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 if (proto == IPPROTO_IPIP) { 260 bool l3slave = ipv6_l3mdev_skb(IP6CB(skb)->flags); 261 int iif = IP6CB(skb)->iif; 262 263 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 264 IPCB(skb)->iif = iif; 265 if (l3slave) 266 IPCB(skb)->flags |= IPSKB_L3SLAVE; 267 } else if (proto == IPPROTO_IPV6) { 268 bool l3slave = ipv6_l3mdev_skb(IP6CB(skb)->flags); 269 int iif = IP6CB(skb)->iif; 270 271 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 272 IP6CB(skb)->iif = iif; 273 IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr); 274 if (l3slave) 275 IP6CB(skb)->flags |= IP6SKB_L3SLAVE; 276 } 277 278 return true; 279 } 280 281 static void advance_nextseg(struct ipv6_sr_hdr *srh, struct in6_addr *daddr) 282 { 283 struct in6_addr *addr; 284 285 srh->segments_left--; 286 addr = srh->segments + srh->segments_left; 287 *daddr = *addr; 288 } 289 290 static int 291 seg6_lookup_any_nexthop(struct sk_buff *skb, struct in6_addr *nhaddr, 292 u32 tbl_id, bool local_delivery, int oif) 293 { 294 struct net *net = dev_net(skb->dev); 295 struct ipv6hdr *hdr = ipv6_hdr(skb); 296 int flags = RT6_LOOKUP_F_HAS_SADDR; 297 struct dst_entry *dst = NULL; 298 struct rt6_info *rt; 299 struct flowi6 fl6; 300 int dev_flags = 0; 301 302 memset(&fl6, 0, sizeof(fl6)); 303 fl6.flowi6_iif = skb->dev->ifindex; 304 fl6.flowi6_oif = oif; 305 fl6.daddr = nhaddr ? *nhaddr : hdr->daddr; 306 fl6.saddr = hdr->saddr; 307 fl6.flowlabel = ip6_flowinfo(hdr); 308 fl6.flowi6_mark = skb->mark; 309 fl6.flowi6_proto = hdr->nexthdr; 310 311 if (nhaddr) 312 fl6.flowi6_flags = FLOWI_FLAG_KNOWN_NH; 313 314 if (!tbl_id && !oif) { 315 dst = ip6_route_input_lookup(net, skb->dev, &fl6, skb, flags); 316 } else if (tbl_id) { 317 struct fib6_table *table; 318 319 table = fib6_get_table(net, tbl_id); 320 if (!table) 321 goto out; 322 323 rt = ip6_pol_route(net, table, oif, &fl6, skb, flags); 324 dst = &rt->dst; 325 } else { 326 dst = ip6_route_output(net, NULL, &fl6); 327 } 328 329 /* we want to discard traffic destined for local packet processing, 330 * if @local_delivery is set to false. 331 */ 332 if (!local_delivery) 333 dev_flags |= IFF_LOOPBACK; 334 335 if (dst && (dst_dev(dst)->flags & dev_flags) && !dst->error) { 336 dst_release(dst); 337 dst = NULL; 338 } 339 340 out: 341 if (!dst) { 342 rt = net->ipv6.ip6_blk_hole_entry; 343 dst = &rt->dst; 344 dst_hold(dst); 345 } 346 347 skb_dst_drop(skb); 348 skb_dst_set(skb, dst); 349 return dst->error; 350 } 351 352 int seg6_lookup_nexthop(struct sk_buff *skb, 353 struct in6_addr *nhaddr, u32 tbl_id) 354 { 355 return seg6_lookup_any_nexthop(skb, nhaddr, tbl_id, false, 0); 356 } 357 358 static __u8 seg6_flv_lcblock_octects(const struct seg6_flavors_info *finfo) 359 { 360 return finfo->lcblock_bits >> 3; 361 } 362 363 static __u8 seg6_flv_lcnode_func_octects(const struct seg6_flavors_info *finfo) 364 { 365 return finfo->lcnode_func_bits >> 3; 366 } 367 368 static bool seg6_next_csid_is_arg_zero(const struct in6_addr *addr, 369 const struct seg6_flavors_info *finfo) 370 { 371 __u8 fnc_octects = seg6_flv_lcnode_func_octects(finfo); 372 __u8 blk_octects = seg6_flv_lcblock_octects(finfo); 373 __u8 arg_octects; 374 int i; 375 376 arg_octects = 16 - blk_octects - fnc_octects; 377 for (i = 0; i < arg_octects; ++i) { 378 if (addr->s6_addr[blk_octects + fnc_octects + i] != 0x00) 379 return false; 380 } 381 382 return true; 383 } 384 385 /* assume that DA.Argument length > 0 */ 386 static void seg6_next_csid_advance_arg(struct in6_addr *addr, 387 const struct seg6_flavors_info *finfo) 388 { 389 __u8 fnc_octects = seg6_flv_lcnode_func_octects(finfo); 390 __u8 blk_octects = seg6_flv_lcblock_octects(finfo); 391 392 /* advance DA.Argument */ 393 memmove(&addr->s6_addr[blk_octects], 394 &addr->s6_addr[blk_octects + fnc_octects], 395 16 - blk_octects - fnc_octects); 396 397 memset(&addr->s6_addr[16 - fnc_octects], 0x00, fnc_octects); 398 } 399 400 static int input_action_end_finish(struct sk_buff *skb, 401 struct seg6_local_lwt *slwt) 402 { 403 seg6_lookup_nexthop(skb, NULL, 0); 404 405 return dst_input(skb); 406 } 407 408 static int input_action_end_core(struct sk_buff *skb, 409 struct seg6_local_lwt *slwt) 410 { 411 struct ipv6_sr_hdr *srh; 412 413 srh = get_and_validate_srh(skb); 414 if (!srh) 415 goto drop; 416 417 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 418 419 return input_action_end_finish(skb, slwt); 420 421 drop: 422 kfree_skb(skb); 423 return -EINVAL; 424 } 425 426 static int end_next_csid_core(struct sk_buff *skb, struct seg6_local_lwt *slwt) 427 { 428 const struct seg6_flavors_info *finfo = &slwt->flv_info; 429 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr; 430 431 if (seg6_next_csid_is_arg_zero(daddr, finfo)) 432 return input_action_end_core(skb, slwt); 433 434 /* update DA */ 435 seg6_next_csid_advance_arg(daddr, finfo); 436 437 return input_action_end_finish(skb, slwt); 438 } 439 440 static int input_action_end_x_finish(struct sk_buff *skb, 441 struct seg6_local_lwt *slwt) 442 { 443 seg6_lookup_any_nexthop(skb, &slwt->nh6, 0, false, slwt->oif); 444 445 return dst_input(skb); 446 } 447 448 static int input_action_end_x_core(struct sk_buff *skb, 449 struct seg6_local_lwt *slwt) 450 { 451 struct ipv6_sr_hdr *srh; 452 453 srh = get_and_validate_srh(skb); 454 if (!srh) 455 goto drop; 456 457 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 458 459 return input_action_end_x_finish(skb, slwt); 460 461 drop: 462 kfree_skb(skb); 463 return -EINVAL; 464 } 465 466 static int end_x_next_csid_core(struct sk_buff *skb, 467 struct seg6_local_lwt *slwt) 468 { 469 const struct seg6_flavors_info *finfo = &slwt->flv_info; 470 struct in6_addr *daddr = &ipv6_hdr(skb)->daddr; 471 472 if (seg6_next_csid_is_arg_zero(daddr, finfo)) 473 return input_action_end_x_core(skb, slwt); 474 475 /* update DA */ 476 seg6_next_csid_advance_arg(daddr, finfo); 477 478 return input_action_end_x_finish(skb, slwt); 479 } 480 481 static bool seg6_next_csid_enabled(__u32 fops) 482 { 483 return fops & SEG6_F_LOCAL_FLV_NEXT_CSID; 484 } 485 486 /* Processing of SRv6 End, End.X, and End.T behaviors can be extended through 487 * the flavors framework. These behaviors must report the subset of (flavor) 488 * operations they currently implement. In this way, if a user specifies a 489 * flavor combination that is not supported by a given End* behavior, the 490 * kernel refuses to instantiate the tunnel reporting the error. 491 */ 492 static int seg6_flv_supp_ops_by_action(int action, __u32 *fops) 493 { 494 switch (action) { 495 case SEG6_LOCAL_ACTION_END: 496 *fops = SEG6_LOCAL_END_FLV_SUPP_OPS; 497 break; 498 case SEG6_LOCAL_ACTION_END_X: 499 *fops = SEG6_LOCAL_END_X_FLV_SUPP_OPS; 500 break; 501 default: 502 return -EOPNOTSUPP; 503 } 504 505 return 0; 506 } 507 508 /* We describe the packet state in relation to the absence/presence of the SRH 509 * and the Segment Left (SL) field. 510 * For our purposes, it is not necessary to record the exact value of the SL 511 * when the SID List consists of two or more segments. 512 */ 513 enum seg6_local_pktinfo { 514 /* the order really matters! */ 515 SEG6_LOCAL_PKTINFO_NOHDR = 0, 516 SEG6_LOCAL_PKTINFO_SL_ZERO, 517 SEG6_LOCAL_PKTINFO_SL_ONE, 518 SEG6_LOCAL_PKTINFO_SL_MORE, 519 __SEG6_LOCAL_PKTINFO_MAX, 520 }; 521 522 #define SEG6_LOCAL_PKTINFO_MAX (__SEG6_LOCAL_PKTINFO_MAX - 1) 523 524 static enum seg6_local_pktinfo seg6_get_srh_pktinfo(struct ipv6_sr_hdr *srh) 525 { 526 __u8 sgl; 527 528 if (!srh) 529 return SEG6_LOCAL_PKTINFO_NOHDR; 530 531 sgl = srh->segments_left; 532 if (sgl < 2) 533 return SEG6_LOCAL_PKTINFO_SL_ZERO + sgl; 534 535 return SEG6_LOCAL_PKTINFO_SL_MORE; 536 } 537 538 enum seg6_local_flv_action { 539 SEG6_LOCAL_FLV_ACT_UNSPEC = 0, 540 SEG6_LOCAL_FLV_ACT_END, 541 SEG6_LOCAL_FLV_ACT_PSP, 542 SEG6_LOCAL_FLV_ACT_USP, 543 SEG6_LOCAL_FLV_ACT_USD, 544 __SEG6_LOCAL_FLV_ACT_MAX 545 }; 546 547 #define SEG6_LOCAL_FLV_ACT_MAX (__SEG6_LOCAL_FLV_ACT_MAX - 1) 548 549 /* The action table for RFC8986 flavors (see the flv8986_act_tbl below) 550 * contains the actions (i.e. processing operations) to be applied on packets 551 * when flavors are configured for an End* behavior. 552 * By combining the pkinfo data and from the flavors mask, the macro 553 * computes the index used to access the elements (actions) stored in the 554 * action table. The index is structured as follows: 555 * 556 * index 557 * _______________/\________________ 558 * / \ 559 * +----------------+----------------+ 560 * | pf | afm | 561 * +----------------+----------------+ 562 * ph-1 ... p1 p0 fk-1 ... f1 f0 563 * MSB LSB 564 * 565 * where: 566 * - 'afm' (adjusted flavor mask) is the mask containing a combination of the 567 * RFC8986 flavors currently supported. 'afm' corresponds to the @fm 568 * argument of the macro whose value is righ-shifted by 1 bit. By doing so, 569 * we discard the SEG6_LOCAL_FLV_OP_UNSPEC flag (bit 0 in @fm) which is 570 * never used here; 571 * - 'pf' encodes the packet info (pktinfo) regarding the presence/absence of 572 * the SRH, SL = 0, etc. 'pf' is set with the value of @pf provided as 573 * argument to the macro. 574 */ 575 #define flv8986_act_tbl_idx(pf, fm) \ 576 ((((pf) << bits_per(SEG6_LOCAL_FLV8986_SUPP_OPS)) | \ 577 ((fm) & SEG6_LOCAL_FLV8986_SUPP_OPS)) >> SEG6_LOCAL_FLV_OP_PSP) 578 579 /* We compute the size of the action table by considering the RFC8986 flavors 580 * actually supported by the kernel. In this way, the size is automatically 581 * adjusted when new flavors are supported. 582 */ 583 #define FLV8986_ACT_TBL_SIZE \ 584 roundup_pow_of_two(flv8986_act_tbl_idx(SEG6_LOCAL_PKTINFO_MAX, \ 585 SEG6_LOCAL_FLV8986_SUPP_OPS)) 586 587 /* tbl_cfg(act, pf, fm) macro is used to easily configure the action 588 * table; it accepts 3 arguments: 589 * i) @act, the suffix from SEG6_LOCAL_FLV_ACT_{act} representing 590 * the action that should be applied on the packet; 591 * ii) @pf, the suffix from SEG6_LOCAL_PKTINFO_{pf} reporting the packet 592 * info about the lack/presence of SRH, SRH with SL = 0, etc; 593 * iii) @fm, the mask of flavors. 594 */ 595 #define tbl_cfg(act, pf, fm) \ 596 [flv8986_act_tbl_idx(SEG6_LOCAL_PKTINFO_##pf, \ 597 (fm))] = SEG6_LOCAL_FLV_ACT_##act 598 599 /* shorthand for improving readability */ 600 #define F_PSP SEG6_F_LOCAL_FLV_PSP 601 602 /* The table contains, for each combination of the pktinfo data and 603 * flavors, the action that should be taken on a packet (e.g. 604 * "standard" Endpoint processing, Penultimate Segment Pop, etc). 605 * 606 * By default, table entries not explicitly configured are initialized with the 607 * SEG6_LOCAL_FLV_ACT_UNSPEC action, which generally has the effect of 608 * discarding the processed packet. 609 */ 610 static const u8 flv8986_act_tbl[FLV8986_ACT_TBL_SIZE] = { 611 /* PSP variant for packet where SRH with SL = 1 */ 612 tbl_cfg(PSP, SL_ONE, F_PSP), 613 /* End for packet where the SRH with SL > 1*/ 614 tbl_cfg(END, SL_MORE, F_PSP), 615 }; 616 617 #undef F_PSP 618 #undef tbl_cfg 619 620 /* For each flavor defined in RFC8986 (or a combination of them) an action is 621 * performed on the packet. The specific action depends on: 622 * - info extracted from the packet (i.e. pktinfo data) regarding the 623 * lack/presence of the SRH, and if the SRH is available, on the value of 624 * Segment Left field; 625 * - the mask of flavors configured for the specific SRv6 End* behavior. 626 * 627 * The function combines both the pkinfo and the flavors mask to evaluate the 628 * corresponding action to be taken on the packet. 629 */ 630 static enum seg6_local_flv_action 631 seg6_local_flv8986_act_lookup(enum seg6_local_pktinfo pinfo, __u32 flvmask) 632 { 633 unsigned long index; 634 635 /* check if the provided mask of flavors is supported */ 636 if (unlikely(flvmask & ~SEG6_LOCAL_FLV8986_SUPP_OPS)) 637 return SEG6_LOCAL_FLV_ACT_UNSPEC; 638 639 index = flv8986_act_tbl_idx(pinfo, flvmask); 640 if (unlikely(index >= FLV8986_ACT_TBL_SIZE)) 641 return SEG6_LOCAL_FLV_ACT_UNSPEC; 642 643 return flv8986_act_tbl[index]; 644 } 645 646 /* skb->data must be aligned with skb->network_header */ 647 static bool seg6_pop_srh(struct sk_buff *skb, int srhoff) 648 { 649 struct ipv6_sr_hdr *srh; 650 struct ipv6hdr *iph; 651 __u8 srh_nexthdr; 652 int thoff = -1; 653 int srhlen; 654 int nhlen; 655 656 if (unlikely(srhoff < sizeof(*iph) || 657 !pskb_may_pull(skb, srhoff + sizeof(*srh)))) 658 return false; 659 660 srh = (struct ipv6_sr_hdr *)(skb->data + srhoff); 661 srhlen = ipv6_optlen(srh); 662 663 /* we are about to mangle the pkt, let's check if we can write on it */ 664 if (unlikely(skb_ensure_writable(skb, srhoff + srhlen))) 665 return false; 666 667 /* skb_ensure_writable() may change skb pointers; evaluate srh again */ 668 srh = (struct ipv6_sr_hdr *)(skb->data + srhoff); 669 srh_nexthdr = srh->nexthdr; 670 671 if (unlikely(!skb_transport_header_was_set(skb))) 672 goto pull; 673 674 nhlen = skb_network_header_len(skb); 675 /* we have to deal with the transport header: it could be set before 676 * the SRH, after the SRH, or within it (which is considered wrong, 677 * however). 678 */ 679 if (likely(nhlen <= srhoff)) 680 thoff = nhlen; 681 else if (nhlen >= srhoff + srhlen) 682 /* transport_header is set after the SRH */ 683 thoff = nhlen - srhlen; 684 else 685 /* transport_header falls inside the SRH; hence, we can't 686 * restore the transport_header pointer properly after 687 * SRH removing operation. 688 */ 689 return false; 690 pull: 691 /* we need to pop the SRH: 692 * 1) first of all, we pull out everything from IPv6 header up to SRH 693 * (included) evaluating also the rcsum; 694 * 2) we overwrite (and then remove) the SRH by properly moving the 695 * IPv6 along with any extension header that precedes the SRH; 696 * 3) At the end, we push back the pulled headers (except for SRH, 697 * obviously). 698 */ 699 skb_pull_rcsum(skb, srhoff + srhlen); 700 memmove(skb_network_header(skb) + srhlen, skb_network_header(skb), 701 srhoff); 702 skb_push(skb, srhoff); 703 704 skb_reset_network_header(skb); 705 skb_mac_header_rebuild(skb); 706 if (likely(thoff >= 0)) 707 skb_set_transport_header(skb, thoff); 708 709 iph = ipv6_hdr(skb); 710 if (iph->nexthdr == NEXTHDR_ROUTING) { 711 iph->nexthdr = srh_nexthdr; 712 } else { 713 /* we must look for the extension header (EXTH, for short) that 714 * immediately precedes the SRH we have just removed. 715 * Then, we update the value of the EXTH nexthdr with the one 716 * contained in the SRH nexthdr. 717 */ 718 unsigned int off = sizeof(*iph); 719 struct ipv6_opt_hdr *hp, _hdr; 720 __u8 nexthdr = iph->nexthdr; 721 722 for (;;) { 723 if (unlikely(!ipv6_ext_hdr(nexthdr) || 724 nexthdr == NEXTHDR_NONE)) 725 return false; 726 727 hp = skb_header_pointer(skb, off, sizeof(_hdr), &_hdr); 728 if (unlikely(!hp)) 729 return false; 730 731 if (hp->nexthdr == NEXTHDR_ROUTING) { 732 hp->nexthdr = srh_nexthdr; 733 break; 734 } 735 736 switch (nexthdr) { 737 case NEXTHDR_FRAGMENT: 738 fallthrough; 739 case NEXTHDR_AUTH: 740 /* we expect SRH before FRAG and AUTH */ 741 return false; 742 default: 743 off += ipv6_optlen(hp); 744 break; 745 } 746 747 nexthdr = hp->nexthdr; 748 } 749 } 750 751 iph->payload_len = htons(skb->len - sizeof(struct ipv6hdr)); 752 753 skb_postpush_rcsum(skb, iph, srhoff); 754 755 return true; 756 } 757 758 /* process the packet on the basis of the RFC8986 flavors set for the given 759 * SRv6 End behavior instance. 760 */ 761 static int end_flv8986_core(struct sk_buff *skb, struct seg6_local_lwt *slwt) 762 { 763 const struct seg6_flavors_info *finfo = &slwt->flv_info; 764 enum seg6_local_flv_action action; 765 enum seg6_local_pktinfo pinfo; 766 struct ipv6_sr_hdr *srh; 767 __u32 flvmask; 768 int srhoff; 769 770 srh = seg6_get_srh(skb, 0); 771 srhoff = srh ? ((unsigned char *)srh - skb->data) : 0; 772 pinfo = seg6_get_srh_pktinfo(srh); 773 #ifdef CONFIG_IPV6_SEG6_HMAC 774 if (srh && !seg6_hmac_validate_skb(skb)) 775 goto drop; 776 #endif 777 flvmask = finfo->flv_ops; 778 if (unlikely(flvmask & ~SEG6_LOCAL_FLV8986_SUPP_OPS)) { 779 pr_warn_once("seg6local: invalid RFC8986 flavors\n"); 780 goto drop; 781 } 782 783 /* retrieve the action triggered by the combination of pktinfo data and 784 * the flavors mask. 785 */ 786 action = seg6_local_flv8986_act_lookup(pinfo, flvmask); 787 switch (action) { 788 case SEG6_LOCAL_FLV_ACT_END: 789 /* process the packet as the "standard" End behavior */ 790 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 791 break; 792 case SEG6_LOCAL_FLV_ACT_PSP: 793 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 794 795 if (unlikely(!seg6_pop_srh(skb, srhoff))) 796 goto drop; 797 break; 798 case SEG6_LOCAL_FLV_ACT_UNSPEC: 799 fallthrough; 800 default: 801 /* by default, we drop the packet since we could not find a 802 * suitable action. 803 */ 804 goto drop; 805 } 806 807 return input_action_end_finish(skb, slwt); 808 809 drop: 810 kfree_skb(skb); 811 return -EINVAL; 812 } 813 814 /* regular endpoint function */ 815 static int input_action_end(struct sk_buff *skb, struct seg6_local_lwt *slwt) 816 { 817 const struct seg6_flavors_info *finfo = &slwt->flv_info; 818 __u32 fops = finfo->flv_ops; 819 820 if (!fops) 821 return input_action_end_core(skb, slwt); 822 823 /* check for the presence of NEXT-C-SID since it applies first */ 824 if (seg6_next_csid_enabled(fops)) 825 return end_next_csid_core(skb, slwt); 826 827 /* the specific processing function to be performed on the packet 828 * depends on the combination of flavors defined in RFC8986 and some 829 * information extracted from the packet, e.g. presence/absence of SRH, 830 * Segment Left = 0, etc. 831 */ 832 return end_flv8986_core(skb, slwt); 833 } 834 835 /* regular endpoint, and forward to specified nexthop */ 836 static int input_action_end_x(struct sk_buff *skb, struct seg6_local_lwt *slwt) 837 { 838 const struct seg6_flavors_info *finfo = &slwt->flv_info; 839 __u32 fops = finfo->flv_ops; 840 841 /* check for the presence of NEXT-C-SID since it applies first */ 842 if (seg6_next_csid_enabled(fops)) 843 return end_x_next_csid_core(skb, slwt); 844 845 return input_action_end_x_core(skb, slwt); 846 } 847 848 static int input_action_end_t(struct sk_buff *skb, struct seg6_local_lwt *slwt) 849 { 850 struct ipv6_sr_hdr *srh; 851 852 srh = get_and_validate_srh(skb); 853 if (!srh) 854 goto drop; 855 856 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 857 858 seg6_lookup_nexthop(skb, NULL, slwt->table); 859 860 return dst_input(skb); 861 862 drop: 863 kfree_skb(skb); 864 return -EINVAL; 865 } 866 867 /* decapsulate and forward inner L2 frame on specified interface */ 868 static int input_action_end_dx2(struct sk_buff *skb, 869 struct seg6_local_lwt *slwt) 870 { 871 struct net *net = dev_net(skb->dev); 872 struct net_device *odev; 873 struct ethhdr *eth; 874 875 if (!decap_and_validate(skb, IPPROTO_ETHERNET)) 876 goto drop; 877 878 if (!pskb_may_pull(skb, ETH_HLEN)) 879 goto drop; 880 881 skb_reset_mac_header(skb); 882 eth = (struct ethhdr *)skb->data; 883 884 /* To determine the frame's protocol, we assume it is 802.3. This avoids 885 * a call to eth_type_trans(), which is not really relevant for our 886 * use case. 887 */ 888 if (!eth_proto_is_802_3(eth->h_proto)) 889 goto drop; 890 891 odev = dev_get_by_index_rcu(net, slwt->oif); 892 if (!odev) 893 goto drop; 894 895 /* As we accept Ethernet frames, make sure the egress device is of 896 * the correct type. 897 */ 898 if (odev->type != ARPHRD_ETHER) 899 goto drop; 900 901 if (!(odev->flags & IFF_UP) || !netif_carrier_ok(odev)) 902 goto drop; 903 904 skb_orphan(skb); 905 906 if (skb_warn_if_lro(skb)) 907 goto drop; 908 909 skb_forward_csum(skb); 910 911 if (skb->len - ETH_HLEN > odev->mtu) 912 goto drop; 913 914 skb->dev = odev; 915 skb->protocol = eth->h_proto; 916 917 return dev_queue_xmit(skb); 918 919 drop: 920 kfree_skb(skb); 921 return -EINVAL; 922 } 923 924 static int input_action_end_dx6_finish(struct net *net, struct sock *sk, 925 struct sk_buff *skb) 926 { 927 struct dst_entry *orig_dst = skb_dst(skb); 928 struct in6_addr *nhaddr = NULL; 929 struct seg6_local_lwt *slwt; 930 931 slwt = seg6_local_lwtunnel(orig_dst->lwtstate); 932 933 /* The inner packet is not associated to any local interface, 934 * so we do not call netif_rx(). 935 * 936 * If slwt->nh6 is set to ::, then lookup the nexthop for the 937 * inner packet's DA. Otherwise, use the specified nexthop. 938 */ 939 if (!ipv6_addr_any(&slwt->nh6)) 940 nhaddr = &slwt->nh6; 941 942 seg6_lookup_nexthop(skb, nhaddr, 0); 943 944 return dst_input(skb); 945 } 946 947 /* decapsulate and forward to specified nexthop */ 948 static int input_action_end_dx6(struct sk_buff *skb, 949 struct seg6_local_lwt *slwt) 950 { 951 /* this function accepts IPv6 encapsulated packets, with either 952 * an SRH with SL=0, or no SRH. 953 */ 954 955 if (!decap_and_validate(skb, IPPROTO_IPV6)) 956 goto drop; 957 958 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 959 goto drop; 960 961 skb_set_transport_header(skb, sizeof(struct ipv6hdr)); 962 nf_reset_ct(skb); 963 964 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled)) 965 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, 966 dev_net(skb->dev), NULL, skb, skb->dev, 967 NULL, input_action_end_dx6_finish); 968 969 return input_action_end_dx6_finish(dev_net(skb->dev), NULL, skb); 970 drop: 971 kfree_skb(skb); 972 return -EINVAL; 973 } 974 975 static int input_action_end_dx4_finish(struct net *net, struct sock *sk, 976 struct sk_buff *skb) 977 { 978 struct dst_entry *orig_dst = skb_dst(skb); 979 enum skb_drop_reason reason; 980 struct seg6_local_lwt *slwt; 981 struct iphdr *iph; 982 __be32 nhaddr; 983 984 slwt = seg6_local_lwtunnel(orig_dst->lwtstate); 985 986 iph = ip_hdr(skb); 987 988 nhaddr = slwt->nh4.s_addr ?: iph->daddr; 989 990 skb_dst_drop(skb); 991 992 reason = ip_route_input(skb, nhaddr, iph->saddr, 0, skb->dev); 993 if (reason) { 994 kfree_skb_reason(skb, reason); 995 return -EINVAL; 996 } 997 998 return dst_input(skb); 999 } 1000 1001 static int input_action_end_dx4(struct sk_buff *skb, 1002 struct seg6_local_lwt *slwt) 1003 { 1004 if (!decap_and_validate(skb, IPPROTO_IPIP)) 1005 goto drop; 1006 1007 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 1008 goto drop; 1009 1010 skb->protocol = htons(ETH_P_IP); 1011 skb_set_transport_header(skb, sizeof(struct iphdr)); 1012 nf_reset_ct(skb); 1013 1014 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled)) 1015 return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, 1016 dev_net(skb->dev), NULL, skb, skb->dev, 1017 NULL, input_action_end_dx4_finish); 1018 1019 return input_action_end_dx4_finish(dev_net(skb->dev), NULL, skb); 1020 drop: 1021 kfree_skb(skb); 1022 return -EINVAL; 1023 } 1024 1025 #ifdef CONFIG_NET_L3_MASTER_DEV 1026 static struct net *fib6_config_get_net(const struct fib6_config *fib6_cfg) 1027 { 1028 const struct nl_info *nli = &fib6_cfg->fc_nlinfo; 1029 1030 return nli->nl_net; 1031 } 1032 1033 static int __seg6_end_dt_vrf_build(struct seg6_local_lwt *slwt, const void *cfg, 1034 u16 family, struct netlink_ext_ack *extack) 1035 { 1036 struct seg6_end_dt_info *info = &slwt->dt_info; 1037 int vrf_ifindex; 1038 struct net *net; 1039 1040 net = fib6_config_get_net(cfg); 1041 1042 /* note that vrf_table was already set by parse_nla_vrftable() */ 1043 vrf_ifindex = l3mdev_ifindex_lookup_by_table_id(L3MDEV_TYPE_VRF, net, 1044 info->vrf_table); 1045 if (vrf_ifindex < 0) { 1046 if (vrf_ifindex == -EPERM) { 1047 NL_SET_ERR_MSG(extack, 1048 "Strict mode for VRF is disabled"); 1049 } else if (vrf_ifindex == -ENODEV) { 1050 NL_SET_ERR_MSG(extack, 1051 "Table has no associated VRF device"); 1052 } else { 1053 pr_debug("seg6local: SRv6 End.DT* creation error=%d\n", 1054 vrf_ifindex); 1055 } 1056 1057 return vrf_ifindex; 1058 } 1059 1060 info->net = net; 1061 info->vrf_ifindex = vrf_ifindex; 1062 1063 info->family = family; 1064 info->mode = DT_VRF_MODE; 1065 1066 return 0; 1067 } 1068 1069 /* The SRv6 End.DT4/DT6 behavior extracts the inner (IPv4/IPv6) packet and 1070 * routes the IPv4/IPv6 packet by looking at the configured routing table. 1071 * 1072 * In the SRv6 End.DT4/DT6 use case, we can receive traffic (IPv6+Segment 1073 * Routing Header packets) from several interfaces and the outer IPv6 1074 * destination address (DA) is used for retrieving the specific instance of the 1075 * End.DT4/DT6 behavior that should process the packets. 1076 * 1077 * However, the inner IPv4/IPv6 packet is not really bound to any receiving 1078 * interface and thus the End.DT4/DT6 sets the VRF (associated with the 1079 * corresponding routing table) as the *receiving* interface. 1080 * In other words, the End.DT4/DT6 processes a packet as if it has been received 1081 * directly by the VRF (and not by one of its slave devices, if any). 1082 * In this way, the VRF interface is used for routing the IPv4/IPv6 packet in 1083 * according to the routing table configured by the End.DT4/DT6 instance. 1084 * 1085 * This design allows you to get some interesting features like: 1086 * 1) the statistics on rx packets; 1087 * 2) the possibility to install a packet sniffer on the receiving interface 1088 * (the VRF one) for looking at the incoming packets; 1089 * 3) the possibility to leverage the netfilter prerouting hook for the inner 1090 * IPv4 packet. 1091 * 1092 * This function returns: 1093 * - the sk_buff* when the VRF rcv handler has processed the packet correctly; 1094 * - NULL when the skb is consumed by the VRF rcv handler; 1095 * - a pointer which encodes a negative error number in case of error. 1096 * Note that in this case, the function takes care of freeing the skb. 1097 */ 1098 static struct sk_buff *end_dt_vrf_rcv(struct sk_buff *skb, u16 family, 1099 struct net_device *dev) 1100 { 1101 /* based on l3mdev_ip_rcv; we are only interested in the master */ 1102 if (unlikely(!netif_is_l3_master(dev) && !netif_has_l3_rx_handler(dev))) 1103 goto drop; 1104 1105 if (unlikely(!dev->l3mdev_ops->l3mdev_l3_rcv)) 1106 goto drop; 1107 1108 /* the decap packet IPv4/IPv6 does not come with any mac header info. 1109 * We must unset the mac header to allow the VRF device to rebuild it, 1110 * just in case there is a sniffer attached on the device. 1111 */ 1112 skb_unset_mac_header(skb); 1113 1114 skb = dev->l3mdev_ops->l3mdev_l3_rcv(dev, skb, family); 1115 if (!skb) 1116 /* the skb buffer was consumed by the handler */ 1117 return NULL; 1118 1119 /* when a packet is received by a VRF or by one of its slaves, the 1120 * master device reference is set into the skb. 1121 */ 1122 if (unlikely(skb->dev != dev || skb->skb_iif != dev->ifindex)) 1123 goto drop; 1124 1125 return skb; 1126 1127 drop: 1128 kfree_skb(skb); 1129 return ERR_PTR(-EINVAL); 1130 } 1131 1132 static struct net_device *end_dt_get_vrf_rcu(struct sk_buff *skb, 1133 struct seg6_end_dt_info *info) 1134 { 1135 int vrf_ifindex = info->vrf_ifindex; 1136 struct net *net = info->net; 1137 1138 if (unlikely(vrf_ifindex < 0)) 1139 goto error; 1140 1141 if (unlikely(!net_eq(dev_net(skb->dev), net))) 1142 goto error; 1143 1144 return dev_get_by_index_rcu(net, vrf_ifindex); 1145 1146 error: 1147 return NULL; 1148 } 1149 1150 static struct sk_buff *end_dt_vrf_core(struct sk_buff *skb, 1151 struct seg6_local_lwt *slwt, u16 family) 1152 { 1153 struct seg6_end_dt_info *info = &slwt->dt_info; 1154 struct net_device *vrf; 1155 __be16 protocol; 1156 int hdrlen; 1157 1158 vrf = end_dt_get_vrf_rcu(skb, info); 1159 if (unlikely(!vrf)) 1160 goto drop; 1161 1162 switch (family) { 1163 case AF_INET: 1164 protocol = htons(ETH_P_IP); 1165 hdrlen = sizeof(struct iphdr); 1166 break; 1167 case AF_INET6: 1168 protocol = htons(ETH_P_IPV6); 1169 hdrlen = sizeof(struct ipv6hdr); 1170 break; 1171 case AF_UNSPEC: 1172 fallthrough; 1173 default: 1174 goto drop; 1175 } 1176 1177 if (unlikely(info->family != AF_UNSPEC && info->family != family)) { 1178 pr_warn_once("seg6local: SRv6 End.DT* family mismatch"); 1179 goto drop; 1180 } 1181 1182 skb->protocol = protocol; 1183 1184 skb_dst_drop(skb); 1185 1186 skb_set_transport_header(skb, hdrlen); 1187 nf_reset_ct(skb); 1188 1189 return end_dt_vrf_rcv(skb, family, vrf); 1190 1191 drop: 1192 kfree_skb(skb); 1193 return ERR_PTR(-EINVAL); 1194 } 1195 1196 static int input_action_end_dt4(struct sk_buff *skb, 1197 struct seg6_local_lwt *slwt) 1198 { 1199 enum skb_drop_reason reason; 1200 struct iphdr *iph; 1201 1202 if (!decap_and_validate(skb, IPPROTO_IPIP)) 1203 goto drop; 1204 1205 if (!pskb_may_pull(skb, sizeof(struct iphdr))) 1206 goto drop; 1207 1208 skb = end_dt_vrf_core(skb, slwt, AF_INET); 1209 if (!skb) 1210 /* packet has been processed and consumed by the VRF */ 1211 return 0; 1212 1213 if (IS_ERR(skb)) 1214 return PTR_ERR(skb); 1215 1216 iph = ip_hdr(skb); 1217 1218 reason = ip_route_input(skb, iph->daddr, iph->saddr, 0, skb->dev); 1219 if (unlikely(reason)) 1220 goto drop; 1221 1222 return dst_input(skb); 1223 1224 drop: 1225 kfree_skb(skb); 1226 return -EINVAL; 1227 } 1228 1229 static int seg6_end_dt4_build(struct seg6_local_lwt *slwt, const void *cfg, 1230 struct netlink_ext_ack *extack) 1231 { 1232 return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET, extack); 1233 } 1234 1235 static enum 1236 seg6_end_dt_mode seg6_end_dt6_parse_mode(struct seg6_local_lwt *slwt) 1237 { 1238 unsigned long parsed_optattrs = slwt->parsed_optattrs; 1239 bool legacy, vrfmode; 1240 1241 legacy = !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE)); 1242 vrfmode = !!(parsed_optattrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE)); 1243 1244 if (!(legacy ^ vrfmode)) 1245 /* both are absent or present: invalid DT6 mode */ 1246 return DT_INVALID_MODE; 1247 1248 return legacy ? DT_LEGACY_MODE : DT_VRF_MODE; 1249 } 1250 1251 static enum seg6_end_dt_mode seg6_end_dt6_get_mode(struct seg6_local_lwt *slwt) 1252 { 1253 struct seg6_end_dt_info *info = &slwt->dt_info; 1254 1255 return info->mode; 1256 } 1257 1258 static int seg6_end_dt6_build(struct seg6_local_lwt *slwt, const void *cfg, 1259 struct netlink_ext_ack *extack) 1260 { 1261 enum seg6_end_dt_mode mode = seg6_end_dt6_parse_mode(slwt); 1262 struct seg6_end_dt_info *info = &slwt->dt_info; 1263 1264 switch (mode) { 1265 case DT_LEGACY_MODE: 1266 info->mode = DT_LEGACY_MODE; 1267 return 0; 1268 case DT_VRF_MODE: 1269 return __seg6_end_dt_vrf_build(slwt, cfg, AF_INET6, extack); 1270 default: 1271 NL_SET_ERR_MSG(extack, "table or vrftable must be specified"); 1272 return -EINVAL; 1273 } 1274 } 1275 #endif 1276 1277 static int input_action_end_dt6(struct sk_buff *skb, 1278 struct seg6_local_lwt *slwt) 1279 { 1280 if (!decap_and_validate(skb, IPPROTO_IPV6)) 1281 goto drop; 1282 1283 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 1284 goto drop; 1285 1286 #ifdef CONFIG_NET_L3_MASTER_DEV 1287 if (seg6_end_dt6_get_mode(slwt) == DT_LEGACY_MODE) 1288 goto legacy_mode; 1289 1290 /* DT6_VRF_MODE */ 1291 skb = end_dt_vrf_core(skb, slwt, AF_INET6); 1292 if (!skb) 1293 /* packet has been processed and consumed by the VRF */ 1294 return 0; 1295 1296 if (IS_ERR(skb)) 1297 return PTR_ERR(skb); 1298 1299 /* note: this time we do not need to specify the table because the VRF 1300 * takes care of selecting the correct table. 1301 */ 1302 seg6_lookup_any_nexthop(skb, NULL, 0, true, 0); 1303 1304 return dst_input(skb); 1305 1306 legacy_mode: 1307 #endif 1308 skb_set_transport_header(skb, sizeof(struct ipv6hdr)); 1309 1310 seg6_lookup_any_nexthop(skb, NULL, slwt->table, true, 0); 1311 1312 return dst_input(skb); 1313 1314 drop: 1315 kfree_skb(skb); 1316 return -EINVAL; 1317 } 1318 1319 #ifdef CONFIG_NET_L3_MASTER_DEV 1320 static int seg6_end_dt46_build(struct seg6_local_lwt *slwt, const void *cfg, 1321 struct netlink_ext_ack *extack) 1322 { 1323 return __seg6_end_dt_vrf_build(slwt, cfg, AF_UNSPEC, extack); 1324 } 1325 1326 static int input_action_end_dt46(struct sk_buff *skb, 1327 struct seg6_local_lwt *slwt) 1328 { 1329 unsigned int off = 0; 1330 int nexthdr; 1331 1332 nexthdr = ipv6_find_hdr(skb, &off, -1, NULL, NULL); 1333 if (unlikely(nexthdr < 0)) 1334 goto drop; 1335 1336 switch (nexthdr) { 1337 case IPPROTO_IPIP: 1338 return input_action_end_dt4(skb, slwt); 1339 case IPPROTO_IPV6: 1340 return input_action_end_dt6(skb, slwt); 1341 } 1342 1343 drop: 1344 kfree_skb(skb); 1345 return -EINVAL; 1346 } 1347 #endif 1348 1349 /* push an SRH on top of the current one */ 1350 static int input_action_end_b6(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1351 { 1352 struct ipv6_sr_hdr *srh; 1353 int err = -EINVAL; 1354 1355 srh = get_and_validate_srh(skb); 1356 if (!srh) 1357 goto drop; 1358 1359 err = seg6_do_srh_inline(skb, slwt->srh); 1360 if (err) 1361 goto drop; 1362 1363 skb_set_transport_header(skb, sizeof(struct ipv6hdr)); 1364 1365 seg6_lookup_nexthop(skb, NULL, 0); 1366 1367 return dst_input(skb); 1368 1369 drop: 1370 kfree_skb(skb); 1371 return err; 1372 } 1373 1374 /* encapsulate within an outer IPv6 header and a specified SRH */ 1375 static int input_action_end_b6_encap(struct sk_buff *skb, 1376 struct seg6_local_lwt *slwt) 1377 { 1378 struct ipv6_sr_hdr *srh; 1379 int err = -EINVAL; 1380 1381 srh = get_and_validate_srh(skb); 1382 if (!srh) 1383 goto drop; 1384 1385 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 1386 1387 skb_reset_inner_headers(skb); 1388 skb->encapsulation = 1; 1389 1390 err = seg6_do_srh_encap(skb, slwt->srh, IPPROTO_IPV6); 1391 if (err) 1392 goto drop; 1393 1394 skb_set_transport_header(skb, sizeof(struct ipv6hdr)); 1395 1396 seg6_lookup_nexthop(skb, NULL, 0); 1397 1398 return dst_input(skb); 1399 1400 drop: 1401 kfree_skb(skb); 1402 return err; 1403 } 1404 1405 DEFINE_PER_CPU(struct seg6_bpf_srh_state, seg6_bpf_srh_states) = { 1406 .bh_lock = INIT_LOCAL_LOCK(bh_lock), 1407 }; 1408 1409 bool seg6_bpf_has_valid_srh(struct sk_buff *skb) 1410 { 1411 struct seg6_bpf_srh_state *srh_state = 1412 this_cpu_ptr(&seg6_bpf_srh_states); 1413 struct ipv6_sr_hdr *srh = srh_state->srh; 1414 1415 lockdep_assert_held(&srh_state->bh_lock); 1416 if (unlikely(srh == NULL)) 1417 return false; 1418 1419 if (unlikely(!srh_state->valid)) { 1420 if ((srh_state->hdrlen & 7) != 0) 1421 return false; 1422 1423 srh->hdrlen = (u8)(srh_state->hdrlen >> 3); 1424 if (!seg6_validate_srh(srh, (srh->hdrlen + 1) << 3, true)) 1425 return false; 1426 1427 srh_state->valid = true; 1428 } 1429 1430 return true; 1431 } 1432 1433 static int input_action_end_bpf(struct sk_buff *skb, 1434 struct seg6_local_lwt *slwt) 1435 { 1436 struct seg6_bpf_srh_state *srh_state; 1437 struct ipv6_sr_hdr *srh; 1438 int ret; 1439 1440 srh = get_and_validate_srh(skb); 1441 if (!srh) { 1442 kfree_skb(skb); 1443 return -EINVAL; 1444 } 1445 advance_nextseg(srh, &ipv6_hdr(skb)->daddr); 1446 1447 /* The access to the per-CPU buffer srh_state is protected by running 1448 * always in softirq context (with disabled BH). On PREEMPT_RT the 1449 * required locking is provided by the following local_lock_nested_bh() 1450 * statement. It is also accessed by the bpf_lwt_seg6_* helpers via 1451 * bpf_prog_run_save_cb(). 1452 */ 1453 local_lock_nested_bh(&seg6_bpf_srh_states.bh_lock); 1454 srh_state = this_cpu_ptr(&seg6_bpf_srh_states); 1455 srh_state->srh = srh; 1456 srh_state->hdrlen = srh->hdrlen << 3; 1457 srh_state->valid = true; 1458 1459 rcu_read_lock(); 1460 bpf_compute_data_pointers(skb); 1461 ret = bpf_prog_run_save_cb(slwt->bpf.prog, skb); 1462 rcu_read_unlock(); 1463 1464 switch (ret) { 1465 case BPF_OK: 1466 case BPF_REDIRECT: 1467 break; 1468 case BPF_DROP: 1469 goto drop; 1470 default: 1471 pr_warn_once("bpf-seg6local: Illegal return value %u\n", ret); 1472 goto drop; 1473 } 1474 1475 if (srh_state->srh && !seg6_bpf_has_valid_srh(skb)) 1476 goto drop; 1477 local_unlock_nested_bh(&seg6_bpf_srh_states.bh_lock); 1478 1479 if (ret != BPF_REDIRECT) 1480 seg6_lookup_nexthop(skb, NULL, 0); 1481 1482 return dst_input(skb); 1483 1484 drop: 1485 local_unlock_nested_bh(&seg6_bpf_srh_states.bh_lock); 1486 kfree_skb(skb); 1487 return -EINVAL; 1488 } 1489 1490 static struct seg6_action_desc seg6_action_table[] = { 1491 { 1492 .action = SEG6_LOCAL_ACTION_END, 1493 .attrs = 0, 1494 .optattrs = SEG6_F_LOCAL_COUNTERS | 1495 SEG6_F_LOCAL_FLAVORS, 1496 .input = input_action_end, 1497 }, 1498 { 1499 .action = SEG6_LOCAL_ACTION_END_X, 1500 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH6), 1501 .optattrs = SEG6_F_LOCAL_COUNTERS | 1502 SEG6_F_LOCAL_FLAVORS | 1503 SEG6_F_ATTR(SEG6_LOCAL_OIF), 1504 .input = input_action_end_x, 1505 }, 1506 { 1507 .action = SEG6_LOCAL_ACTION_END_T, 1508 .attrs = SEG6_F_ATTR(SEG6_LOCAL_TABLE), 1509 .optattrs = SEG6_F_LOCAL_COUNTERS, 1510 .input = input_action_end_t, 1511 }, 1512 { 1513 .action = SEG6_LOCAL_ACTION_END_DX2, 1514 .attrs = SEG6_F_ATTR(SEG6_LOCAL_OIF), 1515 .optattrs = SEG6_F_LOCAL_COUNTERS, 1516 .input = input_action_end_dx2, 1517 }, 1518 { 1519 .action = SEG6_LOCAL_ACTION_END_DX6, 1520 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH6), 1521 .optattrs = SEG6_F_LOCAL_COUNTERS, 1522 .input = input_action_end_dx6, 1523 }, 1524 { 1525 .action = SEG6_LOCAL_ACTION_END_DX4, 1526 .attrs = SEG6_F_ATTR(SEG6_LOCAL_NH4), 1527 .optattrs = SEG6_F_LOCAL_COUNTERS, 1528 .input = input_action_end_dx4, 1529 }, 1530 { 1531 .action = SEG6_LOCAL_ACTION_END_DT4, 1532 .attrs = SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE), 1533 .optattrs = SEG6_F_LOCAL_COUNTERS, 1534 #ifdef CONFIG_NET_L3_MASTER_DEV 1535 .input = input_action_end_dt4, 1536 .slwt_ops = { 1537 .build_state = seg6_end_dt4_build, 1538 }, 1539 #endif 1540 }, 1541 { 1542 .action = SEG6_LOCAL_ACTION_END_DT6, 1543 #ifdef CONFIG_NET_L3_MASTER_DEV 1544 .attrs = 0, 1545 .optattrs = SEG6_F_LOCAL_COUNTERS | 1546 SEG6_F_ATTR(SEG6_LOCAL_TABLE) | 1547 SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE), 1548 .slwt_ops = { 1549 .build_state = seg6_end_dt6_build, 1550 }, 1551 #else 1552 .attrs = SEG6_F_ATTR(SEG6_LOCAL_TABLE), 1553 .optattrs = SEG6_F_LOCAL_COUNTERS, 1554 #endif 1555 .input = input_action_end_dt6, 1556 }, 1557 { 1558 .action = SEG6_LOCAL_ACTION_END_DT46, 1559 .attrs = SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE), 1560 .optattrs = SEG6_F_LOCAL_COUNTERS, 1561 #ifdef CONFIG_NET_L3_MASTER_DEV 1562 .input = input_action_end_dt46, 1563 .slwt_ops = { 1564 .build_state = seg6_end_dt46_build, 1565 }, 1566 #endif 1567 }, 1568 { 1569 .action = SEG6_LOCAL_ACTION_END_B6, 1570 .attrs = SEG6_F_ATTR(SEG6_LOCAL_SRH), 1571 .optattrs = SEG6_F_LOCAL_COUNTERS, 1572 .input = input_action_end_b6, 1573 }, 1574 { 1575 .action = SEG6_LOCAL_ACTION_END_B6_ENCAP, 1576 .attrs = SEG6_F_ATTR(SEG6_LOCAL_SRH), 1577 .optattrs = SEG6_F_LOCAL_COUNTERS, 1578 .input = input_action_end_b6_encap, 1579 .static_headroom = sizeof(struct ipv6hdr), 1580 }, 1581 { 1582 .action = SEG6_LOCAL_ACTION_END_BPF, 1583 .attrs = SEG6_F_ATTR(SEG6_LOCAL_BPF), 1584 .optattrs = SEG6_F_LOCAL_COUNTERS, 1585 .input = input_action_end_bpf, 1586 }, 1587 1588 }; 1589 1590 static struct seg6_action_desc *__get_action_desc(int action) 1591 { 1592 struct seg6_action_desc *desc; 1593 int i, count; 1594 1595 count = ARRAY_SIZE(seg6_action_table); 1596 for (i = 0; i < count; i++) { 1597 desc = &seg6_action_table[i]; 1598 if (desc->action == action) 1599 return desc; 1600 } 1601 1602 return NULL; 1603 } 1604 1605 static bool seg6_lwtunnel_counters_enabled(struct seg6_local_lwt *slwt) 1606 { 1607 return slwt->parsed_optattrs & SEG6_F_LOCAL_COUNTERS; 1608 } 1609 1610 static void seg6_local_update_counters(struct seg6_local_lwt *slwt, 1611 unsigned int len, int err) 1612 { 1613 struct pcpu_seg6_local_counters *pcounters; 1614 1615 pcounters = this_cpu_ptr(slwt->pcpu_counters); 1616 u64_stats_update_begin(&pcounters->syncp); 1617 1618 if (likely(!err)) { 1619 u64_stats_inc(&pcounters->packets); 1620 u64_stats_add(&pcounters->bytes, len); 1621 } else { 1622 u64_stats_inc(&pcounters->errors); 1623 } 1624 1625 u64_stats_update_end(&pcounters->syncp); 1626 } 1627 1628 static int seg6_local_input_core(struct net *net, struct sock *sk, 1629 struct sk_buff *skb) 1630 { 1631 struct dst_entry *orig_dst = skb_dst(skb); 1632 struct seg6_action_desc *desc; 1633 struct seg6_local_lwt *slwt; 1634 unsigned int len = skb->len; 1635 int rc; 1636 1637 slwt = seg6_local_lwtunnel(orig_dst->lwtstate); 1638 desc = slwt->desc; 1639 1640 rc = desc->input(skb, slwt); 1641 1642 if (!seg6_lwtunnel_counters_enabled(slwt)) 1643 return rc; 1644 1645 seg6_local_update_counters(slwt, len, rc); 1646 1647 return rc; 1648 } 1649 1650 static int seg6_local_input(struct sk_buff *skb) 1651 { 1652 if (skb->protocol != htons(ETH_P_IPV6)) { 1653 kfree_skb(skb); 1654 return -EINVAL; 1655 } 1656 1657 if (static_branch_unlikely(&nf_hooks_lwtunnel_enabled)) 1658 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN, 1659 dev_net(skb->dev), NULL, skb, skb->dev, NULL, 1660 seg6_local_input_core); 1661 1662 return seg6_local_input_core(dev_net(skb->dev), NULL, skb); 1663 } 1664 1665 static const struct nla_policy seg6_local_policy[SEG6_LOCAL_MAX + 1] = { 1666 [SEG6_LOCAL_ACTION] = { .type = NLA_U32 }, 1667 [SEG6_LOCAL_SRH] = { .type = NLA_BINARY }, 1668 [SEG6_LOCAL_TABLE] = { .type = NLA_U32 }, 1669 [SEG6_LOCAL_VRFTABLE] = { .type = NLA_U32 }, 1670 [SEG6_LOCAL_NH4] = NLA_POLICY_EXACT_LEN(sizeof(struct in_addr)), 1671 [SEG6_LOCAL_NH6] = NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)), 1672 [SEG6_LOCAL_IIF] = { .type = NLA_U32 }, 1673 [SEG6_LOCAL_OIF] = { .type = NLA_U32 }, 1674 [SEG6_LOCAL_BPF] = { .type = NLA_NESTED }, 1675 [SEG6_LOCAL_COUNTERS] = { .type = NLA_NESTED }, 1676 [SEG6_LOCAL_FLAVORS] = { .type = NLA_NESTED }, 1677 }; 1678 1679 static int parse_nla_srh(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1680 struct netlink_ext_ack *extack) 1681 { 1682 struct ipv6_sr_hdr *srh; 1683 int len; 1684 1685 srh = nla_data(attrs[SEG6_LOCAL_SRH]); 1686 len = nla_len(attrs[SEG6_LOCAL_SRH]); 1687 1688 /* SRH must contain at least one segment */ 1689 if (len < sizeof(*srh) + sizeof(struct in6_addr)) 1690 return -EINVAL; 1691 1692 if (!seg6_validate_srh(srh, len, false)) 1693 return -EINVAL; 1694 1695 slwt->srh = kmemdup(srh, len, GFP_KERNEL); 1696 if (!slwt->srh) 1697 return -ENOMEM; 1698 1699 slwt->headroom += len; 1700 1701 return 0; 1702 } 1703 1704 static int put_nla_srh(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1705 { 1706 struct ipv6_sr_hdr *srh; 1707 struct nlattr *nla; 1708 int len; 1709 1710 srh = slwt->srh; 1711 len = (srh->hdrlen + 1) << 3; 1712 1713 nla = nla_reserve(skb, SEG6_LOCAL_SRH, len); 1714 if (!nla) 1715 return -EMSGSIZE; 1716 1717 memcpy(nla_data(nla), srh, len); 1718 1719 return 0; 1720 } 1721 1722 static int cmp_nla_srh(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1723 { 1724 int len = (a->srh->hdrlen + 1) << 3; 1725 1726 if (len != ((b->srh->hdrlen + 1) << 3)) 1727 return 1; 1728 1729 return memcmp(a->srh, b->srh, len); 1730 } 1731 1732 static void destroy_attr_srh(struct seg6_local_lwt *slwt) 1733 { 1734 kfree(slwt->srh); 1735 } 1736 1737 static int parse_nla_table(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1738 struct netlink_ext_ack *extack) 1739 { 1740 slwt->table = nla_get_u32(attrs[SEG6_LOCAL_TABLE]); 1741 1742 return 0; 1743 } 1744 1745 static int put_nla_table(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1746 { 1747 if (nla_put_u32(skb, SEG6_LOCAL_TABLE, slwt->table)) 1748 return -EMSGSIZE; 1749 1750 return 0; 1751 } 1752 1753 static int cmp_nla_table(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1754 { 1755 if (a->table != b->table) 1756 return 1; 1757 1758 return 0; 1759 } 1760 1761 static struct 1762 seg6_end_dt_info *seg6_possible_end_dt_info(struct seg6_local_lwt *slwt) 1763 { 1764 #ifdef CONFIG_NET_L3_MASTER_DEV 1765 return &slwt->dt_info; 1766 #else 1767 return ERR_PTR(-EOPNOTSUPP); 1768 #endif 1769 } 1770 1771 static int parse_nla_vrftable(struct nlattr **attrs, 1772 struct seg6_local_lwt *slwt, 1773 struct netlink_ext_ack *extack) 1774 { 1775 struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt); 1776 1777 if (IS_ERR(info)) 1778 return PTR_ERR(info); 1779 1780 info->vrf_table = nla_get_u32(attrs[SEG6_LOCAL_VRFTABLE]); 1781 1782 return 0; 1783 } 1784 1785 static int put_nla_vrftable(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1786 { 1787 struct seg6_end_dt_info *info = seg6_possible_end_dt_info(slwt); 1788 1789 if (IS_ERR(info)) 1790 return PTR_ERR(info); 1791 1792 if (nla_put_u32(skb, SEG6_LOCAL_VRFTABLE, info->vrf_table)) 1793 return -EMSGSIZE; 1794 1795 return 0; 1796 } 1797 1798 static int cmp_nla_vrftable(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1799 { 1800 struct seg6_end_dt_info *info_a = seg6_possible_end_dt_info(a); 1801 struct seg6_end_dt_info *info_b = seg6_possible_end_dt_info(b); 1802 1803 if (info_a->vrf_table != info_b->vrf_table) 1804 return 1; 1805 1806 return 0; 1807 } 1808 1809 static int parse_nla_nh4(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1810 struct netlink_ext_ack *extack) 1811 { 1812 memcpy(&slwt->nh4, nla_data(attrs[SEG6_LOCAL_NH4]), 1813 sizeof(struct in_addr)); 1814 1815 return 0; 1816 } 1817 1818 static int put_nla_nh4(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1819 { 1820 struct nlattr *nla; 1821 1822 nla = nla_reserve(skb, SEG6_LOCAL_NH4, sizeof(struct in_addr)); 1823 if (!nla) 1824 return -EMSGSIZE; 1825 1826 memcpy(nla_data(nla), &slwt->nh4, sizeof(struct in_addr)); 1827 1828 return 0; 1829 } 1830 1831 static int cmp_nla_nh4(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1832 { 1833 return memcmp(&a->nh4, &b->nh4, sizeof(struct in_addr)); 1834 } 1835 1836 static int parse_nla_nh6(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1837 struct netlink_ext_ack *extack) 1838 { 1839 memcpy(&slwt->nh6, nla_data(attrs[SEG6_LOCAL_NH6]), 1840 sizeof(struct in6_addr)); 1841 1842 return 0; 1843 } 1844 1845 static int put_nla_nh6(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1846 { 1847 struct nlattr *nla; 1848 1849 nla = nla_reserve(skb, SEG6_LOCAL_NH6, sizeof(struct in6_addr)); 1850 if (!nla) 1851 return -EMSGSIZE; 1852 1853 memcpy(nla_data(nla), &slwt->nh6, sizeof(struct in6_addr)); 1854 1855 return 0; 1856 } 1857 1858 static int cmp_nla_nh6(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1859 { 1860 return memcmp(&a->nh6, &b->nh6, sizeof(struct in6_addr)); 1861 } 1862 1863 static int parse_nla_iif(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1864 struct netlink_ext_ack *extack) 1865 { 1866 slwt->iif = nla_get_u32(attrs[SEG6_LOCAL_IIF]); 1867 1868 return 0; 1869 } 1870 1871 static int put_nla_iif(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1872 { 1873 if (nla_put_u32(skb, SEG6_LOCAL_IIF, slwt->iif)) 1874 return -EMSGSIZE; 1875 1876 return 0; 1877 } 1878 1879 static int cmp_nla_iif(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1880 { 1881 if (a->iif != b->iif) 1882 return 1; 1883 1884 return 0; 1885 } 1886 1887 static int parse_nla_oif(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1888 struct netlink_ext_ack *extack) 1889 { 1890 slwt->oif = nla_get_u32(attrs[SEG6_LOCAL_OIF]); 1891 1892 return 0; 1893 } 1894 1895 static int put_nla_oif(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1896 { 1897 if (nla_put_u32(skb, SEG6_LOCAL_OIF, slwt->oif)) 1898 return -EMSGSIZE; 1899 1900 return 0; 1901 } 1902 1903 static int cmp_nla_oif(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1904 { 1905 if (a->oif != b->oif) 1906 return 1; 1907 1908 return 0; 1909 } 1910 1911 #define MAX_PROG_NAME 256 1912 static const struct nla_policy bpf_prog_policy[SEG6_LOCAL_BPF_PROG_MAX + 1] = { 1913 [SEG6_LOCAL_BPF_PROG] = { .type = NLA_U32, }, 1914 [SEG6_LOCAL_BPF_PROG_NAME] = { .type = NLA_NUL_STRING, 1915 .len = MAX_PROG_NAME }, 1916 }; 1917 1918 static int parse_nla_bpf(struct nlattr **attrs, struct seg6_local_lwt *slwt, 1919 struct netlink_ext_ack *extack) 1920 { 1921 struct nlattr *tb[SEG6_LOCAL_BPF_PROG_MAX + 1]; 1922 struct bpf_prog *p; 1923 int ret; 1924 u32 fd; 1925 1926 ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_BPF_PROG_MAX, 1927 attrs[SEG6_LOCAL_BPF], 1928 bpf_prog_policy, NULL); 1929 if (ret < 0) 1930 return ret; 1931 1932 if (!tb[SEG6_LOCAL_BPF_PROG] || !tb[SEG6_LOCAL_BPF_PROG_NAME]) 1933 return -EINVAL; 1934 1935 slwt->bpf.name = nla_memdup(tb[SEG6_LOCAL_BPF_PROG_NAME], GFP_KERNEL); 1936 if (!slwt->bpf.name) 1937 return -ENOMEM; 1938 1939 fd = nla_get_u32(tb[SEG6_LOCAL_BPF_PROG]); 1940 p = bpf_prog_get_type(fd, BPF_PROG_TYPE_LWT_SEG6LOCAL); 1941 if (IS_ERR(p)) { 1942 kfree(slwt->bpf.name); 1943 return PTR_ERR(p); 1944 } 1945 1946 slwt->bpf.prog = p; 1947 return 0; 1948 } 1949 1950 static int put_nla_bpf(struct sk_buff *skb, struct seg6_local_lwt *slwt) 1951 { 1952 struct nlattr *nest; 1953 1954 if (!slwt->bpf.prog) 1955 return 0; 1956 1957 nest = nla_nest_start_noflag(skb, SEG6_LOCAL_BPF); 1958 if (!nest) 1959 return -EMSGSIZE; 1960 1961 if (nla_put_u32(skb, SEG6_LOCAL_BPF_PROG, slwt->bpf.prog->aux->id)) 1962 return -EMSGSIZE; 1963 1964 if (slwt->bpf.name && 1965 nla_put_string(skb, SEG6_LOCAL_BPF_PROG_NAME, slwt->bpf.name)) 1966 return -EMSGSIZE; 1967 1968 return nla_nest_end(skb, nest); 1969 } 1970 1971 static int cmp_nla_bpf(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 1972 { 1973 if (!a->bpf.name && !b->bpf.name) 1974 return 0; 1975 1976 if (!a->bpf.name || !b->bpf.name) 1977 return 1; 1978 1979 return strcmp(a->bpf.name, b->bpf.name); 1980 } 1981 1982 static void destroy_attr_bpf(struct seg6_local_lwt *slwt) 1983 { 1984 kfree(slwt->bpf.name); 1985 if (slwt->bpf.prog) 1986 bpf_prog_put(slwt->bpf.prog); 1987 } 1988 1989 static const struct 1990 nla_policy seg6_local_counters_policy[SEG6_LOCAL_CNT_MAX + 1] = { 1991 [SEG6_LOCAL_CNT_PACKETS] = { .type = NLA_U64 }, 1992 [SEG6_LOCAL_CNT_BYTES] = { .type = NLA_U64 }, 1993 [SEG6_LOCAL_CNT_ERRORS] = { .type = NLA_U64 }, 1994 }; 1995 1996 static int parse_nla_counters(struct nlattr **attrs, 1997 struct seg6_local_lwt *slwt, 1998 struct netlink_ext_ack *extack) 1999 { 2000 struct pcpu_seg6_local_counters __percpu *pcounters; 2001 struct nlattr *tb[SEG6_LOCAL_CNT_MAX + 1]; 2002 int ret; 2003 2004 ret = nla_parse_nested_deprecated(tb, SEG6_LOCAL_CNT_MAX, 2005 attrs[SEG6_LOCAL_COUNTERS], 2006 seg6_local_counters_policy, NULL); 2007 if (ret < 0) 2008 return ret; 2009 2010 /* basic support for SRv6 Behavior counters requires at least: 2011 * packets, bytes and errors. 2012 */ 2013 if (!tb[SEG6_LOCAL_CNT_PACKETS] || !tb[SEG6_LOCAL_CNT_BYTES] || 2014 !tb[SEG6_LOCAL_CNT_ERRORS]) 2015 return -EINVAL; 2016 2017 /* counters are always zero initialized */ 2018 pcounters = seg6_local_alloc_pcpu_counters(GFP_KERNEL); 2019 if (!pcounters) 2020 return -ENOMEM; 2021 2022 slwt->pcpu_counters = pcounters; 2023 2024 return 0; 2025 } 2026 2027 static int seg6_local_fill_nla_counters(struct sk_buff *skb, 2028 struct seg6_local_counters *counters) 2029 { 2030 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_PACKETS, counters->packets, 2031 SEG6_LOCAL_CNT_PAD)) 2032 return -EMSGSIZE; 2033 2034 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_BYTES, counters->bytes, 2035 SEG6_LOCAL_CNT_PAD)) 2036 return -EMSGSIZE; 2037 2038 if (nla_put_u64_64bit(skb, SEG6_LOCAL_CNT_ERRORS, counters->errors, 2039 SEG6_LOCAL_CNT_PAD)) 2040 return -EMSGSIZE; 2041 2042 return 0; 2043 } 2044 2045 static int put_nla_counters(struct sk_buff *skb, struct seg6_local_lwt *slwt) 2046 { 2047 struct seg6_local_counters counters = { 0, 0, 0 }; 2048 struct nlattr *nest; 2049 int rc, i; 2050 2051 nest = nla_nest_start(skb, SEG6_LOCAL_COUNTERS); 2052 if (!nest) 2053 return -EMSGSIZE; 2054 2055 for_each_possible_cpu(i) { 2056 struct pcpu_seg6_local_counters *pcounters; 2057 u64 packets, bytes, errors; 2058 unsigned int start; 2059 2060 pcounters = per_cpu_ptr(slwt->pcpu_counters, i); 2061 do { 2062 start = u64_stats_fetch_begin(&pcounters->syncp); 2063 2064 packets = u64_stats_read(&pcounters->packets); 2065 bytes = u64_stats_read(&pcounters->bytes); 2066 errors = u64_stats_read(&pcounters->errors); 2067 2068 } while (u64_stats_fetch_retry(&pcounters->syncp, start)); 2069 2070 counters.packets += packets; 2071 counters.bytes += bytes; 2072 counters.errors += errors; 2073 } 2074 2075 rc = seg6_local_fill_nla_counters(skb, &counters); 2076 if (rc < 0) { 2077 nla_nest_cancel(skb, nest); 2078 return rc; 2079 } 2080 2081 return nla_nest_end(skb, nest); 2082 } 2083 2084 static int cmp_nla_counters(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 2085 { 2086 /* a and b are equal if both have pcpu_counters set or not */ 2087 return (!!((unsigned long)a->pcpu_counters)) ^ 2088 (!!((unsigned long)b->pcpu_counters)); 2089 } 2090 2091 static void destroy_attr_counters(struct seg6_local_lwt *slwt) 2092 { 2093 free_percpu(slwt->pcpu_counters); 2094 } 2095 2096 static const 2097 struct nla_policy seg6_local_flavors_policy[SEG6_LOCAL_FLV_MAX + 1] = { 2098 [SEG6_LOCAL_FLV_OPERATION] = { .type = NLA_U32 }, 2099 [SEG6_LOCAL_FLV_LCBLOCK_BITS] = { .type = NLA_U8 }, 2100 [SEG6_LOCAL_FLV_LCNODE_FN_BITS] = { .type = NLA_U8 }, 2101 }; 2102 2103 /* check whether the lengths of the Locator-Block and Locator-Node Function 2104 * are compatible with the dimension of a C-SID container. 2105 */ 2106 static int seg6_chk_next_csid_cfg(__u8 block_len, __u8 func_len) 2107 { 2108 /* Locator-Block and Locator-Node Function cannot exceed 128 bits 2109 * (i.e. C-SID container length). 2110 */ 2111 if (next_csid_chk_cntr_bits(block_len, func_len)) 2112 return -EINVAL; 2113 2114 /* Locator-Block length must be greater than zero and evenly divisible 2115 * by 8. There must be room for a Locator-Node Function, at least. 2116 */ 2117 if (next_csid_chk_lcblock_bits(block_len)) 2118 return -EINVAL; 2119 2120 /* Locator-Node Function length must be greater than zero and evenly 2121 * divisible by 8. There must be room for the Locator-Block. 2122 */ 2123 if (next_csid_chk_lcnode_fn_bits(func_len)) 2124 return -EINVAL; 2125 2126 return 0; 2127 } 2128 2129 static int seg6_parse_nla_next_csid_cfg(struct nlattr **tb, 2130 struct seg6_flavors_info *finfo, 2131 struct netlink_ext_ack *extack) 2132 { 2133 __u8 func_len = SEG6_LOCAL_LCNODE_FN_DBITS; 2134 __u8 block_len = SEG6_LOCAL_LCBLOCK_DBITS; 2135 int rc; 2136 2137 if (tb[SEG6_LOCAL_FLV_LCBLOCK_BITS]) 2138 block_len = nla_get_u8(tb[SEG6_LOCAL_FLV_LCBLOCK_BITS]); 2139 2140 if (tb[SEG6_LOCAL_FLV_LCNODE_FN_BITS]) 2141 func_len = nla_get_u8(tb[SEG6_LOCAL_FLV_LCNODE_FN_BITS]); 2142 2143 rc = seg6_chk_next_csid_cfg(block_len, func_len); 2144 if (rc < 0) { 2145 NL_SET_ERR_MSG(extack, 2146 "Invalid Locator Block/Node Function lengths"); 2147 return rc; 2148 } 2149 2150 finfo->lcblock_bits = block_len; 2151 finfo->lcnode_func_bits = func_len; 2152 2153 return 0; 2154 } 2155 2156 static int parse_nla_flavors(struct nlattr **attrs, struct seg6_local_lwt *slwt, 2157 struct netlink_ext_ack *extack) 2158 { 2159 struct seg6_flavors_info *finfo = &slwt->flv_info; 2160 struct nlattr *tb[SEG6_LOCAL_FLV_MAX + 1]; 2161 int action = slwt->action; 2162 __u32 fops, supp_fops; 2163 int rc; 2164 2165 rc = nla_parse_nested_deprecated(tb, SEG6_LOCAL_FLV_MAX, 2166 attrs[SEG6_LOCAL_FLAVORS], 2167 seg6_local_flavors_policy, NULL); 2168 if (rc < 0) 2169 return rc; 2170 2171 /* this attribute MUST always be present since it represents the Flavor 2172 * operation(s) to be carried out. 2173 */ 2174 if (!tb[SEG6_LOCAL_FLV_OPERATION]) 2175 return -EINVAL; 2176 2177 fops = nla_get_u32(tb[SEG6_LOCAL_FLV_OPERATION]); 2178 rc = seg6_flv_supp_ops_by_action(action, &supp_fops); 2179 if (rc < 0 || (fops & ~supp_fops)) { 2180 NL_SET_ERR_MSG(extack, "Unsupported Flavor operation(s)"); 2181 return -EOPNOTSUPP; 2182 } 2183 2184 finfo->flv_ops = fops; 2185 2186 if (seg6_next_csid_enabled(fops)) { 2187 /* Locator-Block and Locator-Node Function lengths can be 2188 * provided by the user space. Otherwise, default values are 2189 * applied. 2190 */ 2191 rc = seg6_parse_nla_next_csid_cfg(tb, finfo, extack); 2192 if (rc < 0) 2193 return rc; 2194 } 2195 2196 return 0; 2197 } 2198 2199 static int seg6_fill_nla_next_csid_cfg(struct sk_buff *skb, 2200 struct seg6_flavors_info *finfo) 2201 { 2202 if (nla_put_u8(skb, SEG6_LOCAL_FLV_LCBLOCK_BITS, finfo->lcblock_bits)) 2203 return -EMSGSIZE; 2204 2205 if (nla_put_u8(skb, SEG6_LOCAL_FLV_LCNODE_FN_BITS, 2206 finfo->lcnode_func_bits)) 2207 return -EMSGSIZE; 2208 2209 return 0; 2210 } 2211 2212 static int put_nla_flavors(struct sk_buff *skb, struct seg6_local_lwt *slwt) 2213 { 2214 struct seg6_flavors_info *finfo = &slwt->flv_info; 2215 __u32 fops = finfo->flv_ops; 2216 struct nlattr *nest; 2217 int rc; 2218 2219 nest = nla_nest_start(skb, SEG6_LOCAL_FLAVORS); 2220 if (!nest) 2221 return -EMSGSIZE; 2222 2223 if (nla_put_u32(skb, SEG6_LOCAL_FLV_OPERATION, fops)) { 2224 rc = -EMSGSIZE; 2225 goto err; 2226 } 2227 2228 if (seg6_next_csid_enabled(fops)) { 2229 rc = seg6_fill_nla_next_csid_cfg(skb, finfo); 2230 if (rc < 0) 2231 goto err; 2232 } 2233 2234 return nla_nest_end(skb, nest); 2235 2236 err: 2237 nla_nest_cancel(skb, nest); 2238 return rc; 2239 } 2240 2241 static int seg6_cmp_nla_next_csid_cfg(struct seg6_flavors_info *finfo_a, 2242 struct seg6_flavors_info *finfo_b) 2243 { 2244 if (finfo_a->lcblock_bits != finfo_b->lcblock_bits) 2245 return 1; 2246 2247 if (finfo_a->lcnode_func_bits != finfo_b->lcnode_func_bits) 2248 return 1; 2249 2250 return 0; 2251 } 2252 2253 static int cmp_nla_flavors(struct seg6_local_lwt *a, struct seg6_local_lwt *b) 2254 { 2255 struct seg6_flavors_info *finfo_a = &a->flv_info; 2256 struct seg6_flavors_info *finfo_b = &b->flv_info; 2257 2258 if (finfo_a->flv_ops != finfo_b->flv_ops) 2259 return 1; 2260 2261 if (seg6_next_csid_enabled(finfo_a->flv_ops)) { 2262 if (seg6_cmp_nla_next_csid_cfg(finfo_a, finfo_b)) 2263 return 1; 2264 } 2265 2266 return 0; 2267 } 2268 2269 static int encap_size_flavors(struct seg6_local_lwt *slwt) 2270 { 2271 struct seg6_flavors_info *finfo = &slwt->flv_info; 2272 int nlsize; 2273 2274 nlsize = nla_total_size(0) + /* nest SEG6_LOCAL_FLAVORS */ 2275 nla_total_size(4); /* SEG6_LOCAL_FLV_OPERATION */ 2276 2277 if (seg6_next_csid_enabled(finfo->flv_ops)) 2278 nlsize += nla_total_size(1) + /* SEG6_LOCAL_FLV_LCBLOCK_BITS */ 2279 nla_total_size(1); /* SEG6_LOCAL_FLV_LCNODE_FN_BITS */ 2280 2281 return nlsize; 2282 } 2283 2284 struct seg6_action_param { 2285 int (*parse)(struct nlattr **attrs, struct seg6_local_lwt *slwt, 2286 struct netlink_ext_ack *extack); 2287 int (*put)(struct sk_buff *skb, struct seg6_local_lwt *slwt); 2288 int (*cmp)(struct seg6_local_lwt *a, struct seg6_local_lwt *b); 2289 2290 /* optional destroy() callback useful for releasing resources which 2291 * have been previously acquired in the corresponding parse() 2292 * function. 2293 */ 2294 void (*destroy)(struct seg6_local_lwt *slwt); 2295 }; 2296 2297 static struct seg6_action_param seg6_action_params[SEG6_LOCAL_MAX + 1] = { 2298 [SEG6_LOCAL_SRH] = { .parse = parse_nla_srh, 2299 .put = put_nla_srh, 2300 .cmp = cmp_nla_srh, 2301 .destroy = destroy_attr_srh }, 2302 2303 [SEG6_LOCAL_TABLE] = { .parse = parse_nla_table, 2304 .put = put_nla_table, 2305 .cmp = cmp_nla_table }, 2306 2307 [SEG6_LOCAL_NH4] = { .parse = parse_nla_nh4, 2308 .put = put_nla_nh4, 2309 .cmp = cmp_nla_nh4 }, 2310 2311 [SEG6_LOCAL_NH6] = { .parse = parse_nla_nh6, 2312 .put = put_nla_nh6, 2313 .cmp = cmp_nla_nh6 }, 2314 2315 [SEG6_LOCAL_IIF] = { .parse = parse_nla_iif, 2316 .put = put_nla_iif, 2317 .cmp = cmp_nla_iif }, 2318 2319 [SEG6_LOCAL_OIF] = { .parse = parse_nla_oif, 2320 .put = put_nla_oif, 2321 .cmp = cmp_nla_oif }, 2322 2323 [SEG6_LOCAL_BPF] = { .parse = parse_nla_bpf, 2324 .put = put_nla_bpf, 2325 .cmp = cmp_nla_bpf, 2326 .destroy = destroy_attr_bpf }, 2327 2328 [SEG6_LOCAL_VRFTABLE] = { .parse = parse_nla_vrftable, 2329 .put = put_nla_vrftable, 2330 .cmp = cmp_nla_vrftable }, 2331 2332 [SEG6_LOCAL_COUNTERS] = { .parse = parse_nla_counters, 2333 .put = put_nla_counters, 2334 .cmp = cmp_nla_counters, 2335 .destroy = destroy_attr_counters }, 2336 2337 [SEG6_LOCAL_FLAVORS] = { .parse = parse_nla_flavors, 2338 .put = put_nla_flavors, 2339 .cmp = cmp_nla_flavors }, 2340 }; 2341 2342 /* call the destroy() callback (if available) for each set attribute in 2343 * @parsed_attrs, starting from the first attribute up to the @max_parsed 2344 * (excluded) attribute. 2345 */ 2346 static void __destroy_attrs(unsigned long parsed_attrs, int max_parsed, 2347 struct seg6_local_lwt *slwt) 2348 { 2349 struct seg6_action_param *param; 2350 int i; 2351 2352 /* Every required seg6local attribute is identified by an ID which is 2353 * encoded as a flag (i.e: 1 << ID) in the 'attrs' bitmask; 2354 * 2355 * We scan the 'parsed_attrs' bitmask, starting from the first attribute 2356 * up to the @max_parsed (excluded) attribute. 2357 * For each set attribute, we retrieve the corresponding destroy() 2358 * callback. If the callback is not available, then we skip to the next 2359 * attribute; otherwise, we call the destroy() callback. 2360 */ 2361 for (i = SEG6_LOCAL_SRH; i < max_parsed; ++i) { 2362 if (!(parsed_attrs & SEG6_F_ATTR(i))) 2363 continue; 2364 2365 param = &seg6_action_params[i]; 2366 2367 if (param->destroy) 2368 param->destroy(slwt); 2369 } 2370 } 2371 2372 /* release all the resources that may have been acquired during parsing 2373 * operations. 2374 */ 2375 static void destroy_attrs(struct seg6_local_lwt *slwt) 2376 { 2377 unsigned long attrs = slwt->desc->attrs | slwt->parsed_optattrs; 2378 2379 __destroy_attrs(attrs, SEG6_LOCAL_MAX + 1, slwt); 2380 } 2381 2382 static int parse_nla_optional_attrs(struct nlattr **attrs, 2383 struct seg6_local_lwt *slwt, 2384 struct netlink_ext_ack *extack) 2385 { 2386 struct seg6_action_desc *desc = slwt->desc; 2387 unsigned long parsed_optattrs = 0; 2388 struct seg6_action_param *param; 2389 int err, i; 2390 2391 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; ++i) { 2392 if (!(desc->optattrs & SEG6_F_ATTR(i)) || !attrs[i]) 2393 continue; 2394 2395 /* once here, the i-th attribute is provided by the 2396 * userspace AND it is identified optional as well. 2397 */ 2398 param = &seg6_action_params[i]; 2399 2400 err = param->parse(attrs, slwt, extack); 2401 if (err < 0) 2402 goto parse_optattrs_err; 2403 2404 /* current attribute has been correctly parsed */ 2405 parsed_optattrs |= SEG6_F_ATTR(i); 2406 } 2407 2408 /* store in the tunnel state all the optional attributed successfully 2409 * parsed. 2410 */ 2411 slwt->parsed_optattrs = parsed_optattrs; 2412 2413 return 0; 2414 2415 parse_optattrs_err: 2416 __destroy_attrs(parsed_optattrs, i, slwt); 2417 2418 return err; 2419 } 2420 2421 /* call the custom constructor of the behavior during its initialization phase 2422 * and after that all its attributes have been parsed successfully. 2423 */ 2424 static int 2425 seg6_local_lwtunnel_build_state(struct seg6_local_lwt *slwt, const void *cfg, 2426 struct netlink_ext_ack *extack) 2427 { 2428 struct seg6_action_desc *desc = slwt->desc; 2429 struct seg6_local_lwtunnel_ops *ops; 2430 2431 ops = &desc->slwt_ops; 2432 if (!ops->build_state) 2433 return 0; 2434 2435 return ops->build_state(slwt, cfg, extack); 2436 } 2437 2438 /* call the custom destructor of the behavior which is invoked before the 2439 * tunnel is going to be destroyed. 2440 */ 2441 static void seg6_local_lwtunnel_destroy_state(struct seg6_local_lwt *slwt) 2442 { 2443 struct seg6_action_desc *desc = slwt->desc; 2444 struct seg6_local_lwtunnel_ops *ops; 2445 2446 ops = &desc->slwt_ops; 2447 if (!ops->destroy_state) 2448 return; 2449 2450 ops->destroy_state(slwt); 2451 } 2452 2453 static int parse_nla_action(struct nlattr **attrs, struct seg6_local_lwt *slwt, 2454 struct netlink_ext_ack *extack) 2455 { 2456 struct seg6_action_param *param; 2457 struct seg6_action_desc *desc; 2458 unsigned long invalid_attrs; 2459 int i, err; 2460 2461 desc = __get_action_desc(slwt->action); 2462 if (!desc) 2463 return -EINVAL; 2464 2465 if (!desc->input) 2466 return -EOPNOTSUPP; 2467 2468 slwt->desc = desc; 2469 slwt->headroom += desc->static_headroom; 2470 2471 /* Forcing the desc->optattrs *set* and the desc->attrs *set* to be 2472 * disjoined, this allow us to release acquired resources by optional 2473 * attributes and by required attributes independently from each other 2474 * without any interference. 2475 * In other terms, we are sure that we do not release some the acquired 2476 * resources twice. 2477 * 2478 * Note that if an attribute is configured both as required and as 2479 * optional, it means that the user has messed something up in the 2480 * seg6_action_table. Therefore, this check is required for SRv6 2481 * behaviors to work properly. 2482 */ 2483 invalid_attrs = desc->attrs & desc->optattrs; 2484 if (invalid_attrs) { 2485 WARN_ONCE(1, 2486 "An attribute cannot be both required AND optional"); 2487 return -EINVAL; 2488 } 2489 2490 /* parse the required attributes */ 2491 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) { 2492 if (desc->attrs & SEG6_F_ATTR(i)) { 2493 if (!attrs[i]) 2494 return -EINVAL; 2495 2496 param = &seg6_action_params[i]; 2497 2498 err = param->parse(attrs, slwt, extack); 2499 if (err < 0) 2500 goto parse_attrs_err; 2501 } 2502 } 2503 2504 /* parse the optional attributes, if any */ 2505 err = parse_nla_optional_attrs(attrs, slwt, extack); 2506 if (err < 0) 2507 goto parse_attrs_err; 2508 2509 return 0; 2510 2511 parse_attrs_err: 2512 /* release any resource that may have been acquired during the i-1 2513 * parse() operations. 2514 */ 2515 __destroy_attrs(desc->attrs, i, slwt); 2516 2517 return err; 2518 } 2519 2520 static int seg6_local_build_state(struct net *net, struct nlattr *nla, 2521 unsigned int family, const void *cfg, 2522 struct lwtunnel_state **ts, 2523 struct netlink_ext_ack *extack) 2524 { 2525 struct nlattr *tb[SEG6_LOCAL_MAX + 1]; 2526 struct lwtunnel_state *newts; 2527 struct seg6_local_lwt *slwt; 2528 int err; 2529 2530 if (family != AF_INET6) 2531 return -EINVAL; 2532 2533 err = nla_parse_nested_deprecated(tb, SEG6_LOCAL_MAX, nla, 2534 seg6_local_policy, extack); 2535 2536 if (err < 0) 2537 return err; 2538 2539 if (!tb[SEG6_LOCAL_ACTION]) 2540 return -EINVAL; 2541 2542 newts = lwtunnel_state_alloc(sizeof(*slwt)); 2543 if (!newts) 2544 return -ENOMEM; 2545 2546 slwt = seg6_local_lwtunnel(newts); 2547 slwt->action = nla_get_u32(tb[SEG6_LOCAL_ACTION]); 2548 2549 err = parse_nla_action(tb, slwt, extack); 2550 if (err < 0) 2551 goto out_free; 2552 2553 err = seg6_local_lwtunnel_build_state(slwt, cfg, extack); 2554 if (err < 0) 2555 goto out_destroy_attrs; 2556 2557 newts->type = LWTUNNEL_ENCAP_SEG6_LOCAL; 2558 newts->flags = LWTUNNEL_STATE_INPUT_REDIRECT; 2559 newts->headroom = slwt->headroom; 2560 2561 *ts = newts; 2562 2563 return 0; 2564 2565 out_destroy_attrs: 2566 destroy_attrs(slwt); 2567 out_free: 2568 kfree(newts); 2569 return err; 2570 } 2571 2572 static void seg6_local_destroy_state(struct lwtunnel_state *lwt) 2573 { 2574 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt); 2575 2576 seg6_local_lwtunnel_destroy_state(slwt); 2577 2578 destroy_attrs(slwt); 2579 2580 return; 2581 } 2582 2583 static int seg6_local_fill_encap(struct sk_buff *skb, 2584 struct lwtunnel_state *lwt) 2585 { 2586 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt); 2587 struct seg6_action_param *param; 2588 unsigned long attrs; 2589 int i, err; 2590 2591 if (nla_put_u32(skb, SEG6_LOCAL_ACTION, slwt->action)) 2592 return -EMSGSIZE; 2593 2594 attrs = slwt->desc->attrs | slwt->parsed_optattrs; 2595 2596 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) { 2597 if (attrs & SEG6_F_ATTR(i)) { 2598 param = &seg6_action_params[i]; 2599 err = param->put(skb, slwt); 2600 if (err < 0) 2601 return err; 2602 } 2603 } 2604 2605 return 0; 2606 } 2607 2608 static int seg6_local_get_encap_size(struct lwtunnel_state *lwt) 2609 { 2610 struct seg6_local_lwt *slwt = seg6_local_lwtunnel(lwt); 2611 unsigned long attrs; 2612 int nlsize; 2613 2614 nlsize = nla_total_size(4); /* action */ 2615 2616 attrs = slwt->desc->attrs | slwt->parsed_optattrs; 2617 2618 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_SRH)) 2619 nlsize += nla_total_size((slwt->srh->hdrlen + 1) << 3); 2620 2621 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_TABLE)) 2622 nlsize += nla_total_size(4); 2623 2624 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH4)) 2625 nlsize += nla_total_size(4); 2626 2627 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_NH6)) 2628 nlsize += nla_total_size(16); 2629 2630 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_IIF)) 2631 nlsize += nla_total_size(4); 2632 2633 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_OIF)) 2634 nlsize += nla_total_size(4); 2635 2636 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_BPF)) 2637 nlsize += nla_total_size(sizeof(struct nlattr)) + 2638 nla_total_size(MAX_PROG_NAME) + 2639 nla_total_size(4); 2640 2641 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_VRFTABLE)) 2642 nlsize += nla_total_size(4); 2643 2644 if (attrs & SEG6_F_LOCAL_COUNTERS) 2645 nlsize += nla_total_size(0) + /* nest SEG6_LOCAL_COUNTERS */ 2646 /* SEG6_LOCAL_CNT_PACKETS */ 2647 nla_total_size_64bit(sizeof(__u64)) + 2648 /* SEG6_LOCAL_CNT_BYTES */ 2649 nla_total_size_64bit(sizeof(__u64)) + 2650 /* SEG6_LOCAL_CNT_ERRORS */ 2651 nla_total_size_64bit(sizeof(__u64)); 2652 2653 if (attrs & SEG6_F_ATTR(SEG6_LOCAL_FLAVORS)) 2654 nlsize += encap_size_flavors(slwt); 2655 2656 return nlsize; 2657 } 2658 2659 static int seg6_local_cmp_encap(struct lwtunnel_state *a, 2660 struct lwtunnel_state *b) 2661 { 2662 struct seg6_local_lwt *slwt_a, *slwt_b; 2663 struct seg6_action_param *param; 2664 unsigned long attrs_a, attrs_b; 2665 int i; 2666 2667 slwt_a = seg6_local_lwtunnel(a); 2668 slwt_b = seg6_local_lwtunnel(b); 2669 2670 if (slwt_a->action != slwt_b->action) 2671 return 1; 2672 2673 attrs_a = slwt_a->desc->attrs | slwt_a->parsed_optattrs; 2674 attrs_b = slwt_b->desc->attrs | slwt_b->parsed_optattrs; 2675 2676 if (attrs_a != attrs_b) 2677 return 1; 2678 2679 for (i = SEG6_LOCAL_SRH; i < SEG6_LOCAL_MAX + 1; i++) { 2680 if (attrs_a & SEG6_F_ATTR(i)) { 2681 param = &seg6_action_params[i]; 2682 if (param->cmp(slwt_a, slwt_b)) 2683 return 1; 2684 } 2685 } 2686 2687 return 0; 2688 } 2689 2690 static const struct lwtunnel_encap_ops seg6_local_ops = { 2691 .build_state = seg6_local_build_state, 2692 .destroy_state = seg6_local_destroy_state, 2693 .input = seg6_local_input, 2694 .fill_encap = seg6_local_fill_encap, 2695 .get_encap_size = seg6_local_get_encap_size, 2696 .cmp_encap = seg6_local_cmp_encap, 2697 .owner = THIS_MODULE, 2698 }; 2699 2700 int __init seg6_local_init(void) 2701 { 2702 /* If the max total number of defined attributes is reached, then your 2703 * kernel build stops here. 2704 * 2705 * This check is required to avoid arithmetic overflows when processing 2706 * behavior attributes and the maximum number of defined attributes 2707 * exceeds the allowed value. 2708 */ 2709 BUILD_BUG_ON(SEG6_LOCAL_MAX + 1 > BITS_PER_TYPE(unsigned long)); 2710 2711 /* Check whether the number of defined flavors exceeds the maximum 2712 * allowed value. 2713 */ 2714 BUILD_BUG_ON(SEG6_LOCAL_FLV_OP_MAX + 1 > BITS_PER_TYPE(__u32)); 2715 2716 /* If the default NEXT-C-SID Locator-Block/Node Function lengths (in 2717 * bits) have been changed with invalid values, kernel build stops 2718 * here. 2719 */ 2720 BUILD_BUG_ON(next_csid_chk_cntr_bits(SEG6_LOCAL_LCBLOCK_DBITS, 2721 SEG6_LOCAL_LCNODE_FN_DBITS)); 2722 BUILD_BUG_ON(next_csid_chk_lcblock_bits(SEG6_LOCAL_LCBLOCK_DBITS)); 2723 BUILD_BUG_ON(next_csid_chk_lcnode_fn_bits(SEG6_LOCAL_LCNODE_FN_DBITS)); 2724 2725 /* To be memory efficient, we use 'u8' to represent the different 2726 * actions related to RFC8986 flavors. If the kernel build stops here, 2727 * it means that it is not possible to correctly encode these actions 2728 * with the data type chosen for the action table. 2729 */ 2730 BUILD_BUG_ON(SEG6_LOCAL_FLV_ACT_MAX > (typeof(flv8986_act_tbl[0]))~0U); 2731 2732 return lwtunnel_encap_add_ops(&seg6_local_ops, 2733 LWTUNNEL_ENCAP_SEG6_LOCAL); 2734 } 2735 2736 void seg6_local_exit(void) 2737 { 2738 lwtunnel_encap_del_ops(&seg6_local_ops, LWTUNNEL_ENCAP_SEG6_LOCAL); 2739 } 2740