1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of version 2 of the GNU General Public 6 * License as published by the Free Software Foundation. 7 */ 8 #ifndef _UAPI__LINUX_BPF_H__ 9 #define _UAPI__LINUX_BPF_H__ 10 11 #include <linux/types.h> 12 #include <linux/bpf_common.h> 13 14 /* Extended instruction set based on top of classic BPF */ 15 16 /* instruction classes */ 17 #define BPF_ALU64 0x07 /* alu mode in double word width */ 18 19 /* ld/ldx fields */ 20 #define BPF_DW 0x18 /* double word */ 21 #define BPF_XADD 0xc0 /* exclusive add */ 22 23 /* alu/jmp fields */ 24 #define BPF_MOV 0xb0 /* mov reg to reg */ 25 #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 26 27 /* change endianness of a register */ 28 #define BPF_END 0xd0 /* flags for endianness conversion: */ 29 #define BPF_TO_LE 0x00 /* convert to little-endian */ 30 #define BPF_TO_BE 0x08 /* convert to big-endian */ 31 #define BPF_FROM_LE BPF_TO_LE 32 #define BPF_FROM_BE BPF_TO_BE 33 34 /* jmp encodings */ 35 #define BPF_JNE 0x50 /* jump != */ 36 #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 37 #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 38 #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 39 #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 40 #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 41 #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 42 #define BPF_CALL 0x80 /* function call */ 43 #define BPF_EXIT 0x90 /* function return */ 44 45 /* Register numbers */ 46 enum { 47 BPF_REG_0 = 0, 48 BPF_REG_1, 49 BPF_REG_2, 50 BPF_REG_3, 51 BPF_REG_4, 52 BPF_REG_5, 53 BPF_REG_6, 54 BPF_REG_7, 55 BPF_REG_8, 56 BPF_REG_9, 57 BPF_REG_10, 58 __MAX_BPF_REG, 59 }; 60 61 /* BPF has 10 general purpose 64-bit registers and stack frame. */ 62 #define MAX_BPF_REG __MAX_BPF_REG 63 64 struct bpf_insn { 65 __u8 code; /* opcode */ 66 __u8 dst_reg:4; /* dest register */ 67 __u8 src_reg:4; /* source register */ 68 __s16 off; /* signed offset */ 69 __s32 imm; /* signed immediate constant */ 70 }; 71 72 /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 73 struct bpf_lpm_trie_key { 74 __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 75 __u8 data[0]; /* Arbitrary size */ 76 }; 77 78 /* BPF syscall commands, see bpf(2) man-page for details. */ 79 enum bpf_cmd { 80 BPF_MAP_CREATE, 81 BPF_MAP_LOOKUP_ELEM, 82 BPF_MAP_UPDATE_ELEM, 83 BPF_MAP_DELETE_ELEM, 84 BPF_MAP_GET_NEXT_KEY, 85 BPF_PROG_LOAD, 86 BPF_OBJ_PIN, 87 BPF_OBJ_GET, 88 BPF_PROG_ATTACH, 89 BPF_PROG_DETACH, 90 BPF_PROG_TEST_RUN, 91 BPF_PROG_GET_NEXT_ID, 92 BPF_MAP_GET_NEXT_ID, 93 BPF_PROG_GET_FD_BY_ID, 94 BPF_MAP_GET_FD_BY_ID, 95 BPF_OBJ_GET_INFO_BY_FD, 96 BPF_PROG_QUERY, 97 }; 98 99 enum bpf_map_type { 100 BPF_MAP_TYPE_UNSPEC, 101 BPF_MAP_TYPE_HASH, 102 BPF_MAP_TYPE_ARRAY, 103 BPF_MAP_TYPE_PROG_ARRAY, 104 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 105 BPF_MAP_TYPE_PERCPU_HASH, 106 BPF_MAP_TYPE_PERCPU_ARRAY, 107 BPF_MAP_TYPE_STACK_TRACE, 108 BPF_MAP_TYPE_CGROUP_ARRAY, 109 BPF_MAP_TYPE_LRU_HASH, 110 BPF_MAP_TYPE_LRU_PERCPU_HASH, 111 BPF_MAP_TYPE_LPM_TRIE, 112 BPF_MAP_TYPE_ARRAY_OF_MAPS, 113 BPF_MAP_TYPE_HASH_OF_MAPS, 114 BPF_MAP_TYPE_DEVMAP, 115 BPF_MAP_TYPE_SOCKMAP, 116 BPF_MAP_TYPE_CPUMAP, 117 }; 118 119 enum bpf_prog_type { 120 BPF_PROG_TYPE_UNSPEC, 121 BPF_PROG_TYPE_SOCKET_FILTER, 122 BPF_PROG_TYPE_KPROBE, 123 BPF_PROG_TYPE_SCHED_CLS, 124 BPF_PROG_TYPE_SCHED_ACT, 125 BPF_PROG_TYPE_TRACEPOINT, 126 BPF_PROG_TYPE_XDP, 127 BPF_PROG_TYPE_PERF_EVENT, 128 BPF_PROG_TYPE_CGROUP_SKB, 129 BPF_PROG_TYPE_CGROUP_SOCK, 130 BPF_PROG_TYPE_LWT_IN, 131 BPF_PROG_TYPE_LWT_OUT, 132 BPF_PROG_TYPE_LWT_XMIT, 133 BPF_PROG_TYPE_SOCK_OPS, 134 BPF_PROG_TYPE_SK_SKB, 135 BPF_PROG_TYPE_CGROUP_DEVICE, 136 }; 137 138 enum bpf_attach_type { 139 BPF_CGROUP_INET_INGRESS, 140 BPF_CGROUP_INET_EGRESS, 141 BPF_CGROUP_INET_SOCK_CREATE, 142 BPF_CGROUP_SOCK_OPS, 143 BPF_SK_SKB_STREAM_PARSER, 144 BPF_SK_SKB_STREAM_VERDICT, 145 BPF_CGROUP_DEVICE, 146 __MAX_BPF_ATTACH_TYPE 147 }; 148 149 #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 150 151 /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 152 * 153 * NONE(default): No further bpf programs allowed in the subtree. 154 * 155 * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 156 * the program in this cgroup yields to sub-cgroup program. 157 * 158 * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 159 * that cgroup program gets run in addition to the program in this cgroup. 160 * 161 * Only one program is allowed to be attached to a cgroup with 162 * NONE or BPF_F_ALLOW_OVERRIDE flag. 163 * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 164 * release old program and attach the new one. Attach flags has to match. 165 * 166 * Multiple programs are allowed to be attached to a cgroup with 167 * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 168 * (those that were attached first, run first) 169 * The programs of sub-cgroup are executed first, then programs of 170 * this cgroup and then programs of parent cgroup. 171 * When children program makes decision (like picking TCP CA or sock bind) 172 * parent program has a chance to override it. 173 * 174 * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 175 * A cgroup with NONE doesn't allow any programs in sub-cgroups. 176 * Ex1: 177 * cgrp1 (MULTI progs A, B) -> 178 * cgrp2 (OVERRIDE prog C) -> 179 * cgrp3 (MULTI prog D) -> 180 * cgrp4 (OVERRIDE prog E) -> 181 * cgrp5 (NONE prog F) 182 * the event in cgrp5 triggers execution of F,D,A,B in that order. 183 * if prog F is detached, the execution is E,D,A,B 184 * if prog F and D are detached, the execution is E,A,B 185 * if prog F, E and D are detached, the execution is C,A,B 186 * 187 * All eligible programs are executed regardless of return code from 188 * earlier programs. 189 */ 190 #define BPF_F_ALLOW_OVERRIDE (1U << 0) 191 #define BPF_F_ALLOW_MULTI (1U << 1) 192 193 /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 194 * verifier will perform strict alignment checking as if the kernel 195 * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 196 * and NET_IP_ALIGN defined to 2. 197 */ 198 #define BPF_F_STRICT_ALIGNMENT (1U << 0) 199 200 /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */ 201 #define BPF_PSEUDO_MAP_FD 1 202 203 /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 204 * offset to another bpf function 205 */ 206 #define BPF_PSEUDO_CALL 1 207 208 /* flags for BPF_MAP_UPDATE_ELEM command */ 209 #define BPF_ANY 0 /* create new element or update existing */ 210 #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 211 #define BPF_EXIST 2 /* update existing element */ 212 213 /* flags for BPF_MAP_CREATE command */ 214 #define BPF_F_NO_PREALLOC (1U << 0) 215 /* Instead of having one common LRU list in the 216 * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 217 * which can scale and perform better. 218 * Note, the LRU nodes (including free nodes) cannot be moved 219 * across different LRU lists. 220 */ 221 #define BPF_F_NO_COMMON_LRU (1U << 1) 222 /* Specify numa node during map creation */ 223 #define BPF_F_NUMA_NODE (1U << 2) 224 225 /* flags for BPF_PROG_QUERY */ 226 #define BPF_F_QUERY_EFFECTIVE (1U << 0) 227 228 #define BPF_OBJ_NAME_LEN 16U 229 230 /* Flags for accessing BPF object */ 231 #define BPF_F_RDONLY (1U << 3) 232 #define BPF_F_WRONLY (1U << 4) 233 234 union bpf_attr { 235 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 236 __u32 map_type; /* one of enum bpf_map_type */ 237 __u32 key_size; /* size of key in bytes */ 238 __u32 value_size; /* size of value in bytes */ 239 __u32 max_entries; /* max number of entries in a map */ 240 __u32 map_flags; /* BPF_MAP_CREATE related 241 * flags defined above. 242 */ 243 __u32 inner_map_fd; /* fd pointing to the inner map */ 244 __u32 numa_node; /* numa node (effective only if 245 * BPF_F_NUMA_NODE is set). 246 */ 247 char map_name[BPF_OBJ_NAME_LEN]; 248 __u32 map_ifindex; /* ifindex of netdev to create on */ 249 }; 250 251 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 252 __u32 map_fd; 253 __aligned_u64 key; 254 union { 255 __aligned_u64 value; 256 __aligned_u64 next_key; 257 }; 258 __u64 flags; 259 }; 260 261 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 262 __u32 prog_type; /* one of enum bpf_prog_type */ 263 __u32 insn_cnt; 264 __aligned_u64 insns; 265 __aligned_u64 license; 266 __u32 log_level; /* verbosity level of verifier */ 267 __u32 log_size; /* size of user buffer */ 268 __aligned_u64 log_buf; /* user supplied buffer */ 269 __u32 kern_version; /* checked when prog_type=kprobe */ 270 __u32 prog_flags; 271 char prog_name[BPF_OBJ_NAME_LEN]; 272 __u32 prog_ifindex; /* ifindex of netdev to prep for */ 273 }; 274 275 struct { /* anonymous struct used by BPF_OBJ_* commands */ 276 __aligned_u64 pathname; 277 __u32 bpf_fd; 278 __u32 file_flags; 279 }; 280 281 struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 282 __u32 target_fd; /* container object to attach to */ 283 __u32 attach_bpf_fd; /* eBPF program to attach */ 284 __u32 attach_type; 285 __u32 attach_flags; 286 }; 287 288 struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 289 __u32 prog_fd; 290 __u32 retval; 291 __u32 data_size_in; 292 __u32 data_size_out; 293 __aligned_u64 data_in; 294 __aligned_u64 data_out; 295 __u32 repeat; 296 __u32 duration; 297 } test; 298 299 struct { /* anonymous struct used by BPF_*_GET_*_ID */ 300 union { 301 __u32 start_id; 302 __u32 prog_id; 303 __u32 map_id; 304 }; 305 __u32 next_id; 306 __u32 open_flags; 307 }; 308 309 struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 310 __u32 bpf_fd; 311 __u32 info_len; 312 __aligned_u64 info; 313 } info; 314 315 struct { /* anonymous struct used by BPF_PROG_QUERY command */ 316 __u32 target_fd; /* container object to query */ 317 __u32 attach_type; 318 __u32 query_flags; 319 __u32 attach_flags; 320 __aligned_u64 prog_ids; 321 __u32 prog_cnt; 322 } query; 323 } __attribute__((aligned(8))); 324 325 /* BPF helper function descriptions: 326 * 327 * void *bpf_map_lookup_elem(&map, &key) 328 * Return: Map value or NULL 329 * 330 * int bpf_map_update_elem(&map, &key, &value, flags) 331 * Return: 0 on success or negative error 332 * 333 * int bpf_map_delete_elem(&map, &key) 334 * Return: 0 on success or negative error 335 * 336 * int bpf_probe_read(void *dst, int size, void *src) 337 * Return: 0 on success or negative error 338 * 339 * u64 bpf_ktime_get_ns(void) 340 * Return: current ktime 341 * 342 * int bpf_trace_printk(const char *fmt, int fmt_size, ...) 343 * Return: length of buffer written or negative error 344 * 345 * u32 bpf_prandom_u32(void) 346 * Return: random value 347 * 348 * u32 bpf_raw_smp_processor_id(void) 349 * Return: SMP processor ID 350 * 351 * int bpf_skb_store_bytes(skb, offset, from, len, flags) 352 * store bytes into packet 353 * @skb: pointer to skb 354 * @offset: offset within packet from skb->mac_header 355 * @from: pointer where to copy bytes from 356 * @len: number of bytes to store into packet 357 * @flags: bit 0 - if true, recompute skb->csum 358 * other bits - reserved 359 * Return: 0 on success or negative error 360 * 361 * int bpf_l3_csum_replace(skb, offset, from, to, flags) 362 * recompute IP checksum 363 * @skb: pointer to skb 364 * @offset: offset within packet where IP checksum is located 365 * @from: old value of header field 366 * @to: new value of header field 367 * @flags: bits 0-3 - size of header field 368 * other bits - reserved 369 * Return: 0 on success or negative error 370 * 371 * int bpf_l4_csum_replace(skb, offset, from, to, flags) 372 * recompute TCP/UDP checksum 373 * @skb: pointer to skb 374 * @offset: offset within packet where TCP/UDP checksum is located 375 * @from: old value of header field 376 * @to: new value of header field 377 * @flags: bits 0-3 - size of header field 378 * bit 4 - is pseudo header 379 * other bits - reserved 380 * Return: 0 on success or negative error 381 * 382 * int bpf_tail_call(ctx, prog_array_map, index) 383 * jump into another BPF program 384 * @ctx: context pointer passed to next program 385 * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY 386 * @index: 32-bit index inside array that selects specific program to run 387 * Return: 0 on success or negative error 388 * 389 * int bpf_clone_redirect(skb, ifindex, flags) 390 * redirect to another netdev 391 * @skb: pointer to skb 392 * @ifindex: ifindex of the net device 393 * @flags: bit 0 - if set, redirect to ingress instead of egress 394 * other bits - reserved 395 * Return: 0 on success or negative error 396 * 397 * u64 bpf_get_current_pid_tgid(void) 398 * Return: current->tgid << 32 | current->pid 399 * 400 * u64 bpf_get_current_uid_gid(void) 401 * Return: current_gid << 32 | current_uid 402 * 403 * int bpf_get_current_comm(char *buf, int size_of_buf) 404 * stores current->comm into buf 405 * Return: 0 on success or negative error 406 * 407 * u32 bpf_get_cgroup_classid(skb) 408 * retrieve a proc's classid 409 * @skb: pointer to skb 410 * Return: classid if != 0 411 * 412 * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci) 413 * Return: 0 on success or negative error 414 * 415 * int bpf_skb_vlan_pop(skb) 416 * Return: 0 on success or negative error 417 * 418 * int bpf_skb_get_tunnel_key(skb, key, size, flags) 419 * int bpf_skb_set_tunnel_key(skb, key, size, flags) 420 * retrieve or populate tunnel metadata 421 * @skb: pointer to skb 422 * @key: pointer to 'struct bpf_tunnel_key' 423 * @size: size of 'struct bpf_tunnel_key' 424 * @flags: room for future extensions 425 * Return: 0 on success or negative error 426 * 427 * u64 bpf_perf_event_read(map, flags) 428 * read perf event counter value 429 * @map: pointer to perf_event_array map 430 * @flags: index of event in the map or bitmask flags 431 * Return: value of perf event counter read or error code 432 * 433 * int bpf_redirect(ifindex, flags) 434 * redirect to another netdev 435 * @ifindex: ifindex of the net device 436 * @flags: 437 * cls_bpf: 438 * bit 0 - if set, redirect to ingress instead of egress 439 * other bits - reserved 440 * xdp_bpf: 441 * all bits - reserved 442 * Return: cls_bpf: TC_ACT_REDIRECT on success or TC_ACT_SHOT on error 443 * xdp_bfp: XDP_REDIRECT on success or XDP_ABORT on error 444 * int bpf_redirect_map(map, key, flags) 445 * redirect to endpoint in map 446 * @map: pointer to dev map 447 * @key: index in map to lookup 448 * @flags: -- 449 * Return: XDP_REDIRECT on success or XDP_ABORT on error 450 * 451 * u32 bpf_get_route_realm(skb) 452 * retrieve a dst's tclassid 453 * @skb: pointer to skb 454 * Return: realm if != 0 455 * 456 * int bpf_perf_event_output(ctx, map, flags, data, size) 457 * output perf raw sample 458 * @ctx: struct pt_regs* 459 * @map: pointer to perf_event_array map 460 * @flags: index of event in the map or bitmask flags 461 * @data: data on stack to be output as raw data 462 * @size: size of data 463 * Return: 0 on success or negative error 464 * 465 * int bpf_get_stackid(ctx, map, flags) 466 * walk user or kernel stack and return id 467 * @ctx: struct pt_regs* 468 * @map: pointer to stack_trace map 469 * @flags: bits 0-7 - numer of stack frames to skip 470 * bit 8 - collect user stack instead of kernel 471 * bit 9 - compare stacks by hash only 472 * bit 10 - if two different stacks hash into the same stackid 473 * discard old 474 * other bits - reserved 475 * Return: >= 0 stackid on success or negative error 476 * 477 * s64 bpf_csum_diff(from, from_size, to, to_size, seed) 478 * calculate csum diff 479 * @from: raw from buffer 480 * @from_size: length of from buffer 481 * @to: raw to buffer 482 * @to_size: length of to buffer 483 * @seed: optional seed 484 * Return: csum result or negative error code 485 * 486 * int bpf_skb_get_tunnel_opt(skb, opt, size) 487 * retrieve tunnel options metadata 488 * @skb: pointer to skb 489 * @opt: pointer to raw tunnel option data 490 * @size: size of @opt 491 * Return: option size 492 * 493 * int bpf_skb_set_tunnel_opt(skb, opt, size) 494 * populate tunnel options metadata 495 * @skb: pointer to skb 496 * @opt: pointer to raw tunnel option data 497 * @size: size of @opt 498 * Return: 0 on success or negative error 499 * 500 * int bpf_skb_change_proto(skb, proto, flags) 501 * Change protocol of the skb. Currently supported is v4 -> v6, 502 * v6 -> v4 transitions. The helper will also resize the skb. eBPF 503 * program is expected to fill the new headers via skb_store_bytes 504 * and lX_csum_replace. 505 * @skb: pointer to skb 506 * @proto: new skb->protocol type 507 * @flags: reserved 508 * Return: 0 on success or negative error 509 * 510 * int bpf_skb_change_type(skb, type) 511 * Change packet type of skb. 512 * @skb: pointer to skb 513 * @type: new skb->pkt_type type 514 * Return: 0 on success or negative error 515 * 516 * int bpf_skb_under_cgroup(skb, map, index) 517 * Check cgroup2 membership of skb 518 * @skb: pointer to skb 519 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type 520 * @index: index of the cgroup in the bpf_map 521 * Return: 522 * == 0 skb failed the cgroup2 descendant test 523 * == 1 skb succeeded the cgroup2 descendant test 524 * < 0 error 525 * 526 * u32 bpf_get_hash_recalc(skb) 527 * Retrieve and possibly recalculate skb->hash. 528 * @skb: pointer to skb 529 * Return: hash 530 * 531 * u64 bpf_get_current_task(void) 532 * Returns current task_struct 533 * Return: current 534 * 535 * int bpf_probe_write_user(void *dst, void *src, int len) 536 * safely attempt to write to a location 537 * @dst: destination address in userspace 538 * @src: source address on stack 539 * @len: number of bytes to copy 540 * Return: 0 on success or negative error 541 * 542 * int bpf_current_task_under_cgroup(map, index) 543 * Check cgroup2 membership of current task 544 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type 545 * @index: index of the cgroup in the bpf_map 546 * Return: 547 * == 0 current failed the cgroup2 descendant test 548 * == 1 current succeeded the cgroup2 descendant test 549 * < 0 error 550 * 551 * int bpf_skb_change_tail(skb, len, flags) 552 * The helper will resize the skb to the given new size, to be used f.e. 553 * with control messages. 554 * @skb: pointer to skb 555 * @len: new skb length 556 * @flags: reserved 557 * Return: 0 on success or negative error 558 * 559 * int bpf_skb_pull_data(skb, len) 560 * The helper will pull in non-linear data in case the skb is non-linear 561 * and not all of len are part of the linear section. Only needed for 562 * read/write with direct packet access. 563 * @skb: pointer to skb 564 * @len: len to make read/writeable 565 * Return: 0 on success or negative error 566 * 567 * s64 bpf_csum_update(skb, csum) 568 * Adds csum into skb->csum in case of CHECKSUM_COMPLETE. 569 * @skb: pointer to skb 570 * @csum: csum to add 571 * Return: csum on success or negative error 572 * 573 * void bpf_set_hash_invalid(skb) 574 * Invalidate current skb->hash. 575 * @skb: pointer to skb 576 * 577 * int bpf_get_numa_node_id() 578 * Return: Id of current NUMA node. 579 * 580 * int bpf_skb_change_head() 581 * Grows headroom of skb and adjusts MAC header offset accordingly. 582 * Will extends/reallocae as required automatically. 583 * May change skb data pointer and will thus invalidate any check 584 * performed for direct packet access. 585 * @skb: pointer to skb 586 * @len: length of header to be pushed in front 587 * @flags: Flags (unused for now) 588 * Return: 0 on success or negative error 589 * 590 * int bpf_xdp_adjust_head(xdp_md, delta) 591 * Adjust the xdp_md.data by delta 592 * @xdp_md: pointer to xdp_md 593 * @delta: An positive/negative integer to be added to xdp_md.data 594 * Return: 0 on success or negative on error 595 * 596 * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) 597 * Copy a NUL terminated string from unsafe address. In case the string 598 * length is smaller than size, the target is not padded with further NUL 599 * bytes. In case the string length is larger than size, just count-1 600 * bytes are copied and the last byte is set to NUL. 601 * @dst: destination address 602 * @size: maximum number of bytes to copy, including the trailing NUL 603 * @unsafe_ptr: unsafe address 604 * Return: 605 * > 0 length of the string including the trailing NUL on success 606 * < 0 error 607 * 608 * u64 bpf_get_socket_cookie(skb) 609 * Get the cookie for the socket stored inside sk_buff. 610 * @skb: pointer to skb 611 * Return: 8 Bytes non-decreasing number on success or 0 if the socket 612 * field is missing inside sk_buff 613 * 614 * u32 bpf_get_socket_uid(skb) 615 * Get the owner uid of the socket stored inside sk_buff. 616 * @skb: pointer to skb 617 * Return: uid of the socket owner on success or overflowuid if failed. 618 * 619 * u32 bpf_set_hash(skb, hash) 620 * Set full skb->hash. 621 * @skb: pointer to skb 622 * @hash: hash to set 623 * 624 * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen) 625 * Calls setsockopt. Not all opts are available, only those with 626 * integer optvals plus TCP_CONGESTION. 627 * Supported levels: SOL_SOCKET and IPPROTO_TCP 628 * @bpf_socket: pointer to bpf_socket 629 * @level: SOL_SOCKET or IPPROTO_TCP 630 * @optname: option name 631 * @optval: pointer to option value 632 * @optlen: length of optval in bytes 633 * Return: 0 or negative error 634 * 635 * int bpf_getsockopt(bpf_socket, level, optname, optval, optlen) 636 * Calls getsockopt. Not all opts are available. 637 * Supported levels: IPPROTO_TCP 638 * @bpf_socket: pointer to bpf_socket 639 * @level: IPPROTO_TCP 640 * @optname: option name 641 * @optval: pointer to option value 642 * @optlen: length of optval in bytes 643 * Return: 0 or negative error 644 * 645 * int bpf_skb_adjust_room(skb, len_diff, mode, flags) 646 * Grow or shrink room in sk_buff. 647 * @skb: pointer to skb 648 * @len_diff: (signed) amount of room to grow/shrink 649 * @mode: operation mode (enum bpf_adj_room_mode) 650 * @flags: reserved for future use 651 * Return: 0 on success or negative error code 652 * 653 * int bpf_sk_redirect_map(map, key, flags) 654 * Redirect skb to a sock in map using key as a lookup key for the 655 * sock in map. 656 * @map: pointer to sockmap 657 * @key: key to lookup sock in map 658 * @flags: reserved for future use 659 * Return: SK_PASS 660 * 661 * int bpf_sock_map_update(skops, map, key, flags) 662 * @skops: pointer to bpf_sock_ops 663 * @map: pointer to sockmap to update 664 * @key: key to insert/update sock in map 665 * @flags: same flags as map update elem 666 * 667 * int bpf_xdp_adjust_meta(xdp_md, delta) 668 * Adjust the xdp_md.data_meta by delta 669 * @xdp_md: pointer to xdp_md 670 * @delta: An positive/negative integer to be added to xdp_md.data_meta 671 * Return: 0 on success or negative on error 672 * 673 * int bpf_perf_event_read_value(map, flags, buf, buf_size) 674 * read perf event counter value and perf event enabled/running time 675 * @map: pointer to perf_event_array map 676 * @flags: index of event in the map or bitmask flags 677 * @buf: buf to fill 678 * @buf_size: size of the buf 679 * Return: 0 on success or negative error code 680 * 681 * int bpf_perf_prog_read_value(ctx, buf, buf_size) 682 * read perf prog attached perf event counter and enabled/running time 683 * @ctx: pointer to ctx 684 * @buf: buf to fill 685 * @buf_size: size of the buf 686 * Return : 0 on success or negative error code 687 * 688 * int bpf_override_return(pt_regs, rc) 689 * @pt_regs: pointer to struct pt_regs 690 * @rc: the return value to set 691 */ 692 #define __BPF_FUNC_MAPPER(FN) \ 693 FN(unspec), \ 694 FN(map_lookup_elem), \ 695 FN(map_update_elem), \ 696 FN(map_delete_elem), \ 697 FN(probe_read), \ 698 FN(ktime_get_ns), \ 699 FN(trace_printk), \ 700 FN(get_prandom_u32), \ 701 FN(get_smp_processor_id), \ 702 FN(skb_store_bytes), \ 703 FN(l3_csum_replace), \ 704 FN(l4_csum_replace), \ 705 FN(tail_call), \ 706 FN(clone_redirect), \ 707 FN(get_current_pid_tgid), \ 708 FN(get_current_uid_gid), \ 709 FN(get_current_comm), \ 710 FN(get_cgroup_classid), \ 711 FN(skb_vlan_push), \ 712 FN(skb_vlan_pop), \ 713 FN(skb_get_tunnel_key), \ 714 FN(skb_set_tunnel_key), \ 715 FN(perf_event_read), \ 716 FN(redirect), \ 717 FN(get_route_realm), \ 718 FN(perf_event_output), \ 719 FN(skb_load_bytes), \ 720 FN(get_stackid), \ 721 FN(csum_diff), \ 722 FN(skb_get_tunnel_opt), \ 723 FN(skb_set_tunnel_opt), \ 724 FN(skb_change_proto), \ 725 FN(skb_change_type), \ 726 FN(skb_under_cgroup), \ 727 FN(get_hash_recalc), \ 728 FN(get_current_task), \ 729 FN(probe_write_user), \ 730 FN(current_task_under_cgroup), \ 731 FN(skb_change_tail), \ 732 FN(skb_pull_data), \ 733 FN(csum_update), \ 734 FN(set_hash_invalid), \ 735 FN(get_numa_node_id), \ 736 FN(skb_change_head), \ 737 FN(xdp_adjust_head), \ 738 FN(probe_read_str), \ 739 FN(get_socket_cookie), \ 740 FN(get_socket_uid), \ 741 FN(set_hash), \ 742 FN(setsockopt), \ 743 FN(skb_adjust_room), \ 744 FN(redirect_map), \ 745 FN(sk_redirect_map), \ 746 FN(sock_map_update), \ 747 FN(xdp_adjust_meta), \ 748 FN(perf_event_read_value), \ 749 FN(perf_prog_read_value), \ 750 FN(getsockopt), \ 751 FN(override_return), 752 753 /* integer value in 'imm' field of BPF_CALL instruction selects which helper 754 * function eBPF program intends to call 755 */ 756 #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 757 enum bpf_func_id { 758 __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 759 __BPF_FUNC_MAX_ID, 760 }; 761 #undef __BPF_ENUM_FN 762 763 /* All flags used by eBPF helper functions, placed here. */ 764 765 /* BPF_FUNC_skb_store_bytes flags. */ 766 #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 767 #define BPF_F_INVALIDATE_HASH (1ULL << 1) 768 769 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 770 * First 4 bits are for passing the header field size. 771 */ 772 #define BPF_F_HDR_FIELD_MASK 0xfULL 773 774 /* BPF_FUNC_l4_csum_replace flags. */ 775 #define BPF_F_PSEUDO_HDR (1ULL << 4) 776 #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 777 #define BPF_F_MARK_ENFORCE (1ULL << 6) 778 779 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 780 #define BPF_F_INGRESS (1ULL << 0) 781 782 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 783 #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 784 785 /* BPF_FUNC_get_stackid flags. */ 786 #define BPF_F_SKIP_FIELD_MASK 0xffULL 787 #define BPF_F_USER_STACK (1ULL << 8) 788 #define BPF_F_FAST_STACK_CMP (1ULL << 9) 789 #define BPF_F_REUSE_STACKID (1ULL << 10) 790 791 /* BPF_FUNC_skb_set_tunnel_key flags. */ 792 #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 793 #define BPF_F_DONT_FRAGMENT (1ULL << 2) 794 795 /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 796 * BPF_FUNC_perf_event_read_value flags. 797 */ 798 #define BPF_F_INDEX_MASK 0xffffffffULL 799 #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 800 /* BPF_FUNC_perf_event_output for sk_buff input context. */ 801 #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 802 803 /* Mode for BPF_FUNC_skb_adjust_room helper. */ 804 enum bpf_adj_room_mode { 805 BPF_ADJ_ROOM_NET, 806 }; 807 808 /* user accessible mirror of in-kernel sk_buff. 809 * new fields can only be added to the end of this structure 810 */ 811 struct __sk_buff { 812 __u32 len; 813 __u32 pkt_type; 814 __u32 mark; 815 __u32 queue_mapping; 816 __u32 protocol; 817 __u32 vlan_present; 818 __u32 vlan_tci; 819 __u32 vlan_proto; 820 __u32 priority; 821 __u32 ingress_ifindex; 822 __u32 ifindex; 823 __u32 tc_index; 824 __u32 cb[5]; 825 __u32 hash; 826 __u32 tc_classid; 827 __u32 data; 828 __u32 data_end; 829 __u32 napi_id; 830 831 /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 832 __u32 family; 833 __u32 remote_ip4; /* Stored in network byte order */ 834 __u32 local_ip4; /* Stored in network byte order */ 835 __u32 remote_ip6[4]; /* Stored in network byte order */ 836 __u32 local_ip6[4]; /* Stored in network byte order */ 837 __u32 remote_port; /* Stored in network byte order */ 838 __u32 local_port; /* stored in host byte order */ 839 /* ... here. */ 840 841 __u32 data_meta; 842 }; 843 844 struct bpf_tunnel_key { 845 __u32 tunnel_id; 846 union { 847 __u32 remote_ipv4; 848 __u32 remote_ipv6[4]; 849 }; 850 __u8 tunnel_tos; 851 __u8 tunnel_ttl; 852 __u16 tunnel_ext; 853 __u32 tunnel_label; 854 }; 855 856 /* Generic BPF return codes which all BPF program types may support. 857 * The values are binary compatible with their TC_ACT_* counter-part to 858 * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 859 * programs. 860 * 861 * XDP is handled seprately, see XDP_*. 862 */ 863 enum bpf_ret_code { 864 BPF_OK = 0, 865 /* 1 reserved */ 866 BPF_DROP = 2, 867 /* 3-6 reserved */ 868 BPF_REDIRECT = 7, 869 /* >127 are reserved for prog type specific return codes */ 870 }; 871 872 struct bpf_sock { 873 __u32 bound_dev_if; 874 __u32 family; 875 __u32 type; 876 __u32 protocol; 877 __u32 mark; 878 __u32 priority; 879 }; 880 881 #define XDP_PACKET_HEADROOM 256 882 883 /* User return codes for XDP prog type. 884 * A valid XDP program must return one of these defined values. All other 885 * return codes are reserved for future use. Unknown return codes will 886 * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 887 */ 888 enum xdp_action { 889 XDP_ABORTED = 0, 890 XDP_DROP, 891 XDP_PASS, 892 XDP_TX, 893 XDP_REDIRECT, 894 }; 895 896 /* user accessible metadata for XDP packet hook 897 * new fields must be added to the end of this structure 898 */ 899 struct xdp_md { 900 __u32 data; 901 __u32 data_end; 902 __u32 data_meta; 903 /* Below access go through struct xdp_rxq_info */ 904 __u32 ingress_ifindex; /* rxq->dev->ifindex */ 905 __u32 rx_queue_index; /* rxq->queue_index */ 906 }; 907 908 enum sk_action { 909 SK_DROP = 0, 910 SK_PASS, 911 }; 912 913 #define BPF_TAG_SIZE 8 914 915 struct bpf_prog_info { 916 __u32 type; 917 __u32 id; 918 __u8 tag[BPF_TAG_SIZE]; 919 __u32 jited_prog_len; 920 __u32 xlated_prog_len; 921 __aligned_u64 jited_prog_insns; 922 __aligned_u64 xlated_prog_insns; 923 __u64 load_time; /* ns since boottime */ 924 __u32 created_by_uid; 925 __u32 nr_map_ids; 926 __aligned_u64 map_ids; 927 char name[BPF_OBJ_NAME_LEN]; 928 __u32 ifindex; 929 __u64 netns_dev; 930 __u64 netns_ino; 931 } __attribute__((aligned(8))); 932 933 struct bpf_map_info { 934 __u32 type; 935 __u32 id; 936 __u32 key_size; 937 __u32 value_size; 938 __u32 max_entries; 939 __u32 map_flags; 940 char name[BPF_OBJ_NAME_LEN]; 941 } __attribute__((aligned(8))); 942 943 /* User bpf_sock_ops struct to access socket values and specify request ops 944 * and their replies. 945 * Some of this fields are in network (bigendian) byte order and may need 946 * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 947 * New fields can only be added at the end of this structure 948 */ 949 struct bpf_sock_ops { 950 __u32 op; 951 union { 952 __u32 reply; 953 __u32 replylong[4]; 954 }; 955 __u32 family; 956 __u32 remote_ip4; /* Stored in network byte order */ 957 __u32 local_ip4; /* Stored in network byte order */ 958 __u32 remote_ip6[4]; /* Stored in network byte order */ 959 __u32 local_ip6[4]; /* Stored in network byte order */ 960 __u32 remote_port; /* Stored in network byte order */ 961 __u32 local_port; /* stored in host byte order */ 962 __u32 is_fullsock; /* Some TCP fields are only valid if 963 * there is a full socket. If not, the 964 * fields read as zero. 965 */ 966 __u32 snd_cwnd; 967 __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 968 }; 969 970 /* List of known BPF sock_ops operators. 971 * New entries can only be added at the end 972 */ 973 enum { 974 BPF_SOCK_OPS_VOID, 975 BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 976 * -1 if default value should be used 977 */ 978 BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 979 * window (in packets) or -1 if default 980 * value should be used 981 */ 982 BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 983 * active connection is initialized 984 */ 985 BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 986 * active connection is 987 * established 988 */ 989 BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 990 * passive connection is 991 * established 992 */ 993 BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 994 * needs ECN 995 */ 996 BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 997 * based on the path and may be 998 * dependent on the congestion control 999 * algorithm. In general it indicates 1000 * a congestion threshold. RTTs above 1001 * this indicate congestion 1002 */ 1003 }; 1004 1005 #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 1006 #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 1007 1008 struct bpf_perf_event_value { 1009 __u64 counter; 1010 __u64 enabled; 1011 __u64 running; 1012 }; 1013 1014 #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 1015 #define BPF_DEVCG_ACC_READ (1ULL << 1) 1016 #define BPF_DEVCG_ACC_WRITE (1ULL << 2) 1017 1018 #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 1019 #define BPF_DEVCG_DEV_CHAR (1ULL << 1) 1020 1021 struct bpf_cgroup_dev_ctx { 1022 /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 1023 __u32 access_type; 1024 __u32 major; 1025 __u32 minor; 1026 }; 1027 1028 #endif /* _UAPI__LINUX_BPF_H__ */ 1029