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