1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * SR-IPv6 implementation 4 * 5 * Authors: 6 * David Lebrun <david.lebrun@uclouvain.be> 7 * eBPF support: Mathieu Xhonneux <m.xhonneux@gmail.com> 8 */ 9 10 #include <linux/filter.h> 11 #include <linux/types.h> 12 #include <linux/skbuff.h> 13 #include <linux/net.h> 14 #include <linux/module.h> 15 #include <net/ip.h> 16 #include <net/lwtunnel.h> 17 #include <net/netevent.h> 18 #include <net/netns/generic.h> 19 #include <net/ip6_fib.h> 20 #include <net/route.h> 21 #include <net/seg6.h> 22 #include <linux/seg6.h> 23 #include <linux/seg6_local.h> 24 #include <net/addrconf.h> 25 #include <net/ip6_route.h> 26 #include <net/dst_cache.h> 27 #include <net/ip_tunnels.h> 28 #ifdef CONFIG_IPV6_SEG6_HMAC 29 #include <net/seg6_hmac.h> 30 #endif 31 #include <net/seg6_local.h> 32 #include <linux/etherdevice.h> 33 #include <linux/bpf.h> 34 #include <linux/netfilter.h> 35 36 #define SEG6_F_ATTR(i) BIT(i) 37 38 struct seg6_local_lwt; 39 40 /* callbacks used for customizing the creation and destruction of a behavior */ 41 struct seg6_local_lwtunnel_ops { 42 int (*build_state)(struct seg6_local_lwt *slwt, const void *cfg, 43 struct netlink_ext_ack *extack); 44 void (*destroy_state)(struct seg6_local_lwt *slwt); 45 }; 46 47 struct seg6_action_desc { 48 int action; 49 unsigned long attrs; 50 51 /* The optattrs field is used for specifying all the optional 52 * attributes supported by a specific behavior. 53 * It means that if one of these attributes is not provided in the 54 * netlink message during the behavior creation, no errors will be 55 * returned to the userspace. 56 * 57 * Each attribute can be only of two types (mutually exclusive): 58 * 1) required or 2) optional. 59 * Every user MUST obey to this rule! If you set an attribute as 60 * required the same attribute CANNOT be set as optional and vice 61 * versa. 62 */ 63 unsigned long optattrs; 64 65 int (*input)(struct sk_buff *skb, struct seg6_local_lwt *slwt); 66 int static_headroom; 67 68 struct seg6_local_lwtunnel_ops slwt_ops; 69 }; 70 71 struct bpf_lwt_prog { 72 struct bpf_prog *prog; 73 char *name; 74 }; 75 76 /* default length values (expressed in bits) for both Locator-Block and 77 * Locator-Node Function. 78 * 79 * Both SEG6_LOCAL_LCBLOCK_DBITS and SEG6_LOCAL_LCNODE_FN_DBITS *must* be: 80 * i) greater than 0; 81 * ii) evenly divisible by 8. In other terms, the lengths of the 82 * Locator-Block and Locator-Node Function must be byte-aligned (we can 83 * relax this constraint in the future if really needed). 84 * 85 * Moreover, a third condition must hold: 86 * iii) SEG6_LOCAL_LCBLOCK_DBITS + SEG6_LOCAL_LCNODE_FN_DBITS <= 128. 87 * 88 * The correctness of SEG6_LOCAL_LCBLOCK_DBITS and SEG6_LOCAL_LCNODE_FN_DBITS 89 * values are checked during the kernel compilation. If the compilation stops, 90 * check the value of these parameters to see if they meet conditions (i), (ii) 91 * and (iii). 92 */ 93 #define SEG6_LOCAL_LCBLOCK_DBITS 32 94 #define SEG6_LOCAL_LCNODE_FN_DBITS 16 95 96 /* The following next_csid_chk_{cntr,lcblock,lcblock_fn}_bits macros can be 97 * used directly to check whether the lengths (in bits) of Locator-Block and 98 * Locator-Node Function are valid according to (i), (ii), (iii). 99 */ 100 #define next_csid_chk_cntr_bits(blen, flen) \ 101 ((blen) + (flen) > 128) 102 103 #define next_csid_chk_lcblock_bits(blen) \ 104 ({ \ 105 typeof(blen) __tmp = blen; \ 106 (!__tmp || __tmp > 120 || (__tmp & 0x07)); \ 107 }) 108 109 #define next_csid_chk_lcnode_fn_bits(flen) \ 110 next_csid_chk_lcblock_bits(flen) 111 112 /* flag indicating that flavors are set up for a given End* behavior */ 113 #define SEG6_F_LOCAL_FLAVORS SEG6_F_ATTR(SEG6_LOCAL_FLAVORS) 114 115 #define SEG6_F_LOCAL_FLV_OP(flvname) BIT(SEG6_LOCAL_FLV_OP_##flvname) 116 #define SEG6_F_LOCAL_FLV_NEXT_CSID SEG6_F_LOCAL_FLV_OP(NEXT_CSID) 117 #define SEG6_F_LOCAL_FLV_PSP SEG6_F_LOCAL_FLV_OP(PSP) 118 119 /* Supported RFC8986 Flavor operations are reported in this bitmask */ 120 #define SEG6_LOCAL_FLV8986_SUPP_OPS SEG6_F_LOCAL_FLV_PSP 121 122 #define SEG6_LOCAL_END_FLV_SUPP_OPS (SEG6_F_LOCAL_FLV_NEXT_CSID | \ 123 SEG6_LOCAL_FLV8986_SUPP_OPS) 124 #define SEG6_LOCAL_END_X_FLV_SUPP_OPS SEG6_F_LOCAL_FLV_NEXT_CSID 125 126 struct seg6_flavors_info { 127 /* Flavor operations */ 128 __u32 flv_ops; 129 130 /* Locator-Block length, expressed in bits */ 131 __u8 lcblock_bits; 132 /* Locator-Node Function length, expressed in bits*/ 133 __u8 lcnode_func_bits; 134 }; 135 136 enum seg6_end_dt_mode { 137 DT_INVALID_MODE = -EINVAL, 138 DT_LEGACY_MODE = 0, 139 DT_VRF_MODE = 1, 140 }; 141 142 struct seg6_end_dt_info { 143 enum seg6_end_dt_mode mode; 144 145 struct net *net; 146 /* VRF device associated to the routing table used by the SRv6 147 * End.DT4/DT6 behavior for routing IPv4/IPv6 packets. 148 */ 149 int vrf_ifindex; 150 int vrf_table; 151 152 /* tunneled packet family (IPv4 or IPv6). 153 * Protocol and header length are inferred from family. 154 */ 155 u16 family; 156 }; 157 158 struct pcpu_seg6_local_counters { 159 u64_stats_t packets; 160 u64_stats_t bytes; 161 u64_stats_t errors; 162 163 struct u64_stats_sync syncp; 164 }; 165 166 /* This struct groups all the SRv6 Behavior counters supported so far. 167 * 168 * put_nla_counters() makes use of this data structure to collect all counter 169 * values after the per-CPU counter evaluation has been performed. 170 * Finally, each counter value (in seg6_local_counters) is stored in the 171 * corresponding netlink attribute and sent to user space. 172 * 173 * NB: we don't want to expose this structure to user space! 174 */ 175 struct seg6_local_counters { 176 __u64 packets; 177 __u64 bytes; 178 __u64 errors; 179 }; 180 181 #define seg6_local_alloc_pcpu_counters(__gfp) \ 182 __netdev_alloc_pcpu_stats(struct pcpu_seg6_local_counters, \ 183 ((__gfp) | __GFP_ZERO)) 184 185 #define SEG6_F_LOCAL_COUNTERS SEG6_F_ATTR(SEG6_LOCAL_COUNTERS) 186 187 struct seg6_local_lwt { 188 int action; 189 struct ipv6_sr_hdr *srh; 190 int table; 191 struct in_addr nh4; 192 struct in6_addr nh6; 193 int iif; 194 int oif; 195 struct bpf_lwt_prog bpf; 196 #ifdef CONFIG_NET_L3_MASTER_DEV 197 struct seg6_end_dt_info dt_info; 198 #endif 199 struct seg6_flavors_info flv_info; 200 201 struct pcpu_seg6_local_counters __percpu *pcpu_counters; 202 203 int headroom; 204 struct seg6_action_desc *desc; 205 /* unlike the required attrs, we have to track the optional attributes 206 * that have been effectively parsed. 207 */ 208 unsigned long parsed_optattrs; 209 }; 210 211 static struct seg6_local_lwt *seg6_local_lwtunnel(struct lwtunnel_state *lwt) 212 { 213 return (struct seg6_local_lwt *)lwt->data; 214 } 215 216 static struct ipv6_sr_hdr *get_and_validate_srh(struct sk_buff *skb) 217 { 218 struct ipv6_sr_hdr *srh; 219 220 srh = seg6_get_srh(skb, IP6_FH_F_SKIP_RH); 221 if (!srh) 222 return NULL; 223 224 #ifdef CONFIG_IPV6_SEG6_HMAC 225 if (!seg6_hmac_validate_skb(skb)) 226 return NULL; 227 #endif 228 229 return srh; 230 } 231 232 static bool decap_and_validate(struct sk_buff *skb, int proto) 233 { 234 struct ipv6_sr_hdr *srh; 235 unsigned int off = 0; 236 237 srh = seg6_get_srh(skb, 0); 238 if (srh && srh->segments_left > 0) 239 return false; 240 241 #ifdef CONFIG_IPV6_SEG6_HMAC 242 if (srh && !seg6_hmac_validate_skb(skb)) 243 return false; 244 #endif 245 246 if (ipv6_find_hdr(skb, &off, proto, NULL, NULL) < 0) 247 return false; 248 249 if (!pskb_pull(skb, off)) 250 return false; 251 252 skb_postpull_rcsum(skb, skb_network_header(skb), off); 253 254 skb_reset_network_header(skb); 255 skb_reset_transport_header(skb); 256 if (iptunnel_pull_offloads(skb)) 257 return false; 258 259 return true; 260 } 261 262 static void advance_nextseg(struct ipv6_sr_hdr *srh, struct in6_addr *daddr) 263 { 264 struct in6_addr *addr; 265 266 srh->segments_left--; 267 addr = srh->segments + srh->segments_left; 268 *daddr = *addr; 269 } 270 271 static int 272 seg6_lookup_any_nexthop(struct sk_buff *skb, struct in6_addr *nhaddr, 273 u32 tbl_id, bool local_delivery, int oif) 274 { 275 struct net *net = dev_net(skb->dev); 276 struct ipv6hdr *hdr = ipv6_hdr(skb); 277 int flags = RT6_LOOKUP_F_HAS_SADDR; 278 struct dst_entry *dst = NULL; 279 struct rt6_info *rt; 280 struct flowi6 fl6; 281 int dev_flags = 0; 282 283 memset(&fl6, 0, sizeof(fl6)); 284 fl6.flowi6_iif = skb->dev->ifindex; 285 fl6.flowi6_oif = oif; 286 fl6.daddr = nhaddr ? *nhaddr : hdr->daddr; 287 fl6.saddr = hdr->saddr; 288 fl6.flowlabel = ip6_flowinfo(hdr); 289 fl6.flowi6_mark = skb->mark; 290 fl6.flowi6_proto = hdr->nexthdr; 291 292 if (nhaddr) 293 fl6.flowi6_flags = FLOWI_FLAG_KNOWN_NH; 294 295 if (!tbl_id && !oif) { 296 dst = ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags); 297 } else if (tbl_id) { 298 struct fib6_table *table; 299 300 table = fib6_get_table(net, tbl_id); 301 if (!table) 302 goto out; 303 304 rt = ip6_pol_route(net, table, oif, &fl6, skb, flags); 305 dst = &rt->dst; 306 } else { 307 dst = ip6_route_output(net, NULL, &fl6); 308 } 309 310 /* we want to discard traffic destined for local packet processing, 311 * if @local_delivery is set to false. 312 */ 313 if (!local_delivery) 314 dev_flags |= IFF_LOOPBACK; 315 316 if (dst && (dst_dev(dst)->flags & dev_flags) && !dst->error) { 317 dst_release(dst); 318 dst = NULL; 319 } 320 321 out: 322 if (!dst) { 323 rt = net->ipv6.ip6_blk_hole_entry; 324 dst = &rt->dst; 325 dst_hold(dst); 326 } 327 328 skb_dst_drop(skb); 329 skb_dst_set(skb, dst); 330 return dst->error; 331 } 332 333 int seg6_lookup_nexthop(struct sk_buff *skb, 334 struct in6_addr *nhaddr, u32 tbl_id) 335 { 336 return seg6_lookup_any_nexthop(skb, nhaddr, tbl_id, false, 0); 337 } 338 339 static __u8 seg6_flv_lcblock_octects(const struct seg6_flavors_info *finfo) 340 { 341 return finfo->lcblock_bits >> 3; 342 } 343 344 static __u8 seg6_flv_lcnode_func_octects(const struct seg6_flavors_info *finfo) 345 { 346 return finfo->lcnode_func_bits >> 3; 347 } 348 349 static bool seg6_next_csid_is_arg_zero(const struct in6_addr *addr, 350 const struct seg6_flavors_info *finfo) 351 { 352 __u8 fnc_octects = seg6_flv_lcnode_func_octects(finfo); 353 __u8 blk_octects = seg6_flv_lcblock_octects(finfo); 354 __u8 arg_octects; 355 int i; 356 357 arg_octects = 16 - blk_octects - fnc_octects; 358 for (i = 0; i < arg_octects; ++i) { 359 if (addr->s6_addr[blk_octects + fnc_octects + i] != 0x00) 360 return false; 361 } 362 363 return true; 364 } 365 366 /* assume that DA.Argument length > 0 */ 367 static void seg6_next_csid_advance_arg(struct in6_addr *addr, 368 const struct seg6_flavors_info *finfo) 369 { 370 __u8 fnc_octects = seg6_flv_lcnode_func_octects(finfo); 371 __u8 blk_octects = seg6_flv_lcblock_octects(finfo); 372 373 /* advance DA.Argument */ 374 memmove(&addr->s6_addr[blk_octects], 375 &addr->s6_addr[blk_octects + fnc_octects], 376 16 - blk_octects - fnc_octects); 377 378 memset(&addr->s6_addr[16 - fnc_octects], 0x00, fnc_octects); 379 } 380 381 static int input_action_end_finish(struct sk_buff *skb, 382 struct seg6_local_lwt *slwt) 383 { 384 seg6_lookup_nexthop(skb, NULL, 0); 385 386 return dst_input(skb); 387 } 388 389 static int input_action_end_core(struct sk_buff *skb, 390 struct seg6_local_lwt *slwt) 391 { 392 struct ipv6_sr_hdr *srh; 393 394 srh = get_and_validate_srh(skb); 395 if (!srh) 396 goto drop; 397 398 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 399 400 return input_action_end_finish(skb, slwt); 401 402 drop: 403 kfree_skb(skb); 404 return -EINVAL; 405 } 406 407 static int end_next_csid_core(struct sk_buff *skb, struct seg6_local_lwt *slwt) 408 { 409 const struct seg6_flavors_info *finfo = &slwt->flv_info; 410 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr; 411 412 if (seg6_next_csid_is_arg_zero(daddr, finfo)) 413 return input_action_end_core(skb, slwt); 414 415 /* update DA */ 416 seg6_next_csid_advance_arg(daddr, finfo); 417 418 return input_action_end_finish(skb, slwt); 419 } 420 421 static int input_action_end_x_finish(struct sk_buff *skb, 422 struct seg6_local_lwt *slwt) 423 { 424 seg6_lookup_any_nexthop(skb, &slwt->nh6, 0, false, slwt->oif); 425 426 return dst_input(skb); 427 } 428 429 static int input_action_end_x_core(struct sk_buff *skb, 430 struct seg6_local_lwt *slwt) 431 { 432 struct ipv6_sr_hdr *srh; 433 434 srh = get_and_validate_srh(skb); 435 if (!srh) 436 goto drop; 437 438 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 439 440 return input_action_end_x_finish(skb, slwt); 441 442 drop: 443 kfree_skb(skb); 444 return -EINVAL; 445 } 446 447 static int end_x_next_csid_core(struct sk_buff *skb, 448 struct seg6_local_lwt *slwt) 449 { 450 const struct seg6_flavors_info *finfo = &slwt->flv_info; 451 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr; 452 453 if (seg6_next_csid_is_arg_zero(daddr, finfo)) 454 return input_action_end_x_core(skb, slwt); 455 456 /* update DA */ 457 seg6_next_csid_advance_arg(daddr, finfo); 458 459 return input_action_end_x_finish(skb, slwt); 460 } 461 462 static bool seg6_next_csid_enabled(__u32 fops) 463 { 464 return fops & SEG6_F_LOCAL_FLV_NEXT_CSID; 465 } 466 467 /* Processing of SRv6 End, End.X, and End.T behaviors can be extended through 468 * the flavors framework. These behaviors must report the subset of (flavor) 469 * operations they currently implement. In this way, if a user specifies a 470 * flavor combination that is not supported by a given End* behavior, the 471 * kernel refuses to instantiate the tunnel reporting the error. 472 */ 473 static int seg6_flv_supp_ops_by_action(int action, __u32 *fops) 474 { 475 switch (action) { 476 case SEG6_LOCAL_ACTION_END: 477 *fops = SEG6_LOCAL_END_FLV_SUPP_OPS; 478 break; 479 case SEG6_LOCAL_ACTION_END_X: 480 *fops = SEG6_LOCAL_END_X_FLV_SUPP_OPS; 481 break; 482 default: 483 return -EOPNOTSUPP; 484 } 485 486 return 0; 487 } 488 489 /* We describe the packet state in relation to the absence/presence of the SRH 490 * and the Segment Left (SL) field. 491 * For our purposes, it is not necessary to record the exact value of the SL 492 * when the SID List consists of two or more segments. 493 */ 494 enum seg6_local_pktinfo { 495 /* the order really matters! */ 496 SEG6_LOCAL_PKTINFO_NOHDR = 0, 497 SEG6_LOCAL_PKTINFO_SL_ZERO, 498 SEG6_LOCAL_PKTINFO_SL_ONE, 499 SEG6_LOCAL_PKTINFO_SL_MORE, 500 __SEG6_LOCAL_PKTINFO_MAX, 501 }; 502 503 #define SEG6_LOCAL_PKTINFO_MAX (__SEG6_LOCAL_PKTINFO_MAX - 1) 504 505 static enum seg6_local_pktinfo seg6_get_srh_pktinfo(struct ipv6_sr_hdr *srh) 506 { 507 __u8 sgl; 508 509 if (!srh) 510 return SEG6_LOCAL_PKTINFO_NOHDR; 511 512 sgl = srh->segments_left; 513 if (sgl < 2) 514 return SEG6_LOCAL_PKTINFO_SL_ZERO + sgl; 515 516 return SEG6_LOCAL_PKTINFO_SL_MORE; 517 } 518 519 enum seg6_local_flv_action { 520 SEG6_LOCAL_FLV_ACT_UNSPEC = 0, 521 SEG6_LOCAL_FLV_ACT_END, 522 SEG6_LOCAL_FLV_ACT_PSP, 523 SEG6_LOCAL_FLV_ACT_USP, 524 SEG6_LOCAL_FLV_ACT_USD, 525 __SEG6_LOCAL_FLV_ACT_MAX 526 }; 527 528 #define SEG6_LOCAL_FLV_ACT_MAX (__SEG6_LOCAL_FLV_ACT_MAX - 1) 529 530 /* The action table for RFC8986 flavors (see the flv8986_act_tbl below) 531 * contains the actions (i.e. processing operations) to be applied on packets 532 * when flavors are configured for an End* behavior. 533 * By combining the pkinfo data and from the flavors mask, the macro 534 * computes the index used to access the elements (actions) stored in the 535 * action table. The index is structured as follows: 536 * 537 * index 538 * _______________/\________________ 539 * / \ 540 * +----------------+----------------+ 541 * | pf | afm | 542 * +----------------+----------------+ 543 * ph-1 ... p1 p0 fk-1 ... f1 f0 544 * MSB LSB 545 * 546 * where: 547 * - 'afm' (adjusted flavor mask) is the mask containing a combination of the 548 * RFC8986 flavors currently supported. 'afm' corresponds to the @fm 549 * argument of the macro whose value is righ-shifted by 1 bit. By doing so, 550 * we discard the SEG6_LOCAL_FLV_OP_UNSPEC flag (bit 0 in @fm) which is 551 * never used here; 552 * - 'pf' encodes the packet info (pktinfo) regarding the presence/absence of 553 * the SRH, SL = 0, etc. 'pf' is set with the value of @pf provided as 554 * argument to the macro. 555 */ 556 #define flv8986_act_tbl_idx(pf, fm) \ 557 ((((pf) << bits_per(SEG6_LOCAL_FLV8986_SUPP_OPS)) | \ 558 ((fm) & SEG6_LOCAL_FLV8986_SUPP_OPS)) >> SEG6_LOCAL_FLV_OP_PSP) 559 560 /* We compute the size of the action table by considering the RFC8986 flavors 561 * actually supported by the kernel. In this way, the size is automatically 562 * adjusted when new flavors are supported. 563 */ 564 #define FLV8986_ACT_TBL_SIZE \ 565 roundup_pow_of_two(flv8986_act_tbl_idx(SEG6_LOCAL_PKTINFO_MAX, \ 566 SEG6_LOCAL_FLV8986_SUPP_OPS)) 567 568 /* tbl_cfg(act, pf, fm) macro is used to easily configure the action 569 * table; it accepts 3 arguments: 570 * i) @act, the suffix from SEG6_LOCAL_FLV_ACT_{act} representing 571 * the action that should be applied on the packet; 572 * ii) @pf, the suffix from SEG6_LOCAL_PKTINFO_{pf} reporting the packet 573 * info about the lack/presence of SRH, SRH with SL = 0, etc; 574 * iii) @fm, the mask of flavors. 575 */ 576 #define tbl_cfg(act, pf, fm) \ 577 [flv8986_act_tbl_idx(SEG6_LOCAL_PKTINFO_##pf, \ 578 (fm))] = SEG6_LOCAL_FLV_ACT_##act 579 580 /* shorthand for improving readability */ 581 #define F_PSP SEG6_F_LOCAL_FLV_PSP 582 583 /* The table contains, for each combination of the pktinfo data and 584 * flavors, the action that should be taken on a packet (e.g. 585 * "standard" Endpoint processing, Penultimate Segment Pop, etc). 586 * 587 * By default, table entries not explicitly configured are initialized with the 588 * SEG6_LOCAL_FLV_ACT_UNSPEC action, which generally has the effect of 589 * discarding the processed packet. 590 */ 591 static const u8 flv8986_act_tbl[FLV8986_ACT_TBL_SIZE] = { 592 /* PSP variant for packet where SRH with SL = 1 */ 593 tbl_cfg(PSP, SL_ONE, F_PSP), 594 /* End for packet where the SRH with SL > 1*/ 595 tbl_cfg(END, SL_MORE, F_PSP), 596 }; 597 598 #undef F_PSP 599 #undef tbl_cfg 600 601 /* For each flavor defined in RFC8986 (or a combination of them) an action is 602 * performed on the packet. The specific action depends on: 603 * - info extracted from the packet (i.e. pktinfo data) regarding the 604 * lack/presence of the SRH, and if the SRH is available, on the value of 605 * Segment Left field; 606 * - the mask of flavors configured for the specific SRv6 End* behavior. 607 * 608 * The function combines both the pkinfo and the flavors mask to evaluate the 609 * corresponding action to be taken on the packet. 610 */ 611 static enum seg6_local_flv_action 612 seg6_local_flv8986_act_lookup(enum seg6_local_pktinfo pinfo, __u32 flvmask) 613 { 614 unsigned long index; 615 616 /* check if the provided mask of flavors is supported */ 617 if (unlikely(flvmask & ~SEG6_LOCAL_FLV8986_SUPP_OPS)) 618 return SEG6_LOCAL_FLV_ACT_UNSPEC; 619 620 index = flv8986_act_tbl_idx(pinfo, flvmask); 621 if (unlikely(index >= FLV8986_ACT_TBL_SIZE)) 622 return SEG6_LOCAL_FLV_ACT_UNSPEC; 623 624 return flv8986_act_tbl[index]; 625 } 626 627 /* skb->data must be aligned with skb->network_header */ 628 static bool seg6_pop_srh(struct sk_buff *skb, int srhoff) 629 { 630 struct ipv6_sr_hdr *srh; 631 struct ipv6hdr *iph; 632 __u8 srh_nexthdr; 633 int thoff = -1; 634 int srhlen; 635 int nhlen; 636 637 if (unlikely(srhoff < sizeof(*iph) || 638 !pskb_may_pull(skb, srhoff + sizeof(*srh)))) 639 return false; 640 641 srh = (struct ipv6_sr_hdr *)(skb->data + srhoff); 642 srhlen = ipv6_optlen(srh); 643 644 /* we are about to mangle the pkt, let's check if we can write on it */ 645 if (unlikely(skb_ensure_writable(skb, srhoff + srhlen))) 646 return false; 647 648 /* skb_ensure_writable() may change skb pointers; evaluate srh again */ 649 srh = (struct ipv6_sr_hdr *)(skb->data + srhoff); 650 srh_nexthdr = srh->nexthdr; 651 652 if (unlikely(!skb_transport_header_was_set(skb))) 653 goto pull; 654 655 nhlen = skb_network_header_len(skb); 656 /* we have to deal with the transport header: it could be set before 657 * the SRH, after the SRH, or within it (which is considered wrong, 658 * however). 659 */ 660 if (likely(nhlen <= srhoff)) 661 thoff = nhlen; 662 else if (nhlen >= srhoff + srhlen) 663 /* transport_header is set after the SRH */ 664 thoff = nhlen - srhlen; 665 else 666 /* transport_header falls inside the SRH; hence, we can't 667 * restore the transport_header pointer properly after 668 * SRH removing operation. 669 */ 670 return false; 671 pull: 672 /* we need to pop the SRH: 673 * 1) first of all, we pull out everything from IPv6 header up to SRH 674 * (included) evaluating also the rcsum; 675 * 2) we overwrite (and then remove) the SRH by properly moving the 676 * IPv6 along with any extension header that precedes the SRH; 677 * 3) At the end, we push back the pulled headers (except for SRH, 678 * obviously). 679 */ 680 skb_pull_rcsum(skb, srhoff + srhlen); 681 memmove(skb_network_header(skb) + srhlen, skb_network_header(skb), 682 srhoff); 683 skb_push(skb, srhoff); 684 685 skb_reset_network_header(skb); 686 skb_mac_header_rebuild(skb); 687 if (likely(thoff >= 0)) 688 skb_set_transport_header(skb, thoff); 689 690 iph = ipv6_hdr(skb); 691 if (iph->nexthdr == NEXTHDR_ROUTING) { 692 iph->nexthdr = srh_nexthdr; 693 } else { 694 /* we must look for the extension header (EXTH, for short) that 695 * immediately precedes the SRH we have just removed. 696 * Then, we update the value of the EXTH nexthdr with the one 697 * contained in the SRH nexthdr. 698 */ 699 unsigned int off = sizeof(*iph); 700 struct ipv6_opt_hdr *hp, _hdr; 701 __u8 nexthdr = iph->nexthdr; 702 703 for (;;) { 704 if (unlikely(!ipv6_ext_hdr(nexthdr) || 705 nexthdr == NEXTHDR_NONE)) 706 return false; 707 708 hp = skb_header_pointer(skb, off, sizeof(_hdr), &_hdr); 709 if (unlikely(!hp)) 710 return false; 711 712 if (hp->nexthdr == NEXTHDR_ROUTING) { 713 hp->nexthdr = srh_nexthdr; 714 break; 715 } 716 717 switch (nexthdr) { 718 case NEXTHDR_FRAGMENT: 719 fallthrough; 720 case NEXTHDR_AUTH: 721 /* we expect SRH before FRAG and AUTH */ 722 return false; 723 default: 724 off += ipv6_optlen(hp); 725 break; 726 } 727 728 nexthdr = hp->nexthdr; 729 } 730 } 731 732 iph->payload_len = htons(skb->len - sizeof(struct ipv6hdr)); 733 734 skb_postpush_rcsum(skb, iph, srhoff); 735 736 return true; 737 } 738 739 /* process the packet on the basis of the RFC8986 flavors set for the given 740 * SRv6 End behavior instance. 741 */ 742 static int end_flv8986_core(struct sk_buff *skb, struct seg6_local_lwt *slwt) 743 { 744 const struct seg6_flavors_info *finfo = &slwt->flv_info; 745 enum seg6_local_flv_action action; 746 enum seg6_local_pktinfo pinfo; 747 struct ipv6_sr_hdr *srh; 748 __u32 flvmask; 749 int srhoff; 750 751 srh = seg6_get_srh(skb, 0); 752 srhoff = srh ? ((unsigned char *)srh - skb->data) : 0; 753 pinfo = seg6_get_srh_pktinfo(srh); 754 #ifdef CONFIG_IPV6_SEG6_HMAC 755 if (srh && !seg6_hmac_validate_skb(skb)) 756 goto drop; 757 #endif 758 flvmask = finfo->flv_ops; 759 if (unlikely(flvmask & ~SEG6_LOCAL_FLV8986_SUPP_OPS)) { 760 pr_warn_once("seg6local: invalid RFC8986 flavors\n"); 761 goto drop; 762 } 763 764 /* retrieve the action triggered by the combination of pktinfo data and 765 * the flavors mask. 766 */ 767 action = seg6_local_flv8986_act_lookup(pinfo, flvmask); 768 switch (action) { 769 case SEG6_LOCAL_FLV_ACT_END: 770 /* process the packet as the "standard" End behavior */ 771 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 772 break; 773 case SEG6_LOCAL_FLV_ACT_PSP: 774 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 775 776 if (unlikely(!seg6_pop_srh(skb, srhoff))) 777 goto drop; 778 break; 779 case SEG6_LOCAL_FLV_ACT_UNSPEC: 780 fallthrough; 781 default: 782 /* by default, we drop the packet since we could not find a 783 * suitable action. 784 */ 785 goto drop; 786 } 787 788 return input_action_end_finish(skb, slwt); 789 790 drop: 791 kfree_skb(skb); 792 return -EINVAL; 793 } 794 795 /* regular endpoint function */ 796 static int input_action_end(struct sk_buff *skb, struct seg6_local_lwt *slwt) 797 { 798 const struct seg6_flavors_info *finfo = &slwt->flv_info; 799 __u32 fops = finfo->flv_ops; 800 801 if (!fops) 802 return input_action_end_core(skb, slwt); 803 804 /* check for the presence of NEXT-C-SID since it applies first */ 805 if (seg6_next_csid_enabled(fops)) 806 return end_next_csid_core(skb, slwt); 807 808 /* the specific processing function to be performed on the packet 809 * depends on the combination of flavors defined in RFC8986 and some 810 * information extracted from the packet, e.g. presence/absence of SRH, 811 * Segment Left = 0, etc. 812 */ 813 return end_flv8986_core(skb, slwt); 814 } 815 816 /* regular endpoint, and forward to specified nexthop */ 817 static int input_action_end_x(struct sk_buff *skb, struct seg6_local_lwt *slwt) 818 { 819 const struct seg6_flavors_info *finfo = &slwt->flv_info; 820 __u32 fops = finfo->flv_ops; 821 822 /* check for the presence of NEXT-C-SID since it applies first */ 823 if (seg6_next_csid_enabled(fops)) 824 return end_x_next_csid_core(skb, slwt); 825 826 return input_action_end_x_core(skb, slwt); 827 } 828 829 static int input_action_end_t(struct sk_buff *skb, struct seg6_local_lwt *slwt) 830 { 831 struct ipv6_sr_hdr *srh; 832 833 srh = get_and_validate_srh(skb); 834 if (!srh) 835 goto drop; 836 837 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 838 839 seg6_lookup_nexthop(skb, NULL, slwt->table); 840 841 return dst_input(skb); 842 843 drop: 844 kfree_skb(skb); 845 return -EINVAL; 846 } 847 848 /* decapsulate and forward inner L2 frame on specified interface */ 849 static int input_action_end_dx2(struct sk_buff *skb, 850 struct seg6_local_lwt *slwt) 851 { 852 struct net *net = dev_net(skb->dev); 853 struct net_device *odev; 854 struct ethhdr *eth; 855 856 if (!decap_and_validate(skb, IPPROTO_ETHERNET)) 857 goto drop; 858 859 if (!pskb_may_pull(skb, ETH_HLEN)) 860 goto drop; 861 862 skb_reset_mac_header(skb); 863 eth = (struct ethhdr *)skb->data; 864 865 /* To determine the frame's protocol, we assume it is 802.3. This avoids 866 * a call to eth_type_trans(), which is not really relevant for our 867 * use case. 868 */ 869 if (!eth_proto_is_802_3(eth->h_proto)) 870 goto drop; 871 872 odev = dev_get_by_index_rcu(net, slwt->oif); 873 if (!odev) 874 goto drop; 875 876 /* As we accept Ethernet frames, make sure the egress device is of 877 * the correct type. 878 */ 879 if (odev->type != ARPHRD_ETHER) 880 goto drop; 881 882 if (!(odev->flags & IFF_UP) || !netif_carrier_ok(odev)) 883 goto drop; 884 885 skb_orphan(skb); 886 887 if (skb_warn_if_lro(skb)) 888 goto drop; 889 890 skb_forward_csum(skb); 891 892 if (skb->len - ETH_HLEN > odev->mtu) 893 goto drop; 894 895 skb->dev = odev; 896 skb->protocol = eth->h_proto; 897 898 return dev_queue_xmit(skb); 899 900 drop: 901 kfree_skb(skb); 902 return -EINVAL; 903 } 904 905 static int input_action_end_dx6_finish(struct net *net, struct sock *sk, 906 struct sk_buff *skb) 907 { 908 struct dst_entry *orig_dst = skb_dst(skb); 909 struct in6_addr *nhaddr = NULL; 910 struct seg6_local_lwt *slwt; 911 912 slwt = seg6_local_lwtunnel(orig_dst->lwtstate); 913 914 /* The inner packet is not associated to any local interface, 915 * so we do not call netif_rx(). 916 * 917 * If slwt->nh6 is set to ::, then lookup the nexthop for the 918 * inner packet's DA. Otherwise, use the specified nexthop. 919 */ 920 if (!ipv6_addr_any(&slwt->nh6)) 921 nhaddr = &slwt->nh6; 922 923 seg6_lookup_nexthop(skb, nhaddr, 0); 924 925 return dst_input(skb); 926 } 927 928 /* decapsulate and forward to specified nexthop */ 929 static int input_action_end_dx6(struct sk_buff *skb, 930 struct seg6_local_lwt *slwt) 931 { 932 /* this function accepts IPv6 encapsulated packets, with either 933 * an SRH with SL=0, or no SRH. 934 */ 935 936 if (!decap_and_validate(skb, IPPROTO_IPV6)) 937 goto drop; 938 939 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 940 goto drop; 941 942 skb_set_transport_header(skb, sizeof(struct ipv6hdr)); 943 nf_reset_ct(skb); 944 945 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled)) 946 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, 947 dev_net(skb->dev), NULL, skb, skb->dev, 948 NULL, input_action_end_dx6_finish); 949 950 return input_action_end_dx6_finish(dev_net(skb->dev), NULL, skb); 951 drop: 952 kfree_skb(skb); 953 return -EINVAL; 954 } 955 956 static int input_action_end_dx4_finish(struct net *net, struct sock *sk, 957 struct sk_buff *skb) 958 { 959 struct dst_entry *orig_dst = skb_dst(skb); 960 enum skb_drop_reason reason; 961 struct seg6_local_lwt *slwt; 962 struct iphdr *iph; 963 __be32 nhaddr; 964 965 slwt = seg6_local_lwtunnel(orig_dst->lwtstate); 966 967 iph = ip_hdr(skb); 968 969 nhaddr = slwt->nh4.s_addr ?: iph->daddr; 970 971 skb_dst_drop(skb); 972 973 reason = ip_route_input(skb, nhaddr, iph->saddr, 0, skb->dev); 974 if (reason) { 975 kfree_skb_reason(skb, reason); 976 return -EINVAL; 977 } 978 979 return dst_input(skb); 980 } 981 982 static int input_action_end_dx4(struct sk_buff *skb, 983 struct seg6_local_lwt *slwt) 984 { 985 if (!decap_and_validate(skb, IPPROTO_IPIP)) 986 goto drop; 987 988 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 989 goto drop; 990 991 skb->protocol = htons(ETH_P_IP); 992 skb_set_transport_header(skb, sizeof(struct iphdr)); 993 nf_reset_ct(skb); 994 995 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled)) 996 return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, 997 dev_net(skb->dev), NULL, skb, skb->dev, 998 NULL, input_action_end_dx4_finish); 999 1000 return input_action_end_dx4_finish(dev_net(skb->dev), NULL, skb); 1001 drop: 1002 kfree_skb(skb); 1003 return -EINVAL; 1004 } 1005 1006 #ifdef CONFIG_NET_L3_MASTER_DEV 1007 static struct net *fib6_config_get_net(const struct fib6_config *fib6_cfg) 1008 { 1009 const struct nl_info *nli = &fib6_cfg->fc_nlinfo; 1010 1011 return nli->nl_net; 1012 } 1013 1014 static int __seg6_end_dt_vrf_build(struct seg6_local_lwt *slwt, const void *cfg, 1015 u16 family, struct netlink_ext_ack *extack) 1016 { 1017 struct seg6_end_dt_info *info = &slwt->dt_info; 1018 int vrf_ifindex; 1019 struct net *net; 1020 1021 net = fib6_config_get_net(cfg); 1022 1023 /* note that vrf_table was already set by parse_nla_vrftable() */ 1024 vrf_ifindex = l3mdev_ifindex_lookup_by_table_id(L3MDEV_TYPE_VRF, net, 1025 info->vrf_table); 1026 if (vrf_ifindex < 0) { 1027 if (vrf_ifindex == -EPERM) { 1028 NL_SET_ERR_MSG(extack, 1029 "Strict mode for VRF is disabled"); 1030 } else if (vrf_ifindex == -ENODEV) { 1031 NL_SET_ERR_MSG(extack, 1032 "Table has no associated VRF device"); 1033 } else { 1034 pr_debug("seg6local: SRv6 End.DT* creation error=%d\n", 1035 vrf_ifindex); 1036 } 1037 1038 return vrf_ifindex; 1039 } 1040 1041 info->net = net; 1042 info->vrf_ifindex = vrf_ifindex; 1043 1044 info->family = family; 1045 info->mode = DT_VRF_MODE; 1046 1047 return 0; 1048 } 1049 1050 /* The SRv6 End.DT4/DT6 behavior extracts the inner (IPv4/IPv6) packet and 1051 * routes the IPv4/IPv6 packet by looking at the configured routing table. 1052 * 1053 * In the SRv6 End.DT4/DT6 use case, we can receive traffic (IPv6+Segment 1054 * Routing Header packets) from several interfaces and the outer IPv6 1055 * destination address (DA) is used for retrieving the specific instance of the 1056 * End.DT4/DT6 behavior that should process the packets. 1057 * 1058 * However, the inner IPv4/IPv6 packet is not really bound to any receiving 1059 * interface and thus the End.DT4/DT6 sets the VRF (associated with the 1060 * corresponding routing table) as the *receiving* interface. 1061 * In other words, the End.DT4/DT6 processes a packet as if it has been received 1062 * directly by the VRF (and not by one of its slave devices, if any). 1063 * In this way, the VRF interface is used for routing the IPv4/IPv6 packet in 1064 * according to the routing table configured by the End.DT4/DT6 instance. 1065 * 1066 * This design allows you to get some interesting features like: 1067 * 1) the statistics on rx packets; 1068 * 2) the possibility to install a packet sniffer on the receiving interface 1069 * (the VRF one) for looking at the incoming packets; 1070 * 3) the possibility to leverage the netfilter prerouting hook for the inner 1071 * IPv4 packet. 1072 * 1073 * This function returns: 1074 * - the sk_buff* when the VRF rcv handler has processed the packet correctly; 1075 * - NULL when the skb is consumed by the VRF rcv handler; 1076 * - a pointer which encodes a negative error number in case of error. 1077 * Note that in this case, the function takes care of freeing the skb. 1078 */ 1079 static struct sk_buff *end_dt_vrf_rcv(struct sk_buff *skb, u16 family, 1080 struct net_device *dev) 1081 { 1082 /* based on l3mdev_ip_rcv; we are only interested in the master */ 1083 if (unlikely(!netif_is_l3_master(dev) && !netif_has_l3_rx_handler(dev))) 1084 goto drop; 1085 1086 if (unlikely(!dev->l3mdev_ops->l3mdev_l3_rcv)) 1087 goto drop; 1088 1089 /* the decap packet IPv4/IPv6 does not come with any mac header info. 1090 * We must unset the mac header to allow the VRF device to rebuild it, 1091 * just in case there is a sniffer attached on the device. 1092 */ 1093 skb_unset_mac_header(skb); 1094 1095 skb = dev->l3mdev_ops->l3mdev_l3_rcv(dev, skb, family); 1096 if (!skb) 1097 /* the skb buffer was consumed by the handler */ 1098 return NULL; 1099 1100 /* when a packet is received by a VRF or by one of its slaves, the 1101 * master device reference is set into the skb. 1102 */ 1103 if (unlikely(skb->dev != dev || skb->skb_iif != dev->ifindex)) 1104 goto drop; 1105 1106 return skb; 1107 1108 drop: 1109 kfree_skb(skb); 1110 return ERR_PTR(-EINVAL); 1111 } 1112 1113 static struct net_device *end_dt_get_vrf_rcu(struct sk_buff *skb, 1114 struct seg6_end_dt_info *info) 1115 { 1116 int vrf_ifindex = info->vrf_ifindex; 1117 struct net *net = info->net; 1118 1119 if (unlikely(vrf_ifindex < 0)) 1120 goto error; 1121 1122 if (unlikely(!net_eq(dev_net(skb->dev), net))) 1123 goto error; 1124 1125 return dev_get_by_index_rcu(net, vrf_ifindex); 1126 1127 error: 1128 return NULL; 1129 } 1130 1131 static struct sk_buff *end_dt_vrf_core(struct sk_buff *skb, 1132 struct seg6_local_lwt *slwt, u16 family) 1133 { 1134 struct seg6_end_dt_info *info = &slwt->dt_info; 1135 struct net_device *vrf; 1136 __be16 protocol; 1137 int hdrlen; 1138 1139 vrf = end_dt_get_vrf_rcu(skb, info); 1140 if (unlikely(!vrf)) 1141 goto drop; 1142 1143 switch (family) { 1144 case AF_INET: 1145 protocol = htons(ETH_P_IP); 1146 hdrlen = sizeof(struct iphdr); 1147 break; 1148 case AF_INET6: 1149 protocol = htons(ETH_P_IPV6); 1150 hdrlen = sizeof(struct ipv6hdr); 1151 break; 1152 case AF_UNSPEC: 1153 fallthrough; 1154 default: 1155 goto drop; 1156 } 1157 1158 if (unlikely(info->family != AF_UNSPEC && info->family != family)) { 1159 pr_warn_once("seg6local: SRv6 End.DT* family mismatch"); 1160 goto drop; 1161 } 1162 1163 skb->protocol = protocol; 1164 1165 skb_dst_drop(skb); 1166 1167 skb_set_transport_header(skb, hdrlen); 1168 nf_reset_ct(skb); 1169 1170 return end_dt_vrf_rcv(skb, family, vrf); 1171 1172 drop: 1173 kfree_skb(skb); 1174 return ERR_PTR(-EINVAL); 1175 } 1176 1177 static int input_action_end_dt4(struct sk_buff *skb, 1178 struct seg6_local_lwt *slwt) 1179 { 1180 enum skb_drop_reason reason; 1181 struct iphdr *iph; 1182 1183 if (!decap_and_validate(skb, IPPROTO_IPIP)) 1184 goto drop; 1185 1186 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 1187 goto drop; 1188 1189 skb = end_dt_vrf_core(skb, slwt, AF_INET); 1190 if (!skb) 1191 /* packet has been processed and consumed by the VRF */ 1192 return 0; 1193 1194 if (IS_ERR(skb)) 1195 return PTR_ERR(skb); 1196 1197 iph = ip_hdr(skb); 1198 1199 reason = ip_route_input(skb, iph->daddr, iph->saddr, 0, skb->dev); 1200 if (unlikely(reason)) 1201 goto drop; 1202 1203 return dst_input(skb); 1204 1205 drop: 1206 kfree_skb(skb); 1207 return -EINVAL; 1208 } 1209 1210 static int seg6_end_dt4_build(struct seg6_local_lwt *slwt, const void *cfg, 1211 struct netlink_ext_ack *extack) 1212 { 1213 return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET, extack); 1214 } 1215 1216 static enum 1217 seg6_end_dt_mode seg6_end_dt6_parse_mode(struct seg6_local_lwt *slwt) 1218 { 1219 unsigned long parsed_optattrs = slwt->parsed_optattrs; 1220 bool legacy, vrfmode; 1221 1222 legacy = !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE)); 1223 vrfmode = !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE)); 1224 1225 if (!(legacy ^ vrfmode)) 1226 /* both are absent or present: invalid DT6 mode */ 1227 return DT_INVALID_MODE; 1228 1229 return legacy ? DT_LEGACY_MODE : DT_VRF_MODE; 1230 } 1231 1232 static enum seg6_end_dt_mode seg6_end_dt6_get_mode(struct seg6_local_lwt *slwt) 1233 { 1234 struct seg6_end_dt_info *info = &slwt->dt_info; 1235 1236 return info->mode; 1237 } 1238 1239 static int seg6_end_dt6_build(struct seg6_local_lwt *slwt, const void *cfg, 1240 struct netlink_ext_ack *extack) 1241 { 1242 enum seg6_end_dt_mode mode = seg6_end_dt6_parse_mode(slwt); 1243 struct seg6_end_dt_info *info = &slwt->dt_info; 1244 1245 switch (mode) { 1246 case DT_LEGACY_MODE: 1247 info->mode = DT_LEGACY_MODE; 1248 return 0; 1249 case DT_VRF_MODE: 1250 return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET6, extack); 1251 default: 1252 NL_SET_ERR_MSG(extack, "table or vrftable must be specified"); 1253 return -EINVAL; 1254 } 1255 } 1256 #endif 1257 1258 static int input_action_end_dt6(struct sk_buff *skb, 1259 struct seg6_local_lwt *slwt) 1260 { 1261 if (!decap_and_validate(skb, IPPROTO_IPV6)) 1262 goto drop; 1263 1264 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 1265 goto drop; 1266 1267 #ifdef CONFIG_NET_L3_MASTER_DEV 1268 if (seg6_end_dt6_get_mode(slwt) == DT_LEGACY_MODE) 1269 goto legacy_mode; 1270 1271 /* DT6_VRF_MODE */ 1272 skb = end_dt_vrf_core(skb, slwt, AF_INET6); 1273 if (!skb) 1274 /* packet has been processed and consumed by the VRF */ 1275 return 0; 1276 1277 if (IS_ERR(skb)) 1278 return PTR_ERR(skb); 1279 1280 /* note: this time we do not need to specify the table because the VRF 1281 * takes care of selecting the correct table. 1282 */ 1283 seg6_lookup_any_nexthop(skb, NULL, 0, true, 0); 1284 1285 return dst_input(skb); 1286 1287 legacy_mode: 1288 #endif 1289 skb_set_transport_header(skb, sizeof(struct ipv6hdr)); 1290 1291 seg6_lookup_any_nexthop(skb, NULL, slwt->table, true, 0); 1292 1293 return dst_input(skb); 1294 1295 drop: 1296 kfree_skb(skb); 1297 return -EINVAL; 1298 } 1299 1300 #ifdef CONFIG_NET_L3_MASTER_DEV 1301 static int seg6_end_dt46_build(struct seg6_local_lwt *slwt, const void *cfg, 1302 struct netlink_ext_ack *extack) 1303 { 1304 return __seg6_end_dt_vrf_build(slwt, cfg, AF_UNSPEC, extack); 1305 } 1306 1307 static int input_action_end_dt46(struct sk_buff *skb, 1308 struct seg6_local_lwt *slwt) 1309 { 1310 unsigned int off = 0; 1311 int nexthdr; 1312 1313 nexthdr = ipv6_find_hdr(skb, &off, -1, NULL, NULL); 1314 if (unlikely(nexthdr < 0)) 1315 goto drop; 1316 1317 switch (nexthdr) { 1318 case IPPROTO_IPIP: 1319 return input_action_end_dt4(skb, slwt); 1320 case IPPROTO_IPV6: 1321 return input_action_end_dt6(skb, slwt); 1322 } 1323 1324 drop: 1325 kfree_skb(skb); 1326 return -EINVAL; 1327 } 1328 #endif 1329 1330 /* push an SRH on top of the current one */ 1331 static int input_action_end_b6(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1332 { 1333 struct ipv6_sr_hdr *srh; 1334 int err = -EINVAL; 1335 1336 srh = get_and_validate_srh(skb); 1337 if (!srh) 1338 goto drop; 1339 1340 err = seg6_do_srh_inline(skb, slwt->srh); 1341 if (err) 1342 goto drop; 1343 1344 skb_set_transport_header(skb, sizeof(struct ipv6hdr)); 1345 1346 seg6_lookup_nexthop(skb, NULL, 0); 1347 1348 return dst_input(skb); 1349 1350 drop: 1351 kfree_skb(skb); 1352 return err; 1353 } 1354 1355 /* encapsulate within an outer IPv6 header and a specified SRH */ 1356 static int input_action_end_b6_encap(struct sk_buff *skb, 1357 struct seg6_local_lwt *slwt) 1358 { 1359 struct ipv6_sr_hdr *srh; 1360 int err = -EINVAL; 1361 1362 srh = get_and_validate_srh(skb); 1363 if (!srh) 1364 goto drop; 1365 1366 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 1367 1368 skb_reset_inner_headers(skb); 1369 skb->encapsulation = 1; 1370 1371 err = seg6_do_srh_encap(skb, slwt->srh, IPPROTO_IPV6); 1372 if (err) 1373 goto drop; 1374 1375 skb_set_transport_header(skb, sizeof(struct ipv6hdr)); 1376 1377 seg6_lookup_nexthop(skb, NULL, 0); 1378 1379 return dst_input(skb); 1380 1381 drop: 1382 kfree_skb(skb); 1383 return err; 1384 } 1385 1386 DEFINE_PER_CPU(struct seg6_bpf_srh_state, seg6_bpf_srh_states) = { 1387 .bh_lock = INIT_LOCAL_LOCK(bh_lock), 1388 }; 1389 1390 bool seg6_bpf_has_valid_srh(struct sk_buff *skb) 1391 { 1392 struct seg6_bpf_srh_state *srh_state = 1393 this_cpu_ptr(&seg6_bpf_srh_states); 1394 struct ipv6_sr_hdr *srh = srh_state->srh; 1395 1396 lockdep_assert_held(&srh_state->bh_lock); 1397 if (unlikely(srh == NULL)) 1398 return false; 1399 1400 if (unlikely(!srh_state->valid)) { 1401 if ((srh_state->hdrlen & 7) != 0) 1402 return false; 1403 1404 srh->hdrlen = (u8)(srh_state->hdrlen >> 3); 1405 if (!seg6_validate_srh(srh, (srh->hdrlen + 1) << 3, true)) 1406 return false; 1407 1408 srh_state->valid = true; 1409 } 1410 1411 return true; 1412 } 1413 1414 static int input_action_end_bpf(struct sk_buff *skb, 1415 struct seg6_local_lwt *slwt) 1416 { 1417 struct seg6_bpf_srh_state *srh_state; 1418 struct ipv6_sr_hdr *srh; 1419 int ret; 1420 1421 srh = get_and_validate_srh(skb); 1422 if (!srh) { 1423 kfree_skb(skb); 1424 return -EINVAL; 1425 } 1426 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 1427 1428 /* The access to the per-CPU buffer srh_state is protected by running 1429 * always in softirq context (with disabled BH). On PREEMPT_RT the 1430 * required locking is provided by the following local_lock_nested_bh() 1431 * statement. It is also accessed by the bpf_lwt_seg6_* helpers via 1432 * bpf_prog_run_save_cb(). 1433 */ 1434 local_lock_nested_bh(&seg6_bpf_srh_states.bh_lock); 1435 srh_state = this_cpu_ptr(&seg6_bpf_srh_states); 1436 srh_state->srh = srh; 1437 srh_state->hdrlen = srh->hdrlen << 3; 1438 srh_state->valid = true; 1439 1440 rcu_read_lock(); 1441 bpf_compute_data_pointers(skb); 1442 ret = bpf_prog_run_save_cb(slwt->bpf.prog, skb); 1443 rcu_read_unlock(); 1444 1445 switch (ret) { 1446 case BPF_OK: 1447 case BPF_REDIRECT: 1448 break; 1449 case BPF_DROP: 1450 goto drop; 1451 default: 1452 pr_warn_once("bpf-seg6local: Illegal return value %u\n", ret); 1453 goto drop; 1454 } 1455 1456 if (srh_state->srh && !seg6_bpf_has_valid_srh(skb)) 1457 goto drop; 1458 local_unlock_nested_bh(&seg6_bpf_srh_states.bh_lock); 1459 1460 if (ret != BPF_REDIRECT) 1461 seg6_lookup_nexthop(skb, NULL, 0); 1462 1463 return dst_input(skb); 1464 1465 drop: 1466 local_unlock_nested_bh(&seg6_bpf_srh_states.bh_lock); 1467 kfree_skb(skb); 1468 return -EINVAL; 1469 } 1470 1471 static struct seg6_action_desc seg6_action_table[] = { 1472 { 1473 .action = SEG6_LOCAL_ACTION_END, 1474 .attrs = 0, 1475 .optattrs = SEG6_F_LOCAL_COUNTERS | 1476 SEG6_F_LOCAL_FLAVORS, 1477 .input = input_action_end, 1478 }, 1479 { 1480 .action = SEG6_LOCAL_ACTION_END_X, 1481 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH6), 1482 .optattrs = SEG6_F_LOCAL_COUNTERS | 1483 SEG6_F_LOCAL_FLAVORS | 1484 SEG6_F_ATTR(SEG6_LOCAL_OIF), 1485 .input = input_action_end_x, 1486 }, 1487 { 1488 .action = SEG6_LOCAL_ACTION_END_T, 1489 .attrs = SEG6_F_ATTR(SEG6_LOCAL_TABLE), 1490 .optattrs = SEG6_F_LOCAL_COUNTERS, 1491 .input = input_action_end_t, 1492 }, 1493 { 1494 .action = SEG6_LOCAL_ACTION_END_DX2, 1495 .attrs = SEG6_F_ATTR(SEG6_LOCAL_OIF), 1496 .optattrs = SEG6_F_LOCAL_COUNTERS, 1497 .input = input_action_end_dx2, 1498 }, 1499 { 1500 .action = SEG6_LOCAL_ACTION_END_DX6, 1501 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH6), 1502 .optattrs = SEG6_F_LOCAL_COUNTERS, 1503 .input = input_action_end_dx6, 1504 }, 1505 { 1506 .action = SEG6_LOCAL_ACTION_END_DX4, 1507 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH4), 1508 .optattrs = SEG6_F_LOCAL_COUNTERS, 1509 .input = input_action_end_dx4, 1510 }, 1511 { 1512 .action = SEG6_LOCAL_ACTION_END_DT4, 1513 .attrs = SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE), 1514 .optattrs = SEG6_F_LOCAL_COUNTERS, 1515 #ifdef CONFIG_NET_L3_MASTER_DEV 1516 .input = input_action_end_dt4, 1517 .slwt_ops = { 1518 .build_state = seg6_end_dt4_build, 1519 }, 1520 #endif 1521 }, 1522 { 1523 .action = SEG6_LOCAL_ACTION_END_DT6, 1524 #ifdef CONFIG_NET_L3_MASTER_DEV 1525 .attrs = 0, 1526 .optattrs = SEG6_F_LOCAL_COUNTERS | 1527 SEG6_F_ATTR(SEG6_LOCAL_TABLE) | 1528 SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE), 1529 .slwt_ops = { 1530 .build_state = seg6_end_dt6_build, 1531 }, 1532 #else 1533 .attrs = SEG6_F_ATTR(SEG6_LOCAL_TABLE), 1534 .optattrs = SEG6_F_LOCAL_COUNTERS, 1535 #endif 1536 .input = input_action_end_dt6, 1537 }, 1538 { 1539 .action = SEG6_LOCAL_ACTION_END_DT46, 1540 .attrs = SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE), 1541 .optattrs = SEG6_F_LOCAL_COUNTERS, 1542 #ifdef CONFIG_NET_L3_MASTER_DEV 1543 .input = input_action_end_dt46, 1544 .slwt_ops = { 1545 .build_state = seg6_end_dt46_build, 1546 }, 1547 #endif 1548 }, 1549 { 1550 .action = SEG6_LOCAL_ACTION_END_B6, 1551 .attrs = SEG6_F_ATTR(SEG6_LOCAL_SRH), 1552 .optattrs = SEG6_F_LOCAL_COUNTERS, 1553 .input = input_action_end_b6, 1554 }, 1555 { 1556 .action = SEG6_LOCAL_ACTION_END_B6_ENCAP, 1557 .attrs = SEG6_F_ATTR(SEG6_LOCAL_SRH), 1558 .optattrs = SEG6_F_LOCAL_COUNTERS, 1559 .input = input_action_end_b6_encap, 1560 .static_headroom = sizeof(struct ipv6hdr), 1561 }, 1562 { 1563 .action = SEG6_LOCAL_ACTION_END_BPF, 1564 .attrs = SEG6_F_ATTR(SEG6_LOCAL_BPF), 1565 .optattrs = SEG6_F_LOCAL_COUNTERS, 1566 .input = input_action_end_bpf, 1567 }, 1568 1569 }; 1570 1571 static struct seg6_action_desc *__get_action_desc(int action) 1572 { 1573 struct seg6_action_desc *desc; 1574 int i, count; 1575 1576 count = ARRAY_SIZE(seg6_action_table); 1577 for (i = 0; i < count; i++) { 1578 desc = &seg6_action_table[i]; 1579 if (desc->action == action) 1580 return desc; 1581 } 1582 1583 return NULL; 1584 } 1585 1586 static bool seg6_lwtunnel_counters_enabled(struct seg6_local_lwt *slwt) 1587 { 1588 return slwt->parsed_optattrs & SEG6_F_LOCAL_COUNTERS; 1589 } 1590 1591 static void seg6_local_update_counters(struct seg6_local_lwt *slwt, 1592 unsigned int len, int err) 1593 { 1594 struct pcpu_seg6_local_counters *pcounters; 1595 1596 pcounters = this_cpu_ptr(slwt->pcpu_counters); 1597 u64_stats_update_begin(&pcounters->syncp); 1598 1599 if (likely(!err)) { 1600 u64_stats_inc(&pcounters->packets); 1601 u64_stats_add(&pcounters->bytes, len); 1602 } else { 1603 u64_stats_inc(&pcounters->errors); 1604 } 1605 1606 u64_stats_update_end(&pcounters->syncp); 1607 } 1608 1609 static int seg6_local_input_core(struct net *net, struct sock *sk, 1610 struct sk_buff *skb) 1611 { 1612 struct dst_entry *orig_dst = skb_dst(skb); 1613 struct seg6_action_desc *desc; 1614 struct seg6_local_lwt *slwt; 1615 unsigned int len = skb->len; 1616 int rc; 1617 1618 slwt = seg6_local_lwtunnel(orig_dst->lwtstate); 1619 desc = slwt->desc; 1620 1621 rc = desc->input(skb, slwt); 1622 1623 if (!seg6_lwtunnel_counters_enabled(slwt)) 1624 return rc; 1625 1626 seg6_local_update_counters(slwt, len, rc); 1627 1628 return rc; 1629 } 1630 1631 static int seg6_local_input(struct sk_buff *skb) 1632 { 1633 if (skb->protocol != htons(ETH_P_IPV6)) { 1634 kfree_skb(skb); 1635 return -EINVAL; 1636 } 1637 1638 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled)) 1639 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN, 1640 dev_net(skb->dev), NULL, skb, skb->dev, NULL, 1641 seg6_local_input_core); 1642 1643 return seg6_local_input_core(dev_net(skb->dev), NULL, skb); 1644 } 1645 1646 static const struct nla_policy seg6_local_policy[SEG6_LOCAL_MAX + 1] = { 1647 [SEG6_LOCAL_ACTION] = { .type = NLA_U32 }, 1648 [SEG6_LOCAL_SRH] = { .type = NLA_BINARY }, 1649 [SEG6_LOCAL_TABLE] = { .type = NLA_U32 }, 1650 [SEG6_LOCAL_VRFTABLE] = { .type = NLA_U32 }, 1651 [SEG6_LOCAL_NH4] = NLA_POLICY_EXACT_LEN(sizeof(struct in_addr)), 1652 [SEG6_LOCAL_NH6] = NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)), 1653 [SEG6_LOCAL_IIF] = { .type = NLA_U32 }, 1654 [SEG6_LOCAL_OIF] = { .type = NLA_U32 }, 1655 [SEG6_LOCAL_BPF] = { .type = NLA_NESTED }, 1656 [SEG6_LOCAL_COUNTERS] = { .type = NLA_NESTED }, 1657 [SEG6_LOCAL_FLAVORS] = { .type = NLA_NESTED }, 1658 }; 1659 1660 static int parse_nla_srh(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1661 struct netlink_ext_ack *extack) 1662 { 1663 struct ipv6_sr_hdr *srh; 1664 int len; 1665 1666 srh = nla_data(attrs[SEG6_LOCAL_SRH]); 1667 len = nla_len(attrs[SEG6_LOCAL_SRH]); 1668 1669 /* SRH must contain at least one segment */ 1670 if (len < sizeof(*srh) + sizeof(struct in6_addr)) 1671 return -EINVAL; 1672 1673 if (!seg6_validate_srh(srh, len, false)) 1674 return -EINVAL; 1675 1676 slwt->srh = kmemdup(srh, len, GFP_KERNEL); 1677 if (!slwt->srh) 1678 return -ENOMEM; 1679 1680 slwt->headroom += len; 1681 1682 return 0; 1683 } 1684 1685 static int put_nla_srh(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1686 { 1687 struct ipv6_sr_hdr *srh; 1688 struct nlattr *nla; 1689 int len; 1690 1691 srh = slwt->srh; 1692 len = (srh->hdrlen + 1) << 3; 1693 1694 nla = nla_reserve(skb, SEG6_LOCAL_SRH, len); 1695 if (!nla) 1696 return -EMSGSIZE; 1697 1698 memcpy(nla_data(nla), srh, len); 1699 1700 return 0; 1701 } 1702 1703 static int cmp_nla_srh(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1704 { 1705 int len = (a->srh->hdrlen + 1) << 3; 1706 1707 if (len != ((b->srh->hdrlen + 1) << 3)) 1708 return 1; 1709 1710 return memcmp(a->srh, b->srh, len); 1711 } 1712 1713 static void destroy_attr_srh(struct seg6_local_lwt *slwt) 1714 { 1715 kfree(slwt->srh); 1716 } 1717 1718 static int parse_nla_table(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1719 struct netlink_ext_ack *extack) 1720 { 1721 slwt->table = nla_get_u32(attrs[SEG6_LOCAL_TABLE]); 1722 1723 return 0; 1724 } 1725 1726 static int put_nla_table(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1727 { 1728 if (nla_put_u32(skb, SEG6_LOCAL_TABLE, slwt->table)) 1729 return -EMSGSIZE; 1730 1731 return 0; 1732 } 1733 1734 static int cmp_nla_table(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1735 { 1736 if (a->table != b->table) 1737 return 1; 1738 1739 return 0; 1740 } 1741 1742 static struct 1743 seg6_end_dt_info *seg6_possible_end_dt_info(struct seg6_local_lwt *slwt) 1744 { 1745 #ifdef CONFIG_NET_L3_MASTER_DEV 1746 return &slwt->dt_info; 1747 #else 1748 return ERR_PTR(-EOPNOTSUPP); 1749 #endif 1750 } 1751 1752 static int parse_nla_vrftable(struct nlattr **attrs, 1753 struct seg6_local_lwt *slwt, 1754 struct netlink_ext_ack *extack) 1755 { 1756 struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt); 1757 1758 if (IS_ERR(info)) 1759 return PTR_ERR(info); 1760 1761 info->vrf_table = nla_get_u32(attrs[SEG6_LOCAL_VRFTABLE]); 1762 1763 return 0; 1764 } 1765 1766 static int put_nla_vrftable(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1767 { 1768 struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt); 1769 1770 if (IS_ERR(info)) 1771 return PTR_ERR(info); 1772 1773 if (nla_put_u32(skb, SEG6_LOCAL_VRFTABLE, info->vrf_table)) 1774 return -EMSGSIZE; 1775 1776 return 0; 1777 } 1778 1779 static int cmp_nla_vrftable(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1780 { 1781 struct seg6_end_dt_info *info_a = seg6_possible_end_dt_info(a); 1782 struct seg6_end_dt_info *info_b = seg6_possible_end_dt_info(b); 1783 1784 if (info_a->vrf_table != info_b->vrf_table) 1785 return 1; 1786 1787 return 0; 1788 } 1789 1790 static int parse_nla_nh4(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1791 struct netlink_ext_ack *extack) 1792 { 1793 memcpy(&slwt->nh4, nla_data(attrs[SEG6_LOCAL_NH4]), 1794 sizeof(struct in_addr)); 1795 1796 return 0; 1797 } 1798 1799 static int put_nla_nh4(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1800 { 1801 struct nlattr *nla; 1802 1803 nla = nla_reserve(skb, SEG6_LOCAL_NH4, sizeof(struct in_addr)); 1804 if (!nla) 1805 return -EMSGSIZE; 1806 1807 memcpy(nla_data(nla), &slwt->nh4, sizeof(struct in_addr)); 1808 1809 return 0; 1810 } 1811 1812 static int cmp_nla_nh4(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1813 { 1814 return memcmp(&a->nh4, &b->nh4, sizeof(struct in_addr)); 1815 } 1816 1817 static int parse_nla_nh6(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1818 struct netlink_ext_ack *extack) 1819 { 1820 memcpy(&slwt->nh6, nla_data(attrs[SEG6_LOCAL_NH6]), 1821 sizeof(struct in6_addr)); 1822 1823 return 0; 1824 } 1825 1826 static int put_nla_nh6(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1827 { 1828 struct nlattr *nla; 1829 1830 nla = nla_reserve(skb, SEG6_LOCAL_NH6, sizeof(struct in6_addr)); 1831 if (!nla) 1832 return -EMSGSIZE; 1833 1834 memcpy(nla_data(nla), &slwt->nh6, sizeof(struct in6_addr)); 1835 1836 return 0; 1837 } 1838 1839 static int cmp_nla_nh6(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1840 { 1841 return memcmp(&a->nh6, &b->nh6, sizeof(struct in6_addr)); 1842 } 1843 1844 static int parse_nla_iif(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1845 struct netlink_ext_ack *extack) 1846 { 1847 slwt->iif = nla_get_u32(attrs[SEG6_LOCAL_IIF]); 1848 1849 return 0; 1850 } 1851 1852 static int put_nla_iif(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1853 { 1854 if (nla_put_u32(skb, SEG6_LOCAL_IIF, slwt->iif)) 1855 return -EMSGSIZE; 1856 1857 return 0; 1858 } 1859 1860 static int cmp_nla_iif(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1861 { 1862 if (a->iif != b->iif) 1863 return 1; 1864 1865 return 0; 1866 } 1867 1868 static int parse_nla_oif(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1869 struct netlink_ext_ack *extack) 1870 { 1871 slwt->oif = nla_get_u32(attrs[SEG6_LOCAL_OIF]); 1872 1873 return 0; 1874 } 1875 1876 static int put_nla_oif(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1877 { 1878 if (nla_put_u32(skb, SEG6_LOCAL_OIF, slwt->oif)) 1879 return -EMSGSIZE; 1880 1881 return 0; 1882 } 1883 1884 static int cmp_nla_oif(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1885 { 1886 if (a->oif != b->oif) 1887 return 1; 1888 1889 return 0; 1890 } 1891 1892 #define MAX_PROG_NAME 256 1893 static const struct nla_policy bpf_prog_policy[SEG6_LOCAL_BPF_PROG_MAX + 1] = { 1894 [SEG6_LOCAL_BPF_PROG] = { .type = NLA_U32, }, 1895 [SEG6_LOCAL_BPF_PROG_NAME] = { .type = NLA_NUL_STRING, 1896 .len = MAX_PROG_NAME }, 1897 }; 1898 1899 static int parse_nla_bpf(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1900 struct netlink_ext_ack *extack) 1901 { 1902 struct nlattr *tb[SEG6_LOCAL_BPF_PROG_MAX + 1]; 1903 struct bpf_prog *p; 1904 int ret; 1905 u32 fd; 1906 1907 ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_BPF_PROG_MAX, 1908 attrs[SEG6_LOCAL_BPF], 1909 bpf_prog_policy, NULL); 1910 if (ret < 0) 1911 return ret; 1912 1913 if (!tb[SEG6_LOCAL_BPF_PROG] || !tb[SEG6_LOCAL_BPF_PROG_NAME]) 1914 return -EINVAL; 1915 1916 slwt->bpf.name = nla_memdup(tb[SEG6_LOCAL_BPF_PROG_NAME], GFP_KERNEL); 1917 if (!slwt->bpf.name) 1918 return -ENOMEM; 1919 1920 fd = nla_get_u32(tb[SEG6_LOCAL_BPF_PROG]); 1921 p = bpf_prog_get_type(fd, BPF_PROG_TYPE_LWT_SEG6LOCAL); 1922 if (IS_ERR(p)) { 1923 kfree(slwt->bpf.name); 1924 return PTR_ERR(p); 1925 } 1926 1927 slwt->bpf.prog = p; 1928 return 0; 1929 } 1930 1931 static int put_nla_bpf(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1932 { 1933 struct nlattr *nest; 1934 1935 if (!slwt->bpf.prog) 1936 return 0; 1937 1938 nest = nla_nest_start_noflag(skb, SEG6_LOCAL_BPF); 1939 if (!nest) 1940 return -EMSGSIZE; 1941 1942 if (nla_put_u32(skb, SEG6_LOCAL_BPF_PROG, slwt->bpf.prog->aux->id)) 1943 return -EMSGSIZE; 1944 1945 if (slwt->bpf.name && 1946 nla_put_string(skb, SEG6_LOCAL_BPF_PROG_NAME, slwt->bpf.name)) 1947 return -EMSGSIZE; 1948 1949 return nla_nest_end(skb, nest); 1950 } 1951 1952 static int cmp_nla_bpf(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1953 { 1954 if (!a->bpf.name && !b->bpf.name) 1955 return 0; 1956 1957 if (!a->bpf.name || !b->bpf.name) 1958 return 1; 1959 1960 return strcmp(a->bpf.name, b->bpf.name); 1961 } 1962 1963 static void destroy_attr_bpf(struct seg6_local_lwt *slwt) 1964 { 1965 kfree(slwt->bpf.name); 1966 if (slwt->bpf.prog) 1967 bpf_prog_put(slwt->bpf.prog); 1968 } 1969 1970 static const struct 1971 nla_policy seg6_local_counters_policy[SEG6_LOCAL_CNT_MAX + 1] = { 1972 [SEG6_LOCAL_CNT_PACKETS] = { .type = NLA_U64 }, 1973 [SEG6_LOCAL_CNT_BYTES] = { .type = NLA_U64 }, 1974 [SEG6_LOCAL_CNT_ERRORS] = { .type = NLA_U64 }, 1975 }; 1976 1977 static int parse_nla_counters(struct nlattr **attrs, 1978 struct seg6_local_lwt *slwt, 1979 struct netlink_ext_ack *extack) 1980 { 1981 struct pcpu_seg6_local_counters __percpu *pcounters; 1982 struct nlattr *tb[SEG6_LOCAL_CNT_MAX + 1]; 1983 int ret; 1984 1985 ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_CNT_MAX, 1986 attrs[SEG6_LOCAL_COUNTERS], 1987 seg6_local_counters_policy, NULL); 1988 if (ret < 0) 1989 return ret; 1990 1991 /* basic support for SRv6 Behavior counters requires at least: 1992 * packets, bytes and errors. 1993 */ 1994 if (!tb[SEG6_LOCAL_CNT_PACKETS] || !tb[SEG6_LOCAL_CNT_BYTES] || 1995 !tb[SEG6_LOCAL_CNT_ERRORS]) 1996 return -EINVAL; 1997 1998 /* counters are always zero initialized */ 1999 pcounters = seg6_local_alloc_pcpu_counters(GFP_KERNEL); 2000 if (!pcounters) 2001 return -ENOMEM; 2002 2003 slwt->pcpu_counters = pcounters; 2004 2005 return 0; 2006 } 2007 2008 static int seg6_local_fill_nla_counters(struct sk_buff *skb, 2009 struct seg6_local_counters *counters) 2010 { 2011 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_PACKETS, counters->packets, 2012 SEG6_LOCAL_CNT_PAD)) 2013 return -EMSGSIZE; 2014 2015 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_BYTES, counters->bytes, 2016 SEG6_LOCAL_CNT_PAD)) 2017 return -EMSGSIZE; 2018 2019 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_ERRORS, counters->errors, 2020 SEG6_LOCAL_CNT_PAD)) 2021 return -EMSGSIZE; 2022 2023 return 0; 2024 } 2025 2026 static int put_nla_counters(struct sk_buff *skb, struct seg6_local_lwt *slwt) 2027 { 2028 struct seg6_local_counters counters = { 0, 0, 0 }; 2029 struct nlattr *nest; 2030 int rc, i; 2031 2032 nest = nla_nest_start(skb, SEG6_LOCAL_COUNTERS); 2033 if (!nest) 2034 return -EMSGSIZE; 2035 2036 for_each_possible_cpu(i) { 2037 struct pcpu_seg6_local_counters *pcounters; 2038 u64 packets, bytes, errors; 2039 unsigned int start; 2040 2041 pcounters = per_cpu_ptr(slwt->pcpu_counters, i); 2042 do { 2043 start = u64_stats_fetch_begin(&pcounters->syncp); 2044 2045 packets = u64_stats_read(&pcounters->packets); 2046 bytes = u64_stats_read(&pcounters->bytes); 2047 errors = u64_stats_read(&pcounters->errors); 2048 2049 } while (u64_stats_fetch_retry(&pcounters->syncp, start)); 2050 2051 counters.packets += packets; 2052 counters.bytes += bytes; 2053 counters.errors += errors; 2054 } 2055 2056 rc = seg6_local_fill_nla_counters(skb, &counters); 2057 if (rc < 0) { 2058 nla_nest_cancel(skb, nest); 2059 return rc; 2060 } 2061 2062 return nla_nest_end(skb, nest); 2063 } 2064 2065 static int cmp_nla_counters(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 2066 { 2067 /* a and b are equal if both have pcpu_counters set or not */ 2068 return (!!((unsigned long)a->pcpu_counters)) ^ 2069 (!!((unsigned long)b->pcpu_counters)); 2070 } 2071 2072 static void destroy_attr_counters(struct seg6_local_lwt *slwt) 2073 { 2074 free_percpu(slwt->pcpu_counters); 2075 } 2076 2077 static const 2078 struct nla_policy seg6_local_flavors_policy[SEG6_LOCAL_FLV_MAX + 1] = { 2079 [SEG6_LOCAL_FLV_OPERATION] = { .type = NLA_U32 }, 2080 [SEG6_LOCAL_FLV_LCBLOCK_BITS] = { .type = NLA_U8 }, 2081 [SEG6_LOCAL_FLV_LCNODE_FN_BITS] = { .type = NLA_U8 }, 2082 }; 2083 2084 /* check whether the lengths of the Locator-Block and Locator-Node Function 2085 * are compatible with the dimension of a C-SID container. 2086 */ 2087 static int seg6_chk_next_csid_cfg(__u8 block_len, __u8 func_len) 2088 { 2089 /* Locator-Block and Locator-Node Function cannot exceed 128 bits 2090 * (i.e. C-SID container length). 2091 */ 2092 if (next_csid_chk_cntr_bits(block_len, func_len)) 2093 return -EINVAL; 2094 2095 /* Locator-Block length must be greater than zero and evenly divisible 2096 * by 8. There must be room for a Locator-Node Function, at least. 2097 */ 2098 if (next_csid_chk_lcblock_bits(block_len)) 2099 return -EINVAL; 2100 2101 /* Locator-Node Function length must be greater than zero and evenly 2102 * divisible by 8. There must be room for the Locator-Block. 2103 */ 2104 if (next_csid_chk_lcnode_fn_bits(func_len)) 2105 return -EINVAL; 2106 2107 return 0; 2108 } 2109 2110 static int seg6_parse_nla_next_csid_cfg(struct nlattr **tb, 2111 struct seg6_flavors_info *finfo, 2112 struct netlink_ext_ack *extack) 2113 { 2114 __u8 func_len = SEG6_LOCAL_LCNODE_FN_DBITS; 2115 __u8 block_len = SEG6_LOCAL_LCBLOCK_DBITS; 2116 int rc; 2117 2118 if (tb[SEG6_LOCAL_FLV_LCBLOCK_BITS]) 2119 block_len = nla_get_u8(tb[SEG6_LOCAL_FLV_LCBLOCK_BITS]); 2120 2121 if (tb[SEG6_LOCAL_FLV_LCNODE_FN_BITS]) 2122 func_len = nla_get_u8(tb[SEG6_LOCAL_FLV_LCNODE_FN_BITS]); 2123 2124 rc = seg6_chk_next_csid_cfg(block_len, func_len); 2125 if (rc < 0) { 2126 NL_SET_ERR_MSG(extack, 2127 "Invalid Locator Block/Node Function lengths"); 2128 return rc; 2129 } 2130 2131 finfo->lcblock_bits = block_len; 2132 finfo->lcnode_func_bits = func_len; 2133 2134 return 0; 2135 } 2136 2137 static int parse_nla_flavors(struct nlattr **attrs, struct seg6_local_lwt *slwt, 2138 struct netlink_ext_ack *extack) 2139 { 2140 struct seg6_flavors_info *finfo = &slwt->flv_info; 2141 struct nlattr *tb[SEG6_LOCAL_FLV_MAX + 1]; 2142 int action = slwt->action; 2143 __u32 fops, supp_fops; 2144 int rc; 2145 2146 rc = nla_parse_nested_deprecated(tb, SEG6_LOCAL_FLV_MAX, 2147 attrs[SEG6_LOCAL_FLAVORS], 2148 seg6_local_flavors_policy, NULL); 2149 if (rc < 0) 2150 return rc; 2151 2152 /* this attribute MUST always be present since it represents the Flavor 2153 * operation(s) to be carried out. 2154 */ 2155 if (!tb[SEG6_LOCAL_FLV_OPERATION]) 2156 return -EINVAL; 2157 2158 fops = nla_get_u32(tb[SEG6_LOCAL_FLV_OPERATION]); 2159 rc = seg6_flv_supp_ops_by_action(action, &supp_fops); 2160 if (rc < 0 || (fops & ~supp_fops)) { 2161 NL_SET_ERR_MSG(extack, "Unsupported Flavor operation(s)"); 2162 return -EOPNOTSUPP; 2163 } 2164 2165 finfo->flv_ops = fops; 2166 2167 if (seg6_next_csid_enabled(fops)) { 2168 /* Locator-Block and Locator-Node Function lengths can be 2169 * provided by the user space. Otherwise, default values are 2170 * applied. 2171 */ 2172 rc = seg6_parse_nla_next_csid_cfg(tb, finfo, extack); 2173 if (rc < 0) 2174 return rc; 2175 } 2176 2177 return 0; 2178 } 2179 2180 static int seg6_fill_nla_next_csid_cfg(struct sk_buff *skb, 2181 struct seg6_flavors_info *finfo) 2182 { 2183 if (nla_put_u8(skb, SEG6_LOCAL_FLV_LCBLOCK_BITS, finfo->lcblock_bits)) 2184 return -EMSGSIZE; 2185 2186 if (nla_put_u8(skb, SEG6_LOCAL_FLV_LCNODE_FN_BITS, 2187 finfo->lcnode_func_bits)) 2188 return -EMSGSIZE; 2189 2190 return 0; 2191 } 2192 2193 static int put_nla_flavors(struct sk_buff *skb, struct seg6_local_lwt *slwt) 2194 { 2195 struct seg6_flavors_info *finfo = &slwt->flv_info; 2196 __u32 fops = finfo->flv_ops; 2197 struct nlattr *nest; 2198 int rc; 2199 2200 nest = nla_nest_start(skb, SEG6_LOCAL_FLAVORS); 2201 if (!nest) 2202 return -EMSGSIZE; 2203 2204 if (nla_put_u32(skb, SEG6_LOCAL_FLV_OPERATION, fops)) { 2205 rc = -EMSGSIZE; 2206 goto err; 2207 } 2208 2209 if (seg6_next_csid_enabled(fops)) { 2210 rc = seg6_fill_nla_next_csid_cfg(skb, finfo); 2211 if (rc < 0) 2212 goto err; 2213 } 2214 2215 return nla_nest_end(skb, nest); 2216 2217 err: 2218 nla_nest_cancel(skb, nest); 2219 return rc; 2220 } 2221 2222 static int seg6_cmp_nla_next_csid_cfg(struct seg6_flavors_info *finfo_a, 2223 struct seg6_flavors_info *finfo_b) 2224 { 2225 if (finfo_a->lcblock_bits != finfo_b->lcblock_bits) 2226 return 1; 2227 2228 if (finfo_a->lcnode_func_bits != finfo_b->lcnode_func_bits) 2229 return 1; 2230 2231 return 0; 2232 } 2233 2234 static int cmp_nla_flavors(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 2235 { 2236 struct seg6_flavors_info *finfo_a = &a->flv_info; 2237 struct seg6_flavors_info *finfo_b = &b->flv_info; 2238 2239 if (finfo_a->flv_ops != finfo_b->flv_ops) 2240 return 1; 2241 2242 if (seg6_next_csid_enabled(finfo_a->flv_ops)) { 2243 if (seg6_cmp_nla_next_csid_cfg(finfo_a, finfo_b)) 2244 return 1; 2245 } 2246 2247 return 0; 2248 } 2249 2250 static int encap_size_flavors(struct seg6_local_lwt *slwt) 2251 { 2252 struct seg6_flavors_info *finfo = &slwt->flv_info; 2253 int nlsize; 2254 2255 nlsize = nla_total_size(0) + /* nest SEG6_LOCAL_FLAVORS */ 2256 nla_total_size(4); /* SEG6_LOCAL_FLV_OPERATION */ 2257 2258 if (seg6_next_csid_enabled(finfo->flv_ops)) 2259 nlsize += nla_total_size(1) + /* SEG6_LOCAL_FLV_LCBLOCK_BITS */ 2260 nla_total_size(1); /* SEG6_LOCAL_FLV_LCNODE_FN_BITS */ 2261 2262 return nlsize; 2263 } 2264 2265 struct seg6_action_param { 2266 int (*parse)(struct nlattr **attrs, struct seg6_local_lwt *slwt, 2267 struct netlink_ext_ack *extack); 2268 int (*put)(struct sk_buff *skb, struct seg6_local_lwt *slwt); 2269 int (*cmp)(struct seg6_local_lwt *a, struct seg6_local_lwt *b); 2270 2271 /* optional destroy() callback useful for releasing resources which 2272 * have been previously acquired in the corresponding parse() 2273 * function. 2274 */ 2275 void (*destroy)(struct seg6_local_lwt *slwt); 2276 }; 2277 2278 static struct seg6_action_param seg6_action_params[SEG6_LOCAL_MAX + 1] = { 2279 [SEG6_LOCAL_SRH] = { .parse = parse_nla_srh, 2280 .put = put_nla_srh, 2281 .cmp = cmp_nla_srh, 2282 .destroy = destroy_attr_srh }, 2283 2284 [SEG6_LOCAL_TABLE] = { .parse = parse_nla_table, 2285 .put = put_nla_table, 2286 .cmp = cmp_nla_table }, 2287 2288 [SEG6_LOCAL_NH4] = { .parse = parse_nla_nh4, 2289 .put = put_nla_nh4, 2290 .cmp = cmp_nla_nh4 }, 2291 2292 [SEG6_LOCAL_NH6] = { .parse = parse_nla_nh6, 2293 .put = put_nla_nh6, 2294 .cmp = cmp_nla_nh6 }, 2295 2296 [SEG6_LOCAL_IIF] = { .parse = parse_nla_iif, 2297 .put = put_nla_iif, 2298 .cmp = cmp_nla_iif }, 2299 2300 [SEG6_LOCAL_OIF] = { .parse = parse_nla_oif, 2301 .put = put_nla_oif, 2302 .cmp = cmp_nla_oif }, 2303 2304 [SEG6_LOCAL_BPF] = { .parse = parse_nla_bpf, 2305 .put = put_nla_bpf, 2306 .cmp = cmp_nla_bpf, 2307 .destroy = destroy_attr_bpf }, 2308 2309 [SEG6_LOCAL_VRFTABLE] = { .parse = parse_nla_vrftable, 2310 .put = put_nla_vrftable, 2311 .cmp = cmp_nla_vrftable }, 2312 2313 [SEG6_LOCAL_COUNTERS] = { .parse = parse_nla_counters, 2314 .put = put_nla_counters, 2315 .cmp = cmp_nla_counters, 2316 .destroy = destroy_attr_counters }, 2317 2318 [SEG6_LOCAL_FLAVORS] = { .parse = parse_nla_flavors, 2319 .put = put_nla_flavors, 2320 .cmp = cmp_nla_flavors }, 2321 }; 2322 2323 /* call the destroy() callback (if available) for each set attribute in 2324 * @parsed_attrs, starting from the first attribute up to the @max_parsed 2325 * (excluded) attribute. 2326 */ 2327 static void __destroy_attrs(unsigned long parsed_attrs, int max_parsed, 2328 struct seg6_local_lwt *slwt) 2329 { 2330 struct seg6_action_param *param; 2331 int i; 2332 2333 /* Every required seg6local attribute is identified by an ID which is 2334 * encoded as a flag (i.e: 1 << ID) in the 'attrs' bitmask; 2335 * 2336 * We scan the 'parsed_attrs' bitmask, starting from the first attribute 2337 * up to the @max_parsed (excluded) attribute. 2338 * For each set attribute, we retrieve the corresponding destroy() 2339 * callback. If the callback is not available, then we skip to the next 2340 * attribute; otherwise, we call the destroy() callback. 2341 */ 2342 for (i = SEG6_LOCAL_SRH; i < max_parsed; ++i) { 2343 if (!(parsed_attrs & SEG6_F_ATTR(i))) 2344 continue; 2345 2346 param = &seg6_action_params[i]; 2347 2348 if (param->destroy) 2349 param->destroy(slwt); 2350 } 2351 } 2352 2353 /* release all the resources that may have been acquired during parsing 2354 * operations. 2355 */ 2356 static void destroy_attrs(struct seg6_local_lwt *slwt) 2357 { 2358 unsigned long attrs = slwt->desc->attrs | slwt->parsed_optattrs; 2359 2360 __destroy_attrs(attrs, SEG6_LOCAL_MAX + 1, slwt); 2361 } 2362 2363 static int parse_nla_optional_attrs(struct nlattr **attrs, 2364 struct seg6_local_lwt *slwt, 2365 struct netlink_ext_ack *extack) 2366 { 2367 struct seg6_action_desc *desc = slwt->desc; 2368 unsigned long parsed_optattrs = 0; 2369 struct seg6_action_param *param; 2370 int err, i; 2371 2372 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; ++i) { 2373 if (!(desc->optattrs & SEG6_F_ATTR(i)) || !attrs[i]) 2374 continue; 2375 2376 /* once here, the i-th attribute is provided by the 2377 * userspace AND it is identified optional as well. 2378 */ 2379 param = &seg6_action_params[i]; 2380 2381 err = param->parse(attrs, slwt, extack); 2382 if (err < 0) 2383 goto parse_optattrs_err; 2384 2385 /* current attribute has been correctly parsed */ 2386 parsed_optattrs |= SEG6_F_ATTR(i); 2387 } 2388 2389 /* store in the tunnel state all the optional attributed successfully 2390 * parsed. 2391 */ 2392 slwt->parsed_optattrs = parsed_optattrs; 2393 2394 return 0; 2395 2396 parse_optattrs_err: 2397 __destroy_attrs(parsed_optattrs, i, slwt); 2398 2399 return err; 2400 } 2401 2402 /* call the custom constructor of the behavior during its initialization phase 2403 * and after that all its attributes have been parsed successfully. 2404 */ 2405 static int 2406 seg6_local_lwtunnel_build_state(struct seg6_local_lwt *slwt, const void *cfg, 2407 struct netlink_ext_ack *extack) 2408 { 2409 struct seg6_action_desc *desc = slwt->desc; 2410 struct seg6_local_lwtunnel_ops *ops; 2411 2412 ops = &desc->slwt_ops; 2413 if (!ops->build_state) 2414 return 0; 2415 2416 return ops->build_state(slwt, cfg, extack); 2417 } 2418 2419 /* call the custom destructor of the behavior which is invoked before the 2420 * tunnel is going to be destroyed. 2421 */ 2422 static void seg6_local_lwtunnel_destroy_state(struct seg6_local_lwt *slwt) 2423 { 2424 struct seg6_action_desc *desc = slwt->desc; 2425 struct seg6_local_lwtunnel_ops *ops; 2426 2427 ops = &desc->slwt_ops; 2428 if (!ops->destroy_state) 2429 return; 2430 2431 ops->destroy_state(slwt); 2432 } 2433 2434 static int parse_nla_action(struct nlattr **attrs, struct seg6_local_lwt *slwt, 2435 struct netlink_ext_ack *extack) 2436 { 2437 struct seg6_action_param *param; 2438 struct seg6_action_desc *desc; 2439 unsigned long invalid_attrs; 2440 int i, err; 2441 2442 desc = __get_action_desc(slwt->action); 2443 if (!desc) 2444 return -EINVAL; 2445 2446 if (!desc->input) 2447 return -EOPNOTSUPP; 2448 2449 slwt->desc = desc; 2450 slwt->headroom += desc->static_headroom; 2451 2452 /* Forcing the desc->optattrs *set* and the desc->attrs *set* to be 2453 * disjoined, this allow us to release acquired resources by optional 2454 * attributes and by required attributes independently from each other 2455 * without any interference. 2456 * In other terms, we are sure that we do not release some the acquired 2457 * resources twice. 2458 * 2459 * Note that if an attribute is configured both as required and as 2460 * optional, it means that the user has messed something up in the 2461 * seg6_action_table. Therefore, this check is required for SRv6 2462 * behaviors to work properly. 2463 */ 2464 invalid_attrs = desc->attrs & desc->optattrs; 2465 if (invalid_attrs) { 2466 WARN_ONCE(1, 2467 "An attribute cannot be both required AND optional"); 2468 return -EINVAL; 2469 } 2470 2471 /* parse the required attributes */ 2472 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) { 2473 if (desc->attrs & SEG6_F_ATTR(i)) { 2474 if (!attrs[i]) 2475 return -EINVAL; 2476 2477 param = &seg6_action_params[i]; 2478 2479 err = param->parse(attrs, slwt, extack); 2480 if (err < 0) 2481 goto parse_attrs_err; 2482 } 2483 } 2484 2485 /* parse the optional attributes, if any */ 2486 err = parse_nla_optional_attrs(attrs, slwt, extack); 2487 if (err < 0) 2488 goto parse_attrs_err; 2489 2490 return 0; 2491 2492 parse_attrs_err: 2493 /* release any resource that may have been acquired during the i-1 2494 * parse() operations. 2495 */ 2496 __destroy_attrs(desc->attrs, i, slwt); 2497 2498 return err; 2499 } 2500 2501 static int seg6_local_build_state(struct net *net, struct nlattr *nla, 2502 unsigned int family, const void *cfg, 2503 struct lwtunnel_state **ts, 2504 struct netlink_ext_ack *extack) 2505 { 2506 struct nlattr *tb[SEG6_LOCAL_MAX + 1]; 2507 struct lwtunnel_state *newts; 2508 struct seg6_local_lwt *slwt; 2509 int err; 2510 2511 if (family != AF_INET6) 2512 return -EINVAL; 2513 2514 err = nla_parse_nested_deprecated(tb, SEG6_LOCAL_MAX, nla, 2515 seg6_local_policy, extack); 2516 2517 if (err < 0) 2518 return err; 2519 2520 if (!tb[SEG6_LOCAL_ACTION]) 2521 return -EINVAL; 2522 2523 newts = lwtunnel_state_alloc(sizeof(*slwt)); 2524 if (!newts) 2525 return -ENOMEM; 2526 2527 slwt = seg6_local_lwtunnel(newts); 2528 slwt->action = nla_get_u32(tb[SEG6_LOCAL_ACTION]); 2529 2530 err = parse_nla_action(tb, slwt, extack); 2531 if (err < 0) 2532 goto out_free; 2533 2534 err = seg6_local_lwtunnel_build_state(slwt, cfg, extack); 2535 if (err < 0) 2536 goto out_destroy_attrs; 2537 2538 newts->type = LWTUNNEL_ENCAP_SEG6_LOCAL; 2539 newts->flags = LWTUNNEL_STATE_INPUT_REDIRECT; 2540 newts->headroom = slwt->headroom; 2541 2542 *ts = newts; 2543 2544 return 0; 2545 2546 out_destroy_attrs: 2547 destroy_attrs(slwt); 2548 out_free: 2549 kfree(newts); 2550 return err; 2551 } 2552 2553 static void seg6_local_destroy_state(struct lwtunnel_state *lwt) 2554 { 2555 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt); 2556 2557 seg6_local_lwtunnel_destroy_state(slwt); 2558 2559 destroy_attrs(slwt); 2560 2561 return; 2562 } 2563 2564 static int seg6_local_fill_encap(struct sk_buff *skb, 2565 struct lwtunnel_state *lwt) 2566 { 2567 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt); 2568 struct seg6_action_param *param; 2569 unsigned long attrs; 2570 int i, err; 2571 2572 if (nla_put_u32(skb, SEG6_LOCAL_ACTION, slwt->action)) 2573 return -EMSGSIZE; 2574 2575 attrs = slwt->desc->attrs | slwt->parsed_optattrs; 2576 2577 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) { 2578 if (attrs & SEG6_F_ATTR(i)) { 2579 param = &seg6_action_params[i]; 2580 err = param->put(skb, slwt); 2581 if (err < 0) 2582 return err; 2583 } 2584 } 2585 2586 return 0; 2587 } 2588 2589 static int seg6_local_get_encap_size(struct lwtunnel_state *lwt) 2590 { 2591 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt); 2592 unsigned long attrs; 2593 int nlsize; 2594 2595 nlsize = nla_total_size(4); /* action */ 2596 2597 attrs = slwt->desc->attrs | slwt->parsed_optattrs; 2598 2599 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_SRH)) 2600 nlsize += nla_total_size((slwt->srh->hdrlen + 1) << 3); 2601 2602 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE)) 2603 nlsize += nla_total_size(4); 2604 2605 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH4)) 2606 nlsize += nla_total_size(4); 2607 2608 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH6)) 2609 nlsize += nla_total_size(16); 2610 2611 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_IIF)) 2612 nlsize += nla_total_size(4); 2613 2614 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_OIF)) 2615 nlsize += nla_total_size(4); 2616 2617 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_BPF)) 2618 nlsize += nla_total_size(sizeof(struct nlattr)) + 2619 nla_total_size(MAX_PROG_NAME) + 2620 nla_total_size(4); 2621 2622 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE)) 2623 nlsize += nla_total_size(4); 2624 2625 if (attrs & SEG6_F_LOCAL_COUNTERS) 2626 nlsize += nla_total_size(0) + /* nest SEG6_LOCAL_COUNTERS */ 2627 /* SEG6_LOCAL_CNT_PACKETS */ 2628 nla_total_size_64bit(sizeof(__u64)) + 2629 /* SEG6_LOCAL_CNT_BYTES */ 2630 nla_total_size_64bit(sizeof(__u64)) + 2631 /* SEG6_LOCAL_CNT_ERRORS */ 2632 nla_total_size_64bit(sizeof(__u64)); 2633 2634 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_FLAVORS)) 2635 nlsize += encap_size_flavors(slwt); 2636 2637 return nlsize; 2638 } 2639 2640 static int seg6_local_cmp_encap(struct lwtunnel_state *a, 2641 struct lwtunnel_state *b) 2642 { 2643 struct seg6_local_lwt *slwt_a, *slwt_b; 2644 struct seg6_action_param *param; 2645 unsigned long attrs_a, attrs_b; 2646 int i; 2647 2648 slwt_a = seg6_local_lwtunnel(a); 2649 slwt_b = seg6_local_lwtunnel(b); 2650 2651 if (slwt_a->action != slwt_b->action) 2652 return 1; 2653 2654 attrs_a = slwt_a->desc->attrs | slwt_a->parsed_optattrs; 2655 attrs_b = slwt_b->desc->attrs | slwt_b->parsed_optattrs; 2656 2657 if (attrs_a != attrs_b) 2658 return 1; 2659 2660 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) { 2661 if (attrs_a & SEG6_F_ATTR(i)) { 2662 param = &seg6_action_params[i]; 2663 if (param->cmp(slwt_a, slwt_b)) 2664 return 1; 2665 } 2666 } 2667 2668 return 0; 2669 } 2670 2671 static const struct lwtunnel_encap_ops seg6_local_ops = { 2672 .build_state = seg6_local_build_state, 2673 .destroy_state = seg6_local_destroy_state, 2674 .input = seg6_local_input, 2675 .fill_encap = seg6_local_fill_encap, 2676 .get_encap_size = seg6_local_get_encap_size, 2677 .cmp_encap = seg6_local_cmp_encap, 2678 .owner = THIS_MODULE, 2679 }; 2680 2681 int __init seg6_local_init(void) 2682 { 2683 /* If the max total number of defined attributes is reached, then your 2684 * kernel build stops here. 2685 * 2686 * This check is required to avoid arithmetic overflows when processing 2687 * behavior attributes and the maximum number of defined attributes 2688 * exceeds the allowed value. 2689 */ 2690 BUILD_BUG_ON(SEG6_LOCAL_MAX + 1 > BITS_PER_TYPE(unsigned long)); 2691 2692 /* Check whether the number of defined flavors exceeds the maximum 2693 * allowed value. 2694 */ 2695 BUILD_BUG_ON(SEG6_LOCAL_FLV_OP_MAX + 1 > BITS_PER_TYPE(__u32)); 2696 2697 /* If the default NEXT-C-SID Locator-Block/Node Function lengths (in 2698 * bits) have been changed with invalid values, kernel build stops 2699 * here. 2700 */ 2701 BUILD_BUG_ON(next_csid_chk_cntr_bits(SEG6_LOCAL_LCBLOCK_DBITS, 2702 SEG6_LOCAL_LCNODE_FN_DBITS)); 2703 BUILD_BUG_ON(next_csid_chk_lcblock_bits(SEG6_LOCAL_LCBLOCK_DBITS)); 2704 BUILD_BUG_ON(next_csid_chk_lcnode_fn_bits(SEG6_LOCAL_LCNODE_FN_DBITS)); 2705 2706 /* To be memory efficient, we use 'u8' to represent the different 2707 * actions related to RFC8986 flavors. If the kernel build stops here, 2708 * it means that it is not possible to correctly encode these actions 2709 * with the data type chosen for the action table. 2710 */ 2711 BUILD_BUG_ON(SEG6_LOCAL_FLV_ACT_MAX > (typeof(flv8986_act_tbl[0]))~0U); 2712 2713 return lwtunnel_encap_add_ops(&seg6_local_ops, 2714 LWTUNNEL_ENCAP_SEG6_LOCAL); 2715 } 2716 2717 void seg6_local_exit(void) 2718 { 2719 lwtunnel_encap_del_ops(&seg6_local_ops, LWTUNNEL_ENCAP_SEG6_LOCAL); 2720 } 2721