER_KEY_FLAGS] = { .type = NLA_U32 }, 671 [TCA_FLOWER_KEY_FLAGS_MASK] = { .type = NLA_U32 }, 672 [TCA_FLOWER_KEY_ICMPV4_TYPE] = { .type = NLA_U8 }, 673 [TCA_FLOWER_KEY_ICMPV4_TYPE_MASK] = { .type = NLA_U8 }, 674 [TCA_FLOWER_KEY_ICMPV4_CODE] = { .type = NLA_U8 }, 675 [TCA_FLOWER_KEY_ICMPV4_CODE_MASK] = { .type = NLA_U8 }, 676 [TCA_FLOWER_KEY_ICMPV6_TYPE] = { .type = NLA_U8 }, 677 [TCA_FLOWER_KEY_ICMPV6_TYPE_MASK] = { .type = NLA_U8 }, 678 [TCA_FLOWER_KEY_ICMPV6_CODE] = { .type = NLA_U8 }, 679 [TCA_FLOWER_KEY_ICMPV6_CODE_MASK] = { .type = NLA_U8 }, 680 [TCA_FLOWER_KEY_ARP_SIP] = { .type = NLA_U32 }, 681 [TCA_FLOWER_KEY_ARP_SIP_MASK] = { .type = NLA_U32 }, 682 [TCA_FLOWER_KEY_ARP_TIP] = { .type = NLA_U32 }, 683 [TCA_FLOWER_KEY_ARP_TIP_MASK] = { .type = NLA_U32 }, 684 [TCA_FLOWER_KEY_ARP_OP] = { .type = NLA_U8 }, 685 [TCA_FLOWER_KEY_ARP_OP_MASK] = { .type = NLA_U8 }, 686 [TCA_FLOWER_KEY_ARP_SHA] = { .len = ETH_ALEN }, 687 [TCA_FLOWER_KEY_ARP_SHA_MASK] = { .len = ETH_ALEN }, 688 [TCA_FLOWER_KEY_ARP_THA] = { .len = ETH_ALEN }, 689 [TCA_FLOWER_KEY_ARP_THA_MASK] = { .len = ETH_ALEN }, 690 [TCA_FLOWER_KEY_MPLS_TTL] = { .type = NLA_U8 }, 691 [TCA_FLOWER_KEY_MPLS_BOS] = { .type = NLA_U8 }, 692 [TCA_FLOWER_KEY_MPLS_TC] = { .type = NLA_U8 }, 693 [TCA_FLOWER_KEY_MPLS_LABEL] = { .type = NLA_U32 }, 694 [TCA_FLOWER_KEY_MPLS_OPTS] = { .type = NLA_NESTED }, 695 [TCA_FLOWER_KEY_TCP_FLAGS] = { .type = NLA_U16 }, 696 [TCA_FLOWER_KEY_TCP_FLAGS_MASK] = { .type = NLA_U16 }, 697 [TCA_FLOWER_KEY_IP_TOS] = { .type = NLA_U8 }, 698 [TCA_FLOWER_KEY_IP_TOS_MASK] = { .type = NLA_U8 }, 699 [TCA_FLOWER_KEY_IP_TTL] = { .type = NLA_U8 }, 700 [TCA_FLOWER_KEY_IP_TTL_MASK] = { .type = NLA_U8 }, 701 [TCA_FLOWER_KEY_CVLAN_ID] = { .type = NLA_U16 }, 702 [TCA_FLOWER_KEY_CVLAN_PRIO] = { .type = NLA_U8 }, 703 [TCA_FLOWER_KEY_CVLAN_ETH_TYPE] = { .type = NLA_U16 }, 704 [TCA_FLOWER_KEY_ENC_IP_TOS] = { .type = NLA_U8 }, 705 [TCA_FLOWER_KEY_ENC_IP_TOS_MASK] = { .type = NLA_U8 }, 706 [TCA_FLOWER_KEY_ENC_IP_TTL] = { .type = NLA_U8 }, 707 [TCA_FLOWER_KEY_ENC_IP_TTL_MASK] = { .type = NLA_U8 }, 708 [TCA_FLOWER_KEY_ENC_OPTS] = { .type = NLA_NESTED }, 709 [TCA_FLOWER_KEY_ENC_OPTS_MASK] = { .type = NLA_NESTED }, 710 [TCA_FLOWER_KEY_CT_STATE] = 711 NLA_POLICY_MASK(NLA_U16, TCA_FLOWER_KEY_CT_FLAGS_MASK), 712 [TCA_FLOWER_KEY_CT_STATE_MASK] = 713 NLA_POLICY_MASK(NLA_U16, TCA_FLOWER_KEY_CT_FLAGS_MASK), 714 [TCA_FLOWER_KEY_CT_ZONE] = { .type = NLA_U16 }, 715 [TCA_FLOWER_KEY_CT_ZONE_MASK] = { .type = NLA_U16 }, 716 [TCA_FLOWER_KEY_CT_MARK] = { .type = NLA_U32 }, 717 [TCA_FLOWER_KEY_CT_MARK_MASK] = { .type = NLA_U32 }, 718 [TCA_FLOWER_KEY_CT_LABELS] = { .type = NLA_BINARY, 719 .len = 128 / BITS_PER_BYTE }, 720 [TCA_FLOWER_KEY_CT_LABELS_MASK] = { .type = NLA_BINARY, 721 .len = 128 / BITS_PER_BYTE }, 722 [TCA_FLOWER_FLAGS] = { .type = NLA_U32 }, 723 [TCA_FLOWER_KEY_HASH] = { .type = NLA_U32 }, 724 [TCA_FLOWER_KEY_HASH_MASK] = { .type = NLA_U32 }, 725 [TCA_FLOWER_KEY_NUM_OF_VLANS] = { .type = NLA_U8 }, 726 [TCA_FLOWER_KEY_PPPOE_SID] = { .type = NLA_U16 }, 727 [TCA_FLOWER_KEY_PPP_PROTO] = { .type = NLA_U16 }, 728 [TCA_FLOWER_KEY_L2TPV3_SID] = { .type = NLA_U32 }, 729 [TCA_FLOWER_L2_MISS] = NLA_POLICY_MAX(NLA_U8, 1), 730 [TCA_FLOWER_KEY_CFM] = { .type = NLA_NESTED }, 731 }; 732 733 static const struct nla_policy 734 enc_opts_policy[TCA_FLOWER_KEY_ENC_OPTS_MAX + 1] = { 735 [TCA_FLOWER_KEY_ENC_OPTS_UNSPEC] = { 736 .strict_start_type = TCA_FLOWER_KEY_ENC_OPTS_VXLAN }, 737 [TCA_FLOWER_KEY_ENC_OPTS_GENEVE] = { .type = NLA_NESTED }, 738 [TCA_FLOWER_KEY_ENC_OPTS_VXLAN] = { .type = NLA_NESTED }, 739 [TCA_FLOWER_KEY_ENC_OPTS_ERSPAN] = { .type = NLA_NESTED }, 740 [TCA_FLOWER_KEY_ENC_OPTS_GTP] = { .type = NLA_NESTED }, 741 }; 742 743 static const struct nla_policy 744 geneve_opt_policy[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1] = { 745 [TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS] = { .type = NLA_U16 }, 746 [TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE] = { .type = NLA_U8 }, 747 [TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA] = { .type = NLA_BINARY, 748 .len = 128 }, 749 }; 750 751 static const struct nla_policy 752 vxlan_opt_policy[TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX + 1] = { 753 [TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP] = { .type = NLA_U32 }, 754 }; 755 756 static const struct nla_policy 757 erspan_opt_policy[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX + 1] = { 758 [TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER] = { .type = NLA_U8 }, 759 [TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX] = { .type = NLA_U32 }, 760 [TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR] = { .type = NLA_U8 }, 761 [TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID] = { .type = NLA_U8 }, 762 }; 763 764 static const struct nla_policy 765 gtp_opt_policy[TCA_FLOWER_KEY_ENC_OPT_GTP_MAX + 1] = { 766 [TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE] = { .type = NLA_U8 }, 767 [TCA_FLOWER_KEY_ENC_OPT_GTP_QFI] = { .type = NLA_U8 }, 768 }; 769 770 static const struct nla_policy 771 mpls_stack_entry_policy[TCA_FLOWER_KEY_MPLS_OPT_LSE_MAX + 1] = { 772 [TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH] = { .type = NLA_U8 }, 773 [TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL] = { .type = NLA_U8 }, 774 [TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS] = { .type = NLA_U8 }, 775 [TCA_FLOWER_KEY_MPLS_OPT_LSE_TC] = { .type = NLA_U8 }, 776 [TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL] = { .type = NLA_U32 }, 777 }; 778 779 static const struct nla_policy cfm_opt_policy[TCA_FLOWER_KEY_CFM_OPT_MAX] = { 780 [TCA_FLOWER_KEY_CFM_MD_LEVEL] = NLA_POLICY_MAX(NLA_U8, 781 FLOW_DIS_CFM_MDL_MAX), 782 [TCA_FLOWER_KEY_CFM_OPCODE] = { .type = NLA_U8 }, 783 }; 784 785 static void fl_set_key_val(struct nlattr **tb, 786 void *val, int val_type, 787 void *mask, int mask_type, int len) 788 { 789 if (!tb[val_type]) 790 return; 791 nla_memcpy(val, tb[val_type], len); 792 if (mask_type == TCA_FLOWER_UNSPEC || !tb[mask_type]) 793 memset(mask, 0xff, len); 794 else 795 nla_memcpy(mask, tb[mask_type], len); 796 } 797 798 static int fl_set_key_port_range(struct nlattr **tb, struct fl_flow_key *key, 799 struct fl_flow_key *mask, 800 struct netlink_ext_ack *extack) 801 { 802 fl_set_key_val(tb, &key->tp_range.tp_min.dst, 803 TCA_FLOWER_KEY_PORT_DST_MIN, &mask->tp_range.tp_min.dst, 804 TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_min.dst)); 805 fl_set_key_val(tb, &key->tp_range.tp_max.dst, 806 TCA_FLOWER_KEY_PORT_DST_MAX, &mask->tp_range.tp_max.dst, 807 TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_max.dst)); 808 fl_set_key_val(tb, &key->tp_range.tp_min.src, 809 TCA_FLOWER_KEY_PORT_SRC_MIN, &mask->tp_range.tp_min.src, 810 TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_min.src)); 811 fl_set_key_val(tb, &key->tp_range.tp_max.src, 812 TCA_FLOWER_KEY_PORT_SRC_MAX, &mask->tp_range.tp_max.src, 813 TCA_FLOWER_UNSPEC, sizeof(key->tp_range.tp_max.src)); 814 815 if (mask->tp_range.tp_min.dst != mask->tp_range.tp_max.dst) { 816 NL_SET_ERR_MSG(extack, 817 "Both min and max destination ports must be specified"); 818 return -EINVAL; 819 } 820 if (mask->tp_range.tp_min.src != mask->tp_range.tp_max.src) { 821 NL_SET_ERR_MSG(extack, 822 "Both min and max source ports must be specified"); 823 return -EINVAL; 824 } 825 if (mask->tp_range.tp_min.dst && mask->tp_range.tp_max.dst && 826 ntohs(key->tp_range.tp_max.dst) <= 827 ntohs(key->tp_range.tp_min.dst)) { 828 NL_SET_ERR_MSG_ATTR(extack, 829 tb[TCA_FLOWER_KEY_PORT_DST_MIN], 830 "Invalid destination port range (min must be strictly smaller than max)"); 831 return -EINVAL; 832 } 833 if (mask->tp_range.tp_min.src && mask->tp_range.tp_max.src && 834 ntohs(key->tp_range.tp_max.src) <= 835 ntohs(key->tp_range.tp_min.src)) { 836 NL_SET_ERR_MSG_ATTR(extack, 837 tb[TCA_FLOWER_KEY_PORT_SRC_MIN], 838 "Invalid source port range (min must be strictly smaller than max)"); 839 return -EINVAL; 840 } 841 842 return 0; 843 } 844 845 static int fl_set_key_mpls_lse(const struct nlattr *nla_lse, 846 struct flow_dissector_key_mpls *key_val, 847 struct flow_dissector_key_mpls *key_mask, 848 struct netlink_ext_ack *extack) 849 { 850 struct nlattr *tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_MAX + 1]; 851 struct flow_dissector_mpls_lse *lse_mask; 852 struct flow_dissector_mpls_lse *lse_val; 853 u8 lse_index; 854 u8 depth; 855 int err; 856 857 err = nla_parse_nested(tb, TCA_FLOWER_KEY_MPLS_OPT_LSE_MAX, nla_lse, 858 mpls_stack_entry_policy, extack); 859 if (err < 0) 860 return err; 861 862 if (!tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH]) { 863 NL_SET_ERR_MSG(extack, "Missing MPLS option \"depth\""); 864 return -EINVAL; 865 } 866 867 depth = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH]); 868 869 /* LSE depth starts at 1, for consistency with terminology used by 870 * RFC 3031 (section 3.9), where depth 0 refers to unlabeled packets. 871 */ 872 if (depth < 1 || depth > FLOW_DIS_MPLS_MAX) { 873 NL_SET_ERR_MSG_ATTR(extack, 874 tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH], 875 "Invalid MPLS depth"); 876 return -EINVAL; 877 } 878 lse_index = depth - 1; 879 880 dissector_set_mpls_lse(key_val, lse_index); 881 dissector_set_mpls_lse(key_mask, lse_index); 882 883 lse_val = &key_val->ls[lse_index]; 884 lse_mask = &key_mask->ls[lse_index]; 885 886 if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL]) { 887 lse_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL]); 888 lse_mask->mpls_ttl = MPLS_TTL_MASK; 889 } 890 if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS]) { 891 u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS]); 892 893 if (bos & ~MPLS_BOS_MASK) { 894 NL_SET_ERR_MSG_ATTR(extack, 895 tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS], 896 "Bottom Of Stack (BOS) must be 0 or 1"); 897 return -EINVAL; 898 } 899 lse_val->mpls_bos = bos; 900 lse_mask->mpls_bos = MPLS_BOS_MASK; 901 } 902 if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC]) { 903 u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC]); 904 905 if (tc & ~MPLS_TC_MASK) { 906 NL_SET_ERR_MSG_ATTR(extack, 907 tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_TC], 908 "Traffic Class (TC) must be between 0 and 7"); 909 return -EINVAL; 910 } 911 lse_val->mpls_tc = tc; 912 lse_mask->mpls_tc = MPLS_TC_MASK; 913 } 914 if (tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL]) { 915 u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL]); 916 917 if (label & ~MPLS_LABEL_MASK) { 918 NL_SET_ERR_MSG_ATTR(extack, 919 tb[TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL], 920 "Label must be between 0 and 1048575"); 921 return -EINVAL; 922 } 923 lse_val->mpls_label = label; 924 lse_mask->mpls_label = MPLS_LABEL_MASK; 925 } 926 927 return 0; 928 } 929 930 static int fl_set_key_mpls_opts(const struct nlattr *nla_mpls_opts, 931 struct flow_dissector_key_mpls *key_val, 932 struct flow_dissector_key_mpls *key_mask, 933 struct netlink_ext_ack *extack) 934 { 935 struct nlattr *nla_lse; 936 int rem; 937 int err; 938 939 if (!(nla_mpls_opts->nla_type & NLA_F_NESTED)) { 940 NL_SET_ERR_MSG_ATTR(extack, nla_mpls_opts, 941 "NLA_F_NESTED is missing"); 942 return -EINVAL; 943 } 944 945 nla_for_each_nested(nla_lse, nla_mpls_opts, rem) { 946 if (nla_type(nla_lse) != TCA_FLOWER_KEY_MPLS_OPTS_LSE) { 947 NL_SET_ERR_MSG_ATTR(extack, nla_lse, 948 "Invalid MPLS option type"); 949 return -EINVAL; 950 } 951 952 err = fl_set_key_mpls_lse(nla_lse, key_val, key_mask, extack); 953 if (err < 0) 954 return err; 955 } 956 if (rem) { 957 NL_SET_ERR_MSG(extack, 958 "Bytes leftover after parsing MPLS options"); 959 return -EINVAL; 960 } 961 962 return 0; 963 } 964 965 static int fl_set_key_mpls(struct nlattr **tb, 966 struct flow_dissector_key_mpls *key_val, 967 struct flow_dissector_key_mpls *key_mask, 968 struct netlink_ext_ack *extack) 969 { 970 struct flow_dissector_mpls_lse *lse_mask; 971 struct flow_dissector_mpls_lse *lse_val; 972 973 if (tb[TCA_FLOWER_KEY_MPLS_OPTS]) { 974 if (tb[TCA_FLOWER_KEY_MPLS_TTL] || 975 tb[TCA_FLOWER_KEY_MPLS_BOS] || 976 tb[TCA_FLOWER_KEY_MPLS_TC] || 977 tb[TCA_FLOWER_KEY_MPLS_LABEL]) { 978 NL_SET_ERR_MSG_ATTR(extack, 979 tb[TCA_FLOWER_KEY_MPLS_OPTS], 980 "MPLS label, Traffic Class, Bottom Of Stack and Time To Live must be encapsulated in the MPLS options attribute"); 981 return -EBADMSG; 982 } 983 984 return fl_set_key_mpls_opts(tb[TCA_FLOWER_KEY_MPLS_OPTS], 985 key_val, key_mask, extack); 986 } 987 988 lse_val = &key_val->ls[0]; 989 lse_mask = &key_mask->ls[0]; 990 991 if (tb[TCA_FLOWER_KEY_MPLS_TTL]) { 992 lse_val->mpls_ttl = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TTL]); 993 lse_mask->mpls_ttl = MPLS_TTL_MASK; 994 dissector_set_mpls_lse(key_val, 0); 995 dissector_set_mpls_lse(key_mask, 0); 996 } 997 if (tb[TCA_FLOWER_KEY_MPLS_BOS]) { 998 u8 bos = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_BOS]); 999 1000 if (bos & ~MPLS_BOS_MASK) { 1001 NL_SET_ERR_MSG_ATTR(extack, 1002 tb[TCA_FLOWER_KEY_MPLS_BOS], 1003 "Bottom Of Stack (BOS) must be 0 or 1"); 1004 return -EINVAL; 1005 } 1006 lse_val->mpls_bos = bos; 1007 lse_mask->mpls_bos = MPLS_BOS_MASK; 1008 dissector_set_mpls_lse(key_val, 0); 1009 dissector_set_mpls_lse(key_mask, 0); 1010 } 1011 if (tb[TCA_FLOWER_KEY_MPLS_TC]) { 1012 u8 tc = nla_get_u8(tb[TCA_FLOWER_KEY_MPLS_TC]); 1013 1014 if (tc & ~MPLS_TC_MASK) { 1015 NL_SET_ERR_MSG_ATTR(extack, 1016 tb[TCA_FLOWER_KEY_MPLS_TC], 1017 "Traffic Class (TC) must be between 0 and 7"); 1018 return -EINVAL; 1019 } 1020 lse_val->mpls_tc = tc; 1021 lse_mask->mpls_tc = MPLS_TC_MASK; 1022 dissector_set_mpls_lse(key_val, 0); 1023 dissector_set_mpls_lse(key_mask, 0); 1024 } 1025 if (tb[TCA_FLOWER_KEY_MPLS_LABEL]) { 1026 u32 label = nla_get_u32(tb[TCA_FLOWER_KEY_MPLS_LABEL]); 1027 1028 if (label & ~MPLS_LABEL_MASK) { 1029 NL_SET_ERR_MSG_ATTR(extack, 1030 tb[TCA_FLOWER_KEY_MPLS_LABEL], 1031 "Label must be between 0 and 1048575"); 1032 return -EINVAL; 1033 } 1034 lse_val->mpls_label = label; 1035 lse_mask->mpls_label = MPLS_LABEL_MASK; 1036 dissector_set_mpls_lse(key_val, 0); 1037 dissector_set_mpls_lse(key_mask, 0); 1038 } 1039 return 0; 1040 } 1041 1042 static void fl_set_key_vlan(struct nlattr **tb, 1043 __be16 ethertype, 1044 int vlan_id_key, int vlan_prio_key, 1045 int vlan_next_eth_type_key, 1046 struct flow_dissector_key_vlan *key_val, 1047 struct flow_dissector_key_vlan *key_mask) 1048 { 1049 #define VLAN_PRIORITY_MASK 0x7 1050 1051 if (tb[vlan_id_key]) { 1052 key_val->vlan_id = 1053 nla_get_u16(tb[vlan_id_key]) & VLAN_VID_MASK; 1054 key_mask->vlan_id = VLAN_VID_MASK; 1055 } 1056 if (tb[vlan_prio_key]) { 1057 key_val->vlan_priority = 1058 nla_get_u8(tb[vlan_prio_key]) & 1059 VLAN_PRIORITY_MASK; 1060 key_mask->vlan_priority = VLAN_PRIORITY_MASK; 1061 } 1062 if (ethertype) { 1063 key_val->vlan_tpid = ethertype; 1064 key_mask->vlan_tpid = cpu_to_be16(~0); 1065 } 1066 if (tb[vlan_next_eth_type_key]) { 1067 key_val->vlan_eth_type = 1068 nla_get_be16(tb[vlan_next_eth_type_key]); 1069 key_mask->vlan_eth_type = cpu_to_be16(~0); 1070 } 1071 } 1072 1073 static void fl_set_key_pppoe(struct nlattr **tb, 1074 struct flow_dissector_key_pppoe *key_val, 1075 struct flow_dissector_key_pppoe *key_mask, 1076 struct fl_flow_key *key, 1077 struct fl_flow_key *mask) 1078 { 1079 /* key_val::type must be set to ETH_P_PPP_SES 1080 * because ETH_P_PPP_SES was stored in basic.n_proto 1081 * which might get overwritten by ppp_proto 1082 * or might be set to 0, the role of key_val::type 1083 * is similar to vlan_key::tpid 1084 */ 1085 key_val->type = htons(ETH_P_PPP_SES); 1086 key_mask->type = cpu_to_be16(~0); 1087 1088 if (tb[TCA_FLOWER_KEY_PPPOE_SID]) { 1089 key_val->session_id = 1090 nla_get_be16(tb[TCA_FLOWER_KEY_PPPOE_SID]); 1091 key_mask->session_id = cpu_to_be16(~0); 1092 } 1093 if (tb[TCA_FLOWER_KEY_PPP_PROTO]) { 1094 key_val->ppp_proto = 1095 nla_get_be16(tb[TCA_FLOWER_KEY_PPP_PROTO]); 1096 key_mask->ppp_proto = cpu_to_be16(~0); 1097 1098 if (key_val->ppp_proto == htons(PPP_IP)) { 1099 key->basic.n_proto = htons(ETH_P_IP); 1100 mask->basic.n_proto = cpu_to_be16(~0); 1101 } else if (key_val->ppp_proto == htons(PPP_IPV6)) { 1102 key->basic.n_proto = htons(ETH_P_IPV6); 1103 mask->basic.n_proto = cpu_to_be16(~0); 1104 } else if (key_val->ppp_proto == htons(PPP_MPLS_UC)) { 1105 key->basic.n_proto = htons(ETH_P_MPLS_UC); 1106 mask->basic.n_proto = cpu_to_be16(~0); 1107 } else if (key_val->ppp_proto == htons(PPP_MPLS_MC)) { 1108 key->basic.n_proto = htons(ETH_P_MPLS_MC); 1109 mask->basic.n_proto = cpu_to_be16(~0); 1110 } 1111 } else { 1112 key->basic.n_proto = 0; 1113 mask->basic.n_proto = cpu_to_be16(0); 1114 } 1115 } 1116 1117 static void fl_set_key_flag(u32 flower_key, u32 flower_mask, 1118 u32 *dissector_key, u32 *dissector_mask, 1119 u32 flower_flag_bit, u32 dissector_flag_bit) 1120 { 1121 if (flower_mask & flower_flag_bit) { 1122 *dissector_mask |= dissector_flag_bit; 1123 if (flower_key & flower_flag_bit) 1124 *dissector_key |= dissector_flag_bit; 1125 } 1126 } 1127 1128 static int fl_set_key_flags(struct nlattr **tb, u32 *flags_key, 1129 u32 *flags_mask, struct netlink_ext_ack *extack) 1130 { 1131 u32 key, mask; 1132 1133 /* mask is mandatory for flags */ 1134 if (!tb[TCA_FLOWER_KEY_FLAGS_MASK]) { 1135 NL_SET_ERR_MSG(extack, "Missing flags mask"); 1136 return -EINVAL; 1137 } 1138 1139 key = be32_to_cpu(nla_get_be32(tb[TCA_FLOWER_KEY_FLAGS])); 1140 mask = be32_to_cpu(nla_get_be32(tb[TCA_FLOWER_KEY_FLAGS_MASK])); 1141 1142 *flags_key = 0; 1143 *flags_mask = 0; 1144 1145 fl_set_key_flag(key, mask, flags_key, flags_mask, 1146 TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT); 1147 fl_set_key_flag(key, mask, flags_key, flags_mask, 1148 TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST, 1149 FLOW_DIS_FIRST_FRAG); 1150 1151 return 0; 1152 } 1153 1154 static void fl_set_key_ip(struct nlattr **tb, bool encap, 1155 struct flow_dissector_key_ip *key, 1156 struct flow_dissector_key_ip *mask) 1157 { 1158 int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS; 1159 int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL; 1160 int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK; 1161 int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK; 1162 1163 fl_set_key_val(tb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos)); 1164 fl_set_key_val(tb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl)); 1165 } 1166 1167 static int fl_set_geneve_opt(const struct nlattr *nla, struct fl_flow_key *key, 1168 int depth, int option_len, 1169 struct netlink_ext_ack *extack) 1170 { 1171 struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX + 1]; 1172 struct nlattr *class = NULL, *type = NULL, *data = NULL; 1173 struct geneve_opt *opt; 1174 int err, data_len = 0; 1175 1176 if (option_len > sizeof(struct geneve_opt)) 1177 data_len = option_len - sizeof(struct geneve_opt); 1178 1179 if (key->enc_opts.len > FLOW_DIS_TUN_OPTS_MAX - 4) 1180 return -ERANGE; 1181 1182 opt = (struct geneve_opt *)&key->enc_opts.data[key->enc_opts.len]; 1183 memset(opt, 0xff, option_len); 1184 opt->length = data_len / 4; 1185 opt->r1 = 0; 1186 opt->r2 = 0; 1187 opt->r3 = 0; 1188 1189 /* If no mask has been prodived we assume an exact match. */ 1190 if (!depth) 1191 return sizeof(struct geneve_opt) + data_len; 1192 1193 if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_GENEVE) { 1194 NL_SET_ERR_MSG(extack, "Non-geneve option type for mask"); 1195 return -EINVAL; 1196 } 1197 1198 err = nla_parse_nested_deprecated(tb, 1199 TCA_FLOWER_KEY_ENC_OPT_GENEVE_MAX, 1200 nla, geneve_opt_policy, extack); 1201 if (err < 0) 1202 return err; 1203 1204 /* We are not allowed to omit any of CLASS, TYPE or DATA 1205 * fields from the key. 1206 */ 1207 if (!option_len && 1208 (!tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS] || 1209 !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE] || 1210 !tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA])) { 1211 NL_SET_ERR_MSG(extack, "Missing tunnel key geneve option class, type or data"); 1212 return -EINVAL; 1213 } 1214 1215 /* Omitting any of CLASS, TYPE or DATA fields is allowed 1216 * for the mask. 1217 */ 1218 if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA]) { 1219 int new_len = key->enc_opts.len; 1220 1221 data = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA]; 1222 data_len = nla_len(data); 1223 if (data_len < 4) { 1224 NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is less than 4 bytes long"); 1225 return -ERANGE; 1226 } 1227 if (data_len % 4) { 1228 NL_SET_ERR_MSG(extack, "Tunnel key geneve option data is not a multiple of 4 bytes long"); 1229 return -ERANGE; 1230 } 1231 1232 new_len += sizeof(struct geneve_opt) + data_len; 1233 BUILD_BUG_ON(FLOW_DIS_TUN_OPTS_MAX != IP_TUNNEL_OPTS_MAX); 1234 if (new_len > FLOW_DIS_TUN_OPTS_MAX) { 1235 NL_SET_ERR_MSG(extack, "Tunnel options exceeds max size"); 1236 return -ERANGE; 1237 } 1238 opt->length = data_len / 4; 1239 memcpy(opt->opt_data, nla_data(data), data_len); 1240 } 1241 1242 if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS]) { 1243 class = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS]; 1244 opt->opt_class = nla_get_be16(class); 1245 } 1246 1247 if (tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE]) { 1248 type = tb[TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE]; 1249 opt->type = nla_get_u8(type); 1250 } 1251 1252 return sizeof(struct geneve_opt) + data_len; 1253 } 1254 1255 static int fl_set_vxlan_opt(const struct nlattr *nla, struct fl_flow_key *key, 1256 int depth, int option_len, 1257 struct netlink_ext_ack *extack) 1258 { 1259 struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX + 1]; 1260 struct vxlan_metadata *md; 1261 int err; 1262 1263 md = (struct vxlan_metadata *)&key->enc_opts.data[key->enc_opts.len]; 1264 memset(md, 0xff, sizeof(*md)); 1265 1266 if (!depth) 1267 return sizeof(*md); 1268 1269 if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_VXLAN) { 1270 NL_SET_ERR_MSG(extack, "Non-vxlan option type for mask"); 1271 return -EINVAL; 1272 } 1273 1274 err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_VXLAN_MAX, nla, 1275 vxlan_opt_policy, extack); 1276 if (err < 0) 1277 return err; 1278 1279 if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]) { 1280 NL_SET_ERR_MSG(extack, "Missing tunnel key vxlan option gbp"); 1281 return -EINVAL; 1282 } 1283 1284 if (tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]) { 1285 md->gbp = nla_get_u32(tb[TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP]); 1286 md->gbp &= VXLAN_GBP_MASK; 1287 } 1288 1289 return sizeof(*md); 1290 } 1291 1292 static int fl_set_erspan_opt(const struct nlattr *nla, struct fl_flow_key *key, 1293 int depth, int option_len, 1294 struct netlink_ext_ack *extack) 1295 { 1296 struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX + 1]; 1297 struct erspan_metadata *md; 1298 int err; 1299 1300 md = (struct erspan_metadata *)&key->enc_opts.data[key->enc_opts.len]; 1301 memset(md, 0xff, sizeof(*md)); 1302 md->version = 1; 1303 1304 if (!depth) 1305 return sizeof(*md); 1306 1307 if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_ERSPAN) { 1308 NL_SET_ERR_MSG(extack, "Non-erspan option type for mask"); 1309 return -EINVAL; 1310 } 1311 1312 err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_MAX, nla, 1313 erspan_opt_policy, extack); 1314 if (err < 0) 1315 return err; 1316 1317 if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]) { 1318 NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option ver"); 1319 return -EINVAL; 1320 } 1321 1322 if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]) 1323 md->version = nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER]); 1324 1325 if (md->version == 1) { 1326 if (!option_len && !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]) { 1327 NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option index"); 1328 return -EINVAL; 1329 } 1330 if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]) { 1331 nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX]; 1332 memset(&md->u, 0x00, sizeof(md->u)); 1333 md->u.index = nla_get_be32(nla); 1334 } 1335 } else if (md->version == 2) { 1336 if (!option_len && (!tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR] || 1337 !tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID])) { 1338 NL_SET_ERR_MSG(extack, "Missing tunnel key erspan option dir or hwid"); 1339 return -EINVAL; 1340 } 1341 if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR]) { 1342 nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR]; 1343 md->u.md2.dir = nla_get_u8(nla); 1344 } 1345 if (tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID]) { 1346 nla = tb[TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID]; 1347 set_hwid(&md->u.md2, nla_get_u8(nla)); 1348 } 1349 } else { 1350 NL_SET_ERR_MSG(extack, "Tunnel key erspan option ver is incorrect"); 1351 return -EINVAL; 1352 } 1353 1354 return sizeof(*md); 1355 } 1356 1357 static int fl_set_gtp_opt(const struct nlattr *nla, struct fl_flow_key *key, 1358 int depth, int option_len, 1359 struct netlink_ext_ack *extack) 1360 { 1361 struct nlattr *tb[TCA_FLOWER_KEY_ENC_OPT_GTP_MAX + 1]; 1362 struct gtp_pdu_session_info *sinfo; 1363 u8 len = key->enc_opts.len; 1364 int err; 1365 1366 sinfo = (struct gtp_pdu_session_info *)&key->enc_opts.data[len]; 1367 memset(sinfo, 0xff, option_len); 1368 1369 if (!depth) 1370 return sizeof(*sinfo); 1371 1372 if (nla_type(nla) != TCA_FLOWER_KEY_ENC_OPTS_GTP) { 1373 NL_SET_ERR_MSG_MOD(extack, "Non-gtp option type for mask"); 1374 return -EINVAL; 1375 } 1376 1377 err = nla_parse_nested(tb, TCA_FLOWER_KEY_ENC_OPT_GTP_MAX, nla, 1378 gtp_opt_policy, extack); 1379 if (err < 0) 1380 return err; 1381 1382 if (!option_len && 1383 (!tb[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE] || 1384 !tb[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI])) { 1385 NL_SET_ERR_MSG_MOD(extack, 1386 "Missing tunnel key gtp option pdu type or qfi"); 1387 return -EINVAL; 1388 } 1389 1390 if (tb[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE]) 1391 sinfo->pdu_type = 1392 nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE]); 1393 1394 if (tb[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI]) 1395 sinfo->qfi = nla_get_u8(tb[TCA_FLOWER_KEY_ENC_OPT_GTP_QFI]); 1396 1397 return sizeof(*sinfo); 1398 } 1399 1400 static int fl_set_enc_opt(struct nlattr **tb, struct fl_flow_key *key, 1401 struct fl_flow_key *mask, 1402 struct netlink_ext_ack *extack) 1403 { 1404 const struct nlattr *nla_enc_key, *nla_opt_key, *nla_opt_msk = NULL; 1405 int err, option_len, key_depth, msk_depth = 0; 1406 1407 err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS], 1408 TCA_FLOWER_KEY_ENC_OPTS_MAX, 1409 enc_opts_policy, extack); 1410 if (err) 1411 return err; 1412 1413 nla_enc_key = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS]); 1414 1415 if (tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]) { 1416 err = nla_validate_nested_deprecated(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK], 1417 TCA_FLOWER_KEY_ENC_OPTS_MAX, 1418 enc_opts_policy, extack); 1419 if (err) 1420 return err; 1421 1422 nla_opt_msk = nla_data(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]); 1423 msk_depth = nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS_MASK]); 1424 if (!nla_ok(nla_opt_msk, msk_depth)) { 1425 NL_SET_ERR_MSG(extack, "Invalid nested attribute for masks"); 1426 return -EINVAL; 1427 } 1428 } 1429 1430 nla_for_each_attr(nla_opt_key, nla_enc_key, 1431 nla_len(tb[TCA_FLOWER_KEY_ENC_OPTS]), key_depth) { 1432 switch (nla_type(nla_opt_key)) { 1433 case TCA_FLOWER_KEY_ENC_OPTS_GENEVE: 1434 if (key->enc_opts.dst_opt_type && 1435 key->enc_opts.dst_opt_type != TUNNEL_GENEVE_OPT) { 1436 NL_SET_ERR_MSG(extack, "Duplicate type for geneve options"); 1437 return -EINVAL; 1438 } 1439 option_len = 0; 1440 key->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT; 1441 option_len = fl_set_geneve_opt(nla_opt_key, key, 1442 key_depth, option_len, 1443 extack); 1444 if (option_len < 0) 1445 return option_len; 1446 1447 key->enc_opts.len += option_len; 1448 /* At the same time we need to parse through the mask 1449 * in order to verify exact and mask attribute lengths. 1450 */ 1451 mask->enc_opts.dst_opt_type = TUNNEL_GENEVE_OPT; 1452 option_len = fl_set_geneve_opt(nla_opt_msk, mask, 1453 msk_depth, option_len, 1454 extack); 1455 if (option_len < 0) 1456 return option_len; 1457 1458 mask->enc_opts.len += option_len; 1459 if (key->enc_opts.len != mask->enc_opts.len) { 1460 NL_SET_ERR_MSG(extack, "Key and mask miss aligned"); 1461 return -EINVAL; 1462 } 1463 break; 1464 case TCA_FLOWER_KEY_ENC_OPTS_VXLAN: 1465 if (key->enc_opts.dst_opt_type) { 1466 NL_SET_ERR_MSG(extack, "Duplicate type for vxlan options"); 1467 return -EINVAL; 1468 } 1469 option_len = 0; 1470 key->enc_opts.dst_opt_type = TUNNEL_VXLAN_OPT; 1471 option_len = fl_set_vxlan_opt(nla_opt_key, key, 1472 key_depth, option_len, 1473 extack); 1474 if (option_len < 0) 1475 return option_len; 1476 1477 key->enc_opts.len += option_len; 1478 /* At the same time we need to parse through the mask 1479 * in order to verify exact and mask attribute lengths. 1480 */ 1481 mask->enc_opts.dst_opt_type = TUNNEL_VXLAN_OPT; 1482 option_len = fl_set_vxlan_opt(nla_opt_msk, mask, 1483 msk_depth, option_len, 1484 extack); 1485 if (option_len < 0) 1486 return option_len; 1487 1488 mask->enc_opts.len += option_len; 1489 if (key->enc_opts.len != mask->enc_opts.len) { 1490 NL_SET_ERR_MSG(extack, "Key and mask miss aligned"); 1491 return -EINVAL; 1492 } 1493 break; 1494 case TCA_FLOWER_KEY_ENC_OPTS_ERSPAN: 1495 if (key->enc_opts.dst_opt_type) { 1496 NL_SET_ERR_MSG(extack, "Duplicate type for erspan options"); 1497 return -EINVAL; 1498 } 1499 option_len = 0; 1500 key->enc_opts.dst_opt_type = TUNNEL_ERSPAN_OPT; 1501 option_len = fl_set_erspan_opt(nla_opt_key, key, 1502 key_depth, option_len, 1503 extack); 1504 if (option_len < 0) 1505 return option_len; 1506 1507 key->enc_opts.len += option_len; 1508 /* At the same time we need to parse through the mask 1509 * in order to verify exact and mask attribute lengths. 1510 */ 1511 mask->enc_opts.dst_opt_type = TUNNEL_ERSPAN_OPT; 1512 option_len = fl_set_erspan_opt(nla_opt_msk, mask, 1513 msk_depth, option_len, 1514 extack); 1515 if (option_len < 0) 1516 return option_len; 1517 1518 mask->enc_opts.len += option_len; 1519 if (key->enc_opts.len != mask->enc_opts.len) { 1520 NL_SET_ERR_MSG(extack, "Key and mask miss aligned"); 1521 return -EINVAL; 1522 } 1523 break; 1524 case TCA_FLOWER_KEY_ENC_OPTS_GTP: 1525 if (key->enc_opts.dst_opt_type) { 1526 NL_SET_ERR_MSG_MOD(extack, 1527 "Duplicate type for gtp options"); 1528 return -EINVAL; 1529 } 1530 option_len = 0; 1531 key->enc_opts.dst_opt_type = TUNNEL_GTP_OPT; 1532 option_len = fl_set_gtp_opt(nla_opt_key, key, 1533 key_depth, option_len, 1534 extack); 1535 if (option_len < 0) 1536 return option_len; 1537 1538 key->enc_opts.len += option_len; 1539 /* At the same time we need to parse through the mask 1540 * in order to verify exact and mask attribute lengths. 1541 */ 1542 mask->enc_opts.dst_opt_type = TUNNEL_GTP_OPT; 1543 option_len = fl_set_gtp_opt(nla_opt_msk, mask, 1544 msk_depth, option_len, 1545 extack); 1546 if (option_len < 0) 1547 return option_len; 1548 1549 mask->enc_opts.len += option_len; 1550 if (key->enc_opts.len != mask->enc_opts.len) { 1551 NL_SET_ERR_MSG_MOD(extack, 1552 "Key and mask miss aligned"); 1553 return -EINVAL; 1554 } 1555 break; 1556 default: 1557 NL_SET_ERR_MSG(extack, "Unknown tunnel option type"); 1558 return -EINVAL; 1559 } 1560 1561 if (!msk_depth) 1562 continue; 1563 1564 if (!nla_ok(nla_opt_msk, msk_depth)) { 1565 NL_SET_ERR_MSG(extack, "A mask attribute is invalid"); 1566 return -EINVAL; 1567 } 1568 nla_opt_msk = nla_next(nla_opt_msk, &msk_depth); 1569 } 1570 1571 return 0; 1572 } 1573 1574 static int fl_validate_ct_state(u16 state, struct nlattr *tb, 1575 struct netlink_ext_ack *extack) 1576 { 1577 if (state && !(state & TCA_FLOWER_KEY_CT_FLAGS_TRACKED)) { 1578 NL_SET_ERR_MSG_ATTR(extack, tb, 1579 "no trk, so no other flag can be set"); 1580 return -EINVAL; 1581 } 1582 1583 if (state & TCA_FLOWER_KEY_CT_FLAGS_NEW && 1584 state & TCA_FLOWER_KEY_CT_FLAGS_ESTABLISHED) { 1585 NL_SET_ERR_MSG_ATTR(extack, tb, 1586 "new and est are mutually exclusive"); 1587 return -EINVAL; 1588 } 1589 1590 if (state & TCA_FLOWER_KEY_CT_FLAGS_INVALID && 1591 state & ~(TCA_FLOWER_KEY_CT_FLAGS_TRACKED | 1592 TCA_FLOWER_KEY_CT_FLAGS_INVALID)) { 1593 NL_SET_ERR_MSG_ATTR(extack, tb, 1594 "when inv is set, only trk may be set"); 1595 return -EINVAL; 1596 } 1597 1598 if (state & TCA_FLOWER_KEY_CT_FLAGS_NEW && 1599 state & TCA_FLOWER_KEY_CT_FLAGS_REPLY) { 1600 NL_SET_ERR_MSG_ATTR(extack, tb, 1601 "new and rpl are mutually exclusive"); 1602 return -EINVAL; 1603 } 1604 1605 return 0; 1606 } 1607 1608 static int fl_set_key_ct(struct nlattr **tb, 1609 struct flow_dissector_key_ct *key, 1610 struct flow_dissector_key_ct *mask, 1611 struct netlink_ext_ack *extack) 1612 { 1613 if (tb[TCA_FLOWER_KEY_CT_STATE]) { 1614 int err; 1615 1616 if (!IS_ENABLED(CONFIG_NF_CONNTRACK)) { 1617 NL_SET_ERR_MSG(extack, "Conntrack isn't enabled"); 1618 return -EOPNOTSUPP; 1619 } 1620 fl_set_key_val(tb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE, 1621 &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK, 1622 sizeof(key->ct_state)); 1623 1624 err = fl_validate_ct_state(key->ct_state & mask->ct_state, 1625 tb[TCA_FLOWER_KEY_CT_STATE_MASK], 1626 extack); 1627 if (err) 1628 return err; 1629 1630 } 1631 if (tb[TCA_FLOWER_KEY_CT_ZONE]) { 1632 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)) { 1633 NL_SET_ERR_MSG(extack, "Conntrack zones isn't enabled"); 1634 return -EOPNOTSUPP; 1635 } 1636 fl_set_key_val(tb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE, 1637 &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK, 1638 sizeof(key->ct_zone)); 1639 } 1640 if (tb[TCA_FLOWER_KEY_CT_MARK]) { 1641 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)) { 1642 NL_SET_ERR_MSG(extack, "Conntrack mark isn't enabled"); 1643 return -EOPNOTSUPP; 1644 } 1645 fl_set_key_val(tb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK, 1646 &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK, 1647 sizeof(key->ct_mark)); 1648 } 1649 if (tb[TCA_FLOWER_KEY_CT_LABELS]) { 1650 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)) { 1651 NL_SET_ERR_MSG(extack, "Conntrack labels aren't enabled"); 1652 return -EOPNOTSUPP; 1653 } 1654 fl_set_key_val(tb, key->ct_labels, TCA_FLOWER_KEY_CT_LABELS, 1655 mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK, 1656 sizeof(key->ct_labels)); 1657 } 1658 1659 return 0; 1660 } 1661 1662 static bool is_vlan_key(struct nlattr *tb, __be16 *ethertype, 1663 struct fl_flow_key *key, struct fl_flow_key *mask, 1664 int vthresh) 1665 { 1666 const bool good_num_of_vlans = key->num_of_vlans.num_of_vlans > vthresh; 1667 1668 if (!tb) { 1669 *ethertype = 0; 1670 return good_num_of_vlans; 1671 } 1672 1673 *ethertype = nla_get_be16(tb); 1674 if (good_num_of_vlans || eth_type_vlan(*ethertype)) 1675 return true; 1676 1677 key->basic.n_proto = *ethertype; 1678 mask->basic.n_proto = cpu_to_be16(~0); 1679 return false; 1680 } 1681 1682 static void fl_set_key_cfm_md_level(struct nlattr **tb, 1683 struct fl_flow_key *key, 1684 struct fl_flow_key *mask, 1685 struct netlink_ext_ack *extack) 1686 { 1687 u8 level; 1688 1689 if (!tb[TCA_FLOWER_KEY_CFM_MD_LEVEL]) 1690 return; 1691 1692 level = nla_get_u8(tb[TCA_FLOWER_KEY_CFM_MD_LEVEL]); 1693 key->cfm.mdl_ver = FIELD_PREP(FLOW_DIS_CFM_MDL_MASK, level); 1694 mask->cfm.mdl_ver = FLOW_DIS_CFM_MDL_MASK; 1695 } 1696 1697 static void fl_set_key_cfm_opcode(struct nlattr **tb, 1698 struct fl_flow_key *key, 1699 struct fl_flow_key *mask, 1700 struct netlink_ext_ack *extack) 1701 { 1702 fl_set_key_val(tb, &key->cfm.opcode, TCA_FLOWER_KEY_CFM_OPCODE, 1703 &mask->cfm.opcode, TCA_FLOWER_UNSPEC, 1704 sizeof(key->cfm.opcode)); 1705 } 1706 1707 static int fl_set_key_cfm(struct nlattr **tb, 1708 struct fl_flow_key *key, 1709 struct fl_flow_key *mask, 1710 struct netlink_ext_ack *extack) 1711 { 1712 struct nlattr *nla_cfm_opt[TCA_FLOWER_KEY_CFM_OPT_MAX]; 1713 int err; 1714 1715 if (!tb[TCA_FLOWER_KEY_CFM]) 1716 return 0; 1717 1718 err = nla_parse_nested(nla_cfm_opt, TCA_FLOWER_KEY_CFM_OPT_MAX, 1719 tb[TCA_FLOWER_KEY_CFM], cfm_opt_policy, extack); 1720 if (err < 0) 1721 return err; 1722 1723 fl_set_key_cfm_opcode(nla_cfm_opt, key, mask, extack); 1724 fl_set_key_cfm_md_level(nla_cfm_opt, key, mask, extack); 1725 1726 return 0; 1727 } 1728 1729 static int fl_set_key(struct net *net, struct nlattr **tb, 1730 struct fl_flow_key *key, struct fl_flow_key *mask, 1731 struct netlink_ext_ack *extack) 1732 { 1733 __be16 ethertype; 1734 int ret = 0; 1735 1736 if (tb[TCA_FLOWER_INDEV]) { 1737 int err = tcf_change_indev(net, tb[TCA_FLOWER_INDEV], extack); 1738 if (err < 0) 1739 return err; 1740 key->meta.ingress_ifindex = err; 1741 mask->meta.ingress_ifindex = 0xffffffff; 1742 } 1743 1744 fl_set_key_val(tb, &key->meta.l2_miss, TCA_FLOWER_L2_MISS, 1745 &mask->meta.l2_miss, TCA_FLOWER_UNSPEC, 1746 sizeof(key->meta.l2_miss)); 1747 1748 fl_set_key_val(tb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST, 1749 mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK, 1750 sizeof(key->eth.dst)); 1751 fl_set_key_val(tb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC, 1752 mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK, 1753 sizeof(key->eth.src)); 1754 fl_set_key_val(tb, &key->num_of_vlans, 1755 TCA_FLOWER_KEY_NUM_OF_VLANS, 1756 &mask->num_of_vlans, 1757 TCA_FLOWER_UNSPEC, 1758 sizeof(key->num_of_vlans)); 1759 1760 if (is_vlan_key(tb[TCA_FLOWER_KEY_ETH_TYPE], ðertype, key, mask, 0)) { 1761 fl_set_key_vlan(tb, ethertype, TCA_FLOWER_KEY_VLAN_ID, 1762 TCA_FLOWER_KEY_VLAN_PRIO, 1763 TCA_FLOWER_KEY_VLAN_ETH_TYPE, 1764 &key->vlan, &mask->vlan); 1765 1766 if (is_vlan_key(tb[TCA_FLOWER_KEY_VLAN_ETH_TYPE], 1767 ðertype, key, mask, 1)) { 1768 fl_set_key_vlan(tb, ethertype, 1769 TCA_FLOWER_KEY_CVLAN_ID, 1770 TCA_FLOWER_KEY_CVLAN_PRIO, 1771 TCA_FLOWER_KEY_CVLAN_ETH_TYPE, 1772 &key->cvlan, &mask->cvlan); 1773 fl_set_key_val(tb, &key->basic.n_proto, 1774 TCA_FLOWER_KEY_CVLAN_ETH_TYPE, 1775 &mask->basic.n_proto, 1776 TCA_FLOWER_UNSPEC, 1777 sizeof(key->basic.n_proto)); 1778 } 1779 } 1780 1781 if (key->basic.n_proto == htons(ETH_P_PPP_SES)) 1782 fl_set_key_pppoe(tb, &key->pppoe, &mask->pppoe, key, mask); 1783 1784 if (key->basic.n_proto == htons(ETH_P_IP) || 1785 key->basic.n_proto == htons(ETH_P_IPV6)) { 1786 fl_set_key_val(tb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO, 1787 &mask->basic.ip_proto, TCA_FLOWER_UNSPEC, 1788 sizeof(key->basic.ip_proto)); 1789 fl_set_key_ip(tb, false, &key->ip, &mask->ip); 1790 } 1791 1792 if (tb[TCA_FLOWER_KEY_IPV4_SRC] || tb[TCA_FLOWER_KEY_IPV4_DST]) { 1793 key->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 1794 mask->control.addr_type = ~0; 1795 fl_set_key_val(tb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC, 1796 &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK, 1797 sizeof(key->ipv4.src)); 1798 fl_set_key_val(tb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST, 1799 &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK, 1800 sizeof(key->ipv4.dst)); 1801 } else if (tb[TCA_FLOWER_KEY_IPV6_SRC] || tb[TCA_FLOWER_KEY_IPV6_DST]) { 1802 key->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 1803 mask->control.addr_type = ~0; 1804 fl_set_key_val(tb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC, 1805 &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK, 1806 sizeof(key->ipv6.src)); 1807 fl_set_key_val(tb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST, 1808 &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK, 1809 sizeof(key->ipv6.dst)); 1810 } 1811 1812 if (key->basic.ip_proto == IPPROTO_TCP) { 1813 fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC, 1814 &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK, 1815 sizeof(key->tp.src)); 1816 fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST, 1817 &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK, 1818 sizeof(key->tp.dst)); 1819 fl_set_key_val(tb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS, 1820 &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK, 1821 sizeof(key->tcp.flags)); 1822 } else if (key->basic.ip_proto == IPPROTO_UDP) { 1823 fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC, 1824 &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK, 1825 sizeof(key->tp.src)); 1826 fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST, 1827 &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK, 1828 sizeof(key->tp.dst)); 1829 } else if (key->basic.ip_proto == IPPROTO_SCTP) { 1830 fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC, 1831 &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK, 1832 sizeof(key->tp.src)); 1833 fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST, 1834 &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK, 1835 sizeof(key->tp.dst)); 1836 } else if (key->basic.n_proto == htons(ETH_P_IP) && 1837 key->basic.ip_proto == IPPROTO_ICMP) { 1838 fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV4_TYPE, 1839 &mask->icmp.type, 1840 TCA_FLOWER_KEY_ICMPV4_TYPE_MASK, 1841 sizeof(key->icmp.type)); 1842 fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV4_CODE, 1843 &mask->icmp.code, 1844 TCA_FLOWER_KEY_ICMPV4_CODE_MASK, 1845 sizeof(key->icmp.code)); 1846 } else if (key->basic.n_proto == htons(ETH_P_IPV6) && 1847 key->basic.ip_proto == IPPROTO_ICMPV6) { 1848 fl_set_key_val(tb, &key->icmp.type, TCA_FLOWER_KEY_ICMPV6_TYPE, 1849 &mask->icmp.type, 1850 TCA_FLOWER_KEY_ICMPV6_TYPE_MASK, 1851 sizeof(key->icmp.type)); 1852 fl_set_key_val(tb, &key->icmp.code, TCA_FLOWER_KEY_ICMPV6_CODE, 1853 &mask->icmp.code, 1854 TCA_FLOWER_KEY_ICMPV6_CODE_MASK, 1855 sizeof(key->icmp.code)); 1856 } else if (key->basic.n_proto == htons(ETH_P_MPLS_UC) || 1857 key->basic.n_proto == htons(ETH_P_MPLS_MC)) { 1858 ret = fl_set_key_mpls(tb, &key->mpls, &mask->mpls, extack); 1859 if (ret) 1860 return ret; 1861 } else if (key->basic.n_proto == htons(ETH_P_ARP) || 1862 key->basic.n_proto == htons(ETH_P_RARP)) { 1863 fl_set_key_val(tb, &key->arp.sip, TCA_FLOWER_KEY_ARP_SIP, 1864 &mask->arp.sip, TCA_FLOWER_KEY_ARP_SIP_MASK, 1865 sizeof(key->arp.sip)); 1866 fl_set_key_val(tb, &key->arp.tip, TCA_FLOWER_KEY_ARP_TIP, 1867 &mask->arp.tip, TCA_FLOWER_KEY_ARP_TIP_MASK, 1868 sizeof(key->arp.tip)); 1869 fl_set_key_val(tb, &key->arp.op, TCA_FLOWER_KEY_ARP_OP, 1870 &mask->arp.op, TCA_FLOWER_KEY_ARP_OP_MASK, 1871 sizeof(key->arp.op)); 1872 fl_set_key_val(tb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA, 1873 mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK, 1874 sizeof(key->arp.sha)); 1875 fl_set_key_val(tb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA, 1876 mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK, 1877 sizeof(key->arp.tha)); 1878 } else if (key->basic.ip_proto == IPPROTO_L2TP) { 1879 fl_set_key_val(tb, &key->l2tpv3.session_id, 1880 TCA_FLOWER_KEY_L2TPV3_SID, 1881 &mask->l2tpv3.session_id, TCA_FLOWER_UNSPEC, 1882 sizeof(key->l2tpv3.session_id)); 1883 } else if (key->basic.n_proto == htons(ETH_P_CFM)) { 1884 ret = fl_set_key_cfm(tb, key, mask, extack); 1885 if (ret) 1886 return ret; 1887 } 1888 1889 if (key->basic.ip_proto == IPPROTO_TCP || 1890 key->basic.ip_proto == IPPROTO_UDP || 1891 key->basic.ip_proto == IPPROTO_SCTP) { 1892 ret = fl_set_key_port_range(tb, key, mask, extack); 1893 if (ret) 1894 return ret; 1895 } 1896 1897 if (tb[TCA_FLOWER_KEY_ENC_IPV4_SRC] || 1898 tb[TCA_FLOWER_KEY_ENC_IPV4_DST]) { 1899 key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; 1900 mask->enc_control.addr_type = ~0; 1901 fl_set_key_val(tb, &key->enc_ipv4.src, 1902 TCA_FLOWER_KEY_ENC_IPV4_SRC, 1903 &mask->enc_ipv4.src, 1904 TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK, 1905 sizeof(key->enc_ipv4.src)); 1906 fl_set_key_val(tb, &key->enc_ipv4.dst, 1907 TCA_FLOWER_KEY_ENC_IPV4_DST, 1908 &mask->enc_ipv4.dst, 1909 TCA_FLOWER_KEY_ENC_IPV4_DST_MASK, 1910 sizeof(key->enc_ipv4.dst)); 1911 } 1912 1913 if (tb[TCA_FLOWER_KEY_ENC_IPV6_SRC] || 1914 tb[TCA_FLOWER_KEY_ENC_IPV6_DST]) { 1915 key->enc_control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 1916 mask->enc_control.addr_type = ~0; 1917 fl_set_key_val(tb, &key->enc_ipv6.src, 1918 TCA_FLOWER_KEY_ENC_IPV6_SRC, 1919 &mask->enc_ipv6.src, 1920 TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK, 1921 sizeof(key->enc_ipv6.src)); 1922 fl_set_key_val(tb, &key->enc_ipv6.dst, 1923 TCA_FLOWER_KEY_ENC_IPV6_DST, 1924 &mask->enc_ipv6.dst, 1925 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK, 1926 sizeof(key->enc_ipv6.dst)); 1927 } 1928 1929 fl_set_key_val(tb, &key->enc_key_id.keyid, TCA_FLOWER_KEY_ENC_KEY_ID, 1930 &mask->enc_key_id.keyid, TCA_FLOWER_UNSPEC, 1931 sizeof(key->enc_key_id.keyid)); 1932 1933 fl_set_key_val(tb, &key->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT, 1934 &mask->enc_tp.src, TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK, 1935 sizeof(key->enc_tp.src)); 1936 1937 fl_set_key_val(tb, &key->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT, 1938 &mask->enc_tp.dst, TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK, 1939 sizeof(key->enc_tp.dst)); 1940 1941 fl_set_key_ip(tb, true, &key->enc_ip, &mask->enc_ip); 1942 1943 fl_set_key_val(tb, &key->hash.hash, TCA_FLOWER_KEY_HASH, 1944 &mask->hash.hash, TCA_FLOWER_KEY_HASH_MASK, 1945 sizeof(key->hash.hash)); 1946 1947 if (tb[TCA_FLOWER_KEY_ENC_OPTS]) { 1948 ret = fl_set_enc_opt(tb, key, mask, extack); 1949 if (ret) 1950 return ret; 1951 } 1952 1953 ret = fl_set_key_ct(tb, &key->ct, &mask->ct, extack); 1954 if (ret) 1955 return ret; 1956 1957 if (tb[TCA_FLOWER_KEY_FLAGS]) 1958 ret = fl_set_key_flags(tb, &key->control.flags, 1959 &mask->control.flags, extack); 1960 1961 return ret; 1962 } 1963 1964 static void fl_mask_copy(struct fl_flow_mask *dst, 1965 struct fl_flow_mask *src) 1966 { 1967 const void *psrc = fl_key_get_start(&src->key, src); 1968 void *pdst = fl_key_get_start(&dst->key, src); 1969 1970 memcpy(pdst, psrc, fl_mask_range(src)); 1971 dst->range = src->range; 1972 } 1973 1974 static const struct rhashtable_params fl_ht_params = { 1975 .key_offset = offsetof(struct cls_fl_filter, mkey), /* base offset */ 1976 .head_offset = offsetof(struct cls_fl_filter, ht_node), 1977 .automatic_shrinking = true, 1978 }; 1979 1980 static int fl_init_mask_hashtable(struct fl_flow_mask *mask) 1981 { 1982 mask->filter_ht_params = fl_ht_params; 1983 mask->filter_ht_params.key_len = fl_mask_range(mask); 1984 mask->filter_ht_params.key_offset += mask->range.start; 1985 1986 return rhashtable_init(&mask->ht, &mask->filter_ht_params); 1987 } 1988 1989 #define FL_KEY_MEMBER_OFFSET(member) offsetof(struct fl_flow_key, member) 1990 #define FL_KEY_MEMBER_SIZE(member) sizeof_field(struct fl_flow_key, member) 1991 1992 #define FL_KEY_IS_MASKED(mask, member) \ 1993 memchr_inv(((char *)mask) + FL_KEY_MEMBER_OFFSET(member), \ 1994 0, FL_KEY_MEMBER_SIZE(member)) \ 1995 1996 #define FL_KEY_SET(keys, cnt, id, member) \ 1997 do { \ 1998 keys[cnt].key_id = id; \ 1999 keys[cnt].offset = FL_KEY_MEMBER_OFFSET(member); \ 2000 cnt++; \ 2001 } while(0); 2002 2003 #define FL_KEY_SET_IF_MASKED(mask, keys, cnt, id, member) \ 2004 do { \ 2005 if (FL_KEY_IS_MASKED(mask, member)) \ 2006 FL_KEY_SET(keys, cnt, id, member); \ 2007 } while(0); 2008 2009 static void fl_init_dissector(struct flow_dissector *dissector, 2010 struct fl_flow_key *mask) 2011 { 2012 struct flow_dissector_key keys[FLOW_DISSECTOR_KEY_MAX]; 2013 size_t cnt = 0; 2014 2015 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2016 FLOW_DISSECTOR_KEY_META, meta); 2017 FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_CONTROL, control); 2018 FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_BASIC, basic); 2019 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2020 FLOW_DISSECTOR_KEY_ETH_ADDRS, eth); 2021 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2022 FLOW_DISSECTOR_KEY_IPV4_ADDRS, ipv4); 2023 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2024 FLOW_DISSECTOR_KEY_IPV6_ADDRS, ipv6); 2025 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2026 FLOW_DISSECTOR_KEY_PORTS, tp); 2027 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2028 FLOW_DISSECTOR_KEY_PORTS_RANGE, tp_range); 2029 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2030 FLOW_DISSECTOR_KEY_IP, ip); 2031 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2032 FLOW_DISSECTOR_KEY_TCP, tcp); 2033 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2034 FLOW_DISSECTOR_KEY_ICMP, icmp); 2035 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2036 FLOW_DISSECTOR_KEY_ARP, arp); 2037 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2038 FLOW_DISSECTOR_KEY_MPLS, mpls); 2039 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2040 FLOW_DISSECTOR_KEY_VLAN, vlan); 2041 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2042 FLOW_DISSECTOR_KEY_CVLAN, cvlan); 2043 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2044 FLOW_DISSECTOR_KEY_ENC_KEYID, enc_key_id); 2045 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2046 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS, enc_ipv4); 2047 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2048 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS, enc_ipv6); 2049 if (FL_KEY_IS_MASKED(mask, enc_ipv4) || 2050 FL_KEY_IS_MASKED(mask, enc_ipv6)) 2051 FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_ENC_CONTROL, 2052 enc_control); 2053 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2054 FLOW_DISSECTOR_KEY_ENC_PORTS, enc_tp); 2055 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2056 FLOW_DISSECTOR_KEY_ENC_IP, enc_ip); 2057 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2058 FLOW_DISSECTOR_KEY_ENC_OPTS, enc_opts); 2059 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2060 FLOW_DISSECTOR_KEY_CT, ct); 2061 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2062 FLOW_DISSECTOR_KEY_HASH, hash); 2063 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2064 FLOW_DISSECTOR_KEY_NUM_OF_VLANS, num_of_vlans); 2065 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2066 FLOW_DISSECTOR_KEY_PPPOE, pppoe); 2067 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2068 FLOW_DISSECTOR_KEY_L2TPV3, l2tpv3); 2069 FL_KEY_SET_IF_MASKED(mask, keys, cnt, 2070 FLOW_DISSECTOR_KEY_CFM, cfm); 2071 2072 skb_flow_dissector_init(dissector, keys, cnt); 2073 } 2074 2075 static struct fl_flow_mask *fl_create_new_mask(struct cls_fl_head *head, 2076 struct fl_flow_mask *mask) 2077 { 2078 struct fl_flow_mask *newmask; 2079 int err; 2080 2081 newmask = kzalloc(sizeof(*newmask), GFP_KERNEL); 2082 if (!newmask) 2083 return ERR_PTR(-ENOMEM); 2084 2085 fl_mask_copy(newmask, mask); 2086 2087 if ((newmask->key.tp_range.tp_min.dst && 2088 newmask->key.tp_range.tp_max.dst) || 2089 (newmask->key.tp_range.tp_min.src && 2090 newmask->key.tp_range.tp_max.src)) 2091 newmask->flags |= TCA_FLOWER_MASK_FLAGS_RANGE; 2092 2093 err = fl_init_mask_hashtable(newmask); 2094 if (err) 2095 goto errout_free; 2096 2097 fl_init_dissector(&newmask->dissector, &newmask->key); 2098 2099 INIT_LIST_HEAD_RCU(&newmask->filters); 2100 2101 refcount_set(&newmask->refcnt, 1); 2102 err = rhashtable_replace_fast(&head->ht, &mask->ht_node, 2103 &newmask->ht_node, mask_ht_params); 2104 if (err) 2105 goto errout_destroy; 2106 2107 spin_lock(&head->masks_lock); 2108 list_add_tail_rcu(&newmask->list, &head->masks); 2109 spin_unlock(&head->masks_lock); 2110 2111 return newmask; 2112 2113 errout_destroy: 2114 rhashtable_destroy(&newmask->ht); 2115 errout_free: 2116 kfree(newmask); 2117 2118 return ERR_PTR(err); 2119 } 2120 2121 static int fl_check_assign_mask(struct cls_fl_head *head, 2122 struct cls_fl_filter *fnew, 2123 struct cls_fl_filter *fold, 2124 struct fl_flow_mask *mask) 2125 { 2126 struct fl_flow_mask *newmask; 2127 int ret = 0; 2128 2129 rcu_read_lock(); 2130 2131 /* Insert mask as temporary node to prevent concurrent creation of mask 2132 * with same key. Any concurrent lookups with same key will return 2133 * -EAGAIN because mask's refcnt is zero. 2134 */ 2135 fnew->mask = rhashtable_lookup_get_insert_fast(&head->ht, 2136 &mask->ht_node, 2137 mask_ht_params); 2138 if (!fnew->mask) { 2139 rcu_read_unlock(); 2140 2141 if (fold) { 2142 ret = -EINVAL; 2143 goto errout_cleanup; 2144 } 2145 2146 newmask = fl_create_new_mask(head, mask); 2147 if (IS_ERR(newmask)) { 2148 ret = PTR_ERR(newmask); 2149 goto errout_cleanup; 2150 } 2151 2152 fnew->mask = newmask; 2153 return 0; 2154 } else if (IS_ERR(fnew->mask)) { 2155 ret = PTR_ERR(fnew->mask); 2156 } else if (fold && fold->mask != fnew->mask) { 2157 ret = -EINVAL; 2158 } else if (!refcount_inc_not_zero(&fnew->mask->refcnt)) { 2159 /* Mask was deleted concurrently, try again */ 2160 ret = -EAGAIN; 2161 } 2162 rcu_read_unlock(); 2163 return ret; 2164 2165 errout_cleanup: 2166 rhashtable_remove_fast(&head->ht, &mask->ht_node, 2167 mask_ht_params); 2168 return ret; 2169 } 2170 2171 static bool fl_needs_tc_skb_ext(const struct fl_flow_key *mask) 2172 { 2173 return mask->meta.l2_miss; 2174 } 2175 2176 static int fl_set_parms(struct net *net, struct tcf_proto *tp, 2177 struct cls_fl_filter *f, struct fl_flow_mask *mask, 2178 unsigned long base, struct nlattr **tb, 2179 struct nlattr *est, 2180 struct fl_flow_tmplt *tmplt, 2181 u32 flags, u32 fl_flags, 2182 struct netlink_ext_ack *extack) 2183 { 2184 int err; 2185 2186 err = tcf_exts_validate_ex(net, tp, tb, est, &f->exts, flags, 2187 fl_flags, extack); 2188 if (err < 0) 2189 return err; 2190 2191 if (tb[TCA_FLOWER_CLASSID]) { 2192 f->res.classid = nla_get_u32(tb[TCA_FLOWER_CLASSID]); 2193 if (flags & TCA_ACT_FLAGS_NO_RTNL) 2194 rtnl_lock(); 2195 tcf_bind_filter(tp, &f->res, base); 2196 if (flags & TCA_ACT_FLAGS_NO_RTNL) 2197 rtnl_unlock(); 2198 } 2199 2200 err = fl_set_key(net, tb, &f->key, &mask->key, extack); 2201 if (err) 2202 return err; 2203 2204 fl_mask_update_range(mask); 2205 fl_set_masked_key(&f->mkey, &f->key, mask); 2206 2207 if (!fl_mask_fits_tmplt(tmplt, mask)) { 2208 NL_SET_ERR_MSG_MOD(extack, "Mask does not fit the template"); 2209 return -EINVAL; 2210 } 2211 2212 /* Enable tc skb extension if filter matches on data extracted from 2213 * this extension. 2214 */ 2215 if (fl_needs_tc_skb_ext(&mask->key)) { 2216 f->needs_tc_skb_ext = 1; 2217 tc_skb_ext_tc_enable(); 2218 } 2219 2220 return 0; 2221 } 2222 2223 static int fl_ht_insert_unique(struct cls_fl_filter *fnew, 2224 struct cls_fl_filter *fold, 2225 bool *in_ht) 2226 { 2227 struct fl_flow_mask *mask = fnew->mask; 2228 int err; 2229 2230 err = rhashtable_lookup_insert_fast(&mask->ht, 2231 &fnew->ht_node, 2232 mask->filter_ht_params); 2233 if (err) { 2234 *in_ht = false; 2235 /* It is okay if filter with same key exists when 2236 * overwriting. 2237 */ 2238 return fold && err == -EEXIST ? 0 : err; 2239 } 2240 2241 *in_ht = true; 2242 return 0; 2243 } 2244 2245 static int fl_change(struct net *net, struct sk_buff *in_skb, 2246 struct tcf_proto *tp, unsigned long base, 2247 u32 handle, struct nlattr **tca, 2248 void **arg, u32 flags, 2249 struct netlink_ext_ack *extack) 2250 { 2251 struct cls_fl_head *head = fl_head_dereference(tp); 2252 bool rtnl_held = !(flags & TCA_ACT_FLAGS_NO_RTNL); 2253 struct cls_fl_filter *fold = *arg; 2254 struct cls_fl_filter *fnew; 2255 struct fl_flow_mask *mask; 2256 struct nlattr **tb; 2257 bool in_ht; 2258 int err; 2259 2260 if (!tca[TCA_OPTIONS]) { 2261 err = -EINVAL; 2262 goto errout_fold; 2263 } 2264 2265 mask = kzalloc(sizeof(struct fl_flow_mask), GFP_KERNEL); 2266 if (!mask) { 2267 err = -ENOBUFS; 2268 goto errout_fold; 2269 } 2270 2271 tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL); 2272 if (!tb) { 2273 err = -ENOBUFS; 2274 goto errout_mask_alloc; 2275 } 2276 2277 err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX, 2278 tca[TCA_OPTIONS], fl_policy, NULL); 2279 if (err < 0) 2280 goto errout_tb; 2281 2282 if (fold && handle && fold->handle != handle) { 2283 err = -EINVAL; 2284 goto errout_tb; 2285 } 2286 2287 fnew = kzalloc(sizeof(*fnew), GFP_KERNEL); 2288 if (!fnew) { 2289 err = -ENOBUFS; 2290 goto errout_tb; 2291 } 2292 INIT_LIST_HEAD(&fnew->hw_list); 2293 refcount_set(&fnew->refcnt, 1); 2294 2295 if (tb[TCA_FLOWER_FLAGS]) { 2296 fnew->flags = nla_get_u32(tb[TCA_FLOWER_FLAGS]); 2297 2298 if (!tc_flags_valid(fnew->flags)) { 2299 kfree(fnew); 2300 err = -EINVAL; 2301 goto errout_tb; 2302 } 2303 } 2304 2305 if (!fold) { 2306 spin_lock(&tp->lock); 2307 if (!handle) { 2308 handle = 1; 2309 err = idr_alloc_u32(&head->handle_idr, NULL, &handle, 2310 INT_MAX, GFP_ATOMIC); 2311 } else { 2312 err = idr_alloc_u32(&head->handle_idr, NULL, &handle, 2313 handle, GFP_ATOMIC); 2314 2315 /* Filter with specified handle was concurrently 2316 * inserted after initial check in cls_api. This is not 2317 * necessarily an error if NLM_F_EXCL is not set in 2318 * message flags. Returning EAGAIN will cause cls_api to 2319 * try to update concurrently inserted rule. 2320 */ 2321 if (err == -ENOSPC) 2322 err = -EAGAIN; 2323 } 2324 spin_unlock(&tp->lock); 2325 2326 if (err) { 2327 kfree(fnew); 2328 goto errout_tb; 2329 } 2330 } 2331 fnew->handle = handle; 2332 2333 err = tcf_exts_init_ex(&fnew->exts, net, TCA_FLOWER_ACT, 0, tp, handle, 2334 !tc_skip_hw(fnew->flags)); 2335 if (err < 0) 2336 goto errout_idr; 2337 2338 err = fl_set_parms(net, tp, fnew, mask, base, tb, tca[TCA_RATE], 2339 tp->chain->tmplt_priv, flags, fnew->flags, 2340 extack); 2341 if (err) 2342 goto errout_idr; 2343 2344 err = fl_check_assign_mask(head, fnew, fold, mask); 2345 if (err) 2346 goto errout_idr; 2347 2348 err = fl_ht_insert_unique(fnew, fold, &in_ht); 2349 if (err) 2350 goto errout_mask; 2351 2352 if (!tc_skip_hw(fnew->flags)) { 2353 err = fl_hw_replace_filter(tp, fnew, rtnl_held, extack); 2354 if (err) 2355 goto errout_ht; 2356 } 2357 2358 if (!tc_in_hw(fnew->flags)) 2359 fnew->flags |= TCA_CLS_FLAGS_NOT_IN_HW; 2360 2361 spin_lock(&tp->lock); 2362 2363 /* tp was deleted concurrently. -EAGAIN will cause caller to lookup 2364 * proto again or create new one, if necessary. 2365 */ 2366 if (tp->deleting) { 2367 err = -EAGAIN; 2368 goto errout_hw; 2369 } 2370 2371 if (fold) { 2372 /* Fold filter was deleted concurrently. Retry lookup. */ 2373 if (fold->deleted) { 2374 err = -EAGAIN; 2375 goto errout_hw; 2376 } 2377 2378 fnew->handle = handle; 2379 2380 if (!in_ht) { 2381 struct rhashtable_params params = 2382 fnew->mask->filter_ht_params; 2383 2384 err = rhashtable_insert_fast(&fnew->mask->ht, 2385 &fnew->ht_node, 2386 params); 2387 if (err) 2388 goto errout_hw; 2389 in_ht = true; 2390 } 2391 2392 refcount_inc(&fnew->refcnt); 2393 rhashtable_remove_fast(&fold->mask->ht, 2394 &fold->ht_node, 2395 fold->mask->filter_ht_params); 2396 idr_replace(&head->handle_idr, fnew, fnew->handle); 2397 list_replace_rcu(&fold->list, &fnew->list); 2398 fold->deleted = true; 2399 2400 spin_unlock(&tp->lock); 2401 2402 fl_mask_put(head, fold->mask); 2403 if (!tc_skip_hw(fold->flags)) 2404 fl_hw_destroy_filter(tp, fold, rtnl_held, NULL); 2405 tcf_unbind_filter(tp, &fold->res); 2406 /* Caller holds reference to fold, so refcnt is always > 0 2407 * after this. 2408 */ 2409 refcount_dec(&fold->refcnt); 2410 __fl_put(fold); 2411 } else { 2412 idr_replace(&head->handle_idr, fnew, fnew->handle); 2413 2414 refcount_inc(&fnew->refcnt); 2415 list_add_tail_rcu(&fnew->list, &fnew->mask->filters); 2416 spin_unlock(&tp->lock); 2417 } 2418 2419 *arg = fnew; 2420 2421 kfree(tb); 2422 tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work); 2423 return 0; 2424 2425 errout_ht: 2426 spin_lock(&tp->lock); 2427 errout_hw: 2428 fnew->deleted = true; 2429 spin_unlock(&tp->lock); 2430 if (!tc_skip_hw(fnew->flags)) 2431 fl_hw_destroy_filter(tp, fnew, rtnl_held, NULL); 2432 if (in_ht) 2433 rhashtable_remove_fast(&fnew->mask->ht, &fnew->ht_node, 2434 fnew->mask->filter_ht_params); 2435 errout_mask: 2436 fl_mask_put(head, fnew->mask); 2437 errout_idr: 2438 if (!fold) 2439 idr_remove(&head->handle_idr, fnew->handle); 2440 __fl_put(fnew); 2441 errout_tb: 2442 kfree(tb); 2443 errout_mask_alloc: 2444 tcf_queue_work(&mask->rwork, fl_uninit_mask_free_work); 2445 errout_fold: 2446 if (fold) 2447 __fl_put(fold); 2448 return err; 2449 } 2450 2451 static int fl_delete(struct tcf_proto *tp, void *arg, bool *last, 2452 bool rtnl_held, struct netlink_ext_ack *extack) 2453 { 2454 struct cls_fl_head *head = fl_head_dereference(tp); 2455 struct cls_fl_filter *f = arg; 2456 bool last_on_mask; 2457 int err = 0; 2458 2459 err = __fl_delete(tp, f, &last_on_mask, rtnl_held, extack); 2460 *last = list_empty(&head->masks); 2461 __fl_put(f); 2462 2463 return err; 2464 } 2465 2466 static void fl_walk(struct tcf_proto *tp, struct tcf_walker *arg, 2467 bool rtnl_held) 2468 { 2469 struct cls_fl_head *head = fl_head_dereference(tp); 2470 unsigned long id = arg->cookie, tmp; 2471 struct cls_fl_filter *f; 2472 2473 arg->count = arg->skip; 2474 2475 rcu_read_lock(); 2476 idr_for_each_entry_continue_ul(&head->handle_idr, f, tmp, id) { 2477 /* don't return filters that are being deleted */ 2478 if (!f || !refcount_inc_not_zero(&f->refcnt)) 2479 continue; 2480 rcu_read_unlock(); 2481 2482 if (arg->fn(tp, f, arg) < 0) { 2483 __fl_put(f); 2484 arg->stop = 1; 2485 rcu_read_lock(); 2486 break; 2487 } 2488 __fl_put(f); 2489 arg->count++; 2490 rcu_read_lock(); 2491 } 2492 rcu_read_unlock(); 2493 arg->cookie = id; 2494 } 2495 2496 static struct cls_fl_filter * 2497 fl_get_next_hw_filter(struct tcf_proto *tp, struct cls_fl_filter *f, bool add) 2498 { 2499 struct cls_fl_head *head = fl_head_dereference(tp); 2500 2501 spin_lock(&tp->lock); 2502 if (list_empty(&head->hw_filters)) { 2503 spin_unlock(&tp->lock); 2504 return NULL; 2505 } 2506 2507 if (!f) 2508 f = list_entry(&head->hw_filters, struct cls_fl_filter, 2509 hw_list); 2510 list_for_each_entry_continue(f, &head->hw_filters, hw_list) { 2511 if (!(add && f->deleted) && refcount_inc_not_zero(&f->refcnt)) { 2512 spin_unlock(&tp->lock); 2513 return f; 2514 } 2515 } 2516 2517 spin_unlock(&tp->lock); 2518 return NULL; 2519 } 2520 2521 static int fl_reoffload(struct tcf_proto *tp, bool add, flow_setup_cb_t *cb, 2522 void *cb_priv, struct netlink_ext_ack *extack) 2523 { 2524 struct tcf_block *block = tp->chain->block; 2525 struct flow_cls_offload cls_flower = {}; 2526 struct cls_fl_filter *f = NULL; 2527 int err; 2528 2529 /* hw_filters list can only be changed by hw offload functions after 2530 * obtaining rtnl lock. Make sure it is not changed while reoffload is 2531 * iterating it. 2532 */ 2533 ASSERT_RTNL(); 2534 2535 while ((f = fl_get_next_hw_filter(tp, f, add))) { 2536 cls_flower.rule = 2537 flow_rule_alloc(tcf_exts_num_actions(&f->exts)); 2538 if (!cls_flower.rule) { 2539 __fl_put(f); 2540 return -ENOMEM; 2541 } 2542 2543 tc_cls_common_offload_init(&cls_flower.common, tp, f->flags, 2544 extack); 2545 cls_flower.command = add ? 2546 FLOW_CLS_REPLACE : FLOW_CLS_DESTROY; 2547 cls_flower.cookie = (unsigned long)f; 2548 cls_flower.rule->match.dissector = &f->mask->dissector; 2549 cls_flower.rule->match.mask = &f->mask->key; 2550 cls_flower.rule->match.key = &f->mkey; 2551 2552 err = tc_setup_offload_action(&cls_flower.rule->action, &f->exts, 2553 cls_flower.common.extack); 2554 if (err) { 2555 kfree(cls_flower.rule); 2556 if (tc_skip_sw(f->flags)) { 2557 __fl_put(f); 2558 return err; 2559 } 2560 goto next_flow; 2561 } 2562 2563 cls_flower.classid = f->res.classid; 2564 2565 err = tc_setup_cb_reoffload(block, tp, add, cb, 2566 TC_SETUP_CLSFLOWER, &cls_flower, 2567 cb_priv, &f->flags, 2568 &f->in_hw_count); 2569 tc_cleanup_offload_action(&cls_flower.rule->action); 2570 kfree(cls_flower.rule); 2571 2572 if (err) { 2573 __fl_put(f); 2574 return err; 2575 } 2576 next_flow: 2577 __fl_put(f); 2578 } 2579 2580 return 0; 2581 } 2582 2583 static void fl_hw_add(struct tcf_proto *tp, void *type_data) 2584 { 2585 struct flow_cls_offload *cls_flower = type_data; 2586 struct cls_fl_filter *f = 2587 (struct cls_fl_filter *) cls_flower->cookie; 2588 struct cls_fl_head *head = fl_head_dereference(tp); 2589 2590 spin_lock(&tp->lock); 2591 list_add(&f->hw_list, &head->hw_filters); 2592 spin_unlock(&tp->lock); 2593 } 2594 2595 static void fl_hw_del(struct tcf_proto *tp, void *type_data) 2596 { 2597 struct flow_cls_offload *cls_flower = type_data; 2598 struct cls_fl_filter *f = 2599 (struct cls_fl_filter *) cls_flower->cookie; 2600 2601 spin_lock(&tp->lock); 2602 if (!list_empty(&f->hw_list)) 2603 list_del_init(&f->hw_list); 2604 spin_unlock(&tp->lock); 2605 } 2606 2607 static int fl_hw_create_tmplt(struct tcf_chain *chain, 2608 struct fl_flow_tmplt *tmplt) 2609 { 2610 struct flow_cls_offload cls_flower = {}; 2611 struct tcf_block *block = chain->block; 2612 2613 cls_flower.rule = flow_rule_alloc(0); 2614 if (!cls_flower.rule) 2615 return -ENOMEM; 2616 2617 cls_flower.common.chain_index = chain->index; 2618 cls_flower.command = FLOW_CLS_TMPLT_CREATE; 2619 cls_flower.cookie = (unsigned long) tmplt; 2620 cls_flower.rule->match.dissector = &tmplt->dissector; 2621 cls_flower.rule->match.mask = &tmplt->mask; 2622 cls_flower.rule->match.key = &tmplt->dummy_key; 2623 2624 /* We don't care if driver (any of them) fails to handle this 2625 * call. It serves just as a hint for it. 2626 */ 2627 tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, true); 2628 kfree(cls_flower.rule); 2629 2630 return 0; 2631 } 2632 2633 static void fl_hw_destroy_tmplt(struct tcf_chain *chain, 2634 struct fl_flow_tmplt *tmplt) 2635 { 2636 struct flow_cls_offload cls_flower = {}; 2637 struct tcf_block *block = chain->block; 2638 2639 cls_flower.common.chain_index = chain->index; 2640 cls_flower.command = FLOW_CLS_TMPLT_DESTROY; 2641 cls_flower.cookie = (unsigned long) tmplt; 2642 2643 tc_setup_cb_call(block, TC_SETUP_CLSFLOWER, &cls_flower, false, true); 2644 } 2645 2646 static void *fl_tmplt_create(struct net *net, struct tcf_chain *chain, 2647 struct nlattr **tca, 2648 struct netlink_ext_ack *extack) 2649 { 2650 struct fl_flow_tmplt *tmplt; 2651 struct nlattr **tb; 2652 int err; 2653 2654 if (!tca[TCA_OPTIONS]) 2655 return ERR_PTR(-EINVAL); 2656 2657 tb = kcalloc(TCA_FLOWER_MAX + 1, sizeof(struct nlattr *), GFP_KERNEL); 2658 if (!tb) 2659 return ERR_PTR(-ENOBUFS); 2660 err = nla_parse_nested_deprecated(tb, TCA_FLOWER_MAX, 2661 tca[TCA_OPTIONS], fl_policy, NULL); 2662 if (err) 2663 goto errout_tb; 2664 2665 tmplt = kzalloc(sizeof(*tmplt), GFP_KERNEL); 2666 if (!tmplt) { 2667 err = -ENOMEM; 2668 goto errout_tb; 2669 } 2670 tmplt->chain = chain; 2671 err = fl_set_key(net, tb, &tmplt->dummy_key, &tmplt->mask, extack); 2672 if (err) 2673 goto errout_tmplt; 2674 2675 fl_init_dissector(&tmplt->dissector, &tmplt->mask); 2676 2677 err = fl_hw_create_tmplt(chain, tmplt); 2678 if (err) 2679 goto errout_tmplt; 2680 2681 kfree(tb); 2682 return tmplt; 2683 2684 errout_tmplt: 2685 kfree(tmplt); 2686 errout_tb: 2687 kfree(tb); 2688 return ERR_PTR(err); 2689 } 2690 2691 static void fl_tmplt_destroy(void *tmplt_priv) 2692 { 2693 struct fl_flow_tmplt *tmplt = tmplt_priv; 2694 2695 fl_hw_destroy_tmplt(tmplt->chain, tmplt); 2696 kfree(tmplt); 2697 } 2698 2699 static int fl_dump_key_val(struct sk_buff *skb, 2700 void *val, int val_type, 2701 void *mask, int mask_type, int len) 2702 { 2703 int err; 2704 2705 if (!memchr_inv(mask, 0, len)) 2706 return 0; 2707 err = nla_put(skb, val_type, len, val); 2708 if (err) 2709 return err; 2710 if (mask_type != TCA_FLOWER_UNSPEC) { 2711 err = nla_put(skb, mask_type, len, mask); 2712 if (err) 2713 return err; 2714 } 2715 return 0; 2716 } 2717 2718 static int fl_dump_key_port_range(struct sk_buff *skb, struct fl_flow_key *key, 2719 struct fl_flow_key *mask) 2720 { 2721 if (fl_dump_key_val(skb, &key->tp_range.tp_min.dst, 2722 TCA_FLOWER_KEY_PORT_DST_MIN, 2723 &mask->tp_range.tp_min.dst, TCA_FLOWER_UNSPEC, 2724 sizeof(key->tp_range.tp_min.dst)) || 2725 fl_dump_key_val(skb, &key->tp_range.tp_max.dst, 2726 TCA_FLOWER_KEY_PORT_DST_MAX, 2727 &mask->tp_range.tp_max.dst, TCA_FLOWER_UNSPEC, 2728 sizeof(key->tp_range.tp_max.dst)) || 2729 fl_dump_key_val(skb, &key->tp_range.tp_min.src, 2730 TCA_FLOWER_KEY_PORT_SRC_MIN, 2731 &mask->tp_range.tp_min.src, TCA_FLOWER_UNSPEC, 2732 sizeof(key->tp_range.tp_min.src)) || 2733 fl_dump_key_val(skb, &key->tp_range.tp_max.src, 2734 TCA_FLOWER_KEY_PORT_SRC_MAX, 2735 &mask->tp_range.tp_max.src, TCA_FLOWER_UNSPEC, 2736 sizeof(key->tp_range.tp_max.src))) 2737 return -1; 2738 2739 return 0; 2740 } 2741 2742 static int fl_dump_key_mpls_opt_lse(struct sk_buff *skb, 2743 struct flow_dissector_key_mpls *mpls_key, 2744 struct flow_dissector_key_mpls *mpls_mask, 2745 u8 lse_index) 2746 { 2747 struct flow_dissector_mpls_lse *lse_mask = &mpls_mask->ls[lse_index]; 2748 struct flow_dissector_mpls_lse *lse_key = &mpls_key->ls[lse_index]; 2749 int err; 2750 2751 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_DEPTH, 2752 lse_index + 1); 2753 if (err) 2754 return err; 2755 2756 if (lse_mask->mpls_ttl) { 2757 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_TTL, 2758 lse_key->mpls_ttl); 2759 if (err) 2760 return err; 2761 } 2762 if (lse_mask->mpls_bos) { 2763 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_BOS, 2764 lse_key->mpls_bos); 2765 if (err) 2766 return err; 2767 } 2768 if (lse_mask->mpls_tc) { 2769 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_TC, 2770 lse_key->mpls_tc); 2771 if (err) 2772 return err; 2773 } 2774 if (lse_mask->mpls_label) { 2775 err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_OPT_LSE_LABEL, 2776 lse_key->mpls_label); 2777 if (err) 2778 return err; 2779 } 2780 2781 return 0; 2782 } 2783 2784 static int fl_dump_key_mpls_opts(struct sk_buff *skb, 2785 struct flow_dissector_key_mpls *mpls_key, 2786 struct flow_dissector_key_mpls *mpls_mask) 2787 { 2788 struct nlattr *opts; 2789 struct nlattr *lse; 2790 u8 lse_index; 2791 int err; 2792 2793 opts = nla_nest_start(skb, TCA_FLOWER_KEY_MPLS_OPTS); 2794 if (!opts) 2795 return -EMSGSIZE; 2796 2797 for (lse_index = 0; lse_index < FLOW_DIS_MPLS_MAX; lse_index++) { 2798 if (!(mpls_mask->used_lses & 1 << lse_index)) 2799 continue; 2800 2801 lse = nla_nest_start(skb, TCA_FLOWER_KEY_MPLS_OPTS_LSE); 2802 if (!lse) { 2803 err = -EMSGSIZE; 2804 goto err_opts; 2805 } 2806 2807 err = fl_dump_key_mpls_opt_lse(skb, mpls_key, mpls_mask, 2808 lse_index); 2809 if (err) 2810 goto err_opts_lse; 2811 nla_nest_end(skb, lse); 2812 } 2813 nla_nest_end(skb, opts); 2814 2815 return 0; 2816 2817 err_opts_lse: 2818 nla_nest_cancel(skb, lse); 2819 err_opts: 2820 nla_nest_cancel(skb, opts); 2821 2822 return err; 2823 } 2824 2825 static int fl_dump_key_mpls(struct sk_buff *skb, 2826 struct flow_dissector_key_mpls *mpls_key, 2827 struct flow_dissector_key_mpls *mpls_mask) 2828 { 2829 struct flow_dissector_mpls_lse *lse_mask; 2830 struct flow_dissector_mpls_lse *lse_key; 2831 int err; 2832 2833 if (!mpls_mask->used_lses) 2834 return 0; 2835 2836 lse_mask = &mpls_mask->ls[0]; 2837 lse_key = &mpls_key->ls[0]; 2838 2839 /* For backward compatibility, don't use the MPLS nested attributes if 2840 * the rule can be expressed using the old attributes. 2841 */ 2842 if (mpls_mask->used_lses & ~1 || 2843 (!lse_mask->mpls_ttl && !lse_mask->mpls_bos && 2844 !lse_mask->mpls_tc && !lse_mask->mpls_label)) 2845 return fl_dump_key_mpls_opts(skb, mpls_key, mpls_mask); 2846 2847 if (lse_mask->mpls_ttl) { 2848 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TTL, 2849 lse_key->mpls_ttl); 2850 if (err) 2851 return err; 2852 } 2853 if (lse_mask->mpls_tc) { 2854 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_TC, 2855 lse_key->mpls_tc); 2856 if (err) 2857 return err; 2858 } 2859 if (lse_mask->mpls_label) { 2860 err = nla_put_u32(skb, TCA_FLOWER_KEY_MPLS_LABEL, 2861 lse_key->mpls_label); 2862 if (err) 2863 return err; 2864 } 2865 if (lse_mask->mpls_bos) { 2866 err = nla_put_u8(skb, TCA_FLOWER_KEY_MPLS_BOS, 2867 lse_key->mpls_bos); 2868 if (err) 2869 return err; 2870 } 2871 return 0; 2872 } 2873 2874 static int fl_dump_key_ip(struct sk_buff *skb, bool encap, 2875 struct flow_dissector_key_ip *key, 2876 struct flow_dissector_key_ip *mask) 2877 { 2878 int tos_key = encap ? TCA_FLOWER_KEY_ENC_IP_TOS : TCA_FLOWER_KEY_IP_TOS; 2879 int ttl_key = encap ? TCA_FLOWER_KEY_ENC_IP_TTL : TCA_FLOWER_KEY_IP_TTL; 2880 int tos_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TOS_MASK : TCA_FLOWER_KEY_IP_TOS_MASK; 2881 int ttl_mask = encap ? TCA_FLOWER_KEY_ENC_IP_TTL_MASK : TCA_FLOWER_KEY_IP_TTL_MASK; 2882 2883 if (fl_dump_key_val(skb, &key->tos, tos_key, &mask->tos, tos_mask, sizeof(key->tos)) || 2884 fl_dump_key_val(skb, &key->ttl, ttl_key, &mask->ttl, ttl_mask, sizeof(key->ttl))) 2885 return -1; 2886 2887 return 0; 2888 } 2889 2890 static int fl_dump_key_vlan(struct sk_buff *skb, 2891 int vlan_id_key, int vlan_prio_key, 2892 struct flow_dissector_key_vlan *vlan_key, 2893 struct flow_dissector_key_vlan *vlan_mask) 2894 { 2895 int err; 2896 2897 if (!memchr_inv(vlan_mask, 0, sizeof(*vlan_mask))) 2898 return 0; 2899 if (vlan_mask->vlan_id) { 2900 err = nla_put_u16(skb, vlan_id_key, 2901 vlan_key->vlan_id); 2902 if (err) 2903 return err; 2904 } 2905 if (vlan_mask->vlan_priority) { 2906 err = nla_put_u8(skb, vlan_prio_key, 2907 vlan_key->vlan_priority); 2908 if (err) 2909 return err; 2910 } 2911 return 0; 2912 } 2913 2914 static void fl_get_key_flag(u32 dissector_key, u32 dissector_mask, 2915 u32 *flower_key, u32 *flower_mask, 2916 u32 flower_flag_bit, u32 dissector_flag_bit) 2917 { 2918 if (dissector_mask & dissector_flag_bit) { 2919 *flower_mask |= flower_flag_bit; 2920 if (dissector_key & dissector_flag_bit) 2921 *flower_key |= flower_flag_bit; 2922 } 2923 } 2924 2925 static int fl_dump_key_flags(struct sk_buff *skb, u32 flags_key, u32 flags_mask) 2926 { 2927 u32 key, mask; 2928 __be32 _key, _mask; 2929 int err; 2930 2931 if (!memchr_inv(&flags_mask, 0, sizeof(flags_mask))) 2932 return 0; 2933 2934 key = 0; 2935 mask = 0; 2936 2937 fl_get_key_flag(flags_key, flags_mask, &key, &mask, 2938 TCA_FLOWER_KEY_FLAGS_IS_FRAGMENT, FLOW_DIS_IS_FRAGMENT); 2939 fl_get_key_flag(flags_key, flags_mask, &key, &mask, 2940 TCA_FLOWER_KEY_FLAGS_FRAG_IS_FIRST, 2941 FLOW_DIS_FIRST_FRAG); 2942 2943 _key = cpu_to_be32(key); 2944 _mask = cpu_to_be32(mask); 2945 2946 err = nla_put(skb, TCA_FLOWER_KEY_FLAGS, 4, &_key); 2947 if (err) 2948 return err; 2949 2950 return nla_put(skb, TCA_FLOWER_KEY_FLAGS_MASK, 4, &_mask); 2951 } 2952 2953 static int fl_dump_key_geneve_opt(struct sk_buff *skb, 2954 struct flow_dissector_key_enc_opts *enc_opts) 2955 { 2956 struct geneve_opt *opt; 2957 struct nlattr *nest; 2958 int opt_off = 0; 2959 2960 nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_GENEVE); 2961 if (!nest) 2962 goto nla_put_failure; 2963 2964 while (enc_opts->len > opt_off) { 2965 opt = (struct geneve_opt *)&enc_opts->data[opt_off]; 2966 2967 if (nla_put_be16(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_CLASS, 2968 opt->opt_class)) 2969 goto nla_put_failure; 2970 if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_TYPE, 2971 opt->type)) 2972 goto nla_put_failure; 2973 if (nla_put(skb, TCA_FLOWER_KEY_ENC_OPT_GENEVE_DATA, 2974 opt->length * 4, opt->opt_data)) 2975 goto nla_put_failure; 2976 2977 opt_off += sizeof(struct geneve_opt) + opt->length * 4; 2978 } 2979 nla_nest_end(skb, nest); 2980 return 0; 2981 2982 nla_put_failure: 2983 nla_nest_cancel(skb, nest); 2984 return -EMSGSIZE; 2985 } 2986 2987 static int fl_dump_key_vxlan_opt(struct sk_buff *skb, 2988 struct flow_dissector_key_enc_opts *enc_opts) 2989 { 2990 struct vxlan_metadata *md; 2991 struct nlattr *nest; 2992 2993 nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_VXLAN); 2994 if (!nest) 2995 goto nla_put_failure; 2996 2997 md = (struct vxlan_metadata *)&enc_opts->data[0]; 2998 if (nla_put_u32(skb, TCA_FLOWER_KEY_ENC_OPT_VXLAN_GBP, md->gbp)) 2999 goto nla_put_failure; 3000 3001 nla_nest_end(skb, nest); 3002 return 0; 3003 3004 nla_put_failure: 3005 nla_nest_cancel(skb, nest); 3006 return -EMSGSIZE; 3007 } 3008 3009 static int fl_dump_key_erspan_opt(struct sk_buff *skb, 3010 struct flow_dissector_key_enc_opts *enc_opts) 3011 { 3012 struct erspan_metadata *md; 3013 struct nlattr *nest; 3014 3015 nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_ERSPAN); 3016 if (!nest) 3017 goto nla_put_failure; 3018 3019 md = (struct erspan_metadata *)&enc_opts->data[0]; 3020 if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_VER, md->version)) 3021 goto nla_put_failure; 3022 3023 if (md->version == 1 && 3024 nla_put_be32(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_INDEX, md->u.index)) 3025 goto nla_put_failure; 3026 3027 if (md->version == 2 && 3028 (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_DIR, 3029 md->u.md2.dir) || 3030 nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_ERSPAN_HWID, 3031 get_hwid(&md->u.md2)))) 3032 goto nla_put_failure; 3033 3034 nla_nest_end(skb, nest); 3035 return 0; 3036 3037 nla_put_failure: 3038 nla_nest_cancel(skb, nest); 3039 return -EMSGSIZE; 3040 } 3041 3042 static int fl_dump_key_gtp_opt(struct sk_buff *skb, 3043 struct flow_dissector_key_enc_opts *enc_opts) 3044 3045 { 3046 struct gtp_pdu_session_info *session_info; 3047 struct nlattr *nest; 3048 3049 nest = nla_nest_start_noflag(skb, TCA_FLOWER_KEY_ENC_OPTS_GTP); 3050 if (!nest) 3051 goto nla_put_failure; 3052 3053 session_info = (struct gtp_pdu_session_info *)&enc_opts->data[0]; 3054 3055 if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GTP_PDU_TYPE, 3056 session_info->pdu_type)) 3057 goto nla_put_failure; 3058 3059 if (nla_put_u8(skb, TCA_FLOWER_KEY_ENC_OPT_GTP_QFI, session_info->qfi)) 3060 goto nla_put_failure; 3061 3062 nla_nest_end(skb, nest); 3063 return 0; 3064 3065 nla_put_failure: 3066 nla_nest_cancel(skb, nest); 3067 return -EMSGSIZE; 3068 } 3069 3070 static int fl_dump_key_ct(struct sk_buff *skb, 3071 struct flow_dissector_key_ct *key, 3072 struct flow_dissector_key_ct *mask) 3073 { 3074 if (IS_ENABLED(CONFIG_NF_CONNTRACK) && 3075 fl_dump_key_val(skb, &key->ct_state, TCA_FLOWER_KEY_CT_STATE, 3076 &mask->ct_state, TCA_FLOWER_KEY_CT_STATE_MASK, 3077 sizeof(key->ct_state))) 3078 goto nla_put_failure; 3079 3080 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) && 3081 fl_dump_key_val(skb, &key->ct_zone, TCA_FLOWER_KEY_CT_ZONE, 3082 &mask->ct_zone, TCA_FLOWER_KEY_CT_ZONE_MASK, 3083 sizeof(key->ct_zone))) 3084 goto nla_put_failure; 3085 3086 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && 3087 fl_dump_key_val(skb, &key->ct_mark, TCA_FLOWER_KEY_CT_MARK, 3088 &mask->ct_mark, TCA_FLOWER_KEY_CT_MARK_MASK, 3089 sizeof(key->ct_mark))) 3090 goto nla_put_failure; 3091 3092 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) && 3093 fl_dump_key_val(skb, &key->ct_labels, TCA_FLOWER_KEY_CT_LABELS, 3094 &mask->ct_labels, TCA_FLOWER_KEY_CT_LABELS_MASK, 3095 sizeof(key->ct_labels))) 3096 goto nla_put_failure; 3097 3098 return 0; 3099 3100 nla_put_failure: 3101 return -EMSGSIZE; 3102 } 3103 3104 static int fl_dump_key_cfm(struct sk_buff *skb, 3105 struct flow_dissector_key_cfm *key, 3106 struct flow_dissector_key_cfm *mask) 3107 { 3108 struct nlattr *opts; 3109 int err; 3110 u8 mdl; 3111 3112 if (!memchr_inv(mask, 0, sizeof(*mask))) 3113 return 0; 3114 3115 opts = nla_nest_start(skb, TCA_FLOWER_KEY_CFM); 3116 if (!opts) 3117 return -EMSGSIZE; 3118 3119 if (FIELD_GET(FLOW_DIS_CFM_MDL_MASK, mask->mdl_ver)) { 3120 mdl = FIELD_GET(FLOW_DIS_CFM_MDL_MASK, key->mdl_ver); 3121 err = nla_put_u8(skb, TCA_FLOWER_KEY_CFM_MD_LEVEL, mdl); 3122 if (err) 3123 goto err_cfm_opts; 3124 } 3125 3126 if (mask->opcode) { 3127 err = nla_put_u8(skb, TCA_FLOWER_KEY_CFM_OPCODE, key->opcode); 3128 if (err) 3129 goto err_cfm_opts; 3130 } 3131 3132 nla_nest_end(skb, opts); 3133 3134 return 0; 3135 3136 err_cfm_opts: 3137 nla_nest_cancel(skb, opts); 3138 return err; 3139 } 3140 3141 static int fl_dump_key_options(struct sk_buff *skb, int enc_opt_type, 3142 struct flow_dissector_key_enc_opts *enc_opts) 3143 { 3144 struct nlattr *nest; 3145 int err; 3146 3147 if (!enc_opts->len) 3148 return 0; 3149 3150 nest = nla_nest_start_noflag(skb, enc_opt_type); 3151 if (!nest) 3152 goto nla_put_failure; 3153 3154 switch (enc_opts->dst_opt_type) { 3155 case TUNNEL_GENEVE_OPT: 3156 err = fl_dump_key_geneve_opt(skb, enc_opts); 3157 if (err) 3158 goto nla_put_failure; 3159 break; 3160 case TUNNEL_VXLAN_OPT: 3161 err = fl_dump_key_vxlan_opt(skb, enc_opts); 3162 if (err) 3163 goto nla_put_failure; 3164 break; 3165 case TUNNEL_ERSPAN_OPT: 3166 err = fl_dump_key_erspan_opt(skb, enc_opts); 3167 if (err) 3168 goto nla_put_failure; 3169 break; 3170 case TUNNEL_GTP_OPT: 3171 err = fl_dump_key_gtp_opt(skb, enc_opts); 3172 if (err) 3173 goto nla_put_failure; 3174 break; 3175 default: 3176 goto nla_put_failure; 3177 } 3178 nla_nest_end(skb, nest); 3179 return 0; 3180 3181 nla_put_failure: 3182 nla_nest_cancel(skb, nest); 3183 return -EMSGSIZE; 3184 } 3185 3186 static int fl_dump_key_enc_opt(struct sk_buff *skb, 3187 struct flow_dissector_key_enc_opts *key_opts, 3188 struct flow_dissector_key_enc_opts *msk_opts) 3189 { 3190 int err; 3191 3192 err = fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS, key_opts); 3193 if (err) 3194 return err; 3195 3196 return fl_dump_key_options(skb, TCA_FLOWER_KEY_ENC_OPTS_MASK, msk_opts); 3197 } 3198 3199 static int fl_dump_key(struct sk_buff *skb, struct net *net, 3200 struct fl_flow_key *key, struct fl_flow_key *mask) 3201 { 3202 if (mask->meta.ingress_ifindex) { 3203 struct net_device *dev; 3204 3205 dev = __dev_get_by_index(net, key->meta.ingress_ifindex); 3206 if (dev && nla_put_string(skb, TCA_FLOWER_INDEV, dev->name)) 3207 goto nla_put_failure; 3208 } 3209 3210 if (fl_dump_key_val(skb, &key->meta.l2_miss, 3211 TCA_FLOWER_L2_MISS, &mask->meta.l2_miss, 3212 TCA_FLOWER_UNSPEC, sizeof(key->meta.l2_miss))) 3213 goto nla_put_failure; 3214 3215 if (fl_dump_key_val(skb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST, 3216 mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK, 3217 sizeof(key->eth.dst)) || 3218 fl_dump_key_val(skb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC, 3219 mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK, 3220 sizeof(key->eth.src)) || 3221 fl_dump_key_val(skb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE, 3222 &mask->basic.n_proto, TCA_FLOWER_UNSPEC, 3223 sizeof(key->basic.n_proto))) 3224 goto nla_put_failure; 3225 3226 if (mask->num_of_vlans.num_of_vlans) { 3227 if (nla_put_u8(skb, TCA_FLOWER_KEY_NUM_OF_VLANS, key->num_of_vlans.num_of_vlans)) 3228 goto nla_put_failure; 3229 } 3230 3231 if (fl_dump_key_mpls(skb, &key->mpls, &mask->mpls)) 3232 goto nla_put_failure; 3233 3234 if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_VLAN_ID, 3235 TCA_FLOWER_KEY_VLAN_PRIO, &key->vlan, &mask->vlan)) 3236 goto nla_put_failure; 3237 3238 if (fl_dump_key_vlan(skb, TCA_FLOWER_KEY_CVLAN_ID, 3239 TCA_FLOWER_KEY_CVLAN_PRIO, 3240 &key->cvlan, &mask->cvlan) || 3241 (mask->cvlan.vlan_tpid && 3242 nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE, 3243 key->cvlan.vlan_tpid))) 3244 goto nla_put_failure; 3245 3246 if (mask->basic.n_proto) { 3247 if (mask->cvlan.vlan_eth_type) { 3248 if (nla_put_be16(skb, TCA_FLOWER_KEY_CVLAN_ETH_TYPE, 3249 key->basic.n_proto)) 3250 goto nla_put_failure; 3251 } else if (mask->vlan.vlan_eth_type) { 3252 if (nla_put_be16(skb, TCA_FLOWER_KEY_VLAN_ETH_TYPE, 3253 key->vlan.vlan_eth_type)) 3254 goto nla_put_failure; 3255 } 3256 } 3257 3258 if ((key->basic.n_proto == htons(ETH_P_IP) || 3259 key->basic.n_proto == htons(ETH_P_IPV6)) && 3260 (fl_dump_key_val(skb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO, 3261 &mask->basic.ip_proto, TCA_FLOWER_UNSPEC, 3262 sizeof(key->basic.ip_proto)) || 3263 fl_dump_key_ip(skb, false, &key->ip, &mask->ip))) 3264 goto nla_put_failure; 3265 3266 if (mask->pppoe.session_id) { 3267 if (nla_put_be16(skb, TCA_FLOWER_KEY_PPPOE_SID, 3268 key->pppoe.session_id)) 3269 goto nla_put_failure; 3270 } 3271 if (mask->basic.n_proto && mask->pppoe.ppp_proto) { 3272 if (nla_put_be16(skb, TCA_FLOWER_KEY_PPP_PROTO, 3273 key->pppoe.ppp_proto)) 3274 goto nla_put_failure; 3275 } 3276 3277 if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS && 3278 (fl_dump_key_val(skb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC, 3279 &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK, 3280 sizeof(key->ipv4.src)) || 3281 fl_dump_key_val(skb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST, 3282 &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK, 3283 sizeof(key->ipv4.dst)))) 3284 goto nla_put_failure; 3285 else if (key->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS && 3286 (fl_dump_key_val(skb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC, 3287 &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK, 3288 sizeof(key->ipv6.src)) || 3289 fl_dump_key_val(skb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST, 3290 &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK, 3291 sizeof(key->ipv6.dst)))) 3292 goto nla_put_failure; 3293 3294 if (key->basic.ip_proto == IPPROTO_TCP && 3295 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC, 3296 &mask->tp.src, TCA_FLOWER_KEY_TCP_SRC_MASK, 3297 sizeof(key->tp.src)) || 3298 fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST, 3299 &mask->tp.dst, TCA_FLOWER_KEY_TCP_DST_MASK, 3300 sizeof(key->tp.dst)) || 3301 fl_dump_key_val(skb, &key->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS, 3302 &mask->tcp.flags, TCA_FLOWER_KEY_TCP_FLAGS_MASK, 3303 sizeof(key->tcp.flags)))) 3304 goto nla_put_failure; 3305 else if (key->basic.ip_proto == IPPROTO_UDP && 3306 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC, 3307 &mask->tp.src, TCA_FLOWER_KEY_UDP_SRC_MASK, 3308 sizeof(key->tp.src)) || 3309 fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST, 3310 &mask->tp.dst, TCA_FLOWER_KEY_UDP_DST_MASK, 3311 sizeof(key->tp.dst)))) 3312 goto nla_put_failure; 3313 else if (key->basic.ip_proto == IPPROTO_SCTP && 3314 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_SCTP_SRC, 3315 &mask->tp.src, TCA_FLOWER_KEY_SCTP_SRC_MASK, 3316 sizeof(key->tp.src)) || 3317 fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_SCTP_DST, 3318 &mask->tp.dst, TCA_FLOWER_KEY_SCTP_DST_MASK, 3319 sizeof(key->tp.dst)))) 3320 goto nla_put_failure; 3321 else if (key->basic.n_proto == htons(ETH_P_IP) && 3322 key->basic.ip_proto == IPPROTO_ICMP && 3323 (fl_dump_key_val(skb, &key->icmp.type, 3324 TCA_FLOWER_KEY_ICMPV4_TYPE, &mask->icmp.type, 3325 TCA_FLOWER_KEY_ICMPV4_TYPE_MASK, 3326 sizeof(key->icmp.type)) || 3327 fl_dump_key_val(skb, &key->icmp.code, 3328 TCA_FLOWER_KEY_ICMPV4_CODE, &mask->icmp.code, 3329 TCA_FLOWER_KEY_ICMPV4_CODE_MASK, 3330 sizeof(key->icmp.code)))) 3331 goto nla_put_failure; 3332 else if (key->basic.n_proto == htons(ETH_P_IPV6) && 3333 key->basic.ip_proto == IPPROTO_ICMPV6 && 3334 (fl_dump_key_val(skb, &key->icmp.type, 3335 TCA_FLOWER_KEY_ICMPV6_TYPE, &mask->icmp.type, 3336 TCA_FLOWER_KEY_ICMPV6_TYPE_MASK, 3337 sizeof(key->icmp.type)) || 3338 fl_dump_key_val(skb, &key->icmp.code, 3339 TCA_FLOWER_KEY_ICMPV6_CODE, &mask->icmp.code, 3340 TCA_FLOWER_KEY_ICMPV6_CODE_MASK, 3341 sizeof(key->icmp.code)))) 3342 goto nla_put_failure; 3343 else if ((key->basic.n_proto == htons(ETH_P_ARP) || 3344 key->basic.n_proto == htons(ETH_P_RARP)) && 3345 (fl_dump_key_val(skb, &key->arp.sip, 3346 TCA_FLOWER_KEY_ARP_SIP, &mask->arp.sip, 3347 TCA_FLOWER_KEY_ARP_SIP_MASK, 3348 sizeof(key->arp.sip)) || 3349 fl_dump_key_val(skb, &key->arp.tip, 3350 TCA_FLOWER_KEY_ARP_TIP, &mask->arp.tip, 3351 TCA_FLOWER_KEY_ARP_TIP_MASK, 3352 sizeof(key->arp.tip)) || 3353 fl_dump_key_val(skb, &key->arp.op, 3354 TCA_FLOWER_KEY_ARP_OP, &mask->arp.op, 3355 TCA_FLOWER_KEY_ARP_OP_MASK, 3356 sizeof(key->arp.op)) || 3357 fl_dump_key_val(skb, key->arp.sha, TCA_FLOWER_KEY_ARP_SHA, 3358 mask->arp.sha, TCA_FLOWER_KEY_ARP_SHA_MASK, 3359 sizeof(key->arp.sha)) || 3360 fl_dump_key_val(skb, key->arp.tha, TCA_FLOWER_KEY_ARP_THA, 3361 mask->arp.tha, TCA_FLOWER_KEY_ARP_THA_MASK, 3362 sizeof(key->arp.tha)))) 3363 goto nla_put_failure; 3364 else if (key->basic.ip_proto == IPPROTO_L2TP && 3365 fl_dump_key_val(skb, &key->l2tpv3.session_id, 3366 TCA_FLOWER_KEY_L2TPV3_SID, 3367 &mask->l2tpv3.session_id, 3368 TCA_FLOWER_UNSPEC, 3369 sizeof(key->l2tpv3.session_id))) 3370 goto nla_put_failure; 3371 3372 if ((key->basic.ip_proto == IPPROTO_TCP || 3373 key->basic.ip_proto == IPPROTO_UDP || 3374 key->basic.ip_proto == IPPROTO_SCTP) && 3375 fl_dump_key_port_range(skb, key, mask)) 3376 goto nla_put_failure; 3377 3378 if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS && 3379 (fl_dump_key_val(skb, &key->enc_ipv4.src, 3380 TCA_FLOWER_KEY_ENC_IPV4_SRC, &mask->enc_ipv4.src, 3381 TCA_FLOWER_KEY_ENC_IPV4_SRC_MASK, 3382 sizeof(key->enc_ipv4.src)) || 3383 fl_dump_key_val(skb, &key->enc_ipv4.dst, 3384 TCA_FLOWER_KEY_ENC_IPV4_DST, &mask->enc_ipv4.dst, 3385 TCA_FLOWER_KEY_ENC_IPV4_DST_MASK, 3386 sizeof(key->enc_ipv4.dst)))) 3387 goto nla_put_failure; 3388 else if (key->enc_control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS && 3389 (fl_dump_key_val(skb, &key->enc_ipv6.src, 3390 TCA_FLOWER_KEY_ENC_IPV6_SRC, &mask->enc_ipv6.src, 3391 TCA_FLOWER_KEY_ENC_IPV6_SRC_MASK, 3392 sizeof(key->enc_ipv6.src)) || 3393 fl_dump_key_val(skb, &key->enc_ipv6.dst, 3394 TCA_FLOWER_KEY_ENC_IPV6_DST, 3395 &mask->enc_ipv6.dst, 3396 TCA_FLOWER_KEY_ENC_IPV6_DST_MASK, 3397 sizeof(key->enc_ipv6.dst)))) 3398 goto nla_put_failure; 3399 3400 if (fl_dump_key_val(skb, &key->enc_key_id, TCA_FLOWER_KEY_ENC_KEY_ID, 3401 &mask->enc_key_id, TCA_FLOWER_UNSPEC, 3402 sizeof(key->enc_key_id)) || 3403 fl_dump_key_val(skb, &key->enc_tp.src, 3404 TCA_FLOWER_KEY_ENC_UDP_SRC_PORT, 3405 &mask->enc_tp.src, 3406 TCA_FLOWER_KEY_ENC_UDP_SRC_PORT_MASK, 3407 sizeof(key->enc_tp.src)) || 3408 fl_dump_key_val(skb, &key->enc_tp.dst, 3409 TCA_FLOWER_KEY_ENC_UDP_DST_PORT, 3410 &mask->enc_tp.dst, 3411 TCA_FLOWER_KEY_ENC_UDP_DST_PORT_MASK, 3412 sizeof(key->enc_tp.dst)) || 3413 fl_dump_key_ip(skb, true, &key->enc_ip, &mask->enc_ip) || 3414 fl_dump_key_enc_opt(skb, &key->enc_opts, &mask->enc_opts)) 3415 goto nla_put_failure; 3416 3417 if (fl_dump_key_ct(skb, &key->ct, &mask->ct)) 3418 goto nla_put_failure; 3419 3420 if (fl_dump_key_flags(skb, key->control.flags, mask->control.flags)) 3421 goto nla_put_failure; 3422 3423 if (fl_dump_key_val(skb, &key->hash.hash, TCA_FLOWER_KEY_HASH, 3424 &mask->hash.hash, TCA_FLOWER_KEY_HASH_MASK, 3425 sizeof(key->hash.hash))) 3426 goto nla_put_failure; 3427 3428 if (fl_dump_key_cfm(skb, &key->cfm, &mask->cfm)) 3429 goto nla_put_failure; 3430 3431 return 0; 3432 3433 nla_put_failure: 3434 return -EMSGSIZE; 3435 } 3436 3437 static int fl_dump(struct net *net, struct tcf_proto *tp, void *fh, 3438 struct sk_buff *skb, struct tcmsg *t, bool rtnl_held) 3439 { 3440 struct cls_fl_filter *f = fh; 3441 struct nlattr *nest; 3442 struct fl_flow_key *key, *mask; 3443 bool skip_hw; 3444 3445 if (!f) 3446 return skb->len; 3447 3448 t->tcm_handle = f->handle; 3449 3450 nest = nla_nest_start_noflag(skb, TCA_OPTIONS); 3451 if (!nest) 3452 goto nla_put_failure; 3453 3454 spin_lock(&tp->lock); 3455 3456 if (f->res.classid && 3457 nla_put_u32(skb, TCA_FLOWER_CLASSID, f->res.classid)) 3458 goto nla_put_failure_locked; 3459 3460 key = &f->key; 3461 mask = &f->mask->key; 3462 skip_hw = tc_skip_hw(f->flags); 3463 3464 if (fl_dump_key(skb, net, key, mask)) 3465 goto nla_put_failure_locked; 3466 3467 if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags)) 3468 goto nla_put_failure_locked; 3469 3470 spin_unlock(&tp->lock); 3471 3472 if (!skip_hw) 3473 fl_hw_update_stats(tp, f, rtnl_held); 3474 3475 if (nla_put_u32(skb, TCA_FLOWER_IN_HW_COUNT, f->in_hw_count)) 3476 goto nla_put_failure; 3477 3478 if (tcf_exts_dump(skb, &f->exts)) 3479 goto nla_put_failure; 3480 3481 nla_nest_end(skb, nest); 3482 3483 if (tcf_exts_dump_stats(skb, &f->exts) < 0) 3484 goto nla_put_failure; 3485 3486 return skb->len; 3487 3488 nla_put_failure_locked: 3489 spin_unlock(&tp->lock); 3490 nla_put_failure: 3491 nla_nest_cancel(skb, nest); 3492 return -1; 3493 } 3494 3495 static int fl_terse_dump(struct net *net, struct tcf_proto *tp, void *fh, 3496 struct sk_buff *skb, struct tcmsg *t, bool rtnl_held) 3497 { 3498 struct cls_fl_filter *f = fh; 3499 struct nlattr *nest; 3500 bool skip_hw; 3501 3502 if (!f) 3503 return skb->len; 3504 3505 t->tcm_handle = f->handle; 3506 3507 nest = nla_nest_start_noflag(skb, TCA_OPTIONS); 3508 if (!nest) 3509 goto nla_put_failure; 3510 3511 spin_lock(&tp->lock); 3512 3513 skip_hw = tc_skip_hw(f->flags); 3514 3515 if (f->flags && nla_put_u32(skb, TCA_FLOWER_FLAGS, f->flags)) 3516 goto nla_put_failure_locked; 3517 3518 spin_unlock(&tp->lock); 3519 3520 if (!skip_hw) 3521 fl_hw_update_stats(tp, f, rtnl_held); 3522 3523 if (tcf_exts_terse_dump(skb, &f->exts)) 3524 goto nla_put_failure; 3525 3526 nla_nest_end(skb, nest); 3527 3528 return skb->len; 3529 3530 nla_put_failure_locked: 3531 spin_unlock(&tp->lock); 3532 nla_put_failure: 3533 nla_nest_cancel(skb, nest); 3534 return -1; 3535 } 3536 3537 static int fl_tmplt_dump(struct sk_buff *skb, struct net *net, void *tmplt_priv) 3538 { 3539 struct fl_flow_tmplt *tmplt = tmplt_priv; 3540 struct fl_flow_key *key, *mask; 3541 struct nlattr *nest; 3542 3543 nest = nla_nest_start_noflag(skb, TCA_OPTIONS); 3544 if (!nest) 3545 goto nla_put_failure; 3546 3547 key = &tmplt->dummy_key; 3548 mask = &tmplt->mask; 3549 3550 if (fl_dump_key(skb, net, key, mask)) 3551 goto nla_put_failure; 3552 3553 nla_nest_end(skb, nest); 3554 3555 return skb->len; 3556 3557 nla_put_failure: 3558 nla_nest_cancel(skb, nest); 3559 return -EMSGSIZE; 3560 } 3561 3562 static void fl_bind_class(void *fh, u32 classid, unsigned long cl, void *q, 3563 unsigned long base) 3564 { 3565 struct cls_fl_filter *f = fh; 3566 3567 tc_cls_bind_class(classid, cl, q, &f->res, base); 3568 } 3569 3570 static bool fl_delete_empty(struct tcf_proto *tp) 3571 { 3572 struct cls_fl_head *head = fl_head_dereference(tp); 3573 3574 spin_lock(&tp->lock); 3575 tp->deleting = idr_is_empty(&head->handle_idr); 3576 spin_unlock(&tp->lock); 3577 3578 return tp->deleting; 3579 } 3580 3581 static struct tcf_proto_ops cls_fl_ops __read_mostly = { 3582 .kind = "flower", 3583 .classify = fl_classify, 3584 .init = fl_init, 3585 .destroy = fl_destroy, 3586 .get = fl_get, 3587 .put = fl_put, 3588 .change = fl_change, 3589 .delete = fl_delete, 3590 .delete_empty = fl_delete_empty, 3591 .walk = fl_walk, 3592 .reoffload = fl_reoffload, 3593 .hw_add = fl_hw_add, 3594 .hw_del = fl_hw_del, 3595 .dump = fl_dump, 3596 .terse_dump = fl_terse_dump, 3597 .bind_class = fl_bind_class, 3598 .tmplt_create = fl_tmplt_create, 3599 .tmplt_destroy = fl_tmplt_destroy, 3600 .tmplt_dump = fl_tmplt_dump, 3601 .get_exts = fl_get_exts, 3602 .owner = THIS_MODULE, 3603 .flags = TCF_PROTO_OPS_DOIT_UNLOCKED, 3604 }; 3605 3606 static int __init cls_fl_init(void) 3607 { 3608 return register_tcf_proto_ops(&cls_fl_ops); 3609 } 3610 3611 static void __exit cls_fl_exit(void) 3612 { 3613 unregister_tcf_proto_ops(&cls_fl_ops); 3614 } 3615 3616 module_init(cls_fl_init); 3617 module_exit(cls_fl_exit); 3618 3619 MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>"); 3620 MODULE_DESCRIPTION("Flower classifier"); 3621 MODULE_LICENSE("GPL v2"); 3622