1 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 2 * 3 * This program is free software; you can redistribute it and/or 4 * modify it under the terms of version 2 of the GNU General Public 5 * License as published by the Free Software Foundation. 6 */ 7 #ifndef _UAPI__LINUX_BPF_H__ 8 #define _UAPI__LINUX_BPF_H__ 9 10 #include <linux/types.h> 11 #include <linux/bpf_common.h> 12 13 /* Extended instruction set based on top of classic BPF */ 14 15 /* instruction classes */ 16 #define BPF_ALU64 0x07 /* alu mode in double word width */ 17 18 /* ld/ldx fields */ 19 #define BPF_DW 0x18 /* double word */ 20 #define BPF_XADD 0xc0 /* exclusive add */ 21 22 /* alu/jmp fields */ 23 #define BPF_MOV 0xb0 /* mov reg to reg */ 24 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 25 26 /* change endianness of a register */ 27 #define BPF_END 0xd0 /* flags for endianness conversion: */ 28 #define BPF_TO_LE 0x00 /* convert to little-endian */ 29 #define BPF_TO_BE 0x08 /* convert to big-endian */ 30 #define BPF_FROM_LE BPF_TO_LE 31 #define BPF_FROM_BE BPF_TO_BE 32 33 /* jmp encodings */ 34 #define BPF_JNE 0x50 /* jump != */ 35 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 36 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 37 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 38 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 39 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 40 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 41 #define BPF_CALL 0x80 /* function call */ 42 #define BPF_EXIT 0x90 /* function return */ 43 44 /* Register numbers */ 45 enum { 46 BPF_REG_0 = 0, 47 BPF_REG_1, 48 BPF_REG_2, 49 BPF_REG_3, 50 BPF_REG_4, 51 BPF_REG_5, 52 BPF_REG_6, 53 BPF_REG_7, 54 BPF_REG_8, 55 BPF_REG_9, 56 BPF_REG_10, 57 __MAX_BPF_REG, 58 }; 59 60 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 61 #define MAX_BPF_REG __MAX_BPF_REG 62 63 struct bpf_insn { 64 __u8 code; /* opcode */ 65 __u8 dst_reg:4; /* dest register */ 66 __u8 src_reg:4; /* source register */ 67 __s16 off; /* signed offset */ 68 __s32 imm; /* signed immediate constant */ 69 }; 70 71 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 72 struct bpf_lpm_trie_key { 73 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 74 __u8 data[0]; /* Arbitrary size */ 75 }; 76 77 /* BPF syscall commands, see bpf(2) man-page for details. */ 78 enum bpf_cmd { 79 BPF_MAP_CREATE, 80 BPF_MAP_LOOKUP_ELEM, 81 BPF_MAP_UPDATE_ELEM, 82 BPF_MAP_DELETE_ELEM, 83 BPF_MAP_GET_NEXT_KEY, 84 BPF_PROG_LOAD, 85 BPF_OBJ_PIN, 86 BPF_OBJ_GET, 87 BPF_PROG_ATTACH, 88 BPF_PROG_DETACH, 89 BPF_PROG_TEST_RUN, 90 BPF_PROG_GET_NEXT_ID, 91 BPF_MAP_GET_NEXT_ID, 92 BPF_PROG_GET_FD_BY_ID, 93 BPF_MAP_GET_FD_BY_ID, 94 BPF_OBJ_GET_INFO_BY_FD, 95 }; 96 97 enum bpf_map_type { 98 BPF_MAP_TYPE_UNSPEC, 99 BPF_MAP_TYPE_HASH, 100 BPF_MAP_TYPE_ARRAY, 101 BPF_MAP_TYPE_PROG_ARRAY, 102 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 103 BPF_MAP_TYPE_PERCPU_HASH, 104 BPF_MAP_TYPE_PERCPU_ARRAY, 105 BPF_MAP_TYPE_STACK_TRACE, 106 BPF_MAP_TYPE_CGROUP_ARRAY, 107 BPF_MAP_TYPE_LRU_HASH, 108 BPF_MAP_TYPE_LRU_PERCPU_HASH, 109 BPF_MAP_TYPE_LPM_TRIE, 110 BPF_MAP_TYPE_ARRAY_OF_MAPS, 111 BPF_MAP_TYPE_HASH_OF_MAPS, 112 BPF_MAP_TYPE_DEVMAP, 113 }; 114 115 enum bpf_prog_type { 116 BPF_PROG_TYPE_UNSPEC, 117 BPF_PROG_TYPE_SOCKET_FILTER, 118 BPF_PROG_TYPE_KPROBE, 119 BPF_PROG_TYPE_SCHED_CLS, 120 BPF_PROG_TYPE_SCHED_ACT, 121 BPF_PROG_TYPE_TRACEPOINT, 122 BPF_PROG_TYPE_XDP, 123 BPF_PROG_TYPE_PERF_EVENT, 124 BPF_PROG_TYPE_CGROUP_SKB, 125 BPF_PROG_TYPE_CGROUP_SOCK, 126 BPF_PROG_TYPE_LWT_IN, 127 BPF_PROG_TYPE_LWT_OUT, 128 BPF_PROG_TYPE_LWT_XMIT, 129 BPF_PROG_TYPE_SOCK_OPS, 130 }; 131 132 enum bpf_attach_type { 133 BPF_CGROUP_INET_INGRESS, 134 BPF_CGROUP_INET_EGRESS, 135 BPF_CGROUP_INET_SOCK_CREATE, 136 BPF_CGROUP_SOCK_OPS, 137 __MAX_BPF_ATTACH_TYPE 138 }; 139 140 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 141 142 /* If BPF_F_ALLOW_OVERRIDE flag is used in BPF_PROG_ATTACH command 143 * to the given target_fd cgroup the descendent cgroup will be able to 144 * override effective bpf program that was inherited from this cgroup 145 */ 146 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 147 148 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 149 * verifier will perform strict alignment checking as if the kernel 150 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 151 * and NET_IP_ALIGN defined to 2. 152 */ 153 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 154 155 #define BPF_PSEUDO_MAP_FD 1 156 157 /* flags for BPF_MAP_UPDATE_ELEM command */ 158 #define BPF_ANY 0 /* create new element or update existing */ 159 #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 160 #define BPF_EXIST 2 /* update existing element */ 161 162 #define BPF_F_NO_PREALLOC (1U << 0) 163 /* Instead of having one common LRU list in the 164 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 165 * which can scale and perform better. 166 * Note, the LRU nodes (including free nodes) cannot be moved 167 * across different LRU lists. 168 */ 169 #define BPF_F_NO_COMMON_LRU (1U << 1) 170 171 union bpf_attr { 172 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 173 __u32 map_type; /* one of enum bpf_map_type */ 174 __u32 key_size; /* size of key in bytes */ 175 __u32 value_size; /* size of value in bytes */ 176 __u32 max_entries; /* max number of entries in a map */ 177 __u32 map_flags; /* prealloc or not */ 178 __u32 inner_map_fd; /* fd pointing to the inner map */ 179 }; 180 181 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 182 __u32 map_fd; 183 __aligned_u64 key; 184 union { 185 __aligned_u64 value; 186 __aligned_u64 next_key; 187 }; 188 __u64 flags; 189 }; 190 191 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 192 __u32 prog_type; /* one of enum bpf_prog_type */ 193 __u32 insn_cnt; 194 __aligned_u64 insns; 195 __aligned_u64 license; 196 __u32 log_level; /* verbosity level of verifier */ 197 __u32 log_size; /* size of user buffer */ 198 __aligned_u64 log_buf; /* user supplied buffer */ 199 __u32 kern_version; /* checked when prog_type=kprobe */ 200 __u32 prog_flags; 201 }; 202 203 struct { /* anonymous struct used by BPF_OBJ_* commands */ 204 __aligned_u64 pathname; 205 __u32 bpf_fd; 206 }; 207 208 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 209 __u32 target_fd; /* container object to attach to */ 210 __u32 attach_bpf_fd; /* eBPF program to attach */ 211 __u32 attach_type; 212 __u32 attach_flags; 213 }; 214 215 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 216 __u32 prog_fd; 217 __u32 retval; 218 __u32 data_size_in; 219 __u32 data_size_out; 220 __aligned_u64 data_in; 221 __aligned_u64 data_out; 222 __u32 repeat; 223 __u32 duration; 224 } test; 225 226 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 227 union { 228 __u32 start_id; 229 __u32 prog_id; 230 __u32 map_id; 231 }; 232 __u32 next_id; 233 }; 234 235 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 236 __u32 bpf_fd; 237 __u32 info_len; 238 __aligned_u64 info; 239 } info; 240 } __attribute__((aligned(8))); 241 242 /* BPF helper function descriptions: 243 * 244 * void *bpf_map_lookup_elem(&map, &key) 245 * Return: Map value or NULL 246 * 247 * int bpf_map_update_elem(&map, &key, &value, flags) 248 * Return: 0 on success or negative error 249 * 250 * int bpf_map_delete_elem(&map, &key) 251 * Return: 0 on success or negative error 252 * 253 * int bpf_probe_read(void *dst, int size, void *src) 254 * Return: 0 on success or negative error 255 * 256 * u64 bpf_ktime_get_ns(void) 257 * Return: current ktime 258 * 259 * int bpf_trace_printk(const char *fmt, int fmt_size, ...) 260 * Return: length of buffer written or negative error 261 * 262 * u32 bpf_prandom_u32(void) 263 * Return: random value 264 * 265 * u32 bpf_raw_smp_processor_id(void) 266 * Return: SMP processor ID 267 * 268 * int bpf_skb_store_bytes(skb, offset, from, len, flags) 269 * store bytes into packet 270 * @skb: pointer to skb 271 * @offset: offset within packet from skb->mac_header 272 * @from: pointer where to copy bytes from 273 * @len: number of bytes to store into packet 274 * @flags: bit 0 - if true, recompute skb->csum 275 * other bits - reserved 276 * Return: 0 on success or negative error 277 * 278 * int bpf_l3_csum_replace(skb, offset, from, to, flags) 279 * recompute IP checksum 280 * @skb: pointer to skb 281 * @offset: offset within packet where IP checksum is located 282 * @from: old value of header field 283 * @to: new value of header field 284 * @flags: bits 0-3 - size of header field 285 * other bits - reserved 286 * Return: 0 on success or negative error 287 * 288 * int bpf_l4_csum_replace(skb, offset, from, to, flags) 289 * recompute TCP/UDP checksum 290 * @skb: pointer to skb 291 * @offset: offset within packet where TCP/UDP checksum is located 292 * @from: old value of header field 293 * @to: new value of header field 294 * @flags: bits 0-3 - size of header field 295 * bit 4 - is pseudo header 296 * other bits - reserved 297 * Return: 0 on success or negative error 298 * 299 * int bpf_tail_call(ctx, prog_array_map, index) 300 * jump into another BPF program 301 * @ctx: context pointer passed to next program 302 * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY 303 * @index: index inside array that selects specific program to run 304 * Return: 0 on success or negative error 305 * 306 * int bpf_clone_redirect(skb, ifindex, flags) 307 * redirect to another netdev 308 * @skb: pointer to skb 309 * @ifindex: ifindex of the net device 310 * @flags: bit 0 - if set, redirect to ingress instead of egress 311 * other bits - reserved 312 * Return: 0 on success or negative error 313 * 314 * u64 bpf_get_current_pid_tgid(void) 315 * Return: current->tgid << 32 | current->pid 316 * 317 * u64 bpf_get_current_uid_gid(void) 318 * Return: current_gid << 32 | current_uid 319 * 320 * int bpf_get_current_comm(char *buf, int size_of_buf) 321 * stores current->comm into buf 322 * Return: 0 on success or negative error 323 * 324 * u32 bpf_get_cgroup_classid(skb) 325 * retrieve a proc's classid 326 * @skb: pointer to skb 327 * Return: classid if != 0 328 * 329 * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci) 330 * Return: 0 on success or negative error 331 * 332 * int bpf_skb_vlan_pop(skb) 333 * Return: 0 on success or negative error 334 * 335 * int bpf_skb_get_tunnel_key(skb, key, size, flags) 336 * int bpf_skb_set_tunnel_key(skb, key, size, flags) 337 * retrieve or populate tunnel metadata 338 * @skb: pointer to skb 339 * @key: pointer to 'struct bpf_tunnel_key' 340 * @size: size of 'struct bpf_tunnel_key' 341 * @flags: room for future extensions 342 * Return: 0 on success or negative error 343 * 344 * u64 bpf_perf_event_read(map, flags) 345 * read perf event counter value 346 * @map: pointer to perf_event_array map 347 * @flags: index of event in the map or bitmask flags 348 * Return: value of perf event counter read or error code 349 * 350 * int bpf_redirect(ifindex, flags) 351 * redirect to another netdev 352 * @ifindex: ifindex of the net device 353 * @flags: 354 * cls_bpf: 355 * bit 0 - if set, redirect to ingress instead of egress 356 * other bits - reserved 357 * xdp_bpf: 358 * all bits - reserved 359 * Return: cls_bpf: TC_ACT_REDIRECT on success or TC_ACT_SHOT on error 360 * xdp_bfp: XDP_REDIRECT on success or XDP_ABORT on error 361 * int bpf_redirect_map(map, key, flags) 362 * redirect to endpoint in map 363 * @map: pointer to dev map 364 * @key: index in map to lookup 365 * @flags: -- 366 * Return: XDP_REDIRECT on success or XDP_ABORT on error 367 * 368 * u32 bpf_get_route_realm(skb) 369 * retrieve a dst's tclassid 370 * @skb: pointer to skb 371 * Return: realm if != 0 372 * 373 * int bpf_perf_event_output(ctx, map, flags, data, size) 374 * output perf raw sample 375 * @ctx: struct pt_regs* 376 * @map: pointer to perf_event_array map 377 * @flags: index of event in the map or bitmask flags 378 * @data: data on stack to be output as raw data 379 * @size: size of data 380 * Return: 0 on success or negative error 381 * 382 * int bpf_get_stackid(ctx, map, flags) 383 * walk user or kernel stack and return id 384 * @ctx: struct pt_regs* 385 * @map: pointer to stack_trace map 386 * @flags: bits 0-7 - numer of stack frames to skip 387 * bit 8 - collect user stack instead of kernel 388 * bit 9 - compare stacks by hash only 389 * bit 10 - if two different stacks hash into the same stackid 390 * discard old 391 * other bits - reserved 392 * Return: >= 0 stackid on success or negative error 393 * 394 * s64 bpf_csum_diff(from, from_size, to, to_size, seed) 395 * calculate csum diff 396 * @from: raw from buffer 397 * @from_size: length of from buffer 398 * @to: raw to buffer 399 * @to_size: length of to buffer 400 * @seed: optional seed 401 * Return: csum result or negative error code 402 * 403 * int bpf_skb_get_tunnel_opt(skb, opt, size) 404 * retrieve tunnel options metadata 405 * @skb: pointer to skb 406 * @opt: pointer to raw tunnel option data 407 * @size: size of @opt 408 * Return: option size 409 * 410 * int bpf_skb_set_tunnel_opt(skb, opt, size) 411 * populate tunnel options metadata 412 * @skb: pointer to skb 413 * @opt: pointer to raw tunnel option data 414 * @size: size of @opt 415 * Return: 0 on success or negative error 416 * 417 * int bpf_skb_change_proto(skb, proto, flags) 418 * Change protocol of the skb. Currently supported is v4 -> v6, 419 * v6 -> v4 transitions. The helper will also resize the skb. eBPF 420 * program is expected to fill the new headers via skb_store_bytes 421 * and lX_csum_replace. 422 * @skb: pointer to skb 423 * @proto: new skb->protocol type 424 * @flags: reserved 425 * Return: 0 on success or negative error 426 * 427 * int bpf_skb_change_type(skb, type) 428 * Change packet type of skb. 429 * @skb: pointer to skb 430 * @type: new skb->pkt_type type 431 * Return: 0 on success or negative error 432 * 433 * int bpf_skb_under_cgroup(skb, map, index) 434 * Check cgroup2 membership of skb 435 * @skb: pointer to skb 436 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type 437 * @index: index of the cgroup in the bpf_map 438 * Return: 439 * == 0 skb failed the cgroup2 descendant test 440 * == 1 skb succeeded the cgroup2 descendant test 441 * < 0 error 442 * 443 * u32 bpf_get_hash_recalc(skb) 444 * Retrieve and possibly recalculate skb->hash. 445 * @skb: pointer to skb 446 * Return: hash 447 * 448 * u64 bpf_get_current_task(void) 449 * Returns current task_struct 450 * Return: current 451 * 452 * int bpf_probe_write_user(void *dst, void *src, int len) 453 * safely attempt to write to a location 454 * @dst: destination address in userspace 455 * @src: source address on stack 456 * @len: number of bytes to copy 457 * Return: 0 on success or negative error 458 * 459 * int bpf_current_task_under_cgroup(map, index) 460 * Check cgroup2 membership of current task 461 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type 462 * @index: index of the cgroup in the bpf_map 463 * Return: 464 * == 0 current failed the cgroup2 descendant test 465 * == 1 current succeeded the cgroup2 descendant test 466 * < 0 error 467 * 468 * int bpf_skb_change_tail(skb, len, flags) 469 * The helper will resize the skb to the given new size, to be used f.e. 470 * with control messages. 471 * @skb: pointer to skb 472 * @len: new skb length 473 * @flags: reserved 474 * Return: 0 on success or negative error 475 * 476 * int bpf_skb_pull_data(skb, len) 477 * The helper will pull in non-linear data in case the skb is non-linear 478 * and not all of len are part of the linear section. Only needed for 479 * read/write with direct packet access. 480 * @skb: pointer to skb 481 * @len: len to make read/writeable 482 * Return: 0 on success or negative error 483 * 484 * s64 bpf_csum_update(skb, csum) 485 * Adds csum into skb->csum in case of CHECKSUM_COMPLETE. 486 * @skb: pointer to skb 487 * @csum: csum to add 488 * Return: csum on success or negative error 489 * 490 * void bpf_set_hash_invalid(skb) 491 * Invalidate current skb->hash. 492 * @skb: pointer to skb 493 * 494 * int bpf_get_numa_node_id() 495 * Return: Id of current NUMA node. 496 * 497 * int bpf_skb_change_head() 498 * Grows headroom of skb and adjusts MAC header offset accordingly. 499 * Will extends/reallocae as required automatically. 500 * May change skb data pointer and will thus invalidate any check 501 * performed for direct packet access. 502 * @skb: pointer to skb 503 * @len: length of header to be pushed in front 504 * @flags: Flags (unused for now) 505 * Return: 0 on success or negative error 506 * 507 * int bpf_xdp_adjust_head(xdp_md, delta) 508 * Adjust the xdp_md.data by delta 509 * @xdp_md: pointer to xdp_md 510 * @delta: An positive/negative integer to be added to xdp_md.data 511 * Return: 0 on success or negative on error 512 * 513 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) 514 * Copy a NUL terminated string from unsafe address. In case the string 515 * length is smaller than size, the target is not padded with further NUL 516 * bytes. In case the string length is larger than size, just count-1 517 * bytes are copied and the last byte is set to NUL. 518 * @dst: destination address 519 * @size: maximum number of bytes to copy, including the trailing NUL 520 * @unsafe_ptr: unsafe address 521 * Return: 522 * > 0 length of the string including the trailing NUL on success 523 * < 0 error 524 * 525 * u64 bpf_get_socket_cookie(skb) 526 * Get the cookie for the socket stored inside sk_buff. 527 * @skb: pointer to skb 528 * Return: 8 Bytes non-decreasing number on success or 0 if the socket 529 * field is missing inside sk_buff 530 * 531 * u32 bpf_get_socket_uid(skb) 532 * Get the owner uid of the socket stored inside sk_buff. 533 * @skb: pointer to skb 534 * Return: uid of the socket owner on success or overflowuid if failed. 535 * 536 * u32 bpf_set_hash(skb, hash) 537 * Set full skb->hash. 538 * @skb: pointer to skb 539 * @hash: hash to set 540 * 541 * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen) 542 * Calls setsockopt. Not all opts are available, only those with 543 * integer optvals plus TCP_CONGESTION. 544 * Supported levels: SOL_SOCKET and IPROTO_TCP 545 * @bpf_socket: pointer to bpf_socket 546 * @level: SOL_SOCKET or IPROTO_TCP 547 * @optname: option name 548 * @optval: pointer to option value 549 * @optlen: length of optval in byes 550 * Return: 0 or negative error 551 * 552 * int bpf_skb_adjust_room(skb, len_diff, mode, flags) 553 * Grow or shrink room in sk_buff. 554 * @skb: pointer to skb 555 * @len_diff: (signed) amount of room to grow/shrink 556 * @mode: operation mode (enum bpf_adj_room_mode) 557 * @flags: reserved for future use 558 * Return: 0 on success or negative error code 559 */ 560 #define __BPF_FUNC_MAPPER(FN) \ 561 FN(unspec), \ 562 FN(map_lookup_elem), \ 563 FN(map_update_elem), \ 564 FN(map_delete_elem), \ 565 FN(probe_read), \ 566 FN(ktime_get_ns), \ 567 FN(trace_printk), \ 568 FN(get_prandom_u32), \ 569 FN(get_smp_processor_id), \ 570 FN(skb_store_bytes), \ 571 FN(l3_csum_replace), \ 572 FN(l4_csum_replace), \ 573 FN(tail_call), \ 574 FN(clone_redirect), \ 575 FN(get_current_pid_tgid), \ 576 FN(get_current_uid_gid), \ 577 FN(get_current_comm), \ 578 FN(get_cgroup_classid), \ 579 FN(skb_vlan_push), \ 580 FN(skb_vlan_pop), \ 581 FN(skb_get_tunnel_key), \ 582 FN(skb_set_tunnel_key), \ 583 FN(perf_event_read), \ 584 FN(redirect), \ 585 FN(get_route_realm), \ 586 FN(perf_event_output), \ 587 FN(skb_load_bytes), \ 588 FN(get_stackid), \ 589 FN(csum_diff), \ 590 FN(skb_get_tunnel_opt), \ 591 FN(skb_set_tunnel_opt), \ 592 FN(skb_change_proto), \ 593 FN(skb_change_type), \ 594 FN(skb_under_cgroup), \ 595 FN(get_hash_recalc), \ 596 FN(get_current_task), \ 597 FN(probe_write_user), \ 598 FN(current_task_under_cgroup), \ 599 FN(skb_change_tail), \ 600 FN(skb_pull_data), \ 601 FN(csum_update), \ 602 FN(set_hash_invalid), \ 603 FN(get_numa_node_id), \ 604 FN(skb_change_head), \ 605 FN(xdp_adjust_head), \ 606 FN(probe_read_str), \ 607 FN(get_socket_cookie), \ 608 FN(get_socket_uid), \ 609 FN(set_hash), \ 610 FN(setsockopt), \ 611 FN(skb_adjust_room), \ 612 FN(redirect_map), 613 614 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 615 * function eBPF program intends to call 616 */ 617 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 618 enum bpf_func_id { 619 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 620 __BPF_FUNC_MAX_ID, 621 }; 622 #undef __BPF_ENUM_FN 623 624 /* All flags used by eBPF helper functions, placed here. */ 625 626 /* BPF_FUNC_skb_store_bytes flags. */ 627 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 628 #define BPF_F_INVALIDATE_HASH (1ULL << 1) 629 630 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 631 * First 4 bits are for passing the header field size. 632 */ 633 #define BPF_F_HDR_FIELD_MASK 0xfULL 634 635 /* BPF_FUNC_l4_csum_replace flags. */ 636 #define BPF_F_PSEUDO_HDR (1ULL << 4) 637 #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 638 #define BPF_F_MARK_ENFORCE (1ULL << 6) 639 640 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 641 #define BPF_F_INGRESS (1ULL << 0) 642 643 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 644 #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 645 646 /* BPF_FUNC_get_stackid flags. */ 647 #define BPF_F_SKIP_FIELD_MASK 0xffULL 648 #define BPF_F_USER_STACK (1ULL << 8) 649 #define BPF_F_FAST_STACK_CMP (1ULL << 9) 650 #define BPF_F_REUSE_STACKID (1ULL << 10) 651 652 /* BPF_FUNC_skb_set_tunnel_key flags. */ 653 #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 654 #define BPF_F_DONT_FRAGMENT (1ULL << 2) 655 656 /* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */ 657 #define BPF_F_INDEX_MASK 0xffffffffULL 658 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 659 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 660 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 661 662 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 663 enum bpf_adj_room_mode { 664 BPF_ADJ_ROOM_NET, 665 }; 666 667 /* user accessible mirror of in-kernel sk_buff. 668 * new fields can only be added to the end of this structure 669 */ 670 struct __sk_buff { 671 __u32 len; 672 __u32 pkt_type; 673 __u32 mark; 674 __u32 queue_mapping; 675 __u32 protocol; 676 __u32 vlan_present; 677 __u32 vlan_tci; 678 __u32 vlan_proto; 679 __u32 priority; 680 __u32 ingress_ifindex; 681 __u32 ifindex; 682 __u32 tc_index; 683 __u32 cb[5]; 684 __u32 hash; 685 __u32 tc_classid; 686 __u32 data; 687 __u32 data_end; 688 __u32 napi_id; 689 }; 690 691 struct bpf_tunnel_key { 692 __u32 tunnel_id; 693 union { 694 __u32 remote_ipv4; 695 __u32 remote_ipv6[4]; 696 }; 697 __u8 tunnel_tos; 698 __u8 tunnel_ttl; 699 __u16 tunnel_ext; 700 __u32 tunnel_label; 701 }; 702 703 /* Generic BPF return codes which all BPF program types may support. 704 * The values are binary compatible with their TC_ACT_* counter-part to 705 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 706 * programs. 707 * 708 * XDP is handled seprately, see XDP_*. 709 */ 710 enum bpf_ret_code { 711 BPF_OK = 0, 712 /* 1 reserved */ 713 BPF_DROP = 2, 714 /* 3-6 reserved */ 715 BPF_REDIRECT = 7, 716 /* >127 are reserved for prog type specific return codes */ 717 }; 718 719 struct bpf_sock { 720 __u32 bound_dev_if; 721 __u32 family; 722 __u32 type; 723 __u32 protocol; 724 }; 725 726 #define XDP_PACKET_HEADROOM 256 727 728 /* User return codes for XDP prog type. 729 * A valid XDP program must return one of these defined values. All other 730 * return codes are reserved for future use. Unknown return codes will result 731 * in packet drop. 732 */ 733 enum xdp_action { 734 XDP_ABORTED = 0, 735 XDP_DROP, 736 XDP_PASS, 737 XDP_TX, 738 XDP_REDIRECT, 739 }; 740 741 /* user accessible metadata for XDP packet hook 742 * new fields must be added to the end of this structure 743 */ 744 struct xdp_md { 745 __u32 data; 746 __u32 data_end; 747 }; 748 749 #define BPF_TAG_SIZE 8 750 751 struct bpf_prog_info { 752 __u32 type; 753 __u32 id; 754 __u8 tag[BPF_TAG_SIZE]; 755 __u32 jited_prog_len; 756 __u32 xlated_prog_len; 757 __aligned_u64 jited_prog_insns; 758 __aligned_u64 xlated_prog_insns; 759 } __attribute__((aligned(8))); 760 761 struct bpf_map_info { 762 __u32 type; 763 __u32 id; 764 __u32 key_size; 765 __u32 value_size; 766 __u32 max_entries; 767 __u32 map_flags; 768 } __attribute__((aligned(8))); 769 770 /* User bpf_sock_ops struct to access socket values and specify request ops 771 * and their replies. 772 * Some of this fields are in network (bigendian) byte order and may need 773 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 774 * New fields can only be added at the end of this structure 775 */ 776 struct bpf_sock_ops { 777 __u32 op; 778 union { 779 __u32 reply; 780 __u32 replylong[4]; 781 }; 782 __u32 family; 783 __u32 remote_ip4; /* Stored in network byte order */ 784 __u32 local_ip4; /* Stored in network byte order */ 785 __u32 remote_ip6[4]; /* Stored in network byte order */ 786 __u32 local_ip6[4]; /* Stored in network byte order */ 787 __u32 remote_port; /* Stored in network byte order */ 788 __u32 local_port; /* stored in host byte order */ 789 }; 790 791 /* List of known BPF sock_ops operators. 792 * New entries can only be added at the end 793 */ 794 enum { 795 BPF_SOCK_OPS_VOID, 796 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 797 * -1 if default value should be used 798 */ 799 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 800 * window (in packets) or -1 if default 801 * value should be used 802 */ 803 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 804 * active connection is initialized 805 */ 806 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 807 * active connection is 808 * established 809 */ 810 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 811 * passive connection is 812 * established 813 */ 814 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 815 * needs ECN 816 */ 817 }; 818 819 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 820 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 821 822 #endif /* _UAPI__LINUX_BPF_H__ */ 823